diff --git a/.github/workflows/muelu_clang_format.yml b/.github/workflows/muelu_clang_format.yml
new file mode 100644
index 000000000000..fc4ef5e00c34
--- /dev/null
+++ b/.github/workflows/muelu_clang_format.yml
@@ -0,0 +1,29 @@
+name: Check MueLu clang-format
+
+on: [pull_request]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+    - uses: DoozyX/clang-format-lint-action@v0.16.2
+      with:
+        source: './packages/muelu'
+        exclude: ''
+        extensions: 'cpp,hpp'
+        clangFormatVersion: 14
+        inplace: true
+    - run: git diff HEAD > format_patch.txt
+    - run: if [ "$(cat format_patch.txt)" == "" ] ; then rm format_patch.txt ; fi
+    - uses: actions/upload-artifact@v3
+      if: ${{ hashFiles('format_patch.txt') != '' }}
+      with:
+        name: MueLu format patch
+        path: format_patch.txt
+    - uses: actions/github-script@v3
+      if: ${{ hashFiles('format_patch.txt') != '' }}
+      with:
+        script: |
+          core.setFailed('Please download and apply the formatting patch! It is located at the bottom of the summary tab for this workflow.')
diff --git a/packages/muelu/.clang-format b/packages/muelu/.clang-format
new file mode 100644
index 000000000000..137569c6fc4c
--- /dev/null
+++ b/packages/muelu/.clang-format
@@ -0,0 +1,13 @@
+#Official Tool: clang-format version 14.0.0
+#Kokkos options
+BasedOnStyle: google
+SortIncludes: false
+AlignConsecutiveAssignments: true
+AllowShortCaseLabelsOnASingleLine: true
+AllowShortIfStatementsOnASingleLine: true
+#MueLu-specific options
+ColumnLimit: 0
+BreakConstructorInitializersBeforeComma: true
+BreakConstructorInitializers: BeforeComma
+ConstructorInitializerAllOnOneLineOrOnePerLine: false
+ConstructorInitializerIndentWidth: 2
diff --git a/packages/muelu/adapters/amgx/MueLu_AMGXOperator_decl.hpp b/packages/muelu/adapters/amgx/MueLu_AMGXOperator_decl.hpp
index 65d13a03d4a4..b233bbc226f6 100644
--- a/packages/muelu/adapters/amgx/MueLu_AMGXOperator_decl.hpp
+++ b/packages/muelu/adapters/amgx/MueLu_AMGXOperator_decl.hpp
@@ -46,7 +46,7 @@
 #ifndef MUELU_AMGXOPERATOR_DECL_HPP
 #define MUELU_AMGXOPERATOR_DECL_HPP
 
-#if defined (HAVE_MUELU_AMGX)
+#if defined(HAVE_MUELU_AMGX)
 #include <Teuchos_ParameterList.hpp>
 
 #include <Tpetra_Operator.hpp>
@@ -67,468 +67,461 @@
 
 namespace MueLu {
 
-
-  /*! @class AMGXOperator
-      @ingroup MueLuAdapters
-      @brief Adapter for AmgX library from Nvidia.
-
-      This templated version of the class throws errors in all methods as AmgX is not implemented for datatypes where scalar!=double/float and ordinal !=int
- */
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class AMGXOperator : public TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>, public BaseClass {
-  private:
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-    typedef Tpetra::Map<LO,GO,NO>            Map;
-    typedef Tpetra::MultiVector<SC,LO,GO,NO> MultiVector;
-
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    AMGXOperator(const Teuchos::RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > &InA, Teuchos::ParameterList &paramListIn) { }
-
-    //! Destructor.
-    virtual ~AMGXOperator() {}
-
-    //@}
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Map> getDomainMap() const{
-      throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
-    }
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Map> getRangeMap() const{
-      throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
-    }
-
-    //! Returns a solution for the linear system AX=Y in the  Tpetra::MultiVector X.
-    /*!
-      \param[in]  X - Tpetra::MultiVector of dimension NumVectors that contains the solution to the linear system.
-      \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing the RHS of the linear system.
-    */
-    void apply(const MultiVector& X, MultiVector& Y, Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(), Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const {
-      throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
-    }
-
-    //! Indicates whether this operator supports applying the adjoint operator
-    bool hasTransposeApply() const{
-      throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
-    }
-
-    RCP<MueLu::Hierarchy<SC,LO,GO,NO> > GetHierarchy() const {
-      throw Exceptions::RuntimeError("AMGXOperator does not hold a MueLu::Hierarchy object \n");
-    }
-
-  private:
-  };
-
-  /*! @class AMGXOperator
-      @ingroup MueLuAdapters
-      @brief Adapter for AmgX library from Nvidia.
-
-      Creates an AmgX Solver object with a Tpetra Matrix. Partial specialization of the template for data types supported by AmgX.
+/*! @class AMGXOperator
+    @ingroup MueLuAdapters
+    @brief Adapter for AmgX library from Nvidia.
+
+    This templated version of the class throws errors in all methods as AmgX is not implemented for datatypes where scalar!=double/float and ordinal !=int
+*/
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class AMGXOperator : public TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>, public BaseClass {
+ private:
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+  typedef Tpetra::Map<LO, GO, NO> Map;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> MultiVector;
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  AMGXOperator(const Teuchos::RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& InA, Teuchos::ParameterList& paramListIn) {}
+
+  //! Destructor.
+  virtual ~AMGXOperator() {}
+
+  //@}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Map> getDomainMap() const {
+    throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
+  }
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Map> getRangeMap() const {
+    throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
+  }
+
+  //! Returns a solution for the linear system AX=Y in the  Tpetra::MultiVector X.
+  /*!
+    \param[in]  X - Tpetra::MultiVector of dimension NumVectors that contains the solution to the linear system.
+    \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing the RHS of the linear system.
   */
-  template<class Node>
-  class AMGXOperator<double, int, int, Node> : public TpetraOperator<double, int, int, Node> {
-  private:
-    typedef double  SC;
-    typedef int     LO;
-    typedef int     GO;
-    typedef Node    NO;
-
-    typedef Tpetra::Map<LO,GO,NO>            Map;
-    typedef Tpetra::MultiVector<SC,LO,GO,NO> MultiVector;
-
-   
-    void printMaps(Teuchos::RCP<const Teuchos::Comm<int> >& comm, const std::vector<std::vector<int> >& vec, const std::vector<int>& perm,
-                   const int* nbrs, const Map& map, const std::string& label) {
-      for (int p = 0; p < comm->getSize(); p++) {
-        if (comm->getRank() == p) {
-          std::cout << "========\n" << label << ", lid (gid), PID  " << p << "\n========" << std::endl;
-
-          for (size_t i = 0; i < vec.size(); ++i) {
-            std::cout << "   neighbor " << nbrs[i] << " :";
-            for (size_t j = 0; j < vec[i].size(); ++j)
-              std::cout << " " << vec[i][j] << " (" << map.getGlobalElement(perm[vec[i][j]]) << ")";
-            std::cout << std::endl;
-          }
+  void apply(const MultiVector& X, MultiVector& Y, Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(), Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const {
+    throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
+  }
+
+  //! Indicates whether this operator supports applying the adjoint operator
+  bool hasTransposeApply() const {
+    throw Exceptions::RuntimeError("Cannot use AMGXOperator with scalar != double and/or global ordinal != int \n");
+  }
+
+  RCP<MueLu::Hierarchy<SC, LO, GO, NO> > GetHierarchy() const {
+    throw Exceptions::RuntimeError("AMGXOperator does not hold a MueLu::Hierarchy object \n");
+  }
+
+ private:
+};
+
+/*! @class AMGXOperator
+    @ingroup MueLuAdapters
+    @brief Adapter for AmgX library from Nvidia.
+
+    Creates an AmgX Solver object with a Tpetra Matrix. Partial specialization of the template for data types supported by AmgX.
+*/
+template <class Node>
+class AMGXOperator<double, int, int, Node> : public TpetraOperator<double, int, int, Node> {
+ private:
+  typedef double SC;
+  typedef int LO;
+  typedef int GO;
+  typedef Node NO;
+
+  typedef Tpetra::Map<LO, GO, NO> Map;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> MultiVector;
+
+  void printMaps(Teuchos::RCP<const Teuchos::Comm<int> >& comm, const std::vector<std::vector<int> >& vec, const std::vector<int>& perm,
+                 const int* nbrs, const Map& map, const std::string& label) {
+    for (int p = 0; p < comm->getSize(); p++) {
+      if (comm->getRank() == p) {
+        std::cout << "========\n"
+                  << label << ", lid (gid), PID  " << p << "\n========" << std::endl;
+
+        for (size_t i = 0; i < vec.size(); ++i) {
+          std::cout << "   neighbor " << nbrs[i] << " :";
+          for (size_t j = 0; j < vec[i].size(); ++j)
+            std::cout << " " << vec[i][j] << " (" << map.getGlobalElement(perm[vec[i][j]]) << ")";
           std::cout << std::endl;
-        } else {
-          sleep(1);
         }
-        comm->barrier();
+        std::cout << std::endl;
+      } else {
+        sleep(1);
       }
+      comm->barrier();
     }
-
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-    AMGXOperator(const Teuchos::RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > &inA, Teuchos::ParameterList &paramListIn) {
-      RCP<const Teuchos::Comm<int> > comm = inA->getRowMap()->getComm();
-      int numProcs = comm->getSize();
-      int myRank   = comm->getRank();
-     
-
-      RCP<Teuchos::Time> amgxTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: initialize");
-      amgxTimer->start();
-      // Initialize
-      //AMGX_SAFE_CALL(AMGX_initialize());
-      //AMGX_SAFE_CALL(AMGX_initialize_plugins());
-    
-                
-      /*system*/
-      //AMGX_SAFE_CALL(AMGX_register_print_callback(&print_callback));
-      AMGX_SAFE_CALL(AMGX_install_signal_handler());
-      Teuchos::ParameterList configs = paramListIn.sublist("amgx:params", true);
-      if (configs.isParameter("json file")) {
-        AMGX_SAFE_CALL(AMGX_config_create_from_file(&Config_, (const char *) &configs.get<std::string>("json file")[0]));
-      } else {
-        std::ostringstream oss;
-        oss << "";
-        ParameterList::ConstIterator itr;
-        for (itr = configs.begin(); itr != configs.end(); ++itr) {
-          const std::string&    name  = configs.name(itr);
-          const ParameterEntry& entry = configs.entry(itr);
-          oss << name << "=" << filterValueToString(entry) << ", ";
-        }
-        oss << "\0";
-        std::string configString = oss.str();
-        if (configString == "") {
-          //print msg that using defaults
-          //GetOStream(Warnings0) << "Warning: No configuration parameters specified, using default AMGX configuration parameters. \n";
-        }
-        AMGX_SAFE_CALL(AMGX_config_create(&Config_, configString.c_str()));
+  }
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+  AMGXOperator(const Teuchos::RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& inA, Teuchos::ParameterList& paramListIn) {
+    RCP<const Teuchos::Comm<int> > comm = inA->getRowMap()->getComm();
+    int numProcs                        = comm->getSize();
+    int myRank                          = comm->getRank();
+
+    RCP<Teuchos::Time> amgxTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: initialize");
+    amgxTimer->start();
+    // Initialize
+    // AMGX_SAFE_CALL(AMGX_initialize());
+    // AMGX_SAFE_CALL(AMGX_initialize_plugins());
+
+    /*system*/
+    // AMGX_SAFE_CALL(AMGX_register_print_callback(&print_callback));
+    AMGX_SAFE_CALL(AMGX_install_signal_handler());
+    Teuchos::ParameterList configs = paramListIn.sublist("amgx:params", true);
+    if (configs.isParameter("json file")) {
+      AMGX_SAFE_CALL(AMGX_config_create_from_file(&Config_, (const char*)&configs.get<std::string>("json file")[0]));
+    } else {
+      std::ostringstream oss;
+      oss << "";
+      ParameterList::ConstIterator itr;
+      for (itr = configs.begin(); itr != configs.end(); ++itr) {
+        const std::string& name     = configs.name(itr);
+        const ParameterEntry& entry = configs.entry(itr);
+        oss << name << "=" << filterValueToString(entry) << ", ";
+      }
+      oss << "\0";
+      std::string configString = oss.str();
+      if (configString == "") {
+        // print msg that using defaults
+        // GetOStream(Warnings0) << "Warning: No configuration parameters specified, using default AMGX configuration parameters. \n";
       }
+      AMGX_SAFE_CALL(AMGX_config_create(&Config_, configString.c_str()));
+    }
 
-      // TODO: we probably need to add "exception_handling=1" to the parameter list
-      // to switch on internal error handling (with no need for AMGX_SAFE_CALL)
+    // TODO: we probably need to add "exception_handling=1" to the parameter list
+    // to switch on internal error handling (with no need for AMGX_SAFE_CALL)
 
-      //AMGX_SAFE_CALL(AMGX_config_add_parameters(&Config_, "exception_handling=1"))
+    // AMGX_SAFE_CALL(AMGX_config_add_parameters(&Config_, "exception_handling=1"))
 
 #define NEW_COMM
 #ifdef NEW_COMM
-      // NOTE: MPI communicator used in AMGX_resources_create must exist in the scope of AMGX_matrix_comm_from_maps_one_ring
-      // FIXME: fix for serial comm
-      RCP<const Teuchos::MpiComm<int> > tmpic = Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm->duplicate());
-      TEUCHOS_TEST_FOR_EXCEPTION(tmpic.is_null(), Exceptions::RuntimeError, "Communicator is not MpiComm");
+    // NOTE: MPI communicator used in AMGX_resources_create must exist in the scope of AMGX_matrix_comm_from_maps_one_ring
+    // FIXME: fix for serial comm
+    RCP<const Teuchos::MpiComm<int> > tmpic = Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm->duplicate());
+    TEUCHOS_TEST_FOR_EXCEPTION(tmpic.is_null(), Exceptions::RuntimeError, "Communicator is not MpiComm");
 
-      RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
-      MPI_Comm mpiComm = *rawMpiComm;
+    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
+    MPI_Comm mpiComm                                        = *rawMpiComm;
 #endif
 
-      // Construct AMGX resources
-      if (numProcs == 1) {
-        AMGX_resources_create_simple(&Resources_, Config_);
+    // Construct AMGX resources
+    if (numProcs == 1) {
+      AMGX_resources_create_simple(&Resources_, Config_);
 
-      } else {
-        int numGPUDevices;
-        cudaGetDeviceCount(&numGPUDevices);
-        int device[] = {(comm->getRank() % numGPUDevices)};
+    } else {
+      int numGPUDevices;
+      cudaGetDeviceCount(&numGPUDevices);
+      int device[] = {(comm->getRank() % numGPUDevices)};
 
-        AMGX_config_add_parameters(&Config_, "communicator=MPI");
+      AMGX_config_add_parameters(&Config_, "communicator=MPI");
 #ifdef NEW_COMM
-        AMGX_resources_create(&Resources_, Config_, &mpiComm, 1/* number of GPU devices utilized by this rank */, device);
+      AMGX_resources_create(&Resources_, Config_, &mpiComm, 1 /* number of GPU devices utilized by this rank */, device);
 #else
-        AMGX_resources_create(&Resources_, Config_, MPI_COMM_WORLD, 1/* number of GPU devices utilized by this rank */, device);
+      AMGX_resources_create(&Resources_, Config_, MPI_COMM_WORLD, 1 /* number of GPU devices utilized by this rank */, device);
 #endif
-      }
+    }
 
-      AMGX_Mode mode = AMGX_mode_dDDI;
-      AMGX_solver_create(&Solver_, Resources_, mode,  Config_);
-      AMGX_matrix_create(&A_,      Resources_, mode);
-      AMGX_vector_create(&X_,      Resources_, mode);
-      AMGX_vector_create(&Y_,      Resources_, mode);
+    AMGX_Mode mode = AMGX_mode_dDDI;
+    AMGX_solver_create(&Solver_, Resources_, mode, Config_);
+    AMGX_matrix_create(&A_, Resources_, mode);
+    AMGX_vector_create(&X_, Resources_, mode);
+    AMGX_vector_create(&Y_, Resources_, mode);
 
-      amgxTimer->stop();
-      amgxTimer->incrementNumCalls();
+    amgxTimer->stop();
+    amgxTimer->incrementNumCalls();
 
-      std::vector<int> amgx2muelu;
+    std::vector<int> amgx2muelu;
 
-      // Construct AMGX communication pattern
-      if (numProcs > 1) {
-        RCP<const Tpetra::Import<LO,GO,NO> > importer = inA->getCrsGraph()->getImporter();
+    // Construct AMGX communication pattern
+    if (numProcs > 1) {
+      RCP<const Tpetra::Import<LO, GO, NO> > importer = inA->getCrsGraph()->getImporter();
 
-        TEUCHOS_TEST_FOR_EXCEPTION(importer.is_null(), MueLu::Exceptions::RuntimeError, "The matrix A has no Import object.");
+      TEUCHOS_TEST_FOR_EXCEPTION(importer.is_null(), MueLu::Exceptions::RuntimeError, "The matrix A has no Import object.");
 
-        Tpetra::Distributor distributor = importer->getDistributor();
+      Tpetra::Distributor distributor = importer->getDistributor();
 
-        Array<int> sendRanks = distributor.getProcsTo();
-        Array<int> recvRanks = distributor.getProcsFrom();
+      Array<int> sendRanks = distributor.getProcsTo();
+      Array<int> recvRanks = distributor.getProcsFrom();
 
-        std::sort(sendRanks.begin(), sendRanks.end());
-        std::sort(recvRanks.begin(), recvRanks.end());
+      std::sort(sendRanks.begin(), sendRanks.end());
+      std::sort(recvRanks.begin(), recvRanks.end());
 
-        bool match = true;
-        if (sendRanks.size() != recvRanks.size()) {
-          match = false;
-        } else {
-          for (int i = 0; i < sendRanks.size(); i++) {
-            if (recvRanks[i] != sendRanks[i])
-              match = false;
-              break;
-          }
+      bool match = true;
+      if (sendRanks.size() != recvRanks.size()) {
+        match = false;
+      } else {
+        for (int i = 0; i < sendRanks.size(); i++) {
+          if (recvRanks[i] != sendRanks[i])
+            match = false;
+          break;
         }
-        TEUCHOS_TEST_FOR_EXCEPTION(!match, MueLu::Exceptions::RuntimeError, "AMGX requires that the processors that we send to and receive from are the same. "
-                                   "This is not the case: we send to {" << sendRanks << "} and receive from {" << recvRanks << "}");
-
-        int        num_neighbors = sendRanks.size();  // does not include the calling process
-        const int* neighbors     = &sendRanks[0];
-
-        // Later on, we'll have to organize the send and recv data by PIDs,
-        // i.e, a vector V of vectors, where V[i] is PID i's vector of data.
-        // Hence we need to be able to quickly look up  an array index
-        // associated with each PID.
-        Tpetra::Details::HashTable<int,int> hashTable(3*num_neighbors);
-        for (int i = 0; i < num_neighbors; i++)
-          hashTable.add(neighbors[i], i);
-
-        // Get some information out
-        ArrayView<const int> exportLIDs = importer->getExportLIDs();
-        ArrayView<const int> exportPIDs = importer->getExportPIDs();
-        Array<int> importPIDs;
-        Tpetra::Import_Util::getPids(*importer, importPIDs, true/* make local -1 */);
-
-        // Construct the reordering for AMGX as in AMGX_matrix_upload_all documentation
-        RCP<const Map> rowMap = inA->getRowMap();
-        RCP<const Map> colMap = inA->getColMap();
-
-        int N = rowMap->getLocalNumElements(), Nc = colMap->getLocalNumElements();
-        muelu2amgx_.resize(Nc, -1);
-
-        int numUniqExports = 0;
-        for (int i = 0; i < exportLIDs.size(); i++)
-          if (muelu2amgx_[exportLIDs[i]] == -1) {
-            numUniqExports++;
-            muelu2amgx_[exportLIDs[i]] = -2;
-          }
-
-        int localOffset = 0, exportOffset = N - numUniqExports;
-        // Go through exported LIDs and put them at the end of LIDs
-        for (int i = 0; i < exportLIDs.size(); i++)
-          if (muelu2amgx_[exportLIDs[i]] < 0) // exportLIDs are not unique
-            muelu2amgx_[exportLIDs[i]] = exportOffset++;
-        // Go through all non-export LIDs, and put them at the beginning of LIDs
-        for (int i = 0; i < N; i++)
-          if (muelu2amgx_[i] == -1)
-            muelu2amgx_[i] = localOffset++;
-        // Go through the tail (imported LIDs), and order those by neighbors
-        int importOffset = N;
-        for (int k = 0; k < num_neighbors; k++)
-          for (int i = 0; i < importPIDs.size(); i++)
-            if (importPIDs[i] != -1 && hashTable.get(importPIDs[i]) == k)
-              muelu2amgx_[i] = importOffset++;
-
-        amgx2muelu.resize(muelu2amgx_.size());
-        for (int i = 0; i < (int)muelu2amgx_.size(); i++)
-          amgx2muelu[muelu2amgx_[i]] = i;
-
-        // Construct send arrays
-        std::vector<std::vector<int> > sendDatas (num_neighbors);
-        std::vector<int>               send_sizes(num_neighbors, 0);
-        for (int i = 0; i < exportPIDs.size(); i++) {
-          int index = hashTable.get(exportPIDs[i]);
-          sendDatas [index].push_back(muelu2amgx_[exportLIDs[i]]);
-          send_sizes[index]++;
+      }
+      TEUCHOS_TEST_FOR_EXCEPTION(!match, MueLu::Exceptions::RuntimeError,
+                                 "AMGX requires that the processors that we send to and receive from are the same. "
+                                 "This is not the case: we send to {"
+                                     << sendRanks << "} and receive from {" << recvRanks << "}");
+
+      int num_neighbors    = sendRanks.size();  // does not include the calling process
+      const int* neighbors = &sendRanks[0];
+
+      // Later on, we'll have to organize the send and recv data by PIDs,
+      // i.e, a vector V of vectors, where V[i] is PID i's vector of data.
+      // Hence we need to be able to quickly look up  an array index
+      // associated with each PID.
+      Tpetra::Details::HashTable<int, int> hashTable(3 * num_neighbors);
+      for (int i = 0; i < num_neighbors; i++)
+        hashTable.add(neighbors[i], i);
+
+      // Get some information out
+      ArrayView<const int> exportLIDs = importer->getExportLIDs();
+      ArrayView<const int> exportPIDs = importer->getExportPIDs();
+      Array<int> importPIDs;
+      Tpetra::Import_Util::getPids(*importer, importPIDs, true /* make local -1 */);
+
+      // Construct the reordering for AMGX as in AMGX_matrix_upload_all documentation
+      RCP<const Map> rowMap = inA->getRowMap();
+      RCP<const Map> colMap = inA->getColMap();
+
+      int N = rowMap->getLocalNumElements(), Nc = colMap->getLocalNumElements();
+      muelu2amgx_.resize(Nc, -1);
+
+      int numUniqExports = 0;
+      for (int i = 0; i < exportLIDs.size(); i++)
+        if (muelu2amgx_[exportLIDs[i]] == -1) {
+          numUniqExports++;
+          muelu2amgx_[exportLIDs[i]] = -2;
         }
-        // FIXME: sendDatas must be sorted (based on GIDs)
 
-        std::vector<const int*> send_maps(num_neighbors);
-        for (int i = 0; i < num_neighbors; i++)
-          send_maps[i] = &(sendDatas[i][0]);
-
-        // Debugging
-        //        printMaps(comm, sendDatas, amgx2muelu, neighbors, *importer->getTargetMap(), "send_map_vector");
-
-        // Construct recv arrays
-        std::vector<std::vector<int> > recvDatas (num_neighbors);
-        std::vector<int>               recv_sizes(num_neighbors, 0);
+      int localOffset = 0, exportOffset = N - numUniqExports;
+      // Go through exported LIDs and put them at the end of LIDs
+      for (int i = 0; i < exportLIDs.size(); i++)
+        if (muelu2amgx_[exportLIDs[i]] < 0)  // exportLIDs are not unique
+          muelu2amgx_[exportLIDs[i]] = exportOffset++;
+      // Go through all non-export LIDs, and put them at the beginning of LIDs
+      for (int i = 0; i < N; i++)
+        if (muelu2amgx_[i] == -1)
+          muelu2amgx_[i] = localOffset++;
+      // Go through the tail (imported LIDs), and order those by neighbors
+      int importOffset = N;
+      for (int k = 0; k < num_neighbors; k++)
         for (int i = 0; i < importPIDs.size(); i++)
-          if (importPIDs[i] != -1) {
-            int index = hashTable.get(importPIDs[i]);
-            recvDatas [index].push_back(muelu2amgx_[i]);
-            recv_sizes[index]++;
-        }
-        // FIXME: recvDatas must be sorted (based on GIDs)
-
-        std::vector<const int*> recv_maps(num_neighbors);
-        for (int i = 0; i < num_neighbors; i++)
-          recv_maps[i] = &(recvDatas[i][0]);
-
-        // Debugging
-        //        printMaps(comm, recvDatas, amgx2muelu, neighbors, *importer->getTargetMap(), "recv_map_vector");
-
-        AMGX_SAFE_CALL(AMGX_matrix_comm_from_maps_one_ring(A_, 1, num_neighbors, neighbors, &send_sizes[0], &send_maps[0], &recv_sizes[0], &recv_maps[0]));
-
-        AMGX_vector_bind(X_, A_);
-        AMGX_vector_bind(Y_, A_);
+          if (importPIDs[i] != -1 && hashTable.get(importPIDs[i]) == k)
+            muelu2amgx_[i] = importOffset++;
+
+      amgx2muelu.resize(muelu2amgx_.size());
+      for (int i = 0; i < (int)muelu2amgx_.size(); i++)
+        amgx2muelu[muelu2amgx_[i]] = i;
+
+      // Construct send arrays
+      std::vector<std::vector<int> > sendDatas(num_neighbors);
+      std::vector<int> send_sizes(num_neighbors, 0);
+      for (int i = 0; i < exportPIDs.size(); i++) {
+        int index = hashTable.get(exportPIDs[i]);
+        sendDatas[index].push_back(muelu2amgx_[exportLIDs[i]]);
+        send_sizes[index]++;
       }
-
-      RCP<Teuchos::Time> matrixTransformTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transform matrix");
-      matrixTransformTimer->start();
-
-      ArrayRCP<const size_t> ia_s;
-      ArrayRCP<const int>    ja;
-      ArrayRCP<const double> a;
-      inA->getAllValues(ia_s, ja, a);
-
-      ArrayRCP<int> ia(ia_s.size());
-      for (int i = 0; i < ia.size(); i++)
-        ia[i] = Teuchos::as<int>(ia_s[i]);
-
-      N_      = inA->getLocalNumRows();
-      int nnz = inA->getLocalNumEntries();
-
-      matrixTransformTimer->stop();
-      matrixTransformTimer->incrementNumCalls();
-
-
-      // Upload matrix
-      // TODO Do we need to pin memory here through AMGX_pin_memory?
-      RCP<Teuchos::Time> matrixTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer matrix  CPU->GPU");
-      matrixTimer->start();
-      if (numProcs == 1) {
-        AMGX_matrix_upload_all(A_, N_, nnz, 1, 1, &ia[0], &ja[0], &a[0], NULL);
-
-      } else {
-        // Transform the matrix
-        std::vector<int>    ia_new(ia.size());
-        std::vector<int>    ja_new(ja.size());
-        std::vector<double> a_new (a.size());
-
-        ia_new[0] = 0;
-        for (int i = 0; i < N_; i++) {
-          int oldRow = amgx2muelu[i];
-
-          ia_new[i+1] = ia_new[i] + (ia[oldRow+1] - ia[oldRow]);
-
-          for (int j = ia[oldRow]; j < ia[oldRow+1]; j++) {
-            int offset = j - ia[oldRow];
-            ja_new[ia_new[i] + offset] = muelu2amgx_[ja[j]];
-            a_new [ia_new[i] + offset] = a[j];
-          }
-          // Do bubble sort on two arrays
-          // NOTE: There are multiple possible optimizations here (even of bubble sort)
-          bool swapped;
-          do {
-            swapped = false;
-
-            for (int j = ia_new[i]; j < ia_new[i+1]-1; j++)
-              if (ja_new[j] > ja_new[j+1]) {
-                std::swap(ja_new[j], ja_new[j+1]);
-                std::swap(a_new [j], a_new [j+1]);
-                swapped = true;
-              }
-          } while (swapped == true);
+      // FIXME: sendDatas must be sorted (based on GIDs)
+
+      std::vector<const int*> send_maps(num_neighbors);
+      for (int i = 0; i < num_neighbors; i++)
+        send_maps[i] = &(sendDatas[i][0]);
+
+      // Debugging
+      //        printMaps(comm, sendDatas, amgx2muelu, neighbors, *importer->getTargetMap(), "send_map_vector");
+
+      // Construct recv arrays
+      std::vector<std::vector<int> > recvDatas(num_neighbors);
+      std::vector<int> recv_sizes(num_neighbors, 0);
+      for (int i = 0; i < importPIDs.size(); i++)
+        if (importPIDs[i] != -1) {
+          int index = hashTable.get(importPIDs[i]);
+          recvDatas[index].push_back(muelu2amgx_[i]);
+          recv_sizes[index]++;
         }
+      // FIXME: recvDatas must be sorted (based on GIDs)
 
-        AMGX_matrix_upload_all(A_, N_, nnz, 1, 1, &ia_new[0], &ja_new[0], &a_new[0], NULL);
-      }
-      matrixTimer->stop();
-      matrixTimer->incrementNumCalls();
-
-      domainMap_ = inA->getDomainMap();
-      rangeMap_  = inA->getRangeMap();
-
-      RCP<Teuchos::Time> realSetupTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: setup (total)");
-      realSetupTimer->start();
-      AMGX_solver_setup(Solver_, A_);
-      realSetupTimer->stop();
-      realSetupTimer->incrementNumCalls();
-
-      vectorTimer1_ = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer vectors CPU->GPU");
-      vectorTimer2_ = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer vector  GPU->CPU");
-      solverTimer_  = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: Solve (total)");
-    }
-
-    //! Destructor.
-    virtual ~AMGXOperator()
-    {
-      // Comment this out if you need rebuild to work. This causes AMGX_solver_destroy memory issues.
-      AMGX_SAFE_CALL(AMGX_solver_destroy(Solver_));
-      AMGX_SAFE_CALL(AMGX_vector_destroy(X_));
-      AMGX_SAFE_CALL(AMGX_vector_destroy(Y_));
-      AMGX_SAFE_CALL(AMGX_matrix_destroy(A_));
-      AMGX_SAFE_CALL(AMGX_resources_destroy(Resources_));
-      AMGX_SAFE_CALL(AMGX_config_destroy(Config_));
-    }
-
-   
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Map> getDomainMap() const;
+      std::vector<const int*> recv_maps(num_neighbors);
+      for (int i = 0; i < num_neighbors; i++)
+        recv_maps[i] = &(recvDatas[i][0]);
 
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Map> getRangeMap() const;
+      // Debugging
+      //        printMaps(comm, recvDatas, amgx2muelu, neighbors, *importer->getTargetMap(), "recv_map_vector");
 
-    //! Returns in X the solution to the linear system AX=Y.
-    /*!
-       \param[out] X - Tpetra::MultiVector of dimension NumVectors containing the RHS of the linear system
-       \param[in]  Y - Tpetra::MultiVector of dimension NumVectors containing the solution to the linear system
-       */
-    void apply(const MultiVector& X, MultiVector& Y, Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               SC alpha = Teuchos::ScalarTraits<SC>::one(), SC beta = Teuchos::ScalarTraits<SC>::zero()) const;
+      AMGX_SAFE_CALL(AMGX_matrix_comm_from_maps_one_ring(A_, 1, num_neighbors, neighbors, &send_sizes[0], &send_maps[0], &recv_sizes[0], &recv_maps[0]));
 
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
-
-    RCP<MueLu::Hierarchy<SC,LO,GO,NO> > GetHierarchy() const {
-      throw Exceptions::RuntimeError("AMGXOperator does not hold a MueLu::Hierarchy object \n");
+      AMGX_vector_bind(X_, A_);
+      AMGX_vector_bind(Y_, A_);
     }
 
-    std::string filterValueToString(const Teuchos::ParameterEntry& entry ) {
-      return ( entry.isList() ? std::string("...") : toString(entry.getAny()) );
-    }
+    RCP<Teuchos::Time> matrixTransformTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transform matrix");
+    matrixTransformTimer->start();
 
-    int sizeA() {
-      int sizeX, sizeY, n;
-      AMGX_matrix_get_size(A_, &n, &sizeX, &sizeY);
-      return n;
-    }
+    ArrayRCP<const size_t> ia_s;
+    ArrayRCP<const int> ja;
+    ArrayRCP<const double> a;
+    inA->getAllValues(ia_s, ja, a);
 
-    int iters() {
-      int it;
-      AMGX_solver_get_iterations_number(Solver_, &it);
-      return it;
-    }
+    ArrayRCP<int> ia(ia_s.size());
+    for (int i = 0; i < ia.size(); i++)
+      ia[i] = Teuchos::as<int>(ia_s[i]);
 
-    AMGX_SOLVE_STATUS getStatus() {
-      AMGX_SOLVE_STATUS status;
-      AMGX_solver_get_status(Solver_, &status);
-      return status;
-    }
+    N_      = inA->getLocalNumRows();
+    int nnz = inA->getLocalNumEntries();
 
+    matrixTransformTimer->stop();
+    matrixTransformTimer->incrementNumCalls();
 
-  private:
-    AMGX_solver_handle      Solver_;
-    AMGX_resources_handle   Resources_;
-    AMGX_config_handle      Config_;
-    AMGX_matrix_handle      A_;
-    AMGX_vector_handle      X_;
-    AMGX_vector_handle      Y_;
-    int                     N_;
+    // Upload matrix
+    // TODO Do we need to pin memory here through AMGX_pin_memory?
+    RCP<Teuchos::Time> matrixTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer matrix  CPU->GPU");
+    matrixTimer->start();
+    if (numProcs == 1) {
+      AMGX_matrix_upload_all(A_, N_, nnz, 1, 1, &ia[0], &ja[0], &a[0], NULL);
 
-    RCP<const Map>          domainMap_;
-    RCP<const Map>          rangeMap_;
+    } else {
+      // Transform the matrix
+      std::vector<int> ia_new(ia.size());
+      std::vector<int> ja_new(ja.size());
+      std::vector<double> a_new(a.size());
 
-    std::vector<int>        muelu2amgx_;
+      ia_new[0] = 0;
+      for (int i = 0; i < N_; i++) {
+        int oldRow = amgx2muelu[i];
 
-    RCP<Teuchos::Time>      vectorTimer1_;
-    RCP<Teuchos::Time>      vectorTimer2_;
-    RCP<Teuchos::Time>      solverTimer_;
-  };
+        ia_new[i + 1] = ia_new[i] + (ia[oldRow + 1] - ia[oldRow]);
 
-} // namespace
+        for (int j = ia[oldRow]; j < ia[oldRow + 1]; j++) {
+          int offset                 = j - ia[oldRow];
+          ja_new[ia_new[i] + offset] = muelu2amgx_[ja[j]];
+          a_new[ia_new[i] + offset]  = a[j];
+        }
+        // Do bubble sort on two arrays
+        // NOTE: There are multiple possible optimizations here (even of bubble sort)
+        bool swapped;
+        do {
+          swapped = false;
+
+          for (int j = ia_new[i]; j < ia_new[i + 1] - 1; j++)
+            if (ja_new[j] > ja_new[j + 1]) {
+              std::swap(ja_new[j], ja_new[j + 1]);
+              std::swap(a_new[j], a_new[j + 1]);
+              swapped = true;
+            }
+        } while (swapped == true);
+      }
 
-#endif //HAVE_MUELU_AMGX
-#endif // MUELU_AMGXOPERATOR_DECL_HPP
+      AMGX_matrix_upload_all(A_, N_, nnz, 1, 1, &ia_new[0], &ja_new[0], &a_new[0], NULL);
+    }
+    matrixTimer->stop();
+    matrixTimer->incrementNumCalls();
+
+    domainMap_ = inA->getDomainMap();
+    rangeMap_  = inA->getRangeMap();
+
+    RCP<Teuchos::Time> realSetupTimer = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: setup (total)");
+    realSetupTimer->start();
+    AMGX_solver_setup(Solver_, A_);
+    realSetupTimer->stop();
+    realSetupTimer->incrementNumCalls();
+
+    vectorTimer1_ = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer vectors CPU->GPU");
+    vectorTimer2_ = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: transfer vector  GPU->CPU");
+    solverTimer_  = Teuchos::TimeMonitor::getNewTimer("MueLu: AMGX: Solve (total)");
+  }
+
+  //! Destructor.
+  virtual ~AMGXOperator() {
+    // Comment this out if you need rebuild to work. This causes AMGX_solver_destroy memory issues.
+    AMGX_SAFE_CALL(AMGX_solver_destroy(Solver_));
+    AMGX_SAFE_CALL(AMGX_vector_destroy(X_));
+    AMGX_SAFE_CALL(AMGX_vector_destroy(Y_));
+    AMGX_SAFE_CALL(AMGX_matrix_destroy(A_));
+    AMGX_SAFE_CALL(AMGX_resources_destroy(Resources_));
+    AMGX_SAFE_CALL(AMGX_config_destroy(Config_));
+  }
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Map> getDomainMap() const;
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Map> getRangeMap() const;
+
+  //! Returns in X the solution to the linear system AX=Y.
+  /*!
+     \param[out] X - Tpetra::MultiVector of dimension NumVectors containing the RHS of the linear system
+     \param[in]  Y - Tpetra::MultiVector of dimension NumVectors containing the solution to the linear system
+     */
+  void apply(const MultiVector& X, MultiVector& Y, Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             SC alpha = Teuchos::ScalarTraits<SC>::one(), SC beta = Teuchos::ScalarTraits<SC>::zero()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
+
+  RCP<MueLu::Hierarchy<SC, LO, GO, NO> > GetHierarchy() const {
+    throw Exceptions::RuntimeError("AMGXOperator does not hold a MueLu::Hierarchy object \n");
+  }
+
+  std::string filterValueToString(const Teuchos::ParameterEntry& entry) {
+    return (entry.isList() ? std::string("...") : toString(entry.getAny()));
+  }
+
+  int sizeA() {
+    int sizeX, sizeY, n;
+    AMGX_matrix_get_size(A_, &n, &sizeX, &sizeY);
+    return n;
+  }
+
+  int iters() {
+    int it;
+    AMGX_solver_get_iterations_number(Solver_, &it);
+    return it;
+  }
+
+  AMGX_SOLVE_STATUS getStatus() {
+    AMGX_SOLVE_STATUS status;
+    AMGX_solver_get_status(Solver_, &status);
+    return status;
+  }
+
+ private:
+  AMGX_solver_handle Solver_;
+  AMGX_resources_handle Resources_;
+  AMGX_config_handle Config_;
+  AMGX_matrix_handle A_;
+  AMGX_vector_handle X_;
+  AMGX_vector_handle Y_;
+  int N_;
+
+  RCP<const Map> domainMap_;
+  RCP<const Map> rangeMap_;
+
+  std::vector<int> muelu2amgx_;
+
+  RCP<Teuchos::Time> vectorTimer1_;
+  RCP<Teuchos::Time> vectorTimer2_;
+  RCP<Teuchos::Time> solverTimer_;
+};
+
+}  // namespace MueLu
+
+#endif  // HAVE_MUELU_AMGX
+#endif  // MUELU_AMGXOPERATOR_DECL_HPP
diff --git a/packages/muelu/adapters/amgx/MueLu_AMGXOperator_def.hpp b/packages/muelu/adapters/amgx/MueLu_AMGXOperator_def.hpp
index 767ab6a540b9..afaee1ef36a9 100644
--- a/packages/muelu/adapters/amgx/MueLu_AMGXOperator_def.hpp
+++ b/packages/muelu/adapters/amgx/MueLu_AMGXOperator_def.hpp
@@ -47,100 +47,98 @@
 #ifndef MUELU_AMGXOPERATOR_DEF_HPP
 #define MUELU_AMGXOPERATOR_DEF_HPP
 
-
-#if defined (HAVE_MUELU_AMGX)
+#if defined(HAVE_MUELU_AMGX)
 #include "MueLu_AMGXOperator_decl.hpp"
 
 namespace MueLu {
 
-  template<class Node>
-  Teuchos::RCP<const Tpetra::Map<int,int,Node> >
-  AMGXOperator<double,int,int,Node>::getDomainMap() const {
-    return domainMap_;
-  }
+template <class Node>
+Teuchos::RCP<const Tpetra::Map<int, int, Node> >
+AMGXOperator<double, int, int, Node>::getDomainMap() const {
+  return domainMap_;
+}
 
-  template<class Node>
-  Teuchos::RCP<const Tpetra::Map<int,int,Node> > AMGXOperator<double,int,int,Node>::getRangeMap() const {
-    return rangeMap_;
-  }
+template <class Node>
+Teuchos::RCP<const Tpetra::Map<int, int, Node> > AMGXOperator<double, int, int, Node>::getRangeMap() const {
+  return rangeMap_;
+}
 
-  template<class Node>
-  void AMGXOperator<double,int,int,Node>::apply(const Tpetra::MultiVector<double,int,int,Node>& X,
-                                                Tpetra::MultiVector<double,int,int,Node>&       Y,
-                                                Teuchos::ETransp mode, double alpha, double beta) const {
+template <class Node>
+void AMGXOperator<double, int, int, Node>::apply(const Tpetra::MultiVector<double, int, int, Node>& X,
+                                                 Tpetra::MultiVector<double, int, int, Node>& Y,
+                                                 Teuchos::ETransp mode, double alpha, double beta) const {
+  RCP<const Teuchos::Comm<int> > comm = Y.getMap()->getComm();
 
-    RCP<const Teuchos::Comm<int> > comm = Y.getMap()->getComm();
-    
-    ArrayRCP<const double> mueluXdata, amgxXdata;
-    ArrayRCP<double>       mueluYdata, amgxYdata;
+  ArrayRCP<const double> mueluXdata, amgxXdata;
+  ArrayRCP<double> mueluYdata, amgxYdata;
 
-    try {
-      for (int i = 0; i < (int)Y.getNumVectors(); i++) {
-        {
-          vectorTimer1_->start();
+  try {
+    for (int i = 0; i < (int)Y.getNumVectors(); i++) {
+      {
+        vectorTimer1_->start();
 
-          mueluXdata = X.getData(i);
-          mueluYdata = Y.getDataNonConst(i);
+        mueluXdata = X.getData(i);
+        mueluYdata = Y.getDataNonConst(i);
 
-          if (comm->getSize() == 1) {
-            amgxXdata = mueluXdata;
-            amgxYdata = mueluYdata;
+        if (comm->getSize() == 1) {
+          amgxXdata = mueluXdata;
+          amgxYdata = mueluYdata;
 
-          } else {
-            int n = mueluXdata.size();
+        } else {
+          int n = mueluXdata.size();
 
-            amgxXdata.resize(n);
-            amgxYdata.resize(n);
+          amgxXdata.resize(n);
+          amgxYdata.resize(n);
 
-            ArrayRCP<double> amgxXdata_nonConst = Teuchos::arcp_const_cast<double>(amgxXdata);
-            for (int j = 0; j < n; j++) {
-              amgxXdata_nonConst[muelu2amgx_[j]] = mueluXdata[j];
-              amgxYdata         [muelu2amgx_[j]] = mueluYdata[j];
-            }
+          ArrayRCP<double> amgxXdata_nonConst = Teuchos::arcp_const_cast<double>(amgxXdata);
+          for (int j = 0; j < n; j++) {
+            amgxXdata_nonConst[muelu2amgx_[j]] = mueluXdata[j];
+            amgxYdata[muelu2amgx_[j]]          = mueluYdata[j];
           }
+        }
 
-          AMGX_vector_upload(X_, N_, 1, &amgxXdata[0]);
-          AMGX_vector_upload(Y_, N_, 1, &amgxYdata[0]);
+        AMGX_vector_upload(X_, N_, 1, &amgxXdata[0]);
+        AMGX_vector_upload(Y_, N_, 1, &amgxYdata[0]);
 
-          vectorTimer1_->stop();
-          vectorTimer1_->incrementNumCalls();
-        }
-       
-        // Solve the system and time.
-        solverTimer_->start();
-        AMGX_solver_solve(Solver_, X_, Y_);
-        solverTimer_->stop();
-        solverTimer_->incrementNumCalls();
+        vectorTimer1_->stop();
+        vectorTimer1_->incrementNumCalls();
+      }
 
-        {
-          vectorTimer2_->start();
+      // Solve the system and time.
+      solverTimer_->start();
+      AMGX_solver_solve(Solver_, X_, Y_);
+      solverTimer_->stop();
+      solverTimer_->incrementNumCalls();
 
-          AMGX_vector_download(Y_,      &amgxYdata[0]);
+      {
+        vectorTimer2_->start();
 
-          if (comm->getSize() > 1) {
-            int n = mueluYdata.size();
+        AMGX_vector_download(Y_, &amgxYdata[0]);
 
-            for (int j = 0; j < n; j++)
-              mueluYdata[j] = amgxYdata[muelu2amgx_[j]];
-          }
+        if (comm->getSize() > 1) {
+          int n = mueluYdata.size();
 
-          vectorTimer2_->stop();
-          vectorTimer2_->incrementNumCalls();
+          for (int j = 0; j < n; j++)
+            mueluYdata[j] = amgxYdata[muelu2amgx_[j]];
         }
-      }
 
-    } catch (std::exception& e) {
-      std::string errMsg = std::string("Caught an exception in MueLu::AMGXOperator::Apply():\n") + e.what() + "\n";
-      throw Exceptions::RuntimeError(errMsg);
+        vectorTimer2_->stop();
+        vectorTimer2_->incrementNumCalls();
+      }
     }
-  }
 
-  template<class Node>
-  bool AMGXOperator<double,int,int,Node>::hasTransposeApply() const {
-    return false;
+  } catch (std::exception& e) {
+    std::string errMsg = std::string("Caught an exception in MueLu::AMGXOperator::Apply():\n") + e.what() + "\n";
+    throw Exceptions::RuntimeError(errMsg);
   }
+}
+
+template <class Node>
+bool AMGXOperator<double, int, int, Node>::hasTransposeApply() const {
+  return false;
+}
 
-} // namespace
-#endif //if defined(HAVE_MUELU_AMGX)
+}  // namespace MueLu
+#endif  // if defined(HAVE_MUELU_AMGX)
 
-#endif //ifdef MUELU_AMGXOPERATOR_DEF_HPP
+#endif  // ifdef MUELU_AMGXOPERATOR_DEF_HPP
diff --git a/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.cpp b/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.cpp
index e0fe383ca373..2170d252c188 100644
--- a/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.cpp
+++ b/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.cpp
@@ -48,25 +48,20 @@
 
 namespace MueLu {
 
-  void MueLu_AMGX_initialize()
-  {
-    AMGX_SAFE_CALL(AMGX_initialize());
-  }
+void MueLu_AMGX_initialize() {
+  AMGX_SAFE_CALL(AMGX_initialize());
+}
 
-  void MueLu_AMGX_initialize_plugins()
-  {
-    AMGX_SAFE_CALL(AMGX_initialize_plugins());
-  }
+void MueLu_AMGX_initialize_plugins() {
+  AMGX_SAFE_CALL(AMGX_initialize_plugins());
+}
 
-  void MueLu_AMGX_finalize()
-  {
-    AMGX_SAFE_CALL(AMGX_finalize());
-  }
-
-  void MueLu_AMGX_finalize_plugins()
-  {
-   AMGX_print_summary();
-   AMGX_SAFE_CALL(AMGX_finalize_plugins());
-  }
-} // namespace
+void MueLu_AMGX_finalize() {
+  AMGX_SAFE_CALL(AMGX_finalize());
+}
 
+void MueLu_AMGX_finalize_plugins() {
+  AMGX_print_summary();
+  AMGX_SAFE_CALL(AMGX_finalize_plugins());
+}
+}  // namespace MueLu
diff --git a/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.hpp b/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.hpp
index 6b645310dd7d..fe4c1c9791c9 100644
--- a/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.hpp
+++ b/packages/muelu/adapters/amgx/MueLu_AMGX_Setup.hpp
@@ -51,12 +51,12 @@
 
 namespace MueLu {
 
-  void MueLu_AMGX_initialize();
-  void MueLu_AMGX_initialize_plugins();
+void MueLu_AMGX_initialize();
+void MueLu_AMGX_initialize_plugins();
 
-  void MueLu_AMGX_finalize();
-  void MueLu_AMGX_finalize_plugins();
-}
+void MueLu_AMGX_finalize();
+void MueLu_AMGX_finalize_plugins();
+}  // namespace MueLu
 
-#endif //HAVE_MUELU_AMGX
-#endif //ifndef MUELU_AMGX_SETUP_DEF_HPP
+#endif  // HAVE_MUELU_AMGX
+#endif  // ifndef MUELU_AMGX_SETUP_DEF_HPP
diff --git a/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.cpp b/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.cpp
index 515f950ced6e..a0e2efb6d066 100644
--- a/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.cpp
+++ b/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.cpp
@@ -6,7 +6,7 @@
 #include "Xpetra_CrsMatrixWrap.hpp"
 #include "Xpetra_EpetraCrsMatrix.hpp"
 
-#include "MueLu_config.hpp"             // for HAVE_MUELU_DEBUG
+#include "MueLu_config.hpp"  // for HAVE_MUELU_DEBUG
 #include "MueLu_RefMaxwell.hpp"
 #include "MueLu_Exceptions.hpp"
 
@@ -18,30 +18,30 @@ namespace MueLu {
 
 int AztecEpetraOperator::ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const {
   try {
-     // There is no rcpFromRef(const T&), so we need to do const_cast
-     const Xpetra::EpetraMultiVectorT<GO,NO> eX(Teuchos::rcpFromRef(const_cast<Epetra_MultiVector&>(X)));
-     Xpetra::EpetraMultiVectorT<GO,NO>       eY(Teuchos::rcpFromRef(Y));
-     // Generally, we assume two different vectors, but AztecOO uses a single vector
-     if (X.Values() == Y.Values()) {
-       // X and Y point to the same memory, use an additional vector
-       Teuchos::RCP<Xpetra::EpetraMultiVectorT<GO,NO> > tmpY = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GO,NO>(eY.getMap(), eY.getNumVectors()));
-       tmpY->putScalar(0.0);
-       xOp_->apply(eX, *tmpY);
-       // deep copy solution from MueLu
-       eY.update(1.0, *tmpY, 0.0);
-     } else {
-       // X and Y point to different memory, pass the vectors through
-       eY.putScalar(0.0);
-       xOp_->apply(eX, eY);
-     }
-
-   } catch (std::exception& e) {
-     //TODO: error msg directly on std::cerr?
-     std::cerr << "Caught an exception in MueLu::AztecEpetraOperator::ApplyInverse():" << std::endl
-         << e.what() << std::endl;
-     return -1;
-   }
-   return 0;
+    // There is no rcpFromRef(const T&), so we need to do const_cast
+    const Xpetra::EpetraMultiVectorT<GO, NO> eX(Teuchos::rcpFromRef(const_cast<Epetra_MultiVector&>(X)));
+    Xpetra::EpetraMultiVectorT<GO, NO> eY(Teuchos::rcpFromRef(Y));
+    // Generally, we assume two different vectors, but AztecOO uses a single vector
+    if (X.Values() == Y.Values()) {
+      // X and Y point to the same memory, use an additional vector
+      Teuchos::RCP<Xpetra::EpetraMultiVectorT<GO, NO>> tmpY = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GO, NO>(eY.getMap(), eY.getNumVectors()));
+      tmpY->putScalar(0.0);
+      xOp_->apply(eX, *tmpY);
+      // deep copy solution from MueLu
+      eY.update(1.0, *tmpY, 0.0);
+    } else {
+      // X and Y point to different memory, pass the vectors through
+      eY.putScalar(0.0);
+      xOp_->apply(eX, eY);
+    }
+
+  } catch (std::exception& e) {
+    // TODO: error msg directly on std::cerr?
+    std::cerr << "Caught an exception in MueLu::AztecEpetraOperator::ApplyInverse():" << std::endl
+              << e.what() << std::endl;
+    return -1;
+  }
+  return 0;
 }
 
 const Epetra_Comm& AztecEpetraOperator::Comm() const {
@@ -50,12 +50,12 @@ const Epetra_Comm& AztecEpetraOperator::Comm() const {
 }
 
 const Epetra_Map& AztecEpetraOperator::OperatorDomainMap() const {
-  if(Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(xOp_) != Teuchos::null) {
-    RCP<Xpetra::Matrix<SC,LO,GO,NO> > A = Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(xOp_)->getJacobian();
-    RCP<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >(A);
+  if (Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO>>(xOp_) != Teuchos::null) {
+    RCP<Xpetra::Matrix<SC, LO, GO, NO>> A                  = Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO>>(xOp_)->getJacobian();
+    RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(A);
     if (crsOp == Teuchos::null)
       throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO>>(crsOp->getCrsMatrix());
+    const RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(crsOp->getCrsMatrix());
     if (tmp_ECrsMtx == Teuchos::null)
       throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
     return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst()->DomainMap();
@@ -67,14 +67,13 @@ const Epetra_Map& AztecEpetraOperator::OperatorDomainMap() const {
   return Xpetra::toEpetra(map);
 }
 
-const Epetra_Map & AztecEpetraOperator::OperatorRangeMap() const {
-
-  if(Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(xOp_) != Teuchos::null) {
-    RCP<Xpetra::Matrix<SC,LO,GO,NO> > A = Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(xOp_)->getJacobian();
-    RCP<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >(A);
+const Epetra_Map& AztecEpetraOperator::OperatorRangeMap() const {
+  if (Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO>>(xOp_) != Teuchos::null) {
+    RCP<Xpetra::Matrix<SC, LO, GO, NO>> A                  = Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO>>(xOp_)->getJacobian();
+    RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(A);
     if (crsOp == Teuchos::null)
       throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO>>(crsOp->getCrsMatrix());
+    const RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(crsOp->getCrsMatrix());
     if (tmp_ECrsMtx == Teuchos::null)
       throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
     return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst()->RangeMap();
@@ -86,7 +85,7 @@ const Epetra_Map & AztecEpetraOperator::OperatorRangeMap() const {
   return Xpetra::toEpetra(map);
 }
 
-}
+}  // namespace MueLu
 
 #endif /*#if defined(HAVE_MUELU_SERIAL) and defined(HAVE_MUELU_EPETRA)*/
 
diff --git a/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.hpp b/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.hpp
index 5a9fe3d7ef86..feca078f032e 100644
--- a/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.hpp
+++ b/packages/muelu/adapters/aztecoo/MueLu_AztecEpetraOperator.hpp
@@ -16,107 +16,104 @@ namespace MueLu {
 
     Currently only used for RefMaxwell.
 */
-  class AztecEpetraOperator : public Epetra_Operator {
-    typedef double              SC;
-    typedef int                 LO;
-    typedef int                 GO;
-    typedef Xpetra::EpetraNode  NO;
+class AztecEpetraOperator : public Epetra_Operator {
+  typedef double SC;
+  typedef int LO;
+  typedef int GO;
+  typedef Xpetra::EpetraNode NO;
 
-    typedef Xpetra::Map<LO,GO,NO>                        Map;
-    typedef Xpetra::EpetraMapT<GO,NO>                    EpetraMap;
-    typedef Xpetra::Operator<SC,LO,GO,NO>                Operator;
+  typedef Xpetra::Map<LO, GO, NO> Map;
+  typedef Xpetra::EpetraMapT<GO, NO> EpetraMap;
+  typedef Xpetra::Operator<SC, LO, GO, NO> Operator;
 
-  public:
+ public:
+  //! @name Constructor/Destructor
+  //@{
 
-    //! @name Constructor/Destructor
-    //@{
+  //! Constructor
+  AztecEpetraOperator(const Teuchos::RCP<Operator>& Op)
+    : xOp_(Op) {}
 
-    //! Constructor
-    AztecEpetraOperator(const Teuchos::RCP<Operator>& Op) : xOp_(Op) { }
+  //! Destructor.
+  virtual ~AztecEpetraOperator() {}
 
-    //! Destructor.
-    virtual ~AztecEpetraOperator() { }
+  //@}
 
-    //@}
+  int SetUseTranspose(bool /* UseTransposeBool */) { return -1; }
 
-    int SetUseTranspose(bool /* UseTransposeBool */) { return -1; }
+  //! @name Mathematical functions
+  //@{
 
-    //! @name Mathematical functions
-    //@{
+  //! Returns the result of a Epetra_Operator applied to a Epetra_MultiVector X in Y.
+  /*!
+    \param In
+    X - A Epetra_MultiVector of dimension NumVectors to multiply with matrix.
+    \param Out
+    Y -A Epetra_MultiVector of dimension NumVectors containing result.
 
-    //! Returns the result of a Epetra_Operator applied to a Epetra_MultiVector X in Y.
-    /*!
-      \param In
-      X - A Epetra_MultiVector of dimension NumVectors to multiply with matrix.
-      \param Out
-      Y -A Epetra_MultiVector of dimension NumVectors containing result.
+    \return Integer error code, set to 0 if successful.
+  */
+  int Apply(const Epetra_MultiVector& /* X */, Epetra_MultiVector& /* Y */) const { return -1; }
 
-      \return Integer error code, set to 0 if successful.
-    */
-    int Apply(const Epetra_MultiVector& /* X */, Epetra_MultiVector& /* Y */) const { return -1; }
+  //! Returns the result of a Epetra_Operator inverse applied to an Epetra_MultiVector X in Y.
+  /*!
+    \param In
+    X - A Epetra_MultiVector of dimension NumVectors to solve for.
+    \param Out
+    Y -A Epetra_MultiVector of dimension NumVectors containing result.
 
-    //! Returns the result of a Epetra_Operator inverse applied to an Epetra_MultiVector X in Y.
-    /*!
-      \param In
-      X - A Epetra_MultiVector of dimension NumVectors to solve for.
-      \param Out
-      Y -A Epetra_MultiVector of dimension NumVectors containing result.
+    \return Integer error code, set to 0 if successful.
 
-      \return Integer error code, set to 0 if successful.
+    \warning In order to work with AztecOO, any implementation of this method must
+    support the case where X and Y are the same object.
+  */
+  int ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const;
 
-      \warning In order to work with AztecOO, any implementation of this method must
-      support the case where X and Y are the same object.
-    */
-    int ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const;
+  //! Returns the infinity norm of the global matrix.
+  /* Returns the quantity \f$ \| A \|_\infty\f$ such that
+     \f[\| A \|_\infty = \max_{1\lei\lem} \sum_{j=1}^n |a_{ij}| \f].
 
-    //! Returns the infinity norm of the global matrix.
-    /* Returns the quantity \f$ \| A \|_\infty\f$ such that
-       \f[\| A \|_\infty = \max_{1\lei\lem} \sum_{j=1}^n |a_{ij}| \f].
+     \warning This method must not be called unless HasNormInf() returns true.
+  */
+  double NormInf() const { return 0; }
+  //@}
 
-       \warning This method must not be called unless HasNormInf() returns true.
-    */
-    double NormInf() const { return 0; }
-    //@}
+  //! @name Attribute access functions
+  //@{
 
-    //! @name Attribute access functions
-    //@{
+  //! Returns a character string describing the operator
+  const char* Label() const { return "MueLu::AztecEpetraOperator"; }
 
-    //! Returns a character string describing the operator
-    const char * Label() const { return "MueLu::AztecEpetraOperator"; }
+  //! Returns the current UseTranspose setting.
+  bool UseTranspose() const { return false; }
 
-    //! Returns the current UseTranspose setting.
-    bool UseTranspose() const { return false; }
+  //! Returns true if the \e this object can provide an approximate Inf-norm, false otherwise.
+  bool HasNormInf() const { return 0; }
 
-    //! Returns true if the \e this object can provide an approximate Inf-norm, false otherwise.
-    bool HasNormInf() const { return 0; }
+  //! Returns a pointer to the Epetra_Comm communicator associated with this operator.
+  const Epetra_Comm& Comm() const;
 
-    //! Returns a pointer to the Epetra_Comm communicator associated with this operator.
-    const Epetra_Comm & Comm() const;
+  //! Returns the Epetra_Map object associated with the domain of this operator.
+  const Epetra_Map& OperatorDomainMap() const;
 
-    //! Returns the Epetra_Map object associated with the domain of this operator.
-    const Epetra_Map & OperatorDomainMap() const;
+  //! Returns the Epetra_Map object associated with the range of this operator.
+  const Epetra_Map& OperatorRangeMap() const;
 
-    //! Returns the Epetra_Map object associated with the range of this operator.
-    const Epetra_Map & OperatorRangeMap() const;
+  //@}
 
-    //@}
+  //! @name MueLu specific
+  //@{
 
-    //! @name MueLu specific
-    //@{
+  //! Direct access to the underlying Xpetra::Operator.
+  Teuchos::RCP<Operator> GetOperator() const { return xOp_; }
 
-    //! Direct access to the underlying Xpetra::Operator.
-    Teuchos::RCP<Operator> GetOperator() const { return xOp_; }
+  //@}
 
-    //@}
+ private:
+  Teuchos::RCP<Operator> xOp_;
+};
 
-
-  private:
-
-    Teuchos::RCP<Operator> xOp_;
-
-  };
-
-} // namespace
+}  // namespace MueLu
 
 #endif
 
diff --git a/packages/muelu/adapters/belos/BelosMueLuAdapter.hpp b/packages/muelu/adapters/belos/BelosMueLuAdapter.hpp
index ca66967e24d9..1512269be928 100644
--- a/packages/muelu/adapters/belos/BelosMueLuAdapter.hpp
+++ b/packages/muelu/adapters/belos/BelosMueLuAdapter.hpp
@@ -67,397 +67,397 @@
 #include "MueLu_Hierarchy.hpp"
 
 namespace Belos {
-  using Teuchos::RCP;
-  using Teuchos::rcpFromRef;
+using Teuchos::RCP;
+using Teuchos::rcpFromRef;
 
-  //
-  //! @name MueLu Adapter Exceptions
-  //@{
-
-  /** \brief MueLuOpFailure is thrown when a return value from an MueLu
-   * call on an Xpetra::Operator or MueLu::Hierarchy is non-zero.
-   */
-  class MueLuOpFailure : public BelosError {
-  public:
-    MueLuOpFailure(const std::string& what_arg) : BelosError(what_arg) {}
-  };
-
-  /*! @class MueLuOp
-   *
-   * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
-   *        call which represents the effect of the multigrid preconditioner on a given vector.
-   *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
-   *
-   *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
-   *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
-   *
-   *        This is the general implementation for Tpetra only.
-   */
-  template <class Scalar,
-            class LocalOrdinal  = int,
-            class GlobalOrdinal = LocalOrdinal,
-            class Node          = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class MueLuOp :
-      public OperatorT<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+//
+//! @name MueLu Adapter Exceptions
+//@{
+
+/** \brief MueLuOpFailure is thrown when a return value from an MueLu
+ * call on an Xpetra::Operator or MueLu::Hierarchy is non-zero.
+ */
+class MueLuOpFailure : public BelosError {
+ public:
+  MueLuOpFailure(const std::string& what_arg)
+    : BelosError(what_arg) {}
+};
+
+/*! @class MueLuOp
+ *
+ * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
+ *        call which represents the effect of the multigrid preconditioner on a given vector.
+ *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
+ *
+ *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
+ *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
+ *
+ *        This is the general implementation for Tpetra only.
+ */
+template <class Scalar,
+          class LocalOrdinal  = int,
+          class GlobalOrdinal = LocalOrdinal,
+          class Node          = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class MueLuOp : public OperatorT<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
 #ifdef HAVE_XPETRA_TPETRA
-    , public OperatorT<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+  ,
+                public OperatorT<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
 #endif
-  {
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
+{
+ public:
+  //! @name Constructor/Destructor
+  //@{
 
-    //! Default constructor
-    MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >&H) : Hierarchy_(H) {}
+  //! Default constructor
+  MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H) {}
 #ifdef HAVE_MUELU_AMGX
-    MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) : AMGX_(A) {}
+  MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A)
+    : AMGX_(A) {}
 #endif
-    //! Destructor.
-    virtual ~MueLuOp() {}
-    //@}
-
-    //! @name Operator application method
-    //@{
+  //! Destructor.
+  virtual ~MueLuOp() {}
+  //@}
 
-    /*! \brief This routine takes the Xpetra::MultiVector \c x and applies the operator
-      to it resulting in the Xpetra::MultiVector \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS ) const {
+  //! @name Operator application method
+  //@{
 
-      TEUCHOS_TEST_FOR_EXCEPTION(trans!=NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+  /*! \brief This routine takes the Xpetra::MultiVector \c x and applies the operator
+    to it resulting in the Xpetra::MultiVector \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
 
-      // This does not matter for Hierarchy, but matters for AMGX
-      y.putScalar(0.0);
+    // This does not matter for Hierarchy, but matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
-
-        AMGX_->apply(tX, tY);
+    if (!AMGX_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
 
-      }
-#endif
-      if (!Hierarchy_.is_null())
-        Hierarchy_->Iterate(x, y, 1, true);
+      AMGX_->apply(tX, tY);
     }
-    //@}
+#endif
+    if (!Hierarchy_.is_null())
+      Hierarchy_->Iterate(x, y, 1, true);
+  }
+  //@}
 
 #ifdef HAVE_XPETRA_TPETRA
-    // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
-    /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
-      to it resulting in the Tpetra::MultiVector \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS ) const {
-
-      TEUCHOS_TEST_FOR_EXCEPTION(trans!=NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
-
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
-      y.putScalar(0.0);
+  // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
+  /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
+    to it resulting in the Tpetra::MultiVector \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null())
-        AMGX_->apply(x, y);
+    if (!AMGX_.is_null())
+      AMGX_->apply(x, y);
 #endif
 
-      if (!Hierarchy_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &>(x);
+    if (!Hierarchy_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(x);
 
-        const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
-        Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY(rcpFromRef(y));
-        Hierarchy_->Iterate(tX, tY, 1, true);
-      }
+      const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
+      Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY(rcpFromRef(y));
+      Hierarchy_->Iterate(tX, tY, 1, true);
     }
+  }
 #endif
 
-  private:
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >    Hierarchy_;
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
 #ifdef HAVE_MUELU_AMGX
-    RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
+  RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
 #endif
-  };
+};
 
 #ifdef HAVE_XPETRA_EPETRA
 #ifndef EPETRA_NO_32BIT_GLOBAL_INDICES
-  /*! @class MueLuOp
-   *
-   * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
-   *        call which represents the effect of the multigrid preconditioner on a given vector.
-   *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
-   *
-   *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
-   *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
-   *
-   *        This is the specialization for <double,int,int,Xpetra::EpetraNode>
-   */
-  template <>
-  class MueLuOp<double, int, int, Xpetra::EpetraNode> :
-    public OperatorT<Xpetra::MultiVector<double, int, int, Xpetra::EpetraNode> >
+/*! @class MueLuOp
+ *
+ * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
+ *        call which represents the effect of the multigrid preconditioner on a given vector.
+ *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
+ *
+ *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
+ *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
+ *
+ *        This is the specialization for <double,int,int,Xpetra::EpetraNode>
+ */
+template <>
+class MueLuOp<double, int, int, Xpetra::EpetraNode> : public OperatorT<Xpetra::MultiVector<double, int, int, Xpetra::EpetraNode> >
 #ifdef HAVE_XPETRA_TPETRA
-    // check whether Tpetra is instantiated on double,int,int,EpetraNode
+// check whether Tpetra is instantiated on double,int,int,EpetraNode
 #if ((defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))))
-    , public OperatorT<Tpetra::MultiVector<double, int, int, Xpetra::EpetraNode> >
+     (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))))
+  ,
+                                                      public OperatorT<Tpetra::MultiVector<double, int, int, Xpetra::EpetraNode> >
 #endif
 #endif
 #ifdef HAVE_XPETRA_EPETRA
-    , public OperatorT<Epetra_MultiVector>
-    , public Belos::Operator<double>
+  ,
+                                                      public OperatorT<Epetra_MultiVector>,
+                                                      public Belos::Operator<double>
 #endif
-  {
-    typedef double             Scalar;
-    typedef int                LocalOrdinal;
-    typedef int                GlobalOrdinal;
-    typedef Xpetra::EpetraNode Node;
-
-  public:
-
-    MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H) : Hierarchy_(H) {}
+{
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
+
+ public:
+  MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H) {}
 #ifdef HAVE_MUELU_AMGX
-    MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) : AMGX_(A) {}
+  MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A)
+    : AMGX_(A) {}
 #endif
-    virtual ~MueLuOp() {}
-
-    void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS ) const {
+  virtual ~MueLuOp() {}
 
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+  void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
 
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
-      y.putScalar(0.0);
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
+    if (!AMGX_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
 
-        AMGX_->apply(tX, tY);
-      }
-#endif
-      if (!Hierarchy_.is_null())
-        Hierarchy_->Iterate(x, y, 1, true);
+      AMGX_->apply(tX, tY);
     }
+#endif
+    if (!Hierarchy_.is_null())
+      Hierarchy_->Iterate(x, y, 1, true);
+  }
 
 #ifdef HAVE_XPETRA_TPETRA
 #if ((defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))))
-    void Apply ( const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans=NOTRANS ) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+     (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))))
+  void Apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
 
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
-      y.putScalar(0.0);
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null())
-        AMGX_->apply(x, y);
+    if (!AMGX_.is_null())
+      AMGX_->apply(x, y);
 #endif
 
-      if (!Hierarchy_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &>(x);
+    if (!Hierarchy_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(x);
 
-        const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>  tX(rcpFromRef(temp_x));
-        Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>        tY(rcpFromRef(y));
+      const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
+      Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY(rcpFromRef(y));
 
-        tY.putScalar(0.0);
+      tY.putScalar(0.0);
 
-        Hierarchy_->Iterate(tX, tY, 1, true);
-      }
+      Hierarchy_->Iterate(tX, tY, 1, true);
     }
+  }
 #endif
 #endif
 
 #ifdef HAVE_XPETRA_EPETRA
-    // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
-    /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
-      to it resulting in the Tpetra::MultiVector \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Epetra_MultiVector& x, Epetra_MultiVector& y, ETrans trans =  NOTRANS) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
-
-      Epetra_MultiVector& temp_x = const_cast<Epetra_MultiVector&>(x);
-
-      const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> tX(rcpFromRef(temp_x));
-      Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node>       tY(rcpFromRef(y));
-
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate().
-      tY.putScalar(0.0);
-
-      Hierarchy_->Iterate(tX, tY, 1, true);
-    }
-
-    /*! \brief This routine takes the Belos::MultiVec \c x and applies the operator
-      to it resulting in the Belos::MultiVec \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Belos::MultiVec<double>& x, Belos::MultiVec<double>& y, ETrans trans = NOTRANS ) const {
-      const Epetra_MultiVector* vec_x = dynamic_cast<const Epetra_MultiVector*>(&x);
-      Epetra_MultiVector*       vec_y = dynamic_cast<Epetra_MultiVector*>(&y);
-
-      TEUCHOS_TEST_FOR_EXCEPTION(vec_x==NULL || vec_y==NULL, MueLuOpFailure,
-                                 "Belos::MueLuOp::Apply, x and/or y cannot be dynamic cast to an Epetra_MultiVector.");
-
-      Apply(*vec_x, *vec_y, trans);
-    }
+  // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
+  /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
+    to it resulting in the Tpetra::MultiVector \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Epetra_MultiVector& x, Epetra_MultiVector& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+
+    Epetra_MultiVector& temp_x = const_cast<Epetra_MultiVector&>(x);
+
+    const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
+    Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node> tY(rcpFromRef(y));
+
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate().
+    tY.putScalar(0.0);
+
+    Hierarchy_->Iterate(tX, tY, 1, true);
+  }
+
+  /*! \brief This routine takes the Belos::MultiVec \c x and applies the operator
+    to it resulting in the Belos::MultiVec \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Belos::MultiVec<double>& x, Belos::MultiVec<double>& y, ETrans trans = NOTRANS) const {
+    const Epetra_MultiVector* vec_x = dynamic_cast<const Epetra_MultiVector*>(&x);
+    Epetra_MultiVector* vec_y       = dynamic_cast<Epetra_MultiVector*>(&y);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(vec_x == NULL || vec_y == NULL, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, x and/or y cannot be dynamic cast to an Epetra_MultiVector.");
+
+    Apply(*vec_x, *vec_y, trans);
+  }
 #endif
 
-  private:
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >    Hierarchy_;
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
 #ifdef HAVE_MUELU_AMGX
-    RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
+  RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
 #endif
-  };
-#endif // !EPETRA_NO_32BIT_GLOBAL_INDICES
-#endif // HAVE_XPETRA_EPETRA
-
+};
+#endif  // !EPETRA_NO_32BIT_GLOBAL_INDICES
+#endif  // HAVE_XPETRA_EPETRA
 
 #ifdef HAVE_XPETRA_EPETRA
-#ifndef  EPETRA_NO_64BIT_GLOBAL_INDICES
-  /*! @class MueLuOp
-   *
-   * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
-   *        call which represents the effect of the multigrid preconditioner on a given vector.
-   *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
-   *
-   *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
-   *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
-   *
-   *        This is the specialization for <double,int,long long,Xpetra::EpetraNode>
-   */
-  template <>
-  class MueLuOp<double, int, long long, Xpetra::EpetraNode> :
-    public OperatorT<Xpetra::MultiVector<double, int, long long, Xpetra::EpetraNode> >
+#ifndef EPETRA_NO_64BIT_GLOBAL_INDICES
+/*! @class MueLuOp
+ *
+ * @brief MueLuOp derives from Belos::OperatorT and administrates a MueLu::Hierarchy. It implements the apply
+ *        call which represents the effect of the multigrid preconditioner on a given vector.
+ *        Note, in contrast to Belos::XpetraOp this operator has the multigrid hierarchy.
+ *
+ *        The Belos::OperatorT class is a generalization of the Belos::Operator<> class, which
+ *        deals with any kind of vector (not only Belos::MultiVec as the Belos::Operator<> interface does).
+ *
+ *        This is the specialization for <double,int,long long,Xpetra::EpetraNode>
+ */
+template <>
+class MueLuOp<double, int, long long, Xpetra::EpetraNode> : public OperatorT<Xpetra::MultiVector<double, int, long long, Xpetra::EpetraNode> >
 #ifdef HAVE_XPETRA_TPETRA
-    // check whether Tpetra is instantiated on double,int,int,EpetraNode
+// check whether Tpetra is instantiated on double,int,int,EpetraNode
 #if ((defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))) || \
-    (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))))
-    , public OperatorT<Tpetra::MultiVector<double, int, long long, Xpetra::EpetraNode> >
+     (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))))
+  ,
+                                                            public OperatorT<Tpetra::MultiVector<double, int, long long, Xpetra::EpetraNode> >
 #endif
 #endif
 #ifdef HAVE_XPETRA_EPETRA
-    , public OperatorT<Epetra_MultiVector>
-    , public Belos::Operator<double>
+  ,
+                                                            public OperatorT<Epetra_MultiVector>,
+                                                            public Belos::Operator<double>
 #endif
-  {
-    typedef double             Scalar;
-    typedef int                LocalOrdinal;
-    typedef long long          GlobalOrdinal;
-    typedef Xpetra::EpetraNode Node;
-
-  public:
-
-    MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H) : Hierarchy_(H) {}
+{
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef long long GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
+
+ public:
+  MueLuOp(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H) {}
 #ifdef HAVE_MUELU_AMGX
-    MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) : AMGX_(A) {}
+  MueLuOp(const RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A)
+    : AMGX_(A) {}
 #endif
-    virtual ~MueLuOp() {}
+  virtual ~MueLuOp() {}
 
-    void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS ) const {
+  void Apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
 
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
-
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
-      y.putScalar(0.0);
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
+    if (!AMGX_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX = Xpetra::toTpetra(x);
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY = Xpetra::toTpetra(y);
 
-        AMGX_->apply(tX, tY);
-      }
-#endif
-      if (!Hierarchy_.is_null())
-        Hierarchy_->Iterate(x, y, 1, true);
+      AMGX_->apply(tX, tY);
     }
+#endif
+    if (!Hierarchy_.is_null())
+      Hierarchy_->Iterate(x, y, 1, true);
+  }
 
 #ifdef HAVE_XPETRA_TPETRA
 #if ((defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))) || \
-    (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))))
-    void Apply ( const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans=NOTRANS ) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+     (!defined(EPETRA_HAVE_OMP) && (defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG))))
+  void Apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
 
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
-      y.putScalar(0.0);
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate(), but it matters for AMGX
+    y.putScalar(0.0);
 
 #ifdef HAVE_MUELU_AMGX
-      if (!AMGX_.is_null())
-        AMGX_->apply(x, y);
+    if (!AMGX_.is_null())
+      AMGX_->apply(x, y);
 #endif
 
-      if (!Hierarchy_.is_null()) {
-        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &>(x);
+    if (!Hierarchy_.is_null()) {
+      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& temp_x = const_cast<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(x);
 
-        const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>  tX(rcpFromRef(temp_x));
-        Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>        tY(rcpFromRef(y));
+      const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
+      Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tY(rcpFromRef(y));
 
-        tY.putScalar(0.0);
+      tY.putScalar(0.0);
 
-        Hierarchy_->Iterate(tX, tY, 1, true);
-      }
+      Hierarchy_->Iterate(tX, tY, 1, true);
     }
+  }
 #endif
 #endif
 
 #ifdef HAVE_XPETRA_EPETRA
-    // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
-    /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
-      to it resulting in the Tpetra::MultiVector \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Epetra_MultiVector& x, Epetra_MultiVector& y, ETrans trans =  NOTRANS) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
-                         "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
-
-      Epetra_MultiVector& temp_x = const_cast<Epetra_MultiVector&>(x);
-
-      const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> tX(rcpFromRef(temp_x));
-      Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node>       tY(rcpFromRef(y));
-
-      //FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate().
-      tY.putScalar(0.0);
-
-      Hierarchy_->Iterate(tX, tY, 1, true);
-    }
-
-    /*! \brief This routine takes the Belos::MultiVec \c x and applies the operator
-      to it resulting in the Belos::MultiVec \c y, which is returned.
-      \note It is expected that any problem with applying this operator to \c x will be
-      indicated by an std::exception being thrown.
-    */
-    void Apply(const Belos::MultiVec<double>& x, Belos::MultiVec<double>& y, ETrans trans = NOTRANS ) const {
-      const Epetra_MultiVector* vec_x = dynamic_cast<const Epetra_MultiVector*>(&x);
-      Epetra_MultiVector*       vec_y = dynamic_cast<Epetra_MultiVector*>(&y);
-
-      TEUCHOS_TEST_FOR_EXCEPTION(vec_x==NULL || vec_y==NULL, MueLuOpFailure,
-                                 "Belos::MueLuOp::Apply, x and/or y cannot be dynamic cast to an Epetra_MultiVector.");
-
-      Apply(*vec_x, *vec_y, trans);
-    }
+  // TO SKIP THE TRAIT IMPLEMENTATION OF XPETRA::MULTIVECTOR
+  /*! \brief This routine takes the Tpetra::MultiVector \c x and applies the operator
+    to it resulting in the Tpetra::MultiVector \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Epetra_MultiVector& x, Epetra_MultiVector& y, ETrans trans = NOTRANS) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(trans != NOTRANS, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, transpose mode != NOTRANS not supported by MueLu preconditionners.");
+
+    Epetra_MultiVector& temp_x = const_cast<Epetra_MultiVector&>(x);
+
+    const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node> tX(rcpFromRef(temp_x));
+    Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node> tY(rcpFromRef(y));
+
+    // FIXME InitialGuessIsZero currently does nothing in MueLu::Hierarchy.Iterate().
+    tY.putScalar(0.0);
+
+    Hierarchy_->Iterate(tX, tY, 1, true);
+  }
+
+  /*! \brief This routine takes the Belos::MultiVec \c x and applies the operator
+    to it resulting in the Belos::MultiVec \c y, which is returned.
+    \note It is expected that any problem with applying this operator to \c x will be
+    indicated by an std::exception being thrown.
+  */
+  void Apply(const Belos::MultiVec<double>& x, Belos::MultiVec<double>& y, ETrans trans = NOTRANS) const {
+    const Epetra_MultiVector* vec_x = dynamic_cast<const Epetra_MultiVector*>(&x);
+    Epetra_MultiVector* vec_y       = dynamic_cast<Epetra_MultiVector*>(&y);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(vec_x == NULL || vec_y == NULL, MueLuOpFailure,
+                               "Belos::MueLuOp::Apply, x and/or y cannot be dynamic cast to an Epetra_MultiVector.");
+
+    Apply(*vec_x, *vec_y, trans);
+  }
 #endif
 
-  private:
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >    Hierarchy_;
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
 #ifdef HAVE_MUELU_AMGX
-    RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
+  RCP<MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AMGX_;
 #endif
-  };
-#endif // !EPETRA_NO_64BIT_GLOBAL_INDICES
-#endif // HAVE_XPETRA_EPETRA
-} // namespace Belos
+};
+#endif  // !EPETRA_NO_64BIT_GLOBAL_INDICES
+#endif  // HAVE_XPETRA_EPETRA
+}  // namespace Belos
 
-#endif // BELOS_MUELU_ADAPTER_HPP
+#endif  // BELOS_MUELU_ADAPTER_HPP
diff --git a/packages/muelu/adapters/belos/BelosXpetraStatusTestGenResSubNorm.hpp b/packages/muelu/adapters/belos/BelosXpetraStatusTestGenResSubNorm.hpp
index 6f3f401041c2..355d2e346eea 100644
--- a/packages/muelu/adapters/belos/BelosXpetraStatusTestGenResSubNorm.hpp
+++ b/packages/muelu/adapters/belos/BelosXpetraStatusTestGenResSubNorm.hpp
@@ -59,27 +59,25 @@
 #include <BelosXpetraAdapterOperator.hpp>
 #include <BelosStatusTestGenResSubNorm.hpp>
 
-
 namespace Belos {
 
 /*! \brief Template specialization of Belos::StatusTestGenResSubNorm class using the
  * Xpetra::MultiVector and Belos::OperatorT MueLu adapter class.
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>,Belos::OperatorT<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >
-   : public StatusTestResNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>,Belos::OperatorT<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > > {
-
+class StatusTestGenResSubNorm<Scalar, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Belos::OperatorT<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >
+  : public StatusTestResNorm<Scalar, Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Belos::OperatorT<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > > {
  public:
   // Convenience typedefs
-  typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>      MV;
-  typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BCRS;
-  typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node>     ME;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BCRS;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> ME;
   typedef Belos::OperatorT<MV> OP;
 
   typedef Teuchos::ScalarTraits<Scalar> SCT;
   typedef typename SCT::magnitudeType MagnitudeType;
-  typedef MultiVecTraits<Scalar,MV>  MVT;
-  typedef OperatorTraits<Scalar,MV,OP>  OT;
+  typedef MultiVecTraits<Scalar, MV> MVT;
+  typedef OperatorTraits<Scalar, MV, OP> OT;
 
   //! @name Constructors/destructors.
   //@{
@@ -96,27 +94,27 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
     @param showMaxResNormOnly: for output only
 
   */
-  StatusTestGenResSubNorm( MagnitudeType Tolerance, size_t subIdx, int quorum = -1, bool showMaxResNormOnly = false )
-  : tolerance_(Tolerance),
-    subIdx_(subIdx),
-    quorum_(quorum),
-    showMaxResNormOnly_(showMaxResNormOnly),
-    resnormtype_(TwoNorm),
-    scaletype_(NormOfInitRes),
-    scalenormtype_(TwoNorm),
-    scalevalue_(Teuchos::ScalarTraits<MagnitudeType>::one ()),
-    status_(Undefined),
-    curBlksz_(0),
-    curNumRHS_(0),
-    curLSNum_(0),
-    numrhs_(0),
-    firstcallCheckStatus_(true),
-    firstcallDefineResForm_(true),
-    firstcallDefineScaleForm_(true),
-    mapExtractor_(Teuchos::null) { }
+  StatusTestGenResSubNorm(MagnitudeType Tolerance, size_t subIdx, int quorum = -1, bool showMaxResNormOnly = false)
+    : tolerance_(Tolerance)
+    , subIdx_(subIdx)
+    , quorum_(quorum)
+    , showMaxResNormOnly_(showMaxResNormOnly)
+    , resnormtype_(TwoNorm)
+    , scaletype_(NormOfInitRes)
+    , scalenormtype_(TwoNorm)
+    , scalevalue_(Teuchos::ScalarTraits<MagnitudeType>::one())
+    , status_(Undefined)
+    , curBlksz_(0)
+    , curNumRHS_(0)
+    , curLSNum_(0)
+    , numrhs_(0)
+    , firstcallCheckStatus_(true)
+    , firstcallDefineResForm_(true)
+    , firstcallDefineScaleForm_(true)
+    , mapExtractor_(Teuchos::null) {}
 
   //! Destructor
-  virtual ~StatusTestGenResSubNorm() { };
+  virtual ~StatusTestGenResSubNorm(){};
   //@}
 
   //! @name Form and parameter definition methods.
@@ -130,13 +128,13 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
     </ul>
   */
   int defineResForm(NormType TypeOfNorm) {
-    TEUCHOS_TEST_FOR_EXCEPTION(firstcallDefineResForm_==false,StatusTestError,
-          "StatusTestGenResSubNorm::defineResForm(): The residual form has already been defined.");
+    TEUCHOS_TEST_FOR_EXCEPTION(firstcallDefineResForm_ == false, StatusTestError,
+                               "StatusTestGenResSubNorm::defineResForm(): The residual form has already been defined.");
     firstcallDefineResForm_ = false;
 
     resnormtype_ = TypeOfNorm;
 
-    return(0);
+    return (0);
   }
 
   //! Define form of the scaling, its norm, its optional weighting std::vector, or, alternatively, define an explicit value.
@@ -160,35 +158,47 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
     </ul>
     </ol>
   */
-  int defineScaleForm( ScaleType TypeOfScaling, NormType TypeOfNorm, MagnitudeType ScaleValue = Teuchos::ScalarTraits<MagnitudeType>::one()) {
-    TEUCHOS_TEST_FOR_EXCEPTION(firstcallDefineScaleForm_==false,StatusTestError,
-          "StatusTestGenResSubNorm::defineScaleForm(): The scaling type has already been defined.");
+  int defineScaleForm(ScaleType TypeOfScaling, NormType TypeOfNorm, MagnitudeType ScaleValue = Teuchos::ScalarTraits<MagnitudeType>::one()) {
+    TEUCHOS_TEST_FOR_EXCEPTION(firstcallDefineScaleForm_ == false, StatusTestError,
+                               "StatusTestGenResSubNorm::defineScaleForm(): The scaling type has already been defined.");
     firstcallDefineScaleForm_ = false;
 
-    scaletype_ = TypeOfScaling;
+    scaletype_     = TypeOfScaling;
     scalenormtype_ = TypeOfNorm;
-    scalevalue_ = ScaleValue;
+    scalevalue_    = ScaleValue;
 
-    return(0);
+    return (0);
   }
 
   //! Set the value of the tolerance
   /*! We allow the tolerance to be reset for cases where, in the process of testing the residual,
     we find that the initial tolerance was too tight or too lax.
   */
-  int setTolerance(MagnitudeType tolerance) {tolerance_ = tolerance; return(0);}
+  int setTolerance(MagnitudeType tolerance) {
+    tolerance_ = tolerance;
+    return (0);
+  }
 
   //! Set the block index of which we want to check the norm of the sub-residuals
   /*! It does not really make sense to change/reset the index during the solution process
    */
-  int setSubIdx ( size_t subIdx ) { subIdx_ = subIdx; return(0);}
+  int setSubIdx(size_t subIdx) {
+    subIdx_ = subIdx;
+    return (0);
+  }
 
   //! Sets the number of residuals that must pass the convergence test before Passed is returned.
   //! \note If \c quorum=-1 then all residuals must pass the convergence test before Passed is returned.
-  int setQuorum(int quorum) {quorum_ = quorum; return(0);}
+  int setQuorum(int quorum) {
+    quorum_ = quorum;
+    return (0);
+  }
 
   //! Set whether the only maximum residual norm is displayed when the print() method is called
-  int setShowMaxResNormOnly(bool showMaxResNormOnly) {showMaxResNormOnly_ = showMaxResNormOnly; return(0);}
+  int setShowMaxResNormOnly(bool showMaxResNormOnly) {
+    showMaxResNormOnly_ = showMaxResNormOnly;
+    return (0);
+  }
 
   //@}
 
@@ -201,55 +211,57 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
 
     \return StatusType: Passed, Failed, or Undefined.
   */
-  StatusType checkStatus(Iteration<Scalar,MV,OP>* iSolver) {
-    MagnitudeType zero = Teuchos::ScalarTraits<MagnitudeType>::zero();
-    const LinearProblem<Scalar,MV,OP>& lp = iSolver->getProblem();
+  StatusType checkStatus(Iteration<Scalar, MV, OP>* iSolver) {
+    MagnitudeType zero                      = Teuchos::ScalarTraits<MagnitudeType>::zero();
+    const LinearProblem<Scalar, MV, OP>& lp = iSolver->getProblem();
     // Compute scaling term (done once for each block that's being solved)
     if (firstcallCheckStatus_) {
       StatusType status = firstCallCheckStatusSetup(iSolver);
-      if(status==Failed) {
+      if (status == Failed) {
         status_ = Failed;
-        return(status_);
+        return (status_);
       }
     }
 
     //
     // This section computes the norm of the residual std::vector
     //
-    if ( curLSNum_ != lp.getLSNumber() ) {
+    if (curLSNum_ != lp.getLSNumber()) {
       //
       // We have moved on to the next rhs block
       //
-      curLSNum_ = lp.getLSNumber();
-      curLSIdx_ = lp.getLSIndex();
-      curBlksz_ = (int)curLSIdx_.size();
+      curLSNum_   = lp.getLSNumber();
+      curLSIdx_   = lp.getLSIndex();
+      curBlksz_   = (int)curLSIdx_.size();
       int validLS = 0;
-      for (int i=0; i<curBlksz_; ++i) {
+      for (int i = 0; i < curBlksz_; ++i) {
         if (curLSIdx_[i] > -1 && curLSIdx_[i] < numrhs_)
           validLS++;
       }
       curNumRHS_ = validLS;
-      curSoln_ = Teuchos::null;
+      curSoln_   = Teuchos::null;
       //
     } else {
       //
       // We are in the same rhs block, return if we are converged
       //
-      if (status_==Passed) { return status_; }
+      if (status_ == Passed) {
+        return status_;
+      }
     }
 
     //
     // Request the true residual for this block of right-hand sides.
     //
     Teuchos::RCP<MV> cur_update = iSolver->getCurrentUpdate();
-    curSoln_ = lp.updateSolution( cur_update );
-    Teuchos::RCP<MV> cur_res = MVT::Clone( *curSoln_, MVT::GetNumberVecs( *curSoln_ ) );
-    lp.computeCurrResVec( &*cur_res, &*curSoln_ );
-    std::vector<MagnitudeType> tmp_resvector( MVT::GetNumberVecs( *cur_res ) );
-    MvSubNorm( *cur_res, subIdx_, tmp_resvector, resnormtype_ );
+    curSoln_                    = lp.updateSolution(cur_update);
+    Teuchos::RCP<MV> cur_res    = MVT::Clone(*curSoln_, MVT::GetNumberVecs(*curSoln_));
+    lp.computeCurrResVec(&*cur_res, &*curSoln_);
+    std::vector<MagnitudeType> tmp_resvector(MVT::GetNumberVecs(*cur_res));
+    MvSubNorm(*cur_res, subIdx_, tmp_resvector, resnormtype_);
 
     typename std::vector<int>::iterator p = curLSIdx_.begin();
-    for (int i=0; p<curLSIdx_.end(); ++p, ++i) {
+    for (int i = 0; p < curLSIdx_.end(); ++p, ++i) {
       // Check if this index is valid
       if (*p != -1)
         resvector_[*p] = tmp_resvector[i];
@@ -259,60 +271,59 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
     // Compute the new linear system residuals for testing.
     // (if any of them don't meet the tolerance or are NaN, then we exit with that status)
     //
-    if ( scalevector_.size() > 0 ) {
+    if (scalevector_.size() > 0) {
       typename std::vector<int>::iterator pp = curLSIdx_.begin();
-      for (; pp<curLSIdx_.end(); ++pp) {
+      for (; pp < curLSIdx_.end(); ++pp) {
         // Check if this index is valid
         if (*pp != -1) {
           // Scale the std::vector accordingly
-          if ( scalevector_[ *pp ] != zero ) {
+          if (scalevector_[*pp] != zero) {
             // Don't intentionally divide by zero.
-            testvector_[ *pp ] = resvector_[ *pp ] / scalevector_[ *pp ] / scalevalue_;
+            testvector_[*pp] = resvector_[*pp] / scalevector_[*pp] / scalevalue_;
           } else {
-            testvector_[ *pp ] = resvector_[ *pp ] / scalevalue_;
+            testvector_[*pp] = resvector_[*pp] / scalevalue_;
           }
         }
       }
-    }
-    else {
+    } else {
       typename std::vector<int>::iterator pp = curLSIdx_.begin();
-      for (; pp<curLSIdx_.end(); ++pp) {
+      for (; pp < curLSIdx_.end(); ++pp) {
         // Check if this index is valid
         if (*pp != -1)
-          testvector_[ *pp ] = resvector_[ *pp ] / scalevalue_;
+          testvector_[*pp] = resvector_[*pp] / scalevalue_;
       }
     }
     // Check status of new linear system residuals and see if we have the quorum.
     int have = 0;
-    ind_.resize( curLSIdx_.size() );
+    ind_.resize(curLSIdx_.size());
     typename std::vector<int>::iterator p2 = curLSIdx_.begin();
-    for (; p2<curLSIdx_.end(); ++p2) {
+    for (; p2 < curLSIdx_.end(); ++p2) {
       // Check if this index is valid
       if (*p2 != -1) {
         // Check if any of the residuals are larger than the tolerance.
-        if (testvector_[ *p2 ] > tolerance_) {
+        if (testvector_[*p2] > tolerance_) {
           // do nothing.
-        } else if (testvector_[ *p2 ] == Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero())) {
+        } else if (testvector_[*p2] == Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero())) {
           reset();
-        } else if (testvector_[ *p2 ] <= tolerance_) {
+        } else if (testvector_[*p2] <= tolerance_) {
           ind_[have] = *p2;
           have++;
         } else {
           // Throw an std::exception if a NaN is found.
           status_ = Failed;
-          TEUCHOS_TEST_FOR_EXCEPTION(true,StatusTestError,"StatusTestGenResSubNorm::checkStatus(): NaN has been detected.");
+          TEUCHOS_TEST_FOR_EXCEPTION(true, StatusTestError, "StatusTestGenResSubNorm::checkStatus(): NaN has been detected.");
         }
       }
     }
     ind_.resize(have);
-    int need = (quorum_ == -1) ? curNumRHS_: quorum_;
-    status_ = (have >= need) ? Passed : Failed;
+    int need = (quorum_ == -1) ? curNumRHS_ : quorum_;
+    status_  = (have >= need) ? Passed : Failed;
     // Return the current status
     return status_;
   }
 
   //! Return the result of the most recent CheckStatus call.
-  StatusType getStatus() const {return(status_);};
+  StatusType getStatus() const { return (status_); };
   //@}
 
   //! @name Reset methods
@@ -320,14 +331,14 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
 
   //! Resets the internal configuration to the initial state.
   void reset() {
-    status_ = Undefined;
+    status_   = Undefined;
     curBlksz_ = 0;
     curLSNum_ = 0;
     curLSIdx_.resize(0);
     numrhs_ = 0;
     ind_.resize(0);
     firstcallCheckStatus_ = true;
-    curSoln_ = Teuchos::null;
+    curSoln_              = Teuchos::null;
   }
 
   //@}
@@ -338,29 +349,30 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   //! Output formatted description of stopping test to output stream.
   void print(std::ostream& os, int indent = 0) const {
     os.setf(std::ios_base::scientific);
-    for (int j = 0; j < indent; j ++)
+    for (int j = 0; j < indent; j++)
       os << ' ';
     printStatus(os, status_);
     os << resFormStr();
-    if (status_==Undefined)
+    if (status_ == Undefined)
       os << ", tol = " << tolerance_ << std::endl;
     else {
       os << std::endl;
-      if(showMaxResNormOnly_ && curBlksz_ > 1) {
+      if (showMaxResNormOnly_ && curBlksz_ > 1) {
         const MagnitudeType maxRelRes = *std::max_element(
-          testvector_.begin()+curLSIdx_[0],testvector_.begin()+curLSIdx_[curBlksz_-1]
-          );
-        for (int j = 0; j < indent + 13; j ++)
+            testvector_.begin() + curLSIdx_[0], testvector_.begin() + curLSIdx_[curBlksz_ - 1]);
+        for (int j = 0; j < indent + 13; j++)
           os << ' ';
-        os << "max{residual["<<curLSIdx_[0]<<"..."<<curLSIdx_[curBlksz_-1]<<"]} = " << maxRelRes
-           << ( maxRelRes <= tolerance_ ? " <= " : " > " ) << tolerance_ << std::endl;
-      }
-      else {
-        for ( int i=0; i<numrhs_; i++ ) {
-          for (int j = 0; j < indent + 13; j ++)
+        os << "max{residual[" << curLSIdx_[0] << "..." << curLSIdx_[curBlksz_ - 1] << "]} = " << maxRelRes
+           << (maxRelRes <= tolerance_ ? " <= " : " > ") << tolerance_ << std::endl;
+      } else {
+        for (int i = 0; i < numrhs_; i++) {
+          for (int j = 0; j < indent + 13; j++)
             os << ' ';
-          os << "residual [ " << i << " ] = " << testvector_[ i ];
-          os << ((testvector_[i]<tolerance_) ? " < " : (testvector_[i]==tolerance_) ? " == " : (testvector_[i]>tolerance_) ? " > " : " "  ) << tolerance_ << std::endl;
+          os << "residual [ " << i << " ] = " << testvector_[i];
+          os << ((testvector_[i] < tolerance_) ? " < " : (testvector_[i] == tolerance_) ? " == "
+                                                     : (testvector_[i] > tolerance_)    ? " > "
+                                                                                        : " ")
+             << tolerance_ << std::endl;
         }
       }
     }
@@ -371,19 +383,19 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   void printStatus(std::ostream& os, StatusType type) const {
     os << std::left << std::setw(13) << std::setfill('.');
     switch (type) {
-    case  Passed:
-      os << "Converged";
-      break;
-    case  Failed:
-      os << "Unconverged";
-      break;
-    case  Undefined:
-    default:
-      os << "**";
-      break;
+      case Passed:
+        os << "Converged";
+        break;
+      case Failed:
+        os << "Unconverged";
+        break;
+      case Undefined:
+      default:
+        os << "**";
+        break;
     }
     os << std::left << std::setfill(' ');
-      return;
+    return;
   }
   //@}
 
@@ -407,16 +419,16 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   std::vector<int> convIndices() { return ind_; }
 
   //! Returns the value of the tolerance, \f$ \tau \f$, set in the constructor.
-  MagnitudeType getTolerance() const {return(tolerance_);};
+  MagnitudeType getTolerance() const { return (tolerance_); };
 
   //! Returns the test value, \f$ \frac{\|r\|}{\sigma} \f$, computed in most recent call to CheckStatus.
-  const std::vector<MagnitudeType>* getTestValue() const {return(&testvector_);};
+  const std::vector<MagnitudeType>* getTestValue() const { return (&testvector_); };
 
   //! Returns the residual norm value, \f$ \|r\| \f$, computed in most recent call to CheckStatus.
-  const std::vector<MagnitudeType>* getResNormValue() const {return(&resvector_);};
+  const std::vector<MagnitudeType>* getResNormValue() const { return (&resvector_); };
 
   //! Returns the scaled norm value, \f$ \sigma \f$.
-  const std::vector<MagnitudeType>* getScaledNormValue() const {return(&scalevector_);};
+  const std::vector<MagnitudeType>* getScaledNormValue() const { return (&scalevector_); };
 
   //! Returns a boolean indicating a loss of accuracy has been detected in computing the residual.
   //! \note This status test does not check for loss of accuracy, so this method will always return false.
@@ -424,7 +436,6 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
 
   //@}
 
-
   /** @name Misc. */
   //@{
 
@@ -433,11 +444,11 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
    * After this function is called <tt>getScaledNormValue()</tt> can be called
    * to get the scaling std::vector.
    */
-  StatusType firstCallCheckStatusSetup(Iteration<Scalar,MV,OP>* iSolver) {
+  StatusType firstCallCheckStatusSetup(Iteration<Scalar, MV, OP>* iSolver) {
     int i;
-    MagnitudeType zero = Teuchos::ScalarTraits<MagnitudeType>::zero();
-    MagnitudeType one = Teuchos::ScalarTraits<MagnitudeType>::one();
-    const LinearProblem<Scalar,MV,OP>& lp = iSolver->getProblem();
+    MagnitudeType zero                      = Teuchos::ScalarTraits<MagnitudeType>::zero();
+    MagnitudeType one                       = Teuchos::ScalarTraits<MagnitudeType>::one();
+    const LinearProblem<Scalar, MV, OP>& lp = iSolver->getProblem();
     // Compute scaling term (done once for each block that's being solved)
     if (firstcallCheckStatus_) {
       //
@@ -447,9 +458,9 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
 
       // try to access the underlying blocked operator
       Teuchos::RCP<const OP> Op = lp.getOperator();
-      Teuchos::RCP<const Belos::XpetraOp<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xOp =
-          Teuchos::rcp_dynamic_cast<const Belos::XpetraOp<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-      TEUCHOS_TEST_FOR_EXCEPTION(xOp.is_null(), MueLu::Exceptions::BadCast, "Bad cast from \'const Belos::OperatorT\' to \'const Belos::XpetraOp\'. The origin type is "  << typeid(const OP).name() << ".");
+      Teuchos::RCP<const Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xOp =
+          Teuchos::rcp_dynamic_cast<const Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(Op);
+      TEUCHOS_TEST_FOR_EXCEPTION(xOp.is_null(), MueLu::Exceptions::BadCast, "Bad cast from \'const Belos::OperatorT\' to \'const Belos::XpetraOp\'. The origin type is " << typeid(const OP).name() << ".");
       Teuchos::RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xIntOp =
           xOp->getOperator();
       TEUCHOS_TEST_FOR_EXCEPTION(xIntOp.is_null(), MueLu::Exceptions::BadCast, "Cannot access Xpetra::Operator stored in Belos::XpetraOperator.");
@@ -457,77 +468,72 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
           Teuchos::rcp_dynamic_cast<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(xIntOp);
       TEUCHOS_TEST_FOR_EXCEPTION(xMat.is_null(), MueLu::Exceptions::RuntimeError, "Cannot access Xpetra::Matrix stored in Belos::XpetraOp. Error.");
       Teuchos::RCP<const Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bMat = Teuchos::rcp_dynamic_cast<const Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(xMat);
-      TEUCHOS_TEST_FOR_EXCEPTION(bMat.is_null(), MueLu::Exceptions::BadCast, "Bad cast from \'const Xpetra::Matrix\' to \'const Xpetra::BlockedCrsMatrix\'. The origin type is "  << typeid(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>).name() << ". Note: you need a BlockedCrsMatrix object for the StatusTestGenResSubNorm to work!");
+      TEUCHOS_TEST_FOR_EXCEPTION(bMat.is_null(), MueLu::Exceptions::BadCast, "Bad cast from \'const Xpetra::Matrix\' to \'const Xpetra::BlockedCrsMatrix\'. The origin type is " << typeid(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>).name() << ". Note: you need a BlockedCrsMatrix object for the StatusTestGenResSubNorm to work!");
       mapExtractor_ = bMat->getRangeMapExtractor();
       TEUCHOS_TEST_FOR_EXCEPTION(mapExtractor_.is_null(), MueLu::Exceptions::RuntimeError, "Could not extract map extractor from BlockedCrsMatrix. Error.");
-      TEUCHOS_TEST_FOR_EXCEPTION(mapExtractor_->NumMaps()<=subIdx_, MueLu::Exceptions::RuntimeError, "The multivector is only split into " << mapExtractor_->NumMaps() << " sub parts. Cannot access sub-block " << subIdx_ << ".");
+      TEUCHOS_TEST_FOR_EXCEPTION(mapExtractor_->NumMaps() <= subIdx_, MueLu::Exceptions::RuntimeError, "The multivector is only split into " << mapExtractor_->NumMaps() << " sub parts. Cannot access sub-block " << subIdx_ << ".");
 
       // calculate initial norms
-      if (scaletype_== NormOfRHS) {
+      if (scaletype_ == NormOfRHS) {
         Teuchos::RCP<const MV> rhs = lp.getRHS();
-        numrhs_ = MVT::GetNumberVecs( *rhs );
-        scalevector_.resize( numrhs_ );
-        MvSubNorm( *rhs, subIdx_, scalevector_, scalenormtype_ );
-      }
-      else if (scaletype_==NormOfInitRes) {
+        numrhs_                    = MVT::GetNumberVecs(*rhs);
+        scalevector_.resize(numrhs_);
+        MvSubNorm(*rhs, subIdx_, scalevector_, scalenormtype_);
+      } else if (scaletype_ == NormOfInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MvSubNorm( *init_res, subIdx_, scalevector_, scalenormtype_ );
-      }
-      else if (scaletype_==NormOfPrecInitRes) {
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MvSubNorm(*init_res, subIdx_, scalevector_, scalenormtype_);
+      } else if (scaletype_ == NormOfPrecInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitPrecResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MvSubNorm( *init_res, subIdx_, scalevector_, scalenormtype_ );
-      }
-      else if (scaletype_==NormOfFullInitRes) {
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MvSubNorm(*init_res, subIdx_, scalevector_, scalenormtype_);
+      } else if (scaletype_ == NormOfFullInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MVT::MvNorm( *init_res, scalevector_, scalenormtype_ );
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MVT::MvNorm(*init_res, scalevector_, scalenormtype_);
         scalevalue_ = one;
-      }
-      else if (scaletype_==NormOfFullPrecInitRes) {
+      } else if (scaletype_ == NormOfFullPrecInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitPrecResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MVT::MvNorm( *init_res, scalevector_, scalenormtype_ );
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MVT::MvNorm(*init_res, scalevector_, scalenormtype_);
         scalevalue_ = one;
-      }
-      else if (scaletype_==NormOfFullScaledInitRes) {
+      } else if (scaletype_ == NormOfFullScaledInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MVT::MvNorm( *init_res, scalevector_, scalenormtype_ );
-        MvScalingRatio( *init_res, subIdx_, scalevalue_ );
-      }
-      else if (scaletype_==NormOfFullScaledPrecInitRes) {
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MVT::MvNorm(*init_res, scalevector_, scalenormtype_);
+        MvScalingRatio(*init_res, subIdx_, scalevalue_);
+      } else if (scaletype_ == NormOfFullScaledPrecInitRes) {
         Teuchos::RCP<const MV> init_res = lp.getInitPrecResVec();
-        numrhs_ = MVT::GetNumberVecs( *init_res );
-        scalevector_.resize( numrhs_ );
-        MVT::MvNorm( *init_res, scalevector_, scalenormtype_ );
-        MvScalingRatio( *init_res, subIdx_, scalevalue_ );
-      }
-      else {
-        numrhs_ = MVT::GetNumberVecs( *(lp.getRHS()) );
+        numrhs_                         = MVT::GetNumberVecs(*init_res);
+        scalevector_.resize(numrhs_);
+        MVT::MvNorm(*init_res, scalevector_, scalenormtype_);
+        MvScalingRatio(*init_res, subIdx_, scalevalue_);
+      } else {
+        numrhs_ = MVT::GetNumberVecs(*(lp.getRHS()));
       }
 
-      resvector_.resize( numrhs_ );
-      testvector_.resize( numrhs_ );
+      resvector_.resize(numrhs_);
+      testvector_.resize(numrhs_);
 
-      curLSNum_ = lp.getLSNumber();
-      curLSIdx_ = lp.getLSIndex();
-      curBlksz_ = (int)curLSIdx_.size();
+      curLSNum_   = lp.getLSNumber();
+      curLSIdx_   = lp.getLSIndex();
+      curBlksz_   = (int)curLSIdx_.size();
       int validLS = 0;
-      for (i=0; i<curBlksz_; ++i) {
+      for (i = 0; i < curBlksz_; ++i) {
         if (curLSIdx_[i] > -1 && curLSIdx_[i] < numrhs_)
           validLS++;
       }
       curNumRHS_ = validLS;
       //
       // Initialize the testvector.
-      for (i=0; i<numrhs_; i++) { testvector_[i] = one; }
+      for (i = 0; i < numrhs_; i++) {
+        testvector_[i] = one;
+      }
 
       // Return an error if the scaling is zero.
       if (scalevalue_ == zero) {
@@ -542,8 +548,7 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   //@{
 
   /** \brief Method to return description of the maximum iteration status test  */
-  std::string description() const
-  {
+  std::string description() const {
     std::ostringstream oss;
     oss << "Belos::StatusTestGenResSubNorm<>: " << resFormStr();
     oss << ", tol = " << tolerance_;
@@ -552,43 +557,41 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   //@}
 
  protected:
-
  private:
-
   //! @name Private methods.
   //@{
   /** \brief Description of current residual form */
-  std::string resFormStr() const
-  {
+  std::string resFormStr() const {
     std::ostringstream oss;
     oss << "(";
-    oss << ((resnormtype_==OneNorm) ? "1-Norm" : (resnormtype_==TwoNorm) ? "2-Norm" : "Inf-Norm");
+    oss << ((resnormtype_ == OneNorm) ? "1-Norm" : (resnormtype_ == TwoNorm) ? "2-Norm"
+                                                                             : "Inf-Norm");
     oss << " Exp";
     oss << " Res Vec [" << subIdx_ << "]) ";
 
     // If there is no residual scaling, return current string.
-    if (scaletype_!=None)
-    {
+    if (scaletype_ != None) {
       // Insert division sign.
       oss << "/ ";
 
       // Determine output string for scaling, if there is any.
-      if (scaletype_==UserProvided)
+      if (scaletype_ == UserProvided)
         oss << " (User Scale)";
       else {
         oss << "(";
-        oss << ((scalenormtype_==OneNorm) ? "1-Norm" : (resnormtype_==TwoNorm) ? "2-Norm" : "Inf-Norm");
-        if (scaletype_==NormOfInitRes)
+        oss << ((scalenormtype_ == OneNorm) ? "1-Norm" : (resnormtype_ == TwoNorm) ? "2-Norm"
+                                                                                   : "Inf-Norm");
+        if (scaletype_ == NormOfInitRes)
           oss << " Res0 [" << subIdx_ << "]";
-        else if (scaletype_==NormOfPrecInitRes)
+        else if (scaletype_ == NormOfPrecInitRes)
           oss << " Prec Res0 [" << subIdx_ << "]";
-        else if (scaletype_==NormOfFullInitRes)
+        else if (scaletype_ == NormOfFullInitRes)
           oss << " Full Res0 [" << subIdx_ << "]";
-        else if (scaletype_==NormOfFullPrecInitRes)
+        else if (scaletype_ == NormOfFullPrecInitRes)
           oss << " Full Prec Res0 [" << subIdx_ << "]";
-        else if (scaletype_==NormOfFullScaledInitRes)
+        else if (scaletype_ == NormOfFullScaledInitRes)
           oss << " scaled Full Res0 [" << subIdx_ << "]";
-        else if (scaletype_==NormOfFullScaledPrecInitRes)
+        else if (scaletype_ == NormOfFullScaledPrecInitRes)
           oss << " scaled Full Prec Res0 [" << subIdx_ << "]";
         else
           oss << " RHS [" << subIdx_ << "]";
@@ -607,16 +610,15 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   //@{
 
   // calculate norm of partial multivector
-  void MvSubNorm( const MV& mv, size_t block, std::vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& normVec, NormType type = TwoNorm) {
-
+  void MvSubNorm(const MV& mv, size_t block, std::vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& normVec, NormType type = TwoNorm) {
     Teuchos::RCP<const MV> input = Teuchos::rcpFromRef(mv);
 
     Teuchos::RCP<const MV> SubVec = mapExtractor_->ExtractVector(input, block);
-    MVT::MvNorm(*SubVec,normVec,type);
+    MVT::MvNorm(*SubVec, normVec, type);
   }
 
   // calculate ration of sub-vector length to full vector length (for scalevalue_)
-  void MvScalingRatio( const MV& mv, size_t block, MagnitudeType& lengthRatio) {
+  void MvScalingRatio(const MV& mv, size_t block, MagnitudeType& lengthRatio) {
     Teuchos::RCP<const MV> input = Teuchos::rcpFromRef(mv);
 
     Teuchos::RCP<const MV> SubVec = mapExtractor_->ExtractVector(input, block);
@@ -697,9 +699,8 @@ class StatusTestGenResSubNorm<Scalar,Xpetra::MultiVector<Scalar,LocalOrdinal,Glo
   //! MapExtractor for range space
   Teuchos::RCP<const ME> mapExtractor_;
   //@}
-
 };
 
-} // namespace Belos
+}  // namespace Belos
 
 #endif /* BELOS_XPETRA_STATUS_TEST_GEN_RES_SUB_NORM_HPP */
diff --git a/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.cpp b/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.cpp
index 8d5ed9dbdb8b..6536d40be90b 100644
--- a/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.cpp
+++ b/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.cpp
@@ -22,108 +22,104 @@
 #if defined(HAVE_MUELU_EPETRA)
 namespace MueLu {
 
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a EpetraCrs_Matrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] paramListIn Parameter list
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>&   inA,
-                             // FIXME: why is it non-const
-                             Teuchos::ParameterList& paramListIn)
-  {
-    using SC = double;
-    using LO = int;
-    using GO = int;
-    using NO = Xpetra::EpetraNode;
-
-    using Teuchos::ParameterList;
-
-    using MultiVector      = Xpetra::MultiVector<SC, LO, GO, NO>;
-    using Matrix           = Xpetra::Matrix<SC, LO, GO, NO>;
-    using Hierarchy        = Hierarchy<SC,LO,GO,NO>;
-    using HierarchyManager = HierarchyManager<SC,LO,GO,NO>;
-
-    Teuchos::ParameterList& userList = paramListIn.sublist("user data");
-    if (userList.isParameter("Coordinates")) {
-      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType,LO,GO,NO> > coordinates = Teuchos::null;
-      try {
-        coordinates = EpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType,LO,GO,NO>(userList.get<RCP<Epetra_MultiVector> >("Coordinates"));
-      } catch(Teuchos::Exceptions::InvalidParameterType&) {
-        coordinates = userList.get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > >("Coordinates");
-      }
-      if(Teuchos::nonnull(coordinates)){
-        userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType,LO,GO,NO> > >("Coordinates", coordinates);
-      }
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a EpetraCrs_Matrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] paramListIn Parameter list
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA,
+                           // FIXME: why is it non-const
+                           Teuchos::ParameterList& paramListIn) {
+  using SC = double;
+  using LO = int;
+  using GO = int;
+  using NO = Xpetra::EpetraNode;
+
+  using Teuchos::ParameterList;
+
+  using MultiVector      = Xpetra::MultiVector<SC, LO, GO, NO>;
+  using Matrix           = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Hierarchy        = Hierarchy<SC, LO, GO, NO>;
+  using HierarchyManager = HierarchyManager<SC, LO, GO, NO>;
+
+  Teuchos::ParameterList& userList = paramListIn.sublist("user data");
+  if (userList.isParameter("Coordinates")) {
+    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > coordinates = Teuchos::null;
+    try {
+      coordinates = EpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO>(userList.get<RCP<Epetra_MultiVector> >("Coordinates"));
+    } catch (Teuchos::Exceptions::InvalidParameterType&) {
+      coordinates = userList.get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > >("Coordinates");
     }
-    if (userList.isParameter("Nullspace")) {
-      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType,LO,GO,NO> > nullspace = Teuchos::null;
-      try {
-        nullspace = EpetraMultiVector_To_XpetraMultiVector<SC,LO,GO,NO>(userList.get<RCP<Epetra_MultiVector> >("Nullspace"));
-      } catch(Teuchos::Exceptions::InvalidParameterType&) {
-        nullspace = userList.get<RCP<Xpetra::MultiVector<SC, LO, GO, NO> > >("Nullspace");
-      }
-      if(Teuchos::nonnull(nullspace)){
-        userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType,LO,GO,NO> > >("Nullspace", nullspace);
-      }
+    if (Teuchos::nonnull(coordinates)) {
+      userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > >("Coordinates", coordinates);
     }
-
-    RCP<Matrix> A = EpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(inA);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A, paramListIn);
-    return rcp(new EpetraOperator(H));
-  }
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] xmlFileName XML file containing MueLu options.
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>  & A,
-                             const std::string& xmlFileName)
-  {
-    Teuchos::ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *Xpetra::toXpetra(A->Comm()));
-
-    return CreateEpetraPreconditioner(A, paramList);
   }
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix.
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>  & A)
-  {
-    Teuchos::ParameterList paramList;
-    return CreateEpetraPreconditioner(A, paramList);
-  }
-
-  void ReuseEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA, MueLu::EpetraOperator& Op) {
-    using SC = double;
-    using LO = int;
-    using GO = int;
-    using NO = Xpetra::EpetraNode;
-
-    using Teuchos::ParameterList;
-
-    using Matrix           = Xpetra::Matrix<SC, LO, GO, NO>;
-    using Hierarchy        = Hierarchy<SC,LO,GO,NO>;
-
-    RCP<Hierarchy> H = Op.GetHierarchy();
-    RCP<Matrix>    A = EpetraCrs_To_XpetraMatrix<SC,LO,GO,NO>(inA);
-
-    MueLu::ReuseXpetraPreconditioner<SC,LO,GO,NO>(A, H);
+  if (userList.isParameter("Nullspace")) {
+    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > nullspace = Teuchos::null;
+    try {
+      nullspace = EpetraMultiVector_To_XpetraMultiVector<SC, LO, GO, NO>(userList.get<RCP<Epetra_MultiVector> >("Nullspace"));
+    } catch (Teuchos::Exceptions::InvalidParameterType&) {
+      nullspace = userList.get<RCP<Xpetra::MultiVector<SC, LO, GO, NO> > >("Nullspace");
+    }
+    if (Teuchos::nonnull(nullspace)) {
+      userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<SC>::coordinateType, LO, GO, NO> > >("Nullspace", nullspace);
+    }
   }
 
-
-} //namespace
-#endif // HAVE_MUELU_SERIAL and HAVE_MUELU_EPETRA
-
-#endif //ifndef MUELU_CREATE_EPETRA_PRECONDITIONER_CPP
+  RCP<Matrix> A    = EpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(inA);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, paramListIn);
+  return rcp(new EpetraOperator(H));
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] xmlFileName XML file containing MueLu options.
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& A,
+                           const std::string& xmlFileName) {
+  Teuchos::ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *Xpetra::toXpetra(A->Comm()));
+
+  return CreateEpetraPreconditioner(A, paramList);
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix.
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& A) {
+  Teuchos::ParameterList paramList;
+  return CreateEpetraPreconditioner(A, paramList);
+}
+
+void ReuseEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA, MueLu::EpetraOperator& Op) {
+  using SC = double;
+  using LO = int;
+  using GO = int;
+  using NO = Xpetra::EpetraNode;
+
+  using Teuchos::ParameterList;
+
+  using Matrix    = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Hierarchy = Hierarchy<SC, LO, GO, NO>;
+
+  RCP<Hierarchy> H = Op.GetHierarchy();
+  RCP<Matrix> A    = EpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(inA);
+
+  MueLu::ReuseXpetraPreconditioner<SC, LO, GO, NO>(A, H);
+}
+
+}  // namespace MueLu
+#endif  // HAVE_MUELU_SERIAL and HAVE_MUELU_EPETRA
+
+#endif  // ifndef MUELU_CREATE_EPETRA_PRECONDITIONER_CPP
diff --git a/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.hpp b/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.hpp
index 9ffb14c17f45..a06a788afc47 100644
--- a/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.hpp
+++ b/packages/muelu/adapters/epetra/MueLu_CreateEpetraPreconditioner.hpp
@@ -15,43 +15,42 @@
 #if defined(HAVE_MUELU_EPETRA)
 namespace MueLu {
 
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a EpetraCrs_Matrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] paramListIn Parameter list
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>&   inA,
-                             // FIXME: why is it non-const
-                             Teuchos::ParameterList& paramListIn);
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] xmlFileName XML file containing MueLu options.
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>  & A,
-                             const std::string& xmlFileName);
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
-    @ingroup MueLuAdapters
-    Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    */
-  Teuchos::RCP<MueLu::EpetraOperator>
-  CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>  & A,
-                             const std::string& xmlFileName);
-
-  void ReuseEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA, MueLu::EpetraOperator& Op);
-
-
-} //namespace
-#endif // HAVE_MUELU_SERIAL and HAVE_MUELU_EPETRA
-
-#endif //ifndef MUELU_CREATE_EPETRA_PRECONDITIONER_HPP
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a EpetraCrs_Matrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] paramListIn Parameter list
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA,
+                           // FIXME: why is it non-const
+                           Teuchos::ParameterList& paramListIn);
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] xmlFileName XML file containing MueLu options.
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& A,
+                           const std::string& xmlFileName);
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Epetra.
+  @ingroup MueLuAdapters
+  Given a Epetra_CrsMatrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  */
+Teuchos::RCP<MueLu::EpetraOperator>
+CreateEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& A,
+                           const std::string& xmlFileName);
+
+void ReuseEpetraPreconditioner(const Teuchos::RCP<Epetra_CrsMatrix>& inA, MueLu::EpetraOperator& Op);
+
+}  // namespace MueLu
+#endif  // HAVE_MUELU_SERIAL and HAVE_MUELU_EPETRA
+
+#endif  // ifndef MUELU_CREATE_EPETRA_PRECONDITIONER_HPP
diff --git a/packages/muelu/adapters/epetra/MueLu_EpetraOperator.cpp b/packages/muelu/adapters/epetra/MueLu_EpetraOperator.cpp
index 0134537ab2e8..19e49b864c74 100644
--- a/packages/muelu/adapters/epetra/MueLu_EpetraOperator.cpp
+++ b/packages/muelu/adapters/epetra/MueLu_EpetraOperator.cpp
@@ -58,12 +58,12 @@ namespace MueLu {
 int EpetraOperator::ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const {
   try {
     // There is no rcpFromRef(const T&), so we need to do const_cast
-    const Xpetra::EpetraMultiVectorT<GO,NO> eX(rcpFromRef(const_cast<Epetra_MultiVector&>(X)));
-    Xpetra::EpetraMultiVectorT<GO,NO>       eY(rcpFromRef(Y));
+    const Xpetra::EpetraMultiVectorT<GO, NO> eX(rcpFromRef(const_cast<Epetra_MultiVector&>(X)));
+    Xpetra::EpetraMultiVectorT<GO, NO> eY(rcpFromRef(Y));
     // Generally, we assume two different vectors, but AztecOO uses a single vector
     if (X.Values() == Y.Values()) {
       // X and Y point to the same memory, use an additional vector
-      RCP<Xpetra::EpetraMultiVectorT<GO,NO> > tmpY = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GO,NO>(eY.getMap(), eY.getNumVectors()));
+      RCP<Xpetra::EpetraMultiVectorT<GO, NO>> tmpY = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GO, NO>(eY.getMap(), eY.getNumVectors()));
       // InitialGuessIsZero in MueLu::Hierarchy.Iterate() does not zero out components, it
       // only assumes that user provided an already zeroed out vector
       bool initialGuessZero = true;
@@ -83,72 +83,72 @@ int EpetraOperator::ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector
     }
 
   } catch (std::exception& e) {
-    //TODO: error msg directly on std::cerr?
+    // TODO: error msg directly on std::cerr?
     std::cerr << "Caught an exception in MueLu::EpetraOperator::ApplyInverse():" << std::endl
-        << e.what() << std::endl;
+              << e.what() << std::endl;
     return -1;
   }
   return 0;
 }
 
 const Epetra_Comm& EpetraOperator::Comm() const {
-  RCP<Matrix> A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix> >("A");
+  RCP<Matrix> A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix>>("A");
 
-  //TODO: This code is not pretty
-  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> > epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> >(A);
+  // TODO: This code is not pretty
+  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>> epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>>(A);
   if (epbA != Teuchos::null) {
-    RCP<const Xpetra::Matrix<SC, LO, GO, NO> > blockMat = epbA->getMatrix(0,0);
-    RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO> > blockCrsWrap = Teuchos::rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO> >(blockMat);
+    RCP<const Xpetra::Matrix<SC, LO, GO, NO>> blockMat            = epbA->getMatrix(0, 0);
+    RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> blockCrsWrap = Teuchos::rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(blockMat);
     if (blockCrsWrap == Teuchos::null)
       throw Exceptions::BadCast("MueLu::EpetraOperator::Comm(): Cast from block (0,0) to CrsMatrixWrap failed. Could be a block matrix. TODO implement recursive support for block matrices.");
-    RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO> >(blockCrsWrap->getCrsMatrix());
+    RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>> tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(blockCrsWrap->getCrsMatrix());
     if (tmp_ECrsMtx == Teuchos::null)
       throw Exceptions::BadCast("MueLu::EpetraOperator::Comm(): Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
     RCP<Epetra_CrsMatrix> epA = tmp_ECrsMtx->getEpetra_CrsMatrixNonConst();
     return epA->Comm();
   }
 
-  RCP<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >(A);
+  RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(A);
   if (crsOp == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-  const RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO>>(crsOp->getCrsMatrix());
+  const RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(crsOp->getCrsMatrix());
   if (tmp_ECrsMtx == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
   return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst()->Comm();
 }
 
 const Epetra_Map& EpetraOperator::OperatorDomainMap() const {
-  RCP<Xpetra::Matrix<SC,LO,GO,NO> > A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix> >("A");
+  RCP<Xpetra::Matrix<SC, LO, GO, NO>> A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix>>("A");
 
-  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> > epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> >(A);
+  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>> epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>>(A);
   if (epbA != Teuchos::null)
-    return Xpetra::toEpetra(epbA->getFullDomainMap()); // TODO check me
+    return Xpetra::toEpetra(epbA->getFullDomainMap());  // TODO check me
 
-  RCP<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >(A);
+  RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(A);
   if (crsOp == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-  const RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO>>(crsOp->getCrsMatrix());
+  const RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(crsOp->getCrsMatrix());
   if (tmp_ECrsMtx == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
   return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst()->DomainMap();
 }
 
-const Epetra_Map & EpetraOperator::OperatorRangeMap() const {
-  RCP<Xpetra::Matrix<SC,LO,GO,NO> > A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix> >("A");
+const Epetra_Map& EpetraOperator::OperatorRangeMap() const {
+  RCP<Xpetra::Matrix<SC, LO, GO, NO>> A = Hierarchy_->GetLevel(0)->Get<RCP<Matrix>>("A");
 
-  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> > epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> >(A);
+  RCP<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>> epbA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<SC, LO, GO, NO>>(A);
   if (epbA != Teuchos::null)
     return Xpetra::toEpetra(epbA->getFullRangeMap());
 
-  RCP<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >(A);
+  RCP<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<SC, LO, GO, NO>>(A);
   if (crsOp == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-  const RCP<const Xpetra::EpetraCrsMatrixT<GO,NO>> &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO,NO>>(crsOp->getCrsMatrix());
+  const RCP<const Xpetra::EpetraCrsMatrixT<GO, NO>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GO, NO>>(crsOp->getCrsMatrix());
   if (tmp_ECrsMtx == Teuchos::null)
     throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
   return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst()->RangeMap();
 }
 
-} // namespace
+}  // namespace MueLu
 
-#endif // #if defined(HAVE_MUELU_SERIAL) and defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) and defined(HAVE_MUELU_EPETRA)
diff --git a/packages/muelu/adapters/epetra/MueLu_EpetraOperator.hpp b/packages/muelu/adapters/epetra/MueLu_EpetraOperator.hpp
index e20f8650b2da..0b325cc5ffbb 100644
--- a/packages/muelu/adapters/epetra/MueLu_EpetraOperator.hpp
+++ b/packages/muelu/adapters/epetra/MueLu_EpetraOperator.hpp
@@ -50,7 +50,7 @@
 
 #include <Epetra_Operator.h>
 #include "MueLu_Hierarchy.hpp"
-//TODO: Kokkos headers
+// TODO: Kokkos headers
 
 #if defined(HAVE_MUELU_SERIAL) and defined(HAVE_MUELU_EPETRA)
 
@@ -60,108 +60,105 @@ namespace MueLu {
     @brief Turns a MueLu::Hierarchy into a Epetra_Operator.
     It allows MueLu to be used as a preconditioner for AztecOO (for instance).
 */
-  class EpetraOperator : public Epetra_Operator {
-    typedef double              SC;
-    typedef int                 LO;
-    typedef int                 GO;
-    typedef Xpetra::EpetraNode  NO;
+class EpetraOperator : public Epetra_Operator {
+  typedef double SC;
+  typedef int LO;
+  typedef int GO;
+  typedef Xpetra::EpetraNode NO;
 
-    typedef Xpetra::Matrix<SC,LO,GO,NO>                     Matrix;
-    typedef MueLu::Hierarchy<SC,LO,GO,NO>                   Hierarchy;
-    typedef MueLu::Utilities<SC,LO,GO,NO>                       Utils;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef MueLu::Hierarchy<SC, LO, GO, NO> Hierarchy;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
 
-  public:
+ public:
+  //! @name Constructor/Destructor
+  //@{
 
-    //! @name Constructor/Destructor
-    //@{
+  //! Constructor
+  EpetraOperator(const RCP<Hierarchy>& H)
+    : Hierarchy_(H) {}
 
-    //! Constructor
-    EpetraOperator(const RCP<Hierarchy>& H) : Hierarchy_(H) { }
+  //! Destructor.
+  virtual ~EpetraOperator() {}
 
-    //! Destructor.
-    virtual ~EpetraOperator() { }
+  //@}
 
-    //@}
+  int SetUseTranspose(bool /* UseTransposeBool */) { return -1; }
 
-    int SetUseTranspose(bool /* UseTransposeBool */) { return -1; }
+  //! @name Mathematical functions
+  //@{
 
-    //! @name Mathematical functions
-    //@{
+  //! Returns the result of a Epetra_Operator applied to a Epetra_MultiVector X in Y.
+  /*!
+    \param In
+    X - A Epetra_MultiVector of dimension NumVectors to multiply with matrix.
+    \param Out
+    Y -A Epetra_MultiVector of dimension NumVectors containing result.
 
-    //! Returns the result of a Epetra_Operator applied to a Epetra_MultiVector X in Y.
-    /*!
-      \param In
-      X - A Epetra_MultiVector of dimension NumVectors to multiply with matrix.
-      \param Out
-      Y -A Epetra_MultiVector of dimension NumVectors containing result.
+    \return Integer error code, set to 0 if successful.
+  */
+  int Apply(const Epetra_MultiVector& /* X */, Epetra_MultiVector& /* Y */) const { return -1; }
 
-      \return Integer error code, set to 0 if successful.
-    */
-    int Apply(const Epetra_MultiVector& /* X */, Epetra_MultiVector& /* Y */) const { return -1; }
+  //! Returns the result of a Epetra_Operator inverse applied to an Epetra_MultiVector X in Y.
+  /*!
+    \param In
+    X - A Epetra_MultiVector of dimension NumVectors to solve for.
+    \param Out
+    Y -A Epetra_MultiVector of dimension NumVectors containing result.
 
-    //! Returns the result of a Epetra_Operator inverse applied to an Epetra_MultiVector X in Y.
-    /*!
-      \param In
-      X - A Epetra_MultiVector of dimension NumVectors to solve for.
-      \param Out
-      Y -A Epetra_MultiVector of dimension NumVectors containing result.
+    \return Integer error code, set to 0 if successful.
 
-      \return Integer error code, set to 0 if successful.
+    \warning In order to work with AztecOO, any implementation of this method must
+    support the case where X and Y are the same object.
+  */
+  int ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const;
 
-      \warning In order to work with AztecOO, any implementation of this method must
-      support the case where X and Y are the same object.
-    */
-    int ApplyInverse(const Epetra_MultiVector& X, Epetra_MultiVector& Y) const;
+  //! Returns the infinity norm of the global matrix.
+  /* Returns the quantity \f$ \| A \|_\infty\f$ such that
+     \f[\| A \|_\infty = \max_{1\lei\lem} \sum_{j=1}^n |a_{ij}| \f].
 
-    //! Returns the infinity norm of the global matrix.
-    /* Returns the quantity \f$ \| A \|_\infty\f$ such that
-       \f[\| A \|_\infty = \max_{1\lei\lem} \sum_{j=1}^n |a_{ij}| \f].
+     \warning This method must not be called unless HasNormInf() returns true.
+  */
+  double NormInf() const { return 0; }
+  //@}
 
-       \warning This method must not be called unless HasNormInf() returns true.
-    */
-    double NormInf() const { return 0; }
-    //@}
+  //! @name Attribute access functions
+  //@{
 
-    //! @name Attribute access functions
-    //@{
+  //! Returns a character string describing the operator
+  const char* Label() const { return "MueLu::Hierarchy"; }
 
-    //! Returns a character string describing the operator
-    const char * Label() const { return "MueLu::Hierarchy"; }
+  //! Returns the current UseTranspose setting.
+  bool UseTranspose() const { return false; }
 
-    //! Returns the current UseTranspose setting.
-    bool UseTranspose() const { return false; }
+  //! Returns true if the \e this object can provide an approximate Inf-norm, false otherwise.
+  bool HasNormInf() const { return 0; }
 
-    //! Returns true if the \e this object can provide an approximate Inf-norm, false otherwise.
-    bool HasNormInf() const { return 0; }
+  //! Returns a pointer to the Epetra_Comm communicator associated with this operator.
+  const Epetra_Comm& Comm() const;
 
-    //! Returns a pointer to the Epetra_Comm communicator associated with this operator.
-    const Epetra_Comm & Comm() const;
+  //! Returns the Epetra_Map object associated with the domain of this operator.
+  const Epetra_Map& OperatorDomainMap() const;
 
-    //! Returns the Epetra_Map object associated with the domain of this operator.
-    const Epetra_Map & OperatorDomainMap() const;
+  //! Returns the Epetra_Map object associated with the range of this operator.
+  const Epetra_Map& OperatorRangeMap() const;
 
-    //! Returns the Epetra_Map object associated with the range of this operator.
-    const Epetra_Map & OperatorRangeMap() const;
+  //@}
 
-    //@}
+  //! @name MueLu specific
+  //@{
 
-    //! @name MueLu specific
-    //@{
+  //! Direct access to the underlying MueLu::Hierarchy.
+  RCP<Hierarchy> GetHierarchy() const { return Hierarchy_; }
 
-    //! Direct access to the underlying MueLu::Hierarchy.
-    RCP<Hierarchy> GetHierarchy() const { return Hierarchy_; }
+  //@}
 
-    //@}
+ private:
+  RCP<Hierarchy> Hierarchy_;
+};
 
+}  // namespace MueLu
 
-  private:
+#endif  // HAVE_MUELU_EPETRA and HAVE_MUELU_SERIAL
 
-    RCP<Hierarchy> Hierarchy_;
-
-  };
-
-} // namespace
-
-#endif // HAVE_MUELU_EPETRA and HAVE_MUELU_SERIAL
-
-#endif // MUELU_EPETRAOPERATOR_HPP
+#endif  // MUELU_EPETRAOPERATOR_HPP
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.cpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.cpp
index f414dd518bee..399cd351599c 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.cpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.cpp
@@ -57,7 +57,7 @@
 // Do explicit instantiation of MueLu::Details::LinearSolverFactory,
 // for Epetra objects.
 template class MueLu::Details::LinearSolverFactory<Epetra_MultiVector, Epetra_Operator, double>;
-#endif // HAVE_MUELU_EPETRA
+#endif  // HAVE_MUELU_EPETRA
 
 // Define typedefs that make the Tpetra macros work
 TPETRA_ETI_MANGLING_TYPEDEFS()
@@ -65,8 +65,8 @@ TPETRA_ETI_MANGLING_TYPEDEFS()
 // Do explicit instantiation of MueLu::Details::LinearSolverFactory, for
 // Tpetra objects, for all combinations of Tpetra template parameters
 // for which Tpetra does explicit template instantiation (ETI).
-TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR( MUELU_DETAILS_LINEARSOLVERFACTORY_INSTANT )
+TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_DETAILS_LINEARSOLVERFACTORY_INSTANT)
 
 // TODO amk: do we also have to do this for Xpetra?
 
-#endif // HAVE_MUELU_EXPLICIT_INSTANTIATION
\ No newline at end of file
+#endif  // HAVE_MUELU_EXPLICIT_INSTANTIATION
\ No newline at end of file
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.hpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.hpp
index ffe7875c07e6..2bc5ba10c4cf 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.hpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory.hpp
@@ -46,7 +46,7 @@
 
 #include "MueLu_Details_LinearSolverFactory_decl.hpp"
 #ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-  #include "MueLu_Details_LinearSolverFactory_def.hpp"
+#include "MueLu_Details_LinearSolverFactory_def.hpp"
 #endif
 
-#endif // MUELU_DETAILS_LINEARSOLVERFACTORY_HPP
\ No newline at end of file
+#endif  // MUELU_DETAILS_LINEARSOLVERFACTORY_HPP
\ No newline at end of file
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_decl.hpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_decl.hpp
index f74b4f341442..df25f6680749 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_decl.hpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_decl.hpp
@@ -54,58 +54,57 @@
 namespace MueLu {
 namespace Details {
 
-  /// \class LinearSolverFactory
-  /// \brief Interface for a "factory" that creates MueLu solvers.
+/// \class LinearSolverFactory
+/// \brief Interface for a "factory" that creates MueLu solvers.
+///
+/// \tparam MV Type of a (multi)vector, representing either the
+///   solution(s) X or the right-hand side(s) B of a linear system
+///   AX=B.  For example, with Tpetra, use a Tpetra::MultiVector
+///   specialization.  A <i>multivector</i> is a single data structure
+///   containing zero or more vectors with the same dimensions and
+///   layout.
+///
+/// \tparam OP Type of a matrix or linear operator that this Solver
+///   understands.  For example, for Tpetra, use a Tpetra::Operator
+///   specialization.  Always use the most abstract interface
+///   possible; solvers should dynamic_cast to the subclass they
+///   need.  Also, be consistent: using different classes here
+///   (e.g., Tpetra::RowMatrix instead of Tpetra::Operator) means
+///   more expensive explicit template instantiation.
+///
+/// \tparam NormType Type of the norm of a residual.
+template <class MV, class OP, class NormType>
+class LinearSolverFactory : public Trilinos::Details::LinearSolverFactory<MV, OP, NormType> {
+ public:
+  /// \brief Get an instance of a MueLu solver.
   ///
-  /// \tparam MV Type of a (multi)vector, representing either the
-  ///   solution(s) X or the right-hand side(s) B of a linear system
-  ///   AX=B.  For example, with Tpetra, use a Tpetra::MultiVector
-  ///   specialization.  A <i>multivector</i> is a single data structure
-  ///   containing zero or more vectors with the same dimensions and
-  ///   layout.
+  /// The solver is wrapped in a Trilinos::Details::LinearSolver
+  /// interface.
   ///
-  /// \tparam OP Type of a matrix or linear operator that this Solver
-  ///   understands.  For example, for Tpetra, use a Tpetra::Operator
-  ///   specialization.  Always use the most abstract interface
-  ///   possible; solvers should dynamic_cast to the subclass they
-  ///   need.  Also, be consistent: using different classes here
-  ///   (e.g., Tpetra::RowMatrix instead of Tpetra::Operator) means
-  ///   more expensive explicit template instantiation.
+  /// \param solverName [in] The solver's name.  Names are case
+  ///   sensitive
+  /// \return A pointer to the solver, if the name was valid; else,
+  ///   a null pointer (Teuchos::null).
+  virtual Teuchos::RCP<Trilinos::Details::LinearSolver<MV, OP, NormType> >
+  getLinearSolver(const std::string& solverName);
+
+  /// \brief Register this LinearSolverFactory with the central registry.
+  ///
+  /// Register this LinearSolverFactory with the central registry, for
+  /// the given SC, LO, GO, NT template parameters.  This will let any
+  /// clients of Trilinos::Details::getLinearSolver create MueLu
+  /// solvers with those template parameters.
   ///
-  /// \tparam NormType Type of the norm of a residual.
-  template<class MV, class OP, class NormType>
-  class LinearSolverFactory :
-    public Trilinos::Details::LinearSolverFactory<MV, OP, NormType> {
-  public:
-    /// \brief Get an instance of a MueLu solver.
-    ///
-    /// The solver is wrapped in a Trilinos::Details::LinearSolver
-    /// interface.
-    ///
-    /// \param solverName [in] The solver's name.  Names are case
-    ///   sensitive
-    /// \return A pointer to the solver, if the name was valid; else,
-    ///   a null pointer (Teuchos::null).
-    virtual Teuchos::RCP<Trilinos::Details::LinearSolver<MV, OP, NormType> >
-    getLinearSolver (const std::string& solverName);
-    
-    /// \brief Register this LinearSolverFactory with the central registry.
-    ///
-    /// Register this LinearSolverFactory with the central registry, for
-    /// the given SC, LO, GO, NT template parameters.  This will let any
-    /// clients of Trilinos::Details::getLinearSolver create MueLu
-    /// solvers with those template parameters.
-    ///
-    /// You may call this function multiple times; it will only have an
-    /// effect the first time (it is idempotent).
-    ///
-    /// Users do not normally have to call this function.  MueLu
-    /// automatically registers its LinearSolverFactory with the central
-    /// repository, for all enabled template parameter combinations.
-    static void registerLinearSolverFactory ();
-  };
-    
-} // namespace Details
-} // namespace MueLu
+  /// You may call this function multiple times; it will only have an
+  /// effect the first time (it is idempotent).
+  ///
+  /// Users do not normally have to call this function.  MueLu
+  /// automatically registers its LinearSolverFactory with the central
+  /// repository, for all enabled template parameter combinations.
+  static void registerLinearSolverFactory();
+};
+
+}  // namespace Details
+}  // namespace MueLu
 
-#endif // MUELU_DETAILS_LINEARSOLVERFACTORY_DECL_HPP
+#endif  // MUELU_DETAILS_LINEARSOLVERFACTORY_DECL_HPP
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_def.hpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_def.hpp
index 39e707cb0938..0e8f151dbecf 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_def.hpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_LinearSolverFactory_def.hpp
@@ -54,169 +54,149 @@
 #include <type_traits>
 
 #ifdef HAVE_MUELU_EPETRA
-#  include "Epetra_CrsMatrix.h"
+#include "Epetra_CrsMatrix.h"
 #include "MueLu_CreateEpetraPreconditioner.hpp"
-#endif // HAVE_MUELU_EPETRA
+#endif  // HAVE_MUELU_EPETRA
 
-#  include "Tpetra_Operator.hpp"
-#  include "MueLu_CreateTpetraPreconditioner.hpp"
+#include "Tpetra_Operator.hpp"
+#include "MueLu_CreateTpetraPreconditioner.hpp"
 
 namespace MueLu {
 namespace Details {
 
-template<class MV, class OP, class NormType>
-class LinearSolver :
-    public Trilinos::Details::LinearSolver<MV, OP, NormType>,
-    virtual public Teuchos::Describable
-{
-
-public:
-
+template <class MV, class OP, class NormType>
+class LinearSolver : public Trilinos::Details::LinearSolver<MV, OP, NormType>,
+                     virtual public Teuchos::Describable {
+ public:
   /// \brief Constructor.
-  LinearSolver () {}
+  LinearSolver() {}
 
   //! Destructor (virtual for memory safety).
-  virtual ~LinearSolver () {}
+  virtual ~LinearSolver() {}
 
   /// \brief Set the Solver's matrix.
   ///
   /// \param A [in] Pointer to the matrix A in the linear system(s)
   ///   AX=B to solve.
-  void setMatrix (const Teuchos::RCP<const OP>& A);
+  void setMatrix(const Teuchos::RCP<const OP>& A);
 
   //! Get a pointer to this Solver's matrix.
-  Teuchos::RCP<const OP> getMatrix () const {
+  Teuchos::RCP<const OP> getMatrix() const {
     return A_;
   }
 
   //! Solve the linear system(s) AX=B.
-  void solve (MV& X, const MV& B);
+  void solve(MV& X, const MV& B);
 
   //! Set this solver's parameters.
-  void setParameters (const Teuchos::RCP<Teuchos::ParameterList>& params);
+  void setParameters(const Teuchos::RCP<Teuchos::ParameterList>& params);
 
   /// \brief Set up any part of the solve that depends on the
   ///   structure of the input matrix, but not its numerical values.
-  void symbolic () {}
+  void symbolic() {}
 
   /// \brief Set up any part of the solve that depends on both the
   ///   structure and the numerical values of the input matrix.
-  void numeric ();
+  void numeric();
 
   //! Implementation of Teuchos::Describable::description.
-  std::string description () const;
+  std::string description() const;
 
   //! Implementation of Teuchos::Describable::describe.
   void
-  describe (Teuchos::FancyOStream& out,
-            const Teuchos::EVerbosityLevel verbLevel =
-            Teuchos::Describable::verbLevel_default) const;
+  describe(Teuchos::FancyOStream& out,
+           const Teuchos::EVerbosityLevel verbLevel =
+               Teuchos::Describable::verbLevel_default) const;
 
-private:
+ private:
   Teuchos::RCP<const OP> A_;
   Teuchos::RCP<Teuchos::ParameterList> params_;
 };
-    
+
 // Why does MueLu_EpetraOperator insist on HAVE_MUELU_SERIAL?
 #if defined(HAVE_MUELU_SERIAL) and defined(HAVE_MUELU_EPETRA)
-template<>
-class LinearSolver<Epetra_MultiVector, Epetra_Operator, double> :
-    public Trilinos::Details::LinearSolver<Epetra_MultiVector, Epetra_Operator, double>,
-    virtual public Teuchos::Describable
-{
-
-public:
-
+template <>
+class LinearSolver<Epetra_MultiVector, Epetra_Operator, double> : public Trilinos::Details::LinearSolver<Epetra_MultiVector, Epetra_Operator, double>,
+                                                                  virtual public Teuchos::Describable {
+ public:
   /// \brief Constructor.
-  LinearSolver () :
-    changedA_(false),
-    changedParams_(false)
-  {}
+  LinearSolver()
+    : changedA_(false)
+    , changedParams_(false) {}
 
   //! Destructor (virtual for memory safety).
-  virtual ~LinearSolver () {}
+  virtual ~LinearSolver() {}
 
   /// \brief Set the Solver's matrix.
   ///
   /// \param A [in] Pointer to the matrix A in the linear system(s)
   ///   AX=B to solve.
-  void setMatrix (const Teuchos::RCP<const Epetra_Operator>& A)
-  {
+  void setMatrix(const Teuchos::RCP<const Epetra_Operator>& A) {
     const char prefix[] = "MueLu::Details::LinearSolver::setMatrix: ";
-    
-    if(A != A_)
-    {
-      if(solver_ != Teuchos::null)
+
+    if (A != A_) {
+      if (solver_ != Teuchos::null)
         changedA_ = true;
-      
+
       A_ = rcp_dynamic_cast<const Epetra_CrsMatrix>(A);
-      TEUCHOS_TEST_FOR_EXCEPTION
-        (A_.is_null(), std::runtime_error, prefix << "MueLu requires "
-         "an Epetra_CrsMatrix, but the matrix you provided is of a "
-         "different type.  Please provide an Epetra_CrsMatrix instead.");
+      TEUCHOS_TEST_FOR_EXCEPTION(A_.is_null(), std::runtime_error, prefix << "MueLu requires "
+                                                                             "an Epetra_CrsMatrix, but the matrix you provided is of a "
+                                                                             "different type.  Please provide an Epetra_CrsMatrix instead.");
     }
   }
 
   //! Get a pointer to this Solver's matrix.
-  Teuchos::RCP<const Epetra_Operator> getMatrix () const {
+  Teuchos::RCP<const Epetra_Operator> getMatrix() const {
     return A_;
   }
 
   //! Solve the linear system(s) AX=B.
-  void solve (Epetra_MultiVector& X, const Epetra_MultiVector& B)
-  {
+  void solve(Epetra_MultiVector& X, const Epetra_MultiVector& B) {
     // TODO amk: Do we assume the user has called numeric before solve, or should we call it for them?
     const char prefix[] = "MueLu::Details::LinearSolver::solve: ";
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (solver_.is_null (), std::runtime_error, prefix << "The solver does not "
-       "exist yet.  You must call numeric() before you may call this method.");
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (changedA_, std::runtime_error, prefix << "The matrix A has been reset "
-       "since the last call to numeric().  Please call numeric() again.");
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (changedParams_, std::runtime_error, prefix << "The parameters have been reset "
-       "since the last call to numeric().  Please call numeric() again.");
-    
+    TEUCHOS_TEST_FOR_EXCEPTION(solver_.is_null(), std::runtime_error, prefix << "The solver does not "
+                                                                                "exist yet.  You must call numeric() before you may call this method.");
+    TEUCHOS_TEST_FOR_EXCEPTION(changedA_, std::runtime_error, prefix << "The matrix A has been reset "
+                                                                        "since the last call to numeric().  Please call numeric() again.");
+    TEUCHOS_TEST_FOR_EXCEPTION(changedParams_, std::runtime_error, prefix << "The parameters have been reset "
+                                                                             "since the last call to numeric().  Please call numeric() again.");
+
     int err = solver_->ApplyInverse(B, X);
-    
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (err != 0, std::runtime_error, prefix << "EpetraOperator::ApplyInverse returned "
-       "nonzero error code " << err);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(err != 0, std::runtime_error, prefix << "EpetraOperator::ApplyInverse returned "
+                                                                       "nonzero error code "
+                                                                    << err);
   }
 
   //! Set this solver's parameters.
-  void setParameters (const Teuchos::RCP<Teuchos::ParameterList>& params)
-  {
-    if(solver_ != Teuchos::null && params != params_)
+  void setParameters(const Teuchos::RCP<Teuchos::ParameterList>& params) {
+    if (solver_ != Teuchos::null && params != params_)
       changedParams_ = true;
-    
+
     params_ = params;
   }
 
   /// \brief Set up any part of the solve that depends on the
   ///   structure of the input matrix, but not its numerical values.
-  void symbolic () {}
+  void symbolic() {}
 
   /// \brief Set up any part of the solve that depends on both the
   ///   structure and the numerical values of the input matrix.
-  void numeric ()
-  {
+  void numeric() {
     const char prefix[] = "MueLu::Details::LinearSolver::numeric: ";
-    
+
     // If the solver is up-to-date, leave it alone
-    if(solver_ == Teuchos::null || changedA_ || changedParams_)
-    {
-      changedA_ = false;
+    if (solver_ == Teuchos::null || changedA_ || changedParams_) {
+      changedA_      = false;
       changedParams_ = false;
-      
-      TEUCHOS_TEST_FOR_EXCEPTION
-        (A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
-         "set yet.  You must call setMatrix() with a nonnull matrix before you may "
-         "call this method.");
-      
+
+      TEUCHOS_TEST_FOR_EXCEPTION(A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
+                                                                                    "set yet.  You must call setMatrix() with a nonnull matrix before you may "
+                                                                                    "call this method.");
+
       // TODO: We should not have to cast away the constness here
       // TODO: See bug 6462
-      if(params_ != Teuchos::null)
+      if (params_ != Teuchos::null)
         solver_ = CreateEpetraPreconditioner(rcp_const_cast<Epetra_CrsMatrix>(A_), *params_);
       else
         solver_ = CreateEpetraPreconditioner(rcp_const_cast<Epetra_CrsMatrix>(A_));
@@ -224,239 +204,209 @@ class LinearSolver<Epetra_MultiVector, Epetra_Operator, double> :
   }
 
   //! Implementation of Teuchos::Describable::description.
-  std::string description () const
-  {
+  std::string description() const {
     if (solver_.is_null()) {
       return "\"MueLu::Details::LinearSolver\": {MV: Epetra_MultiVector, OP: Epetra_Operator, NormType: double}";
-    }
-    else {
-      return solver_->GetHierarchy()->description ();
+    } else {
+      return solver_->GetHierarchy()->description();
     }
   }
 
   //! Implementation of Teuchos::Describable::describe.
   void
-  describe (Teuchos::FancyOStream& out,
-            const Teuchos::EVerbosityLevel verbLevel =
-            Teuchos::Describable::verbLevel_default) const
-  {
+  describe(Teuchos::FancyOStream& out,
+           const Teuchos::EVerbosityLevel verbLevel =
+               Teuchos::Describable::verbLevel_default) const {
     using std::endl;
     if (solver_.is_null()) {
-      if(verbLevel > Teuchos::VERB_NONE) {
-        Teuchos::OSTab tab0 (out);
+      if (verbLevel > Teuchos::VERB_NONE) {
+        Teuchos::OSTab tab0(out);
         out << "\"MueLu::Details::LinearSolver\":" << endl;
-        Teuchos::OSTab tab1 (out);
+        Teuchos::OSTab tab1(out);
         out << "MV: Epetra_MultiVector" << endl
             << "OP: Epetra_Operator" << endl
             << "NormType: double" << endl;
       }
-    }
-    else {
-      solver_->GetHierarchy()->describe (out, verbLevel);
+    } else {
+      solver_->GetHierarchy()->describe(out, verbLevel);
     }
   }
 
-private:
+ private:
   Teuchos::RCP<const Epetra_CrsMatrix> A_;
   Teuchos::RCP<Teuchos::ParameterList> params_;
   Teuchos::RCP<EpetraOperator> solver_;
   bool changedA_;
   bool changedParams_;
 };
-#endif // HAVE_MUELU_EPETRA
-
-template<class Scalar, class LO, class GO, class Node>
-class LinearSolver<Tpetra::MultiVector<Scalar,LO,GO,Node>,
-                   Tpetra::Operator<Scalar,LO,GO,Node>,
-                   typename Teuchos::ScalarTraits<Scalar>::magnitudeType> :
-    public Trilinos::Details::LinearSolver<Tpetra::MultiVector<Scalar,LO,GO,Node>,
-                                           Tpetra::Operator<Scalar,LO,GO,Node>,
-                                           typename Teuchos::ScalarTraits<Scalar>::magnitudeType>,
-    virtual public Teuchos::Describable
-{
-
-public:
-
+#endif  // HAVE_MUELU_EPETRA
+
+template <class Scalar, class LO, class GO, class Node>
+class LinearSolver<Tpetra::MultiVector<Scalar, LO, GO, Node>,
+                   Tpetra::Operator<Scalar, LO, GO, Node>,
+                   typename Teuchos::ScalarTraits<Scalar>::magnitudeType> : public Trilinos::Details::LinearSolver<Tpetra::MultiVector<Scalar, LO, GO, Node>,
+                                                                                                                   Tpetra::Operator<Scalar, LO, GO, Node>,
+                                                                                                                   typename Teuchos::ScalarTraits<Scalar>::magnitudeType>,
+                                                                            virtual public Teuchos::Describable {
+ public:
   /// \brief Constructor.
-  LinearSolver () :
-    changedA_(false),
-    changedParams_(false)
-  {}
+  LinearSolver()
+    : changedA_(false)
+    , changedParams_(false) {}
 
   //! Destructor (virtual for memory safety).
-  virtual ~LinearSolver () {}
+  virtual ~LinearSolver() {}
 
   /// \brief Set the Solver's matrix.
   ///
   /// \param A [in] Pointer to the matrix A in the linear system(s)
   ///   AX=B to solve.
-  void setMatrix (const Teuchos::RCP<const Tpetra::Operator<Scalar,LO,GO,Node> >& A)
-  {
-    if(A != A_)
-    {
-      if(solver_ != Teuchos::null)
+  void setMatrix(const Teuchos::RCP<const Tpetra::Operator<Scalar, LO, GO, Node> >& A) {
+    if (A != A_) {
+      if (solver_ != Teuchos::null)
         changedA_ = true;
-      
+
       A_ = A;
     }
   }
 
   //! Get a pointer to this Solver's matrix.
-  Teuchos::RCP<const Tpetra::Operator<Scalar,LO,GO,Node> > getMatrix () const {
+  Teuchos::RCP<const Tpetra::Operator<Scalar, LO, GO, Node> > getMatrix() const {
     return A_;
   }
 
   //! Solve the linear system(s) AX=B.
-  void solve (Tpetra::MultiVector<Scalar,LO,GO,Node>& X, const Tpetra::MultiVector<Scalar,LO,GO,Node>& B)
-  {
+  void solve(Tpetra::MultiVector<Scalar, LO, GO, Node>& X, const Tpetra::MultiVector<Scalar, LO, GO, Node>& B) {
     // TODO amk: Do we assume the user has called numeric before solve, or should we call it for them?
     const char prefix[] = "MueLu::Details::LinearSolver::solve: ";
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (solver_.is_null (), std::runtime_error, prefix << "The solver does not "
-       "exist yet.  You must call numeric() before you may call this method.");
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (changedA_, std::runtime_error, prefix << "The matrix A has been reset "
-       "since the last call to numeric().  Please call numeric() again.");
-    TEUCHOS_TEST_FOR_EXCEPTION
-      (changedParams_, std::runtime_error, prefix << "The parameters have been reset "
-       "since the last call to numeric().  Please call numeric() again.");
-    
+    TEUCHOS_TEST_FOR_EXCEPTION(solver_.is_null(), std::runtime_error, prefix << "The solver does not "
+                                                                                "exist yet.  You must call numeric() before you may call this method.");
+    TEUCHOS_TEST_FOR_EXCEPTION(changedA_, std::runtime_error, prefix << "The matrix A has been reset "
+                                                                        "since the last call to numeric().  Please call numeric() again.");
+    TEUCHOS_TEST_FOR_EXCEPTION(changedParams_, std::runtime_error, prefix << "The parameters have been reset "
+                                                                             "since the last call to numeric().  Please call numeric() again.");
+
     solver_->apply(B, X);
   }
 
   //! Set this solver's parameters.
-  void setParameters (const Teuchos::RCP<Teuchos::ParameterList>& params)
-  {
-    if(solver_ != Teuchos::null && params != params_)
+  void setParameters(const Teuchos::RCP<Teuchos::ParameterList>& params) {
+    if (solver_ != Teuchos::null && params != params_)
       changedParams_ = true;
-    
+
     params_ = params;
   }
 
   /// \brief Set up any part of the solve that depends on the
   ///   structure of the input matrix, but not its numerical values.
-  void symbolic () {}
+  void symbolic() {}
 
   /// \brief Set up any part of the solve that depends on both the
   ///   structure and the numerical values of the input matrix.
-  void numeric ()
-  {
+  void numeric() {
     const char prefix[] = "MueLu::Details::LinearSolver::numeric: ";
-    
+
     // If the solver is up-to-date, leave it alone
-    if(solver_ == Teuchos::null || changedParams_)
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION
-        (A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
-         "set yet.  You must call setMatrix() with a nonnull matrix before you may "
-         "call this method.");
-      
+    if (solver_ == Teuchos::null || changedParams_) {
+      TEUCHOS_TEST_FOR_EXCEPTION(A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
+                                                                                    "set yet.  You must call setMatrix() with a nonnull matrix before you may "
+                                                                                    "call this method.");
+
       // TODO: We should not have to cast away the constness here
       // TODO: See bug 6462
-      if(params_ != Teuchos::null)
-        solver_ = CreateTpetraPreconditioner(rcp_const_cast<Tpetra::Operator<Scalar,LO,GO,Node> >(A_), *params_);
+      if (params_ != Teuchos::null)
+        solver_ = CreateTpetraPreconditioner(rcp_const_cast<Tpetra::Operator<Scalar, LO, GO, Node> >(A_), *params_);
       else
-        solver_ = CreateTpetraPreconditioner(rcp_const_cast<Tpetra::Operator<Scalar,LO,GO,Node> >(A_));
-    }
-    else if(changedA_)
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION
-        (A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
-         "set yet.  You must call setMatrix() with a nonnull matrix before you may "
-         "call this method.");
-      
+        solver_ = CreateTpetraPreconditioner(rcp_const_cast<Tpetra::Operator<Scalar, LO, GO, Node> >(A_));
+    } else if (changedA_) {
+      TEUCHOS_TEST_FOR_EXCEPTION(A_ == Teuchos::null, std::runtime_error, prefix << "The matrix has not been "
+                                                                                    "set yet.  You must call setMatrix() with a nonnull matrix before you may "
+                                                                                    "call this method.");
+
       // TODO: We should not have to cast away the constness here
       // TODO: See bug 6462
-      RCP<const Tpetra::CrsMatrix<Scalar,LO,GO,Node> > helperMat;
-      helperMat = rcp_dynamic_cast<const Tpetra::CrsMatrix<Scalar,LO,GO,Node> >(A_);
-      TEUCHOS_TEST_FOR_EXCEPTION
-        (helperMat.is_null(), std::runtime_error, prefix << "MueLu requires "
-         "a Tpetra::CrsMatrix, but the matrix you provided is of a "
-         "different type.  Please provide a Tpetra::CrsMatrix instead.");
-      ReuseTpetraPreconditioner(rcp_const_cast<Tpetra::CrsMatrix<Scalar,LO,GO,Node> >(helperMat), *solver_);
+      RCP<const Tpetra::CrsMatrix<Scalar, LO, GO, Node> > helperMat;
+      helperMat = rcp_dynamic_cast<const Tpetra::CrsMatrix<Scalar, LO, GO, Node> >(A_);
+      TEUCHOS_TEST_FOR_EXCEPTION(helperMat.is_null(), std::runtime_error, prefix << "MueLu requires "
+                                                                                    "a Tpetra::CrsMatrix, but the matrix you provided is of a "
+                                                                                    "different type.  Please provide a Tpetra::CrsMatrix instead.");
+      ReuseTpetraPreconditioner(rcp_const_cast<Tpetra::CrsMatrix<Scalar, LO, GO, Node> >(helperMat), *solver_);
     }
-    
-    changedA_ = false;
+
+    changedA_      = false;
     changedParams_ = false;
   }
 
   //! Implementation of Teuchos::Describable::description.
-  std::string description () const
-  {
+  std::string description() const {
     using Teuchos::TypeNameTraits;
     if (solver_.is_null()) {
       std::ostringstream os;
       os << "\"MueLu::Details::LinearSolver\": {"
-         << "MV: " << TypeNameTraits<Tpetra::MultiVector<Scalar,LO,GO,Node> >::name()
-         << "OP: " << TypeNameTraits<Tpetra::Operator<Scalar,LO,GO,Node> >::name()
+         << "MV: " << TypeNameTraits<Tpetra::MultiVector<Scalar, LO, GO, Node> >::name()
+         << "OP: " << TypeNameTraits<Tpetra::Operator<Scalar, LO, GO, Node> >::name()
          << "NormType: " << TypeNameTraits<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>::name()
          << "}";
-      return os.str ();
-    }
-    else {
-      return solver_->GetHierarchy()->description ();
+      return os.str();
+    } else {
+      return solver_->GetHierarchy()->description();
     }
   }
 
   //! Implementation of Teuchos::Describable::describe.
   void
-  describe (Teuchos::FancyOStream& out,
-            const Teuchos::EVerbosityLevel verbLevel =
-            Teuchos::Describable::verbLevel_default) const
-  {
-    using Teuchos::TypeNameTraits;
+  describe(Teuchos::FancyOStream& out,
+           const Teuchos::EVerbosityLevel verbLevel =
+               Teuchos::Describable::verbLevel_default) const {
     using std::endl;
+    using Teuchos::TypeNameTraits;
     if (solver_.is_null()) {
-      if(verbLevel > Teuchos::VERB_NONE) {
-        Teuchos::OSTab tab0 (out);
+      if (verbLevel > Teuchos::VERB_NONE) {
+        Teuchos::OSTab tab0(out);
         out << "\"MueLu::Details::LinearSolver\":" << endl;
-        Teuchos::OSTab tab1 (out);
-        out << "MV: " << TypeNameTraits<Tpetra::MultiVector<Scalar,LO,GO,Node> >::name() << endl
-            << "OP: " << TypeNameTraits<Tpetra::Operator<Scalar,LO,GO,Node> >::name() << endl
+        Teuchos::OSTab tab1(out);
+        out << "MV: " << TypeNameTraits<Tpetra::MultiVector<Scalar, LO, GO, Node> >::name() << endl
+            << "OP: " << TypeNameTraits<Tpetra::Operator<Scalar, LO, GO, Node> >::name() << endl
             << "NormType: " << TypeNameTraits<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>::name() << endl;
       }
-    }
-    else {
-      solver_->GetHierarchy()->describe (out, verbLevel);
+    } else {
+      solver_->GetHierarchy()->describe(out, verbLevel);
     }
   }
 
-private:
-  Teuchos::RCP<const Tpetra::Operator<Scalar,LO,GO,Node> > A_;
+ private:
+  Teuchos::RCP<const Tpetra::Operator<Scalar, LO, GO, Node> > A_;
   Teuchos::RCP<Teuchos::ParameterList> params_;
-  Teuchos::RCP<TpetraOperator<Scalar,LO,GO,Node> > solver_;
+  Teuchos::RCP<TpetraOperator<Scalar, LO, GO, Node> > solver_;
   bool changedA_;
   bool changedParams_;
 };
 
-template<class MV, class OP, class NormType>
+template <class MV, class OP, class NormType>
 Teuchos::RCP<Trilinos::Details::LinearSolver<MV, OP, NormType> >
 LinearSolverFactory<MV, OP, NormType>::
-getLinearSolver (const std::string& solverName)
-{
+    getLinearSolver(const std::string& solverName) {
   using Teuchos::rcp;
-  return rcp (new MueLu::Details::LinearSolver<MV, OP, NormType> ());
+  return rcp(new MueLu::Details::LinearSolver<MV, OP, NormType>());
 }
 
-template<class MV, class OP, class NormType>
-void
-LinearSolverFactory<MV, OP, NormType>::
-registerLinearSolverFactory ()
-{
+template <class MV, class OP, class NormType>
+void LinearSolverFactory<MV, OP, NormType>::
+    registerLinearSolverFactory() {
 #ifdef HAVE_TEUCHOSCORE_CXX11
   typedef std::shared_ptr<MueLu::Details::LinearSolverFactory<MV, OP, NormType> > ptr_type;
-  //typedef std::shared_ptr<Trilinos::Details::LinearSolverFactory<MV, OP> > base_ptr_type;
+  // typedef std::shared_ptr<Trilinos::Details::LinearSolverFactory<MV, OP> > base_ptr_type;
 #else
   typedef Teuchos::RCP<MueLu::Details::LinearSolverFactory<MV, OP, NormType> > ptr_type;
-  //typedef Teuchos::RCP<Trilinos::Details::LinearSolverFactory<MV, OP> > base_ptr_type;
-#endif // HAVE_TEUCHOSCORE_CXX11
+  // typedef Teuchos::RCP<Trilinos::Details::LinearSolverFactory<MV, OP> > base_ptr_type;
+#endif  // HAVE_TEUCHOSCORE_CXX11
 
-  ptr_type factory (new MueLu::Details::LinearSolverFactory<MV, OP, NormType> ());
-  Trilinos::Details::registerLinearSolverFactory<MV, OP, NormType> ("MueLu", factory);
+  ptr_type factory(new MueLu::Details::LinearSolverFactory<MV, OP, NormType>());
+  Trilinos::Details::registerLinearSolverFactory<MV, OP, NormType>("MueLu", factory);
 }
 
-} // namespace Details
-} // namespace MueLu
+}  // namespace Details
+}  // namespace MueLu
 
 // Macro for doing explicit instantiation of
 // MueLu::Details::LinearSolverFactory, for Tpetra objects, with
@@ -465,9 +415,9 @@ registerLinearSolverFactory ()
 //
 // We don't have to protect use of Tpetra objects here, or include
 // any header files for them, because this is a macro definition.
-#define MUELU_DETAILS_LINEARSOLVERFACTORY_INSTANT(SC, LO, GO, NT) \
+#define MUELU_DETAILS_LINEARSOLVERFACTORY_INSTANT(SC, LO, GO, NT)                         \
   template class MueLu::Details::LinearSolverFactory<Tpetra::MultiVector<SC, LO, GO, NT>, \
-                                                     Tpetra::Operator<SC, LO, GO, NT>, \
+                                                     Tpetra::Operator<SC, LO, GO, NT>,    \
                                                      typename Tpetra::MultiVector<SC, LO, GO, NT>::mag_type>;
 
-#endif // MUELU_DETAILS_LINEARSOLVERFACTORY_DEF_HPP
+#endif  // MUELU_DETAILS_LINEARSOLVERFACTORY_DEF_HPP
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.cpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.cpp
index 0c3e228d69e4..de8cd6659de0 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.cpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.cpp
@@ -46,8 +46,8 @@
 #include "Tpetra_MultiVector.hpp"
 #include "Tpetra_Operator.hpp"
 #ifdef HAVE_MUELU_EPETRA
-#  include "Epetra_MultiVector.h"
-#  include "Epetra_Operator.h"
+#include "Epetra_MultiVector.h"
+#include "Epetra_Operator.h"
 #endif
 #include "TpetraCore_ETIHelperMacros.h"
 
@@ -66,29 +66,27 @@ TPETRA_ETI_MANGLING_TYPEDEFS()
 // call it LCL_CALL and not LCL_INST.  We are just using the macros to
 // invoke this class method over the set of enabled template
 // parameters.
-#define LCL_CALL( SC, LO, GO, NT ) \
+#define LCL_CALL(SC, LO, GO, NT)                                             \
   ::MueLu::Details::LinearSolverFactory<Tpetra::MultiVector<SC, LO, GO, NT>, \
-                                        Tpetra::Operator<SC, LO, GO, NT>, \
-                                        typename Tpetra::MultiVector<SC, LO, GO, NT>::mag_type>::registerLinearSolverFactory ();
+                                        Tpetra::Operator<SC, LO, GO, NT>,    \
+                                        typename Tpetra::MultiVector<SC, LO, GO, NT>::mag_type>::registerLinearSolverFactory();
 
 namespace MueLu {
 namespace Details {
 
-void
-registerLinearSolverFactory ()
-{
+void registerLinearSolverFactory() {
   // Fill in the body of the function with all the type-specific
   // run-time registration functions, for registering MueLu's
   // LinearSolverFactory with Tpetra objects.
-  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR( LCL_CALL )
-      
+  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(LCL_CALL)
+
   // If Epetra is enabled in MueLu, also register MueLu's
   // LinearSolverFactory for Epetra objects.
 #ifdef HAVE_MUELU_EPETRA
   ::MueLu::Details::LinearSolverFactory<Epetra_MultiVector,
-    Epetra_Operator, double>::registerLinearSolverFactory ();
-#endif // HAVE_MUELU_EPETRA
+                                        Epetra_Operator, double>::registerLinearSolverFactory();
+#endif  // HAVE_MUELU_EPETRA
 }
 
-} // namespace Details
-} // namespace MueLu
\ No newline at end of file
+}  // namespace Details
+}  // namespace MueLu
\ No newline at end of file
diff --git a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.hpp b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.hpp
index 5732cf2f562f..26136f1338de 100644
--- a/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.hpp
+++ b/packages/muelu/adapters/linear_solver_factory/MueLu_Details_registerLinearSolverFactory.hpp
@@ -79,12 +79,12 @@ namespace Details {
 /// If you need to register MueLu's LinearSolverFactory for a set of
 /// template parameters that is <i>not</i> enabled, see
 /// MueLu_Details_LinearSolverFactory.hpp (in this directory).
-void registerLinearSolverFactory ();
+void registerLinearSolverFactory();
 
-} // namespace Details
-} // namespace MueLu
+}  // namespace Details
+}  // namespace MueLu
 
-namespace { // (anonymous)
+namespace {  // (anonymous)
 
 // \class RegisterLinearSolverFactory
 // \brief Register MueLu's solver factory/ies with the central registry.
@@ -99,9 +99,9 @@ namespace { // (anonymous)
 // <tt>__attribute__((constructor))</tt>, without actually requiring
 // the syntax extension.)
 class RegisterLinearSolverFactory {
-public:
-  RegisterLinearSolverFactory () {
-    MueLu::Details::registerLinearSolverFactory ();
+ public:
+  RegisterLinearSolverFactory() {
+    MueLu::Details::registerLinearSolverFactory();
   }
 };
 
@@ -110,6 +110,6 @@ class RegisterLinearSolverFactory {
 // MueLu::Details::registerLinearSolverFactory().
 RegisterLinearSolverFactory registerIt;
 
-} // namespace (anonymous)
+}  // namespace
 
-#endif // MUELU_DETAILS_REGISTERLINEARSOLVERFACTORY_HPP
\ No newline at end of file
+#endif  // MUELU_DETAILS_REGISTERLINEARSOLVERFACTORY_HPP
\ No newline at end of file
diff --git a/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.cpp b/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.cpp
index be6d55d36364..a60ea0d89092 100644
--- a/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.cpp
+++ b/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.cpp
@@ -77,4 +77,4 @@ namespace Stratimikos {
 #endif
 #endif
 
-} // namespace Stratimikos
+}  // namespace Stratimikos
diff --git a/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.hpp b/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.hpp
index 080953dcc678..af6306f1c5ca 100644
--- a/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.hpp
+++ b/packages/muelu/adapters/stratimikos/Stratimikos_MueLuHelpers.hpp
@@ -67,71 +67,65 @@
 
 namespace Stratimikos {
 
-  template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
-  void enableMueLu(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLu")
-  {
+template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
+void enableMueLu(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLu") {
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
+  const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
-                               "Stratimikos::enableMueLu cannot add \"" + stratName +"\" because it is already included in builder!");
+  TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
+                             "Stratimikos::enableMueLu cannot add \"" + stratName + "\" because it is already included in builder!");
 
-    typedef Thyra::PreconditionerFactoryBase<Scalar>                                     Base;
-    typedef Thyra::MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
+  typedef Thyra::PreconditionerFactoryBase<Scalar> Base;
+  typedef Thyra::MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
 
-    builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
+  builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
 #endif
-  }
+}
 
-  template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-  MUELU_DEPRECATED void enableMueLu(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLu")
-  {
-    enableMueLu<double,LocalOrdinal,GlobalOrdinal,Node>(builder,stratName);
-  }
+template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+MUELU_DEPRECATED void enableMueLu(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLu") {
+  enableMueLu<double, LocalOrdinal, GlobalOrdinal, Node>(builder, stratName);
+}
 
-  template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
-  void enableMueLuRefMaxwell(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLuRefMaxwell")
-  {
+template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
+void enableMueLuRefMaxwell(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLuRefMaxwell") {
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
+  const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
-                               "Stratimikos::enableMueLuRefMaxwell cannot add \"" + stratName +"\" because it is already included in builder!");
+  TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
+                             "Stratimikos::enableMueLuRefMaxwell cannot add \"" + stratName + "\" because it is already included in builder!");
 
-    typedef Thyra::PreconditionerFactoryBase<Scalar>                                     Base;
-    typedef Thyra::MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
+  typedef Thyra::PreconditionerFactoryBase<Scalar> Base;
+  typedef Thyra::MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
 
-    builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
+  builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
 #endif
-  }
+}
 
-  template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-  MUELU_DEPRECATED void enableMueLuRefMaxwell(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLuRefMaxwell")
-  {
-    enableMueLuRefMaxwell<double,LocalOrdinal,GlobalOrdinal,Node>(builder,stratName);
-  }
+template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+MUELU_DEPRECATED void enableMueLuRefMaxwell(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLuRefMaxwell") {
+  enableMueLuRefMaxwell<double, LocalOrdinal, GlobalOrdinal, Node>(builder, stratName);
+}
 
-  template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
-  void enableMueLuMaxwell1(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLuMaxwell1")
-  {
+template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
+void enableMueLuMaxwell1(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLuMaxwell1") {
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
+  const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
-                               "Stratimikos::enableMueLuRefMaxwell cannot add \"" + stratName +"\" because it is already included in builder!");
+  TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
+                             "Stratimikos::enableMueLuRefMaxwell cannot add \"" + stratName + "\" because it is already included in builder!");
 
-    typedef Thyra::PreconditionerFactoryBase<Scalar>                                     Base;
-    typedef Thyra::MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
+  typedef Thyra::PreconditionerFactoryBase<Scalar> Base;
+  typedef Thyra::MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
 
-    builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
+  builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
 #endif
-  }
+}
 
-  template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-  MUELU_DEPRECATED void enableMueLuMaxwell1(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLuMaxwell1")
-  {
-    enableMueLuMaxwell1<double,LocalOrdinal,GlobalOrdinal,Node>(builder,stratName);
-  }
+template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+MUELU_DEPRECATED void enableMueLuMaxwell1(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLuMaxwell1") {
+  enableMueLuMaxwell1<double, LocalOrdinal, GlobalOrdinal, Node>(builder, stratName);
+}
 
 #if defined(HAVE_MUELU_EXPERIMENTAL) && defined(HAVE_MUELU_TEKO)
 #if 0
@@ -139,26 +133,25 @@ namespace Stratimikos {
   void enableMueLuTpetraQ2Q1(DefaultLinearSolverBuilder &builder, const std::string &stratName = "MueLu");
 #endif
 
-  template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
-  void enableMueLuTpetraQ2Q1(LinearSolverBuilder<Scalar>& builder, const std::string &stratName = "MueLu") {
-    const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
+template <typename Scalar = MueLu::DefaultScalar, typename LocalOrdinal = MueLu::DefaultLocalOrdinal, typename GlobalOrdinal = MueLu::DefaultGlobalOrdinal, typename Node = MueLu::DefaultNode>
+void enableMueLuTpetraQ2Q1(LinearSolverBuilder<Scalar>& builder, const std::string& stratName = "MueLu") {
+  const Teuchos::RCP<const Teuchos::ParameterList> precValidParams = Teuchos::sublist(builder.getValidParameters(), "Preconditioner Types");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
-                               "Stratimikos::enableMueLuTpetraQ2Q1 cannot add \"" + stratName +"\" because it is already included in builder!");
+  TEUCHOS_TEST_FOR_EXCEPTION(precValidParams->isParameter(stratName), std::logic_error,
+                             "Stratimikos::enableMueLuTpetraQ2Q1 cannot add \"" + stratName + "\" because it is already included in builder!");
 
-    typedef Thyra::PreconditionerFactoryBase<Scalar>                                              Base;
-    typedef Thyra::MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal,Node> Impl;
+  typedef Thyra::PreconditionerFactoryBase<Scalar> Base;
+  typedef Thyra::MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> Impl;
 
-    builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
-  }
+  builder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, Impl>(), stratName);
+}
 
-  template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-  MUELU_DEPRECATED void enableMueLuTpetraQ2Q1(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLu")
-  {
-    enableMueLuTpetraQ2Q1<double,LocalOrdinal,GlobalOrdinal,Node>(builder,stratName);
-  }
+template <typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+MUELU_DEPRECATED void enableMueLuTpetraQ2Q1(LinearSolverBuilder<double>& builder, const std::string& stratName = "MueLu") {
+  enableMueLuTpetraQ2Q1<double, LocalOrdinal, GlobalOrdinal, Node>(builder, stratName);
+}
 #endif
 
-} // namespace Stratimikos
+}  // namespace Stratimikos
 
 #endif
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_decl.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_decl.hpp
index 7ba561cd8c68..139244813cbc 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_decl.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_decl.hpp
@@ -57,77 +57,71 @@
 #include "Thyra_XpetraLinearOp.hpp"
 #include <Thyra_MueLuPreconditionerFactory.hpp>
 
-
 namespace Thyra {
 
-  /** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
-      @ingroup MueLuAdapters
-      Add support for MueLu's Maxwell1 preconditioner in Thyra.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class MueLuMaxwell1PreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
-  public:
-
-    /** @name Constructors/initializers/accessors */
-    //@{
-
-    /** \brief . */
-    MueLuMaxwell1PreconditionerFactory();
-    //@}
-
-    /** @name Overridden from PreconditionerFactoryBase */
-    //@{
-
-    /** \brief . */
-    bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
-    /** \brief . */
-    Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
-    /** \brief . */
-    void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
-                        PreconditionerBase<Scalar>* prec,
-                        const ESupportSolveUse supportSolveUse
-                       ) const;
-    /** \brief . */
-    void uninitializePrec(PreconditionerBase<Scalar>* prec,
-                          Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
-                          ESupportSolveUse* supportSolveUse
-                         ) const;
-
-    //@}
-
-    /** @name Overridden from Teuchos::ParameterListAcceptor */
-    //@{
-
-    /** \brief . */
-    void                                          setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          unsetParameterList();
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          getNonconstParameterList();
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getParameterList() const;
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getValidParameters() const;
-    //@}
-
-    /** \name Public functions overridden from Describable. */
-    //@{
-
-    /** \brief . */
-    std::string description() const;
-
-    // ToDo: Add an override of describe(...) to give more detail!
-
-    //@}
-
-  private:
-
-    Teuchos::RCP<Teuchos::ParameterList> paramList_;
-
-  };
-
-} // namespace Thyra
-
-#endif // #ifdef HAVE_MUELU_STRATIMIKOS
-
-#endif // THYRA_MUELU_MAXWELL1_PRECONDITIONER_FACTORY_DECL_HPP
+/** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
+    @ingroup MueLuAdapters
+    Add support for MueLu's Maxwell1 preconditioner in Thyra.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class MueLuMaxwell1PreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
+ public:
+  /** @name Constructors/initializers/accessors */
+  //@{
+
+  /** \brief . */
+  MueLuMaxwell1PreconditionerFactory();
+  //@}
+
+  /** @name Overridden from PreconditionerFactoryBase */
+  //@{
+
+  /** \brief . */
+  bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
+  /** \brief . */
+  Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
+  /** \brief . */
+  void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
+                      PreconditionerBase<Scalar>* prec,
+                      const ESupportSolveUse supportSolveUse) const;
+  /** \brief . */
+  void uninitializePrec(PreconditionerBase<Scalar>* prec,
+                        Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
+                        ESupportSolveUse* supportSolveUse) const;
+
+  //@}
+
+  /** @name Overridden from Teuchos::ParameterListAcceptor */
+  //@{
+
+  /** \brief . */
+  void setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> unsetParameterList();
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> getNonconstParameterList();
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getParameterList() const;
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getValidParameters() const;
+  //@}
+
+  /** \name Public functions overridden from Describable. */
+  //@{
+
+  /** \brief . */
+  std::string description() const;
+
+  // ToDo: Add an override of describe(...) to give more detail!
+
+  //@}
+
+ private:
+  Teuchos::RCP<Teuchos::ParameterList> paramList_;
+};
+
+}  // namespace Thyra
+
+#endif  // #ifdef HAVE_MUELU_STRATIMIKOS
+
+#endif  // THYRA_MUELU_MAXWELL1_PRECONDITIONER_FACTORY_DECL_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_def.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_def.hpp
index 2fd17754fff1..1476bfc3e5d5 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_def.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuMaxwell1PreconditionerFactory_def.hpp
@@ -55,276 +55,268 @@
 #include <MueLu_Maxwell1.hpp>
 #include <Xpetra_TpetraHalfPrecisionOperator.hpp>
 
-
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
 
 // This is not as general as possible, but should be good enough for most builds.
-#if((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
-# define MUELU_CAN_USE_MIXED_PRECISION
+#if ((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
+#define MUELU_CAN_USE_MIXED_PRECISION
 #endif
 
 namespace Thyra {
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::ParameterList;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcp_const_cast;
+using Teuchos::ParameterList;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_const_cast;
+using Teuchos::rcp_dynamic_cast;
 
-  // Constructors/initializers/accessors
+// Constructors/initializers/accessors
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::MueLuMaxwell1PreconditionerFactory() :
-      paramList_(rcp(new ParameterList()))
-  {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MueLuMaxwell1PreconditionerFactory()
+  : paramList_(rcp(new ParameterList())) {}
 
-  // Overridden from PreconditionerFactoryBase
+// Overridden from PreconditionerFactoryBase
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
-    const RCP<const LinearOpBase<Scalar> > fwdOp = fwdOpSrc.getOp();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
+  const RCP<const LinearOpBase<Scalar>> fwdOp = fwdOpSrc.getOp();
 
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isTpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isTpetra(fwdOp)) return true;
 
 #ifdef HAVE_MUELU_EPETRA
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isEpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isEpetra(fwdOp)) return true;
 #endif
 
-    return false;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<PreconditionerBase<Scalar> > MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createPrec() const {
-    return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initializePrec(const RCP<const LinearOpSourceBase<Scalar> >& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
-
-    // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
-    typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>      XpOp;
-    typedef Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>       XpThyUtils;
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMat;
-    typedef Thyra::LinearOpBase<Scalar>                                      ThyLinOpBase;
-    typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<PreconditionerBase<Scalar>> MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createPrec() const {
+  return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initializePrec(const RCP<const LinearOpSourceBase<Scalar>>& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
+  // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpOp;
+  typedef Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpThyUtils;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMat;
+  typedef Thyra::LinearOpBase<Scalar> ThyLinOpBase;
+  typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-    typedef Xpetra::TpetraHalfPrecisionOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> XpHalfPrecOp;
-    typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMV;
-    typedef typename XpHalfPrecOp::HalfScalar                                     HalfScalar;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType                 Magnitude;
-    typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision              HalfMagnitude;
-    typedef Xpetra::MultiVector<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>       XphMV;
-    typedef Xpetra::MultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node >       XpmMV;
-    typedef Xpetra::MultiVector<HalfMagnitude,LocalOrdinal,GlobalOrdinal,Node >   XphmMV;
-    typedef Xpetra::Matrix<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>            XphMat;
+  typedef Xpetra::TpetraHalfPrecisionOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpHalfPrecOp;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMV;
+  typedef typename XpHalfPrecOp::HalfScalar HalfScalar;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision HalfMagnitude;
+  typedef Xpetra::MultiVector<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMV;
+  typedef Xpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> XpmMV;
+  typedef Xpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> XphmMV;
+  typedef Xpetra::Matrix<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMat;
 #endif
-    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLuMaxwell1::initializePrec")));
-
-    // Check precondition
-    TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
-    TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
-    TEUCHOS_ASSERT(prec);
-
-    // Create a copy, as we may remove some things from the list
-    ParameterList paramList = *paramList_;
-
-    // Retrieve wrapped concrete Xpetra matrix from FwdOp
-    const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
-
-    // Check whether it is Epetra/Tpetra
-    bool bIsEpetra  = XpThyUtils::isEpetra(fwdOp);
-    bool bIsTpetra  = XpThyUtils::isTpetra(fwdOp);
-    TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true  && bIsTpetra == true));
-
-    // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
-    // MueLu needs a non-const object as input
-    RCP<XpMat> A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    // extract preconditioner operator
-    RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
-    thyra_precOp = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar> >(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
-
-    // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
-    // rebuild preconditioner if startingOver == true
-    // reuse preconditioner if startingOver == false
-    const bool startingOver = (thyra_precOp.is_null() || !paramList.isParameter("Maxwell1: enable reuse") || !paramList.get<bool>("Maxwell1: enable reuse"));
-    const bool useHalfPrecision = paramList.get<bool>("half precision", false) && bIsTpetra;
-
-    RCP<XpOp> xpPrecOp;
-    if (startingOver == true) {
-
-      // Convert to Xpetra
-      std::list<std::string> convertXpetra = {"Coordinates", "Nullspace", "Kn", "D0"};
-      for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
-        Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(paramList,*it);
-
-      std::list<std::string> sublists = {"maxwell1: 11list", "maxwell1: 22list"};
-      for (auto itSublist = sublists.begin(); itSublist != sublists.end(); ++itSublist)
+  Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLuMaxwell1::initializePrec")));
+
+  // Check precondition
+  TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
+  TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
+  TEUCHOS_ASSERT(prec);
+
+  // Create a copy, as we may remove some things from the list
+  ParameterList paramList = *paramList_;
+
+  // Retrieve wrapped concrete Xpetra matrix from FwdOp
+  const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
+
+  // Check whether it is Epetra/Tpetra
+  bool bIsEpetra = XpThyUtils::isEpetra(fwdOp);
+  bool bIsTpetra = XpThyUtils::isTpetra(fwdOp);
+  TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true && bIsTpetra == true));
+
+  // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
+  // MueLu needs a non-const object as input
+  RCP<XpMat> A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
+
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar>> defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
+
+  // extract preconditioner operator
+  RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
+  thyra_precOp                   = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar>>(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
+
+  // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
+  // rebuild preconditioner if startingOver == true
+  // reuse preconditioner if startingOver == false
+  const bool startingOver     = (thyra_precOp.is_null() || !paramList.isParameter("Maxwell1: enable reuse") || !paramList.get<bool>("Maxwell1: enable reuse"));
+  const bool useHalfPrecision = paramList.get<bool>("half precision", false) && bIsTpetra;
+
+  RCP<XpOp> xpPrecOp;
+  if (startingOver == true) {
+    // Convert to Xpetra
+    std::list<std::string> convertXpetra = {"Coordinates", "Nullspace", "Kn", "D0"};
+    for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
+      Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(paramList, *it);
+
+    std::list<std::string> sublists = {"maxwell1: 11list", "maxwell1: 22list"};
+    for (auto itSublist = sublists.begin(); itSublist != sublists.end(); ++itSublist)
       if (paramList.isSublist(*itSublist)) {
         ParameterList& sublist = paramList.sublist(*itSublist);
-        for (int lvlNo=0; lvlNo < 10; ++lvlNo) {
+        for (int lvlNo = 0; lvlNo < 10; ++lvlNo) {
           if (sublist.isSublist("level " + std::to_string(lvlNo) + " user data")) {
             ParameterList& lvlList = sublist.sublist("level " + std::to_string(lvlNo) + " user data");
             std::list<std::string> convertKeys;
             for (auto it = lvlList.begin(); it != lvlList.end(); ++it)
               convertKeys.push_back(lvlList.name(it));
             for (auto it = convertKeys.begin(); it != convertKeys.end(); ++it)
-              Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(lvlList,*it);
+              Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(lvlList, *it);
           }
         }
       }
 
-      ParameterList& sublist = paramList.sublist("maxwell1: 11list");
-      if (sublist.isParameter("D0")) {
-        Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(sublist,"D0");
-      }
+    ParameterList& sublist = paramList.sublist("maxwell1: 11list");
+    if (sublist.isParameter("D0")) {
+      Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(sublist, "D0");
+    }
 
-      paramList.set<bool>("Maxwell1: use as preconditioner", true);
-      if (useHalfPrecision) {
+    paramList.set<bool>("Maxwell1: use as preconditioner", true);
+    if (useHalfPrecision) {
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
 
-        // convert to half precision
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-        if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
-          RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
-          paramList.remove("Coordinates");
-          RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
-          paramList.set("Coordinates",halfCoords);
-        }
-        if (paramList.isType<RCP<XpMV> >("Nullspace")) {
-          RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
-          paramList.remove("Nullspace");
-          RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
-          paramList.set("Nullspace",halfNullspace);
-        }
-        std::list<std::string> convertMat = {"Kn", "D0"};
-        for (auto it = convertMat.begin(); it != convertMat.end(); ++it) {
-          if (paramList.isType<RCP<XpMat> >(*it)) {
-            RCP<XpMat> M = paramList.get<RCP<XpMat> >(*it);
-            paramList.remove(*it);
-            RCP<XphMat> halfM = Xpetra::convertToHalfPrecision(M);
-            paramList.set(*it,halfM);
-          }
+      // convert to half precision
+      RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
+      if (paramList.isType<RCP<XpmMV>>("Coordinates")) {
+        RCP<XpmMV> coords = paramList.get<RCP<XpmMV>>("Coordinates");
+        paramList.remove("Coordinates");
+        RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
+        paramList.set("Coordinates", halfCoords);
+      }
+      if (paramList.isType<RCP<XpMV>>("Nullspace")) {
+        RCP<XpMV> nullspace = paramList.get<RCP<XpMV>>("Nullspace");
+        paramList.remove("Nullspace");
+        RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
+        paramList.set("Nullspace", halfNullspace);
+      }
+      std::list<std::string> convertMat = {"Kn", "D0"};
+      for (auto it = convertMat.begin(); it != convertMat.end(); ++it) {
+        if (paramList.isType<RCP<XpMat>>(*it)) {
+          RCP<XpMat> M = paramList.get<RCP<XpMat>>(*it);
+          paramList.remove(*it);
+          RCP<XphMat> halfM = Xpetra::convertToHalfPrecision(M);
+          paramList.set(*it, halfM);
         }
+      }
 
-        // build a new half-precision MueLu Maxwell1 preconditioner
-        RCP<MueLu::Maxwell1<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > halfPrec = rcp(new MueLu::Maxwell1<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>(halfA, paramList, true));
-        xpPrecOp = rcp(new XpHalfPrecOp(halfPrec));
+      // build a new half-precision MueLu Maxwell1 preconditioner
+      RCP<MueLu::Maxwell1<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>> halfPrec = rcp(new MueLu::Maxwell1<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>(halfA, paramList, true));
+      xpPrecOp                                                                     = rcp(new XpHalfPrecOp(halfPrec));
 #else
-        TEUCHOS_TEST_FOR_EXCEPT(true);
+      TEUCHOS_TEST_FOR_EXCEPT(true);
 #endif
-      } else
-        {
-          // build a new MueLu Maxwell1 preconditioner
-          RCP<MueLu::Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp(new MueLu::Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A, paramList, true));
-          xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-        }
     } else {
-      // reuse old MueLu preconditioner stored in MueLu Xpetra operator and put in new matrix
+      // build a new MueLu Maxwell1 preconditioner
+      RCP<MueLu::Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp(new MueLu::Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, paramList, true));
+      xpPrecOp                                                                       = rcp_dynamic_cast<XpOp>(preconditioner);
+    }
+  } else {
+    // reuse old MueLu preconditioner stored in MueLu Xpetra operator and put in new matrix
 
-      RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
-      RCP<XpOp>    xpOp    = thyXpOp->getXpetraOperator();
+    RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
+    RCP<XpOp> xpOp       = thyXpOp->getXpetraOperator();
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-      RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpOp);
-      if (!xpHalfPrecOp.is_null()) {
-        RCP<MueLu::Maxwell1<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp_dynamic_cast<MueLu::Maxwell1<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>>(xpHalfPrecOp->GetHalfPrecisionOperator(), true);
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-        preconditioner->resetMatrix(halfA);
-        xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-      } else
+    RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpOp);
+    if (!xpHalfPrecOp.is_null()) {
+      RCP<MueLu::Maxwell1<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp_dynamic_cast<MueLu::Maxwell1<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>>(xpHalfPrecOp->GetHalfPrecisionOperator(), true);
+      RCP<XphMat> halfA                                                                  = Xpetra::convertToHalfPrecision(A);
+      preconditioner->resetMatrix(halfA);
+      xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
+    } else
 #endif
-        {
-          RCP<MueLu::Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp_dynamic_cast<MueLu::Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>>(xpOp, true);
-          preconditioner->resetMatrix(A);
-          xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-        }
+    {
+      RCP<MueLu::Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp_dynamic_cast<MueLu::Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(xpOp, true);
+      preconditioner->resetMatrix(A);
+      xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
     }
-
-    // wrap preconditioner in thyraPrecOp
-    RCP<const VectorSpaceBase<Scalar> > thyraRangeSpace  = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getRangeMap());
-    RCP<const VectorSpaceBase<Scalar> > thyraDomainSpace = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getDomainMap());
-
-    RCP<ThyLinOpBase > thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
-
-    defaultPrec->initializeUnspecified(thyraPrecOp);
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
-    TEUCHOS_ASSERT(prec);
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    if (fwdOp) {
-      // TODO: Implement properly instead of returning default value
-      *fwdOp = Teuchos::null;
-    }
-
-    if (supportSolveUse) {
-      // TODO: Implement properly instead of returning default value
-      *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
-    }
-
-    defaultPrec->uninitialize();
   }
 
+  // wrap preconditioner in thyraPrecOp
+  RCP<const VectorSpaceBase<Scalar>> thyraRangeSpace  = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getRangeMap());
+  RCP<const VectorSpaceBase<Scalar>> thyraDomainSpace = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getDomainMap());
 
-  // Overridden from ParameterListAcceptor
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameterList(RCP<ParameterList> const& paramList) {
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
-    paramList_ = paramList;
-  }
+  RCP<ThyLinOpBase> thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getNonconstParameterList() {
-    return paramList_;
-  }
+  defaultPrec->initializeUnspecified(thyraPrecOp);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::unsetParameterList() {
-    RCP<ParameterList> savedParamList = paramList_;
-    paramList_ = Teuchos::null;
-    return savedParamList;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar>>* fwdOp, ESupportSolveUse* supportSolveUse) const {
+  TEUCHOS_ASSERT(prec);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getParameterList() const {
-    return paramList_;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getValidParameters() const {
-    static RCP<const ParameterList> validPL;
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar>> defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
 
-    if (Teuchos::is_null(validPL))
-      validPL = rcp(new ParameterList());
-
-    return validPL;
+  if (fwdOp) {
+    // TODO: Implement properly instead of returning default value
+    *fwdOp = Teuchos::null;
   }
 
-  // Public functions overridden from Teuchos::Describable
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MueLuMaxwell1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::description() const {
-    return "Thyra::MueLuMaxwell1PreconditionerFactory";
+  if (supportSolveUse) {
+    // TODO: Implement properly instead of returning default value
+    *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
   }
-} // namespace Thyra
-
-#endif // HAVE_MUELU_STRATIMIKOS
 
-#endif // ifdef THYRA_MUELU_MAXWELL1_PRECONDITIONER_FACTORY_DEF_HPP
+  defaultPrec->uninitialize();
+}
+
+// Overridden from ParameterListAcceptor
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameterList(RCP<ParameterList> const& paramList) {
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
+  paramList_ = paramList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNonconstParameterList() {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::unsetParameterList() {
+  RCP<ParameterList> savedParamList = paramList_;
+  paramList_                        = Teuchos::null;
+  return savedParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getParameterList() const {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getValidParameters() const {
+  static RCP<const ParameterList> validPL;
+
+  if (Teuchos::is_null(validPL))
+    validPL = rcp(new ParameterList());
+
+  return validPL;
+}
+
+// Public functions overridden from Teuchos::Describable
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MueLuMaxwell1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  return "Thyra::MueLuMaxwell1PreconditionerFactory";
+}
+}  // namespace Thyra
+
+#endif  // HAVE_MUELU_STRATIMIKOS
+
+#endif  // ifdef THYRA_MUELU_MAXWELL1_PRECONDITIONER_FACTORY_DEF_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_decl.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_decl.hpp
index 31b68376f2fa..4bc13452fc68 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_decl.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_decl.hpp
@@ -78,7 +78,7 @@
 #include <MueLu_Utilities.hpp>
 #include <MueLu_ParameterListInterpreter.hpp>
 #include <MueLu_MasterList.hpp>
-#include <MueLu_XpetraOperator_decl.hpp> // todo fix me
+#include <MueLu_XpetraOperator_decl.hpp>  // todo fix me
 #include <MueLu_CreateXpetraPreconditioner.hpp>
 #include <MueLu_TpetraOperator.hpp>
 #include <Xpetra_TpetraHalfPrecisionOperator.hpp>
@@ -94,92 +94,87 @@
 
 namespace Thyra {
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
+using Teuchos::RCP;
+using Teuchos::rcp;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  struct Converters {
-    static bool replaceWithXpetra(ParameterList& paramList, std::string parameterName);
-  };
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct Converters {
+  static bool replaceWithXpetra(ParameterList& paramList, std::string parameterName);
+};
 
 #ifdef HAVE_MUELU_EPETRA
-  template <class GlobalOrdinal>
-  struct Converters<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
-    static bool replaceWithXpetra(ParameterList& paramList, std::string parameterName);
-  };
+template <class GlobalOrdinal>
+struct Converters<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
+  static bool replaceWithXpetra(ParameterList& paramList, std::string parameterName);
+};
 #endif
 
-  /** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
-      @ingroup MueLuAdapters
-      Add support for MueLu preconditioners in Thyra. This class provides an interface both
-      for Epetra and Tpetra.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class MueLuPreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
-  public:
-
-    /** @name Constructors/initializers/accessors */
-    //@{
-
-    /** \brief . */
-    MueLuPreconditionerFactory();
-    //@}
-
-    /** @name Overridden from PreconditionerFactoryBase */
-    //@{
-
-    /** \brief . */
-    bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
-    /** \brief . */
-    Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
-    /** \brief . */
-    void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
-                        PreconditionerBase<Scalar>* prec,
-                        const ESupportSolveUse supportSolveUse
-                       ) const;
-    /** \brief . */
-    void uninitializePrec(PreconditionerBase<Scalar>* prec,
-                          Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
-                          ESupportSolveUse* supportSolveUse
-                         ) const;
-
-    //@}
-
-    /** @name Overridden from Teuchos::ParameterListAcceptor */
-    //@{
-
-    /** \brief . */
-    void                                          setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          unsetParameterList();
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          getNonconstParameterList();
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getParameterList() const;
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getValidParameters() const;
-    //@}
-
-    /** \name Public functions overridden from Describable. */
-    //@{
-
-    /** \brief . */
-    std::string description() const;
-
-    // ToDo: Add an override of describe(...) to give more detail!
-
-    //@}
-
-  private:
-
-    //Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op, const Teuchos::ParameterList& paramList, Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > coords, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > nullspace) const;
-
-    Teuchos::RCP<Teuchos::ParameterList> paramList_;
-
-  };
-
-} // namespace Thyra
-
-#endif // #ifdef HAVE_MUELU_STRATIMIKOS
-
-#endif // THYRA_MUELU_PRECONDITIONER_FACTORY_DECL_HPP
+/** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
+    @ingroup MueLuAdapters
+    Add support for MueLu preconditioners in Thyra. This class provides an interface both
+    for Epetra and Tpetra.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class MueLuPreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
+ public:
+  /** @name Constructors/initializers/accessors */
+  //@{
+
+  /** \brief . */
+  MueLuPreconditionerFactory();
+  //@}
+
+  /** @name Overridden from PreconditionerFactoryBase */
+  //@{
+
+  /** \brief . */
+  bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
+  /** \brief . */
+  Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
+  /** \brief . */
+  void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
+                      PreconditionerBase<Scalar>* prec,
+                      const ESupportSolveUse supportSolveUse) const;
+  /** \brief . */
+  void uninitializePrec(PreconditionerBase<Scalar>* prec,
+                        Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
+                        ESupportSolveUse* supportSolveUse) const;
+
+  //@}
+
+  /** @name Overridden from Teuchos::ParameterListAcceptor */
+  //@{
+
+  /** \brief . */
+  void setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> unsetParameterList();
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> getNonconstParameterList();
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getParameterList() const;
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getValidParameters() const;
+  //@}
+
+  /** \name Public functions overridden from Describable. */
+  //@{
+
+  /** \brief . */
+  std::string description() const;
+
+  // ToDo: Add an override of describe(...) to give more detail!
+
+  //@}
+
+ private:
+  // Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op, const Teuchos::ParameterList& paramList, Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > coords, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > nullspace) const;
+
+  Teuchos::RCP<Teuchos::ParameterList> paramList_;
+};
+
+}  // namespace Thyra
+
+#endif  // #ifdef HAVE_MUELU_STRATIMIKOS
+
+#endif  // THYRA_MUELU_PRECONDITIONER_FACTORY_DECL_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_def.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_def.hpp
index 7b541dc6924c..0213059d9e89 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_def.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuPreconditionerFactory_def.hpp
@@ -52,695 +52,675 @@
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
 
 // This is not as general as possible, but should be good enough for most builds.
-#if((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
-# define MUELU_CAN_USE_MIXED_PRECISION
+#if ((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
+#define MUELU_CAN_USE_MIXED_PRECISION
 #endif
 
 namespace Thyra {
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::ParameterList;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcp_const_cast;
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(ParameterList& paramList, std::string parameterName) {
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType            Magnitude;
-    typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>      XpOp;
-    typedef Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>       XpThyUtils;
-    // typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>    XpCrsMatWrap;
-    // typedef Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XpCrsMat;
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMat;
-    typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>      XpMultVec;
-    typedef Xpetra::MultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>   XpMagMultVec;
-    typedef Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpVec;
-
-    typedef Thyra::LinearOpBase<Scalar>                                      ThyLinOpBase;
-    typedef Thyra::DiagonalLinearOpBase<Scalar>                              ThyDiagLinOpBase;
-    // typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
-    // typedef Thyra::SpmdVectorSpaceBase<Scalar>                               ThyVSBase;
-
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>         TpCrsMat;
-    typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>       tOp;
-    typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>         tV;
-    typedef Thyra::TpetraVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>       thyTpV;
-    typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>    tMV;
-    typedef Tpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> tMagMV;
+using Teuchos::ParameterList;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_const_cast;
+using Teuchos::rcp_dynamic_cast;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(ParameterList& paramList, std::string parameterName) {
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpOp;
+  typedef Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpThyUtils;
+  // typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>    XpCrsMatWrap;
+  // typedef Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XpCrsMat;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMat;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMultVec;
+  typedef Xpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> XpMagMultVec;
+  typedef Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpVec;
+
+  typedef Thyra::LinearOpBase<Scalar> ThyLinOpBase;
+  typedef Thyra::DiagonalLinearOpBase<Scalar> ThyDiagLinOpBase;
+  // typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
+  // typedef Thyra::SpmdVectorSpaceBase<Scalar>                               ThyVSBase;
+
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> TpCrsMat;
+  typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> tOp;
+  typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tV;
+  typedef Thyra::TpetraVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> thyTpV;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tMV;
+  typedef Tpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> tMagMV;
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-    typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision          HalfMagnitude;
-    typedef Tpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> tHalfMagMV;
+  typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision HalfMagnitude;
+  typedef Tpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> tHalfMagMV;
 #endif
 
-    if (paramList.isParameter(parameterName)) {
-      if (paramList.isType<RCP<XpMat> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMat> >(parameterName)) {
-        RCP<const XpMat> constM = paramList.get<RCP<const XpMat> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMat> M = rcp_const_cast<XpMat>(constM);
-        paramList.set<RCP<XpMat> >(parameterName, M);
-        return true;
-      }
-      else if (paramList.isType<RCP<XpMultVec> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMultVec> >(parameterName)) {
-        RCP<const XpMultVec> constX = paramList.get<RCP<const XpMultVec> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMultVec> X = rcp_const_cast<XpMultVec>(constX);
-        paramList.set<RCP<XpMultVec> >(parameterName, X);
-        return true;
-      }
-      else if (paramList.isType<RCP<XpMagMultVec> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMagMultVec> >(parameterName)) {
-        RCP<const XpMagMultVec> constX = paramList.get<RCP<const XpMagMultVec> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMagMultVec> X = rcp_const_cast<XpMagMultVec>(constX);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        return true;
-      }
-      else if (paramList.isType<RCP<TpCrsMat> >(parameterName)) {
-        RCP<TpCrsMat> tM = paramList.get<RCP<TpCrsMat> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMat> xM = MueLu::TpetraCrs_To_XpetraMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tM);
-        paramList.set<RCP<XpMat> >(parameterName, xM);
-        return true;
-      } else if (paramList.isType<RCP<tMV> >(parameterName)) {
-        RCP<tMV> tpetra_X = paramList.get<RCP<tMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      } else if (paramList.isType<RCP<tMagMV> >(parameterName)) {
-        RCP<tMagMV> tpetra_X = paramList.get<RCP<tMagMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      }
+  if (paramList.isParameter(parameterName)) {
+    if (paramList.isType<RCP<XpMat> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMat> >(parameterName)) {
+      RCP<const XpMat> constM = paramList.get<RCP<const XpMat> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMat> M = rcp_const_cast<XpMat>(constM);
+      paramList.set<RCP<XpMat> >(parameterName, M);
+      return true;
+    } else if (paramList.isType<RCP<XpMultVec> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMultVec> >(parameterName)) {
+      RCP<const XpMultVec> constX = paramList.get<RCP<const XpMultVec> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMultVec> X = rcp_const_cast<XpMultVec>(constX);
+      paramList.set<RCP<XpMultVec> >(parameterName, X);
+      return true;
+    } else if (paramList.isType<RCP<XpMagMultVec> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMagMultVec> >(parameterName)) {
+      RCP<const XpMagMultVec> constX = paramList.get<RCP<const XpMagMultVec> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMagMultVec> X = rcp_const_cast<XpMagMultVec>(constX);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      return true;
+    } else if (paramList.isType<RCP<TpCrsMat> >(parameterName)) {
+      RCP<TpCrsMat> tM = paramList.get<RCP<TpCrsMat> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMat> xM = MueLu::TpetraCrs_To_XpetraMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tM);
+      paramList.set<RCP<XpMat> >(parameterName, xM);
+      return true;
+    } else if (paramList.isType<RCP<tMV> >(parameterName)) {
+      RCP<tMV> tpetra_X = paramList.get<RCP<tMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    } else if (paramList.isType<RCP<tMagMV> >(parameterName)) {
+      RCP<tMagMV> tpetra_X = paramList.get<RCP<tMagMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    }
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-      else if (paramList.isType<RCP<tHalfMagMV> >(parameterName)) {
-        RCP<tHalfMagMV> tpetra_hX = paramList.get<RCP<tHalfMagMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<tMagMV> tpetra_X = rcp(new tMagMV(tpetra_hX->getMap(),tpetra_hX->getNumVectors()));
-        Tpetra::deep_copy(*tpetra_X,*tpetra_hX);
-        RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      }
+    else if (paramList.isType<RCP<tHalfMagMV> >(parameterName)) {
+      RCP<tHalfMagMV> tpetra_hX = paramList.get<RCP<tHalfMagMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<tMagMV> tpetra_X = rcp(new tMagMV(tpetra_hX->getMap(), tpetra_hX->getNumVectors()));
+      Tpetra::deep_copy(*tpetra_X, *tpetra_hX);
+      RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    }
 #endif
-      else if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName) ||
-               (paramList.isType<RCP<const ThyLinOpBase> >(parameterName) && !
-                rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName)).is_null())) {
-        RCP<const ThyDiagLinOpBase> thyM;
-        if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName))
-         thyM = paramList.get<RCP<const ThyDiagLinOpBase> >(parameterName);
-        else
-          thyM = rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName), true);
-        paramList.remove(parameterName);
-        RCP<const Thyra::VectorBase<Scalar> > diag = thyM->getDiag();
-
-        RCP<const XpVec> xpDiag;
-        if (!rcp_dynamic_cast<const thyTpV>(diag).is_null()) {
-          RCP<const tV> tDiag = Thyra::TpetraOperatorVectorExtraction<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getConstTpetraVector(diag);
-          if (!tDiag.is_null())
-            xpDiag = Xpetra::toXpetra(tDiag);
-        }
-        TEUCHOS_ASSERT(!xpDiag.is_null());
-        RCP<XpMat> M = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(xpDiag);
-        paramList.set<RCP<XpMat> >(parameterName, M);
-        return true;
+    else if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName) ||
+             (paramList.isType<RCP<const ThyLinOpBase> >(parameterName) && !rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName)).is_null())) {
+      RCP<const ThyDiagLinOpBase> thyM;
+      if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName))
+        thyM = paramList.get<RCP<const ThyDiagLinOpBase> >(parameterName);
+      else
+        thyM = rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName), true);
+      paramList.remove(parameterName);
+      RCP<const Thyra::VectorBase<Scalar> > diag = thyM->getDiag();
+
+      RCP<const XpVec> xpDiag;
+      if (!rcp_dynamic_cast<const thyTpV>(diag).is_null()) {
+        RCP<const tV> tDiag = Thyra::TpetraOperatorVectorExtraction<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getConstTpetraVector(diag);
+        if (!tDiag.is_null())
+          xpDiag = Xpetra::toXpetra(tDiag);
       }
-      else if (paramList.isType<RCP<const ThyLinOpBase> >(parameterName)) {
-        RCP<const ThyLinOpBase> thyM = paramList.get<RCP<const ThyLinOpBase> >(parameterName);
-        paramList.remove(parameterName);
-        try {
-          RCP<XpMat> M = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
-          paramList.set<RCP<XpMat> >(parameterName, M);
-        } catch (std::exception& e) {
-          RCP<XpOp> M = XpThyUtils::toXpetraOperator(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
-          RCP<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(M, true);
-          RCP<tOp> tO = tpOp->getOperator();
-          RCP<tV> diag;
-          if (tO->hasDiagonal()) {
-            diag = rcp(new tV(tO->getRangeMap()));
-            tO->getLocalDiagCopy(*diag);
-          }
-          auto fTpRow = rcp(new MueLu::TpetraOperatorAsRowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tO, diag));
-          RCP<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpFOp = rcp(new Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> (fTpRow));
-          auto op = rcp_dynamic_cast<XpOp>(tpFOp);
-          paramList.set<RCP<XpOp> >(parameterName, op);
+      TEUCHOS_ASSERT(!xpDiag.is_null());
+      RCP<XpMat> M = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(xpDiag);
+      paramList.set<RCP<XpMat> >(parameterName, M);
+      return true;
+    } else if (paramList.isType<RCP<const ThyLinOpBase> >(parameterName)) {
+      RCP<const ThyLinOpBase> thyM = paramList.get<RCP<const ThyLinOpBase> >(parameterName);
+      paramList.remove(parameterName);
+      try {
+        RCP<XpMat> M = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
+        paramList.set<RCP<XpMat> >(parameterName, M);
+      } catch (std::exception& e) {
+        RCP<XpOp> M                                                                  = XpThyUtils::toXpetraOperator(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
+        RCP<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(M, true);
+        RCP<tOp> tO                                                                  = tpOp->getOperator();
+        RCP<tV> diag;
+        if (tO->hasDiagonal()) {
+          diag = rcp(new tV(tO->getRangeMap()));
+          tO->getLocalDiagCopy(*diag);
         }
-        return true;
-      }
-      else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Parameter " << parameterName << " has wrong type.");
-        return false;
+        auto fTpRow                                                                   = rcp(new MueLu::TpetraOperatorAsRowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tO, diag));
+        RCP<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpFOp = rcp(new Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>(fTpRow));
+        auto op                                                                       = rcp_dynamic_cast<XpOp>(tpFOp);
+        paramList.set<RCP<XpOp> >(parameterName, op);
       }
-    } else
+      return true;
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Parameter " << parameterName << " has wrong type.");
       return false;
-  }
-
+    }
+  } else
+    return false;
+}
 
 #ifdef HAVE_MUELU_EPETRA
-  template <class GlobalOrdinal>
-  bool Converters<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>::replaceWithXpetra(ParameterList& paramList, std::string parameterName) {
-    typedef double Scalar;
-    typedef int LocalOrdinal;
-    typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode Node;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType            Magnitude;
-    typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>      XpOp;
-    typedef Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>       XpThyUtils;
-    typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>    XpCrsMatWrap;
-    typedef Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XpCrsMat;
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMat;
-    typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>      XpMultVec;
-    typedef Xpetra::MultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>   XpMagMultVec;
-    typedef Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpVec;
-
-    typedef Thyra::LinearOpBase<Scalar>                                      ThyLinOpBase;
-    typedef Thyra::DiagonalLinearOpBase<Scalar>                              ThyDiagLinOpBase;
-    typedef Thyra::SpmdVectorSpaceBase<Scalar>                               ThyVSBase;
-
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>         TpCrsMat;
-    typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>       tOp;
-    typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>         tV;
-    typedef Thyra::TpetraVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>       thyTpV;
-    typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>    tMV;
-    typedef Tpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> tMagMV;
+template <class GlobalOrdinal>
+bool Converters<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>::replaceWithXpetra(ParameterList& paramList, std::string parameterName) {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode Node;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpOp;
+  typedef Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpThyUtils;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpCrsMatWrap;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpCrsMat;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMat;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMultVec;
+  typedef Xpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> XpMagMultVec;
+  typedef Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpVec;
+
+  typedef Thyra::LinearOpBase<Scalar> ThyLinOpBase;
+  typedef Thyra::DiagonalLinearOpBase<Scalar> ThyDiagLinOpBase;
+  typedef Thyra::SpmdVectorSpaceBase<Scalar> ThyVSBase;
+
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> TpCrsMat;
+  typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> tOp;
+  typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tV;
+  typedef Thyra::TpetraVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> thyTpV;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> tMV;
+  typedef Tpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> tMagMV;
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-    typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision          HalfMagnitude;
-    typedef Tpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> tHalfMagMV;
+  typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision HalfMagnitude;
+  typedef Tpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> tHalfMagMV;
 #endif
 #if defined(HAVE_MUELU_EPETRA)
-    typedef Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>                  XpEpCrsMat;
+  typedef Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node> XpEpCrsMat;
 #endif
 
-    if (paramList.isParameter(parameterName)) {
-      if (paramList.isType<RCP<XpMat> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMat> >(parameterName)) {
-        RCP<const XpMat> constM = paramList.get<RCP<const XpMat> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMat> M = rcp_const_cast<XpMat>(constM);
-        paramList.set<RCP<XpMat> >(parameterName, M);
-        return true;
-      }
-      else if (paramList.isType<RCP<XpMultVec> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMultVec> >(parameterName)) {
-        RCP<const XpMultVec> constX = paramList.get<RCP<const XpMultVec> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMultVec> X = rcp_const_cast<XpMultVec>(constX);
-        paramList.set<RCP<XpMultVec> >(parameterName, X);
-        return true;
-      }
-      else if (paramList.isType<RCP<XpMagMultVec> >(parameterName))
-        return true;
-      else if (paramList.isType<RCP<const XpMagMultVec> >(parameterName)) {
-        RCP<const XpMagMultVec> constX = paramList.get<RCP<const XpMagMultVec> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMagMultVec> X = rcp_const_cast<XpMagMultVec>(constX);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        return true;
-      }
-      else if (paramList.isType<RCP<TpCrsMat> >(parameterName)) {
-        RCP<TpCrsMat> tM = paramList.get<RCP<TpCrsMat> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMat> xM = MueLu::TpetraCrs_To_XpetraMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tM);
-        paramList.set<RCP<XpMat> >(parameterName, xM);
-        return true;
-      } else if (paramList.isType<RCP<tMV> >(parameterName)) {
-        RCP<tMV> tpetra_X = paramList.get<RCP<tMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      } else if (paramList.isType<RCP<tMagMV> >(parameterName)) {
-        RCP<tMagMV> tpetra_X = paramList.get<RCP<tMagMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      }
+  if (paramList.isParameter(parameterName)) {
+    if (paramList.isType<RCP<XpMat> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMat> >(parameterName)) {
+      RCP<const XpMat> constM = paramList.get<RCP<const XpMat> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMat> M = rcp_const_cast<XpMat>(constM);
+      paramList.set<RCP<XpMat> >(parameterName, M);
+      return true;
+    } else if (paramList.isType<RCP<XpMultVec> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMultVec> >(parameterName)) {
+      RCP<const XpMultVec> constX = paramList.get<RCP<const XpMultVec> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMultVec> X = rcp_const_cast<XpMultVec>(constX);
+      paramList.set<RCP<XpMultVec> >(parameterName, X);
+      return true;
+    } else if (paramList.isType<RCP<XpMagMultVec> >(parameterName))
+      return true;
+    else if (paramList.isType<RCP<const XpMagMultVec> >(parameterName)) {
+      RCP<const XpMagMultVec> constX = paramList.get<RCP<const XpMagMultVec> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMagMultVec> X = rcp_const_cast<XpMagMultVec>(constX);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      return true;
+    } else if (paramList.isType<RCP<TpCrsMat> >(parameterName)) {
+      RCP<TpCrsMat> tM = paramList.get<RCP<TpCrsMat> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMat> xM = MueLu::TpetraCrs_To_XpetraMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tM);
+      paramList.set<RCP<XpMat> >(parameterName, xM);
+      return true;
+    } else if (paramList.isType<RCP<tMV> >(parameterName)) {
+      RCP<tMV> tpetra_X = paramList.get<RCP<tMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    } else if (paramList.isType<RCP<tMagMV> >(parameterName)) {
+      RCP<tMagMV> tpetra_X = paramList.get<RCP<tMagMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    }
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-      else if (paramList.isType<RCP<tHalfMagMV> >(parameterName)) {
-        RCP<tHalfMagMV> tpetra_hX = paramList.get<RCP<tHalfMagMV> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<tMagMV> tpetra_X = rcp(new tMagMV(tpetra_hX->getMap(),tpetra_hX->getNumVectors()));
-        Tpetra::deep_copy(*tpetra_X,*tpetra_hX);
-        RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node>(tpetra_X);
-        paramList.set<RCP<XpMagMultVec> >(parameterName, X);
-        TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
-        return true;
-      }
+    else if (paramList.isType<RCP<tHalfMagMV> >(parameterName)) {
+      RCP<tHalfMagMV> tpetra_hX = paramList.get<RCP<tHalfMagMV> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<tMagMV> tpetra_X = rcp(new tMagMV(tpetra_hX->getMap(), tpetra_hX->getNumVectors()));
+      Tpetra::deep_copy(*tpetra_X, *tpetra_hX);
+      RCP<XpMagMultVec> X = MueLu::TpetraMultiVector_To_XpetraMultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node>(tpetra_X);
+      paramList.set<RCP<XpMagMultVec> >(parameterName, X);
+      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(X));
+      return true;
+    }
 #endif
 #ifdef HAVE_MUELU_EPETRA
-      else if (paramList.isType<RCP<Epetra_CrsMatrix> >(parameterName)) {
-        RCP<Epetra_CrsMatrix> eM = paramList.get<RCP<Epetra_CrsMatrix> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<XpEpCrsMat> xeM = rcp(new XpEpCrsMat(eM));
-        RCP<XpCrsMat> xCrsM = rcp_dynamic_cast<XpCrsMat>(xeM, true);
-        RCP<XpCrsMatWrap> xwM = rcp(new XpCrsMatWrap(xCrsM));
-        RCP<XpMat> xM = rcp_dynamic_cast<XpMat>(xwM);
-        paramList.set<RCP<XpMat> >(parameterName, xM);
-        return true;
-      } else if (paramList.isType<RCP<Epetra_MultiVector> >(parameterName)) {
-        RCP<Epetra_MultiVector> epetra_X = Teuchos::null;
-        epetra_X = paramList.get<RCP<Epetra_MultiVector> >(parameterName);
-        paramList.remove(parameterName);
-        RCP<Xpetra::EpetraMultiVectorT<int,Node> > xpEpX = rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epetra_X));
-        RCP<Xpetra::MultiVector<Scalar,int,int,Node> > xpEpXMult = rcp_dynamic_cast<Xpetra::MultiVector<Scalar,int,int,Node> >(xpEpX, true);
-        RCP<XpMultVec> X = rcp_dynamic_cast<XpMultVec>(xpEpXMult, true);
-        paramList.set<RCP<XpMultVec> >(parameterName, X);
-        return true;
-      }
+    else if (paramList.isType<RCP<Epetra_CrsMatrix> >(parameterName)) {
+      RCP<Epetra_CrsMatrix> eM = paramList.get<RCP<Epetra_CrsMatrix> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<XpEpCrsMat> xeM   = rcp(new XpEpCrsMat(eM));
+      RCP<XpCrsMat> xCrsM   = rcp_dynamic_cast<XpCrsMat>(xeM, true);
+      RCP<XpCrsMatWrap> xwM = rcp(new XpCrsMatWrap(xCrsM));
+      RCP<XpMat> xM         = rcp_dynamic_cast<XpMat>(xwM);
+      paramList.set<RCP<XpMat> >(parameterName, xM);
+      return true;
+    } else if (paramList.isType<RCP<Epetra_MultiVector> >(parameterName)) {
+      RCP<Epetra_MultiVector> epetra_X = Teuchos::null;
+      epetra_X                         = paramList.get<RCP<Epetra_MultiVector> >(parameterName);
+      paramList.remove(parameterName);
+      RCP<Xpetra::EpetraMultiVectorT<int, Node> > xpEpX           = rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epetra_X));
+      RCP<Xpetra::MultiVector<Scalar, int, int, Node> > xpEpXMult = rcp_dynamic_cast<Xpetra::MultiVector<Scalar, int, int, Node> >(xpEpX, true);
+      RCP<XpMultVec> X                                            = rcp_dynamic_cast<XpMultVec>(xpEpXMult, true);
+      paramList.set<RCP<XpMultVec> >(parameterName, X);
+      return true;
+    }
 #endif
-      else if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName) ||
-               (paramList.isType<RCP<const ThyLinOpBase> >(parameterName) && !
-                rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName)).is_null())) {
-        RCP<const ThyDiagLinOpBase> thyM;
-        if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName))
-         thyM = paramList.get<RCP<const ThyDiagLinOpBase> >(parameterName);
-        else
-          thyM = rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName), true);
-        paramList.remove(parameterName);
-        RCP<const Thyra::VectorBase<Scalar> > diag = thyM->getDiag();
-
-        RCP<const XpVec> xpDiag;
-        if (!rcp_dynamic_cast<const thyTpV>(diag).is_null()) {
-          RCP<const tV> tDiag = Thyra::TpetraOperatorVectorExtraction<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getConstTpetraVector(diag);
-          if (!tDiag.is_null())
-            xpDiag = Xpetra::toXpetra(tDiag);
-        }
+    else if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName) ||
+             (paramList.isType<RCP<const ThyLinOpBase> >(parameterName) && !rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName)).is_null())) {
+      RCP<const ThyDiagLinOpBase> thyM;
+      if (paramList.isType<RCP<const ThyDiagLinOpBase> >(parameterName))
+        thyM = paramList.get<RCP<const ThyDiagLinOpBase> >(parameterName);
+      else
+        thyM = rcp_dynamic_cast<const ThyDiagLinOpBase>(paramList.get<RCP<const ThyLinOpBase> >(parameterName), true);
+      paramList.remove(parameterName);
+      RCP<const Thyra::VectorBase<Scalar> > diag = thyM->getDiag();
+
+      RCP<const XpVec> xpDiag;
+      if (!rcp_dynamic_cast<const thyTpV>(diag).is_null()) {
+        RCP<const tV> tDiag = Thyra::TpetraOperatorVectorExtraction<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getConstTpetraVector(diag);
+        if (!tDiag.is_null())
+          xpDiag = Xpetra::toXpetra(tDiag);
+      }
 #ifdef HAVE_MUELU_EPETRA
-        if (xpDiag.is_null()) {
-          RCP<const Epetra_Comm> comm = Thyra::get_Epetra_Comm(*rcp_dynamic_cast<const ThyVSBase>(thyM->range())->getComm());
-          RCP<const Epetra_Map> map = Thyra::get_Epetra_Map(*(thyM->range()), comm);
-          if (!map.is_null()) {
-            RCP<const Epetra_Vector> eDiag = Thyra::get_Epetra_Vector(*map, diag);
-            RCP<Epetra_Vector> nceDiag = rcp_const_cast<Epetra_Vector>(eDiag);
-            RCP<Xpetra::EpetraVectorT<int,Node> > xpEpDiag = rcp(new Xpetra::EpetraVectorT<int,Node>(nceDiag));
-            xpDiag = rcp_dynamic_cast<XpVec>(xpEpDiag, true);
-          }
+      if (xpDiag.is_null()) {
+        RCP<const Epetra_Comm> comm = Thyra::get_Epetra_Comm(*rcp_dynamic_cast<const ThyVSBase>(thyM->range())->getComm());
+        RCP<const Epetra_Map> map   = Thyra::get_Epetra_Map(*(thyM->range()), comm);
+        if (!map.is_null()) {
+          RCP<const Epetra_Vector> eDiag                  = Thyra::get_Epetra_Vector(*map, diag);
+          RCP<Epetra_Vector> nceDiag                      = rcp_const_cast<Epetra_Vector>(eDiag);
+          RCP<Xpetra::EpetraVectorT<int, Node> > xpEpDiag = rcp(new Xpetra::EpetraVectorT<int, Node>(nceDiag));
+          xpDiag                                          = rcp_dynamic_cast<XpVec>(xpEpDiag, true);
         }
+      }
 #endif
-        TEUCHOS_ASSERT(!xpDiag.is_null());
-        RCP<XpMat> M = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(xpDiag);
+      TEUCHOS_ASSERT(!xpDiag.is_null());
+      RCP<XpMat> M = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(xpDiag);
+      paramList.set<RCP<XpMat> >(parameterName, M);
+      return true;
+    } else if (paramList.isType<RCP<const ThyLinOpBase> >(parameterName)) {
+      RCP<const ThyLinOpBase> thyM = paramList.get<RCP<const ThyLinOpBase> >(parameterName);
+      paramList.remove(parameterName);
+      try {
+        RCP<XpMat> M = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
         paramList.set<RCP<XpMat> >(parameterName, M);
-        return true;
-      }
-      else if (paramList.isType<RCP<const ThyLinOpBase> >(parameterName)) {
-        RCP<const ThyLinOpBase> thyM = paramList.get<RCP<const ThyLinOpBase> >(parameterName);
-        paramList.remove(parameterName);
-        try {
-          RCP<XpMat> M = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
-          paramList.set<RCP<XpMat> >(parameterName, M);
-        } catch (std::exception& e) {
-          RCP<XpOp> M = XpThyUtils::toXpetraOperator(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
-          RCP<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(M, true);
-          RCP<tOp> tO = tpOp->getOperator();
-          RCP<tV> diag;
-          if (tO->hasDiagonal()) {
-            diag = rcp(new tV(tO->getRangeMap()));
-            tO->getLocalDiagCopy(*diag);
-          }
-          auto fTpRow = rcp(new MueLu::TpetraOperatorAsRowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(tO, diag));
-          RCP<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpFOp = rcp(new Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> (fTpRow));
-          auto op = rcp_dynamic_cast<XpOp>(tpFOp);
-          paramList.set<RCP<XpOp> >(parameterName, op);
+      } catch (std::exception& e) {
+        RCP<XpOp> M                                                                  = XpThyUtils::toXpetraOperator(Teuchos::rcp_const_cast<ThyLinOpBase>(thyM));
+        RCP<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(M, true);
+        RCP<tOp> tO                                                                  = tpOp->getOperator();
+        RCP<tV> diag;
+        if (tO->hasDiagonal()) {
+          diag = rcp(new tV(tO->getRangeMap()));
+          tO->getLocalDiagCopy(*diag);
         }
-        return true;
-      }
-      else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Parameter " << parameterName << " has wrong type.");
-        return false;
+        auto fTpRow                                                                   = rcp(new MueLu::TpetraOperatorAsRowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(tO, diag));
+        RCP<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpFOp = rcp(new Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>(fTpRow));
+        auto op                                                                       = rcp_dynamic_cast<XpOp>(tpFOp);
+        paramList.set<RCP<XpOp> >(parameterName, op);
       }
-    } else
+      return true;
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Parameter " << parameterName << " has wrong type.");
       return false;
-  }
+    }
+  } else
+    return false;
+}
 #endif
 
-  // Constructors/initializers/accessors
+// Constructors/initializers/accessors
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::MueLuPreconditionerFactory() :
-      paramList_(rcp(new ParameterList()))
-  {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MueLuPreconditionerFactory()
+  : paramList_(rcp(new ParameterList())) {}
 
-  // Overridden from PreconditionerFactoryBase
+// Overridden from PreconditionerFactoryBase
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
-    const RCP<const LinearOpBase<Scalar> > fwdOp = fwdOpSrc.getOp();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
+  const RCP<const LinearOpBase<Scalar> > fwdOp = fwdOpSrc.getOp();
 
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isTpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isTpetra(fwdOp)) return true;
 
 #ifdef HAVE_MUELU_EPETRA
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isEpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isEpetra(fwdOp)) return true;
 #endif
 
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isBlockedOperator(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isBlockedOperator(fwdOp)) return true;
 
-    return false;
-  }
+  return false;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<PreconditionerBase<Scalar> > MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createPrec() const {
+  return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<PreconditionerBase<Scalar> > MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createPrec() const {
-    return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initializePrec(const RCP<const LinearOpSourceBase<Scalar> >& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
+  Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLu::initializePrec")));
+
+  using Teuchos::rcp_dynamic_cast;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initializePrec(const RCP<const LinearOpSourceBase<Scalar> >& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
-    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLu::initializePrec")));
-
-    using Teuchos::rcp_dynamic_cast;
-
-    // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>                     XpMap;
-    typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>      XpOp;
-    typedef MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> MueLuXpOp;
-    typedef Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>       XpThyUtils;
-    // typedef Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XpCrsMat;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> XpBlockedCrsMat;
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMat;
-    // typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>      XpMultVec;
-    // typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>      XpMultVecDouble;
-    typedef Thyra::LinearOpBase<Scalar>                                      ThyLinOpBase;
-    typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
-    typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMV;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType                 Magnitude;
-    typedef Xpetra::MultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node >       XpmMV;
+  // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> XpMap;
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpOp;
+  typedef MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> MueLuXpOp;
+  typedef Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpThyUtils;
+  // typedef Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XpCrsMat;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpBlockedCrsMat;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMat;
+  // typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>      XpMultVec;
+  // typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>      XpMultVecDouble;
+  typedef Thyra::LinearOpBase<Scalar> ThyLinOpBase;
+  typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMV;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  typedef Xpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> XpmMV;
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-    typedef Xpetra::TpetraHalfPrecisionOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> XpHalfPrecOp;
-    typedef typename XpHalfPrecOp::HalfScalar                                     HalfScalar;
-    typedef Xpetra::Operator<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>          XpHalfOp;
-    typedef MueLu::XpetraOperator<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>  MueLuHalfXpOp;
-    typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision              HalfMagnitude;
-    typedef Xpetra::MultiVector<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>       XphMV;
-    typedef Xpetra::MultiVector<HalfMagnitude,LocalOrdinal,GlobalOrdinal,Node >   XphmMV;
-    typedef Xpetra::Matrix<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>            XphMat;
+  typedef Xpetra::TpetraHalfPrecisionOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpHalfPrecOp;
+  typedef typename XpHalfPrecOp::HalfScalar HalfScalar;
+  typedef Xpetra::Operator<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XpHalfOp;
+  typedef MueLu::XpetraOperator<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> MueLuHalfXpOp;
+  typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision HalfMagnitude;
+  typedef Xpetra::MultiVector<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMV;
+  typedef Xpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> XphmMV;
+  typedef Xpetra::Matrix<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMat;
 #endif
 
-
-    // Check precondition
-    TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
-    TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
-    TEUCHOS_ASSERT(prec);
-
-    // Create a copy, as we may remove some things from the list
-    ParameterList paramList = *paramList_;
-
-    // Retrieve wrapped concrete Xpetra matrix from FwdOp
-    const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
-
-    // Check whether it is Epetra/Tpetra
-    bool bIsEpetra  = XpThyUtils::isEpetra(fwdOp);
-    bool bIsTpetra  = XpThyUtils::isTpetra(fwdOp);
-    bool bIsBlocked = XpThyUtils::isBlockedOperator(fwdOp);
-    TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true  && bIsTpetra == true));
-    TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == bIsTpetra) && bIsBlocked == false);
-    TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra != bIsTpetra) && bIsBlocked == true);
-
-    RCP<XpMat> A = Teuchos::null;
-    if(bIsBlocked) {
-      Teuchos::RCP<const Thyra::BlockedLinearOpBase<Scalar> > ThyBlockedOp =
-          Teuchos::rcp_dynamic_cast<const Thyra::BlockedLinearOpBase<Scalar> >(fwdOp);
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(ThyBlockedOp));
-
-      TEUCHOS_TEST_FOR_EXCEPT(ThyBlockedOp->blockExists(0,0)==false);
-
-      Teuchos::RCP<const LinearOpBase<Scalar> > b00 = ThyBlockedOp->getBlock(0,0);
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(b00));
-
-      // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
-      // MueLu needs a non-const object as input
-      RCP<XpMat> A00 = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<LinearOpBase<Scalar> >(b00));
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A00));
-
-      RCP<const XpMap> rowmap00 = A00->getRowMap();
-      RCP< const Teuchos::Comm< int > > comm = rowmap00->getComm();
-
-      // create a Xpetra::BlockedCrsMatrix which derives from Xpetra::Matrix that MueLu can work with
-      RCP<XpBlockedCrsMat> bMat = Teuchos::rcp(new XpBlockedCrsMat(ThyBlockedOp, comm));
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(bMat));
-
-      // save blocked matrix
-      A = bMat;
-    } else {
-      // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
-      // MueLu needs a non-const object as input
-      A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
-    }
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    // extract preconditioner operator
-    RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
-    thyra_precOp = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar> >(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
-
-    // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
-    // rebuild preconditioner if startingOver == true
-    // reuse preconditioner if startingOver == false
-    const bool startingOver = (thyra_precOp.is_null() || !paramList.isParameter("reuse: type") || paramList.get<std::string>("reuse: type") == "none");
-    bool useHalfPrecision = false;
-    if (paramList.isParameter("half precision"))
-      useHalfPrecision = paramList.get<bool>("half precision");
-    else if (paramList.isSublist("Hierarchy") && paramList.sublist("Hierarchy").isParameter("half precision"))
-      useHalfPrecision = paramList.sublist("Hierarchy").get<bool>("half precision");
-    if (useHalfPrecision)
-      TEUCHOS_TEST_FOR_EXCEPTION(!bIsTpetra, MueLu::Exceptions::RuntimeError, "The only scalar type Epetra knows is double, so a half precision preconditioner cannot be constructed.");
-
-    RCP<XpOp> xpPrecOp;
-    if (startingOver == true) {
-      // Convert to Xpetra
-      std::list<std::string> convertXpetra = {"Coordinates", "Nullspace"};
-      for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
-        Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(paramList,*it);
-
-      for (int lvlNo=0; lvlNo < 10; ++lvlNo) {
-        if (paramList.isSublist("level " + std::to_string(lvlNo) + " user data")) {
-          ParameterList& lvlList = paramList.sublist("level " + std::to_string(lvlNo) + " user data");
-          std::list<std::string> convertKeys;
-          for (auto it = lvlList.begin(); it != lvlList.end(); ++it)
-            convertKeys.push_back(lvlList.name(it));
-          for (auto it = convertKeys.begin(); it != convertKeys.end(); ++it)
-            Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(lvlList,*it);
-        }
+  // Check precondition
+  TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
+  TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
+  TEUCHOS_ASSERT(prec);
+
+  // Create a copy, as we may remove some things from the list
+  ParameterList paramList = *paramList_;
+
+  // Retrieve wrapped concrete Xpetra matrix from FwdOp
+  const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
+
+  // Check whether it is Epetra/Tpetra
+  bool bIsEpetra  = XpThyUtils::isEpetra(fwdOp);
+  bool bIsTpetra  = XpThyUtils::isTpetra(fwdOp);
+  bool bIsBlocked = XpThyUtils::isBlockedOperator(fwdOp);
+  TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true && bIsTpetra == true));
+  TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == bIsTpetra) && bIsBlocked == false);
+  TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra != bIsTpetra) && bIsBlocked == true);
+
+  RCP<XpMat> A = Teuchos::null;
+  if (bIsBlocked) {
+    Teuchos::RCP<const Thyra::BlockedLinearOpBase<Scalar> > ThyBlockedOp =
+        Teuchos::rcp_dynamic_cast<const Thyra::BlockedLinearOpBase<Scalar> >(fwdOp);
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(ThyBlockedOp));
+
+    TEUCHOS_TEST_FOR_EXCEPT(ThyBlockedOp->blockExists(0, 0) == false);
+
+    Teuchos::RCP<const LinearOpBase<Scalar> > b00 = ThyBlockedOp->getBlock(0, 0);
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(b00));
+
+    // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
+    // MueLu needs a non-const object as input
+    RCP<XpMat> A00 = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<LinearOpBase<Scalar> >(b00));
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A00));
+
+    RCP<const XpMap> rowmap00           = A00->getRowMap();
+    RCP<const Teuchos::Comm<int> > comm = rowmap00->getComm();
+
+    // create a Xpetra::BlockedCrsMatrix which derives from Xpetra::Matrix that MueLu can work with
+    RCP<XpBlockedCrsMat> bMat = Teuchos::rcp(new XpBlockedCrsMat(ThyBlockedOp, comm));
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(bMat));
+
+    // save blocked matrix
+    A = bMat;
+  } else {
+    // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
+    // MueLu needs a non-const object as input
+    A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
+  }
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
+
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
+
+  // extract preconditioner operator
+  RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
+  thyra_precOp                   = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar> >(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
+
+  // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
+  // rebuild preconditioner if startingOver == true
+  // reuse preconditioner if startingOver == false
+  const bool startingOver = (thyra_precOp.is_null() || !paramList.isParameter("reuse: type") || paramList.get<std::string>("reuse: type") == "none");
+  bool useHalfPrecision   = false;
+  if (paramList.isParameter("half precision"))
+    useHalfPrecision = paramList.get<bool>("half precision");
+  else if (paramList.isSublist("Hierarchy") && paramList.sublist("Hierarchy").isParameter("half precision"))
+    useHalfPrecision = paramList.sublist("Hierarchy").get<bool>("half precision");
+  if (useHalfPrecision)
+    TEUCHOS_TEST_FOR_EXCEPTION(!bIsTpetra, MueLu::Exceptions::RuntimeError, "The only scalar type Epetra knows is double, so a half precision preconditioner cannot be constructed.");
+
+  RCP<XpOp> xpPrecOp;
+  if (startingOver == true) {
+    // Convert to Xpetra
+    std::list<std::string> convertXpetra = {"Coordinates", "Nullspace"};
+    for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
+      Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(paramList, *it);
+
+    for (int lvlNo = 0; lvlNo < 10; ++lvlNo) {
+      if (paramList.isSublist("level " + std::to_string(lvlNo) + " user data")) {
+        ParameterList& lvlList = paramList.sublist("level " + std::to_string(lvlNo) + " user data");
+        std::list<std::string> convertKeys;
+        for (auto it = lvlList.begin(); it != lvlList.end(); ++it)
+          convertKeys.push_back(lvlList.name(it));
+        for (auto it = convertKeys.begin(); it != convertKeys.end(); ++it)
+          Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(lvlList, *it);
       }
+    }
 
-      if (useHalfPrecision) {
+    if (useHalfPrecision) {
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
 
-        // CAG: There is nothing special about the combination double-float,
-        //      except that I feel somewhat confident that Trilinos builds
-        //      with both scalar types.
-
-        // convert to half precision
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-        const std::string userName = "user data";
-        Teuchos::ParameterList& userParamList = paramList.sublist(userName);
-        if (userParamList.isType<RCP<XpmMV> >("Coordinates")) {
-          RCP<XpmMV> coords = userParamList.get<RCP<XpmMV> >("Coordinates");
-          userParamList.remove("Coordinates");
-          RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
-          userParamList.set("Coordinates",halfCoords);
-        }
-        if (userParamList.isType<RCP<XpMV> >("Nullspace")) {
-          RCP<XpMV> nullspace = userParamList.get<RCP<XpMV> >("Nullspace");
-          userParamList.remove("Nullspace");
-          RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
-          userParamList.set("Nullspace",halfNullspace);
-        }
-        if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
-          RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
-          paramList.remove("Coordinates");
-          RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
-          userParamList.set("Coordinates",halfCoords);
-        }
-        if (paramList.isType<RCP<XpMV> >("Nullspace")) {
-          RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
-          paramList.remove("Nullspace");
-          RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
-          userParamList.set("Nullspace",halfNullspace);
-        }
-
+      // CAG: There is nothing special about the combination double-float,
+      //      except that I feel somewhat confident that Trilinos builds
+      //      with both scalar types.
+
+      // convert to half precision
+      RCP<XphMat> halfA                     = Xpetra::convertToHalfPrecision(A);
+      const std::string userName            = "user data";
+      Teuchos::ParameterList& userParamList = paramList.sublist(userName);
+      if (userParamList.isType<RCP<XpmMV> >("Coordinates")) {
+        RCP<XpmMV> coords = userParamList.get<RCP<XpmMV> >("Coordinates");
+        userParamList.remove("Coordinates");
+        RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
+        userParamList.set("Coordinates", halfCoords);
+      }
+      if (userParamList.isType<RCP<XpMV> >("Nullspace")) {
+        RCP<XpMV> nullspace = userParamList.get<RCP<XpMV> >("Nullspace");
+        userParamList.remove("Nullspace");
+        RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
+        userParamList.set("Nullspace", halfNullspace);
+      }
+      if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
+        RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
+        paramList.remove("Coordinates");
+        RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
+        userParamList.set("Coordinates", halfCoords);
+      }
+      if (paramList.isType<RCP<XpMV> >("Nullspace")) {
+        RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
+        paramList.remove("Nullspace");
+        RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
+        userParamList.set("Nullspace", halfNullspace);
+      }
 
-        // build a new half-precision MueLu preconditioner
+      // build a new half-precision MueLu preconditioner
 
-        RCP<MueLu::Hierarchy<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > H = MueLu::CreateXpetraPreconditioner(halfA, paramList);
-        RCP<MueLuHalfXpOp> xpOp = rcp(new MueLuHalfXpOp(H));
-        xpPrecOp = rcp(new XpHalfPrecOp(xpOp));
+      RCP<MueLu::Hierarchy<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> > H = MueLu::CreateXpetraPreconditioner(halfA, paramList);
+      RCP<MueLuHalfXpOp> xpOp                                                 = rcp(new MueLuHalfXpOp(H));
+      xpPrecOp                                                                = rcp(new XpHalfPrecOp(xpOp));
 #else
-        TEUCHOS_TEST_FOR_EXCEPT(true);
+      TEUCHOS_TEST_FOR_EXCEPT(true);
 #endif
-      } else
-        {
-          const std::string userName = "user data";
-          Teuchos::ParameterList& userParamList = paramList.sublist(userName);
-          if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
-            RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
-            paramList.remove("Coordinates");
-            userParamList.set("Coordinates",coords);
-          }
-          if (paramList.isType<RCP<XpMV> >("Nullspace")) {
-            RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
-            paramList.remove("Nullspace");
-            userParamList.set("Nullspace",nullspace);
-          }
-
-          // build a new MueLu RefMaxwell preconditioner
-          RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > H = MueLu::CreateXpetraPreconditioner(A, paramList);
-          xpPrecOp = rcp(new MueLuXpOp(H));
-        }
     } else {
-      // reuse old MueLu hierarchy stored in MueLu Xpetra operator and put in new matrix
-      RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
-      xpPrecOp = rcp_dynamic_cast<XpOp>(thyXpOp->getXpetraOperator(), true);
+      const std::string userName            = "user data";
+      Teuchos::ParameterList& userParamList = paramList.sublist(userName);
+      if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
+        RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
+        paramList.remove("Coordinates");
+        userParamList.set("Coordinates", coords);
+      }
+      if (paramList.isType<RCP<XpMV> >("Nullspace")) {
+        RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
+        paramList.remove("Nullspace");
+        userParamList.set("Nullspace", nullspace);
+      }
+
+      // build a new MueLu RefMaxwell preconditioner
+      RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > H = MueLu::CreateXpetraPreconditioner(A, paramList);
+      xpPrecOp                                                            = rcp(new MueLuXpOp(H));
+    }
+  } else {
+    // reuse old MueLu hierarchy stored in MueLu Xpetra operator and put in new matrix
+    RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
+    xpPrecOp             = rcp_dynamic_cast<XpOp>(thyXpOp->getXpetraOperator(), true);
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-      RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpPrecOp);
-      if (!xpHalfPrecOp.is_null()) {
-        RCP<MueLu::Hierarchy<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > H = rcp_dynamic_cast<MueLuHalfXpOp>(xpHalfPrecOp->GetHalfPrecisionOperator(), true)->GetHierarchy();
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-
-        TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), MueLu::Exceptions::RuntimeError,
-                                   "Thyra::MueLuPreconditionerFactory: Hierarchy has no levels in it");
-        TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), MueLu::Exceptions::RuntimeError,
-                                   "Thyra::MueLuPreconditionerFactory: Hierarchy has no fine level operator");
-        RCP<MueLu::Level> level0 = H->GetLevel(0);
-        RCP<XpHalfOp> O0 = level0->Get<RCP<XpHalfOp> >("A");
-        RCP<XphMat>   A0 = rcp_dynamic_cast<XphMat>(O0, true);
-
-        if (!A0.is_null()) {
-          // If a user provided a "number of equations" argument in a parameter list
-          // during the initial setup, we must honor that settings and reuse it for
-          // all consequent setups.
-          halfA->SetFixedBlockSize(A0->GetFixedBlockSize());
-        }
+    RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpPrecOp);
+    if (!xpHalfPrecOp.is_null()) {
+      RCP<MueLu::Hierarchy<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> > H = rcp_dynamic_cast<MueLuHalfXpOp>(xpHalfPrecOp->GetHalfPrecisionOperator(), true)->GetHierarchy();
+      RCP<XphMat> halfA                                                       = Xpetra::convertToHalfPrecision(A);
+
+      TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), MueLu::Exceptions::RuntimeError,
+                                 "Thyra::MueLuPreconditionerFactory: Hierarchy has no levels in it");
+      TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), MueLu::Exceptions::RuntimeError,
+                                 "Thyra::MueLuPreconditionerFactory: Hierarchy has no fine level operator");
+      RCP<MueLu::Level> level0 = H->GetLevel(0);
+      RCP<XpHalfOp> O0         = level0->Get<RCP<XpHalfOp> >("A");
+      RCP<XphMat> A0           = rcp_dynamic_cast<XphMat>(O0, true);
+
+      if (!A0.is_null()) {
+        // If a user provided a "number of equations" argument in a parameter list
+        // during the initial setup, we must honor that settings and reuse it for
+        // all consequent setups.
+        halfA->SetFixedBlockSize(A0->GetFixedBlockSize());
+      }
 
-        // set new matrix
-        level0->Set("A", halfA);
+      // set new matrix
+      level0->Set("A", halfA);
 
-        H->SetupRe();
-      } else
+      H->SetupRe();
+    } else
 #endif
-      {
-        // get old MueLu hierarchy
-        RCP<MueLuXpOp>    xpOp    = rcp_dynamic_cast<MueLuXpOp>(thyXpOp->getXpetraOperator(), true);
-        RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > H = xpOp->GetHierarchy();;
-
-        TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), MueLu::Exceptions::RuntimeError,
-                                   "Thyra::MueLuPreconditionerFactory: Hierarchy has no levels in it");
-        TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), MueLu::Exceptions::RuntimeError,
-                                   "Thyra::MueLuPreconditionerFactory: Hierarchy has no fine level operator");
-        RCP<MueLu::Level> level0 = H->GetLevel(0);
-        RCP<XpOp>    O0 = level0->Get<RCP<XpOp> >("A");
-        RCP<XpMat>   A0 = rcp_dynamic_cast<XpMat>(O0);
-
-        if (!A0.is_null()) {
-          // If a user provided a "number of equations" argument in a parameter list
-          // during the initial setup, we must honor that settings and reuse it for
-          // all consequent setups.
-          A->SetFixedBlockSize(A0->GetFixedBlockSize());
-        }
-
-        // set new matrix
-        level0->Set("A", A);
-
-        H->SetupRe();
+    {
+      // get old MueLu hierarchy
+      RCP<MueLuXpOp> xpOp                                                 = rcp_dynamic_cast<MueLuXpOp>(thyXpOp->getXpetraOperator(), true);
+      RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > H = xpOp->GetHierarchy();
+      ;
+
+      TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), MueLu::Exceptions::RuntimeError,
+                                 "Thyra::MueLuPreconditionerFactory: Hierarchy has no levels in it");
+      TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), MueLu::Exceptions::RuntimeError,
+                                 "Thyra::MueLuPreconditionerFactory: Hierarchy has no fine level operator");
+      RCP<MueLu::Level> level0 = H->GetLevel(0);
+      RCP<XpOp> O0             = level0->Get<RCP<XpOp> >("A");
+      RCP<XpMat> A0            = rcp_dynamic_cast<XpMat>(O0);
+
+      if (!A0.is_null()) {
+        // If a user provided a "number of equations" argument in a parameter list
+        // during the initial setup, we must honor that settings and reuse it for
+        // all consequent setups.
+        A->SetFixedBlockSize(A0->GetFixedBlockSize());
       }
-    }
-
-    // wrap preconditioner in thyraPrecOp
-    RCP<const VectorSpaceBase<Scalar> > thyraRangeSpace  = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getRangeMap());
-    RCP<const VectorSpaceBase<Scalar> > thyraDomainSpace = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getDomainMap());
-
-    RCP<ThyLinOpBase > thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
-
-    defaultPrec->initializeUnspecified(thyraPrecOp);
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
-    TEUCHOS_ASSERT(prec);
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
 
-    if (fwdOp) {
-      // TODO: Implement properly instead of returning default value
-      *fwdOp = Teuchos::null;
-    }
+      // set new matrix
+      level0->Set("A", A);
 
-    if (supportSolveUse) {
-      // TODO: Implement properly instead of returning default value
-      *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
+      H->SetupRe();
     }
-
-    defaultPrec->uninitialize();
-  }
-
-
-  // Overridden from ParameterListAcceptor
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameterList(RCP<ParameterList> const& paramList) {
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
-    paramList_ = paramList;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getNonconstParameterList() {
-    return paramList_;
-  }
+  // wrap preconditioner in thyraPrecOp
+  RCP<const VectorSpaceBase<Scalar> > thyraRangeSpace  = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getRangeMap());
+  RCP<const VectorSpaceBase<Scalar> > thyraDomainSpace = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getDomainMap());
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::unsetParameterList() {
-    RCP<ParameterList> savedParamList = paramList_;
-    paramList_ = Teuchos::null;
-    return savedParamList;
-  }
+  RCP<ThyLinOpBase> thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getParameterList() const {
-    return paramList_;
-  }
+  defaultPrec->initializeUnspecified(thyraPrecOp);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getValidParameters() const {
-    static RCP<const ParameterList> validPL;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
+  TEUCHOS_ASSERT(prec);
 
-    if (Teuchos::is_null(validPL))
-      validPL = rcp(new ParameterList());
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
 
-    return validPL;
+  if (fwdOp) {
+    // TODO: Implement properly instead of returning default value
+    *fwdOp = Teuchos::null;
   }
 
-  // Public functions overridden from Teuchos::Describable
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MueLuPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::description() const {
-    return "Thyra::MueLuPreconditionerFactory";
+  if (supportSolveUse) {
+    // TODO: Implement properly instead of returning default value
+    *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
   }
-} // namespace Thyra
-
-#endif // HAVE_MUELU_STRATIMIKOS
 
-#endif // ifdef THYRA_MUELU_PRECONDITIONER_FACTORY_DEF_HPP
+  defaultPrec->uninitialize();
+}
+
+// Overridden from ParameterListAcceptor
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameterList(RCP<ParameterList> const& paramList) {
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
+  paramList_ = paramList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNonconstParameterList() {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::unsetParameterList() {
+  RCP<ParameterList> savedParamList = paramList_;
+  paramList_                        = Teuchos::null;
+  return savedParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getParameterList() const {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getValidParameters() const {
+  static RCP<const ParameterList> validPL;
+
+  if (Teuchos::is_null(validPL))
+    validPL = rcp(new ParameterList());
+
+  return validPL;
+}
+
+// Public functions overridden from Teuchos::Describable
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  return "Thyra::MueLuPreconditionerFactory";
+}
+}  // namespace Thyra
+
+#endif  // HAVE_MUELU_STRATIMIKOS
+
+#endif  // ifdef THYRA_MUELU_PRECONDITIONER_FACTORY_DEF_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_decl.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_decl.hpp
index 46702974d0bc..e158803bf6ba 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_decl.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_decl.hpp
@@ -74,7 +74,7 @@
 #include <Xpetra_Matrix.hpp>
 #include <Xpetra_ThyraUtils.hpp>
 
-#include <MueLu_XpetraOperator_decl.hpp> // todo fix me
+#include <MueLu_XpetraOperator_decl.hpp>  // todo fix me
 #include <MueLu_RefMaxwell.hpp>
 #include <MueLu_TpetraOperator.hpp>
 #include <Xpetra_TpetraOperator.hpp>
@@ -92,78 +92,73 @@
 
 namespace Thyra {
 
-  /** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
-      @ingroup MueLuAdapters
-      Add support for MueLu preconditioners in Thyra. This class provides an interface both
-      for Epetra and Tpetra.
-
-      The general implementation only handles Tpetra. For Epetra there is a specialization
-      on SC=double, LO=int, GO=int and NO=EpetraNode.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class MueLuRefMaxwellPreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
-  public:
-
-    /** @name Constructors/initializers/accessors */
-    //@{
-
-    /** \brief . */
-    MueLuRefMaxwellPreconditionerFactory();
-    //@}
-
-    /** @name Overridden from PreconditionerFactoryBase */
-    //@{
-
-    /** \brief . */
-    bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
-    /** \brief . */
-    Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
-    /** \brief . */
-    void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
-                        PreconditionerBase<Scalar>* prec,
-                        const ESupportSolveUse supportSolveUse
-                       ) const;
-    /** \brief . */
-    void uninitializePrec(PreconditionerBase<Scalar>* prec,
-                          Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
-                          ESupportSolveUse* supportSolveUse
-                         ) const;
-
-    //@}
-
-    /** @name Overridden from Teuchos::ParameterListAcceptor */
-    //@{
-
-    /** \brief . */
-    void                                          setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          unsetParameterList();
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          getNonconstParameterList();
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getParameterList() const;
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getValidParameters() const;
-    //@}
-
-    /** \name Public functions overridden from Describable. */
-    //@{
-
-    /** \brief . */
-    std::string description() const;
-
-    // ToDo: Add an override of describe(...) to give more detail!
-
-    //@}
-
-  private:
-
-    Teuchos::RCP<Teuchos::ParameterList> paramList_;
-
-  };
-
-} // namespace Thyra
-
-#endif // #ifdef HAVE_MUELU_STRATIMIKOS
-
-#endif // THYRA_MUELU_REFMAXWELL_PRECONDITIONER_FACTORY_DECL_HPP
+/** @brief Concrete preconditioner factory subclass for Thyra based on MueLu.
+    @ingroup MueLuAdapters
+    Add support for MueLu preconditioners in Thyra. This class provides an interface both
+    for Epetra and Tpetra.
+
+    The general implementation only handles Tpetra. For Epetra there is a specialization
+    on SC=double, LO=int, GO=int and NO=EpetraNode.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class MueLuRefMaxwellPreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
+ public:
+  /** @name Constructors/initializers/accessors */
+  //@{
+
+  /** \brief . */
+  MueLuRefMaxwellPreconditionerFactory();
+  //@}
+
+  /** @name Overridden from PreconditionerFactoryBase */
+  //@{
+
+  /** \brief . */
+  bool isCompatible(const LinearOpSourceBase<Scalar>& fwdOp) const;
+  /** \brief . */
+  Teuchos::RCP<PreconditionerBase<Scalar> > createPrec() const;
+  /** \brief . */
+  void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<Scalar> >& fwdOp,
+                      PreconditionerBase<Scalar>* prec,
+                      const ESupportSolveUse supportSolveUse) const;
+  /** \brief . */
+  void uninitializePrec(PreconditionerBase<Scalar>* prec,
+                        Teuchos::RCP<const LinearOpSourceBase<Scalar> >* fwdOp,
+                        ESupportSolveUse* supportSolveUse) const;
+
+  //@}
+
+  /** @name Overridden from Teuchos::ParameterListAcceptor */
+  //@{
+
+  /** \brief . */
+  void setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> unsetParameterList();
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> getNonconstParameterList();
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getParameterList() const;
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getValidParameters() const;
+  //@}
+
+  /** \name Public functions overridden from Describable. */
+  //@{
+
+  /** \brief . */
+  std::string description() const;
+
+  // ToDo: Add an override of describe(...) to give more detail!
+
+  //@}
+
+ private:
+  Teuchos::RCP<Teuchos::ParameterList> paramList_;
+};
+
+}  // namespace Thyra
+
+#endif  // #ifdef HAVE_MUELU_STRATIMIKOS
+
+#endif  // THYRA_MUELU_REFMAXWELL_PRECONDITIONER_FACTORY_DECL_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_def.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_def.hpp
index d79c5bb5da1e..f310be29b742 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_def.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuRefMaxwellPreconditionerFactory_def.hpp
@@ -52,261 +52,252 @@
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
 
 // This is not as general as possible, but should be good enough for most builds.
-#if((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
-    (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
-# define MUELU_CAN_USE_MIXED_PRECISION
+#if ((defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && !defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && !defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (!defined(HAVE_TPETRA_INST_DOUBLE) && !defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)) || \
+     (defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)))
+#define MUELU_CAN_USE_MIXED_PRECISION
 #endif
 
-
 namespace Thyra {
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::ParameterList;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcp_const_cast;
+using Teuchos::ParameterList;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_const_cast;
+using Teuchos::rcp_dynamic_cast;
 
-  // Constructors/initializers/accessors
+// Constructors/initializers/accessors
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::MueLuRefMaxwellPreconditionerFactory() :
-      paramList_(rcp(new ParameterList()))
-  {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MueLuRefMaxwellPreconditionerFactory()
+  : paramList_(rcp(new ParameterList())) {}
 
-  // Overridden from PreconditionerFactoryBase
+// Overridden from PreconditionerFactoryBase
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
-    const RCP<const LinearOpBase<Scalar> > fwdOp = fwdOpSrc.getOp();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
+  const RCP<const LinearOpBase<Scalar>> fwdOp = fwdOpSrc.getOp();
 
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isTpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isTpetra(fwdOp)) return true;
 
 #ifdef HAVE_MUELU_EPETRA
-    if (Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isEpetra(fwdOp)) return true;
+  if (Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isEpetra(fwdOp)) return true;
 #endif
 
-    return false;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<PreconditionerBase<Scalar> > MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createPrec() const {
-    return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initializePrec(const RCP<const LinearOpSourceBase<Scalar> >& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
-
-    // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
-    typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>      XpOp;
-    typedef Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>       XpThyUtils;
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMat;
-    typedef Thyra::LinearOpBase<Scalar>                                      ThyLinOpBase;
-    typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<PreconditionerBase<Scalar>> MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createPrec() const {
+  return Teuchos::rcp(new DefaultPreconditioner<Scalar>);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initializePrec(const RCP<const LinearOpSourceBase<Scalar>>& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
+  // we are using typedefs here, since we are using objects from different packages (Xpetra, Thyra,...)
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpOp;
+  typedef Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpThyUtils;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMat;
+  typedef Thyra::LinearOpBase<Scalar> ThyLinOpBase;
+  typedef Thyra::XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> ThyXpOp;
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-    typedef Xpetra::TpetraHalfPrecisionOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> XpHalfPrecOp;
-    typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>           XpMV;
-    typedef typename XpHalfPrecOp::HalfScalar                                     HalfScalar;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType                 Magnitude;
-    typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision              HalfMagnitude;
-    typedef Xpetra::MultiVector<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>       XphMV;
-    typedef Xpetra::MultiVector<Magnitude,LocalOrdinal,GlobalOrdinal,Node >       XpmMV;
-    typedef Xpetra::MultiVector<HalfMagnitude,LocalOrdinal,GlobalOrdinal,Node >   XphmMV;
-    typedef Xpetra::Matrix<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>            XphMat;
+  typedef Xpetra::TpetraHalfPrecisionOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpHalfPrecOp;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XpMV;
+  typedef typename XpHalfPrecOp::HalfScalar HalfScalar;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  typedef typename Teuchos::ScalarTraits<Magnitude>::halfPrecision HalfMagnitude;
+  typedef Xpetra::MultiVector<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMV;
+  typedef Xpetra::MultiVector<Magnitude, LocalOrdinal, GlobalOrdinal, Node> XpmMV;
+  typedef Xpetra::MultiVector<HalfMagnitude, LocalOrdinal, GlobalOrdinal, Node> XphmMV;
+  typedef Xpetra::Matrix<HalfScalar, LocalOrdinal, GlobalOrdinal, Node> XphMat;
 #endif
-    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLuRefMaxwell::initializePrec")));
-
-    // Check precondition
-    TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
-    TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
-    TEUCHOS_ASSERT(prec);
-
-    // Create a copy, as we may remove some things from the list
-    ParameterList paramList = *paramList_;
-
-    // Retrieve wrapped concrete Xpetra matrix from FwdOp
-    const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
-
-    // Check whether it is Epetra/Tpetra
-    bool bIsEpetra  = XpThyUtils::isEpetra(fwdOp);
-    bool bIsTpetra  = XpThyUtils::isTpetra(fwdOp);
-    TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true  && bIsTpetra == true));
-
-    // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
-    // MueLu needs a non-const object as input
-    RCP<XpMat> A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    // extract preconditioner operator
-    RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
-    thyra_precOp = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar> >(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
-
-    // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
-    // rebuild preconditioner if startingOver == true
-    // reuse preconditioner if startingOver == false
-    const bool startingOver = (thyra_precOp.is_null() || !paramList.isParameter("refmaxwell: enable reuse") || !paramList.get<bool>("refmaxwell: enable reuse"));
-    const bool useHalfPrecision = paramList.get<bool>("half precision", false) && bIsTpetra;
-
-    RCP<XpOp> xpPrecOp;
-    if (startingOver == true) {
-
-      // Convert to Xpetra
-      std::list<std::string> convertMat = {
+  Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("ThyraMueLuRefMaxwell::initializePrec")));
+
+  // Check precondition
+  TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
+  TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
+  TEUCHOS_ASSERT(prec);
+
+  // Create a copy, as we may remove some things from the list
+  ParameterList paramList = *paramList_;
+
+  // Retrieve wrapped concrete Xpetra matrix from FwdOp
+  const RCP<const ThyLinOpBase> fwdOp = fwdOpSrc->getOp();
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
+
+  // Check whether it is Epetra/Tpetra
+  bool bIsEpetra = XpThyUtils::isEpetra(fwdOp);
+  bool bIsTpetra = XpThyUtils::isTpetra(fwdOp);
+  TEUCHOS_TEST_FOR_EXCEPT((bIsEpetra == true && bIsTpetra == true));
+
+  // wrap the forward operator as an Xpetra::Matrix that MueLu can work with
+  // MueLu needs a non-const object as input
+  RCP<XpMat> A = XpThyUtils::toXpetra(Teuchos::rcp_const_cast<ThyLinOpBase>(fwdOp));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
+
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar>> defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
+
+  // extract preconditioner operator
+  RCP<ThyLinOpBase> thyra_precOp = Teuchos::null;
+  thyra_precOp                   = rcp_dynamic_cast<Thyra::LinearOpBase<Scalar>>(defaultPrec->getNonconstUnspecifiedPrecOp(), true);
+
+  // make a decision whether to (re)build the multigrid preconditioner or reuse the old one
+  // rebuild preconditioner if startingOver == true
+  // reuse preconditioner if startingOver == false
+  const bool startingOver     = (thyra_precOp.is_null() || !paramList.isParameter("refmaxwell: enable reuse") || !paramList.get<bool>("refmaxwell: enable reuse"));
+  const bool useHalfPrecision = paramList.get<bool>("half precision", false) && bIsTpetra;
+
+  RCP<XpOp> xpPrecOp;
+  if (startingOver == true) {
+    // Convert to Xpetra
+    std::list<std::string> convertMat = {
         "Dk_1", "Dk_2", "D0",
         "Mk_one", "Mk_1_one", "M1_beta", "M1_alpha",
         "invMk_1_invBeta", "invMk_2_invAlpha",
         // for backwards compatibility
-         "M1", "Ms", "M0inv"
-      };
-      std::list<std::string> convertMV = {"Coordinates", "Nullspace"};
-      std::list<std::string> convertXpetra;
-      convertXpetra.insert(convertXpetra.end(), convertMV.begin(), convertMV.end());
-      convertXpetra.insert(convertXpetra.end(), convertMat.begin(), convertMat.end());
-      for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
-        Converters<Scalar,LocalOrdinal,GlobalOrdinal,Node>::replaceWithXpetra(paramList,*it);
-
-      paramList.set<bool>("refmaxwell: use as preconditioner", true);
-      if (useHalfPrecision) {
+        "M1", "Ms", "M0inv"};
+    std::list<std::string> convertMV = {"Coordinates", "Nullspace"};
+    std::list<std::string> convertXpetra;
+    convertXpetra.insert(convertXpetra.end(), convertMV.begin(), convertMV.end());
+    convertXpetra.insert(convertXpetra.end(), convertMat.begin(), convertMat.end());
+    for (auto it = convertXpetra.begin(); it != convertXpetra.end(); ++it)
+      Converters<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceWithXpetra(paramList, *it);
+
+    paramList.set<bool>("refmaxwell: use as preconditioner", true);
+    if (useHalfPrecision) {
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
 
-        // convert to half precision
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-        if (paramList.isType<RCP<XpmMV> >("Coordinates")) {
-          RCP<XpmMV> coords = paramList.get<RCP<XpmMV> >("Coordinates");
-          paramList.remove("Coordinates");
-          RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
-          paramList.set("Coordinates",halfCoords);
-        }
-        if (paramList.isType<RCP<XpMV> >("Nullspace")) {
-          RCP<XpMV> nullspace = paramList.get<RCP<XpMV> >("Nullspace");
-          paramList.remove("Nullspace");
-          RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
-          paramList.set("Nullspace",halfNullspace);
-        }
-        for (auto it = convertMat.begin(); it != convertMat.end(); ++it) {
-          if (paramList.isType<RCP<XpMat> >(*it)) {
-            RCP<XpMat> M = paramList.get<RCP<XpMat> >(*it);
-            paramList.remove(*it);
-            RCP<XphMat> halfM = Xpetra::convertToHalfPrecision(M);
-            paramList.set(*it,halfM);
-          }
+      // convert to half precision
+      RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
+      if (paramList.isType<RCP<XpmMV>>("Coordinates")) {
+        RCP<XpmMV> coords = paramList.get<RCP<XpmMV>>("Coordinates");
+        paramList.remove("Coordinates");
+        RCP<XphmMV> halfCoords = Xpetra::convertToHalfPrecision(coords);
+        paramList.set("Coordinates", halfCoords);
+      }
+      if (paramList.isType<RCP<XpMV>>("Nullspace")) {
+        RCP<XpMV> nullspace = paramList.get<RCP<XpMV>>("Nullspace");
+        paramList.remove("Nullspace");
+        RCP<XphMV> halfNullspace = Xpetra::convertToHalfPrecision(nullspace);
+        paramList.set("Nullspace", halfNullspace);
+      }
+      for (auto it = convertMat.begin(); it != convertMat.end(); ++it) {
+        if (paramList.isType<RCP<XpMat>>(*it)) {
+          RCP<XpMat> M = paramList.get<RCP<XpMat>>(*it);
+          paramList.remove(*it);
+          RCP<XphMat> halfM = Xpetra::convertToHalfPrecision(M);
+          paramList.set(*it, halfM);
         }
+      }
 
-        // build a new half-precision MueLu RefMaxwell preconditioner
-        RCP<MueLu::RefMaxwell<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > halfPrec = rcp(new MueLu::RefMaxwell<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>(halfA, paramList, true));
-        xpPrecOp = rcp(new XpHalfPrecOp(halfPrec));
+      // build a new half-precision MueLu RefMaxwell preconditioner
+      RCP<MueLu::RefMaxwell<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>> halfPrec = rcp(new MueLu::RefMaxwell<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>(halfA, paramList, true));
+      xpPrecOp                                                                       = rcp(new XpHalfPrecOp(halfPrec));
 #else
-        TEUCHOS_TEST_FOR_EXCEPT(true);
+      TEUCHOS_TEST_FOR_EXCEPT(true);
 #endif
-      } else
-        {
-          // build a new MueLu RefMaxwell preconditioner
-          RCP<MueLu::RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp(new MueLu::RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A, paramList, true));
-          xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-        }
     } else {
-      // reuse old MueLu preconditioner stored in MueLu Xpetra operator and put in new matrix
+      // build a new MueLu RefMaxwell preconditioner
+      RCP<MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp(new MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, paramList, true));
+      xpPrecOp                                                                         = rcp_dynamic_cast<XpOp>(preconditioner);
+    }
+  } else {
+    // reuse old MueLu preconditioner stored in MueLu Xpetra operator and put in new matrix
 
-      RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
-      RCP<XpOp>    xpOp    = thyXpOp->getXpetraOperator();
+    RCP<ThyXpOp> thyXpOp = rcp_dynamic_cast<ThyXpOp>(thyra_precOp, true);
+    RCP<XpOp> xpOp       = thyXpOp->getXpetraOperator();
 #if defined(MUELU_CAN_USE_MIXED_PRECISION)
-      RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpOp);
-      if (!xpHalfPrecOp.is_null()) {
-        RCP<MueLu::RefMaxwell<HalfScalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp_dynamic_cast<MueLu::RefMaxwell<HalfScalar,LocalOrdinal,GlobalOrdinal,Node>>(xpHalfPrecOp->GetHalfPrecisionOperator(), true);
-        RCP<XphMat> halfA = Xpetra::convertToHalfPrecision(A);
-        preconditioner->resetMatrix(halfA);
-        xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-      } else
+    RCP<XpHalfPrecOp> xpHalfPrecOp = rcp_dynamic_cast<XpHalfPrecOp>(xpOp);
+    if (!xpHalfPrecOp.is_null()) {
+      RCP<MueLu::RefMaxwell<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp_dynamic_cast<MueLu::RefMaxwell<HalfScalar, LocalOrdinal, GlobalOrdinal, Node>>(xpHalfPrecOp->GetHalfPrecisionOperator(), true);
+      RCP<XphMat> halfA                                                                    = Xpetra::convertToHalfPrecision(A);
+      preconditioner->resetMatrix(halfA);
+      xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
+    } else
 #endif
-        {
-          RCP<MueLu::RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node> > preconditioner = rcp_dynamic_cast<MueLu::RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>>(xpOp, true);
-          preconditioner->resetMatrix(A);
-          xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
-        }
-    }
-
-    // wrap preconditioner in thyraPrecOp
-    RCP<const VectorSpaceBase<Scalar> > thyraRangeSpace  = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getRangeMap());
-    RCP<const VectorSpaceBase<Scalar> > thyraDomainSpace = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpPrecOp->getDomainMap());
-
-    RCP<ThyLinOpBase > thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
-
-    defaultPrec->initializeUnspecified(thyraPrecOp);
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
-    TEUCHOS_ASSERT(prec);
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<Scalar> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    if (fwdOp) {
-      // TODO: Implement properly instead of returning default value
-      *fwdOp = Teuchos::null;
+    {
+      RCP<MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>> preconditioner = rcp_dynamic_cast<MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(xpOp, true);
+      preconditioner->resetMatrix(A);
+      xpPrecOp = rcp_dynamic_cast<XpOp>(preconditioner);
     }
-
-    if (supportSolveUse) {
-      // TODO: Implement properly instead of returning default value
-      *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
-    }
-
-    defaultPrec->uninitialize();
-  }
-
-
-  // Overridden from ParameterListAcceptor
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameterList(RCP<ParameterList> const& paramList) {
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
-    paramList_ = paramList;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getNonconstParameterList() {
-    return paramList_;
-  }
+  // wrap preconditioner in thyraPrecOp
+  RCP<const VectorSpaceBase<Scalar>> thyraRangeSpace  = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getRangeMap());
+  RCP<const VectorSpaceBase<Scalar>> thyraDomainSpace = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpPrecOp->getDomainMap());
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::unsetParameterList() {
-    RCP<ParameterList> savedParamList = paramList_;
-    paramList_ = Teuchos::null;
-    return savedParamList;
-  }
+  RCP<ThyLinOpBase> thyraPrecOp = Thyra::xpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(thyraRangeSpace, thyraDomainSpace, xpPrecOp);
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraPrecOp));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getParameterList() const {
-    return paramList_;
-  }
+  defaultPrec->initializeUnspecified(thyraPrecOp);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getValidParameters() const {
-    static RCP<const ParameterList> validPL;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar>>* fwdOp, ESupportSolveUse* supportSolveUse) const {
+  TEUCHOS_ASSERT(prec);
 
-    if (Teuchos::is_null(validPL))
-      validPL = rcp(new ParameterList());
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<Scalar>> defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<Scalar>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
 
-    return validPL;
+  if (fwdOp) {
+    // TODO: Implement properly instead of returning default value
+    *fwdOp = Teuchos::null;
   }
 
-  // Public functions overridden from Teuchos::Describable
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MueLuRefMaxwellPreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::description() const {
-    return "Thyra::MueLuRefMaxwellPreconditionerFactory";
+  if (supportSolveUse) {
+    // TODO: Implement properly instead of returning default value
+    *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
   }
-} // namespace Thyra
-
-#endif // HAVE_MUELU_STRATIMIKOS
 
-#endif // ifdef THYRA_MUELU_REFMAXWELL_PRECONDITIONER_FACTORY_DEF_HPP
+  defaultPrec->uninitialize();
+}
+
+// Overridden from ParameterListAcceptor
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameterList(RCP<ParameterList> const& paramList) {
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
+  paramList_ = paramList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNonconstParameterList() {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::unsetParameterList() {
+  RCP<ParameterList> savedParamList = paramList_;
+  paramList_                        = Teuchos::null;
+  return savedParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getParameterList() const {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getValidParameters() const {
+  static RCP<const ParameterList> validPL;
+
+  if (Teuchos::is_null(validPL))
+    validPL = rcp(new ParameterList());
+
+  return validPL;
+}
+
+// Public functions overridden from Teuchos::Describable
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MueLuRefMaxwellPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  return "Thyra::MueLuRefMaxwellPreconditionerFactory";
+}
+}  // namespace Thyra
+
+#endif  // HAVE_MUELU_STRATIMIKOS
+
+#endif  // ifdef THYRA_MUELU_REFMAXWELL_PRECONDITIONER_FACTORY_DEF_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_decl.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_decl.hpp
index 2f73ff9c52cd..cffc273eb42c 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_decl.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_decl.hpp
@@ -47,7 +47,6 @@
 #define THYRA_MUELU_TPETRA_Q2Q1PRECONDITIONER_FACTORY_DECL_HPP
 #ifdef HAVE_MUELU_EXPERIMENTAL
 
-
 #include "Thyra_PreconditionerFactoryBase.hpp"
 
 #include <Tpetra_KokkosCompat_DefaultNode.hpp>
@@ -59,86 +58,83 @@
 
 namespace Thyra {
 
-  /** \brief Concrete preconditioner factory subclass based on MueLu.
-   *
-   * ToDo: Finish documentation!
-   */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class MueLuTpetraQ2Q1PreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
-  private:
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-  public:
-
-    /** @name Constructors/initializers/accessors */
-    //@{
-
-    /** \brief . */
-    MueLuTpetraQ2Q1PreconditionerFactory();
-    //@}
-
-    /** @name Overridden from PreconditionerFactoryBase */
-    //@{
-
-    /** \brief . */
-    bool isCompatible( const LinearOpSourceBase<SC> &fwdOp ) const;
-    /** \brief . */
-    Teuchos::RCP<PreconditionerBase<SC> > createPrec() const;
-    /** \brief . */
-    void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<SC> > &fwdOp, PreconditionerBase<SC> *prec, const ESupportSolveUse supportSolveUse) const;
-    /** \brief . */
-    void uninitializePrec(PreconditionerBase<SC> *prec, Teuchos::RCP<const LinearOpSourceBase<SC> > *fwdOp, ESupportSolveUse *supportSolveUse) const;
-    //@}
-
-    /** @name Overridden from Teuchos::ParameterListAcceptor */
-    //@{
-
-    /** \brief . */
-    void                                          setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          unsetParameterList();
-    /** \brief . */
-    Teuchos::RCP<Teuchos::ParameterList>          getNonconstParameterList();
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getParameterList() const;
-    /** \brief . */
-    Teuchos::RCP<const Teuchos::ParameterList>    getValidParameters() const;
-    //@}
-
-    /** \name Public functions overridden from Describable. */
-    //@{
-
-    /** \brief . */
-    std::string description() const;
-
-    // ToDo: Add an override of describe(...) to give more detail!
-
-    //@}
-
-  private:
-
-    Teuchos::RCP<MueLu::TpetraOperator<SC,LO,GO,NO> >
-    Q2Q1MkPrecond(const ParameterList& paramList,
-                  const Teuchos::RCP<Tpetra::MultiVector<SC,LO,GO,NO> >& velCoords,
-                  const Teuchos::RCP<Tpetra::MultiVector<SC,LO,GO,NO> >& presCoords,
-                  const Teuchos::ArrayRCP<LO>& p2vMap,
-                  const Teko::LinearOp& thA11, const Teko::LinearOp& thA12, const Teko::LinearOp& thA21, const Teko::LinearOp& thA11_9Pt) const;
-
-    Teuchos::RCP<Xpetra::Matrix<SC,LO,GO,NO> > Absolute    (const Xpetra::Matrix<SC,LO,GO,NO>& A) const;
-    Teuchos::RCP<Xpetra::Matrix<SC,LO,GO,NO> > FilterMatrix(Xpetra::Matrix<SC,LO,GO,NO>& A, Xpetra::Matrix<SC,LO,GO,NO>& Pattern, SC dropTol) const;
-
-    void SetDependencyTree     (MueLu::FactoryManager<SC,LO,GO,NO>& M, const ParameterList& paramList) const;
-    void SetBlockDependencyTree(MueLu::FactoryManager<SC,LO,GO,NO>& M, LO row, LO col, const std::string& mode, const ParameterList& paramList)  const;
-
-    RCP<MueLu::FactoryBase> GetSmoother(const std::string& type, const ParameterList& paramList, bool coarseSolver) const;
-
-    Teuchos::RCP<Teuchos::ParameterList> paramList_;
-
-  };
-
-} // namespace Thyra
+/** \brief Concrete preconditioner factory subclass based on MueLu.
+ *
+ * ToDo: Finish documentation!
+ */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class MueLuTpetraQ2Q1PreconditionerFactory : public PreconditionerFactoryBase<Scalar> {
+ private:
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+ public:
+  /** @name Constructors/initializers/accessors */
+  //@{
+
+  /** \brief . */
+  MueLuTpetraQ2Q1PreconditionerFactory();
+  //@}
+
+  /** @name Overridden from PreconditionerFactoryBase */
+  //@{
+
+  /** \brief . */
+  bool isCompatible(const LinearOpSourceBase<SC>& fwdOp) const;
+  /** \brief . */
+  Teuchos::RCP<PreconditionerBase<SC> > createPrec() const;
+  /** \brief . */
+  void initializePrec(const Teuchos::RCP<const LinearOpSourceBase<SC> >& fwdOp, PreconditionerBase<SC>* prec, const ESupportSolveUse supportSolveUse) const;
+  /** \brief . */
+  void uninitializePrec(PreconditionerBase<SC>* prec, Teuchos::RCP<const LinearOpSourceBase<SC> >* fwdOp, ESupportSolveUse* supportSolveUse) const;
+  //@}
+
+  /** @name Overridden from Teuchos::ParameterListAcceptor */
+  //@{
+
+  /** \brief . */
+  void setParameterList(const Teuchos::RCP<Teuchos::ParameterList>& paramList);
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> unsetParameterList();
+  /** \brief . */
+  Teuchos::RCP<Teuchos::ParameterList> getNonconstParameterList();
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getParameterList() const;
+  /** \brief . */
+  Teuchos::RCP<const Teuchos::ParameterList> getValidParameters() const;
+  //@}
+
+  /** \name Public functions overridden from Describable. */
+  //@{
+
+  /** \brief . */
+  std::string description() const;
+
+  // ToDo: Add an override of describe(...) to give more detail!
+
+  //@}
+
+ private:
+  Teuchos::RCP<MueLu::TpetraOperator<SC, LO, GO, NO> >
+  Q2Q1MkPrecond(const ParameterList& paramList,
+                const Teuchos::RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& velCoords,
+                const Teuchos::RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& presCoords,
+                const Teuchos::ArrayRCP<LO>& p2vMap,
+                const Teko::LinearOp& thA11, const Teko::LinearOp& thA12, const Teko::LinearOp& thA21, const Teko::LinearOp& thA11_9Pt) const;
+
+  Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, NO> > Absolute(const Xpetra::Matrix<SC, LO, GO, NO>& A) const;
+  Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, NO> > FilterMatrix(Xpetra::Matrix<SC, LO, GO, NO>& A, Xpetra::Matrix<SC, LO, GO, NO>& Pattern, SC dropTol) const;
+
+  void SetDependencyTree(MueLu::FactoryManager<SC, LO, GO, NO>& M, const ParameterList& paramList) const;
+  void SetBlockDependencyTree(MueLu::FactoryManager<SC, LO, GO, NO>& M, LO row, LO col, const std::string& mode, const ParameterList& paramList) const;
+
+  RCP<MueLu::FactoryBase> GetSmoother(const std::string& type, const ParameterList& paramList, bool coarseSolver) const;
+
+  Teuchos::RCP<Teuchos::ParameterList> paramList_;
+};
+
+}  // namespace Thyra
 #endif
-#endif // THYRA_MUELU_TPETRA_Q2Q1PRECONDITIONER_FACTORY_DECL_HPP
+#endif  // THYRA_MUELU_TPETRA_Q2Q1PRECONDITIONER_FACTORY_DECL_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_def.hpp b/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_def.hpp
index 0c69bd698027..44c70302b9e3 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_def.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_MueLuTpetraQ2Q1PreconditionerFactory_def.hpp
@@ -105,249 +105,272 @@ namespace Thyra {
 #define MUELU_GPD(name, type, defaultValue) \
   (paramList.isParameter(name) ? paramList.get<type>(name) : defaultValue)
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::rcp_const_cast;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::ParameterList;
-  using Teuchos::ArrayView;
-  using Teuchos::ArrayRCP;
-  using Teuchos::as;
-  using Teuchos::Array;
-
-  // Constructors/initializers/accessors
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::MueLuTpetraQ2Q1PreconditionerFactory() {}
-
-
-  // Overridden from PreconditionerFactoryBase
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
-    typedef Thyra ::TpetraLinearOp<SC,LO,GO,NO> ThyraTpetraLinOp;
-    typedef Tpetra::Operator      <SC,LO,GO,NO> TpetraLinOp;
-    typedef Tpetra::CrsMatrix     <SC,LO,GO,NO> TpetraCrsMat;
-
-    const RCP<const LinearOpBase<SC> > fwdOp            = fwdOpSrc.getOp();
-    const RCP<const ThyraTpetraLinOp>  thyraTpetraFwdOp = rcp_dynamic_cast<const ThyraTpetraLinOp>(fwdOp);
-    const RCP<const TpetraLinOp>       tpetraFwdOp      = Teuchos::nonnull(thyraTpetraFwdOp) ? thyraTpetraFwdOp->getConstTpetraOperator() : Teuchos::null;
-    const RCP<const TpetraCrsMat>      tpetraFwdCrsMat  = rcp_dynamic_cast<const TpetraCrsMat>(tpetraFwdOp);
-
-    return Teuchos::nonnull(tpetraFwdCrsMat);
+using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::ArrayView;
+using Teuchos::as;
+using Teuchos::ParameterList;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_const_cast;
+using Teuchos::rcp_dynamic_cast;
+
+// Constructors/initializers/accessors
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MueLuTpetraQ2Q1PreconditionerFactory() {}
+
+// Overridden from PreconditionerFactoryBase
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isCompatible(const LinearOpSourceBase<Scalar>& fwdOpSrc) const {
+  typedef Thyra ::TpetraLinearOp<SC, LO, GO, NO> ThyraTpetraLinOp;
+  typedef Tpetra::Operator<SC, LO, GO, NO> TpetraLinOp;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TpetraCrsMat;
+
+  const RCP<const LinearOpBase<SC> > fwdOp           = fwdOpSrc.getOp();
+  const RCP<const ThyraTpetraLinOp> thyraTpetraFwdOp = rcp_dynamic_cast<const ThyraTpetraLinOp>(fwdOp);
+  const RCP<const TpetraLinOp> tpetraFwdOp           = Teuchos::nonnull(thyraTpetraFwdOp) ? thyraTpetraFwdOp->getConstTpetraOperator() : Teuchos::null;
+  const RCP<const TpetraCrsMat> tpetraFwdCrsMat      = rcp_dynamic_cast<const TpetraCrsMat>(tpetraFwdOp);
+
+  return Teuchos::nonnull(tpetraFwdCrsMat);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<PreconditionerBase<Scalar> >
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createPrec() const {
+  return rcp(new DefaultPreconditioner<SC>);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initializePrec(const RCP<const LinearOpSourceBase<Scalar> >& fwdOpSrc, PreconditionerBase<Scalar>* prec, const ESupportSolveUse supportSolveUse) const {
+  // Check precondition
+  TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
+  TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
+  TEUCHOS_ASSERT(prec);
+
+  // Retrieve wrapped concrete Tpetra matrix from FwdOp
+  const RCP<const LinearOpBase<SC> > fwdOp = fwdOpSrc->getOp();
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
+
+  typedef Thyra::TpetraLinearOp<SC, LO, GO, NO> ThyraTpetraLinOp;
+  const RCP<const ThyraTpetraLinOp> thyraTpetraFwdOp = rcp_dynamic_cast<const ThyraTpetraLinOp>(fwdOp);
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraTpetraFwdOp));
+
+  typedef Tpetra::Operator<SC, LO, GO, NO> TpetraLinOp;
+  const RCP<const TpetraLinOp> tpetraFwdOp = thyraTpetraFwdOp->getConstTpetraOperator();
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(tpetraFwdOp));
+
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TpetraCrsMat;
+  const RCP<const TpetraCrsMat> tpetraFwdCrsMat = rcp_dynamic_cast<const TpetraCrsMat>(tpetraFwdOp);
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(tpetraFwdCrsMat));
+
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<SC> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<SC>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
+
+  // Workaround since MueLu interface does not accept const matrix as input
+  const RCP<TpetraCrsMat> tpetraFwdCrsMatNonConst = rcp_const_cast<TpetraCrsMat>(tpetraFwdCrsMat);
+
+  // Create and compute the initial preconditioner
+
+  // Create a copy, as we may remove some things from the list
+  ParameterList paramList = *paramList_;
+
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> MultiVector;
+  RCP<MultiVector> coords, nullspace, velCoords, presCoords;
+  ArrayRCP<LO> p2vMap;
+  Teko::LinearOp thA11, thA12, thA21, thA11_9Pt;
+  if (paramList.isType<RCP<MultiVector> >("Coordinates")) {
+    coords = paramList.get<RCP<MultiVector> >("Coordinates");
+    paramList.remove("Coordinates");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<PreconditionerBase<Scalar> >
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createPrec() const {
-    return rcp(new DefaultPreconditioner<SC>);
+  if (paramList.isType<RCP<MultiVector> >("Nullspace")) {
+    nullspace = paramList.get<RCP<MultiVector> >("Nullspace");
+    paramList.remove("Nullspace");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initializePrec(const RCP<const LinearOpSourceBase<Scalar> > &fwdOpSrc, PreconditionerBase<Scalar> *prec, const ESupportSolveUse supportSolveUse) const {
-    // Check precondition
-    TEUCHOS_ASSERT(Teuchos::nonnull(fwdOpSrc));
-    TEUCHOS_ASSERT(this->isCompatible(*fwdOpSrc));
-    TEUCHOS_ASSERT(prec);
-
-    // Retrieve wrapped concrete Tpetra matrix from FwdOp
-    const RCP<const LinearOpBase<SC> > fwdOp = fwdOpSrc->getOp();
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(fwdOp));
-
-    typedef Thyra::TpetraLinearOp<SC,LO,GO,NO> ThyraTpetraLinOp;
-    const RCP<const ThyraTpetraLinOp> thyraTpetraFwdOp = rcp_dynamic_cast<const ThyraTpetraLinOp>(fwdOp);
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(thyraTpetraFwdOp));
-
-    typedef Tpetra::Operator<SC,LO,GO,NO> TpetraLinOp;
-    const RCP<const TpetraLinOp> tpetraFwdOp = thyraTpetraFwdOp->getConstTpetraOperator();
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(tpetraFwdOp));
-
-    typedef Tpetra::CrsMatrix<SC,LO,GO,NO> TpetraCrsMat;
-    const RCP<const TpetraCrsMat> tpetraFwdCrsMat = rcp_dynamic_cast<const TpetraCrsMat>(tpetraFwdOp);
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(tpetraFwdCrsMat));
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<SC> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<SC> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    // Workaround since MueLu interface does not accept const matrix as input
-    const RCP<TpetraCrsMat> tpetraFwdCrsMatNonConst = rcp_const_cast<TpetraCrsMat>(tpetraFwdCrsMat);
-
-    // Create and compute the initial preconditioner
-
-    // Create a copy, as we may remove some things from the list
-    ParameterList paramList = *paramList_;
-
-    typedef Tpetra::MultiVector<SC,LO,GO,NO> MultiVector;
-    RCP<MultiVector> coords, nullspace, velCoords, presCoords;
-    ArrayRCP<LO>     p2vMap;
-    Teko::LinearOp   thA11, thA12, thA21, thA11_9Pt;
-    if (paramList.isType<RCP<MultiVector> >("Coordinates")) { coords     = paramList.get<RCP<MultiVector> >("Coordinates"); paramList.remove("Coordinates"); }
-    if (paramList.isType<RCP<MultiVector> >("Nullspace"))   { nullspace  = paramList.get<RCP<MultiVector> >("Nullspace");   paramList.remove("Nullspace"); }
-    if (paramList.isType<RCP<MultiVector> >("Velcoords"))   { velCoords  = paramList.get<RCP<MultiVector> >("Velcoords");   paramList.remove("Velcoords"); }
-    if (paramList.isType<RCP<MultiVector> >("Prescoords"))  { presCoords = paramList.get<RCP<MultiVector> >("Prescoords");  paramList.remove("Prescoords"); }
-    if (paramList.isType<ArrayRCP<LO> >    ("p2vMap"))      { p2vMap     = paramList.get<ArrayRCP<LO> >    ("p2vMap");      paramList.remove("p2vMap"); }
-    if (paramList.isType<Teko::LinearOp>   ("A11"))         { thA11      = paramList.get<Teko::LinearOp>   ("A11");         paramList.remove("A11"); }
-    if (paramList.isType<Teko::LinearOp>   ("A12"))         { thA12      = paramList.get<Teko::LinearOp>   ("A12");         paramList.remove("A12"); }
-    if (paramList.isType<Teko::LinearOp>   ("A21"))         { thA21      = paramList.get<Teko::LinearOp>   ("A21");         paramList.remove("A21"); }
-    if (paramList.isType<Teko::LinearOp>   ("A11_9Pt"))     { thA11_9Pt  = paramList.get<Teko::LinearOp>   ("A11_9Pt");     paramList.remove("A11_9Pt"); }
-
-    typedef MueLu::TpetraOperator<SC,LO,GO,NO> MueLuOperator;
-    const RCP<MueLuOperator> mueluPrecOp = Q2Q1MkPrecond(paramList, velCoords, presCoords, p2vMap, thA11, thA12, thA21, thA11_9Pt);
-
-    const RCP<LinearOpBase<SC> > thyraPrecOp = Thyra::createLinearOp(RCP<TpetraLinOp>(mueluPrecOp));
-    defaultPrec->initializeUnspecified(thyraPrecOp);
+  if (paramList.isType<RCP<MultiVector> >("Velcoords")) {
+    velCoords = paramList.get<RCP<MultiVector> >("Velcoords");
+    paramList.remove("Velcoords");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  uninitializePrec(PreconditionerBase<Scalar> *prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
-    // Check precondition
-    TEUCHOS_ASSERT(prec);
-
-    // Retrieve concrete preconditioner object
-    const Teuchos::Ptr<DefaultPreconditioner<SC> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<SC> *>(prec));
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
-
-    if (fwdOp) {
-      // TODO: Implement properly instead of returning default value
-      *fwdOp = Teuchos::null;
-    }
-
-    if (supportSolveUse) {
-      // TODO: Implement properly instead of returning default value
-      *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
-    }
-
-    defaultPrec->uninitialize();
+  if (paramList.isType<RCP<MultiVector> >("Prescoords")) {
+    presCoords = paramList.get<RCP<MultiVector> >("Prescoords");
+    paramList.remove("Prescoords");
   }
-
-
-  // Overridden from ParameterListAcceptor
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameterList(const RCP<ParameterList>& paramList) {
-    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
-    paramList_ = paramList;
+  if (paramList.isType<ArrayRCP<LO> >("p2vMap")) {
+    p2vMap = paramList.get<ArrayRCP<LO> >("p2vMap");
+    paramList.remove("p2vMap");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getNonconstParameterList() {
-    return paramList_;
+  if (paramList.isType<Teko::LinearOp>("A11")) {
+    thA11 = paramList.get<Teko::LinearOp>("A11");
+    paramList.remove("A11");
+  }
+  if (paramList.isType<Teko::LinearOp>("A12")) {
+    thA12 = paramList.get<Teko::LinearOp>("A12");
+    paramList.remove("A12");
+  }
+  if (paramList.isType<Teko::LinearOp>("A21")) {
+    thA21 = paramList.get<Teko::LinearOp>("A21");
+    paramList.remove("A21");
+  }
+  if (paramList.isType<Teko::LinearOp>("A11_9Pt")) {
+    thA11_9Pt = paramList.get<Teko::LinearOp>("A11_9Pt");
+    paramList.remove("A11_9Pt");
   }
 
+  typedef MueLu::TpetraOperator<SC, LO, GO, NO> MueLuOperator;
+  const RCP<MueLuOperator> mueluPrecOp = Q2Q1MkPrecond(paramList, velCoords, presCoords, p2vMap, thA11, thA12, thA21, thA11_9Pt);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::unsetParameterList() {
-    RCP<ParameterList> savedParamList = paramList_;
-    paramList_ = Teuchos::null;
-    return savedParamList;
-  }
+  const RCP<LinearOpBase<SC> > thyraPrecOp = Thyra::createLinearOp(RCP<TpetraLinOp>(mueluPrecOp));
+  defaultPrec->initializeUnspecified(thyraPrecOp);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getParameterList() const {
-    return paramList_;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    uninitializePrec(PreconditionerBase<Scalar>* prec, RCP<const LinearOpSourceBase<Scalar> >* fwdOp, ESupportSolveUse* supportSolveUse) const {
+  // Check precondition
+  TEUCHOS_ASSERT(prec);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getValidParameters() const {
-    static RCP<const ParameterList> validPL;
+  // Retrieve concrete preconditioner object
+  const Teuchos::Ptr<DefaultPreconditioner<SC> > defaultPrec = Teuchos::ptr(dynamic_cast<DefaultPreconditioner<SC>*>(prec));
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(defaultPrec));
 
-    if (validPL.is_null())
-      validPL = rcp(new ParameterList());
+  if (fwdOp) {
+    // TODO: Implement properly instead of returning default value
+    *fwdOp = Teuchos::null;
+  }
 
-    return validPL;
+  if (supportSolveUse) {
+    // TODO: Implement properly instead of returning default value
+    *supportSolveUse = Thyra::SUPPORT_SOLVE_UNSPECIFIED;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  Q2Q1MkPrecond(const ParameterList& paramList,
-                const RCP<Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& velCoords,
-                const RCP<Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& presCoords,
-                const ArrayRCP<LocalOrdinal>& p2vMap,
-                const Teko::LinearOp& thA11, const Teko::LinearOp& thA12, const Teko::LinearOp& thA21, const Teko::LinearOp& thA11_9Pt) const
-  {
-    using Teuchos::null;
-
-    typedef Tpetra::CrsMatrix           <SC,LO,GO,NO> TP_Crs;
-    typedef Tpetra::Operator            <SC,LO,GO,NO> TP_Op;
-
-    typedef Xpetra::BlockedCrsMatrix    <SC,LO,GO,NO> BlockedCrsMatrix;
-    typedef Xpetra::CrsMatrix           <SC,LO,GO,NO> CrsMatrix;
-    typedef Xpetra::CrsMatrixWrap       <SC,LO,GO,NO> CrsMatrixWrap;
-    typedef Xpetra::MapExtractorFactory <SC,LO,GO,NO> MapExtractorFactory;
-    typedef Xpetra::MapExtractor        <SC,LO,GO,NO> MapExtractor;
-    typedef Xpetra::Map                    <LO,GO,NO> Map;
-    typedef Xpetra::MapFactory             <LO,GO,NO> MapFactory;
-    typedef Xpetra::Matrix              <SC,LO,GO,NO> Matrix;
-    typedef Xpetra::MatrixFactory       <SC,LO,GO,NO> MatrixFactory;
-    typedef Xpetra::StridedMapFactory      <LO,GO,NO> StridedMapFactory;
-
-    typedef MueLu::Hierarchy            <SC,LO,GO,NO> Hierarchy;
-
-    const RCP<const Teuchos::Comm<int> > comm = velCoords->getMap()->getComm();
-
-    // Pull out Tpetra matrices
-    RCP<Thyra::LinearOpBase<SC> > ThNonConstA11     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA11);
-    RCP<Thyra::LinearOpBase<SC> > ThNonConstA21     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA21);
-    RCP<Thyra::LinearOpBase<SC> > ThNonConstA12     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA12);
-    RCP<Thyra::LinearOpBase<SC> > ThNonConstA11_9Pt = rcp_const_cast<Thyra::LinearOpBase<double> >(thA11_9Pt);
-
-    RCP<TP_Op>  TpetA11     = Thyra::TpetraOperatorVectorExtraction<SC,LO,GO,NO>::getTpetraOperator(ThNonConstA11);
-    RCP<TP_Op>  TpetA21     = Thyra::TpetraOperatorVectorExtraction<SC,LO,GO,NO>::getTpetraOperator(ThNonConstA21);
-    RCP<TP_Op>  TpetA12     = Thyra::TpetraOperatorVectorExtraction<SC,LO,GO,NO>::getTpetraOperator(ThNonConstA12);
-    RCP<TP_Op>  TpetA11_9Pt = Thyra::TpetraOperatorVectorExtraction<SC,LO,GO,NO>::getTpetraOperator(ThNonConstA11_9Pt);
-
-    RCP<TP_Crs> TpetCrsA11      = rcp_dynamic_cast<TP_Crs>(TpetA11);
-    RCP<TP_Crs> TpetCrsA21      = rcp_dynamic_cast<TP_Crs>(TpetA21);
-    RCP<TP_Crs> TpetCrsA12      = rcp_dynamic_cast<TP_Crs>(TpetA12);
-    RCP<TP_Crs> TpetCrsA11_9Pt  = rcp_dynamic_cast<TP_Crs>(TpetA11_9Pt);
-
-    RCP<Matrix> A_11      = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA11);
-    RCP<Matrix> tmp_A_21  = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA21); // needs map modification
-    RCP<Matrix> tmp_A_12  = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA12); // needs map modification
-    RCP<Matrix> A_11_9Pt  = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA11_9Pt);
-
-    Xpetra::global_size_t numVel  =     A_11->getRowMap()->getLocalNumElements();
-    Xpetra::global_size_t numPres = tmp_A_21->getRowMap()->getLocalNumElements();
-
-    // Create new A21 with map so that the global indices of the row map starts
-    // from numVel+1 (where numVel is the number of rows in the A11 block)
-    RCP<const Map>          domainMap2  = tmp_A_12->getDomainMap();
-    RCP<const Map>          rangeMap2   = tmp_A_21->getRangeMap();
-    Xpetra::global_size_t   numRows2    = rangeMap2->getLocalNumElements();
-    Xpetra::global_size_t   numCols2    = domainMap2->getLocalNumElements();
-    ArrayView<const GO>     rangeElem2  = rangeMap2->getLocalElementList();
-    ArrayView<const GO>     domainElem2 = domainMap2->getLocalElementList();
-    ArrayView<const GO>     rowElem1    = tmp_A_12->getRowMap()->getLocalElementList();
-    ArrayView<const GO>     colElem1    = tmp_A_21->getColMap()->getLocalElementList();
-
-    Xpetra::UnderlyingLib lib = domainMap2->lib();
-    GO indexBase = domainMap2->getIndexBase();
-
-    Array<GO> newRowElem2(numRows2, 0);
-    for (Xpetra::global_size_t i = 0; i < numRows2; i++)
-      newRowElem2[i] = numVel + rangeElem2[i];
-
-    RCP<const Map> newRangeMap2 = MapFactory::Build(lib, numRows2, newRowElem2, indexBase, comm);
-
-    // maybe should be column map???
-    Array<GO> newColElem2(numCols2, 0);
-    for (Xpetra::global_size_t i = 0; i < numCols2; i++)
-      newColElem2[i] = numVel + domainElem2[i];
-
-    RCP<const Map> newDomainMap2 = MapFactory::Build(lib, numCols2, newColElem2, indexBase, comm);
-
-    RCP<Matrix>    A_12         = MatrixFactory::Build(tmp_A_12->getRangeMap(), newDomainMap2,            tmp_A_12->getLocalMaxNumRowEntries());
-    RCP<Matrix>    A_21         = MatrixFactory::Build(newRangeMap2,            tmp_A_21->getDomainMap(), tmp_A_21->getLocalMaxNumRowEntries());
-
-    RCP<CrsMatrix> A_11_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_11)    ->getCrsMatrix();
-    RCP<CrsMatrix> A_12_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_12)    ->getCrsMatrix();
-    RCP<CrsMatrix> A_21_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_21)    ->getCrsMatrix();
-    RCP<CrsMatrix> A_11_crs_9Pt = rcp_dynamic_cast<CrsMatrixWrap>(A_11_9Pt)->getCrsMatrix();
+  defaultPrec->uninitialize();
+}
+
+// Overridden from ParameterListAcceptor
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameterList(const RCP<ParameterList>& paramList) {
+  TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(paramList));
+  paramList_ = paramList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNonconstParameterList() {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::unsetParameterList() {
+  RCP<ParameterList> savedParamList = paramList_;
+  paramList_                        = Teuchos::null;
+  return savedParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getParameterList() const {
+  return paramList_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getValidParameters() const {
+  static RCP<const ParameterList> validPL;
+
+  if (validPL.is_null())
+    validPL = rcp(new ParameterList());
+
+  return validPL;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Q2Q1MkPrecond(const ParameterList& paramList,
+                  const RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& velCoords,
+                  const RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& presCoords,
+                  const ArrayRCP<LocalOrdinal>& p2vMap,
+                  const Teko::LinearOp& thA11, const Teko::LinearOp& thA12, const Teko::LinearOp& thA21, const Teko::LinearOp& thA11_9Pt) const {
+  using Teuchos::null;
+
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TP_Crs;
+  typedef Tpetra::Operator<SC, LO, GO, NO> TP_Op;
+
+  typedef Xpetra::BlockedCrsMatrix<SC, LO, GO, NO> BlockedCrsMatrix;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> CrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> CrsMatrixWrap;
+  typedef Xpetra::MapExtractorFactory<SC, LO, GO, NO> MapExtractorFactory;
+  typedef Xpetra::MapExtractor<SC, LO, GO, NO> MapExtractor;
+  typedef Xpetra::Map<LO, GO, NO> Map;
+  typedef Xpetra::MapFactory<LO, GO, NO> MapFactory;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef Xpetra::MatrixFactory<SC, LO, GO, NO> MatrixFactory;
+  typedef Xpetra::StridedMapFactory<LO, GO, NO> StridedMapFactory;
+
+  typedef MueLu::Hierarchy<SC, LO, GO, NO> Hierarchy;
+
+  const RCP<const Teuchos::Comm<int> > comm = velCoords->getMap()->getComm();
+
+  // Pull out Tpetra matrices
+  RCP<Thyra::LinearOpBase<SC> > ThNonConstA11     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA11);
+  RCP<Thyra::LinearOpBase<SC> > ThNonConstA21     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA21);
+  RCP<Thyra::LinearOpBase<SC> > ThNonConstA12     = rcp_const_cast<Thyra::LinearOpBase<double> >(thA12);
+  RCP<Thyra::LinearOpBase<SC> > ThNonConstA11_9Pt = rcp_const_cast<Thyra::LinearOpBase<double> >(thA11_9Pt);
+
+  RCP<TP_Op> TpetA11     = Thyra::TpetraOperatorVectorExtraction<SC, LO, GO, NO>::getTpetraOperator(ThNonConstA11);
+  RCP<TP_Op> TpetA21     = Thyra::TpetraOperatorVectorExtraction<SC, LO, GO, NO>::getTpetraOperator(ThNonConstA21);
+  RCP<TP_Op> TpetA12     = Thyra::TpetraOperatorVectorExtraction<SC, LO, GO, NO>::getTpetraOperator(ThNonConstA12);
+  RCP<TP_Op> TpetA11_9Pt = Thyra::TpetraOperatorVectorExtraction<SC, LO, GO, NO>::getTpetraOperator(ThNonConstA11_9Pt);
+
+  RCP<TP_Crs> TpetCrsA11     = rcp_dynamic_cast<TP_Crs>(TpetA11);
+  RCP<TP_Crs> TpetCrsA21     = rcp_dynamic_cast<TP_Crs>(TpetA21);
+  RCP<TP_Crs> TpetCrsA12     = rcp_dynamic_cast<TP_Crs>(TpetA12);
+  RCP<TP_Crs> TpetCrsA11_9Pt = rcp_dynamic_cast<TP_Crs>(TpetA11_9Pt);
+
+  RCP<Matrix> A_11     = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA11);
+  RCP<Matrix> tmp_A_21 = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA21);  // needs map modification
+  RCP<Matrix> tmp_A_12 = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA12);  // needs map modification
+  RCP<Matrix> A_11_9Pt = MueLu::TpetraCrs_To_XpetraMatrix(TpetCrsA11_9Pt);
+
+  Xpetra::global_size_t numVel  = A_11->getRowMap()->getLocalNumElements();
+  Xpetra::global_size_t numPres = tmp_A_21->getRowMap()->getLocalNumElements();
+
+  // Create new A21 with map so that the global indices of the row map starts
+  // from numVel+1 (where numVel is the number of rows in the A11 block)
+  RCP<const Map> domainMap2       = tmp_A_12->getDomainMap();
+  RCP<const Map> rangeMap2        = tmp_A_21->getRangeMap();
+  Xpetra::global_size_t numRows2  = rangeMap2->getLocalNumElements();
+  Xpetra::global_size_t numCols2  = domainMap2->getLocalNumElements();
+  ArrayView<const GO> rangeElem2  = rangeMap2->getLocalElementList();
+  ArrayView<const GO> domainElem2 = domainMap2->getLocalElementList();
+  ArrayView<const GO> rowElem1    = tmp_A_12->getRowMap()->getLocalElementList();
+  ArrayView<const GO> colElem1    = tmp_A_21->getColMap()->getLocalElementList();
+
+  Xpetra::UnderlyingLib lib = domainMap2->lib();
+  GO indexBase              = domainMap2->getIndexBase();
+
+  Array<GO> newRowElem2(numRows2, 0);
+  for (Xpetra::global_size_t i = 0; i < numRows2; i++)
+    newRowElem2[i] = numVel + rangeElem2[i];
+
+  RCP<const Map> newRangeMap2 = MapFactory::Build(lib, numRows2, newRowElem2, indexBase, comm);
+
+  // maybe should be column map???
+  Array<GO> newColElem2(numCols2, 0);
+  for (Xpetra::global_size_t i = 0; i < numCols2; i++)
+    newColElem2[i] = numVel + domainElem2[i];
+
+  RCP<const Map> newDomainMap2 = MapFactory::Build(lib, numCols2, newColElem2, indexBase, comm);
+
+  RCP<Matrix> A_12 = MatrixFactory::Build(tmp_A_12->getRangeMap(), newDomainMap2, tmp_A_12->getLocalMaxNumRowEntries());
+  RCP<Matrix> A_21 = MatrixFactory::Build(newRangeMap2, tmp_A_21->getDomainMap(), tmp_A_21->getLocalMaxNumRowEntries());
+
+  RCP<CrsMatrix> A_11_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_11)->getCrsMatrix();
+  RCP<CrsMatrix> A_12_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_12)->getCrsMatrix();
+  RCP<CrsMatrix> A_21_crs     = rcp_dynamic_cast<CrsMatrixWrap>(A_21)->getCrsMatrix();
+  RCP<CrsMatrix> A_11_crs_9Pt = rcp_dynamic_cast<CrsMatrixWrap>(A_11_9Pt)->getCrsMatrix();
 
 #if 0
     RCP<Matrix>    A_22         = MatrixFactory::Build(newRangeMap2,            newDomainMap2,            1);
@@ -375,109 +398,109 @@ namespace Thyra {
     A_21_crs->fillComplete(tmp_A_21->getDomainMap(), newRangeMap2);
     A_22_crs->fillComplete(newDomainMap2,            newRangeMap2);
 #else
-    RCP<Matrix>    A_22     = Teuchos::null;
-    RCP<CrsMatrix> A_22_crs = Teuchos::null;
+  RCP<Matrix> A_22        = Teuchos::null;
+  RCP<CrsMatrix> A_22_crs = Teuchos::null;
 
-    ArrayView<const LO> inds;
-    ArrayView<const SC> vals;
-    for (LO row = 0; row < as<LO>(numRows2); ++row) {
-      tmp_A_21->getLocalRowView(row, inds, vals);
+  ArrayView<const LO> inds;
+  ArrayView<const SC> vals;
+  for (LO row = 0; row < as<LO>(numRows2); ++row) {
+    tmp_A_21->getLocalRowView(row, inds, vals);
 
-      size_t nnz = inds.size();
-      Array<GO> newInds(nnz, 0);
-      for (LO colID = 0; colID < as<LO>(nnz); colID++)
-        newInds[colID] = colElem1[inds[colID]];
+    size_t nnz = inds.size();
+    Array<GO> newInds(nnz, 0);
+    for (LO colID = 0; colID < as<LO>(nnz); colID++)
+      newInds[colID] = colElem1[inds[colID]];
 
-      A_21_crs->insertGlobalValues(newRowElem2[row], newInds, vals);
-    }
-    A_21_crs->fillComplete(tmp_A_21->getDomainMap(), newRangeMap2);
+    A_21_crs->insertGlobalValues(newRowElem2[row], newInds, vals);
+  }
+  A_21_crs->fillComplete(tmp_A_21->getDomainMap(), newRangeMap2);
 #endif
 
-    // Create new A12 with map so that the global indices of the ColMap starts
-    // from numVel+1 (where numVel is the number of rows in the A11 block)
-    for (LO row = 0; row < as<LO>(tmp_A_12->getRowMap()->getLocalNumElements()); ++row) {
-      tmp_A_12->getLocalRowView(row, inds, vals);
+  // Create new A12 with map so that the global indices of the ColMap starts
+  // from numVel+1 (where numVel is the number of rows in the A11 block)
+  for (LO row = 0; row < as<LO>(tmp_A_12->getRowMap()->getLocalNumElements()); ++row) {
+    tmp_A_12->getLocalRowView(row, inds, vals);
 
-      size_t nnz = inds.size();
-      Array<GO> newInds(nnz, 0);
-      for (LO colID = 0; colID < as<LO>(nnz); colID++)
-        newInds[colID] = newColElem2[inds[colID]];
+    size_t nnz = inds.size();
+    Array<GO> newInds(nnz, 0);
+    for (LO colID = 0; colID < as<LO>(nnz); colID++)
+      newInds[colID] = newColElem2[inds[colID]];
 
-      A_12_crs->insertGlobalValues(rowElem1[row], newInds, vals);
-    }
-    A_12_crs->fillComplete(newDomainMap2, tmp_A_12->getRangeMap());
-
-    RCP<Matrix> A_12_abs = Absolute(*A_12);
-    RCP<Matrix> A_21_abs = Absolute(*A_21);
-
-    // =========================================================================
-    // Preconditioner construction - I (block)
-    // =========================================================================
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
-    out.setOutputToRootOnly(0);
-    RCP<Matrix> BBt     = Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*A_21,     false, *A_12,     false, out);
-    RCP<Matrix> BBt_abs = Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*A_21_abs, false, *A_12_abs, false, out);
-
-    SC dropTol = (paramList.get<int>("useFilters") ? paramList.get<double>("tau_1") : 0.00);
-    RCP<Matrix> filteredA = FilterMatrix(*A_11, *A_11,    dropTol);
-    RCP<Matrix> filteredB = FilterMatrix(*BBt,  *BBt_abs, dropTol);
-
-    RCP<Matrix> fA_11_crs = rcp_dynamic_cast<CrsMatrixWrap>(filteredA);
-    RCP<Matrix> fA_12_crs = Teuchos::null;
-    RCP<Matrix> fA_21_crs = Teuchos::null;
-    RCP<Matrix> fA_22_crs = rcp_dynamic_cast<CrsMatrixWrap>(filteredB);
-
-    // Build the large filtered matrix which requires strided maps
-    std::vector<size_t> stridingInfo(1, 1);
-    int stridedBlockId = -1;
-
-    Array<GO> elementList(numVel+numPres); // Not RCP ...  does this get cleared ?
-    Array<GO> velElem  = A_12_crs->getRangeMap()->getLocalElementList();
-    Array<GO> presElem = A_21_crs->getRangeMap()->getLocalElementList();
-
-    for (Xpetra::global_size_t i =    0  ; i < numVel;         i++) elementList[i] = velElem[i];
-    for (Xpetra::global_size_t i = numVel; i < numVel+numPres; i++) elementList[i] = presElem[i-numVel];
-    RCP<const Map> fullMap = StridedMapFactory::Build(Xpetra::UseTpetra, numVel+numPres, elementList(), indexBase, stridingInfo, comm);
-
-    std::vector<RCP<const Map> > partMaps(2);
-    partMaps[0] = StridedMapFactory::Build(Xpetra::UseTpetra, numVel,  velElem,  indexBase, stridingInfo, comm);
-    partMaps[1] = StridedMapFactory::Build(Xpetra::UseTpetra, numPres, presElem, indexBase, stridingInfo, comm, stridedBlockId, numVel);
-
-    // Map extractors are necessary for Xpetra's block operators
-    RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(fullMap, partMaps);
-    RCP<BlockedCrsMatrix> fA = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
-    fA->setMatrix(0, 0, fA_11_crs);
-    fA->setMatrix(0, 1, fA_12_crs);
-    fA->setMatrix(1, 0, fA_21_crs);
-    fA->setMatrix(1, 1, fA_22_crs);
-    fA->fillComplete();
-
-    // -------------------------------------------------------------------------
-    // Preconditioner construction - I.a (filtered hierarchy)
-    // -------------------------------------------------------------------------
-    MueLu::FactoryManager<SC,LO,GO,NO> M;
-    SetDependencyTree(M, paramList);
-
-    RCP<Hierarchy> H = rcp(new Hierarchy);
-    RCP<MueLu::Level> finestLevel = H->GetLevel(0);
-    finestLevel->Set("A",                     rcp_dynamic_cast<Matrix>(fA));
-    finestLevel->Set("p2vMap",                p2vMap);
-    finestLevel->Set("CoordinatesVelocity",   Xpetra::toXpetra(velCoords));
-    finestLevel->Set("CoordinatesPressure",   Xpetra::toXpetra(presCoords));
-    finestLevel->Set("AForPat",               A_11_9Pt);
-    H->SetMaxCoarseSize(MUELU_GPD("coarse: max size", int, 1));
-
-    // The first invocation of Setup() builds the hierarchy using the filtered
-    // matrix. This build includes the grid transfers but not the creation of the
-    // smoothers.
-    // NOTE: we need to indicate what should be kept from the first invocation
-    // for the second invocation, which then focuses on building the smoothers
-    // for the unfiltered matrix.
-    H->Keep("P",     M.GetFactory("P")    .get());
-    H->Keep("R",     M.GetFactory("R")    .get());
-    H->Keep("Ptent", M.GetFactory("Ptent").get());
-    H->Setup(M, 0, MUELU_GPD("max levels", int, 3));
+    A_12_crs->insertGlobalValues(rowElem1[row], newInds, vals);
+  }
+  A_12_crs->fillComplete(newDomainMap2, tmp_A_12->getRangeMap());
+
+  RCP<Matrix> A_12_abs = Absolute(*A_12);
+  RCP<Matrix> A_21_abs = Absolute(*A_21);
+
+  // =========================================================================
+  // Preconditioner construction - I (block)
+  // =========================================================================
+  RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  Teuchos::FancyOStream& out       = *fancy;
+  out.setOutputToRootOnly(0);
+  RCP<Matrix> BBt     = Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*A_21, false, *A_12, false, out);
+  RCP<Matrix> BBt_abs = Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*A_21_abs, false, *A_12_abs, false, out);
+
+  SC dropTol            = (paramList.get<int>("useFilters") ? paramList.get<double>("tau_1") : 0.00);
+  RCP<Matrix> filteredA = FilterMatrix(*A_11, *A_11, dropTol);
+  RCP<Matrix> filteredB = FilterMatrix(*BBt, *BBt_abs, dropTol);
+
+  RCP<Matrix> fA_11_crs = rcp_dynamic_cast<CrsMatrixWrap>(filteredA);
+  RCP<Matrix> fA_12_crs = Teuchos::null;
+  RCP<Matrix> fA_21_crs = Teuchos::null;
+  RCP<Matrix> fA_22_crs = rcp_dynamic_cast<CrsMatrixWrap>(filteredB);
+
+  // Build the large filtered matrix which requires strided maps
+  std::vector<size_t> stridingInfo(1, 1);
+  int stridedBlockId = -1;
+
+  Array<GO> elementList(numVel + numPres);  // Not RCP ...  does this get cleared ?
+  Array<GO> velElem  = A_12_crs->getRangeMap()->getLocalElementList();
+  Array<GO> presElem = A_21_crs->getRangeMap()->getLocalElementList();
+
+  for (Xpetra::global_size_t i = 0; i < numVel; i++) elementList[i] = velElem[i];
+  for (Xpetra::global_size_t i = numVel; i < numVel + numPres; i++) elementList[i] = presElem[i - numVel];
+  RCP<const Map> fullMap = StridedMapFactory::Build(Xpetra::UseTpetra, numVel + numPres, elementList(), indexBase, stridingInfo, comm);
+
+  std::vector<RCP<const Map> > partMaps(2);
+  partMaps[0] = StridedMapFactory::Build(Xpetra::UseTpetra, numVel, velElem, indexBase, stridingInfo, comm);
+  partMaps[1] = StridedMapFactory::Build(Xpetra::UseTpetra, numPres, presElem, indexBase, stridingInfo, comm, stridedBlockId, numVel);
+
+  // Map extractors are necessary for Xpetra's block operators
+  RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(fullMap, partMaps);
+  RCP<BlockedCrsMatrix> fA             = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
+  fA->setMatrix(0, 0, fA_11_crs);
+  fA->setMatrix(0, 1, fA_12_crs);
+  fA->setMatrix(1, 0, fA_21_crs);
+  fA->setMatrix(1, 1, fA_22_crs);
+  fA->fillComplete();
+
+  // -------------------------------------------------------------------------
+  // Preconditioner construction - I.a (filtered hierarchy)
+  // -------------------------------------------------------------------------
+  MueLu::FactoryManager<SC, LO, GO, NO> M;
+  SetDependencyTree(M, paramList);
+
+  RCP<Hierarchy> H              = rcp(new Hierarchy);
+  RCP<MueLu::Level> finestLevel = H->GetLevel(0);
+  finestLevel->Set("A", rcp_dynamic_cast<Matrix>(fA));
+  finestLevel->Set("p2vMap", p2vMap);
+  finestLevel->Set("CoordinatesVelocity", Xpetra::toXpetra(velCoords));
+  finestLevel->Set("CoordinatesPressure", Xpetra::toXpetra(presCoords));
+  finestLevel->Set("AForPat", A_11_9Pt);
+  H->SetMaxCoarseSize(MUELU_GPD("coarse: max size", int, 1));
+
+  // The first invocation of Setup() builds the hierarchy using the filtered
+  // matrix. This build includes the grid transfers but not the creation of the
+  // smoothers.
+  // NOTE: we need to indicate what should be kept from the first invocation
+  // for the second invocation, which then focuses on building the smoothers
+  // for the unfiltered matrix.
+  H->Keep("P", M.GetFactory("P").get());
+  H->Keep("R", M.GetFactory("R").get());
+  H->Keep("Ptent", M.GetFactory("Ptent").get());
+  H->Setup(M, 0, MUELU_GPD("max levels", int, 3));
 
 #if 0
     for (int i = 1; i < H->GetNumLevels(); i++) {
@@ -491,393 +514,390 @@ namespace Thyra {
     }
 #endif
 
-    // -------------------------------------------------------------------------
-    // Preconditioner construction - I.b (smoothers for unfiltered matrix)
-    // -------------------------------------------------------------------------
-    std::string   smootherType   = MUELU_GPD("smoother: type", std::string, "vanka");
-    ParameterList smootherParams;
-    if (paramList.isSublist("smoother: params"))
-      smootherParams = paramList.sublist("smoother: params");
-    M.SetFactory("Smoother", GetSmoother(smootherType, smootherParams, false/*coarseSolver?*/));
-
-    std::string   coarseType   = MUELU_GPD("coarse: type", std::string, "direct");
-    ParameterList coarseParams;
-    if (paramList.isSublist("coarse: params"))
-      coarseParams = paramList.sublist("coarse: params");
-    M.SetFactory("CoarseSolver", GetSmoother(coarseType, coarseParams, true/*coarseSolver?*/));
+  // -------------------------------------------------------------------------
+  // Preconditioner construction - I.b (smoothers for unfiltered matrix)
+  // -------------------------------------------------------------------------
+  std::string smootherType = MUELU_GPD("smoother: type", std::string, "vanka");
+  ParameterList smootherParams;
+  if (paramList.isSublist("smoother: params"))
+    smootherParams = paramList.sublist("smoother: params");
+  M.SetFactory("Smoother", GetSmoother(smootherType, smootherParams, false /*coarseSolver?*/));
+
+  std::string coarseType = MUELU_GPD("coarse: type", std::string, "direct");
+  ParameterList coarseParams;
+  if (paramList.isSublist("coarse: params"))
+    coarseParams = paramList.sublist("coarse: params");
+  M.SetFactory("CoarseSolver", GetSmoother(coarseType, coarseParams, true /*coarseSolver?*/));
 
 #ifdef HAVE_MUELU_DEBUG
-    M.ResetDebugData();
+  M.ResetDebugData();
 #endif
 
-    RCP<BlockedCrsMatrix> A = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
-    A->setMatrix(0, 0, A_11);
-    A->setMatrix(0, 1, A_12);
-    A->setMatrix(1, 0, A_21);
-    A->setMatrix(1, 1, A_22);
-    A->fillComplete();
+  RCP<BlockedCrsMatrix> A = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
+  A->setMatrix(0, 0, A_11);
+  A->setMatrix(0, 1, A_12);
+  A->setMatrix(1, 0, A_21);
+  A->setMatrix(1, 1, A_22);
+  A->fillComplete();
 
-    H->GetLevel(0)->Set("A", rcp_dynamic_cast<Matrix>(A));
+  H->GetLevel(0)->Set("A", rcp_dynamic_cast<Matrix>(A));
 
-    H->Setup(M, 0, H->GetNumLevels());
+  H->Setup(M, 0, H->GetNumLevels());
 
-    return rcp(new MueLu::TpetraOperator<SC,LO,GO,NO>(H));
-  }
+  return rcp(new MueLu::TpetraOperator<SC, LO, GO, NO>(H));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  FilterMatrix(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Pattern, Scalar dropTol) const {
-    typedef Xpetra::Matrix<SC,LO,GO,NO>             Matrix;
-    typedef MueLu::AmalgamationFactory<SC,LO,GO,NO> AmalgamationFactory;
-    typedef MueLu::CoalesceDropFactory<SC,LO,GO,NO> CoalesceDropFactory;
-    typedef MueLu::FilteredAFactory<SC,LO,GO,NO>    FilteredAFactory;
-    typedef MueLu::GraphBase<LO,GO,NO>              GraphBase;
-
-    RCP<GraphBase> filteredGraph;
-    {
-      // Get graph pattern for the pattern matrix
-      MueLu::Level level;
-      level.SetLevelID(1);
-
-      level.Set<RCP<Matrix> >("A", rcpFromRef(Pattern));
-
-      RCP<AmalgamationFactory> amalgFactory = rcp(new AmalgamationFactory());
-
-      RCP<CoalesceDropFactory> dropFactory = rcp(new CoalesceDropFactory());
-      ParameterList dropParams = *(dropFactory->GetValidParameterList());
-      dropParams.set("lightweight wrap",          true);
-      dropParams.set("aggregation: drop scheme",  "classical");
-      dropParams.set("aggregation: drop tol",     dropTol);
-      // dropParams.set("Dirichlet detection threshold", <>);
-      dropFactory->SetParameterList(dropParams);
-      dropFactory->SetFactory("UnAmalgamationInfo", amalgFactory);
-
-      // Build
-      level.Request("Graph", dropFactory.get());
-      dropFactory->Build(level);
-
-      level.Get("Graph", filteredGraph, dropFactory.get());
-    }
-
-    RCP<Matrix> filteredA;
-    {
-      // Filter the original matrix, not the pattern one
-      MueLu::Level level;
-      level.SetLevelID(1);
-
-      level.Set("A",         rcpFromRef(A));
-      level.Set("Graph",     filteredGraph);
-      level.Set("Filtering", true);
-
-      RCP<FilteredAFactory> filterFactory = rcp(new FilteredAFactory());
-      ParameterList filterParams = *(filterFactory->GetValidParameterList());
-      // We need a graph that has proper structure in it. Therefore, we need to
-      // drop older pattern, i.e. not to reuse it
-      filterParams.set("filtered matrix: reuse graph", false);
-      filterParams.set("filtered matrix: use lumping", false);
-      filterFactory->SetParameterList(filterParams);
-
-      // Build
-      level.Request("A", filterFactory.get());
-      filterFactory->Build(level);
-
-      level.Get("A", filteredA, filterFactory.get());
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FilterMatrix(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Pattern, Scalar dropTol) const {
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef MueLu::AmalgamationFactory<SC, LO, GO, NO> AmalgamationFactory;
+  typedef MueLu::CoalesceDropFactory<SC, LO, GO, NO> CoalesceDropFactory;
+  typedef MueLu::FilteredAFactory<SC, LO, GO, NO> FilteredAFactory;
+  typedef MueLu::GraphBase<LO, GO, NO> GraphBase;
 
-    // Zero out row sums by fixing the diagonal
-    filteredA->resumeFill();
-    size_t numRows = filteredA->getRowMap()->getLocalNumElements();
-    for (size_t i = 0; i < numRows; i++) {
-      ArrayView<const LO> inds;
-      ArrayView<const SC> vals;
-      filteredA->getLocalRowView(i, inds, vals);
+  RCP<GraphBase> filteredGraph;
+  {
+    // Get graph pattern for the pattern matrix
+    MueLu::Level level;
+    level.SetLevelID(1);
 
-      size_t nnz = inds.size();
+    level.Set<RCP<Matrix> >("A", rcpFromRef(Pattern));
 
-      Array<SC> valsNew = vals;
+    RCP<AmalgamationFactory> amalgFactory = rcp(new AmalgamationFactory());
 
-      LO diagIndex = -1;
-      SC diag = Teuchos::ScalarTraits<SC>::zero();
-      for (size_t j = 0; j < nnz; j++) {
-        diag += vals[j];
-        if (inds[j] == Teuchos::as<int>(i))
-          diagIndex = j;
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION(diagIndex == -1, MueLu::Exceptions::RuntimeError,
-                                 "No diagonal found");
-      if (nnz <= 1)
-        continue;
+    RCP<CoalesceDropFactory> dropFactory = rcp(new CoalesceDropFactory());
+    ParameterList dropParams             = *(dropFactory->GetValidParameterList());
+    dropParams.set("lightweight wrap", true);
+    dropParams.set("aggregation: drop scheme", "classical");
+    dropParams.set("aggregation: drop tol", dropTol);
+    // dropParams.set("Dirichlet detection threshold", <>);
+    dropFactory->SetParameterList(dropParams);
+    dropFactory->SetFactory("UnAmalgamationInfo", amalgFactory);
 
-      valsNew[diagIndex] -= diag;
+    // Build
+    level.Request("Graph", dropFactory.get());
+    dropFactory->Build(level);
 
-      filteredA->replaceLocalValues(i, inds, valsNew);
-    }
-    filteredA->fillComplete();
+    level.Get("Graph", filteredGraph, dropFactory.get());
+  }
 
-    return filteredA;
+  RCP<Matrix> filteredA;
+  {
+    // Filter the original matrix, not the pattern one
+    MueLu::Level level;
+    level.SetLevelID(1);
+
+    level.Set("A", rcpFromRef(A));
+    level.Set("Graph", filteredGraph);
+    level.Set("Filtering", true);
+
+    RCP<FilteredAFactory> filterFactory = rcp(new FilteredAFactory());
+    ParameterList filterParams          = *(filterFactory->GetValidParameterList());
+    // We need a graph that has proper structure in it. Therefore, we need to
+    // drop older pattern, i.e. not to reuse it
+    filterParams.set("filtered matrix: reuse graph", false);
+    filterParams.set("filtered matrix: use lumping", false);
+    filterFactory->SetParameterList(filterParams);
+
+    // Build
+    level.Request("A", filterFactory.get());
+    filterFactory->Build(level);
+
+    level.Get("A", filteredA, filterFactory.get());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  SetDependencyTree(MueLu::FactoryManager<Scalar,LocalOrdinal,GlobalOrdinal,Node>& M, const ParameterList& paramList) const {
-    typedef MueLu::BlockedPFactory      <SC,LO,GO,NO> BlockedPFactory;
-    typedef MueLu::GenericRFactory      <SC,LO,GO,NO> GenericRFactory;
-    typedef MueLu::BlockedRAPFactory    <SC,LO,GO,NO> BlockedRAPFactory;
-    typedef MueLu::SmootherFactory      <SC,LO,GO,NO> SmootherFactory;
-    typedef MueLu::BlockedDirectSolver  <SC,LO,GO,NO> BlockedDirectSolver;
-    typedef MueLu::FactoryManager       <SC,LO,GO,NO> FactoryManager;
-
-    // Pressure and velocity dependency trees are identical. The only
-    // difference is that pressure has to go first, so that velocity can use
-    // some of pressure data
-    RCP<FactoryManager> M11 = rcp(new FactoryManager()), M22 = rcp(new FactoryManager());
-    M11->SetKokkosRefactor(paramList.get<bool>("use kokkos refactor"));
-    M22->SetKokkosRefactor(paramList.get<bool>("use kokkos refactor"));
-    SetBlockDependencyTree(*M11, 0, 0, "velocity", paramList);
-    SetBlockDependencyTree(*M22, 1, 1, "pressure", paramList);
-
-    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
-    ParameterList pParamList = *(PFact->GetValidParameterList());
-    pParamList.set("backwards", true);      // do pressure first
-    PFact->SetParameterList(pParamList);
-    PFact->AddFactoryManager(M11);
-    PFact->AddFactoryManager(M22);
-    M.SetFactory("P", PFact);
+  // Zero out row sums by fixing the diagonal
+  filteredA->resumeFill();
+  size_t numRows = filteredA->getRowMap()->getLocalNumElements();
+  for (size_t i = 0; i < numRows; i++) {
+    ArrayView<const LO> inds;
+    ArrayView<const SC> vals;
+    filteredA->getLocalRowView(i, inds, vals);
+
+    size_t nnz = inds.size();
 
-    RCP<GenericRFactory> RFact = rcp(new GenericRFactory());
-    RFact->SetFactory("P", PFact);
-    M.SetFactory("R", RFact);
+    Array<SC> valsNew = vals;
 
-    RCP<MueLu::Factory > AcFact = rcp(new BlockedRAPFactory());
-    AcFact->SetFactory("R", RFact);
-    AcFact->SetFactory("P", PFact);
-    M.SetFactory("A", AcFact);
+    LO diagIndex = -1;
+    SC diag      = Teuchos::ScalarTraits<SC>::zero();
+    for (size_t j = 0; j < nnz; j++) {
+      diag += vals[j];
+      if (inds[j] == Teuchos::as<int>(i))
+        diagIndex = j;
+    }
+    TEUCHOS_TEST_FOR_EXCEPTION(diagIndex == -1, MueLu::Exceptions::RuntimeError,
+                               "No diagonal found");
+    if (nnz <= 1)
+      continue;
 
-    // Smoothers will be set later
-    M.SetFactory("Smoother", Teuchos::null);
+    valsNew[diagIndex] -= diag;
 
-    RCP<MueLu::Factory> coarseFact = rcp(new SmootherFactory(rcp(new BlockedDirectSolver()), Teuchos::null));
-    // M.SetFactory("CoarseSolver", coarseFact);
-    M.SetFactory("CoarseSolver", Teuchos::null);
+    filteredA->replaceLocalValues(i, inds, valsNew);
   }
+  filteredA->fillComplete();
+
+  return filteredA;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SetDependencyTree(MueLu::FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>& M, const ParameterList& paramList) const {
+  typedef MueLu::BlockedPFactory<SC, LO, GO, NO> BlockedPFactory;
+  typedef MueLu::GenericRFactory<SC, LO, GO, NO> GenericRFactory;
+  typedef MueLu::BlockedRAPFactory<SC, LO, GO, NO> BlockedRAPFactory;
+  typedef MueLu::SmootherFactory<SC, LO, GO, NO> SmootherFactory;
+  typedef MueLu::BlockedDirectSolver<SC, LO, GO, NO> BlockedDirectSolver;
+  typedef MueLu::FactoryManager<SC, LO, GO, NO> FactoryManager;
+
+  // Pressure and velocity dependency trees are identical. The only
+  // difference is that pressure has to go first, so that velocity can use
+  // some of pressure data
+  RCP<FactoryManager> M11 = rcp(new FactoryManager()), M22 = rcp(new FactoryManager());
+  M11->SetKokkosRefactor(paramList.get<bool>("use kokkos refactor"));
+  M22->SetKokkosRefactor(paramList.get<bool>("use kokkos refactor"));
+  SetBlockDependencyTree(*M11, 0, 0, "velocity", paramList);
+  SetBlockDependencyTree(*M22, 1, 1, "pressure", paramList);
+
+  RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
+  ParameterList pParamList   = *(PFact->GetValidParameterList());
+  pParamList.set("backwards", true);  // do pressure first
+  PFact->SetParameterList(pParamList);
+  PFact->AddFactoryManager(M11);
+  PFact->AddFactoryManager(M22);
+  M.SetFactory("P", PFact);
+
+  RCP<GenericRFactory> RFact = rcp(new GenericRFactory());
+  RFact->SetFactory("P", PFact);
+  M.SetFactory("R", RFact);
+
+  RCP<MueLu::Factory> AcFact = rcp(new BlockedRAPFactory());
+  AcFact->SetFactory("R", RFact);
+  AcFact->SetFactory("P", PFact);
+  M.SetFactory("A", AcFact);
+
+  // Smoothers will be set later
+  M.SetFactory("Smoother", Teuchos::null);
+
+  RCP<MueLu::Factory> coarseFact = rcp(new SmootherFactory(rcp(new BlockedDirectSolver()), Teuchos::null));
+  // M.SetFactory("CoarseSolver", coarseFact);
+  M.SetFactory("CoarseSolver", Teuchos::null);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SetBlockDependencyTree(MueLu::FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>& M, LocalOrdinal row, LocalOrdinal col, const std::string& mode, const ParameterList& paramList) const {
+  typedef MueLu::ConstraintFactory<SC, LO, GO, NO> ConstraintFactory;
+  typedef MueLu::EminPFactory<SC, LO, GO, NO> EminPFactory;
+  typedef MueLu::GenericRFactory<SC, LO, GO, NO> GenericRFactory;
+  typedef MueLu::PatternFactory<SC, LO, GO, NO> PatternFactory;
+  typedef MueLu::Q2Q1PFactory<SC, LO, GO, NO> Q2Q1PFactory;
+  typedef MueLu::Q2Q1uPFactory<SC, LO, GO, NO> Q2Q1uPFactory;
+  typedef MueLu::SubBlockAFactory<SC, LO, GO, NO> SubBlockAFactory;
+
+  RCP<SubBlockAFactory> AFact = rcp(new SubBlockAFactory());
+  AFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  AFact->SetParameter("block row", Teuchos::ParameterEntry(row));
+  AFact->SetParameter("block col", Teuchos::ParameterEntry(col));
+  M.SetFactory("A", AFact);
+
+  RCP<MueLu::Factory> Q2Q1Fact;
+
+  const bool isStructured = false;
+
+  if (isStructured) {
+    Q2Q1Fact = rcp(new Q2Q1PFactory);
+
+  } else {
+    Q2Q1Fact                    = rcp(new Q2Q1uPFactory);
+    ParameterList q2q1ParamList = *(Q2Q1Fact->GetValidParameterList());
+    q2q1ParamList.set("mode", mode);
+    if (paramList.isParameter("dump status"))
+      q2q1ParamList.set("dump status", paramList.get<bool>("dump status"));
+    if (paramList.isParameter("phase2"))
+      q2q1ParamList.set("phase2", paramList.get<bool>("phase2"));
+    if (paramList.isParameter("tau_2"))
+      q2q1ParamList.set("tau_2", paramList.get<double>("tau_2"));
+    Q2Q1Fact->SetParameterList(q2q1ParamList);
+  }
+  Q2Q1Fact->SetFactory("A", AFact);
+  M.SetFactory("Ptent", Q2Q1Fact);
+
+  RCP<PatternFactory> patternFact = rcp(new PatternFactory);
+  ParameterList patternParams     = *(patternFact->GetValidParameterList());
+  // Our prolongator constructs the exact pattern we are going to use,
+  // therefore we do not expand it
+  patternParams.set("emin: pattern order", 0);
+  patternFact->SetParameterList(patternParams);
+  patternFact->SetFactory("A", AFact);
+  patternFact->SetFactory("P", Q2Q1Fact);
+  M.SetFactory("Ppattern", patternFact);
+
+  RCP<ConstraintFactory> CFact = rcp(new ConstraintFactory);
+  CFact->SetFactory("Ppattern", patternFact);
+  M.SetFactory("Constraint", CFact);
+
+  RCP<EminPFactory> EminPFact = rcp(new EminPFactory());
+  ParameterList eminParams    = *(EminPFact->GetValidParameterList());
+  if (paramList.isParameter("emin: num iterations"))
+    eminParams.set("emin: num iterations", paramList.get<int>("emin: num iterations"));
+  if (mode == "pressure") {
+    eminParams.set("emin: iterative method", "cg");
+  } else {
+    eminParams.set("emin: iterative method", "gmres");
+    if (paramList.isParameter("emin: iterative method"))
+      eminParams.set("emin: iterative method", paramList.get<std::string>("emin: iterative method"));
+  }
+  EminPFact->SetParameterList(eminParams);
+  EminPFact->SetFactory("A", AFact);
+  EminPFact->SetFactory("Constraint", CFact);
+  EminPFact->SetFactory("P", Q2Q1Fact);
+  M.SetFactory("P", EminPFact);
+
+  if (mode == "velocity" && (!paramList.isParameter("velocity: use transpose") || paramList.get<bool>("velocity: use transpose") == false)) {
+    // Pressure system is symmetric, so it does not matter
+    // Velocity system may benefit from running emin in restriction mode (with A^T)
+    RCP<GenericRFactory> RFact = rcp(new GenericRFactory());
+    RFact->SetFactory("P", EminPFact);
+    M.SetFactory("R", RFact);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::FactoryBase>
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetSmoother(const std::string& type, const ParameterList& paramList, bool coarseSolver) const {
+  typedef Teuchos::ParameterEntry ParameterEntry;
+
+  typedef MueLu::BlockedDirectSolver<SC, LO, GO, NO> BlockedDirectSolver;
+  typedef MueLu::BraessSarazinSmoother<SC, LO, GO, NO> BraessSarazinSmoother;
+  typedef MueLu::DirectSolver<SC, LO, GO, NO> DirectSolver;
+  typedef MueLu::FactoryManager<SC, LO, GO, NO> FactoryManager;
+  typedef MueLu::SchurComplementFactory<SC, LO, GO, NO> SchurComplementFactory;
+  typedef MueLu::SmootherFactory<SC, LO, GO, NO> SmootherFactory;
+  typedef MueLu::SmootherPrototype<SC, LO, GO, NO> SmootherPrototype;
+  typedef MueLu::TrilinosSmoother<SC, LO, GO, NO> TrilinosSmoother;
+
+  RCP<SmootherPrototype> smootherPrototype;
+  if (type == "none") {
+    return Teuchos::null;
+
+  } else if (type == "vanka") {
+    // Set up Vanka smoothing via a combination of Schwarz and block relaxation.
+    ParameterList schwarzList;
+    schwarzList.set("schwarz: overlap level", as<int>(0));
+    schwarzList.set("schwarz: zero starting solution", false);
+    schwarzList.set("subdomain solver name", "Block_Relaxation");
+
+    ParameterList& innerSolverList = schwarzList.sublist("subdomain solver parameters");
+    innerSolverList.set("partitioner: type", "user");
+    innerSolverList.set("partitioner: overlap", MUELU_GPD("partitioner: overlap", int, 1));
+    innerSolverList.set("relaxation: type", MUELU_GPD("relaxation: type", std::string, "Gauss-Seidel"));
+    innerSolverList.set("relaxation: sweeps", MUELU_GPD("relaxation: sweeps", int, 1));
+    innerSolverList.set("relaxation: damping factor", MUELU_GPD("relaxation: damping factor", double, 0.5));
+    innerSolverList.set("relaxation: zero starting solution", false);
+    // innerSolverList.set("relaxation: backward mode",          MUELU_GPD("relaxation: backward mode",  bool,           true);  NOT SUPPORTED YET
+
+    std::string ifpackType = "SCHWARZ";
+
+    smootherPrototype = rcp(new TrilinosSmoother(ifpackType, schwarzList));
+
+  } else if (type == "schwarz") {
+    std::string ifpackType = "SCHWARZ";
+
+    smootherPrototype = rcp(new TrilinosSmoother(ifpackType, paramList));
+
+  } else if (type == "braess-sarazin") {
+    // Define smoother/solver for BraessSarazin
+    SC omega     = MUELU_GPD("bs: omega", double, 1.0);
+    bool lumping = MUELU_GPD("bs: lumping", bool, false);
+
+    RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
+    schurFact->SetParameter("omega", ParameterEntry(omega));
+    schurFact->SetParameter("lumping", ParameterEntry(lumping));
+    schurFact->SetFactory("A", MueLu::NoFactory::getRCP());
+
+    // Schur complement solver
+    RCP<SmootherPrototype> schurSmootherPrototype;
+    std::string schurSmootherType = (paramList.isParameter("schur smoother: type") ? paramList.get<std::string>("schur smoother: type") : "RELAXATION");
+    if (schurSmootherType == "RELAXATION") {
+      ParameterList schurSmootherParams = paramList.sublist("schur smoother: params");
+      // schurSmootherParams.set("relaxation: damping factor", omega);
+      schurSmootherPrototype = rcp(new TrilinosSmoother(schurSmootherType, schurSmootherParams));
+    } else {
+      schurSmootherPrototype = rcp(new DirectSolver());
+    }
+    schurSmootherPrototype->SetFactory("A", schurFact);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  SetBlockDependencyTree(MueLu::FactoryManager<Scalar,LocalOrdinal,GlobalOrdinal,Node>& M, LocalOrdinal row, LocalOrdinal col, const std::string& mode, const ParameterList& paramList) const {
-    typedef MueLu::ConstraintFactory <SC,LO,GO,NO> ConstraintFactory;
-    typedef MueLu::EminPFactory      <SC,LO,GO,NO> EminPFactory;
-    typedef MueLu::GenericRFactory   <SC,LO,GO,NO> GenericRFactory;
-    typedef MueLu::PatternFactory    <SC,LO,GO,NO> PatternFactory;
-    typedef MueLu::Q2Q1PFactory      <SC,LO,GO,NO> Q2Q1PFactory;
-    typedef MueLu::Q2Q1uPFactory     <SC,LO,GO,NO> Q2Q1uPFactory;
-    typedef MueLu::SubBlockAFactory  <SC,LO,GO,NO> SubBlockAFactory;
+    RCP<SmootherFactory> schurSmootherFact = rcp(new SmootherFactory(schurSmootherPrototype));
 
-    RCP<SubBlockAFactory> AFact = rcp(new SubBlockAFactory());
-    AFact->SetFactory  ("A",         MueLu::NoFactory::getRCP());
-    AFact->SetParameter("block row", Teuchos::ParameterEntry(row));
-    AFact->SetParameter("block col", Teuchos::ParameterEntry(col));
-    M.SetFactory("A", AFact);
+    // Define temporary FactoryManager that is used as input for BraessSarazin smoother
+    RCP<FactoryManager> braessManager = rcp(new FactoryManager());
+    braessManager->SetFactory("A", schurFact);                 // SchurComplement operator for correction step (defined as "A")
+    braessManager->SetFactory("Smoother", schurSmootherFact);  // solver/smoother for correction step
+    braessManager->SetFactory("PreSmoother", schurSmootherFact);
+    braessManager->SetFactory("PostSmoother", schurSmootherFact);
+    braessManager->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
-    RCP<MueLu::Factory> Q2Q1Fact;
+    smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetParameter("Sweeps", ParameterEntry(MUELU_GPD("bs: sweeps", int, 1)));
+    smootherPrototype->SetParameter("lumping", ParameterEntry(lumping));
+    smootherPrototype->SetParameter("Damping factor", ParameterEntry(omega));
+    smootherPrototype->SetParameter("q2q1 mode", ParameterEntry(true));
+    rcp_dynamic_cast<BraessSarazinSmoother>(smootherPrototype)->AddFactoryManager(braessManager, 0);  // set temporary factory manager in BraessSarazin smoother
 
-    const bool isStructured = false;
+  } else if (type == "ilu") {
+    std::string ifpackType = "RILUK";
 
-    if (isStructured) {
-      Q2Q1Fact = rcp(new Q2Q1PFactory);
+    smootherPrototype = rcp(new TrilinosSmoother(ifpackType, paramList));
 
-    } else {
-      Q2Q1Fact = rcp(new Q2Q1uPFactory);
-      ParameterList q2q1ParamList = *(Q2Q1Fact->GetValidParameterList());
-      q2q1ParamList.set("mode", mode);
-      if (paramList.isParameter("dump status"))
-        q2q1ParamList.set("dump status", paramList.get<bool>("dump status"));
-      if (paramList.isParameter("phase2"))
-        q2q1ParamList.set("phase2", paramList.get<bool>("phase2"));
-      if (paramList.isParameter("tau_2"))
-        q2q1ParamList.set("tau_2", paramList.get<double>("tau_2"));
-      Q2Q1Fact->SetParameterList(q2q1ParamList);
-    }
-    Q2Q1Fact->SetFactory("A", AFact);
-    M.SetFactory("Ptent", Q2Q1Fact);
-
-    RCP<PatternFactory> patternFact = rcp(new PatternFactory);
-    ParameterList patternParams = *(patternFact->GetValidParameterList());
-    // Our prolongator constructs the exact pattern we are going to use,
-    // therefore we do not expand it
-    patternParams.set("emin: pattern order", 0);
-    patternFact->SetParameterList(patternParams);
-    patternFact->SetFactory("A", AFact);
-    patternFact->SetFactory("P", Q2Q1Fact);
-    M.SetFactory("Ppattern", patternFact);
-
-    RCP<ConstraintFactory> CFact = rcp(new ConstraintFactory);
-    CFact->SetFactory("Ppattern", patternFact);
-    M.SetFactory("Constraint", CFact);
-
-    RCP<EminPFactory> EminPFact = rcp(new EminPFactory());
-    ParameterList eminParams = *(EminPFact->GetValidParameterList());
-    if (paramList.isParameter("emin: num iterations"))
-      eminParams.set("emin: num iterations", paramList.get<int>("emin: num iterations"));
-    if (mode == "pressure") {
-      eminParams.set("emin: iterative method", "cg");
-    } else {
-      eminParams.set("emin: iterative method", "gmres");
-      if (paramList.isParameter("emin: iterative method"))
-        eminParams.set("emin: iterative method", paramList.get<std::string>("emin: iterative method"));
-    }
-    EminPFact->SetParameterList(eminParams);
-    EminPFact->SetFactory("A",          AFact);
-    EminPFact->SetFactory("Constraint", CFact);
-    EminPFact->SetFactory("P",          Q2Q1Fact);
-    M.SetFactory("P", EminPFact);
-
-    if (mode == "velocity" && (!paramList.isParameter("velocity: use transpose") || paramList.get<bool>("velocity: use transpose") == false)) {
-      // Pressure system is symmetric, so it does not matter
-      // Velocity system may benefit from running emin in restriction mode (with A^T)
-      RCP<GenericRFactory> RFact = rcp(new GenericRFactory());
-      RFact->SetFactory("P", EminPFact);
-      M.SetFactory("R", RFact);
-    }
+  } else if (type == "direct") {
+    smootherPrototype = rcp(new BlockedDirectSolver());
+
+  } else {
+    throw MueLu::Exceptions::RuntimeError("Unknown smoother type: \"" + type + "\"");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::FactoryBase>
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  GetSmoother(const std::string& type, const ParameterList& paramList, bool coarseSolver) const {
-    typedef Teuchos::ParameterEntry                   ParameterEntry;
-
-    typedef MueLu::BlockedDirectSolver   <SC,LO,GO,NO> BlockedDirectSolver;
-    typedef MueLu::BraessSarazinSmoother <SC,LO,GO,NO> BraessSarazinSmoother;
-    typedef MueLu::DirectSolver          <SC,LO,GO,NO> DirectSolver;
-    typedef MueLu::FactoryManager        <SC,LO,GO,NO> FactoryManager;
-    typedef MueLu::SchurComplementFactory<SC,LO,GO,NO> SchurComplementFactory;
-    typedef MueLu::SmootherFactory       <SC,LO,GO,NO> SmootherFactory;
-    typedef MueLu::SmootherPrototype     <SC,LO,GO,NO> SmootherPrototype;
-    typedef MueLu::TrilinosSmoother      <SC,LO,GO,NO> TrilinosSmoother;
-
-    RCP<SmootherPrototype> smootherPrototype;
-    if (type == "none") {
-      return Teuchos::null;
-
-    } else if (type == "vanka") {
-      // Set up Vanka smoothing via a combination of Schwarz and block relaxation.
-      ParameterList schwarzList;
-      schwarzList.set("schwarz: overlap level",                 as<int>(0));
-      schwarzList.set("schwarz: zero starting solution",        false);
-      schwarzList.set("subdomain solver name",                  "Block_Relaxation");
-
-      ParameterList& innerSolverList = schwarzList.sublist("subdomain solver parameters");
-      innerSolverList.set("partitioner: type",                  "user");
-      innerSolverList.set("partitioner: overlap",               MUELU_GPD("partitioner: overlap",       int,            1));
-      innerSolverList.set("relaxation: type",                   MUELU_GPD("relaxation: type",           std::string,    "Gauss-Seidel"));
-      innerSolverList.set("relaxation: sweeps",                 MUELU_GPD("relaxation: sweeps",         int,            1));
-      innerSolverList.set("relaxation: damping factor",         MUELU_GPD("relaxation: damping factor", double,         0.5));
-      innerSolverList.set("relaxation: zero starting solution", false);
-      // innerSolverList.set("relaxation: backward mode",          MUELU_GPD("relaxation: backward mode",  bool,           true);  NOT SUPPORTED YET
-
-      std::string ifpackType = "SCHWARZ";
-
-      smootherPrototype = rcp(new TrilinosSmoother(ifpackType, schwarzList));
-
-    } else if (type == "schwarz") {
-
-      std::string ifpackType = "SCHWARZ";
-
-      smootherPrototype = rcp(new TrilinosSmoother(ifpackType, paramList));
-
-    } else if (type == "braess-sarazin") {
-      // Define smoother/solver for BraessSarazin
-      SC   omega   = MUELU_GPD("bs: omega",     double, 1.0);
-      bool lumping = MUELU_GPD("bs: lumping",   bool,   false);
-
-      RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
-      schurFact->SetParameter("omega",      ParameterEntry(omega));
-      schurFact->SetParameter("lumping",    ParameterEntry(lumping));
-      schurFact->SetFactory  ("A",          MueLu::NoFactory::getRCP());
-
-      // Schur complement solver
-      RCP<SmootherPrototype> schurSmootherPrototype;
-      std::string schurSmootherType = (paramList.isParameter("schur smoother: type") ? paramList.get<std::string>("schur smoother: type") : "RELAXATION");
-      if (schurSmootherType == "RELAXATION") {
-        ParameterList schurSmootherParams = paramList.sublist("schur smoother: params");
-        // schurSmootherParams.set("relaxation: damping factor", omega);
-        schurSmootherPrototype = rcp(new TrilinosSmoother(schurSmootherType, schurSmootherParams));
-      } else {
-        schurSmootherPrototype = rcp(new DirectSolver());
-      }
-      schurSmootherPrototype->SetFactory("A", schurFact);
-
-      RCP<SmootherFactory> schurSmootherFact = rcp(new SmootherFactory(schurSmootherPrototype));
-
-      // Define temporary FactoryManager that is used as input for BraessSarazin smoother
-      RCP<FactoryManager> braessManager = rcp(new FactoryManager());
-      braessManager->SetFactory("A",            schurFact);             // SchurComplement operator for correction step (defined as "A")
-      braessManager->SetFactory("Smoother",     schurSmootherFact);     // solver/smoother for correction step
-      braessManager->SetFactory("PreSmoother",  schurSmootherFact);
-      braessManager->SetFactory("PostSmoother", schurSmootherFact);
-      braessManager->SetIgnoreUserData(true);                           // always use data from factories defined in factory manager
-
-      smootherPrototype = rcp(new BraessSarazinSmoother());
-      smootherPrototype->SetParameter("Sweeps",         ParameterEntry(MUELU_GPD("bs: sweeps", int, 1)));
-      smootherPrototype->SetParameter("lumping",        ParameterEntry(lumping));
-      smootherPrototype->SetParameter("Damping factor", ParameterEntry(omega));
-      smootherPrototype->SetParameter("q2q1 mode",      ParameterEntry(true));
-      rcp_dynamic_cast<BraessSarazinSmoother>(smootherPrototype)->AddFactoryManager(braessManager, 0);   // set temporary factory manager in BraessSarazin smoother
-
-    } else if (type == "ilu") {
-      std::string ifpackType = "RILUK";
-
-      smootherPrototype = rcp(new TrilinosSmoother(ifpackType, paramList));
-
-    } else if (type == "direct") {
-      smootherPrototype = rcp(new BlockedDirectSolver());
+  return coarseSolver ? rcp(new SmootherFactory(smootherPrototype, Teuchos::null)) : rcp(new SmootherFactory(smootherPrototype));
+}
 
-    } else {
-      throw MueLu::Exceptions::RuntimeError("Unknown smoother type: \"" + type + "\"");
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Absolute(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A) const {
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> CrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> CrsMatrixWrap;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
 
-    return coarseSolver ? rcp(new SmootherFactory(smootherPrototype, Teuchos::null)) : rcp(new SmootherFactory(smootherPrototype));
-  }
+  const CrsMatrixWrap& Awrap = dynamic_cast<const CrsMatrixWrap&>(A);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Absolute(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A) const {
-    typedef Xpetra::CrsMatrix    <SC,LO,GO,NO> CrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO> CrsMatrixWrap;
-    typedef Xpetra::Matrix       <SC,LO,GO,NO> Matrix;
-
-    const CrsMatrixWrap& Awrap = dynamic_cast<const CrsMatrixWrap&>(A);
-
-    ArrayRCP<const size_t> iaA;
-    ArrayRCP<const LO>     jaA;
-    ArrayRCP<const SC>     valA;
-    Awrap.getCrsMatrix()->getAllValues(iaA, jaA, valA);
-
-    ArrayRCP<size_t> iaB (iaA .size());
-    ArrayRCP<LO>     jaB (jaA .size());
-    ArrayRCP<SC>     valB(valA.size());
-    for (int i = 0; i < iaA .size(); i++) iaB [i] = iaA[i];
-    for (int i = 0; i < jaA .size(); i++) jaB [i] = jaA[i];
-    for (int i = 0; i < valA.size(); i++) valB[i] = Teuchos::ScalarTraits<SC>::magnitude(valA[i]);
-
-    RCP<Matrix> B = rcp(new CrsMatrixWrap(A.getRowMap(), A.getColMap(), 0));
-    RCP<CrsMatrix> Bcrs = rcp_dynamic_cast<CrsMatrixWrap>(B)->getCrsMatrix();
-    Bcrs->setAllValues(iaB, jaB, valB);
-    Bcrs->expertStaticFillComplete(A.getDomainMap(), A.getRangeMap());
-
-    return B;
-  }
+  ArrayRCP<const size_t> iaA;
+  ArrayRCP<const LO> jaA;
+  ArrayRCP<const SC> valA;
+  Awrap.getCrsMatrix()->getAllValues(iaA, jaA, valA);
 
-  // Public functions overridden from Teuchos::Describable
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MueLuTpetraQ2Q1PreconditionerFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::description() const {
-    return "Thyra::MueLuTpetraQ2Q1PreconditionerFactory";
-  }
+  ArrayRCP<size_t> iaB(iaA.size());
+  ArrayRCP<LO> jaB(jaA.size());
+  ArrayRCP<SC> valB(valA.size());
+  for (int i = 0; i < iaA.size(); i++) iaB[i] = iaA[i];
+  for (int i = 0; i < jaA.size(); i++) jaB[i] = jaA[i];
+  for (int i = 0; i < valA.size(); i++) valB[i] = Teuchos::ScalarTraits<SC>::magnitude(valA[i]);
+
+  RCP<Matrix> B       = rcp(new CrsMatrixWrap(A.getRowMap(), A.getColMap(), 0));
+  RCP<CrsMatrix> Bcrs = rcp_dynamic_cast<CrsMatrixWrap>(B)->getCrsMatrix();
+  Bcrs->setAllValues(iaB, jaB, valB);
+  Bcrs->expertStaticFillComplete(A.getDomainMap(), A.getRangeMap());
+
+  return B;
+}
+
+// Public functions overridden from Teuchos::Describable
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MueLuTpetraQ2Q1PreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  return "Thyra::MueLuTpetraQ2Q1PreconditionerFactory";
+}
 
-} // namespace Thyra
+}  // namespace Thyra
 
 #endif
-#endif // ifdef THYRA_MUELU_TPETRA_Q2Q1PRECONDITIONER_FACTORY_DEF_HPP
+#endif  // ifdef THYRA_MUELU_TPETRA_Q2Q1PRECONDITIONER_FACTORY_DEF_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_decl.hpp b/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_decl.hpp
index ac0fcb2c3d4e..4ab8c289873b 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_decl.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_decl.hpp
@@ -51,23 +51,19 @@
 #include "Xpetra_Operator.hpp"
 #include "Teuchos_ConstNonconstObjectContainer.hpp"
 
-
 namespace Thyra {
 
-
 /** \brief Concrete Thyra::LinearOpBase subclass for Xpetra::Operator.
  *
  * \todo Move this to Thyra??
  *
  * \ingroup Xpetra_Thyra_Op_Vec_adapters_grp
  */
-template <class Scalar, class LocalOrdinal, class GlobalOrdinal=LocalOrdinal,
-  class Node=Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal = LocalOrdinal,
+          class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
 class XpetraLinearOp
-  : virtual public Thyra::LinearOpDefaultBase<Scalar>
-{
-public:
-
+  : virtual public Thyra::LinearOpDefaultBase<Scalar> {
+ public:
   /** \name Constructors/initializers. */
   //@{
 
@@ -76,24 +72,22 @@ class XpetraLinearOp
 
   /** \brief Initialize. */
   void initialize(
-    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-    const RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-    );
+      const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+      const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+      const RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator);
 
   /** \brief Initialize. */
   void constInitialize(
-    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-    const RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-    );
+      const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+      const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+      const RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator);
 
   /** \brief Get embedded non-const Xpetra::Operator. */
-  RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
+  RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   getXpetraOperator();
 
   /** \brief Get embedded const Xpetra::Operator. */
-  RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
+  RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   getConstXpetraOperator() const;
 
   //@}
@@ -109,8 +103,7 @@ class XpetraLinearOp
 
   //@}
 
-protected:
-
+ protected:
   /** \name Protected Overridden functions from LinearOpBase. */
   //@{
 
@@ -119,36 +112,31 @@ class XpetraLinearOp
 
   /** \brief . */
   void applyImpl(
-    const Thyra::EOpTransp M_trans,
-    const Thyra::MultiVectorBase<Scalar> &X_in,
-    const Teuchos::Ptr<Thyra::MultiVectorBase<Scalar> > &Y_inout,
-    const Scalar alpha,
-    const Scalar beta
-    ) const;
+      const Thyra::EOpTransp M_trans,
+      const Thyra::MultiVectorBase<Scalar> &X_in,
+      const Teuchos::Ptr<Thyra::MultiVectorBase<Scalar> > &Y_inout,
+      const Scalar alpha,
+      const Scalar beta) const;
 
   //@}
 
-private:
-
+ private:
   RCP<const VectorSpaceBase<Scalar> >
-  rangeSpace_;
+      rangeSpace_;
 
   RCP<const VectorSpaceBase<Scalar> >
-  domainSpace_;
+      domainSpace_;
 
-  Teuchos::ConstNonconstObjectContainer<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  xpetraOperator_;
+  Teuchos::ConstNonconstObjectContainer<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+      xpetraOperator_;
 
-  template<class XpetraOperator_t>
+  template <class XpetraOperator_t>
   void initializeImpl(
-    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-    const RCP<XpetraOperator_t> &xpetraOperator
-  );
-
+      const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+      const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+      const RCP<XpetraOperator_t> &xpetraOperator);
 };
 
-
 /** \brief Nonmmeber constructor for XpetraLinearOp.
  *
  * \relates XpetraLinearOp
@@ -156,18 +144,15 @@ class XpetraLinearOp
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
 xpetraLinearOp(
-  const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-  const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-  const RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-  )
-{
+    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+    const RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator) {
   const RCP<XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op =
-    Teuchos::rcp(new XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>);
+      Teuchos::rcp(new XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>);
   op->initialize(rangeSpace, domainSpace, xpetraOperator);
   return op;
 }
 
-
 /** \brief Nonmmeber constructor for XpetraLinearOp.
  *
  * \relates XpetraLinearOp
@@ -175,17 +160,15 @@ xpetraLinearOp(
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
 constXpetraLinearOp(
-  const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-  const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-  const RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-  )
-{
+    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+    const RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator) {
   const RCP<XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op =
-    Teuchos::rcp(new XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>);
+      Teuchos::rcp(new XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>);
   op->constInitialize(rangeSpace, domainSpace, xpetraOperator);
   return op;
 }
 
 }  // namespace Thyra
 
-#endif // THYRA_XPETRA_LINEAR_OP_DECL_HPP
+#endif  // THYRA_XPETRA_LINEAR_OP_DECL_HPP
diff --git a/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_def.hpp b/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_def.hpp
index 9a7517e1e280..05f1cf52a364 100644
--- a/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_def.hpp
+++ b/packages/muelu/adapters/stratimikos/Thyra_XpetraLinearOp_def.hpp
@@ -56,77 +56,58 @@
 
 namespace Thyra {
 
-
 // Constructors/initializers
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::XpetraLinearOp()
-{}
-
+XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::XpetraLinearOp() {}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::initialize(
-  const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-  const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-  const RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-  )
-{
+void XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::initialize(
+    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+    const RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator) {
   initializeImpl(rangeSpace, domainSpace, xpetraOperator);
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::constInitialize(
-  const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-  const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-  const RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &xpetraOperator
-  )
-{
+void XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::constInitialize(
+    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+    const RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &xpetraOperator) {
   initializeImpl(rangeSpace, domainSpace, xpetraOperator);
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getXpetraOperator()
-{
+RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getXpetraOperator() {
   return xpetraOperator_.getNonconstObj();
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getConstXpetraOperator() const
-{
+RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getConstXpetraOperator() const {
   return xpetraOperator_;
 }
 
-
 // Public Overridden functions from LinearOpBase
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const Thyra::VectorSpaceBase<Scalar> >
-XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::range() const
-{
+XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::range() const {
   return rangeSpace_;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const Thyra::VectorSpaceBase<Scalar> >
-XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::domain() const
-{
+XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::domain() const {
   return domainSpace_;
 }
 
 // Protected Overridden functions from LinearOpBase
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-bool XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::opSupportedImpl(
-  Thyra::EOpTransp M_trans) const
-{
+bool XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::opSupportedImpl(
+    Thyra::EOpTransp M_trans) const {
   if (is_null(xpetraOperator_))
     return false;
 
@@ -142,30 +123,27 @@ bool XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::opSupportedImpl(
   return xpetraOperator_->hasTransposeApply();
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyImpl(
-  const Thyra::EOpTransp M_trans,
-  const Thyra::MultiVectorBase<Scalar> &X_in,
-  const Teuchos::Ptr<Thyra::MultiVectorBase<Scalar> > &Y_inout,
-  const Scalar alpha,
-  const Scalar beta
-  ) const
-{
-  using Teuchos::rcpFromRef;
+void XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::applyImpl(
+    const Thyra::EOpTransp M_trans,
+    const Thyra::MultiVectorBase<Scalar> &X_in,
+    const Teuchos::Ptr<Thyra::MultiVectorBase<Scalar> > &Y_inout,
+    const Scalar alpha,
+    const Scalar beta) const {
   using Teuchos::rcpFromPtr;
+  using Teuchos::rcpFromRef;
 
   TEUCHOS_TEST_FOR_EXCEPTION(getConstXpetraOperator() == Teuchos::null, MueLu::Exceptions::RuntimeError, "XpetraLinearOp::applyImpl: internal Xpetra::Operator is null.");
-  RCP< const Teuchos::Comm< int > > comm = getConstXpetraOperator()->getRangeMap()->getComm();
+  RCP<const Teuchos::Comm<int> > comm = getConstXpetraOperator()->getRangeMap()->getComm();
 
-  const RCP<const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tX_in =
-      Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(rcpFromRef(X_in), comm);
-  RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tY_inout =
-      Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(rcpFromPtr(Y_inout), comm);
+  const RCP<const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tX_in =
+      Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(rcpFromRef(X_in), comm);
+  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tY_inout =
+      Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(rcpFromPtr(Y_inout), comm);
   Teuchos::ETransp transp;
   switch (M_trans) {
-    case NOTRANS:   transp = Teuchos::NO_TRANS;   break;
-    case TRANS:     transp = Teuchos::TRANS;      break;
+    case NOTRANS: transp = Teuchos::NO_TRANS; break;
+    case TRANS: transp = Teuchos::TRANS; break;
     case CONJTRANS: transp = Teuchos::CONJ_TRANS; break;
     default: TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::NotImplemented, "Thyra::XpetraLinearOp::apply. Unknown value for M_trans. Only NOTRANS, TRANS and CONJTRANS are supported.");
   }
@@ -173,25 +151,25 @@ void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyImpl(
   xpetraOperator_->apply(*tX_in, *tY_inout, transp, alpha, beta);
 
   // check whether Y is a product vector
-  RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal,Node> > rgMapExtractor = Teuchos::null;
+  RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rgMapExtractor = Teuchos::null;
   Teuchos::Ptr<Thyra::ProductMultiVectorBase<Scalar> > prodY_inout =
       Teuchos::ptr_dynamic_cast<Thyra::ProductMultiVectorBase<Scalar> >(Y_inout);
-  if(prodY_inout != Teuchos::null) {
+  if (prodY_inout != Teuchos::null) {
     // If Y is a product vector we split up the data from tY and merge them
     // into the product vector. The necessary Xpetra::MapExtractor is extracted
     // from the fine level operator (not this!)
 
     // get underlying fine level operator (BlockedCrsMatrix)
     // to extract the range MapExtractor
-    RCP<MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal,Node> > mueXop =
-        Teuchos::rcp_dynamic_cast<MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal,Node> >(xpetraOperator_.getNonconstObj());
+    RCP<MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > mueXop =
+        Teuchos::rcp_dynamic_cast<MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(xpetraOperator_.getNonconstObj());
 
-    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal,Node> > A =
-        mueXop->GetHierarchy()->GetLevel(0)->template Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal,Node> > >("A");
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A =
+        mueXop->GetHierarchy()->GetLevel(0)->template Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("A");
     TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(A));
 
-    RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal,Node> > bA =
-        Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal,Node> >(A);
+    RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bA =
+        Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
     TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(bA));
 
     rgMapExtractor = bA->getRangeMapExtractor();
@@ -199,30 +177,24 @@ void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyImpl(
   }
 }
 
-
 // private
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-template<class XpetraOperator_t>
-void XpetraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>::initializeImpl(
-  const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
-  const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
-  const RCP<XpetraOperator_t> &xpetraOperator
-  )
-{
+template <class XpetraOperator_t>
+void XpetraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>::initializeImpl(
+    const RCP<const VectorSpaceBase<Scalar> > &rangeSpace,
+    const RCP<const VectorSpaceBase<Scalar> > &domainSpace,
+    const RCP<XpetraOperator_t> &xpetraOperator) {
 #ifdef THYRA_DEBUG
   TEUCHOS_ASSERT(nonnull(rangeSpace));
   TEUCHOS_ASSERT(nonnull(domainSpace));
   TEUCHOS_ASSERT(nonnull(xpetraOperator));
 #endif
-  rangeSpace_ = rangeSpace;
-  domainSpace_ = domainSpace;
+  rangeSpace_     = rangeSpace;
+  domainSpace_    = domainSpace;
   xpetraOperator_ = xpetraOperator;
 }
 
-
-} // namespace Thyra
-
+}  // namespace Thyra
 
 #endif  // THYRA_XPETRA_LINEAR_OP_HPP
diff --git a/packages/muelu/adapters/stratimikos/Xpetra_ThyraLinearOp.hpp b/packages/muelu/adapters/stratimikos/Xpetra_ThyraLinearOp.hpp
index 0ffa0e9e9cbc..60d8261c7063 100644
--- a/packages/muelu/adapters/stratimikos/Xpetra_ThyraLinearOp.hpp
+++ b/packages/muelu/adapters/stratimikos/Xpetra_ThyraLinearOp.hpp
@@ -61,168 +61,164 @@
 #include <Stratimikos_LinearSolverBuilder.hpp>
 #include <Thyra_MueLuPreconditionerFactory.hpp>
 
-
 namespace MueLu {
 
-  /*!  @brief Wraps an existing Thyra::LinearOp as a Xpetra::Operator.
-*/
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class XpetraThyraLinearOp : public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-  protected:
-    XpetraThyraLinearOp() = default;
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    XpetraThyraLinearOp(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A,
-                        RCP<ParameterList> params) : A_(A) {
-      throw Exceptions::RuntimeError("Interface not supported");
-    };
-
-    //! Destructor.
-    ~XpetraThyraLinearOp() = default;
-
-    //@}
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const {
-      throw Exceptions::RuntimeError("Interface not supported");
-    }
-
-    // //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const {
-      throw Exceptions::RuntimeError("Interface not supported");
-    }
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    /*!
-      \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::one()) const {
-      throw Exceptions::RuntimeError("Interface not supported");
-    }
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const { return false; }
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & X,
-                  const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & B,
-                  Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & R) const {
-      throw Exceptions::RuntimeError("Interface not supported");
-    }
-
-
-  private:
-    RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A_;
+/*!  @brief Wraps an existing Thyra::LinearOp as a Xpetra::Operator.
+ */
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class XpetraThyraLinearOp : public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ protected:
+  XpetraThyraLinearOp() = default;
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  XpetraThyraLinearOp(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> A,
+                      RCP<ParameterList> params)
+    : A_(A) {
+    throw Exceptions::RuntimeError("Interface not supported");
   };
 
-
-  // Partial specialization for Scalar == double.
-  // Allows to avoid issues with Stokhos instantiating Thyra objects.
-  template <class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class XpetraThyraLinearOp<double,LocalOrdinal,GlobalOrdinal,Node> : public Xpetra::Operator<double,LocalOrdinal,GlobalOrdinal,Node> {
-
-    using Scalar = double;
-
-  protected:
-    XpetraThyraLinearOp() = default;
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    XpetraThyraLinearOp(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A,
-                        RCP<ParameterList> params) : A_(A) {
-      // Build Thyra linear algebra objects
-      RCP<const Thyra::LinearOpBase<Scalar> > thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>>(A)->getCrsMatrix());
-
-      Stratimikos::LinearSolverBuilder<Scalar> linearSolverBuilder;
-      typedef Thyra::PreconditionerFactoryBase<Scalar>                                     Base;
-      typedef Thyra::MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> ImplMueLu;
-      linearSolverBuilder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, ImplMueLu>(), "MueLu");
-
-      linearSolverBuilder.setParameterList(params);
-
-      // Build a new "solver factory" according to the previously specified parameter list.
-      // RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
-
-      auto precFactory = Thyra::createPreconditioningStrategy(linearSolverBuilder);
-      auto prec = precFactory->createPrec();
-
-      precFactory->initializePrec(Thyra::defaultLinearOpSource(thyraA), prec.get(), Thyra::SUPPORT_SOLVE_UNSPECIFIED);
-      prec_ = prec;
-    };
-
-    //! Destructor.
-    ~XpetraThyraLinearOp() = default;
-
-    //@}
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const {
-      return A_->getDomainMap();
-    }
-
-    // //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const {
-      return A_->getRangeMap();
-    }
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    /*!
-      \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::one()) const {
-
-      RCP<const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rcpX = Teuchos::rcpFromRef(X);
-      RCP<const Thyra::MultiVectorBase<Scalar> > thyraX = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(rcpX);
-
-      RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rcpY = Teuchos::rcpFromRef(Y);
-      RCP<Thyra::MultiVectorBase<Scalar> > thyraY = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(rcpY));
-
-      prec_->getUnspecifiedPrecOp()->apply(Thyra::NOTRANS, *thyraX, thyraY.ptr(), alpha, beta);
-      Y = *Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(thyraY, Y.getMap()->getComm());
-    }
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const { return false; }
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & X,
-                  const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & B,
-                  Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & R) const {
-      using STS = Teuchos::ScalarTraits<Scalar>;
-      R.update(STS::one(),B,STS::zero());
-      this->apply (X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
-    }
-
-
-  private:
-    RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A_;
-    RCP<Thyra::PreconditionerBase<Scalar> > prec_;
+  //! Destructor.
+  ~XpetraThyraLinearOp() = default;
+
+  //@}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> getDomainMap() const {
+    throw Exceptions::RuntimeError("Interface not supported");
+  }
+
+  // //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> getRangeMap() const {
+    throw Exceptions::RuntimeError("Interface not supported");
+  }
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  /*!
+    \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::one()) const {
+    throw Exceptions::RuntimeError("Interface not supported");
+  }
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const { return false; }
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const {
+    throw Exceptions::RuntimeError("Interface not supported");
+  }
+
+ private:
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> A_;
+};
+
+// Partial specialization for Scalar == double.
+// Allows to avoid issues with Stokhos instantiating Thyra objects.
+template <class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class XpetraThyraLinearOp<double, LocalOrdinal, GlobalOrdinal, Node> : public Xpetra::Operator<double, LocalOrdinal, GlobalOrdinal, Node> {
+  using Scalar = double;
+
+ protected:
+  XpetraThyraLinearOp() = default;
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  XpetraThyraLinearOp(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> A,
+                      RCP<ParameterList> params)
+    : A_(A) {
+    // Build Thyra linear algebra objects
+    RCP<const Thyra::LinearOpBase<Scalar>> thyraA = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(A)->getCrsMatrix());
+
+    Stratimikos::LinearSolverBuilder<Scalar> linearSolverBuilder;
+    typedef Thyra::PreconditionerFactoryBase<Scalar> Base;
+    typedef Thyra::MueLuPreconditionerFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> ImplMueLu;
+    linearSolverBuilder.setPreconditioningStrategyFactory(Teuchos::abstractFactoryStd<Base, ImplMueLu>(), "MueLu");
+
+    linearSolverBuilder.setParameterList(params);
+
+    // Build a new "solver factory" according to the previously specified parameter list.
+    // RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
+
+    auto precFactory = Thyra::createPreconditioningStrategy(linearSolverBuilder);
+    auto prec        = precFactory->createPrec();
+
+    precFactory->initializePrec(Thyra::defaultLinearOpSource(thyraA), prec.get(), Thyra::SUPPORT_SOLVE_UNSPECIFIED);
+    prec_ = prec;
   };
 
-} // namespace
-
-#endif // defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-
-#endif // MUELU_XPETRA_THYRALINEAROP_HPP
+  //! Destructor.
+  ~XpetraThyraLinearOp() = default;
+
+  //@}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> getDomainMap() const {
+    return A_->getDomainMap();
+  }
+
+  // //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> getRangeMap() const {
+    return A_->getRangeMap();
+  }
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  /*!
+    \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::one()) const {
+    RCP<const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> rcpX = Teuchos::rcpFromRef(X);
+    RCP<const Thyra::MultiVectorBase<Scalar>> thyraX                               = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(rcpX);
+
+    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> rcpY = Teuchos::rcpFromRef(Y);
+    RCP<Thyra::MultiVectorBase<Scalar>> thyraY                               = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar>>(Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(rcpY));
+
+    prec_->getUnspecifiedPrecOp()->apply(Thyra::NOTRANS, *thyraX, thyraY.ptr(), alpha, beta);
+    Y = *Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(thyraY, Y.getMap()->getComm());
+  }
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const { return false; }
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const {
+    using STS = Teuchos::ScalarTraits<Scalar>;
+    R.update(STS::one(), B, STS::zero());
+    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+  }
+
+ private:
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> A_;
+  RCP<Thyra::PreconditionerBase<Scalar>> prec_;
+};
+
+}  // namespace MueLu
+
+#endif  // defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
+
+#endif  // MUELU_XPETRA_THYRALINEAROP_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_CreateTpetraPreconditioner.hpp b/packages/muelu/adapters/tpetra/MueLu_CreateTpetraPreconditioner.hpp
index e1fe95d2bdc5..90be6d89c4a5 100644
--- a/packages/muelu/adapters/tpetra/MueLu_CreateTpetraPreconditioner.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_CreateTpetraPreconditioner.hpp
@@ -24,7 +24,6 @@
 #include <MueLu_Utilities.hpp>
 #include <MueLu_HierarchyUtils.hpp>
 
-
 #if defined(HAVE_MUELU_AMGX)
 #include <MueLu_AMGXOperator.hpp>
 #include <amgx_c.h>
@@ -33,165 +32,154 @@
 
 namespace MueLu {
 
-
-  /*!
-    @brief Helper function to create a MueLu or AMGX preconditioner that can be used by Tpetra.
-    @ingroup MueLuAdapters
-    Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] inParamList Parameter list
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &inA,
-                             Teuchos::ParameterList& inParamList)
-  {
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-    using   Teuchos::ParameterList;
-
-    typedef Xpetra::MultiVector<SC,LO,GO,NO>            MultiVector;
-    typedef Xpetra::Matrix<SC,LO,GO,NO>                 Matrix;
-    typedef Hierarchy<SC,LO,GO,NO>                      Hierarchy;
-    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
-    typedef Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> block_crs_matrix_type;
+/*!
+  @brief Helper function to create a MueLu or AMGX preconditioner that can be used by Tpetra.
+  @ingroup MueLuAdapters
+  Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] inParamList Parameter list
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
+                           Teuchos::ParameterList& inParamList) {
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+  using Teuchos::ParameterList;
+
+  typedef Xpetra::MultiVector<SC, LO, GO, NO> MultiVector;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef Hierarchy<SC, LO, GO, NO> Hierarchy;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> block_crs_matrix_type;
 
 #if defined(HAVE_MUELU_AMGX)
-    std::string externalMG = "use external multigrid package";
-    if (inParamList.isParameter(externalMG) && inParamList.get<std::string>(externalMG) == "amgx"){
-      const RCP<crs_matrix_type> constCrsA = rcp_dynamic_cast<crs_matrix_type>(inA);
-      TEUCHOS_TEST_FOR_EXCEPTION(constCrsA == Teuchos::null, Exceptions::RuntimeError, "CreateTpetraPreconditioner: failed to dynamic cast to Tpetra::CrsMatrix, which is required to be able to use AmgX.");
-      return rcp(new AMGXOperator<SC,LO,GO,NO>(constCrsA,inParamList));
-    }
-#endif
-
-    // Wrap A
-    RCP<Matrix> A;
-    RCP<block_crs_matrix_type> bcrsA = rcp_dynamic_cast<block_crs_matrix_type>(inA);
-    RCP<crs_matrix_type> crsA = rcp_dynamic_cast<crs_matrix_type>(inA);
-    if (crsA != Teuchos::null)
-      A = TpetraCrs_To_XpetraMatrix<SC,LO,GO,NO>(crsA);
-    else if (bcrsA != Teuchos::null) {
-      RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO>(bcrsA));
-      TEUCHOS_TEST_FOR_EXCEPTION(temp==Teuchos::null, Exceptions::RuntimeError, "CreateTpetraPreconditioner: cast from Tpetra::BlockCrsMatrix to Xpetra::TpetraBlockCrsMatrix failed.");
-      A = rcp(new Xpetra::CrsMatrixWrap<SC,LO,GO,NO>(temp));
-    }
-    else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "CreateTpetraPreconditioner: only Tpetra CrsMatrix and BlockCrsMatrix types are supported.");
-    }
-
-    Teuchos::ParameterList& userList = inParamList.sublist("user data");
-    if (userList.isParameter("Coordinates")) {
-      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> > coordinates = Teuchos::null;
-      try {
-        coordinates = TpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>(userList.get<RCP<Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >("Coordinates"));
-      } catch(Teuchos::Exceptions::InvalidParameterType&) {
-        coordinates = userList.get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >("Coordinates");
-      }
-      userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> > >("Coordinates", coordinates);
-    }
-
-    if (userList.isParameter("Nullspace")) {
-      RCP<MultiVector> nullspace = Teuchos::null;
-      try {
-        nullspace = TpetraMultiVector_To_XpetraMultiVector<SC,LO,GO,NO>(userList.get<RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("Nullspace"));
-      } catch(Teuchos::Exceptions::InvalidParameterType&) {
-        nullspace = userList.get<RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("Nullspace");
-      }
-      userList.set<RCP<MultiVector> >("Nullspace", nullspace);
-    }
-
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A, inParamList);
-    return rcp(new TpetraOperator<SC,LO,GO,NO>(H));
+  std::string externalMG = "use external multigrid package";
+  if (inParamList.isParameter(externalMG) && inParamList.get<std::string>(externalMG) == "amgx") {
+    const RCP<crs_matrix_type> constCrsA = rcp_dynamic_cast<crs_matrix_type>(inA);
+    TEUCHOS_TEST_FOR_EXCEPTION(constCrsA == Teuchos::null, Exceptions::RuntimeError, "CreateTpetraPreconditioner: failed to dynamic cast to Tpetra::CrsMatrix, which is required to be able to use AmgX.");
+    return rcp(new AMGXOperator<SC, LO, GO, NO>(constCrsA, inParamList));
   }
+#endif
 
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Tpetra.
-    @ingroup MueLuAdapters
-
-    Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
-
-    @param[in] inA Matrix
-    @param[in] xmlFileName XML file containing MueLu options
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
-                             const std::string& xmlFileName)
-  {
-    Teuchos::ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *inA->getDomainMap()->getComm());
-    return CreateTpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(inA, paramList);
-  }
-
-
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Tpetra.
-    @ingroup MueLuAdapters
-
-    Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
-
-    @param[in] inA Matrix
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA)
-  {
-    Teuchos::ParameterList paramList;
-    return CreateTpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(inA, paramList);
+  // Wrap A
+  RCP<Matrix> A;
+  RCP<block_crs_matrix_type> bcrsA = rcp_dynamic_cast<block_crs_matrix_type>(inA);
+  RCP<crs_matrix_type> crsA        = rcp_dynamic_cast<crs_matrix_type>(inA);
+  if (crsA != Teuchos::null)
+    A = TpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(crsA);
+  else if (bcrsA != Teuchos::null) {
+    RCP<Xpetra::CrsMatrix<SC, LO, GO, NO> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO>(bcrsA));
+    TEUCHOS_TEST_FOR_EXCEPTION(temp == Teuchos::null, Exceptions::RuntimeError, "CreateTpetraPreconditioner: cast from Tpetra::BlockCrsMatrix to Xpetra::TpetraBlockCrsMatrix failed.");
+    A = rcp(new Xpetra::CrsMatrixWrap<SC, LO, GO, NO>(temp));
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "CreateTpetraPreconditioner: only Tpetra CrsMatrix and BlockCrsMatrix types are supported.");
   }
 
-
-  /*!
-    @brief Helper function to reuse an existing MueLu preconditioner.
-    @ingroup MueLuAdapters
-
-    @param[in] inA Matrix
-    @param[in] Op  Existing MueLu preconditioner.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReuseTpetraPreconditioner(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
-                                 MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-    typedef Xpetra::Matrix<SC,LO,GO,NO>     Matrix;
-    typedef MueLu ::Hierarchy<SC,LO,GO,NO>  Hierarchy;
-
-    RCP<Hierarchy> H = Op.GetHierarchy();
-    RCP<Matrix>    A = TpetraCrs_To_XpetraMatrix<SC,LO,GO,NO>(inA);
-
-    MueLu::ReuseXpetraPreconditioner<SC,LO,GO,NO>(A, H);
+  Teuchos::ParameterList& userList = inParamList.sublist("user data");
+  if (userList.isParameter("Coordinates")) {
+    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> > coordinates = Teuchos::null;
+    try {
+      coordinates = TpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>(userList.get<RCP<Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >("Coordinates"));
+    } catch (Teuchos::Exceptions::InvalidParameterType&) {
+      coordinates = userList.get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >("Coordinates");
+    }
+    userList.set<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> > >("Coordinates", coordinates);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReuseTpetraPreconditioner(const Teuchos::RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
-                                 MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-    typedef Xpetra::Matrix<SC,LO,GO,NO>     Matrix;
-    typedef MueLu ::Hierarchy<SC,LO,GO,NO>  Hierarchy;
-
-    RCP<Hierarchy> H = Op.GetHierarchy();
-    RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO>(inA));
-    TEUCHOS_TEST_FOR_EXCEPTION(temp==Teuchos::null, Exceptions::RuntimeError, "ReuseTpetraPreconditioner: cast from Tpetra::BlockCrsMatrix to Xpetra::TpetraBlockCrsMatrix failed.");
-    RCP<Matrix> A = rcp(new Xpetra::CrsMatrixWrap<SC,LO,GO,NO>(temp));
-
-    MueLu::ReuseXpetraPreconditioner<SC,LO,GO,NO>(A, H);
+  if (userList.isParameter("Nullspace")) {
+    RCP<MultiVector> nullspace = Teuchos::null;
+    try {
+      nullspace = TpetraMultiVector_To_XpetraMultiVector<SC, LO, GO, NO>(userList.get<RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("Nullspace"));
+    } catch (Teuchos::Exceptions::InvalidParameterType&) {
+      nullspace = userList.get<RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("Nullspace");
+    }
+    userList.set<RCP<MultiVector> >("Nullspace", nullspace);
   }
 
-
-
-} //namespace
-
-#endif //ifndef MUELU_CREATE_TPETRA_PRECONDITIONER_HPP
-
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, inParamList);
+  return rcp(new TpetraOperator<SC, LO, GO, NO>(H));
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Tpetra.
+  @ingroup MueLuAdapters
+
+  Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
+
+  @param[in] inA Matrix
+  @param[in] xmlFileName XML file containing MueLu options
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
+                           const std::string& xmlFileName) {
+  Teuchos::ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *inA->getDomainMap()->getComm());
+  return CreateTpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(inA, paramList);
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Tpetra.
+  @ingroup MueLuAdapters
+
+  Given a Tpetra::Operator, this function returns a constructed MueLu preconditioner.
+
+  @param[in] inA Matrix
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+CreateTpetraPreconditioner(const Teuchos::RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA) {
+  Teuchos::ParameterList paramList;
+  return CreateTpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(inA, paramList);
+}
+
+/*!
+  @brief Helper function to reuse an existing MueLu preconditioner.
+  @ingroup MueLuAdapters
+
+  @param[in] inA Matrix
+  @param[in] Op  Existing MueLu preconditioner.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReuseTpetraPreconditioner(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
+                               MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef MueLu ::Hierarchy<SC, LO, GO, NO> Hierarchy;
+
+  RCP<Hierarchy> H = Op.GetHierarchy();
+  RCP<Matrix> A    = TpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(inA);
+
+  MueLu::ReuseXpetraPreconditioner<SC, LO, GO, NO>(A, H);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReuseTpetraPreconditioner(const Teuchos::RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& inA,
+                               MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef MueLu ::Hierarchy<SC, LO, GO, NO> Hierarchy;
+
+  RCP<Hierarchy> H                             = Op.GetHierarchy();
+  RCP<Xpetra::CrsMatrix<SC, LO, GO, NO> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO>(inA));
+  TEUCHOS_TEST_FOR_EXCEPTION(temp == Teuchos::null, Exceptions::RuntimeError, "ReuseTpetraPreconditioner: cast from Tpetra::BlockCrsMatrix to Xpetra::TpetraBlockCrsMatrix failed.");
+  RCP<Matrix> A = rcp(new Xpetra::CrsMatrixWrap<SC, LO, GO, NO>(temp));
+
+  MueLu::ReuseXpetraPreconditioner<SC, LO, GO, NO>(A, H);
+}
+
+}  // namespace MueLu
+
+#endif  // ifndef MUELU_CREATE_TPETRA_PRECONDITIONER_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_decl.hpp b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_decl.hpp
index dfd15d15f89b..2f7f822cf614 100644
--- a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_decl.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_decl.hpp
@@ -58,106 +58,105 @@
 
 namespace MueLu {
 
-  /*! @brief Wraps an existing MueLu::Hierarchy as a Tpetra::Operator, with an optional two-level correction.
-      Intended to be used with MueLu::ShiftedLaplacian.
+/*! @brief Wraps an existing MueLu::Hierarchy as a Tpetra::Operator, with an optional two-level correction.
+    Intended to be used with MueLu::ShiftedLaplacian.
+*/
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class ShiftedLaplacianOperator
+  : public Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> OP;
+  typedef MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node> MUtils;
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  ShiftedLaplacianOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H)
+    , option_(0) {}
+
+  //! Auxiliary Constructor
+  ShiftedLaplacianOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H,
+                           const RCP<Matrix> A, int cycles, int iters, int option, double tol)
+    : Hierarchy_(H)
+    , A_(A)
+    , cycles_(cycles)
+    , iters_(iters)
+    , option_(option)
+    , tol_(tol) {
+    // setup 2-level correction
+    /*RCP< MueLu::Level > Level1 = H -> GetLevel(1);
+    R_ = Level1 -> Get< RCP<Matrix> >("R");
+    P_ = Level1 -> Get< RCP<Matrix> >("P");
+    //RCP<Matrix> AP = Level1 -> Get< RCP<Matrix> >("AP graph");
+    RCP<Matrix> AP;
+    AP = MUtils::Multiply(*A_, false, *P_, false, AP);
+    // Optimization storage option. If matrix is not changing later, allow this.
+    bool doOptimizedStorage = true;
+    // Reuse coarse matrix memory if available (multiple solve)
+    //RCP<Matrix> Ac = Level1 -> Get< RCP<Matrix> >("RAP graph");
+    RCP<Matrix> Ac;
+    Ac = MUtils::Multiply(*R_, false, *AP, false, Ac, true, doOptimizedStorage);
+    Ac_ = MUtils::Op2NonConstTpetraCrs(Ac);
+
+    // Setup Belos for two-level correction
+    BelosList_ = rcp( new Teuchos::ParameterList("GMRES") );
+    BelosList_ -> set("Maximum Iterations", iters_ );
+    BelosList_ -> set("Convergence Tolerance", tol_ );
+    BelosLP_   = rcp( new Belos::LinearProblem<Scalar,MV,OP> );
+    BelosLP_   -> setOperator ( Ac_ );
+    BelosSM_   = rcp( new Belos::BlockGmresSolMgr<Scalar,MV,OP>(BelosLP_, BelosList_) );*/
+  }
+
+  //! Destructor.
+  virtual ~ShiftedLaplacianOperator() {}
+
+  //@}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const;
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const;
+
+  //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
+  /*!
+    \param[in] X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
+
   */
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class ShiftedLaplacianOperator
-    : public Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-
-    typedef Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>      Matrix;
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>   CrsMatrix;
-    typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> MV;
-    typedef Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>    OP;
-    typedef MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>        MUtils;
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    ShiftedLaplacianOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & H) : Hierarchy_(H), option_(0) { }
-
-    //! Auxiliary Constructor
-    ShiftedLaplacianOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & H,
-                             const RCP<Matrix> A, int cycles, int iters, int option, double tol) : Hierarchy_(H), A_(A), cycles_(cycles), iters_(iters), option_(option), tol_(tol)
-    {
-
-      // setup 2-level correction
-      /*RCP< MueLu::Level > Level1 = H -> GetLevel(1);
-      R_ = Level1 -> Get< RCP<Matrix> >("R");
-      P_ = Level1 -> Get< RCP<Matrix> >("P");
-      //RCP<Matrix> AP = Level1 -> Get< RCP<Matrix> >("AP graph");
-      RCP<Matrix> AP;
-      AP = MUtils::Multiply(*A_, false, *P_, false, AP);
-      // Optimization storage option. If matrix is not changing later, allow this.
-      bool doOptimizedStorage = true;
-      // Reuse coarse matrix memory if available (multiple solve)
-      //RCP<Matrix> Ac = Level1 -> Get< RCP<Matrix> >("RAP graph");
-      RCP<Matrix> Ac;
-      Ac = MUtils::Multiply(*R_, false, *AP, false, Ac, true, doOptimizedStorage);
-      Ac_ = MUtils::Op2NonConstTpetraCrs(Ac);
-
-      // Setup Belos for two-level correction
-      BelosList_ = rcp( new Teuchos::ParameterList("GMRES") );
-      BelosList_ -> set("Maximum Iterations", iters_ );
-      BelosList_ -> set("Convergence Tolerance", tol_ );
-      BelosLP_   = rcp( new Belos::LinearProblem<Scalar,MV,OP> );
-      BelosLP_   -> setOperator ( Ac_ );
-      BelosSM_   = rcp( new Belos::BlockGmresSolMgr<Scalar,MV,OP>(BelosLP_, BelosList_) );*/
-
-    }
-
-
-
-    //! Destructor.
-    virtual ~ShiftedLaplacianOperator() { }
-
-    //@}
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const;
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const;
-
-    //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
-    /*!
-      \param[in] X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
-
-    */
-    void apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                         Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                         Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                                         Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                                         Scalar beta = Teuchos::ScalarTraits<Scalar>::one()) const;
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
-
-  private:
-
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
-    RCP< Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > R_, P_, A_;
-    RCP< Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Ac_;
-    RCP< Teuchos::ParameterList >  BelosList_;
-
-    //RCP< Belos::LinearProblem<Scalar,MV,OP> > BelosLP_;
-    //RCP< Belos::SolverManager<Scalar,MV,OP> > BelosSM_;
-
-    // cycles -> number of V-cycles
-    // iters  -> number of GMRES iterations per correction
-    // option -> 0 if no correction is desired
-    int cycles_, iters_, option_;
-    double tol_;
-
-  };
-
-} // namespace
-
-#endif // MUELU_SHIFTEDLAPLACIANOPERATOR_DECL_HPP
+  void apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::one()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
+
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > R_, P_, A_;
+  RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Ac_;
+  RCP<Teuchos::ParameterList> BelosList_;
+
+  // RCP< Belos::LinearProblem<Scalar,MV,OP> > BelosLP_;
+  // RCP< Belos::SolverManager<Scalar,MV,OP> > BelosSM_;
+
+  // cycles -> number of V-cycles
+  // iters  -> number of GMRES iterations per correction
+  // option -> 0 if no correction is desired
+  int cycles_, iters_, option_;
+  double tol_;
+};
+
+}  // namespace MueLu
+
+#endif  // MUELU_SHIFTEDLAPLACIANOPERATOR_DECL_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_def.hpp b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_def.hpp
index 69512d6e8f45..fce9b5f4ec8b 100644
--- a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_def.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacianOperator_def.hpp
@@ -59,78 +59,74 @@
 #include "MueLu_Hierarchy.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLu {
 
 // ------------- getDomainMap -----------------------
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >
-ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-getDomainMap () const
-{
+Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >
+ShiftedLaplacianOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getDomainMap() const {
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XMatrix;
 
-  RCP<MueLu::Level> L0 = Hierarchy_->GetLevel (0);
-  RCP<XMatrix> A = L0->Get<RCP<XMatrix> > ("A");
+  RCP<MueLu::Level> L0 = Hierarchy_->GetLevel(0);
+  RCP<XMatrix> A       = L0->Get<RCP<XMatrix> >("A");
 
   RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpbA =
-    Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
+      Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
   if (tpbA != Teuchos::null) {
-    return Xpetra::toTpetraNonZero (tpbA->getDomainMap ());
+    return Xpetra::toTpetraNonZero(tpbA->getDomainMap());
   }
 
   RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpA =
-    Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs (A);
-  return tpA->getDomainMap ();
+      Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(A);
+  return tpA->getDomainMap();
 }
 
 // ------------- getRangeMap -----------------------
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >
-ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-getRangeMap () const
-{
+Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >
+ShiftedLaplacianOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getRangeMap() const {
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XMatrix;
 
-  RCP<MueLu::Level>  L0 = Hierarchy_->GetLevel(0);
-  RCP<XMatrix> A = L0->Get< RCP<XMatrix> >("A");
+  RCP<MueLu::Level> L0 = Hierarchy_->GetLevel(0);
+  RCP<XMatrix> A       = L0->Get<RCP<XMatrix> >("A");
 
   RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpbA =
-    Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
-  if(tpbA != Teuchos::null)
+      Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
+  if (tpbA != Teuchos::null)
     return Xpetra::toTpetraNonZero(tpbA->getRangeMap());
 
-  RCP< Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpA =
-    Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(A);
+  RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpA =
+      Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(A);
   return tpA->getRangeMap();
 }
 
 // ------------- apply -----------------------
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                                                               Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                                                               Teuchos::ETransp /* mode */, Scalar /* alpha */, Scalar /* beta */) const {
-
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>        TMV;
-  typedef Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>  XTMV;
+void ShiftedLaplacianOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                                                                                Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                                                                                Teuchos::ETransp /* mode */, Scalar /* alpha */, Scalar /* beta */) const {
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> TMV;
+  typedef Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTMV;
   // typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>        XMV; // unused
 
-  TMV & temp_x = const_cast<TMV &>(X);
+  TMV& temp_x = const_cast<TMV&>(X);
   const XTMV tX(rcpFromRef(temp_x));
-  XTMV       tY(rcpFromRef(Y));
+  XTMV tY(rcpFromRef(Y));
 
   try {
     tY.putScalar(0.0);
     Hierarchy_->Iterate(tX, tY, cycles_, true);
   }
 
-  catch(std::exception& e) {
-    //FIXME add message and rethrow
+  catch (std::exception& e) {
+    // FIXME add message and rethrow
     std::cerr << "Caught an exception in MueLu::ShiftedLaplacianOperator::ApplyInverse():" << std::endl
-                                                                                              << e.what() << std::endl;
+              << e.what() << std::endl;
   }
 
   // update solution with 2-grid error correction
@@ -160,15 +156,14 @@ void ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply(con
     std::cerr << "Caught an exception in MueLu::ShiftedLaplacianOperator::ApplyInverse():" << std::endl
     << e.what() << std::endl;
     }*/
-
 }
 
 // ------------- apply -----------------------
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-bool ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::hasTransposeApply() const {
+bool ShiftedLaplacianOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasTransposeApply() const {
   return false;
 }
 
-} // namespace
+}  // namespace MueLu
 
-#endif //ifdef MUELU_SHIFTEDLAPLACIANOPERATOR_DEF_HPP
+#endif  // ifdef MUELU_SHIFTEDLAPLACIANOPERATOR_DEF_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_decl.hpp b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_decl.hpp
index 0a4edf9c5a1b..c35bd00f661d 100644
--- a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_decl.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_decl.hpp
@@ -91,226 +91,221 @@
 
 namespace MueLu {
 
-  /*!
-    @brief Shifted Laplacian Helmholtz solver
-
-    This class provides a black box solver for indefinite Helmholtz problems.
-    An AMG-Shifted Laplacian is used as a preconditioner for Krylov iterative
-    solvers in Belos.
-
-    @ingroup MueLuAdapters
-  */
-  template <class Scalar        = DefaultScalar,
-            class LocalOrdinal  = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node          = DefaultNode>
-  class ShiftedLaplacian : public BaseClass {
+/*!
+  @brief Shifted Laplacian Helmholtz solver
+
+  This class provides a black box solver for indefinite Helmholtz problems.
+  An AMG-Shifted Laplacian is used as a preconditioner for Krylov iterative
+  solvers in Belos.
+
+  @ingroup MueLuAdapters
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ShiftedLaplacian : public BaseClass {
 #undef MUELU_SHIFTEDLAPLACIAN_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-    typedef Tpetra::Vector<SC,LO,GO,NO>                  TVEC;
-    typedef Tpetra::MultiVector<SC,LO,GO,NO>             TMV;
-    typedef Tpetra::Operator<SC,LO,GO,NO>                OP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::Operator<SC, LO, GO, NO> OP;
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-    typedef Belos::LinearProblem<SC,TMV,OP>              LinearProblem;
-    typedef Belos::SolverManager<SC,TMV,OP>              SolverManager;
-    typedef Belos::SolverFactory<SC,TMV,OP>              SolverFactory;
+  typedef Belos::LinearProblem<SC, TMV, OP> LinearProblem;
+  typedef Belos::SolverManager<SC, TMV, OP> SolverManager;
+  typedef Belos::SolverFactory<SC, TMV, OP> SolverFactory;
 #endif
 
-  public:
-
+ public:
   /*
     FIXME  26-June-2015 JJH:  This contructor is setting numerous defaults.  However, they don't match the defaults
     FIXME  int the method setParameters().  There also isn't any parameter validation that I can see.
   */
 
-    //! Constructors
-    ShiftedLaplacian():
-      numPDEs_(1),
-      Smoother_("schwarz"),
-      Aggregation_("uncoupled"),
-      Nullspace_("constant"),
-      numLevels_(5),
-      coarseGridSize_(100),
-      omega_(2.0*M_PI),
-      iters_(500),
-      blksize_(1),
-      tol_(1.0e-4),
-      nsweeps_(5),
-      ncycles_(1),
-      cycles_(8),
-      subiters_(10),
-      option_(1),
-      nproblems_(0),
-      solverType_(1),
-      restart_size_(100),
-      recycle_size_(25),
-      smoother_sweeps_(4),
-      smoother_damping_((SC)1.0),
-      krylov_type_(1),
-      krylov_iterations_(5),
-      krylov_preconditioner_(1),
-      ilu_leveloffill_(5.0),
-      ilu_abs_thresh_(0.0),
-      ilu_rel_thresh_(1.0),
-      ilu_diagpivotthresh_(0.1),
-      ilu_drop_tol_(0.01),
-      ilu_fill_tol_(0.01),
-      ilu_relax_val_(1.0),
-      ilu_rowperm_("LargeDiag"),
-      ilu_colperm_("COLAMD"),
-      ilu_drop_rule_("DROP_BASIC"),
-      ilu_normtype_("INF_NORM"),
-      ilu_milutype_("SILU"),
-      schwarz_overlap_(0),
-      schwarz_usereorder_(true),
-      schwarz_combinemode_(Tpetra::ADD),
-      schwarz_ordermethod_("rcm"),
-      GridTransfersExist_(false),
-      isSymmetric_(true)
-    { }
-
-    // Destructor
-    virtual ~ShiftedLaplacian();
-
-    // Parameters
-    void setParameters(Teuchos::RCP< Teuchos::ParameterList > paramList);
-
-    // Set matrices
-    void setProblemMatrix(RCP<Matrix>& A);
-    void setProblemMatrix(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraA);
-    void setPreconditioningMatrix(RCP<Matrix>& P);
-    void setPreconditioningMatrix(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraP);
-    void setstiff(RCP<Matrix>& K);
-    void setstiff(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraK);
-    void setmass(RCP<Matrix>& M);
-    void setmass(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraM);
-    void setcoords(RCP<MultiVector>& Coords);
-    void setNullSpace(RCP<MultiVector> NullSpace);
-    void setLevelShifts(std::vector<Scalar> levelshifts);
-
-    // initialize: set parameters and factories, construct
-    // prolongation and restriction matrices
-    void initialize();
-    // setupFastRAP: setup hierarchy with
-    // prolongators of the stiffness matrix
-    // constant complex shifts
-    void setupFastRAP();
-    // setupSlowRAP: setup hierarchy with
-    // prolongators of the stiffness matrix
-    // variable complex shifts
-    void setupSlowRAP();
-    // setupNormalRAP: setup hierarchy with
-    // prolongators of the preconditioning matrix
-    void setupNormalRAP();
-    // setupSolver: initialize Belos solver
-    void setupSolver();
-    // resetLinearProblem: for multiple frequencies;
-    // reset the Belos operator if the frequency changes
-    void resetLinearProblem();
-
-
-    // Solve phase
-    int solve(const RCP<TMV> B, RCP<TMV>& X);
-    void multigrid_apply(const RCP<MultiVector> B,
-			 RCP<MultiVector>& X);
-    void multigrid_apply(const RCP<Tpetra::MultiVector<SC,LO,GO,NO> > B,
-			 RCP<Tpetra::MultiVector<SC,LO,GO,NO> >& X);
-    int GetIterations();
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType GetResidual();
-
-    RCP<FactoryManager>               Manager_;
-
-  private:
-
-    // Problem options
-    // numPDEs_ -> number of DOFs at each node
-    int numPDEs_;
-
-    // Multigrid options
-    // numLevels_      -> number of Multigrid levels
-    // coarseGridSize_ -> size of coarsest grid (if current level has less DOFs, stop coarsening)
-    std::string Smoother_, Aggregation_, Nullspace_;
-    int numLevels_, coarseGridSize_;
-
-    // Shifted Laplacian/Helmholtz parameters
-    double          omega_;
-    std::vector<SC> levelshifts_;
-
-    // Krylov solver inputs
-    // iters  -> max number of iterations
-    // tol    -> residual tolerance
-    int    iters_, blksize_;
-    double tol_;
-    int    nsweeps_, ncycles_;
-    int    cycles_, subiters_, option_, nproblems_, solverType_;
-    int    restart_size_, recycle_size_;
-
-    // Smoother parameters
-    int    smoother_sweeps_;
-    Scalar smoother_damping_;
-    int    krylov_type_;
-    int    krylov_iterations_;
-    int    krylov_preconditioner_;
-    double ilu_leveloffill_, ilu_abs_thresh_, ilu_rel_thresh_, ilu_diagpivotthresh_;
-    double ilu_drop_tol_, ilu_fill_tol_, ilu_relax_val_;
-    std::string ilu_rowperm_, ilu_colperm_, ilu_drop_rule_, ilu_normtype_, ilu_milutype_;
-    int    schwarz_overlap_;
-    bool   schwarz_usereorder_;
-    Tpetra::CombineMode schwarz_combinemode_;
-    std::string schwarz_ordermethod_;
-
-    // flags for setup
-    bool GridTransfersExist_;
-    bool isSymmetric_;
-
-    // Xpetra matrices
-    // K_ -> stiffness matrix
-    // M_ -> mass matrix
-    // A_ -> Problem matrix
-    // P_ -> Preconditioning matrix
-    RCP<Matrix>                       K_, M_, A_, P_;
-    RCP<MultiVector>                  Coords_, NullSpace_;
-
-    // Multigrid Hierarchy
-    RCP<Hierarchy>                    Hierarchy_;
-
-    // Factories and prototypes
-    RCP<TentativePFactory>            TentPfact_;
-    RCP<PFactory>                     Pfact_;
-    RCP<PgPFactory>                   PgPfact_;
-    RCP<TransPFactory>                TransPfact_;
-    RCP<GenericRFactory>              Rfact_;
-    RCP<RAPFactory>                   Acfact_;
-    RCP<RAPShiftFactory>              Acshift_;
-    RCP<AmalgamationFactory>          Amalgfact_;
-    RCP<CoalesceDropFactory>          Dropfact_;
-    RCP<UncoupledAggregationFactory>  UCaggfact_;
-    RCP<CoarseMapFactory>             CoarseMapfact_;
-    RCP<SmootherPrototype>            smooProto_, coarsestSmooProto_;
-    RCP<SmootherFactory>              smooFact_,  coarsestSmooFact_;
-    Teuchos::ParameterList            coarsestSmooList_;
-    std::string                       precType_;
-    Teuchos::ParameterList            precList_;
-
-    // Operator and Preconditioner
-    RCP< MueLu::ShiftedLaplacianOperator<SC,LO,GO,NO> > MueLuOp_;
-    RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >               TpetraA_;
+  //! Constructors
+  ShiftedLaplacian()
+    : numPDEs_(1)
+    , Smoother_("schwarz")
+    , Aggregation_("uncoupled")
+    , Nullspace_("constant")
+    , numLevels_(5)
+    , coarseGridSize_(100)
+    , omega_(2.0 * M_PI)
+    , iters_(500)
+    , blksize_(1)
+    , tol_(1.0e-4)
+    , nsweeps_(5)
+    , ncycles_(1)
+    , cycles_(8)
+    , subiters_(10)
+    , option_(1)
+    , nproblems_(0)
+    , solverType_(1)
+    , restart_size_(100)
+    , recycle_size_(25)
+    , smoother_sweeps_(4)
+    , smoother_damping_((SC)1.0)
+    , krylov_type_(1)
+    , krylov_iterations_(5)
+    , krylov_preconditioner_(1)
+    , ilu_leveloffill_(5.0)
+    , ilu_abs_thresh_(0.0)
+    , ilu_rel_thresh_(1.0)
+    , ilu_diagpivotthresh_(0.1)
+    , ilu_drop_tol_(0.01)
+    , ilu_fill_tol_(0.01)
+    , ilu_relax_val_(1.0)
+    , ilu_rowperm_("LargeDiag")
+    , ilu_colperm_("COLAMD")
+    , ilu_drop_rule_("DROP_BASIC")
+    , ilu_normtype_("INF_NORM")
+    , ilu_milutype_("SILU")
+    , schwarz_overlap_(0)
+    , schwarz_usereorder_(true)
+    , schwarz_combinemode_(Tpetra::ADD)
+    , schwarz_ordermethod_("rcm")
+    , GridTransfersExist_(false)
+    , isSymmetric_(true) {}
+
+  // Destructor
+  virtual ~ShiftedLaplacian();
+
+  // Parameters
+  void setParameters(Teuchos::RCP<Teuchos::ParameterList> paramList);
+
+  // Set matrices
+  void setProblemMatrix(RCP<Matrix>& A);
+  void setProblemMatrix(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraA);
+  void setPreconditioningMatrix(RCP<Matrix>& P);
+  void setPreconditioningMatrix(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraP);
+  void setstiff(RCP<Matrix>& K);
+  void setstiff(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraK);
+  void setmass(RCP<Matrix>& M);
+  void setmass(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraM);
+  void setcoords(RCP<MultiVector>& Coords);
+  void setNullSpace(RCP<MultiVector> NullSpace);
+  void setLevelShifts(std::vector<Scalar> levelshifts);
+
+  // initialize: set parameters and factories, construct
+  // prolongation and restriction matrices
+  void initialize();
+  // setupFastRAP: setup hierarchy with
+  // prolongators of the stiffness matrix
+  // constant complex shifts
+  void setupFastRAP();
+  // setupSlowRAP: setup hierarchy with
+  // prolongators of the stiffness matrix
+  // variable complex shifts
+  void setupSlowRAP();
+  // setupNormalRAP: setup hierarchy with
+  // prolongators of the preconditioning matrix
+  void setupNormalRAP();
+  // setupSolver: initialize Belos solver
+  void setupSolver();
+  // resetLinearProblem: for multiple frequencies;
+  // reset the Belos operator if the frequency changes
+  void resetLinearProblem();
+
+  // Solve phase
+  int solve(const RCP<TMV> B, RCP<TMV>& X);
+  void multigrid_apply(const RCP<MultiVector> B,
+                       RCP<MultiVector>& X);
+  void multigrid_apply(const RCP<Tpetra::MultiVector<SC, LO, GO, NO> > B,
+                       RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& X);
+  int GetIterations();
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType GetResidual();
+
+  RCP<FactoryManager> Manager_;
+
+ private:
+  // Problem options
+  // numPDEs_ -> number of DOFs at each node
+  int numPDEs_;
+
+  // Multigrid options
+  // numLevels_      -> number of Multigrid levels
+  // coarseGridSize_ -> size of coarsest grid (if current level has less DOFs, stop coarsening)
+  std::string Smoother_, Aggregation_, Nullspace_;
+  int numLevels_, coarseGridSize_;
+
+  // Shifted Laplacian/Helmholtz parameters
+  double omega_;
+  std::vector<SC> levelshifts_;
+
+  // Krylov solver inputs
+  // iters  -> max number of iterations
+  // tol    -> residual tolerance
+  int iters_, blksize_;
+  double tol_;
+  int nsweeps_, ncycles_;
+  int cycles_, subiters_, option_, nproblems_, solverType_;
+  int restart_size_, recycle_size_;
+
+  // Smoother parameters
+  int smoother_sweeps_;
+  Scalar smoother_damping_;
+  int krylov_type_;
+  int krylov_iterations_;
+  int krylov_preconditioner_;
+  double ilu_leveloffill_, ilu_abs_thresh_, ilu_rel_thresh_, ilu_diagpivotthresh_;
+  double ilu_drop_tol_, ilu_fill_tol_, ilu_relax_val_;
+  std::string ilu_rowperm_, ilu_colperm_, ilu_drop_rule_, ilu_normtype_, ilu_milutype_;
+  int schwarz_overlap_;
+  bool schwarz_usereorder_;
+  Tpetra::CombineMode schwarz_combinemode_;
+  std::string schwarz_ordermethod_;
+
+  // flags for setup
+  bool GridTransfersExist_;
+  bool isSymmetric_;
+
+  // Xpetra matrices
+  // K_ -> stiffness matrix
+  // M_ -> mass matrix
+  // A_ -> Problem matrix
+  // P_ -> Preconditioning matrix
+  RCP<Matrix> K_, M_, A_, P_;
+  RCP<MultiVector> Coords_, NullSpace_;
+
+  // Multigrid Hierarchy
+  RCP<Hierarchy> Hierarchy_;
+
+  // Factories and prototypes
+  RCP<TentativePFactory> TentPfact_;
+  RCP<PFactory> Pfact_;
+  RCP<PgPFactory> PgPfact_;
+  RCP<TransPFactory> TransPfact_;
+  RCP<GenericRFactory> Rfact_;
+  RCP<RAPFactory> Acfact_;
+  RCP<RAPShiftFactory> Acshift_;
+  RCP<AmalgamationFactory> Amalgfact_;
+  RCP<CoalesceDropFactory> Dropfact_;
+  RCP<UncoupledAggregationFactory> UCaggfact_;
+  RCP<CoarseMapFactory> CoarseMapfact_;
+  RCP<SmootherPrototype> smooProto_, coarsestSmooProto_;
+  RCP<SmootherFactory> smooFact_, coarsestSmooFact_;
+  Teuchos::ParameterList coarsestSmooList_;
+  std::string precType_;
+  Teuchos::ParameterList precList_;
+
+  // Operator and Preconditioner
+  RCP<MueLu::ShiftedLaplacianOperator<SC, LO, GO, NO> > MueLuOp_;
+  RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > TpetraA_;
 
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-    // Belos Linear Problem and Solver
-    RCP<LinearProblem>                LinearProblem_;
-    RCP<SolverManager>                SolverManager_;
-    RCP<SolverFactory>                SolverFactory_;
-    RCP<Teuchos::ParameterList>       BelosList_;
+  // Belos Linear Problem and Solver
+  RCP<LinearProblem> LinearProblem_;
+  RCP<SolverManager> SolverManager_;
+  RCP<SolverFactory> SolverFactory_;
+  RCP<Teuchos::ParameterList> BelosList_;
 #endif
+};
 
-  };
-
-}
+}  // namespace MueLu
 
 #define MUELU_SHIFTEDLAPLACIAN_SHORT
 
-#endif //if defined(HAVE_MUELU_IFPACK2) and defined(HAVE_MUELU_TPETRA)
+#endif  // if defined(HAVE_MUELU_IFPACK2) and defined(HAVE_MUELU_TPETRA)
 
-#endif // MUELU_SHIFTEDLAPLACIAN_DECL_HPP
+#endif  // MUELU_SHIFTEDLAPLACIAN_DECL_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_def.hpp b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_def.hpp
index 231849305e7a..79751844e0d6 100644
--- a/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_def.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_ShiftedLaplacian_def.hpp
@@ -75,278 +75,255 @@
 namespace MueLu {
 
 // Destructor
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::~ShiftedLaplacian() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~ShiftedLaplacian() {}
 
 // Input
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameters(Teuchos::RCP< Teuchos::ParameterList > paramList) {
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameters(Teuchos::RCP<Teuchos::ParameterList> paramList) {
   // Parameters
-  coarseGridSize_      = paramList->get("MueLu: coarse size", 1000);
-  numLevels_           = paramList->get("MueLu: levels",         3);
-  int stype            = paramList->get("MueLu: smoother",       8);
-  if(stype==1)         { Smoother_="jacobi";                      }
-  else if(stype==2)    { Smoother_="gauss-seidel";                }
-  else if(stype==3)    { Smoother_="symmetric gauss-seidel";      }
-  else if(stype==4)    { Smoother_="chebyshev";                   }
-  else if(stype==5)    { Smoother_="krylov";                      }
-  else if(stype==6)    { Smoother_="ilut";                        }
-  else if(stype==7)    { Smoother_="riluk";                       }
-  else if(stype==8)    { Smoother_="schwarz";                     }
-  else if(stype==9)    { Smoother_="superilu";                    }
-  else if(stype==10)   { Smoother_="superlu";                     }
-  else                 { Smoother_="schwarz";                     }
-  smoother_sweeps_     = paramList->get("MueLu: sweeps",         5);
-  smoother_damping_    = paramList->get("MueLu: relax val",    1.0);
-  ncycles_             = paramList->get("MueLu: cycles",         1);
-  iters_               = paramList->get("MueLu: iterations",   500);
-  solverType_          = paramList->get("MueLu: solver type",    1);
+  coarseGridSize_ = paramList->get("MueLu: coarse size", 1000);
+  numLevels_      = paramList->get("MueLu: levels", 3);
+  int stype       = paramList->get("MueLu: smoother", 8);
+  if (stype == 1) {
+    Smoother_ = "jacobi";
+  } else if (stype == 2) {
+    Smoother_ = "gauss-seidel";
+  } else if (stype == 3) {
+    Smoother_ = "symmetric gauss-seidel";
+  } else if (stype == 4) {
+    Smoother_ = "chebyshev";
+  } else if (stype == 5) {
+    Smoother_ = "krylov";
+  } else if (stype == 6) {
+    Smoother_ = "ilut";
+  } else if (stype == 7) {
+    Smoother_ = "riluk";
+  } else if (stype == 8) {
+    Smoother_ = "schwarz";
+  } else if (stype == 9) {
+    Smoother_ = "superilu";
+  } else if (stype == 10) {
+    Smoother_ = "superlu";
+  } else {
+    Smoother_ = "schwarz";
+  }
+  smoother_sweeps_     = paramList->get("MueLu: sweeps", 5);
+  smoother_damping_    = paramList->get("MueLu: relax val", 1.0);
+  ncycles_             = paramList->get("MueLu: cycles", 1);
+  iters_               = paramList->get("MueLu: iterations", 500);
+  solverType_          = paramList->get("MueLu: solver type", 1);
   restart_size_        = paramList->get("MueLu: restart size", 100);
-  recycle_size_        = paramList->get("MueLu: recycle size",  25);
-  isSymmetric_         = paramList->get("MueLu: symmetric",   true);
-  ilu_leveloffill_     = paramList->get("MueLu: level-of-fill",  5);
-  ilu_abs_thresh_      = paramList->get("MueLu: abs thresh",   0.0);
-  ilu_rel_thresh_      = paramList->get("MueLu: rel thresh",   1.0);
-  ilu_diagpivotthresh_ = paramList->get("MueLu: piv thresh",   0.1);
-  ilu_drop_tol_        = paramList->get("MueLu: drop tol",    0.01);
-  ilu_fill_tol_        = paramList->get("MueLu: fill tol",    0.01);
-  schwarz_overlap_     = paramList->get("MueLu: overlap",        0);
+  recycle_size_        = paramList->get("MueLu: recycle size", 25);
+  isSymmetric_         = paramList->get("MueLu: symmetric", true);
+  ilu_leveloffill_     = paramList->get("MueLu: level-of-fill", 5);
+  ilu_abs_thresh_      = paramList->get("MueLu: abs thresh", 0.0);
+  ilu_rel_thresh_      = paramList->get("MueLu: rel thresh", 1.0);
+  ilu_diagpivotthresh_ = paramList->get("MueLu: piv thresh", 0.1);
+  ilu_drop_tol_        = paramList->get("MueLu: drop tol", 0.01);
+  ilu_fill_tol_        = paramList->get("MueLu: fill tol", 0.01);
+  schwarz_overlap_     = paramList->get("MueLu: overlap", 0);
   schwarz_usereorder_  = paramList->get("MueLu: use reorder", true);
-  int combinemode      = paramList->get("MueLu: combine mode",   1);
-  if(combinemode==0)   { schwarz_combinemode_ = Tpetra::ZERO;     }
-  else                 { schwarz_combinemode_ = Tpetra::ADD;      }
-  tol_                 = paramList->get("MueLu: tolerance",  0.001);
-
+  int combinemode      = paramList->get("MueLu: combine mode", 1);
+  if (combinemode == 0) {
+    schwarz_combinemode_ = Tpetra::ZERO;
+  } else {
+    schwarz_combinemode_ = Tpetra::ADD;
+  }
+  tol_ = paramList->get("MueLu: tolerance", 0.001);
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setProblemMatrix(RCP<Matrix>& A) {
-
-  A_=A;
-  if(A_!=Teuchos::null)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setProblemMatrix(RCP<Matrix>& A) {
+  A_ = A;
+  if (A_ != Teuchos::null)
     TpetraA_ = Utilities::Op2NonConstTpetraCrs(A_);
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-  if(LinearProblem_!=Teuchos::null)
-    LinearProblem_ -> setOperator ( TpetraA_ );
+  if (LinearProblem_ != Teuchos::null)
+    LinearProblem_->setOperator(TpetraA_);
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
 #endif
-
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setProblemMatrix(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraA) {
-
-  TpetraA_=TpetraA;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setProblemMatrix(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraA) {
+  TpetraA_ = TpetraA;
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-  if(LinearProblem_!=Teuchos::null)
-    LinearProblem_ -> setOperator ( TpetraA_ );
+  if (LinearProblem_ != Teuchos::null)
+    LinearProblem_->setOperator(TpetraA_);
 #endif
-
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setPreconditioningMatrix(RCP<Matrix>& P) {
-
-  P_=P;
-  GridTransfersExist_=false;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setPreconditioningMatrix(RCP<Matrix>& P) {
+  P_                  = P;
+  GridTransfersExist_ = false;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setPreconditioningMatrix(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraP) {
-
-  RCP< Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp
-    = rcp( new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraP) );
-  P_= rcp( new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp) );
-  GridTransfersExist_=false;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setPreconditioningMatrix(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraP) {
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraP));
+  P_                                                                      = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp));
+  GridTransfersExist_                                                     = false;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setstiff(RCP<Matrix>& K) {
-
-  K_=K;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setstiff(RCP<Matrix>& K) {
+  K_ = K;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setstiff(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraK) {
-
-  RCP< Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp
-    = rcp( new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraK) );
-  K_= rcp( new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp) );
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setstiff(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraK) {
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraK));
+  K_                                                                      = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp));
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setmass(RCP<Matrix>& M) {
-
-  M_=M;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setmass(RCP<Matrix>& M) {
+  M_ = M;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setmass(RCP< Tpetra::CrsMatrix<SC,LO,GO,NO> >& TpetraM) {
-
-  RCP< Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp
-    = rcp( new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraM) );
-  M_= rcp( new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp) );
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setmass(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& TpetraM) {
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Atmp = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(TpetraM));
+  M_                                                                      = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Atmp));
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setcoords(RCP<MultiVector>& Coords) {
-
-  Coords_=Coords;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setcoords(RCP<MultiVector>& Coords) {
+  Coords_ = Coords;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setNullSpace(RCP<MultiVector> NullSpace) {
-
-  NullSpace_=NullSpace;
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setNullSpace(RCP<MultiVector> NullSpace) {
+  NullSpace_ = NullSpace;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setLevelShifts(std::vector<Scalar> levelshifts) {
-
-  levelshifts_=levelshifts;
-  numLevels_=levelshifts_.size();
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setLevelShifts(std::vector<Scalar> levelshifts) {
+  levelshifts_ = levelshifts;
+  numLevels_   = levelshifts_.size();
 }
 
 // initialize
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::initialize() {
-
-  TentPfact_     = rcp( new TentativePFactory           );
-  Pfact_         = rcp( new SaPFactory                  );
-  PgPfact_       = rcp( new PgPFactory                  );
-  TransPfact_    = rcp( new TransPFactory               );
-  Rfact_         = rcp( new GenericRFactory             );
-  Acfact_        = rcp( new RAPFactory                  );
-  Acshift_       = rcp( new RAPShiftFactory             );
-  Amalgfact_     = rcp( new AmalgamationFactory         );
-  Dropfact_      = rcp( new CoalesceDropFactory         );
-  UCaggfact_     = rcp( new UncoupledAggregationFactory );
-  CoarseMapfact_ = rcp( new CoarseMapFactory            );
-  Manager_       = rcp( new FactoryManager              );
-  Manager_   -> SetFactory("UnAmalgamationInfo", Amalgfact_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::initialize() {
+  TentPfact_     = rcp(new TentativePFactory);
+  Pfact_         = rcp(new SaPFactory);
+  PgPfact_       = rcp(new PgPFactory);
+  TransPfact_    = rcp(new TransPFactory);
+  Rfact_         = rcp(new GenericRFactory);
+  Acfact_        = rcp(new RAPFactory);
+  Acshift_       = rcp(new RAPShiftFactory);
+  Amalgfact_     = rcp(new AmalgamationFactory);
+  Dropfact_      = rcp(new CoalesceDropFactory);
+  UCaggfact_     = rcp(new UncoupledAggregationFactory);
+  CoarseMapfact_ = rcp(new CoarseMapFactory);
+  Manager_       = rcp(new FactoryManager);
+  Manager_->SetFactory("UnAmalgamationInfo", Amalgfact_);
   Teuchos::ParameterList params;
-  params.set("lightweight wrap",true);
-  params.set("aggregation: drop scheme","classical");
-  Dropfact_  -> SetParameterList(params);
-  Manager_   -> SetFactory("Graph", Dropfact_);
-  Manager_   -> SetFactory("Aggregates", UCaggfact_ );
-  Manager_     -> SetFactory("CoarseMap", CoarseMapfact_);
-  Manager_     -> SetFactory("Ptent", TentPfact_);
-  if(isSymmetric_==true) {
-    Manager_   -> SetFactory("P", Pfact_);
-    Manager_   -> SetFactory("R", TransPfact_);
-  }
-  else {
-    Manager_   -> SetFactory("P", PgPfact_);
-    Manager_   -> SetFactory("R", Rfact_);
+  params.set("lightweight wrap", true);
+  params.set("aggregation: drop scheme", "classical");
+  Dropfact_->SetParameterList(params);
+  Manager_->SetFactory("Graph", Dropfact_);
+  Manager_->SetFactory("Aggregates", UCaggfact_);
+  Manager_->SetFactory("CoarseMap", CoarseMapfact_);
+  Manager_->SetFactory("Ptent", TentPfact_);
+  if (isSymmetric_ == true) {
+    Manager_->SetFactory("P", Pfact_);
+    Manager_->SetFactory("R", TransPfact_);
+  } else {
+    Manager_->SetFactory("P", PgPfact_);
+    Manager_->SetFactory("R", Rfact_);
     solverType_ = 10;
   }
 
   // choose smoother
-  if(Smoother_=="jacobi") {
+  if (Smoother_ == "jacobi") {
     precType_ = "RELAXATION";
     precList_.set("relaxation: type", "Jacobi");
     precList_.set("relaxation: sweeps", smoother_sweeps_);
     precList_.set("relaxation: damping factor", smoother_damping_);
-  }
-  else if(Smoother_=="gauss-seidel") {
+  } else if (Smoother_ == "gauss-seidel") {
     precType_ = "RELAXATION";
     precList_.set("relaxation: type", "Gauss-Seidel");
     precList_.set("relaxation: sweeps", smoother_sweeps_);
     precList_.set("relaxation: damping factor", smoother_damping_);
-  }
-  else if(Smoother_=="symmetric gauss-seidel") {
+  } else if (Smoother_ == "symmetric gauss-seidel") {
     precType_ = "RELAXATION";
     precList_.set("relaxation: type", "Symmetric Gauss-Seidel");
     precList_.set("relaxation: sweeps", smoother_sweeps_);
     precList_.set("relaxation: damping factor", smoother_damping_);
-  }
-  else if(Smoother_=="chebyshev") {
+  } else if (Smoother_ == "chebyshev") {
     precType_ = "CHEBYSHEV";
-  }
-  else if(Smoother_=="krylov") {
+  } else if (Smoother_ == "krylov") {
     precType_ = "KRYLOV";
     precList_.set("krylov: iteration type", krylov_type_);
     precList_.set("krylov: number of iterations", krylov_iterations_);
-    precList_.set("krylov: residual tolerance",1.0e-8);
-    precList_.set("krylov: block size",1);
+    precList_.set("krylov: residual tolerance", 1.0e-8);
+    precList_.set("krylov: block size", 1);
     precList_.set("krylov: preconditioner type", krylov_preconditioner_);
-    precList_.set("relaxation: sweeps",1);
-    solverType_=10;
-  }
-  else if(Smoother_=="ilut") {
+    precList_.set("relaxation: sweeps", 1);
+    solverType_ = 10;
+  } else if (Smoother_ == "ilut") {
     precType_ = "ILUT";
     precList_.set("fact: ilut level-of-fill", ilu_leveloffill_);
     precList_.set("fact: absolute threshold", ilu_abs_thresh_);
     precList_.set("fact: relative threshold", ilu_rel_thresh_);
-    precList_.set("fact: drop tolerance",     ilu_drop_tol_);
-    precList_.set("fact: relax value",        ilu_relax_val_);
-  }
-  else if(Smoother_=="riluk") {
+    precList_.set("fact: drop tolerance", ilu_drop_tol_);
+    precList_.set("fact: relax value", ilu_relax_val_);
+  } else if (Smoother_ == "riluk") {
     precType_ = "RILUK";
     precList_.set("fact: iluk level-of-fill", ilu_leveloffill_);
     precList_.set("fact: absolute threshold", ilu_abs_thresh_);
     precList_.set("fact: relative threshold", ilu_rel_thresh_);
-    precList_.set("fact: drop tolerance",     ilu_drop_tol_);
-    precList_.set("fact: relax value",        ilu_relax_val_);
-  }
-  else if(Smoother_=="schwarz") {
+    precList_.set("fact: drop tolerance", ilu_drop_tol_);
+    precList_.set("fact: relax value", ilu_relax_val_);
+  } else if (Smoother_ == "schwarz") {
     precType_ = "SCHWARZ";
     precList_.set("schwarz: overlap level", schwarz_overlap_);
     precList_.set("schwarz: combine mode", schwarz_combinemode_);
     precList_.set("schwarz: use reordering", schwarz_usereorder_);
     //    precList_.set("schwarz: filter singletons", true); // Disabled due to issues w/ Ifpack2/Zoltan2 w.r.t. Issue #560 - CMS 8/26/16
-    precList_.set("order_method",schwarz_ordermethod_);
-    precList_.sublist("schwarz: reordering list").set("order_method",schwarz_ordermethod_);
+    precList_.set("order_method", schwarz_ordermethod_);
+    precList_.sublist("schwarz: reordering list").set("order_method", schwarz_ordermethod_);
     precList_.sublist("schwarz: subdomain solver parameters").set("fact: ilut level-of-fill", ilu_leveloffill_);
     precList_.sublist("schwarz: subdomain solver parameters").set("fact: absolute threshold", ilu_abs_thresh_);
     precList_.sublist("schwarz: subdomain solver parameters").set("fact: relative threshold", ilu_rel_thresh_);
-    precList_.sublist("schwarz: subdomain solver parameters").set("fact: drop tolerance",     ilu_drop_tol_);
-    precList_.sublist("schwarz: subdomain solver parameters").set("fact: relax value",        ilu_relax_val_);
-  }
-  else if(Smoother_=="superilu") {
+    precList_.sublist("schwarz: subdomain solver parameters").set("fact: drop tolerance", ilu_drop_tol_);
+    precList_.sublist("schwarz: subdomain solver parameters").set("fact: relax value", ilu_relax_val_);
+  } else if (Smoother_ == "superilu") {
     precType_ = "superlu";
     precList_.set("RowPerm", ilu_rowperm_);
     precList_.set("ColPerm", ilu_colperm_);
     precList_.set("DiagPivotThresh", ilu_diagpivotthresh_);
-    precList_.set("ILU_DropRule",ilu_drop_rule_);
-    precList_.set("ILU_DropTol",ilu_drop_tol_);
-    precList_.set("ILU_FillFactor",ilu_leveloffill_);
-    precList_.set("ILU_Norm",ilu_normtype_);
-    precList_.set("ILU_MILU",ilu_milutype_);
-    precList_.set("ILU_FillTol",ilu_fill_tol_);
-    precList_.set("ILU_Flag",true);
-  }
-  else if(Smoother_=="superlu") {
+    precList_.set("ILU_DropRule", ilu_drop_rule_);
+    precList_.set("ILU_DropTol", ilu_drop_tol_);
+    precList_.set("ILU_FillFactor", ilu_leveloffill_);
+    precList_.set("ILU_Norm", ilu_normtype_);
+    precList_.set("ILU_MILU", ilu_milutype_);
+    precList_.set("ILU_FillTol", ilu_fill_tol_);
+    precList_.set("ILU_Flag", true);
+  } else if (Smoother_ == "superlu") {
     precType_ = "superlu";
     precList_.set("ColPerm", ilu_colperm_);
     precList_.set("DiagPivotThresh", ilu_diagpivotthresh_);
   }
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
   // construct smoother
-  smooProto_ = rcp( new Ifpack2Smoother(precType_,precList_) );
-  smooFact_  = rcp( new SmootherFactory(smooProto_) );
+  smooProto_ = rcp(new Ifpack2Smoother(precType_, precList_));
+  smooFact_  = rcp(new SmootherFactory(smooProto_));
 #if defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_SUPERLU)
-  coarsestSmooProto_ = rcp( new DirectSolver("Superlu",coarsestSmooList_) );
+  coarsestSmooProto_ = rcp(new DirectSolver("Superlu", coarsestSmooList_));
 #elif defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_KLU2)
-  coarsestSmooProto_ = rcp( new DirectSolver("Klu",coarsestSmooList_) );
+  coarsestSmooProto_ = rcp(new DirectSolver("Klu", coarsestSmooList_));
 #elif defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_SUPERLUDIST)
-  coarsestSmooProto_ = rcp( new DirectSolver("Superludist",coarsestSmooList_) );
+  coarsestSmooProto_ = rcp(new DirectSolver("Superludist", coarsestSmooList_));
 #else
-  coarsestSmooProto_ = rcp( new Ifpack2Smoother(precType_,precList_) );
+  coarsestSmooProto_ = rcp(new Ifpack2Smoother(precType_, precList_));
 #endif
-  coarsestSmooFact_  = rcp( new SmootherFactory(coarsestSmooProto_, Teuchos::null) );
+  coarsestSmooFact_ = rcp(new SmootherFactory(coarsestSmooProto_, Teuchos::null));
 
   // For setupSlowRAP and setupFastRAP, the prolongation/restriction matrices
   // are constructed with the stiffness matrix. These matrices are kept for future
@@ -354,150 +331,143 @@ void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::initialize() {
   // useful for multiple frequency problems - when the frequency/preconditioner
   // changes, you only compute coarse grids (RAPs) and setup level smoothers when
   // you call Hierarchy->Setup().
-  if(K_!=Teuchos::null) {
-    Manager_ -> SetFactory("Smoother", Teuchos::null);
-    Manager_ -> SetFactory("CoarseSolver", Teuchos::null);
-    Hierarchy_ = rcp( new Hierarchy(K_)  );
-    if(NullSpace_!=Teuchos::null)
-      Hierarchy_ -> GetLevel(0) -> Set("Nullspace", NullSpace_);
-    if(isSymmetric_==true) {
-      Hierarchy_ -> Keep("P", Pfact_.get());
-      Hierarchy_ -> Keep("R", TransPfact_.get());
-      Hierarchy_ -> SetImplicitTranspose(true);
-    }
-    else {
-      Hierarchy_ -> Keep("P", PgPfact_.get());
-      Hierarchy_ -> Keep("R", Rfact_.get());
+  if (K_ != Teuchos::null) {
+    Manager_->SetFactory("Smoother", Teuchos::null);
+    Manager_->SetFactory("CoarseSolver", Teuchos::null);
+    Hierarchy_ = rcp(new Hierarchy(K_));
+    if (NullSpace_ != Teuchos::null)
+      Hierarchy_->GetLevel(0)->Set("Nullspace", NullSpace_);
+    if (isSymmetric_ == true) {
+      Hierarchy_->Keep("P", Pfact_.get());
+      Hierarchy_->Keep("R", TransPfact_.get());
+      Hierarchy_->SetImplicitTranspose(true);
+    } else {
+      Hierarchy_->Keep("P", PgPfact_.get());
+      Hierarchy_->Keep("R", Rfact_.get());
     }
-    Hierarchy_ -> Keep("Ptent", TentPfact_.get());
-    Hierarchy_ -> SetMaxCoarseSize( coarseGridSize_ );
-    Hierarchy_ -> Setup(*Manager_, 0, numLevels_);
-    GridTransfersExist_=true;
+    Hierarchy_->Keep("Ptent", TentPfact_.get());
+    Hierarchy_->SetMaxCoarseSize(coarseGridSize_);
+    Hierarchy_->Setup(*Manager_, 0, numLevels_);
+    GridTransfersExist_ = true;
   }
   // Use preconditioning matrix to setup prolongation/restriction operators
   else {
-    Manager_ -> SetFactory("Smoother", smooFact_);
-    Manager_ -> SetFactory("CoarseSolver", coarsestSmooFact_);
-    Hierarchy_ = rcp( new Hierarchy(P_)  );
-    if(NullSpace_!=Teuchos::null)
-      Hierarchy_ -> GetLevel(0) -> Set("Nullspace", NullSpace_);
-    if(isSymmetric_==true)
-      Hierarchy_ -> SetImplicitTranspose(true);
-    Hierarchy_ -> SetMaxCoarseSize( coarseGridSize_ );
-    Hierarchy_ -> Setup(*Manager_, 0, numLevels_);
-    GridTransfersExist_=true;
+    Manager_->SetFactory("Smoother", smooFact_);
+    Manager_->SetFactory("CoarseSolver", coarsestSmooFact_);
+    Hierarchy_ = rcp(new Hierarchy(P_));
+    if (NullSpace_ != Teuchos::null)
+      Hierarchy_->GetLevel(0)->Set("Nullspace", NullSpace_);
+    if (isSymmetric_ == true)
+      Hierarchy_->SetImplicitTranspose(true);
+    Hierarchy_->SetMaxCoarseSize(coarseGridSize_);
+    Hierarchy_->Setup(*Manager_, 0, numLevels_);
+    GridTransfersExist_ = true;
   }
 
   // Belos Linear Problem and Solver Manager
-  BelosList_ = rcp( new Teuchos::ParameterList("GMRES") );
-  BelosList_ -> set("Maximum Iterations",iters_ );
-  BelosList_ -> set("Convergence Tolerance",tol_ );
-  BelosList_ -> set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  BelosList_ -> set("Output Frequency",1);
-  BelosList_ -> set("Output Style",Belos::Brief);
-  BelosList_ -> set("Num Blocks",restart_size_);
-  BelosList_ -> set("Num Recycled Blocks",recycle_size_);
+  BelosList_ = rcp(new Teuchos::ParameterList("GMRES"));
+  BelosList_->set("Maximum Iterations", iters_);
+  BelosList_->set("Convergence Tolerance", tol_);
+  BelosList_->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  BelosList_->set("Output Frequency", 1);
+  BelosList_->set("Output Style", Belos::Brief);
+  BelosList_->set("Num Blocks", restart_size_);
+  BelosList_->set("Num Recycled Blocks", recycle_size_);
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
 #endif
 }
 
 // setup coarse grids for new frequency
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setupFastRAP() {
-
-  int numLevels = Hierarchy_ -> GetNumLevels();
-  Manager_ -> SetFactory("Smoother", smooFact_);
-  Manager_ -> SetFactory("CoarseSolver", coarsestSmooFact_);
-  Hierarchy_ -> GetLevel(0) -> Set("A", P_);
-  Hierarchy_ -> Setup(*Manager_, 0, numLevels);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setupFastRAP() {
+  int numLevels = Hierarchy_->GetNumLevels();
+  Manager_->SetFactory("Smoother", smooFact_);
+  Manager_->SetFactory("CoarseSolver", coarsestSmooFact_);
+  Hierarchy_->GetLevel(0)->Set("A", P_);
+  Hierarchy_->Setup(*Manager_, 0, numLevels);
   setupSolver();
-
 }
 
 // setup coarse grids for new frequency
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setupSlowRAP() {
-
-  int numLevels = Hierarchy_ -> GetNumLevels();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setupSlowRAP() {
+  int numLevels = Hierarchy_->GetNumLevels();
   Acshift_->SetShifts(levelshifts_);
-  Manager_ -> SetFactory("Smoother", smooFact_);
-  Manager_ -> SetFactory("CoarseSolver", coarsestSmooFact_);
-  Manager_ -> SetFactory("A", Acshift_);
-  Manager_ -> SetFactory("K", Acshift_);
-  Manager_ -> SetFactory("M", Acshift_);
-  Hierarchy_ -> GetLevel(0) -> Set("A", P_);
-  Hierarchy_ -> GetLevel(0) -> Set("K", K_);
-  Hierarchy_ -> GetLevel(0) -> Set("M", M_);
-  Hierarchy_ -> Setup(*Manager_, 0, numLevels);
+  Manager_->SetFactory("Smoother", smooFact_);
+  Manager_->SetFactory("CoarseSolver", coarsestSmooFact_);
+  Manager_->SetFactory("A", Acshift_);
+  Manager_->SetFactory("K", Acshift_);
+  Manager_->SetFactory("M", Acshift_);
+  Hierarchy_->GetLevel(0)->Set("A", P_);
+  Hierarchy_->GetLevel(0)->Set("K", K_);
+  Hierarchy_->GetLevel(0)->Set("M", M_);
+  Hierarchy_->Setup(*Manager_, 0, numLevels);
   setupSolver();
-
 }
 
 // setup coarse grids for new frequency
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setupNormalRAP() {
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setupNormalRAP() {
   // Only setup hierarchy again if preconditioning matrix has changed
-  if( GridTransfersExist_ == false ) {
-    Hierarchy_ = rcp( new Hierarchy(P_)  );
-    if(NullSpace_!=Teuchos::null)
-      Hierarchy_ -> GetLevel(0) -> Set("Nullspace", NullSpace_);
-    if(isSymmetric_==true)
-      Hierarchy_ -> SetImplicitTranspose(true);
-    Hierarchy_ -> SetMaxCoarseSize( coarseGridSize_ );
-    Hierarchy_ -> Setup(*Manager_, 0, numLevels_);
-    GridTransfersExist_=true;
+  if (GridTransfersExist_ == false) {
+    Hierarchy_ = rcp(new Hierarchy(P_));
+    if (NullSpace_ != Teuchos::null)
+      Hierarchy_->GetLevel(0)->Set("Nullspace", NullSpace_);
+    if (isSymmetric_ == true)
+      Hierarchy_->SetImplicitTranspose(true);
+    Hierarchy_->SetMaxCoarseSize(coarseGridSize_);
+    Hierarchy_->Setup(*Manager_, 0, numLevels_);
+    GridTransfersExist_ = true;
   }
   setupSolver();
-
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setupSolver() {
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setupSolver() {
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
   // Define Preconditioner and Operator
-  MueLuOp_ = rcp( new MueLu::ShiftedLaplacianOperator<SC,LO,GO,NO>
-                  (Hierarchy_, A_, ncycles_, subiters_, option_, tol_) );
+  MueLuOp_ = rcp(new MueLu::ShiftedLaplacianOperator<SC, LO, GO, NO>(Hierarchy_, A_, ncycles_, subiters_, option_, tol_));
   // Belos Linear Problem
-  if(LinearProblem_==Teuchos::null)
-    LinearProblem_ = rcp( new LinearProblem );
-  LinearProblem_ -> setOperator (  TpetraA_  );
-  LinearProblem_ -> setRightPrec(  MueLuOp_  );
-  if(SolverManager_==Teuchos::null) {
+  if (LinearProblem_ == Teuchos::null)
+    LinearProblem_ = rcp(new LinearProblem);
+  LinearProblem_->setOperator(TpetraA_);
+  LinearProblem_->setRightPrec(MueLuOp_);
+  if (SolverManager_ == Teuchos::null) {
     std::string solverName;
-    SolverFactory_= rcp( new SolverFactory() );
-    if(solverType_==1)      { solverName="Block GMRES";      }
-    else if(solverType_==2) { solverName="Recycling GMRES";  }
-    else                    { solverName="Flexible GMRES";   }
-    SolverManager_ = SolverFactory_->create( solverName, BelosList_ );
-    SolverManager_ -> setProblem( LinearProblem_ );
+    SolverFactory_ = rcp(new SolverFactory());
+    if (solverType_ == 1) {
+      solverName = "Block GMRES";
+    } else if (solverType_ == 2) {
+      solverName = "Recycling GMRES";
+    } else {
+      solverName = "Flexible GMRES";
+    }
+    SolverManager_ = SolverFactory_->create(solverName, BelosList_);
+    SolverManager_->setProblem(LinearProblem_);
   }
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
 #endif
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::resetLinearProblem()
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::resetLinearProblem() {
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-  LinearProblem_ -> setOperator (  TpetraA_  );
+  LinearProblem_->setOperator(TpetraA_);
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
 #endif
 }
 
 // Solve phase
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::solve(const RCP<TMV> B, RCP<TMV>& X)
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::solve(const RCP<TMV> B, RCP<TMV>& X) {
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
   // Set left and right hand sides for Belos
-  LinearProblem_ -> setProblem(X, B);
+  LinearProblem_->setProblem(X, B);
   // iterative solve
-  SolverManager_ -> solve();
+  SolverManager_->solve();
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
 #endif
@@ -505,33 +475,28 @@ int ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::solve(const RCP<TM
 }
 
 // Solve phase
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::multigrid_apply(const RCP<MultiVector> B,
-                                                                               RCP<MultiVector>& X)
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::multigrid_apply(const RCP<MultiVector> B,
+                                                                                  RCP<MultiVector>& X) {
   // Set left and right hand sides for Belos
-  Hierarchy_ -> Iterate(*B, *X, 1, true, 0);
+  Hierarchy_->Iterate(*B, *X, 1, true, 0);
 }
 
 // Solve phase
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::multigrid_apply(const RCP<Tpetra::MultiVector<SC,LO,GO,NO> > B,
-                                                                               RCP<Tpetra::MultiVector<SC,LO,GO,NO> >& X)
-{
-  Teuchos::RCP< Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > XpetraX
-    = Teuchos::rcp( new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(X) );
-  Teuchos::RCP< Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > XpetraB
-    = Teuchos::rcp( new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(B) );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::multigrid_apply(const RCP<Tpetra::MultiVector<SC, LO, GO, NO> > B,
+                                                                                  RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& X) {
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > XpetraX = Teuchos::rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(X));
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > XpetraB = Teuchos::rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(B));
   // Set left and right hand sides for Belos
-  Hierarchy_ -> Iterate(*XpetraB, *XpetraX, 1, true, 0);
+  Hierarchy_->Iterate(*XpetraB, *XpetraX, 1, true, 0);
 }
 
 // Get most recent iteration count
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetIterations()
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetIterations() {
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-  int numiters = SolverManager_ -> getNumIters();
+  int numiters = SolverManager_->getNumIters();
   return numiters;
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
@@ -540,13 +505,12 @@ int ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetIterations()
 }
 
 // Get most recent solver tolerance achieved
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetResidual()
-{
+ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetResidual() {
   typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-  MT residual = SolverManager_ -> achievedTol();
+  MT residual = SolverManager_->achievedTol();
   return residual;
 #else
   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "ShiftedLaplacian only available with Tpetra and GO=int enabled.");
@@ -554,9 +518,9 @@ ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetResidual()
 #endif
 }
 
-}
+}  // namespace MueLu
 
 #define MUELU_SHIFTEDLAPLACIAN_SHORT
 
-#endif //if defined(HAVE_MUELU_IFPACK2)
-#endif // MUELU_SHIFTEDLAPLACIAN_DEF_HPP
+#endif  // if defined(HAVE_MUELU_IFPACK2)
+#endif  // MUELU_SHIFTEDLAPLACIAN_DEF_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_TpetraOperatorAsRowMatrix.hpp b/packages/muelu/adapters/tpetra/MueLu_TpetraOperatorAsRowMatrix.hpp
index 1852785c6a0c..19cc68d7abe1 100644
--- a/packages/muelu/adapters/tpetra/MueLu_TpetraOperatorAsRowMatrix.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_TpetraOperatorAsRowMatrix.hpp
@@ -46,242 +46,236 @@
 
 namespace MueLu {
 
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class TpetraOperatorAsRowMatrix : public Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-
-  public:
-    using op_type = Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using vec_type = Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-
-    //! The RowMatrix representing the base class of CrsMatrix
-    using row_matrix_type = Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-
-    using impl_scalar_type = typename row_matrix_type::impl_scalar_type;
-    using mag_type = typename Kokkos::ArithTraits<impl_scalar_type>::mag_type;
-
-    using local_inds_device_view_type =
-          typename row_matrix_type::local_inds_device_view_type;
-    using local_inds_host_view_type =
-          typename row_matrix_type::local_inds_host_view_type;
-    using nonconst_local_inds_host_view_type =
-          typename row_matrix_type::nonconst_local_inds_host_view_type;
-
-    using global_inds_device_view_type =
-          typename row_matrix_type::global_inds_device_view_type;
-    using global_inds_host_view_type =
-          typename row_matrix_type::global_inds_host_view_type;
-    using nonconst_global_inds_host_view_type =
-          typename row_matrix_type::nonconst_global_inds_host_view_type;
-
-    using values_device_view_type =
-          typename row_matrix_type::values_device_view_type;
-    using values_host_view_type =
-          typename row_matrix_type::values_host_view_type;
-    using nonconst_values_host_view_type =
-          typename row_matrix_type::nonconst_values_host_view_type;
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    TpetraOperatorAsRowMatrix(const RCP<op_type>& op)
-      :
-      op_(op),
-      diag_(Teuchos::null)
-    { }
-
-    TpetraOperatorAsRowMatrix(const RCP<op_type>& op,
-                        const RCP<vec_type>& diag)
-      :
-      op_(op),
-      diag_(diag)
-    { }
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const {
-      return op_->getDomainMap();
-    }
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const {
-      return op_->getRangeMap();
-    }
-
-    //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
-    /*!
-      \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const {
-      op_->apply(X, Y, mode, alpha, beta);
-    }
-
-    // Fake RowMatrix interface
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRowMap() const {
-      return op_->getRangeMap();
-    }
-
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getColMap() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    typename row_matrix_type::local_ordinal_type getBlockSize() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    Teuchos::RCP<const Teuchos::Comm<int> > getComm() const {
-      return op_->getDomainMap()->getComm();
-    }
-
-    Teuchos::RCP<const Tpetra::RowGraph<LocalOrdinal,GlobalOrdinal,Node> > getGraph() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    Tpetra::global_size_t getGlobalNumRows() const {
-      return getRowMap()->getGlobalNumElements();
-    }
-
-    Tpetra::global_size_t getGlobalNumCols() const {
-      return getDomainMap()->getGlobalNumElements();
-    }
-
-    size_t getLocalNumRows() const {
-      return getRowMap()->getLocalNumElements();
-    }
-
-    size_t getLocalNumCols() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    GlobalOrdinal getIndexBase() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    Tpetra::global_size_t getGlobalNumEntries() const {
-      return 0;
-    }
-
-    size_t getLocalNumEntries() const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    size_t getNumEntriesInGlobalRow (GlobalOrdinal globalRow) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    size_t getNumEntriesInLocalRow (LocalOrdinal localRow) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    size_t getGlobalMaxNumRowEntries () const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    size_t getLocalMaxNumRowEntries () const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    bool hasColMap () const {
-      return false;
-    }
-
-    bool isLocallyIndexed() const {
-      return true;
-    }
-
-    bool isGloballyIndexed() const {
-      return true;
-    }
-
-    bool isFillComplete() const {
-      return true;
-    }
-
-    bool supportsRowViews() const {
-      return false;
-    }
-
-    void
-    getGlobalRowCopy (GlobalOrdinal GlobalRow,
-                      nonconst_global_inds_host_view_type &Indices,
-                      nonconst_values_host_view_type &Values,
-                      size_t& NumEntries) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    void
-    getLocalRowCopy (LocalOrdinal LocalRow,
-                     nonconst_local_inds_host_view_type &Indices,
-                     nonconst_values_host_view_type &Values,
-                     size_t& NumEntries) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    void
-    getGlobalRowView (GlobalOrdinal GlobalRow,
-                      global_inds_host_view_type &indices,
-                      values_host_view_type &values) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    void
-    getLocalRowView (LocalOrdinal LocalRow,
-                     local_inds_host_view_type & indices,
-                     values_host_view_type & values) const {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    void getLocalDiagCopy (Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &diag) const {
-      if (diag_.is_null())
-        throw MueLu::Exceptions::RuntimeError("No diagonal available.");
-      else
-        diag = *diag_;
-    }
-
-    void leftScale (const Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    void rightScale (const Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
-      throw MueLu::Exceptions::RuntimeError("Not implemented.");
-    }
-
-    mag_type getFrobeniusNorm() const {
-      return 0.;
-    }
-
-    // void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
-    //   using std::setw;
-    //   using std::endl;
-    //   const size_t numRows = nearField_->getRowMap()->getGlobalNumElements();
-    //   const size_t nnzNearField = nearField_->getGlobalNumEntries();
-    //   const double nnzNearPerRow = Teuchos::as<double>(nnzNearField)/numRows;
-    //   const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
-    //   const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
-    //   const size_t nnzBasis = basisMatrix_->getGlobalNumEntries();
-    //   size_t nnzTransfer = 0;
-    //   for (size_t i = 0; i<transferMatrices_.size(); i++)
-    //     nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
-    //   const size_t nnzTotal = nnzNearField+nnzKernelApprox+nnzBasis+nnzTransfer;
-    //   const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal)/numRows;
-    //   std::ostringstream oss;
-    //   oss << std::left;
-    //   oss << setw(9) << "rows"  << setw(12) << "nnz(near)"  << setw(14) << "nnz(near)/row" << setw(12) << "nnz(basis)" << setw(15) << "#cluster pairs" << setw(12)<< "nnz(kernel)"    << setw(14) << "nnz(transfer)" << setw(12) << "nnz(total)" << setw(14) << "nnz(total)/row" << endl;
-    //   oss << setw(9) << numRows << setw(12) << nnzNearField << setw(14) << nnzNearPerRow   << setw(12) << nnzBasis     << setw(15) << numClusterPairs  << setw(12) << nnzKernelApprox << setw(14) << nnzTransfer     << setw(12) << nnzTotal     << setw(14) << nnzTotalPerRow   << endl;
-    //   out << oss.str();
-    // }
-
-  private:
-
-    RCP<op_type>  op_;
-    RCP<vec_type> diag_;
-  };
-
-}
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class TpetraOperatorAsRowMatrix : public Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  using op_type  = Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using vec_type = Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  //! The RowMatrix representing the base class of CrsMatrix
+  using row_matrix_type = Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  using impl_scalar_type = typename row_matrix_type::impl_scalar_type;
+  using mag_type         = typename Kokkos::ArithTraits<impl_scalar_type>::mag_type;
+
+  using local_inds_device_view_type =
+      typename row_matrix_type::local_inds_device_view_type;
+  using local_inds_host_view_type =
+      typename row_matrix_type::local_inds_host_view_type;
+  using nonconst_local_inds_host_view_type =
+      typename row_matrix_type::nonconst_local_inds_host_view_type;
+
+  using global_inds_device_view_type =
+      typename row_matrix_type::global_inds_device_view_type;
+  using global_inds_host_view_type =
+      typename row_matrix_type::global_inds_host_view_type;
+  using nonconst_global_inds_host_view_type =
+      typename row_matrix_type::nonconst_global_inds_host_view_type;
+
+  using values_device_view_type =
+      typename row_matrix_type::values_device_view_type;
+  using values_host_view_type =
+      typename row_matrix_type::values_host_view_type;
+  using nonconst_values_host_view_type =
+      typename row_matrix_type::nonconst_values_host_view_type;
+
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  TpetraOperatorAsRowMatrix(const RCP<op_type>& op)
+    : op_(op)
+    , diag_(Teuchos::null) {}
+
+  TpetraOperatorAsRowMatrix(const RCP<op_type>& op,
+                            const RCP<vec_type>& diag)
+    : op_(op)
+    , diag_(diag) {}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const {
+    return op_->getDomainMap();
+  }
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const {
+    return op_->getRangeMap();
+  }
+
+  //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
+  /*!
+    \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
+    op_->apply(X, Y, mode, alpha, beta);
+  }
+
+  // Fake RowMatrix interface
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRowMap() const {
+    return op_->getRangeMap();
+  }
+
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getColMap() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  typename row_matrix_type::local_ordinal_type getBlockSize() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  Teuchos::RCP<const Teuchos::Comm<int> > getComm() const {
+    return op_->getDomainMap()->getComm();
+  }
+
+  Teuchos::RCP<const Tpetra::RowGraph<LocalOrdinal, GlobalOrdinal, Node> > getGraph() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  Tpetra::global_size_t getGlobalNumRows() const {
+    return getRowMap()->getGlobalNumElements();
+  }
+
+  Tpetra::global_size_t getGlobalNumCols() const {
+    return getDomainMap()->getGlobalNumElements();
+  }
+
+  size_t getLocalNumRows() const {
+    return getRowMap()->getLocalNumElements();
+  }
+
+  size_t getLocalNumCols() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  GlobalOrdinal getIndexBase() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  Tpetra::global_size_t getGlobalNumEntries() const {
+    return 0;
+  }
+
+  size_t getLocalNumEntries() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  size_t getNumEntriesInGlobalRow(GlobalOrdinal globalRow) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  size_t getNumEntriesInLocalRow(LocalOrdinal localRow) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  size_t getGlobalMaxNumRowEntries() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  size_t getLocalMaxNumRowEntries() const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  bool hasColMap() const {
+    return false;
+  }
+
+  bool isLocallyIndexed() const {
+    return true;
+  }
+
+  bool isGloballyIndexed() const {
+    return true;
+  }
+
+  bool isFillComplete() const {
+    return true;
+  }
+
+  bool supportsRowViews() const {
+    return false;
+  }
+
+  void
+  getGlobalRowCopy(GlobalOrdinal GlobalRow,
+                   nonconst_global_inds_host_view_type& Indices,
+                   nonconst_values_host_view_type& Values,
+                   size_t& NumEntries) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  void
+  getLocalRowCopy(LocalOrdinal LocalRow,
+                  nonconst_local_inds_host_view_type& Indices,
+                  nonconst_values_host_view_type& Values,
+                  size_t& NumEntries) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  void
+  getGlobalRowView(GlobalOrdinal GlobalRow,
+                   global_inds_host_view_type& indices,
+                   values_host_view_type& values) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  void
+  getLocalRowView(LocalOrdinal LocalRow,
+                  local_inds_host_view_type& indices,
+                  values_host_view_type& values) const {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  void getLocalDiagCopy(Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& diag) const {
+    if (diag_.is_null())
+      throw MueLu::Exceptions::RuntimeError("No diagonal available.");
+    else
+      diag = *diag_;
+  }
+
+  void leftScale(const Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  void rightScale(const Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+    throw MueLu::Exceptions::RuntimeError("Not implemented.");
+  }
+
+  mag_type getFrobeniusNorm() const {
+    return 0.;
+  }
+
+  // void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
+  //   using std::setw;
+  //   using std::endl;
+  //   const size_t numRows = nearField_->getRowMap()->getGlobalNumElements();
+  //   const size_t nnzNearField = nearField_->getGlobalNumEntries();
+  //   const double nnzNearPerRow = Teuchos::as<double>(nnzNearField)/numRows;
+  //   const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
+  //   const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
+  //   const size_t nnzBasis = basisMatrix_->getGlobalNumEntries();
+  //   size_t nnzTransfer = 0;
+  //   for (size_t i = 0; i<transferMatrices_.size(); i++)
+  //     nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
+  //   const size_t nnzTotal = nnzNearField+nnzKernelApprox+nnzBasis+nnzTransfer;
+  //   const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal)/numRows;
+  //   std::ostringstream oss;
+  //   oss << std::left;
+  //   oss << setw(9) << "rows"  << setw(12) << "nnz(near)"  << setw(14) << "nnz(near)/row" << setw(12) << "nnz(basis)" << setw(15) << "#cluster pairs" << setw(12)<< "nnz(kernel)"    << setw(14) << "nnz(transfer)" << setw(12) << "nnz(total)" << setw(14) << "nnz(total)/row" << endl;
+  //   oss << setw(9) << numRows << setw(12) << nnzNearField << setw(14) << nnzNearPerRow   << setw(12) << nnzBasis     << setw(15) << numClusterPairs  << setw(12) << nnzKernelApprox << setw(14) << nnzTransfer     << setw(12) << nnzTotal     << setw(14) << nnzTotalPerRow   << endl;
+  //   out << oss.str();
+  // }
+
+ private:
+  RCP<op_type> op_;
+  RCP<vec_type> diag_;
+};
+
+}  // namespace MueLu
diff --git a/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_decl.hpp b/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_decl.hpp
index 4b5628140372..ecaa3be9ba9d 100644
--- a/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_decl.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_decl.hpp
@@ -56,67 +56,68 @@
 namespace MueLu {
 
 /*!  @brief Wraps an existing MueLu::Hierarchy as a Tpetra::Operator.
-*/
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class TpetraOperator : public Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-  protected:
-    TpetraOperator() = delete;
-  public:
+ */
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class TpetraOperator : public Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ protected:
+  TpetraOperator() = delete;
 
-    //! @name Constructor/Destructor
-    //@{
+ public:
+  //! @name Constructor/Destructor
+  //@{
 
-    //! Constructor
-    TpetraOperator(const RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op) : Operator_(Op){ }
+  //! Constructor
+  TpetraOperator(const RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op)
+    : Operator_(Op) {}
 
-    //! Constructor
-    TpetraOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H) : Hierarchy_(H){ }
+  //! Constructor
+  TpetraOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H) {}
 
-    //! Destructor.
-    virtual ~TpetraOperator() { }
+  //! Destructor.
+  virtual ~TpetraOperator() {}
 
-    //@}
+  //@}
 
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const;
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const;
 
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const;
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const;
 
-    //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
-    /*!
-      \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                         Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                         Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                                         Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                                         Scalar beta  = Teuchos::ScalarTraits<Scalar>::one()) const;
+  //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
+  /*!
+    \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::one()) const;
 
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
 
-    //! @name MueLu specific
-    //@{
+  //! @name MueLu specific
+  //@{
 
-    //! Direct access to the underlying MueLu::Hierarchy.
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetHierarchy() const;
+  //! Direct access to the underlying MueLu::Hierarchy.
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetHierarchy() const;
 
-    //! Direct access to the underlying MueLu::Operator
-    RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetOperator() const;
+  //! Direct access to the underlying MueLu::Operator
+  RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetOperator() const;
 
-    //@}
+  //@}
 
-  private:
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
-    RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Operator_;
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
+  RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Operator_;
+};
 
-  };
+}  // namespace MueLu
 
-} // namespace
-
-#endif // MUELU_TPETRAOPERATOR_DECL_HPP
+#endif  // MUELU_TPETRAOPERATOR_DECL_HPP
diff --git a/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_def.hpp b/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_def.hpp
index 0d2ae1581503..9f029b47d7af 100644
--- a/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_def.hpp
+++ b/packages/muelu/adapters/tpetra/MueLu_TpetraOperator_def.hpp
@@ -60,89 +60,90 @@
 #include "MueLu_Hierarchy.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLu {
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >
-TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getDomainMap() const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >
+TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getDomainMap() const {
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
   typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
   typedef Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node> BlockedMap;
 
   RCP<const Map> domainMap;
-  if(!Hierarchy_.is_null()) domainMap = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A")->getDomainMap();
-  else domainMap = Operator_->getDomainMap();
-
+  if (!Hierarchy_.is_null())
+    domainMap = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A")->getDomainMap();
+  else
+    domainMap = Operator_->getDomainMap();
 
   RCP<const BlockedMap> bDomainMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(domainMap);
-  if(bDomainMap.is_null() == false) {
+  if (bDomainMap.is_null() == false) {
     return Xpetra::toTpetraNonZero(bDomainMap->getFullMap());
   }
   return Xpetra::toTpetraNonZero(domainMap);
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getRangeMap() const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRangeMap() const {
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
   typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
   typedef Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node> BlockedMap;
 
-
   RCP<const Map> rangeMap;
-  if(!Hierarchy_.is_null()) rangeMap = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A")->getRangeMap();
-  else rangeMap = Operator_->getRangeMap();
+  if (!Hierarchy_.is_null())
+    rangeMap = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A")->getRangeMap();
+  else
+    rangeMap = Operator_->getRangeMap();
 
   RCP<const BlockedMap> bRangeMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rangeMap);
-  if(bRangeMap.is_null() == false) {
+  if (bRangeMap.is_null() == false) {
     return Xpetra::toTpetraNonZero(bRangeMap->getFullMap());
   }
   return Xpetra::toTpetraNonZero(rangeMap);
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                                                               Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                                                               Teuchos::ETransp mode, Scalar /* alpha */, Scalar /* beta */) const {
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>       TMV;
-  typedef Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> XTMV;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                                                                      Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                                                                      Teuchos::ETransp mode, Scalar /* alpha */, Scalar /* beta */) const {
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> TMV;
+  typedef Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTMV;
 
-  TEUCHOS_TEST_FOR_EXCEPTION(mode!=Teuchos::NO_TRANS,std::logic_error,"MueLu::TpetraOperator does not support applying the adjoint operator");
+  TEUCHOS_TEST_FOR_EXCEPTION(mode != Teuchos::NO_TRANS, std::logic_error, "MueLu::TpetraOperator does not support applying the adjoint operator");
 
   try {
-    TMV& temp_x = const_cast<TMV &>(X);
+    TMV& temp_x = const_cast<TMV&>(X);
     const XTMV tX(rcpFromRef(temp_x));
-    XTMV       tY(rcpFromRef(Y));
+    XTMV tY(rcpFromRef(Y));
 
-    if(!Hierarchy_.is_null())
+    if (!Hierarchy_.is_null())
       Hierarchy_->Iterate(tX, tY, 1, true);
     else
       Operator_->apply(tX, tY);
 
   } catch (std::exception& e) {
     std::cerr << "MueLu::TpetraOperator::apply : detected an exception" << std::endl
-        << e.what() << std::endl;
+              << e.what() << std::endl;
     throw;
   }
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-bool TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::hasTransposeApply() const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasTransposeApply() const {
   return false;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetHierarchy() const {
+TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetHierarchy() const {
   return Hierarchy_;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetOperator() const {
+TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetOperator() const {
   return Operator_;
 }
 
-} // namespace
+}  // namespace MueLu
 
-#endif //ifdef MUELU_TPETRAOPERATOR_DEF_HPP
+#endif  // ifdef MUELU_TPETRAOPERATOR_DEF_HPP
diff --git a/packages/muelu/doc/MueLu_DoxygenDocumentation.hpp b/packages/muelu/doc/MueLu_DoxygenDocumentation.hpp
index ca314b66e204..270b288d0c03 100644
--- a/packages/muelu/doc/MueLu_DoxygenDocumentation.hpp
+++ b/packages/muelu/doc/MueLu_DoxygenDocumentation.hpp
@@ -115,4 +115,4 @@ The MueLu User's Guide is located in muelu/doc/UsersGuide and at the
 /* ************************************************************************ */
 /* ************************************************************************ */
 
-#endif //ifndef MUELU_DOXYGEN_DOCUMENTATION_HPP
+#endif  // ifndef MUELU_DOXYGEN_DOCUMENTATION_HPP
diff --git a/packages/muelu/doc/Tutorial/src/Challenge.cpp b/packages/muelu/doc/Tutorial/src/Challenge.cpp
index 0ae754397f61..a3db1e26519a 100644
--- a/packages/muelu/doc/Tutorial/src/Challenge.cpp
+++ b/packages/muelu/doc/Tutorial/src/Challenge.cpp
@@ -88,7 +88,6 @@
 #if defined(HAVE_MUELU_EPETRA)
 #include <MueLu_EpetraOperator.hpp>
 
-
 // prescribe types
 // run plain Epetra
 typedef double Scalar;
@@ -105,20 +104,19 @@ typedef Xpetra::EpetraNode Node;
  *
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA)
 #include "MueLu_UseShortNames.hpp"
   using Teuchos::RCP;
   using Teuchos::rcp;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
 
 #ifndef HAVE_XPETRA_INT_LONG_LONG
@@ -130,13 +128,14 @@ int main(int argc, char *argv[]) {
     // =========================================================================
     Teuchos::CommandLineProcessor clp(false);
 
-    std::string xmlFileName = "xml/muelu_ParameterList.xml"; clp.setOption("xml", &xmlFileName, "read parameters from a file [default = 'xml/muelu_ParameterList.xml']");
+    std::string xmlFileName = "xml/muelu_ParameterList.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file [default = 'xml/muelu_ParameterList.xml']");
 
-    int globalNumDofs = 0; //7020;
+    int globalNumDofs = 0;  // 7020;
     clp.setOption("globalNumDofs", &globalNumDofs, "global number of degrees of freedom [has to be set by user, default = 0 -> error]");
     int nDofsPerNode = 1;
     clp.setOption("nDofsPerNode", &nDofsPerNode, "number of degrees of freedom per node [has to be set by user, default = 1]");
-    int nProcs = comm->getSize();
+    int nProcs             = comm->getSize();
     std::string dsolveType = "cg";
     clp.setOption("solver", &dsolveType, "solve type: (none | cg | gmres | standalone) [default = cg]");
     double dtol = 1e-12;
@@ -146,90 +145,88 @@ int main(int argc, char *argv[]) {
     std::string coordsFile = "";
     clp.setOption("coordinates", &coordsFile, "file name containing coordinates in matrix market format");
 
-
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    if(globalNumDofs == 0) {
+    if (globalNumDofs == 0) {
       std::cout << "Please specify '--globalNumDofs'! Simulation cannot run without that parameter correctly set" << std::endl;
       return EXIT_FAILURE;
     }
 
-    int nLocalDofs = (int) globalNumDofs / nProcs;
-    nLocalDofs = nLocalDofs - (nLocalDofs % nDofsPerNode);
+    int nLocalDofs     = (int)globalNumDofs / nProcs;
+    nLocalDofs         = nLocalDofs - (nLocalDofs % nDofsPerNode);
     int nCumulatedDofs = 0;
-    MueLu_sumAll(comm,nLocalDofs, nCumulatedDofs);
+    MueLu_sumAll(comm, nLocalDofs, nCumulatedDofs);
 
-    if(comm->getRank() == nProcs-1) {
+    if (comm->getRank() == nProcs - 1) {
       nLocalDofs += globalNumDofs - nCumulatedDofs;
     }
 
     // read in problem
-    Epetra_Map emap (globalNumDofs, nLocalDofs, 0, *Xpetra::toEpetra(comm));
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_Map emap(globalNumDofs, nLocalDofs, 0, *Xpetra::toEpetra(comm));
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     std::cout << "Reading matrix market file" << std::endl;
 
     std::stringstream ssA, ssB, ssNS;
-    ssA  << problemFile << "_A.txt";
-    ssB  << problemFile << "_b.txt";
+    ssA << problemFile << "_A.txt";
+    ssB << problemFile << "_b.txt";
     ssNS << problemFile << "_ns.txt";
-    std::string fileA = ssA.str();
-    std::string fileB = ssB.str();
+    std::string fileA  = ssA.str();
+    std::string fileB  = ssB.str();
     std::string fileNS = ssNS.str();
-    EpetraExt::MatrixMarketFileToCrsMatrix(fileA.c_str(),emap,emap,emap,ptrA);
-    EpetraExt::MatrixMarketFileToVector(fileB.c_str(),emap,ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix(fileA.c_str(), emap, emap, emap, ptrA);
+    EpetraExt::MatrixMarketFileToVector(fileB.c_str(), emap, ptrf);
     EpetraExt::MatrixMarketFileToMultiVector(fileNS.c_str(), emap, ptrNS);
-    RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    RCP<Epetra_Vector> epB = Teuchos::rcp(ptrf);
+    RCP<Epetra_CrsMatrix> epA    = Teuchos::rcp(ptrA);
+    RCP<Epetra_Vector> epB       = Teuchos::rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = Teuchos::rcp(ptrNS);
 
     // read in coordinates
     RCP<MultiVector> xCoords = Teuchos::null;
-    if(coordsFile != "") {
+    if (coordsFile != "") {
       Epetra_MultiVector* ptrcoords = 0;
-      Epetra_Map coords_emap (globalNumDofs/nDofsPerNode, nLocalDofs/nDofsPerNode, 0, *Xpetra::toEpetra(comm));
+      Epetra_Map coords_emap(globalNumDofs / nDofsPerNode, nLocalDofs / nDofsPerNode, 0, *Xpetra::toEpetra(comm));
       EpetraExt::MatrixMarketFileToMultiVector(coordsFile.c_str(), coords_emap, ptrcoords);
       RCP<Epetra_MultiVector> epCoords = Teuchos::rcp(ptrcoords);
-      xCoords = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epCoords));
+      xCoords                          = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epCoords));
     }
 
     // Epetra_CrsMatrix -> Xpetra::Matrix
-    RCP<CrsMatrix> exA = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int,Node>(epA));
+    RCP<CrsMatrix> exA       = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int, Node>(epA));
     RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
-    RCP<Matrix> Op = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
+    RCP<Matrix> Op           = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
     Op->SetFixedBlockSize(nDofsPerNode);
 
-    RCP<MultiVector> xNS = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epNS));
+    RCP<MultiVector> xNS = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epNS));
 
     // Epetra_Map -> Xpetra::Map
-    const RCP< const Map> map = Xpetra::toXpetra<GO,Node>(emap);
+    const RCP<const Map> map = Xpetra::toXpetra<GO, Node>(emap);
 
-    ParameterListInterpreter mueLuFactory(xmlFileName,*comm);
-    RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
+    ParameterListInterpreter mueLuFactory(xmlFileName, *comm);
+    RCP<Hierarchy> H         = mueLuFactory.CreateHierarchy();
     RCP<MueLu::Level> Finest = H->GetLevel(0);
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);
-    Finest->Set("Nullspace",xNS);
-    if(xCoords != Teuchos::null) Finest->Set("Coordinates",xCoords);
+    Finest->Set("A", Op);
+    Finest->Set("Nullspace", xNS);
+    if (xCoords != Teuchos::null) Finest->Set("Coordinates", xCoords);
 
     mueLuFactory.SetupHierarchy(*H);
 
-
 #ifdef HAVE_MUELU_AZTECOO
 
     H->IsPreconditioner(true);
-    MueLu::EpetraOperator mueluPrec(H); // Wrap MueLu preconditioner into an Epetra Operator
+    MueLu::EpetraOperator mueluPrec(H);  // Wrap MueLu preconditioner into an Epetra Operator
 
     // create a solution vector
     RCP<Epetra_Vector> epX = rcp(new Epetra_Vector(epA->RowMap()));
-    epX->PutScalar((Scalar) 0.0);
+    epX->PutScalar((Scalar)0.0);
 
     Epetra_LinearProblem eProblem(epA.get(), epX.get(), epB.get());
 
@@ -240,28 +237,28 @@ int main(int argc, char *argv[]) {
       solver.SetAztecOption(AZ_solver, AZ_cg);
     else if (dsolveType == "gmres")
       solver.SetAztecOption(AZ_solver, AZ_gmres);
-    else { // use fix point method instead
+    else {  // use fix point method instead
       solver.SetAztecOption(AZ_solver, AZ_fixed_pt);
     }
     solver.SetAztecOption(AZ_output, 1);
 
     solver.Iterate(500, dtol);
 
-    { //TODO: simplify this
-      RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int,Node>(epX));
-      RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int,Node>(epB));
+    {  // TODO: simplify this
+      RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int, Node>(epX));
+      RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int, Node>(epB));
       // Print relative residual norm
       Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = Utilities::ResidualNorm(*Op, *mueluX, *mueluB)[0];
       if (comm->getRank() == 0)
         std::cout << "||Residual|| = " << residualNorms << std::endl;
     }
-#endif // HAVE_MUELU_AZTECOO
+#endif  // HAVE_MUELU_AZTECOO
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   return EXIT_SUCCESS;
 #endif  // #ifdef defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
diff --git a/packages/muelu/doc/Tutorial/src/MLParameterList.cpp b/packages/muelu/doc/Tutorial/src/MLParameterList.cpp
index a4a713670df9..dc3e7469200d 100644
--- a/packages/muelu/doc/Tutorial/src/MLParameterList.cpp
+++ b/packages/muelu/doc/Tutorial/src/MLParameterList.cpp
@@ -47,7 +47,7 @@
 
 #include <MueLu_ConfigDefs.hpp>
 
-#include <Teuchos_XMLParameterListHelpers.hpp> // getParametersFromXmlFile()
+#include <Teuchos_XMLParameterListHelpers.hpp>  // getParametersFromXmlFile()
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
 #include <Epetra_CrsMatrix.h>
 #include <ml_MultiLevelPreconditioner.h>
@@ -91,38 +91,42 @@ int main(int argc, char *argv[]) {
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   //
   // Parameters
   //
 
-  //TODO: FIXME: option by default does not work for MueLu/Tpetra
+  // TODO: FIXME: option by default does not work for MueLu/Tpetra
 
   int nIts = 9;
 
-  Teuchos::CommandLineProcessor clp(false); // Note:
+  Teuchos::CommandLineProcessor clp(false);  // Note:
 
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 256); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);      // manage parameters of xpetra
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 256);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                   // manage parameters of xpetra
 
-  std::string xmlFileName; clp.setOption("xml",   &xmlFileName, "read parameters from a file. Otherwise, this example uses by default an hard-coded parameter list.");
-  int muelu = true;        clp.setOption("muelu", &muelu,       "use muelu"); //TODO: bool instead of int
-  int ml    = true;
+  std::string xmlFileName;
+  clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default an hard-coded parameter list.");
+  int muelu = true;
+  clp.setOption("muelu", &muelu, "use muelu");  // TODO: bool instead of int
+  int ml = true;
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
-  clp.setOption("ml",    &ml,          "use ml");
+  clp.setOption("ml", &ml, "use ml");
 #endif
 
-  switch (clp.parse(argc,argv)) {
-  case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
-  case Teuchos::CommandLineProcessor::PARSE_ERROR:
-  case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-  case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
+    case Teuchos::CommandLineProcessor::PARSE_ERROR:
+    case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   // TODO: check -ml and --linAlgebra
 
-  if (comm->getRank() == 0) { std::cout << xpetraParameters << matrixParameters; }
+  if (comm->getRank() == 0) {
+    std::cout << xpetraParameters << matrixParameters;
+  }
   if (ml && xpetraParameters.GetLib() == Xpetra::UseTpetra) {
     ml = false;
     std::cout << "ML preconditionner can only be built if --linAlgebra=Epetra. Option --ml ignored" << std::endl;
@@ -134,9 +138,9 @@ int main(int argc, char *argv[]) {
 
   // TUTORIALSPLIT ===========================================================
   RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
-  RCP<Matrix>  A = Pr->BuildMatrix();
+  RCP<Matrix> A = Pr->BuildMatrix();
   // TUTORIALSPLIT ===========================================================
 
   //
@@ -146,35 +150,31 @@ int main(int argc, char *argv[]) {
   // ML parameter list
   RCP<Teuchos::ParameterList> params;
   if (xmlFileName != "") {
-
     std::cout << "Reading " << xmlFileName << " ..." << std::endl;
     // TUTORIALSPLIT ===========================================================
     params = Teuchos::getParametersFromXmlFile(xmlFileName);
     // TUTORIALSPLIT ===========================================================
 
   } else {
-
     std::cout << "Using hard-coded parameter list:" << std::endl;
     // TUTORIALSPLIT ===========================================================
     params = rcp(new Teuchos::ParameterList());
 
-    params->set("ML output",  10);
+    params->set("ML output", 10);
     params->set("max levels", 2);
     params->set("smoother: type", "symmetric Gauss-Seidel");
 
     if (xpetraParameters.GetLib() == Xpetra::UseTpetra)
-      params->set("coarse: type","Amesos-Superlu");
+      params->set("coarse: type", "Amesos-Superlu");
     else
-      params->set("coarse: type","Amesos-KLU");
+      params->set("coarse: type", "Amesos-KLU");
     // TUTORIALSPLIT ===========================================================
-
   }
 
   std::cout << "Initial parameter list" << std::endl;
   std::cout << *params << std::endl;
 
   if (muelu) {
-
     //
     // Construct a multigrid preconditioner
     //
@@ -182,27 +182,27 @@ int main(int argc, char *argv[]) {
     // Multigrid Hierarchy
     // TUTORIALSPLIT ===========================================================
     std::string paramXML = MueLu::ML2MueLuParameterTranslator::translate(*params, "");
-    params = Teuchos::getParametersFromXmlString(paramXML);
+    params               = Teuchos::getParametersFromXmlString(paramXML);
     ParameterListInterpreter mueLuFactory(*params);
     RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
     // TUTORIALSPLIT ===========================================================
 
     // build default null space
     LocalOrdinal numPDEs = 1;
-    if(A->IsView("stridedMaps")==true) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
-      oldView = A->SwitchToView(oldView);
+    if (A->IsView("stridedMaps") == true) {
+      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+      numPDEs                     = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
+      oldView                     = A->SwitchToView(oldView);
     }
 
     // TUTORIALSPLIT ===========================================================
     RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
 
-    for (int i=0; i<numPDEs; ++i) {
+    for (int i = 0; i < numPDEs; ++i) {
       Teuchos::ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-      int numBlocks = nsValues.size() / numPDEs;
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*numPDEs + i] = 1.0;
+      int numBlocks                      = nsValues.size() / numPDEs;
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * numPDEs + i] = 1.0;
       }
     }
     // TUTORIALSPLIT ===========================================================
@@ -224,8 +224,9 @@ int main(int argc, char *argv[]) {
     RCP<Vector> X = VectorFactory::Build(map);
     RCP<Vector> B = VectorFactory::Build(map);
 
-    X->putScalar((Scalar) 0.0);
-    B->setSeed(846930886); B->randomize();
+    X->putScalar((Scalar)0.0);
+    B->setSeed(846930886);
+    B->randomize();
 
     // AMG as a standalone solver
     H->IsPreconditioner(false);
@@ -237,31 +238,32 @@ int main(int argc, char *argv[]) {
       std::cout << "||Residual|| = " << residualNorms << std::endl;
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AZTECOO)
-    if (xpetraParameters.GetLib() == Xpetra::UseEpetra) { //TODO: should be doable with Tpetra too
+    if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {  // TODO: should be doable with Tpetra too
 
       // AMG as a preconditioner
 
-      //TODO: name mueluPrec and mlPrec not
+      // TODO: name mueluPrec and mlPrec not
 
       H->IsPreconditioner(true);
-      MueLu::EpetraOperator mueluPrec(H); // Wrap MueLu preconditioner into an Epetra Operator
+      MueLu::EpetraOperator mueluPrec(H);  // Wrap MueLu preconditioner into an Epetra Operator
 
       //
       // Solve Ax = b
       //
-      RCP<Epetra_CrsMatrix> eA; //duplicate code
-      { // TODO: simplify this
-        RCP<CrsMatrixWrap>     xCrsOp  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-        RCP<CrsMatrix>         xCrsMtx = xCrsOp->getCrsMatrix();
-        RCP<EpetraCrsMatrixT<int,Node> >   eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int,Node> >(xCrsMtx, true);
-        eA = eCrsMtx->getEpetra_CrsMatrixNonConst();
+      RCP<Epetra_CrsMatrix> eA;  // duplicate code
+      {                          // TODO: simplify this
+        RCP<CrsMatrixWrap> xCrsOp                 = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+        RCP<CrsMatrix> xCrsMtx                    = xCrsOp->getCrsMatrix();
+        RCP<EpetraCrsMatrixT<int, Node> > eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int, Node> >(xCrsMtx, true);
+        eA                                        = eCrsMtx->getEpetra_CrsMatrixNonConst();
       }
 
       RCP<Epetra_Vector> eX = rcp(new Epetra_Vector(eA->RowMap()));
       RCP<Epetra_Vector> eB = rcp(new Epetra_Vector(eA->RowMap()));
 
-      eX->PutScalar((Scalar) 0.0);
-      eB->SetSeed(846930886); eB->Random();
+      eX->PutScalar((Scalar)0.0);
+      eB->SetSeed(846930886);
+      eB->Random();
 
       Epetra_LinearProblem eProblem(eA.get(), eX.get(), eB.get());
 
@@ -273,7 +275,7 @@ int main(int argc, char *argv[]) {
 
       solver.Iterate(nIts, 1e-10);
 
-      { //TODO: simplify this
+      {  // TODO: simplify this
         RCP<Vector> mueluX = rcp(new Xpetra::EpetraVector(eX));
         RCP<Vector> mueluB = rcp(new Xpetra::EpetraVector(eB));
         // Print relative residual norm
@@ -284,14 +286,20 @@ int main(int argc, char *argv[]) {
 
       // TODO: AMG as a preconditioner (AZ_cg)
     }
-#endif // HAVE_MUELU_AZTECOO
+#endif  // HAVE_MUELU_AZTECOO
 
-  } // if (muelu)
+  }  // if (muelu)
 
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
   if (ml) {
-
-    std::cout << std::endl << std::endl << std::endl << std::endl << "**** ML ml ML ml ML" << std::endl << std::endl << std::endl << std::endl;
+    std::cout << std::endl
+              << std::endl
+              << std::endl
+              << std::endl
+              << "**** ML ml ML ml ML" << std::endl
+              << std::endl
+              << std::endl
+              << std::endl;
 
     //
     // Construct a multigrid preconditioner
@@ -299,18 +307,18 @@ int main(int argc, char *argv[]) {
 
     // Multigrid Hierarchy
     // TUTORIALSPLIT ===========================================================
-    RCP<CrsMatrixWrap>                crsOp         = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-    RCP<CrsMatrix>                    crsMtx        = crsOp->getCrsMatrix();
-    RCP<EpetraCrsMatrixT<int,Node> >  epetraCrsMtx  = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int,Node> >(crsMtx, true);
-    RCP<const Epetra_CrsMatrix> epetra_CrsMtx = epetraCrsMtx->getEpetra_CrsMatrix();
+    RCP<CrsMatrixWrap> crsOp                       = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+    RCP<CrsMatrix> crsMtx                          = crsOp->getCrsMatrix();
+    RCP<EpetraCrsMatrixT<int, Node> > epetraCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int, Node> >(crsMtx, true);
+    RCP<const Epetra_CrsMatrix> epetra_CrsMtx      = epetraCrsMtx->getEpetra_CrsMatrix();
 
     RCP<Epetra_CrsMatrix> eA;
     {
       // TUTORIALSPLIT ===========================================================
-      RCP<CrsMatrixWrap>                 xCrsOp  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-      RCP<CrsMatrix>                     xCrsMtx = xCrsOp->getCrsMatrix();
-      RCP<EpetraCrsMatrixT<int,Node> >   eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int,Node> >(xCrsMtx, true);
-      eA = eCrsMtx->getEpetra_CrsMatrixNonConst();
+      RCP<CrsMatrixWrap> xCrsOp                 = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+      RCP<CrsMatrix> xCrsMtx                    = xCrsOp->getCrsMatrix();
+      RCP<EpetraCrsMatrixT<int, Node> > eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrixT<int, Node> >(xCrsMtx, true);
+      eA                                        = eCrsMtx->getEpetra_CrsMatrixNonConst();
       // TUTORIALSPLIT ===========================================================
     }
     // TUTORIALSPLIT ===========================================================
@@ -325,8 +333,9 @@ int main(int argc, char *argv[]) {
     RCP<Epetra_Vector> eX = rcp(new Epetra_Vector(eA->RowMap()));
     RCP<Epetra_Vector> eB = rcp(new Epetra_Vector(eA->RowMap()));
 
-    eX->PutScalar((Scalar) 0.0);
-    eB->SetSeed(846930886); eB->Random();
+    eX->PutScalar((Scalar)0.0);
+    eB->SetSeed(846930886);
+    eB->Random();
 
     Epetra_LinearProblem eProblem(eA.get(), eX.get(), eB.get());
 
@@ -338,7 +347,7 @@ int main(int argc, char *argv[]) {
 
     solver.Iterate(nIts, 1e-10);
 
-    { //TODO: simplify this
+    {  // TODO: simplify this
       RCP<Vector> mueluX = rcp(new Xpetra::EpetraVector(eX));
       RCP<Vector> mueluB = rcp(new Xpetra::EpetraVector(eB));
       // Print relative residual norm
@@ -350,16 +359,15 @@ int main(int argc, char *argv[]) {
     // TODO: AMG as a preconditioner (AZ_cg)
 #else
     std::cout << "Enable AztecOO to see solution" << std::endl;
-#endif // HAVE_MUELU_AZTECOO
+#endif  // HAVE_MUELU_AZTECOO
 
     std::cout << "Parameter list after ML run" << std::endl;
-    const Teuchos::ParameterList & paramsAfterML = mlPrec->GetList();
+    const Teuchos::ParameterList &paramsAfterML = mlPrec->GetList();
     std::cout << paramsAfterML << std::endl;
 
-  } // if (ml)
-
+  }  // if (ml)
 
-#endif // HAVE_MUELU_ML && HAVE_MUELU_EPETRA
-#endif // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
+#endif  // HAVE_MUELU_ML && HAVE_MUELU_EPETRA
+#endif  // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/doc/Tutorial/src/ScalingTest.cpp b/packages/muelu/doc/Tutorial/src/ScalingTest.cpp
index 742c0a026e65..e60398e8ddec 100644
--- a/packages/muelu/doc/Tutorial/src/ScalingTest.cpp
+++ b/packages/muelu/doc/Tutorial/src/ScalingTest.cpp
@@ -93,8 +93,8 @@
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosXpetraAdapter.hpp" // this header defines Belos::XpetraOp()
-#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
+#include "BelosXpetraAdapter.hpp"  // this header defines Belos::XpetraOp()
+#include "BelosMueLuAdapter.hpp"   // this header defines Belos::MueLuOp()
 #endif
 
 #ifdef HAVE_MUELU_ISORROPIA
@@ -103,10 +103,10 @@
 
 //
 typedef double Scalar;
-typedef int    LocalOrdinal;
-//FIXME we need a HAVE_MUELU_LONG_LONG_INT option
+typedef int LocalOrdinal;
+// FIXME we need a HAVE_MUELU_LONG_LONG_INT option
 //
-// NOTE (mfh 11 Aug 2015) I just added a HAVE_XPETRA_INT_LONG_LONG option.
+//  NOTE (mfh 11 Aug 2015) I just added a HAVE_XPETRA_INT_LONG_LONG option.
 
 #ifdef HAVE_XPETRA_INT_LONG_LONG
 typedef long long int GlobalOrdinal;
@@ -120,9 +120,10 @@ typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 int main(int argc, char *argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  //using Galeri::Xpetra::CreateCartesianCoordinates;
+  // using Galeri::Xpetra::CreateCartesianCoordinates;
 
   Teuchos::oblackholestream blackhole;
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
@@ -134,16 +135,16 @@ int main(int argc, char *argv[]) {
   out->setOutputToRootOnly(0);
   *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
-  // out->setOutputToRootOnly(-1);
-  // out->precision(12);
+// out->setOutputToRootOnly(-1);
+// out->precision(12);
 
-  //FIXME we need a HAVE_MUELU_LONG_LONG_INT option
-  //
-  // NOTE (mfh 11 Aug 2015) I just added a HAVE_XPETRA_INT_LONG_LONG option.
-  //
-  #ifndef HAVE_XPETRA_INT_LONG_LONG
+// FIXME we need a HAVE_MUELU_LONG_LONG_INT option
+//
+//  NOTE (mfh 11 Aug 2015) I just added a HAVE_XPETRA_INT_LONG_LONG option.
+//
+#ifndef HAVE_XPETRA_INT_LONG_LONG
   *out << "Warning: scaling test was not compiled with long long int support" << std::endl;
-  #endif
+#endif
 
   //
   // SET TEST PARAMETERS
@@ -153,56 +154,74 @@ int main(int argc, char *argv[]) {
 
   // Default is Laplace1D with nx = 8748.
   // It's a nice size for 1D and perfect aggregation. (6561 = 3^8)
-  //Nice size for 1D and perfect aggregation on small numbers of processors. (8748 = 4*3^7)
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-  Xpetra::Parameters xpetraParameters(clp);                   // manage parameters of xpetra
+  // Nice size for 1D and perfect aggregation on small numbers of processors. (8748 = 4*3^7)
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
   // Custom command line parameters
-  int optDump    = 0;                     clp.setOption("dump",           &optDump,               "write matrix to file");
-  int optTimings = 0;                     clp.setOption("timings",        &optTimings,            "print timings to screen");
+  int optDump = 0;
+  clp.setOption("dump", &optDump, "write matrix to file");
+  int optTimings = 0;
+  clp.setOption("timings", &optTimings, "print timings to screen");
 
   // - Levels
-  LO  optMaxLevels     = 10;              clp.setOption("maxLevels",      &optMaxLevels,          "maximum number of levels allowed");
-  int optMaxCoarseSize = 50;              clp.setOption("maxCoarseSize",  &optMaxCoarseSize,      "maximum #dofs in coarse operator"); //FIXME clp doesn't like long long int
+  LO optMaxLevels = 10;
+  clp.setOption("maxLevels", &optMaxLevels, "maximum number of levels allowed");
+  int optMaxCoarseSize = 50;
+  clp.setOption("maxCoarseSize", &optMaxCoarseSize, "maximum #dofs in coarse operator");  // FIXME clp doesn't like long long int
 
   // - Smoothed-Aggregation
-  Scalar optSaDamping = 4./3;             clp.setOption("saDamping",      &optSaDamping,          "prolongator damping factor");
+  Scalar optSaDamping = 4. / 3;
+  clp.setOption("saDamping", &optSaDamping, "prolongator damping factor");
 
   // - Aggregation
-  std::string optAggOrdering = "natural"; clp.setOption("aggOrdering",    &optAggOrdering,        "aggregation ordering strategy (natural, random, graph)");
-  int optMinPerAgg = 2;                   clp.setOption("minPerAgg",      &optMinPerAgg,          "minimum #DOFs per aggregate");
-  int optMaxNbrSel = 0;                   clp.setOption("maxNbrSel",      &optMaxNbrSel,          "maximum # of nbrs allowed to be in other aggregates");
+  std::string optAggOrdering = "natural";
+  clp.setOption("aggOrdering", &optAggOrdering, "aggregation ordering strategy (natural, random, graph)");
+  int optMinPerAgg = 2;
+  clp.setOption("minPerAgg", &optMinPerAgg, "minimum #DOFs per aggregate");
+  int optMaxNbrSel = 0;
+  clp.setOption("maxNbrSel", &optMaxNbrSel, "maximum # of nbrs allowed to be in other aggregates");
 
   // - R
-  int optExplicitR = 1;                   clp.setOption("explicitR",      &optExplicitR,          "restriction will be explicitly stored as transpose of prolongator");
+  int optExplicitR = 1;
+  clp.setOption("explicitR", &optExplicitR, "restriction will be explicitly stored as transpose of prolongator");
 
   // - Smoothers
-  std::string optSmooType = "sgs";        clp.setOption("smooType",       &optSmooType,           "smoother type ('l1-sgs', 'sgs 'or 'cheby')");
-  int optSweeps = 2;                      clp.setOption("sweeps",         &optSweeps,             "sweeps to be used in SGS (or Chebyshev degree)");
+  std::string optSmooType = "sgs";
+  clp.setOption("smooType", &optSmooType, "smoother type ('l1-sgs', 'sgs 'or 'cheby')");
+  int optSweeps = 2;
+  clp.setOption("sweeps", &optSweeps, "sweeps to be used in SGS (or Chebyshev degree)");
 
   // - Repartitioning
 #if defined(HAVE_MPI) && defined(HAVE_MUELU_ZOLTAN)
-  int optRepartition = 1;                 clp.setOption("repartition",    &optRepartition,        "enable repartitioning (0=no repartitioning, 1=Zoltan RCB, 2=Isorropia+Zoltan PHG");
-  LO optMinRowsPerProc = 2000;            clp.setOption("minRowsPerProc", &optMinRowsPerProc,     "min #rows allowable per proc before repartitioning occurs");
-  double optNnzImbalance = 1.2;           clp.setOption("nnzImbalance",   &optNnzImbalance,       "max allowable nonzero imbalance before repartitioning occurs");
+  int optRepartition = 1;
+  clp.setOption("repartition", &optRepartition, "enable repartitioning (0=no repartitioning, 1=Zoltan RCB, 2=Isorropia+Zoltan PHG");
+  LO optMinRowsPerProc = 2000;
+  clp.setOption("minRowsPerProc", &optMinRowsPerProc, "min #rows allowable per proc before repartitioning occurs");
+  double optNnzImbalance = 1.2;
+  clp.setOption("nnzImbalance", &optNnzImbalance, "max allowable nonzero imbalance before repartitioning occurs");
 #else
   int optRepartition = 0;
-#endif // HAVE_MPI && HAVE_MUELU_ZOLTAN
+#endif  // HAVE_MPI && HAVE_MUELU_ZOLTAN
 
   // - Solve
-  int    optFixPoint = 1;                 clp.setOption("fixPoint",       &optFixPoint,           "apply multigrid as solver");
-  int    optPrecond  = 1;                 clp.setOption("precond",        &optPrecond,            "apply multigrid as preconditioner");
-  LO     optIts      = 10;                clp.setOption("its",            &optIts,                "number of multigrid cycles");
-  double optTol      = 1e-7;              clp.setOption("tol",            &optTol,                "stopping tolerance for Krylov method");
+  int optFixPoint = 1;
+  clp.setOption("fixPoint", &optFixPoint, "apply multigrid as solver");
+  int optPrecond = 1;
+  clp.setOption("precond", &optPrecond, "apply multigrid as preconditioner");
+  LO optIts = 10;
+  clp.setOption("its", &optIts, "number of multigrid cycles");
+  double optTol = 1e-7;
+  clp.setOption("tol", &optTol, "stopping tolerance for Krylov method");
 
   switch (clp.parse(argc, argv)) {
-  case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
-  case Teuchos::CommandLineProcessor::PARSE_ERROR:
-  case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-  case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
+    case Teuchos::CommandLineProcessor::PARSE_ERROR:
+    case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
-  RCP<TimeMonitor> globalTimeMonitor = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
+  RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
 
   matrixParameters.check();
   xpetraParameters.check();
@@ -226,17 +245,15 @@ int main(int argc, char *argv[]) {
     TimeMonitor tm(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build"));
 
     map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
-    Teuchos::RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+    Teuchos::RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
     A = Pr->BuildMatrix();
 
     if (matrixParameters.GetMatrixType() == "Laplace1D") {
       coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, matrixParameters.GetParameterList());
-    }
-    else if (matrixParameters.GetMatrixType() == "Laplace2D") {
+    } else if (matrixParameters.GetMatrixType() == "Laplace2D") {
       coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, matrixParameters.GetParameterList());
-    }
-    else if (matrixParameters.GetMatrixType() == "Laplace3D") {
+    } else if (matrixParameters.GetMatrixType() == "Laplace3D") {
       coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, matrixParameters.GetParameterList());
     }
   }
@@ -253,7 +270,7 @@ int main(int argc, char *argv[]) {
 
   // USER GUIDE   // define near null space
   RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
-  nullspace->putScalar( (SC) 1.0);
+  nullspace->putScalar((SC)1.0);
   // USER GUIDE   //
   Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
 
@@ -281,7 +298,7 @@ int main(int argc, char *argv[]) {
     // USER GUIDE     // instantiate new Hierarchy object
     H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    H->SetMaxCoarseSize((GO) optMaxCoarseSize);
+    H->SetMaxCoarseSize((GO)optMaxCoarseSize);
     // USER GUIDE     //
 
     //
@@ -291,9 +308,9 @@ int main(int argc, char *argv[]) {
     // USER GUIDE     // create a fine level object
     RCP<Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",           A);
-    Finest->Set("Nullspace",   nullspace);
-    Finest->Set("Coordinates", coordinates); //FIXME: XCoordinates, YCoordinates, ..
+    Finest->Set("A", A);
+    Finest->Set("Nullspace", nullspace);
+    Finest->Set("Coordinates", coordinates);  // FIXME: XCoordinates, YCoordinates, ..
     // USER GUIDE //
 
     //
@@ -314,20 +331,25 @@ int main(int argc, char *argv[]) {
       *out << "========================= Aggregate option summary =========================" << std::endl;
       *out << "min DOFs per aggregate :                " << optMinPerAgg << std::endl;
       *out << "min # of root nbrs already aggregated : " << optMaxNbrSel << std::endl;
-      AggregationFact->SetMinNodesPerAggregate(optMinPerAgg);  //TODO should increase if run anything othpermRFacter than 1D
+      AggregationFact->SetMinNodesPerAggregate(optMinPerAgg);  // TODO should increase if run anything othpermRFacter than 1D
       AggregationFact->SetMaxNeighAlreadySelected(optMaxNbrSel);
       std::transform(optAggOrdering.begin(), optAggOrdering.end(), optAggOrdering.begin(), ::tolower);
       if (optAggOrdering == "natural" || optAggOrdering == "random" || optAggOrdering == "graph") {
         *out << "aggregate ordering :                    " << optAggOrdering << std::endl;
         AggregationFact->SetOrdering(optAggOrdering);
       } else {
-        std::string msg = "main: bad aggregation option """ + optAggOrdering + """.";
+        std::string msg =
+            "main: bad aggregation option "
+            "" +
+            optAggOrdering +
+            ""
+            ".";
         throw(MueLu::Exceptions::RuntimeError(msg));
       }
-      //AggregationFact->SetPhase3AggCreation(0.5);
+      // AggregationFact->SetPhase3AggCreation(0.5);
       M.SetFactory("Aggregates", AggregationFact);
 
-    *out << "=============================================================================" << std::endl;
+      *out << "=============================================================================" << std::endl;
     }
 
     //
@@ -343,7 +365,7 @@ int main(int argc, char *argv[]) {
       RCP<SaPFactory> PFact = rcp(new SaPFactory());
       PFact->SetParameter("sa: damping factor", ParameterEntry(optSaDamping));
 
-      RCP<Factory>    RFact = rcp(new TransPFactory());
+      RCP<Factory> RFact = rcp(new TransPFactory());
 
       RCP<RAPFactory> AFact = rcp(new RAPFactory());
       AFact->setVerbLevel(Teuchos::VERB_HIGH);
@@ -354,7 +376,8 @@ int main(int argc, char *argv[]) {
         ParameterList Aclist = *(AFact->GetValidParameterList());
         Aclist.set("transpose: use implicit", true);
         AFact->SetParameterList(Aclist);
-        if (comm->getRank() == 0) std::cout << "\n\n* ***** USING IMPLICIT RESTRICTION OPERATOR ***** *\n" << std::endl;
+        if (comm->getRank() == 0) std::cout << "\n\n* ***** USING IMPLICIT RESTRICTION OPERATOR ***** *\n"
+                                            << std::endl;
       }
 
       //
@@ -383,13 +406,12 @@ int main(int argc, char *argv[]) {
 
         // Transfer coordinates
         RCP<CoordinatesTransferFactory> TransferCoordinatesFact = rcp(new CoordinatesTransferFactory());
-        AFact->AddTransferFactory(TransferCoordinatesFact); // FIXME REMOVE
+        AFact->AddTransferFactory(TransferCoordinatesFact);  // FIXME REMOVE
 
         // Compute partition (creates "Partition" object)
-        if(optRepartition == 1) { // use plain Zoltan Interface
-
-        } else if (optRepartition == 2) { // use Isorropia + Zoltan interface
+        if (optRepartition == 1) {  // use plain Zoltan Interface
 
+        } else if (optRepartition == 2) {  // use Isorropia + Zoltan interface
         }
 
         // Repartitioning (creates "Importer" from "Partition")
@@ -402,13 +424,12 @@ int main(int argc, char *argv[]) {
         }
         RepartitionFact->SetFactory("A", AFact);
 
-        if(optRepartition == 1) {
+        if (optRepartition == 1) {
           RCP<Factory> ZoltanFact = rcp(new ZoltanInterface());
           ZoltanFact->SetFactory("A", AFact);
           ZoltanFact->SetFactory("Coordinates", TransferCoordinatesFact);
           RepartitionFact->SetFactory("Partition", ZoltanFact);
-        }
-        else if(optRepartition == 2) {
+        } else if (optRepartition == 2) {
 #if defined(HAVE_MPI) && defined(HAVE_MUELU_ISORROPIA)
           RCP<MueLu::IsorropiaInterface<LO, GO, NO> > isoInterface = rcp(new MueLu::IsorropiaInterface<LO, GO, NO>());
           isoInterface->SetFactory("A", AFact);
@@ -421,13 +442,12 @@ int main(int argc, char *argv[]) {
 #endif
         }
 
-
         // Reordering of the transfer operators
         RCP<Factory> RebalancedPFact = rcp(new RebalanceTransferFactory());
         RebalancedPFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Interpolation")));
         RebalancedPFact->SetFactory("P", PFact);
         RebalancedPFact->SetFactory("Coordinates", TransferCoordinatesFact);
-        RebalancedPFact->SetFactory("Nullspace", M.GetFactory("Ptent")); // TODO
+        RebalancedPFact->SetFactory("Nullspace", M.GetFactory("Ptent"));  // TODO
 
         RCP<Factory> RebalancedRFact = rcp(new RebalanceTransferFactory());
         RebalancedRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
@@ -441,16 +461,16 @@ int main(int argc, char *argv[]) {
         M.SetFactory("A", RebalancedAFact);
         M.SetFactory("P", RebalancedPFact);
         M.SetFactory("R", RebalancedRFact);
-        M.SetFactory("Nullspace",   RebalancedPFact);
+        M.SetFactory("Nullspace", RebalancedPFact);
         M.SetFactory("Coordinates", RebalancedPFact);
-        M.SetFactory("Importer",    RepartitionFact);
+        M.SetFactory("Importer", RepartitionFact);
 
 #else
         TEUCHOS_TEST_FOR_EXCEPT(true);
 #endif
-      } // optRepartition
+      }  // optRepartition
 
-    } // Transfer
+    }  // Transfer
 
     //
     // Smoothers
@@ -460,8 +480,8 @@ int main(int argc, char *argv[]) {
       // USER GUIDE       // define smoother object
       std::string ifpackType;
       Teuchos::ParameterList ifpackList;
-      ifpackList.set("relaxation: sweeps", (LO) optSweeps);
-      ifpackList.set("relaxation: damping factor", (SC) 1.0);
+      ifpackList.set("relaxation: sweeps", (LO)optSweeps);
+      ifpackList.set("relaxation: damping factor", (SC)1.0);
       if (optSmooType == "sgs") {
         ifpackType = "RELAXATION";
         ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
@@ -473,16 +493,14 @@ int main(int argc, char *argv[]) {
         ifpackList.set("relaxation: use l1", true);
       } else if (optSmooType == "cheby") {
         ifpackType = "CHEBYSHEV";
-        ifpackList.set("chebyshev: degree", (LO) optSweeps);
+        ifpackList.set("chebyshev: degree", (LO)optSweeps);
 
         if (matrixParameters.GetMatrixType() == "Laplace1D") {
-          ifpackList.set("chebyshev: ratio eigenvalue", (SC) 3);
-        }
-        else if (matrixParameters.GetMatrixType() == "Laplace2D") {
-          ifpackList.set("chebyshev: ratio eigenvalue", (SC) 7);
-        }
-        else if (matrixParameters.GetMatrixType() == "Laplace3D") {
-          ifpackList.set("chebyshev: ratio eigenvalue", (SC) 20);
+          ifpackList.set("chebyshev: ratio eigenvalue", (SC)3);
+        } else if (matrixParameters.GetMatrixType() == "Laplace2D") {
+          ifpackList.set("chebyshev: ratio eigenvalue", (SC)7);
+        } else if (matrixParameters.GetMatrixType() == "Laplace3D") {
+          ifpackList.set("chebyshev: ratio eigenvalue", (SC)20);
         }
         // ifpackList.set("chebyshev: max eigenvalue", (double) -1.0);
         // ifpackList.set("chebyshev: min eigenvalue", (double) 1.0);
@@ -503,7 +521,7 @@ int main(int argc, char *argv[]) {
     H->Setup(M, startLevel, optMaxLevels);
     // USER GUIDE     //
 
-  } // end of Setup TimeMonitor
+  }  // end of Setup TimeMonitor
 
   /*{ // some debug output
     // print out content of levels
@@ -529,7 +547,7 @@ int main(int argc, char *argv[]) {
   X->randomize();
   A->apply(*X, *B, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
   B->norm2(norms);
-  B->scale(1.0/norms[0]);
+  B->scale(1.0 / norms[0]);
   // USER GUIDE   //
 
   //
@@ -537,15 +555,14 @@ int main(int argc, char *argv[]) {
   //
 
   if (optFixPoint) {
-
-    X->putScalar( (SC) 0.0);
+    X->putScalar((SC)0.0);
 
     TimeMonitor tm(*TimeMonitor::getNewTimer("ScalingTest: 3 - Fixed Point Solve"));
 
     H->IsPreconditioner(false);
     H->Iterate(*B, *X, optIts);
 
-  } // optFixedPt
+  }  // optFixedPt
 
   //
   // Use AMG as a preconditioner in Belos
@@ -554,27 +571,27 @@ int main(int argc, char *argv[]) {
 #ifdef HAVE_MUELU_BELOS
 
   if (optPrecond) {
-
     RCP<TimeMonitor> tm;
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Belos Solve")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Belos Solve")));
 
     // USER GUIDE     // Operator and Multivector type that will be used with Belos
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
     H->IsPreconditioner(true);
 
     // Define Operator and Preconditioner
-    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Operator object into a Belos operator
-    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Operator object into a Belos operator
+    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
-    RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+    RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
     belosProblem->setLeftPrec(belosPrec);
 
     bool set = belosProblem->setProblem();
     if (set == false) {
       if (comm->getRank() == 0)
-        std::cout << std::endl << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
+        std::cout << std::endl
+                  << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
       return EXIT_FAILURE;
     }
     // USER GUIDE     //
@@ -582,15 +599,15 @@ int main(int argc, char *argv[]) {
     // USER GUIDE     // Belos parameter list
     int maxIts = 100;
     Teuchos::ParameterList belosList;
-    belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-    belosList.set("Convergence Tolerance", optTol);    // Relative convergence tolerance requested
-    //belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails);
+    belosList.set("Maximum Iterations", maxIts);     // Maximum number of iterations allowed
+    belosList.set("Convergence Tolerance", optTol);  // Relative convergence tolerance requested
+    // belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails);
     belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
     belosList.set("Output Frequency", 1);
     belosList.set("Output Style", Belos::Brief);
 
     // Create an iterative solver manager
-    RCP< Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+    RCP<Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
     // USER GUIDE     //
 
     // Perform solve
@@ -601,7 +618,7 @@ int main(int argc, char *argv[]) {
         // USER GUIDE         // solve linear system
         ret = solver->solve();
         // USER GUIDE         //
-      } // end of TimeMonitor
+      }  // end of TimeMonitor
 
       // Get the number of iterations for this solve.
       if (comm->getRank() == 0)
@@ -609,49 +626,53 @@ int main(int argc, char *argv[]) {
 
       // Compute actual residuals.
       int numrhs = 1;
-      std::vector<double> actual_resids( numrhs ); //TODO: double?
-      std::vector<double> rhs_norm( numrhs );
+      std::vector<double> actual_resids(numrhs);  // TODO: double?
+      std::vector<double> rhs_norm(numrhs);
       RCP<MultiVector> resid = MultiVectorFactory::Build(map, numrhs);
 
-      typedef Belos::OperatorTraits<SC, MV, OP>  OPT;
-      typedef Belos::MultiVecTraits<SC, MV>     MVT;
-
-      OPT::Apply( *belosOp, *X, *resid );
-      MVT::MvAddMv( -1.0, *resid, 1.0, *B, *resid );
-      MVT::MvNorm( *resid, actual_resids );
-      MVT::MvNorm( *B, rhs_norm );
-      *out<< "---------- Actual Residuals (normalized) ----------"<<std::endl<<std::endl;
-      for ( int i = 0; i<numrhs; i++) {
-        double actRes = actual_resids[i]/rhs_norm[i];
-        *out<<"Problem "<<i<<" : \t"<< actRes <<std::endl;
-        //if (actRes > tol) { badRes = true; }
+      typedef Belos::OperatorTraits<SC, MV, OP> OPT;
+      typedef Belos::MultiVecTraits<SC, MV> MVT;
+
+      OPT::Apply(*belosOp, *X, *resid);
+      MVT::MvAddMv(-1.0, *resid, 1.0, *B, *resid);
+      MVT::MvNorm(*resid, actual_resids);
+      MVT::MvNorm(*B, rhs_norm);
+      *out << "---------- Actual Residuals (normalized) ----------" << std::endl
+           << std::endl;
+      for (int i = 0; i < numrhs; i++) {
+        double actRes = actual_resids[i] / rhs_norm[i];
+        *out << "Problem " << i << " : \t" << actRes << std::endl;
+        // if (actRes > tol) { badRes = true; }
       }
 
-    } //try
+    }  // try
 
-    catch(...) {
+    catch (...) {
       if (comm->getRank() == 0)
-        std::cout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+        std::cout << std::endl
+                  << "ERROR:  Belos threw an error! " << std::endl;
     }
 
     // USER GUIDE     // Check convergence
     if (ret != Belos::Converged) {
-      if (comm->getRank() == 0) std::cout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "ERROR:  Belos did not converge! " << std::endl;
     } else {
-      if (comm->getRank() == 0) std::cout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "SUCCESS:  Belos converged!" << std::endl;
     }
     // USER GUIDE     //
     tm = Teuchos::null;
 
-  } //if (optPrecond)
+  }  // if (optPrecond)
 
-#endif // HAVE_MUELU_BELOS
+#endif  // HAVE_MUELU_BELOS
 
   //
   // Timer final summaries
   //
 
-  globalTimeMonitor = Teuchos::null; // stop this timer before summary
+  globalTimeMonitor = Teuchos::null;  // stop this timer before summary
 
   if (optTimings)
     TimeMonitor::summarize();
diff --git a/packages/muelu/doc/Tutorial/src/ScalingTestParamList.cpp b/packages/muelu/doc/Tutorial/src/ScalingTestParamList.cpp
index bf6be6ec1e0d..8285963777d5 100644
--- a/packages/muelu/doc/Tutorial/src/ScalingTestParamList.cpp
+++ b/packages/muelu/doc/Tutorial/src/ScalingTestParamList.cpp
@@ -49,7 +49,7 @@
 
 #include <Teuchos_XMLParameterListHelpers.hpp>
 
-#include <Tpetra_KokkosCompat_DefaultNode.hpp> // For Epetra only runs this points to FakeKokkos in Xpetra
+#include <Tpetra_KokkosCompat_DefaultNode.hpp>  // For Epetra only runs this points to FakeKokkos in Xpetra
 
 #include "Xpetra_ConfigDefs.hpp"
 #include <Xpetra_MultiVectorFactory.hpp>
@@ -65,7 +65,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
@@ -77,8 +77,8 @@
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
 // Define default data types
@@ -87,20 +87,20 @@ typedef int LocalOrdinal;
 typedef int GlobalOrdinal;
 typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   // =========================================================================
   // Convenient definitions
@@ -114,28 +114,40 @@ int main(int argc, char *argv[]) {
   Teuchos::CommandLineProcessor clp(false);
 
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "scalingTest.xml"; clp.setOption("xml",                   &xmlFileName,      "read parameters from a file [default = 'scalingTest.xml']");
-  bool        printTimings      = true;              clp.setOption("timings", "notimings",  &printTimings,     "print timings to screen");
-  int         writeMatricesOPT  = -2;                clp.setOption("write",                 &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
-  std::string dsolveType        = "cg", solveType;   clp.setOption("solver",                &dsolveType,       "solve type: (none | cg | gmres | standalone)");
-  double      dtol              = 1e-12, tol;        clp.setOption("tol",                   &dtol,             "solver convergence tolerance");
-
-  std::string mapFile;                               clp.setOption("map",                   &mapFile,          "map data file");
-  std::string matrixFile;                            clp.setOption("matrix",                &matrixFile,       "matrix data file");
-  std::string coordFile;                             clp.setOption("coords",                &coordFile,        "coordinates data file");
-  std::string nullFile;                              clp.setOption("nullspace",             &nullFile,         "nullspace data file");
-  int         numRebuilds       = 0;                 clp.setOption("rebuild",               &numRebuilds,      "#times to rebuild hierarchy");
-  int         maxIts            = 200;               clp.setOption("its",                   &maxIts,           "maximum number of solver iterations");
-  bool        scaleResidualHistory = true;              clp.setOption("scale", "noscale",  &scaleResidualHistory, "scaled Krylov residual history");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "scalingTest.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file [default = 'scalingTest.xml']");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  int writeMatricesOPT = -2;
+  clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+  std::string dsolveType = "cg", solveType;
+  clp.setOption("solver", &dsolveType, "solve type: (none | cg | gmres | standalone)");
+  double dtol = 1e-12, tol;
+  clp.setOption("tol", &dtol, "solver convergence tolerance");
+
+  std::string mapFile;
+  clp.setOption("map", &mapFile, "map data file");
+  std::string matrixFile;
+  clp.setOption("matrix", &matrixFile, "matrix data file");
+  std::string coordFile;
+  clp.setOption("coords", &coordFile, "coordinates data file");
+  std::string nullFile;
+  clp.setOption("nullspace", &nullFile, "nullspace data file");
+  int numRebuilds = 0;
+  clp.setOption("rebuild", &numRebuilds, "#times to rebuild hierarchy");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  bool scaleResidualHistory = true;
+  clp.setOption("scale", "noscale", &scaleResidualHistory, "scaled Krylov residual history");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
@@ -166,12 +178,13 @@ int main(int argc, char *argv[]) {
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
-  RCP<Matrix>      A;
-  RCP<const Map>   map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<MultiVector> coordinates;
   RCP<MultiVector> nullspace;
   if (matrixFile.empty()) {
-    galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+    galeriStream << "========================================================\n"
+                 << xpetraParameters << galeriParameters;
 
     // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
     //                                 d1  d2  d3
@@ -188,24 +201,24 @@ int main(int argc, char *argv[]) {
     // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
     // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
     if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, galeriList);
 
     } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
                matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, galeriList);
 
     } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, galeriList);
     }
 
     // Expand map to do multiple DOF per node for block problems
     if (matrixType == "Elasticity2D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
     if (matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
     galeriStream << "Processor subdomains in x direction: " << galeriList.get<int>("mx") << std::endl
                  << "Processor subdomains in y direction: " << galeriList.get<int>("my") << std::endl
@@ -214,15 +227,15 @@ int main(int argc, char *argv[]) {
 
     if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
       // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-      galeriList.set("right boundary" , "Neumann");
+      galeriList.set("right boundary", "Neumann");
       galeriList.set("bottom boundary", "Neumann");
-      galeriList.set("top boundary"   , "Neumann");
-      galeriList.set("front boundary" , "Neumann");
-      galeriList.set("back boundary"  , "Neumann");
+      galeriList.set("top boundary", "Neumann");
+      galeriList.set("front boundary", "Neumann");
+      galeriList.set("back boundary", "Neumann");
     }
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
     A = Pr->BuildMatrix();
 
     if (matrixType == "Elasticity2D" ||
@@ -243,7 +256,7 @@ int main(int argc, char *argv[]) {
       // Tpetra matrix reader is still broken, so instead we read in
       // a matrix in a binary format and then redistribute it
       const bool binaryFormat = true;
-      A = Utils::Read(matrixFile, lib, comm, binaryFormat);
+      A                       = Utils::Read(matrixFile, lib, comm, binaryFormat);
 
       if (!map.is_null()) {
         RCP<Matrix> newMatrix = MatrixFactory::Build(map, 1);
@@ -280,11 +293,11 @@ int main(int argc, char *argv[]) {
 
     bool stop = false;
     if (isDriver) {
-      runList   = paramList.sublist("Run1",  mustAlreadyExist);
-      mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+      runList   = paramList.sublist("Run1", mustAlreadyExist);
+      mueluList = runList.sublist("MueLu", mustAlreadyExist);
     } else {
       mueluList = paramList;
-      stop = true;
+      stop      = true;
     }
 
     if (nullspace.is_null()) {
@@ -303,13 +316,13 @@ int main(int argc, char *argv[]) {
       nullspace = MultiVectorFactory::Build(map, blkSize);
       for (int i = 0; i < blkSize; i++) {
         RCP<const Map> domainMap = A->getDomainMap();
-        GO             indexBase = domainMap->getIndexBase();
+        GO indexBase             = domainMap->getIndexBase();
 
         ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
         for (int j = 0; j < nsData.size(); j++) {
           GO GID = domainMap->getGlobalElement(j) - indexBase;
 
-          if ((GID-i) % blkSize == 0)
+          if ((GID - i) % blkSize == 0)
             nsData[j] = Teuchos::ScalarTraits<SC>::one();
         }
       }
@@ -322,7 +335,7 @@ int main(int argc, char *argv[]) {
       solveType = dsolveType;
       tol       = dtol;
 
-      int   savedOut  = -1;
+      int savedOut    = -1;
       FILE* openedOut = NULL;
       if (isDriver) {
         if (runList.isParameter("filename")) {
@@ -339,12 +352,12 @@ int main(int argc, char *argv[]) {
           dup2(fileno(openedOut), STDOUT_FILENO);
         }
         if (runList.isParameter("solver")) solveType = runList.get<std::string>("solver");
-        if (runList.isParameter("tol"))    tol       = runList.get<double>     ("tol");
+        if (runList.isParameter("tol")) tol = runList.get<double>("tol");
       }
 
       // Instead of checking each time for rank, create a rank 0 stream
       RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      Teuchos::FancyOStream& out = *fancy;
+      Teuchos::FancyOStream& out       = *fancy;
       out.setOutputToRootOnly(0);
 
       out << galeriStream.str();
@@ -366,12 +379,12 @@ int main(int argc, char *argv[]) {
         A->SetMaxEigenvalueEstimate(-one);
         //============================================ SPLIT
         H = mueLuFactory->CreateHierarchy();
-        H->GetLevel(0)->Set("A",           A);
-        H->GetLevel(0)->Set("Nullspace",   nullspace);
+        H->GetLevel(0)->Set("A", A);
+        H->GetLevel(0)->Set("Nullspace", nullspace);
         if (!coordinates.is_null())
           H->GetLevel(0)->Set("Coordinates", coordinates);
         mueLuFactory->SetupHierarchy(*H);
-	//============================================ SPLIT
+        //============================================ SPLIT
       }
 
       comm->barrier();
@@ -394,7 +407,7 @@ int main(int argc, char *argv[]) {
 
         Teuchos::Array<STS::magnitudeType> norms(1);
         B->norm2(norms);
-        B->scale(one/norms[0]);
+        B->scale(one / norms[0]);
         X->putScalar(zero);
       }
       tm = Teuchos::null;
@@ -410,7 +423,7 @@ int main(int argc, char *argv[]) {
         // Do not perform a solve
 
       } else if (solveType == "standalone") {
-        tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
+        tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
 
         H->IsPreconditioner(false);
         H->Iterate(*B, *X, maxIts);
@@ -420,17 +433,17 @@ int main(int argc, char *argv[]) {
         tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Belos Solve")));
 
         // Operator and Multivector type that will be used with Belos
-        typedef MultiVector          MV;
+        typedef MultiVector MV;
         typedef Belos::OperatorT<MV> OP;
 
         H->IsPreconditioner(true);
 
         // Define Operator and Preconditioner
-        Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-        Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+        Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+        Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
         // Construct a Belos LinearProblem object
-        RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+        RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
         belosProblem->setRightPrec(belosPrec);
 
         bool set = belosProblem->setProblem();
@@ -441,18 +454,18 @@ int main(int argc, char *argv[]) {
 
         // Belos parameter list
         Teuchos::ParameterList belosList;
-        belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-        belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-        belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-        belosList.set("Output Frequency",      1);
-        belosList.set("Output Style",          Belos::Brief);
+        belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+        belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+        belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+        belosList.set("Output Frequency", 1);
+        belosList.set("Output Style", Belos::Brief);
         if (!scaleResidualHistory)
           belosList.set("Implicit Residual Scaling", "None");
 
         // Create an iterative solver manager
-        RCP< Belos::SolverManager<SC, MV, OP> > solver;
+        RCP<Belos::SolverManager<SC, MV, OP> > solver;
         if (solveType == "cg") {
-          solver = rcp(new Belos::PseudoBlockCGSolMgr   <SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+          solver = rcp(new Belos::PseudoBlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
         } else if (solveType == "gmres") {
           solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
         }
@@ -465,28 +478,30 @@ int main(int argc, char *argv[]) {
           // Get the number of iterations for this solve.
           out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-        }
-        catch(const std::exception& ex)
-        {
-          out << std::endl << "ERROR:  Belos threw an error!  The exception message is:" << std::endl;
+        } catch (const std::exception& ex) {
+          out << std::endl
+              << "ERROR:  Belos threw an error!  The exception message is:" << std::endl;
           std::cout << ex.what() << std::endl;
         }
-        
-        catch(...) {
-          out << std::endl << "ERROR:  Belos threw an unknown error! " << std::endl;
+
+        catch (...) {
+          out << std::endl
+              << "ERROR:  Belos threw an unknown error! " << std::endl;
         }
 
         // Check convergence
         if (ret != Belos::Converged)
-          out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+          out << std::endl
+              << "ERROR:  Belos did not converge! " << std::endl;
         else
-          out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-#endif //ifdef HAVE_MUELU_BELOS
+          out << std::endl
+              << "SUCCESS:  Belos converged!" << std::endl;
+#endif  // ifdef HAVE_MUELU_BELOS
       } else {
         throw MueLu::Exceptions::RuntimeError("Unknown solver type: \"" + solveType + "\"");
       }
       comm->barrier();
-      tm = Teuchos::null;
+      tm                = Teuchos::null;
       globalTimeMonitor = Teuchos::null;
 
       if (printTimings)
@@ -502,7 +517,7 @@ int main(int argc, char *argv[]) {
         }
         try {
           runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
-          mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+          mueluList = runList.sublist("MueLu", mustAlreadyExist);
         } catch (std::exception) {
           stop = true;
         }
@@ -511,6 +526,5 @@ int main(int argc, char *argv[]) {
     } while (stop == false);
   }
 
-
   return 0;
-} //main
+}  // main
diff --git a/packages/muelu/doc/Tutorial/src/laplace2d.cpp b/packages/muelu/doc/Tutorial/src/laplace2d.cpp
index 93f2cdad2302..6d7ae40be5fc 100644
--- a/packages/muelu/doc/Tutorial/src/laplace2d.cpp
+++ b/packages/muelu/doc/Tutorial/src/laplace2d.cpp
@@ -73,7 +73,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
@@ -95,11 +95,11 @@ typedef int GlobalOrdinal;
 typedef Xpetra::EpetraNode Node;
 #endif
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA)
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
@@ -108,9 +108,9 @@ int main(int argc, char *argv[]) {
 
   bool success = false;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int MyPID   = comm->getRank();
-    int NumProc = comm->getSize();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int MyPID                           = comm->getRank();
+    int NumProc                         = comm->getSize();
 
     const Teuchos::RCP<Epetra_Comm> epComm = Teuchos::rcp_const_cast<Epetra_Comm>(Xpetra::toEpetra(comm));
 
@@ -118,33 +118,38 @@ int main(int argc, char *argv[]) {
     // ================================
     // Convenient definitions
     // ================================
-    //SC zero = Teuchos::ScalarTraits<SC>::zero();
+    // SC zero = Teuchos::ScalarTraits<SC>::zero();
     SC one = Teuchos::ScalarTraits<SC>::one();
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& fancyout = *fancy;
+    Teuchos::FancyOStream& fancyout  = *fancy;
     fancyout.setOutputToRootOnly(0);
 
-
-
     // ================================
     // Parameters initialization
     // ================================
     Teuchos::CommandLineProcessor clp(false);
-    GO nx                    = 100;           clp.setOption("nx",                       &nx, "mesh size in x direction");
-    GO ny                    = 100;           clp.setOption("ny",                       &ny, "mesh size in y direction");
-    std::string xmlFileName  = "xml/s2a.xml"; clp.setOption("xml",             &xmlFileName, "read parameters from a file");
-    int mgridSweeps          = 1;             clp.setOption("mgridSweeps",     &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
-    std::string printTimings = "no";          clp.setOption("timings",        &printTimings, "print timings to screen [yes/no]");
-    double tol               = 1e-12;         clp.setOption("tol",                     &tol, "solver convergence tolerance");
-    int importOldData = 0;                    clp.setOption("importOldData", &importOldData, "import map and matrix from previous run (highly experimental).");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    GO nx = 100;
+    clp.setOption("nx", &nx, "mesh size in x direction");
+    GO ny = 100;
+    clp.setOption("ny", &ny, "mesh size in y direction");
+    std::string xmlFileName = "xml/s2a.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file");
+    int mgridSweeps = 1;
+    clp.setOption("mgridSweeps", &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
+    std::string printTimings = "no";
+    clp.setOption("timings", &printTimings, "print timings to screen [yes/no]");
+    double tol = 1e-12;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+    int importOldData = 0;
+    clp.setOption("importOldData", &importOldData, "import map and matrix from previous run (highly experimental).");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // ================================
@@ -160,14 +165,14 @@ int main(int argc, char *argv[]) {
     GaleriList.set("ny", ny);
     GaleriList.set("mx", epComm->NumProc());
     GaleriList.set("my", 1);
-    GaleriList.set("lx", 1.0); // length of x-axis
-    GaleriList.set("ly", 1.0); // length of y-axis
+    GaleriList.set("lx", 1.0);  // length of x-axis
+    GaleriList.set("ly", 1.0);  // length of y-axis
 
-    Teuchos::RCP<Epetra_Map> epMap = Teuchos::null;
+    Teuchos::RCP<Epetra_Map> epMap           = Teuchos::null;
     Teuchos::RCP<Epetra_MultiVector> epCoord = Teuchos::null;
-    Teuchos::RCP<Epetra_CrsMatrix> epA = Teuchos::null;
+    Teuchos::RCP<Epetra_CrsMatrix> epA       = Teuchos::null;
 
-    if(importOldData==0) {
+    if (importOldData == 0) {
       // TUTORIALSPLIT ===========================================================
       // create map
       epMap = Teuchos::rcp(Galeri::CreateMap("Cartesian2D", *epComm, GaleriList));
@@ -178,8 +183,9 @@ int main(int argc, char *argv[]) {
       // create matrix
       epA = Teuchos::rcp(Galeri::CreateCrsMatrix("Laplace2D", epMap.get(), GaleriList));
 
-      double hx = 1./(nx-1); double hy = 1./(ny-1);
-      epA->Scale(1./(hx*hy));
+      double hx = 1. / (nx - 1);
+      double hy = 1. / (ny - 1);
+      epA->Scale(1. / (hx * hy));
       // TUTORIALSPLIT ===========================================================
     } else {
       std::cout << "Import old data" << std::endl;
@@ -192,14 +198,14 @@ int main(int argc, char *argv[]) {
       epCoord = Teuchos::rcp(myEpVector);
       comm->barrier();
       Epetra_CrsMatrix* myEpMatrix;
-      EpetraExt::MatrixMarketFileToCrsMatrix("A.mat",*(Xpetra::toEpetra(comm)), myEpMatrix);
+      EpetraExt::MatrixMarketFileToCrsMatrix("A.mat", *(Xpetra::toEpetra(comm)), myEpMatrix);
       epA = Teuchos::rcp(myEpMatrix);
       comm->barrier();
     }
 
     // TUTORIALSPLIT ===========================================================
     // Epetra -> Xpetra
-    Teuchos::RCP<CrsMatrix> exA = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int,Node>(epA));
+    Teuchos::RCP<CrsMatrix> exA         = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int, Node>(epA));
     Teuchos::RCP<CrsMatrixWrap> exAWrap = Teuchos::rcp(new CrsMatrixWrap(exA));
 
     RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(exAWrap);
@@ -213,16 +219,16 @@ int main(int argc, char *argv[]) {
     X->PutScalar(0.0);
 
     // Epetra -> Xpetra
-    RCP<Vector> xB = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(B));
-    RCP<Vector> xX = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(X));
-    RCP<MultiVector> coords = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epCoord));
+    RCP<Vector> xB          = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(B));
+    RCP<Vector> xX          = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(X));
+    RCP<MultiVector> coords = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epCoord));
 
     xX->setSeed(100);
     xX->randomize();
 
     // TUTORIALSPLIT ===========================================================
     // build null space vector
-    RCP<const Map> map = A->getRowMap();
+    RCP<const Map> map         = A->getRowMap();
     RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
     nullspace->putScalar(one);
 
@@ -248,8 +254,8 @@ int main(int argc, char *argv[]) {
     // TUTORIALSPLIT ===========================================================
     RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
 
-    H->GetLevel(0)->Set("A",           A);
-    H->GetLevel(0)->Set("Nullspace",   nullspace);
+    H->GetLevel(0)->Set("A", A);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
     H->GetLevel(0)->Set("Coordinates", coords);
 
     mueLuFactory.SetupHierarchy(*H);
@@ -270,7 +276,7 @@ int main(int argc, char *argv[]) {
       fancyout << "========================================================\nCalculate exact solution." << std::endl;
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - direct solve")));
       exactLsgVec->PutScalar(0.0);
-      exactLsgVec->Update(1.0,*X,1.0);
+      exactLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), exactLsgVec.get(), B.get());
 
       Amesos amesosFactory;
@@ -295,7 +301,7 @@ int main(int argc, char *argv[]) {
 
       // TUTORIALSPLIT ===========================================================
       precLsgVec->PutScalar(0.0);
-      precLsgVec->Update(1.0,*X,1.0);
+      precLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), precLsgVec.get(), B.get());
 
       AztecOO aztecSolver(epetraProblem);
@@ -319,8 +325,8 @@ int main(int argc, char *argv[]) {
     mgridLsgVec->putScalar(0.0);
     {
       fancyout << "========================================================\nUse multigrid hierarchy as solver." << std::endl;
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
-      mgridLsgVec->update(1.0,*xX,1.0);
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
+      mgridLsgVec->update(1.0, *xX, 1.0);
       H->IsPreconditioner(false);
       H->Iterate(*xB, *mgridLsgVec, mgridSweeps);
       comm->barrier();
@@ -331,68 +337,65 @@ int main(int argc, char *argv[]) {
 
     fancyout << "========================================================\nExport results.\n========================================================" << std::endl;
     std::ofstream myfile;
-    std::stringstream ss; ss << "example" << MyPID << ".txt";
-    myfile.open (ss.str().c_str());
+    std::stringstream ss;
+    ss << "example" << MyPID << ".txt";
+    myfile.open(ss.str().c_str());
 
     //////////////////
 
     // loop over all procs
-    for (int iproc=0; iproc < NumProc; iproc++) {
-      if (MyPID==iproc) {
-        int NumVectors1 = 2;
-        int NumMyElements1 = epCoord->Map(). NumMyElements();
-        int MaxElementSize1 = epCoord->Map().MaxElementSize();
-        int * FirstPointInElementList1 = NULL;
-        if (MaxElementSize1!=1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
-        double ** A_Pointers = epCoord->Pointers();
-
-        if (MyPID==0) {
+    for (int iproc = 0; iproc < NumProc; iproc++) {
+      if (MyPID == iproc) {
+        int NumVectors1               = 2;
+        int NumMyElements1            = epCoord->Map().NumMyElements();
+        int MaxElementSize1           = epCoord->Map().MaxElementSize();
+        int* FirstPointInElementList1 = NULL;
+        if (MaxElementSize1 != 1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
+        double** A_Pointers = epCoord->Pointers();
+
+        if (MyPID == 0) {
           myfile.width(8);
-          myfile <<  "#     MyPID"; myfile << "    ";
+          myfile << "#     MyPID";
+          myfile << "    ";
           myfile.width(12);
-          if (MaxElementSize1==1)
-            myfile <<  "GID  ";
+          if (MaxElementSize1 == 1)
+            myfile << "GID  ";
           else
-            myfile <<  "     GID/Point";
-          for (int j = 0; j < NumVectors1 ; j++)
-          {
+            myfile << "     GID/Point";
+          for (int j = 0; j < NumVectors1; j++) {
             myfile.width(20);
-            myfile <<  "Value  ";
+            myfile << "Value  ";
           }
           myfile << std::endl;
         }
-        for (int i=0; i < NumMyElements1; i++) {
-          for (int ii=0; ii< epCoord->Map().ElementSize(i); ii++) {
+        for (int i = 0; i < NumMyElements1; i++) {
+          for (int ii = 0; ii < epCoord->Map().ElementSize(i); ii++) {
             int iii;
             myfile.width(10);
-            myfile <<  MyPID; myfile << "    ";
+            myfile << MyPID;
+            myfile << "    ";
             myfile.width(10);
-            if (MaxElementSize1==1) {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+            if (MaxElementSize1 == 1) {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
                 myfile << MyGlobalElements1[i] << "    ";
               }
 
               iii = i;
-            }
-            else {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
-                myfile <<  MyGlobalElements1[i]<< "/" << ii << "    ";
+            } else {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+                myfile << MyGlobalElements1[i] << "/" << ii << "    ";
               }
 
-              iii = FirstPointInElementList1[i]+ii;
+              iii = FirstPointInElementList1[i] + ii;
             }
-            for (int j = 0; j < NumVectors1 ; j++)
-            {
+            for (int j = 0; j < NumVectors1; j++) {
               myfile.width(20);
-              myfile <<  A_Pointers[j][iii];
+              myfile << A_Pointers[j][iii];
             }
 
-            myfile.precision(18); // set high precision for output
+            myfile.precision(18);  // set high precision for output
 
             // add solution vector entry
             myfile.width(25);
@@ -406,10 +409,10 @@ int main(int argc, char *argv[]) {
             Teuchos::ArrayRCP<SC> mgridLsgVecData = mgridLsgVec->getDataNonConst(0);
             myfile << mgridLsgVecData[iii];
 
-            myfile.precision(6); // set default precision
+            myfile.precision(6);  // set default precision
             myfile << std::endl;
           }
-        } // end loop over all lines on current proc
+        }  // end loop over all lines on current proc
         myfile << std::flush;
 
         // syncronize procs
@@ -417,19 +420,19 @@ int main(int argc, char *argv[]) {
         comm->barrier();
         comm->barrier();
 
-      } // end myProc
+      }  // end myProc
     }
 
     // export map
-    RCP<const Map> Amap = A->getRowMap();
-    RCP<const Xpetra::EpetraMapT<int,Node> > epAmap = Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int,Node> >(Amap);
+    RCP<const Map> Amap                              = A->getRowMap();
+    RCP<const Xpetra::EpetraMapT<int, Node> > epAmap = Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(Amap);
 
-    //Epetra_Map* eMap;
-    //int rv = EpetraExt::MatrixMarketFileToMap(fileName.c_str(), *(Xpetra::toEpetra(comm)), eMap);
-    EpetraExt::BlockMapToMatrixMarketFile( "ARowMap.mat", epAmap->getEpetra_BlockMap(),
-        "ARowMap",
-        "Row map of matrix A",
-        true);
+    // Epetra_Map* eMap;
+    // int rv = EpetraExt::MatrixMarketFileToMap(fileName.c_str(), *(Xpetra::toEpetra(comm)), eMap);
+    EpetraExt::BlockMapToMatrixMarketFile("ARowMap.mat", epAmap->getEpetra_BlockMap(),
+                                          "ARowMap",
+                                          "Row map of matrix A",
+                                          true);
 
     EpetraExt::MultiVectorToMatrixMarketFile("ACoordVector.mat", *epCoord, "Coordinate multi vector", "Multi vector with mesh coordinates", true);
     EpetraExt::RowMatrixToMatrixMarketFile("A.mat", *epA, "A matrix", "Matrix A", true);
@@ -438,7 +441,7 @@ int main(int argc, char *argv[]) {
     myfile.close();
 
     comm->barrier();
-    tm = Teuchos::null;
+    tm                = Teuchos::null;
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings == "yes") {
@@ -449,8 +452,8 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
-} //main
+#endif  // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
+}  // main
diff --git a/packages/muelu/doc/Tutorial/src/recirc2d.cpp b/packages/muelu/doc/Tutorial/src/recirc2d.cpp
index 92f6d08b47a5..735da741555e 100644
--- a/packages/muelu/doc/Tutorial/src/recirc2d.cpp
+++ b/packages/muelu/doc/Tutorial/src/recirc2d.cpp
@@ -65,7 +65,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
@@ -84,14 +84,14 @@
 typedef double Scalar;
 typedef int LocalOrdinal;
 typedef int GlobalOrdinal;
-typedef Xpetra::EpetraNode Node; // Epetra needs SerialNode
+typedef Xpetra::EpetraNode Node;  // Epetra needs SerialNode
 #endif
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA)
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
@@ -103,41 +103,43 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int MyPID   = comm->getRank();
-    int NumProc = comm->getSize();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int MyPID                           = comm->getRank();
+    int NumProc                         = comm->getSize();
 
     const Teuchos::RCP<Epetra_Comm> epComm = Teuchos::rcp_const_cast<Epetra_Comm>(Xpetra::toEpetra(comm));
 
     // =========================================================================
     // Convenient definitions
     // =========================================================================
-    //SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+    // SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& fancyout = *fancy;
+    Teuchos::FancyOStream& fancyout  = *fancy;
     fancyout.setOutputToRootOnly(0);
 
-
-
     // =========================================================================
     // Parameters initialization
     // =========================================================================
     Teuchos::CommandLineProcessor clp(false);
     GO nx = 100, ny = 100;
-    clp.setOption("nx",                   &nx,              "mesh size in x direction");
-    clp.setOption("ny",                   &ny,              "mesh size in y direction");
-    std::string xmlFileName = "xml/s3a.xml"; clp.setOption("xml", &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'tutorial1a.xml'");
-    int mgridSweeps = 1; clp.setOption("mgridSweeps", &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
-    std::string printTimings = "no";   clp.setOption("timings", &printTimings,     "print timings to screen [yes/no]");
-    double tol               = 1e-12;  clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    clp.setOption("nx", &nx, "mesh size in x direction");
+    clp.setOption("ny", &ny, "mesh size in y direction");
+    std::string xmlFileName = "xml/s3a.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'tutorial1a.xml'");
+    int mgridSweeps = 1;
+    clp.setOption("mgridSweeps", &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
+    std::string printTimings = "no";
+    clp.setOption("timings", &printTimings, "print timings to screen [yes/no]");
+    double tol = 1e-12;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // =========================================================================
@@ -153,8 +155,8 @@ int main(int argc, char *argv[]) {
     GaleriList.set("ny", ny);
     GaleriList.set("mx", epComm->NumProc());
     GaleriList.set("my", 1);
-    GaleriList.set("lx", 1.0); // length of x-axis
-    GaleriList.set("ly", 1.0); // length of y-axis
+    GaleriList.set("lx", 1.0);  // length of x-axis
+    GaleriList.set("ly", 1.0);  // length of y-axis
     GaleriList.set("diff", 1e-5);
     GaleriList.set("conv", 1.0);
 
@@ -168,11 +170,11 @@ int main(int argc, char *argv[]) {
     Teuchos::RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(Galeri::CreateCrsMatrix("Recirc2D", epMap.get(), GaleriList));
 
     // Epetra -> Xpetra
-    Teuchos::RCP<CrsMatrix> exA = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int,Node>(epA));
+    Teuchos::RCP<CrsMatrix> exA         = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int, Node>(epA));
     Teuchos::RCP<CrsMatrixWrap> exAWrap = Teuchos::rcp(new CrsMatrixWrap(exA));
 
     RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(exAWrap);
-    int numPDEs = 1;
+    int numPDEs   = 1;
     A->SetFixedBlockSize(numPDEs);
 
     // set rhs and solution vector
@@ -182,22 +184,22 @@ int main(int argc, char *argv[]) {
     X->PutScalar(0.0);
 
     // Epetra -> Xpetra
-    RCP<Vector> xB = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(B));
-    RCP<Vector> xX = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(X));
-    RCP<MultiVector> coords = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epCoord));
+    RCP<Vector> xB          = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(B));
+    RCP<Vector> xX          = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(X));
+    RCP<MultiVector> coords = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epCoord));
 
     xX->setSeed(100);
     xX->randomize();
 
     // build null space vector
-    RCP<const Map> map = A->getRowMap();
+    RCP<const Map> map         = A->getRowMap();
     RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, numPDEs);
 
-    for (int i=0; i<numPDEs; ++i) {
+    for (int i = 0; i < numPDEs; ++i) {
       Teuchos::ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-      int numBlocks = nsValues.size() / numPDEs;
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*numPDEs + i] = 1.0;
+      int numBlocks                      = nsValues.size() / numPDEs;
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * numPDEs + i] = 1.0;
       }
     }
 
@@ -219,8 +221,8 @@ int main(int argc, char *argv[]) {
 
     RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
 
-    H->GetLevel(0)->Set("A",           A);
-    H->GetLevel(0)->Set("Nullspace",   nullspace);
+    H->GetLevel(0)->Set("A", A);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
     H->GetLevel(0)->Set("Coordinates", coords);
 
     mueLuFactory.SetupHierarchy(*H);
@@ -240,7 +242,7 @@ int main(int argc, char *argv[]) {
       fancyout << "========================================================\nCalculate exact solution." << std::endl;
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - direct solve")));
       exactLsgVec->PutScalar(0.0);
-      exactLsgVec->Update(1.0,*X,1.0);
+      exactLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), exactLsgVec.get(), B.get());
 
       Amesos amesosFactory;
@@ -263,7 +265,7 @@ int main(int argc, char *argv[]) {
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - AMG as preconditioner")));
 
       precLsgVec->PutScalar(0.0);
-      precLsgVec->Update(1.0,*X,1.0);
+      precLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), precLsgVec.get(), B.get());
 
       AztecOO aztecSolver(epetraProblem);
@@ -287,8 +289,8 @@ int main(int argc, char *argv[]) {
     mgridLsgVec->putScalar(0.0);
     {
       fancyout << "========================================================\nUse multigrid hierarchy as solver." << std::endl;
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
-      mgridLsgVec->update(1.0,*xX,1.0);
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
+      mgridLsgVec->update(1.0, *xX, 1.0);
       H->IsPreconditioner(false);
       H->Iterate(*xB, *mgridLsgVec, mgridSweeps);
       comm->barrier();
@@ -299,83 +301,82 @@ int main(int argc, char *argv[]) {
 
     fancyout << "========================================================\nExport results.\n========================================================" << std::endl;
     std::ofstream myfile;
-    std::stringstream ss; ss << "example" << MyPID << ".txt";
-    myfile.open (ss.str().c_str());
+    std::stringstream ss;
+    ss << "example" << MyPID << ".txt";
+    myfile.open(ss.str().c_str());
 
     //////////////////
 
     // loop over all procs
-    for (int iproc=0; iproc < NumProc; iproc++) {
-      if (MyPID==iproc) {
-        int NumVectors1 = 2;
-        int NumMyElements1 = epCoord->Map(). NumMyElements();
-        int MaxElementSize1 = epCoord->Map().MaxElementSize();
-        int * FirstPointInElementList1 = NULL;
-        if (MaxElementSize1!=1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
-        double ** A_Pointers = epCoord->Pointers();
-
-        if (MyPID==0) {
+    for (int iproc = 0; iproc < NumProc; iproc++) {
+      if (MyPID == iproc) {
+        int NumVectors1               = 2;
+        int NumMyElements1            = epCoord->Map().NumMyElements();
+        int MaxElementSize1           = epCoord->Map().MaxElementSize();
+        int* FirstPointInElementList1 = NULL;
+        if (MaxElementSize1 != 1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
+        double** A_Pointers = epCoord->Pointers();
+
+        if (MyPID == 0) {
           myfile.width(8);
-          myfile <<  "#     MyPID"; myfile << "    ";
+          myfile << "#     MyPID";
+          myfile << "    ";
           myfile.width(12);
-          if (MaxElementSize1==1)
-            myfile <<  "GID  ";
+          if (MaxElementSize1 == 1)
+            myfile << "GID  ";
           else
-            myfile <<  "     GID/Point";
-          for (int j = 0; j < NumVectors1 ; j++)
-          {
+            myfile << "     GID/Point";
+          for (int j = 0; j < NumVectors1; j++) {
             myfile.width(20);
-            myfile <<  "Value  ";
+            myfile << "Value  ";
           }
           myfile << std::endl;
         }
-        for (int i=0; i < NumMyElements1; i++) {
-          for (int ii=0; ii< epCoord->Map().ElementSize(i); ii++) {
+        for (int i = 0; i < NumMyElements1; i++) {
+          for (int ii = 0; ii < epCoord->Map().ElementSize(i); ii++) {
             int iii;
             myfile.width(10);
-            myfile <<  MyPID; myfile << "    ";
+            myfile << MyPID;
+            myfile << "    ";
             myfile.width(10);
-            if (MaxElementSize1==1) {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+            if (MaxElementSize1 == 1) {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
                 myfile << MyGlobalElements1[i] << "    ";
               }
 
               iii = i;
-            }
-            else {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
-                myfile <<  MyGlobalElements1[i]<< "/" << ii << "    ";
+            } else {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+                myfile << MyGlobalElements1[i] << "/" << ii << "    ";
               }
 
-              iii = FirstPointInElementList1[i]+ii;
+              iii = FirstPointInElementList1[i] + ii;
             }
-            for (int j = 0; j < NumVectors1 ; j++)
-            {
+            for (int j = 0; j < NumVectors1; j++) {
               myfile.width(20);
-              myfile <<  A_Pointers[j][iii];
+              myfile << A_Pointers[j][iii];
             }
 
-            myfile.precision(18); // set high precision for output
+            myfile.precision(18);  // set high precision for output
 
             // add solution vector entry
-            myfile.width(25); myfile << (*exactLsgVec)[iii];
+            myfile.width(25);
+            myfile << (*exactLsgVec)[iii];
 
             // add preconditioned solution vector entry
-            myfile.width(25); myfile << (*precLsgVec)[iii];
+            myfile.width(25);
+            myfile << (*precLsgVec)[iii];
 
             Teuchos::ArrayRCP<SC> mgridLsgVecData = mgridLsgVec->getDataNonConst(0);
-            myfile.width(25); myfile << mgridLsgVecData[iii];
-
+            myfile.width(25);
+            myfile << mgridLsgVecData[iii];
 
-            myfile.precision(6); // set default precision
+            myfile.precision(6);  // set default precision
             myfile << std::endl;
           }
-        } // end loop over all lines on current proc
+        }  // end loop over all lines on current proc
         myfile << std::flush;
 
         // syncronize procs
@@ -383,14 +384,14 @@ int main(int argc, char *argv[]) {
         comm->barrier();
         comm->barrier();
 
-      } // end myProc
+      }  // end myProc
     }
 
     ////////////
     myfile.close();
 
     comm->barrier();
-    tm = Teuchos::null;
+    tm                = Teuchos::null;
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings == "yes") {
@@ -401,8 +402,8 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
-} //main
+#endif  // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
+}  // main
diff --git a/packages/muelu/doc/Tutorial/src/recirc2d_api.cpp b/packages/muelu/doc/Tutorial/src/recirc2d_api.cpp
index 5379a352b627..26c5a9e0287d 100644
--- a/packages/muelu/doc/Tutorial/src/recirc2d_api.cpp
+++ b/packages/muelu/doc/Tutorial/src/recirc2d_api.cpp
@@ -84,14 +84,14 @@
 typedef double Scalar;
 typedef int LocalOrdinal;
 typedef int GlobalOrdinal;
-typedef Xpetra::EpetraNode Node; // Epetra needs SerialNode
+typedef Xpetra::EpetraNode Node;  // Epetra needs SerialNode
 #endif
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA)
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
@@ -102,20 +102,20 @@ int main(int argc, char *argv[]) {
 
   bool success = false;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int MyPID   = comm->getRank();
-    int NumProc = comm->getSize();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int MyPID                           = comm->getRank();
+    int NumProc                         = comm->getSize();
 
     const Teuchos::RCP<Epetra_Comm> epComm = Teuchos::rcp_const_cast<Epetra_Comm>(Xpetra::toEpetra(comm));
 
     // =========================================================================
     // Convenient definitions
     // =========================================================================
-    //SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+    // SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& fancyout = *fancy;
+    Teuchos::FancyOStream& fancyout  = *fancy;
     fancyout.setOutputToRootOnly(0);
 
     // =========================================================================
@@ -124,20 +124,23 @@ int main(int argc, char *argv[]) {
     Teuchos::CommandLineProcessor clp(false);
     GO nx = 100, ny = 100;
     GO maxCoarseSize = 10;
-    LO maxLevels = 4;
-    clp.setOption("nx",                   &nx,              "mesh size in x direction");
-    clp.setOption("ny",                   &ny,              "mesh size in y direction");
-    clp.setOption("maxCoarseSize",        &maxCoarseSize,   "maximum coarse size");
-    clp.setOption("maxLevels",            &maxLevels,       "maximum number of multigrid levels");
-    int mgridSweeps = 1; clp.setOption("mgridSweeps", &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
-    std::string printTimings = "no";   clp.setOption("timings", &printTimings,     "print timings to screen [yes/no]");
-    double tol               = 1e-12;  clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    LO maxLevels     = 4;
+    clp.setOption("nx", &nx, "mesh size in x direction");
+    clp.setOption("ny", &ny, "mesh size in y direction");
+    clp.setOption("maxCoarseSize", &maxCoarseSize, "maximum coarse size");
+    clp.setOption("maxLevels", &maxLevels, "maximum number of multigrid levels");
+    int mgridSweeps = 1;
+    clp.setOption("mgridSweeps", &mgridSweeps, "number of multigrid sweeps within Multigrid solver.");
+    std::string printTimings = "no";
+    clp.setOption("timings", &printTimings, "print timings to screen [yes/no]");
+    double tol = 1e-12;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // =========================================================================
@@ -153,8 +156,8 @@ int main(int argc, char *argv[]) {
     GaleriList.set("ny", ny);
     GaleriList.set("mx", epComm->NumProc());
     GaleriList.set("my", 1);
-    GaleriList.set("lx", 1.0); // length of x-axis
-    GaleriList.set("ly", 1.0); // length of y-axis
+    GaleriList.set("lx", 1.0);  // length of x-axis
+    GaleriList.set("ly", 1.0);  // length of y-axis
     GaleriList.set("diff", 1e-5);
     GaleriList.set("conv", 1.0);
 
@@ -168,11 +171,11 @@ int main(int argc, char *argv[]) {
     Teuchos::RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(Galeri::CreateCrsMatrix("Recirc2D", epMap.get(), GaleriList));
 
     // Epetra -> Xpetra
-    Teuchos::RCP<CrsMatrix> exA = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int,Node>(epA));
+    Teuchos::RCP<CrsMatrix> exA         = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<int, Node>(epA));
     Teuchos::RCP<CrsMatrixWrap> exAWrap = Teuchos::rcp(new CrsMatrixWrap(exA));
 
     RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(exAWrap);
-    int numPDEs = 1;
+    int numPDEs   = 1;
     A->SetFixedBlockSize(numPDEs);
 
     // set rhs and solution vector
@@ -182,21 +185,21 @@ int main(int argc, char *argv[]) {
     X->PutScalar(0.0);
 
     // Epetra -> Xpetra
-    RCP<Vector> xB = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(B));
-    RCP<Vector> xX = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(X));
+    RCP<Vector> xB = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(B));
+    RCP<Vector> xX = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(X));
 
     xX->setSeed(100);
     xX->randomize();
 
     // build null space vector
-    RCP<const Map> map = A->getRowMap();
+    RCP<const Map> map         = A->getRowMap();
     RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, numPDEs);
 
-    for (int i=0; i<numPDEs; ++i) {
+    for (int i = 0; i < numPDEs; ++i) {
       Teuchos::ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-      int numBlocks = nsValues.size() / numPDEs;
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*numPDEs + i] = 1.0;
+      int numBlocks                      = nsValues.size() / numPDEs;
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * numPDEs + i] = 1.0;
       }
     }
 
@@ -211,19 +214,19 @@ int main(int argc, char *argv[]) {
     comm->barrier();
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1.5 - MueLu read XML")));
 
-    RCP<Hierarchy> H = rcp ( new Hierarchy() );
+    RCP<Hierarchy> H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     // build finest Level
     RCP<MueLu::Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",A);
-    Finest->Set("Nullspace",nullspace);
+    Finest->Set("A", A);
+    Finest->Set("Nullspace", nullspace);
 
     // create factories for transfer operators
     RCP<TentativePFactory> PFact = Teuchos::rcp(new TentativePFactory());
-    RCP<TransPFactory>     RFact = Teuchos::rcp(new TransPFactory());
+    RCP<TransPFactory> RFact     = Teuchos::rcp(new TransPFactory());
     RFact->SetFactory("P", PFact);
 
     // build level smoothers
@@ -231,15 +234,15 @@ int main(int argc, char *argv[]) {
     RCP<SmootherPrototype> smooProto;
     std::string ifpackType;
     Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 1);
-    ifpackList.set("relaxation: damping factor", (SC) 1.0);
+    ifpackList.set("relaxation: sweeps", (LO)1);
+    ifpackList.set("relaxation: damping factor", (SC)1.0);
     ifpackType = "RELAXATION";
     ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
 
-    smooProto = Teuchos::rcp( new TrilinosSmoother(ifpackType, ifpackList) );
+    smooProto = Teuchos::rcp(new TrilinosSmoother(ifpackType, ifpackList));
     RCP<SmootherFactory> SmooFact;
     if (maxLevels > 1)
-      SmooFact = rcp( new SmootherFactory(smooProto) );
+      SmooFact = rcp(new SmootherFactory(smooProto));
 
     // design multigrid hierarchy
     FactoryManager M;
@@ -266,7 +269,7 @@ int main(int argc, char *argv[]) {
       fancyout << "========================================================\nCalculate exact solution." << std::endl;
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - direct solve")));
       exactLsgVec->PutScalar(0.0);
-      exactLsgVec->Update(1.0,*X,1.0);
+      exactLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), exactLsgVec.get(), B.get());
 
       Amesos amesosFactory;
@@ -289,7 +292,7 @@ int main(int argc, char *argv[]) {
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - AMG as preconditioner")));
 
       precLsgVec->PutScalar(0.0);
-      precLsgVec->Update(1.0,*X,1.0);
+      precLsgVec->Update(1.0, *X, 1.0);
       Epetra_LinearProblem epetraProblem(epA.get(), precLsgVec.get(), B.get());
 
       AztecOO aztecSolver(epetraProblem);
@@ -299,7 +302,7 @@ int main(int argc, char *argv[]) {
       aztecSolver.SetPrecOperator(&aztecPrec);
 
       int maxIts = 100;
-      //double tol2 = 1e-8;
+      // double tol2 = 1e-8;
 
       aztecSolver.Iterate(maxIts, tol);
 
@@ -314,8 +317,8 @@ int main(int argc, char *argv[]) {
     mgridLsgVec->putScalar(0.0);
     {
       fancyout << "========================================================\nUse multigrid hierarchy as solver." << std::endl;
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
-      mgridLsgVec->update(1.0,*xX,1.0);
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Multigrid Solve")));
+      mgridLsgVec->update(1.0, *xX, 1.0);
       H->IsPreconditioner(false);
       H->Iterate(*xB, *mgridLsgVec, mgridSweeps);
       comm->barrier();
@@ -326,83 +329,82 @@ int main(int argc, char *argv[]) {
 
     fancyout << "========================================================\nExport results.\n========================================================" << std::endl;
     std::ofstream myfile;
-    std::stringstream ss; ss << "example" << MyPID << ".txt";
-    myfile.open (ss.str().c_str());
+    std::stringstream ss;
+    ss << "example" << MyPID << ".txt";
+    myfile.open(ss.str().c_str());
 
     //////////////////
 
     // loop over all procs
-    for (int iproc=0; iproc < NumProc; iproc++) {
-      if (MyPID==iproc) {
-        int NumVectors1 = 2;
-        int NumMyElements1 = epCoord->Map(). NumMyElements();
-        int MaxElementSize1 = epCoord->Map().MaxElementSize();
-        int * FirstPointInElementList1 = NULL;
-        if (MaxElementSize1!=1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
-        double ** A_Pointers = epCoord->Pointers();
-
-        if (MyPID==0) {
+    for (int iproc = 0; iproc < NumProc; iproc++) {
+      if (MyPID == iproc) {
+        int NumVectors1               = 2;
+        int NumMyElements1            = epCoord->Map().NumMyElements();
+        int MaxElementSize1           = epCoord->Map().MaxElementSize();
+        int* FirstPointInElementList1 = NULL;
+        if (MaxElementSize1 != 1) FirstPointInElementList1 = epCoord->Map().FirstPointInElementList();
+        double** A_Pointers = epCoord->Pointers();
+
+        if (MyPID == 0) {
           myfile.width(8);
-          myfile <<  "#     MyPID"; myfile << "    ";
+          myfile << "#     MyPID";
+          myfile << "    ";
           myfile.width(12);
-          if (MaxElementSize1==1)
-            myfile <<  "GID  ";
+          if (MaxElementSize1 == 1)
+            myfile << "GID  ";
           else
-            myfile <<  "     GID/Point";
-          for (int j = 0; j < NumVectors1 ; j++)
-          {
+            myfile << "     GID/Point";
+          for (int j = 0; j < NumVectors1; j++) {
             myfile.width(20);
-            myfile <<  "Value  ";
+            myfile << "Value  ";
           }
           myfile << std::endl;
         }
-        for (int i=0; i < NumMyElements1; i++) {
-          for (int ii=0; ii< epCoord->Map().ElementSize(i); ii++) {
+        for (int i = 0; i < NumMyElements1; i++) {
+          for (int ii = 0; ii < epCoord->Map().ElementSize(i); ii++) {
             int iii;
             myfile.width(10);
-            myfile <<  MyPID; myfile << "    ";
+            myfile << MyPID;
+            myfile << "    ";
             myfile.width(10);
-            if (MaxElementSize1==1) {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+            if (MaxElementSize1 == 1) {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
                 myfile << MyGlobalElements1[i] << "    ";
               }
 
               iii = i;
-            }
-            else {
-              if(epCoord->Map().GlobalIndicesInt())
-              {
-
-                int * MyGlobalElements1 = epCoord->Map().MyGlobalElements();
-                myfile <<  MyGlobalElements1[i]<< "/" << ii << "    ";
+            } else {
+              if (epCoord->Map().GlobalIndicesInt()) {
+                int* MyGlobalElements1 = epCoord->Map().MyGlobalElements();
+                myfile << MyGlobalElements1[i] << "/" << ii << "    ";
               }
 
-              iii = FirstPointInElementList1[i]+ii;
+              iii = FirstPointInElementList1[i] + ii;
             }
-            for (int j = 0; j < NumVectors1 ; j++)
-            {
+            for (int j = 0; j < NumVectors1; j++) {
               myfile.width(20);
-              myfile <<  A_Pointers[j][iii];
+              myfile << A_Pointers[j][iii];
             }
 
-            myfile.precision(18); // set high precision for output
+            myfile.precision(18);  // set high precision for output
 
             // add solution vector entry
-            myfile.width(25); myfile << (*exactLsgVec)[iii];
+            myfile.width(25);
+            myfile << (*exactLsgVec)[iii];
 
             // add preconditioned solution vector entry
-            myfile.width(25); myfile << (*precLsgVec)[iii];
+            myfile.width(25);
+            myfile << (*precLsgVec)[iii];
 
             Teuchos::ArrayRCP<SC> mgridLsgVecData = mgridLsgVec->getDataNonConst(0);
-            myfile.width(25); myfile << mgridLsgVecData[iii];
-
+            myfile.width(25);
+            myfile << mgridLsgVecData[iii];
 
-            myfile.precision(6); // set default precision
+            myfile.precision(6);  // set default precision
             myfile << std::endl;
           }
-        } // end loop over all lines on current proc
+        }  // end loop over all lines on current proc
         myfile << std::flush;
 
         // syncronize procs
@@ -410,14 +412,14 @@ int main(int argc, char *argv[]) {
         comm->barrier();
         comm->barrier();
 
-      } // end myProc
+      }  // end myProc
     }
 
     ////////////
     myfile.close();
 
     comm->barrier();
-    tm = Teuchos::null;
+    tm                = Teuchos::null;
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings == "yes") {
@@ -428,9 +430,8 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
-} //main
-
+#endif  // #if defined(HAVE_MUELU_EPETRA) and defined(HAVE_MUELU_SERIAL)
+}  // main
diff --git a/packages/muelu/doc/UsersGuide/update_params.sh b/packages/muelu/doc/UsersGuide/update_params.sh
index dfca374e23af..21a647662e6d 100755
--- a/packages/muelu/doc/UsersGuide/update_params.sh
+++ b/packages/muelu/doc/UsersGuide/update_params.sh
@@ -63,6 +63,8 @@ echo '// @HEADER
 // ***********************************************************************
 //
 // @HEADER
+
+// clang-format off
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
 #include "MueLu_Exceptions.hpp"
diff --git a/packages/muelu/example/ParameterList/MLParameterList.cpp b/packages/muelu/example/ParameterList/MLParameterList.cpp
index 3e39d0eecd9b..7a373ef9a616 100644
--- a/packages/muelu/example/ParameterList/MLParameterList.cpp
+++ b/packages/muelu/example/ParameterList/MLParameterList.cpp
@@ -47,9 +47,9 @@
 
 #include <MueLu_ConfigDefs.hpp>
 
-#include <Tpetra_KokkosCompat_DefaultNode.hpp> // For Epetra only runs this points to FakeKokkos in Xpetra
+#include <Tpetra_KokkosCompat_DefaultNode.hpp>  // For Epetra only runs this points to FakeKokkos in Xpetra
 
-#include <Teuchos_XMLParameterListHelpers.hpp> // getParametersFromXmlFile()
+#include <Teuchos_XMLParameterListHelpers.hpp>  // getParametersFromXmlFile()
 
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
 #include <Epetra_CrsMatrix.h>
@@ -80,7 +80,7 @@
 
 // Default problem is Laplace1D with nx = 8748. Use --help to list available options.
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -94,34 +94,38 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     //
     // Parameters
     //
 
-    //TODO: FIXME: option by default does not work for MueLu/Tpetra
+    // TODO: FIXME: option by default does not work for MueLu/Tpetra
 
     int nIts = 9;
 
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 256); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);      // manage parameters of xpetra
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 256);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                   // manage parameters of xpetra
 
-    std::string xmlFileName; clp.setOption("xml",   &xmlFileName, "read parameters from a file. Otherwise, this example uses by default an hard-coded parameter list.");
-    int translatedmuelu = true;  clp.setOption("muelu2", &translatedmuelu,   "use muelu through XML parameter translation and ParameterListInterpreter");
-    int ml    = true;
+    std::string xmlFileName;
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default an hard-coded parameter list.");
+    int translatedmuelu = true;
+    clp.setOption("muelu2", &translatedmuelu, "use muelu through XML parameter translation and ParameterListInterpreter");
+    int ml = true;
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
-    clp.setOption("ml",    &ml,          "use ml");
+    clp.setOption("ml", &ml, "use ml");
 #endif
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    if (comm->getRank() == 0) { std::cout << xpetraParameters << matrixParameters; }
+    if (comm->getRank() == 0) {
+      std::cout << xpetraParameters << matrixParameters;
+    }
 
     // choose ML and Tpetra
     if (ml && xpetraParameters.GetLib() == Xpetra::UseTpetra) {
@@ -134,9 +138,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     //
 
     RCP<const Map> map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
-    RCP<Matrix>  A = Pr->BuildMatrix();
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
+    RCP<Matrix> A = Pr->BuildMatrix();
 
     //
     // Preconditioner configuration
@@ -145,31 +149,28 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // ML parameter list
     RCP<Teuchos::ParameterList> params;
     if (xmlFileName != "") {
-
       std::cout << "Reading " << xmlFileName << " ..." << std::endl;
       params = Teuchos::getParametersFromXmlFile(xmlFileName);
 
     } else {
-
       std::cout << "Using hard-coded parameter list:" << std::endl;
       params = rcp(new Teuchos::ParameterList());
 
-      params->set("ML output",  10);
+      params->set("ML output", 10);
       params->set("max levels", 2);
       params->set("smoother: type", "symmetric Gauss-Seidel");
 
-      if (xpetraParameters.GetLib() == Xpetra::UseTpetra) // TODO: remove 'if' when Amesos2-KLU becomes available
-        params->set("coarse: type","Amesos-Superlu");
+      if (xpetraParameters.GetLib() == Xpetra::UseTpetra)  // TODO: remove 'if' when Amesos2-KLU becomes available
+        params->set("coarse: type", "Amesos-Superlu");
       else
-        params->set("coarse: type","Amesos-KLU");
-
+        params->set("coarse: type", "Amesos-KLU");
     }
 
     std::cout << "Initial parameter list" << std::endl;
     std::cout << *params << std::endl;
 
-    if ( translatedmuelu ) {
-       //
+    if (translatedmuelu) {
+      //
       // Construct a multigrid preconditioner
       //
 
@@ -186,19 +187,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       // build default null space
       LocalOrdinal numPDEs = 1;
-      if(A->IsView("stridedMaps")==true) {
-        Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-        numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
-        oldView = A->SwitchToView(oldView);
+      if (A->IsView("stridedMaps") == true) {
+        Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+        numPDEs                     = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
+        oldView                     = A->SwitchToView(oldView);
       }
 
       RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
 
-      for (int i=0; i<numPDEs; ++i) {
+      for (int i = 0; i < numPDEs; ++i) {
         Teuchos::ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-        int numBlocks = nsValues.size() / numPDEs;
-        for (int j=0; j< numBlocks; ++j) {
-          nsValues[j*numPDEs + i] = 1.0;
+        int numBlocks                      = nsValues.size() / numPDEs;
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + i] = 1.0;
         }
       }
 
@@ -217,8 +218,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       RCP<Vector> X = VectorFactory::Build(map);
       RCP<Vector> B = VectorFactory::Build(map);
 
-      X->putScalar((Scalar) 0.0);
-      B->setSeed(846930886); B->randomize();
+      X->putScalar((Scalar)0.0);
+      B->setSeed(846930886);
+      B->randomize();
 
       // AMG as a standalone solver
       H->IsPreconditioner(false);
@@ -237,31 +239,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // probably a fix necessary in EpetraOperator (which only supports
       // SERIAL or OPENMP, but not PTHREAD of course).
 
-      if (xpetraParameters.GetLib() == Xpetra::UseEpetra) { //TODO: should be doable with Tpetra too
+      if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {  // TODO: should be doable with Tpetra too
 
         // AMG as a preconditioner
 
-        //TODO: name mueluPrec and mlPrec not
+        // TODO: name mueluPrec and mlPrec not
 
         H->IsPreconditioner(true);
-        MueLu::EpetraOperator mueluPrec(H); // Wrap MueLu preconditioner into an Epetra Operator
+        MueLu::EpetraOperator mueluPrec(H);  // Wrap MueLu preconditioner into an Epetra Operator
 
         //
         // Solve Ax = b
         //
-        RCP<Epetra_CrsMatrix> eA; //duplicate code
-        { // TODO: simplify this
-          RCP<CrsMatrixWrap>     xCrsOp  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-          RCP<CrsMatrix>         xCrsMtx = xCrsOp->getCrsMatrix();
-          RCP<EpetraCrsMatrix>   eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(xCrsMtx, true);
-          eA = eCrsMtx->getEpetra_CrsMatrixNonConst();
+        RCP<Epetra_CrsMatrix> eA;  // duplicate code
+        {                          // TODO: simplify this
+          RCP<CrsMatrixWrap> xCrsOp    = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+          RCP<CrsMatrix> xCrsMtx       = xCrsOp->getCrsMatrix();
+          RCP<EpetraCrsMatrix> eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(xCrsMtx, true);
+          eA                           = eCrsMtx->getEpetra_CrsMatrixNonConst();
         }
 
         RCP<Epetra_Vector> eX = rcp(new Epetra_Vector(eA->RowMap()));
         RCP<Epetra_Vector> eB = rcp(new Epetra_Vector(eA->RowMap()));
 
-        eX->PutScalar((Scalar) 0.0);
-        eB->SetSeed(846930886); eB->Random();
+        eX->PutScalar((Scalar)0.0);
+        eB->SetSeed(846930886);
+        eB->Random();
 
         Epetra_LinearProblem eProblem(eA.get(), eX.get(), eB.get());
 
@@ -273,9 +276,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
         solver.Iterate(nIts, 1e-10);
 
-        { //TODO: simplify this
-          RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int,Node>(eX));
-          RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int,Node>(eB));
+        {  // TODO: simplify this
+          RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int, Node>(eX));
+          RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int, Node>(eB));
           // Print relative residual norm
           typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms2 = Utilities::ResidualNorm(*A, *mueluX, *mueluB)[0];
           if (comm->getRank() == 0)
@@ -284,30 +287,36 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
         // TODO: AMG as a preconditioner (AZ_cg)
       }
-#endif // HAVE_MUELU_AZTECOO
-    } // if (translatedmuelu)
+#endif  // HAVE_MUELU_AZTECOO
+    }   // if (translatedmuelu)
 
 #if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AZTECOO) && !defined(HAVE_MUELU_CUDA) && !defined(HAVE_MUELU_HIP) && !defined(HAVE_MUELU_SYCL)
     if (ml) {
-
-      std::cout << std::endl << std::endl << std::endl << std::endl << "**** ML ml ML ml ML" << std::endl << std::endl << std::endl << std::endl;
+      std::cout << std::endl
+                << std::endl
+                << std::endl
+                << std::endl
+                << "**** ML ml ML ml ML" << std::endl
+                << std::endl
+                << std::endl
+                << std::endl;
 
       //
       // Construct a multigrid preconditioner
       //
 
       // Multigrid Hierarchy
-      RCP<CrsMatrixWrap>      crsOp         = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-      RCP<CrsMatrix>        crsMtx        = crsOp->getCrsMatrix();
-      RCP<EpetraCrsMatrix>  epetraCrsMtx  = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(crsMtx, true);
+      RCP<CrsMatrixWrap> crsOp                  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+      RCP<CrsMatrix> crsMtx                     = crsOp->getCrsMatrix();
+      RCP<EpetraCrsMatrix> epetraCrsMtx         = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(crsMtx, true);
       RCP<const Epetra_CrsMatrix> epetra_CrsMtx = epetraCrsMtx->getEpetra_CrsMatrix();
 
       RCP<Epetra_CrsMatrix> eA;
-      { // TODO: simplify this
-        RCP<CrsMatrixWrap>       xCrsOp  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
-        RCP<CrsMatrix>         xCrsMtx = xCrsOp->getCrsMatrix();
-        RCP<EpetraCrsMatrix>   eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(xCrsMtx, true);
-        eA = eCrsMtx->getEpetra_CrsMatrixNonConst();
+      {  // TODO: simplify this
+        RCP<CrsMatrixWrap> xCrsOp    = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A, true);
+        RCP<CrsMatrix> xCrsMtx       = xCrsOp->getCrsMatrix();
+        RCP<EpetraCrsMatrix> eCrsMtx = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(xCrsMtx, true);
+        eA                           = eCrsMtx->getEpetra_CrsMatrixNonConst();
       }
 
       RCP<ML_Epetra::MultiLevelPreconditioner> mlPrec = rcp(new ML_Epetra::MultiLevelPreconditioner(*eA, *params));
@@ -319,8 +328,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       RCP<Epetra_Vector> eX = rcp(new Epetra_Vector(eA->RowMap()));
       RCP<Epetra_Vector> eB = rcp(new Epetra_Vector(eA->RowMap()));
 
-      eX->PutScalar((Scalar) 0.0);
-      eB->SetSeed(846930886); eB->Random();
+      eX->PutScalar((Scalar)0.0);
+      eB->SetSeed(846930886);
+      eB->Random();
 
       Epetra_LinearProblem eProblem(eA.get(), eX.get(), eB.get());
 
@@ -332,9 +342,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       solver.Iterate(nIts, 1e-10);
 
-      { //TODO: simplify this
-        RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int,Node>(eX));
-        RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int,Node>(eB));
+      {  // TODO: simplify this
+        RCP<Vector> mueluX = rcp(new Xpetra::EpetraVectorT<int, Node>(eX));
+        RCP<Vector> mueluB = rcp(new Xpetra::EpetraVectorT<int, Node>(eB));
         // Print relative residual norm
         typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = Utilities::ResidualNorm(*A, *mueluX, *mueluB)[0];
         if (comm->getRank() == 0)
@@ -342,28 +352,26 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       }
 
       std::cout << "Parameter list after ML run" << std::endl;
-      const Teuchos::ParameterList & paramsAfterML = mlPrec->GetList();
+      const Teuchos::ParameterList &paramsAfterML = mlPrec->GetList();
       std::cout << paramsAfterML << std::endl;
 
-    } // if (ml)
-
-#endif // HAVE_MUELU_ML && HAVE_MUELU_EPETRA
+    }  // if (ml)
 
+#endif  // HAVE_MUELU_ML && HAVE_MUELU_EPETRA
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
-int main(int argc, char* argv[]) {
+int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
 
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv);
-  Kokkos::initialize(argc,argv);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
+  Kokkos::initialize(argc, argv);
   try {
     const bool throwExceptions     = false;
     const bool recogniseAllOptions = false;
@@ -371,20 +379,21 @@ int main(int argc, char* argv[]) {
     Teuchos::CommandLineProcessor clp(throwExceptions, recogniseAllOptions);
     Xpetra::Parameters xpetraParameters(clp);
 
-    std::string node = "";  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+    std::string node = "";
+    clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
 
     switch (clp.parse(argc, argv, NULL)) {
-      case Teuchos::CommandLineProcessor::PARSE_ERROR:               return EXIT_FAILURE;
+      case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
 
     if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      return main_<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
+      return main_<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
 #else
       throw MueLu::Exceptions::RuntimeError("Epetra is not available");
 #endif
@@ -396,5 +405,5 @@ int main(int argc, char* argv[]) {
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
   Kokkos::finalize();
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/example/advanced/blockcrs/BlockCrs.cpp b/packages/muelu/example/advanced/blockcrs/BlockCrs.cpp
index 55487c018a5a..157177172ed5 100644
--- a/packages/muelu/example/advanced/blockcrs/BlockCrs.cpp
+++ b/packages/muelu/example/advanced/blockcrs/BlockCrs.cpp
@@ -79,7 +79,6 @@
 using Teuchos::RCP;
 using Teuchos::rcp;
 
-
 //----------------------------------------------------------------------------------------------------------
 //
 // This example demonstrates how to use MueLu in a fashion that looks like ML's LevelWrap
@@ -94,158 +93,155 @@ const std::string thinSeparator  = "--------------------------------------------
 
 const std::string prefSeparator = "=====================================";
 
-
 namespace MueLuExamples {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType diff_vectors(
-      const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> & X,
-      const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> & Y) {
-    RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diff = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(X.getMap());
-    diff->update(1.0,X,-1.0,Y,0.0);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType mgn;
-    Teuchos::Array<mgn> mt(1);
-    diff->norm2(mt);
-    return mt[0];
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType diff_vectors(
+    const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &X,
+    const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &Y) {
+  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diff = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(X.getMap());
+  diff->update(1.0, X, -1.0, Y, 0.0);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType mgn;
+  Teuchos::Array<mgn> mt(1);
+  diff->norm2(mt);
+  return mt[0];
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType compute_resid_norm(
-      const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A,
-      const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> & X,
-      const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> & B) {
-    RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > temp = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(X.getMap());
-    A.apply(X,*temp);
-    temp->update(1.0,B,-1.0);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType mgn;
-    Teuchos::Array<mgn> mt(1);
-    temp->norm2(mt);
-    return mt[0];
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType compute_resid_norm(
+    const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A,
+    const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &X,
+    const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B) {
+  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > temp = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(X.getMap());
+  A.apply(X, *temp);
+  temp->update(1.0, B, -1.0);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType mgn;
+  Teuchos::Array<mgn> mt(1);
+  temp->norm2(mt);
+  return mt[0];
+}
 
 // --------------------------------------------------------------------------------------
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void solve_system_belos(
-      RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A,
-      RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & X,
-      RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & B,
-      Teuchos::ParameterList & MueLuList,
-      const std::string & belos_solver,
-      RCP<Teuchos::ParameterList> & SList) {
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    typedef Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_Operator;
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_CrsMatrix;
-    typedef Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_BlockCrsMatrix;
-    typedef Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Xpetra_TpetraBlockCrsMatrix;
-    typedef Tpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_Vector;
-    typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_MultiVector;
-
-    RCP<Tpetra_Operator>    At = MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Op2NonConstTpetraRow(A);
-    RCP<Tpetra_Operator>    Mt = MueLu::CreateTpetraPreconditioner(At,MueLuList);
-    RCP<Tpetra_MultiVector> Xt = Xpetra::toTpetra(*X);
-    RCP<Tpetra_MultiVector> Bt = Xpetra::toTpetra(*B);
-
-    if(Xt.is_null() || Bt.is_null() || At.is_null() || Mt.is_null()) throw std::runtime_error("ERROR: Xpetra to Tpetra conversion failed");
-
-    typedef Tpetra_MultiVector MV;
-    typedef Tpetra_Operator OP;
-    RCP<Belos::LinearProblem<Scalar,MV,OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar,MV,OP>(At, Xt, Bt));
-    belosProblem->setLeftPrec(Mt);
-    belosProblem->setProblem(Xt,Bt);
-
-    Belos::SolverFactory<Scalar, MV, OP> BelosFactory;
-    Teuchos::RCP<Belos::SolverManager<Scalar, MV, OP> > BelosSolver = BelosFactory.create(belos_solver, SList);
-    BelosSolver->setProblem(belosProblem);
-    BelosSolver->solve();
-  }
-
- // --------------------------------------------------------------------------------------
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void solve_system_ifpack2(
-      RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A,
-      RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & X,
-      RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & B,
-      const std::string & ifpack2_solver,
-      Teuchos::ParameterList & Ifpack2List) {
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    typedef Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_Operator;
-    typedef Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> Ifpack2_Preconditioner;
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_CrsMatrix;
-    typedef Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_RowMatrix;
-    typedef Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_BlockCrsMatrix;
-    typedef Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> Xpetra_TpetraBlockCrsMatrix;
-    typedef Tpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_Vector;
-    typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> Tpetra_MultiVector;
-
-    RCP<Tpetra_RowMatrix>    At = MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Op2NonConstTpetraRow(A);
-    RCP<Tpetra_MultiVector> Xt = Xpetra::toTpetra(*X);
-    RCP<Tpetra_MultiVector> Bt = Xpetra::toTpetra(*B);
-
-    RCP<Ifpack2_Preconditioner> Solver = Ifpack2::Factory::create<Tpetra_RowMatrix>(ifpack2_solver,At);
-    Solver->setParameters(Ifpack2List);
-    Solver->initialize();
-    Solver->compute();
-
-    Solver->apply(*Bt,*Xt);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void solve_system_belos(
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &A,
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &X,
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &B,
+    Teuchos::ParameterList &MueLuList,
+    const std::string &belos_solver,
+    RCP<Teuchos::ParameterList> &SList) {
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_Operator;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_CrsMatrix;
+  typedef Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_BlockCrsMatrix;
+  typedef Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Xpetra_TpetraBlockCrsMatrix;
+  typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_Vector;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_MultiVector;
+
+  RCP<Tpetra_Operator> At    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraRow(A);
+  RCP<Tpetra_Operator> Mt    = MueLu::CreateTpetraPreconditioner(At, MueLuList);
+  RCP<Tpetra_MultiVector> Xt = Xpetra::toTpetra(*X);
+  RCP<Tpetra_MultiVector> Bt = Xpetra::toTpetra(*B);
+
+  if (Xt.is_null() || Bt.is_null() || At.is_null() || Mt.is_null()) throw std::runtime_error("ERROR: Xpetra to Tpetra conversion failed");
+
+  typedef Tpetra_MultiVector MV;
+  typedef Tpetra_Operator OP;
+  RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar, MV, OP>(At, Xt, Bt));
+  belosProblem->setLeftPrec(Mt);
+  belosProblem->setProblem(Xt, Bt);
+
+  Belos::SolverFactory<Scalar, MV, OP> BelosFactory;
+  Teuchos::RCP<Belos::SolverManager<Scalar, MV, OP> > BelosSolver = BelosFactory.create(belos_solver, SList);
+  BelosSolver->setProblem(belosProblem);
+  BelosSolver->solve();
+}
 
-  // --------------------------------------------------------------------------------------
-  // This routine generate's the user's original A matrix and nullspace
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void generate_user_matrix_and_nullspace(std::string &matrixType,  Xpetra::UnderlyingLib & lib,Teuchos::ParameterList &galeriList,  RCP<const Teuchos::Comm<int> > &comm, RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A, RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & nullspace){
-    using Teuchos::RCP;
+// --------------------------------------------------------------------------------------
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void solve_system_ifpack2(
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &A,
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &X,
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &B,
+    const std::string &ifpack2_solver,
+    Teuchos::ParameterList &Ifpack2List) {
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  typedef Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_Operator;
+  typedef Ifpack2::Preconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node> Ifpack2_Preconditioner;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_CrsMatrix;
+  typedef Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_RowMatrix;
+  typedef Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_BlockCrsMatrix;
+  typedef Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Xpetra_TpetraBlockCrsMatrix;
+  typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_Vector;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_MultiVector;
+
+  RCP<Tpetra_RowMatrix> At   = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraRow(A);
+  RCP<Tpetra_MultiVector> Xt = Xpetra::toTpetra(*X);
+  RCP<Tpetra_MultiVector> Bt = Xpetra::toTpetra(*B);
+
+  RCP<Ifpack2_Preconditioner> Solver = Ifpack2::Factory::create<Tpetra_RowMatrix>(ifpack2_solver, At);
+  Solver->setParameters(Ifpack2List);
+  Solver->initialize();
+  Solver->compute();
+
+  Solver->apply(*Bt, *Xt);
+}
 
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
-
-    typedef typename Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> map_type;
-    typedef typename Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> multivector_type;
-    typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> realvaluedmultivector_type;
-    typedef typename Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> matrixwrap_type;
-    RCP<const map_type >   map;
-    RCP<multivector_type > coordinates;
-    if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,map_type,multivector_type>("1D", map, galeriList);
-
-    } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,map_type,multivector_type>("2D", map, galeriList);
-
-    } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,map_type,multivector_type>("3D", map, galeriList);
-    }
+// --------------------------------------------------------------------------------------
+// This routine generate's the user's original A matrix and nullspace
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void generate_user_matrix_and_nullspace(std::string &matrixType, Xpetra::UnderlyingLib &lib, Teuchos::ParameterList &galeriList, RCP<const Teuchos::Comm<int> > &comm, RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &A, RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &nullspace) {
+  using Teuchos::RCP;
 
-    // Expand map to do multiple DOF per node for block problems
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
+  RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  Teuchos::FancyOStream &out       = *fancy;
+
+  typedef typename Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+  typedef typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> multivector_type;
+  typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> realvaluedmultivector_type;
+  typedef typename Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> matrixwrap_type;
+  RCP<const map_type> map;
+  RCP<multivector_type> coordinates;
+  if (matrixType == "Laplace1D") {
+    map         = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar, LocalOrdinal, GlobalOrdinal, map_type, multivector_type>("1D", map, galeriList);
+
+  } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+    map         = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar, LocalOrdinal, GlobalOrdinal, map_type, multivector_type>("2D", map, galeriList);
+
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+    map         = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar, LocalOrdinal, GlobalOrdinal, map_type, multivector_type>("3D", map, galeriList);
+  }
 
-    out << "Processor subdomains in x direction: " << galeriList.get<GlobalOrdinal>("mx") << std::endl
-        << "Processor subdomains in y direction: " << galeriList.get<GlobalOrdinal>("my") << std::endl
-        << "Processor subdomains in z direction: " << galeriList.get<GlobalOrdinal>("mz") << std::endl
-        << "========================================================" << std::endl;
+  // Expand map to do multiple DOF per node for block problems
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
+    map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
 
-    RCP<Galeri::Xpetra::Problem<map_type,matrixwrap_type,multivector_type> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar,LocalOrdinal,GlobalOrdinal,map_type,matrixwrap_type,multivector_type>(matrixType, map, galeriList);
+  out << "Processor subdomains in x direction: " << galeriList.get<GlobalOrdinal>("mx") << std::endl
+      << "Processor subdomains in y direction: " << galeriList.get<GlobalOrdinal>("my") << std::endl
+      << "Processor subdomains in z direction: " << galeriList.get<GlobalOrdinal>("mz") << std::endl
+      << "========================================================" << std::endl;
 
-    A = Pr->BuildMatrix();
+  RCP<Galeri::Xpetra::Problem<map_type, matrixwrap_type, multivector_type> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, map_type, matrixwrap_type, multivector_type>(matrixType, map, galeriList);
 
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
-      nullspace = Pr->BuildNullspace();
-      A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
-    }
+  A = Pr->BuildMatrix();
 
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
+    nullspace = Pr->BuildNullspace();
+    A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
   }
-
 }
 
+}  // namespace MueLuExamples
 
 // --------------------------------------------------------------------------------------
-//int main(int argc, char *argv[]) {
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// int main(int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
@@ -257,47 +253,49 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream &out       = *fancy;
 
     typedef Teuchos::ScalarTraits<SC> STS;
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
-    //Teuchos::CommandLineProcessor clp(false);
+    // Teuchos::CommandLineProcessor clp(false);
 
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-    std::string matFileName       = "";  clp.setOption("matrix",&matFileName,"read matrix from a file");
-    LO blocksize = 1;                    clp.setOption("blocksize",&blocksize,"block size");
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+    std::string matFileName = "";
+    clp.setOption("matrix", &matFileName, "read matrix from a file");
+    LO blocksize = 1;
+    clp.setOption("blocksize", &blocksize, "block size");
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    //Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
+    // Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
     Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
 
-    if(lib!=Xpetra::UseTpetra)
+    if (lib != Xpetra::UseTpetra)
       throw std::runtime_error("This test only works with Tpetra linear algebra");
 
     // =========================================================================
     // Problem construction
     // =========================================================================
-    RCP<const Map>   map;
+    RCP<const Map> map;
     RCP<Matrix> A;
     RCP<MultiVector> nullspace;
 
-    typedef Tpetra::CrsMatrix<SC,LO,GO,NO> Tpetra_CrsMatrix;
-    typedef Tpetra::Operator<SC,LO,GO,NO> Tpetra_Operator;
-    typedef Tpetra::BlockCrsMatrix<SC,LO,GO,NO> Tpetra_BlockCrsMatrix;
-    typedef Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> Xpetra_TpetraBlockCrsMatrix;
-    typedef Xpetra::CrsMatrix<SC,LO,GO,NO> Xpetra_CrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO> Xpetra_CrsMatrixWrap;
+    typedef Tpetra::CrsMatrix<SC, LO, GO, NO> Tpetra_CrsMatrix;
+    typedef Tpetra::Operator<SC, LO, GO, NO> Tpetra_Operator;
+    typedef Tpetra::BlockCrsMatrix<SC, LO, GO, NO> Tpetra_BlockCrsMatrix;
+    typedef Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> Xpetra_TpetraBlockCrsMatrix;
+    typedef Xpetra::CrsMatrix<SC, LO, GO, NO> Xpetra_CrsMatrix;
+    typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> Xpetra_CrsMatrixWrap;
     typedef typename Teuchos::ScalarTraits<SC>::magnitudeType SCN;
 
     SC one = Teuchos::ScalarTraits<SC>::one();
@@ -305,94 +303,95 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<Tpetra_CrsMatrix> Acrs;
     RCP<Tpetra_BlockCrsMatrix> Ablock;
 
-    if(matFileName.length() > 0) {
+    if (matFileName.length() > 0) {
       // Read matrix from disk
-      out << thickSeparator << std::endl << "Reading matrix from disk" <<std::endl;
+      out << thickSeparator << std::endl
+          << "Reading matrix from disk" << std::endl;
       // TODO
       typedef Tpetra::MatrixMarket::Reader<Tpetra_CrsMatrix> reader_type;
-      Acrs = reader_type::readSparseFile(matFileName,comm);
-    }
-    else{
+      Acrs = reader_type::readSparseFile(matFileName, comm);
+    } else {
       // Use Galeri
-      out << thickSeparator << std::endl << xpetraParameters << galeriParameters;
+      out << thickSeparator << std::endl
+          << xpetraParameters << galeriParameters;
       std::string matrixType = galeriParameters.GetMatrixType();
-      RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Axp;
-      MueLuExamples::generate_user_matrix_and_nullspace<Scalar,LocalOrdinal,GlobalOrdinal,Node>(matrixType,lib,galeriList,comm,Axp,nullspace);
-      Acrs = Xpetra::Helpers<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Axp);
+      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Axp;
+      MueLuExamples::generate_user_matrix_and_nullspace<Scalar, LocalOrdinal, GlobalOrdinal, Node>(matrixType, lib, galeriList, comm, Axp, nullspace);
+      Acrs = Xpetra::Helpers<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Axp);
     }
     // Block this bad boy
-    Ablock = Tpetra::convertToBlockCrsMatrix<SC,LO,GO,NO>(*Acrs,blocksize);
+    Ablock = Tpetra::convertToBlockCrsMatrix<SC, LO, GO, NO>(*Acrs, blocksize);
 
     // Now wrap BlockCrs to Xpetra::Matrix
     RCP<Xpetra_CrsMatrix> Axt = rcp(new Xpetra_TpetraBlockCrsMatrix(Ablock));
-    A = rcp(new Xpetra_CrsMatrixWrap(Axt));
+    A                         = rcp(new Xpetra_CrsMatrixWrap(Axt));
 
     // =========================================================================
     // Setups and solves
     // =========================================================================
-    map=Xpetra::toXpetra(Acrs->getRowMap());
+    map = Xpetra::toXpetra(Acrs->getRowMap());
 
     RCP<Vector> X1 = VectorFactory::Build(map);
     RCP<Vector> X2 = VectorFactory::Build(map);
-    RCP<Vector> B = VectorFactory::Build(map);
+    RCP<Vector> B  = VectorFactory::Build(map);
     B->setSeed(846930886);
     B->randomize();
     RCP<TimeMonitor> tm;
 
     // Belos Options
-    RCP<Teuchos::ParameterList> SList = rcp(new Teuchos::ParameterList );
-    SList->set("Verbosity",Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    SList->set("Output Frequency",10);
-    SList->set("Output Style",Belos::Brief);
-    SList->set("Maximum Iterations",10);
-    SList->set("Convergence Tolerance",5e-2);
+    RCP<Teuchos::ParameterList> SList = rcp(new Teuchos::ParameterList);
+    SList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    SList->set("Output Frequency", 10);
+    SList->set("Output Style", Belos::Brief);
+    SList->set("Maximum Iterations", 10);
+    SList->set("Convergence Tolerance", 5e-2);
 
     // =========================================================================
     // Solve #1 (fixed point + Jacobi)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 1: Fixed Point + Jacobi"<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 1: Fixed Point + Jacobi" << prefSeparator << std::endl;
     {
       Teuchos::ParameterList MueList;
-      MueList.set("max levels",1);
+      MueList.set("max levels", 1);
       MueList.set("coarse: type", "RELAXATION");
 
       std::string belos_solver("Fixed Point");
-      MueLuExamples::solve_system_belos<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A,X1,B,MueList,belos_solver,SList);
+      MueLuExamples::solve_system_belos<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, X1, B, MueList, belos_solver, SList);
       std::cout << "I" << std::endl;
-      SCN result = MueLuExamples::compute_resid_norm<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*A,*X1,*B);
-      out<<"Solve #1: Residual Norm = "<<result<<std::endl;
+      SCN result = MueLuExamples::compute_resid_norm<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*A, *X1, *B);
+      out << "Solve #1: Residual Norm = " << result << std::endl;
     }
 
     // =========================================================================
     // Solve #2 (striaght up Jacobi)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 2: Fixed Jacobi"<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 2: Fixed Jacobi" << prefSeparator << std::endl;
     {
       Teuchos::ParameterList IList;
-      IList.set("relaxation: type","Jacobi");
-      IList.set("relaxation: damping factor",one);
-      IList.set("relaxation: sweeps",10);
+      IList.set("relaxation: type", "Jacobi");
+      IList.set("relaxation: damping factor", one);
+      IList.set("relaxation: sweeps", 10);
       std::string ifpack2_precond("RELAXATION");
 
-      MueLuExamples::solve_system_ifpack2(A,X2,B,ifpack2_precond,IList);
+      MueLuExamples::solve_system_ifpack2(A, X2, B, ifpack2_precond, IList);
 
-      SCN result = MueLuExamples::compute_resid_norm<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*A,*X2,*B);
-      out<<"Solve #2: Residual Norm = "<<result<<std::endl;
+      SCN result = MueLuExamples::compute_resid_norm<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*A, *X2, *B);
+      out << "Solve #2: Residual Norm = " << result << std::endl;
     }
 
     // Compare 1 & 2
-    SCN norm = MueLuExamples::diff_vectors<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*X1,*X2);
-    if(norm > 1e-10) {
-      out<<"ERROR: Norm of Solve #1 and Solve #2 differs by "<<norm<<std::endl;
-      success=false;
+    SCN norm = MueLuExamples::diff_vectors<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*X1, *X2);
+    if (norm > 1e-10) {
+      out << "ERROR: Norm of Solve #1 and Solve #2 differs by " << norm << std::endl;
+      success = false;
     }
 
-  }//end try
+  }  // end try
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
@@ -400,7 +399,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
diff --git a/packages/muelu/example/advanced/clone/Clone-Tpetra.cpp b/packages/muelu/example/advanced/clone/Clone-Tpetra.cpp
index 46e701a4ccd9..8334ea68e3a0 100644
--- a/packages/muelu/example/advanced/clone/Clone-Tpetra.cpp
+++ b/packages/muelu/example/advanced/clone/Clone-Tpetra.cpp
@@ -61,7 +61,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 #include "MueLu_SmootherFactory.hpp"
 #include "MueLu_TrilinosSmoother.hpp"
 #include "MueLu_Ifpack2Smoother.hpp"
@@ -74,15 +74,15 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-int main(int argc, char *argv[]) {
-  #include <MueLu_UseDefaultTypes.hpp>
-  #include <MueLu_UseShortNames.hpp>
+int main(int argc, char* argv[]) {
+#include <MueLu_UseDefaultTypes.hpp>
+#include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
@@ -90,7 +90,7 @@ int main(int argc, char *argv[]) {
   // MPI initialization using Teuchos
   // =========================================================================
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   // =========================================================================
   // Convenient definitions
@@ -99,38 +99,46 @@ int main(int argc, char *argv[]) {
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& fancyout = *fancy;
+  Teuchos::FancyOStream& fancyout  = *fancy;
   fancyout.setOutputToRootOnly(0);
 
-
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   Teuchos::CommandLineProcessor clp(false);
 
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName = "scalingTest.xml"; clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
-  int    amgAsPrecond     = 1;                 clp.setOption("precond",               &amgAsPrecond,     "apply multigrid as preconditioner");
-  int    amgAsSolver      = 0;                 clp.setOption("fixPoint",              &amgAsSolver,      "apply multigrid as solver");
-  bool   printTimings     = true;              clp.setOption("timings", "notimings",  &printTimings,     "print timings to screen");
-  int    writeMatricesOPT = -2;                clp.setOption("write",                 &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
-  double tol              = 1e-12;             clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-  std::string krylovMethod = "cg"; clp.setOption("krylov",                   &krylovMethod,     "outer Krylov method");
-   std::string optSmooType = "cheby";        clp.setOption("smooType",       &optSmooType,           "smoother type ('l1-sgs', 'sgs 'or 'cheby')");
-  int optSweeps = 2;                      clp.setOption("sweeps",         &optSweeps,             "sweeps to be used in SGS (or Chebyshev degree)");
-
-
-  switch (clp.parse(argc,argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "scalingTest.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+  int amgAsPrecond = 1;
+  clp.setOption("precond", &amgAsPrecond, "apply multigrid as preconditioner");
+  int amgAsSolver = 0;
+  clp.setOption("fixPoint", &amgAsSolver, "apply multigrid as solver");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  int writeMatricesOPT = -2;
+  clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  std::string krylovMethod = "cg";
+  clp.setOption("krylov", &krylovMethod, "outer Krylov method");
+  std::string optSmooType = "cheby";
+  clp.setOption("smooType", &optSmooType, "smoother type ('l1-sgs', 'sgs 'or 'cheby')");
+  int optSweeps = 2;
+  clp.setOption("sweeps", &optSweeps, "sweeps to be used in SGS (or Chebyshev degree)");
+
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
-  fancyout << "========================================================\n" << xpetraParameters << matrixParameters;
+  fancyout << "========================================================\n"
+           << xpetraParameters << matrixParameters;
 
   // =========================================================================
   // Problem construction
@@ -140,7 +148,7 @@ int main(int argc, char *argv[]) {
   comm->barrier();
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
-  RCP<const Map>   map;
+  RCP<const Map> map;
   RCP<MultiVector> coordinates;
 
   // Retrieve matrix parameters (they may have been changed on the command line), and pass them to Galeri.
@@ -155,9 +163,9 @@ int main(int argc, char *argv[]) {
   // If you don't want Galeri to do this, specify mx or my on the galeriList.
   Teuchos::ParameterList pl = matrixParameters.GetParameterList();
   Teuchos::ParameterList galeriList;
-  galeriList.set("nx", pl.get("nx",nx));
-  galeriList.set("ny", pl.get("ny",ny));
-  galeriList.set("nz", pl.get("nz",nz));
+  galeriList.set("nx", pl.get("nx", nx));
+  galeriList.set("ny", pl.get("ny", ny));
+  galeriList.set("nz", pl.get("nz", nz));
   // galeriList.set("mx", comm->getSize());
   // galeriList.set("my", 1);
 
@@ -165,31 +173,29 @@ int main(int argc, char *argv[]) {
   // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
   // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
   if (matrixParameters.GetMatrixType() == "Laplace1D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D",map,matrixParameters.GetParameterList());
-  }
-  else if (matrixParameters.GetMatrixType() == "Laplace2D" || matrixParameters.GetMatrixType() == "Star2D" || matrixParameters.GetMatrixType() == "Elasticity2D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D",map,matrixParameters.GetParameterList());
-  }
-  else if (matrixParameters.GetMatrixType() == "Laplace3D" || matrixParameters.GetMatrixType() == "Elasticity3D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D",map,matrixParameters.GetParameterList());
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, matrixParameters.GetParameterList());
+  } else if (matrixParameters.GetMatrixType() == "Laplace2D" || matrixParameters.GetMatrixType() == "Star2D" || matrixParameters.GetMatrixType() == "Elasticity2D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, matrixParameters.GetParameterList());
+  } else if (matrixParameters.GetMatrixType() == "Laplace3D" || matrixParameters.GetMatrixType() == "Elasticity3D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, matrixParameters.GetParameterList());
   }
   // Expand map to do multiple DOF per node for block problems
   if (matrixParameters.GetMatrixType() == "Elasticity2D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
   if (matrixParameters.GetMatrixType() == "Elasticity3D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
   if (comm->getRank() == 0) {
     GO mx = galeriList.get("mx", -1);
     GO my = galeriList.get("my", -1);
     GO mz = galeriList.get("mz", -1);
     fancyout << "Processor subdomains in x direction: " << mx << std::endl
-         << "Processor subdomains in y direction: " << my << std::endl
-         << "Processor subdomains in z direction: " << mz << std::endl
-         << "========================================================" << std::endl;
+             << "Processor subdomains in y direction: " << my << std::endl
+             << "Processor subdomains in z direction: " << mz << std::endl
+             << "========================================================" << std::endl;
   }
 
   Teuchos::ParameterList matrixParams = matrixParameters.GetParameterList();
@@ -198,18 +204,18 @@ int main(int argc, char *argv[]) {
   matrixParams.set("mz", galeriList.get("mz", -1));
   if (matrixParameters.GetMatrixType() == "Elasticity2D" || matrixParameters.GetMatrixType() == "Elasticity3D") {
     // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-    matrixParams.set("right boundary" , "Neumann");
+    matrixParams.set("right boundary", "Neumann");
     matrixParams.set("bottom boundary", "Neumann");
-    matrixParams.set("top boundary"   , "Neumann");
-    matrixParams.set("front boundary" , "Neumann");
-    matrixParams.set("back boundary"  , "Neumann");
+    matrixParams.set("top boundary", "Neumann");
+    matrixParams.set("front boundary", "Neumann");
+    matrixParams.set("back boundary", "Neumann");
   }
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParams);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParams);
   RCP<Matrix> A = Pr->BuildMatrix();
 
-  RCP<MultiVector> nullspace = MultiVectorFactory::Build(map,1);
+  RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
   if (matrixParameters.GetMatrixType() == "Elasticity2D" ||
       matrixParameters.GetMatrixType() == "Elasticity3D") {
     nullspace = Pr->BuildNullspace();
@@ -219,57 +225,52 @@ int main(int argc, char *argv[]) {
     nullspace->putScalar(one);
   }
 
-
   fancyout << "Galeri complete.\n========================================================" << std::endl;
 
   // =========================================================================
   // Preconditioner construction
   // =========================================================================
 
-  //Multigrid Hierarchy
+  // Multigrid Hierarchy
   RCP<Hierarchy> H = rcp(new Hierarchy(A));
   H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
   FactoryManager M;
 
-  //Smoothers
+  // Smoothers
   std::string ifpackType;
   Teuchos::ParameterList ifpackList;
-  ifpackList.set("relaxation: sweeps", (LO) optSweeps);
-  ifpackList.set("relaxation: damping factor", (SC) 1.0);
+  ifpackList.set("relaxation: sweeps", (LO)optSweeps);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
   if (optSmooType == "sgs") {
     ifpackType = "RELAXATION";
     ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-  }
-  else if (optSmooType == "l1-sgs") {
+  } else if (optSmooType == "l1-sgs") {
     ifpackType = "RELAXATION";
     ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
     ifpackList.set("relaxation: use l1", true);
   } else if (optSmooType == "cheby") {
     ifpackType = "CHEBYSHEV";
-    ifpackList.set("chebyshev: degree", (LO) optSweeps);
+    ifpackList.set("chebyshev: degree", (LO)optSweeps);
 
     if (matrixParameters.GetMatrixType() == "Laplace1D") {
-      ifpackList.set("chebyshev: ratio eigenvalue", (SC) 3);
-    }
-    else if (matrixParameters.GetMatrixType() == "Laplace2D") {
-      ifpackList.set("chebyshev: ratio eigenvalue", (SC) 7);
-    }
-    else if (matrixParameters.GetMatrixType() == "Laplace3D") {
-      ifpackList.set("chebyshev: ratio eigenvalue", (SC) 20);
+      ifpackList.set("chebyshev: ratio eigenvalue", (SC)3);
+    } else if (matrixParameters.GetMatrixType() == "Laplace2D") {
+      ifpackList.set("chebyshev: ratio eigenvalue", (SC)7);
+    } else if (matrixParameters.GetMatrixType() == "Laplace3D") {
+      ifpackList.set("chebyshev: ratio eigenvalue", (SC)20);
     }
   }
   RCP<SmootherPrototype> smootherPrototype = rcp(new Ifpack2Smoother(ifpackType, ifpackList));
   M.SetFactory("Smoother", rcp(new SmootherFactory(smootherPrototype)));
 
-
   // create coarsest smoother
   RCP<SmootherPrototype> coarsestSmooProto;
   Teuchos::ParameterList coarsestSmooList;
-  coarsestSmooProto = rcp( new Ifpack2Smoother("RILUK",coarsestSmooList) );
+  coarsestSmooProto                     = rcp(new Ifpack2Smoother("RILUK", coarsestSmooList));
   RCP<SmootherFactory> coarsestSmooFact = rcp(new SmootherFactory(coarsestSmooProto, Teuchos::null));
   M.SetFactory("CoarseSolver", coarsestSmooFact);
 
-  int startLevel = 0;
+  int startLevel   = 0;
   int optMaxLevels = 10;
   H->Setup(M, startLevel, optMaxLevels);
 
@@ -292,29 +293,27 @@ int main(int argc, char *argv[]) {
 
     Teuchos::Array<ST::magnitudeType> norms(1);
     B->norm2(norms);
-    B->scale(1.0/norms[0]);
+    B->scale(1.0 / norms[0]);
     X->putScalar(zero);
   }
 
   if (amgAsSolver) {
-
     H->IsPreconditioner(false);
     H->Iterate(*B, *X, 25);
 
   } else if (amgAsPrecond) {
-
 #ifdef HAVE_MUELU_BELOS
     // Operator and Multivector type that will be used with Belos
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
     H->IsPreconditioner(true);
 
     // Define Operator and Preconditioner
-    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
-    RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+    RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
     belosProblem->setLeftPrec(belosPrec);
 
     bool set = belosProblem->setProblem();
@@ -326,14 +325,14 @@ int main(int argc, char *argv[]) {
     // Belos parameter list
     int maxIts = 2000;
     Teuchos::ParameterList belosList;
-    belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-    belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-    belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    belosList.set("Output Frequency",      1);
-    belosList.set("Output Style",          Belos::Brief);
+    belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+    belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+    belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    belosList.set("Output Frequency", 1);
+    belosList.set("Output Style", Belos::Brief);
 
     // Create an iterative solver manager
-    RCP< Belos::SolverManager<SC, MV, OP> > solver;
+    RCP<Belos::SolverManager<SC, MV, OP> > solver;
     if (krylovMethod == "cg") {
       solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
     } else if (krylovMethod == "gmres") {
@@ -350,41 +349,43 @@ int main(int argc, char *argv[]) {
       // Get the number of iterations for this solve.
       fancyout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-    } catch(...) {
-      fancyout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+    } catch (...) {
+      fancyout << std::endl
+               << "ERROR:  Belos threw an error! " << std::endl;
     }
 
     // Check convergence
     if (ret != Belos::Converged)
-      fancyout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      fancyout << std::endl
+               << "ERROR:  Belos did not converge! " << std::endl;
     else
-      fancyout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      fancyout << std::endl
+               << "SUCCESS:  Belos converged!" << std::endl;
 
-    //Clone the preconditioner to ThrustGPU node type
+    // Clone the preconditioner to ThrustGPU node type
     typedef KokkosClassic::ThrustGPUNode NO2;
-    typedef MueLu::Hierarchy<SC,LO,GO,NO2> Hierarchy2;
-    typedef Xpetra::MultiVector<SC,LO,GO,NO2> MV2;
+    typedef MueLu::Hierarchy<SC, LO, GO, NO2> Hierarchy2;
+    typedef Xpetra::MultiVector<SC, LO, GO, NO2> MV2;
     typedef Belos::OperatorT<MV2> OP2;
 
     ParameterList plClone;
     plClone.set<LocalOrdinal>("Verbose", 1);
-    RCP<NO2> node = rcp(new NO2(plClone));
+    RCP<NO2> node           = rcp(new NO2(plClone));
     RCP<Hierarchy2> clonedH = H->clone<NO2>(node);
 
-    //Clone A, X, B to new node type
-    RCP< Xpetra::Matrix<SC,LO,GO,NO2 > > clonedA = Xpetra::clone(*A, node);
-    RCP< MV2 > clonedX = Xpetra::clone(*X, node);
+    // Clone A, X, B to new node type
+    RCP<Xpetra::Matrix<SC, LO, GO, NO2> > clonedA = Xpetra::clone(*A, node);
+    RCP<MV2> clonedX                              = Xpetra::clone(*X, node);
     clonedX->putScalar(zero);
-    RCP< MV2 > clonedB = Xpetra::clone(*B, node);
+    RCP<MV2> clonedB = Xpetra::clone(*B, node);
     clonedH->IsPreconditioner(true);
 
-
     // Define Operator and Preconditioner
-    Teuchos::RCP<OP2> belosOp2   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO2>(clonedA)); // Turns a Xpetra::Matrix object into a Belos operator
-    Teuchos::RCP<OP2> belosPrec2 = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO2>(clonedH));  // Turns a MueLu::Hierarchy object into a Belos operator
+    Teuchos::RCP<OP2> belosOp2   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO2>(clonedA));  // Turns a Xpetra::Matrix object into a Belos operator
+    Teuchos::RCP<OP2> belosPrec2 = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO2>(clonedH));   // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
-    RCP< Belos::LinearProblem<SC, MV2, OP2> > belosProblem2 = rcp(new Belos::LinearProblem<SC, MV2, OP2>(belosOp2, clonedX, clonedB));
+    RCP<Belos::LinearProblem<SC, MV2, OP2> > belosProblem2 = rcp(new Belos::LinearProblem<SC, MV2, OP2>(belosOp2, clonedX, clonedB));
     belosProblem2->setLeftPrec(belosPrec2);
 
     bool set2 = belosProblem2->setProblem();
@@ -393,7 +394,7 @@ int main(int argc, char *argv[]) {
       return EXIT_FAILURE;
     }
     // Create an iterative solver manager
-    RCP< Belos::SolverManager<SC, MV2, OP2> > solver2;
+    RCP<Belos::SolverManager<SC, MV2, OP2> > solver2;
     if (krylovMethod == "cg") {
       solver2 = rcp(new Belos::BlockCGSolMgr<SC, MV2, OP2>(belosProblem2, rcp(&belosList, false)));
     } else if (krylovMethod == "gmres") {
@@ -401,30 +402,33 @@ int main(int argc, char *argv[]) {
     } else {
       TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Invalid Krylov method.  Options are \"cg\" or \" gmres\".");
     }
-    //Perform solve
+    // Perform solve
     Belos::ReturnType ret2 = Belos::Unconverged;
     try {
-       ret2 = solver2->solve();
-       // Get the number of iterations for this solve.
-       fancyout << "Number of iterations performed for this solve: " << solver2->getNumIters() << std::endl;
-    } catch(...) {
-       fancyout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+      ret2 = solver2->solve();
+      // Get the number of iterations for this solve.
+      fancyout << "Number of iterations performed for this solve: " << solver2->getNumIters() << std::endl;
+    } catch (...) {
+      fancyout << std::endl
+               << "ERROR:  Belos threw an error! " << std::endl;
     }
     // Check convergence
     if (ret2 != Belos::Converged)
-    fancyout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      fancyout << std::endl
+               << "ERROR:  Belos did not converge! " << std::endl;
     else
-    fancyout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      fancyout << std::endl
+               << "SUCCESS:  Belos converged!" << std::endl;
 
-    //Determine if example passed
+    // Determine if example passed
     RCP<KokkosClassic::DefaultNode::DefaultNodeType> defaultNode =
-      rcp (new Tpetra::KokkosClassic::DefaultNode::DefaultNodeType (pl));
-    RCP<MV> clonedXcpu = Xpetra::clone (*clonedX, defaultNode);
-    clonedXcpu->update (1.0, *X, -1.0);
+        rcp(new Tpetra::KokkosClassic::DefaultNode::DefaultNodeType(pl));
+    RCP<MV> clonedXcpu = Xpetra::clone(*clonedX, defaultNode);
+    clonedXcpu->update(1.0, *X, -1.0);
     Scalar norm;
-    clonedXcpu->norm2(Teuchos::arrayView(&norm,1));
-    std::cout <<"\nNorm of serial node soln - ThrustGPU node soln = "
-                << norm << std::endl;
+    clonedXcpu->norm2(Teuchos::arrayView(&norm, 1));
+    std::cout << "\nNorm of serial node soln - ThrustGPU node soln = "
+              << norm << std::endl;
 
     bool passed = false;
     if (norm <= Scalar(1e-10))
@@ -434,7 +438,7 @@ int main(int argc, char *argv[]) {
     else
       std::cout << "Example Failed!" << std::endl;
   }
-  #endif //ifdef HAVE_MUELU_BELOS
+#endif  // ifdef HAVE_MUELU_BELOS
 
   return 0;
-} //main
+}  // main
diff --git a/packages/muelu/example/advanced/levelwrap/LevelWrap.cpp b/packages/muelu/example/advanced/levelwrap/LevelWrap.cpp
index b37d0b5a7822..b44f23b7d1f9 100644
--- a/packages/muelu/example/advanced/levelwrap/LevelWrap.cpp
+++ b/packages/muelu/example/advanced/levelwrap/LevelWrap.cpp
@@ -55,7 +55,7 @@
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-#include <Tpetra_KokkosCompat_DefaultNode.hpp> // For Epetra only runs this points to FakeKokkos in Xpetra
+#include <Tpetra_KokkosCompat_DefaultNode.hpp>  // For Epetra only runs this points to FakeKokkos in Xpetra
 
 #include "Xpetra_ConfigDefs.hpp"
 #include <Xpetra_Map.hpp>
@@ -95,139 +95,136 @@ const std::string prefSeparator = "=====================================";
 
 namespace MueLuExamples {
 
-  template<class SC, class LO, class GO, class NO>
-  MueLu::ParameterListInterpreter<SC,LO,GO,NO> makeFactory(Teuchos::ParameterList & paramList) {
-    std::string paramXML = MueLu::ML2MueLuParameterTranslator::translate(paramList, "");
-    paramList = *Teuchos::getParametersFromXmlString(paramXML);
-    return MueLu::ParameterListInterpreter<SC,LO,GO,NO>(paramList);
-  }
-
-
+template <class SC, class LO, class GO, class NO>
+MueLu::ParameterListInterpreter<SC, LO, GO, NO> makeFactory(Teuchos::ParameterList& paramList) {
+  std::string paramXML = MueLu::ML2MueLuParameterTranslator::translate(paramList, "");
+  paramList            = *Teuchos::getParametersFromXmlString(paramXML);
+  return MueLu::ParameterListInterpreter<SC, LO, GO, NO>(paramList);
+}
 
 #ifdef HAVE_MUELU_BELOS
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void solve_system_hierarchy(Xpetra::UnderlyingLib& lib,
-                              Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>>&   A,
-                              Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>>&   X,
-                              Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>>&   B,
-                              Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& H,
-                              Teuchos::RCP<Teuchos::ParameterList>& SList) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void solve_system_hierarchy(Xpetra::UnderlyingLib& lib,
+                            Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                            Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                            Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B,
+                            Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& H,
+                            Teuchos::RCP<Teuchos::ParameterList>& SList) {
 #include "MueLu_UseShortNames.hpp"
-    using Teuchos::rcp;
-
-    typedef Xpetra::MultiVector <Scalar,LocalOrdinal,GlobalOrdinal,Node> MV;
-    typedef Belos::OperatorT<MV>                                         OP;
-
-    // Construct a Belos LinearProblem object
-    RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp <Scalar,LocalOrdinal,GlobalOrdinal,Node>(H));
-
-    RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem =
-        rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
-    belosProblem->setRightPrec(belosPrec);
-    belosProblem->setProblem(X,B);
-    Belos::SolverFactory<Scalar, MV, OP> BelosFactory;
-    RCP<Belos::SolverManager<Scalar, MV, OP> > BelosSolver =
-        BelosFactory.create(std::string("CG"), SList);
-    BelosSolver->setProblem(belosProblem);
-    Belos::ReturnType result = BelosSolver->solve();
-    TEUCHOS_TEST_FOR_EXCEPTION(result == Belos::Unconverged, std::runtime_error, "Belos failed to converge");
-  }
+  using Teuchos::rcp;
+
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef Belos::OperatorT<MV> OP;
+
+  // Construct a Belos LinearProblem object
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(H));
+
+  RCP<Belos::LinearProblem<Scalar, MV, OP>> belosProblem =
+      rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
+  belosProblem->setRightPrec(belosPrec);
+  belosProblem->setProblem(X, B);
+  Belos::SolverFactory<Scalar, MV, OP> BelosFactory;
+  RCP<Belos::SolverManager<Scalar, MV, OP>> BelosSolver =
+      BelosFactory.create(std::string("CG"), SList);
+  BelosSolver->setProblem(belosProblem);
+  Belos::ReturnType result = BelosSolver->solve();
+  TEUCHOS_TEST_FOR_EXCEPTION(result == Belos::Unconverged, std::runtime_error, "Belos failed to converge");
+}
 
-  // --------------------------------------------------------------------------------------
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void solve_system_list(Xpetra::UnderlyingLib& lib,
-                         Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& A,
-                         Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& X,
-                         Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& B,
-                         Teuchos::ParameterList& MueLuList,
-                         Teuchos::RCP<Teuchos::ParameterList>& SList) {
+// --------------------------------------------------------------------------------------
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void solve_system_list(Xpetra::UnderlyingLib& lib,
+                       Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                       Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                       Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B,
+                       Teuchos::ParameterList& MueLuList,
+                       Teuchos::RCP<Teuchos::ParameterList>& SList) {
 #include "MueLu_UseShortNames.hpp"
-    using Teuchos::rcp;
-
-    if(lib == Xpetra::UseEpetra) {MueLuList.set("use kokkos refactor", false);}
-    Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > H =
-        MueLu::CreateXpetraPreconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A, MueLuList);
+  using Teuchos::rcp;
 
-    typedef Xpetra::MultiVector <Scalar,LocalOrdinal,GlobalOrdinal,Node> MV;
-    typedef Belos::OperatorT<MV>                                         OP;
-
-    // Construct a Belos LinearProblem object
-    RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp <Scalar,LocalOrdinal,GlobalOrdinal,Node>(H));
-
-    RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem =
-        rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
-    belosProblem->setRightPrec(belosPrec);
-    belosProblem->setProblem(X,B);
-
-    Belos::SolverFactory<SC, MV, OP> BelosFactory;
-    Teuchos::RCP<Belos::SolverManager<SC, MV, OP> > BelosSolver = BelosFactory.create(std::string("CG"), SList);
-    BelosSolver->setProblem(belosProblem);
-    Belos::ReturnType result = BelosSolver->solve();
-    TEUCHOS_TEST_FOR_EXCEPTION(result == Belos::Unconverged, std::runtime_error, "Belos failed to converge");
+  if (lib == Xpetra::UseEpetra) {
+    MueLuList.set("use kokkos refactor", false);
   }
+  Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>> H =
+      MueLu::CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, MueLuList);
+
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef Belos::OperatorT<MV> OP;
+
+  // Construct a Belos LinearProblem object
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(H));
+
+  RCP<Belos::LinearProblem<Scalar, MV, OP>> belosProblem =
+      rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
+  belosProblem->setRightPrec(belosPrec);
+  belosProblem->setProblem(X, B);
+
+  Belos::SolverFactory<SC, MV, OP> BelosFactory;
+  Teuchos::RCP<Belos::SolverManager<SC, MV, OP>> BelosSolver = BelosFactory.create(std::string("CG"), SList);
+  BelosSolver->setProblem(belosProblem);
+  Belos::ReturnType result = BelosSolver->solve();
+  TEUCHOS_TEST_FOR_EXCEPTION(result == Belos::Unconverged, std::runtime_error, "Belos failed to converge");
+}
 #endif
 
-
-  // --------------------------------------------------------------------------------------
-  // This routine generate's the user's original A matrix and nullspace
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void generate_user_matrix_and_nullspace(std::string& matrixType,
-                                          Xpetra::UnderlyingLib& lib,
-                                          Teuchos::ParameterList& galeriList,
-                                          Teuchos::RCP<const Teuchos::Comm<int>>& comm,
-                                          Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>>&      A,
-                                          Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& nullspace) {
+// --------------------------------------------------------------------------------------
+// This routine generate's the user's original A matrix and nullspace
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void generate_user_matrix_and_nullspace(std::string& matrixType,
+                                        Xpetra::UnderlyingLib& lib,
+                                        Teuchos::ParameterList& galeriList,
+                                        Teuchos::RCP<const Teuchos::Comm<int>>& comm,
+                                        Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                        Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& nullspace) {
 #include "MueLu_UseShortNames.hpp"
 
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+  RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  Teuchos::FancyOStream& out       = *fancy;
 
-    RCP<const Map>   map;
-    RCP<MultiVector> coordinates;
-    if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", map, galeriList);
+  RCP<const Map> map;
+  RCP<MultiVector> coordinates;
+  if (matrixType == "Laplace1D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, galeriList);
 
-    } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D", map, galeriList);
+  } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, galeriList);
 
-    } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D", map, galeriList);
-    }
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, galeriList);
+  }
 
-    // Expand map to do multiple DOF per node for block problems
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
+  // Expand map to do multiple DOF per node for block problems
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
 
-    out << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
-        << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
-        << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
-        << "========================================================" << std::endl;
+  out << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
+      << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
+      << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
+      << "========================================================" << std::endl;
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, galeriList);
 
-    A = Pr->BuildMatrix();
+  A = Pr->BuildMatrix();
 
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
-      nullspace = Pr->BuildNullspace();
-      A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
-    }
-    else {
-      nullspace = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A->getRowMap(),1);
-      nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    }      
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
+    nullspace = Pr->BuildNullspace();
+    A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
+  } else {
+    nullspace = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A->getRowMap(), 1);
+    nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
   }
 }
-
+}  // namespace MueLuExamples
 
 // --------------------------------------------------------------------------------------
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::TimeMonitor;
 
@@ -235,10 +232,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool verbose = true;
   try {
 #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
 
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream& out       = *fancy;
 
     typedef Teuchos::ScalarTraits<SC> STS;
 
@@ -246,28 +243,29 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-    out << thickSeparator << std::endl << xpetraParameters << galeriParameters;
+    out << thickSeparator << std::endl
+        << xpetraParameters << galeriParameters;
 
     // =========================================================================
     // Problem construction
     // =========================================================================
-    RCP<const Map>   map;
-    RCP<Matrix> A,P,R, Ac;
-    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal,Node> > nullspace;
+    RCP<const Map> map;
+    RCP<Matrix> A, P, R, Ac;
+    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> nullspace;
     std::string matrixType = galeriParameters.GetMatrixType();
-    MueLuExamples::generate_user_matrix_and_nullspace(matrixType,lib,galeriList,comm,A,nullspace);
-    map=A->getRowMap();
+    MueLuExamples::generate_user_matrix_and_nullspace(matrixType, lib, galeriList, comm, A, nullspace);
+    map = A->getRowMap();
 
     // =========================================================================
     // Setups and solves
@@ -280,54 +278,52 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
 #ifdef HAVE_MUELU_BELOS
     // Belos Options
-    RCP<Teuchos::ParameterList> SList = rcp(new Teuchos::ParameterList );
-    SList->set("Verbosity",Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    SList->set("Output Frequency",10);
-    SList->set("Output Style",Belos::Brief);
-    SList->set("Maximum Iterations",200);
-    SList->set("Convergence Tolerance",1e-10);
+    RCP<Teuchos::ParameterList> SList = rcp(new Teuchos::ParameterList);
+    SList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    SList->set("Output Frequency", 10);
+    SList->set("Output Style", Belos::Brief);
+    SList->set("Maximum Iterations", 200);
+    SList->set("Convergence Tolerance", 1e-10);
 #endif
 
-
     // =========================================================================
     // Solve #1 (standard MueLu)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 1: Standard "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 1: Standard " << prefSeparator << std::endl;
     {
       // Use an MueLu-style parameter list for variety
       Teuchos::ParameterList MueList;
-      MueList.set("multigrid algorithm","unsmoothed");
+      MueList.set("multigrid algorithm", "unsmoothed");
       MueList.set("verbosity", "high");
-      MueList.set("coarse: type","Amesos-Superlu");
-      MueList.set("coarse: max size",10);
-      MueList.set("max levels",3);
+      MueList.set("coarse: type", "Amesos-Superlu");
+      MueList.set("coarse: max size", 10);
+      MueList.set("max levels", 3);
 #ifdef HAVE_AMESOS2_KLU2
-      MueList.set("coarse: type","Amesos-KLU");
+      MueList.set("coarse: type", "Amesos-KLU");
 #endif
-      MueLu::ParameterListInterpreter<SC,LO,GO,NO> mueLuFactory(MueList);
-      RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
+      MueLu::ParameterListInterpreter<SC, LO, GO, NO> mueLuFactory(MueList);
+      RCP<Hierarchy> H                              = mueLuFactory.CreateHierarchy();
       Teuchos::RCP<FactoryManagerBase> LevelFactory = mueLuFactory.GetFactoryManager(1);
       H->setlib(lib);
       H->AddNewLevel();
-      H->GetLevel(1)->Keep("Nullspace",LevelFactory->GetFactory("Nullspace").get());
+      H->GetLevel(1)->Keep("Nullspace", LevelFactory->GetFactory("Nullspace").get());
       H->GetLevel(0)->Set("A", A);
       H->GetLevel(0)->Set("Nullspace", nullspace);
       mueLuFactory.SetupHierarchy(*H);
 
-
 #ifdef HAVE_MUELU_BELOS
       // Solve
-      MueLuExamples::solve_system_hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node>(lib,A,X,B,H,SList);
+      MueLuExamples::solve_system_hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib, A, X, B, H, SList);
 #endif
 
       // Extract R,P & Ac for LevelWrap Usage
-      H->GetLevel(1)->Get("R",R);
-      H->GetLevel(1)->Get("P",P);
-      H->GetLevel(1)->Get("A",Ac);
+      H->GetLevel(1)->Get("R", R);
+      H->GetLevel(1)->Get("P", P);
+      H->GetLevel(1)->Get("A", Ac);
 
       // extract coarse level null space from level 1 that we have to inject for the next runs...
-      nullspace = H->GetLevel(1)->template Get<RCP<MultiVector> >("Nullspace",LevelFactory->GetFactory("Nullspace").get());
+      nullspace = H->GetLevel(1)->template Get<RCP<MultiVector>>("Nullspace", LevelFactory->GetFactory("Nullspace").get());
     }
     out << thickSeparator << std::endl;
 
@@ -335,22 +331,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Solve #2 (level wrap, the long way, using pre-done Ac)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 2: LevelWrap, Long Way, P, R, Ac "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 2: LevelWrap, Long Way, P, R, Ac " << prefSeparator << std::endl;
     {
       // Start w/ an ML-style parameter list
       Teuchos::ParameterList MLList;
       MLList.set("ML output", 10);
-      MLList.set("coarse: type","Amesos-Superlu");
+      MLList.set("coarse: type", "Amesos-Superlu");
 #ifdef HAVE_AMESOS2_KLU2
-      MLList.set("coarse: type","Amesos-KLU");
+      MLList.set("coarse: type", "Amesos-KLU");
 #endif
       FactoryManager M1;
       M1.SetKokkosRefactor(false);
-      M1.SetFactory("A",        MueLu::NoFactory::getRCP());
-      M1.SetFactory("P",        MueLu::NoFactory::getRCP());
-      M1.SetFactory("R",        MueLu::NoFactory::getRCP());
+      M1.SetFactory("A", MueLu::NoFactory::getRCP());
+      M1.SetFactory("P", MueLu::NoFactory::getRCP());
+      M1.SetFactory("R", MueLu::NoFactory::getRCP());
 
-      MueLu::ParameterListInterpreter<SC,LO,GO,NO> mueLuFactory = MueLuExamples::makeFactory<SC,LO,GO,NO>(MLList);
+      MueLu::ParameterListInterpreter<SC, LO, GO, NO> mueLuFactory = MueLuExamples::makeFactory<SC, LO, GO, NO>(MLList);
       mueLuFactory.AddFactoryManager(1, 1, Teuchos::rcpFromRef(M1));
       RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
       H->setlib(lib);
@@ -363,33 +359,30 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       mueLuFactory.SetupHierarchy(*H);
 
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_hierarchy(lib,A,X,B,H,SList);
+      MueLuExamples::solve_system_hierarchy(lib, A, X, B, H, SList);
 #endif
-
     }
     out << thickSeparator << std::endl;
 
-
     // =========================================================================
     // Solve #3 (level wrap, the long way, using P, R and nullspace)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 3: LevelWrap, Long Way, P, R "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 3: LevelWrap, Long Way, P, R " << prefSeparator << std::endl;
     {
-
       // Start w/ an ML-style parameter list
       Teuchos::ParameterList MLList;
       MLList.set("ML output", 10);
-      MLList.set("coarse: type","Amesos-Superlu");
+      MLList.set("coarse: type", "Amesos-Superlu");
 #ifdef HAVE_AMESOS2_KLU2
-      MLList.set("coarse: type","Amesos-KLU");
+      MLList.set("coarse: type", "Amesos-KLU");
 #endif
       FactoryManager M1;
       M1.SetKokkosRefactor(false);
-      M1.SetFactory("P",        MueLu::NoFactory::getRCP());
-      M1.SetFactory("R",        MueLu::NoFactory::getRCP());
+      M1.SetFactory("P", MueLu::NoFactory::getRCP());
+      M1.SetFactory("R", MueLu::NoFactory::getRCP());
 
-      MueLu::ParameterListInterpreter<SC,LO,GO,NO> mueLuFactory = MueLuExamples::makeFactory<SC,LO,GO,NO>(MLList);
+      MueLu::ParameterListInterpreter<SC, LO, GO, NO> mueLuFactory = MueLuExamples::makeFactory<SC, LO, GO, NO>(MLList);
       mueLuFactory.AddFactoryManager(1, 1, Teuchos::rcpFromRef(M1));
       RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
       H->setlib(lib);
@@ -400,9 +393,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       H->GetLevel(1)->Set("Nullspace", nullspace);
       mueLuFactory.SetupHierarchy(*H);
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_hierarchy(lib,A,X,B,H,SList);
+      MueLuExamples::solve_system_hierarchy(lib, A, X, B, H, SList);
 #endif
-
     }
     out << thickSeparator << std::endl;
 
@@ -410,19 +402,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Solve #4 (level wrap, the fast way, everything)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 4: LevelWrap, Fast Way, P, R, Ac "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 4: LevelWrap, Fast Way, P, R, Ac " << prefSeparator << std::endl;
     {
       Teuchos::ParameterList MueLuList, level1;
-      level1.set("A",Ac);
-      level1.set("R",R);
-      level1.set("P",P);
-      level1.set("Nullspace",nullspace);
-      MueLuList.set("level 1",level1);
-      MueLuList.set("verbosity","high");
-      MueLuList.set("coarse: max size",100);
-      MueLuList.set("max levels",4);
+      level1.set("A", Ac);
+      level1.set("R", R);
+      level1.set("P", P);
+      level1.set("Nullspace", nullspace);
+      MueLuList.set("level 1", level1);
+      MueLuList.set("verbosity", "high");
+      MueLuList.set("coarse: max size", 100);
+      MueLuList.set("max levels", 4);
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_list(lib,A,X,B,MueLuList,SList);
+      MueLuExamples::solve_system_list(lib, A, X, B, MueLuList, SList);
 #endif
     }
 
@@ -430,57 +422,55 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Solve #5 (level wrap, the fast way, P, R + Nullspace)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 5: LevelWrap, Fast Way, P, R "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 5: LevelWrap, Fast Way, P, R " << prefSeparator << std::endl;
     {
       Teuchos::ParameterList MueLuList, level1;
-      level1.set("R",R);
-      level1.set("P",P);
-      level1.set("Nullspace",nullspace);
-      MueLuList.set("level 1",level1);
-      MueLuList.set("verbosity","high");
-      MueLuList.set("coarse: max size",100);
+      level1.set("R", R);
+      level1.set("P", P);
+      level1.set("Nullspace", nullspace);
+      MueLuList.set("level 1", level1);
+      MueLuList.set("verbosity", "high");
+      MueLuList.set("coarse: max size", 100);
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_list(lib,A,X,B,MueLuList,SList);
+      MueLuExamples::solve_system_list(lib, A, X, B, MueLuList, SList);
 #endif
     }
 
-
     // =========================================================================
     // Solve #6 (level wrap, the fast way, P only, explicit transpose)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 6: LevelWrap, Fast Way, P only, explicit transpose "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 6: LevelWrap, Fast Way, P only, explicit transpose " << prefSeparator << std::endl;
     {
       Teuchos::ParameterList MueLuList, level1;
-      level1.set("P",P);
-      level1.set("Nullspace",nullspace);
-      MueLuList.set("level 1",level1);
-      MueLuList.set("verbosity","high");
-      MueLuList.set("coarse: max size",100);
-      MueLuList.set("transpose: use implicit",false);
-      MueLuList.set("max levels",4);
+      level1.set("P", P);
+      level1.set("Nullspace", nullspace);
+      MueLuList.set("level 1", level1);
+      MueLuList.set("verbosity", "high");
+      MueLuList.set("coarse: max size", 100);
+      MueLuList.set("transpose: use implicit", false);
+      MueLuList.set("max levels", 4);
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_list(lib,A,X,B,MueLuList,SList);
+      MueLuExamples::solve_system_list(lib, A, X, B, MueLuList, SList);
 #endif
     }
 
-
     // =========================================================================
     // Solve #7 (level wrap, the fast way, P only, implicit transpose)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 7: LevelWrap, Fast Way, P only, implicit transpose "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 7: LevelWrap, Fast Way, P only, implicit transpose " << prefSeparator << std::endl;
     {
       Teuchos::ParameterList MueLuList, level1;
-      level1.set("P",P);
-      level1.set("Nullspace",nullspace);
-      MueLuList.set("level 1",level1);
-      MueLuList.set("verbosity","high");
-      MueLuList.set("coarse: max size",100);
-      MueLuList.set("transpose: use implicit",true);
-      MueLuList.set("max levels",2);
+      level1.set("P", P);
+      level1.set("Nullspace", nullspace);
+      MueLuList.set("level 1", level1);
+      MueLuList.set("verbosity", "high");
+      MueLuList.set("coarse: max size", 100);
+      MueLuList.set("transpose: use implicit", true);
+      MueLuList.set("max levels", 2);
 #ifdef HAVE_MUELU_BELOS
-      MueLuExamples::solve_system_list(lib,A,X,B,MueLuList,SList);
+      MueLuExamples::solve_system_list(lib, A, X, B, MueLuList, SList);
 #endif
     }
 #endif
@@ -488,17 +478,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
diff --git a/packages/muelu/example/advanced/memory/Epetra1DLaplace.cpp b/packages/muelu/example/advanced/memory/Epetra1DLaplace.cpp
index 8997b8f3ff3f..dcba6cf12a3b 100644
--- a/packages/muelu/example/advanced/memory/Epetra1DLaplace.cpp
+++ b/packages/muelu/example/advanced/memory/Epetra1DLaplace.cpp
@@ -60,13 +60,12 @@
 
 #include "MueLu_MemoryProfiler.hpp"
 
-int main(int argc, char *argv[])
-{
+int main(int argc, char *argv[]) {
   int ierr, i;
 
 #ifdef HAVE_MPI
-  MPI_Init(&argc,&argv);
-  Epetra_MpiComm Comm( MPI_COMM_WORLD );
+  MPI_Init(&argc, &argv);
+  Epetra_MpiComm Comm(MPI_COMM_WORLD);
 #else
   Epetra_SerialComm Comm;
 #endif
@@ -74,14 +73,14 @@ int main(int argc, char *argv[])
   bool success = false;
   bool verbose = true;
   try {
-    //int myRank = Comm.MyPID();
+    // int myRank = Comm.MyPID();
 
-    //int numGlobalElements = 10000000;
+    // int numGlobalElements = 10000000;
     int numGlobalElements = 100;
 
-    Teuchos::CommandLineProcessor cmdp(false,true);
-    cmdp.setOption("numGlobalElements",&numGlobalElements,"Global problem size.");
-    if (cmdp.parse(argc,argv) != Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL) {
+    Teuchos::CommandLineProcessor cmdp(false, true);
+    cmdp.setOption("numGlobalElements", &numGlobalElements, "Global problem size.");
+    if (cmdp.parse(argc, argv) != Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL) {
       throw -1;
     }
 
@@ -109,31 +108,31 @@ int main(int argc, char *argv[])
     PrintMemoryUsage("Memory after CrsMatrix constructor", "epetra-after-ctor.heap");
 
     std::vector<double> Values(2);
-    Values[0] = -1.0; Values[1] = -1.0;
+    Values[0] = -1.0;
+    Values[1] = -1.0;
     std::vector<int> Indices(2);
     double two = 2.0;
     int NumEntries;
 
-    for (i=0; i<NumMyElements; i++) {
-
-      if (MyGlobalElements[i]==0) {
+    for (i = 0; i < NumMyElements; i++) {
+      if (MyGlobalElements[i] == 0) {
         Indices[0] = 1;
         NumEntries = 1;
-      } else if (MyGlobalElements[i] == numGlobalElements-1) {
-        Indices[0] = numGlobalElements-2;
+      } else if (MyGlobalElements[i] == numGlobalElements - 1) {
+        Indices[0] = numGlobalElements - 2;
         NumEntries = 1;
       } else {
-        Indices[0] = MyGlobalElements[i]-1;
-        Indices[1] = MyGlobalElements[i]+1;
+        Indices[0] = MyGlobalElements[i] - 1;
+        Indices[1] = MyGlobalElements[i] + 1;
         NumEntries = 2;
       }
 
       ierr = A.InsertGlobalValues(MyGlobalElements[i], NumEntries, &Values[0], &Indices[0]);
-      assert(ierr==0);
+      assert(ierr == 0);
 
       // Put in the diagonal entry
       ierr = A.InsertGlobalValues(MyGlobalElements[i], 1, &two, &MyGlobalElements[i]);
-      assert(ierr==0);
+      assert(ierr == 0);
     }
 
     PrintMemoryUsage("Memory after InsertGlobalValues()", "epetra-after-insert.heap");
@@ -145,8 +144,10 @@ int main(int argc, char *argv[])
 
     MemoryUsageStop();
 
-    if(ierr==0) success = true;
-    else success=false;
+    if (ierr == 0)
+      success = true;
+    else
+      success = false;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
@@ -154,5 +155,5 @@ int main(int argc, char *argv[])
   MPI_Finalize();
 #endif
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/example/advanced/memory/MueLu_MatrixDriver.cpp b/packages/muelu/example/advanced/memory/MueLu_MatrixDriver.cpp
index d6b8fe929774..591ff3aac2fb 100644
--- a/packages/muelu/example/advanced/memory/MueLu_MatrixDriver.cpp
+++ b/packages/muelu/example/advanced/memory/MueLu_MatrixDriver.cpp
@@ -67,10 +67,9 @@
 
 #ifdef _MSC_VER
 #define NOMINMAX
-#include <windows.h> // for Sleep
+#include <windows.h>  // for Sleep
 #endif
 
-
 /*
    This driver simply generates a Tpetra matrix, prints it to screen, and exits.
 
@@ -79,16 +78,15 @@
    Use the "--help" option to get verbose help.
    */
 
-int main(int argc, char** argv)
-{
+int main(int argc, char** argv) {
   using Teuchos::RCP;
 
-  typedef typename Tpetra::Map<>::local_ordinal_type    LO; // LocalOrdinal
-  typedef typename Tpetra::Map<>::global_ordinal_type   GO; // GlobalOrdinal
-  typedef MueLu::DefaultScalar                                 SC;
+  typedef typename Tpetra::Map<>::local_ordinal_type LO;   // LocalOrdinal
+  typedef typename Tpetra::Map<>::global_ordinal_type GO;  // GlobalOrdinal
+  typedef MueLu::DefaultScalar SC;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
@@ -101,13 +99,13 @@ int main(int argc, char** argv)
     // Note: use --help to list available options.
     Teuchos::CommandLineProcessor clp(false);
 
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp);   // manage parameters of the test case
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp);  // manage parameters of the test case
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     matrixParameters.check();
@@ -116,11 +114,10 @@ int main(int argc, char** argv)
     /**********************************************************************************/
     /* CREATE INITAL MATRIX                                                           */
     /**********************************************************************************/
-    RCP<const Tpetra::Map<LO,GO> > map = Teuchos::rcp( new Tpetra::Map<LO,GO>(matrixParameters.GetNumGlobalElements(), 0, comm) );
-    RCP<Galeri::Xpetra::Problem<Tpetra::Map<LO,GO>,Tpetra::CrsMatrix<SC,LO,GO>,Tpetra::MultiVector<SC,LO,GO> > > problem =
-      Galeri::Xpetra::BuildProblem<SC, LO, GO, Tpetra::Map<LO,GO>, Tpetra::CrsMatrix<SC,LO,GO>, Tpetra::MultiVector<SC,LO,GO> >
-      (matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
-    RCP<Tpetra::CrsMatrix<SC,LO,GO> > A = problem->BuildMatrix();
+    RCP<const Tpetra::Map<LO, GO> > map = Teuchos::rcp(new Tpetra::Map<LO, GO>(matrixParameters.GetNumGlobalElements(), 0, comm));
+    RCP<Galeri::Xpetra::Problem<Tpetra::Map<LO, GO>, Tpetra::CrsMatrix<SC, LO, GO>, Tpetra::MultiVector<SC, LO, GO> > > problem =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Tpetra::Map<LO, GO>, Tpetra::CrsMatrix<SC, LO, GO>, Tpetra::MultiVector<SC, LO, GO> >(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
+    RCP<Tpetra::CrsMatrix<SC, LO, GO> > A = problem->BuildMatrix();
 
     /**********************************************************************************/
     /*                                                                                */
@@ -128,7 +125,8 @@ int main(int argc, char** argv)
 
     RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     if (comm->getRank() == 0)
-      std::cout << "\n================ MAP =====================================================\n" << std::endl;
+      std::cout << "\n================ MAP =====================================================\n"
+                << std::endl;
     map->describe(*out, Teuchos::VERB_EXTREME);
     comm->barrier();
 #ifdef _MSC_VER
@@ -137,12 +135,13 @@ int main(int argc, char** argv)
     sleep(1);
 #endif
     if (comm->getRank() == 0)
-      std::cout << "\n================ MATRIX ==================================================\n" << std::endl;
+      std::cout << "\n================ MATRIX ==================================================\n"
+                << std::endl;
     A->describe(*out, Teuchos::VERB_EXTREME);
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.cpp b/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.cpp
index 63d1f6dc93c7..efc50b5632c3 100644
--- a/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.cpp
+++ b/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.cpp
@@ -53,35 +53,34 @@
 
 #ifdef HAVE_MUELU_GOOGLE_PERFTOOLS
 //#include <google/heap-profiler.h>
-#include </home/jngaida/google-perftools-1.8.3/src/google/heap-profiler.h> // TODO :-)
+#include </home/jngaida/google-perftools-1.8.3/src/google/heap-profiler.h>  // TODO :-)
 #endif
 
 #include <MueLu_MemoryProfiler.hpp>
 
 std::string GetMemoryUsage() {
-
   std::ostringstream mem;
 
 #ifdef HAVE_MUELU_PROC_SELF_STATUS
-  //TODO: test if /proc/self/status exist on the system instead of #ifdef
+  // TODO: test if /proc/self/status exist on the system instead of #ifdef
 
   std::ifstream proc("/proc/self/status");
   std::string s;
-  while(getline(proc, s), !proc.fail()) {
-    if(s.substr(0, 6) == "VmSize") {
+  while (getline(proc, s), !proc.fail()) {
+    if (s.substr(0, 6) == "VmSize") {
       mem << s;
       return mem.str();
     }
   }
 
-#endif // HAVE_MUELU_PROC_SELF_STATUS
+#endif  // HAVE_MUELU_PROC_SELF_STATUS
 
   return mem.str();
 }
 
 void MemoryUsageStart(const std::string& autoLogPrefix) {
 #ifdef HAVE_MUELU_GOOGLE_PERFTOOLS
-   HeapProfilerStart("auto-profiling");
+  HeapProfilerStart("auto-profiling");
 #endif
 }
 
@@ -98,18 +97,20 @@ void PrintMemoryUsage(const std::string& description, const std::string& filenam
 
 #ifdef HAVE_MUELU_GOOGLE_PERFTOOLS
   if (IsHeapProfilerRunning()) {
-
     char* profile = GetHeapProfile();
 
     std::istringstream iss(profile);
     std::string sub;
-    iss >> sub; iss >> sub; iss >> sub; // skip 3 first substring
     iss >> sub;
-    double MB = atof(sub.c_str()) / (1024*1024);
+    iss >> sub;
+    iss >> sub;  // skip 3 first substring
+    iss >> sub;
+    double MB = atof(sub.c_str()) / (1024 * 1024);
 
     // print
     if (description != "") {
-      std::ostringstream sname; sname.precision(1);
+      std::ostringstream sname;
+      sname.precision(1);
       sname << description << ": " << std::fixed << MB << " MB";
       std::cout << sname.str() << std::endl;
     }
@@ -117,7 +118,10 @@ void PrintMemoryUsage(const std::string& description, const std::string& filenam
     // dump to file
     if (filename != "") {
       std::ofstream out(filename.c_str(), std::ios::out | std::ios::binary);
-      if(!out) { std::cout << "Cannot open output file: " << filename << std::endl; return; }
+      if (!out) {
+        std::cout << "Cannot open output file: " << filename << std::endl;
+        return;
+      }
       out.write(profile, strlen(profile));
       out.close();
 
@@ -126,8 +130,6 @@ void PrintMemoryUsage(const std::string& description, const std::string& filenam
     }
 
     free(profile);
-
   }
 #endif
-
 }
diff --git a/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.hpp b/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.hpp
index b180ae5cb656..6ddf8fad40c9 100644
--- a/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.hpp
+++ b/packages/muelu/example/advanced/memory/MueLu_MemoryProfiler.hpp
@@ -54,5 +54,4 @@ void PrintMemoryUsage(const std::string& description = "Memory Usage:", const st
 void MemoryUsageStart(const std::string& autoLogPrefix = "memorylog-");
 void MemoryUsageStop();
 
-#endif // MUELU_MEMORY_PROFILER_HPP
-
+#endif  // MUELU_MEMORY_PROFILER_HPP
diff --git a/packages/muelu/example/advanced/memory/Tpetra1DLaplace.cpp b/packages/muelu/example/advanced/memory/Tpetra1DLaplace.cpp
index d356b0c3e991..e893655e348f 100644
--- a/packages/muelu/example/advanced/memory/Tpetra1DLaplace.cpp
+++ b/packages/muelu/example/advanced/memory/Tpetra1DLaplace.cpp
@@ -57,72 +57,72 @@
 #include "MueLu_MemoryProfiler.hpp"
 
 int main(int argc, char *argv[]) {
-  Tpetra::ScopeGuard mpiSession(&argc,&argv);
+  Tpetra::ScopeGuard mpiSession(&argc, &argv);
 
   bool success = false;
   bool verbose = true;
   try {
-    typedef Tpetra::CrsMatrix<>::scalar_type        Scalar;
-    typedef Tpetra::Map<>::local_ordinal_type       LO;
+    typedef Tpetra::CrsMatrix<>::scalar_type Scalar;
+    typedef Tpetra::Map<>::local_ordinal_type LO;
 #if defined(HAVE_TPETRA_INST_INT_INT)
     // mfh 07 Aug 2015: Prefer GO = int, for consistency with Epetra,
     // but use the default GO type if GO = int is not enabled.
-    typedef int                                     GO;
+    typedef int GO;
 #else
-    typedef Tpetra::Map<LO>::global_ordinal_type    GO;
-#endif // HAVE_TPETRA_INT_INT
-    typedef Tpetra::Map<LO, GO>::node_type          Node;
-    typedef Tpetra::Map<LO, GO, Node>               Map;
+    typedef Tpetra::Map<LO>::global_ordinal_type GO;
+#endif  // HAVE_TPETRA_INT_INT
+    typedef Tpetra::Map<LO, GO>::node_type Node;
+    typedef Tpetra::Map<LO, GO, Node> Map;
     typedef Tpetra::CrsMatrix<Scalar, LO, GO, Node> CrsMatrix;
     using Teuchos::RCP;
     using Teuchos::rcp;
     using Teuchos::tuple;
 
     RCP<const Teuchos::Comm<int> > comm = Tpetra::getDefaultComm();
-    //const int myRank = comm->getRank();
+    // const int myRank = comm->getRank();
 
-    //int numGlobalElements = 10000000;
+    // int numGlobalElements = 10000000;
     int numGlobalElements = 100;
 
-    Teuchos::CommandLineProcessor cmdp(false,true);
-    cmdp.setOption("numGlobalElements",&numGlobalElements,"Global problem size.");
-    if (cmdp.parse(argc,argv) != Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL) {
+    Teuchos::CommandLineProcessor cmdp(false, true);
+    cmdp.setOption("numGlobalElements", &numGlobalElements, "Global problem size.");
+    if (cmdp.parse(argc, argv) != Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL) {
       return -1;
     }
 
     const GO indexBase = 0;
     RCP<const Map> map =
-      rcp (new Map (static_cast<Tpetra::global_size_t> (numGlobalElements),
+        rcp(new Map(static_cast<Tpetra::global_size_t>(numGlobalElements),
                     indexBase, comm));
-    const size_t numMyElements = map->getLocalNumElements();
+    const size_t numMyElements                    = map->getLocalNumElements();
     Teuchos::ArrayView<const GO> myGlobalElements = map->getLocalElementList();
 
     MemoryUsageStart("Tpetra");
     PrintMemoryUsage("Initial memory usage", "tpetra-init.heap");
 
-    RCP<CrsMatrix> A = Tpetra::createCrsMatrix<Scalar>(map,3);
+    RCP<CrsMatrix> A = Tpetra::createCrsMatrix<Scalar>(map, 3);
 
     PrintMemoryUsage("Memory after CrsMatrix constructor", "tpetra-after-ctor.heap");
 
-    for (size_t i=0; i<numMyElements; i++) {
+    for (size_t i = 0; i < numMyElements; i++) {
       if (myGlobalElements[i] == 0) {
-        A->insertGlobalValues( myGlobalElements[i],
-            tuple<GO>( myGlobalElements[i], myGlobalElements[i]+1 ),
-            tuple<Scalar> ( 2.0, -1.0 ) );
-      } else if (myGlobalElements[i] == numGlobalElements-1) {
-        A->insertGlobalValues( myGlobalElements[i],
-            tuple<GO>( myGlobalElements[i]-1, myGlobalElements[i] ),
-            tuple<Scalar> ( -1.0, 2.0 ) );
+        A->insertGlobalValues(myGlobalElements[i],
+                              tuple<GO>(myGlobalElements[i], myGlobalElements[i] + 1),
+                              tuple<Scalar>(2.0, -1.0));
+      } else if (myGlobalElements[i] == numGlobalElements - 1) {
+        A->insertGlobalValues(myGlobalElements[i],
+                              tuple<GO>(myGlobalElements[i] - 1, myGlobalElements[i]),
+                              tuple<Scalar>(-1.0, 2.0));
       } else {
-        A->insertGlobalValues( myGlobalElements[i],
-            tuple<GO>( myGlobalElements[i]-1, myGlobalElements[i], myGlobalElements[i]+1 ),
-            tuple<Scalar> ( -1.0, 2.0, -1.0 ) );
+        A->insertGlobalValues(myGlobalElements[i],
+                              tuple<GO>(myGlobalElements[i] - 1, myGlobalElements[i], myGlobalElements[i] + 1),
+                              tuple<Scalar>(-1.0, 2.0, -1.0));
       }
     }
 
     PrintMemoryUsage("Memory after insertGlobalValues()", "tpetra-after-insert.heap");
 
-    A->fillComplete(); // DoOptimizeStorage by default
+    A->fillComplete();  // DoOptimizeStorage by default
 
     PrintMemoryUsage("Memory after fillComplete()", "tpetra-after-fillcomplete.heap");
 
@@ -132,5 +132,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/example/advanced/multiplesolve/FixedMatrixPattern.cpp b/packages/muelu/example/advanced/multiplesolve/FixedMatrixPattern.cpp
index d6cda610af83..9bae72e25a28 100644
--- a/packages/muelu/example/advanced/multiplesolve/FixedMatrixPattern.cpp
+++ b/packages/muelu/example/advanced/multiplesolve/FixedMatrixPattern.cpp
@@ -68,7 +68,7 @@
 //
 // The resulting preconditioners are identical to multigrid preconditioners built without recycling the parts described above.
 // This can be verified by using the --no-recycling option.
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -79,7 +79,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     //
     // Parameters
     //
@@ -87,7 +87,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);
     Xpetra::Parameters xpetraParameters(clp);
 
-    bool optRecycling           = true;  clp.setOption("recycling",             "no-recycling",             &optRecycling,           "Enable recycling of the multigrid preconditioner");
+    bool optRecycling = true;
+    clp.setOption("recycling", "no-recycling", &optRecycling, "Enable recycling of the multigrid preconditioner");
 
     /* DO NOT WORK YET
        bool optRecyclingRAPpattern = true;  clp.setOption("recycling-rap-pattern", "no-recycling-rap-pattern", &optRecyclingRAPpattern, "Enable recycling of Ac=RAP pattern");
@@ -97,10 +98,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     bool optRecyclingAPpattern  = false;
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // option dependencies
@@ -114,11 +115,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     //
 
     RCP<const Map> map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
-    Teuchos::RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
+    Teuchos::RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
     RCP<Matrix> A1 = Pr->BuildMatrix();
 
-    RCP<Matrix> A2 = Pr->BuildMatrix(); // TODO: generate another problem would be more meaningful (ex: scale A1)
+    RCP<Matrix> A2 = Pr->BuildMatrix();  // TODO: generate another problem would be more meaningful (ex: scale A1)
 
     //
     // First solve
@@ -142,7 +143,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // PTENT:
       RCP<Factory> PtentFact = rcp(new TentativePFactory());
       M.SetFactory("Ptent", PtentFact);
-      H.Keep("P",           PtentFact.get());
+      H.Keep("P", PtentFact.get());
     }
 
     RCP<Factory> AcFact = rcp(new RAPFactory());
@@ -162,8 +163,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       RCP<Vector> X = VectorFactory::Build(map);
       RCP<Vector> B = VectorFactory::Build(map);
 
-      X->putScalar((Scalar) 0.0);
-      B->setSeed(846930886); B->randomize();
+      X->putScalar((Scalar)0.0);
+      B->setSeed(846930886);
+      B->randomize();
 
       int nIts = 9;
       H.Iterate(*B, *X, nIts);
@@ -197,8 +199,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       RCP<Vector> X = VectorFactory::Build(map);
       RCP<Vector> B = VectorFactory::Build(map);
 
-      X->putScalar((Scalar) 0.0);
-      B->setSeed(846930886); B->randomize();
+      X->putScalar((Scalar)0.0);
+      B->setSeed(846930886);
+      B->randomize();
 
       int nIts = 9;
       H.Iterate(*B, *X, nIts);
@@ -215,8 +218,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Remove kept data from the preconditioner. This will force recomputation on future runs. "Keep" flags are also removed.
 
     if (optRecycling) {
-      //if aggregates explicitly kept: H.Delete("Aggregates", M.GetFactory("Aggregates").get());
-      H.Delete("P",           M.GetFactory("Ptent").get());
+      // if aggregates explicitly kept: H.Delete("Aggregates", M.GetFactory("Aggregates").get());
+      H.Delete("P", M.GetFactory("Ptent").get());
     }
     if (optRecyclingRAPpattern) {
       H.Delete("RAP graph", M.GetFactory("A").get());
@@ -232,16 +235,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
diff --git a/packages/muelu/example/advanced/multiplesolve/ReuseSequence.cpp b/packages/muelu/example/advanced/multiplesolve/ReuseSequence.cpp
index 7d3570de6d04..784315ede482 100644
--- a/packages/muelu/example/advanced/multiplesolve/ReuseSequence.cpp
+++ b/packages/muelu/example/advanced/multiplesolve/ReuseSequence.cpp
@@ -74,8 +74,8 @@
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
 #include <MueLu_CreateXpetraPreconditioner.hpp>
@@ -95,18 +95,21 @@
 // The resulting preconditioners are identical to multigrid preconditioners built without recycling the parts described above.
 // This can be verified by using the --no-recycling option.
 
-
-template<class Scalar>
+template <class Scalar>
 class Tensor {
-private:
-  typedef Scalar                    SC;
+ private:
+  typedef Scalar SC;
   typedef Teuchos::ScalarTraits<SC> STS;
 
-public:
-  Tensor() : useSigmaRTC_(false), is3D_(true) { }
+ public:
+  Tensor()
+    : useSigmaRTC_(false)
+    , is3D_(true) {}
 
 #ifdef HAVE_MUELU_PAMGEN
-  Tensor(const std::string& rtcString, bool is3D = true) : useSigmaRTC_(true), is3D_(is3D) {
+  Tensor(const std::string& rtcString, bool is3D = true)
+    : useSigmaRTC_(true)
+    , is3D_(is3D) {
     sigmaRTC_ = Teuchos::rcp(new PG_RuntimeCompiler::Function);
     std::string variableType;
     if (TYPE_EQUAL(Scalar, float) || TYPE_EQUAL(Scalar, std::complex<float>))
@@ -114,17 +117,17 @@ class Tensor {
     else
       variableType = "double";
 
-    if (!sigmaRTC_->addVar(variableType, "x"))          throw std::runtime_error("Error setting RTC input argument \"x\"");
-    if (!sigmaRTC_->addVar(variableType, "y"))          throw std::runtime_error("Error setting RTC input argument \"y\"");
+    if (!sigmaRTC_->addVar(variableType, "x")) throw std::runtime_error("Error setting RTC input argument \"x\"");
+    if (!sigmaRTC_->addVar(variableType, "y")) throw std::runtime_error("Error setting RTC input argument \"y\"");
     if (is3D_ &&
-        !sigmaRTC_->addVar(variableType, "z"))          throw std::runtime_error("Error setting RTC input argument \"z\"");
-    if (!sigmaRTC_->addVar(variableType, "t"))          throw std::runtime_error("Error setting RTC input argument \"t\"");
-    if (!sigmaRTC_->addVar(variableType, "sigmax"))     throw std::runtime_error("Error setting RTC input argument \"sigmax\"");
-    if (!sigmaRTC_->addVar(variableType, "sigmay"))     throw std::runtime_error("Error setting RTC input argument \"sigmay\"");
+        !sigmaRTC_->addVar(variableType, "z")) throw std::runtime_error("Error setting RTC input argument \"z\"");
+    if (!sigmaRTC_->addVar(variableType, "t")) throw std::runtime_error("Error setting RTC input argument \"t\"");
+    if (!sigmaRTC_->addVar(variableType, "sigmax")) throw std::runtime_error("Error setting RTC input argument \"sigmax\"");
+    if (!sigmaRTC_->addVar(variableType, "sigmay")) throw std::runtime_error("Error setting RTC input argument \"sigmay\"");
     if (is3D_ &&
-        !sigmaRTC_->addVar(variableType, "sigmaz"))     throw std::runtime_error("Error setting RTC input argument \"sigmaz\"");
+        !sigmaRTC_->addVar(variableType, "sigmaz")) throw std::runtime_error("Error setting RTC input argument \"sigmaz\"");
 
-    if (!sigmaRTC_->addBody(rtcString))             throw std::runtime_error("Error in RTC function compilation");
+    if (!sigmaRTC_->addBody(rtcString)) throw std::runtime_error("Error in RTC function compilation");
   }
 #endif
 
@@ -143,11 +146,11 @@ class Tensor {
     is3D_        = tensor.is3D_;
     t_           = tensor.t_;
 #ifdef HAVE_MUELU_PAMGEN
-    sigmaRTC_    = tensor.sigmaRTC_;
+    sigmaRTC_ = tensor.sigmaRTC_;
 #endif
   }
 
-private:
+ private:
   SC tensorDefault(char c, SC x, SC y, SC z) const {
     // isotropic tensor
     return STS::one();
@@ -158,15 +161,15 @@ class Tensor {
     SC sigmax, sigmay, sigmaz;
 
     int cnt = 0;
-    if (!sigmaRTC_->varValueFill(cnt++, x))       throw std::runtime_error("Could not fill \"x\"");
-    if (!sigmaRTC_->varValueFill(cnt++, y))       throw std::runtime_error("Could not fill \"y\"");
+    if (!sigmaRTC_->varValueFill(cnt++, x)) throw std::runtime_error("Could not fill \"x\"");
+    if (!sigmaRTC_->varValueFill(cnt++, y)) throw std::runtime_error("Could not fill \"y\"");
     if (is3D_ &&
-        !sigmaRTC_->varValueFill(cnt++, z))       throw std::runtime_error("Could not fill \"z\"");
-    if (!sigmaRTC_->varValueFill(cnt++, t_))      throw std::runtime_error("Could not fill \"t\"");
-    if (!sigmaRTC_->varAddrFill (cnt++, &sigmax)) throw std::runtime_error("Could not fill \"sigmax\"");
-    if (!sigmaRTC_->varAddrFill (cnt++, &sigmay)) throw std::runtime_error("Could not fill \"sigmay\"");
+        !sigmaRTC_->varValueFill(cnt++, z)) throw std::runtime_error("Could not fill \"z\"");
+    if (!sigmaRTC_->varValueFill(cnt++, t_)) throw std::runtime_error("Could not fill \"t\"");
+    if (!sigmaRTC_->varAddrFill(cnt++, &sigmax)) throw std::runtime_error("Could not fill \"sigmax\"");
+    if (!sigmaRTC_->varAddrFill(cnt++, &sigmay)) throw std::runtime_error("Could not fill \"sigmay\"");
     if (is3D_ &&
-        !sigmaRTC_->varAddrFill (cnt++, &sigmaz)) throw std::runtime_error("Could not fill \"sigmaz\"");
+        !sigmaRTC_->varAddrFill(cnt++, &sigmaz)) throw std::runtime_error("Could not fill \"sigmaz\"");
 
     sigmaRTC_->execute();
 
@@ -179,60 +182,60 @@ class Tensor {
   }
 #endif
 
-private:
-  bool   useSigmaRTC_;
-  bool   is3D_;
+ private:
+  bool useSigmaRTC_;
+  bool is3D_;
   double t_;
 
 #ifdef HAVE_MUELU_PAMGEN
-  mutable
-    Teuchos::RCP<PG_RuntimeCompiler::Function> sigmaRTC_;
+  mutable Teuchos::RCP<PG_RuntimeCompiler::Function> sigmaRTC_;
 #endif
 };
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Map, class Matrix, class MultiVector>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Map, class Matrix, class MultiVector>
 Teuchos::RCP<Matrix> BuildMatrix(bool is3D, const Tensor<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& tensor, Teuchos::ParameterList& list,
                                  const Teuchos::RCP<const Map>& map, const Teuchos::RCP<const MultiVector>& coords) {
   typedef GlobalOrdinal GO;
-  typedef LocalOrdinal  LO;
-  typedef Scalar        SC;
-  using Teuchos::ArrayView;
+  typedef LocalOrdinal LO;
+  typedef Scalar SC;
   using Teuchos::ArrayRCP;
+  using Teuchos::ArrayView;
   typedef typename MultiVector::scalar_type Real;
 
-  GO nx = list.get("nx", (GO) -1);
-  GO ny = list.get("ny", (GO) -1);
+  GO nx = list.get("nx", (GO)-1);
+  GO ny = list.get("ny", (GO)-1);
   GO nz = -1;
   if (is3D) {
     // 3D
-    nz = list.get("nz", (GO) -1);
+    nz = list.get("nz", (GO)-1);
 
     if (nx == -1 || ny == -1 || nz == -1) {
       GO n = map->getGlobalNumElements();
-      nx = (GO) Teuchos::ScalarTraits<double>::pow(n, 0.33334);
-      ny = nx; nz = nx;
+      nx   = (GO)Teuchos::ScalarTraits<double>::pow(n, 0.33334);
+      ny   = nx;
+      nz   = nx;
       TEUCHOS_TEST_FOR_EXCEPTION(nx * ny * nz != n, std::logic_error, "You need to specify nx, ny, and nz");
     }
   } else {
     // 2D
     if (nx == -1 || ny == -1) {
       GO n = map->getGlobalNumElements();
-      nx = (GO) Teuchos::ScalarTraits<double>::pow(n, 0.5);
-      ny = nx;
+      nx   = (GO)Teuchos::ScalarTraits<double>::pow(n, 0.5);
+      ny   = nx;
       TEUCHOS_TEST_FOR_EXCEPTION(nx * ny != n, std::logic_error, "You need to specify nx, ny, and nz");
     }
   }
 
-  double  one = 1.0;
-  SC  stretchx = list.get("stretchx", one);
-  SC  stretchy = list.get("stretchy", one);
-  SC  stretchz = list.get("stretchz", one);
+  double one  = 1.0;
+  SC stretchx = list.get("stretchx", one);
+  SC stretchy = list.get("stretchy", one);
+  SC stretchz = list.get("stretchz", one);
 
   // bool keepBCs = list.get("keepBCs", false);
 
   LO nnz = (is3D ? 7 : 5);
 
-  Teuchos::RCP<Matrix> A = Galeri::Xpetra::MatrixTraits<Map,Matrix>::Build(map, nnz);
+  Teuchos::RCP<Matrix> A = Galeri::Xpetra::MatrixTraits<Map, Matrix>::Build(map, nnz);
 
   LO numMyElements = map->getLocalNumElements();
   GO indexBase     = map->getIndexBase();
@@ -262,24 +265,42 @@ Teuchos::RCP<Matrix> BuildMatrix(bool is3D, const Tensor<typename Teuchos::Scala
 
       SC w;
 
-      w = tensor('x', x[center], y[center], z[center]);
-      SC b = -((left   != -1) ? tensor('x', 0.5*(x[center] + x[left]),   0.5*(y[center] + y[left]),   0.5*(z[center] + z[left]))  : w) / (stretchx*stretchx);
-      SC c = -((right  != -1) ? tensor('x', 0.5*(x[center] + x[right]),  0.5*(y[center] + y[right]),  0.5*(z[center] + z[right])) : w) / (stretchx*stretchx);
-      w = tensor('y', x[center], y[center], z[center]);
-      SC d = -((front  != -1) ? tensor('y', 0.5*(x[center] + x[front]),  0.5*(y[center] + y[front]),  0.5*(z[center] + z[front])) : w) / (stretchy*stretchy);
-      SC e = -((back   != -1) ? tensor('y', 0.5*(x[center] + x[back]),   0.5*(y[center] + y[back]),   0.5*(z[center] + z[back]))  : w) / (stretchy*stretchy);
-      w = tensor('z', x[center], y[center], z[center]);
-      SC f = -((bottom != -1) ? tensor('z', 0.5*(x[center] + x[bottom]), 0.5*(y[center] + y[bottom]), 0.5*(z[center] + z[bottom])): w) / (stretchz*stretchz);
-      SC g = -((top    != -1) ? tensor('z', 0.5*(x[center] + x[top]),    0.5*(y[center] + y[top]),    0.5*(z[center] + z[top]))   : w) / (stretchz*stretchz);
+      w    = tensor('x', x[center], y[center], z[center]);
+      SC b = -((left != -1) ? tensor('x', 0.5 * (x[center] + x[left]), 0.5 * (y[center] + y[left]), 0.5 * (z[center] + z[left])) : w) / (stretchx * stretchx);
+      SC c = -((right != -1) ? tensor('x', 0.5 * (x[center] + x[right]), 0.5 * (y[center] + y[right]), 0.5 * (z[center] + z[right])) : w) / (stretchx * stretchx);
+      w    = tensor('y', x[center], y[center], z[center]);
+      SC d = -((front != -1) ? tensor('y', 0.5 * (x[center] + x[front]), 0.5 * (y[center] + y[front]), 0.5 * (z[center] + z[front])) : w) / (stretchy * stretchy);
+      SC e = -((back != -1) ? tensor('y', 0.5 * (x[center] + x[back]), 0.5 * (y[center] + y[back]), 0.5 * (z[center] + z[back])) : w) / (stretchy * stretchy);
+      w    = tensor('z', x[center], y[center], z[center]);
+      SC f = -((bottom != -1) ? tensor('z', 0.5 * (x[center] + x[bottom]), 0.5 * (y[center] + y[bottom]), 0.5 * (z[center] + z[bottom])) : w) / (stretchz * stretchz);
+      SC g = -((top != -1) ? tensor('z', 0.5 * (x[center] + x[top]), 0.5 * (y[center] + y[top]), 0.5 * (z[center] + z[top])) : w) / (stretchz * stretchz);
 
       SC a = -(b + c + d + e + f + g);
 
-      if (left   != -1) { inds[n] = left;   vals[n++] = b; }
-      if (right  != -1) { inds[n] = right;  vals[n++] = c; }
-      if (front  != -1) { inds[n] = front;  vals[n++] = d; }
-      if (back   != -1) { inds[n] = back;   vals[n++] = e; }
-      if (bottom != -1) { inds[n] = bottom; vals[n++] = f; }
-      if (top    != -1) { inds[n] = top;    vals[n++] = g; }
+      if (left != -1) {
+        inds[n]   = left;
+        vals[n++] = b;
+      }
+      if (right != -1) {
+        inds[n]   = right;
+        vals[n++] = c;
+      }
+      if (front != -1) {
+        inds[n]   = front;
+        vals[n++] = d;
+      }
+      if (back != -1) {
+        inds[n]   = back;
+        vals[n++] = e;
+      }
+      if (bottom != -1) {
+        inds[n]   = bottom;
+        vals[n++] = f;
+      }
+      if (top != -1) {
+        inds[n]   = top;
+        vals[n++] = g;
+      }
 
       if (bottom != -1 && Galeri::Xpetra::IsBoundary3d(center, nx, ny, nz)) {
         // Neumann boundary unknown (diagonal = sum of all offdiagonal)
@@ -296,19 +317,31 @@ Teuchos::RCP<Matrix> BuildMatrix(bool is3D, const Tensor<typename Teuchos::Scala
 
       SC w;
 
-      w = tensor('x', x[center], y[center]);
-      SC b = -((left   != -1) ? tensor('x', 0.5*(x[center] + x[left]),   0.5*(y[center] + y[left]))  : w) / (stretchx*stretchx);
-      SC c = -((right  != -1) ? tensor('x', 0.5*(x[center] + x[right]),  0.5*(y[center] + y[right])) : w) / (stretchx*stretchx);
-      w = tensor('y', x[center], y[center]);
-      SC d = -((front  != -1) ? tensor('y', 0.5*(x[center] + x[front]),  0.5*(y[center] + y[front])) : w) / (stretchy*stretchy);
-      SC e = -((back   != -1) ? tensor('y', 0.5*(x[center] + x[back]),   0.5*(y[center] + y[back]))  : w) / (stretchy*stretchy);
+      w    = tensor('x', x[center], y[center]);
+      SC b = -((left != -1) ? tensor('x', 0.5 * (x[center] + x[left]), 0.5 * (y[center] + y[left])) : w) / (stretchx * stretchx);
+      SC c = -((right != -1) ? tensor('x', 0.5 * (x[center] + x[right]), 0.5 * (y[center] + y[right])) : w) / (stretchx * stretchx);
+      w    = tensor('y', x[center], y[center]);
+      SC d = -((front != -1) ? tensor('y', 0.5 * (x[center] + x[front]), 0.5 * (y[center] + y[front])) : w) / (stretchy * stretchy);
+      SC e = -((back != -1) ? tensor('y', 0.5 * (x[center] + x[back]), 0.5 * (y[center] + y[back])) : w) / (stretchy * stretchy);
 
       SC a = -(b + c + d + e);
 
-      if (left   != -1) { inds[n] = left;   vals[n++] = b; }
-      if (right  != -1) { inds[n] = right;  vals[n++] = c; }
-      if (front  != -1) { inds[n] = front;  vals[n++] = d; }
-      if (back   != -1) { inds[n] = back;   vals[n++] = e; }
+      if (left != -1) {
+        inds[n]   = left;
+        vals[n++] = b;
+      }
+      if (right != -1) {
+        inds[n]   = right;
+        vals[n++] = c;
+      }
+      if (front != -1) {
+        inds[n]   = front;
+        vals[n++] = d;
+      }
+      if (back != -1) {
+        inds[n]   = back;
+        vals[n++] = e;
+      }
 
       if (front != -1 && Galeri::Xpetra::IsBoundary2d(center, nx, ny)) {
         // Neumann boundary unknown (diagonal = sum of all offdiagonal)
@@ -333,49 +366,48 @@ Teuchos::RCP<Matrix> BuildMatrix(bool is3D, const Tensor<typename Teuchos::Scala
   return A;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ConstructData(bool is3D, const Tensor<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& tensor, const std::string& matrixType, Teuchos::ParameterList& galeriList,
-                   Xpetra::UnderlyingLib lib, Teuchos::RCP<const Teuchos::Comm<int> >& comm,
-                   Teuchos::RCP<Xpetra::Matrix      <Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                   Teuchos::RCP<const Xpetra::Map   <LocalOrdinal,GlobalOrdinal, Node> >&       map,
-                   Teuchos::RCP<Xpetra::MultiVector <typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >& coordinates,
-                   Teuchos::RCP<Xpetra::MultiVector <Scalar,LocalOrdinal,GlobalOrdinal,Node> >& nullspace) {
+                   Xpetra::UnderlyingLib lib, Teuchos::RCP<const Teuchos::Comm<int>>& comm,
+                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                   Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& map,
+                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>>& coordinates,
+                   Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& nullspace) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   if (is3D) {
     // 3D
     map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
 
   } else {
     // 2D
     map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
   }
 
-  A = BuildMatrix<SC,LO,GO,Map,CrsMatrixWrap,RealValuedMultiVector>(is3D, tensor, galeriList, map, coordinates);
+  A = BuildMatrix<SC, LO, GO, Map, CrsMatrixWrap, RealValuedMultiVector>(is3D, tensor, galeriList, map, coordinates);
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
   using namespace std::chrono;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
-  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
 
   // =========================================================================
   // Convenient definitions
@@ -384,28 +416,33 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
   SC one = STS::one(), zero = STS::zero();
 
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 20, ny = 20, nz = 20;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName = "reuse_seq.xml";    clp.setOption("xml",    &xmlFileName, "read parameters from a file");
-  std::string solveType   = "cg";               clp.setOption("solver", &solveType,   "solve type: (none | cg | standalone)");
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType      tol         = 1e-6;               clp.setOption("tol",    &tol,         "solver convergence tolerance");
-  int         maxIts      = 200;                clp.setOption("its",    &maxIts,      "maximum number of solver iterations");
-  int         dim         = 3;                  clp.setOption("dim",    &dim,         "space dimension");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "reuse_seq.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file");
+  std::string solveType = "cg";
+  clp.setOption("solver", &solveType, "solve type: (none | cg | standalone)");
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol = 1e-6;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  int dim = 3;
+  clp.setOption("dim", &dim, "space dimension");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   bool is3D = (dim == 3);
@@ -418,7 +455,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
   // Problem construction
   // =========================================================================
   // For comments, see Driver.cpp
-  out << "========================================================\n" << xpetraParameters << galeriParameters;
+  out << "========================================================\n"
+      << xpetraParameters << galeriParameters;
   std::string matrixType = galeriParameters.GetMatrixType();
 
   out << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
@@ -438,7 +476,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
 
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
   Tensor<real_type> tensor;
   if (paramList.isParameter("sigma")) {
     std::string sigmaString = paramList.get<std::string>("sigma");
@@ -447,38 +485,44 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
     out << "Switching to RTC" << std::endl;
     tensor = Tensor<real_type>(sigmaString, is3D);
 #else
-    (void)sigmaString; // fix compiler warning
+    (void)sigmaString;  // fix compiler warning
 #endif
   }
 
-  out << "Parameter list:" << std::endl << paramList << std::endl;
+  out << "Parameter list:" << std::endl
+      << paramList << std::endl;
 
   // =========================================================================
   // The LOOP
   // =========================================================================
   std::vector<std::string> reuseTypes, reuseNames;
-  reuseTypes.push_back("none"); reuseNames.push_back("none");
-  reuseTypes.push_back("S");    reuseNames.push_back("smoothers");
-  reuseTypes.push_back("tP");   reuseNames.push_back("tentative P");
-  reuseTypes.push_back("RP");   reuseNames.push_back("smoothed P and R");
-  reuseTypes.push_back("RAP");  reuseNames.push_back("coarse grids");
+  reuseTypes.push_back("none");
+  reuseNames.push_back("none");
+  reuseTypes.push_back("S");
+  reuseNames.push_back("smoothers");
+  reuseTypes.push_back("tP");
+  reuseNames.push_back("tentative P");
+  reuseTypes.push_back("RP");
+  reuseNames.push_back("smoothed P and R");
+  reuseTypes.push_back("RAP");
+  reuseNames.push_back("coarse grids");
 
   const size_t numSteps = 8;
 
   high_resolution_clock::time_point tc;
-  std::vector<duration<double>> setup_time(reuseTypes.size()*numSteps);
-  std::vector<duration<double>> solve_time(reuseTypes.size()*numSteps);
-  std::vector<int>              num_its   (reuseTypes.size()*numSteps);
+  std::vector<duration<double>> setup_time(reuseTypes.size() * numSteps);
+  std::vector<duration<double>> solve_time(reuseTypes.size() * numSteps);
+  std::vector<int> num_its(reuseTypes.size() * numSteps);
 
   for (size_t k = 0; k < reuseTypes.size(); k++) {
     out << thickSeparator << " " << reuseTypes[k] << " " << thickSeparator << std::endl;
 
     paramList.set("reuse: type", reuseTypes[k]);
 
-    RCP<Matrix>           A;
-    RCP<const Map>        map;
+    RCP<Matrix> A;
+    RCP<const Map> map;
     RCP<RealValuedMultiVector> coordinates;
-    RCP<MultiVector>      nullspace;
+    RCP<MultiVector> nullspace;
 
     tensor.setT(0);
 
@@ -492,7 +536,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
     A->apply(*X, *B);
 
     Teuchos::ParameterList userParamList = paramList.sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+    userParamList.set<RCP<RealValuedMultiVector>>("Coordinates", coordinates);
     RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner(A, paramList);
 
     for (size_t t = 1; t < numSteps; t++) {
@@ -508,7 +552,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
         H = MueLu::CreateXpetraPreconditioner(A, paramList);
       else
         MueLu::ReuseXpetraPreconditioner(A, H);
-      setup_time[k*numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
+      setup_time[k * numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
 
       X->putScalar(zero);
 
@@ -520,22 +564,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
 
         tc = high_resolution_clock::now();
         H->Iterate(*B, *X, tol);
-        solve_time[k*numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
+        solve_time[k * numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
 
       } else if (solveType == "cg" || solveType == "gmres") {
         H->IsPreconditioner(true);
 
 #ifdef HAVE_MUELU_BELOS
         // Operator and Multivector type that will be used with Belos
-        typedef MultiVector          MV;
+        typedef MultiVector MV;
         typedef Belos::OperatorT<MV> OP;
 
         // Define Operator and Preconditioner
-        Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-        Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+        Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+        Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
         // Construct a Belos LinearProblem object
-        RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+        RCP<Belos::LinearProblem<SC, MV, OP>> belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
         belosProblem->setRightPrec(belosPrec);
 
         bool set = belosProblem->setProblem();
@@ -546,16 +590,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
 
         // Belos parameter list
         Teuchos::ParameterList belosList;
-        belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-        belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-        belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-        belosList.set("Output Frequency",      1);
-        belosList.set("Output Style",          Belos::Brief);
+        belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+        belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+        belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+        belosList.set("Output Frequency", 1);
+        belosList.set("Output Style", Belos::Brief);
 
         // Create an iterative solver manager
-        RCP< Belos::SolverManager<SC, MV, OP> > solver;
+        RCP<Belos::SolverManager<SC, MV, OP>> solver;
         if (solveType == "cg") {
-          solver = rcp(new Belos::PseudoBlockCGSolMgr   <SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+          solver = rcp(new Belos::PseudoBlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
         } else if (solveType == "gmres") {
           solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
         }
@@ -563,22 +607,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
         // Perform solve
         Belos::ReturnType ret = Belos::Unconverged;
 
-        tc = high_resolution_clock::now();
-        ret = solver->solve();
-        solve_time[k*numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
+        tc                           = high_resolution_clock::now();
+        ret                          = solver->solve();
+        solve_time[k * numSteps + t] = duration_cast<duration<double>>(high_resolution_clock::now() - tc);
 
         // Get the number of iterations for this solve.
         out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
         // Check convergence
         if (ret != Belos::Converged) {
-          out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
-          num_its[k*numSteps+t] = -1;
+          out << std::endl
+              << "ERROR:  Belos did not converge! " << std::endl;
+          num_its[k * numSteps + t] = -1;
 
         } else {
-          out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-          num_its[k*numSteps+t] = solver->getNumIters();
+          out << std::endl
+              << "SUCCESS:  Belos converged!" << std::endl;
+          num_its[k * numSteps + t] = solver->getNumIters();
         }
-#endif //ifdef HAVE_MUELU_BELOS
+#endif  // ifdef HAVE_MUELU_BELOS
       } else {
         throw MueLu::Exceptions::RuntimeError("Unknown solver type: \"" + solveType + "\"");
       }
@@ -590,18 +636,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib,  int a
     out << thinSeparator << std::endl;
     for (size_t k = 0; k < reuseTypes.size(); k++)
       printf("step #%d reuse \"%20s\": setup = %5.2e, solve = %5.2e [%3d], total = %5.2e\n", static_cast<int>(t), reuseNames[k].c_str(),
-             setup_time[k*numSteps+t].count(), solve_time[k*numSteps+t].count(), num_its[k*numSteps+t],
-             setup_time[k*numSteps+t].count() + solve_time[k*numSteps+t].count());
+             setup_time[k * numSteps + t].count(), solve_time[k * numSteps + t].count(), num_its[k * numSteps + t],
+             setup_time[k * numSteps + t].count() + solve_time[k * numSteps + t].count());
   }
 
   return EXIT_SUCCESS;
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/example/advanced/multiplesolve/StandardReuse.cpp b/packages/muelu/example/advanced/multiplesolve/StandardReuse.cpp
index 1e17f6d41337..8ff37de279ee 100644
--- a/packages/muelu/example/advanced/multiplesolve/StandardReuse.cpp
+++ b/packages/muelu/example/advanced/multiplesolve/StandardReuse.cpp
@@ -79,21 +79,20 @@
 // The resulting preconditioners are identical to multigrid preconditioners built without recycling the parts described above.
 // This can be verified by using the --no-recycling option.
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ConstructData(const std::string& matrixType, Teuchos::ParameterList& galeriList,
                    Xpetra::UnderlyingLib lib, Teuchos::RCP<const Teuchos::Comm<int> >& comm,
-                   Teuchos::RCP<Xpetra::Matrix      <Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                   Teuchos::RCP<const Xpetra::Map   <LocalOrdinal,GlobalOrdinal, Node> >&       map,
-                   Teuchos::RCP<Xpetra::MultiVector <typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >& coordinates,
-                   Teuchos::RCP<Xpetra::MultiVector <Scalar,LocalOrdinal,GlobalOrdinal,Node> >& nullspace) {
+                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
+                   Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
+                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& coordinates,
+                   Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& nullspace) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
   //                                 d1  d2  d3
@@ -109,36 +108,36 @@ void ConstructData(const std::string& matrixType, Teuchos::ParameterList& galeri
   // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
   // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
   if (matrixType == "Laplace1D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
   } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
              matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
   } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
   }
 
   // Expand map to do multiple DOF per node for block problems
   if (matrixType == "Elasticity2D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
   if (matrixType == "Elasticity3D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
   if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
     // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-    galeriList.set("right boundary" , "Neumann");
+    galeriList.set("right boundary", "Neumann");
     galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
+    galeriList.set("top boundary", "Neumann");
+    galeriList.set("front boundary", "Neumann");
+    galeriList.set("back boundary", "Neumann");
   }
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, galeriList);
   A = Pr->BuildMatrix();
 
   if (matrixType == "Elasticity2D" ||
@@ -148,16 +147,16 @@ void ConstructData(const std::string& matrixType, Teuchos::ParameterList& galeri
   }
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
 
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -171,27 +170,31 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
   SC one = STS::one(), zero = STS::zero();
 
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName = "";     clp.setOption("xml",                &xmlFileName, "read parameters from a file");
-  int         numRebuilds = 0;      clp.setOption("rebuild",            &numRebuilds, "#times to rebuild hierarchy");
-  bool        useFilter   = true;   clp.setOption("filter", "nofilter", &useFilter,   "Print out only Setup times");
-  bool        modify      = true;   clp.setOption("modify", "nomodify", &modify,      "Change values of the matrix used for reuse");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file");
+  int numRebuilds = 0;
+  clp.setOption("rebuild", &numRebuilds, "#times to rebuild hierarchy");
+  bool useFilter = true;
+  clp.setOption("filter", "nofilter", &useFilter, "Print out only Setup times");
+  bool modify = true;
+  clp.setOption("modify", "nomodify", &modify, "Change values of the matrix used for reuse");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   // Retrieve matrix parameters (they may have been changed on the command line)
@@ -202,12 +205,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
   // Problem construction
   // =========================================================================
   // For comments, see Driver.cpp
-  out << "========================================================\n" << xpetraParameters << galeriParameters;
+  out << "========================================================\n"
+      << xpetraParameters << galeriParameters;
   std::string matrixType = galeriParameters.GetMatrixType();
-  RCP<Matrix>           A, B;
-  RCP<const Map>        map;
-  RCP<RealValuedMultiVector>  coordinates;
-  RCP<MultiVector>      nullspace;
+  RCP<Matrix> A, B;
+  RCP<const Map> map;
+  RCP<RealValuedMultiVector> coordinates;
+  RCP<MultiVector> nullspace;
   ConstructData(matrixType, galeriList, lib, comm, A, map, coordinates, nullspace);
 
   if (modify) {
@@ -237,7 +241,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
 
   Teuchos::ParameterList paramList;
   paramList.set("verbosity", "none");
-  if(lib == Xpetra::UseEpetra) {
+  if (lib == Xpetra::UseEpetra) {
     out << "Setting: \"use kokkos refactor\" to: false" << std::endl;
     paramList.set("use kokkos refactor", false);
   }
@@ -289,10 +293,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
   // Setup #2-inf (reuse)
   // =========================================================================
   std::vector<std::string> reuseTypes, reuseNames;
-  reuseTypes.push_back("none"); reuseNames.push_back("none");
-  reuseTypes.push_back("S");    reuseNames.push_back("smoothers");
-  reuseTypes.push_back("tP");   reuseNames.push_back("tentative P");
-  reuseTypes.push_back("RP");   reuseNames.push_back("smoothed P and R");
+  reuseTypes.push_back("none");
+  reuseNames.push_back("none");
+  reuseTypes.push_back("S");
+  reuseNames.push_back("smoothers");
+  reuseTypes.push_back("tP");
+  reuseNames.push_back("tentative P");
+  reuseTypes.push_back("RP");
+  reuseNames.push_back("smoothed P and R");
 
   for (size_t k = 0; k < reuseTypes.size(); k++) {
     out << thickSeparator << " " << reuseTypes[k] << " " << thickSeparator << std::endl;
@@ -310,7 +318,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
     // Reuse setup
     RCP<Matrix> Bcopy = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(B);
 
-    RCP<Teuchos::Time> tm = TimeMonitor::getNewTimer("Setup #" + MueLu::toString(k+2) + ": reuse " + reuseNames[k]);
+    RCP<Teuchos::Time> tm = TimeMonitor::getNewTimer("Setup #" + MueLu::toString(k + 2) + ": reuse " + reuseNames[k]);
     for (int i = 0; i <= numRebuilds; i++) {
       out << thinSeparator << " " << reuseTypes[k] << " (rebuild #" << i << ") " << thinSeparator << std::endl;
 
@@ -350,11 +358,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int ar
   return EXIT_SUCCESS;
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/example/advanced/separatesetups/SeparateSetups.cpp b/packages/muelu/example/advanced/separatesetups/SeparateSetups.cpp
index c4e0bef32ade..3935c3f75a10 100644
--- a/packages/muelu/example/advanced/separatesetups/SeparateSetups.cpp
+++ b/packages/muelu/example/advanced/separatesetups/SeparateSetups.cpp
@@ -72,7 +72,7 @@
 #include <Galeri_XpetraParameters.hpp>
 #include <Galeri_XpetraProblemFactory.hpp>
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -88,19 +88,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     //
     // Process command line arguments
     //
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 81); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 81);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                  // manage parameters of xpetra
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
       default:;
     }
 
@@ -114,46 +114,47 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<const Map> map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
 
     // Matrix
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());
     RCP<Matrix> A = Pr->BuildMatrix();
 
     // User defined nullspace
-    RCP<MultiVector> nullSpace = VectorFactory::Build(map,1); nullSpace->putScalar((SC) 1.0);
+    RCP<MultiVector> nullSpace = VectorFactory::Build(map, 1);
+    nullSpace->putScalar((SC)1.0);
 
     // Define B
-    RCP<Vector> X = VectorFactory::Build(map,1);
-    RCP<Vector> B = VectorFactory::Build(map,1);
+    RCP<Vector> X = VectorFactory::Build(map, 1);
+    RCP<Vector> B = VectorFactory::Build(map, 1);
     X->setSeed(846930886);
     X->randomize();
     A->apply(*X, *B, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
     // X = 0
-    X->putScalar((SC) 0.0);
+    X->putScalar((SC)0.0);
 
     //
     // Create a multigrid configuration
     //
 
     // Transfer operators
-    RCP<TentativePFactory> TentativePFact = rcp( new TentativePFactory() );
-    RCP<SaPFactory>        SaPFact        = rcp( new SaPFactory() );
-    RCP<TransPFactory>     RFact          = rcp( new TransPFactory());
+    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+    RCP<SaPFactory> SaPFact               = rcp(new SaPFactory());
+    RCP<TransPFactory> RFact              = rcp(new TransPFactory());
 
     FactoryManager M;
     M.SetFactory("Ptent", TentativePFact);
-    M.SetFactory("P",     SaPFact);
-    M.SetFactory("R",     RFact);
+    M.SetFactory("P", SaPFact);
+    M.SetFactory("R", RFact);
 
-    M.SetFactory("Smoother", Teuchos::null);      //skips smoother setup
-    M.SetFactory("CoarseSolver", Teuchos::null);  //skips coarsest solve setup
+    M.SetFactory("Smoother", Teuchos::null);      // skips smoother setup
+    M.SetFactory("CoarseSolver", Teuchos::null);  // skips coarsest solve setup
 
     //
     // Multigrid setup phase
     //
 
     int startLevel = 0;
-    int maxLevels = 10;
+    int maxLevels  = 10;
 
     std::cout << "=============== Setup transfers only ====================" << std::endl;
 
@@ -165,11 +166,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     finestLevel->Set("Nullspace", nullSpace);
 
     // Indicate which Hierarchy operators we want to keep
-    H.Keep("P", SaPFact.get());  //SaPFact is the generating factory for P.
-    H.Keep("R", RFact.get());    //RFact is the generating factory for R.
-    H.Keep("Ptent", TentativePFact.get());  //SaPFact is the generating factory for P.
+    H.Keep("P", SaPFact.get());             // SaPFact is the generating factory for P.
+    H.Keep("R", RFact.get());               // RFact is the generating factory for R.
+    H.Keep("Ptent", TentativePFact.get());  // SaPFact is the generating factory for P.
 
-    H.Setup(M,startLevel,maxLevels);
+    H.Setup(M, startLevel, maxLevels);
 
     std::cout << "=============== Setup smoothers only ====================" << std::endl;
 
@@ -180,19 +181,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Create Gauss-Seidel smoother.
     std::string ifpackType = "RELAXATION";
     Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 3);
-    ifpackList.set("relaxation: damping factor", (SC) 1.0);
+    ifpackList.set("relaxation: sweeps", (LO)3);
+    ifpackList.set("relaxation: damping factor", (SC)1.0);
     RCP<SmootherPrototype> smootherPrototype = rcp(new TrilinosSmoother(ifpackType, ifpackList));
 
     M.SetFactory("Smoother", rcp(new SmootherFactory(smootherPrototype)));
 
     // Create coarsest solver.
-    RCP<SmootherPrototype> coarseSolverPrototype = rcp( new DirectSolver() );
-    RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, Teuchos::null) );
+    RCP<SmootherPrototype> coarseSolverPrototype = rcp(new DirectSolver());
+    RCP<SmootherFactory> coarseSolverFact        = rcp(new SmootherFactory(coarseSolverPrototype, Teuchos::null));
     M.SetFactory("CoarseSolver", coarseSolverFact);
 
     // Note that we pass the number of levels back in.
-    H.Setup(M,startLevel, H.GetNumLevels());
+    H.Setup(M, startLevel, H.GetNumLevels());
 
     std::cout << "=============== Solve ====================" << std::endl;
 
@@ -211,23 +212,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     if (comm->getRank() == 0) {
       std::ios::fmtflags f(std::cout.flags());
       std::cout << "||Residual|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorms << std::endl;
-       std::cout.flags(f);
+      std::cout.flags(f);
     }
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
diff --git a/packages/muelu/example/basic/Simple.cpp b/packages/muelu/example/basic/Simple.cpp
index d8dc0fb308e3..6a16c8650772 100644
--- a/packages/muelu/example/basic/Simple.cpp
+++ b/packages/muelu/example/basic/Simple.cpp
@@ -77,17 +77,16 @@
 #include <MatrixLoad.hpp>
 #include <DriverCore.hpp>
 
-
 /*********************************************************************/
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -99,45 +98,68 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // =========================================================================
   typedef Teuchos::ScalarTraits<SC> STS;
   typedef typename STS::coordinateType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "";                 clp.setOption("xml",                   &xmlFileName,       "read parameters from an xml file");
-  std::string yamlFileName      = "";                 clp.setOption("yaml",                  &yamlFileName,      "read parameters from a yaml file");
-  bool        printTimings      = true;               clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-  std::string timingsFormat     = "table-fixed";      clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-  std::string solveType         = "belos";            clp.setOption("solver",                &solveType,         "solve type: (none | belos)");
-  std::string belosType         = "cg";               clp.setOption("belosType",             &belosType,         "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
-  double      tol               = 1e-12;              clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-  bool        binaryFormat      = false;              clp.setOption("binary", "ascii",       &binaryFormat,      "read matrices in binary format");
-  std::string rowMapFile;                             clp.setOption("rowmap",                &rowMapFile,        "map data file");
-  std::string colMapFile;                             clp.setOption("colmap",                &colMapFile,        "colmap data file");
-  std::string domainMapFile;                          clp.setOption("domainmap",             &domainMapFile,     "domainmap data file");
-  std::string rangeMapFile;                           clp.setOption("rangemap",              &rangeMapFile,      "rangemap data file");
-  std::string matrixFile;                             clp.setOption("matrix",                &matrixFile,        "matrix data file");
-  std::string rhsFile;                                clp.setOption("rhs",                   &rhsFile,           "rhs data file");
-  std::string coordFile;                              clp.setOption("coords",                &coordFile,         "coordinates data file");
-  std::string coordMapFile;                           clp.setOption("coordsmap",             &coordMapFile,      "coordinates map data file");
-  std::string nullFile;                               clp.setOption("nullspace",             &nullFile,          "nullspace data file");
-  std::string materialFile;                           clp.setOption("material",              &materialFile,      "material data file");
-  int         maxIts            = 200;                clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-  int         numVectors        = 1;                  clp.setOption("multivector",           &numVectors,        "number of rhs to solve simultaneously");
-  bool        scaleResidualHist = true;               clp.setOption("scale", "noscale",      &scaleResidualHist, "scaled Krylov residual history");
-  bool        solvePreconditioned = true;             clp.setOption("solve-preconditioned","no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
-  int         cacheSize = 0;                          clp.setOption("cachesize",             &cacheSize,         "cache size (in KB)");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+  std::string yamlFileName = "";
+  clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  std::string timingsFormat = "table-fixed";
+  clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+  std::string solveType = "belos";
+  clp.setOption("solver", &solveType, "solve type: (none | belos)");
+  std::string belosType = "cg";
+  clp.setOption("belosType", &belosType, "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  bool binaryFormat = false;
+  clp.setOption("binary", "ascii", &binaryFormat, "read matrices in binary format");
+  std::string rowMapFile;
+  clp.setOption("rowmap", &rowMapFile, "map data file");
+  std::string colMapFile;
+  clp.setOption("colmap", &colMapFile, "colmap data file");
+  std::string domainMapFile;
+  clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+  std::string rangeMapFile;
+  clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+  std::string matrixFile;
+  clp.setOption("matrix", &matrixFile, "matrix data file");
+  std::string rhsFile;
+  clp.setOption("rhs", &rhsFile, "rhs data file");
+  std::string coordFile;
+  clp.setOption("coords", &coordFile, "coordinates data file");
+  std::string coordMapFile;
+  clp.setOption("coordsmap", &coordMapFile, "coordinates map data file");
+  std::string nullFile;
+  clp.setOption("nullspace", &nullFile, "nullspace data file");
+  std::string materialFile;
+  clp.setOption("material", &materialFile, "material data file");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  int numVectors = 1;
+  clp.setOption("multivector", &numVectors, "number of rhs to solve simultaneously");
+  bool scaleResidualHist = true;
+  clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
+  bool solvePreconditioned = true;
+  clp.setOption("solve-preconditioned", "no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
+  int cacheSize = 0;
+  clp.setOption("cachesize", &cacheSize, "cache size (in KB)");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   TEUCHOS_TEST_FOR_EXCEPTION(xmlFileName != "" && yamlFileName != "", std::runtime_error,
@@ -145,7 +167,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream &out       = *fancy;
   out.setOutputToRootOnly(0);
 
   ParameterList paramList;
@@ -160,7 +182,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   if (inst == Xpetra::COMPLEX_INT_INT && solveType == "belos") {
     belosType = "gmres";
-    out << "WARNING: CG will not work with COMPLEX scalars, switching to GMRES"<<std::endl;
+    out << "WARNING: CG will not work with COMPLEX scalars, switching to GMRES" << std::endl;
   }
 
   // Retrieve matrix parameters (they may have been changed on the command line)
@@ -177,15 +199,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-  RCP<Matrix>                 A;
-  RCP<const Map>              map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<RealValuedMultiVector> coordinates;
   RCP<MultiVector> nullspace;
   RCP<MultiVector> material;
   RCP<MultiVector> X, B;
 
   // Load the matrix off disk (or generate it via Galeri)
-  MatrixLoad<SC,LO,GO,NO>(comm,lib,binaryFormat,matrixFile,rhsFile,rowMapFile,colMapFile,domainMapFile,rangeMapFile,coordFile,coordMapFile,nullFile,materialFile,map,A,coordinates,nullspace,material,X,B,numVectors,galeriParameters,xpetraParameters,galeriStream);
+  MatrixLoad<SC, LO, GO, NO>(comm, lib, binaryFormat, matrixFile, rhsFile, rowMapFile, colMapFile, domainMapFile, rangeMapFile, coordFile, coordMapFile, nullFile, materialFile, map, A, coordinates, nullspace, material, X, B, numVectors, galeriParameters, xpetraParameters, galeriStream);
   comm->barrier();
   tm = Teuchos::null;
   out << galeriStream.str();
@@ -193,17 +215,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // =========================================================================
   // Preconditioner construction
   // =========================================================================
-  bool useML   = paramList.isParameter("use external multigrid package") && (paramList.get<std::string>("use external multigrid package") == "ml");
-  out<<"*********** MueLu ParameterList ***********"<<std::endl;
-  out<<paramList;
-  out<<"*******************************************"<<std::endl;
+  bool useML = paramList.isParameter("use external multigrid package") && (paramList.get<std::string>("use external multigrid package") == "ml");
+  out << "*********** MueLu ParameterList ***********" << std::endl;
+  out << paramList;
+  out << "*******************************************" << std::endl;
 
   RCP<Hierarchy> H;
   RCP<Operator> Prec;
   {
     comm->barrier();
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup")));
-    PreconditionerSetup(A,coordinates,nullspace,material,paramList,false,false,useML,false,0,H,Prec);
+    PreconditionerSetup(A, coordinates, nullspace, material, paramList, false, false, useML, false, 0, H, Prec);
     comm->barrier();
     tm = Teuchos::null;
   }
@@ -213,29 +235,31 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // =========================================================================
   {
     comm->barrier();
-    SystemSolve(A,X,B,H,Prec,out,solveType,belosType,false,false,useML,cacheSize,0,scaleResidualHist,solvePreconditioned,maxIts,tol);
+    SystemSolve(A, X, B, H, Prec, out, solveType, belosType, false, false, useML, cacheSize, 0, scaleResidualHist, solvePreconditioned, maxIts, tol);
     comm->barrier();
   }
 
-  tm = Teuchos::null;
+  tm                = Teuchos::null;
   globalTimeMonitor = Teuchos::null;
 
   if (printTimings) {
     RCP<ParameterList> reportParams = rcp(new ParameterList);
     if (timingsFormat == "yaml") {
-      reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-      reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
+      reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+      reportParams->set("YAML style", "compact");  // "spacious" or "compact"
     }
-    reportParams->set("How to merge timer sets",   "Union");
-    reportParams->set("alwaysWriteLocal",          false);
-    reportParams->set("writeGlobalStats",          true);
-    reportParams->set("writeZeroTimers",           false);
+    reportParams->set("How to merge timer sets", "Union");
+    reportParams->set("alwaysWriteLocal", false);
+    reportParams->set("writeGlobalStats", true);
+    reportParams->set("writeZeroTimers", false);
 
     const std::string filter = "";
 
     std::ios_base::fmtflags ff(out.flags());
-    if (timingsFormat == "table-fixed") out << std::fixed;
-    else                                out << std::scientific;
+    if (timingsFormat == "table-fixed")
+      out << std::fixed;
+    else
+      out << std::scientific;
     TimeMonitor::report(comm.ptr(), out, filter, reportParams);
     out << std::setiosflags(ff);
   }
@@ -254,5 +278,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/example/basic/Stratimikos.cpp b/packages/muelu/example/basic/Stratimikos.cpp
index 998d5039194b..b7e242743250 100644
--- a/packages/muelu/example/basic/Stratimikos.cpp
+++ b/packages/muelu/example/basic/Stratimikos.cpp
@@ -82,26 +82,24 @@ The source code is not MueLu specific and can be used with any Stratimikos strat
 // Galeri includes
 #include <Galeri_XpetraParameters.hpp>
 
-
-template<typename Scalar,class LocalOrdinal,class GlobalOrdinal,class Node>
+template <typename Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
-  #include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
   typedef Teuchos::ScalarTraits<Scalar> STS;
   typedef typename STS::coordinateType real_type;
-  typedef Xpetra::MultiVector<real_type,LocalOrdinal,GlobalOrdinal,Node> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, Node> RealValuedMultiVector;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ParameterList;
   using Teuchos::TimeMonitor;
 
   bool success = false;
   bool verbose = true;
   try {
-
     //
     // MPI initialization
     //
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     //
     // Parameters
@@ -109,37 +107,55 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // manage parameters of the test case
     Galeri::Xpetra::Parameters<GlobalOrdinal> matrixParameters(clp, 100, 100, 100, "Laplace2D");
     // manage parameters of Xpetra
-    Xpetra::Parameters                        xpetraParameters(clp);
+    Xpetra::Parameters xpetraParameters(clp);
 
     // command line parameters
-    std::string xmlFileName       = "stratimikos_ParameterList.xml"; clp.setOption("xml",      &xmlFileName,       "read parameters from an xml file");
-    std::string yamlFileName      = "";                 clp.setOption("yaml",                  &yamlFileName,      "read parameters from a yaml file");
-    bool        printTimings      = false;              clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-    bool        use_stacked_timer = false;              clp.setOption("stacked-timer", "no-stacked-timer", &use_stacked_timer, "Run with or without stacked timer output");
-    std::string timingsFormat     = "table-fixed";      clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-    bool        binaryFormat      = false;              clp.setOption("binary", "ascii",       &binaryFormat,      "read matrices in binary format");
-    std::string rowMapFile;                             clp.setOption("rowmap",                &rowMapFile,        "map data file");
-    std::string colMapFile;                             clp.setOption("colmap",                &colMapFile,        "colmap data file");
-    std::string domainMapFile;                          clp.setOption("domainmap",             &domainMapFile,     "domainmap data file");
-    std::string rangeMapFile;                           clp.setOption("rangemap",              &rangeMapFile,      "rangemap data file");
-    std::string matrixFile;                             clp.setOption("matrix",                &matrixFile,        "matrix data file");
-    std::string rhsFile;                                clp.setOption("rhs",                   &rhsFile,           "rhs data file");
-    std::string coordFile;                              clp.setOption("coords",                &coordFile,         "coordinates data file");
-    std::string coordMapFile;                           clp.setOption("coordsmap",             &coordMapFile,      "coordinates map data file");
-    std::string nullFile;                               clp.setOption("nullspace",             &nullFile,          "nullspace data file");
-    std::string materialFile;                           clp.setOption("material",              &materialFile,      "material data file");
-    int         numVectors        = 1;                  clp.setOption("multivector",           &numVectors,        "number of rhs to solve simultaneously");
-    int         numSolves         = 1;                  clp.setOption("numSolves",             &numSolves,         "number of times the system should be solved");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    std::string xmlFileName = "stratimikos_ParameterList.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+    std::string yamlFileName = "";
+    clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+    bool printTimings = false;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    bool use_stacked_timer = false;
+    clp.setOption("stacked-timer", "no-stacked-timer", &use_stacked_timer, "Run with or without stacked timer output");
+    std::string timingsFormat = "table-fixed";
+    clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+    bool binaryFormat = false;
+    clp.setOption("binary", "ascii", &binaryFormat, "read matrices in binary format");
+    std::string rowMapFile;
+    clp.setOption("rowmap", &rowMapFile, "map data file");
+    std::string colMapFile;
+    clp.setOption("colmap", &colMapFile, "colmap data file");
+    std::string domainMapFile;
+    clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+    std::string rangeMapFile;
+    clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+    std::string matrixFile;
+    clp.setOption("matrix", &matrixFile, "matrix data file");
+    std::string rhsFile;
+    clp.setOption("rhs", &rhsFile, "rhs data file");
+    std::string coordFile;
+    clp.setOption("coords", &coordFile, "coordinates data file");
+    std::string coordMapFile;
+    clp.setOption("coordsmap", &coordMapFile, "coordinates map data file");
+    std::string nullFile;
+    clp.setOption("nullspace", &nullFile, "nullspace data file");
+    std::string materialFile;
+    clp.setOption("material", &materialFile, "material data file");
+    int numVectors = 1;
+    clp.setOption("multivector", &numVectors, "number of rhs to solve simultaneously");
+    int numSolves = 1;
+    clp.setOption("numSolves", &numSolves, "number of times the system should be solved");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream &out       = *fancy;
     out.setOutputToRootOnly(0);
 
     // Set up timers
@@ -161,15 +177,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Construct the problem
     //
 
-    RCP<Matrix>                A;
-    RCP<const Map>             map;
+    RCP<Matrix> A;
+    RCP<const Map> map;
     RCP<RealValuedMultiVector> coordinates;
-    RCP<MultiVector>           nullspace;
-    RCP<MultiVector>           material;
-    RCP<MultiVector>           X, B;
+    RCP<MultiVector> nullspace;
+    RCP<MultiVector> material;
+    RCP<MultiVector> X, B;
 
     std::ostringstream galeriStream;
-    MatrixLoad<SC,LocalOrdinal,GlobalOrdinal,Node>(comm,lib,binaryFormat,matrixFile,rhsFile,rowMapFile,colMapFile,domainMapFile,rangeMapFile,coordFile,coordMapFile,nullFile,materialFile,map,A,coordinates,nullspace,material,X,B,numVectors,matrixParameters,xpetraParameters,galeriStream);
+    MatrixLoad<SC, LocalOrdinal, GlobalOrdinal, Node>(comm, lib, binaryFormat, matrixFile, rhsFile, rowMapFile, colMapFile, domainMapFile, rangeMapFile, coordFile, coordMapFile, nullFile, materialFile, map, A, coordinates, nullspace, material, X, B, numVectors, matrixParameters, xpetraParameters, galeriStream);
     out << galeriStream.str();
     X->putScalar(0);
 
@@ -177,11 +193,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Build Thyra linear algebra objects
     //
 
-    RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xpCrsA = Teuchos::rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(A);
+    RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xpCrsA = Teuchos::rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(A);
 
-    RCP<const Thyra::LinearOpBase<Scalar> >    thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(xpCrsA->getCrsMatrix());
-    RCP<      Thyra::MultiVectorBase<Scalar> > thyraX = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(X));
-    RCP<const Thyra::MultiVectorBase<Scalar> > thyraB = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(B);
+    RCP<const Thyra::LinearOpBase<Scalar> > thyraA    = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(xpCrsA->getCrsMatrix());
+    RCP<Thyra::MultiVectorBase<Scalar> > thyraX       = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(X));
+    RCP<const Thyra::MultiVectorBase<Scalar> > thyraB = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(B);
 
     //
     // Build Stratimikos solver
@@ -190,22 +206,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // This is the Stratimikos main class (= factory of solver factory).
     Stratimikos::LinearSolverBuilder<Scalar> linearSolverBuilder;
     // Register MueLu as a Stratimikos preconditioner strategy.
-    Stratimikos::enableMueLu<Scalar,LocalOrdinal,GlobalOrdinal,Node>(linearSolverBuilder);
+    Stratimikos::enableMueLu<Scalar, LocalOrdinal, GlobalOrdinal, Node>(linearSolverBuilder);
 
     // add coordinates and nullspace to parameter list
     if (paramList->isSublist("Preconditioner Types") &&
         paramList->sublist("Preconditioner Types").isSublist("MueLu")) {
-        ParameterList& userParamList = paramList->sublist("Preconditioner Types").sublist("MueLu").sublist("user data");
-        userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-        userParamList.set<RCP<MultiVector> >("Nullspace", nullspace);
-      }
+      ParameterList &userParamList = paramList->sublist("Preconditioner Types").sublist("MueLu").sublist("user data");
+      userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+      userParamList.set<RCP<MultiVector> >("Nullspace", nullspace);
+    }
 
     // Setup solver parameters using a Stratimikos parameter list.
     linearSolverBuilder.setParameterList(paramList);
 
     // Build a new "solver factory" according to the previously specified parameter list.
     RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
-    auto precFactory = solverFactory->getPreconditionerFactory();
+    auto precFactory                                                = solverFactory->getPreconditionerFactory();
     RCP<Thyra::PreconditionerBase<Scalar> > prec;
     Teuchos::RCP<Thyra::LinearOpWithSolveBase<Scalar> > thyraInverseA;
     if (!precFactory.is_null()) {
@@ -244,19 +260,21 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       } else {
         RCP<ParameterList> reportParams = rcp(new ParameterList);
         if (timingsFormat == "yaml") {
-          reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-          reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
+          reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+          reportParams->set("YAML style", "compact");  // "spacious" or "compact"
         }
-        reportParams->set("How to merge timer sets",   "Union");
-        reportParams->set("alwaysWriteLocal",          false);
-        reportParams->set("writeGlobalStats",          true);
-        reportParams->set("writeZeroTimers",           false);
+        reportParams->set("How to merge timer sets", "Union");
+        reportParams->set("alwaysWriteLocal", false);
+        reportParams->set("writeGlobalStats", true);
+        reportParams->set("writeZeroTimers", false);
 
         const std::string filter = "";
 
         std::ios_base::fmtflags ff(out.flags());
-        if (timingsFormat == "table-fixed") out << std::fixed;
-        else                                out << std::scientific;
+        if (timingsFormat == "table-fixed")
+          out << std::fixed;
+        else
+          out << std::scientific;
         TimeMonitor::report(comm.ptr(), out, filter, reportParams);
         out << std::setiosflags(ff);
       }
@@ -264,18 +282,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     TimeMonitor::clearCounters();
     out << std::endl;
-
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/example/basic/Teko.cpp b/packages/muelu/example/basic/Teko.cpp
index a2c75895118a..29ab168ef2d6 100644
--- a/packages/muelu/example/basic/Teko.cpp
+++ b/packages/muelu/example/basic/Teko.cpp
@@ -75,12 +75,11 @@
 #include "Stratimikos_DefaultLinearSolverBuilder.hpp"
 #include "Stratimikos_MueLuHelpers.hpp"
 
-
 // Belos includes
 #include "BelosConfigDefs.hpp"
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosThyraAdapter.hpp" // Requires Stratimikos...
+#include "BelosThyraAdapter.hpp"  // Requires Stratimikos...
 
 #include <iostream> /*@ \label{lned:end-includes} @*/
 
@@ -91,80 +90,81 @@ using Teuchos::rcpFromRef;
 
 RCP<Teuchos::ParameterList> buildLibPL();
 
-template<class SC,class LO, class GO, class NO>
-void GenerateDefault_2x2_Splitting(const RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> >& crsMat,
-                                   Teko::LinearOp & A,
-                                   Teko::MultiVector & x,
-                                   Teko::MultiVector & b)
-{
-   typedef Tpetra::Vector<SC,LO,GO,NO>        TP_Vec;
-   typedef Tpetra::CrsMatrix<SC,LO,GO,NO>     TP_Crs;
-   typedef Tpetra::Operator<SC,LO,GO,NO>      TP_Op;
-   RCP<TP_Crs> zeroCrsMat = rcp(new TP_Crs(*crsMat, Teuchos::Copy));
-   zeroCrsMat->setAllToScalar(0.0);
-   
-   RCP<TP_Op> Mat = crsMat;
-   RCP<TP_Op> zeroMat = zeroCrsMat;
-
-   // Allocate some right handside vectors
-   RCP<TP_Vec> x0_tp = rcp(new TP_Vec(Mat->getDomainMap()));
-   RCP<TP_Vec> x1_tp = rcp(new TP_Vec(Mat->getDomainMap()));
-   RCP<TP_Vec> b0_tp = rcp(new TP_Vec(Mat->getRangeMap()));
-   RCP<TP_Vec> b1_tp = rcp(new TP_Vec(Mat->getRangeMap()));
-   b0_tp->randomize();
-   b1_tp->randomize();
-
-   RCP<const Thyra::TpetraVectorSpace<SC,LO,GO,NO> > domain = Thyra::tpetraVectorSpace<SC>(Mat->getDomainMap());
-   RCP<const Thyra::TpetraVectorSpace<SC,LO,GO,NO> > range  = Thyra::tpetraVectorSpace<SC>(Mat->getRangeMap());
-
-   /////////////////////////////////////////////////////////
-   // Build Teko compatible matrices and vectors
-   /////////////////////////////////////////////////////////
-
-   // convert them to teko compatible sub vectors
-   Teko::MultiVector x0_th = Thyra::tpetraVector(domain, x0_tp);
-   Teko::MultiVector x1_th = Thyra::tpetraVector(domain, x1_tp);
-   Teko::MultiVector b0_th = Thyra::tpetraVector( range, b0_tp);
-   Teko::MultiVector b1_th = Thyra::tpetraVector( range, b1_tp);
-   std::vector<Teko::MultiVector> x_vec; x_vec.push_back(x0_th); x_vec.push_back(x1_th);
-   std::vector<Teko::MultiVector> b_vec; b_vec.push_back(b0_th); b_vec.push_back(b1_th);
-
-   x = Teko::buildBlockedMultiVector(x_vec); // these will be used in the Teko solve
-   b = Teko::buildBlockedMultiVector(b_vec);
-
-   // Build the Teko compatible linear system
-   Teko::LinearOp thMat = Thyra::tpetraLinearOp<double>(range,domain,Mat);
-   Teko::LinearOp thZero = Thyra::tpetraLinearOp<double>(range,domain,zeroMat);
-   A = Thyra::block2x2(thMat,thZero,thZero,thMat); // build an upper triangular 2x2
-}
+template <class SC, class LO, class GO, class NO>
+void GenerateDefault_2x2_Splitting(const RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > &crsMat,
+                                   Teko::LinearOp &A,
+                                   Teko::MultiVector &x,
+                                   Teko::MultiVector &b) {
+  typedef Tpetra::Vector<SC, LO, GO, NO> TP_Vec;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TP_Crs;
+  typedef Tpetra::Operator<SC, LO, GO, NO> TP_Op;
+  RCP<TP_Crs> zeroCrsMat = rcp(new TP_Crs(*crsMat, Teuchos::Copy));
+  zeroCrsMat->setAllToScalar(0.0);
+
+  RCP<TP_Op> Mat     = crsMat;
+  RCP<TP_Op> zeroMat = zeroCrsMat;
+
+  // Allocate some right handside vectors
+  RCP<TP_Vec> x0_tp = rcp(new TP_Vec(Mat->getDomainMap()));
+  RCP<TP_Vec> x1_tp = rcp(new TP_Vec(Mat->getDomainMap()));
+  RCP<TP_Vec> b0_tp = rcp(new TP_Vec(Mat->getRangeMap()));
+  RCP<TP_Vec> b1_tp = rcp(new TP_Vec(Mat->getRangeMap()));
+  b0_tp->randomize();
+  b1_tp->randomize();
+
+  RCP<const Thyra::TpetraVectorSpace<SC, LO, GO, NO> > domain = Thyra::tpetraVectorSpace<SC>(Mat->getDomainMap());
+  RCP<const Thyra::TpetraVectorSpace<SC, LO, GO, NO> > range  = Thyra::tpetraVectorSpace<SC>(Mat->getRangeMap());
 
+  /////////////////////////////////////////////////////////
+  // Build Teko compatible matrices and vectors
+  /////////////////////////////////////////////////////////
 
-int extract_int(std::istringstream & iss) {
+  // convert them to teko compatible sub vectors
+  Teko::MultiVector x0_th = Thyra::tpetraVector(domain, x0_tp);
+  Teko::MultiVector x1_th = Thyra::tpetraVector(domain, x1_tp);
+  Teko::MultiVector b0_th = Thyra::tpetraVector(range, b0_tp);
+  Teko::MultiVector b1_th = Thyra::tpetraVector(range, b1_tp);
+  std::vector<Teko::MultiVector> x_vec;
+  x_vec.push_back(x0_th);
+  x_vec.push_back(x1_th);
+  std::vector<Teko::MultiVector> b_vec;
+  b_vec.push_back(b0_th);
+  b_vec.push_back(b1_th);
+
+  x = Teko::buildBlockedMultiVector(x_vec);  // these will be used in the Teko solve
+  b = Teko::buildBlockedMultiVector(b_vec);
+
+  // Build the Teko compatible linear system
+  Teko::LinearOp thMat  = Thyra::tpetraLinearOp<double>(range, domain, Mat);
+  Teko::LinearOp thZero = Thyra::tpetraLinearOp<double>(range, domain, zeroMat);
+  A                     = Thyra::block2x2(thMat, thZero, thZero, thMat);  // build an upper triangular 2x2
+}
+
+int extract_int(std::istringstream &iss) {
   std::string s;
   iss >> s;
-  if(s != "") {
-    return stoi(s); }
-  else
+  if (s != "") {
+    return stoi(s);
+  } else
     return -1;
 }
 
-
 // Generates proc-by-proc block gid lists
-template<class LO, class GO, class NO>
-std::vector<std::vector<GO> > read_block_gids(std::string partitionFile, RCP<const Tpetra::Map<LO,GO,NO> > & rowMap) {
-  using V   = Tpetra::Vector<LO,LO,GO,NO>;
-  using CRS = Tpetra::CrsMatrix<LO,LO,GO,NO>;
+template <class LO, class GO, class NO>
+std::vector<std::vector<GO> > read_block_gids(std::string partitionFile, RCP<const Tpetra::Map<LO, GO, NO> > &rowMap) {
+  using V   = Tpetra::Vector<LO, LO, GO, NO>;
+  using CRS = Tpetra::CrsMatrix<LO, LO, GO, NO>;
 
-  RCP<V> pfile = Tpetra::MatrixMarket::Reader<CRS>::readVectorFile(partitionFile,rowMap->getComm(),rowMap);
+  RCP<V> pfile = Tpetra::MatrixMarket::Reader<CRS>::readVectorFile(partitionFile, rowMap->getComm(), rowMap);
 
   int num_blocks = 1 + pfile->normInf();
 
-  if(!rowMap->getComm()->getRank())
-    std::cout<<"Reading partition file: Found "<<num_blocks<<" blocks"<<std::endl;
+  if (!rowMap->getComm()->getRank())
+    std::cout << "Reading partition file: Found " << num_blocks << " blocks" << std::endl;
 
   std::vector<std::vector<GO> > block_gids(num_blocks);
   auto vv = pfile->get1dView();
-  for(LO i=0; i<(LO)vv.size(); i++){
+  for (LO i = 0; i < (LO)vv.size(); i++) {
     LO block_id = vv[i];
     block_gids[block_id].push_back(rowMap->getGlobalElement(i));
   }
@@ -172,314 +172,296 @@ std::vector<std::vector<GO> > read_block_gids(std::string partitionFile, RCP<con
   return block_gids;
 }
 
-
-
-
-template<class SC,class LO, class GO, class NO>
-void ReadSplittingFromDisk(const std::string & partitionFile,
-                           const RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> >& crsMat,
-                           RCP< Tpetra::Operator<SC,LO,GO,NO> > & A) {
+template <class SC, class LO, class GO, class NO>
+void ReadSplittingFromDisk(const std::string &partitionFile,
+                           const RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > &crsMat,
+                           RCP<Tpetra::Operator<SC, LO, GO, NO> > &A) {
   // The format for the partition file is a number of lines of the form:
   // proc_id block_id gid0, gid1,...
   //
   // Blocks do not need to be uniquely owned by a processor.
 
   //  Each rank is going to read the file, one at a time (to avoid hammering on the disk)
-  auto comm = crsMat->getRowMap()->getComm();
-  RCP<const Tpetra::Map<LO,GO,NO> > rowmap = crsMat->getRowMap();
-  std::vector<std::vector<GO> > my_blocks_and_gids = read_block_gids<LO,GO,NO>(partitionFile,rowmap);
+  auto comm                                        = crsMat->getRowMap()->getComm();
+  RCP<const Tpetra::Map<LO, GO, NO> > rowmap       = crsMat->getRowMap();
+  std::vector<std::vector<GO> > my_blocks_and_gids = read_block_gids<LO, GO, NO>(partitionFile, rowmap);
 
-  RCP<Teko::TpetraHelpers::BlockedTpetraOperator> rA = rcp( new Teko::TpetraHelpers::BlockedTpetraOperator(my_blocks_and_gids,crsMat));
+  RCP<Teko::TpetraHelpers::BlockedTpetraOperator> rA = rcp(new Teko::TpetraHelpers::BlockedTpetraOperator(my_blocks_and_gids, crsMat));
 
   A = rA;
 }
 
+template <class SC, class LO, class GO, class NO>
+int solve_thyra(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > &crsMat, const std::string &xmlFile) {
+  typedef Tpetra::CrsMatrix<SC> TP_Crs;
 
-template<class SC,class LO, class GO, class NO>
-  int solve_thyra(RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > & crsMat, const std::string &xmlFile) {
-  typedef Tpetra::CrsMatrix<SC>       TP_Crs;
-
-  typedef Thyra::MultiVectorBase<SC>  MV;
-  typedef Thyra::LinearOpBase<SC>     OP;
+  typedef Thyra::MultiVectorBase<SC> MV;
+  typedef Thyra::LinearOpBase<SC> OP;
   auto comm = crsMat->getRowMap()->getComm();
 
   RCP<Stratimikos::DefaultLinearSolverBuilder> linearSolverBuilder = Teuchos::rcp(new Stratimikos::DefaultLinearSolverBuilder);
- 
+
   /////////////////////////////////////////////////////////
   // Build the Thyra operators
   /////////////////////////////////////////////////////////
-   Teko::LinearOp A;
-   Teko::MultiVector x,b;
-   GenerateDefault_2x2_Splitting<SC,LO,GO,NO>(crsMat,A,x,b);
-
-
-   /////////////////////////////////////////////////////////
-   // Build the preconditioner
-   /////////////////////////////////////////////////////////
-
-   // build an InverseLibrary
-   RCP<Teko::InverseLibrary> invLib;
-   RCP<Teko::InverseFactory> inverse;
-   if(xmlFile == "") {
-     invLib = Teko::InverseLibrary::buildFromParameterList(*buildLibPL(),linearSolverBuilder);
-     inverse = invLib->getInverseFactory("Gauss-Seidel");
-   }
-   else {
-     Teuchos::ParameterList xmlList;
-     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFile, Teuchos::Ptr<Teuchos::ParameterList>(&xmlList),*comm);
-     invLib = Teko::InverseLibrary::buildFromParameterList(xmlList,linearSolverBuilder);
-     inverse = invLib->getInverseFactory("MyTekoPreconditioner");
-   }
-   // build the inverse factory needed by the example preconditioner
-
-
-   // build the preconditioner from the jacobian
-   Teko::LinearOp prec = Teko::buildInverse(*inverse,A);
-
-   // Setup the Belos solver
-   /////////////////////////////////////////////////////////
-
-   Teuchos::ParameterList belosList;
-   belosList.set( "Num Blocks", 200 );               // Maximum number of blocks in Krylov factorization
-   belosList.set( "Block Size",1 );                  // Blocksize to be used by iterative solver
-   belosList.set( "Maximum Iterations", 200 );       // Maximum number of iterations allowed
-   belosList.set( "Maximum Restarts", 1 );           // Maximum number of restarts allowed
-   belosList.set( "Convergence Tolerance", 1e-5 );    // Relative convergence tolerance requested
-   belosList.set( "Verbosity", 33);//Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails );
-   belosList.set( "Output Frequency", 1 );
-   belosList.set( "Output Style", 1 );
-
-   RCP<Belos::LinearProblem<double,MV,OP> > problem = rcp(new Belos::LinearProblem<double,MV,OP>( A, x, b ) );
-   problem->setLeftPrec(prec);
-   problem->setProblem(); // should check the return type!!!
-
-   RCP<Belos::SolverManager<double,MV,OP> > solver
-    = rcp(new Belos::BlockGmresSolMgr<double,MV,OP>(problem, rcpFromRef(belosList)));
-
-   //
-   // Perform solve
-   //
-   Belos::ReturnType ret = solver->solve();
-
-   if (ret!=Belos::Converged) {
-     std::cout << std::endl << "ERROR:  Belos did not converge!" << std::endl;
-     return -1;
-   }
-
-   return 0;
-}
+  Teko::LinearOp A;
+  Teko::MultiVector x, b;
+  GenerateDefault_2x2_Splitting<SC, LO, GO, NO>(crsMat, A, x, b);
+
+  /////////////////////////////////////////////////////////
+  // Build the preconditioner
+  /////////////////////////////////////////////////////////
+
+  // build an InverseLibrary
+  RCP<Teko::InverseLibrary> invLib;
+  RCP<Teko::InverseFactory> inverse;
+  if (xmlFile == "") {
+    invLib  = Teko::InverseLibrary::buildFromParameterList(*buildLibPL(), linearSolverBuilder);
+    inverse = invLib->getInverseFactory("Gauss-Seidel");
+  } else {
+    Teuchos::ParameterList xmlList;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFile, Teuchos::Ptr<Teuchos::ParameterList>(&xmlList), *comm);
+    invLib  = Teko::InverseLibrary::buildFromParameterList(xmlList, linearSolverBuilder);
+    inverse = invLib->getInverseFactory("MyTekoPreconditioner");
+  }
+  // build the inverse factory needed by the example preconditioner
 
+  // build the preconditioner from the jacobian
+  Teko::LinearOp prec = Teko::buildInverse(*inverse, A);
+
+  // Setup the Belos solver
+  /////////////////////////////////////////////////////////
 
+  Teuchos::ParameterList belosList;
+  belosList.set("Num Blocks", 200);              // Maximum number of blocks in Krylov factorization
+  belosList.set("Block Size", 1);                // Blocksize to be used by iterative solver
+  belosList.set("Maximum Iterations", 200);      // Maximum number of iterations allowed
+  belosList.set("Maximum Restarts", 1);          // Maximum number of restarts allowed
+  belosList.set("Convergence Tolerance", 1e-5);  // Relative convergence tolerance requested
+  belosList.set("Verbosity", 33);                // Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails );
+  belosList.set("Output Frequency", 1);
+  belosList.set("Output Style", 1);
 
+  RCP<Belos::LinearProblem<double, MV, OP> > problem = rcp(new Belos::LinearProblem<double, MV, OP>(A, x, b));
+  problem->setLeftPrec(prec);
+  problem->setProblem();  // should check the return type!!!
 
-template<class SC,class LO, class GO, class NO>
-  int solve_tpetra(RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > & crsMat, RCP<Tpetra::Vector<SC,LO,GO,NO> > & b,  RCP<Tpetra::MultiVector<SC,LO,GO,NO> > & coords, const std::string &xmlFile,const std::string &partitionFile) {
-  //typedef Tpetra::Map<>               TP_Map;
-  //typedef Tpetra::Vector<SC>          TP_Vec;
-  typedef Tpetra::CrsMatrix<SC>       TP_Crs;
-  typedef Thyra::PreconditionerFactoryBase<SC>        Base;
+  RCP<Belos::SolverManager<double, MV, OP> > solver = rcp(new Belos::BlockGmresSolMgr<double, MV, OP>(problem, rcpFromRef(belosList)));
 
-  typedef Thyra::MultiVectorBase<SC>  MV;
-  typedef Thyra::LinearOpBase<SC>     OP;
+  //
+  // Perform solve
+  //
+  Belos::ReturnType ret = solver->solve();
+
+  if (ret != Belos::Converged) {
+    std::cout << std::endl
+              << "ERROR:  Belos did not converge!" << std::endl;
+    return -1;
+  }
+
+  return 0;
+}
+
+template <class SC, class LO, class GO, class NO>
+int solve_tpetra(RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > &crsMat, RCP<Tpetra::Vector<SC, LO, GO, NO> > &b, RCP<Tpetra::MultiVector<SC, LO, GO, NO> > &coords, const std::string &xmlFile, const std::string &partitionFile) {
+  // typedef Tpetra::Map<>               TP_Map;
+  // typedef Tpetra::Vector<SC>          TP_Vec;
+  typedef Tpetra::CrsMatrix<SC> TP_Crs;
+  typedef Thyra::PreconditionerFactoryBase<SC> Base;
+
+  typedef Thyra::MultiVectorBase<SC> MV;
+  typedef Thyra::LinearOpBase<SC> OP;
   //  typedef Tpetra::Vector<SC,LO,GO,NO>  MV;
-//    typedef Tpetra::Operator<SC,LO,GO,NO>     OP;
+  //    typedef Tpetra::Operator<SC,LO,GO,NO>     OP;
   auto comm = crsMat->getRowMap()->getComm();
 
   // tell Stratimikos => Teko about MueLu
   RCP<Stratimikos::DefaultLinearSolverBuilder> linearSolverBuilder = Teuchos::rcp(new Stratimikos::DefaultLinearSolverBuilder);
-  Stratimikos::enableMueLu<SC,LO,GO,NO>(*linearSolverBuilder);
- 
+  Stratimikos::enableMueLu<SC, LO, GO, NO>(*linearSolverBuilder);
+
   /////////////////////////////////////////////////////////
   // Build the Thyra operators
   /////////////////////////////////////////////////////////
-  RCP<Tpetra::Operator<SC,LO,GO,NO> > A;
-  RCP<Tpetra::Vector<SC,LO,GO,NO> > x = rcp(new Tpetra::Vector<SC,LO,GO,NO>(b->getMap()));
-  ReadSplittingFromDisk<SC,LO,GO,NO>(partitionFile,crsMat,A);
+  RCP<Tpetra::Operator<SC, LO, GO, NO> > A;
+  RCP<Tpetra::Vector<SC, LO, GO, NO> > x = rcp(new Tpetra::Vector<SC, LO, GO, NO>(b->getMap()));
+  ReadSplittingFromDisk<SC, LO, GO, NO>(partitionFile, crsMat, A);
   x->putScalar(Teuchos::ScalarTraits<SC>::zero());
- 
+
   RCP<Thyra::MultiVectorBase<SC> > xt = Thyra::createVector(x);
   RCP<Thyra::MultiVectorBase<SC> > bt = Thyra::createVector(b);
 
+  /////////////////////////////////////////////////////////
+  // Build the preconditioner
+  /////////////////////////////////////////////////////////
+
+  // build an InverseLibrary
+  RCP<Teko::InverseLibrary> invLib;
+  RCP<Teko::InverseFactory> inverse;
+  if (xmlFile == "") {
+    invLib  = Teko::InverseLibrary::buildFromParameterList(*buildLibPL(), linearSolverBuilder);
+    inverse = invLib->getInverseFactory("Gauss-Seidel");
+  } else {
+    Teuchos::ParameterList xmlList;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFile, Teuchos::Ptr<Teuchos::ParameterList>(&xmlList), *comm);
+
+    // Add coordinates if we need to
+    if (!coords.is_null()) {
+      // FIXME:  Please do this more generally
+      auto &sublist = xmlList.sublist("MueluScalar").sublist("user data");
+      sublist.set("Coordinates", Xpetra::toXpetra(coords));
+    }
+
+    invLib  = Teko::InverseLibrary::buildFromParameterList(xmlList, linearSolverBuilder);
+    inverse = invLib->getInverseFactory("MyTekoPreconditioner");
+  }
 
-   /////////////////////////////////////////////////////////
-   // Build the preconditioner
-   /////////////////////////////////////////////////////////
-
-   // build an InverseLibrary
-   RCP<Teko::InverseLibrary> invLib;
-   RCP<Teko::InverseFactory> inverse;
-   if(xmlFile == "") {
-     invLib = Teko::InverseLibrary::buildFromParameterList(*buildLibPL(),linearSolverBuilder);
-     inverse = invLib->getInverseFactory("Gauss-Seidel");
-   }
-   else {
-     Teuchos::ParameterList xmlList;
-     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFile, Teuchos::Ptr<Teuchos::ParameterList>(&xmlList),*comm);
-
-     // Add coordinates if we need to
-     if(!coords.is_null()){
-       // FIXME:  Please do this more generally
-       auto & sublist = xmlList.sublist("MueluScalar").sublist("user data");
-       sublist.set("Coordinates",Xpetra::toXpetra(coords));
-     }
-
-     invLib = Teko::InverseLibrary::buildFromParameterList(xmlList,linearSolverBuilder);
-     inverse = invLib->getInverseFactory("MyTekoPreconditioner");
-   }
-      
-   // build the inverse factory needed by the example preconditioner
-   Teko::TpetraHelpers::InverseFactoryOperator ifo(inverse);
-   ifo.buildInverseOperator(A);
-
-
-   // build the preconditioner from the jacobian
-   Teko::LinearOp At = Thyra::tpetraLinearOp<SC,LO,GO,NO>(Thyra::tpetraVectorSpace<SC,LO,GO,NO>(A->getRangeMap()),Thyra::tpetraVectorSpace<SC,LO,GO,NO>(A->getDomainMap()),A);
-
-
-   Teko::LinearOp prec = Teko::buildInverse(*inverse,At);
-
-   // Setup the Belos solver
-   /////////////////////////////////////////////////////////
-
-   Teuchos::ParameterList belosList;
-   belosList.set( "Num Blocks", 200 );               // Maximum number of blocks in Krylov factorization
-   belosList.set( "Block Size",1 );                  // Blocksize to be used by iterative solver
-   belosList.set( "Maximum Iterations", 200 );       // Maximum number of iterations allowed
-   belosList.set( "Maximum Restarts", 1 );           // Maximum number of restarts allowed
-   belosList.set( "Convergence Tolerance", 1e-5 );    // Relative convergence tolerance requested
-   belosList.set( "Verbosity", 33);//Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails );
-   belosList.set( "Output Frequency", 1 );
-   belosList.set( "Output Style", 1 );
-
-   RCP<Belos::LinearProblem<double,MV,OP> > problem = rcp(new Belos::LinearProblem<double,MV,OP>( At, xt, bt ) );
-   problem->setLeftPrec(prec);
-   problem->setProblem(); // should check the return type!!!
-
-   RCP<Belos::SolverManager<double,MV,OP> > solver
-    = rcp(new Belos::BlockGmresSolMgr<double,MV,OP>(problem, rcpFromRef(belosList)));
-
-   //
-   // Perform solve
-   //
-   Belos::ReturnType ret = solver->solve();
-
-   if (ret!=Belos::Converged) {
-     std::cout << std::endl << "ERROR:  Belos did not converge!" << std::endl;
-     return -1;
-   }
-
-   return 0;
-}
+  // build the inverse factory needed by the example preconditioner
+  Teko::TpetraHelpers::InverseFactoryOperator ifo(inverse);
+  ifo.buildInverseOperator(A);
 
+  // build the preconditioner from the jacobian
+  Teko::LinearOp At = Thyra::tpetraLinearOp<SC, LO, GO, NO>(Thyra::tpetraVectorSpace<SC, LO, GO, NO>(A->getRangeMap()), Thyra::tpetraVectorSpace<SC, LO, GO, NO>(A->getDomainMap()), A);
 
-int main(int argc,char * argv[])
-{
-   typedef double                      SC;
-
-   typedef Tpetra::Map<>               TP_Map;
-   typedef Tpetra::Vector<SC>          TP_Vec;
-   typedef Tpetra::MultiVector<SC>     TP_MV;
-   typedef Tpetra::CrsMatrix<SC>       TP_Crs;
-   //typedef Tpetra::Operator<SC>        TP_Op;
-
-   typedef TP_Vec::local_ordinal_type  LO;
-   typedef TP_Vec::global_ordinal_type GO;
-   typedef TP_Vec::node_type           NO;
-
-
-   // calls MPI_Init and MPI_Finalize
-   Teuchos::GlobalMPISession mpiSession(&argc,&argv);
-
-   // Parse CLI options
-   Teuchos::CommandLineProcessor clp(false);
-   std::string rhsFile;                                clp.setOption("rhs",                   &rhsFile,           "rhs data file");
-  
-   std::string rowMapFile;                             clp.setOption("rowmap",                &rowMapFile,        "map data file");
-   std::string xmlFile           = "";                 clp.setOption("xml",                   &xmlFile,           "read Tekko parameters from an xml file");
-   std::string matrixFile = "../data/nsjac.mm";        clp.setOption("matrix",                &matrixFile,        "matrix data file");
-
-   std::string partitionFile = "";                     clp.setOption("partition",             &partitionFile,      "partition file which defines the blocks");
-
- 
-   std::string coordsFile;                             clp.setOption("coords",                &coordsFile,         "coords data file");
-
-   std::string coordsMapFile;                          clp.setOption("coordsmap",             &coordsMapFile,      "coords data file");
-
-   clp.recogniseAllOptions(true);
-   switch (clp.parse(argc, argv)) {
-   case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
-   case Teuchos::CommandLineProcessor::PARSE_ERROR:
-   case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-   case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
-   }
-   
-   /////////////////////////////////////////////////////////
-   // Build the Tpetra matrices and vectors
-   /////////////////////////////////////////////////////////
-
-   // read in the CRS matrix
-   RCP<TP_Crs> crsMat;
-   RCP<TP_Vec> rhs;
-   RCP<TP_MV> coords;
-   if(rowMapFile == "") {
-     crsMat = Tpetra::MatrixMarket::Reader<TP_Crs>::readSparseFile(matrixFile, Tpetra::getDefaultComm());
-   }
-   else {
-     RCP<const TP_Map> rowMap = Tpetra::MatrixMarket::Reader<TP_Crs>::readMapFile(rowMapFile,Tpetra::getDefaultComm());
-     RCP<const TP_Map> colMap;
-     crsMat = Tpetra::MatrixMarket::Reader<TP_Crs>::readSparseFile(matrixFile,rowMap,colMap,rowMap,rowMap);
-
-     if(rhsFile!= "")  {
-       rhs = Tpetra::MatrixMarket::Reader<TP_Crs>::readVectorFile(rhsFile,rowMap->getComm(),rowMap);
-     }
-
-     if(coordsFile!= "") {
-       RCP<const TP_Map> coordsMap = rowMap;
-       if(coordsMapFile != "")
-         coordsMap = Tpetra::MatrixMarket::Reader<TP_Crs>::readMapFile(coordsMapFile,rowMap->getComm());
-       coords = Tpetra::MatrixMarket::Reader<TP_Crs>::readDenseFile(coordsFile,coordsMap->getComm(),coordsMap);
-     }
-   }
-
-   // Sanity checks
-   if(rhs.is_null())    throw std::runtime_error("rhs is null");
-   if(crsMat.is_null()) throw std::runtime_error("crsMat is null");
-
-
-   auto comm = crsMat->getRowMap()->getComm();
-
-
-   /////////////////////////////////////////////////////////
-   // Build the Thyra operators
-   /////////////////////////////////////////////////////////
-   Teko::LinearOp A;
-   Teko::MultiVector x,b;
-
-   int rv;
-   if(partitionFile == "")
-     rv = solve_thyra(crsMat,xmlFile);
-   else
-     rv = solve_tpetra<SC,LO,GO,NO>(crsMat,rhs,coords,xmlFile,partitionFile);
-   
-   return rv;
+  Teko::LinearOp prec = Teko::buildInverse(*inverse, At);
 
+  // Setup the Belos solver
+  /////////////////////////////////////////////////////////
 
+  Teuchos::ParameterList belosList;
+  belosList.set("Num Blocks", 200);              // Maximum number of blocks in Krylov factorization
+  belosList.set("Block Size", 1);                // Blocksize to be used by iterative solver
+  belosList.set("Maximum Iterations", 200);      // Maximum number of iterations allowed
+  belosList.set("Maximum Restarts", 1);          // Maximum number of restarts allowed
+  belosList.set("Convergence Tolerance", 1e-5);  // Relative convergence tolerance requested
+  belosList.set("Verbosity", 33);                // Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails );
+  belosList.set("Output Frequency", 1);
+  belosList.set("Output Style", 1);
+
+  RCP<Belos::LinearProblem<double, MV, OP> > problem = rcp(new Belos::LinearProblem<double, MV, OP>(At, xt, bt));
+  problem->setLeftPrec(prec);
+  problem->setProblem();  // should check the return type!!!
+
+  RCP<Belos::SolverManager<double, MV, OP> > solver = rcp(new Belos::BlockGmresSolMgr<double, MV, OP>(problem, rcpFromRef(belosList)));
+
+  //
+  // Perform solve
+  //
+  Belos::ReturnType ret = solver->solve();
+
+  if (ret != Belos::Converged) {
+    std::cout << std::endl
+              << "ERROR:  Belos did not converge!" << std::endl;
+    return -1;
+  }
+
+  return 0;
 }
 
-RCP<Teuchos::ParameterList> buildLibPL()
-{
-   RCP<Teuchos::ParameterList> pl = rcp(new Teuchos::ParameterList());
-
-   {
-      Teuchos::ParameterList & sub_jac = pl->sublist("Jacobi");
-      sub_jac.set("Type","Block Jacobi");
-      sub_jac.set("Inverse Type","Ifpack2");
-
-      Teuchos::ParameterList & sub_gs = pl->sublist("Gauss-Seidel");
-      sub_gs.set("Type","Block Gauss-Seidel");
-      sub_gs.set("Use Upper Triangle",true);
-      sub_gs.set("Inverse Type","Ifpack2");
-   }
-   return pl;
+int main(int argc, char *argv[]) {
+  typedef double SC;
+
+  typedef Tpetra::Map<> TP_Map;
+  typedef Tpetra::Vector<SC> TP_Vec;
+  typedef Tpetra::MultiVector<SC> TP_MV;
+  typedef Tpetra::CrsMatrix<SC> TP_Crs;
+  // typedef Tpetra::Operator<SC>        TP_Op;
+
+  typedef TP_Vec::local_ordinal_type LO;
+  typedef TP_Vec::global_ordinal_type GO;
+  typedef TP_Vec::node_type NO;
+
+  // calls MPI_Init and MPI_Finalize
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
+
+  // Parse CLI options
+  Teuchos::CommandLineProcessor clp(false);
+  std::string rhsFile;
+  clp.setOption("rhs", &rhsFile, "rhs data file");
+
+  std::string rowMapFile;
+  clp.setOption("rowmap", &rowMapFile, "map data file");
+  std::string xmlFile = "";
+  clp.setOption("xml", &xmlFile, "read Tekko parameters from an xml file");
+  std::string matrixFile = "../data/nsjac.mm";
+  clp.setOption("matrix", &matrixFile, "matrix data file");
+
+  std::string partitionFile = "";
+  clp.setOption("partition", &partitionFile, "partition file which defines the blocks");
+
+  std::string coordsFile;
+  clp.setOption("coords", &coordsFile, "coords data file");
+
+  std::string coordsMapFile;
+  clp.setOption("coordsmap", &coordsMapFile, "coords data file");
+
+  clp.recogniseAllOptions(true);
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_ERROR:
+    case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
+  }
+
+  /////////////////////////////////////////////////////////
+  // Build the Tpetra matrices and vectors
+  /////////////////////////////////////////////////////////
+
+  // read in the CRS matrix
+  RCP<TP_Crs> crsMat;
+  RCP<TP_Vec> rhs;
+  RCP<TP_MV> coords;
+  if (rowMapFile == "") {
+    crsMat = Tpetra::MatrixMarket::Reader<TP_Crs>::readSparseFile(matrixFile, Tpetra::getDefaultComm());
+  } else {
+    RCP<const TP_Map> rowMap = Tpetra::MatrixMarket::Reader<TP_Crs>::readMapFile(rowMapFile, Tpetra::getDefaultComm());
+    RCP<const TP_Map> colMap;
+    crsMat = Tpetra::MatrixMarket::Reader<TP_Crs>::readSparseFile(matrixFile, rowMap, colMap, rowMap, rowMap);
+
+    if (rhsFile != "") {
+      rhs = Tpetra::MatrixMarket::Reader<TP_Crs>::readVectorFile(rhsFile, rowMap->getComm(), rowMap);
+    }
+
+    if (coordsFile != "") {
+      RCP<const TP_Map> coordsMap = rowMap;
+      if (coordsMapFile != "")
+        coordsMap = Tpetra::MatrixMarket::Reader<TP_Crs>::readMapFile(coordsMapFile, rowMap->getComm());
+      coords = Tpetra::MatrixMarket::Reader<TP_Crs>::readDenseFile(coordsFile, coordsMap->getComm(), coordsMap);
+    }
+  }
+
+  // Sanity checks
+  if (rhs.is_null()) throw std::runtime_error("rhs is null");
+  if (crsMat.is_null()) throw std::runtime_error("crsMat is null");
+
+  auto comm = crsMat->getRowMap()->getComm();
+
+  /////////////////////////////////////////////////////////
+  // Build the Thyra operators
+  /////////////////////////////////////////////////////////
+  Teko::LinearOp A;
+  Teko::MultiVector x, b;
+
+  int rv;
+  if (partitionFile == "")
+    rv = solve_thyra(crsMat, xmlFile);
+  else
+    rv = solve_tpetra<SC, LO, GO, NO>(crsMat, rhs, coords, xmlFile, partitionFile);
+
+  return rv;
 }
 
+RCP<Teuchos::ParameterList> buildLibPL() {
+  RCP<Teuchos::ParameterList> pl = rcp(new Teuchos::ParameterList());
+
+  {
+    Teuchos::ParameterList &sub_jac = pl->sublist("Jacobi");
+    sub_jac.set("Type", "Block Jacobi");
+    sub_jac.set("Inverse Type", "Ifpack2");
+
+    Teuchos::ParameterList &sub_gs = pl->sublist("Gauss-Seidel");
+    sub_gs.set("Type", "Block Gauss-Seidel");
+    sub_gs.set("Use Upper Triangle", true);
+    sub_gs.set("Inverse Type", "Ifpack2");
+  }
+  return pl;
+}
diff --git a/packages/muelu/matlab/bin/muemex.cpp b/packages/muelu/matlab/bin/muemex.cpp
index e167c16d6201..b75100351ecc 100644
--- a/packages/muelu/matlab/bin/muemex.cpp
+++ b/packages/muelu/matlab/bin/muemex.cpp
@@ -50,7 +50,7 @@
 #define IS_TRUE 1
 #define MUEMEX_ERROR -1
 
-//Do not compile MueMex if any of these aren't available
+// Do not compile MueMex if any of these aren't available
 #if !defined HAVE_MUELU_EPETRA || !defined HAVE_MUELU_MATLAB
 #error "MueMex requires Epetra, Tpetra and MATLAB."
 #endif
@@ -77,142 +77,134 @@ extern void _main();
 #define MUEMEX_DEFAULT_NUMPDES 1
 #define MUEMEX_DEFAULT_ADAPTIVEVECS 0
 #define MUEMEX_DEFAULT_USEDEFAULTNS true
-#define MMABS(x)   ((x)>0?(x):(-(x)))
-#define MMISINT(x) ((x)==0?(((x-(int)(x))<1e-15)?true:false):(((x-(int)(x))<1e-15*MMABS(x))?true:false))
+#define MMABS(x) ((x) > 0 ? (x) : (-(x)))
+#define MMISINT(x) ((x) == 0 ? (((x - (int)(x)) < 1e-15) ? true : false) : (((x - (int)(x)) < 1e-15 * MMABS(x)) ? true : false))
 
 /* Debugging */
 //#define VERBOSE_OUTPUT
 
 namespace MueLu {
 
-//Need subclass of Hierarchy that gives public access to list of FactoryManagers
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-class OpenHierarchy : public Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>
-{
-  public:
-    const RCP<const FactoryManagerBase>& GetFactoryManager(const int levelID) const;
+// Need subclass of Hierarchy that gives public access to list of FactoryManagers
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+class OpenHierarchy : public Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  const RCP<const FactoryManagerBase>& GetFactoryManager(const int levelID) const;
 };
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-const RCP<const FactoryManagerBase>& OpenHierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactoryManager(const int levelID) const
-{
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+const RCP<const FactoryManagerBase>& OpenHierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactoryManager(const int levelID) const {
   TEUCHOS_TEST_FOR_EXCEPTION(levelID < 0 || levelID > this->GetNumLevels(), Exceptions::RuntimeError,
-                            "MueLu::Hierarchy::GetFactoryManager(): invalid input parameter value: LevelID = " << levelID);
+                             "MueLu::Hierarchy::GetFactoryManager(): invalid input parameter value: LevelID = " << levelID);
   return this->GetLevelManager(levelID);
 }
 
-//Declare and call default constructor for data_pack_list vector (starts empty)
+// Declare and call default constructor for data_pack_list vector (starts empty)
 vector<RCP<MuemexSystem>> MuemexSystemList::list;
 int MuemexSystemList::nextID = 0;
 
-//Need a global flag to keep track of Epetra vs. Tpetra for constructing multivectors for param lists
+// Need a global flag to keep track of Epetra vs. Tpetra for constructing multivectors for param lists
 bool useEpetra = false;
 
-//Parse a string to get each bit of Belos verbosity
-int strToMsgType(const char* str)
-{
-  if(str == NULL)
+// Parse a string to get each bit of Belos verbosity
+int strToMsgType(const char* str) {
+  if (str == NULL)
     return Belos::Errors;
-  else if(strstr(str, "Warnings") != NULL)
+  else if (strstr(str, "Warnings") != NULL)
     return Belos::Warnings;
-  else if(strstr(str, "IterationDetails") != NULL)
+  else if (strstr(str, "IterationDetails") != NULL)
     return Belos::IterationDetails;
-  else if(strstr(str, "OrthoDetails") != NULL)
+  else if (strstr(str, "OrthoDetails") != NULL)
     return Belos::OrthoDetails;
-  else if(strstr(str, "FinalSummary") != NULL)
+  else if (strstr(str, "FinalSummary") != NULL)
     return Belos::FinalSummary;
-  else if(strstr(str, "TimingDetails") != NULL)
+  else if (strstr(str, "TimingDetails") != NULL)
     return Belos::TimingDetails;
-  else if(strstr(str, "StatusTestDetails") != NULL)
+  else if (strstr(str, "StatusTestDetails") != NULL)
     return Belos::StatusTestDetails;
-  else if(strstr(str, "Debug") != NULL)
+  else if (strstr(str, "Debug") != NULL)
     return Belos::Debug;
-  //This has no effect when added/OR'd with flags
+  // This has no effect when added/OR'd with flags
   return Belos::Errors;
 }
 
-MuemexType strToDataType(const char* str, char* typeName, bool complexFlag = false)
-{
+MuemexType strToDataType(const char* str, char* typeName, bool complexFlag = false) {
   std::string temp(str);
   std::string myStr = trim(temp);
   MuemexType matrixType, multivectorType, scalarType;
-  if(!complexFlag)
-  {
-    matrixType = XPETRA_MATRIX_DOUBLE;
+  if (!complexFlag) {
+    matrixType      = XPETRA_MATRIX_DOUBLE;
     multivectorType = XPETRA_MULTIVECTOR_DOUBLE;
-    scalarType = DOUBLE;
-  }
-  else
-  {
-    matrixType = XPETRA_MATRIX_COMPLEX;
+    scalarType      = DOUBLE;
+  } else {
+    matrixType      = XPETRA_MATRIX_COMPLEX;
     multivectorType = XPETRA_MULTIVECTOR_COMPLEX;
-    scalarType = COMPLEX;
+    scalarType      = COMPLEX;
   }
   size_t npos = string::npos;
-  if(myStr == "A" ||
-     myStr == "P" || myStr == "Ptent" ||
-     myStr == "R")
+  if (myStr == "A" ||
+      myStr == "P" || myStr == "Ptent" ||
+      myStr == "R")
     return matrixType;
-  if(myStr == "Nullspace")
+  if (myStr == "Nullspace")
     return multivectorType;
-  if(myStr == "Aggregates")
+  if (myStr == "Aggregates")
     return AGGREGATES;
-  if(myStr == "Graph")
+  if (myStr == "Graph")
     return GRAPH;
-  if(myStr == "Coordinates")
+  if (myStr == "Coordinates")
     return XPETRA_MULTIVECTOR_DOUBLE;
 #ifdef HAVE_MUELU_INTREPID2
-  if(myStr == "pcoarsen: element to node map")
+  if (myStr == "pcoarsen: element to node map")
     return FIELDCONTAINER_ORDINAL;
 #endif
-  //Check for custom variable
+  // Check for custom variable
   size_t firstWordStart = myStr.find_first_not_of(' ');
   size_t firstWordEnd   = myStr.find(' ', firstWordStart);
   std::string firstWord = myStr.substr(firstWordStart, firstWordEnd - firstWordStart);
-  if(firstWord.length() > 0) {
-    temp = myStr.substr(firstWordEnd,myStr.length()-firstWordEnd);
+  if (firstWord.length() > 0) {
+    temp                   = myStr.substr(firstWordEnd, myStr.length() - firstWordEnd);
     std::string secondWord = trim(temp);
-    if(secondWord.length() > 0) {
-      //make first word lowercase
+    if (secondWord.length() > 0) {
+      // make first word lowercase
       std::transform(firstWord.begin(), firstWord.end(), firstWord.begin(), ::tolower);
-      //compare first word with possible values
-      if(firstWord.find("matrix") != npos)
+      // compare first word with possible values
+      if (firstWord.find("matrix") != npos)
         return matrixType;
-      if(firstWord.find("multivector") != npos)
+      if (firstWord.find("multivector") != npos)
         return multivectorType;
-      if(firstWord.find("map") != npos)
+      if (firstWord.find("map") != npos)
         return XPETRA_MAP;
-      if(firstWord.find("ordinalvector") != npos)
+      if (firstWord.find("ordinalvector") != npos)
         return XPETRA_ORDINAL_VECTOR;
-      if(firstWord.find("int") != npos)
+      if (firstWord.find("int") != npos)
         return INT;
-      if(firstWord.find("scalar") != npos)
+      if (firstWord.find("scalar") != npos)
         return scalarType;
-      if(firstWord.find("double") != npos)
+      if (firstWord.find("double") != npos)
         return DOUBLE;
-      if(firstWord.find("complex") != npos)
+      if (firstWord.find("complex") != npos)
         return COMPLEX;
     }
   }
-  if(typeName)
-  {
+  if (typeName) {
     std::string typeString(typeName);
     std::transform(typeString.begin(), typeString.end(), typeString.begin(), ::tolower);
-    if(typeString.find("matrix") != npos)
+    if (typeString.find("matrix") != npos)
       return matrixType;
-    if(typeString.find("multivector") != npos)
+    if (typeString.find("multivector") != npos)
       return multivectorType;
-    if(typeString.find("map") != npos)
+    if (typeString.find("map") != npos)
       return XPETRA_MAP;
-    if(typeString.find("ordinalvector") != npos)
+    if (typeString.find("ordinalvector") != npos)
       return XPETRA_ORDINAL_VECTOR;
-    if(typeString.find("int") != npos)
+    if (typeString.find("int") != npos)
       return INT;
-    if(typeString.find("scalar") != npos)
+    if (typeString.find("scalar") != npos)
       return scalarType;
-    if(typeString.find("double") != npos)
+    if (typeString.find("double") != npos)
       return DOUBLE;
-    if(typeString.find("complex") != npos)
+    if (typeString.find("complex") != npos)
       return COMPLEX;
     string errMsg = typeString + " is not a valid type.";
     throw runtime_error(errMsg);
@@ -220,53 +212,48 @@ MuemexType strToDataType(const char* str, char* typeName, bool complexFlag = fal
   throw runtime_error("Could not determine type of data.");
 }
 
-//Parse a string to get Belos output style (Brief is default)
-int strToOutputStyle(const char* str)
-{
-  if(strstr("General", str) != NULL)
+// Parse a string to get Belos output style (Brief is default)
+int strToOutputStyle(const char* str) {
+  if (strstr("General", str) != NULL)
     return Belos::General;
   else
     return Belos::Brief;
 }
 
-//Get Belos "Verbosity" setting for its ParameterList
-int getBelosVerbosity(const char* input)
-{
+// Get Belos "Verbosity" setting for its ParameterList
+int getBelosVerbosity(const char* input) {
   int result = 0;
-  char* str = (char*) input;
+  char* str  = (char*)input;
   char* pch;
   pch = strtok(str, " +,");
-  if(pch == NULL)
+  if (pch == NULL)
     return result;
   result |= strToMsgType(pch);
-  while(pch != NULL)
-    {
-      pch = strtok(NULL, " +,");
-      if(pch == NULL)
-        return result;
-      result |= strToMsgType(pch);
-    }
+  while (pch != NULL) {
+    pch = strtok(NULL, " +,");
+    if (pch == NULL)
+      return result;
+    result |= strToMsgType(pch);
+  }
   return result;
 }
 
-template<typename Scalar>
-mxArray* TpetraSystem<Scalar>::solve(RCP<ParameterList> params, RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> matrix, const mxArray* b, int& iters)
-{
+template <typename Scalar>
+mxArray* TpetraSystem<Scalar>::solve(RCP<ParameterList> params, RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> matrix, const mxArray* b, int& iters) {
   mxArray* output;
-  try
-  {
+  try {
     int matSize = A->getGlobalNumRows();
-    //Define Tpetra vector/multivector types for convenience
+    // Define Tpetra vector/multivector types for convenience
     typedef Tpetra::Vector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Tpetra_Vector;
     typedef Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Tpetra_MultiVector;
     typedef Tpetra::Operator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Tpetra_Operator;
     RCP<const Teuchos::Comm<int>> comm = Tpetra::getDefaultComm();
-    //numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
-    RCP<const muemex_map_type> map = rcp(new muemex_map_type(matSize, (mm_GlobalOrd) 0, comm));
-    RCP<Tpetra_MultiVector> rhs = loadDataFromMatlab<RCP<Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(b);
-    RCP<Tpetra_MultiVector> lhs = rcp(new Tpetra_MultiVector(map, rhs->getNumVectors()));
-    //rhs is initialized, lhs is not
-    // Default params
+    // numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
+    RCP<const muemex_map_type> map = rcp(new muemex_map_type(matSize, (mm_GlobalOrd)0, comm));
+    RCP<Tpetra_MultiVector> rhs    = loadDataFromMatlab<RCP<Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(b);
+    RCP<Tpetra_MultiVector> lhs    = rcp(new Tpetra_MultiVector(map, rhs->getNumVectors()));
+    // rhs is initialized, lhs is not
+    //  Default params
     params->get("Output Frequency", 1);
     params->get("Output Style", Belos::Brief);
 
@@ -275,66 +262,51 @@ mxArray* TpetraSystem<Scalar>::solve(RCP<ParameterList> params, RCP<Tpetra::CrsM
 #else
     params->get("Verbosity", Belos::Errors | Belos::Warnings | Belos::IterationDetails | Belos::Warnings | Belos::StatusTestDetails);
 #endif
-    //register all possible solvers
+    // register all possible solvers
     auto problem = rcp(new Belos::LinearProblem<Scalar, Tpetra_MultiVector, Tpetra_Operator>(matrix, lhs, rhs));
     problem->setRightPrec(prec);
-    if(!problem->setProblem())
-    {
+    if (!problem->setProblem()) {
       throw std::runtime_error("ERROR: failed to set up Belos problem.");
     }
     std::string solverName = "GMRES";
-    if(params->isParameter("solver"))
-    {
+    if (params->isParameter("solver")) {
       solverName = params->template get<std::string>("solver");
     }
-    //Convert from basic MueMex solver names to the official Belos names.
-    //At the same time, check that solverName is in the valid set.
+    // Convert from basic MueMex solver names to the official Belos names.
+    // At the same time, check that solverName is in the valid set.
     std::string belosSolverName;
-    if(solverName == "GMRES")
-    {
+    if (solverName == "GMRES") {
       belosSolverName = "PseudoBlock GMRES";
-    }
-    else if(solverName == "CG")
-    {
+    } else if (solverName == "CG") {
       belosSolverName = "PseudoBlock CG";
-    }
-    else
-    {
+    } else {
       std::string msg = std::string("ERROR: requested solver \"") + solverName + "\" not supported. Currently supported solvers: CG, GMRES";
       mexPrintf("%s\n", msg.c_str());
       output = mxCreateDoubleScalar(0);
       return output;
     }
-    Teuchos::RCP<Belos::SolverManager<Scalar, Tpetra_MultiVector, Tpetra_Operator> > solver;
+    Teuchos::RCP<Belos::SolverManager<Scalar, Tpetra_MultiVector, Tpetra_Operator>> solver;
     Belos::SolverFactory<Scalar, Tpetra_MultiVector, Tpetra_Operator> factory;
-    try
-    {
-      //Just use the default parameters for the solver
-      solver = factory.create (belosSolverName, params);
-    }
-    catch(std::exception& e)
-    {
+    try {
+      // Just use the default parameters for the solver
+      solver = factory.create(belosSolverName, params);
+    } catch (std::exception& e) {
       mexPrintf("%s\n", e.what());
       output = mxCreateDoubleScalar(0);
       return output;
     }
     solver->setProblem(problem);
     Belos::ReturnType ret = solver->solve();
-    iters = solver->getNumIters();
-    if(ret == Belos::Converged)
-    {
+    iters                 = solver->getNumIters();
+    if (ret == Belos::Converged) {
       mexPrintf("Success, Belos converged!\n");
       output = saveDataToMatlab(lhs);
-    }
-    else
-    {
+    } else {
       mexPrintf("Belos failed to converge.\n");
-      iters = 0;
+      iters  = 0;
       output = mxCreateDoubleScalar(0);
     }
-  }
-  catch(exception& e)
-  {
+  } catch (exception& e) {
     mexPrintf("Error occurred while running Belos solver:\n");
     cout << e.what() << endl;
     output = mxCreateDoubleScalar(0);
@@ -342,44 +314,35 @@ mxArray* TpetraSystem<Scalar>::solve(RCP<ParameterList> params, RCP<Tpetra::CrsM
   return output;
 }
 
-template<typename Scalar>
-mxArray* TpetraSystem<Scalar>::apply(const mxArray* r)
-{
+template <typename Scalar>
+mxArray* TpetraSystem<Scalar>::apply(const mxArray* r) {
   typedef Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Tpetra_MultiVector;
   RCP<Tpetra_MultiVector> rhs = loadDataFromMatlab<RCP<Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(r);
   RCP<Tpetra_MultiVector> lhs = rcp(new Tpetra_MultiVector(rhs->getMap(), rhs->getNumVectors()));
-  try
-  {
+  try {
     this->prec->apply(*rhs, *lhs);
-  }
-  catch(exception& e)
-  {
+  } catch (exception& e) {
     mexPrintf("Error occurred while applying MueLu-Tpetra preconditioner:\n");
     cout << e.what() << endl;
   }
   return saveDataToMatlab(lhs);
 }
 
-template<>
-RCP<Hierarchy_double> getDatapackHierarchy<double>(MuemexSystem* dp)
-{
+template <>
+RCP<Hierarchy_double> getDatapackHierarchy<double>(MuemexSystem* dp) {
   RCP<MueLu::Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier;
-  switch(dp->type)
-  {
-    case EPETRA:
-    {
-      EpetraSystem* pack = (EpetraSystem*) dp;
-      hier = pack->getHierarchy();
+  switch (dp->type) {
+    case EPETRA: {
+      EpetraSystem* pack = (EpetraSystem*)dp;
+      hier               = pack->getHierarchy();
       break;
     }
-    case TPETRA:
-    {
-      TpetraSystem<double>* pack = (TpetraSystem<double>*) dp;
-      hier = pack->getHierarchy();
+    case TPETRA: {
+      TpetraSystem<double>* pack = (TpetraSystem<double>*)dp;
+      hier                       = pack->getHierarchy();
       break;
     }
-    default:
-    {
+    default: {
       throw runtime_error("Got unexpected linear system type for real-valued functions.");
     }
   }
@@ -387,113 +350,95 @@ RCP<Hierarchy_double> getDatapackHierarchy<double>(MuemexSystem* dp)
 }
 
 #ifdef HAVE_COMPLEX_SCALARS
-template<> RCP<Hierarchy_complex> getDatapackHierarchy<complex_t>(MuemexSystem* dp)
-{
-  return ((TpetraSystem<complex_t>*) dp)->getHierarchy();
+template <>
+RCP<Hierarchy_complex> getDatapackHierarchy<complex_t>(MuemexSystem* dp) {
+  return ((TpetraSystem<complex_t>*)dp)->getHierarchy();
 }
 #endif
 
-template<typename Scalar, typename T>
-void setHierarchyData(MuemexSystem* problem, int levelID, T& data, string& dataName)
-{
+template <typename Scalar, typename T>
+void setHierarchyData(MuemexSystem* problem, int levelID, T& data, string& dataName) {
   RCP<Level> level;
-  if(problem->type == EPETRA)
-  {
-    RCP<Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((EpetraSystem*) problem)->getHierarchy();
-    level = hier->GetLevel(levelID);
-  }
-  else if(problem->type == TPETRA)
-  {
-    RCP<Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((TpetraSystem<double>*) problem)->getHierarchy();
-    level = hier->GetLevel(levelID);
-  }
-  else if(problem->type == TPETRA_COMPLEX)
-  {
+  if (problem->type == EPETRA) {
+    RCP<Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((EpetraSystem*)problem)->getHierarchy();
+    level                                                             = hier->GetLevel(levelID);
+  } else if (problem->type == TPETRA) {
+    RCP<Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((TpetraSystem<double>*)problem)->getHierarchy();
+    level                                                             = hier->GetLevel(levelID);
+  } else if (problem->type == TPETRA_COMPLEX) {
 #ifdef HAVE_COMPLEX_SCALARS
-    RCP<Hierarchy<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((TpetraSystem<complex_t>*) problem)->getHierarchy();
-    level = hier->GetLevel(levelID);
+    RCP<Hierarchy<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = ((TpetraSystem<complex_t>*)problem)->getHierarchy();
+    level                                                                = hier->GetLevel(levelID);
 #else
     throw std::runtime_error("setHierarchyData(): complex scalars not supported.");
 #endif
   }
-  if(level.is_null())
+  if (level.is_null())
     throw runtime_error("Error getting level when setting custom level data.");
   level->Set(dataName, data);
   level->AddKeepFlag(dataName, NoFactory::get(), UserData);
 }
 
-//data pack base class implementation
+// data pack base class implementation
 
-MuemexSystem::MuemexSystem(DataPackType probType) : id(MUEMEX_ERROR), type(probType) {}
+MuemexSystem::MuemexSystem(DataPackType probType)
+  : id(MUEMEX_ERROR)
+  , type(probType) {}
 MuemexSystem::~MuemexSystem() {}
 
-mxArray* MuemexSystem::getHierarchyData(string dataName, MuemexType dataType, int levelID)
-{
+mxArray* MuemexSystem::getHierarchyData(string dataName, MuemexType dataType, int levelID) {
   mxArray* output = NULL;
-  try
-  {
-    //First, get Level, which doesn't depend on Epetra vs. Tpetra
+  try {
+    // First, get Level, which doesn't depend on Epetra vs. Tpetra
     RCP<MueLu::Level> level;
     RCP<const FactoryManagerBase> fmb;
-    if(this->type == TPETRA)
-    {
-      TpetraSystem<double>* tsys = (TpetraSystem<double>*) this;
-      if(tsys->keepAll)
+    if (this->type == TPETRA) {
+      TpetraSystem<double>* tsys = (TpetraSystem<double>*)this;
+      if (tsys->keepAll)
         fmb = tsys->systemManagers[levelID];
-    }
-    else if(this->type == TPETRA_COMPLEX)
-    {
+    } else if (this->type == TPETRA_COMPLEX) {
 #ifdef HAVE_COMPLEX_SCALARS
-      TpetraSystem<complex_t>* tsys = (TpetraSystem<complex_t>*) this;
-      if(tsys->keepAll)
+      TpetraSystem<complex_t>* tsys = (TpetraSystem<complex_t>*)this;
+      if (tsys->keepAll)
         fmb = tsys->systemManagers[levelID];
 #else
       throw std::runtime_error("getHierarchyData(): complex scalars not supported.");
 #endif
     }
-    const FactoryBase* factory = NoFactory::get(); //(ptr to constant)
-    bool needFMB = true;
-    if(dataName == "A" || dataName == "P") //these are kept by default, don't use actual factory pointer
-      //Otherwise would break getting A and P when 'keep' is off
+    const FactoryBase* factory = NoFactory::get();  //(ptr to constant)
+    bool needFMB               = true;
+    if (dataName == "A" || dataName == "P")  // these are kept by default, don't use actual factory pointer
+      // Otherwise would break getting A and P when 'keep' is off
       needFMB = false;
-    switch(this->type)
-    {
+    switch (this->type) {
       case EPETRA:
-      case TPETRA:
-      {
+      case TPETRA: {
         RCP<OpenHierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = rcp_static_cast<OpenHierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>(getDatapackHierarchy<double>(this));
-        level = hier->GetLevel(levelID);
-        if(needFMB)
-        {
-          if(fmb.is_null())
-            fmb = (RCP<const FactoryManagerBase>) hier->GetFactoryManager(levelID);
-          if(!fmb.is_null())
-          {
-            try
-            {
+        level                                                                 = hier->GetLevel(levelID);
+        if (needFMB) {
+          if (fmb.is_null())
+            fmb = (RCP<const FactoryManagerBase>)hier->GetFactoryManager(levelID);
+          if (!fmb.is_null()) {
+            try {
               factory = fmb->GetFactory(dataName).get();
-            }
-            catch(exception& e) {} //forced to try using NoFactory (which will work with default keeps A, P)
+            } catch (exception& e) {
+            }  // forced to try using NoFactory (which will work with default keeps A, P)
           }
         }
         break;
       }
-      case TPETRA_COMPLEX:
-      {
+      case TPETRA_COMPLEX: {
 #ifdef HAVE_COMPLEX_SCALARS
         RCP<OpenHierarchy<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = rcp_static_cast<OpenHierarchy<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>(getDatapackHierarchy<complex_t>(this));
-        level = hier->GetLevel(levelID);
-        if(needFMB)
-        {
-          if(fmb.is_null())
-            fmb = (RCP<const FactoryManagerBase>) hier->GetFactoryManager(levelID);
-          if(!fmb.is_null())
-          {
-            try
-            {
+        level                                                                    = hier->GetLevel(levelID);
+        if (needFMB) {
+          if (fmb.is_null())
+            fmb = (RCP<const FactoryManagerBase>)hier->GetFactoryManager(levelID);
+          if (!fmb.is_null()) {
+            try {
               factory = fmb->GetFactory(dataName).get();
-            }
-            catch(exception& e) {} //attempt to use NoFactory
+            } catch (exception& e) {
+            }  // attempt to use NoFactory
           }
         }
         break;
@@ -502,70 +447,58 @@ mxArray* MuemexSystem::getHierarchyData(string dataName, MuemexType dataType, in
 #endif
       }
     }
-    if(level.is_null())
+    if (level.is_null())
       throw runtime_error("Can't get level data because level is null.");
     bool dataIsAvailable = level->IsAvailable(dataName, factory);
-    if(!dataIsAvailable)
-    {
-      //Give the level the FactoryManager again so it can provide the data (by re-creating it, if necessary)
+    if (!dataIsAvailable) {
+      // Give the level the FactoryManager again so it can provide the data (by re-creating it, if necessary)
       level->SetFactoryManager(fmb);
     }
-    //Given the dataName and factory pointer, all data in the level should now be accessible
-    switch(dataType)
-    {
+    // Given the dataName and factory pointer, all data in the level should now be accessible
+    switch (dataType) {
       case XPETRA_MATRIX_DOUBLE:
         return saveDataToMatlab(level->Get<RCP<Xpetra_Matrix_double>>(dataName, factory));
-      case XPETRA_MATRIX_COMPLEX:
-      {
+      case XPETRA_MATRIX_COMPLEX: {
 #ifdef HAVE_COMPLEX_SCALARS
         return saveDataToMatlab(level->Get<RCP<Xpetra_Matrix_complex>>(dataName, factory));
 #endif
       }
       case XPETRA_MULTIVECTOR_DOUBLE:
-        if(dataName == "Coordinates")
-        {
-          //Coordinates is special because it's always user-provided on level 0, not always provided at all, not always kept in the level (only kept if doing agg viz, etc), and is always MV<double> regardless of problem scalar type
+        if (dataName == "Coordinates") {
+          // Coordinates is special because it's always user-provided on level 0, not always provided at all, not always kept in the level (only kept if doing agg viz, etc), and is always MV<double> regardless of problem scalar type
           double errReturn = -1;
-          if(level->GetLevelID() == 0)
-          {
-            //Try to get coordinates as if it's user data, but don't be surprised if it's not there at all.
-            try
-            {
+          if (level->GetLevelID() == 0) {
+            // Try to get coordinates as if it's user data, but don't be surprised if it's not there at all.
+            try {
               RCP<Xpetra_MultiVector_double> coords = level->Get<RCP<Xpetra_MultiVector_double>>(dataName, NoFactory::get());
-              if(coords.is_null())
-                throw runtime_error("Coordinates were not available (Level 0)."); //just print the message below and return -1
+              if (coords.is_null())
+                throw runtime_error("Coordinates were not available (Level 0).");  // just print the message below and return -1
               return saveDataToMatlab(coords);
-            }
-            catch(exception& e)
-            {
-              cout << endl << "Coordinates were not available on Level 0." << endl;
+            } catch (exception& e) {
+              cout << endl
+                   << "Coordinates were not available on Level 0." << endl;
               cout << "They must be provided by the user and aren't generated or kept by default (even in MueMex 'keep' mode)." << endl;
               cout << "User-provided coordinates for Level 0 will be kept and passed to other levels if something requires them:" << endl;
-              cout << "aggregate visualization, brick aggregation, repartitioning or distance laplacian filtering." << endl << endl;
+              cout << "aggregate visualization, brick aggregation, repartitioning or distance laplacian filtering." << endl
+                   << endl;
               return saveDataToMatlab(errReturn);
             }
-          }
-          else
-          {
-            //If coords are provided & kept, they are produced by CoordinatesTransferFactory in levels > 0.
-            try
-            {
+          } else {
+            // If coords are provided & kept, they are produced by CoordinatesTransferFactory in levels > 0.
+            try {
               RCP<Xpetra_MultiVector_double> coords = level->Get<RCP<Xpetra_MultiVector_double>>(dataName, factory);
-              if(coords.is_null())
+              if (coords.is_null())
                 throw runtime_error("Coordinates were not available (Level > 0).");
               return saveDataToMatlab(coords);
-            }
-            catch(exception& e)
-            {
+            } catch (exception& e) {
               cout << "Coordinates must be provided by the user and aren't generated or kept by default (even in MueMex 'keep' mode)." << endl;
               cout << "User-provided coordinates for Level 0 will be kept and passed to other levels if something requires them:" << endl;
-              cout << "aggregate visualization, brick aggregation, repartitioning or distance laplacian filtering." << endl << endl;
+              cout << "aggregate visualization, brick aggregation, repartitioning or distance laplacian filtering." << endl
+                   << endl;
               return saveDataToMatlab(errReturn);
             }
           }
-        }
-        else
-        {
+        } else {
           return saveDataToMatlab(level->Get<RCP<Xpetra_MultiVector_double>>(dataName, factory));
         }
       case XPETRA_MULTIVECTOR_COMPLEX:
@@ -590,61 +523,53 @@ mxArray* MuemexSystem::getHierarchyData(string dataName, MuemexType dataType, in
       default:
         throw runtime_error("Invalid MuemexType for getting hierarchy data.");
     }
-  }
-  catch(exception& e)
-  {
+  } catch (exception& e) {
     mexPrintf("Error occurred while getting hierarchy data.\n");
     cout << e.what() << endl;
   }
   return output;
 }
 
-//EpetraSystem impl
+// EpetraSystem impl
 
-EpetraSystem::EpetraSystem() : MuemexSystem(EPETRA) {}
+EpetraSystem::EpetraSystem()
+  : MuemexSystem(EPETRA) {}
 EpetraSystem::~EpetraSystem() {}
 
-int EpetraSystem::status()
-{
+int EpetraSystem::status() {
   mexPrintf("**** Problem ID %d [MueLu_Epetra] ****\n", id);
-  if(!A.is_null())
+  if (!A.is_null())
     mexPrintf("Matrix: %dx%d w/ %d nnz\n", A->NumGlobalRows(), A->NumGlobalCols(), A->NumMyNonzeros());
   mexPrintf("Operator Complexity: %f\n", operatorComplexity);
-  if(!List.is_null())
-    {
-      mexPrintf("Parameter List:\n");
-      List->print();
-    }
+  if (!List.is_null()) {
+    mexPrintf("Parameter List:\n");
+    List->print();
+  }
   mexPrintf("\n");
   return IS_TRUE;
-}/*end status*/
+} /*end status*/
 
-int EpetraSystem::setup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords)
-{
+int EpetraSystem::setup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords) {
   bool success = false;
-  try
-    {
-      /* Matrix Fill */
-      A = loadDataFromMatlab<RCP<Epetra_CrsMatrix>>(matlabA);
-      if(haveCoords)
-      {
-        //Create 'user data' sublist if it doesn't already exist
-        auto userData = Teuchos::sublist(List, "user data");
-        userData->set("Coordinates", loadDataFromMatlab<RCP<Epetra_MultiVector>>(matlabCoords));
-      }
-      prec = MueLu::CreateEpetraPreconditioner(A, *List);
-      //underlying the Epetra_Operator prec is a MueLu::EpetraOperator
-      RCP<MueLu::EpetraOperator> meo = rcp_static_cast<MueLu::EpetraOperator, Epetra_Operator>(prec);
-      operatorComplexity = meo->GetHierarchy()->GetOperatorComplexity();
-      success = true;
-    }
-  catch(exception& e)
-    {
-      mexPrintf("Error occurred while setting up epetra problem:\n");
-      cout << e.what() << endl;
+  try {
+    /* Matrix Fill */
+    A = loadDataFromMatlab<RCP<Epetra_CrsMatrix>>(matlabA);
+    if (haveCoords) {
+      // Create 'user data' sublist if it doesn't already exist
+      auto userData = Teuchos::sublist(List, "user data");
+      userData->set("Coordinates", loadDataFromMatlab<RCP<Epetra_MultiVector>>(matlabCoords));
     }
+    prec = MueLu::CreateEpetraPreconditioner(A, *List);
+    // underlying the Epetra_Operator prec is a MueLu::EpetraOperator
+    RCP<MueLu::EpetraOperator> meo = rcp_static_cast<MueLu::EpetraOperator, Epetra_Operator>(prec);
+    operatorComplexity             = meo->GetHierarchy()->GetOperatorComplexity();
+    success                        = true;
+  } catch (exception& e) {
+    mexPrintf("Error occurred while setting up epetra problem:\n");
+    cout << e.what() << endl;
+  }
   return success ? IS_TRUE : IS_FALSE;
-}/*end setup*/
+} /*end setup*/
 
 /* EpetraSystem::solve - Given two Teuchos lists, one in the EpetraSystem, and one of
    solve-time options, this routine calls the relevant solver and returns the solution.
@@ -655,13 +580,11 @@ int EpetraSystem::setup(const mxArray* matlabA, bool haveCoords, const mxArray*
    iters   - number of iterations taken [O]
    Returns: IS_TRUE if solve was succesful, IS_FALSE otherwise
 */
-mxArray* EpetraSystem::solve(RCP<ParameterList> TPL, RCP<Epetra_CrsMatrix> matrix, const mxArray* b, int& iters)
-{
+mxArray* EpetraSystem::solve(RCP<ParameterList> TPL, RCP<Epetra_CrsMatrix> matrix, const mxArray* b, int& iters) {
   mxArray* output;
-  try
-  {
-    //Set up X and B
-    Epetra_Map map = matrix->DomainMap();
+  try {
+    // Set up X and B
+    Epetra_Map map              = matrix->DomainMap();
     RCP<Epetra_MultiVector> rhs = loadDataFromMatlab<RCP<Epetra_MultiVector>>(b);
     RCP<Epetra_MultiVector> lhs = rcp(new Epetra_MultiVector(map, rhs->NumVectors(), true));
     // Default params
@@ -672,33 +595,28 @@ mxArray* EpetraSystem::solve(RCP<ParameterList> TPL, RCP<Epetra_CrsMatrix> matri
 #else
     TPL->get("Verbosity", Belos::Errors + Belos::Warnings + Belos::IterationDetails + Belos::Warnings + Belos::StatusTestDetails);
 #endif
-   RCP<Belos::LinearProblem<double, Epetra_MultiVector, Epetra_Operator>> problem = rcp(new    Belos::LinearProblem<double, Epetra_MultiVector, Epetra_Operator>(matrix, lhs, rhs));
-    RCP<Belos::EpetraPrecOp> epo = rcp(new Belos::EpetraPrecOp(prec));
+    RCP<Belos::LinearProblem<double, Epetra_MultiVector, Epetra_Operator>> problem = rcp(new Belos::LinearProblem<double, Epetra_MultiVector, Epetra_Operator>(matrix, lhs, rhs));
+    RCP<Belos::EpetraPrecOp> epo                                                   = rcp(new Belos::EpetraPrecOp(prec));
     problem->setRightPrec(epo);
     bool set = problem->setProblem();
     TEUCHOS_TEST_FOR_EXCEPTION(!set, runtime_error, "Linear Problem failed to set up correctly!");
     Belos::SolverFactory<double, Epetra_MultiVector, Epetra_Operator> factory;
-    //Get the solver name from the parameter list, default to PseudoBlockGmres if none specified by user
-    string solverName = TPL->get("solver", "GMRES");
+    // Get the solver name from the parameter list, default to PseudoBlockGmres if none specified by user
+    string solverName                                                             = TPL->get("solver", "GMRES");
     RCP<Belos::SolverManager<double, Epetra_MultiVector, Epetra_Operator>> solver = factory.create(solverName, TPL);
     solver->setProblem(problem);
     Belos::ReturnType ret = solver->solve();
-    if(ret == Belos::Converged)
-    {
+    if (ret == Belos::Converged) {
       mexPrintf("Success, Belos converged!\n");
-      iters = solver->getNumIters();
+      iters  = solver->getNumIters();
       output = saveDataToMatlab(lhs);
-    }
-    else
-    {
+    } else {
       mexPrintf("Belos failed to converge.\n");
-      iters = 0;
+      iters  = 0;
       output = mxCreateDoubleScalar(0);
     }
     output = saveDataToMatlab(lhs);
-  }
-  catch(exception& e)
-  {
+  } catch (exception& e) {
     mexPrintf("Error occurred during Belos solve:\n");
     cout << e.what() << endl;
     output = mxCreateDoubleScalar(0);
@@ -706,65 +624,53 @@ mxArray* EpetraSystem::solve(RCP<ParameterList> TPL, RCP<Epetra_CrsMatrix> matri
   return output;
 }
 
-mxArray* EpetraSystem::apply(const mxArray* r)
-{
+mxArray* EpetraSystem::apply(const mxArray* r) {
   RCP<Epetra_MultiVector> rhs = loadDataFromMatlab<RCP<Epetra_MultiVector>>(r);
   Epetra_SerialComm Comm;
   Epetra_Map map(rhs->GlobalLength(), 0, Comm);
   RCP<Epetra_MultiVector> lhs = rcp(new Epetra_MultiVector(map, rhs->NumVectors(), true));
-  try
-  {
+  try {
     this->prec->Apply(*rhs, *lhs);
-  }
-  catch(exception& e)
-  {
+  } catch (exception& e) {
     mexPrintf("Error occurred while applying MueLu-Epetra preconditioner:\n");
     cout << e.what() << endl;
   }
   return saveDataToMatlab(lhs);
 }
 
-RCP<Hierarchy_double> EpetraSystem::getHierarchy()
-{
-  RCP<MueLu::EpetraOperator> meo = rcp_static_cast<MueLu::EpetraOperator, Epetra_Operator>(prec);
+RCP<Hierarchy_double> EpetraSystem::getHierarchy() {
+  RCP<MueLu::EpetraOperator> meo                                           = rcp_static_cast<MueLu::EpetraOperator, Epetra_Operator>(prec);
   RCP<MueLu::Hierarchy<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> hier = meo->GetHierarchy();
-  if(hier.is_null())
+  if (hier.is_null())
     throw runtime_error("Hierarchy from Epetra problem was null.");
   return hier;
 }
 
-//tpetra_double_data_pack implementation
+// tpetra_double_data_pack implementation
 
-template<> TpetraSystem<double>::TpetraSystem() : MuemexSystem(TPETRA) {}
-template<> TpetraSystem<double>::~TpetraSystem() {}
+template <>
+TpetraSystem<double>::TpetraSystem()
+  : MuemexSystem(TPETRA) {}
+template <>
+TpetraSystem<double>::~TpetraSystem() {}
 
-template<typename Scalar>
-int TpetraSystem<Scalar>::setup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords)
-{
-  //decide whether do do default or custom setup
+template <typename Scalar>
+int TpetraSystem<Scalar>::setup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords) {
+  // decide whether do do default or custom setup
   bool doCustomSetup = List->isParameter("keep") && List->isType<bool>("keep") && List->get<bool>("keep");
-  List->remove("keep", false); //"keep" would cause Plist validation to fail if left in
-  if(doCustomSetup)
-  {
-    try
-    {
+  List->remove("keep", false);  //"keep" would cause Plist validation to fail if left in
+  if (doCustomSetup) {
+    try {
       customSetup(matlabA, haveCoords, matlabCoords);
-    }
-    catch(exception& e)
-    {
+    } catch (exception& e) {
       cout << "An error occurred during Tpetra custom problem setup:" << endl;
       cout << e.what() << endl;
       return IS_FALSE;
     }
-  }
-  else
-  {
-    try
-    {
+  } else {
+    try {
       normalSetup(matlabA, haveCoords, matlabCoords);
-    }
-    catch(exception& e)
-    {
+    } catch (exception& e) {
       cout << "An error occurred during Tpetra preconditioner setup:" << endl;
       cout << e.what();
       return IS_FALSE;
@@ -773,50 +679,38 @@ int TpetraSystem<Scalar>::setup(const mxArray* matlabA, bool haveCoords, const m
   return IS_TRUE;
 }
 
-template<typename Scalar>
-void TpetraSystem<Scalar>::normalSetup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords)
-{
+template <typename Scalar>
+void TpetraSystem<Scalar>::normalSetup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords) {
   keepAll = false;
-  A = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(matlabA);
-  RCP<Tpetra::Operator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > opA(A);
+  A       = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(matlabA);
+  RCP<Tpetra::Operator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> opA(A);
   RCP<MueLu::TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> mop;
-  if(haveCoords)
-  {
+  if (haveCoords) {
     auto userData = Teuchos::sublist(List, "user data");
     userData->set("Coordinates", loadDataFromMatlab<RCP<Tpetra_MultiVector_double>>(matlabCoords));
   }
-  //Create the nullspace if not already set by user through XML
-  if(!(List->isSublist("level 0") && List->sublist("level 0", true).isParameter("Nullspace"))
-    && !(List->isSublist("user data") && List->sublist("user data", true).isParameter("Nullspace")))
-  {
+  // Create the nullspace if not already set by user through XML
+  if (!(List->isSublist("level 0") && List->sublist("level 0", true).isParameter("Nullspace")) && !(List->isSublist("user data") && List->sublist("user data", true).isParameter("Nullspace"))) {
     int nPDE = MasterList::getDefault<int>("number of equations");
-    if (List->isSublist("Matrix"))
-    {
+    if (List->isSublist("Matrix")) {
       // Factory style parameter list
       const Teuchos::ParameterList& operatorList = List->sublist("Matrix");
       if (operatorList.isParameter("PDE equations"))
         nPDE = operatorList.get<int>("PDE equations");
-    }
-    else if (List->isParameter("number of equations"))
-    {
+    } else if (List->isParameter("number of equations")) {
       // Easy style parameter list
       nPDE = List->get<int>("number of equations");
     }
     mexPrintf("** Constructing nullspace for %d PDEs\n", nPDE);
     auto domainMap = A->getDomainMap();
     auto nullspace = rcp(new Tpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(domainMap, nPDE));
-    if (nPDE == 1)
-    {
+    if (nPDE == 1) {
       nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    }
-    else
-    {
+    } else {
       typedef typename Teuchos::ArrayRCP<Scalar>::size_type arrayRCPSizeType;
-      for (int i = 0; i < nPDE; i++)
-      {
+      for (int i = 0; i < nPDE; i++) {
         Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
-        for (arrayRCPSizeType j = 0; j < nsData.size(); j++)
-        {
+        for (arrayRCPSizeType j = 0; j < nsData.size(); j++) {
           mm_GlobalOrd GID = domainMap->getGlobalElement(j) - domainMap->getIndexBase();
           if ((GID - i) % nPDE == 0)
             nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
@@ -826,69 +720,58 @@ void TpetraSystem<Scalar>::normalSetup(const mxArray* matlabA, bool haveCoords,
     auto userData = Teuchos::sublist(List, "user data");
     userData->set("Nullspace", nullspace);
   }
-  mop = MueLu::CreateTpetraPreconditioner<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(opA, *List);
+  mop  = MueLu::CreateTpetraPreconditioner<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(opA, *List);
   prec = rcp_implicit_cast<Tpetra::Operator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>(mop);
 
   // print data??
-  //mop->GetHierarchy()->GetLevel(0)->print(std::cout, MueLu::Debug);
+  // mop->GetHierarchy()->GetLevel(0)->print(std::cout, MueLu::Debug);
 
   operatorComplexity = mop->GetHierarchy()->GetOperatorComplexity();
 }
 
-template<typename Scalar>
-void TpetraSystem<Scalar>::customSetup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords)
-{
+template <typename Scalar>
+void TpetraSystem<Scalar>::customSetup(const mxArray* matlabA, bool haveCoords, const mxArray* matlabCoords) {
   keepAll = true;
-  A = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(matlabA);
+  A       = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(matlabA);
   RCP<MueLu::TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> mop;
-  //Now modify CreateTpetraPreconditioner to set keep flags on all factories
+  // Now modify CreateTpetraPreconditioner to set keep flags on all factories
   typedef Xpetra::MultiVector<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> MultiVector;
   typedef Xpetra::Matrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Matrix;
   typedef Hierarchy<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> Hierarchy;
   typedef HierarchyManager<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t> HierarchyManager;
   RCP<HierarchyManager> mueluFactory = rcp(new ParameterListInterpreter<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(*List, A->getComm()));
-  RCP<Hierarchy> H = mueluFactory->CreateHierarchy();
+  RCP<Hierarchy> H                   = mueluFactory->CreateHierarchy();
   H->setlib(Xpetra::UseTpetra);
   RCP<Matrix> xA = TpetraCrs_To_XpetraMatrix<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(A);
   H->GetLevel(0)->Set("A", xA);
-  if(haveCoords)
-  {
+  if (haveCoords) {
     RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> coords = loadDataFromMatlab<RCP<Xpetra_MultiVector_double>>(matlabCoords);
     H->GetLevel(0)->Set("Coordinates", coords);
   }
-  //Decide whether user passed level 0 Nullspace in parameter list. If not, make it here.
-  if(!List->isSublist("level 0") || !List->sublist("level 0", true).isParameter("Nullspace"))
-  {
+  // Decide whether user passed level 0 Nullspace in parameter list. If not, make it here.
+  if (!List->isSublist("level 0") || !List->sublist("level 0", true).isParameter("Nullspace")) {
     int nPDE = MasterList::getDefault<int>("number of equations");
-    if (List->isSublist("Matrix"))
-    {
+    if (List->isSublist("Matrix")) {
       // Factory style parameter list
       const Teuchos::ParameterList& operatorList = List->sublist("Matrix");
       if (operatorList.isParameter("PDE equations"))
         nPDE = operatorList.get<int>("PDE equations");
-    }
-    else if (List->isParameter("number of equations"))
-    {
+    } else if (List->isParameter("number of equations")) {
       // Easy style parameter list
       nPDE = List->get<int>("number of equations");
     }
     RCP<MultiVector> nullspace = Xpetra::MultiVectorFactory<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(xA->getDomainMap(), nPDE);
-    if (nPDE == 1)
-    {
+    if (nPDE == 1) {
       nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    }
-    else
-    {
+    } else {
       typedef typename Teuchos::ArrayRCP<Scalar>::size_type arrayRCPSizeType;
-      for (int i = 0; i < nPDE; i++)
-      {
+      for (int i = 0; i < nPDE; i++) {
         Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
-        for (arrayRCPSizeType j = 0; j < nsData.size(); j++)
-        {
-          //TODO optimizations:
-          //TODO This can be optimized by getting the domain map and index base outside the loop.
-          //TODO Also, the whole local-to-global lookup table can be fetched one time, instead of repeatedly
-          //TODO calling getGlobalElement.
+        for (arrayRCPSizeType j = 0; j < nsData.size(); j++) {
+          // TODO optimizations:
+          // TODO This can be optimized by getting the domain map and index base outside the loop.
+          // TODO Also, the whole local-to-global lookup table can be fetched one time, instead of repeatedly
+          // TODO calling getGlobalElement.
           mm_GlobalOrd GID = A->getDomainMap()->getGlobalElement(j) - A->getDomainMap()->getIndexBase();
           if ((GID - i) % nPDE == 0)
             nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
@@ -900,99 +783,93 @@ void TpetraSystem<Scalar>::customSetup(const mxArray* matlabA, bool haveCoords,
   Teuchos::ParameterList nonSerialList, dummyList;
   ExtractNonSerializableData(*List, dummyList, nonSerialList);
   HierarchyUtils<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>::AddNonSerializableDataToHierarchy(*mueluFactory, *H, nonSerialList);
-  //Set up dummy levels in hierarchy
-  for(int i = 0; i < 5; i++)
-  {
+  // Set up dummy levels in hierarchy
+  for (int i = 0; i < 5; i++) {
     RCP<Level> l = rcp(new Level());
     H->AddLevel(l);
   }
-  //Set keep flags on ALL factories in ALL levels
-  //We have access to H's list of FactoryManagers so we know how to get factory pointer given the name of the factory.
-  //We don't know which names are in the FactoryManagers though so a brute force approach is needed...
-  vector<string> keepItems = {"A", "P", "R", "Ptent", "Aggregates", "Coordinates", "UnAmalgamationInfo", "Smoother", "PreSmoother", "PostSmoother", "CoarseSolver", "Graph", "CoarseMap", "Nullspace", "Ppattern", "Constraint", "CoarseNumZLayers", "LineDetection_Layers", "LineDetection_VertLineIds", "Partition", "Importer", "DofsPerNode", "Filtering" "pcoarsen: element to node map"};
+  // Set keep flags on ALL factories in ALL levels
+  // We have access to H's list of FactoryManagers so we know how to get factory pointer given the name of the factory.
+  // We don't know which names are in the FactoryManagers though so a brute force approach is needed...
+  vector<string> keepItems                                               = {"A", "P", "R", "Ptent", "Aggregates", "Coordinates", "UnAmalgamationInfo", "Smoother", "PreSmoother", "PostSmoother", "CoarseSolver", "Graph", "CoarseMap", "Nullspace", "Ppattern", "Constraint", "CoarseNumZLayers", "LineDetection_Layers", "LineDetection_VertLineIds", "Partition", "Importer", "DofsPerNode",
+                                                                            "Filtering"
+                                                                                                                          "pcoarsen: element to node map"};
   RCP<OpenHierarchy<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> openH = rcp_static_cast<OpenHierarchy<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>, Hierarchy>(H);
-  if(openH.is_null())
+  if (openH.is_null())
     throw runtime_error("Could not cast RCP<Hierarchy> to subclass.");
-  for(int lvl = 0; lvl < H->GetNumLevels(); lvl++)
-  {
-    RCP<const FactoryManagerBase> fman = (RCP<const FactoryManagerBase>) mueluFactory->GetFactoryManager(lvl);
+  for (int lvl = 0; lvl < H->GetNumLevels(); lvl++) {
+    RCP<const FactoryManagerBase> fman = (RCP<const FactoryManagerBase>)mueluFactory->GetFactoryManager(lvl);
     systemManagers.push_back(fman);
-    for(auto s : keepItems)
-    {
-      try
-      {
-        const RCP<const FactoryBase> fact = fman->GetFactory(s); //will throw if factory doesn't exist, ignore in that case
-        if(!fact.is_null())
-        {
-          FactoryBase* factPtr = (FactoryBase*) fact.get();
-          //Add keep flag to level
+    for (auto s : keepItems) {
+      try {
+        const RCP<const FactoryBase> fact = fman->GetFactory(s);  // will throw if factory doesn't exist, ignore in that case
+        if (!fact.is_null()) {
+          FactoryBase* factPtr = (FactoryBase*)fact.get();
+          // Add keep flag to level
           H->GetLevel(lvl)->Keep(s, factPtr);
         }
+      } catch (exception& e) {
       }
-      catch(exception& e) {}
     }
   }
   mueluFactory->SetupHierarchy(*H);
   operatorComplexity = H->GetOperatorComplexity();
-  prec = rcp(new TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(H));
+  prec               = rcp(new TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(H));
 }
 
-template<>
-int TpetraSystem<double>::status()
-{
+template <>
+int TpetraSystem<double>::status() {
   mexPrintf("**** Problem ID %d [MueLu_Tpetra] ****\n", id);
-  if(!A.is_null())
+  if (!A.is_null())
     mexPrintf("Matrix: %dx%d w/ %d nnz\n", A->getGlobalNumRows(), A->getGlobalNumCols(), A->getGlobalNumEntries());
   mexPrintf("Operator Complexity: %f\n", operatorComplexity);
-  if(!List.is_null())
-    {
-      mexPrintf("Parameter List:\n");
-      List->print();
-    }
+  if (!List.is_null()) {
+    mexPrintf("Parameter List:\n");
+    List->print();
+  }
   mexPrintf("\n");
   return IS_TRUE;
 }
 
-template<typename Scalar>
-RCP<MueLu::Hierarchy<Scalar, mm_GlobalOrd, mm_LocalOrd, mm_node_t>> TpetraSystem<Scalar>::getHierarchy()
-{
+template <typename Scalar>
+RCP<MueLu::Hierarchy<Scalar, mm_GlobalOrd, mm_LocalOrd, mm_node_t>> TpetraSystem<Scalar>::getHierarchy() {
   RCP<MueLu::TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> mueluOp = rcp_static_cast<MueLu::TpetraOperator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>, Tpetra::Operator<Scalar, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>(prec);
-  if(mueluOp.is_null())
+  if (mueluOp.is_null())
     throw runtime_error("Tpetra precondition operator was null.");
   RCP<MueLu::Hierarchy<Scalar, mm_GlobalOrd, mm_LocalOrd, mm_node_t>> hier = mueluOp->GetHierarchy();
-  if(hier.is_null())
+  if (hier.is_null())
     throw runtime_error("Hierarchy from Tpetra problem was null.");
   return hier;
 }
 
-//tpetra_complex_data_pack implementation
+// tpetra_complex_data_pack implementation
 
 #ifdef HAVE_COMPLEX_SCALARS
-template<> TpetraSystem<complex_t>::TpetraSystem() : MuemexSystem(TPETRA_COMPLEX) {}
-template<> TpetraSystem<complex_t>::~TpetraSystem() {}
-
-template<>
-int TpetraSystem<complex_t>::status()
-{
+template <>
+TpetraSystem<complex_t>::TpetraSystem()
+  : MuemexSystem(TPETRA_COMPLEX) {}
+template <>
+TpetraSystem<complex_t>::~TpetraSystem() {}
+
+template <>
+int TpetraSystem<complex_t>::status() {
   mexPrintf("**** Problem ID %d [MueLu_Tpetra (Complex Scalars)] ****\n", id);
-  if(!A.is_null())
+  if (!A.is_null())
     mexPrintf("Matrix: %dx%d w/ %d nnz\n", A->getGlobalNumRows(), A->getGlobalNumCols(), A->getGlobalNumEntries());
   mexPrintf("Operator Complexity: %f\n", operatorComplexity);
-  if(!List.is_null())
-  {
+  if (!List.is_null()) {
     mexPrintf("Parameter List:\n");
     List->print();
   }
   mexPrintf("\n");
   return IS_TRUE;
 }
-#endif  //HAVE_COMPLEX_SCALARS
+#endif  // HAVE_COMPLEX_SCALARS
 
-//MuemexSystemList namespace implementation
+// MuemexSystemList namespace implementation
 
-void MuemexSystemList::clearAll()
-{
-  //When items are cleared, RCPs will auto-delete the datapacks
+void MuemexSystemList::clearAll() {
+  // When items are cleared, RCPs will auto-delete the datapacks
   list.clear();
 }
 
@@ -1001,14 +878,13 @@ void MuemexSystemList::clearAll()
    D       - The MuemexSystem. [I]
    Returns: problem id number of D
 */
-int MuemexSystemList::add(RCP<MuemexSystem> D)
-{
+int MuemexSystemList::add(RCP<MuemexSystem> D) {
   TEUCHOS_ASSERT(!D.is_null());
   D->id = nextID;
   nextID++;
   list.push_back(D);
   return D->id;
-}       /*end add*/
+} /*end add*/
 
 /* find - Finds problem by id
    Parameters:
@@ -1016,78 +892,63 @@ int MuemexSystemList::add(RCP<MuemexSystem> D)
    Returns: pointer to MuemexSystem matching 'id', if found, NULL if not
    found.
 */
-RCP<MuemexSystem> MuemexSystemList::find(int id)
-{
-  if(isInList(id))
-  {
-    for(auto problem : list)
-    {
-      if(problem->id == id)
+RCP<MuemexSystem> MuemexSystemList::find(int id) {
+  if (isInList(id)) {
+    for (auto problem : list) {
+      if (problem->id == id)
         return problem;
     }
   }
   RCP<MuemexSystem> notFound;
   return notFound;
-}/*end find*/
+} /*end find*/
 
 /* remove - Removes problem by id
    Parameters:
    id      - ID number [I]
    Returns: IS_TRUE if remove was succesful, IS_FALSE otherwise
 */
-int MuemexSystemList::remove(int id)
-{
+int MuemexSystemList::remove(int id) {
   int index = -1;
-  for(int i = 0; i < int(list.size()); i++)
-    {
-      if(list[i]->id == id)
-        {
-          index = i;
-          break;
-        }
-    }
-  if(index == -1)
-    {
-      mexErrMsgTxt("Error: Tried to clean up a problem that doesn't exist.");
-      return IS_FALSE;
+  for (int i = 0; i < int(list.size()); i++) {
+    if (list[i]->id == id) {
+      index = i;
+      break;
     }
+  }
+  if (index == -1) {
+    mexErrMsgTxt("Error: Tried to clean up a problem that doesn't exist.");
+    return IS_FALSE;
+  }
   list.erase(list.begin() + index);
   return IS_TRUE;
-}/*end remove*/
+} /*end remove*/
 
 /* size - Number of stored problems */
-int MuemexSystemList::size()
-{
+int MuemexSystemList::size() {
   return list.size();
 }
 
 /* Returns the status of all members of the list
    Returns IS_TRUE
 */
-int MuemexSystemList::status_all()
-{
-  //This prints all the existing problems in ascending order by ID
-  for(int i = 0; i < nextID; i++)
-    {
-      for(auto problem : list)
-        {
-          if(problem->id == i)
-            {
-              problem->status();
-              break;
-            }
-        }
+int MuemexSystemList::status_all() {
+  // This prints all the existing problems in ascending order by ID
+  for (int i = 0; i < nextID; i++) {
+    for (auto problem : list) {
+      if (problem->id == i) {
+        problem->status();
+        break;
+      }
     }
+  }
   return IS_TRUE;
-}/*end status_all */
+} /*end status_all */
 
-bool MuemexSystemList::isInList(int id)
-{
+bool MuemexSystemList::isInList(int id) {
   bool rv = false;
-  for(auto problem : list)
-  {
-    if(problem->id == id)
-    {
+  for (auto problem : list) {
+    if (problem->id == id) {
       rv = true;
       break;
     }
@@ -1103,107 +964,97 @@ bool MuemexSystemList::isInList(int id)
    Return value: Which mode to run the program in.
 */
 
-MODE_TYPE sanity_check(int nrhs, const mxArray *prhs[])
-{
+MODE_TYPE sanity_check(int nrhs, const mxArray* prhs[]) {
   MODE_TYPE rv = MODE_ERROR;
   /* Check for mode */
-  if(nrhs == 0)
+  if (nrhs == 0)
     mexErrMsgTxt("Error: muelu() expects at least one argument\n");
   /* Pull mode data from 1st Input */
-  MODE_TYPE mode = (MODE_TYPE) loadDataFromMatlab<int>(prhs[0]);
-  switch (mode)
-  {
+  MODE_TYPE mode = (MODE_TYPE)loadDataFromMatlab<int>(prhs[0]);
+  switch (mode) {
     case MODE_SETUP:
-      if(nrhs > 1 && mxIsSparse(prhs[1]))
-      {
-        if(nrhs > 3 && mxIsSparse(prhs[2]) && mxIsSparse(prhs[3]))
+      if (nrhs > 1 && mxIsSparse(prhs[1])) {
+        if (nrhs > 3 && mxIsSparse(prhs[2]) && mxIsSparse(prhs[3]))
           rv = MODE_ERROR;
         else
           rv = MODE_SETUP;
-      }
-      else
-      {
+      } else {
         mexErrMsgTxt("Error: Invalid input for setup\n");
       }
       break;
     case MODE_SOLVE:
-      //problem ID and matrix or rhs must be numeric
-      if(nrhs >= 2 && mxIsNumeric(prhs[1]) && mxIsNumeric(prhs[2]))
+      // problem ID and matrix or rhs must be numeric
+      if (nrhs >= 2 && mxIsNumeric(prhs[1]) && mxIsNumeric(prhs[2]))
         rv = MODE_SOLVE;
       else
         mexErrMsgTxt("Error: Invalid input for solve\n");
       break;
     case MODE_APPLY:
-      //problem ID and RHS must be numeric
-      if(nrhs == 3 && mxIsNumeric(prhs[1]) && mxIsNumeric(prhs[2]))
+      // problem ID and RHS must be numeric
+      if (nrhs == 3 && mxIsNumeric(prhs[1]) && mxIsNumeric(prhs[2]))
         rv = MODE_APPLY;
       else
         mexErrMsgTxt("Error: Invalid input for apply\n");
       break;
     case MODE_CLEANUP:
-      if(nrhs == 1 || nrhs == 2)
+      if (nrhs == 1 || nrhs == 2)
         rv = MODE_CLEANUP;
       else
         mexErrMsgTxt("Error: Extraneous args for cleanup\n");
       break;
     case MODE_STATUS:
-      if(nrhs == 1 || nrhs == 2)
+      if (nrhs == 1 || nrhs == 2)
         rv = MODE_STATUS;
       else
         mexErrMsgTxt("Error: Extraneous args for status\n");
       break;
     case MODE_AGGREGATE:
-      if(nrhs > 1 && mxIsSparse(prhs[1]))
-        //Uncomment the next line and remove one after when implementing aggregate mode
-        //rv = MODE_AGGREGATE;
+      if (nrhs > 1 && mxIsSparse(prhs[1]))
+        // Uncomment the next line and remove one after when implementing aggregate mode
+        // rv = MODE_AGGREGATE;
         rv = MODE_ERROR;
       else
         mexErrMsgTxt("Error: Invalid input for aggregate\n");
       break;
     case MODE_GET:
-      if(nrhs < 4 || nrhs > 5)
+      if (nrhs < 4 || nrhs > 5)
         mexErrMsgTxt("Error: Wrong number of args for get\n");
       else
         rv = MODE_GET;
       break;
     default:
-      printf("Mode number = %d\n", (int) mode);
+      printf("Mode number = %d\n", (int)mode);
       mexErrMsgTxt("Error: Invalid input mode\n");
-    };
+  };
   return rv;
 }
 /*end sanity_check*/
 
-void csc_print(int n, int* rowind, int* colptr, double* vals)
-{
+void csc_print(int n, int* rowind, int* colptr, double* vals) {
   int i, j;
-  for(i = 0; i < n; i++)
-  {
-    for(j = colptr[i]; j < colptr[i + 1]; j++)
-    {
+  for (i = 0; i < n; i++) {
+    for (j = colptr[i]; j < colptr[i + 1]; j++) {
       mexPrintf("%d %d %20.16e\n", rowind[j], i, vals[j]);
     }
   }
 }
 
-void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *prhs)
-{
-  //List shouldn't be NULL but if it is, initialize here
-  if(List.is_null())
-  {
+void parse_list_item(RCP<ParameterList> List, char* option_name, const mxArray* prhs) {
+  // List shouldn't be NULL but if it is, initialize here
+  if (List.is_null()) {
     List = rcp(new ParameterList);
   }
   mxClassID cid;
   int i, M, N, *opt_int;
-  char *opt_char;
-  double *opt_float;
+  char* opt_char;
+  double* opt_float;
   string opt_str;
   RCP<ParameterList> sublist = rcp(new ParameterList);
   mxArray *cell1, *cell2;
   /* Pull relevant info the the option value */
   cid = mxGetClassID(prhs);
-  M = mxGetM(prhs);
-  N = mxGetN(prhs);
+  M   = mxGetM(prhs);
+  N   = mxGetN(prhs);
   /* Add to the Teuchos list */
 
   // extract potential typeStr. The code is based on the assumption that
@@ -1211,24 +1062,22 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
   // between "map" type (representing a Xpetra::Map) and multivector (default)
   vector<string> typestring = tokenizeList(option_name);
   std::transform(typestring[0].begin(), typestring[0].end(), typestring[0].begin(), ::tolower);
-  size_t WordStart = typestring[0].find_first_not_of(' ');
-  size_t WordEnd   = typestring[0].find(' ', WordStart);
+  size_t WordStart    = typestring[0].find_first_not_of(' ');
+  size_t WordEnd      = typestring[0].find(' ', WordStart);
   std::string typeStr = typestring[0].substr(WordStart, WordEnd - WordStart);
 
   /////
 
-  switch(cid)
-  {
+  switch (cid) {
     case mxCHAR_CLASS:
       // String
       opt_char = mxArrayToString(prhs);
-      opt_str = opt_char;
+      opt_str  = opt_char;
       List->set(option_name, opt_str);
-      if(strcmp(option_name, MUEMEX_INTERFACE) == 0)
-      {
-        if(strcmp(opt_str.c_str(), "epetra") == 0)
+      if (strcmp(option_name, MUEMEX_INTERFACE) == 0) {
+        if (strcmp(opt_str.c_str(), "epetra") == 0)
           useEpetra = true;
-        else if(strcmp(opt_str.c_str(), "tpetra") == 0)
+        else if (strcmp(opt_str.c_str(), "tpetra") == 0)
           useEpetra = false;
       }
       mxFree(opt_char);
@@ -1236,23 +1085,17 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
     case mxDOUBLE_CLASS:
     case mxSINGLE_CLASS:
       // Single or double, real or complex
-      if(mxIsComplex(prhs))
-      {
+      if (mxIsComplex(prhs)) {
 #ifndef HAVE_COMPLEX_SCALARS
-        opt_float = mxGetPr(prhs);
+        opt_float              = mxGetPr(prhs);
         double* opt_float_imag = mxGetPi(prhs);
-        //assuming user wants std::complex<double> here...
-        if(M == 1 && N == 1)
-        {
+        // assuming user wants std::complex<double> here...
+        if (M == 1 && N == 1) {
           List->set(option_name, complex_t(*opt_float, *opt_float_imag));
-        }
-        else if(M == 0 || N == 0)
-        {
-          List->set(option_name, (complex_t*) NULL);
-        }
-        else
-        {
-          if(mxIsSparse(prhs))
+        } else if (M == 0 || N == 0) {
+          List->set(option_name, (complex_t*)NULL);
+        } else {
+          if (mxIsSparse(prhs))
             List->set(option_name, loadDataFromMatlab<RCP<Xpetra_Matrix_complex>>(prhs));
           else
             List->set(option_name, loadDataFromMatlab<RCP<Xpetra_MultiVector_complex>>(prhs));
@@ -1261,30 +1104,21 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
         std::cerr << "Error: cannot load argument \"" << option_name << "\" because complex is not instantiated in this build.\n";
         throw std::invalid_argument("Complex not supported");
 #endif
-      }
-      else
-      {
+      } else {
         opt_float = mxGetPr(prhs);
-        if(M == 1 && N == 1 && MMISINT(opt_float[0]))
-        {
-          List->set(option_name, (int) opt_float[0]);
-        }
-        else if(M == 1 && N == 1)
-        {
+        if (M == 1 && N == 1 && MMISINT(opt_float[0])) {
+          List->set(option_name, (int)opt_float[0]);
+        } else if (M == 1 && N == 1) {
           List->set(option_name, opt_float[0]);
-        }
-        else if(M == 0 || N == 0)
-        {
-          List->set(option_name, (double*) NULL);
-        }
-        else
-        {
-          if(mxIsSparse(prhs))
+        } else if (M == 0 || N == 0) {
+          List->set(option_name, (double*)NULL);
+        } else {
+          if (mxIsSparse(prhs))
             List->set(option_name, loadDataFromMatlab<RCP<Xpetra_Matrix_double>>(prhs));
           else {
-            if(typeStr == "map") // data stored as Xpetra::Map type
+            if (typeStr == "map")  // data stored as Xpetra::Map type
               List->set(option_name, loadDataFromMatlab<RCP<Xpetra_map>>(prhs));
-            else // data stored as MultiVector
+            else  // data stored as MultiVector
               List->set(option_name, loadDataFromMatlab<RCP<Xpetra_MultiVector_double>>(prhs));
           }
         }
@@ -1292,11 +1126,11 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
       break;
     case mxLOGICAL_CLASS:
       // Bool
-      if(M == 1 && N == 1)
+      if (M == 1 && N == 1)
         List->set(option_name, mxIsLogicalScalarTrue(prhs));
       else
         List->set(option_name, mxGetLogicals(prhs));
-      //NTS: The else probably doesn't work.
+      // NTS: The else probably doesn't work.
       break;
     case mxINT8_CLASS:
     case mxUINT8_CLASS:
@@ -1305,8 +1139,8 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
     case mxINT32_CLASS:
     case mxUINT32_CLASS:
       // Integer
-      opt_int = (int*) mxGetData(prhs);
-      if(M == 1 && N == 1)
+      opt_int = (int*)mxGetData(prhs);
+      if (M == 1 && N == 1)
         List->set(option_name, *opt_int);
 #ifdef HAVE_MUELU_INTREPID2
       else if (strcmp(option_name, "pcoarsen: element to node map") == 0)
@@ -1321,11 +1155,10 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
     case mxCELL_CLASS:
       // Interpret a cell list as a nested teuchos list.
       // NTS: Assuming that it's a 1D row ordered array
-      for(i = 0; i < N; i += 2)
-      {
+      for (i = 0; i < N; i += 2) {
         cell1 = mxGetCell(prhs, i);
         cell2 = mxGetCell(prhs, i + 1);
-        if(!mxIsChar(cell1))
+        if (!mxIsChar(cell1))
           mexErrMsgTxt("Error: Input options are not in ['parameter',value] format!\n");
         opt_char = mxArrayToString(cell1);
         parse_list_item(sublist, opt_char, cell2);
@@ -1338,29 +1171,20 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
     case mxFUNCTION_CLASS:
     case mxUNKNOWN_CLASS:
     case mxSTRUCT_CLASS:
-      //Currently Graph and Aggregates are stored as structures
-      if(isValidMatlabAggregates(prhs))
-      {
-        try
-        {
+      // Currently Graph and Aggregates are stored as structures
+      if (isValidMatlabAggregates(prhs)) {
+        try {
           List->set(option_name, loadDataFromMatlab<RCP<MAggregates>>(prhs));
           break;
-        }
-        catch(exception& e)
-        {
+        } catch (exception& e) {
           cout << e.what();
           throw runtime_error("Parsing aggregates in parameter list failed.");
         }
-      }
-      else if(isValidMatlabGraph(prhs))
-      {
-        try
-        {
+      } else if (isValidMatlabGraph(prhs)) {
+        try {
           List->set(option_name, loadDataFromMatlab<RCP<MGraph>>(prhs));
           break;
-        }
-        catch(exception& e)
-        {
+        } catch (exception& e) {
           cout << e.what();
           throw runtime_error("Parsing graph in parameter list failed.");
         }
@@ -1368,7 +1192,7 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
     default:
       mexPrintf("Error parsing input option: %s [type=%d]\n", option_name, cid);
       mexErrMsgTxt("Error: An input option is invalid!\n");
-    };
+  };
 }
 
 /**************************************************************/
@@ -1381,13 +1205,11 @@ void parse_list_item(RCP<ParameterList> List, char *option_name, const mxArray *
    prhs    - The problem inputs [I]
    Return value: Teuchos list containing all parameters passed in by the user.
 */
-RCP<ParameterList> build_teuchos_list(int nrhs, const mxArray *prhs[])
-{
+RCP<ParameterList> build_teuchos_list(int nrhs, const mxArray* prhs[]) {
   RCP<ParameterList> TPL = rcp(new ParameterList);
   char* option_name;
-  for(int i = 0; i < nrhs; i += 2)
-  {
-    if(i == nrhs - 1 || !mxIsChar(prhs[i]))
+  for (int i = 0; i < nrhs; i += 2) {
+    if (i == nrhs - 1 || !mxIsChar(prhs[i]))
       mexErrMsgTxt("Error: Input options are not in ['parameter',value] format!\n");
     /* What option are we setting? */
     option_name = mxArrayToString(prhs[i]);
@@ -1401,22 +1223,20 @@ RCP<ParameterList> build_teuchos_list(int nrhs, const mxArray *prhs[])
 }
 /*end build_teuchos_list*/
 
-} //end MueLu namespace (mexFunction must be in global namespace)
+}  // namespace MueLu
 
-using namespace MueLu; //...but give mexFunction access to all MueLu members defined above
+using namespace MueLu;  //...but give mexFunction access to all MueLu members defined above
 
-void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
-{
-  //Lazily initialize Tpetra
-  if(!Tpetra::isInitialized())
-  {
-    int argc = 0;
+void mexFunction(int nlhs, mxArray* plhs[], int nrhs, const mxArray* prhs[]) {
+  // Lazily initialize Tpetra
+  if (!Tpetra::isInitialized()) {
+    int argc    = 0;
     char** argv = NULL;
     Tpetra::initialize(&argc, &argv);
   }
   double* id;
   int rv;
-  //Arrays representing vectors
+  // Arrays representing vectors
   string intf;
   RCP<ParameterList> List;
   MODE_TYPE mode;
@@ -1428,207 +1248,173 @@ void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
 #ifdef FORCE_FACTORIES_TO_COMPILE
   {
     // debug
-    MueLu::TwoLevelMatlabFactory<double,int,int> f1;
-    MueLu::SingleLevelMatlabFactory<double,int,int> f2;
+    MueLu::TwoLevelMatlabFactory<double, int, int> f1;
+    MueLu::SingleLevelMatlabFactory<double, int, int> f2;
   }
 #endif
 
-  switch(mode)
-  {
-    case MODE_SETUP:
-    {
-      try
-      {
-        double oc = 0;
+  switch (mode) {
+    case MODE_SETUP: {
+      try {
+        double oc       = 0;
         bool haveCoords = false;
-        if(nrhs == 2)
+        if (nrhs == 2)
           List = rcp(new ParameterList);
-        else if(nrhs == 3)
-        {
-          if(!mxIsNumeric(prhs[2]) || mxIsSparse(prhs[2]) || mxIsComplex(prhs[2]))
+        else if (nrhs == 3) {
+          if (!mxIsNumeric(prhs[2]) || mxIsSparse(prhs[2]) || mxIsComplex(prhs[2]))
             throw runtime_error("Expected real-valued, dense Coordinates array as the third muelu argument");
-          else
-          {
+          else {
             haveCoords = true;
-            List = rcp(new ParameterList);
+            List       = rcp(new ParameterList);
           }
-        }
-        else
-        {
-          if(mxIsNumeric(prhs[2]) && !mxIsSparse(prhs[2]) && !mxIsComplex(prhs[2]))
-          {
-            List = build_teuchos_list(nrhs - 3, &(prhs[3]));
+        } else {
+          if (mxIsNumeric(prhs[2]) && !mxIsSparse(prhs[2]) && !mxIsComplex(prhs[2])) {
+            List       = build_teuchos_list(nrhs - 3, &(prhs[3]));
             haveCoords = true;
-          }
-          else
-          {
-            //assume that the parameters start at third argument if it doesn't seem like coords are there
+          } else {
+            // assume that the parameters start at third argument if it doesn't seem like coords are there
             List = build_teuchos_list(nrhs - 2, &(prhs[2]));
           }
         }
-        //Combine xml and easy parameter lists
-        if(List->isParameter("xml parameter file"))
-        {
+        // Combine xml and easy parameter lists
+        if (List->isParameter("xml parameter file")) {
           RCP<ParameterList> xmlParams = Teuchos::getParametersFromXmlFile(List->get<string>("xml parameter file"));
           List->remove("xml parameter file");
           List->setParametersNotAlreadySet(*xmlParams);
         }
-        if(mxIsComplex(prhs[1]))
-        {
-          //Abort if input is complex but complex isn't supported
+        if (mxIsComplex(prhs[1])) {
+          // Abort if input is complex but complex isn't supported
 #ifndef HAVE_COMPLEX_SCALARS
           mexPrintf("Error: Complex scalars unsupported by this build of Trilinos.\n");
           throw runtime_error("Complex scalars not supported.");
 #endif
         }
         intf = List->get(MUEMEX_INTERFACE, "tpetra");
-        List->remove(MUEMEX_INTERFACE);             //no longer need this parameter
-        if(intf == "epetra")
-        {
-          if(mxIsComplex(prhs[1]))
-          {
+        List->remove(MUEMEX_INTERFACE);  // no longer need this parameter
+        if (intf == "epetra") {
+          if (mxIsComplex(prhs[1])) {
             mexPrintf("Error: Attempting to use complex-valued matrix with Epetra, which is unsupported.\n");
             mexPrintf("Use Tpetra with complex matrices instead.\n");
             throw runtime_error("Tried to use complex matrix with Epetra");
           }
           RCP<EpetraSystem> dp = rcp(new EpetraSystem());
-          dp->List = List;
-          dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*) NULL);
+          dp->List             = List;
+          dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*)NULL);
           oc = dp->operatorComplexity;
-          D = rcp_implicit_cast<MuemexSystem>(dp);
-        }
-        else if(intf == "tpetra")
-        {
-          //infer scalar type from prhs (can be double or complex<double>)
-          if(mxIsComplex(prhs[1]))
-          {
+          D  = rcp_implicit_cast<MuemexSystem>(dp);
+        } else if (intf == "tpetra") {
+          // infer scalar type from prhs (can be double or complex<double>)
+          if (mxIsComplex(prhs[1])) {
 #ifdef HAVE_COMPLEX_SCALARS
             RCP<TpetraSystem<complex_t>> dp = rcp(new TpetraSystem<complex_t>());
-            dp->List = List;
-            dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*) NULL);
+            dp->List                        = List;
+            dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*)NULL);
             oc = dp->operatorComplexity;
-            D = rcp_implicit_cast<MuemexSystem>(dp);
+            D  = rcp_implicit_cast<MuemexSystem>(dp);
 #else
             throw runtime_error("Complex scalars not supported.");
 #endif
-          }
-          else
-          {
+          } else {
             RCP<TpetraSystem<double>> dp = rcp(new TpetraSystem<double>());
-            dp->List = List;
-            dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*) NULL);
+            dp->List                     = List;
+            dp->setup(prhs[1], haveCoords, haveCoords ? prhs[2] : (mxArray*)NULL);
             oc = dp->operatorComplexity;
-            D = rcp_implicit_cast<MuemexSystem>(dp);
+            D  = rcp_implicit_cast<MuemexSystem>(dp);
           }
         }
         rv = MuemexSystemList::add(D);
         mexPrintf("Set up problem #%d\n", rv);
-        if(nlhs > 0)
-        {
-          plhs[0] = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
-          *((int*) mxGetData(plhs[0])) = rv;
-          //output OC also
-          if(nlhs > 1)
+        if (nlhs > 0) {
+          plhs[0]                     = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
+          *((int*)mxGetData(plhs[0])) = rv;
+          // output OC also
+          if (nlhs > 1)
             plhs[1] = mxCreateDoubleScalar(oc);
         }
         mexLock();
-      }
-      catch(exception& e)
-      {
+      } catch (exception& e) {
         mexPrintf("An error occurred during setup routine:\n");
         cout << e.what() << endl;
-        if(nlhs > 0)
-        {
-          plhs[0] = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
-          *((int*) mxGetData(plhs[0])) = -1;
+        if (nlhs > 0) {
+          plhs[0]                     = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
+          *((int*)mxGetData(plhs[0])) = -1;
         }
-        if(nlhs > 1)
+        if (nlhs > 1)
           plhs[1] = mxCreateDoubleScalar(0);
       }
       break;
     }
-    case MODE_SOLVE:
-    {
+    case MODE_SOLVE: {
       int iters;
-      try
-      {
+      try {
         bool reuse;
-        //MODE_SOLVE, probID, matrix, bVec, params OR
-        //MODE_SOLVE, probID, bVec, params
-        //prhs[0] holds the MODE_SOLVE enum value
-        //if reusing, either nrhs == 3 or prhs[3] is not a string
+        // MODE_SOLVE, probID, matrix, bVec, params OR
+        // MODE_SOLVE, probID, bVec, params
+        // prhs[0] holds the MODE_SOLVE enum value
+        // if reusing, either nrhs == 3 or prhs[3] is not a string
         //(because bVec can't be a string)
-        if(MuemexSystemList::size() == 0)
+        if (MuemexSystemList::size() == 0)
           throw runtime_error("No linear systems are set up.");
-        if(nrhs == 3 || (nrhs > 3 && mxGetClassID(prhs[3]) == mxCHAR_CLASS))
-        {
-          //No matrix supplied as argument, use one from setup
-          if(nrhs > 3)
+        if (nrhs == 3 || (nrhs > 3 && mxGetClassID(prhs[3]) == mxCHAR_CLASS)) {
+          // No matrix supplied as argument, use one from setup
+          if (nrhs > 3)
             List = build_teuchos_list(nrhs - 3, &prhs[3]);
           else
             List = rcp(new ParameterList);
           reuse = true;
-        }
-        else
-        {
-          if(nrhs > 4)
+        } else {
+          if (nrhs > 4)
             List = build_teuchos_list(nrhs - 4, &prhs[4]);
           else
             List = rcp(new ParameterList);
           reuse = false;
         }
-        if(List->isType<string>("Output Style"))
-        {
+        if (List->isType<string>("Output Style")) {
           int type = strToOutputStyle(List->get("Output Style", "Belos::Brief").c_str());
           List->remove("Output Style");
-          //Reset the ParameterList entry to be of type int instead of string
+          // Reset the ParameterList entry to be of type int instead of string
           List->set("Output Style", type);
         }
-        //Convert Belos msg type string to int in List
-        //Note: if the parameter value is already an int, don't touch it.
-        if(List->isType<string>("Verbosity"))
-        {
-          //Errors + Warnings is already the default Belos verbosity setting
+        // Convert Belos msg type string to int in List
+        // Note: if the parameter value is already an int, don't touch it.
+        if (List->isType<string>("Verbosity")) {
+          // Errors + Warnings is already the default Belos verbosity setting
           int verb = getBelosVerbosity(List->get("Verbosity", "Belos::Errors + Belos::Warnings").c_str());
           List->remove("Verbosity");
           List->set("Verbosity", verb);
         }
-        int probID = loadDataFromMatlab<int>(prhs[1]);
+        int probID           = loadDataFromMatlab<int>(prhs[1]);
         RCP<MuemexSystem> dp = MuemexSystemList::find(probID);
-        if(dp.is_null())
+        if (dp.is_null())
           throw runtime_error("Problem handle not allocated.");
-        //get pointer to MATLAB array that will be "B" or "rhs" multivector
+        // get pointer to MATLAB array that will be "B" or "rhs" multivector
         const mxArray* rhs = reuse ? prhs[2] : prhs[3];
-        switch(dp->type)
-        {
-          case EPETRA:
-          {
+        switch (dp->type) {
+          case EPETRA: {
             RCP<EpetraSystem> esys = rcp_static_cast<EpetraSystem, MuemexSystem>(dp);
             RCP<Epetra_CrsMatrix> matrix;
-            if(reuse)
+            if (reuse)
               matrix = esys->GetMatrix();
             else
               matrix = loadDataFromMatlab<RCP<Epetra_CrsMatrix>>(prhs[2]);
             plhs[0] = esys->solve(List, matrix, rhs, iters);
             break;
           }
-          case TPETRA:
-          {
+          case TPETRA: {
             RCP<TpetraSystem<double>> tsys = rcp_static_cast<TpetraSystem<double>, MuemexSystem>(dp);
             RCP<Tpetra_CrsMatrix_double> matrix;
-            if(reuse)
+            if (reuse)
               matrix = tsys->GetMatrix();
             else
               matrix = loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double>>(prhs[2]);
             plhs[0] = tsys->solve(List, matrix, rhs, iters);
             break;
           }
-          case TPETRA_COMPLEX:
-          {
+          case TPETRA_COMPLEX: {
 #ifdef HAVE_COMPLEX_SCALARS
             RCP<TpetraSystem<complex_t>> tsys = rcp_static_cast<TpetraSystem<complex_t>, MuemexSystem>(dp);
             RCP<Tpetra_CrsMatrix_complex> matrix;
-            if(reuse)
+            if (reuse)
               matrix = tsys->GetMatrix();
             else
               matrix = loadDataFromMatlab<RCP<Tpetra_CrsMatrix_complex>>(prhs[2]);
@@ -1640,49 +1426,41 @@ void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
 #endif
           }
         }
-        if(nlhs > 1)
+        if (nlhs > 1)
           plhs[1] = saveDataToMatlab<int>(iters);
-      }
-      catch(exception& e)
-      {
+      } catch (exception& e) {
         mexPrintf("An error occurred during the solve routine:\n");
         cout << e.what() << endl;
       }
       break;
     }
-    case MODE_APPLY:
-    {
-      try
-      {
-        //MODE_APPLY, probID, rhsVec
-        //prhs[0] holds the MODE_APPLY enum value
-        if(MuemexSystemList::size() == 0)
+    case MODE_APPLY: {
+      try {
+        // MODE_APPLY, probID, rhsVec
+        // prhs[0] holds the MODE_APPLY enum value
+        if (MuemexSystemList::size() == 0)
           throw runtime_error("No linear systems are set up.");
-        int probID = loadDataFromMatlab<int>(prhs[1]);
+        int probID           = loadDataFromMatlab<int>(prhs[1]);
         RCP<MuemexSystem> dp = MuemexSystemList::find(probID);
-        if(dp.is_null())
+        if (dp.is_null())
           throw runtime_error("Problem handle not allocated.");
-        //get pointer to MATLAB array that will be "B" or "rhs" multivector
+        // get pointer to MATLAB array that will be "B" or "rhs" multivector
         const mxArray* rhs = prhs[2];
-        switch(dp->type)
-        {
-          case EPETRA:
-          {
+        switch (dp->type) {
+          case EPETRA: {
             RCP<EpetraSystem> esys = rcp_static_cast<EpetraSystem, MuemexSystem>(dp);
-            plhs[0] = esys->apply(rhs);
+            plhs[0]                = esys->apply(rhs);
             break;
           }
-          case TPETRA:
-          {
+          case TPETRA: {
             RCP<TpetraSystem<double>> tsys = rcp_static_cast<TpetraSystem<double>, MuemexSystem>(dp);
-            plhs[0] = tsys->apply(rhs);
+            plhs[0]                        = tsys->apply(rhs);
             break;
           }
-          case TPETRA_COMPLEX:
-          {
+          case TPETRA_COMPLEX: {
 #ifdef HAVE_COMPLEX_SCALARS
             RCP<TpetraSystem<complex_t>> tsys = rcp_static_cast<TpetraSystem<complex_t>, MuemexSystem>(dp);
-            plhs[0] = tsys->apply(rhs);
+            plhs[0]                           = tsys->apply(rhs);
             break;
 #else
             std::cerr << "Cannot solve complex-valued system because complex is not enabled in this build.\n";
@@ -1690,122 +1468,99 @@ void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
 #endif
           }
         }
-      }
-      catch(exception& e)
-      {
+      } catch (exception& e) {
         mexPrintf("An error occurred during the apply routine:\n");
         cout << e.what() << endl;
       }
       break;
     }
-    case MODE_CLEANUP:
-    {
-      try
-      {
+    case MODE_CLEANUP: {
+      try {
         mexPrintf("MueMex in cleanup mode.\n");
-        if(MuemexSystemList::size() > 0 && nrhs == 1)
-        {
+        if (MuemexSystemList::size() > 0 && nrhs == 1) {
           /* Cleanup all problems */
-          for(int i = 0; i < MuemexSystemList::size(); i++)
+          for (int i = 0; i < MuemexSystemList::size(); i++)
             mexUnlock();
           MuemexSystemList::clearAll();
           rv = 1;
-        }
-        else if(MuemexSystemList::size() > 0 && nrhs == 2)
-        {
+        } else if (MuemexSystemList::size() > 0 && nrhs == 2) {
           /* Cleanup one problem */
           int probID = loadDataFromMatlab<int>(prhs[1]);
           mexPrintf("Cleaning up problem #%d\n", probID);
           rv = MuemexSystemList::remove(probID);
-          if(rv)
+          if (rv)
             mexUnlock();
         } /*end elseif*/
-        else
-        {
+        else {
           rv = 0;
         }
         /* Set return value */
         plhs[0] = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
-        id = (double*) mxGetData(plhs[0]);
-        *id = double(rv);
-      }
-      catch(exception& e)
-      {
+        id      = (double*)mxGetData(plhs[0]);
+        *id     = double(rv);
+      } catch (exception& e) {
         mexPrintf("An error occurred during the cleanup routine:\n");
         cout << e.what() << endl;
-        if(nlhs > 0)
-        {
-          plhs[0] = mxCreateNumericMatrix(1,1, mxINT32_CLASS, mxREAL);
-          *(int*) mxGetData(plhs[0]) = 0;
+        if (nlhs > 0) {
+          plhs[0]                   = mxCreateNumericMatrix(1, 1, mxINT32_CLASS, mxREAL);
+          *(int*)mxGetData(plhs[0]) = 0;
         }
       }
       break;
     }
-    case MODE_STATUS:
-    {
-      try
-      {
-        //mexPrintf("MueMex in status checking mode.\n");
-        if(MuemexSystemList::size() > 0 && nrhs == 1)
-        {
+    case MODE_STATUS: {
+      try {
+        // mexPrintf("MueMex in status checking mode.\n");
+        if (MuemexSystemList::size() > 0 && nrhs == 1) {
           /* Status check on all problems */
           rv = MuemexSystemList::status_all();
-        }/*end if*/
-        else if(MuemexSystemList::size() > 0 && nrhs == 2)
-        {
+        } /*end if*/
+        else if (MuemexSystemList::size() > 0 && nrhs == 2) {
           /* Status check one problem */
           int probID = loadDataFromMatlab<int>(prhs[1]);
-          D = MuemexSystemList::find(probID);
-          if(D.is_null())
+          D          = MuemexSystemList::find(probID);
+          if (D.is_null())
             throw runtime_error("Error: Problem handle not allocated.\n");
           rv = D->status();
-        }/*end elseif*/
+        } /*end elseif*/
         else
           mexPrintf("No problems set up.\n");
-        if(nlhs > 0)
-        {
+        if (nlhs > 0) {
           int outVal = 0;
-          plhs[0] = saveDataToMatlab<int>(outVal);
+          plhs[0]    = saveDataToMatlab<int>(outVal);
         }
-      }
-      catch(exception& e)
-      {
+      } catch (exception& e) {
         mexPrintf("An error occurred during the status routine:\n");
         cout << e.what() << endl;
         int outVal = -1;
-        if(nlhs > 0)
+        if (nlhs > 0)
           plhs[0] = saveDataToMatlab<int>(outVal);
       }
       break;
     }
-    case MODE_GET:
-    {
-      try
-      {
-        int probID = loadDataFromMatlab<int>(prhs[1]);
-        int levelID = loadDataFromMatlab<int>(prhs[2]);
-        char* dataName = mxArrayToString(prhs[3]);
+    case MODE_GET: {
+      try {
+        int probID            = loadDataFromMatlab<int>(prhs[1]);
+        int levelID           = loadDataFromMatlab<int>(prhs[2]);
+        char* dataName        = mxArrayToString(prhs[3]);
         MuemexType outputType = INT;
-        RCP<MuemexSystem> dp = MuemexSystemList::find(probID);
-        if(dp.is_null())
-        {
+        RCP<MuemexSystem> dp  = MuemexSystemList::find(probID);
+        if (dp.is_null()) {
           throw runtime_error("Problem handle not allocated.");
         }
 
-        //See if typeHint was given
+        // See if typeHint was given
         char* paramTypeName = NULL;
-        if(nrhs > 4) {
+        if (nrhs > 4) {
           paramTypeName = mxArrayToString(prhs[4]);
           mexPrintf("paramTypeName %s", paramTypeName);
         }
         bool complexFlag = dp->type == TPETRA_COMPLEX;
-        //std::cout << "before strToDataType dataName=" << dataName << " paramTypeName " << paramTypeName << std::endl;
+        // std::cout << "before strToDataType dataName=" << dataName << " paramTypeName " << paramTypeName << std::endl;
 
         outputType = strToDataType(dataName, paramTypeName, complexFlag);
-        plhs[0] = dp->getHierarchyData(string(dataName), outputType, levelID);
-      }
-      catch(exception& e)
-      {
+        plhs[0]    = dp->getHierarchyData(string(dataName), outputType, levelID);
+      } catch (exception& e) {
         mexPrintf("An error occurred during the get routine:\n");
         cout << e.what() << endl;
         plhs[0] = mxCreateDoubleScalar(0);
@@ -1818,5 +1573,5 @@ void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
     case MODE_AGGREGATE:
     default:
       mexPrintf("Mode not supported yet.");
-    }
+  }
 }
diff --git a/packages/muelu/matlab/src/MueLu_MatlabSmoother.cpp b/packages/muelu/matlab/src/MueLu_MatlabSmoother.cpp
index e6490ac3fadb..efaec72740ab 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabSmoother.cpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabSmoother.cpp
@@ -44,7 +44,6 @@
 //
 // @HEADER
 
-
 #include "MueLu_ExplicitInstantiation.hpp"
 
 #include "MueLu_MatlabSmoother_def.hpp"
@@ -52,8 +51,8 @@
 #include "TpetraCore_ETIHelperMacros.h"
 
 #ifdef HAVE_MUELU_MATLAB
-#define MUELU_LOCAL_INSTANT(S,LO,GO,N) \
-        template class MueLu::MatlabSmoother<S,LO,GO,N>;
+#define MUELU_LOCAL_INSTANT(S, LO, GO, N) \
+  template class MueLu::MatlabSmoother<S, LO, GO, N>;
 
 TPETRA_ETI_MANGLING_TYPEDEFS()
 
diff --git a/packages/muelu/matlab/src/MueLu_MatlabSmoother_decl.hpp b/packages/muelu/matlab/src/MueLu_MatlabSmoother_decl.hpp
index b31b7823d731..6dd3d6b79c30 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabSmoother_decl.hpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabSmoother_decl.hpp
@@ -58,155 +58,150 @@
 
 namespace MueLu {
 
-  /*!
-    @class MatlabSmoother
-    @ingroup MueMexClasses
-    @brief Class that encapsulates Matlab smoothers.
-
-    //   This class creates an Matlab preconditioner factory. The factory creates a smoother based on the
-    //   type and ParameterList passed into the constructor. See the constructor for more information.
-    */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class MatlabSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>
-  {
+/*!
+  @class MatlabSmoother
+  @ingroup MueMexClasses
+  @brief Class that encapsulates Matlab smoothers.
+
+  //   This class creates an Matlab preconditioner factory. The factory creates a smoother based on the
+  //   type and ParameterList passed into the constructor. See the constructor for more information.
+  */
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class MatlabSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_MATLABSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors / destructors
+  //@{
+  // TODO: update doc for Matlab.
+  /*! @brief Constructor
 
-    //! @name Constructors / destructors
-    //@{
-    //TODO: update doc for Matlab.
-    /*! @brief Constructor
+    ADD DOCUMENTATION HERE
 
-      ADD DOCUMENTATION HERE
-
-    */
+  */
 
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class MatlabSmoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class MatlabSmoother;
 #endif
 
-    MatlabSmoother(const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
+  MatlabSmoother(const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    //! Destructor
-    virtual ~MatlabSmoother() { }
+  //! Destructor
+  virtual ~MatlabSmoother() {}
 
-    //@}
+  //@}
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Computational methods.
-    //@{
+  //! @name Computational methods.
+  //@{
 
-    /*! @brief Set up the smoother.
+  /*! @brief Set up the smoother.
 
-    This creates the underlying Matlab smoother object, copies any parameter list options
-    supplied to the constructor to the Matlab object, and computes the preconditioner.
+  This creates the underlying Matlab smoother object, copies any parameter list options
+  supplied to the constructor to the Matlab object, and computes the preconditioner.
 
-    TODO The eigenvalue estimate should come from A_, not the Matlab parameter list.
-    */
-    void Setup(Level &currentLevel);
+  TODO The eigenvalue estimate should come from A_, not the Matlab parameter list.
+  */
+  void Setup(Level& currentLevel);
 
-    /*! @brief Apply the preconditioner.
+  /*! @brief Apply the preconditioner.
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Matlab implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Matlab implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //@}
+  //@}
 
-    //! @name Utilities
-    //@{
+  //! @name Utilities
+  //@{
 
-    RCP<SmootherPrototype> Copy() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //@}
+  //@}
 
-    //! Clone the smoother to a different node type
-    template<typename Node2>
-    RCP<MueLu::MatlabSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node2> >
-    clone(const RCP<Node2>& node2, const Teuchos::RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node2> >& A_newnode) const;
+  //! Clone the smoother to a different node type
+  template <typename Node2>
+  RCP<MueLu::MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node2> >
+  clone(const RCP<Node2>& node2, const Teuchos::RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node2> >& A_newnode) const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    size_t getNodeSmootherComplexity() const {return Teuchos::OrdinalTraits<size_t>::invalid();}
+  size_t getNodeSmootherComplexity() const { return Teuchos::OrdinalTraits<size_t>::invalid(); }
 
+  //@}
 
-    //@}
+ private:
+  //! List of arguments to the MATLAB setup function besides "A", in order
+  mutable std::string needsSetup_;
 
-  private:
+  //! Amount of solve data (besides A, LHS & RHS)
+  size_t solveDataSize_;
 
-    //! List of arguments to the MATLAB setup function besides "A", in order
-    mutable std::string needsSetup_;
+  //! List of data generated by setup which will be sent to solve after "A", "LHS" and "RHS"
+  std::vector<Teuchos::RCP<MuemexArg> > solveData_;
 
-    //! Amount of solve data (besides A, LHS & RHS)
-    size_t solveDataSize_;
+  //! Matlab setup function
+  std::string setupFunction_;
 
-    //! List of data generated by setup which will be sent to solve after "A", "LHS" and "RHS"
-    std::vector<Teuchos::RCP<MuemexArg> > solveData_;
-    
-    //! Matlab setup function
-    std::string setupFunction_;
+  //! Matlab solve function
+  std::string solveFunction_;
 
-    //! Matlab solve function
-    std::string solveFunction_;
+  //! Matrix, (maybe) used in apply
+  mutable RCP<Matrix> A_;
 
-    //! Matrix, (maybe) used in apply
-    mutable RCP<Matrix> A_;
-    
-  }; // class MatlabSmoother
+};  // class MatlabSmoother
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  template<typename Node2>
-  RCP<MueLu::MatlabSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node2> >
-  MatlabSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>::clone(const RCP<Node2>& node2, const RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node2> >& A_newnode) const {
-    const ParameterList& paramList = this->GetParameterList();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <typename Node2>
+RCP<MueLu::MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node2> >
+MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::clone(const RCP<Node2>& node2, const RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node2> >& A_newnode) const {
+  const ParameterList& paramList = this->GetParameterList();
 
-    RCP<MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node2> > cloneSmoother =
+  RCP<MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node2> > cloneSmoother =
       rcp(new MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node2>(paramList));
 
-    cloneSmoother->needsSetup_    = needsSetup_;
-    cloneSmoother->setupFunction_ = setupFunction_;
-    cloneSmoother->solveFunction_  = solveFunction_;
-    cloneSmoother->A_             = A_;
-    
-    for(size_t i=0; i< solveData_.size(); i++)
-      cloneSmoother->solveData_->push_back(solveData_[i]);
-    cloneSmoother->SetParameterList(paramList);
-    cloneSmoother->IsSetup(this->IsSetup());
-    return cloneSmoother;
-  }
+  cloneSmoother->needsSetup_    = needsSetup_;
+  cloneSmoother->setupFunction_ = setupFunction_;
+  cloneSmoother->solveFunction_ = solveFunction_;
+  cloneSmoother->A_             = A_;
 
+  for (size_t i = 0; i < solveData_.size(); i++)
+    cloneSmoother->solveData_->push_back(solveData_[i]);
+  cloneSmoother->SetParameterList(paramList);
+  cloneSmoother->IsSetup(this->IsSetup());
+  return cloneSmoother;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MATLABSMOOTHER_SHORT
-#endif // HAVE_MUELU_MATLAB
-#endif // MUELU_MATLABSMOOTHER_DECL_HPP
+#endif  // HAVE_MUELU_MATLAB
+#endif  // MUELU_MATLABSMOOTHER_DECL_HPP
diff --git a/packages/muelu/matlab/src/MueLu_MatlabSmoother_def.hpp b/packages/muelu/matlab/src/MueLu_MatlabSmoother_def.hpp
index 0e5a0cbf8fa1..d5d99a41891d 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabSmoother_def.hpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabSmoother_def.hpp
@@ -51,125 +51,116 @@
 #if defined(HAVE_MUELU_MATLAB)
 #include "MueLu_Monitor.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MatlabSmoother(const Teuchos::ParameterList& paramList)
-  {
-    SetParameterList(paramList);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MatlabSmoother(const Teuchos::ParameterList& paramList) {
+  SetParameterList(paramList);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList)
-  {
-    Factory::SetParameterList(paramList);
-    ParameterList& pL = const_cast<ParameterList&>(this->GetParameterList());
-    setupFunction_ = pL.get("Setup Function","");
-    solveFunction_ = pL.get("Solve Function","");
-    solveDataSize_ = pL.get("Number of Solver Args", 0);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+  ParameterList& pL = const_cast<ParameterList&>(this->GetParameterList());
+  setupFunction_    = pL.get("Setup Function", "");
+  solveFunction_    = pL.get("Solve Function", "");
+  solveDataSize_    = pL.get("Number of Solver Args", 0);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-  {
-    using namespace std;
-    this->Input(currentLevel, "A");
-    ParameterList& pL = const_cast<ParameterList&>(this->GetParameterList());
-    needsSetup_ = pL.get<string>("Needs");
-    vector<string> needsList = tokenizeList(needsSetup_);
-    for(size_t i = 0; i < needsList.size(); i++)
-    {
-      if(!IsParamMuemexVariable(needsList[i]) && needsList[i] != "Level")
-        this->Input(currentLevel, needsList[i]);
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  using namespace std;
+  this->Input(currentLevel, "A");
+  ParameterList& pL        = const_cast<ParameterList&>(this->GetParameterList());
+  needsSetup_              = pL.get<string>("Needs");
+  vector<string> needsList = tokenizeList(needsSetup_);
+  for (size_t i = 0; i < needsList.size(); i++) {
+    if (!IsParamMuemexVariable(needsList[i]) && needsList[i] != "Level")
+      this->Input(currentLevel, needsList[i]);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel)
-  {
-    using namespace std;
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-    if (this->IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::MatlabSmoother::Setup(): Setup() has already been called";
-    vector<RCP<MuemexArg>> InputArgs = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsSetup_, currentLevel);
-    A_ = Factory::Get<RCP<Matrix>>(currentLevel, "A");
-    RCP<MuemexArg> AmatArg = rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Matrix>>(A_)));
-    //Always add A to the beginning of InputArgs
-    InputArgs.insert(InputArgs.begin(), AmatArg);
-    // Call mex function
-    if(!setupFunction_.length())
-      throw runtime_error("Invalid matlab function name");
-    solveData_= callMatlab(setupFunction_, solveDataSize_, InputArgs);
-    this->GetOStream(Statistics1) << description() << endl;
-    this->IsSetup(true); //mark the smoother as set up
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  using namespace std;
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+  if (this->IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::MatlabSmoother::Setup(): Setup() has already been called";
+  vector<RCP<MuemexArg>> InputArgs = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsSetup_, currentLevel);
+  A_                               = Factory::Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<MuemexArg> AmatArg           = rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Matrix>>(A_)));
+  // Always add A to the beginning of InputArgs
+  InputArgs.insert(InputArgs.begin(), AmatArg);
+  // Call mex function
+  if (!setupFunction_.length())
+    throw runtime_error("Invalid matlab function name");
+  solveData_ = callMatlab(setupFunction_, solveDataSize_, InputArgs);
+  this->GetOStream(Statistics1) << description() << endl;
+  this->IsSetup(true);  // mark the smoother as set up
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
-        "MueLu::MatlabSmoother::Apply(): Setup() has not been called");
-    using namespace Teuchos;
-    using namespace std;
-    if(InitialGuessIsZero)
-      X.putScalar(0.0);
-    // Push on A as first input
-    vector<RCP<MuemexArg>> InputArgs;
-    InputArgs.push_back(rcp(new MuemexData<RCP<Matrix>>(A_)));
-    // Push on LHS & RHS
-    RCP<MultiVector> Xrcp(&X, false);
-    MultiVector* BPtrNonConst = (MultiVector*) &B;
-    RCP<MultiVector> Brcp = rcp<MultiVector>(BPtrNonConst, false);
-    RCP<MuemexData<RCP<MultiVector>>> XData = rcp(new MuemexData<RCP<MultiVector>>(Xrcp));
-    RCP<MuemexData<RCP<MultiVector>>> BData = rcp(new MuemexData<RCP<MultiVector>>(Brcp));
-    InputArgs.push_back(XData);
-    InputArgs.push_back(BData);
-    for(size_t i = 0; i < solveData_.size(); i++)
-      InputArgs.push_back(solveData_[i]);
-    if(!solveFunction_.length()) throw std::runtime_error("Invalid matlab function name");
-    vector<Teuchos::RCP<MuemexArg>> mexOutput = callMatlab(solveFunction_, 1, InputArgs);
-    RCP<MuemexData<RCP<MultiVector>>> mydata = Teuchos::rcp_static_cast<MuemexData<RCP<MultiVector>>>(mexOutput[0]);
-    X = *(mydata->getData());
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
+                             "MueLu::MatlabSmoother::Apply(): Setup() has not been called");
+  using namespace Teuchos;
+  using namespace std;
+  if (InitialGuessIsZero)
+    X.putScalar(0.0);
+  // Push on A as first input
+  vector<RCP<MuemexArg>> InputArgs;
+  InputArgs.push_back(rcp(new MuemexData<RCP<Matrix>>(A_)));
+  // Push on LHS & RHS
+  RCP<MultiVector> Xrcp(&X, false);
+  MultiVector* BPtrNonConst               = (MultiVector*)&B;
+  RCP<MultiVector> Brcp                   = rcp<MultiVector>(BPtrNonConst, false);
+  RCP<MuemexData<RCP<MultiVector>>> XData = rcp(new MuemexData<RCP<MultiVector>>(Xrcp));
+  RCP<MuemexData<RCP<MultiVector>>> BData = rcp(new MuemexData<RCP<MultiVector>>(Brcp));
+  InputArgs.push_back(XData);
+  InputArgs.push_back(BData);
+  for (size_t i = 0; i < solveData_.size(); i++)
+    InputArgs.push_back(solveData_[i]);
+  if (!solveFunction_.length()) throw std::runtime_error("Invalid matlab function name");
+  vector<Teuchos::RCP<MuemexArg>> mexOutput = callMatlab(solveFunction_, 1, InputArgs);
+  RCP<MuemexData<RCP<MultiVector>>> mydata  = Teuchos::rcp_static_cast<MuemexData<RCP<MultiVector>>>(mexOutput[0]);
+  X                                         = *(mydata->getData());
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>> MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const
-  {
-    RCP<MatlabSmoother> smoother = rcp(new MatlabSmoother(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return smoother;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>> MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<MatlabSmoother> smoother = rcp(new MatlabSmoother(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return smoother;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (SmootherPrototype::IsSetup()) {
-      out << "Matlab Smoother("<<setupFunction_<<"/"<<solveFunction_<<")";
-    } else {
-      out << SmootherPrototype::description();
-    }
-    return out.str();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (SmootherPrototype::IsSetup()) {
+    out << "Matlab Smoother(" << setupFunction_ << "/" << solveFunction_ << ")";
+  } else {
+    out << SmootherPrototype::description();
   }
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0)
-      out << "Matlab Smoother("<<setupFunction_<<"/"<<solveFunction_<<")";
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-    }
+  if (verbLevel & Parameters0)
+    out << "Matlab Smoother(" << setupFunction_ << "/" << solveFunction_ << ")";
 
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
   }
 
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  }
+}
 
 // Dummy specializations for GO = long long
 /*template <>
@@ -190,8 +181,7 @@ void MatlabSmoother<std::complex<double>,int,long long>::Apply(MultiVector& X, c
   throw std::runtime_error("MatlabSmoother does not support GlobalOrdinal == long long.");
 }*/
 
+}  // namespace MueLu
 
-} // namespace MueLu
-
-#endif // HAVE_MUELU_MATLAB
-#endif // MUELU_MATLABSMOOTHER_DEF_HPP
+#endif  // HAVE_MUELU_MATLAB
+#endif  // MUELU_MATLABSMOOTHER_DEF_HPP
diff --git a/packages/muelu/matlab/src/MueLu_MatlabUtils.cpp b/packages/muelu/matlab/src/MueLu_MatlabUtils.cpp
index 4824c411dd27..bbd4a2d38615 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabUtils.cpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabUtils.cpp
@@ -50,7 +50,7 @@
 #else
 
 /* Stuff for MATLAB R2006b vs. previous versions */
-#if(defined(MX_API_VER) && MX_API_VER >= 0x07030000)
+#if (defined(MX_API_VER) && MX_API_VER >= 0x07030000)
 #else
 typedef int mwIndex;
 #endif
@@ -61,33 +61,32 @@ using namespace Teuchos;
 namespace MueLu {
 
 /* Explicit instantiation of MuemexData variants */
-  template class MuemexData<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<MAggregates>>;
-  template class MuemexData<RCP<MAmalInfo>>;
-  template class MuemexData<int>;
-  template class MuemexData<bool>;
-  template class MuemexData<complex_t>;
-  template class MuemexData<string>;
-  template class MuemexData<double>;
-  template class MuemexData<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Epetra_MultiVector> >;
-  template class MuemexData<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
-  template class MuemexData<RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>;
+template class MuemexData<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<MAggregates> >;
+template class MuemexData<RCP<MAmalInfo> >;
+template class MuemexData<int>;
+template class MuemexData<bool>;
+template class MuemexData<complex_t>;
+template class MuemexData<string>;
+template class MuemexData<double>;
+template class MuemexData<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Epetra_MultiVector> >;
+template class MuemexData<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
+template class MuemexData<RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >;
 
-//Flag set to true if MATLAB's CSC matrix index type is not int (usually false)
+// Flag set to true if MATLAB's CSC matrix index type is not int (usually false)
 bool rewrap_ints = sizeof(int) != sizeof(mwIndex);
 
-int* mwIndex_to_int(int N, mwIndex* mwi_array)
-{
-  //int* rv = (int*) malloc(N * sizeof(int));
-  int* rv = new int[N]; // not really better but may avoid confusion for valgrind
-  for(int i = 0; i < N; i++)
-    rv[i] = (int) mwi_array[i];
+int* mwIndex_to_int(int N, mwIndex* mwi_array) {
+  // int* rv = (int*) malloc(N * sizeof(int));
+  int* rv = new int[N];  // not really better but may avoid confusion for valgrind
+  for (int i = 0; i < N; i++)
+    rv[i] = (int)mwi_array[i];
   return rv;
 }
 
@@ -95,27 +94,26 @@ int* mwIndex_to_int(int N, mwIndex* mwi_array)
 /* Specializations                 */
 /* ******************************* */
 
-template<> mxArray* createMatlabSparse<double>(int numRows, int numCols, int nnz)
-{
+template <>
+mxArray* createMatlabSparse<double>(int numRows, int numCols, int nnz) {
   return mxCreateSparse(numRows, numCols, nnz, mxREAL);
 }
 
-template<> mxArray* createMatlabSparse<complex_t>(int numRows, int numCols, int nnz)
-{
+template <>
+mxArray* createMatlabSparse<complex_t>(int numRows, int numCols, int nnz) {
   return mxCreateSparse(numRows, numCols, nnz, mxCOMPLEX);
 }
 
-template<> void fillMatlabArray<double>(double* array, const mxArray* mxa, int n)
-{
+template <>
+void fillMatlabArray<double>(double* array, const mxArray* mxa, int n) {
   memcpy(mxGetPr(mxa), array, n * sizeof(double));
 }
 
-template<> void fillMatlabArray<complex_t>(complex_t* array, const mxArray* mxa, int n)
-{
+template <>
+void fillMatlabArray<complex_t>(complex_t* array, const mxArray* mxa, int n) {
   double* pr = mxGetPr(mxa);
   double* pi = mxGetPi(mxa);
-  for(int i = 0; i < n; i++)
-  {
+  for (int i = 0; i < n; i++) {
     pr[i] = std::real<double>(array[i]);
     pi[i] = std::imag<double>(array[i]);
   }
@@ -125,26 +123,21 @@ template<> void fillMatlabArray<complex_t>(complex_t* array, const mxArray* mxa,
 /* Callback Functions         */
 /******************************/
 
-void callMatlabNoArgs(std::string function)
-{
+void callMatlabNoArgs(std::string function) {
   int result = mexEvalString(function.c_str());
-  if(result != 0)
-    mexPrintf("An error occurred while running a MATLAB command.");\
+  if (result != 0)
+    mexPrintf("An error occurred while running a MATLAB command.");
 }
 
-std::vector<RCP<MuemexArg>> callMatlab(std::string function, int numOutputs, std::vector<RCP<MuemexArg>> args)
-{
+std::vector<RCP<MuemexArg> > callMatlab(std::string function, int numOutputs, std::vector<RCP<MuemexArg> > args) {
   using Teuchos::rcp_static_cast;
-  mxArray** matlabArgs = new mxArray* [args.size()];
-  mxArray** matlabOutput = new mxArray* [numOutputs];
-  std::vector<RCP<MuemexArg>> output;
+  mxArray** matlabArgs   = new mxArray*[args.size()];
+  mxArray** matlabOutput = new mxArray*[numOutputs];
+  std::vector<RCP<MuemexArg> > output;
 
-  for(int i = 0; i < int(args.size()); i++)
-  {
-    try
-    {
-      switch(args[i]->type)
-      {
+  for (int i = 0; i < int(args.size()); i++) {
+    try {
+      switch (args[i]->type) {
         case BOOL:
           matlabArgs[i] = rcp_static_cast<MuemexData<bool>, MuemexArg>(args[i])->convertToMatlab();
           break;
@@ -161,77 +154,71 @@ std::vector<RCP<MuemexArg>> callMatlab(std::string function, int numOutputs, std
           matlabArgs[i] = rcp_static_cast<MuemexData<complex_t>, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_MAP:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_map>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_map> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_ORDINAL_VECTOR:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_ordinal_vector>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_ordinal_vector> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case TPETRA_MULTIVECTOR_DOUBLE:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case TPETRA_MULTIVECTOR_COMPLEX:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case TPETRA_MATRIX_DOUBLE:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case TPETRA_MATRIX_COMPLEX:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_MATRIX_DOUBLE:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_Matrix_double>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_Matrix_double> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_MATRIX_COMPLEX:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_Matrix_complex>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_Matrix_complex> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_MULTIVECTOR_DOUBLE:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_MultiVector_double>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_MultiVector_double> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case XPETRA_MULTIVECTOR_COMPLEX:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_MultiVector_complex>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Xpetra_MultiVector_complex> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case EPETRA_CRSMATRIX:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Epetra_CrsMatrix>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Epetra_CrsMatrix> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case EPETRA_MULTIVECTOR:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Epetra_MultiVector>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<Epetra_MultiVector> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case AGGREGATES:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MAggregates>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MAggregates> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case AMALGAMATION_INFO:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MAmalInfo>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MAmalInfo> >, MuemexArg>(args[i])->convertToMatlab();
           break;
         case GRAPH:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MGraph>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<MGraph> >, MuemexArg>(args[i])->convertToMatlab();
 #ifdef HAVE_MUELU_INTREPID2
         case FIELDCONTAINER_ORDINAL:
-          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<FieldContainer_ordinal>>, MuemexArg>(args[i])->convertToMatlab();
+          matlabArgs[i] = rcp_static_cast<MuemexData<RCP<FieldContainer_ordinal> >, MuemexArg>(args[i])->convertToMatlab();
           break;
 #endif
       }
-    }
-    catch (std::exception& e)
-    {
+    } catch (std::exception& e) {
       mexPrintf("An error occurred while converting arg #%d to MATLAB:\n", i);
       std::cout << e.what() << std::endl;
       mexPrintf("Passing 0 instead.\n");
       matlabArgs[i] = mxCreateDoubleScalar(0);
     }
   }
-  //now matlabArgs is populated with MATLAB data types
+  // now matlabArgs is populated with MATLAB data types
   int result = mexCallMATLAB(numOutputs, matlabOutput, args.size(), matlabArgs, function.c_str());
-  if(result != 0)
+  if (result != 0)
     mexPrintf("Matlab encountered an error while running command through muemexCallbacks.\n");
-  //now, if all went well, matlabOutput contains all the output to return to user
-  for(int i = 0; i < numOutputs; i++)
-  {
-    try
-    {
+  // now, if all went well, matlabOutput contains all the output to return to user
+  for (int i = 0; i < numOutputs; i++) {
+    try {
       output.push_back(convertMatlabVar(matlabOutput[i]));
-    }
-    catch(std::exception& e)
-    {
+    } catch (std::exception& e) {
       mexPrintf("An error occurred while converting output #%d from MATLAB:\n", i);
       std::cout << e.what() << std::endl;
     }
@@ -245,90 +232,83 @@ std::vector<RCP<MuemexArg>> callMatlab(std::string function, int numOutputs, std
 /* More utility functions     */
 /******************************/
 
-template<> mxArray* createMatlabMultiVector<double>(int numRows, int numCols)
-{
+template <>
+mxArray* createMatlabMultiVector<double>(int numRows, int numCols) {
   return mxCreateDoubleMatrix(numRows, numCols, mxREAL);
 }
 
-template<> mxArray* createMatlabMultiVector<complex_t>(int numRows, int numCols)
-{
+template <>
+mxArray* createMatlabMultiVector<complex_t>(int numRows, int numCols) {
   return mxCreateDoubleMatrix(numRows, numCols, mxCOMPLEX);
 }
 
-mxArray* saveAmalInfo(RCP<MAmalInfo>& amalInfo)
-{
+mxArray* saveAmalInfo(RCP<MAmalInfo>& amalInfo) {
   throw runtime_error("AmalgamationInfo not supported in MueMex yet.");
   return mxCreateDoubleScalar(0);
 }
 
-bool isValidMatlabAggregates(const mxArray* mxa)
-{
+bool isValidMatlabAggregates(const mxArray* mxa) {
   bool isValidAggregates = true;
-  if(!mxIsStruct(mxa))
+  if (!mxIsStruct(mxa))
     return false;
-  int numFields = mxGetNumberOfFields(mxa); //check that struct has correct # of fields
-  if(numFields != 5)
+  int numFields = mxGetNumberOfFields(mxa);  // check that struct has correct # of fields
+  if (numFields != 5)
     isValidAggregates = false;
-  if(isValidAggregates)
-  {
+  if (isValidAggregates) {
     const char* mem1 = mxGetFieldNameByNumber(mxa, 0);
-    if(mem1 == NULL || strcmp(mem1, "nVertices") != 0)
+    if (mem1 == NULL || strcmp(mem1, "nVertices") != 0)
       isValidAggregates = false;
     const char* mem2 = mxGetFieldNameByNumber(mxa, 1);
-    if(mem2 == NULL || strcmp(mem2, "nAggregates") != 0)
+    if (mem2 == NULL || strcmp(mem2, "nAggregates") != 0)
       isValidAggregates = false;
     const char* mem3 = mxGetFieldNameByNumber(mxa, 2);
-    if(mem3 == NULL || strcmp(mem3, "vertexToAggID") != 0)
+    if (mem3 == NULL || strcmp(mem3, "vertexToAggID") != 0)
       isValidAggregates = false;
     const char* mem4 = mxGetFieldNameByNumber(mxa, 3);
-    if(mem3 == NULL || strcmp(mem4, "rootNodes") != 0)
+    if (mem3 == NULL || strcmp(mem4, "rootNodes") != 0)
       isValidAggregates = false;
     const char* mem5 = mxGetFieldNameByNumber(mxa, 4);
-    if(mem4 == NULL || strcmp(mem5, "aggSizes") != 0)
+    if (mem4 == NULL || strcmp(mem5, "aggSizes") != 0)
       isValidAggregates = false;
   }
   return isValidAggregates;
 }
 
-bool isValidMatlabGraph(const mxArray* mxa)
-{
+bool isValidMatlabGraph(const mxArray* mxa) {
   bool isValidGraph = true;
-  if(!mxIsStruct(mxa))
+  if (!mxIsStruct(mxa))
     return false;
-  int numFields = mxGetNumberOfFields(mxa); //check that struct has correct # of fields
-  if(numFields != 2)
+  int numFields = mxGetNumberOfFields(mxa);  // check that struct has correct # of fields
+  if (numFields != 2)
     isValidGraph = false;
-  if(isValidGraph)
-  {
+  if (isValidGraph) {
     const char* mem1 = mxGetFieldNameByNumber(mxa, 0);
-    if(mem1 == NULL || strcmp(mem1, "edges") != 0)
+    if (mem1 == NULL || strcmp(mem1, "edges") != 0)
       isValidGraph = false;
     const char* mem2 = mxGetFieldNameByNumber(mxa, 1);
-    if(mem2 == NULL || strcmp(mem2, "boundaryNodes") != 0)
-      isValidGraph  = false;
+    if (mem2 == NULL || strcmp(mem2, "boundaryNodes") != 0)
+      isValidGraph = false;
   }
   return isValidGraph;
 }
 
-std::vector<std::string> tokenizeList(const std::string& params)
-{
+std::vector<std::string> tokenizeList(const std::string& params) {
   using namespace std;
   vector<string> rlist;
   const char* delims = ",";
-  char* copy = (char*) malloc(params.length() + 1);
+  char* copy         = (char*)malloc(params.length() + 1);
   strcpy(copy, params.c_str());
-  char* mark = (char*) strtok(copy, delims);
-  while(mark != NULL)
-  {
-    //Remove leading and trailing whitespace in token
+  char* mark = (char*)strtok(copy, delims);
+  while (mark != NULL) {
+    // Remove leading and trailing whitespace in token
     char* tail = mark + strlen(mark) - 1;
-    while(*mark == ' ')
+    while (*mark == ' ')
       mark++;
-    while(*tail == ' ' && tail > mark)
+    while (*tail == ' ' && tail > mark)
       tail--;
     tail++;
     *tail = 0;
-    string tok(mark); //copies the characters to string object
+    string tok(mark);  // copies the characters to string object
     rlist.push_back(tok);
     mark = strtok(NULL, delims);
   }
@@ -336,84 +316,76 @@ std::vector<std::string> tokenizeList(const std::string& params)
   return rlist;
 }
 
-Teuchos::RCP<Teuchos::ParameterList> getInputParamList()
-{
+Teuchos::RCP<Teuchos::ParameterList> getInputParamList() {
   using namespace Teuchos;
   RCP<ParameterList> validParamList = rcp(new ParameterList());
-  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Factory for the matrix A.");
-  validParamList->set<RCP<const FactoryBase>>("P", Teuchos::null, "Factory for the prolongator.");
-  validParamList->set<RCP<const FactoryBase>>("R", Teuchos::null, "Factory for the restrictor.");
-  validParamList->set<RCP<const FactoryBase>>("Ptent", Teuchos::null, "Factory for the tentative (unsmoothed) prolongator.");
-  validParamList->set<RCP<const FactoryBase>>("Coordinates", Teuchos::null, "Factory for the node coordinates.");
-  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Factory for the nullspace.");
-  validParamList->set<RCP<const FactoryBase>>("Aggregates",  Teuchos::null, "Factory for the aggregates.");
-  validParamList->set<RCP<const FactoryBase>>("UnamalgamationInfo", Teuchos::null, "Factory for amalgamation.");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory for the matrix A.");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Factory for the prolongator.");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Factory for the restrictor.");
+  validParamList->set<RCP<const FactoryBase> >("Ptent", Teuchos::null, "Factory for the tentative (unsmoothed) prolongator.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for the node coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Factory for the nullspace.");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Factory for the aggregates.");
+  validParamList->set<RCP<const FactoryBase> >("UnamalgamationInfo", Teuchos::null, "Factory for amalgamation.");
 #ifdef HAVE_MUELU_INTREPID2
-  validParamList->set<RCP<const FactoryBase>>("pcoarsen: element to node map", Teuchos::null, "Generating factory of the element to node map");
+  validParamList->set<RCP<const FactoryBase> >("pcoarsen: element to node map", Teuchos::null, "Generating factory of the element to node map");
 #endif
   return validParamList;
 }
 
-Teuchos::RCP<MuemexArg> convertMatlabVar(const mxArray* mxa)
-{
-  switch(mxGetClassID(mxa))
-  {
+Teuchos::RCP<MuemexArg> convertMatlabVar(const mxArray* mxa) {
+  switch (mxGetClassID(mxa)) {
     case mxCHAR_CLASS:
-      //string
+      // string
       return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<std::string>(mxa)));
       break;
     case mxLOGICAL_CLASS:
-      //boolean
+      // boolean
       return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<bool>(mxa)));
       break;
     case mxINT32_CLASS:
-      if(mxGetM(mxa) == 1 && mxGetN(mxa) == 1)
-        //individual integer
+      if (mxGetM(mxa) == 1 && mxGetN(mxa) == 1)
+        // individual integer
         return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<int>(mxa)));
-      else if(mxGetM(mxa) != 1 || mxGetN(mxa) != 1)
-        //ordinal vector
-        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_ordinal_vector>>(mxa)));
+      else if (mxGetM(mxa) != 1 || mxGetN(mxa) != 1)
+        // ordinal vector
+        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_ordinal_vector> >(mxa)));
       else
         throw std::runtime_error("Error: Don't know what to do with integer array.\n");
       break;
     case mxDOUBLE_CLASS:
-      if(mxGetM(mxa) == 1 && mxGetN(mxa) == 1)
-      {
-        if(mxIsComplex(mxa))
-          //single double (scalar, real)
+      if (mxGetM(mxa) == 1 && mxGetN(mxa) == 1) {
+        if (mxIsComplex(mxa))
+          // single double (scalar, real)
           return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<complex_t>(mxa)));
         else
-          //single complex scalar
+          // single complex scalar
           return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<double>(mxa)));
-      }
-      else if(mxIsSparse(mxa)) //use a CRS matrix
+      } else if (mxIsSparse(mxa))  // use a CRS matrix
       {
-        //Default to Tpetra matrix for this
-        if(mxIsComplex(mxa))
-          //complex matrix
-          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_Matrix_complex>>(mxa)));
+        // Default to Tpetra matrix for this
+        if (mxIsComplex(mxa))
+          // complex matrix
+          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_Matrix_complex> >(mxa)));
         else
-          //real-valued matrix
-          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_Matrix_double>>(mxa)));
-      }
-      else
-      {
-        //Default to Xpetra multivector for this case
-        if(mxIsComplex(mxa))
-          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa)));
+          // real-valued matrix
+          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra_Matrix_double> >(mxa)));
+      } else {
+        // Default to Xpetra multivector for this case
+        if (mxIsComplex(mxa))
+          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa)));
         else
-          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa)));
+          return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa)));
       }
       break;
-    case mxSTRUCT_CLASS:
-    {
-      //the only thing that should get here currently is an Aggregates struct or Graph struct
-      //verify that it has the correct fields with the correct types
-      //also assume that aggregates data will not be stored in an array of more than 1 element.
-      if(isValidMatlabAggregates(mxa)) {
-        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<MAggregates>>(mxa)));
-      } else if(isValidMatlabGraph(mxa)) {
-        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<MGraph>>(mxa)));
+    case mxSTRUCT_CLASS: {
+      // the only thing that should get here currently is an Aggregates struct or Graph struct
+      // verify that it has the correct fields with the correct types
+      // also assume that aggregates data will not be stored in an array of more than 1 element.
+      if (isValidMatlabAggregates(mxa)) {
+        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<MAggregates> >(mxa)));
+      } else if (isValidMatlabGraph(mxa)) {
+        return rcp_implicit_cast<MuemexArg>(rcp(new MuemexData<RCP<MGraph> >(mxa)));
       } else {
         throw runtime_error("Invalid aggregates or graph struct passed in from MATLAB.");
         return Teuchos::null;
@@ -435,19 +407,19 @@ template int loadDataFromMatlab<int>(const mxArray* mxa);
 template double loadDataFromMatlab<double>(const mxArray* mxa);
 template complex_t loadDataFromMatlab<complex_t>(const mxArray* mxa);
 template string loadDataFromMatlab<string>(const mxArray* mxa);
-template RCP<Xpetra_ordinal_vector> loadDataFromMatlab<RCP<Xpetra_ordinal_vector>>(const mxArray* mxa);
-template RCP<Tpetra_MultiVector_double> loadDataFromMatlab<RCP<Tpetra_MultiVector_double>>(const mxArray* mxa);
-template RCP<Tpetra_MultiVector_complex> loadDataFromMatlab<RCP<Tpetra_MultiVector_complex>>(const mxArray* mxa);
-template RCP<Tpetra_CrsMatrix_double> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double>>(const mxArray* mxa);
-template RCP<Tpetra_CrsMatrix_complex> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_complex>>(const mxArray* mxa);
-template RCP<Xpetra_Matrix_double> loadDataFromMatlab<RCP<Xpetra_Matrix_double>>(const mxArray* mxa);
-template RCP<Xpetra_Matrix_complex> loadDataFromMatlab<RCP<Xpetra_Matrix_complex>>(const mxArray* mxa);
-template RCP<Xpetra_MultiVector_double> loadDataFromMatlab<RCP<Xpetra_MultiVector_double>>(const mxArray* mxa);
-template RCP<Xpetra_MultiVector_complex> loadDataFromMatlab<RCP<Xpetra_MultiVector_complex>>(const mxArray* mxa);
-template RCP<Epetra_CrsMatrix> loadDataFromMatlab<RCP<Epetra_CrsMatrix>>(const mxArray* mxa);
-template RCP<Epetra_MultiVector> loadDataFromMatlab<RCP<Epetra_MultiVector>>(const mxArray* mxa);
-template RCP<MAggregates> loadDataFromMatlab<RCP<MAggregates>>(const mxArray* mxa);
-template RCP<MAmalInfo> loadDataFromMatlab<RCP<MAmalInfo>>(const mxArray* mxa);
+template RCP<Xpetra_ordinal_vector> loadDataFromMatlab<RCP<Xpetra_ordinal_vector> >(const mxArray* mxa);
+template RCP<Tpetra_MultiVector_double> loadDataFromMatlab<RCP<Tpetra_MultiVector_double> >(const mxArray* mxa);
+template RCP<Tpetra_MultiVector_complex> loadDataFromMatlab<RCP<Tpetra_MultiVector_complex> >(const mxArray* mxa);
+template RCP<Tpetra_CrsMatrix_double> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double> >(const mxArray* mxa);
+template RCP<Tpetra_CrsMatrix_complex> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_complex> >(const mxArray* mxa);
+template RCP<Xpetra_Matrix_double> loadDataFromMatlab<RCP<Xpetra_Matrix_double> >(const mxArray* mxa);
+template RCP<Xpetra_Matrix_complex> loadDataFromMatlab<RCP<Xpetra_Matrix_complex> >(const mxArray* mxa);
+template RCP<Xpetra_MultiVector_double> loadDataFromMatlab<RCP<Xpetra_MultiVector_double> >(const mxArray* mxa);
+template RCP<Xpetra_MultiVector_complex> loadDataFromMatlab<RCP<Xpetra_MultiVector_complex> >(const mxArray* mxa);
+template RCP<Epetra_CrsMatrix> loadDataFromMatlab<RCP<Epetra_CrsMatrix> >(const mxArray* mxa);
+template RCP<Epetra_MultiVector> loadDataFromMatlab<RCP<Epetra_MultiVector> >(const mxArray* mxa);
+template RCP<MAggregates> loadDataFromMatlab<RCP<MAggregates> >(const mxArray* mxa);
+template RCP<MAmalInfo> loadDataFromMatlab<RCP<MAmalInfo> >(const mxArray* mxa);
 
 template mxArray* saveDataToMatlab(bool& data);
 template mxArray* saveDataToMatlab(int& data);
@@ -468,10 +440,10 @@ template mxArray* saveDataToMatlab(RCP<Epetra_MultiVector>& data);
 template mxArray* saveDataToMatlab(RCP<MAggregates>& data);
 template mxArray* saveDataToMatlab(RCP<MAmalInfo>& data);
 
-template vector<RCP<MuemexArg>> processNeeds<double>(const Factory* factory, string& needsParam, Level& lvl);
-template vector<RCP<MuemexArg>> processNeeds<complex_t>(const Factory* factory, string& needsParam, Level& lvl);
-template void processProvides<double>(vector<RCP<MuemexArg>>& mexOutput, const Factory* factory, string& providesParam, Level& lvl);
-template void processProvides<complex_t>(vector<RCP<MuemexArg>>& mexOutput, const Factory* factory, string& providesParam, Level& lvl);
+template vector<RCP<MuemexArg> > processNeeds<double>(const Factory* factory, string& needsParam, Level& lvl);
+template vector<RCP<MuemexArg> > processNeeds<complex_t>(const Factory* factory, string& needsParam, Level& lvl);
+template void processProvides<double>(vector<RCP<MuemexArg> >& mexOutput, const Factory* factory, string& providesParam, Level& lvl);
+template void processProvides<complex_t>(vector<RCP<MuemexArg> >& mexOutput, const Factory* factory, string& providesParam, Level& lvl);
 
-}//end namespace
-#endif // HAVE_MUELU_MATLAB
+}  // namespace MueLu
+#endif  // HAVE_MUELU_MATLAB
diff --git a/packages/muelu/matlab/src/MueLu_MatlabUtils_decl.hpp b/packages/muelu/matlab/src/MueLu_MatlabUtils_decl.hpp
index f75a89a92d2c..391ccca1acf8 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabUtils_decl.hpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabUtils_decl.hpp
@@ -77,12 +77,9 @@
 
 #include "Kokkos_DynRankView.hpp"
 
+namespace MueLu {
 
-namespace MueLu
-{
-
-enum MuemexType
-{
+enum MuemexType {
   INT,
   BOOL,
   DOUBLE,
@@ -104,12 +101,13 @@ enum MuemexType
   AMALGAMATION_INFO,
   GRAPH
 #ifdef HAVE_MUELU_INTREPID2
-, FIELDCONTAINER_ORDINAL
+  ,
+  FIELDCONTAINER_ORDINAL
 #endif
 };
 
 typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<Kokkos::Serial, Kokkos::HostSpace> mm_node_t;
-typedef typename Tpetra::Map<>::local_ordinal_type mm_LocalOrd;  //these are used for LocalOrdinal and GlobalOrdinal of all xpetra/tpetra templated types
+typedef typename Tpetra::Map<>::local_ordinal_type mm_LocalOrd;  // these are used for LocalOrdinal and GlobalOrdinal of all xpetra/tpetra templated types
 typedef typename Tpetra::Map<>::global_ordinal_type mm_GlobalOrd;
 typedef std::complex<double> complex_t;
 typedef Tpetra::Map<> muemex_map_type;
@@ -131,91 +129,93 @@ typedef MueLu::AmalgamationInfo<mm_LocalOrd, mm_GlobalOrd, mm_node_t> MAmalInfo;
 typedef MueLu::GraphBase<mm_LocalOrd, mm_GlobalOrd, mm_node_t> MGraph;
 
 #ifdef HAVE_MUELU_INTREPID2
-  typedef Kokkos::DynRankView<mm_LocalOrd,typename mm_node_t::device_type> FieldContainer_ordinal;
+typedef Kokkos::DynRankView<mm_LocalOrd, typename mm_node_t::device_type> FieldContainer_ordinal;
 #endif
 
-class MuemexArg
-{
-  public:
-    MuemexArg(MuemexType dataType) {type = dataType;}
-    MuemexType type;
+class MuemexArg {
+ public:
+  MuemexArg(MuemexType dataType) { type = dataType; }
+  MuemexType type;
 };
 
-template<typename T>
-MuemexType getMuemexType(const T & data);
-
-template<typename T>
-class MuemexData : public MuemexArg
-{
-  public:
-    MuemexData(T& data); //Construct from pre-existing data, to pass to MATLAB.
-    MuemexData(T& data, MuemexType type);        //Construct from pre-existing data, to pass to MATLAB.
-    MuemexData(const mxArray* mxa); //Construct from MATLAB array, to get from MATLAB.
-    mxArray* convertToMatlab(); //Create a MATLAB object and copy this data to it
-    T& getData();                         //Set and get methods
-    void setData(T& data);
-  private:
-    T data;
+template <typename T>
+MuemexType getMuemexType(const T& data);
+
+template <typename T>
+class MuemexData : public MuemexArg {
+ public:
+  MuemexData(T& data);                   // Construct from pre-existing data, to pass to MATLAB.
+  MuemexData(T& data, MuemexType type);  // Construct from pre-existing data, to pass to MATLAB.
+  MuemexData(const mxArray* mxa);        // Construct from MATLAB array, to get from MATLAB.
+  mxArray* convertToMatlab();            // Create a MATLAB object and copy this data to it
+  T& getData();                          // Set and get methods
+  void setData(T& data);
+
+ private:
+  T data;
 };
 
-template<typename T>
-MuemexType getMuemexType(const T & data);
+template <typename T>
+MuemexType getMuemexType(const T& data);
 
-template<typename T>
+template <typename T>
 MuemexType getMuemexType();
 
-template<typename T>
+template <typename T>
 T loadDataFromMatlab(const mxArray* mxa);
 
-template<typename T>
+template <typename T>
 mxArray* saveDataToMatlab(T& data);
 
-//Add data to level. Set the keep flag on the data to "user-provided" so it's not deleted.
-template<typename T>
-void addLevelVariable(const T& data, std::string& name, Level& lvl, const FactoryBase *fact = NoFactory::get());
+// Add data to level. Set the keep flag on the data to "user-provided" so it's not deleted.
+template <typename T>
+void addLevelVariable(const T& data, std::string& name, Level& lvl, const FactoryBase* fact = NoFactory::get());
 
-template<typename T>
+template <typename T>
 const T& getLevelVariable(std::string& name, Level& lvl);
 
-//Functions used to put data through matlab factories - first arg is "this" pointer of matlab factory
-template<typename Scalar = double, typename LocalOrdinal = mm_LocalOrd, typename GlobalOrdinal = mm_GlobalOrd, typename Node = mm_node_t>
-std::vector<Teuchos::RCP<MuemexArg>> processNeeds(const Factory* factory, std::string& needsParam, Level& lvl);
+// Functions used to put data through matlab factories - first arg is "this" pointer of matlab factory
+template <typename Scalar = double, typename LocalOrdinal = mm_LocalOrd, typename GlobalOrdinal = mm_GlobalOrd, typename Node = mm_node_t>
+std::vector<Teuchos::RCP<MuemexArg> > processNeeds(const Factory* factory, std::string& needsParam, Level& lvl);
 
-template<typename Scalar = double, typename LocalOrdinal = mm_LocalOrd, typename GlobalOrdinal = mm_GlobalOrd, typename Node = mm_node_t>
-void processProvides(std::vector<Teuchos::RCP<MuemexArg>>& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl);
+template <typename Scalar = double, typename LocalOrdinal = mm_LocalOrd, typename GlobalOrdinal = mm_GlobalOrd, typename Node = mm_node_t>
+void processProvides(std::vector<Teuchos::RCP<MuemexArg> >& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl);
 
-//create a sparse array in Matlab
-template<typename Scalar> mxArray* createMatlabSparse(int numRows, int numCols, int nnz);
-template<typename Scalar> mxArray* createMatlabMultiVector(int numRows, int numCols);
-template<typename Scalar> void fillMatlabArray(Scalar* array, const mxArray* mxa, int n);
+// create a sparse array in Matlab
+template <typename Scalar>
+mxArray* createMatlabSparse(int numRows, int numCols, int nnz);
+template <typename Scalar>
+mxArray* createMatlabMultiVector(int numRows, int numCols);
+template <typename Scalar>
+void fillMatlabArray(Scalar* array, const mxArray* mxa, int n);
 int* mwIndex_to_int(int N, mwIndex* mwi_array);
 bool isValidMatlabAggregates(const mxArray* mxa);
 bool isValidMatlabGraph(const mxArray* mxa);
 std::vector<std::string> tokenizeList(const std::string& param);
-//The two callback functions that MueLu can call to run anything in MATLAB
+// The two callback functions that MueLu can call to run anything in MATLAB
 void callMatlabNoArgs(std::string function);
-std::vector<Teuchos::RCP<MuemexArg>> callMatlab(std::string function, int numOutputs, std::vector<Teuchos::RCP<MuemexArg>> args);
+std::vector<Teuchos::RCP<MuemexArg> > callMatlab(std::string function, int numOutputs, std::vector<Teuchos::RCP<MuemexArg> > args);
 Teuchos::RCP<Teuchos::ParameterList> getInputParamList();
 Teuchos::RCP<MuemexArg> convertMatlabVar(const mxArray* mxa);
 
 // trim from start
-static inline std::string &ltrim(std::string &s) {
+static inline std::string& ltrim(std::string& s) {
   s.erase(s.begin(), std::find_if(s.begin(), s.end(), std::not1(std::ptr_fun<int, int>(std::isspace))));
   return s;
 }
 
 // trim from end
-static inline std::string &rtrim(std::string &s) {
+static inline std::string& rtrim(std::string& s) {
   s.erase(std::find_if(s.rbegin(), s.rend(), std::not1(std::ptr_fun<int, int>(std::isspace))).base(), s.end());
   return s;
 }
 
 // trim from both ends
-static inline std::string &trim(std::string &s) {
+static inline std::string& trim(std::string& s) {
   return ltrim(rtrim(s));
 }
 
-}//end namespace
+}  // namespace MueLu
 
-#endif //HAVE_MUELU_MATLAB error handler
-#endif //MUELU_MATLABUTILS_DECL_HPP guard
+#endif  // HAVE_MUELU_MATLAB error handler
+#endif  // MUELU_MATLABUTILS_DECL_HPP guard
diff --git a/packages/muelu/matlab/src/MueLu_MatlabUtils_def.hpp b/packages/muelu/matlab/src/MueLu_MatlabUtils_def.hpp
index 88c542d77ed0..541650def661 100644
--- a/packages/muelu/matlab/src/MueLu_MatlabUtils_def.hpp
+++ b/packages/muelu/matlab/src/MueLu_MatlabUtils_def.hpp
@@ -65,301 +65,316 @@ extern bool rewrap_ints;
 /* getMuemexType                   */
 /* ******************************* */
 
-template<typename T> MuemexType getMuemexType(const T & data) {throw std::runtime_error("Unknown Type");}
-
-template<> MuemexType getMuemexType(const int & data) {return INT;}
-template<> MuemexType getMuemexType<int>() {return INT;}
-template<> MuemexType getMuemexType<bool>() {return BOOL;}
-
-template<> MuemexType getMuemexType(const double & data) {return DOUBLE;}
-template<> MuemexType getMuemexType<double>() {return DOUBLE;}
-
-template<> MuemexType getMuemexType(const std::string & data) {return STRING;}
-template<> MuemexType getMuemexType<string>() {return STRING;}
-
-template<> MuemexType getMuemexType(const complex_t& data) {return COMPLEX;}
-template<> MuemexType getMuemexType<complex_t>() {return COMPLEX;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_map> & data) {return XPETRA_MAP;}
-template<> MuemexType getMuemexType<RCP<Xpetra_map> >() {return XPETRA_MAP;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_ordinal_vector> & data) {return XPETRA_ORDINAL_VECTOR;}
-template<> MuemexType getMuemexType<RCP<Xpetra_ordinal_vector>>() {return XPETRA_ORDINAL_VECTOR;}
-
-template<> MuemexType getMuemexType(const RCP<Tpetra_MultiVector_double> & data) {return TPETRA_MULTIVECTOR_DOUBLE;}
-template<> MuemexType getMuemexType<RCP<Tpetra_MultiVector_double>>() {return TPETRA_MULTIVECTOR_DOUBLE;}
-
-template<> MuemexType getMuemexType(const RCP<Tpetra_MultiVector_complex>& data) {return TPETRA_MULTIVECTOR_COMPLEX;}
-template<> MuemexType getMuemexType<RCP<Tpetra_MultiVector_complex>>() {return TPETRA_MULTIVECTOR_COMPLEX;}
-
-template<> MuemexType getMuemexType(const RCP<Tpetra_CrsMatrix_double> & data) {return TPETRA_MATRIX_DOUBLE;}
-template<> MuemexType getMuemexType<RCP<Tpetra_CrsMatrix_double>>() {return TPETRA_MATRIX_DOUBLE;}
-
-template<> MuemexType getMuemexType(const RCP<Tpetra_CrsMatrix_complex> & data) {return TPETRA_MATRIX_COMPLEX;}
-template<> MuemexType getMuemexType<RCP<Tpetra_CrsMatrix_complex>>() {return TPETRA_MATRIX_COMPLEX;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_MultiVector_double> & data) {return XPETRA_MULTIVECTOR_DOUBLE;}
-template<> MuemexType getMuemexType<RCP<Xpetra_MultiVector_double>>() {return XPETRA_MULTIVECTOR_DOUBLE;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_MultiVector_complex> & data) {return XPETRA_MULTIVECTOR_COMPLEX;}
-template<> MuemexType getMuemexType<RCP<Xpetra_MultiVector_complex>>() {return XPETRA_MULTIVECTOR_COMPLEX;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_Matrix_double> & data) {return XPETRA_MATRIX_DOUBLE;}
-template<> MuemexType getMuemexType<RCP<Xpetra_Matrix_double>>() {return XPETRA_MATRIX_DOUBLE;}
-
-template<> MuemexType getMuemexType(const RCP<Xpetra_Matrix_complex> & data) {return XPETRA_MATRIX_COMPLEX;}
-template<> MuemexType getMuemexType<RCP<Xpetra_Matrix_complex>>() {return XPETRA_MATRIX_COMPLEX;}
-
-template<> MuemexType getMuemexType(const RCP<Epetra_CrsMatrix> & data) {return EPETRA_CRSMATRIX;}
-template<> MuemexType getMuemexType<RCP<Epetra_CrsMatrix>>() {return EPETRA_CRSMATRIX;}
-
-template<> MuemexType getMuemexType(const RCP<Epetra_MultiVector> & data) {return EPETRA_MULTIVECTOR;}
-template<> MuemexType getMuemexType<RCP<Epetra_MultiVector>>() {return EPETRA_MULTIVECTOR;}
-
-template<> MuemexType getMuemexType(const RCP<MAggregates>& data) {return AGGREGATES;}
-template<> MuemexType getMuemexType<RCP<MAggregates>>() {return AGGREGATES;}
-
-template<> MuemexType getMuemexType(const RCP<MAmalInfo>& data) {return AMALGAMATION_INFO;}
-template<> MuemexType getMuemexType<RCP<MAmalInfo>>() {return AMALGAMATION_INFO;}
-
-template<> MuemexType getMuemexType(const RCP<MGraph>& data) {return GRAPH;}
-template<> MuemexType getMuemexType<RCP<MGraph>>() {return GRAPH;}
+template <typename T>
+MuemexType getMuemexType(const T& data) { throw std::runtime_error("Unknown Type"); }
+
+template <>
+MuemexType getMuemexType(const int& data) { return INT; }
+template <>
+MuemexType getMuemexType<int>() { return INT; }
+template <>
+MuemexType getMuemexType<bool>() { return BOOL; }
+
+template <>
+MuemexType getMuemexType(const double& data) { return DOUBLE; }
+template <>
+MuemexType getMuemexType<double>() { return DOUBLE; }
+
+template <>
+MuemexType getMuemexType(const std::string& data) { return STRING; }
+template <>
+MuemexType getMuemexType<string>() { return STRING; }
+
+template <>
+MuemexType getMuemexType(const complex_t& data) { return COMPLEX; }
+template <>
+MuemexType getMuemexType<complex_t>() { return COMPLEX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_map>& data) { return XPETRA_MAP; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_map> >() { return XPETRA_MAP; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_ordinal_vector>& data) { return XPETRA_ORDINAL_VECTOR; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_ordinal_vector> >() { return XPETRA_ORDINAL_VECTOR; }
+
+template <>
+MuemexType getMuemexType(const RCP<Tpetra_MultiVector_double>& data) { return TPETRA_MULTIVECTOR_DOUBLE; }
+template <>
+MuemexType getMuemexType<RCP<Tpetra_MultiVector_double> >() { return TPETRA_MULTIVECTOR_DOUBLE; }
+
+template <>
+MuemexType getMuemexType(const RCP<Tpetra_MultiVector_complex>& data) { return TPETRA_MULTIVECTOR_COMPLEX; }
+template <>
+MuemexType getMuemexType<RCP<Tpetra_MultiVector_complex> >() { return TPETRA_MULTIVECTOR_COMPLEX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Tpetra_CrsMatrix_double>& data) { return TPETRA_MATRIX_DOUBLE; }
+template <>
+MuemexType getMuemexType<RCP<Tpetra_CrsMatrix_double> >() { return TPETRA_MATRIX_DOUBLE; }
+
+template <>
+MuemexType getMuemexType(const RCP<Tpetra_CrsMatrix_complex>& data) { return TPETRA_MATRIX_COMPLEX; }
+template <>
+MuemexType getMuemexType<RCP<Tpetra_CrsMatrix_complex> >() { return TPETRA_MATRIX_COMPLEX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_MultiVector_double>& data) { return XPETRA_MULTIVECTOR_DOUBLE; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_MultiVector_double> >() { return XPETRA_MULTIVECTOR_DOUBLE; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_MultiVector_complex>& data) { return XPETRA_MULTIVECTOR_COMPLEX; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_MultiVector_complex> >() { return XPETRA_MULTIVECTOR_COMPLEX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_Matrix_double>& data) { return XPETRA_MATRIX_DOUBLE; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_Matrix_double> >() { return XPETRA_MATRIX_DOUBLE; }
+
+template <>
+MuemexType getMuemexType(const RCP<Xpetra_Matrix_complex>& data) { return XPETRA_MATRIX_COMPLEX; }
+template <>
+MuemexType getMuemexType<RCP<Xpetra_Matrix_complex> >() { return XPETRA_MATRIX_COMPLEX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Epetra_CrsMatrix>& data) { return EPETRA_CRSMATRIX; }
+template <>
+MuemexType getMuemexType<RCP<Epetra_CrsMatrix> >() { return EPETRA_CRSMATRIX; }
+
+template <>
+MuemexType getMuemexType(const RCP<Epetra_MultiVector>& data) { return EPETRA_MULTIVECTOR; }
+template <>
+MuemexType getMuemexType<RCP<Epetra_MultiVector> >() { return EPETRA_MULTIVECTOR; }
+
+template <>
+MuemexType getMuemexType(const RCP<MAggregates>& data) { return AGGREGATES; }
+template <>
+MuemexType getMuemexType<RCP<MAggregates> >() { return AGGREGATES; }
+
+template <>
+MuemexType getMuemexType(const RCP<MAmalInfo>& data) { return AMALGAMATION_INFO; }
+template <>
+MuemexType getMuemexType<RCP<MAmalInfo> >() { return AMALGAMATION_INFO; }
+
+template <>
+MuemexType getMuemexType(const RCP<MGraph>& data) { return GRAPH; }
+template <>
+MuemexType getMuemexType<RCP<MGraph> >() { return GRAPH; }
 
 #ifdef HAVE_MUELU_INTREPID2
-template<> MuemexType getMuemexType(const RCP<FieldContainer_ordinal>& data) {return FIELDCONTAINER_ORDINAL;}
-template<> MuemexType getMuemexType<RCP<FieldContainer_ordinal>>() {return FIELDCONTAINER_ORDINAL;}
+template <>
+MuemexType getMuemexType(const RCP<FieldContainer_ordinal>& data) { return FIELDCONTAINER_ORDINAL; }
+template <>
+MuemexType getMuemexType<RCP<FieldContainer_ordinal> >() { return FIELDCONTAINER_ORDINAL; }
 #endif
 
 /* "prototypes" for specialized functions used in other specialized functions */
 
-template<> mxArray* createMatlabSparse<double>(int numRows, int numCols, int nnz);
-template<> mxArray* createMatlabSparse<complex_t>(int numRows, int numCols, int nnz);
-template<> mxArray* createMatlabMultiVector<double>(int numRows, int numCols);
-template<> mxArray* createMatlabMultiVector<complex_t>(int numRows, int numCols);
-template<> void fillMatlabArray<double>(double* array, const mxArray* mxa, int n);
-template<> void fillMatlabArray<complex_t>(complex_t* array, const mxArray* mxa, int n);
-template<> mxArray* saveDataToMatlab(RCP<Xpetra_MultiVector_double>& data);
-template<> mxArray* saveDataToMatlab(RCP<Xpetra_MultiVector_complex>& data);
-template<> mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_double>& data);
-template<> mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_complex>& data);
+template <>
+mxArray* createMatlabSparse<double>(int numRows, int numCols, int nnz);
+template <>
+mxArray* createMatlabSparse<complex_t>(int numRows, int numCols, int nnz);
+template <>
+mxArray* createMatlabMultiVector<double>(int numRows, int numCols);
+template <>
+mxArray* createMatlabMultiVector<complex_t>(int numRows, int numCols);
+template <>
+void fillMatlabArray<double>(double* array, const mxArray* mxa, int n);
+template <>
+void fillMatlabArray<complex_t>(complex_t* array, const mxArray* mxa, int n);
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_MultiVector_double>& data);
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_MultiVector_complex>& data);
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_double>& data);
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_complex>& data);
 
 /* ******************************* */
 /* loadDataFromMatlab              */
 /* ******************************* */
 
-template<>
-int loadDataFromMatlab<int>(const mxArray* mxa)
-{
+template <>
+int loadDataFromMatlab<int>(const mxArray* mxa) {
   mxClassID probIDtype = mxGetClassID(mxa);
   int rv;
-  if(probIDtype == mxINT32_CLASS)
-  {
-    rv = *((int*) mxGetData(mxa));
-  }
-  else if(probIDtype == mxLOGICAL_CLASS)
-  {
-    rv = (int) *((bool*) mxGetData(mxa));
-  }
-  else if(probIDtype == mxDOUBLE_CLASS)
-  {
-    rv = (int) *((double*) mxGetData(mxa));
-  }
-  else if(probIDtype == mxUINT32_CLASS)
-  {
-    rv = (int) *((unsigned int*) mxGetData(mxa));
-  }
-  else
-  {
+  if (probIDtype == mxINT32_CLASS) {
+    rv = *((int*)mxGetData(mxa));
+  } else if (probIDtype == mxLOGICAL_CLASS) {
+    rv = (int)*((bool*)mxGetData(mxa));
+  } else if (probIDtype == mxDOUBLE_CLASS) {
+    rv = (int)*((double*)mxGetData(mxa));
+  } else if (probIDtype == mxUINT32_CLASS) {
+    rv = (int)*((unsigned int*)mxGetData(mxa));
+  } else {
     rv = -1;
     throw std::runtime_error("Error: Unrecognized numerical type.");
   }
   return rv;
 }
 
-template<>
-bool loadDataFromMatlab<bool>(const mxArray* mxa)
-{
-  return *((bool*) mxGetData(mxa));
+template <>
+bool loadDataFromMatlab<bool>(const mxArray* mxa) {
+  return *((bool*)mxGetData(mxa));
 }
 
-template<>
-double loadDataFromMatlab<double>(const mxArray* mxa)
-{
-  return *((double*) mxGetPr(mxa));
+template <>
+double loadDataFromMatlab<double>(const mxArray* mxa) {
+  return *((double*)mxGetPr(mxa));
 }
 
-template<>
-complex_t loadDataFromMatlab<complex_t>(const mxArray* mxa)
-{
-  double realpart = real<double>(*((double*) mxGetPr(mxa)));
-  double imagpart = imag<double>(*((double*) mxGetPi(mxa)));
+template <>
+complex_t loadDataFromMatlab<complex_t>(const mxArray* mxa) {
+  double realpart = real<double>(*((double*)mxGetPr(mxa)));
+  double imagpart = imag<double>(*((double*)mxGetPi(mxa)));
   return complex_t(realpart, imagpart);
 }
 
-template<>
-string loadDataFromMatlab<string>(const mxArray* mxa)
-{
+template <>
+string loadDataFromMatlab<string>(const mxArray* mxa) {
   string rv = "";
-  if (mxGetClassID(mxa) != mxCHAR_CLASS)
-  {
+  if (mxGetClassID(mxa) != mxCHAR_CLASS) {
     throw runtime_error("Can't construct string from anything but a char array.");
   }
   rv = string(mxArrayToString(mxa));
   return rv;
 }
 
-template<>
-RCP<Xpetra_map> loadDataFromMatlab<RCP<Xpetra_map>>(const mxArray* mxa)
-{
+template <>
+RCP<Xpetra_map> loadDataFromMatlab<RCP<Xpetra_map> >(const mxArray* mxa) {
   RCP<const Teuchos::Comm<int> > comm = rcp(new Teuchos::SerialComm<int>());
-  int nr = mxGetM(mxa);
-  int nc = mxGetN(mxa);
-  if(nr != 1)
+  int nr                              = mxGetM(mxa);
+  int nc                              = mxGetN(mxa);
+  if (nr != 1)
     throw std::runtime_error("A Xpetra::Map representation from MATLAB must be a single row vector.");
-  double* pr = mxGetPr(mxa);
+  double* pr                    = mxGetPr(mxa);
   mm_GlobalOrd numGlobalIndices = nc;
 
   std::vector<mm_GlobalOrd> localGIDs(numGlobalIndices);
-  for(int i = 0; i < int(numGlobalIndices); i++) {
+  for (int i = 0; i < int(numGlobalIndices); i++) {
     localGIDs[i] = Teuchos::as<mm_GlobalOrd>(pr[i]);
   }
 
-  const Teuchos::ArrayView<const mm_GlobalOrd> localGIDs_view(&localGIDs[0],localGIDs.size());
+  const Teuchos::ArrayView<const mm_GlobalOrd> localGIDs_view(&localGIDs[0], localGIDs.size());
   RCP<Xpetra_map> map =
       Xpetra::MapFactory<mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(
-      Xpetra::UseTpetra,
-      Teuchos::OrdinalTraits<mm_GlobalOrd>::invalid(),
-      localGIDs_view,
-      0, comm);
+          Xpetra::UseTpetra,
+          Teuchos::OrdinalTraits<mm_GlobalOrd>::invalid(),
+          localGIDs_view,
+          0, comm);
 
-  if(map.is_null())
+  if (map.is_null())
     throw runtime_error("Failed to create Xpetra::Map.");
   return map;
 }
 
-template<>
-RCP<Xpetra_ordinal_vector> loadDataFromMatlab<RCP<Xpetra_ordinal_vector>>(const mxArray* mxa)
-{
+template <>
+RCP<Xpetra_ordinal_vector> loadDataFromMatlab<RCP<Xpetra_ordinal_vector> >(const mxArray* mxa) {
   RCP<const Teuchos::Comm<int> > comm = rcp(new Teuchos::SerialComm<int>());
-  if(mxGetN(mxa) != 1 && mxGetM(mxa) != 1)
+  if (mxGetN(mxa) != 1 && mxGetM(mxa) != 1)
     throw std::runtime_error("An OrdinalVector from MATLAB must be a single row or column vector.");
-  mm_GlobalOrd numGlobalIndices = mxGetM(mxa) * mxGetN(mxa);
-  RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t>> map = Xpetra::MapFactory<mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(Xpetra::UseTpetra, numGlobalIndices, 0, comm);
-  if(mxGetClassID(mxa) != mxINT32_CLASS)
+  mm_GlobalOrd numGlobalIndices                               = mxGetM(mxa) * mxGetN(mxa);
+  RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t> > map = Xpetra::MapFactory<mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(Xpetra::UseTpetra, numGlobalIndices, 0, comm);
+  if (mxGetClassID(mxa) != mxINT32_CLASS)
     throw std::runtime_error("Can only construct LOVector with int32 data.");
-  int* array = (int*) mxGetData(mxa);
-  if(map.is_null())
+  int* array = (int*)mxGetData(mxa);
+  if (map.is_null())
     throw runtime_error("Failed to create map for Xpetra ordinal vector.");
   RCP<Xpetra_ordinal_vector> loVec = Xpetra::VectorFactory<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(map, false);
-  if(loVec.is_null())
+  if (loVec.is_null())
     throw runtime_error("Failed to create ordinal vector with Xpetra::VectorFactory.");
-  for(int i = 0; i < int(numGlobalIndices); i++)
-  {
+  for (int i = 0; i < int(numGlobalIndices); i++) {
     loVec->replaceGlobalValue(i, 0, array[i]);
   }
   return loVec;
 }
 
-template<>
-RCP<Tpetra_MultiVector_double> loadDataFromMatlab<RCP<Tpetra_MultiVector_double>>(const mxArray* mxa)
-{
-  RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> mv;
-  try
-  {
-    int nr = mxGetM(mxa);
-    int nc = mxGetN(mxa);
-    double* pr = mxGetPr(mxa);
-    RCP<const Teuchos::Comm<int>> comm = Tpetra::getDefaultComm();
-    //numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
-    RCP<const muemex_map_type> map = rcp(new muemex_map_type(nr, (mm_GlobalOrd) 0, comm));
-    //Allocate a new array of complex values to use with the multivector
+template <>
+RCP<Tpetra_MultiVector_double> loadDataFromMatlab<RCP<Tpetra_MultiVector_double> >(const mxArray* mxa) {
+  RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > mv;
+  try {
+    int nr                              = mxGetM(mxa);
+    int nc                              = mxGetN(mxa);
+    double* pr                          = mxGetPr(mxa);
+    RCP<const Teuchos::Comm<int> > comm = Tpetra::getDefaultComm();
+    // numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
+    RCP<const muemex_map_type> map = rcp(new muemex_map_type(nr, (mm_GlobalOrd)0, comm));
+    // Allocate a new array of complex values to use with the multivector
     Teuchos::ArrayView<const double> arrView(pr, nr * nc);
     mv = rcp(new Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(map, arrView, size_t(nr), size_t(nc)));
-  }
-  catch(std::exception& e)
-  {
+  } catch (std::exception& e) {
     mexPrintf("Error constructing Tpetra MultiVector.\n");
     std::cout << e.what() << std::endl;
   }
   return mv;
 }
 
-template<>
-RCP<Tpetra_MultiVector_complex> loadDataFromMatlab<RCP<Tpetra_MultiVector_complex>>(const mxArray* mxa)
-{
-  RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> mv;
-  try
-  {
-    int nr = mxGetM(mxa);
-    int nc = mxGetN(mxa);
-    double* pr = mxGetPr(mxa);
-    double* pi = mxGetPi(mxa);
-    RCP<const Teuchos::Comm<int>> comm = Tpetra::getDefaultComm();
-    //numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
-    RCP<const muemex_map_type> map = rcp(new muemex_map_type(nr, (mm_GlobalOrd) 0, comm));
-    //Allocate a new array of complex values to use with the multivector
+template <>
+RCP<Tpetra_MultiVector_complex> loadDataFromMatlab<RCP<Tpetra_MultiVector_complex> >(const mxArray* mxa) {
+  RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > mv;
+  try {
+    int nr                              = mxGetM(mxa);
+    int nc                              = mxGetN(mxa);
+    double* pr                          = mxGetPr(mxa);
+    double* pi                          = mxGetPi(mxa);
+    RCP<const Teuchos::Comm<int> > comm = Tpetra::getDefaultComm();
+    // numGlobalIndices for map constructor is the number of rows in matrix/vectors, right?
+    RCP<const muemex_map_type> map = rcp(new muemex_map_type(nr, (mm_GlobalOrd)0, comm));
+    // Allocate a new array of complex values to use with the multivector
     complex_t* myArr = new complex_t[nr * nc];
-    for(int n = 0; n < nc; n++)
-    {
-      for(int m = 0; m < nr; m++)
-      {
+    for (int n = 0; n < nc; n++) {
+      for (int m = 0; m < nr; m++) {
         myArr[n * nr + m] = complex_t(pr[n * nr + m], pi[n * nr + m]);
       }
     }
     Teuchos::ArrayView<complex_t> arrView(myArr, nr * nc);
     mv = rcp(new Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(map, arrView, nr, nc));
-  }
-  catch(std::exception& e)
-  {
+  } catch (std::exception& e) {
     mexPrintf("Error constructing Tpetra MultiVector.\n");
     std::cout << e.what() << std::endl;
   }
   return mv;
 }
 
-template<>
-RCP<Tpetra_CrsMatrix_double> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double>>(const mxArray* mxa)
-{
+template <>
+RCP<Tpetra_CrsMatrix_double> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double> >(const mxArray* mxa) {
   bool success = false;
   RCP<Tpetra_CrsMatrix_double> A;
 
   int* colptr = NULL;
   int* rowind = NULL;
 
-  try
-  {
-    RCP<const Teuchos::Comm<int>> comm = rcp(new Teuchos::SerialComm<int>());
-    //numGlobalIndices is just the number of rows in the matrix
-    const size_t numGlobalIndices = mxGetM(mxa);
-    RCP<const muemex_map_type> rowMap = rcp(new muemex_map_type(numGlobalIndices, 0, comm));
+  try {
+    RCP<const Teuchos::Comm<int> > comm = rcp(new Teuchos::SerialComm<int>());
+    // numGlobalIndices is just the number of rows in the matrix
+    const size_t numGlobalIndices        = mxGetM(mxa);
+    RCP<const muemex_map_type> rowMap    = rcp(new muemex_map_type(numGlobalIndices, 0, comm));
     RCP<const muemex_map_type> domainMap = rcp(new muemex_map_type(mxGetN(mxa), 0, comm));
-    double* valueArray = mxGetPr(mxa);
-    int nc = mxGetN(mxa);
-    if(rewrap_ints)
-    {
-      //mwIndex_to_int allocates memory so must delete[] later
+    double* valueArray                   = mxGetPr(mxa);
+    int nc                               = mxGetN(mxa);
+    if (rewrap_ints) {
+      // mwIndex_to_int allocates memory so must delete[] later
       colptr = mwIndex_to_int(nc + 1, mxGetJc(mxa));
       rowind = mwIndex_to_int(colptr[nc], mxGetIr(mxa));
+    } else {
+      rowind = (int*)mxGetIr(mxa);
+      colptr = (int*)mxGetJc(mxa);
     }
-    else
-    {
-      rowind = (int*) mxGetIr(mxa);
-      colptr = (int*) mxGetJc(mxa);
-    }
-    //Need this to convert CSC colptrs to CRS row counts
+    // Need this to convert CSC colptrs to CRS row counts
     Teuchos::Array<size_t> rowCounts(numGlobalIndices);
-    for(int i = 0; i < nc; i++)
-    {
-      for(int j = colptr[i]; j < colptr[i + 1]; j++)
-      {
+    for (int i = 0; i < nc; i++) {
+      for (int j = colptr[i]; j < colptr[i + 1]; j++) {
         rowCounts[rowind[j]]++;
       }
     }
     A = rcp(new Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(rowMap, rowCounts()));
-    for(int i = 0; i < nc; i++)
-    {
-      for(int j = colptr[i]; j < colptr[i + 1]; j++)
-      {
+    for (int i = 0; i < nc; i++) {
+      for (int j = colptr[i]; j < colptr[i + 1]; j++) {
         //'array' of 1 element, containing column (in global matrix).
         Teuchos::ArrayView<mm_GlobalOrd> cols = Teuchos::ArrayView<mm_GlobalOrd>(&i, 1);
         //'array' of 1 element, containing value
@@ -368,175 +383,146 @@ RCP<Tpetra_CrsMatrix_double> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_double>>(co
       }
     }
     A->fillComplete(domainMap, rowMap);
-    if(rewrap_ints)
-    {
-      delete[] rowind; rowind = NULL;
-      delete[] colptr; colptr = NULL;
+    if (rewrap_ints) {
+      delete[] rowind;
+      rowind = NULL;
+      delete[] colptr;
+      colptr = NULL;
     }
     success = true;
-  }
-  catch(std::exception& e)
-  {
-    if(rewrap_ints)
-    {
-      if(rowind!=NULL) delete[] rowind;
-      if(colptr!=NULL) delete[] colptr;
+  } catch (std::exception& e) {
+    if (rewrap_ints) {
+      if (rowind != NULL) delete[] rowind;
+      if (colptr != NULL) delete[] colptr;
       rowind = NULL;
       colptr = NULL;
     }
     mexPrintf("Error while constructing Tpetra matrix:\n");
     std::cout << e.what() << std::endl;
   }
-  if(!success)
+  if (!success)
     mexErrMsgTxt("An error occurred while setting up a Tpetra matrix.\n");
   return A;
 }
 
-template<>
-RCP<Tpetra_CrsMatrix_complex> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_complex>>(const mxArray* mxa)
-{
+template <>
+RCP<Tpetra_CrsMatrix_complex> loadDataFromMatlab<RCP<Tpetra_CrsMatrix_complex> >(const mxArray* mxa) {
   RCP<Tpetra_CrsMatrix_complex> A;
-  //Create a map in order to create the matrix (taken from muelu basic example - complex)
-  try
-  {
-    RCP<const Teuchos::Comm<int>> comm = Tpetra::getDefaultComm();
+  // Create a map in order to create the matrix (taken from muelu basic example - complex)
+  try {
+    RCP<const Teuchos::Comm<int> > comm          = Tpetra::getDefaultComm();
     const Tpetra::global_size_t numGlobalIndices = mxGetM(mxa);
-    const mm_GlobalOrd indexBase = 0;
-    RCP<const muemex_map_type> rowMap = rcp(new muemex_map_type(numGlobalIndices, indexBase, comm));
-    RCP<const muemex_map_type> domainMap = rcp(new muemex_map_type(mxGetN(mxa), indexBase, comm));
-    double* realArray = mxGetPr(mxa);
-    double* imagArray = mxGetPi(mxa);
+    const mm_GlobalOrd indexBase                 = 0;
+    RCP<const muemex_map_type> rowMap            = rcp(new muemex_map_type(numGlobalIndices, indexBase, comm));
+    RCP<const muemex_map_type> domainMap         = rcp(new muemex_map_type(mxGetN(mxa), indexBase, comm));
+    double* realArray                            = mxGetPr(mxa);
+    double* imagArray                            = mxGetPi(mxa);
     int* colptr;
     int* rowind;
     int nc = mxGetN(mxa);
-    if(rewrap_ints)
-    {
-      //mwIndex_to_int allocates memory so must delete[] later
+    if (rewrap_ints) {
+      // mwIndex_to_int allocates memory so must delete[] later
       colptr = mwIndex_to_int(nc + 1, mxGetJc(mxa));
       rowind = mwIndex_to_int(colptr[nc], mxGetIr(mxa));
+    } else {
+      rowind = (int*)mxGetIr(mxa);
+      colptr = (int*)mxGetJc(mxa);
     }
-    else
-    {
-      rowind = (int*) mxGetIr(mxa);
-      colptr = (int*) mxGetJc(mxa);
-    }
-    //Need this to convert CSC colptrs to CRS row counts
+    // Need this to convert CSC colptrs to CRS row counts
     Teuchos::Array<size_t> rowCounts(numGlobalIndices);
-    for(int i = 0; i < nc; i++)
-    {
-      for(int j = colptr[i]; j < colptr[i + 1]; j++)
-      {
+    for (int i = 0; i < nc; i++) {
+      for (int j = colptr[i]; j < colptr[i + 1]; j++) {
         rowCounts[rowind[j]]++;
       }
     }
     A = rcp(new Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(rowMap, rowCounts()));
-    for(int i = 0; i < nc; i++)
-    {
-      for(int j = colptr[i]; j < colptr[i + 1]; j++)
-      {
-        //here assuming that complex_t will always be defined as std::complex<double>
-        //use 'value' over and over again with Teuchos::ArrayViews to insert into matrix
-        complex_t value = std::complex<double>(realArray[j], imagArray[j]);
+    for (int i = 0; i < nc; i++) {
+      for (int j = colptr[i]; j < colptr[i + 1]; j++) {
+        // here assuming that complex_t will always be defined as std::complex<double>
+        // use 'value' over and over again with Teuchos::ArrayViews to insert into matrix
+        complex_t value                       = std::complex<double>(realArray[j], imagArray[j]);
         Teuchos::ArrayView<mm_GlobalOrd> cols = Teuchos::ArrayView<mm_GlobalOrd>(&i, 1);
-        Teuchos::ArrayView<complex_t> vals = Teuchos::ArrayView<complex_t>(&value, 1);
+        Teuchos::ArrayView<complex_t> vals    = Teuchos::ArrayView<complex_t>(&value, 1);
         A->insertGlobalValues(rowind[j], cols, vals);
       }
     }
     A->fillComplete(domainMap, rowMap);
-    if(rewrap_ints)
-    {
+    if (rewrap_ints) {
       delete[] rowind;
       delete[] colptr;
     }
-  }
-  catch(std::exception& e)
-  {
+  } catch (std::exception& e) {
     mexPrintf("Error while constructing tpetra matrix:\n");
     std::cout << e.what() << std::endl;
   }
   return A;
 }
 
-template<>
-RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> loadDataFromMatlab<RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(const mxArray* mxa)
-{
-  RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> tmat = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa);
+template <>
+RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > loadDataFromMatlab<RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(const mxArray* mxa) {
+  RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > tmat = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa);
   return MueLu::TpetraCrs_To_XpetraMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(tmat);
 }
 
-template<>
-RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> loadDataFromMatlab<RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(const mxArray* mxa)
-{
-  RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> tmat = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa);
+template <>
+RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > loadDataFromMatlab<RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(const mxArray* mxa) {
+  RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > tmat = loadDataFromMatlab<RCP<Tpetra::CrsMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa);
   return MueLu::TpetraCrs_To_XpetraMatrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(tmat);
 }
 
-template<>
-RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> loadDataFromMatlab<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(const mxArray* mxa)
-{
-  RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> tpetraMV = loadDataFromMatlab<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa);
+template <>
+RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > loadDataFromMatlab<RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(const mxArray* mxa) {
+  RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > tpetraMV = loadDataFromMatlab<RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa);
   return MueLu::TpetraMultiVector_To_XpetraMultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(tpetraMV);
 }
 
-template<>
-RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> loadDataFromMatlab<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(const mxArray* mxa)
-{
-  RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> tpetraMV = loadDataFromMatlab<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>>(mxa);
+template <>
+RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > loadDataFromMatlab<RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(const mxArray* mxa) {
+  RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > tpetraMV = loadDataFromMatlab<RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > >(mxa);
   return MueLu::TpetraMultiVector_To_XpetraMultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(tpetraMV);
 }
 
-template<>
-RCP<Epetra_CrsMatrix> loadDataFromMatlab<RCP<Epetra_CrsMatrix>>(const mxArray* mxa)
-{
+template <>
+RCP<Epetra_CrsMatrix> loadDataFromMatlab<RCP<Epetra_CrsMatrix> >(const mxArray* mxa) {
   RCP<Epetra_CrsMatrix> matrix;
-  try
-  {
+  try {
     int* colptr;
     int* rowind;
     double* vals = mxGetPr(mxa);
-    int nr = mxGetM(mxa);
-    int nc = mxGetN(mxa);
-    if(rewrap_ints)
-    {
+    int nr       = mxGetM(mxa);
+    int nc       = mxGetN(mxa);
+    if (rewrap_ints) {
       colptr = mwIndex_to_int(nc + 1, mxGetJc(mxa));
       rowind = mwIndex_to_int(colptr[nc], mxGetIr(mxa));
-    }
-    else
-    {
-      rowind = (int*) mxGetIr(mxa);
-      colptr = (int*) mxGetJc(mxa);
+    } else {
+      rowind = (int*)mxGetIr(mxa);
+      colptr = (int*)mxGetJc(mxa);
     }
     Epetra_SerialComm Comm;
     Epetra_Map RangeMap(nr, 0, Comm);
     Epetra_Map DomainMap(nc, 0, Comm);
     matrix = rcp(new Epetra_CrsMatrix(Epetra_DataAccess::Copy, RangeMap, DomainMap, 0));
     /* Do the matrix assembly */
-    for(int i = 0; i < nc; i++)
-    {
-      for(int j = colptr[i]; j < colptr[i + 1]; j++)
-      {
-        //global row, # of entries, value array, column indices array
+    for (int i = 0; i < nc; i++) {
+      for (int j = colptr[i]; j < colptr[i + 1]; j++) {
+        // global row, # of entries, value array, column indices array
         matrix->InsertGlobalValues(rowind[j], 1, &vals[j], &i);
       }
     }
     matrix->FillComplete(DomainMap, RangeMap);
-    if(rewrap_ints)
-    {
-      delete [] rowind;
-      delete [] colptr;
+    if (rewrap_ints) {
+      delete[] rowind;
+      delete[] colptr;
     }
-  }
-  catch(std::exception& e)
-  {
+  } catch (std::exception& e) {
     mexPrintf("An error occurred while setting up an Epetra matrix:\n");
     std::cout << e.what() << std::endl;
   }
   return matrix;
 }
 
-template<>
-RCP<Epetra_MultiVector> loadDataFromMatlab<RCP<Epetra_MultiVector>>(const mxArray* mxa)
-{
+template <>
+RCP<Epetra_MultiVector> loadDataFromMatlab<RCP<Epetra_MultiVector> >(const mxArray* mxa) {
   int nr = mxGetM(mxa);
   int nc = mxGetN(mxa);
   Epetra_SerialComm Comm;
@@ -544,133 +530,127 @@ RCP<Epetra_MultiVector> loadDataFromMatlab<RCP<Epetra_MultiVector>>(const mxArra
   return rcp(new Epetra_MultiVector(Epetra_DataAccess::Copy, map, mxGetPr(mxa), nr, nc));
 }
 
-template<>
-RCP<MAggregates> loadDataFromMatlab<RCP<MAggregates>>(const mxArray* mxa)
-{
-  if(mxGetNumberOfElements(mxa) != 1)
+template <>
+RCP<MAggregates> loadDataFromMatlab<RCP<MAggregates> >(const mxArray* mxa) {
+  if (mxGetNumberOfElements(mxa) != 1)
     throw runtime_error("Aggregates must be individual structs in MATLAB.");
-  if(!mxIsStruct(mxa))
+  if (!mxIsStruct(mxa))
     throw runtime_error("Trying to pull aggregates from non-struct MATLAB object.");
-  //assume that in matlab aggregate structs will only be stored in a 1x1 array
-  //mxa must have the same fields as the ones declared in constructAggregates function in muelu.m for this to work
-  const int correctNumFields = 5; //change if more fields are added to the aggregates representation in constructAggregates in muelu.m
-  if(mxGetNumberOfFields(mxa) != correctNumFields)
+  // assume that in matlab aggregate structs will only be stored in a 1x1 array
+  // mxa must have the same fields as the ones declared in constructAggregates function in muelu.m for this to work
+  const int correctNumFields = 5;  // change if more fields are added to the aggregates representation in constructAggregates in muelu.m
+  if (mxGetNumberOfFields(mxa) != correctNumFields)
     throw runtime_error("Aggregates structure has wrong number of fields.");
-  //Pull MuemexData types back out
-  int nVert = *(int*) mxGetData(mxGetField(mxa, 0, "nVertices"));
-  int nAgg = *(int*) mxGetData(mxGetField(mxa, 0, "nAggregates"));
-  //Now have all the data needed to fully reconstruct the aggregate
-  //Use similar approach as UserAggregationFactory (which is written for >1 thread but will just be serial here)
-  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
-  int myRank = comm->getRank();
-  Xpetra::UnderlyingLib lib = Xpetra::UseTpetra;
-  RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t>> map = Xpetra::MapFactory<mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(lib, nVert, 0, comm);
-  RCP<MAggregates> agg = rcp(new MAggregates(map));
+  // Pull MuemexData types back out
+  int nVert = *(int*)mxGetData(mxGetField(mxa, 0, "nVertices"));
+  int nAgg  = *(int*)mxGetData(mxGetField(mxa, 0, "nAggregates"));
+  // Now have all the data needed to fully reconstruct the aggregate
+  // Use similar approach as UserAggregationFactory (which is written for >1 thread but will just be serial here)
+  RCP<const Teuchos::Comm<int> > comm                         = Teuchos::DefaultComm<int>::getComm();
+  int myRank                                                  = comm->getRank();
+  Xpetra::UnderlyingLib lib                                   = Xpetra::UseTpetra;
+  RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t> > map = Xpetra::MapFactory<mm_LocalOrd, mm_GlobalOrd, mm_node_t>::Build(lib, nVert, 0, comm);
+  RCP<MAggregates> agg                                        = rcp(new MAggregates(map));
   agg->SetNumAggregates(nAgg);
-  //Get handles for the vertex2AggId and procwinner arrays in reconstituted aggregates object
-  //this is serial so all procwinner values will be same (0)
-  ArrayRCP<mm_LocalOrd> vertex2AggId = agg->GetVertex2AggId()->getDataNonConst(0);  //the '0' means first (and only) column of multivector, since is just vector
-  ArrayRCP<mm_LocalOrd> procWinner = agg->GetProcWinner()->getDataNonConst(0);
-  //mm_LocalOrd and int are equivalent, so is ok to talk about aggSize with just 'int'
-  //Deep copy the entire vertex2AggID and isRoot arrays, which are both nVert items long
-  //At the same time, set ProcWinner
-  mxArray* vertToAggID_in = mxGetField(mxa, 0, "vertexToAggID");
-  int* vertToAggID_inArray = (int*) mxGetData(vertToAggID_in);
-  mxArray* rootNodes_in = mxGetField(mxa, 0, "rootNodes");
-  int* rootNodes_inArray = (int*) mxGetData(rootNodes_in);
-  for(int i = 0; i < nVert; i++)
-  {
+  // Get handles for the vertex2AggId and procwinner arrays in reconstituted aggregates object
+  // this is serial so all procwinner values will be same (0)
+  ArrayRCP<mm_LocalOrd> vertex2AggId = agg->GetVertex2AggId()->getDataNonConst(0);  // the '0' means first (and only) column of multivector, since is just vector
+  ArrayRCP<mm_LocalOrd> procWinner   = agg->GetProcWinner()->getDataNonConst(0);
+  // mm_LocalOrd and int are equivalent, so is ok to talk about aggSize with just 'int'
+  // Deep copy the entire vertex2AggID and isRoot arrays, which are both nVert items long
+  // At the same time, set ProcWinner
+  mxArray* vertToAggID_in  = mxGetField(mxa, 0, "vertexToAggID");
+  int* vertToAggID_inArray = (int*)mxGetData(vertToAggID_in);
+  mxArray* rootNodes_in    = mxGetField(mxa, 0, "rootNodes");
+  int* rootNodes_inArray   = (int*)mxGetData(rootNodes_in);
+  for (int i = 0; i < nVert; i++) {
     vertex2AggId[i] = vertToAggID_inArray[i];
-    procWinner[i] = myRank; //all nodes are going to be on the same proc
-    agg->SetIsRoot(i, false); //the ones that are root will be set in next loop
+    procWinner[i]   = myRank;  // all nodes are going to be on the same proc
+    agg->SetIsRoot(i, false);  // the ones that are root will be set in next loop
   }
-  for(int i = 0; i < nAgg; i++) //rootNodesToCopy is an array of node IDs which are the roots of their aggs
+  for (int i = 0; i < nAgg; i++)  // rootNodesToCopy is an array of node IDs which are the roots of their aggs
   {
     agg->SetIsRoot(rootNodes_inArray[i], true);
   }
-  //Now recompute the aggSize array the results in the object
+  // Now recompute the aggSize array the results in the object
   agg->ComputeAggregateSizes(true);
   agg->AggregatesCrossProcessors(false);
   return agg;
 }
 
-template<>
-RCP<MAmalInfo> loadDataFromMatlab<RCP<MAmalInfo>>(const mxArray* mxa)
-{
+template <>
+RCP<MAmalInfo> loadDataFromMatlab<RCP<MAmalInfo> >(const mxArray* mxa) {
   RCP<MAmalInfo> amal;
   throw runtime_error("AmalgamationInfo not supported in Muemex yet.");
   return amal;
 }
 
-template<>
-RCP<MGraph> loadDataFromMatlab<RCP<MGraph>>(const mxArray* mxa)
-{
-  //mxa must be struct with logical sparse matrix called 'edges' and Nx1 int32 array 'boundaryNodes'
-  mxArray* edges = mxGetField(mxa, 0, "edges");
+template <>
+RCP<MGraph> loadDataFromMatlab<RCP<MGraph> >(const mxArray* mxa) {
+  // mxa must be struct with logical sparse matrix called 'edges' and Nx1 int32 array 'boundaryNodes'
+  mxArray* edges         = mxGetField(mxa, 0, "edges");
   mxArray* boundaryNodes = mxGetField(mxa, 0, "boundaryNodes");
-  if(edges == NULL)
+  if (edges == NULL)
     throw runtime_error("Graph structure in MATLAB must have a field called 'edges' (logical sparse matrix)");
-  if(boundaryNodes == NULL)
+  if (boundaryNodes == NULL)
     throw runtime_error("Graph structure in MATLAB must have a field called 'boundaryNodes' (int32 array containing list of boundary nodes)");
-  int* boundaryList = (int*) mxGetData(boundaryNodes);
-  if(!mxIsSparse(edges) || mxGetClassID(edges) != mxLOGICAL_CLASS)
+  int* boundaryList = (int*)mxGetData(boundaryNodes);
+  if (!mxIsSparse(edges) || mxGetClassID(edges) != mxLOGICAL_CLASS)
     throw runtime_error("Graph edges must be stored as a logical sparse matrix.");
   // Note that Matlab stores sparse matrices in column major format.
-  mwIndex* matlabColPtrs = mxGetJc(edges);
+  mwIndex* matlabColPtrs    = mxGetJc(edges);
   mwIndex* matlabRowIndices = mxGetIr(edges);
-  mm_GlobalOrd nRows = (mm_GlobalOrd) mxGetM(edges);
+  mm_GlobalOrd nRows        = (mm_GlobalOrd)mxGetM(edges);
 
   // Create and populate row-major CRS data structures for Xpetra::TpetraCrsGraph.
 
   // calculate number of nonzeros in each row
   Teuchos::Array<int> entriesPerRow(nRows);
-  int nnz = matlabColPtrs[mxGetN(edges)]; //last entry in matlabColPtrs
-  for(int i = 0; i < nnz; i++)
+  int nnz = matlabColPtrs[mxGetN(edges)];  // last entry in matlabColPtrs
+  for (int i = 0; i < nnz; i++)
     entriesPerRow[matlabRowIndices[i]]++;
   // Populate usual row index array.  We don't need this for the Xpetra Graph ctor, but
   // it's convenient for building up the column index array, which the ctor does need.
-  Teuchos::Array<int> rows(nRows+1);
+  Teuchos::Array<int> rows(nRows + 1);
   rows[0] = 0;
-  for(int i = 0; i < nRows; i++)
-    rows[i+1] = rows[i] + entriesPerRow[i];
-  Teuchos::Array<int> cols(nnz);     //column index array
-  Teuchos::Array<int> insertionsPerRow(nRows,0); //track of #insertions done per row
+  for (int i = 0; i < nRows; i++)
+    rows[i + 1] = rows[i] + entriesPerRow[i];
+  Teuchos::Array<int> cols(nnz);                   // column index array
+  Teuchos::Array<int> insertionsPerRow(nRows, 0);  // track of #insertions done per row
   int ncols = mxGetN(edges);
-  for (int colNum=0; colNum<ncols; ++colNum) {
+  for (int colNum = 0; colNum < ncols; ++colNum) {
     int ci = matlabColPtrs[colNum];
-    for (int j=ci; j<Teuchos::as<int>(matlabColPtrs[colNum+1]); ++j) {
-      int rowNum = matlabRowIndices[j];
-      cols[ rows[rowNum] + insertionsPerRow[rowNum] ] = colNum;
+    for (int j = ci; j < Teuchos::as<int>(matlabColPtrs[colNum + 1]); ++j) {
+      int rowNum                                    = matlabRowIndices[j];
+      cols[rows[rowNum] + insertionsPerRow[rowNum]] = colNum;
       insertionsPerRow[rowNum]++;
     }
   }
-  //Find maximum
+  // Find maximum
   int maxNzPerRow = 0;
-  for(int i = 0; i < nRows; i++) {
-    if(maxNzPerRow < entriesPerRow[i])
+  for (int i = 0; i < nRows; i++) {
+    if (maxNzPerRow < entriesPerRow[i])
       maxNzPerRow = entriesPerRow[i];
   }
 
-  RCP<const Teuchos::Comm<mm_GlobalOrd>> comm = rcp(new Teuchos::SerialComm<mm_GlobalOrd>());
+  RCP<const Teuchos::Comm<mm_GlobalOrd> > comm = rcp(new Teuchos::SerialComm<mm_GlobalOrd>());
   typedef Xpetra::TpetraMap<mm_LocalOrd, mm_GlobalOrd, mm_node_t> MMap;
   RCP<MMap> map = rcp(new MMap(nRows, 0, comm));
   typedef Xpetra::TpetraCrsGraph<mm_LocalOrd, mm_GlobalOrd, mm_node_t> TpetraGraph;
-  RCP<TpetraGraph> tgraph = rcp(new TpetraGraph(map, (size_t) maxNzPerRow));
-  //Populate tgraph in compressed-row format. Must get each row individually...
-  for(int i = 0; i < nRows; ++i) {
-    tgraph->insertGlobalIndices((mm_GlobalOrd) i, cols(rows[i],entriesPerRow[i]));
+  RCP<TpetraGraph> tgraph = rcp(new TpetraGraph(map, (size_t)maxNzPerRow));
+  // Populate tgraph in compressed-row format. Must get each row individually...
+  for (int i = 0; i < nRows; ++i) {
+    tgraph->insertGlobalIndices((mm_GlobalOrd)i, cols(rows[i], entriesPerRow[i]));
   }
   tgraph->fillComplete(map, map);
   RCP<MGraph> mgraph = rcp(new MueLu::Graph<mm_LocalOrd, mm_GlobalOrd, mm_node_t>(tgraph));
-  //Set boundary nodes
+  // Set boundary nodes
   int numBoundaryNodes = mxGetNumberOfElements(boundaryNodes);
-  bool* boundaryFlags = new bool[nRows];
-  for(int i = 0; i < nRows; i++)
-  {
+  bool* boundaryFlags  = new bool[nRows];
+  for (int i = 0; i < nRows; i++) {
     boundaryFlags[i] = false;
   }
-  for(int i = 0; i < numBoundaryNodes; i++)
-  {
+  for (int i = 0; i < numBoundaryNodes; i++) {
     boundaryFlags[boundaryList[i]] = true;
   }
   ArrayRCP<bool> boundaryNodesInput(boundaryFlags, 0, nRows, true);
@@ -678,24 +658,20 @@ RCP<MGraph> loadDataFromMatlab<RCP<MGraph>>(const mxArray* mxa)
   return mgraph;
 }
 
-
 #ifdef HAVE_MUELU_INTREPID2
-template<>
-RCP<FieldContainer_ordinal> loadDataFromMatlab<RCP<FieldContainer_ordinal>>(const mxArray* mxa)
-{
-  if(mxGetClassID(mxa) != mxINT32_CLASS)
+template <>
+RCP<FieldContainer_ordinal> loadDataFromMatlab<RCP<FieldContainer_ordinal> >(const mxArray* mxa) {
+  if (mxGetClassID(mxa) != mxINT32_CLASS)
     throw runtime_error("FieldContainer must have integer storage entries");
 
-  int *data = (int *) mxGetData(mxa);
-  int nr = mxGetM(mxa);
-  int nc = mxGetN(mxa);
+  int* data = (int*)mxGetData(mxa);
+  int nr    = mxGetM(mxa);
+  int nc    = mxGetN(mxa);
 
-  RCP<FieldContainer_ordinal> fc = rcp(new FieldContainer_ordinal("FC from Matlab",nr,nc));
-  for(int col = 0; col < nc; col++)
-  {
-    for(int row = 0; row < nr; row++)
-    {
-      (*fc)(row,col) = data[col * nr + row];
+  RCP<FieldContainer_ordinal> fc = rcp(new FieldContainer_ordinal("FC from Matlab", nr, nc));
+  for (int col = 0; col < nc; col++) {
+    for (int row = 0; row < nr; row++) {
+      (*fc)(row, col) = data[col * nr + row];
     }
   }
   return fc;
@@ -706,336 +682,301 @@ RCP<FieldContainer_ordinal> loadDataFromMatlab<RCP<FieldContainer_ordinal>>(cons
 /* saveDataToMatlab                */
 /* ******************************* */
 
-template<>
-mxArray* saveDataToMatlab(int& data)
-{
-  mwSize dims[] = {1, 1};
-  mxArray* mxa = mxCreateNumericArray(2, dims, mxINT32_CLASS, mxREAL);
-  *((int*) mxGetData(mxa)) = data;
+template <>
+mxArray* saveDataToMatlab(int& data) {
+  mwSize dims[]           = {1, 1};
+  mxArray* mxa            = mxCreateNumericArray(2, dims, mxINT32_CLASS, mxREAL);
+  *((int*)mxGetData(mxa)) = data;
   return mxa;
 }
 
-template<>
-mxArray* saveDataToMatlab(bool& data)
-{
-  mwSize dims[] = {1, 1};
-  mxArray* mxa = mxCreateLogicalArray(2, dims);
-  *((bool*) mxGetData(mxa)) = data;
+template <>
+mxArray* saveDataToMatlab(bool& data) {
+  mwSize dims[]            = {1, 1};
+  mxArray* mxa             = mxCreateLogicalArray(2, dims);
+  *((bool*)mxGetData(mxa)) = data;
   return mxa;
 }
 
-template<>
-mxArray* saveDataToMatlab(double& data)
-{
+template <>
+mxArray* saveDataToMatlab(double& data) {
   return mxCreateDoubleScalar(data);
 }
 
-template<>
-mxArray* saveDataToMatlab(complex_t& data)
-{
-  mwSize dims[] = {1, 1};
-  mxArray* mxa = mxCreateNumericArray(2, dims, mxDOUBLE_CLASS, mxCOMPLEX);
-  *((double*) mxGetPr(mxa)) = real<double>(data);
-  *((double*) mxGetPi(mxa)) = imag<double>(data);
+template <>
+mxArray* saveDataToMatlab(complex_t& data) {
+  mwSize dims[]            = {1, 1};
+  mxArray* mxa             = mxCreateNumericArray(2, dims, mxDOUBLE_CLASS, mxCOMPLEX);
+  *((double*)mxGetPr(mxa)) = real<double>(data);
+  *((double*)mxGetPi(mxa)) = imag<double>(data);
   return mxa;
 }
 
-template<>
-mxArray* saveDataToMatlab(string& data)
-{
+template <>
+mxArray* saveDataToMatlab(string& data) {
   return mxCreateString(data.c_str());
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra_map>& data)
-{
-  //Precondition: Memory has already been allocated by MATLAB for the array.
-  int nc = data->getGlobalNumElements();
-  int nr = 1;
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_map>& data) {
+  // Precondition: Memory has already been allocated by MATLAB for the array.
+  int nc          = data->getGlobalNumElements();
+  int nr          = 1;
   mxArray* output = createMatlabMultiVector<double>(nr, nc);
-  double* array = (double*) malloc(sizeof(double) * nr * nc);
-  for(int col = 0; col < nc; col++)
-  {
+  double* array   = (double*)malloc(sizeof(double) * nr * nc);
+  for (int col = 0; col < nc; col++) {
     mm_GlobalOrd gid = data->getGlobalElement(col);
-    array[col] = Teuchos::as<double>(gid);
+    array[col]       = Teuchos::as<double>(gid);
   }
   fillMatlabArray<double>(array, output, nc * nr);
   free(array);
   return output;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra_ordinal_vector>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_ordinal_vector>& data) {
   mwSize len = data->getGlobalLength();
-  //create a single column vector
-  mwSize dimensions[] = {len, 1};
-  mxArray* rv = mxCreateNumericArray(2, dimensions, mxINT32_CLASS, mxREAL);
-  int* dataPtr = (int*) mxGetData(rv);
+  // create a single column vector
+  mwSize dimensions[]             = {len, 1};
+  mxArray* rv                     = mxCreateNumericArray(2, dimensions, mxINT32_CLASS, mxREAL);
+  int* dataPtr                    = (int*)mxGetData(rv);
   ArrayRCP<const mm_LocalOrd> arr = data->getData(0);
-  for(int i = 0; i < int(data->getGlobalLength()); i++)
-  {
+  for (int i = 0; i < int(data->getGlobalLength()); i++) {
     dataPtr[i] = arr[i];
   }
   return rv;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Tpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> >& data) {
   RCP<Xpetra_MultiVector_double> xmv = MueLu::TpetraMultiVector_To_XpetraMultiVector(data);
   return saveDataToMatlab(xmv);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Tpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> >& data) {
   RCP<Xpetra_MultiVector_complex> xmv = MueLu::TpetraMultiVector_To_XpetraMultiVector(data);
   return saveDataToMatlab(xmv);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Tpetra_CrsMatrix_double>& data)
-{
-  RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> xmat = TpetraCrs_To_XpetraMatrix(data);
+template <>
+mxArray* saveDataToMatlab(RCP<Tpetra_CrsMatrix_double>& data) {
+  RCP<Xpetra::Matrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > xmat = TpetraCrs_To_XpetraMatrix(data);
   return saveDataToMatlab(xmat);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Tpetra_CrsMatrix_complex>& data)
-{
-  RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> xmat = TpetraCrs_To_XpetraMatrix(data);
+template <>
+mxArray* saveDataToMatlab(RCP<Tpetra_CrsMatrix_complex>& data) {
+  RCP<Xpetra::Matrix<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > xmat = TpetraCrs_To_XpetraMatrix(data);
   return saveDataToMatlab(xmat);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_double>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_double>& data) {
   typedef double Scalar;
   // Compute global constants, if we need them
   Teuchos::rcp_const_cast<Xpetra_CrsGraph>(data->getCrsGraph())->computeGlobalConstants();
 
-  int nr = data->getGlobalNumRows();
-  int nc = data->getGlobalNumCols();
-  int nnz = data->getGlobalNumEntries();
+  int nr                                 = data->getGlobalNumRows();
+  int nc                                 = data->getGlobalNumCols();
+  int nnz                                = data->getGlobalNumEntries();
 
 #ifdef VERBOSE_OUTPUT
   RCP<Teuchos::FancyOStream> fancyStream = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
   mat->describe(*fancyStream, Teuchos::VERB_EXTREME);
 #endif
   mxArray* mxa = createMatlabSparse<Scalar>(nr, nc, nnz);
-  mwIndex* ir = mxGetIr(mxa);
-  mwIndex* jc = mxGetJc(mxa);
-  for(int i = 0; i < nc + 1; i++)
-  {
+  mwIndex* ir  = mxGetIr(mxa);
+  mwIndex* jc  = mxGetJc(mxa);
+  for (int i = 0; i < nc + 1; i++) {
     jc[i] = 0;
   }
 
   size_t maxEntriesPerRow = data->getGlobalMaxNumRowEntries();
-  if(maxEntriesPerRow == Teuchos::OrdinalTraits<size_t>::invalid() || maxEntriesPerRow == 0) maxEntriesPerRow = data->getLocalMaxNumRowEntries();
+  if (maxEntriesPerRow == Teuchos::OrdinalTraits<size_t>::invalid() || maxEntriesPerRow == 0) maxEntriesPerRow = data->getLocalMaxNumRowEntries();
 
   int* rowProgress = new int[nc];
-  //The array that will be copied to Pr and (if complex) Pi later
+  // The array that will be copied to Pr and (if complex) Pi later
   Scalar* sparseVals = new Scalar[nnz];
   size_t numEntries;
-  if(data->isLocallyIndexed())
-  {
+  if (data->isLocallyIndexed()) {
     Scalar* rowValArray = new Scalar[maxEntriesPerRow];
     Teuchos::ArrayView<Scalar> rowVals(rowValArray, maxEntriesPerRow);
     mm_LocalOrd* rowIndicesArray = new mm_LocalOrd[maxEntriesPerRow];
     Teuchos::ArrayView<mm_LocalOrd> rowIndices(rowIndicesArray, maxEntriesPerRow);
-    for(mm_LocalOrd m = 0; m < nr; m++)       //All rows in the Xpetra matrix
+    for (mm_LocalOrd m = 0; m < nr; m++)  // All rows in the Xpetra matrix
     {
-      data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);     //Get the row
-      for(mm_LocalOrd entry = 0; entry < int(numEntries); entry++)  //All entries in row
+      data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);     // Get the row
+      for (mm_LocalOrd entry = 0; entry < int(numEntries); entry++)  // All entries in row
       {
-        jc[rowIndices[entry] + 1]++; //for each entry, increase jc for the entry's column
+        jc[rowIndices[entry] + 1]++;  // for each entry, increase jc for the entry's column
       }
     }
 
-    //now jc holds the number of elements in each column, but needs cumulative sum over all previous columns also
+    // now jc holds the number of elements in each column, but needs cumulative sum over all previous columns also
     int entriesAccum = 0;
-    for(int n = 0; n <= nc; n++)
-    {
+    for (int n = 0; n <= nc; n++) {
       int temp = entriesAccum;
       entriesAccum += jc[n];
       jc[n] += temp;
     }
-    //Jc now populated with colptrs
-    for(int i = 0; i < nc; i++)
-    {
+    // Jc now populated with colptrs
+    for (int i = 0; i < nc; i++) {
       rowProgress[i] = 0;
     }
-    //Row progress values like jc but keep track as the MATLAB matrix is being filled in
-    for(mm_LocalOrd m = 0; m < nr; m++)       //rows
+    // Row progress values like jc but keep track as the MATLAB matrix is being filled in
+    for (mm_LocalOrd m = 0; m < nr; m++)  // rows
     {
       data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);
-      for(mm_LocalOrd i = 0; i < int(numEntries); i++)      //entries in row m (NOT columns)
+      for (mm_LocalOrd i = 0; i < int(numEntries); i++)  // entries in row m (NOT columns)
       {
-        //row is m, col is rowIndices[i], val is rowVals[i]
-        mm_LocalOrd col = rowIndices[i];
-        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];      //Set value
-        ir[jc[col] + rowProgress[col]] = m;                                               //Set row at which value occurs
+        // row is m, col is rowIndices[i], val is rowVals[i]
+        mm_LocalOrd col                        = rowIndices[i];
+        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];  // Set value
+        ir[jc[col] + rowProgress[col]]         = m;           // Set row at which value occurs
         rowProgress[col]++;
       }
     }
     delete[] rowIndicesArray;
-  }
-  else
-  {
+  } else {
     Teuchos::ArrayView<const mm_GlobalOrd> rowIndices;
     Teuchos::ArrayView<const Scalar> rowVals;
-    for(mm_GlobalOrd m = 0; m < nr; m++)
-    {
+    for (mm_GlobalOrd m = 0; m < nr; m++) {
       data->getGlobalRowView(m, rowIndices, rowVals);
-      for(mm_GlobalOrd n = 0; n < rowIndices.size(); n++)
-      {
+      for (mm_GlobalOrd n = 0; n < rowIndices.size(); n++) {
         jc[rowIndices[n] + 1]++;
       }
     }
-    //Last element of jc is just nnz
+    // Last element of jc is just nnz
     jc[nc] = nnz;
-    //Jc now populated with colptrs
-    for(int i = 0; i < nc; i++)
-    {
+    // Jc now populated with colptrs
+    for (int i = 0; i < nc; i++) {
       rowProgress[i] = 0;
     }
     int entriesAccum = 0;
-    for(int n = 0; n <= nc; n++)
-    {
+    for (int n = 0; n <= nc; n++) {
       int temp = entriesAccum;
       entriesAccum += jc[n];
       jc[n] += temp;
     }
-    //Row progress values like jc but keep track as the MATLAB matrix is being filled in
-    for(mm_GlobalOrd m = 0; m < nr; m++)                     //rows
+    // Row progress values like jc but keep track as the MATLAB matrix is being filled in
+    for (mm_GlobalOrd m = 0; m < nr; m++)  // rows
     {
       data->getGlobalRowView(m, rowIndices, rowVals);
-      for(mm_LocalOrd i = 0; i < rowIndices.size(); i++)     //entries in row m
+      for (mm_LocalOrd i = 0; i < rowIndices.size(); i++)  // entries in row m
       {
-        mm_GlobalOrd col = rowIndices[i];                    //row is m, col is rowIndices[i], val is rowVals[i]
-        sparseVals[jc[col] + rowProgress[col]] = rowVals[i]; //Set value
-        ir[jc[col] + rowProgress[col]] = m;                  //Set row at which value occurs
+        mm_GlobalOrd col                       = rowIndices[i];  // row is m, col is rowIndices[i], val is rowVals[i]
+        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];     // Set value
+        ir[jc[col] + rowProgress[col]]         = m;              // Set row at which value occurs
         rowProgress[col]++;
       }
     }
   }
-  //finally, copy sparseVals into pr (and pi, if complex)
+  // finally, copy sparseVals into pr (and pi, if complex)
   fillMatlabArray<Scalar>(sparseVals, mxa, nnz);
   delete[] sparseVals;
   delete[] rowProgress;
   return mxa;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_complex>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra_Matrix_complex>& data) {
   typedef complex_t Scalar;
 
   // Compute global constants, if we need them
   Teuchos::rcp_const_cast<Xpetra_CrsGraph>(data->getCrsGraph())->computeGlobalConstants();
 
-  int nr = data->getGlobalNumRows();
-  int nc = data->getGlobalNumCols();
-  int nnz = data->getGlobalNumEntries();
+  int nr                                 = data->getGlobalNumRows();
+  int nc                                 = data->getGlobalNumCols();
+  int nnz                                = data->getGlobalNumEntries();
 #ifdef VERBOSE_OUTPUT
   RCP<Teuchos::FancyOStream> fancyStream = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
   mat->describe(*fancyStream, Teuchos::VERB_EXTREME);
 #endif
   mxArray* mxa = createMatlabSparse<Scalar>(nr, nc, nnz);
-  mwIndex* ir = mxGetIr(mxa);
-  mwIndex* jc = mxGetJc(mxa);
-  for(int i = 0; i < nc + 1; i++)
-  {
+  mwIndex* ir  = mxGetIr(mxa);
+  mwIndex* jc  = mxGetJc(mxa);
+  for (int i = 0; i < nc + 1; i++) {
     jc[i] = 0;
   }
   size_t maxEntriesPerRow = data->getGlobalMaxNumRowEntries();
-  int* rowProgress = new int[nc];
-  //The array that will be copied to Pr and (if complex) Pi later
+  int* rowProgress        = new int[nc];
+  // The array that will be copied to Pr and (if complex) Pi later
   Scalar* sparseVals = new Scalar[nnz];
   size_t numEntries;
-  if(data->isLocallyIndexed())
-  {
+  if (data->isLocallyIndexed()) {
     Scalar* rowValArray = new Scalar[maxEntriesPerRow];
     Teuchos::ArrayView<Scalar> rowVals(rowValArray, maxEntriesPerRow);
     mm_LocalOrd* rowIndicesArray = new mm_LocalOrd[maxEntriesPerRow];
     Teuchos::ArrayView<mm_LocalOrd> rowIndices(rowIndicesArray, maxEntriesPerRow);
-    for(mm_LocalOrd m = 0; m < nr; m++)       //All rows in the Xpetra matrix
+    for (mm_LocalOrd m = 0; m < nr; m++)  // All rows in the Xpetra matrix
     {
-      data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);     //Get the row
-      for(mm_LocalOrd entry = 0; entry < int(numEntries); entry++)  //All entries in row
+      data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);     // Get the row
+      for (mm_LocalOrd entry = 0; entry < int(numEntries); entry++)  // All entries in row
       {
-        jc[rowIndices[entry] + 1]++; //for each entry, increase jc for the entry's column
+        jc[rowIndices[entry] + 1]++;  // for each entry, increase jc for the entry's column
       }
     }
-    //now jc holds the number of elements in each column, but needs cumulative sum over all previous columns also
+    // now jc holds the number of elements in each column, but needs cumulative sum over all previous columns also
     int entriesAccum = 0;
-    for(int n = 0; n <= nc; n++)
-    {
+    for (int n = 0; n <= nc; n++) {
       int temp = entriesAccum;
       entriesAccum += jc[n];
       jc[n] += temp;
     }
-    //Jc now populated with colptrs
-    for(int i = 0; i < nc; i++)
-    {
+    // Jc now populated with colptrs
+    for (int i = 0; i < nc; i++) {
       rowProgress[i] = 0;
     }
-    //Row progress values like jc but keep track as the MATLAB matrix is being filled in
-    for(mm_LocalOrd m = 0; m < nr; m++)       //rows
+    // Row progress values like jc but keep track as the MATLAB matrix is being filled in
+    for (mm_LocalOrd m = 0; m < nr; m++)  // rows
     {
       data->getLocalRowCopy(m, rowIndices, rowVals, numEntries);
-      for(mm_LocalOrd i = 0; i < int(numEntries); i++)      //entries in row m (NOT columns)
+      for (mm_LocalOrd i = 0; i < int(numEntries); i++)  // entries in row m (NOT columns)
       {
-        //row is m, col is rowIndices[i], val is rowVals[i]
-        mm_LocalOrd col = rowIndices[i];
-        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];      //Set value
-        ir[jc[col] + rowProgress[col]] = m;                                               //Set row at which value occurs
+        // row is m, col is rowIndices[i], val is rowVals[i]
+        mm_LocalOrd col                        = rowIndices[i];
+        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];  // Set value
+        ir[jc[col] + rowProgress[col]]         = m;           // Set row at which value occurs
         rowProgress[col]++;
       }
     }
     delete[] rowIndicesArray;
-  }
-  else
-  {
+  } else {
     Teuchos::ArrayView<const mm_GlobalOrd> rowIndices;
     Teuchos::ArrayView<const Scalar> rowVals;
-    for(mm_GlobalOrd m = 0; m < nr; m++)
-    {
+    for (mm_GlobalOrd m = 0; m < nr; m++) {
       data->getGlobalRowView(m, rowIndices, rowVals);
-      for(mm_GlobalOrd n = 0; n < rowIndices.size(); n++)
-      {
+      for (mm_GlobalOrd n = 0; n < rowIndices.size(); n++) {
         jc[rowIndices[n] + 1]++;
       }
     }
-    //Last element of jc is just nnz
+    // Last element of jc is just nnz
     jc[nc] = nnz;
-    //Jc now populated with colptrs
-    for(int i = 0; i < nc; i++)
-    {
+    // Jc now populated with colptrs
+    for (int i = 0; i < nc; i++) {
       rowProgress[i] = 0;
     }
     int entriesAccum = 0;
-    for(int n = 0; n <= nc; n++)
-    {
+    for (int n = 0; n <= nc; n++) {
       int temp = entriesAccum;
       entriesAccum += jc[n];
       jc[n] += temp;
     }
-    //Row progress values like jc but keep track as the MATLAB matrix is being filled in
-    for(mm_GlobalOrd m = 0; m < nr; m++)                     //rows
+    // Row progress values like jc but keep track as the MATLAB matrix is being filled in
+    for (mm_GlobalOrd m = 0; m < nr; m++)  // rows
     {
       data->getGlobalRowView(m, rowIndices, rowVals);
-      for(mm_LocalOrd i = 0; i < rowIndices.size(); i++)     //entries in row m
+      for (mm_LocalOrd i = 0; i < rowIndices.size(); i++)  // entries in row m
       {
-        mm_GlobalOrd col = rowIndices[i];                    //row is m, col is rowIndices[i], val is rowVals[i]
-        sparseVals[jc[col] + rowProgress[col]] = rowVals[i]; //Set value
-        ir[jc[col] + rowProgress[col]] = m;                  //Set row at which value occurs
+        mm_GlobalOrd col                       = rowIndices[i];  // row is m, col is rowIndices[i], val is rowVals[i]
+        sparseVals[jc[col] + rowProgress[col]] = rowVals[i];     // Set value
+        ir[jc[col] + rowProgress[col]]         = m;              // Set row at which value occurs
         rowProgress[col]++;
       }
     }
   }
-  //finally, copy sparseVals into pr (and pi, if complex)
+  // finally, copy sparseVals into pr (and pi, if complex)
   fillMatlabArray<Scalar>(sparseVals, mxa, nnz);
   delete[] sparseVals;
   delete[] rowProgress;
@@ -1065,19 +1006,16 @@ mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<Scalar, mm_LocalOrd, mm_Global
 }
 */
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>& data)
-{
-  //Precondition: Memory has already been allocated by MATLAB for the array.
-  int nr = data->getGlobalLength();
-  int nc = data->getNumVectors();
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t> >& data) {
+  // Precondition: Memory has already been allocated by MATLAB for the array.
+  int nr          = data->getGlobalLength();
+  int nc          = data->getNumVectors();
   mxArray* output = createMatlabMultiVector<double>(nr, nc);
-  double* array = (double*) malloc(sizeof(double) * nr * nc);
-  for(int col = 0; col < nc; col++)
-  {
+  double* array   = (double*)malloc(sizeof(double) * nr * nc);
+  for (int col = 0; col < nc; col++) {
     Teuchos::ArrayRCP<const double> colData = data->getData(col);
-    for(int row = 0; row < nr; row++)
-    {
+    for (int row = 0; row < nr; row++) {
       array[col * nr + row] = colData[row];
     }
   }
@@ -1086,19 +1024,16 @@ mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<double, mm_LocalOrd, mm_Global
   return output;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t>>& data)
-{
-  //Precondition: Memory has already been allocated by MATLAB for the array.
-  int nr = data->getGlobalLength();
-  int nc = data->getNumVectors();
-  mxArray* output = createMatlabMultiVector<complex_t>(nr, nc);
-  complex_t* array = (complex_t*) malloc(sizeof(complex_t) * nr * nc);
-  for(int col = 0; col < nc; col++)
-  {
+template <>
+mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_GlobalOrd, mm_node_t> >& data) {
+  // Precondition: Memory has already been allocated by MATLAB for the array.
+  int nr           = data->getGlobalLength();
+  int nc           = data->getNumVectors();
+  mxArray* output  = createMatlabMultiVector<complex_t>(nr, nc);
+  complex_t* array = (complex_t*)malloc(sizeof(complex_t) * nr * nc);
+  for (int col = 0; col < nc; col++) {
     Teuchos::ArrayRCP<const complex_t> colData = data->getData(col);
-    for(int row = 0; row < nr; row++)
-    {
+    for (int row = 0; row < nr; row++) {
       array[col * nr + row] = colData[row];
     }
   }
@@ -1107,206 +1042,180 @@ mxArray* saveDataToMatlab(RCP<Xpetra::MultiVector<complex_t, mm_LocalOrd, mm_Glo
   return output;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Epetra_CrsMatrix>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Epetra_CrsMatrix>& data) {
   RCP<Xpetra_Matrix_double> xmat = EpetraCrs_To_XpetraMatrix<double, mm_LocalOrd, mm_GlobalOrd, mm_node_t>(data);
   return saveDataToMatlab(xmat);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<Epetra_MultiVector>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<Epetra_MultiVector>& data) {
   mxArray* output = mxCreateDoubleMatrix(data->GlobalLength(), data->NumVectors(), mxREAL);
   double* dataPtr = mxGetPr(output);
   data->ExtractCopy(dataPtr, data->GlobalLength());
   return output;
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<MAggregates>& data)
-{
-  //Set up array of inputs for matlab constructAggregates
+template <>
+mxArray* saveDataToMatlab(RCP<MAggregates>& data) {
+  // Set up array of inputs for matlab constructAggregates
   int numNodes = data->GetVertex2AggId()->getData(0).size();
-  int numAggs = data->GetNumAggregates();
+  int numAggs  = data->GetNumAggregates();
   mxArray* dataIn[5];
-  mwSize singleton[] = {1, 1};
-  dataIn[0] = mxCreateNumericArray(2, singleton, mxINT32_CLASS, mxREAL);
-  *((int*) mxGetData(dataIn[0])) = numNodes;
-  dataIn[1] = mxCreateNumericArray(2, singleton, mxINT32_CLASS, mxREAL);
-  *((int*) mxGetData(dataIn[1])) = numAggs;
-  mwSize nodeArrayDims[] = {(mwSize) numNodes, 1}; //dimensions for Nx1 array, where N is number of nodes (vert2Agg)
-  dataIn[2] = mxCreateNumericArray(2, nodeArrayDims, mxINT32_CLASS, mxREAL);
-  int* vtaid = (int*) mxGetData(dataIn[2]);
+  mwSize singleton[]                        = {1, 1};
+  dataIn[0]                                 = mxCreateNumericArray(2, singleton, mxINT32_CLASS, mxREAL);
+  *((int*)mxGetData(dataIn[0]))             = numNodes;
+  dataIn[1]                                 = mxCreateNumericArray(2, singleton, mxINT32_CLASS, mxREAL);
+  *((int*)mxGetData(dataIn[1]))             = numAggs;
+  mwSize nodeArrayDims[]                    = {(mwSize)numNodes, 1};  // dimensions for Nx1 array, where N is number of nodes (vert2Agg)
+  dataIn[2]                                 = mxCreateNumericArray(2, nodeArrayDims, mxINT32_CLASS, mxREAL);
+  int* vtaid                                = (int*)mxGetData(dataIn[2]);
   ArrayRCP<const mm_LocalOrd> vertexToAggID = data->GetVertex2AggId()->getData(0);
-  for(int i = 0; i < numNodes; i++)
-  {
+  for (int i = 0; i < numNodes; i++) {
     vtaid[i] = vertexToAggID[i];
   }
-  mwSize aggArrayDims[] = {(mwSize) numAggs, 1}; //dims for Nx1 array, where N is number of aggregates (rootNodes, aggSizes)
-  dataIn[3] = mxCreateNumericArray(2, aggArrayDims, mxINT32_CLASS, mxREAL);
-  //First, find out if the aggregates even have 1 root node per aggregate. If not, assume roots are invalid and assign ourselves
+  mwSize aggArrayDims[] = {(mwSize)numAggs, 1};  // dims for Nx1 array, where N is number of aggregates (rootNodes, aggSizes)
+  dataIn[3]             = mxCreateNumericArray(2, aggArrayDims, mxINT32_CLASS, mxREAL);
+  // First, find out if the aggregates even have 1 root node per aggregate. If not, assume roots are invalid and assign ourselves
   int totalRoots = 0;
-  for(int i = 0; i < numNodes; i++)
-  {
-    if(data->IsRoot(i))
+  for (int i = 0; i < numNodes; i++) {
+    if (data->IsRoot(i))
       totalRoots++;
   }
   bool reassignRoots = false;
-  if(totalRoots != numAggs)
-  {
-    cout << endl << "Warning: Number of root nodes and number of aggregates do not match." << endl;
-    cout << "Will reassign root nodes when writing aggregates to matlab." << endl << endl;
+  if (totalRoots != numAggs) {
+    cout << endl
+         << "Warning: Number of root nodes and number of aggregates do not match." << endl;
+    cout << "Will reassign root nodes when writing aggregates to matlab." << endl
+         << endl;
     reassignRoots = true;
   }
-  int* rn = (int*) mxGetData(dataIn[3]); //list of root nodes (in no particular order)
+  int* rn = (int*)mxGetData(dataIn[3]);  // list of root nodes (in no particular order)
   {
-    if(reassignRoots)
-    {
-      //For each aggregate, just pick the first node we see in it and set it as root
-      int lastFoundNode = 0; //heuristic for speed, a node in aggregate N+1 is likely to come very soon after a node in agg N
-      for(int i = 0; i < numAggs; i++)
-      {
+    if (reassignRoots) {
+      // For each aggregate, just pick the first node we see in it and set it as root
+      int lastFoundNode = 0;  // heuristic for speed, a node in aggregate N+1 is likely to come very soon after a node in agg N
+      for (int i = 0; i < numAggs; i++) {
         rn[i] = -1;
-        for(int j = lastFoundNode; j < lastFoundNode + numNodes; j++)
-        {
+        for (int j = lastFoundNode; j < lastFoundNode + numNodes; j++) {
           int index = j % numNodes;
-          if(vertexToAggID[index] == i)
-          {
-            rn[i] = index;
+          if (vertexToAggID[index] == i) {
+            rn[i]         = index;
             lastFoundNode = index;
           }
         }
         TEUCHOS_TEST_FOR_EXCEPTION(rn[i] == -1, runtime_error, "Invalid aggregates: Couldn't find any node in aggregate #" << i << ".");
       }
-    }
-    else
-    {
-      int i = 0; //iterates over aggregate IDs
-      for(int j = 0; j < numNodes; j++)
-      {
-        if(data->IsRoot(j))
-        {
-          if(i == numAggs)
+    } else {
+      int i = 0;  // iterates over aggregate IDs
+      for (int j = 0; j < numNodes; j++) {
+        if (data->IsRoot(j)) {
+          if (i == numAggs)
             throw runtime_error("Cannot store invalid aggregates in MATLAB - more root nodes than aggregates.");
-          rn[i] = j; //now we know this won't go out of bounds (rn's underlying matlab array is numAggs in length)
+          rn[i] = j;  // now we know this won't go out of bounds (rn's underlying matlab array is numAggs in length)
           i++;
         }
       }
-      if(i + 1 < numAggs)
+      if (i + 1 < numAggs)
         throw runtime_error("Cannot store invalid aggregates in MATLAB - fewer root nodes than aggregates.");
     }
   }
-  dataIn[4] = mxCreateNumericArray(1, aggArrayDims, mxINT32_CLASS, mxREAL);
-  int* as = (int*) mxGetData(dataIn[4]); //list of aggregate sizes
+  dataIn[4]                      = mxCreateNumericArray(1, aggArrayDims, mxINT32_CLASS, mxREAL);
+  int* as                        = (int*)mxGetData(dataIn[4]);  // list of aggregate sizes
   ArrayRCP<mm_LocalOrd> aggSizes = data->ComputeAggregateSizes();
-  for(int i = 0; i < numAggs; i++)
-  {
+  for (int i = 0; i < numAggs; i++) {
     as[i] = aggSizes[i];
   }
   mxArray* matlabAggs[1];
   int result = mexCallMATLAB(1, matlabAggs, 5, dataIn, "constructAggregates");
-  if(result != 0)
+  if (result != 0)
     throw runtime_error("Matlab encountered an error while constructing aggregates struct.");
   return matlabAggs[0];
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<MAmalInfo>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<MAmalInfo>& data) {
   throw runtime_error("AmalgamationInfo not supported in MueMex yet.");
   return mxCreateDoubleScalar(0);
 }
 
-template<>
-mxArray* saveDataToMatlab(RCP<MGraph>& data)
-{
-  int numEntries = (int) data->GetGlobalNumEdges();
-  int numRows = (int) data->GetDomainMap()->getGlobalNumElements(); //assume numRows == numCols
-  mxArray* mat = mxCreateSparseLogicalMatrix(numRows, numRows, numEntries);
-  mxLogical* outData = (mxLogical*) mxGetData(mat);
-  mwIndex* rowInds = mxGetIr(mat);
-  mwIndex* colPtrs = mxGetJc(mat);
+template <>
+mxArray* saveDataToMatlab(RCP<MGraph>& data) {
+  int numEntries        = (int)data->GetGlobalNumEdges();
+  int numRows           = (int)data->GetDomainMap()->getGlobalNumElements();  // assume numRows == numCols
+  mxArray* mat          = mxCreateSparseLogicalMatrix(numRows, numRows, numEntries);
+  mxLogical* outData    = (mxLogical*)mxGetData(mat);
+  mwIndex* rowInds      = mxGetIr(mat);
+  mwIndex* colPtrs      = mxGetJc(mat);
   mm_LocalOrd* dataCopy = new mm_LocalOrd[numEntries];
-  mm_LocalOrd* iter = dataCopy;
-  int* entriesPerRow = new int[numRows];
-  int* entriesPerCol = new int[numRows];
-  for(int i = 0; i < numRows; i++)
-  {
+  mm_LocalOrd* iter     = dataCopy;
+  int* entriesPerRow    = new int[numRows];
+  int* entriesPerCol    = new int[numRows];
+  for (int i = 0; i < numRows; i++) {
     entriesPerRow[i] = 0;
     entriesPerCol[i] = 0;
   }
-  for(int i = 0; i < numRows; i++)
-  {
-    ArrayView<const mm_LocalOrd> neighbors = data->getNeighborVertices(i); //neighbors has the column indices for row i
+  for (int i = 0; i < numRows; i++) {
+    ArrayView<const mm_LocalOrd> neighbors = data->getNeighborVertices(i);  // neighbors has the column indices for row i
     memcpy(iter, neighbors.getRawPtr(), sizeof(mm_LocalOrd) * neighbors.size());
     entriesPerRow[i] = neighbors.size();
-    for(int j = 0; j < neighbors.size(); j++)
-    {
+    for (int j = 0; j < neighbors.size(); j++) {
       entriesPerCol[neighbors[j]]++;
     }
     iter += neighbors.size();
   }
-  mwIndex** rowIndsByColumn = new mwIndex*[numRows]; //rowIndsByColumn[0] points to array of row indices in column 1
+  mwIndex** rowIndsByColumn  = new mwIndex*[numRows];  // rowIndsByColumn[0] points to array of row indices in column 1
   mxLogical** valuesByColumn = new mxLogical*[numRows];
-  int* numEnteredPerCol = new int[numRows];
-  int accum = 0;
-  for(int i = 0; i < numRows; i++)
-  {
+  int* numEnteredPerCol      = new int[numRows];
+  int accum                  = 0;
+  for (int i = 0; i < numRows; i++) {
     rowIndsByColumn[i] = &rowInds[accum];
-    //cout << "Entries in column " << i << " start at offset " << accum << endl;
+    // cout << "Entries in column " << i << " start at offset " << accum << endl;
     valuesByColumn[i] = &outData[accum];
     accum += entriesPerCol[i];
-    if(accum > numEntries)
+    if (accum > numEntries)
       throw runtime_error("potato");
   }
-  for(int i = 0; i < numRows; i++)
-  {
-    numEnteredPerCol[i] = 0; //rowIndsByColumn[n][numEnteredPerCol[n]] gives the next place to put a row index
+  for (int i = 0; i < numRows; i++) {
+    numEnteredPerCol[i] = 0;  // rowIndsByColumn[n][numEnteredPerCol[n]] gives the next place to put a row index
   }
-  //entriesPerCol now has Jc information (col offsets)
-  accum = 0; //keep track of cumulative index in dataCopy
-  for(int row = 0; row < numRows; row++)
-  {
-    for(int entryInRow = 0; entryInRow < entriesPerRow[row]; entryInRow++)
-    {
+  // entriesPerCol now has Jc information (col offsets)
+  accum = 0;  // keep track of cumulative index in dataCopy
+  for (int row = 0; row < numRows; row++) {
+    for (int entryInRow = 0; entryInRow < entriesPerRow[row]; entryInRow++) {
       int col = dataCopy[accum];
       accum++;
       rowIndsByColumn[col][numEnteredPerCol[col]] = row;
-      valuesByColumn[col][numEnteredPerCol[col]] = (mxLogical) 1;
+      valuesByColumn[col][numEnteredPerCol[col]]  = (mxLogical)1;
       numEnteredPerCol[col]++;
     }
   }
-  accum = 0; //keep track of total entries over all columns
-  for(int col = 0; col < numRows; col++)
-  {
+  accum = 0;  // keep track of total entries over all columns
+  for (int col = 0; col < numRows; col++) {
     colPtrs[col] = accum;
     accum += entriesPerCol[col];
   }
-  colPtrs[numRows] = accum; //the last entry in jc, which is equivalent to numEntries
+  colPtrs[numRows] = accum;  // the last entry in jc, which is equivalent to numEntries
   delete[] numEnteredPerCol;
   delete[] rowIndsByColumn;
   delete[] valuesByColumn;
   delete[] dataCopy;
   delete[] entriesPerRow;
   delete[] entriesPerCol;
-  //Construct list of boundary nodes
+  // Construct list of boundary nodes
   const ArrayRCP<const bool> boundaryFlags = data->GetBoundaryNodeMap();
-  int numBoundaryNodes = 0;
-  for(int i = 0; i < boundaryFlags.size(); i++)
-  {
-    if(boundaryFlags[i])
+  int numBoundaryNodes                     = 0;
+  for (int i = 0; i < boundaryFlags.size(); i++) {
+    if (boundaryFlags[i])
       numBoundaryNodes++;
   }
   cout << "Graph has " << numBoundaryNodes << " Dirichlet boundary nodes." << endl;
-  mwSize dims[] = {(mwSize) numBoundaryNodes, 1};
+  mwSize dims[]         = {(mwSize)numBoundaryNodes, 1};
   mxArray* boundaryList = mxCreateNumericArray(2, dims, mxINT32_CLASS, mxREAL);
-  int* dest = (int*) mxGetData(boundaryList);
-  int* destIter = dest;
-  for(int i = 0; i < boundaryFlags.size(); i++)
-  {
-    if(boundaryFlags[i])
-    {
+  int* dest             = (int*)mxGetData(boundaryList);
+  int* destIter         = dest;
+  for (int i = 0; i < boundaryFlags.size(); i++) {
+    if (boundaryFlags[i]) {
       *destIter = i;
       destIter++;
     }
@@ -1318,63 +1227,57 @@ mxArray* saveDataToMatlab(RCP<MGraph>& data)
 }
 
 #ifdef HAVE_MUELU_INTREPID2
-template<>
-mxArray* saveDataToMatlab(RCP<FieldContainer_ordinal>& data)
-{
+template <>
+mxArray* saveDataToMatlab(RCP<FieldContainer_ordinal>& data) {
   int rank = data->rank();
   // NOTE: Only supports rank 2 arrays
-  if(rank!=2)
+  if (rank != 2)
     throw std::runtime_error("Error: Only rank two FieldContainers are supported.");
 
   int nr = data->extent(0);
   int nc = data->extent(1);
 
-  mwSize dims[]={(mwSize)nr,(mwSize)nc};
-  mxArray* mxa = mxCreateNumericArray(2,dims, mxINT32_CLASS, mxREAL);
-  int *array = (int*) mxGetData(mxa);
+  mwSize dims[] = {(mwSize)nr, (mwSize)nc};
+  mxArray* mxa  = mxCreateNumericArray(2, dims, mxINT32_CLASS, mxREAL);
+  int* array    = (int*)mxGetData(mxa);
 
-  for(int col = 0; col < nc; col++)
-  {
-    for(int row = 0; row < nr; row++)
-    {
-      array[col * nr + row] = (*data)(row,col);
+  for (int col = 0; col < nc; col++) {
+    for (int row = 0; row < nr; row++) {
+      array[col * nr + row] = (*data)(row, col);
     }
   }
   return mxa;
 }
 #endif
 
-
-template<typename T>
-MuemexData<T>::MuemexData(const mxArray* mxa) : MuemexArg(getMuemexType<T>())
-{
+template <typename T>
+MuemexData<T>::MuemexData(const mxArray* mxa)
+  : MuemexArg(getMuemexType<T>()) {
   data = loadDataFromMatlab<T>(mxa);
 }
 
-template<typename T>
-mxArray* MuemexData<T>::convertToMatlab()
-{
+template <typename T>
+mxArray* MuemexData<T>::convertToMatlab() {
   return saveDataToMatlab<T>(data);
 }
 
-template<typename T>
-MuemexData<T>::MuemexData(T& dataToCopy, MuemexType dataType) : MuemexArg(dataType)
-{
+template <typename T>
+MuemexData<T>::MuemexData(T& dataToCopy, MuemexType dataType)
+  : MuemexArg(dataType) {
   data = dataToCopy;
 }
 
-template<typename T>
-MuemexData<T>::MuemexData(T& dataToCopy) : MuemexData(dataToCopy, getMuemexType(dataToCopy)) {}
+template <typename T>
+MuemexData<T>::MuemexData(T& dataToCopy)
+  : MuemexData(dataToCopy, getMuemexType(dataToCopy)) {}
 
-template<typename T>
-T& MuemexData<T>::getData()
-{
+template <typename T>
+T& MuemexData<T>::getData() {
   return data;
 }
 
-template<typename T>
-void MuemexData<T>::setData(T& newData)
-{
+template <typename T>
+void MuemexData<T>::setData(T& newData) {
   this->data = newData;
 }
 
@@ -1382,150 +1285,107 @@ void MuemexData<T>::setData(T& newData)
 /* More Template Functions       */
 /* ***************************** */
 
-template<typename T>
-void addLevelVariable(const T& data, std::string& name, Level& lvl, const Factory * fact)
-{
+template <typename T>
+void addLevelVariable(const T& data, std::string& name, Level& lvl, const Factory* fact) {
   lvl.AddKeepFlag(name, fact, MueLu::UserData);
   lvl.Set<T>(name, data, fact);
 }
 
-template<typename T>
-const T& getLevelVariable(std::string& name, Level& lvl)
-{
-  try
-  {
+template <typename T>
+const T& getLevelVariable(std::string& name, Level& lvl) {
+  try {
     return lvl.Get<T>(name);
-  }
-  catch(std::exception& e)
-  {
+  } catch (std::exception& e) {
     throw std::runtime_error("Requested custom variable " + name + " is not in the level.");
   }
 }
 
-//Functions used to put data through matlab factories - first arg is "this" pointer of matlab factory
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-std::vector<Teuchos::RCP<MuemexArg>> processNeeds(const Factory* factory, std::string& needsParam, Level& lvl)
-{
+// Functions used to put data through matlab factories - first arg is "this" pointer of matlab factory
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+std::vector<Teuchos::RCP<MuemexArg> > processNeeds(const Factory* factory, std::string& needsParam, Level& lvl) {
   using namespace std;
   using namespace Teuchos;
-  typedef RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Matrix_t;
-  typedef RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> MultiVector_t;
-  typedef RCP<Aggregates<LocalOrdinal, GlobalOrdinal, Node>> Aggregates_t;
-  typedef RCP<AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>> AmalgamationInfo_t;
+  typedef RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Matrix_t;
+  typedef RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MultiVector_t;
+  typedef RCP<Aggregates<LocalOrdinal, GlobalOrdinal, Node> > Aggregates_t;
+  typedef RCP<AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo_t;
   typedef RCP<MGraph> Graph_t;
   vector<string> needsList = tokenizeList(needsParam);
-  vector<RCP<MuemexArg>> args;
-  for(size_t i = 0; i < needsList.size(); i++)
-  {
-    if(needsList[i] == "A" || needsList[i] == "P" || needsList[i] == "R" || needsList[i]=="Ptent")
-    {
+  vector<RCP<MuemexArg> > args;
+  for (size_t i = 0; i < needsList.size(); i++) {
+    if (needsList[i] == "A" || needsList[i] == "P" || needsList[i] == "R" || needsList[i] == "Ptent") {
       Matrix_t mydata = lvl.Get<Matrix_t>(needsList[i], factory->GetFactory(needsList[i]).get());
       args.push_back(rcp(new MuemexData<Matrix_t>(mydata)));
-    }
-    else if(needsList[i] == "Nullspace" || needsList[i] == "Coordinates")
-    {
+    } else if (needsList[i] == "Nullspace" || needsList[i] == "Coordinates") {
       MultiVector_t mydata = lvl.Get<MultiVector_t>(needsList[i], factory->GetFactory(needsList[i]).get());
       args.push_back(rcp(new MuemexData<MultiVector_t>(mydata)));
-    }
-    else if(needsList[i] == "Aggregates")
-    {
+    } else if (needsList[i] == "Aggregates") {
       Aggregates_t mydata = lvl.Get<Aggregates_t>(needsList[i], factory->GetFactory(needsList[i]).get());
       args.push_back(rcp(new MuemexData<Aggregates_t>(mydata)));
-    }
-    else if(needsList[i] == "UnAmalgamationInfo")
-    {
+    } else if (needsList[i] == "UnAmalgamationInfo") {
       AmalgamationInfo_t mydata = lvl.Get<AmalgamationInfo_t>(needsList[i], factory->GetFactory(needsList[i]).get());
       args.push_back(rcp(new MuemexData<AmalgamationInfo_t>(mydata)));
-    }
-    else if(needsList[i] == "Level")
-    {
+    } else if (needsList[i] == "Level") {
       int levelNum = lvl.GetLevelID();
       args.push_back(rcp(new MuemexData<int>(levelNum)));
-    }
-    else if(needsList[i] == "Graph")
-    {
+    } else if (needsList[i] == "Graph") {
       Graph_t mydata = lvl.Get<Graph_t>(needsList[i], factory->GetFactory(needsList[i]).get());
       args.push_back(rcp(new MuemexData<Graph_t>(mydata)));
-    }
-    else
-    {
+    } else {
       vector<string> words;
       string badNameMsg = "Custom Muemex variables in \"Needs\" list require a type and a name, e.g. \"double myVal\". \n Leading and trailing spaces are OK. \n Don't know how to handle \"" + needsList[i] + "\".\n";
-      //compare type without case sensitivity
-      char* buf = (char*) malloc(needsList[i].size() + 1);
+      // compare type without case sensitivity
+      char* buf = (char*)malloc(needsList[i].size() + 1);
       strcpy(buf, needsList[i].c_str());
-      for(char* iter = buf; *iter != ' '; iter++)
-      {
-        if(*iter == 0)
-        {
+      for (char* iter = buf; *iter != ' '; iter++) {
+        if (*iter == 0) {
           free(buf);
           throw runtime_error(badNameMsg);
         }
-        *iter = (char) tolower(*iter);
+        *iter = (char)tolower(*iter);
       }
       const char* wordDelim = " ";
-      char* mark = strtok(buf, wordDelim);
-      while(mark != NULL)
-      {
+      char* mark            = strtok(buf, wordDelim);
+      while (mark != NULL) {
         string wordStr(mark);
         words.push_back(wordStr);
         mark = strtok(NULL, wordDelim);
       }
-      if(words.size() != 2)
-      {
+      if (words.size() != 2) {
         free(buf);
         throw runtime_error(badNameMsg);
       }
-      char* typeStr = (char*) words[0].c_str();
-      if(strstr(typeStr, "ordinalvector"))
-      {
-        typedef RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> LOVector_t;
+      char* typeStr = (char*)words[0].c_str();
+      if (strstr(typeStr, "ordinalvector")) {
+        typedef RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > LOVector_t;
         LOVector_t mydata = getLevelVariable<LOVector_t>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<LOVector_t>(mydata)));
-      }
-      else if(strstr(typeStr, "map"))
-      {
-        typedef RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t>> Map_t;
+      } else if (strstr(typeStr, "map")) {
+        typedef RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t> > Map_t;
         Map_t mydata = getLevelVariable<Map_t>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<Map_t>(mydata)));
-      }
-      else if(strstr(typeStr, "scalar"))
-      {
+      } else if (strstr(typeStr, "scalar")) {
         Scalar mydata = getLevelVariable<Scalar>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<Scalar>(mydata)));
-      }
-      else if(strstr(typeStr, "double"))
-      {
+      } else if (strstr(typeStr, "double")) {
         double mydata = getLevelVariable<double>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<double>(mydata)));
-      }
-      else if(strstr(typeStr, "complex"))
-      {
+      } else if (strstr(typeStr, "complex")) {
         complex_t mydata = getLevelVariable<complex_t>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<complex_t>(mydata)));
-      }
-      else if(strstr(typeStr, "matrix"))
-      {
+      } else if (strstr(typeStr, "matrix")) {
         Matrix_t mydata = getLevelVariable<Matrix_t>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<Matrix_t>(mydata)));
-      }
-      else if(strstr(typeStr, "multivector"))
-      {
+      } else if (strstr(typeStr, "multivector")) {
         MultiVector_t mydata = getLevelVariable<MultiVector_t>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<MultiVector_t>(mydata)));
-      }
-      else if(strstr(typeStr, "int"))
-      {
+      } else if (strstr(typeStr, "int")) {
         int mydata = getLevelVariable<int>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<int>(mydata)));
-      }
-      else if(strstr(typeStr, "string"))
-      {
+      } else if (strstr(typeStr, "string")) {
         string mydata = getLevelVariable<string>(needsList[i], lvl);
         args.push_back(rcp(new MuemexData<string>(mydata)));
-      }
-      else
-      {
+      } else {
         free(buf);
         throw std::runtime_error(words[0] + " is not a known variable type.");
       }
@@ -1535,128 +1395,90 @@ std::vector<Teuchos::RCP<MuemexArg>> processNeeds(const Factory* factory, std::s
   return args;
 }
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
-void processProvides(std::vector<Teuchos::RCP<MuemexArg>>& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl)
-{
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+void processProvides(std::vector<Teuchos::RCP<MuemexArg> >& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl) {
   using namespace std;
   using namespace Teuchos;
-  typedef RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Matrix_t;
-  typedef RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> MultiVector_t;
-  typedef RCP<Aggregates<LocalOrdinal, GlobalOrdinal, Node>> Aggregates_t;
-  typedef RCP<AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>> AmalgamationInfo_t;
+  typedef RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Matrix_t;
+  typedef RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MultiVector_t;
+  typedef RCP<Aggregates<LocalOrdinal, GlobalOrdinal, Node> > Aggregates_t;
+  typedef RCP<AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo_t;
   typedef RCP<MGraph> Graph_t;
   vector<string> provides = tokenizeList(providesParam);
-  for(size_t i = 0; i < size_t(provides.size()); i++)
-  {
-    if(provides[i] == "A" || provides[i] == "P" || provides[i] == "R" || provides[i]=="Ptent")
-    {
-      RCP<MuemexData<Matrix_t>> mydata = Teuchos::rcp_static_cast<MuemexData<Matrix_t>>(mexOutput[i]);
+  for (size_t i = 0; i < size_t(provides.size()); i++) {
+    if (provides[i] == "A" || provides[i] == "P" || provides[i] == "R" || provides[i] == "Ptent") {
+      RCP<MuemexData<Matrix_t> > mydata = Teuchos::rcp_static_cast<MuemexData<Matrix_t> >(mexOutput[i]);
       lvl.Set(provides[i], mydata->getData(), factory);
-    }
-    else if(provides[i] == "Nullspace" || provides[i] == "Coordinates")
-    {
-      RCP<MuemexData<MultiVector_t>> mydata = Teuchos::rcp_static_cast<MuemexData<MultiVector_t>>(mexOutput[i]);
+    } else if (provides[i] == "Nullspace" || provides[i] == "Coordinates") {
+      RCP<MuemexData<MultiVector_t> > mydata = Teuchos::rcp_static_cast<MuemexData<MultiVector_t> >(mexOutput[i]);
       lvl.Set(provides[i], mydata->getData(), factory);
-    }
-    else if(provides[i] == "Aggregates")
-    {
-      RCP<MuemexData<Aggregates_t>> mydata = Teuchos::rcp_static_cast<MuemexData<Aggregates_t>>(mexOutput[i]);
+    } else if (provides[i] == "Aggregates") {
+      RCP<MuemexData<Aggregates_t> > mydata = Teuchos::rcp_static_cast<MuemexData<Aggregates_t> >(mexOutput[i]);
       lvl.Set(provides[i], mydata->getData(), factory);
-    }
-    else if(provides[i] == "UnAmalgamationInfo")
-    {
-      RCP<MuemexData<AmalgamationInfo_t>> mydata = Teuchos::rcp_static_cast<MuemexData<AmalgamationInfo_t>>(mexOutput[i]);
+    } else if (provides[i] == "UnAmalgamationInfo") {
+      RCP<MuemexData<AmalgamationInfo_t> > mydata = Teuchos::rcp_static_cast<MuemexData<AmalgamationInfo_t> >(mexOutput[i]);
       lvl.Set(provides[i], mydata->getData(), factory);
-    }
-    else if(provides[i] == "Graph")
-    {
-      RCP<MuemexData<Graph_t>> mydata = Teuchos::rcp_static_cast<MuemexData<Graph_t>>(mexOutput[i]);
+    } else if (provides[i] == "Graph") {
+      RCP<MuemexData<Graph_t> > mydata = Teuchos::rcp_static_cast<MuemexData<Graph_t> >(mexOutput[i]);
       lvl.Set(provides[i], mydata->getData(), factory);
-    }
-    else
-    {
+    } else {
       vector<string> words;
       string badNameMsg = "Custom Muemex variables in \"Provides\" list require a type and a name, e.g. \"double myVal\". \n Leading and trailing spaces are OK. \n Don't know how to handle \"" + provides[i] + "\".\n";
-      //compare type without case sensitivity
-      char* buf = (char*) malloc(provides[i].size() + 1);
+      // compare type without case sensitivity
+      char* buf = (char*)malloc(provides[i].size() + 1);
       strcpy(buf, provides[i].c_str());
-      for(char* iter = buf; *iter != ' '; iter++)
-      {
-        if(*iter == 0)
-        {
+      for (char* iter = buf; *iter != ' '; iter++) {
+        if (*iter == 0) {
           free(buf);
           throw runtime_error(badNameMsg);
         }
-        *iter = (char) tolower(*iter);
+        *iter = (char)tolower(*iter);
       }
       const char* wordDelim = " ";
-      char* mark = strtok(buf, wordDelim);
-      while(mark != NULL)
-      {
+      char* mark            = strtok(buf, wordDelim);
+      while (mark != NULL) {
         string wordStr(mark);
         words.push_back(wordStr);
         mark = strtok(NULL, wordDelim);
       }
-      if(words.size() != 2)
-      {
+      if (words.size() != 2) {
         free(buf);
         throw runtime_error(badNameMsg);
       }
       const char* typeStr = words[0].c_str();
-      if(strstr(typeStr, "ordinalvector"))
-      {
-        typedef RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t>> LOVector_t;
-        RCP<MuemexData<LOVector_t>> mydata = Teuchos::rcp_static_cast<MuemexData<LOVector_t>>(mexOutput[i]);
+      if (strstr(typeStr, "ordinalvector")) {
+        typedef RCP<Xpetra::Vector<mm_LocalOrd, mm_LocalOrd, mm_GlobalOrd, mm_node_t> > LOVector_t;
+        RCP<MuemexData<LOVector_t> > mydata = Teuchos::rcp_static_cast<MuemexData<LOVector_t> >(mexOutput[i]);
         addLevelVariable<LOVector_t>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "map"))
-      {
-        typedef RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t>> Map_t;
-        RCP<MuemexData<Map_t>> mydata = Teuchos::rcp_static_cast<MuemexData<Map_t>>(mexOutput[i]);
+      } else if (strstr(typeStr, "map")) {
+        typedef RCP<Xpetra::Map<mm_LocalOrd, mm_GlobalOrd, mm_node_t> > Map_t;
+        RCP<MuemexData<Map_t> > mydata = Teuchos::rcp_static_cast<MuemexData<Map_t> >(mexOutput[i]);
         addLevelVariable<Map_t>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "scalar"))
-      {
-        RCP<MuemexData<Scalar>> mydata = Teuchos::rcp_static_cast<MuemexData<Scalar>>(mexOutput[i]);
+      } else if (strstr(typeStr, "scalar")) {
+        RCP<MuemexData<Scalar> > mydata = Teuchos::rcp_static_cast<MuemexData<Scalar> >(mexOutput[i]);
         addLevelVariable<Scalar>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "double"))
-      {
-        RCP<MuemexData<double>> mydata = Teuchos::rcp_static_cast<MuemexData<double>>(mexOutput[i]);
+      } else if (strstr(typeStr, "double")) {
+        RCP<MuemexData<double> > mydata = Teuchos::rcp_static_cast<MuemexData<double> >(mexOutput[i]);
         addLevelVariable<double>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "complex"))
-      {
-        RCP<MuemexData<complex_t>> mydata = Teuchos::rcp_static_cast<MuemexData<complex_t>>(mexOutput[i]);
+      } else if (strstr(typeStr, "complex")) {
+        RCP<MuemexData<complex_t> > mydata = Teuchos::rcp_static_cast<MuemexData<complex_t> >(mexOutput[i]);
         addLevelVariable<complex_t>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "matrix"))
-      {
-        RCP<MuemexData<Matrix_t>> mydata = Teuchos::rcp_static_cast<MuemexData<Matrix_t>>(mexOutput[i]);
+      } else if (strstr(typeStr, "matrix")) {
+        RCP<MuemexData<Matrix_t> > mydata = Teuchos::rcp_static_cast<MuemexData<Matrix_t> >(mexOutput[i]);
         addLevelVariable<Matrix_t>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "multivector"))
-      {
-        RCP<MuemexData<MultiVector_t>> mydata = Teuchos::rcp_static_cast<MuemexData<MultiVector_t>>(mexOutput[i]);
+      } else if (strstr(typeStr, "multivector")) {
+        RCP<MuemexData<MultiVector_t> > mydata = Teuchos::rcp_static_cast<MuemexData<MultiVector_t> >(mexOutput[i]);
         addLevelVariable<MultiVector_t>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "int"))
-      {
-        RCP<MuemexData<int>> mydata = Teuchos::rcp_static_cast<MuemexData<int>>(mexOutput[i]);
+      } else if (strstr(typeStr, "int")) {
+        RCP<MuemexData<int> > mydata = Teuchos::rcp_static_cast<MuemexData<int> >(mexOutput[i]);
         addLevelVariable<int>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "bool"))
-      {
+      } else if (strstr(typeStr, "bool")) {
         RCP<MuemexData<bool> > mydata = Teuchos::rcp_static_cast<MuemexData<bool> >(mexOutput[i]);
         addLevelVariable<bool>(mydata->getData(), words[1], lvl, factory);
-      }
-      else if(strstr(typeStr, "string"))
-      {
-        RCP<MuemexData<string>> mydata = Teuchos::rcp_static_cast<MuemexData<string>>(mexOutput[i]);
+      } else if (strstr(typeStr, "string")) {
+        RCP<MuemexData<string> > mydata = Teuchos::rcp_static_cast<MuemexData<string> >(mexOutput[i]);
         addLevelVariable<string>(mydata->getData(), words[1], lvl, factory);
-      }
-      else
-      {
+      } else {
         free(buf);
         throw std::runtime_error(words[0] + " is not a known variable type.");
       }
@@ -1667,27 +1489,26 @@ void processProvides(std::vector<Teuchos::RCP<MuemexArg>>& mexOutput, const Fact
 
 // Throwable Stubs for long long
 
-template<>
-std::vector<Teuchos::RCP<MuemexArg>> processNeeds<double,mm_LocalOrd,long long,mm_node_t>(const Factory* factory, std::string& needsParam, Level& lvl) {
+template <>
+std::vector<Teuchos::RCP<MuemexArg> > processNeeds<double, mm_LocalOrd, long long, mm_node_t>(const Factory* factory, std::string& needsParam, Level& lvl) {
   throw std::runtime_error("Muemex does not support long long for global indices");
 }
 
-template<>
-std::vector<Teuchos::RCP<MuemexArg>> processNeeds<complex_t,mm_LocalOrd,long long,mm_node_t>(const Factory* factory, std::string& needsParam, Level& lvl) {
+template <>
+std::vector<Teuchos::RCP<MuemexArg> > processNeeds<complex_t, mm_LocalOrd, long long, mm_node_t>(const Factory* factory, std::string& needsParam, Level& lvl) {
   throw std::runtime_error("Muemex does not support long long for global indices");
 }
 
-template<>
-void processProvides<double,mm_LocalOrd,long long,mm_node_t>(std::vector<Teuchos::RCP<MuemexArg>>& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl) {
+template <>
+void processProvides<double, mm_LocalOrd, long long, mm_node_t>(std::vector<Teuchos::RCP<MuemexArg> >& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl) {
   throw std::runtime_error("Muemex does not support long long for global indices");
 }
 
-template<>
-void processProvides<complex_t,mm_LocalOrd,long long,mm_node_t>(std::vector<Teuchos::RCP<MuemexArg>>& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl) {
+template <>
+void processProvides<complex_t, mm_LocalOrd, long long, mm_node_t>(std::vector<Teuchos::RCP<MuemexArg> >& mexOutput, const Factory* factory, std::string& providesParam, Level& lvl) {
   throw std::runtime_error("Muemex does not support long long for global indices");
 }
 
-
-}// end namespace
-#endif //HAVE_MUELU_MATLAB error handler
-#endif //MUELU_MATLABUTILS_DEF_HPP guard
+}  // namespace MueLu
+#endif  // HAVE_MUELU_MATLAB error handler
+#endif  // MUELU_MATLABUTILS_DEF_HPP guard
diff --git a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory.cpp b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory.cpp
index 338992b88dec..36fbf7a0e607 100644
--- a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory.cpp
+++ b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory.cpp
@@ -44,7 +44,6 @@
 //
 // @HEADER
 
-
 #include "MueLu_ExplicitInstantiation.hpp"
 
 #include "MueLu_SingleLevelMatlabFactory_def.hpp"
@@ -52,8 +51,8 @@
 #include "TpetraCore_ETIHelperMacros.h"
 
 #ifdef HAVE_MUELU_MATLAB
-#define MUELU_LOCAL_INSTANT(S,LO,GO,N) \
-        template class MueLu::SingleLevelMatlabFactory<S,LO,GO,N>;
+#define MUELU_LOCAL_INSTANT(S, LO, GO, N) \
+  template class MueLu::SingleLevelMatlabFactory<S, LO, GO, N>;
 
 TPETRA_ETI_MANGLING_TYPEDEFS()
 
diff --git a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_decl.hpp b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_decl.hpp
index 6e6298bf8284..fb0be250c6e5 100644
--- a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_decl.hpp
+++ b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_decl.hpp
@@ -65,65 +65,64 @@
 #ifdef HAVE_MUELU_MATLAB
 #include "mex.h"
 
-
 namespace MueLu {
-  /*!
-    @class SingleLevelMatlabFactory
-    @ingroup MueMexClasses
-    @brief Factory for interacting with Matlab
-  */
-  template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class SingleLevelMatlabFactory : public SingleLevelFactoryBase {
+/*!
+  @class SingleLevelMatlabFactory
+  @ingroup MueMexClasses
+  @brief Factory for interacting with Matlab
+*/
+template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class SingleLevelMatlabFactory : public SingleLevelFactoryBase {
 #undef MUELU_SINGLELEVELMATLABFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    SingleLevelMatlabFactory();
+  SingleLevelMatlabFactory();
 
-    virtual ~SingleLevelMatlabFactory() { }
+  virtual ~SingleLevelMatlabFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& currentLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level& currentLevel) const;
+  //@}
 
-    //! @ name Description
-    //@{
-    std::string description() const;
-    //@}
+  //! @ name Description
+  //@{
+  std::string description() const;
+  //@}
 
-  private:
-    //@{
+ private:
+  //@{
 
-    mutable bool hasDeclaredInput_;
+  mutable bool hasDeclaredInput_;
 
-    //@}
+  //@}
 
-    //@{
+  //@{
 
-    //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects.
-    mutable std::vector<std::string> needs_;
+  //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects.
+  mutable std::vector<std::string> needs_;
 
-    //@}
+  //@}
 
-  }; //class SingleLevelMatlabFactory
+};  // class SingleLevelMatlabFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SINGLELEVELMATLABFACTORY_SHORT
 
-#endif // HAVE_MUELU_MATLAB
-#endif // MUELU SINGLELEVELMATLABFACTORY_DECL_HPP
+#endif  // HAVE_MUELU_MATLAB
+#endif  // MUELU SINGLELEVELMATLABFACTORY_DECL_HPP
diff --git a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_def.hpp b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_def.hpp
index fe1a943c8688..849e059ed673 100644
--- a/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_def.hpp
+++ b/packages/muelu/matlab/src/MueLu_SingleLevelMatlabFactory_def.hpp
@@ -54,76 +54,68 @@
 #include "MueLu_SingleLevelMatlabFactory_decl.hpp"
 #include "MueLu_MatlabUtils_decl.hpp"
 
-
 #ifdef HAVE_MUELU_MATLAB
 #include "mex.h"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SingleLevelMatlabFactory()
-    : hasDeclaredInput_(false) { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SingleLevelMatlabFactory()
+  : hasDeclaredInput_(false) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-  {
-    RCP<ParameterList> validParamList = getInputParamList();
-    validParamList->set<std::string>("Provides"     , "" ,"A comma-separated list of objects provided by the SingleLevelMatlabFactory");
-    validParamList->set<std::string>("Needs"        , "", "A comma-separated list of objects needed by the SingleLevelMatlabFactory");
-    validParamList->set<std::string>("Function"     , "" , "The name of the Matlab MEX function to call for Build()");
-    return validParamList;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = getInputParamList();
+  validParamList->set<std::string>("Provides", "", "A comma-separated list of objects provided by the SingleLevelMatlabFactory");
+  validParamList->set<std::string>("Needs", "", "A comma-separated list of objects needed by the SingleLevelMatlabFactory");
+  validParamList->set<std::string>("Function", "", "The name of the Matlab MEX function to call for Build()");
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-  {
-    const Teuchos::ParameterList& pL = GetParameterList();
-    needs_ = tokenizeList(pL.get<std::string>("Needs"));
-    // Declare inputs
-    for(size_t i = 0; i < needs_.size(); i++)
-    {
-      if(!IsParamMuemexVariable(needs_[i]) && needs_[i] != "Level")
-        this->Input(currentLevel, needs_[i]);
-    }
-    hasDeclaredInput_ = true;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const Teuchos::ParameterList& pL = GetParameterList();
+  needs_                           = tokenizeList(pL.get<std::string>("Needs"));
+  // Declare inputs
+  for (size_t i = 0; i < needs_.size(); i++) {
+    if (!IsParamMuemexVariable(needs_[i]) && needs_[i] != "Level")
+      this->Input(currentLevel, needs_[i]);
   }
+  hasDeclaredInput_ = true;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const
-  {
-    FactoryMonitor m(*this, "Build", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
 
-    const Teuchos::ParameterList& pL = GetParameterList();
-    using Teuchos::rcp;
-    using Teuchos::rcp;
-    using namespace std;
-    // NOTE: mexOutput[0] is the "Provides."  Might want to modify to allow for additional outputs
-    string needsList = pL.get<string>("Needs");
-    vector<RCP<MuemexArg>> InputArgs = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsList, currentLevel);
-    string providesList = pL.get<std::string>("Provides");
-    size_t numProvides = tokenizeList(providesList).size();
-    // Call mex function
-    string matlabFunction = pL.get<std::string>("Function");
-    if(!matlabFunction.length())
-      throw std::runtime_error("Invalid matlab function name");
-    vector<Teuchos::RCP<MuemexArg> > mexOutput = callMatlab(matlabFunction, numProvides, InputArgs);
-    // Set output in level
-    processProvides<Scalar, LocalOrdinal, GlobalOrdinal, Node>(mexOutput, this, providesList, currentLevel);
-  }
-  
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    const Teuchos::ParameterList& pL = GetParameterList();
-    out << "SingleLevelMatlabFactory["<<pL.get<std::string>("Function")<<"]";
-    return out.str();
-  }
+  const Teuchos::ParameterList& pL = GetParameterList();
+  using Teuchos::rcp;
+  using namespace std;
+  // NOTE: mexOutput[0] is the "Provides."  Might want to modify to allow for additional outputs
+  string needsList                 = pL.get<string>("Needs");
+  vector<RCP<MuemexArg>> InputArgs = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsList, currentLevel);
+  string providesList              = pL.get<std::string>("Provides");
+  size_t numProvides               = tokenizeList(providesList).size();
+  // Call mex function
+  string matlabFunction = pL.get<std::string>("Function");
+  if (!matlabFunction.length())
+    throw std::runtime_error("Invalid matlab function name");
+  vector<Teuchos::RCP<MuemexArg>> mexOutput = callMatlab(matlabFunction, numProvides, InputArgs);
+  // Set output in level
+  processProvides<Scalar, LocalOrdinal, GlobalOrdinal, Node>(mexOutput, this, providesList, currentLevel);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  const Teuchos::ParameterList& pL = GetParameterList();
+  out << "SingleLevelMatlabFactory[" << pL.get<std::string>("Function") << "]";
+  return out.str();
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SINGLELEVELMATLABFACTORY_SHORT
-#endif // HAVE_MUELU_MATLAB
-
-#endif // MUELU_SINGLELEVELMATLABFACTORY_DEF_HPP
+#endif  // HAVE_MUELU_MATLAB
 
+#endif  // MUELU_SINGLELEVELMATLABFACTORY_DEF_HPP
diff --git a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory.cpp b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory.cpp
index 6e3e836d0203..38249d1da71e 100644
--- a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory.cpp
+++ b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory.cpp
@@ -44,7 +44,6 @@
 //
 // @HEADER
 
-
 #include "MueLu_ExplicitInstantiation.hpp"
 
 #include "MueLu_TwoLevelMatlabFactory_def.hpp"
@@ -52,8 +51,8 @@
 #include "TpetraCore_ETIHelperMacros.h"
 
 #ifdef HAVE_MUELU_MATLAB
-#define MUELU_LOCAL_INSTANT(S,LO,GO,N) \
-        template class MueLu::TwoLevelMatlabFactory<S,LO,GO,N>;
+#define MUELU_LOCAL_INSTANT(S, LO, GO, N) \
+  template class MueLu::TwoLevelMatlabFactory<S, LO, GO, N>;
 
 TPETRA_ETI_MANGLING_TYPEDEFS()
 
diff --git a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_decl.hpp b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_decl.hpp
index dde2b72a60a8..2bf2e3a9df98 100644
--- a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_decl.hpp
+++ b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_decl.hpp
@@ -66,67 +66,66 @@
 #ifdef HAVE_MUELU_MATLAB
 #include "mex.h"
 
-
 namespace MueLu {
-  /*!
-    @class TwoLevelMatlabFactory
-    @ingroup MueMexClasses
-    @brief Factory for interacting with Matlab
-  */
-  template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class TwoLevelMatlabFactory : public TwoLevelFactoryBase {
+/*!
+  @class TwoLevelMatlabFactory
+  @ingroup MueMexClasses
+  @brief Factory for interacting with Matlab
+*/
+template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class TwoLevelMatlabFactory : public TwoLevelFactoryBase {
 #undef MUELU_TWOLEVELMATLABFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    TwoLevelMatlabFactory();
+  TwoLevelMatlabFactory();
 
-    virtual ~TwoLevelMatlabFactory() { }
+  virtual ~TwoLevelMatlabFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& fineLevel, Level& coarseLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //! @ name Description
-    //@{
-    std::string description() const;
-    //@}
-  private:
-    //@{
+  //! @ name Description
+  //@{
+  std::string description() const;
+  //@}
+ private:
+  //@{
 
-    mutable bool hasDeclaredInput_;
+  mutable bool hasDeclaredInput_;
 
-    //@}
+  //@}
 
-    //@{
+  //@{
 
-    //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects for the fine level.  These must be listed before coarse needs.
-    mutable std::vector<std::string> needsFine_;
+  //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects for the fine level.  These must be listed before coarse needs.
+  mutable std::vector<std::string> needsFine_;
 
-    //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects for the coarse level.  These must be listed after fine needs.
-    mutable std::vector<std::string> needsCoarse_;
+  //! List of arguments to the MATLAB function, in order.  These args must correspond to MueLu "Needs" objects for the coarse level.  These must be listed after fine needs.
+  mutable std::vector<std::string> needsCoarse_;
 
-    //@}
+  //@}
 
-  }; //class TwoLevelMatlabFactory
+};  // class TwoLevelMatlabFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TWOLEVELMATLABFACTORY_SHORT
 
-#endif // HAVE_MUELU_MATLAB
-#endif // MUELU TWOLEVELMATLABFACTORY_DECL_HPP
+#endif  // HAVE_MUELU_MATLAB
+#endif  // MUELU TWOLEVELMATLABFACTORY_DECL_HPP
diff --git a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_def.hpp b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_def.hpp
index ce98877e7bc8..d221e0cd6dcf 100644
--- a/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_def.hpp
+++ b/packages/muelu/matlab/src/MueLu_TwoLevelMatlabFactory_def.hpp
@@ -56,88 +56,83 @@
 
 #include <iostream>
 
-
 #ifdef HAVE_MUELU_MATLAB
 #include "mex.h"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoLevelMatlabFactory()
-    : hasDeclaredInput_(false) { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = getInputParamList();
-    validParamList->set<std::string>("Provides"     , "" ,"A comma-separated list of objects provided on the coarse level by the TwoLevelMatlabFactory");
-    validParamList->set<std::string>("Needs Fine"   , "", "A comma-separated list of objects needed on the fine level by the TwoLevelMatlabFactory");
-    validParamList->set<std::string>("Needs Coarse" , "", "A comma-separated list of objects needed on the coarse level by the TwoLevelMatlabFactory");
-    validParamList->set<std::string>("Function"     , "" , "The name of the Matlab MEX function to call for Build()");
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    using namespace std;
-    const Teuchos::ParameterList& pL = GetParameterList();
-    // Get needs strings
-    const std::string str_nf = pL.get<std::string>("Needs Fine");
-    const std::string str_nc = pL.get<std::string>("Needs Coarse");
-    needsFine_ = tokenizeList(str_nf);
-    needsCoarse_ = tokenizeList(str_nc);
-    for(auto fineNeed : needsFine_)
-    {
-      if(!IsParamMuemexVariable(fineNeed) && fineNeed != "Level")
-        this->Input(fineLevel, fineNeed);
-    }
-    for(auto coarseNeed : needsCoarse_)
-    {
-      if(!IsParamMuemexVariable(coarseNeed) && coarseNeed != "Level")
-        this->Input(coarseLevel, coarseNeed);
-    }
-    hasDeclaredInput_ = true;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-    using Teuchos::rcp;
-    using Teuchos::RCP;
-    using namespace std;
-    string needsFine = pL.get<string>("Needs Fine");
-    string needsCoarse = pL.get<string>("Needs Coarse");
-    vector<RCP<MuemexArg>> InputArgs = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsFine, fineLevel);
-    vector<RCP<MuemexArg>> InputArgsCoarse = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsCoarse, coarseLevel);
-    //Add coarse args to the end of InputArgs
-    InputArgs.reserve(InputArgs.size() + InputArgsCoarse.size());
-    InputArgs.insert(InputArgs.begin(), InputArgsCoarse.begin(), InputArgsCoarse.end());
-
-    // Determine output
-    string provides = pL.get<string>("Provides");
-    size_t numProvides = tokenizeList(provides).size();
-    // Call mex function
-    string matlabFunction = pL.get<string>("Function");
-    if(!matlabFunction.length())
-      throw runtime_error("Invalid matlab function name");
-    vector<RCP<MuemexArg>> mexOutput = callMatlab(matlabFunction, numProvides, InputArgs);
-    processProvides<Scalar, LocalOrdinal, GlobalOrdinal, Node>(mexOutput, this, provides, coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoLevelMatlabFactory()
+  : hasDeclaredInput_(false) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = getInputParamList();
+  validParamList->set<std::string>("Provides", "", "A comma-separated list of objects provided on the coarse level by the TwoLevelMatlabFactory");
+  validParamList->set<std::string>("Needs Fine", "", "A comma-separated list of objects needed on the fine level by the TwoLevelMatlabFactory");
+  validParamList->set<std::string>("Needs Coarse", "", "A comma-separated list of objects needed on the coarse level by the TwoLevelMatlabFactory");
+  validParamList->set<std::string>("Function", "", "The name of the Matlab MEX function to call for Build()");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  using namespace std;
+  const Teuchos::ParameterList& pL = GetParameterList();
+  // Get needs strings
+  const std::string str_nf = pL.get<std::string>("Needs Fine");
+  const std::string str_nc = pL.get<std::string>("Needs Coarse");
+  needsFine_               = tokenizeList(str_nf);
+  needsCoarse_             = tokenizeList(str_nc);
+  for (auto fineNeed : needsFine_) {
+    if (!IsParamMuemexVariable(fineNeed) && fineNeed != "Level")
+      this->Input(fineLevel, fineNeed);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    const Teuchos::ParameterList& pL = GetParameterList();
-    out << "TwoLevelMatlabFactory["<<pL.get<std::string>("Function")<<"]";
-    return out.str();
+  for (auto coarseNeed : needsCoarse_) {
+    if (!IsParamMuemexVariable(coarseNeed) && coarseNeed != "Level")
+      this->Input(coarseLevel, coarseNeed);
   }
-
-
-} //namespace MueLu
+  hasDeclaredInput_ = true;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+  using Teuchos::rcp;
+  using Teuchos::RCP;
+  using namespace std;
+  string needsFine                       = pL.get<string>("Needs Fine");
+  string needsCoarse                     = pL.get<string>("Needs Coarse");
+  vector<RCP<MuemexArg>> InputArgs       = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsFine, fineLevel);
+  vector<RCP<MuemexArg>> InputArgsCoarse = processNeeds<Scalar, LocalOrdinal, GlobalOrdinal, Node>(this, needsCoarse, coarseLevel);
+  // Add coarse args to the end of InputArgs
+  InputArgs.reserve(InputArgs.size() + InputArgsCoarse.size());
+  InputArgs.insert(InputArgs.begin(), InputArgsCoarse.begin(), InputArgsCoarse.end());
+
+  // Determine output
+  string provides    = pL.get<string>("Provides");
+  size_t numProvides = tokenizeList(provides).size();
+  // Call mex function
+  string matlabFunction = pL.get<string>("Function");
+  if (!matlabFunction.length())
+    throw runtime_error("Invalid matlab function name");
+  vector<RCP<MuemexArg>> mexOutput = callMatlab(matlabFunction, numProvides, InputArgs);
+  processProvides<Scalar, LocalOrdinal, GlobalOrdinal, Node>(mexOutput, this, provides, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  const Teuchos::ParameterList& pL = GetParameterList();
+  out << "TwoLevelMatlabFactory[" << pL.get<std::string>("Function") << "]";
+  return out.str();
+}
+
+}  // namespace MueLu
 
 #define MUELU_TWOLEVELMATLABFACTORY_SHORT
-#endif // HAVE_MUELU_MATLAB
-
-#endif // MUELU_TWOLEVELMATLABFACTORY_DEF_HPP
+#endif  // HAVE_MUELU_MATLAB
 
+#endif  // MUELU_TWOLEVELMATLABFACTORY_DEF_HPP
diff --git a/packages/muelu/research/caglusa/MueLu_IOhelpers.hpp b/packages/muelu/research/caglusa/MueLu_IOhelpers.hpp
index 4b67209a05a2..4bd682c4844a 100644
--- a/packages/muelu/research/caglusa/MueLu_IOhelpers.hpp
+++ b/packages/muelu/research/caglusa/MueLu_IOhelpers.hpp
@@ -4,137 +4,131 @@
 #include <Teuchos_ParameterList.hpp>
 #include <Xpetra_IO.hpp>
 
-
 namespace MueLu {
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::rcp_dynamic_cast;
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  struct IOhelpers {
-
-    static
-    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-    Read(const std::string&   filename,
-         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap,
-         RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > colMap,
-         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainMap        = Teuchos::null,
-         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangeMap         = Teuchos::null,
-         const bool           callFillComplete = true,
-         const bool           binary           = false,
-         const bool           readLocal        = false) {
-      using IO = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
-      if (readLocal)
-        A = IO::ReadLocal(filename, rowMap, colMap, domainMap, rangeMap, callFillComplete, binary);
-      else
-        A = IO::Read(filename, rowMap, colMap, domainMap, rangeMap, callFillComplete, binary);
-      return A;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_dynamic_cast;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct IOhelpers {
+  static Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+  Read(const std::string& filename,
+       const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap,
+       RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > colMap,
+       const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainMap = Teuchos::null,
+       const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangeMap  = Teuchos::null,
+       const bool callFillComplete                                                = true,
+       const bool binary                                                          = false,
+       const bool readLocal                                                       = false) {
+    using IO = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
+    if (readLocal)
+      A = IO::ReadLocal(filename, rowMap, colMap, domainMap, rangeMap, callFillComplete, binary);
+    else
+      A = IO::Read(filename, rowMap, colMap, domainMap, rangeMap, callFillComplete, binary);
+    return A;
+  }
+
+  static Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+  Read(std::string& filename,
+       RCP<const Teuchos::Comm<int> >& comm) {
+    Teuchos::ParameterList hierarchicalParams;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&hierarchicalParams), *comm);
+    auto op = Read(hierarchicalParams, comm);
+    return op;
+  }
+
+  static Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+  Read(Teuchos::ParameterList& hierarchicalParams,
+       RCP<const Teuchos::Comm<int> >& comm) {
+    using HOp                 = Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    using blocked_matrix_type = typename HOp::blocked_matrix_type;
+    using blocked_map_type    = typename blocked_matrix_type::blocked_map_type;
+    using matrix_type         = typename HOp::matrix_type;
+    using map_type            = typename HOp::map_type;
+    using lo_vec_type         = typename blocked_map_type::lo_vec_type;
+
+    auto lib = Xpetra::UseTpetra;
+    RCP<HOp> op;
+    RCP<const map_type> map, near_colmap, clusterCoeffMap, ghosted_clusterCoeffMap, clusterMap, ghosted_clusterMap;
+    RCP<matrix_type> nearField, basisMatrix, kernelApproximations, kernelBlockGraph;
+
+    std::vector<RCP<blocked_matrix_type> > transferMatrices;
+    RCP<lo_vec_type> clusterSizes;
+    RCP<blocked_map_type> blockedClusterMap, ghosted_blockedClusterMap;
+    RCP<blocked_matrix_type> blockKernelApproximations;
+
+    const bool readBinary = hierarchicalParams.get<bool>("read binary", false);
+    const bool readLocal  = hierarchicalParams.get<bool>("read local", false);
+
+    using IO = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+    // row, domain and range map of the operator
+    map = IO::ReadMap(hierarchicalParams.get<std::string>("map"), lib, comm, readBinary);
+    // colmap of near field
+    near_colmap = IO::ReadMap(hierarchicalParams.get<std::string>("near colmap"), lib, comm, readBinary);
+    if (hierarchicalParams.isType<std::string>("coefficient map")) {
+      // 1-to-1 map for the cluster coefficients
+      clusterCoeffMap = IO::ReadMap(hierarchicalParams.get<std::string>("coefficient map"), lib, comm, readBinary);
+      // overlapping map for the cluster coefficients
+      ghosted_clusterCoeffMap = IO::ReadMap(hierarchicalParams.get<std::string>("ghosted coefficient map"), lib, comm, readBinary);
+      // 1-to-1 map for the clusters
+      clusterMap = IO::ReadMap(hierarchicalParams.get<std::string>("cluster map"), lib, comm, readBinary);
+      // overlapping map for the clusters
+      ghosted_clusterMap = IO::ReadMap(hierarchicalParams.get<std::string>("ghosted cluster map"), lib, comm, readBinary);
+
+      // blocked cluster map
+      clusterSizes      = Xpetra::IO<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(hierarchicalParams.get<std::string>("gid_cluster_to_gid_coeff"), clusterMap)->getVectorNonConst(0);
+      blockedClusterMap = rcp(new blocked_map_type(clusterCoeffMap, clusterSizes));
     }
 
-    static
-    Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-    Read(std::string& filename,
-         RCP< const Teuchos::Comm<int> >& comm) {
-      Teuchos::ParameterList hierarchicalParams;
-      Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&hierarchicalParams), *comm);
-      auto op = Read(hierarchicalParams, comm);
-      return op;
-    }
+    // near field interactions
+    nearField = Read(hierarchicalParams.get<std::string>("near field matrix"), map, near_colmap, map, map, true, readBinary, readLocal);
 
-    static
-    Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-    Read(Teuchos::ParameterList& hierarchicalParams,
-         RCP< const Teuchos::Comm<int> >& comm) {
-      using HOp = Xpetra::HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      using blocked_matrix_type = typename HOp::blocked_matrix_type;
-      using blocked_map_type = typename blocked_matrix_type::blocked_map_type;
-      using matrix_type = typename HOp::matrix_type;
-      using map_type = typename HOp::map_type;
-      using lo_vec_type = typename blocked_map_type::lo_vec_type;
-
-      auto  lib = Xpetra::UseTpetra;
-      RCP<HOp>                       op;
-      RCP<const map_type>            map, near_colmap, clusterCoeffMap, ghosted_clusterCoeffMap, clusterMap, ghosted_clusterMap;
-      RCP<matrix_type>               nearField, basisMatrix, kernelApproximations, kernelBlockGraph;
-
-      std::vector<RCP<blocked_matrix_type> > transferMatrices;
-      RCP<lo_vec_type>               clusterSizes;
-      RCP<blocked_map_type>          blockedClusterMap, ghosted_blockedClusterMap;
-      RCP<blocked_matrix_type>       blockKernelApproximations;
-
-      const bool readBinary = hierarchicalParams.get<bool>("read binary", false);
-      const bool readLocal = hierarchicalParams.get<bool>("read local", false);
-
-      using IO = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-
-      // row, domain and range map of the operator
-      map = IO::ReadMap(hierarchicalParams.get<std::string>("map"), lib, comm, readBinary);
-      // colmap of near field
-      near_colmap = IO::ReadMap(hierarchicalParams.get<std::string>("near colmap"), lib, comm, readBinary);
-      if (hierarchicalParams.isType<std::string>("coefficient map")) {
-        // 1-to-1 map for the cluster coefficients
-        clusterCoeffMap = IO::ReadMap(hierarchicalParams.get<std::string>("coefficient map"), lib, comm, readBinary);
-        // overlapping map for the cluster coefficients
-        ghosted_clusterCoeffMap = IO::ReadMap(hierarchicalParams.get<std::string>("ghosted coefficient map"), lib, comm, readBinary);
-        // 1-to-1 map for the clusters
-        clusterMap = IO::ReadMap(hierarchicalParams.get<std::string>("cluster map"), lib, comm, readBinary);
-        // overlapping map for the clusters
-        ghosted_clusterMap = IO::ReadMap(hierarchicalParams.get<std::string>("ghosted cluster map"), lib, comm, readBinary);
-
-        // blocked cluster map
-        clusterSizes = Xpetra::IO<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::ReadMultiVector(hierarchicalParams.get<std::string>("gid_cluster_to_gid_coeff"), clusterMap)->getVectorNonConst(0);
-        blockedClusterMap = rcp(new blocked_map_type(clusterCoeffMap, clusterSizes));
-      }
+    if (hierarchicalParams.isType<std::string>("coefficient map")) {
+      // far field basis expansion coefficients
+      basisMatrix = IOhelpers::Read(hierarchicalParams.get<std::string>("basis expansion coefficient matrix"), map, clusterCoeffMap, clusterCoeffMap, map, true, readBinary, readLocal);
 
-      // near field interactions
-      nearField = Read(hierarchicalParams.get<std::string>("near field matrix"), map, near_colmap, map, map, true, readBinary, readLocal);
-
-      if (hierarchicalParams.isType<std::string>("coefficient map")) {
-        // far field basis expansion coefficients
-        basisMatrix = IOhelpers::Read(hierarchicalParams.get<std::string>("basis expansion coefficient matrix"), map, clusterCoeffMap, clusterCoeffMap, map, true, readBinary, readLocal);
-
-        // far field interactions
-        kernelApproximations = IOhelpers::Read(hierarchicalParams.get<std::string>("far field interaction matrix"), clusterCoeffMap, ghosted_clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, true, readBinary, readLocal);
-        // block graph of far field interactions
-        kernelBlockGraph = IOhelpers::Read(hierarchicalParams.get<std::string>("far field interaction matrix")+".block", clusterMap, ghosted_clusterMap, clusterMap, clusterMap, true, readBinary, readLocal);
-
-        {
-          auto import = kernelBlockGraph->getCrsGraph()->getImporter();
-          RCP<lo_vec_type> ghosted_clusterSizes = Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(ghosted_clusterMap);
-          ghosted_clusterSizes->doImport(*clusterSizes, *import, Xpetra::INSERT);
-          ghosted_blockedClusterMap = rcp(new blocked_map_type(ghosted_clusterCoeffMap, ghosted_clusterSizes));
-        }
-
-        blockKernelApproximations = rcp(new blocked_matrix_type(kernelApproximations, kernelBlockGraph, blockedClusterMap, ghosted_blockedClusterMap));
-
-        // Transfer matrices
-        auto transfersList = hierarchicalParams.sublist("shift coefficient matrices");
-        for (int i = 0; i < transfersList.numParams(); i++) {
-          std::string filename = transfersList.get<std::string>(std::to_string(i));
-          auto transferPoint = IOhelpers::Read(filename, clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, true, readBinary, readLocal);
-          auto transferBlock = IOhelpers::Read(filename+".block", clusterMap, clusterMap, clusterMap, clusterMap, true, readBinary, readLocal);
-          auto transfer = rcp(new blocked_matrix_type(transferPoint, transferBlock, blockedClusterMap));
-          transferMatrices.push_back(transfer);
-        }
-      }
+      // far field interactions
+      kernelApproximations = IOhelpers::Read(hierarchicalParams.get<std::string>("far field interaction matrix"), clusterCoeffMap, ghosted_clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, true, readBinary, readLocal);
+      // block graph of far field interactions
+      kernelBlockGraph = IOhelpers::Read(hierarchicalParams.get<std::string>("far field interaction matrix") + ".block", clusterMap, ghosted_clusterMap, clusterMap, clusterMap, true, readBinary, readLocal);
 
-      RCP<Teuchos::ParameterList> params;
-      if (hierarchicalParams.isSublist("params")) {
-        params = rcp(new Teuchos::ParameterList(hierarchicalParams.sublist("params")));
+      {
+        auto import                           = kernelBlockGraph->getCrsGraph()->getImporter();
+        RCP<lo_vec_type> ghosted_clusterSizes = Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(ghosted_clusterMap);
+        ghosted_clusterSizes->doImport(*clusterSizes, *import, Xpetra::INSERT);
+        ghosted_blockedClusterMap = rcp(new blocked_map_type(ghosted_clusterCoeffMap, ghosted_clusterSizes));
       }
 
-      if (hierarchicalParams.isType<std::string>("coefficient map")) {
-        op = rcp(new HOp(nearField, blockKernelApproximations, basisMatrix, transferMatrices, params));
+      blockKernelApproximations = rcp(new blocked_matrix_type(kernelApproximations, kernelBlockGraph, blockedClusterMap, ghosted_blockedClusterMap));
 
-        return op;
-      } else
-        return nearField;
+      // Transfer matrices
+      auto transfersList = hierarchicalParams.sublist("shift coefficient matrices");
+      for (int i = 0; i < transfersList.numParams(); i++) {
+        std::string filename = transfersList.get<std::string>(std::to_string(i));
+        auto transferPoint   = IOhelpers::Read(filename, clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, clusterCoeffMap, true, readBinary, readLocal);
+        auto transferBlock   = IOhelpers::Read(filename + ".block", clusterMap, clusterMap, clusterMap, clusterMap, true, readBinary, readLocal);
+        auto transfer        = rcp(new blocked_matrix_type(transferPoint, transferBlock, blockedClusterMap));
+        transferMatrices.push_back(transfer);
+      }
+    }
+
+    RCP<Teuchos::ParameterList> params;
+    if (hierarchicalParams.isSublist("params")) {
+      params = rcp(new Teuchos::ParameterList(hierarchicalParams.sublist("params")));
     }
 
-  };
+    if (hierarchicalParams.isType<std::string>("coefficient map")) {
+      op = rcp(new HOp(nearField, blockKernelApproximations, basisMatrix, transferMatrices, params));
+
+      return op;
+    } else
+      return nearField;
+  }
+};
 
-}
+}  // namespace MueLu
 
 #endif
diff --git a/packages/muelu/research/caglusa/Tpetra_BlockedMap_decl.hpp b/packages/muelu/research/caglusa/Tpetra_BlockedMap_decl.hpp
index 497294a2e77e..a04d34127bd2 100644
--- a/packages/muelu/research/caglusa/Tpetra_BlockedMap_decl.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_BlockedMap_decl.hpp
@@ -1,37 +1,33 @@
 #ifndef TPETRA_BLOCKEDMAP_DECL_HPP
 #define TPETRA_BLOCKEDMAP_DECL_HPP
 
-
 #include <Teuchos_RCP.hpp>
 #include <Kokkos_Core.hpp>
 #include <Tpetra_Map.hpp>
 #include <Tpetra_Vector.hpp>
 
-
 namespace Tpetra {
 
-  template <class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class BlockedMap {
-
-  public:
-    using map_type = Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
-    using lo_vec_type = Tpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>;
-
-    BlockedMap(const Teuchos::RCP<const map_type>& pointMap,
-               const Teuchos::RCP<lo_vec_type>& blockSizes);
-
-    // private:
-    Teuchos::RCP<const map_type> pointMap_;
-    Teuchos::RCP<const map_type> blockMap_;
-    Teuchos::RCP<lo_vec_type>    blockSizes_;
-    Kokkos::View<size_t*> offsets_;
-    LocalOrdinal minClusterSize_;
-    LocalOrdinal maxClusterSize_;
-
-  };
-
-}
-
-#endif // TPETRA_BLOCKEDMAP_DECL_HPP
+template <class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class BlockedMap {
+ public:
+  using map_type    = Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+  using lo_vec_type = Tpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>;
+
+  BlockedMap(const Teuchos::RCP<const map_type>& pointMap,
+             const Teuchos::RCP<lo_vec_type>& blockSizes);
+
+  // private:
+  Teuchos::RCP<const map_type> pointMap_;
+  Teuchos::RCP<const map_type> blockMap_;
+  Teuchos::RCP<lo_vec_type> blockSizes_;
+  Kokkos::View<size_t*> offsets_;
+  LocalOrdinal minClusterSize_;
+  LocalOrdinal maxClusterSize_;
+};
+
+}  // namespace Tpetra
+
+#endif  // TPETRA_BLOCKEDMAP_DECL_HPP
diff --git a/packages/muelu/research/caglusa/Tpetra_BlockedMap_def.hpp b/packages/muelu/research/caglusa/Tpetra_BlockedMap_def.hpp
index 49e63cc99023..e654a557db41 100644
--- a/packages/muelu/research/caglusa/Tpetra_BlockedMap_def.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_BlockedMap_def.hpp
@@ -3,37 +3,34 @@
 
 #include <Kokkos_Core.hpp>
 
-
 namespace Tpetra {
 
-  template <class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  BlockedMap<LocalOrdinal,GlobalOrdinal,Node>::
-  BlockedMap(const Teuchos::RCP<const map_type>& pointMap,
-             const Teuchos::RCP<lo_vec_type>& blockSizes)
-    :
-    pointMap_(pointMap),
-    blockMap_(blockSizes->getMap()),
-    blockSizes_(blockSizes)
-  {
-    auto lclBlockSizes = blockSizes_->getLocalViewHost(Tpetra::Access::ReadOnly);
-    LocalOrdinal minClusterSize = Teuchos::OrdinalTraits<LocalOrdinal>::max();
-    LocalOrdinal maxClusterSize = 0;
-    offsets_ = Kokkos::View<size_t*>("offsets", blockMap_->getLocalNumElements()+1);
-    auto offsets_h = Kokkos::create_mirror_view(offsets_);
-    offsets_h(0) = 0;
-    for (size_t blockNum = 0; blockNum < blockMap_->getLocalNumElements(); ++blockNum) {
-      offsets_h(blockNum+1) = offsets_h(blockNum) + lclBlockSizes(blockNum, 0);
-      minClusterSize = std::min(minClusterSize, lclBlockSizes(blockNum, 0));
-      maxClusterSize = std::max(maxClusterSize, lclBlockSizes(blockNum, 0));
-    }
-    Kokkos::deep_copy(offsets_, offsets_h);
-    TEUCHOS_ASSERT_EQUALITY(offsets_h(blockMap_->getLocalNumElements()), pointMap->getLocalNumElements());
-    minClusterSize_ = minClusterSize;
-    maxClusterSize_ = maxClusterSize;
+template <class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+BlockedMap<LocalOrdinal, GlobalOrdinal, Node>::
+    BlockedMap(const Teuchos::RCP<const map_type>& pointMap,
+               const Teuchos::RCP<lo_vec_type>& blockSizes)
+  : pointMap_(pointMap)
+  , blockMap_(blockSizes->getMap())
+  , blockSizes_(blockSizes) {
+  auto lclBlockSizes          = blockSizes_->getLocalViewHost(Tpetra::Access::ReadOnly);
+  LocalOrdinal minClusterSize = Teuchos::OrdinalTraits<LocalOrdinal>::max();
+  LocalOrdinal maxClusterSize = 0;
+  offsets_                    = Kokkos::View<size_t*>("offsets", blockMap_->getLocalNumElements() + 1);
+  auto offsets_h              = Kokkos::create_mirror_view(offsets_);
+  offsets_h(0)                = 0;
+  for (size_t blockNum = 0; blockNum < blockMap_->getLocalNumElements(); ++blockNum) {
+    offsets_h(blockNum + 1) = offsets_h(blockNum) + lclBlockSizes(blockNum, 0);
+    minClusterSize          = std::min(minClusterSize, lclBlockSizes(blockNum, 0));
+    maxClusterSize          = std::max(maxClusterSize, lclBlockSizes(blockNum, 0));
   }
-
+  Kokkos::deep_copy(offsets_, offsets_h);
+  TEUCHOS_ASSERT_EQUALITY(offsets_h(blockMap_->getLocalNumElements()), pointMap->getLocalNumElements());
+  minClusterSize_ = minClusterSize;
+  maxClusterSize_ = maxClusterSize;
 }
 
-#endif // TPETRA_BLOCKEDMAP_DEF_HPP
+}  // namespace Tpetra
+
+#endif  // TPETRA_BLOCKEDMAP_DEF_HPP
diff --git a/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_decl.hpp b/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_decl.hpp
index 8a4554d7ea5e..5242f96853fa 100644
--- a/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_decl.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_decl.hpp
@@ -6,45 +6,42 @@
 #include <Tpetra_CrsMatrix.hpp>
 #include <Tpetra_MultiVector.hpp>
 
-
 namespace Tpetra {
 
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BlockedMatrix {
-
-  public:
-    using matrix_type = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using blocked_map_type = BlockedMap<LocalOrdinal,GlobalOrdinal,Node>;
-    using lo_vec_type = typename blocked_map_type::lo_vec_type;
-
-    BlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
-                  const Teuchos::RCP<matrix_type>& blockA,
-                  const Teuchos::RCP<blocked_map_type>& blockMap,
-                  const Teuchos::RCP<blocked_map_type>& ghosted_blockMap=Teuchos::null);
-
-    void apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    void localApply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                    Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                    Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                    Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                    Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const;
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BlockedMatrix {
+ public:
+  using matrix_type      = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using blocked_map_type = BlockedMap<LocalOrdinal, GlobalOrdinal, Node>;
+  using lo_vec_type      = typename blocked_map_type::lo_vec_type;
+
+  BlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
+                const Teuchos::RCP<matrix_type>& blockA,
+                const Teuchos::RCP<blocked_map_type>& blockMap,
+                const Teuchos::RCP<blocked_map_type>& ghosted_blockMap = Teuchos::null);
+
+  void apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  void localApply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                  Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                  Teuchos::ETransp mode = Teuchos::NO_TRANS,
+                  Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+                  Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
 
   // private:
-    Teuchos::RCP<matrix_type>      pointA_;
-    Teuchos::RCP<matrix_type>      blockA_;
-    Teuchos::RCP<blocked_map_type> blockMap_;
-    Teuchos::RCP<blocked_map_type> ghosted_blockMap_;
-
-  };
+  Teuchos::RCP<matrix_type> pointA_;
+  Teuchos::RCP<matrix_type> blockA_;
+  Teuchos::RCP<blocked_map_type> blockMap_;
+  Teuchos::RCP<blocked_map_type> ghosted_blockMap_;
+};
 
-}
+}  // namespace Tpetra
 
-#endif // TPETRA_BLOCKEDMATRIX_DECL_HPP
+#endif  // TPETRA_BLOCKEDMATRIX_DECL_HPP
diff --git a/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_def.hpp b/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_def.hpp
index 398a2fdff680..392df660e046 100644
--- a/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_def.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_BlockedMatrix_def.hpp
@@ -1,61 +1,55 @@
 #ifndef TPETRA_BLOCKEDMATRIX_DEF_HPP
 #define TPETRA_BLOCKEDMATRIX_DEF_HPP
 
-
 namespace Tpetra {
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  BlockedMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  BlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
-                const Teuchos::RCP<matrix_type>& blockA,
-                const Teuchos::RCP<blocked_map_type>& blockMap,
-                const Teuchos::RCP<blocked_map_type>& ghosted_blockMap)
-    :
-    pointA_(pointA),
-    blockA_(blockA),
-    blockMap_(blockMap),
-    ghosted_blockMap_(ghosted_blockMap)
-  {
-    TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockA_->getRangeMap()));
-    TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockA_->getRowMap()));
-    TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockMap_->blockMap_));
-
-    TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*pointA_->getRangeMap()));
-    TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*pointA_->getRowMap()));
-    TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*blockMap_->pointMap_));
-
-  }
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
+                  const Teuchos::RCP<matrix_type>& blockA,
+                  const Teuchos::RCP<blocked_map_type>& blockMap,
+                  const Teuchos::RCP<blocked_map_type>& ghosted_blockMap)
+  : pointA_(pointA)
+  , blockA_(blockA)
+  , blockMap_(blockMap)
+  , ghosted_blockMap_(ghosted_blockMap) {
+  TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockA_->getRangeMap()));
+  TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockA_->getRowMap()));
+  TEUCHOS_ASSERT(blockA_->getDomainMap()->isSameAs(*blockMap_->blockMap_));
+
+  TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*pointA_->getRangeMap()));
+  TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*pointA_->getRowMap()));
+  TEUCHOS_ASSERT(pointA_->getDomainMap()->isSameAs(*blockMap_->pointMap_));
+}
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  BlockedMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-        Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-        Teuchos::ETransp mode,
-        Scalar alpha,
-        Scalar beta) const {
-    pointA_->apply(X, Y, mode, alpha, beta);
-  }
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+          Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+          Teuchos::ETransp mode,
+          Scalar alpha,
+          Scalar beta) const {
+  pointA_->apply(X, Y, mode, alpha, beta);
+}
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  BlockedMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  localApply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-             Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-             Teuchos::ETransp mode,
-             Scalar alpha,
-             Scalar beta) const {
-    pointA_->localApply(X, Y, mode, alpha, beta);
-  }
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    localApply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+               Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+               Teuchos::ETransp mode,
+               Scalar alpha,
+               Scalar beta) const {
+  pointA_->localApply(X, Y, mode, alpha, beta);
 }
+}  // namespace Tpetra
 
-#endif // TPETRA_BLOCKEDMATRIX_DEF_HPP
+#endif  // TPETRA_BLOCKEDMATRIX_DEF_HPP
diff --git a/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_decl.hpp b/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_decl.hpp
index c96c286a439e..c583d7e05127 100644
--- a/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_decl.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_decl.hpp
@@ -9,368 +9,364 @@
 #include <Tpetra_BlockedMap_decl.hpp>
 #include <Tpetra_BlockedMatrix_decl.hpp>
 
-
 namespace Tpetra {
 
-    /*
-
-    A container class for hierarchical matrices of different types.
-    In particular, both H- and H2-matrices are supported.
-
-    The unknowns of the kernel approximations are collected in the clusterMap.
-    For H-matrices, this is just a concatenation.
-    For H2-matrices, the map also contains the intermediate clusters that might be needed in upward/downward pass.
-
-
-    H = nearField
-        + basisMatrix *
-          ((I+transferMatrices[K-1]^T) * ... * (I+transferMatrices[0]^T)) *
-          kernelApproximations *
-          ((I+transferMatrices[0]) * ... * (I+transferMatrices[K-1])) *
-          basisMatrix^T
-
-   nearField and basisMatrix are standard (point) CRS matrices.
-   kernelApproximations and transferMatrices[.] are blocked CRS matrices
-
-   I is the identity matrix and is not explicitely saved.
-
-   Maps:
-   map (standard): domain and range of H;
-                   domain, range, row of nearField
-   clusterMap (blocked map): domain, range, row, column of transferMatrices;
-                             domain, range, row of kernelApproximations
-   ghosted_clusterMap (blocked map): column of kernelApproximations
-
-
-   For H-matrices:
-   K = 0, i.e. there are no transfer matrices
-
-   For H2-matrices:
-   upward and downward pass in the cluster hierarchy are encoded in transfer matrices
-   */
-
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class HierarchicalOperator : public Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-
-  public:
-    using matrix_type = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using mv_type = Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using map_type = Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node>;
-
-    //! The RowMatrix representing the base class of CrsMatrix
-    using row_matrix_type = RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-
-    using impl_scalar_type = typename row_matrix_type::impl_scalar_type;
-    using mag_type = typename Kokkos::ArithTraits<impl_scalar_type>::mag_type;
-
-    using local_inds_device_view_type =
-          typename row_matrix_type::local_inds_device_view_type;
-    using local_inds_host_view_type =
-          typename row_matrix_type::local_inds_host_view_type;
-    using nonconst_local_inds_host_view_type =
-          typename row_matrix_type::nonconst_local_inds_host_view_type;
-
-    using global_inds_device_view_type =
-          typename row_matrix_type::global_inds_device_view_type;
-    using global_inds_host_view_type =
-          typename row_matrix_type::global_inds_host_view_type;
-    using nonconst_global_inds_host_view_type =
-          typename row_matrix_type::nonconst_global_inds_host_view_type;
-
-    using values_device_view_type =
-          typename row_matrix_type::values_device_view_type;
-    using values_host_view_type =
-          typename row_matrix_type::values_host_view_type;
-    using nonconst_values_host_view_type =
-          typename row_matrix_type::nonconst_values_host_view_type;
-
-    using blocked_matrix_type = BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using blocked_map_type = BlockedMap<LocalOrdinal, GlobalOrdinal, Node>;
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
-                         const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
-                         const Teuchos::RCP<matrix_type>& basisMatrix,
-                         std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
-                         const Teuchos::RCP<Teuchos::ParameterList>& params=Teuchos::null);
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const {
-      return nearField_->getDomainMap();
-    }
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const {
-      return nearField_->getRangeMap();
-    }
-
-    //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
-    /*!
-      \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-
-    Teuchos::RCP<HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > restrict(const Teuchos::RCP<matrix_type>& P);
-
-    Teuchos::RCP<matrix_type> toMatrix();
-
-    double getCompression() {
-      size_t nnz = (nearField_->getGlobalNumEntries() +
-                    kernelApproximations_->pointA_->getGlobalNumEntries() +
-                    basisMatrix_->getGlobalNumEntries());
-      for (size_t i = 0; i < transferMatrices_.size(); i++)
-        nnz += transferMatrices_[i]->pointA_->getGlobalNumEntries();
-      return Teuchos::as<double>(nnz) / (getDomainMap()->getGlobalNumElements()*getDomainMap()->getGlobalNumElements());
-    }
-
-    Teuchos::RCP<matrix_type> nearFieldMatrix() {
-      return nearField_;
-    }
-
-    // Fake RowMatrix interface
-    Teuchos::RCP<const Map<LocalOrdinal,GlobalOrdinal,Node> > getRowMap() const {
-      return nearField_->getRowMap();
-    }
-
-    Teuchos::RCP<const Map<LocalOrdinal,GlobalOrdinal,Node> > getColMap() const {
-      return nearField_->getColMap();
-    }
-
-    Teuchos::RCP<const Teuchos::Comm<int> > getComm() const {
-      return nearField_->getDomainMap()->getComm();
-    }
-
-    Teuchos::RCP<const RowGraph<LocalOrdinal,GlobalOrdinal,Node> > getGraph() const {
-      return nearField_->getCrsGraph();
-    }
-
-    global_size_t getGlobalNumRows() const {
-      return nearField_->getGlobalNumRows();
-    }
-
-    global_size_t getGlobalNumCols() const {
-      return nearField_->getGlobalNumCols();
-    }
-
-    size_t getLocalNumRows() const {
-      return nearField_->getLocalNumRows();
-    }
-
-    size_t getLocalNumCols() const {
-      return nearField_->getLocalNumCols();
-    }
-
-    GlobalOrdinal getIndexBase() const {
-      return nearField_->getIndexBase();
-    }
-
-    global_size_t getGlobalNumEntries() const {
-      return nearField_->getGlobalNumEntries();
-    }
-
-    size_t getLocalNumEntries() const {
-      return nearField_->getLocalNumEntries();
-    }
-
-    size_t getNumEntriesInGlobalRow (GlobalOrdinal globalRow) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    size_t getNumEntriesInLocalRow (LocalOrdinal localRow) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    size_t getGlobalMaxNumRowEntries () const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    LocalOrdinal getBlockSize () const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    size_t getLocalMaxNumRowEntries () const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    bool hasColMap () const {
-      return false;
-    }
-
-    bool isLocallyIndexed() const {
-      return true;
-    }
-
-    bool isGloballyIndexed() const {
-      return true;
-    }
-
-    bool isFillComplete() const {
-      return true;
-    }
-
-    bool supportsRowViews() const {
-      return false;
-    }
-
-    void
-    getGlobalRowCopy (GlobalOrdinal GlobalRow,
-                      nonconst_global_inds_host_view_type &Indices,
-                      nonconst_values_host_view_type &Values,
-                      size_t& NumEntries) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    void
-    getLocalRowCopy (LocalOrdinal LocalRow,
-                     nonconst_local_inds_host_view_type &Indices,
-                     nonconst_values_host_view_type &Values,
-                     size_t& NumEntries) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    void
-    getGlobalRowView (GlobalOrdinal GlobalRow,
-                      global_inds_host_view_type &indices,
-                      values_host_view_type &values) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    void
-    getLocalRowView (LocalOrdinal LocalRow,
-                     local_inds_host_view_type & indices,
-                     values_host_view_type & values) const {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    void getLocalDiagCopy (Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &diag) const {
-      nearField_->getLocalDiagCopy(diag);
-    }
-
-    void leftScale (const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    void rightScale (const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
-      throw std::runtime_error("Not implemented.");
-    }
-
-    mag_type getFrobeniusNorm() const {
-      return 0.;
-    }
-
-    void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
-      describe(out, verbLevel, true);
-    }
-
-    void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel, const bool printHeader) const {
-      using std::setw;
-      using std::endl;
-      const size_t numRows = nearField_->getRowMap()->getGlobalNumElements();
-      const size_t nnzNearField = nearField_->getGlobalNumEntries();
-      const double nnzNearPerRow = Teuchos::as<double>(nnzNearField)/numRows;
-      const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
-      const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
-      const size_t nnzBasis = basisMatrix_->getGlobalNumEntries();
-      size_t numTransfers = transferMatrices_.size();
-      size_t nnzTransfer = 0;
-      for (size_t i = 0; i<transferMatrices_.size(); i++)
-        nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
-      const size_t nnzTotal = nnzNearField+nnzKernelApprox+nnzBasis+nnzTransfer;
-      const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal)/numRows;
-      std::ostringstream oss;
-      oss << std::left;
-      if (printHeader)
-        oss << setw(9) << "rows"  << setw(12)
-            << "nnz(near)"  << setw(14)
-            << "nnz(near)/row" << setw(12)
-            << "nnz(basis)" << setw(15)
-            << "#cluster pairs" << setw(12)
-            << "nnz(kernel)"    << setw(14)
-            << "#transfers"    << setw(14)
-            << "nnz(transfer)" << setw(12)
-            << "nnz(total)" << setw(14)
-            << "nnz(total)/row" << endl;
-      oss << setw(9) << numRows << setw(12)
-          << nnzNearField << setw(14)
-          << nnzNearPerRow   << setw(12)
-          << nnzBasis     << setw(15)
-          << numClusterPairs  << setw(12)
-          << nnzKernelApprox << setw(14)
-          << numTransfers    << setw(14)
-          << nnzTransfer     << setw(12)
-          << nnzTotal     << setw(14)
-          << nnzTotalPerRow   << endl;
-      out << oss.str();
-    }
-
-    bool hasFarField() const {
-      return kernelApproximations_->blockA_->getGlobalNumEntries() > 0;
-    }
-
-    bool hasTransferMatrices() const {
-      return transferMatrices_.size() > 0;
-    }
-
-    bool denserThanDenseMatrix() const {
-      const size_t numRows = nearField_->getRowMap()->getGlobalNumElements();
-      const size_t nnzNearField = nearField_->getGlobalNumEntries();
-      // const double nnzNearPerRow = Teuchos::as<double>(nnzNearField)/numRows;
-      const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
-      // const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
-      const size_t nnzBasis = basisMatrix_->getGlobalNumEntries();
-      size_t nnzTransfer = 0;
-      for (size_t i = 0; i<transferMatrices_.size(); i++)
-        nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
-      const size_t nnzTotal = nnzNearField+nnzKernelApprox+nnzBasis+nnzTransfer;
-      const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal)/numRows;
-
-      return (nnzTotalPerRow >= numRows);
-    }
-
-  private:
-
-    void allocateMemory(size_t numVectors) const;
-
-    void applyWithTransposes(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                             Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                             Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                             Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                             Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    void applyWithoutTransposes(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                                Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                                Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    bool canApplyWithoutTransposes_;
-    std::string coarseningCriterion_;
-    bool debugOutput_;
-
-    Teuchos::RCP<matrix_type> nearField_;
-    Teuchos::RCP<blocked_matrix_type> kernelApproximations_;
-    Teuchos::RCP<matrix_type> basisMatrix_;
-    Teuchos::RCP<matrix_type> basisMatrixT_;
-    std::vector<Teuchos::RCP<blocked_matrix_type> > transferMatrices_;
-    std::vector<Teuchos::RCP<blocked_matrix_type> > transferMatricesT_;
-    Teuchos::RCP<const map_type> clusterCoeffMap_;
-    mutable Teuchos::RCP<mv_type> coefficients_, coefficients2_;
-    mutable Teuchos::RCP<mv_type> X_colmap_, coefficients_colmap_;
-
-    Teuchos::RCP<Teuchos::ParameterList> params_;
-  };
-}
-
-#endif // TPETRA_HIERARCHICALOPERATOR_DECL_HPP
+/*
+
+A container class for hierarchical matrices of different types.
+In particular, both H- and H2-matrices are supported.
+
+The unknowns of the kernel approximations are collected in the clusterMap.
+For H-matrices, this is just a concatenation.
+For H2-matrices, the map also contains the intermediate clusters that might be needed in upward/downward pass.
+
+
+H = nearField
+    + basisMatrix *
+      ((I+transferMatrices[K-1]^T) * ... * (I+transferMatrices[0]^T)) *
+      kernelApproximations *
+      ((I+transferMatrices[0]) * ... * (I+transferMatrices[K-1])) *
+      basisMatrix^T
+
+nearField and basisMatrix are standard (point) CRS matrices.
+kernelApproximations and transferMatrices[.] are blocked CRS matrices
+
+I is the identity matrix and is not explicitely saved.
+
+Maps:
+map (standard): domain and range of H;
+               domain, range, row of nearField
+clusterMap (blocked map): domain, range, row, column of transferMatrices;
+                         domain, range, row of kernelApproximations
+ghosted_clusterMap (blocked map): column of kernelApproximations
+
+
+For H-matrices:
+K = 0, i.e. there are no transfer matrices
+
+For H2-matrices:
+upward and downward pass in the cluster hierarchy are encoded in transfer matrices
+*/
+
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class HierarchicalOperator : public Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  using matrix_type = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using mv_type     = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using map_type    = Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+
+  //! The RowMatrix representing the base class of CrsMatrix
+  using row_matrix_type = RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  using impl_scalar_type = typename row_matrix_type::impl_scalar_type;
+  using mag_type         = typename Kokkos::ArithTraits<impl_scalar_type>::mag_type;
+
+  using local_inds_device_view_type =
+      typename row_matrix_type::local_inds_device_view_type;
+  using local_inds_host_view_type =
+      typename row_matrix_type::local_inds_host_view_type;
+  using nonconst_local_inds_host_view_type =
+      typename row_matrix_type::nonconst_local_inds_host_view_type;
+
+  using global_inds_device_view_type =
+      typename row_matrix_type::global_inds_device_view_type;
+  using global_inds_host_view_type =
+      typename row_matrix_type::global_inds_host_view_type;
+  using nonconst_global_inds_host_view_type =
+      typename row_matrix_type::nonconst_global_inds_host_view_type;
+
+  using values_device_view_type =
+      typename row_matrix_type::values_device_view_type;
+  using values_host_view_type =
+      typename row_matrix_type::values_host_view_type;
+  using nonconst_values_host_view_type =
+      typename row_matrix_type::nonconst_values_host_view_type;
+
+  using blocked_matrix_type = BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using blocked_map_type    = BlockedMap<LocalOrdinal, GlobalOrdinal, Node>;
+
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
+                       const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
+                       const Teuchos::RCP<matrix_type>& basisMatrix,
+                       std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
+                       const Teuchos::RCP<Teuchos::ParameterList>& params = Teuchos::null);
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const {
+    return nearField_->getDomainMap();
+  }
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const {
+    return nearField_->getRangeMap();
+  }
+
+  //! Returns in Y the result of a Tpetra::Operator applied to a Tpetra::MultiVector X.
+  /*!
+    \param[in]  X - Tpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y -Tpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  Teuchos::RCP<HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > restrict(const Teuchos::RCP<matrix_type>& P);
+
+  Teuchos::RCP<matrix_type> toMatrix();
+
+  double getCompression() {
+    size_t nnz = (nearField_->getGlobalNumEntries() +
+                  kernelApproximations_->pointA_->getGlobalNumEntries() +
+                  basisMatrix_->getGlobalNumEntries());
+    for (size_t i = 0; i < transferMatrices_.size(); i++)
+      nnz += transferMatrices_[i]->pointA_->getGlobalNumEntries();
+    return Teuchos::as<double>(nnz) / (getDomainMap()->getGlobalNumElements() * getDomainMap()->getGlobalNumElements());
+  }
+
+  Teuchos::RCP<matrix_type> nearFieldMatrix() {
+    return nearField_;
+  }
+
+  // Fake RowMatrix interface
+  Teuchos::RCP<const Map<LocalOrdinal, GlobalOrdinal, Node> > getRowMap() const {
+    return nearField_->getRowMap();
+  }
+
+  Teuchos::RCP<const Map<LocalOrdinal, GlobalOrdinal, Node> > getColMap() const {
+    return nearField_->getColMap();
+  }
+
+  Teuchos::RCP<const Teuchos::Comm<int> > getComm() const {
+    return nearField_->getDomainMap()->getComm();
+  }
+
+  Teuchos::RCP<const RowGraph<LocalOrdinal, GlobalOrdinal, Node> > getGraph() const {
+    return nearField_->getCrsGraph();
+  }
+
+  global_size_t getGlobalNumRows() const {
+    return nearField_->getGlobalNumRows();
+  }
+
+  global_size_t getGlobalNumCols() const {
+    return nearField_->getGlobalNumCols();
+  }
+
+  size_t getLocalNumRows() const {
+    return nearField_->getLocalNumRows();
+  }
+
+  size_t getLocalNumCols() const {
+    return nearField_->getLocalNumCols();
+  }
+
+  GlobalOrdinal getIndexBase() const {
+    return nearField_->getIndexBase();
+  }
+
+  global_size_t getGlobalNumEntries() const {
+    return nearField_->getGlobalNumEntries();
+  }
+
+  size_t getLocalNumEntries() const {
+    return nearField_->getLocalNumEntries();
+  }
+
+  size_t getNumEntriesInGlobalRow(GlobalOrdinal globalRow) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  size_t getNumEntriesInLocalRow(LocalOrdinal localRow) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  size_t getGlobalMaxNumRowEntries() const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  LocalOrdinal getBlockSize() const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  size_t getLocalMaxNumRowEntries() const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  bool hasColMap() const {
+    return false;
+  }
+
+  bool isLocallyIndexed() const {
+    return true;
+  }
+
+  bool isGloballyIndexed() const {
+    return true;
+  }
+
+  bool isFillComplete() const {
+    return true;
+  }
+
+  bool supportsRowViews() const {
+    return false;
+  }
+
+  void
+  getGlobalRowCopy(GlobalOrdinal GlobalRow,
+                   nonconst_global_inds_host_view_type& Indices,
+                   nonconst_values_host_view_type& Values,
+                   size_t& NumEntries) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  void
+  getLocalRowCopy(LocalOrdinal LocalRow,
+                  nonconst_local_inds_host_view_type& Indices,
+                  nonconst_values_host_view_type& Values,
+                  size_t& NumEntries) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  void
+  getGlobalRowView(GlobalOrdinal GlobalRow,
+                   global_inds_host_view_type& indices,
+                   values_host_view_type& values) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  void
+  getLocalRowView(LocalOrdinal LocalRow,
+                  local_inds_host_view_type& indices,
+                  values_host_view_type& values) const {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  void getLocalDiagCopy(Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& diag) const {
+    nearField_->getLocalDiagCopy(diag);
+  }
+
+  void leftScale(const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  void rightScale(const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+    throw std::runtime_error("Not implemented.");
+  }
+
+  mag_type getFrobeniusNorm() const {
+    return 0.;
+  }
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
+    describe(out, verbLevel, true);
+  }
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel, const bool printHeader) const {
+    using std::endl;
+    using std::setw;
+    const size_t numRows         = nearField_->getRowMap()->getGlobalNumElements();
+    const size_t nnzNearField    = nearField_->getGlobalNumEntries();
+    const double nnzNearPerRow   = Teuchos::as<double>(nnzNearField) / numRows;
+    const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
+    const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
+    const size_t nnzBasis        = basisMatrix_->getGlobalNumEntries();
+    size_t numTransfers          = transferMatrices_.size();
+    size_t nnzTransfer           = 0;
+    for (size_t i = 0; i < transferMatrices_.size(); i++)
+      nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
+    const size_t nnzTotal       = nnzNearField + nnzKernelApprox + nnzBasis + nnzTransfer;
+    const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal) / numRows;
+    std::ostringstream oss;
+    oss << std::left;
+    if (printHeader)
+      oss << setw(9) << "rows" << setw(12)
+          << "nnz(near)" << setw(14)
+          << "nnz(near)/row" << setw(12)
+          << "nnz(basis)" << setw(15)
+          << "#cluster pairs" << setw(12)
+          << "nnz(kernel)" << setw(14)
+          << "#transfers" << setw(14)
+          << "nnz(transfer)" << setw(12)
+          << "nnz(total)" << setw(14)
+          << "nnz(total)/row" << endl;
+    oss << setw(9) << numRows << setw(12)
+        << nnzNearField << setw(14)
+        << nnzNearPerRow << setw(12)
+        << nnzBasis << setw(15)
+        << numClusterPairs << setw(12)
+        << nnzKernelApprox << setw(14)
+        << numTransfers << setw(14)
+        << nnzTransfer << setw(12)
+        << nnzTotal << setw(14)
+        << nnzTotalPerRow << endl;
+    out << oss.str();
+  }
+
+  bool hasFarField() const {
+    return kernelApproximations_->blockA_->getGlobalNumEntries() > 0;
+  }
+
+  bool hasTransferMatrices() const {
+    return transferMatrices_.size() > 0;
+  }
+
+  bool denserThanDenseMatrix() const {
+    const size_t numRows      = nearField_->getRowMap()->getGlobalNumElements();
+    const size_t nnzNearField = nearField_->getGlobalNumEntries();
+    // const double nnzNearPerRow = Teuchos::as<double>(nnzNearField)/numRows;
+    const size_t nnzKernelApprox = kernelApproximations_->pointA_->getGlobalNumEntries();
+    // const size_t numClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
+    const size_t nnzBasis = basisMatrix_->getGlobalNumEntries();
+    size_t nnzTransfer    = 0;
+    for (size_t i = 0; i < transferMatrices_.size(); i++)
+      nnzTransfer += transferMatrices_[i]->pointA_->getGlobalNumEntries();
+    const size_t nnzTotal       = nnzNearField + nnzKernelApprox + nnzBasis + nnzTransfer;
+    const double nnzTotalPerRow = Teuchos::as<double>(nnzTotal) / numRows;
+
+    return (nnzTotalPerRow >= numRows);
+  }
+
+ private:
+  void allocateMemory(size_t numVectors) const;
+
+  void applyWithTransposes(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                           Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                           Teuchos::ETransp mode = Teuchos::NO_TRANS,
+                           Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+                           Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  void applyWithoutTransposes(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                              Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                              Teuchos::ETransp mode = Teuchos::NO_TRANS,
+                              Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+                              Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  bool canApplyWithoutTransposes_;
+  std::string coarseningCriterion_;
+  bool debugOutput_;
+
+  Teuchos::RCP<matrix_type> nearField_;
+  Teuchos::RCP<blocked_matrix_type> kernelApproximations_;
+  Teuchos::RCP<matrix_type> basisMatrix_;
+  Teuchos::RCP<matrix_type> basisMatrixT_;
+  std::vector<Teuchos::RCP<blocked_matrix_type> > transferMatrices_;
+  std::vector<Teuchos::RCP<blocked_matrix_type> > transferMatricesT_;
+  Teuchos::RCP<const map_type> clusterCoeffMap_;
+  mutable Teuchos::RCP<mv_type> coefficients_, coefficients2_;
+  mutable Teuchos::RCP<mv_type> X_colmap_, coefficients_colmap_;
+
+  Teuchos::RCP<Teuchos::ParameterList> params_;
+};
+}  // namespace Tpetra
+
+#endif  // TPETRA_HIERARCHICALOPERATOR_DECL_HPP
diff --git a/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_def.hpp b/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_def.hpp
index c1abd28bdd81..67af4b36c4e6 100644
--- a/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_def.hpp
+++ b/packages/muelu/research/caglusa/Tpetra_HierarchicalOperator_def.hpp
@@ -4,951 +4,936 @@
 #include <TpetraExt_MatrixMatrix.hpp>
 #include <Tpetra_RowMatrixTransposer.hpp>
 
-
 namespace Tpetra {
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  removeSmallEntries(Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-                     typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol) {
-
-    using crs_matrix = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using row_ptr_type = typename crs_matrix::local_graph_device_type::row_map_type::non_const_type;
-    using col_idx_type = typename crs_matrix::local_graph_device_type::entries_type::non_const_type;
-    using vals_type = typename crs_matrix::local_matrix_device_type::values_type;
-
-    typedef Kokkos::ArithTraits<Scalar> ATS;
-    using impl_SC = typename ATS::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_SC>;
-
-    auto lclA = A->getLocalMatrixDevice();
-
-    auto rowptr = row_ptr_type("rowptr", lclA.numRows()+1);
-
-    Kokkos::parallel_for("removeSmallEntries::rowptr1",
-                         Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
-                         KOKKOS_LAMBDA(const LocalOrdinal rlid) {
-                           auto row = lclA.row(rlid);
-                           for (LocalOrdinal k = 0; k<row.length; ++k) {
-                             if (impl_ATS::magnitude(row.value(k)) > tol) {
-                               rowptr(rlid+1) += 1;
-                             }
-                           }
-                         });
-    LocalOrdinal nnz;
-    Kokkos::parallel_scan("removeSmallEntries::rowptr2",
-                          Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
-                          KOKKOS_LAMBDA(const LocalOrdinal rlid, LocalOrdinal& partial_nnz, bool is_final) {
-
-                            partial_nnz += rowptr(rlid+1);
-                            if (is_final)
-                              rowptr(rlid+1) = partial_nnz;
-
-                          }, nnz);
-
-    // auto nnz = rowptr(lclA.numRows());
-
-    auto idx = col_idx_type("idx", nnz);
-    auto vals = vals_type("vals", nnz);
-
-    Kokkos::parallel_for("removeSmallEntries::indicesValues",
-                         Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
-                         KOKKOS_LAMBDA(const LocalOrdinal rlid) {
-                           auto row = lclA.row(rlid);
-                           auto I = rowptr(rlid);
-                           for (LocalOrdinal k = 0; k<row.length; ++k) {
-                             if (impl_ATS::magnitude(row.value(k)) > tol) {
-                               idx(I) = row.colidx(k);
-                               vals(I) = row.value(k);
-                               I += 1;
-                             }
-                           }
-                         });
-
-    auto newA =  Teuchos::rcp(new crs_matrix(A->getRowMap(), A->getColMap(), rowptr, idx, vals));
-    newA->fillComplete(A->getDomainMap(),
-                       A->getRangeMap());
-    return newA;
-  }
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+removeSmallEntries(Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
+                   typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol) {
+  using crs_matrix   = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using row_ptr_type = typename crs_matrix::local_graph_device_type::row_map_type::non_const_type;
+  using col_idx_type = typename crs_matrix::local_graph_device_type::entries_type::non_const_type;
+  using vals_type    = typename crs_matrix::local_matrix_device_type::values_type;
+
+  typedef Kokkos::ArithTraits<Scalar> ATS;
+  using impl_SC  = typename ATS::val_type;
+  using impl_ATS = Kokkos::ArithTraits<impl_SC>;
+
+  auto lclA = A->getLocalMatrixDevice();
+
+  auto rowptr = row_ptr_type("rowptr", lclA.numRows() + 1);
+
+  Kokkos::parallel_for(
+      "removeSmallEntries::rowptr1",
+      Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
+      KOKKOS_LAMBDA(const LocalOrdinal rlid) {
+        auto row = lclA.row(rlid);
+        for (LocalOrdinal k = 0; k < row.length; ++k) {
+          if (impl_ATS::magnitude(row.value(k)) > tol) {
+            rowptr(rlid + 1) += 1;
+          }
+        }
+      });
+  LocalOrdinal nnz;
+  Kokkos::parallel_scan(
+      "removeSmallEntries::rowptr2",
+      Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
+      KOKKOS_LAMBDA(const LocalOrdinal rlid, LocalOrdinal& partial_nnz, bool is_final) {
+        partial_nnz += rowptr(rlid + 1);
+        if (is_final)
+          rowptr(rlid + 1) = partial_nnz;
+      },
+      nnz);
+
+  // auto nnz = rowptr(lclA.numRows());
+
+  auto idx  = col_idx_type("idx", nnz);
+  auto vals = vals_type("vals", nnz);
+
+  Kokkos::parallel_for(
+      "removeSmallEntries::indicesValues",
+      Kokkos::RangePolicy<LocalOrdinal>(0, lclA.numRows()),
+      KOKKOS_LAMBDA(const LocalOrdinal rlid) {
+        auto row = lclA.row(rlid);
+        auto I   = rowptr(rlid);
+        for (LocalOrdinal k = 0; k < row.length; ++k) {
+          if (impl_ATS::magnitude(row.value(k)) > tol) {
+            idx(I)  = row.colidx(k);
+            vals(I) = row.value(k);
+            I += 1;
+          }
+        }
+      });
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  Teuchos::RCP<Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  transpose(Teuchos::RCP<Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) {
+  auto newA = Teuchos::rcp(new crs_matrix(A->getRowMap(), A->getColMap(), rowptr, idx, vals));
+  newA->fillComplete(A->getDomainMap(),
+                     A->getRangeMap());
+  return newA;
+}
 
-    TEUCHOS_ASSERT(A->ghosted_blockMap_.is_null());
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+Teuchos::RCP<Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+transpose(Teuchos::RCP<Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) {
+  TEUCHOS_ASSERT(A->ghosted_blockMap_.is_null());
 
-    Teuchos::RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
+  Teuchos::RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
+
+  Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerPoint(A->pointA_);
+  auto pointAT = transposerPoint.createTranspose(transposeParams);
 
-    Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerPoint(A->pointA_);
-    auto pointAT = transposerPoint.createTranspose(transposeParams);
+  Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerBlock(A->blockA_);
+  auto blockAT = transposerBlock.createTranspose(transposeParams);
 
-    Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerBlock(A->blockA_);
-    auto blockAT = transposerBlock.createTranspose(transposeParams);
+  auto AT = Teuchos::rcp(new Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(pointAT, blockAT, A->blockMap_));
 
-    auto AT = Teuchos::rcp(new Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(pointAT, blockAT, A->blockMap_));
+  return AT;
+}
 
-    return AT;
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+buildIdentityMatrix(Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map) {
+  using matrix_type                               = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  Teuchos::RCP<matrix_type> identity              = Teuchos::rcp(new matrix_type(map, 1));
+  Teuchos::ArrayView<const GlobalOrdinal> gblRows = map->getLocalElementList();
+  for (auto it = gblRows.begin(); it != gblRows.end(); ++it) {
+    Teuchos::Array<GlobalOrdinal> col(1, *it);
+    Teuchos::Array<Scalar> val(1, Teuchos::ScalarTraits<Scalar>::one());
+    identity->insertGlobalValues(*it, col(), val());
   }
+  identity->fillComplete();
+  return identity;
+}
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  buildIdentityMatrix(Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map) {
-    using matrix_type = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    Teuchos::RCP<matrix_type> identity = Teuchos::rcp(new matrix_type(map, 1));
-    Teuchos::ArrayView<const GlobalOrdinal> gblRows = map->getLocalElementList ();
-    for (auto it = gblRows.begin (); it != gblRows.end (); ++it) {
-      Teuchos::Array<GlobalOrdinal> col (1, *it);
-      Teuchos::Array<Scalar> val (1, Teuchos::ScalarTraits<Scalar>::one());
-      identity->insertGlobalValues (*it, col (), val ());
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
+                         const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
+                         const Teuchos::RCP<matrix_type>& basisMatrix,
+                         std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
+                         const Teuchos::RCP<Teuchos::ParameterList>& params)
+  : nearField_(nearField)
+  , kernelApproximations_(kernelApproximations)
+  , basisMatrix_(basisMatrix)
+  , transferMatrices_(transferMatrices)
+  , params_(params) {
+  auto map         = nearField_->getDomainMap();
+  clusterCoeffMap_ = basisMatrix_->getDomainMap();
+
+  bool setupTransposes;
+  bool doDebugChecks;
+  std::string sendTypeNearField;
+  std::string sendTypeBasisMatrix;
+  std::string sendTypeKernelApproximations;
+
+  Teuchos::ParameterList defaultParams("Default params");
+  defaultParams.set("setupTransposes", true);
+  defaultParams.set("doDebugChecks", true);
+  defaultParams.set("Send type nearField", "Isend");
+  defaultParams.set("Send type basisMatrix", "Isend");
+  defaultParams.set("Send type kernelApproximations", "Alltoall");
+  defaultParams.set("Coarsening criterion", "transferLevels");
+  defaultParams.set("debugOutput", false);
+  defaultParams.set("keepTransfers", -1);
+  defaultParams.set("treeCoarseningFactor", 2.0);
+  defaultParams.set("leftOverFactor", 1.0);
+  if (params_.is_null())
+    params_ = Teuchos::rcp(new Teuchos::ParameterList(""));
+  params_->validateParametersAndSetDefaults(defaultParams);
+
+  setupTransposes              = params_->get<bool>("setupTransposes");
+  doDebugChecks                = params_->get<bool>("doDebugChecks");
+  sendTypeNearField            = params_->get<std::string>("Send type nearField");
+  sendTypeBasisMatrix          = params_->get<std::string>("Send type basisMatrix");
+  sendTypeKernelApproximations = params_->get<std::string>("Send type kernelApproximations");
+  coarseningCriterion_         = params_->get<std::string>("Coarsening criterion");
+  TEUCHOS_ASSERT((coarseningCriterion_ == "numClusters") || (coarseningCriterion_ == "equivalentDense") || (coarseningCriterion_ == "transferLevels"));
+  debugOutput_ = params_->get<bool>("debugOutput");
+
+  if (doDebugChecks) {
+    // near field matrix lives on map and is nonlocal
+    TEUCHOS_ASSERT(map->isSameAs(*nearField_->getRangeMap()));
+    TEUCHOS_ASSERT(map->isSameAs(*nearField_->getRowMap()));
+
+    // basis matrix is entirely local and maps from clusterCoeffMap_ to map.
+    TEUCHOS_ASSERT(map->isSameAs(*basisMatrix->getRangeMap()));
+    TEUCHOS_ASSERT(map->isSameAs(*basisMatrix->getRowMap()));
+    // TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*basisMatrix->getDomainMap()));
+
+    // kernel approximations live on clusterCoeffMap and are nonlocal
+    TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getDomainMap()));
+    TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getRangeMap()));
+    TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getRowMap()));
+
+    for (size_t i = 0; i < transferMatrices_.size(); i++) {
+      // transfer matrices are entirely local, block diagonal on clusterCoeffMap
+      TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getDomainMap()));
+      TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getColMap()));
+      TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getRowMap()));
+      TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getRangeMap()));
     }
-    identity->fillComplete ();
-    return identity;
   }
 
+  // Set the send types
+  Teuchos::RCP<Teuchos::ParameterList> distParams = rcp(new Teuchos::ParameterList());
+  {
+    distParams->set("Send type", sendTypeNearField);
+    Teuchos::RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
+    nearFieldImporter->getDistributor().setParameterList(distParams);
+    auto revDistor = nearFieldImporter->getDistributor().getReverse(false);
+    if (!revDistor.is_null())
+      revDistor->setParameterList(distParams);
+  }
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
-                       const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
-                       const Teuchos::RCP<matrix_type>& basisMatrix,
-                       std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
-                       const Teuchos::RCP<Teuchos::ParameterList>& params)
-      :
-      nearField_(nearField),
-      kernelApproximations_(kernelApproximations),
-      basisMatrix_(basisMatrix),
-      transferMatrices_(transferMatrices),
-      params_(params)
-    {
-      auto map = nearField_->getDomainMap();
-      clusterCoeffMap_ = basisMatrix_->getDomainMap();
-
-      bool setupTransposes;
-      bool doDebugChecks;
-      std::string sendTypeNearField;
-      std::string sendTypeBasisMatrix;
-      std::string sendTypeKernelApproximations;
-
-      Teuchos::ParameterList defaultParams("Default params");
-      defaultParams.set("setupTransposes", true);
-      defaultParams.set("doDebugChecks", true);
-      defaultParams.set("Send type nearField", "Isend");
-      defaultParams.set("Send type basisMatrix", "Isend");
-      defaultParams.set("Send type kernelApproximations", "Alltoall");
-      defaultParams.set("Coarsening criterion", "transferLevels");
-      defaultParams.set("debugOutput", false);
-      defaultParams.set("keepTransfers", -1);
-      defaultParams.set("treeCoarseningFactor", 2.0);
-      defaultParams.set("leftOverFactor", 1.0);
-      if (params_.is_null())
-        params_ = Teuchos::rcp(new Teuchos::ParameterList(""));
-      params_->validateParametersAndSetDefaults(defaultParams);
-
-      setupTransposes = params_->get<bool>("setupTransposes");
-      doDebugChecks = params_->get<bool>("doDebugChecks");
-      sendTypeNearField = params_->get<std::string>("Send type nearField");
-      sendTypeBasisMatrix = params_->get<std::string>("Send type basisMatrix");
-      sendTypeKernelApproximations = params_->get<std::string>("Send type kernelApproximations");
-      coarseningCriterion_ = params_->get<std::string>("Coarsening criterion");
-      TEUCHOS_ASSERT((coarseningCriterion_ == "numClusters") || (coarseningCriterion_ == "equivalentDense") || (coarseningCriterion_ == "transferLevels"));
-      debugOutput_ = params_->get<bool>("debugOutput");
-
-      if (doDebugChecks) {
-        // near field matrix lives on map and is nonlocal
-        TEUCHOS_ASSERT(map->isSameAs(*nearField_->getRangeMap()));
-        TEUCHOS_ASSERT(map->isSameAs(*nearField_->getRowMap()));
-
-        // basis matrix is entirely local and maps from clusterCoeffMap_ to map.
-        TEUCHOS_ASSERT(map->isSameAs(*basisMatrix->getRangeMap()));
-        TEUCHOS_ASSERT(map->isSameAs(*basisMatrix->getRowMap()));
-        // TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*basisMatrix->getDomainMap()));
-
-        // kernel approximations live on clusterCoeffMap and are nonlocal
-        TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getDomainMap()));
-        TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getRangeMap()));
-        TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*kernelApproximations_->pointA_->getRowMap()));
-
-        for (size_t i = 0; i<transferMatrices_.size(); i++) {
-          // transfer matrices are entirely local, block diagonal on clusterCoeffMap
-          TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getDomainMap()));
-          TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getColMap()));
-          TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getRowMap()));
-          TEUCHOS_ASSERT(clusterCoeffMap_->isSameAs(*transferMatrices_[i]->pointA_->getRangeMap()));
-        }
-      }
-
-      // Set the send types
-      Teuchos::RCP<Teuchos::ParameterList> distParams = rcp(new Teuchos::ParameterList());
-      {
-        distParams->set("Send type", sendTypeNearField);
-        Teuchos::RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
-        nearFieldImporter->getDistributor().setParameterList(distParams);
-        auto revDistor = nearFieldImporter->getDistributor().getReverse(false);
-        if (!revDistor.is_null())
-          revDistor->setParameterList(distParams);
-      }
-
-      {
-        distParams->set("Send type", sendTypeBasisMatrix);
-        Teuchos::RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > basisMatrixImporter = basisMatrix_->getGraph()->getImporter();
-        if (!basisMatrixImporter.is_null()) {
-          basisMatrixImporter->getDistributor().setParameterList(distParams);
-          auto revDistor = basisMatrixImporter->getDistributor().getReverse(false);
-          if (!revDistor.is_null())
-            revDistor->setParameterList(distParams);
-        }
-      }
-
-      {
-        distParams->set("Send type", sendTypeKernelApproximations);
-        Teuchos::RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
-        kernelApproximationsImporter->getDistributor().setParameterList(distParams);
-        auto revDistor = kernelApproximationsImporter->getDistributor().getReverse(false);
-        if (!revDistor.is_null())
-          revDistor->setParameterList(distParams);
-      }
-
-      if (setupTransposes) {
-        Teuchos::RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
-
-        Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerBasisMatrix(basisMatrix_);
-        basisMatrixT_ = transposerBasisMatrix.createTranspose(transposeParams);
-
-        for (size_t i = 0; i<transferMatrices_.size(); i++) {
-          transferMatricesT_.push_back(transpose(transferMatrices_[i]));
-        }
-
-        canApplyWithoutTransposes_ = true;
-      } else
-        canApplyWithoutTransposes_ = false;
-
-      // Allocate memory for apply with vectors
-      allocateMemory(1);
+  {
+    distParams->set("Send type", sendTypeBasisMatrix);
+    Teuchos::RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > basisMatrixImporter = basisMatrix_->getGraph()->getImporter();
+    if (!basisMatrixImporter.is_null()) {
+      basisMatrixImporter->getDistributor().setParameterList(distParams);
+      auto revDistor = basisMatrixImporter->getDistributor().getReverse(false);
+      if (!revDistor.is_null())
+        revDistor->setParameterList(distParams);
     }
+  }
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  apply(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-        Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-        Teuchos::ETransp mode,
-        Scalar alpha,
-        Scalar beta) const {
-    if (canApplyWithoutTransposes_)
-      applyWithoutTransposes(X, Y, mode, alpha, beta);
-    else
-      applyWithTransposes(X, Y, mode, alpha, beta);
+  {
+    distParams->set("Send type", sendTypeKernelApproximations);
+    Teuchos::RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+    kernelApproximationsImporter->getDistributor().setParameterList(distParams);
+    auto revDistor = kernelApproximationsImporter->getDistributor().getReverse(false);
+    if (!revDistor.is_null())
+      revDistor->setParameterList(distParams);
   }
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  applyWithTransposes(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                      Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                      Teuchos::ETransp mode,
-                      Scalar alpha,
-                      Scalar beta) const {
-    using Teuchos::RCP;
-    const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    bool flip = true;
-
-    allocateMemory(X.getNumVectors());
-
-    // upward pass
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("upward pass")));
+  if (setupTransposes) {
+    Teuchos::RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
 
-      basisMatrix_->apply(X, *coefficients_, Teuchos::TRANS);
+    Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposerBasisMatrix(basisMatrix_);
+    basisMatrixT_ = transposerBasisMatrix.createTranspose(transposeParams);
 
-      for (int i = Teuchos::as<int>(transferMatrices_.size())-1; i>=0; i--)
-        if (flip) {
-          coefficients2_->assign(*coefficients_);
-          transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
-          flip = false;
-        } else {
-          coefficients_->assign(*coefficients2_);
-          transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
-          flip = true;
-        }
+    for (size_t i = 0; i < transferMatrices_.size(); i++) {
+      transferMatricesT_.push_back(transpose(transferMatrices_[i]));
     }
 
-    // far field interactions - part 1
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 1")));
+    canApplyWithoutTransposes_ = true;
+  } else
+    canApplyWithoutTransposes_ = false;
+
+  // Allocate memory for apply with vectors
+  allocateMemory(1);
+}
+
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    apply(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+          Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+          Teuchos::ETransp mode,
+          Scalar alpha,
+          Scalar beta) const {
+  if (canApplyWithoutTransposes_)
+    applyWithoutTransposes(X, Y, mode, alpha, beta);
+  else
+    applyWithTransposes(X, Y, mode, alpha, beta);
+}
 
-      RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    applyWithTransposes(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                        Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                        Teuchos::ETransp mode,
+                        Scalar alpha,
+                        Scalar beta) const {
+  using Teuchos::RCP;
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+  bool flip         = true;
+
+  allocateMemory(X.getNumVectors());
+
+  // upward pass
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("upward pass")));
+
+    basisMatrix_->apply(X, *coefficients_, Teuchos::TRANS);
+
+    for (int i = Teuchos::as<int>(transferMatrices_.size()) - 1; i >= 0; i--)
       if (flip) {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->beginImport(*coefficients_, *kernelApproximationsImporter, INSERT);
-        } else if (mode == Teuchos::TRANS) {
-          kernelApproximations_->localApply(*coefficients_, *coefficients_colmap_, mode, alpha);
-          coefficients2_->putScalar(zero);
-          coefficients2_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
+        coefficients2_->assign(*coefficients_);
+        transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
+        flip = false;
       } else {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->beginImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
-        } else if (mode == Teuchos::TRANS) {
-          kernelApproximations_->localApply(*coefficients2_, *coefficients_colmap_, mode, alpha);
-          coefficients_->putScalar(zero);
-          coefficients_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
+        coefficients_->assign(*coefficients2_);
+        transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
+        flip = true;
       }
-    }
+  }
 
-    // near field - part 1
-    RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 1")));
+  // far field interactions - part 1
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 1")));
+
+    RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+    if (flip) {
       if (mode == Teuchos::NO_TRANS) {
-        X_colmap_->beginImport(X, *nearFieldImporter, INSERT);
+        coefficients_colmap_->beginImport(*coefficients_, *kernelApproximationsImporter, INSERT);
       } else if (mode == Teuchos::TRANS) {
-        nearField_->localApply(X, *X_colmap_, mode, alpha, zero);
-        Y.scale (beta);
-        Y.beginExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
+        kernelApproximations_->localApply(*coefficients_, *coefficients_colmap_, mode, alpha);
+        coefficients2_->putScalar(zero);
+        coefficients2_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
+      }
+    } else {
+      if (mode == Teuchos::NO_TRANS) {
+        coefficients_colmap_->beginImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
+      } else if (mode == Teuchos::TRANS) {
+        kernelApproximations_->localApply(*coefficients2_, *coefficients_colmap_, mode, alpha);
+        coefficients_->putScalar(zero);
+        coefficients_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
       }
     }
+  }
 
-    // far field interactions - part 2
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 2")));
-
-      RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
-      if (flip) {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->endImport(*coefficients_, *kernelApproximationsImporter, INSERT);
-          kernelApproximations_->localApply(*coefficients_colmap_, *coefficients2_, mode, alpha);
-        } else if (mode == Teuchos::TRANS) {
-          coefficients2_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
-      } else {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->endImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
-          kernelApproximations_->localApply(*coefficients_colmap_, *coefficients_, mode, alpha);
-        } else if (mode == Teuchos::TRANS) {
-          coefficients_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
-      }
+  // near field - part 1
+  RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 1")));
+    if (mode == Teuchos::NO_TRANS) {
+      X_colmap_->beginImport(X, *nearFieldImporter, INSERT);
+    } else if (mode == Teuchos::TRANS) {
+      nearField_->localApply(X, *X_colmap_, mode, alpha, zero);
+      Y.scale(beta);
+      Y.beginExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
     }
+  }
 
-    // near field - part 2
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 2")));
+  // far field interactions - part 2
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 2")));
 
+    RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+    if (flip) {
       if (mode == Teuchos::NO_TRANS) {
-        X_colmap_->endImport(X, *nearFieldImporter, INSERT);
-        nearField_->localApply(*X_colmap_, Y, mode, alpha, beta);
+        coefficients_colmap_->endImport(*coefficients_, *kernelApproximationsImporter, INSERT);
+        kernelApproximations_->localApply(*coefficients_colmap_, *coefficients2_, mode, alpha);
       } else if (mode == Teuchos::TRANS) {
-        Y.endExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
+        coefficients2_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
+      }
+    } else {
+      if (mode == Teuchos::NO_TRANS) {
+        coefficients_colmap_->endImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
+        kernelApproximations_->localApply(*coefficients_colmap_, *coefficients_, mode, alpha);
+      } else if (mode == Teuchos::TRANS) {
+        coefficients_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
       }
     }
+  }
 
-    // downward pass
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("downward pass")));
+  // near field - part 2
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 2")));
 
-      for (size_t i = 0; i<transferMatrices_.size(); i++)
-        if (flip) {
-          coefficients_->assign(*coefficients2_);
-          transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::TRANS, one, one);
-          flip = false;
-        } else {
-          coefficients2_->assign(*coefficients_);
-          transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::TRANS, one, one);
-          flip = true;
-        }
-      if (flip)
-        basisMatrix_->apply(*coefficients2_, Y, Teuchos::NO_TRANS, one, one);
-      else
-        basisMatrix_->apply(*coefficients_, Y, Teuchos::NO_TRANS, one, one);
+    if (mode == Teuchos::NO_TRANS) {
+      X_colmap_->endImport(X, *nearFieldImporter, INSERT);
+      nearField_->localApply(*X_colmap_, Y, mode, alpha, beta);
+    } else if (mode == Teuchos::TRANS) {
+      Y.endExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
     }
   }
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  applyWithoutTransposes(const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                         Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                         Teuchos::ETransp mode,
-                         Scalar alpha,
-                         Scalar beta) const {
-    using Teuchos::RCP;
-    const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    bool flip = true;
-
-    allocateMemory(X.getNumVectors());
-
-    // upward pass
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("upward pass")));
-
-      basisMatrixT_->apply(X, *coefficients_, Teuchos::NO_TRANS);
+  // downward pass
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("downward pass")));
 
-      for (int i = Teuchos::as<int>(transferMatrices_.size())-1; i>=0; i--)
-        if (flip) {
-          coefficients2_->assign(*coefficients_);
-          transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
-          flip = false;
-        } else {
-          coefficients_->assign(*coefficients2_);
-          transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
-          flip = true;
-        }
-    }
-
-    // far field interactions - part 1
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 1")));
+    for (size_t i = 0; i < transferMatrices_.size(); i++)
+      if (flip) {
+        coefficients_->assign(*coefficients2_);
+        transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::TRANS, one, one);
+        flip = false;
+      } else {
+        coefficients2_->assign(*coefficients_);
+        transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::TRANS, one, one);
+        flip = true;
+      }
+    if (flip)
+      basisMatrix_->apply(*coefficients2_, Y, Teuchos::NO_TRANS, one, one);
+    else
+      basisMatrix_->apply(*coefficients_, Y, Teuchos::NO_TRANS, one, one);
+  }
+}
 
-      RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    applyWithoutTransposes(const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                           Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                           Teuchos::ETransp mode,
+                           Scalar alpha,
+                           Scalar beta) const {
+  using Teuchos::RCP;
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+  bool flip         = true;
+
+  allocateMemory(X.getNumVectors());
+
+  // upward pass
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("upward pass")));
+
+    basisMatrixT_->apply(X, *coefficients_, Teuchos::NO_TRANS);
+
+    for (int i = Teuchos::as<int>(transferMatrices_.size()) - 1; i >= 0; i--)
       if (flip) {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->beginImport(*coefficients_, *kernelApproximationsImporter, INSERT);
-        } else if (mode == Teuchos::TRANS) {
-          kernelApproximations_->localApply(*coefficients_, *coefficients_colmap_, mode, alpha);
-          coefficients2_->putScalar(zero);
-          coefficients2_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
+        coefficients2_->assign(*coefficients_);
+        transferMatrices_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
+        flip = false;
       } else {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->beginImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
-        } else if (mode == Teuchos::TRANS) {
-          kernelApproximations_->localApply(*coefficients2_, *coefficients_colmap_, mode, alpha);
-          coefficients_->putScalar(zero);
-          coefficients_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
+        coefficients_->assign(*coefficients2_);
+        transferMatrices_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
+        flip = true;
       }
-    }
+  }
 
-    // near field - part 1
-    RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 1")));
+  // far field interactions - part 1
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 1")));
+
+    RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+    if (flip) {
       if (mode == Teuchos::NO_TRANS) {
-        X_colmap_->beginImport(X, *nearFieldImporter, INSERT);
+        coefficients_colmap_->beginImport(*coefficients_, *kernelApproximationsImporter, INSERT);
       } else if (mode == Teuchos::TRANS) {
-        nearField_->localApply(X, *X_colmap_, mode, alpha, zero);
-        Y.scale (beta);
-        Y.beginExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
+        kernelApproximations_->localApply(*coefficients_, *coefficients_colmap_, mode, alpha);
+        coefficients2_->putScalar(zero);
+        coefficients2_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
+      }
+    } else {
+      if (mode == Teuchos::NO_TRANS) {
+        coefficients_colmap_->beginImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
+      } else if (mode == Teuchos::TRANS) {
+        kernelApproximations_->localApply(*coefficients2_, *coefficients_colmap_, mode, alpha);
+        coefficients_->putScalar(zero);
+        coefficients_->beginExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
       }
     }
+  }
 
-    // far field interactions - part 2
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 2")));
-
-      RCP<const Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
-      if (flip) {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->endImport(*coefficients_, *kernelApproximationsImporter, INSERT);
-          kernelApproximations_->localApply(*coefficients_colmap_, *coefficients2_, mode, alpha);
-        } else if (mode == Teuchos::TRANS) {
-          coefficients2_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
-      } else {
-        if (mode == Teuchos::NO_TRANS) {
-          coefficients_colmap_->endImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
-          kernelApproximations_->localApply(*coefficients_colmap_, *coefficients_, mode, alpha);
-        } else if (mode == Teuchos::TRANS) {
-          coefficients_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
-        }
-      }
+  // near field - part 1
+  RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > nearFieldImporter = nearField_->getGraph()->getImporter();
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 1")));
+    if (mode == Teuchos::NO_TRANS) {
+      X_colmap_->beginImport(X, *nearFieldImporter, INSERT);
+    } else if (mode == Teuchos::TRANS) {
+      nearField_->localApply(X, *X_colmap_, mode, alpha, zero);
+      Y.scale(beta);
+      Y.beginExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
     }
+  }
 
-    // near field - part 2
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 2")));
+  // far field interactions - part 2
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("far field 2")));
 
+    RCP<const Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > kernelApproximationsImporter = kernelApproximations_->pointA_->getGraph()->getImporter();
+    if (flip) {
       if (mode == Teuchos::NO_TRANS) {
-        X_colmap_->endImport(X, *nearFieldImporter, INSERT);
-        nearField_->localApply(*X_colmap_, Y, mode, alpha, beta);
+        coefficients_colmap_->endImport(*coefficients_, *kernelApproximationsImporter, INSERT);
+        kernelApproximations_->localApply(*coefficients_colmap_, *coefficients2_, mode, alpha);
       } else if (mode == Teuchos::TRANS) {
-        Y.endExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
+        coefficients2_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
+      }
+    } else {
+      if (mode == Teuchos::NO_TRANS) {
+        coefficients_colmap_->endImport(*coefficients2_, *kernelApproximationsImporter, INSERT);
+        kernelApproximations_->localApply(*coefficients_colmap_, *coefficients_, mode, alpha);
+      } else if (mode == Teuchos::TRANS) {
+        coefficients_->endExport(*coefficients_colmap_, *kernelApproximationsImporter, ADD_ASSIGN);
       }
     }
+  }
 
-    // downward pass
-    {
-      Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("downward pass")));
+  // near field - part 2
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("near field 2")));
 
-      for (size_t i = 0; i<transferMatrices_.size(); i++)
-        if (flip) {
-          coefficients_->assign(*coefficients2_);
-          transferMatricesT_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
-          flip = false;
-        } else {
-          coefficients2_->assign(*coefficients_);
-          transferMatricesT_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
-          flip = true;
-        }
-      if (flip)
-        basisMatrix_->apply(*coefficients2_, Y, Teuchos::NO_TRANS, one, one);
-      else
-        basisMatrix_->apply(*coefficients_, Y, Teuchos::NO_TRANS, one, one);
+    if (mode == Teuchos::NO_TRANS) {
+      X_colmap_->endImport(X, *nearFieldImporter, INSERT);
+      nearField_->localApply(*X_colmap_, Y, mode, alpha, beta);
+    } else if (mode == Teuchos::TRANS) {
+      Y.endExport(*X_colmap_, *nearFieldImporter, ADD_ASSIGN);
     }
   }
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  Teuchos::RCP<HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  restrict(const Teuchos::RCP<matrix_type>& P) {
-    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("Galerkin product")));
-
-    // H_c = P^T * H * P
-    using lo_vec_type = typename blocked_map_type::lo_vec_type;
-    using vec_type = typename Tpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    const Scalar HALF = ONE/(ONE+ONE);
-
-    // newBasisMatrix = P^T * basisMatrix
-    RCP<matrix_type> newBasisMatrix = rcp(new matrix_type(P->getDomainMap(), clusterCoeffMap_, 0));
-    MatrixMatrix::Multiply(*P, true, *basisMatrix_, false, *newBasisMatrix);
-
-    //
-    auto clusterSizes = kernelApproximations_->blockMap_->blockSizes_;
-    auto ghosted_clusterMap = kernelApproximations_->blockA_->getColMap();
-    auto ghosted_clusterSizes = kernelApproximations_->ghosted_blockMap_->blockSizes_;
-
-    // Get number of unknowns associated with each cluster via new basisMatrix.
-    // numUnknownsPerCluster = \prod transfer_k * graph(newBasisMatrix)^T * ones
-    RCP<vec_type> numUnknownsPerCluster;
-    RCP<vec_type> ghosted_numUnknownsPerCluster;
-    if ((coarseningCriterion_ == "equivalentDense") ||
-        (coarseningCriterion_ == "numClusters")) {
-      {
-        numUnknownsPerCluster = rcp(new vec_type(kernelApproximations_->blockA_->getRowMap(), false));
-        auto lcl_clusterSizes = clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
-        auto lcl_numUnknownsPerCluster = numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::OverwriteAll);
-        // Compute the transpose of the newBasisMatrix.
-        RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > newBasisMatrixT;
-        Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(newBasisMatrix);
-        RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
-        newBasisMatrixT = transposer.createTranspose(transposeParams);
-
-        // TODO: parallel_for
-        auto rowptr = newBasisMatrixT->getLocalRowPtrsHost();
-        LocalOrdinal clusterStart = 0;
-        LocalOrdinal clusterEnd = 0;
-        for (LocalOrdinal cluster = 0; cluster < lcl_clusterSizes.extent_int(0); ++cluster) {
-          clusterStart = clusterEnd;
-          clusterEnd += lcl_clusterSizes(cluster, 0);
-          LocalOrdinal maxEntries = 0;
-          for (LocalOrdinal row = clusterStart; row < clusterEnd; ++row) {
-            LocalOrdinal numEntriesPerRow = rowptr(row+1)-rowptr(row);
-            maxEntries = std::max(maxEntries, numEntriesPerRow);
-          }
-          lcl_numUnknownsPerCluster(cluster, 0) = maxEntries;
-        }
-        TEUCHOS_ASSERT_EQUALITY(clusterEnd+1, rowptr.extent_int(0));
-      }
+  // downward pass
+  {
+    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("downward pass")));
 
+    for (size_t i = 0; i < transferMatrices_.size(); i++)
+      if (flip) {
+        coefficients_->assign(*coefficients2_);
+        transferMatricesT_[i]->localApply(*coefficients2_, *coefficients_, Teuchos::NO_TRANS, one, one);
+        flip = false;
+      } else {
+        coefficients2_->assign(*coefficients_);
+        transferMatricesT_[i]->localApply(*coefficients_, *coefficients2_, Teuchos::NO_TRANS, one, one);
+        flip = true;
+      }
+    if (flip)
+      basisMatrix_->apply(*coefficients2_, Y, Teuchos::NO_TRANS, one, one);
+    else
+      basisMatrix_->apply(*coefficients_, Y, Teuchos::NO_TRANS, one, one);
+  }
+}
 
-      // sum from child nodes to parents via transfer operators
-      for (int i = Teuchos::as<int>(transferMatrices_.size())-1; i>=0; i--)
-        transferMatrices_[i]->blockA_->apply(*numUnknownsPerCluster, *numUnknownsPerCluster, Teuchos::NO_TRANS, ONE, ONE);
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+Teuchos::RCP<HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+    HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+        restrict(const Teuchos::RCP<matrix_type>& P) {
+  Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("Galerkin product")));
+
+  // H_c = P^T * H * P
+  using lo_vec_type   = typename blocked_map_type::lo_vec_type;
+  using vec_type      = typename Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  const Scalar ONE  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
+  const Scalar HALF = ONE / (ONE + ONE);
+
+  // newBasisMatrix = P^T * basisMatrix
+  RCP<matrix_type> newBasisMatrix = rcp(new matrix_type(P->getDomainMap(), clusterCoeffMap_, 0));
+  MatrixMatrix::Multiply(*P, true, *basisMatrix_, false, *newBasisMatrix);
+
+  //
+  auto clusterSizes         = kernelApproximations_->blockMap_->blockSizes_;
+  auto ghosted_clusterMap   = kernelApproximations_->blockA_->getColMap();
+  auto ghosted_clusterSizes = kernelApproximations_->ghosted_blockMap_->blockSizes_;
+
+  // Get number of unknowns associated with each cluster via new basisMatrix.
+  // numUnknownsPerCluster = \prod transfer_k * graph(newBasisMatrix)^T * ones
+  RCP<vec_type> numUnknownsPerCluster;
+  RCP<vec_type> ghosted_numUnknownsPerCluster;
+  if ((coarseningCriterion_ == "equivalentDense") ||
+      (coarseningCriterion_ == "numClusters")) {
+    {
+      numUnknownsPerCluster          = rcp(new vec_type(kernelApproximations_->blockA_->getRowMap(), false));
+      auto lcl_clusterSizes          = clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
+      auto lcl_numUnknownsPerCluster = numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::OverwriteAll);
+      // Compute the transpose of the newBasisMatrix.
+      RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > newBasisMatrixT;
+      Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(newBasisMatrix);
+      RCP<Teuchos::ParameterList> transposeParams = rcp(new Teuchos::ParameterList);
+      newBasisMatrixT                             = transposer.createTranspose(transposeParams);
 
-      // get ghosted numUnknownsPerCluster
-      ghosted_numUnknownsPerCluster = rcp(new vec_type(ghosted_clusterMap, false));
-      auto import = kernelApproximations_->blockA_->getCrsGraph()->getImporter();
-      ghosted_numUnknownsPerCluster->doImport(*numUnknownsPerCluster, *import, Tpetra::INSERT);
+      // TODO: parallel_for
+      auto rowptr               = newBasisMatrixT->getLocalRowPtrsHost();
+      LocalOrdinal clusterStart = 0;
+      LocalOrdinal clusterEnd   = 0;
+      for (LocalOrdinal cluster = 0; cluster < lcl_clusterSizes.extent_int(0); ++cluster) {
+        clusterStart = clusterEnd;
+        clusterEnd += lcl_clusterSizes(cluster, 0);
+        LocalOrdinal maxEntries = 0;
+        for (LocalOrdinal row = clusterStart; row < clusterEnd; ++row) {
+          LocalOrdinal numEntriesPerRow = rowptr(row + 1) - rowptr(row);
+          maxEntries                    = std::max(maxEntries, numEntriesPerRow);
+        }
+        lcl_numUnknownsPerCluster(cluster, 0) = maxEntries;
+      }
+      TEUCHOS_ASSERT_EQUALITY(clusterEnd + 1, rowptr.extent_int(0));
     }
 
-    // coarse cluster pair graph
-    RCP<matrix_type> newKernelBlockGraph = rcp(new matrix_type(kernelApproximations_->blockA_->getCrsGraph()));
-    newKernelBlockGraph->resumeFill();
-    // point entries of cluster pairs that should be moved to the near field
-    RCP<matrix_type> diffKernelApprox = rcp(new matrix_type(kernelApproximations_->pointA_->getCrsGraph()));
+    // sum from child nodes to parents via transfer operators
+    for (int i = Teuchos::as<int>(transferMatrices_.size()) - 1; i >= 0; i--)
+      transferMatrices_[i]->blockA_->apply(*numUnknownsPerCluster, *numUnknownsPerCluster, Teuchos::NO_TRANS, ONE, ONE);
 
-    // Determine which cluster pairs should be moved to the near field.
-    // We are constructing the coarse block matrix newKernelBlockGraph
-    // and the point matrix diffKernelApprox.
-    {
-      typename vec_type::dual_view_type::t_host::const_type lcl_numUnknownsPerCluster;
-      typename vec_type::dual_view_type::t_host::const_type lcl_ghosted_numUnknownsPerCluster;
-      auto lcl_offsets = Kokkos::create_mirror_view(kernelApproximations_->blockMap_->offsets_);
-      auto lcl_ghosted_offsets = Kokkos::create_mirror_view(kernelApproximations_->ghosted_blockMap_->offsets_);
-      Kokkos::deep_copy(lcl_offsets, kernelApproximations_->blockMap_->offsets_);
-      Kokkos::deep_copy(lcl_ghosted_offsets, kernelApproximations_->ghosted_blockMap_->offsets_);
-      auto lcl_clusterSizes = clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
-      auto lcl_ghosted_clusterSizes = ghosted_clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
-
-      if ((coarseningCriterion_ == "equivalentDense") ||
+    // get ghosted numUnknownsPerCluster
+    ghosted_numUnknownsPerCluster = rcp(new vec_type(ghosted_clusterMap, false));
+    auto import                   = kernelApproximations_->blockA_->getCrsGraph()->getImporter();
+    ghosted_numUnknownsPerCluster->doImport(*numUnknownsPerCluster, *import, Tpetra::INSERT);
+  }
+
+  // coarse cluster pair graph
+  RCP<matrix_type> newKernelBlockGraph = rcp(new matrix_type(kernelApproximations_->blockA_->getCrsGraph()));
+  newKernelBlockGraph->resumeFill();
+  // point entries of cluster pairs that should be moved to the near field
+  RCP<matrix_type> diffKernelApprox = rcp(new matrix_type(kernelApproximations_->pointA_->getCrsGraph()));
+
+  // Determine which cluster pairs should be moved to the near field.
+  // We are constructing the coarse block matrix newKernelBlockGraph
+  // and the point matrix diffKernelApprox.
+  {
+    typename vec_type::dual_view_type::t_host::const_type lcl_numUnknownsPerCluster;
+    typename vec_type::dual_view_type::t_host::const_type lcl_ghosted_numUnknownsPerCluster;
+    auto lcl_offsets         = Kokkos::create_mirror_view(kernelApproximations_->blockMap_->offsets_);
+    auto lcl_ghosted_offsets = Kokkos::create_mirror_view(kernelApproximations_->ghosted_blockMap_->offsets_);
+    Kokkos::deep_copy(lcl_offsets, kernelApproximations_->blockMap_->offsets_);
+    Kokkos::deep_copy(lcl_ghosted_offsets, kernelApproximations_->ghosted_blockMap_->offsets_);
+    auto lcl_clusterSizes         = clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
+    auto lcl_ghosted_clusterSizes = ghosted_clusterSizes->getLocalViewHost(Tpetra::Access::ReadOnly);
+
+    if ((coarseningCriterion_ == "equivalentDense") ||
         (coarseningCriterion_ == "numClusters")) {
-        lcl_numUnknownsPerCluster = numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::ReadOnly);
-        lcl_ghosted_numUnknownsPerCluster = ghosted_numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::ReadOnly);
-      }
+      lcl_numUnknownsPerCluster         = numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::ReadOnly);
+      lcl_ghosted_numUnknownsPerCluster = ghosted_numUnknownsPerCluster->getLocalViewHost(Tpetra::Access::ReadOnly);
+    }
 
-      // Criterion: "numClusters"
-      // Compute all cluster pair sizes, sort them, and pick cut-off
-      // so that the number of cluster pairs decreases propotionally
-      // to the number of unknowns.
-      size_t tgt_clusterPairSize = 0;
-      if (coarseningCriterion_ == "numClusters") {
-        auto lcl_BlockGraph = kernelApproximations_->blockA_->getLocalMatrixHost();
-        std::vector<size_t> clusterPairSizes;
-        for (LocalOrdinal brlid = 0; brlid < lcl_BlockGraph.numRows(); ++brlid) {
-          auto brow = lcl_BlockGraph.row(brlid);
-          for (LocalOrdinal k = 0; k < brow.length; ++k) {
-            // Entries of the block matrix for kernelApproximations
-            // decide whether the cluster pair is present and only take
-            // values 1 or 0.
-            if (brow.value(k) > HALF) {
-              LocalOrdinal bclid = brow.colidx(k);
-              clusterPairSizes.push_back(lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0));
-            }
+    // Criterion: "numClusters"
+    // Compute all cluster pair sizes, sort them, and pick cut-off
+    // so that the number of cluster pairs decreases propotionally
+    // to the number of unknowns.
+    size_t tgt_clusterPairSize = 0;
+    if (coarseningCriterion_ == "numClusters") {
+      auto lcl_BlockGraph = kernelApproximations_->blockA_->getLocalMatrixHost();
+      std::vector<size_t> clusterPairSizes;
+      for (LocalOrdinal brlid = 0; brlid < lcl_BlockGraph.numRows(); ++brlid) {
+        auto brow = lcl_BlockGraph.row(brlid);
+        for (LocalOrdinal k = 0; k < brow.length; ++k) {
+          // Entries of the block matrix for kernelApproximations
+          // decide whether the cluster pair is present and only take
+          // values 1 or 0.
+          if (brow.value(k) > HALF) {
+            LocalOrdinal bclid = brow.colidx(k);
+            clusterPairSizes.push_back(lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0));
           }
         }
-        std::sort(clusterPairSizes.begin(), clusterPairSizes.end());
-        double coarseningRate = Teuchos::as<double>(P->getGlobalNumCols())/Teuchos::as<double>(P->getGlobalNumRows());
-        tgt_clusterPairSize = clusterPairSizes[Teuchos::as<size_t>(clusterPairSizes.size()*(1-coarseningRate))];
-        // std::cout << "HERE " << clusterPairSizes[0] << " " << tgt_clusterPairSize << " " << clusterPairSizes[clusterPairSizes.size()-1] << std::endl;
       }
+      std::sort(clusterPairSizes.begin(), clusterPairSizes.end());
+      double coarseningRate = Teuchos::as<double>(P->getGlobalNumCols()) / Teuchos::as<double>(P->getGlobalNumRows());
+      tgt_clusterPairSize   = clusterPairSizes[Teuchos::as<size_t>(clusterPairSizes.size() * (1 - coarseningRate))];
+      // std::cout << "HERE " << clusterPairSizes[0] << " " << tgt_clusterPairSize << " " << clusterPairSizes[clusterPairSizes.size()-1] << std::endl;
+    }
 
-      // Criterion: "transferLevels"
-      // Drop cluster pairs by level in the tree.
-      auto comm = getComm();
-      std::set<LocalOrdinal> blidsToDrop;
-      if (coarseningCriterion_ == "transferLevels") {
-        double coarseningRate = Teuchos::as<double>(P->getGlobalNumCols())/Teuchos::as<double>(P->getGlobalNumRows());
-        size_t droppedClusterPairs = 0;
-        size_t totalNumClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
-        RCP<vec_type> tempV = Teuchos::rcp(new vec_type(kernelApproximations_->blockMap_->blockMap_, false));
-        RCP<vec_type> tempV2 = Teuchos::rcp(new vec_type(kernelApproximations_->blockMap_->blockMap_, false));
-        int keepTransfers = params_->get<int>("keepTransfers",-1);
-        if (keepTransfers == -1) {
-          double leftOverFactor = params_->get<double>("leftOverFactor");
-          keepTransfers = transferMatrices_.size();
-          double temp = (1.0 / coarseningRate) * leftOverFactor;
-          const double treeCoarseningFactor = params_->get<double>("treeCoarseningFactor");
-          while (temp >= 2.0) {
-            --keepTransfers;
-            temp /= treeCoarseningFactor;
-          }
-          keepTransfers = std::max(keepTransfers, 0);
-          params_->set("leftOverFactor", temp);
+    // Criterion: "transferLevels"
+    // Drop cluster pairs by level in the tree.
+    auto comm = getComm();
+    std::set<LocalOrdinal> blidsToDrop;
+    if (coarseningCriterion_ == "transferLevels") {
+      double coarseningRate       = Teuchos::as<double>(P->getGlobalNumCols()) / Teuchos::as<double>(P->getGlobalNumRows());
+      size_t droppedClusterPairs  = 0;
+      size_t totalNumClusterPairs = kernelApproximations_->blockA_->getGlobalNumEntries();
+      RCP<vec_type> tempV         = Teuchos::rcp(new vec_type(kernelApproximations_->blockMap_->blockMap_, false));
+      RCP<vec_type> tempV2        = Teuchos::rcp(new vec_type(kernelApproximations_->blockMap_->blockMap_, false));
+      int keepTransfers           = params_->get<int>("keepTransfers", -1);
+      if (keepTransfers == -1) {
+        double leftOverFactor             = params_->get<double>("leftOverFactor");
+        keepTransfers                     = transferMatrices_.size();
+        double temp                       = (1.0 / coarseningRate) * leftOverFactor;
+        const double treeCoarseningFactor = params_->get<double>("treeCoarseningFactor");
+        while (temp >= 2.0) {
+          --keepTransfers;
+          temp /= treeCoarseningFactor;
         }
+        keepTransfers = std::max(keepTransfers, 0);
+        params_->set("leftOverFactor", temp);
+      }
 
-        for (int k = Teuchos::as<int>(transferMatrices_.size())-1; k>=0; --k) {
-
-          size_t clustersInLevel = transferMatrices_[k]->blockA_->getGlobalNumEntries();
+      for (int k = Teuchos::as<int>(transferMatrices_.size()) - 1; k >= 0; --k) {
+        size_t clustersInLevel = transferMatrices_[k]->blockA_->getGlobalNumEntries();
 
-          if (debugOutput_ && (comm->getRank() == 0))
-            std::cout << "level " << k << " clustersInLevel " << clustersInLevel << std::endl;
+        if (debugOutput_ && (comm->getRank() == 0))
+          std::cout << "level " << k << " clustersInLevel " << clustersInLevel << std::endl;
 
-          tempV->putScalar(ONE);
-          transferMatrices_[k]->blockA_->apply(*tempV, *tempV2, Teuchos::TRANS);
+        tempV->putScalar(ONE);
+        transferMatrices_[k]->blockA_->apply(*tempV, *tempV2, Teuchos::TRANS);
 
-          size_t numClusters = tempV2->norm1();
-          if (debugOutput_ && (comm->getRank() == 0))
-            std::cout << "numClusters " << numClusters << std::endl;
-          tempV->putScalar(ZERO);
-          kernelApproximations_->blockA_->apply(*tempV2, *tempV);
+        size_t numClusters = tempV2->norm1();
+        if (debugOutput_ && (comm->getRank() == 0))
+          std::cout << "numClusters " << numClusters << std::endl;
+        tempV->putScalar(ZERO);
+        kernelApproximations_->blockA_->apply(*tempV2, *tempV);
 
-          Scalar numClusterPairs = tempV->dot(*tempV2);
-          if (debugOutput_ && (comm->getRank() == 0))
-            std::cout << "numClusterPairs " << numClusterPairs << std::endl;
+        Scalar numClusterPairs = tempV->dot(*tempV2);
+        if (debugOutput_ && (comm->getRank() == 0))
+          std::cout << "numClusterPairs " << numClusterPairs << std::endl;
 
-          bool doDrop;
-          if (keepTransfers >= 0) {
-            doDrop = (keepTransfers<=k);
-          } else {
-            doDrop = (droppedClusterPairs + numClusterPairs < (1.0-coarseningRate) * totalNumClusterPairs);
-          }
-          if (doDrop) {
-            auto lcl_transfer = transferMatrices_[k]->blockA_->getLocalMatrixHost();
-            auto lcl_transfer_graph = lcl_transfer.graph;
-            for (LocalOrdinal j = 0; j < lcl_transfer_graph.entries.extent_int(0); j++)
-              blidsToDrop.insert(lcl_transfer_graph.entries(j));
+        bool doDrop;
+        if (keepTransfers >= 0) {
+          doDrop = (keepTransfers <= k);
+        } else {
+          doDrop = (droppedClusterPairs + numClusterPairs < (1.0 - coarseningRate) * totalNumClusterPairs);
+        }
+        if (doDrop) {
+          auto lcl_transfer       = transferMatrices_[k]->blockA_->getLocalMatrixHost();
+          auto lcl_transfer_graph = lcl_transfer.graph;
+          for (LocalOrdinal j = 0; j < lcl_transfer_graph.entries.extent_int(0); j++)
+            blidsToDrop.insert(lcl_transfer_graph.entries(j));
 
-            droppedClusterPairs += numClusterPairs;
-          } else {
-            if (debugOutput_ && (comm->getRank() == 0))
-              std::cout << "Dropped " << transferMatrices_.size()-1-k << " transfers of " << transferMatrices_.size() << " dropped cp: " << droppedClusterPairs <<std::endl;
-            break;
-          }
+          droppedClusterPairs += numClusterPairs;
+        } else {
+          if (debugOutput_ && (comm->getRank() == 0))
+            std::cout << "Dropped " << transferMatrices_.size() - 1 - k << " transfers of " << transferMatrices_.size() << " dropped cp: " << droppedClusterPairs << std::endl;
+          break;
         }
       }
+    }
 
-      // number of cluster pairs dropped
-      int dropped = 0;
-      // number of cluster pairs we kept
-      int kept = 0;
-      // number of cluster pairs that were no longer present
-      int ignored = 0;
-      // loop over cluster pairs
-      // TODO: parallel_for
-      auto lcl_BlockGraph = kernelApproximations_->blockA_->getLocalMatrixHost();
-      auto lcl_newBlockGraph = newKernelBlockGraph->getLocalMatrixHost();
-      auto lcl_KernelApprox = kernelApproximations_->pointA_->getLocalMatrixHost();
-      auto lcl_diffKernelApprox = diffKernelApprox->getLocalMatrixHost();
-      for (LocalOrdinal brlid = 0; brlid < lcl_BlockGraph.numRows(); ++brlid) {
-        size_t brsize = lcl_clusterSizes(brlid, 0);
-        auto brow = lcl_BlockGraph.row(brlid);
-        auto new_brow = lcl_newBlockGraph.row(brlid);
-        for (LocalOrdinal k = 0; k < brow.length; ++k) {
-          // Entries of the block matrix for kernelApproximations
-          // decide whether the cluster pair is present and only take
-          // values 1 or 0.
-          if (brow.value(k) > HALF) {
-            LocalOrdinal bclid = brow.colidx(k);
-            size_t bcsize = lcl_ghosted_clusterSizes(bclid, 0);
-
-            // criterium for removing a cluster pair from the far field
-            bool removeCluster = false;
-            if (coarseningCriterion_ == "equivalentDense") {
-              // Size of the sparse cluster approximation >= size of dense equivalent
-              removeCluster = (brsize * bcsize
-                               >= lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0));
-            } else if (coarseningCriterion_ == "numClusters") {
-              removeCluster = (lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0) < tgt_clusterPairSize);
-            } else if (coarseningCriterion_ == "transferLevels") {
-              removeCluster = ((blidsToDrop.find(brlid) != blidsToDrop.end()) ||
-                               (blidsToDrop.find(bclid) != blidsToDrop.end()));
-            }
-            if (removeCluster) {
-              // we are dropping the cluster pair from the far field
-              ++dropped;
-              new_brow.value(k) = ZERO;
-
-              // loop over the point matrix and add the entries to diffKernelApprox
-              const LocalOrdinal row_start = lcl_offsets(brlid);
-              const LocalOrdinal row_end = lcl_offsets(brlid+1);
-              const LocalOrdinal col_start = lcl_ghosted_offsets(bclid);
-              const LocalOrdinal col_end = lcl_ghosted_offsets(bclid+1);
-              TEUCHOS_ASSERT_EQUALITY(Teuchos::as<size_t>(row_end-row_start), brsize);
-              TEUCHOS_ASSERT_EQUALITY(Teuchos::as<size_t>(col_end-col_start), bcsize);
-              for (LocalOrdinal rlid = row_start; rlid <  row_end; ++rlid) {
-                auto diff_row = lcl_diffKernelApprox.row(rlid);
-                auto row = lcl_KernelApprox.row(rlid);
-                size_t removed = 0;
-                for (LocalOrdinal n = 0; n < row.length; ++n) {
-                  if ((col_start <= row.colidx(n)) && (col_end > row.colidx(n))) {
-                    diff_row.value(n) = row.value(n);
-                    ++removed;
-                  }
+    // number of cluster pairs dropped
+    int dropped = 0;
+    // number of cluster pairs we kept
+    int kept = 0;
+    // number of cluster pairs that were no longer present
+    int ignored = 0;
+    // loop over cluster pairs
+    // TODO: parallel_for
+    auto lcl_BlockGraph       = kernelApproximations_->blockA_->getLocalMatrixHost();
+    auto lcl_newBlockGraph    = newKernelBlockGraph->getLocalMatrixHost();
+    auto lcl_KernelApprox     = kernelApproximations_->pointA_->getLocalMatrixHost();
+    auto lcl_diffKernelApprox = diffKernelApprox->getLocalMatrixHost();
+    for (LocalOrdinal brlid = 0; brlid < lcl_BlockGraph.numRows(); ++brlid) {
+      size_t brsize = lcl_clusterSizes(brlid, 0);
+      auto brow     = lcl_BlockGraph.row(brlid);
+      auto new_brow = lcl_newBlockGraph.row(brlid);
+      for (LocalOrdinal k = 0; k < brow.length; ++k) {
+        // Entries of the block matrix for kernelApproximations
+        // decide whether the cluster pair is present and only take
+        // values 1 or 0.
+        if (brow.value(k) > HALF) {
+          LocalOrdinal bclid = brow.colidx(k);
+          size_t bcsize      = lcl_ghosted_clusterSizes(bclid, 0);
+
+          // criterium for removing a cluster pair from the far field
+          bool removeCluster = false;
+          if (coarseningCriterion_ == "equivalentDense") {
+            // Size of the sparse cluster approximation >= size of dense equivalent
+            removeCluster = (brsize * bcsize >= lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0));
+          } else if (coarseningCriterion_ == "numClusters") {
+            removeCluster = (lcl_numUnknownsPerCluster(brlid, 0) * lcl_ghosted_numUnknownsPerCluster(bclid, 0) < tgt_clusterPairSize);
+          } else if (coarseningCriterion_ == "transferLevels") {
+            removeCluster = ((blidsToDrop.find(brlid) != blidsToDrop.end()) ||
+                             (blidsToDrop.find(bclid) != blidsToDrop.end()));
+          }
+          if (removeCluster) {
+            // we are dropping the cluster pair from the far field
+            ++dropped;
+            new_brow.value(k) = ZERO;
+
+            // loop over the point matrix and add the entries to diffKernelApprox
+            const LocalOrdinal row_start = lcl_offsets(brlid);
+            const LocalOrdinal row_end   = lcl_offsets(brlid + 1);
+            const LocalOrdinal col_start = lcl_ghosted_offsets(bclid);
+            const LocalOrdinal col_end   = lcl_ghosted_offsets(bclid + 1);
+            TEUCHOS_ASSERT_EQUALITY(Teuchos::as<size_t>(row_end - row_start), brsize);
+            TEUCHOS_ASSERT_EQUALITY(Teuchos::as<size_t>(col_end - col_start), bcsize);
+            for (LocalOrdinal rlid = row_start; rlid < row_end; ++rlid) {
+              auto diff_row  = lcl_diffKernelApprox.row(rlid);
+              auto row       = lcl_KernelApprox.row(rlid);
+              size_t removed = 0;
+              for (LocalOrdinal n = 0; n < row.length; ++n) {
+                if ((col_start <= row.colidx(n)) && (col_end > row.colidx(n))) {
+                  diff_row.value(n) = row.value(n);
+                  ++removed;
                 }
-                if (removed != bcsize) {
-                  std::ostringstream oss;
-                  oss << "brlid " << brlid << " row " << rlid << std::endl;
-                  oss << "col_start "<< col_start << " col_end " << col_end << std::endl;
-                  for (LocalOrdinal n = 0; n < row.length; ++n) {
-                    oss << row.colidx(n) << " " << row.value(n) << std::endl;
-                  }
-                  std::cout << oss.str();
+              }
+              if (removed != bcsize) {
+                std::ostringstream oss;
+                oss << "brlid " << brlid << " row " << rlid << std::endl;
+                oss << "col_start " << col_start << " col_end " << col_end << std::endl;
+                for (LocalOrdinal n = 0; n < row.length; ++n) {
+                  oss << row.colidx(n) << " " << row.value(n) << std::endl;
                 }
-                TEUCHOS_ASSERT_EQUALITY(removed, bcsize);
+                std::cout << oss.str();
               }
-            } else {
-              // We are keeping the cluster pair.
-              ++kept;
-              new_brow.value(k) = brow.value(k);
+              TEUCHOS_ASSERT_EQUALITY(removed, bcsize);
             }
           } else {
-            // The cluster pair has already been dropped on the fine level.
-            ++ignored;
+            // We are keeping the cluster pair.
+            ++kept;
             new_brow.value(k) = brow.value(k);
           }
+        } else {
+          // The cluster pair has already been dropped on the fine level.
+          ++ignored;
+          new_brow.value(k) = brow.value(k);
         }
       }
-      if (debugOutput_) {
-        // number of cluster pairs dropped
-        int gbl_dropped = 0;
-        // number of cluster pairs we kept
-        int gbl_kept = 0;
-        // number of cluster pairs that were no longer present
-        int gbl_ignored = 0;
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &dropped, &gbl_dropped);
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &kept, &gbl_kept);
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &ignored, &gbl_ignored);
-        if (comm->getRank() == 0)
-          std::cout << "dropped " << gbl_dropped << " kept " << gbl_kept << " ignored " << gbl_ignored << std::endl;
-      }
     }
+    if (debugOutput_) {
+      // number of cluster pairs dropped
+      int gbl_dropped = 0;
+      // number of cluster pairs we kept
+      int gbl_kept = 0;
+      // number of cluster pairs that were no longer present
+      int gbl_ignored = 0;
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &dropped, &gbl_dropped);
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &kept, &gbl_kept);
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &ignored, &gbl_ignored);
+      if (comm->getRank() == 0)
+        std::cout << "dropped " << gbl_dropped << " kept " << gbl_kept << " ignored " << gbl_ignored << std::endl;
+    }
+  }
 
-    newKernelBlockGraph->fillComplete(kernelApproximations_->blockA_->getDomainMap(),
-                                      kernelApproximations_->blockA_->getRangeMap());
-    newKernelBlockGraph = removeSmallEntries(newKernelBlockGraph, Teuchos::ScalarTraits<MagnitudeType>::eps());
-    diffKernelApprox->fillComplete(clusterCoeffMap_,
-                                   clusterCoeffMap_);
+  newKernelBlockGraph->fillComplete(kernelApproximations_->blockA_->getDomainMap(),
+                                    kernelApproximations_->blockA_->getRangeMap());
+  newKernelBlockGraph = removeSmallEntries(newKernelBlockGraph, Teuchos::ScalarTraits<MagnitudeType>::eps());
+  diffKernelApprox->fillComplete(clusterCoeffMap_,
+                                 clusterCoeffMap_);
+
+  // coarse point matrix of cluster pairs
+  Teuchos::RCP<matrix_type> newKernelApprox;
+  {
+    Teuchos::RCP<matrix_type> temp = MatrixMatrix::add(ONE, false, *kernelApproximations_->pointA_, -ONE, false, *diffKernelApprox);
+    newKernelApprox                = removeSmallEntries(temp, Teuchos::ScalarTraits<MagnitudeType>::eps());
+  }
 
-    // coarse point matrix of cluster pairs
-    Teuchos::RCP<matrix_type> newKernelApprox;
-    {
-      Teuchos::RCP<matrix_type> temp = MatrixMatrix::add(ONE, false, *kernelApproximations_->pointA_, -ONE, false, *diffKernelApprox);
-      newKernelApprox = removeSmallEntries(temp, Teuchos::ScalarTraits<MagnitudeType>::eps());
+  // construct identity on clusterCoeffMap_
+  Teuchos::RCP<matrix_type> identity = buildIdentityMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(clusterCoeffMap_);
+
+  Teuchos::RCP<blocked_matrix_type> newBlockedKernelApproximation = rcp(new blocked_matrix_type(newKernelApprox, newKernelBlockGraph, kernelApproximations_->blockMap_, kernelApproximations_->ghosted_blockMap_));
+
+  // select subset of transfer matrices for coarse operator
+  std::vector<Teuchos::RCP<blocked_matrix_type> > newTransferMatrices;
+  {
+    auto comm = getComm();
+
+    RCP<vec_type> v_temp          = rcp(new vec_type(newKernelBlockGraph->getDomainMap()));
+    RCP<vec_type> clusterUseCount = rcp(new vec_type(newKernelBlockGraph->getDomainMap()));
+    v_temp->putScalar(ONE);
+    clusterUseCount->putScalar(ZERO);
+    newKernelBlockGraph->apply(*v_temp, *clusterUseCount, Teuchos::NO_TRANS);
+    newKernelBlockGraph->apply(*v_temp, *clusterUseCount, Teuchos::TRANS, ONE, ONE);
+
+    for (int i = Teuchos::as<int>(transferMatrices_.size()) - 1; i >= 0; i--) {
+      // We drop a transfer operator T_i when
+      //  sum(T_i * clusterUseCount) == 0
+      // Since we need to use Scalar, we instead check for < 0.5
+      transferMatrices_[i]->blockA_->localApply(*clusterUseCount, *v_temp, Teuchos::NO_TRANS);
+      Scalar gbl_use_count = v_temp->norm1();
+      // if (comm->getRank() == 0)
+      //   std::cout << "Transfer " << i << " count " << gbl_use_count << std::endl;
+
+      if (gbl_use_count < HALF) {
+        // We do not keep the i-th transfer for the coarse operator.
+        // newBasisMatrix := newBasisMatrix * (I+transferMatrices_[i])^T
+        Teuchos::RCP<matrix_type> temp2 = MatrixMatrix::add(ONE, false, *identity, ONE, false, *transferMatrices_[i]->pointA_);
+        RCP<matrix_type> temp           = rcp(new matrix_type(newBasisMatrix->getRowMap(), clusterCoeffMap_, 0));
+        MatrixMatrix::Multiply(*newBasisMatrix, false, *temp2, true, *temp);
+        newBasisMatrix = temp;
+      } else {
+        // We keep the i-th transfer for the coarse operator.
+        newTransferMatrices.insert(newTransferMatrices.begin(), transferMatrices_[i]);
+      }
     }
+  }
 
-    // construct identity on clusterCoeffMap_
-    Teuchos::RCP<matrix_type> identity = buildIdentityMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(clusterCoeffMap_);
-
-    Teuchos::RCP<blocked_matrix_type> newBlockedKernelApproximation = rcp(new blocked_matrix_type(newKernelApprox, newKernelBlockGraph, kernelApproximations_->blockMap_, kernelApproximations_->ghosted_blockMap_));
+  // Coarse near field
+  RCP<matrix_type> newNearField;
+  {
+    // transfer = newBasisMatrix * (identity + newTransferMatrices[K-1]^T) * ... * (identity + newTransferMatrices[0])^T
+    Teuchos::RCP<matrix_type> transfer = rcp(new matrix_type(*newBasisMatrix));
+    for (int i = Teuchos::as<int>(newTransferMatrices.size()) - 1; i >= 0; i--) {
+      Teuchos::RCP<matrix_type> temp  = MatrixMatrix::add(ONE, false, *identity, ONE, false, *newTransferMatrices[i]->pointA_);
+      Teuchos::RCP<matrix_type> temp2 = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
+      MatrixMatrix::Multiply(*transfer, false, *temp, true, *temp2);
+      transfer = temp2;
+    }
 
-    // select subset of transfer matrices for coarse operator
-    std::vector<Teuchos::RCP<blocked_matrix_type> > newTransferMatrices;
+    // diffFarField = transfer * diffKernelApprox * transfer^T
+    RCP<matrix_type> diffFarField;
     {
-      auto comm = getComm();
-
-      RCP<vec_type> v_temp         = rcp(new vec_type(newKernelBlockGraph->getDomainMap()));
-      RCP<vec_type> clusterUseCount = rcp(new vec_type(newKernelBlockGraph->getDomainMap()));
-      v_temp->putScalar(ONE);
-      clusterUseCount->putScalar(ZERO);
-      newKernelBlockGraph->apply(*v_temp, *clusterUseCount, Teuchos::NO_TRANS);
-      newKernelBlockGraph->apply(*v_temp, *clusterUseCount, Teuchos::TRANS, ONE, ONE);
-
-      for (int i = Teuchos::as<int>(transferMatrices_.size())-1; i>=0; i--) {
-        // We drop a transfer operator T_i when
-        //  sum(T_i * clusterUseCount) == 0
-        // Since we need to use Scalar, we instead check for < 0.5
-        transferMatrices_[i]->blockA_->localApply(*clusterUseCount, *v_temp, Teuchos::NO_TRANS);
-        Scalar gbl_use_count = v_temp->norm1();
-        // if (comm->getRank() == 0)
-        //   std::cout << "Transfer " << i << " count " << gbl_use_count << std::endl;
-
-        if (gbl_use_count < HALF) {
-          // We do not keep the i-th transfer for the coarse operator.
-          // newBasisMatrix := newBasisMatrix * (I+transferMatrices_[i])^T
-          Teuchos::RCP<matrix_type> temp2 = MatrixMatrix::add(ONE, false, *identity, ONE, false, *transferMatrices_[i]->pointA_);
-          RCP<matrix_type> temp = rcp(new matrix_type(newBasisMatrix->getRowMap(), clusterCoeffMap_, 0));
-          MatrixMatrix::Multiply(*newBasisMatrix, false, *temp2, true, *temp);
-          newBasisMatrix = temp;
-        } else {
-          // We keep the i-th transfer for the coarse operator.
-          newTransferMatrices.insert(newTransferMatrices.begin(), transferMatrices_[i]);
-        }
-      }
+      Teuchos::RCP<matrix_type> temp = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
+      MatrixMatrix::Multiply(*transfer, false, *diffKernelApprox, false, *temp);
+      diffFarField = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
+      MatrixMatrix::Multiply(*temp, false, *transfer, true, *diffFarField);
     }
 
-    // Coarse near field
-    RCP<matrix_type> newNearField;
+    // newNearField = P^T * nearField * P + diffFarField
     {
-      // transfer = newBasisMatrix * (identity + newTransferMatrices[K-1]^T) * ... * (identity + newTransferMatrices[0])^T
-      Teuchos::RCP<matrix_type> transfer = rcp(new matrix_type(*newBasisMatrix));
-      for (int i = Teuchos::as<int>(newTransferMatrices.size())-1; i>=0; i--) {
-        Teuchos::RCP<matrix_type> temp = MatrixMatrix::add(ONE, false, *identity, ONE, false, *newTransferMatrices[i]->pointA_);
-        Teuchos::RCP<matrix_type> temp2 = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
-        MatrixMatrix::Multiply(*transfer, false, *temp, true, *temp2);
-        transfer = temp2;
-      }
-
-      // diffFarField = transfer * diffKernelApprox * transfer^T
-      RCP<matrix_type> diffFarField;
-      {
-        Teuchos::RCP<matrix_type> temp = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
-        MatrixMatrix::Multiply(*transfer, false, *diffKernelApprox, false, *temp);
-        diffFarField = rcp(new matrix_type(newBasisMatrix->getRowMap(), 0));
-        MatrixMatrix::Multiply(*temp, false, *transfer, true, *diffFarField);
-      }
-
-      // newNearField = P^T * nearField * P + diffFarField
-      {
-        RCP<matrix_type> temp = rcp(new matrix_type(nearField_->getRowMap(), 0));
-        MatrixMatrix::Multiply(*nearField_, false, *P, false, *temp);
-        RCP<matrix_type> temp2 = rcp(new matrix_type(P->getDomainMap(), 0));
-        MatrixMatrix::Multiply(*P, true, *temp, false, *temp2);
-        newNearField = MatrixMatrix::add(ONE, false, *temp2, ONE, false, *diffFarField);
-        newNearField = removeSmallEntries(newNearField, Teuchos::ScalarTraits<MagnitudeType>::eps());
-      }
+      RCP<matrix_type> temp = rcp(new matrix_type(nearField_->getRowMap(), 0));
+      MatrixMatrix::Multiply(*nearField_, false, *P, false, *temp);
+      RCP<matrix_type> temp2 = rcp(new matrix_type(P->getDomainMap(), 0));
+      MatrixMatrix::Multiply(*P, true, *temp, false, *temp2);
+      newNearField = MatrixMatrix::add(ONE, false, *temp2, ONE, false, *diffFarField);
+      newNearField = removeSmallEntries(newNearField, Teuchos::ScalarTraits<MagnitudeType>::eps());
     }
-
-    return Teuchos::rcp(new HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>(newNearField,
-                                                                                         newBlockedKernelApproximation,
-                                                                                         newBasisMatrix,
-                                                                                         newTransferMatrices,
-                                                                                         params_));
   }
 
+  return Teuchos::rcp(new HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>(newNearField,
+                                                                                          newBlockedKernelApproximation,
+                                                                                          newBasisMatrix,
+                                                                                          newTransferMatrices,
+                                                                                          params_));
+}
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  toMatrix() {
-    Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("Conversion from H-matrix to CSR")));
-
-    using Teuchos::RCP;
-    using Teuchos::rcp;
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    toMatrix() {
+  Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("Conversion from H-matrix to CSR")));
 
-    const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
+  using Teuchos::RCP;
+  using Teuchos::rcp;
 
-    if (hasFarField()) {
+  const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
 
-      // transfer = basisMatrix_ * (identity + transferMatrices_[K-1]) * ... * (identity + transferMatrices_[0])
-      RCP<matrix_type> transfer = rcp(new matrix_type(*basisMatrix_));
+  if (hasFarField()) {
+    // transfer = basisMatrix_ * (identity + transferMatrices_[K-1]) * ... * (identity + transferMatrices_[0])
+    RCP<matrix_type> transfer = rcp(new matrix_type(*basisMatrix_));
 
-      if (hasTransferMatrices()) {
-        // construct identity on clusterCoeffMap_
-        Teuchos::RCP<matrix_type> identity = buildIdentityMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(clusterCoeffMap_);
+    if (hasTransferMatrices()) {
+      // construct identity on clusterCoeffMap_
+      Teuchos::RCP<matrix_type> identity = buildIdentityMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(clusterCoeffMap_);
 
-        for (int i = Teuchos::as<int>(transferMatrices_.size())-1; i>=0; i--) {
-          RCP<matrix_type> temp = MatrixMatrix::add(ONE, false, *identity, ONE, false, *transferMatrices_[i]->pointA_);
-          RCP<matrix_type> temp2 = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
-          MatrixMatrix::Multiply(*transfer, false, *temp, true, *temp2);
-          transfer = temp2;
-        }
+      for (int i = Teuchos::as<int>(transferMatrices_.size()) - 1; i >= 0; i--) {
+        RCP<matrix_type> temp  = MatrixMatrix::add(ONE, false, *identity, ONE, false, *transferMatrices_[i]->pointA_);
+        RCP<matrix_type> temp2 = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
+        MatrixMatrix::Multiply(*transfer, false, *temp, true, *temp2);
+        transfer = temp2;
       }
+    }
 
-      // farField = transfer * kernelApproximations_ * transfer^T
-      RCP<matrix_type> temp = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
-      MatrixMatrix::Multiply(*transfer, false, *kernelApproximations_->pointA_, false, *temp);
-      RCP<matrix_type> farField = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
-      MatrixMatrix::Multiply(*temp, false, *transfer, true, *farField);
-
-      // nearField_ + farField
-      return MatrixMatrix::add(ONE, false, *nearField_, ONE, false, *farField);
-
-    } else
+    // farField = transfer * kernelApproximations_ * transfer^T
+    RCP<matrix_type> temp = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
+    MatrixMatrix::Multiply(*transfer, false, *kernelApproximations_->pointA_, false, *temp);
+    RCP<matrix_type> farField = rcp(new matrix_type(basisMatrix_->getRowMap(), 0));
+    MatrixMatrix::Multiply(*temp, false, *transfer, true, *farField);
 
-      return nearField_;
+    // nearField_ + farField
+    return MatrixMatrix::add(ONE, false, *nearField_, ONE, false, *farField);
 
-  }
+  } else
 
+    return nearField_;
+}
 
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  void
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  allocateMemory(size_t numVectors) const {
-    if (coefficients_.is_null() || coefficients_->getNumVectors() != numVectors) {
-      coefficients_  = Teuchos::rcp(new mv_type(clusterCoeffMap_, numVectors));
-      coefficients2_ = Teuchos::rcp(new mv_type(clusterCoeffMap_, numVectors));
-      X_colmap_ = Teuchos::rcp(new mv_type(nearField_->getColMap(), numVectors));
-      coefficients_colmap_  = Teuchos::rcp(new mv_type(kernelApproximations_->pointA_->getColMap(), numVectors));
-    }
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+void HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    allocateMemory(size_t numVectors) const {
+  if (coefficients_.is_null() || coefficients_->getNumVectors() != numVectors) {
+    coefficients_        = Teuchos::rcp(new mv_type(clusterCoeffMap_, numVectors));
+    coefficients2_       = Teuchos::rcp(new mv_type(clusterCoeffMap_, numVectors));
+    X_colmap_            = Teuchos::rcp(new mv_type(nearField_->getColMap(), numVectors));
+    coefficients_colmap_ = Teuchos::rcp(new mv_type(kernelApproximations_->pointA_->getColMap(), numVectors));
   }
 }
+}  // namespace Tpetra
 
-#endif // TPETRA_HIERARCHICALOPERATOR_DEF_HPP
+#endif  // TPETRA_HIERARCHICALOPERATOR_DEF_HPP
diff --git a/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_decl.hpp b/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_decl.hpp
index f511e0dfded5..f65d3a7a8bca 100644
--- a/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_decl.hpp
+++ b/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_decl.hpp
@@ -9,111 +9,110 @@
 #include <Xpetra_Matrix.hpp>
 #include <Xpetra_TpetraBlockedMatrix.hpp>
 
-
 namespace Xpetra {
 
-  template<class Scalar,
-           class LocalOrdinal,
-           class GlobalOrdinal,
-           class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class HierarchicalOperator : public TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-
-  public:
-    using tHOp = Tpetra::HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using map_type = Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>;
-    using mv_type = Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using matrix_type = Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using blocked_matrix_type = Xpetra::TpetraBlockedMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    HierarchicalOperator(const RCP<tHOp>& op) : op_(op) { }
-
-    HierarchicalOperator(const RCP<matrix_type>& nearField,
-                         const RCP<blocked_matrix_type>& kernelApproximations,
-                         const RCP<matrix_type>& basisMatrix,
-                         std::vector<RCP<blocked_matrix_type> >& transferMatrices,
-                         const Teuchos::RCP<Teuchos::ParameterList>& params=Teuchos::null);
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const map_type> getDomainMap() const {
-      return toXpetra(op_->getDomainMap());
-    }
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const map_type> getRangeMap() const {
-      return toXpetra(op_->getRangeMap());
-    }
-
-    //! \brief Computes the operator-multivector application.
-    /*! Loosely, performs \f$Y = \alpha \cdot A^{\textrm{mode}} \cdot X + \beta \cdot Y\f$. However, the details of operation
-      vary according to the values of \c alpha and \c beta. Specifically
-      - if <tt>beta == 0</tt>, apply() <b>must</b> overwrite \c Y, so that any values in \c Y (including NaNs) are ignored.
-      - if <tt>alpha == 0</tt>, apply() <b>may</b> short-circuit the operator, so that any values in \c X (including NaNs) are ignored.
-    */
-    void apply (const mv_type& X, mv_type& Y,
-                Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const {
-      op_->apply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
-    }
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const mv_type & X,
-                  const mv_type & B,
-                  mv_type& R) const {
-      Tpetra::Details::residual(*op_, toTpetra(X), toTpetra(B), toTpetra(R));
-    }
-
-    RCP<HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > restrict(const RCP<matrix_type>& P) {
-      using TpCrs = TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      using CrsWrap = CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      return Teuchos::rcp(new HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>(op_->restrict(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(P)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst())));
-    }
-
-    RCP<matrix_type> toMatrix() {
-      using TpCrs = TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      using CrsWrap = CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      auto tpMat = Teuchos::rcp(new TpCrs(op_->toMatrix()));
-      return Teuchos::rcp(new CrsWrap(Teuchos::rcp_dynamic_cast<CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tpMat)));
-    }
-
-    double getCompression() {
-      return op_->getCompression();
-    }
-
-    RCP<matrix_type> nearFieldMatrix() {
-      auto tpMat = Teuchos::rcp(new TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(op_->nearFieldMatrix()));
-      return Teuchos::rcp(new CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(Teuchos::rcp_dynamic_cast<CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tpMat)));
-    }
-
-    //! Gets the operator out
-    RCP<Tpetra::Operator< Scalar, LocalOrdinal, GlobalOrdinal, Node> > getOperator() { return op_; }
-
-    RCP<const Tpetra::Operator< Scalar, LocalOrdinal, GlobalOrdinal, Node> > getOperatorConst() const { return op_; }
-
-    void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
-      describe(out, verbLevel, true);
-    }
-
-    void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel, const bool printHeader) const {
-      op_->describe(out, verbLevel, printHeader);
-    }
-
-    bool hasFarField() const {
-      return op_->hasFarField();
-    }
-
-    bool denserThanDenseMatrix() const {
-      return op_->denserThanDenseMatrix();
-    }
-
-  private:
-    RCP<tHOp> op_;
-  };
-
-}
-
-#endif // XPETRA_HIERARCHICALOPERATOR_DECL_HPP
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class HierarchicalOperator : public TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  using tHOp                = Tpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using map_type            = Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+  using mv_type             = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using matrix_type         = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using blocked_matrix_type = Xpetra::TpetraBlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  HierarchicalOperator(const RCP<tHOp>& op)
+    : op_(op) {}
+
+  HierarchicalOperator(const RCP<matrix_type>& nearField,
+                       const RCP<blocked_matrix_type>& kernelApproximations,
+                       const RCP<matrix_type>& basisMatrix,
+                       std::vector<RCP<blocked_matrix_type> >& transferMatrices,
+                       const Teuchos::RCP<Teuchos::ParameterList>& params = Teuchos::null);
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const map_type> getDomainMap() const {
+    return toXpetra(op_->getDomainMap());
+  }
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const map_type> getRangeMap() const {
+    return toXpetra(op_->getRangeMap());
+  }
+
+  //! \brief Computes the operator-multivector application.
+  /*! Loosely, performs \f$Y = \alpha \cdot A^{\textrm{mode}} \cdot X + \beta \cdot Y\f$. However, the details of operation
+    vary according to the values of \c alpha and \c beta. Specifically
+    - if <tt>beta == 0</tt>, apply() <b>must</b> overwrite \c Y, so that any values in \c Y (including NaNs) are ignored.
+    - if <tt>alpha == 0</tt>, apply() <b>may</b> short-circuit the operator, so that any values in \c X (including NaNs) are ignored.
+  */
+  void apply(const mv_type& X, mv_type& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
+    op_->apply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
+  }
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const mv_type& X,
+                const mv_type& B,
+                mv_type& R) const {
+    Tpetra::Details::residual(*op_, toTpetra(X), toTpetra(B), toTpetra(R));
+  }
+
+  RCP<HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > restrict(const RCP<matrix_type>& P) {
+    using TpCrs   = TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    using CrsWrap = CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    return Teuchos::rcp(new HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op_->restrict(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(P)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst())));
+  }
+
+  RCP<matrix_type> toMatrix() {
+    using TpCrs   = TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    using CrsWrap = CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    auto tpMat    = Teuchos::rcp(new TpCrs(op_->toMatrix()));
+    return Teuchos::rcp(new CrsWrap(Teuchos::rcp_dynamic_cast<CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(tpMat)));
+  }
+
+  double getCompression() {
+    return op_->getCompression();
+  }
+
+  RCP<matrix_type> nearFieldMatrix() {
+    auto tpMat = Teuchos::rcp(new TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op_->nearFieldMatrix()));
+    return Teuchos::rcp(new CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Teuchos::rcp_dynamic_cast<CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(tpMat)));
+  }
+
+  //! Gets the operator out
+  RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > getOperator() { return op_; }
+
+  RCP<const Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > getOperatorConst() const { return op_; }
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
+    describe(out, verbLevel, true);
+  }
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel, const bool printHeader) const {
+    op_->describe(out, verbLevel, printHeader);
+  }
+
+  bool hasFarField() const {
+    return op_->hasFarField();
+  }
+
+  bool denserThanDenseMatrix() const {
+    return op_->denserThanDenseMatrix();
+  }
+
+ private:
+  RCP<tHOp> op_;
+};
+
+}  // namespace Xpetra
+
+#endif  // XPETRA_HIERARCHICALOPERATOR_DECL_HPP
diff --git a/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_def.hpp b/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_def.hpp
index ef53bf383c6b..f59c48e3738a 100644
--- a/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_def.hpp
+++ b/packages/muelu/research/caglusa/Xpetra_HierarchicalOperator_def.hpp
@@ -3,35 +3,35 @@
 
 namespace Xpetra {
 
-  template<class Scalar,
-           class LocalOrdinal,
-           class GlobalOrdinal,
-           class Node>
-  HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
-                       const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
-                       const Teuchos::RCP<matrix_type>& basisMatrix,
-                       std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
-                       const Teuchos::RCP<Teuchos::ParameterList>& params) {
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    using TpCrs = TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using TpGOVec = TpetraMultiVector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>;
-    using CrsWrap = CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    HierarchicalOperator(const Teuchos::RCP<matrix_type>& nearField,
+                         const Teuchos::RCP<blocked_matrix_type>& kernelApproximations,
+                         const Teuchos::RCP<matrix_type>& basisMatrix,
+                         std::vector<Teuchos::RCP<blocked_matrix_type> >& transferMatrices,
+                         const Teuchos::RCP<Teuchos::ParameterList>& params) {
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  using TpCrs   = TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TpGOVec = TpetraMultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>;
+  using CrsWrap = CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    std::vector<RCP<typename tHOp::blocked_matrix_type> > tTransferMatrices;
-    for (size_t i = 0; i<transferMatrices.size(); i++) {
-      auto transferT = transferMatrices[i]->getTpetra_BlockedMatrix();
-      tTransferMatrices.push_back(transferT);
-    }
-
-    op_ = rcp(new tHOp(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(nearField)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
-                       kernelApproximations->getTpetra_BlockedMatrix(),
-                       Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(basisMatrix)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
-                       tTransferMatrices,
-                       params));
+  std::vector<RCP<typename tHOp::blocked_matrix_type> > tTransferMatrices;
+  for (size_t i = 0; i < transferMatrices.size(); i++) {
+    auto transferT = transferMatrices[i]->getTpetra_BlockedMatrix();
+    tTransferMatrices.push_back(transferT);
   }
 
+  op_ = rcp(new tHOp(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(nearField)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
+                     kernelApproximations->getTpetra_BlockedMatrix(),
+                     Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(basisMatrix)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
+                     tTransferMatrices,
+                     params));
 }
 
-#endif // XPETRA_HIERARCHICALOPERATOR_DEF_HPP
+}  // namespace Xpetra
+
+#endif  // XPETRA_HIERARCHICALOPERATOR_DEF_HPP
diff --git a/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMap.hpp b/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMap.hpp
index c0259d922220..fca269d36458 100644
--- a/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMap.hpp
+++ b/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMap.hpp
@@ -8,36 +8,32 @@
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_TpetraVector.hpp>
 
-
 namespace Xpetra {
 
-  template <class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class TpetraBlockedMap {
-
-  public:
-    using map_type = Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
-    using lo_vec_type = Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>;
-    using tpetra_blocked_map_type = Tpetra::BlockedMap<LocalOrdinal,GlobalOrdinal,Node>;
-
-    TpetraBlockedMap(const Teuchos::RCP<const map_type>& pointMap,
-                     const Teuchos::RCP<lo_vec_type>& blockSizes)
-    {
-      using TpLOVec = TpetraVector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>;
-      tpBlockedMap_ = Teuchos::rcp(new tpetra_blocked_map_type(toTpetra(pointMap),
-                                                               Teuchos::rcp_dynamic_cast<TpLOVec>(blockSizes)->getTpetra_Vector()));
-    }
-
-    RCP<tpetra_blocked_map_type> getTpetra_BlockedMap() const {
-      return tpBlockedMap_;
-    }
-
-  private:
-    RCP<tpetra_blocked_map_type> tpBlockedMap_;
-
-  };
-
-}
-
-#endif // XPETRA_TPETRABLOCKEDMAP_HPP
+template <class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class TpetraBlockedMap {
+ public:
+  using map_type                = Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+  using lo_vec_type             = Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>;
+  using tpetra_blocked_map_type = Tpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>;
+
+  TpetraBlockedMap(const Teuchos::RCP<const map_type>& pointMap,
+                   const Teuchos::RCP<lo_vec_type>& blockSizes) {
+    using TpLOVec = TpetraVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>;
+    tpBlockedMap_ = Teuchos::rcp(new tpetra_blocked_map_type(toTpetra(pointMap),
+                                                             Teuchos::rcp_dynamic_cast<TpLOVec>(blockSizes)->getTpetra_Vector()));
+  }
+
+  RCP<tpetra_blocked_map_type> getTpetra_BlockedMap() const {
+    return tpBlockedMap_;
+  }
+
+ private:
+  RCP<tpetra_blocked_map_type> tpBlockedMap_;
+};
+
+}  // namespace Xpetra
+
+#endif  // XPETRA_TPETRABLOCKEDMAP_HPP
diff --git a/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMatrix.hpp b/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMatrix.hpp
index 85d18c207e0d..0145219ee0f3 100644
--- a/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMatrix.hpp
+++ b/packages/muelu/research/caglusa/Xpetra_TpetraBlockedMatrix.hpp
@@ -9,61 +9,57 @@
 #include <Tpetra_BlockedMatrix_decl.hpp>
 #include <Tpetra_BlockedMatrix_def.hpp>
 
-
 namespace Xpetra {
 
-  template <class Scalar = Tpetra::Operator<>::scalar_type,
-            class LocalOrdinal = typename Tpetra::Operator<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class TpetraBlockedMatrix {
-
-  public:
-    using matrix_type = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using blocked_map_type = TpetraBlockedMap<LocalOrdinal,GlobalOrdinal,Node>;
-    using tpetra_blocked_matrix_type = Tpetra::BlockedMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-
-    TpetraBlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
-                        const Teuchos::RCP<matrix_type>& blockA,
-                        const Teuchos::RCP<blocked_map_type>& blockMap,
-                        const Teuchos::RCP<blocked_map_type>& ghosted_blockMap=Teuchos::null)
-    {
-      using TpCrs = TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      using CrsWrap = CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      RCP<typename blocked_map_type::tpetra_blocked_map_type> tp_ghosted_blockMap;
-      if (!ghosted_blockMap.is_null())
-        tp_ghosted_blockMap = ghosted_blockMap->getTpetra_BlockedMap();
-      blockMatrix_ = Teuchos::rcp(new tpetra_blocked_matrix_type(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(pointA)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
-                                                                 Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(blockA)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
-                                                                 blockMap->getTpetra_BlockedMap(),
-                                                                 tp_ghosted_blockMap));
-    }
+template <class Scalar        = Tpetra::Operator<>::scalar_type,
+          class LocalOrdinal  = typename Tpetra::Operator<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename Tpetra::Operator<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class TpetraBlockedMatrix {
+ public:
+  using matrix_type                = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using blocked_map_type           = TpetraBlockedMap<LocalOrdinal, GlobalOrdinal, Node>;
+  using tpetra_blocked_matrix_type = Tpetra::BlockedMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    void apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-               Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-               Teuchos::ETransp mode = Teuchos::NO_TRANS,
-               Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-               Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const {
-      blockMatrix_->apply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
-    }
+  TpetraBlockedMatrix(const Teuchos::RCP<matrix_type>& pointA,
+                      const Teuchos::RCP<matrix_type>& blockA,
+                      const Teuchos::RCP<blocked_map_type>& blockMap,
+                      const Teuchos::RCP<blocked_map_type>& ghosted_blockMap = Teuchos::null) {
+    using TpCrs   = TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    using CrsWrap = CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    RCP<typename blocked_map_type::tpetra_blocked_map_type> tp_ghosted_blockMap;
+    if (!ghosted_blockMap.is_null())
+      tp_ghosted_blockMap = ghosted_blockMap->getTpetra_BlockedMap();
+    blockMatrix_ = Teuchos::rcp(new tpetra_blocked_matrix_type(Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(pointA)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
+                                                               Teuchos::rcp_dynamic_cast<TpCrs>(Teuchos::rcp_dynamic_cast<CrsWrap>(blockA)->getCrsMatrix(), true)->getTpetra_CrsMatrixNonConst(),
+                                                               blockMap->getTpetra_BlockedMap(),
+                                                               tp_ghosted_blockMap));
+  }
 
-    void localApply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                    Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                    Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                    Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                    Scalar beta  = Teuchos::ScalarTraits<Scalar>::zero()) const {
-      blockMatrix_->localApply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
-    }
+  void apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
+    blockMatrix_->apply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
+  }
 
-    Teuchos::RCP<tpetra_blocked_matrix_type> getTpetra_BlockedMatrix() const {
-      return blockMatrix_;
-    }
+  void localApply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                  Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+                  Teuchos::ETransp mode = Teuchos::NO_TRANS,
+                  Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+                  Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
+    blockMatrix_->localApply(Xpetra::toTpetra(X), Xpetra::toTpetra(Y), mode, alpha, beta);
+  }
 
-  private:
-    Teuchos::RCP<tpetra_blocked_matrix_type> blockMatrix_;
+  Teuchos::RCP<tpetra_blocked_matrix_type> getTpetra_BlockedMatrix() const {
+    return blockMatrix_;
+  }
 
-  };
+ private:
+  Teuchos::RCP<tpetra_blocked_matrix_type> blockMatrix_;
+};
 
-}
+}  // namespace Xpetra
 
-#endif // XPETRA_TPETRABLOCKEDMATRIX_HPP
+#endif  // XPETRA_TPETRABLOCKEDMATRIX_HPP
diff --git a/packages/muelu/research/caglusa/auxiliaryOperators.hpp b/packages/muelu/research/caglusa/auxiliaryOperators.hpp
index faa61c24eab5..a8b274e8b444 100644
--- a/packages/muelu/research/caglusa/auxiliaryOperators.hpp
+++ b/packages/muelu/research/caglusa/auxiliaryOperators.hpp
@@ -19,287 +19,282 @@
 #include <MueLu_CoalesceDropFactory_kokkos.hpp>
 #include <MueLu_ThresholdAFilterFactory.hpp>
 
-
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  buildDistanceLaplacian(RCP<const Xpetra::CrsGraph<LocalOrdinal,GlobalOrdinal,Node> >& graph,
-                         RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> >& coords)
-  {
-
-    const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    const LocalOrdinal INV = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
-    auto distLapl = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(graph);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+buildDistanceLaplacian(RCP<const Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> >& graph,
+                       RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coords) {
+  const Scalar ONE       = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar ZERO      = Teuchos::ScalarTraits<Scalar>::zero();
+  const LocalOrdinal INV = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
+  auto distLapl          = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(graph);
 
-    auto rowMap = graph->getRowMap();
-    auto colMap = graph->getColMap();
-    auto ghosted_coords = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap, coords->getNumVectors());
-    ghosted_coords->doImport(*coords, *graph->getImporter(), Xpetra::INSERT);
+  auto rowMap         = graph->getRowMap();
+  auto colMap         = graph->getColMap();
+  auto ghosted_coords = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(colMap, coords->getNumVectors());
+  ghosted_coords->doImport(*coords, *graph->getImporter(), Xpetra::INSERT);
 
-    {
-      auto lcl_coords = coords->getHostLocalView(Xpetra::Access::ReadOnly);
-      auto lcl_ghosted_coords = ghosted_coords->getHostLocalView(Xpetra::Access::ReadOnly);
-      auto lcl_distLapl = distLapl->getLocalMatrixHost();
-
-      // TODO: parallel_for
-      for (LocalOrdinal rlid = 0; rlid < lcl_distLapl.numRows(); ++rlid) {
-        auto row = lcl_distLapl.row(rlid);
-        Scalar diag = ZERO;
-        LocalOrdinal diagIndex = INV;
-        for (LocalOrdinal k = 0; k < row.length; ++k) {
-          LocalOrdinal clid = row.colidx(k);
-          if (rowMap->getGlobalElement(rlid) == colMap->getGlobalElement(clid)) {
-            diagIndex = k;
-          } else {
-            Scalar dist = ZERO;
-            for (size_t j = 0; j < lcl_coords.extent(1); j++) {
-              auto s = lcl_coords(rlid,j) - lcl_ghosted_coords(clid,j);
-              dist += s*s;
-            }
-            row.value(k) = ONE/std::sqrt(dist);
-            diag -= row.value(k);
+  {
+    auto lcl_coords         = coords->getHostLocalView(Xpetra::Access::ReadOnly);
+    auto lcl_ghosted_coords = ghosted_coords->getHostLocalView(Xpetra::Access::ReadOnly);
+    auto lcl_distLapl       = distLapl->getLocalMatrixHost();
+
+    // TODO: parallel_for
+    for (LocalOrdinal rlid = 0; rlid < lcl_distLapl.numRows(); ++rlid) {
+      auto row               = lcl_distLapl.row(rlid);
+      Scalar diag            = ZERO;
+      LocalOrdinal diagIndex = INV;
+      for (LocalOrdinal k = 0; k < row.length; ++k) {
+        LocalOrdinal clid = row.colidx(k);
+        if (rowMap->getGlobalElement(rlid) == colMap->getGlobalElement(clid)) {
+          diagIndex = k;
+        } else {
+          Scalar dist = ZERO;
+          for (size_t j = 0; j < lcl_coords.extent(1); j++) {
+            auto s = lcl_coords(rlid, j) - lcl_ghosted_coords(clid, j);
+            dist += s * s;
           }
+          row.value(k) = ONE / std::sqrt(dist);
+          diag -= row.value(k);
         }
-        TEUCHOS_ASSERT(diagIndex != INV);
-        row.value(diagIndex) = diag;
       }
+      TEUCHOS_ASSERT(diagIndex != INV);
+      row.value(diagIndex) = diag;
     }
-    distLapl->fillComplete();
-    return distLapl;
   }
+  distLapl->fillComplete();
+  return distLapl;
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  constructAuxiliaryOperator(RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op,
-                             Teuchos::ParameterList& problemParams) {
-
-    using IO = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    using IOhelpers = MueLu::IOhelpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+constructAuxiliaryOperator(RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op,
+                           Teuchos::ParameterList& problemParams) {
+  using IO        = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using IOhelpers = MueLu::IOhelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    RCP<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > hop = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(op);
+  RCP<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > hop = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(op);
 
-    RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > auxOp;
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > auxOp;
 
-    const std::string auxOpStr = problemParams.get<std::string>("auxiliary operator");
+  const std::string auxOpStr = problemParams.get<std::string>("auxiliary operator");
 
-    if ((auxOpStr == "near") || (auxOpStr == "distanceLaplacian")) {
-      if (hop.is_null())
-        auxOp = rcp_dynamic_cast<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(op, true);
-      else
-        auxOp = hop->nearFieldMatrix();
+  if ((auxOpStr == "near") || (auxOpStr == "distanceLaplacian")) {
+    if (hop.is_null())
+      auxOp = rcp_dynamic_cast<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(op, true);
+    else
+      auxOp = hop->nearFieldMatrix();
 #ifdef MUELU_HIERARCHICAL_DEBUG
-      // CoalesceDropFactory_kokkos assumes fitted row and column maps
-      Xpetra::MatrixUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::checkLocalRowMapMatchesColMap(*auxOp);
+    // CoalesceDropFactory_kokkos assumes fitted row and column maps
+    Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkLocalRowMapMatchesColMap(*auxOp);
 #endif
-      // coordinates
-      auto coords = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>::ReadMultiVector(problemParams.get<std::string>("coordinates"), op->getRangeMap());
+    // coordinates
+    auto coords = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(problemParams.get<std::string>("coordinates"), op->getRangeMap());
 
-      {
-        // apply dropping to auxOp
-        Level fineLevel;
-        fineLevel.SetFactoryManager(Teuchos::null);
-        fineLevel.SetLevelID(0);
-        fineLevel.Set("A",auxOp);
-        fineLevel.Set("Coordinates",coords);
-        fineLevel.Set("DofsPerNode",1);
-        fineLevel.setlib(auxOp->getDomainMap()->lib());
-        auto amalgFact = rcp(new AmalgamationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-        auto dropFact = rcp(new CoalesceDropFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-        dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-        double dropTol = problemParams.get<double>("drop tolerance");
-        // double dropTol = 0.1; // 1D
-        // double dropTol = 0.03; // 2D
-        std::string dropScheme = "classical";
-        dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-        dropFact->SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(dropScheme));
-
-        fineLevel.Request("A",dropFact.get());
-        fineLevel.Get("A", auxOp, dropFact.get());
-      }
+    {
+      // apply dropping to auxOp
+      Level fineLevel;
+      fineLevel.SetFactoryManager(Teuchos::null);
+      fineLevel.SetLevelID(0);
+      fineLevel.Set("A", auxOp);
+      fineLevel.Set("Coordinates", coords);
+      fineLevel.Set("DofsPerNode", 1);
+      fineLevel.setlib(auxOp->getDomainMap()->lib());
+      auto amalgFact = rcp(new AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+      auto dropFact  = rcp(new CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+      double dropTol = problemParams.get<double>("drop tolerance");
+      // double dropTol = 0.1; // 1D
+      // double dropTol = 0.03; // 2D
+      std::string dropScheme = "classical";
+      dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+      dropFact->SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(dropScheme));
+
+      fineLevel.Request("A", dropFact.get());
+      fineLevel.Get("A", auxOp, dropFact.get());
+    }
 
-      {
-        // filter out small entries in auxOp
-        Level fineLevel;
-        fineLevel.SetFactoryManager(Teuchos::null);
-        fineLevel.SetLevelID(0);
-        fineLevel.Set("A",auxOp);
-        auto filterFact = rcp(new ThresholdAFilterFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>("A", 1.0e-8, true, -1));
-        fineLevel.Request("A",filterFact.get());
-        filterFact->Build(fineLevel);
-        auxOp = fineLevel.Get< RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >("A",filterFact.get());
-      }
+    {
+      // filter out small entries in auxOp
+      Level fineLevel;
+      fineLevel.SetFactoryManager(Teuchos::null);
+      fineLevel.SetLevelID(0);
+      fineLevel.Set("A", auxOp);
+      auto filterFact = rcp(new ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>("A", 1.0e-8, true, -1));
+      fineLevel.Request("A", filterFact.get());
+      filterFact->Build(fineLevel);
+      auxOp = fineLevel.Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("A", filterFact.get());
+    }
 
-      if (auxOpStr == "distanceLaplacian") {
-        // build distance Laplacian using graph of auxOp and coordinates
-        auto graph = auxOp->getCrsGraph();
-        auxOp = buildDistanceLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>(graph, coords);
-      }
+    if (auxOpStr == "distanceLaplacian") {
+      // build distance Laplacian using graph of auxOp and coordinates
+      auto graph = auxOp->getCrsGraph();
+      auxOp      = buildDistanceLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>(graph, coords);
+    }
 
-    } else {
-      const bool readBinary = problemParams.get<bool>("read binary", false);
-      const bool readLocal = problemParams.get<bool>("read local", false);
+  } else {
+    const bool readBinary = problemParams.get<bool>("read binary", false);
+    const bool readLocal  = problemParams.get<bool>("read local", false);
+
+    // colmap of auxiliary operator
+    auto aux_colmap = IO::ReadMap(problemParams.get<std::string>("aux colmap"), op->getRangeMap()->lib(), op->getRangeMap()->getComm(), readBinary);
 
-      // colmap of auxiliary operator
-      auto aux_colmap = IO::ReadMap(problemParams.get<std::string>("aux colmap"), op->getRangeMap()->lib(), op->getRangeMap()->getComm(), readBinary);
+    auxOp = IOhelpers::Read(auxOpStr, op->getRangeMap(), aux_colmap, op->getRangeMap(), op->getRangeMap(), true, readBinary, readLocal);
+  }
+
+  return auxOp;
+}
 
-      auxOp = IOhelpers::Read(auxOpStr, op->getRangeMap(), aux_colmap, op->getRangeMap(), op->getRangeMap(), true, readBinary, readLocal);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+constructHierarchyFromAuxiliary(RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > op,
+                                RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > auxH,
+                                Teuchos::ParameterList& params,
+                                Teuchos::FancyOStream& out) {
+  params.set("coarse: max size", 1);
+  params.set("max levels", auxH->GetNumLevels());
+
+  const bool implicitTranspose = params.get("transpose: use implicit", MueLu::MasterList::getDefault<bool>("transpose: use implicit"));
+
+  auto hop = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(op);
+  if (!hop.is_null())
+    op->describe(out, Teuchos::VERB_EXTREME);
+
+  RCP<Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > H = rcp(new Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+  RCP<Level> lvl                                               = H->GetLevel(0);
+  lvl->Set("A", op);
+  // lvl->Set("Coordinates", coords);
+  for (int lvlNo = 1; lvlNo < auxH->GetNumLevels(); lvlNo++) {
+    RCP<Level> fineLvl = H->GetLevel(lvlNo - 1);
+    H->AddNewLevel();
+    lvl               = H->GetLevel(lvlNo);
+    RCP<Level> auxLvl = auxH->GetLevel(lvlNo);
+    // auto mgr = auxLvl->GetFactoryManager();
+    // auxLvl->print(std::cout, MueLu::Debug);
+
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > P         = auxLvl->Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("P");
+    RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > fineAOp = fineLvl->Get<RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("A");
+    lvl->Set("P", P);
+    params.sublist("level " + std::to_string(lvlNo)).set("P", P);
+
+    if (!implicitTranspose) {
+      TEUCHOS_ASSERT(auxLvl->IsAvailable("R"));
+      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > R = auxLvl->Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("R");
+      lvl->Set("R", R);
+      params.sublist("level " + std::to_string(lvlNo)).set("R", R);
     }
 
-    return auxOp;
-  }
+    auto fineA = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(fineAOp);
+    if (!fineA.is_null()) {
+      auto coarseA = fineA->restrict(P);
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  constructHierarchyFromAuxiliary(RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > op,
-                                  RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > auxH,
-                                  Teuchos::ParameterList& params,
-                                  Teuchos::FancyOStream& out) {
-
-    params.set("coarse: max size", 1);
-    params.set("max levels", auxH->GetNumLevels());
-
-    const bool implicitTranspose = params.get("transpose: use implicit", MueLu::MasterList::getDefault<bool>("transpose: use implicit"));
-
-    auto hop = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(op);
-    if (!hop.is_null())
-      op->describe(out, Teuchos::VERB_EXTREME);
-
-    RCP<Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > H = rcp(new Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-    RCP<Level> lvl = H->GetLevel(0);
-    lvl->Set("A", op);
-    // lvl->Set("Coordinates", coords);
-    for(int lvlNo = 1; lvlNo < auxH->GetNumLevels(); lvlNo++) {
-      RCP<Level> fineLvl = H->GetLevel(lvlNo-1);
-      H->AddNewLevel();
-      lvl = H->GetLevel(lvlNo);
-      RCP<Level> auxLvl = auxH->GetLevel(lvlNo);
-      // auto mgr = auxLvl->GetFactoryManager();
-      // auxLvl->print(std::cout, MueLu::Debug);
-
-      RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > P = auxLvl->Get<RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >("P");
-      RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > fineAOp = fineLvl->Get<RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >("A");
-      lvl->Set("P", P);
-      params.sublist("level "+std::to_string(lvlNo)).set("P", P);
-
-      if (!implicitTranspose) {
-        TEUCHOS_ASSERT(auxLvl->IsAvailable("R"));
-        RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > R = auxLvl->Get<RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >("R");
-        lvl->Set("R", R);
-        params.sublist("level "+std::to_string(lvlNo)).set("R", R);
+#ifdef MUELU_HIERARCHICAL_DEBUG
+      {
+        // Test that the Galerkin product worked
+        using MagnitudeType     = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+        const Scalar one        = Teuchos::ScalarTraits<Scalar>::one();
+        const MagnitudeType tol = 10000 * Teuchos::ScalarTraits<MagnitudeType>::eps();
+        auto testLHS            = MultiVectorFactory::Build(coarseA->getDomainMap(), 1);
+        auto testRHS_HOp_coarse = MultiVectorFactory::Build(coarseA->getRangeMap(), 1);
+        auto testRHS_HOp_fine   = MultiVectorFactory::Build(coarseA->getRangeMap(), 1);
+        auto temp1              = MultiVectorFactory::Build(fineA->getDomainMap(), 1);
+        auto temp2              = MultiVectorFactory::Build(fineA->getRangeMap(), 1);
+        testLHS->putScalar(one);
+        coarseA->apply(*testLHS, *testRHS_HOp_coarse);
+        P->apply(*testLHS, *temp1);
+        fineA->apply(*temp1, *temp2);
+        P->apply(*temp2, *testRHS_HOp_fine, Teuchos::TRANS);
+        testRHS_HOp_fine->update(one, *testRHS_HOp_coarse, -one);
+        auto norm = testRHS_HOp_fine->getVector(0)->norm2();
+        out << "|P^T*op_fine*P*1 - op_H_coarse*1| = " << norm << std::endl;
+        TEUCHOS_ASSERT(norm < tol);
       }
+#endif
+
+      if ((lvlNo + 1 == auxH->GetNumLevels()) || !coarseA->hasFarField() || coarseA->denserThanDenseMatrix()) {
+        // coarseA->describe(out, Teuchos::VERB_EXTREME);
 
-      auto fineA = rcp_dynamic_cast<Xpetra::HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(fineAOp);
-      if (!fineA.is_null()) {
-        auto coarseA = fineA->restrict(P);
+        auto matA = coarseA->toMatrix();
 
 #ifdef MUELU_HIERARCHICAL_DEBUG
         {
-          // Test that the Galerkin product worked
-          using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-          const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-          const MagnitudeType tol = 10000*Teuchos::ScalarTraits<MagnitudeType>::eps();
+          // test that the conversion to Crs format worked
+          using MagnitudeType     = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+          const Scalar one        = Teuchos::ScalarTraits<Scalar>::one();
+          const MagnitudeType tol = 10000 * Teuchos::ScalarTraits<MagnitudeType>::eps();
           auto testLHS            = MultiVectorFactory::Build(coarseA->getDomainMap(), 1);
-          auto testRHS_HOp_coarse = MultiVectorFactory::Build(coarseA->getRangeMap(),  1);
-          auto testRHS_HOp_fine   = MultiVectorFactory::Build(coarseA->getRangeMap(),  1);
-          auto temp1              = MultiVectorFactory::Build(fineA->getDomainMap(),   1);
-          auto temp2              = MultiVectorFactory::Build(fineA->getRangeMap(),    1);
+          auto testRHS_HOp        = MultiVectorFactory::Build(coarseA->getRangeMap(), 1);
+          auto testRHS_dense      = MultiVectorFactory::Build(coarseA->getRangeMap(), 1);
           testLHS->putScalar(one);
-          coarseA->apply(*testLHS, *testRHS_HOp_coarse);
-          P->apply(*testLHS, *temp1);
-          fineA->apply(*temp1, *temp2);
-          P->apply(*temp2, *testRHS_HOp_fine, Teuchos::TRANS);
-          testRHS_HOp_fine->update(one, *testRHS_HOp_coarse, -one);
-          auto norm = testRHS_HOp_fine->getVector(0)->norm2();
-          out << "|P^T*op_fine*P*1 - op_H_coarse*1| = " << norm << std::endl;
+          coarseA->apply(*testLHS, *testRHS_HOp);
+          matA->apply(*testLHS, *testRHS_dense);
+          testRHS_dense->update(one, *testRHS_HOp, -one);
+          auto norm = testRHS_dense->getVector(0)->norm2();
+          out << "|op_dense*1 - op_H*1| = " << norm << std::endl;
           TEUCHOS_ASSERT(norm < tol);
         }
 #endif
 
-        if ((lvlNo+1 == auxH->GetNumLevels()) || !coarseA->hasFarField() || coarseA->denserThanDenseMatrix()) {
-          // coarseA->describe(out, Teuchos::VERB_EXTREME);
-
-          auto matA = coarseA->toMatrix();
-
-#ifdef MUELU_HIERARCHICAL_DEBUG
-          {
-            // test that the conversion to Crs format worked
-            using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-            const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-            const MagnitudeType tol = 10000*Teuchos::ScalarTraits<MagnitudeType>::eps();
-            auto testLHS       = MultiVectorFactory::Build(coarseA->getDomainMap(), 1);
-            auto testRHS_HOp   = MultiVectorFactory::Build(coarseA->getRangeMap(),  1);
-            auto testRHS_dense = MultiVectorFactory::Build(coarseA->getRangeMap(),  1);
-            testLHS->putScalar(one);
-            coarseA->apply(*testLHS, *testRHS_HOp);
-            matA->apply(*testLHS, *testRHS_dense);
-            testRHS_dense->update(one, *testRHS_HOp, -one);
-            auto norm = testRHS_dense->getVector(0)->norm2();
-            out << "|op_dense*1 - op_H*1| = " << norm << std::endl;
-            TEUCHOS_ASSERT(norm < tol);
-          }
-#endif
-
-          using std::setw;
-          using std::endl;
-          const size_t numRows = matA->getRowMap()->getGlobalNumElements();
-          const size_t nnz = matA->getGlobalNumEntries();
-          const double nnzPerRow = Teuchos::as<double>(nnz)/numRows;
-          std::ostringstream oss;
-          oss << std::left;
-          // oss << setw(9) << "rows"  << setw(12) << "nnz"  << setw(14) << "nnz/row" << setw(12)  << endl;
-          oss << setw(9) << numRows << setw(12) << nnz << setw(14) << nnzPerRow << endl;
-          out << oss.str();
-
-          lvl->Set("A", matA);
-        } else {
-          coarseA->describe(out, Teuchos::VERB_EXTREME, /*printHeader=*/false);
-          lvl->Set("A", rcp_dynamic_cast<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(coarseA));
-        }
-      } else {
-        // classical RAP
-        auto fineAmat = rcp_dynamic_cast<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(fineAOp, true);
-        Level fineLevel, coarseLevel;
-        fineLevel.SetFactoryManager(Teuchos::null);
-        coarseLevel.SetFactoryManager(Teuchos::null);
-        coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
-        fineLevel.SetLevelID(0);
-        coarseLevel.SetLevelID(1);
-        fineLevel.Set("A", fineAmat);
-        coarseLevel.Set("P", P);
-        RCP<RAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rapFact = rcp(new RAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-        Teuchos::ParameterList rapList = *(rapFact->GetValidParameterList());
-        rapList.set("transpose: use implicit", true);
-        rapFact->SetParameterList(rapList);
-        coarseLevel.Request("A", rapFact.get());
-        RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > matA = coarseLevel.Get<RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >("A", rapFact.get());
-
-        using std::setw;
         using std::endl;
-        const size_t numRows = matA->getRowMap()->getGlobalNumElements();
-        const size_t nnz = matA->getGlobalNumEntries();
-        const double nnzPerRow = Teuchos::as<double>(nnz)/numRows;
+        using std::setw;
+        const size_t numRows   = matA->getRowMap()->getGlobalNumElements();
+        const size_t nnz       = matA->getGlobalNumEntries();
+        const double nnzPerRow = Teuchos::as<double>(nnz) / numRows;
         std::ostringstream oss;
         oss << std::left;
-        oss << setw(9) << "rows"  << setw(12) << "nnz"  << setw(14) << "nnz/row" << setw(12)  << endl;
+        // oss << setw(9) << "rows"  << setw(12) << "nnz"  << setw(14) << "nnz/row" << setw(12)  << endl;
         oss << setw(9) << numRows << setw(12) << nnz << setw(14) << nnzPerRow << endl;
         out << oss.str();
 
         lvl->Set("A", matA);
+      } else {
+        coarseA->describe(out, Teuchos::VERB_EXTREME, /*printHeader=*/false);
+        lvl->Set("A", rcp_dynamic_cast<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(coarseA));
       }
+    } else {
+      // classical RAP
+      auto fineAmat = rcp_dynamic_cast<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(fineAOp, true);
+      Level fineLevel, coarseLevel;
+      fineLevel.SetFactoryManager(Teuchos::null);
+      coarseLevel.SetFactoryManager(Teuchos::null);
+      coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
+      fineLevel.SetLevelID(0);
+      coarseLevel.SetLevelID(1);
+      fineLevel.Set("A", fineAmat);
+      coarseLevel.Set("P", P);
+      RCP<RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rapFact = rcp(new RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+      Teuchos::ParameterList rapList                                      = *(rapFact->GetValidParameterList());
+      rapList.set("transpose: use implicit", true);
+      rapFact->SetParameterList(rapList);
+      coarseLevel.Request("A", rapFact.get());
+      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > matA = coarseLevel.Get<RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("A", rapFact.get());
+
+      using std::endl;
+      using std::setw;
+      const size_t numRows   = matA->getRowMap()->getGlobalNumElements();
+      const size_t nnz       = matA->getGlobalNumEntries();
+      const double nnzPerRow = Teuchos::as<double>(nnz) / numRows;
+      std::ostringstream oss;
+      oss << std::left;
+      oss << setw(9) << "rows" << setw(12) << "nnz" << setw(14) << "nnz/row" << setw(12) << endl;
+      oss << setw(9) << numRows << setw(12) << nnz << setw(14) << nnzPerRow << endl;
+      out << oss.str();
+
+      lvl->Set("A", matA);
     }
-
-    RCP<HierarchyManager<Scalar,LocalOrdinal,GlobalOrdinal,Node> > mueLuFactory = rcp(new ParameterListInterpreter<Scalar,LocalOrdinal,GlobalOrdinal,Node>(params,op->getDomainMap()->getComm()));
-    H->setlib(op->getDomainMap()->lib());
-    H->SetProcRankVerbose(op->getDomainMap()->getComm()->getRank());
-    mueLuFactory->SetupHierarchy(*H);
-    H->IsPreconditioner(true);
-
-    return H;
   }
 
+  RCP<HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> > mueLuFactory = rcp(new ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>(params, op->getDomainMap()->getComm()));
+  H->setlib(op->getDomainMap()->lib());
+  H->SetProcRankVerbose(op->getDomainMap()->getComm()->getRank());
+  mueLuFactory->SetupHierarchy(*H);
+  H->IsPreconditioner(true);
+
+  return H;
 }
 
+}  // namespace MueLu
+
 #endif
diff --git a/packages/muelu/research/caglusa/main.cpp b/packages/muelu/research/caglusa/main.cpp
index e0e2e607b551..0de4ac0a8ef6 100644
--- a/packages/muelu/research/caglusa/main.cpp
+++ b/packages/muelu/research/caglusa/main.cpp
@@ -68,81 +68,89 @@
 using Teuchos::RCP;
 using Teuchos::rcp;
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
-  #include "MueLu_UseShortNames.hpp"
-
-  std::string xmlHierarchical = "1d-binary/hierarchical.xml"; clp.setOption("xmlHierarchical", &xmlHierarchical, "XML describing the hierarchical operator");
-  std::string xmlProblem      = "1d-binary/problem.xml";      clp.setOption("xmlProblem",      &xmlProblem,      "XML describing the problem");
-  std::string xmlBelos        = "belos.xml";                  clp.setOption("xmlBelos",        &xmlBelos,        "XML with Belos parameters");
-  std::string xmlMueLu        = "muelu.xml";                  clp.setOption("xmlMueLu",        &xmlMueLu,        "XML with MueLu parameters");
-  std::string xmlAuxHierarchy = "auxiliary.xml";              clp.setOption("xmlAux",          &xmlAuxHierarchy, "XML with MueLu parameters for the auxiliary hierarchy");
-  bool printTimings  = true; clp.setOption("timings", "notimings", &printTimings,  "print timings to screen");
-  bool doTests       = true; clp.setOption("tests",   "notests",   &doTests,       "Test operator using known LHS & RHS.");
-  bool doUnPrecSolve = true; clp.setOption("unPrec",  "noUnPrec",  &doUnPrecSolve, "Solve unpreconditioned");
-  bool doPrecSolve   = true; clp.setOption("prec",    "noPrec",    &doPrecSolve,   "Solve preconditioned with AMG");
+#include "MueLu_UseShortNames.hpp"
+
+  std::string xmlHierarchical = "1d-binary/hierarchical.xml";
+  clp.setOption("xmlHierarchical", &xmlHierarchical, "XML describing the hierarchical operator");
+  std::string xmlProblem = "1d-binary/problem.xml";
+  clp.setOption("xmlProblem", &xmlProblem, "XML describing the problem");
+  std::string xmlBelos = "belos.xml";
+  clp.setOption("xmlBelos", &xmlBelos, "XML with Belos parameters");
+  std::string xmlMueLu = "muelu.xml";
+  clp.setOption("xmlMueLu", &xmlMueLu, "XML with MueLu parameters");
+  std::string xmlAuxHierarchy = "auxiliary.xml";
+  clp.setOption("xmlAux", &xmlAuxHierarchy, "XML with MueLu parameters for the auxiliary hierarchy");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  bool doTests = true;
+  clp.setOption("tests", "notests", &doTests, "Test operator using known LHS & RHS.");
+  bool doUnPrecSolve = true;
+  clp.setOption("unPrec", "noUnPrec", &doUnPrecSolve, "Solve unpreconditioned");
+  bool doPrecSolve = true;
+  clp.setOption("prec", "noPrec", &doPrecSolve, "Solve preconditioned with AMG");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   Teuchos::RCP<Teuchos::StackedTimer> stacked_timer = rcp(new Teuchos::StackedTimer("Hierarchical Driver"));
-  Teuchos::RCP<Teuchos::FancyOStream> verbose_out = Teuchos::rcp(new Teuchos::FancyOStream(Teuchos::rcpFromRef(std::cout)));
+  Teuchos::RCP<Teuchos::FancyOStream> verbose_out   = Teuchos::rcp(new Teuchos::FancyOStream(Teuchos::rcpFromRef(std::cout)));
   verbose_out->setShowProcRank(true);
   stacked_timer->setVerboseOstream(verbose_out);
   Teuchos::TimeMonitor::setStackedTimer(stacked_timer);
 
-  using HOp = Xpetra::HierarchicalOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-  using op_type = Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+  using HOp                 = Xpetra::HierarchicalOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using op_type             = Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using blocked_matrix_type = typename HOp::blocked_matrix_type;
-  using blocked_map_type = typename blocked_matrix_type::blocked_map_type;
-  using matrix_type = typename HOp::matrix_type;
-  using map_type = typename HOp::map_type;
-  using mv_type = typename HOp::mv_type;
-  using lo_vec_type = typename blocked_map_type::lo_vec_type;
-  using coord_mv = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>;
-  using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-  using IO = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-  using IOhelpers = MueLu::IOhelpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+  using blocked_map_type    = typename blocked_matrix_type::blocked_map_type;
+  using matrix_type         = typename HOp::matrix_type;
+  using map_type            = typename HOp::map_type;
+  using mv_type             = typename HOp::mv_type;
+  using lo_vec_type         = typename blocked_map_type::lo_vec_type;
+  using coord_mv            = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>;
+  using MagnitudeType       = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using IO                  = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using IOhelpers           = MueLu::IOhelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream &out       = *fancy;
   out.setOutputToRootOnly(0);
-  bool success = true;
-  const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-  const MagnitudeType tol = 100000*Teuchos::ScalarTraits<MagnitudeType>::eps();
+  bool success            = true;
+  const Scalar one        = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar zero       = Teuchos::ScalarTraits<Scalar>::zero();
+  const MagnitudeType tol = 100000 * Teuchos::ScalarTraits<MagnitudeType>::eps();
 
   RCP<op_type> op;
   RCP<HOp> hop;
   {
     Teuchos::TimeMonitor tM(*Teuchos::TimeMonitor::getNewTimer(std::string("Read hierarchical matrix")));
 
-    op = IOhelpers::Read(xmlHierarchical, comm);
+    op  = IOhelpers::Read(xmlHierarchical, comm);
     hop = Teuchos::rcp_dynamic_cast<HOp>(op);
   }
 
   if (!hop.is_null())
-    out << "Compression: " << hop->getCompression() << " of dense matrix."<< std::endl;
+    out << "Compression: " << hop->getCompression() << " of dense matrix." << std::endl;
 
   Teuchos::ParameterList problemParams;
   Teuchos::updateParametersFromXmlFileAndBroadcast(xmlProblem, Teuchos::Ptr<Teuchos::ParameterList>(&problemParams), *comm);
 
   RCP<const map_type> map = op->getDomainMap();
-  RCP<matrix_type>    auxOp, mass;
-  RCP<mv_type>        X_ex, RHS, X;
-  RCP<coord_mv>       coords;
+  RCP<matrix_type> auxOp, mass;
+  RCP<mv_type> X_ex, RHS, X;
+  RCP<coord_mv> coords;
   {
     // Read in auxiliary stuff
 
     // coordinates
-    coords = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>::ReadMultiVector(problemParams.get<std::string>("coordinates"), map);
+    coords = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(problemParams.get<std::string>("coordinates"), map);
 
     // Auxiliary matrix used for multigrid construction
     {
@@ -154,7 +162,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Mass matrix for L2 error computation
     {
       const bool readBinary = problemParams.get<bool>("read binary", false);
-      const bool readLocal = problemParams.get<bool>("read local", false);
+      const bool readLocal  = problemParams.get<bool>("read local", false);
       // colmap of auxiliary operator
       RCP<const map_type> aux_colmap = IO::ReadMap(problemParams.get<std::string>("aux colmap"), lib, comm, readBinary);
 
@@ -165,8 +173,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     X_ex = IO::ReadMultiVector(problemParams.get<std::string>("exact solution"), map);
     RHS  = IO::ReadMultiVector(problemParams.get<std::string>("right-hand side"), map);
     // solution vector
-    X    = MultiVectorFactory::Build(map, 1);
-
+    X = MultiVectorFactory::Build(map, 1);
   }
 
   if (doTests) {
@@ -222,11 +229,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     using OP = Belos::OperatorT<MV>;
 
     X->putScalar(zero);
-    RCP<OP> belosOp = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op));
+    RCP<OP> belosOp                                         = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op));
     RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, RHS));
 
     std::string belosType = "Pseudoblock CG";
-    auto belosSolverList = rcpFromRef(belosParams.sublist(belosType));
+    auto belosSolverList  = rcpFromRef(belosParams.sublist(belosType));
 
     bool set = belosProblem->setProblem();
     if (set == false) {
@@ -235,24 +242,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Create an iterative solver manager
     Belos::SolverFactory<Scalar, MV, OP> solverFactory;
-    RCP< Belos::SolverManager<Scalar, MV, OP> > solver = solverFactory.create(belosType, belosSolverList);
+    RCP<Belos::SolverManager<Scalar, MV, OP> > solver = solverFactory.create(belosType, belosSolverList);
     solver->setProblem(belosProblem);
 
     // Perform solve
     Belos::ReturnType ret = solver->solve();
-    int numIts = solver->getNumIters();
+    int numIts            = solver->getNumIters();
 
     // Get the number of iterations for this solve.
     out << "Number of iterations performed for this solve: " << numIts << std::endl;
 
     // Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("X.mtx", *X);
     X->update(one, *X_ex, -one);
-    out << "|X-X_ex| = " << X->getVector(0)->norm2() << std::endl << std::endl;
+    out << "|X-X_ex| = " << X->getVector(0)->norm2() << std::endl
+        << std::endl;
 
     success &= (ret == Belos::Converged);
-
   }
-#endif // HAVE_MUELU_BELOS
+#endif  // HAVE_MUELU_BELOS
 
   if (doPrecSolve) {
     // Solve linear system using a AMG preconditioned Krylov method
@@ -273,8 +280,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       auxParams.set("hierarchy label", "Auxiliary");
       auxParams.sublist("user data").set("Coordinates", coords);
       // No rebalancing yet.
-      auxParams.set("coarse: max size", std::max(auxParams.get("coarse: max size", 2*comm->getSize()),
-                                                 2*comm->getSize()));
+      auxParams.set("coarse: max size", std::max(auxParams.get("coarse: max size", 2 * comm->getSize()),
+                                                 2 * comm->getSize()));
 
       auxH = MueLu::CreateXpetraPreconditioner(auxOp, auxParams);
     }
@@ -295,7 +302,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       H = MueLu::constructHierarchyFromAuxiliary(Teuchos::rcp_dynamic_cast<Operator>(op, true), auxH, params, out);
     }
 
-
 #ifdef HAVE_MUELU_BELOS
     {
       ////////////////////////////////////////////////////////////////
@@ -307,12 +313,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       using OP = Belos::OperatorT<MV>;
 
       X->putScalar(zero);
-      RCP<OP> belosOp = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op));
-      RCP<OP> belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H));
+      RCP<OP> belosOp                                         = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op));
+      RCP<OP> belosPrec                                       = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));
       RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, RHS));
 
       std::string belosType = "Pseudoblock CG";
-      auto belosSolverList = rcpFromRef(belosParams.sublist(belosType));
+      auto belosSolverList  = rcpFromRef(belosParams.sublist(belosType));
 
       belosProblem->setRightPrec(belosPrec);
 
@@ -323,12 +329,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       // Create an iterative solver manager
       Belos::SolverFactory<Scalar, MV, OP> solverFactory;
-      RCP< Belos::SolverManager<Scalar, MV, OP> > solver = solverFactory.create(belosType, belosSolverList);
+      RCP<Belos::SolverManager<Scalar, MV, OP> > solver = solverFactory.create(belosType, belosSolverList);
       solver->setProblem(belosProblem);
 
       // Perform solve
       Belos::ReturnType ret = solver->solve();
-      int numIts = solver->getNumIters();
+      int numIts            = solver->getNumIters();
 
       // Get the number of iterations for this solve.
       out << "Number of iterations performed for this solve: " << numIts << std::endl;
@@ -343,8 +349,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       success &= (ret == Belos::Converged);
     }
-#endif // HAVE_MUELU_BELOS
-
+#endif  // HAVE_MUELU_BELOS
   }
 
   stacked_timer->stop("Hierarchical Driver");
@@ -353,13 +358,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   if (printTimings)
     stacked_timer->report(out, comm, options);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/research/graham/mf_example_01.cpp b/packages/muelu/research/graham/mf_example_01.cpp
index a689dae838e1..d06286f63d92 100644
--- a/packages/muelu/research/graham/mf_example_01.cpp
+++ b/packages/muelu/research/graham/mf_example_01.cpp
@@ -44,7 +44,7 @@
 //
 // @HEADER
 
-// Matrix-free example 01: Xpetra operator that generates 
+// Matrix-free example 01: Xpetra operator that generates
 // a simple tridiagonal finite difference poisson matrix
 //
 // This solves the problem f''(x) = h^2*(-pi^2)*sin(pi*x)
@@ -52,7 +52,6 @@
 // This supplies n global DOFs, and breaks it into
 // contiguous segments depending on the number of MPI ranks.
 
-
 // STL includes
 #include <iostream>
 #include <vector>
@@ -73,7 +72,7 @@
 #include "Kokkos_Core.hpp"
 //#include "kokkosTools.hpp"
 
-//Tpetra includes
+// Tpetra includes
 #include "Tpetra_Map.hpp"
 #include "Tpetra_MultiVector.hpp"
 #include "Tpetra_CrsMatrix.hpp"
@@ -81,7 +80,7 @@
 #include "Tpetra_Export.hpp"
 #include "MatrixMarket_Tpetra.hpp"
 
-//Xpetra includes
+// Xpetra includes
 #include "Xpetra_TpetraMultiVector.hpp"
 #include "Xpetra_Map.hpp"
 #include "Xpetra_MultiVector.hpp"
@@ -110,8 +109,8 @@
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosXpetraAdapter.hpp" // this header defines Belos::XpetraOp()
-#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
+#include "BelosXpetraAdapter.hpp"  // this header defines Belos::XpetraOp()
+#include "BelosMueLuAdapter.hpp"   // this header defines Belos::MueLuOp()
 #include "BelosTpetraAdapter.hpp"
 #include "BelosTpetraOperator.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
@@ -128,67 +127,67 @@
 #endif
 
 // some things like multivectors are "2D views" but only appear as 1D in practice, so we macro a print statement
-#define PRINT_VIEW2_LINEAR(view)                                                               \
-std::cout << #view << " (" << view.extent(0) << "," << view.extent(1) << ") = [" << std::endl;  \
-for(unsigned int i=0; i<view.extent(0); ++i)                                      \
-  for(unsigned int j=0; j<view.extent(1); ++j)                                    \
-    std::cout << view(i,j) << " ";                                                \
-std::cout << "]" << std::endl;
+#define PRINT_VIEW2_LINEAR(view)                                                                 \
+  std::cout << #view << " (" << view.extent(0) << "," << view.extent(1) << ") = [" << std::endl; \
+  for (unsigned int i = 0; i < view.extent(0); ++i)                                              \
+    for (unsigned int j = 0; j < view.extent(1); ++j)                                            \
+      std::cout << view(i, j) << " ";                                                            \
+  std::cout << "]" << std::endl;
 
 /**
- * This class defines an operator corresponding to the 
- * [-1 2 -1] tridiagonal matrix usually obtained from 
+ * This class defines an operator corresponding to the
+ * [-1 2 -1] tridiagonal matrix usually obtained from
  * a 1D finite difference discretization.
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 class TridiagonalOperator : public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-public:
+ public:
   // Xpetra::Operator subclasses should always define typedefs according to Xpetra
-  typedef typename Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::node_type            node_type;
-  typedef typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>                    MV;
-  typedef typename Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>                                    map_type;
-  typedef typename Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>                                 import_type;
-  typedef typename Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node>                                 export_type;
-public:
+  typedef typename Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::node_type node_type;
+  typedef typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef typename Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+  typedef typename Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> import_type;
+  typedef typename Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node> export_type;
+
+ public:
   /** Constructor
    * \param[in] n The number of global DOFs
    * \param[in] comm The Teuchos::Comm for the object
    */
   TridiagonalOperator(const GlobalOrdinal n,
-                      const Teuchos::RCP<const Teuchos::Comm<int> > comm)
-  {
+                      const Teuchos::RCP<const Teuchos::Comm<int>> comm) {
     TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), std::invalid_argument, "TridiagonalOperator constructor: The input Comm object must be nonnull.");
-    
-    const int my_rank = comm->getRank();
+
+    const int my_rank   = comm->getRank();
     const int num_procs = comm->getSize();
 
     // Construct a default map and let it choose how DOFs are divided
     // Note: This assumes the map constructor is generating contiguous local DOFs
     const GlobalOrdinal index_base = 0;
-    opMap_ = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, n, index_base, comm);
-    LocalOrdinal nlocal = opMap_->getLocalNumElements();
+    opMap_                         = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, n, index_base, comm);
+    LocalOrdinal nlocal            = opMap_->getLocalNumElements();
 
     // Ghosting: procs 0,1,...,n_p-1 are ordered left to right on [0,1]
-    if(my_rank > 0)
+    if (my_rank > 0)
       ++nlocal;
-    if(my_rank < num_procs - 1)
+    if (my_rank < num_procs - 1)
       ++nlocal;
 
-    // Construct a list of columns where this process has nonzero elements  
+    // Construct a list of columns where this process has nonzero elements
     // For this tridiagonal matrix, this is firstRowItOwns-1:lastRowItOwns+1
     std::vector<GlobalOrdinal> indices;
     indices.reserve(nlocal);
-    if(my_rank > 0)
+    if (my_rank > 0)
       indices.push_back(opMap_->getMinGlobalIndex() - 1);
-    for(GlobalOrdinal i = opMap_->getMinGlobalIndex(); i <= opMap_->getMaxGlobalIndex(); ++i)
+    for (GlobalOrdinal i = opMap_->getMinGlobalIndex(); i <= opMap_->getMaxGlobalIndex(); ++i)
       indices.push_back(i);
-    if(my_rank < num_procs - 1)
+    if (my_rank < num_procs - 1)
       indices.push_back(opMap_->getMaxGlobalIndex() + 1);
     Teuchos::ArrayView<const GlobalOrdinal> element_list(indices);
 
     // column Map for handling the redistribution
-    const GlobalOrdinal num_global_elements = n + 2*(num_procs - 1);
-    redistMap_ = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, num_global_elements, element_list, index_base, comm);
+    const GlobalOrdinal num_global_elements = n + 2 * (num_procs - 1);
+    redistMap_                              = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, num_global_elements, element_list, index_base, comm);
 
     // import object that describes how data will be redistributed
     importer_ = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(opMap_, redistMap_);
@@ -197,7 +196,7 @@ class TridiagonalOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Global
   // Required since we inherit from Xpetra::Operator
   // Destructor
   virtual ~TridiagonalOperator() {}
-  
+
   /**
    * \brief Compute Y := alpha Op X + beta Y.
    * \param[in] X Vector to apply the operator to
@@ -210,16 +209,15 @@ class TridiagonalOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Global
   apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
         Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
         Teuchos::ETransp mode = Teuchos::NO_TRANS,
-        Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-        Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const
-  {
+        Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+        Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
     // Setup: get comms, ranks, procs
-    Teuchos::RCP<const Teuchos::Comm<int> > comm = opMap_->getComm();
-    const int my_rank = comm->getRank();
-    const int num_procs = comm->getSize();
-    const size_t num_vecs = X.getNumVectors();
-    const LocalOrdinal numlocrows = static_cast<LocalOrdinal>(X.getLocalLength());
-    
+    Teuchos::RCP<const Teuchos::Comm<int>> comm = opMap_->getComm();
+    const int my_rank                           = comm->getRank();
+    const int num_procs                         = comm->getSize();
+    const size_t num_vecs                       = X.getNumVectors();
+    const LocalOrdinal numlocrows               = static_cast<LocalOrdinal>(X.getLocalLength());
+
     // Make a temporary multivector for holding the redistributed data and then redistribute
     Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> redistDataX = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(redistMap_, num_vecs);
     redistDataX->doImport(X, *importer_, Xpetra::INSERT);
@@ -230,42 +228,41 @@ class TridiagonalOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Global
 
     // Perform the matvec with the locally owned data
     // For each column
-    for(size_t c = 0; c < num_vecs; ++c) {
+    for (size_t c = 0; c < num_vecs; ++c) {
       LocalOrdinal offset;
       // On ranks greater than 0, we need the ghosted X values for the computation
       // Y[0,c] = beta*Y[0,c] + alpha*(-colViewX[0] + 2*colViewX[1] - colViewX[2])
-      if(my_rank > 0) {
-        KokkosViewY(0,c) = beta*KokkosViewY(0, c) + alpha*(-KokkosViewX(0, c) + 2*KokkosViewX(1, c) - KokkosViewX(2, c));
-        offset = 0;
+      if (my_rank > 0) {
+        KokkosViewY(0, c) = beta * KokkosViewY(0, c) + alpha * (-KokkosViewX(0, c) + 2 * KokkosViewX(1, c) - KokkosViewX(2, c));
+        offset            = 0;
       }
       // On rank 0, we only have two entries in the first row
       // Y[0,c] = beta*Y[0,c] + alpha*(2*colViewX[1] - colViewX[2])
       else {
-        KokkosViewY(0,c) = beta*KokkosViewY(0, c) + alpha*(2*KokkosViewX(0, c) - KokkosViewX(1, c));
-        offset = 1;
+        KokkosViewY(0, c) = beta * KokkosViewY(0, c) + alpha * (2 * KokkosViewX(0, c) - KokkosViewX(1, c));
+        offset            = 1;
       }
       // For all other rows, we need the full stencil
       // Y[r,c] = beta*Y[r,c] + alpha*(-colViewX[r-offset] + 2*colViewX[r+1-offset] - colViewX[r+2-offset])
-      for(LocalOrdinal r = 1; r < numlocrows - 1; ++r) {
-        const Scalar newVal = beta*KokkosViewY(r, c) + 
-                                   alpha*(-KokkosViewX(r-offset, c) + 2*KokkosViewX(r+1-offset, c) - KokkosViewX(r+2-offset, c));
-        KokkosViewY(r,c) = newVal;
+      for (LocalOrdinal r = 1; r < numlocrows - 1; ++r) {
+        const Scalar newVal = beta * KokkosViewY(r, c) +
+                              alpha * (-KokkosViewX(r - offset, c) + 2 * KokkosViewX(r + 1 - offset, c) - KokkosViewX(r + 2 - offset, c));
+        KokkosViewY(r, c) = newVal;
       }
       // On ranks other than the last rank, we need the ghosted X values for the computation
       // Y[numlocrows-1,c] = beta*Y[numlocrows-1,c] + alpha*(-colViewX[numlocrows-1-offset] + 2*colViewX[numlocrows-offset]
       //                   - colViewX[numlocrows+1-offset])
-      if(my_rank < num_procs - 1) {
-        const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) + 
-                                   alpha*(-KokkosViewX(numlocrows-1-offset, c) + 2*KokkosViewX(numlocrows-offset, c)
-                                          - KokkosViewX(numlocrows+1-offset, c));
-        KokkosViewY(numlocrows-1,c) = newVal;
+      if (my_rank < num_procs - 1) {
+        const Scalar newVal = beta * KokkosViewY(numlocrows - 1, c) +
+                              alpha * (-KokkosViewX(numlocrows - 1 - offset, c) + 2 * KokkosViewX(numlocrows - offset, c) - KokkosViewX(numlocrows + 1 - offset, c));
+        KokkosViewY(numlocrows - 1, c) = newVal;
       }
       // On the last rank, we only have two entries in the last row
       // Y[numlocrows-1,c] = beta*Y[numlocrows-1,c] + alpha*(-colViewX[numlocrows-1-offset] + 2*colViewX[numlocrows-offset])
       else {
-        const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) + 
-          alpha*(-KokkosViewX(numlocrows-1-offset, c) + 2*KokkosViewX(numlocrows-offset, c));
-        KokkosViewY(numlocrows-1,c) = newVal;
+        const Scalar newVal = beta * KokkosViewY(numlocrows - 1, c) +
+                              alpha * (-KokkosViewX(numlocrows - 1 - offset, c) + 2 * KokkosViewX(numlocrows - offset, c));
+        KokkosViewY(numlocrows - 1, c) = newVal;
       }
     }
   }
@@ -280,46 +277,45 @@ class TridiagonalOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Global
   bool hasTransposeApply() const { return true; }
 
   //! Compute a residual R = B - (*this) * X
-  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & X,
-                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & B,
-                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & R) const {
+  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const {
     typedef Teuchos::ScalarTraits<Scalar> STS;
-    R.update(STS::one(), B, STS::zero()); // R = 1*B + 0*R
-    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one()); // R = R - Op*X
+    R.update(STS::one(), B, STS::zero());                           // R = 1*B + 0*R
+    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());  // R = R - Op*X
   }
 
-private:
+ private:
   Teuchos::RCP<const map_type> opMap_, redistMap_;
   Teuchos::RCP<const import_type> importer_;
 };
 
-
 /**
- * This class defines an operator corresponding to the 
+ * This class defines an operator corresponding to the
  * [1 2 1] interpolation stencil. For interpolation to
  * behave correctly, each MPI rank should have a
- * number of DOFs that is divisible by 3, as the 
+ * number of DOFs that is divisible by 3, as the
  * coarsening ratio is 3->1. Communication is ignored.
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-public:
+ public:
   // Xpetra::Operator subclasses should always define typedefs according to Xpetra
-  typedef typename Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::node_type            node_type;
-  typedef typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>                    MV;
-  typedef typename Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>                                    map_type;
-  typedef typename Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>                                 import_type;
-  typedef typename Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node>                                 export_type;
-public:
+  typedef typename Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::node_type node_type;
+  typedef typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef typename Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+  typedef typename Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> import_type;
+  typedef typename Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node> export_type;
+
+ public:
   /** Constructor
    * \param[in] n The number of global DOFs
    * \param[in] comm The Teuchos::Comm for the object
    */
-  MFProlongatorOperator(const Teuchos::RCP<const map_type> fine_map)
-  {
-    const GlobalOrdinal n = fine_map->getGlobalNumElements();
-    const LocalOrdinal n_local = fine_map->getLocalNumElements();
-    const Teuchos::RCP<const Teuchos::Comm<int> > comm = fine_map->getComm();
+  MFProlongatorOperator(const Teuchos::RCP<const map_type> fine_map) {
+    const GlobalOrdinal n                             = fine_map->getGlobalNumElements();
+    const LocalOrdinal n_local                        = fine_map->getLocalNumElements();
+    const Teuchos::RCP<const Teuchos::Comm<int>> comm = fine_map->getComm();
 
     TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), std::invalid_argument, "MFProlongatorOperator constructor: The input Comm object must be nonnull.");
     TEUCHOS_TEST_FOR_EXCEPTION(n_local % 3 != 0, std::invalid_argument, "MFProlongatorOperator constructor: The number of local DOFs is not divisible by 3.");
@@ -327,16 +323,16 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
     // Construct a default map and let it choose how DOFs are divided
     // Note: This assumes the map constructor is generating contiguous local DOFs
     const GlobalOrdinal index_base = 0;
-    rangeMap_ = fine_map;
-    domainMap_ = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, n/3, index_base, comm);
+    rangeMap_                      = fine_map;
+    domainMap_                     = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, n / 3, index_base, comm);
 
-    TEUCHOS_TEST_FOR_EXCEPTION(domainMap_->getLocalNumElements() != ((size_t) n_local)/3, std::invalid_argument, "MFProlongatorOperator constructor: The operator is not coarsening by 3.");
+    TEUCHOS_TEST_FOR_EXCEPTION(domainMap_->getLocalNumElements() != ((size_t)n_local) / 3, std::invalid_argument, "MFProlongatorOperator constructor: The operator is not coarsening by 3.");
   };
 
   // Required since we inherit from Xpetra::Operator
   // Destructor
   virtual ~MFProlongatorOperator() {}
-  
+
   /**
    * \brief Compute Y := alpha Op X + beta Y.
    * \param[in] X Vector to apply the operator to
@@ -349,16 +345,15 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
   apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
         Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
         Teuchos::ETransp mode = Teuchos::NO_TRANS,
-        Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-        Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const
-  {
+        Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+        Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {
     // // Setup: get comms, ranks, procs
     // Teuchos::RCP<const Teuchos::Comm<int> > comm = domainMap_->getComm();
     // const int my_rank = comm->getRank();
     // const int num_procs = comm->getSize();
     // const size_t num_vecs = X.getNumVectors();
     // const LocalOrdinal numlocrows = static_cast<LocalOrdinal>(X.getLocalLength());
-    
+
     // // Make a temporary multivector for holding the redistributed data and then redistribute
     // Teuchos::RCP<MV> redistDataX = Teuchos::rcp(new MV(redistMap_, num_vecs));
     // redistDataX->doImport(X, *importer_, Xpetra::INSERT);
@@ -386,7 +381,7 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
     //   // For all other rows, we need the full stencil
     //   // Y[r,c] = beta*Y[r,c] + alpha*(-colViewX[r-offset] + 2*colViewX[r+1-offset] - colViewX[r+2-offset])
     //   for(LocalOrdinal r = 1; r < numlocrows - 1; ++r) {
-    //     const Scalar newVal = beta*KokkosViewY(r, c) + 
+    //     const Scalar newVal = beta*KokkosViewY(r, c) +
     //                                alpha*(-KokkosViewX(r-offset, c) + 2*KokkosViewX(r+1-offset, c) - KokkosViewX(r+2-offset, c));
     //     KokkosViewY(r,c) = newVal;
     //   }
@@ -394,7 +389,7 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
     //   // Y[numlocrows-1,c] = beta*Y[numlocrows-1,c] + alpha*(-colViewX[numlocrows-1-offset] + 2*colViewX[numlocrows-offset]
     //   //                   - colViewX[numlocrows+1-offset])
     //   if(my_rank < num_procs - 1) {
-    //     const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) + 
+    //     const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) +
     //                                alpha*(-KokkosViewX(numlocrows-1-offset, c) + 2*KokkosViewX(numlocrows-offset, c)
     //                                       - KokkosViewX(numlocrows+1-offset, c));
     //     KokkosViewY(numlocrows-1,c) = newVal;
@@ -402,7 +397,7 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
     //   // On the last rank, we only have two entries in the last row
     //   // Y[numlocrows-1,c] = beta*Y[numlocrows-1,c] + alpha*(-colViewX[numlocrows-1-offset] + 2*colViewX[numlocrows-offset])
     //   else {
-    //     const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) + 
+    //     const Scalar newVal = beta*KokkosViewY(numlocrows-1, c) +
     //       alpha*(-KokkosViewX(numlocrows-1-offset, c) + 2*KokkosViewX(numlocrows-offset, c));
     //     KokkosViewY(numlocrows-1,c) = newVal;
     //   }
@@ -419,43 +414,41 @@ class MFProlongatorOperator : public Xpetra::Operator<Scalar, LocalOrdinal, Glob
   bool hasTransposeApply() const { return true; }
 
   //! Compute a residual R = B - (*this) * X
-  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & X,
-                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & B,
-                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & R) const {
-    //throw Exceptions::RuntimeError("Interface not supported");
+  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const {
+    // throw Exceptions::RuntimeError("Interface not supported");
   }
 
-private:
+ private:
   Teuchos::RCP<const map_type> domainMap_, rangeMap_, redistMap_;
   Teuchos::RCP<const import_type> importer_;
 };
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[])
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
   Teuchos::RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
-  const int my_rank = comm->getRank();
-  const int num_procs = comm->getSize();
+  const int my_rank                           = comm->getRank();
+  const int num_procs                         = comm->getSize();
   {
     // Necessary typedefs
     using SC = Scalar;
     using LO = LocalOrdinal;
     using GO = GlobalOrdinal;
     using NO = Node;
-    //using map_type = Xpetra::Map<>; // unused
-    using MV = Xpetra::MultiVector<SC,LO,GO,NO>;
+    // using map_type = Xpetra::Map<>; // unused
+    using MV = Xpetra::MultiVector<SC, LO, GO, NO>;
 
     // Set a command line processor and parse it
-    int n = 300;
-    int max_iterations = 1000;
-    double tol = 1e-10;
-    bool do_multigrid = true;
-    bool show_timer_summary = false;
-    bool belos_verbose = false;
-    bool show_kokkos = false; 
+    int n                       = 300;
+    int max_iterations          = 1000;
+    double tol                  = 1e-10;
+    bool do_multigrid           = true;
+    bool show_timer_summary     = false;
+    bool belos_verbose          = false;
+    bool show_kokkos            = false;
     bool print_RHS_and_solution = false;
-    bool dump_matrix_market = false;
+    bool dump_matrix_market     = false;
     // problem config
     clp.setOption("n", &n, "Size of the n-by-n operator (default: 300)");
     clp.setOption("verbose", "no-verbose", &belos_verbose, "Use Belos verbose output (default: false)");
@@ -472,36 +465,36 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     clp.recogniseAllOptions(true);
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: 
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
         return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
-      case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: 
+      case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
         return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:
         break;
     }
-    if(my_rank == 0)
-      std::cout << "Running example-01 with n=" << n << " verbose=" << belos_verbose << " config=" << show_kokkos << "..." << std::endl; 
+    if (my_rank == 0)
+      std::cout << "Running example-01 with n=" << n << " verbose=" << belos_verbose << " config=" << show_kokkos << "..." << std::endl;
 
     // print configuration details if needed
-    Kokkos::Serial().print_configuration(std::cout, true/*details*/);
-    //Kokkos::OpenMP().print_configuration(std::cout, true/*details*/);
-    //std::cout << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
-    //Kokkos::Cuda().print_configuration(std::cout, true/*details*/);
-    //Kokkos::HIP().print_configuration(std::cout, true/*details*/);
+    Kokkos::Serial().print_configuration(std::cout, true /*details*/);
+    // Kokkos::OpenMP().print_configuration(std::cout, true/*details*/);
+    // std::cout << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
+    // Kokkos::Cuda().print_configuration(std::cout, true/*details*/);
+    // Kokkos::HIP().print_configuration(std::cout, true/*details*/);
 
     // Create the operator
-    Teuchos::RCP<TridiagonalOperator<SC,LO,GO,NO>> matrix = Teuchos::rcp(new TridiagonalOperator<SC,LO,GO,NO>(n, comm));
+    Teuchos::RCP<TridiagonalOperator<SC, LO, GO, NO>> matrix = Teuchos::rcp(new TridiagonalOperator<SC, LO, GO, NO>(n, comm));
 
     // A useful class name means excruciating template arguments... but I want to avoid headaches
     std::cout << "Creating the vectors..." << std::endl;
-    
+
     // Construct the right-hand side
-    Teuchos::RCP<MV> rhs = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(matrix->getRangeMap(),1);
+    Teuchos::RCP<MV> rhs = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(matrix->getRangeMap(), 1);
     rhs->putScalar(0.0);
 
     // Construct the initial guess (seedrandom is not always reproducible across machines, see Github)
-    Teuchos::RCP<MV> solution = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(matrix->getDomainMap(),1);
+    Teuchos::RCP<MV> solution = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(matrix->getDomainMap(), 1);
     Teuchos::ScalarTraits<SC>::seedrandom(314159);
     solution->randomize();
     // solution->putScalar(1.0) is an alternative if the random seed isn't reproducible
@@ -510,50 +503,51 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // f''(x) = h^2*(-pi^2)*sin(pi*x)
     {
       // divide domain up according to MPI ranks
-      const SC dx = 1.0/num_procs;
-      const SC x_left = dx*my_rank;
-      const SC x_right = dx*(my_rank+1);
+      const SC dx          = 1.0 / num_procs;
+      const SC x_left      = dx * my_rank;
+      const SC x_right     = dx * (my_rank + 1);
       const size_t n_local = matrix->getRangeMap()->getLocalNumElements();
-      const SC h = (x_right - x_left)/(n_local-1);
+      const SC h           = (x_right - x_left) / (n_local - 1);
 
       // fill the RHS appropriately
       auto rhs_2d = rhs->getHostLocalView(Xpetra::Access::OverwriteAll);
-      auto rhs_1d = Kokkos::subview (rhs_2d, Kokkos::ALL(), 0);
-      SC x = x_left;
-      for(size_t i=0; i<n_local; ++i) {
-        rhs_1d(i) = 4*h*h*M_PI*M_PI*sin(2*M_PI*x);
+      auto rhs_1d = Kokkos::subview(rhs_2d, Kokkos::ALL(), 0);
+      SC x        = x_left;
+      for (size_t i = 0; i < n_local; ++i) {
+        rhs_1d(i) = 4 * h * h * M_PI * M_PI * sin(2 * M_PI * x);
         x += h;
       }
     }
 
-    if(!do_multigrid) {
+    if (!do_multigrid) {
       std::cout << "Creating Belos solver..." << std::endl;
 
       // Create solver settings
       Teuchos::RCP<Teuchos::ParameterList> belos_settings = Teuchos::parameterList("Belos");
       belos_settings->set("Verbosity", belos_verbose ? 1 : 0);
       belos_settings->set("Output Style", 1);
-      belos_settings->set("Maximum Iterations",    max_iterations);
-      //belos_settings->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-      //belos_settings->set("Output Frequency",      1);
-      //belos_settings->set("Output Style",          Belos::Brief);
+      belos_settings->set("Maximum Iterations", max_iterations);
+      // belos_settings->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
+      // belos_settings->set("Output Frequency",      1);
+      // belos_settings->set("Output Style",          Belos::Brief);
 
       // Define and set the linear problem
-      Teuchos::RCP<Belos::OperatorT<MV>> belos_operator = Teuchos::rcp(new Belos::XpetraOp<SC,LO,GO,NO>(matrix)); // Turns a Xpetra::Operator object into a Belos operator
-      Teuchos::RCP<Belos::LinearProblem<SC,MV,Belos::OperatorT<MV>>> belos_problem = Teuchos::rcp(new Belos::LinearProblem<SC,MV,Belos::OperatorT<MV>>(belos_operator, solution, rhs));
-      bool set = belos_problem->setProblem();
+      Teuchos::RCP<Belos::OperatorT<MV>> belos_operator                              = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(matrix));  // Turns a Xpetra::Operator object into a Belos operator
+      Teuchos::RCP<Belos::LinearProblem<SC, MV, Belos::OperatorT<MV>>> belos_problem = Teuchos::rcp(new Belos::LinearProblem<SC, MV, Belos::OperatorT<MV>>(belos_operator, solution, rhs));
+      bool set                                                                       = belos_problem->setProblem();
       if (set == false) {
         if (comm->getRank() == 0)
-          std::cout << std::endl << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
+          std::cout << std::endl
+                    << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
         return EXIT_FAILURE;
       }
 
       // Solve the problem
       std::cout << "Solving the problem..." << std::endl;
-      Teuchos::RCP<Belos::SolverManager<SC,MV,Belos::OperatorT<MV>>> solver = Teuchos::rcp(new Belos::BlockCGSolMgr<SC,MV,Belos::OperatorT<MV>>(belos_problem, belos_settings));
-      const Belos::ReturnType belos_result = solver->solve();
+      Teuchos::RCP<Belos::SolverManager<SC, MV, Belos::OperatorT<MV>>> solver = Teuchos::rcp(new Belos::BlockCGSolMgr<SC, MV, Belos::OperatorT<MV>>(belos_problem, belos_settings));
+      const Belos::ReturnType belos_result                                    = solver->solve();
 
-      if(my_rank == 0) {
+      if (my_rank == 0) {
         std::cout << "Belos solver wrapper results: "
                   << (belos_result == Belos::Converged ? "Converged" : "Unconverged")
                   << std::endl
@@ -567,30 +561,30 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       params.set("coarse: max size", 1);
       params.set("max levels", 2);
       params.set("transpose: use implicit", true);
-      
+
       // generate coarse matrix-free operator
-      Teuchos::RCP<TridiagonalOperator<SC,LO,GO,NO>> coarse_matrix = Teuchos::rcp(new TridiagonalOperator<SC,LO,GO,NO>(n/3, comm));
-      Teuchos::RCP<MFProlongatorOperator<SC,LO,GO,NO>> P = Teuchos::rcp(new MFProlongatorOperator<SC,LO,GO,NO>(matrix->getDomainMap()));
+      Teuchos::RCP<TridiagonalOperator<SC, LO, GO, NO>> coarse_matrix = Teuchos::rcp(new TridiagonalOperator<SC, LO, GO, NO>(n / 3, comm));
+      Teuchos::RCP<MFProlongatorOperator<SC, LO, GO, NO>> P           = Teuchos::rcp(new MFProlongatorOperator<SC, LO, GO, NO>(matrix->getDomainMap()));
 
       // create MueLu hierarchy and levels
-      Teuchos::RCP<MueLu::Hierarchy<SC,LO,GO,NO>> hierarchy = Teuchos::rcp(new typename MueLu::Hierarchy<SC,LO,GO,NO>());
+      Teuchos::RCP<MueLu::Hierarchy<SC, LO, GO, NO>> hierarchy = Teuchos::rcp(new typename MueLu::Hierarchy<SC, LO, GO, NO>());
       hierarchy->SetProcRankVerbose(matrix->getDomainMap()->getComm()->getRank());
-      Teuchos::RCP<MueLu::HierarchyManager<SC,LO,GO,NO>> hierarchyManager = Teuchos::rcp(new MueLu::ParameterListInterpreter<SC,LO,GO,NO>(params,matrix->getDomainMap()->getComm()));
-      Teuchos::RCP<MueLu::FactoryManager<SC,LO,GO,NO>> factoryManager = Teuchos::rcp(new MueLu::FactoryManager<SC,LO,GO,NO>());
-      
+      Teuchos::RCP<MueLu::HierarchyManager<SC, LO, GO, NO>> hierarchyManager = Teuchos::rcp(new MueLu::ParameterListInterpreter<SC, LO, GO, NO>(params, matrix->getDomainMap()->getComm()));
+      Teuchos::RCP<MueLu::FactoryManager<SC, LO, GO, NO>> factoryManager     = Teuchos::rcp(new MueLu::FactoryManager<SC, LO, GO, NO>());
+
       // set A on fine level
       Teuchos::RCP<MueLu::Level> fineLevel = hierarchy->GetLevel(0);
       fineLevel->SetFactoryManager(factoryManager);
-      Teuchos::RCP<Xpetra::Operator<SC,LO,GO,NO>> matrix_op = matrix;
+      Teuchos::RCP<Xpetra::Operator<SC, LO, GO, NO>> matrix_op = matrix;
       fineLevel->Set("A", matrix_op);
-      
+
       // handle the smoother
       std::string ifpackType = "RELAXATION";
       Teuchos::ParameterList ifpackList;
-      ifpackList.set("relaxation: sweeps", (LO) 1);
-      ifpackList.set("relaxation: damping factor", (SC) 1.0);
-      Teuchos::RCP<MueLu::SmootherPrototype<SC,LO,GO,NO>> smootherPrototype = Teuchos::rcp(new MueLu::TrilinosSmoother<SC,LO,GO,NO>(ifpackType, ifpackList));
-      Teuchos::RCP<MueLu::SmootherFactory<SC,LO,GO,NO>> smootherFact = Teuchos::rcp(new MueLu::SmootherFactory<SC,LO,GO,NO>(smootherPrototype));
+      ifpackList.set("relaxation: sweeps", (LO)1);
+      ifpackList.set("relaxation: damping factor", (SC)1.0);
+      Teuchos::RCP<MueLu::SmootherPrototype<SC, LO, GO, NO>> smootherPrototype = Teuchos::rcp(new MueLu::TrilinosSmoother<SC, LO, GO, NO>(ifpackType, ifpackList));
+      Teuchos::RCP<MueLu::SmootherFactory<SC, LO, GO, NO>> smootherFact        = Teuchos::rcp(new MueLu::SmootherFactory<SC, LO, GO, NO>(smootherPrototype));
       factoryManager->SetFactory("Smoother", smootherFact);
 
       // work on the next level
@@ -600,38 +594,37 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // coarseLevel->Set("A", coarse_matrix);
 
       // hierarchyManager->SetupHierarchy(*hierarchy);
-      //hierarchy->Setup(factoryManager,fineLevel,coarseLevel);
-      //hierarchy->setlib(matrix->getDomainMap()->lib());
-      //hierarchyManager->SetupHierarchy(*hierarchy);
+      // hierarchy->Setup(factoryManager,fineLevel,coarseLevel);
+      // hierarchy->setlib(matrix->getDomainMap()->lib());
+      // hierarchyManager->SetupHierarchy(*hierarchy);
 
       std::cout << "Finished hierarchy!" << std::endl;
 
       Teuchos::ParameterList status;
-      //status = hierarchy->FullPopulate(PRfact,Acfact,SmooFact,0,maxLevels);
+      // status = hierarchy->FullPopulate(PRfact,Acfact,SmooFact,0,maxLevels);
       if (comm->getRank() == 0) {
-        std::cout  << "======================\n Multigrid statistics \n======================" << std::endl;
+        std::cout << "======================\n Multigrid statistics \n======================" << std::endl;
         status.print(std::cout, Teuchos::ParameterList::PrintOptions().indent(2));
       }
 
-      //hierarchy->Iterate(*rhs, *solution, max_iterations);
+      // hierarchy->Iterate(*rhs, *solution, max_iterations);
     }
 
     // output the RHS and solution for validation
     // (sleep my_rank is very hacky here, but keeps prints contiguous)
-    if(print_RHS_and_solution) {
+    if (print_RHS_and_solution) {
       sleep(my_rank);
       auto rhs_2d = rhs->getHostLocalView(Xpetra::Access::ReadOnly);
       PRINT_VIEW2_LINEAR(rhs_2d)
       auto solution_2d = solution->getHostLocalView(Xpetra::Access::ReadOnly);
-      PRINT_VIEW2_LINEAR(solution_2d) 
+      PRINT_VIEW2_LINEAR(solution_2d)
     }
-    
+
     // can't dump MatrixMarket if it's not a matrix
-    if(dump_matrix_market) {
-      //Xpetra::MatrixMarket::Writer<vector_type>::writeDenseFile("example_01_solution.mm", *solution);
-      //Xpetra::MatrixMarket::Writer<vector_type>::writeDenseFile("example_01_rhs.mm", *rhs);
+    if (dump_matrix_market) {
+      // Xpetra::MatrixMarket::Writer<vector_type>::writeDenseFile("example_01_solution.mm", *solution);
+      // Xpetra::MatrixMarket::Writer<vector_type>::writeDenseFile("example_01_rhs.mm", *rhs);
     }
-
   }
   Kokkos::finalize();
   return 0;
@@ -641,6 +634,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/research/luc/region_algorithms/Driver.cpp b/packages/muelu/research/luc/region_algorithms/Driver.cpp
index 5a903f5ca8ea..6e9251a3dac0 100644
--- a/packages/muelu/research/luc/region_algorithms/Driver.cpp
+++ b/packages/muelu/research/luc/region_algorithms/Driver.cpp
@@ -50,7 +50,7 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 
 #include <MueLu.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 #include <MueLu_Utilities.hpp>
 #include <MueLu_AmalgamationFactory.hpp>
 #include <MueLu_CoalesceDropFactory.hpp>
@@ -65,25 +65,27 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void createTwoLevelHierarchy(MueLu::Level& fineLevel, MueLu::Level& coarseLevel,
                              Teuchos::RCP<typename Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A);
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::Array; using Teuchos::ArrayView;
-  using Teuchos::RCP; using Teuchos::rcp;
-  using Teuchos::tuple;
+  using Teuchos::Array;
+  using Teuchos::ArrayView;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
+  using Teuchos::tuple;
 
-  typedef Tpetra::Map<LO,GO,NO> map_type;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO> crs_matrix_type;
+  typedef Tpetra::Map<LO, GO, NO> map_type;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> crs_matrix_type;
 
   typedef typename crs_matrix_type::scalar_type scalar_type;
   typedef typename crs_matrix_type::local_ordinal_type local_ordinal_type;
@@ -95,8 +97,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool verbose = true;
 
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    const local_ordinal_type myRank = comm->getRank();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    const local_ordinal_type myRank     = comm->getRank();
 
     // Manage the way output stream works
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
@@ -111,24 +113,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Parameters initialization
     // =========================================================================
 
-    //GO nx = 100, ny = 100, nz = 100;
-    //Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    // GO nx = 100, ny = 100, nz = 100;
+    // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
 
-    std::string xmlFileName = "driver.xml";      clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+    std::string xmlFileName = "driver.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     global_ordinal_type gNumCompFineGIDs = 10, lCompFineGIDOffset = 0;
-    global_ordinal_type lRegFineGIDOffset = 0; // gNumRegFineGIDs  = 11,
-    local_ordinal_type  lNumCompFineGIDs =  0, lNumRegFineGIDs = 0;
-    local_ordinal_type lNumRegCoarseGIDs = 0; // lNumCompCoarseGIDs =  0,
-    if(myRank == 0) {
+    global_ordinal_type lRegFineGIDOffset = 0;  // gNumRegFineGIDs  = 11,
+    local_ordinal_type lNumCompFineGIDs = 0, lNumRegFineGIDs = 0;
+    local_ordinal_type lNumRegCoarseGIDs = 0;  // lNumCompCoarseGIDs =  0,
+    if (myRank == 0) {
       lNumCompFineGIDs = 7;
       lNumRegFineGIDs  = 7;
 
@@ -136,8 +139,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       lRegFineGIDOffset  = 0;
 
       // lNumCompCoarseGIDs = 3;
-      lNumRegCoarseGIDs  = 3;
-    } else if(myRank == 1) {
+      lNumRegCoarseGIDs = 3;
+    } else if (myRank == 1) {
       lNumCompFineGIDs = 3;
       lNumRegFineGIDs  = 4;
 
@@ -145,13 +148,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       lRegFineGIDOffset  = 6;
 
       // lNumCompCoarseGIDs = 1;
-      lNumRegCoarseGIDs  = 2;
+      lNumRegCoarseGIDs = 2;
     }
 
     // The initial focus is on getting rowMaps as they are also used as rangeMap and potentially
     // domainMaps...
     Array<global_ordinal_type> fineCompRowGIDs(lNumCompFineGIDs);
-    for(local_ordinal_type dof = 0; dof < lNumCompFineGIDs; ++dof) {
+    for (local_ordinal_type dof = 0; dof < lNumCompFineGIDs; ++dof) {
       fineCompRowGIDs[dof] = lCompFineGIDOffset + dof;
     }
     out << "fineCompRowGIDs: " << fineCompRowGIDs << std::endl;
@@ -159,12 +162,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Now the columnMap for the initial composite operator is needed
     Array<global_ordinal_type> fineCompColGIDs(lNumCompFineGIDs + 1);
-    if(myRank == 0) {
-      for(local_ordinal_type dof = 0; dof < lNumCompFineGIDs + 1; ++dof) {
+    if (myRank == 0) {
+      for (local_ordinal_type dof = 0; dof < lNumCompFineGIDs + 1; ++dof) {
         fineCompColGIDs[dof] = lCompFineGIDOffset + dof;
       }
-    } else if(myRank == 1) {
-      for(local_ordinal_type dof = 0; dof < lNumCompFineGIDs + 1; ++dof) {
+    } else if (myRank == 1) {
+      for (local_ordinal_type dof = 0; dof < lNumCompFineGIDs + 1; ++dof) {
         fineCompColGIDs[dof] = lCompFineGIDOffset - 1 + dof;
       }
     }
@@ -172,40 +175,41 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<map_type> fineCompColMap = rcp(new map_type(12, fineCompColGIDs, 0, comm));
 
     // Create the matrix using our maps and assuming at most 3 entries per row
-    RCP<crs_matrix_type> compA (new crs_matrix_type (fineCompRowMap, fineCompColMap, 3));
+    RCP<crs_matrix_type> compA(new crs_matrix_type(fineCompRowMap, fineCompColMap, 3));
 
     // Now each row needs to be filled
-    const scalar_type one = static_cast<scalar_type> (1.0);
-    const scalar_type two = static_cast<scalar_type> (2.0);
-    const scalar_type negOne = static_cast<scalar_type> (-1.0);
-    for(local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumCompFineGIDs); ++lclRow) {
-      const global_ordinal_type gblRow = fineCompRowMap->getGlobalElement (lclRow);
+    const scalar_type one    = static_cast<scalar_type>(1.0);
+    const scalar_type two    = static_cast<scalar_type>(2.0);
+    const scalar_type negOne = static_cast<scalar_type>(-1.0);
+    for (local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumCompFineGIDs); ++lclRow) {
+      const global_ordinal_type gblRow = fineCompRowMap->getGlobalElement(lclRow);
       // A(0, 0) = [1]
       if (gblRow == 0) {
-        compA->insertGlobalValues (gblRow,
-                                   tuple<global_ordinal_type> (gblRow),
-                                   tuple<scalar_type> (one));
+        compA->insertGlobalValues(gblRow,
+                                  tuple<global_ordinal_type>(gblRow),
+                                  tuple<scalar_type>(one));
       }
       // A(N-1, N-2:N-1) = [-1, 1]
       else if (gblRow == gNumCompFineGIDs - 1) {
-        compA->insertGlobalValues (gblRow,
-                                   tuple<global_ordinal_type> (gblRow - 1, gblRow),
-                                   tuple<scalar_type> (negOne, one));
+        compA->insertGlobalValues(gblRow,
+                                  tuple<global_ordinal_type>(gblRow - 1, gblRow),
+                                  tuple<scalar_type>(negOne, one));
       }
       // A(i, i-1:i+1) = [-1, 2, -1]
       else {
-        compA->insertGlobalValues (gblRow,
-                                   tuple<global_ordinal_type> (gblRow - 1, gblRow, gblRow + 1),
-                                   tuple<scalar_type> (negOne, two, negOne));
+        compA->insertGlobalValues(gblRow,
+                                  tuple<global_ordinal_type>(gblRow - 1, gblRow, gblRow + 1),
+                                  tuple<scalar_type>(negOne, two, negOne));
       }
     }
-    compA->fillComplete (fineCompRowMap, fineCompRowMap);
+    compA->fillComplete(fineCompRowMap, fineCompRowMap);
     // compA->print(out);
 
-    out << std::endl << "Now switching to the region matrix" << std::endl;
+    out << std::endl
+        << "Now switching to the region matrix" << std::endl;
 
-    Array<global_ordinal_type> fineRegRowGIDs (lNumRegFineGIDs);
-    for(local_ordinal_type dof = 0; dof < lNumRegFineGIDs; ++dof) {
+    Array<global_ordinal_type> fineRegRowGIDs(lNumRegFineGIDs);
+    for (local_ordinal_type dof = 0; dof < lNumRegFineGIDs; ++dof) {
       fineRegRowGIDs[dof] = (lRegFineGIDOffset + dof == 6 && myRank == 1) ? 10 : lRegFineGIDOffset + dof;
     }
     out << "fineRegRowGIDs: " << fineRegRowGIDs << std::endl;
@@ -213,116 +217,119 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Create the matrix using the simplest constructor as we assume that the region matrix
     // has now off processor entries, hence no column map is required.
-    RCP<crs_matrix_type> regA (new crs_matrix_type (fineRegRowMap, 3));
+    RCP<crs_matrix_type> regA(new crs_matrix_type(fineRegRowMap, 3));
 
     // Now each row needs to be filled
-    for(local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumRegFineGIDs); ++lclRow) {
-      const global_ordinal_type prevGblRow = (lclRow > 0) ? fineRegRowMap->getGlobalElement (lclRow - 1) : -1;
-      const global_ordinal_type curGblRow  = fineRegRowMap->getGlobalElement (lclRow);
-      const global_ordinal_type nextGblRow = (lclRow < lNumRegFineGIDs - 1) ? fineRegRowMap->getGlobalElement (lclRow + 1) : -1;
+    for (local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumRegFineGIDs); ++lclRow) {
+      const global_ordinal_type prevGblRow = (lclRow > 0) ? fineRegRowMap->getGlobalElement(lclRow - 1) : -1;
+      const global_ordinal_type curGblRow  = fineRegRowMap->getGlobalElement(lclRow);
+      const global_ordinal_type nextGblRow = (lclRow < lNumRegFineGIDs - 1) ? fineRegRowMap->getGlobalElement(lclRow + 1) : -1;
       // A(0, 0) = [1]
       if (lclRow == 0) {
-        if (curGblRow == 0) { // Apply Dirichlet BC to the left of the mesh
-          regA->insertGlobalValues (curGblRow,
-                                    tuple<global_ordinal_type> (curGblRow),
-                                    tuple<scalar_type> (one));
-        } else { // Here I hard code the column associated with "-1" because I'm lazy...
-          regA->insertGlobalValues (curGblRow,
-                                    tuple<global_ordinal_type> (curGblRow, nextGblRow),
-                                    tuple<scalar_type> (one, negOne));
+        if (curGblRow == 0) {  // Apply Dirichlet BC to the left of the mesh
+          regA->insertGlobalValues(curGblRow,
+                                   tuple<global_ordinal_type>(curGblRow),
+                                   tuple<scalar_type>(one));
+        } else {  // Here I hard code the column associated with "-1" because I'm lazy...
+          regA->insertGlobalValues(curGblRow,
+                                   tuple<global_ordinal_type>(curGblRow, nextGblRow),
+                                   tuple<scalar_type>(one, negOne));
         }
       }
       // A(N-1, N-2:N-1) = [-1, 1]
       else if (lclRow == static_cast<local_ordinal_type>(lNumRegFineGIDs - 1)) {
-        regA->insertGlobalValues (curGblRow,
-                                   tuple<global_ordinal_type> (prevGblRow, curGblRow),
-                                   tuple<scalar_type> (negOne, one));
+        regA->insertGlobalValues(curGblRow,
+                                 tuple<global_ordinal_type>(prevGblRow, curGblRow),
+                                 tuple<scalar_type>(negOne, one));
       }
       // A(i, i-1:i+1) = [-1, 2, -1]
       else {
-        regA->insertGlobalValues (curGblRow,
-                                   tuple<global_ordinal_type> (prevGblRow, curGblRow, nextGblRow),
-                                   tuple<scalar_type> (negOne, two, negOne));
+        regA->insertGlobalValues(curGblRow,
+                                 tuple<global_ordinal_type>(prevGblRow, curGblRow, nextGblRow),
+                                 tuple<scalar_type>(negOne, two, negOne));
       }
     }
-    regA->fillComplete ();
+    regA->fillComplete();
 
     Tpetra::MatrixMarket::Writer<crs_matrix_type>::
-      writeSparseFile("regA.m", regA, "regA",
-                      "region representation of the operator.");
+        writeSparseFile("regA.m", regA, "regA",
+                        "region representation of the operator.");
 
-    out << std::endl << "Forming the prolongator" << std::endl;
+    out << std::endl
+        << "Forming the prolongator" << std::endl;
 
     MueLu::Level fineLevel, coarseLevel;
-    RCP<Xpetra::Matrix<SC,LO,GO,NO> > matA = MueLu::TpetraCrs_To_XpetraMatrix<SC,LO,GO,NO>(regA);
-    createTwoLevelHierarchy<SC,LO,GO,NO>(fineLevel, coarseLevel, matA);
+    RCP<Xpetra::Matrix<SC, LO, GO, NO> > matA = MueLu::TpetraCrs_To_XpetraMatrix<SC, LO, GO, NO>(regA);
+    createTwoLevelHierarchy<SC, LO, GO, NO>(fineLevel, coarseLevel, matA);
 
     // Now the prolongator needs to be created
-    Array<global_ordinal_type> coarseRegGIDs (lNumRegCoarseGIDs);
-    if(myRank == 0) {
+    Array<global_ordinal_type> coarseRegGIDs(lNumRegCoarseGIDs);
+    if (myRank == 0) {
       coarseRegGIDs[0] = 0;
       coarseRegGIDs[1] = 3;
       coarseRegGIDs[2] = 6;
-    } else if(myRank == 1) {
+    } else if (myRank == 1) {
       coarseRegGIDs[0] = 10;
       coarseRegGIDs[1] = 9;
     }
     out << "coarseRegGIDs: " << coarseRegGIDs << std::endl;
     RCP<map_type> colMapP = rcp(new map_type(5, coarseRegGIDs, 0, comm));
     // RCP<crs_matrix_type> regP (new crs_matrix_type (fineRegRowMap, 1));
-    RCP<crs_matrix_type> regP (new crs_matrix_type (fineRegRowMap, colMapP, 1));
+    RCP<crs_matrix_type> regP(new crs_matrix_type(fineRegRowMap, colMapP, 1));
 
     // Now each row needs to be filled
-    for(local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumRegFineGIDs); ++lclRow) {
-      const global_ordinal_type gblRow = fineRegRowMap->getGlobalElement (lclRow);
-      const global_ordinal_type gblCol = colMapP->getGlobalElement ((lclRow + 1) / 3);
-      regP->insertGlobalValues (gblRow,
-                                tuple<global_ordinal_type> (gblCol),
-                                tuple<scalar_type> (one));
+    for (local_ordinal_type lclRow = 0; lclRow < static_cast<local_ordinal_type>(lNumRegFineGIDs); ++lclRow) {
+      const global_ordinal_type gblRow = fineRegRowMap->getGlobalElement(lclRow);
+      const global_ordinal_type gblCol = colMapP->getGlobalElement((lclRow + 1) / 3);
+      regP->insertGlobalValues(gblRow,
+                               tuple<global_ordinal_type>(gblCol),
+                               tuple<scalar_type>(one));
     }
-    regP->fillComplete ();
+    regP->fillComplete();
 
     Tpetra::MatrixMarket::Writer<crs_matrix_type>::
-      writeSparseFile("regP.m", regP, "regP",
-                      "region representation of the prolongator.");
+        writeSparseFile("regP.m", regP, "regP",
+                        "region representation of the prolongator.");
 
-    out << std::endl << "Computing AP = AxP" << std::endl;
+    out << std::endl
+        << "Computing AP = AxP" << std::endl;
 
     // The number of non-zeros per row is set to zero because it's hard to guess what it will be...
-    RCP<crs_matrix_type> regAP = rcp( new crs_matrix_type(fineRegRowMap, 0));
+    RCP<crs_matrix_type> regAP = rcp(new crs_matrix_type(fineRegRowMap, 0));
     Tpetra::MatrixMatrix::
-      Multiply<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(*regA, false, *regP,
-                                                                                false, *regAP);
+        Multiply<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(*regA, false, *regP,
+                                                                                  false, *regAP);
 
     Tpetra::MatrixMarket::Writer<crs_matrix_type>::
-      writeSparseFile("regAP.m", regAP, "regAP",
-                      "region representation of AP.");
+        writeSparseFile("regAP.m", regAP, "regAP",
+                        "region representation of AP.");
 
-    out << std::endl << "Computing Ac = P'xAP" << std::endl;
+    out << std::endl
+        << "Computing Ac = P'xAP" << std::endl;
 
     // The number of non-zeros per row is set to zero because it's hard to guess what it will be...
-    RCP<crs_matrix_type> regAc = rcp( new crs_matrix_type(regP->getDomainMap(), 0));
+    RCP<crs_matrix_type> regAc = rcp(new crs_matrix_type(regP->getDomainMap(), 0));
     Tpetra::MatrixMatrix::
-      Multiply<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(*regP, true, *regAP,
-                                                                                false, *regAc);
+        Multiply<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(*regP, true, *regAP,
+                                                                                  false, *regAc);
 
     Tpetra::MatrixMarket::Writer<crs_matrix_type>::
-      writeSparseFile("regAc.m", regAc, "regAc",
-                      "region representation of Ac.");
+        writeSparseFile("regAc.m", regAc, "regAc",
+                        "region representation of Ac.");
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void createTwoLevelHierarchy(MueLu::Level& fineLevel, MueLu::Level& coarseLevel,
                              Teuchos::RCP<typename Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A) {
 #include <MueLu_UseShortNames.hpp>
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
   fineLevel.SetFactoryManager(factoryHandler);
   coarseLevel.SetFactoryManager(factoryHandler);
@@ -338,14 +345,14 @@ void createTwoLevelHierarchy(MueLu::Level& fineLevel, MueLu::Level& coarseLevel,
   A->SetFixedBlockSize(1);
   fineLevel.Request("A");
   fineLevel.Set("A", A);
-  fineLevel.Set("DofsPerNode",2);
+  fineLevel.Set("DofsPerNode", 2);
 
   RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
   nullSpace->randomize();
-  fineLevel.Set("Nullspace",nullSpace);
+  fineLevel.Set("Nullspace", nullSpace);
 
   RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-  RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
   dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
   RCP<UncoupledAggregationFactory> UnCoupledAggFact = rcp(new UncoupledAggregationFactory());
   UnCoupledAggFact->SetFactory("Graph", dropFact);
@@ -357,22 +364,22 @@ void createTwoLevelHierarchy(MueLu::Level& fineLevel, MueLu::Level& coarseLevel,
   TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
   TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-  coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-  coarseLevel.Request("Nullspace",TentativePFact.get());
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
   coarseLevel.Request(*TentativePFact);
   Teuchos::ParameterList paramList;
-  paramList.set("tentative: calculate qr",       false);
+  paramList.set("tentative: calculate qr", false);
   TentativePFact->SetParameterList(paramList);
-  TentativePFact->Build(fineLevel,coarseLevel);
+  TentativePFact->Build(fineLevel, coarseLevel);
 
   RCP<Matrix> Ptent;
-  coarseLevel.Get("P",Ptent,TentativePFact.get());
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
 }
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/research/luc/region_algorithms/Driver_Structured_Regions.cpp b/packages/muelu/research/luc/region_algorithms/Driver_Structured_Regions.cpp
index 5f08211d71c3..a60731e64a4e 100644
--- a/packages/muelu/research/luc/region_algorithms/Driver_Structured_Regions.cpp
+++ b/packages/muelu/research/luc/region_algorithms/Driver_Structured_Regions.cpp
@@ -84,12 +84,11 @@
 #include <BelosBlockGmresSolMgr.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
-#include <BelosTpetraAdapter.hpp>    // => This header defines Belos::TpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
+#include <BelosTpetraAdapter.hpp>  // => This header defines Belos::TpetraOp
 #endif
 
-
 #ifdef HAVE_MUELU_CUDA
 #include "cuda_profiler_api.h"
 #endif
@@ -108,21 +107,20 @@
 
 #include <MueLu_CreateXpetraPreconditioner.hpp>
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  const int myRank = comm->getRank();
+  const int myRank                    = comm->getRank();
 
   // =========================================================================
   // Convenient definitions
@@ -130,34 +128,44 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   typedef Teuchos::ScalarTraits<SC> STS;
   SC zero = STS::zero(), one = STS::one();
   typedef typename STS::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 10, ny = 10, nz = 10;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "";                clp.setOption("xml",                   &xmlFileName,       "read parameters from an xml file");
-  std::string yamlFileName      = "";                clp.setOption("yaml",                  &yamlFileName,      "read parameters from a yaml file");
-  int         maxIts            = 200;               clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-  double      tol              = 1e-12;              clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-  bool        scaleResidualHist = true;              clp.setOption("scale", "noscale",      &scaleResidualHist, "scaled Krylov residual history");
-  bool        solvePreconditioned = true;            clp.setOption("solve-preconditioned","no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
-  std::string equilibrate = "no" ;                   clp.setOption("equilibrate",           &equilibrate,       "equilibrate the system (no | diag | 1-norm)");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+  std::string yamlFileName = "";
+  clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  bool scaleResidualHist = true;
+  clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
+  bool solvePreconditioned = true;
+  clp.setOption("solve-preconditioned", "no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
+  std::string equilibrate = "no";
+  clp.setOption("equilibrate", &equilibrate, "equilibrate the system (no | diag | 1-norm)");
 #ifdef HAVE_MUELU_CUDA
-  bool profileSetup = false;                         clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
-  bool profileSolve = false;                         clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
+  bool profileSetup = false;
+  clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
+  bool profileSolve = false;
+  clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
 #endif
-  int  cacheSize = 0;                                clp.setOption("cachesize",               &cacheSize,       "cache size (in KB)");
+  int cacheSize = 0;
+  clp.setOption("cachesize", &cacheSize, "cache size (in KB)");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   TEUCHOS_TEST_FOR_EXCEPTION(xmlFileName != "" && yamlFileName != "", std::runtime_error,
@@ -165,7 +173,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream &out       = *fancy;
   out.setOutputToRootOnly(0);
 
   ParameterList paramList;
@@ -181,7 +189,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), *comm);
   }
 
-
   // Retrieve matrix parameters (they may have been changed on the command line)
   // [for instance, if we changed matrix type from 2D to 3D we need to update nz]
   ParameterList galeriList = galeriParameters.GetParameterList();
@@ -192,25 +199,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   std::ostringstream galeriStream;
 #ifdef HAVE_MUELU_OPENMP
   std::string node_name = Node::name();
-  if(!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
-    galeriStream<<"OpenMP Max Threads = "<<omp_get_max_threads()<<std::endl;
+  if (!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
+    galeriStream << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
 #endif
 
-
   comm->barrier();
   Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-
-  RCP<Matrix>      A;
-  RCP<const Map>   map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<RealValuedMultiVector> coordinates;
   typedef typename RealValuedMultiVector::scalar_type Real;
-  RCP<Xpetra::MultiVector<SC,LO,GO,NO> > nullspace;
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > nullspace;
   RCP<MultiVector> X, B;
 
-  galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+  galeriStream << "========================================================\n"
+               << xpetraParameters << galeriParameters;
 
   // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
   //                                 d1  d2  d3
@@ -222,8 +228,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // size. For example, np=14 will give a 7-by-2 distribution.
   // If you don't want Galeri to do this, specify mx or my on the galeriList.
   std::string matrixType = galeriParameters.GetMatrixType();
-  int numDimensions  = 0;
-  int numDofsPerNode = 1;
+  int numDimensions      = 0;
+  int numDofsPerNode     = 1;
   Teuchos::Array<GO> procsPerDim(3);
   Teuchos::Array<GO> gNodesPerDim(3);
   Teuchos::Array<LO> lNodesPerDim(3);
@@ -233,26 +239,26 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
   if (matrixType == "Laplace1D") {
     numDimensions = 1;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
   } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
              matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
     numDimensions = 2;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
   } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
     numDimensions = 3;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
   }
 
   // Expand map to do multiple DOF per node for block problems
   if (matrixType == "Elasticity2D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
   if (matrixType == "Elasticity3D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
   galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -261,20 +267,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
     // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-    galeriList.set("right boundary" , "Neumann");
+    galeriList.set("right boundary", "Neumann");
     galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
+    galeriList.set("top boundary", "Neumann");
+    galeriList.set("front boundary", "Neumann");
+    galeriList.set("back boundary", "Neumann");
   }
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
   A = Pr->BuildMatrix();
 
-  if(matrixType == "Elasticity2D") {
+  if (matrixType == "Elasticity2D") {
     numDofsPerNode = 2;
-  } else if(matrixType == "Elasticity3D") {
+  } else if (matrixType == "Elasticity3D") {
     numDofsPerNode = 3;
   }
 
@@ -290,7 +296,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   Teuchos::Array<typename STS::magnitudeType> norms(1);
   B->norm2(norms);
-  B->scale(one/norms[0]);
+  B->scale(one / norms[0]);
   galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
   out << galeriStream.str();
@@ -301,7 +307,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - Compute region data")));
 
   // Loading geometric info from galeri
-  if(numDimensions == 1) {
+  if (numDimensions == 1) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = 1;
     gNodesPerDim[2] = 1;
@@ -313,7 +319,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     procsPerDim[0] = galeriList.get<GO>("mx");
     procsPerDim[1] = 1;
     procsPerDim[2] = 1;
-  } else if(numDimensions == 2) {
+  } else if (numDimensions == 2) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = galeriList.get<GO>("ny");
     gNodesPerDim[2] = 1;
@@ -325,7 +331,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     procsPerDim[0] = galeriList.get<GO>("mx");
     procsPerDim[1] = galeriList.get<GO>("my");
     procsPerDim[2] = 1;
-  } else if(numDimensions == 3) {
+  } else if (numDimensions == 3) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = galeriList.get<GO>("ny");
     gNodesPerDim[2] = galeriList.get<GO>("nz");
@@ -342,22 +348,34 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   Teuchos::Array<GO> startIndices(3);
   Teuchos::Array<GO> endIndices(3);
   const GO startGID = map->getMinGlobalIndex();
-  startIndices[2] = startGID / (gNodesPerDim[1]*gNodesPerDim[0]);
-  const GO rem    = startGID % (gNodesPerDim[1]*gNodesPerDim[0]);
-  startIndices[1] = rem / gNodesPerDim[0];
-  startIndices[0] = rem % gNodesPerDim[0];
-  endIndices[0] = startIndices[0] + lNodesPerDim[0] - 1;
-  endIndices[1] = startIndices[1] + lNodesPerDim[1] - 1;
-  endIndices[2] = startIndices[2] + lNodesPerDim[2] - 1;
+  startIndices[2]   = startGID / (gNodesPerDim[1] * gNodesPerDim[0]);
+  const GO rem      = startGID % (gNodesPerDim[1] * gNodesPerDim[0]);
+  startIndices[1]   = rem / gNodesPerDim[0];
+  startIndices[0]   = rem % gNodesPerDim[0];
+  endIndices[0]     = startIndices[0] + lNodesPerDim[0] - 1;
+  endIndices[1]     = startIndices[1] + lNodesPerDim[1] - 1;
+  endIndices[2]     = startIndices[2] + lNodesPerDim[2] - 1;
 
   int leftBC = 0, rightBC = 0, frontBC = 0, backBC = 0, bottomBC = 0, topBC = 0;
-  if(startIndices[0] == 0) {leftBC = 1;}
-  if(startIndices[1] == 0) {frontBC = 1;}
-  if(startIndices[2] == 0) {bottomBC = 1;}
+  if (startIndices[0] == 0) {
+    leftBC = 1;
+  }
+  if (startIndices[1] == 0) {
+    frontBC = 1;
+  }
+  if (startIndices[2] == 0) {
+    bottomBC = 1;
+  }
 
-  if(endIndices[0] == gNodesPerDim[0] - 1) {rightBC = 1;}
-  if(endIndices[1] == gNodesPerDim[1] - 1) {backBC = 1;}
-  if(endIndices[2] == gNodesPerDim[2] - 1) {topBC = 1;}
+  if (endIndices[0] == gNodesPerDim[0] - 1) {
+    rightBC = 1;
+  }
+  if (endIndices[1] == gNodesPerDim[1] - 1) {
+    backBC = 1;
+  }
+  if (endIndices[2] == gNodesPerDim[2] - 1) {
+    topBC = 1;
+  }
 
   std::cout << "p=" << myRank << " | startGID= " << startGID
             << ", startIndices: " << startIndices
@@ -374,12 +392,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // First we count how many nodes the region needs to send and receive
   // and allocate arrays accordingly
   LO numReceive = 0, numSend = 0;
-  Teuchos::Array<GO>  receiveGIDs;
+  Teuchos::Array<GO> receiveGIDs;
   Teuchos::Array<int> receivePIDs;
-  Teuchos::Array<GO>  sendGIDs;
+  Teuchos::Array<GO> sendGIDs;
   Teuchos::Array<int> sendPIDs;
-  if(numDimensions == 1) {
-    if(leftBC == 0) {
+  if (numDimensions == 1) {
+    if (leftBC == 0) {
       numReceive = 1;
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -387,7 +405,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       receiveGIDs[0] = startIndices[0] - 1;
       receivePIDs[0] = myRank - 1;
     }
-    if(rightBC == 0) {
+    if (rightBC == 0) {
       numSend = 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
@@ -395,9 +413,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       sendGIDs[0] = endIndices[0];
       sendGIDs[0] = myRank + 1;
     }
-  } else if(numDimensions == 2) {
+  } else if (numDimensions == 2) {
     // Received nodes
-    if(frontBC == 0 && leftBC == 0) {
+    if (frontBC == 0 && leftBC == 0) {
       numReceive = lNodesPerDim[0] + lNodesPerDim[1] + 1;
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -408,67 +426,67 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
       ++countIDs;
       // Receive front edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
         receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i;
         receivePIDs[countIDs] = myRank - procsPerDim[0];
         ++countIDs;
       }
       // Receive left edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - 1 + j*gNodesPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0];
         receivePIDs[countIDs] = myRank - 1;
         ++countIDs;
       }
-    } else if(frontBC == 0) {
+    } else if (frontBC == 0) {
       numReceive = lNodesPerDim[0];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
         receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i;
         receivePIDs[countIDs] = myRank - procsPerDim[0];
         ++countIDs;
       }
-    } else if(leftBC == 0) {
+    } else if (leftBC == 0) {
       numReceive = lNodesPerDim[1];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive left edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - 1 + j*gNodesPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0];
         receivePIDs[countIDs] = myRank - 1;
         ++countIDs;
       }
     }
 
     // Sent nodes
-    if(rightBC == 0 && backBC == 0) {
+    if (rightBC == 0 && backBC == 0) {
       numSend = lNodesPerDim[0] + lNodesPerDim[1] + 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + 1;
         ++countIDs;
       }
       // Send nodes of back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
         sendPIDs[countIDs] = myRank + procsPerDim[0];
         ++countIDs;
       }
       // Send node of back-right corner
-      sendGIDs[countIDs] = startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
+      sendGIDs[countIDs] = startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
       sendPIDs[countIDs] = myRank + procsPerDim[1] + 1;
       ++countIDs;
-    } else if(backBC == 0) {
+    } else if (backBC == 0) {
       numSend = lNodesPerDim[0];
 
       sendGIDs.resize(numSend);
@@ -476,460 +494,406 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       LO countIDs = 0;
       // Send nodes of back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
         sendPIDs[countIDs] = myRank + procsPerDim[0];
         ++countIDs;
       }
-    } else if(rightBC == 0) {
+    } else if (rightBC == 0) {
       numSend = lNodesPerDim[1];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + 1;
         ++countIDs;
       }
     }
-  } else if(numDimensions == 3) {
+  } else if (numDimensions == 3) {
     // Received nodes
-    if( (bottomBC == 0) && (frontBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[1]     // bottom face
-        + (lNodesPerDim[0] + 1)*lNodesPerDim[2]        // front face
-        + (lNodesPerDim[1] + 1)*(lNodesPerDim[2] + 1); // left face
+    if ((bottomBC == 0) && (frontBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[1]                 // bottom face
+                   + (lNodesPerDim[0] + 1) * lNodesPerDim[2]         // front face
+                   + (lNodesPerDim[1] + 1) * (lNodesPerDim[2] + 1);  // left face
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-left-bottom corner node
-      receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1
-          - gNodesPerDim[1]*gNodesPerDim[0];
-      receivePIDs[countIDs] = myRank - procsPerDim[0] - 1
-          - procsPerDim[1]*procsPerDim[0];
+      receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1 - gNodesPerDim[1] * gNodesPerDim[0];
+      receivePIDs[countIDs] = myRank - procsPerDim[0] - 1 - procsPerDim[1] * procsPerDim[0];
       ++countIDs;
       // Receive front-bottom edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - gNodesPerDim[0] + i;
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - gNodesPerDim[0] + i;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - procsPerDim[0];
         ++countIDs;
       }
       // Receive left-bottom edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - 1 + j*gNodesPerDim[0];
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - 1 + j * gNodesPerDim[0];
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - 1;
         ++countIDs;
       }
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              + i
-              + j*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] + i + j * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1];
           ++countIDs;
         }
       }
       // Receive front-left edge nodes
-      for(LO k = 0; k < lNodesPerDim[1]; ++k) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]
-              - 1 + k*gNodesPerDim[0]*gNodesPerDim[1];
+      for (LO k = 0; k < lNodesPerDim[1]; ++k) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1 + k * gNodesPerDim[0] * gNodesPerDim[1];
         receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
         ++countIDs;
       }
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
       // Receive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = startGID - 1
-              + j*gNodesPerDim[0]
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0] + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-    // Two faces received
-    } else if( (bottomBC == 0) && (frontBC == 0) ) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[1]     // bottom face
-        + lNodesPerDim[0]*(lNodesPerDim[2] + 1);       // front face;
+      // Two faces received
+    } else if ((bottomBC == 0) && (frontBC == 0)) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[1]           // bottom face
+                   + lNodesPerDim[0] * (lNodesPerDim[2] + 1);  // front face;
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-bottom edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - gNodesPerDim[0] + i;
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - gNodesPerDim[0] + i;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - procsPerDim[0];
         ++countIDs;
       }
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              + i
-              + j*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] + i + j * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1];
           ++countIDs;
         }
       }
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if( (bottomBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[1]*(lNodesPerDim[0] + lNodesPerDim[2] + 1);
+    } else if ((bottomBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[1] * (lNodesPerDim[0] + lNodesPerDim[2] + 1);
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive left-bottom edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] =  j*gNodesPerDim[0]
-          + startGID - gNodesPerDim[1]*gNodesPerDim[0] - 1;
-        receivePIDs[countIDs] =  myRank - procsPerDim[1]*procsPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = j * gNodesPerDim[0] + startGID - gNodesPerDim[1] * gNodesPerDim[0] - 1;
+        receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0] - 1;
         ++countIDs;
       }
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[1]*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID - gNodesPerDim[1] * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Receive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-    } else if( (frontBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[2]*(lNodesPerDim[1] + lNodesPerDim[0] + 1);
+    } else if ((frontBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[2] * (lNodesPerDim[1] + lNodesPerDim[0] + 1);
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-left edge nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        receiveGIDs[countIDs] =  k*gNodesPerDim[1]*gNodesPerDim[0]
-          + startGID - gNodesPerDim[0] - 1;
-        receivePIDs[countIDs] =  myRank - procsPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + startGID - gNodesPerDim[0] - 1;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
         ++countIDs;
       }
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID - gNodesPerDim[0];
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
       // Receive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-    // Single face received
-    } else if(bottomBC == 0) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[1];
+      // Single face received
+    } else if (bottomBC == 0) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[1];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[1]*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID - gNodesPerDim[1] * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(frontBC == 0) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[2];
+    } else if (frontBC == 0) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[2];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID - gNodesPerDim[0];
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(leftBC == 0) {
-      numReceive = lNodesPerDim[1]*lNodesPerDim[2];
+    } else if (leftBC == 0) {
+      numReceive = lNodesPerDim[1] * lNodesPerDim[2];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Recive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + j*gNodesPerDim[0] + startGID - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
-
     }
 
     // Sent nodes
-    if( (topBC == 0) && (backBC == 0) && (rightBC == 0) ) {
-      numSend = (lNodesPerDim[0])*(lNodesPerDim[1])
-        + (lNodesPerDim[0])*(lNodesPerDim[2])
-        + (lNodesPerDim[1])*(lNodesPerDim[2])
-        + lNodesPerDim[0]
-        + lNodesPerDim[1]
-        + lNodesPerDim[2]
-        + 1;
+    if ((topBC == 0) && (backBC == 0) && (rightBC == 0)) {
+      numSend = (lNodesPerDim[0]) * (lNodesPerDim[1]) + (lNodesPerDim[0]) * (lNodesPerDim[2]) + (lNodesPerDim[1]) * (lNodesPerDim[2]) + lNodesPerDim[0] + lNodesPerDim[1] + lNodesPerDim[2] + 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-            + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0]) + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of right-back edge
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-          + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
         ++countIDs;
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0]
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + 1;
         ++countIDs;
       }
       // Send nodes of top-back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-          + (lNodesPerDim[0] - 1)*gNodesPerDim[0];
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[1];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[0] - 1) * gNodesPerDim[0];
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[1];
         ++countIDs;
       }
       // Send node of top-back-right corner
-      sendGIDs[countIDs] = startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-        + (lNodesPerDim[0] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-      sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[1] + 1;
+      sendGIDs[countIDs] = startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[0] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+      sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[1] + 1;
       ++countIDs;
 
-    } else if( (topBC == 0) && (backBC == 0) ) {
-      numSend = (lNodesPerDim[0]*lNodesPerDim[2]) // back face
-              + (lNodesPerDim[0]*lNodesPerDim[1]) // Top face
-              + (lNodesPerDim[0]); // top-back edge
+    } else if ((topBC == 0) && (backBC == 0)) {
+      numSend = (lNodesPerDim[0] * lNodesPerDim[2])    // back face
+                + (lNodesPerDim[0] * lNodesPerDim[1])  // Top face
+                + (lNodesPerDim[0]);                   // top-back edge
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0]) + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-          + (lNodesPerDim[0] - 1)*gNodesPerDim[0];
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[1];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[0] - 1) * gNodesPerDim[0];
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[1];
         ++countIDs;
       }
 
-    } else if( (topBC == 0) && (rightBC == 0) ) {
-      numSend = (lNodesPerDim[1]*lNodesPerDim[2]) // right face
-              + (lNodesPerDim[0]*lNodesPerDim[1]) // Top face
-              + (lNodesPerDim[1]); // top-right edge
+    } else if ((topBC == 0) && (rightBC == 0)) {
+      numSend = (lNodesPerDim[1] * lNodesPerDim[2])    // right face
+                + (lNodesPerDim[0] * lNodesPerDim[1])  // Top face
+                + (lNodesPerDim[1]);                   // top-right edge
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-            + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0]
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + 1;
         ++countIDs;
       }
 
-    } else if( (backBC == 0) && (rightBC == 0) ) {
-      numSend = lNodesPerDim[2]*(lNodesPerDim[0] + lNodesPerDim[1] + 1);
+    } else if ((backBC == 0) && (rightBC == 0)) {
+      numSend = lNodesPerDim[2] * (lNodesPerDim[0] + lNodesPerDim[1] + 1);
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of back-right edge
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-          sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
-          ++countIDs;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
+        ++countIDs;
       }
 
-    } else if(topBC == 0) {
-      numSend = lNodesPerDim[0]*lNodesPerDim[1];
+    } else if (topBC == 0) {
+      numSend = lNodesPerDim[0] * lNodesPerDim[1];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[0]*gNodesPerDim[0];
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[0] * gNodesPerDim[0];
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(backBC == 0) {
-      numSend = lNodesPerDim[0]*lNodesPerDim[2];
+    } else if (backBC == 0) {
+      numSend = lNodesPerDim[0] * lNodesPerDim[2];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(rightBC == 0) {
-      numSend = lNodesPerDim[1]*lNodesPerDim[2];
+    } else if (rightBC == 0) {
+      numSend = lNodesPerDim[1] * lNodesPerDim[2];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
-
     }
   }
 
@@ -946,94 +910,94 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   comm->barrier();
   tm = Teuchos::null;
 
-// #ifdef HAVE_MUELU_CUDA
-//   if(profileSetup) cudaProfilerStart();
-// #endif
-
-//   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup")));
-//   RCP<Hierarchy> H;
-//   RCP<Operator> Prec;
-//   A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
-
-//   const std::string userName = "user data";
-//   Teuchos::ParameterList& userParamList = paramList.sublist(userName);
-//   userParamList.set<int>("int numDimensions", numDimensions);
-//   userParamList.set<Teuchos::Array<LO> >("Array<LO> lNodesPerDim", lNodesPerDim);
-//   userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-//   H = MueLu::CreateXpetraPreconditioner(A, paramList, paramList);
-
-//   comm->barrier();
-//   tm = Teuchos::null;
-
-// #ifdef HAVE_MUELU_CUDA
-//   if(profileSolve) cudaProfilerStop();
-// #endif
-
-//   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - LHS and RHS initialization")));
-//   X->putScalar(zero);
-//   tm = Teuchos::null;
-
-// #ifdef HAVE_MUELU_BELOS
-//   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Belos Solve")));
-// #ifdef HAVE_MUELU_CUDA
-//   if(profileSolve) cudaProfilerStart();
-// #endif
-//   // Operator and Multivector type that will be used with Belos
-//   typedef MultiVector          MV;
-//   typedef Belos::OperatorT<MV> OP;
-
-//   // Define Operator and Preconditioner
-//   Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-//   Teuchos::RCP<OP> belosPrec; // Turns a MueLu::Hierarchy object into a Belos operator
-//   H->IsPreconditioner(true);
-//   belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
-
-//   // Construct a Belos LinearProblem object
-//   RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
-//   if(solvePreconditioned) belosProblem->setRightPrec(belosPrec);
-
-//   bool set = belosProblem->setProblem();
-//   if (set == false) {
-//     out << "\nERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
-//     return EXIT_FAILURE;
-//   }
-
-//   // Belos parameter list
-//   Teuchos::ParameterList belosList;
-//   belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-//   belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-//   belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-//   belosList.set("Output Frequency",      1);
-//   belosList.set("Output Style",          Belos::Brief);
-//   if (!scaleResidualHist)
-//     belosList.set("Implicit Residual Scaling", "None");
-
-//   // Create an iterative solver manager
-//   RCP< Belos::SolverManager<SC, MV, OP> > solver;
-//   solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
-
-//   // Perform solve
-//   Belos::ReturnType retStatus = Belos::Unconverged;
-//   retStatus = solver->solve();
-
-//   // Get the number of iterations for this solve.
-//   out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
-//   // Check convergence
-//   if (retStatus != Belos::Converged)
-//     out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
-//   else
-//     out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-// #ifdef HAVE_MUELU_CUDA
-//   if(profileSolve) cudaProfilerStop();
-// #endif
-// #endif //ifdef HAVE_MUELU_BELOS
+  // #ifdef HAVE_MUELU_CUDA
+  //   if(profileSetup) cudaProfilerStart();
+  // #endif
+
+  //   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup")));
+  //   RCP<Hierarchy> H;
+  //   RCP<Operator> Prec;
+  //   A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
+
+  //   const std::string userName = "user data";
+  //   Teuchos::ParameterList& userParamList = paramList.sublist(userName);
+  //   userParamList.set<int>("int numDimensions", numDimensions);
+  //   userParamList.set<Teuchos::Array<LO> >("Array<LO> lNodesPerDim", lNodesPerDim);
+  //   userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  //   H = MueLu::CreateXpetraPreconditioner(A, paramList, paramList);
+
+  //   comm->barrier();
+  //   tm = Teuchos::null;
+
+  // #ifdef HAVE_MUELU_CUDA
+  //   if(profileSolve) cudaProfilerStop();
+  // #endif
+
+  //   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - LHS and RHS initialization")));
+  //   X->putScalar(zero);
+  //   tm = Teuchos::null;
+
+  // #ifdef HAVE_MUELU_BELOS
+  //   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Belos Solve")));
+  // #ifdef HAVE_MUELU_CUDA
+  //   if(profileSolve) cudaProfilerStart();
+  // #endif
+  //   // Operator and Multivector type that will be used with Belos
+  //   typedef MultiVector          MV;
+  //   typedef Belos::OperatorT<MV> OP;
+
+  //   // Define Operator and Preconditioner
+  //   Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
+  //   Teuchos::RCP<OP> belosPrec; // Turns a MueLu::Hierarchy object into a Belos operator
+  //   H->IsPreconditioner(true);
+  //   belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+
+  //   // Construct a Belos LinearProblem object
+  //   RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+  //   if(solvePreconditioned) belosProblem->setRightPrec(belosPrec);
+
+  //   bool set = belosProblem->setProblem();
+  //   if (set == false) {
+  //     out << "\nERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
+  //     return EXIT_FAILURE;
+  //   }
+
+  //   // Belos parameter list
+  //   Teuchos::ParameterList belosList;
+  //   belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
+  //   belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
+  //   belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  //   belosList.set("Output Frequency",      1);
+  //   belosList.set("Output Style",          Belos::Brief);
+  //   if (!scaleResidualHist)
+  //     belosList.set("Implicit Residual Scaling", "None");
+
+  //   // Create an iterative solver manager
+  //   RCP< Belos::SolverManager<SC, MV, OP> > solver;
+  //   solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+
+  //   // Perform solve
+  //   Belos::ReturnType retStatus = Belos::Unconverged;
+  //   retStatus = solver->solve();
+
+  //   // Get the number of iterations for this solve.
+  //   out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
+  //   // Check convergence
+  //   if (retStatus != Belos::Converged)
+  //     out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+  //   else
+  //     out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+  // #ifdef HAVE_MUELU_CUDA
+  //   if(profileSolve) cudaProfilerStop();
+  // #endif
+  // #endif //ifdef HAVE_MUELU_BELOS
 
   comm->barrier();
-  tm = Teuchos::null;
+  tm                = Teuchos::null;
   globalTimeMonitor = Teuchos::null;
 
   RCP<ParameterList> reportParams = rcp(new ParameterList);
-  const std::string filter = "";
+  const std::string filter        = "";
   std::ios_base::fmtflags ff(out.flags());
   TimeMonitor::report(comm.ptr(), out, filter, reportParams);
   out << std::setiosflags(ff);
@@ -1048,5 +1012,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/research/max/AdditiveMG/ADRProblemFactory.hpp b/packages/muelu/research/max/AdditiveMG/ADRProblemFactory.hpp
index 21a7ce6c8195..3574d40dba79 100644
--- a/packages/muelu/research/max/AdditiveMG/ADRProblemFactory.hpp
+++ b/packages/muelu/research/max/AdditiveMG/ADRProblemFactory.hpp
@@ -59,30 +59,33 @@
 
 namespace ADR {
 
-  namespace Xpetra {
+namespace Xpetra {
 
-    using Teuchos::RCP;
+using Teuchos::RCP;
 
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    RCP<Problem<Map,Matrix,MultiVector> > BuildProblem(const std::string &MatrixType, const RCP<const Map>& map, Teuchos::ParameterList& list) {
-      RCP<Problem<Map,Matrix,MultiVector> > P;
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+RCP<Problem<Map, Matrix, MultiVector> > BuildProblem(const std::string& MatrixType, const RCP<const Map>& map, Teuchos::ParameterList& list) {
+  RCP<Problem<Map, Matrix, MultiVector> > P;
 
-      if      (MatrixType == "ADR1D")    P.reset(new ADR1DProblem   <Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>(list, map));
-      else if (MatrixType == "ADR2D")    P.reset(new ADR2DProblem   <Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>(list, map));
-      else if (MatrixType == "ADR3D")    P.reset(new ADR3DProblem   <Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>(list, map));
- 
-      else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, std::logic_error,
-                                   "`MatrixType' has incorrect value (" << MatrixType << ") in input to function CreateCrsMatrix()."
-                                   << "Check the documentation for a list of valid choices");
+  if (MatrixType == "ADR1D")
+    P.reset(new ADR1DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>(list, map));
+  else if (MatrixType == "ADR2D")
+    P.reset(new ADR2DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>(list, map));
+  else if (MatrixType == "ADR3D")
+    P.reset(new ADR3DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>(list, map));
 
-      P->setObjectLabel(MatrixType);
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::logic_error,
+                               "`MatrixType' has incorrect value (" << MatrixType << ") in input to function CreateCrsMatrix()."
+                                                                    << "Check the documentation for a list of valid choices");
 
-      return P;
-    }
+  P->setObjectLabel(MatrixType);
 
-  } // namespace Xpetra
+  return P;
+}
 
-} // namespace ADR
+}  // namespace Xpetra
 
-#endif //ifndef ADRPROBLEMFACTORY_HPP
+}  // namespace ADR
+
+#endif  // ifndef ADRPROBLEMFACTORY_HPP
diff --git a/packages/muelu/research/max/AdditiveMG/ADR_Problem.hpp b/packages/muelu/research/max/AdditiveMG/ADR_Problem.hpp
index 3c41e11eafd7..731ca4cad57c 100644
--- a/packages/muelu/research/max/AdditiveMG/ADR_Problem.hpp
+++ b/packages/muelu/research/max/AdditiveMG/ADR_Problem.hpp
@@ -7,71 +7,73 @@
 
 namespace ADR {
 
-  namespace Xpetra {
+namespace Xpetra {
 
-    enum {
-      DIR_LEFT   = 0x01,
-      DIR_RIGHT  = 0x02,
-      DIR_BOTTOM = 0x04,
-      DIR_TOP    = 0x08,
-      DIR_FRONT  = 0x10,
-      DIR_BACK   = 0x20,
-      DIR_ALL    = DIR_LEFT | DIR_RIGHT | DIR_BOTTOM | DIR_TOP | DIR_FRONT | DIR_BACK
-    };
-    typedef size_t DirBC;
+enum {
+  DIR_LEFT   = 0x01,
+  DIR_RIGHT  = 0x02,
+  DIR_BOTTOM = 0x04,
+  DIR_TOP    = 0x08,
+  DIR_FRONT  = 0x10,
+  DIR_BACK   = 0x20,
+  DIR_ALL    = DIR_LEFT | DIR_RIGHT | DIR_BOTTOM | DIR_TOP | DIR_FRONT | DIR_BACK
+};
+typedef size_t DirBC;
 
-    template<typename Map, typename Matrix, typename MultiVector>
-    class Problem : public Teuchos::Describable {
-    public:
-      Problem(Teuchos::ParameterList& list) : list_(list) {
-        SetBoundary();
-      };
-      Problem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map) : list_(list) {
-        Map_ = map;
-        SetBoundary();
-      };
-      virtual ~Problem() { }
+template <typename Map, typename Matrix, typename MultiVector>
+class Problem : public Teuchos::Describable {
+ public:
+  Problem(Teuchos::ParameterList& list)
+    : list_(list) {
+    SetBoundary();
+  };
+  Problem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map)
+    : list_(list) {
+    Map_ = map;
+    SetBoundary();
+  };
+  virtual ~Problem() {}
 
-      virtual Teuchos::RCP<Matrix>      BuildMatrix() = 0;
-      virtual Teuchos::RCP<MultiVector> BuildCoords() {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error, "Coordinates construction is not implemented for this problem");
-      }
-      virtual Teuchos::RCP<MultiVector> BuildNullspace() {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error, "Nullspace construction is not implemented for this problem");
-      }
+  virtual Teuchos::RCP<Matrix> BuildMatrix() = 0;
+  virtual Teuchos::RCP<MultiVector> BuildCoords() {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error, "Coordinates construction is not implemented for this problem");
+  }
+  virtual Teuchos::RCP<MultiVector> BuildNullspace() {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error, "Nullspace construction is not implemented for this problem");
+  }
 
-      // Get methods
-      Teuchos::RCP<const Map>           getMap()                                   const { return Map_; }
-      Teuchos::RCP<const Matrix>        getMatrix()                                const { return A_; }
-      Teuchos::RCP<const MultiVector>   getNullspace()                             const { return Nullspace_; }
-      Teuchos::RCP<const MultiVector>   getCoords()                                const { return Coords_; }
+  // Get methods
+  Teuchos::RCP<const Map> getMap() const { return Map_; }
+  Teuchos::RCP<const Matrix> getMatrix() const { return A_; }
+  Teuchos::RCP<const MultiVector> getNullspace() const { return Nullspace_; }
+  Teuchos::RCP<const MultiVector> getCoords() const { return Coords_; }
 
-      // Set methods
-      Teuchos::RCP<const Map>           setMap(const Teuchos::RCP<const Map>& map)       { Map_ = map; }
+  // Set methods
+  Teuchos::RCP<const Map> setMap(const Teuchos::RCP<const Map>& map) { Map_ = map; }
 
-    protected:
-      Teuchos::ParameterList&           list_;
-      Teuchos::RCP<const Map>           Map_;
-      Teuchos::RCP<Matrix>              A_;
-      Teuchos::RCP<MultiVector>         Nullspace_;
-      Teuchos::RCP<MultiVector>         Coords_;
+ protected:
+  Teuchos::ParameterList& list_;
+  Teuchos::RCP<const Map> Map_;
+  Teuchos::RCP<Matrix> A_;
+  Teuchos::RCP<MultiVector> Nullspace_;
+  Teuchos::RCP<MultiVector> Coords_;
 
-      DirBC                             DirichletBC_;
+  DirBC DirichletBC_;
 
-    private:
-      void SetBoundary() {
-        DirichletBC_ = DIR_ALL;
-        if (this->list_.get("left boundary",   "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_LEFT;
-        if (this->list_.get("right boundary",  "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_RIGHT;
-        if (this->list_.get("bottom boundary", "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_BOTTOM;
-        if (this->list_.get("top boundary",    "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_TOP;
-        if (this->list_.get("front boundary",  "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_FRONT;
-        if (this->list_.get("back boundary",   "Dirichlet") == "Neumann")   DirichletBC_ ^= DIR_BACK;
-      }
-    };
+ private:
+  void SetBoundary() {
+    DirichletBC_ = DIR_ALL;
+    if (this->list_.get("left boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_LEFT;
+    if (this->list_.get("right boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_RIGHT;
+    if (this->list_.get("bottom boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_BOTTOM;
+    if (this->list_.get("top boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_TOP;
+    if (this->list_.get("front boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_FRONT;
+    if (this->list_.get("back boundary", "Dirichlet") == "Neumann") DirichletBC_ ^= DIR_BACK;
+  }
+};
 
-  } // namespace Xpetra
+}  // namespace Xpetra
 
-} // namespace ADR
+}  // namespace ADR
 
-#endif // ADR_PROBLEM_HPP
+#endif  // ADR_PROBLEM_HPP
diff --git a/packages/muelu/research/max/AdditiveMG/ADR_XpetraParameters.hpp b/packages/muelu/research/max/AdditiveMG/ADR_XpetraParameters.hpp
index 66705fc2055c..98c3ca2e7595 100644
--- a/packages/muelu/research/max/AdditiveMG/ADR_XpetraParameters.hpp
+++ b/packages/muelu/research/max/AdditiveMG/ADR_XpetraParameters.hpp
@@ -13,183 +13,203 @@
 
 namespace ADR {
 
-  namespace Xpetra {
-
-    // TODO nx/ny/nz == GO or global_size_t ? But what is the best to do?
-
-    template<typename GO>
-    class Parameters : public Teuchos::VerboseObject<Parameters<GO> >, public Teuchos::Describable {
-    public:
-
-      Parameters(Teuchos::CommandLineProcessor& clp, GO nx = 16, GO ny = -1, GO nz = -1, const std::string& matrixType = "ADR1D",
-                 int keepBCs = 0, double stretchx = 1.0, double stretchy = 1.0, double stretchz = 1.0, double h = 1.0, double delta = 0.0,
-                 int PMLXL = 0, int PMLXR = 0, int PMLYL = 0, int PMLYR = 0, int PMLZL = 0, int PMLZR = 0,
-                 double omega = 2.0*M_PI, double shift = 0.5, GO mx = -1, GO my = -1, GO mz = -1, int model = 0)
-	: nx_(nx), ny_(ny), nz_(nz), mx_(mx), my_(my), mz_(mz), stretchx_(stretchx), stretchy_(stretchy), stretchz_(stretchz), matrixType_(matrixType), keepBCs_(keepBCs),
-      h_(h), delta_(delta), PMLx_left(PMLXL), PMLx_right(PMLXR), PMLy_left(PMLYL), PMLy_right(PMLYR), PMLz_left(PMLZL), PMLz_right(PMLZR),
-      omega_(omega), shift_(shift), model_(model) {
-        clp.setOption("nx",         &nx_,           "mesh points in x-direction.");
-        clp.setOption("ny",         &ny_,           "mesh points in y-direction.");
-        clp.setOption("nz",         &nz_,           "mesh points in z-direction.");
-        clp.setOption("mx",         &mx_,           "processors in x-direction.");
-        clp.setOption("my",         &my_,           "processors in y-direction.");
-        clp.setOption("mz",         &mz_,           "processors in z-direction.");
-        clp.setOption("stretchx",   &stretchx_,     "stretch mesh in x-direction.");
-        clp.setOption("stretchy",   &stretchy_,     "stretch mesh in y-direction.");
-        clp.setOption("stretchz",   &stretchz_,     "stretch mesh in z-direction.");
-        clp.setOption("keepBCs",    &keepBCs_,      "keep Dirichlet boundary rows in matrix (0=false,1=true)");
-        clp.setOption("matrixType", &matrixType_,   "matrix type: Laplace1D, Laplace2D, Laplace3D, ..."); //TODO: Star2D, numGlobalElements=...
-        clp.setOption("h",          &h_,            "mesh width for uniform h");
-        clp.setOption("delta",      &delta_,        "maximum PML damping value");
-        clp.setOption("PMLx_left",  &PMLx_left,     "PML grid points in x-direction (left boundary)");
-        clp.setOption("PMLx_right", &PMLx_right,    "PML grid points in x-direction (right boundary)");
-        clp.setOption("PMLy_left",  &PMLy_left,     "PML grid points in y-direction (left boundary)");
-        clp.setOption("PMLy_right", &PMLy_right,    "PML grid points in y-direction (right boundary)");
-        clp.setOption("PMLz_left",  &PMLz_left,     "PML grid points in z-direction (left boundary)");
-        clp.setOption("PMLz_right", &PMLz_right,    "PML grid points in z-direction (right boundary)");
-        clp.setOption("omega",      &omega_,        "angular frequency omega");
-        clp.setOption("shift",      &shift_,        "complex frequency shift");
-        clp.setOption("mx",         &mx_,           "processors in x-direction.");
-        clp.setOption("my",         &my_,           "processors in y-direction.");
-        clp.setOption("mz",         &mz_,           "processors in z-direction.");
-        clp.setOption("model",      &model_,        "velocity model");
-      }
-
-      GO GetNumGlobalElements() const {
-        const Teuchos::ParameterList& pL = GetParameterList();
-
-        const std::string& matrixType = pL.get<std::string>("matrixType");
-        const GO nx = pL.get<GO>("nx");
-        const GO ny = pL.get<GO>("ny");
-        const GO nz = pL.get<GO>("nz");
-
-        GO numGlobalElements = -1;
-        if (matrixType == "ADR1D")
-          numGlobalElements = nx;
-
-        else if (matrixType == "ADR2D")
-          numGlobalElements = nx*ny;
-
-        else if (matrixType == "ADR3D")
-          numGlobalElements = nx*ny*nz;
-
-        TEUCHOS_TEST_FOR_EXCEPTION(numGlobalElements < 0, std::runtime_error,
-                                   "Gallery: numGlobalElements < 0 (did you forget --ny (or --nz) for 2D (3D) problems?)");
-
-        return numGlobalElements;
-      }
-
-      const std::string& GetMatrixType() const {
-        const Teuchos::ParameterList& paramList = GetParameterList();
-        return paramList.get<std::string>("matrixType");
-      }
-
-      void check() const { }
-
-      Teuchos::ParameterList& GetParameterList() const {
-        if (!paramList_.is_null())
-          return *paramList_;
-
-        paramList_ = rcp(new Teuchos::ParameterList());
-
-        paramList_->set("nx",          nx_);
-        paramList_->set("ny",          ny_);
-        paramList_->set("nz",          nz_);
-        paramList_->set("mx",          mx_);
-        paramList_->set("my",          my_);
-        paramList_->set("mz",          mz_);
-        paramList_->set("model",       model_);
-        paramList_->set("stretchx",    stretchx_);
-        paramList_->set("stretchy",    stretchy_);
-        paramList_->set("stretchz",    stretchz_);
-        paramList_->set("keepBCs",     static_cast<bool>(keepBCs_));
-        paramList_->set("matrixType",  matrixType_);
-        paramList_->set("h",           h_);
-        paramList_->set("delta",       delta_);
-        paramList_->set("PMLx_left",   PMLx_left);
-        paramList_->set("PMLx_right",  PMLx_right);
-        paramList_->set("PMLy_left",   PMLy_left);
-        paramList_->set("PMLy_right",  PMLy_right);
-        paramList_->set("PMLz_left",   PMLz_left);
-        paramList_->set("PMLz_right",  PMLz_right);
-        paramList_->set("omega",       omega_);
-        paramList_->set("shift",       shift_);
-
-        check();
-
-        return *paramList_;
-      }
-
-      //! @name Overridden from Teuchos::Describable
-      //@{
-
-      //! Return a simple one-line description of this object.
-      std::string description() const {
-        std::ostringstream out;
-        out << Teuchos::Describable::description();
-        out << "{type = "  << GetMatrixType() << ", size = " << GetNumGlobalElements() << "} ";
-
-        return out.str();
-      }
-
-      //! Print the object with some verbosity level to an FancyOStream object.
-      void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = verbLevel_default) const {
-        int vl = (verbLevel == Teuchos::VERB_DEFAULT) ? Teuchos::VERB_LOW : verbLevel;
-        if (vl == Teuchos::VERB_NONE)
-          return;
-
-        if (vl == Teuchos::VERB_LOW)
-          out << description() << std::endl;
-        else
-          out << Teuchos::Describable::description() << std::endl;
-
-        if (vl == Teuchos:: VERB_MEDIUM || vl == Teuchos::VERB_HIGH || vl == Teuchos::VERB_EXTREME) {
-          Teuchos::OSTab tab1(out);
-
-          const Teuchos::ParameterList& paramList = GetParameterList();
-          std::string matrixType = paramList.get<std::string>("matrixType");
-          GO nx = paramList.get<GO>("nx");
-          GO ny = paramList.get<GO>("ny");
-          GO nz = paramList.get<GO>("nz");
-
-          out << "Matrix type: "  << matrixType << std::endl
-              << "Problem size: " << GetNumGlobalElements();
-
-          if      (matrixType == "Laplace2D" || matrixType == "Elasticity2D" || matrixType == "Helmholtz2D")  out << " (" << nx << "x" << ny << ")";
-          else if (matrixType == "Laplace3D" || matrixType == "Elasticity3D" || matrixType == "Helmholtz3D")  out << " (" << nx << "x" << ny << "x" << nz << ")";
-
-          out << std::endl;
-        }
-      }
-
-      //@}
-
-    private:
-      // See Teuchos BUG 5249: https://software.sandia.gov/bugzilla/show_bug.cgi?id=5249
-      mutable GO     nx_, ny_, nz_;
-      mutable GO     mx_, my_, mz_;
-      mutable double stretchx_, stretchy_, stretchz_;
-
-      std::string    matrixType_;
-
-      mutable int    keepBCs_;
-
-      mutable double    h_;
-      mutable double    delta_;
-      mutable int       PMLx_left, PMLx_right;
-      mutable int       PMLy_left, PMLy_right;
-      mutable int       PMLz_left, PMLz_right;
-      mutable double    omega_;
-      mutable double    shift_;
-      mutable int       model_;
-
-      // There is a major assumption here:
-      // As soon as somebody call GetParameterList(), we freeze all other variables into the list,
-      // and ignore them. This allows us to make the modification of the list from outside.
-      mutable Teuchos::RCP<Teuchos::ParameterList> paramList_;
-    };
+namespace Xpetra {
+
+// TODO nx/ny/nz == GO or global_size_t ? But what is the best to do?
+
+template <typename GO>
+class Parameters : public Teuchos::VerboseObject<Parameters<GO> >, public Teuchos::Describable {
+ public:
+  Parameters(Teuchos::CommandLineProcessor& clp, GO nx = 16, GO ny = -1, GO nz = -1, const std::string& matrixType = "ADR1D",
+             int keepBCs = 0, double stretchx = 1.0, double stretchy = 1.0, double stretchz = 1.0, double h = 1.0, double delta = 0.0,
+             int PMLXL = 0, int PMLXR = 0, int PMLYL = 0, int PMLYR = 0, int PMLZL = 0, int PMLZR = 0,
+             double omega = 2.0 * M_PI, double shift = 0.5, GO mx = -1, GO my = -1, GO mz = -1, int model = 0)
+    : nx_(nx)
+    , ny_(ny)
+    , nz_(nz)
+    , mx_(mx)
+    , my_(my)
+    , mz_(mz)
+    , stretchx_(stretchx)
+    , stretchy_(stretchy)
+    , stretchz_(stretchz)
+    , matrixType_(matrixType)
+    , keepBCs_(keepBCs)
+    , h_(h)
+    , delta_(delta)
+    , PMLx_left(PMLXL)
+    , PMLx_right(PMLXR)
+    , PMLy_left(PMLYL)
+    , PMLy_right(PMLYR)
+    , PMLz_left(PMLZL)
+    , PMLz_right(PMLZR)
+    , omega_(omega)
+    , shift_(shift)
+    , model_(model) {
+    clp.setOption("nx", &nx_, "mesh points in x-direction.");
+    clp.setOption("ny", &ny_, "mesh points in y-direction.");
+    clp.setOption("nz", &nz_, "mesh points in z-direction.");
+    clp.setOption("mx", &mx_, "processors in x-direction.");
+    clp.setOption("my", &my_, "processors in y-direction.");
+    clp.setOption("mz", &mz_, "processors in z-direction.");
+    clp.setOption("stretchx", &stretchx_, "stretch mesh in x-direction.");
+    clp.setOption("stretchy", &stretchy_, "stretch mesh in y-direction.");
+    clp.setOption("stretchz", &stretchz_, "stretch mesh in z-direction.");
+    clp.setOption("keepBCs", &keepBCs_, "keep Dirichlet boundary rows in matrix (0=false,1=true)");
+    clp.setOption("matrixType", &matrixType_, "matrix type: Laplace1D, Laplace2D, Laplace3D, ...");  // TODO: Star2D, numGlobalElements=...
+    clp.setOption("h", &h_, "mesh width for uniform h");
+    clp.setOption("delta", &delta_, "maximum PML damping value");
+    clp.setOption("PMLx_left", &PMLx_left, "PML grid points in x-direction (left boundary)");
+    clp.setOption("PMLx_right", &PMLx_right, "PML grid points in x-direction (right boundary)");
+    clp.setOption("PMLy_left", &PMLy_left, "PML grid points in y-direction (left boundary)");
+    clp.setOption("PMLy_right", &PMLy_right, "PML grid points in y-direction (right boundary)");
+    clp.setOption("PMLz_left", &PMLz_left, "PML grid points in z-direction (left boundary)");
+    clp.setOption("PMLz_right", &PMLz_right, "PML grid points in z-direction (right boundary)");
+    clp.setOption("omega", &omega_, "angular frequency omega");
+    clp.setOption("shift", &shift_, "complex frequency shift");
+    clp.setOption("mx", &mx_, "processors in x-direction.");
+    clp.setOption("my", &my_, "processors in y-direction.");
+    clp.setOption("mz", &mz_, "processors in z-direction.");
+    clp.setOption("model", &model_, "velocity model");
+  }
+
+  GO GetNumGlobalElements() const {
+    const Teuchos::ParameterList& pL = GetParameterList();
+
+    const std::string& matrixType = pL.get<std::string>("matrixType");
+    const GO nx                   = pL.get<GO>("nx");
+    const GO ny                   = pL.get<GO>("ny");
+    const GO nz                   = pL.get<GO>("nz");
+
+    GO numGlobalElements = -1;
+    if (matrixType == "ADR1D")
+      numGlobalElements = nx;
+
+    else if (matrixType == "ADR2D")
+      numGlobalElements = nx * ny;
+
+    else if (matrixType == "ADR3D")
+      numGlobalElements = nx * ny * nz;
+
+    TEUCHOS_TEST_FOR_EXCEPTION(numGlobalElements < 0, std::runtime_error,
+                               "Gallery: numGlobalElements < 0 (did you forget --ny (or --nz) for 2D (3D) problems?)");
+
+    return numGlobalElements;
+  }
+
+  const std::string& GetMatrixType() const {
+    const Teuchos::ParameterList& paramList = GetParameterList();
+    return paramList.get<std::string>("matrixType");
+  }
+
+  void check() const {}
+
+  Teuchos::ParameterList& GetParameterList() const {
+    if (!paramList_.is_null())
+      return *paramList_;
+
+    paramList_ = rcp(new Teuchos::ParameterList());
+
+    paramList_->set("nx", nx_);
+    paramList_->set("ny", ny_);
+    paramList_->set("nz", nz_);
+    paramList_->set("mx", mx_);
+    paramList_->set("my", my_);
+    paramList_->set("mz", mz_);
+    paramList_->set("model", model_);
+    paramList_->set("stretchx", stretchx_);
+    paramList_->set("stretchy", stretchy_);
+    paramList_->set("stretchz", stretchz_);
+    paramList_->set("keepBCs", static_cast<bool>(keepBCs_));
+    paramList_->set("matrixType", matrixType_);
+    paramList_->set("h", h_);
+    paramList_->set("delta", delta_);
+    paramList_->set("PMLx_left", PMLx_left);
+    paramList_->set("PMLx_right", PMLx_right);
+    paramList_->set("PMLy_left", PMLy_left);
+    paramList_->set("PMLy_right", PMLy_right);
+    paramList_->set("PMLz_left", PMLz_left);
+    paramList_->set("PMLz_right", PMLz_right);
+    paramList_->set("omega", omega_);
+    paramList_->set("shift", shift_);
+
+    check();
+
+    return *paramList_;
+  }
+
+  //! @name Overridden from Teuchos::Describable
+  //@{
+
+  //! Return a simple one-line description of this object.
+  std::string description() const {
+    std::ostringstream out;
+    out << Teuchos::Describable::description();
+    out << "{type = " << GetMatrixType() << ", size = " << GetNumGlobalElements() << "} ";
 
+    return out.str();
   }
-}
+
+  //! Print the object with some verbosity level to an FancyOStream object.
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = verbLevel_default) const {
+    int vl = (verbLevel == Teuchos::VERB_DEFAULT) ? Teuchos::VERB_LOW : verbLevel;
+    if (vl == Teuchos::VERB_NONE)
+      return;
+
+    if (vl == Teuchos::VERB_LOW)
+      out << description() << std::endl;
+    else
+      out << Teuchos::Describable::description() << std::endl;
+
+    if (vl == Teuchos::VERB_MEDIUM || vl == Teuchos::VERB_HIGH || vl == Teuchos::VERB_EXTREME) {
+      Teuchos::OSTab tab1(out);
+
+      const Teuchos::ParameterList& paramList = GetParameterList();
+      std::string matrixType                  = paramList.get<std::string>("matrixType");
+      GO nx                                   = paramList.get<GO>("nx");
+      GO ny                                   = paramList.get<GO>("ny");
+      GO nz                                   = paramList.get<GO>("nz");
+
+      out << "Matrix type: " << matrixType << std::endl
+          << "Problem size: " << GetNumGlobalElements();
+
+      if (matrixType == "Laplace2D" || matrixType == "Elasticity2D" || matrixType == "Helmholtz2D")
+        out << " (" << nx << "x" << ny << ")";
+      else if (matrixType == "Laplace3D" || matrixType == "Elasticity3D" || matrixType == "Helmholtz3D")
+        out << " (" << nx << "x" << ny << "x" << nz << ")";
+
+      out << std::endl;
+    }
+  }
+
+  //@}
+
+ private:
+  // See Teuchos BUG 5249: https://software.sandia.gov/bugzilla/show_bug.cgi?id=5249
+  mutable GO nx_, ny_, nz_;
+  mutable GO mx_, my_, mz_;
+  mutable double stretchx_, stretchy_, stretchz_;
+
+  std::string matrixType_;
+
+  mutable int keepBCs_;
+
+  mutable double h_;
+  mutable double delta_;
+  mutable int PMLx_left, PMLx_right;
+  mutable int PMLy_left, PMLy_right;
+  mutable int PMLz_left, PMLz_right;
+  mutable double omega_;
+  mutable double shift_;
+  mutable int model_;
+
+  // There is a major assumption here:
+  // As soon as somebody call GetParameterList(), we freeze all other variables into the list,
+  // and ignore them. This allows us to make the modification of the list from outside.
+  mutable Teuchos::RCP<Teuchos::ParameterList> paramList_;
+};
+
+}  // namespace Xpetra
+}  // namespace ADR
 
 #endif
diff --git a/packages/muelu/research/max/AdditiveMG/BAP.hpp b/packages/muelu/research/max/AdditiveMG/BAP.hpp
index 72344805ce6d..62faea3a20de 100644
--- a/packages/muelu/research/max/AdditiveMG/BAP.hpp
+++ b/packages/muelu/research/max/AdditiveMG/BAP.hpp
@@ -3,250 +3,222 @@
 
 #include "neighbours.hpp"
 
-
-typedef Tpetra::CrsMatrix<double,int,int,KokkosClassic::DefaultNode::DefaultNodeType>     tpetra_matrix_type;
-typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,int,int,KokkosClassic::DefaultNode::DefaultNodeType>   tpetra_multivector_type;
-typedef typename Teuchos::ArrayView<const int>::const_iterator                            iterator_type;
+typedef Tpetra::CrsMatrix<double, int, int, KokkosClassic::DefaultNode::DefaultNodeType> tpetra_matrix_type;
+typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, int, int, KokkosClassic::DefaultNode::DefaultNodeType> tpetra_multivector_type;
+typedef typename Teuchos::ArrayView<const int>::const_iterator iterator_type;
 typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type2;
 
-void BAP1D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP< const Teuchos::Comm<int> > comm)
-{
-
-  //INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
-  //INPUT: tpetra_prolong = Ptentative
-  //INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  
- Teuchos::ArrayView<const int> myLocalElements = BAP->getRowMap()->getLocalElementList();
- int mypid = comm->getRank();
-
-	for(int color = 0; color<3; ++color)
-	{
-	  Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
-
-		for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it)
-		{
-		 	const int i_local = *it;
-			const int aux = BAP->getRowMap()->getLocalElement (i_local);
-
-			std::vector<int> BAP_inds;
-			std::vector<double> BAP_vals;
-
-			int aux2;
-
-			if( (mypid-1)%3==color && (mypid-1)>= 0 && (mypid-1)<tpetra_prolong->getGlobalNumCols() )
-				aux2 = BAP->getColMap()->getLocalElement (mypid-1);
-			else if( (mypid-2)%3==color && (mypid-2)>= 0 && (mypid-2)<tpetra_prolong->getGlobalNumCols() )
-				aux2 = BAP->getColMap()->getLocalElement (mypid-2);
-			else if( (mypid)%3==color && (mypid)>= 0 && (mypid)<tpetra_prolong->getGlobalNumCols() )
-				aux2 = BAP->getColMap()->getLocalElement (mypid);
-
-			if(aux2>=0)
-			{
-				BAP_inds.emplace_back(aux2);
-				BAP_vals.emplace_back(localBAP[aux]);
-				BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
-			}
-		}
-
-	}
+void BAP1D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+  // INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
+  // INPUT: tpetra_prolong = Ptentative
+  // INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
 
+  Teuchos::ArrayView<const int> myLocalElements = BAP->getRowMap()->getLocalElementList();
+  int mypid                                     = comm->getRank();
+
+  for (int color = 0; color < 3; ++color) {
+    Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
+
+    for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it) {
+      const int i_local = *it;
+      const int aux     = BAP->getRowMap()->getLocalElement(i_local);
+
+      std::vector<int> BAP_inds;
+      std::vector<double> BAP_vals;
+
+      int aux2;
+
+      if ((mypid - 1) % 3 == color && (mypid - 1) >= 0 && (mypid - 1) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid - 1);
+      else if ((mypid - 2) % 3 == color && (mypid - 2) >= 0 && (mypid - 2) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid - 2);
+      else if ((mypid) % 3 == color && (mypid) >= 0 && (mypid) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid);
+
+      if (aux2 >= 0) {
+        BAP_inds.emplace_back(aux2);
+        BAP_vals.emplace_back(localBAP[aux]);
+        BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
+      }
+    }
+  }
 }
 
+void BAP2D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP<const Teuchos::Comm<int> > comm, int ndx) {
+  // INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
+  // INPUT: tpetra_prolong = Ptentative
+  // INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
+  // INPUT: ndx = number of domains along x-direction
 
-void BAP2D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP< const Teuchos::Comm<int> > comm, int ndx)
-{
-
-  //INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
-  //INPUT: tpetra_prolong = Ptentative
-  //INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  //INPUT: ndx = number of domains along x-direction
-
- Teuchos::ArrayView<const int> myLocalElements = BAP->getRowMap()->getLocalElementList();
- int mypid = comm->getRank();
- int brick_id = mypid;
- int shifted_id = brick_id-1;
+  Teuchos::ArrayView<const int> myLocalElements = BAP->getRowMap()->getLocalElementList();
+  int mypid                                     = comm->getRank();
+  int brick_id                                  = mypid;
+  int shifted_id                                = brick_id - 1;
 
- if(mypid>0)
- {
-	for(int color = 0; color<9; ++color)
-	{
-                int neighbour = -1;
-	        Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
+  if (mypid > 0) {
+    for (int color = 0; color < 9; ++color) {
+      int neighbour                            = -1;
+      Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
 
-                //The following if statements control the neighbours of a subdomain in a 2D brick partitioned mesh
-		if( coloring2D( brick_id, ndx )==color && (shifted_id)>= 0 && (shifted_id)<tpetra_prolong->getGlobalNumCols() )
-		  neighbour = (shifted_id);
+      // The following if statements control the neighbours of a subdomain in a 2D brick partitioned mesh
+      if (coloring2D(brick_id, ndx) == color && (shifted_id) >= 0 && (shifted_id) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id);
 
-		else if( coloring2D( brick_id-1, ndx )==color && (shifted_id-1)>= 0 && (shifted_id-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1);
+      else if (coloring2D(brick_id - 1, ndx) == color && (shifted_id - 1) >= 0 && (shifted_id - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1);
 
-		else if( coloring2D( brick_id+1, ndx )==color && (shifted_id+1)>= 0 && (shifted_id+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1);
+      else if (coloring2D(brick_id + 1, ndx) == color && (shifted_id + 1) >= 0 && (shifted_id + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1);
 
-		else if( coloring2D( brick_id-ndx, ndx )==color && (shifted_id-ndx)>= 0 && (shifted_id-ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx);
+      else if (coloring2D(brick_id - ndx, ndx) == color && (shifted_id - ndx) >= 0 && (shifted_id - ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx);
 
-		else if( coloring2D( brick_id-ndx-1, ndx )==color && (shifted_id-ndx-1)>= 0 && (shifted_id-ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1);
+      else if (coloring2D(brick_id - ndx - 1, ndx) == color && (shifted_id - ndx - 1) >= 0 && (shifted_id - ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1);
 
-		else if( coloring2D( brick_id-ndx+1, ndx )==color && (shifted_id-ndx+1)>= 0 && (shifted_id-ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1);
+      else if (coloring2D(brick_id - ndx + 1, ndx) == color && (shifted_id - ndx + 1) >= 0 && (shifted_id - ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1);
 
-		else if( coloring2D( brick_id+ndx, ndx )==color && (shifted_id+ndx)>= 0 && (shifted_id+ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx);
+      else if (coloring2D(brick_id + ndx, ndx) == color && (shifted_id + ndx) >= 0 && (shifted_id + ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx);
 
-		else if( coloring2D( brick_id+ndx-1, ndx )==color && (shifted_id+ndx-1)>= 0 && (shifted_id+ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1);
+      else if (coloring2D(brick_id + ndx - 1, ndx) == color && (shifted_id + ndx - 1) >= 0 && (shifted_id + ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1);
 
-		else if( coloring2D( brick_id+ndx+1, ndx )==color && (shifted_id+ndx+1)>= 0 && (shifted_id+ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1);
+      else if (coloring2D(brick_id + ndx + 1, ndx) == color && (shifted_id + ndx + 1) >= 0 && (shifted_id + ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1);
 
-                //in case neighbour>=0, it means that the current MPI processor (=subdomain) has a neighbour with the color analyzed at the current for-loop iteration
-                //otherwise the current MPI processor sits on the boundary of the mesh and it has less than 8 neighbours.
-                if(neighbour>=0)
-                {
-			for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it)
-			{
-				const int i_local = *it;
-				const int aux = BAP->getRowMap()->getLocalElement (i_local);
+      // in case neighbour>=0, it means that the current MPI processor (=subdomain) has a neighbour with the color analyzed at the current for-loop iteration
+      // otherwise the current MPI processor sits on the boundary of the mesh and it has less than 8 neighbours.
+      if (neighbour >= 0) {
+        for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it) {
+          const int i_local = *it;
+          const int aux     = BAP->getRowMap()->getLocalElement(i_local);
 
-				std::vector<int> BAP_inds;
-				std::vector<double> BAP_vals;
+          std::vector<int> BAP_inds;
+          std::vector<double> BAP_vals;
 
-				int aux2;
+          int aux2;
 
-			       aux2 = BAP->getColMap()->getLocalElement (neighbour);
-				
-				if(aux2>=0)
-				{
-					BAP_inds.emplace_back(aux2);
-					BAP_vals.emplace_back(localBAP[aux]);
-					BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
-				}
-			}
-		}
+          aux2 = BAP->getColMap()->getLocalElement(neighbour);
 
-	}
- }
+          if (aux2 >= 0) {
+            BAP_inds.emplace_back(aux2);
+            BAP_vals.emplace_back(localBAP[aux]);
+            BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
+          }
+        }
+      }
+    }
+  }
 }
 
-void BAP3D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP< const Teuchos::Comm<int> > comm, int ndx, int ndy)
-{
-
-  //INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
-  //INPUT: tpetra_prolong = Ptentative
-  //INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  //INPUT: ndx = number of domains along x-direction
-  //INPUT: ndy = number of domains along y-direction
+void BAP3D(Teuchos::RCP<tpetra_matrix_type> BAP, Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, Teuchos::RCP<tpetra_multivector_type> BAP_shrunk, Teuchos::RCP<const Teuchos::Comm<int> > comm, int ndx, int ndy) {
+  // INPUT: BAP = matrix where the uncompressed version of B_DD * A_h * Ptentative must be stored
+  // INPUT: tpetra_prolong = Ptentative
+  // INPUT: BAP_shrunk = Tpetra:MultiVector contatining the shrunk version of B_DD * A_h * Ptentative resulting from domain decomposition with coloring
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
+  // INPUT: ndx = number of domains along x-direction
+  // INPUT: ndy = number of domains along y-direction
 
   Teuchos::ArrayView<const int> myLocalElements = BAP->getRowMap()->getLocalElementList();
-  int mypid = comm->getRank();
-  int brick_id = mypid;
-  int shifted_id = brick_id - 1;
-
-  if(mypid>0)
-  {
-	for(int color = 0; color<27; ++color)
-	{
-		int neighbour = -1;
-		Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
-
-                //The following if statements control the neighbours of a subdomain in a 3D brick partitioned mesh
-                //Each subdomains is incorporated in a 3x3x3 cube which is sliced into 3 squares living on three different planes
-                //The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
-                //
-                //Identification of neighbours that are on the current plane
-		if( coloring3D( brick_id, ndx, ndy )==color && (shifted_id)>= 0 && (shifted_id)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id);
-		else if( coloring3D( brick_id-1, ndx, ndy )==color && (shifted_id-1)>= 0 && (shifted_id-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1);
-		else if( coloring3D( brick_id+1, ndx, ndy )==color && (shifted_id+1)>= 0 && (shifted_id+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1);
-		else if( coloring3D( brick_id-ndx, ndx, ndy )==color && (shifted_id-ndx)>= 0 && (shifted_id-ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx);
-		else if( coloring3D( brick_id-ndx-1, ndx, ndy )==color && (shifted_id-ndx-1)>= 0 && (shifted_id-ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1);
-		else if( coloring3D( brick_id-ndx+1, ndx, ndy )==color && (shifted_id-ndx+1)>= 0 && (shifted_id-ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1);
-		else if( coloring3D( brick_id+ndx, ndx, ndy )==color && (shifted_id+ndx)>= 0 && (shifted_id+ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx);
-		else if( coloring3D( brick_id+ndx-1, ndx, ndy )==color && (shifted_id+ndx-1)>= 0 && (shifted_id+ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1);
-		else if( coloring3D( brick_id+ndx+1, ndx, ndy )==color && (shifted_id+ndx+1)>= 0 && (shifted_id+ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1);
-
-                //Identification of the neighbours that are on the plane below
-		else if( coloring3D( brick_id-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx*ndy)>= 0 && (shifted_id-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx*ndy);
-		else if( coloring3D( brick_id-1-ndx*ndy, ndx, ndy )==color && (shifted_id-1-ndx*ndy)>= 0 && (shifted_id-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1-ndx*ndy);
-		else if( coloring3D( brick_id+1-ndx*ndy, ndx, ndy )==color && (shifted_id+1-ndx*ndy)>= 0 && (shifted_id+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1-ndx*ndy);
-		else if( coloring3D( brick_id-ndx-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-ndx*ndy)>= 0 && (shifted_id-ndx-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-ndx*ndy);
-		else if( coloring3D( brick_id-ndx-1-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-1-ndx*ndy)>= 0 && (shifted_id-ndx-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1-ndx*ndy);
-		else if( coloring3D( brick_id-ndx+1-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+1-ndx*ndy)>= 0 && (shifted_id-ndx+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1-ndx*ndy);
-		else if( coloring3D( brick_id+ndx-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-ndx*ndy)>= 0 && (shifted_id+ndx-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-ndx*ndy);
-		else if( coloring3D( brick_id+ndx-1-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-1-ndx*ndy)>= 0 && (shifted_id+ndx-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1-ndx*ndy);
-		else if( coloring3D( brick_id+ndx+1-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+1-ndx*ndy)>= 0 && (shifted_id+ndx+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1-ndx*ndy);
-
-                //Identification of the neighbours that are on the plane above
-		else if( coloring3D( brick_id+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx*ndy)>= 0 && (shifted_id+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx*ndy);
-		else if( coloring3D( brick_id-1+ndx*ndy, ndx, ndy )==color && (shifted_id-1+ndx*ndy)>= 0 && (shifted_id-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1+ndx*ndy);
-		else if( coloring3D( brick_id+1+ndx*ndy, ndx, ndy )==color && (shifted_id+1+ndx*ndy)>= 0 && (shifted_id+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1+ndx*ndy);
-		else if( coloring3D( brick_id-ndx+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+ndx*ndy)>= 0 && (shifted_id-ndx+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+ndx*ndy);
-		else if( coloring3D( brick_id-ndx-1+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-1+ndx*ndy)>= 0 && (shifted_id-ndx-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1+ndx*ndy);
-		else if( coloring3D( brick_id-ndx+1+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+1+ndx*ndy)>= 0 && (shifted_id-ndx+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1+ndx*ndy);
-		else if( coloring3D( brick_id+ndx+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+ndx*ndy)>= 0 && (shifted_id+ndx+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+ndx*ndy);
-		else if( coloring3D( brick_id+ndx-1+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-1+ndx*ndy)>= 0 && (shifted_id+ndx-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1+ndx*ndy);
-		else if( coloring3D( brick_id+ndx+1+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+1+ndx*ndy)>= 0 && (shifted_id+ndx+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1+ndx*ndy);
-
-                //in case neighbour>=0, it means that the current MPI processor (=subdomain) has a neighbour with the color analyzed at the current for-loop iteration
-                //otherwise the current MPI processor sits on the boundary of the mesh and it has less than 26 neighbours.
-                if(neighbour>=0)
-                {
-			for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it)
-			{
-				const int i_local = *it;
-				const int aux = BAP->getRowMap()->getLocalElement (i_local);
-
-				std::vector<int> BAP_inds;
-				std::vector<double> BAP_vals;
-
-				int aux2;
-
-			       aux2 = BAP->getColMap()->getLocalElement (neighbour);
-				
-				if(aux2>=0)
-				{
-					BAP_inds.emplace_back(aux2);
-					BAP_vals.emplace_back(localBAP[aux]);
-					BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
-				}
-                                else
-				std::cout<<"ID: "<<mypid<<" does not reach "<<neighbour+1<<std::endl;
-			}
-		}
-
-	}
-   }
+  int mypid                                     = comm->getRank();
+  int brick_id                                  = mypid;
+  int shifted_id                                = brick_id - 1;
+
+  if (mypid > 0) {
+    for (int color = 0; color < 27; ++color) {
+      int neighbour                            = -1;
+      Teuchos::ArrayRCP<const double> localBAP = BAP_shrunk->getData(color);
+
+      // The following if statements control the neighbours of a subdomain in a 3D brick partitioned mesh
+      // Each subdomains is incorporated in a 3x3x3 cube which is sliced into 3 squares living on three different planes
+      // The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
+      //
+      // Identification of neighbours that are on the current plane
+      if (coloring3D(brick_id, ndx, ndy) == color && (shifted_id) >= 0 && (shifted_id) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id);
+      else if (coloring3D(brick_id - 1, ndx, ndy) == color && (shifted_id - 1) >= 0 && (shifted_id - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1);
+      else if (coloring3D(brick_id + 1, ndx, ndy) == color && (shifted_id + 1) >= 0 && (shifted_id + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1);
+      else if (coloring3D(brick_id - ndx, ndx, ndy) == color && (shifted_id - ndx) >= 0 && (shifted_id - ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx);
+      else if (coloring3D(brick_id - ndx - 1, ndx, ndy) == color && (shifted_id - ndx - 1) >= 0 && (shifted_id - ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1);
+      else if (coloring3D(brick_id - ndx + 1, ndx, ndy) == color && (shifted_id - ndx + 1) >= 0 && (shifted_id - ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1);
+      else if (coloring3D(brick_id + ndx, ndx, ndy) == color && (shifted_id + ndx) >= 0 && (shifted_id + ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx);
+      else if (coloring3D(brick_id + ndx - 1, ndx, ndy) == color && (shifted_id + ndx - 1) >= 0 && (shifted_id + ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1);
+      else if (coloring3D(brick_id + ndx + 1, ndx, ndy) == color && (shifted_id + ndx + 1) >= 0 && (shifted_id + ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1);
+
+      // Identification of the neighbours that are on the plane below
+      else if (coloring3D(brick_id - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx * ndy) >= 0 && (shifted_id - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx * ndy);
+      else if (coloring3D(brick_id - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - 1 - ndx * ndy) >= 0 && (shifted_id - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1 - ndx * ndy);
+      else if (coloring3D(brick_id + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + 1 - ndx * ndy) >= 0 && (shifted_id + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1 - ndx * ndy);
+      else if (coloring3D(brick_id - ndx - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - ndx * ndy) >= 0 && (shifted_id - ndx - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - ndx * ndy);
+      else if (coloring3D(brick_id - ndx - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - 1 - ndx * ndy) >= 0 && (shifted_id - ndx - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1 - ndx * ndy);
+      else if (coloring3D(brick_id - ndx + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + 1 - ndx * ndy) >= 0 && (shifted_id - ndx + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1 - ndx * ndy);
+      else if (coloring3D(brick_id + ndx - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - ndx * ndy) >= 0 && (shifted_id + ndx - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - ndx * ndy);
+      else if (coloring3D(brick_id + ndx - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - 1 - ndx * ndy) >= 0 && (shifted_id + ndx - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1 - ndx * ndy);
+      else if (coloring3D(brick_id + ndx + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + 1 - ndx * ndy) >= 0 && (shifted_id + ndx + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1 - ndx * ndy);
+
+      // Identification of the neighbours that are on the plane above
+      else if (coloring3D(brick_id + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx * ndy) >= 0 && (shifted_id + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx * ndy);
+      else if (coloring3D(brick_id - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - 1 + ndx * ndy) >= 0 && (shifted_id - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1 + ndx * ndy);
+      else if (coloring3D(brick_id + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + 1 + ndx * ndy) >= 0 && (shifted_id + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1 + ndx * ndy);
+      else if (coloring3D(brick_id - ndx + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + ndx * ndy) >= 0 && (shifted_id - ndx + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + ndx * ndy);
+      else if (coloring3D(brick_id - ndx - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - 1 + ndx * ndy) >= 0 && (shifted_id - ndx - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1 + ndx * ndy);
+      else if (coloring3D(brick_id - ndx + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + 1 + ndx * ndy) >= 0 && (shifted_id - ndx + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1 + ndx * ndy);
+      else if (coloring3D(brick_id + ndx + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + ndx * ndy) >= 0 && (shifted_id + ndx + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + ndx * ndy);
+      else if (coloring3D(brick_id + ndx - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - 1 + ndx * ndy) >= 0 && (shifted_id + ndx - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1 + ndx * ndy);
+      else if (coloring3D(brick_id + ndx + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + 1 + ndx * ndy) >= 0 && (shifted_id + ndx + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1 + ndx * ndy);
+
+      // in case neighbour>=0, it means that the current MPI processor (=subdomain) has a neighbour with the color analyzed at the current for-loop iteration
+      // otherwise the current MPI processor sits on the boundary of the mesh and it has less than 26 neighbours.
+      if (neighbour >= 0) {
+        for (iterator_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it) {
+          const int i_local = *it;
+          const int aux     = BAP->getRowMap()->getLocalElement(i_local);
+
+          std::vector<int> BAP_inds;
+          std::vector<double> BAP_vals;
+
+          int aux2;
+
+          aux2 = BAP->getColMap()->getLocalElement(neighbour);
+
+          if (aux2 >= 0) {
+            BAP_inds.emplace_back(aux2);
+            BAP_vals.emplace_back(localBAP[aux]);
+            BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
+          } else
+            std::cout << "ID: " << mypid << " does not reach " << neighbour + 1 << std::endl;
+        }
+      }
+    }
+  }
 }
-
-
diff --git a/packages/muelu/research/max/AdditiveMG/CreateADRMatrix.hpp b/packages/muelu/research/max/AdditiveMG/CreateADRMatrix.hpp
index e936baf87926..f233e57f4cc9 100644
--- a/packages/muelu/research/max/AdditiveMG/CreateADRMatrix.hpp
+++ b/packages/muelu/research/max/AdditiveMG/CreateADRMatrix.hpp
@@ -54,439 +54,462 @@
 
 namespace ADR {
 
-  namespace Xpetra {
+namespace Xpetra {
 
-    // =============================================  ADR1D  =============================================
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    class ADR1DProblem : public Problem<Map,Matrix,MultiVector> {
-    public:
-      ADR1DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map) : Problem<Map,Matrix,MultiVector>(list, map) { }
-      Teuchos::RCP<Matrix> BuildMatrix();
+// =============================================  ADR1D  =============================================
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+class ADR1DProblem : public Problem<Map, Matrix, MultiVector> {
+ public:
+  ADR1DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map)
+    : Problem<Map, Matrix, MultiVector>(list, map) {}
+  Teuchos::RCP<Matrix> BuildMatrix();
 
-		private:
+ private:
+  // domain definition
+  Scalar xleft  = 0.0;
+  Scalar xright = 1.0;
 
-			//domain definition
-			Scalar xleft = 0.0;
-			Scalar xright = 1.0;
+  // Function that defines the diffusion coefficient
+  inline Scalar diff(Scalar x) { return 1.0 + x; };
 
-			//Function that defines the diffusion coefficient
-			inline Scalar diff( Scalar  x){return 1.0 + x;};
+  // Function that defines the first derivative of the diffusion coefficient
+  inline Scalar diff_prime(Scalar x) { return 1.0; };
 
-			//Function that defines the first derivative of the diffusion coefficient
-			inline Scalar diff_prime( Scalar x ){return 1.0;};
+  // Function that defines the advection coefficient
+  inline Scalar adv(Scalar x) { return 10.0 * x; };
 
-			//Function that defines the advection coefficient
-			inline Scalar adv( Scalar x){return 10.0 * x;};
+  // Function taht defines the reaction coefficient
+  inline Scalar reac(Scalar x) { return 0.0; };
+};
 
-			//Function taht defines the reaction coefficient
-			inline Scalar reac( Scalar x ){return 0.0;};
-    };
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+Teuchos::RCP<Matrix> ADR1DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>::BuildMatrix() {
+  GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal)-1);
 
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    Teuchos::RCP<Matrix> ADR1DProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>::BuildMatrix() {
-      GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal) -1);
+  if (nx == -1)
+    nx = this->Map_->getGlobalNumElements();
 
-      if (nx == -1)
-        nx = this->Map_->getGlobalNumElements();
+  const Scalar dx = (Scalar)(xright - xleft) / static_cast<Scalar>(nx - 1);
 
-			const Scalar dx = (Scalar) (xright - xleft)/static_cast<Scalar>(nx - 1);
+  const Scalar a = 2.0;
+  const Scalar b = -1.0;
+  const Scalar c = -1.0;
 
-      const Scalar a = 2.0;
-      const Scalar b = -1.0;
-      const Scalar c = -1.0;
+  // this->A_ = Galeri::Xpetra::TriDiag<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, 2.0, -1.0, -1.0);
+  Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map, Matrix>::Build(this->Map_, 3);
 
-      //this->A_ = Galeri::Xpetra::TriDiag<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, 2.0, -1.0, -1.0);
-      Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map,Matrix>::Build(this->Map_, 3);
+  LocalOrdinal NumMyElements                               = this->Map_->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = this->Map_->getLocalElementList();
+  GlobalOrdinal indexBase                                  = this->Map_->getIndexBase();
 
-      LocalOrdinal NumMyElements = this->Map_->getLocalNumElements();
-      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = this->Map_->getLocalElementList();
-      GlobalOrdinal indexBase = this->Map_->getIndexBase();
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = this->Map_->getComm();
 
-      Teuchos::RCP<const Teuchos::Comm<int> > comm = this->Map_->getComm();
+  GlobalOrdinal NumGlobalElements = this->Map_->getGlobalNumElements();
 
-      GlobalOrdinal NumGlobalElements = this->Map_->getGlobalNumElements();
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
 
-      GlobalOrdinal NumEntries;
-      LocalOrdinal  nnz = 2;
-      std::vector<Scalar>        Values(nnz);
-      std::vector<GlobalOrdinal> Indices(nnz);
+  comm->barrier();
 
-      comm->barrier();
+  Teuchos::RCP<Teuchos::Time> timer = Teuchos::rcp(new Teuchos::Time("1D Assembler global insert"));
+  timer->start(true);
 
-      Teuchos::RCP<Teuchos::Time> timer = Teuchos::rcp(new Teuchos::Time("1D Assembler global insert"));
-      timer->start(true);
+  // c a b
+  for (LocalOrdinal i = 0; i < NumMyElements; i++) {
+    if (MyGlobalElements[i] == indexBase) {
+      // off-diagonal for first row
+      Indices[0] = 1 + indexBase;
 
-      // c a b
-      for (LocalOrdinal i = 0; i < NumMyElements; i++) {
-        if (MyGlobalElements[i] == indexBase) {
-          // off-diagonal for first row
-          Indices[0] = 1 + indexBase;
+      // Diffusive term
+      Values[0] = b * diff(xleft + MyGlobalElements[i] * dx) / (dx * dx);
+      Values[0] = Values[0] - diff_prime(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-					//Diffusive term
-          Values [0] = b * diff( xleft + MyGlobalElements[i]*dx )/(dx * dx);
-					Values [0] = Values[0] -  diff_prime( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
+      // Advective term
+      Values[0] = Values[0] + adv(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-					//Advective term
-					Values [0] = Values[0] + adv( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
+      NumEntries = 1;
 
-          NumEntries = 1;
+    } else if (MyGlobalElements[i] == NumGlobalElements + indexBase - 1) {
+      // off-diagonal for last row
+      Indices[0] = NumGlobalElements - 2 + indexBase;
 
-        } else if (MyGlobalElements[i] == NumGlobalElements + indexBase - 1) {
-          // off-diagonal for last row
-          Indices[0] = NumGlobalElements - 2 + indexBase;
+      // Diffusive term
+      Values[0] = c * diff(xleft + MyGlobalElements[i] * dx) / (dx * dx);
+      Values[0] = Values[0] + diff_prime(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-					//Diffusive term
-          Values [0] = c * diff( xleft + MyGlobalElements[i]*dx )/(dx * dx);
-					Values [0] = Values[0] +  diff_prime( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
+      // Advective term
+      Values[0] = Values[0] - adv(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-					//Advective term
-					Values [0] = Values[0] - adv( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
+      NumEntries = 1;
 
-          NumEntries = 1;
+    } else {
+      // off-diagonal for internal row
+      Indices[0] = MyGlobalElements[i] - 1;
+      Indices[1] = MyGlobalElements[i] + 1;
 
-        } else {
-          // off-diagonal for internal row
-          Indices[0] = MyGlobalElements[i] - 1;
-					Indices[1] = MyGlobalElements[i] + 1;
+      // Diffusive term
+      Values[0] = c * diff(xleft + MyGlobalElements[i] * dx) / (dx * dx);
+      Values[0] = Values[0] + diff_prime(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
+      Values[1] = b * diff(xleft + MyGlobalElements[i] * dx) / (dx * dx);
+      Values[1] = Values[1] - diff_prime(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-					//Diffusive term
-          Values [0] = c * diff( xleft + MyGlobalElements[i]*dx )/(dx * dx);
-					Values [0] = Values[0] +  diff_prime( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
-          Values [1] = b * diff( xleft + MyGlobalElements[i]*dx )/(dx * dx);
-					Values [1] = Values[1] -  diff_prime( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
-  
-					//Advective term
-					Values [0] = Values[0] - adv( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
-					Values [1] = Values[1] + adv( xleft + MyGlobalElements[i]*dx )/(2.0 * dx);
+      // Advective term
+      Values[0] = Values[0] - adv(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
+      Values[1] = Values[1] + adv(xleft + MyGlobalElements[i] * dx) / (2.0 * dx);
 
-          NumEntries = 2;
-        }
-
-        // put the off-diagonal entries
-        // Xpetra wants ArrayViews (sigh)
-        Teuchos::ArrayView<Scalar>        av(&Values [0], NumEntries);
-        Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
-        mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
-
-        // Put in the diagonal entry
-				//Diffusion
-				Scalar diag_entry = a * diff( xleft + MyGlobalElements[i]*dx )/(dx * dx);
-				//Reaction
-				diag_entry = diag_entry + reac( xleft + MyGlobalElements[i]*dx );
-
-        mtx->insertGlobalValues(MyGlobalElements[i],
-                                Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-                                Teuchos::tuple<Scalar>( diag_entry ));
-      }
-
-      timer->stop();
-
-      timer = Teuchos::rcp(new Teuchos::Time("1D Assembler fillComplete"));
-      timer->start(true);
-
-      mtx->fillComplete();
-      this->A_ = mtx;
-
-      timer->stop();
-
-      this->A_->setObjectLabel(this->getObjectLabel());
-      return this->A_;
+      NumEntries = 2;
     }
 
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+
+    // Put in the diagonal entry
+    // Diffusion
+    Scalar diag_entry = a * diff(xleft + MyGlobalElements[i] * dx) / (dx * dx);
+    // Reaction
+    diag_entry = diag_entry + reac(xleft + MyGlobalElements[i] * dx);
+
+    mtx->insertGlobalValues(MyGlobalElements[i],
+                            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                            Teuchos::tuple<Scalar>(diag_entry));
+  }
 
-    // =============================================  ADR2D  =============================================
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    class ADR2DProblem : public Problem<Map,Matrix,MultiVector> {
-    public:
-      ADR2DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map) : Problem<Map,Matrix,MultiVector>(list, map) { }
-      Teuchos::RCP<Matrix> BuildMatrix();
-
-		private:
-
-			//Function that defines the diffusion coefficient
-			inline Scalar diff( Scalar  x, Scalar y ){return 1.0 + 0.0 * x + 0.0 * y;};
-
-			//Function that defines the first derivative in x-direction of the diffusion coefficient
-			inline Scalar diff_primex( Scalar  x, Scalar y ){return 0.0 + 0.0 * x + 0.0 * y;};
+  timer->stop();
+
+  timer = Teuchos::rcp(new Teuchos::Time("1D Assembler fillComplete"));
+  timer->start(true);
 
-			//Function that defines the first derivative in x-direction of the diffusion coefficient
-			inline Scalar diff_primey( Scalar  x, Scalar y ){return 0.0 + 0.0 * x + 0.0 * y;};
-
-			//Function that defines the advection coefficient in the x-direction
-			inline Scalar advx( Scalar  x, Scalar y ){return 10.0 + 0.0 * x + 0.0 * y;};
-
-			//Function that defines the advection coefficient in the x-direction
-			inline Scalar advy( Scalar  x, Scalar y ){return 10.0 + 0.0 * x + 0.0 * y;};
-
-			//Function taht defines the reaction coefficient
-			inline Scalar reac( Scalar  x, Scalar y ){return 0.0 + 0.0 * x + 0.0 * y;};
-
-    };
-
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    Teuchos::RCP<Matrix> ADR2DProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>::BuildMatrix() {
-      GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal) -1);
-      GlobalOrdinal ny = this->list_.get("ny", (GlobalOrdinal) -1);
-      double  one = 1.0;
-      Scalar  stretchx = (Scalar) this->list_.get("stretchx", one);
-      Scalar  stretchy = (Scalar) this->list_.get("stretchy", one);
-
-      if (nx == -1 || ny == -1) {
-        GlobalOrdinal n = this->Map_->getGlobalNumElements();
-        nx = (GlobalOrdinal)sqrt((double)n);
-        ny = nx;
-        TEUCHOS_TEST_FOR_EXCEPTION(nx*ny != n, std::logic_error, "You need to specify nx and ny.");
+  mtx->fillComplete();
+  this->A_ = mtx;
+
+  timer->stop();
+
+  this->A_->setObjectLabel(this->getObjectLabel());
+  return this->A_;
+}
+
+// =============================================  ADR2D  =============================================
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+class ADR2DProblem : public Problem<Map, Matrix, MultiVector> {
+ public:
+  ADR2DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map)
+    : Problem<Map, Matrix, MultiVector>(list, map) {}
+  Teuchos::RCP<Matrix> BuildMatrix();
+
+ private:
+  // Function that defines the diffusion coefficient
+  inline Scalar diff(Scalar x, Scalar y) { return 1.0 + 0.0 * x + 0.0 * y; };
+
+  // Function that defines the first derivative in x-direction of the diffusion coefficient
+  inline Scalar diff_primex(Scalar x, Scalar y) { return 0.0 + 0.0 * x + 0.0 * y; };
+
+  // Function that defines the first derivative in x-direction of the diffusion coefficient
+  inline Scalar diff_primey(Scalar x, Scalar y) { return 0.0 + 0.0 * x + 0.0 * y; };
+
+  // Function that defines the advection coefficient in the x-direction
+  inline Scalar advx(Scalar x, Scalar y) { return 10.0 + 0.0 * x + 0.0 * y; };
+
+  // Function that defines the advection coefficient in the x-direction
+  inline Scalar advy(Scalar x, Scalar y) { return 10.0 + 0.0 * x + 0.0 * y; };
+
+  // Function taht defines the reaction coefficient
+  inline Scalar reac(Scalar x, Scalar y) { return 0.0 + 0.0 * x + 0.0 * y; };
+};
+
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+Teuchos::RCP<Matrix> ADR2DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>::BuildMatrix() {
+  GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal)-1);
+  GlobalOrdinal ny = this->list_.get("ny", (GlobalOrdinal)-1);
+  double one       = 1.0;
+  Scalar stretchx  = (Scalar)this->list_.get("stretchx", one);
+  Scalar stretchy  = (Scalar)this->list_.get("stretchy", one);
+
+  if (nx == -1 || ny == -1) {
+    GlobalOrdinal n = this->Map_->getGlobalNumElements();
+    nx              = (GlobalOrdinal)sqrt((double)n);
+    ny              = nx;
+    TEUCHOS_TEST_FOR_EXCEPTION(nx * ny != n, std::logic_error, "You need to specify nx and ny.");
+  }
+  bool keepBCs = this->list_.get("keepBCs", false);
+
+  // Diffusion stencil
+  Scalar c1 = (Scalar)-one / (stretchx * stretchx);
+  Scalar b1 = (Scalar)-one / (stretchx * stretchx);
+  Scalar e1 = (Scalar)-one / (stretchy * stretchy);
+  Scalar d1 = (Scalar)-one / (stretchy * stretchy);
+  Scalar a1 = -(b1 + c1 + d1 + e1);
+
+  // Advection stencil
+  Scalar c2 = (Scalar)one / (2.0 * stretchx);
+  Scalar b2 = (Scalar)one / (2.0 * stretchx);
+  Scalar e2 = (Scalar)one / (2.0 * stretchy);
+  Scalar d2 = (Scalar)one / (2.0 * stretchy);
+
+  // this->A_ = Cross2D<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, ny, center, west, east, south, north, this->DirichletBC_, keepBCs);
+  LocalOrdinal nnz = 5;
+
+  Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map, Matrix>::Build(this->Map_, nnz);
+
+  LocalOrdinal numMyElements = (this->Map_)->getLocalNumElements();
+  GlobalOrdinal indexBase    = (this->Map_)->getIndexBase();
+
+  Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = (this->Map_)->getLocalElementList();
+
+  GlobalOrdinal center, left, right, lower, upper;
+  std::vector<Scalar> vals(nnz);
+  std::vector<GlobalOrdinal> inds(nnz);
+
+  //    e
+  //  b a c
+  //    d
+  for (LocalOrdinal i = 0; i < numMyElements; ++i) {
+    size_t n = 0;
+
+    center = myGlobalElements[i] - indexBase;
+
+    // Determine coordinates
+    Scalar x1 = (Scalar)(center % nx) * stretchx;
+    Scalar x2 = (Scalar)(std::floor(center / nx)) * stretchy;
+
+    Galeri::Xpetra::GetNeighboursCartesian2d(center, nx, ny, left, right, lower, upper);
+
+    bool isDirichlet = (left == -1 && (this->DirichletBC_ & DIR_LEFT)) ||
+                       (right == -1 && (this->DirichletBC_ & DIR_RIGHT)) ||
+                       (lower == -1 && (this->DirichletBC_ & DIR_BOTTOM)) ||
+                       (upper == -1 && (this->DirichletBC_ & DIR_TOP));
+
+    if (isDirichlet && keepBCs) {
+      // Dirichlet unknown we want to keep
+      inds[n]   = center;
+      vals[n++] = Teuchos::ScalarTraits<Scalar>::one();
+
+    } else {
+      // The Neumann b.c. are treated in a sane way. The Dirichlet b.c., however, are treated
+      // insane when the option keepBCs=false. Speicifically, in this case we don't want to keep
+      // Dirichlet b.c., but that would result in inconsistency between the map and the number of
+      // degrees of freedom, plus the problem with GIDs. Therefore, we virtually expand domain by
+      // one node in the direction of the Dirichlet b.c., and then assume that that node was
+      // not kept. But we use an old GIDs. So yes, that's weird.
+
+      if (left != -1) {
+        inds[n]   = left;
+        vals[n++] = b1 * diff(x1, x2) + b2 * diff_primex(x1, x2) - b2 * advx(x1, x2);
+      }
+      if (right != -1) {
+        inds[n]   = right;
+        vals[n++] = c1 * diff(x1, x2) - c2 * diff_primex(x1, x2) + c2 * advx(x1, x2);
+      }
+      if (lower != -1) {
+        inds[n]   = lower;
+        vals[n++] = d1 * diff(x1, x2) + d2 * diff_primey(x1, x2) - d2 * advy(x1, x2);
+      }
+      if (upper != -1) {
+        inds[n]   = upper;
+        vals[n++] = e1 * diff(x1, x2) - e2 * diff_primey(x1, x2) + e2 * advy(x1, x2);
       }
-      bool keepBCs = this->list_.get("keepBCs", false);
-
-			//Diffusion stencil
-      Scalar c1   = (Scalar) -one / (stretchx*stretchx);
-      Scalar b1   = (Scalar) -one / (stretchx*stretchx);
-      Scalar e1   = (Scalar) -one / (stretchy*stretchy);
-      Scalar d1   = (Scalar) -one / (stretchy*stretchy);
-      Scalar a1   = -( b1 + c1 + d1 + e1 );
-
-			//Advection stencil
-      Scalar c2   = (Scalar) one / (2.0 * stretchx);
-      Scalar b2   = (Scalar) one / (2.0 * stretchx);
-      Scalar e2   = (Scalar) one / (2.0 * stretchy);
-      Scalar d2   = (Scalar) one / (2.0 * stretchy);
-      
-      //this->A_ = Cross2D<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, ny, center, west, east, south, north, this->DirichletBC_, keepBCs);
-      LocalOrdinal nnz = 5;
-
-      Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map,Matrix>::Build(this->Map_, nnz);
-
-      LocalOrdinal  numMyElements = (this->Map_)->getLocalNumElements();
-      GlobalOrdinal indexBase     = (this->Map_)->getIndexBase();
-
-      Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = (this->Map_)->getLocalElementList();
-
-      GlobalOrdinal center, left, right, lower, upper;
-      std::vector<Scalar>        vals(nnz);
-      std::vector<GlobalOrdinal> inds(nnz);
-
-      //    e
-      //  b a c
-      //    d
-      for (LocalOrdinal i = 0; i < numMyElements; ++i)  {
-        size_t n = 0;
-
-        center = myGlobalElements[i] - indexBase;
-
-				//Determine coordinates
-				Scalar x1 = (Scalar) (center % nx) * stretchx;
-				Scalar x2 = (Scalar) (std::floor( center/nx )) * stretchy;
-
-        Galeri::Xpetra::GetNeighboursCartesian2d(center, nx, ny, left, right, lower, upper);
-
-        bool isDirichlet = (left  == -1 && (this->DirichletBC_ & DIR_LEFT))   ||
-                           (right == -1 && (this->DirichletBC_ & DIR_RIGHT))  ||
-                           (lower == -1 && (this->DirichletBC_ & DIR_BOTTOM)) ||
-                           (upper == -1 && (this->DirichletBC_ & DIR_TOP));
-
-        if (isDirichlet && keepBCs) {
-          // Dirichlet unknown we want to keep
-          inds[n]   = center;
-          vals[n++] = Teuchos::ScalarTraits<Scalar>::one();
-
-        } else {
-          // The Neumann b.c. are treated in a sane way. The Dirichlet b.c., however, are treated
-          // insane when the option keepBCs=false. Speicifically, in this case we don't want to keep
-          // Dirichlet b.c., but that would result in inconsistency between the map and the number of
-          // degrees of freedom, plus the problem with GIDs. Therefore, we virtually expand domain by
-          // one node in the direction of the Dirichlet b.c., and then assume that that node was
-          // not kept. But we use an old GIDs. So yes, that's weird.
-
-
-          if (left  != -1) { inds[n] = left;  vals[n++] = b1 * diff(x1,x2) + b2 * diff_primex(x1,x2) - b2 * advx(x1,x2); }
-          if (right != -1) { inds[n] = right; vals[n++] = c1 * diff(x1,x2) - c2 * diff_primex(x1,x2) + c2 * advx(x1,x2); }
-          if (lower != -1) { inds[n] = lower; vals[n++] = d1 * diff(x1,x2) + d2 * diff_primey(x1,x2) - d2 * advy(x1,x2); }
-          if (upper != -1) { inds[n] = upper; vals[n++] = e1 * diff(x1,x2) - e2 * diff_primey(x1,x2) + e2 * advy(x1,x2); }
-
-          // diagonal
-          Scalar z = a1 * diff(x1,x2) + reac(x1,x2);
-          if (Galeri::Xpetra::IsBoundary2d(center, nx, ny) && !isDirichlet) {
-            // Neumann boundary unknown (diagonal = sum of all offdiagonal)
-            z = Teuchos::ScalarTraits<Scalar>::zero();
-            for (size_t j = 0; j < n; j++)
-              z -= vals[j];
-          }
-          inds[n]   = center;
-          vals[n++] = z;
-        }
 
+      // diagonal
+      Scalar z = a1 * diff(x1, x2) + reac(x1, x2);
+      if (Galeri::Xpetra::IsBoundary2d(center, nx, ny) && !isDirichlet) {
+        // Neumann boundary unknown (diagonal = sum of all offdiagonal)
+        z = Teuchos::ScalarTraits<Scalar>::zero();
         for (size_t j = 0; j < n; j++)
-          inds[j] += indexBase;
-
-        Teuchos::ArrayView<GlobalOrdinal> iv(&inds[0], n);
-        Teuchos::ArrayView<Scalar>        av(&vals[0], n);
-        mtx->insertGlobalValues(myGlobalElements[i], iv, av);
+          z -= vals[j];
       }
-
-      mtx->fillComplete();
-      this->A_ = mtx;
-
-      this->A_->setObjectLabel(this->getObjectLabel());
-      return this->A_;
+      inds[n]   = center;
+      vals[n++] = z;
     }
 
-    // =============================================  ADR3D  =============================================
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    class ADR3DProblem : public Problem<Map,Matrix,MultiVector> {
-    public:
-      ADR3DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map) : Problem<Map,Matrix,MultiVector>(list, map) { }
-      Teuchos::RCP<Matrix> BuildMatrix();
-
-		private:
-
-			//Function that defines the diffusion coefficient
-			inline Scalar diff( Scalar  x, Scalar y, Scalar z ){return 1.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the first derivative in x-direction of the diffusion coefficient
-			inline Scalar diff_primex( Scalar  x, Scalar y, Scalar z ){return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the first derivative in x-direction of the diffusion coefficient
-			inline Scalar diff_primey( Scalar  x, Scalar y, Scalar z ){return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the first derivative in x-direction of the diffusion coefficient
-			inline Scalar diff_primez( Scalar  x, Scalar y, Scalar z ){return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the advection coefficient in the x-direction
-			inline Scalar advx( Scalar  x, Scalar y, Scalar z ){return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the advection coefficient in the x-direction
-			inline Scalar advy( Scalar  x, Scalar y, Scalar z ){return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function that defines the advection coefficient in the x-direction
-			inline Scalar advz( Scalar  x, Scalar y, Scalar z ){return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-			//Function taht defines the reaction coefficient
-			inline Scalar reac( Scalar  x, Scalar y, Scalar z ){return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z;};
-
-    };
-
-    template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
-    Teuchos::RCP<Matrix> ADR3DProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix,MultiVector>::BuildMatrix() {
-      GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal) -1);
-      GlobalOrdinal ny = this->list_.get("ny", (GlobalOrdinal) -1);
-      GlobalOrdinal nz = this->list_.get("nz", (GlobalOrdinal) -1);
-      double  one = 1.0;
-      Scalar  stretchx = (Scalar) this->list_.get("stretchx", one);
-      Scalar  stretchy = (Scalar) this->list_.get("stretchy", one);
-      Scalar  stretchz = (Scalar) this->list_.get("stretchz", one);
-
-      if (nx == -1 || ny == -1 || nz == -1) {
-        GlobalOrdinal n = this->Map_->getGlobalNumElements();
-        nx = (GlobalOrdinal) Teuchos::ScalarTraits<double>::pow(n, 0.33334);
-        ny = nx; nz = nx;
-        TEUCHOS_TEST_FOR_EXCEPTION(nx * ny * nz != n, std::logic_error, "You need to specify nx, ny, and nz");
+    for (size_t j = 0; j < n; j++)
+      inds[j] += indexBase;
+
+    Teuchos::ArrayView<GlobalOrdinal> iv(&inds[0], n);
+    Teuchos::ArrayView<Scalar> av(&vals[0], n);
+    mtx->insertGlobalValues(myGlobalElements[i], iv, av);
+  }
+
+  mtx->fillComplete();
+  this->A_ = mtx;
+
+  this->A_->setObjectLabel(this->getObjectLabel());
+  return this->A_;
+}
+
+// =============================================  ADR3D  =============================================
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+class ADR3DProblem : public Problem<Map, Matrix, MultiVector> {
+ public:
+  ADR3DProblem(Teuchos::ParameterList& list, const Teuchos::RCP<const Map>& map)
+    : Problem<Map, Matrix, MultiVector>(list, map) {}
+  Teuchos::RCP<Matrix> BuildMatrix();
+
+ private:
+  // Function that defines the diffusion coefficient
+  inline Scalar diff(Scalar x, Scalar y, Scalar z) { return 1.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the first derivative in x-direction of the diffusion coefficient
+  inline Scalar diff_primex(Scalar x, Scalar y, Scalar z) { return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the first derivative in x-direction of the diffusion coefficient
+  inline Scalar diff_primey(Scalar x, Scalar y, Scalar z) { return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the first derivative in x-direction of the diffusion coefficient
+  inline Scalar diff_primez(Scalar x, Scalar y, Scalar z) { return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the advection coefficient in the x-direction
+  inline Scalar advx(Scalar x, Scalar y, Scalar z) { return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the advection coefficient in the x-direction
+  inline Scalar advy(Scalar x, Scalar y, Scalar z) { return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function that defines the advection coefficient in the x-direction
+  inline Scalar advz(Scalar x, Scalar y, Scalar z) { return 10.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+
+  // Function taht defines the reaction coefficient
+  inline Scalar reac(Scalar x, Scalar y, Scalar z) { return 0.0 + 0.0 * x + 0.0 * y + 0.0 * z; };
+};
+
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Map, typename Matrix, typename MultiVector>
+Teuchos::RCP<Matrix> ADR3DProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, MultiVector>::BuildMatrix() {
+  GlobalOrdinal nx = this->list_.get("nx", (GlobalOrdinal)-1);
+  GlobalOrdinal ny = this->list_.get("ny", (GlobalOrdinal)-1);
+  GlobalOrdinal nz = this->list_.get("nz", (GlobalOrdinal)-1);
+  double one       = 1.0;
+  Scalar stretchx  = (Scalar)this->list_.get("stretchx", one);
+  Scalar stretchy  = (Scalar)this->list_.get("stretchy", one);
+  Scalar stretchz  = (Scalar)this->list_.get("stretchz", one);
+
+  if (nx == -1 || ny == -1 || nz == -1) {
+    GlobalOrdinal n = this->Map_->getGlobalNumElements();
+    nx              = (GlobalOrdinal)Teuchos::ScalarTraits<double>::pow(n, 0.33334);
+    ny              = nx;
+    nz              = nx;
+    TEUCHOS_TEST_FOR_EXCEPTION(nx * ny * nz != n, std::logic_error, "You need to specify nx, ny, and nz");
+  }
+  bool keepBCs = this->list_.get("keepBCs", false);
+
+  // Diffusion stencil
+  Scalar c1 = (Scalar)-one / (stretchx * stretchx);
+  Scalar b1 = (Scalar)-one / (stretchx * stretchx);
+  Scalar g1 = (Scalar)-one / (stretchy * stretchy);
+  Scalar f1 = (Scalar)-one / (stretchy * stretchy);
+  Scalar e1 = (Scalar)-one / (stretchz * stretchz);
+  Scalar d1 = (Scalar)-one / (stretchz * stretchz);
+  Scalar a1 = -(c1 + b1 + g1 + f1 + e1 + d1);
+
+  // Advection stencil
+  Scalar c2 = (Scalar)one / (2.0 * stretchx);
+  Scalar b2 = (Scalar)one / (2.0 * stretchx);
+  Scalar g2 = (Scalar)one / (2.0 * stretchy);
+  Scalar f2 = (Scalar)one / (2.0 * stretchy);
+  Scalar e2 = (Scalar)one / (2.0 * stretchz);
+  Scalar d2 = (Scalar)one / (2.0 * stretchz);
+
+  // this->A_ = Cross3D<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, ny, nz, center, left, right, front, back, down, up, this->DirichletBC_, keepBCs);
+  LocalOrdinal nnz = 7;
+
+  Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map, Matrix>::Build(this->Map_, nnz);
+
+  LocalOrdinal numMyElements = this->Map_->getLocalNumElements();
+  GlobalOrdinal indexBase    = this->Map_->getIndexBase();
+
+  Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = (this->Map_)->getLocalElementList();
+
+  GlobalOrdinal center, left, right, bottom, top, front, back;
+  std::vector<GlobalOrdinal> inds(nnz);
+  std::vector<Scalar> vals(nnz);
+
+  //    e
+  //  b a c
+  //    d
+  // + f bottom and g top
+  for (LocalOrdinal i = 0; i < numMyElements; ++i) {
+    size_t n = 0;
+
+    center = myGlobalElements[i] - indexBase;
+
+    // Determine coordinates
+    Scalar x3 = (Scalar)(std::floor(center / (nx * ny))) * stretchz;
+    int plane = center % (nx * ny);
+    Scalar x1 = (Scalar)(plane % nx) * stretchx;
+    Scalar x2 = (Scalar)(std::floor(plane / nx)) * stretchy;
+
+    Galeri::Xpetra::GetNeighboursCartesian3d(center, nx, ny, nz,
+                                             left, right, front, back, bottom, top);
+
+    bool isDirichlet = (left == -1 && (this->DirichletBC_ & DIR_LEFT)) ||
+                       (right == -1 && (this->DirichletBC_ & DIR_RIGHT)) ||
+                       (front == -1 && (this->DirichletBC_ & DIR_BOTTOM)) ||
+                       (back == -1 && (this->DirichletBC_ & DIR_TOP)) ||
+                       (front == -1 && (this->DirichletBC_ & DIR_FRONT)) ||
+                       (back == -1 && (this->DirichletBC_ & DIR_BACK));
+
+    if (isDirichlet && keepBCs) {
+      // Dirichlet unknown we want to keep
+      inds[n]   = center;
+      vals[n++] = Teuchos::ScalarTraits<Scalar>::one();
+
+    } else {
+      // See comments about weird in Cross2D
+      if (left != -1) {
+        inds[n]   = left;
+        vals[n++] = b1 * diff(x1, x2, x3) + b2 * diff_primex(x1, x2, x3) - b2 * advx(x1, x2, x3);
+      }
+      if (right != -1) {
+        inds[n]   = right;
+        vals[n++] = c1 * diff(x1, x2, x3) - c2 * diff_primex(x1, x2, x3) + c2 * advx(x1, x2, x3);
+      }
+      if (front != -1) {
+        inds[n]   = front;
+        vals[n++] = d1 * diff(x1, x2, x3) + d2 * diff_primey(x1, x2, x3) - d2 * advy(x1, x2, x3);
+      }
+      if (back != -1) {
+        inds[n]   = back;
+        vals[n++] = e1 * diff(x1, x2, x3) - e2 * diff_primey(x1, x2, x3) + e2 * advy(x1, x2, x3);
+      }
+      if (bottom != -1) {
+        inds[n]   = bottom;
+        vals[n++] = f1 * diff(x1, x2, x3) + f2 * diff_primez(x1, x2, x3) - f2 * advz(x1, x2, x3);
+      }
+      if (top != -1) {
+        inds[n]   = top;
+        vals[n++] = g1 * diff(x1, x2, x3) - g2 * diff_primez(x1, x2, x3) + g2 * advz(x1, x2, x3);
       }
-      bool keepBCs = this->list_.get("keepBCs", false);
-
-			//Diffusion stencil
-      Scalar c1   = (Scalar) -one / (stretchx*stretchx);
-      Scalar b1   = (Scalar) -one / (stretchx*stretchx);
-      Scalar g1   = (Scalar) -one / (stretchy*stretchy);
-      Scalar f1   = (Scalar) -one / (stretchy*stretchy);
-      Scalar e1   = (Scalar) -one / (stretchz*stretchz);
-      Scalar d1   = (Scalar) -one / (stretchz*stretchz);
-      Scalar a1	 = -(c1 + b1 + g1 + f1 + e1 + d1);
-
-			//Advection stencil
-      Scalar c2   = (Scalar) one / (2.0 * stretchx);
-      Scalar b2   = (Scalar) one / (2.0 * stretchx);
-      Scalar g2   = (Scalar) one / (2.0 * stretchy);
-      Scalar f2   = (Scalar) one / (2.0 * stretchy);
-      Scalar e2   = (Scalar) one / (2.0 * stretchz);
-      Scalar d2   = (Scalar) one / (2.0 * stretchz);
-
-      //this->A_ = Cross3D<Scalar,LocalOrdinal,GlobalOrdinal,Map,Matrix>(this->Map_, nx, ny, nz, center, left, right, front, back, down, up, this->DirichletBC_, keepBCs);
-      LocalOrdinal nnz = 7;
-
-      Teuchos::RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map,Matrix>::Build(this->Map_, nnz);
-
-      LocalOrdinal  numMyElements = this->Map_->getLocalNumElements();
-      GlobalOrdinal indexBase     = this->Map_->getIndexBase();
-
-      Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = (this->Map_)->getLocalElementList();
-
-      GlobalOrdinal center, left, right, bottom, top, front, back;
-      std::vector<GlobalOrdinal> inds(nnz);
-      std::vector<Scalar>        vals(nnz);
-
-      //    e
-      //  b a c
-      //    d
-      // + f bottom and g top
-      for (LocalOrdinal i = 0; i < numMyElements; ++i) {
-        size_t n = 0;
-
-        center = myGlobalElements[i] - indexBase;
-
-				//Determine coordinates
-				Scalar x3 = (Scalar) (std::floor( center/( nx*ny ) )) * stretchz;
-				int plane = center % (nx*ny);
-				Scalar x1 = (Scalar) (plane % nx) * stretchx;
-				Scalar x2 = (Scalar) (std::floor( plane/nx )) * stretchy;
-
-
-        Galeri::Xpetra::GetNeighboursCartesian3d(center, nx, ny, nz,
-                                 left, right, front, back, bottom, top);
-
-        bool isDirichlet = (left  == -1 && (this->DirichletBC_ & DIR_LEFT))   ||
-                           (right == -1 && (this->DirichletBC_ & DIR_RIGHT))  ||
-                           (front == -1 && (this->DirichletBC_ & DIR_BOTTOM)) ||
-                           (back  == -1 && (this->DirichletBC_ & DIR_TOP))    ||
-                           (front == -1 && (this->DirichletBC_ & DIR_FRONT))  ||
-                           (back  == -1 && (this->DirichletBC_ & DIR_BACK));
-
-        if (isDirichlet && keepBCs) {
-          // Dirichlet unknown we want to keep
-          inds[n]   = center;
-          vals[n++] = Teuchos::ScalarTraits<Scalar>::one();
-
-        } else {
-          // See comments about weird in Cross2D
-          if (left   != -1) { inds[n] = left;   vals[n++] = b1 * diff(x1,x2,x3) + b2 * diff_primex(x1,x2,x3) - b2 * advx(x1,x2,x3); }
-          if (right  != -1) { inds[n] = right;  vals[n++] = c1 * diff(x1,x2,x3) - c2 * diff_primex(x1,x2,x3) + c2 * advx(x1,x2,x3); }
-          if (front  != -1) { inds[n] = front;  vals[n++] = d1 * diff(x1,x2,x3) + d2 * diff_primey(x1,x2,x3) - d2 * advy(x1,x2,x3); }
-          if (back   != -1) { inds[n] = back;   vals[n++] = e1 * diff(x1,x2,x3) - e2 * diff_primey(x1,x2,x3) + e2 * advy(x1,x2,x3); }
-          if (bottom != -1) { inds[n] = bottom; vals[n++] = f1 * diff(x1,x2,x3) + f2 * diff_primez(x1,x2,x3) - f2 * advz(x1,x2,x3); }
-          if (top    != -1) { inds[n] = top;    vals[n++] = g1 * diff(x1,x2,x3) - g2 * diff_primez(x1,x2,x3) + g2 * advz(x1,x2,x3); }
-
-          // diagonal
-          Scalar z = a1 * diff(x1,x2,x3) + reac(x1,x2,x3);
-          if (Galeri::Xpetra::IsBoundary3d(center, nx, ny, nz) && !isDirichlet) {
-            // Neumann boundary unknown (diagonal = sum of all offdiagonal)
-            z = Teuchos::ScalarTraits<Scalar>::zero();
-            for (size_t j = 0; j < n; j++)
-              z -= vals[j];
-          }
-          inds[n]   = center;
-          vals[n++] = z;
-        }
 
+      // diagonal
+      Scalar z = a1 * diff(x1, x2, x3) + reac(x1, x2, x3);
+      if (Galeri::Xpetra::IsBoundary3d(center, nx, ny, nz) && !isDirichlet) {
+        // Neumann boundary unknown (diagonal = sum of all offdiagonal)
+        z = Teuchos::ScalarTraits<Scalar>::zero();
         for (size_t j = 0; j < n; j++)
-          inds[j] += indexBase;
-
-        Teuchos::ArrayView<GlobalOrdinal> iv(&inds[0], n);
-        Teuchos::ArrayView<Scalar>        av(&vals[0], n);
-        mtx->insertGlobalValues(myGlobalElements[i], iv, av);
+          z -= vals[j];
       }
-
-      mtx->fillComplete();
-			this->A_ = mtx;
-
-      this->A_->setObjectLabel(this->getObjectLabel());
-      return this->A_;
+      inds[n]   = center;
+      vals[n++] = z;
     }
 
+    for (size_t j = 0; j < n; j++)
+      inds[j] += indexBase;
+
+    Teuchos::ArrayView<GlobalOrdinal> iv(&inds[0], n);
+    Teuchos::ArrayView<Scalar> av(&vals[0], n);
+    mtx->insertGlobalValues(myGlobalElements[i], iv, av);
+  }
 
+  mtx->fillComplete();
+  this->A_ = mtx;
 
+  this->A_->setObjectLabel(this->getObjectLabel());
+  return this->A_;
+}
 
-  } // namespace Xpetra
+}  // namespace Xpetra
 
-} // namespace ADR
+}  // namespace ADR
 
-#endif // CREATEADRMATRIX_HPP
+#endif  // CREATEADRMATRIX_HPP
diff --git a/packages/muelu/research/max/AdditiveMG/CreateBrickMap.hpp b/packages/muelu/research/max/AdditiveMG/CreateBrickMap.hpp
index 96ec3ba27c69..4a544e827ff8 100644
--- a/packages/muelu/research/max/AdditiveMG/CreateBrickMap.hpp
+++ b/packages/muelu/research/max/AdditiveMG/CreateBrickMap.hpp
@@ -1,103 +1,89 @@
 
-void createBrickMap1D( int numGlobalElements, std::vector<int>& ind, Teuchos::RCP< const Teuchos::Comm<int> > comm)
-{
-    //INPUT: numGlobalElements = number of finite difference nodes along x
-    //INPUT: ind = vector containing indices of the rows owned by the current MPI processor
-    //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-
-    int mypid = comm->getRank();
-
-    ind.reserve (static_cast<int>( numGlobalElements/( comm->getSize()-1 ) + 1 ));
-		if( mypid !=0 && mypid!=comm->getSize()-1 )
-			for(int i = 0; i <= ( static_cast<int>(numGlobalElements/( comm->getSize()-1 ))) - 1; ++i)
-				ind.emplace_back( (mypid-1) * static_cast<int>(numGlobalElements/( comm->getSize()-1 )) + i );
-		
-			if( mypid==comm->getSize()-1 )
-				for(int i = (mypid-1) * static_cast<int>(numGlobalElements/( comm->getSize()-1 )); i != numGlobalElements; ++i)\
-					ind.emplace_back( i );
-}
-
-
-
+void createBrickMap1D(int numGlobalElements, std::vector<int>& ind, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+  // INPUT: numGlobalElements = number of finite difference nodes along x
+  // INPUT: ind = vector containing indices of the rows owned by the current MPI processor
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
 
-void createBrickMap2D( int nx, int brick_sizex, int brick_sizey, std::vector<int>& ind, Teuchos::RCP< const Teuchos::Comm<int> > comm)
-{
+  int mypid = comm->getRank();
 
-    //INPUT: nx = number of finite difference nodes along x
-    //INPUT: brick_sizex = size of a brick aggregate along x-direction
-    //INPUT: brick_sizey = size of a brick aggregate along y-direction
-    //INPUT: ind = vector containing indices of the rows owned by the current MPI processor
-    //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-    
-    int mypid = comm->getRank();
+  ind.reserve(static_cast<int>(numGlobalElements / (comm->getSize() - 1) + 1));
+  if (mypid != 0 && mypid != comm->getSize() - 1)
+    for (int i = 0; i <= (static_cast<int>(numGlobalElements / (comm->getSize() - 1))) - 1; ++i)
+      ind.emplace_back((mypid - 1) * static_cast<int>(numGlobalElements / (comm->getSize() - 1)) + i);
 
-    ind.reserve(brick_sizex * brick_sizey);
-   
-	int ndx = nx /brick_sizex;
+  if (mypid == comm->getSize() - 1)
+    for (int i = (mypid - 1) * static_cast<int>(numGlobalElements / (comm->getSize() - 1)); i != numGlobalElements; ++i)
+      ind.emplace_back(i);
+}
 
-	if( mypid !=0 )
-	{
-                  int grid_row  = std::ceil(static_cast<double>(mypid)/ndx);
-                  int ypos = grid_row;
-                  int xpos;
+void createBrickMap2D(int nx, int brick_sizex, int brick_sizey, std::vector<int>& ind, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+  // INPUT: nx = number of finite difference nodes along x
+  // INPUT: brick_sizex = size of a brick aggregate along x-direction
+  // INPUT: brick_sizey = size of a brick aggregate along y-direction
+  // INPUT: ind = vector containing indices of the rows owned by the current MPI processor
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
 
-                  if( 0 != mypid%ndx )
-	            xpos = mypid%ndx;
-                  else
-	            xpos = ndx;
+  int mypid = comm->getRank();
 
-                  int preliminary = nx * brick_sizey * (ypos - 1);
+  ind.reserve(brick_sizex * brick_sizey);
 
-		  for( int row=0; row<brick_sizey; ++row )
-			for( int col=brick_sizex*(xpos-1)+1; col<=brick_sizex*xpos; ++col )
-				  ind.emplace_back( preliminary + row*nx + col - 1 );
-        }
-}
+  int ndx = nx / brick_sizex;
 
+  if (mypid != 0) {
+    int grid_row = std::ceil(static_cast<double>(mypid) / ndx);
+    int ypos     = grid_row;
+    int xpos;
 
+    if (0 != mypid % ndx)
+      xpos = mypid % ndx;
+    else
+      xpos = ndx;
 
+    int preliminary = nx * brick_sizey * (ypos - 1);
 
-void createBrickMap3D( int nx, int ny, int brick_sizex, int brick_sizey, int brick_sizez, std::vector<int>& ind, Teuchos::RCP< const Teuchos::Comm<int> > comm)
-{
+    for (int row = 0; row < brick_sizey; ++row)
+      for (int col = brick_sizex * (xpos - 1) + 1; col <= brick_sizex * xpos; ++col)
+        ind.emplace_back(preliminary + row * nx + col - 1);
+  }
+}
 
-    //INPUT: nx = number of finite difference nodes along x
-    //INPUT: ny = number of finite difference nodes along y
-    //INPUT: brick_sizex = size of a brick aggregate along x-direction
-    //INPUT: brick_sizey = size of a brick aggregate along y-direction
-    //INPUT: ind = vector containing indices of the rows owned by the current MPI processor
-    //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
+void createBrickMap3D(int nx, int ny, int brick_sizex, int brick_sizey, int brick_sizez, std::vector<int>& ind, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+  // INPUT: nx = number of finite difference nodes along x
+  // INPUT: ny = number of finite difference nodes along y
+  // INPUT: brick_sizex = size of a brick aggregate along x-direction
+  // INPUT: brick_sizey = size of a brick aggregate along y-direction
+  // INPUT: ind = vector containing indices of the rows owned by the current MPI processor
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
 
-    int mypid = comm->getRank();
+  int mypid = comm->getRank();
 
-    ind.reserve(brick_sizex * brick_sizey * brick_sizez);
+  ind.reserve(brick_sizex * brick_sizey * brick_sizez);
 
-    // determine the number of subdomains along x and y directions
-    int ndx = nx /brick_sizex;
-    int ndy = ny /brick_sizey;
+  // determine the number of subdomains along x and y directions
+  int ndx = nx / brick_sizex;
+  int ndy = ny / brick_sizey;
 
-    if( mypid !=0 )
-    {
-	int grid_plane = std::ceil( static_cast<double>(mypid)/(ndx*ndy) );
-	int plane_id = mypid % ( ndx*ndy );
+  if (mypid != 0) {
+    int grid_plane = std::ceil(static_cast<double>(mypid) / (ndx * ndy));
+    int plane_id   = mypid % (ndx * ndy);
 
-	if(0 == plane_id)
-	  plane_id = ndx*ndy;
+    if (0 == plane_id)
+      plane_id = ndx * ndy;
 
-	int plane_row  = std::ceil(static_cast<double>(plane_id)/ndx);
-	int ypos = plane_row;
-	int xpos;
+    int plane_row = std::ceil(static_cast<double>(plane_id) / ndx);
+    int ypos      = plane_row;
+    int xpos;
 
-	  if( 0 != plane_id%ndx )
-	    xpos = plane_id%ndx;
-	  else
-	    xpos = ndx;
+    if (0 != plane_id % ndx)
+      xpos = plane_id % ndx;
+    else
+      xpos = ndx;
 
-      int preliminary = nx * ny * brick_sizez * (grid_plane - 1) + nx * brick_sizey * (ypos - 1);
+    int preliminary = nx * ny * brick_sizez * (grid_plane - 1) + nx * brick_sizey * (ypos - 1);
 
-      for(int l = 0; l<brick_sizez; ++l)
-	  for( int row=0; row<brick_sizey; ++row )
-		for( int col=brick_sizex*(xpos-1)+1; col<=brick_sizex*xpos; ++col )			
-			ind.emplace_back( preliminary + l*nx*ny + row*nx + col - 1);
-						
-   }
+    for (int l = 0; l < brick_sizez; ++l)
+      for (int row = 0; row < brick_sizey; ++row)
+        for (int col = brick_sizex * (xpos - 1) + 1; col <= brick_sizex * xpos; ++col)
+          ind.emplace_back(preliminary + l * nx * ny + row * nx + col - 1);
+  }
 }
diff --git a/packages/muelu/research/max/AdditiveMG/MultiplicativeBricks.cpp b/packages/muelu/research/max/AdditiveMG/MultiplicativeBricks.cpp
index dcb088dbcb0a..e89604dc3f45 100644
--- a/packages/muelu/research/max/AdditiveMG/MultiplicativeBricks.cpp
+++ b/packages/muelu/research/max/AdditiveMG/MultiplicativeBricks.cpp
@@ -60,7 +60,6 @@
 #include <iostream>
 #include <cmath>
 
-
 // Belos provides Krylov solvers
 #include <BelosConfigDefs.hpp>
 #include <BelosLinearProblem.hpp>
@@ -76,7 +75,7 @@
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -87,24 +86,23 @@
 #include "BAP.hpp"
 #include "CreateBrickMap.hpp"
 
-int main(int argc, char *argv[]) {
-
+int main(int argc, char* argv[]) {
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-	typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -112,50 +110,50 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-	typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
+  typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-  typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                           = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "final_parser.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "final_parser.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Teuchos::ParameterList xmlParams;
@@ -166,7 +164,7 @@ int main(int argc, char *argv[]) {
   problemParams = xmlParams.sublist(static_cast<const std::string>("Problem"));
 
   // Problem definition
-   std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
+  std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
 
   // Parameters
 
@@ -183,14 +181,12 @@ int main(int argc, char *argv[]) {
   MueLu::DomainPartitioning domain;
 
   int keep_boundary = 0;
-  Scalar stretchx = (Scalar) Lx/nx;
-  Scalar stretchy = (Scalar) Ly/ny;
-  Scalar stretchz = (Scalar) Lz/nz;
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
-
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
- 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
   //
   // Construct the problem
@@ -200,57 +196,55 @@ int main(int argc, char *argv[]) {
   RCP<const Map> xpetraMap;
   std::vector<global_ordinal_type> ind;
 
-  //BRICK SIZE
+  // BRICK SIZE
   int brick_sizex = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick x size"));
   int brick_sizey = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick y size"));
   int brick_sizez = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick z size"));
 
-  //Creation of the map where processor 0 gets nothing at the fine level
-  if( comm->getSize()>1 )
-  {
-  	if(problem_type == "ADR1D")
-		createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm );
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    if (problem_type == "ADR1D")
+      createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm);
 
-	else if(problem_type == "ADR2D")
-		createBrickMap2D( nx, brick_sizex, brick_sizey, ind, comm );
+    else if (problem_type == "ADR2D")
+      createBrickMap2D(nx, brick_sizex, brick_sizey, ind, comm);
 
-	else if(problem_type == "ADR3D")
-		createBrickMap3D( nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm );
+    else if (problem_type == "ADR3D")
+      createBrickMap3D(nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm);
 
-	ind.shrink_to_fit();
+    ind.shrink_to_fit();
 
-	Teuchos::ArrayView<const global_ordinal_type> elementList (ind);
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-  }
-  else if( comm->getSize()==1 )
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+    Teuchos::ArrayView<const global_ordinal_type> elementList(ind);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1)
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
 
   RCP<MultiVector> coordinates;
 
   if (problem_type == "ADR1D")
-	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR2D")
-	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR3D")
-	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
 
   RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-  RCP<Matrix>  xpetraA = Pr->BuildMatrix();
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
 
-  RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
   RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
 
- // ===================================================
- // 	Domain Decomposition Preconditioner
- // 	===================================
+  // ===================================================
+  // 	Domain Decomposition Preconditioner
+  // 	===================================
 
-  //Creation of the MueLu list for the DD preconditioner
+  // Creation of the MueLu list for the DD preconditioner
   RCP<Teuchos::ParameterList> dd_list = rcp(new Teuchos::ParameterList());
   dd_list->setName("MueLu");
   dd_list->set("verbosity", "low");
   dd_list->set("number of equations", 1);
   dd_list->set("max levels", 1);
-  dd_list->set("coarse: type", "SCHWARZ"); //FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
+  dd_list->set("coarse: type", "SCHWARZ");  // FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
 
   Teuchos::ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
   dd_smooth_sublist.set("schwarz: overlap level", 0);
@@ -263,30 +257,30 @@ int main(int argc, char *argv[]) {
   coarse_subdomain_solver.set("fact: relative threshold", 1.);
   coarse_subdomain_solver.set("fact: relax value", 0.);
 
-  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, *dd_list );
+  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, *dd_list);
+
+  // ===================================================
+  // 	Multi Grid Preconditioner
+  // 	===================================
 
- // ===================================================
- // 	Multi Grid Preconditioner
- // 	===================================
- 
   RCP<muelu_tpetra_operator_type> M;
-  M = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates) );
+  M = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates));
 
-  RCP<multivector_type> X_muelu = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
+  RCP<multivector_type> X_muelu = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B       = rcp(new multivector_type(map, 1));
   RCP<linear_problem_type> Problem_muelu;
-  X_muelu->putScalar((scalar_type) 0.0);
+  X_muelu->putScalar((scalar_type)0.0);
   B->randomize();
 
   Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
 
   RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  //belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Verbosity",             Belos::Errors);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
+  belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+  belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+  // belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  belosList->set("Verbosity", Belos::Errors);
+  belosList->set("Output Frequency", 1);
+  belosList->set("Output Style", Belos::Brief);
   belosList->set("Implicit Residual Scaling", "None");
   RCP<belos_solver_manager_type> solver;
   if (krylovSolverType == "cg")
@@ -298,38 +292,35 @@ int main(int argc, char *argv[]) {
   else
     throw std::invalid_argument("bad Krylov solver type");
 
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-
-  X_muelu->putScalar((scalar_type) 0.0);
-  B->randomize();
-
-  //
-  // Set up Krylov solver and iterate.
-  //
- 
-  Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
-  Problem_muelu->setRightPrec(M);
-  Problem_muelu->setProblem();
-
-  solver->setProblem(Problem_muelu);
-  solver->solve();
-  int numIterations_muelu = solver->getNumIters();
-
-  Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
-  multivector_type residual_muelu(B->getMap(),1);
-  A->apply(*X_muelu, residual_muelu);
-  residual_muelu.update(1.0, *B, -1.0);
-  residual_muelu.norm2(normVec_muelu);
-  if (mypid == 0) {
-    std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
-    std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
- }
-}
-
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    X_muelu->putScalar((scalar_type)0.0);
+    B->randomize();
+
+    //
+    // Set up Krylov solver and iterate.
+    //
+
+    Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
+    Problem_muelu->setRightPrec(M);
+    Problem_muelu->setProblem();
+
+    solver->setProblem(Problem_muelu);
+    solver->solve();
+    int numIterations_muelu = solver->getNumIters();
+
+    Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
+    multivector_type residual_muelu(B->getMap(), 1);
+    A->apply(*X_muelu, residual_muelu);
+    residual_muelu.update(1.0, *B, -1.0);
+    residual_muelu.norm2(normVec_muelu);
+    if (mypid == 0) {
+      std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
+      std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
+    }
+  }
 
-  #include <Teuchos_TimeMonitor.hpp>
-  Teuchos::TimeMonitor::summarize ();
+#include <Teuchos_TimeMonitor.hpp>
+  Teuchos::TimeMonitor::summarize();
 
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/MultiplicativeStride.cpp b/packages/muelu/research/max/AdditiveMG/MultiplicativeStride.cpp
index 141915443d21..3355a1d8bf37 100644
--- a/packages/muelu/research/max/AdditiveMG/MultiplicativeStride.cpp
+++ b/packages/muelu/research/max/AdditiveMG/MultiplicativeStride.cpp
@@ -60,7 +60,6 @@
 
 #include <iostream>
 
-
 // Belos provides Krylov solvers
 #include <BelosConfigDefs.hpp>
 #include <BelosLinearProblem.hpp>
@@ -76,7 +75,7 @@
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -85,23 +84,22 @@
 #include "Smooth_Prolongation.cpp"
 
 int main(int argc, char *argv[]) {
-
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-	typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -109,50 +107,50 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-	typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
+  typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-  typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                           = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "final_parser.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "final_parser.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Teuchos::ParameterList xmlParams;
@@ -163,7 +161,7 @@ int main(int argc, char *argv[]) {
   problemParams = xmlParams.sublist(static_cast<const std::string>("Problem"));
 
   // Problem definition
-   std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
+  std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
 
   // Parameters
 
@@ -180,14 +178,12 @@ int main(int argc, char *argv[]) {
   MueLu::DomainPartitioning domain;
 
   int keep_boundary = 0;
-  Scalar stretchx = (Scalar) Lx/nx;
-  Scalar stretchy = (Scalar) Ly/ny;
-  Scalar stretchz = (Scalar) Lz/nz;
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
-
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
- 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
   //
   // Construct the problem
@@ -197,94 +193,91 @@ int main(int argc, char *argv[]) {
   RCP<const Map> xpetraMap;
   std::vector<global_ordinal_type> ind;
 
-  //Creation of the map where processor 0 gets nothing at the fine level
-  if( comm->getSize()>1 )
-  {
-  	ind.reserve (static_cast<int>( matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ) + 1 ));
-  	if( mypid !=0 && mypid!=comm->getSize()-1 )
-  		for(int i = 0; i <= ( static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ))) - 1; ++i)
-  			ind.emplace_back( (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )) + i );
-		
-		if( mypid==comm->getSize()-1 )
-			for(int i = (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )); i != matrixParameters.GetNumGlobalElements(); ++i)
-				ind.emplace_back( i );
-
-	 ind.shrink_to_fit();
-
-	Teuchos::ArrayView<const global_ordinal_type> elementList (ind);
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-  }
-  else if( comm->getSize()==1 )
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    ind.reserve(static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1) + 1));
+    if (mypid != 0 && mypid != comm->getSize() - 1)
+      for (int i = 0; i <= (static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1))) - 1; ++i)
+        ind.emplace_back((mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)) + i);
+
+    if (mypid == comm->getSize() - 1)
+      for (int i = (mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)); i != matrixParameters.GetNumGlobalElements(); ++i)
+        ind.emplace_back(i);
+
+    ind.shrink_to_fit();
+
+    Teuchos::ArrayView<const global_ordinal_type> elementList(ind);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1)
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
 
   RCP<MultiVector> coordinates;
 
   if (problem_type == "ADR1D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR2D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR3D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
 
   RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-  RCP<Matrix>  xpetraA = Pr->BuildMatrix();
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
 
-  RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
   RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
 
   // Construct a multigrid preconditioner
-  RCP<muelu_tpetra_operator_type> M = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates) );
-
-  RCP<multivector_type> X = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
-
-for(int iter = 1; iter <= number_runs; ++iter)
-{
-  X->putScalar((scalar_type) 0.0);
-  B->randomize();
-
-  //
-  // Set up Krylov solver and iterate.
-  //
- 
-  RCP<multivector_type> X_muelu = rcp(new multivector_type(map,1));
-  RCP<linear_problem_type> Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
-  Problem_muelu->setRightPrec(M);
-  Problem_muelu->setProblem();
-
-  RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  //belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Verbosity",             Belos::Errors);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
-  belosList->set("Implicit Residual Scaling", "None");
-  RCP<belos_solver_manager_type> solver;
-  if (krylovSolverType == "cg")
-    solver = rcp(new belos_pseudocg_manager_type(Problem_muelu, belosList));
-  else if (krylovSolverType == "gmres")
-    solver = rcp(new belos_gmres_manager_type(Problem_muelu, belosList));
-  else if (krylovSolverType == "bicgstab")
-    solver = rcp(new belos_bicgstab_manager_type(Problem_muelu, belosList));
-  else
-    throw std::invalid_argument("bad Krylov solver type");
-
-  solver->solve();
-  int numIterations_muelu = solver->getNumIters();
-
-  Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
-  multivector_type residual_muelu(B->getMap(),1);
-  A->apply(*X_muelu, residual_muelu);
-  residual_muelu.update(1.0, *B, -1.0);
-  residual_muelu.norm2(normVec_muelu);
-  if (mypid == 0) {
-    std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
-    std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
- }
-}
-  #include <Teuchos_TimeMonitor.hpp>
-  Teuchos::TimeMonitor::summarize ();
+  RCP<muelu_tpetra_operator_type> M = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates));
+
+  RCP<multivector_type> X = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B = rcp(new multivector_type(map, 1));
+
+  for (int iter = 1; iter <= number_runs; ++iter) {
+    X->putScalar((scalar_type)0.0);
+    B->randomize();
+
+    //
+    // Set up Krylov solver and iterate.
+    //
+
+    RCP<multivector_type> X_muelu          = rcp(new multivector_type(map, 1));
+    RCP<linear_problem_type> Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
+    Problem_muelu->setRightPrec(M);
+    Problem_muelu->setProblem();
+
+    RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
+    belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+    belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+    // belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    belosList->set("Verbosity", Belos::Errors);
+    belosList->set("Output Frequency", 1);
+    belosList->set("Output Style", Belos::Brief);
+    belosList->set("Implicit Residual Scaling", "None");
+    RCP<belos_solver_manager_type> solver;
+    if (krylovSolverType == "cg")
+      solver = rcp(new belos_pseudocg_manager_type(Problem_muelu, belosList));
+    else if (krylovSolverType == "gmres")
+      solver = rcp(new belos_gmres_manager_type(Problem_muelu, belosList));
+    else if (krylovSolverType == "bicgstab")
+      solver = rcp(new belos_bicgstab_manager_type(Problem_muelu, belosList));
+    else
+      throw std::invalid_argument("bad Krylov solver type");
+
+    solver->solve();
+    int numIterations_muelu = solver->getNumIters();
+
+    Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
+    multivector_type residual_muelu(B->getMap(), 1);
+    A->apply(*X_muelu, residual_muelu);
+    residual_muelu.update(1.0, *B, -1.0);
+    residual_muelu.norm2(normVec_muelu);
+    if (mypid == 0) {
+      std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
+      std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
+    }
+  }
+#include <Teuchos_TimeMonitor.hpp>
+  Teuchos::TimeMonitor::summarize();
 
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/Repartition_ADR.cpp b/packages/muelu/research/max/AdditiveMG/Repartition_ADR.cpp
index 0b50560d27c8..89a8148d94e4 100644
--- a/packages/muelu/research/max/AdditiveMG/Repartition_ADR.cpp
+++ b/packages/muelu/research/max/AdditiveMG/Repartition_ADR.cpp
@@ -79,7 +79,7 @@
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -98,22 +98,21 @@
 #include <Xpetra_MatrixFactory.hpp>
 
 int main(int argc, char *argv[]) {
-
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -121,73 +120,71 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-	typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                           = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
   // Problem definition
-   std::string problem_type = "ADR2D";
+  std::string problem_type = "ADR2D";
 
   // Parameters
 
-	Scalar Lx = 6.0;
-	Scalar Ly = 6.0;
-	Scalar Lz = 6.0;
+  Scalar Lx = 6.0;
+  Scalar Ly = 6.0;
+  Scalar Lz = 6.0;
 
   global_ordinal_type nx = 50;
   global_ordinal_type ny = 50;
   global_ordinal_type nz = 50;
 
   int keep_boundary = 0;
-	Scalar stretchx = (Scalar) Lx/nx;
-	Scalar stretchy = (Scalar) Ly/ny;
-	Scalar stretchz = (Scalar) Lz/nz;
-
-
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "dd.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "dd.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
-  
+
   if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {
     throw std::invalid_argument("This example only supports Tpetra.");
   }
@@ -200,60 +197,55 @@ int main(int argc, char *argv[]) {
   //
 
   global_ordinal_type indexBase = 0;
-	RCP<const Map> xpetraMap;
-	std::vector<global_ordinal_type> indices;
-
-	//Creation of the map where processor 0 gets nothing at the fine level
-	if( comm->getSize()>1 )
-	{
-		indices.reserve (static_cast<int>( matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ) + 1 ));
-		if( mypid !=0 && mypid!=comm->getSize()-1 )
-			for(int i = 0; i <= ( static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ))) - 1; ++i)
-				indices.emplace_back( (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )) + i );
-		
-		if( mypid==comm->getSize()-1 )
-			for(int i = (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )); i != matrixParameters.GetNumGlobalElements(); ++i)\
-				indices.emplace_back( i );
-
-		 indices.shrink_to_fit();
-
-		Teuchos::ArrayView<const global_ordinal_type> elementList (indices);
-		xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-	}
-	else if( comm->getSize()==1 )
-	{
-		xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
-	}
-
-	//============================================================================================================
+  RCP<const Map> xpetraMap;
+  std::vector<global_ordinal_type> indices;
+
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    indices.reserve(static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1) + 1));
+    if (mypid != 0 && mypid != comm->getSize() - 1)
+      for (int i = 0; i <= (static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1))) - 1; ++i)
+        indices.emplace_back((mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)) + i);
+
+    if (mypid == comm->getSize() - 1)
+      for (int i = (mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)); i != matrixParameters.GetNumGlobalElements(); ++i)
+        indices.emplace_back(i);
+
+    indices.shrink_to_fit();
+
+    Teuchos::ArrayView<const global_ordinal_type> elementList(indices);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1) {
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+  }
+
+  //============================================================================================================
 
   RCP<MultiVector> coordinates;
 
-  //RCP<const Map> xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
-	//RCP<const Map> xpetraMap = Galeri::Xpetra::CreateMap<local_ordinal_type, global_ordinal_type, node_type>(Xpetra::UseTpetra, "Cartesian2D", comm, matrixParameters.GetParameterList());
+  // RCP<const Map> xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+  // RCP<const Map> xpetraMap = Galeri::Xpetra::CreateMap<local_ordinal_type, global_ordinal_type, node_type>(Xpetra::UseTpetra, "Cartesian2D", comm, matrixParameters.GetParameterList());
 
-	std::cout<<"Before coordinates"<<std::endl;
-	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
-	std::cout<<"After coordinates"<<std::endl;
+  std::cout << "Before coordinates" << std::endl;
+  coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+  std::cout << "After coordinates" << std::endl;
 
   RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-  	RCP<Matrix>  xpetraA = Pr->BuildMatrix();
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
 
-		RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
-    RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
 
   //
   // Construct a multigrid preconditioner
   //
 
-	RCP<muelu_tpetra_operator_type> M = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates) );
-
-
+  RCP<muelu_tpetra_operator_type> M = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, mueluParams, Utilities::MV2NonConstTpetraMV(coordinates));
 
-  RCP<multivector_type> X = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
+  RCP<multivector_type> X = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B = rcp(new multivector_type(map, 1));
 
-  X->putScalar((scalar_type) 0.0);
+  X->putScalar((scalar_type)0.0);
   B->randomize();
 
   RCP<linear_problem_type> Problem = rcp(new linear_problem_type(A, X, B));
@@ -263,13 +255,13 @@ int main(int argc, char *argv[]) {
   //
   // Set up Krylov solver and iterate.
   //
- 
+
   RCP<ParameterList> belosList = rcp(new ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
+  belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+  belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+  belosList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  belosList->set("Output Frequency", 1);
+  belosList->set("Output Style", Belos::Brief);
   belosList->set("Implicit Residual Scaling", "None");
   RCP<belos_solver_manager_type> solver;
   if (krylovSolverType == "cg")
@@ -285,8 +277,8 @@ int main(int argc, char *argv[]) {
   int numIterations = solver->getNumIters();
 
   Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec(1);
-  multivector_type Ax(B->getMap(),1);
-  multivector_type residual(B->getMap(),1);
+  multivector_type Ax(B->getMap(), 1);
+  multivector_type residual(B->getMap(), 1);
   A->apply(*X, residual);
   residual.update(1.0, *B, -1.0);
   residual.norm2(normVec);
@@ -294,6 +286,6 @@ int main(int argc, char *argv[]) {
     std::cout << "number of iterations = " << numIterations << std::endl;
     std::cout << "||Residual|| = " << normVec[0] << std::endl;
   }
- 
+
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.cpp b/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.cpp
index 2b2e3148ee3d..50fef96d0c62 100644
--- a/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.cpp
+++ b/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.cpp
@@ -1,255 +1,234 @@
 #include "Smooth_Prolongation.hpp"
 
-namespace MueLu{
+namespace MueLu {
 
-AdditiveVariant::AdditiveVariant( RCP<crs_matrix_type> A, RCP<multivector_type> coords, DomainPartitioning domain)
-{
+AdditiveVariant::AdditiveVariant(RCP<crs_matrix_type> A, RCP<multivector_type> coords, DomainPartitioning domain) {
+  domain_ = domain;
 
-	domain_ = domain;
+  TEUCHOS_TEST_FOR_EXCEPT(!(GlobalComm_.is_null()));
+  GlobalComm_ = A->getComm();
+  TEUCHOS_TEST_FOR_EXCEPT(GlobalComm_.is_null());
 
-	TEUCHOS_TEST_FOR_EXCEPT( !( GlobalComm_.is_null() ) );
-	GlobalComm_ = A->getComm();
-	TEUCHOS_TEST_FOR_EXCEPT( GlobalComm_.is_null() );
+  TEUCHOS_TEST_FOR_EXCEPT(!(coords_.is_null()));
+  coords_ = coords;
+  TEUCHOS_TEST_FOR_EXCEPT((coords_.is_null()));
 
-	TEUCHOS_TEST_FOR_EXCEPT( !( coords_.is_null() ) );
-	coords_ = coords;
-	TEUCHOS_TEST_FOR_EXCEPT( ( coords_.is_null() ) );
+  TEUCHOS_TEST_FOR_EXCEPT(!(DomainMap_.is_null()));
+  DomainMap_ = A->getDomainMap();
+  TEUCHOS_TEST_FOR_EXCEPT(DomainMap_.is_null());
 
-	TEUCHOS_TEST_FOR_EXCEPT( !( DomainMap_.is_null() ) );
-	DomainMap_ = A->getDomainMap();
-	TEUCHOS_TEST_FOR_EXCEPT( DomainMap_.is_null() );
+  TEUCHOS_TEST_FOR_EXCEPT(!(RangeMap_.is_null()));
+  RangeMap_ = A->getRangeMap();
+  TEUCHOS_TEST_FOR_EXCEPT(RangeMap_.is_null());
 
-	TEUCHOS_TEST_FOR_EXCEPT( !( RangeMap_.is_null() ) );
-	RangeMap_ = A->getRangeMap();
-	TEUCHOS_TEST_FOR_EXCEPT( RangeMap_.is_null() );
-
-	TEUCHOS_TEST_FOR_EXCEPT( !( B_fine_.is_null() ) || !( B_coarse_.is_null() ) );
-	AdditiveFineSmoother( A );
-	AdditiveCoarseSolver( A );
-	TEUCHOS_TEST_FOR_EXCEPT( B_fine_.is_null() || B_coarse_.is_null() );
-
-	/*TEUCHOS_TEST_FOR_EXCEPT( !( B_DD_.is_null() ) || !( B_coarse_.is_null() ) );
-	AdditiveFineSmoother( A );
-	AdditiveCoarseSolver( A );
-	TEUCHOS_TEST_FOR_EXCEPT( B_DD_.is_null() || B_coarse_.is_null() );
-	TEUCHOS_TEST_FOR_EXCEPT( !B_DD_->isInitialized() || !B_DD_->isComputed() );*/
+  TEUCHOS_TEST_FOR_EXCEPT(!(B_fine_.is_null()) || !(B_coarse_.is_null()));
+  AdditiveFineSmoother(A);
+  AdditiveCoarseSolver(A);
+  TEUCHOS_TEST_FOR_EXCEPT(B_fine_.is_null() || B_coarse_.is_null());
 
+  /*TEUCHOS_TEST_FOR_EXCEPT( !( B_DD_.is_null() ) || !( B_coarse_.is_null() ) );
+  AdditiveFineSmoother( A );
+  AdditiveCoarseSolver( A );
+  TEUCHOS_TEST_FOR_EXCEPT( B_DD_.is_null() || B_coarse_.is_null() );
+  TEUCHOS_TEST_FOR_EXCEPT( !B_DD_->isInitialized() || !B_DD_->isComputed() );*/
 }
 
-
-void
-AdditiveVariant::AdditiveFineSmoother( RCP<crs_matrix_type> A )
-{
-		
-	  //Creation of the MueLu list for the DD preconditioner
-	  RCP<ParameterList> dd_list = rcp(new Teuchos::ParameterList());
-	  dd_list->setName("MueLu");
-	  dd_list->set("verbosity", "low");
-	  dd_list->set("number of equations", 1);
-	  dd_list->set("max levels", 1);
-	  dd_list->set("coarse: type", "SCHWARZ"); //FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
-		
-	  ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
-	  dd_smooth_sublist.set("schwarz: overlap level", 0);
-	  dd_smooth_sublist.set("schwarz: combine mode", "Zero");
-	  dd_smooth_sublist.set("subdomain solver name", "RILUK");
-
-	  ParameterList& coarse_subdomain_solver = dd_smooth_sublist.sublist("subdomain solver parameters");
-	  coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
-	  coarse_subdomain_solver.set("fact: absolute threshold", 0.);
-	  coarse_subdomain_solver.set("fact: relative threshold", 1.);
-	  coarse_subdomain_solver.set("fact: relax value", 0.);
-
-	  B_fine_ = CreateTpetraPreconditioner( (RCP<operator_type>)A, *dd_list );
-		
-	  /*B_DD_ = rcp( new Ifpack2::AdditiveSchwarz< row_matrix_type, precond_type >(A, 0) );
-	  B_DD_->initialize();
-	  B_DD_->compute();*/
-
+void AdditiveVariant::AdditiveFineSmoother(RCP<crs_matrix_type> A) {
+  // Creation of the MueLu list for the DD preconditioner
+  RCP<ParameterList> dd_list = rcp(new Teuchos::ParameterList());
+  dd_list->setName("MueLu");
+  dd_list->set("verbosity", "low");
+  dd_list->set("number of equations", 1);
+  dd_list->set("max levels", 1);
+  dd_list->set("coarse: type", "SCHWARZ");  // FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
+
+  ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
+  dd_smooth_sublist.set("schwarz: overlap level", 0);
+  dd_smooth_sublist.set("schwarz: combine mode", "Zero");
+  dd_smooth_sublist.set("subdomain solver name", "RILUK");
+
+  ParameterList& coarse_subdomain_solver = dd_smooth_sublist.sublist("subdomain solver parameters");
+  coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
+  coarse_subdomain_solver.set("fact: absolute threshold", 0.);
+  coarse_subdomain_solver.set("fact: relative threshold", 1.);
+  coarse_subdomain_solver.set("fact: relax value", 0.);
+
+  B_fine_ = CreateTpetraPreconditioner((RCP<operator_type>)A, *dd_list);
+
+  /*B_DD_ = rcp( new Ifpack2::AdditiveSchwarz< row_matrix_type, precond_type >(A, 0) );
+  B_DD_->initialize();
+  B_DD_->compute();*/
 }
 
+void AdditiveVariant::AdditiveCoarseSolver(RCP<crs_matrix_type> A) {
+  // Creation of the MueLu list for the DD preconditioner
+  RCP<ParameterList> coarse_list = rcp(new Teuchos::ParameterList());
+  coarse_list->setName("MueLu");
+  coarse_list->set("verbosity", "low");
+  coarse_list->set("number of equations", 1);
+  coarse_list->set("max levels", 2);
+  coarse_list->set("multigrid algorithm", "unsmoothed");
+  coarse_list->set("aggregation: type", "brick");
+  coarse_list->set("aggregation: brick x size", domain_.bricksize_x);
+  coarse_list->set("aggregation: brick y size", domain_.bricksize_y);
+  coarse_list->set("aggregation: brick z size", domain_.bricksize_z);
+  coarse_list->set("aggregation: drop scheme", "classical");
+  coarse_list->set("smoother: pre or post", "none");
+  coarse_list->set("repartition: enable", true);
+  coarse_list->set("repartition: partitioner", "zoltan");
+  coarse_list->set("repartition: start level", 1);
+  coarse_list->set("repartition: min rows per proc", static_cast<int>(A->getGlobalNumRows()));
+  coarse_list->set("repartition: max imbalance", 1.2);
+  coarse_list->set("repartition: remap parts", false);
+  coarse_list->set("coarse: type", "SCHWARZ");
 
-void
-AdditiveVariant::AdditiveCoarseSolver( RCP<crs_matrix_type> A )
-{
-	  //Creation of the MueLu list for the DD preconditioner
-		RCP<ParameterList> coarse_list = rcp(new Teuchos::ParameterList());
-		coarse_list->setName("MueLu");
-		coarse_list->set("verbosity", "low");
-		coarse_list->set("number of equations", 1);
-		coarse_list->set("max levels", 2);
-		coarse_list->set("multigrid algorithm", "unsmoothed");
-		coarse_list->set("aggregation: type", "brick");
-		coarse_list->set("aggregation: brick x size", domain_.bricksize_x);
-		coarse_list->set("aggregation: brick y size", domain_.bricksize_y);
-		coarse_list->set("aggregation: brick z size", domain_.bricksize_z);
-		coarse_list->set("aggregation: drop scheme", "classical");
-		coarse_list->set("smoother: pre or post", "none");
-		coarse_list->set("repartition: enable", true);
-		coarse_list->set("repartition: partitioner", "zoltan");
-		coarse_list->set("repartition: start level", 1);
-		coarse_list->set("repartition: min rows per proc", static_cast<int>(A->getGlobalNumRows()) );
-		coarse_list->set("repartition: max imbalance", 1.2);
-		coarse_list->set("repartition: remap parts", false);
-		coarse_list->set("coarse: type", "SCHWARZ");
-
-		//Creation of Sublist for smoother	
-	  ParameterList& coarse_smooth_sublist = coarse_list->sublist("coarse: params");
-		coarse_smooth_sublist.set("schwarz: overlap level", 0);
-		coarse_smooth_sublist.set("schwarz: combine mode", "Zero");
-		coarse_smooth_sublist.set("subdomain solver name", "RILUK");
-
-		//Creation of the sublist for the subdomain solver
-		ParameterList& coarse_subdomain_solver = coarse_smooth_sublist.sublist("subdomain solver parameters");
-		coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
-		coarse_subdomain_solver.set("fact: absolute threshold", 0.);
-		coarse_subdomain_solver.set("fact: relative threshold", 1.);
-		coarse_subdomain_solver.set("fact: relax value", 0.);
-
-  	/*ParameterList& coarse_file_sublist = coarse_list->sublist("export data");
-    coarse_file_sublist.set("A", "{0}");
-    coarse_file_sublist.set("P", "{0,1}");
-    coarse_file_sublist.set("R", "{1}");*/
-
-    //Manual set up of the prolongation and restriction
-    MueLu::ParameterListInterpreter<scalar_type> mueLuFactory( *coarse_list );
-    RCP<MueLu::Hierarchy<scalar_type> > H = mueLuFactory.CreateHierarchy();
-    H->setVerbLevel(Teuchos::VERB_HIGH);
-    RCP<xpetra_matrix> mueluA = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>(A);
-	
-    RCP<Xpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type > > coords = Xpetra::toXpetra( coords_	); 	
-
-    H->GetLevel(0)->Set("A", mueluA);
-    H->GetLevel(0)->Set("Coordinates", coords);
-
-    // Multigrid setup phase
-    mueLuFactory.SetupHierarchy(*H);	
-
-  	RCP<Level> L = H->GetLevel(1);
-
-  	RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> prolong, restr;
-
-  	if (L->IsAvailable("P"))
-    	  prolong = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("P");
-
-  	if (L->IsAvailable("R"))
-    	  restr = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("R");
-
-		RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
-		RCP<crs_matrix_type> tpetra_restr = MueLuUtilities::Op2NonConstTpetraCrs(restr);
-
-		int mypid = GlobalComm_->getRank();
-		GlobalComm_->barrier();
-
-	  // We have to transform P into a condensed multivector
-	  RCP<multivector_type> identity_shrunk = rcp( new multivector_type(tpetra_prolong->getDomainMap(), 3) );
-	  Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
-	  typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
-
-	  int my_color = (mypid-1)%3;
-	  Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(my_color);
-
-    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-  	  const local_ordinal_type i_local = *it;
-   	  const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-		  localMV[aux] = 1.0;	
-	  }
-
-	  RCP<multivector_type> P_shrunk = rcp( new multivector_type(tpetra_prolong->getRangeMap(), 3) );
-	  tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
-	  RCP<multivector_type> AP_shrunk = rcp( new multivector_type(A->getRangeMap(), 3) );
-	  A->apply(*P_shrunk, *AP_shrunk);
-
-		TEUCHOS_TEST_FOR_EXCEPT( B_fine_.is_null() );
-
-	//========================================================================================================
-
-	// CREATION OF BAP
-
-	  RCP<multivector_type> BAP_multivector = rcp(new multivector_type(B_fine_->getRangeMap(),AP_shrunk->getNumVectors()));
-		RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_fine_->getRangeMap(),AP_shrunk->getNumVectors()));
-	  B_fine_->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits< scalar_type >::one(), Teuchos::ScalarTraits< scalar_type >::zero());
-
-	  //I just need this to generate the right colMap to populate BAP
-	  RCP<crs_matrix_type> AP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), tpetra_prolong->getColMap(), tpetra_prolong->getGlobalNumRows() ) );
-	  Tpetra::MatrixMatrix::Multiply(*A, false, *tpetra_prolong, false,  *AP, true);
-
-	  GlobalComm_->barrier();
-
-	  RCP<crs_matrix_type> BAP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), AP->getColMap(), tpetra_prolong->getGlobalNumCols() ) );
-	  Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
-
-	  for(int color = 0; color<3; ++color)
-	  {
-		    Teuchos::ArrayRCP<const scalar_type> localBAP = BAP_shrunk->getData(color);
-		
-			  for (iter_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it)
-			  {
-				  const local_ordinal_type i_local = *it;
-    	 	  const local_ordinal_type aux = BAP->getRowMap()->getLocalElement (i_local);
-
-			  	std::vector<global_ordinal_type> BAP_inds;
-				  std::vector<scalar_type> BAP_vals;
-			
-				  local_ordinal_type aux2;
+  // Creation of Sublist for smoother
+  ParameterList& coarse_smooth_sublist = coarse_list->sublist("coarse: params");
+  coarse_smooth_sublist.set("schwarz: overlap level", 0);
+  coarse_smooth_sublist.set("schwarz: combine mode", "Zero");
+  coarse_smooth_sublist.set("subdomain solver name", "RILUK");
+
+  // Creation of the sublist for the subdomain solver
+  ParameterList& coarse_subdomain_solver = coarse_smooth_sublist.sublist("subdomain solver parameters");
+  coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
+  coarse_subdomain_solver.set("fact: absolute threshold", 0.);
+  coarse_subdomain_solver.set("fact: relative threshold", 1.);
+  coarse_subdomain_solver.set("fact: relax value", 0.);
+
+  /*ParameterList& coarse_file_sublist = coarse_list->sublist("export data");
+coarse_file_sublist.set("A", "{0}");
+coarse_file_sublist.set("P", "{0,1}");
+coarse_file_sublist.set("R", "{1}");*/
+
+  // Manual set up of the prolongation and restriction
+  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory(*coarse_list);
+  RCP<MueLu::Hierarchy<scalar_type>> H = mueLuFactory.CreateHierarchy();
+  H->setVerbLevel(Teuchos::VERB_HIGH);
+  RCP<xpetra_matrix> mueluA = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(A);
+
+  RCP<Xpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type>> coords = Xpetra::toXpetra(coords_);
+
+  H->GetLevel(0)->Set("A", mueluA);
+  H->GetLevel(0)->Set("Coordinates", coords);
+
+  // Multigrid setup phase
+  mueLuFactory.SetupHierarchy(*H);
+
+  RCP<Level> L = H->GetLevel(1);
+
+  RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>> prolong, restr;
+
+  if (L->IsAvailable("P"))
+    prolong = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("P");
+
+  if (L->IsAvailable("R"))
+    restr = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("R");
+
+  RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
+  RCP<crs_matrix_type> tpetra_restr   = MueLuUtilities::Op2NonConstTpetraCrs(restr);
+
+  int mypid = GlobalComm_->getRank();
+  GlobalComm_->barrier();
+
+  // We have to transform P into a condensed multivector
+  RCP<multivector_type> identity_shrunk                                  = rcp(new multivector_type(tpetra_prolong->getDomainMap(), 3));
+  Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
+  typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
+
+  int my_color                           = (mypid - 1) % 3;
+  Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(my_color);
+
+  for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+    const local_ordinal_type i_local = *it;
+    const local_ordinal_type aux     = identity_shrunk->getMap()->getLocalElement(i_local);
+    localMV[aux]                     = 1.0;
+  }
+
+  RCP<multivector_type> P_shrunk = rcp(new multivector_type(tpetra_prolong->getRangeMap(), 3));
+  tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
+  RCP<multivector_type> AP_shrunk = rcp(new multivector_type(A->getRangeMap(), 3));
+  A->apply(*P_shrunk, *AP_shrunk);
+
+  TEUCHOS_TEST_FOR_EXCEPT(B_fine_.is_null());
+
+  //========================================================================================================
+
+  // CREATION OF BAP
+
+  RCP<multivector_type> BAP_multivector = rcp(new multivector_type(B_fine_->getRangeMap(), AP_shrunk->getNumVectors()));
+  RCP<multivector_type> BAP_shrunk      = rcp(new multivector_type(B_fine_->getRangeMap(), AP_shrunk->getNumVectors()));
+  B_fine_->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits<scalar_type>::one(), Teuchos::ScalarTraits<scalar_type>::zero());
+
+  // I just need this to generate the right colMap to populate BAP
+  RCP<crs_matrix_type> AP = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), tpetra_prolong->getColMap(), tpetra_prolong->getGlobalNumRows()));
+  Tpetra::MatrixMatrix::Multiply(*A, false, *tpetra_prolong, false, *AP, true);
+
+  GlobalComm_->barrier();
+
+  RCP<crs_matrix_type> BAP                                      = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), AP->getColMap(), tpetra_prolong->getGlobalNumCols()));
+  Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
 
-				  if( (mypid-1)%3==color && (mypid-1)>= 0 && (mypid-1)<tpetra_prolong->getGlobalNumCols() )
-					  aux2 = BAP->getColMap()->getLocalElement (mypid-1);
-				  else if( (mypid-2)%3==color && (mypid-2)>= 0 && (mypid-2)<tpetra_prolong->getGlobalNumCols() )
-					  aux2 = BAP->getColMap()->getLocalElement (mypid-2);
-				  else if( (mypid)%3==color && (mypid)>= 0 && (mypid)<tpetra_prolong->getGlobalNumCols() )
-					  aux2 = BAP->getColMap()->getLocalElement (mypid);
+  for (int color = 0; color < 3; ++color) {
+    Teuchos::ArrayRCP<const scalar_type> localBAP = BAP_shrunk->getData(color);
 
-				  if(aux2>=0)
-				  {
-					  BAP_inds.emplace_back(aux2);
-					  BAP_vals.emplace_back(localBAP[aux]);
-					  BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
-				  }
-			  }
-		
-	  }
-	  BAP->fillComplete( tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap() );
+    for (iter_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it) {
+      const local_ordinal_type i_local = *it;
+      const local_ordinal_type aux     = BAP->getRowMap()->getLocalElement(i_local);
 
-	//=============================================================================================================
+      std::vector<global_ordinal_type> BAP_inds;
+      std::vector<scalar_type> BAP_vals;
 
-    RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
-    RCP<xpetra_matrix> mueluPbar = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>(Pbar);
+      local_ordinal_type aux2;
 
-    H->GetLevel(1)->Set("Pbar", mueluPbar);
+      if ((mypid - 1) % 3 == color && (mypid - 1) >= 0 && (mypid - 1) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid - 1);
+      else if ((mypid - 2) % 3 == color && (mypid - 2) >= 0 && (mypid - 2) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid - 2);
+      else if ((mypid) % 3 == color && (mypid) >= 0 && (mypid) < tpetra_prolong->getGlobalNumCols())
+        aux2 = BAP->getColMap()->getLocalElement(mypid);
 
-    H->IsPreconditioner(true);
-    B_coarse_ = rcp(new muelu_tpetra_operator_type(H));
+      if (aux2 >= 0) {
+        BAP_inds.emplace_back(aux2);
+        BAP_vals.emplace_back(localBAP[aux]);
+        BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
+      }
+    }
+  }
+  BAP->fillComplete(tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap());
 
-}
+  //=============================================================================================================
 
+  RCP<crs_matrix_type> Pbar    = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
+  RCP<xpetra_matrix> mueluPbar = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(Pbar);
 
-void AdditiveVariant::apply( const multivector_type &r, multivector_type &Pr, Teuchos::ETransp mode = Teuchos::NO_TRANS, scalar_type alpha = Teuchos::ScalarTraits< scalar_type >::one(), scalar_type beta = Teuchos::ScalarTraits< scalar_type >::zero() ) const
-{
+  H->GetLevel(1)->Set("Pbar", mueluPbar);
 
+  H->IsPreconditioner(true);
+  B_coarse_ = rcp(new muelu_tpetra_operator_type(H));
+}
 
-	Tpetra::Export<local_ordinal_type, global_ordinal_type, node_type> Export_fine1(r.getMap(), DomainMap_);
-	Tpetra::Export<local_ordinal_type, global_ordinal_type, node_type> Export_fine2(RangeMap_, r.getMap());	
+void AdditiveVariant::apply(const multivector_type& r, multivector_type& Pr, Teuchos::ETransp mode = Teuchos::NO_TRANS, scalar_type alpha = Teuchos::ScalarTraits<scalar_type>::one(), scalar_type beta = Teuchos::ScalarTraits<scalar_type>::zero()) const {
+  Tpetra::Export<local_ordinal_type, global_ordinal_type, node_type> Export_fine1(r.getMap(), DomainMap_);
+  Tpetra::Export<local_ordinal_type, global_ordinal_type, node_type> Export_fine2(RangeMap_, r.getMap());
 
-	multivector_type r_fine(DomainMap_, 1);
-	r_fine.doImport(r, Export_fine1, Tpetra::INSERT);	
-	multivector_type B_fine_Pr(RangeMap_,1);
-	B_fine_->apply(r_fine, B_fine_Pr, mode, alpha, beta);
-	multivector_type B_Pr1(r.getMap(),1);	
-	B_Pr1.doImport(B_fine_Pr, Export_fine2, Tpetra::INSERT);
+  multivector_type r_fine(DomainMap_, 1);
+  r_fine.doImport(r, Export_fine1, Tpetra::INSERT);
+  multivector_type B_fine_Pr(RangeMap_, 1);
+  B_fine_->apply(r_fine, B_fine_Pr, mode, alpha, beta);
+  multivector_type B_Pr1(r.getMap(), 1);
+  B_Pr1.doImport(B_fine_Pr, Export_fine2, Tpetra::INSERT);
 
-	multivector_type r_coarse(DomainMap_, 1);
-	r_coarse.doImport(r, Export_fine1, Tpetra::INSERT);	
-	multivector_type B_coarse_Pr(RangeMap_,1);
-	B_coarse_->apply(r_coarse, B_coarse_Pr, mode, alpha, beta);
-	multivector_type B_Pr2(r.getMap(),1);
-	B_Pr2.doImport(B_coarse_Pr, Export_fine2, Tpetra::INSERT);
+  multivector_type r_coarse(DomainMap_, 1);
+  r_coarse.doImport(r, Export_fine1, Tpetra::INSERT);
+  multivector_type B_coarse_Pr(RangeMap_, 1);
+  B_coarse_->apply(r_coarse, B_coarse_Pr, mode, alpha, beta);
+  multivector_type B_Pr2(r.getMap(), 1);
+  B_Pr2.doImport(B_coarse_Pr, Export_fine2, Tpetra::INSERT);
 
-  multivector_type B_Pr_sum(r.getMap(),1);
-	B_Pr_sum.update(1.0, B_Pr1, 1.0, B_Pr2, 0.0);//Careful to set the correct coefficients!!!
+  multivector_type B_Pr_sum(r.getMap(), 1);
+  B_Pr_sum.update(1.0, B_Pr1, 1.0, B_Pr2, 0.0);  // Careful to set the correct coefficients!!!
 
-	Pr = B_Pr_sum;
+  Pr = B_Pr_sum;
 }
 
-
-}
+}  // namespace MueLu
diff --git a/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.hpp b/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.hpp
index de1ec4059d26..a2e478819805 100644
--- a/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.hpp
+++ b/packages/muelu/research/max/AdditiveMG/Smooth_Prolongation.hpp
@@ -17,81 +17,72 @@
 
 #include <Ifpack2_AdditiveSchwarz_decl.hpp>
 
-namespace MueLu
-{
+namespace MueLu {
 
 // Define default types
-typedef double                                      scalar_type;
-typedef int                                         local_ordinal_type;
-typedef int                                         global_ordinal_type;
+typedef double scalar_type;
+typedef int local_ordinal_type;
+typedef int global_ordinal_type;
 typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
-typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
+typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
 
-typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
 typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
-typedef Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type> xpetra_matrix;
+typedef Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> xpetra_matrix;
 
 typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
+struct DomainPartitioning {
+  global_ordinal_type nx = 0;
+  global_ordinal_type ny = 0;
+  global_ordinal_type nz = 0;
 
-struct DomainPartitioning{
-
-	global_ordinal_type nx = 0;
-	global_ordinal_type ny = 0;
-	global_ordinal_type nz = 0;
-
-	global_ordinal_type bricksize_x = 0;
-	global_ordinal_type bricksize_y = 0;
-	global_ordinal_type bricksize_z = 0;
-	
+  global_ordinal_type bricksize_x = 0;
+  global_ordinal_type bricksize_y = 0;
+  global_ordinal_type bricksize_z = 0;
 };
 
+class AdditiveVariant : public operator_type {
+ public:
+  AdditiveVariant(RCP<crs_matrix_type>, RCP<multivector_type>, DomainPartitioning);
 
-class AdditiveVariant: public operator_type{
-
-	public:
-
-		AdditiveVariant( RCP<crs_matrix_type>, RCP<multivector_type>, DomainPartitioning );
-
-		void apply( const multivector_type &, multivector_type &, Teuchos::ETransp, scalar_type, scalar_type ) const;
-
-		bool hasTransposeApply()const{return false;}
-	
-		RCP< const driver_map_type > getDomainMap() const{return DomainMap_;};
+  void apply(const multivector_type &, multivector_type &, Teuchos::ETransp, scalar_type, scalar_type) const;
 
-		RCP< const driver_map_type > getRangeMap() const{return RangeMap_;};
+  bool hasTransposeApply() const { return false; }
 
-	private:		
+  RCP<const driver_map_type> getDomainMap() const { return DomainMap_; };
 
-		DomainPartitioning domain_;
+  RCP<const driver_map_type> getRangeMap() const { return RangeMap_; };
 
-		RCP< const Teuchos::Comm<int> > GlobalComm_;
+ private:
+  DomainPartitioning domain_;
 
-		RCP< multivector_type > coords_;
+  RCP<const Teuchos::Comm<int> > GlobalComm_;
 
-		RCP< const driver_map_type > DomainMap_;
+  RCP<multivector_type> coords_;
 
-		RCP< const driver_map_type > RangeMap_;
+  RCP<const driver_map_type> DomainMap_;
 
-		//MueLu Preconditioner to store the smoother at the fine level
-		RCP<muelu_tpetra_operator_type>	B_fine_ = Teuchos::null;
+  RCP<const driver_map_type> RangeMap_;
 
-	  //RCP<Ifpack2::AdditiveSchwarz< row_matrix_type, precond_type > > B_DD_;
+  // MueLu Preconditioner to store the smoother at the fine level
+  RCP<muelu_tpetra_operator_type> B_fine_ = Teuchos::null;
 
-		//MueLu Preconditioner to store
-		RCP<muelu_tpetra_operator_type>	B_coarse_ = Teuchos::null;
+  // RCP<Ifpack2::AdditiveSchwarz< row_matrix_type, precond_type > > B_DD_;
 
-		void AdditiveFineSmoother( RCP<crs_matrix_type> );
+  // MueLu Preconditioner to store
+  RCP<muelu_tpetra_operator_type> B_coarse_ = Teuchos::null;
 
-	  void AdditiveCoarseSolver( RCP<crs_matrix_type> );
+  void AdditiveFineSmoother(RCP<crs_matrix_type>);
 
+  void AdditiveCoarseSolver(RCP<crs_matrix_type>);
 };
 
-}
+}  // namespace MueLu
diff --git a/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveBricks.cpp b/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveBricks.cpp
index 3cc47a5d7616..edd37f3e5c90 100644
--- a/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveBricks.cpp
+++ b/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveBricks.cpp
@@ -77,7 +77,7 @@ extern clock_t Timers_Max[4];
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -88,24 +88,23 @@ extern clock_t Timers_Max[4];
 #include "BAP.hpp"
 #include "CreateBrickMap.hpp"
 
-int main(int argc, char *argv[]) {
-
+int main(int argc, char* argv[]) {
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-	typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -113,50 +112,50 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-	typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
+  typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-  typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                          = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "final_parser.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "final_parser.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Teuchos::ParameterList xmlParams;
@@ -167,7 +166,7 @@ int main(int argc, char *argv[]) {
   problemParams = xmlParams.sublist(static_cast<const std::string>("Problem"));
 
   // Problem definition
-   std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
+  std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
 
   // Parameters
 
@@ -184,14 +183,13 @@ int main(int argc, char *argv[]) {
   MueLu::DomainPartitioning domain;
 
   int keep_boundary = 0;
-  Scalar stretchx = (Scalar) Lx/nx;
-  Scalar stretchy = (Scalar) Ly/ny;
-  Scalar stretchz = (Scalar) Lz/nz;
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
- 
   if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {
     throw std::invalid_argument("This example only supports Tpetra.");
   }
@@ -204,59 +202,56 @@ int main(int argc, char *argv[]) {
   RCP<const Map> xpetraMap;
   std::vector<global_ordinal_type> ind;
 
-  //BRICK SIZE
+  // BRICK SIZE
   int brick_sizex = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick x size"));
   int brick_sizey = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick y size"));
   int brick_sizez = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick z size"));
 
-  //Creation of the map where processor 0 gets nothing at the fine level
-  if( comm->getSize()>1 )
-  {
-	if(problem_type == "ADR1D")
-		createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm );
-
-	else if(problem_type == "ADR2D")
-		createBrickMap2D( nx, brick_sizex, brick_sizey, ind, comm );
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    if (problem_type == "ADR1D")
+      createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm);
 
-	else if(problem_type == "ADR3D")
-		createBrickMap3D( nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm );
+    else if (problem_type == "ADR2D")
+      createBrickMap2D(nx, brick_sizex, brick_sizey, ind, comm);
 
-	ind.shrink_to_fit();
+    else if (problem_type == "ADR3D")
+      createBrickMap3D(nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm);
 
-	Teuchos::ArrayView<const global_ordinal_type> elementList (ind);
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-  }
-  else if( comm->getSize()==1 )
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+    ind.shrink_to_fit();
 
+    Teuchos::ArrayView<const global_ordinal_type> elementList(ind);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1)
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
 
   RCP<MultiVector> coordinates;
 
   if (problem_type == "ADR1D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR2D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR3D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
 
- RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-  RCP<Matrix>  xpetraA = Pr->BuildMatrix();
+  RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
 
-  RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
   RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
 
- // ===================================================
- // 	Domain Decomposition Preconditioner
- // 	===================================
+  // ===================================================
+  // 	Domain Decomposition Preconditioner
+  // 	===================================
 
-  //Creation of the MueLu list for the DD preconditioner
+  // Creation of the MueLu list for the DD preconditioner
   RCP<Teuchos::ParameterList> dd_list = rcp(new Teuchos::ParameterList());
   dd_list->setName("MueLu");
   dd_list->set("verbosity", "low");
   dd_list->set("number of equations", 1);
   dd_list->set("max levels", 1);
-  dd_list->set("coarse: type", "SCHWARZ"); //FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
-		
+  dd_list->set("coarse: type", "SCHWARZ");  // FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
+
   Teuchos::ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
   dd_smooth_sublist.set("schwarz: overlap level", 0);
   dd_smooth_sublist.set("schwarz: combine mode", "Zero");
@@ -268,152 +263,137 @@ int main(int argc, char *argv[]) {
   coarse_subdomain_solver.set("fact: relative threshold", 1.);
   coarse_subdomain_solver.set("fact: relax value", 0.);
 
-  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, *dd_list );
+  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, *dd_list);
 
- // ===================================================
- // 	Multi Grid Preconditioner
- // 	===================================
+  // ===================================================
+  // 	Multi Grid Preconditioner
+  // 	===================================
 
   RCP<muelu_tpetra_operator_type> M;
 
-  //Manual set up of the prolongation and restriction
-  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory( mueluParams );
-  RCP<MueLu::Hierarchy<scalar_type> > H = mueLuFactory.CreateHierarchy();
+  // Manual set up of the prolongation and restriction
+  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory(mueluParams);
+  RCP<MueLu::Hierarchy<scalar_type>> H = mueLuFactory.CreateHierarchy();
   H->setVerbLevel(Teuchos::VERB_HIGH);
-	
+
   H->GetLevel(0)->Set("A", xpetraA);
   H->GetLevel(0)->Set("Coordinates", coordinates);
 
   // Multigrid setup phase
-  mueLuFactory.SetupHierarchy(*H);	
+  mueLuFactory.SetupHierarchy(*H);
 
   RCP<Level> L = H->GetLevel(1);
 
-  RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> prolong, restr;
+  RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>> prolong, restr;
 
   if (L->IsAvailable("P"))
-    prolong = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("P");
+    prolong = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("P");
 
   if (L->IsAvailable("R"))
-    restr = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("R");
+    restr = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("R");
 
   RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
-  RCP<crs_matrix_type> tpetra_restr = MueLuUtilities::Op2NonConstTpetraCrs(restr);
+  RCP<crs_matrix_type> tpetra_restr   = MueLuUtilities::Op2NonConstTpetraCrs(restr);
 
-  #include <Teuchos_TimeMonitor.hpp>
-  RCP<Teuchos::Time> PbarSetUp    = Teuchos::TimeMonitor::getNewCounter("Pbar: SetUp");
+#include <Teuchos_TimeMonitor.hpp>
+  RCP<Teuchos::Time> PbarSetUp = Teuchos::TimeMonitor::getNewCounter("Pbar: SetUp");
   PbarSetUp->start();
   RCP<Matrix> mueluPbar;
- 
+
   // We have to transform P into a condensed multivector
-  RCP<multivector_type> identity_shrunk = rcp( new multivector_type(tpetra_prolong->getDomainMap(), std::pow(3, coordinates->getNumVectors()) ) );
+  RCP<multivector_type> identity_shrunk                                  = rcp(new multivector_type(tpetra_prolong->getDomainMap(), std::pow(3, coordinates->getNumVectors())));
   Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
   typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
 
-  for(int trial = 1; trial<=number_runs; ++trial)
-  {
-        if(1 == coordinates->getNumVectors())
-	{
-
-	  for(int j = 0; j<3; ++j)
-	  {
-	    int color = j;
-	
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-		    const local_ordinal_type i_local = *it;
-		    const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-                    int local_color = (i_local)%3;
-
-	      if( local_color == color )
-			localMV[aux] = 1.0;	
-		
-	     }
-	   }
-	}
-	else if(2 == coordinates->getNumVectors())
-	{
-
-	  for(int j = 0; j<9; ++j)
-	  {
-	    int color = j;
-			
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-		    const local_ordinal_type i_local = *it;
-		    const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-                    const local_ordinal_type local_color = coloring2D( i_local+1, std::sqrt( tpetra_prolong->getGlobalNumCols() ) );
-
-	      if( local_color == color  )
-			    localMV[aux] = 1.0;	
-		
-	     }
-	   }
-	}
-	else if(3 == coordinates->getNumVectors())
-	{
-          //const local_ordinal_type local_color = coloring3D( mypid, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
-
-	  for(int j = 0; j<27; ++j)
-	  {
-	    int color = j;
-	
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-              const local_ordinal_type i_local = *it;
-	      const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-              const local_ordinal_type local_color = coloring3D( i_local+1, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
-
-	      if( local_color == color  )
-	         localMV[aux] = 1.0;	
-	    }
-	  }
-	}
-	
-
-  RCP<multivector_type> P_shrunk = rcp( new multivector_type( tpetra_prolong->getRangeMap(), std::pow(3, coordinates->getNumVectors()) ) );
-  tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
-  RCP<multivector_type> AP_shrunk = rcp( new multivector_type( A->getRangeMap(), std::pow(3, coordinates->getNumVectors()) ) );
-  A->apply(*P_shrunk, *AP_shrunk);
-
-  //========================================================================================================
-
-  // CREATION OF BAP
-
-  RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_DD->getRangeMap(),AP_shrunk->getNumVectors()));
-  B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits< scalar_type >::one(), Teuchos::ScalarTraits< scalar_type >::zero());
-
-   //Columns Map for BAP
-   std::vector<int> indBAPcolMap;
-    if(1 == coordinates->getNumVectors())
-      neighbours1D( tpetra_prolong, indBAPcolMap, comm );
-    else if(2 == coordinates->getNumVectors())
-      neighbours2D( tpetra_prolong, indBAPcolMap, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()) );
-    else if(3 == coordinates->getNumVectors())
-      neighbours3D( tpetra_prolong, indBAPcolMap, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()) );
-
-  Teuchos::ArrayView<const global_ordinal_type> elementListBAP (indBAPcolMap);
-  Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > BAPcolMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm ) );
-
-  RCP<crs_matrix_type> BAP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols() ) );
-  Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
-
-
-  if(1 == coordinates->getNumVectors())
-    BAP1D(BAP, tpetra_prolong, BAP_shrunk, comm);
-  else if(2 == coordinates->getNumVectors())
-   BAP2D( BAP, tpetra_prolong, BAP_shrunk, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()) );
-  else if(3 == coordinates->getNumVectors())
-   BAP3D( BAP, tpetra_prolong, BAP_shrunk, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()) );
-
-  BAP->fillComplete( tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap() );
-	
-//=============================================================================================================
-  RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
-  mueluPbar = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>(Pbar);
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    if (1 == coordinates->getNumVectors()) {
+      for (int j = 0; j < 3; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local = *it;
+          const local_ordinal_type aux     = identity_shrunk->getMap()->getLocalElement(i_local);
+          int local_color                  = (i_local) % 3;
+
+          if (local_color == color)
+            localMV[aux] = 1.0;
+        }
+      }
+    } else if (2 == coordinates->getNumVectors()) {
+      for (int j = 0; j < 9; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local     = *it;
+          const local_ordinal_type aux         = identity_shrunk->getMap()->getLocalElement(i_local);
+          const local_ordinal_type local_color = coloring2D(i_local + 1, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+
+          if (local_color == color)
+            localMV[aux] = 1.0;
+        }
+      }
+    } else if (3 == coordinates->getNumVectors()) {
+      // const local_ordinal_type local_color = coloring3D( mypid, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
+
+      for (int j = 0; j < 27; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local     = *it;
+          const local_ordinal_type aux         = identity_shrunk->getMap()->getLocalElement(i_local);
+          const local_ordinal_type local_color = coloring3D(i_local + 1, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+          if (local_color == color)
+            localMV[aux] = 1.0;
+        }
+      }
+    }
+
+    RCP<multivector_type> P_shrunk = rcp(new multivector_type(tpetra_prolong->getRangeMap(), std::pow(3, coordinates->getNumVectors())));
+    tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
+    RCP<multivector_type> AP_shrunk = rcp(new multivector_type(A->getRangeMap(), std::pow(3, coordinates->getNumVectors())));
+    A->apply(*P_shrunk, *AP_shrunk);
+
+    //========================================================================================================
+
+    // CREATION OF BAP
+
+    RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_DD->getRangeMap(), AP_shrunk->getNumVectors()));
+    B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits<scalar_type>::one(), Teuchos::ScalarTraits<scalar_type>::zero());
+
+    // Columns Map for BAP
+    std::vector<int> indBAPcolMap;
+    if (1 == coordinates->getNumVectors())
+      neighbours1D(tpetra_prolong, indBAPcolMap, comm);
+    else if (2 == coordinates->getNumVectors())
+      neighbours2D(tpetra_prolong, indBAPcolMap, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+    else if (3 == coordinates->getNumVectors())
+      neighbours3D(tpetra_prolong, indBAPcolMap, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+    Teuchos::ArrayView<const global_ordinal_type> elementListBAP(indBAPcolMap);
+    Teuchos::RCP<Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>> BAPcolMap = rcp(new Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>(static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm));
+
+    RCP<crs_matrix_type> BAP                                      = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols()));
+    Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
+
+    if (1 == coordinates->getNumVectors())
+      BAP1D(BAP, tpetra_prolong, BAP_shrunk, comm);
+    else if (2 == coordinates->getNumVectors())
+      BAP2D(BAP, tpetra_prolong, BAP_shrunk, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+    else if (3 == coordinates->getNumVectors())
+      BAP3D(BAP, tpetra_prolong, BAP_shrunk, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+    BAP->fillComplete(tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap());
+
+    //=============================================================================================================
+    RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
+    mueluPbar                 = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(Pbar);
   }
   PbarSetUp->stop();
 
@@ -423,28 +403,28 @@ int main(int argc, char *argv[]) {
 
   M = rcp(new muelu_tpetra_operator_type(H));
 
-  //Intermediate print before zeroing out the global timers (needed to split set up timing and solve timing)
+  // Intermediate print before zeroing out the global timers (needed to split set up timing and solve timing)
   Teuchos::TimeMonitor::summarize();
   Teuchos::TimeMonitor::zeroOutTimers();
   //===============================================================================================================
 
   // Linear Solver
 
-  RCP<multivector_type> X_muelu = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
+  RCP<multivector_type> X_muelu = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B       = rcp(new multivector_type(map, 1));
   RCP<linear_problem_type> Problem_muelu;
-  X_muelu->putScalar((scalar_type) 0.0);
+  X_muelu->putScalar((scalar_type)0.0);
   B->randomize();
 
   Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
 
   RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  //belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Verbosity",             Belos::Errors);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
+  belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+  belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+  // belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  belosList->set("Verbosity", Belos::Errors);
+  belosList->set("Output Frequency", 1);
+  belosList->set("Output Style", Belos::Brief);
   belosList->set("Implicit Residual Scaling", "None");
   RCP<belos_solver_manager_type> solver;
   if (krylovSolverType == "cg")
@@ -456,64 +436,62 @@ int main(int argc, char *argv[]) {
   else
     throw std::invalid_argument("bad Krylov solver type");
 
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-
-  X_muelu->putScalar((scalar_type) 0.0);
-  B->randomize();
-
-  //
-  // Set up Krylov solver and iterate.
-  //
- 
-  Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
-  Problem_muelu->setRightPrec(M);
-  Problem_muelu->setProblem();
-
-  solver->setProblem(Problem_muelu);
-  solver->solve();
-  int numIterations_muelu = solver->getNumIters();
-
-  Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
-  multivector_type residual_muelu(B->getMap(),1);
-  A->apply(*X_muelu, residual_muelu);
-  residual_muelu.update(1.0, *B, -1.0);
-  residual_muelu.norm2(normVec_muelu);
-  if (mypid == 0) {
-    std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
-    std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
- }
-}
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    X_muelu->putScalar((scalar_type)0.0);
+    B->randomize();
+
+    //
+    // Set up Krylov solver and iterate.
+    //
+
+    Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
+    Problem_muelu->setRightPrec(M);
+    Problem_muelu->setProblem();
+
+    solver->setProblem(Problem_muelu);
+    solver->solve();
+    int numIterations_muelu = solver->getNumIters();
+
+    Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
+    multivector_type residual_muelu(B->getMap(), 1);
+    A->apply(*X_muelu, residual_muelu);
+    residual_muelu.update(1.0, *B, -1.0);
+    residual_muelu.norm2(normVec_muelu);
+    if (mypid == 0) {
+      std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
+      std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
+    }
+  }
 
-  Teuchos::TimeMonitor::summarize ();
-/*
-   //Row Map for Timers vectors
-   std::vector<int> indTimerMap;
-   
-   indTimerMap.emplace_back(mypid);
-
-   Teuchos::ArrayView<const global_ordinal_type> elementListTimer (indTimerMap);
-   Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > TimerMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(comm->getSize()), elementListTimer, indexBase, comm ) );
-
-  RCP<multivector_type> TimerRestr = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerProlong = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerFine = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerCoarse = rcp(new multivector_type(TimerMap,1));
-
-  Teuchos::ArrayRCP<scalar_type> localMV = TimerRestr->getDataNonConst(0);
-  const local_ordinal_type aux = TimerMap->getLocalElement (mypid);
-  localMV[aux] = (static_cast<double>(Timers_Max[0])/CLOCKS_PER_SEC/number_runs);
-  localMV = TimerProlong->getDataNonConst(0);
-  localMV[aux] = (static_cast<double>(Timers_Max[1])/CLOCKS_PER_SEC/number_runs);
-  localMV = TimerFine->getDataNonConst(0);
-  localMV[aux] = (static_cast<double>(Timers_Max[2])/CLOCKS_PER_SEC/number_runs);
-  localMV = TimerCoarse->getDataNonConst(0);
-  localMV[aux] = (static_cast<double>(Timers_Max[3])/CLOCKS_PER_SEC/number_runs);
-
-  Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeRestr.mtx", TimerRestr);
-  Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeProlong.mtx", TimerProlong);
-  Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeFine.mtx", TimerFine);
-  Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeCoarse.mtx", TimerCoarse);
-*/
+  Teuchos::TimeMonitor::summarize();
+  /*
+     //Row Map for Timers vectors
+     std::vector<int> indTimerMap;
+
+     indTimerMap.emplace_back(mypid);
+
+     Teuchos::ArrayView<const global_ordinal_type> elementListTimer (indTimerMap);
+     Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > TimerMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(comm->getSize()), elementListTimer, indexBase, comm ) );
+
+    RCP<multivector_type> TimerRestr = rcp(new multivector_type(TimerMap,1));
+    RCP<multivector_type> TimerProlong = rcp(new multivector_type(TimerMap,1));
+    RCP<multivector_type> TimerFine = rcp(new multivector_type(TimerMap,1));
+    RCP<multivector_type> TimerCoarse = rcp(new multivector_type(TimerMap,1));
+
+    Teuchos::ArrayRCP<scalar_type> localMV = TimerRestr->getDataNonConst(0);
+    const local_ordinal_type aux = TimerMap->getLocalElement (mypid);
+    localMV[aux] = (static_cast<double>(Timers_Max[0])/CLOCKS_PER_SEC/number_runs);
+    localMV = TimerProlong->getDataNonConst(0);
+    localMV[aux] = (static_cast<double>(Timers_Max[1])/CLOCKS_PER_SEC/number_runs);
+    localMV = TimerFine->getDataNonConst(0);
+    localMV[aux] = (static_cast<double>(Timers_Max[2])/CLOCKS_PER_SEC/number_runs);
+    localMV = TimerCoarse->getDataNonConst(0);
+    localMV[aux] = (static_cast<double>(Timers_Max[3])/CLOCKS_PER_SEC/number_runs);
+
+    Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeRestr.mtx", TimerRestr);
+    Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeProlong.mtx", TimerProlong);
+    Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeFine.mtx", TimerFine);
+    Tpetra::MatrixMarket::Writer<multivector_type>::writeDenseFile("TimeCoarse.mtx", TimerCoarse);
+  */
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveStride.cpp b/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveStride.cpp
index 903af95b2380..40e12892d214 100644
--- a/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveStride.cpp
+++ b/packages/muelu/research/max/AdditiveMG/SmoothedAdditiveStride.cpp
@@ -61,7 +61,6 @@
 #include <iostream>
 #include <cmath>
 
-
 // Belos provides Krylov solvers
 #include <BelosConfigDefs.hpp>
 #include <BelosLinearProblem.hpp>
@@ -77,7 +76,7 @@
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -87,24 +86,23 @@
 //#include "coloring.hpp"
 #include "BAP.hpp"
 
-int main(int argc, char *argv[]) {
-
+int main(int argc, char* argv[]) {
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-	typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -112,50 +110,50 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-	typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
+  typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-  typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                          = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "final_parser.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "final_parser.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Teuchos::ParameterList xmlParams;
@@ -166,7 +164,7 @@ int main(int argc, char *argv[]) {
   problemParams = xmlParams.sublist(static_cast<const std::string>("Problem"));
 
   // Problem definition
-   std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
+  std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
 
   // Parameters
 
@@ -183,14 +181,12 @@ int main(int argc, char *argv[]) {
   MueLu::DomainPartitioning domain;
 
   int keep_boundary = 0;
-  Scalar stretchx = (Scalar) Lx/nx;
-  Scalar stretchy = (Scalar) Ly/ny;
-  Scalar stretchz = (Scalar) Lz/nz;
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
-
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
- 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
   //
   // Construct the problem
@@ -200,54 +196,51 @@ int main(int argc, char *argv[]) {
   RCP<const Map> xpetraMap;
   std::vector<global_ordinal_type> ind;
 
-  //Creation of the map where processor 0 gets nothing at the fine level
-  if( comm->getSize()>1 )
-  {
-	ind.reserve (static_cast<int>( matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ) + 1 ));
-	if( mypid !=0 && mypid!=comm->getSize()-1 )
-		for(int i = 0; i <= ( static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 ))) - 1; ++i)
-			ind.emplace_back( (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )) + i );
-		
-		if( mypid==comm->getSize()-1 )
-			for(int i = (mypid-1) * static_cast<int>(matrixParameters.GetNumGlobalElements()/( comm->getSize()-1 )); i != matrixParameters.GetNumGlobalElements(); ++i)
-				ind.emplace_back( i );
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    ind.reserve(static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1) + 1));
+    if (mypid != 0 && mypid != comm->getSize() - 1)
+      for (int i = 0; i <= (static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1))) - 1; ++i)
+        ind.emplace_back((mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)) + i);
 
-		ind.shrink_to_fit();
+    if (mypid == comm->getSize() - 1)
+      for (int i = (mypid - 1) * static_cast<int>(matrixParameters.GetNumGlobalElements() / (comm->getSize() - 1)); i != matrixParameters.GetNumGlobalElements(); ++i)
+        ind.emplace_back(i);
 
+    ind.shrink_to_fit();
 
-		Teuchos::ArrayView<const global_ordinal_type> elementList (ind);
-		xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-   }
-   else if( comm->getSize()==1 )
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+    Teuchos::ArrayView<const global_ordinal_type> elementList(ind);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1)
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
 
   RCP<MultiVector> coordinates;
 
   if (problem_type == "ADR1D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR2D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
   else if (problem_type == "ADR3D")
-  	coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
 
   RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-   RCP<Matrix>  xpetraA = Pr->BuildMatrix();
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
 
-  RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
   RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
 
- // ===================================================
- // 	Domain Decomposition Preconditioner
- // 	===================================
+  // ===================================================
+  // 	Domain Decomposition Preconditioner
+  // 	===================================
 
-  //Creation of the MueLu list for the DD preconditioner
+  // Creation of the MueLu list for the DD preconditioner
   RCP<Teuchos::ParameterList> dd_list = rcp(new Teuchos::ParameterList());
   dd_list->setName("MueLu");
   dd_list->set("verbosity", "low");
   dd_list->set("number of equations", 1);
   dd_list->set("max levels", 1);
-  dd_list->set("coarse: type", "SCHWARZ"); //FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
-		
+  dd_list->set("coarse: type", "SCHWARZ");  // FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
+
   Teuchos::ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
   dd_smooth_sublist.set("schwarz: overlap level", 0);
   dd_smooth_sublist.set("schwarz: combine mode", "Zero");
@@ -259,157 +252,150 @@ int main(int argc, char *argv[]) {
   coarse_subdomain_solver.set("fact: relative threshold", 1.);
   coarse_subdomain_solver.set("fact: relax value", 0.);
 
-  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, *dd_list );
+  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, *dd_list);
 
- // ===================================================
- // 	Multi Grid Preconditioner
- // 	===================================
+  // ===================================================
+  // 	Multi Grid Preconditioner
+  // 	===================================
   RCP<muelu_tpetra_operator_type> M;
 
-  //Manual set up of the prolongation and restriction
-  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory( mueluParams );
-  RCP<MueLu::Hierarchy<scalar_type> > H = mueLuFactory.CreateHierarchy();
+  // Manual set up of the prolongation and restriction
+  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory(mueluParams);
+  RCP<MueLu::Hierarchy<scalar_type>> H = mueLuFactory.CreateHierarchy();
   H->setVerbLevel(Teuchos::VERB_HIGH);
-	
+
   H->GetLevel(0)->Set("A", xpetraA);
   H->GetLevel(0)->Set("Coordinates", coordinates);
 
   // Multigrid setup phase
-  mueLuFactory.SetupHierarchy(*H);	
+  mueLuFactory.SetupHierarchy(*H);
 
   RCP<Level> L = H->GetLevel(1);
 
-  RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> prolong, restr;
+  RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>> prolong, restr;
 
   if (L->IsAvailable("P"))
-    prolong = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("P");
+    prolong = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("P");
 
   if (L->IsAvailable("R"))
-    restr = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("R");
+    restr = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("R");
 
   RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
-  RCP<crs_matrix_type> tpetra_restr = MueLuUtilities::Op2NonConstTpetraCrs(restr);
+  RCP<crs_matrix_type> tpetra_restr   = MueLuUtilities::Op2NonConstTpetraCrs(restr);
 
-  Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("P.mtx", tpetra_prolong); //Auxiliary prints introduced to generate pictures
+  Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("P.mtx", tpetra_prolong);  // Auxiliary prints introduced to generate pictures
 
   RCP<Matrix> mueluPbar;
 
   // We have to transform P into a condensed multivector
-  RCP<multivector_type> identity_shrunk = rcp( new multivector_type(tpetra_prolong->getDomainMap(), 3) );
+  RCP<multivector_type> identity_shrunk                                  = rcp(new multivector_type(tpetra_prolong->getDomainMap(), 3));
   Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
   typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
 
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-    for(int j = 0; j<tpetra_prolong->getGlobalNumCols(); ++j)
-    {
-        int color = j%3;
-	
-        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-    	    const local_ordinal_type i_local = *it;
-    	    const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-
-          if(i_local%3 == color)
-    		localMV[aux] = 1.0;	
-		
-         }
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    for (int j = 0; j < tpetra_prolong->getGlobalNumCols(); ++j) {
+      int color = j % 3;
+
+      Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+      for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+        const local_ordinal_type i_local = *it;
+        const local_ordinal_type aux     = identity_shrunk->getMap()->getLocalElement(i_local);
+
+        if (i_local % 3 == color)
+          localMV[aux] = 1.0;
+      }
     }
 
-  RCP<multivector_type> P_shrunk = rcp( new multivector_type(tpetra_prolong->getRangeMap(), 3) );
-  tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
-  RCP<multivector_type> AP_shrunk = rcp( new multivector_type(A->getRangeMap(), 3) );
-  A->apply(*P_shrunk, *AP_shrunk);
+    RCP<multivector_type> P_shrunk = rcp(new multivector_type(tpetra_prolong->getRangeMap(), 3));
+    tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
+    RCP<multivector_type> AP_shrunk = rcp(new multivector_type(A->getRangeMap(), 3));
+    A->apply(*P_shrunk, *AP_shrunk);
 
-//========================================================================================================
+    //========================================================================================================
 
-// CREATION OF BAP
+    // CREATION OF BAP
 
-  RCP<multivector_type> BAP_multivector = rcp(new multivector_type(B_DD->getRangeMap(),AP_shrunk->getNumVectors()));
-	RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_DD->getRangeMap(),AP_shrunk->getNumVectors()));
-  B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits< scalar_type >::one(), Teuchos::ScalarTraits< scalar_type >::zero());
+    RCP<multivector_type> BAP_multivector = rcp(new multivector_type(B_DD->getRangeMap(), AP_shrunk->getNumVectors()));
+    RCP<multivector_type> BAP_shrunk      = rcp(new multivector_type(B_DD->getRangeMap(), AP_shrunk->getNumVectors()));
+    B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits<scalar_type>::one(), Teuchos::ScalarTraits<scalar_type>::zero());
 
-  //I just need this to generate the right colMap to populate BAP
-  RCP<crs_matrix_type> AP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), tpetra_prolong->getColMap(), tpetra_prolong->getGlobalNumRows() ) );
-  Tpetra::MatrixMatrix::Multiply(*A, false, *tpetra_prolong, false,  *AP, true);
+    // I just need this to generate the right colMap to populate BAP
+    RCP<crs_matrix_type> AP = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), tpetra_prolong->getColMap(), tpetra_prolong->getGlobalNumRows()));
+    Tpetra::MatrixMatrix::Multiply(*A, false, *tpetra_prolong, false, *AP, true);
 
-  //Columns Map for BAP
-  std::vector<int> indBAPcolMap;
-  for(int i = 0; i<static_cast<size_t>(tpetra_prolong->getGlobalNumCols()); ++i)
-    indBAPcolMap.emplace_back(i);
+    // Columns Map for BAP
+    std::vector<int> indBAPcolMap;
+    for (int i = 0; i < static_cast<size_t>(tpetra_prolong->getGlobalNumCols()); ++i)
+      indBAPcolMap.emplace_back(i);
 
-  Teuchos::ArrayView<const global_ordinal_type> elementListBAP (indBAPcolMap);
-  Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > BAPcolMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm ) );
+    Teuchos::ArrayView<const global_ordinal_type> elementListBAP(indBAPcolMap);
+    Teuchos::RCP<Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>> BAPcolMap = rcp(new Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>(static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm));
 
-   RCP<crs_matrix_type> BAP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols() ) );
-   Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
+    RCP<crs_matrix_type> BAP                                      = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols()));
+    Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
 
-  for(int color = 0; color<3; ++color)
-  {
+    for (int color = 0; color < 3; ++color) {
       Teuchos::ArrayRCP<const scalar_type> localBAP = BAP_shrunk->getData(color);
-	
-      for (iter_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it)
-      {
-   	  const local_ordinal_type i_local = *it;
-	  const local_ordinal_type aux = BAP->getRowMap()->getLocalElement (i_local);
-
-          std::vector<global_ordinal_type> BAP_inds;
-	  std::vector<scalar_type> BAP_vals;
-		
-	  local_ordinal_type aux2;
-
-	  if( (mypid-1)%3==color && (mypid-1)>= 0 && (mypid-1)<tpetra_prolong->getGlobalNumCols() )
-	  	  aux2 = BAP->getColMap()->getLocalElement (mypid-1);
-	  else if( (mypid-2)%3==color && (mypid-2)>= 0 && (mypid-2)<tpetra_prolong->getGlobalNumCols() )
-	  	  aux2 = BAP->getColMap()->getLocalElement (mypid-2);
-	  else if( (mypid)%3==color && (mypid)>= 0 && (mypid)<tpetra_prolong->getGlobalNumCols() )
-	  	  aux2 = BAP->getColMap()->getLocalElement (mypid);
- 
-          if(aux2>=0)
-	  {
-	  	  BAP_inds.emplace_back(aux2);
-	  	  BAP_vals.emplace_back(localBAP[aux]);
-	  	  BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
-	  }
+
+      for (iter_type it = myLocalElements.begin(); it != myLocalElements.end(); ++it) {
+        const local_ordinal_type i_local = *it;
+        const local_ordinal_type aux     = BAP->getRowMap()->getLocalElement(i_local);
+
+        std::vector<global_ordinal_type> BAP_inds;
+        std::vector<scalar_type> BAP_vals;
+
+        local_ordinal_type aux2;
+
+        if ((mypid - 1) % 3 == color && (mypid - 1) >= 0 && (mypid - 1) < tpetra_prolong->getGlobalNumCols())
+          aux2 = BAP->getColMap()->getLocalElement(mypid - 1);
+        else if ((mypid - 2) % 3 == color && (mypid - 2) >= 0 && (mypid - 2) < tpetra_prolong->getGlobalNumCols())
+          aux2 = BAP->getColMap()->getLocalElement(mypid - 2);
+        else if ((mypid) % 3 == color && (mypid) >= 0 && (mypid) < tpetra_prolong->getGlobalNumCols())
+          aux2 = BAP->getColMap()->getLocalElement(mypid);
+
+        if (aux2 >= 0) {
+          BAP_inds.emplace_back(aux2);
+          BAP_vals.emplace_back(localBAP[aux]);
+          BAP->insertLocalValues(aux, BAP_inds, BAP_vals);
+        }
       }
-	
-  }
-  BAP->fillComplete( tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap() );
+    }
+    BAP->fillComplete(tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap());
 
-  //=============================================================================================================
-  RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
-  mueluPbar = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>(Pbar);
-  Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("Pbar.mtx", Pbar); //Auxiliary prints introduced to generate pictures
-}
+    //=============================================================================================================
+    RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
+    mueluPbar                 = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(Pbar);
+    Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("Pbar.mtx", Pbar);  // Auxiliary prints introduced to generate pictures
+  }
   H->GetLevel(1)->Set("Pbar", mueluPbar);
 
   H->IsPreconditioner(true);
 
   M = rcp(new muelu_tpetra_operator_type(H));
 
-  //Intermediate print before zeroing out the global timers (needed to split set up timing and solve timing)
+  // Intermediate print before zeroing out the global timers (needed to split set up timing and solve timing)
   Teuchos::TimeMonitor::summarize();
   Teuchos::TimeMonitor::zeroOutTimers();
   //======================================================================================================================
 
   // Linear Solver
-    
-  RCP<multivector_type> X_muelu = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
+
+  RCP<multivector_type> X_muelu = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B       = rcp(new multivector_type(map, 1));
   RCP<linear_problem_type> Problem_muelu;
-  X_muelu->putScalar((scalar_type) 0.0);
+  X_muelu->putScalar((scalar_type)0.0);
   B->randomize();
 
   Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
 
   RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  //belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Verbosity",             Belos::Errors);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
+  belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+  belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+  // belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  belosList->set("Verbosity", Belos::Errors);
+  belosList->set("Output Frequency", 1);
+  belosList->set("Output Style", Belos::Brief);
   belosList->set("Implicit Residual Scaling", "None");
   RCP<belos_solver_manager_type> solver;
   if (krylovSolverType == "cg")
@@ -421,36 +407,34 @@ for(int trial = 1; trial<=number_runs; ++trial)
   else
     throw std::invalid_argument("bad Krylov solver type");
 
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-
-  X_muelu->putScalar((scalar_type) 0.0);
-  B->randomize();
-
-  //
-  // Set up Krylov solver and iterate.
-  //
- 
-  Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
-  Problem_muelu->setRightPrec(M);
-  Problem_muelu->setProblem();
-
-  solver->setProblem(Problem_muelu);
-  solver->solve();
-  int numIterations_muelu = solver->getNumIters();
-
-  Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
-  multivector_type residual_muelu(B->getMap(),1);
-  A->apply(*X_muelu, residual_muelu);
-  residual_muelu.update(1.0, *B, -1.0);
-  residual_muelu.norm2(normVec_muelu);
-  if (mypid == 0) {
-    std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
-    std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
- }
-}
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    X_muelu->putScalar((scalar_type)0.0);
+    B->randomize();
+
+    //
+    // Set up Krylov solver and iterate.
+    //
+
+    Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
+    Problem_muelu->setRightPrec(M);
+    Problem_muelu->setProblem();
+
+    solver->setProblem(Problem_muelu);
+    solver->solve();
+    int numIterations_muelu = solver->getNumIters();
+
+    Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
+    multivector_type residual_muelu(B->getMap(), 1);
+    A->apply(*X_muelu, residual_muelu);
+    residual_muelu.update(1.0, *B, -1.0);
+    residual_muelu.norm2(normVec_muelu);
+    if (mypid == 0) {
+      std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
+      std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
+    }
+  }
 
-  Teuchos::TimeMonitor::summarize ();
+  Teuchos::TimeMonitor::summarize();
 
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/Test.cpp b/packages/muelu/research/max/AdditiveMG/Test.cpp
index ae74f6890fd7..4bbd26304e8b 100644
--- a/packages/muelu/research/max/AdditiveMG/Test.cpp
+++ b/packages/muelu/research/max/AdditiveMG/Test.cpp
@@ -5,13 +5,11 @@
 #include <Teuchos_DefaultComm.hpp>
 #include <Teuchos_StandardCatchMacros.hpp>
 
-
-//Galeri
+// Galeri
 #include "Galeri_Maps.h"
 #include "Galeri_CrsMatrices.h"
 #include "Galeri_Utils.h"
 
-
 // MueLu
 #include "MueLu.hpp"
 #include <MueLu_Level.hpp>
@@ -26,8 +24,6 @@
 #include <BelosXpetraAdapter.hpp>
 #include <BelosMueLuAdapter.hpp>
 
-
-
 #ifdef HAVE_MPI
 #include "Epetra_MpiComm.h"
 #include "mpi.h"
@@ -46,8 +42,7 @@ using namespace Galeri;
 // main driver //
 // =========== //
 
-int main(int argc, char* argv[])
-{
+int main(int argc, char* argv[]) {
 #ifdef HAVE_MPI
   MPI_Init(&argc, &argv);
   Epetra_MpiComm Comm(MPI_COMM_WORLD);
@@ -60,9 +55,9 @@ int main(int argc, char* argv[])
   //   Matrix * LHS = RHS
   //
   // with Matrix arising from a 5-point formula discretization.
-  
-  Epetra_Map*         Map = 0;
-  Epetra_RowMatrix*   Matrix = 0;
+
+  Epetra_Map* Map          = 0;
+  Epetra_RowMatrix* Matrix = 0;
 
   Teuchos::ParameterList GaleriList;
   // dimension of the problem is nx x ny
@@ -72,12 +67,13 @@ int main(int argc, char* argv[])
   GaleriList.set("mx", Comm.NumProc());
   GaleriList.set("my", 1);
 
-  try
-  {
-    Map = CreateMap("Cartesian2D", Comm, GaleriList);
+  try {
+    Map    = CreateMap("Cartesian2D", Comm, GaleriList);
     Matrix = CreateCrsMatrix("Laplace2D", Map, GaleriList);
-    Epetra_Vector ExactSolution(*Map); ExactSolution.Random();
-    Epetra_Vector LHS(*Map); LHS.PutScalar(0.0);
+    Epetra_Vector ExactSolution(*Map);
+    ExactSolution.Random();
+    Epetra_Vector LHS(*Map);
+    LHS.PutScalar(0.0);
     Epetra_Vector RHS(*Map);
 
     Matrix->Multiply(false, ExactSolution, RHS);
@@ -88,7 +84,7 @@ int main(int argc, char* argv[])
     // used, for example AztecOO, Amesos. IFPACK and ML can be used to define a
     // preconditioner for Matrix. Here we use a simple solver, based on
     // LAPACK, that is meant for simple testing only.
-    
+
     Solve(Problem);
 
     // and we compute the norm of the true residual.
@@ -99,9 +95,7 @@ int main(int argc, char* argv[])
 
     delete Map;
     delete Matrix;
-  }
-  catch (Galeri::Exception& rhs)
-  {
+  } catch (Galeri::Exception& rhs) {
     if (Comm.MyPID() == 0)
       rhs.Print();
     exit(EXIT_FAILURE);
@@ -111,5 +105,5 @@ int main(int argc, char* argv[])
   MPI_Finalize();
 #endif
 
-  return(EXIT_SUCCESS);
+  return (EXIT_SUCCESS);
 }
diff --git a/packages/muelu/research/max/AdditiveMG/coloring.hpp b/packages/muelu/research/max/AdditiveMG/coloring.hpp
index 831e57754f14..2c8246d3c7c3 100644
--- a/packages/muelu/research/max/AdditiveMG/coloring.hpp
+++ b/packages/muelu/research/max/AdditiveMG/coloring.hpp
@@ -1,93 +1,87 @@
 #include <math.h>
 
+int coloring2D(int mypid, int ndx) {
+  // INPUT: mypid is the MPI id of the processor
+  // INPUT: ndx is the nuber of domains along x-direction
+  // INPUT: ndy is the nuber of domains along y-direction
+  // OUTPUT: color label associated with the current MPI processor (=subdomain)
 
-int coloring2D(int mypid, int ndx)
-{
+  // detect the row of the domain grid where the current subdomain is located
+  int grid_row = std::ceil(static_cast<double>(mypid) / ndx);
 
-  //INPUT: mypid is the MPI id of the processor
-  //INPUT: ndx is the nuber of domains along x-direction
-  //INPUT: ndy is the nuber of domains along y-direction
-  //OUTPUT: color label associated with the current MPI processor (=subdomain)
-
-  //detect the row of the domain grid where the current subdomain is located
-  int grid_row  = std::ceil(static_cast<double>(mypid)/ndx);
-   
-  //detect the y-coordinate in the domain grid associated with the current subdomain
+  // detect the y-coordinate in the domain grid associated with the current subdomain
   int ypos = grid_row % 3;
   int xpos;
   int color = -1;
 
-  //detect the x-coordinate in the domain grid associated with current subdomain
-  if( 0 != mypid%ndx )
-	xpos = static_cast<int>( mypid - std::floor(static_cast<double>(mypid)/ndx)*ndx )%3;
+  // detect the x-coordinate in the domain grid associated with current subdomain
+  if (0 != mypid % ndx)
+    xpos = static_cast<int>(mypid - std::floor(static_cast<double>(mypid) / ndx) * ndx) % 3;
   else
-	xpos = (mypid - ((ypos - 1)*ndx))%3;
-
-  //use x and y coordinates to determine the color of the current subdomain
-  if ( xpos>0 && ypos>0 )
-	color = ( ypos-1 ) * 3 + xpos;
-  else if ( xpos>0 && ypos==0 )
-	color = 6 + xpos;
-  else if ( xpos==0 && ypos>0 )
-	color = ypos * 3;
+    xpos = (mypid - ((ypos - 1) * ndx)) % 3;
+
+  // use x and y coordinates to determine the color of the current subdomain
+  if (xpos > 0 && ypos > 0)
+    color = (ypos - 1) * 3 + xpos;
+  else if (xpos > 0 && ypos == 0)
+    color = 6 + xpos;
+  else if (xpos == 0 && ypos > 0)
+    color = ypos * 3;
   else
-	color = 9;
+    color = 9;
 
-  //TEUCHOS_TEST_FOR_EXCEPT( color < 1 );
-  return color-1;
+  // TEUCHOS_TEST_FOR_EXCEPT( color < 1 );
+  return color - 1;
 }
 
+int coloring3D(int mypid, int ndx, int ndy) {
+  // INPUT: mypid is the MPI id of the processor
+  // INPUT: ndx is the nuber of domains along x-direction
+  // INPUT: ndy is the nuber of domains along y-direction
+  // OUTPUT: color label associated with the current MPI processor (=subdomain)
 
-int coloring3D(int mypid, int ndx, int ndy)
-{
-
-  //INPUT: mypid is the MPI id of the processor
-  //INPUT: ndx is the nuber of domains along x-direction
-  //INPUT: ndy is the nuber of domains along y-direction
-  //OUTPUT: color label associated with the current MPI processor (=subdomain)
- 
-  //detect the plane of the domain grid where the current subdomain resides
-  int grid_plane = std::ceil( static_cast<double>(mypid)/(ndx*ndy) );
+  // detect the plane of the domain grid where the current subdomain resides
+  int grid_plane = std::ceil(static_cast<double>(mypid) / (ndx * ndy));
 
-  //On the given plane, find the local id of the current subdomain
-  int plane_id = mypid % ( ndx*ndy );
+  // On the given plane, find the local id of the current subdomain
+  int plane_id = mypid % (ndx * ndy);
 
-  //detect the row on the current two-dimensional grid where the current subdomain is located
-  int plane_row  = std::ceil(static_cast<double>(plane_id)/ndx);
+  // detect the row on the current two-dimensional grid where the current subdomain is located
+  int plane_row = std::ceil(static_cast<double>(plane_id) / ndx);
 
-  //detect the y-coordinate on the given plane associated with the current subdomain
+  // detect the y-coordinate on the given plane associated with the current subdomain
   int ypos = plane_row % 3;
   int xpos;
   int label_plane;
   int color = -1;
 
-  //detect the x-coordinate on the given plane associated with the current subdomain
-  if( 0 != plane_id%ndx )
-	xpos = static_cast<int>( mypid - std::floor(static_cast<double>(plane_id)/ndx)*ndx ) % 3;
+  // detect the x-coordinate on the given plane associated with the current subdomain
+  if (0 != plane_id % ndx)
+    xpos = static_cast<int>(mypid - std::floor(static_cast<double>(plane_id) / ndx) * ndx) % 3;
   else
-	xpos = (plane_id - ((ypos - 1)*ndx))%3;
-
-  //find a two-dimensional coloring to give to the current subdomain according to its position
-  //in the plane where it resides
-  if ( xpos>0 && ypos>0 )
-	label_plane = ( ypos-1 ) * 3 + xpos;
-  else if ( xpos>0 && ypos==0 )
-	label_plane = 6 + xpos;
-  else if ( xpos==0 && ypos>0 )
-	label_plane = ypos * 3;
+    xpos = (plane_id - ((ypos - 1) * ndx)) % 3;
+
+  // find a two-dimensional coloring to give to the current subdomain according to its position
+  // in the plane where it resides
+  if (xpos > 0 && ypos > 0)
+    label_plane = (ypos - 1) * 3 + xpos;
+  else if (xpos > 0 && ypos == 0)
+    label_plane = 6 + xpos;
+  else if (xpos == 0 && ypos > 0)
+    label_plane = ypos * 3;
   else
-	label_plane = 9;
+    label_plane = 9;
 
-  //use the two-dimensional coloring to determine the three-dimensional coloring
-  if ( grid_plane%3 ==1 )
+  // use the two-dimensional coloring to determine the three-dimensional coloring
+  if (grid_plane % 3 == 1)
     color = label_plane;
 
-  else if ( grid_plane%3 == 2 )
+  else if (grid_plane % 3 == 2)
     color = 9 + label_plane;
 
-  else if ( grid_plane%3 == 0 )
+  else if (grid_plane % 3 == 0)
     color = 18 + label_plane;
 
-  //TEUCHOS_TEST_FOR_EXCEPT( color < 1 );
-  return color-1;
+  // TEUCHOS_TEST_FOR_EXCEPT( color < 1 );
+  return color - 1;
 }
diff --git a/packages/muelu/research/max/AdditiveMG/neighbours.hpp b/packages/muelu/research/max/AdditiveMG/neighbours.hpp
index 21b0180d34e6..154351934f85 100644
--- a/packages/muelu/research/max/AdditiveMG/neighbours.hpp
+++ b/packages/muelu/research/max/AdditiveMG/neighbours.hpp
@@ -3,195 +3,172 @@
 
 #include "coloring.hpp"
 
-
-typedef Tpetra::CrsMatrix<double,int,int,KokkosClassic::DefaultNode::DefaultNodeType>     tpetra_matrix_type;
-typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,int,int,KokkosClassic::DefaultNode::DefaultNodeType>   tpetra_multivector_type;
-typedef typename Teuchos::ArrayView<const int>::const_iterator                            iterator_type;
+typedef Tpetra::CrsMatrix<double, int, int, KokkosClassic::DefaultNode::DefaultNodeType> tpetra_matrix_type;
+typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, int, int, KokkosClassic::DefaultNode::DefaultNodeType> tpetra_multivector_type;
+typedef typename Teuchos::ArrayView<const int>::const_iterator iterator_type;
 typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type2;
 
-void neighbours1D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP< const Teuchos::Comm<int> > comm)
-{
-
-  //INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
-  //INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  
- neighbours.clear();
-
- int mypid = comm->getRank();
+void neighbours1D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+  // INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
+  // INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
 
- for(int color = 0; color<3; ++color)
- {
-
-	if( (mypid-1)%3==color && (mypid-1)>= 0 && (mypid-1)<tpetra_prolong->getGlobalNumCols() )
-		neighbours.emplace_back(mypid-1);
-	else if( (mypid-2)%3==color && (mypid-2)>= 0 && (mypid-2)<tpetra_prolong->getGlobalNumCols() )
-		neighbours.emplace_back(mypid-2);
-	else if( (mypid)%3==color && (mypid)>= 0 && (mypid)<tpetra_prolong->getGlobalNumCols() )
-		neighbours.emplace_back(mypid);
+  neighbours.clear();
 
- }
+  int mypid = comm->getRank();
 
- neighbours.shrink_to_fit();
+  for (int color = 0; color < 3; ++color) {
+    if ((mypid - 1) % 3 == color && (mypid - 1) >= 0 && (mypid - 1) < tpetra_prolong->getGlobalNumCols())
+      neighbours.emplace_back(mypid - 1);
+    else if ((mypid - 2) % 3 == color && (mypid - 2) >= 0 && (mypid - 2) < tpetra_prolong->getGlobalNumCols())
+      neighbours.emplace_back(mypid - 2);
+    else if ((mypid) % 3 == color && (mypid) >= 0 && (mypid) < tpetra_prolong->getGlobalNumCols())
+      neighbours.emplace_back(mypid);
+  }
 
+  neighbours.shrink_to_fit();
 }
 
+void neighbours2D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP<const Teuchos::Comm<int> > comm, int ndx) {
+  // INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
+  // INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
+  // INPUT: ndx is the number of domains along the x-direction
 
-void neighbours2D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP< const Teuchos::Comm<int> > comm, int ndx)
-{
-
-  //INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
-  //INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  //INPUT: ndx is the number of domains along the x-direction
-  
- neighbours.clear();
- int mypid = comm->getRank();
- int brick_id = mypid;
- int shifted_id = brick_id-1;
-
- if(mypid>0)
- {
-	for(int color = 0; color<9; ++color)
-	{
-                int neighbour = -1;
-
-                //The following if statements control the neighbours of a subdomain in a 2D brick partitioned mesh
-                //Each subdomains is incorporated in a 3x3 square which is sliced into 3 stripes
-                //The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
-                // In case the subdomains sit on a boundary, there are missing neighbours for a specific color
-                // (this is what the last two conditions of each if statement take care of)
-		if( coloring2D( brick_id, ndx )==color && (shifted_id)>= 0 && (shifted_id)<tpetra_prolong->getGlobalNumCols() )
-		        neighbour = (shifted_id);
+  neighbours.clear();
+  int mypid      = comm->getRank();
+  int brick_id   = mypid;
+  int shifted_id = brick_id - 1;
 
-		else if( coloring2D( brick_id-1, ndx )==color && (shifted_id-1)>= 0 && (shifted_id-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1);
+  if (mypid > 0) {
+    for (int color = 0; color < 9; ++color) {
+      int neighbour = -1;
 
-		else if( coloring2D( brick_id+1, ndx )==color && (shifted_id+1)>= 0 && (shifted_id+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1);
+      // The following if statements control the neighbours of a subdomain in a 2D brick partitioned mesh
+      // Each subdomains is incorporated in a 3x3 square which is sliced into 3 stripes
+      // The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
+      //  In case the subdomains sit on a boundary, there are missing neighbours for a specific color
+      //  (this is what the last two conditions of each if statement take care of)
+      if (coloring2D(brick_id, ndx) == color && (shifted_id) >= 0 && (shifted_id) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id);
 
-		else if( coloring2D( brick_id-ndx, ndx )==color && (shifted_id-ndx)>= 0 && (shifted_id-ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx);
+      else if (coloring2D(brick_id - 1, ndx) == color && (shifted_id - 1) >= 0 && (shifted_id - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1);
 
-		else if( coloring2D( brick_id-ndx-1, ndx )==color && (shifted_id-ndx-1)>= 0 && (shifted_id-ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1);
+      else if (coloring2D(brick_id + 1, ndx) == color && (shifted_id + 1) >= 0 && (shifted_id + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1);
 
-		else if( coloring2D( brick_id-ndx+1, ndx )==color && (shifted_id-ndx+1)>= 0 && (shifted_id-ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1);
+      else if (coloring2D(brick_id - ndx, ndx) == color && (shifted_id - ndx) >= 0 && (shifted_id - ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx);
 
-		else if( coloring2D( brick_id+ndx, ndx )==color && (shifted_id+ndx)>= 0 && (shifted_id+ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx);
+      else if (coloring2D(brick_id - ndx - 1, ndx) == color && (shifted_id - ndx - 1) >= 0 && (shifted_id - ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1);
 
-		else if( coloring2D( brick_id+ndx-1, ndx )==color && (shifted_id+ndx-1)>= 0 && (shifted_id+ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1);
+      else if (coloring2D(brick_id - ndx + 1, ndx) == color && (shifted_id - ndx + 1) >= 0 && (shifted_id - ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1);
 
-		else if( coloring2D( brick_id+ndx+1, ndx )==color && (shifted_id+ndx+1)>= 0 && (shifted_id+ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1);
+      else if (coloring2D(brick_id + ndx, ndx) == color && (shifted_id + ndx) >= 0 && (shifted_id + ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx);
 
-                if(neighbour>=0)
-                   neighbours.emplace_back(neighbour);
+      else if (coloring2D(brick_id + ndx - 1, ndx) == color && (shifted_id + ndx - 1) >= 0 && (shifted_id + ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1);
 
-	}
- }
+      else if (coloring2D(brick_id + ndx + 1, ndx) == color && (shifted_id + ndx + 1) >= 0 && (shifted_id + ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1);
 
- neighbours.shrink_to_fit();
+      if (neighbour >= 0)
+        neighbours.emplace_back(neighbour);
+    }
+  }
 
+  neighbours.shrink_to_fit();
 }
 
-
-void neighbours3D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP< const Teuchos::Comm<int> > comm, int ndx, int ndy)
-{
-
-  //INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
-  //INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
-  //INPUT: comm = MPI communicator (MPI_COMM_WORLD)
-  //INPUT: ndx is the number of domains along the x-direction
-  //INPUT: ndy is the nubmer of domains along the y-direction
+void neighbours3D(Teuchos::RCP<tpetra_matrix_type> tpetra_prolong, std::vector<int>& neighbours, Teuchos::RCP<const Teuchos::Comm<int> > comm, int ndx, int ndy) {
+  // INPUT: tpetra_prolong is the tentative prolongator, here it is just used to get the Total number of aggregates given by the nubmer of columns
+  // INPUT: neighoubrs is a vector where the neoghbours of the current MPI processor (=subdomain) are stored
+  // INPUT: comm = MPI communicator (MPI_COMM_WORLD)
+  // INPUT: ndx is the number of domains along the x-direction
+  // INPUT: ndy is the nubmer of domains along the y-direction
 
   neighbours.clear();
-  int mypid = comm->getRank();
-  int brick_id = mypid;
+  int mypid      = comm->getRank();
+  int brick_id   = mypid;
   int shifted_id = brick_id - 1;
 
-  if(mypid>0)
-  {
-	for(int color = 0; color<27; ++color)
-	{
-		int neighbour = -1;
-
-                //The following if statements control the neighbours of a subdomain in a 3D brick partitioned mesh
-                //Each subdomains is incorporated in a 3x3x3 cube which is sliced into 3 squares living on three different planes
-                //The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
-                // In case the subdomains sit on a boundary, there are missing neighbours for a specific color
-                // (this is what the last two conditions of each if statement take care of)
-                //Identification of neighbours that are on the current plane
-		if( coloring3D( brick_id, ndx, ndy )==color && (shifted_id)>= 0 && (shifted_id)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id);
-		else if( coloring3D( brick_id-1, ndx, ndy )==color && (shifted_id-1)>= 0 && (shifted_id-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1);
-		else if( coloring3D( brick_id+1, ndx, ndy )==color && (shifted_id+1)>= 0 && (shifted_id+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1);
-		else if( coloring3D( brick_id-ndx, ndx, ndy )==color && (shifted_id-ndx)>= 0 && (shifted_id-ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx);
-		else if( coloring3D( brick_id-ndx-1, ndx, ndy )==color && (shifted_id-ndx-1)>= 0 && (shifted_id-ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1);
-		else if( coloring3D( brick_id-ndx+1, ndx, ndy )==color && (shifted_id-ndx+1)>= 0 && (shifted_id-ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1);
-		else if( coloring3D( brick_id+ndx, ndx, ndy )==color && (shifted_id+ndx)>= 0 && (shifted_id+ndx)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx);
-		else if( coloring3D( brick_id+ndx-1, ndx, ndy )==color && (shifted_id+ndx-1)>= 0 && (shifted_id+ndx-1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1);
-		else if( coloring3D( brick_id+ndx+1, ndx, ndy )==color && (shifted_id+ndx+1)>= 0 && (shifted_id+ndx+1)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1);
-
-                //Identification of the neighbours that are on the plane below
-		else if( coloring3D( brick_id-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx*ndy)>= 0 && (shifted_id-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx*ndy);
-		else if( coloring3D( brick_id-1-ndx*ndy, ndx, ndy )==color && (shifted_id-1-ndx*ndy)>= 0 && (shifted_id-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1-ndx*ndy);
-		else if( coloring3D( brick_id+1-ndx*ndy, ndx, ndy )==color && (shifted_id+1-ndx*ndy)>= 0 && (shifted_id+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1-ndx*ndy);
-		else if( coloring3D( brick_id-ndx-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-ndx*ndy)>= 0 && (shifted_id-ndx-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-ndx*ndy);
-		else if( coloring3D( brick_id-ndx-1-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-1-ndx*ndy)>= 0 && (shifted_id-ndx-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1-ndx*ndy);
-		else if( coloring3D( brick_id-ndx+1-ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+1-ndx*ndy)>= 0 && (shifted_id-ndx+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1-ndx*ndy);
-		else if( coloring3D( brick_id+ndx-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-ndx*ndy)>= 0 && (shifted_id+ndx-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-ndx*ndy);
-		else if( coloring3D( brick_id+ndx-1-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-1-ndx*ndy)>= 0 && (shifted_id+ndx-1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1-ndx*ndy);
-		else if( coloring3D( brick_id+ndx+1-ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+1-ndx*ndy)>= 0 && (shifted_id+ndx+1-ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1-ndx*ndy);
-
-                //Identification of the neighbours that are on the plane above
-		else if( coloring3D( brick_id+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx*ndy)>= 0 && (shifted_id+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx*ndy);
-		else if( coloring3D( brick_id-1+ndx*ndy, ndx, ndy )==color && (shifted_id-1+ndx*ndy)>= 0 && (shifted_id-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-1+ndx*ndy);
-		else if( coloring3D( brick_id+1+ndx*ndy, ndx, ndy )==color && (shifted_id+1+ndx*ndy)>= 0 && (shifted_id+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+1+ndx*ndy);
-		else if( coloring3D( brick_id-ndx+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+ndx*ndy)>= 0 && (shifted_id-ndx+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+ndx*ndy);
-		else if( coloring3D( brick_id-ndx-1+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx-1+ndx*ndy)>= 0 && (shifted_id-ndx-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx-1+ndx*ndy);
-		else if( coloring3D( brick_id-ndx+1+ndx*ndy, ndx, ndy )==color && (shifted_id-ndx+1+ndx*ndy)>= 0 && (shifted_id-ndx+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id-ndx+1+ndx*ndy);
-		else if( coloring3D( brick_id+ndx+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+ndx*ndy)>= 0 && (shifted_id+ndx+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+ndx*ndy);
-		else if( coloring3D( brick_id+ndx-1+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx-1+ndx*ndy)>= 0 && (shifted_id+ndx-1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx-1+ndx*ndy);
-		else if( coloring3D( brick_id+ndx+1+ndx*ndy, ndx, ndy )==color && (shifted_id+ndx+1+ndx*ndy)>= 0 && (shifted_id+ndx+1+ndx*ndy)<tpetra_prolong->getGlobalNumCols() )
-			neighbour = (shifted_id+ndx+1+ndx*ndy);
-
-                if(neighbour>=0)
-                  neighbours.emplace_back(neighbour);
-
-	}
-   }
-
- neighbours.shrink_to_fit();
-
+  if (mypid > 0) {
+    for (int color = 0; color < 27; ++color) {
+      int neighbour = -1;
+
+      // The following if statements control the neighbours of a subdomain in a 3D brick partitioned mesh
+      // Each subdomains is incorporated in a 3x3x3 cube which is sliced into 3 squares living on three different planes
+      // The neighbours of a subdomain are checked plane by plane: in total there are three planes to span
+      //  In case the subdomains sit on a boundary, there are missing neighbours for a specific color
+      //  (this is what the last two conditions of each if statement take care of)
+      // Identification of neighbours that are on the current plane
+      if (coloring3D(brick_id, ndx, ndy) == color && (shifted_id) >= 0 && (shifted_id) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id);
+      else if (coloring3D(brick_id - 1, ndx, ndy) == color && (shifted_id - 1) >= 0 && (shifted_id - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1);
+      else if (coloring3D(brick_id + 1, ndx, ndy) == color && (shifted_id + 1) >= 0 && (shifted_id + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1);
+      else if (coloring3D(brick_id - ndx, ndx, ndy) == color && (shifted_id - ndx) >= 0 && (shifted_id - ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx);
+      else if (coloring3D(brick_id - ndx - 1, ndx, ndy) == color && (shifted_id - ndx - 1) >= 0 && (shifted_id - ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1);
+      else if (coloring3D(brick_id - ndx + 1, ndx, ndy) == color && (shifted_id - ndx + 1) >= 0 && (shifted_id - ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1);
+      else if (coloring3D(brick_id + ndx, ndx, ndy) == color && (shifted_id + ndx) >= 0 && (shifted_id + ndx) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx);
+      else if (coloring3D(brick_id + ndx - 1, ndx, ndy) == color && (shifted_id + ndx - 1) >= 0 && (shifted_id + ndx - 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1);
+      else if (coloring3D(brick_id + ndx + 1, ndx, ndy) == color && (shifted_id + ndx + 1) >= 0 && (shifted_id + ndx + 1) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1);
+
+      // Identification of the neighbours that are on the plane below
+      else if (coloring3D(brick_id - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx * ndy) >= 0 && (shifted_id - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx * ndy);
+      else if (coloring3D(brick_id - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - 1 - ndx * ndy) >= 0 && (shifted_id - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1 - ndx * ndy);
+      else if (coloring3D(brick_id + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + 1 - ndx * ndy) >= 0 && (shifted_id + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1 - ndx * ndy);
+      else if (coloring3D(brick_id - ndx - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - ndx * ndy) >= 0 && (shifted_id - ndx - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - ndx * ndy);
+      else if (coloring3D(brick_id - ndx - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - 1 - ndx * ndy) >= 0 && (shifted_id - ndx - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1 - ndx * ndy);
+      else if (coloring3D(brick_id - ndx + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + 1 - ndx * ndy) >= 0 && (shifted_id - ndx + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1 - ndx * ndy);
+      else if (coloring3D(brick_id + ndx - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - ndx * ndy) >= 0 && (shifted_id + ndx - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - ndx * ndy);
+      else if (coloring3D(brick_id + ndx - 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - 1 - ndx * ndy) >= 0 && (shifted_id + ndx - 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1 - ndx * ndy);
+      else if (coloring3D(brick_id + ndx + 1 - ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + 1 - ndx * ndy) >= 0 && (shifted_id + ndx + 1 - ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1 - ndx * ndy);
+
+      // Identification of the neighbours that are on the plane above
+      else if (coloring3D(brick_id + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx * ndy) >= 0 && (shifted_id + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx * ndy);
+      else if (coloring3D(brick_id - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - 1 + ndx * ndy) >= 0 && (shifted_id - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - 1 + ndx * ndy);
+      else if (coloring3D(brick_id + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + 1 + ndx * ndy) >= 0 && (shifted_id + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + 1 + ndx * ndy);
+      else if (coloring3D(brick_id - ndx + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + ndx * ndy) >= 0 && (shifted_id - ndx + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + ndx * ndy);
+      else if (coloring3D(brick_id - ndx - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx - 1 + ndx * ndy) >= 0 && (shifted_id - ndx - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx - 1 + ndx * ndy);
+      else if (coloring3D(brick_id - ndx + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id - ndx + 1 + ndx * ndy) >= 0 && (shifted_id - ndx + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id - ndx + 1 + ndx * ndy);
+      else if (coloring3D(brick_id + ndx + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + ndx * ndy) >= 0 && (shifted_id + ndx + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + ndx * ndy);
+      else if (coloring3D(brick_id + ndx - 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx - 1 + ndx * ndy) >= 0 && (shifted_id + ndx - 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx - 1 + ndx * ndy);
+      else if (coloring3D(brick_id + ndx + 1 + ndx * ndy, ndx, ndy) == color && (shifted_id + ndx + 1 + ndx * ndy) >= 0 && (shifted_id + ndx + 1 + ndx * ndy) < tpetra_prolong->getGlobalNumCols())
+        neighbour = (shifted_id + ndx + 1 + ndx * ndy);
+
+      if (neighbour >= 0)
+        neighbours.emplace_back(neighbour);
+    }
+  }
+
+  neighbours.shrink_to_fit();
 }
-
-
diff --git a/packages/muelu/research/max/AdditiveMG/tentative.cpp b/packages/muelu/research/max/AdditiveMG/tentative.cpp
index b7834756e02a..a5e759379a97 100644
--- a/packages/muelu/research/max/AdditiveMG/tentative.cpp
+++ b/packages/muelu/research/max/AdditiveMG/tentative.cpp
@@ -77,7 +77,7 @@ extern clock_t Timers_Max[4];
 #include <Galeri_XpetraUtils.hpp>
 #include <Galeri_XpetraMaps.hpp>
 
-//ADR subdirectory
+// ADR subdirectory
 #include "CreateADRMatrix.hpp"
 #include <CreateADRMatrix.hpp>
 #include <ADRProblemFactory.hpp>
@@ -88,24 +88,23 @@ extern clock_t Timers_Max[4];
 #include "BAP.hpp"
 #include "CreateBrickMap.hpp"
 
-int main(int argc, char *argv[]) {
-
+int main(int argc, char* argv[]) {
   // Define default types
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType node_type;
 
   // Convenient typedef's
-  typedef Tpetra::Operator<scalar_type,local_ordinal_type,global_ordinal_type,node_type>    operator_type;
-  typedef Tpetra::CrsMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   crs_matrix_type;
-	typedef Tpetra::RowMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>   row_matrix_type;
-  typedef Tpetra::Vector<scalar_type,local_ordinal_type,global_ordinal_type,node_type>      vector_type;
-  typedef Tpetra::MultiVector<scalar_type,local_ordinal_type,global_ordinal_type,node_type> multivector_type;
-  typedef Tpetra::Map<local_ordinal_type,global_ordinal_type,node_type>                     driver_map_type;
+  typedef Tpetra::Operator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> operator_type;
+  typedef Tpetra::CrsMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> crs_matrix_type;
+  typedef Tpetra::RowMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type> row_matrix_type;
+  typedef Tpetra::Vector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> vector_type;
+  typedef Tpetra::MultiVector<scalar_type, local_ordinal_type, global_ordinal_type, node_type> multivector_type;
+  typedef Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type> driver_map_type;
 
   typedef MueLu::TpetraOperator<scalar_type, local_ordinal_type, global_ordinal_type, node_type> muelu_tpetra_operator_type;
-  typedef MueLu::Utilities<scalar_type,local_ordinal_type,global_ordinal_type,node_type> MueLuUtilities;
+  typedef MueLu::Utilities<scalar_type, local_ordinal_type, global_ordinal_type, node_type> MueLuUtilities;
 
   typedef Belos::LinearProblem<scalar_type, multivector_type, operator_type> linear_problem_type;
   typedef Belos::SolverManager<scalar_type, multivector_type, operator_type> belos_solver_manager_type;
@@ -113,50 +112,50 @@ int main(int argc, char *argv[]) {
   typedef Belos::BlockGmresSolMgr<scalar_type, multivector_type, operator_type> belos_gmres_manager_type;
   typedef Belos::BiCGStabSolMgr<scalar_type, multivector_type, operator_type> belos_bicgstab_manager_type;
 
-	typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
+  typedef Ifpack2::Preconditioner<scalar_type, local_ordinal_type, global_ordinal_type, node_type> precond_type;
 
-  //MueLu_UseShortNames.hpp wants these typedefs.
-  typedef scalar_type         Scalar;
-  typedef local_ordinal_type  LocalOrdinal;
+  // MueLu_UseShortNames.hpp wants these typedefs.
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
   typedef global_ordinal_type GlobalOrdinal;
-  typedef node_type           Node;
-# include <MueLu_UseShortNames.hpp>
+  typedef node_type Node;
+#include <MueLu_UseShortNames.hpp>
 
-  typedef Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> GaleriXpetraProblem;
-  typedef ADR::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> ADRXpetraProblem;
+  typedef Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> GaleriXpetraProblem;
+  typedef ADR::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> ADRXpetraProblem;
 
-  using Teuchos::RCP; // reference count pointers
-  using Teuchos::rcp; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
+  using Teuchos::rcp;  // reference count pointers
 
   //
   // MPI initialization using Teuchos
   //
 
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int mypid = comm->getRank();
-/*
-  int subCommRank[3]={0,1,2};
-  Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
-  auto subComm = comm->createSubcommunicator(arraySubCommRank);
-*/
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+  int mypid                          = comm->getRank();
+  /*
+    int subCommRank[3]={0,1,2};
+    Teuchos::ArrayView<int> arraySubCommRank(subCommRank, 3);
+    auto subComm = comm->createSubcommunicator(arraySubCommRank);
+  */
   Teuchos::CommandLineProcessor clp(false);
 
-  global_ordinal_type maxIts            = 10000;
-  scalar_type tol                       = 1e-10;
-  std::string solverOptionsFile         = "final_parser.xml";
-  std::string krylovSolverType          = "bicgstab";
+  global_ordinal_type maxIts    = 10000;
+  scalar_type tol               = 1e-10;
+  std::string solverOptionsFile = "final_parser.xml";
+  std::string krylovSolverType  = "bicgstab";
 
-  clp.setOption("xmlFile",    &solverOptionsFile, "XML file containing MueLu solver parameters");
-  clp.setOption("maxits",     &maxIts,            "maximum number of Krylov iterations");
-  clp.setOption("tol",        &tol,               "tolerance for Krylov solver");
-  clp.setOption("krylovType", &krylovSolverType,  "cg or gmres solver");
+  clp.setOption("xmlFile", &solverOptionsFile, "XML file containing MueLu solver parameters");
+  clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+  clp.setOption("tol", &tol, "tolerance for Krylov solver");
+  clp.setOption("krylovType", &krylovSolverType, "cg or gmres solver");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Teuchos::ParameterList xmlParams;
@@ -167,7 +166,7 @@ int main(int argc, char *argv[]) {
   problemParams = xmlParams.sublist(static_cast<const std::string>("Problem"));
 
   // Problem definition
-   std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
+  std::string problem_type = problemParams.get<std::string>(static_cast<const std::string>("problem type"));
 
   // Parameters
 
@@ -182,14 +181,13 @@ int main(int argc, char *argv[]) {
   MueLu::DomainPartitioning domain;
 
   int keep_boundary = 0;
-  Scalar stretchx = (Scalar) Lx/nx;
-  Scalar stretchy = (Scalar) Ly/ny;
-  Scalar stretchz = (Scalar) Lz/nz;
+  Scalar stretchx   = (Scalar)Lx / nx;
+  Scalar stretchy   = (Scalar)Ly / ny;
+  Scalar stretchz   = (Scalar)Lz / nz;
 
+  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary, stretchx, stretchy, stretchz);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                                                                  // manage parameters of xpetra
 
-  ADR::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, problem_type, keep_boundary , stretchx, stretchy, stretchz); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);     // manage parameters of xpetra
- 
   if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {
     throw std::invalid_argument("This example only supports Tpetra.");
   }
@@ -202,226 +200,206 @@ int main(int argc, char *argv[]) {
   RCP<const Map> xpetraMap;
   std::vector<global_ordinal_type> ind;
 
-  //BRICK SIZE
+  // BRICK SIZE
   int brick_sizex = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick x size"));
   int brick_sizey = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick y size"));
   int brick_sizez = mueluParams.get<int>(static_cast<const std::string>("aggregation: brick z size"));
 
-  //Creation of the map where processor 0 gets nothing at the fine level
-  if( comm->getSize()>1 )
-  {
-	if(problem_type == "ADR1D")
-		createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm );
+  // Creation of the map where processor 0 gets nothing at the fine level
+  if (comm->getSize() > 1) {
+    if (problem_type == "ADR1D")
+      createBrickMap1D(matrixParameters.GetNumGlobalElements(), ind, comm);
 
-	else if(problem_type == "ADR2D")
-		createBrickMap2D( nx, brick_sizex, brick_sizey, ind, comm );
+    else if (problem_type == "ADR2D")
+      createBrickMap2D(nx, brick_sizex, brick_sizey, ind, comm);
 
-	else if(problem_type == "ADR3D")
-		createBrickMap3D( nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm );
+    else if (problem_type == "ADR3D")
+      createBrickMap3D(nx, ny, brick_sizex, brick_sizey, brick_sizez, ind, comm);
 
-	ind.shrink_to_fit();
+    ind.shrink_to_fit();
 
-	Teuchos::ArrayView<const global_ordinal_type> elementList (ind);
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra,matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
-  }
-  else if( comm->getSize()==1 )
-  {
-	xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
+    Teuchos::ArrayView<const global_ordinal_type> elementList(ind);
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), elementList, indexBase, comm);
+  } else if (comm->getSize() == 1) {
+    xpetraMap = MapFactory::Build(Xpetra::UseTpetra, matrixParameters.GetNumGlobalElements(), indexBase, comm);
   }
 
   RCP<MultiVector> coordinates;
 
-	if (problem_type == "ADR1D")
-		coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
-	else if (problem_type == "ADR2D")
-		coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
-	else if (problem_type == "ADR3D")
-		coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type,local_ordinal_type,global_ordinal_type,Map,MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
-
- RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
-  	RCP<Matrix>  xpetraA = Pr->BuildMatrix();
-
-    RCP<crs_matrix_type> A = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
-    RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
-
- // ===================================================
- // 	Domain Decomposition Preconditioner
- // 	===================================
-
-    //Creation of the MueLu list for the DD preconditioner
-    RCP<Teuchos::ParameterList> dd_list = rcp(new Teuchos::ParameterList());
-    dd_list->setName("MueLu");
-    dd_list->set("verbosity", "low");
-    dd_list->set("number of equations", 1);
-    dd_list->set("max levels", 1);
-    dd_list->set("coarse: type", "SCHWARZ"); //FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
-		
-    Teuchos::ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
-    dd_smooth_sublist.set("schwarz: overlap level", 0);
-    dd_smooth_sublist.set("schwarz: combine mode", "Zero");
-    dd_smooth_sublist.set("subdomain solver name", "RILUK");
-
-    Teuchos::ParameterList& coarse_subdomain_solver = dd_smooth_sublist.sublist("subdomain solver parameters");
-    coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
-    coarse_subdomain_solver.set("fact: absolute thresh/old", 0.);
-    coarse_subdomain_solver.set("fact: relative threshold", 1.);
-    coarse_subdomain_solver.set("fact: relax value", 0.);
-
-    RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner( (RCP<operator_type>)A, *dd_list );
-
-    // Construct a multigrid preconditioner
-    RCP<muelu_tpetra_operator_type> M;
-
-    //Manual set up of the prolongation and restriction
-    MueLu::ParameterListInterpreter<scalar_type> mueLuFactory( mueluParams );
-    RCP<MueLu::Hierarchy<scalar_type> > H = mueLuFactory.CreateHierarchy();
-    H->setVerbLevel(Teuchos::VERB_HIGH);
-	
-    H->GetLevel(0)->Set("A", xpetraA);
-    H->GetLevel(0)->Set("Coordinates", coordinates);
-
-    // Multigrid setup phase
-    mueLuFactory.SetupHierarchy(*H);	
-
-    RCP<Level> L = H->GetLevel(1);
-
-    RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> prolong, restr;
-
-    if (L->IsAvailable("P"))
-      prolong = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("P");
-
-    if (L->IsAvailable("R"))
-      restr = L->template Get< RCP<Xpetra::Matrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>> >("R");
-
-
-    RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
-    RCP<crs_matrix_type> tpetra_restr = MueLuUtilities::Op2NonConstTpetraCrs(restr);
-
-    #include <Teuchos_TimeMonitor.hpp>
-    RCP<Teuchos::Time> PbarSetUp    = Teuchos::TimeMonitor::getNewCounter("Pbar: SetUp");
-    PbarSetUp->start();
-    RCP<Matrix> mueluPbar;
- 
-    // We have to transform P into a condensed multivector
-    RCP<multivector_type> identity_shrunk = rcp( new multivector_type(tpetra_prolong->getDomainMap(), std::pow(3, coordinates->getNumVectors()) ) );
-    Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
-    typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
-
-   int number_runs = 1;
-
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-        if(1 == coordinates->getNumVectors())
-	{
-
-	  for(int j = 0; j<3; ++j)
-	  {
-	    int color = j;
-	
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-		    const local_ordinal_type i_local = *it;
-		    const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-                    int local_color = (i_local)%3;
-
-	      if( local_color == color )
-			localMV[aux] = 1.0;	
-		
-	     }
-	   }
-	}
-	else if(2 == coordinates->getNumVectors())
-	{
-
-	  for(int j = 0; j<9; ++j)
-	  {
-	    int color = j;
-			
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-		    const local_ordinal_type i_local = *it;
-		    const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-                    const local_ordinal_type local_color = coloring2D( i_local+1, std::sqrt( tpetra_prolong->getGlobalNumCols() ) );
-	
-	      if( local_color == color  )
-              {
-			    localMV[aux] = 1.0;	
-               //std::cout<<"ID: "<<i_local+1<<" -- color: "<<color+1<<std::endl;
-              }		
-	     }
-	   }
-	}
-	else if(3 == coordinates->getNumVectors())
-	{
-          //const local_ordinal_type local_color = coloring3D( mypid, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
-
-	  for(int j = 0; j<27; ++j)
-	  {
-	    int color = j;
-	
-	    Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
-
-	    for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
-              const local_ordinal_type i_local = *it;
-	      const local_ordinal_type aux = identity_shrunk->getMap()->getLocalElement (i_local);
-              const local_ordinal_type local_color = coloring3D( i_local+1, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
-
-	      if( local_color == color  )
-              {
-                            //std::cout<<"i_local: "<<i_local+1<<" has color: "<<color<<std::endl;
-                            //sleep(1);
-			    localMV[aux] = 1.0;	
-	      }	
-	    }
-	  }
-	}
-	
-
-     RCP<multivector_type> P_shrunk = rcp( new multivector_type( tpetra_prolong->getRangeMap(), std::pow(3, coordinates->getNumVectors()) ) );
-     tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
-     RCP<multivector_type> AP_shrunk = rcp( new multivector_type( A->getRangeMap(), std::pow(3, coordinates->getNumVectors()) ) );
-     A->apply(*P_shrunk, *AP_shrunk);
-
-	//========================================================================================================
-
-	// CREATION OF BAP
-
-    RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_DD->getRangeMap(),AP_shrunk->getNumVectors()));
-	  B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits< scalar_type >::one(), Teuchos::ScalarTraits< scalar_type >::zero());
-    
-   //Columns Map for BAP
-   std::vector<int> indBAPcolMap;
-    if(1 == coordinates->getNumVectors())
-      neighbours1D( tpetra_prolong, indBAPcolMap, comm );
-    else if(2 == coordinates->getNumVectors())
-      neighbours2D( tpetra_prolong, indBAPcolMap, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()) );
-    else if(3 == coordinates->getNumVectors())
-      neighbours3D( tpetra_prolong, indBAPcolMap, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()) );
-
-   Teuchos::ArrayView<const global_ordinal_type> elementListBAP (indBAPcolMap);
-   Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > BAPcolMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm ) );
-
-    RCP<crs_matrix_type> BAP = rcp( new crs_matrix_type( tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols() ) );
-    Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
+  if (problem_type == "ADR1D")
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("1D", xpetraMap, matrixParameters.GetParameterList());
+  else if (problem_type == "ADR2D")
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("2D", xpetraMap, matrixParameters.GetParameterList());
+  else if (problem_type == "ADR3D")
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<scalar_type, local_ordinal_type, global_ordinal_type, Map, MultiVector>("3D", xpetraMap, matrixParameters.GetParameterList());
+
+  RCP<ADRXpetraProblem> Pr = ADR::Xpetra::BuildProblem<scalar_type, local_ordinal_type, global_ordinal_type, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), xpetraMap, matrixParameters.GetParameterList());
+  RCP<Matrix> xpetraA      = Pr->BuildMatrix();
+
+  RCP<crs_matrix_type> A         = MueLuUtilities::Op2NonConstTpetraCrs(xpetraA);
+  RCP<const driver_map_type> map = MueLuUtilities::Map2TpetraMap(*xpetraMap);
+
+  // ===================================================
+  // 	Domain Decomposition Preconditioner
+  // 	===================================
+
+  // Creation of the MueLu list for the DD preconditioner
+  RCP<Teuchos::ParameterList> dd_list = rcp(new Teuchos::ParameterList());
+  dd_list->setName("MueLu");
+  dd_list->set("verbosity", "low");
+  dd_list->set("number of equations", 1);
+  dd_list->set("max levels", 1);
+  dd_list->set("coarse: type", "SCHWARZ");  // FOR A ONE LEVEL PRECONDITIONER THE COARSE LEVEL IS INTERPRETED AS SMOOTHING LEVEL
+
+  Teuchos::ParameterList& dd_smooth_sublist = dd_list->sublist("coarse: params");
+  dd_smooth_sublist.set("schwarz: overlap level", 0);
+  dd_smooth_sublist.set("schwarz: combine mode", "Zero");
+  dd_smooth_sublist.set("subdomain solver name", "RILUK");
+
+  Teuchos::ParameterList& coarse_subdomain_solver = dd_smooth_sublist.sublist("subdomain solver parameters");
+  coarse_subdomain_solver.set("fact: iluk level-of-fill", 3);
+  coarse_subdomain_solver.set("fact: absolute thresh/old", 0.);
+  coarse_subdomain_solver.set("fact: relative threshold", 1.);
+  coarse_subdomain_solver.set("fact: relax value", 0.);
+
+  RCP<muelu_tpetra_operator_type> B_DD = MueLu::CreateTpetraPreconditioner((RCP<operator_type>)A, *dd_list);
+
+  // Construct a multigrid preconditioner
+  RCP<muelu_tpetra_operator_type> M;
+
+  // Manual set up of the prolongation and restriction
+  MueLu::ParameterListInterpreter<scalar_type> mueLuFactory(mueluParams);
+  RCP<MueLu::Hierarchy<scalar_type>> H = mueLuFactory.CreateHierarchy();
+  H->setVerbLevel(Teuchos::VERB_HIGH);
+
+  H->GetLevel(0)->Set("A", xpetraA);
+  H->GetLevel(0)->Set("Coordinates", coordinates);
+
+  // Multigrid setup phase
+  mueLuFactory.SetupHierarchy(*H);
+
+  RCP<Level> L = H->GetLevel(1);
+
+  RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>> prolong, restr;
+
+  if (L->IsAvailable("P"))
+    prolong = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("P");
+
+  if (L->IsAvailable("R"))
+    restr = L->template Get<RCP<Xpetra::Matrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>>>("R");
+
+  RCP<crs_matrix_type> tpetra_prolong = MueLuUtilities::Op2NonConstTpetraCrs(prolong);
+  RCP<crs_matrix_type> tpetra_restr   = MueLuUtilities::Op2NonConstTpetraCrs(restr);
+
+#include <Teuchos_TimeMonitor.hpp>
+  RCP<Teuchos::Time> PbarSetUp = Teuchos::TimeMonitor::getNewCounter("Pbar: SetUp");
+  PbarSetUp->start();
+  RCP<Matrix> mueluPbar;
+
+  // We have to transform P into a condensed multivector
+  RCP<multivector_type> identity_shrunk                                  = rcp(new multivector_type(tpetra_prolong->getDomainMap(), std::pow(3, coordinates->getNumVectors())));
+  Teuchos::ArrayView<const global_ordinal_type> myIdentityGlobalElements = tpetra_prolong->getDomainMap()->getLocalElementList();
+  typedef typename Teuchos::ArrayView<const global_ordinal_type>::const_iterator iter_type;
+
+  int number_runs = 1;
+
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    if (1 == coordinates->getNumVectors()) {
+      for (int j = 0; j < 3; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local = *it;
+          const local_ordinal_type aux     = identity_shrunk->getMap()->getLocalElement(i_local);
+          int local_color                  = (i_local) % 3;
+
+          if (local_color == color)
+            localMV[aux] = 1.0;
+        }
+      }
+    } else if (2 == coordinates->getNumVectors()) {
+      for (int j = 0; j < 9; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local     = *it;
+          const local_ordinal_type aux         = identity_shrunk->getMap()->getLocalElement(i_local);
+          const local_ordinal_type local_color = coloring2D(i_local + 1, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+
+          if (local_color == color) {
+            localMV[aux] = 1.0;
+            // std::cout<<"ID: "<<i_local+1<<" -- color: "<<color+1<<std::endl;
+          }
+        }
+      }
+    } else if (3 == coordinates->getNumVectors()) {
+      // const local_ordinal_type local_color = coloring3D( mypid, std::cbrt( tpetra_prolong->getGlobalNumCols() ), std::cbrt( tpetra_prolong->getGlobalNumCols() ) );
+
+      for (int j = 0; j < 27; ++j) {
+        int color = j;
+
+        Teuchos::ArrayRCP<scalar_type> localMV = identity_shrunk->getDataNonConst(color);
+
+        for (iter_type it = myIdentityGlobalElements.begin(); it != myIdentityGlobalElements.end(); ++it) {
+          const local_ordinal_type i_local     = *it;
+          const local_ordinal_type aux         = identity_shrunk->getMap()->getLocalElement(i_local);
+          const local_ordinal_type local_color = coloring3D(i_local + 1, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+          if (local_color == color) {
+            // std::cout<<"i_local: "<<i_local+1<<" has color: "<<color<<std::endl;
+            // sleep(1);
+            localMV[aux] = 1.0;
+          }
+        }
+      }
+    }
+
+    RCP<multivector_type> P_shrunk = rcp(new multivector_type(tpetra_prolong->getRangeMap(), std::pow(3, coordinates->getNumVectors())));
+    tpetra_prolong->apply(*identity_shrunk, *P_shrunk);
+    RCP<multivector_type> AP_shrunk = rcp(new multivector_type(A->getRangeMap(), std::pow(3, coordinates->getNumVectors())));
+    A->apply(*P_shrunk, *AP_shrunk);
+
+    //========================================================================================================
+
+    // CREATION OF BAP
 
+    RCP<multivector_type> BAP_shrunk = rcp(new multivector_type(B_DD->getRangeMap(), AP_shrunk->getNumVectors()));
+    B_DD->apply(*AP_shrunk, *BAP_shrunk, Teuchos::NO_TRANS, Teuchos::ScalarTraits<scalar_type>::one(), Teuchos::ScalarTraits<scalar_type>::zero());
 
-    if(1 == coordinates->getNumVectors())
+    // Columns Map for BAP
+    std::vector<int> indBAPcolMap;
+    if (1 == coordinates->getNumVectors())
+      neighbours1D(tpetra_prolong, indBAPcolMap, comm);
+    else if (2 == coordinates->getNumVectors())
+      neighbours2D(tpetra_prolong, indBAPcolMap, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+    else if (3 == coordinates->getNumVectors())
+      neighbours3D(tpetra_prolong, indBAPcolMap, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+    Teuchos::ArrayView<const global_ordinal_type> elementListBAP(indBAPcolMap);
+    Teuchos::RCP<Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>> BAPcolMap = rcp(new Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>(static_cast<size_t>(tpetra_prolong->getGlobalNumCols()), elementListBAP, indexBase, comm));
+
+    RCP<crs_matrix_type> BAP                                      = rcp(new crs_matrix_type(tpetra_prolong->getRowMap(), BAPcolMap, tpetra_prolong->getGlobalNumCols()));
+    Teuchos::ArrayView<const global_ordinal_type> myLocalElements = BAP->getRowMap()->getLocalElementList();
+
+    if (1 == coordinates->getNumVectors())
       BAP1D(BAP, tpetra_prolong, BAP_shrunk, comm);
-    else if(2 == coordinates->getNumVectors())
-     BAP2D( BAP, tpetra_prolong, BAP_shrunk, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()) );
-    else if(3 == coordinates->getNumVectors())
-     BAP3D( BAP, tpetra_prolong, BAP_shrunk, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()) );
-
-    BAP->fillComplete( tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap() );
-	
-    //Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("BAP.mtx", BAP);
-//=============================================================================================================
-  RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
-  mueluPbar = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type,local_ordinal_type,global_ordinal_type,node_type>(Pbar);
-}
-PbarSetUp->stop();
+    else if (2 == coordinates->getNumVectors())
+      BAP2D(BAP, tpetra_prolong, BAP_shrunk, comm, std::sqrt(tpetra_prolong->getGlobalNumCols()));
+    else if (3 == coordinates->getNumVectors())
+      BAP3D(BAP, tpetra_prolong, BAP_shrunk, comm, std::cbrt(tpetra_prolong->getGlobalNumCols()), std::cbrt(tpetra_prolong->getGlobalNumCols()));
+
+    BAP->fillComplete(tpetra_prolong->getDomainMap(), tpetra_prolong->getRangeMap());
+
+    // Tpetra::MatrixMarket::Writer<crs_matrix_type>::writeSparseFile("BAP.mtx", BAP);
+    //=============================================================================================================
+    RCP<crs_matrix_type> Pbar = Tpetra::MatrixMatrix::add(1.0, false, *tpetra_prolong, -1.0, false, *BAP);
+    mueluPbar                 = MueLu::TpetraCrs_To_XpetraMatrix<scalar_type, local_ordinal_type, global_ordinal_type, node_type>(Pbar);
+  }
+  PbarSetUp->stop();
 
   H->GetLevel(1)->Set("Pbar", mueluPbar);
 
@@ -429,70 +407,68 @@ PbarSetUp->stop();
 
   M = rcp(new muelu_tpetra_operator_type(H));
 
-  RCP<multivector_type> X = rcp(new multivector_type(map,1));
-  RCP<multivector_type> B = rcp(new multivector_type(map,1));
-
-for(int trial = 1; trial<=number_runs; ++trial)
-{
-
-  X->putScalar((scalar_type) 0.0);
-  B->randomize();
-
-  //
-  // Set up Krylov solver and iterate.
-  //
- 
-  RCP<multivector_type> X_muelu = rcp(new multivector_type(map,1));
-  RCP<linear_problem_type> Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
-  Problem_muelu->setRightPrec(M);
-  Problem_muelu->setProblem();
-
-  RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
-  belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  //belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList->set("Verbosity",             Belos::Errors);
-  belosList->set("Output Frequency",      1);
-  belosList->set("Output Style",          Belos::Brief);
-  belosList->set("Implicit Residual Scaling", "None");
-  RCP<belos_solver_manager_type> solver;
-  if (krylovSolverType == "cg")
-    solver = rcp(new belos_pseudocg_manager_type(Problem_muelu, belosList));
-  else if (krylovSolverType == "gmres")
-    solver = rcp(new belos_gmres_manager_type(Problem_muelu, belosList));
-  else if (krylovSolverType == "bicgstab")
-    solver = rcp(new belos_bicgstab_manager_type(Problem_muelu, belosList));
-  else
-    throw std::invalid_argument("bad Krylov solver type");
-
-  solver->solve();
-  int numIterations_muelu = solver->getNumIters();
-
-  Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
-  multivector_type residual_muelu(B->getMap(),1);
-  A->apply(*X_muelu, residual_muelu);
-  residual_muelu.update(1.0, *B, -1.0);
-  residual_muelu.norm2(normVec_muelu);
-  if (mypid == 0) {
-    std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
-    std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
- }
-}
+  RCP<multivector_type> X = rcp(new multivector_type(map, 1));
+  RCP<multivector_type> B = rcp(new multivector_type(map, 1));
+
+  for (int trial = 1; trial <= number_runs; ++trial) {
+    X->putScalar((scalar_type)0.0);
+    B->randomize();
+
+    //
+    // Set up Krylov solver and iterate.
+    //
+
+    RCP<multivector_type> X_muelu          = rcp(new multivector_type(map, 1));
+    RCP<linear_problem_type> Problem_muelu = rcp(new linear_problem_type(A, X_muelu, B));
+    Problem_muelu->setRightPrec(M);
+    Problem_muelu->setProblem();
+
+    RCP<Teuchos::ParameterList> belosList = rcp(new Teuchos::ParameterList());
+    belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+    belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+    // belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    belosList->set("Verbosity", Belos::Errors);
+    belosList->set("Output Frequency", 1);
+    belosList->set("Output Style", Belos::Brief);
+    belosList->set("Implicit Residual Scaling", "None");
+    RCP<belos_solver_manager_type> solver;
+    if (krylovSolverType == "cg")
+      solver = rcp(new belos_pseudocg_manager_type(Problem_muelu, belosList));
+    else if (krylovSolverType == "gmres")
+      solver = rcp(new belos_gmres_manager_type(Problem_muelu, belosList));
+    else if (krylovSolverType == "bicgstab")
+      solver = rcp(new belos_bicgstab_manager_type(Problem_muelu, belosList));
+    else
+      throw std::invalid_argument("bad Krylov solver type");
+
+    solver->solve();
+    int numIterations_muelu = solver->getNumIters();
+
+    Teuchos::Array<typename Teuchos::ScalarTraits<scalar_type>::magnitudeType> normVec_muelu(1);
+    multivector_type residual_muelu(B->getMap(), 1);
+    A->apply(*X_muelu, residual_muelu);
+    residual_muelu.update(1.0, *B, -1.0);
+    residual_muelu.norm2(normVec_muelu);
+    if (mypid == 0) {
+      std::cout << "number of iterations with MueLu preconditioner= " << numIterations_muelu << std::endl;
+      std::cout << "||Residual|| = " << normVec_muelu[0] << std::endl;
+    }
+  }
 
-  //Teuchos::TimeMonitor::summarize ();
+  // Teuchos::TimeMonitor::summarize ();
 
-   //Columns Map for BAP
-   std::vector<int> indTimerMap;
-   if(0!=mypid)
-     indTimerMap.emplace_back(mypid-1);
+  // Columns Map for BAP
+  std::vector<int> indTimerMap;
+  if (0 != mypid)
+    indTimerMap.emplace_back(mypid - 1);
 
-   Teuchos::ArrayView<const global_ordinal_type> elementListTimer (indTimerMap);
-   Teuchos::RCP< Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type > > TimerMap = rcp( new Tpetra::Map< local_ordinal_type, global_ordinal_type, node_type >( static_cast<size_t>(comm->getSize()), elementListTimer, indexBase, comm ) );
+  Teuchos::ArrayView<const global_ordinal_type> elementListTimer(indTimerMap);
+  Teuchos::RCP<Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>> TimerMap = rcp(new Tpetra::Map<local_ordinal_type, global_ordinal_type, node_type>(static_cast<size_t>(comm->getSize()), elementListTimer, indexBase, comm));
 
-  RCP<multivector_type> TimerRestr = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerProlong = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerFine = rcp(new multivector_type(TimerMap,1));
-  RCP<multivector_type> TimerCoarse = rcp(new multivector_type(TimerMap,1));
+  RCP<multivector_type> TimerRestr   = rcp(new multivector_type(TimerMap, 1));
+  RCP<multivector_type> TimerProlong = rcp(new multivector_type(TimerMap, 1));
+  RCP<multivector_type> TimerFine    = rcp(new multivector_type(TimerMap, 1));
+  RCP<multivector_type> TimerCoarse  = rcp(new multivector_type(TimerMap, 1));
 
   return EXIT_SUCCESS;
 }
diff --git a/packages/muelu/research/max/XpetraSplitting/Test_muelu.cpp b/packages/muelu/research/max/XpetraSplitting/Test_muelu.cpp
index 271490c23681..c06dabe58572 100644
--- a/packages/muelu/research/max/XpetraSplitting/Test_muelu.cpp
+++ b/packages/muelu/research/max/XpetraSplitting/Test_muelu.cpp
@@ -40,53 +40,46 @@
 #include <MueLu_CreateTpetraPreconditioner.hpp>
 #include <MueLu_Utilities.hpp>
 
-
 // =========== //
 // main driver //
 // =========== //
 
-int main(int argc, char* argv[])
-{
-
-
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
-  typedef scalar_type         												Scalar;
-  typedef local_ordinal_type  												LocalOrdinal;
-  typedef global_ordinal_type 												GlobalOrdinal;
+int main(int argc, char* argv[]) {
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
+  typedef global_ordinal_type GlobalOrdinal;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
-  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> 											Matrix;
-  typedef Xpetra::MatrixSplitting<Scalar, LocalOrdinal, GlobalOrdinal, Node>							MatrixSplitting;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+  typedef Xpetra::MatrixSplitting<Scalar, LocalOrdinal, GlobalOrdinal, Node> MatrixSplitting;
   typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Xpetra::EpetraNode> EpCrsMatrix;
 
-
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(argc<2, "\nInvalid name for input matrix\n");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(argc < 2, "\nInvalid name for input matrix\n");
 
   int numGlobalElements = 1;
 
   Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-  //Create Xpetra map
-  Teuchos::RCP<const Xpetra::Map<int,GlobalOrdinal,Xpetra::EpetraNode> > xpetraMap;
-  xpetraMap = Xpetra::MapFactory<int,GlobalOrdinal,Xpetra::EpetraNode>::Build(Xpetra::UseEpetra, numGlobalElements, 0, comm);
+  // Create Xpetra map
+  Teuchos::RCP<const Xpetra::Map<int, GlobalOrdinal, Xpetra::EpetraNode> > xpetraMap;
+  xpetraMap = Xpetra::MapFactory<int, GlobalOrdinal, Xpetra::EpetraNode>::Build(Xpetra::UseEpetra, numGlobalElements, 0, comm);
 
-  //Import matrix from an .mtx file into an Xpetra wrapper for an Epetra matrix
-  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xpetraMatrix = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(argv[1], Xpetra::UseEpetra, comm);
-  //Export matrix from an Xpetra wrapper into an .mtx file
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("A_write.mtx", *xpetraMatrix);
+  // Import matrix from an .mtx file into an Xpetra wrapper for an Epetra matrix
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xpetraMatrix = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(argv[1], Xpetra::UseEpetra, comm);
+  // Export matrix from an Xpetra wrapper into an .mtx file
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A_write.mtx", *xpetraMatrix);
 
   Teuchos::RCP<MatrixSplitting> xpetraMatrixSplitting;
 
   Teuchos::ParameterList xmlParams;
-  Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Hierarchy = MueLu::CreateXpetraPreconditioner( (Teuchos::RCP<Matrix>)xpetraMatrixSplitting, xmlParams );
-
+  Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy = MueLu::CreateXpetraPreconditioner((Teuchos::RCP<Matrix>)xpetraMatrixSplitting, xmlParams);
 
 #ifdef HAVE_MPI
   MPI_Finalize();
 #endif
 
-  return(EXIT_SUCCESS);
+  return (EXIT_SUCCESS);
 }
-
diff --git a/packages/muelu/research/max/XpetraSplitting/Test_xpetra.cpp b/packages/muelu/research/max/XpetraSplitting/Test_xpetra.cpp
index 287f2aed5646..e712344b5f95 100644
--- a/packages/muelu/research/max/XpetraSplitting/Test_xpetra.cpp
+++ b/packages/muelu/research/max/XpetraSplitting/Test_xpetra.cpp
@@ -36,25 +36,21 @@
 #include "Epetra_CrsMatrix.h"
 #include "Epetra_LinearProblem.h"
 
-
 // =========== //
 // main driver //
 // =========== //
 
-int main(int argc, char* argv[])
-{
-
-
-  typedef double                                      scalar_type;
-  typedef int                                         local_ordinal_type;
-  typedef int                                         global_ordinal_type;
-  typedef scalar_type         												Scalar;
-  typedef local_ordinal_type  												LocalOrdinal;
-  typedef global_ordinal_type 												GlobalOrdinal;
+int main(int argc, char* argv[]) {
+  typedef double scalar_type;
+  typedef int local_ordinal_type;
+  typedef int global_ordinal_type;
+  typedef scalar_type Scalar;
+  typedef local_ordinal_type LocalOrdinal;
+  typedef global_ordinal_type GlobalOrdinal;
   typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
-  typedef Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> multivector_type;
-  typedef Xpetra::MatrixSplitting<Scalar,LocalOrdinal,GlobalOrdinal,Node,Xpetra::UseTpetra, false> tpetra_splitting;
+  typedef Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> multivector_type;
+  typedef Xpetra::MatrixSplitting<Scalar, LocalOrdinal, GlobalOrdinal, Node, Xpetra::UseTpetra, false> tpetra_splitting;
 
 #ifdef HAVE_MPI
   MPI_Init(&argc, &argv);
@@ -69,13 +65,13 @@ int main(int argc, char* argv[])
   //
   // with Matrix arising from a 5-point formula discretization.
 
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(argc<4, "\nInvalid name for input matrix and output file\n");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(argc < 4, "\nInvalid name for input matrix and output file\n");
 
   Teuchos::RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
   // process command line arguments
-  const char* xmlFileName = argv[1];
-  const char* matrixFileName = argv[2];
+  const char* xmlFileName     = argv[1];
+  const char* matrixFileName  = argv[2];
   const char* mappingFileName = argv[3];
 
   Teuchos::ParameterList xmlParams;
@@ -84,52 +80,49 @@ int main(int argc, char* argv[])
 
   Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
   if (CommEpetra.MyPID() == 0)
-    std::cout<<"Number of processors: "<<CommEpetra.NumProc()<<std::endl;
+    std::cout << "Number of processors: " << CommEpetra.NumProc() << std::endl;
 
-
-  //SplittingDriver
+  // SplittingDriver
   /*Xpetra::SplittingDriver<Scalar, LocalOrdinal, GlobalOrdinal, Node> driver("node.txt", comm);
-	Teuchos::Array<GlobalOrdinal> elementlist = driver.GetGlobalRowMap();
-	driver.printInactive();
-	Xpetra::MatrixSplitting<Scalar,LocalOrdinal,GlobalOrdinal,Node,Xpetra::UseTpetra> xpetraWrapper( argv[1], argv[2], comm );
-	std::string output_file="A_write.mtx";
-	xpetraWrapper.writeGlobalMatrix();
-	xpetraWrapper.writeRegionMatrices();*/
-
+        Teuchos::Array<GlobalOrdinal> elementlist = driver.GetGlobalRowMap();
+        driver.printInactive();
+        Xpetra::MatrixSplitting<Scalar,LocalOrdinal,GlobalOrdinal,Node,Xpetra::UseTpetra> xpetraWrapper( argv[1], argv[2], comm );
+        std::string output_file="A_write.mtx";
+        xpetraWrapper.writeGlobalMatrix();
+        xpetraWrapper.writeRegionMatrices();*/
 
-  //Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
-  //A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(argv[2], Xpetra::UseTpetra, comm);
+  // Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
+  // A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(argv[2], Xpetra::UseTpetra, comm);
 
   // Create the RegionHandler to deal with mappings of nodes to regions etc.
-  Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler = Teuchos::rcp(new Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> (mappingFileName, comm));
-  Teuchos::Array<GlobalOrdinal> elementlist = regionHandler->GetGlobalRowMap();
-  std::size_t num_total_elements = regionHandler->GetNumGlobalElements();
-  std::size_t num_total_regions = regionHandler->GetNumTotalRegions();
+  Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler = Teuchos::rcp(new Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>(mappingFileName, comm));
+  Teuchos::Array<GlobalOrdinal> elementlist                                                     = regionHandler->GetGlobalRowMap();
+  std::size_t num_total_elements                                                                = regionHandler->GetNumGlobalElements();
+  std::size_t num_total_regions                                                                 = regionHandler->GetNumTotalRegions();
 
   // Read and split the matrix
   Teuchos::RCP<tpetra_splitting> matrixSplitting = Teuchos::rcp(new tpetra_splitting(matrixFileName, regionHandler, comm));
 
   // Create region-wise AMG hierarchy
-  int max_num_levels = 4;
+  int max_num_levels    = 4;
   int coarsening_factor = 3;
-  Xpetra::RegionAMG<Scalar,LocalOrdinal,GlobalOrdinal,Node> preconditioner(matrixSplitting, regionHandler, comm, mueluParams, max_num_levels, coarsening_factor);
-
-//  // Setup vectors for test problem
-//  Teuchos::RCP<multivector_type> X = Xpetra::MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(preconditioner.getDomainMap(), 1) ;
-//  Teuchos::RCP<multivector_type> Y = Xpetra::MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(preconditioner.getRangeMap(), 1) ;
-//  X->randomize();
-//  Y->putScalar((scalar_type) 0.0);
-//
-//  // Apply the preconditioner
-//  preconditioner.apply(*X,*Y);
-//
-//  // Output result to screen
-//  Y->describe(*out, Teuchos::VERB_EXTREME);
+  Xpetra::RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node> preconditioner(matrixSplitting, regionHandler, comm, mueluParams, max_num_levels, coarsening_factor);
+
+  //  // Setup vectors for test problem
+  //  Teuchos::RCP<multivector_type> X = Xpetra::MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(preconditioner.getDomainMap(), 1) ;
+  //  Teuchos::RCP<multivector_type> Y = Xpetra::MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(preconditioner.getRangeMap(), 1) ;
+  //  X->randomize();
+  //  Y->putScalar((scalar_type) 0.0);
+  //
+  //  // Apply the preconditioner
+  //  preconditioner.apply(*X,*Y);
+  //
+  //  // Output result to screen
+  //  Y->describe(*out, Teuchos::VERB_EXTREME);
 
 #ifdef HAVE_MPI
   MPI_Finalize();
 #endif
 
-  return(EXIT_SUCCESS);
+  return (EXIT_SUCCESS);
 }
-
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_decl.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_decl.hpp
index 7207c1cae481..49e3b8b48c7b 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_decl.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_decl.hpp
@@ -52,119 +52,114 @@
 #ifndef XPETRA_LEVEL_DECL_HPP
 #define XPETRA_LEVEL_DECL_HPP
 
-namespace Xpetra{
+namespace Xpetra {
 
-	template<class Scalar					= MultiVector<>::scalar_type,
-		class LocalOrdinal   				= typename MultiVector<Scalar>::local_ordinal_type,
-		class GlobalOrdinal  				= typename MultiVector<Scalar, LocalOrdinal>::global_ordinal_type,
-		class Node           				= typename MultiVector<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-		class Level {
+template <class Scalar        = MultiVector<>::scalar_type,
+          class LocalOrdinal  = typename MultiVector<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename MultiVector<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename MultiVector<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class Level {
+  typedef Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+  typedef Vector<Scalar, LocalOrdinal, GlobalOrdinal> vector_type;
+  typedef MultiVector<Scalar, LocalOrdinal, GlobalOrdinal> multivector_type;
+  typedef Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> matrix_type;
 
-		typedef Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
-		typedef Vector<Scalar, LocalOrdinal, GlobalOrdinal> vector_type;
-		typedef MultiVector<Scalar, LocalOrdinal, GlobalOrdinal> multivector_type;
-		typedef Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node > matrix_type;
+ public:
+  //! Constructors/destructors
+  //@{
+  // Constructor
+  Level(int, int);
 
-		public:
+  // Default destructor
+  virtual ~Level(){};
 
-			//! Constructors/destructors
-			//@{	
-			// Constructor
-			Level( int, int );
+  //@}
 
-			//Default destructor
-			virtual ~Level(){};
+  //! Public methods
+  //@{
+  virtual void
+  apply(const multivector_type& X, multivector_type& Y,
+        Teuchos::ETransp mode = Teuchos::NO_TRANS,
+        Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+        Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const {};
 
-			//@}
+  virtual bool hasTransposeApply() const { return false; }
 
-			//! Public methods
-			//@{
-			virtual void
-				apply (const multivector_type& X, multivector_type& Y,
-				Teuchos::ETransp mode = Teuchos::NO_TRANS,
-				Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-				Scalar beta = Teuchos::ScalarTraits<Scalar>::zero())const{};
+  //! Get methods
+  //@{
 
-			virtual bool hasTransposeApply() const { return false; }
+  //! Extract the number of regions instantiated in the level
+  int GetNumRegions() const;
 
-			//! Get methods
-			//@{
+  GlobalOrdinal GetNumRegionNodes(GlobalOrdinal region_idx) const { return regionA_[region_idx]->getCrsGraph()->getGlobalNumRows(); };
 
-			//!Extract the number of regions instantiated in the level
-			int GetNumRegions() const;
+  RCP<matrix_type> GetRegionMatrix(GlobalOrdinal region_idx) const { return regionA_[region_idx]; };
 
-			GlobalOrdinal GetNumRegionNodes( GlobalOrdinal region_idx)const{return regionA_[region_idx]->getCrsGraph()->getGlobalNumRows();};
+  //! Extract regionToAll from the level
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > GetRegionToAll() const;
 
-			RCP<matrix_type> GetRegionMatrix( GlobalOrdinal region_idx)const{return regionA_[region_idx];};
+  // Extract the level ID
+  GlobalOrdinal GetLevelID() const { return levelID_; }
 
-			//! Extract regionToAll from the level
-			Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > GetRegionToAll() const;
+  //! Take a region index and returns the composite index for a given mesh node
+  GlobalOrdinal GetCompositeIndex(int, GlobalOrdinal) const;
 
-			// Extract the level ID
-			GlobalOrdinal GetLevelID() const {return levelID_;}
+  //@}
 
-			//! Take a region index and returns the composite index for a given mesh node
-			GlobalOrdinal GetCompositeIndex(int, GlobalOrdinal)const;
+  //@}
 
-			//@}
-			
-			//@}
+  //! Set up methods
+  //@{
+  //! Set the region galerkin operators
+  void SetA(Array<RCP<matrix_type> >&);
 
-			//! Set up methods
-			//@{
-			//! Set the region galerkin operators
-			void SetA( Array<RCP<matrix_type> >& );
+  //! Set the region prolongators
+  void SetP(Array<RCP<matrix_type> >&);
 
-			//! Set the region prolongators
-			void SetP( Array<RCP<matrix_type> >& );
+  //! Set the region restrioctions
+  void SetR(Array<RCP<matrix_type> >&);
 
-			//! Set the region restrioctions
-			void SetR( Array<RCP<matrix_type> >& );
+  //! Set the region smoothers
+  void SetSmoother(Array<RCP<multivector_type> >&);
 
-			//! Set the region smoothers
-			void SetSmoother( Array<RCP<multivector_type> >& );
+  //! Set the regionToAll structure for the current level by using information coming from the finer level
+  void SetRegionToAll(Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > >);
 
-			//! Set the regionToAll structure for the current level by using information coming from the finer level
-			void SetRegionToAll( Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > );
+  //@}
 
-			//@}
-			
-			//! Check method
-			//@{
-			//! Control that algebraic quantities have matching dimensions
-			void checkConsistency() const;
-			//@}
+  //! Check method
+  //@{
+  //! Control that algebraic quantities have matching dimensions
+  void checkConsistency() const;
+  //@}
 
-			//! Constrcution of region smoothers
-			//@{			
-			void ComputeRegionJacobi();
-			//@}
+  //! Constrcution of region smoothers
+  //@{
+  void ComputeRegionJacobi();
+  //@}
 
-		private:
+ private:
+  //! Private variables
+  //@{
+  GlobalOrdinal levelID_     = -1;
+  GlobalOrdinal num_regions_ = -1;
 
-			//! Private variables
-			//@{
-			GlobalOrdinal levelID_ = -1;
-			GlobalOrdinal num_regions_ = -1;
+  //! Region operators
+  Array<RCP<matrix_type> > regionA_;
 
-			//! Region operators
-			Array<RCP<matrix_type> > regionA_;
+  //! Region grid transfers
+  Array<RCP<matrix_type> > regionP_;
+  Array<RCP<matrix_type> > regionR_;
 
-			//! Region grid transfers
-			Array<RCP<matrix_type> > regionP_;
-			Array<RCP<matrix_type> > regionR_;
+  //! Smoother
+  Array<RCP<vector_type> > regionSmoother_;
 
-			//! Smoother
-			Array<RCP<vector_type> > regionSmoother_;
+  //! Auxiliary quantities to handle regions at each coarsening level
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > level_regionToAll_;
 
-			//! Auxiliary quantities to handle regions at each coarsening level
-			Array< Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > level_regionToAll_;
+  //@}
+};
 
-			//@}
-
-
-	};
-
-} //namespace Xpetra
+}  // namespace Xpetra
 
 #endif
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_def.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_def.hpp
index a4436045e78d..c263236c0ee2 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_def.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_Level_def.hpp
@@ -54,134 +54,109 @@
 
 #include "Xpetra_Level_decl.hpp"
 
-namespace Xpetra
-{
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Level( int levelID, int num_regions ):
-levelID_(levelID),
-num_regions_(num_regions){}
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetP( Array<RCP<matrix_type> >& P )
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( regionP_.size()!=0, Exceptions::RuntimeError, "Current level already has prolongators \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( P.size()!=num_regions_, Exceptions::RuntimeError, "Number of region prolongators is "<<P.size()<<"and does not math the number of regions which is "<<num_regions_<<" \n" );
+namespace Xpetra {
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Level(int levelID, int num_regions)
+  : levelID_(levelID)
+  , num_regions_(num_regions) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetP(Array<RCP<matrix_type> >& P) {
+  TEUCHOS_TEST_FOR_EXCEPTION(regionP_.size() != 0, Exceptions::RuntimeError, "Current level already has prolongators \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(P.size() != num_regions_, Exceptions::RuntimeError, "Number of region prolongators is " << P.size() << "and does not math the number of regions which is " << num_regions_ << " \n");
   regionP_ = P;
 }
 
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetR( Array<RCP<matrix_type> >& R )
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( regionR_.size()!=0, Exceptions::RuntimeError, "Current level already has restrictions \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( R.size()!=num_regions_, Exceptions::RuntimeError, "Number of region restrictions is "<<R.size()<<"and does not math the number of regions which is "<<num_regions_<<" \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetR(Array<RCP<matrix_type> >& R) {
+  TEUCHOS_TEST_FOR_EXCEPTION(regionR_.size() != 0, Exceptions::RuntimeError, "Current level already has restrictions \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(R.size() != num_regions_, Exceptions::RuntimeError, "Number of region restrictions is " << R.size() << "and does not math the number of regions which is " << num_regions_ << " \n");
   regionR_ = R;
 }
 
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetA( Array<RCP<matrix_type> >& A )
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( regionA_.size()!=0, Exceptions::RuntimeError, "Current level already has operators \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( A.size()!=num_regions_, Exceptions::RuntimeError, "Number of region operators is "<<A.size()<<"and does not math the number of regions which is "<<num_regions_<<" \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetA(Array<RCP<matrix_type> >& A) {
+  TEUCHOS_TEST_FOR_EXCEPTION(regionA_.size() != 0, Exceptions::RuntimeError, "Current level already has operators \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(A.size() != num_regions_, Exceptions::RuntimeError, "Number of region operators is " << A.size() << "and does not math the number of regions which is " << num_regions_ << " \n");
   regionA_ = A;
 }
 
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmoother( Array<RCP<multivector_type> >& S )
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( regionSmoother_.size()!=0, Exceptions::RuntimeError, "Current level already has smoothers \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( S.size()!=num_regions_, Exceptions::RuntimeError, "Number of region smoothers is "<<S.size()<<"and does not math the number of regions which is "<<num_regions_<<" \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmoother(Array<RCP<multivector_type> >& S) {
+  TEUCHOS_TEST_FOR_EXCEPTION(regionSmoother_.size() != 0, Exceptions::RuntimeError, "Current level already has smoothers \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(S.size() != num_regions_, Exceptions::RuntimeError, "Number of region smoothers is " << S.size() << "and does not math the number of regions which is " << num_regions_ << " \n");
   regionSmoother_ = S;
 }
 
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetRegionToAll( Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll )
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( regionToAll.size()!=num_regions_, Exceptions::RuntimeError, "Passed regionToAll has number of regions equal to "<<regionToAll.size()<<"which does not match the number of regions in level ID: <<"<<levelID_<<" with number of regions equal to "<<num_regions_<<"\n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetRegionToAll(Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll) {
+  TEUCHOS_TEST_FOR_EXCEPTION(regionToAll.size() != num_regions_, Exceptions::RuntimeError, "Passed regionToAll has number of regions equal to " << regionToAll.size() << "which does not match the number of regions in level ID: <<" << levelID_ << " with number of regions equal to " << num_regions_ << "\n");
   level_regionToAll_ = regionToAll;
 }
 
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-int Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNumRegions( ) const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_==0, Exceptions::RuntimeError, "level ID: <<"<<levelID_<<" does NOT have defined regions \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNumRegions() const {
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ == 0, Exceptions::RuntimeError, "level ID: <<" << levelID_ << " does NOT have defined regions \n");
   return num_regions_;
 }
 
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll( ) const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( level_regionToAll_.size()==0, Exceptions::RuntimeError, "level ID: "<<levelID_<<" does NOT have level_regionToAll_ initialized yet \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll() const {
+  TEUCHOS_TEST_FOR_EXCEPTION(level_regionToAll_.size() == 0, Exceptions::RuntimeError, "level ID: " << levelID_ << " does NOT have level_regionToAll_ initialized yet \n");
   return level_regionToAll_;
 }
 
-
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-GlobalOrdinal Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetCompositeIndex(int region_idx, GlobalOrdinal region_node_idx ) const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( level_regionToAll_.size()==0, Exceptions::RuntimeError, "level ID: "<<levelID_<<" does NOT have level_regionToAll_ initialized yet \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( level_regionToAll_.size()!=num_regions_, Exceptions::RuntimeError, "level ID: "<<levelID_<<" has information stored for a number of regions that does NOT match with the declared number of regions \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( region_idx>=num_regions_, Exceptions::RuntimeError, "level ID: "<<levelID_<<" Invalid region index \n" );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetCompositeIndex(int region_idx, GlobalOrdinal region_node_idx) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(level_regionToAll_.size() == 0, Exceptions::RuntimeError, "level ID: " << levelID_ << " does NOT have level_regionToAll_ initialized yet \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(level_regionToAll_.size() != num_regions_, Exceptions::RuntimeError, "level ID: " << levelID_ << " has information stored for a number of regions that does NOT match with the declared number of regions \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(region_idx >= num_regions_, Exceptions::RuntimeError, "level ID: " << levelID_ << " Invalid region index \n");
 
   GlobalOrdinal composite_index = -1;
 
   checkerRegionToAll<GlobalOrdinal> unaryPredicate(region_node_idx);
-  typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator global_iterator;
-  Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >   regionToAll = level_regionToAll_[region_idx];
-  global_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
-  TEUCHOS_TEST_FOR_EXCEPTION( global_iterator==level_regionToAll_[region_idx].end(), Exceptions::RuntimeError, " - Region: "<<region_idx<<" - "<<" node with region index: "<<region_idx<<" is not in regionToAll["<<region_idx<<"]"<<"\n" );
-  composite_index = std::get<1>( *global_iterator );
+  typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator global_iterator;
+  Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > regionToAll = level_regionToAll_[region_idx];
+  global_iterator                                              = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
+  TEUCHOS_TEST_FOR_EXCEPTION(global_iterator == level_regionToAll_[region_idx].end(), Exceptions::RuntimeError, " - Region: " << region_idx << " - "
+                                                                                                                              << " node with region index: " << region_idx << " is not in regionToAll[" << region_idx << "]"
+                                                                                                                              << "\n");
+  composite_index = std::get<1>(*global_iterator);
 
   return composite_index;
 }
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkConsistency() const {
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ <= 0, Exceptions::RuntimeError, "level ID: " << levelID_ << " does not have any regions \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ != regionA_.size(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - number of region matrices " << regionA_.size() << " does NOT match the number of regions " << num_regions_ << "\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ != level_regionToAll_.size(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - size of level_regionToAll_ " << level_regionToAll_.size() << " does NOT match the number of regions " << num_regions_ << "\n");
 
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkConsistency( ) const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_<=0, Exceptions::RuntimeError, "level ID: "<<levelID_<<" does not have any regions \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_!=regionA_.size(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - number of region matrices "<<regionA_.size()<< " does NOT match the number of regions "<<num_regions_<<"\n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_!=level_regionToAll_.size(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - size of level_regionToAll_ "<<level_regionToAll_.size()<<" does NOT match the number of regions "<<num_regions_<<"\n" );
-
-  if( levelID_>0 )
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION( num_regions_!=regionP_.size(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - number of prolongators "<<regionP_.size()<<" does NOT match the number of regions "<<num_regions_<<"\n" );
-    TEUCHOS_TEST_FOR_EXCEPTION( num_regions_!=regionR_.size(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - number of restrictions "<<regionR_.size()<<" does NOT match the number of regions "<<num_regions_<<"\n" );
+  if (levelID_ > 0) {
+    TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ != regionP_.size(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - number of prolongators " << regionP_.size() << " does NOT match the number of regions " << num_regions_ << "\n");
+    TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ != regionR_.size(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - number of restrictions " << regionR_.size() << " does NOT match the number of regions " << num_regions_ << "\n");
   }
 
-
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-
-    if( levelID_>0 )
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION( regionP_[region_idx]->getGlobalNumCols()!=regionA_[region_idx]->getGlobalNumCols(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - Region: "<<region_idx<<" , numCols(P) = "<<regionP_[region_idx]->getGlobalNumCols()<<" BUT numCols(A) = "<<regionA_[region_idx]->getGlobalNumCols()<<"\n" );
-      TEUCHOS_TEST_FOR_EXCEPTION( regionR_[region_idx]->getGlobalNumRows()!=regionA_[region_idx]->getGlobalNumCols(), Exceptions::RuntimeError, "level ID: "<<levelID_<<" - Region: "<<region_idx<<" , numRows(R) = "<<regionR_[region_idx]->getGlobalNumRows()<<" BUT numCols(A) = "<<regionA_[region_idx]->getGlobalNumCols()<<"\n" );
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    if (levelID_ > 0) {
+      TEUCHOS_TEST_FOR_EXCEPTION(regionP_[region_idx]->getGlobalNumCols() != regionA_[region_idx]->getGlobalNumCols(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - Region: " << region_idx << " , numCols(P) = " << regionP_[region_idx]->getGlobalNumCols() << " BUT numCols(A) = " << regionA_[region_idx]->getGlobalNumCols() << "\n");
+      TEUCHOS_TEST_FOR_EXCEPTION(regionR_[region_idx]->getGlobalNumRows() != regionA_[region_idx]->getGlobalNumCols(), Exceptions::RuntimeError, "level ID: " << levelID_ << " - Region: " << region_idx << " , numRows(R) = " << regionR_[region_idx]->getGlobalNumRows() << " BUT numCols(A) = " << regionA_[region_idx]->getGlobalNumCols() << "\n");
     }
 
-    TEUCHOS_TEST_FOR_EXCEPTION( regionA_[region_idx]->getRowMap()->getLocalNumElements()>0 && level_regionToAll_[region_idx].size()!=regionA_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: "<<regionA_[region_idx]->getRowMap()->getComm()->getRank()<<" - level ID: "<<levelID_<<" - Region: "<<region_idx<<" , size(regionToAll) = "<<level_regionToAll_[region_idx].size()<<" BUT numCols(A) = "<<regionA_[region_idx]->getGlobalNumCols()<<"\n" );
-
+    TEUCHOS_TEST_FOR_EXCEPTION(regionA_[region_idx]->getRowMap()->getLocalNumElements() > 0 && level_regionToAll_[region_idx].size() != regionA_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: " << regionA_[region_idx]->getRowMap()->getComm()->getRank() << " - level ID: " << levelID_ << " - Region: " << region_idx << " , size(regionToAll) = " << level_regionToAll_[region_idx].size() << " BUT numCols(A) = " << regionA_[region_idx]->getGlobalNumCols() << "\n");
   }
-
 }
 
-
-
-template< class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node >
-void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeRegionJacobi( )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Level<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeRegionJacobi() {
   regionSmoother_.resize(num_regions_);
 
-  for( int i = 0; i<num_regions_; ++i )
-  {
-    regionSmoother_[i] =  VectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(regionA_[i]->getRowMap()) ;
-    regionA_[i]->getLocalDiagCopy( *(regionSmoother_[i]) );
+  for (int i = 0; i < num_regions_; ++i) {
+    regionSmoother_[i] = VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(regionA_[i]->getRowMap());
+    regionA_[i]->getLocalDiagCopy(*(regionSmoother_[i]));
   }
-
 }
 
-}
+}  // namespace Xpetra
 #endif
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_MatrixSplitting.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_MatrixSplitting.hpp
index 5b28ded63dd6..c2fd9d96d805 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_MatrixSplitting.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_MatrixSplitting.hpp
@@ -52,7 +52,7 @@
 #ifndef XPETRA_MATRIXSPLITTING_HPP
 #define XPETRA_MATRIXSPLITTING_HPP
 
-//Xpetra
+// Xpetra
 #include "Xpetra_Map.hpp"
 #include "Xpetra_MapFactory.hpp"
 #include "Xpetra_Matrix.hpp"
@@ -60,10 +60,10 @@
 #include "Xpetra_IO.hpp"
 #include "Xpetra_RegionHandler_def.hpp"
 
-//Ifpack2
+// Ifpack2
 #include "Ifpack2_OverlappingRowMatrix_def.hpp"
 
-//MueLu
+// MueLu
 #include <MueLu_Utilities.hpp>
 
 /** \file Xpetra_MatrixSplitting.hpp
@@ -72,157 +72,134 @@ Declarations for the class Xpetra::MatrixSplitting.
  */
 namespace Xpetra {
 
-
-//Definition of the predicate for the regionToAll structure.
-//Given a tuple made of node index and a specific region it belongs to,
-//this predicate returns true if the node has composite index which coincides with the index specified in input.
-template<class GlobalOrdinal>
+// Definition of the predicate for the regionToAll structure.
+// Given a tuple made of node index and a specific region it belongs to,
+// this predicate returns true if the node has composite index which coincides with the index specified in input.
+template <class GlobalOrdinal>
 class checkerRegionToAll {
+ public:
+  // Constructor
+  checkerRegionToAll(GlobalOrdinal node_index) { node_index_ = node_index; };
 
-public:
-
-  //Constructor
-  checkerRegionToAll( GlobalOrdinal node_index){node_index_ = node_index;};
-
-  //Unary Operator
-  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal > &node)
-  { return (std::get<0>(node) == node_index_); }
-
-private:
+  // Unary Operator
+  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal> &node) { return (std::get<0>(node) == node_index_); }
 
+ private:
   GlobalOrdinal node_index_;
-
 };
 
-
-//Definition of the predicate for the node_ structure.
-//Given a tuple made of node index and a specific region it belongs to,
-//this predicate returns true if the node has composite index which coincides with the index specified in input.
-//It does the same thing as checkerRegionToAll but it works on a different data structure
-template<class GlobalOrdinal>
+// Definition of the predicate for the node_ structure.
+// Given a tuple made of node index and a specific region it belongs to,
+// this predicate returns true if the node has composite index which coincides with the index specified in input.
+// It does the same thing as checkerRegionToAll but it works on a different data structure
+template <class GlobalOrdinal>
 class checkerAllToRegion {
+ public:
+  // Constructor
+  checkerAllToRegion(GlobalOrdinal node_index) { node_index_ = node_index; };
 
-public:
-
-  //Constructor
-  checkerAllToRegion( GlobalOrdinal node_index){node_index_ = node_index;};
-
-  //Unary Operator
-  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal > &node)
-  { return (std::get<1>(node) == node_index_); }
-
-private:
+  // Unary Operator
+  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal> &node) { return (std::get<1>(node) == node_index_); }
 
+ private:
   GlobalOrdinal node_index_;
-
 };
 
-
-//Definition of the predicate for the node_ structure.
-//Given a tuple made of node index and a specific region it belongs to,
-//this predicate returns true if the node has composite index which coincides with the index specified in input.
-//This checker is specifically used only for nodes lying on the interface
-template<class GlobalOrdinal>
+// Definition of the predicate for the node_ structure.
+// Given a tuple made of node index and a specific region it belongs to,
+// this predicate returns true if the node has composite index which coincides with the index specified in input.
+// This checker is specifically used only for nodes lying on the interface
+template <class GlobalOrdinal>
 class checkerInterfaceNodes {
+ public:
+  // Constructor
+  checkerInterfaceNodes(GlobalOrdinal node_index) { node_index_ = node_index; };
 
-public:
-
-  //Constructor
-  checkerInterfaceNodes( GlobalOrdinal node_index){node_index_ = node_index;};
-
-  //Unary Operator
-  bool operator()(const std::tuple<int, Array<GlobalOrdinal> >  &node)
-  { return (std::get<0>(node) == node_index_); }
-
-private:
+  // Unary Operator
+  bool operator()(const std::tuple<int, Array<GlobalOrdinal> > &node) { return (std::get<0>(node) == node_index_); }
 
+ private:
   GlobalOrdinal node_index_;
-
 };
 
 /*!
-	@class Xpetra::MatrixSplitting class.
-	@brief Xpetra-specific matrix class.
+        @class Xpetra::MatrixSplitting class.
+        @brief Xpetra-specific matrix class.
 
-	This class is specific to Xpetra and has no analogue in Epetra or Tpetra.  The main motivation for this class is to be able to access matrix data in a manner different than how it is stored.
-	For example, it might be more convenient to treat ("view") a matrix stored in compressed row storage as if it were a block matrix.  The Xpetra::MatrixSplitting class is intended to manage these "views".
+        This class is specific to Xpetra and has no analogue in Epetra or Tpetra.  The main motivation for this class is to be able to access matrix data in a manner different than how it is stored.
+        For example, it might be more convenient to treat ("view") a matrix stored in compressed row storage as if it were a block matrix.  The Xpetra::MatrixSplitting class is intended to manage these "views".
 
-	<B>How to create a Matrix from an existing CrsMatrix</B>
+        <B>How to create a Matrix from an existing CrsMatrix</B>
 
  */
 
 typedef std::string viewLabel_t;
 
 template <class Scalar        = Operator<>::scalar_type,
-    class LocalOrdinal  = Operator<>::local_ordinal_type,
-    class GlobalOrdinal = typename Operator<LocalOrdinal>::global_ordinal_type,
-    class Node          = typename Operator<LocalOrdinal, GlobalOrdinal>::node_type,
-    UnderlyingLib lib   = Xpetra::UseEpetra,
-    bool collapse       = false>
-class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node > {
-
+          class LocalOrdinal  = Operator<>::local_ordinal_type,
+          class GlobalOrdinal = typename Operator<LocalOrdinal>::global_ordinal_type,
+          class Node          = typename Operator<LocalOrdinal, GlobalOrdinal>::node_type,
+          UnderlyingLib lib   = Xpetra::UseEpetra,
+          bool collapse       = false>
+class MatrixSplitting : public Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
   typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
-  typedef Xpetra::Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node > Matrix;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
   typedef Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> CrsGraph;
   typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
   typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixWrap;
   typedef Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixFactory;
   typedef Xpetra::MatrixView<Scalar, LocalOrdinal, GlobalOrdinal, Node> MatrixView;
 
-  //Xpetra structures must be converted into Tpetra specialized ones to construct an Ifpack2::OverlappingRowMatrix object
-  //Once the Ifpack2::OverlappingRowMatrix class is transferred into the Xpetra directory, the following 6 lines can be changed/removed
+  // Xpetra structures must be converted into Tpetra specialized ones to construct an Ifpack2::OverlappingRowMatrix object
+  // Once the Ifpack2::OverlappingRowMatrix class is transferred into the Xpetra directory, the following 6 lines can be changed/removed
   typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> tpetra_crs_matrix;
   typedef Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> tpetra_row_matrix;
 
-public:
-
+ public:
   //! @name Constructor/Destructor Methods
   //@{
 
   //! Constructor specifying fixed number of entries for each row.
-  MatrixSplitting(RCP<Matrix> matrix, RCP< Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > nodes)
-  {
-    std::cout<<"This version of MatrixSplitting constructor is NOT currently supported \n";
+  MatrixSplitting(RCP<Matrix> matrix, RCP<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > nodes) {
+    std::cout << "This version of MatrixSplitting constructor is NOT currently supported \n";
   }
   //
   //
-  MatrixSplitting(const char* matrix_file_name,
-      Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler,
-      RCP<const Teuchos::Comm<int> > comm
-      )
-  {
-    comm_ = comm;
-    regionHandler_ = regionHandler;
+  MatrixSplitting(const char *matrix_file_name,
+                  Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler,
+                  RCP<const Teuchos::Comm<int> > comm) {
+    comm_                            = comm;
+    regionHandler_                   = regionHandler;
     Array<GlobalOrdinal> elementlist = regionHandler_->GetGlobalRowMap();
-    num_total_elements_ = regionHandler_->GetNumGlobalElements();
-    num_total_regions_ = regionHandler_->GetNumTotalRegions();
+    num_total_elements_              = regionHandler_->GetNumGlobalElements();
+    num_total_regions_               = regionHandler_->GetNumTotalRegions();
 
-    if(comm_->getRank()==0)
-      std::cout<<"MatrixSplitting constructor initialized"<<std::endl;
+    if (comm_->getRank() == 0)
+      std::cout << "MatrixSplitting constructor initialized" << std::endl;
     region_matrix_initialized_.clear();
-    for( int i = 0; i<num_total_regions_; ++i )
+    for (int i = 0; i < num_total_regions_; ++i)
       region_matrix_initialized_.push_back(false);
 
-    //Create Xpetra map for composite stiffness matrix
-    if(comm_->getRank()==0)
-      std::cout<<"Starting construction of Composite Map"<<std::endl;
-    RCP<const Xpetra::Map<int,GlobalOrdinal,Node> > xpetraMap;
-    xpetraMap = Xpetra::MapFactory<int,GlobalOrdinal,Node>::Build(lib, num_total_elements_, elementlist, 0, comm);
-    if(comm_->getRank()==0)
-      std::cout<<"Finished construction of Composite Map"<<std::endl;
-
-    if(comm_->getRank()==0)
-      std::cout<<"Started reading composite matrix"<<std::endl;
-    //Import matrix from an .mm file into an Xpetra wrapper for an Epetra matrix
-    compositeMatrixData_ = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(matrix_file_name, xpetraMap);
-    if(comm_->getRank()==0)
-      std::cout<<"Finished reading composite matrix"<<std::endl;
-
-    CreateRegionMatrices( regionHandler_->GetRegionRowMaps() );
+    // Create Xpetra map for composite stiffness matrix
+    if (comm_->getRank() == 0)
+      std::cout << "Starting construction of Composite Map" << std::endl;
+    RCP<const Xpetra::Map<int, GlobalOrdinal, Node> > xpetraMap;
+    xpetraMap = Xpetra::MapFactory<int, GlobalOrdinal, Node>::Build(lib, num_total_elements_, elementlist, 0, comm);
+    if (comm_->getRank() == 0)
+      std::cout << "Finished construction of Composite Map" << std::endl;
+
+    if (comm_->getRank() == 0)
+      std::cout << "Started reading composite matrix" << std::endl;
+    // Import matrix from an .mm file into an Xpetra wrapper for an Epetra matrix
+    compositeMatrixData_ = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(matrix_file_name, xpetraMap);
+    if (comm_->getRank() == 0)
+      std::cout << "Finished reading composite matrix" << std::endl;
+
+    CreateRegionMatrices(regionHandler_->GetRegionRowMaps());
   }
 
   //! Destructor
-  virtual ~MatrixSplitting() { }
+  virtual ~MatrixSplitting() {}
   //@}
 
   //! @name Insertion/Removal Methods
@@ -263,16 +240,16 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
     \note If (globalRow,cols[i]) corresponds to an entry that is duplicated in this matrix row (likely because it was inserted more than once and fillComplete() has not been called in the interim), the behavior of this function is not defined. */
   void replaceGlobalValues(GlobalOrdinal globalRow,
-      const ArrayView<const GlobalOrdinal> &cols,
-      const ArrayView<const Scalar>        &vals) { compositeMatrixData_->replaceGlobalValues(globalRow, cols, vals); }
+                           const ArrayView<const GlobalOrdinal> &cols,
+                           const ArrayView<const Scalar> &vals) { compositeMatrixData_->replaceGlobalValues(globalRow, cols, vals); }
 
   //! Replace matrix entries, using local IDs.
   /** All index values must be in the local space.
         Note that if a value is not already present for the specified location in the matrix, the input value will be ignored silently.
    */
   void replaceLocalValues(LocalOrdinal localRow,
-      const ArrayView<const LocalOrdinal> &cols,
-      const ArrayView<const Scalar>       &vals) { compositeMatrixData_->replaceLocalValues(localRow, cols, vals); }
+                          const ArrayView<const LocalOrdinal> &cols,
+                          const ArrayView<const Scalar> &vals) { compositeMatrixData_->replaceLocalValues(localRow, cols, vals); }
 
   //! Set all matrix entries equal to scalar
   virtual void setAllToScalar(const Scalar &alpha) { compositeMatrixData_->setAllToScalar(alpha); }
@@ -288,27 +265,27 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   //@{
 
   /*! Resume fill operations.
-		  After calling fillComplete(), resumeFill() must be called before initiating any changes to the matrix.
+                  After calling fillComplete(), resumeFill() must be called before initiating any changes to the matrix.
 
-		  resumeFill() may be called repeatedly.
+                  resumeFill() may be called repeatedly.
 
-		  \post  <tt>isFillActive() == true<tt>
-		  \post  <tt>isFillComplete() == false<tt>
+                  \post  <tt>isFillActive() == true<tt>
+                  \post  <tt>isFillComplete() == false<tt>
    */
-  void resumeFill(const RCP< ParameterList > &params=null) {
+  void resumeFill(const RCP<ParameterList> &params = null) {
     compositeMatrixData_->resumeFill(params);
   }
 
   /*! \brief Signal that data entry is complete, specifying domain and range maps.
 
-		Off-node indices are distributed (via globalAssemble()), indices are sorted, redundant indices are eliminated, and global indices are transformed to local indices.
+                Off-node indices are distributed (via globalAssemble()), indices are sorted, redundant indices are eliminated, and global indices are transformed to local indices.
 
-		\pre  <tt>isFillActive() == true<tt>
-		\pre <tt>isFillComplete()() == false<tt>
+                \pre  <tt>isFillActive() == true<tt>
+                \pre <tt>isFillComplete()() == false<tt>
 
-		\post <tt>isFillActive() == false<tt>
-		\post <tt>isFillComplete() == true<tt>
-		\post if <tt>os == DoOptimizeStorage<tt>, then <tt>isStorageOptimized() == true</tt>
+                \post <tt>isFillActive() == false<tt>
+                \post <tt>isFillComplete() == true<tt>
+                \post if <tt>os == DoOptimizeStorage<tt>, then <tt>isStorageOptimized() == true</tt>
    */
   void fillComplete(const RCP<const Map> &domainMap, const RCP<const Map> &rangeMap, const RCP<Teuchos::ParameterList> &params = null) {
     compositeMatrixData_->fillComplete(domainMap, rangeMap, params);
@@ -319,18 +296,18 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   /*! \brief Signal that data entry is complete.
 
-		Off-node entries are distributed (via globalAssemble()), repeated entries are summed, and global indices are transformed to local indices.
+                Off-node entries are distributed (via globalAssemble()), repeated entries are summed, and global indices are transformed to local indices.
 
-		\note This method calls fillComplete( getRowMap(), getRowMap(), os ).
+                \note This method calls fillComplete( getRowMap(), getRowMap(), os ).
 
-		\pre  <tt>isFillActive() == true<tt>
-		\pre <tt>isFillComplete()() == false<tt>
+                \pre  <tt>isFillActive() == true<tt>
+                \pre <tt>isFillComplete()() == false<tt>
 
-		\post <tt>isFillActive() == false<tt>
-		\post <tt>isFillComplete() == true<tt>
-		\post if <tt>os == DoOptimizeStorage<tt>, then <tt>isStorageOptimized() == true</tt>
+                \post <tt>isFillActive() == false<tt>
+                \post <tt>isFillComplete() == true<tt>
+                \post if <tt>os == DoOptimizeStorage<tt>, then <tt>isStorageOptimized() == true</tt>
    */
-  //TODO : Get ride of "Tpetra"::OptimizeOption
+  // TODO : Get ride of "Tpetra"::OptimizeOption
   void fillComplete(const RCP<ParameterList> &params = null) {
     compositeMatrixData_->fillComplete(params);
 
@@ -374,7 +351,7 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   size_t getNumEntriesInLocalRow(LocalOrdinal localRow) const {
     return compositeMatrixData_->getNumEntriesInLocalRow(localRow);
   }
- 
+
   //! \brief Returns the maximum number of entries across all rows/columns on all nodes.
   /** Undefined if isFillActive().
    */
@@ -391,14 +368,13 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! \brief Returns the number of regions in the composite domain.
   global_size_t getNumRegions() const {
-    TEUCHOS_TEST_FOR_EXCEPTION( region_matrix_initialized_.size()==0 , Exceptions::RuntimeError, "Regions have not been initialized yet \n");
-    for( int i = 0; i<region_matrix_initialized_.size(); ++i )
-      TEUCHOS_TEST_FOR_EXCEPTION( !region_matrix_initialized_[i] , Exceptions::RuntimeError, "Region matrix for region "<< i+1 << "has not been initialized yet \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(region_matrix_initialized_.size() == 0, Exceptions::RuntimeError, "Regions have not been initialized yet \n");
+    for (int i = 0; i < region_matrix_initialized_.size(); ++i)
+      TEUCHOS_TEST_FOR_EXCEPTION(!region_matrix_initialized_[i], Exceptions::RuntimeError, "Region matrix for region " << i + 1 << "has not been initialized yet \n");
 
     return num_total_regions_;
   }
 
-
   //! \brief If matrix indices are in the local range, this function returns true. Otherwise, this function returns false. */
   bool isLocallyIndexed() const {
     return compositeMatrixData_->isLocallyIndexed();
@@ -416,34 +392,33 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! Extract a list of entries in a specified local row of the matrix. Put into storage allocated by calling routine.
   /*!
-		  \param LocalRow - (In) Local row number for which indices are desired.
-		  \param Indices - (Out) Local column indices corresponding to values.
-		  \param Values - (Out) Matrix values.
-		  \param NumIndices - (Out) Number of indices.
+                  \param LocalRow - (In) Local row number for which indices are desired.
+                  \param Indices - (Out) Local column indices corresponding to values.
+                  \param Values - (Out) Matrix values.
+                  \param NumIndices - (Out) Number of indices.
 
-		  Note: A std::runtime_error exception is thrown if either \c Indices or \c Values is not large enough to hold the data associated
-		  with row \c LocalRow. If \c LocalRow is not valid for this node, then \c Indices and \c Values are unchanged and \c NumIndices is
-		  returned as OrdinalTraits<size_t>::invalid().
+                  Note: A std::runtime_error exception is thrown if either \c Indices or \c Values is not large enough to hold the data associated
+                  with row \c LocalRow. If \c LocalRow is not valid for this node, then \c Indices and \c Values are unchanged and \c NumIndices is
+                  returned as OrdinalTraits<size_t>::invalid().
 
-		  \pre <tt>isLocallyIndexed()==true</tt> or <tt>hasColMap() == true</tt>
+                  \pre <tt>isLocallyIndexed()==true</tt> or <tt>hasColMap() == true</tt>
    */
   void getLocalRowCopy(LocalOrdinal LocalRow,
-      const ArrayView<LocalOrdinal> &Indices,
-      const ArrayView<Scalar> &Values,
-      size_t &NumEntries
-  ) const {
+                       const ArrayView<LocalOrdinal> &Indices,
+                       const ArrayView<Scalar> &Values,
+                       size_t &NumEntries) const {
     compositeMatrixData_->getLocalRowCopy(LocalRow, Indices, Values, NumEntries);
   }
 
   //! Extract a const, non-persisting view of global indices in a specified row of the matrix.
   /*!
-		  \param GlobalRow - (In) Global row number for which indices are desired.
-		  \param Indices   - (Out) Global column indices corresponding to values.
-		  \param Values    - (Out) Row values
-		  \pre <tt>isLocallyIndexed() == false</tt>
-		  \post <tt>indices.size() == getNumEntriesInGlobalRow(GlobalRow)</tt>
+                  \param GlobalRow - (In) Global row number for which indices are desired.
+                  \param Indices   - (Out) Global column indices corresponding to values.
+                  \param Values    - (Out) Row values
+                  \pre <tt>isLocallyIndexed() == false</tt>
+                  \post <tt>indices.size() == getNumEntriesInGlobalRow(GlobalRow)</tt>
 
-		  Note: If \c GlobalRow does not belong to this node, then \c indices is set to null.
+                  Note: If \c GlobalRow does not belong to this node, then \c indices is set to null.
    */
   void getGlobalRowView(GlobalOrdinal GlobalRow, ArrayView<const GlobalOrdinal> &indices, ArrayView<const Scalar> &values) const {
     compositeMatrixData_->getGlobalRowView(GlobalRow, indices, values);
@@ -451,13 +426,13 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! Extract a const, non-persisting view of local indices in a specified row of the matrix.
   /*!
-		  \param LocalRow - (In) Local row number for which indices are desired.
-		  \param Indices  - (Out) Global column indices corresponding to values.
-		  \param Values   - (Out) Row values
-		  \pre <tt>isGloballyIndexed() == false</tt>
-		  \post <tt>indices.size() == getNumEntriesInLocalRow(LocalRow)</tt>
+                  \param LocalRow - (In) Local row number for which indices are desired.
+                  \param Indices  - (Out) Global column indices corresponding to values.
+                  \param Values   - (Out) Row values
+                  \pre <tt>isGloballyIndexed() == false</tt>
+                  \post <tt>indices.size() == getNumEntriesInLocalRow(LocalRow)</tt>
 
-		  Note: If \c LocalRow does not belong to this node, then \c indices is set to null.
+                  Note: If \c LocalRow does not belong to this node, then \c indices is set to null.
    */
   void getLocalRowView(LocalOrdinal LocalRow, ArrayView<const LocalOrdinal> &indices, ArrayView<const Scalar> &values) const {
     compositeMatrixData_->getLocalRowView(LocalRow, indices, values);
@@ -465,8 +440,8 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! \brief Get a copy of the diagonal entries owned by this node, with local row idices.
   /*! Returns a distributed Vector object partitioned according to this matrix's row map, containing the
-		  the zero and non-zero diagonals owned by this node. */
-  void getLocalDiagCopy(Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &diag) const {
+                  the zero and non-zero diagonals owned by this node. */
+  void getLocalDiagCopy(Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &diag) const {
     compositeMatrixData_->getLocalDiagCopy(diag);
   }
 
@@ -476,8 +451,8 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   }
 
   //! Get a copy of the diagonal entries owned by this node, with local row indices, using row offsets.
-  void getLocalDiagCopy(Xpetra::Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node > &diag, const ArrayView<const size_t> &offsets) const {
-    compositeMatrixData_->getLocalDiagCopy(diag,offsets);
+  void getLocalDiagCopy(Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &diag, const ArrayView<const size_t> &offsets) const {
+    compositeMatrixData_->getLocalDiagCopy(diag, offsets);
   }
 
   //! Get Frobenius norm of the matrix
@@ -486,12 +461,12 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   }
 
   //! Left scale matrix using the given vector entries
-  void leftScale (const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+  void leftScale(const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &x) {
     compositeMatrixData_->leftScale(x);
   }
 
   //! Neighbor2 scale matrix using the given vector entries
-  void rightScale (const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x) {
+  void rightScale(const Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &x) {
     compositeMatrixData_->rightScale(x);
   }
 
@@ -507,41 +482,40 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! \brief Computes the sparse matrix-multivector multiplication.
   /*! Performs \f$Y = \alpha A^{\textrm{mode}} X + \beta Y\f$, with one special exceptions:
-		- if <tt>beta == 0</tt>, apply() overwrites \c Y, so that any values in \c Y (including NaNs) are ignored.
+                - if <tt>beta == 0</tt>, apply() overwrites \c Y, so that any values in \c Y (including NaNs) are ignored.
    */
-  virtual void apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-      Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-      Teuchos::ETransp mode = Teuchos::NO_TRANS,
-      Scalar alpha = ScalarTraits<Scalar>::one(),
-      Scalar beta = ScalarTraits<Scalar>::zero()) const {
-
-    compositeMatrixData_->apply(X,Y,mode,alpha,beta);
+  virtual void apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &X,
+                     Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &Y,
+                     Teuchos::ETransp mode = Teuchos::NO_TRANS,
+                     Scalar alpha          = ScalarTraits<Scalar>::one(),
+                     Scalar beta           = ScalarTraits<Scalar>::zero()) const {
+    compositeMatrixData_->apply(X, Y, mode, alpha, beta);
   }
 
   //! \brief Returns the Map associated with the domain of this operator.
   //! This will be <tt>null</tt> until fillComplete() is called.
-  RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const {
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const {
     return compositeMatrixData_->getDomainMap();
   }
 
   //! Returns the Map associated with the domain of this operator.
   //! This will be <tt>null</tt> until fillComplete() is called.
-  RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const {
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const {
     return compositeMatrixData_->getRangeMap();
   }
 
   //! \brief Returns the Map that describes the column distribution in this matrix.
   //! This might be <tt>null</tt> until fillComplete() is called.
-  const RCP<const Map> & getColMap() const { return getColMap(Matrix::GetCurrentViewLabel()); }
+  const RCP<const Map> &getColMap() const { return getColMap(Matrix::GetCurrentViewLabel()); }
 
   //! \brief Returns the Map that describes the column distribution in this matrix.
-  const RCP<const Map> & getColMap(viewLabel_t viewLabel) const {
+  const RCP<const Map> &getColMap(viewLabel_t viewLabel) const {
     TEUCHOS_TEST_FOR_EXCEPTION(Matrix::operatorViewTable_.containsKey(viewLabel) == false, Xpetra::Exceptions::RuntimeError, "Xpetra::Matrix.GetColMap(): view '" + viewLabel + "' does not exist.");
-    updateDefaultView(); // If CrsMatrix::fillComplete() have been used instead of MatrixSplitting::fillComplete(), the default view is updated.
+    updateDefaultView();  // If CrsMatrix::fillComplete() have been used instead of MatrixSplitting::fillComplete(), the default view is updated.
     return Matrix::operatorViewTable_.get(viewLabel)->GetColMap();
   }
 
-  void removeEmptyProcessesInPlace(const RCP<const Map>& newMap) {
+  void removeEmptyProcessesInPlace(const RCP<const Map> &newMap) {
     compositeMatrixData_->removeEmptyProcessesInPlace(newMap);
     this->operatorViewTable_.get(this->GetCurrentViewLabel())->SetRowMap(compositeMatrixData_->getRowMap());
     this->operatorViewTable_.get(this->GetCurrentViewLabel())->SetColMap(compositeMatrixData_->getColMap());
@@ -553,45 +527,44 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   //{@
 
   //! Access function for the Tpetra::Map this DistObject was constructed with.
-  const RCP< const Xpetra::Map< LocalOrdinal, GlobalOrdinal, Node > > getMap() const {
+  const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getMap() const {
     return compositeMatrixData_->getMap();
   }
 
   //! Import.
   void doImport(const Matrix &source,
-      const Xpetra::Import< LocalOrdinal, GlobalOrdinal, Node > &importer, CombineMode CM) {
-    std::cout<<"Import not implemented"<<std::endl;
-    //const MatrixSplitting & sourceWrp = dynamic_cast<const MatrixSplitting &>(source);
-    //compositeMatrixData_->doImport(*sourceWrp.getCrsMatrix(), importer, CM);
+                const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> &importer, CombineMode CM) {
+    std::cout << "Import not implemented" << std::endl;
+    // const MatrixSplitting & sourceWrp = dynamic_cast<const MatrixSplitting &>(source);
+    // compositeMatrixData_->doImport(*sourceWrp.getCrsMatrix(), importer, CM);
   }
 
   //! Export.
   void doExport(const Matrix &dest,
-      const Xpetra::Import< LocalOrdinal, GlobalOrdinal, Node >& importer, CombineMode CM) {
-    std::cout<<"Export not implemented"<<std::endl;
-    //const MatrixSplitting & destWrp = dynamic_cast<const MatrixSplitting &>(dest);
-    //compositeMatrixData_->doExport(*destWrp.getCrsMatrix(), importer, CM);
+                const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> &importer, CombineMode CM) {
+    std::cout << "Export not implemented" << std::endl;
+    // const MatrixSplitting & destWrp = dynamic_cast<const MatrixSplitting &>(dest);
+    // compositeMatrixData_->doExport(*destWrp.getCrsMatrix(), importer, CM);
   }
 
   //! Import (using an Exporter).
   void doImport(const Matrix &source,
-      const Xpetra::Export< LocalOrdinal, GlobalOrdinal, Node >& exporter, CombineMode CM) {
-    std::cout<<"Import not implemented"<<std::endl;
-    //const MatrixSplitting & sourceWrp = dynamic_cast<const MatrixSplitting &>(source);
-    //compositeMatrixData_->doImport(*sourceWrp.getCrsMatrix(), exporter, CM);
+                const Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node> &exporter, CombineMode CM) {
+    std::cout << "Import not implemented" << std::endl;
+    // const MatrixSplitting & sourceWrp = dynamic_cast<const MatrixSplitting &>(source);
+    // compositeMatrixData_->doImport(*sourceWrp.getCrsMatrix(), exporter, CM);
   }
 
   //! Export (using an Importer).
   void doExport(const Matrix &dest,
-      const Xpetra::Export< LocalOrdinal, GlobalOrdinal, Node >& exporter, CombineMode CM) {
-    std::cout<<"Export not implemented"<<std::endl;
-    //const MatrixSplitting & destWrp = dynamic_cast<const MatrixSplitting &>(dest);
-    //compositeMatrixData_->doExport(*destWrp.getCrsMatrix(), exporter, CM);
+                const Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node> &exporter, CombineMode CM) {
+    std::cout << "Export not implemented" << std::endl;
+    // const MatrixSplitting & destWrp = dynamic_cast<const MatrixSplitting &>(dest);
+    // compositeMatrixData_->doExport(*destWrp.getCrsMatrix(), exporter, CM);
   }
 
   // @}
 
-
   //! @name Overridden from Teuchos::Describable
   //@{
 
@@ -601,8 +574,8 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   }
 
   /** \brief Print the object with some verbosity level to an FancyOStream object. */
-  void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel=Teuchos::Describable::verbLevel_default) const {
-    compositeMatrixData_->describe(out,verbLevel);
+  void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::Describable::verbLevel_default) const {
+    compositeMatrixData_->describe(out, verbLevel);
   }
   //@}
 
@@ -613,58 +586,50 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   RCP<const CrsGraph> getCrsGraph() const { return compositeMatrixData_->getCrsGraph(); }
 
   //! Returns an Xpetra::CrsMatrix pointer to the the composite matrix
-  RCP<CrsMatrix> getCrsMatrix() const {  return compositeMatrixData_; }
+  RCP<CrsMatrix> getCrsMatrix() const { return compositeMatrixData_; }
 
   //! Returns an Xpetra::Matrix pointer to the composite matrix
-  RCP<Matrix> getMatrix() const {  return compositeMatrixData_; }
+  RCP<Matrix> getMatrix() const { return compositeMatrixData_; }
 
   //! Returns an Xpetra::Matrix pointer to the region matrix associated with a specific region index
-  RCP<Matrix> getRegionMatrix( GlobalOrdinal region_idx ) const
-		        {
-    //The region index is assumed to start from 0
-    TEUCHOS_TEST_FOR_EXCEPTION( num_total_regions_<=0, Exceptions::RuntimeError, "Regions not initialized yet ( total number of regions is <=0 ) \n");
-    TEUCHOS_TEST_FOR_EXCEPTION( region_idx >= num_total_regions_, Exceptions::RuntimeError, "Region index not valid \n");
+  RCP<Matrix> getRegionMatrix(GlobalOrdinal region_idx) const {
+    // The region index is assumed to start from 0
+    TEUCHOS_TEST_FOR_EXCEPTION(num_total_regions_ <= 0, Exceptions::RuntimeError, "Regions not initialized yet ( total number of regions is <=0 ) \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(region_idx >= num_total_regions_, Exceptions::RuntimeError, "Region index not valid \n");
 
-    return regionMatrixData_[ region_idx ];
-		        }
+    return regionMatrixData_[region_idx];
+  }
 
   //! Return a ppointer to the underlying regionHandler object used for the matrix splitting
-  RCP<RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > getRegionHandler() const
-		        {
+  RCP<RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > getRegionHandler() const {
     return regionHandler_;
-		        }
+  }
   //@}
 
   //! Write methods
   //{@
-  void writeGlobalMatrix()
-  {
+  void writeGlobalMatrix() {
     std::string file_name;
     file_name += "./output/A_composite.mm";
-    Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write(file_name, *compositeMatrixData_);
+    Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(file_name, *compositeMatrixData_);
   }
 
-
-  void writeRegionMatrices()
-  {
-    for( int i = 0; i<num_total_regions_; ++i )
-    {
-      std::string region_str = std::to_string(i+1);
+  void writeRegionMatrices() {
+    for (int i = 0; i < num_total_regions_; ++i) {
+      std::string region_str = std::to_string(i + 1);
       std::string file_name;
       file_name += "./output/A_region_";
       file_name += region_str;
       file_name += ".mm";
-      Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write(file_name.c_str(), *regionMatrixData_[i]);
+      Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(file_name.c_str(), *regionMatrixData_[i]);
     }
   }
   // @}
 
-private:
-
+ private:
   // Default view is created after fillComplete()
   // Because ColMap might not be available before fillComplete().
   void CreateDefaultView() {
-
     // Create default view
     this->defaultViewLabel_ = "point";
     this->CreateView(this->GetDefaultViewLabel(), compositeMatrixData_->getRowMap(), compositeMatrixData_->getColMap());
@@ -676,7 +641,7 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   // The colMap can be <tt>null</tt> until fillComplete() is called. The default view of the Matrix have to be updated when fillComplete() is called.
   // If CrsMatrix::fillComplete() have been used instead of MatrixSplitting::fillComplete(), the default view is updated when getColMap() is called.
   void updateDefaultView() const {
-    if ((finalDefaultView_ == false) &&  compositeMatrixData_->isFillComplete() ) {
+    if ((finalDefaultView_ == false) && compositeMatrixData_->isFillComplete()) {
       // Update default view with the colMap
       Matrix::operatorViewTable_.get(Matrix::GetDefaultViewLabel())->SetColMap(compositeMatrixData_->getColMap());
       finalDefaultView_ = true;
@@ -688,188 +653,176 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
   //! Creation of region matrices
   //@{
-  void RegionMatrix(GlobalOrdinal region_idx)
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION( num_total_regions_!=regionMatrixData_.size(), Exceptions::RuntimeError, "Number of regions does not match with the size of regionMatrixData_ structure \n");
+  void RegionMatrix(GlobalOrdinal region_idx) {
+    TEUCHOS_TEST_FOR_EXCEPTION(num_total_regions_ != regionMatrixData_.size(), Exceptions::RuntimeError, "Number of regions does not match with the size of regionMatrixData_ structure \n");
     RCP<Matrix> region_matrix = regionMatrixData_[region_idx];
 
-    RCP<tpetra_crs_matrix > tpetraGlobalMatrix = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(compositeMatrixData_);
+    RCP<tpetra_crs_matrix> tpetraGlobalMatrix = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(compositeMatrixData_);
     Ifpack2::OverlappingRowMatrix<tpetra_row_matrix> enlargedMatrix(tpetraGlobalMatrix, 2);
 
     region_matrix->resumeFill();
 
-    //Region matrices are initially built to be a chopped version of the composite matrix
-    InitializeRegionMatrices( region_idx, region_matrix, enlargedMatrix );
+    // Region matrices are initially built to be a chopped version of the composite matrix
+    InitializeRegionMatrices(region_idx, region_matrix, enlargedMatrix);
 
-    //If the template paramater is set to collapse by the user, then interface entries of the region matrix are modified to collapse
-    //information coming from adjacent regions. If the collapsing is not done, then the splitting is calculated
-    if( collapse )
-      RegionCollapse( region_idx, region_matrix, enlargedMatrix );
+    // If the template paramater is set to collapse by the user, then interface entries of the region matrix are modified to collapse
+    // information coming from adjacent regions. If the collapsing is not done, then the splitting is calculated
+    if (collapse)
+      RegionCollapse(region_idx, region_matrix, enlargedMatrix);
     else
-      RegionSplitting( region_idx, region_matrix, enlargedMatrix );
+      RegionSplitting(region_idx, region_matrix, enlargedMatrix);
 
     region_matrix->fillComplete();
   };
   // @}
 
-
   //! @name Initialization of Region matrices
   //@{
-  void InitializeRegionMatrices(GlobalOrdinal region_idx, RCP<Matrix>& region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix>& enlargedMatrix)
-  {
-
-    //Region matrices are initially built to be a chopped version of the composite matrix
-    TEUCHOS_TEST_FOR_EXCEPTION( region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "Surrogate region stiffness matrices are already initialized by chopping the composite stiffness matrix \n");
+  void InitializeRegionMatrices(GlobalOrdinal region_idx, RCP<Matrix> &region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix> &enlargedMatrix) {
+    // Region matrices are initially built to be a chopped version of the composite matrix
+    TEUCHOS_TEST_FOR_EXCEPTION(region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "Surrogate region stiffness matrices are already initialized by chopping the composite stiffness matrix \n");
 
-    Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >   regionToAll = regionHandler_->GetRegionToAll(region_idx);
+    Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > regionToAll = regionHandler_->GetRegionToAll(region_idx);
 
-    //THIS IS THE CORE OF THE PROBLEM WHERE ONE NEEDS TO POPULATE THE REGIONAL MATRICES BY ACCESSING ENTRIES OF THE GLOBAL MATRIX
+    // THIS IS THE CORE OF THE PROBLEM WHERE ONE NEEDS TO POPULATE THE REGIONAL MATRICES BY ACCESSING ENTRIES OF THE GLOBAL MATRIX
     //
-    ArrayView<const GlobalOrdinal> MyRegionElements =region_matrix->getRowMap()->getLocalElementList();
-    for( typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter!=MyRegionElements.end(); ++iter )
-    {
-      //Nodes are saved in data structures with 1 as base index
-      checkerRegionToAll<GlobalOrdinal> unaryPredicate(*iter+1);
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator;
-      composite_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
-      TEUCHOS_TEST_FOR_EXCEPTION( composite_iterator==regionToAll.end(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with region index: "<<*iter+1<<" is not in regionToAll["<<region_idx<<"]"<<"\n" );
-      GlobalOrdinal node_idx = std::get<1>( *composite_iterator );
-      LocalOrdinal node_local_idx = enlargedMatrix.getRowMap()->getLocalElement(node_idx-1);
+    ArrayView<const GlobalOrdinal> MyRegionElements = region_matrix->getRowMap()->getLocalElementList();
+    for (typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter != MyRegionElements.end(); ++iter) {
+      // Nodes are saved in data structures with 1 as base index
+      checkerRegionToAll<GlobalOrdinal> unaryPredicate(*iter + 1);
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator;
+      composite_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
+      TEUCHOS_TEST_FOR_EXCEPTION(composite_iterator == regionToAll.end(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                                   << " node with region index: " << *iter + 1 << " is not in regionToAll[" << region_idx << "]"
+                                                                                                                   << "\n");
+      GlobalOrdinal node_idx      = std::get<1>(*composite_iterator);
+      LocalOrdinal node_local_idx = enlargedMatrix.getRowMap()->getLocalElement(node_idx - 1);
 
       ArrayView<const GlobalOrdinal> inds;
       ArrayView<const Scalar> vals;
-      enlargedMatrix.getLocalRowView( node_local_idx, inds, vals );
+      enlargedMatrix.getLocalRowView(node_local_idx, inds, vals);
 
       std::vector<GlobalOrdinal> region_inds_vector(0);
-      std::vector<Scalar>	region_vals_vector(0);
+      std::vector<Scalar> region_vals_vector(0);
 
-      for( LocalOrdinal i = 0; i < inds.size(); ++i )
-      {
-        //Nodes are saved in data structures with 1 as base index
+      for (LocalOrdinal i = 0; i < inds.size(); ++i) {
+        // Nodes are saved in data structures with 1 as base index
         GlobalOrdinal composite_col_ind = enlargedMatrix.getColMap()->getGlobalElement(inds[i]) + 1;
         checkerAllToRegion<GlobalOrdinal> unaryPredicate2(composite_col_ind);
-        typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator region_iterator;
-        region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate2);
-        if( region_iterator!=regionToAll.end() )
-        {
-          region_inds_vector.push_back( std::get<0>(*region_iterator)-1 );
-          region_vals_vector.push_back( vals[i] );
+        typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator region_iterator;
+        region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate2);
+        if (region_iterator != regionToAll.end()) {
+          region_inds_vector.push_back(std::get<0>(*region_iterator) - 1);
+          region_vals_vector.push_back(vals[i]);
         }
       }
       ArrayView<GlobalOrdinal> region_inds(region_inds_vector);
       ArrayView<Scalar> region_vals(region_vals_vector);
-      region_matrix -> insertGlobalValues( *iter,region_inds,region_vals );
+      region_matrix->insertGlobalValues(*iter, region_inds, region_vals);
     }
     region_matrix_initialized_[region_idx] = true;
   }
   //@}
 
-
   //! @name Collapse of External neighbouring nodes information on Region matrices
   //@{
-  void RegionCollapse(GlobalOrdinal region_idx, RCP<Matrix>& region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix>& enlargedMatrix)
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION( !region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "The composite stiffness matrix must be chopped into surrogate region matrices before collapsing \n");
-    TEUCHOS_TEST_FOR_EXCEPTION( regionHandler_->GetNumRegionNodes(region_idx)!=regionMatrixData_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Number of region nodes in region "<<region_idx+1<<" does not coincide with the value returned by regionMatrixData_["<<region_idx+1<<"]->getGlobalNumRows() \n");
+  void RegionCollapse(GlobalOrdinal region_idx, RCP<Matrix> &region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix> &enlargedMatrix) {
+    TEUCHOS_TEST_FOR_EXCEPTION(!region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "The composite stiffness matrix must be chopped into surrogate region matrices before collapsing \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(regionHandler_->GetNumRegionNodes(region_idx) != regionMatrixData_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Number of region nodes in region " << region_idx + 1 << " does not coincide with the value returned by regionMatrixData_[" << region_idx + 1 << "]->getGlobalNumRows() \n");
 
-    Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >   regionToAll = regionHandler_->GetRegionToAll(region_idx);
+    Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > regionToAll = regionHandler_->GetRegionToAll(region_idx);
 
-    //This portion of the code assumes that the number of region nodes is the same on each direction of the domain
-    //Foir a 2D problem we have then nx = ny = sqrt( num_region_ndoes_ )
+    // This portion of the code assumes that the number of region nodes is the same on each direction of the domain
+    // Foir a 2D problem we have then nx = ny = sqrt( num_region_ndoes_ )
     GlobalOrdinal n;
     GlobalOrdinal nx;
     GlobalOrdinal ny;
 
-    n = regionHandler_->GetNumRegionNodes(region_idx);
+    n  = regionHandler_->GetNumRegionNodes(region_idx);
     nx = std::sqrt(n);
     ny = nx;
-    TEUCHOS_TEST_FOR_EXCEPTION( static_cast<double>( nx - std::floor(static_cast<double>( std::sqrt(static_cast<double>(n)) )))!=0.0 , Exceptions::RuntimeError, "The code assumes that the regions are 2D and that the number of region nodes is the same on each direction of the domain \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(static_cast<double>(nx - std::floor(static_cast<double>(std::sqrt(static_cast<double>(n))))) != 0.0, Exceptions::RuntimeError, "The code assumes that the regions are 2D and that the number of region nodes is the same on each direction of the domain \n");
 
-    //interfaceNodes contains nodes on an interface between any regions
+    // interfaceNodes contains nodes on an interface between any regions
     Array<std::tuple<int, Array<GlobalOrdinal> > > interfaceNodes = regionHandler_->GetInterfaceNodes();
 
-    ArrayView<const GlobalOrdinal> MyRegionElements =region_matrix->getRowMap()->getLocalElementList();
-    for( typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter!=MyRegionElements.end(); ++iter )
-    {
-
-      //Nodes are saved in data structures with 1 as base index
-      GlobalOrdinal region_node_idx	= *iter+1;
+    ArrayView<const GlobalOrdinal> MyRegionElements = region_matrix->getRowMap()->getLocalElementList();
+    for (typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter != MyRegionElements.end(); ++iter) {
+      // Nodes are saved in data structures with 1 as base index
+      GlobalOrdinal region_node_idx = *iter + 1;
       checkerRegionToAll<GlobalOrdinal> unaryPredicate(region_node_idx);
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator;
-      composite_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
-      TEUCHOS_TEST_FOR_EXCEPTION( composite_iterator==regionToAll.end(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with region index: "<<region_node_idx<<" is not in regionToAll["<<region_idx<<"]"<<"\n" );
-
-      GlobalOrdinal composite_node_idx = std::get<1>( *composite_iterator );
-      checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2( composite_node_idx );
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator;
-      interface_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2);
-
-      //Here we assuming that a specific labeling choice is adopted region wise and we use it to distinguish coarse node from fine nodes
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator;
+      composite_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
+      TEUCHOS_TEST_FOR_EXCEPTION(composite_iterator == regionToAll.end(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                                   << " node with region index: " << region_node_idx << " is not in regionToAll[" << region_idx << "]"
+                                                                                                                   << "\n");
+
+      GlobalOrdinal composite_node_idx = std::get<1>(*composite_iterator);
+      checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2(composite_node_idx);
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator;
+      interface_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2);
+
+      // Here we assuming that a specific labeling choice is adopted region wise and we use it to distinguish coarse node from fine nodes
       bool coarse_point = false;
-      if( region_node_idx%3==1 )
+      if (region_node_idx % 3 == 1)
         coarse_point = true;
 
       GlobalOrdinal region_node_idx_neighbor1 = 0;
       GlobalOrdinal region_node_idx_neighbor2 = 0;
 
-      //Horizontal-Vertical Collapse
-      if( interface_iterator!=interfaceNodes.end() && region_node_idx>ny && region_node_idx<=(nx-1)*ny && !coarse_point )
-      {
-        region_node_idx_neighbor1 = region_node_idx-ny;
-        region_node_idx_neighbor2 = region_node_idx+ny;
-      }
-      else if( interface_iterator!=interfaceNodes.end() && region_node_idx%ny>1 && !coarse_point )
-      {
-        region_node_idx_neighbor1 = region_node_idx-1;
-        region_node_idx_neighbor2 = region_node_idx+1;
+      // Horizontal-Vertical Collapse
+      if (interface_iterator != interfaceNodes.end() && region_node_idx > ny && region_node_idx <= (nx - 1) * ny && !coarse_point) {
+        region_node_idx_neighbor1 = region_node_idx - ny;
+        region_node_idx_neighbor2 = region_node_idx + ny;
+      } else if (interface_iterator != interfaceNodes.end() && region_node_idx % ny > 1 && !coarse_point) {
+        region_node_idx_neighbor1 = region_node_idx - 1;
+        region_node_idx_neighbor2 = region_node_idx + 1;
       }
 
-      if( region_node_idx_neighbor1!=0 && region_node_idx_neighbor2!=0 )
-      {
-        //Computation of composite index for neighbor1 node
+      if (region_node_idx_neighbor1 != 0 && region_node_idx_neighbor2 != 0) {
+        // Computation of composite index for neighbor1 node
         checkerRegionToAll<GlobalOrdinal> unaryPredicateLeft(region_node_idx_neighbor1);
-        typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor1;
-        composite_iterator_neighbor1 = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateLeft);
+        typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor1;
+        composite_iterator_neighbor1 = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateLeft);
 
-        //Computation of composite index for neighbor2 node
+        // Computation of composite index for neighbor2 node
         checkerRegionToAll<GlobalOrdinal> unaryPredicateRight(region_node_idx_neighbor2);
-        typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor2;
-        composite_iterator_neighbor2 = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateRight);
-
-        TEUCHOS_TEST_FOR_EXCEPTION( composite_iterator_neighbor1 == regionToAll.end() || composite_iterator_neighbor2 == regionToAll.end(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with region index: "<<region_node_idx<<" lies on the interface between regions: "<<std::get<1>(*interface_iterator)<<" BUT has compositely mislabeled neighbouring nodes missing from regionToAll \n" );
-
-        //Check to see if neighbor1 node lies on a coarse line
-        GlobalOrdinal composite_node_idx_neighbor1 = std::get<1>( *composite_iterator_neighbor1 );
-        checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighbor1( composite_node_idx_neighbor1 );
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor1;
-        interface_iterator_neighbor1 = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighbor1);
-
-        //Check to see if neighbor2 node lies on a coarse line
-        GlobalOrdinal composite_node_idx_neighbor2 = std::get<1>( *composite_iterator_neighbor2 );
-        checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighbor2( composite_node_idx_neighbor2 );
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor2;
-        interface_iterator_neighbor2 = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighbor2);
-
-        //I apply the collapse only if the current node is a fine node which lies on a coarse line
-        //This means that the neighbor1 node and neighbor2 node must both lie on the coarse line as well
-        if( interface_iterator_neighbor1!=interfaceNodes.end() && interface_iterator_neighbor2!=interfaceNodes.end() )
-        {
-
-          //For each fine node on a horixontal coarse line on the interface, I extract the rows from the composite matrix
-          LocalOrdinal node_idx = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx-1);
-          LocalOrdinal node_idx_neighbor1 = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx_neighbor1-1);
-          LocalOrdinal node_idx_neighbor2 = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx_neighbor2-1);
+        typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor2;
+        composite_iterator_neighbor2 = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateRight);
+
+        TEUCHOS_TEST_FOR_EXCEPTION(composite_iterator_neighbor1 == regionToAll.end() || composite_iterator_neighbor2 == regionToAll.end(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                                                                                                    << " node with region index: " << region_node_idx << " lies on the interface between regions: " << std::get<1>(*interface_iterator) << " BUT has compositely mislabeled neighbouring nodes missing from regionToAll \n");
+
+        // Check to see if neighbor1 node lies on a coarse line
+        GlobalOrdinal composite_node_idx_neighbor1 = std::get<1>(*composite_iterator_neighbor1);
+        checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighbor1(composite_node_idx_neighbor1);
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor1;
+        interface_iterator_neighbor1 = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighbor1);
+
+        // Check to see if neighbor2 node lies on a coarse line
+        GlobalOrdinal composite_node_idx_neighbor2 = std::get<1>(*composite_iterator_neighbor2);
+        checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighbor2(composite_node_idx_neighbor2);
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor2;
+        interface_iterator_neighbor2 = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighbor2);
+
+        // I apply the collapse only if the current node is a fine node which lies on a coarse line
+        // This means that the neighbor1 node and neighbor2 node must both lie on the coarse line as well
+        if (interface_iterator_neighbor1 != interfaceNodes.end() && interface_iterator_neighbor2 != interfaceNodes.end()) {
+          // For each fine node on a horixontal coarse line on the interface, I extract the rows from the composite matrix
+          LocalOrdinal node_idx           = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx - 1);
+          LocalOrdinal node_idx_neighbor1 = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx_neighbor1 - 1);
+          LocalOrdinal node_idx_neighbor2 = enlargedMatrix.getRowMap()->getLocalElement(composite_node_idx_neighbor2 - 1);
           ArrayView<const LocalOrdinal> inds;
           ArrayView<const Scalar> vals;
-          enlargedMatrix.getLocalRowView( node_idx, inds, vals );
+          enlargedMatrix.getLocalRowView(node_idx, inds, vals);
           ArrayView<const LocalOrdinal> inds_neighbor1;
           ArrayView<const Scalar> vals_neighbor1;
-          enlargedMatrix.getLocalRowView( node_idx_neighbor1, inds_neighbor1, vals_neighbor1 );
+          enlargedMatrix.getLocalRowView(node_idx_neighbor1, inds_neighbor1, vals_neighbor1);
           ArrayView<const LocalOrdinal> inds_neighbor2;
           ArrayView<const Scalar> vals_neighbor2;
-          enlargedMatrix.getLocalRowView( node_idx_neighbor2, inds_neighbor2, vals_neighbor2 );
+          enlargedMatrix.getLocalRowView(node_idx_neighbor2, inds_neighbor2, vals_neighbor2);
 
-          std::vector<LocalOrdinal> inds_vector = createVector(inds);
+          std::vector<LocalOrdinal> inds_vector           = createVector(inds);
           std::vector<LocalOrdinal> inds_neighbor1_vector = createVector(inds_neighbor1);
           std::vector<LocalOrdinal> inds_neighbor2_vector = createVector(inds_neighbor2);
 
@@ -877,282 +830,273 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
           std::vector<GlobalOrdinal> composite_inds_neighbor1_vector(0);
           std::vector<GlobalOrdinal> composite_inds_neighbor2_vector(0);
 
-          for( typename std::vector<LocalOrdinal>::iterator iter_node = inds_vector.begin(); iter_node!=inds_vector.end(); ++iter_node )
-            composite_inds_vector.push_back( enlargedMatrix.getRowMap()->getGlobalElement(*iter_node) );
-          std::sort( composite_inds_vector.begin(), composite_inds_vector.end() );
+          for (typename std::vector<LocalOrdinal>::iterator iter_node = inds_vector.begin(); iter_node != inds_vector.end(); ++iter_node)
+            composite_inds_vector.push_back(enlargedMatrix.getRowMap()->getGlobalElement(*iter_node));
+          std::sort(composite_inds_vector.begin(), composite_inds_vector.end());
 
-          for( typename std::vector<LocalOrdinal>::iterator iter_node = inds_neighbor1_vector.begin(); iter_node!=inds_neighbor1_vector.end(); ++iter_node )
-            composite_inds_neighbor1_vector.push_back( enlargedMatrix.getRowMap()->getGlobalElement(*iter_node) );
+          for (typename std::vector<LocalOrdinal>::iterator iter_node = inds_neighbor1_vector.begin(); iter_node != inds_neighbor1_vector.end(); ++iter_node)
+            composite_inds_neighbor1_vector.push_back(enlargedMatrix.getRowMap()->getGlobalElement(*iter_node));
 
-          std::sort( composite_inds_neighbor1_vector.begin(), composite_inds_neighbor1_vector.end() );
+          std::sort(composite_inds_neighbor1_vector.begin(), composite_inds_neighbor1_vector.end());
 
-          for( typename std::vector<LocalOrdinal>::iterator iter_node = inds_neighbor2_vector.begin(); iter_node!=inds_neighbor2_vector.end(); ++iter_node )
-            composite_inds_neighbor2_vector.push_back( enlargedMatrix.getRowMap()->getGlobalElement(*iter_node) );
+          for (typename std::vector<LocalOrdinal>::iterator iter_node = inds_neighbor2_vector.begin(); iter_node != inds_neighbor2_vector.end(); ++iter_node)
+            composite_inds_neighbor2_vector.push_back(enlargedMatrix.getRowMap()->getGlobalElement(*iter_node));
 
-          std::sort( composite_inds_neighbor2_vector.begin(), composite_inds_neighbor2_vector.end() );
+          std::sort(composite_inds_neighbor2_vector.begin(), composite_inds_neighbor2_vector.end());
 
-          //IDENTIFICATION OF EXTERNAL NODES THROUGH COMPOSITE INDICES STARTS HERE
+          // IDENTIFICATION OF EXTERNAL NODES THROUGH COMPOSITE INDICES STARTS HERE
           std::vector<GlobalOrdinal> composite_node_idx_neighbor1_extra;
           std::vector<GlobalOrdinal> composite_node_idx_neighbor2_extra;
           std::vector<GlobalOrdinal> composite_node_idx_extra(composite_inds_vector);
 
-          //The follolwing triple of vector is expected to EVENTUALLY contain only one entry:
-          //the label of the external node with information to collapse close to neighbor1, neighbor2 and central node
+          // The follolwing triple of vector is expected to EVENTUALLY contain only one entry:
+          // the label of the external node with information to collapse close to neighbor1, neighbor2 and central node
           std::vector<GlobalOrdinal> diff_neighbor1;
           std::vector<GlobalOrdinal> diff_neighbor2;
           std::vector<GlobalOrdinal> diff_center;
 
-          //Identification of external node from the side of neighbor1
+          // Identification of external node from the side of neighbor1
           {
-            //Compute the intersection between neoghbourhood of neighbor1 node and neighbourhood of central node
+            // Compute the intersection between neoghbourhood of neighbor1 node and neighbourhood of central node
             std::set_intersection(composite_inds_vector.begin(), composite_inds_vector.end(), composite_inds_neighbor1_vector.begin(), composite_inds_neighbor1_vector.end(), std::back_inserter(composite_node_idx_neighbor1_extra));
-            for( typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_neighbor1_extra.begin(); iter_node!=composite_node_idx_neighbor1_extra.end(); ++iter_node )
-            {
-              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node+1);
-              typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator region_iterator_extra;
-              region_iterator_extra = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
-
-              //Invalidation of node indices for nodes belonging to the current region
-              if( region_iterator_extra!=regionToAll.end() )
+            for (typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_neighbor1_extra.begin(); iter_node != composite_node_idx_neighbor1_extra.end(); ++iter_node) {
+              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node + 1);
+              typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator region_iterator_extra;
+              region_iterator_extra = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
+
+              // Invalidation of node indices for nodes belonging to the current region
+              if (region_iterator_extra != regionToAll.end())
                 *iter_node = -1;
             }
 
-            //Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
-            composite_node_idx_neighbor1_extra.erase(std::remove(composite_node_idx_neighbor1_extra.begin(), composite_node_idx_neighbor1_extra.end(), -1),composite_node_idx_neighbor1_extra.end());
+            // Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
+            composite_node_idx_neighbor1_extra.erase(std::remove(composite_node_idx_neighbor1_extra.begin(), composite_node_idx_neighbor1_extra.end(), -1), composite_node_idx_neighbor1_extra.end());
 
-            //External node from neighbor1 side does not belong to the neighborhood of neighbor2
+            // External node from neighbor1 side does not belong to the neighborhood of neighbor2
             std::set_difference(composite_node_idx_neighbor1_extra.begin(), composite_node_idx_neighbor1_extra.end(), composite_inds_neighbor2_vector.begin(), composite_inds_neighbor2_vector.end(), std::inserter(diff_neighbor1, diff_neighbor1.begin()));
-            TEUCHOS_TEST_FOR_EXCEPTION( diff_neighbor1.size()!=1 , Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<"Mislabeling of nodes obstructed the identification of the extra node: region node "<< region_node_idx<<" leads to diff_neighbor1.size()= "<<diff_neighbor1.size()<<" \n");
+            TEUCHOS_TEST_FOR_EXCEPTION(diff_neighbor1.size() != 1, Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                            << "Mislabeling of nodes obstructed the identification of the extra node: region node " << region_node_idx << " leads to diff_neighbor1.size()= " << diff_neighbor1.size() << " \n");
           }
 
-          //Identification of external node from the side of neighbor2
+          // Identification of external node from the side of neighbor2
           {
-            //Compute the intersection between neighbourhood of neighbor2 node and neighbourhood of central node
+            // Compute the intersection between neighbourhood of neighbor2 node and neighbourhood of central node
             std::set_intersection(composite_inds_vector.begin(), composite_inds_vector.end(), composite_inds_neighbor2_vector.begin(), composite_inds_neighbor2_vector.end(), std::back_inserter(composite_node_idx_neighbor2_extra));
 
-            for( typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_neighbor2_extra.begin(); iter_node!=composite_node_idx_neighbor2_extra.end(); ++iter_node )
-            {
-              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node+1);
-              typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator region_iterator_extra;
-              region_iterator_extra = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
+            for (typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_neighbor2_extra.begin(); iter_node != composite_node_idx_neighbor2_extra.end(); ++iter_node) {
+              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node + 1);
+              typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator region_iterator_extra;
+              region_iterator_extra = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
 
-              //Invalidation of node indices for nodes belonging to the current region
-              if( region_iterator_extra!=regionToAll.end() )
+              // Invalidation of node indices for nodes belonging to the current region
+              if (region_iterator_extra != regionToAll.end())
                 *iter_node = -1;
             }
 
-            //Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
-            composite_node_idx_neighbor2_extra.erase(std::remove(composite_node_idx_neighbor2_extra.begin(), composite_node_idx_neighbor2_extra.end(), -1),composite_node_idx_neighbor2_extra.end());
+            // Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
+            composite_node_idx_neighbor2_extra.erase(std::remove(composite_node_idx_neighbor2_extra.begin(), composite_node_idx_neighbor2_extra.end(), -1), composite_node_idx_neighbor2_extra.end());
 
-            //External node from neighbor2 side does not belong to the neighborhood of neighbor1
+            // External node from neighbor2 side does not belong to the neighborhood of neighbor1
             std::set_difference(composite_node_idx_neighbor2_extra.begin(), composite_node_idx_neighbor2_extra.end(), composite_inds_neighbor1_vector.begin(), composite_inds_neighbor1_vector.end(), std::inserter(diff_neighbor2, diff_neighbor2.begin()));
-            TEUCHOS_TEST_FOR_EXCEPTION( diff_neighbor2.size()!=1 , Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<"Mislabeling of nodes obstructed the identification of the extra node: region node "<< region_node_idx<<" leads to diff_neighbor2.size()= "<<diff_neighbor2.size()<<" \n");
+            TEUCHOS_TEST_FOR_EXCEPTION(diff_neighbor2.size() != 1, Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                            << "Mislabeling of nodes obstructed the identification of the extra node: region node " << region_node_idx << " leads to diff_neighbor2.size()= " << diff_neighbor2.size() << " \n");
           }
 
-          //Identification of external node from the side of central node
+          // Identification of external node from the side of central node
           {
-            for( typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_extra.begin(); iter_node!=composite_node_idx_extra.end(); ++iter_node )
-            {
-              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node+1);
-              typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator region_iterator_extra;
-              region_iterator_extra = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
-
-              //Invalidation of node indices for nodes belonging to the current region
-              if( region_iterator_extra!=regionToAll.end() )
+            for (typename std::vector<GlobalOrdinal>::iterator iter_node = composite_node_idx_extra.begin(); iter_node != composite_node_idx_extra.end(); ++iter_node) {
+              checkerAllToRegion<GlobalOrdinal> unaryPredicateExtra(*iter_node + 1);
+              typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator region_iterator_extra;
+              region_iterator_extra = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerAllToRegion<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateExtra);
+
+              // Invalidation of node indices for nodes belonging to the current region
+              if (region_iterator_extra != regionToAll.end())
                 *iter_node = -1;
             }
 
-            //Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
-            composite_node_idx_extra.erase(std::remove(composite_node_idx_extra.begin(), composite_node_idx_extra.end(), -1),composite_node_idx_extra.end());
+            // Removal of invalidated indices associated with nodes belonging to current region: (external nodes do not belong to this region)
+            composite_node_idx_extra.erase(std::remove(composite_node_idx_extra.begin(), composite_node_idx_extra.end(), -1), composite_node_idx_extra.end());
             std::vector<GlobalOrdinal> diff_center_temp;
 
-            //At thie point composite_node_idx_extra contains indices of all the three external nodes: two of these must be removed since they are already tracked
-            //External nodes from neighbors1's and neighbor2's side must be removed
+            // At thie point composite_node_idx_extra contains indices of all the three external nodes: two of these must be removed since they are already tracked
+            // External nodes from neighbors1's and neighbor2's side must be removed
             std::set_difference(composite_node_idx_extra.begin(), composite_node_idx_extra.end(), diff_neighbor1.begin(), diff_neighbor1.end(), std::inserter(diff_center_temp, diff_center_temp.begin()));
             std::set_difference(diff_center_temp.begin(), diff_center_temp.end(), diff_neighbor2.begin(), diff_neighbor2.end(), std::inserter(diff_center, diff_center.begin()));
-            TEUCHOS_TEST_FOR_EXCEPTION( diff_center.size()!=1 , Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<"Mislabeling of nodes obstructed the identification of the extra node: region node "<< region_node_idx<<" leads to diff_center.size()= "<<diff_center.size()<<" \n");
+            TEUCHOS_TEST_FOR_EXCEPTION(diff_center.size() != 1, Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                         << "Mislabeling of nodes obstructed the identification of the extra node: region node " << region_node_idx << " leads to diff_center.size()= " << diff_center.size() << " \n");
           }
 
-          //Computation of local indices for central node and its neighbors
-          LocalOrdinal local_region_node_idx = region_matrix->getRowMap()->getLocalElement( region_node_idx );
-          LocalOrdinal local_region_node_idx_neighbor1 = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor1 );
-          LocalOrdinal local_region_node_idx_neighbor2 = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor2 );
+          // Computation of local indices for central node and its neighbors
+          LocalOrdinal local_region_node_idx           = region_matrix->getRowMap()->getLocalElement(region_node_idx);
+          LocalOrdinal local_region_node_idx_neighbor1 = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor1);
+          LocalOrdinal local_region_node_idx_neighbor2 = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor2);
 
-          //Computation of local indices for external nodes
-          LocalOrdinal local_extra_central = enlargedMatrix.getRowMap()->getLocalElement( diff_center[0] );
-          LocalOrdinal local_extra_neighbor1 = enlargedMatrix.getRowMap()->getLocalElement( diff_neighbor1[0] );
-          LocalOrdinal local_extra_neighbor2 = enlargedMatrix.getRowMap()->getLocalElement( diff_neighbor2[0] );
+          // Computation of local indices for external nodes
+          LocalOrdinal local_extra_central   = enlargedMatrix.getRowMap()->getLocalElement(diff_center[0]);
+          LocalOrdinal local_extra_neighbor1 = enlargedMatrix.getRowMap()->getLocalElement(diff_neighbor1[0]);
+          LocalOrdinal local_extra_neighbor2 = enlargedMatrix.getRowMap()->getLocalElement(diff_neighbor2[0]);
 
           ArrayView<const GlobalOrdinal> region_row;
           ArrayView<const GlobalOrdinal> region_col;
           ArrayView<const Scalar> region_val;
 
-          //Extract Row view of the region matrix before collapsing
-          if( region_matrix -> isLocallyIndexed() )
-            region_matrix -> getLocalRowView( local_region_node_idx, region_col, region_val );
+          // Extract Row view of the region matrix before collapsing
+          if (region_matrix->isLocallyIndexed())
+            region_matrix->getLocalRowView(local_region_node_idx, region_col, region_val);
           else
-            region_matrix -> getGlobalRowView( region_node_idx, region_col, region_val );
+            region_matrix->getGlobalRowView(region_node_idx, region_col, region_val);
 
-          //Extract Row of overlapped composite matrix to detect node with information to collapse
+          // Extract Row of overlapped composite matrix to detect node with information to collapse
           ArrayView<const GlobalOrdinal> external_row;
           ArrayView<const GlobalOrdinal> external_col;
           ArrayView<const Scalar> external_val;
-          enlargedMatrix.getLocalRowView( node_idx, external_col, external_val );
+          enlargedMatrix.getLocalRowView(node_idx, external_col, external_val);
 
-          //neighbor1 collapse
+          // neighbor1 collapse
           {
             Scalar initial_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view!=region_col.end(); ++iter_view  )
-            {
-              if( region_matrix -> isLocallyIndexed() )
-                if( *iter_view==local_region_node_idx_neighbor1 )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view != region_col.end(); ++iter_view) {
+              if (region_matrix->isLocallyIndexed())
+                if (*iter_view == local_region_node_idx_neighbor1)
                   initial_value = region_val[iter_view - region_col.begin()];
-              if( region_matrix -> isGloballyIndexed() )
-                if( *iter_view==region_node_idx_neighbor1 )
+              if (region_matrix->isGloballyIndexed())
+                if (*iter_view == region_node_idx_neighbor1)
                   initial_value = region_val[iter_view - region_col.begin()];
 
-              if(initial_value!=0)
+              if (initial_value != 0)
                 break;
             }
 
             Scalar external_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view!=external_col.end(); ++iter_view  )
-            {
-              if( *iter_view==local_extra_neighbor1 )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view != external_col.end(); ++iter_view) {
+              if (*iter_view == local_extra_neighbor1)
                 external_value = external_val[iter_view - external_col.begin()];
 
-              if(external_value!=0)
+              if (external_value != 0)
                 break;
             }
 
-            Scalar new_value = external_value;// new matrix entry generated with the collapsing
+            Scalar new_value = external_value;  // new matrix entry generated with the collapsing
             std::vector<GlobalOrdinal> new_entry_ind;
             std::vector<Scalar> new_entry_val;
-            new_entry_ind.push_back(region_node_idx_neighbor1-1);
+            new_entry_ind.push_back(region_node_idx_neighbor1 - 1);
             new_entry_val.push_back(new_value);
 
-            //If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
-            //See description of insertGlobalValues(...)
-            region_matrix -> insertGlobalValues( region_node_idx-1, new_entry_ind, new_entry_val );
+            // If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
+            // See description of insertGlobalValues(...)
+            region_matrix->insertGlobalValues(region_node_idx - 1, new_entry_ind, new_entry_val);
           }
-          //neighbor2 collapse
+          // neighbor2 collapse
           {
             Scalar initial_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view!=region_col.end(); ++iter_view  )
-            {
-              if( region_matrix -> isLocallyIndexed() )
-                if( *iter_view==local_region_node_idx_neighbor2 )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view != region_col.end(); ++iter_view) {
+              if (region_matrix->isLocallyIndexed())
+                if (*iter_view == local_region_node_idx_neighbor2)
                   initial_value = region_val[iter_view - region_col.begin()];
-              if( region_matrix -> isGloballyIndexed() )
-                if( *iter_view==region_node_idx_neighbor2 )
+              if (region_matrix->isGloballyIndexed())
+                if (*iter_view == region_node_idx_neighbor2)
                   initial_value = region_val[iter_view - region_col.begin()];
 
-              if(initial_value!=0)
+              if (initial_value != 0)
                 break;
             }
 
             Scalar external_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view!=external_col.end(); ++iter_view  )
-            {
-              if( *iter_view==local_extra_neighbor2 )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view != external_col.end(); ++iter_view) {
+              if (*iter_view == local_extra_neighbor2)
                 external_value = external_val[iter_view - external_col.begin()];
 
-              if(external_value!=0)
+              if (external_value != 0)
                 break;
             }
 
-            Scalar new_value = external_value;// new matrix entry generated with the collapsing
+            Scalar new_value = external_value;  // new matrix entry generated with the collapsing
             std::vector<GlobalOrdinal> new_entry_ind;
             std::vector<Scalar> new_entry_val;
-            new_entry_ind.push_back(region_node_idx_neighbor2-1);
+            new_entry_ind.push_back(region_node_idx_neighbor2 - 1);
             new_entry_val.push_back(new_value);
 
-            //If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
-            //See description of insertGlobalValues(...)
-            region_matrix -> insertGlobalValues( region_node_idx-1, new_entry_ind, new_entry_val );
+            // If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
+            // See description of insertGlobalValues(...)
+            region_matrix->insertGlobalValues(region_node_idx - 1, new_entry_ind, new_entry_val);
           }
-          //central node collapse
+          // central node collapse
           {
             Scalar initial_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view!=region_col.end(); ++iter_view  )
-            {
-              if( region_matrix -> isLocallyIndexed() )
-                if( *iter_view==local_region_node_idx )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = region_col.begin(); iter_view != region_col.end(); ++iter_view) {
+              if (region_matrix->isLocallyIndexed())
+                if (*iter_view == local_region_node_idx)
                   initial_value = region_val[iter_view - region_col.begin()];
-              if( region_matrix -> isGloballyIndexed() )
-                if( *iter_view==region_node_idx )
+              if (region_matrix->isGloballyIndexed())
+                if (*iter_view == region_node_idx)
                   initial_value = region_val[iter_view - region_col.begin()];
 
-              if(initial_value!=0)
+              if (initial_value != 0)
                 break;
             }
 
             Scalar external_value = 0;
-            for( typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view!=external_col.end(); ++iter_view  )
-            {
-              if( *iter_view==local_extra_central )
+            for (typename ArrayView<const GlobalOrdinal>::iterator iter_view = external_col.begin(); iter_view != external_col.end(); ++iter_view) {
+              if (*iter_view == local_extra_central)
                 external_value = external_val[iter_view - external_col.begin()];
 
-              if(external_value!=0)
+              if (external_value != 0)
                 break;
             }
 
-            Scalar new_value = external_value;// new matrix entry generated with the collapsing
+            Scalar new_value = external_value;  // new matrix entry generated with the collapsing
             std::vector<GlobalOrdinal> new_entry_ind;
             std::vector<Scalar> new_entry_val;
-            new_entry_ind.push_back(region_node_idx-1);
+            new_entry_ind.push_back(region_node_idx - 1);
             new_entry_val.push_back(new_value);
 
-            //If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
-            //See description of insertGlobalValues(...)
-            region_matrix -> insertGlobalValues( region_node_idx-1, new_entry_ind, new_entry_val );
+            // If a nonzero value is already stored in the specified position, the new values is SUMMED to the already existing one
+            // See description of insertGlobalValues(...)
+            region_matrix->insertGlobalValues(region_node_idx - 1, new_entry_ind, new_entry_val);
           }
         }
       }
-
     }
   }
   //@}
 
-
   //! @name Creation of Region Splitting
   //@{
-  void RegionSplitting(GlobalOrdinal region_idx, RCP<Matrix>& region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix>& enlargedMatrix)
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION( !region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "The composite stiffness matrix must be chopped into surrogate region matrices before collapsing \n");
-    TEUCHOS_TEST_FOR_EXCEPTION( regionHandler_->GetNumRegionNodes(region_idx)!=regionMatrixData_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Number of region nodes in region "<<region_idx+1<<" does not coincide with the value returned by regionMatrixData_["<<region_idx+1<<"]->getGlobalNumRows() \n");
+  void RegionSplitting(GlobalOrdinal region_idx, RCP<Matrix> &region_matrix, Ifpack2::OverlappingRowMatrix<tpetra_row_matrix> &enlargedMatrix) {
+    TEUCHOS_TEST_FOR_EXCEPTION(!region_matrix_initialized_[region_idx], Exceptions::RuntimeError, "The composite stiffness matrix must be chopped into surrogate region matrices before collapsing \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(regionHandler_->GetNumRegionNodes(region_idx) != regionMatrixData_[region_idx]->getGlobalNumRows(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Number of region nodes in region " << region_idx + 1 << " does not coincide with the value returned by regionMatrixData_[" << region_idx + 1 << "]->getGlobalNumRows() \n");
 
-    Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >   regionToAll = regionHandler_->GetRegionToAll(region_idx);
+    Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > regionToAll = regionHandler_->GetRegionToAll(region_idx);
 
-    //This portion of the code assumes that the number of region nodes is the same on each direction of the domain
-    //Foir a 2D problem we have then nx = ny = sqrt( num_region_ndoes_ )
+    // This portion of the code assumes that the number of region nodes is the same on each direction of the domain
+    // Foir a 2D problem we have then nx = ny = sqrt( num_region_ndoes_ )
     GlobalOrdinal n;
     GlobalOrdinal nx;
     GlobalOrdinal ny;
 
-    n = regionHandler_->GetNumRegionNodes(region_idx);
+    n  = regionHandler_->GetNumRegionNodes(region_idx);
     nx = std::sqrt(n);
     ny = nx;
-    TEUCHOS_TEST_FOR_EXCEPTION( static_cast<double>( nx - std::floor(static_cast<double>( std::sqrt(static_cast<double>(n)) )))!=0.0 , Exceptions::RuntimeError, "The code assumes that the regions are 2D and that the number of region nodes is the same on each direction of the domain \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(static_cast<double>(nx - std::floor(static_cast<double>(std::sqrt(static_cast<double>(n))))) != 0.0, Exceptions::RuntimeError, "The code assumes that the regions are 2D and that the number of region nodes is the same on each direction of the domain \n");
 
-    //interfaceNodes contains nodes on an interface between any regions
+    // interfaceNodes contains nodes on an interface between any regions
     Array<std::tuple<int, Array<GlobalOrdinal> > > interfaceNodes = regionHandler_->GetInterfaceNodes();
 
-    ArrayView<const GlobalOrdinal> MyRegionElements =region_matrix->getRowMap()->getLocalElementList();
-    for( typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter!=MyRegionElements.end(); ++iter )
-    {
-
-      //Nodes are saved in data structures with 1 as base index
-      GlobalOrdinal region_node_idx	= *iter+1;
+    ArrayView<const GlobalOrdinal> MyRegionElements = region_matrix->getRowMap()->getLocalElementList();
+    for (typename ArrayView<const GlobalOrdinal>::iterator iter = MyRegionElements.begin(); iter != MyRegionElements.end(); ++iter) {
+      // Nodes are saved in data structures with 1 as base index
+      GlobalOrdinal region_node_idx = *iter + 1;
       checkerRegionToAll<GlobalOrdinal> unaryPredicate(region_node_idx);
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator;
-      composite_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
-      TEUCHOS_TEST_FOR_EXCEPTION( composite_iterator==regionToAll.end(), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with region index: "<<region_node_idx<<" is not in regionToAll["<<region_idx<<"]"<<"\n" );
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator;
+      composite_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicate);
+      TEUCHOS_TEST_FOR_EXCEPTION(composite_iterator == regionToAll.end(), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                                   << " node with region index: " << region_node_idx << " is not in regionToAll[" << region_idx << "]"
+                                                                                                                   << "\n");
 
-      GlobalOrdinal composite_node_idx = std::get<1>( *composite_iterator );
-      checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2( composite_node_idx );
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator;
-      interface_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2);
+      GlobalOrdinal composite_node_idx = std::get<1>(*composite_iterator);
+      checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2(composite_node_idx);
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator;
+      interface_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2);
 
       GlobalOrdinal region_node_idx_neighbor_w = 0;
       GlobalOrdinal region_node_idx_neighbor_e = 0;
@@ -1161,287 +1105,252 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
 
       int count_neighbours = 0;
 
-      if( interface_iterator!=interfaceNodes.end() && region_node_idx>ny )
-      {
-        region_node_idx_neighbor_w = region_node_idx-ny;
+      if (interface_iterator != interfaceNodes.end() && region_node_idx > ny) {
+        region_node_idx_neighbor_w = region_node_idx - ny;
         count_neighbours++;
       }
 
-      if( interface_iterator!=interfaceNodes.end() && region_node_idx<=(nx-1)*ny )
-      {
-        region_node_idx_neighbor_e = region_node_idx+ny;
+      if (interface_iterator != interfaceNodes.end() && region_node_idx <= (nx - 1) * ny) {
+        region_node_idx_neighbor_e = region_node_idx + ny;
         count_neighbours++;
       }
 
-      if( interface_iterator!=interfaceNodes.end() && region_node_idx%ny!=1 )
-      {
-        region_node_idx_neighbor_s = region_node_idx-1;
+      if (interface_iterator != interfaceNodes.end() && region_node_idx % ny != 1) {
+        region_node_idx_neighbor_s = region_node_idx - 1;
         count_neighbours++;
       }
 
-      if( interface_iterator!=interfaceNodes.end() && region_node_idx%ny!=0 )
-      {
-        region_node_idx_neighbor_n= region_node_idx+1;
+      if (interface_iterator != interfaceNodes.end() && region_node_idx % ny != 0) {
+        region_node_idx_neighbor_n = region_node_idx + 1;
         count_neighbours++;
       }
 
-      bool interface_line = false;
+      bool interface_line   = false;
       bool interface_corner = false;
 
-      if( 3==count_neighbours )
+      if (3 == count_neighbours)
         interface_line = true;
-      else if( 2==count_neighbours )
+      else if (2 == count_neighbours)
         interface_corner = true;
 
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor_e;
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_e;
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor_w;
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_w;
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor_s;
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_s;
-      typename Array< std::tuple<GlobalOrdinal, GlobalOrdinal > >::iterator composite_iterator_neighbor_n;
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_n;
-
-      if( interface_line || interface_corner )
-      {
-        //Computation of composite index for East node
-        if( region_node_idx_neighbor_e!=0 )
-        {
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor_e;
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_e;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor_w;
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_w;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor_s;
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_s;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator composite_iterator_neighbor_n;
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator interface_iterator_neighbor_n;
+
+      if (interface_line || interface_corner) {
+        // Computation of composite index for East node
+        if (region_node_idx_neighbor_e != 0) {
           checkerRegionToAll<GlobalOrdinal> unaryPredicateEast(region_node_idx_neighbor_e);
-          composite_iterator_neighbor_e = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateEast);
+          composite_iterator_neighbor_e = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateEast);
 
-          //Check to see if neighbor_e node lies on a coarse line
-          GlobalOrdinal composite_node_idx_neighbor_e = std::get<1>( *composite_iterator_neighbor_e );
-          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborEast( composite_node_idx_neighbor_e );
-          interface_iterator_neighbor_e = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborEast);
-        }
-        else
+          // Check to see if neighbor_e node lies on a coarse line
+          GlobalOrdinal composite_node_idx_neighbor_e = std::get<1>(*composite_iterator_neighbor_e);
+          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborEast(composite_node_idx_neighbor_e);
+          interface_iterator_neighbor_e = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborEast);
+        } else
           interface_iterator_neighbor_e = interfaceNodes.end();
 
-        //Computation of composite index for West node
-        if( region_node_idx_neighbor_w!=0 )
-        {
+        // Computation of composite index for West node
+        if (region_node_idx_neighbor_w != 0) {
           checkerRegionToAll<GlobalOrdinal> unaryPredicateWest(region_node_idx_neighbor_w);
-          composite_iterator_neighbor_w = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateWest);
+          composite_iterator_neighbor_w = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateWest);
 
-          //Check to see if neighbor_w node lies on a coarse line
-          GlobalOrdinal composite_node_idx_neighbor_w = std::get<1>( *composite_iterator_neighbor_w );
-          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborWest( composite_node_idx_neighbor_w );
-          interface_iterator_neighbor_w = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborWest);
-        }
-        else
+          // Check to see if neighbor_w node lies on a coarse line
+          GlobalOrdinal composite_node_idx_neighbor_w = std::get<1>(*composite_iterator_neighbor_w);
+          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborWest(composite_node_idx_neighbor_w);
+          interface_iterator_neighbor_w = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborWest);
+        } else
           interface_iterator_neighbor_w = interfaceNodes.end();
 
-        //Computation of composite index for South node
-        if( region_node_idx_neighbor_s!=0 )
-        {
+        // Computation of composite index for South node
+        if (region_node_idx_neighbor_s != 0) {
           checkerRegionToAll<GlobalOrdinal> unaryPredicateSouth(region_node_idx_neighbor_s);
-          composite_iterator_neighbor_s = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateSouth);
+          composite_iterator_neighbor_s = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateSouth);
 
-          //Check to see if neighbor_s node lies on a coarse line
-          GlobalOrdinal composite_node_idx_neighbor_s = std::get<1>( *composite_iterator_neighbor_s );
-          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborSouth( composite_node_idx_neighbor_s );
-          interface_iterator_neighbor_s = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborSouth);
-        }
-        else
+          // Check to see if neighbor_s node lies on a coarse line
+          GlobalOrdinal composite_node_idx_neighbor_s = std::get<1>(*composite_iterator_neighbor_s);
+          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborSouth(composite_node_idx_neighbor_s);
+          interface_iterator_neighbor_s = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborSouth);
+        } else
           interface_iterator_neighbor_s = interfaceNodes.end();
 
-        //Computation of composite index for North node
-        if( region_node_idx_neighbor_n!=0 )
-        {
+        // Computation of composite index for North node
+        if (region_node_idx_neighbor_n != 0) {
           checkerRegionToAll<GlobalOrdinal> unaryPredicateNorth(region_node_idx_neighbor_n);
-          composite_iterator_neighbor_n = std::find_if<typename Array< std::tuple< GlobalOrdinal,GlobalOrdinal > >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateNorth);
+          composite_iterator_neighbor_n = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerRegionToAll<GlobalOrdinal> >(regionToAll.begin(), regionToAll.end(), unaryPredicateNorth);
 
-          //Check to see if neighbor_n node lies on a coarse line
-          GlobalOrdinal composite_node_idx_neighbor_n = std::get<1>( *composite_iterator_neighbor_n );
-          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborNorth( composite_node_idx_neighbor_n );
-          interface_iterator_neighbor_n = std::find_if<typename Array< std::tuple< GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborNorth);
-        }
-        else
+          // Check to see if neighbor_n node lies on a coarse line
+          GlobalOrdinal composite_node_idx_neighbor_n = std::get<1>(*composite_iterator_neighbor_n);
+          checkerInterfaceNodes<GlobalOrdinal> unaryPredicate2neighborNorth(composite_node_idx_neighbor_n);
+          interface_iterator_neighbor_n = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerInterfaceNodes<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicate2neighborNorth);
+        } else
           interface_iterator_neighbor_n = interfaceNodes.end();
 
         int count_neighbours_interface = 0;
-        if( interface_iterator_neighbor_e!=interfaceNodes.end() )
+        if (interface_iterator_neighbor_e != interfaceNodes.end())
           count_neighbours_interface++;
-        if( interface_iterator_neighbor_w!=interfaceNodes.end() )
+        if (interface_iterator_neighbor_w != interfaceNodes.end())
           count_neighbours_interface++;
-        if( interface_iterator_neighbor_s!=interfaceNodes.end() )
+        if (interface_iterator_neighbor_s != interfaceNodes.end())
           count_neighbours_interface++;
-        if( interface_iterator_neighbor_n!=interfaceNodes.end() )
+        if (interface_iterator_neighbor_n != interfaceNodes.end())
           count_neighbours_interface++;
 
-        TEUCHOS_TEST_FOR_EXCEPTION( count_neighbours_interface>count_neighbours, Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with region index: "<<region_node_idx<<" has inconsistent information on the number of neighbours: count_neighbours = "<<count_neighbours<<"but count_neighbours_interface ="<<count_neighbours_interface<<" \n" );
+        TEUCHOS_TEST_FOR_EXCEPTION(count_neighbours_interface > count_neighbours, Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                                           << " node with region index: " << region_node_idx << " has inconsistent information on the number of neighbours: count_neighbours = " << count_neighbours << "but count_neighbours_interface =" << count_neighbours_interface << " \n");
 
-        //First the splitting is applied on extradiagonal entries
+        // First the splitting is applied on extradiagonal entries
 
-        //Computation of local indices for central node and its neighbors
-        //Node index base start from 1 in the structures used, but Trilinos maps start from 0, so
-        //indices must be shifted by 1
-        LocalOrdinal local_region_node_idx = region_matrix->getRowMap()->getLocalElement( region_node_idx-1 );
-        LocalOrdinal local_region_node_idx_neighbor_e = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor_e-1 );
-        LocalOrdinal local_region_node_idx_neighbor_w = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor_w-1 );
-        LocalOrdinal local_region_node_idx_neighbor_s = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor_s-1 );
-        LocalOrdinal local_region_node_idx_neighbor_n = region_matrix->getRowMap()->getLocalElement( region_node_idx_neighbor_n-1 );
+        // Computation of local indices for central node and its neighbors
+        // Node index base start from 1 in the structures used, but Trilinos maps start from 0, so
+        // indices must be shifted by 1
+        LocalOrdinal local_region_node_idx            = region_matrix->getRowMap()->getLocalElement(region_node_idx - 1);
+        LocalOrdinal local_region_node_idx_neighbor_e = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor_e - 1);
+        LocalOrdinal local_region_node_idx_neighbor_w = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor_w - 1);
+        LocalOrdinal local_region_node_idx_neighbor_s = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor_s - 1);
+        LocalOrdinal local_region_node_idx_neighbor_n = region_matrix->getRowMap()->getLocalElement(region_node_idx_neighbor_n - 1);
 
         ArrayView<const GlobalOrdinal> region_col;
         ArrayView<const Scalar> region_val;
 
-        //Extract Row view of the region matrix
-        if( region_matrix -> isLocallyIndexed() )
-          region_matrix -> getLocalRowView( local_region_node_idx, region_col, region_val );
+        // Extract Row view of the region matrix
+        if (region_matrix->isLocallyIndexed())
+          region_matrix->getLocalRowView(local_region_node_idx, region_col, region_val);
         else
-          region_matrix -> getGlobalRowView( *iter, region_col, region_val );
+          region_matrix->getGlobalRowView(*iter, region_col, region_val);
 
         std::vector<LocalOrdinal> region_col_vector = createVector(region_col);
         std::vector<LocalOrdinal> ind_vector(0);
         std::vector<Scalar> val_vector(0);
 
-        //Extraction of the info about East neighbour to halve the associated entry in the matrix
-        if( interface_iterator_neighbor_e!=interfaceNodes.end() )
-        {
+        // Extraction of the info about East neighbour to halve the associated entry in the matrix
+        if (interface_iterator_neighbor_e != interfaceNodes.end()) {
           typename std::vector<GlobalOrdinal>::iterator iter_east_vector;
           GlobalOrdinal east_ind;
-          if( region_matrix -> isLocallyIndexed() )
-          {
-            iter_east_vector = std::find( region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_e );
-            east_ind = region_matrix->getRowMap()->getGlobalElement( *iter_east_vector );
+          if (region_matrix->isLocallyIndexed()) {
+            iter_east_vector = std::find(region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_e);
+            east_ind         = region_matrix->getRowMap()->getGlobalElement(*iter_east_vector);
+          } else {
+            iter_east_vector = std::find(region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_e - 1);
+            east_ind         = *iter_east_vector;
           }
-          else
-          {
-            iter_east_vector = std::find( region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_e-1 );
-            east_ind = *iter_east_vector;
-          }
-          Scalar east_val = - 0.5 * region_val[iter_east_vector-region_col_vector.begin()];
-          ind_vector.push_back( east_ind );
-          val_vector.push_back( east_val );
+          Scalar east_val = -0.5 * region_val[iter_east_vector - region_col_vector.begin()];
+          ind_vector.push_back(east_ind);
+          val_vector.push_back(east_val);
         }
 
-        //Extraction of the info about West neighbour to halve the associated entry in the matrix
-        if( interface_iterator_neighbor_w!=interfaceNodes.end() )
-        {
+        // Extraction of the info about West neighbour to halve the associated entry in the matrix
+        if (interface_iterator_neighbor_w != interfaceNodes.end()) {
           typename std::vector<GlobalOrdinal>::iterator iter_west_vector;
           GlobalOrdinal west_ind;
-          if( region_matrix -> isLocallyIndexed() )
-          {
-            iter_west_vector = std::find( region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_w );
-            west_ind = region_matrix->getRowMap()->getGlobalElement( *iter_west_vector );
+          if (region_matrix->isLocallyIndexed()) {
+            iter_west_vector = std::find(region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_w);
+            west_ind         = region_matrix->getRowMap()->getGlobalElement(*iter_west_vector);
+          } else {
+            iter_west_vector = std::find(region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_w - 1);
+            west_ind         = *iter_west_vector;
           }
-          else
-          {
-            iter_west_vector = std::find( region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_w-1 );
-            west_ind = *iter_west_vector;
-          }
-          Scalar west_val = - 0.5 * region_val[iter_west_vector-region_col_vector.begin()];
-          ind_vector.push_back( west_ind );
-          val_vector.push_back( west_val );
+          Scalar west_val = -0.5 * region_val[iter_west_vector - region_col_vector.begin()];
+          ind_vector.push_back(west_ind);
+          val_vector.push_back(west_val);
         }
 
-        //Extraction of the info about South neighbour to halve the associated entry in the matrix
-        if( interface_iterator_neighbor_s!=interfaceNodes.end() )
-        {
+        // Extraction of the info about South neighbour to halve the associated entry in the matrix
+        if (interface_iterator_neighbor_s != interfaceNodes.end()) {
           typename std::vector<GlobalOrdinal>::iterator iter_south_vector;
           GlobalOrdinal south_ind;
-          if( region_matrix -> isLocallyIndexed() )
-          {
-            iter_south_vector = std::find( region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_s );
-            south_ind = region_matrix->getRowMap()->getGlobalElement( *iter_south_vector );
+          if (region_matrix->isLocallyIndexed()) {
+            iter_south_vector = std::find(region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_s);
+            south_ind         = region_matrix->getRowMap()->getGlobalElement(*iter_south_vector);
+          } else {
+            iter_south_vector = std::find(region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_s - 1);
+            south_ind         = *iter_south_vector;
           }
-          else
-          {
-            iter_south_vector = std::find( region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_s-1 );
-            south_ind = *iter_south_vector;
-          }
-          Scalar south_val = - 0.5 * region_val[iter_south_vector-region_col_vector.begin()];
-          ind_vector.push_back( south_ind );
-          val_vector.push_back( south_val );
+          Scalar south_val = -0.5 * region_val[iter_south_vector - region_col_vector.begin()];
+          ind_vector.push_back(south_ind);
+          val_vector.push_back(south_val);
         }
 
-        //Extraction of the info about North neighbour to halve the associated entry in the matrix
-        if( interface_iterator_neighbor_n!=interfaceNodes.end() )
-        {
+        // Extraction of the info about North neighbour to halve the associated entry in the matrix
+        if (interface_iterator_neighbor_n != interfaceNodes.end()) {
           typename std::vector<GlobalOrdinal>::iterator iter_north_vector;
           GlobalOrdinal north_ind;
-          if( region_matrix -> isLocallyIndexed() )
-          {
-            iter_north_vector = std::find( region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_n );
-            north_ind = region_matrix->getRowMap()->getGlobalElement( *iter_north_vector );
+          if (region_matrix->isLocallyIndexed()) {
+            iter_north_vector = std::find(region_col_vector.begin(), region_col_vector.end(), local_region_node_idx_neighbor_n);
+            north_ind         = region_matrix->getRowMap()->getGlobalElement(*iter_north_vector);
+          } else {
+            iter_north_vector = std::find(region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_n - 1);
+            north_ind         = *iter_north_vector;
           }
-          else
-          {
-            iter_north_vector = std::find( region_col_vector.begin(), region_col_vector.end(), region_node_idx_neighbor_n-1 );
-            north_ind = *iter_north_vector;
-          }
-          Scalar north_val = - 0.5 * region_val[iter_north_vector-region_col_vector.begin()];
-          ind_vector.push_back( north_ind );
-          val_vector.push_back( north_val );
+          Scalar north_val = -0.5 * region_val[iter_north_vector - region_col_vector.begin()];
+          ind_vector.push_back(north_ind);
+          val_vector.push_back(north_val);
         }
 
-        //Extraction of the info about my Node ID to split the associated entry in the matrix
-        //The ratio used for the splitting depends on the num ber of regions this current node
-        //belongs to
+        // Extraction of the info about my Node ID to split the associated entry in the matrix
+        // The ratio used for the splitting depends on the num ber of regions this current node
+        // belongs to
         typename std::vector<LocalOrdinal>::iterator iter_center_vector;
         GlobalOrdinal center_ind;
-        if( region_matrix -> isLocallyIndexed() )
-        {
-          iter_center_vector = std::find( region_col_vector.begin(), region_col_vector.end(), local_region_node_idx );
-          center_ind = region_matrix->getRowMap()->getGlobalElement( *iter_center_vector );
-        }
-        else
-        {
-          iter_center_vector = std::find( region_col_vector.begin(), region_col_vector.end(), region_node_idx-1 );
-          center_ind = *iter_center_vector;
+        if (region_matrix->isLocallyIndexed()) {
+          iter_center_vector = std::find(region_col_vector.begin(), region_col_vector.end(), local_region_node_idx);
+          center_ind         = region_matrix->getRowMap()->getGlobalElement(*iter_center_vector);
+        } else {
+          iter_center_vector = std::find(region_col_vector.begin(), region_col_vector.end(), region_node_idx - 1);
+          center_ind         = *iter_center_vector;
         }
 
-        //Count of the nubmer of regions the current node belogns to
+        // Count of the nubmer of regions the current node belogns to
         GlobalOrdinal region_belonging = std::get<1>(*interface_iterator).size();
-        TEUCHOS_TEST_FOR_EXCEPTION( region_belonging<2, Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Region: "<<region_idx<<" - "<<" node with composite index: "<<std::get<0>(*interface_iterator)<<" should lie on an interface between regions but the nubmer of regions it belongs to is only "<<region_belonging<<"\n" );
+        TEUCHOS_TEST_FOR_EXCEPTION(region_belonging < 2, Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Region: " << region_idx << " - "
+                                                                                                  << " node with composite index: " << std::get<0>(*interface_iterator) << " should lie on an interface between regions but the nubmer of regions it belongs to is only " << region_belonging << "\n");
 
         Scalar center_val;
-        //If a node is on a corner between four itnerfaces, then each the entry A(node_idx,node_idx) must be split in four parts
-        //otherwise the entry must be divided by two, similarly to what done for the neighbours
+        // If a node is on a corner between four itnerfaces, then each the entry A(node_idx,node_idx) must be split in four parts
+        // otherwise the entry must be divided by two, similarly to what done for the neighbours
 
-        center_val = - (1 - static_cast<Scalar>(1/static_cast<Scalar>(region_belonging))) * region_val[iter_center_vector-region_col_vector.begin()];
-        ind_vector.push_back( center_ind );
-        val_vector.push_back( center_val );
-
-        region_matrix -> insertGlobalValues( region_node_idx-1, ind_vector, val_vector );
+        center_val = -(1 - static_cast<Scalar>(1 / static_cast<Scalar>(region_belonging))) * region_val[iter_center_vector - region_col_vector.begin()];
+        ind_vector.push_back(center_ind);
+        val_vector.push_back(center_val);
 
+        region_matrix->insertGlobalValues(region_node_idx - 1, ind_vector, val_vector);
       }
     }
   }
   //@}
 
-
   //! @name Creation of Region matrices
   //@{
-  void CreateRegionMatrices( Array<Array<GlobalOrdinal> > region_maps){
+  void CreateRegionMatrices(Array<Array<GlobalOrdinal> > region_maps) {
+    TEUCHOS_TEST_FOR_EXCEPTION(num_total_regions_ != region_maps.size(), Exceptions::RuntimeError, "Number of regions does not match with the size of region_maps structure \n");
 
-    TEUCHOS_TEST_FOR_EXCEPTION( num_total_regions_!=region_maps.size(), Exceptions::RuntimeError, "Number of regions does not match with the size of region_maps structure \n");
+    regionMatrixData_.clear();
 
-    regionMatrixData_.clear( );
-
-    for( int i = 0; i<num_total_regions_; ++i )
-    {
-      //Create Xpetra map for region stiffness matrix
-      RCP<const Xpetra::Map<int,GlobalOrdinal,Node> > xpetraMap;
-      xpetraMap = Xpetra::MapFactory<int,GlobalOrdinal,Node>::Build(lib, regionHandler_->GetNumRegionNodes(i), region_maps[i], 0, comm_);
+    for (int i = 0; i < num_total_regions_; ++i) {
+      // Create Xpetra map for region stiffness matrix
+      RCP<const Xpetra::Map<int, GlobalOrdinal, Node> > xpetraMap;
+      xpetraMap        = Xpetra::MapFactory<int, GlobalOrdinal, Node>::Build(lib, regionHandler_->GetNumRegionNodes(i), region_maps[i], 0, comm_);
       int num_elements = xpetraMap->getGlobalNumElements();
 
       RCP<CrsMatrix> crs_matrix;
-      if( Xpetra::UseEpetra==lib )
-        crs_matrix = rcp( new EpetraCrsMatrix(xpetraMap, num_elements) );
-      else if( Xpetra::UseTpetra==lib )
-        crs_matrix = rcp( new TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>( xpetraMap, num_elements ) );
+      if (Xpetra::UseEpetra == lib)
+        crs_matrix = rcp(new EpetraCrsMatrix(xpetraMap, num_elements));
+      else if (Xpetra::UseTpetra == lib)
+        crs_matrix = rcp(new TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xpetraMap, num_elements));
       else
-        std::cerr<<" The library to build matrices must be either Epetra or Tpetra \n";
+        std::cerr << " The library to build matrices must be either Epetra or Tpetra \n";
 
       RCP<Matrix> matrixPointer = rcp(new CrsMatrixWrap(crs_matrix));
-      regionMatrixData_.push_back( matrixPointer );
+      regionMatrixData_.push_back(matrixPointer);
     }
 
-    for( GlobalOrdinal i = 0; i<num_total_regions_; ++i )
+    for (GlobalOrdinal i = 0; i < num_total_regions_; ++i)
       RegionMatrix(i);
-
   };
   //@}
 
@@ -1468,12 +1377,12 @@ class MatrixSplitting : public Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node
   Array<RCP<Matrix> > regionMatrixData_;
 
   GlobalOrdinal num_total_elements_ = 0;
-  GlobalOrdinal num_total_regions_ = 0;
+  GlobalOrdinal num_total_regions_  = 0;
 
   //@}
 
-}; //class MatrixSplitting
+};  // class MatrixSplitting
 
-} //namespace Xpetra
+}  // namespace Xpetra
 
-#endif //XPETRA_MATRIXSPLITTING_HPP
+#endif  // XPETRA_MATRIXSPLITTING_HPP
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_decl.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_decl.hpp
index 555feb9e754e..4b5a526a58ff 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_decl.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_decl.hpp
@@ -53,19 +53,18 @@
 #ifndef XPETRA_REGIONAMG_DECL_HPP
 #define XPETRA_REGIONAMG_DECL_HPP
 
-//Xpetra
+// Xpetra
 #include <Xpetra_Operator.hpp>
 #include "Xpetra_MatrixSplitting.hpp"
 #include "Xpetra_Level_def.hpp"
 
-//MueLu
+// MueLu
 #include <MueLu.hpp>
 #include <MueLu_TpetraOperator.hpp>
 #include <MueLu_CreateTpetraPreconditioner.hpp>
 #include <MueLu_Utilities.hpp>
 
-
-namespace Xpetra{
+namespace Xpetra {
 
 /*!
  * class RegionAMG
@@ -80,121 +79,116 @@ namespace Xpetra{
  * we explains here below the rationale behind the approach adopted in this class.
  *  */
 
-template<class Scalar					= MultiVector<>::scalar_type,
-    class LocalOrdinal   				= typename MultiVector<Scalar>::local_ordinal_type,
-    class GlobalOrdinal  				= typename MultiVector<Scalar, LocalOrdinal>::global_ordinal_type,
-    class Node           				= typename MultiVector<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+template <class Scalar        = MultiVector<>::scalar_type,
+          class LocalOrdinal  = typename MultiVector<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename MultiVector<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename MultiVector<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
 class RegionAMG : Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-
   typedef Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
   typedef MultiVector<Scalar, LocalOrdinal, GlobalOrdinal> multivector_type;
-  typedef Matrix< Scalar, LocalOrdinal, GlobalOrdinal, Node > matrix_type;
-  typedef MatrixSplitting<Scalar,LocalOrdinal,GlobalOrdinal,Node,Xpetra::UseTpetra, false> tpetra_splitting;
-  typedef MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> Hierarchy;
+  typedef Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> matrix_type;
+  typedef MatrixSplitting<Scalar, LocalOrdinal, GlobalOrdinal, Node, Xpetra::UseTpetra, false> tpetra_splitting;
+  typedef MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> Hierarchy;
   typedef MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> mv_factory_type;
-  typedef Level<Scalar,LocalOrdinal,GlobalOrdinal,Node> level;
-
-public:
+  typedef Level<Scalar, LocalOrdinal, GlobalOrdinal, Node> level;
 
+ public:
   //! Constructors
   //@{
 
-  RegionAMG()
-{
-    std::cout<<"This version of constructor is not implemented yet"<<std::endl;
-};
+  RegionAMG() {
+    std::cout << "This version of constructor is not implemented yet" << std::endl;
+  };
 
   //! This is the actual constructor we are interested in calling
   RegionAMG(Teuchos::RCP<tpetra_splitting> matrixSplitting,
-      Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler,
-      RCP<const Teuchos::Comm<int> > comm, Teuchos::ParameterList muelu,
-      GlobalOrdinal num_levels, GlobalOrdinal coarsening_factor);
+            Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler,
+            RCP<const Teuchos::Comm<int> > comm, Teuchos::ParameterList muelu,
+            GlobalOrdinal num_levels, GlobalOrdinal coarsening_factor);
 
   //@}
 
-  GlobalOrdinal GetNumLevels(){return num_levels_;}
+  GlobalOrdinal GetNumLevels() { return num_levels_; }
 
   //! Methods to extract Map information
   //@{
   //! For now, the domain Map coincides with the Domain Map of the composite matrix at the fine level
-  virtual RCP<const map_type > getDomainMap()const{return domainMap_;}
+  virtual RCP<const map_type> getDomainMap() const { return domainMap_; }
 
   //! For now, the domain Map coincides with the Range Map of the composite matrix at the fine level
-  virtual RCP<const map_type > getRangeMap()const{return rangeMap_;}
+  virtual RCP<const map_type> getRangeMap() const { return rangeMap_; }
 
   //@}
 
   //! Apply method
   //@{
-  //!N.B.: The implementation still has to be finished
+  //! N.B.: The implementation still has to be finished
   virtual void
-  apply (const multivector_type& X, multivector_type& Y,
-      Teuchos::ETransp mode = Teuchos::NO_TRANS,
-      Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-      Scalar beta = Teuchos::ScalarTraits<Scalar>::zero())const;
+  apply(const multivector_type& X, multivector_type& Y,
+        Teuchos::ETransp mode = Teuchos::NO_TRANS,
+        Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+        Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
   //@}
 
   //! hasTransposeApply should not be needed
   virtual bool hasTransposeApply() const { return false; }
 
-private:
-
+ private:
   //! Private variables
   //@{
 
-  //Total number of levels in the hierarchy
+  // Total number of levels in the hierarchy
   GlobalOrdinal num_levels_ = -1;
 
-  //Total number of regions
+  // Total number of regions
   GlobalOrdinal num_regions_ = -1;
 
-  //Coarsening factor to transfer quantities across levels
+  // Coarsening factor to transfer quantities across levels
   int coarsening_factor_ = -1;
 
   RCP<const Teuchos::Comm<int> > comm_;
 
   Teuchos::ParameterList muelu_;
 
-  RCP<const map_type > domainMap_;
-  RCP<const map_type > rangeMap_;
+  RCP<const map_type> domainMap_;
+  RCP<const map_type> rangeMap_;
 
-  //matrixSplitting associated with the composite matrix at the fine level
+  // matrixSplitting associated with the composite matrix at the fine level
   RCP<tpetra_splitting> matrixSplitting_;
 
-  //Array of MueLu hierarchies (one for each region)
+  // Array of MueLu hierarchies (one for each region)
   Array<RCP<Hierarchy> > regionHierarchies_;
 
-  //Array of levels in the new hierarchy (each levels contains quantities associaed with every region for that level)
+  // Array of levels in the new hierarchy (each levels contains quantities associaed with every region for that level)
   Array<RCP<level> > levels_;
 
   //@}
 
   //@Private Methods
   //
-  //This methods construct the Hierarchy using the quantities stored in the MueLu::Hierarchy objects
-  //There are as many MueLu::Hierarchy objects as the number of geometric regions partitioning the domain
+  // This methods construct the Hierarchy using the quantities stored in the MueLu::Hierarchy objects
+  // There are as many MueLu::Hierarchy objects as the number of geometric regions partitioning the domain
   void SetUpHierarchy();
 
-  //This method is called in SetUpHierarchy and it is in chanrge of extracting quantities from MueLu::Hierarchy objects
-  //and store them inside Xpetra::Level objects
+  // This method is called in SetUpHierarchy and it is in chanrge of extracting quantities from MueLu::Hierarchy objects
+  // and store them inside Xpetra::Level objects
   void DefineLevels();
 
-  //Method to create region input multivectors from the composite input multivector
-  void computeRegionX(const multivector_type& , Array<RCP<multivector_type> > )const;
+  // Method to create region input multivectors from the composite input multivector
+  void computeRegionX(const multivector_type&, Array<RCP<multivector_type> >) const;
 
-  //Method to create composite output multivector from region output multivectors
-  void computeCompositeY(Array<RCP<const multivector_type> >, multivector_type& )const;
+  // Method to create composite output multivector from region output multivectors
+  void computeCompositeY(Array<RCP<const multivector_type> >, multivector_type&) const;
 
-  //This method detects entries in region output mutlivectors associated with mesh nodes that lie on interregion interfaces
-  //and it scales them. The scaling factor is equal to the number of regions that the given degrees of freedom are shared by
-  void rescaleInterfaceEntries( Array<RCP<multivector_type> > )const;
+  // This method detects entries in region output mutlivectors associated with mesh nodes that lie on interregion interfaces
+  // and it scales them. The scaling factor is equal to the number of regions that the given degrees of freedom are shared by
+  void rescaleInterfaceEntries(Array<RCP<multivector_type> >) const;
 
-  virtual void regionToAllCoarsen(const level&, level& );
+  virtual void regionToAllCoarsen(const level&, level&);
 
   //@}
-
 };
 
-} //namespace Xpetra
+}  // namespace Xpetra
 
 #endif
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_def.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_def.hpp
index 022b51e32b76..c10e467ec0c4 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_def.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionAMG_def.hpp
@@ -54,41 +54,41 @@
 
 #include "Xpetra_RegionAMG_decl.hpp"
 
-namespace Xpetra{
+namespace Xpetra {
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RegionAMG(
     Teuchos::RCP<tpetra_splitting> matrixSplitting,
     Teuchos::RCP<Xpetra::RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionHandler,
     RCP<const Teuchos::Comm<int> > comm, Teuchos::ParameterList muelu,
-    GlobalOrdinal num_levels, GlobalOrdinal coarsening_factor) :
-    num_levels_(num_levels), coarsening_factor_(coarsening_factor), comm_(comm), muelu_(
-        muelu)
-{
-
+    GlobalOrdinal num_levels, GlobalOrdinal coarsening_factor)
+  : num_levels_(num_levels)
+  , coarsening_factor_(coarsening_factor)
+  , comm_(comm)
+  , muelu_(
+        muelu) {
   TEUCHOS_TEST_FOR_EXCEPT(matrixSplitting.is_null());
   matrixSplitting_ = matrixSplitting;
 
-  //The maps defined here below are used to interface with outer applications (e.g. Belos-type solvers)
-  //and guarantee that DomainMap and RangeMap partition input multivectors and output multivectors in a proper way.
-  //the maps stored in doaminMap and rangeMap coincide with the maps of the composite matrix
-  domainMap_ = matrixSplitting_->getDomainMap();
-  rangeMap_ = matrixSplitting_->getRangeMap();
+  // The maps defined here below are used to interface with outer applications (e.g. Belos-type solvers)
+  // and guarantee that DomainMap and RangeMap partition input multivectors and output multivectors in a proper way.
+  // the maps stored in doaminMap and rangeMap coincide with the maps of the composite matrix
+  domainMap_   = matrixSplitting_->getDomainMap();
+  rangeMap_    = matrixSplitting_->getRangeMap();
   num_regions_ = matrixSplitting_->getNumRegions();
 
   SetUpHierarchy();
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetUpHierarchy()
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( num_levels_<=0, Exceptions::RuntimeError, "Number of levels must be a positive integer number \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_<=0 , Exceptions::RuntimeError, "No existing regions \n");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetUpHierarchy() {
+  TEUCHOS_TEST_FOR_EXCEPTION(num_levels_ <= 0, Exceptions::RuntimeError, "Number of levels must be a positive integer number \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ <= 0, Exceptions::RuntimeError, "No existing regions \n");
 
-  //Creation of the MueLu list for the region multigrid preconditioner
-  //It would be ideal to have this parameter list read by an .xml file
-  //at the final stage of the implementation.
-  //As for now, we hard code it insice of SetUpHierarchy because we want standard setups for almost everything
+  // Creation of the MueLu list for the region multigrid preconditioner
+  // It would be ideal to have this parameter list read by an .xml file
+  // at the final stage of the implementation.
+  // As for now, we hard code it insice of SetUpHierarchy because we want standard setups for almost everything
   RCP<ParameterList> list = rcp(new Teuchos::ParameterList());
   list->setName("MueLu");
 
@@ -101,29 +101,27 @@ void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetUpHierarchy()
   list->set("coarse: max size", 16);
 
   GlobalOrdinal num_regions = matrixSplitting_->getNumRegions();
-  regionHierarchies_.resize( num_regions );
-
-  //Instantiation of MueLu Hierarchies for each region of the domain
-  for( GlobalOrdinal region_idx = 0; region_idx<num_regions; ++region_idx )
-  {
+  regionHierarchies_.resize(num_regions);
 
-    //The parameter n is the total number of mesh nodes in a region
+  // Instantiation of MueLu Hierarchies for each region of the domain
+  for (GlobalOrdinal region_idx = 0; region_idx < num_regions; ++region_idx) {
+    // The parameter n is the total number of mesh nodes in a region
     GlobalOrdinal n = matrixSplitting_->getRegionHandler()->GetNumRegionNodes(region_idx);
 
     // N.B. The following instructions is based on the assumption that each region is a square. If not, then another instruciton
     // must be used to specify the size of the region mesh along teh x-direction
     GlobalOrdinal nx = std::sqrt(n);
 
-    RCP<const map_type> map = matrixSplitting_->getRegionMatrix(region_idx)->getRowMap();
-    size_t NumMyElements = map->getLocalNumElements();
+    RCP<const map_type> map                                  = matrixSplitting_->getRegionMatrix(region_idx)->getRowMap();
+    size_t NumMyElements                                     = map->getLocalNumElements();
     Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
 
-    RCP<multivector_type> coords = mv_factory_type::Build (map, 2);
+    RCP<multivector_type> coords = mv_factory_type::Build(map, 2);
 
-    //Coordinates are created to pass them as additional input the MueLu::Hierarchy constructors
-    //The only thing we care aboput is to create a unique mapping between coordinates and degrees of freedom associated with mesh nodes
-    //Therefore we base the creation of coordinates upon a fake region mesh, which does not coincide with its real gemetric shape
-    //The fake region we use for coordinates is a [0,(number_nodes_x-1)]X[0,(number_nodes_y-1)] rectangle
+    // Coordinates are created to pass them as additional input the MueLu::Hierarchy constructors
+    // The only thing we care aboput is to create a unique mapping between coordinates and degrees of freedom associated with mesh nodes
+    // Therefore we base the creation of coordinates upon a fake region mesh, which does not coincide with its real gemetric shape
+    // The fake region we use for coordinates is a [0,(number_nodes_x-1)]X[0,(number_nodes_y-1)] rectangle
     Teuchos::ArrayRCP<Teuchos::ArrayRCP<Scalar> > Coord(2);
     Coord[0] = coords->getDataNonConst(0);
     Coord[1] = coords->getDataNonConst(1);
@@ -136,26 +134,23 @@ void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetUpHierarchy()
       Coord[1][i] = Teuchos::as<Scalar>(iy);
     }
 
-    //We create as many MueLu::Hierarchy objects as the nubmer of region that partition the composite domain
-    regionHierarchies_[region_idx] = MueLu::CreateXpetraPreconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node>( matrixSplitting_->getRegionMatrix( region_idx), *list, coords );
+    // We create as many MueLu::Hierarchy objects as the nubmer of region that partition the composite domain
+    regionHierarchies_[region_idx] = MueLu::CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(matrixSplitting_->getRegionMatrix(region_idx), *list, coords);
 
     // // The following commented line here below should replace the line right above if the MueLu parameter list is passed through an .xml file
-    //regionHierarchies_[region_idx] = MueLu::CreateXpetraPreconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node>( matrixSplitting_->getRegionMatrix( region_idx), muelu_, coords );
+    // regionHierarchies_[region_idx] = MueLu::CreateXpetraPreconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node>( matrixSplitting_->getRegionMatrix( region_idx), muelu_, coords );
 
-    //Different regions may have meshes with different size, or the number of levels the user wants to create may be too big
-    //with respect to the number of levels MueLu allows (i.e. the minimum coarse size may be reached for a smaller number of levels)
-    //In this case, the value of num_levels_ is readjusted to the minimum number of levels instantied across all the regions of the domain
-    num_levels_ = std::min( num_levels_, regionHierarchies_[region_idx]->GetNumLevels() );
+    // Different regions may have meshes with different size, or the number of levels the user wants to create may be too big
+    // with respect to the number of levels MueLu allows (i.e. the minimum coarse size may be reached for a smaller number of levels)
+    // In this case, the value of num_levels_ is readjusted to the minimum number of levels instantied across all the regions of the domain
+    num_levels_ = std::min(num_levels_, regionHierarchies_[region_idx]->GetNumLevels());
   }
 
   DefineLevels();
-
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DefineLevels ( )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DefineLevels() {
   levels_.clear();
   Array<RCP<matrix_type> > P;
   P.clear();
@@ -164,46 +159,41 @@ void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DefineLevels ( )
   Array<RCP<matrix_type> > A;
   A.clear();
 
-  TEUCHOS_TEST_FOR_EXCEPTION( levels_.size()!=0, Exceptions::RuntimeError, "Levels structure is already initialized \n" );
-  for( int i = 0; i<num_levels_; ++i )
-  {
+  TEUCHOS_TEST_FOR_EXCEPTION(levels_.size() != 0, Exceptions::RuntimeError, "Levels structure is already initialized \n");
+  for (int i = 0; i < num_levels_; ++i) {
     P.clear();
     R.clear();
     A.clear();
 
-    RCP<level> new_level = rcp( new level(i,num_regions_) );
-    for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION( !regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("A") , Exceptions::RuntimeError, "No existing operator at level "<<i<<" of region "<<region_idx<<"\n");
-      A.push_back( regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("A") );
-
-      //MueLu::Hierarchy objects store prolongator and restriction operators only in levels that are the result of a coarsening (they do not exist at the fine level)
-      if( i>0 )
-      {
-        TEUCHOS_TEST_FOR_EXCEPTION( !regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("P") , Exceptions::RuntimeError, "No existing prolongator at level "<<i<<" of region "<<region_idx<<"\n");
-        TEUCHOS_TEST_FOR_EXCEPTION( !regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("R") , Exceptions::RuntimeError, "No existing restriction at level "<<i<<" of region "<<region_idx<<"\n");
-        P.push_back( regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("P") );
-        R.push_back( regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("R") );
+    RCP<level> new_level = rcp(new level(i, num_regions_));
+    for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+      TEUCHOS_TEST_FOR_EXCEPTION(!regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("A"), Exceptions::RuntimeError, "No existing operator at level " << i << " of region " << region_idx << "\n");
+      A.push_back(regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("A"));
+
+      // MueLu::Hierarchy objects store prolongator and restriction operators only in levels that are the result of a coarsening (they do not exist at the fine level)
+      if (i > 0) {
+        TEUCHOS_TEST_FOR_EXCEPTION(!regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("P"), Exceptions::RuntimeError, "No existing prolongator at level " << i << " of region " << region_idx << "\n");
+        TEUCHOS_TEST_FOR_EXCEPTION(!regionHierarchies_[region_idx]->GetLevel(i)->IsAvailable("R"), Exceptions::RuntimeError, "No existing restriction at level " << i << " of region " << region_idx << "\n");
+        P.push_back(regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("P"));
+        R.push_back(regionHierarchies_[region_idx]->GetLevel(i)->template Get<RCP<matrix_type> >("R"));
       }
     }
-    //Following the same policy adopted in MueLu::Hierarchy, prolongator and restriction operators are stored at levels produced with a coarsening
-    if( i>0 )
-    {
-      new_level->SetP( P );
-      new_level->SetR( R );
+    // Following the same policy adopted in MueLu::Hierarchy, prolongator and restriction operators are stored at levels produced with a coarsening
+    if (i > 0) {
+      new_level->SetP(P);
+      new_level->SetR(R);
     }
 
-    //The Galerkin operator is define dat every level (regardless of coarsening)
-    new_level->SetA( A );
+    // The Galerkin operator is define dat every level (regardless of coarsening)
+    new_level->SetA(A);
 
-    //The structure regionToAll is passed to the fine level as it is
-    if( 0==i )
-      new_level->SetRegionToAll( matrixSplitting_->getRegionHandler()->GetRegionToAll() );
-    else
-    {
-      //Coarsening level own a regionToAll structure whcih is the result of an injected coarsened regionToAll from the upper level of the hierarchy
+    // The structure regionToAll is passed to the fine level as it is
+    if (0 == i)
+      new_level->SetRegionToAll(matrixSplitting_->getRegionHandler()->GetRegionToAll());
+    else {
+      // Coarsening level own a regionToAll structure whcih is the result of an injected coarsened regionToAll from the upper level of the hierarchy
       Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > coarse_regionToAll;
-      regionToAllCoarsen( *(levels_[i-1]), *new_level );
+      regionToAllCoarsen(*(levels_[i - 1]), *new_level);
     }
 
     new_level->checkConsistency();
@@ -214,133 +204,106 @@ void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DefineLevels ( )
     // (THIS HAS TO BE CHANGED SO THAT A PORTION OF THE COMPOSITE SMOOTHER IS STORED INSTEAD)
     new_level->ComputeRegionJacobi();
 
-    levels_.push_back( new_level );
+    levels_.push_back(new_level);
   }
 }
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::regionToAllCoarsen(const level& fine, level& coarse) {
+  TEUCHOS_TEST_FOR_EXCEPTION(fine.GetNumRegions() != coarse.GetNumRegions(), Exceptions::RuntimeError, "Level " << fine.GetLevelID() << "has " << fine.GetNumRegions() << " regions instantiated whereas level " << coarse.GetLevelID() << "has " << coarse.GetNumRegions() << " regions instantiated \n");
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::regionToAllCoarsen (const level& fine, level& coarse)
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( fine.GetNumRegions()!=coarse.GetNumRegions(), Exceptions::RuntimeError, "Level "<<fine.GetLevelID()<<"has "<<fine.GetNumRegions()<<" regions instantiated whereas level "<<coarse.GetLevelID()<<"has "<<coarse.GetNumRegions()<<" regions instantiated \n" );
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > fine_regionToAll = fine.GetRegionToAll();
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > coarse_regionToAll(num_regions_);
 
-  Array<Array<std::tuple<GlobalOrdinal,GlobalOrdinal> > > fine_regionToAll = fine.GetRegionToAll();
-  Array<Array<std::tuple<GlobalOrdinal,GlobalOrdinal> > > coarse_regionToAll(num_regions_);
-
-  for( GlobalOrdinal region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-
-    //Coordinates needed to apply the coarsening
-    GlobalOrdinal n = fine.GetNumRegionNodes(region_idx);
+  for (GlobalOrdinal region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    // Coordinates needed to apply the coarsening
+    GlobalOrdinal n  = fine.GetNumRegionNodes(region_idx);
     GlobalOrdinal nx = std::sqrt(n);
     GlobalOrdinal ny = nx;
 
-    TEUCHOS_TEST_FOR_EXCEPTION( (nx-1)%coarsening_factor_!=0 || (ny-1)%coarsening_factor_!=0, Exceptions::RuntimeError, "Region: "<<region_idx<<" cannot be coarsened by factor equal to "<<coarsening_factor_<<"n: "<<n<<" - nx: "<<nx<<" - ny: "<<ny<<" \n" );
+    TEUCHOS_TEST_FOR_EXCEPTION((nx - 1) % coarsening_factor_ != 0 || (ny - 1) % coarsening_factor_ != 0, Exceptions::RuntimeError, "Region: " << region_idx << " cannot be coarsened by factor equal to " << coarsening_factor_ << "n: " << n << " - nx: " << nx << " - ny: " << ny << " \n");
 
-    //here below we exploit the fact that the regionToAll has been sorted in ascending order for the region node index
+    // here below we exploit the fact that the regionToAll has been sorted in ascending order for the region node index
     GlobalOrdinal count = 1;
 
-    while( count<=fine_regionToAll[region_idx].size() )
-    {
+    while (count <= fine_regionToAll[region_idx].size()) {
       coarse_regionToAll[region_idx].push_back(fine_regionToAll[region_idx][count]);
-      if( count%ny!=0 )
+      if (count % ny != 0)
         count = count + coarsening_factor_;
       else
-        count = count + ( coarsening_factor_-1 ) * ny + 1;
-
+        count = count + (coarsening_factor_ - 1) * ny + 1;
     }
 
-    coarse.SetRegionToAll( coarse_regionToAll );
-
+    coarse.SetRegionToAll(coarse_regionToAll);
   }
-
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::computeRegionX (const multivector_type& X, Array<RCP<multivector_type> > regionX)const
-{
-
-  //Array to store extended region Maps (needed to copy composite entries of the input vector into region partitioning of it)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::computeRegionX(const multivector_type& X, Array<RCP<multivector_type> > regionX) const {
+  // Array to store extended region Maps (needed to copy composite entries of the input vector into region partitioning of it)
   Teuchos::Array<Teuchos::Array<GlobalOrdinal> > overlapping_composite_array;
 
-  //Create Overlapping composite maps
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
+  // Create Overlapping composite maps
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
     Teuchos::Array<GlobalOrdinal> overlapped_composite;
     LocalOrdinal num_elements = regionX[region_idx]->getMap()->getLocalNumElements();
-    for( LocalOrdinal local_region_index = 0; local_region_index<num_elements; ++local_region_index )
-    {
-      GlobalOrdinal composite_region_index = regionX[region_idx]->getMap()->getGlobalElement(local_region_index);
-      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index+1);
-      overlapped_composite.push_back( composite_composite_index-1 );
+    for (LocalOrdinal local_region_index = 0; local_region_index < num_elements; ++local_region_index) {
+      GlobalOrdinal composite_region_index    = regionX[region_idx]->getMap()->getGlobalElement(local_region_index);
+      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index + 1);
+      overlapped_composite.push_back(composite_composite_index - 1);
     }
-    overlapping_composite_array.push_back( overlapped_composite );
+    overlapping_composite_array.push_back(overlapped_composite);
   }
 
-  //We split at first the input and output multivectors into region ones
+  // We split at first the input and output multivectors into region ones
   Array<RCP<multivector_type> > composite_overlapping_X;
-  composite_overlapping_X.resize( num_regions_ );
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-    //The new Map associated with a process must contain entries currently owned plus new ones previously owned only by neighbouring processes
+  composite_overlapping_X.resize(num_regions_);
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    // The new Map associated with a process must contain entries currently owned plus new ones previously owned only by neighbouring processes
     Teuchos::Array<GlobalOrdinal> aux;
-    if( overlapping_composite_array[region_idx].size()>0 )
-    {
+    if (overlapping_composite_array[region_idx].size() > 0) {
       Teuchos::Array<GlobalOrdinal> aux1 = overlapping_composite_array[region_idx];
       Teuchos::Array<GlobalOrdinal> aux2 = X.getMap()->getLocalElementList();
-      std::set_union(aux1.begin(), aux1.end(), aux2.begin(), aux2.end(),std::back_inserter(aux));
-    }
-    else
-    {
+      std::set_union(aux1.begin(), aux1.end(), aux2.begin(), aux2.end(), std::back_inserter(aux));
+    } else {
       aux = X.getMap()->getLocalElementList();
     }
-    RCP<map_type> overlapping_composite_map = MapFactory< LocalOrdinal, GlobalOrdinal, Node >::Build( Xpetra::UseTpetra, X.getGlobalLength(), aux, 0, comm_ );
-    composite_overlapping_X[region_idx] = MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(overlapping_composite_map, X.getNumVectors());
-    RCP<Import<LocalOrdinal, GlobalOrdinal, Node> > Import1 = ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build( X.getMap(), overlapping_composite_map );
-    TEUCHOS_TEST_FOR_EXCEPTION( X.getMap()->getMinAllGlobalIndex()!=composite_overlapping_X[region_idx]->getMap()->getMinAllGlobalIndex(), Exceptions::RuntimeError, "Minimal index in old an new maps do not coincide \n" );
-    TEUCHOS_TEST_FOR_EXCEPTION( X.getMap()->getMaxAllGlobalIndex()!=composite_overlapping_X[region_idx]->getMap()->getMaxAllGlobalIndex(), Exceptions::RuntimeError, "Maximal index in old an new maps do not coincide \n" );
-    composite_overlapping_X[region_idx]->doImport( X, *Import1, Xpetra::INSERT );
+    RCP<map_type> overlapping_composite_map                 = MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, X.getGlobalLength(), aux, 0, comm_);
+    composite_overlapping_X[region_idx]                     = MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(overlapping_composite_map, X.getNumVectors());
+    RCP<Import<LocalOrdinal, GlobalOrdinal, Node> > Import1 = ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(X.getMap(), overlapping_composite_map);
+    TEUCHOS_TEST_FOR_EXCEPTION(X.getMap()->getMinAllGlobalIndex() != composite_overlapping_X[region_idx]->getMap()->getMinAllGlobalIndex(), Exceptions::RuntimeError, "Minimal index in old an new maps do not coincide \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(X.getMap()->getMaxAllGlobalIndex() != composite_overlapping_X[region_idx]->getMap()->getMaxAllGlobalIndex(), Exceptions::RuntimeError, "Maximal index in old an new maps do not coincide \n");
+    composite_overlapping_X[region_idx]->doImport(X, *Import1, Xpetra::INSERT);
   }
 
-  //Copy values from composite input multivector X into region multivectors regionX
-  for( int i = 0; i<X.getNumVectors(); ++i )
-  {
-    for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-    {
-      ArrayRCP<const Scalar> composite_column = composite_overlapping_X[region_idx]->getData( i );
-      LocalOrdinal num_elements = regionX[region_idx]->getMap()->getLocalNumElements();
-      ArrayRCP<Scalar> region_column = regionX[region_idx]->getDataNonConst( i );
-      for( LocalOrdinal local_region_index = 0; local_region_index<num_elements; ++local_region_index )
-      {
-        GlobalOrdinal composite_region_index = regionX[region_idx]->getMap()->getGlobalElement(local_region_index);
-        GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index+1);
-        LocalOrdinal local_composite_index =  composite_overlapping_X[region_idx]->getMap()->getLocalElement( composite_composite_index-1 );
-        region_column[local_region_index] = composite_column[local_composite_index];
+  // Copy values from composite input multivector X into region multivectors regionX
+  for (int i = 0; i < X.getNumVectors(); ++i) {
+    for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+      ArrayRCP<const Scalar> composite_column = composite_overlapping_X[region_idx]->getData(i);
+      LocalOrdinal num_elements               = regionX[region_idx]->getMap()->getLocalNumElements();
+      ArrayRCP<Scalar> region_column          = regionX[region_idx]->getDataNonConst(i);
+      for (LocalOrdinal local_region_index = 0; local_region_index < num_elements; ++local_region_index) {
+        GlobalOrdinal composite_region_index    = regionX[region_idx]->getMap()->getGlobalElement(local_region_index);
+        GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index + 1);
+        LocalOrdinal local_composite_index      = composite_overlapping_X[region_idx]->getMap()->getLocalElement(composite_composite_index - 1);
+        region_column[local_region_index]       = composite_column[local_composite_index];
       }
     }
   }
-
 }
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::computeCompositeY(Array<RCP<const multivector_type> > regionY, multivector_type& Y) const {
+  // Array to store extended region Maps (needed to copy composite entries of the input vector into region partitioning of it)
+  Teuchos::Array<GlobalOrdinal> overlapping_composite;
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::computeCompositeY (Array<RCP<const multivector_type> > regionY, multivector_type& Y)const
-{
-
-  //Array to store extended region Maps (needed to copy composite entries of the input vector into region partitioning of it)
-  Teuchos::Array<GlobalOrdinal>  overlapping_composite;
-
-  //Create Overlapping composite maps
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
+  // Create Overlapping composite maps
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
     LocalOrdinal num_elements = regionY[region_idx]->getMap()->getLocalNumElements();
-    for( LocalOrdinal local_region_index = 0; local_region_index<num_elements; ++local_region_index )
-    {
-      GlobalOrdinal composite_region_index = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
-      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index+1);
-      overlapping_composite.push_back( composite_composite_index-1 );
+    for (LocalOrdinal local_region_index = 0; local_region_index < num_elements; ++local_region_index) {
+      GlobalOrdinal composite_region_index    = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
+      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index + 1);
+      overlapping_composite.push_back(composite_composite_index - 1);
     }
   }
 
@@ -349,136 +312,116 @@ void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::computeCompositeY (Ar
   last = std::unique(overlapping_composite.begin(), overlapping_composite.end());
   overlapping_composite.erase(last, overlapping_composite.end());
 
-  //We split at first the input and output multivectors into region ones
+  // We split at first the input and output multivectors into region ones
   RCP<multivector_type> composite_overlapping_Y;
-  RCP<map_type> overlapping_composite_map = MapFactory< LocalOrdinal, GlobalOrdinal, Node >::Build( Xpetra::UseTpetra, Y.getGlobalLength(), overlapping_composite, 0, comm_ );
-  composite_overlapping_Y = MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(overlapping_composite_map, Y.getNumVectors());
+  RCP<map_type> overlapping_composite_map = MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Xpetra::UseTpetra, Y.getGlobalLength(), overlapping_composite, 0, comm_);
+  composite_overlapping_Y                 = MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(overlapping_composite_map, Y.getNumVectors());
 
-  //Safety checks to guarantee the legitimacy of non-unique source composite map and unique target composite map
-  //The unique target composite map must be a subset of the non-unique source composite map
-  //Before running the std::includes algorithm it is very important to sort the ElementList in ascending order
-  Teuchos::Array<GlobalOrdinal> original_map = Y.getMap()->getLocalElementList();
+  // Safety checks to guarantee the legitimacy of non-unique source composite map and unique target composite map
+  // The unique target composite map must be a subset of the non-unique source composite map
+  // Before running the std::includes algorithm it is very important to sort the ElementList in ascending order
+  Teuchos::Array<GlobalOrdinal> original_map    = Y.getMap()->getLocalElementList();
   Teuchos::Array<GlobalOrdinal> overlapping_map = composite_overlapping_Y->getMap()->getLocalElementList();
   std::sort(original_map.begin(), original_map.end());
   std::sort(overlapping_map.begin(), overlapping_map.end());
-  TEUCHOS_TEST_FOR_EXCEPTION( !( std::includes(overlapping_map.begin(), overlapping_map.end(), original_map.begin(), original_map.end()) ), Exceptions::RuntimeError, "Overlapping (non-unique) composite map does NOT include original (unique) composite map \n" );
-
-  //Copy values from output region multivectors regionY into output composite multivector Y
-  for( int i = 0; i<Y.getNumVectors(); ++i )
-  {
-    for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-    {
-      ArrayRCP<Scalar> composite_column = composite_overlapping_Y->getDataNonConst( i );
-      LocalOrdinal num_elements = regionY[region_idx]->getMap()->getLocalNumElements();
-      ArrayRCP<const Scalar> region_column = regionY[region_idx]->getData( i );
-      for( LocalOrdinal local_region_index = 0; local_region_index<num_elements; ++local_region_index )
-      {
-        GlobalOrdinal composite_region_index = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
-        GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index+1);
-        LocalOrdinal local_composite_index =  composite_overlapping_Y->getMap()->getLocalElement( composite_composite_index-1 );
+  TEUCHOS_TEST_FOR_EXCEPTION(!(std::includes(overlapping_map.begin(), overlapping_map.end(), original_map.begin(), original_map.end())), Exceptions::RuntimeError, "Overlapping (non-unique) composite map does NOT include original (unique) composite map \n");
+
+  // Copy values from output region multivectors regionY into output composite multivector Y
+  for (int i = 0; i < Y.getNumVectors(); ++i) {
+    for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+      ArrayRCP<Scalar> composite_column    = composite_overlapping_Y->getDataNonConst(i);
+      LocalOrdinal num_elements            = regionY[region_idx]->getMap()->getLocalNumElements();
+      ArrayRCP<const Scalar> region_column = regionY[region_idx]->getData(i);
+      for (LocalOrdinal local_region_index = 0; local_region_index < num_elements; ++local_region_index) {
+        GlobalOrdinal composite_region_index    = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
+        GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index + 1);
+        LocalOrdinal local_composite_index      = composite_overlapping_Y->getMap()->getLocalElement(composite_composite_index - 1);
         composite_column[local_composite_index] += region_column[local_region_index];
       }
     }
   }
 
-  //Import object needed to transfer output composite multivector from non-unique map to unique map
-  RCP<Import<LocalOrdinal, GlobalOrdinal, Node> > Import1 = ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build( composite_overlapping_Y->getMap(), Y.getMap() );
-
-  //IMPORTANT: ALWAYS USE IMPORT-EXPORT OBJECTS IN FORWARD MODE (i.e. doImport methods must use Import objects and doExport methods must use Export objects)
-  //Reverse mode (i.e. crossed use of doImport/doExport methods with Export/Import objects makes the code possibly crash due to undefined behavior)
-  Y.doImport( *composite_overlapping_Y, *Import1, Xpetra::ADD );
+  // Import object needed to transfer output composite multivector from non-unique map to unique map
+  RCP<Import<LocalOrdinal, GlobalOrdinal, Node> > Import1 = ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(composite_overlapping_Y->getMap(), Y.getMap());
 
+  // IMPORTANT: ALWAYS USE IMPORT-EXPORT OBJECTS IN FORWARD MODE (i.e. doImport methods must use Import objects and doExport methods must use Export objects)
+  // Reverse mode (i.e. crossed use of doImport/doExport methods with Export/Import objects makes the code possibly crash due to undefined behavior)
+  Y.doImport(*composite_overlapping_Y, *Import1, Xpetra::ADD);
 }
 
-
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::rescaleInterfaceEntries (Array<RCP<multivector_type> > regionY)const
-{
-
-  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll = levels_[0]->GetRegionToAll();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::rescaleInterfaceEntries(Array<RCP<multivector_type> > regionY) const {
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll    = levels_[0]->GetRegionToAll();
   Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > > interfaceNodes = matrixSplitting_->getRegionHandler()->GetInterfaceNodes();
 
-  TEUCHOS_TEST_FOR_EXCEPTION( num_regions_!=regionToAll.size(), Exceptions::RuntimeError, "Regions stored in Level 0 do not match with total number of regions in RegionAMG class \n" );
+  TEUCHOS_TEST_FOR_EXCEPTION(num_regions_ != regionToAll.size(), Exceptions::RuntimeError, "Regions stored in Level 0 do not match with total number of regions in RegionAMG class \n");
 
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
     LocalOrdinal num_elements = regionY[region_idx]->getMap()->getLocalNumElements();
-    for( LocalOrdinal local_region_index = 0; local_region_index<num_elements; ++local_region_index )
-    {
-      GlobalOrdinal composite_region_index = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
-      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index+1);
+    for (LocalOrdinal local_region_index = 0; local_region_index < num_elements; ++local_region_index) {
+      GlobalOrdinal composite_region_index    = regionY[region_idx]->getMap()->getGlobalElement(local_region_index);
+      GlobalOrdinal composite_composite_index = levels_[0]->GetCompositeIndex(region_idx, composite_region_index + 1);
       checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(composite_composite_index);
-      typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-      nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicateNode);
-      //The rescaling must be applied only to entries associated with mesh on the interface
-      if( nodes_to_region_iterator!=interfaceNodes.end() )
-      {
-        Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
-        for( int i = 0; i<regionY[region_idx]->getNumVectors(); ++i)
-        {
-          ArrayRCP<Scalar> region_column = regionY[region_idx]->getDataNonConst( i );
-          if( nodal_regions.size()>1 )
-            region_column[ local_region_index ] = region_column[ local_region_index ]/( nodal_regions.size()  );
+      typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+      nodes_to_region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(interfaceNodes.begin(), interfaceNodes.end(), unaryPredicateNode);
+      // The rescaling must be applied only to entries associated with mesh on the interface
+      if (nodes_to_region_iterator != interfaceNodes.end()) {
+        Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
+        for (int i = 0; i < regionY[region_idx]->getNumVectors(); ++i) {
+          ArrayRCP<Scalar> region_column = regionY[region_idx]->getDataNonConst(i);
+          if (nodal_regions.size() > 1)
+            region_column[local_region_index] = region_column[local_region_index] / (nodal_regions.size());
         }
       }
     }
   }
-
 }
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const multivector_type& X, multivector_type& Y, Teuchos::ETransp mode, Scalar alpha, Scalar beta) const {
+  // N.B.: currently Scalar quantities alpha and beta are passed as input parameters to have the apply method signature match with the apply signature of an Xpetra::Operator
+  // however we are not currently using them (for us alpha=0 and beta=1)
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionAMG<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply (const multivector_type& X, multivector_type& Y, Teuchos::ETransp mode, Scalar alpha, Scalar beta)const
-{
-
-  //N.B.: currently Scalar quantities alpha and beta are passed as input parameters to have the apply method signature match with the apply signature of an Xpetra::Operator
-  //however we are not currently using them (for us alpha=0 and beta=1)
-
-  //At first we check that input and output vector have matching maps with the composite matrix
-  //The Map of X must coincide with the Domain map of compositeA
-  //The Map of Y must coincide with the Range map of compositeA
-  TEUCHOS_TEST_FOR_EXCEPTION( !(X.getMap()->isSameAs( *(matrixSplitting_->getMatrix()->getDomainMap()) ) ), Exceptions::RuntimeError, "Map of composite input multivector X does not coincide with Domain Map of composite matrix \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( !(Y.getMap()->isSameAs( *(matrixSplitting_->getMatrix()->getRangeMap()) ) ), Exceptions::RuntimeError, "Map of composite input multivector X does not coincide with Range Map of composite matrix \n" );
-  TEUCHOS_TEST_FOR_EXCEPTION( X.getNumVectors()!=Y.getNumVectors(), Exceptions::RuntimeError, "Number of vectors in input numltivector X does NOT match number of vectors in output multivector Y \n" );
+  // At first we check that input and output vector have matching maps with the composite matrix
+  // The Map of X must coincide with the Domain map of compositeA
+  // The Map of Y must coincide with the Range map of compositeA
+  TEUCHOS_TEST_FOR_EXCEPTION(!(X.getMap()->isSameAs(*(matrixSplitting_->getMatrix()->getDomainMap()))), Exceptions::RuntimeError, "Map of composite input multivector X does not coincide with Domain Map of composite matrix \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(!(Y.getMap()->isSameAs(*(matrixSplitting_->getMatrix()->getRangeMap()))), Exceptions::RuntimeError, "Map of composite input multivector X does not coincide with Range Map of composite matrix \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != Y.getNumVectors(), Exceptions::RuntimeError, "Number of vectors in input numltivector X does NOT match number of vectors in output multivector Y \n");
 
-  //We split at first the input and output multivectors into region ones
+  // We split at first the input and output multivectors into region ones
   Array<RCP<multivector_type> > regionX;
   Array<RCP<multivector_type> > regionY;
-  regionX.resize( num_regions_ );
-  regionY.resize( num_regions_ );
-
-  //Associate Maps to region input (regionX) and output vectors (regionY)
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-    regionX[region_idx] =  MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(levels_[0]->GetRegionMatrix(region_idx)->getDomainMap(), X.getNumVectors());
-    regionY[region_idx] =  MultiVectorFactory< Scalar, LocalOrdinal, GlobalOrdinal, Node >::Build(levels_[0]->GetRegionMatrix(region_idx)->getRangeMap(), Y.getNumVectors());
+  regionX.resize(num_regions_);
+  regionY.resize(num_regions_);
+
+  // Associate Maps to region input (regionX) and output vectors (regionY)
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    regionX[region_idx] = MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(levels_[0]->GetRegionMatrix(region_idx)->getDomainMap(), X.getNumVectors());
+    regionY[region_idx] = MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(levels_[0]->GetRegionMatrix(region_idx)->getRangeMap(), Y.getNumVectors());
   }
 
-  //Split the composite input multivector	X into region multivector regionX
-  computeRegionX( X, regionX );
+  // Split the composite input multivector	X into region multivector regionX
+  computeRegionX(X, regionX);
 
-  //This is the portion where the V-cycle is executed (for now we only have region V-cycles)
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-    regionHierarchies_[region_idx]->Iterate( *regionX[region_idx],*regionY[region_idx] );
+  // This is the portion where the V-cycle is executed (for now we only have region V-cycles)
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    regionHierarchies_[region_idx]->Iterate(*regionX[region_idx], *regionY[region_idx]);
   }
 
-  //We rescale entries of region multivector regionY that are associated with mesh nodes on an interface
-  rescaleInterfaceEntries( regionY );
+  // We rescale entries of region multivector regionY that are associated with mesh nodes on an interface
+  rescaleInterfaceEntries(regionY);
 
-  //We create const view of the region multivector regionY because we want to guarantee that the composite output multivector Y
-  //is constructed without modifying any information given from each region
+  // We create const view of the region multivector regionY because we want to guarantee that the composite output multivector Y
+  // is constructed without modifying any information given from each region
   Array<RCP<const multivector_type> > regionYconst;
-  for( int region_idx = 0; region_idx<num_regions_; ++region_idx )
-  {
-    regionYconst.push_back( regionY[region_idx].getConst() );
+  for (int region_idx = 0; region_idx < num_regions_; ++region_idx) {
+    regionYconst.push_back(regionY[region_idx].getConst());
   }
 
-  //Assemble composite output multivector Y from region multivector regionY
-  computeCompositeY( regionYconst, Y );
-
+  // Assemble composite output multivector Y from region multivector regionY
+  computeCompositeY(regionYconst, Y);
 }
 
-}
+}  // namespace Xpetra
 #endif
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_decl.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_decl.hpp
index 4c748dbe5148..1216e22530d8 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_decl.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_decl.hpp
@@ -57,93 +57,78 @@
 #include <fstream>
 #include <vector>
 
-namespace Xpetra{
+namespace Xpetra {
 
-template<class GlobalOrdinal>
+template <class GlobalOrdinal>
 bool compareRegions(const std::tuple<GlobalOrdinal, GlobalOrdinal> &, const std::tuple<GlobalOrdinal, GlobalOrdinal> &);
 
-template<class GlobalOrdinal>
+template <class GlobalOrdinal>
 bool compareNodes(const std::tuple<GlobalOrdinal, GlobalOrdinal> &, const std::tuple<GlobalOrdinal, GlobalOrdinal> &);
 
-//Definition of the predicate for the node_ structure.
-//Given a tuple made of node index and a specific region it belongs to,
-//this predicate returns true if the node belongs to the region specified in input to the predicate.
-template<class GlobalOrdinal>
+// Definition of the predicate for the node_ structure.
+// Given a tuple made of node index and a specific region it belongs to,
+// this predicate returns true if the node belongs to the region specified in input to the predicate.
+template <class GlobalOrdinal>
 class checkerNode {
+ public:
+  // Constructor
+  checkerNode(GlobalOrdinal region_index) { region_index_ = region_index; };
 
-public:
-
-  //Constructor
-  checkerNode( GlobalOrdinal region_index){region_index_ = region_index;};
-
-  //Unary Operator
-  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal> &node)
-  { return (std::get<1>(node) == region_index_); }
-
-private:
+  // Unary Operator
+  bool operator()(const std::tuple<GlobalOrdinal, GlobalOrdinal> &node) { return (std::get<1>(node) == region_index_); }
 
+ private:
   GlobalOrdinal region_index_;
-
 };
 
-
-//Definition of the predicate for the nodesToRegion_ sitructure
-//Given a tuple made of node index and a vector with labels of regions it belongs to,
-//this predicate returns true if the node coincides with the node specified in input to the predicate.
-template<class GlobalOrdinal>
+// Definition of the predicate for the nodesToRegion_ sitructure
+// Given a tuple made of node index and a vector with labels of regions it belongs to,
+// this predicate returns true if the node coincides with the node specified in input to the predicate.
+template <class GlobalOrdinal>
 class checkerNodesToRegion {
+ public:
+  // Constructor
+  checkerNodesToRegion(GlobalOrdinal node_index) { node_index_ = node_index; };
 
-public:
-
-  //Constructor
-  checkerNodesToRegion( GlobalOrdinal node_index){node_index_ = node_index;};
-
-  //Unary Operator
-  bool operator()(const std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > &node)
-  { return (std::get<0>(node) == node_index_); }
-
-private:
+  // Unary Operator
+  bool operator()(const std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > &node) { return (std::get<0>(node) == node_index_); }
 
+ private:
   GlobalOrdinal node_index_;
-
 };
 
-
-//This is an auxiliary class to store row maps for the composite matrix, region matrices and
-//a regionToAll map to link region node indices with the composite ones
+// This is an auxiliary class to store row maps for the composite matrix, region matrices and
+// a regionToAll map to link region node indices with the composite ones
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-class Splitting_MapsInfo{
-public:
-  Array<Array< std::tuple<GlobalOrdinal,GlobalOrdinal> > > regionToAll_;//used as a map for a RegionToAll node index
-  Array<GlobalOrdinal> composite_map_; //used as RowMap for composite matrices
-  Array<Array<GlobalOrdinal> > region_maps_; //used as RowMap for region matrices
+class Splitting_MapsInfo {
+ public:
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll_;  // used as a map for a RegionToAll node index
+  Array<GlobalOrdinal> composite_map_;                                    // used as RowMap for composite matrices
+  Array<Array<GlobalOrdinal> > region_maps_;                              // used as RowMap for region matrices
 };
 
-
 // This is the actual class that defines the regionHandler
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-class RegionHandler{
-
-public:
-
+class RegionHandler {
+ public:
   //! @name Constructor/Destructor Methods
   //@{
 
   //! Constructor specifying the file name containing region information.
-  RegionHandler (const std::string &, RCP< const Teuchos::Comm<int> >);
+  RegionHandler(const std::string &, RCP<const Teuchos::Comm<int> >);
 
   //}
   //! @Interface methods
   //@{
-  GlobalOrdinal GetNumGlobalElements()const{return num_total_nodes_;};
-  GlobalOrdinal GetNumTotalRegions()const{return num_total_regions_;};
-  GlobalOrdinal GetNumRegionNodes(GlobalOrdinal region_idx)const{return num_region_nodes_[region_idx];};
-  Array<GlobalOrdinal> GetGlobalRowMap()const{return maps_.composite_map_;};
-  Array<GlobalOrdinal> GetRegionRowMap(GlobalOrdinal region_index)const;
-  Array<Array<GlobalOrdinal> > GetRegionRowMaps()const{return maps_.region_maps_;};
-  Array<Array< std::tuple<GlobalOrdinal,GlobalOrdinal> > >  GetRegionToAll()const;//used as a map for a RegionToAll node index
-  Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >   GetRegionToAll(GlobalOrdinal)const;//used as a map for a RegionToAll node index
-  Array<std::tuple<int, Array<GlobalOrdinal> > > GetInterfaceNodes()const{return interfaceNodes_;};
+  GlobalOrdinal GetNumGlobalElements() const { return num_total_nodes_; };
+  GlobalOrdinal GetNumTotalRegions() const { return num_total_regions_; };
+  GlobalOrdinal GetNumRegionNodes(GlobalOrdinal region_idx) const { return num_region_nodes_[region_idx]; };
+  Array<GlobalOrdinal> GetGlobalRowMap() const { return maps_.composite_map_; };
+  Array<GlobalOrdinal> GetRegionRowMap(GlobalOrdinal region_index) const;
+  Array<Array<GlobalOrdinal> > GetRegionRowMaps() const { return maps_.region_maps_; };
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > GetRegionToAll() const;       // used as a map for a RegionToAll node index
+  Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > GetRegionToAll(GlobalOrdinal) const;  // used as a map for a RegionToAll node index
+  Array<std::tuple<int, Array<GlobalOrdinal> > > GetInterfaceNodes() const { return interfaceNodes_; };
   //}
   //! @Printout methods
   void printView() const;
@@ -151,29 +136,28 @@ class RegionHandler{
   void printInactive() const;
   //}
 
-private:
-
+ private:
   //! @Private variables
   //@{
 
-  RCP< const Teuchos::Comm<int> > comm_;
+  RCP<const Teuchos::Comm<int> > comm_;
   bool nodes_sorted_by_regions_ = false;
 
-  //Global information
-  GlobalOrdinal num_total_nodes_ = 0;
+  // Global information
+  GlobalOrdinal num_total_nodes_   = 0;
   GlobalOrdinal num_total_regions_ = 0;
-  Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > nodes_;//basic structure that imports the information from the input file
+  Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > nodes_;  // basic structure that imports the information from the input file
 
-  //the following two Array are used to handle the situation where either the number of processes exceeds the number of regions or viceversa
-  Array<GlobalOrdinal> regions_per_proc_;//if num_proc > num_regions, then it says how many regions are owned by a single process, empty otherwise
-  Array<std::tuple<int, Array<GlobalOrdinal> > > procs_per_region_; //lists of processes instantiated for each region
+  // the following two Array are used to handle the situation where either the number of processes exceeds the number of regions or viceversa
+  Array<GlobalOrdinal> regions_per_proc_;                            // if num_proc > num_regions, then it says how many regions are owned by a single process, empty otherwise
+  Array<std::tuple<int, Array<GlobalOrdinal> > > procs_per_region_;  // lists of processes instantiated for each region
 
-  Array<std::tuple<int, Array<GlobalOrdinal> > > nodesToRegion_; //for each node it lists the regions it belongs to
-  Array<std::tuple<int, Array<GlobalOrdinal> > > interfaceNodes_; //for each node on the interface it lists the regions it belongs to
-  //vector which contains the number of region nodes for each domain region
+  Array<std::tuple<int, Array<GlobalOrdinal> > > nodesToRegion_;   // for each node it lists the regions it belongs to
+  Array<std::tuple<int, Array<GlobalOrdinal> > > interfaceNodes_;  // for each node on the interface it lists the regions it belongs to
+  // vector which contains the number of region nodes for each domain region
   Array<GlobalOrdinal> num_region_nodes_;
 
-  //Maps used for composite and region operators
+  // Maps used for composite and region operators
   Splitting_MapsInfo<Scalar, LocalOrdinal, GlobalOrdinal, Node> maps_;
   //@}
 
@@ -185,32 +169,30 @@ class RegionHandler{
   void CreateRowMaps();
   //@}
 
-}; //class RegionHandler
-
-//This compare class is used to run the sorting algorithm on the list of nodes with associated regions they belong to.
-//First, nodes are sorted in ascending order for region labels. Then, the sorting shuffles the nodes in ascending node index for
-//each given region
-template<class GlobalOrdinal>
-bool compareRegions(const std::tuple<GlobalOrdinal, GlobalOrdinal> &lhs, const std::tuple<GlobalOrdinal, GlobalOrdinal> &rhs)
-{
-  //First we prioritize the sorting according to the region label
-  //If the region is the same, then the sorting looks at the composite node index
-  if( std::get<1>(lhs) < std::get<1>(rhs) )
+};  // class RegionHandler
+
+// This compare class is used to run the sorting algorithm on the list of nodes with associated regions they belong to.
+// First, nodes are sorted in ascending order for region labels. Then, the sorting shuffles the nodes in ascending node index for
+// each given region
+template <class GlobalOrdinal>
+bool compareRegions(const std::tuple<GlobalOrdinal, GlobalOrdinal> &lhs, const std::tuple<GlobalOrdinal, GlobalOrdinal> &rhs) {
+  // First we prioritize the sorting according to the region label
+  // If the region is the same, then the sorting looks at the composite node index
+  if (std::get<1>(lhs) < std::get<1>(rhs))
     return true;
-  else if( std::get<1>(lhs) == std::get<1>(rhs) )
+  else if (std::get<1>(lhs) == std::get<1>(rhs))
     return std::get<0>(lhs) < std::get<0>(rhs);
   else
     return false;
 }
 
-//This compare is sed to run the sorting algorithm where the nodes are ordered in ascendin order for thei node indes, regardless of the
-//associated region index
-template<class GlobalOrdinal>
-bool compareNodes(const std::tuple<GlobalOrdinal, GlobalOrdinal> &lhs, const std::tuple<GlobalOrdinal, GlobalOrdinal> &rhs)
-{
+// This compare is sed to run the sorting algorithm where the nodes are ordered in ascendin order for thei node indes, regardless of the
+// associated region index
+template <class GlobalOrdinal>
+bool compareNodes(const std::tuple<GlobalOrdinal, GlobalOrdinal> &lhs, const std::tuple<GlobalOrdinal, GlobalOrdinal> &rhs) {
   return std::get<0>(lhs) < std::get<0>(rhs);
 }
 
-} //namespace Xpetra
+}  // namespace Xpetra
 
 #endif
diff --git a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_def.hpp b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_def.hpp
index 8e7b264c11c3..516f090f102e 100644
--- a/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_def.hpp
+++ b/packages/muelu/research/max/XpetraSplitting/Xpetra_RegionHandler_def.hpp
@@ -55,60 +55,55 @@
 #include "Xpetra_RegionHandler_decl.hpp"
 #include <algorithm>
 
-namespace Xpetra{
+namespace Xpetra {
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RegionHandler(const std::string &file_name, RCP< const Teuchos::Comm<int> > comm): comm_(comm)
-{
+RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RegionHandler(const std::string &file_name, RCP<const Teuchos::Comm<int> > comm)
+  : comm_(comm) {
   ReadFileInfo(file_name);
 
-  //Nodes are shuffled so that regions are sorted in ascending labeling order
+  // Nodes are shuffled so that regions are sorted in ascending labeling order
   std::sort(nodes_.begin(), nodes_.end(), compareRegions<GlobalOrdinal>);
   nodes_sorted_by_regions_ = true;
 
-  if(comm_->getRank()==0)
-    std::cout<<"Started NodesToRegion"<<std::endl;
+  if (comm_->getRank() == 0)
+    std::cout << "Started NodesToRegion" << std::endl;
   NodesToRegion();
-  if(comm_->getRank()==0)
-    std::cout<<"Finished NodesToRegion"<<std::endl;
+  if (comm_->getRank() == 0)
+    std::cout << "Finished NodesToRegion" << std::endl;
   ComputeProcRegions();
-  if(comm_->getRank()==0)
-    std::cout<<"Started RowMaps"<<std::endl;
+  if (comm_->getRank() == 0)
+    std::cout << "Started RowMaps" << std::endl;
   CreateRowMaps();
-  if(comm_->getRank()==0)
-    std::cout<<"Finished RowMaps"<<std::endl;
+  if (comm_->getRank() == 0)
+    std::cout << "Finished RowMaps" << std::endl;
 
   num_region_nodes_.clear();
   num_region_nodes_.resize(num_total_regions_);
 
-  //For each region, the following loop counts the number of region nodes and stores them
-  for( GlobalOrdinal region_idx = 1; region_idx<=num_total_regions_; ++region_idx )
-  {
+  // For each region, the following loop counts the number of region nodes and stores them
+  for (GlobalOrdinal region_idx = 1; region_idx <= num_total_regions_; ++region_idx) {
     checkerNode<GlobalOrdinal> unaryPredicate(region_idx);
-    typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator1;
-    typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator2;
-    nodes_iterator1 = std::find_if<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
-    nodes_iterator2 = std::find_if_not<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
-    num_region_nodes_[region_idx-1] = nodes_iterator2 - nodes_iterator1;
+    typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator1;
+    typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator2;
+    nodes_iterator1                   = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
+    nodes_iterator2                   = std::find_if_not<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
+    num_region_nodes_[region_idx - 1] = nodes_iterator2 - nodes_iterator1;
   }
-
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadFileInfo(const std::string &file_name)
-{
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadFileInfo(const std::string &file_name) {
   std::ifstream input_file_(file_name, std::ifstream::in);
-  std::string  line;
-  TEUCHOS_TEST_FOR_EXCEPTION( !input_file_.good(), Exceptions::RuntimeError, "Cannot read \"" << file_name << "\"");
+  std::string line;
+  TEUCHOS_TEST_FOR_EXCEPTION(!input_file_.good(), Exceptions::RuntimeError, "Cannot read \"" << file_name << "\"");
 
   GlobalOrdinal line_index = 0;
 
-  //The information contained in the file is imported and stored in a Teuchos::Array of tuples.
-  //The first field of the tuple is the composite node index, the second field of the tuple is the region index
-  while ( std::getline (input_file_,line) )
-  {
-    std::istringstream is( line );
+  // The information contained in the file is imported and stored in a Teuchos::Array of tuples.
+  // The first field of the tuple is the composite node index, the second field of the tuple is the region index
+  while (std::getline(input_file_, line)) {
+    std::istringstream is(line);
     GlobalOrdinal number;
     Array<GlobalOrdinal> node;
     std::tuple<GlobalOrdinal, GlobalOrdinal> node_region;
@@ -117,22 +112,18 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadFileInfo(cons
     node.clear();
     composite_info.clear();
 
-    if( 1==line_index )
-    {
-      while(is>>number)
+    if (1 == line_index) {
+      while (is >> number)
         composite_info.push_back(number);
 
-      TEUCHOS_TEST_FOR_EXCEPTION( composite_info.size()!=2, Exceptions::RuntimeError, "The composite information must be a couple of integers: nTotal of nodes + nTotal of regions \n");
-      num_total_nodes_ = composite_info[0];
+      TEUCHOS_TEST_FOR_EXCEPTION(composite_info.size() != 2, Exceptions::RuntimeError, "The composite information must be a couple of integers: nTotal of nodes + nTotal of regions \n");
+      num_total_nodes_   = composite_info[0];
       num_total_regions_ = composite_info[1];
-    }
-    else if( line_index>2 )
-    {
-      while(is>>number)
-      {
+    } else if (line_index > 2) {
+      while (is >> number) {
         node.push_back(number);
       }
-      TEUCHOS_TEST_FOR_EXCEPTION( node.size()!=2, Exceptions::RuntimeError, "The node information must be a couple of integers: Node index + Region idnex \n");
+      TEUCHOS_TEST_FOR_EXCEPTION(node.size() != 2, Exceptions::RuntimeError, "The node information must be a couple of integers: Node index + Region idnex \n");
       node_region = std::make_tuple(node[0], node[1]);
       nodes_.push_back(node_region);
       node.clear();
@@ -142,97 +133,86 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadFileInfo(cons
   input_file_.close();
 }
 
-//This routines computes the way regions are partitioned across processes
-//The partitioning policies of course depends on whether the number of processes
-//exceeds the number of regions or not.
-//ASSUMPTION: A PROCESS CANNOT OWN CHUNKS OF MULTIPLE REGIONS. EITHER A PROCESS IS CONFINED INSIDE A SINGLE REGION
-//OR IT MUST POSSESS ENTIRE REGIONS.
-//The distribution of regions (or portions of them) across processes is conductes so to guarantee load balancing
+// This routines computes the way regions are partitioned across processes
+// The partitioning policies of course depends on whether the number of processes
+// exceeds the number of regions or not.
+// ASSUMPTION: A PROCESS CANNOT OWN CHUNKS OF MULTIPLE REGIONS. EITHER A PROCESS IS CONFINED INSIDE A SINGLE REGION
+// OR IT MUST POSSESS ENTIRE REGIONS.
+// The distribution of regions (or portions of them) across processes is conductes so to guarantee load balancing
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeProcRegions()
-{
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeProcRegions() {
   int tot_num_proc = comm_->getSize();
-  int myPID = comm_->getRank();
+  int myPID        = comm_->getRank();
 
   regions_per_proc_.clear();
   procs_per_region_.clear();
 
-  //If the number of processes instantiate is smaller than the total number of regions,
-  //then each process owns entire regions. The number of regions per process is calculates so to guarantee
-  //load balancing. After an initial distribution of regions, leftover regions that have not been assigned to any process yet are
-  //distributed in a round-robin fashion
-  if( tot_num_proc < num_total_regions_ )
-  {
-    int min_nregions_proc = std::floor( static_cast<double>(num_total_regions_)/static_cast<double>(tot_num_proc) );
+  // If the number of processes instantiate is smaller than the total number of regions,
+  // then each process owns entire regions. The number of regions per process is calculates so to guarantee
+  // load balancing. After an initial distribution of regions, leftover regions that have not been assigned to any process yet are
+  // distributed in a round-robin fashion
+  if (tot_num_proc < num_total_regions_) {
+    int min_nregions_proc    = std::floor(static_cast<double>(num_total_regions_) / static_cast<double>(tot_num_proc));
     int num_leftover_regions = num_total_regions_ % tot_num_proc;
 
-    for( int i=1; i<=min_nregions_proc; ++i )
-      regions_per_proc_.push_back( myPID*min_nregions_proc+i );
+    for (int i = 1; i <= min_nregions_proc; ++i)
+      regions_per_proc_.push_back(myPID * min_nregions_proc + i);
 
-    if( num_leftover_regions>=myPID+1 && num_leftover_regions!=0 )
-      regions_per_proc_.push_back( min_nregions_proc*tot_num_proc + (myPID+1) );
+    if (num_leftover_regions >= myPID + 1 && num_leftover_regions != 0)
+      regions_per_proc_.push_back(min_nregions_proc * tot_num_proc + (myPID + 1));
 
-    for( int procID = 0; procID<comm_->getSize(); ++procID )
-    {
+    for (int procID = 0; procID < comm_->getSize(); ++procID) {
       Array<GlobalOrdinal> proc;
       proc.clear();
       proc.push_back(procID);
-      for( int i = 1; i<=min_nregions_proc; ++i )
-      {
-        GlobalOrdinal region_index = (procID)*min_nregions_proc + i;
+      for (int i = 1; i <= min_nregions_proc; ++i) {
+        GlobalOrdinal region_index                                 = (procID)*min_nregions_proc + i;
         std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > tuple_aux = std::make_tuple(region_index, proc);
-        procs_per_region_.push_back( tuple_aux );
+        procs_per_region_.push_back(tuple_aux);
       }
 
-      if( num_leftover_regions>=procID+1 && num_leftover_regions!=0 )
-      {
-        GlobalOrdinal region_index = min_nregions_proc*tot_num_proc + (procID+1);
+      if (num_leftover_regions >= procID + 1 && num_leftover_regions != 0) {
+        GlobalOrdinal region_index                                 = min_nregions_proc * tot_num_proc + (procID + 1);
         std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > tuple_aux = std::make_tuple(region_index, proc);
-        procs_per_region_.push_back( tuple_aux );
+        procs_per_region_.push_back(tuple_aux);
       }
     }
-    TEUCHOS_TEST_FOR_EXCEPTION( !( procs_per_region_.size()==num_total_regions_ ), Exceptions::RuntimeError, "PID: "<<comm_->getRank()<<" - Number of regions detected does not match with the initially declared one \n procs_per_region_ tracks "<<procs_per_region_.size()<<" regions whereas num_total_regions_ = "<<num_total_regions_<<"\n");
+    TEUCHOS_TEST_FOR_EXCEPTION(!(procs_per_region_.size() == num_total_regions_), Exceptions::RuntimeError, "PID: " << comm_->getRank() << " - Number of regions detected does not match with the initially declared one \n procs_per_region_ tracks " << procs_per_region_.size() << " regions whereas num_total_regions_ = " << num_total_regions_ << "\n");
   }
-  //This is easy: if the number of regions coincides with the total number of processes instantiated, then
-  //a one-to-one relation between processes and regions is created
-  else if( tot_num_proc == num_total_regions_ )
-  {
-    regions_per_proc_.push_back( myPID+1 );
-
-    for( int i = 0; i<num_total_regions_; ++i )
-    {
+  // This is easy: if the number of regions coincides with the total number of processes instantiated, then
+  // a one-to-one relation between processes and regions is created
+  else if (tot_num_proc == num_total_regions_) {
+    regions_per_proc_.push_back(myPID + 1);
+
+    for (int i = 0; i < num_total_regions_; ++i) {
       GlobalOrdinal region_index = i + 1;
       Array<GlobalOrdinal> proc;
       proc.clear();
       proc.push_back(i);
       std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > tuple_aux = std::make_tuple(region_index, proc);
-      procs_per_region_.push_back( tuple_aux );
+      procs_per_region_.push_back(tuple_aux);
     }
   }
-  //If the number of processes exceeds the number of regions in the domain,
-  //then each process is given a subset of a region.
-  //N.B.: A SINGLE PROCESS IS NOT ALLOWED TO OWN CHUNCKS OF MULTIPLE REGIONS.
-  //IN THIS CONFIGURATION EACH PROCESS IS CONFINED TO A SINGLE REGION
-  else if( tot_num_proc > num_total_regions_ )
-  {
-    int num_procs_region = std::ceil( static_cast<double>(tot_num_proc)/static_cast<double>(num_total_regions_) );
+  // If the number of processes exceeds the number of regions in the domain,
+  // then each process is given a subset of a region.
+  // N.B.: A SINGLE PROCESS IS NOT ALLOWED TO OWN CHUNCKS OF MULTIPLE REGIONS.
+  // IN THIS CONFIGURATION EACH PROCESS IS CONFINED TO A SINGLE REGION
+  else if (tot_num_proc > num_total_regions_) {
+    int num_procs_region       = std::ceil(static_cast<double>(tot_num_proc) / static_cast<double>(num_total_regions_));
     int num_regions_extra_proc = tot_num_proc % num_total_regions_;
-    int proc_count = 0;
+    int proc_count             = 0;
     std::tuple<int, Array<GlobalOrdinal> > region_tuple;
 
-    for( int i = 1; i<=num_total_regions_; ++i )
-    {
+    for (int i = 1; i <= num_total_regions_; ++i) {
       Array<GlobalOrdinal> procs;
       procs.clear();
-      if( i<=num_regions_extra_proc || num_regions_extra_proc==0 )
-        for( int j=1; j<=num_procs_region; ++j )
-        {
+      if (i <= num_regions_extra_proc || num_regions_extra_proc == 0)
+        for (int j = 1; j <= num_procs_region; ++j) {
           procs.push_back(proc_count);
           proc_count++;
         }
       else
-        for( int j=1; j<=num_procs_region-1; ++j )
-        {
+        for (int j = 1; j <= num_procs_region - 1; ++j) {
           procs.push_back(proc_count);
           proc_count++;
         }
@@ -244,13 +224,11 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeProcRegion
   }
 }
 
-
-//This routine associates a globally indexed node with the list of regions it belongs to
-//This is helpful to spot which nodes lie on a interregion interface. In fact, these nodes must have
-//the list of regions with more than one element
+// This routine associates a globally indexed node with the list of regions it belongs to
+// This is helpful to spot which nodes lie on a interregion interface. In fact, these nodes must have
+// the list of regions with more than one element
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodesToRegion()
-{
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodesToRegion() {
   nodesToRegion_.clear();
   interfaceNodes_.clear();
   Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > nodes_reordered;
@@ -260,29 +238,24 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodesToRegion()
   typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator node_iterator;
   node_iterator = nodes_reordered.begin();
 
-  while( node_iterator != nodes_reordered.end() )
-  {
-    GlobalOrdinal current_node = std::get<0>( *(node_iterator) );
+  while (node_iterator != nodes_reordered.end()) {
+    GlobalOrdinal current_node = std::get<0>(*(node_iterator));
     Array<GlobalOrdinal> regions;
     regions.clear();
-    regions.push_back( std::get<1>(*node_iterator) );
+    regions.push_back(std::get<1>(*node_iterator));
 
     typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator next_node_iterator = node_iterator + 1;
 
-    while( next_node_iterator != nodes_reordered.end() )
-    {
-      GlobalOrdinal next_node = std::get<0>( *(next_node_iterator) );
-      if( current_node == next_node )
-      {
-        //As long as the information spanned regards the same node,
-        //the algorithm keeps on increasing the list of regions the given mesh node belong to
-        regions.push_back( std::get<1>( *(next_node_iterator) ) );
+    while (next_node_iterator != nodes_reordered.end()) {
+      GlobalOrdinal next_node = std::get<0>(*(next_node_iterator));
+      if (current_node == next_node) {
+        // As long as the information spanned regards the same node,
+        // the algorithm keeps on increasing the list of regions the given mesh node belong to
+        regions.push_back(std::get<1>(*(next_node_iterator)));
         next_node_iterator++;
-      }
-      else
-      {
-        //When the mesh node label changes, then the algorithm
-        //stops recording information about the previous node and it starts recording information for the new one
+      } else {
+        // When the mesh node label changes, then the algorithm
+        // stops recording information about the previous node and it starts recording information for the new one
         node_iterator = next_node_iterator;
         break;
       }
@@ -292,24 +265,24 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodesToRegion()
     new_tuple = std::make_tuple(current_node, regions);
     nodesToRegion_.push_back(new_tuple);
 
-    if( regions.size()>1 )
+    if (regions.size() > 1)
       interfaceNodes_.push_back(new_tuple);
 
-    if( next_node_iterator == nodes_reordered.end() )
+    if (next_node_iterator == nodes_reordered.end())
       break;
   }
-  TEUCHOS_TEST_FOR_EXCEPTION( !( nodesToRegion_.size()==num_total_nodes_ ), Exceptions::RuntimeError, "Number of nodes detected does not match with the initially declared one \n"<<"nodesToRegion tracks "<<nodesToRegion_.size()<<" nodes whereas num_total_nodes_ ="<<num_total_nodes_<<"\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(!(nodesToRegion_.size() == num_total_nodes_), Exceptions::RuntimeError, "Number of nodes detected does not match with the initially declared one \n"
+                                                                                                         << "nodesToRegion tracks " << nodesToRegion_.size() << " nodes whereas num_total_nodes_ =" << num_total_nodes_ << "\n");
 }
 
-//This routine creates row maps for composite and region matrices
+// This routine creates row maps for composite and region matrices
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateRowMaps()
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( ( procs_per_region_.empty() && regions_per_proc_.empty() ), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Information about region partitioning across processors is not consistent: incorrect values for number of processors or number of regions \n");
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateRowMaps() {
+  TEUCHOS_TEST_FOR_EXCEPTION((procs_per_region_.empty() && regions_per_proc_.empty()), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Information about region partitioning across processors is not consistent: incorrect values for number of processors or number of regions \n");
   Array<GlobalOrdinal> elements;
   Array<GlobalOrdinal> region_elements;
   Array<Array<GlobalOrdinal> > elements_per_region;
-  Array<Array< std::tuple<GlobalOrdinal,GlobalOrdinal> > > regionToAll;
+  Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > regionToAll;
   int myPID = comm_->getRank();
 
   elements.clear();
@@ -319,285 +292,258 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateRowMaps()
   regionToAll.clear();
   regionToAll.resize(num_total_regions_);
 
-  TEUCHOS_TEST_FOR_EXCEPTION( !nodes_sorted_by_regions_, Exceptions::RuntimeError, "Nodes are not sorted by regions in ascending order \n");
-  TEUCHOS_TEST_FOR_EXCEPTION( num_total_nodes_>nodes_.size(), Exceptions::RuntimeError, "Number of nodes declared in input file does not match with the effective number of nodes provided\n"<<"num_total_nodes_ ="<<num_total_nodes_<<" whereas nodes_ tracks "<<nodes_.size()<<" nodes \n");
-  TEUCHOS_TEST_FOR_EXCEPTION( num_total_regions_!=std::get<1>( *(nodes_.end()-1) ), Exceptions::RuntimeError, "Number of regions declared in input file does not match with the effective number of regions provided\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(!nodes_sorted_by_regions_, Exceptions::RuntimeError, "Nodes are not sorted by regions in ascending order \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(num_total_nodes_ > nodes_.size(), Exceptions::RuntimeError, "Number of nodes declared in input file does not match with the effective number of nodes provided\n"
+                                                                                             << "num_total_nodes_ =" << num_total_nodes_ << " whereas nodes_ tracks " << nodes_.size() << " nodes \n");
+  TEUCHOS_TEST_FOR_EXCEPTION(num_total_regions_ != std::get<1>(*(nodes_.end() - 1)), Exceptions::RuntimeError, "Number of regions declared in input file does not match with the effective number of regions provided\n");
 
-  if( !( regions_per_proc_.empty() ) )
-  {
+  if (!(regions_per_proc_.empty())) {
     typename Array<GlobalOrdinal>::iterator iter_array;
-    for( iter_array=regions_per_proc_.begin(); iter_array!=regions_per_proc_.end(); ++iter_array )
-    {
+    for (iter_array = regions_per_proc_.begin(); iter_array != regions_per_proc_.end(); ++iter_array) {
       region_elements.clear();
       checkerNode<GlobalOrdinal> unaryPredicate(*iter_array);
-      typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator1;
-      typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator2;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator1;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator2;
 
-      //We position an iterator at the beginning of the information associated with the a region owned by the calling process
-      //and another iterator right at the end of the information
-      nodes_iterator1 = std::find_if<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
-      nodes_iterator2 = std::find_if_not<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
+      // We position an iterator at the beginning of the information associated with the a region owned by the calling process
+      // and another iterator right at the end of the information
+      nodes_iterator1 = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
+      nodes_iterator2 = std::find_if_not<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
 
-      //Coun the number of mesh nodes inside a region
+      // Coun the number of mesh nodes inside a region
       int num_region_nodes = nodes_iterator2 - nodes_iterator1;
 
-      typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator_aux;
+      typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator_aux;
 
-      //The policy assumes that in the input file the indexBase for the node label is 1
+      // The policy assumes that in the input file the indexBase for the node label is 1
       GlobalOrdinal region_node_label = 1;
-      for( nodes_iterator_aux=nodes_iterator1; nodes_iterator_aux!=nodes_iterator2; ++nodes_iterator_aux )
-      {
+      for (nodes_iterator_aux = nodes_iterator1; nodes_iterator_aux != nodes_iterator2; ++nodes_iterator_aux) {
         GlobalOrdinal node = std::get<0>(*nodes_iterator_aux);
         checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-        nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-        Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+        nodes_to_region_iterator           = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+        Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
 
-        //By default, I choose that a node is owned by the process associated with the region that shows up first in its list of beloning
-        //This guarantees that each row of the composite stiffness matrix is owned only by a single process, as Trilinos requires
-        if( *iter_array==nodal_regions[0] )
-          elements.push_back( node );
+        // By default, I choose that a node is owned by the process associated with the region that shows up first in its list of beloning
+        // This guarantees that each row of the composite stiffness matrix is owned only by a single process, as Trilinos requires
+        if (*iter_array == nodal_regions[0])
+          elements.push_back(node);
 
-        //Nodes on the interface still belong to multiple regions, so
-        //it is important to keep track of this for the row maps of region matrices
+        // Nodes on the interface still belong to multiple regions, so
+        // it is important to keep track of this for the row maps of region matrices
         region_elements.push_back(region_node_label);
 
-        //If a process owns a region (or even a portion of it), we provide to it a map
-        //from region indices to composite indices for all the nodes inside that region,
-        //even if a specific node is not owned by the calling process
-        regionToAll[*iter_array-1].push_back( std::make_tuple( region_node_label,std::get<0>(*nodes_iterator_aux) ) );
+        // If a process owns a region (or even a portion of it), we provide to it a map
+        // from region indices to composite indices for all the nodes inside that region,
+        // even if a specific node is not owned by the calling process
+        regionToAll[*iter_array - 1].push_back(std::make_tuple(region_node_label, std::get<0>(*nodes_iterator_aux)));
         region_node_label++;
       }
 
-      //C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
-      for( typename Array<GlobalOrdinal>::iterator iter = region_elements.begin(); iter!=region_elements.end(); ++iter )
+      // C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
+      for (typename Array<GlobalOrdinal>::iterator iter = region_elements.begin(); iter != region_elements.end(); ++iter)
         *iter = *iter - 1;
 
-      elements_per_region[*iter_array-1] = region_elements;
-      TEUCHOS_TEST_FOR_EXCEPTION( ( num_region_nodes!=regionToAll[*iter_array-1].size() ), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Number of region nodes does not match with number of nodes stored in regionToAll \n"<<"num_region_nodes= "<< num_region_nodes<<" whereas regionToAll["<<*iter_array-1<<"].size()= "<<regionToAll[*iter_array-1].size()<<"\n");
+      elements_per_region[*iter_array - 1] = region_elements;
+      TEUCHOS_TEST_FOR_EXCEPTION((num_region_nodes != regionToAll[*iter_array - 1].size()), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Number of region nodes does not match with number of nodes stored in regionToAll \n"
+                                                                                                                                     << "num_region_nodes= " << num_region_nodes << " whereas regionToAll[" << *iter_array - 1 << "].size()= " << regionToAll[*iter_array - 1].size() << "\n");
     }
-    TEUCHOS_TEST_FOR_EXCEPTION( ( num_total_regions_!=regionToAll.size() ), Exceptions::RuntimeError, "regionToAll size has been corrupted\n"<<"num_total_regions_ = "<<num_total_regions_<<" whereas regionToAll.size() = "<<regionToAll.size()<<"\n");
-  }
-  else
-  {
+    TEUCHOS_TEST_FOR_EXCEPTION((num_total_regions_ != regionToAll.size()), Exceptions::RuntimeError, "regionToAll size has been corrupted\n"
+                                                                                                         << "num_total_regions_ = " << num_total_regions_ << " whereas regionToAll.size() = " << regionToAll.size() << "\n");
+  } else {
     // The code enters the scope of these curly brackets if the number of processes exceeds the number of regions in the domain.
     // Therefore, each process owns only a subregion (or at most a single entire region).
     // In this situation, the calling process must identify the region it is associated with
-    bool region_found = false;
+    bool region_found      = false;
     GlobalOrdinal myRegion = -1;
-    TEUCHOS_TEST_FOR_EXCEPTION( !( procs_per_region_.size() == num_total_regions_ ), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - Number of total regions does not match with regionHandler structures \n");
+    TEUCHOS_TEST_FOR_EXCEPTION(!(procs_per_region_.size() == num_total_regions_), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - Number of total regions does not match with regionHandler structures \n");
 
     Array<GlobalOrdinal> region_procs;
-    while( !region_found )
-    {
+    while (!region_found) {
       typename Array<GlobalOrdinal>::iterator iter_proc;
-      for( GlobalOrdinal region_index=1; region_index<=procs_per_region_.size(); ++region_index )
-      {
-        region_procs = std::get<1>(procs_per_region_[region_index-1]);
-        iter_proc = std::find( region_procs.begin(), region_procs.end(), myPID );
-        if( iter_proc!=region_procs.end() )
-        {
-          myRegion = region_index;
+      for (GlobalOrdinal region_index = 1; region_index <= procs_per_region_.size(); ++region_index) {
+        region_procs = std::get<1>(procs_per_region_[region_index - 1]);
+        iter_proc    = std::find(region_procs.begin(), region_procs.end(), myPID);
+        if (iter_proc != region_procs.end()) {
+          myRegion     = region_index;
           region_found = true;
         }
       }
     }
 
-    TEUCHOS_TEST_FOR_EXCEPTION( ( myRegion == -1 || !region_found ), Exceptions::RuntimeError, ( "Region containing PROC ID: "+  std::to_string(myPID) + " NOT FOUND \n" ) );
-    region_procs = std::get<1>(procs_per_region_[myRegion-1]);
+    TEUCHOS_TEST_FOR_EXCEPTION((myRegion == -1 || !region_found), Exceptions::RuntimeError, ("Region containing PROC ID: " + std::to_string(myPID) + " NOT FOUND \n"));
+    region_procs = std::get<1>(procs_per_region_[myRegion - 1]);
 
     checkerNode<GlobalOrdinal> unaryPredicate(myRegion);
-    typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator1;
-    typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator nodes_iterator2;
-    nodes_iterator1 = std::find_if<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
-    nodes_iterator2 = std::find_if_not<typename Array< std::tuple<GlobalOrdinal,GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
+    typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator1;
+    typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator nodes_iterator2;
+    nodes_iterator1 = std::find_if<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_.begin(), nodes_.end(), unaryPredicate);
+    nodes_iterator2 = std::find_if_not<typename Array<std::tuple<GlobalOrdinal, GlobalOrdinal> >::iterator, checkerNode<GlobalOrdinal> >(nodes_iterator1, nodes_.end(), unaryPredicate);
 
     int num_region_nodes = nodes_iterator2 - nodes_iterator1;
     int num_region_procs = region_procs.size();
 
-    if( num_region_nodes < num_region_procs )
-    {
+    if (num_region_nodes < num_region_procs) {
       Array<GlobalOrdinal> region_procs_reduced;
       region_procs_reduced.clear();
-      for(int i = 0; i<num_region_nodes; ++i)
-        region_procs_reduced.push_back( region_procs[i] );
+      for (int i = 0; i < num_region_nodes; ++i)
+        region_procs_reduced.push_back(region_procs[i]);
 
       typename Array<GlobalOrdinal>::iterator proc_iterator;
       proc_iterator = std::find<typename Array<GlobalOrdinal>::iterator, GlobalOrdinal>(region_procs_reduced.begin(), region_procs_reduced.end(), myPID);
 
-      if( proc_iterator!=region_procs_reduced.end() )//This reasoning works because the PROC ID for each region has been previously sorted in ascending order
+      if (proc_iterator != region_procs_reduced.end())  // This reasoning works because the PROC ID for each region has been previously sorted in ascending order
       {
-        GlobalOrdinal node = std::get<0>( *( nodes_iterator1+(proc_iterator-region_procs_reduced.begin()+1) ) );
-        GlobalOrdinal region_node_label = proc_iterator-region_procs_reduced.begin()+1;
+        GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + (proc_iterator - region_procs_reduced.begin() + 1)));
+        GlobalOrdinal region_node_label = proc_iterator - region_procs_reduced.begin() + 1;
         checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-        nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-        Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+        nodes_to_region_iterator           = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+        Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
 
-        //The follolwing if statement is necessary to guarantee uniqueness of the RowMap used for the composite matrix
-        if( myRegion == nodal_regions[0] )
-          elements.push_back( node );
+        // The follolwing if statement is necessary to guarantee uniqueness of the RowMap used for the composite matrix
+        if (myRegion == nodal_regions[0])
+          elements.push_back(node);
 
-        //Although a process does not own a row in the composite matrix, it may still happen that it owns the row
-        //from a region matrix perspective
+        // Although a process does not own a row in the composite matrix, it may still happen that it owns the row
+        // from a region matrix perspective
         region_elements.push_back(region_node_label);
       }
 
-      //If a process owns a region (or even a portion of it), we provide to it a map
-      //from region indices to composite indices for all the nodes inside that region,
-      //even if a specific node is not owned by the calling process
-      //If a process owns something of a region, then the process has a global view of who owns what for that region
-      //Although this may seem more information than what actually needed, it is important for the computation of the collapsing.
-      //If the collapsing is not calculated, then this structure actually overestimates what a process needs to know.
-      for( proc_iterator=region_procs_reduced.begin(); proc_iterator!=region_procs_reduced.end(); ++proc_iterator )
-      {
-        GlobalOrdinal node = std::get<0>( *( nodes_iterator1+(proc_iterator-region_procs_reduced.begin()+1) ) );
-        GlobalOrdinal region_node_label = proc_iterator-region_procs_reduced.begin()+1;
+      // If a process owns a region (or even a portion of it), we provide to it a map
+      // from region indices to composite indices for all the nodes inside that region,
+      // even if a specific node is not owned by the calling process
+      // If a process owns something of a region, then the process has a global view of who owns what for that region
+      // Although this may seem more information than what actually needed, it is important for the computation of the collapsing.
+      // If the collapsing is not calculated, then this structure actually overestimates what a process needs to know.
+      for (proc_iterator = region_procs_reduced.begin(); proc_iterator != region_procs_reduced.end(); ++proc_iterator) {
+        GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + (proc_iterator - region_procs_reduced.begin() + 1)));
+        GlobalOrdinal region_node_label = proc_iterator - region_procs_reduced.begin() + 1;
         checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-        nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-        regionToAll[myRegion-1].push_back( std::make_tuple(region_node_label, node) );
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+        nodes_to_region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+        regionToAll[myRegion - 1].push_back(std::make_tuple(region_node_label, node));
       }
 
-    }
-    else if( num_region_nodes == num_region_procs )
-    {
+    } else if (num_region_nodes == num_region_procs) {
       typename Array<GlobalOrdinal>::iterator proc_iterator;
       proc_iterator = std::find<typename Array<GlobalOrdinal>::iterator, GlobalOrdinal>(region_procs.begin(), region_procs.end(), myPID);
 
-      if( proc_iterator!=region_procs.end() )//This reasoning works because the PROC ID for each region has been previously sorted in ascending order
+      if (proc_iterator != region_procs.end())  // This reasoning works because the PROC ID for each region has been previously sorted in ascending order
       {
-        GlobalOrdinal node = std::get<0>( *( nodes_iterator1+(proc_iterator-region_procs.begin()+1) ) );
-        GlobalOrdinal region_node_label = proc_iterator-region_procs.begin()+1;
+        GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + (proc_iterator - region_procs.begin() + 1)));
+        GlobalOrdinal region_node_label = proc_iterator - region_procs.begin() + 1;
         checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-        nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-        Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
-        if( myRegion == nodal_regions[0] )
-          elements.push_back( node );
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+        nodes_to_region_iterator           = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+        Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
+        if (myRegion == nodal_regions[0])
+          elements.push_back(node);
 
         region_elements.push_back(region_node_label);
       }
 
-      //If a process owns a region (or even a portion of it), we provide to it a map
-      //from region indices to composite indices for all the nodes inside that region,
-      //even if a specific node is not owned by the calling process
-      for( proc_iterator=region_procs.begin(); proc_iterator!=region_procs.end(); ++proc_iterator )
-      {
-        GlobalOrdinal node = std::get<0>( *( nodes_iterator1+(proc_iterator-region_procs.begin()+1) ) );
-        GlobalOrdinal region_node_label = proc_iterator-region_procs.begin()+1;
+      // If a process owns a region (or even a portion of it), we provide to it a map
+      // from region indices to composite indices for all the nodes inside that region,
+      // even if a specific node is not owned by the calling process
+      for (proc_iterator = region_procs.begin(); proc_iterator != region_procs.end(); ++proc_iterator) {
+        GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + (proc_iterator - region_procs.begin() + 1)));
+        GlobalOrdinal region_node_label = proc_iterator - region_procs.begin() + 1;
         checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-        typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-        nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-        regionToAll[myRegion-1].push_back( std::make_tuple(region_node_label, node) );
+        typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+        nodes_to_region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+        regionToAll[myRegion - 1].push_back(std::make_tuple(region_node_label, node));
       }
-    }
-    else
-    {
+    } else {
       typename Array<GlobalOrdinal>::iterator proc_iterator;
       proc_iterator = std::find<typename Array<GlobalOrdinal>::iterator, GlobalOrdinal>(region_procs.begin(), region_procs.end(), myPID);
 
-      int num_nodes_proc = std::ceil( static_cast<double>(num_region_nodes)/static_cast<double>(num_region_procs) );
+      int num_nodes_proc       = std::ceil(static_cast<double>(num_region_nodes) / static_cast<double>(num_region_procs));
       int num_procs_extra_node = num_region_nodes % num_region_procs;
 
-      if( proc_iterator-region_procs.begin()+1 <= num_procs_extra_node || num_procs_extra_node == 0 )
-      {
-        int init_node = num_nodes_proc * ( proc_iterator-region_procs.begin() );
-        for( int i=0; i<num_nodes_proc; ++i )
-        {
-          GlobalOrdinal node = std::get<0>( *( nodes_iterator1 + init_node + i ) );
+      if (proc_iterator - region_procs.begin() + 1 <= num_procs_extra_node || num_procs_extra_node == 0) {
+        int init_node = num_nodes_proc * (proc_iterator - region_procs.begin());
+        for (int i = 0; i < num_nodes_proc; ++i) {
+          GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + init_node + i));
           GlobalOrdinal region_node_label = init_node + i + 1;
           checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-          typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-          nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-          Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
-          if( myRegion == nodal_regions[0] )
-            elements.push_back( node );
+          typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+          nodes_to_region_iterator           = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+          Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
+          if (myRegion == nodal_regions[0])
+            elements.push_back(node);
 
           region_elements.push_back(region_node_label);
         }
-      }
-      else
-      {
-        int init_node = num_nodes_proc * num_procs_extra_node + (proc_iterator - region_procs.begin() - num_procs_extra_node) * (num_nodes_proc-1);
-        for( int i=0; i<num_nodes_proc-1; ++i )
-        {
-          GlobalOrdinal node = std::get<0>( *( nodes_iterator1 + init_node + i ) );
+      } else {
+        int init_node = num_nodes_proc * num_procs_extra_node + (proc_iterator - region_procs.begin() - num_procs_extra_node) * (num_nodes_proc - 1);
+        for (int i = 0; i < num_nodes_proc - 1; ++i) {
+          GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + init_node + i));
           GlobalOrdinal region_node_label = init_node + i + 1;
           checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-          typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-          nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-          Array<GlobalOrdinal> nodal_regions =  std::get<1>(*nodes_to_region_iterator);
-          if( myRegion == nodal_regions[0] )
-            elements.push_back( node );
+          typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+          nodes_to_region_iterator           = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+          Array<GlobalOrdinal> nodal_regions = std::get<1>(*nodes_to_region_iterator);
+          if (myRegion == nodal_regions[0])
+            elements.push_back(node);
 
           region_elements.push_back(region_node_label);
         }
       }
 
-      //If a process owns a region (or even a portion of it), we provide to it a map
-      //from region indices to composite indices for all the nodes inside that region,
-      //even if a specific node is not owned by the calling process
-      for( proc_iterator=region_procs.begin(); proc_iterator!=region_procs.end(); ++proc_iterator )
-      {
-        if( proc_iterator-region_procs.begin()+1 <= num_procs_extra_node || num_procs_extra_node == 0 )
-        {
-          int init_node = num_nodes_proc * ( proc_iterator-region_procs.begin() );
-          for( int i=0; i<num_nodes_proc; ++i )
-          {
-            GlobalOrdinal node = std::get<0>( *( nodes_iterator1 + init_node + i ) );
+      // If a process owns a region (or even a portion of it), we provide to it a map
+      // from region indices to composite indices for all the nodes inside that region,
+      // even if a specific node is not owned by the calling process
+      for (proc_iterator = region_procs.begin(); proc_iterator != region_procs.end(); ++proc_iterator) {
+        if (proc_iterator - region_procs.begin() + 1 <= num_procs_extra_node || num_procs_extra_node == 0) {
+          int init_node = num_nodes_proc * (proc_iterator - region_procs.begin());
+          for (int i = 0; i < num_nodes_proc; ++i) {
+            GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + init_node + i));
             GlobalOrdinal region_node_label = init_node + i + 1;
             checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-            typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-            nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-            regionToAll[myRegion-1].push_back( std::make_tuple(region_node_label, node) );
+            typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+            nodes_to_region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+            regionToAll[myRegion - 1].push_back(std::make_tuple(region_node_label, node));
           }
-        }
-        else
-        {
-          int init_node = num_nodes_proc * num_procs_extra_node + (proc_iterator - region_procs.begin() - num_procs_extra_node) * (num_nodes_proc-1);
-          for( int i=0; i<num_nodes_proc-1; ++i )
-          {
-            GlobalOrdinal node = std::get<0>( *( nodes_iterator1 + init_node + i ) );
+        } else {
+          int init_node = num_nodes_proc * num_procs_extra_node + (proc_iterator - region_procs.begin() - num_procs_extra_node) * (num_nodes_proc - 1);
+          for (int i = 0; i < num_nodes_proc - 1; ++i) {
+            GlobalOrdinal node              = std::get<0>(*(nodes_iterator1 + init_node + i));
             GlobalOrdinal region_node_label = init_node + i + 1;
             checkerNodesToRegion<GlobalOrdinal> unaryPredicateNode(node);
-            typename Array< std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
-            nodes_to_region_iterator = std::find_if<typename Array< std::tuple< GlobalOrdinal,Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
-            regionToAll[myRegion-1].push_back( std::make_tuple(region_node_label, node) );
+            typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator nodes_to_region_iterator;
+            nodes_to_region_iterator = std::find_if<typename Array<std::tuple<GlobalOrdinal, Array<GlobalOrdinal> > >::iterator, checkerNodesToRegion<GlobalOrdinal> >(nodesToRegion_.begin(), nodesToRegion_.end(), unaryPredicateNode);
+            regionToAll[myRegion - 1].push_back(std::make_tuple(region_node_label, node));
           }
         }
       }
-
     }
-    TEUCHOS_TEST_FOR_EXCEPTION( ( num_total_regions_!=regionToAll.size() ), Exceptions::RuntimeError, "Process ID: "<<comm_->getRank()<<" - regionToAll size has been corrupted\n"<<"num_total_regions_ = "<<num_total_regions_<<" whereas regionToAll.size()= "<<regionToAll.size()<<"\n");
+    TEUCHOS_TEST_FOR_EXCEPTION((num_total_regions_ != regionToAll.size()), Exceptions::RuntimeError, "Process ID: " << comm_->getRank() << " - regionToAll size has been corrupted\n"
+                                                                                                                    << "num_total_regions_ = " << num_total_regions_ << " whereas regionToAll.size()= " << regionToAll.size() << "\n");
 
-    //C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
-    for( typename Array<GlobalOrdinal>::iterator iter = region_elements.begin(); iter!=region_elements.end(); ++iter )
+    // C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
+    for (typename Array<GlobalOrdinal>::iterator iter = region_elements.begin(); iter != region_elements.end(); ++iter)
       *iter = *iter - 1;
 
-    elements_per_region[myRegion-1] = region_elements;
+    elements_per_region[myRegion - 1] = region_elements;
   }
 
-  //C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
-  for( typename Array<GlobalOrdinal>::iterator iter = elements.begin(); iter!=elements.end(); ++iter )
+  // C++ indexing starts from 0, so everything is shofted backward by one to make it consistent with programming language's policies
+  for (typename Array<GlobalOrdinal>::iterator iter = elements.begin(); iter != elements.end(); ++iter)
     *iter = *iter - 1;
 
   maps_.composite_map_ = elements;
-  maps_.region_maps_ = elements_per_region;
-  maps_.regionToAll_ = regionToAll;
+  maps_.region_maps_   = elements_per_region;
+  maps_.regionToAll_   = regionToAll;
 
-  for (int i = 0; i < comm_->getSize(); ++i)
-  {
+  for (int i = 0; i < comm_->getSize(); ++i) {
     comm_->barrier();
 
-    if (comm_->getRank() == i)
-    {
+    if (comm_->getRank() == i) {
       std::cout << "Proc " << i << std::endl;
-      for (int i = 0; i < regionToAll.size(); ++i)
-      {
-        Teuchos::Array<std::tuple<GlobalOrdinal,GlobalOrdinal> > currArray = regionToAll[i];
-        for (int j = 0; j < currArray.size(); ++j)
-        {
+      for (int i = 0; i < regionToAll.size(); ++i) {
+        Teuchos::Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > currArray = regionToAll[i];
+        for (int j = 0; j < currArray.size(); ++j) {
           std::cout << std::get<0>(currArray[j]) << "\t" << std::get<1>(currArray[j]) << std::endl;
         }
       }
@@ -607,68 +553,57 @@ void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateRowMaps()
   }
 }
 
-
 // Get methods to allow a user to interface with private members of the RegionHandler class
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Array<GlobalOrdinal> RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionRowMap(GlobalOrdinal region_index)const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( region_index>=num_total_regions_, Exceptions::RuntimeError, "Value of region index exceeds total number of regions stored \n"<<"Trying to access informaiton about region "<<region_index<<" when the total number of regions stored is "<<num_total_regions_<<"\n");
+Array<GlobalOrdinal> RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionRowMap(GlobalOrdinal region_index) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(region_index >= num_total_regions_, Exceptions::RuntimeError, "Value of region index exceeds total number of regions stored \n"
+                                                                                               << "Trying to access informaiton about region " << region_index << " when the total number of regions stored is " << num_total_regions_ << "\n");
   return maps_.region_maps_[region_index];
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll()const
-{
+Array<Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > > RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll() const {
   return maps_.regionToAll_;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll(GlobalOrdinal region_index)const
-{
-  TEUCHOS_TEST_FOR_EXCEPTION( region_index>=num_total_regions_, Exceptions::RuntimeError, "Value of region index exceeds total number of regions stored \n"<<"Trying to access informaiton about region "<<region_index<<" when the total number of regions stored is "<<num_total_regions_<<"\n");
+Array<std::tuple<GlobalOrdinal, GlobalOrdinal> > RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetRegionToAll(GlobalOrdinal region_index) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(region_index >= num_total_regions_, Exceptions::RuntimeError, "Value of region index exceeds total number of regions stored \n"
+                                                                                               << "Trying to access informaiton about region " << region_index << " when the total number of regions stored is " << num_total_regions_ << "\n");
   return maps_.regionToAll_[region_index];
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printView() const
-{
-  if( 0==comm_->getRank() )
-  {
-    std::cout<<"Total number of mesh nodes: "<<num_total_nodes_<<std::endl;
-    std::cout<<"Total number of mesh regions: "<<num_total_regions_<<std::endl;
-    std::cout<<"Number of rows in nodes_ structure: "<<nodes_.size()<<std::endl;
-    for( int i = 0; i < nodes_.size(); ++i )
-    {
-      std::cout<< std::get<0>(nodes_[i]) <<"\t"<< std::get<1>(nodes_[i]) <<std::endl;
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printView() const {
+  if (0 == comm_->getRank()) {
+    std::cout << "Total number of mesh nodes: " << num_total_nodes_ << std::endl;
+    std::cout << "Total number of mesh regions: " << num_total_regions_ << std::endl;
+    std::cout << "Number of rows in nodes_ structure: " << nodes_.size() << std::endl;
+    for (int i = 0; i < nodes_.size(); ++i) {
+      std::cout << std::get<0>(nodes_[i]) << "\t" << std::get<1>(nodes_[i]) << std::endl;
     }
   }
 }
 
-//Print methods
+// Print methods
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printNodesToRegion() const
-{
-  if( 0==comm_->getRank() )
-  {
-    std::cout<<"Total number of mesh nodes: "<<num_total_nodes_<<std::endl;
-    std::cout<<"Total number of mesh regions: "<<num_total_regions_<<std::endl;
-    std::cout<<"Number of rows in nodes_ structure: "<<nodes_.size()<<std::endl;
-    for( int i = 0; i < nodesToRegion_.size(); ++i )
-    {
-      std::cout<<"Node "<< std::get<0>(nodesToRegion_[i]) <<"\t belongs to regions: "<< std::get<1>(nodesToRegion_[i]) <<std::endl;
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printNodesToRegion() const {
+  if (0 == comm_->getRank()) {
+    std::cout << "Total number of mesh nodes: " << num_total_nodes_ << std::endl;
+    std::cout << "Total number of mesh regions: " << num_total_regions_ << std::endl;
+    std::cout << "Number of rows in nodes_ structure: " << nodes_.size() << std::endl;
+    for (int i = 0; i < nodesToRegion_.size(); ++i) {
+      std::cout << "Node " << std::get<0>(nodesToRegion_[i]) << "\t belongs to regions: " << std::get<1>(nodesToRegion_[i]) << std::endl;
     }
   }
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printInactive() const
-{
-  if( maps_.composite_map_.empty() )
-    std::cout<<"INACTIVE PROC ID: "<<comm_->getRank()<<std::endl;
+void RegionHandler<Scalar, LocalOrdinal, GlobalOrdinal, Node>::printInactive() const {
+  if (maps_.composite_map_.empty())
+    std::cout << "INACTIVE PROC ID: " << comm_->getRank() << std::endl;
 }
 
-} //namespace Xpetra
+}  // namespace Xpetra
 
 #endif
diff --git a/packages/muelu/research/q2q1/MueLu_Q2Q1PFactory.hpp b/packages/muelu/research/q2q1/MueLu_Q2Q1PFactory.hpp
index 53865418746d..5db995886422 100644
--- a/packages/muelu/research/q2q1/MueLu_Q2Q1PFactory.hpp
+++ b/packages/muelu/research/q2q1/MueLu_Q2Q1PFactory.hpp
@@ -60,182 +60,180 @@
 
 namespace MueLu {
 
-  template <class Scalar = MueLu::DefaultScalar,
-            class LocalOrdinal = MueLu::DefaultLocalOrdinal,
-            class GlobalOrdinal = MueLu::DefaultGlobalOrdinal,
-            class Node = MueLu::DefaultNode>
-  class Q2Q1PFactory : public PFactory {
+template <class Scalar        = MueLu::DefaultScalar,
+          class LocalOrdinal  = MueLu::DefaultLocalOrdinal,
+          class GlobalOrdinal = MueLu::DefaultGlobalOrdinal,
+          class Node          = MueLu::DefaultNode>
+class Q2Q1PFactory : public PFactory {
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    Q2Q1PFactory() { }
+  //! Constructor
+  Q2Q1PFactory() {}
 
-    //! Destructor.
-    virtual ~Q2Q1PFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~Q2Q1PFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
-  };
+  //@}
+};
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(fineLevel, "A");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1PFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel, "A");
+  RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
 
-    RCP<const Map>        rowMap  = A->getRowMap();
+  RCP<const Map> rowMap = A->getRowMap();
 
-    Xpetra::global_size_t N = rowMap->getGlobalNumElements();
+  Xpetra::global_size_t N = rowMap->getGlobalNumElements();
 
-    int V;
-    size_t n = as<size_t>(sqrt(N));
-    if (N == n*n) {
-      // pressure mode
-      V = 1;
-      GetOStream(Runtime1) << "Pressure mode" << std::endl;
+  int V;
+  size_t n = as<size_t>(sqrt(N));
+  if (N == n * n) {
+    // pressure mode
+    V = 1;
+    GetOStream(Runtime1) << "Pressure mode" << std::endl;
 
+  } else {
+    n = as<size_t>(sqrt(N / 2));
+
+    if (N == 2 * n * n) {
+      // velocity mode
+      V = 2;
+      GetOStream(Runtime1) << "Velocity mode" << std::endl;
     } else {
-      n = as<size_t>(sqrt(N/2));
-
-      if (N == 2*n*n) {
-        // velocity mode
-        V = 2;
-        GetOStream(Runtime1) << "Velocity mode" << std::endl;
-      } else {
-        throw Exceptions::RuntimeError("Matrix size (" + toString(N) + ") is incompatible with both velocity and pressure");
-      }
+      throw Exceptions::RuntimeError("Matrix size (" + toString(N) + ") is incompatible with both velocity and pressure");
     }
+  }
+
+  const int C              = 4;
+  Xpetra::global_size_t nc = (n - 1) / C + 1;
+  TEUCHOS_TEST_FOR_EXCEPTION(C * (nc - 1) + 1 != n, Exceptions::InvalidArgument, "Incorrect dim size: " << n);
 
-    const int             C = 4;
-    Xpetra::global_size_t nc = (n-1)/C + 1;
-    TEUCHOS_TEST_FOR_EXCEPTION(C*(nc-1)+1 != n, Exceptions::InvalidArgument, "Incorrect dim size: " << n);
+  ArrayView<const GO> elementList = rowMap->getLocalElementList();
+  GO indexBase                    = rowMap->getIndexBase();
 
+  // Calculate offsets
+  GO offset       = (V == 2 ? 0 : 2 * (2 * n - 1) * (2 * n - 1));
+  GO coarseOffset = (V == 2 ? 0 : 2 * (2 * nc - 1) * (2 * nc - 1));
 
-    ArrayView<const GO> elementList = rowMap->getLocalElementList();
-    GO indexBase = rowMap->getIndexBase();
+  GetOStream(Runtime1) << "offset = " << offset << ", coarseOffset = " << coarseOffset << std::endl;
 
-    // Calculate offsets
-    GO offset       = (V == 2 ? 0 : 2*(2*n -1)*(2*n -1));
-    GO coarseOffset = (V == 2 ? 0 : 2*(2*nc-1)*(2*nc-1));
+  Array<GO> coarseList;
+  for (LO k = 0; k < elementList.size(); k += V) {
+    GO GID = elementList[k] - offset - indexBase;
+    GO i = (GID / V) % n, ii = i / C;
+    GO j = (GID / V) / n, jj = j / C;
 
-    GetOStream(Runtime1) << "offset = " << offset << ", coarseOffset = " << coarseOffset << std::endl;
+    if (i % C == 0 && j % C == 0)
+      for (int q = 0; q < V; q++)
+        coarseList.push_back(V * (jj * nc + ii) + q + coarseOffset);
+  }
 
-    Array<GO> coarseList;
-    for (LO k = 0; k < elementList.size(); k += V) {
-      GO GID = elementList[k] - offset - indexBase;
-      GO i = (GID / V) % n, ii = i/C;
-      GO j = (GID / V) / n, jj = j/C;
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC one = STS::one();
+
+  Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+  std::vector<size_t> stridingInfo(1, 1);
+  const int stridedBlockId         = -1;
+  RCP<Map> coarseMap               = StridedMapFactory ::Build(rowMap->lib(), INVALID, coarseList, indexBase, stridingInfo, rowMap->getComm(), stridedBlockId, coarseOffset);
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, 1);
+  coarseNullspace->putScalar(one);
+
+  int nnzEstimate = 4;
+  RCP<Matrix> P   = MatrixFactory::Build(rowMap, coarseMap, nnzEstimate);
+
+  Array<GO> inds(nnzEstimate), inds1(nnzEstimate);
+  Array<SC> vals(nnzEstimate, one);
+  int sz;
+  for (LO k = 0; k < elementList.size(); k += V) {
+    GO GID = elementList[k] - offset - indexBase;
+    GO i = (GID / V) % n, ii = i / C;
+    GO j = (GID / V) / n, jj = j / C;
+
+    if (i % C == 0 && j % C == 0) {
+      sz      = 1;
+      inds[0] = jj * nc + ii;
+
+    } else if (i % C == 0 && j % C != 0) {
+      sz      = 2;
+      inds[0] = jj * nc + ii;
+      inds[1] = (jj + 1) * nc + ii;
+
+    } else if (i % C != 0 && j % C == 0) {
+      sz      = 2;
+      inds[0] = jj * nc + ii;
+      inds[1] = jj * nc + ii + 1;
 
-      if (i % C == 0 && j % C == 0)
-        for (int q = 0; q < V; q++)
-          coarseList.push_back(V*(jj*nc + ii) + q + coarseOffset);
+    } else {
+      sz      = 4;
+      inds[0] = jj * nc + ii;
+      inds[1] = jj * nc + ii + 1;
+      inds[2] = (jj + 1) * nc + ii;
+      inds[3] = (jj + 1) * nc + ii + 1;
     }
 
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC one = STS::one();
-
-    Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-    std::vector<size_t> stridingInfo(1,1);
-    const int           stridedBlockId = -1;
-    RCP<Map>         coarseMap       = StridedMapFactory ::Build(rowMap->lib(), INVALID, coarseList, indexBase, stridingInfo, rowMap->getComm(), stridedBlockId, coarseOffset);
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, 1);
-    coarseNullspace->putScalar(one);
-
-    int nnzEstimate = 4;
-    RCP<Matrix> P = MatrixFactory::Build(rowMap, coarseMap, nnzEstimate);
-
-    Array<GO> inds(nnzEstimate), inds1(nnzEstimate);
-    Array<SC> vals(nnzEstimate, one);
-    int sz;
-    for (LO k = 0; k < elementList.size(); k += V) {
-      GO GID = elementList[k] - offset - indexBase;
-      GO i = (GID/V) % n, ii = i/C;
-      GO j = (GID/V) / n, jj = j/C;
-
-      if        (i % C == 0 && j % C == 0) {
-        sz = 1;
-        inds[0] =  jj   *nc + ii  ;
-
-      } else if (i % C == 0 && j % C != 0) {
-        sz = 2;
-        inds[0] =  jj   *nc + ii  ;
-        inds[1] = (jj+1)*nc + ii  ;
-
-      } else if (i % C != 0 && j % C == 0) {
-        sz = 2;
-        inds[0] =  jj   *nc + ii  ;
-        inds[1] =  jj   *nc + ii+1;
-
-      } else {
-        sz = 4;
-        inds[0] =  jj   *nc + ii  ;
-        inds[1] =  jj   *nc + ii+1;
-        inds[2] = (jj+1)*nc + ii  ;
-        inds[3] = (jj+1)*nc + ii+1;
-      }
-
-      for (int q = 0; q < V; q++) {
-        for (int p = 0; p < sz; p++)
-          inds1[p] = V*inds[p]+q + coarseOffset;
-
-        P->insertGlobalValues(elementList[k]+q, inds1.view(0,sz), vals.view(0,sz));
-      }
-    }
+    for (int q = 0; q < V; q++) {
+      for (int p = 0; p < sz; p++)
+        inds1[p] = V * inds[p] + q + coarseOffset;
 
-    P->fillComplete(coarseMap, A->getDomainMap());
+      P->insertGlobalValues(elementList[k] + q, inds1.view(0, sz), vals.view(0, sz));
+    }
+  }
 
-    // Level Set
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-    Set(coarseLevel, "P",         P);
-    Set(fineLevel,   "CoarseMap", coarseMap);
+  P->fillComplete(coarseMap, A->getDomainMap());
 
-    if (IsPrint(Statistics1)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
-    }
+  // Level Set
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "P", P);
+  Set(fineLevel, "CoarseMap", coarseMap);
 
+  if (IsPrint(Statistics1)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+    GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_Q2Q1PFACTORY_DECL_HPP
+#endif  // MUELU_Q2Q1PFACTORY_DECL_HPP
diff --git a/packages/muelu/research/q2q1/MueLu_Q2Q1uPFactory.hpp b/packages/muelu/research/q2q1/MueLu_Q2Q1uPFactory.hpp
index a98daccd10f4..dc3f51a0e0b9 100644
--- a/packages/muelu/research/q2q1/MueLu_Q2Q1uPFactory.hpp
+++ b/packages/muelu/research/q2q1/MueLu_Q2Q1uPFactory.hpp
@@ -68,703 +68,695 @@
 
 namespace MueLu {
 
-  template<class LocalOrdinal = int>
-  class MyCptList_ {
-    typedef LocalOrdinal LO;
+template <class LocalOrdinal = int>
+class MyCptList_ {
+  typedef LocalOrdinal LO;
 
-  public:
-    MyCptList_(int n, int nnzPerRow = 100) {
-      TEUCHOS_TEST_FOR_EXCEPTION(nnzPerRow <= 0, Exceptions::RuntimeError, "Why is nnzPerRow " << nnzPerRow << "?");
+ public:
+  MyCptList_(int n, int nnzPerRow = 100) {
+    TEUCHOS_TEST_FOR_EXCEPTION(nnzPerRow <= 0, Exceptions::RuntimeError, "Why is nnzPerRow " << nnzPerRow << "?");
 
-      nnzPerRow_ = nnzPerRow;
-      storage_.resize(n * nnzPerRow);
-      numCpts_.resize(n, 0);
-
-      list_.resize(n, NULL);
-      for (int i = 0; i < n; i++)
-        list_[i] = &storage_[i*nnzPerRow];
-    }
-
-    size_t                    getLocalNumRows()  const { return list_.size(); }
-    int                       getNnzPerRow()    const { return nnzPerRow_;   }
-    std::vector<short>&       getNumCpts()            { return numCpts_;     }
-    Teuchos::Array<LO>&       getCList()              { return cptlist_;     }
-    const Teuchos::Array<LO>& getCList()        const { return cptlist_;     }
-    const std::vector<short>& getNumCpts()      const { return numCpts_;     }
-    LO*                       operator()(int i)       { return list_[i];     }
-    const LO*                 operator()(int i) const { return list_[i];     }
-
-  private:
-    std::vector<LO*>    list_;          // list[k] gives the CPOINTs that interpolate to the k-th fine point
-                                        // These CPOINTs are given as fine grid *local* indices
-    std::vector<short>  numCpts_;       // Number of CPOINTs for each point
-    Teuchos::Array<LO>  cptlist_;       // List of CPOINTs for each point
-    int                 nnzPerRow_;     // Max number of CPOINTs per row in order for a row to use storage_
-    std::vector<LO>     storage_;       // Large data array used to store most CPOINT information
-  };
+    nnzPerRow_ = nnzPerRow;
+    storage_.resize(n * nnzPerRow);
+    numCpts_.resize(n, 0);
 
+    list_.resize(n, NULL);
+    for (int i = 0; i < n; i++)
+      list_[i] = &storage_[i * nnzPerRow];
+  }
 
-  template <class Scalar        = MueLu::DefaultScalar,
-            class LocalOrdinal  = MueLu::DefaultLocalOrdinal,
-            class GlobalOrdinal = MueLu::DefaultGlobalOrdinal,
-            class Node          = MueLu::DefaultNode>
-  class Q2Q1uPFactory : public PFactory {
+  size_t getLocalNumRows() const { return list_.size(); }
+  int getNnzPerRow() const { return nnzPerRow_; }
+  std::vector<short>& getNumCpts() { return numCpts_; }
+  Teuchos::Array<LO>& getCList() { return cptlist_; }
+  const Teuchos::Array<LO>& getCList() const { return cptlist_; }
+  const std::vector<short>& getNumCpts() const { return numCpts_; }
+  LO* operator()(int i) { return list_[i]; }
+  const LO* operator()(int i) const { return list_[i]; }
+
+ private:
+  std::vector<LO*> list_;       // list[k] gives the CPOINTs that interpolate to the k-th fine point
+                                // These CPOINTs are given as fine grid *local* indices
+  std::vector<short> numCpts_;  // Number of CPOINTs for each point
+  Teuchos::Array<LO> cptlist_;  // List of CPOINTs for each point
+  int nnzPerRow_;               // Max number of CPOINTs per row in order for a row to use storage_
+  std::vector<LO> storage_;     // Large data array used to store most CPOINT information
+};
+
+template <class Scalar        = MueLu::DefaultScalar,
+          class LocalOrdinal  = MueLu::DefaultLocalOrdinal,
+          class GlobalOrdinal = MueLu::DefaultGlobalOrdinal,
+          class Node          = MueLu::DefaultNode>
+class Q2Q1uPFactory : public PFactory {
 #include "MueLu_UseShortNames.hpp"
-    typedef MyCptList_<LocalOrdinal> MyCptList;
-
-  private:
-    enum Status {
-      UNASSIGNED  = '0',
-      CANDIDATE   = '1',
-      FPOINT      = '2',
-      TWOTIMER    = '3',
-      CPOINT      = '4',
-      CPOINT_U    = '5'
-    };
+  typedef MyCptList_<LocalOrdinal> MyCptList;
+
+ private:
+  enum Status {
+    UNASSIGNED = '0',
+    CANDIDATE  = '1',
+    FPOINT     = '2',
+    TWOTIMER   = '3',
+    CPOINT     = '4',
+    CPOINT_U   = '5'
+  };
 
   std::string getStatusString(char status) const {
     switch (status) {
       case UNASSIGNED: return "UNASSIGNED";
-      case CANDIDATE : return "CANDIDATE";
-      case FPOINT    : return "FPOINT";
-      case TWOTIMER  : return "TWOTIMER";
-      case CPOINT    : return "CPOINT";
-      case CPOINT_U  : return "CPOINT_U";
-      default:         return "UNKNOWN";
+      case CANDIDATE: return "CANDIDATE";
+      case FPOINT: return "FPOINT";
+      case TWOTIMER: return "TWOTIMER";
+      case CPOINT: return "CPOINT";
+      case CPOINT_U: return "CPOINT_U";
+      default: return "UNKNOWN";
     }
   }
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    Q2Q1uPFactory() { }
+  //! Constructor
+  Q2Q1uPFactory() {}
 
-    //! Destructor.
-    virtual ~Q2Q1uPFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~Q2Q1uPFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    void FindDist4Cpts     (const Matrix& A, const MultiVector& coords, const Array<LO>& userCpts, std::vector<char>& status, MyCptList& myCpts, int levelID) const;
-    void PhaseTwoPattern   (const Matrix& A, const MultiVector& coords, const std::vector<char>& status, MyCptList& myCpts) const;
-    void FindMidPoints     (const Matrix& A, const MultiVector& coords, Array<LO>& Cptlist, const MyCptList& myCpts) const;
-    void CompDistances     (const Matrix& A, LO start, int numDist, std::vector<LO>& dist1, std::vector<LO>& dist2,
-                            std::vector<LO>& dist3, std::vector<LO>& dist4) const;
-    void CreateCrsPointers (const Matrix& A, ArrayRCP<const size_t>& ia, ArrayRCP<const LO>& ja) const;
-    void CptDepends2Pattern(const Matrix& A, const MyCptList& myCpts, RCP<Matrix>& P, LO offset) const;
+ private:
+  void FindDist4Cpts(const Matrix& A, const MultiVector& coords, const Array<LO>& userCpts, std::vector<char>& status, MyCptList& myCpts, int levelID) const;
+  void PhaseTwoPattern(const Matrix& A, const MultiVector& coords, const std::vector<char>& status, MyCptList& myCpts) const;
+  void FindMidPoints(const Matrix& A, const MultiVector& coords, Array<LO>& Cptlist, const MyCptList& myCpts) const;
+  void CompDistances(const Matrix& A, LO start, int numDist, std::vector<LO>& dist1, std::vector<LO>& dist2,
+                     std::vector<LO>& dist3, std::vector<LO>& dist4) const;
+  void CreateCrsPointers(const Matrix& A, ArrayRCP<const size_t>& ia, ArrayRCP<const LO>& ja) const;
+  void CptDepends2Pattern(const Matrix& A, const MyCptList& myCpts, RCP<Matrix>& P, LO offset) const;
 
-    void DumpStatus(const std::string& filename, const std::vector<char>& status, int NDim, bool isAmalgamated = true) const;
-    void DumpCoords(const std::string& filename, const MultiVector& coords) const;
-  };
+  void DumpStatus(const std::string& filename, const std::vector<char>& status, int NDim, bool isAmalgamated = true) const;
+  void DumpCoords(const std::string& filename, const MultiVector& coords) const;
+};
 
-  // Sort a double array and move along list2 to match sorted array
-  template<class SC>
-  void Muelu_az_dsort2(std::vector<SC>& dlist, std::vector<int>& list2) {
-    int l, r, j, i, flag;
-    int RR2;
-    SC  dRR, dK;
-
-    int N = dlist.size();
-    if (N <= 1) return;
-
-    l    = N / 2 + 1;
-    r    = N - 1;
-    l    = l - 1;
-    dRR  = dlist[l - 1];
-    dK   = dlist[l - 1];
-
-    if (list2.size()) {
-      RR2 = list2[l - 1];
-      while (r != 0) {
-        j = l;
-        flag = 1;
+// Sort a double array and move along list2 to match sorted array
+template <class SC>
+void Muelu_az_dsort2(std::vector<SC>& dlist, std::vector<int>& list2) {
+  int l, r, j, i, flag;
+  int RR2;
+  SC dRR, dK;
+
+  int N = dlist.size();
+  if (N <= 1) return;
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+  l   = N / 2 + 1;
+  r   = N - 1;
+  l   = l - 1;
+  dRR = dlist[l - 1];
+  dK  = dlist[l - 1];
 
-          if (j > r + 1)
+  if (list2.size()) {
+    RR2 = list2[l - 1];
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (dlist[j] > dlist[j - 1]) j = j + 1;
+
+          if (dlist[j - 1] > dK) {
+            dlist[i - 1] = dlist[j - 1];
+            list2[i - 1] = list2[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (dlist[j] > dlist[j - 1]) j = j + 1;
-
-            if (dlist[j - 1] > dK) {
-              dlist[i - 1] = dlist[j - 1];
-              list2[i - 1] = list2[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
-        dlist[i - 1] = dRR;
-        list2[i - 1] = RR2;
-
-        if (l == 1) {
-          dRR = dlist[r];
-          RR2 = list2[r];
-          dK = dlist[r];
-          dlist[r] = dlist[0];
-          list2[r] = list2[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          dRR = dlist[l - 1];
-          RR2 = list2[l - 1];
-          dK  = dlist[l - 1];
-        }
       }
-      dlist[0] = dRR;
-      list2[0] = RR2;
+      dlist[i - 1] = dRR;
+      list2[i - 1] = RR2;
+
+      if (l == 1) {
+        dRR      = dlist[r];
+        RR2      = list2[r];
+        dK       = dlist[r];
+        dlist[r] = dlist[0];
+        list2[r] = list2[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        dRR = dlist[l - 1];
+        RR2 = list2[l - 1];
+        dK  = dlist[l - 1];
+      }
     }
-    else {
-      while (r != 0) {
-        j = l;
-        flag = 1;
-        while (flag == 1) {
-          i = j;
-          j = j + j;
-          if (j > r + 1)
+    dlist[0] = dRR;
+    list2[0] = RR2;
+  } else {
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (dlist[j] > dlist[j - 1]) j = j + 1;
+          if (dlist[j - 1] > dK) {
+            dlist[i - 1] = dlist[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (dlist[j] > dlist[j - 1]) j = j + 1;
-            if (dlist[j - 1] > dK) {
-              dlist[i - 1] = dlist[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
-        dlist[i - 1] = dRR;
-        if (l == 1) {
-          dRR = dlist[r];
-          dK  = dlist[r];
-          dlist[r] = dlist[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          dRR  = dlist[l - 1];
-          dK   = dlist[l - 1];
-        }
       }
-      dlist[0] = dRR;
+      dlist[i - 1] = dRR;
+      if (l == 1) {
+        dRR      = dlist[r];
+        dK       = dlist[r];
+        dlist[r] = dlist[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        dRR = dlist[l - 1];
+        dK  = dlist[l - 1];
+      }
     }
+    dlist[0] = dRR;
   }
+}
+
+/* ******************************************************************* */
+/* sort an array and move along list2 and/or list to match sorted array*/
+/* ------------------------------------------------------------------- */
+template <class SC>
+void Muelu_az_sort(int list[], int N, int list2[], SC list3[]) {
+  int l, r, RR, K, j, i, flag;
+  int RR2;
+  SC RR3;
+
+  if (N <= 1) return;
+
+  l  = N / 2 + 1;
+  r  = N - 1;
+  l  = l - 1;
+  RR = list[l - 1];
+  K  = list[l - 1];
+
+  if ((list2 != NULL) && (list3 != NULL)) {
+    RR2 = list2[l - 1];
+    RR3 = list3[l - 1];
+    while (r != 0) {
+      j    = l;
+      flag = 1;
 
-  /* ******************************************************************* */
-  /* sort an array and move along list2 and/or list to match sorted array*/
-  /* ------------------------------------------------------------------- */
-  template<class SC>
-  void Muelu_az_sort(int list[], int N, int list2[], SC list3[]) {
-    int l, r, RR, K, j, i, flag;
-    int RR2;
-    SC  RR3;
-
-    if (N <= 1) return;
-
-    l   = N / 2 + 1;
-    r   = N - 1;
-    l   = l - 1;
-    RR  = list[l - 1];
-    K   = list[l - 1];
-
-    if ((list2 != NULL) && (list3 != NULL)) {
-      RR2 = list2[l - 1];
-      RR3 = list3[l - 1];
-      while (r != 0) {
-        j = l;
-        flag = 1;
+      while (flag == 1) {
+        i = j;
+        j = j + j;
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (list[j] > list[j - 1]) j = j + 1;
 
-          if (j > r + 1)
+          if (list[j - 1] > K) {
+            list[i - 1]  = list[j - 1];
+            list2[i - 1] = list2[j - 1];
+            list3[i - 1] = list3[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (list[j] > list[j - 1]) j = j + 1;
-
-            if (list[j - 1] > K) {
-              list [i - 1] = list [j - 1];
-              list2[i - 1] = list2[j - 1];
-              list3[i - 1] = list3[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
+      }
 
-        list [i - 1] = RR;
-        list2[i - 1] = RR2;
-        list3[i - 1] = RR3;
+      list[i - 1]  = RR;
+      list2[i - 1] = RR2;
+      list3[i - 1] = RR3;
 
-        if (l == 1) {
-          RR  = list [r];
-          RR2 = list2[r];
-          RR3 = list3[r];
+      if (l == 1) {
+        RR  = list[r];
+        RR2 = list2[r];
+        RR3 = list3[r];
 
-          K = list[r];
-          list[r ] = list[0];
-          list2[r] = list2[0];
-          list3[r] = list3[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          RR  = list [l - 1];
-          RR2 = list2[l - 1];
-          RR3 = list3[l - 1];
-          K   = list [l - 1];
-        }
+        K        = list[r];
+        list[r]  = list[0];
+        list2[r] = list2[0];
+        list3[r] = list3[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        RR  = list[l - 1];
+        RR2 = list2[l - 1];
+        RR3 = list3[l - 1];
+        K   = list[l - 1];
       }
-
-      list [0] = RR;
-      list2[0] = RR2;
-      list3[0] = RR3;
     }
-    else if (list2 != NULL) {
-      RR2 = list2[l - 1];
-      while (r != 0) {
-        j = l;
-        flag = 1;
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+    list[0]  = RR;
+    list2[0] = RR2;
+    list3[0] = RR3;
+  } else if (list2 != NULL) {
+    RR2 = list2[l - 1];
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (list[j] > list[j - 1]) j = j + 1;
 
-          if (j > r + 1)
+          if (list[j - 1] > K) {
+            list[i - 1]  = list[j - 1];
+            list2[i - 1] = list2[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (list[j] > list[j - 1]) j = j + 1;
-
-            if (list[j - 1] > K) {
-              list [i - 1] = list [j - 1];
-              list2[i - 1] = list2[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
+      }
 
-        list [i - 1] = RR;
-        list2[i - 1] = RR2;
+      list[i - 1]  = RR;
+      list2[i - 1] = RR2;
 
-        if (l == 1) {
-          RR  = list [r];
-          RR2 = list2[r];
+      if (l == 1) {
+        RR  = list[r];
+        RR2 = list2[r];
 
-          K = list[r];
-          list[r ] = list[0];
-          list2[r] = list2[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          RR  = list [l - 1];
-          RR2 = list2[l - 1];
-          K   = list [l - 1];
-        }
+        K        = list[r];
+        list[r]  = list[0];
+        list2[r] = list2[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        RR  = list[l - 1];
+        RR2 = list2[l - 1];
+        K   = list[l - 1];
       }
-
-      list [0] = RR;
-      list2[0] = RR2;
     }
-    else if (list3 != NULL) {
-      RR3 = list3[l - 1];
-      while (r != 0) {
-        j = l;
-        flag = 1;
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+    list[0]  = RR;
+    list2[0] = RR2;
+  } else if (list3 != NULL) {
+    RR3 = list3[l - 1];
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (list[j] > list[j - 1]) j = j + 1;
 
-          if (j > r + 1)
+          if (list[j - 1] > K) {
+            list[i - 1]  = list[j - 1];
+            list3[i - 1] = list3[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (list[j] > list[j - 1]) j = j + 1;
-
-            if (list[j - 1] > K) {
-              list [i - 1] = list [j - 1];
-              list3[i - 1] = list3[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
+      }
 
-        list [i - 1] = RR;
-        list3[i - 1] = RR3;
+      list[i - 1]  = RR;
+      list3[i - 1] = RR3;
 
-        if (l == 1) {
-          RR  = list [r];
-          RR3 = list3[r];
+      if (l == 1) {
+        RR  = list[r];
+        RR3 = list3[r];
 
-          K = list[r];
-          list[r ] = list[0];
-          list3[r] = list3[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          RR  = list [l - 1];
-          RR3 = list3[l - 1];
-          K   = list [l - 1];
-        }
+        K        = list[r];
+        list[r]  = list[0];
+        list3[r] = list3[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        RR  = list[l - 1];
+        RR3 = list3[l - 1];
+        K   = list[l - 1];
       }
+    }
 
-      list [0] = RR;
-      list3[0] = RR3;
+    list[0]  = RR;
+    list3[0] = RR3;
 
-    }
-    else {
-      while (r != 0) {
-        j = l;
-        flag = 1;
+  } else {
+    while (r != 0) {
+      j    = l;
+      flag = 1;
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (list[j] > list[j - 1]) j = j + 1;
 
-          if (j > r + 1)
+          if (list[j - 1] > K) {
+            list[i - 1] = list[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (list[j] > list[j - 1]) j = j + 1;
-
-            if (list[j - 1] > K) {
-              list[i - 1] = list[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
+      }
 
-        list[i - 1] = RR;
+      list[i - 1] = RR;
 
-        if (l == 1) {
-          RR  = list [r];
+      if (l == 1) {
+        RR = list[r];
 
-          K = list[r];
-          list[r ] = list[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          RR  = list[l - 1];
-          K   = list[l - 1];
-        }
+        K       = list[r];
+        list[r] = list[0];
+        r       = r - 1;
+      } else {
+        l  = l - 1;
+        RR = list[l - 1];
+        K  = list[l - 1];
       }
-
-      list[0] = RR;
     }
+
+    list[0] = RR;
   }
+}
+
+// Merge two already sorted lists into one combined sorted list.
+// NOTE: lists are given as integer arrays. These integer arrays give
+// locations in CoordDist[] defining the list values. That the ith value
+// associated with the Candidates list is actually CoordDist[Candidates[i]].
+template <class T>
+void MergeSort(std::vector<int>& oldCandidates, size_t numOldCandidates, const std::vector<int>& newCandidates, const std::vector<T>& coordDist, ArrayRCP<const size_t> ia) {
+  size_t numNewCandidates = newCandidates.size();
+  size_t numCandidates    = numOldCandidates + numNewCandidates;
+
+  oldCandidates.resize(numCandidates);
+
+  int i = numOldCandidates - 1;
+  int j = numNewCandidates - 1;
+  int k = numCandidates - 1;
+  while ((i >= 0) || (j >= 0)) {
+    if (i < 0)
+      oldCandidates[k--] = newCandidates[j--];
+    else if (j < 0)
+      oldCandidates[k--] = oldCandidates[i--];
+    else {
+      int ii = oldCandidates[i];
+      int jj = newCandidates[j];
 
-  // Merge two already sorted lists into one combined sorted list.
-  // NOTE: lists are given as integer arrays. These integer arrays give
-  // locations in CoordDist[] defining the list values. That the ith value
-  // associated with the Candidates list is actually CoordDist[Candidates[i]].
-  template<class T>
-  void MergeSort(std::vector<int>& oldCandidates, size_t numOldCandidates, const std::vector<int>& newCandidates, const std::vector<T>& coordDist, ArrayRCP<const size_t> ia) {
-    size_t numNewCandidates = newCandidates.size();
-    size_t numCandidates    = numOldCandidates + numNewCandidates;
-
-    oldCandidates.resize(numCandidates);
-
-    int i = numOldCandidates - 1;
-    int j = numNewCandidates - 1;
-    int k = numCandidates    - 1;
-    while ((i >= 0) || (j >= 0)) {
-      if      (i < 0) oldCandidates[k--] = newCandidates[j--];
-      else if (j < 0) oldCandidates[k--] = oldCandidates[i--];
-      else {
-        int ii = oldCandidates[i];
-        int jj = newCandidates[j];
-
-        // Must match code above. There is something arbitrary
-        // and crappy about the current weighting.
+      // Must match code above. There is something arbitrary
+      // and crappy about the current weighting.
 
 #ifdef optimal
-        if (-coordDist[ii] - .01*(ia[ii+1]-ia[ii]) + 1.e-10*(ii+1) <
-            -coordDist[jj] - .01*(ia[jj+1]-ia[jj]) + 1.e-10*(jj+1))
+      if (-coordDist[ii] - .01 * (ia[ii + 1] - ia[ii]) + 1.e-10 * (ii + 1) <
+          -coordDist[jj] - .01 * (ia[jj + 1] - ia[jj]) + 1.e-10 * (jj + 1))
 #else
-        if (coordDist[ii] - .0*(ia[ii+1]-ia[ii]) + 1.e-3*(ii+1) <
-            coordDist[jj] - .0*(ia[jj+1]-ia[jj]) + 1.e-3*(jj+1))
-        // if (ii < jj)
+      if (coordDist[ii] - .0 * (ia[ii + 1] - ia[ii]) + 1.e-3 * (ii + 1) <
+          coordDist[jj] - .0 * (ia[jj + 1] - ia[jj]) + 1.e-3 * (jj + 1))
+      // if (ii < jj)
 #endif
-          oldCandidates[k--] = newCandidates[j--];
-        else
-          oldCandidates[k--] = oldCandidates[i--];
-      }
+        oldCandidates[k--] = newCandidates[j--];
+      else
+        oldCandidates[k--] = oldCandidates[i--];
     }
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",                          null, "Generating factory of the matrix A");
-
-    RCP<const FactoryBase> rcpThis = rcpFromRef(*this);
-    validParamList->set< RCP<const FactoryBase> >("CoordinatesVelocity",     rcpThis, "Generating factory of the coordinates");
-    validParamList->set< RCP<const FactoryBase> >("AForPat",                 rcpThis, "Generating factory for Apattern");
-    validParamList->set< RCP<const FactoryBase> >("CoordinatesPressure",     rcpThis, "Generating factory of the coordinates");
-    validParamList->set< RCP<const FactoryBase> >("p2vMap",                  rcpThis, "Mapping of pressure coords to u-velocity coords");
-
-    validParamList->set< std::string >           ("mode",                 "pressure", "Mode");
-    validParamList->set< bool >                  ("phase2",                    false, "Use extra phase to improve pattern");
-    validParamList->set< bool >                  ("dump status",               false, "Output status");
-
-    validParamList->set< double >                ("tau_2",              sqrt(0.0015), "tau_2 parameter from the paper (used for mid points)");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(fineLevel, "A");
-
-    const ParameterList& pL = GetParameterList();
-    bool pressureMode = (pL.get<std::string>("mode") == "pressure");
-
-    // NOTE: we cannot simply do Input(fineLevel, "CoordinatePressure", as in
-    // valid parameter list we specified *this as the generating factory
-    if (fineLevel.GetLevelID()) {
-      if (pressureMode) {
-        Input(fineLevel, "CoordinatesPressure");
-      } else {
-        Input(fineLevel, "CoordinatesVelocity");
-        Input(fineLevel, "AForPat");
-        Input(fineLevel, "p2vMap");
-      }
-    }
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+  RCP<const FactoryBase> rcpThis = rcpFromRef(*this);
+  validParamList->set<RCP<const FactoryBase> >("CoordinatesVelocity", rcpThis, "Generating factory of the coordinates");
+  validParamList->set<RCP<const FactoryBase> >("AForPat", rcpThis, "Generating factory for Apattern");
+  validParamList->set<RCP<const FactoryBase> >("CoordinatesPressure", rcpThis, "Generating factory of the coordinates");
+  validParamList->set<RCP<const FactoryBase> >("p2vMap", rcpThis, "Mapping of pressure coords to u-velocity coords");
 
-    typedef Teuchos::ScalarTraits<SC> STS;
+  validParamList->set<std::string>("mode", "pressure", "Mode");
+  validParamList->set<bool>("phase2", false, "Use extra phase to improve pattern");
+  validParamList->set<bool>("dump status", false, "Output status");
 
-    const ParameterList& pL = GetParameterList();
-    bool pressureMode = (pL.get<std::string>("mode") == "pressure");
-    GetOStream(Runtime0) << (pressureMode ? "Pressure" : "Velocity") << " mode" << std::endl;
+  validParamList->set<double>("tau_2", sqrt(0.0015), "tau_2 parameter from the paper (used for mid points)");
 
-    bool fineLevelID = fineLevel.GetLevelID();
+  return validParamList;
+}
 
-    RCP<Matrix>           A = Get< RCP<Matrix> >(fineLevel, "A");
-    Xpetra::global_size_t N = A->getRowMap()->getGlobalNumElements();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
 
-    RCP<MyCptList>    myCpts = rcp(new MyCptList(N));
-    std::vector<char> status(N, UNASSIGNED);
+  const ParameterList& pL = GetParameterList();
+  bool pressureMode       = (pL.get<std::string>("mode") == "pressure");
 
-    RCP<MultiVector> coords;
-    RCP<Matrix>      AForPat;
-    int NDim = -1;
+  // NOTE: we cannot simply do Input(fineLevel, "CoordinatePressure", as in
+  // valid parameter list we specified *this as the generating factory
+  if (fineLevel.GetLevelID()) {
     if (pressureMode) {
-      if (fineLevelID == 0) coords  = fineLevel.Get< RCP<MultiVector> > ("CoordinatesPressure", NoFactory::get());
-      else                  coords  = Get< RCP<MultiVector> >(fineLevel, "CoordinatesPressure");
-      NDim = coords->getNumVectors();
-
-      Array<LO> userCpts; // pressure does not reuse any CPOINTs
-      FindDist4Cpts(*A, *coords, userCpts, status, *myCpts, fineLevelID);
-
-      if (pL.get<bool>("phase2")) {
-        // Beef up any limited pattern
-        PhaseTwoPattern(*A, *coords, status, *myCpts);
-      }
-
+      Input(fineLevel, "CoordinatesPressure");
     } else {
-      // Do all the coarsening/pattern stuff on amalgamated velocities.
-      // We need to guarantee that all velocity dofs are treated identically
-      // This means that we must amalgmate AForPat and the velocity coordinates
-      if (fineLevelID == 0) coords  = fineLevel.Get< RCP<MultiVector> > ("CoordinatesVelocity", NoFactory::get());
-      else                  coords  = Get< RCP<MultiVector> >(fineLevel, "CoordinatesVelocity");
-      if (fineLevelID == 0) AForPat = fineLevel.Get< RCP<Matrix> >      ("AForPat",             NoFactory::get());
-      else                  AForPat = Get< RCP<Matrix> >     (fineLevel, "AForPat");
-      NDim = coords->getNumVectors();
+      Input(fineLevel, "CoordinatesVelocity");
+      Input(fineLevel, "AForPat");
+      Input(fineLevel, "p2vMap");
+    }
+  }
+}
 
-      TEUCHOS_TEST_FOR_EXCEPTION(!coarseLevel.IsAvailable("PresCptsAndMids"), Exceptions::RuntimeError,
-                                 "Pressure points are not available");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
-      Array<LO> userCpts = coarseLevel.Get<Array<LO> >("PresCptsAndMids");
-      GetOStream(Runtime1) << "Found stored pressure C-points: " << userCpts.size() << " " << userCpts << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-      TEUCHOS_TEST_FOR_EXCEPTION(N % NDim, Exceptions::RuntimeError, "Number of velocity DOFs is odd");
-      Xpetra::global_size_t NN = N/NDim;
+  typedef Teuchos::ScalarTraits<SC> STS;
 
-      std::vector<GO> gNodeIds(NN);
-      for (size_t k = 0; k < NN; k++)
-        gNodeIds[k] = k;
+  const ParameterList& pL = GetParameterList();
+  bool pressureMode       = (pL.get<std::string>("mode") == "pressure");
+  GetOStream(Runtime0) << (pressureMode ? "Pressure" : "Velocity") << " mode" << std::endl;
 
-      RCP<Map> nodeMap = MapFactory::Build(AForPat->getRowMap()->lib(), NN, gNodeIds, 0, AForPat->getRowMap()->getComm());
+  bool fineLevelID = fineLevel.GetLevelID();
 
-      // FIXME: remove magic number 30
-      RCP<Matrix>    amalgA     = MatrixFactory::Build(nodeMap, nodeMap, 30);
-      RCP<CrsMatrix> amalgA_crs = rcp_dynamic_cast<CrsMatrixWrap>(amalgA)->getCrsMatrix();
+  RCP<Matrix> A           = Get<RCP<Matrix> >(fineLevel, "A");
+  Xpetra::global_size_t N = A->getRowMap()->getGlobalNumElements();
 
-      // FIXME: this should be written similar to CoalesceDropFactory Merge
-      for (LO row = 0; row < as<LO>(AForPat->getRowMap()->getLocalNumElements()); row += NDim) {
-        GO grid = AForPat->getRowMap()->getGlobalElement(row);
-        GO currentId = grid/NDim;
+  RCP<MyCptList> myCpts = rcp(new MyCptList(N));
+  std::vector<char> status(N, UNASSIGNED);
 
-        Teuchos::ArrayView<const LO> inds;
-        Teuchos::ArrayView<const SC> vals;
-        AForPat->getLocalRowView(row, inds, vals);
+  RCP<MultiVector> coords;
+  RCP<Matrix> AForPat;
+  int NDim = -1;
+  if (pressureMode) {
+    if (fineLevelID == 0)
+      coords = fineLevel.Get<RCP<MultiVector> >("CoordinatesPressure", NoFactory::get());
+    else
+      coords = Get<RCP<MultiVector> >(fineLevel, "CoordinatesPressure");
+    NDim = coords->getNumVectors();
 
-        size_t nnz = inds.size();
+    Array<LO> userCpts;  // pressure does not reuse any CPOINTs
+    FindDist4Cpts(*A, *coords, userCpts, status, *myCpts, fineLevelID);
 
-        // Count the number of nonzero block columns in this row
-        // NOTE: this assumes that blocks are dense, i.e. that if one column is
-        // nonzero, then all columns in the same block are nonzeros
-        LO realnnz = 0;
-        for (LO col = 0; col < Teuchos::as<LO>(nnz); col++)
-          if (inds[col] % NDim == 0)
-            realnnz++;
+    if (pL.get<bool>("phase2")) {
+      // Beef up any limited pattern
+      PhaseTwoPattern(*A, *coords, status, *myCpts);
+    }
 
-        if (realnnz == 0)
-          continue;
+  } else {
+    // Do all the coarsening/pattern stuff on amalgamated velocities.
+    // We need to guarantee that all velocity dofs are treated identically
+    // This means that we must amalgmate AForPat and the velocity coordinates
+    if (fineLevelID == 0)
+      coords = fineLevel.Get<RCP<MultiVector> >("CoordinatesVelocity", NoFactory::get());
+    else
+      coords = Get<RCP<MultiVector> >(fineLevel, "CoordinatesVelocity");
+    if (fineLevelID == 0)
+      AForPat = fineLevel.Get<RCP<Matrix> >("AForPat", NoFactory::get());
+    else
+      AForPat = Get<RCP<Matrix> >(fineLevel, "AForPat");
+    NDim = coords->getNumVectors();
+
+    TEUCHOS_TEST_FOR_EXCEPTION(!coarseLevel.IsAvailable("PresCptsAndMids"), Exceptions::RuntimeError,
+                               "Pressure points are not available");
+
+    Array<LO> userCpts = coarseLevel.Get<Array<LO> >("PresCptsAndMids");
+    GetOStream(Runtime1) << "Found stored pressure C-points: " << userCpts.size() << " " << userCpts << std::endl;
+
+    TEUCHOS_TEST_FOR_EXCEPTION(N % NDim, Exceptions::RuntimeError, "Number of velocity DOFs is odd");
+    Xpetra::global_size_t NN = N / NDim;
+
+    std::vector<GO> gNodeIds(NN);
+    for (size_t k = 0; k < NN; k++)
+      gNodeIds[k] = k;
+
+    RCP<Map> nodeMap = MapFactory::Build(AForPat->getRowMap()->lib(), NN, gNodeIds, 0, AForPat->getRowMap()->getComm());
+
+    // FIXME: remove magic number 30
+    RCP<Matrix> amalgA        = MatrixFactory::Build(nodeMap, nodeMap, 30);
+    RCP<CrsMatrix> amalgA_crs = rcp_dynamic_cast<CrsMatrixWrap>(amalgA)->getCrsMatrix();
+
+    // FIXME: this should be written similar to CoalesceDropFactory Merge
+    for (LO row = 0; row < as<LO>(AForPat->getRowMap()->getLocalNumElements()); row += NDim) {
+      GO grid      = AForPat->getRowMap()->getGlobalElement(row);
+      GO currentId = grid / NDim;
+
+      Teuchos::ArrayView<const LO> inds;
+      Teuchos::ArrayView<const SC> vals;
+      AForPat->getLocalRowView(row, inds, vals);
+
+      size_t nnz = inds.size();
+
+      // Count the number of nonzero block columns in this row
+      // NOTE: this assumes that blocks are dense, i.e. that if one column is
+      // nonzero, then all columns in the same block are nonzeros
+      LO realnnz = 0;
+      for (LO col = 0; col < Teuchos::as<LO>(nnz); col++)
+        if (inds[col] % NDim == 0)
+          realnnz++;
+
+      if (realnnz == 0)
+        continue;
 
-        Teuchos::Array<GO>  cnodeIds(realnnz, 0);
-        Teuchos::Array<SC>  ones(realnnz, STS::one()); //Pattern has all 1's
+      Teuchos::Array<GO> cnodeIds(realnnz, 0);
+      Teuchos::Array<SC> ones(realnnz, STS::one());  // Pattern has all 1's
 
-        realnnz = 0;
-        for (LO col = 0; col < Teuchos::as<LO>(nnz); col++) {
-          if (inds[col] % NDim == 0) {
-            GO gcid = AForPat->getColMap()->getGlobalElement(inds[col]);
-            cnodeIds[realnnz++] = gcid/NDim;
-          }
+      realnnz = 0;
+      for (LO col = 0; col < Teuchos::as<LO>(nnz); col++) {
+        if (inds[col] % NDim == 0) {
+          GO gcid             = AForPat->getColMap()->getGlobalElement(inds[col]);
+          cnodeIds[realnnz++] = gcid / NDim;
         }
-        amalgA_crs->insertGlobalValues(currentId, cnodeIds, ones);
       }
-      amalgA_crs->fillComplete(nodeMap, nodeMap);
+      amalgA_crs->insertGlobalValues(currentId, cnodeIds, ones);
+    }
+    amalgA_crs->fillComplete(nodeMap, nodeMap);
 
-      // Amalgmate the velocity coordinates
-      // NOTE: This assumes that the original coords vector contains duplicated (x NDim) degrees of freedom
-      RCP<MultiVector> amalgCoords = Xpetra::MultiVectorFactory<SC,LO,GO,Node>::Build(nodeMap, NDim);
+    // Amalgmate the velocity coordinates
+    // NOTE: This assumes that the original coords vector contains duplicated (x NDim) degrees of freedom
+    RCP<MultiVector> amalgCoords = Xpetra::MultiVectorFactory<SC, LO, GO, Node>::Build(nodeMap, NDim);
 
-      for (int j = 0; j < NDim; j++) {
-        ArrayRCP<SC> coordView      = coords     ->getDataNonConst(j);
-        ArrayRCP<SC> amalgCoordView = amalgCoords->getDataNonConst(j);
-        for (size_t k = 0; k < NN; k++)
-          amalgCoordView[k] = coordView[k*NDim];
-      }
+    for (int j = 0; j < NDim; j++) {
+      ArrayRCP<SC> coordView      = coords->getDataNonConst(j);
+      ArrayRCP<SC> amalgCoordView = amalgCoords->getDataNonConst(j);
+      for (size_t k = 0; k < NN; k++)
+        amalgCoordView[k] = coordView[k * NDim];
+    }
 
-      // On the finest level, we must map userCpts (which corresponds to
-      // pressure cpts and pressure mid-points) to the velocity variables
-      //
-      // NOTE: on coarser levels the lower numbered velocity dofs correspond
-      // to points that are co-located with pressures and the two numberings
-      // are identical so no translation is needed.
-      if (fineLevelID == 0) {
-        ArrayRCP<LO> p2vMap = fineLevel.Get< ArrayRCP<LO> >("p2vMap", NoFactory::get());
-
-        for (int k = 0; k < userCpts.size(); k++)
-          userCpts[k] = p2vMap[userCpts[k]]/NDim;
-      }
+    // On the finest level, we must map userCpts (which corresponds to
+    // pressure cpts and pressure mid-points) to the velocity variables
+    //
+    // NOTE: on coarser levels the lower numbered velocity dofs correspond
+    // to points that are co-located with pressures and the two numberings
+    // are identical so no translation is needed.
+    if (fineLevelID == 0) {
+      ArrayRCP<LO> p2vMap = fineLevel.Get<ArrayRCP<LO> >("p2vMap", NoFactory::get());
+
+      for (int k = 0; k < userCpts.size(); k++)
+        userCpts[k] = p2vMap[userCpts[k]] / NDim;
+    }
 
-      GetOStream(Runtime1) << "Amalgamated velocity C-points: " << userCpts.size() << " " << userCpts << std::endl;
+    GetOStream(Runtime1) << "Amalgamated velocity C-points: " << userCpts.size() << " " << userCpts << std::endl;
 
-      // Now determine velocity CPOINTs for amalgamated system
-      RCP<MyCptList>    amalgCpts  = rcp(new MyCptList(NN));
-      std::vector<char> amalgStatus(NN, UNASSIGNED);
+    // Now determine velocity CPOINTs for amalgamated system
+    RCP<MyCptList> amalgCpts = rcp(new MyCptList(NN));
+    std::vector<char> amalgStatus(NN, UNASSIGNED);
 
-      FindDist4Cpts(*amalgA, *amalgCoords, userCpts, amalgStatus, *amalgCpts, fineLevelID);
+    FindDist4Cpts(*amalgA, *amalgCoords, userCpts, amalgStatus, *amalgCpts, fineLevelID);
 
-      if (pL.get<bool>("phase2")) {
-        // Beef up any limited pattern
-        PhaseTwoPattern(*amalgA, *amalgCoords, amalgStatus, *amalgCpts);
-      }
+    if (pL.get<bool>("phase2")) {
+      // Beef up any limited pattern
+      PhaseTwoPattern(*amalgA, *amalgCoords, amalgStatus, *amalgCpts);
+    }
 
-      // Unamalgamate data
-      Array<LO>&          Cptlist      = myCpts   ->getCList();
-      Array<LO>&          amalgCptlist = amalgCpts->getCList();
-      std::vector<short>& numCpts      = myCpts   ->getNumCpts();
-      std::vector<short>& amalgNumCpts = amalgCpts->getNumCpts();
+    // Unamalgamate data
+    Array<LO>& Cptlist               = myCpts->getCList();
+    Array<LO>& amalgCptlist          = amalgCpts->getCList();
+    std::vector<short>& numCpts      = myCpts->getNumCpts();
+    std::vector<short>& amalgNumCpts = amalgCpts->getNumCpts();
 
-      int p = amalgCptlist.size();
+    int p = amalgCptlist.size();
 
-      Cptlist.resize(p*NDim);
-      for (int k = 0; k < p; k++) {
-        Cptlist[k*NDim] = amalgCptlist[k] * NDim;
-        for (int j = 1; j < NDim; j++)
-          Cptlist[k*NDim+j] = Cptlist[k*NDim] + j;
-      }
+    Cptlist.resize(p * NDim);
+    for (int k = 0; k < p; k++) {
+      Cptlist[k * NDim] = amalgCptlist[k] * NDim;
+      for (int j = 1; j < NDim; j++)
+        Cptlist[k * NDim + j] = Cptlist[k * NDim] + j;
+    }
 
-      for (Xpetra::global_size_t i = 0; i < NN; i++) {
-        for (int j = 0; j < NDim; j++) {
-          status [N-1-(i*NDim+j)] = amalgStatus [NN-1-(i)];
-          numCpts[i*NDim+j]       = amalgNumCpts[i];
-          for (int k = 0; k < amalgNumCpts[i]; k++)
-            (*myCpts)(i*NDim+j)[k] = (*amalgCpts)(i)[k]*NDim + j;
-        }
+    for (Xpetra::global_size_t i = 0; i < NN; i++) {
+      for (int j = 0; j < NDim; j++) {
+        status[N - 1 - (i * NDim + j)] = amalgStatus[NN - 1 - (i)];
+        numCpts[i * NDim + j]          = amalgNumCpts[i];
+        for (int k = 0; k < amalgNumCpts[i]; k++)
+          (*myCpts)(i * NDim + j)[k] = (*amalgCpts)(i)[k] * NDim + j;
       }
     }
+  }
 
-    const bool doStatusOutput = pL.get<bool>("dump status");
-    if (doStatusOutput) {
-      const Array<LO>&    Cptlist = myCpts->getCList();
-      std::vector<short>& numCpts = myCpts->getNumCpts();
-
-      std::string depPrefix = std::string("dep0-l") + toString(fineLevel.GetLevelID()) + (pressureMode ? "-p-" : "-v-");
-
-      std::vector<char> depStatus(N);
-      // Graph is unamalgamted, so we need to skip some CPOINTs as they are
-      // essentially duplicated for different velocities
-      for (int k = 0; k < Cptlist.size(); k += NDim) {
+  const bool doStatusOutput = pL.get<bool>("dump status");
+  if (doStatusOutput) {
+    const Array<LO>& Cptlist    = myCpts->getCList();
+    std::vector<short>& numCpts = myCpts->getNumCpts();
 
-        for (Xpetra::global_size_t i = 0; i < N; i++) {
-          bool isPresent = false;
-          for (int j = 0; j < numCpts[i]; j++)
-            if ((*myCpts)(i)[j] == Cptlist[k])
-              isPresent = true;
-          depStatus[i] = (isPresent ? FPOINT : UNASSIGNED);
-        }
-        depStatus[Cptlist[k]] = CPOINT;
+    std::string depPrefix = std::string("dep0-l") + toString(fineLevel.GetLevelID()) + (pressureMode ? "-p-" : "-v-");
 
-        DumpStatus(depPrefix + toString(k), depStatus, NDim, false);
+    std::vector<char> depStatus(N);
+    // Graph is unamalgamted, so we need to skip some CPOINTs as they are
+    // essentially duplicated for different velocities
+    for (int k = 0; k < Cptlist.size(); k += NDim) {
+      for (Xpetra::global_size_t i = 0; i < N; i++) {
+        bool isPresent = false;
+        for (int j = 0; j < numCpts[i]; j++)
+          if ((*myCpts)(i)[j] == Cptlist[k])
+            isPresent = true;
+        depStatus[i] = (isPresent ? FPOINT : UNASSIGNED);
       }
+      depStatus[Cptlist[k]] = CPOINT;
+
+      DumpStatus(depPrefix + toString(k), depStatus, NDim, false);
     }
+  }
 
-    RCP<Matrix> P;
-    // FIXME :hardwired hack, pressure gids must not overlap with velocity gids
-    if (pressureMode) CptDepends2Pattern(*A,       *myCpts, P, 999999);
-    else              CptDepends2Pattern(*AForPat, *myCpts, P, 0);
+  RCP<Matrix> P;
+  // FIXME :hardwired hack, pressure gids must not overlap with velocity gids
+  if (pressureMode)
+    CptDepends2Pattern(*A, *myCpts, P, 999999);
+  else
+    CptDepends2Pattern(*AForPat, *myCpts, P, 0);
 
 #if 0
     if (pressureMode) {
@@ -776,1080 +768,1079 @@ namespace MueLu {
     }
 #endif
 
-    // Construct coarse map
-    RCP<const Map> coarseMap = P->getDomainMap();
+  // Construct coarse map
+  RCP<const Map> coarseMap = P->getDomainMap();
 
-    // Construct coarse nullspace
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, 1);
-    coarseNullspace->putScalar(STS::one());
+  // Construct coarse nullspace
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, 1);
+  coarseNullspace->putScalar(STS::one());
 
-    // Construct coarse coordinates
-    const Array<LO>& Cptlist = myCpts->getCList();
-    RCP<MultiVector> coarseCoords    = MultiVectorFactory::Build(coarseMap, NDim);
-    for (int k = 0; k < NDim; k++) {
-      ArrayRCP<const SC> coords1D       = coords      ->getData(k);
-      ArrayRCP<SC>       coarseCoords1D = coarseCoords->getDataNonConst(k);
+  // Construct coarse coordinates
+  const Array<LO>& Cptlist      = myCpts->getCList();
+  RCP<MultiVector> coarseCoords = MultiVectorFactory::Build(coarseMap, NDim);
+  for (int k = 0; k < NDim; k++) {
+    ArrayRCP<const SC> coords1D = coords->getData(k);
+    ArrayRCP<SC> coarseCoords1D = coarseCoords->getDataNonConst(k);
 
-      for (int i = 0; i < coarseCoords1D.size(); i++)
-        coarseCoords1D[i] = coords1D[Cptlist[i]];
-    }
-
-    // Level Set
-    Set(coarseLevel, "P",           P);
-    Set(fineLevel,   "CoarseMap",   coarseMap);
-    if (pressureMode) {
-      Set(coarseLevel, "CoordinatesPressure", coarseCoords);
-
-    } else {
-      Set(coarseLevel, "CoordinatesVelocity", coarseCoords);
-      // FIXME: why does coarse pattern matrix look like?
-      RCP<Matrix> AP  = Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*AForPat, false, *P, false, GetOStream(Statistics2), true, true);
-      RCP<Matrix> RAP = Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*P,       true, *AP, false, GetOStream(Statistics2), true, true);
-      Set(coarseLevel, "AForPat", RAP);
-    }
-    Set(coarseLevel, "Nullspace",   coarseNullspace);
-
-    // Compute data for velocity
-    if (pressureMode) {
-      Array<LO> velCptlist = Cptlist;
-      FindMidPoints(*A, *coords, velCptlist, *myCpts);
-      coarseLevel.Set<Array<LO> >("PresCptsAndMids", velCptlist, NoFactory::get());
-    }
+    for (int i = 0; i < coarseCoords1D.size(); i++)
+      coarseCoords1D[i] = coords1D[Cptlist[i]];
   }
 
-  template<class T>
-  void PrintVector(const std::vector<T>& v, const std::string& name, int n = -1) {
-    std::cout << "======================" << std::endl;
-    if (!name.empty())
-      std::cout << "=== " << name << " ===" << std::endl;
-    if (n == -1)
-      n = v.size();
-    for (int i = 0; i < n; i++)
-      std::cout << i << ": " << v[i] << std::endl;
-    std::cout << "======================" << std::endl;
+  // Level Set
+  Set(coarseLevel, "P", P);
+  Set(fineLevel, "CoarseMap", coarseMap);
+  if (pressureMode) {
+    Set(coarseLevel, "CoordinatesPressure", coarseCoords);
+
+  } else {
+    Set(coarseLevel, "CoordinatesVelocity", coarseCoords);
+    // FIXME: why does coarse pattern matrix look like?
+    RCP<Matrix> AP  = Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*AForPat, false, *P, false, GetOStream(Statistics2), true, true);
+    RCP<Matrix> RAP = Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*P, true, *AP, false, GetOStream(Statistics2), true, true);
+    Set(coarseLevel, "AForPat", RAP);
   }
+  Set(coarseLevel, "Nullspace", coarseNullspace);
 
-  // distance2 returns _squared_ distance
-  // i.e. the final sqrt calculation is not done
-  template<class SC>
-  SC distance2(const ArrayRCP<ArrayRCP<const SC> >& coords1D, int i, int j) {
-    const int NDim = coords1D.size();
-
-    SC d = Teuchos::ScalarTraits<SC>::zero();
-    for (int k = 0; k < NDim; k++) {
-      SC dtmp = coords1D[k][j] - coords1D[k][i];
-      d += dtmp*dtmp;
-    }
-
-    return d;
+  // Compute data for velocity
+  if (pressureMode) {
+    Array<LO> velCptlist = Cptlist;
+    FindMidPoints(*A, *coords, velCptlist, *myCpts);
+    coarseLevel.Set<Array<LO> >("PresCptsAndMids", velCptlist, NoFactory::get());
+  }
+}
+
+template <class T>
+void PrintVector(const std::vector<T>& v, const std::string& name, int n = -1) {
+  std::cout << "======================" << std::endl;
+  if (!name.empty())
+    std::cout << "=== " << name << " ===" << std::endl;
+  if (n == -1)
+    n = v.size();
+  for (int i = 0; i < n; i++)
+    std::cout << i << ": " << v[i] << std::endl;
+  std::cout << "======================" << std::endl;
+}
+
+// distance2 returns _squared_ distance
+// i.e. the final sqrt calculation is not done
+template <class SC>
+SC distance2(const ArrayRCP<ArrayRCP<const SC> >& coords1D, int i, int j) {
+  const int NDim = coords1D.size();
+
+  SC d = Teuchos::ScalarTraits<SC>::zero();
+  for (int k = 0; k < NDim; k++) {
+    SC dtmp = coords1D[k][j] - coords1D[k][i];
+    d += dtmp * dtmp;
   }
 
-  template<int N = 10000>
-  std::string i2s(int i) {
-    std::ostringstream os;
-    if (i >= N) {
-      os << i;
-    } else {
-      if (i < 10)           os << "0";
-      if (i < 100)          os << "0";
-      if (i < 1000)         os << "0";
-      if (i < 10000)        os << "0";
-      if (i < 100000)       os << "0";
-      if (i < 1000000)      os << "0";
-      if (i < 10000000)     os << "0";
-      if (i < 100000000)    os << "0";
-      if (i < 1000000000)   os << "0";
-    }
+  return d;
+}
+
+template <int N = 10000>
+std::string i2s(int i) {
+  std::ostringstream os;
+  if (i >= N) {
     os << i;
-    return os.str();
+  } else {
+    if (i < 10) os << "0";
+    if (i < 100) os << "0";
+    if (i < 1000) os << "0";
+    if (i < 10000) os << "0";
+    if (i < 100000) os << "0";
+    if (i < 1000000) os << "0";
+    if (i < 10000000) os << "0";
+    if (i < 100000000) os << "0";
+    if (i < 1000000000) os << "0";
+  }
+  os << i;
+  return os.str();
+}
+
+// Initial fill Cptlist with a set of distance 4 points (during phase one).
+// Additional Cpts are then determined looking for large gaps between the
+// phase one Cpts. Candidate additional Cpts corresponds to phase one FPOINTs
+// that have only 1 or 2 Cpts within a graph distance of 3 and are generally
+// far (via graph or coordinate distances) from existing Cpts. We also define
+// a sparsity pattern. An initial pattern is computed which basically
+// includes all FPOINTs within a distance 3 from a Cpt. Additional entries
+// are added to the initial sparsity pattern via PhaseTwoPattern(). These
+// points correspond to Fpoints that only interpolate from 2 or less Cpts,
+// are also far from existing Cpoints, and where the orientation of the
+// interpolation Cpts is skewed to one side of the Fpoint (see
+// PhaseTwoPattern for more details).
+//
+// NOTE: inefficiencies
+// The main inefficiency is the sorting of the CandidateList. This is
+// expensive and it is not clear how important it really is. My guess is that
+// we could actually leave the CandidateList unsorted and things would work
+// almost as well. We could also do some kind of hybrid where we sort the
+// first 100 candidates. Use this data to create something like 10 bins and
+// for any further candidates just put them in the right bin. When we need to
+// choose a new Cpt from the CandidateList, just pick any one from the lowest
+// populated bin. There are also potential inefficiences with respect to
+// malloc(). I doubt that these are a big problem, but one could allocate
+// some workspaces ahead of time to avoid the constant malloc/free cycle in
+// CompDistances().
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindDist4Cpts(const Matrix& A, const MultiVector& coords, const Array<LO>& userCpts, std::vector<char>& status, MyCptList& myCpts, int levelID) const {
+  int NDim       = coords.getNumVectors();
+  size_t numRows = A.getLocalNumRows();
+
+  ArrayRCP<const size_t> ia;
+  ArrayRCP<const LO> ja;
+  CreateCrsPointers(A, ia, ja);
+
+  ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
+  for (int k = 0; k < NDim; k++)
+    coords1D[k] = coords.getData(k);
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero();
+
+  // Initialize coordDist to some large value.
+  // coordDist is an attempt to measure an average distance from a given
+  // point to all the CPOINTs that it depends on. The averages are harmonic, so
+  // basically the initial large coordDist will be averaged away with the 1st
+  // harmonic average. The initial big value computed here is
+  //     (Max(x)-Min(x))^2 + (Max(y)-Min(y))^2 + (Max(z)-Min(z))^2
+  SC big = zero;
+  for (int i = 0; i < NDim; i++) {
+    SC dmin = *std::min_element(coords1D[i].begin(), coords1D[i].end());
+    SC dmax = *std::max_element(coords1D[i].begin(), coords1D[i].end());
+
+    big += ((dmax - dmin) * (dmax - dmin));
+  }
+  // MATCH_MATLAB
+  std::vector<SC> coordDist(numRows, 10000 * big);
+
+  const ParameterList& pL   = GetParameterList();
+  const bool doStatusOutput = pL.get<bool>("dump status");
+  const bool pressureMode   = (pL.get<std::string>("mode") == "pressure");
+
+  // Set all Dirichlet points as Fpoints FIXME: why FPOINTs?
+  // However, if a Dirichlet point is in userCpts, it will be added to the
+  // Cpt list later
+  for (size_t i = 0; i < numRows; i++)
+    if (ia[i + 1] - ia[i] == 1)
+      status[i] = FPOINT;
+
+  // userCpts have already been fixed to be CPOINTs so we want to first mark
+  // them appropriately, and put them first in the CPOINT list, but still go
+  // through loops below to update distances and FPOINTs. Initialization is
+  // done here so that these points do not end up with more than 1 nnz in the
+  // associated sparsity pattern row.
+  TEUCHOS_TEST_FOR_EXCEPTION(myCpts.getCList().size(), Exceptions::RuntimeError,
+                             "myCpts in FindDist4Points must be uninitialized");
+  Array<LO>& Cptlist = myCpts.getCList();
+  for (int i = 0; i < userCpts.size(); i++) {
+    status[userCpts[i]] = CPOINT_U;
+    Cptlist.push_back(userCpts[i]);
   }
 
+  std::string st = std::string("status-l") + toString(levelID) + (pressureMode ? "-p-" : "-v-");
+  int dumpCount  = 0;
+  if (doStatusOutput) {
+    DumpCoords("coord-l" + toString(levelID) + (pressureMode ? "-p" : "-v"), coords);
+    DumpStatus(st + i2s(dumpCount++) + "-A", status, NDim);
+  }
 
-  // Initial fill Cptlist with a set of distance 4 points (during phase one).
-  // Additional Cpts are then determined looking for large gaps between the
-  // phase one Cpts. Candidate additional Cpts corresponds to phase one FPOINTs
-  // that have only 1 or 2 Cpts within a graph distance of 3 and are generally
-  // far (via graph or coordinate distances) from existing Cpts. We also define
-  // a sparsity pattern. An initial pattern is computed which basically
-  // includes all FPOINTs within a distance 3 from a Cpt. Additional entries
-  // are added to the initial sparsity pattern via PhaseTwoPattern(). These
-  // points correspond to Fpoints that only interpolate from 2 or less Cpts,
-  // are also far from existing Cpoints, and where the orientation of the
-  // interpolation Cpts is skewed to one side of the Fpoint (see
-  // PhaseTwoPattern for more details).
-  //
-  // NOTE: inefficiencies
-  // The main inefficiency is the sorting of the CandidateList. This is
-  // expensive and it is not clear how important it really is. My guess is that
-  // we could actually leave the CandidateList unsorted and things would work
-  // almost as well. We could also do some kind of hybrid where we sort the
-  // first 100 candidates. Use this data to create something like 10 bins and
-  // for any further candidates just put them in the right bin. When we need to
-  // choose a new Cpt from the CandidateList, just pick any one from the lowest
-  // populated bin. There are also potential inefficiences with respect to
-  // malloc(). I doubt that these are a big problem, but one could allocate
-  // some workspaces ahead of time to avoid the constant malloc/free cycle in
-  // CompDistances().
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindDist4Cpts(const Matrix& A, const MultiVector& coords, const Array<LO>& userCpts, std::vector<char>& status, MyCptList& myCpts, int levelID) const {
-    int    NDim    = coords.getNumVectors();
-    size_t numRows = A.getLocalNumRows();
-
-    ArrayRCP<const size_t> ia;
-    ArrayRCP<const LO>     ja;
-    CreateCrsPointers(A, ia, ja);
-
-    ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
-    for (int k = 0; k < NDim; k++)
-      coords1D[k] = coords.getData(k);
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero();
-
-    // Initialize coordDist to some large value.
-    // coordDist is an attempt to measure an average distance from a given
-    // point to all the CPOINTs that it depends on. The averages are harmonic, so
-    // basically the initial large coordDist will be averaged away with the 1st
-    // harmonic average. The initial big value computed here is
-    //     (Max(x)-Min(x))^2 + (Max(y)-Min(y))^2 + (Max(z)-Min(z))^2
-    SC big = zero;
-    for (int i = 0; i < NDim; i++) {
-      SC dmin = *std::min_element(coords1D[i].begin(), coords1D[i].end());
-      SC dmax = *std::max_element(coords1D[i].begin(), coords1D[i].end());
-
-      big += ((dmax - dmin)*(dmax - dmin));
+  std::vector<short>& numCpts = myCpts.getNumCpts();
+
+  // Determine CPOINTs
+  int userCcount       = 0;
+  size_t numCandidates = 0;
+  std::vector<char> distIncrement(numRows, 0);
+  std::vector<double> cumGraphDist(numRows, 0.);
+  std::vector<LO> candidateList(numRows, 0);
+  size_t i = 0;
+  while (i < numRows) {
+    LO newCpt = -1;
+
+    // Check userCpts list
+    //
+    // These were essentially already determined to be CPOINTs by some other
+    // functions. We want to then add them one-by-one, updating distances,
+    // and look to see if further CPOINTs can be added once we have finished
+    // all of the userCpts
+    if (userCcount < userCpts.size())
+      newCpt = userCpts[userCcount++];
+
+    // Check for possible CPOINT on candidate list
+    // FIXME: Could CANDIDATE list contain non-CANDIDATE statuses?
+    while ((newCpt == -1) && (numCandidates > 0)) {
+      if (status[candidateList[numCandidates - 1]] <= CANDIDATE) {
+        newCpt         = candidateList[numCandidates - 1];
+        status[newCpt] = CPOINT;
+
+        if (doStatusOutput)
+          DumpStatus(st + i2s(dumpCount++) + "-B", status, NDim);
+      }
+      numCandidates--;
+      // FIXME: Why is there no i++ here?
     }
-    // MATCH_MATLAB
-    std::vector<SC> coordDist(numRows, 10000*big);
 
-    const ParameterList& pL = GetParameterList();
-    const bool doStatusOutput = pL.get<bool>("dump status");
-    const bool pressureMode   = (pL.get<std::string>("mode") == "pressure");
+    // If no new CPOINT identified in candidate list, check the unassigned list
+    while ((newCpt == -1) && (i < numRows)) {
+      if (status[i] == UNASSIGNED) {
+        newCpt         = i;
+        status[newCpt] = CPOINT;
 
-    // Set all Dirichlet points as Fpoints FIXME: why FPOINTs?
-    // However, if a Dirichlet point is in userCpts, it will be added to the
-    // Cpt list later
-    for (size_t i = 0; i < numRows; i++)
-      if (ia[i+1] - ia[i] == 1)
-        status[i] = FPOINT;
-
-    // userCpts have already been fixed to be CPOINTs so we want to first mark
-    // them appropriately, and put them first in the CPOINT list, but still go
-    // through loops below to update distances and FPOINTs. Initialization is
-    // done here so that these points do not end up with more than 1 nnz in the
-    // associated sparsity pattern row.
-    TEUCHOS_TEST_FOR_EXCEPTION(myCpts.getCList().size(), Exceptions::RuntimeError,
-                               "myCpts in FindDist4Points must be uninitialized");
-    Array<LO>& Cptlist = myCpts.getCList();
-    for (int i = 0; i < userCpts.size(); i++) {
-      status[userCpts[i]] = CPOINT_U;
-      Cptlist.push_back(userCpts[i]);
-    }
-
-    std::string st = std::string("status-l") + toString(levelID) + (pressureMode ? "-p-" : "-v-");
-    int dumpCount = 0;
-    if (doStatusOutput) {
-      DumpCoords("coord-l" + toString(levelID) + (pressureMode ? "-p" : "-v"), coords);
-      DumpStatus(st + i2s(dumpCount++) + "-A", status, NDim);
+        if (doStatusOutput)
+          DumpStatus(st + i2s(dumpCount++) + "-C", status, NDim);
+      }
+      i++;
     }
 
-    std::vector<short>& numCpts = myCpts.getNumCpts();
-
-    // Determine CPOINTs
-    int    userCcount    = 0;
-    size_t numCandidates = 0;
-    std::vector<char>   distIncrement(numRows, 0);
-    std::vector<double> cumGraphDist (numRows, 0.);
-    std::vector<LO>     candidateList(numRows, 0);
-    size_t i = 0;
-    while (i < numRows) {
-      LO newCpt = -1;
-
-      // Check userCpts list
-      //
-      // These were essentially already determined to be CPOINTs by some other
-      // functions. We want to then add them one-by-one, updating distances,
-      // and look to see if further CPOINTs can be added once we have finished
-      // all of the userCpts
-      if (userCcount < userCpts.size())
-        newCpt = userCpts[userCcount++];
-
-      // Check for possible CPOINT on candidate list
-      // FIXME: Could CANDIDATE list contain non-CANDIDATE statuses?
-      while ((newCpt == -1) && (numCandidates > 0)) {
-        if (status[candidateList[numCandidates-1]] <= CANDIDATE) {
-          newCpt         = candidateList[numCandidates-1];
-          status[newCpt] = CPOINT;
-
-          if (doStatusOutput)
-            DumpStatus(st + i2s(dumpCount++) + "-B", status, NDim);
-        }
-        numCandidates--;
-        // FIXME: Why is there no i++ here?
+    // Update distances and the status of neighbors neighbors to reflect a
+    // newly found CPOINT
+    if (newCpt != -1) {
+      std::vector<LO> dist1, dist2, dist3, dist4;
+      // FIXME: Should CompDistances automatically exclude other CPOINTs?
+      CompDistances(A, newCpt, 4, dist1, dist2, dist3, dist4);
+
+      // Make sure that the only CPOINT in dist3 is newCpt. All others should be excluded.
+      int numDist3 = 0;
+      for (size_t k = 0; k < dist3.size(); k++) {
+        LO j = dist3[k];
+        if (status[j] < CPOINT)
+          dist3[numDist3++] = j;
       }
-
-      // If no new CPOINT identified in candidate list, check the unassigned list
-      while ((newCpt == -1) && (i < numRows)) {
-        if (status[i] == UNASSIGNED) {
-          newCpt         = i;
-          status[newCpt] = CPOINT;
-
-          if (doStatusOutput)
-            DumpStatus(st + i2s(dumpCount++) + "-C", status, NDim);
+      dist3.resize(numDist3);
+      dist3.push_back(newCpt);
+
+      // UNASSIGNED or CANDIDATE distance 3 and closer neighbors are put into FPOINT list
+      // FIXME: why not put TWOTIMER there too?
+      bool dumpStatus = false;
+      for (size_t k = 0; k < dist3.size(); k++) {
+        LO j = dist3[k];
+        if (status[j] == UNASSIGNED || status[j] == CANDIDATE) {
+          status[j]  = FPOINT;
+          dumpStatus = true;
         }
-        i++;
       }
+      if (dumpStatus && doStatusOutput)
+        DumpStatus(st + i2s(dumpCount++) + "-D", status, NDim);
 
-      // Update distances and the status of neighbors neighbors to reflect a
-      // newly found CPOINT
-      if (newCpt != -1) {
-        std::vector<LO> dist1, dist2, dist3, dist4;
-        // FIXME: Should CompDistances automatically exclude other CPOINTs?
-        CompDistances(A, newCpt, 4, dist1, dist2, dist3, dist4);
-
-        // Make sure that the only CPOINT in dist3 is newCpt. All others should be excluded.
-        int numDist3 = 0;
-        for (size_t k = 0; k < dist3.size(); k++) {
-          LO j = dist3[k];
-          if (status[j] < CPOINT)
-            dist3[numDist3++] = j;
-        }
-        dist3.resize(numDist3);
-        dist3.push_back(newCpt);
-
-        // UNASSIGNED or CANDIDATE distance 3 and closer neighbors are put into FPOINT list
-        // FIXME: why not put TWOTIMER there too?
-        bool dumpStatus = false;
-        for (size_t k = 0; k < dist3.size(); k++) {
-          LO j = dist3[k];
-          if (status[j] == UNASSIGNED || status[j] == CANDIDATE) {
-            status[j] = FPOINT;
-            dumpStatus = true;
-          }
-        }
-        if (dumpStatus && doStatusOutput)
-          DumpStatus(st + i2s(dumpCount++) + "-D", status, NDim);
+      // Update myCpts() to reflect dependence of neighbors on newCpt
+      for (size_t k = 0; k < dist3.size(); k++) {
+        LO j = dist3[k];
 
-        // Update myCpts() to reflect dependence of neighbors on newCpt
-        for (size_t k = 0; k < dist3.size(); k++) {
-          LO j = dist3[k];
+        TEUCHOS_TEST_FOR_EXCEPTION(numCpts[j] >= myCpts.getNnzPerRow(), Exceptions::RuntimeError,
+                                   "Increase max number of C points per row");
+        myCpts(j)[numCpts[j]++] = newCpt;
+      }
 
-          TEUCHOS_TEST_FOR_EXCEPTION(numCpts[j] >= myCpts.getNnzPerRow(), Exceptions::RuntimeError,
-                                     "Increase max number of C points per row");
-          myCpts(j)[numCpts[j]++] = newCpt;
-        }
+      // Update cumGraphDist
+      // NOTE: order matters as dist2 is contained within dist3, etc.
+      // FIXME: Do dist2 and dist1 contain CPOINTs?
+      for (size_t k = 0; k < dist3.size(); k++) distIncrement[dist3[k]] = 3;
+      for (size_t k = 0; k < dist2.size(); k++) distIncrement[dist2[k]] = 2;
+      for (size_t k = 0; k < dist1.size(); k++) distIncrement[dist1[k]] = 1;
+      distIncrement[newCpt] = 0;
+
+      for (size_t k = 0; k < dist3.size(); k++) {
+        LO j = dist3[k];
+        // MATCH_MATLAB: (numCpts[j]-1) is to match Matlab, where numCpts is updated after distance calculation
+        cumGraphDist[j] = (cumGraphDist[j] * (numCpts[j] - 1) + distIncrement[j]) / numCpts[j];
+      }
+      cumGraphDist[newCpt] = 0;
 
-        // Update cumGraphDist
-        // NOTE: order matters as dist2 is contained within dist3, etc.
-        // FIXME: Do dist2 and dist1 contain CPOINTs?
-        for (size_t k = 0; k < dist3.size(); k++) distIncrement[dist3[k]] = 3;
-        for (size_t k = 0; k < dist2.size(); k++) distIncrement[dist2[k]] = 2;
-        for (size_t k = 0; k < dist1.size(); k++) distIncrement[dist1[k]] = 1;
-        distIncrement[newCpt] = 0;
-
-        for (size_t k = 0; k < dist3.size(); k++) {
-          LO j = dist3[k];
-          // MATCH_MATLAB: (numCpts[j]-1) is to match Matlab, where numCpts is updated after distance calculation
-          cumGraphDist[j] = (cumGraphDist[j]*(numCpts[j]-1) + distIncrement[j])/numCpts[j];
-        }
-        cumGraphDist[newCpt] = 0;
-
-        // Compute coordinate distance to CPOINT
-        //
-        // Distance of CANDIDATEs to CPOINTs will be used to determine the next
-        // chosen CPOINT from the candidate list. Distances are also used to
-        // decide where new CPOINTs should be added.
-        for (size_t k = 0; k < dist4.size(); k++) {
-          LO j = dist4[k];
-
-          SC distance = distance2(coords1D, newCpt, j);
-
-          // Harmonic average new distance with old distance
-          // There should really be a '2' in front of this expression. This
-          // is actually a bug in the code. However, if I put a '2', I don't
-          // get the perfect coarsening for a uniform mesh ... so I'm leaving
-          // if for now without the 2.
-          // MATCH_MATLAB
-          coordDist[j] = 2.0*(coordDist[j]*distance) / (coordDist[j] + distance);
+      // Compute coordinate distance to CPOINT
+      //
+      // Distance of CANDIDATEs to CPOINTs will be used to determine the next
+      // chosen CPOINT from the candidate list. Distances are also used to
+      // decide where new CPOINTs should be added.
+      for (size_t k = 0; k < dist4.size(); k++) {
+        LO j = dist4[k];
+
+        SC distance = distance2(coords1D, newCpt, j);
+
+        // Harmonic average new distance with old distance
+        // There should really be a '2' in front of this expression. This
+        // is actually a bug in the code. However, if I put a '2', I don't
+        // get the perfect coarsening for a uniform mesh ... so I'm leaving
+        // if for now without the 2.
+        // MATCH_MATLAB
+        coordDist[j] = 2.0 * (coordDist[j] * distance) / (coordDist[j] + distance);
 #if 0
           SC kkk = 10.;
           if (coordDist[j] > distance)
             coordDist[j] = (kkk*coordDist[j]*distance)/(coordDist[j]*(kkk-1)+ distance);
           coordDist[j] = (kkk*coordDist[j]*distance)/(coordDist[j]        + distance*(kkk-1));
 #endif
-        }
+      }
 
-        // Mark all unassigned dist4 points as CANDIDATE and compress
-        // dist4 so that it only contains entries for the candidate list.
-        size_t numNewCandidates = 0;
-        dumpStatus = false;
-        for (size_t k = 0; k < dist4.size(); k++) {
-          LO j = dist4[k];
-
-          // NOTE: numNewCandidates is always <= k, so we don't overwrite the
-          // dist4 before reading
-          if (status[j] == CANDIDATE) {
-            // Mark as already being assigned again to candidate list so that
-            // entry in old 'sorted' candidate list can be removed and a new
-            // 'unsorted' entry can be created. This new entry will later be
-            // sorted reflecting the new coordinate distance.
-            status[j] = TWOTIMER;
-            dist4[numNewCandidates++] = j;
-            dumpStatus = true;
-
-          } else if (status[j] == UNASSIGNED) {
-            status[j] = CANDIDATE;
-            dist4[numNewCandidates++] = j;
-            dumpStatus = true;
-          }
+      // Mark all unassigned dist4 points as CANDIDATE and compress
+      // dist4 so that it only contains entries for the candidate list.
+      size_t numNewCandidates = 0;
+      dumpStatus              = false;
+      for (size_t k = 0; k < dist4.size(); k++) {
+        LO j = dist4[k];
+
+        // NOTE: numNewCandidates is always <= k, so we don't overwrite the
+        // dist4 before reading
+        if (status[j] == CANDIDATE) {
+          // Mark as already being assigned again to candidate list so that
+          // entry in old 'sorted' candidate list can be removed and a new
+          // 'unsorted' entry can be created. This new entry will later be
+          // sorted reflecting the new coordinate distance.
+          status[j]                 = TWOTIMER;
+          dist4[numNewCandidates++] = j;
+          dumpStatus                = true;
+
+        } else if (status[j] == UNASSIGNED) {
+          status[j]                 = CANDIDATE;
+          dist4[numNewCandidates++] = j;
+          dumpStatus                = true;
         }
-        dist4.resize(numNewCandidates);
-        if (dumpStatus && doStatusOutput)
-          DumpStatus(st + i2s(dumpCount++) + "-E", status, pressureMode);
-
-        // Now remove all TWOTIMERs from the old candidate list
-        size_t numOldCandidates = 0;
-        dumpStatus = false;
-        for (size_t k = 0; k < numCandidates; k++) {
-          LO j = candidateList[k];
-
-          if (status[j] == CANDIDATE) { candidateList[numOldCandidates++] = j; }
-          if (status[j] == TWOTIMER ) {
-            status[j] = CANDIDATE;
-            dumpStatus = true;
-          }
+      }
+      dist4.resize(numNewCandidates);
+      if (dumpStatus && doStatusOutput)
+        DumpStatus(st + i2s(dumpCount++) + "-E", status, pressureMode);
+
+      // Now remove all TWOTIMERs from the old candidate list
+      size_t numOldCandidates = 0;
+      dumpStatus              = false;
+      for (size_t k = 0; k < numCandidates; k++) {
+        LO j = candidateList[k];
+
+        if (status[j] == CANDIDATE) {
+          candidateList[numOldCandidates++] = j;
         }
-        if (dumpStatus && doStatusOutput)
-          DumpStatus(st + i2s(dumpCount++) + "-F", status, NDim);
-
-        // Sort the candidates based on distances (breaking ties via degrees,
-        // encouraging points near boundary). First, we order new candidates
-        // and then we merge together two sorted lists.
-        //
-        // NOTE: to match matlab (and break ties), I added the  1.e-10 term
-        std::vector<double> ddtemp(numNewCandidates);
-        for (size_t k = 0; k < numNewCandidates; k++) {
-          LO j = dist4[k];
-          // MATCH_MATLAB
+        if (status[j] == TWOTIMER) {
+          status[j]  = CANDIDATE;
+          dumpStatus = true;
+        }
+      }
+      if (dumpStatus && doStatusOutput)
+        DumpStatus(st + i2s(dumpCount++) + "-F", status, NDim);
+
+      // Sort the candidates based on distances (breaking ties via degrees,
+      // encouraging points near boundary). First, we order new candidates
+      // and then we merge together two sorted lists.
+      //
+      // NOTE: to match matlab (and break ties), I added the  1.e-10 term
+      std::vector<double> ddtemp(numNewCandidates);
+      for (size_t k = 0; k < numNewCandidates; k++) {
+        LO j = dist4[k];
+        // MATCH_MATLAB
 #ifdef optimal
-          // This one is better, but we are trying to replicate Matlab now
-          ddtemp[k] = -coordDist[j] - .01*(ia[j+1]-ia[j]) + 1e-10*(j+1);
+        // This one is better, but we are trying to replicate Matlab now
+        ddtemp[k] = -coordDist[j] - .01 * (ia[j + 1] - ia[j]) + 1e-10 * (j + 1);
 #else
-          ddtemp[k] = +coordDist[j] - 0.0*(ia[j+1]-ia[j]) + 1e-3*(j+1);
+        ddtemp[k] = +coordDist[j] - 0.0 * (ia[j + 1] - ia[j]) + 1e-3 * (j + 1);
 #endif
-        }
-        Muelu_az_dsort2(ddtemp, dist4);
-
-        MergeSort(candidateList, numOldCandidates, dist4, coordDist, ia);
-        numCandidates = numOldCandidates + numNewCandidates;
       }
+      Muelu_az_dsort2(ddtemp, dist4);
+
+      MergeSort(candidateList, numOldCandidates, dist4, coordDist, ia);
+      numCandidates = numOldCandidates + numNewCandidates;
     }
+  }
 
-    // Add additional CPOINTs based on some score which includes the number of
-    // CPOINTs that an FPOINT depends on as well as its distance (both graph
-    // and coordinate) to nearby CPOINTs.
-    const double graphWeight  = 0.8;
-    const double orientWeight = 0.5;
-    for (int numCDepends = 1; numCDepends <= 2; numCDepends++) {
-      numCandidates = 0;
-
-      std::vector<int> candidates;
-      for (i = 0; i < numRows; i++)
-        if (status[i] < CPOINT && numCpts[i] == numCDepends) {
-          candidates.push_back(i);
-          numCandidates++;
-        }
+  // Add additional CPOINTs based on some score which includes the number of
+  // CPOINTs that an FPOINT depends on as well as its distance (both graph
+  // and coordinate) to nearby CPOINTs.
+  const double graphWeight  = 0.8;
+  const double orientWeight = 0.5;
+  for (int numCDepends = 1; numCDepends <= 2; numCDepends++) {
+    numCandidates = 0;
 
-      if (numCandidates != 0) {
-        // Sort FPOINTs based on distance to CPOINTs and orientation
-        double maxGraphDist = -1e20;
-        double maxCoordDist = -1e20;
-        for (size_t p = 0; p < numCandidates; p++) {
-          LO j = candidates[p];
+    std::vector<int> candidates;
+    for (i = 0; i < numRows; i++)
+      if (status[i] < CPOINT && numCpts[i] == numCDepends) {
+        candidates.push_back(i);
+        numCandidates++;
+      }
 
-          maxGraphDist = std::max(maxGraphDist, cumGraphDist[j]);
-          maxCoordDist = std::max(maxCoordDist, coordDist[j]);
-        }
+    if (numCandidates != 0) {
+      // Sort FPOINTs based on distance to CPOINTs and orientation
+      double maxGraphDist = -1e20;
+      double maxCoordDist = -1e20;
+      for (size_t p = 0; p < numCandidates; p++) {
+        LO j = candidates[p];
 
-        std::vector<double> score      (numCandidates);
-        std::vector<double> orientation(numCandidates);
-        for (size_t p = 0; p < numCandidates; p++) {
-          LO j = candidates[p];
+        maxGraphDist = std::max(maxGraphDist, cumGraphDist[j]);
+        maxCoordDist = std::max(maxCoordDist, coordDist[j]);
+      }
 
-          double graphScore = cumGraphDist[j] / maxGraphDist;
-          double coordScore = coordDist   [j] / maxCoordDist;
-          // MATCH_MATLAB
-          score[p] = -(graphWeight*graphScore + (1-graphWeight)*coordScore + 1e-6*(j+1));
-
-          if (numCDepends == 2) {
-            // Orientation of -1 means that we have two CPOINTs on opposite
-            // sides of the FPOINT. Orientation of 0 implies that things are
-            // orthogonal. orientation of 1 implies that Cpoints are on the
-            // same side
-            SC norm = zero, vec1[3], vec2[3];
-            for (int k = 0; k < NDim; k++) {
-              vec1[k] =  coords1D[k][j] - coords1D[k][myCpts(j)[0]];
-              norm   += vec1[k]*vec1[k];
-            }
-            norm = sqrt(norm);
-            for (int k = 0; k < NDim; k++)
-              vec1[k] /= norm;
-
-            norm = zero;
-            for (int k = 0; k < NDim; k++) {
-              vec2[k] =  coords1D[k][j] - coords1D[k][myCpts(j)[1]];
-              norm   += vec2[k]*vec2[k];
-            }
-            norm = sqrt(norm);
-            for (int k = 0; k < NDim; k++)
-              vec2[k] /= norm;
-
-            orientation[p] = zero;
-            for (int k = 0; k < NDim; k++)
-              orientation[p] += vec1[k]*vec2[k];
-
-            score[p] = -(orientWeight*orientation[p] - (1-orientWeight)*score[p]);
+      std::vector<double> score(numCandidates);
+      std::vector<double> orientation(numCandidates);
+      for (size_t p = 0; p < numCandidates; p++) {
+        LO j = candidates[p];
 
-          } else {
-            orientation[p] = 1.0;
+        double graphScore = cumGraphDist[j] / maxGraphDist;
+        double coordScore = coordDist[j] / maxCoordDist;
+        // MATCH_MATLAB
+        score[p] = -(graphWeight * graphScore + (1 - graphWeight) * coordScore + 1e-6 * (j + 1));
+
+        if (numCDepends == 2) {
+          // Orientation of -1 means that we have two CPOINTs on opposite
+          // sides of the FPOINT. Orientation of 0 implies that things are
+          // orthogonal. orientation of 1 implies that Cpoints are on the
+          // same side
+          SC norm = zero, vec1[3], vec2[3];
+          for (int k = 0; k < NDim; k++) {
+            vec1[k] = coords1D[k][j] - coords1D[k][myCpts(j)[0]];
+            norm += vec1[k] * vec1[k];
           }
-        }
+          norm = sqrt(norm);
+          for (int k = 0; k < NDim; k++)
+            vec1[k] /= norm;
 
-        std::vector<LO> index(numCandidates);
-        for (size_t p = 0; p < numCandidates; p++)
-          index[p] = p;
-        Muelu_az_dsort2(score, index);
-
-        for (size_t p = 0; p < numCandidates; p++) {
-          int newCpt = candidates[index[p]];
-
-          if (numCpts[newCpt] == numCDepends &&
-              cumGraphDist[newCpt]  >= 2.6 &&
-              orientation[index[p]] > -0.2) {
-            status [newCpt] = CPOINT;
-            numCpts[newCpt] = 1;
-            myCpts(newCpt)[0] = newCpt;
-
-            if (doStatusOutput)
-              DumpStatus(st + i2s(dumpCount++) + "-G", status, NDim);
-
-            std::vector<LO> dist1, dist2, dist3, dist4;
-            CompDistances(A, newCpt, 3, dist1, dist2, dist3, dist4);
-
-            // Make sure that there are no CPOINTs in dist1, dist2, dist3.
-            int numDist1 = 0;
-            for (size_t k = 0; k < dist1.size(); k++) {
-              LO j = dist1[k];
-              if (status[j] < CPOINT)
-                dist1[numDist1++] = j;
-            }
-            dist1.resize(numDist1);
-            int numDist2 = 0;
-            for (size_t k = 0; k < dist2.size(); k++) {
-              LO j = dist2[k];
-              if (status[j] < CPOINT)
-                dist2[numDist2++] = j;
-            }
-            dist2.resize(numDist2);
-            int numDist3 = 0;
-            for (size_t k = 0; k < dist3.size(); k++) {
-              LO j = dist3[k];
-              if (status[j] < CPOINT)
-                dist3[numDist3++] = j;
-            }
-            dist3.resize(numDist3);
-
-            // Update cumGraphDist
-            // NOTE: order matters as dist2 is contained within dist3, etc.
-            for (size_t k = 0; k < dist3.size(); k++) distIncrement[dist3[k]] = 3;
-            for (size_t k = 0; k < dist2.size(); k++) distIncrement[dist2[k]] = 2;
-            for (size_t k = 0; k < dist1.size(); k++) distIncrement[dist1[k]] = 1;
-            distIncrement[newCpt] = 0;
-
-            // Update myCpts() to reflect dependence of neighbors on newCpt
-            for (size_t k = 0; k < dist3.size(); k++) {
-              LO j = dist3[k];
-
-              TEUCHOS_TEST_FOR_EXCEPTION(numCpts[j] >= myCpts.getNnzPerRow(), Exceptions::RuntimeError,
-                                         "Increase max number of C points per row");
-              myCpts(j)[numCpts[j]++] = newCpt;
-            }
-
-            for (size_t k = 0; k < dist3.size(); k++) {
-              LO j = dist3[k];
-              // (numCpts[j]-1) is to match Matlab, where numCpts is updated after distance calculation
-              cumGraphDist[j] = (cumGraphDist[j]*(numCpts[j]-1) + distIncrement[j])/numCpts[j];
-            }
-            cumGraphDist[newCpt] = 0;
+          norm = zero;
+          for (int k = 0; k < NDim; k++) {
+            vec2[k] = coords1D[k][j] - coords1D[k][myCpts(j)[1]];
+            norm += vec2[k] * vec2[k];
           }
+          norm = sqrt(norm);
+          for (int k = 0; k < NDim; k++)
+            vec2[k] /= norm;
+
+          orientation[p] = zero;
+          for (int k = 0; k < NDim; k++)
+            orientation[p] += vec1[k] * vec2[k];
+
+          score[p] = -(orientWeight * orientation[p] - (1 - orientWeight) * score[p]);
+
+        } else {
+          orientation[p] = 1.0;
         }
       }
-    }
 
-    // Build up the CPOINT list
-    for (i = 0; i < numRows; i++)
-      if (status[i] == CPOINT)
-        Cptlist.push_back(i);
-      else if (status[i] == CPOINT_U)
-        status[i] = CPOINT;
-  }
+      std::vector<LO> index(numCandidates);
+      for (size_t p = 0; p < numCandidates; p++)
+        index[p] = p;
+      Muelu_az_dsort2(score, index);
 
+      for (size_t p = 0; p < numCandidates; p++) {
+        int newCpt = candidates[index[p]];
 
-  // Look at pattern rows which have only one or two nonzeros and consider
-  // adding additional nonzero entries. New nonzero possibilities for row k are
-  // obtained by looking at k's neighbors (as determined by the matrix) to see
-  // what CPOINTs these neighbors interpolate from.  The final determination is
-  // based on a composite score that considers the distance between k and the
-  // possible new CPOINT as well as the orientation of the new possible CPOINT
-  // with respect to k's current CPOINTs. Generally, points which are on the
-  // opposite side of k' current CPOINTs are favored.
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  PhaseTwoPattern(const Matrix& A, const MultiVector& coords, const std::vector<char>& status, MyCptList& myCpts) const {
-    GetOStream(Runtime0) << "Starting phase 2" << std::endl;
+        if (numCpts[newCpt] == numCDepends &&
+            cumGraphDist[newCpt] >= 2.6 &&
+            orientation[index[p]] > -0.2) {
+          status[newCpt]    = CPOINT;
+          numCpts[newCpt]   = 1;
+          myCpts(newCpt)[0] = newCpt;
 
-    int    NDim    = coords.getNumVectors();
-    size_t numRows = A.getLocalNumRows();
+          if (doStatusOutput)
+            DumpStatus(st + i2s(dumpCount++) + "-G", status, NDim);
 
-    ArrayRCP<const size_t> ia;
-    ArrayRCP<const LO>     ja;
-    CreateCrsPointers(A, ia, ja);
+          std::vector<LO> dist1, dist2, dist3, dist4;
+          CompDistances(A, newCpt, 3, dist1, dist2, dist3, dist4);
 
-    ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
-    for (int k = 0; k < NDim; k++)
-      coords1D[k] = coords.getData(k);
-
-    std::vector<short>& numCpts = myCpts.getNumCpts();
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero();
-
-    size_t N = myCpts.getCList().size();
-    std::vector<int>    nearbyCs(N);
-    std::vector<double> score   (N);
-    std::vector<double> dists   (N);
-
-    std::vector<char> scratch   (numRows, 'n');
-    std::vector<LO>   candidates(numRows);
-
-    for (int numCDepends = 1; numCDepends <= 2; numCDepends++) {
-      int numCandidates = 0;
-      for (size_t i = 0; i < numRows; i++)
-        if (numCpts[i] == numCDepends && status[i] < CPOINT)
-            candidates[numCandidates++] = i;
-
-      for (int p = 0; p < numCandidates; p++) {
-        // Mark already existing CPOINT dependencies
-        LO  i    = candidates[p];
-        LO* cpts = myCpts(i);
-        for (int k = 0; k < numCpts[i]; k++)
-          scratch[cpts[k]] = 'y';
-
-        // Make a list of my neighbors' CPOINT dependencies, excluding all
-        // already existing CPOINT dependencies for candidates[p]
-        const LO* neighs = &ja[ia[i]];
-        int numNeighbors = ia[i+1] - ia[i];
-        int numNearbyCs  = 0;
-        for (int k = 0; k < numNeighbors; k++) {
-          LO        curNeigh = neighs[k];
-          const LO* neighCs  = myCpts(curNeigh);
-
-          for (int j = 0; j < numCpts[curNeigh]; j++) {
-            LO neighNeighC = neighCs[j];
-
-            if (scratch[neighNeighC] != 'y') {
-              scratch[neighNeighC] = 'y';
-              nearbyCs[numNearbyCs++] = neighNeighC;
-            }
+          // Make sure that there are no CPOINTs in dist1, dist2, dist3.
+          int numDist1 = 0;
+          for (size_t k = 0; k < dist1.size(); k++) {
+            LO j = dist1[k];
+            if (status[j] < CPOINT)
+              dist1[numDist1++] = j;
           }
+          dist1.resize(numDist1);
+          int numDist2 = 0;
+          for (size_t k = 0; k < dist2.size(); k++) {
+            LO j = dist2[k];
+            if (status[j] < CPOINT)
+              dist2[numDist2++] = j;
+          }
+          dist2.resize(numDist2);
+          int numDist3 = 0;
+          for (size_t k = 0; k < dist3.size(); k++) {
+            LO j = dist3[k];
+            if (status[j] < CPOINT)
+              dist3[numDist3++] = j;
+          }
+          dist3.resize(numDist3);
+
+          // Update cumGraphDist
+          // NOTE: order matters as dist2 is contained within dist3, etc.
+          for (size_t k = 0; k < dist3.size(); k++) distIncrement[dist3[k]] = 3;
+          for (size_t k = 0; k < dist2.size(); k++) distIncrement[dist2[k]] = 2;
+          for (size_t k = 0; k < dist1.size(); k++) distIncrement[dist1[k]] = 1;
+          distIncrement[newCpt] = 0;
+
+          // Update myCpts() to reflect dependence of neighbors on newCpt
+          for (size_t k = 0; k < dist3.size(); k++) {
+            LO j = dist3[k];
+
+            TEUCHOS_TEST_FOR_EXCEPTION(numCpts[j] >= myCpts.getNnzPerRow(), Exceptions::RuntimeError,
+                                       "Increase max number of C points per row");
+            myCpts(j)[numCpts[j]++] = newCpt;
+          }
+
+          for (size_t k = 0; k < dist3.size(); k++) {
+            LO j = dist3[k];
+            // (numCpts[j]-1) is to match Matlab, where numCpts is updated after distance calculation
+            cumGraphDist[j] = (cumGraphDist[j] * (numCpts[j] - 1) + distIncrement[j]) / numCpts[j];
+          }
+          cumGraphDist[newCpt] = 0;
         }
+      }
+    }
+  }
 
-        // Reset scratch
-        for (int k = 0; k < numCpts[i]; k++)
-          scratch[cpts[k]] = 'n';
-        for (int k = 0; k < numNearbyCs; k++)
-          scratch[nearbyCs[k]] = 'n';
+  // Build up the CPOINT list
+  for (i = 0; i < numRows; i++)
+    if (status[i] == CPOINT)
+      Cptlist.push_back(i);
+    else if (status[i] == CPOINT_U)
+      status[i] = CPOINT;
+}
+
+// Look at pattern rows which have only one or two nonzeros and consider
+// adding additional nonzero entries. New nonzero possibilities for row k are
+// obtained by looking at k's neighbors (as determined by the matrix) to see
+// what CPOINTs these neighbors interpolate from.  The final determination is
+// based on a composite score that considers the distance between k and the
+// possible new CPOINT as well as the orientation of the new possible CPOINT
+// with respect to k's current CPOINTs. Generally, points which are on the
+// opposite side of k' current CPOINTs are favored.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    PhaseTwoPattern(const Matrix& A, const MultiVector& coords, const std::vector<char>& status, MyCptList& myCpts) const {
+  GetOStream(Runtime0) << "Starting phase 2" << std::endl;
+
+  int NDim       = coords.getNumVectors();
+  size_t numRows = A.getLocalNumRows();
+
+  ArrayRCP<const size_t> ia;
+  ArrayRCP<const LO> ja;
+  CreateCrsPointers(A, ia, ja);
+
+  ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
+  for (int k = 0; k < NDim; k++)
+    coords1D[k] = coords.getData(k);
+
+  std::vector<short>& numCpts = myCpts.getNumCpts();
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero();
+
+  size_t N = myCpts.getCList().size();
+  std::vector<int> nearbyCs(N);
+  std::vector<double> score(N);
+  std::vector<double> dists(N);
+
+  std::vector<char> scratch(numRows, 'n');
+  std::vector<LO> candidates(numRows);
+
+  for (int numCDepends = 1; numCDepends <= 2; numCDepends++) {
+    int numCandidates = 0;
+    for (size_t i = 0; i < numRows; i++)
+      if (numCpts[i] == numCDepends && status[i] < CPOINT)
+        candidates[numCandidates++] = i;
+
+    for (int p = 0; p < numCandidates; p++) {
+      // Mark already existing CPOINT dependencies
+      LO i     = candidates[p];
+      LO* cpts = myCpts(i);
+      for (int k = 0; k < numCpts[i]; k++)
+        scratch[cpts[k]] = 'y';
+
+      // Make a list of my neighbors' CPOINT dependencies, excluding all
+      // already existing CPOINT dependencies for candidates[p]
+      const LO* neighs = &ja[ia[i]];
+      int numNeighbors = ia[i + 1] - ia[i];
+      int numNearbyCs  = 0;
+      for (int k = 0; k < numNeighbors; k++) {
+        LO curNeigh       = neighs[k];
+        const LO* neighCs = myCpts(curNeigh);
+
+        for (int j = 0; j < numCpts[curNeigh]; j++) {
+          LO neighNeighC = neighCs[j];
+
+          if (scratch[neighNeighC] != 'y') {
+            scratch[neighNeighC]    = 'y';
+            nearbyCs[numNearbyCs++] = neighNeighC;
+          }
+        }
+      }
+
+      // Reset scratch
+      for (int k = 0; k < numCpts[i]; k++)
+        scratch[cpts[k]] = 'n';
+      for (int k = 0; k < numNearbyCs; k++)
+        scratch[nearbyCs[k]] = 'n';
 
-        // MATCH_MATLB
-        std::sort(nearbyCs.begin(), nearbyCs.begin() + numNearbyCs);
+      // MATCH_MATLB
+      std::sort(nearbyCs.begin(), nearbyCs.begin() + numNearbyCs);
 
-        if (numNearbyCs != 0) {
-          SC norm = zero, vec1[3] = {0.0, 0.0, 0.0}, vec2[3] = {0.0, 0.0, 0.0};
+      if (numNearbyCs != 0) {
+        SC norm = zero, vec1[3] = {0.0, 0.0, 0.0}, vec2[3] = {0.0, 0.0, 0.0};
+        for (int k = 0; k < NDim; k++) {
+          vec1[k] = coords1D[k][i] - coords1D[k][cpts[0]];
+          norm += vec1[k] * vec1[k];
+        }
+        norm = sqrt(norm);
+        for (int k = 0; k < NDim; k++)
+          vec1[k] /= norm;
+
+        if (numCDepends == 2) {
+          norm = zero;
           for (int k = 0; k < NDim; k++) {
-            vec1[k] =  coords1D[k][i] - coords1D[k][cpts[0]];
-            norm   += vec1[k]*vec1[k];
+            vec2[k] = coords1D[k][i] - coords1D[k][cpts[1]];
+            norm += vec2[k] * vec2[k];
           }
           norm = sqrt(norm);
           for (int k = 0; k < NDim; k++)
-            vec1[k] /= norm;
+            vec2[k] /= norm;
 
-          if (numCDepends == 2) {
-            norm = zero;
-            for (int k = 0; k < NDim; k++) {
-              vec2[k] =  coords1D[k][i] - coords1D[k][cpts[1]];
-              norm   += vec2[k]*vec2[k];
-            }
-            norm = sqrt(norm);
-            for (int k = 0; k < NDim; k++)
-              vec2[k] /= norm;
-
-          } else {
-            for (int k = 0; k < NDim; k++)
-              vec2[k] = vec1[k];
-          }
+        } else {
+          for (int k = 0; k < NDim; k++)
+            vec2[k] = vec1[k];
+        }
 
-          for (int k = 0; k < numNearbyCs; k++) score[k] = 0;
-          for (int k = 0; k < numNearbyCs; k++) dists[k] = 0;
+        for (int k = 0; k < numNearbyCs; k++) score[k] = 0;
+        for (int k = 0; k < numNearbyCs; k++) dists[k] = 0;
 
-          for (int j = 0; j < numNearbyCs; j++) {
-            SC newVec[3];
+        for (int j = 0; j < numNearbyCs; j++) {
+          SC newVec[3];
 
-            norm = 0;
-            for (int k = 0; k < NDim; k++) {
-              newVec[k] = coords1D[k][nearbyCs[j]] - coords1D[k][i];
-              norm += newVec[k]*newVec[k];
-            }
-            norm = sqrt(norm);
-            for (int k = 0; k < NDim; k++)
-              newVec[k] /= norm;
+          norm = 0;
+          for (int k = 0; k < NDim; k++) {
+            newVec[k] = coords1D[k][nearbyCs[j]] - coords1D[k][i];
+            norm += newVec[k] * newVec[k];
+          }
+          norm = sqrt(norm);
+          for (int k = 0; k < NDim; k++)
+            newVec[k] /= norm;
 
-            score[j] = 0;
-            for (int k = 0; k < NDim; k++)
-              score[j] += newVec[k]*(vec1[k] + vec2[k]);
-            // Why??
-            score[j] /= 2;
+          score[j] = 0;
+          for (int k = 0; k < NDim; k++)
+            score[j] += newVec[k] * (vec1[k] + vec2[k]);
+          // Why??
+          score[j] /= 2;
 
-            dists[j] = norm;
-          }
+          dists[j] = norm;
+        }
 
-          // Normalize distances
-          double maxDist = 0.;
-          for (int j = 0; j < numNearbyCs; j++)
-            if (maxDist < dists[j])
-              maxDist = dists[j];
-
-          for (int j = 0; j < numNearbyCs; j++)
-            dists[j] /= maxDist;
-
-          const double distWeight   =  0.3;
-          double       maxComposite = -10000;
-          int          maxIndex     = -1;
-          for (int j = 0; j < numNearbyCs; j++) {
-            // The formula is
-            //     if (score[j] - distWeight*dists[j] > maxComposite)
-            // MATCH_MATLAB
-            double composite = score[j] - distWeight*dists[j] + 1.0e-7*(nearbyCs[j]-1);
-            if (maxComposite < composite) {
-              maxComposite = composite;
-              maxIndex     = j;
-            }
+        // Normalize distances
+        double maxDist = 0.;
+        for (int j = 0; j < numNearbyCs; j++)
+          if (maxDist < dists[j])
+            maxDist = dists[j];
+
+        for (int j = 0; j < numNearbyCs; j++)
+          dists[j] /= maxDist;
+
+        const double distWeight = 0.3;
+        double maxComposite     = -10000;
+        int maxIndex            = -1;
+        for (int j = 0; j < numNearbyCs; j++) {
+          // The formula is
+          //     if (score[j] - distWeight*dists[j] > maxComposite)
+          // MATCH_MATLAB
+          double composite = score[j] - distWeight * dists[j] + 1.0e-7 * (nearbyCs[j] - 1);
+          if (maxComposite < composite) {
+            maxComposite = composite;
+            maxIndex     = j;
           }
+        }
 
-          if (score[maxIndex] - 0.2*numCDepends > -0.3) {
-            TEUCHOS_TEST_FOR_EXCEPTION(numCpts[i] >= myCpts.getNnzPerRow(),
-                 Exceptions::RuntimeError, "Increase max number of C points per row");
-            myCpts(i)[numCpts[i]++] = nearbyCs[maxIndex];
-          }
+        if (score[maxIndex] - 0.2 * numCDepends > -0.3) {
+          TEUCHOS_TEST_FOR_EXCEPTION(numCpts[i] >= myCpts.getNnzPerRow(),
+                                     Exceptions::RuntimeError, "Increase max number of C points per row");
+          myCpts(i)[numCpts[i]++] = nearbyCs[maxIndex];
         }
       }
     }
   }
+}
 
-  // Compute mid-points associated with adjacent Cpts.
-  //
-  // Basically, the location of each Fpoint is compared with the average
-  // location of each Cpoint in Pat(Fpoint,:). If the location of the Fpoint is
-  // close to this average it is considered as a possible mid-point. In
-  // addition, however, we look to see if a possible mid-point is "close" or
-  // not to an already computed mid-point. If it is NOT too close, then this
-  // possible mid-point is declared to be an actual mid-point.
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindMidPoints(const Matrix& A, const MultiVector& coords, Array<LO>& Cptlist, const MyCptList& myCpts) const {
-    int    NDim    = coords.getNumVectors();
-    size_t numRows = A.getLocalNumRows();
-
-    const ParameterList& pL = GetParameterList();
-    double tau_2 = pL.get<double>("tau_2");
-    // In calculations, we use tau_2^2
-    tau_2 = tau_2 * tau_2;
-
-    const std::vector<short>& numCpts = myCpts.getNumCpts();
-
-    ArrayRCP<const size_t> ia;
-    ArrayRCP<const LO>     ja;
-    CreateCrsPointers(A, ia, ja);
-
-    ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
-    for (int k = 0; k < NDim; k++)
-      coords1D[k] = coords.getData(k);
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero();
+// Compute mid-points associated with adjacent Cpts.
+//
+// Basically, the location of each Fpoint is compared with the average
+// location of each Cpoint in Pat(Fpoint,:). If the location of the Fpoint is
+// close to this average it is considered as a possible mid-point. In
+// addition, however, we look to see if a possible mid-point is "close" or
+// not to an already computed mid-point. If it is NOT too close, then this
+// possible mid-point is declared to be an actual mid-point.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindMidPoints(const Matrix& A, const MultiVector& coords, Array<LO>& Cptlist, const MyCptList& myCpts) const {
+  int NDim       = coords.getNumVectors();
+  size_t numRows = A.getLocalNumRows();
+
+  const ParameterList& pL = GetParameterList();
+  double tau_2            = pL.get<double>("tau_2");
+  // In calculations, we use tau_2^2
+  tau_2 = tau_2 * tau_2;
+
+  const std::vector<short>& numCpts = myCpts.getNumCpts();
+
+  ArrayRCP<const size_t> ia;
+  ArrayRCP<const LO> ja;
+  CreateCrsPointers(A, ia, ja);
+
+  ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
+  for (int k = 0; k < NDim; k++)
+    coords1D[k] = coords.getData(k);
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero();
+
+  // Calculate number of nonzeros per row, make it negative, and then sort.
+  // The idea is that when assigning midpoints, we want to start by looking
+  // at points which have many coarse point dependencies
+  std::vector<int> nnzPerRow(numRows);
+  std::vector<int> index(numRows);
+  for (size_t i = 0; i < numRows; i++) {
+    nnzPerRow[i] = (numCpts[i] ? numCpts[i] : 1);
+    nnzPerRow[i] = -100000 * numCpts[i] + i;
+    index[i]     = i;
+  }
 
-    // Calculate number of nonzeros per row, make it negative, and then sort.
-    // The idea is that when assigning midpoints, we want to start by looking
-    // at points which have many coarse point dependencies
-    std::vector<int> nnzPerRow(numRows);
-    std::vector<int> index    (numRows);
-    for (size_t i = 0; i < numRows; i++) {
-      nnzPerRow[i] = (numCpts[i] ? numCpts[i] : 1);
-      nnzPerRow[i] = -100000*numCpts[i] + i;
-      index    [i] = i;
-    }
+  // Sort only for the purposes of filling 'index', which determines the
+  // order that we search for possible midpoints
+  Muelu_az_sort<SC>(&nnzPerRow[0], numRows, &index[0], NULL);
+
+  // Reset so that we have unsorted version of nnzPerRow and also mark points
+  // which cannot be mid points
+  std::vector<char> lookedAt(numRows, 'n');
+  for (size_t i = 0; i < numRows; i++) {
+    nnzPerRow[i] = (numCpts[i] ? numCpts[i] : 1);
+    if (nnzPerRow[i] == 1)
+      lookedAt[i] = 'y';
+  }
+  for (int i = 0; i < Cptlist.size(); i++)
+    lookedAt[Cptlist[i]] = 'y';
 
-    // Sort only for the purposes of filling 'index', which determines the
-    // order that we search for possible midpoints
-    Muelu_az_sort<SC>(&nnzPerRow[0], numRows, &index[0], NULL);
+  // Compute some target midpoints based on taking averages associated with
+  // the sparsity pattern and coarse grid point locations.
+  ArrayRCP<ArrayRCP<SC> > targetMidCoords1D(NDim);
+  for (int k = 0; k < NDim; k++) {
+    ArrayRCP<SC>& target1D = targetMidCoords1D[k];
 
-    // Reset so that we have unsorted version of nnzPerRow and also mark points
-    // which cannot be mid points
-    std::vector<char> lookedAt(numRows, 'n');
+    target1D.resize(numRows);
     for (size_t i = 0; i < numRows; i++) {
-      nnzPerRow[i] = (numCpts[i] ? numCpts[i] : 1);
-      if (nnzPerRow[i] == 1)
-        lookedAt[i] = 'y';
-    }
-    for (int i = 0; i < Cptlist.size(); i++)
-      lookedAt[Cptlist[i]] = 'y';
-
-    // Compute some target midpoints based on taking averages associated with
-    // the sparsity pattern and coarse grid point locations.
-    ArrayRCP<ArrayRCP<SC> > targetMidCoords1D(NDim);
-    for (int k = 0; k < NDim; k++) {
-      ArrayRCP<SC>& target1D = targetMidCoords1D[k];
+      target1D[i] = zero;
 
-      target1D.resize(numRows);
-      for (size_t i = 0; i < numRows; i++) {
-        target1D[i] = zero;
-
-        for (int j = 0; j < numCpts[i]; j++)
-          target1D[i] += coords1D[k][myCpts(i)[j]];
+      for (int j = 0; j < numCpts[i]; j++)
+        target1D[i] += coords1D[k][myCpts(i)[j]];
 
-        target1D[i] /= nnzPerRow[i];
-      }
+      target1D[i] /= nnzPerRow[i];
     }
+  }
 
-    std::vector<char> isMidPoint(numRows, 'n');
-    std::vector<char> inNearbyCs(numRows, 'n');
-    std::vector<char> inNeighs  (numRows, 'n');
-    std::vector<int>  neighs(numRows);
-    std::vector<int>  sameCGroup(50);
+  std::vector<char> isMidPoint(numRows, 'n');
+  std::vector<char> inNearbyCs(numRows, 'n');
+  std::vector<char> inNeighs(numRows, 'n');
+  std::vector<int> neighs(numRows);
+  std::vector<int> sameCGroup(50);
 
-    int numMidPoints = 0;
-    for (size_t i = 0; i < numRows; i++) {
-      int curF = index[i];
+  int numMidPoints = 0;
+  for (size_t i = 0; i < numRows; i++) {
+    int curF = index[i];
 
-      if (lookedAt[curF] == 'y')
-        continue;
-      lookedAt[curF] = 'y';
+    if (lookedAt[curF] == 'y')
+      continue;
+    lookedAt[curF] = 'y';
 
-      const LO* curFCs = myCpts(curF);
+    const LO* curFCs = myCpts(curF);
 
-      for (int j = 0; j < numCpts[curF]; j++)
-        inNearbyCs[curFCs[j]] = 'y';
+    for (int j = 0; j < numCpts[curF]; j++)
+      inNearbyCs[curFCs[j]] = 'y';
 
-      // Find all FPOINTs with the same curFCs (perhaps
-      // containing additional curFCs) as curF and
-      // put them in sameCGroup
-      int numNeigh = 0;
-      neighs[numNeigh++] = curF;
-      inNeighs[curF] = 'y';
+    // Find all FPOINTs with the same curFCs (perhaps
+    // containing additional curFCs) as curF and
+    // put them in sameCGroup
+    int numNeigh       = 0;
+    neighs[numNeigh++] = curF;
+    inNeighs[curF]     = 'y';
 
-      int nextLayerStart = 0;
-      int nextLayerEnd   = 0;
-      int numSameGrp     = 0;
+    int nextLayerStart = 0;
+    int nextLayerEnd   = 0;
+    int numSameGrp     = 0;
 
-      int flag           = 1;
-      while (flag == 1) {
-        flag = 0;
-
-        for (int k = nextLayerStart; k <= nextLayerEnd; k++) {
-          LO        curNeigh = neighs[k];
-          const LO* neighCs  = myCpts(curNeigh);
+    int flag = 1;
+    while (flag == 1) {
+      flag = 0;
 
-          // Check if subset of this neighbor's CPOINT dependencies include all
-          // the CPOINT dependencies of curF
-          int sum = 0;
-          for (int j = 0; j < numCpts[curNeigh]; j++)
-            if (inNearbyCs[neighCs[j]] == 'y')
-              sum++;
+      for (int k = nextLayerStart; k <= nextLayerEnd; k++) {
+        LO curNeigh       = neighs[k];
+        const LO* neighCs = myCpts(curNeigh);
 
-          if (sum == nnzPerRow[curF]) {
-            lookedAt[curNeigh] = 'y';
+        // Check if subset of this neighbor's CPOINT dependencies include all
+        // the CPOINT dependencies of curF
+        int sum = 0;
+        for (int j = 0; j < numCpts[curNeigh]; j++)
+          if (inNearbyCs[neighCs[j]] == 'y')
+            sum++;
 
-            // Make sure we have enough space
-            if (Teuchos::as<int>(sameCGroup.size()) <= numSameGrp)
-              sameCGroup.resize(2*numSameGrp);
+        if (sum == nnzPerRow[curF]) {
+          lookedAt[curNeigh] = 'y';
 
-            sameCGroup[numSameGrp++] = curNeigh;
-            flag = 1;
-          }
+          // Make sure we have enough space
+          if (Teuchos::as<int>(sameCGroup.size()) <= numSameGrp)
+            sameCGroup.resize(2 * numSameGrp);
 
-          // Add neighbors of curNeigh that haven't already been
-          // add to the neighbor list while processing curF
-          for (size_t j = ia[curNeigh]; j < ia[curNeigh+1]; j++)
-            if (inNeighs[ja[j]] == 'n') {
-              neighs[numNeigh++] = ja[j];
-              inNeighs[ja[j]]    = 'y';
-            }
+          sameCGroup[numSameGrp++] = curNeigh;
+          flag                     = 1;
         }
 
-        nextLayerStart = nextLayerEnd + 1;
-        nextLayerEnd   = numNeigh     - 1;
+        // Add neighbors of curNeigh that haven't already been
+        // add to the neighbor list while processing curF
+        for (size_t j = ia[curNeigh]; j < ia[curNeigh + 1]; j++)
+          if (inNeighs[ja[j]] == 'n') {
+            neighs[numNeigh++] = ja[j];
+            inNeighs[ja[j]]    = 'y';
+          }
       }
 
-      // Reset status arrays
-      for (int j = 0; j < numNeigh; j++)
-        inNeighs[neighs[j]] = 'n';
-      for (int j = 0; j < numCpts[curF]; j++)
-        inNearbyCs[curFCs[j]] = 'n';
-
-      // At this point we have now constructed a group of possible mid points
-      // all with the same Cpt dependencies. Now, we need to find the one in
-      // this group which is closest to the target midpoint coordinates.
-      double smallest      = 1e30;
-      int    smallestIndex = -1;
-      for (int j = 0; j < numSameGrp; j++) {
-        // MATCH_MATLAB
-        double dist = 1e-8*(sameCGroup[j]+1);
+      nextLayerStart = nextLayerEnd + 1;
+      nextLayerEnd   = numNeigh - 1;
+    }
 
-        for (int k = 0; k < NDim; k++) {
-          double dtemp = coords1D[k][sameCGroup[j]] - targetMidCoords1D[k][curF];
-          dist += dtemp*dtemp;
-        }
-        if (dist < smallest) {
-          smallest      = dist;
-          smallestIndex = sameCGroup[j];
-        }
+    // Reset status arrays
+    for (int j = 0; j < numNeigh; j++)
+      inNeighs[neighs[j]] = 'n';
+    for (int j = 0; j < numCpts[curF]; j++)
+      inNearbyCs[curFCs[j]] = 'n';
+
+    // At this point we have now constructed a group of possible mid points
+    // all with the same Cpt dependencies. Now, we need to find the one in
+    // this group which is closest to the target midpoint coordinates.
+    double smallest   = 1e30;
+    int smallestIndex = -1;
+    for (int j = 0; j < numSameGrp; j++) {
+      // MATCH_MATLAB
+      double dist = 1e-8 * (sameCGroup[j] + 1);
+
+      for (int k = 0; k < NDim; k++) {
+        double dtemp = coords1D[k][sameCGroup[j]] - targetMidCoords1D[k][curF];
+        dist += dtemp * dtemp;
       }
+      if (dist < smallest) {
+        smallest      = dist;
+        smallestIndex = sameCGroup[j];
+      }
+    }
 
-      // So now smallestIndex is the best midpoint candidate within sameCGroup.
-      // We now need to check if smallestIndex is really close to an already
-      // existing mid-point. In fact, we could have multiple copies of
-      // mid-points or some very close mid-points. To see this, consider
-      //
-      //                    P1              P2
-      //
-      //
-      //
-      //                    P3              P4
-      // where P1/P4's midpoint is the same as P2/P3's midpoint. We get rid of
-      // these by checking if a new potential midpoint is close to any previous
-      // midpoints.
-
-      // Check if anybody in sameCGroup is already a midpoint. If so, check
-      // each one of these midpoints to see if any of these is real close to
-      // the curF.
-      flag = 0;
-      for (int j = 0; j < numSameGrp; j++)
-        if (isMidPoint[sameCGroup[j]] == 'y')
-          flag = 1;
-
-      if (flag == 1) {
-        // Get an idea of the spacing between curFCs
-        double delta = 0.0;
-        for (int k = 0; k < NDim; k++) {
-          double dmin = coords1D[k][curFCs[0]];
-          double dmax = dmin;
-
-          for (int j = 1; j < numCpts[curF]; j++) {
-            SC c = coords1D[k][curFCs[j]];
-            if (c < dmin) dmin = c;
-            if (c > dmax) dmax = c;
-          }
-          delta += ((dmax-dmin)*(dmax-dmin));
-        }
+    // So now smallestIndex is the best midpoint candidate within sameCGroup.
+    // We now need to check if smallestIndex is really close to an already
+    // existing mid-point. In fact, we could have multiple copies of
+    // mid-points or some very close mid-points. To see this, consider
+    //
+    //                    P1              P2
+    //
+    //
+    //
+    //                    P3              P4
+    // where P1/P4's midpoint is the same as P2/P3's midpoint. We get rid of
+    // these by checking if a new potential midpoint is close to any previous
+    // midpoints.
+
+    // Check if anybody in sameCGroup is already a midpoint. If so, check
+    // each one of these midpoints to see if any of these is real close to
+    // the curF.
+    flag = 0;
+    for (int j = 0; j < numSameGrp; j++)
+      if (isMidPoint[sameCGroup[j]] == 'y')
+        flag = 1;
 
-        // Now find the closest point among all sameCGroup midPoints to
-        // smallestIndex. If this point is not too close, we make smallestIndex
-        // a new mid-point.
-        double close = 1000000.;
-        for (int j = 0; j < numSameGrp; j++) {
-          int t = sameCGroup[j];
-          if (isMidPoint[t] == 'y') {
-            double current = distance2(coords1D, smallestIndex, t);
-            if (current < close)
-              close = current;
-          }
+    if (flag == 1) {
+      // Get an idea of the spacing between curFCs
+      double delta = 0.0;
+      for (int k = 0; k < NDim; k++) {
+        double dmin = coords1D[k][curFCs[0]];
+        double dmax = dmin;
+
+        for (int j = 1; j < numCpts[curF]; j++) {
+          SC c = coords1D[k][curFCs[j]];
+          if (c < dmin) dmin = c;
+          if (c > dmax) dmax = c;
         }
+        delta += ((dmax - dmin) * (dmax - dmin));
+      }
 
-        if  (close/delta > tau_2) {
-          isMidPoint[smallestIndex] = 'y';
-          numMidPoints++;
+      // Now find the closest point among all sameCGroup midPoints to
+      // smallestIndex. If this point is not too close, we make smallestIndex
+      // a new mid-point.
+      double close = 1000000.;
+      for (int j = 0; j < numSameGrp; j++) {
+        int t = sameCGroup[j];
+        if (isMidPoint[t] == 'y') {
+          double current = distance2(coords1D, smallestIndex, t);
+          if (current < close)
+            close = current;
         }
+      }
 
-      } else {
+      if (close / delta > tau_2) {
         isMidPoint[smallestIndex] = 'y';
         numMidPoints++;
       }
+
+    } else {
+      isMidPoint[smallestIndex] = 'y';
+      numMidPoints++;
     }
+  }
 
-    // This loop also sorts mid points
-    int count = 0;
-    for (size_t i = 0; i < numRows; i++)
-      if (isMidPoint[i] == 'y') {
-        Cptlist.push_back(i);
-        count++;
-      }
+  // This loop also sorts mid points
+  int count = 0;
+  for (size_t i = 0; i < numRows; i++)
+    if (isMidPoint[i] == 'y') {
+      Cptlist.push_back(i);
+      count++;
+    }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(count != numMidPoints, Exceptions::RuntimeError,
-                               "Wrong with the number of mid points: " << count << " vs. " << numMidPoints);
-  }
+  TEUCHOS_TEST_FOR_EXCEPTION(count != numMidPoints, Exceptions::RuntimeError,
+                             "Wrong with the number of mid points: " << count << " vs. " << numMidPoints);
+}
 
-  // Convert information in Cptlist, myCpts into a sparsity pattern matrix
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CptDepends2Pattern(const Matrix& A, const MyCptList& myCpts, RCP<Matrix>& P, LocalOrdinal offset) const {
-    RCP<const Map> rowMap = A.getRowMap();
-    size_t numRows = myCpts.getLocalNumRows();
+// Convert information in Cptlist, myCpts into a sparsity pattern matrix
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CptDepends2Pattern(const Matrix& A, const MyCptList& myCpts, RCP<Matrix>& P, LocalOrdinal offset) const {
+  RCP<const Map> rowMap = A.getRowMap();
+  size_t numRows        = myCpts.getLocalNumRows();
 
-    // FIXME: how does offset play here?
-    const Array<LO>& Cptlist = myCpts.getCList();
-    RCP<const Map> coarseMap = MapFactory::Build(rowMap->lib(), Cptlist.size(), rowMap->getIndexBase() + offset, rowMap->getComm());
+  // FIXME: how does offset play here?
+  const Array<LO>& Cptlist = myCpts.getCList();
+  RCP<const Map> coarseMap = MapFactory::Build(rowMap->lib(), Cptlist.size(), rowMap->getIndexBase() + offset, rowMap->getComm());
 
-    P = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
-    RCP<CrsMatrix> Pcrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  P                   = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
+  RCP<CrsMatrix> Pcrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
 
-    ArrayRCP<size_t>  iaP;
-    ArrayRCP<LO>      jaP;
-    ArrayRCP<SC>     valP;
+  ArrayRCP<size_t> iaP;
+  ArrayRCP<LO> jaP;
+  ArrayRCP<SC> valP;
 
-    const std::vector<short>& numCpts = myCpts.getNumCpts();
-    size_t nnzEstimate = std::accumulate(numCpts.begin(), numCpts.end(), 0);
+  const std::vector<short>& numCpts = myCpts.getNumCpts();
+  size_t nnzEstimate                = std::accumulate(numCpts.begin(), numCpts.end(), 0);
 
-    Pcrs->allocateAllValues(nnzEstimate, iaP, jaP, valP);
+  Pcrs->allocateAllValues(nnzEstimate, iaP, jaP, valP);
 
-    ArrayView<size_t> ia  =  iaP();
-    ArrayView<LO>     ja  =  jaP();
-    ArrayView<SC>     val = valP();
+  ArrayView<size_t> ia = iaP();
+  ArrayView<LO> ja     = jaP();
+  ArrayView<SC> val    = valP();
 
-    std::vector<GO> coarseCmap(numRows, -1);
-    for (int i = 0; i < Cptlist.size(); i++)
-      coarseCmap[Cptlist[i]] = i;
+  std::vector<GO> coarseCmap(numRows, -1);
+  for (int i = 0; i < Cptlist.size(); i++)
+    coarseCmap[Cptlist[i]] = i;
 
-    SC one = Teuchos::ScalarTraits<SC>::one();
+  SC one = Teuchos::ScalarTraits<SC>::one();
 
-    // Build up the prolongator sparsity pattern and the initial
-    // guess used for Emin (which must have row sums equal to one)
+  // Build up the prolongator sparsity pattern and the initial
+  // guess used for Emin (which must have row sums equal to one)
 
-    ia[0] = 0;
-    size_t nnzCount = 0;
-    for (size_t i = 0; i < numRows; i++) {
-      const LO* cpts = myCpts(i);
-
-      for (int j = 0; j < numCpts[i]; j++) {
-        ja [nnzCount] = coarseCmap[cpts[j]];
-        TEUCHOS_TEST_FOR_EXCEPTION(ja[nnzCount] == -1, Exceptions::RuntimeError,
-            "Point " << cpts[j] << " is not in the global list of cpoints, but it is in the list for " << i);
-        val[nnzCount] = one/((SC) numCpts[i]);
-        nnzCount++;
-      }
-      // NOTE: we could theoretically sort here
-      // Do we need to, though?
-      ia[i+1] = nnzCount;
-    }
+  ia[0]           = 0;
+  size_t nnzCount = 0;
+  for (size_t i = 0; i < numRows; i++) {
+    const LO* cpts = myCpts(i);
 
-    if (rowMap->lib() == Xpetra::UseTpetra) {
-      // - Cannot resize for Epetra, as it checks for same pointers
-      // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
-      // NOTE: these invalidate ja and val views
-      jaP .resize(nnzCount);
-      valP.resize(nnzCount);
+    for (int j = 0; j < numCpts[i]; j++) {
+      ja[nnzCount] = coarseCmap[cpts[j]];
+      TEUCHOS_TEST_FOR_EXCEPTION(ja[nnzCount] == -1, Exceptions::RuntimeError,
+                                 "Point " << cpts[j] << " is not in the global list of cpoints, but it is in the list for " << i);
+      val[nnzCount] = one / ((SC)numCpts[i]);
+      nnzCount++;
     }
+    // NOTE: we could theoretically sort here
+    // Do we need to, though?
+    ia[i + 1] = nnzCount;
+  }
 
-    Pcrs->setAllValues(iaP, jaP, valP);
-    Pcrs->expertStaticFillComplete(coarseMap, A.getDomainMap());
+  if (rowMap->lib() == Xpetra::UseTpetra) {
+    // - Cannot resize for Epetra, as it checks for same pointers
+    // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
+    // NOTE: these invalidate ja and val views
+    jaP.resize(nnzCount);
+    valP.resize(nnzCount);
   }
 
-  // Compute all points which are within a distance 1-4 from StartPt
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CompDistances(const Matrix& A, LO start, int numDist, std::vector<LO>& dist1, std::vector<LO>& dist2, std::vector<LO>& dist3, std::vector<LO>& dist4) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(numDist < 1 || numDist > 4, Exceptions::InvalidArgument, "CompDistances() cannot compute " << numDist << " distances");
+  Pcrs->setAllValues(iaP, jaP, valP);
+  Pcrs->expertStaticFillComplete(coarseMap, A.getDomainMap());
+}
 
-    size_t numRows = A.getGlobalNumRows();
+// Compute all points which are within a distance 1-4 from StartPt
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CompDistances(const Matrix& A, LO start, int numDist, std::vector<LO>& dist1, std::vector<LO>& dist2, std::vector<LO>& dist3, std::vector<LO>& dist4) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(numDist < 1 || numDist > 4, Exceptions::InvalidArgument, "CompDistances() cannot compute " << numDist << " distances");
 
-    ArrayRCP<const size_t> ia;
-    ArrayRCP<const LO>     ja;
-    CreateCrsPointers(A, ia, ja);
+  size_t numRows = A.getGlobalNumRows();
 
-    std::vector<char> added(numRows, 'n');
-    std::vector<LO>   neighs;
-    neighs.reserve(100);
+  ArrayRCP<const size_t> ia;
+  ArrayRCP<const LO> ja;
+  CreateCrsPointers(A, ia, ja);
 
-    neighs.push_back(start);
-    added[start] = 'y';
+  std::vector<char> added(numRows, 'n');
+  std::vector<LO> neighs;
+  neighs.reserve(100);
 
-    for (int k = 1; k <= numDist; k++) {
+  neighs.push_back(start);
+  added[start] = 'y';
 
-      int numNeighs = neighs.size();
-      for (int i = 0; i < numNeighs; i++)
-        for (size_t j = ia[neighs[i]]; j < ia[neighs[i]+1]; j++)
-          if (added[ja[j]] == 'n') {
-            added[ja[j]] = 'y';
-            neighs.push_back(ja[j]);
-          }
+  for (int k = 1; k <= numDist; k++) {
+    int numNeighs = neighs.size();
+    for (int i = 0; i < numNeighs; i++)
+      for (size_t j = ia[neighs[i]]; j < ia[neighs[i] + 1]; j++)
+        if (added[ja[j]] == 'n') {
+          added[ja[j]] = 'y';
+          neighs.push_back(ja[j]);
+        }
 
-      if (k == 1) dist1 = neighs;
-      if (k == 2) dist2 = neighs;
-      if (k == 3) dist3 = neighs;
-      if (k == 4) dist4 = neighs;
-    }
+    if (k == 1) dist1 = neighs;
+    if (k == 2) dist2 = neighs;
+    if (k == 3) dist3 = neighs;
+    if (k == 4) dist4 = neighs;
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CreateCrsPointers(const Matrix& A, ArrayRCP<const size_t>& ia, ArrayRCP<const LO>& ja) const {
-    RCP<const CrsMatrixWrap> Awrap = rcp_dynamic_cast<const CrsMatrixWrap>(rcpFromRef(A));
-    TEUCHOS_TEST_FOR_EXCEPTION(Awrap.is_null(), Exceptions::RuntimeError, "A is not of CrsMatrixWrap type");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CreateCrsPointers(const Matrix& A, ArrayRCP<const size_t>& ia, ArrayRCP<const LO>& ja) const {
+  RCP<const CrsMatrixWrap> Awrap = rcp_dynamic_cast<const CrsMatrixWrap>(rcpFromRef(A));
+  TEUCHOS_TEST_FOR_EXCEPTION(Awrap.is_null(), Exceptions::RuntimeError, "A is not of CrsMatrixWrap type");
 
-    ArrayRCP<const SC> val;
-    Awrap->getCrsMatrix()->getAllValues(ia, ja, val);
-  }
+  ArrayRCP<const SC> val;
+  Awrap->getCrsMatrix()->getAllValues(ia, ja, val);
+}
 
-  const std::string OUTPUT_DIR = "status/";
+const std::string OUTPUT_DIR = "status/";
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DumpStatus(const std::string& filename, const std::vector<char>& status, int NDim, bool isAmalgamated) const {
-    const std::string dirName = OUTPUT_DIR;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DumpStatus(const std::string& filename, const std::vector<char>& status, int NDim, bool isAmalgamated) const {
+  const std::string dirName = OUTPUT_DIR;
 
-    struct stat sb;
-    TEUCHOS_TEST_FOR_EXCEPTION(stat(dirName.c_str(), &sb) != 0 || !S_ISDIR(sb.st_mode), Exceptions::RuntimeError,
-                               "Please create a \"" << dirName << "\" directory");
+  struct stat sb;
+  TEUCHOS_TEST_FOR_EXCEPTION(stat(dirName.c_str(), &sb) != 0 || !S_ISDIR(sb.st_mode), Exceptions::RuntimeError,
+                             "Please create a \"" << dirName << "\" directory");
 
-    std::ofstream ofs((dirName + filename).c_str());
-    size_t step = (isAmalgamated ? 1 : NDim);
-    for (size_t i = 0; i < status.size(); i += step)
-      ofs << status[i] << std::endl;
-  }
+  std::ofstream ofs((dirName + filename).c_str());
+  size_t step = (isAmalgamated ? 1 : NDim);
+  for (size_t i = 0; i < status.size(); i += step)
+    ofs << status[i] << std::endl;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DumpCoords(const std::string& filename, const MultiVector& coords) const {
-    const std::string dirName = OUTPUT_DIR;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1uPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DumpCoords(const std::string& filename, const MultiVector& coords) const {
+  const std::string dirName = OUTPUT_DIR;
 
-    struct stat sb;
-    if (stat(dirName.c_str(), &sb) != 0 || !S_ISDIR(sb.st_mode))
-      GetOStream(Errors) << "Please create a \"" << dirName << "\" directory" << std::endl;
+  struct stat sb;
+  if (stat(dirName.c_str(), &sb) != 0 || !S_ISDIR(sb.st_mode))
+    GetOStream(Errors) << "Please create a \"" << dirName << "\" directory" << std::endl;
 
-    const int NDim = coords.getNumVectors();
-    const int n    = coords.getLocalLength();
+  const int NDim = coords.getNumVectors();
+  const int n    = coords.getLocalLength();
 
-    ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
-    for (int k = 0; k < NDim; k++)
-      coords1D[k] = coords.getData(k);
+  ArrayRCP<ArrayRCP<const SC> > coords1D(NDim);
+  for (int k = 0; k < NDim; k++)
+    coords1D[k] = coords.getData(k);
 
-    std::ofstream ofs((dirName + filename).c_str());
-    for (int i = 0; i < n; i++) {
-      for (int k = 0; k < NDim; k++)
-        ofs << " " << coords1D[k][i];
-      ofs << std::endl;
-    }
+  std::ofstream ofs((dirName + filename).c_str());
+  for (int i = 0; i < n; i++) {
+    for (int k = 0; k < NDim; k++)
+      ofs << " " << coords1D[k][i];
+    ofs << std::endl;
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_Q2Q1UPFACTORY_DECL_HPP
+#endif  // MUELU_Q2Q1UPFACTORY_DECL_HPP
diff --git a/packages/muelu/research/q2q1/Q2Q1.cpp b/packages/muelu/research/q2q1/Q2Q1.cpp
index 8608c62f6906..ef80b7dff5a6 100644
--- a/packages/muelu/research/q2q1/Q2Q1.cpp
+++ b/packages/muelu/research/q2q1/Q2Q1.cpp
@@ -101,30 +101,31 @@
 // FIXME
 #include "MueLu_UseDefaultTypes.hpp"
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
 ReadBinary(const std::string& fileName, const Teuchos::RCP<const Teuchos::Comm<int> >& comm) {
-  typedef Scalar        SC;
-  typedef LocalOrdinal  LO;
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
   typedef GlobalOrdinal GO;
-  typedef Node          NO;
+  typedef Node NO;
   TEUCHOS_TEST_FOR_EXCEPTION(comm->getSize() != 1, MueLu::Exceptions::RuntimeError, "Serial read only");
 
   std::ifstream ifs(fileName.c_str(), std::ios::binary);
   TEUCHOS_TEST_FOR_EXCEPTION(!ifs.good(), MueLu::Exceptions::RuntimeError, "Can not read \"" << fileName << "\"");
 
   int m, n, nnz;
-  ifs.read(reinterpret_cast<char*>(&m),   sizeof(m));
-  ifs.read(reinterpret_cast<char*>(&n),   sizeof(n));
+  ifs.read(reinterpret_cast<char*>(&m), sizeof(m));
+  ifs.read(reinterpret_cast<char*>(&n), sizeof(n));
   ifs.read(reinterpret_cast<char*>(&nnz), sizeof(nnz));
 
-  typedef Tpetra::Map      <LO,GO,NO>       tMap;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>    tCrsMatrix;
+  typedef Tpetra::Map<LO, GO, NO> tMap;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> tCrsMatrix;
 
-  GO indexBase = 0;
-  Teuchos::RCP<tMap>       rowMap = rcp(new tMap(m, indexBase, comm)), rangeMap  = rowMap;
-  Teuchos::RCP<tMap>       colMap = rcp(new tMap(n, indexBase, comm)), domainMap = colMap;;
-  Teuchos::RCP<tCrsMatrix> A      = rcp(new tCrsMatrix(rowMap, colMap, 9));
+  GO indexBase              = 0;
+  Teuchos::RCP<tMap> rowMap = rcp(new tMap(m, indexBase, comm)), rangeMap = rowMap;
+  Teuchos::RCP<tMap> colMap = rcp(new tMap(n, indexBase, comm)), domainMap = colMap;
+  ;
+  Teuchos::RCP<tCrsMatrix> A = rcp(new tCrsMatrix(rowMap, colMap, 9));
 
   TEUCHOS_TEST_FOR_EXCEPTION(sizeof(int) != sizeof(GO), MueLu::Exceptions::RuntimeError, "Incompatible sizes");
 
@@ -132,7 +133,7 @@ ReadBinary(const std::string& fileName, const Teuchos::RCP<const Teuchos::Comm<i
   Teuchos::Array<SC> vals;
   for (int i = 0; i < m; i++) {
     int row, rownnz;
-    ifs.read(reinterpret_cast<char*>(&row),    sizeof(row));
+    ifs.read(reinterpret_cast<char*>(&row), sizeof(row));
     ifs.read(reinterpret_cast<char*>(&rownnz), sizeof(rownnz));
     inds.resize(rownnz);
     vals.resize(rownnz);
@@ -154,20 +155,20 @@ ReadBinary(const std::string& fileName, const Teuchos::RCP<const Teuchos::Comm<i
   return A;
 }
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::ParameterList;
   using Teuchos::Array;
   using Teuchos::ArrayRCP;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::null;
   using Teuchos::as;
+  using Teuchos::null;
+  using Teuchos::ParameterList;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  using Teuchos::rcp_dynamic_cast;
   using Teuchos::TimeMonitor;
-  using Tpetra::MatrixMarket::Reader;
   using Thyra::tpetraVectorSpace;
+  using Tpetra::MatrixMarket::Reader;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -190,36 +191,47 @@ int main(int argc, char *argv[]) {
     Xpetra::Parameters xpetraParameters(clp);
 
     // configure problem
-    std::string prefix = "./Q2Q1_9x9_";      clp.setOption("prefix",     &prefix,        "prefix for data files");
-    std::string rhs    = "";                    clp.setOption("rhs",        &rhs,           "rhs");
+    std::string prefix = "./Q2Q1_9x9_";
+    clp.setOption("prefix", &prefix, "prefix for data files");
+    std::string rhs = "";
+    clp.setOption("rhs", &rhs, "rhs");
 
     // configure run
-    std::string xmlFileName  = "driver.xml";    clp.setOption("xml",        &xmlFileName,   "read parameters from a file [default = 'driver.xml']");
-    double      tol          = 1e-8;            clp.setOption("tol",        &tol,           "solver convergence tolerance");
-    std::string type         = "unstructured";  clp.setOption("type",       &type,          "structured/unstructured");
-    int         use9ptPatA   = 1;               clp.setOption("use9pt",     &use9ptPatA,    "use 9-point stencil matrix for velocity prolongator construction");
-    int         useFilters   = 1;               clp.setOption("usefilters", &useFilters,    "use filters on A and BB^T");
-
-    double      tau_1        = 0.06;            clp.setOption("tau_1",      &tau_1,         "tau_1 parameter from paper (for matrix filtering)");
-    double      tau_2        = sqrt(0.0015);    clp.setOption("tau_2",      &tau_2,         "tau_2 parameter from paper (for mid point detection)");
-
-    int         binary       = 0;               clp.setOption("binary",     &binary,        "read matrix in binary format");
+    std::string xmlFileName = "driver.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file [default = 'driver.xml']");
+    double tol = 1e-8;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+    std::string type = "unstructured";
+    clp.setOption("type", &type, "structured/unstructured");
+    int use9ptPatA = 1;
+    clp.setOption("use9pt", &use9ptPatA, "use 9-point stencil matrix for velocity prolongator construction");
+    int useFilters = 1;
+    clp.setOption("usefilters", &useFilters, "use filters on A and BB^T");
+
+    double tau_1 = 0.06;
+    clp.setOption("tau_1", &tau_1, "tau_1 parameter from paper (for matrix filtering)");
+    double tau_2 = sqrt(0.0015);
+    clp.setOption("tau_2", &tau_2, "tau_2 parameter from paper (for mid point detection)");
+
+    int binary = 0;
+    clp.setOption("binary", &binary, "read matrix in binary format");
 
     // configure misc
-    int         printTimings = 0;               clp.setOption("timings",    &printTimings,  "print timings to screen");
+    int printTimings = 0;
+    clp.setOption("timings", &printTimings, "print timings to screen");
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    typedef Tpetra::CrsMatrix<SC,LO,GO,NO>        tCrsMatrix;
-    typedef Tpetra::Operator<SC,LO,GO,NO>         tOperator;
-    typedef Tpetra::MultiVector<SC,LO,GO,NO>      tMultiVector;
-    typedef Tpetra::Map<LO,GO,NO>                 tMap;
-    typedef Thyra::TpetraVectorSpace<SC,LO,GO,NO> THTP_Vs;
+    typedef Tpetra::CrsMatrix<SC, LO, GO, NO> tCrsMatrix;
+    typedef Tpetra::Operator<SC, LO, GO, NO> tOperator;
+    typedef Tpetra::MultiVector<SC, LO, GO, NO> tMultiVector;
+    typedef Tpetra::Map<LO, GO, NO> tMap;
+    typedef Thyra::TpetraVectorSpace<SC, LO, GO, NO> THTP_Vs;
 
     // Read data from files
     RCP<tOperator> A11, A119Pt, A21, A12;
@@ -230,36 +242,37 @@ int main(int argc, char *argv[]) {
     try {
       if (!binary) {
         filename = prefix + "A.mm";
-        A11     = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
-        A119Pt  = A11;
+        A11      = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
+        A119Pt   = A11;
         if (use9ptPatA) {
           filename = prefix + "AForPat.mm";
-          A119Pt = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
+          A119Pt   = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
         }
       } else {
         filename = prefix + "A.dat";
-        A11     = ReadBinary<SC,LO,GO,NO>(filename.c_str(), comm);
-        A119Pt  = A11;
+        A11      = ReadBinary<SC, LO, GO, NO>(filename.c_str(), comm);
+        A119Pt   = A11;
         if (use9ptPatA) {
           filename = prefix + "AForPat.dat";
-          A119Pt = ReadBinary<SC,LO,GO,NO>(filename.c_str(), comm);
+          A119Pt   = ReadBinary<SC, LO, GO, NO>(filename.c_str(), comm);
         }
       }
       filename = prefix + "B.mm";
-      A21 = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
+      A21      = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
       filename = prefix + "Bt.mm";
-      A12 = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
+      A12      = Reader<tCrsMatrix>::readSparseFile(filename.c_str(), comm);
 
       RCP<const tMap> cmap1 = A11->getDomainMap(), cmap2 = A12->getDomainMap();
       filename = prefix + "VelCoords.mm";
-      Vcoords = Reader<tCrsMatrix>::readDenseFile(filename.c_str(),  comm, cmap1);
+      Vcoords  = Reader<tCrsMatrix>::readDenseFile(filename.c_str(), comm, cmap1);
       filename = prefix + "PresCoords.mm";
-      Pcoords = Reader<tCrsMatrix>::readDenseFile(filename.c_str(), comm, cmap2);
+      Pcoords  = Reader<tCrsMatrix>::readDenseFile(filename.c_str(), comm, cmap2);
 
       // For now, we assume that p2v maps local pressure DOF to a local x-velocity DOF
-      filename = prefix + "p2vMap.mm";
-      ArrayRCP<const SC> slop = Xpetra::IO<SC,LO,GO,NO>::ReadMultiVector(filename.c_str(),
-                                                        Xpetra::toXpetra(A21->getRangeMap()))->getData(0);
+      filename                = prefix + "p2vMap.mm";
+      ArrayRCP<const SC> slop = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(filename.c_str(),
+                                                                            Xpetra::toXpetra(A21->getRangeMap()))
+                                    ->getData(0);
       p2vMap.resize(slop.size());
       for (int i = 0; i < slop.size(); i++)
         p2vMap[i] = as<LO>(slop[i]);
@@ -272,14 +285,14 @@ int main(int argc, char *argv[]) {
     RCP<const THTP_Vs> domain12 = tpetraVectorSpace<SC>(A12->getDomainMap());
     RCP<const THTP_Vs> domain21 = tpetraVectorSpace<SC>(A21->getDomainMap());
 
-    RCP<const THTP_Vs> range11  = tpetraVectorSpace<SC>(A11->getRangeMap());
-    RCP<const THTP_Vs> range12  = tpetraVectorSpace<SC>(A12->getRangeMap());
-    RCP<const THTP_Vs> range21  = tpetraVectorSpace<SC>(A21->getRangeMap());
+    RCP<const THTP_Vs> range11 = tpetraVectorSpace<SC>(A11->getRangeMap());
+    RCP<const THTP_Vs> range12 = tpetraVectorSpace<SC>(A12->getRangeMap());
+    RCP<const THTP_Vs> range21 = tpetraVectorSpace<SC>(A21->getRangeMap());
 
-    Teko::LinearOp thA11        = Thyra::tpetraLinearOp<double>(range11, domain11, A11);
-    Teko::LinearOp thA12        = Thyra::tpetraLinearOp<double>(range12, domain12, A12);
-    Teko::LinearOp thA21        = Thyra::tpetraLinearOp<double>(range21, domain21, A21);
-    Teko::LinearOp thA11_9Pt    = Thyra::tpetraLinearOp<double>(range11, domain11, A119Pt);
+    Teko::LinearOp thA11     = Thyra::tpetraLinearOp<double>(range11, domain11, A11);
+    Teko::LinearOp thA12     = Thyra::tpetraLinearOp<double>(range12, domain12, A12);
+    Teko::LinearOp thA21     = Thyra::tpetraLinearOp<double>(range21, domain21, A21);
+    Teko::LinearOp thA11_9Pt = Thyra::tpetraLinearOp<double>(range11, domain11, A119Pt);
 
     // Bang together the parameter list. Right now, all the MueLu details is
     // hardwired in the MueLu-TpetraQ2Q1 classes. We probably want to switch
@@ -287,65 +300,65 @@ int main(int argc, char *argv[]) {
     // via parameter lists.
 
     RCP<ParameterList> stratimikosList = rcp(new ParameterList);
-    stratimikosList->set("Linear Solver Type",  "Belos");
+    stratimikosList->set("Linear Solver Type", "Belos");
     stratimikosList->set("Preconditioner Type", "MueLu-TpetraQ2Q1");
 
     ParameterList& BelosList = stratimikosList->sublist("Linear Solver Types").sublist("Belos");
-    BelosList.set("Solver Type", "Block GMRES"); // FIXME: should it be "Pseudo Block GMRES"?
-    BelosList.sublist("VerboseObject").set("Verbosity Level", "low"); // this is needed, as otherwise Stratimikos ignores Belos output
+    BelosList.set("Solver Type", "Block GMRES");                       // FIXME: should it be "Pseudo Block GMRES"?
+    BelosList.sublist("VerboseObject").set("Verbosity Level", "low");  // this is needed, as otherwise Stratimikos ignores Belos output
 
     ParameterList& GmresDetails = BelosList.sublist("Solver Types").sublist("Block GMRES");
-    GmresDetails.set("Maximum Iterations",      100);
-    GmresDetails.set("Convergence Tolerance",   tol);
-    GmresDetails.set("Verbosity",               Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    GmresDetails.set("Output Frequency",        1);
-    GmresDetails.set("Output Style",            Belos::Brief);
+    GmresDetails.set("Maximum Iterations", 100);
+    GmresDetails.set("Convergence Tolerance", tol);
+    GmresDetails.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    GmresDetails.set("Output Frequency", 1);
+    GmresDetails.set("Output Style", Belos::Brief);
 
     ParameterList& Q2Q1List = stratimikosList->sublist("Preconditioner Types").sublist("MueLu-TpetraQ2Q1");
-    Q2Q1List.set("useFilters",  useFilters);
-    Q2Q1List.set("tau_1",       tau_1);
-    Q2Q1List.set("tau_2",       tau_2);
-    Q2Q1List.set("Velcoords",   Vcoords);
-    Q2Q1List.set("Prescoords",  Pcoords);
-    Q2Q1List.set("p2vMap",      p2vMap);
-    Q2Q1List.set("A11",         thA11);
-    Q2Q1List.set("A12",         thA12);
-    Q2Q1List.set("A21",         thA21);
-    Q2Q1List.set("A11_9Pt",     thA11_9Pt);
+    Q2Q1List.set("useFilters", useFilters);
+    Q2Q1List.set("tau_1", tau_1);
+    Q2Q1List.set("tau_2", tau_2);
+    Q2Q1List.set("Velcoords", Vcoords);
+    Q2Q1List.set("Prescoords", Pcoords);
+    Q2Q1List.set("p2vMap", p2vMap);
+    Q2Q1List.set("A11", thA11);
+    Q2Q1List.set("A12", thA12);
+    Q2Q1List.set("A21", thA21);
+    Q2Q1List.set("A11_9Pt", thA11_9Pt);
 
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&Q2Q1List), *comm);
 
     std::cout << "Input parameters: " << *stratimikosList << std::endl;
 
     // Stratimikos vodou
-    typedef Thyra::PreconditionerFactoryBase<SC>             Base;
-    typedef Thyra::LinearOpWithSolveFactoryBase<SC>          LOWSFB;
-    typedef Thyra::LinearOpWithSolveBase<SC>                 LOWSB;
-    typedef Thyra::MultiVectorBase<SC>                       TH_Mvb;
+    typedef Thyra::PreconditionerFactoryBase<SC> Base;
+    typedef Thyra::LinearOpWithSolveFactoryBase<SC> LOWSFB;
+    typedef Thyra::LinearOpWithSolveBase<SC> LOWSB;
+    typedef Thyra::MultiVectorBase<SC> TH_Mvb;
 
     Stratimikos::LinearSolverBuilder<SC> linearSolverBuilder;
 
-    //Thyra::addMueLuToStratimikosBuilder(linearSolverBuilder);
+    // Thyra::addMueLuToStratimikosBuilder(linearSolverBuilder);
 
-    Stratimikos::enableMueLu<SC,LO,GO,NO>(linearSolverBuilder);                                          // Epetra
-    Stratimikos::enableMueLuTpetraQ2Q1<SC,LO,GO,NO>(linearSolverBuilder, "MueLu-TpetraQ2Q1");  // Tpetra
+    Stratimikos::enableMueLu<SC, LO, GO, NO>(linearSolverBuilder);                                // Epetra
+    Stratimikos::enableMueLuTpetraQ2Q1<SC, LO, GO, NO>(linearSolverBuilder, "MueLu-TpetraQ2Q1");  // Tpetra
 
     linearSolverBuilder.setParameterList(stratimikosList);
     RCP<const LOWSFB> lowsFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
-    RCP<LOWSB       > nsA = lowsFactory->createOp();
+    RCP<LOWSB> nsA                = lowsFactory->createOp();
 
     // I've hacked together a big matrix that does not use strided maps by
     // simply reading the data again. Normally, this would be supplied by Eric
     // Cyr and would be Teko operators.
 
-    int numElem = A12->getRangeMap()->getLocalNumElements() + A21->getRangeMap()->getLocalNumElements();
+    int numElem             = A12->getRangeMap()->getLocalNumElements() + A21->getRangeMap()->getLocalNumElements();
     RCP<const tMap> fullMap = Utilities::Map2TpetraMap(*(MapFactory::createUniformContigMap(Xpetra::UseTpetra, numElem, comm)));
 
     RCP<tOperator> A;
     if (!binary)
       A = Reader<tCrsMatrix>::readSparseFile((prefix + "BigA.mm").c_str(), fullMap, fullMap, fullMap, fullMap, true, true, false);
     else
-      A = ReadBinary<SC,LO,GO,NO>((prefix + "BigA.dat").c_str(), comm);
+      A = ReadBinary<SC, LO, GO, NO>((prefix + "BigA.dat").c_str(), comm);
 
     const RCP<Thyra::LinearOpBase<SC> > thA = Thyra::createLinearOp(A);
     Thyra::initializeOp<SC>(*lowsFactory, thA, nsA.ptr());
@@ -366,8 +379,8 @@ int main(int argc, char *argv[]) {
     // Set the initial guess Dirichlet points to the proper value.
     // This step is pretty important as the preconditioner may return zero at Dirichlet points
     ArrayRCP<const bool> dirBCs = Utilities::DetectDirichletRows(*MueLu::TpetraCrs_To_XpetraMatrix(rcp_dynamic_cast<tCrsMatrix>(A)));
-    ArrayRCP<SC>         tXdata = tX->getDataNonConst(0);
-    ArrayRCP<const SC>   tBdata = tB->getData(0);
+    ArrayRCP<SC> tXdata         = tX->getDataNonConst(0);
+    ArrayRCP<const SC> tBdata   = tB->getData(0);
     for (LO i = 0; i < tXdata.size(); i++)
       if (dirBCs[i])
         tXdata[i] = tBdata[i];
@@ -389,5 +402,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/research/regionMG/src/SetupRegionHierarchy_def.hpp b/packages/muelu/research/regionMG/src/SetupRegionHierarchy_def.hpp
index f760776a9f38..56cad37696e0 100644
--- a/packages/muelu/research/regionMG/src/SetupRegionHierarchy_def.hpp
+++ b/packages/muelu/research/regionMG/src/SetupRegionHierarchy_def.hpp
@@ -73,17 +73,16 @@
 #include "SetupRegionMatrix_def.hpp"
 #include "SetupRegionSmoothers_def.hpp"
 
-
 #if defined(HAVE_MUELU_AMESOS2)
 #include <Amesos2_config.h>
 #include <Amesos2.hpp>
 #endif
 
-using Teuchos::RCP;
-using Teuchos::ArrayRCP;
 using Teuchos::Array;
+using Teuchos::ArrayRCP;
 using Teuchos::ArrayView;
 using Teuchos::ParameterList;
+using Teuchos::RCP;
 
 /*! \brief Create coarse level maps with continuous GIDs
  *
@@ -95,12 +94,11 @@ using Teuchos::ParameterList;
  *  to deal with them in particular.
  */
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void createContinuousCoarseLevelMaps(const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap, ///< row map
-                                     const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > colMap, ///< column map
-                                     RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& contRowMap, ///< row map with continuous GIDs
-                                     RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& contColMap ///< column map with continuous GIDs
-                                     )
-{
+void createContinuousCoarseLevelMaps(const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap,  ///< row map
+                                     const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap,  ///< column map
+                                     RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& contRowMap,         ///< row map with continuous GIDs
+                                     RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& contColMap          ///< column map with continuous GIDs
+) {
 #include "Xpetra_UseShortNamesOrdinal.hpp"
   //   /!\ This function is pure ordinal, no scalar type is passed as input
   //       This means that use only three template paramters and that we
@@ -114,8 +112,7 @@ void createContinuousCoarseLevelMaps(const RCP<const Xpetra::Map<LocalOrdinal, G
                                  rowMap->getComm());
 
   return;
-} // createContinuousCoarseLevelMaps
-
+}  // createContinuousCoarseLevelMaps
 
 /* Reconstruct coarse-level maps (assuming fully structured grids)
  *
@@ -133,8 +130,7 @@ void createContinuousCoarseLevelMaps(const RCP<const Xpetra::Map<LocalOrdinal, G
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void MakeCoarseLevelMaps(const int maxRegPerGID,
-                         Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regHierarchy) {
-
+                         Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regHierarchy) {
 #include "Xpetra_UseShortNames.hpp"
 #include "MueLu_UseShortNames.hpp"
 
@@ -142,39 +138,39 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
 
   RCP<Level> level0 = regHierarchy->GetLevel(0);
 
-  const GO GO_INV = Teuchos::OrdinalTraits<GO>::invalid();
-  const int  numLevels = regHierarchy->GetNumLevels();
+  const GO GO_INV     = Teuchos::OrdinalTraits<GO>::invalid();
+  const int numLevels = regHierarchy->GetNumLevels();
 
   // RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
   // Teuchos::FancyOStream& out = *fancy;
   const int myRank = level0->GetComm()->getRank();
 
   Teuchos::Array<LO> coarseCompositeToRegionLIDs;
-  Teuchos::ArrayView<LO> compositeToRegionLIDs = level0->Get<ArrayView<LO> > ("compositeToRegionLIDs");
+  Teuchos::ArrayView<LO> compositeToRegionLIDs = level0->Get<ArrayView<LO>>("compositeToRegionLIDs");
 
-  for(int currentLevel = 1; currentLevel < numLevels; ++currentLevel) {
-    RCP<Level> level        = regHierarchy->GetLevel(currentLevel);
-    RCP<Matrix> regProlong  = level->Get<RCP<Matrix> >("P");
-    RCP<const Map> regRowMap      = regProlong->getColMap();
+  for (int currentLevel = 1; currentLevel < numLevels; ++currentLevel) {
+    RCP<Level> level         = regHierarchy->GetLevel(currentLevel);
+    RCP<Matrix> regProlong   = level->Get<RCP<Matrix>>("P");
+    RCP<const Map> regRowMap = regProlong->getColMap();
 
-    RCP<Level> levelFine    = regHierarchy->GetLevel(currentLevel-1);
-    RCP<Import> regRowImportFine   = levelFine->Get<RCP<Import> >("rowImport");
-    RCP<Matrix> regMatFine         = levelFine->Get<RCP<Matrix> >("A");
-    RCP<const Map> regRowMapFine         = regMatFine->getRowMap();
+    RCP<Level> levelFine         = regHierarchy->GetLevel(currentLevel - 1);
+    RCP<Import> regRowImportFine = levelFine->Get<RCP<Import>>("rowImport");
+    RCP<Matrix> regMatFine       = levelFine->Get<RCP<Matrix>>("A");
+    RCP<const Map> regRowMapFine = regMatFine->getRowMap();
 
     // Extracting some basic information about local mesh in composite/region format
     const size_t numFineRegionNodes    = regProlong->getLocalNumRows();
     const size_t numFineCompositeNodes = compositeToRegionLIDs.size();
     const size_t numFineDuplicateNodes = numFineRegionNodes - numFineCompositeNodes;
 
-    const size_t numCoarseRegionNodes  = regProlong->getColMap()->getLocalNumElements();
+    const size_t numCoarseRegionNodes = regProlong->getColMap()->getLocalNumElements();
 
     // Find the regionLIDs associated with local duplicated nodes
     // This will allow us to later loop only on duplicated nodes
     size_t countComposites = 0, countDuplicates = 0;
     Array<LO> fineDuplicateLIDs(numFineDuplicateNodes);
-    for(size_t regionIdx = 0; regionIdx < numFineRegionNodes; ++regionIdx) {
-      if(compositeToRegionLIDs[countComposites] == static_cast<LO>(regionIdx)) {
+    for (size_t regionIdx = 0; regionIdx < numFineRegionNodes; ++regionIdx) {
+      if (compositeToRegionLIDs[countComposites] == static_cast<LO>(regionIdx)) {
         ++countComposites;
       } else {
         fineDuplicateLIDs[countDuplicates] = regionIdx;
@@ -183,17 +179,16 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
     }
 
     // We gather the coarse GIDs associated with each fine point in the local composite mesh part.
-    RCP<Xpetra::Vector<MT,LO,GO,NO> > coarseCompositeGIDs
-      = Xpetra::VectorFactory<MT,LO,GO,NO>::Build(regRowImportFine->getSourceMap(), false);
-    Teuchos::ArrayRCP<MT> coarseCompositeGIDsData = coarseCompositeGIDs->getDataNonConst(0);
+    RCP<Xpetra::Vector<MT, LO, GO, NO>> coarseCompositeGIDs = Xpetra::VectorFactory<MT, LO, GO, NO>::Build(regRowImportFine->getSourceMap(), false);
+    Teuchos::ArrayRCP<MT> coarseCompositeGIDsData           = coarseCompositeGIDs->getDataNonConst(0);
 
-    for(size_t compositeNodeIdx = 0; compositeNodeIdx < numFineCompositeNodes; ++compositeNodeIdx) {
-      ArrayView<const LO> coarseRegionLID; // Should contain a single value
-      ArrayView<const SC> dummyData; // Should contain a single value
+    for (size_t compositeNodeIdx = 0; compositeNodeIdx < numFineCompositeNodes; ++compositeNodeIdx) {
+      ArrayView<const LO> coarseRegionLID;  // Should contain a single value
+      ArrayView<const SC> dummyData;        // Should contain a single value
       regProlong->getLocalRowView(compositeToRegionLIDs[compositeNodeIdx],
-                                                   coarseRegionLID,
-                                                   dummyData);
-      if(coarseRegionLID.size() == 1) {
+                                  coarseRegionLID,
+                                  dummyData);
+      if (coarseRegionLID.size() == 1) {
         coarseCompositeGIDsData[compositeNodeIdx] = regProlong->getColMap()->getGlobalElement(coarseRegionLID[0]);
       } else {
         coarseCompositeGIDsData[compositeNodeIdx] = -1;
@@ -202,42 +197,42 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
 
     // We communicate the above GIDs to their duplicate so that we can replace GIDs of the region
     // column map and form the quasiRegion column map.
-    RCP<Xpetra::Vector<MT, LO, GO, NO> > coarseQuasiregionGIDs;
-    RCP<Xpetra::Vector<MT, LO, GO, NO> > coarseRegionGIDs;
+    RCP<Xpetra::Vector<MT, LO, GO, NO>> coarseQuasiregionGIDs;
+    RCP<Xpetra::Vector<MT, LO, GO, NO>> coarseRegionGIDs;
     compositeToRegional(coarseCompositeGIDs,
                         coarseQuasiregionGIDs,
                         coarseRegionGIDs,
                         regRowMapFine,
                         regRowImportFine);
 
-    RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > regionsPerGIDWithGhosts =
-      Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(regRowMapFine,
-                                                        maxRegPerGID,
-                                                        false);
-    RCP<Xpetra::MultiVector<GO,LO,GO,NO>> interfaceGIDs = Xpetra::MultiVectorFactory<GO, LO, GO, NO>::Build(regRowMapFine,
-                                                        maxRegPerGID,
-                                                        false);
-
-    Array<ArrayRCP<const LO> > regionsPerGIDWithGhostsFine(maxRegPerGID);
-    Array<ArrayRCP<LO> > regionsPerGIDWithGhostsCoarse(maxRegPerGID);
-    Array<ArrayRCP<GO> > interfaceGIDsCoarse(maxRegPerGID);
-    for(size_t idx = 0; idx < static_cast<size_t>(maxRegPerGID); ++idx) {
-      regionsPerGIDWithGhostsFine[idx]   = levelFine->Get<RCP<Xpetra::MultiVector<LO,LO,GO,NO> > >("regionsPerGIDWithGhosts")->getData(idx);
+    RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> regionsPerGIDWithGhosts =
+        Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(regRowMapFine,
+                                                          maxRegPerGID,
+                                                          false);
+    RCP<Xpetra::MultiVector<GO, LO, GO, NO>> interfaceGIDs = Xpetra::MultiVectorFactory<GO, LO, GO, NO>::Build(regRowMapFine,
+                                                                                                               maxRegPerGID,
+                                                                                                               false);
+
+    Array<ArrayRCP<const LO>> regionsPerGIDWithGhostsFine(maxRegPerGID);
+    Array<ArrayRCP<LO>> regionsPerGIDWithGhostsCoarse(maxRegPerGID);
+    Array<ArrayRCP<GO>> interfaceGIDsCoarse(maxRegPerGID);
+    for (size_t idx = 0; idx < static_cast<size_t>(maxRegPerGID); ++idx) {
+      regionsPerGIDWithGhostsFine[idx]   = levelFine->Get<RCP<Xpetra::MultiVector<LO, LO, GO, NO>>>("regionsPerGIDWithGhosts")->getData(idx);
       regionsPerGIDWithGhostsCoarse[idx] = regionsPerGIDWithGhosts->getDataNonConst(idx);
-      interfaceGIDsCoarse[idx]          = interfaceGIDs->getDataNonConst(idx);
-      for(size_t coarseIdx = 0;
-          coarseIdx < regionsPerGIDWithGhosts->getLocalLength(); ++coarseIdx) {
+      interfaceGIDsCoarse[idx]           = interfaceGIDs->getDataNonConst(idx);
+      for (size_t coarseIdx = 0;
+           coarseIdx < regionsPerGIDWithGhosts->getLocalLength(); ++coarseIdx) {
         regionsPerGIDWithGhostsCoarse[idx][coarseIdx] = -1;
-        interfaceGIDsCoarse[idx][coarseIdx] = 0;
+        interfaceGIDsCoarse[idx][coarseIdx]           = 0;
       }
     }
 
-    for(size_t fineIdx = 0; fineIdx < numFineRegionNodes; ++fineIdx) {
-      ArrayView<const LO> coarseRegionLID; // Should contain a single value
-      ArrayView<const SC> dummyData;       // Should contain a single value
+    for (size_t fineIdx = 0; fineIdx < numFineRegionNodes; ++fineIdx) {
+      ArrayView<const LO> coarseRegionLID;  // Should contain a single value
+      ArrayView<const SC> dummyData;        // Should contain a single value
       regProlong->getLocalRowView(fineIdx,
-                                   coarseRegionLID,
-                                   dummyData);
+                                  coarseRegionLID,
+                                  dummyData);
       const LO coarseIdx = coarseRegionLID[0];
 
       // Now fill regionPerGIDWithGhostsCoarse[:][coarseRegionLID]
@@ -251,15 +246,19 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
       // For now let us assume that maxRegPerGID is constant and hope for the best.
       LO countFinePIDs   = 0;
       LO countCoarsePIDs = 0;
-      for(LO idx = 0; idx < maxRegPerGID; ++idx) {
-        if(-1 < regionsPerGIDWithGhostsFine[idx][fineIdx]) {++countFinePIDs;}
-        if(-1 < regionsPerGIDWithGhostsCoarse[idx][coarseIdx]) {++countCoarsePIDs;}
+      for (LO idx = 0; idx < maxRegPerGID; ++idx) {
+        if (-1 < regionsPerGIDWithGhostsFine[idx][fineIdx]) {
+          ++countFinePIDs;
+        }
+        if (-1 < regionsPerGIDWithGhostsCoarse[idx][coarseIdx]) {
+          ++countCoarsePIDs;
+        }
       }
 
-      if(countCoarsePIDs < countFinePIDs) {
-        for(LO idx = 0; idx < countFinePIDs; ++idx) {
+      if (countCoarsePIDs < countFinePIDs) {
+        for (LO idx = 0; idx < countFinePIDs; ++idx) {
           regionsPerGIDWithGhostsCoarse[idx][coarseIdx] = regionsPerGIDWithGhostsFine[idx][fineIdx];
-          if(regionsPerGIDWithGhostsCoarse[idx][coarseIdx] == myRank) {
+          if (regionsPerGIDWithGhostsCoarse[idx][coarseIdx] == myRank) {
             interfaceGIDsCoarse[idx][coarseIdx] = regRowMap->getGlobalElement(coarseIdx);
           }
         }
@@ -268,12 +267,12 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
 
     Array<GO> fineRegionDuplicateCoarseLIDs(numFineDuplicateNodes);
     Array<GO> fineRegionDuplicateCoarseGIDs(numFineDuplicateNodes);
-    for(size_t duplicateIdx = 0; duplicateIdx < numFineDuplicateNodes; ++duplicateIdx) {
-      ArrayView<const LO> coarseRegionLID; // Should contain a single value
-      ArrayView<const SC> dummyData; // Should contain a single value
+    for (size_t duplicateIdx = 0; duplicateIdx < numFineDuplicateNodes; ++duplicateIdx) {
+      ArrayView<const LO> coarseRegionLID;  // Should contain a single value
+      ArrayView<const SC> dummyData;        // Should contain a single value
       regProlong->getLocalRowView(fineDuplicateLIDs[duplicateIdx],
-                                                   coarseRegionLID,
-                                                   dummyData);
+                                  coarseRegionLID,
+                                  dummyData);
       fineRegionDuplicateCoarseLIDs[duplicateIdx] = regProlong->getColMap()->getGlobalElement(coarseRegionLID[0]);
       fineRegionDuplicateCoarseGIDs[duplicateIdx] = (coarseQuasiregionGIDs->getDataNonConst(0))[fineDuplicateLIDs[duplicateIdx]];
     }
@@ -283,17 +282,17 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
     coarseCompositeToRegionLIDs.resize(numCoarseRegionNodes);
     Array<GO> coarseQuasiregRowMapData = regProlong->getColMap()->getLocalElementList();
     Array<GO> coarseCompRowMapData(numCoarseRegionNodes, -1);
-    for(size_t regionIdx = 0; regionIdx < numCoarseRegionNodes; ++regionIdx) {
+    for (size_t regionIdx = 0; regionIdx < numCoarseRegionNodes; ++regionIdx) {
       const GO initialValue = coarseQuasiregRowMapData[regionIdx];
-      for(size_t duplicateIdx = 0; duplicateIdx < numFineDuplicateNodes; ++duplicateIdx) {
-        if((initialValue == fineRegionDuplicateCoarseLIDs[duplicateIdx]) &&
-           (fineRegionDuplicateCoarseGIDs[duplicateIdx] < coarseQuasiregRowMapData[regionIdx]) &&
-           (-1 < fineRegionDuplicateCoarseGIDs[duplicateIdx])){
+      for (size_t duplicateIdx = 0; duplicateIdx < numFineDuplicateNodes; ++duplicateIdx) {
+        if ((initialValue == fineRegionDuplicateCoarseLIDs[duplicateIdx]) &&
+            (fineRegionDuplicateCoarseGIDs[duplicateIdx] < coarseQuasiregRowMapData[regionIdx]) &&
+            (-1 < fineRegionDuplicateCoarseGIDs[duplicateIdx])) {
           coarseQuasiregRowMapData[regionIdx] = fineRegionDuplicateCoarseGIDs[duplicateIdx];
         }
       }
-      if(initialValue == coarseQuasiregRowMapData[regionIdx]) {
-        coarseCompRowMapData[countCoarseComposites] = coarseQuasiregRowMapData[regionIdx];
+      if (initialValue == coarseQuasiregRowMapData[regionIdx]) {
+        coarseCompRowMapData[countCoarseComposites]        = coarseQuasiregRowMapData[regionIdx];
         coarseCompositeToRegionLIDs[countCoarseComposites] = regionIdx;
         ++countCoarseComposites;
       }
@@ -302,59 +301,57 @@ void MakeCoarseLevelMaps(const int maxRegPerGID,
     coarseCompositeToRegionLIDs.resize(countCoarseComposites);
 
     // We are now ready to fill up the outputs
-    RCP<const Map> regRowMapCurrent      = regProlong->getColMap();
+    RCP<const Map> regRowMapCurrent = regProlong->getColMap();
 
-    RCP<Map> quasiRegRowMap =  MapFactory::Build(regProlong->getColMap()->lib(),
-                                                 GO_INV,
-                                                 coarseQuasiregRowMapData(),
-                                                 regProlong->getColMap()->getIndexBase(),
-                                                 regProlong->getColMap()->getComm());
+    RCP<Map> quasiRegRowMap = MapFactory::Build(regProlong->getColMap()->lib(),
+                                                GO_INV,
+                                                coarseQuasiregRowMapData(),
+                                                regProlong->getColMap()->getIndexBase(),
+                                                regProlong->getColMap()->getComm());
 
-    RCP<Map> compRowMap =  MapFactory::Build(regProlong->getColMap()->lib(),
-                                             GO_INV,
-                                             coarseCompRowMapData(),
-                                             regProlong->getColMap()->getIndexBase(),
-                                             regProlong->getColMap()->getComm());
+    RCP<Map> compRowMap = MapFactory::Build(regProlong->getColMap()->lib(),
+                                            GO_INV,
+                                            coarseCompRowMapData(),
+                                            regProlong->getColMap()->getIndexBase(),
+                                            regProlong->getColMap()->getComm());
 
     RCP<Import> regRowImportCurrent = ImportFactory::Build(compRowMap, quasiRegRowMap);
 
     // Now generate matvec data
     Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
-    Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > regionInterfaceImporter;
+    Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> regionInterfaceImporter;
     SetupMatVec(interfaceGIDs, regionsPerGIDWithGhosts,
                 regRowMapCurrent, regRowImportCurrent,
                 regionMatVecLIDs, regionInterfaceImporter);
 
     // Fill level with the outputs
-    level->Set<Teuchos::ArrayRCP<LO> >("regionMatVecLIDs",regionMatVecLIDs);
-    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("regionInterfaceImporter", regionInterfaceImporter);
-    level->Set<RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >("interfaceGIDs", interfaceGIDs);
-    level->Set<RCP<Xpetra::MultiVector<LO,LO,GO,NO> > >("regionsPerGIDWithGhosts", regionsPerGIDWithGhosts);
-    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport",regRowImportCurrent);
+    level->Set<Teuchos::ArrayRCP<LO>>("regionMatVecLIDs", regionMatVecLIDs);
+    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>>("regionInterfaceImporter", regionInterfaceImporter);
+    level->Set<RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>>("interfaceGIDs", interfaceGIDs);
+    level->Set<RCP<Xpetra::MultiVector<LO, LO, GO, NO>>>("regionsPerGIDWithGhosts", regionsPerGIDWithGhosts);
+    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>>("rowImport", regRowImportCurrent);
 
     // Finally reset compositeToRegionLIDs
     compositeToRegionLIDs = coarseCompositeToRegionLIDs();
-  } // Loop over numLevels
-} // MakeCoarseLevelMaps
-
+  }  // Loop over numLevels
+}  // MakeCoarseLevelMaps
 
 // Form the composite coarse level operator
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& compRowMap,
-                                 RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& quasiRegRowMap,
-                                 RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& quasiRegColMap,
-                                 RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& regRowImporter,
-                                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regMatrix,
-                                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& regCoarseCoordinates,
-                                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coarseCompOp,
-                                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& compCoarseCoordinates,
-                                 const bool makeCompCoords)
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& compRowMap,
+                                 RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& quasiRegRowMap,
+                                 RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& quasiRegColMap,
+                                 RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& regRowImporter,
+                                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regMatrix,
+                                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>>& regCoarseCoordinates,
+                                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& coarseCompOp,
+                                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>>& compCoarseCoordinates,
+                                 const bool makeCompCoords) {
 #include "Xpetra_UseShortNames.hpp"
   using CoordType = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
-  coarseCompOp = MatrixFactory::Build(compRowMap,
-                                       // This estimate is very conservative and probably costs us lots of memory...
-                                      8*regMatrix->getCrsGraph()->getLocalMaxNumRowEntries());
+  coarseCompOp    = MatrixFactory::Build(compRowMap,
+                                         // This estimate is very conservative and probably costs us lots of memory...
+                                         8 * regMatrix->getCrsGraph()->getLocalMaxNumRowEntries());
   regionalToComposite(regMatrix,
                       quasiRegRowMap,
                       quasiRegColMap,
@@ -366,21 +363,21 @@ void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdin
   coarseCompOp->setObjectLabel("coarse composite operator");
 
   // Create coarse composite coordinates for repartitioning
-  if(makeCompCoords) {
-    const int check       = regMatrix->getRowMap()->getLocalNumElements() % regCoarseCoordinates->getMap()->getLocalNumElements();
+  if (makeCompCoords) {
+    const int check = regMatrix->getRowMap()->getLocalNumElements() % regCoarseCoordinates->getMap()->getLocalNumElements();
     TEUCHOS_ASSERT(check == 0);
 
     RCP<const Map> compCoordMap;
-    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > regCoordImporter;
-    if(dofsPerNode == 1) {
-      compCoordMap = compRowMap;
+    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> regCoordImporter;
+    if (dofsPerNode == 1) {
+      compCoordMap     = compRowMap;
       regCoordImporter = regRowImporter;
     } else {
       using size_type = typename Teuchos::Array<GlobalOrdinal>::size_type;
       Array<GlobalOrdinal> compCoordMapData(compRowMap->getLocalNumElements() / dofsPerNode);
       ArrayView<const GlobalOrdinal> compRowMapData = compRowMap->getLocalElementList();
-      for(size_type nodeIdx = 0; nodeIdx < compCoordMapData.size(); ++nodeIdx) {
-        compCoordMapData[nodeIdx] = compRowMapData[nodeIdx*dofsPerNode] / dofsPerNode;
+      for (size_type nodeIdx = 0; nodeIdx < compCoordMapData.size(); ++nodeIdx) {
+        compCoordMapData[nodeIdx] = compRowMapData[nodeIdx * dofsPerNode] / dofsPerNode;
       }
       compCoordMap = MapFactory::Build(compRowMap->lib(),
                                        compRowMap->getGlobalNumElements() / dofsPerNode,
@@ -388,21 +385,20 @@ void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdin
                                        compRowMap->getIndexBase(),
                                        compRowMap->getComm());
 
-      RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > quasiRegCoordMap;
-        Array<GlobalOrdinal> quasiRegCoordMapData(quasiRegRowMap->getLocalNumElements() / dofsPerNode);
-        ArrayView<const GlobalOrdinal> quasiRegRowMapData = quasiRegRowMap->getLocalElementList();
-        for(size_type nodeIdx = 0; nodeIdx < quasiRegCoordMapData.size(); ++nodeIdx) {
-          quasiRegCoordMapData[nodeIdx] = quasiRegRowMapData[nodeIdx*dofsPerNode] / dofsPerNode;
-        }
-        quasiRegCoordMap = MapFactory::Build(quasiRegRowMap->lib(),
-                                             quasiRegRowMap->getGlobalNumElements() / dofsPerNode,
-                                             quasiRegCoordMapData(),
-                                             quasiRegRowMap->getIndexBase(),
-                                             quasiRegRowMap->getComm());
-        regCoordImporter = ImportFactory::Build(compCoordMap, quasiRegCoordMap);
+      RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> quasiRegCoordMap;
+      Array<GlobalOrdinal> quasiRegCoordMapData(quasiRegRowMap->getLocalNumElements() / dofsPerNode);
+      ArrayView<const GlobalOrdinal> quasiRegRowMapData = quasiRegRowMap->getLocalElementList();
+      for (size_type nodeIdx = 0; nodeIdx < quasiRegCoordMapData.size(); ++nodeIdx) {
+        quasiRegCoordMapData[nodeIdx] = quasiRegRowMapData[nodeIdx * dofsPerNode] / dofsPerNode;
+      }
+      quasiRegCoordMap = MapFactory::Build(quasiRegRowMap->lib(),
+                                           quasiRegRowMap->getGlobalNumElements() / dofsPerNode,
+                                           quasiRegCoordMapData(),
+                                           quasiRegRowMap->getIndexBase(),
+                                           quasiRegRowMap->getComm());
+      regCoordImporter = ImportFactory::Build(compCoordMap, quasiRegCoordMap);
     }
-    compCoarseCoordinates = Xpetra::MultiVectorFactory<CoordType, LocalOrdinal, GlobalOrdinal, Node>
-      ::Build(compCoordMap, regCoarseCoordinates->getNumVectors());
+    compCoarseCoordinates = Xpetra::MultiVectorFactory<CoordType, LocalOrdinal, GlobalOrdinal, Node>::Build(compCoordMap, regCoarseCoordinates->getNumVectors());
     TEUCHOS_ASSERT(Teuchos::nonnull(compCoarseCoordinates));
 
     // The following looks like regionalToComposite for Vector
@@ -415,8 +411,7 @@ void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdin
     quasiRegCoarseCoordinates->replaceMap(regCoordImporter->getTargetMap());
     compCoarseCoordinates->doExport(*quasiRegCoarseCoordinates, *(regCoordImporter), Xpetra::INSERT);
   }
-} // MakeCoarseCompositeOperator
-
+}  // MakeCoarseCompositeOperator
 
 /* Create a direct solver for a composite operator
  *
@@ -425,16 +420,15 @@ void MakeCoarseCompositeOperator(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdin
  *
  * \note For now, we're limited to Tpetra/Amesos2. From Amesos2, we use KLU as direct solver.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >
-MakeCompositeDirectSolver(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& compOp)
-{
-  using Tpetra_CrsMatrix = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>>
+MakeCompositeDirectSolver(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& compOp) {
+  using Tpetra_CrsMatrix   = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using Tpetra_MultiVector = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-  using Utilities = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Utilities          = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using Teuchos::TimeMonitor;
 
-  RCP<Amesos2::Solver<Tpetra_CrsMatrix, Tpetra_MultiVector> > coarseSolver;
+  RCP<Amesos2::Solver<Tpetra_CrsMatrix, Tpetra_MultiVector>> coarseSolver;
   {
     RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MakeCompositeDirectSolver: 1 - Setup")));
 
@@ -444,7 +438,7 @@ MakeCompositeDirectSolver(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
     // Amesos2-specific key phrase that denote smoother type
     std::string amesos2SolverName = "KLU2";
     TEUCHOS_ASSERT(Amesos2::query(amesos2SolverName));
-    coarseSolver = Amesos2::create<Tpetra_CrsMatrix,Tpetra_MultiVector>(amesos2SolverName, tMat);
+    coarseSolver = Amesos2::create<Tpetra_CrsMatrix, Tpetra_MultiVector>(amesos2SolverName, tMat);
 
     Teuchos::ParameterList amesos2_params("Amesos2");
     amesos2_params.sublist(amesos2SolverName).set("IsContiguous", false, "Are GIDs Contiguous");
@@ -459,20 +453,19 @@ MakeCompositeDirectSolver(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
   }
 
   return coarseSolver;
-} // MakeCorseCompositeDirectSolver
+}  // MakeCorseCompositeDirectSolver
 
 /* Rebalance coarse operator
  *
  */
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceCoarseCompositeOperator(const int rebalanceNumPartitions,
-              RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coarseCompOp,
-              RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& compCoarseCoordinates,
-              RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& rebalancedCompOp,
-              RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& rebalancedCoordinates,
-              RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rebalanceImporter)
-{
+                                      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& coarseCompOp,
+                                      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>>& compCoarseCoordinates,
+                                      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& rebalancedCompOp,
+                                      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>>& rebalancedCoordinates,
+                                      RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& rebalanceImporter) {
 #include "MueLu_UseShortNames.hpp"
   using CoordType = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
   using Teuchos::TimeMonitor;
@@ -496,21 +489,21 @@ void RebalanceCoarseCompositeOperator(const int rebalanceNumPartitions,
 
   RCP<Zoltan2Interface> zoltan = rcp(new Zoltan2Interface());
 
-  level.Set<RCP<Matrix> >     ("A",                    coarseCompOp);
-  level.Set<RCP<MultiVector> >("Coordinates",          compCoarseCoordinates);
-//  int numPartitions = Get<int>(level, "number of partitions");
+  level.Set<RCP<Matrix>>("A", coarseCompOp);
+  level.Set<RCP<MultiVector>>("Coordinates", compCoarseCoordinates);
+  //  int numPartitions = Get<int>(level, "number of partitions");
 
   RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
   Teuchos::ParameterList paramList;
-  paramList.set("repartition: remap parts",       false);
-  if( numPartitions > 0 ){ // If number of coarse rebalance partitions was provided by the user.
-    level.Set<int>              ("number of partitions", numPartitions);
+  paramList.set("repartition: remap parts", false);
+  if (numPartitions > 0) {  // If number of coarse rebalance partitions was provided by the user.
+    level.Set<int>("number of partitions", numPartitions);
   } else {
     Teuchos::ParameterList paramListHeuristic;
-    paramListHeuristic.set("repartition: start level",       1);
+    paramListHeuristic.set("repartition: start level", 1);
     RCP<RepartitionHeuristicFactory> repartHeuristic = rcp(new RepartitionHeuristicFactory());
     repartHeuristic->SetParameterList(paramListHeuristic);
-    repart->SetFactory("number of partitions", repartHeuristic );
+    repart->SetFactory("number of partitions", repartHeuristic);
   }
   repart->SetParameterList(paramList);
   repart->SetFactory("Partition", zoltan);
@@ -526,13 +519,13 @@ void RebalanceCoarseCompositeOperator(const int rebalanceNumPartitions,
 
   ParameterList XpetraList;
   XpetraList.set("Restrict Communicator", true);
-  XpetraList.set("Timer Label","MueLu::RebalanceAc-for-coarseAMG");
+  XpetraList.set("Timer Label", "MueLu::RebalanceAc-for-coarseAMG");
 
   // Build rebalanced coarse composite operator
-  rebalancedCompOp = MatrixFactory::Build(coarseCompOp, *rebalanceImporter, *rebalanceImporter, rebalanceImporter->getTargetMap(), rebalanceImporter->getTargetMap(), rcp(&XpetraList,false));
-        if (!rebalancedCompOp.is_null()) {
-          rebalancedCompOp->SetFixedBlockSize(coarseCompOp->GetFixedBlockSize());
-        }
+  rebalancedCompOp = MatrixFactory::Build(coarseCompOp, *rebalanceImporter, *rebalanceImporter, rebalanceImporter->getTargetMap(), rebalanceImporter->getTargetMap(), rcp(&XpetraList, false));
+  if (!rebalancedCompOp.is_null()) {
+    rebalancedCompOp->SetFixedBlockSize(coarseCompOp->GetFixedBlockSize());
+  }
 
   // Build rebalanced coarse coordinates (The following code is borrowed from MueLu_RebalanceTransferFactory_def.hpp)
   LO blkSize = coarseCompOp->GetFixedBlockSize();
@@ -544,24 +537,24 @@ void RebalanceCoarseCompositeOperator(const int rebalanceNumPartitions,
     // NOTE: there is an implicit assumption here: we assume that dof any node are enumerated consequently
     // Proper fix would require using decomposition similar to how we construct importer in the
     // RepartitionFactory
-    RCP<const Map> origMap   = compCoarseCoordinates->getMap();
-    GO             indexBase = origMap->getIndexBase();
+    RCP<const Map> origMap = compCoarseCoordinates->getMap();
+    GO indexBase           = origMap->getIndexBase();
 
-    ArrayView<const GO> OEntries   = rebalanceImporter->getTargetMap()->getLocalElementList();
-    LO                  numEntries = OEntries.size()/blkSize;
+    ArrayView<const GO> OEntries = rebalanceImporter->getTargetMap()->getLocalElementList();
+    LO numEntries                = OEntries.size() / blkSize;
     ArrayRCP<GO> Entries(numEntries);
     for (LO i = 0; i < numEntries; i++)
-      Entries[i] = (OEntries[i*blkSize]-indexBase)/blkSize + indexBase;
+      Entries[i] = (OEntries[i * blkSize] - indexBase) / blkSize + indexBase;
 
     RCP<const Map> targetMap = MapFactory::Build(origMap->lib(), origMap->getGlobalNumElements(), Entries(), indexBase, origMap->getComm());
-    coordImporter = ImportFactory::Build(origMap, targetMap);
+    coordImporter            = ImportFactory::Build(origMap, targetMap);
   }
-  rebalancedCoordinates = Xpetra::MultiVectorFactory<CoordType,LocalOrdinal,GlobalOrdinal,Node>::Build(coordImporter->getTargetMap(), compCoarseCoordinates->getNumVectors());
+  rebalancedCoordinates = Xpetra::MultiVectorFactory<CoordType, LocalOrdinal, GlobalOrdinal, Node>::Build(coordImporter->getTargetMap(), compCoarseCoordinates->getNumVectors());
   rebalancedCoordinates->doImport(*compCoarseCoordinates, *coordImporter, Xpetra::INSERT);
   rebalancedCoordinates->replaceMap(rebalancedCoordinates->getMap()->removeEmptyProcesses());
 
   return;
-} // RebalanceCoarseCompositeOperator
+}  // RebalanceCoarseCompositeOperator
 #endif
 
 /* Create an AMG hierarchy for a composite operator
@@ -569,12 +562,11 @@ void RebalanceCoarseCompositeOperator(const int rebalanceNumPartitions,
  * Create the hierarchy object and perform the multigrid setup.
  * Finally, the hierarhcy object will be ready to be applied during the region MG V-cycle call.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& compOp,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& compOp,
                           const std::string& xmlFileName,
-                          RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > coordinates)
-{
+                          RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>> coordinates) {
 #include "MueLu_UseShortNames.hpp"
   using coordinates_type = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
 
@@ -587,24 +579,23 @@ MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
   // Get parameter list for AMG hierarchy
   RCP<ParameterList> mueluParams = Teuchos::rcp(new ParameterList());
   Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, mueluParams.ptr(),
-      *compOp->getRowMap()->getComm());
-
+                                                   *compOp->getRowMap()->getComm());
 
   // Get the user data sublist
-  const std::string userName = "user data";
+  const std::string userName            = "user data";
   Teuchos::ParameterList& userParamList = mueluParams->sublist(userName);
 
   // Add nullspace information
   {
     // Compute nullspace
     RCP<MultiVector> nullspace;
-    if((compOp->GetFixedBlockSize() == 1) || Teuchos::is_null(coordinates)) { // Scalar problem, constant nullspace
+    if ((compOp->GetFixedBlockSize() == 1) || Teuchos::is_null(coordinates)) {  // Scalar problem, constant nullspace
       nullspace = MultiVectorFactory::Build(compOp->getRowMap(), 1);
       nullspace->putScalar(one);
-    } else if(compOp->GetFixedBlockSize() == 2) { // 2D Elasticity
+    } else if (compOp->GetFixedBlockSize() == 2) {  // 2D Elasticity
       nullspace = MultiVectorFactory::Build(compOp->getRowMap(), 3);
-      Array<ArrayRCP<SC> > nullspaceData(3);
-      Array<ArrayRCP<const coordinates_type> > coordinateData(2);
+      Array<ArrayRCP<SC>> nullspaceData(3);
+      Array<ArrayRCP<const coordinates_type>> coordinateData(2);
 
       // Calculate center
       const coordinates_type cx = coordinates->getVector(0)->meanValue();
@@ -613,24 +604,24 @@ MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
       coordinateData[0] = coordinates->getData(0);
       coordinateData[1] = coordinates->getData(1);
 
-      for(int vecIdx = 0; vecIdx < 3; ++vecIdx) {
+      for (int vecIdx = 0; vecIdx < 3; ++vecIdx) {
         nullspaceData[vecIdx] = nullspace->getDataNonConst(vecIdx);
       }
 
-      for(size_t nodeIdx = 0; nodeIdx < coordinates->getLocalLength(); ++nodeIdx) {
+      for (size_t nodeIdx = 0; nodeIdx < coordinates->getLocalLength(); ++nodeIdx) {
         // translations
-        nullspaceData[0][2*nodeIdx + 0] = one;
-        nullspaceData[1][2*nodeIdx + 1] = one;
+        nullspaceData[0][2 * nodeIdx + 0] = one;
+        nullspaceData[1][2 * nodeIdx + 1] = one;
 
         // rotation about z axis
-        nullspaceData[2][2*nodeIdx + 0] = -(coordinateData[1][nodeIdx] - cy);
-        nullspaceData[2][2*nodeIdx + 1] =  (coordinateData[0][nodeIdx] - cx);
+        nullspaceData[2][2 * nodeIdx + 0] = -(coordinateData[1][nodeIdx] - cy);
+        nullspaceData[2][2 * nodeIdx + 1] = (coordinateData[0][nodeIdx] - cx);
       }
 
-    } else if(compOp->GetFixedBlockSize() == 3) { // 3D Elasticity
+    } else if (compOp->GetFixedBlockSize() == 3) {  // 3D Elasticity
       nullspace = MultiVectorFactory::Build(compOp->getRowMap(), 6);
-      Array<ArrayRCP<SC> > nullspaceData(6);
-      Array<ArrayRCP<const coordinates_type> > coordinateData(3);
+      Array<ArrayRCP<SC>> nullspaceData(6);
+      Array<ArrayRCP<const coordinates_type>> coordinateData(3);
 
       // Calculate center
       const coordinates_type cx = coordinates->getVector(0)->meanValue();
@@ -641,27 +632,27 @@ MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
       coordinateData[1] = coordinates->getData(1);
       coordinateData[2] = coordinates->getData(2);
 
-      for(int vecIdx = 0; vecIdx < 6; ++vecIdx) {
+      for (int vecIdx = 0; vecIdx < 6; ++vecIdx) {
         nullspaceData[vecIdx] = nullspace->getDataNonConst(vecIdx);
       }
 
-      for(size_t nodeIdx = 0; nodeIdx < coordinates->getLocalLength(); ++nodeIdx) {
+      for (size_t nodeIdx = 0; nodeIdx < coordinates->getLocalLength(); ++nodeIdx) {
         // translations
-        nullspaceData[0][3*nodeIdx + 0] = one;
-        nullspaceData[1][3*nodeIdx + 1] = one;
-        nullspaceData[2][3*nodeIdx + 2] = one;
+        nullspaceData[0][3 * nodeIdx + 0] = one;
+        nullspaceData[1][3 * nodeIdx + 1] = one;
+        nullspaceData[2][3 * nodeIdx + 2] = one;
 
         // rotation about z axis
-        nullspaceData[3][3*nodeIdx + 0] = -(coordinateData[1][nodeIdx] - cy);
-        nullspaceData[3][3*nodeIdx + 1] =  (coordinateData[0][nodeIdx] - cx);
+        nullspaceData[3][3 * nodeIdx + 0] = -(coordinateData[1][nodeIdx] - cy);
+        nullspaceData[3][3 * nodeIdx + 1] = (coordinateData[0][nodeIdx] - cx);
 
         // rotation about x axis
-        nullspaceData[4][3*nodeIdx + 1] = -(coordinateData[2][nodeIdx] - cz);
-        nullspaceData[4][3*nodeIdx + 2] =  (coordinateData[1][nodeIdx] - cy);
+        nullspaceData[4][3 * nodeIdx + 1] = -(coordinateData[2][nodeIdx] - cz);
+        nullspaceData[4][3 * nodeIdx + 2] = (coordinateData[1][nodeIdx] - cy);
 
         // rotation about y axis
-        nullspaceData[5][3*nodeIdx + 0] =  (coordinateData[2][nodeIdx] - cz);
-        nullspaceData[5][3*nodeIdx + 2] = -(coordinateData[0][nodeIdx] - cx);
+        nullspaceData[5][3 * nodeIdx + 0] = (coordinateData[2][nodeIdx] - cz);
+        nullspaceData[5][3 * nodeIdx + 2] = -(coordinateData[0][nodeIdx] - cx);
       }
     }
 
@@ -681,7 +672,7 @@ MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
 
   // Add coordinate information for rebalancing
   {
-    if(Teuchos::nonnull(coordinates)) {
+    if (Teuchos::nonnull(coordinates)) {
       userParamList.set("Coordinates", coordinates);
     }
   }
@@ -694,68 +685,64 @@ MakeCompositeAMGHierarchy(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal
   compOpHiearchy->SetVerbLevel(MueLu::VERB_NONE);
 
   return compOpHiearchy;
-} // MakeCompositeAMGHierarchy
-
+}  // MakeCompositeAMGHierarchy
 
-  // Make interface scaling factors recursively
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// Make interface scaling factors recursively
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void MakeInterfaceScalingFactors(const int numLevels,
-                                 Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regHierarchy)
-{
+                                 Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regHierarchy) {
 #include "Xpetra_UseShortNames.hpp"
 #include "MueLu_UseShortNames.hpp"
 
   const SC SC_ONE = Teuchos::ScalarTraits<SC>::one();
 
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(numLevels>0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(numLevels > 0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
 
   for (int l = 0; l < numLevels; l++) {
-    RCP<Level> level        = regHierarchy->GetLevel(l);
-    RCP<Matrix> regMat      = level->Get<RCP<Matrix> >("A");
-    RCP<const Map> regRowMap      = regMat->getRowMap();
-    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > regRowImporters = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport");
+    RCP<Level> level                                                       = regHierarchy->GetLevel(l);
+    RCP<Matrix> regMat                                                     = level->Get<RCP<Matrix>>("A");
+    RCP<const Map> regRowMap                                               = regMat->getRowMap();
+    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> regRowImporters = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>>("rowImport");
     // initialize region vector with all ones.
-    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regInterfaceScalings = VectorFactory::Build(regRowMap);
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regInterfaceScalings = VectorFactory::Build(regRowMap);
     regInterfaceScalings->putScalar(SC_ONE);
 
     // transform to composite layout while adding interface values via the Export() combine mode
-    RCP<Vector> compInterfaceScalingSum = VectorFactory::Build( regRowImporters->getSourceMap() , true);
+    RCP<Vector> compInterfaceScalingSum = VectorFactory::Build(regRowImporters->getSourceMap(), true);
     regionalToComposite(regInterfaceScalings, compInterfaceScalingSum, regRowImporters);
 
     /* transform composite layout back to regional layout. Now, GIDs associated
      * with region interface should carry a scaling factor (!= 1).
      */
-    RCP<Vector> quasiRegInterfaceScaling; // Is that vector really needed?
+    RCP<Vector> quasiRegInterfaceScaling;  // Is that vector really needed?
     compositeToRegional(compInterfaceScalingSum, quasiRegInterfaceScaling,
                         regInterfaceScalings,
                         regRowMap, regRowImporters);
 
-    level->Set<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("regInterfaceScalings", regInterfaceScalings);
+    level->Set<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>>("regInterfaceScalings", regInterfaceScalings);
   }
-} // MakeInterfaceScalingFactors
+}  // MakeInterfaceScalingFactors
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void createRegionHierarchy(const int numDimensions,
                            const Array<int> lNodesPerDim,
                            const std::string aggregationRegionType,
                            RCP<Teuchos::ParameterList>& interfaceParams,
                            const int maxRegPerGID,
                            RCP<Teuchos::ParameterList>& coarseSolverData,
-                           Array<RCP<Teuchos::ParameterList> >& smootherParams,
+                           Array<RCP<Teuchos::ParameterList>>& smootherParams,
                            RCP<Teuchos::ParameterList> hierarchyData,
-                           RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-                           const bool keepCoarseCoords)
-{
+                           RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+                           const bool keepCoarseCoords) {
 #include "Xpetra_UseShortNames.hpp"
 #include "MueLu_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   // This monitor times everything and gets the overall setting cost
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("createRegionHierarchy")));
 
-  using Hierarchy = MueLu::Hierarchy<SC, LO, GO, NO>;
-  using Utilities = MueLu::Utilities<SC, LO, GO, NO>;
-  using DirectCoarseSolver = Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >;
+  using Hierarchy          = MueLu::Hierarchy<SC, LO, GO, NO>;
+  using Utilities          = MueLu::Utilities<SC, LO, GO, NO>;
+  using DirectCoarseSolver = Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>;
 
   // std::cout << mapComp->getComm()->getRank() << " | Setting up MueLu hierarchies ..." << std::endl;
   int numLevels = 0;
@@ -770,37 +757,39 @@ void createRegionHierarchy(const int numDimensions,
     // resize group containers on each level
     for (int l = 0; l < numLevels; ++l) {
       // Also doing some initialization in the smootherParams
-      if(l > 0) {smootherParams[l] = rcp(new Teuchos::ParameterList(*smootherParams[0]));}
+      if (l > 0) {
+        smootherParams[l] = rcp(new Teuchos::ParameterList(*smootherParams[0]));
+      }
     }
   }
 
-  RCP<Level> level0 = regHierarchy->GetLevel(0);
-  RCP<Matrix> regMat       = level0->Get<RCP<Matrix> >("A");
-  RCP<const Map> revisedRowMap  = regMat->getRowMap();
+  RCP<Level> level0            = regHierarchy->GetLevel(0);
+  RCP<Matrix> regMat           = level0->Get<RCP<Matrix>>("A");
+  RCP<const Map> revisedRowMap = regMat->getRowMap();
   /* Get coarse level matrices and prolongators from MueLu hierarchy
    * Note: fine level has been dealt with previously, so we start at level 1 here.
    */
-  using real_type = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
+  using real_type                  = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
   using realvaluedmultivector_type = Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, Node>;
   RCP<realvaluedmultivector_type> regCoarseCoordinates;
-  for (int l = 1; l < numLevels; ++l) { // Note: we start at level 1 (which is the first coarse level)
+  for (int l = 1; l < numLevels; ++l) {  // Note: we start at level 1 (which is the first coarse level)
     RCP<Level> level = regHierarchy->GetLevel(l);
 
-    if(keepCoarseCoords && (l == numLevels - 1)) {
-      regCoarseCoordinates = level->Get<RCP<realvaluedmultivector_type> >("Coordinates2", MueLu::NoFactory::get());
+    if (keepCoarseCoords && (l == numLevels - 1)) {
+      regCoarseCoordinates = level->Get<RCP<realvaluedmultivector_type>>("Coordinates2", MueLu::NoFactory::get());
     }
 
-    RCP<Matrix> regMatrices = level->Get<RCP<Matrix> >("A", MueLu::NoFactory::get());
+    RCP<Matrix> regMatrices = level->Get<RCP<Matrix>>("A", MueLu::NoFactory::get());
 
     // Create residual and solution vectors and cache them for vCycle apply
     std::string levelName("level");
     levelName += std::to_string(l);
     ParameterList& levelList = hierarchyData->sublist(levelName, false, "list of data on current level");
-    RCP<Vector> regRes = VectorFactory::Build(revisedRowMap, true);
-    RCP<Vector> regSol = VectorFactory::Build(revisedRowMap, true);
+    RCP<Vector> regRes       = VectorFactory::Build(revisedRowMap, true);
+    RCP<Vector> regSol       = VectorFactory::Build(revisedRowMap, true);
 
-    levelList.set<RCP<Vector> >("residual", regRes, "Cached residual vector");
-    levelList.set<RCP<Vector> >("solution", regSol, "Cached solution vector");
+    levelList.set<RCP<Vector>>("residual", regRes, "Cached residual vector");
+    levelList.set<RCP<Vector>>("solution", regSol, "Cached solution vector");
   }
 
   // std::cout << mapComp->getComm()->getRank() << " | MakeCoarseLevelMaps ..." << std::endl;
@@ -824,12 +813,12 @@ void createRegionHierarchy(const int numDimensions,
   // Only set the smoother up to the last but one level
   // if we want to use a smoother on the coarse level
   // we will handle that separately with "coarse solver type"
-  for(int levelIdx = 0; levelIdx < numLevels - 1; ++levelIdx) {
-    RCP<Level> level = regHierarchy->GetLevel(levelIdx);
-    RCP<Matrix> regMatrix = level->Get<RCP<Matrix> >("A", MueLu::NoFactory::get());
-    RCP<const Map>    regRowMap = regMatrix->getRowMap();
-    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > regRowImporter = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport");
-    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regInterfaceScalings = level->Get<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("regInterfaceScalings");
+  for (int levelIdx = 0; levelIdx < numLevels - 1; ++levelIdx) {
+    RCP<Level> level                                                                    = regHierarchy->GetLevel(levelIdx);
+    RCP<Matrix> regMatrix                                                               = level->Get<RCP<Matrix>>("A", MueLu::NoFactory::get());
+    RCP<const Map> regRowMap                                                            = regMatrix->getRowMap();
+    RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> regRowImporter               = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>>("rowImport");
+    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regInterfaceScalings = level->Get<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>>("regInterfaceScalings");
 
     smootherParams[levelIdx]->set("smoother: level", levelIdx);
     smootherSetup(smootherParams[levelIdx], regRowMap,
@@ -839,40 +828,39 @@ void createRegionHierarchy(const int numDimensions,
 
   // std::cout << mapComp->getComm()->getRank() << " | CreateCoarseSolver ..." << std::endl;
 
-  tmLocal = Teuchos::null;
-  tmLocal = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("createRegionHierarchy: CreateCoarseSolver")));
+  tmLocal                            = Teuchos::null;
+  tmLocal                            = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("createRegionHierarchy: CreateCoarseSolver")));
   const std::string coarseSolverType = coarseSolverData->get<std::string>("coarse solver type");
   if (coarseSolverType == "smoother") {
-    RCP<Level> level = regHierarchy->GetLevel(numLevels - 1);
-    RCP<Matrix> regMatrix = level->Get<RCP<Matrix> >("A", MueLu::NoFactory::get());
-    RCP<const Map> regRowMap = regMatrix->getRowMap();
-    RCP<Import> regRowImporter       = level->Get<RCP<Import> >("rowImport");
-    RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector> >("regInterfaceScalings");
+    RCP<Level> level                 = regHierarchy->GetLevel(numLevels - 1);
+    RCP<Matrix> regMatrix            = level->Get<RCP<Matrix>>("A", MueLu::NoFactory::get());
+    RCP<const Map> regRowMap         = regMatrix->getRowMap();
+    RCP<Import> regRowImporter       = level->Get<RCP<Import>>("rowImport");
+    RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector>>("regInterfaceScalings");
 
     // Set the smoother on the coarsest level.
-    const std::string smootherXMLFileName = coarseSolverData->get<std::string>("smoother xml file");
+    const std::string smootherXMLFileName   = coarseSolverData->get<std::string>("smoother xml file");
     RCP<ParameterList> coarseSmootherParams = smootherParams[numLevels - 1];
     Teuchos::updateParametersFromXmlFileAndBroadcast(smootherXMLFileName, coarseSmootherParams.ptr(), *level->GetComm());
     coarseSmootherParams->set("smoother: level", numLevels - 1);
     coarseSmootherParams->print();
 
-
     smootherSetup(smootherParams[numLevels - 1], regRowMap,
                   regMatrix, regInterfaceScalings,
                   regRowImporter);
-  } else if( (coarseSolverType == "direct") || (coarseSolverType == "amg") ) {
+  } else if ((coarseSolverType == "direct") || (coarseSolverType == "amg")) {
     // A composite coarse matrix is needed
 
     // std::cout << mapComp->getComm()->getRank() << " | MakeCoarseCompositeOperator ..." << std::endl;
 
-    RCP<Level> level = regHierarchy->GetLevel(numLevels - 1);
-    RCP<Matrix>  regMatrix      = level->Get<RCP<Matrix> >("A", MueLu::NoFactory::get());
-    RCP<Import> regRowImporter  = level->Get<RCP<Import> >("rowImport");
-    RCP<const Map> compRowMap         = regRowImporter->getSourceMap();
-    RCP<const Map> quasiRegRowMap     = regRowImporter->getTargetMap();
-    RCP<const Map> quasiRegColMap     = regRowImporter->getTargetMap();// col map same as row map.
+    RCP<Level> level              = regHierarchy->GetLevel(numLevels - 1);
+    RCP<Matrix> regMatrix         = level->Get<RCP<Matrix>>("A", MueLu::NoFactory::get());
+    RCP<Import> regRowImporter    = level->Get<RCP<Import>>("rowImport");
+    RCP<const Map> compRowMap     = regRowImporter->getSourceMap();
+    RCP<const Map> quasiRegRowMap = regRowImporter->getTargetMap();
+    RCP<const Map> quasiRegColMap = regRowImporter->getTargetMap();  // col map same as row map.
 
-    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > coarseCompOp;
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> coarseCompOp;
     RCP<realvaluedmultivector_type> compCoarseCoordinates;
     MakeCoarseCompositeOperator(compRowMap,
                                 quasiRegRowMap,
@@ -884,14 +872,14 @@ void createRegionHierarchy(const int numDimensions,
                                 compCoarseCoordinates,
                                 keepCoarseCoords);
 
-    coarseSolverData->set<RCP<const Map> >("compCoarseRowMap", coarseCompOp->getRowMap());
+    coarseSolverData->set<RCP<const Map>>("compCoarseRowMap", coarseCompOp->getRowMap());
 
     // std::cout << mapComp->getComm()->getRank() << " | MakeCoarseCompositeSolver ..." << std::endl;
     if (coarseSolverType == "direct") {
       RCP<DirectCoarseSolver> coarseDirectSolver = MakeCompositeDirectSolver(coarseCompOp);
-      coarseSolverData->set<RCP<DirectCoarseSolver> >("direct solver object", coarseDirectSolver);
+      coarseSolverData->set<RCP<DirectCoarseSolver>>("direct solver object", coarseDirectSolver);
     } else if (coarseSolverType == "amg") {
-      if(keepCoarseCoords == false) {
+      if (keepCoarseCoords == false) {
         std::cout << "WARNING: you requested a coarse AMG solver but you did not request coarse coordinates to be kept, repartitioning is not possible!" << std::endl;
       }
 
@@ -899,21 +887,21 @@ void createRegionHierarchy(const int numDimensions,
       std::string amgXmlFileName = coarseSolverData->get<std::string>("amg xml file");
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
       const bool coarseSolverRebalance = coarseSolverData->get<bool>("coarse solver rebalance");
-      if(keepCoarseCoords == true && coarseSolverRebalance == true ){
-        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rebalancedCompOp;
-        RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > rebalancedCoordinates;
+      if (keepCoarseCoords == true && coarseSolverRebalance == true) {
+        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> rebalancedCompOp;
+        RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>> rebalancedCoordinates;
         RCP<const Import> rebalanceImporter;
 
         const int rebalanceNumPartitions = coarseSolverData->get<int>("coarse rebalance num partitions");
         RebalanceCoarseCompositeOperator(rebalanceNumPartitions,
-                                    coarseCompOp,
-                                    compCoarseCoordinates,
-                                    rebalancedCompOp,
-                                    rebalancedCoordinates,
-                                    rebalanceImporter);
-        coarseSolverData->set<RCP<const Import> >("rebalanceImporter", rebalanceImporter);
-
-        if( !rebalancedCompOp.is_null() )
+                                         coarseCompOp,
+                                         compCoarseCoordinates,
+                                         rebalancedCompOp,
+                                         rebalancedCoordinates,
+                                         rebalanceImporter);
+        coarseSolverData->set<RCP<const Import>>("rebalanceImporter", rebalanceImporter);
+
+        if (!rebalancedCompOp.is_null())
           coarseAMGHierarchy = MakeCompositeAMGHierarchy(rebalancedCompOp, amgXmlFileName, rebalancedCoordinates);
 
       } else {
@@ -922,12 +910,12 @@ void createRegionHierarchy(const int numDimensions,
 #else
       coarseAMGHierarchy = MakeCompositeAMGHierarchy(coarseCompOp, amgXmlFileName, compCoarseCoordinates);
 #endif
-      coarseSolverData->set<RCP<Hierarchy> >("amg hierarchy object", coarseAMGHierarchy);
+      coarseSolverData->set<RCP<Hierarchy>>("amg hierarchy object", coarseAMGHierarchy);
     }
-  } else  {
+  } else {
     TEUCHOS_TEST_FOR_EXCEPT_MSG(false, "Unknown coarse solver type.");
   }
 
-} // createRegionHierarchy
+}  // createRegionHierarchy
 
-#endif // MUELU_SETUPREGIONHIERARCHY_DEF_HPP
+#endif  // MUELU_SETUPREGIONHIERARCHY_DEF_HPP
diff --git a/packages/muelu/research/regionMG/src/SetupRegionMatrix_def.hpp b/packages/muelu/research/regionMG/src/SetupRegionMatrix_def.hpp
index 73650a565d6f..587607edc716 100644
--- a/packages/muelu/research/regionMG/src/SetupRegionMatrix_def.hpp
+++ b/packages/muelu/research/regionMG/src/SetupRegionMatrix_def.hpp
@@ -48,8 +48,8 @@
 #include <vector>
 #include <iostream>
 
-#define RegionsSpanProcs  1
-#define MultipleRegionsPerProc  2
+#define RegionsSpanProcs 1
+#define MultipleRegionsPerProc 2
 
 #include <Tpetra_KokkosCompat_DefaultNode.hpp>
 #include <KokkosSparse_spmv.hpp>
@@ -62,20 +62,18 @@
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_CrsMatrixWrap.hpp>
 
-using Teuchos::RCP;
-using Teuchos::ArrayRCP;
 using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::RCP;
 
 /*! \brief Find common regions of two nodes
  *
  */
-template<class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::Array<int> findCommonRegions(const GlobalOrdinal nodeA, ///< GID of first node
-                                      const GlobalOrdinal nodeB, ///< GID of second node
-                                      const Array<ArrayRCP<const LocalOrdinal> > nodesToRegions, ///< mapping of nodes to regions
-                                      RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > nodesToRegionsMap
-                                      )
-{
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::Array<int> findCommonRegions(const GlobalOrdinal nodeA,                                 ///< GID of first node
+                                      const GlobalOrdinal nodeB,                                 ///< GID of second node
+                                      const Array<ArrayRCP<const LocalOrdinal>> nodesToRegions,  ///< mapping of nodes to regions
+                                      RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> nodesToRegionsMap) {
 #include "Xpetra_UseShortNamesOrdinal.hpp"
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("findCommonRegions: 1 - Extract regions")));
@@ -103,7 +101,7 @@ Teuchos::Array<int> findCommonRegions(const GlobalOrdinal nodeA, ///< GID of fir
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("findCommonRegions: 3 - Find commons")));
 
   std::vector<int>::iterator it = std::set_intersection(regionsA.begin(),
-      regionsA.end(), regionsB.begin(), regionsB.end(), commonRegions.begin());
+                                                        regionsA.end(), regionsB.begin(), regionsB.end(), commonRegions.begin());
   commonRegions.resize(it - commonRegions.begin());
 
   tm = Teuchos::null;
@@ -121,21 +119,20 @@ Teuchos::Array<int> findCommonRegions(const GlobalOrdinal nodeA, ///< GID of fir
   return finalCommonRegions;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MakeQuasiregionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AComp,
-                             RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > regionsPerGIDWithGhosts,
-                             RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap,
-                             RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > colMap,
-                             RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
-                             RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegionMats,
+void MakeQuasiregionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AComp,
+                             RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> regionsPerGIDWithGhosts,
+                             RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap,
+                             RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap,
+                             RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& rowImport,
+                             RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& quasiRegionMats,
                              const Teuchos::ArrayRCP<LocalOrdinal>& regionMatVecLIDs) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::RCP;
   using Teuchos::TimeMonitor;
 
-  Array<ArrayRCP<const LO> > regionPerGIDWithGhostsData(regionsPerGIDWithGhosts->getNumVectors());
-  for(size_t vecIdx = 0; vecIdx < regionsPerGIDWithGhosts->getNumVectors(); ++vecIdx) {
+  Array<ArrayRCP<const LO>> regionPerGIDWithGhostsData(regionsPerGIDWithGhosts->getNumVectors());
+  for (size_t vecIdx = 0; vecIdx < regionsPerGIDWithGhosts->getNumVectors(); ++vecIdx) {
     regionPerGIDWithGhostsData[vecIdx] = regionsPerGIDWithGhosts->getData(vecIdx);
   }
 
@@ -166,12 +163,12 @@ void MakeQuasiregionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdina
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MakeQuasiregionMatrices: 3 - Scale interface entries")));
 
   RCP<CrsMatrixWrap> quasiRegionCrsWrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(quasiRegionMats);
-  RCP<CrsMatrix> quasiRegionCrs = quasiRegionCrsWrap->getCrsMatrix();
+  RCP<CrsMatrix> quasiRegionCrs         = quasiRegionCrsWrap->getCrsMatrix();
 
   // Grab first and last element of sorted interface LIDs
   Array<LO> interfaceLIDs(regionMatVecLIDs());
   std::sort(interfaceLIDs.begin(), interfaceLIDs.end());
-  auto vecEnd = std::unique(interfaceLIDs.begin(), interfaceLIDs.end());
+  auto vecEnd   = std::unique(interfaceLIDs.begin(), interfaceLIDs.end());
   auto vecStart = interfaceLIDs.begin();
 
   GO rowGID;
@@ -179,18 +176,18 @@ void MakeQuasiregionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdina
   GlobalOrdinal colGID;
   std::size_t sizeOfCommonRegions;
   std::size_t numEntries = 0;
-  for(auto row = vecStart; row < vecEnd; ++row) {
-    rowGID = rowMap->getGlobalElement(*row);
-    numEntries = quasiRegionMats->getNumEntriesInLocalRow(*row); // number of entries in this row
-    Array<SC> values(numEntries); // non-zeros in this row
-    Array<LO> colInds(numEntries); // local column indices
+  for (auto row = vecStart; row < vecEnd; ++row) {
+    rowGID     = rowMap->getGlobalElement(*row);
+    numEntries = quasiRegionMats->getNumEntriesInLocalRow(*row);  // number of entries in this row
+    Array<SC> values(numEntries);                                 // non-zeros in this row
+    Array<LO> colInds(numEntries);                                // local column indices
     quasiRegionMats->getLocalRowCopy(*row, colInds, values, numEntries);
 
-    for (std::size_t entryIdx = 0; entryIdx < numEntries; ++entryIdx) { // loop over all entries in this row
-      col = colInds[entryIdx];
+    for (std::size_t entryIdx = 0; entryIdx < numEntries; ++entryIdx) {  // loop over all entries in this row
+      col    = colInds[entryIdx];
       colGID = colMap->getGlobalElement(col);
       Array<int> commonRegions;
-      if (rowGID != colGID) { // Skip the diagonal entry. It will be processed later.
+      if (rowGID != colGID) {  // Skip the diagonal entry. It will be processed later.
         commonRegions = findCommonRegions(rowGID, colGID, regionPerGIDWithGhostsData, regionsPerGIDWithGhosts->getMap());
       }
 
@@ -208,18 +205,17 @@ void MakeQuasiregionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdina
   quasiRegionMats->fillComplete();
 
   tm = Teuchos::null;
-} // MakeQuasiregionMatrices
-
+}  // MakeQuasiregionMatrices
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AComp,
-                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > mapComp,
-                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap,
-                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
-                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedColMap,
-                        RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
-                        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegionMats,
-                        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regionMats) {
+void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AComp,
+                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> mapComp,
+                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap,
+                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> revisedRowMap,
+                        RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> revisedColMap,
+                        RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& rowImport,
+                        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& quasiRegionMats,
+                        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regionMats) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::RCP;
   using Teuchos::TimeMonitor;
@@ -230,40 +226,40 @@ void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, Gl
 
   // Copy data from quasiRegionMats, but into new map layout
   {
-      regionMats = rcp(new CrsMatrixWrap(revisedRowMap, revisedColMap, 9));
+    regionMats = rcp(new CrsMatrixWrap(revisedRowMap, revisedColMap, 9));
 
-      // Extract current region CrsMatrix
-      RCP<CrsMatrix> regionCrsMat = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(regionMats)->getCrsMatrix();
+    // Extract current region CrsMatrix
+    RCP<CrsMatrix> regionCrsMat = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(regionMats)->getCrsMatrix();
 
-      // Extract current quasi-region CrsMatrix
-      RCP<CrsMatrix> quasiRegionCrsMat = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(quasiRegionMats)->getCrsMatrix();
+    // Extract current quasi-region CrsMatrix
+    RCP<CrsMatrix> quasiRegionCrsMat = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(quasiRegionMats)->getCrsMatrix();
 
-      // Pull out the data from the quasi-region CrsMatrix
-      ArrayRCP<const size_t> rowptrQuasiRegion;
-      ArrayRCP<const LocalOrdinal> colindQuasiRegion;
-      ArrayRCP<const Scalar> valuesQuasiRegion;
-      quasiRegionCrsMat->getAllValues(rowptrQuasiRegion, colindQuasiRegion, valuesQuasiRegion);
+    // Pull out the data from the quasi-region CrsMatrix
+    ArrayRCP<const size_t> rowptrQuasiRegion;
+    ArrayRCP<const LocalOrdinal> colindQuasiRegion;
+    ArrayRCP<const Scalar> valuesQuasiRegion;
+    quasiRegionCrsMat->getAllValues(rowptrQuasiRegion, colindQuasiRegion, valuesQuasiRegion);
 
-      // Do a deep copy of values
-      // (at least we've been doing deep copies so far, maybe we could do shallow copies to save time?)
-      ArrayRCP<size_t> rowptrRegion(rowptrQuasiRegion.size());
-      ArrayRCP<LocalOrdinal> colindRegion(colindQuasiRegion.size());
-      ArrayRCP<Scalar> valuesRegion(valuesQuasiRegion.size());
+    // Do a deep copy of values
+    // (at least we've been doing deep copies so far, maybe we could do shallow copies to save time?)
+    ArrayRCP<size_t> rowptrRegion(rowptrQuasiRegion.size());
+    ArrayRCP<LocalOrdinal> colindRegion(colindQuasiRegion.size());
+    ArrayRCP<Scalar> valuesRegion(valuesQuasiRegion.size());
 
-      regionCrsMat->allocateAllValues(valuesRegion.size(), rowptrRegion, colindRegion, valuesRegion);
+    regionCrsMat->allocateAllValues(valuesRegion.size(), rowptrRegion, colindRegion, valuesRegion);
 
-      for(LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(rowptrRegion.size()); ++idx) {
-        rowptrRegion[idx] = rowptrQuasiRegion[idx];
-      }
+    for (LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(rowptrRegion.size()); ++idx) {
+      rowptrRegion[idx] = rowptrQuasiRegion[idx];
+    }
 
-      for(LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(colindRegion.size()); ++idx) {
-        colindRegion[idx] = colindQuasiRegion[idx];
-        valuesRegion[idx] = valuesQuasiRegion[idx];
-      }
+    for (LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(colindRegion.size()); ++idx) {
+      colindRegion[idx] = colindQuasiRegion[idx];
+      valuesRegion[idx] = valuesQuasiRegion[idx];
+    }
 
-      // Set and fillComplete the region CrsMatrix
-      regionCrsMat->setAllValues(rowptrRegion, colindRegion, valuesRegion);
-      regionCrsMat->expertStaticFillComplete(revisedRowMap, revisedRowMap);
+    // Set and fillComplete the region CrsMatrix
+    regionCrsMat->setAllValues(rowptrRegion, colindRegion, valuesRegion);
+    regionCrsMat->expertStaticFillComplete(revisedRowMap, revisedRowMap);
   }
 
   tm = Teuchos::null;
@@ -280,7 +276,7 @@ void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, Gl
 
     // move to regional layout
     RCP<Vector> quasiRegNspViolation = VectorFactory::Build(rowMap, true);
-    regNspViolation = VectorFactory::Build(revisedRowMap, true);
+    regNspViolation                  = VectorFactory::Build(revisedRowMap, true);
     compositeToRegional(nspViolation, quasiRegNspViolation, regNspViolation,
                         revisedRowMap, rowImport);
 
@@ -328,9 +324,8 @@ void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, Gl
   regionMats->apply(*regNsp, *regCorrection);
   regionMats->SetFixedBlockSize(AComp->GetFixedBlockSize());
 
-
   RCP<Vector> regDiag = Teuchos::null;
-  regDiag = VectorFactory::Build(revisedRowMap, true);
+  regDiag             = VectorFactory::Build(revisedRowMap, true);
   regionMats->getLocalDiagCopy(*regDiag);
   regDiag->update(-SC_ONE, *regCorrection, SC_ONE, *regNspViolation, SC_ONE);
 
@@ -339,8 +334,7 @@ void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, Gl
   regionCrsMat->replaceDiag(*regDiag);
 
   tm = Teuchos::null;
-} // MakeRegionMatrices
-
+}  // MakeRegionMatrices
 
 /*! \brief Transform regional matrix to composite layout
  *
@@ -356,18 +350,17 @@ void MakeRegionMatrices(const RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, Gl
  *  \return Composite matrix that is fill-completed
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regMat, ///< Matrix in region layout [in]
-                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap, ///< row maps in quasiRegion layout [in]
-                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > colMap, ///< col maps in quasiRegion layout [in]
-                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport, ///< row importer in region layout [in]
-                         const Xpetra::CombineMode combineMode, ///< Combine mode for import/export [in]
-                         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& compMat ///< Matrix in composite layout [in/out]
-                         )
-{
+void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regMat,  ///< Matrix in region layout [in]
+                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap,        ///< row maps in quasiRegion layout [in]
+                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap,        ///< col maps in quasiRegion layout [in]
+                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> rowImport,        ///< row importer in region layout [in]
+                         const Xpetra::CombineMode combineMode,                                         ///< Combine mode for import/export [in]
+                         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& compMat        ///< Matrix in composite layout [in/out]
+) {
 #include "Xpetra_UseShortNames.hpp"
-  using Teuchos::TimeMonitor;
-  using Teuchos::rcp;
   using std::size_t;
+  using Teuchos::rcp;
+  using Teuchos::TimeMonitor;
 
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: Matrix")));
 
@@ -386,8 +379,8 @@ void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOr
   RCP<Matrix> quasiRegMat;
   {
     quasiRegMat = rcp(new CrsMatrixWrap(rowMap,
-                                           colMap,
-                                           regMat->getCrsGraph()->getLocalMaxNumRowEntries()));
+                                        colMap,
+                                        regMat->getCrsGraph()->getLocalMaxNumRowEntries()));
 
     // Extract current quasi-region CrsMatrix
     RCP<CrsMatrix> quasiRegionCrsMat = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(quasiRegMat)->getCrsMatrix();
@@ -409,11 +402,11 @@ void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOr
 
     quasiRegionCrsMat->allocateAllValues(valuesQuasiRegion.size(), rowptrQuasiRegion, colindQuasiRegion, valuesQuasiRegion);
 
-    for(LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(rowptrQuasiRegion.size()); ++idx) {
+    for (LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(rowptrQuasiRegion.size()); ++idx) {
       rowptrQuasiRegion[idx] = rowptrRegion[idx];
     }
 
-    for(LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(colindQuasiRegion.size()); ++idx) {
+    for (LocalOrdinal idx = 0; idx < static_cast<LocalOrdinal>(colindQuasiRegion.size()); ++idx) {
       colindQuasiRegion[idx] = colindRegion[idx];
       valuesQuasiRegion[idx] = valuesRegion[idx];
     }
@@ -426,7 +419,7 @@ void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOr
   // Export from quasiRegional format to composite layout
   RCP<Matrix> partialCompMat;
   partialCompMat = MatrixFactory::Build(compMat->getRowMap(),
-                                           8*regMat->getCrsGraph()->getLocalMaxNumRowEntries());
+                                        8 * regMat->getCrsGraph()->getLocalMaxNumRowEntries());
   partialCompMat->doExport(*(quasiRegMat), *(rowImport), Xpetra::INSERT);
   partialCompMat->fillComplete();
 
@@ -436,29 +429,28 @@ void regionalToComposite(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOr
   compMat->fillComplete();
 
   return;
-} // regionalToComposite
-
+}  // regionalToComposite
 
 /*! \brief Compute local data needed to perform a MatVec in region format
 
  */
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> >& interfaceGIDsMV,
-                 const Teuchos::RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> >& regionsPerGIDWithGhosts,
-                 const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& regionRowMap,
-                 const Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
+void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>& interfaceGIDsMV,
+                 const Teuchos::RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>& regionsPerGIDWithGhosts,
+                 const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>& regionRowMap,
+                 const Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& rowImport,
                  Teuchos::ArrayRCP<LocalOrdinal>& regionMatVecLIDs,
-                 Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& regionInterfaceImporter) {
+                 Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>& regionInterfaceImporter) {
 #include "Xpetra_UseShortNamesOrdinal.hpp"
   using Teuchos::TimeMonitor;
 
   RCP<TimeMonitor> tm;
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("SetupMatVec: 1 - sumInterfaceValues")));
 
-  const LO  maxRegPerGID  = static_cast<LO>(regionsPerGIDWithGhosts->getNumVectors());
-  const int myRank        = regionRowMap->getComm()->getRank();
+  const LO maxRegPerGID = static_cast<LO>(regionsPerGIDWithGhosts->getNumVectors());
+  const int myRank      = regionRowMap->getComm()->getRank();
   interfaceGIDsMV->replaceMap(regionRowMap);
-  RCP<Xpetra::MultiVector<GO, LO, GO, NO> > interfaceGIDs;
+  RCP<Xpetra::MultiVector<GO, LO, GO, NO>> interfaceGIDs;
   interfaceGIDs = interfaceGIDsMV;
   sumInterfaceValues(interfaceGIDs, regionRowMap, rowImport);
 
@@ -467,14 +459,13 @@ void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdi
 
   Teuchos::Array<LO> regionMatVecLIDstmp;
   Teuchos::Array<GO> regionMatVecGIDs;
-  Array<ArrayRCP<const LO> > regionsPerGIDWithGhostsData(maxRegPerGID);
-  Array<ArrayRCP<const GO> > interfaceGIDsData(maxRegPerGID);
-  for(LO regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
+  Array<ArrayRCP<const LO>> regionsPerGIDWithGhostsData(maxRegPerGID);
+  Array<ArrayRCP<const GO>> interfaceGIDsData(maxRegPerGID);
+  for (LO regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
     regionsPerGIDWithGhostsData[regionIdx] = regionsPerGIDWithGhosts->getData(regionIdx);
-    interfaceGIDsData[regionIdx] = interfaceGIDs->getData(regionIdx);
-    for(LO idx = 0; idx < static_cast<LO>(regionsPerGIDWithGhostsData[regionIdx].size()); ++idx) {
-      if((regionsPerGIDWithGhostsData[regionIdx][idx] != -1)
-         && (regionsPerGIDWithGhostsData[regionIdx][idx] != myRank)) {
+    interfaceGIDsData[regionIdx]           = interfaceGIDs->getData(regionIdx);
+    for (LO idx = 0; idx < static_cast<LO>(regionsPerGIDWithGhostsData[regionIdx].size()); ++idx) {
+      if ((regionsPerGIDWithGhostsData[regionIdx][idx] != -1) && (regionsPerGIDWithGhostsData[regionIdx][idx] != myRank)) {
         regionMatVecLIDstmp.push_back(idx);
         regionMatVecGIDs.push_back(interfaceGIDsData[regionIdx][idx]);
       }
@@ -485,11 +476,11 @@ void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdi
   // so we can store it and retrieve it easily later on.
   regionMatVecLIDs.deepCopy(regionMatVecLIDstmp());
 
-  RCP<Map> regionInterfaceMap = Xpetra::MapFactory<LO,GO,Node>::Build(regionRowMap->lib(),
-                                                                      Teuchos::OrdinalTraits<GO>::invalid(),
-                                                                      regionMatVecGIDs(),
-                                                                      regionRowMap->getIndexBase(),
-                                                                      regionRowMap->getComm());
+  RCP<Map> regionInterfaceMap = Xpetra::MapFactory<LO, GO, Node>::Build(regionRowMap->lib(),
+                                                                        Teuchos::OrdinalTraits<GO>::invalid(),
+                                                                        regionMatVecGIDs(),
+                                                                        regionRowMap->getIndexBase(),
+                                                                        regionRowMap->getComm());
 
   tm = Teuchos::null;
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("SetupMatVec: 3 - Build importer")));
@@ -497,7 +488,7 @@ void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdi
   regionInterfaceImporter = ImportFactory::Build(regionRowMap, regionInterfaceMap);
 
   tm = Teuchos::null;
-} // SetupMatVec
+}  // SetupMatVec
 
 /*! \brief Compute the residual \f$r = b - Ax\f$ with pre-computed communication patterns
  *
@@ -506,14 +497,12 @@ void SetupMatVec(const Teuchos::RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdi
  *  2. Compute residual via r = b - y
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void
-computeResidual(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regRes, ///< residual (to be evaluated)
-                const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regX, ///< left-hand side (solution)
-                const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB, ///< right-hand side (forcing term)
-                const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats, ///< matrix in region format
-                const Teuchos::ParameterList& params ///< parameter with fast MatVec parameters and pre-computed communication patterns
-    )
-{
+void computeResidual(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regRes,           ///< residual (to be evaluated)
+                     const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regX,        ///< left-hand side (solution)
+                     const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regB,        ///< right-hand side (forcing term)
+                     const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regionMats,  ///< matrix in region format
+                     const Teuchos::ParameterList& params                                              ///< parameter with fast MatVec parameters and pre-computed communication patterns
+) {
 #include "Xpetra_UseShortNames.hpp"
   using TST = Teuchos::ScalarTraits<Scalar>;
   using Teuchos::TimeMonitor;
@@ -522,7 +511,7 @@ computeResidual(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >&
 
   // Get pre-communicated communication patterns for the fast MatVec
   const ArrayRCP<LocalOrdinal> regionInterfaceLIDs = params.get<ArrayRCP<LO>>("Fast MatVec: interface LIDs");
-  const RCP<Import> regionInterfaceImporter = params.get<RCP<Import>>("Fast MatVec: interface importer");
+  const RCP<Import> regionInterfaceImporter        = params.get<RCP<Import>>("Fast MatVec: interface importer");
 
   // Step 1: Compute region version of y = Ax and store it in regRes
   regionMats->apply(*regX, *regRes, Teuchos::NO_TRANS, TST::one(), TST::zero(), true, regionInterfaceImporter, regionInterfaceLIDs);
@@ -531,6 +520,6 @@ computeResidual(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >&
   regRes->update(TST::one(), *regB, -TST::one(), *regRes, TST::zero());
 
   tm = Teuchos::null;
-} // computeResidual
+}  // computeResidual
 
-#endif // MUELU_SETUPREGIONMATRIX_DEF_HPP
+#endif  // MUELU_SETUPREGIONMATRIX_DEF_HPP
diff --git a/packages/muelu/research/regionMG/src/SetupRegionSmoothers_def.hpp b/packages/muelu/research/regionMG/src/SetupRegionSmoothers_def.hpp
index 079feb75c5a9..563f337d2787 100644
--- a/packages/muelu/research/regionMG/src/SetupRegionSmoothers_def.hpp
+++ b/packages/muelu/research/regionMG/src/SetupRegionSmoothers_def.hpp
@@ -59,15 +59,14 @@
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_CrsMatrixWrap.hpp>
 
-
 #include "SetupRegionMatrix_def.hpp"
 #include "SetupRegionVector_def.hpp"
 
-using Teuchos::RCP;
-using Teuchos::ArrayRCP;
 using Teuchos::Array;
+using Teuchos::ArrayRCP;
 using Teuchos::ArrayView;
 using Teuchos::ParameterList;
+using Teuchos::RCP;
 
 /*! \brief Create list of valid smoother types
  *
@@ -76,13 +75,12 @@ using Teuchos::ParameterList;
  *
  * ToDo: replace this list by an enum when we migrate to actual code.
  */
-std::map<std::string, int> getListOfValidSmootherTypes()
-{
+std::map<std::string, int> getListOfValidSmootherTypes() {
   std::map<std::string, int> smootherTypes;
-  smootherTypes.insert(std::pair<std::string, int>("None",      0));
-  smootherTypes.insert(std::pair<std::string, int>("Jacobi",    1));
-  smootherTypes.insert(std::pair<std::string, int>("Gauss",     2));
-  smootherTypes.insert(std::pair<std::string, int>("SymmetricGauss",     3));
+  smootherTypes.insert(std::pair<std::string, int>("None", 0));
+  smootherTypes.insert(std::pair<std::string, int>("Jacobi", 1));
+  smootherTypes.insert(std::pair<std::string, int>("Gauss", 2));
+  smootherTypes.insert(std::pair<std::string, int>("SymmetricGauss", 3));
   smootherTypes.insert(std::pair<std::string, int>("Chebyshev", 4));
 
   return smootherTypes;
@@ -92,11 +90,11 @@ std::map<std::string, int> getListOfValidSmootherTypes()
  *
  * Computes the inverse of the diagonal in region format and with interface scaling
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void computeInverseDiagonal(RCP<Teuchos::ParameterList> params,
-                 const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
-                 const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
-                 const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) ///< row importer in region layout [in]
+                            const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
+                            const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
+                            const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport)  ///< row importer in region layout [in]
 {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
@@ -115,22 +113,19 @@ void computeInverseDiagonal(RCP<Teuchos::ParameterList> params,
   params->set<RCP<Vector> >("smoothers: inverse diagonal", diagReg);
 }
 
-
 /*! \brief Do Jacobi smoothing
  *
  *  Perform Jacobi smoothing in the region layout using the true diagonal value
  *  recovered from the splitted matrix.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void jacobiIterate(RCP<Teuchos::ParameterList> smootherParams,
-                   RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX, // left-hand side (or solution)
-                   const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB, // right-hand side (or residual)
-                   const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats, // matrices in true region layout
-                   const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap, ///< revised row maps in region layout [in] (actually extracted from regionMats)
-                   const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,///< row importer in region layout [in]
-		   bool& zeroInitGuess
-    )
-{
+                   RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX,             // left-hand side (or solution)
+                   const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB,        // right-hand side (or residual)
+                   const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,  // matrices in true region layout
+                   const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,    ///< revised row maps in region layout [in] (actually extracted from regionMats)
+                   const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,           ///< row importer in region layout [in]
+                   bool& zeroInitGuess) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Jacobi Iterate")));
@@ -138,19 +133,17 @@ void jacobiIterate(RCP<Teuchos::ParameterList> smootherParams,
   // const Scalar SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
   const Scalar SC_ONE = Teuchos::ScalarTraits<Scalar>::one();
 
-  const int maxIter    = smootherParams->get<int>   ("smoother: sweeps");
+  const int maxIter    = smootherParams->get<int>("smoother: sweeps");
   const double damping = smootherParams->get<double>("smoother: damping");
   RCP<Vector> diag_inv = smootherParams->get<RCP<Vector> >("smoothers: inverse diagonal");
 
   RCP<Vector> regRes = VectorFactory::Build(revisedRowMap, true);
 
   for (int iter = 0; iter < maxIter; ++iter) {
-
     // Update the residual vector
     if (zeroInitGuess) {
       regX->elementWiseMultiply(damping, *diag_inv, *regB, SC_ONE);
-    }
-    else {
+    } else {
       computeResidual(regRes, regX, regB, regionMats, *smootherParams);
 
       // update solution according to Jacobi's method
@@ -160,8 +153,7 @@ void jacobiIterate(RCP<Teuchos::ParameterList> smootherParams,
   }
 
   return;
-} // jacobiIterate
-
+}  // jacobiIterate
 
 /*! \brief Do Gauss-Seidel smoothing
  *
@@ -169,23 +161,21 @@ void jacobiIterate(RCP<Teuchos::ParameterList> smootherParams,
  *  recovered from the splitted matrix. Off-diagonal values are just taken as they are
  *  in region format.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void GSIterate(RCP<Teuchos::ParameterList> smootherParams,
-               RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX, // left-hand side (or solution)
-               const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB, // right-hand side (or residual)
-               const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats, // matrices in true region layout
-               const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap, ///< revised row maps in region layout [in] (actually extracted from regionMats)
-               const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,///< row importer in region layout [in]
+               RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX,             // left-hand side (or solution)
+               const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB,        // right-hand side (or residual)
+               const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,  // matrices in true region layout
+               const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,    ///< revised row maps in region layout [in] (actually extracted from regionMats)
+               const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,           ///< row importer in region layout [in]
                bool& zeroInitGuess,
-               bool sgs = false
-               )
-{
+               bool sgs = false) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Gauss-Seidel Iterate")));
 
   // Extract user-given and pre-computed data from paremter list
-  const int maxIter = smootherParams->get<int>("smoother: sweeps");
+  const int maxIter    = smootherParams->get<int>("smoother: sweeps");
   const double damping = smootherParams->get<double>("smoother: damping");
   RCP<Vector> diag_inv = smootherParams->get<RCP<Vector> >("smoothers: inverse diagonal");
 
@@ -195,21 +185,19 @@ void GSIterate(RCP<Teuchos::ParameterList> smootherParams,
 
   // GS iteration loop
   for (int iter = 0; iter < maxIter; ++iter) {
-
     // Update the residual vector
-    if (!zeroInitGuess)
-    {
+    if (!zeroInitGuess) {
       computeResidual(regRes, regX, regB, regionMats, *smootherParams);
     }
 
     // update the solution and the residual
 
-    using MT = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-    RCP<Vector> delta = VectorFactory::Build(regionMats->getRowMap(), true);
-    ArrayRCP<SC> ldelta = delta->getDataNonConst(0);
-    ArrayRCP<SC> OneregX = regX->getDataNonConst(0);
+    using MT               = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+    RCP<Vector> delta      = VectorFactory::Build(regionMats->getRowMap(), true);
+    ArrayRCP<SC> ldelta    = delta->getDataNonConst(0);
+    ArrayRCP<SC> OneregX   = regX->getDataNonConst(0);
     ArrayRCP<SC> OneregRes = regRes->getDataNonConst(0);
-    if (zeroInitGuess) { // copy regB to regRes
+    if (zeroInitGuess) {  // copy regB to regRes
       ArrayRCP<SC> rhs = regB->getDataNonConst(0);
       for (size_t k = 0; k < numRows; ++k) OneregRes[k] = rhs[k];
     }
@@ -221,58 +209,55 @@ void GSIterate(RCP<Teuchos::ParameterList> smootherParams,
       ArrayView<const LO> AAcols;
       ArrayView<const SC> AAvals;
       regionMats->getLocalRowView(k, AAcols, AAvals);
-      const int *Acols = AAcols.getRawPtr();
-      const SC  *Avals = AAvals.getRawPtr();
+      const int* Acols = AAcols.getRawPtr();
+      const SC* Avals  = AAvals.getRawPtr();
       const LO RowLeng = AAvals.size();
 
       // Loop over entries in row k and perform GS iteration
       for (LO kk = 0; kk < RowLeng; kk++) {
-        OneregRes[k] -= Avals[kk]*ldelta[Acols[kk]];
+        OneregRes[k] -= Avals[kk] * ldelta[Acols[kk]];
       }
-      ldelta[k] = damping*Onediag[k]*OneregRes[k];
+      ldelta[k] = damping * Onediag[k] * OneregRes[k];
       OneregX[k] += ldelta[k];
     }
     zeroInitGuess = false;
 
-    if( sgs ){
+    if (sgs) {
       for (size_t k = numRows; k--;) {
         // Extract a single row
         ArrayView<const LO> AAcols;
         ArrayView<const SC> AAvals;
         regionMats->getLocalRowView(k, AAcols, AAvals);
-        const int *Acols = AAcols.getRawPtr();
-        const SC  *Avals = AAvals.getRawPtr();
+        const int* Acols = AAcols.getRawPtr();
+        const SC* Avals  = AAvals.getRawPtr();
         const LO RowLeng = AAvals.size();
 
         // Loop over entries in row k and perform GS iteration
         for (LO kk = 0; kk < RowLeng; kk++) {
-          OneregRes[k] -= Avals[kk]*ldelta[Acols[kk]];
+          OneregRes[k] -= Avals[kk] * ldelta[Acols[kk]];
         }
-        ldelta[k] = damping*Onediag[k]*OneregRes[k];
+        ldelta[k] = damping * Onediag[k] * OneregRes[k];
         OneregX[k] += ldelta[k];
       }
     }
   }
 
   return;
-} // GS
-
+}  // GS
 
 //! Transfer region vector to composite format and compute its 2-norm
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 typename Teuchos::ScalarTraits<Scalar>::magnitudeType
 calcNorm2(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,
-          const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport)
-{
+          const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) {
 #include "Xpetra_UseShortNames.hpp"
   const RCP<const Map> mapComp = rowImport->getSourceMap();
-  RCP<Vector> compVec = VectorFactory::Build(mapComp, true);
+  RCP<Vector> compVec          = VectorFactory::Build(mapComp, true);
   regionalToComposite(regVec, compVec, rowImport);
   typename Teuchos::ScalarTraits<Scalar>::magnitudeType norm = compVec->norm2();
 
   return norm;
-} // calcNorm2
-
+}  // calcNorm2
 
 /*! Compute inner product of two region vectors
  *
@@ -285,38 +270,35 @@ calcNorm2(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVe
  *
  * @return Inner product of regX and regY
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 Scalar
 dotProd(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX,
         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regY,
-        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport)
-{
+        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) {
 #include "Xpetra_UseShortNames.hpp"
   const RCP<const Map> mapComp = rowImport->getSourceMap();
-  RCP<Vector> compX = VectorFactory::Build(mapComp, true);
-  RCP<Vector> compY = VectorFactory::Build(mapComp, true);
+  RCP<Vector> compX            = VectorFactory::Build(mapComp, true);
+  RCP<Vector> compY            = VectorFactory::Build(mapComp, true);
   regionalToComposite(regX, compX, rowImport);
   regionalToComposite(regY, compY, rowImport);
   SC dotVal = compX->dot(*compY);
 
   return dotVal;
-} // dotProd
-
+}  // dotProd
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 Scalar
 powerMethod(RCP<Teuchos::ParameterList> params,
             const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
             const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
             const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,
-            const int numIters)
-{
+            const int numIters) {
 #include "Xpetra_UseShortNames.hpp"
 
   RCP<Vector> diag_inv = params->get<RCP<Vector> >("smoothers: inverse diagonal");
-  const SC SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-  const SC SC_ONE  = Teuchos::ScalarTraits<Scalar>::one();
-  SC lambdaMax = SC_ZERO;
+  const SC SC_ZERO     = Teuchos::ScalarTraits<Scalar>::zero();
+  const SC SC_ONE      = Teuchos::ScalarTraits<Scalar>::one();
+  SC lambdaMax         = SC_ZERO;
   SC RQ_top, RQ_bottom, norm;
 
   RCP<Vector> regX = VectorFactory::Build(revisedRowMap, true);
@@ -325,100 +307,96 @@ powerMethod(RCP<Teuchos::ParameterList> params,
   regX->randomize();
 
   norm = calcNorm2(regX, rowImport);
-  regX->scale( SC_ONE / norm );
+  regX->scale(SC_ONE / norm);
 
   for (int iter = 0; iter < numIters; ++iter) {
-
-    regionMats->apply(*regX, *regY); // A.apply (x, y);
-    sumInterfaceValues(regY, revisedRowMap, rowImport); // step 2
+    regionMats->apply(*regX, *regY);                     // A.apply (x, y);
+    sumInterfaceValues(regY, revisedRowMap, rowImport);  // step 2
 
     // Scale by inverse of diagonal
     regY->elementWiseMultiply(SC_ONE, *diag_inv, *regY, SC_ZERO);
 
-    RQ_top = dotProd(regY, regX, rowImport);
+    RQ_top    = dotProd(regY, regX, rowImport);
     RQ_bottom = dotProd(regX, regX, rowImport);
     lambdaMax = RQ_top / RQ_bottom;
 
     norm = calcNorm2(regY, rowImport);
 
-    if (norm == SC_ZERO) { // Return something reasonable.
+    if (norm == SC_ZERO) {  // Return something reasonable.
       return SC_ZERO;
     }
-    regX->update( SC_ONE / norm, *regY, SC_ZERO);
-
+    regX->update(SC_ONE / norm, *regY, SC_ZERO);
   }
 
   return lambdaMax;
-} // powerMethod
+}  // powerMethod
 
 /*! \brief Performs Chebyshev specific setup
  *
  * Use power method to estimate lambdaMx
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void chebyshevSetup(RCP<Teuchos::ParameterList> params,
-                   const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
-                   const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionInterfaceScaling,
-                   const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
-                   const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) {
+                    const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
+                    const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionInterfaceScaling,
+                    const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
+                    const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Chebyshev Setup")));
 
   // Calculate lambdaMax
   Scalar lambdaMax = 1;
-  lambdaMax = powerMethod(params,
-                          regionMats,
-                          revisedRowMap,
-                          rowImport,
-                          10);
-  params->set< Scalar >("chebyshev: lambda max", lambdaMax );
-
-} // chebyshevSetup
+  lambdaMax        = powerMethod(params,
+                                 regionMats,
+                                 revisedRowMap,
+                                 rowImport,
+                                 10);
+  params->set<Scalar>("chebyshev: lambda max", lambdaMax);
 
+}  // chebyshevSetup
 
 /*! \brief The textbook Chebyshev algorithm from Ifpack2 translated into the region format
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void chebyshevIterate(RCP<Teuchos::ParameterList> smootherParams,
-                      RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX, ///< left-hand side (or solution)
-                      const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB, ///< right-hand side (or residual)
-                      const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats, ///< matrices in true region layout
-                      const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap, ///< revised row maps in region layout [in] (actually extracted from regionMats)
-                      const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,///< row importer in region layout [in]
-                      bool& zeroInitGuess ///< Use a zero vector as initial guess?
-                      )
-{
+                      RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX,             ///< left-hand side (or solution)
+                      const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB,        ///< right-hand side (or residual)
+                      const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,  ///< matrices in true region layout
+                      const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,    ///< revised row maps in region layout [in] (actually extracted from regionMats)
+                      const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,           ///< row importer in region layout [in]
+                      bool& zeroInitGuess                                                                ///< Use a zero vector as initial guess?
+) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Chebyshev Iterate")));
 
   // Extract input data from parameter list
-  const int maxIter = smootherParams->get<int>("smoother: sweeps");
-  const Scalar eigRatio = smootherParams->get<double>("smoother: Chebyshev eigRatio");
-  const Scalar lambdaMax = smootherParams->get<Scalar>("chebyshev: lambda max");
+  const int maxIter        = smootherParams->get<int>("smoother: sweeps");
+  const Scalar eigRatio    = smootherParams->get<double>("smoother: Chebyshev eigRatio");
+  const Scalar lambdaMax   = smootherParams->get<Scalar>("chebyshev: lambda max");
   const Scalar boostFactor = smootherParams->get<double>("smoother: Chebyshev boost factor");
-  RCP<Vector> diag_inv = smootherParams->get<RCP<Vector> >("smoothers: inverse diagonal");
+  RCP<Vector> diag_inv     = smootherParams->get<RCP<Vector> >("smoothers: inverse diagonal");
 
   // Define some constants for convenience
   const Scalar SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-  const Scalar SC_ONE = Teuchos::ScalarTraits<Scalar>::one();
-  const Scalar SC_TWO = Teuchos::as<Scalar> (2);
+  const Scalar SC_ONE  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar SC_TWO  = Teuchos::as<Scalar>(2);
 
   // Estimation of eigenvalue interval of interest: [alpha, beta]
-  const Scalar alpha = lambdaMax / eigRatio; // lower bound (estimate via given max-to-min ratio)
-  const Scalar beta = boostFactor * lambdaMax; // upper bound (estimated via boost factor)
+  const Scalar alpha = lambdaMax / eigRatio;     // lower bound (estimate via given max-to-min ratio)
+  const Scalar beta  = boostFactor * lambdaMax;  // upper bound (estimated via boost factor)
 
   // Algorithmic constants
   const Scalar delta = SC_TWO / (beta - alpha);
   const Scalar theta = (beta + alpha) / SC_TWO;
-  const Scalar s1 = theta * delta;
+  const Scalar s1    = theta * delta;
 
   // Algorithmic parameters
   Scalar dtemp1 = SC_ZERO;
   Scalar dtemp2 = SC_ZERO;
   Scalar rhokp1 = SC_ZERO;
-  Scalar rhok = SC_ONE / s1;
+  Scalar rhok   = SC_ONE / s1;
 
   RCP<Vector> regRes = VectorFactory::Build(revisedRowMap, true);
 
@@ -427,17 +405,16 @@ void chebyshevIterate(RCP<Teuchos::ParameterList> smootherParams,
 
   // First Iteration
   if (zeroInitGuess) {
-    regZ->elementWiseMultiply(SC_ONE, *diag_inv, *regB, SC_ZERO); // Z = D_inv * b
-    regP->update(SC_ONE/theta, *regZ, SC_ZERO); // P = 1/theta Z
-    regX->update(SC_ONE, *regP, SC_ZERO); // X = 0 + P
-  }
-  else {
+    regZ->elementWiseMultiply(SC_ONE, *diag_inv, *regB, SC_ZERO);  // Z = D_inv * b
+    regP->update(SC_ONE / theta, *regZ, SC_ZERO);                  // P = 1/theta Z
+    regX->update(SC_ONE, *regP, SC_ZERO);                          // X = 0 + P
+  } else {
     // Compute residual vector
     computeResidual(regRes, regX, regB, regionMats, *smootherParams);
 
-    regZ->elementWiseMultiply(SC_ONE, *diag_inv, *regRes, SC_ZERO); // z = D_inv * R, that is, D \ R.
-    regP->update(SC_ONE/theta, *regZ, SC_ZERO); // P = 1/theta Z
-    regX->update(SC_ONE, *regP, SC_ONE); // X = X + P
+    regZ->elementWiseMultiply(SC_ONE, *diag_inv, *regRes, SC_ZERO);  // z = D_inv * R, that is, D \ R.
+    regP->update(SC_ONE / theta, *regZ, SC_ZERO);                    // P = 1/theta Z
+    regX->update(SC_ONE, *regP, SC_ONE);                             // X = X + P
   }
 
   // The rest of the iterations
@@ -451,9 +428,9 @@ void chebyshevIterate(RCP<Teuchos::ParameterList> smootherParams,
     rhokp1 = SC_ONE / (SC_TWO * s1 - rhok);
     dtemp1 = rhokp1 * rhok;
     dtemp2 = SC_TWO * rhokp1 * delta;
-    rhok = rhokp1;
-    regP->update(dtemp2, *regZ, dtemp1);// P = dtemp2*Z + dtemp1*P
-    regX->update(SC_ONE, *regP, SC_ONE);// X = X + P
+    rhok   = rhokp1;
+    regP->update(dtemp2, *regZ, dtemp1);  // P = dtemp2*Z + dtemp1*P
+    regX->update(SC_ONE, *regP, SC_ONE);  // X = X + P
 
     // If we compute the residual here, we could either do R = B -
     // A*X, or R = R - alpha*A*P.  Since we choose the former, we
@@ -461,15 +438,14 @@ void chebyshevIterate(RCP<Teuchos::ParameterList> smootherParams,
   }
 
   zeroInitGuess = false;
-} // chebyshevIterate
+}  // chebyshevIterate
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void smootherSetup(RCP<Teuchos::ParameterList> params,
                    const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
                    const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
                    const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionInterfaceScaling,
-                   const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport) ///< row importer in region layout [in]
+                   const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport)  ///< row importer in region layout [in]
 {
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Smoother: 1 - Setup")));
@@ -478,42 +454,39 @@ void smootherSetup(RCP<Teuchos::ParameterList> params,
 
   std::map<std::string, int> smootherTypes = getListOfValidSmootherTypes();
 
-  switch(smootherTypes[type]) {
-  case 0: // None
-  {
-    break;
-  }
-  case 1: // Jacobi
-  case 2: // Gauss-Seidel
-  case 3: // Symmetric Gauss-Seidel
-  {
-    computeInverseDiagonal(params, revisedRowMap, regionMats, rowImport);
-    break;
-  }
-  case 4: // Chebyshev
-  {
-    computeInverseDiagonal(params, revisedRowMap, regionMats, rowImport);
-    chebyshevSetup(params, regionMats, regionInterfaceScaling, revisedRowMap, rowImport);
-    break;
-  }
-  default:
-  {
-    std::cout << "Unknown smoother: " << type << "!" << std::endl;
-    throw;
-  }
+  switch (smootherTypes[type]) {
+    case 0:  // None
+    {
+      break;
+    }
+    case 1:  // Jacobi
+    case 2:  // Gauss-Seidel
+    case 3:  // Symmetric Gauss-Seidel
+    {
+      computeInverseDiagonal(params, revisedRowMap, regionMats, rowImport);
+      break;
+    }
+    case 4:  // Chebyshev
+    {
+      computeInverseDiagonal(params, revisedRowMap, regionMats, rowImport);
+      chebyshevSetup(params, regionMats, regionInterfaceScaling, revisedRowMap, rowImport);
+      break;
+    }
+    default: {
+      std::cout << "Unknown smoother: " << type << "!" << std::endl;
+      throw;
+    }
   }
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void smootherApply(RCP<Teuchos::ParameterList> params,
                    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regX,
                    const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regB,
                    const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regionMats,
                    const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,
                    const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport,
-                   bool& zeroInitGuess)
-{
+                   bool& zeroInitGuess) {
   using Teuchos::TimeMonitor;
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Region Smoother: 2 - Apply")));
 
@@ -521,38 +494,37 @@ void smootherApply(RCP<Teuchos::ParameterList> params,
 
   std::map<std::string, int> smootherTypes = getListOfValidSmootherTypes();
 
-  switch(smootherTypes[type]) {
-  case 0: // None
-  {
-    break;
-  }
-  case 1: // Jacobi
-  {
-    jacobiIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
-    break;
-  }
-  case 2: // Gauss-Seidel
-  {
-    GSIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
-    break;
-  }
-  case 3: // Symmetric Gauss-Seidel
-  {
-    GSIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess, true);
-    break;
-  }
-  case 4: // Chebyshev
-    chebyshevIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
-  {
-    break;
-  }
-  default:
-  {
-    std::cout << "Unknown smoother: " << type << "!" << std::endl;
-    throw;
-  }
+  switch (smootherTypes[type]) {
+    case 0:  // None
+    {
+      break;
+    }
+    case 1:  // Jacobi
+    {
+      jacobiIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
+      break;
+    }
+    case 2:  // Gauss-Seidel
+    {
+      GSIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
+      break;
+    }
+    case 3:  // Symmetric Gauss-Seidel
+    {
+      GSIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess, true);
+      break;
+    }
+    case 4:  // Chebyshev
+      chebyshevIterate(params, regX, regB, regionMats, revisedRowMap, rowImport, zeroInitGuess);
+      {
+        break;
+      }
+    default: {
+      std::cout << "Unknown smoother: " << type << "!" << std::endl;
+      throw;
+    }
   }
 
-} // smootherApply
+}  // smootherApply
 
-#endif // MUELU_SETUPREGIONSMOOTHERS_DEF_HPP
+#endif  // MUELU_SETUPREGIONSMOOTHERS_DEF_HPP
diff --git a/packages/muelu/research/regionMG/src/SetupRegionUtilities.hpp b/packages/muelu/research/regionMG/src/SetupRegionUtilities.hpp
index 5c650dfd7e8c..f76d1ed140f8 100644
--- a/packages/muelu/research/regionMG/src/SetupRegionUtilities.hpp
+++ b/packages/muelu/research/regionMG/src/SetupRegionUtilities.hpp
@@ -47,12 +47,11 @@
 
 #include "Teuchos_Array.hpp"
 
-template<class LocalOrdinal>
+template <class LocalOrdinal>
 void findInterface(const int numDimensions, Teuchos::Array<LocalOrdinal> nodesPerDim,
                    const Teuchos::Array<int> boundaryConditions,
                    Teuchos::Array<LocalOrdinal>& interfacesDimensions,
                    Teuchos::Array<LocalOrdinal>& interfacesLIDs) {
-
   using LO = LocalOrdinal;
 
   std::cout << "nodesPerDim: " << nodesPerDim << std::endl;
@@ -64,162 +63,153 @@ void findInterface(const int numDimensions, Teuchos::Array<LocalOrdinal> nodesPe
   // Step 1: determine what edges/faces are needed based on BC.
   LO numInterfaces = 0, numTotalLIDs = 0, numInterfaceLIDs;
   interfacesDimensions.resize(18, 1);
-  for(LO dim = 0; dim < 3; ++dim) {
+  for (LO dim = 0; dim < 3; ++dim) {
     // Check for nodes and no boundary conditions in this direction, otherwise skip it.
-    if((nodesPerDim[dim] == 1) || (boundaryConditions[2*dim] + boundaryConditions[2*dim + 1] == 2)) {
+    if ((nodesPerDim[dim] == 1) || (boundaryConditions[2 * dim] + boundaryConditions[2 * dim + 1] == 2)) {
       continue;
     }
 
     // Since we are not skipping this direction
     // we at least need to store data for one surface
     numInterfaceLIDs = 1;
-    for(LO dimIdx = 0; dimIdx < 3; ++dimIdx) {
-      interfacesDimensions[numInterfaces*3 + dimIdx] = (dimIdx == dim ? 1 : nodesPerDim[dimIdx]);
+    for (LO dimIdx = 0; dimIdx < 3; ++dimIdx) {
+      interfacesDimensions[numInterfaces * 3 + dimIdx] = (dimIdx == dim ? 1 : nodesPerDim[dimIdx]);
       numInterfaceLIDs *= (dimIdx == dim ? 1 : nodesPerDim[dimIdx]);
     }
     numTotalLIDs += numInterfaceLIDs;
     ++numInterfaces;
 
     // If there are no BC we need to store the surface twice.
-    if(boundaryConditions[2*dim] + boundaryConditions[2*dim + 1] == 0) {
-      for(LO dimIdx = 0; dimIdx < 3; ++dimIdx) {
-        interfacesDimensions[numInterfaces*3 + dimIdx]
-          = interfacesDimensions[(numInterfaces - 1)*3 + dimIdx];
+    if (boundaryConditions[2 * dim] + boundaryConditions[2 * dim + 1] == 0) {
+      for (LO dimIdx = 0; dimIdx < 3; ++dimIdx) {
+        interfacesDimensions[numInterfaces * 3 + dimIdx] = interfacesDimensions[(numInterfaces - 1) * 3 + dimIdx];
       }
       numTotalLIDs += numInterfaceLIDs;
       ++numInterfaces;
     }
   }
-  interfacesDimensions.resize(3*numInterfaces);
+  interfacesDimensions.resize(3 * numInterfaces);
   interfacesLIDs.resize(numTotalLIDs, -1);
 
   // Step 2 lazy implementation of all geometrical cases.
 
   LO nodeOffset = 0;
-  if(numDimensions == 2) {
+  if (numDimensions == 2) {
     // left interface
-    if(boundaryConditions[0] == 0) {
-      for(LO nodeIdx = 0; nodeIdx < nodesPerDim[1]; ++nodeIdx) {
-        interfacesLIDs[nodeOffset + nodeIdx] = nodeIdx*nodesPerDim[0];
+    if (boundaryConditions[0] == 0) {
+      for (LO nodeIdx = 0; nodeIdx < nodesPerDim[1]; ++nodeIdx) {
+        interfacesLIDs[nodeOffset + nodeIdx] = nodeIdx * nodesPerDim[0];
       }
       nodeOffset += nodesPerDim[1];
     }
 
     // right interface
-    if(boundaryConditions[1] == 0) {
-      for(LO nodeIdx = 0; nodeIdx < nodesPerDim[1]; ++nodeIdx) {
-        interfacesLIDs[nodeOffset + nodeIdx] = (nodeIdx + 1)*nodesPerDim[0] - 1;
+    if (boundaryConditions[1] == 0) {
+      for (LO nodeIdx = 0; nodeIdx < nodesPerDim[1]; ++nodeIdx) {
+        interfacesLIDs[nodeOffset + nodeIdx] = (nodeIdx + 1) * nodesPerDim[0] - 1;
       }
       nodeOffset += nodesPerDim[1];
     }
 
     // front interface
-    if(boundaryConditions[2] == 0) {
-      for(LO nodeIdx = 0; nodeIdx < nodesPerDim[0]; ++nodeIdx) {
+    if (boundaryConditions[2] == 0) {
+      for (LO nodeIdx = 0; nodeIdx < nodesPerDim[0]; ++nodeIdx) {
         interfacesLIDs[nodeOffset + nodeIdx] = nodeIdx;
       }
       nodeOffset += nodesPerDim[0];
     }
 
     // back interface
-    if(boundaryConditions[3] == 0) {
-      for(LO nodeIdx = 0; nodeIdx < nodesPerDim[0]; ++nodeIdx) {
-        interfacesLIDs[nodeOffset + nodeIdx] = (nodesPerDim[1] - 1)*nodesPerDim[0] + nodeIdx;
+    if (boundaryConditions[3] == 0) {
+      for (LO nodeIdx = 0; nodeIdx < nodesPerDim[0]; ++nodeIdx) {
+        interfacesLIDs[nodeOffset + nodeIdx] = (nodesPerDim[1] - 1) * nodesPerDim[0] + nodeIdx;
       }
       nodeOffset += nodesPerDim[0];
     }
   }
 
-  if(numDimensions == 3) {
+  if (numDimensions == 3) {
     // left interface
-    if(boundaryConditions[0] == 0) {
-      for(LO k = 0; k < nodesPerDim[2]; ++k) {
-        for(LO j = 0; j < nodesPerDim[1]; ++j) {
-          interfacesLIDs[nodeOffset + k*nodesPerDim[1] + j]
-            = k*nodesPerDim[1]*nodesPerDim[0] + j*nodesPerDim[0];
+    if (boundaryConditions[0] == 0) {
+      for (LO k = 0; k < nodesPerDim[2]; ++k) {
+        for (LO j = 0; j < nodesPerDim[1]; ++j) {
+          interfacesLIDs[nodeOffset + k * nodesPerDim[1] + j] = k * nodesPerDim[1] * nodesPerDim[0] + j * nodesPerDim[0];
         }
       }
-      nodeOffset += nodesPerDim[2]*nodesPerDim[1];
+      nodeOffset += nodesPerDim[2] * nodesPerDim[1];
     }
 
     // right interface
-    if(boundaryConditions[1] == 0) {
-      for(LO k = 0; k < nodesPerDim[2]; ++k) {
-        for(LO j = 0; j < nodesPerDim[1]; ++j) {
-          interfacesLIDs[nodeOffset + k*nodesPerDim[1] + j]
-            = k*nodesPerDim[1]*nodesPerDim[0] + (j + 1)*nodesPerDim[0] - 1;
+    if (boundaryConditions[1] == 0) {
+      for (LO k = 0; k < nodesPerDim[2]; ++k) {
+        for (LO j = 0; j < nodesPerDim[1]; ++j) {
+          interfacesLIDs[nodeOffset + k * nodesPerDim[1] + j] = k * nodesPerDim[1] * nodesPerDim[0] + (j + 1) * nodesPerDim[0] - 1;
         }
       }
-      nodeOffset += nodesPerDim[2]*nodesPerDim[1];
+      nodeOffset += nodesPerDim[2] * nodesPerDim[1];
     }
 
     // front interface
-    if(boundaryConditions[2] == 0) {
-      for(LO k = 0; k < nodesPerDim[2]; ++k) {
-        for(LO i = 0; i < nodesPerDim[0]; ++i) {
-          interfacesLIDs[nodeOffset + k*nodesPerDim[0] + i]
-            = k*nodesPerDim[1]*nodesPerDim[0] + i;
+    if (boundaryConditions[2] == 0) {
+      for (LO k = 0; k < nodesPerDim[2]; ++k) {
+        for (LO i = 0; i < nodesPerDim[0]; ++i) {
+          interfacesLIDs[nodeOffset + k * nodesPerDim[0] + i] = k * nodesPerDim[1] * nodesPerDim[0] + i;
         }
       }
-      nodeOffset += nodesPerDim[2]*nodesPerDim[0];
+      nodeOffset += nodesPerDim[2] * nodesPerDim[0];
     }
 
     // back interface
-    if(boundaryConditions[3] == 0) {
-      for(LO k = 0; k < nodesPerDim[2]; ++k) {
-        for(LO i = 0; i < nodesPerDim[0]; ++i) {
-          interfacesLIDs[nodeOffset + k*nodesPerDim[0] + i]
-            = k*nodesPerDim[1]*nodesPerDim[0] + (nodesPerDim[1] - 1)*nodesPerDim[0] + i;
+    if (boundaryConditions[3] == 0) {
+      for (LO k = 0; k < nodesPerDim[2]; ++k) {
+        for (LO i = 0; i < nodesPerDim[0]; ++i) {
+          interfacesLIDs[nodeOffset + k * nodesPerDim[0] + i] = k * nodesPerDim[1] * nodesPerDim[0] + (nodesPerDim[1] - 1) * nodesPerDim[0] + i;
         }
       }
-      nodeOffset += nodesPerDim[2]*nodesPerDim[0];
+      nodeOffset += nodesPerDim[2] * nodesPerDim[0];
     }
 
     // bottom interface
-    if(boundaryConditions[4] == 0) {
-      for(LO j = 0; j < nodesPerDim[1]; ++j) {
-        for(LO i = 0; i < nodesPerDim[0]; ++i) {
-          interfacesLIDs[nodeOffset + j*nodesPerDim[0] + i]
-            = j*nodesPerDim[0] + i;
+    if (boundaryConditions[4] == 0) {
+      for (LO j = 0; j < nodesPerDim[1]; ++j) {
+        for (LO i = 0; i < nodesPerDim[0]; ++i) {
+          interfacesLIDs[nodeOffset + j * nodesPerDim[0] + i] = j * nodesPerDim[0] + i;
         }
       }
-      nodeOffset += nodesPerDim[1]*nodesPerDim[0];
+      nodeOffset += nodesPerDim[1] * nodesPerDim[0];
     }
 
-    //top interface
-    if(boundaryConditions[5] == 0) {
-      for(LO j = 0; j < nodesPerDim[1]; ++j) {
-        for(LO i = 0; i < nodesPerDim[0]; ++i) {
-          interfacesLIDs[nodeOffset + j*nodesPerDim[0] + i]
-            = (nodesPerDim[2] - 1)*nodesPerDim[1]*nodesPerDim[0] + j*nodesPerDim[0] + i;
+    // top interface
+    if (boundaryConditions[5] == 0) {
+      for (LO j = 0; j < nodesPerDim[1]; ++j) {
+        for (LO i = 0; i < nodesPerDim[0]; ++i) {
+          interfacesLIDs[nodeOffset + j * nodesPerDim[0] + i] = (nodesPerDim[2] - 1) * nodesPerDim[1] * nodesPerDim[0] + j * nodesPerDim[0] + i;
         }
       }
-      nodeOffset += nodesPerDim[1]*nodesPerDim[0];
+      nodeOffset += nodesPerDim[1] * nodesPerDim[0];
     }
   }
-} // findInterface
-
+}  // findInterface
 
-template<class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
 void createRegionData(const int numDimensions,
                       const bool useUnstructured, const int numDofsPerNode,
                       const Teuchos::ArrayView<GlobalOrdinal> gNodesPerDim,
-                      const Teuchos::ArrayView<LocalOrdinal>  lNodesPerDim,
+                      const Teuchos::ArrayView<LocalOrdinal> lNodesPerDim,
                       const Teuchos::ArrayView<GlobalOrdinal> procsPerDim,
                       const Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > nodeMap,
                       const Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > dofMap,
                       int& maxRegPerGID, LocalOrdinal& numLocalRegionNodes,
                       Teuchos::Array<int>& boundaryConditions,
-                      Teuchos::Array<GlobalOrdinal>& sendGIDs, ///< GIDs of nodes
+                      Teuchos::Array<GlobalOrdinal>& sendGIDs,  ///< GIDs of nodes
                       Teuchos::Array<int>& sendPIDs,
                       int& numInterfaces,
-                      Teuchos::Array<LocalOrdinal>&  rNodesPerDim,
+                      Teuchos::Array<LocalOrdinal>& rNodesPerDim,
                       Teuchos::Array<GlobalOrdinal>& quasiRegionGIDs,
                       Teuchos::Array<GlobalOrdinal>& quasiRegionCoordGIDs,
-                      Teuchos::Array<LocalOrdinal>&  compositeToRegionLIDs,
+                      Teuchos::Array<LocalOrdinal>& compositeToRegionLIDs,
                       Teuchos::Array<GlobalOrdinal>& interfaceGIDs,
-                      Teuchos::Array<LocalOrdinal>&  interfaceLIDsData) {
-
+                      Teuchos::Array<LocalOrdinal>& interfaceLIDsData) {
   using GO = GlobalOrdinal;
   using LO = LocalOrdinal;
 
@@ -228,40 +218,52 @@ void createRegionData(const int numDimensions,
   Teuchos::Array<GO> endIndices(3);
   const GO startGID = dofMap->getMinGlobalIndex() / numDofsPerNode;
   {
-    startIndices[2] = startGID / (gNodesPerDim[1]*gNodesPerDim[0]);
-    const GO rem    = startGID % (gNodesPerDim[1]*gNodesPerDim[0]);
+    startIndices[2] = startGID / (gNodesPerDim[1] * gNodesPerDim[0]);
+    const GO rem    = startGID % (gNodesPerDim[1] * gNodesPerDim[0]);
     startIndices[1] = rem / gNodesPerDim[0];
     startIndices[0] = rem % gNodesPerDim[0];
-    endIndices[0] = startIndices[0] + lNodesPerDim[0] - 1;
-    endIndices[1] = startIndices[1] + lNodesPerDim[1] - 1;
-    endIndices[2] = startIndices[2] + lNodesPerDim[2] - 1;
+    endIndices[0]   = startIndices[0] + lNodesPerDim[0] - 1;
+    endIndices[1]   = startIndices[1] + lNodesPerDim[1] - 1;
+    endIndices[2]   = startIndices[2] + lNodesPerDim[2] - 1;
   }
 
   int leftBC = 0, rightBC = 0, frontBC = 0, backBC = 0, bottomBC = 0, topBC = 0;
-  if(startIndices[0] == 0) {leftBC = 1;}
-  if(startIndices[1] == 0) {frontBC = 1;}
-  if(startIndices[2] == 0) {bottomBC = 1;}
+  if (startIndices[0] == 0) {
+    leftBC = 1;
+  }
+  if (startIndices[1] == 0) {
+    frontBC = 1;
+  }
+  if (startIndices[2] == 0) {
+    bottomBC = 1;
+  }
 
-  if(endIndices[0] == gNodesPerDim[0] - 1) {rightBC = 1;}
-  if(endIndices[1] == gNodesPerDim[1] - 1) {backBC = 1;}
-  if(endIndices[2] == gNodesPerDim[2] - 1) {topBC = 1;}
+  if (endIndices[0] == gNodesPerDim[0] - 1) {
+    rightBC = 1;
+  }
+  if (endIndices[1] == gNodesPerDim[1] - 1) {
+    backBC = 1;
+  }
+  if (endIndices[2] == gNodesPerDim[2] - 1) {
+    topBC = 1;
+  }
 
   boundaryConditions.resize(6);
-  boundaryConditions[0] = leftBC  ;
-  boundaryConditions[1] = rightBC ;
-  boundaryConditions[2] = frontBC ;
-  boundaryConditions[3] = backBC  ;
+  boundaryConditions[0] = leftBC;
+  boundaryConditions[1] = rightBC;
+  boundaryConditions[2] = frontBC;
+  boundaryConditions[3] = backBC;
   boundaryConditions[4] = bottomBC;
-  boundaryConditions[5] = topBC   ;
+  boundaryConditions[5] = topBC;
 
   LO numReceive = 0, numSend = 0;
-  Teuchos::Array<GO>  receiveGIDs;
+  Teuchos::Array<GO> receiveGIDs;
   Teuchos::Array<int> receivePIDs;
-  Teuchos::Array<LO>  receiveLIDs, sendLIDs, interfaceLIDs;
+  Teuchos::Array<LO> receiveLIDs, sendLIDs, interfaceLIDs;
 
-  if(numDimensions == 1) {
+  if (numDimensions == 1) {
     maxRegPerGID = 2;
-    if(leftBC == 0) {
+    if (leftBC == 0) {
       numReceive = 1;
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -270,7 +272,7 @@ void createRegionData(const int numDimensions,
       receiveGIDs[0] = startIndices[0] - 1;
       receivePIDs[0] = myRank - 1;
     }
-    if(rightBC == 0) {
+    if (rightBC == 0) {
       numSend = 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
@@ -280,10 +282,10 @@ void createRegionData(const int numDimensions,
       sendGIDs[0] = myRank + 1;
       sendLIDs[0] = lNodesPerDim[0] - 1;
     }
-  } else if(numDimensions == 2) {
+  } else if (numDimensions == 2) {
     maxRegPerGID = 4;
     // Received nodes
-    if(frontBC == 0 && leftBC == 0) {
+    if (frontBC == 0 && leftBC == 0) {
       numReceive = lNodesPerDim[0] + lNodesPerDim[1] + 1;
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -295,31 +297,31 @@ void createRegionData(const int numDimensions,
       receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
       ++countIDs;
       // Receive front edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
         receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i;
         receivePIDs[countIDs] = myRank - procsPerDim[0];
         ++countIDs;
       }
       // Receive left edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - 1 + j*gNodesPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0];
         receivePIDs[countIDs] = myRank - 1;
         ++countIDs;
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 2;
 
         interfaceLIDs.resize((lNodesPerDim[0] + 1) + (lNodesPerDim[1] + 1));
-        for(LO nodeIdx = 0; nodeIdx < lNodesPerDim[0] + 1; ++nodeIdx) {
+        for (LO nodeIdx = 0; nodeIdx < lNodesPerDim[0] + 1; ++nodeIdx) {
           interfaceLIDs[nodeIdx] = nodeIdx;
         }
-        for(LO nodeIdx = 0; nodeIdx < lNodesPerDim[1] + 1; ++nodeIdx) {
-          interfaceLIDs[lNodesPerDim[0] + 1 + nodeIdx] = nodeIdx*(lNodesPerDim[1] + 1);
+        for (LO nodeIdx = 0; nodeIdx < lNodesPerDim[1] + 1; ++nodeIdx) {
+          interfaceLIDs[lNodesPerDim[0] + 1 + nodeIdx] = nodeIdx * (lNodesPerDim[1] + 1);
         }
       }
 
-    } else if(frontBC == 0) {
+    } else if (frontBC == 0) {
       numReceive = lNodesPerDim[0];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -327,22 +329,22 @@ void createRegionData(const int numDimensions,
 
       LO countIDs = 0;
       // Receive front edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
         receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i;
         receivePIDs[countIDs] = myRank - procsPerDim[0];
         ++countIDs;
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 1;
 
         interfaceLIDs.resize(lNodesPerDim[0] + 1);
-        for(LO nodeIdx = 0; nodeIdx < lNodesPerDim[0] + 1; ++nodeIdx) {
+        for (LO nodeIdx = 0; nodeIdx < lNodesPerDim[0] + 1; ++nodeIdx) {
           interfaceLIDs[nodeIdx] = nodeIdx;
         }
       }
 
-    } else if(leftBC == 0) {
+    } else if (leftBC == 0) {
       numReceive = lNodesPerDim[1];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
@@ -350,25 +352,24 @@ void createRegionData(const int numDimensions,
 
       LO countIDs = 0;
       // Receive left edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - 1 + j*gNodesPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0];
         receivePIDs[countIDs] = myRank - 1;
         ++countIDs;
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 1;
 
         interfaceLIDs.resize(lNodesPerDim[1] + 1);
-        for(LO nodeIdx = 0; nodeIdx < lNodesPerDim[1] + 1; ++nodeIdx) {
-          interfaceLIDs[nodeIdx] = nodeIdx*(lNodesPerDim[1] + 1);
+        for (LO nodeIdx = 0; nodeIdx < lNodesPerDim[1] + 1; ++nodeIdx) {
+          interfaceLIDs[nodeIdx] = nodeIdx * (lNodesPerDim[1] + 1);
         }
       }
-
     }
 
     // Sent nodes
-    if(rightBC == 0 && backBC == 0) {
+    if (rightBC == 0 && backBC == 0) {
       numSend = lNodesPerDim[0] + lNodesPerDim[1] + 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
@@ -376,25 +377,25 @@ void createRegionData(const int numDimensions,
 
       LO countIDs = 0;
       // Send nodes of right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + 1;
-        sendLIDs[countIDs] = (j + 1)*lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = (j + 1) * lNodesPerDim[0] - 1;
         ++countIDs;
       }
       // Send nodes of back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
         sendPIDs[countIDs] = myRank + procsPerDim[0];
-        sendLIDs[countIDs] = (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+        sendLIDs[countIDs] = (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
         ++countIDs;
       }
       // Send node of back-right corner
-      sendGIDs[countIDs] = startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
+      sendGIDs[countIDs] = startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
       sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
-      sendLIDs[countIDs] = lNodesPerDim[1]*lNodesPerDim[0] - 1;
+      sendLIDs[countIDs] = lNodesPerDim[1] * lNodesPerDim[0] - 1;
       ++countIDs;
-    } else if(backBC == 0) {
+    } else if (backBC == 0) {
       numSend = lNodesPerDim[0];
 
       sendGIDs.resize(numSend);
@@ -403,13 +404,13 @@ void createRegionData(const int numDimensions,
 
       LO countIDs = 0;
       // Send nodes of back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
         sendPIDs[countIDs] = myRank + procsPerDim[0];
-        sendLIDs[countIDs] = (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+        sendLIDs[countIDs] = (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
         ++countIDs;
       }
-    } else if(rightBC == 0) {
+    } else if (rightBC == 0) {
       numSend = lNodesPerDim[1];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
@@ -417,642 +418,567 @@ void createRegionData(const int numDimensions,
 
       LO countIDs = 0;
       // Send nodes of right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + 1;
-        sendLIDs[countIDs] = (j + 1)*lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = (j + 1) * lNodesPerDim[0] - 1;
         ++countIDs;
       }
     }
-  } else if(numDimensions == 3) {
+  } else if (numDimensions == 3) {
     maxRegPerGID = 8;
     // Received nodes
-    if( (bottomBC == 0) && (frontBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[1]     // bottom face
-        + lNodesPerDim[0]*(lNodesPerDim[2] + 1)        // front face
-        + (lNodesPerDim[1] + 1)*(lNodesPerDim[2] + 1); // left face
+    if ((bottomBC == 0) && (frontBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[1]                 // bottom face
+                   + lNodesPerDim[0] * (lNodesPerDim[2] + 1)         // front face
+                   + (lNodesPerDim[1] + 1) * (lNodesPerDim[2] + 1);  // left face
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-left-bottom corner node
-      receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1
-          - gNodesPerDim[1]*gNodesPerDim[0];
-      receivePIDs[countIDs] = myRank - procsPerDim[0] - 1
-          - procsPerDim[1]*procsPerDim[0];
+      receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1 - gNodesPerDim[1] * gNodesPerDim[0];
+      receivePIDs[countIDs] = myRank - procsPerDim[0] - 1 - procsPerDim[1] * procsPerDim[0];
       ++countIDs;
 
       // Receive front-bottom edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - gNodesPerDim[0] + i;
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - gNodesPerDim[0] + i;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - procsPerDim[0];
         ++countIDs;
       }
 
       // Recieve left-bottom edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - 1 + j*gNodesPerDim[0];
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - 1 + j * gNodesPerDim[0];
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - 1;
         ++countIDs;
         // Recieve bottom face nodes
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              + i
-              + j*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1];
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] + i + j * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1];
           ++countIDs;
         }
       }
 
       // Receive front-left edge nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]
-              - 1 + k*gNodesPerDim[0]*gNodesPerDim[1];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] - 1 + k * gNodesPerDim[0] * gNodesPerDim[1];
         receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
         ++countIDs;
         // Receive front face nodes
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
         // Receive left face nodes
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = startGID - 1
-              + j*gNodesPerDim[0]
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = startGID - 1 + j * gNodesPerDim[0] + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 3;
 
-        interfaceLIDs.resize((lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1)
-                             + (lNodesPerDim[0] + 1)*(lNodesPerDim[2] + 1)
-                             + (lNodesPerDim[1] + 1)*(lNodesPerDim[2] + 1));
+        interfaceLIDs.resize((lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1) + (lNodesPerDim[0] + 1) * (lNodesPerDim[2] + 1) + (lNodesPerDim[1] + 1) * (lNodesPerDim[2] + 1));
 
         LO nodeOffset = 0, nodeIdx, nodeLID;
         // Bottom face
-        for(nodeIdx = 0; nodeIdx < (lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1); ++nodeIdx) {
+        for (nodeIdx = 0; nodeIdx < (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1); ++nodeIdx) {
           interfaceLIDs[nodeIdx] = nodeIdx;
         }
         // Front face
-        nodeOffset += (lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1);
-        for(LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
-          for(LO i = 0; i < lNodesPerDim[0] + 1; ++i) {
-            nodeIdx = k*(lNodesPerDim[0] + 1) + i + nodeOffset;
-            nodeLID = k*(lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1) + i;
+        nodeOffset += (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1);
+        for (LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
+          for (LO i = 0; i < lNodesPerDim[0] + 1; ++i) {
+            nodeIdx                = k * (lNodesPerDim[0] + 1) + i + nodeOffset;
+            nodeLID                = k * (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1) + i;
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
         // Left face
-        nodeOffset += (lNodesPerDim[0] + 1)*(lNodesPerDim[2] + 1);
-        for(LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
-          for(LO j = 0; j < lNodesPerDim[1] + 1; ++j) {
-            nodeIdx = k*(lNodesPerDim[1] + 1) + j + nodeOffset;
-            nodeLID = k*(lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1) + j*(lNodesPerDim[0] + 1);
+        nodeOffset += (lNodesPerDim[0] + 1) * (lNodesPerDim[2] + 1);
+        for (LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
+          for (LO j = 0; j < lNodesPerDim[1] + 1; ++j) {
+            nodeIdx                = k * (lNodesPerDim[1] + 1) + j + nodeOffset;
+            nodeLID                = k * (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1) + j * (lNodesPerDim[0] + 1);
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
       }
 
-    // Two faces received
-    } else if( (bottomBC == 0) && (frontBC == 0) ) {
-      numReceive = lNodesPerDim[0]*(lNodesPerDim[1] + lNodesPerDim[2] + 1);
+      // Two faces received
+    } else if ((bottomBC == 0) && (frontBC == 0)) {
+      numReceive = lNodesPerDim[0] * (lNodesPerDim[1] + lNodesPerDim[2] + 1);
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-bottom edge nodes
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              - gNodesPerDim[0] + i;
-        receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1] - procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] - gNodesPerDim[0] + i;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1] - procsPerDim[0];
         ++countIDs;
       }
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0]*gNodesPerDim[1]
-              + i
-              + j*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[0]*procsPerDim[1];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] * gNodesPerDim[1] + i + j * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[0] * procsPerDim[1];
           ++countIDs;
         }
       }
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i
-              + k*(gNodesPerDim[1]*gNodesPerDim[0]);
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = startGID - gNodesPerDim[0] + i + k * (gNodesPerDim[1] * gNodesPerDim[0]);
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 2;
 
-        interfaceLIDs.resize(lNodesPerDim[0]*(lNodesPerDim[1] + 1)
-                             + lNodesPerDim[0]*(lNodesPerDim[2] + 1));
+        interfaceLIDs.resize(lNodesPerDim[0] * (lNodesPerDim[1] + 1) + lNodesPerDim[0] * (lNodesPerDim[2] + 1));
 
         LO nodeOffset = 0, nodeIdx, nodeLID;
         // Bottom face
-        for(nodeIdx = 0; nodeIdx < lNodesPerDim[0]*(lNodesPerDim[1] + 1); ++nodeIdx) {
+        for (nodeIdx = 0; nodeIdx < lNodesPerDim[0] * (lNodesPerDim[1] + 1); ++nodeIdx) {
           interfaceLIDs[nodeIdx] = nodeIdx;
         }
         // Front face
-        nodeOffset += lNodesPerDim[0]*(lNodesPerDim[1] + 1);
-        for(LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
-          for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-            nodeIdx = k*lNodesPerDim[0] + i + nodeOffset;
-            nodeLID = k*lNodesPerDim[0]*(lNodesPerDim[1] + 1) + i;
+        nodeOffset += lNodesPerDim[0] * (lNodesPerDim[1] + 1);
+        for (LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
+          for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+            nodeIdx                = k * lNodesPerDim[0] + i + nodeOffset;
+            nodeLID                = k * lNodesPerDim[0] * (lNodesPerDim[1] + 1) + i;
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
       }
 
-    } else if( (bottomBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[1]*(lNodesPerDim[0] + lNodesPerDim[2] + 1);
+    } else if ((bottomBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[1] * (lNodesPerDim[0] + lNodesPerDim[2] + 1);
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive left-bottom edge nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        receiveGIDs[countIDs] =  j*gNodesPerDim[0]
-          + startGID - gNodesPerDim[1]*gNodesPerDim[0] - 1;
-        receivePIDs[countIDs] =  myRank - procsPerDim[1]*procsPerDim[0] - 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        receiveGIDs[countIDs] = j * gNodesPerDim[0] + startGID - gNodesPerDim[1] * gNodesPerDim[0] - 1;
+        receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0] - 1;
         ++countIDs;
         // Receive bottom face nodes
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[1]*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[1]*procsPerDim[0];
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID - gNodesPerDim[1] * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Receive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 2;
 
-        interfaceLIDs.resize((lNodesPerDim[0] + 1)*lNodesPerDim[1]
-                             + lNodesPerDim[1]*(lNodesPerDim[2] + 1));
+        interfaceLIDs.resize((lNodesPerDim[0] + 1) * lNodesPerDim[1] + lNodesPerDim[1] * (lNodesPerDim[2] + 1));
 
         LO nodeOffset = 0, nodeIdx, nodeLID;
         // Bottom face
-        for(nodeIdx = 0; nodeIdx < (lNodesPerDim[0] + 1)*lNodesPerDim[1]; ++nodeIdx) {
+        for (nodeIdx = 0; nodeIdx < (lNodesPerDim[0] + 1) * lNodesPerDim[1]; ++nodeIdx) {
           interfaceLIDs[nodeIdx] = nodeIdx;
         }
         // Left face
-        nodeOffset += (lNodesPerDim[0] + 1)*lNodesPerDim[1];
-        for(LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
-          for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-            nodeIdx = k*lNodesPerDim[1] + j + nodeOffset;
-            nodeLID = k*(lNodesPerDim[0] + 1)*lNodesPerDim[1] + j*(lNodesPerDim[0] + 1);
+        nodeOffset += (lNodesPerDim[0] + 1) * lNodesPerDim[1];
+        for (LO k = 0; k < lNodesPerDim[2] + 1; ++k) {
+          for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+            nodeIdx                = k * lNodesPerDim[1] + j + nodeOffset;
+            nodeLID                = k * (lNodesPerDim[0] + 1) * lNodesPerDim[1] + j * (lNodesPerDim[0] + 1);
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
       }
 
-    } else if( (frontBC == 0) && (leftBC == 0) ) {
-      numReceive = lNodesPerDim[2]*(lNodesPerDim[0] + lNodesPerDim[1] + 1);
+    } else if ((frontBC == 0) && (leftBC == 0)) {
+      numReceive = lNodesPerDim[2] * (lNodesPerDim[0] + lNodesPerDim[1] + 1);
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front-left edge nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        receiveGIDs[countIDs] =  k*gNodesPerDim[1]*gNodesPerDim[0]
-          + startGID - gNodesPerDim[0] - 1;
-        receivePIDs[countIDs] =  myRank - procsPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + startGID - gNodesPerDim[0] - 1;
+        receivePIDs[countIDs] = myRank - procsPerDim[0] - 1;
         ++countIDs;
         // Receive front face nodes
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[0];
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID - gNodesPerDim[0];
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
         // Receive left face nodes
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID - 1;
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
 
-      if(useUnstructured) { // Set parameters for interface aggregation
+      if (useUnstructured) {  // Set parameters for interface aggregation
         numInterfaces = 2;
 
-        interfaceLIDs.resize((lNodesPerDim[0] + 1)*lNodesPerDim[2]
-                             + (lNodesPerDim[1] + 1)*lNodesPerDim[2]);
+        interfaceLIDs.resize((lNodesPerDim[0] + 1) * lNodesPerDim[2] + (lNodesPerDim[1] + 1) * lNodesPerDim[2]);
 
         LO nodeOffset = 0, nodeIdx, nodeLID;
         // Front face
-        for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-          for(LO i = 0; i < lNodesPerDim[0] + 1; ++i) {
-            nodeIdx = k*(lNodesPerDim[0] + 1) + i;
-            nodeLID = k*(lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1) + i;
+        for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+          for (LO i = 0; i < lNodesPerDim[0] + 1; ++i) {
+            nodeIdx                = k * (lNodesPerDim[0] + 1) + i;
+            nodeLID                = k * (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1) + i;
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
         // Left face
-        nodeOffset += (lNodesPerDim[0] + 1)*lNodesPerDim[2];
-        for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-          for(LO j = 0; j < lNodesPerDim[1] + 1; ++j) {
-            nodeIdx = k*(lNodesPerDim[1] + 1) + j + nodeOffset;
-            nodeLID = k*(lNodesPerDim[0] + 1)*(lNodesPerDim[1] + 1) + j*(lNodesPerDim[0] + 1);
+        nodeOffset += (lNodesPerDim[0] + 1) * lNodesPerDim[2];
+        for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+          for (LO j = 0; j < lNodesPerDim[1] + 1; ++j) {
+            nodeIdx                = k * (lNodesPerDim[1] + 1) + j + nodeOffset;
+            nodeLID                = k * (lNodesPerDim[0] + 1) * (lNodesPerDim[1] + 1) + j * (lNodesPerDim[0] + 1);
             interfaceLIDs[nodeIdx] = nodeLID;
           }
         }
       }
 
-    // Single face received
-    } else if(bottomBC == 0) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[1];
+      // Single face received
+    } else if (bottomBC == 0) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[1];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive bottom face nodes
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[1]*gNodesPerDim[0];
-          receivePIDs[countIDs] = myRank - procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID - gNodesPerDim[1] * gNodesPerDim[0];
+          receivePIDs[countIDs] = myRank - procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(frontBC == 0) {
-      numReceive = lNodesPerDim[0]*lNodesPerDim[2];
+    } else if (frontBC == 0) {
+      numReceive = lNodesPerDim[0] * lNodesPerDim[2];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Receive front face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID - gNodesPerDim[0];
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID - gNodesPerDim[0];
           receivePIDs[countIDs] = myRank - procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(leftBC == 0) {
-      numReceive = lNodesPerDim[1]*lNodesPerDim[2];
+    } else if (leftBC == 0) {
+      numReceive = lNodesPerDim[1] * lNodesPerDim[2];
       receiveGIDs.resize(numReceive);
       receivePIDs.resize(numReceive);
       receiveLIDs.resize(numReceive);
 
       LO countIDs = 0;
       // Recive left face nodes
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          receiveGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + j*gNodesPerDim[0] + startGID - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          receiveGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID - 1;
           receivePIDs[countIDs] = myRank - 1;
           ++countIDs;
         }
       }
-
     }
 
     // Sent nodes
-    if( (topBC == 0) && (backBC == 0) && (rightBC == 0) ) {
-      numSend = (lNodesPerDim[0])*(lNodesPerDim[1])
-        + (lNodesPerDim[0])*(lNodesPerDim[2])
-        + (lNodesPerDim[1])*(lNodesPerDim[2])
-        + lNodesPerDim[0]
-        + lNodesPerDim[1]
-        + lNodesPerDim[2]
-        + 1;
+    if ((topBC == 0) && (backBC == 0) && (rightBC == 0)) {
+      numSend = (lNodesPerDim[0]) * (lNodesPerDim[1]) + (lNodesPerDim[0]) * (lNodesPerDim[2]) + (lNodesPerDim[1]) * (lNodesPerDim[2]) + lNodesPerDim[0] + lNodesPerDim[1] + lNodesPerDim[2] + 1;
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-            + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
-          sendLIDs[countIDs] = k*(lNodesPerDim[1]*lNodesPerDim[0])
-            + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+          sendLIDs[countIDs] = k * (lNodesPerDim[1] * lNodesPerDim[0]) + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0]) + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-          sendLIDs[countIDs] = k*(lNodesPerDim[1]*lNodesPerDim[0])
-            + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+          sendLIDs[countIDs] = k * (lNodesPerDim[1] * lNodesPerDim[0]) + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of right-back edge
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-          + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendLIDs[countIDs] = k*(lNodesPerDim[1]*lNodesPerDim[0])
-          + lNodesPerDim[1]*lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = k * (lNodesPerDim[1] * lNodesPerDim[0]) + lNodesPerDim[1] * lNodesPerDim[0] - 1;
         sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
         ++countIDs;
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + i;
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i;
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0]
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*(lNodesPerDim[1]*lNodesPerDim[0])
-          + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * (lNodesPerDim[1] * lNodesPerDim[0]) + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + 1;
         ++countIDs;
       }
       // Send nodes of top-back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-          + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-        sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*(lNodesPerDim[1]*lNodesPerDim[0])
-          + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+        sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * (lNodesPerDim[1] * lNodesPerDim[0]) + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[0];
         ++countIDs;
       }
       // Send node of top-back-right corner
-      sendGIDs[countIDs] = startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-        + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-      sendLIDs[countIDs] = lNodesPerDim[2]*lNodesPerDim[1]*lNodesPerDim[0] - 1;
-      sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[0] + 1;
+      sendGIDs[countIDs] = startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+      sendLIDs[countIDs] = lNodesPerDim[2] * lNodesPerDim[1] * lNodesPerDim[0] - 1;
+      sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[0] + 1;
       ++countIDs;
 
-    } else if( (topBC == 0) && (backBC == 0) ) {
-      numSend = (lNodesPerDim[0]*lNodesPerDim[2]) // back face
-              + (lNodesPerDim[0]*lNodesPerDim[1]) // Top face
-              + (lNodesPerDim[0]); // top-back edge
+    } else if ((topBC == 0) && (backBC == 0)) {
+      numSend = (lNodesPerDim[0] * lNodesPerDim[2])    // back face
+                + (lNodesPerDim[0] * lNodesPerDim[1])  // Top face
+                + (lNodesPerDim[0]);                   // top-back edge
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0]) + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-          sendLIDs[countIDs] = k*lNodesPerDim[1]*lNodesPerDim[0]
-            + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+          sendLIDs[countIDs] = k * lNodesPerDim[1] * lNodesPerDim[0] + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + i;
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i;
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-back edge
-      for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-        sendGIDs[countIDs] = i
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0]
-          + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-        sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-          + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + procsPerDim[0];
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        sendGIDs[countIDs] = i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+        sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + procsPerDim[0];
         ++countIDs;
       }
 
-    } else if( (topBC == 0) && (rightBC == 0) ) {
-      numSend = (lNodesPerDim[1]*lNodesPerDim[2]) // right face
-              + (lNodesPerDim[0]*lNodesPerDim[1]) // Top face
-              + (lNodesPerDim[1]); // top-right edge
+    } else if ((topBC == 0) && (rightBC == 0)) {
+      numSend = (lNodesPerDim[1] * lNodesPerDim[2])    // right face
+                + (lNodesPerDim[0] * lNodesPerDim[1])  // Top face
+                + (lNodesPerDim[1]);                   // top-right edge
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*(gNodesPerDim[1]*gNodesPerDim[0])
-            + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
-          sendLIDs[countIDs] = k*(lNodesPerDim[1]*lNodesPerDim[0])
-            + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * (gNodesPerDim[1] * gNodesPerDim[0]) + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+          sendLIDs[countIDs] = k * (lNodesPerDim[1] * lNodesPerDim[0]) + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0];
-          sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + i;
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0];
+          sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i;
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of top-right edge
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        sendGIDs[countIDs] = j*gNodesPerDim[0]
-          + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[1]*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-          + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
-        sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0] + 1;
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        sendGIDs[countIDs] = j * gNodesPerDim[0] + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[1] * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0] + 1;
         ++countIDs;
       }
 
-    } else if( (backBC == 0) && (rightBC == 0) ) {
-      numSend = lNodesPerDim[2]*(lNodesPerDim[0] + lNodesPerDim[1] + 1);
+    } else if ((backBC == 0) && (rightBC == 0)) {
+      numSend = lNodesPerDim[2] * (lNodesPerDim[0] + lNodesPerDim[1] + 1);
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0]
-            + startGID + lNodesPerDim[0] - 1;
-          sendLIDs[countIDs] = k*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+          sendLIDs[countIDs] = k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-          sendLIDs[countIDs] = k*lNodesPerDim[1]*lNodesPerDim[0]
-            + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+          sendLIDs[countIDs] = k * lNodesPerDim[1] * lNodesPerDim[0] + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
       // Send nodes of back-right edge
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0] + lNodesPerDim[0] - 1;
-          sendLIDs[countIDs] = k*(lNodesPerDim[1]*lNodesPerDim[0])
-            + lNodesPerDim[1]*lNodesPerDim[0] - 1;
-          sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
-          ++countIDs;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0] + lNodesPerDim[0] - 1;
+        sendLIDs[countIDs] = k * (lNodesPerDim[1] * lNodesPerDim[0]) + lNodesPerDim[1] * lNodesPerDim[0] - 1;
+        sendPIDs[countIDs] = myRank + procsPerDim[0] + 1;
+        ++countIDs;
       }
 
-    } else if(topBC == 0) {
-      numSend = lNodesPerDim[0]*lNodesPerDim[1];
+    } else if (topBC == 0) {
+      numSend = lNodesPerDim[0] * lNodesPerDim[1];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of top face
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = j*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[2] - 1)*gNodesPerDim[0]*gNodesPerDim[1];
-          sendLIDs[countIDs] = (lNodesPerDim[2] - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + i;
-          sendPIDs[countIDs] = myRank + procsPerDim[1]*procsPerDim[0];
+      for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = j * gNodesPerDim[0] + i + startGID + (lNodesPerDim[2] - 1) * gNodesPerDim[0] * gNodesPerDim[1];
+          sendLIDs[countIDs] = (lNodesPerDim[2] - 1) * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i;
+          sendPIDs[countIDs] = myRank + procsPerDim[1] * procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(backBC == 0) {
-      numSend = lNodesPerDim[0]*lNodesPerDim[2];
+    } else if (backBC == 0) {
+      numSend = lNodesPerDim[0] * lNodesPerDim[2];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of back face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0] + i
-            + startGID + (lNodesPerDim[1] - 1)*gNodesPerDim[0];
-          sendLIDs[countIDs] = k*lNodesPerDim[1]*lNodesPerDim[0]
-            + (lNodesPerDim[1] - 1)*lNodesPerDim[0] + i;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + i + startGID + (lNodesPerDim[1] - 1) * gNodesPerDim[0];
+          sendLIDs[countIDs] = k * lNodesPerDim[1] * lNodesPerDim[0] + (lNodesPerDim[1] - 1) * lNodesPerDim[0] + i;
           sendPIDs[countIDs] = myRank + procsPerDim[0];
           ++countIDs;
         }
       }
 
-    } else if(rightBC == 0) {
-      numSend = lNodesPerDim[1]*lNodesPerDim[2];
+    } else if (rightBC == 0) {
+      numSend = lNodesPerDim[1] * lNodesPerDim[2];
       sendGIDs.resize(numSend);
       sendPIDs.resize(numSend);
       sendLIDs.resize(numSend);
 
       LO countIDs = 0;
       // Send nodes of right face
-      for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-          sendGIDs[countIDs] = k*gNodesPerDim[1]*gNodesPerDim[0]
-            + j*gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
-          sendLIDs[countIDs] = k*lNodesPerDim[1]*lNodesPerDim[0]
-            + j*lNodesPerDim[0] + lNodesPerDim[0] - 1;
+      for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+          sendGIDs[countIDs] = k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + startGID + lNodesPerDim[0] - 1;
+          sendLIDs[countIDs] = k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + lNodesPerDim[0] - 1;
           sendPIDs[countIDs] = myRank + 1;
           ++countIDs;
         }
       }
-
     }
   }
 
-
   // The code below could probably be separated and performed in its own function.
   // It pretty much tries to go from the more geometric based information generated
   // above to the quasiRegion map which means finding the GID of nodes-off rank on the interface.
-  const LO numLocalCompositeNodes = lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2];
-  numLocalRegionNodes = numLocalCompositeNodes + numReceive;
-  quasiRegionGIDs.resize(numLocalRegionNodes*numDofsPerNode);
+  const LO numLocalCompositeNodes = lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2];
+  numLocalRegionNodes             = numLocalCompositeNodes + numReceive;
+  quasiRegionGIDs.resize(numLocalRegionNodes * numDofsPerNode);
   quasiRegionCoordGIDs.resize(numLocalRegionNodes);
 
   rNodesPerDim[0] = lNodesPerDim[0];
   rNodesPerDim[1] = lNodesPerDim[1];
   rNodesPerDim[2] = lNodesPerDim[2];
-  if(leftBC   == 0) {rNodesPerDim[0] += 1;}
-  if(frontBC  == 0) {rNodesPerDim[1] += 1;}
-  if(bottomBC == 0) {rNodesPerDim[2] += 1;}
+  if (leftBC == 0) {
+    rNodesPerDim[0] += 1;
+  }
+  if (frontBC == 0) {
+    rNodesPerDim[1] += 1;
+  }
+  if (bottomBC == 0) {
+    rNodesPerDim[2] += 1;
+  }
 
   // Using receiveGIDs, rNodesPerDim and numLocalRegionNodes, build quasi-region row map
   // This will potentially be done by the application or in a MueLu interface but for now
   // let us keep it in this utility function.
   LO interfaceCount = 0, compositeIdx = 0;
   Teuchos::Array<LO> regionIJK(3);
-  for(LO nodeRegionIdx = 0; nodeRegionIdx < numLocalRegionNodes; ++nodeRegionIdx) {
-    regionIJK[2] = nodeRegionIdx / (rNodesPerDim[1]*rNodesPerDim[0]);
-    LO tmp       = nodeRegionIdx % (rNodesPerDim[1]*rNodesPerDim[0]);
+  for (LO nodeRegionIdx = 0; nodeRegionIdx < numLocalRegionNodes; ++nodeRegionIdx) {
+    regionIJK[2] = nodeRegionIdx / (rNodesPerDim[1] * rNodesPerDim[0]);
+    LO tmp       = nodeRegionIdx % (rNodesPerDim[1] * rNodesPerDim[0]);
     regionIJK[1] = tmp / rNodesPerDim[0];
     regionIJK[0] = tmp % rNodesPerDim[0];
 
-    if( (regionIJK[0] == 0 && leftBC   == 0) ||
-        (regionIJK[1] == 0 && frontBC  == 0) ||
-        (regionIJK[2] == 0 && bottomBC == 0) ) {
+    if ((regionIJK[0] == 0 && leftBC == 0) ||
+        (regionIJK[1] == 0 && frontBC == 0) ||
+        (regionIJK[2] == 0 && bottomBC == 0)) {
       quasiRegionCoordGIDs[nodeRegionIdx] = receiveGIDs[interfaceCount];
-      for(int dof = 0; dof < numDofsPerNode; ++dof) {
-        quasiRegionGIDs[nodeRegionIdx*numDofsPerNode + dof] =
-          receiveGIDs[interfaceCount]*numDofsPerNode + dof;
+      for (int dof = 0; dof < numDofsPerNode; ++dof) {
+        quasiRegionGIDs[nodeRegionIdx * numDofsPerNode + dof] =
+            receiveGIDs[interfaceCount] * numDofsPerNode + dof;
       }
       receiveLIDs[interfaceCount] = nodeRegionIdx;
       ++interfaceCount;
     } else {
-      compositeIdx = (regionIJK[2] + bottomBC - 1)*lNodesPerDim[1]*lNodesPerDim[0]
-        + (regionIJK[1] + frontBC  - 1)*lNodesPerDim[0]
-        + (regionIJK[0] + leftBC - 1);
+      compositeIdx                        = (regionIJK[2] + bottomBC - 1) * lNodesPerDim[1] * lNodesPerDim[0] + (regionIJK[1] + frontBC - 1) * lNodesPerDim[0] + (regionIJK[0] + leftBC - 1);
       quasiRegionCoordGIDs[nodeRegionIdx] = nodeMap->getGlobalElement(compositeIdx);
-      for(int dof = 0; dof < numDofsPerNode; ++dof) {
-        quasiRegionGIDs[nodeRegionIdx*numDofsPerNode + dof]
-          = dofMap->getGlobalElement(compositeIdx*numDofsPerNode + dof);
-        compositeToRegionLIDs[compositeIdx*numDofsPerNode + dof] = nodeRegionIdx*numDofsPerNode + dof;
+      for (int dof = 0; dof < numDofsPerNode; ++dof) {
+        quasiRegionGIDs[nodeRegionIdx * numDofsPerNode + dof]      = dofMap->getGlobalElement(compositeIdx * numDofsPerNode + dof);
+        compositeToRegionLIDs[compositeIdx * numDofsPerNode + dof] = nodeRegionIdx * numDofsPerNode + dof;
       }
     }
   }
@@ -1062,17 +988,17 @@ void createRegionData(const int numDimensions,
   interfaceLIDsData.resize((sendGIDs.size() + receiveGIDs.size()) * numDofsPerNode);
   interfaceGIDs.resize((sendGIDs.size() + receiveGIDs.size()) * numDofsPerNode);
   using size_type = typename Teuchos::Array<GO>::size_type;
-  for(size_type nodeIdx = 0; nodeIdx < sendGIDs.size(); ++nodeIdx) {
-    for(int dof = 0; dof < numDofsPerNode; ++dof) {
-      LO dofIdx = nodeIdx*numDofsPerNode + dof;
-      interfaceGIDs[dofIdx] = sendGIDs[nodeIdx] * numDofsPerNode + dof;
+  for (size_type nodeIdx = 0; nodeIdx < sendGIDs.size(); ++nodeIdx) {
+    for (int dof = 0; dof < numDofsPerNode; ++dof) {
+      LO dofIdx                 = nodeIdx * numDofsPerNode + dof;
+      interfaceGIDs[dofIdx]     = sendGIDs[nodeIdx] * numDofsPerNode + dof;
       interfaceLIDsData[dofIdx] = compositeToRegionLIDs[sendLIDs[nodeIdx] * numDofsPerNode + dof];
     }
   }
-  for(size_type nodeIdx = 0; nodeIdx < receiveGIDs.size(); ++nodeIdx) {
-    for(int dof = 0; dof < numDofsPerNode; ++dof) {
-      LO dofIdx = nodeIdx*numDofsPerNode + dof;
-      interfaceGIDs[dofIdx + sendGIDs.size() * numDofsPerNode] = receiveGIDs[nodeIdx] * numDofsPerNode + dof;
+  for (size_type nodeIdx = 0; nodeIdx < receiveGIDs.size(); ++nodeIdx) {
+    for (int dof = 0; dof < numDofsPerNode; ++dof) {
+      LO dofIdx                                                    = nodeIdx * numDofsPerNode + dof;
+      interfaceGIDs[dofIdx + sendGIDs.size() * numDofsPerNode]     = receiveGIDs[nodeIdx] * numDofsPerNode + dof;
       interfaceLIDsData[dofIdx + sendLIDs.size() * numDofsPerNode] = receiveLIDs[nodeIdx] * numDofsPerNode + dof;
     }
   }
@@ -1087,7 +1013,7 @@ void createRegionData(const int numDimensions,
   interfaceGIDs.erase(std::unique(interfaceGIDs.begin(), interfaceGIDs.end()),
                       interfaceGIDs.end());
 
-} // createRegionData
+}  // createRegionData
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
 void MakeRegionPerGIDWithGhosts(const Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& nodeMap,
@@ -1095,7 +1021,7 @@ void MakeRegionPerGIDWithGhosts(const Teuchos::RCP<Xpetra::Map<LocalOrdinal, Glo
                                 const Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
                                 const int maxRegPerGID,
                                 const LocalOrdinal numDofsPerNode,
-                                const Teuchos::Array<LocalOrdinal>&  lNodesPerDir,
+                                const Teuchos::Array<LocalOrdinal>& lNodesPerDir,
                                 const Teuchos::Array<GlobalOrdinal>& sendGIDs,
                                 const Teuchos::Array<int>& sendPIDs,
                                 const Teuchos::Array<LocalOrdinal>& interfaceRegionLIDs,
@@ -1104,41 +1030,41 @@ void MakeRegionPerGIDWithGhosts(const Teuchos::RCP<Xpetra::Map<LocalOrdinal, Glo
   using LO = LocalOrdinal;
   using GO = GlobalOrdinal;
   using NO = Node;
-  using Teuchos::RCP;
   using Teuchos::Array;
   using Teuchos::ArrayRCP;
+  using Teuchos::RCP;
 
-  const RCP<const Xpetra::Map<LO,GO,NO> > dofMap       = rowImport->getSourceMap();
-  const RCP<const Xpetra::Map<LO,GO,NO> > quasiRegionRowMap = rowImport->getTargetMap();
-  const int myRank = dofMap->getComm()->getRank();
+  const RCP<const Xpetra::Map<LO, GO, NO> > dofMap            = rowImport->getSourceMap();
+  const RCP<const Xpetra::Map<LO, GO, NO> > quasiRegionRowMap = rowImport->getTargetMap();
+  const int myRank                                            = dofMap->getComm()->getRank();
 
-  RCP<Xpetra::MultiVector<LO, LO, GO, NO> >regionsPerGID =
-    Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(dofMap, maxRegPerGID, false);
+  RCP<Xpetra::MultiVector<LO, LO, GO, NO> > regionsPerGID =
+      Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(dofMap, maxRegPerGID, false);
   regionsPerGIDWithGhosts =
-    Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(quasiRegionRowMap, maxRegPerGID, false);
+      Xpetra::MultiVectorFactory<LO, LO, GO, NO>::Build(quasiRegionRowMap, maxRegPerGID, false);
 
-  { // Scope for regionsPerGIDView
+  {  // Scope for regionsPerGIDView
     Array<ArrayRCP<LO> > regionsPerGIDView(maxRegPerGID);
-    for(int regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
+    for (int regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
       regionsPerGIDView[regionIdx] = regionsPerGID->getDataNonConst(regionIdx);
     }
 
     // Initialize all entries to myRank in first column and to -1 in other columns
-    for(LO dofIdx = 0; dofIdx < lNodesPerDir[0]*lNodesPerDir[1]*lNodesPerDir[2]*numDofsPerNode; ++dofIdx) {
+    for (LO dofIdx = 0; dofIdx < lNodesPerDir[0] * lNodesPerDir[1] * lNodesPerDir[2] * numDofsPerNode; ++dofIdx) {
       regionsPerGIDView[0][dofIdx] = myRank;
-      for(int regionIdx = 1; regionIdx < maxRegPerGID; ++regionIdx) {
+      for (int regionIdx = 1; regionIdx < maxRegPerGID; ++regionIdx) {
         regionsPerGIDView[regionIdx][dofIdx] = -1;
       }
     }
 
     // Now loop over the sendGIDs array to fill entries with values in sendPIDs
     LO nodeIdx = 0;
-    for(LO sendIdx = 0; sendIdx < static_cast<LO>(sendPIDs.size()); ++sendIdx) {
+    for (LO sendIdx = 0; sendIdx < static_cast<LO>(sendPIDs.size()); ++sendIdx) {
       nodeIdx = nodeMap->getLocalElement(sendGIDs[sendIdx]);
-      for(int dof = 0; dof < numDofsPerNode; ++dof) {
-        LO dofIdx = nodeIdx*numDofsPerNode + dof;
-        for(int regionIdx = 1; regionIdx < maxRegPerGID; ++regionIdx) {
-          if(regionsPerGIDView[regionIdx][dofIdx] == -1) {
+      for (int dof = 0; dof < numDofsPerNode; ++dof) {
+        LO dofIdx = nodeIdx * numDofsPerNode + dof;
+        for (int regionIdx = 1; regionIdx < maxRegPerGID; ++regionIdx) {
+          if (regionsPerGIDView[regionIdx][dofIdx] == -1) {
             regionsPerGIDView[regionIdx][dofIdx] = sendPIDs[sendIdx];
             break;
           }
@@ -1152,23 +1078,22 @@ void MakeRegionPerGIDWithGhosts(const Teuchos::RCP<Xpetra::Map<LocalOrdinal, Glo
   interfaceGIDsMV = Xpetra::MultiVectorFactory<GO, LO, GO, NO>::Build(quasiRegionRowMap, maxRegPerGID, false);
   interfaceGIDsMV->putScalar(Teuchos::OrdinalTraits<GO>::zero());
   const LO numRegionInterfaceLIDs = static_cast<LO>(interfaceRegionLIDs.size());
-  { // Scope for interfaceGIDsPerRegion
+  {  // Scope for interfaceGIDsPerRegion
     Array<ArrayRCP<LO> > regionsPerGIDWithGhostsData(maxRegPerGID);
     Array<ArrayRCP<GO> > interfaceGIDsMVData(maxRegPerGID);
-    for(int regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
+    for (int regionIdx = 0; regionIdx < maxRegPerGID; ++regionIdx) {
       regionsPerGIDWithGhostsData[regionIdx] = regionsPerGIDWithGhosts->getDataNonConst(regionIdx);
-      interfaceGIDsMVData[regionIdx] = interfaceGIDsMV->getDataNonConst(regionIdx);
-      for(LO idx = 0; idx < numRegionInterfaceLIDs; ++idx) {
+      interfaceGIDsMVData[regionIdx]         = interfaceGIDsMV->getDataNonConst(regionIdx);
+      for (LO idx = 0; idx < numRegionInterfaceLIDs; ++idx) {
         LO LID = interfaceRegionLIDs[idx];
-        if(regionsPerGIDWithGhostsData[regionIdx][LID] == myRank) {
+        if (regionsPerGIDWithGhostsData[regionIdx][LID] == myRank) {
           interfaceGIDsMVData[regionIdx][LID] = regionRowMap->getGlobalElement(LID);
         }
       }
     }
-
   }
 
-} // MakeRegionPerGIDWithGhosts
+}  // MakeRegionPerGIDWithGhosts
 
 /*!
 \brief Extract list of region GIDs of all interface DOFs from the region row map
@@ -1179,13 +1104,12 @@ refert to interface DOFs, so we can grab them and stick them into the list of \c
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
 void ExtractListOfInterfaceRegionGIDs(
     Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > regionRowMap,
-    const Teuchos::Array<LocalOrdinal>& interfaceRegionLIDs, Teuchos::Array<GlobalOrdinal>& interfaceRegionGIDs)
-{
+    const Teuchos::Array<LocalOrdinal>& interfaceRegionLIDs, Teuchos::Array<GlobalOrdinal>& interfaceRegionGIDs) {
   interfaceRegionGIDs.resize(interfaceRegionLIDs.size());
-  for(LocalOrdinal interfaceIdx = 0; interfaceIdx < static_cast<LocalOrdinal>(interfaceRegionLIDs.size()); ++interfaceIdx) {
+  for (LocalOrdinal interfaceIdx = 0; interfaceIdx < static_cast<LocalOrdinal>(interfaceRegionLIDs.size()); ++interfaceIdx) {
     interfaceRegionGIDs[interfaceIdx] =
-      regionRowMap->getGlobalElement(interfaceRegionLIDs[interfaceIdx]);
+        regionRowMap->getGlobalElement(interfaceRegionLIDs[interfaceIdx]);
   }
-} // ExtractListOfInterfaceRegionGIDs
+}  // ExtractListOfInterfaceRegionGIDs
 
-#endif // MUELU_SETUPREGIONUTILITIES_HPP
+#endif  // MUELU_SETUPREGIONUTILITIES_HPP
diff --git a/packages/muelu/research/regionMG/src/SetupRegionVector_def.hpp b/packages/muelu/research/regionMG/src/SetupRegionVector_def.hpp
index a935f62922cc..e3dc90fcec28 100644
--- a/packages/muelu/research/regionMG/src/SetupRegionVector_def.hpp
+++ b/packages/muelu/research/regionMG/src/SetupRegionVector_def.hpp
@@ -62,9 +62,9 @@
 #include <Xpetra_MultiVector.hpp>
 #include <Xpetra_Vector.hpp>
 
-using Teuchos::RCP;
-using Teuchos::ArrayRCP;
 using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::RCP;
 
 /*! \brief Transform composite vector to regional layout
  *
@@ -72,14 +72,13 @@ using Teuchos::Array;
  *  1. import it into an auxiliary vector in the quasiRegional layout
  *  2. replace the quasiRegional map of the auxiliary vector with the regional map
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void compositeToRegional(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec, ///< Vector in composite layout [in]
-                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegVecs, ///< Vector in quasiRegional layout [in/out]
-                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVecs, ///< Vector in regional layout [in/out]
-                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap, ///< revised row maps in region layout [in]
-                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport ///< row importer in region layout [in]
-                         )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void compositeToRegional(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec,         ///< Vector in composite layout [in]
+                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegVecs,   ///< Vector in quasiRegional layout [in/out]
+                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVecs,        ///< Vector in regional layout [in/out]
+                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,  ///< revised row maps in region layout [in]
+                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport          ///< row importer in region layout [in]
+) {
 #include "Xpetra_UseShortNames.hpp"
 
   // quasiRegional layout
@@ -90,12 +89,12 @@ void compositeToRegional(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal,
 
   // regional layout
   // create regVecs vector (copy from quasiRegVecs and swap the map)
-  regVecs = quasiRegVecs; // assignment operator= does deep copy in Xpetra
+  regVecs = quasiRegVecs;  // assignment operator= does deep copy in Xpetra
   TEUCHOS_ASSERT(!regVecs.is_null());
   regVecs->replaceMap(revisedRowMap);
 
   return;
-} // compositeToRegional
+}  // compositeToRegional
 
 /*! \brief Transform composite vector to regional layout
  *
@@ -103,14 +102,13 @@ void compositeToRegional(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal,
  *  1. import it into an auxiliary vector in the quasiRegional layout
  *  2. replace the quasiRegional map of the auxiliary vector with the regional map
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void compositeToRegional(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec, ///< Vector in composite layout [in]
-                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegVecs, ///< Vector in quasiRegional layout [in/out]
-                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVecs, ///< Vector in regional layout [in/out]
-                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap, ///< revised row maps in region layout [in]
-                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport ///< row importer in region layout [in]
-                         )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void compositeToRegional(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec,        ///< Vector in composite layout [in]
+                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& quasiRegVecs,  ///< Vector in quasiRegional layout [in/out]
+                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVecs,       ///< Vector in regional layout [in/out]
+                         const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,      ///< revised row maps in region layout [in]
+                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport              ///< row importer in region layout [in]
+) {
 #include "Xpetra_UseShortNames.hpp"
 
   // quasiRegional layout
@@ -121,13 +119,12 @@ void compositeToRegional(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrd
 
   // regional layout
   // create regVecs vector (copy from quasiRegVecs and swap the map)
-  regVecs = quasiRegVecs; // assignment operator= does deep copy in Xpetra
+  regVecs = quasiRegVecs;  // assignment operator= does deep copy in Xpetra
   TEUCHOS_ASSERT(!regVecs.is_null());
   regVecs->replaceMap(revisedRowMap);
 
   return;
-} // compositeToRegional
-
+}  // compositeToRegional
 
 /*! \brief Transform regional vector to composite layout
  *
@@ -139,12 +136,11 @@ void compositeToRegional(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrd
  *  \note We also need the capability to add processor-local values. This is not supported by
  *  available CombineMode options in Xpetra/Tpetra, so we use a manual implementation here.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void regionalToComposite(const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec, ///< Vector in region layout [in]
-                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec, ///< Vector in composite layout [in/out]
-                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport ///< row importer in region layout [in]
-                         )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void regionalToComposite(const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,  ///< Vector in region layout [in]
+                         RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec,        ///< Vector in composite layout [in/out]
+                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport         ///< row importer in region layout [in]
+) {
   /* Let's fake an ADD combine mode that also adds local values by
    * 1. exporting quasiRegional vectors to auxiliary composite vectors (1 per group)
    * 2. add all auxiliary vectors together
@@ -163,34 +159,33 @@ void regionalToComposite(const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOr
 
   {
     RCP<Vector> quasiRegVec;
-      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 2 - quasiRegVec")));
-      quasiRegVec = regVec;
-      TEUCHOS_ASSERT(Teuchos::nonnull(quasiRegVec));
-      quasiRegVec->replaceMap(rowImport->getTargetMap());
+    tm          = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 2 - quasiRegVec")));
+    quasiRegVec = regVec;
+    TEUCHOS_ASSERT(Teuchos::nonnull(quasiRegVec));
+    quasiRegVec->replaceMap(rowImport->getTargetMap());
 
-      tm = Teuchos::null;
-      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 3 - partialCompVec")));
+    tm = Teuchos::null;
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 3 - partialCompVec")));
 
-      RCP<Vector> partialCompVec = VectorFactory::Build(rowImport->getSourceMap(), true);
-      TEUCHOS_ASSERT(Teuchos::nonnull(partialCompVec));
-      TEUCHOS_ASSERT(partialCompVec->getLocalLength() == compVecLocalLength);
-      partialCompVec->doExport(*quasiRegVec, *(rowImport), Xpetra::ADD);
+    RCP<Vector> partialCompVec = VectorFactory::Build(rowImport->getSourceMap(), true);
+    TEUCHOS_ASSERT(Teuchos::nonnull(partialCompVec));
+    TEUCHOS_ASSERT(partialCompVec->getLocalLength() == compVecLocalLength);
+    partialCompVec->doExport(*quasiRegVec, *(rowImport), Xpetra::ADD);
 
-      tm = Teuchos::null;
-      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 4 - compVec->sumIntoLocalValue")));
+    tm = Teuchos::null;
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 4 - compVec->sumIntoLocalValue")));
 
-      ArrayRCP<const SC> partialCompVecData = partialCompVec->getData(0);
-      ArrayRCP<SC>       compVecData        = compVec->getDataNonConst(0);
-      for(size_t entryIdx = 0; entryIdx < compVecLocalLength; ++entryIdx) {
-        compVecData[entryIdx] += partialCompVecData[entryIdx];
-      }
+    ArrayRCP<const SC> partialCompVecData = partialCompVec->getData(0);
+    ArrayRCP<SC> compVecData              = compVec->getDataNonConst(0);
+    for (size_t entryIdx = 0; entryIdx < compVecLocalLength; ++entryIdx) {
+      compVecData[entryIdx] += partialCompVecData[entryIdx];
+    }
 
-      tm = Teuchos::null;
+    tm = Teuchos::null;
   }
 
   return;
-} // regionalToComposite
-
+}  // regionalToComposite
 
 /*! \brief Transform regional vector to composite layout
  *
@@ -202,12 +197,11 @@ void regionalToComposite(const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOr
  *  \note We also need the capability to add processor-local values. This is not supported by
  *  available CombineMode options in Xpetra/Tpetra, so we use a manual implementation here.
  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec, ///< Vector in region layout [in]
-                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec, ///< Vector in composite layout [in/out]
-                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport ///< row importer in region layout [in]
-                         )
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,  ///< Vector in region layout [in]
+                         RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > compVec,        ///< Vector in composite layout [in/out]
+                         const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport              ///< row importer in region layout [in]
+) {
   /* Let's fake an ADD combine mode that also adds local values by
    * 1. exporting quasiRegional vectors to auxiliary composite vectors (1 per group)
    * 2. add all auxiliary vectors together
@@ -226,7 +220,7 @@ void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, Glo
 
   {
     RCP<MultiVector> quasiRegVec;
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 2 - quasiRegVec")));
+    tm          = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 2 - quasiRegVec")));
     quasiRegVec = regVec;
     TEUCHOS_ASSERT(Teuchos::nonnull(quasiRegVec));
     quasiRegVec->replaceMap(rowImport->getTargetMap());
@@ -234,8 +228,7 @@ void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, Glo
     tm = Teuchos::null;
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 3 - partialCompVec")));
 
-    RCP<MultiVector> partialCompVec
-      = MultiVectorFactory::Build(rowImport->getSourceMap(), quasiRegVec->getNumVectors(), true);
+    RCP<MultiVector> partialCompVec = MultiVectorFactory::Build(rowImport->getSourceMap(), quasiRegVec->getNumVectors(), true);
     TEUCHOS_ASSERT(Teuchos::nonnull(partialCompVec));
     TEUCHOS_ASSERT(partialCompVec->getLocalLength() == compVecLocalLength);
     partialCompVec->doExport(*quasiRegVec, *(rowImport), Xpetra::ADD);
@@ -243,10 +236,10 @@ void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, Glo
     tm = Teuchos::null;
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("regionalToComposite: 4 - compVec->sumIntoLocalValue")));
 
-    for(LO vecIdx = 0; vecIdx < static_cast<LO>(partialCompVec->getNumVectors()); ++vecIdx) {
+    for (LO vecIdx = 0; vecIdx < static_cast<LO>(partialCompVec->getNumVectors()); ++vecIdx) {
       ArrayRCP<const SC> partialCompVecData = partialCompVec->getData(vecIdx);
-      ArrayRCP<SC>       compVecData        = compVec->getDataNonConst(vecIdx);
-      for(size_t entryIdx = 0; entryIdx < compVecLocalLength; ++entryIdx) {
+      ArrayRCP<SC> compVecData              = compVec->getDataNonConst(vecIdx);
+      for (size_t entryIdx = 0; entryIdx < compVecLocalLength; ++entryIdx) {
         compVecData[entryIdx] += partialCompVecData[entryIdx];
       }
     }
@@ -255,7 +248,7 @@ void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, Glo
   }
 
   return;
-} // regionalToComposite
+}  // regionalToComposite
 
 /*! \brief Sum region interface values
  *
@@ -266,10 +259,9 @@ void regionalToComposite(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, Glo
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void sumInterfaceValues(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,
-                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >  revisedRowMap,///< revised row maps in region layout [in]
-                        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >  rowImport ///< row importer in region layout [in])
-                        )
-{
+                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,  ///< revised row maps in region layout [in]
+                        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport          ///< row importer in region layout [in])
+) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
 
@@ -292,8 +284,7 @@ void sumInterfaceValues(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal,
   tm = Teuchos::null;
 
   return;
-} // sumInterfaceValues
-
+}  // sumInterfaceValues
 
 /*! \brief Sum region interface values
  *
@@ -304,10 +295,9 @@ void sumInterfaceValues(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal,
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void sumInterfaceValues(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,
-                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,///< revised row maps in region layout [in]
-                        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport ///< row importer in region layout [in])
-                        )
-{
+                        const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > revisedRowMap,  ///< revised row maps in region layout [in]
+                        const RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > rowImport          ///< row importer in region layout [in])
+) {
 #include "Xpetra_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
 
@@ -330,8 +320,7 @@ void sumInterfaceValues(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdi
   tm = Teuchos::null;
 
   return;
-} // sumInterfaceValues
-
+}  // sumInterfaceValues
 
 /*! \brief Apply scaling to interface DOFs
  *
@@ -342,27 +331,23 @@ void sumInterfaceValues(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdi
  * This can be achieved by setting \c inverseScaling to \c true.
  */
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void scaleInterfaceDOFs(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec, ///< Vector to be scaled
-                        const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& scalingFactors, ///< Vector with scaling factors
-                        bool inverseScaling ///< Divide by scaling factors (yes/no?)
-                        )
-{
-  using Vector = Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+void scaleInterfaceDOFs(RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regVec,                ///< Vector to be scaled
+                        const RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& scalingFactors,  ///< Vector with scaling factors
+                        bool inverseScaling                                                                     ///< Divide by scaling factors (yes/no?)
+) {
+  using Vector        = Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using VectorFactory = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
   const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-  const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
 
-  if (inverseScaling)
-  {
+  if (inverseScaling) {
     RCP<Vector> inverseScalingFactors = VectorFactory::Build(scalingFactors->getMap());
     inverseScalingFactors->reciprocal(*scalingFactors);
     regVec->elementWiseMultiply(one, *regVec, *inverseScalingFactors, zero);
-  }
-  else
-  {
+  } else {
     regVec->elementWiseMultiply(one, *regVec, *scalingFactors, zero);
   }
-}// scaleInterfaceDOFs
+}  // scaleInterfaceDOFs
 
-#endif // MUELU_SETUPREGIONVECTOR_DEF_HPP
+#endif  // MUELU_SETUPREGIONVECTOR_DEF_HPP
diff --git a/packages/muelu/research/regionMG/src/SolveRegionHierarchy_def.hpp b/packages/muelu/research/regionMG/src/SolveRegionHierarchy_def.hpp
index 18616ae101b3..44d5347377a3 100644
--- a/packages/muelu/research/regionMG/src/SolveRegionHierarchy_def.hpp
+++ b/packages/muelu/research/regionMG/src/SolveRegionHierarchy_def.hpp
@@ -47,44 +47,43 @@
 
 #include "SetupRegionHierarchy_def.hpp"
 
+using Teuchos::Array;
 using Teuchos::RCP;
 using Teuchos::rcp;
-using Teuchos::Array;
 
 //! Recursive multigrid cycle (V or W) in region fashion
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MgCycle(const int levelID, ///< ID of current level
-            const std::string cycleType,
-            RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-            RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& fineRegX, ///< solution
-            RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > fineRegB, ///< right hand side
-            Array<RCP<Teuchos::ParameterList> > smootherParams, ///< region smoother parameter list
-            bool& zeroInitGuess,
-            RCP<ParameterList> coarseSolverData = Teuchos::null,
-            RCP<ParameterList> hierarchyData = Teuchos::null)
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MgCycle(const int levelID,  ///< ID of current level
+             const std::string cycleType,
+             RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+             RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& fineRegX,  ///< solution
+             RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> fineRegB,   ///< right hand side
+             Array<RCP<Teuchos::ParameterList>> smootherParams,                         ///< region smoother parameter list
+             bool& zeroInitGuess,
+             RCP<ParameterList> coarseSolverData = Teuchos::null,
+             RCP<ParameterList> hierarchyData    = Teuchos::null) {
 #include "MueLu_UseShortNames.hpp"
   using Teuchos::TimeMonitor;
   const Scalar SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-  const Scalar SC_ONE = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar SC_ONE  = Teuchos::ScalarTraits<Scalar>::one();
 
-  RCP<Level> level = regHierarchy->GetLevel(levelID);
-  RCP<Matrix> regMatrix = level->Get<RCP<Matrix> >("A", MueLu::NoFactory::get());
-  RCP<const Map> regRowMap = regMatrix->getRowMap();
-  RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > regRowImporter = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport");
-  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > regInterfaceScalings = level->Get<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >("regInterfaceScalings");
+  RCP<Level> level                                                                    = regHierarchy->GetLevel(levelID);
+  RCP<Matrix> regMatrix                                                               = level->Get<RCP<Matrix>>("A", MueLu::NoFactory::get());
+  RCP<const Map> regRowMap                                                            = regMatrix->getRowMap();
+  RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> regRowImporter               = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>>>("rowImport");
+  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> regInterfaceScalings = level->Get<RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>>("regInterfaceScalings");
 
   // Setup recursive cycling to represent either V- or W-cycles
   int cycleCount = 1;
-  if (cycleType == "W" && levelID > 0 ){ // W cycle and not on finest level.
+  if (cycleType == "W" && levelID > 0) {  // W cycle and not on finest level.
     const std::string coarseSolverType = coarseSolverData->get<std::string>("coarse solver type");
-    if (coarseSolverType == "direct" && levelID == regHierarchy->GetNumLevels()-2 ) // Only call coarse level solve once if direct solve
+    if (coarseSolverType == "direct" && levelID == regHierarchy->GetNumLevels() - 2)  // Only call coarse level solve once if direct solve
       cycleCount = 1;
     else
       cycleCount = 2;
   }
 
-  if (levelID < regHierarchy->GetNumLevels() - 1) // fine or intermediate levels
+  if (levelID < regHierarchy->GetNumLevels() - 1)  // fine or intermediate levels
   {
     // extract data from hierarchy parameterlist
     std::string levelName("level" + std::to_string(levelID));
@@ -107,8 +106,8 @@ void MgCycle(const int levelID, ///< ID of current level
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MgCycle: 2 - compute residual")));
 
     RCP<Vector> regRes;
-    if(useCachedVectors) {
-      regRes = levelList.get<RCP<Vector> >("residual");
+    if (useCachedVectors) {
+      regRes = levelList.get<RCP<Vector>>("residual");
     } else {
       regRes = VectorFactory::Build(regRowMap, true);
     }
@@ -127,49 +126,47 @@ void MgCycle(const int levelID, ///< ID of current level
     RCP<Vector> coarseRegB;
 
     {
-      RCP<Level> levelCoarse = regHierarchy->GetLevel(levelID+1);
-      RCP<Matrix> regProlongCoarse = levelCoarse->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-      RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > regRowMapCoarse = regProlongCoarse->getColMap();
+      RCP<Level> levelCoarse                                                    = regHierarchy->GetLevel(levelID + 1);
+      RCP<Matrix> regProlongCoarse                                              = levelCoarse->Get<RCP<Matrix>>("P", MueLu::NoFactory::get());
+      RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> regRowMapCoarse = regProlongCoarse->getColMap();
       // Get pre-communicated communication patterns for the fast MatVec
-      const ArrayRCP<LocalOrdinal> regionInterfaceLIDs = smootherParams[levelID+1]->get<ArrayRCP<LO>>("Fast MatVec: interface LIDs");
-      const RCP<Import> regionInterfaceImporter = smootherParams[levelID+1]->get<RCP<Import>>("Fast MatVec: interface importer");
+      const ArrayRCP<LocalOrdinal> regionInterfaceLIDs = smootherParams[levelID + 1]->get<ArrayRCP<LO>>("Fast MatVec: interface LIDs");
+      const RCP<Import> regionInterfaceImporter        = smootherParams[levelID + 1]->get<RCP<Import>>("Fast MatVec: interface importer");
 
       coarseRegX = VectorFactory::Build(regRowMapCoarse, true);
       coarseRegB = VectorFactory::Build(regRowMapCoarse, true);
 
       regProlongCoarse->apply(*regRes, *coarseRegB, Teuchos::TRANS, SC_ONE, SC_ZERO, true, regionInterfaceImporter, regionInterfaceLIDs);
-        // TEUCHOS_ASSERT(regProlong[l+1][j]->getRangeMap()->isSameAs(*regRes[j]->getMap()));
-        // TEUCHOS_ASSERT(regProlong[l+1][j]->getDomainMap()->isSameAs(*coarseRegB[j]->getMap()));
+      // TEUCHOS_ASSERT(regProlong[l+1][j]->getRangeMap()->isSameAs(*regRes[j]->getMap()));
+      // TEUCHOS_ASSERT(regProlong[l+1][j]->getDomainMap()->isSameAs(*coarseRegB[j]->getMap()));
     }
 
-    tm = Teuchos::null;
+    tm                       = Teuchos::null;
     bool coarseZeroInitGuess = true;
 
-    for(int cycle = 0; cycle < cycleCount; cycle++)
-    {
-
+    for (int cycle = 0; cycle < cycleCount; cycle++) {
       // Call V-cycle recursively
-      MgCycle(levelID+1, cycleType, regHierarchy,
-             coarseRegX, coarseRegB,
-             smootherParams, coarseZeroInitGuess, coarseSolverData, hierarchyData);
+      MgCycle(levelID + 1, cycleType, regHierarchy,
+              coarseRegX, coarseRegB,
+              smootherParams, coarseZeroInitGuess, coarseSolverData, hierarchyData);
 
-    } //cycleCount
+    }  // cycleCount
 
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MgCycle: 6 - transfer coarse to fine")));
 
     // Transfer coarse level correction to fine level
     RCP<Vector> regCorrection;
     {
-      RCP<Level> levelCoarse = regHierarchy->GetLevel(levelID+1);
-      RCP<Matrix> regProlongCoarse = levelCoarse->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+      RCP<Level> levelCoarse       = regHierarchy->GetLevel(levelID + 1);
+      RCP<Matrix> regProlongCoarse = levelCoarse->Get<RCP<Matrix>>("P", MueLu::NoFactory::get());
       // Get pre-communicated communication patterns for the fast MatVec
       const ArrayRCP<LocalOrdinal> regionInterfaceLIDs = smootherParams[levelID]->get<ArrayRCP<LO>>("Fast MatVec: interface LIDs");
-      const RCP<Import> regionInterfaceImporter = smootherParams[levelID]->get<RCP<Import>>("Fast MatVec: interface importer");
+      const RCP<Import> regionInterfaceImporter        = smootherParams[levelID]->get<RCP<Import>>("Fast MatVec: interface importer");
 
       regCorrection = VectorFactory::Build(regRowMap, true);
       regProlongCoarse->apply(*coarseRegX, *regCorrection, Teuchos::NO_TRANS, SC_ONE, SC_ZERO, false, regionInterfaceImporter, regionInterfaceLIDs);
-        // TEUCHOS_ASSERT(regProlong[l+1][j]->getDomainMap()->isSameAs(*coarseRegX[j]->getMap()));
-        // TEUCHOS_ASSERT(regProlong[l+1][j]->getRangeMap()->isSameAs(*regCorrection[j]->getMap()));
+      // TEUCHOS_ASSERT(regProlong[l+1][j]->getDomainMap()->isSameAs(*coarseRegX[j]->getMap()));
+      // TEUCHOS_ASSERT(regProlong[l+1][j]->getRangeMap()->isSameAs(*regCorrection[j]->getMap()));
     }
 
     tm = Teuchos::null;
@@ -189,7 +186,6 @@ void MgCycle(const int levelID, ///< ID of current level
     tm = Teuchos::null;
 
   } else {
-
     // Coarsest grid solve
 
     RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
@@ -202,13 +198,13 @@ void MgCycle(const int levelID, ///< ID of current level
     const std::string coarseSolverType = coarseSolverData->get<std::string>("coarse solver type");
     if (coarseSolverType == "smoother") {
       smootherApply(smootherParams[levelID], fineRegX, fineRegB, regMatrix,
-                  regRowMap, regRowImporter, zeroInitGuess);
+                    regRowMap, regRowImporter, zeroInitGuess);
     } else {
       zeroInitGuess = false;
       // First get the Xpetra vectors from region to composite format
-      RCP<const Map> coarseRowMap = coarseSolverData->get<RCP<const Map> >("compCoarseRowMap");
-      RCP<Vector> compX = VectorFactory::Build(coarseRowMap, true);
-      RCP<Vector> compRhs = VectorFactory::Build(coarseRowMap, true);
+      RCP<const Map> coarseRowMap = coarseSolverData->get<RCP<const Map>>("compCoarseRowMap");
+      RCP<Vector> compX           = VectorFactory::Build(coarseRowMap, true);
+      RCP<Vector> compRhs         = VectorFactory::Build(coarseRowMap, true);
       {
         RCP<Vector> inverseInterfaceScaling = VectorFactory::Build(regInterfaceScalings->getMap());
         inverseInterfaceScaling->reciprocal(*regInterfaceScalings);
@@ -217,15 +213,14 @@ void MgCycle(const int levelID, ///< ID of current level
         regionalToComposite(fineRegB, compRhs, regRowImporter);
       }
 
-      if (coarseSolverType == "direct")
-      {
+      if (coarseSolverType == "direct") {
 #if defined(HAVE_MUELU_AMESOS2)
 
-        using DirectCoarseSolver = Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >;
-        RCP<DirectCoarseSolver> coarseSolver = coarseSolverData->get<RCP<DirectCoarseSolver> >("direct solver object");
+        using DirectCoarseSolver             = Amesos2::Solver<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>;
+        RCP<DirectCoarseSolver> coarseSolver = coarseSolverData->get<RCP<DirectCoarseSolver>>("direct solver object");
 
-        TEUCHOS_TEST_FOR_EXCEPT_MSG(coarseRowMap->lib()!=Xpetra::UseTpetra,
-            "Coarse solver requires Tpetra/Amesos2 stack.");
+        TEUCHOS_TEST_FOR_EXCEPT_MSG(coarseRowMap->lib() != Xpetra::UseTpetra,
+                                    "Coarse solver requires Tpetra/Amesos2 stack.");
         TEUCHOS_ASSERT(!coarseSolver.is_null());
 
         // using Utilities = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
@@ -234,8 +229,8 @@ void MgCycle(const int levelID, ///< ID of current level
         // we could also implement a similar Epetra branch
         using Tpetra_MultiVector = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-      //    *fos << "Attempting to use Amesos2 to solve the coarse grid problem" << std::endl;
-        RCP<Tpetra_MultiVector> tX = Utilities::MV2NonConstTpetraMV2(*compX);
+        //    *fos << "Attempting to use Amesos2 to solve the coarse grid problem" << std::endl;
+        RCP<Tpetra_MultiVector> tX       = Utilities::MV2NonConstTpetraMV2(*compX);
         RCP<const Tpetra_MultiVector> tB = Utilities::MV2TpetraMV(compRhs);
 
         /* Solve!
@@ -249,10 +244,12 @@ void MgCycle(const int levelID, ///< ID of current level
          */
         if (not coarseSolver->getStatus().symbolicFactorizationDone())
           *fos << "Symbolic factorization should have been done during hierarchy setup, "
-              "but actually is missing. Anyway ... just do it right now." << std::endl;
+                  "but actually is missing. Anyway ... just do it right now."
+               << std::endl;
         if (not coarseSolver->getStatus().numericFactorizationDone())
           *fos << "Numeric factorization should have been done during hierarchy setup, "
-              "but actually is missing. Anyway ... just do it right now." << std::endl;
+                  "but actually is missing. Anyway ... just do it right now."
+               << std::endl;
         coarseSolver->solve(tX.ptr(), tB.ptr());
 #else
         *fos << "+++++++++++++++++++++++++++ WARNING +++++++++++++++++++++++++\n"
@@ -261,8 +258,7 @@ void MgCycle(const int levelID, ///< ID of current level
              << "+++++++++++++++++++++++++++ WARNING +++++++++++++++++++++++++"
              << std::endl;
 #endif
-      }
-      else if (coarseSolverType == "amg") // use AMG as coarse level solver
+      } else if (coarseSolverType == "amg")  // use AMG as coarse level solver
       {
         const bool coarseSolverRebalance = coarseSolverData->get<bool>("coarse solver rebalance");
 
@@ -270,22 +266,22 @@ void MgCycle(const int levelID, ///< ID of current level
         RCP<Hierarchy> amgHierarchy = coarseSolverData->get<RCP<Hierarchy>>("amg hierarchy object");
 
         // Run a single V-cycle
-        if(coarseSolverRebalance==false){
+        if (coarseSolverRebalance == false) {
           amgHierarchy->Iterate(*compRhs, *compX, 1, true);
 
         } else {
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
-          RCP<const Import> rebalanceImporter = coarseSolverData->get<RCP<const Import> >("rebalanceImporter");
+          RCP<const Import> rebalanceImporter = coarseSolverData->get<RCP<const Import>>("rebalanceImporter");
 
           // TODO: These vectors could be cached to improve performance
           RCP<Vector> rebalancedRhs = VectorFactory::Build(rebalanceImporter->getTargetMap());
-          RCP<Vector> rebalancedX = VectorFactory::Build(rebalanceImporter->getTargetMap(), true);
+          RCP<Vector> rebalancedX   = VectorFactory::Build(rebalanceImporter->getTargetMap(), true);
           rebalancedRhs->doImport(*compRhs, *rebalanceImporter, Xpetra::INSERT);
 
           rebalancedRhs->replaceMap(rebalancedRhs->getMap()->removeEmptyProcesses());
           rebalancedX->replaceMap(rebalancedX->getMap()->removeEmptyProcesses());
 
-          if(!amgHierarchy.is_null()){
+          if (!amgHierarchy.is_null()) {
             amgHierarchy->Iterate(*rebalancedRhs, *rebalancedX, 1, true);
           }
 
@@ -295,9 +291,7 @@ void MgCycle(const int levelID, ///< ID of current level
           amgHierarchy->Iterate(*compRhs, *compX, 1, true);
 #endif
         }
-      }
-      else
-      {
+      } else {
         TEUCHOS_TEST_FOR_EXCEPT_MSG(false, "Unknown coarse solver type.");
       }
 
@@ -312,39 +306,37 @@ void MgCycle(const int levelID, ///< ID of current level
   }
 
   return;
-} // MgCycle
+}  // MgCycle
 
 //! Adapter that uses composite vectors and a region hierarchy
 //  and performs a region MG cycle on them.
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RegionMgCycleAdapter(const std::string cycleType,
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X, ///< solution
-    RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > B, ///< right hand side
-    Array<RCP<Teuchos::ParameterList> > smootherParams, ///< region smoother parameter list
-    bool& zeroInitGuess,
-    RCP<ParameterList> coarseSolverData = Teuchos::null,
-    RCP<ParameterList> hierarchyData = Teuchos::null)
-{
-
+                          RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+                          RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,  ///< solution
+                          RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> B,   ///< right hand side
+                          Array<RCP<Teuchos::ParameterList>> smootherParams,                  ///< region smoother parameter list
+                          bool& zeroInitGuess,
+                          RCP<ParameterList> coarseSolverData = Teuchos::null,
+                          RCP<ParameterList> hierarchyData    = Teuchos::null) {
   using LO = LocalOrdinal;
   using GO = GlobalOrdinal;
   using NO = Node;
   using SC = Scalar;
 
-  using Level = MueLu::Level;
-  using Map    = Xpetra::Map<LO,GO,NO>;
-  using Import = Xpetra::Import<LO,GO,NO>;
-  using Matrix = Xpetra::Matrix<SC,LO,GO,NO>;
-  using Vector = Xpetra::Vector<SC,LO,GO,NO>;
-  using VectorFactory = Xpetra::VectorFactory<SC,LO,GO,NO>;
+  using Level         = MueLu::Level;
+  using Map           = Xpetra::Map<LO, GO, NO>;
+  using Import        = Xpetra::Import<LO, GO, NO>;
+  using Matrix        = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Vector        = Xpetra::Vector<SC, LO, GO, NO>;
+  using VectorFactory = Xpetra::VectorFactory<SC, LO, GO, NO>;
 
   // Extract some info from the hierarchy
   // to convert vectors from composite to regional and back
   RCP<Level> level0                = regHierarchy->GetLevel(0);
-  RCP<Import> rowImport            = level0->Get<RCP<Import> >("rowImport");
-  RCP<Vector> regInterfaceScalings = level0->Get<RCP<Vector> >("regInterfaceScalings");
-  RCP<Matrix> regMat               = level0->Get<RCP<Matrix> >("A");
+  RCP<Import> rowImport            = level0->Get<RCP<Import>>("rowImport");
+  RCP<Vector> regInterfaceScalings = level0->Get<RCP<Vector>>("regInterfaceScalings");
+  RCP<Matrix> regMat               = level0->Get<RCP<Matrix>>("A");
   RCP<const Map> revisedRowMap     = regMat->getRowMap();
 
   // Compute region vectors for B and X
@@ -359,66 +351,64 @@ void RegionMgCycleAdapter(const std::string cycleType,
                       revisedRowMap, rowImport);
 
   MgCycle(0, cycleType, regHierarchy,
-         regX, regB,
-         smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
+          regX, regB,
+          smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
 
   // Bring solution back to composite format
   scaleInterfaceDOFs(regX, regInterfaceScalings, true);
   regionalToComposite(regX, X, rowImport);
-} // RegionMgCycleAdapter
+}  // RegionMgCycleAdapter
 
 // Solve via Richardson iteration with region MG preconditioning, hand in matrix in region format
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist, const int maxIts,
                                   const std::string cycleType, const std::string convergenceLog,
                                   RCP<Teuchos::ParameterList>& coarseSolverData,
-                                  Array<RCP<Teuchos::ParameterList> >& smootherParams,
+                                  Array<RCP<Teuchos::ParameterList>>& smootherParams,
                                   RCP<Teuchos::ParameterList> hierarchyData,
-                                  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-                                  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X,
-                                  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B) {
-
+                                  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+                                  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                                  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B) {
   using LO = LocalOrdinal;
   using GO = GlobalOrdinal;
   using NO = Node;
   using SC = Scalar;
 
-  using Map    = Xpetra::Map<LO,GO,NO>;
-  using Import = Xpetra::Import<LO,GO,NO>;
-  using Matrix = Xpetra::Matrix<SC,LO,GO,NO>;
-  using Vector = Xpetra::Vector<SC,LO,GO,NO>;
-  using VectorFactory = Xpetra::VectorFactory<SC,LO,GO,NO>;
+  using Map           = Xpetra::Map<LO, GO, NO>;
+  using Import        = Xpetra::Import<LO, GO, NO>;
+  using Matrix        = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Vector        = Xpetra::Vector<SC, LO, GO, NO>;
+  using VectorFactory = Xpetra::VectorFactory<SC, LO, GO, NO>;
 
   using Level = MueLu::Level;
 
-  using STS = Teuchos::ScalarTraits<Scalar>;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
   using magnitude_type = typename STS::magnitudeType;
   const Scalar SC_zero = STS::zero();
   const Scalar SC_one  = STS::one();
 
   // we start by extracting some basic data from the hierarchy
-  RCP<Level> level0 = regHierarchy->GetLevel(0);
-  RCP<Matrix> regMat  = level0->Get<RCP<Matrix> >("A");
-  RCP<const Map> revisedRowMap  = regMat->getRowMap();
-  RCP<Import> rowImport = level0->Get<RCP<Import> >("rowImport");
-  RCP<const Map> dofMap = X->getMap();
-  const int myRank = dofMap->getComm()->getRank();
+  RCP<Level> level0            = regHierarchy->GetLevel(0);
+  RCP<Matrix> regMat           = level0->Get<RCP<Matrix>>("A");
+  RCP<const Map> revisedRowMap = regMat->getRowMap();
+  RCP<Import> rowImport        = level0->Get<RCP<Import>>("rowImport");
+  RCP<const Map> dofMap        = X->getMap();
+  const int myRank             = dofMap->getComm()->getRank();
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   // Prepare output of residual norm to file
   RCP<std::ofstream> log;
-  if (myRank == 0)
-    {
-      log = rcp(new std::ofstream(convergenceLog.c_str()));
-      (*log) << "# num procs = " << dofMap->getComm()->getSize() << "\n"
-             << "# iteration | res-norm (scaled=" << scaleResidualHist << ")\n"
-             << "#\n";
-      *log << std::setprecision(16) << std::scientific;
-    }
+  if (myRank == 0) {
+    log = rcp(new std::ofstream(convergenceLog.c_str()));
+    (*log) << "# num procs = " << dofMap->getComm()->getSize() << "\n"
+           << "# iteration | res-norm (scaled=" << scaleResidualHist << ")\n"
+           << "#\n";
+    *log << std::setprecision(16) << std::scientific;
+  }
 
   // Print type of residual norm to the screen
   out << "Using region solver" << std::endl;
@@ -427,7 +417,7 @@ void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist
   else
     out << "Using unscaled residual norm." << std::endl;
 
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels()>0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels() > 0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
 
   // We first use the non-level container variables to setup the fine grid problem.
   // This is ok since the initial setup just mimics the application and the outer
@@ -439,7 +429,7 @@ void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist
 
   // Composite residual vector
   RCP<Vector> compRes = VectorFactory::Build(dofMap, true);
-  compRes = VectorFactory::Build(dofMap, true);
+  compRes             = VectorFactory::Build(dofMap, true);
 
   // transform composite vectors to regional layout
   Teuchos::RCP<Vector> quasiRegX;
@@ -462,17 +452,16 @@ void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist
   // SWITCH TO RECURSIVE STYLE --> USE LEVEL CONTAINER VARIABLES
   /////////////////////////////////////////////////////////////////////////
 
-
   // Richardson iterations
   magnitude_type normResIni = Teuchos::ScalarTraits<magnitude_type>::zero();
-  const int old_precision = std::cout.precision();
+  const int old_precision   = std::cout.precision();
   std::cout << std::setprecision(8) << std::scientific;
   int cycle = 0;
 
   // Get Stuff out of Hierarchy
-  RCP<Level> level = regHierarchy->GetLevel(0);
-  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector> >("regInterfaceScalings");
-  bool zeroInitGuess = true;
+  RCP<Level> level                 = regHierarchy->GetLevel(0);
+  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector>>("regInterfaceScalings");
+  bool zeroInitGuess               = true;
   for (cycle = 0; cycle < maxIts; ++cycle) {
     regCorrect->putScalar(SC_zero);
     // check for convergence
@@ -485,8 +474,12 @@ void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist
       regionalToComposite(regRes, compRes, rowImport);
 
       typename Teuchos::ScalarTraits<Scalar>::magnitudeType normRes = compRes->norm2();
-      if(cycle == 0) { normResIni = normRes; }// out << "NormResIni = " << normResIni << std::endl; }
-      if(scaleResidualHist) { normRes /= normResIni; }
+      if (cycle == 0) {
+        normResIni = normRes;
+      }  // out << "NormResIni = " << normResIni << std::endl; }
+      if (scaleResidualHist) {
+        normRes /= normResIni;
+      }
 
       // Output current residual norm to screen (on proc 0 only)
       out << cycle << "\t" << normRes << std::endl;
@@ -505,70 +498,67 @@ void solveRegionProblemRichardson(const double tol, const bool scaleResidualHist
 
     // std::cout << "regB->norm2() " << regRes->norm2() << std::endl;
     MgCycle(0, cycleType, regHierarchy,
-           regCorrect, regRes,
-           smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
+            regCorrect, regRes,
+            smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
 
     // std::cout << "regX->norm2() " << regCorrect->norm2() << std::endl;
 
     regX->update(SC_one, *regCorrect, SC_one);
-
   }
   out << "Number of iterations performed for this solve: " << cycle << std::endl;
 
   std::cout << std::setprecision(old_precision);
   std::cout.unsetf(std::ios::fixed | std::ios::scientific);
-} // solveRegionProblemRichardson
+}  // solveRegionProblemRichardson
 
 // Solve via Conjugate Gradient with region MG preconditioning, hand in matrix in composite format
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, const int maxIts,
                               const std::string cycleType, const std::string convergenceLog,
                               RCP<Teuchos::ParameterList>& coarseSolverData,
-                              Array<RCP<Teuchos::ParameterList> >& smootherParams,
+                              Array<RCP<Teuchos::ParameterList>>& smootherParams,
                               RCP<Teuchos::ParameterList> hierarchyData,
-                              RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-                              RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-                              RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X,
-                              RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B) {
-
+                              RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+                              RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                              RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                              RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B) {
   using LO = LocalOrdinal;
   using GO = GlobalOrdinal;
   using NO = Node;
   using SC = Scalar;
 
-  using Map    = Xpetra::Map<LO,GO,NO>;
-  using Import = Xpetra::Import<LO,GO,NO>;
-  using Matrix = Xpetra::Matrix<SC,LO,GO,NO>;
-  using Vector = Xpetra::Vector<SC,LO,GO,NO>;
-  using VectorFactory = Xpetra::VectorFactory<SC,LO,GO,NO>;
+  using Map           = Xpetra::Map<LO, GO, NO>;
+  using Import        = Xpetra::Import<LO, GO, NO>;
+  using Matrix        = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Vector        = Xpetra::Vector<SC, LO, GO, NO>;
+  using VectorFactory = Xpetra::VectorFactory<SC, LO, GO, NO>;
 
   using Level = MueLu::Level;
 
-  using STS = Teuchos::ScalarTraits<Scalar>;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
   using magnitude_type = typename STS::magnitudeType;
   const Scalar SC_zero = STS::zero();
   const Scalar SC_one  = STS::one();
 
   // we start by extracting some basic data from the hierarchy
-  RCP<Level> level0 = regHierarchy->GetLevel(0);
+  RCP<Level> level0     = regHierarchy->GetLevel(0);
   RCP<const Map> dofMap = X->getMap();
-  const int myRank = dofMap->getComm()->getRank();
+  const int myRank      = dofMap->getComm()->getRank();
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   // Prepare output of residual norm to file
   RCP<std::ofstream> log;
-  if (myRank == 0)
-    {
-      log = rcp(new std::ofstream(convergenceLog.c_str()));
-      (*log) << "# num procs = " << dofMap->getComm()->getSize() << "\n"
-             << "# iteration | res-norm (scaled=" << scaleResidualHist << ")\n"
-             << "#\n";
-      *log << std::setprecision(16) << std::scientific;
-    }
+  if (myRank == 0) {
+    log = rcp(new std::ofstream(convergenceLog.c_str()));
+    (*log) << "# num procs = " << dofMap->getComm()->getSize() << "\n"
+           << "# iteration | res-norm (scaled=" << scaleResidualHist << ")\n"
+           << "#\n";
+    *log << std::setprecision(16) << std::scientific;
+  }
 
   // Print type of residual norm to the screen
   out << "Using CG solver" << std::endl;
@@ -577,23 +567,23 @@ void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, co
   else
     out << "Using unscaled residual norm." << std::endl;
 
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels()>0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels() > 0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
 
   // PCG iterations
   const int old_precision = std::cout.precision();
   std::cout << std::setprecision(8) << std::scientific;
 
   // Get Stuff out of Hierarchy
-  RCP<Level> level = regHierarchy->GetLevel(0);
-  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector> >("regInterfaceScalings");
-  bool zeroInitGuess = true;
+  RCP<Level> level                 = regHierarchy->GetLevel(0);
+  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector>>("regInterfaceScalings");
+  bool zeroInitGuess               = true;
 
   // Set variables for iterations
-  int cycle = 0;
-  RCP<Vector> Res     = VectorFactory::Build(dofMap, true);
-  RCP<Vector> Z       = VectorFactory::Build(dofMap, true);
-  RCP<Vector> P       = VectorFactory::Build(dofMap, true);
-  RCP<Vector> AP      = VectorFactory::Build(dofMap, true);
+  int cycle                 = 0;
+  RCP<Vector> Res           = VectorFactory::Build(dofMap, true);
+  RCP<Vector> Z             = VectorFactory::Build(dofMap, true);
+  RCP<Vector> P             = VectorFactory::Build(dofMap, true);
+  RCP<Vector> AP            = VectorFactory::Build(dofMap, true);
   magnitude_type normResIni = Teuchos::ScalarTraits<magnitude_type>::zero();
   magnitude_type normRes    = Teuchos::ScalarTraits<magnitude_type>::zero();
 
@@ -603,9 +593,9 @@ void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, co
   Z->putScalar(SC_zero);
 
   RegionMgCycleAdapter(cycleType, regHierarchy,
-                Z, Res,
-                smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
-  P->update(SC_one, *Z, SC_zero); // deep copy values of Z into P
+                       Z, Res,
+                       smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
+  P->update(SC_one, *Z, SC_zero);  // deep copy values of Z into P
 
   Scalar alpha = SC_zero, beta_old = SC_zero, beta_new = SC_zero, PAP = SC_zero;
   for (cycle = 0; cycle < maxIts; ++cycle) {
@@ -613,19 +603,21 @@ void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, co
     PAP = P->dot(*AP);
 
     TEUCHOS_TEST_FOR_EXCEPTION(PAP <= SC_zero, std::runtime_error,
-	 "At iteration " << (cycle) << " out of " << maxIts
-         << ", P.dot(AP) = " << PAP << " <= 0.  This usually means that "
-	 "the matrix A is not symmetric (Hermitian) positive definite.");
+                               "At iteration " << (cycle) << " out of " << maxIts
+                                               << ", P.dot(AP) = " << PAP << " <= 0.  This usually means that "
+                                                                             "the matrix A is not symmetric (Hermitian) positive definite.");
 
     beta_old = Res->dot(*Z);
-    alpha = beta_old / PAP;
+    alpha    = beta_old / PAP;
     X->update(alpha, *P, SC_one);
     Res->update(-alpha, *AP, SC_one);
 
     // check for convergence
     {
       normRes = Res->norm2();
-      if(scaleResidualHist) { normRes /= normResIni; }
+      if (scaleResidualHist) {
+        normRes /= normResIni;
+      }
 
       // Output current residual norm to screen (on proc 0 only)
       out << cycle << "\t" << normRes << std::endl;
@@ -638,8 +630,8 @@ void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, co
 
     Z->putScalar(SC_zero);
     RegionMgCycleAdapter(cycleType, regHierarchy,
-                  Z, Res,
-                  smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
+                         Z, Res,
+                         smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
 
     beta_new = Res->dot(*Z);
     P->update(SC_one, *Z, (beta_new / beta_old));
@@ -649,52 +641,50 @@ void solveCompositeProblemPCG(const double tol, const bool scaleResidualHist, co
 
   std::cout << std::setprecision(old_precision);
   std::cout.unsetf(std::ios::fixed | std::ios::scientific);
-} // solveCompositeProblemPCG
+}  // solveCompositeProblemPCG
 
 // Solve via Richardson iteration with region MG preconditioning, hand in matrix in composite format
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void solveCompositeProblemRichardson(const double tol, const bool scaleResidualHist, const int maxIts,
                                      const std::string cycleType, const std::string convergenceLog,
                                      RCP<Teuchos::ParameterList>& coarseSolverData,
-                                     Array<RCP<Teuchos::ParameterList> >& smootherParams,
+                                     Array<RCP<Teuchos::ParameterList>>& smootherParams,
                                      RCP<Teuchos::ParameterList> hierarchyData,
-                                     RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & regHierarchy,
-                                     RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-                                     RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X,
-                                     RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B) {
-
+                                     RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& regHierarchy,
+                                     RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                     RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                                     RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B) {
   using LO = LocalOrdinal;
   using GO = GlobalOrdinal;
   using NO = Node;
   using SC = Scalar;
 
-  using Map    = Xpetra::Map<LO,GO,NO>;
-  using Import = Xpetra::Import<LO,GO,NO>;
-  using Matrix = Xpetra::Matrix<SC,LO,GO,NO>;
-  using Vector = Xpetra::Vector<SC,LO,GO,NO>;
-  using VectorFactory = Xpetra::VectorFactory<SC,LO,GO,NO>;
+  using Map           = Xpetra::Map<LO, GO, NO>;
+  using Import        = Xpetra::Import<LO, GO, NO>;
+  using Matrix        = Xpetra::Matrix<SC, LO, GO, NO>;
+  using Vector        = Xpetra::Vector<SC, LO, GO, NO>;
+  using VectorFactory = Xpetra::VectorFactory<SC, LO, GO, NO>;
 
   using Level = MueLu::Level;
 
-  using STS = Teuchos::ScalarTraits<Scalar>;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
   using magnitude_type = typename STS::magnitudeType;
   const Scalar SC_zero = STS::zero();
   const Scalar SC_one  = STS::one();
 
   // we start by extracting some basic data from the hierarchy
-  RCP<Level> level0 = regHierarchy->GetLevel(0);
+  RCP<Level> level0     = regHierarchy->GetLevel(0);
   RCP<const Map> dofMap = X->getMap();
-  const int myRank = dofMap->getComm()->getRank();
+  const int myRank      = dofMap->getComm()->getRank();
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   // Prepare output of residual norm to file
   RCP<std::ofstream> log;
-  if (myRank == 0)
-  {
+  if (myRank == 0) {
     log = rcp(new std::ofstream(convergenceLog.c_str()));
     (*log) << "# num procs = " << dofMap->getComm()->getSize() << "\n"
            << "# iteration | res-norm (scaled=" << scaleResidualHist << ")\n"
@@ -709,25 +699,25 @@ void solveCompositeProblemRichardson(const double tol, const bool scaleResidualH
   else
     out << "Using unscaled residual norm." << std::endl;
 
-  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels()>0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
+  TEUCHOS_TEST_FOR_EXCEPT_MSG(!(regHierarchy->GetNumLevels() > 0), "We require numLevel > 0. Probably, numLevel has not been set, yet.");
 
   // Richardson iterations
   const int old_precision = std::cout.precision();
   std::cout << std::setprecision(8) << std::scientific;
 
   // Set variables for iterations
-  int cycle = 0;
-  RCP<Vector> Correct = VectorFactory::Build(dofMap, true);
-  RCP<Vector> Res = VectorFactory::Build(dofMap, true);
+  int cycle                 = 0;
+  RCP<Vector> Correct       = VectorFactory::Build(dofMap, true);
+  RCP<Vector> Res           = VectorFactory::Build(dofMap, true);
   magnitude_type normResIni = Teuchos::ScalarTraits<magnitude_type>::zero();
   magnitude_type normRes    = Teuchos::ScalarTraits<magnitude_type>::zero();
 
   // out << "X->norm2() " << X->norm2() << std::endl;
 
   // Get Stuff out of Hierarchy
-  RCP<Level> level = regHierarchy->GetLevel(0);
-  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector> >("regInterfaceScalings");
-  bool zeroInitGuess = true;
+  RCP<Level> level                 = regHierarchy->GetLevel(0);
+  RCP<Vector> regInterfaceScalings = level->Get<RCP<Vector>>("regInterfaceScalings");
+  bool zeroInitGuess               = true;
   for (cycle = 0; cycle < maxIts; ++cycle) {
     Correct->putScalar(SC_zero);
     // check for convergence
@@ -736,8 +726,12 @@ void solveCompositeProblemRichardson(const double tol, const bool scaleResidualH
       Res->update(SC_one, *B, SC_one);
       normRes = Res->norm2();
 
-      if(cycle == 0) { normResIni = normRes; }// out << "NormResIni = " << normResIni << std::endl;}
-      if(scaleResidualHist) { normRes /= normResIni; }
+      if (cycle == 0) {
+        normResIni = normRes;
+      }  // out << "NormResIni = " << normResIni << std::endl;}
+      if (scaleResidualHist) {
+        normRes /= normResIni;
+      }
 
       // Output current residual norm to screen (on proc 0 only)
       out << cycle << "\t" << normRes << std::endl;
@@ -749,8 +743,8 @@ void solveCompositeProblemRichardson(const double tol, const bool scaleResidualH
     }
 
     RegionMgCycleAdapter(cycleType, regHierarchy,
-                  Correct, Res,
-                  smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
+                         Correct, Res,
+                         smootherParams, zeroInitGuess, coarseSolverData, hierarchyData);
 
     X->update(SC_one, *Correct, SC_one);
   }
@@ -758,6 +752,6 @@ void solveCompositeProblemRichardson(const double tol, const bool scaleResidualH
 
   std::cout << std::setprecision(old_precision);
   std::cout.unsetf(std::ios::fixed | std::ios::scientific);
-} // solveCompositeProblemRichardson
+}  // solveCompositeProblemRichardson
 
-#endif // MUELU_SOLVEREGIONHIERARCHY_DEF_HPP
+#endif  // MUELU_SOLVEREGIONHIERARCHY_DEF_HPP
diff --git a/packages/muelu/research/regionMG/test/structured/Driver_Structured_Regions.cpp b/packages/muelu/research/regionMG/test/structured/Driver_Structured_Regions.cpp
index a2a3625d7101..35008bdb4fcc 100644
--- a/packages/muelu/research/regionMG/test/structured/Driver_Structured_Regions.cpp
+++ b/packages/muelu/research/regionMG/test/structured/Driver_Structured_Regions.cpp
@@ -88,12 +88,11 @@
 #include <BelosBlockGmresSolMgr.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
-#include <BelosTpetraAdapter.hpp>    // => This header defines Belos::TpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
+#include <BelosTpetraAdapter.hpp>  // => This header defines Belos::TpetraOp
 #endif
 
-
 #ifdef HAVE_MUELU_CUDA
 #include "cuda_profiler_api.h"
 #endif
@@ -126,75 +125,101 @@
 #include "SetupRegionHierarchy_def.hpp"
 #include "SolveRegionHierarchy_def.hpp"
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
   // const int numRanks = comm->getSize();
-  const int myRank   = comm->getRank();
+  const int myRank = comm->getRank();
 
   // =========================================================================
   // Convenient definitions
   // =========================================================================
   using STS = Teuchos::ScalarTraits<SC>;
   SC zero = STS::zero(), one = STS::one();
-  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-  using real_type = typename STS::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using real_type             = typename STS::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 10, ny = 10, nz = 10;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName           = "";                  clp.setOption("xml",                   &xmlFileName,           "read parameters from an xml file");
-  std::string yamlFileName          = "";                  clp.setOption("yaml",                  &yamlFileName,          "read parameters from a yaml file");
-  std::string solverType            = "region";            clp.setOption("solverType",            &solverType,            "iterative solver to be used: (region | Richardson | CG)");
-  std::string convergenceLog        = "residual_norm.txt"; clp.setOption("convergence-log",       &convergenceLog,        "file in which the convergence history of the linear solver is stored");
-  int         maxIts                = 200;                 clp.setOption("its",                   &maxIts,                "maximum number of solver iterations");
-  double      tol                   = 1e-12;               clp.setOption("tol",                   &tol,                   "solver convergence tolerance");
-  bool        scaleResidualHist     = true;                clp.setOption("scale", "noscale",      &scaleResidualHist,     "scaled Krylov residual history");
-  bool        serialRandom          = false;               clp.setOption("use-serial-random", "no-use-serial-random", &serialRandom, "generate the random vector serially and then broadcast it");
-  std::string cycleType = "V";                             clp.setOption("cycleType", &cycleType, "{Multigrid cycle type. Possible values: V, W.");
-  std::string smootherType          = "Jacobi";            clp.setOption("smootherType",          &smootherType,          "smoother to be used: (None | Jacobi | Gauss | Chebyshev)");
-  int         smootherIts           = 2;                   clp.setOption("smootherIts",           &smootherIts,           "number of smoother iterations");
-  double      smootherDamp          = 0.67;                clp.setOption("smootherDamp",          &smootherDamp,          "damping parameter for the level smoother");
-  double      smootherChebyEigRatio = 2.0;                 clp.setOption("smootherChebyEigRatio", &smootherChebyEigRatio, "eigenvalue ratio max/min used to approximate the smallest eigenvalue for Chebyshev relaxation");
-  double      smootherChebyBoostFactor = 1.1;              clp.setOption("smootherChebyBoostFactor", &smootherChebyBoostFactor, "boost factor for Chebyshev smoother");
-  bool        keepCoarseCoords      = false;               clp.setOption("keep-coarse-coords", "no-keep-coarse-coords", &keepCoarseCoords, "keep coordinates on coarsest level of region hierarchy");
-  bool        coarseSolverRebalance = false;               clp.setOption("rebalance-coarse", "no-rebalance-coarse", &coarseSolverRebalance, "rebalance before AMG coarse grid solve");
-  int         rebalanceNumPartitions = -1;                   clp.setOption("numPartitions",         &rebalanceNumPartitions, "number of partitions for rebalancing the coarse grid AMG solve");
-  std::string coarseSolverType      = "direct";            clp.setOption("coarseSolverType",      &coarseSolverType,      "Type of solver for (composite) coarse level operator (smoother | direct | amg)");
-  std::string unstructured          = "{}";                clp.setOption("unstructured",          &unstructured,          "List of ranks to be treated as unstructured, e.g. {0, 2, 5}");
-  std::string coarseAmgXmlFile      = "";                  clp.setOption("coarseAmgXml",          &coarseAmgXmlFile,      "Read parameters for AMG as coarse level solve from this xml file.");
-  std::string coarseSmootherXMLFile = "";                  clp.setOption("coarseSmootherXML",     &coarseSmootherXMLFile, "File containing the parameters to use with the coarse level smoother.");
-  std::string equilibrate = "no" ;                         clp.setOption("equilibrate",           &equilibrate,           "equilibrate the system (no | diag | 1-norm)");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+  std::string yamlFileName = "";
+  clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+  std::string solverType = "region";
+  clp.setOption("solverType", &solverType, "iterative solver to be used: (region | Richardson | CG)");
+  std::string convergenceLog = "residual_norm.txt";
+  clp.setOption("convergence-log", &convergenceLog, "file in which the convergence history of the linear solver is stored");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  bool scaleResidualHist = true;
+  clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
+  bool serialRandom = false;
+  clp.setOption("use-serial-random", "no-use-serial-random", &serialRandom, "generate the random vector serially and then broadcast it");
+  std::string cycleType = "V";
+  clp.setOption("cycleType", &cycleType, "{Multigrid cycle type. Possible values: V, W.");
+  std::string smootherType = "Jacobi";
+  clp.setOption("smootherType", &smootherType, "smoother to be used: (None | Jacobi | Gauss | Chebyshev)");
+  int smootherIts = 2;
+  clp.setOption("smootherIts", &smootherIts, "number of smoother iterations");
+  double smootherDamp = 0.67;
+  clp.setOption("smootherDamp", &smootherDamp, "damping parameter for the level smoother");
+  double smootherChebyEigRatio = 2.0;
+  clp.setOption("smootherChebyEigRatio", &smootherChebyEigRatio, "eigenvalue ratio max/min used to approximate the smallest eigenvalue for Chebyshev relaxation");
+  double smootherChebyBoostFactor = 1.1;
+  clp.setOption("smootherChebyBoostFactor", &smootherChebyBoostFactor, "boost factor for Chebyshev smoother");
+  bool keepCoarseCoords = false;
+  clp.setOption("keep-coarse-coords", "no-keep-coarse-coords", &keepCoarseCoords, "keep coordinates on coarsest level of region hierarchy");
+  bool coarseSolverRebalance = false;
+  clp.setOption("rebalance-coarse", "no-rebalance-coarse", &coarseSolverRebalance, "rebalance before AMG coarse grid solve");
+  int rebalanceNumPartitions = -1;
+  clp.setOption("numPartitions", &rebalanceNumPartitions, "number of partitions for rebalancing the coarse grid AMG solve");
+  std::string coarseSolverType = "direct";
+  clp.setOption("coarseSolverType", &coarseSolverType, "Type of solver for (composite) coarse level operator (smoother | direct | amg)");
+  std::string unstructured = "{}";
+  clp.setOption("unstructured", &unstructured, "List of ranks to be treated as unstructured, e.g. {0, 2, 5}");
+  std::string coarseAmgXmlFile = "";
+  clp.setOption("coarseAmgXml", &coarseAmgXmlFile, "Read parameters for AMG as coarse level solve from this xml file.");
+  std::string coarseSmootherXMLFile = "";
+  clp.setOption("coarseSmootherXML", &coarseSmootherXMLFile, "File containing the parameters to use with the coarse level smoother.");
+  std::string equilibrate = "no";
+  clp.setOption("equilibrate", &equilibrate, "equilibrate the system (no | diag | 1-norm)");
 #ifdef HAVE_MUELU_CUDA
-  bool profileSetup = false;                               clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
-  bool profileSolve = false;                               clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
+  bool profileSetup = false;
+  clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
+  bool profileSolve = false;
+  clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
 #endif
-  int  cacheSize = 0;                                      clp.setOption("cachesize",               &cacheSize,           "cache size (in KB)");
-  bool useStackedTimer   = false;                          clp.setOption("stacked-timer","no-stacked-timer", &useStackedTimer, "use stacked timer");
-  bool showTimerSummary = true;                            clp.setOption("show-timer-summary", "no-show-timer-summary", &showTimerSummary, "Switch on/off the timer summary at the end of the run.");
+  int cacheSize = 0;
+  clp.setOption("cachesize", &cacheSize, "cache size (in KB)");
+  bool useStackedTimer = false;
+  clp.setOption("stacked-timer", "no-stacked-timer", &useStackedTimer, "use stacked timer");
+  bool showTimerSummary = true;
+  clp.setOption("show-timer-summary", "no-show-timer-summary", &showTimerSummary, "Switch on/off the timer summary at the end of the run.");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   TEUCHOS_TEST_FOR_EXCEPTION(xmlFileName != "" && yamlFileName != "", std::runtime_error,
@@ -202,7 +227,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream &out       = *fancy;
   out.setOutputToRootOnly(0);
 
   ParameterList paramList;
@@ -218,18 +243,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), *comm);
   }
 
-  Array<RCP<Teuchos::ParameterList> > smootherParams(1); //TODO: this is good, resized to numlevel
+  Array<RCP<Teuchos::ParameterList> > smootherParams(1);  // TODO: this is good, resized to numlevel
   smootherParams[0] = rcp(new Teuchos::ParameterList());
-  smootherParams[0]->set("smoother: type",    smootherType);
-  smootherParams[0]->set("smoother: sweeps",  smootherIts);
+  smootherParams[0]->set("smoother: type", smootherType);
+  smootherParams[0]->set("smoother: sweeps", smootherIts);
   smootherParams[0]->set("smoother: damping", smootherDamp);
   smootherParams[0]->set("smoother: Chebyshev eigRatio", smootherChebyEigRatio);
   smootherParams[0]->set("smoother: Chebyshev boost factor", smootherChebyBoostFactor);
 
-  bool useUnstructured = false;
+  bool useUnstructured        = false;
   Array<LO> unstructuredRanks = Teuchos::fromStringToArray<LO>(unstructured);
-  for(int idx = 0; idx < unstructuredRanks.size(); ++idx) {
-    if(unstructuredRanks[idx] == myRank) {useUnstructured = true;}
+  for (int idx = 0; idx < unstructuredRanks.size(); ++idx) {
+    if (unstructuredRanks[idx] == myRank) {
+      useUnstructured = true;
+    }
   }
 
   // Retrieve matrix parameters (they may have been changed on the command line)
@@ -242,27 +269,26 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   std::ostringstream galeriStream;
 #ifdef HAVE_MUELU_OPENMP
   std::string node_name = Node::name();
-  if(!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
-    galeriStream<<"OpenMP Max Threads = "<<omp_get_max_threads()<<std::endl;
+  if (!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
+    galeriStream << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
 #endif
 
-
   comm->barrier();
   Teuchos::RCP<Teuchos::StackedTimer> stacked_timer;
-  if(useStackedTimer)
+  if (useStackedTimer)
     stacked_timer = rcp(new Teuchos::StackedTimer("MueLu_Driver"));
   Teuchos::TimeMonitor::setStackedTimer(stacked_timer);
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Build Composite Matrix")));
 
-
   RCP<Matrix> A;
-  RCP<Map>    nodeMap, dofMap;
+  RCP<Map> nodeMap, dofMap;
   RCP<Vector> X, B;
-  RCP<MultiVector>           nullspace;
+  RCP<MultiVector> nullspace;
   RCP<RealValuedMultiVector> coordinates;
 
-  galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+  galeriStream << "========================================================\n"
+               << xpetraParameters << galeriParameters;
 
   // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
   //                                 d1  d2  d3
@@ -274,8 +300,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // size. For example, np=14 will give a 7-by-2 distribution.
   // If you don't want Galeri to do this, specify mx or my on the galeriList.
   std::string matrixType = galeriParameters.GetMatrixType();
-  int numDimensions  = 0;
-  int numDofsPerNode = 0;
+  int numDimensions      = 0;
+  int numDofsPerNode     = 0;
   Teuchos::Array<GO> procsPerDim(3);
   Teuchos::Array<GO> gNodesPerDim(3);
   Teuchos::Array<LO> lNodesPerDim(3);
@@ -285,19 +311,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
   if (matrixType == "Laplace1D") {
     numDimensions = 1;
-    nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("1D", nodeMap, galeriList);
+    nodeMap       = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("1D", nodeMap, galeriList);
 
   } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
              matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
     numDimensions = 2;
-    nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
+    nodeMap       = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
 
   } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
     numDimensions = 3;
-    nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("3D", nodeMap, galeriList);
+    nodeMap       = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("3D", nodeMap, galeriList);
   }
 
   // Expand map to do multiple DOF per node for block problems
@@ -308,7 +334,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   } else {
     numDofsPerNode = 1;
   }
-  dofMap = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  dofMap = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -317,15 +343,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
     // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-    galeriList.set("right boundary" , "Neumann");
+    galeriList.set("right boundary", "Neumann");
     galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
+    galeriList.set("top boundary", "Neumann");
+    galeriList.set("front boundary", "Neumann");
+    galeriList.set("back boundary", "Neumann");
   }
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
   A = Pr->BuildMatrix();
 
   A->SetFixedBlockSize(numDofsPerNode);
@@ -335,40 +361,40 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   X = VectorFactory::Build(dofMap);
   B = VectorFactory::Build(dofMap);
 
-  if(serialRandom) {
-    //Build the seed on rank zero and broadcast it.
+  if (serialRandom) {
+    // Build the seed on rank zero and broadcast it.
     size_t localNumElements = 0;
-    if(comm->getRank() == 0) {
+    if (comm->getRank() == 0) {
       localNumElements = static_cast<size_t>(dofMap->getGlobalNumElements());
     }
-    RCP<Map> serialMap = MapFactory::Build(dofMap->lib(),
-                                           dofMap->getGlobalNumElements(),
-                                           localNumElements,
-                                           0,
-                                           comm);
+    RCP<Map> serialMap  = MapFactory::Build(dofMap->lib(),
+                                            dofMap->getGlobalNumElements(),
+                                            localNumElements,
+                                            0,
+                                            comm);
     RCP<Vector> Xserial = VectorFactory::Build(serialMap);
     Xserial->setSeed(251743369);
-    Xserial->randomize(true);// using xpetra's randomize. Otherwise random vector is only consistent for first 128 entries
+    Xserial->randomize(true);  // using xpetra's randomize. Otherwise random vector is only consistent for first 128 entries
     RCP<Import> randomnessImporter = ImportFactory::Build(serialMap, dofMap);
     X->doImport(*Xserial, *randomnessImporter, Xpetra::INSERT);
   } else {
     // we set seed for reproducibility
     Utilities::SetRandomSeed(*comm);
-    X->randomize(true);// using xpetra's randomize. Otherwise random vector is only consistent for first 128 entries
+    X->randomize(true);  // using xpetra's randomize. Otherwise random vector is only consistent for first 128 entries
   }
   A->apply(*X, *B, Teuchos::NO_TRANS, one, zero);
 
   Teuchos::Array<typename STS::magnitudeType> norms(1);
   B->norm2(norms);
-  B->scale(one/norms[0]);
+  B->scale(one / norms[0]);
   galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
 #ifdef MATLAB_COMPARE
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("Ax.mm",*B);
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("A.mm",*A);
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("Ax.mm", *B);
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mm", *A);
   B->putScalar(zero);
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("rhs.mm",*B);
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("x.mm",*X);
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("rhs.mm", *B);
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("x.mm", *X);
 #endif
   out << galeriStream.str();
 
@@ -380,14 +406,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Set aggregation type for each region
   std::string aggregationRegionType;
   RCP<ParameterList> interfaceParams = rcp(new ParameterList());
-  if(useUnstructured) {
+  if (useUnstructured) {
     aggregationRegionType = "uncoupled";
   } else {
     aggregationRegionType = "structured";
   }
 
   // Loading geometric info from galeri
-  if(numDimensions == 1) {
+  if (numDimensions == 1) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = 1;
     gNodesPerDim[2] = 1;
@@ -399,7 +425,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     procsPerDim[0] = galeriList.get<GO>("mx");
     procsPerDim[1] = 1;
     procsPerDim[2] = 1;
-  } else if(numDimensions == 2) {
+  } else if (numDimensions == 2) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = galeriList.get<GO>("ny");
     gNodesPerDim[2] = 1;
@@ -411,7 +437,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     procsPerDim[0] = galeriList.get<GO>("mx");
     procsPerDim[1] = galeriList.get<GO>("my");
     procsPerDim[2] = 1;
-  } else if(numDimensions == 3) {
+  } else if (numDimensions == 3) {
     gNodesPerDim[0] = galeriList.get<GO>("nx");
     gNodesPerDim[1] = galeriList.get<GO>("ny");
     gNodesPerDim[2] = galeriList.get<GO>("nz");
@@ -446,17 +472,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // First we count how many nodes the region needs to send and receive
   // and allocate arrays accordingly
   Array<int> boundaryConditions;
-  int maxRegPerGID = 0;
-  int numInterfaces = 0;
+  int maxRegPerGID       = 0;
+  int numInterfaces      = 0;
   LO numLocalRegionNodes = 0;
-  Array<GO>  sendGIDs;
+  Array<GO> sendGIDs;
   Array<int> sendPIDs;
-  Array<LO>  rNodesPerDim(3);
-  Array<LO>  compositeToRegionLIDs(nodeMap->getLocalNumElements()*numDofsPerNode);
-  Array<GO>  quasiRegionGIDs;
-  Array<GO>  quasiRegionCoordGIDs;
-  Array<GO>  interfaceGIDs;
-  Array<LO>  interfaceLIDsData;
+  Array<LO> rNodesPerDim(3);
+  Array<LO> compositeToRegionLIDs(nodeMap->getLocalNumElements() * numDofsPerNode);
+  Array<GO> quasiRegionGIDs;
+  Array<GO> quasiRegionCoordGIDs;
+  Array<GO> interfaceGIDs;
+  Array<LO> interfaceLIDsData;
 
   createRegionData(numDimensions, useUnstructured, numDofsPerNode,
                    gNodesPerDim(), lNodesPerDim(), procsPerDim(), nodeMap, dofMap,
@@ -466,7 +492,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
                    interfaceGIDs, interfaceLIDsData);
 
   // std::cout << "p=" << myRank << " | numSend=" << numSend << std::endl;
-            // << ", numReceive=" << numReceive << std::endl;
+  // << ", numReceive=" << numReceive << std::endl;
   // std::cout << "p=" << myRank << " | receiveGIDs: " << receiveGIDs << std::endl;
   // std::cout << "p=" << myRank << " | receivePIDs: " << receivePIDs << std::endl;
   // std::cout << "p=" << myRank << " | sendGIDs: " << sendGIDs << std::endl;
@@ -475,8 +501,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Second we actually fill the send and receive arrays with appropriate data
   // which will allow us to compute the region and composite maps.
   // Now we can construct a list of GIDs that corresponds to rowMap
-  Array<LO>  interfacesDimensions, interfacesLIDs;
-  if(useUnstructured) {
+  Array<LO> interfacesDimensions, interfacesLIDs;
+  if (useUnstructured) {
     findInterface(numDimensions, rNodesPerDim, boundaryConditions,
                   interfacesDimensions, interfacesLIDs);
 
@@ -488,8 +514,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     //           << "p=" << myRank << " | interfacesLIDs: " << interfacesLIDs << std::endl;
   }
 
-  interfaceParams->set<Array<LO> >("interfaces: nodes per dimensions", interfacesDimensions); // nodesPerDimensions);
-  interfaceParams->set<Array<LO> >("interfaces: interface nodes",      interfacesLIDs); // interfaceLIDs);
+  interfaceParams->set<Array<LO> >("interfaces: nodes per dimensions", interfacesDimensions);  // nodesPerDimensions);
+  interfaceParams->set<Array<LO> >("interfaces: interface nodes", interfacesLIDs);             // interfaceLIDs);
 
   // std::cout << "p=" << myRank << " | compositeToRegionLIDs: " << compositeToRegionLIDs << std::endl;
   // std::cout << "p=" << myRank << " | quasiRegionGIDs: " << quasiRegionGIDs << std::endl;
@@ -509,34 +535,34 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   RCP<TimeMonitor> tmLocal = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3.1 - Build Region Maps")));
 
-  Teuchos::RCP<const Xpetra::Map<LO,GO,NO> > rowMap, colMap;
-  Teuchos::RCP<const Xpetra::Map<LO,GO,NO> > revisedRowMap, revisedColMap;
-  rowMap = Xpetra::MapFactory<LO,GO,Node>::Build(dofMap->lib(),
+  Teuchos::RCP<const Xpetra::Map<LO, GO, NO> > rowMap, colMap;
+  Teuchos::RCP<const Xpetra::Map<LO, GO, NO> > revisedRowMap, revisedColMap;
+  rowMap        = Xpetra::MapFactory<LO, GO, Node>::Build(dofMap->lib(),
                                                           Teuchos::OrdinalTraits<GO>::invalid(),
                                                           quasiRegionGIDs(),
                                                           dofMap->getIndexBase(),
                                                           dofMap->getComm());
-  colMap = rowMap;
-  revisedRowMap = Xpetra::MapFactory<LO,GO,Node>::Build(dofMap->lib(),
-                                                                 Teuchos::OrdinalTraits<GO>::invalid(),
-                                                                 numLocalRegionNodes*numDofsPerNode,
-                                                                 dofMap->getIndexBase(),
-                                                                 dofMap->getComm());
+  colMap        = rowMap;
+  revisedRowMap = Xpetra::MapFactory<LO, GO, Node>::Build(dofMap->lib(),
+                                                          Teuchos::OrdinalTraits<GO>::invalid(),
+                                                          numLocalRegionNodes * numDofsPerNode,
+                                                          dofMap->getIndexBase(),
+                                                          dofMap->getComm());
   revisedColMap = revisedRowMap;
 
   // Build objects needed to construct the region coordinates
-  Teuchos::RCP<Xpetra::Map<LO,GO,NO> > quasiRegCoordMap = Xpetra::MapFactory<LO,GO,Node>::
-    Build(nodeMap->lib(),
-          Teuchos::OrdinalTraits<GO>::invalid(),
-          quasiRegionCoordGIDs(),
-          nodeMap->getIndexBase(),
-          nodeMap->getComm());
-  Teuchos::RCP<Xpetra::Map<LO,GO,NO> > regCoordMap = Xpetra::MapFactory<LO,GO,Node>::
-    Build(nodeMap->lib(),
-          Teuchos::OrdinalTraits<GO>::invalid(),
-          numLocalRegionNodes,
-          nodeMap->getIndexBase(),
-          nodeMap->getComm());
+  Teuchos::RCP<Xpetra::Map<LO, GO, NO> > quasiRegCoordMap = Xpetra::MapFactory<LO, GO, Node>::
+      Build(nodeMap->lib(),
+            Teuchos::OrdinalTraits<GO>::invalid(),
+            quasiRegionCoordGIDs(),
+            nodeMap->getIndexBase(),
+            nodeMap->getComm());
+  Teuchos::RCP<Xpetra::Map<LO, GO, NO> > regCoordMap = Xpetra::MapFactory<LO, GO, Node>::
+      Build(nodeMap->lib(),
+            Teuchos::OrdinalTraits<GO>::invalid(),
+            numLocalRegionNodes,
+            nodeMap->getIndexBase(),
+            nodeMap->getComm());
 
   comm->barrier();
   tmLocal = Teuchos::null;
@@ -545,15 +571,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Setup importers
   RCP<Import> rowImport;
   RCP<Import> colImport;
-  rowImport = ImportFactory::Build(dofMap, rowMap);
-  colImport = ImportFactory::Build(dofMap, colMap);
+  rowImport                 = ImportFactory::Build(dofMap, rowMap);
+  colImport                 = ImportFactory::Build(dofMap, colMap);
   RCP<Import> coordImporter = ImportFactory::Build(nodeMap, quasiRegCoordMap);
 
   comm->barrier();
   tmLocal = Teuchos::null;
   tmLocal = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3.3 - Import ghost GIDs")));
 
-  Array<GO>  interfaceCompositeGIDs, interfaceRegionGIDs;
+  Array<GO> interfaceCompositeGIDs, interfaceRegionGIDs;
   ExtractListOfInterfaceRegionGIDs(revisedRowMap, interfaceLIDsData, interfaceRegionGIDs);
 
   RCP<Xpetra::MultiVector<LO, LO, GO, NO> > regionsPerGIDWithGhosts;
@@ -589,7 +615,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
                      rowImport, quasiRegionMats, regionMats);
 
   // If we don't need the composite operator on the fine level anymore, free it!
-  if(solverType == "region") A = Teuchos::null;
+  if (solverType == "region") A = Teuchos::null;
 
   comm->barrier();
   tmLocal = Teuchos::null;
@@ -614,12 +640,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   regionNullspace->replaceMap(revisedRowMap);
 
   // create region coordinates vector
-  regionCoordinates = Xpetra::MultiVectorFactory<real_type,LO,GO,NO>::Build(quasiRegCoordMap,
+  regionCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(quasiRegCoordMap,
                                                                                coordinates->getNumVectors());
   regionCoordinates->doImport(*coordinates, *coordImporter, Xpetra::INSERT);
   regionCoordinates->replaceMap(regCoordMap);
 
-  using Tpetra_CrsMatrix = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Tpetra_CrsMatrix   = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using Tpetra_MultiVector = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
   /* Stuff for multi-level algorithm
@@ -639,46 +665,44 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   coarseSolverData->set<std::string>("smoother xml file", coarseSmootherXMLFile);
   RCP<ParameterList> hierarchyData = rcp(new ParameterList());
 
-
   // Create MueLu Hierarchy Initially...
   // Read MueLu parameter list form xml file
   RCP<ParameterList> mueluParams = Teuchos::rcp(new ParameterList());
   Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, mueluParams.ptr(), *dofMap->getComm());
 
   // Insert region-specific data into parameter list
-  const std::string userName = "user data";
-  Teuchos::ParameterList& userParamList = mueluParams->sublist(userName);
-  userParamList.set<int>        ("int numDimensions", numDimensions);
-  userParamList.set<Array<LO> > ("Array<LO> lNodesPerDim", regionNodesPerDim);
+  const std::string userName            = "user data";
+  Teuchos::ParameterList &userParamList = mueluParams->sublist(userName);
+  userParamList.set<int>("int numDimensions", numDimensions);
+  userParamList.set<Array<LO> >("Array<LO> lNodesPerDim", regionNodesPerDim);
   userParamList.set<std::string>("string aggregationRegionType", aggregationRegionType);
-  userParamList.set<Array<LO> > ("Array<LO> nodeOnInterface", interfaceParams->get<Array<LO> >("interfaces: interface nodes"));
-  userParamList.set<Array<LO> > ("Array<LO> interfacesDimensions", interfaceParams->get<Array<LO> >("interfaces: nodes per dimensions"));
-  if(Teuchos::nonnull(regionCoordinates)) {
+  userParamList.set<Array<LO> >("Array<LO> nodeOnInterface", interfaceParams->get<Array<LO> >("interfaces: interface nodes"));
+  userParamList.set<Array<LO> >("Array<LO> interfacesDimensions", interfaceParams->get<Array<LO> >("interfaces: nodes per dimensions"));
+  if (Teuchos::nonnull(regionCoordinates)) {
     userParamList.set("Coordinates", regionCoordinates);
   }
-  if(Teuchos::nonnull(regionNullspace)) {
+  if (Teuchos::nonnull(regionNullspace)) {
     userParamList.set("Nullspace", regionNullspace);
   }
 
   tmLocal = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("CreateXpetraPreconditioner: Hierarchy")));
 
   // Create multigrid hierarchy part 1
-  RCP<Hierarchy> regHierarchy  = MueLu::CreateXpetraPreconditioner(regionMats, *mueluParams);
+  RCP<Hierarchy> regHierarchy = MueLu::CreateXpetraPreconditioner(regionMats, *mueluParams);
 
   {
     RCP<MueLu::Level> level = regHierarchy->GetLevel(0);
-    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport",rowImport);
-    level->Set<ArrayView<LocalOrdinal> > ("compositeToRegionLIDs", compositeToRegionLIDs() );
+    level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("rowImport", rowImport);
+    level->Set<ArrayView<LocalOrdinal> >("compositeToRegionLIDs", compositeToRegionLIDs());
     level->Set<RCP<Xpetra::MultiVector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >("interfaceGIDs", interfaceGIDsMV);
     level->Set<RCP<Xpetra::MultiVector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >("regionsPerGIDWithGhosts", regionsPerGIDWithGhosts);
     level->Set<Teuchos::ArrayRCP<LocalOrdinal> >("regionMatVecLIDs", regionMatVecLIDs);
     level->Set<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("regionInterfaceImporter", regionInterfaceImporter);
-    level->print( std::cout, MueLu::Extreme );
+    level->print(std::cout, MueLu::Extreme);
   }
 
   tmLocal = Teuchos::null;
 
-
   // Create multigrid hierarchy part 2
   createRegionHierarchy(numDimensions,
                         regionNodesPerDim,
@@ -693,8 +717,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   hierarchyData->print();
 
-
-
   comm->barrier();
   tm = Teuchos::null;
 
@@ -702,10 +724,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   const int numLevels = regHierarchy->GetNumLevels();
 
   // Set data for fast MatVec
-  for(LO levelIdx = 0; levelIdx < numLevels; ++levelIdx) {
-    RCP<MueLu::Level> level = regHierarchy->GetLevel(levelIdx);
+  for (LO levelIdx = 0; levelIdx < numLevels; ++levelIdx) {
+    RCP<MueLu::Level> level                                = regHierarchy->GetLevel(levelIdx);
     RCP<Xpetra::Import<LO, GO, NO> > regionInterfaceImport = level->Get<RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > >("regionInterfaceImporter");
-    Teuchos::ArrayRCP<LO>            regionMatVecLIDs1     = level->Get<Teuchos::ArrayRCP<LO> >("regionMatVecLIDs");
+    Teuchos::ArrayRCP<LO> regionMatVecLIDs1                = level->Get<Teuchos::ArrayRCP<LO> >("regionMatVecLIDs");
     smootherParams[levelIdx]->set("Fast MatVec: interface LIDs",
                                   regionMatVecLIDs1);
     smootherParams[levelIdx]->set("Fast MatVec: interface importer",
@@ -722,68 +744,67 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Solve with V-cycle")));
 
 #ifdef DUMP_LOCALX_AND_A
-    FILE *fp;
-    char str[80];
-    sprintf(str,"theMatrix.%d",myRank);
-    fp = fopen(str,"w");
-    fprintf(fp, "%%%%MatrixMarket matrix coordinate real general\n");
-    LO numNzs = 0;
-    for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) {
-      ArrayView<const LO> AAcols;
-      ArrayView<const SC> AAvals;
-      regionMats->getLocalRowView(kkk, AAcols, AAvals);
-      const int *Acols    = AAcols.getRawPtr();
-      const SC  *Avals = AAvals.getRawPtr();
-      numNzs += AAvals.size();
-    }
-    fprintf(fp, "%d %d %d\n",regionMats->getLocalNumRows(),regionMats->getLocalNumRows(),numNzs);
-
-    for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) {
-      ArrayView<const LO> AAcols;
-      ArrayView<const SC> AAvals;
-      regionMats->getLocalRowView(kkk, AAcols, AAvals);
-      const int *Acols    = AAcols.getRawPtr();
-      const SC  *Avals = AAvals.getRawPtr();
-      LO RowLeng = AAvals.size();
-      for (LO kk = 0; kk < RowLeng; kk++) {
-          fprintf(fp, "%d %d %22.16e\n",kkk+1,Acols[kk]+1,Avals[kk]);
-      }
+  FILE *fp;
+  char str[80];
+  sprintf(str, "theMatrix.%d", myRank);
+  fp = fopen(str, "w");
+  fprintf(fp, "%%%%MatrixMarket matrix coordinate real general\n");
+  LO numNzs = 0;
+  for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) {
+    ArrayView<const LO> AAcols;
+    ArrayView<const SC> AAvals;
+    regionMats->getLocalRowView(kkk, AAcols, AAvals);
+    const int *Acols = AAcols.getRawPtr();
+    const SC *Avals  = AAvals.getRawPtr();
+    numNzs += AAvals.size();
+  }
+  fprintf(fp, "%d %d %d\n", regionMats->getLocalNumRows(), regionMats->getLocalNumRows(), numNzs);
+
+  for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) {
+    ArrayView<const LO> AAcols;
+    ArrayView<const SC> AAvals;
+    regionMats->getLocalRowView(kkk, AAcols, AAvals);
+    const int *Acols = AAcols.getRawPtr();
+    const SC *Avals  = AAvals.getRawPtr();
+    LO RowLeng       = AAvals.size();
+    for (LO kk = 0; kk < RowLeng; kk++) {
+      fprintf(fp, "%d %d %22.16e\n", kkk + 1, Acols[kk] + 1, Avals[kk]);
     }
-    fclose(fp);
-    // sprintf(str,"theX.%d",myRank);
-    // fp = fopen(str,"w");
-    // ArrayRCP<SC> lX= regX->getDataNonConst(0);
-    // for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) fprintf(fp, "%22.16e\n",lX[kkk]);
-    // fclose(fp);
+  }
+  fclose(fp);
+  // sprintf(str,"theX.%d",myRank);
+  // fp = fopen(str,"w");
+  // ArrayRCP<SC> lX= regX->getDataNonConst(0);
+  // for (size_t kkk = 0; kkk < regionMats->getLocalNumRows(); kkk++) fprintf(fp, "%22.16e\n",lX[kkk]);
+  // fclose(fp);
 #endif
 
-  if(solverType == "region") {
+  if (solverType == "region") {
     solveRegionProblemRichardson(tol, scaleResidualHist, maxIts,
-        cycleType, convergenceLog,
-        coarseSolverData, smootherParams, hierarchyData,
-        regHierarchy, X, B);
-  } else if(solverType == "Richardson") {
+                                 cycleType, convergenceLog,
+                                 coarseSolverData, smootherParams, hierarchyData,
+                                 regHierarchy, X, B);
+  } else if (solverType == "Richardson") {
     solveCompositeProblemRichardson(tol, scaleResidualHist, maxIts,
-        cycleType, convergenceLog,
-        coarseSolverData, smootherParams, hierarchyData,
-        regHierarchy, A, X, B);
-  } else if(solverType == "CG") {
+                                    cycleType, convergenceLog,
+                                    coarseSolverData, smootherParams, hierarchyData,
+                                    regHierarchy, A, X, B);
+  } else if (solverType == "CG") {
     solveCompositeProblemPCG(tol, scaleResidualHist, maxIts,
-        cycleType, convergenceLog,
-        coarseSolverData, smootherParams, hierarchyData,
-        regHierarchy, A, X, B);
+                             cycleType, convergenceLog,
+                             coarseSolverData, smootherParams, hierarchyData,
+                             regHierarchy, A, X, B);
   } else {
-    throw std::runtime_error("Unknown solverType: "+solverType);
+    throw std::runtime_error("Unknown solverType: " + solverType);
   }
 
   comm->barrier();
-  tm = Teuchos::null;
+  tm                = Teuchos::null;
   globalTimeMonitor = Teuchos::null;
 
-  if (showTimerSummary)
-  {
+  if (showTimerSummary) {
     RCP<ParameterList> reportParams = rcp(new ParameterList);
-    const std::string filter = "";
+    const std::string filter        = "";
     if (useStackedTimer) {
       Teuchos::StackedTimer::OutputOptions options;
       options.output_fraction = options.output_histogram = options.output_minmax = true;
@@ -805,5 +826,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/research/semicoarsening/driver.cpp b/packages/muelu/research/semicoarsening/driver.cpp
index b6c5687d748d..1da5a2793418 100644
--- a/packages/muelu/research/semicoarsening/driver.cpp
+++ b/packages/muelu/research/semicoarsening/driver.cpp
@@ -64,10 +64,10 @@ this directory.
 // Define default data types
 //#include <Tpetra_KokkosCompat_DefaultNode.hpp>
 
-//typedef double                                                              Scalar;
-//typedef int                                                                 LocalOrdinal;
-//typedef int                                                                 GlobalOrdinal;
-//typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType                         Node;
+// typedef double                                                              Scalar;
+// typedef int                                                                 LocalOrdinal;
+// typedef int                                                                 GlobalOrdinal;
+// typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType                         Node;
 
 #include <Xpetra_MultiVectorFactory.hpp>
 #include <Xpetra_ImportFactory.hpp>
@@ -95,7 +95,7 @@ this directory.
 #include <BelosMueLuAdapter.hpp>
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
   // Most MueLu and Xpetra classes are templated on some or all of the
   // following template types: Scalar, LocalOrdinal, GlobalOrdinal,
@@ -122,9 +122,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   // you don't have to write the namespace along with the class or
   // object name. This is especially helpful with commonly used things
   // like std::endl or Teuchos::RCP.
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ParameterList;
 
   // Start up MPI, if using MPI. Trilinos doesn't have to be built
   // with MPI; it's called a "serial" build if you build without MPI.
@@ -158,7 +158,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   // Make an output stream (for verbose output) that only prints on
   // Process 0 of the communicator.
   Teuchos::oblackholestream blackHole;
-  std::ostream& out = (myRank == 0) ? std::cout : blackHole;
+  std::ostream &out = (myRank == 0) ? std::cout : blackHole;
 
   // Teuchos provides an interface to get arguments from the command
   // line. The first argument indicates that we don't want any
@@ -171,36 +171,43 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   //  - Xpetra parameters: Allow switching between Epetra and Tpetra.
   //  - Driver parameters: These are specified below.
   GO nx = 6, ny = 5, nz = 163;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName = "driver1.xml";   clp.setOption("xml",                   &xmlFileName,     "read parameters from a file [default = 'scalingTest.xml']");
-
-  std::string solveType    = "cg";              clp.setOption("solver",                &solveType,       "solve type: (cg | gmres)");
-  double      tol          = 1e-12;             clp.setOption("tol",                   &tol,             "solver convergence tolerance");
-  int         maxIts       = 200;               clp.setOption("its",                   &maxIts,           "maximum number of solver iterations");
-
-  std::string mapFile;                          clp.setOption("map",                   &mapFile,          "map data file");
-  std::string matrixFile;                       clp.setOption("matrix",                &matrixFile,       "matrix data file");
-  std::string coordFile;                        clp.setOption("coords",                &coordFile,        "coordinates data file");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "driver1.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file [default = 'scalingTest.xml']");
+
+  std::string solveType = "cg";
+  clp.setOption("solver", &solveType, "solve type: (cg | gmres)");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+
+  std::string mapFile;
+  clp.setOption("map", &mapFile, "map data file");
+  std::string matrixFile;
+  clp.setOption("matrix", &matrixFile, "matrix data file");
+  std::string coordFile;
+  clp.setOption("coords", &coordFile, "coordinates data file");
 
   // Command line processor parsing stage
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   // Retrieve matrix parameters (they may have been changed on the
   // command line).
   ParameterList galeriList = galeriParameters.GetParameterList();
-//galeriList.set("mx",6); galeriList.set("my",5); galeriList.set("mz",1);
+  // galeriList.set("mx",6); galeriList.set("my",5); galeriList.set("mz",1);
 
   // Construct the problem data.  Typically, we construct the matrix,
   // coordinates, and nullspace, though only the matrix is mandatory.
-  RCP<Matrix>      A;
-  RCP<const Map>   map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<MultiVector> coordinates;
   if (matrixFile.empty()) {
     // Use Galeri to construct the data
@@ -208,7 +215,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Galeri provides several pre-defined problem types, including
     // some stencil based ones (like Laplace matrices in 2D and 3D),
     // and some others (like elasticity in 2D and 3D)
-    out << "========================================================\n" << xpetraParameters << galeriParameters;
+    out << "========================================================\n"
+        << xpetraParameters << galeriParameters;
 
     // Galeri will attempt to create a square-as-possible distribution of
     // subdomains di, e.g.,
@@ -229,16 +237,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Create map (Map) and coordinates (MultiVector).  Xpetra's Map
     // and MultiVector imitate Tpetra's interface.
     if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, galeriList);
 
     } else if (matrixType == "Laplace2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, galeriList);
 
     } else if (matrixType == "Laplace3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, galeriList);
     }
 
     out << "Processor subdomains in x direction: " << galeriList.get<int>("mx") << std::endl
@@ -248,8 +256,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Construct the matrix based on the problem name and provided
     // parameters.
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
     A = Pr->BuildMatrix();
 
   } else {
@@ -260,12 +268,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // (though you may avoid that in a serial run), a matrix (in a
     // MatrixMarket format), and a file with coordinates.
     if (!mapFile.empty())
-      map = Xpetra::IO<SC,LO,GO,Node>::ReadMap(mapFile, xpetraParameters.GetLib(), comm);
+      map = Xpetra::IO<SC, LO, GO, Node>::ReadMap(mapFile, xpetraParameters.GetLib(), comm);
 
-    A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, map);
+    A = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, map);
 
     if (!coordFile.empty())
-      coordinates = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(coordFile, map);
+      coordinates = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(coordFile, map);
   }
 
   // For scalar equations, we assume that the constant vector is a
@@ -276,7 +284,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   nullspace->putScalar(one);
 
   out << "Galeri complete.\n========================================================" << std::endl;
-std::cout << coordinates << std::endl;
+  std::cout << coordinates << std::endl;
 
   // Hierarchy is a key concept in MueLu. Once set up, it has all the
   // data required to apply the preconditioner. MueLu provides
@@ -302,7 +310,7 @@ std::cout << coordinates << std::endl;
   // subsequent coarser levels have increasing indices.  This way, if
   // we have L levels, level 0 corresponds to the finest level, and
   // level L-1 is the coarsest level.
-  RCP<Level>     L = H->GetLevel(0);
+  RCP<Level> L = H->GetLevel(0);
 
   // Setting the finest level matrix is mandatory in MueLu. However,
   // setting the nullspace and coordinates is optional. If nullspace
@@ -310,23 +318,23 @@ std::cout << coordinates << std::endl;
   // vector. Coordinates are needed only in two cases: if one uses
   // distance Laplacian filtering, or one uses repartitioning (which
   // is based on geometric partitioning).
-  L->Set("A",         A);
+  L->Set("A", A);
   L->Set("Nullspace", nullspace);
 
-/*
-  if (!coordinates.is_null())
-    L->Set("Coordinates", coordinates);
-*/
+  /*
+    if (!coordinates.is_null())
+      L->Set("Coordinates", coordinates);
+  */
 
-Teuchos::ArrayRCP<LO> SemiCoarsenInfo = Teuchos::arcp<LO>(3);
-SemiCoarsenInfo[NUM_ZPTS] = nz;
-SemiCoarsenInfo[ORIENTATION] = VERTICAL;
-//L->Set("SemiCoarsenInfo",SemiCoarsenInfo, MueLu::NoFactory::get() );
-L->Set("SemiCoarsenInfo",SemiCoarsenInfo);
+  Teuchos::ArrayRCP<LO> SemiCoarsenInfo = Teuchos::arcp<LO>(3);
+  SemiCoarsenInfo[NUM_ZPTS]             = nz;
+  SemiCoarsenInfo[ORIENTATION]          = VERTICAL;
+  // L->Set("SemiCoarsenInfo",SemiCoarsenInfo, MueLu::NoFactory::get() );
+  L->Set("SemiCoarsenInfo", SemiCoarsenInfo);
 
-printf("before level print\n");
-L->print(std::cout,Teuchos::VERB_EXTREME);
-printf("after level print\n");
+  printf("before level print\n");
+  L->print(std::cout, Teuchos::VERB_EXTREME);
+  printf("after level print\n");
 
   // Now that we have set the mandatory data, we can construct the
   // preconditioner.  The result of this is a fully constructed
@@ -357,11 +365,11 @@ printf("after level print\n");
 
     Teuchos::Array<MT> norms(1);
     B->norm2(norms);
-    B->scale(one/norms[0]);
+    B->scale(one / norms[0]);
     X->putScalar(zero);
   }
 #ifdef HAVE_MUELU_BELOS
-  typedef MultiVector          MV;
+  typedef MultiVector MV;
   typedef Belos::OperatorT<MV> OP;
 
   // Internally, Belos uses its own types. We use special wrappers to
@@ -369,18 +377,18 @@ printf("after level print\n");
   // one of many different ways one could choose to wrap MueLu and
   // Xpetra objects in order to get them to work with Belos.
   RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));
-  RCP<OP> belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));
 
   // Construct a Belos LinearProblem object. This is a complete
   // problem formulation. All the data necessary to solve the system
   // are now inside that problem.
-  RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem =
-    rcp (new Belos::LinearProblem<SC, MV, OP> (belosOp, X, B));
-  belosProblem->setRightPrec (belosPrec);
+  RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem =
+      rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+  belosProblem->setRightPrec(belosPrec);
 
   // Prepare the linear problem to solve the linear system that was
   // already passed in.
-  bool set = belosProblem->setProblem ();
+  bool set = belosProblem->setProblem();
   if (!set) {
     out << "\nERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
     return EXIT_FAILURE;
@@ -391,11 +399,11 @@ printf("after level print\n");
   // the number of iterations or its verbosity. For a full list of
   // Belos parameters, please consult the Belos package documentation.
   ParameterList belosList;
-  belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-  belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-  belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-  belosList.set("Output Frequency",      1);
-  belosList.set("Output Style",          Belos::Brief);
+  belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+  belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+  belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+  belosList.set("Output Frequency", 1);
+  belosList.set("Output Style", Belos::Brief);
 
   // Create a Belos iterative linear solver
   //
@@ -406,9 +414,9 @@ printf("after level print\n");
   // interface in Belos.
   RCP<Belos::SolverManager<SC, MV, OP> > solver;
   if (solveType == "cg")
-    solver = rcp(new Belos::PseudoBlockCGSolMgr<SC,MV,OP>(belosProblem, rcp(&belosList, false)));
+    solver = rcp(new Belos::PseudoBlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
   else if (solveType == "gmres")
-    solver = rcp(new Belos::BlockGmresSolMgr   <SC,MV,OP>(belosProblem, rcp(&belosList, false)));
+    solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
 
   // Finally, perform the solve.  We wrap it in a try-catch block,
   // since Belos indicates erros by throwing exceptions.
@@ -419,17 +427,21 @@ printf("after level print\n");
     // Get the number of iterations for this solve.
     out << "Converged in " << solver->getNumIters() << " iterations" << std::endl;
 
-  } catch(...) {
-    out << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+  } catch (...) {
+    out << std::endl
+        << "ERROR:  Belos threw an error! " << std::endl;
   }
 
   // Check convergence
   if (ret != Belos::Converged)
-    out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+    out << std::endl
+        << "ERROR:  Belos did not converge! " << std::endl;
   else
-    out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+    out << std::endl
+        << "SUCCESS:  Belos converged!" << std::endl;
 #else
-  out << std::endl << "MueLu has been compiled without Belos support!" << std::endl;
+  out << std::endl
+      << "MueLu has been compiled without Belos support!" << std::endl;
 #endif
   // GlobalMPISession calls MPI_Finalize() in its destructor, if
   // appropriate. You don't have to do anything here!  Just return
@@ -437,14 +449,10 @@ printf("after level print\n");
   return 0;
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
-
diff --git a/packages/muelu/research/tawiesn/aria/Driver.cpp b/packages/muelu/research/tawiesn/aria/Driver.cpp
index 10a0f563808c..05cdc1077191 100644
--- a/packages/muelu/research/tawiesn/aria/Driver.cpp
+++ b/packages/muelu/research/tawiesn/aria/Driver.cpp
@@ -50,7 +50,7 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 
 #include <MueLu.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 #include <MueLu_VisualizationHelpers.hpp>
 
 #include <MueLu_CreateXpetraPreconditioner.hpp>
@@ -60,94 +60,90 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-  public:
-    ExportVTK() {};
-
-  public:
-
-    void writeFile(std::ofstream& fout, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol)
-    {
-      using namespace std;
-      typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
-      Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > nodeMap = coordinates->getMap();
-      std::vector<LocalOrdinal> vertices;
-      std::vector<LocalOrdinal> geomSize;
-      LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
-
-      vertices.reserve(numFineNodes);
-      geomSize.reserve(numFineNodes);
-      for(LocalOrdinal i = 0; i < numFineNodes; i++)
-      {
-        vertices.push_back(i);
-        geomSize.push_back(1);
-      }
-
-      Teuchos::ArrayRCP<const double> xCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(0));
-      Teuchos::ArrayRCP<const double> yCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(1));
-      Teuchos::ArrayRCP<const double> zCoords = Teuchos::null;
-      if(coordinates->getNumVectors() == 3) {
-        zCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(2));
-      }
-      Teuchos::ArrayRCP<const double> solData = Teuchos::arcp_reinterpret_cast<const double>(sol->getData(0));
-
-      std::vector<int> uniqueFine = this->makeUnique(vertices);
-      this->writeFileVTKOpening(fout, uniqueFine, geomSize);
-      this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
-
-      std::string indent = "      ";
-      fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      int myRank = coordinates->getMap()->getComm()->getRank();
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        fout << myRank << " ";
-        if(i % 20 == 19)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      // solution vector
-      fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        size_t numVec = coordinates->getNumVectors();
-        for(int d=0; d<numVec; d++)
-          fout << solData[i*numVec+d] << " ";
-        if(i % 3 == 0)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      fout << "      </PointData>" << std::endl; // that is annoying
-
-      this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
-      this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
-      this->writeFileVTKClosing(fout);
-      fout.close();
-
-    };
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  ExportVTK(){};
+
+ public:
+  void writeFile(std::ofstream& fout, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol) {
+    using namespace std;
+    typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
+    Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > nodeMap = coordinates->getMap();
+    std::vector<LocalOrdinal> vertices;
+    std::vector<LocalOrdinal> geomSize;
+    LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
+
+    vertices.reserve(numFineNodes);
+    geomSize.reserve(numFineNodes);
+    for (LocalOrdinal i = 0; i < numFineNodes; i++) {
+      vertices.push_back(i);
+      geomSize.push_back(1);
+    }
 
+    Teuchos::ArrayRCP<const double> xCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(0));
+    Teuchos::ArrayRCP<const double> yCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(1));
+    Teuchos::ArrayRCP<const double> zCoords = Teuchos::null;
+    if (coordinates->getNumVectors() == 3) {
+      zCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(2));
+    }
+    Teuchos::ArrayRCP<const double> solData = Teuchos::arcp_reinterpret_cast<const double>(sol->getData(0));
+
+    std::vector<int> uniqueFine = this->makeUnique(vertices);
+    this->writeFileVTKOpening(fout, uniqueFine, geomSize);
+    this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
+
+    std::string indent = "      ";
+    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    int myRank = coordinates->getMap()->getComm()->getRank();
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      fout << myRank << " ";
+      if (i % 20 == 19)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    // solution vector
+    fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      size_t numVec = coordinates->getNumVectors();
+      for (int d = 0; d < numVec; d++)
+        fout << solData[i * numVec + d] << " ";
+      if (i % 3 == 0)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    fout << "      </PointData>" << std::endl;  // that is annoying
+
+    this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
+    this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
+    this->writeFileVTKClosing(fout);
+    fout.close();
   };
+};
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   bool success = false;
   bool verbose = true;
 
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -156,24 +152,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream& out       = *fancy;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
 
-    //GO nx = 100, ny = 100, nz = 100;
-    //Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    // GO nx = 100, ny = 100, nz = 100;
+    // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
 
-    std::string xmlFileName = "driver.xml";      clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+    std::string xmlFileName = "driver.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // =========================================================================
@@ -181,200 +178,197 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // =========================================================================
     RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatrixRead: S - Global Time"))), tm;
 
-
     GlobalOrdinal nGlobalNodes = 0;
-    if(comm->getRank() == 0) {
+    if (comm->getRank() == 0) {
       std::ifstream data_file;
       data_file.open("xxx.mm");
-      TEUCHOS_TEST_FOR_EXCEPTION(data_file.good() == false, MueLu::Exceptions::RuntimeError,"Problem opening file xxx.mm");
+      TEUCHOS_TEST_FOR_EXCEPTION(data_file.good() == false, MueLu::Exceptions::RuntimeError, "Problem opening file xxx.mm");
       std::string line;
-      getline(data_file,line);
+      getline(data_file, line);
       data_file >> nGlobalNodes;
       data_file.close();
     }
-    Teuchos::broadcast(*comm,0,1,&nGlobalNodes);
+    Teuchos::broadcast(*comm, 0, 1, &nGlobalNodes);
 
     out << "Found " << nGlobalNodes << " nodes " << std::endl;
 
     // each processor reads in its dofPresent array
     LocalOrdinal nNodes = 0;
     LocalOrdinal nDofs  = 0;
-    int maxDofPerNode = -1;
+    int maxDofPerNode   = -1;
     Teuchos::ArrayRCP<LocalOrdinal> dofPresent;
     {
-      std::stringstream ss; ss << comm->getSize() << "dofPresent" << comm->getRank();
+      std::stringstream ss;
+      ss << comm->getSize() << "dofPresent" << comm->getRank();
 
-      try{
+      try {
         std::ifstream data_file(ss.str());
         // read in first line containing number of local nodes and maxDofPerNode
         data_file >> nNodes >> maxDofPerNode;
-        dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(nNodes * maxDofPerNode,0);
+        dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(nNodes * maxDofPerNode, 0);
         // loop over all local nodes
-        for(LocalOrdinal i = 0; i < nNodes; i++) {
-          for(int j=0; j<maxDofPerNode; j++) {
+        for (LocalOrdinal i = 0; i < nNodes; i++) {
+          for (int j = 0; j < maxDofPerNode; j++) {
             int tmp = -1;
             data_file >> tmp;
-            if(tmp == 1) {
-              dofPresent[i*maxDofPerNode+j] = 1; nDofs++;
+            if (tmp == 1) {
+              dofPresent[i * maxDofPerNode + j] = 1;
+              nDofs++;
             } else {
-              dofPresent[i*maxDofPerNode+j] = 0;
+              dofPresent[i * maxDofPerNode + j] = 0;
             }
           }
         }
+      } catch (const std::exception& e) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception& e) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-
     }
 
     out << "PROC " << comm->getRank() << "/" << comm->getSize() << " " << nNodes << " " << maxDofPerNode << std::endl;
 
-
     /*for(LocalOrdinal i = 0; i < nNodes; i++) {
       for(int j=0; j<maxDofPerNode; j++) {
         std::cout << i << " " << dofPresent[i*maxDofPerNode+j] << std::endl;
       }
     }*/
 
-
     // read in dofs
 
     Teuchos::Array<GlobalOrdinal> dofGlobals;
     Teuchos::Array<GlobalOrdinal> nodalGlobals;  // nodal GIDs for laplacian (with holes)
     {
-      std::stringstream ss; ss << comm->getSize() << "proc" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "proc" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
 
         // read in first line containing number of local nodes and maxDofPerNode
         LocalOrdinal tmpDofs = 0;
         data_file >> tmpDofs;
-        TEUCHOS_TEST_FOR_EXCEPTION(tmpDofs != nDofs, MueLu::Exceptions::RuntimeError,"Number of gid entries in map file is " << tmpDofs << " but should be " << nDofs);
+        TEUCHOS_TEST_FOR_EXCEPTION(tmpDofs != nDofs, MueLu::Exceptions::RuntimeError, "Number of gid entries in map file is " << tmpDofs << " but should be " << nDofs);
 
         dofGlobals = Teuchos::Array<GlobalOrdinal>(nDofs);
 
         // loop over all local nodes
-        for(GlobalOrdinal i = 0; i < nDofs; i++) {
+        for (GlobalOrdinal i = 0; i < nDofs; i++) {
           int data;
           data_file >> data;
           dofGlobals[i] = Teuchos::as<GlobalOrdinal>(data);
         }
-      }
-      catch(const std::exception& e) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
+      } catch (const std::exception& e) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
 
       nodalGlobals = Teuchos::Array<GlobalOrdinal>(nNodes);
 
       GlobalOrdinal count = nDofs - 1;
-      for(GlobalOrdinal i = nNodes - 1; i>=0; i--) {
-        for(int j = maxDofPerNode-1; j >=0; j--) {
-          if(dofPresent[i*maxDofPerNode+j] == 1)
+      for (GlobalOrdinal i = nNodes - 1; i >= 0; i--) {
+        for (int j = maxDofPerNode - 1; j >= 0; j--) {
+          if (dofPresent[i * maxDofPerNode + j] == 1)
             nodalGlobals[i] = dofGlobals[count--];
         }
       }
     }
 
-    Teuchos::RCP<Map> LapMap = MapFactory::Build(lib,nGlobalNodes,nodalGlobals(),0,comm);
+    Teuchos::RCP<Map> LapMap = MapFactory::Build(lib, nGlobalNodes, nodalGlobals(), 0, comm);
 
-    //LapMap->describe(out, Teuchos::VERB_EXTREME);
+    // LapMap->describe(out, Teuchos::VERB_EXTREME);
 
     //
 
     {
-      std::stringstream ss; ss << comm->getSize() << "ProcLinear" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "ProcLinear" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
 
         // loop over all local nodes
-        for(GlobalOrdinal i = 0; i < nNodes; i++) {
+        for (GlobalOrdinal i = 0; i < nNodes; i++) {
           int data;
           data_file >> data;
           nodalGlobals[i] = Teuchos::as<GlobalOrdinal>(data);
         }
+      } catch (const std::exception& e) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception& e) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-      for(GlobalOrdinal i = 0; i < nNodes; i++) {
+      for (GlobalOrdinal i = 0; i < nNodes; i++) {
         nodalGlobals[i] = nodalGlobals[i] - 1;
       }
     }
 
-    Teuchos::RCP<Map> LinearMap = MapFactory::Build(lib,nGlobalNodes,nodalGlobals(),0,comm);
-    //LinearMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
+    Teuchos::RCP<Map> LinearMap = MapFactory::Build(lib, nGlobalNodes, nodalGlobals(), 0, comm);
+    // LinearMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
 
-    //LapMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
+    // LapMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
 
     Teuchos::RCP<Import> Importer = ImportFactory::Build(LinearMap, LapMap);
 
-    Teuchos::RCP<MultiVector> xpetraXXX = MultiVectorFactory::Build(LapMap,1);
-    Teuchos::RCP<MultiVector> xpetraYYY = MultiVectorFactory::Build(LapMap,1);
+    Teuchos::RCP<MultiVector> xpetraXXX = MultiVectorFactory::Build(LapMap, 1);
+    Teuchos::RCP<MultiVector> xpetraYYY = MultiVectorFactory::Build(LapMap, 1);
 
-    RCP<MultiVector> temp = IO::ReadMultiVector ("xxx.mm", LinearMap);
+    RCP<MultiVector> temp = IO::ReadMultiVector("xxx.mm", LinearMap);
     xpetraXXX->doImport(*temp, *Importer, Xpetra::INSERT);
-    temp = IO::ReadMultiVector ("yyy.mm", LinearMap);
+    temp = IO::ReadMultiVector("yyy.mm", LinearMap);
     xpetraYYY->doImport(*temp, *Importer, Xpetra::INSERT);
 
-    Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(LapMap,2);
-    Teuchos::ArrayRCP< const Scalar > srcX = xpetraXXX->getData(0);
-    Teuchos::ArrayRCP< const Scalar > srcY = xpetraYYY->getData(0);
-    Teuchos::ArrayRCP< Scalar > dataX = coordinates->getDataNonConst(0);
-    Teuchos::ArrayRCP< Scalar > dataY = coordinates->getDataNonConst(1);
-    for(decltype(coordinates->getLocalLength()) i = 0; i < coordinates->getLocalLength(); i++) {
+    Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(LapMap, 2);
+    Teuchos::ArrayRCP<const Scalar> srcX  = xpetraXXX->getData(0);
+    Teuchos::ArrayRCP<const Scalar> srcY  = xpetraYYY->getData(0);
+    Teuchos::ArrayRCP<Scalar> dataX       = coordinates->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> dataY       = coordinates->getDataNonConst(1);
+    for (decltype(coordinates->getLocalLength()) i = 0; i < coordinates->getLocalLength(); i++) {
       dataX[i] = srcX[i];
       dataY[i] = srcY[i];
     }
 
     // read in matrix
-    Teuchos::RCP<const Map>    DistributedMap  = Teuchos::null;
-    Teuchos::RCP<Matrix>       DistributedMatrix = Teuchos::null;
+    Teuchos::RCP<const Map> DistributedMap = Teuchos::null;
+    Teuchos::RCP<Matrix> DistributedMatrix = Teuchos::null;
 
     // read in matrix to determine number of rows
-    Teuchos::RCP<Matrix> SerialMatrix = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("theMatrix.m",lib,comm);
-    GlobalOrdinal NumRows = SerialMatrix->getRowMap()->getGlobalNumElements();
+    Teuchos::RCP<Matrix> SerialMatrix = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("theMatrix.m", lib, comm);
+    GlobalOrdinal NumRows             = SerialMatrix->getRowMap()->getGlobalNumElements();
 
     // re-read in matrix and distribute it using the user-given distribution over processors
-    DistributedMap    = MapFactory::Build(lib,NumRows,dofGlobals(),0,comm);
-    DistributedMatrix = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("theMatrix.m",
-              DistributedMap,
-              Teuchos::null,
-              DistributedMap,
-              DistributedMap,
-              true,   //           callFillComplete = true,
-              false,  //           binary           = false,
-              false,  //           tolerant         = false,
-              false); //           debug            = false)
+    DistributedMap    = MapFactory::Build(lib, NumRows, dofGlobals(), 0, comm);
+    DistributedMatrix = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("theMatrix.m",
+                                                                                    DistributedMap,
+                                                                                    Teuchos::null,
+                                                                                    DistributedMap,
+                                                                                    DistributedMap,
+                                                                                    true,    //           callFillComplete = true,
+                                                                                    false,   //           binary           = false,
+                                                                                    false,   //           tolerant         = false,
+                                                                                    false);  //           debug            = false)
 
     // read in global vectors (e.g. rhs)
     GlobalOrdinal nGlobalDof = 0;
     GlobalOrdinal nLocalDofs = Teuchos::as<GlobalOrdinal>(nDofs);
 
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_SUM,comm->getSize(),&nLocalDofs,&nGlobalDof);
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, comm->getSize(), &nLocalDofs, &nGlobalDof);
 
     Teuchos::RCP<const Map> dofLinearMap = Teuchos::null;
     {
-      std::stringstream ss; ss << comm->getSize() << "ProcLinear" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "ProcLinear" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
         // loop over all local nodes
         data_file >> dofGlobals[0];
+      } catch (const std::exception& e) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception& e) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-      for(decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = i + dofGlobals[0];
-      for(decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = dofGlobals[i] - 1;
-      dofLinearMap = MapFactory::Build(lib,nGlobalDof,dofGlobals(),0,comm);
+      for (decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = i + dofGlobals[0];
+      for (decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = dofGlobals[i] - 1;
+      dofLinearMap = MapFactory::Build(lib, nGlobalDof, dofGlobals(), 0, comm);
     }
 
     Teuchos::RCP<Import> dofImporter = ImportFactory::Build(dofLinearMap, DistributedMap);
 
-    RCP<MultiVector> RHS  = MultiVectorFactory::Build(DistributedMap,1);
-    RCP<MultiVector> LHS  = MultiVectorFactory::Build(DistributedMap,1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(DistributedMap, 1);
+    RCP<MultiVector> LHS = MultiVectorFactory::Build(DistributedMap, 1);
 
-    temp = IO::ReadMultiVector ("rhs.mm", dofLinearMap);
+    temp = IO::ReadMultiVector("rhs.mm", dofLinearMap);
     RHS->doImport(*temp, *dofImporter, Xpetra::INSERT);
     LHS->putScalar(zero);
 
@@ -383,9 +377,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     Teuchos::ParameterList paramList;
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
 
-    const std::string userName = "user data";
+    const std::string userName            = "user data";
     Teuchos::ParameterList& userParamList = paramList.sublist(userName);
-    userParamList.set("multivector Coordinates",coordinates);
+    userParamList.set("multivector Coordinates", coordinates);
     userParamList.set("ArrayRCP<LO> DofPresent", dofPresent);
 
     RCP<Hierarchy> H;
@@ -395,15 +389,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     {
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - Belos Solve")));
       // Operator and Multivector type that will be used with Belos
-      typedef MultiVector          MV;
+      typedef MultiVector MV;
       typedef Belos::OperatorT<MV> OP;
       H->IsPreconditioner(true);
 
       // Define Operator and Preconditioner
-      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(DistributedMatrix)); // Turns a Xpetra::Matrix object into a Belos operator
-      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(DistributedMatrix));  // Turns a Xpetra::Matrix object into a Belos operator
+      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));                   // Turns a MueLu::Hierarchy object into a Belos operator
       // Construct a Belos LinearProblem object
-      RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, LHS, RHS));
+      RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, LHS, RHS));
       belosProblem->setRightPrec(belosPrec);
       bool set = belosProblem->setProblem();
       if (set == false) {
@@ -413,14 +407,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       // Belos parameter list
       Teuchos::ParameterList belosList;
-      belosList.set("Maximum Iterations",    300); // Maximum number of iterations allowed
-      belosList.set("Convergence Tolerance", 1e-8);    // Relative convergence tolerance requested
-      belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-      belosList.set("Output Frequency",      1);
-      belosList.set("Output Style",          Belos::Brief);
+      belosList.set("Maximum Iterations", 300);      // Maximum number of iterations allowed
+      belosList.set("Convergence Tolerance", 1e-8);  // Relative convergence tolerance requested
+      belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+      belosList.set("Output Frequency", 1);
+      belosList.set("Output Style", Belos::Brief);
 
       // Create an iterative solver manager
-      RCP< Belos::SolverManager<SC, MV, OP> > solver;
+      RCP<Belos::SolverManager<SC, MV, OP> > solver;
       solver = rcp(new Belos::PseudoBlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
 
       // Perform solve
@@ -431,14 +425,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         // Get the number of iterations for this solve.
         out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-      } catch(...) {
-        out << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+      } catch (...) {
+        out << std::endl
+            << "ERROR:  Belos threw an error! " << std::endl;
       }
       // Check convergence
       if (ret != Belos::Converged)
-        out << std::endl << "ERROR:  Belos did not converge or a warning occured! " << std::endl;
+        out << std::endl
+            << "ERROR:  Belos did not converge or a warning occured! " << std::endl;
       else
-        out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+        out << std::endl
+            << "SUCCESS:  Belos converged!" << std::endl;
     }
 #endif
 
@@ -446,14 +443,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 int main(int argc, char* argv[]) {
   bool success = false;
   bool verbose = true;
 
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
   try {
     const bool throwExceptions     = false;
@@ -462,20 +459,21 @@ int main(int argc, char* argv[]) {
     Teuchos::CommandLineProcessor clp(throwExceptions, recogniseAllOptions);
     Xpetra::Parameters xpetraParameters(clp);
 
-    std::string node = "";  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+    std::string node = "";
+    clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
 
     switch (clp.parse(argc, argv, NULL)) {
-      case Teuchos::CommandLineProcessor::PARSE_ERROR:               return EXIT_FAILURE;
+      case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
 
     if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      return main_<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
+      return main_<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
 #else
       throw MueLu::Exceptions::RuntimeError("Epetra is not available");
 #endif
@@ -486,35 +484,35 @@ int main(int argc, char* argv[]) {
         typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
 #ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
 #else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#  else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#  endif
+#endif
 #endif
       } else if (node == "serial") {
 #ifdef KOKKOS_ENABLE_SERIAL
         typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("Serial node type is disabled");
 #endif
@@ -522,19 +520,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_OPENMP
         typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("OpenMP node type is disabled");
 #endif
@@ -542,19 +540,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_CUDA
         typedef Tpetra::KokkosCompat::KokkosCudaWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("CUDA node type is disabled");
 #endif
@@ -562,19 +560,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_HIP
         typedef Tpetra::KokkosCompat::KokkosHIPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("HIP node type is disabled");
 #endif
@@ -582,22 +580,22 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_SYCL
         typedef Tpetra::KokkosCompatKokkosSYCLWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("SYCL node type is disabled");
-#endif	
+#endif
       } else {
         throw MueLu::Exceptions::RuntimeError("Unrecognized node type");
       }
@@ -605,6 +603,5 @@ int main(int argc, char* argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
-
diff --git a/packages/muelu/research/tawiesn/crada/Driver.cpp b/packages/muelu/research/tawiesn/crada/Driver.cpp
index 60d7d410adb9..f6efdc1886a8 100644
--- a/packages/muelu/research/tawiesn/crada/Driver.cpp
+++ b/packages/muelu/research/tawiesn/crada/Driver.cpp
@@ -70,7 +70,7 @@
 #include <MueLu_Exceptions.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 #include <MueLu_VisualizationHelpers.hpp>
 
 #include <MueLu_Utilities.hpp>
@@ -83,95 +83,90 @@
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockGmresSolMgr.hpp>
 #include <BelosXpetraStatusTestGenResSubNorm.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-  public:
-    using real_type = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
-    using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, Node>;
-
-    ExportVTK() {};
-
-  public:
-
-    void writeFile(std::ofstream& fout, Teuchos::RCP<RealValuedMultiVector>& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol)
-    {
-      using namespace std;
-      typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
-      Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > nodeMap = coordinates->getMap();
-      std::vector<LocalOrdinal> vertices;
-      std::vector<LocalOrdinal> geomSize;
-      LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
-
-      vertices.reserve(numFineNodes);
-      geomSize.reserve(numFineNodes);
-      for(LocalOrdinal i = 0; i < numFineNodes; i++)
-      {
-        vertices.push_back(i);
-        geomSize.push_back(1);
-      }
-
-      Teuchos::ArrayRCP<const real_type> xCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(0));
-      Teuchos::ArrayRCP<const real_type> yCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(1));
-      Teuchos::ArrayRCP<const real_type> zCoords = Teuchos::null;
-      if(coordinates->getNumVectors() == 3) {
-        zCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(2));
-      }
-      Teuchos::ArrayRCP<const Scalar> solData = Teuchos::arcp_reinterpret_cast<const Scalar>(sol->getData(0));
-
-      std::vector<int> uniqueFine = this->makeUnique(vertices);
-      this->writeFileVTKOpening(fout, uniqueFine, geomSize);
-      this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
-
-      std::string indent = "      ";
-      fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      int myRank = coordinates->getMap()->getComm()->getRank();
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        fout << myRank << " ";
-        if(i % 20 == 19)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      // solution vector
-      fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        size_t numVec = coordinates->getNumVectors();
-        for(int d=0; d<(int)numVec; d++)
-          fout << solData[i*numVec+d] << " ";
-        if(i % 3 == 0)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      fout << "      </PointData>" << std::endl; // that is annoying
-
-      this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
-      this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
-      this->writeFileVTKClosing(fout);
-      fout.close();
-
-    };
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  using real_type             = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
+  using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, Node>;
+
+  ExportVTK(){};
+
+ public:
+  void writeFile(std::ofstream& fout, Teuchos::RCP<RealValuedMultiVector>& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol) {
+    using namespace std;
+    typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
+    Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > nodeMap = coordinates->getMap();
+    std::vector<LocalOrdinal> vertices;
+    std::vector<LocalOrdinal> geomSize;
+    LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
+
+    vertices.reserve(numFineNodes);
+    geomSize.reserve(numFineNodes);
+    for (LocalOrdinal i = 0; i < numFineNodes; i++) {
+      vertices.push_back(i);
+      geomSize.push_back(1);
+    }
 
+    Teuchos::ArrayRCP<const real_type> xCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(0));
+    Teuchos::ArrayRCP<const real_type> yCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(1));
+    Teuchos::ArrayRCP<const real_type> zCoords = Teuchos::null;
+    if (coordinates->getNumVectors() == 3) {
+      zCoords = Teuchos::arcp_reinterpret_cast<const real_type>(coordinates->getData(2));
+    }
+    Teuchos::ArrayRCP<const Scalar> solData = Teuchos::arcp_reinterpret_cast<const Scalar>(sol->getData(0));
+
+    std::vector<int> uniqueFine = this->makeUnique(vertices);
+    this->writeFileVTKOpening(fout, uniqueFine, geomSize);
+    this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
+
+    std::string indent = "      ";
+    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    int myRank = coordinates->getMap()->getComm()->getRank();
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      fout << myRank << " ";
+      if (i % 20 == 19)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    // solution vector
+    fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      size_t numVec = coordinates->getNumVectors();
+      for (int d = 0; d < (int)numVec; d++)
+        fout << solData[i * numVec + d] << " ";
+      if (i % 3 == 0)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    fout << "      </PointData>" << std::endl;  // that is annoying
+
+    this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
+    this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
+    this->writeFileVTKClosing(fout);
+    fout.close();
   };
+};
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  using TST       = Teuchos::ScalarTraits<SC>;
-  using MT        = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-  using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using MT                    = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
   using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   // =========================================================================
   // MPI initialization using Teuchos
@@ -180,7 +175,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -189,44 +184,65 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& fancyout = *fancy;
+    Teuchos::FancyOStream& fancyout  = *fancy;
     fancyout.setOutputToRootOnly(0);
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
 
-    //GO nx = 100, ny = 100, nz = 100;
-    //Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName = "driver.xml";      clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
-    std::string belosFileName = "";              clp.setOption("belosXml",              &belosFileName,   "read parameters for Belos from a file.");
-    int    amgAsPrecond     = 1;                 clp.setOption("precond",               &amgAsPrecond,     "apply multigrid as preconditioner");
-    int    amgAsSolver      = 0;                 clp.setOption("fixPoint",              &amgAsSolver,      "apply multigrid as solver");
-    bool   printTimings     = true;              clp.setOption("timings", "notimings",  &printTimings,     "print timings to screen");
-    int    blockedSystem    = 1;                 clp.setOption("split",                 &blockedSystem,    "split system matrix into 2x2 system (default=0)");
-    int    useThyraGIDs     = 0;                 clp.setOption("thyra",                 &useThyraGIDs,     "use Thyra style numbering of GIDs.");
-    int    writeMatricesOPT = -2;                clp.setOption("write",                 &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
-    double tol              = 1e-6;             clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-    std::string krylovMethod = "gmres"; clp.setOption("krylov",                   &krylovMethod,     "outer Krylov method");
-    int maxIts = 100; clp.setOption("maxits",           &maxIts,   "maximum number of Krylov iterations");
-    int output = 1; clp.setOption("output",           &output,   "how often to print Krylov residual history");
-    std::string matrixFileName = "crada1/crada_A.mm"; clp.setOption("matrixfile",            &matrixFileName,   "matrix market file containing matrix");
-    std::string rhsFileName = "crada1/crada_b.mm";    clp.setOption("rhsfile",               &rhsFileName,      "matrix market file containing right-hand side");
-    std::string nspFileName = "crada1/crada_ns.mm";   clp.setOption("nspfile",               &nspFileName,      "matrix market file containing fine level null space");
-    std::string cooFileName = "crada1/crada_coordinates.mm"; clp.setOption("coordinatesfile",&cooFileName,      "matrix market file containing fine level coordinates");
-    std::string spcFileName = "crada1/crada_special.mm"; clp.setOption("specialfile",        &spcFileName,      "matrix market file containing fine level special dofs");
-    int nPDE = 3; clp.setOption("numpdes",           &nPDE,   "number of PDE equations");
-    int nNspVectors = 6; clp.setOption("numnsp", &nNspVectors, "number of nullspace vectors. Only used if null space is read from file. Must be smaller or equal than the number of null space vectors read in from file.");
-    std::string convType = "r0"; clp.setOption("convtype",       &convType,         "convergence type (r0 or none)");
-    std::string strOutputFilename = ""; clp.setOption("output",  &strOutputFilename,"filename prefix for output file name. If empty, no output is written.");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS; break;
+    // GO nx = 100, ny = 100, nz = 100;
+    // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
+
+    std::string xmlFileName = "driver.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+    std::string belosFileName = "";
+    clp.setOption("belosXml", &belosFileName, "read parameters for Belos from a file.");
+    int amgAsPrecond = 1;
+    clp.setOption("precond", &amgAsPrecond, "apply multigrid as preconditioner");
+    int amgAsSolver = 0;
+    clp.setOption("fixPoint", &amgAsSolver, "apply multigrid as solver");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int blockedSystem = 1;
+    clp.setOption("split", &blockedSystem, "split system matrix into 2x2 system (default=0)");
+    int useThyraGIDs = 0;
+    clp.setOption("thyra", &useThyraGIDs, "use Thyra style numbering of GIDs.");
+    int writeMatricesOPT = -2;
+    clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+    double tol = 1e-6;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+    std::string krylovMethod = "gmres";
+    clp.setOption("krylov", &krylovMethod, "outer Krylov method");
+    int maxIts = 100;
+    clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+    int output = 1;
+    clp.setOption("output", &output, "how often to print Krylov residual history");
+    std::string matrixFileName = "crada1/crada_A.mm";
+    clp.setOption("matrixfile", &matrixFileName, "matrix market file containing matrix");
+    std::string rhsFileName = "crada1/crada_b.mm";
+    clp.setOption("rhsfile", &rhsFileName, "matrix market file containing right-hand side");
+    std::string nspFileName = "crada1/crada_ns.mm";
+    clp.setOption("nspfile", &nspFileName, "matrix market file containing fine level null space");
+    std::string cooFileName = "crada1/crada_coordinates.mm";
+    clp.setOption("coordinatesfile", &cooFileName, "matrix market file containing fine level coordinates");
+    std::string spcFileName = "crada1/crada_special.mm";
+    clp.setOption("specialfile", &spcFileName, "matrix market file containing fine level special dofs");
+    int nPDE = 3;
+    clp.setOption("numpdes", &nPDE, "number of PDE equations");
+    int nNspVectors = 6;
+    clp.setOption("numnsp", &nNspVectors, "number of nullspace vectors. Only used if null space is read from file. Must be smaller or equal than the number of null space vectors read in from file.");
+    std::string convType = "r0";
+    clp.setOption("convtype", &convType, "convergence type (r0 or none)");
+    std::string strOutputFilename = "";
+    clp.setOption("output", &strOutputFilename, "filename prefix for output file name. If empty, no output is written.");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS; break;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE; break;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // =========================================================================
@@ -240,42 +256,41 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<Matrix> A = Teuchos::null;
     if (matrixFileName != "") {
       fancyout << "Read matrix from file " << matrixFileName << std::endl;
-      RCP<Matrix> Atest = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm);
-      RCP<const Map>   maptest = Atest->getRowMap();
+      RCP<Matrix> Atest      = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm);
+      RCP<const Map> maptest = Atest->getRowMap();
 
       // re-read matrix and make sure it is properly distributed over all processors
       // make sure that all DOF rows per node are located on same processor.
       GO numGlobalNodes = maptest->getGlobalNumElements() / nPDE;
       GO numMyNodes     = numGlobalNodes / comm->getSize();
-      if (comm->getRank() == comm->getSize()-1)
-        numMyNodes = numGlobalNodes - (comm->getSize()-1) * numMyNodes;
+      if (comm->getRank() == comm->getSize() - 1)
+        numMyNodes = numGlobalNodes - (comm->getSize() - 1) * numMyNodes;
       LO numMyDofs = Teuchos::as<LocalOrdinal>(numMyNodes) * nPDE;
 
       // construct new row map for equally distributing the matrix rows without
       // splitting dofs that belong to the same node
       Teuchos::Array<GO> myDofGIDs(numMyDofs);
-      for(LO i = 0; i < numMyDofs; i++) {
-        if (comm->getRank() == comm->getSize()-1)
+      for (LO i = 0; i < numMyDofs; i++) {
+        if (comm->getRank() == comm->getSize() - 1)
           myDofGIDs[i] = Teuchos::as<GO>(maptest->getGlobalNumElements()) - Teuchos::as<GO>(numMyDofs - i);
         else
           myDofGIDs[i] = Teuchos::as<GO>(comm->getRank() * numMyDofs) + Teuchos::as<GO>(i);
       }
-      RCP<const Map> Arowmap = MapFactory::Build (xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),myDofGIDs(),0,comm);
+      RCP<const Map> Arowmap = MapFactory::Build(xpetraParameters.GetLib(), Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), myDofGIDs(), 0, comm);
 
-      A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFileName,Arowmap);
+      A = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFileName, Arowmap);
       A->SetFixedBlockSize(Teuchos::as<LocalOrdinal>(nPDE));
     }
-    RCP<const Map>   map = A->getRowMap();
-    RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(),nPDE);
+    RCP<const Map> map         = A->getRowMap();
+    RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(), nPDE);
     A->SetFixedBlockSize(nPDE);
     fancyout << "#pdes = " << A->GetFixedBlockSize() << std::endl;
 
     if (nspFileName != "") {
       fancyout << "Read null space from file " << nspFileName << std::endl;
 
-
-      nullspace = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(std::string(nspFileName), A->getRowMap());
-            //nullspace = MultiVectorFactory::Build(A->getRowMap(),6);//haq
+      nullspace = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(std::string(nspFileName), A->getRowMap());
+      // nullspace = MultiVectorFactory::Build(A->getRowMap(),6);//haq
       fancyout << "Found " << nullspace->getNumVectors() << " null space vectors" << std::endl;
       if (nNspVectors > Teuchos::as<int>(nullspace->getNumVectors())) {
         fancyout << "Set number of null space vectors from " << nNspVectors << " to " << nullspace->getNumVectors() << " as only " << nullspace->getNumVectors() << " are provided by " << nspFileName << std::endl;
@@ -286,11 +301,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         nNspVectors = nullspace->getNumVectors();
       }
       if (nNspVectors < Teuchos::as<int>(nullspace->getNumVectors())) {
-        RCP<MultiVector> temp = MultiVectorFactory::Build(A->getDomainMap(),nNspVectors);
-        for(int j=0; j<nNspVectors; j++) {
-          Teuchos::ArrayRCP<SC> tempData = temp->getDataNonConst(j);
-          Teuchos::ArrayRCP<const SC> nsData   = nullspace->getData(j);
-          for (int i=0; i<nsData.size(); ++i) {
+        RCP<MultiVector> temp = MultiVectorFactory::Build(A->getDomainMap(), nNspVectors);
+        for (int j = 0; j < nNspVectors; j++) {
+          Teuchos::ArrayRCP<SC> tempData     = temp->getDataNonConst(j);
+          Teuchos::ArrayRCP<const SC> nsData = nullspace->getData(j);
+          for (int i = 0; i < nsData.size(); ++i) {
             tempData[i] = nsData[i];
           }
         }
@@ -299,46 +314,46 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       }
     } else {
       if (nPDE == 1)
-        nullspace->putScalar( Teuchos::ScalarTraits<SC>::one() );
+        nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
       else {
-        for (int i=0; i<nPDE; ++i) {
+        for (int i = 0; i < nPDE; ++i) {
           Teuchos::ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
-          for (int j=0; j<nsData.size(); ++j) {
+          for (int j = 0; j < nsData.size(); ++j) {
             GO gel = A->getDomainMap()->getGlobalElement(j) - A->getDomainMap()->getIndexBase();
-            if ((gel-i) % nPDE == 0)
+            if ((gel - i) % nPDE == 0)
               nsData[j] = Teuchos::ScalarTraits<SC>::one();
           }
         }
       }
     }
 
-    RCP<RealValuedMultiVector> coordinates = Teuchos::null; //MultiVectorFactory::Build(A->getDomainMap(),1);
+    RCP<RealValuedMultiVector> coordinates = Teuchos::null;  // MultiVectorFactory::Build(A->getDomainMap(),1);
     if (cooFileName != "") {
       TEUCHOS_TEST_FOR_EXCEPTION(map->getLocalNumElements() % A->GetFixedBlockSize() != 0, MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs on proc " << comm->getRank() << " is " << map->getLocalNumElements() << " and not divisible by 3.");
 
-      Teuchos::ArrayView<const GO> dofGidList   = map->getLocalElementList();
-      GlobalOrdinal                indexBase    = map->getIndexBase();
-      LocalOrdinal                 blkSize      = A->GetFixedBlockSize();
+      Teuchos::ArrayView<const GO> dofGidList = map->getLocalElementList();
+      GlobalOrdinal indexBase                 = map->getIndexBase();
+      LocalOrdinal blkSize                    = A->GetFixedBlockSize();
       TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(dofGidList.size()) != Teuchos::as<size_t>(map->getLocalNumElements()), MueLu::Exceptions::RuntimeError, "Driver: Number of local DOFs inconsistent.");
 
-      size_t              numNodes  = dofGidList.size() / blkSize;
-      Teuchos::Array<GO>  nodeList(numNodes);
+      size_t numNodes = dofGidList.size() / blkSize;
+      Teuchos::Array<GO> nodeList(numNodes);
 
       // Amalgamate the map
       for (LO i = 0; i < Teuchos::as<LO>(numNodes); i++)
-        nodeList[i] = (dofGidList[i*blkSize]-indexBase)/blkSize + indexBase;
+        nodeList[i] = (dofGidList[i * blkSize] - indexBase) / blkSize + indexBase;
 
       TEUCHOS_TEST_FOR_EXCEPTION(dofGidList.size() / blkSize != nodeList.size(), MueLu::Exceptions::RuntimeError, "Driver: Number of local DOFs and local Nodes inconsistent.");
 
       GO gCntGIDs  = 0;
       GO glCntGIDs = Teuchos::as<GlobalOrdinal>(nodeList.size());
-      MueLu_sumAll(comm,glCntGIDs,gCntGIDs);
+      MueLu_sumAll(comm, glCntGIDs, gCntGIDs);
 
-      //Teuchos::Array<GlobalOrdinal> eltList(myGIDs);
-      RCP<const Map> myCoordMap = MapFactory::Build (xpetraParameters.GetLib(),gCntGIDs,nodeList(),indexBase,comm);
+      // Teuchos::Array<GlobalOrdinal> eltList(myGIDs);
+      RCP<const Map> myCoordMap = MapFactory::Build(xpetraParameters.GetLib(), gCntGIDs, nodeList(), indexBase, comm);
 
       fancyout << "Read fine level coordinates from file " << cooFileName << std::endl;
-      coordinates = Xpetra::IO<real_type,LO,GO,Node>::ReadMultiVector(std::string(cooFileName), myCoordMap);
+      coordinates = Xpetra::IO<real_type, LO, GO, Node>::ReadMultiVector(std::string(cooFileName), myCoordMap);
       fancyout << "Found " << coordinates->getNumVectors() << " coordinate vectors of length " << myCoordMap->getGlobalNumElements() << std::endl;
       /*TEUCHOS_TEST_FOR_EXCEPTION(myCoordMap->getMinGlobalIndex() != map->getMinGlobalIndex() / blkSize, MueLu::Exceptions::RuntimeError,
           "Driver: Inconsistent minGlobalIndex on proc " << comm->getRank());
@@ -354,55 +369,55 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       std::string line;
       Teuchos::Array<GlobalOrdinal> mySpecialGids;
       Teuchos::Array<GlobalOrdinal> nonSpecialGids;
-      GlobalOrdinal cnt = 0;   // count overall number of gids
-      //GlobalOrdinal mycnt = 0; // count only local gids
-      while ( std::getline(infile, line)) {
-        if(0 == line.find("%")) continue;
-        if(0 == line.find(" ")) {
+      GlobalOrdinal cnt = 0;  // count overall number of gids
+      // GlobalOrdinal mycnt = 0; // count only local gids
+      while (std::getline(infile, line)) {
+        if (0 == line.find("%")) continue;
+        if (0 == line.find(" ")) {
           cnt++;
           GlobalOrdinal gid;
           std::istringstream iss(line);
           iss >> gid;
-          gid--; // note, that the matlab vector starts counting at 1 and not 0!
-          if(map->isNodeGlobalElement(gid)) {
+          gid--;  // note, that the matlab vector starts counting at 1 and not 0!
+          if (map->isNodeGlobalElement(gid)) {
             mySpecialGids.push_back(gid);
-            //mycnt++;
+            // mycnt++;
           }
         }
       }
 
       std::vector<GlobalOrdinal> mySpecialNodeGids;
-      for(size_t k = 0; k < Teuchos::as<size_t>(mySpecialGids.size()); k++) {
-        mySpecialNodeGids.push_back(mySpecialGids[k]/3);
+      for (size_t k = 0; k < Teuchos::as<size_t>(mySpecialGids.size()); k++) {
+        mySpecialNodeGids.push_back(mySpecialGids[k] / 3);
       }
 
-      std::sort(mySpecialNodeGids.begin(),mySpecialNodeGids.end());
+      std::sort(mySpecialNodeGids.begin(), mySpecialNodeGids.end());
       mySpecialNodeGids.erase(std::unique(mySpecialNodeGids.begin(), mySpecialNodeGids.end()), mySpecialNodeGids.end());
 
       cnt = 0;
       Teuchos::Array<GlobalOrdinal> myFinalSpecialGids;
-      for(size_t k = 0; k < mySpecialNodeGids.size(); k++) {
-        myFinalSpecialGids.push_back(mySpecialNodeGids[k]*3);
-        myFinalSpecialGids.push_back(mySpecialNodeGids[k]*3+1);
-        myFinalSpecialGids.push_back(mySpecialNodeGids[k]*3+2);
+      for (size_t k = 0; k < mySpecialNodeGids.size(); k++) {
+        myFinalSpecialGids.push_back(mySpecialNodeGids[k] * 3);
+        myFinalSpecialGids.push_back(mySpecialNodeGids[k] * 3 + 1);
+        myFinalSpecialGids.push_back(mySpecialNodeGids[k] * 3 + 2);
         cnt += 3;
       }
 
       std::cout << "Number of special gids read: " << mySpecialGids.size() << " Final number of special gids: " << myFinalSpecialGids.size() << std::endl;
 
-      //Teuchos::Array<GlobalOrdinal> eltList(mySpecialGids);
-      mySpecialMap    = MapFactory::Build (xpetraParameters.GetLib(),cnt,myFinalSpecialGids(),0,comm);
+      // Teuchos::Array<GlobalOrdinal> eltList(mySpecialGids);
+      mySpecialMap = MapFactory::Build(xpetraParameters.GetLib(), cnt, myFinalSpecialGids(), 0, comm);
 
       // empty processors
-      std::vector<size_t> lelePerProc(comm->getSize(),0);
-      std::vector<size_t> gelePerProc(comm->getSize(),0);
+      std::vector<size_t> lelePerProc(comm->getSize(), 0);
+      std::vector<size_t> gelePerProc(comm->getSize(), 0);
       lelePerProc[comm->getRank()] = mySpecialMap->getLocalNumElements();
-      Teuchos::reduceAll(*comm,Teuchos::REDUCE_MAX,comm->getSize(),&lelePerProc[0],&gelePerProc[0]);
-      if(comm->getRank() == 0) {
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, comm->getSize(), &lelePerProc[0], &gelePerProc[0]);
+      if (comm->getRank() == 0) {
         fancyout << "Distribution of " << cnt << " special dofs over processors:" << std::endl;
         fancyout << "Proc   #DOFs" << std::endl;
-        for(int i=0; i<comm->getSize(); i++) {
-         fancyout  << i << "      " << gelePerProc[i] << std::endl;
+        for (int i = 0; i < comm->getSize(); i++) {
+          fancyout << i << "      " << gelePerProc[i] << std::endl;
         }
       }
     }
@@ -411,15 +426,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     tm = Teuchos::null;
 
     comm->barrier();
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1.2 - LHS and RHS initialization")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1.2 - LHS and RHS initialization")));
 
-    RCP<Vector> X = VectorFactory::Build(map,1);
-    RCP<MultiVector> B = VectorFactory::Build(map,1);
+    RCP<Vector> X      = VectorFactory::Build(map, 1);
+    RCP<MultiVector> B = VectorFactory::Build(map, 1);
 
     if (rhsFileName != "")
-      B = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(std::string(rhsFileName), A->getRowMap());
-    else
-    {
+      B = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(std::string(rhsFileName), A->getRowMap());
+    else {
       // we set seed for reproducibility
       X->setSeed(846930886);
       bool useSameRandomGen = false;
@@ -428,7 +442,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       Teuchos::Array<MT> norms(1);
       B->norm2(norms);
-      //B->scale(1.0/norms[0]);
+      // B->scale(1.0/norms[0]);
     }
 
     X->putScalar(zero);
@@ -443,7 +457,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1.5 - MueLu read XML and feed MueLu")));
     ParameterListInterpreter mueLuFactory(xmlFileName, *comm);
 
-
     RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
 
     // By default, we use Extreme. However, typically the xml file contains verbosity parameter
@@ -457,7 +470,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // - split the linear system in a 2x2 multiphysics problem using Xpetra style gids
     // - split the linear system in a 2x2 multiphysics problem using Thyra style gids
     // - solve the problem as a monolithic linear system
-    if(blockedSystem == 1) {
+    if (blockedSystem == 1) {
       // split matrix and vectors
 
       // create map extractor
@@ -474,16 +487,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       std::cout << "non special gids: " << nonSpecialGids.size() << std::endl;
 
-      //MapFactory::Build (xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),nonSpecialGids(),0,comm);
-      std::vector<size_t> strInfo(1,nPDE);
-      RCP<const Map> myStridedNonSpecialMap = StridedMapFactory::Build(xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),nonSpecialGids(),0,strInfo,comm);
-      RCP<const Map> myStridedSpecialMap    = StridedMapFactory::Build(xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),specialGids()   ,0,strInfo,comm);
+      // MapFactory::Build (xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),nonSpecialGids(),0,comm);
+      std::vector<size_t> strInfo(1, nPDE);
+      RCP<const Map> myStridedNonSpecialMap = StridedMapFactory::Build(xpetraParameters.GetLib(), Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), nonSpecialGids(), 0, strInfo, comm);
+      RCP<const Map> myStridedSpecialMap    = StridedMapFactory::Build(xpetraParameters.GetLib(), Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), specialGids(), 0, strInfo, comm);
       // RCP<const Map> myStridedSpecialMap    = StridedMapFactory::Build(mySpecialMap, strInfo);
 
-      //std::cout << "map " << map->getMaxAllGlobalIndex() << " nonspecial " << myStridedNonSpecialMap->getMinAllGlobalIndex() << " " << myStridedNonSpecialMap->getMaxAllGlobalIndex() << " (" << myStridedNonSpecialMap->getGlobalNumElements() << ") special " << mySpecialMap->getMinAllGlobalIndex() << " " << mySpecialMap->getMaxAllGlobalIndex() << "(" << myStridedSpecialMap->getGlobalNumElements() << ")" << std::endl;
-      //std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myNonSpecialMap)->getEpetra_Map() << std::endl;
-      //std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myStridedSpecialMap)->getEpetra_Map() << std::endl;
-
+      // std::cout << "map " << map->getMaxAllGlobalIndex() << " nonspecial " << myStridedNonSpecialMap->getMinAllGlobalIndex() << " " << myStridedNonSpecialMap->getMaxAllGlobalIndex() << " (" << myStridedNonSpecialMap->getGlobalNumElements() << ") special " << mySpecialMap->getMinAllGlobalIndex() << " " << mySpecialMap->getMaxAllGlobalIndex() << "(" << myStridedSpecialMap->getGlobalNumElements() << ")" << std::endl;
+      // std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myNonSpecialMap)->getEpetra_Map() << std::endl;
+      // std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myStridedSpecialMap)->getEpetra_Map() << std::endl;
 
       // We always build an Xpetra style map extractor with unique global ids
       TEUCHOS_TEST_FOR_EXCEPTION(map->getLocalNumElements() != myStridedNonSpecialMap->getLocalNumElements() + myStridedSpecialMap->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs on proc " << comm->getRank() << " is " << map->getLocalNumElements() << " and does not match the sum of the partial maps of size " << myStridedNonSpecialMap->getLocalNumElements() << " and " << myStridedSpecialMap->getLocalNumElements());
@@ -493,15 +505,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       xmaps.push_back(myStridedSpecialMap);
 
       // Xpetra mode
-      Teuchos::RCP<const Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map,xmaps);
+      Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, xmaps);
 
       // split null space vectors
-      RCP<MultiVector> nullspace1 = map_extractor->ExtractVector(nullspace,0);
-      RCP<MultiVector> nullspace2 = map_extractor->ExtractVector(nullspace,1);
+      RCP<MultiVector> nullspace1 = map_extractor->ExtractVector(nullspace, 0);
+      RCP<MultiVector> nullspace2 = map_extractor->ExtractVector(nullspace, 1);
 
-      bool bThyraMode = (useThyraGIDs==1) ? true : false;
-      Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bOp =
-          Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SplitMatrix(*A,map_extractor,map_extractor,Teuchos::null,bThyraMode);
+      bool bThyraMode = (useThyraGIDs == 1) ? true : false;
+      Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOp =
+          Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SplitMatrix(*A, map_extractor, map_extractor, Teuchos::null, bThyraMode);
 
       // TODO plausibility checks
       // TODO set number of dofs per node
@@ -513,8 +525,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       RCP<RealValuedMultiVector> coordinates1 = Teuchos::null;
       RCP<RealValuedMultiVector> coordinates2 = Teuchos::null;
       if (coordinates != Teuchos::null) {
-        TEUCHOS_TEST_FOR_EXCEPTION( myStridedNonSpecialMap->getLocalNumElements() % 3 != 0, MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs of non-special map on proc " << comm->getRank() << " is " << myStridedNonSpecialMap->getLocalNumElements() << " and cannot divided by 3");
-        TEUCHOS_TEST_FOR_EXCEPTION( myStridedSpecialMap->getLocalNumElements() % 3 != 0, MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs of special map on proc " << comm->getRank() << " is " << myStridedSpecialMap->getLocalNumElements() << " and cannot divided by 3");
+        TEUCHOS_TEST_FOR_EXCEPTION(myStridedNonSpecialMap->getLocalNumElements() % 3 != 0, MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs of non-special map on proc " << comm->getRank() << " is " << myStridedNonSpecialMap->getLocalNumElements() << " and cannot divided by 3");
+        TEUCHOS_TEST_FOR_EXCEPTION(myStridedSpecialMap->getLocalNumElements() % 3 != 0, MueLu::Exceptions::RuntimeError, "Driver: Number of DOFs of special map on proc " << comm->getRank() << " is " << myStridedSpecialMap->getLocalNumElements() << " and cannot divided by 3");
 
         Teuchos::Array<GlobalOrdinal> nonSpecialCoordGids;
         Teuchos::Array<GlobalOrdinal> SpecialCoordGids;
@@ -530,64 +542,63 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
               nonSpecialCoordGids.append(gid);
               break;
             } else
-              TEUCHOS_TEST_FOR_EXCEPTION( true, MueLu::Exceptions::RuntimeError, "Driver: DofGid " << dofgid << " is neither contained in special nor in non-special map.");
+              TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Driver: DofGid " << dofgid << " is neither contained in special nor in non-special map.");
           }
         }
 
-        RCP<const Map> myNonSpecialCoordsMap = MapFactory::Build(xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),nonSpecialCoordGids(),0,comm);
-        TEUCHOS_TEST_FOR_EXCEPTION( myStridedNonSpecialMap->getLocalNumElements() / 3 != myNonSpecialCoordsMap->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "Driver: Number of entries in non-special node map is inconsistent");
+        RCP<const Map> myNonSpecialCoordsMap = MapFactory::Build(xpetraParameters.GetLib(), Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), nonSpecialCoordGids(), 0, comm);
+        TEUCHOS_TEST_FOR_EXCEPTION(myStridedNonSpecialMap->getLocalNumElements() / 3 != myNonSpecialCoordsMap->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "Driver: Number of entries in non-special node map is inconsistent");
 
-        RCP<const Map> mySpecialCoordsMap = MapFactory::Build(xpetraParameters.GetLib(),Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),SpecialCoordGids(),0,comm);
-        TEUCHOS_TEST_FOR_EXCEPTION( myStridedSpecialMap->getLocalNumElements() / 3 != mySpecialCoordsMap->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "Driver: Number of entries in non-special node map is inconsistent");
+        RCP<const Map> mySpecialCoordsMap = MapFactory::Build(xpetraParameters.GetLib(), Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), SpecialCoordGids(), 0, comm);
+        TEUCHOS_TEST_FOR_EXCEPTION(myStridedSpecialMap->getLocalNumElements() / 3 != mySpecialCoordsMap->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "Driver: Number of entries in non-special node map is inconsistent");
 
         std::vector<Teuchos::RCP<const Map> > nodexmaps;
         nodexmaps.push_back(myNonSpecialCoordsMap);
         nodexmaps.push_back(mySpecialCoordsMap);
 
-        Teuchos::RCP<const Xpetra::MapExtractor<real_type,LO,GO,NO> > nodemap_extractor = Xpetra::MapExtractorFactory<real_type,LO,GO,NO>::Build(coordinates->getMap(),nodexmaps);
+        Teuchos::RCP<const Xpetra::MapExtractor<real_type, LO, GO, NO> > nodemap_extractor = Xpetra::MapExtractorFactory<real_type, LO, GO, NO>::Build(coordinates->getMap(), nodexmaps);
 
         // split coordinate vectors
-        coordinates1 = nodemap_extractor->ExtractVector(coordinates,0);
-        coordinates2 = nodemap_extractor->ExtractVector(coordinates,1);
-        //std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myNonSpecialCoordsMap)->getEpetra_Map() << std::endl;
-        //std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(mySpecialCoordsMap)->getEpetra_Map() << std::endl;
+        coordinates1 = nodemap_extractor->ExtractVector(coordinates, 0);
+        coordinates2 = nodemap_extractor->ExtractVector(coordinates, 1);
+        // std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(myNonSpecialCoordsMap)->getEpetra_Map() << std::endl;
+        // std::cout << Teuchos::rcp_dynamic_cast<const Xpetra::EpetraMapT<int, Node> >(mySpecialCoordsMap)->getEpetra_Map() << std::endl;
       }
 
-
-      if(bThyraMode == false) {
+      if (bThyraMode == false) {
         // use Xpetra style GIDs
-        H->GetLevel(0)->Set("A",            Teuchos::rcp_dynamic_cast<Matrix>(bOp));
-        H->GetLevel(0)->Set("Nullspace1",   nullspace1);
-        H->GetLevel(0)->Set("Nullspace2",   nullspace2);
-        H->GetLevel(0)->Set("Coordinates",  coordinates); // TODO split coordinates for rebalancing! (or provide the full vector in the right map and split it in the factories!)
+        H->GetLevel(0)->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));
+        H->GetLevel(0)->Set("Nullspace1", nullspace1);
+        H->GetLevel(0)->Set("Nullspace2", nullspace2);
+        H->GetLevel(0)->Set("Coordinates", coordinates);  // TODO split coordinates for rebalancing! (or provide the full vector in the right map and split it in the factories!)
         H->GetLevel(0)->Set("Coordinates1", coordinates1);
         H->GetLevel(0)->Set("Coordinates2", coordinates2);
-        if(mySpecialMap!=Teuchos::null) H->GetLevel(0)->Set("map SpecialMap", mySpecialMap);
+        if (mySpecialMap != Teuchos::null) H->GetLevel(0)->Set("map SpecialMap", mySpecialMap);
       } else {
         // use Thyra style GIDs
-        RCP<MultiVector> nsp1shrinked             = Xpetra::MatrixUtils<SC,LO,GO,NO>::xpetraGidNumbering2ThyraGidNumbering(*nullspace1);
-        RCP<MultiVector> nsp2shrinked             = Xpetra::MatrixUtils<SC,LO,GO,NO>::xpetraGidNumbering2ThyraGidNumbering(*nullspace2);
-        RCP<RealValuedMultiVector> coordsshrinked = Xpetra::MatrixUtils<real_type,LO,GO,NO>::xpetraGidNumbering2ThyraGidNumbering(*coordinates);
-        H->GetLevel(0)->Set("A",           Teuchos::rcp_dynamic_cast<Matrix>(bOp));
-        H->GetLevel(0)->Set("Nullspace1",  nsp1shrinked);
-        H->GetLevel(0)->Set("Nullspace2",  nsp2shrinked);
+        RCP<MultiVector> nsp1shrinked             = Xpetra::MatrixUtils<SC, LO, GO, NO>::xpetraGidNumbering2ThyraGidNumbering(*nullspace1);
+        RCP<MultiVector> nsp2shrinked             = Xpetra::MatrixUtils<SC, LO, GO, NO>::xpetraGidNumbering2ThyraGidNumbering(*nullspace2);
+        RCP<RealValuedMultiVector> coordsshrinked = Xpetra::MatrixUtils<real_type, LO, GO, NO>::xpetraGidNumbering2ThyraGidNumbering(*coordinates);
+        H->GetLevel(0)->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));
+        H->GetLevel(0)->Set("Nullspace1", nsp1shrinked);
+        H->GetLevel(0)->Set("Nullspace2", nsp2shrinked);
         H->GetLevel(0)->Set("Coordinates", coordsshrinked);  // TODO split coordinates for rebalancing! (or provide the full vector in the right map and split it in the factories!)
-        if(mySpecialMap!=Teuchos::null) {
-          RCP<const Map> specialmapshrinked = Xpetra::MapUtils<LocalOrdinal, GlobalOrdinal, Node>::shrinkMapGIDs(*mySpecialMap,*mySpecialMap);
+        if (mySpecialMap != Teuchos::null) {
+          RCP<const Map> specialmapshrinked = Xpetra::MapUtils<LocalOrdinal, GlobalOrdinal, Node>::shrinkMapGIDs(*mySpecialMap, *mySpecialMap);
           H->GetLevel(0)->Set("map SpecialMap", Teuchos::rcp_const_cast<Map>(specialmapshrinked));
         }
 
         // rearrange contents of rhs vector B
         RCP<const MapExtractor> thy_map_extractor = bOp->getRangeMapExtractor();
-        RCP<MultiVector> Bshrinked = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(thy_map_extractor->getFullMap(),B->getNumVectors(),true);
+        RCP<MultiVector> Bshrinked                = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(thy_map_extractor->getFullMap(), B->getNumVectors(), true);
 
         size_t numMaps = map_extractor->NumMaps();
-        for(size_t k = 0; k < numMaps; k++) {
-          RCP<MultiVector> partVec         = map_extractor->ExtractVector(B, k);
-          RCP<MultiVector> newPartVec      = thy_map_extractor->getVector(k, B->getNumVectors(), false); // we need real GIDs (not zero based Thyra GIDs)
+        for (size_t k = 0; k < numMaps; k++) {
+          RCP<MultiVector> partVec    = map_extractor->ExtractVector(B, k);
+          RCP<MultiVector> newPartVec = thy_map_extractor->getVector(k, B->getNumVectors(), false);  // we need real GIDs (not zero based Thyra GIDs)
           // copy data
           for (size_t c = 0; c < partVec->getNumVectors(); c++) {
-            Teuchos::ArrayRCP< const Scalar > data = partVec->getData(c);
+            Teuchos::ArrayRCP<const Scalar> data = partVec->getData(c);
             for (size_t r = 0; r < partVec->getLocalLength(); r++) {
               newPartVec->replaceLocalValue(Teuchos::as<LocalOrdinal>(r), c, data[r]);
             }
@@ -596,17 +607,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         }
 
         B = Bshrinked;
-        X = VectorFactory::Build(thy_map_extractor->getFullMap(),1);
+        X = VectorFactory::Build(thy_map_extractor->getFullMap(), 1);
         X->putScalar(zero);
 
         // TODO the ordering of the solution vector X is different!
       }
     } else {
       // standard (non-blocked) case
-      H->GetLevel(0)->Set("A",           A);
-      H->GetLevel(0)->Set("Nullspace",   nullspace);
+      H->GetLevel(0)->Set("A", A);
+      H->GetLevel(0)->Set("Nullspace", nullspace);
       H->GetLevel(0)->Set("Coordinates", coordinates);
-      if(mySpecialMap!=Teuchos::null) H->GetLevel(0)->Set("map SpecialMap", mySpecialMap);
+      if (mySpecialMap != Teuchos::null) H->GetLevel(0)->Set("map SpecialMap", mySpecialMap);
     }
 
     comm->barrier();
@@ -632,31 +643,31 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     comm->barrier();
     if (amgAsSolver) {
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - Fixed Point Solve")));
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - Fixed Point Solve")));
 
       H->IsPreconditioner(false);
       Teuchos::Array<MT> norms(1);
-      norms = Utilities::ResidualNorm(*A,*X,*B);
+      norms = Utilities::ResidualNorm(*A, *X, *B);
       std::cout << "                iter:    0           residual = " << norms[0] << std::endl;
-      for (int i=0; i< maxIts; ++i) {
+      for (int i = 0; i < maxIts; ++i) {
         H->Iterate(*B, *X);
-        norms = Utilities::ResidualNorm(*A,*X,*B);
-        std::cout << "                iter:    " << i+1 << "           residual = " << norms[0] << std::endl;
+        norms = Utilities::ResidualNorm(*A, *X, *B);
+        std::cout << "                iter:    " << i + 1 << "           residual = " << norms[0] << std::endl;
       }
 
     } else if (amgAsPrecond) {
 #ifdef HAVE_MUELU_BELOS
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - Belos Solve")));
       // Operator and Multivector type that will be used with Belos
-      typedef MultiVector          MV;
+      typedef MultiVector MV;
       typedef Belos::OperatorT<MV> OP;
       H->IsPreconditioner(true);
 
       // Define Operator and Preconditioner
-      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
       // Construct a Belos LinearProblem object
-      RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+      RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
       belosProblem->setRightPrec(belosPrec);
       bool set = belosProblem->setProblem();
       if (set == false) {
@@ -666,15 +677,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       // Belos parameter list
       Teuchos::ParameterList belosList;
-      if(belosFileName == "") {
-        belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-        belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-        belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-        belosList.set("Output Frequency",      output);
-        belosList.set("Output Style",          Belos::Brief);
+      if (belosFileName == "") {
+        belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+        belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+        belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+        belosList.set("Output Frequency", output);
+        belosList.set("Output Style", Belos::Brief);
         if (convType == "none") {
-          belosList.set("Explicit Residual Scaling",  "None");
-          belosList.set("Implicit Residual Scaling",  "None");
+          belosList.set("Explicit Residual Scaling", "None");
+          belosList.set("Implicit Residual Scaling", "None");
         }
       } else {
         Teuchos::updateParametersFromXmlFileAndBroadcast(belosFileName, Teuchos::Ptr<Teuchos::ParameterList>(&belosList), *comm);
@@ -711,7 +722,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       }
 
       // Create an iterative solver manager
-      RCP< Belos::SolverManager<SC, MV, OP> > solver;
+      RCP<Belos::SolverManager<SC, MV, OP> > solver;
       if (krylovMethod == "cg") {
         solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
       } else if (krylovMethod == "gmres") {
@@ -728,22 +739,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         // Get the number of iterations for this solve.
         fancyout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-      } catch(...) {
-        fancyout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+      } catch (...) {
+        fancyout << std::endl
+                 << "ERROR:  Belos threw an error! " << std::endl;
       }
       // Check convergence
       if (ret != Belos::Converged)
-        fancyout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+        fancyout << std::endl
+                 << "ERROR:  Belos did not converge! " << std::endl;
       else
-        fancyout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-#endif //ifdef HAVE_MUELU_BELOS
+        fancyout << std::endl
+                 << "SUCCESS:  Belos converged!" << std::endl;
+#endif  // ifdef HAVE_MUELU_BELOS
     }
     comm->barrier();
     tm = Teuchos::null;
 
     if (strOutputFilename.empty() == false) {
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Export solution in VTK")));
-      if(comm->getSize() > 1) {
+      if (comm->getSize() > 1) {
         std::stringstream ss;
         ss << "-proc" << comm->getRank();
         strOutputFilename.append(ss.str());
@@ -751,19 +765,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       strOutputFilename.append(".vtu");
 
       std::ofstream fout(strOutputFilename);
-      ExportVTK<Scalar,LocalOrdinal,GlobalOrdinal,Node> expVTK;
-      expVTK.writeFile(fout,coordinates,X);
+      ExportVTK<Scalar, LocalOrdinal, GlobalOrdinal, Node> expVTK;
+      expVTK.writeFile(fout, coordinates, X);
       fout.close();
 
       size_t start_pos = strOutputFilename.find(".vtu");
       strOutputFilename.replace(start_pos, 4, ".m");
-      Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write(strOutputFilename,*X);
+      Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(strOutputFilename, *X);
     }
 
     // print timings
 
     comm->barrier();
-    tm = Teuchos::null;
+    tm                = Teuchos::null;
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings) {
@@ -775,17 +789,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
-
diff --git a/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_decl.hpp b/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_decl.hpp
index 6b38a9f96750..588762b95df3 100644
--- a/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_decl.hpp
@@ -70,105 +70,103 @@
 /*!
   @class BrickAggregationFactory
   @brief Aggregation method for generating "brick" aggregates.  It also does "hotdogs" and "pancakes."
-  
+
   This factory can generate aggregates of size 1, 2 or 3 in each dimension, in any combination.
 
 */
 
 namespace MueLu {
 
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class BrickAggregationFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class BrickAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_BRICKAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
-  private:
-    typedef Teuchos::ScalarTraits<Scalar>                                  STS;
-
-    // Comparator for doubles
-    // Generally, the coordinates for coarser levels would come out of averaging of fine level coordinates
-    // It is possible that the result of the averaging differs slightly between clusters, as we might have
-    // 3x2 and 2x2 cluster which would result in averaging 6 and 4 y-coordinates respectively, leading to
-    // slightly different results.
-    // Therefore, we hardcode a constant so that close points are considered the same.
-    class compare {
-    public:
-      bool operator()(const Scalar& x, const Scalar& y) const {
-        if (STS::magnitude(x - y) < 1e-14)
-          return false;
-        return STS::real(x) < STS::real(y);
-      }
-    };
-    typedef std::map<Scalar,GlobalOrdinal,compare> container;
+ private:
+  typedef Teuchos::ScalarTraits<Scalar> STS;
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+  // Comparator for doubles
+  // Generally, the coordinates for coarser levels would come out of averaging of fine level coordinates
+  // It is possible that the result of the averaging differs slightly between clusters, as we might have
+  // 3x2 and 2x2 cluster which would result in averaging 6 and 4 y-coordinates respectively, leading to
+  // slightly different results.
+  // Therefore, we hardcode a constant so that close points are considered the same.
+  class compare {
+   public:
+    bool operator()(const Scalar& x, const Scalar& y) const {
+      if (STS::magnitude(x - y) < 1e-14)
+        return false;
+      return STS::real(x) < STS::real(y);
+    }
+  };
+  typedef std::map<Scalar, GlobalOrdinal, compare> container;
 
-    //! Constructor.
-    BrickAggregationFactory() : nDim_(-1), nx_(-1), ny_(-1), nz_(-1), bx_(-1), by_(-1), bz_(-1) { };
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~BrickAggregationFactory() { }
+  //! Constructor.
+  BrickAggregationFactory()
+    : nDim_(-1)
+    , nx_(-1)
+    , ny_(-1)
+    , nz_(-1)
+    , bx_(-1)
+    , by_(-1)
+    , bz_(-1){};
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~BrickAggregationFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    // Options shared by all aggregation algorithms
+  //@}
 
-    //! Input
-    //@{
+  // Options shared by all aggregation algorithms
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*! @brief Build aggregates. */
-    void Build(Level &currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  /*! @brief Build aggregates. */
+  void Build(Level& currentLevel) const;
 
-  private:
-    void Setup(const RCP<const Teuchos::Comm<int> >& comm, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> >& coords, const RCP<const Map>& map) const;
-    RCP<container> Construct1DMap(const RCP<const Teuchos::Comm<int> >& comm, const ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& x) const;
+  //@}
 
-    void BuildGraph(Level& currentLevel, const RCP<Matrix>& A) const;
+ private:
+  void Setup(const RCP<const Teuchos::Comm<int> >& comm, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coords, const RCP<const Map>& map) const;
+  RCP<container> Construct1DMap(const RCP<const Teuchos::Comm<int> >& comm, const ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& x) const;
 
+  void BuildGraph(Level& currentLevel, const RCP<Matrix>& A) const;
 
-    bool        isDirichlet(LocalOrdinal LID) const;
-    bool           isRoot  (LocalOrdinal LID) const;
-    GlobalOrdinal getRoot  (LocalOrdinal LID) const;
-    GlobalOrdinal getAggGID(LocalOrdinal LID) const;
+  bool isDirichlet(LocalOrdinal LID) const;
+  bool isRoot(LocalOrdinal LID) const;
+  GlobalOrdinal getRoot(LocalOrdinal LID) const;
+  GlobalOrdinal getAggGID(LocalOrdinal LID) const;
 
-    void getIJK(LocalOrdinal LID, int &i, int &j, int &k) const;
-    void getAggIJK(LocalOrdinal LID, int &i, int &j, int &k) const;
+  void getIJK(LocalOrdinal LID, int& i, int& j, int& k) const;
+  void getAggIJK(LocalOrdinal LID, int& i, int& j, int& k) const;
 
-    mutable
-     int nDim_;
-    mutable
-     RCP<container> xMap_, yMap_, zMap_;
-    mutable
-     ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> x_, y_, z_;
-    mutable
-     int nx_, ny_, nz_;
-    mutable
-     int bx_, by_, bz_;
-    mutable
-     bool dirichletX_,dirichletY_,dirichletZ_;
-    mutable
-     int naggx_, naggy_, naggz_;
+  mutable int nDim_;
+  mutable RCP<container> xMap_, yMap_, zMap_;
+  mutable ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> x_, y_, z_;
+  mutable int nx_, ny_, nz_;
+  mutable int bx_, by_, bz_;
+  mutable bool dirichletX_, dirichletY_, dirichletZ_;
+  mutable int naggx_, naggy_, naggz_;
 
-    mutable
-     std::map<GlobalOrdinal,GlobalOrdinal> revMap_;
-  }; // class BrickAggregationFactory
+  mutable std::map<GlobalOrdinal, GlobalOrdinal> revMap_;
+};  // class BrickAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_BRICKAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_BRICKAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_def.hpp b/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_def.hpp
index aa6863becd89..7b7ab9f9e07e 100644
--- a/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/BrickAggregation/MueLu_BrickAggregationFactory_def.hpp
@@ -70,518 +70,506 @@
 #include "MueLu_Graph.hpp"
 #include "MueLu_LWGraph.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: brick x size");
-    SET_VALID_ENTRY("aggregation: brick y size");
-    SET_VALID_ENTRY("aggregation: brick z size");
-    SET_VALID_ENTRY("aggregation: brick x Dirichlet");
-    SET_VALID_ENTRY("aggregation: brick y Dirichlet");
-    SET_VALID_ENTRY("aggregation: brick z Dirichlet");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",           Teuchos::null, "Generating factory for matrix");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for coordinates");
-    return validParamList;
+  SET_VALID_ENTRY("aggregation: brick x size");
+  SET_VALID_ENTRY("aggregation: brick y size");
+  SET_VALID_ENTRY("aggregation: brick z size");
+  SET_VALID_ENTRY("aggregation: brick x Dirichlet");
+  SET_VALID_ENTRY("aggregation: brick y Dirichlet");
+  SET_VALID_ENTRY("aggregation: brick z Dirichlet");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for matrix");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for coordinates");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "Coordinates");
+}
+
+// The current implementation cannot deal with bricks larger than 3x3(x3) in
+// parallel. The reason is that aggregation infrastructure in place has
+// major drawbacks.
+//
+// Aggregates class is constructed with a help of a provided map, either
+// taken from a graph, or provided directly. This map is usually taken to be
+// a column map of a matrix. The reason for that is that if we have an
+// overlapped aggregation, we want the processor owning aggregates to store
+// agg id for all nodes in this aggregate. If we used row map, there would
+// be no way for the processor to know whether there are some other nodes on
+// a different processor which belong to its aggregate. On the other hand,
+// using column map allows both vertex2AggId and procWinner arrays in
+// Aggregates class to store some extra data, such as whether nodes belonging
+// to a different processor belong to this processor aggregate.
+//
+// The drawback of this is that it stores only overlap=1 data. For aggressive
+// coarsening, such a brick aggregation with a large single dimension of
+// brick, it could happen that we need to know depth two or more extra nodes
+// in the other processor subdomain.
+//
+// Another issue is that we may have some implicit connection between
+// aggregate map and maps of A used in the construction of a tentative
+// prolongator.
+//
+// Another issue is that it seems that some info is unused or not required.
+// Specifically, it seems that if a node belongs to an aggregate on a
+// different processor, we don't actually need to set vertex2AggId and
+// procWinner, despite the following comment in
+// Aggregates decl:
+//      vertex2AggId[k] gives a local id
+//      corresponding to the aggregate to which
+//      local id k has been assigned.  While k
+//      is the local id on my processor (MyPID)
+//      vertex2AggId[k] is the local id on the
+//      processor which actually owns the
+//      aggregate. This owning processor has id
+//      given by procWinner[k].
+// It is possible that that info is only used during arbitration in
+// CoupledAggregationFactory.
+//
+// The steps that we need to do to resolve this issue:
+//   - Break the link between maps in TentativePFactory, allowing any maps in Aggregates
+//   - Allow Aggregates to construct their own maps, if necessary, OR
+//   - construct aggregates based on row map
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> MultiVector_d;
+
+  const ParameterList& pL   = GetParameterList();
+  RCP<MultiVector_d> coords = Get<RCP<MultiVector_d> >(currentLevel, "Coordinates");
+  RCP<Matrix> A             = Get<RCP<Matrix> >(currentLevel, "A");
+  RCP<const Map> rowMap     = A->getRowMap();
+  RCP<const Map> colMap     = A->getColMap();
+  GO GO_INVALID             = Teuchos::OrdinalTraits<GO>::invalid();
+
+  RCP<const Teuchos::Comm<int> > comm = rowMap->getComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  int numPoints = colMap->getLocalNumElements();
+
+  bx_ = pL.get<int>("aggregation: brick x size");
+  by_ = pL.get<int>("aggregation: brick y size");
+  bz_ = pL.get<int>("aggregation: brick z size");
+
+  dirichletX_ = pL.get<bool>("aggregation: brick x Dirichlet");
+  dirichletY_ = pL.get<bool>("aggregation: brick y Dirichlet");
+  dirichletZ_ = pL.get<bool>("aggregation: brick z Dirichlet");
+  if (dirichletX_) GetOStream(Runtime0) << "Dirichlet boundaries in the x direction" << std::endl;
+  if (dirichletY_) GetOStream(Runtime0) << "Dirichlet boundaries in the y direction" << std::endl;
+  if (dirichletZ_) GetOStream(Runtime0) << "Dirichlet boundaries in the z direction" << std::endl;
+
+  if (numProcs > 1) {
+    // TODO: deal with block size > 1  (see comments above)
+    // TEUCHOS_TEST_FOR_EXCEPTION(bx_ > 3 || by_ > 3 || bz_ > 3, Exceptions::RuntimeError, "Currently cannot deal with brick size > 3");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "Coordinates");
+  RCP<MultiVector_d> overlappedCoords = coords;
+  RCP<const Import> importer          = ImportFactory::Build(coords->getMap(), colMap);
+  if (!importer.is_null()) {
+    overlappedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(colMap, coords->getNumVectors());
+    overlappedCoords->doImport(*coords, *importer, Xpetra::INSERT);
   }
 
-  // The current implementation cannot deal with bricks larger than 3x3(x3) in
-  // parallel. The reason is that aggregation infrastructure in place has
-  // major drawbacks.
-  //
-  // Aggregates class is constructed with a help of a provided map, either
-  // taken from a graph, or provided directly. This map is usually taken to be
-  // a column map of a matrix. The reason for that is that if we have an
-  // overlapped aggregation, we want the processor owning aggregates to store
-  // agg id for all nodes in this aggregate. If we used row map, there would
-  // be no way for the processor to know whether there are some other nodes on
-  // a different processor which belong to its aggregate. On the other hand,
-  // using column map allows both vertex2AggId and procWinner arrays in
-  // Aggregates class to store some extra data, such as whether nodes belonging
-  // to a different processor belong to this processor aggregate.
-  //
-  // The drawback of this is that it stores only overlap=1 data. For aggressive
-  // coarsening, such a brick aggregation with a large single dimension of
-  // brick, it could happen that we need to know depth two or more extra nodes
-  // in the other processor subdomain.
-  //
-  // Another issue is that we may have some implicit connection between
-  // aggregate map and maps of A used in the construction of a tentative
-  // prolongator.
-  //
-  // Another issue is that it seems that some info is unused or not required.
-  // Specifically, it seems that if a node belongs to an aggregate on a
-  // different processor, we don't actually need to set vertex2AggId and
-  // procWinner, despite the following comment in
-  // Aggregates decl:
-  //      vertex2AggId[k] gives a local id
-  //      corresponding to the aggregate to which
-  //      local id k has been assigned.  While k
-  //      is the local id on my processor (MyPID)
-  //      vertex2AggId[k] is the local id on the
-  //      processor which actually owns the
-  //      aggregate. This owning processor has id
-  //      given by procWinner[k].
-  // It is possible that that info is only used during arbitration in
-  // CoupledAggregationFactory.
-  //
-  // The steps that we need to do to resolve this issue:
-  //   - Break the link between maps in TentativePFactory, allowing any maps in Aggregates
-  //   - Allow Aggregates to construct their own maps, if necessary, OR
-  //   - construct aggregates based on row map
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> MultiVector_d;
-
-    const ParameterList& pL = GetParameterList();
-    RCP<MultiVector_d> coords = Get<RCP<MultiVector_d> >(currentLevel, "Coordinates");
-    RCP<Matrix>        A      = Get< RCP<Matrix> >      (currentLevel, "A");
-    RCP<const Map>     rowMap = A->getRowMap();
-    RCP<const Map>     colMap = A->getColMap();
-    GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
-
-    RCP<const Teuchos::Comm<int> > comm = rowMap->getComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    int numPoints = colMap->getLocalNumElements();
-
-    bx_ = pL.get<int>("aggregation: brick x size");
-    by_ = pL.get<int>("aggregation: brick y size");
-    bz_ = pL.get<int>("aggregation: brick z size");
-
-    dirichletX_ = pL.get<bool>("aggregation: brick x Dirichlet");
-    dirichletY_ = pL.get<bool>("aggregation: brick y Dirichlet");
-    dirichletZ_ = pL.get<bool>("aggregation: brick z Dirichlet");
-    if(dirichletX_) GetOStream(Runtime0) << "Dirichlet boundaries in the x direction"<<std::endl;
-    if(dirichletY_) GetOStream(Runtime0) << "Dirichlet boundaries in the y direction"<<std::endl;
-    if(dirichletZ_) GetOStream(Runtime0) << "Dirichlet boundaries in the z direction"<<std::endl;
-
-    if (numProcs > 1) {
-      // TODO: deal with block size > 1  (see comments above)
-      //TEUCHOS_TEST_FOR_EXCEPTION(bx_ > 3 || by_ > 3 || bz_ > 3, Exceptions::RuntimeError, "Currently cannot deal with brick size > 3");
-    }
-
-    RCP<MultiVector_d> overlappedCoords = coords;
-    RCP<const Import> importer = ImportFactory::Build(coords->getMap(), colMap);
-    if (!importer.is_null()) {
-      overlappedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(colMap, coords->getNumVectors());
-      overlappedCoords->doImport(*coords, *importer, Xpetra::INSERT);
-    }
-
-    // Setup misc structures
-    // Logically, we construct enough data to query topological information of a rectangular grid
-    Setup(comm, overlappedCoords, colMap);
-
-    GetOStream(Runtime0) << "Using brick size: " << bx_
-                                                 << (nDim_ > 1 ? "x " + toString(by_) : "")
-                                                 << (nDim_ > 2 ? "x " + toString(bz_) : "") << std::endl;
-
-    // Build the graph
-    BuildGraph(currentLevel,A);
-
-    // Construct aggregates
-    RCP<Aggregates> aggregates = rcp(new Aggregates(colMap));
-    aggregates->setObjectLabel("Brick");
-
-    ArrayRCP<LO> vertex2AggId = aggregates->GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates->GetProcWinner()  ->getDataNonConst(0);
-
-    // In the first pass, we set a mapping from a vertex to aggregate global id. We deal with a structured
-    // rectangular mesh, therefore we know the structure of aggregates. For each vertex we can tell exactly
-    // which aggregate it belongs to.
-    // If we determine that the aggregate does not belong to us (i.e. the root vertex does not belong to this
-    // processor, or is outside and we lost "" arbitration), we record the global aggregate id in order to
-    // fetch the local info from the processor owning the aggregate. This is required for aggregates, as it
-    // uses the local aggregate ids of the owning processor.
-    std::set<GO> myAggGIDs, remoteAggGIDs;
-    for (LO LID = 0; LID < numPoints; LID++) {
-      GO aggGID = getAggGID(LID);
-      //      printf("[%d] (%d,%d,%d) => agg %d\n",LID,(int)(*xMap_)[x_[LID]],nDim_ > 1 ? (int)(*yMap_)[y_[LID]] : -1,nDim_ > 2 ? (int)(*zMap_)[z_[LID]] : -1,(int)aggGID);
-      if(aggGID == GO_INVALID) continue;
-      //      printf("[%d] getRoot = %d\n",(int)LID,(int)getRoot(LID));
-
-      if ((revMap_.find(getRoot(LID)) != revMap_.end()) && rowMap->isNodeGlobalElement(colMap->getGlobalElement(revMap_[getRoot(LID)]))) {
-        // Root of the brick aggregate containing GID (<- LID) belongs to us
-        vertex2AggId[LID] = aggGID;
-        myAggGIDs.insert(aggGID);
-
-        if (isRoot(LID))
-          aggregates->SetIsRoot(LID);
-	//	printf("[%d] initial vertex2AggId = %d\n",(int)LID,(int)vertex2AggId[LID]);
-      } else {
-        remoteAggGIDs.insert(aggGID);
-      }
-    }
-    size_t numAggregates = myAggGIDs    .size();
-    size_t numRemote     = remoteAggGIDs.size();
-    aggregates->SetNumAggregates(numAggregates);
-
-    std::map<GO,LO>  AggG2L;  // Map: Agg GID -> Agg LID (possibly on a different processor)
-    std::map<GO,int> AggG2R;  // Map: Agg GID -> processor rank owning aggregate
-
-    Array<GO> myAggGIDsArray(numAggregates), remoteAggGIDsArray(numRemote);
-
-    // Fill in the maps for aggregates that we own
-    size_t ind = 0;
-    for (typename std::set<GO>::const_iterator it = myAggGIDs.begin(); it != myAggGIDs.end(); it++) {
-      AggG2L[*it] = ind;
-      AggG2R[*it] = myRank;
-
-      myAggGIDsArray[ind++] = *it;
-    }
-
-    // The map is a convenient way to fetch remote local indices from global indices.
-    RCP<Map> aggMap = MapFactory::Build(rowMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                        myAggGIDsArray, 0, comm);
-
-    ind = 0;
-    for (typename std::set<GO>::const_iterator it = remoteAggGIDs.begin(); it != remoteAggGIDs.end(); it++)
-      remoteAggGIDsArray[ind++] = *it;
-
-    // Fetch the required aggregate local ids and ranks
-    Array<int> remoteProcIDs(numRemote);
-    Array<LO>  remoteLIDs   (numRemote);
-    aggMap->getRemoteIndexList(remoteAggGIDsArray, remoteProcIDs, remoteLIDs);
-
-    // Fill in the maps for aggregates that we don't own but which have some of our vertices
-    for (size_t i = 0; i < numRemote; i++) {
-      AggG2L[remoteAggGIDsArray[i]] = remoteLIDs   [i];
-      AggG2R[remoteAggGIDsArray[i]] = remoteProcIDs[i];
+  // Setup misc structures
+  // Logically, we construct enough data to query topological information of a rectangular grid
+  Setup(comm, overlappedCoords, colMap);
+
+  GetOStream(Runtime0) << "Using brick size: " << bx_
+                       << (nDim_ > 1 ? "x " + toString(by_) : "")
+                       << (nDim_ > 2 ? "x " + toString(bz_) : "") << std::endl;
+
+  // Build the graph
+  BuildGraph(currentLevel, A);
+
+  // Construct aggregates
+  RCP<Aggregates> aggregates = rcp(new Aggregates(colMap));
+  aggregates->setObjectLabel("Brick");
+
+  ArrayRCP<LO> vertex2AggId = aggregates->GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates->GetProcWinner()->getDataNonConst(0);
+
+  // In the first pass, we set a mapping from a vertex to aggregate global id. We deal with a structured
+  // rectangular mesh, therefore we know the structure of aggregates. For each vertex we can tell exactly
+  // which aggregate it belongs to.
+  // If we determine that the aggregate does not belong to us (i.e. the root vertex does not belong to this
+  // processor, or is outside and we lost "" arbitration), we record the global aggregate id in order to
+  // fetch the local info from the processor owning the aggregate. This is required for aggregates, as it
+  // uses the local aggregate ids of the owning processor.
+  std::set<GO> myAggGIDs, remoteAggGIDs;
+  for (LO LID = 0; LID < numPoints; LID++) {
+    GO aggGID = getAggGID(LID);
+    //      printf("[%d] (%d,%d,%d) => agg %d\n",LID,(int)(*xMap_)[x_[LID]],nDim_ > 1 ? (int)(*yMap_)[y_[LID]] : -1,nDim_ > 2 ? (int)(*zMap_)[z_[LID]] : -1,(int)aggGID);
+    if (aggGID == GO_INVALID) continue;
+    //      printf("[%d] getRoot = %d\n",(int)LID,(int)getRoot(LID));
+
+    if ((revMap_.find(getRoot(LID)) != revMap_.end()) && rowMap->isNodeGlobalElement(colMap->getGlobalElement(revMap_[getRoot(LID)]))) {
+      // Root of the brick aggregate containing GID (<- LID) belongs to us
+      vertex2AggId[LID] = aggGID;
+      myAggGIDs.insert(aggGID);
+
+      if (isRoot(LID))
+        aggregates->SetIsRoot(LID);
+      //	printf("[%d] initial vertex2AggId = %d\n",(int)LID,(int)vertex2AggId[LID]);
+    } else {
+      remoteAggGIDs.insert(aggGID);
     }
-
-    // Remap aggregate GIDs to LIDs and set up owning processors
-    for (LO LID = 0; LID < numPoints; LID++) {
-      if (revMap_.find(getRoot(LID)) != revMap_.end() && rowMap->isNodeGlobalElement(colMap->getGlobalElement(revMap_[getRoot(LID)]))) {
-        GO aggGID = vertex2AggId[LID];
-        if(aggGID != MUELU_UNAGGREGATED) {
-          vertex2AggId[LID] = AggG2L[aggGID];
-          procWinner  [LID] = AggG2R[aggGID];
-        }
-      }
-    }
-
-
-    GO numGlobalRemote;
-    MueLu_sumAll(comm, as<GO>(numRemote), numGlobalRemote);
-    aggregates->AggregatesCrossProcessors(numGlobalRemote);
-
-    Set(currentLevel, "Aggregates", aggregates);
-
-    GetOStream(Statistics1) << aggregates->description() << std::endl;
   }
+  size_t numAggregates = myAggGIDs.size();
+  size_t numRemote     = remoteAggGIDs.size();
+  aggregates->SetNumAggregates(numAggregates);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Setup(const RCP<const Teuchos::Comm<int> >& comm, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> >& coords, const RCP<const Map>& /* map */) const {
-    nDim_ = coords->getNumVectors();
-
-    x_    = coords->getData(0);
-    xMap_ = Construct1DMap(comm, x_);
-    nx_   = xMap_->size();
+  std::map<GO, LO> AggG2L;   // Map: Agg GID -> Agg LID (possibly on a different processor)
+  std::map<GO, int> AggG2R;  // Map: Agg GID -> processor rank owning aggregate
 
-    ny_   = 1;
-    if (nDim_ > 1) {
-      y_    = coords->getData(1);
-      yMap_ = Construct1DMap(comm, y_);
-      ny_   = yMap_->size();
-    }
+  Array<GO> myAggGIDsArray(numAggregates), remoteAggGIDsArray(numRemote);
 
-    nz_   = 1;
-    if (nDim_ > 2) {
-      z_    = coords->getData(2);
-      zMap_ = Construct1DMap(comm, z_);
-      nz_   = zMap_->size();
-    }
+  // Fill in the maps for aggregates that we own
+  size_t ind = 0;
+  for (typename std::set<GO>::const_iterator it = myAggGIDs.begin(); it != myAggGIDs.end(); it++) {
+    AggG2L[*it] = ind;
+    AggG2R[*it] = myRank;
 
-    for (size_t ind = 0; ind < coords->getLocalLength(); ind++) {
-      GO i = (*xMap_)[(coords->getData(0))[ind]], j = 0, k = 0;
-      if (nDim_ > 1)
-        j = (*yMap_)[(coords->getData(1))[ind]];
-      if (nDim_ > 2)
-        k = (*zMap_)[(coords->getData(2))[ind]];
-
-      revMap_[k*ny_*nx_ + j*nx_ + i] = ind;
-    }
+    myAggGIDsArray[ind++] = *it;
+  }
 
-   
-    // Get the number of aggregates in each direction, correcting for Dirichlet
-    int xboost = dirichletX_ ? 1 : 0;
-    int yboost = dirichletY_ ? 1 : 0;
-    int zboost = dirichletZ_ ? 1 : 0;
-    naggx_ = (nx_-2*xboost)/bx_ + ((nx_-2*xboost) % bx_ ? 1 : 0);
+  // The map is a convenient way to fetch remote local indices from global indices.
+  RCP<Map> aggMap = MapFactory::Build(rowMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                      myAggGIDsArray, 0, comm);
 
-    if(nDim_ > 1)
-      naggy_ = (ny_-2*yboost)/by_ + ( (ny_-2*yboost) % by_ ? 1 : 0);
-    else
-      naggy_ = 1;
+  ind = 0;
+  for (typename std::set<GO>::const_iterator it = remoteAggGIDs.begin(); it != remoteAggGIDs.end(); it++)
+    remoteAggGIDsArray[ind++] = *it;
 
-    if(nDim_ > 2)
-      naggz_ = (nz_-2*zboost)/bz_ + ( (nz_-2*zboost) % bz_ ? 1 : 0);
-    else
-      naggz_ = 1;
+  // Fetch the required aggregate local ids and ranks
+  Array<int> remoteProcIDs(numRemote);
+  Array<LO> remoteLIDs(numRemote);
+  aggMap->getRemoteIndexList(remoteAggGIDsArray, remoteProcIDs, remoteLIDs);
 
+  // Fill in the maps for aggregates that we don't own but which have some of our vertices
+  for (size_t i = 0; i < numRemote; i++) {
+    AggG2L[remoteAggGIDsArray[i]] = remoteLIDs[i];
+    AggG2R[remoteAggGIDsArray[i]] = remoteProcIDs[i];
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<typename BrickAggregationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::container>
-  BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Construct1DMap (const RCP<const Teuchos::Comm<int> >& comm,
-                  const ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& x) const
-  {
-    int n = x.size();
-
-    // Step 1: Create a local vector with unique coordinate points
-    RCP<container> gMap = rcp(new container);
-    for (int i = 0; i < n; i++)
-      (*gMap)[x[i]] = 0;
-
-#ifdef HAVE_MPI
-    // Step 2: exchange coordinates
-    // NOTE: we assume the coordinates are double, or double compatible
-    // That means that for complex case, we assume that all imaginary parts are zeros
-    int numProcs = comm->getSize();
-    if (numProcs > 1) {
-      RCP<const Teuchos::MpiComm<int> > dupMpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm->duplicate());
-
-      MPI_Comm rawComm = (*dupMpiComm->getRawMpiComm())();
-
-      int           sendCnt = gMap->size(), cnt = 0, recvSize;
-      Array<int>    recvCnt(numProcs), Displs(numProcs);
-      Array<double> sendBuf, recvBuf;
-
-      sendBuf.resize(sendCnt);
-      for (typename container::const_iterator cit = gMap->begin(); cit != gMap->end(); cit++)
-        sendBuf[cnt++] = Teuchos::as<double>(STS::real(cit->first));
-
-      MPI_Allgather(&sendCnt, 1, MPI_INT, recvCnt.getRawPtr(), 1, MPI_INT, rawComm);
-      Displs[0] = 0;
-      for (int i = 0; i < numProcs-1; i++)
-        Displs[i+1] = Displs[i] + recvCnt[i];
-      recvSize = Displs[numProcs-1] + recvCnt[numProcs-1];
-      recvBuf.resize(recvSize);
-      MPI_Allgatherv(sendBuf.getRawPtr(), sendCnt, MPI_DOUBLE, recvBuf.getRawPtr(), recvCnt.getRawPtr(), Displs.getRawPtr(), MPI_DOUBLE, rawComm);
-
-      for (int i = 0; i < recvSize; i++)
-        (*gMap)[as<SC>(recvBuf[i])] = 0;
+  // Remap aggregate GIDs to LIDs and set up owning processors
+  for (LO LID = 0; LID < numPoints; LID++) {
+    if (revMap_.find(getRoot(LID)) != revMap_.end() && rowMap->isNodeGlobalElement(colMap->getGlobalElement(revMap_[getRoot(LID)]))) {
+      GO aggGID = vertex2AggId[LID];
+      if (aggGID != MUELU_UNAGGREGATED) {
+        vertex2AggId[LID] = AggG2L[aggGID];
+        procWinner[LID]   = AggG2R[aggGID];
+      }
     }
-#endif
-
-    GO cnt = 0;
-    for (typename container::iterator it = gMap->begin(); it != gMap->end(); it++)
-      it->second = cnt++;
-
-    return gMap;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isRoot(LocalOrdinal LID) const {
-    int i,j,k;
-    getIJK(LID,i,j,k);
+  GO numGlobalRemote;
+  MueLu_sumAll(comm, as<GO>(numRemote), numGlobalRemote);
+  aggregates->AggregatesCrossProcessors(numGlobalRemote);
 
-    return (k*ny_*nx_ + j*nx_ + i) == getRoot(LID);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isDirichlet(LocalOrdinal LID) const {
-    bool boundary = false;
-    int i,j,k;
-    getIJK(LID,i,j,k);
-    if(             dirichletX_  &&  (i == 0 || i == nx_-1) )
-      boundary = true;
-    if(nDim_ > 1 && dirichletY_  &&  (j == 0 || j == ny_-1) )
-      boundary = true;
-    if(nDim_ > 2 && dirichletZ_  &&  (k == 0 || k == nz_-1) )
-      boundary = true;
-    
-    return boundary;
-  }
+  Set(currentLevel, "Aggregates", aggregates);
 
+  GetOStream(Statistics1) << aggregates->description() << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRoot(LocalOrdinal LID) const {
-    if(isDirichlet(LID))
-      return Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Setup(const RCP<const Teuchos::Comm<int> >& comm, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coords, const RCP<const Map>& /* map */) const {
+  nDim_ = coords->getNumVectors();
 
-    int aggI,aggJ,aggK;
-    getAggIJK(LID,aggI,aggJ,aggK);
-    int xboost = dirichletX_ ? 1 : 0;
-    int yboost = dirichletY_ ? 1 : 0;
-    int zboost = dirichletZ_ ? 1 : 0;
-   
-    int i =             xboost + aggI*bx_ + (bx_-1)/2;
-    int j = (nDim_>1) ? yboost + aggJ*by_ + (by_-1)/2 : 0;
-    int k = (nDim_>2) ? zboost + aggK*bz_ + (bz_-1)/2 : 0;
+  x_    = coords->getData(0);
+  xMap_ = Construct1DMap(comm, x_);
+  nx_   = xMap_->size();
 
-    return k*ny_*nx_ + j*nx_ + i;
+  ny_ = 1;
+  if (nDim_ > 1) {
+    y_    = coords->getData(1);
+    yMap_ = Construct1DMap(comm, y_);
+    ny_   = yMap_->size();
   }
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void  BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getIJK(LocalOrdinal LID, int &i, int &j, int &k) const {
-   i = (*xMap_)[x_[LID]];
-   j = (nDim_>1) ? (*yMap_)[y_[LID]] : 0;
-   k =  (nDim_>2) ? (*zMap_)[z_[LID]] : 0;
+  nz_ = 1;
+  if (nDim_ > 2) {
+    z_    = coords->getData(2);
+    zMap_ = Construct1DMap(comm, z_);
+    nz_   = zMap_->size();
   }
 
+  for (size_t ind = 0; ind < coords->getLocalLength(); ind++) {
+    GO i = (*xMap_)[(coords->getData(0))[ind]], j = 0, k = 0;
+    if (nDim_ > 1)
+      j = (*yMap_)[(coords->getData(1))[ind]];
+    if (nDim_ > 2)
+      k = (*zMap_)[(coords->getData(2))[ind]];
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void  BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getAggIJK(LocalOrdinal LID, int &i, int &j, int &k) const {
-    int xboost = dirichletX_ ? 1 : 0;
-    int yboost = dirichletY_ ? 1 : 0;
-    int zboost = dirichletZ_ ? 1 : 0;
-    int pointI, pointJ, pointK;
-    getIJK(LID,pointI,pointJ,pointK);
-    i = (pointI-xboost)/bx_;
-
-    if (nDim_ > 1) j = (pointJ-yboost)/by_;
-    else j = 0;
-
-    if (nDim_ > 2) k = (pointK-zboost)/bz_;
-    else k = 0;
+    revMap_[k * ny_ * nx_ + j * nx_ + i] = ind;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getAggGID(LocalOrdinal LID) const {
-    bool boundary = false;
-
-    int i, j, k;
-    getIJK(LID,i,j,k);
-    int ii , jj, kk;
-    getAggIJK(LID,ii,jj,kk);
-
-    if(              dirichletX_ && (i == 0 || i == nx_ - 1)) boundary = true;
-    if (nDim_ > 1 && dirichletY_ && (j == 0 || j == ny_ - 1)) boundary = true;
-    if (nDim_ > 2 && dirichletZ_ && (k == 0 || k == nz_ - 1)) boundary = true;
-
-    /*
-    if(boundary)
-      printf("[%d] coord = (%d,%d,%d) {%d,%d,%d} agg = (%d,%d,%d) {%d,%d,%d} => agg %s\n",LID,i,j,k,nx_,ny_,nz_,ii,jj,kk,naggx_,naggy_,naggz_,"BOUNDARY");
-    else
-      printf("[%d] coord = (%d,%d,%d) {%d,%d,%d} agg = (%d,%d,%d) {%d,%d,%d} => agg %d\n",LID,i,j,k,nx_,ny_,nz_,ii,jj,kk,naggx_,naggy_,naggz_,kk*naggy_*naggx_ + jj*naggx_ + ii);
-    */
-
-    if (boundary)
-      return Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
-    else
-      return Teuchos::as<GlobalOrdinal>(kk*naggy_*naggx_) + Teuchos::as<GlobalOrdinal>(jj*naggx_) + ii;
-    
-  }
+  // Get the number of aggregates in each direction, correcting for Dirichlet
+  int xboost = dirichletX_ ? 1 : 0;
+  int yboost = dirichletY_ ? 1 : 0;
+  int zboost = dirichletZ_ ? 1 : 0;
+  naggx_     = (nx_ - 2 * xboost) / bx_ + ((nx_ - 2 * xboost) % bx_ ? 1 : 0);
+
+  if (nDim_ > 1)
+    naggy_ = (ny_ - 2 * yboost) / by_ + ((ny_ - 2 * yboost) % by_ ? 1 : 0);
+  else
+    naggy_ = 1;
+
+  if (nDim_ > 2)
+    naggz_ = (nz_ - 2 * zboost) / bz_ + ((nz_ - 2 * zboost) % bz_ ? 1 : 0);
+  else
+    naggz_ = 1;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<typename BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::container>
+BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Construct1DMap(const RCP<const Teuchos::Comm<int> >& comm,
+                   const ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType>& x) const {
+  int n = x.size();
+
+  // Step 1: Create a local vector with unique coordinate points
+  RCP<container> gMap = rcp(new container);
+  for (int i = 0; i < n; i++)
+    (*gMap)[x[i]] = 0;
 
+#ifdef HAVE_MPI
+  // Step 2: exchange coordinates
+  // NOTE: we assume the coordinates are double, or double compatible
+  // That means that for complex case, we assume that all imaginary parts are zeros
+  int numProcs = comm->getSize();
+  if (numProcs > 1) {
+    RCP<const Teuchos::MpiComm<int> > dupMpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm->duplicate());
+
+    MPI_Comm rawComm = (*dupMpiComm->getRawMpiComm())();
+
+    int sendCnt = gMap->size(), cnt = 0, recvSize;
+    Array<int> recvCnt(numProcs), Displs(numProcs);
+    Array<double> sendBuf, recvBuf;
+
+    sendBuf.resize(sendCnt);
+    for (typename container::const_iterator cit = gMap->begin(); cit != gMap->end(); cit++)
+      sendBuf[cnt++] = Teuchos::as<double>(STS::real(cit->first));
+
+    MPI_Allgather(&sendCnt, 1, MPI_INT, recvCnt.getRawPtr(), 1, MPI_INT, rawComm);
+    Displs[0] = 0;
+    for (int i = 0; i < numProcs - 1; i++)
+      Displs[i + 1] = Displs[i] + recvCnt[i];
+    recvSize = Displs[numProcs - 1] + recvCnt[numProcs - 1];
+    recvBuf.resize(recvSize);
+    MPI_Allgatherv(sendBuf.getRawPtr(), sendCnt, MPI_DOUBLE, recvBuf.getRawPtr(), recvCnt.getRawPtr(), Displs.getRawPtr(), MPI_DOUBLE, rawComm);
+
+    for (int i = 0; i < recvSize; i++)
+      (*gMap)[as<SC>(recvBuf[i])] = 0;
+  }
+#endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildGraph(Level& currentLevel, const RCP<Matrix>& A) const {
-    // TODO: Currently only works w/ 1 DOF per node
-    double dirichletThreshold = 0.0;
-
-    if(bx_ > 1 && (nDim_ <= 1 || by_ > 1) && (nDim_ <=2 || bz_>1) ) {
-      FactoryMonitor m(*this, "Generating Graph (trivial)", currentLevel);
-      /*** Case 1: Use the matrix is the graph ***/
-      // Bricks are of non-trivial size in all active dimensions
-      RCP<GraphBase> graph = rcp(new Graph(A->getCrsGraph(), "graph of A"));
-      ArrayRCP<bool > boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-      graph->SetBoundaryNodeMap(boundaryNodes);
-      
-      if (GetVerbLevel() & Statistics1) {
-        GO numLocalBoundaryNodes  = 0;
-        GO numGlobalBoundaryNodes = 0;
-        for (LO i = 0; i < boundaryNodes.size(); ++i)
-          if (boundaryNodes[i])
-            numLocalBoundaryNodes++;
-        RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-        MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-        GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-      }
-      Set(currentLevel, "DofsPerNode", 1);
-      Set(currentLevel, "Graph", graph);
-      Set(currentLevel, "Filtering",false);
+  GO cnt = 0;
+  for (typename container::iterator it = gMap->begin(); it != gMap->end(); it++)
+    it->second = cnt++;
+
+  return gMap;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isRoot(LocalOrdinal LID) const {
+  int i, j, k;
+  getIJK(LID, i, j, k);
+
+  return (k * ny_ * nx_ + j * nx_ + i) == getRoot(LID);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isDirichlet(LocalOrdinal LID) const {
+  bool boundary = false;
+  int i, j, k;
+  getIJK(LID, i, j, k);
+  if (dirichletX_ && (i == 0 || i == nx_ - 1))
+    boundary = true;
+  if (nDim_ > 1 && dirichletY_ && (j == 0 || j == ny_ - 1))
+    boundary = true;
+  if (nDim_ > 2 && dirichletZ_ && (k == 0 || k == nz_ - 1))
+    boundary = true;
+
+  return boundary;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRoot(LocalOrdinal LID) const {
+  if (isDirichlet(LID))
+    return Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
+
+  int aggI, aggJ, aggK;
+  getAggIJK(LID, aggI, aggJ, aggK);
+  int xboost = dirichletX_ ? 1 : 0;
+  int yboost = dirichletY_ ? 1 : 0;
+  int zboost = dirichletZ_ ? 1 : 0;
+
+  int i = xboost + aggI * bx_ + (bx_ - 1) / 2;
+  int j = (nDim_ > 1) ? yboost + aggJ * by_ + (by_ - 1) / 2 : 0;
+  int k = (nDim_ > 2) ? zboost + aggK * bz_ + (bz_ - 1) / 2 : 0;
+
+  return k * ny_ * nx_ + j * nx_ + i;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getIJK(LocalOrdinal LID, int& i, int& j, int& k) const {
+  i = (*xMap_)[x_[LID]];
+  j = (nDim_ > 1) ? (*yMap_)[y_[LID]] : 0;
+  k = (nDim_ > 2) ? (*zMap_)[z_[LID]] : 0;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getAggIJK(LocalOrdinal LID, int& i, int& j, int& k) const {
+  int xboost = dirichletX_ ? 1 : 0;
+  int yboost = dirichletY_ ? 1 : 0;
+  int zboost = dirichletZ_ ? 1 : 0;
+  int pointI, pointJ, pointK;
+  getIJK(LID, pointI, pointJ, pointK);
+  i = (pointI - xboost) / bx_;
+
+  if (nDim_ > 1)
+    j = (pointJ - yboost) / by_;
+  else
+    j = 0;
+
+  if (nDim_ > 2)
+    k = (pointK - zboost) / bz_;
+  else
+    k = 0;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getAggGID(LocalOrdinal LID) const {
+  bool boundary = false;
+
+  int i, j, k;
+  getIJK(LID, i, j, k);
+  int ii, jj, kk;
+  getAggIJK(LID, ii, jj, kk);
+
+  if (dirichletX_ && (i == 0 || i == nx_ - 1)) boundary = true;
+  if (nDim_ > 1 && dirichletY_ && (j == 0 || j == ny_ - 1)) boundary = true;
+  if (nDim_ > 2 && dirichletZ_ && (k == 0 || k == nz_ - 1)) boundary = true;
+
+  /*
+  if(boundary)
+    printf("[%d] coord = (%d,%d,%d) {%d,%d,%d} agg = (%d,%d,%d) {%d,%d,%d} => agg %s\n",LID,i,j,k,nx_,ny_,nz_,ii,jj,kk,naggx_,naggy_,naggz_,"BOUNDARY");
+  else
+    printf("[%d] coord = (%d,%d,%d) {%d,%d,%d} agg = (%d,%d,%d) {%d,%d,%d} => agg %d\n",LID,i,j,k,nx_,ny_,nz_,ii,jj,kk,naggx_,naggy_,naggz_,kk*naggy_*naggx_ + jj*naggx_ + ii);
+  */
+
+  if (boundary)
+    return Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
+  else
+    return Teuchos::as<GlobalOrdinal>(kk * naggy_ * naggx_) + Teuchos::as<GlobalOrdinal>(jj * naggx_) + ii;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildGraph(Level& currentLevel, const RCP<Matrix>& A) const {
+  // TODO: Currently only works w/ 1 DOF per node
+  double dirichletThreshold = 0.0;
+
+  if (bx_ > 1 && (nDim_ <= 1 || by_ > 1) && (nDim_ <= 2 || bz_ > 1)) {
+    FactoryMonitor m(*this, "Generating Graph (trivial)", currentLevel);
+    /*** Case 1: Use the matrix is the graph ***/
+    // Bricks are of non-trivial size in all active dimensions
+    RCP<GraphBase> graph         = rcp(new Graph(A->getCrsGraph(), "graph of A"));
+    ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+    graph->SetBoundaryNodeMap(boundaryNodes);
+
+    if (GetVerbLevel() & Statistics1) {
+      GO numLocalBoundaryNodes  = 0;
+      GO numGlobalBoundaryNodes = 0;
+      for (LO i = 0; i < boundaryNodes.size(); ++i)
+        if (boundaryNodes[i])
+          numLocalBoundaryNodes++;
+      RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+      MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+      GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
     }
-    else {
-      FactoryMonitor m(*this, "Generating Graph", currentLevel);
-      /*** Case 2: Dropping required ***/
-      // There is at least one active dimension in which we are not coarsening.
-      // Those connections need to be dropped
-      bool drop_x = (bx_ == 1);
-      bool drop_y = (nDim_> 1 && by_ == 1);
-      bool drop_z = (nDim_> 2 && bz_ == 1);
-      
-      ArrayRCP<LO> rows   (A->getLocalNumRows()+1);
-      ArrayRCP<LO> columns(A->getLocalNumEntries());
-      
-      size_t N = A->getRowMap()->getLocalNumElements();
-
-      // FIXME: Do this on the host because indexing functions are host functions
-      auto G = A->getLocalMatrixHost().graph;
-      auto rowptr = G.row_map;
-      auto colind = G.entries;
-
-      int ct=0;
-      rows[0] = 0;
-      for(size_t row=0; row<N; row++) {
-        // NOTE: Assumes that the first part of the colmap is the rowmap
-        int ir,jr,kr;
-        LO row2 = A->getColMap()->getLocalElement(A->getRowMap()->getGlobalElement(row));
-        getIJK(row2,ir,jr,kr);
-
-        for(size_t cidx=rowptr[row]; cidx<rowptr[row+1]; cidx++) {
-          int ic,jc,kc;
-          LO col = colind[cidx];
-          getIJK(col,ic,jc,kc);
-
-          if( (row2 !=col) && ((drop_x && ir != ic) || (drop_y && jr != jc) || (drop_z && kr != kc) )) {
-            // Drop it
-            //            printf("[%4d] DROP row = (%d,%d,%d) col = (%d,%d,%d)\n",(int)row,ir,jr,kr,ic,jc,kc);
-          }
-          else {
-            // Keep it
-            //            printf("[%4d] KEEP row = (%d,%d,%d) col = (%d,%d,%d)\n",(int)row,ir,jr,kr,ic,jc,kc);
-            columns[ct] = col;
-            ct++;
-          }
+    Set(currentLevel, "DofsPerNode", 1);
+    Set(currentLevel, "Graph", graph);
+    Set(currentLevel, "Filtering", false);
+  } else {
+    FactoryMonitor m(*this, "Generating Graph", currentLevel);
+    /*** Case 2: Dropping required ***/
+    // There is at least one active dimension in which we are not coarsening.
+    // Those connections need to be dropped
+    bool drop_x = (bx_ == 1);
+    bool drop_y = (nDim_ > 1 && by_ == 1);
+    bool drop_z = (nDim_ > 2 && bz_ == 1);
+
+    ArrayRCP<LO> rows(A->getLocalNumRows() + 1);
+    ArrayRCP<LO> columns(A->getLocalNumEntries());
+
+    size_t N = A->getRowMap()->getLocalNumElements();
+
+    // FIXME: Do this on the host because indexing functions are host functions
+    auto G      = A->getLocalMatrixHost().graph;
+    auto rowptr = G.row_map;
+    auto colind = G.entries;
+
+    int ct  = 0;
+    rows[0] = 0;
+    for (size_t row = 0; row < N; row++) {
+      // NOTE: Assumes that the first part of the colmap is the rowmap
+      int ir, jr, kr;
+      LO row2 = A->getColMap()->getLocalElement(A->getRowMap()->getGlobalElement(row));
+      getIJK(row2, ir, jr, kr);
+
+      for (size_t cidx = rowptr[row]; cidx < rowptr[row + 1]; cidx++) {
+        int ic, jc, kc;
+        LO col = colind[cidx];
+        getIJK(col, ic, jc, kc);
+
+        if ((row2 != col) && ((drop_x && ir != ic) || (drop_y && jr != jc) || (drop_z && kr != kc))) {
+          // Drop it
+          //            printf("[%4d] DROP row = (%d,%d,%d) col = (%d,%d,%d)\n",(int)row,ir,jr,kr,ic,jc,kc);
+        } else {
+          // Keep it
+          //            printf("[%4d] KEEP row = (%d,%d,%d) col = (%d,%d,%d)\n",(int)row,ir,jr,kr,ic,jc,kc);
+          columns[ct] = col;
+          ct++;
         }
-        rows[row+1] = ct;
-      }//end for
-
-      RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, A->getRowMap(), A->getColMap(), "thresholded graph of A"));
-
-
-      ArrayRCP<bool > boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-      graph->SetBoundaryNodeMap(boundaryNodes);
-      
-      if (GetVerbLevel() & Statistics1) {
-        GO numLocalBoundaryNodes  = 0;
-        GO numGlobalBoundaryNodes = 0;
-        for (LO i = 0; i < boundaryNodes.size(); ++i)
-          if (boundaryNodes[i])
-            numLocalBoundaryNodes++;
-        RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-        MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-        GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
       }
-      Set(currentLevel, "DofsPerNode", 1);
-      Set(currentLevel, "Graph", graph);
-      Set(currentLevel, "Filtering",true);
-    }//end else
-
-
-  }//end BuildGraph
-
-  
+      rows[row + 1] = ct;
+    }  // end for
+
+    RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, A->getRowMap(), A->getColMap(), "thresholded graph of A"));
+
+    ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+    graph->SetBoundaryNodeMap(boundaryNodes);
+
+    if (GetVerbLevel() & Statistics1) {
+      GO numLocalBoundaryNodes  = 0;
+      GO numGlobalBoundaryNodes = 0;
+      for (LO i = 0; i < boundaryNodes.size(); ++i)
+        if (boundaryNodes[i])
+          numLocalBoundaryNodes++;
+      RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+      MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+      GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+    }
+    Set(currentLevel, "DofsPerNode", 1);
+    Set(currentLevel, "Graph", graph);
+    Set(currentLevel, "Filtering", true);
+  }  // end else
 
+}  // end BuildGraph
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_BRICKAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/Containers/MueLu_Aggregates_decl.hpp b/packages/muelu/src/Graph/Containers/MueLu_Aggregates_decl.hpp
index 056673b7b9db..b11e8584442e 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_Aggregates_decl.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_Aggregates_decl.hpp
@@ -66,20 +66,20 @@
 #include "MueLu_IndexManager.hpp"
 #include "MueLu_IndexManager_kokkos.hpp"
 
-#define MUELU_UNAGGREGATED  -1   /* indicates that a node is unassigned to  */
-                                 /* any aggregate.                          */
-
-#define MUELU_UNASSIGNED    -1   /* indicates a vertex is not yet claimed   */
-                                 /* by a processor during aggregation.      */
-                                 /* Note, it is possible at                 */
-                                 /* this stage that some processors may have*/
-                                 /* claimed their copy of a vertex for one  */
-                                 /* of their aggregates.  However, some     */
-                                 /* arbitration still needs to occur.       */
-                                 /* The corresponding procWinner[]'s remain */
-                                 /* as MUELU_UNASSIGNED until               */
-                                 /* ArbitrateAndCommunicate() is            */
-                                 /* invoked to arbitrate.                   */
+#define MUELU_UNAGGREGATED -1 /* indicates that a node is unassigned to  */
+                              /* any aggregate.                          */
+
+#define MUELU_UNASSIGNED -1 /* indicates a vertex is not yet claimed   */
+                            /* by a processor during aggregation.      */
+                            /* Note, it is possible at                 */
+                            /* this stage that some processors may have*/
+                            /* claimed their copy of a vertex for one  */
+                            /* of their aggregates.  However, some     */
+                            /* arbitration still needs to occur.       */
+                            /* The corresponding procWinner[]'s remain */
+                            /* as MUELU_UNASSIGNED until               */
+                            /* ArbitrateAndCommunicate() is            */
+                            /* invoked to arbitrate.                   */
 
 /*****************************************************************************
 
@@ -102,275 +102,272 @@ namespace MueLu {
     where rows (or vertices) correspond to aggregates and colunmns (or edges)
     correspond to nodes. While not strictly necessary, it might be convenient.
 */
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Aggregates;
-
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public BaseClass {
-  public:
-    using local_ordinal_type  = LocalOrdinal;
-    using global_ordinal_type = GlobalOrdinal;
-    using execution_space     = typename DeviceType::execution_space;
-    using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
-    using device_type         = DeviceType;
-    using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
-    using LO_view             = Kokkos::View<local_ordinal_type*, device_type>;
-
-    using aggregates_sizes_type = Kokkos::View<LocalOrdinal*, device_type>;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class Aggregates;
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public BaseClass {
+ public:
+  using local_ordinal_type  = LocalOrdinal;
+  using global_ordinal_type = GlobalOrdinal;
+  using execution_space     = typename DeviceType::execution_space;
+  using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
+  using device_type         = DeviceType;
+  using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
+  using LO_view             = Kokkos::View<local_ordinal_type*, device_type>;
+
+  using aggregates_sizes_type = Kokkos::View<LocalOrdinal*, device_type>;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_AGGREGATES_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-
-    // Defining types that require the short names included above
-    using local_graph_type = typename LWGraph_kokkos::local_graph_type;
-    using colors_view_type = Kokkos::View<typename local_graph_type::entries_type::data_type,
-                                          typename local_graph_type::device_type::memory_space>;
-
-    /*! @brief Standard constructor for Aggregates structure
-     *
-     * Standard constructor of aggregates takes a Graph object as parameter.
-     * Uses the graph.GetImportMap() to initialize the internal vector for mapping nodes to (local) aggregate ids as well as
-     * the mapping of node to the owning processor id.
-     *
-     */
-    Aggregates(const GraphBase & graph);
-
-    /*! @brief Standard constructor for Aggregates structure
-     *
-     * Standard constructor of aggregates takes a LWGraph object as parameter.
-     * Uses the graph.GetImportMap() to initialize the internal vector for mapping nodes to (local) aggregate ids as well as
-     * the mapping of node to the owning processor id.
-     *
-     */
-    Aggregates(LWGraph_kokkos graph);
-
-    /*! @brief Constructor for Aggregates structure
-     *
-     * This constructor takes a RCP pointer to a map which is used for the internal mappings of nodes to the (local) aggregate ids and the owning processor.
-     *
-     */
-    Aggregates(const RCP<const Map>& map);
-
-    /*! @brief Destructor
-     *
-     */
-    virtual ~Aggregates() { }
-
-    //! @name Set/Get Methods for specific aggregation data
-    //@{
-
-    /*! @brief Get the index manager used by structured aggregation algorithms.
-        This has to be done by the aggregation factory.
-    */
-    RCP<IndexManager_kokkos>& GetIndexManagerKokkos() { return geoDataKokkos_; }
-
-    /*! @brief Set the index manager used by structured aggregation algorithms.
-        This has to be done by the aggregation factory.
-    */
-    void SetIndexManagerKokkos(RCP<IndexManager_kokkos> & geoDataKokkos) { geoDataKokkos_ = geoDataKokkos; }
-
-    /*! @brief Get the index manager used by various aggregation algorithms.
-        This has to be done by the aggregation factory.
-    */
-    RCP<IndexManager>& GetIndexManager() { return geoData_; }
-
-    /*! @brief Set the index manager used by various aggregation algorithms.
-        This has to be done by the aggregation factory.
-    */
-    void SetIndexManager(RCP<IndexManager> & geoData) { geoData_ = geoData; }
-
-    /*! @brief Get a distance 2 coloring of the underlying graph.
-        The coloring is computed and set during Phase1 of aggregation.
-    */
-    colors_view_type& GetGraphColors() { return graphColors_; }
-
-    /*! @brief Set a distance 2 coloring of the underlying graph.
-        The coloring is computed and set during Phase1 of aggregation.
-    */
-    void SetGraphColors(colors_view_type graphColors) { graphColors_ = graphColors; }
+ public:
+  // Defining types that require the short names included above
+  using local_graph_type = typename LWGraph_kokkos::local_graph_type;
+  using colors_view_type = Kokkos::View<typename local_graph_type::entries_type::data_type,
+                                        typename local_graph_type::device_type::memory_space>;
+
+  /*! @brief Standard constructor for Aggregates structure
+   *
+   * Standard constructor of aggregates takes a Graph object as parameter.
+   * Uses the graph.GetImportMap() to initialize the internal vector for mapping nodes to (local) aggregate ids as well as
+   * the mapping of node to the owning processor id.
+   *
+   */
+  Aggregates(const GraphBase& graph);
+
+  /*! @brief Standard constructor for Aggregates structure
+   *
+   * Standard constructor of aggregates takes a LWGraph object as parameter.
+   * Uses the graph.GetImportMap() to initialize the internal vector for mapping nodes to (local) aggregate ids as well as
+   * the mapping of node to the owning processor id.
+   *
+   */
+  Aggregates(LWGraph_kokkos graph);
+
+  /*! @brief Constructor for Aggregates structure
+   *
+   * This constructor takes a RCP pointer to a map which is used for the internal mappings of nodes to the (local) aggregate ids and the owning processor.
+   *
+   */
+  Aggregates(const RCP<const Map>& map);
+
+  /*! @brief Destructor
+   *
+   */
+  virtual ~Aggregates() {}
+
+  //! @name Set/Get Methods for specific aggregation data
+  //@{
+
+  /*! @brief Get the index manager used by structured aggregation algorithms.
+      This has to be done by the aggregation factory.
+  */
+  RCP<IndexManager_kokkos>& GetIndexManagerKokkos() { return geoDataKokkos_; }
+
+  /*! @brief Set the index manager used by structured aggregation algorithms.
+      This has to be done by the aggregation factory.
+  */
+  void SetIndexManagerKokkos(RCP<IndexManager_kokkos>& geoDataKokkos) { geoDataKokkos_ = geoDataKokkos; }
+
+  /*! @brief Get the index manager used by various aggregation algorithms.
+      This has to be done by the aggregation factory.
+  */
+  RCP<IndexManager>& GetIndexManager() { return geoData_; }
 
-    /*! @brief Get the number of colors needed by the distance 2 coloring.
-    */
-    LO GetGraphNumColors() { return graphNumColors_; }
+  /*! @brief Set the index manager used by various aggregation algorithms.
+      This has to be done by the aggregation factory.
+  */
+  void SetIndexManager(RCP<IndexManager>& geoData) { geoData_ = geoData; }
+
+  /*! @brief Get a distance 2 coloring of the underlying graph.
+      The coloring is computed and set during Phase1 of aggregation.
+  */
+  colors_view_type& GetGraphColors() { return graphColors_; }
+
+  /*! @brief Set a distance 2 coloring of the underlying graph.
+      The coloring is computed and set during Phase1 of aggregation.
+  */
+  void SetGraphColors(colors_view_type graphColors) { graphColors_ = graphColors; }
 
-    /*! @brief Set the number of colors needed by the distance 2 coloring.
-    */
-    void SetGraphNumColors(const LO graphNumColors) { graphNumColors_ = graphNumColors; }
+  /*! @brief Get the number of colors needed by the distance 2 coloring.
+   */
+  LO GetGraphNumColors() { return graphNumColors_; }
 
-    //@}
-
-    /*! @brief Set number of local aggregates on current processor.
+  /*! @brief Set the number of colors needed by the distance 2 coloring.
+   */
+  void SetGraphNumColors(const LO graphNumColors) { graphNumColors_ = graphNumColors; }
 
-        This has to be done by the aggregation routines.
-    */
-    void SetNumAggregates(LO nAggregates) { numAggregates_ = nAggregates; }
+  //@}
+
+  /*! @brief Set number of local aggregates on current processor.
 
-    /*! @brief Set number of global aggregates on current processor.
-  
-        This has to be done by the aggregation routines.
-    */
-    void SetNumGlobalAggregates(GO nGlobalAggregates) { numGlobalAggregates_ = nGlobalAggregates; }
+      This has to be done by the aggregation routines.
+  */
+  void SetNumAggregates(LO nAggregates) { numAggregates_ = nAggregates; }
 
-    ///< returns the number of aggregates of the current processor. Note: could/should be renamed to GetNumLocalAggregates?
-    KOKKOS_INLINE_FUNCTION LO GetNumAggregates() const {
-      return numAggregates_;
-    }
-
-    //! @brief Record whether aggregates include DOFs from other processes.
-    KOKKOS_INLINE_FUNCTION void AggregatesCrossProcessors(const bool& flag) {
-      aggregatesIncludeGhosts_ = flag;
-    }
-
-    /*! @brief Return false if and only if no aggregates include DOFs from other processes.
-
-        Used in construction of tentative prolongator to skip a communication phase.
-    */
-    KOKKOS_INLINE_FUNCTION bool AggregatesCrossProcessors() const {
-      return aggregatesIncludeGhosts_;
-    }
-
-    /*! @brief Returns a nonconstant vector that maps local node IDs to local aggregates IDs.
-
-        For local node ID i, the corresponding vector entry v[i] is the local aggregate id to which i belongs on the current processor.
-    */
-    RCP<LOMultiVector>& GetVertex2AggIdNonConst()     { return vertex2AggId_;       }
-
-    /*! @brief Returns nonconstant vector that maps local node IDs to owning processor IDs.
-
-        For local node ID i, the corresponding vector entry v[i] is the owning processor ID.
-    */
-    RCP<LOVector>& GetProcWinnerNonConst()       { return procWinner_;         }
-    /*! @brief Returns constant vector that maps local node IDs to local aggregates IDs.
-
-        For local node ID i, the corresponding vector entry v[i] is the local aggregate id to which i belongs on the current processor.
-    */
-    const RCP<LOMultiVector>& GetVertex2AggId() const { return vertex2AggId_;       }
+  /*! @brief Set number of global aggregates on current processor.
 
-    /*! @brief Returns constant vector that maps local node IDs to owning processor IDs.
+      This has to be done by the aggregation routines.
+  */
+  void SetNumGlobalAggregates(GO nGlobalAggregates) { numGlobalAggregates_ = nGlobalAggregates; }
 
-        For local node ID i, the corresponding vector entry v[i] is the owning processor ID.
-    */
-    const RCP<LOVector>& GetProcWinner() const   { return procWinner_;         }
+  ///< returns the number of aggregates of the current processor. Note: could/should be renamed to GetNumLocalAggregates?
+  KOKKOS_INLINE_FUNCTION LO GetNumAggregates() const {
+    return numAggregates_;
+  }
+
+  //! @brief Record whether aggregates include DOFs from other processes.
+  KOKKOS_INLINE_FUNCTION void AggregatesCrossProcessors(const bool& flag) {
+    aggregatesIncludeGhosts_ = flag;
+  }
+
+  /*! @brief Return false if and only if no aggregates include DOFs from other processes.
+
+      Used in construction of tentative prolongator to skip a communication phase.
+  */
+  KOKKOS_INLINE_FUNCTION bool AggregatesCrossProcessors() const {
+    return aggregatesIncludeGhosts_;
+  }
+
+  /*! @brief Returns a nonconstant vector that maps local node IDs to local aggregates IDs.
+
+      For local node ID i, the corresponding vector entry v[i] is the local aggregate id to which i belongs on the current processor.
+  */
+  RCP<LOMultiVector>& GetVertex2AggIdNonConst() { return vertex2AggId_; }
 
-    //! Returns true if node with given local node id is marked to be a root node
-    inline bool IsRoot(LO i) const                      { return isRoot_[i];          }
+  /*! @brief Returns nonconstant vector that maps local node IDs to owning processor IDs.
+
+      For local node ID i, the corresponding vector entry v[i] is the owning processor ID.
+  */
+  RCP<LOVector>& GetProcWinnerNonConst() { return procWinner_; }
+  /*! @brief Returns constant vector that maps local node IDs to local aggregates IDs.
 
-    /*! @brief Set root node information.
+      For local node ID i, the corresponding vector entry v[i] is the local aggregate id to which i belongs on the current processor.
+  */
+  const RCP<LOMultiVector>& GetVertex2AggId() const { return vertex2AggId_; }
 
-    Used by aggregation methods only.
-    */
-    inline void SetIsRoot(LO i, bool value = true)      { isRoot_[i] = value;         }
+  /*! @brief Returns constant vector that maps local node IDs to owning processor IDs.
 
-    const RCP<const Map> GetMap() const; ///< returns (overlapping) map of aggregate/node distribution
+      For local node ID i, the corresponding vector entry v[i] is the owning processor ID.
+  */
+  const RCP<LOVector>& GetProcWinner() const { return procWinner_; }
 
-    /*! @brief Compute sizes of aggregates
+  //! Returns true if node with given local node id is marked to be a root node
+  inline bool IsRoot(LO i) const { return isRoot_[i]; }
 
-      Returns the number of nodes in each aggregate in an array.
-      If the aggregate sizes are not stored internally (which is the default), they are computed and returned.
-      If the aggregate sizes have been stored internally, then they are *not* recomputed, but instead the
-      stored sizes are returned.
+  /*! @brief Set root node information.
 
-      @param[in] forceRecompute if true, force recomputation of the aggregate sizes.
-     */
-    typename aggregates_sizes_type::const_type ComputeAggregateSizes(bool forceRecompute = false) const;
+  Used by aggregation methods only.
+  */
+  inline void SetIsRoot(LO i, bool value = true) { isRoot_[i] = value; }
 
-    /*! @brief Compute sizes of aggregates
+  const RCP<const Map> GetMap() const;  ///< returns (overlapping) map of aggregate/node distribution
 
-      Returns the number of nodes in each aggregate in an array.
-      If the aggregate sizes are not stored internally (which is the default), they are computed and returned.
-      If the aggregate sizes have been stored internally, then they are *not* recomputed, but instead the
-      stored sizes are returned.
+  /*! @brief Compute sizes of aggregates
 
-      @param[in] forceRecompute if true, force recomputation of the aggregate sizes.
-     */
-    Teuchos::ArrayRCP<LocalOrdinal> ComputeAggregateSizesArrayRCP(bool forceRecompute = false) const;
+    Returns the number of nodes in each aggregate in an array.
+    If the aggregate sizes are not stored internally (which is the default), they are computed and returned.
+    If the aggregate sizes have been stored internally, then they are *not* recomputed, but instead the
+    stored sizes are returned.
 
-    local_graph_type GetGraph() const;
+    @param[in] forceRecompute if true, force recomputation of the aggregate sizes.
+   */
+  typename aggregates_sizes_type::const_type ComputeAggregateSizes(bool forceRecompute = false) const;
 
-    /*! @brief Generates a compressed list of nodes in each aggregate, where
-      the entries in aggNodes[aggPtr[i]] up to aggNodes[aggPtr[i+1]-1] contain the nodes in aggregate i.
-      unaggregated contains the list of nodes which are, for whatever reason, not aggregated (e.g. Dirichlet)
-     */
-    void ComputeNodesInAggregate(LO_view & aggPtr, LO_view & aggNodes, LO_view & unaggregated) const;
+  /*! @brief Compute sizes of aggregates
 
-    //! Get global number of aggregates
-    //  If # of global aggregates is unknown, this method does coummunication and internally record the value
-    GO GetNumGlobalAggregatesComputeIfNeeded();
+    Returns the number of nodes in each aggregate in an array.
+    If the aggregate sizes are not stored internally (which is the default), they are computed and returned.
+    If the aggregate sizes have been stored internally, then they are *not* recomputed, but instead the
+    stored sizes are returned.
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+    @param[in] forceRecompute if true, force recomputation of the aggregate sizes.
+   */
+  Teuchos::ArrayRCP<LocalOrdinal> ComputeAggregateSizesArrayRCP(bool forceRecompute = false) const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  local_graph_type GetGraph() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = verbLevel_default) const;
+  /*! @brief Generates a compressed list of nodes in each aggregate, where
+    the entries in aggNodes[aggPtr[i]] up to aggNodes[aggPtr[i+1]-1] contain the nodes in aggregate i.
+    unaggregated contains the list of nodes which are, for whatever reason, not aggregated (e.g. Dirichlet)
+   */
+  void ComputeNodesInAggregate(LO_view& aggPtr, LO_view& aggNodes, LO_view& unaggregated) const;
 
-  private:
-    LO   numAggregates_;              ///< Number of aggregates on this processor
-    GO   numGlobalAggregates_;        ///< Number of global aggregates
+  //! Get global number of aggregates
+  //  If # of global aggregates is unknown, this method does coummunication and internally record the value
+  GO GetNumGlobalAggregatesComputeIfNeeded();
 
-    /*! vertex2AggId[k] gives a local id corresponding to the aggregate to which
-     * local id k has been assigned. While k is the local id on my processor (MyPID),
-     * vertex2AggId[k] is the local id on the processor which actually owns the aggregate.
-     */
-    RCP<LOMultiVector> vertex2AggId_;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    /*!
-     * If k is the local id on my processor (MyPID), the owning processor has the
-     * id given by procWinner[k]
-     */
-    RCP<LOVector> procWinner_;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    /*! geoData stores an index manager object that is used to perform structured aggreation
-     *  on a problem.
-     */
-    RCP<IndexManager_kokkos> geoDataKokkos_;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = verbLevel_default) const;
 
-    /*! geoData stores an index manager object that is used to perform structured aggreation
-     *  on a problem.
-     */
-    RCP<IndexManager> geoData_;
+ private:
+  LO numAggregates_;        ///< Number of aggregates on this processor
+  GO numGlobalAggregates_;  ///< Number of global aggregates
 
-    /*! graphColors_ stores a view that assigns a color to each node in the graph
-     *  These colors are used to parallelize the aggregation process in UncoupledAggregation
-     */
-    colors_view_type graphColors_;
+  /*! vertex2AggId[k] gives a local id corresponding to the aggregate to which
+   * local id k has been assigned. While k is the local id on my processor (MyPID),
+   * vertex2AggId[k] is the local id on the processor which actually owns the aggregate.
+   */
+  RCP<LOMultiVector> vertex2AggId_;
 
-    /*! graphNumColors_ stores the number of colors that are needed to perform a distance 2
-     *  coloring of the underlying graph.
-     */
-    LO graphNumColors_;
+  /*!
+   * If k is the local id on my processor (MyPID), the owning processor has the
+   * id given by procWinner[k]
+   */
+  RCP<LOVector> procWinner_;
 
-    //! An ArrayRCP of booleans specifying if a local entry is an aggregate root.
-    Teuchos::ArrayRCP<bool> isRoot_;
+  /*! geoData stores an index manager object that is used to perform structured aggreation
+   *  on a problem.
+   */
+  RCP<IndexManager_kokkos> geoDataKokkos_;
 
-    //! Set to false iff aggregates do not include any DOFs belong to other processes.
-    bool aggregatesIncludeGhosts_;
+  /*! geoData stores an index manager object that is used to perform structured aggreation
+   *  on a problem.
+   */
+  RCP<IndexManager> geoData_;
 
-    //! Array of sizes of each local aggregate.
-    mutable
-    aggregates_sizes_type aggregateSizes_;
+  /*! graphColors_ stores a view that assigns a color to each node in the graph
+   *  These colors are used to parallelize the aggregation process in UncoupledAggregation
+   */
+  colors_view_type graphColors_;
 
-    /*! aggragateSizesHost_ is a host copy of aggregate sizes, which
-     * helps slightly reduce the cost of calling ComputeAggregateSizes
-     * from different parts of MueLu that require such data on the host device.
-     */
-    mutable
-    typename aggregates_sizes_type::HostMirror aggregateSizesHost_;
+  /*! graphNumColors_ stores the number of colors that are needed to perform a distance 2
+   *  coloring of the underlying graph.
+   */
+  LO graphNumColors_;
 
-    //! Aggregates represented as Kokkos graph type
-    mutable
-    local_graph_type graph_;
-  };
+  //! An ArrayRCP of booleans specifying if a local entry is an aggregate root.
+  Teuchos::ArrayRCP<bool> isRoot_;
 
-} //namespace MueLu
+  //! Set to false iff aggregates do not include any DOFs belong to other processes.
+  bool aggregatesIncludeGhosts_;
+
+  //! Array of sizes of each local aggregate.
+  mutable aggregates_sizes_type aggregateSizes_;
+
+  /*! aggragateSizesHost_ is a host copy of aggregate sizes, which
+   * helps slightly reduce the cost of calling ComputeAggregateSizes
+   * from different parts of MueLu that require such data on the host device.
+   */
+  mutable
+      typename aggregates_sizes_type::HostMirror aggregateSizesHost_;
+
+  //! Aggregates represented as Kokkos graph type
+  mutable local_graph_type graph_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_AGGREGATES_SHORT
-#endif // MUELU_AGGREGATES_DECL_HPP
+#endif  // MUELU_AGGREGATES_DECL_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_Aggregates_def.hpp b/packages/muelu/src/Graph/Containers/MueLu_Aggregates_def.hpp
index c9940f5524ac..aec11bce17c3 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_Aggregates_def.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_Aggregates_def.hpp
@@ -59,254 +59,262 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::Aggregates(const GraphBase & graph) {
-    numAggregates_  = 0;
-    numGlobalAggregates_ = 0;
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Aggregates(const GraphBase& graph) {
+  numAggregates_       = 0;
+  numGlobalAggregates_ = 0;
 
-    vertex2AggId_ = LOMultiVectorFactory::Build(graph.GetImportMap(), 1);
-    vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
+  vertex2AggId_ = LOMultiVectorFactory::Build(graph.GetImportMap(), 1);
+  vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
 
-    procWinner_ = LOVectorFactory::Build(graph.GetImportMap());
-    procWinner_->putScalar(MUELU_UNASSIGNED);
+  procWinner_ = LOVectorFactory::Build(graph.GetImportMap());
+  procWinner_->putScalar(MUELU_UNASSIGNED);
 
-    isRoot_ = Teuchos::ArrayRCP<bool>(graph.GetImportMap()->getLocalNumElements(), false);
+  isRoot_ = Teuchos::ArrayRCP<bool>(graph.GetImportMap()->getLocalNumElements(), false);
 
-    // slow but safe, force TentativePFactory to build column map for P itself
-    aggregatesIncludeGhosts_ = true;
-  }
+  // slow but safe, force TentativePFactory to build column map for P itself
+  aggregatesIncludeGhosts_ = true;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::
-  Aggregates(LWGraph_kokkos graph) {
-    numAggregates_  = 0;
-    numGlobalAggregates_  = 0;
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    Aggregates(LWGraph_kokkos graph) {
+  numAggregates_       = 0;
+  numGlobalAggregates_ = 0;
 
-    vertex2AggId_ = LOMultiVectorFactory::Build(graph.GetImportMap(), 1);
-    vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
+  vertex2AggId_ = LOMultiVectorFactory::Build(graph.GetImportMap(), 1);
+  vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
 
-    procWinner_ = LOVectorFactory::Build(graph.GetImportMap());
-    procWinner_->putScalar(MUELU_UNASSIGNED);
+  procWinner_ = LOVectorFactory::Build(graph.GetImportMap());
+  procWinner_->putScalar(MUELU_UNASSIGNED);
 
-    isRoot_ = Teuchos::ArrayRCP<bool>(graph.GetImportMap()->getLocalNumElements(), false);
+  isRoot_ = Teuchos::ArrayRCP<bool>(graph.GetImportMap()->getLocalNumElements(), false);
 
-    // slow but safe, force TentativePFactory to build column map for P itself
-    aggregatesIncludeGhosts_ = true;
-  }
+  // slow but safe, force TentativePFactory to build column map for P itself
+  aggregatesIncludeGhosts_ = true;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::
-  Aggregates(const RCP<const Map>& map) {
-    numAggregates_ = 0;
-    numGlobalAggregates_  = 0;
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    Aggregates(const RCP<const Map>& map) {
+  numAggregates_       = 0;
+  numGlobalAggregates_ = 0;
 
-    vertex2AggId_ = LOMultiVectorFactory::Build(map, 1);
-    vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
+  vertex2AggId_ = LOMultiVectorFactory::Build(map, 1);
+  vertex2AggId_->putScalar(MUELU_UNAGGREGATED);
 
-    procWinner_ = LOVectorFactory::Build(map);
-    procWinner_->putScalar(MUELU_UNASSIGNED);
+  procWinner_ = LOVectorFactory::Build(map);
+  procWinner_->putScalar(MUELU_UNASSIGNED);
 
-    isRoot_ = Teuchos::ArrayRCP<bool>(map->getLocalNumElements(), false);
+  isRoot_ = Teuchos::ArrayRCP<bool>(map->getLocalNumElements(), false);
 
-    // slow but safe, force TentativePFactory to build column map for P itself
-    aggregatesIncludeGhosts_ = true;
-  }
+  // slow but safe, force TentativePFactory to build column map for P itself
+  aggregatesIncludeGhosts_ = true;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  typename Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::aggregates_sizes_type::const_type
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::ComputeAggregateSizes(bool forceRecompute) const {
-    if (aggregateSizes_.size() && !forceRecompute) {
-      return aggregateSizes_;
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+typename Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::aggregates_sizes_type::const_type
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::ComputeAggregateSizes(bool forceRecompute) const {
+  if (aggregateSizes_.size() && !forceRecompute) {
+    return aggregateSizes_;
 
-    } else {
-      // It is necessary to initialize this to 0
-      aggregates_sizes_type aggregateSizes("aggregates", numAggregates_);
+  } else {
+    // It is necessary to initialize this to 0
+    aggregates_sizes_type aggregateSizes("aggregates", numAggregates_);
 
-      int myPID = GetMap()->getComm()->getRank();
+    int myPID = GetMap()->getComm()->getRank();
 
-      auto vertex2AggId = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      auto procWinner   = procWinner_  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto vertex2AggId = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto procWinner   = procWinner_->getDeviceLocalView(Xpetra::Access::ReadOnly);
 
-      typename AppendTrait<decltype(aggregateSizes_), Kokkos::Atomic>::type aggregateSizesAtomic = aggregateSizes;
-      Kokkos::parallel_for("MueLu:Aggregates:ComputeAggregateSizes:for", range_type(0,procWinner.size()),
+    typename AppendTrait<decltype(aggregateSizes_), Kokkos::Atomic>::type aggregateSizesAtomic = aggregateSizes;
+    Kokkos::parallel_for(
+        "MueLu:Aggregates:ComputeAggregateSizes:for", range_type(0, procWinner.size()),
         KOKKOS_LAMBDA(const LO i) {
           if (procWinner(i, 0) == myPID)
             aggregateSizesAtomic(vertex2AggId(i, 0))++;
         });
 
-      aggregateSizes_ = aggregateSizes;
-
-      return aggregateSizes;
-    }
+    aggregateSizes_ = aggregateSizes;
 
+    return aggregateSizes;
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  typename Teuchos::ArrayRCP<LocalOrdinal>
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::
-  ComputeAggregateSizesArrayRCP(bool forceRecompute) const {
-    auto aggregateSizes = this->ComputeAggregateSizes(forceRecompute);
-    
-    // if this is the first time this is called, setup the host mirror and fill it
-    if(!aggregateSizesHost_.is_allocated()) {
-      aggregateSizesHost_ = Kokkos::create_mirror_view(aggregateSizes);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+typename Teuchos::ArrayRCP<LocalOrdinal>
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    ComputeAggregateSizesArrayRCP(bool forceRecompute) const {
+  auto aggregateSizes = this->ComputeAggregateSizes(forceRecompute);
+
+  // if this is the first time this is called, setup the host mirror and fill it
+  if (!aggregateSizesHost_.is_allocated()) {
+    aggregateSizesHost_ = Kokkos::create_mirror_view(aggregateSizes);
+    Kokkos::deep_copy(aggregateSizesHost_, aggregateSizes);
+  } else {
+    // otherwise, only update if we forced a recompute
+    if (forceRecompute)
       Kokkos::deep_copy(aggregateSizesHost_, aggregateSizes);
-    } else {
-      // otherwise, only update if we forced a recompute
-      if(forceRecompute)
-        Kokkos::deep_copy(aggregateSizesHost_, aggregateSizes);
-    }
+  }
 
-    // put the data in an ArrayRCP, but do not give it ownership of the data
-    Teuchos::ArrayRCP<LocalOrdinal> aggregateSizesArrayRCP(aggregateSizesHost_.data(),0,aggregateSizesHost_.extent(0),false);
+  // put the data in an ArrayRCP, but do not give it ownership of the data
+  Teuchos::ArrayRCP<LocalOrdinal> aggregateSizesArrayRCP(aggregateSizesHost_.data(), 0, aggregateSizesHost_.extent(0), false);
 
-    return aggregateSizesArrayRCP;
-  }
+  return aggregateSizesArrayRCP;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  typename Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::local_graph_type
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::GetGraph() const {
-    using row_map_type = typename local_graph_type::row_map_type;
-    using entries_type = typename local_graph_type::entries_type;
-    using size_type    = typename local_graph_type::size_type;
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+typename Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::local_graph_type
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetGraph() const {
+  using row_map_type = typename local_graph_type::row_map_type;
+  using entries_type = typename local_graph_type::entries_type;
+  using size_type    = typename local_graph_type::size_type;
 
-    auto numAggregates = numAggregates_;
+  auto numAggregates = numAggregates_;
 
-    if (static_cast<LO>(graph_.numRows()) == numAggregates)
-      return graph_;
+  if (static_cast<LO>(graph_.numRows()) == numAggregates)
+    return graph_;
 
-    auto vertex2AggId = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto procWinner   = procWinner_  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto sizes        = ComputeAggregateSizes();
+  auto vertex2AggId = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto procWinner   = procWinner_->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto sizes        = ComputeAggregateSizes();
 
-    // FIXME_KOKKOS: replace by ViewAllocateWithoutInitializing + rows(0) = 0.
-    typename row_map_type::non_const_type rows("Agg_rows", numAggregates+1);  // rows(0) = 0 automatically
+  // FIXME_KOKKOS: replace by ViewAllocateWithoutInitializing + rows(0) = 0.
+  typename row_map_type::non_const_type rows("Agg_rows", numAggregates + 1);  // rows(0) = 0 automatically
 
-    // parallel_scan (exclusive)
-    Kokkos::parallel_scan("MueLu:Aggregates:GetGraph:compute_rows", range_type(0, numAggregates),
+  // parallel_scan (exclusive)
+  Kokkos::parallel_scan(
+      "MueLu:Aggregates:GetGraph:compute_rows", range_type(0, numAggregates),
       KOKKOS_LAMBDA(const LO i, LO& update, const bool& final_pass) {
         update += sizes(i);
         if (final_pass)
-          rows(i+1) = update;
+          rows(i + 1) = update;
       });
 
-    decltype(rows) offsets(Kokkos::ViewAllocateWithoutInitializing("Agg_offsets"), numAggregates+1); // +1 is just for ease
-    Kokkos::deep_copy(offsets, rows);
+  decltype(rows) offsets(Kokkos::ViewAllocateWithoutInitializing("Agg_offsets"), numAggregates + 1);  // +1 is just for ease
+  Kokkos::deep_copy(offsets, rows);
 
-    int myPID = GetMap()->getComm()->getRank();
+  int myPID = GetMap()->getComm()->getRank();
 
-    size_type numNNZ;
-    {
-      Kokkos::View<size_type, device_type> numNNZ_device = Kokkos::subview(rows, numAggregates);
-      typename Kokkos::View<size_type, device_type>::HostMirror numNNZ_host = Kokkos::create_mirror_view(numNNZ_device);
-      Kokkos::deep_copy(numNNZ_host, numNNZ_device);
-      numNNZ = numNNZ_host();
-    }
-    typename entries_type::non_const_type cols(Kokkos::ViewAllocateWithoutInitializing("Agg_cols"), numNNZ);
-    size_t realnnz = 0;
-    Kokkos::parallel_reduce("MueLu:Aggregates:GetGraph:compute_cols", range_type(0, procWinner.size()),
+  size_type numNNZ;
+  {
+    Kokkos::View<size_type, device_type> numNNZ_device                    = Kokkos::subview(rows, numAggregates);
+    typename Kokkos::View<size_type, device_type>::HostMirror numNNZ_host = Kokkos::create_mirror_view(numNNZ_device);
+    Kokkos::deep_copy(numNNZ_host, numNNZ_device);
+    numNNZ = numNNZ_host();
+  }
+  typename entries_type::non_const_type cols(Kokkos::ViewAllocateWithoutInitializing("Agg_cols"), numNNZ);
+  size_t realnnz = 0;
+  Kokkos::parallel_reduce(
+      "MueLu:Aggregates:GetGraph:compute_cols", range_type(0, procWinner.size()),
       KOKKOS_LAMBDA(const LO i, size_t& nnz) {
         if (procWinner(i, 0) == myPID) {
-          typedef typename std::remove_reference< decltype( offsets(0) ) >::type atomic_incr_type;
-          auto idx = Kokkos::atomic_fetch_add( &offsets(vertex2AggId(i,0)), atomic_incr_type(1));
+          typedef typename std::remove_reference<decltype(offsets(0))>::type atomic_incr_type;
+          auto idx  = Kokkos::atomic_fetch_add(&offsets(vertex2AggId(i, 0)), atomic_incr_type(1));
           cols(idx) = i;
           nnz++;
         }
-      }, realnnz);
-    TEUCHOS_TEST_FOR_EXCEPTION(realnnz != numNNZ, Exceptions::RuntimeError,
-                               "MueLu: Internal error: Something is wrong with aggregates graph construction: numNNZ = " << numNNZ << " != " << realnnz << " = realnnz");
-
-    graph_ = local_graph_type(cols, rows);
-
-    return graph_;
-  }
-  
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void 
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::ComputeNodesInAggregate(LO_view & aggPtr, LO_view & aggNodes, LO_view & unaggregated) const {
-    LO numAggs  = GetNumAggregates();
-    LO numNodes = vertex2AggId_->getLocalLength();
-    auto vertex2AggId = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    typename aggregates_sizes_type::const_type aggSizes = ComputeAggregateSizes(true);
-    LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-
-    aggPtr = LO_view("aggPtr",numAggs+1);
-    aggNodes = LO_view("aggNodes",numNodes);
-    LO_view aggCurr("agg curr",numAggs+1);
-
-    // Construct the "rowptr" and the counter
-    Kokkos::parallel_scan("MueLu:Aggregates:ComputeNodesInAggregate:scan", range_type(0,numAggs+1),
+      },
+      realnnz);
+  TEUCHOS_TEST_FOR_EXCEPTION(realnnz != numNNZ, Exceptions::RuntimeError,
+                             "MueLu: Internal error: Something is wrong with aggregates graph construction: numNNZ = " << numNNZ << " != " << realnnz << " = realnnz");
+
+  graph_ = local_graph_type(cols, rows);
+
+  return graph_;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::ComputeNodesInAggregate(LO_view& aggPtr, LO_view& aggNodes, LO_view& unaggregated) const {
+  LO numAggs                                          = GetNumAggregates();
+  LO numNodes                                         = vertex2AggId_->getLocalLength();
+  auto vertex2AggId                                   = vertex2AggId_->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  typename aggregates_sizes_type::const_type aggSizes = ComputeAggregateSizes(true);
+  LO INVALID                                          = Teuchos::OrdinalTraits<LO>::invalid();
+
+  aggPtr   = LO_view("aggPtr", numAggs + 1);
+  aggNodes = LO_view("aggNodes", numNodes);
+  LO_view aggCurr("agg curr", numAggs + 1);
+
+  // Construct the "rowptr" and the counter
+  Kokkos::parallel_scan(
+      "MueLu:Aggregates:ComputeNodesInAggregate:scan", range_type(0, numAggs + 1),
       KOKKOS_LAMBDA(const LO aggIdx, LO& aggOffset, bool final_pass) {
         LO count = 0;
-        if(aggIdx < numAggs)
+        if (aggIdx < numAggs)
           count = aggSizes(aggIdx);
-        if(final_pass) {
-          aggPtr(aggIdx) = aggOffset;
+        if (final_pass) {
+          aggPtr(aggIdx)  = aggOffset;
           aggCurr(aggIdx) = aggOffset;
-          if(aggIdx==numAggs)
-            aggCurr(numAggs) = 0; // use this for counting unaggregated nodes
+          if (aggIdx == numAggs)
+            aggCurr(numAggs) = 0;  // use this for counting unaggregated nodes
         }
         aggOffset += count;
       });
 
-    // Preallocate unaggregated to the correct size
-    LO numUnaggregated = 0;
-    Kokkos::parallel_reduce("MueLu:Aggregates:ComputeNodesInAggregate:unaggregatedSize", range_type(0,numNodes),
-      KOKKOS_LAMBDA(const LO nodeIdx, LO & count) { 
-        if(vertex2AggId(nodeIdx,0)==INVALID) 
+  // Preallocate unaggregated to the correct size
+  LO numUnaggregated = 0;
+  Kokkos::parallel_reduce(
+      "MueLu:Aggregates:ComputeNodesInAggregate:unaggregatedSize", range_type(0, numNodes),
+      KOKKOS_LAMBDA(const LO nodeIdx, LO& count) {
+        if (vertex2AggId(nodeIdx, 0) == INVALID)
           count++;
-      }, numUnaggregated);
-    unaggregated = LO_view("unaggregated",numUnaggregated);
+      },
+      numUnaggregated);
+  unaggregated = LO_view("unaggregated", numUnaggregated);
 
-    // Stick the nodes in each aggregate's spot
-    Kokkos::parallel_for("MueLu:Aggregates:ComputeNodesInAggregate:for", range_type(0,numNodes),
+  // Stick the nodes in each aggregate's spot
+  Kokkos::parallel_for(
+      "MueLu:Aggregates:ComputeNodesInAggregate:for", range_type(0, numNodes),
       KOKKOS_LAMBDA(const LO nodeIdx) {
-        LO aggIdx = vertex2AggId(nodeIdx,0);
-        if(aggIdx != INVALID) {
+        LO aggIdx = vertex2AggId(nodeIdx, 0);
+        if (aggIdx != INVALID) {
           // atomic postincrement aggCurr(aggIdx) each time
-          aggNodes(Kokkos::atomic_fetch_add(&aggCurr(aggIdx),1)) = nodeIdx;
+          aggNodes(Kokkos::atomic_fetch_add(&aggCurr(aggIdx), 1)) = nodeIdx;
         } else {
           // same, but using last entry of aggCurr for unaggregated nodes
-          unaggregated(Kokkos::atomic_fetch_add(&aggCurr(numAggs),1)) = nodeIdx;
+          unaggregated(Kokkos::atomic_fetch_add(&aggCurr(numAggs), 1)) = nodeIdx;
         }
       });
-    
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  std::string Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::description() const {
-    if (numGlobalAggregates_ == -1) return BaseClass::description() + "{nGlobalAggregates = not computed}"; 
-    else                            return BaseClass::description() + "{nGlobalAggregates = " + toString(numGlobalAggregates_) + "}";
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+std::string Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::description() const {
+  if (numGlobalAggregates_ == -1)
+    return BaseClass::description() + "{nGlobalAggregates = not computed}";
+  else
+    return BaseClass::description() + "{nGlobalAggregates = " + toString(numGlobalAggregates_) + "}";
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::print(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Statistics1) {
+    if (numGlobalAggregates_ == -1)
+      out0 << "Global number of aggregates: not computed " << std::endl;
+    else
+      out0 << "Global number of aggregates: " << numGlobalAggregates_ << std::endl;
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::print(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Statistics1) {
-      if (numGlobalAggregates_ == -1) out0 << "Global number of aggregates: not computed " << std::endl;
-      else                            out0 << "Global number of aggregates: " << numGlobalAggregates_ << std::endl;
-    }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+GlobalOrdinal Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetNumGlobalAggregatesComputeIfNeeded() {
+  if (numGlobalAggregates_ != -1) {
+    LO nAggregates = GetNumAggregates();
+    GO nGlobalAggregates;
+    MueLu_sumAll(vertex2AggId_->getMap()->getComm(), (GO)nAggregates, nGlobalAggregates);
+    SetNumGlobalAggregates(nGlobalAggregates);
   }
+  return numGlobalAggregates_;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  GlobalOrdinal Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> >::GetNumGlobalAggregatesComputeIfNeeded() {
-
-    if (numGlobalAggregates_ != -1) {
-      LO nAggregates = GetNumAggregates();
-      GO nGlobalAggregates;
-      MueLu_sumAll(vertex2AggId_->getMap()->getComm(), (GO)nAggregates, nGlobalAggregates);
-      SetNumGlobalAggregates(nGlobalAggregates);
-    }
-    return numGlobalAggregates_;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  const RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> >
-  Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetMap() const {
-    return vertex2AggId_->getMap();
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+const RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>>
+Aggregates<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetMap() const {
+  return vertex2AggId_->getMap();
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_AGGREGATES_DEF_HPP
+#endif  // MUELU_AGGREGATES_DEF_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_GraphBase.hpp b/packages/muelu/src/Graph/Containers/MueLu_GraphBase.hpp
index 462f8aef3153..eadb45c6fa36 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_GraphBase.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_GraphBase.hpp
@@ -46,7 +46,7 @@
 #ifndef MUELU_GRAPHBASE_HPP
 #define MUELU_GRAPHBASE_HPP
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
+#include <Xpetra_ConfigDefs.hpp>  // global_size_t
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_ConfigDefs.hpp"
@@ -61,69 +61,68 @@ namespace MueLu {
 
    Pure virtual base class for MueLu representations of graphs.
 */
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class GraphBase
-    : public BaseClass {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GraphBase
+  : public BaseClass {
 #undef MUELU_GRAPHBASE_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    // For Zoltan2 compatibility
-    using lno_t  = LocalOrdinal;
-    using gno_t  = GlobalOrdinal;
-    using node_t = Node;
+ public:
+  // For Zoltan2 compatibility
+  using lno_t  = LocalOrdinal;
+  using gno_t  = GlobalOrdinal;
+  using node_t = Node;
 
-    //! @name Constructors/Destructors.
-    //@{
-    virtual ~GraphBase() {};
-    //@}
+  //! @name Constructors/Destructors.
+  //@{
+  virtual ~GraphBase(){};
+  //@}
 
-    virtual const RCP<const Teuchos::Comm<int> > GetComm() const = 0;
-    virtual const RCP<const Map> GetDomainMap() const = 0;
-    virtual const RCP<const Map> GetImportMap() const = 0;
+  virtual const RCP<const Teuchos::Comm<int> > GetComm() const = 0;
+  virtual const RCP<const Map> GetDomainMap() const            = 0;
+  virtual const RCP<const Map> GetImportMap() const            = 0;
 
-    //! @name Query graph attributes.
-    //@{
+  //! @name Query graph attributes.
+  //@{
 
-    //! Return number of vertices owned by the calling node.
-    virtual size_t GetNodeNumVertices() const = 0;
+  //! Return number of vertices owned by the calling node.
+  virtual size_t GetNodeNumVertices() const = 0;
 
-    //! Return number of edges owned by the calling node.
-    virtual size_t GetNodeNumEdges()    const = 0;
+  //! Return number of edges owned by the calling node.
+  virtual size_t GetNodeNumEdges() const = 0;
 
-    virtual void   SetBoundaryNodeMap(const ArrayRCP<const bool > & boundaryArray) = 0;
+  virtual void SetBoundaryNodeMap(const ArrayRCP<const bool> &boundaryArray) = 0;
 
-    virtual size_t getLocalMaxNumRowEntries() const = 0;
+  virtual size_t getLocalMaxNumRowEntries() const = 0;
 
-    virtual const ArrayRCP<const bool> GetBoundaryNodeMap() const = 0;
+  virtual const ArrayRCP<const bool> GetBoundaryNodeMap() const = 0;
 
-    //FIXME is this necessary?
-    //! Return number of global edges in the graph.
-    virtual Xpetra::global_size_t GetGlobalNumEdges() const = 0;
+  // FIXME is this necessary?
+  //! Return number of global edges in the graph.
+  virtual Xpetra::global_size_t GetGlobalNumEdges() const = 0;
 
-    //! Return the list of vertices adjacent to the vertex 'v'.
-    virtual Teuchos::ArrayView<const LocalOrdinal> getNeighborVertices(LocalOrdinal v) const = 0;
+  //! Return the list of vertices adjacent to the vertex 'v'.
+  virtual Teuchos::ArrayView<const LocalOrdinal> getNeighborVertices(LocalOrdinal v) const = 0;
 
-    //! Return true if vertex with local id 'v' is on current process.
-    virtual bool isLocalNeighborVertex(LocalOrdinal v) const = 0;
-    //@}
+  //! Return true if vertex with local id 'v' is on current process.
+  virtual bool isLocalNeighborVertex(LocalOrdinal v) const = 0;
+  //@}
 
-    //! @name Print graph.
-    //@{
-    /// Return a simple one-line description of the Graph.
-    virtual std::string description() const = 0;
+  //! @name Print graph.
+  //@{
+  /// Return a simple one-line description of the Graph.
+  virtual std::string description() const = 0;
 
-    //! Print the Graph with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
-    virtual void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const = 0;
-    //@}
+  //! Print the Graph with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
+  virtual void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const = 0;
+  //@}
+};
 
-  };
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GRAPHBASE_SHORT
-#endif // MUELU_GRAPHBASE_HPP
+#endif  // MUELU_GRAPHBASE_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_Graph_decl.hpp b/packages/muelu/src/Graph/Containers/MueLu_Graph_decl.hpp
index 5ecc7c87e000..2136b9320d7f 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_Graph_decl.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_Graph_decl.hpp
@@ -46,8 +46,8 @@
 #ifndef MUELU_GRAPH_DECL_HPP
 #define MUELU_GRAPH_DECL_HPP
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
-#include <Xpetra_CrsGraph_fwd.hpp>     // inline functions requires class declaration
+#include <Xpetra_ConfigDefs.hpp>    // global_size_t
+#include <Xpetra_CrsGraph_fwd.hpp>  // inline functions requires class declaration
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_ConfigDefs.hpp"
@@ -64,78 +64,76 @@ namespace MueLu {
    This class holds an underlying Xpetra_CrsGraph.
    This class can be considered a facade, as MueLu needs only limited functionality for aggregation.
 */
-  template <class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Graph
-    : public MueLu::GraphBase<LocalOrdinal,GlobalOrdinal,Node> { //FIXME  shortnames isn't working
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Graph
+  : public MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node> {  // FIXME  shortnames isn't working
 #undef MUELU_GRAPH_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+  Graph(const RCP<const CrsGraph>& graph, const std::string& /* objectLabel */ = "");
 
-    //! @name Constructors/Destructors.
-    //@{
-    Graph(const RCP<const CrsGraph> & graph, const std::string & /* objectLabel */="");
+  virtual ~Graph() {}
+  //@}
 
-    virtual ~Graph() {}
-    //@}
+  size_t GetNodeNumVertices() const { return graph_->getLocalNumRows(); }
+  size_t GetNodeNumEdges() const { return graph_->getLocalNumEntries(); }
 
-    size_t GetNodeNumVertices() const                                        { return graph_->getLocalNumRows(); }
-    size_t GetNodeNumEdges()    const                                        { return graph_->getLocalNumEntries(); }
+  Xpetra::global_size_t GetGlobalNumEdges() const { return graph_->getGlobalNumEntries(); }
 
-    Xpetra::global_size_t GetGlobalNumEdges() const                          { return graph_->getGlobalNumEntries(); }
+  const RCP<const Teuchos::Comm<int> > GetComm() const { return graph_->getComm(); }
+  const RCP<const Map> GetDomainMap() const { return graph_->getDomainMap(); }
+  //! Returns overlapping import map (nodes).
+  const RCP<const Map> GetImportMap() const { return graph_->getColMap(); }
 
-    const RCP<const Teuchos::Comm<int> > GetComm() const                     { return graph_->getComm(); }
-    const RCP<const Map> GetDomainMap() const                                { return graph_->getDomainMap(); }
-    //! Returns overlapping import map (nodes).
-    const RCP<const Map> GetImportMap() const                                { return graph_->getColMap();    }
+  const RCP<const CrsGraph> GetGraph() const { return graph_; }
 
-    const RCP<const CrsGraph> GetGraph() const {return graph_;}
+  //! Set map with local ids of boundary nodes.
+  void SetBoundaryNodeMap(const ArrayRCP<const bool>& localDirichletNodes) { localDirichletNodes_ = localDirichletNodes; }
 
-    //! Set map with local ids of boundary nodes.
-    void SetBoundaryNodeMap(const ArrayRCP<const bool>& localDirichletNodes) { localDirichletNodes_ = localDirichletNodes; }
+  //! Returns map with local ids of boundary nodes.
+  const ArrayRCP<const bool> GetBoundaryNodeMap() const { return localDirichletNodes_; }
 
-    //! Returns map with local ids of boundary nodes.
-    const ArrayRCP<const bool> GetBoundaryNodeMap() const                    { return localDirichletNodes_; }
+  //! Returns the maximum number of entries across all rows/columns on this node
+  size_t getLocalMaxNumRowEntries() const { return graph_->getLocalMaxNumRowEntries(); }
 
-    //! Returns the maximum number of entries across all rows/columns on this node
-    size_t getLocalMaxNumRowEntries () const                                  { return graph_->getLocalMaxNumRowEntries(); }
+  //! Return the list of vertices adjacent to the vertex 'v'.
+  ArrayView<const LO> getNeighborVertices(LO i) const {
+    ArrayView<const LO> rowView;
+    graph_->getLocalRowView(i, rowView);
+    return rowView;
+  }
 
-    //! Return the list of vertices adjacent to the vertex 'v'.
-    ArrayView<const LO> getNeighborVertices(LO i) const {
-      ArrayView<const LO> rowView;
-      graph_->getLocalRowView(i, rowView);
-      return rowView;
-    }
-
-    //! Return true if vertex with local id 'v' is on current process.
-    bool isLocalNeighborVertex(LO i) const                                   { return i >= minLocalIndex_ && i <= maxLocalIndex_; }
+  //! Return true if vertex with local id 'v' is on current process.
+  bool isLocalNeighborVertex(LO i) const { return i >= minLocalIndex_ && i <= maxLocalIndex_; }
 
 #ifdef MUELU_UNUSED
-    size_t GetNodeNumGhost() const;
+  size_t GetNodeNumGhost() const;
 #endif
 
-    /// Return a simple one-line description of the Graph.
-    std::string description() const                                          { return "MueLu.description()"; }
-
-    //! Print the Graph with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  /// Return a simple one-line description of the Graph.
+  std::string description() const { return "MueLu.description()"; }
 
-  private:
+  //! Print the Graph with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    RCP<const CrsGraph> graph_;
+ private:
+  RCP<const CrsGraph> graph_;
 
-    //! Vector of Dirichlet boundary node IDs on current process.
-    ArrayRCP<const bool> localDirichletNodes_;
+  //! Vector of Dirichlet boundary node IDs on current process.
+  ArrayRCP<const bool> localDirichletNodes_;
 
-    // local index boundaries (cached from domain map)
-    LO minLocalIndex_, maxLocalIndex_;
-  };
+  // local index boundaries (cached from domain map)
+  LO minLocalIndex_, maxLocalIndex_;
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GRAPH_SHORT
-#endif // MUELU_GRAPH_DECL_HPP
+#endif  // MUELU_GRAPH_DECL_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_Graph_def.hpp b/packages/muelu/src/Graph/Containers/MueLu_Graph_def.hpp
index 7d9bf76e1cf3..d2c7240beafb 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_Graph_def.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_Graph_def.hpp
@@ -54,51 +54,52 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  Graph<LocalOrdinal,GlobalOrdinal,Node>::Graph(const RCP<const CrsGraph> & graph, const std::string & /* objectLabel */) : graph_(graph) {
-      minLocalIndex_ = graph_->getDomainMap()->getMinLocalIndex();
-      maxLocalIndex_ = graph_->getDomainMap()->getMaxLocalIndex();
-    }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Graph<LocalOrdinal, GlobalOrdinal, Node>::Graph(const RCP<const CrsGraph> &graph, const std::string & /* objectLabel */)
+  : graph_(graph) {
+  minLocalIndex_ = graph_->getDomainMap()->getMinLocalIndex();
+  maxLocalIndex_ = graph_->getDomainMap()->getMaxLocalIndex();
+}
 
 #ifdef MUELU_UNUSED
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t Graph<LocalOrdinal, GlobalOrdinal, Node>::GetNodeNumGhost() const {
-    /*
-      Ray's comments about nGhost:
-      Graph->NGhost == graph_->RowMatrixColMap()->NumMyElements() - graph_->MatrixDomainMap()->NumMyElements()
-      is basically right. But we've had some issues about how epetra handles empty columns.
-      Probably worth discussing this with Jonathan and Chris to see if this is ALWAYS right.
-    */
-    size_t nGhost = graph_->getColMap()->getLocalNumElements() - graph_->getDomainMap()->getLocalNumElements();
-    if (nGhost < 0) nGhost = 0; // FIXME: size_t is unsigned.
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t Graph<LocalOrdinal, GlobalOrdinal, Node>::GetNodeNumGhost() const {
+  /*
+    Ray's comments about nGhost:
+    Graph->NGhost == graph_->RowMatrixColMap()->NumMyElements() - graph_->MatrixDomainMap()->NumMyElements()
+    is basically right. But we've had some issues about how epetra handles empty columns.
+    Probably worth discussing this with Jonathan and Chris to see if this is ALWAYS right.
+  */
+  size_t nGhost = graph_->getColMap()->getLocalNumElements() - graph_->getDomainMap()->getLocalNumElements();
+  if (nGhost < 0) nGhost = 0;  // FIXME: size_t is unsigned.
 
-    return nGhost;
-  }
+  return nGhost;
+}
 #endif
 
-  //! Print the object with some verbosity level to an FancyOStream object.
-  //using MueLu::Describable::describe; // overloading, not hiding
-  //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Graph<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      //out0 << "Prec. type: " << type_ << std::endl;
-    }
+//! Print the object with some verbosity level to an FancyOStream object.
+// using MueLu::Describable::describe; // overloading, not hiding
+// void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void Graph<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    if (verbLevel & Parameters1) {
-      //out0 << "Linear Algebra: " << toString(lib_) << std::endl;
-      //out0 << "PrecType: " << type_ << std::endl;
-      //out0 << "Parameter list: " << std::endl; { Teuchos::OSTab tab2(out); out << paramList_; }
-      //out0 << "Overlap: " << overlap_ << std::endl;
-    }
+  if (verbLevel & Parameters0) {
+    // out0 << "Prec. type: " << type_ << std::endl;
+  }
 
-    if (verbLevel & Debug) {
-      graph_->describe(out0, Teuchos::VERB_EXTREME);
-    }
+  if (verbLevel & Parameters1) {
+    // out0 << "Linear Algebra: " << toString(lib_) << std::endl;
+    // out0 << "PrecType: " << type_ << std::endl;
+    // out0 << "Parameter list: " << std::endl; { Teuchos::OSTab tab2(out); out << paramList_; }
+    // out0 << "Overlap: " << overlap_ << std::endl;
   }
 
+  if (verbLevel & Debug) {
+    graph_->describe(out0, Teuchos::VERB_EXTREME);
+  }
 }
 
-#endif // MUELU_GRAPH_DEF_HPP
+}  // namespace MueLu
+
+#endif  // MUELU_GRAPH_DEF_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_decl.hpp b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_decl.hpp
index f6b3c8338cbe..2292b98a2795 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_decl.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_decl.hpp
@@ -46,8 +46,8 @@
 #ifndef MUELU_LWGRAPH_DECL_HPP
 #define MUELU_LWGRAPH_DECL_HPP
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
-#include <Xpetra_CrsGraph_fwd.hpp>     // inline functions requires class declaration
+#include <Xpetra_ConfigDefs.hpp>    // global_size_t
+#include <Xpetra_CrsGraph_fwd.hpp>  // inline functions requires class declaration
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_ConfigDefs.hpp"
@@ -66,117 +66,117 @@ namespace MueLu {
    fillComplete.
    TODO handle systems
 */
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class LWGraph : public MueLu::GraphBase<LocalOrdinal,GlobalOrdinal,Node> {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class LWGraph : public MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_LWGRAPH_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! LWGraph constructor
-    //
-    // @param[in] rowPtrs: Array containing row offsets (CSR format)
-    // @param[in] colPtrs: Array containing local column indices (CSR format)
-    // @param[in] domainMap: non-overlapping (domain) map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
-    // @param[in] importMap: overlapping map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
-    // @param[in] objectLabel: label string
-    LWGraph(const ArrayRCP<const LO>& rowPtrs, const ArrayRCP<const LO>& colPtrs,
-            const RCP<const Map>& domainMap, const RCP<const Map>& importMap, const std::string& objectLabel = "")
-            : rows_(rowPtrs), columns_(colPtrs), domainMap_(domainMap), importMap_(importMap), domainMapRef_(*domainMap), objectLabel_(objectLabel)
-    {
-      minLocalIndex_ = domainMapRef_.getMinLocalIndex();
-      maxLocalIndex_ = domainMapRef_.getMaxLocalIndex();
-
-      maxNumRowEntries_ = 0;
-
-      LO nRows = as<LO>(rowPtrs.size()-1);
-      for (LO i = 0; i < nRows; i++)
-        maxNumRowEntries_ = std::max(maxNumRowEntries_, as<size_t>(rowPtrs[i+1] - rowPtrs[i]));
-    }
-
-    virtual ~LWGraph() {}
-    //@}
-
-    size_t GetNodeNumVertices() const { return rows_.size()-1; }
-    size_t GetNodeNumEdges()    const { return rows_[rows_.size()-1]; }
-
-    // TODO: do we really need this function
-    // It is being called from CoupledAggregation, but do we need it there?
-    Xpetra::global_size_t GetGlobalNumEdges() const {
-      Xpetra::global_size_t in = GetNodeNumEdges(), out;
-      Teuchos::reduceAll(*domainMap_->getComm(), Teuchos::REDUCE_SUM, in, Teuchos::outArg(out));
-      return out;
-    }
-
-    const RCP<const Teuchos::Comm<int> > GetComm()      const    { return domainMap_->getComm(); }
-    const RCP<const Map>                 GetDomainMap() const    { return domainMap_; }
-    //! Returns overlapping import map (nodes).
-    const RCP<const Map>                 GetImportMap() const    { return importMap_; }
-
-    void SetBoundaryNodeMap(RCP<const Map> const &/* map */)           { throw Exceptions::NotImplemented("LWGraph: Boundary node map not implemented."); }
-
-    //! Return the list of vertices adjacent to the vertex 'v'.
-    Teuchos::ArrayView<const LO> getNeighborVertices(LO i) const { return columns_.view(rows_[i], rows_[i+1]-rows_[i]); }
-
-    //! Return true if vertex with local id 'v' is on current process.
-    bool isLocalNeighborVertex(LO i) const                       { return i >= minLocalIndex_ && i <= maxLocalIndex_; }
-
-    //! Set boolean array indicating which rows correspond to Dirichlet boundaries.
-    void SetBoundaryNodeMap(const ArrayRCP<const bool>& bndry)   { dirichletBoundaries_ = bndry; }
-
-    //! Returns the maximum number of entries across all rows/columns on this node
-    size_t getLocalMaxNumRowEntries () const                      { return maxNumRowEntries_; }
-
-    //! Returns map with global ids of boundary nodes.
-    const ArrayRCP<const bool> GetBoundaryNodeMap() const        { return dirichletBoundaries_; }
-
-
-    /// Return a simple one-line description of the Graph.
-    std::string description() const                              { return "MueLu.description()"; } //FIXME use object's label
-
-    //! Return the row pointers of the local graph
-    const ArrayRCP<const LO> getRowPtrs() const {
-      return rows_;
-    }
-
-    //! Return the list entries in the local graph
-    const ArrayRCP<const LO> getEntries() const {
-      return columns_;
-    }
-    
-    //! Print the Graph with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
-
-
-    RCP<CrsGraph> GetCrsGraph() const;
-
-  private:
-
-    //! Indices into columns_ array.  Part of local graph information.
-    const ArrayRCP<const LO> rows_;
-    //! Columns corresponding to connections.  Part of local graph information.
-    const ArrayRCP<const LO> columns_;
-    //! Graph maps
-    const RCP<const Map> domainMap_, importMap_;
-    const Map& domainMapRef_;
-    //! Name of this graph.
-    const std::string objectLabel_;
-    //! Boolean array marking Dirichlet rows.
-    ArrayRCP<const bool> dirichletBoundaries_;
-
-    // local index boundaries (cached from domain map)
-    LO     minLocalIndex_, maxLocalIndex_;
-    size_t maxNumRowEntries_;
-  };
-
-} // namespace MueLu
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! LWGraph constructor
+  //
+  // @param[in] rowPtrs: Array containing row offsets (CSR format)
+  // @param[in] colPtrs: Array containing local column indices (CSR format)
+  // @param[in] domainMap: non-overlapping (domain) map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
+  // @param[in] importMap: overlapping map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
+  // @param[in] objectLabel: label string
+  LWGraph(const ArrayRCP<const LO>& rowPtrs, const ArrayRCP<const LO>& colPtrs,
+          const RCP<const Map>& domainMap, const RCP<const Map>& importMap, const std::string& objectLabel = "")
+    : rows_(rowPtrs)
+    , columns_(colPtrs)
+    , domainMap_(domainMap)
+    , importMap_(importMap)
+    , domainMapRef_(*domainMap)
+    , objectLabel_(objectLabel) {
+    minLocalIndex_ = domainMapRef_.getMinLocalIndex();
+    maxLocalIndex_ = domainMapRef_.getMaxLocalIndex();
+
+    maxNumRowEntries_ = 0;
+
+    LO nRows = as<LO>(rowPtrs.size() - 1);
+    for (LO i = 0; i < nRows; i++)
+      maxNumRowEntries_ = std::max(maxNumRowEntries_, as<size_t>(rowPtrs[i + 1] - rowPtrs[i]));
+  }
+
+  virtual ~LWGraph() {}
+  //@}
+
+  size_t GetNodeNumVertices() const { return rows_.size() - 1; }
+  size_t GetNodeNumEdges() const { return rows_[rows_.size() - 1]; }
+
+  // TODO: do we really need this function
+  // It is being called from CoupledAggregation, but do we need it there?
+  Xpetra::global_size_t GetGlobalNumEdges() const {
+    Xpetra::global_size_t in = GetNodeNumEdges(), out;
+    Teuchos::reduceAll(*domainMap_->getComm(), Teuchos::REDUCE_SUM, in, Teuchos::outArg(out));
+    return out;
+  }
+
+  const RCP<const Teuchos::Comm<int> > GetComm() const { return domainMap_->getComm(); }
+  const RCP<const Map> GetDomainMap() const { return domainMap_; }
+  //! Returns overlapping import map (nodes).
+  const RCP<const Map> GetImportMap() const { return importMap_; }
+
+  void SetBoundaryNodeMap(RCP<const Map> const& /* map */) { throw Exceptions::NotImplemented("LWGraph: Boundary node map not implemented."); }
+
+  //! Return the list of vertices adjacent to the vertex 'v'.
+  Teuchos::ArrayView<const LO> getNeighborVertices(LO i) const { return columns_.view(rows_[i], rows_[i + 1] - rows_[i]); }
+
+  //! Return true if vertex with local id 'v' is on current process.
+  bool isLocalNeighborVertex(LO i) const { return i >= minLocalIndex_ && i <= maxLocalIndex_; }
+
+  //! Set boolean array indicating which rows correspond to Dirichlet boundaries.
+  void SetBoundaryNodeMap(const ArrayRCP<const bool>& bndry) { dirichletBoundaries_ = bndry; }
+
+  //! Returns the maximum number of entries across all rows/columns on this node
+  size_t getLocalMaxNumRowEntries() const { return maxNumRowEntries_; }
+
+  //! Returns map with global ids of boundary nodes.
+  const ArrayRCP<const bool> GetBoundaryNodeMap() const { return dirichletBoundaries_; }
+
+  /// Return a simple one-line description of the Graph.
+  std::string description() const { return "MueLu.description()"; }  // FIXME use object's label
+
+  //! Return the row pointers of the local graph
+  const ArrayRCP<const LO> getRowPtrs() const {
+    return rows_;
+  }
+
+  //! Return the list entries in the local graph
+  const ArrayRCP<const LO> getEntries() const {
+    return columns_;
+  }
+
+  //! Print the Graph with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+
+  RCP<CrsGraph> GetCrsGraph() const;
+
+ private:
+  //! Indices into columns_ array.  Part of local graph information.
+  const ArrayRCP<const LO> rows_;
+  //! Columns corresponding to connections.  Part of local graph information.
+  const ArrayRCP<const LO> columns_;
+  //! Graph maps
+  const RCP<const Map> domainMap_, importMap_;
+  const Map& domainMapRef_;
+  //! Name of this graph.
+  const std::string objectLabel_;
+  //! Boolean array marking Dirichlet rows.
+  ArrayRCP<const bool> dirichletBoundaries_;
+
+  // local index boundaries (cached from domain map)
+  LO minLocalIndex_, maxLocalIndex_;
+  size_t maxNumRowEntries_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_LWGRAPH_SHORT
-#endif // MUELU_LWGRAPH_DECL_HPP
+#endif  // MUELU_LWGRAPH_DECL_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_def.hpp b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_def.hpp
index 230a6c908587..1f0e907adaa9 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_def.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_def.hpp
@@ -52,46 +52,45 @@
 
 namespace MueLu {
 
-  //! Print the object with some verbosity level to an FancyOStream object.
-  //using MueLu::Describable::describe; // overloading, not hiding
-  //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LWGraph<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    //    MUELU_DESCRIBE;
+//! Print the object with some verbosity level to an FancyOStream object.
+// using MueLu::Describable::describe; // overloading, not hiding
+// void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LWGraph<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  //    MUELU_DESCRIBE;
 
-    if (verbLevel & Parameters0) {
-      //out0 << "Prec. type: " << type_ << std::endl;
-    }
-
-    if (verbLevel & Parameters1) {
-      //out0 << "Linear Algebra: " << toString(lib_) << std::endl;
-      //out0 << "PrecType: " << type_ << std::endl;
-      //out0 << "Parameter list: " << std::endl; { Teuchos::OSTab tab2(out); out << paramList_; }
-      //out0 << "Overlap: " << overlap_ << std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      RCP<const Map> col_map = importMap_.is_null() ? domainMap_ : importMap_;
+  if (verbLevel & Parameters0) {
+    // out0 << "Prec. type: " << type_ << std::endl;
+  }
 
-      for (LO i = 0; i < rows_.size()-1; i++) {
-        for (LO j = rows_[i]; j < rows_[i+1]; j++)
-          out<< domainMap_->getGlobalElement(i) << " " << col_map->getGlobalElement(columns_[j])<<std::endl;
-      }
-    }
+  if (verbLevel & Parameters1) {
+    // out0 << "Linear Algebra: " << toString(lib_) << std::endl;
+    // out0 << "PrecType: " << type_ << std::endl;
+    // out0 << "Parameter list: " << std::endl; { Teuchos::OSTab tab2(out); out << paramList_; }
+    // out0 << "Overlap: " << overlap_ << std::endl;
   }
 
+  if (verbLevel & Debug) {
+    RCP<const Map> col_map = importMap_.is_null() ? domainMap_ : importMap_;
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::CrsGraph<LocalOrdinal,GlobalOrdinal,Node> > LWGraph<LocalOrdinal, GlobalOrdinal, Node>::GetCrsGraph() const {
-      ArrayRCP<size_t> rowPtrs;
-      rowPtrs.resize(rows_.size());
-      for (size_t i=0; i<Teuchos::as<size_t>(rows_.size()); i++)
-        rowPtrs[i] = rows_[i];
-      auto graph =  Xpetra::CrsGraphFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(GetDomainMap(), GetImportMap(), rowPtrs, Teuchos::arcp_const_cast<LO>(getEntries()));
-      graph->fillComplete();
-      return graph;
+    for (LO i = 0; i < rows_.size() - 1; i++) {
+      for (LO j = rows_[i]; j < rows_[i + 1]; j++)
+        out << domainMap_->getGlobalElement(i) << " " << col_map->getGlobalElement(columns_[j]) << std::endl;
     }
+  }
+}
 
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > LWGraph<LocalOrdinal, GlobalOrdinal, Node>::GetCrsGraph() const {
+  ArrayRCP<size_t> rowPtrs;
+  rowPtrs.resize(rows_.size());
+  for (size_t i = 0; i < Teuchos::as<size_t>(rows_.size()); i++)
+    rowPtrs[i] = rows_[i];
+  auto graph = Xpetra::CrsGraphFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(GetDomainMap(), GetImportMap(), rowPtrs, Teuchos::arcp_const_cast<LO>(getEntries()));
+  graph->fillComplete();
+  return graph;
 }
 
-#endif // MUELU_LWGRAPH_DEF_HPP
+}  // namespace MueLu
+
+#endif  // MUELU_LWGRAPH_DEF_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_decl.hpp b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_decl.hpp
index dd8a0231b15a..6debffc42f98 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_decl.hpp
@@ -51,7 +51,7 @@
 #include <Kokkos_StaticCrsGraph.hpp>
 #include <Tpetra_KokkosCompat_ClassicNodeAPI_Wrapper.hpp>
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
+#include <Xpetra_ConfigDefs.hpp>  // global_size_t
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_VerbosityLevel.hpp"
@@ -62,110 +62,111 @@
 
 namespace MueLu {
 
-  /*!
-    @class LWGraph_kokkos
-    @brief Lightweight MueLu representation of a compressed row storage graph
-
-    This class is lightweight in the sense that it holds to local graph
-    information. These were built without using fillComplete.
-   */
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LWGraph_kokkos;
-
-  // Partial specialization for DeviceType
-  template<class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
-  public:
-    using local_ordinal_type  = LocalOrdinal;
-    using global_ordinal_type = GlobalOrdinal;
-    using execution_space     = typename DeviceType::execution_space;
-    using memory_space        = typename DeviceType::memory_space;
-    using device_type         = Kokkos::Device<execution_space, memory_space>;
-    using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
-    using local_lw_graph_type = MueLu::LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, node_type>;
-    using size_type           = size_t;
-
-    using map_type            = Xpetra::Map<LocalOrdinal, GlobalOrdinal, node_type>;
-    using local_graph_type    = typename local_lw_graph_type::local_graph_type;
-    using boundary_nodes_type = typename local_lw_graph_type::boundary_nodes_type;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class LWGraph_kokkos
+  @brief Lightweight MueLu representation of a compressed row storage graph
+
+  This class is lightweight in the sense that it holds to local graph
+  information. These were built without using fillComplete.
+ */
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LWGraph_kokkos;
+
+// Partial specialization for DeviceType
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
+ public:
+  using local_ordinal_type  = LocalOrdinal;
+  using global_ordinal_type = GlobalOrdinal;
+  using execution_space     = typename DeviceType::execution_space;
+  using memory_space        = typename DeviceType::memory_space;
+  using device_type         = Kokkos::Device<execution_space, memory_space>;
+  using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
+  using local_lw_graph_type = MueLu::LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, node_type>;
+  using size_type           = size_t;
+
+  using map_type            = Xpetra::Map<LocalOrdinal, GlobalOrdinal, node_type>;
+  using local_graph_type    = typename local_lw_graph_type::local_graph_type;
+  using boundary_nodes_type = typename local_lw_graph_type::boundary_nodes_type;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_LWGRAPH_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! LWGraph constructor
-    //
-    // @param[in] graph: local graph of type Kokkos::StaticCrsGraph containing CRS data
-    // @param[in] domainMap: non-overlapping (domain) map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
-    // @param[in] importMap: overlapping map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
-    // @param[in] objectLabel: label string
-    LWGraph_kokkos(const local_graph_type&    graph,
-                   const RCP<const map_type>& domainMap,
-                   const RCP<const map_type>& importMap,
-                   const std::string&         objectLabel = "")
-      : lclLWGraph_(graph, domainMap), domainMap_(domainMap), importMap_(importMap), objectLabel_(objectLabel) { }
-
-    ~LWGraph_kokkos() = default;
-    //@}
-
-    const RCP<const Teuchos::Comm<int> > GetComm() const {
-      return domainMap_->getComm();
-    }
-    const RCP<const Map> GetDomainMap() const {
-      return domainMap_;
-    }
-    //! Return overlapping import map (nodes).
-    const RCP<const Map> GetImportMap() const {
-      return importMap_;
-    }
-
-    //! Return number of graph vertices
-    KOKKOS_INLINE_FUNCTION size_type GetNodeNumVertices() const {
-      return lclLWGraph_.GetNodeNumVertices();
-    }
-    //! Return number of graph edges
-    KOKKOS_INLINE_FUNCTION size_type GetNodeNumEdges() const {
-      return lclLWGraph_.GetNodeNumEdges();
-    }
-
-    //! Returns the maximum number of entries across all rows/columns on this node
-    KOKKOS_INLINE_FUNCTION size_type getLocalMaxNumRowEntries () const {
-      return lclLWGraph_.getLocalMaxNumRowEntries();
-    }
-
-    /// Return a simple one-line description of the Graph.
-    std::string description() const {
-      return "LWGraph (" + objectLabel_ + ")";
-    }
-
-    //! Print the Graph with some verbosity level to an FancyOStream object.
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
-
-    local_lw_graph_type& getLocalLWGraph() const {
-      return lclLWGraph_;
-    }
-
-  private:
-
-    //! Underlying graph (with label)
-    mutable local_lw_graph_type   lclLWGraph_;
-
-    //! Graph maps
-    const RCP<const map_type>   domainMap_;
-    const RCP<const map_type>   importMap_;
-
-    //! Name of this graph.
-    const std::string objectLabel_;
-  };
-
-}
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! LWGraph constructor
+  //
+  // @param[in] graph: local graph of type Kokkos::StaticCrsGraph containing CRS data
+  // @param[in] domainMap: non-overlapping (domain) map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
+  // @param[in] importMap: overlapping map for graph. Usually provided by AmalgamationFactory stored in UnAmalgamationInfo container
+  // @param[in] objectLabel: label string
+  LWGraph_kokkos(const local_graph_type& graph,
+                 const RCP<const map_type>& domainMap,
+                 const RCP<const map_type>& importMap,
+                 const std::string& objectLabel = "")
+    : lclLWGraph_(graph, domainMap)
+    , domainMap_(domainMap)
+    , importMap_(importMap)
+    , objectLabel_(objectLabel) {}
+
+  ~LWGraph_kokkos() = default;
+  //@}
+
+  const RCP<const Teuchos::Comm<int>> GetComm() const {
+    return domainMap_->getComm();
+  }
+  const RCP<const Map> GetDomainMap() const {
+    return domainMap_;
+  }
+  //! Return overlapping import map (nodes).
+  const RCP<const Map> GetImportMap() const {
+    return importMap_;
+  }
+
+  //! Return number of graph vertices
+  KOKKOS_INLINE_FUNCTION size_type GetNodeNumVertices() const {
+    return lclLWGraph_.GetNodeNumVertices();
+  }
+  //! Return number of graph edges
+  KOKKOS_INLINE_FUNCTION size_type GetNodeNumEdges() const {
+    return lclLWGraph_.GetNodeNumEdges();
+  }
+
+  //! Returns the maximum number of entries across all rows/columns on this node
+  KOKKOS_INLINE_FUNCTION size_type getLocalMaxNumRowEntries() const {
+    return lclLWGraph_.getLocalMaxNumRowEntries();
+  }
+
+  /// Return a simple one-line description of the Graph.
+  std::string description() const {
+    return "LWGraph (" + objectLabel_ + ")";
+  }
+
+  //! Print the Graph with some verbosity level to an FancyOStream object.
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+
+  local_lw_graph_type& getLocalLWGraph() const {
+    return lclLWGraph_;
+  }
+
+ private:
+  //! Underlying graph (with label)
+  mutable local_lw_graph_type lclLWGraph_;
+
+  //! Graph maps
+  const RCP<const map_type> domainMap_;
+  const RCP<const map_type> importMap_;
+
+  //! Name of this graph.
+  const std::string objectLabel_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_LWGRAPH_KOKKOS_SHORT
-#endif // MUELU_LWGRAPH_KOKKOS_DECL_HPP
+#endif  // MUELU_LWGRAPH_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_def.hpp b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_def.hpp
index 4d164f1b8f50..03a55d4fa301 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LWGraph_kokkos_def.hpp
@@ -55,38 +55,37 @@
 
 namespace MueLu {
 
-  template<class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void LWGraph_kokkos<LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  if (verbLevel & Debug) {
+    auto graph             = lclLWGraph_.getGraph();
+    RCP<const Map> col_map = importMap_.is_null() ? domainMap_ : importMap_;
+    int mypid              = col_map->getComm()->getRank();
 
-    if (verbLevel & Debug) {
-      auto graph = lclLWGraph_.getGraph();
-      RCP<const Map> col_map = importMap_.is_null() ? domainMap_ : importMap_;
-      int mypid = col_map->getComm()->getRank();
-
-      {
+    {
       std::ostringstream ss;
       ss << "[pid " << mypid << "] num entries=" << graph.entries.size();
       out << ss.str() << std::endl;
-      }
+    }
 
-      const size_t numRows = graph.numRows();
-      auto rowPtrs = graph.row_map;
-      auto columns = graph.entries;
-      for (size_t i=0; i < numRows; ++i) {
-        std::ostringstream ss;
-        ss << "[pid " << mypid << "] row " << domainMap_->getGlobalElement(i) << ":";
-        ss << " (numEntries=" << rowPtrs(i+1)-rowPtrs(i) << ")";
+    const size_t numRows = graph.numRows();
+    auto rowPtrs         = graph.row_map;
+    auto columns         = graph.entries;
+    for (size_t i = 0; i < numRows; ++i) {
+      std::ostringstream ss;
+      ss << "[pid " << mypid << "] row " << domainMap_->getGlobalElement(i) << ":";
+      ss << " (numEntries=" << rowPtrs(i + 1) - rowPtrs(i) << ")";
 
-        auto rowView = graph.rowConst(i);
-        for (LO j = 0; j < rowView.length; j++) {
-          ss << " " << col_map->getGlobalElement(rowView.colidx(j));
-        }
-        out << ss.str() << std::endl;
+      auto rowView = graph.rowConst(i);
+      for (LO j = 0; j < rowView.length; j++) {
+        ss << " " << col_map->getGlobalElement(rowView.colidx(j));
       }
+      out << ss.str() << std::endl;
     }
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_LWGRAPH_KOKKOS_DEF_HPP
+#endif  // MUELU_LWGRAPH_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LinkedList.cpp b/packages/muelu/src/Graph/Containers/MueLu_LinkedList.cpp
index 04192eee8d72..b1c15756ccb3 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LinkedList.cpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LinkedList.cpp
@@ -43,51 +43,53 @@
 // ***********************************************************************
 //
 // @HEADER
-#include <stddef.h>                     // for NULL
+#include <stddef.h>  // for NULL
 #include "MueLu_LinkedList.hpp"
 
 namespace MueLu {
 
-  LinkedList::LinkedList() : nodeHead(NULL), nodeTail(NULL) { }
+LinkedList::LinkedList()
+  : nodeHead(NULL)
+  , nodeTail(NULL) {}
 
-  LinkedList::~LinkedList() {
-    while (nodeHead != NULL)
-      DeleteHead();
-  }
+LinkedList::~LinkedList() {
+  while (nodeHead != NULL)
+    DeleteHead();
+}
 
-  bool LinkedList::IsEmpty() {
-    return nodeHead == NULL;
-  }
+bool LinkedList::IsEmpty() {
+  return nodeHead == NULL;
+}
 
-  void LinkedList::Add(int iNode) {
-    MueLu_Node *newNode = new MueLu_Node;
-    newNode->nodeId = iNode;
-    newNode->next = NULL;
-    if (nodeHead == NULL) {
-      nodeHead = newNode;
-      nodeTail = newNode;
-    } else {
-      nodeTail->next = newNode;
-      nodeTail = newNode;
-    }
+void LinkedList::Add(int iNode) {
+  MueLu_Node *newNode = new MueLu_Node;
+  newNode->nodeId     = iNode;
+  newNode->next       = NULL;
+  if (nodeHead == NULL) {
+    nodeHead = newNode;
+    nodeTail = newNode;
+  } else {
+    nodeTail->next = newNode;
+    nodeTail       = newNode;
   }
+}
 
-  int LinkedList::Pop() { // get head and remove first node
-    if (IsEmpty()) return -1;
-
-    int iNode = nodeHead->nodeId;
-    DeleteHead();
-    return iNode;
-  }
+int LinkedList::Pop() {  // get head and remove first node
+  if (IsEmpty()) return -1;
 
-  void LinkedList::DeleteHead() {
-    if (IsEmpty()) return;
+  int iNode = nodeHead->nodeId;
+  DeleteHead();
+  return iNode;
+}
 
-    MueLu_Node *newNode = nodeHead;
-    nodeHead = newNode->next;
-    delete newNode;
-  }
+void LinkedList::DeleteHead() {
+  if (IsEmpty()) return;
 
+  MueLu_Node *newNode = nodeHead;
+  nodeHead            = newNode->next;
+  delete newNode;
 }
 
-//TODO: nodeTail unused -> remove?
+}  // namespace MueLu
+
+// TODO: nodeTail unused -> remove?
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LinkedList.hpp b/packages/muelu/src/Graph/Containers/MueLu_LinkedList.hpp
index ba8d95e30751..8546ec3ccfec 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LinkedList.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LinkedList.hpp
@@ -52,33 +52,30 @@
 
 namespace MueLu {
 
-  typedef struct MueLu_Node_Struct
-  {
-    int nodeId;
-    struct MueLu_Node_Struct *next;
-  } MueLu_Node;
+typedef struct MueLu_Node_Struct {
+  int nodeId;
+  struct MueLu_Node_Struct *next;
+} MueLu_Node;
 
-  class LinkedList {
+class LinkedList {
+ public:
+  LinkedList();
 
-  public:
-    LinkedList();
+  ~LinkedList();
 
-    ~LinkedList();
+  bool IsEmpty();
 
-    bool IsEmpty();
+  void Add(int iNode);
 
-    void Add(int iNode);
+  int Pop();
 
-    int Pop();
+ private:
+  MueLu_Node *nodeHead;
+  MueLu_Node *nodeTail;
 
-  private:
-    MueLu_Node *nodeHead;
-    MueLu_Node *nodeTail;
+  void DeleteHead();
+};
 
-    void DeleteHead();
+}  // namespace MueLu
 
-  };
-
-}
-
-#endif // MUELU_LINKEDLIST_HPP
+#endif  // MUELU_LINKEDLIST_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_decl.hpp b/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_decl.hpp
index 294fe160e530..80f24eea4bbb 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_decl.hpp
@@ -51,7 +51,7 @@
 #include <Kokkos_StaticCrsGraph.hpp>
 #include <Tpetra_KokkosCompat_ClassicNodeAPI_Wrapper.hpp>
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
+#include <Xpetra_ConfigDefs.hpp>  // global_size_t
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_VerbosityLevel.hpp"
@@ -61,125 +61,122 @@
 
 namespace MueLu {
 
-  /*!
-    @class LocalLWGraph_kokkos
-    @brief Lightweight MueLu representation of a compressed row storage graph
-
-    This class is lightweight in the sense that it holds to local graph
-    information. These were built without using fillComplete.
-   */
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalLWGraph_kokkos;
-
-  // Partial specialization for DeviceType
-  template<class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
-  public:
-    using local_ordinal_type  = LocalOrdinal;
-    using global_ordinal_type = GlobalOrdinal;
-    using execution_space     = typename DeviceType::execution_space;
-    using memory_space        = typename DeviceType::memory_space;
-    using device_type         = Kokkos::Device<execution_space, memory_space>;
-    using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
-    using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
-    using size_type           = size_t;
-
-    using local_graph_type    = Kokkos::StaticCrsGraph<LocalOrdinal,
-                                                       Kokkos::LayoutLeft,
-                                                       device_type, void, size_t>;
-    using boundary_nodes_type = Kokkos::View<const bool*, memory_space>;
-    using row_type            = Kokkos::View<const LocalOrdinal*, memory_space>;
-    using map_type            = Xpetra::Map<LocalOrdinal, GlobalOrdinal, node_type>;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class LocalLWGraph_kokkos
+  @brief Lightweight MueLu representation of a compressed row storage graph
+
+  This class is lightweight in the sense that it holds to local graph
+  information. These were built without using fillComplete.
+ */
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalLWGraph_kokkos;
+
+// Partial specialization for DeviceType
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
+ public:
+  using local_ordinal_type  = LocalOrdinal;
+  using global_ordinal_type = GlobalOrdinal;
+  using execution_space     = typename DeviceType::execution_space;
+  using memory_space        = typename DeviceType::memory_space;
+  using device_type         = Kokkos::Device<execution_space, memory_space>;
+  using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
+  using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
+  using size_type           = size_t;
+
+  using local_graph_type    = Kokkos::StaticCrsGraph<LocalOrdinal,
+                                                  Kokkos::LayoutLeft,
+                                                  device_type, void, size_t>;
+  using boundary_nodes_type = Kokkos::View<const bool*, memory_space>;
+  using row_type            = Kokkos::View<const LocalOrdinal*, memory_space>;
+  using map_type            = Xpetra::Map<LocalOrdinal, GlobalOrdinal, node_type>;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_LOCALLWGRAPH_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! LocalLWGraph constructor
-    //
-    // @param[in] graph: local graph of type Kokkos::StaticCrsGraph containing CRS data
-    LocalLWGraph_kokkos(const local_graph_type& graph,
-                        const RCP<const map_type>& domainMap);
-
-    ~LocalLWGraph_kokkos() = default;
-    //@}
-
-    //! Return number of graph vertices
-    KOKKOS_INLINE_FUNCTION size_type GetNodeNumVertices() const {
-      return graph_.numRows();
-    }
-    //! Return number of graph edges
-    KOKKOS_INLINE_FUNCTION size_type GetNodeNumEdges() const {
-      return graph_.row_map(GetNodeNumVertices());
-    }
-
-    //! Returns the maximum number of entries across all rows/columns on this node
-    KOKKOS_INLINE_FUNCTION size_type getLocalMaxNumRowEntries () const {
-      return maxNumRowEntries_;
-    }
-
-    //! Return the row pointers of the local graph
-    KOKKOS_INLINE_FUNCTION typename local_graph_type::row_map_type getRowPtrs() const {
-      return graph_.row_map;
-    }
-
-    //! Return the list entries in the local graph
-    KOKKOS_INLINE_FUNCTION typename local_graph_type::entries_type getEntries() const {
-      return graph_.entries;
-    }
-
-    //! Return the list of vertices adjacent to the vertex 'v'.
-    // Unfortunately, C++11 does not support the following:
-    //    auto getNeighborVertices(LO i) const -> decltype(rowView)
-    // auto return with decltype was only introduced in C++14
-    KOKKOS_INLINE_FUNCTION
-    Kokkos::GraphRowViewConst<local_graph_type> getNeighborVertices(LO i) const {
-      auto rowView = graph_.rowConst(i);
-
-      return rowView;
-    }
-
-    //! Return true if vertex with local id 'v' is on current process.
-    KOKKOS_INLINE_FUNCTION bool isLocalNeighborVertex(LO i) const {
-      return i >= minLocalIndex_ && i <= maxLocalIndex_;
-    }
-
-    //! Set boolean array indicating which rows correspond to Dirichlet boundaries.
-    KOKKOS_INLINE_FUNCTION void SetBoundaryNodeMap(const boundary_nodes_type bndry) {
-      dirichletBoundaries_ = bndry;
-    }
-
-    //! Returns map with global ids of boundary nodes.
-    KOKKOS_INLINE_FUNCTION const boundary_nodes_type GetBoundaryNodeMap() const {
-      return dirichletBoundaries_;
-    }
-
-    const local_graph_type& getGraph() const {
-      return graph_;
-    }
-
-  private:
-
-    //! Underlying graph (with label)
-    const local_graph_type      graph_;
-
-    //! Boolean array marking Dirichlet rows.
-    boundary_nodes_type         dirichletBoundaries_;
-
-    //! Local index boundaries (cached from domain map)
-    LO        minLocalIndex_, maxLocalIndex_;
-    size_type maxNumRowEntries_;
-
-  };
-
-}
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! LocalLWGraph constructor
+  //
+  // @param[in] graph: local graph of type Kokkos::StaticCrsGraph containing CRS data
+  LocalLWGraph_kokkos(const local_graph_type& graph,
+                      const RCP<const map_type>& domainMap);
+
+  ~LocalLWGraph_kokkos() = default;
+  //@}
+
+  //! Return number of graph vertices
+  KOKKOS_INLINE_FUNCTION size_type GetNodeNumVertices() const {
+    return graph_.numRows();
+  }
+  //! Return number of graph edges
+  KOKKOS_INLINE_FUNCTION size_type GetNodeNumEdges() const {
+    return graph_.row_map(GetNodeNumVertices());
+  }
+
+  //! Returns the maximum number of entries across all rows/columns on this node
+  KOKKOS_INLINE_FUNCTION size_type getLocalMaxNumRowEntries() const {
+    return maxNumRowEntries_;
+  }
+
+  //! Return the row pointers of the local graph
+  KOKKOS_INLINE_FUNCTION typename local_graph_type::row_map_type getRowPtrs() const {
+    return graph_.row_map;
+  }
+
+  //! Return the list entries in the local graph
+  KOKKOS_INLINE_FUNCTION typename local_graph_type::entries_type getEntries() const {
+    return graph_.entries;
+  }
+
+  //! Return the list of vertices adjacent to the vertex 'v'.
+  // Unfortunately, C++11 does not support the following:
+  //    auto getNeighborVertices(LO i) const -> decltype(rowView)
+  // auto return with decltype was only introduced in C++14
+  KOKKOS_INLINE_FUNCTION
+  Kokkos::GraphRowViewConst<local_graph_type> getNeighborVertices(LO i) const {
+    auto rowView = graph_.rowConst(i);
+
+    return rowView;
+  }
+
+  //! Return true if vertex with local id 'v' is on current process.
+  KOKKOS_INLINE_FUNCTION bool isLocalNeighborVertex(LO i) const {
+    return i >= minLocalIndex_ && i <= maxLocalIndex_;
+  }
+
+  //! Set boolean array indicating which rows correspond to Dirichlet boundaries.
+  KOKKOS_INLINE_FUNCTION void SetBoundaryNodeMap(const boundary_nodes_type bndry) {
+    dirichletBoundaries_ = bndry;
+  }
+
+  //! Returns map with global ids of boundary nodes.
+  KOKKOS_INLINE_FUNCTION const boundary_nodes_type GetBoundaryNodeMap() const {
+    return dirichletBoundaries_;
+  }
+
+  const local_graph_type& getGraph() const {
+    return graph_;
+  }
+
+ private:
+  //! Underlying graph (with label)
+  const local_graph_type graph_;
+
+  //! Boolean array marking Dirichlet rows.
+  boundary_nodes_type dirichletBoundaries_;
+
+  //! Local index boundaries (cached from domain map)
+  LO minLocalIndex_, maxLocalIndex_;
+  size_type maxNumRowEntries_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_LOCALLWGRAPH_KOKKOS_SHORT
-#endif // MUELU_LOCALLWGRAPH_KOKKOS_DECL_HPP
+#endif  // MUELU_LOCALLWGRAPH_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_def.hpp b/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_def.hpp
index 785706b1a002..2edf949396a4 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_LocalLWGraph_kokkos_def.hpp
@@ -55,49 +55,49 @@
 
 namespace MueLu {
 
-  namespace { // anonymous
+namespace {  // anonymous
 
-    template<class LocalOrdinal, class RowType>
-    class MaxNumRowEntriesFunctor {
-    public:
-      MaxNumRowEntriesFunctor(RowType rowPointers) : rowPointers_(rowPointers) { }
+template <class LocalOrdinal, class RowType>
+class MaxNumRowEntriesFunctor {
+ public:
+  MaxNumRowEntriesFunctor(RowType rowPointers)
+    : rowPointers_(rowPointers) {}
 
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LocalOrdinal i, size_t& maxLength) const {
-        size_t d = rowPointers_(i+1) - rowPointers_(i);
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LocalOrdinal i, size_t& maxLength) const {
+    size_t d = rowPointers_(i + 1) - rowPointers_(i);
 
-        maxLength = (d > maxLength ? d : maxLength);
-      }
+    maxLength = (d > maxLength ? d : maxLength);
+  }
+
+  KOKKOS_INLINE_FUNCTION
+  void join(volatile size_t& dest, const volatile size_t& src) {
+    dest = (dest > src ? dest : src);
+  }
 
-      KOKKOS_INLINE_FUNCTION
-      void join(volatile size_t& dest, const volatile size_t& src) {
-        dest = (dest > src ? dest : src);
-      }
+  KOKKOS_INLINE_FUNCTION
+  void init(size_t& initValue) {
+    initValue = 0;
+  }
 
-      KOKKOS_INLINE_FUNCTION
-      void init(size_t& initValue) {
-        initValue = 0;
-      }
+ private:
+  RowType rowPointers_;
+};
 
-    private:
-      RowType rowPointers_;
-    };
+}  // namespace
 
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    LocalLWGraph_kokkos(const local_graph_type& graph,
+                        const RCP<const map_type>& domainMap)
+  : graph_(graph) {
+  minLocalIndex_ = domainMap->getMinLocalIndex();
+  maxLocalIndex_ = domainMap->getMaxLocalIndex();
 
-  template<class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  LocalLWGraph_kokkos<LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  LocalLWGraph_kokkos(const local_graph_type& graph,
-                      const RCP<const map_type>& domainMap)
-    : graph_(graph)
-  {
-    minLocalIndex_ = domainMap->getMinLocalIndex();
-    maxLocalIndex_ = domainMap->getMaxLocalIndex();
-
-    MaxNumRowEntriesFunctor<LO,typename local_graph_type::row_map_type> maxNumRowEntriesFunctor(graph_.row_map);
-    Kokkos::parallel_reduce("MueLu:LocalLWGraph:LWGraph:maxnonzeros", range_type(0,graph_.numRows()), maxNumRowEntriesFunctor, maxNumRowEntries_);
-  }
+  MaxNumRowEntriesFunctor<LO, typename local_graph_type::row_map_type> maxNumRowEntriesFunctor(graph_.row_map);
+  Kokkos::parallel_reduce("MueLu:LocalLWGraph:LWGraph:maxnonzeros", range_type(0, graph_.numRows()), maxNumRowEntriesFunctor, maxNumRowEntries_);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_LWGRAPH_KOKKOS_DEF_HPP
+#endif  // MUELU_LWGRAPH_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/Containers/MueLu_Zoltan2GraphAdapter.hpp b/packages/muelu/src/Graph/Containers/MueLu_Zoltan2GraphAdapter.hpp
index 655e8663550d..e8398e08809b 100644
--- a/packages/muelu/src/Graph/Containers/MueLu_Zoltan2GraphAdapter.hpp
+++ b/packages/muelu/src/Graph/Containers/MueLu_Zoltan2GraphAdapter.hpp
@@ -61,70 +61,63 @@
 #include <Zoltan2_PartitioningSolution.hpp>
 #include "MueLu_GraphBase.hpp"
 
-
-
 // Zoltab2 InputTraits for MueLu Graph objects
 namespace Zoltan2 {
 
 template <typename LocalOrdinal,
           typename GlobalOrdinal,
           typename Node>
-struct InputTraits<MueLu::GraphBase<LocalOrdinal,GlobalOrdinal,Node> >
-{
+struct InputTraits<MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node> > {
   typedef Zoltan2::default_scalar_t scalar_t;
-  typedef LocalOrdinal  lno_t;
+  typedef LocalOrdinal lno_t;
   typedef GlobalOrdinal gno_t;
   typedef size_t offset_t;
-  typedef Zoltan2::default_part_t  part_t;
-  typedef Node          node_t;
-  static inline std::string name() {return "MueLu::Graph";}
+  typedef Zoltan2::default_part_t part_t;
+  typedef Node node_t;
+  static inline std::string name() { return "MueLu::Graph"; }
 
-  Z2_STATIC_ASSERT_TYPES // validate the types
+  Z2_STATIC_ASSERT_TYPES  // validate the types
 };
-}//end namespace Zoltan2
-
+}  // end namespace Zoltan2
 
 namespace MueLu {
 
-template <typename User, typename UserCoord=User>
-class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User,UserCoord> {
-public:
-
+template <typename User, typename UserCoord = User>
+class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User, UserCoord> {
+ public:
 #ifndef DOXYGEN_SHOULD_SKIP_THIS
-  typedef typename Zoltan2::InputTraits<User>::scalar_t    scalar_t;
-  typedef typename Zoltan2::InputTraits<User>::offset_t    offset_t;
-  typedef typename Zoltan2::InputTraits<User>::lno_t    lno_t;
-  typedef typename Zoltan2::InputTraits<User>::gno_t    gno_t;
-  typedef typename Zoltan2::InputTraits<User>::part_t   part_t;
-  typedef typename Zoltan2::InputTraits<User>::node_t   node_t;
+  typedef typename Zoltan2::InputTraits<User>::scalar_t scalar_t;
+  typedef typename Zoltan2::InputTraits<User>::offset_t offset_t;
+  typedef typename Zoltan2::InputTraits<User>::lno_t lno_t;
+  typedef typename Zoltan2::InputTraits<User>::gno_t gno_t;
+  typedef typename Zoltan2::InputTraits<User>::part_t part_t;
+  typedef typename Zoltan2::InputTraits<User>::node_t node_t;
   typedef User xgraph_t;
   typedef User user_t;
   typedef UserCoord userCoord_t;
 #endif
 
   //! MueLu::GraphBase Compatibility Layer
-  const Teuchos::RCP< const Teuchos::Comm< int > >  getComm() const { return graph_->GetComm();}
-  const Teuchos::RCP< const Xpetra::Map<lno_t, gno_t, node_t> > getRowMap() const { return graph_->GetDomainMap();}
-  const RCP< const Xpetra::Map<lno_t, gno_t, node_t> > getColMap() const {
+  const Teuchos::RCP<const Teuchos::Comm<int> > getComm() const { return graph_->GetComm(); }
+  const Teuchos::RCP<const Xpetra::Map<lno_t, gno_t, node_t> > getRowMap() const { return graph_->GetDomainMap(); }
+  const RCP<const Xpetra::Map<lno_t, gno_t, node_t> > getColMap() const {
     // For some GraphBases' this is a ColMap, in others it is a seperate map that is
     // only non-null in parallel.
-    Teuchos::RCP<const Xpetra::Map<lno_t,gno_t,node_t> > map =  graph_->GetImportMap();
-    if(map.is_null()) map = graph_->GetDomainMap();
+    Teuchos::RCP<const Xpetra::Map<lno_t, gno_t, node_t> > map = graph_->GetImportMap();
+    if (map.is_null()) map = graph_->GetDomainMap();
     return map;
   }
-  size_t getLocalNumEntries() const { return graph_->GetNodeNumEdges();}
-  size_t getLocalNumRows() const { return getRowMap()->getLocalNumElements();}
-  size_t getLocalNumCols() const { return getColMap()->getLocalNumElements();}
+  size_t getLocalNumEntries() const { return graph_->GetNodeNumEdges(); }
+  size_t getLocalNumRows() const { return getRowMap()->getLocalNumElements(); }
+  size_t getLocalNumCols() const { return getColMap()->getLocalNumElements(); }
 
-  void getLocalRowView(lno_t LocalRow, Teuchos::ArrayView< const lno_t > &indices) const {
-   indices = graph_->getNeighborVertices(LocalRow);
+  void getLocalRowView(lno_t LocalRow, Teuchos::ArrayView<const lno_t> &indices) const {
+    indices = graph_->getNeighborVertices(LocalRow);
   }
 
-
-
   /*! \brief Destructor
    */
-  ~MueLuGraphBaseAdapter() { }
+  ~MueLuGraphBaseAdapter() {}
 
   /*! \brief Constructor for graph with no weights or coordinates.
    *  \param ingraph the Epetra_CrsGraph, Tpetra::CrsGraph or Xpetra::CrsGraph
@@ -135,8 +128,8 @@ class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User,UserCoord> {
    * one does because the user is obviously a Trilinos user.
    */
 
-   MueLuGraphBaseAdapter(const RCP<const User> &ingraph,
-                      int nVtxWeights=0, int nEdgeWeights=0);
+  MueLuGraphBaseAdapter(const RCP<const User> &ingraph,
+                        int nVtxWeights = 0, int nEdgeWeights = 0);
 
   /*! \brief Provide a pointer to weights for the primary entity type.
    *    \param val A pointer to the weights for index \c idx.
@@ -227,8 +220,7 @@ class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User,UserCoord> {
   // TODO:  Need to add option for columns or nonzeros?
   size_t getLocalNumVertices() const { return getLocalNumRows(); }
 
-  void getVertexIDsView(const gno_t *&ids) const
-  {
+  void getVertexIDsView(const gno_t *&ids) const {
     ids = NULL;
     if (getLocalNumVertices())
       ids = getRowMap()->getLocalElementList().getRawPtr();
@@ -236,67 +228,57 @@ class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User,UserCoord> {
 
   size_t getLocalNumEdges() const { return getLocalNumEntries(); }
 
-  void getEdgesView(const offset_t *&offsets, const gno_t *&adjIds) const
-  {
+  void getEdgesView(const offset_t *&offsets, const gno_t *&adjIds) const {
     offsets = offs_.getRawPtr();
-    adjIds = (getLocalNumEdges() ? adjids_.getRawPtr() : NULL);
+    adjIds  = (getLocalNumEdges() ? adjids_.getRawPtr() : NULL);
   }
 
-  int getNumWeightsPerVertex() const { return nWeightsPerVertex_;}
+  int getNumWeightsPerVertex() const { return nWeightsPerVertex_; }
 
   void getVertexWeightsView(const scalar_t *&weights, int &stride,
-                            int idx) const
-  {
-    if(idx<0 || idx >= nWeightsPerVertex_)
-    {
+                            int idx) const {
+    if (idx < 0 || idx >= nWeightsPerVertex_) {
       std::ostringstream emsg;
       emsg << __FILE__ << ":" << __LINE__
            << "  Invalid vertex weight index " << idx << std::endl;
       throw std::runtime_error(emsg.str());
     }
 
-
     size_t length;
     vertexWeights_[idx].getStridedList(length, weights, stride);
   }
 
-  bool useDegreeAsVertexWeight(int idx) const {return vertexDegreeWeight_[idx];}
+  bool useDegreeAsVertexWeight(int idx) const { return vertexDegreeWeight_[idx]; }
 
-  int getNumWeightsPerEdge() const { return nWeightsPerEdge_;}
+  int getNumWeightsPerEdge() const { return nWeightsPerEdge_; }
 
-  void getEdgeWeightsView(const scalar_t *&weights, int &stride, int idx) const
-  {
-    if(idx<0 || idx >= nWeightsPerEdge_)
-    {
+  void getEdgeWeightsView(const scalar_t *&weights, int &stride, int idx) const {
+    if (idx < 0 || idx >= nWeightsPerEdge_) {
       std::ostringstream emsg;
       emsg << __FILE__ << ":" << __LINE__
            << "  Invalid edge weight index " << idx << std::endl;
       throw std::runtime_error(emsg.str());
     }
 
-
     size_t length;
     edgeWeights_[idx].getStridedList(length, weights, stride);
   }
 
-
   template <typename Adapter>
   void applyPartitioningSolution(const User &in, User *&out,
                                  const Zoltan2::PartitioningSolution<Adapter> &solution) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(1, std::invalid_argument,"applyPartitionlingSolution not implemeneted");
-}
+    TEUCHOS_TEST_FOR_EXCEPTION(1, std::invalid_argument, "applyPartitionlingSolution not implemeneted");
+  }
 
   template <typename Adapter>
   void applyPartitioningSolution(const User &in, RCP<User> &out,
                                  const Zoltan2::PartitioningSolution<Adapter> &solution) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(1, std::invalid_argument,"applyPartitionlingSolution not implemeneted");
+    TEUCHOS_TEST_FOR_EXCEPTION(1, std::invalid_argument, "applyPartitionlingSolution not implemeneted");
   }
 
-
-private:
-
-  RCP<const User > ingraph_;
-  RCP<const xgraph_t > graph_;
+ private:
+  RCP<const User> ingraph_;
+  RCP<const xgraph_t> graph_;
   RCP<const Teuchos::Comm<int> > comm_;
 
   ArrayRCP<const offset_t> offs_;
@@ -311,67 +293,69 @@ class MueLuGraphBaseAdapter : public Zoltan2::GraphAdapter<User,UserCoord> {
 
   int coordinateDim_;
   ArrayRCP<Zoltan2::StridedData<lno_t, scalar_t> > coords_;
-
 };
 
-
 /////////////////////////////////////////////////////////////////
 // Definitions
 /////////////////////////////////////////////////////////////////
 
 template <typename User, typename UserCoord>
-  MueLuGraphBaseAdapter<User,UserCoord>::MueLuGraphBaseAdapter(
-    const RCP<const User> &ingraph, int nVtxWgts, int nEdgeWgts):
-      ingraph_(ingraph), graph_(), comm_() , offs_(), adjids_(),
-      nWeightsPerVertex_(nVtxWgts), vertexWeights_(), vertexDegreeWeight_(),
-      nWeightsPerEdge_(nEdgeWgts), edgeWeights_(),
-      coordinateDim_(0), coords_()
-{
-  typedef Zoltan2::StridedData<lno_t,scalar_t> input_t;
+MueLuGraphBaseAdapter<User, UserCoord>::MueLuGraphBaseAdapter(
+    const RCP<const User> &ingraph, int nVtxWgts, int nEdgeWgts)
+  : ingraph_(ingraph)
+  , graph_()
+  , comm_()
+  , offs_()
+  , adjids_()
+  , nWeightsPerVertex_(nVtxWgts)
+  , vertexWeights_()
+  , vertexDegreeWeight_()
+  , nWeightsPerEdge_(nEdgeWgts)
+  , edgeWeights_()
+  , coordinateDim_(0)
+  , coords_() {
+  typedef Zoltan2::StridedData<lno_t, scalar_t> input_t;
   graph_ = ingraph;
 
-  comm_ = getRowMap()->getComm();
-  size_t nvtx = getLocalNumRows();
+  comm_         = getRowMap()->getComm();
+  size_t nvtx   = getLocalNumRows();
   size_t nedges = getLocalNumEntries();
 
   // Unfortunately we have to copy the offsets and edge Ids
   // because edge Ids are not usually stored in vertex id order.
   size_t n = nvtx + 1;
   offs_.resize(n);
-  offset_t* offs = const_cast<offset_t*>(offs_.getRawPtr());
-  gno_t* adjids=0;
-  if(nedges > 0) {
+  offset_t *offs = const_cast<offset_t *>(offs_.getRawPtr());
+  gno_t *adjids  = 0;
+  if (nedges > 0) {
     adjids_.resize(nedges);
-    adjids = const_cast<gno_t*>(adjids_.getRawPtr());
+    adjids = const_cast<gno_t *>(adjids_.getRawPtr());
   }
 
   offs[0] = 0;
-  for (size_t v=0; v < nvtx; v++){
+  for (size_t v = 0; v < nvtx; v++) {
     ArrayView<const lno_t> nbors;
     getLocalRowView(v, nbors);
-    offs[v+1] = offs[v] + nbors.size();
-    for (offset_t e=offs[v], i=0; e < offs[v+1]; e++) {
+    offs[v + 1] = offs[v] + nbors.size();
+    for (offset_t e = offs[v], i = 0; e < offs[v + 1]; e++) {
       adjids[e] = getColMap()->getGlobalElement(nbors[i++]);
     }
   }
 
   if (nWeightsPerVertex_ > 0) {
     vertexWeights_ =
-          arcp(new input_t[nWeightsPerVertex_], 0, nWeightsPerVertex_, true);
+        arcp(new input_t[nWeightsPerVertex_], 0, nWeightsPerVertex_, true);
     vertexDegreeWeight_ =
-          arcp(new bool[nWeightsPerVertex_], 0, nWeightsPerVertex_, true);
-    for (int i=0; i < nWeightsPerVertex_; i++)
+        arcp(new bool[nWeightsPerVertex_], 0, nWeightsPerVertex_, true);
+    for (int i = 0; i < nWeightsPerVertex_; i++)
       vertexDegreeWeight_[i] = false;
   }
-
-  
 }
 
 ////////////////////////////////////////////////////////////////////////////
 template <typename User, typename UserCoord>
-  void MueLuGraphBaseAdapter<User,UserCoord>::setWeights(
-    const scalar_t *weightVal, int stride, int idx)
-{
+void MueLuGraphBaseAdapter<User, UserCoord>::setWeights(
+    const scalar_t *weightVal, int stride, int idx) {
   if (this->getPrimaryEntityType() == Zoltan2::GRAPH_VERTEX)
     setVertexWeights(weightVal, stride, idx);
   else
@@ -380,29 +364,26 @@ template <typename User, typename UserCoord>
 
 ////////////////////////////////////////////////////////////////////////////
 template <typename User, typename UserCoord>
-  void MueLuGraphBaseAdapter<User,UserCoord>::setVertexWeights(
-    const scalar_t *weightVal, int stride, int idx)
-{
-  typedef Zoltan2::StridedData<lno_t,scalar_t> input_t;
+void MueLuGraphBaseAdapter<User, UserCoord>::setVertexWeights(
+    const scalar_t *weightVal, int stride, int idx) {
+  typedef Zoltan2::StridedData<lno_t, scalar_t> input_t;
 
-  if(idx<0 || idx >= nWeightsPerVertex_)
-  {
-      std::ostringstream emsg;
-      emsg << __FILE__ << ":" << __LINE__
-           << "  Invalid vertex weight index " << idx << std::endl;
-      throw std::runtime_error(emsg.str());
+  if (idx < 0 || idx >= nWeightsPerVertex_) {
+    std::ostringstream emsg;
+    emsg << __FILE__ << ":" << __LINE__
+         << "  Invalid vertex weight index " << idx << std::endl;
+    throw std::runtime_error(emsg.str());
   }
 
   size_t nvtx = getLocalNumVertices();
-  ArrayRCP<const scalar_t> weightV(weightVal, 0, nvtx*stride, false);
+  ArrayRCP<const scalar_t> weightV(weightVal, 0, nvtx * stride, false);
   vertexWeights_[idx] = input_t(weightV, stride);
 }
 
 ////////////////////////////////////////////////////////////////////////////
 template <typename User, typename UserCoord>
-  void MueLuGraphBaseAdapter<User,UserCoord>::setWeightIsDegree(
-    int idx)
-{
+void MueLuGraphBaseAdapter<User, UserCoord>::setWeightIsDegree(
+    int idx) {
   if (this->getPrimaryEntityType() == Zoltan2::GRAPH_VERTEX)
     setVertexWeightIsDegree(idx);
   else {
@@ -416,15 +397,13 @@ template <typename User, typename UserCoord>
 
 ////////////////////////////////////////////////////////////////////////////
 template <typename User, typename UserCoord>
-  void MueLuGraphBaseAdapter<User,UserCoord>::setVertexWeightIsDegree(
-    int idx)
-{
-  if(idx<0 || idx >= nWeightsPerVertex_)
-  {
-      std::ostringstream emsg;
-      emsg << __FILE__ << ":" << __LINE__
-           << "  Invalid vertex weight index " << idx << std::endl;
-      throw std::runtime_error(emsg.str());
+void MueLuGraphBaseAdapter<User, UserCoord>::setVertexWeightIsDegree(
+    int idx) {
+  if (idx < 0 || idx >= nWeightsPerVertex_) {
+    std::ostringstream emsg;
+    emsg << __FILE__ << ":" << __LINE__
+         << "  Invalid vertex weight index " << idx << std::endl;
+    throw std::runtime_error(emsg.str());
   }
 
   vertexDegreeWeight_[idx] = true;
@@ -432,28 +411,24 @@ template <typename User, typename UserCoord>
 
 ////////////////////////////////////////////////////////////////////////////
 template <typename User, typename UserCoord>
-  void MueLuGraphBaseAdapter<User,UserCoord>::setEdgeWeights(
-    const scalar_t *weightVal, int stride, int idx)
-{
-  typedef Zoltan2::StridedData<lno_t,scalar_t> input_t;
+void MueLuGraphBaseAdapter<User, UserCoord>::setEdgeWeights(
+    const scalar_t *weightVal, int stride, int idx) {
+  typedef Zoltan2::StridedData<lno_t, scalar_t> input_t;
 
-  if(idx<0 || idx >= nWeightsPerEdge_)
-  {
-      std::ostringstream emsg;
-      emsg << __FILE__ << ":" << __LINE__
-           << "  Invalid edge weight index " << idx << std::endl;
-      throw std::runtime_error(emsg.str());
+  if (idx < 0 || idx >= nWeightsPerEdge_) {
+    std::ostringstream emsg;
+    emsg << __FILE__ << ":" << __LINE__
+         << "  Invalid edge weight index " << idx << std::endl;
+    throw std::runtime_error(emsg.str());
   }
 
   size_t nedges = getLocalNumEdges();
-  ArrayRCP<const scalar_t> weightV(weightVal, 0, nedges*stride, false);
+  ArrayRCP<const scalar_t> weightV(weightVal, 0, nedges * stride, false);
   edgeWeights_[idx] = input_t(weightV, stride);
 }
 
+}  // namespace MueLu
 
-}  //namespace MueLu
-
+#endif  // HAVE_MUELU_ZOLTAN2
 
-#endif// HAVE_MUELU_ZOLTAN2
-  
 #endif
diff --git a/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_decl.hpp b/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_decl.hpp
index 879332ca0d44..4e2c5302bbd1 100644
--- a/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_decl.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_HYBRIDAGGREGATIONFACTORY_DECL_HPP_
 #define MUELU_HYBRIDAGGREGATIONFACTORY_DECL_HPP_
 
-
 #include <Xpetra_Map_fwd.hpp>
 
 #include "MueLu_ConfigDefs.hpp"
@@ -129,59 +128,58 @@ namespace MueLu {
     | Aggregates   | HybridAggregationFactory   | Container class with aggregation information. See also Aggregates.
 */
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class HybridAggregationFactory : public SingleLevelFactoryBase {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class HybridAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_HYBRIDAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    HybridAggregationFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~HybridAggregationFactory() { }
+  //! Constructor.
+  HybridAggregationFactory();
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~HybridAggregationFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*! @brief Build aggregates. */
-    void Build(Level &currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    /*! @brief Specifically build aggregates along interfaces */
-    void BuildInterfaceAggregates(Level& currentLevel, RCP<Aggregates> aggregates,
-                                  std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes,
-                                  Array<LO> coarseRate) const;
+  /*! @brief Build aggregates. */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  /*! @brief Specifically build aggregates along interfaces */
+  void BuildInterfaceAggregates(Level& currentLevel, RCP<Aggregates> aggregates,
+                                std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes,
+                                Array<LO> coarseRate) const;
 
-  private:
+  //@}
 
-    //! aggregation algorithms
-    // will be filled in Build routine
-    mutable std::vector<RCP<MueLu::AggregationAlgorithmBase<LO, GO, Node> > > algos_;
+ private:
+  //! aggregation algorithms
+  // will be filled in Build routine
+  mutable std::vector<RCP<MueLu::AggregationAlgorithmBase<LO, GO, Node> > > algos_;
 
-    //! boolean flag: definition phase
-    //! if true, the aggregation algorithms still can be set and changed.
-    //! if false, no change in aggregation algorithms is possible any more
-    mutable bool bDefinitionPhase_;
+  //! boolean flag: definition phase
+  //! if true, the aggregation algorithms still can be set and changed.
+  //! if false, no change in aggregation algorithms is possible any more
+  mutable bool bDefinitionPhase_;
 
-  }; // class HybridAggregationFactory
+};  // class HybridAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_HYBRIDAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_HYBRIDAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_def.hpp b/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_def.hpp
index d0ac7bcc7496..36ba2ec572f8 100644
--- a/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/HybridAggregation/MueLu_HybridAggregationFactory_def.hpp
@@ -77,265 +77,258 @@
 #include "MueLu_MasterList.hpp"
 #include "MueLu_Monitor.hpp"
 
-
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
-  HybridAggregationFactory() : bDefinitionPhase_(true)
-  { }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
+    HybridAggregationFactory()
+  : bDefinitionPhase_(true) {}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
-  GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
+    GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+  typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    // From UncoupledAggregationFactory
-    SET_VALID_ENTRY("aggregation: max agg size");
-    SET_VALID_ENTRY("aggregation: min agg size");
-    SET_VALID_ENTRY("aggregation: max selected neighbors");
-    SET_VALID_ENTRY("aggregation: ordering");
-    validParamList->getEntry("aggregation: ordering").setValidator(
-      rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
-    SET_VALID_ENTRY("aggregation: enable phase 1");
-    SET_VALID_ENTRY("aggregation: enable phase 2a");
-    SET_VALID_ENTRY("aggregation: enable phase 2b");
-    SET_VALID_ENTRY("aggregation: enable phase 3");
-    SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
-    SET_VALID_ENTRY("aggregation: allow user-specified singletons");
-    SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
-    SET_VALID_ENTRY("aggregation: match ML phase2a");
-    SET_VALID_ENTRY("aggregation: phase2a agg factor");
-    SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
-
-    // From StructuredAggregationFactory
-    SET_VALID_ENTRY("aggregation: coarsening rate");
-    SET_VALID_ENTRY("aggregation: coarsening order");
-    SET_VALID_ENTRY("aggregation: number of spatial dimensions");
-
-    // From HybridAggregationFactory
-    SET_VALID_ENTRY("aggregation: use interface aggregation");
-#undef  SET_VALID_ENTRY
-
-    /* From UncoupledAggregation */
-    // general variables needed in AggregationFactory
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
-    // special variables necessary for OnePtAggregationAlgorithm
-    validParamList->set<std::string>             ("OnePt aggregate map name",                    "",
-                                                  "Name of input map for single node aggregates. (default='')");
-    validParamList->set<std::string>             ("OnePt aggregate map factory",                 "",
-                                                  "Generating factory of (DOF) map for single node aggregates.");
-
-    // InterfaceAggregation parameters
-    validParamList->set<std::string>             ("Interface aggregate map name",                "",
-                                                  "Name of input map for interface aggregates. (default='')");
-    validParamList->set<std::string>             ("Interface aggregate map factory",             "",
-                                                  "Generating factory of (DOF) map for interface aggregates.");
-    validParamList->set<RCP<const FactoryBase> > ("interfacesDimensions",             Teuchos::null,
-                                                  "Describes the dimensions of all the interfaces on this rank.");
-    validParamList->set<RCP<const FactoryBase> > ("nodeOnInterface",                  Teuchos::null,
-                                                  "List the LIDs of the nodes on any interface.");
-
-    /* From StructuredAggregation */
-    // general variables needed in AggregationFactory
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",           Teuchos::null,
-                                                 "Number of spatial dimension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",                     Teuchos::null,
-                                                 "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-
-
-    // Hybrid Aggregation Params
-    validParamList->set<RCP<const FactoryBase> > ("aggregationRegionType",            Teuchos::null,
-                                                  "Type of aggregation to use on the region (\"structured\" or \"uncoupled\")");
-
-    return validParamList;
+  // From UncoupledAggregationFactory
+  SET_VALID_ENTRY("aggregation: max agg size");
+  SET_VALID_ENTRY("aggregation: min agg size");
+  SET_VALID_ENTRY("aggregation: max selected neighbors");
+  SET_VALID_ENTRY("aggregation: ordering");
+  validParamList->getEntry("aggregation: ordering").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
+  SET_VALID_ENTRY("aggregation: enable phase 1");
+  SET_VALID_ENTRY("aggregation: enable phase 2a");
+  SET_VALID_ENTRY("aggregation: enable phase 2b");
+  SET_VALID_ENTRY("aggregation: enable phase 3");
+  SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
+  SET_VALID_ENTRY("aggregation: allow user-specified singletons");
+  SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
+  SET_VALID_ENTRY("aggregation: match ML phase2a");
+  SET_VALID_ENTRY("aggregation: phase2a agg factor");
+  SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
+
+  // From StructuredAggregationFactory
+  SET_VALID_ENTRY("aggregation: coarsening rate");
+  SET_VALID_ENTRY("aggregation: coarsening order");
+  SET_VALID_ENTRY("aggregation: number of spatial dimensions");
+
+  // From HybridAggregationFactory
+  SET_VALID_ENTRY("aggregation: use interface aggregation");
+#undef SET_VALID_ENTRY
+
+  /* From UncoupledAggregation */
+  // general variables needed in AggregationFactory
+  validParamList->set<RCP<const FactoryBase> >("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+  // special variables necessary for OnePtAggregationAlgorithm
+  validParamList->set<std::string>("OnePt aggregate map name", "",
+                                   "Name of input map for single node aggregates. (default='')");
+  validParamList->set<std::string>("OnePt aggregate map factory", "",
+                                   "Generating factory of (DOF) map for single node aggregates.");
+
+  // InterfaceAggregation parameters
+  validParamList->set<std::string>("Interface aggregate map name", "",
+                                   "Name of input map for interface aggregates. (default='')");
+  validParamList->set<std::string>("Interface aggregate map factory", "",
+                                   "Generating factory of (DOF) map for interface aggregates.");
+  validParamList->set<RCP<const FactoryBase> >("interfacesDimensions", Teuchos::null,
+                                               "Describes the dimensions of all the interfaces on this rank.");
+  validParamList->set<RCP<const FactoryBase> >("nodeOnInterface", Teuchos::null,
+                                               "List the LIDs of the nodes on any interface.");
+
+  /* From StructuredAggregation */
+  // general variables needed in AggregationFactory
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spatial dimension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+
+  // Hybrid Aggregation Params
+  validParamList->set<RCP<const FactoryBase> >("aggregationRegionType", Teuchos::null,
+                                               "Type of aggregation to use on the region (\"structured\" or \"uncoupled\")");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Graph");
+
+  ParameterList pL = GetParameterList();
+
+  /* StructuredAggregation */
+
+  // Request the local number of nodes per dimensions
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("aggregationRegionType", NoFactory::get())) {
+      currentLevel.DeclareInput("aggregationRegionType", NoFactory::get(), this);
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("aggregationRegionType", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "Aggregation region type was not provided by the user!");
+    }
+    if (currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
+      currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "numDimensions was not provided by the user on level0!");
+    }
+    if (currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
+    }
+  } else {
+    Input(currentLevel, "aggregationRegionType");
+    Input(currentLevel, "numDimensions");
+    Input(currentLevel, "lNodesPerDim");
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "Graph");
-
-    ParameterList pL = GetParameterList();
-
-
+  /* UncoupledAggregation */
+  Input(currentLevel, "DofsPerNode");
 
-    /* StructuredAggregation */
-
-    // Request the local number of nodes per dimensions
-    if(currentLevel.GetLevelID() == 0) {
-      if(currentLevel.IsAvailable("aggregationRegionType", NoFactory::get())) {
-        currentLevel.DeclareInput("aggregationRegionType", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("aggregationRegionType",NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "Aggregation region type was not provided by the user!");
-      }
-      if(currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
-        currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
+  // request special data necessary for InterfaceAggregation
+  if (pL.get<bool>("aggregation: use interface aggregation") == true) {
+    if (currentLevel.GetLevelID() == 0) {
+      if (currentLevel.IsAvailable("interfacesDimensions", NoFactory::get())) {
+        currentLevel.DeclareInput("interfacesDimensions", NoFactory::get(), this);
       } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
+        TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("interfacesDimensions", NoFactory::get()),
                                    Exceptions::RuntimeError,
-                                   "numDimensions was not provided by the user on level0!");
+                                   "interfacesDimensions was not provided by the user on level0!");
       }
-      if(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
+      if (currentLevel.IsAvailable("nodeOnInterface", NoFactory::get())) {
+        currentLevel.DeclareInput("nodeOnInterface", NoFactory::get(), this);
       } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+        TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("nodeOnInterface", NoFactory::get()),
                                    Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
+                                   "nodeOnInterface was not provided by the user on level0!");
       }
     } else {
-      Input(currentLevel, "aggregationRegionType");
-      Input(currentLevel, "numDimensions");
-      Input(currentLevel, "lNodesPerDim");
-    }
-
-
-
-    /* UncoupledAggregation */
-    Input(currentLevel, "DofsPerNode");
-
-    // request special data necessary for InterfaceAggregation
-    if (pL.get<bool>("aggregation: use interface aggregation") == true){
-      if(currentLevel.GetLevelID() == 0) {
-        if(currentLevel.IsAvailable("interfacesDimensions", NoFactory::get())) {
-          currentLevel.DeclareInput("interfacesDimensions", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("interfacesDimensions", NoFactory::get()),
-                                       Exceptions::RuntimeError,
-                                       "interfacesDimensions was not provided by the user on level0!");
-        }
-        if(currentLevel.IsAvailable("nodeOnInterface", NoFactory::get())) {
-          currentLevel.DeclareInput("nodeOnInterface", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("nodeOnInterface", NoFactory::get()),
-                                       Exceptions::RuntimeError,
-                                       "nodeOnInterface was not provided by the user on level0!");
-        }
-      } else {
-        Input(currentLevel, "interfacesDimensions");
-        Input(currentLevel, "nodeOnInterface");
-      }
+      Input(currentLevel, "interfacesDimensions");
+      Input(currentLevel, "nodeOnInterface");
     }
+  }
 
-    // request special data necessary for OnePtAggregationAlgorithm
-    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-    if (mapOnePtName.length() > 0) {
-      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-        currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
-      } else {
-        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-        currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
-      }
+  // request special data necessary for OnePtAggregationAlgorithm
+  std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+  if (mapOnePtName.length() > 0) {
+    std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+    if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+      currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
+    } else {
+      RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+      currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
     }
-  } // DeclareInput()
+  }
+}  // DeclareInput()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_HYBRIDAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
+  *out << "Entering hybrid aggregation" << std::endl;
 
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_HYBRIDAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
+  ParameterList pL  = GetParameterList();
+  bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
 
-    *out << "Entering hybrid aggregation" << std::endl;
+  if (pL.get<int>("aggregation: max agg size") == -1)
+    pL.set("aggregation: max agg size", INT_MAX);
 
-    ParameterList pL = GetParameterList();
-    bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
+  // define aggregation algorithms
+  RCP<const FactoryBase> graphFact = GetFactory("Graph");
 
-    if (pL.get<int>("aggregation: max agg size") == -1)
-      pL.set("aggregation: max agg size", INT_MAX);
+  // General problem informations are gathered from data stored in the problem matix.
+  RCP<const GraphBase> graph = Get<RCP<GraphBase> >(currentLevel, "Graph");
+  RCP<const Map> fineMap     = graph->GetDomainMap();
+  const int myRank           = fineMap->getComm()->getRank();
+  const int numRanks         = fineMap->getComm()->getSize();
 
-    // define aggregation algorithms
-    RCP<const FactoryBase> graphFact = GetFactory("Graph");
+  out->setProcRankAndSize(graph->GetImportMap()->getComm()->getRank(),
+                          graph->GetImportMap()->getComm()->getSize());
 
-    // General problem informations are gathered from data stored in the problem matix.
-    RCP<const GraphBase> graph  = Get< RCP<GraphBase> >(currentLevel, "Graph");
-    RCP<const Map> fineMap      = graph->GetDomainMap();
-    const int myRank            = fineMap->getComm()->getRank();
-    const int numRanks          = fineMap->getComm()->getSize();
+  // Build aggregates
+  RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("HB");
 
-    out->setProcRankAndSize(graph->GetImportMap()->getComm()->getRank(),
-                            graph->GetImportMap()->getComm()->getSize());
+  // construct aggStat information
+  const LO numRows = graph->GetNodeNumVertices();
+  std::vector<unsigned> aggStat(numRows, READY);
 
-    // Build aggregates
-    RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("HB");
+  // Get aggregation type for region
+  std::string regionType;
+  if (currentLevel.GetLevelID() == 0) {
+    // On level 0, data is provided by applications and has no associated factory.
+    regionType = currentLevel.Get<std::string>("aggregationRegionType", NoFactory::get());
+  } else {
+    // On level > 0, data is provided directly by generating factories.
+    regionType = Get<std::string>(currentLevel, "aggregationRegionType");
+  }
 
-    // construct aggStat information
-    const LO numRows = graph->GetNodeNumVertices();
-    std::vector<unsigned> aggStat(numRows, READY);
+  int numDimensions = 0;
+  if (currentLevel.GetLevelID() == 0) {
+    // On level 0, data is provided by applications and has no associated factory.
+    numDimensions = currentLevel.Get<int>("numDimensions", NoFactory::get());
+  } else {
+    // On level > 0, data is provided directly by generating factories.
+    numDimensions = Get<int>(currentLevel, "numDimensions");
+  }
 
-    // Get aggregation type for region
-    std::string regionType;
-    if(currentLevel.GetLevelID() == 0) {
+  // Get the coarsening rate (potentially used for both structured and uncoupled aggregation if interface)
+  std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
+  Teuchos::Array<LO> coarseRate;
+  try {
+    coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
+  } catch (const Teuchos::InvalidArrayStringRepresentation& e) {
+    GetOStream(Errors, -1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
+                           << std::endl;
+    throw e;
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
+                             Exceptions::RuntimeError,
+                             "\"aggregation: coarsening rate\" must have at least as many"
+                             " components as the number of spatial dimensions in the problem.");
+
+  algos_.clear();
+  LO numNonAggregatedNodes = numRows;
+  if (regionType == "structured") {
+    // Add AggregationStructuredAlgorithm
+    algos_.push_back(rcp(new AggregationStructuredAlgorithm(graphFact)));
+
+    // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
+    // obtain a nodeMap.
+    const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
+    Array<LO> lFineNodesPerDir(3);
+    if (currentLevel.GetLevelID() == 0) {
       // On level 0, data is provided by applications and has no associated factory.
-      regionType = currentLevel.Get<std::string>("aggregationRegionType", NoFactory::get());
+      lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
     } else {
       // On level > 0, data is provided directly by generating factories.
-      regionType = Get< std::string >(currentLevel, "aggregationRegionType");
+      lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
     }
 
-    int numDimensions = 0;
-    if(currentLevel.GetLevelID() == 0) {
-      // On level 0, data is provided by applications and has no associated factory.
-      numDimensions = currentLevel.Get<int>("numDimensions", NoFactory::get());
-    } else {
-      // On level > 0, data is provided directly by generating factories.
-      numDimensions = Get<int>(currentLevel, "numDimensions");
+    // Set lFineNodesPerDir to 1 for directions beyond numDimensions
+    for (int dim = numDimensions; dim < 3; ++dim) {
+      lFineNodesPerDir[dim] = 1;
     }
 
-    // Get the coarsening rate (potentially used for both structured and uncoupled aggregation if interface)
-    std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
-    Teuchos::Array<LO> coarseRate;
-    try {
-      coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
-    } catch(const Teuchos::InvalidArrayStringRepresentation& e) {
-      GetOStream(Errors,-1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
-                            << std::endl;
-      throw e;
-    }
-    TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
-                               Exceptions::RuntimeError,
-                               "\"aggregation: coarsening rate\" must have at least as many"
-                               " components as the number of spatial dimensions in the problem.");
-
-    algos_.clear();
-    LO numNonAggregatedNodes = numRows;
-    if (regionType == "structured") {
-      // Add AggregationStructuredAlgorithm
-      algos_.push_back(rcp(new AggregationStructuredAlgorithm(graphFact)));
-
-      // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
-      // obtain a nodeMap.
-      const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
-      Array<LO> lFineNodesPerDir(3);
-      if(currentLevel.GetLevelID() == 0) {
-        // On level 0, data is provided by applications and has no associated factory.
-        lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
-      } else {
-        // On level > 0, data is provided directly by generating factories.
-        lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
-      }
-
-      // Set lFineNodesPerDir to 1 for directions beyond numDimensions
-      for(int dim = numDimensions; dim < 3; ++dim) {
-        lFineNodesPerDir[dim] = 1;
-      }
-
-      // Now that we have extracted info from the level, create the IndexManager
-      RCP<MueLu::IndexManager<LO,GO,NO> > geoData;
-      geoData = rcp(new MueLu::UncoupledIndexManager<LO,GO,NO>(fineMap->getComm(),
+    // Now that we have extracted info from the level, create the IndexManager
+    RCP<MueLu::IndexManager<LO, GO, NO> > geoData;
+    geoData = rcp(new MueLu::UncoupledIndexManager<LO, GO, NO>(fineMap->getComm(),
                                                                false,
                                                                numDimensions,
                                                                interpolationOrder,
@@ -345,241 +338,245 @@ namespace MueLu {
                                                                lFineNodesPerDir,
                                                                coarseRate, false));
 
-      TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements()
-                                 != static_cast<size_t>(geoData->getNumLocalFineNodes()),
-                                 Exceptions::RuntimeError,
-                                 "The local number of elements in the graph's map is not equal to "
-                                 "the number of nodes given by: lNodesPerDim!");
-
-      aggregates->SetIndexManager(geoData);
-      aggregates->SetNumAggregates(geoData->getNumLocalCoarseNodes());
-
-      Set(currentLevel, "lCoarseNodesPerDim", geoData->getLocalCoarseNodesPerDir());
-
-    } // end structured aggregation setup
-
-    if (regionType == "uncoupled"){
-      // Add unstructred aggregation phases
-      algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm(graphFact)));
-      if (pL.get<bool>("aggregation: use interface aggregation")       == true)   algos_.push_back(rcp(new InterfaceAggregationAlgorithm         (graphFact)));
-      if (pL.get<bool>("aggregation: allow user-specified singletons") == true)   algos_.push_back(rcp(new OnePtAggregationAlgorithm             (graphFact)));
-      if (pL.get<bool>("aggregation: enable phase 1" )                 == true)   algos_.push_back(rcp(new AggregationPhase1Algorithm            (graphFact)));
-      if (pL.get<bool>("aggregation: enable phase 2a")                 == true)   algos_.push_back(rcp(new AggregationPhase2aAlgorithm           (graphFact)));
-      if (pL.get<bool>("aggregation: enable phase 2b")                 == true)   algos_.push_back(rcp(new AggregationPhase2bAlgorithm           (graphFact)));
-      if (pL.get<bool>("aggregation: enable phase 3" )                 == true)   algos_.push_back(rcp(new AggregationPhase3Algorithm            (graphFact)));
-
-      *out << " Build interface aggregates" << std::endl;
-      // interface
-      if (pL.get<bool>("aggregation: use interface aggregation") == true) {
-        BuildInterfaceAggregates(currentLevel, aggregates, aggStat, numNonAggregatedNodes,
-                                 coarseRate);
-      }
+    TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements() != static_cast<size_t>(geoData->getNumLocalFineNodes()),
+                               Exceptions::RuntimeError,
+                               "The local number of elements in the graph's map is not equal to "
+                               "the number of nodes given by: lNodesPerDim!");
+
+    aggregates->SetIndexManager(geoData);
+    aggregates->SetNumAggregates(geoData->getNumLocalCoarseNodes());
+
+    Set(currentLevel, "lCoarseNodesPerDim", geoData->getLocalCoarseNodesPerDir());
+
+  }  // end structured aggregation setup
+
+  if (regionType == "uncoupled") {
+    // Add unstructred aggregation phases
+    algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm(graphFact)));
+    if (pL.get<bool>("aggregation: use interface aggregation") == true) algos_.push_back(rcp(new InterfaceAggregationAlgorithm(graphFact)));
+    if (pL.get<bool>("aggregation: allow user-specified singletons") == true) algos_.push_back(rcp(new OnePtAggregationAlgorithm(graphFact)));
+    if (pL.get<bool>("aggregation: enable phase 1") == true) algos_.push_back(rcp(new AggregationPhase1Algorithm(graphFact)));
+    if (pL.get<bool>("aggregation: enable phase 2a") == true) algos_.push_back(rcp(new AggregationPhase2aAlgorithm(graphFact)));
+    if (pL.get<bool>("aggregation: enable phase 2b") == true) algos_.push_back(rcp(new AggregationPhase2bAlgorithm(graphFact)));
+    if (pL.get<bool>("aggregation: enable phase 3") == true) algos_.push_back(rcp(new AggregationPhase3Algorithm(graphFact)));
+
+    *out << " Build interface aggregates" << std::endl;
+    // interface
+    if (pL.get<bool>("aggregation: use interface aggregation") == true) {
+      BuildInterfaceAggregates(currentLevel, aggregates, aggStat, numNonAggregatedNodes,
+                               coarseRate);
+    }
 
-      *out << "Treat Dirichlet BC" << std::endl;
-      // Dirichlet boundary
-      ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
-      if (dirichletBoundaryMap != Teuchos::null)
-        for (LO i = 0; i < numRows; i++)
-          if (dirichletBoundaryMap[i] == true)
-            aggStat[i] = BOUNDARY;
-
-      // OnePt aggregation
-      std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-      RCP<Map> OnePtMap = Teuchos::null;
-      if (mapOnePtName.length()) {
-        std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-        if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-          OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, NoFactory::get());
-        } else {
-          RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-          OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, mapOnePtFact.get());
-        }
-      }
+    *out << "Treat Dirichlet BC" << std::endl;
+    // Dirichlet boundary
+    ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
+    if (dirichletBoundaryMap != Teuchos::null)
+      for (LO i = 0; i < numRows; i++)
+        if (dirichletBoundaryMap[i] == true)
+          aggStat[i] = BOUNDARY;
 
-      LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
-      GO indexBase = graph->GetDomainMap()->getIndexBase();
-      if (OnePtMap != Teuchos::null) {
-        for (LO i = 0; i < numRows; i++) {
-          // reconstruct global row id (FIXME only works for contiguous maps)
-          GO grid = (graph->GetDomainMap()->getGlobalElement(i)-indexBase) * nDofsPerNode + indexBase;
-          for (LO kr = 0; kr < nDofsPerNode; kr++)
-            if (OnePtMap->isNodeGlobalElement(grid + kr))
-              aggStat[i] = ONEPT;
-        }
+    // OnePt aggregation
+    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+    RCP<Map> OnePtMap        = Teuchos::null;
+    if (mapOnePtName.length()) {
+      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+        OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, NoFactory::get());
+      } else {
+        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+        OnePtMap                            = currentLevel.Get<RCP<Map> >(mapOnePtName, mapOnePtFact.get());
       }
+    }
 
-      // Create a fake lCoarseNodesPerDir for CoordinatesTranferFactory
-      Array<LO> lCoarseNodesPerDir(3,-1);
-      Set(currentLevel, "lCoarseNodesPerDim", lCoarseNodesPerDir);
-    } // end uncoupled aggregation setup
-
-    aggregates->AggregatesCrossProcessors(false); // No coupled aggregation
+    LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
+    GO indexBase    = graph->GetDomainMap()->getIndexBase();
+    if (OnePtMap != Teuchos::null) {
+      for (LO i = 0; i < numRows; i++) {
+        // reconstruct global row id (FIXME only works for contiguous maps)
+        GO grid = (graph->GetDomainMap()->getGlobalElement(i) - indexBase) * nDofsPerNode + indexBase;
+        for (LO kr = 0; kr < nDofsPerNode; kr++)
+          if (OnePtMap->isNodeGlobalElement(grid + kr))
+            aggStat[i] = ONEPT;
+      }
+    }
 
-    *out << "Run all the algorithms on the local rank" << std::endl;
-    for (size_t a = 0; a < algos_.size(); a++) {
-      std::string phase = algos_[a]->description();
-      SubFactoryMonitor sfm(*this, "Algo \"" + phase + "\"", currentLevel);
-      *out << regionType <<" | Executing phase " << a << std::endl;
+    // Create a fake lCoarseNodesPerDir for CoordinatesTranferFactory
+    Array<LO> lCoarseNodesPerDir(3, -1);
+    Set(currentLevel, "lCoarseNodesPerDim", lCoarseNodesPerDir);
+  }  // end uncoupled aggregation setup
 
-      int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
-      algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-      algos_[a]->SetProcRankVerbose(oldRank);
-      *out << regionType <<" | Done Executing phase " << a << std::endl;
-    }
+  aggregates->AggregatesCrossProcessors(false);  // No coupled aggregation
 
-    *out << "Compute statistics on aggregates" << std::endl;
-    aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
+  *out << "Run all the algorithms on the local rank" << std::endl;
+  for (size_t a = 0; a < algos_.size(); a++) {
+    std::string phase = algos_[a]->description();
+    SubFactoryMonitor sfm(*this, "Algo \"" + phase + "\"", currentLevel);
+    *out << regionType << " | Executing phase " << a << std::endl;
 
-    Set(currentLevel, "Aggregates",                  aggregates);
-    Set(currentLevel, "numDimensions",               numDimensions);
-    Set(currentLevel, "aggregationRegionTypeCoarse", regionType);
+    int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
+    algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+    algos_[a]->SetProcRankVerbose(oldRank);
+    *out << regionType << " | Done Executing phase " << a << std::endl;
+  }
 
-    GetOStream(Statistics1) << aggregates->description() << std::endl;
-    *out << "HybridAggregation done!" << std::endl;
+  *out << "Compute statistics on aggregates" << std::endl;
+  aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
+
+  Set(currentLevel, "Aggregates", aggregates);
+  Set(currentLevel, "numDimensions", numDimensions);
+  Set(currentLevel, "aggregationRegionTypeCoarse", regionType);
+
+  GetOStream(Statistics1) << aggregates->description() << std::endl;
+  *out << "HybridAggregation done!" << std::endl;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildInterfaceAggregates(Level& currentLevel, RCP<Aggregates> aggregates,
+                             std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes,
+                             Array<LO> coarseRate) const {
+  FactoryMonitor m(*this, "BuildInterfaceAggregates", currentLevel);
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_HYBRIDAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildInterfaceAggregates(Level& currentLevel, RCP<Aggregates> aggregates,
-                           std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes,
-                           Array<LO> coarseRate) const {
-    FactoryMonitor m(*this, "BuildInterfaceAggregates", currentLevel);
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_HYBRIDAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  // Extract and format input data for algo
+  if (coarseRate.size() == 1) {
+    coarseRate.resize(3, coarseRate[0]);
+  }
+  ArrayRCP<LO> vertex2AggId      = aggregates->GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner        = aggregates->GetProcWinner()->getDataNonConst(0);
+  Array<LO> interfacesDimensions = Get<Array<LO> >(currentLevel, "interfacesDimensions");
+  Array<LO> nodesOnInterfaces    = Get<Array<LO> >(currentLevel, "nodeOnInterface");
+  const int numInterfaces        = interfacesDimensions.size() / 3;
+  const int myRank               = aggregates->GetMap()->getComm()->getRank();
+
+  // Create coarse level container to gather data on the fly
+  Array<LO> coarseInterfacesDimensions(interfacesDimensions.size());
+  Array<LO> nodesOnCoarseInterfaces;
+  {  // Scoping the temporary variables...
+    LO endRate, totalNumCoarseNodes = 0, numCoarseNodes;
+    for (int interfaceIdx = 0; interfaceIdx < numInterfaces; ++interfaceIdx) {
+      numCoarseNodes = 1;
+      for (int dim = 0; dim < 3; ++dim) {
+        endRate = (interfacesDimensions[3 * interfaceIdx + dim] - 1) % coarseRate[dim];
+        if (interfacesDimensions[3 * interfaceIdx + dim] == 1) {
+          coarseInterfacesDimensions[3 * interfaceIdx + dim] = 1;
+        } else {
+          coarseInterfacesDimensions[3 * interfaceIdx + dim] = (interfacesDimensions[3 * interfaceIdx + dim] - 1) / coarseRate[dim] + 2;
+          if (endRate == 0) {
+            coarseInterfacesDimensions[3 * interfaceIdx + dim]--;
+          }
+        }
+        numCoarseNodes *= coarseInterfacesDimensions[3 * interfaceIdx + dim];
+      }
+      totalNumCoarseNodes += numCoarseNodes;
     }
+    nodesOnCoarseInterfaces.resize(totalNumCoarseNodes, -1);
+  }
 
-    // Extract and format input data for algo
-    if(coarseRate.size() == 1) {coarseRate.resize(3, coarseRate[0]);}
-    ArrayRCP<LO> vertex2AggId      = aggregates->GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner        = aggregates->GetProcWinner()  ->getDataNonConst(0);
-    Array<LO> interfacesDimensions = Get<Array<LO> >(currentLevel, "interfacesDimensions");
-    Array<LO> nodesOnInterfaces    = Get<Array<LO> >(currentLevel, "nodeOnInterface");
-    const int numInterfaces        = interfacesDimensions.size() / 3;
-    const int myRank               = aggregates->GetMap()->getComm()->getRank();
-
-    // Create coarse level container to gather data on the fly
-    Array<LO> coarseInterfacesDimensions(interfacesDimensions.size());
-    Array<LO> nodesOnCoarseInterfaces;
-    { // Scoping the temporary variables...
-      LO endRate, totalNumCoarseNodes = 0, numCoarseNodes;
-      for(int interfaceIdx = 0; interfaceIdx < numInterfaces; ++interfaceIdx) {
-        numCoarseNodes = 1;
-        for(int dim = 0; dim < 3; ++dim) {
-          endRate = (interfacesDimensions[3*interfaceIdx + dim] - 1) % coarseRate[dim];
-          if(interfacesDimensions[3*interfaceIdx + dim] == 1) {
-            coarseInterfacesDimensions[3*interfaceIdx + dim] = 1;
-          } else {
-            coarseInterfacesDimensions[3*interfaceIdx + dim]
-              = (interfacesDimensions[3*interfaceIdx+dim]-1) / coarseRate[dim] + 2;
-            if(endRate==0){ coarseInterfacesDimensions[3*interfaceIdx + dim]--;}
-          }
-          numCoarseNodes *= coarseInterfacesDimensions[3*interfaceIdx + dim];
+  Array<LO> endRate(3);
+  LO interfaceOffset = 0, aggregateCount = 0, coarseNodeCount = 0;
+  for (int interfaceIdx = 0; interfaceIdx < numInterfaces; ++interfaceIdx) {
+    ArrayView<LO> fineNodesPerDim   = interfacesDimensions(3 * interfaceIdx, 3);
+    ArrayView<LO> coarseNodesPerDim = coarseInterfacesDimensions(3 * interfaceIdx, 3);
+    LO numInterfaceNodes = 1, numCoarseNodes = 1;
+    for (int dim = 0; dim < 3; ++dim) {
+      numInterfaceNodes *= fineNodesPerDim[dim];
+      numCoarseNodes *= coarseNodesPerDim[dim];
+      endRate[dim] = (fineNodesPerDim[dim] - 1) % coarseRate[dim];
+    }
+    ArrayView<LO> interfaceNodes = nodesOnInterfaces(interfaceOffset, numInterfaceNodes);
+
+    interfaceOffset += numInterfaceNodes;
+
+    LO rem, rate, fineNodeIdx;
+    Array<LO> nodeIJK(3), coarseIJK(3), rootIJK(3);
+    // First find treat coarse nodes as they generate the aggregate IDs
+    // and they might be repeated on multiple interfaces (think corners and edges).
+    for (LO coarseNodeIdx = 0; coarseNodeIdx < numCoarseNodes; ++coarseNodeIdx) {
+      coarseIJK[2] = coarseNodeIdx / (coarseNodesPerDim[0] * coarseNodesPerDim[1]);
+      rem          = coarseNodeIdx % (coarseNodesPerDim[0] * coarseNodesPerDim[1]);
+      coarseIJK[1] = rem / coarseNodesPerDim[0];
+      coarseIJK[0] = rem % coarseNodesPerDim[0];
+
+      for (LO dim = 0; dim < 3; ++dim) {
+        if (coarseIJK[dim] == coarseNodesPerDim[dim] - 1) {
+          nodeIJK[dim] = fineNodesPerDim[dim] - 1;
+        } else {
+          nodeIJK[dim] = coarseIJK[dim] * coarseRate[dim];
         }
-        totalNumCoarseNodes += numCoarseNodes;
       }
-      nodesOnCoarseInterfaces.resize(totalNumCoarseNodes, -1);
+      fineNodeIdx = (nodeIJK[2] * fineNodesPerDim[1] + nodeIJK[1]) * fineNodesPerDim[0] + nodeIJK[0];
+
+      if (aggStat[interfaceNodes[fineNodeIdx]] == READY) {
+        vertex2AggId[interfaceNodes[fineNodeIdx]] = aggregateCount;
+        procWinner[interfaceNodes[fineNodeIdx]]   = myRank;
+        aggStat[interfaceNodes[fineNodeIdx]]      = AGGREGATED;
+        ++aggregateCount;
+        --numNonAggregatedNodes;
+      }
+      nodesOnCoarseInterfaces[coarseNodeCount] = vertex2AggId[interfaceNodes[fineNodeIdx]];
+      ++coarseNodeCount;
     }
 
-    Array<LO> endRate(3);
-    LO interfaceOffset = 0, aggregateCount = 0, coarseNodeCount = 0;
-    for(int interfaceIdx = 0; interfaceIdx < numInterfaces; ++interfaceIdx) {
-      ArrayView<LO> fineNodesPerDim   = interfacesDimensions(3*interfaceIdx, 3);
-      ArrayView<LO> coarseNodesPerDim = coarseInterfacesDimensions(3*interfaceIdx, 3);
-      LO numInterfaceNodes = 1, numCoarseNodes = 1;
-      for(int dim = 0; dim < 3; ++dim) {
-        numInterfaceNodes *= fineNodesPerDim[dim];
-        numCoarseNodes    *= coarseNodesPerDim[dim];
-        endRate[dim]       = (fineNodesPerDim[dim]-1) % coarseRate[dim];
+    // Now loop over all the node on the interface
+    // skip the coarse nodes as they are already aggregated
+    // and find the appropriate aggregate ID for the fine nodes.
+    for (LO nodeIdx = 0; nodeIdx < numInterfaceNodes; ++nodeIdx) {
+      // If the node is already aggregated skip it!
+      if (aggStat[interfaceNodes[nodeIdx]] == AGGREGATED) {
+        continue;
       }
-      ArrayView<LO> interfaceNodes = nodesOnInterfaces(interfaceOffset, numInterfaceNodes);
-
-      interfaceOffset += numInterfaceNodes;
-
-      LO rem, rate, fineNodeIdx;
-      Array<LO> nodeIJK(3), coarseIJK(3), rootIJK(3);
-      // First find treat coarse nodes as they generate the aggregate IDs
-      // and they might be repeated on multiple interfaces (think corners and edges).
-      for(LO coarseNodeIdx = 0; coarseNodeIdx < numCoarseNodes; ++coarseNodeIdx) {
-        coarseIJK[2] = coarseNodeIdx / (coarseNodesPerDim[0]*coarseNodesPerDim[1]);
-        rem          = coarseNodeIdx % (coarseNodesPerDim[0]*coarseNodesPerDim[1]);
-        coarseIJK[1] = rem / coarseNodesPerDim[0];
-        coarseIJK[0] = rem % coarseNodesPerDim[0];
-
-        for(LO dim = 0; dim < 3; ++dim) {
-          if(coarseIJK[dim] == coarseNodesPerDim[dim] - 1) {
-            nodeIJK[dim] = fineNodesPerDim[dim] - 1;
-          } else {
-            nodeIJK[dim] = coarseIJK[dim]*coarseRate[dim];
-          }
+
+      nodeIJK[2] = nodeIdx / (fineNodesPerDim[0] * fineNodesPerDim[1]);
+      rem        = nodeIdx % (fineNodesPerDim[0] * fineNodesPerDim[1]);
+      nodeIJK[1] = rem / fineNodesPerDim[0];
+      nodeIJK[0] = rem % fineNodesPerDim[0];
+
+      for (int dim = 0; dim < 3; ++dim) {
+        coarseIJK[dim] = nodeIJK[dim] / coarseRate[dim];
+        rem            = nodeIJK[dim] % coarseRate[dim];
+        if (nodeIJK[dim] < fineNodesPerDim[dim] - endRate[dim]) {
+          rate = coarseRate[dim];
+        } else {
+          rate = endRate[dim];
         }
-        fineNodeIdx = (nodeIJK[2]*fineNodesPerDim[1] + nodeIJK[1])*fineNodesPerDim[0] + nodeIJK[0];
-
-        if(aggStat[interfaceNodes[fineNodeIdx]] == READY) {
-          vertex2AggId[interfaceNodes[fineNodeIdx]] = aggregateCount;
-          procWinner[interfaceNodes[fineNodeIdx]]   = myRank;
-          aggStat[interfaceNodes[fineNodeIdx]]      = AGGREGATED;
-          ++aggregateCount;
-          --numNonAggregatedNodes;
+        if (rem > (rate / 2)) {
+          ++coarseIJK[dim];
         }
-        nodesOnCoarseInterfaces[coarseNodeCount] = vertex2AggId[interfaceNodes[fineNodeIdx]];
-        ++coarseNodeCount;
       }
 
-      // Now loop over all the node on the interface
-      // skip the coarse nodes as they are already aggregated
-      // and find the appropriate aggregate ID for the fine nodes.
-      for(LO nodeIdx = 0; nodeIdx < numInterfaceNodes; ++nodeIdx) {
-
-        // If the node is already aggregated skip it!
-        if(aggStat[interfaceNodes[nodeIdx]] == AGGREGATED) {continue;}
-
-        nodeIJK[2] = nodeIdx / (fineNodesPerDim[0]*fineNodesPerDim[1]);
-        rem        = nodeIdx % (fineNodesPerDim[0]*fineNodesPerDim[1]);
-        nodeIJK[1] = rem / fineNodesPerDim[0];
-        nodeIJK[0] = rem % fineNodesPerDim[0];
-
-        for(int dim = 0; dim < 3; ++dim) {
-          coarseIJK[dim] = nodeIJK[dim] / coarseRate[dim];
-          rem            = nodeIJK[dim] % coarseRate[dim];
-          if(nodeIJK[dim] < fineNodesPerDim[dim] - endRate[dim]) {
-            rate = coarseRate[dim];
-          } else {
-            rate = endRate[dim];
-          }
-          if(rem > (rate / 2)) {++coarseIJK[dim];}
-        }
-
-        for(LO dim = 0; dim < 3; ++dim) {
-          if(coarseIJK[dim] == coarseNodesPerDim[dim] - 1) {
-            nodeIJK[dim] = fineNodesPerDim[dim] - 1;
-          } else {
-            nodeIJK[dim] = coarseIJK[dim]*coarseRate[dim];
-          }
+      for (LO dim = 0; dim < 3; ++dim) {
+        if (coarseIJK[dim] == coarseNodesPerDim[dim] - 1) {
+          nodeIJK[dim] = fineNodesPerDim[dim] - 1;
+        } else {
+          nodeIJK[dim] = coarseIJK[dim] * coarseRate[dim];
         }
-        fineNodeIdx = (nodeIJK[2]*fineNodesPerDim[1] + nodeIJK[1])*fineNodesPerDim[0] + nodeIJK[0];
-
-        vertex2AggId[interfaceNodes[nodeIdx]] = vertex2AggId[interfaceNodes[fineNodeIdx]];
-        procWinner[interfaceNodes[nodeIdx]]   = myRank;
-        aggStat[interfaceNodes[nodeIdx]]      = AGGREGATED;
-        --numNonAggregatedNodes;
-      } // Loop over interface nodes
-    } // Loop over the interfaces
+      }
+      fineNodeIdx = (nodeIJK[2] * fineNodesPerDim[1] + nodeIJK[1]) * fineNodesPerDim[0] + nodeIJK[0];
 
-    // Update aggregates information before subsequent aggregation algorithms
-    aggregates->SetNumAggregates(aggregateCount);
+      vertex2AggId[interfaceNodes[nodeIdx]] = vertex2AggId[interfaceNodes[fineNodeIdx]];
+      procWinner[interfaceNodes[nodeIdx]]   = myRank;
+      aggStat[interfaceNodes[nodeIdx]]      = AGGREGATED;
+      --numNonAggregatedNodes;
+    }  // Loop over interface nodes
+  }    // Loop over the interfaces
 
-    // Set coarse data for next level
-    Set(currentLevel, "coarseInterfacesDimensions", coarseInterfacesDimensions);
-    Set(currentLevel, "nodeOnCoarseInterface",      nodesOnCoarseInterfaces);
+  // Update aggregates information before subsequent aggregation algorithms
+  aggregates->SetNumAggregates(aggregateCount);
 
-  } // BuildInterfaceAggregates()
+  // Set coarse data for next level
+  Set(currentLevel, "coarseInterfacesDimensions", coarseInterfacesDimensions);
+  Set(currentLevel, "nodeOnCoarseInterface", nodesOnCoarseInterfaces);
 
-} //namespace MueLu
+}  // BuildInterfaceAggregates()
 
+}  // namespace MueLu
 
 #endif /* MUELU_HYBRIDAGGREGATIONFACTORY_DEF_HPP */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_decl.hpp
index 7cca1e45f86e..b6daf462bb42 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_decl.hpp
@@ -58,76 +58,74 @@
 
 namespace MueLu {
 
-  /*!
-    @class AmalgamationFactory
-    @brief AmalgamationFactory for subblocks of strided map based amalgamation data
+/*!
+  @class AmalgamationFactory
+  @brief AmalgamationFactory for subblocks of strided map based amalgamation data
 
-    Class generates unamalgamation information using matrix A with strided maps.
-    It stores the output information within an AmalgamationInfo object as "UnAmalgamationInfo".
-    This object contains
+  Class generates unamalgamation information using matrix A with strided maps.
+  It stores the output information within an AmalgamationInfo object as "UnAmalgamationInfo".
+  This object contains
 
-    \li \c nodegid2dofgids_ a map of all node ids of which the current proc has corresponding DOF gids (used by \c TentativePFactory).
-    \li \c gNodeIds vector of all node ids on the current proc (may be less than nodegid2dofgids_.size()). These nodes are stored on the current proc.
+  \li \c nodegid2dofgids_ a map of all node ids of which the current proc has corresponding DOF gids (used by \c TentativePFactory).
+  \li \c gNodeIds vector of all node ids on the current proc (may be less than nodegid2dofgids_.size()). These nodes are stored on the current proc.
 
-  */
+*/
 
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AmalgamationFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AmalgamationFactory : public SingleLevelFactoryBase {
 #undef MUELU_AMALGAMATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  AmalgamationFactory() = default;
 
-    //! Constructor
-    AmalgamationFactory() = default;
+  //! Destructor
+  virtual ~AmalgamationFactory() = default;
 
-    //! Destructor
-    virtual ~AmalgamationFactory() = default;
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const override;
 
-    void DeclareInput(Level &currentLevel) const override;
+  //@}
 
-    //@}
+  void Build(Level& currentLevel) const override;
 
-    void Build(Level &currentLevel) const override;
+  /*! @brief Translate global (row/column) id to global amalgamation block id
+   *
+   * @note Assume that the node map has the same \c indexBase as the dof map
+   *
+   * @param gid (GlobalOrdinal): input global id (row gid or column gid)
+   * @param blockSize (LocalOrdinal): block size (needed for constant block size)
+   * @param offset (GlobalOrdinal): global offset for dofs (stored in strided map, default = 0)
+   * @param indexBase (GlobalOrdinal): indexBase for DOF map (and node map, default = 0)
+   */
+  static const GlobalOrdinal DOFGid2NodeId(GlobalOrdinal gid, LocalOrdinal blockSize, const GlobalOrdinal offset /*= 0*/,
+                                           const GlobalOrdinal indexBase /* = 0*/);
 
-    /*! @brief Translate global (row/column) id to global amalgamation block id
-     *
-     * @note Assume that the node map has the same \c indexBase as the dof map
-     *
-     * @param gid (GlobalOrdinal): input global id (row gid or column gid)
-     * @param blockSize (LocalOrdinal): block size (needed for constant block size)
-     * @param offset (GlobalOrdinal): global offset for dofs (stored in strided map, default = 0)
-     * @param indexBase (GlobalOrdinal): indexBase for DOF map (and node map, default = 0)
-     */
-    static const GlobalOrdinal DOFGid2NodeId(GlobalOrdinal gid, LocalOrdinal blockSize, const GlobalOrdinal offset /*= 0*/,
-        const GlobalOrdinal indexBase/* = 0*/);
+  /*! @brief Method to create merged  map for systems of PDEs.
+   *
+   * @param sourceMap (const Map&): source map with dofs which shall be amalgamated to a node map
+   * @param A (const Matrix&): operator A (matrix) with striding information (if available)
+   * @param amalgamatedMap (const Map&): amalgamated node based map
+   * @param translation (Array<LO>&): array storing local node ids given local dof ids (needed in CoalesceDropFactory)
+   */
+  static void AmalgamateMap(const Map& sourceMap, const Matrix& A, RCP<const Map>& amalgamatedMap, Array<LO>& translation);
 
-    /*! @brief Method to create merged  map for systems of PDEs.
-     *
-     * @param sourceMap (const Map&): source map with dofs which shall be amalgamated to a node map
-     * @param A (const Matrix&): operator A (matrix) with striding information (if available)
-     * @param amalgamatedMap (const Map&): amalgamated node based map
-     * @param translation (Array<LO>&): array storing local node ids given local dof ids (needed in CoalesceDropFactory)
-     */
-    static void AmalgamateMap(const Map& sourceMap, const Matrix& A, RCP<const Map>& amalgamatedMap, Array<LO>& translation);
+};  // class AmalgamationFactory
 
-
-  }; //class AmalgamationFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AMALGAMATIONFACTORY_SHORT
-#endif // MUELU_AMALGAMATIONFACTORY_DECL_HPP
+#endif  // MUELU_AMALGAMATIONFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_def.hpp
index 60baf9efe0fe..b8dbfb162304 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationFactory_def.hpp
@@ -56,186 +56,182 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A"); // sub-block from blocked A
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const
-  {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-
-    /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
-       fullblocksize is the number of storage blocks that must kept together during the amalgamation process.
-
-       Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
-
-       numPDEs = fullblocksize * storageblocksize.
-       
-       If numPDEs==1
-         Matrix is point storage (classical CRS storage).  storageblocksize=1 and fullblocksize=1
-         No other values makes sense.
-
-       If numPDEs>1
-         If matrix uses point storage, then storageblocksize=1  and fullblockssize=numPDEs.
-         If matrix uses block storage, with block size of n, then storageblocksize=n, and fullblocksize=numPDEs/n.
-         Thus far, only storageblocksize=numPDEs and fullblocksize=1 has been tested.
-    */
-
-
-    LO fullblocksize    = 1;   // block dim for fixed size blocks
-    GO offset           = 0;   // global offset of dof gids
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    LO stridedblocksize = fullblocksize; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
-    LO storageblocksize = A->GetStorageBlockSize();
-    // GO indexBase        = A->getRowMap()->getIndexBase();  // index base for maps (unused)
-
-    // 1) check for blocking/striding information
-
-    if (A->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // NOTE: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      RCP<const StridedMap> stridedRowMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedRowMap == Teuchos::null,Exceptions::BadCast,"MueLu::CoalesceFactory::Build: cast to strided row map failed.");
-      fullblocksize = stridedRowMap->getFixedBlockSize();
-      offset        = stridedRowMap->getOffset();
-      blockid       = stridedRowMap->getStridedBlockId();
-
-      if (blockid > -1) {
-        std::vector<size_t> stridingInfo = stridedRowMap->getStridingData();
-        for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
-          nStridedOffset += stridingInfo[j];
-        stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
-
-      } else {
-        stridedblocksize = fullblocksize;
-      }
-      // Correct for the storageblocksize
-      // NOTE:  Before this point fullblocksize is actually numPDEs
-      TEUCHOS_TEST_FOR_EXCEPTION(fullblocksize % storageblocksize != 0,Exceptions::RuntimeError,"AmalgamationFactory: fullblocksize needs to be a multiple of A->GetStorageBlockSize()");
-      fullblocksize /= storageblocksize;
-      stridedblocksize /= storageblocksize;
-
-      oldView = A->SwitchToView(oldView);
-      GetOStream(Runtime1) << "AmalagamationFactory::Build():" << " found fullblocksize=" << fullblocksize << " and stridedblocksize=" << stridedblocksize << " from strided maps. offset=" << offset << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");  // sub-block from blocked A
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+
+  /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
+     fullblocksize is the number of storage blocks that must kept together during the amalgamation process.
+
+     Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
+
+     numPDEs = fullblocksize * storageblocksize.
+
+     If numPDEs==1
+       Matrix is point storage (classical CRS storage).  storageblocksize=1 and fullblocksize=1
+       No other values makes sense.
+
+     If numPDEs>1
+       If matrix uses point storage, then storageblocksize=1  and fullblockssize=numPDEs.
+       If matrix uses block storage, with block size of n, then storageblocksize=n, and fullblocksize=numPDEs/n.
+       Thus far, only storageblocksize=numPDEs and fullblocksize=1 has been tested.
+  */
+
+  LO fullblocksize    = 1;              // block dim for fixed size blocks
+  GO offset           = 0;              // global offset of dof gids
+  LO blockid          = -1;             // block id in strided map
+  LO nStridedOffset   = 0;              // DOF offset for strided block id "blockid" (default = 0)
+  LO stridedblocksize = fullblocksize;  // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+  LO storageblocksize = A->GetStorageBlockSize();
+  // GO indexBase        = A->getRowMap()->getIndexBase();  // index base for maps (unused)
+
+  // 1) check for blocking/striding information
+
+  if (A->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+    Xpetra::viewLabel_t oldView         = A->SwitchToView("stridedMaps");  // NOTE: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+    RCP<const StridedMap> stridedRowMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedRowMap == Teuchos::null, Exceptions::BadCast, "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
+    fullblocksize = stridedRowMap->getFixedBlockSize();
+    offset        = stridedRowMap->getOffset();
+    blockid       = stridedRowMap->getStridedBlockId();
+
+    if (blockid > -1) {
+      std::vector<size_t> stridingInfo = stridedRowMap->getStridingData();
+      for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
+        nStridedOffset += stridingInfo[j];
+      stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
 
     } else {
-      GetOStream(Warnings0) << "AmalagamationFactory::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
-    }
-
-
-    // build node row map (uniqueMap) and node column map (nonUniqueMap)
-    // the arrays rowTranslation and colTranslation contain the local node id
-    // given a local dof id. They are only necessary for the CoalesceDropFactory if
-    // fullblocksize > 1
-    RCP<const Map> uniqueMap, nonUniqueMap;
-    RCP<AmalgamationInfo> amalgamationData;
-    RCP<Array<LO> > rowTranslation = Teuchos::null;
-    RCP<Array<LO> > colTranslation = Teuchos::null;
-
-    if (fullblocksize > 1) {
-      // mfh 14 Apr 2015: These need to have different names than
-      // rowTranslation and colTranslation, in order to avoid
-      // shadowing warnings (-Wshadow with GCC).  Alternately, it
-      // looks like you could just assign to the existing variables in
-      // this scope, rather than creating new ones.
-      RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
-      RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
-      AmalgamateMap(*(A->getRowMap()), *A, uniqueMap,    *theRowTranslation);
-      AmalgamateMap(*(A->getColMap()), *A, nonUniqueMap, *theColTranslation);
-
-      amalgamationData = rcp(new AmalgamationInfo(theRowTranslation,
-                                                  theColTranslation,
-                                                  uniqueMap,
-                                                  nonUniqueMap,
-                                                  A->getColMap(),
-                                                  fullblocksize,
-                                                  offset,
-                                                  blockid,
-                                                  nStridedOffset,
-                                                  stridedblocksize) );
-    } else {
-      amalgamationData = rcp(new AmalgamationInfo(rowTranslation, // Teuchos::null
-                                                  colTranslation, // Teuchos::null
-                                                  A->getRowMap(), // unique map of graph
-                                                  A->getColMap(), // non-unique map of graph
-                                                  A->getColMap(), // column map of A
-                                                  fullblocksize,
-                                                  offset,
-                                                  blockid,
-                                                  nStridedOffset,
-                                                  stridedblocksize) );
+      stridedblocksize = fullblocksize;
     }
+    // Correct for the storageblocksize
+    // NOTE:  Before this point fullblocksize is actually numPDEs
+    TEUCHOS_TEST_FOR_EXCEPTION(fullblocksize % storageblocksize != 0, Exceptions::RuntimeError, "AmalgamationFactory: fullblocksize needs to be a multiple of A->GetStorageBlockSize()");
+    fullblocksize /= storageblocksize;
+    stridedblocksize /= storageblocksize;
+
+    oldView = A->SwitchToView(oldView);
+    GetOStream(Runtime1) << "AmalagamationFactory::Build():"
+                         << " found fullblocksize=" << fullblocksize << " and stridedblocksize=" << stridedblocksize << " from strided maps. offset=" << offset << std::endl;
+
+  } else {
+    GetOStream(Warnings0) << "AmalagamationFactory::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
+  }
 
-    // store (un)amalgamation information on current level
-    Set(currentLevel, "UnAmalgamationInfo", amalgamationData);
+  // build node row map (uniqueMap) and node column map (nonUniqueMap)
+  // the arrays rowTranslation and colTranslation contain the local node id
+  // given a local dof id. They are only necessary for the CoalesceDropFactory if
+  // fullblocksize > 1
+  RCP<const Map> uniqueMap, nonUniqueMap;
+  RCP<AmalgamationInfo> amalgamationData;
+  RCP<Array<LO> > rowTranslation = Teuchos::null;
+  RCP<Array<LO> > colTranslation = Teuchos::null;
+
+  if (fullblocksize > 1) {
+    // mfh 14 Apr 2015: These need to have different names than
+    // rowTranslation and colTranslation, in order to avoid
+    // shadowing warnings (-Wshadow with GCC).  Alternately, it
+    // looks like you could just assign to the existing variables in
+    // this scope, rather than creating new ones.
+    RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
+    RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
+    AmalgamateMap(*(A->getRowMap()), *A, uniqueMap, *theRowTranslation);
+    AmalgamateMap(*(A->getColMap()), *A, nonUniqueMap, *theColTranslation);
+
+    amalgamationData = rcp(new AmalgamationInfo(theRowTranslation,
+                                                theColTranslation,
+                                                uniqueMap,
+                                                nonUniqueMap,
+                                                A->getColMap(),
+                                                fullblocksize,
+                                                offset,
+                                                blockid,
+                                                nStridedOffset,
+                                                stridedblocksize));
+  } else {
+    amalgamationData = rcp(new AmalgamationInfo(rowTranslation,  // Teuchos::null
+                                                colTranslation,  // Teuchos::null
+                                                A->getRowMap(),  // unique map of graph
+                                                A->getColMap(),  // non-unique map of graph
+                                                A->getColMap(),  // column map of A
+                                                fullblocksize,
+                                                offset,
+                                                blockid,
+                                                nStridedOffset,
+                                                stridedblocksize));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AmalgamateMap(const Map& sourceMap, const Matrix& A, RCP<const Map>& amalgamatedMap, Array<LO>& translation) {
-    typedef typename ArrayView<const GO>::size_type size_type;
-    typedef std::unordered_map<GO,size_type> container;
-
-    GO                      indexBase = sourceMap.getIndexBase();
-    ArrayView<const GO>     elementAList = sourceMap.getLocalElementList();
-    size_type               numElements  = elementAList.size();
-    container               filter;
-
-    GO offset = 0;
-    LO blkSize = A.GetFixedBlockSize() / A.GetStorageBlockSize();
-    if (A.IsView("stridedMaps") == true) {
-      Teuchos::RCP<const Map> myMap = A.getRowMap("stridedMaps");
-      Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-      offset = strMap->getOffset();
-      blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
-    }
+  // store (un)amalgamation information on current level
+  Set(currentLevel, "UnAmalgamationInfo", amalgamationData);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AmalgamateMap(const Map& sourceMap, const Matrix& A, RCP<const Map>& amalgamatedMap, Array<LO>& translation) {
+  typedef typename ArrayView<const GO>::size_type size_type;
+  typedef std::unordered_map<GO, size_type> container;
+
+  GO indexBase                     = sourceMap.getIndexBase();
+  ArrayView<const GO> elementAList = sourceMap.getLocalElementList();
+  size_type numElements            = elementAList.size();
+  container filter;
+
+  GO offset  = 0;
+  LO blkSize = A.GetFixedBlockSize() / A.GetStorageBlockSize();
+  if (A.IsView("stridedMaps") == true) {
+    Teuchos::RCP<const Map> myMap         = A.getRowMap("stridedMaps");
+    Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+    offset  = strMap->getOffset();
+    blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
+  }
 
-    Array<GO> elementList(numElements);
-    translation.resize(numElements);
+  Array<GO> elementList(numElements);
+  translation.resize(numElements);
 
-    size_type numRows = 0;
-    for (size_type id = 0; id < numElements; id++) {
-      GO dofID  = elementAList[id];
-      GO nodeID = AmalgamationFactory::DOFGid2NodeId(dofID, blkSize, offset, indexBase);
+  size_type numRows = 0;
+  for (size_type id = 0; id < numElements; id++) {
+    GO dofID  = elementAList[id];
+    GO nodeID = AmalgamationFactory::DOFGid2NodeId(dofID, blkSize, offset, indexBase);
 
-      typename container::iterator it = filter.find(nodeID);
-      if (it == filter.end()) {
-        filter[nodeID] = numRows;
+    typename container::iterator it = filter.find(nodeID);
+    if (it == filter.end()) {
+      filter[nodeID] = numRows;
 
-        translation[id]      = numRows;
-        elementList[numRows] = nodeID;
+      translation[id]      = numRows;
+      elementList[numRows] = nodeID;
 
-        numRows++;
+      numRows++;
 
-      } else {
-        translation[id]      = it->second;
-      }
+    } else {
+      translation[id] = it->second;
     }
-    elementList.resize(numRows);
-
-    amalgamatedMap = MapFactory::Build(sourceMap.lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, sourceMap.getComm());
-
   }
+  elementList.resize(numRows);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  const GlobalOrdinal AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DOFGid2NodeId(GlobalOrdinal gid, LocalOrdinal blockSize,
-      const GlobalOrdinal offset, const GlobalOrdinal indexBase) {
-    GlobalOrdinal globalblockid = ((GlobalOrdinal) gid - offset - indexBase) / blockSize + indexBase;
-    return globalblockid;
-  }
+  amalgamatedMap = MapFactory::Build(sourceMap.lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, sourceMap.getComm());
+}
 
-} //namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const GlobalOrdinal AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DOFGid2NodeId(GlobalOrdinal gid, LocalOrdinal blockSize,
+                                                                                                  const GlobalOrdinal offset, const GlobalOrdinal indexBase) {
+  GlobalOrdinal globalblockid = ((GlobalOrdinal)gid - offset - indexBase) / blockSize + indexBase;
+  return globalblockid;
+}
 
-#endif /* MUELU_SUBBLOCKUNAMALGAMATIONFACTORY_DEF_HPP */
+}  // namespace MueLu
 
+#endif /* MUELU_SUBBLOCKUNAMALGAMATIONFACTORY_DEF_HPP */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_decl.hpp
index b14a14a1cc10..0d15843f3049 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_decl.hpp
@@ -53,7 +53,7 @@
 #ifndef MUELU_AMALGAMATIONINFO_DECL_HPP_
 #define MUELU_AMALGAMATIONINFO_DECL_HPP_
 
-#include <Xpetra_ConfigDefs.hpp>   // global_size_t
+#include <Xpetra_ConfigDefs.hpp>  // global_size_t
 #include <Xpetra_Map_fwd.hpp>
 #include <Xpetra_Vector_fwd.hpp>
 #include <Xpetra_MapFactory_fwd.hpp>
@@ -75,164 +75,158 @@ namespace MueLu {
   current processor.  That mapping is used for unamalgamation.
 */
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AmalgamationInfo
-    : public BaseClass {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AmalgamationInfo
+  : public BaseClass {
 #undef MUELU_AMALGAMATIONINFO_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-
-    /// Constructor
-    AmalgamationInfo(RCP<Array<LO> > rowTranslation,
-                     RCP<Array<LO> > colTranslation,
-                     RCP<const Map> nodeRowMap,
-                     RCP<const Map> nodeColMap,
-                     RCP<const Map> const &columnMap,
-                     LO fullblocksize, GO offset, LO blockid, LO nStridedOffset, LO stridedblocksize) :
-                     rowTranslation_(rowTranslation),
-                     colTranslation_(colTranslation),
-                     nodeRowMap_(nodeRowMap),
-                     nodeColMap_(nodeColMap),
-                     columnMap_(columnMap),
-                     fullblocksize_(fullblocksize),
-                     offset_(offset),
-                     blockid_(blockid),
-                     nStridedOffset_(nStridedOffset),
-                     stridedblocksize_(stridedblocksize),
-                     indexBase_(columnMap->getIndexBase())
-    {}
-
-    /// Destructor
-    virtual ~AmalgamationInfo() {}
-
-    /// Return a simple one-line description of this object.
-    std::string description() const { return "AmalgamationInfo"; }
-
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
-
-    RCP<const Map> getNodeRowMap() const { return nodeRowMap_; } //! < returns the node row map for the graph
-    RCP<const Map> getNodeColMap() const { return nodeColMap_; } //! < returns the node column map for the graph
-
-    /* @brief Translation arrays
-     *
-     * Returns translation arrays providing local node ids given local dof ids built from either
-     * the non-overlapping (unique) row map or the overlapping (non-unique) column map.
-     * The getColTranslation routine, e.g., is used for the MergeRows routine in CoalesceDropFactory.
-     */
-    //@{
-    RCP<Array<LO> > getRowTranslation() const { return rowTranslation_; }
-    RCP<Array<LO> > getColTranslation() const { return colTranslation_; }
-    //@}
-
-    /*! @brief UnamalgamateAggregates
-
-      Puts all dofs for aggregate \c i in aggToRowMap[\c i].  Also calculate aggregate sizes.
-    */
-    void UnamalgamateAggregates(const Aggregates& aggregates, Teuchos::ArrayRCP<LocalOrdinal>& aggStart, Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const;
-    void UnamalgamateAggregatesLO(const Aggregates& aggregates, Teuchos::ArrayRCP<LocalOrdinal>& aggStart, Teuchos::ArrayRCP<LO>& aggToRowMap) const;
-
-    /*! @brief ComputeUnamalgamatedImportDofMap
-     * build overlapping dof row map from aggregates needed for overlapping null space
-     */
-    Teuchos::RCP< Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > ComputeUnamalgamatedImportDofMap(const Aggregates& aggregates) const;
-
-  private:
-
-    void UnamalgamateAggregates(const Teuchos::RCP<const Map> &nodeMap,
-                                const RCP<LOVector> &procWinnerVec,
-                                const RCP<LOMultiVector> &vertex2AggIdVec,
+ public:
+  /// Constructor
+  AmalgamationInfo(RCP<Array<LO> > rowTranslation,
+                   RCP<Array<LO> > colTranslation,
+                   RCP<const Map> nodeRowMap,
+                   RCP<const Map> nodeColMap,
+                   RCP<const Map> const& columnMap,
+                   LO fullblocksize, GO offset, LO blockid, LO nStridedOffset, LO stridedblocksize)
+    : rowTranslation_(rowTranslation)
+    , colTranslation_(colTranslation)
+    , nodeRowMap_(nodeRowMap)
+    , nodeColMap_(nodeColMap)
+    , columnMap_(columnMap)
+    , fullblocksize_(fullblocksize)
+    , offset_(offset)
+    , blockid_(blockid)
+    , nStridedOffset_(nStridedOffset)
+    , stridedblocksize_(stridedblocksize)
+    , indexBase_(columnMap->getIndexBase()) {}
+
+  /// Destructor
+  virtual ~AmalgamationInfo() {}
+
+  /// Return a simple one-line description of this object.
+  std::string description() const { return "AmalgamationInfo"; }
+
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;;
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+
+  RCP<const Map> getNodeRowMap() const { return nodeRowMap_; }  //! < returns the node row map for the graph
+  RCP<const Map> getNodeColMap() const { return nodeColMap_; }  //! < returns the node column map for the graph
+
+  /* @brief Translation arrays
+   *
+   * Returns translation arrays providing local node ids given local dof ids built from either
+   * the non-overlapping (unique) row map or the overlapping (non-unique) column map.
+   * The getColTranslation routine, e.g., is used for the MergeRows routine in CoalesceDropFactory.
+   */
+  //@{
+  RCP<Array<LO> > getRowTranslation() const { return rowTranslation_; }
+  RCP<Array<LO> > getColTranslation() const { return colTranslation_; }
+  //@}
+
+  /*! @brief UnamalgamateAggregates
+
+    Puts all dofs for aggregate \c i in aggToRowMap[\c i].  Also calculate aggregate sizes.
+  */
+  void UnamalgamateAggregates(const Aggregates& aggregates, Teuchos::ArrayRCP<LocalOrdinal>& aggStart, Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const;
+  void UnamalgamateAggregatesLO(const Aggregates& aggregates, Teuchos::ArrayRCP<LocalOrdinal>& aggStart, Teuchos::ArrayRCP<LO>& aggToRowMap) const;
+
+  /*! @brief ComputeUnamalgamatedImportDofMap
+   * build overlapping dof row map from aggregates needed for overlapping null space
+   */
+  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > ComputeUnamalgamatedImportDofMap(const Aggregates& aggregates) const;
+
+ private:
+  void UnamalgamateAggregates(const Teuchos::RCP<const Map>& nodeMap,
+                              const RCP<LOVector>& procWinnerVec,
+                              const RCP<LOMultiVector>& vertex2AggIdVec,
+                              const GO numAggregates,
+                              Teuchos::ArrayRCP<LocalOrdinal>& aggStart,
+                              Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const;
+
+  void UnamalgamateAggregatesLO(const Teuchos::RCP<const Map>& nodeMap,
+                                const RCP<LOVector>& procWinnerVec,
+                                const RCP<LOMultiVector>& vertex2AggIdVec,
                                 const GO numAggregates,
                                 Teuchos::ArrayRCP<LocalOrdinal>& aggStart,
-                                Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const;
-
-    void UnamalgamateAggregatesLO(const Teuchos::RCP<const Map> &nodeMap,
-                                  const RCP<LOVector> &procWinnerVec,
-                                  const RCP<LOMultiVector> &vertex2AggIdVec,
-                                  const GO numAggregates,
-                                  Teuchos::ArrayRCP<LocalOrdinal>& aggStart,
-                                  Teuchos::ArrayRCP<LO>& aggToRowMap) const;
-
-    Teuchos::RCP< Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > ComputeUnamalgamatedImportDofMap(const Teuchos::RCP<const Map> &nodeMap) const;
-
-  public:
-
-    /*! @brief ComputeGlobalDOF
-     *
-     * Return global dof id associated with global node id gNodeID and dof index k
-     *
-     * \note We assume that \c indexBase_ is valid for both the node and the dof map.
-     *
-     * @param (GO): global node id
-     * @param (LO): local dof index within node
-     * @return (GO): global dof id
-     */
-    GO ComputeGlobalDOF(GO const &gNodeID, LO const &k=0) const;
-
-    /*! @brief ComputeLocalDOF
-     * return locbal dof id associated with local node id lNodeID and dof index k
-     *
-     * @param (LO): local node id
-     * @param (LO): local dof index within node
-     * @return (LO): local dof id
-     */
-    LO ComputeLocalDOF(LocalOrdinal const &lNodeID, LocalOrdinal const &k) const;
-
-    LO ComputeLocalNode(LocalOrdinal const &ldofID) const;
-
-    /*! Access routines */
-
-    /// returns offset of global dof ids
-    GO GlobalOffset() { return offset_; }
-
-    /// returns striding information
-    void GetStridingInformation(LO& fullBlockSize, LO& blockID, LO& stridingOffset, LO& stridedBlockSize, GO& indexBase) {
-      fullBlockSize = fullblocksize_;
-      blockID = blockid_;
-      stridingOffset = nStridedOffset_;
-      stridedBlockSize = stridedblocksize_;
-      indexBase = indexBase_;
-    }
-
-  private:
-
-    //! @name amalgamation information variables
-    //@{
-
-    //! Arrays containing local node ids given local dof ids
-    RCP<Array<LO> > rowTranslation_;
-    RCP<Array<LO> > colTranslation_;
-
-    //! node row and column map of graph (built from row and column map of A)
-    RCP<const Map> nodeRowMap_;
-    RCP<const Map> nodeColMap_;
-
-    /*! @brief DOF map (really column map of A)
-
-    We keep a RCP on the column map to make sure that the map is still valid when it is used.
-    */
-    RCP<const Map> columnMap_;
-
-    //@}
-
-    //! @name Strided map information.
-    //@{
-    LO fullblocksize_;
-    GO offset_;
-    LO blockid_;
-    LO nStridedOffset_;
-    LO stridedblocksize_;
-    GO indexBase_;
-    //@}
-
-  };
-
-} // namespace MueLu
+                                Teuchos::ArrayRCP<LO>& aggToRowMap) const;
+
+  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > ComputeUnamalgamatedImportDofMap(const Teuchos::RCP<const Map>& nodeMap) const;
+
+ public:
+  /*! @brief ComputeGlobalDOF
+   *
+   * Return global dof id associated with global node id gNodeID and dof index k
+   *
+   * \note We assume that \c indexBase_ is valid for both the node and the dof map.
+   *
+   * @param (GO): global node id
+   * @param (LO): local dof index within node
+   * @return (GO): global dof id
+   */
+  GO ComputeGlobalDOF(GO const& gNodeID, LO const& k = 0) const;
+
+  /*! @brief ComputeLocalDOF
+   * return locbal dof id associated with local node id lNodeID and dof index k
+   *
+   * @param (LO): local node id
+   * @param (LO): local dof index within node
+   * @return (LO): local dof id
+   */
+  LO ComputeLocalDOF(LocalOrdinal const& lNodeID, LocalOrdinal const& k) const;
+
+  LO ComputeLocalNode(LocalOrdinal const& ldofID) const;
+
+  /*! Access routines */
+
+  /// returns offset of global dof ids
+  GO GlobalOffset() { return offset_; }
+
+  /// returns striding information
+  void GetStridingInformation(LO& fullBlockSize, LO& blockID, LO& stridingOffset, LO& stridedBlockSize, GO& indexBase) {
+    fullBlockSize    = fullblocksize_;
+    blockID          = blockid_;
+    stridingOffset   = nStridedOffset_;
+    stridedBlockSize = stridedblocksize_;
+    indexBase        = indexBase_;
+  }
+
+ private:
+  //! @name amalgamation information variables
+  //@{
+
+  //! Arrays containing local node ids given local dof ids
+  RCP<Array<LO> > rowTranslation_;
+  RCP<Array<LO> > colTranslation_;
+
+  //! node row and column map of graph (built from row and column map of A)
+  RCP<const Map> nodeRowMap_;
+  RCP<const Map> nodeColMap_;
+
+  /*! @brief DOF map (really column map of A)
+
+  We keep a RCP on the column map to make sure that the map is still valid when it is used.
+  */
+  RCP<const Map> columnMap_;
+
+  //@}
+
+  //! @name Strided map information.
+  //@{
+  LO fullblocksize_;
+  GO offset_;
+  LO blockid_;
+  LO nStridedOffset_;
+  LO stridedblocksize_;
+  GO indexBase_;
+  //@}
+};
+
+}  // namespace MueLu
 
 #define MUELU_AMALGAMATIONINFO_SHORT
 #endif /* MUELU_AMALGAMATIONINFO_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_def.hpp
index f46746657142..28ba61ec70b3 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_AmalgamationInfo_def.hpp
@@ -63,265 +63,257 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  UnamalgamateAggregates(const Aggregates& aggregates,
-                         Teuchos::ArrayRCP<LocalOrdinal>& aggStart,
-                         Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const {
-
-    UnamalgamateAggregates(aggregates.GetMap(),
-                           aggregates.GetProcWinner(),
-                           aggregates.GetVertex2AggId(),
-                           aggregates.GetNumAggregates(),
-                           aggStart,
-                           aggToRowMap);
-
-  } //UnamalgamateAggregates
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  UnamalgamateAggregates(const Teuchos::RCP<const Map> &nodeMap,
-                         const RCP<LOVector> &procWinnerVec,
-                         const RCP<LOMultiVector> &vertex2AggIdVec,
-                         const GO numAggregates,
-                         Teuchos::ArrayRCP<LocalOrdinal>& aggStart,
-                         Teuchos::ArrayRCP<GlobalOrdinal>& aggToRowMap) const {
-
-    int myPid = nodeMap->getComm()->getRank();
-    Teuchos::ArrayView<const GO> nodeGlobalElts = nodeMap->getLocalElementList();
-    Teuchos::ArrayRCP<LO> procWinner   = procWinnerVec->getDataNonConst(0);
-    Teuchos::ArrayRCP<LO> vertex2AggId = vertex2AggIdVec->getDataNonConst(0);
-    const LO size = procWinner.size();
-
-    std::vector<LO> sizes(numAggregates);
-    if (stridedblocksize_ == 1) {
-      for (LO lnode = 0; lnode < size; ++lnode) {
-        LO myAgg = vertex2AggId[lnode];
-        if (procWinner[lnode] == myPid)
-          sizes[myAgg] += 1;
-      }
-    } else {
-      for (LO lnode = 0; lnode < size; ++lnode) {
-        LO myAgg = vertex2AggId[lnode];
-        if (procWinner[lnode] == myPid) {
-          GO gnodeid = nodeGlobalElts[lnode];
-          for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
-            GlobalOrdinal gDofIndex = ComputeGlobalDOF(gnodeid,k);
-            if (columnMap_->isNodeGlobalElement(gDofIndex))
-              sizes[myAgg] += 1;
-          }
-        }
-      }
-    }
-    aggStart = ArrayRCP<LO>(numAggregates+1,0);
-    aggStart[0] = Teuchos::ScalarTraits<LO>::zero();
-    for (GO i=0; i<numAggregates; ++i) {
-      aggStart[i+1] = aggStart[i] + sizes[i];
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    UnamalgamateAggregates(const Aggregates &aggregates,
+                           Teuchos::ArrayRCP<LocalOrdinal> &aggStart,
+                           Teuchos::ArrayRCP<GlobalOrdinal> &aggToRowMap) const {
+  UnamalgamateAggregates(aggregates.GetMap(),
+                         aggregates.GetProcWinner(),
+                         aggregates.GetVertex2AggId(),
+                         aggregates.GetNumAggregates(),
+                         aggStart,
+                         aggToRowMap);
+
+}  // UnamalgamateAggregates
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    UnamalgamateAggregates(const Teuchos::RCP<const Map> &nodeMap,
+                           const RCP<LOVector> &procWinnerVec,
+                           const RCP<LOMultiVector> &vertex2AggIdVec,
+                           const GO numAggregates,
+                           Teuchos::ArrayRCP<LocalOrdinal> &aggStart,
+                           Teuchos::ArrayRCP<GlobalOrdinal> &aggToRowMap) const {
+  int myPid                                   = nodeMap->getComm()->getRank();
+  Teuchos::ArrayView<const GO> nodeGlobalElts = nodeMap->getLocalElementList();
+  Teuchos::ArrayRCP<LO> procWinner            = procWinnerVec->getDataNonConst(0);
+  Teuchos::ArrayRCP<LO> vertex2AggId          = vertex2AggIdVec->getDataNonConst(0);
+  const LO size                               = procWinner.size();
+
+  std::vector<LO> sizes(numAggregates);
+  if (stridedblocksize_ == 1) {
+    for (LO lnode = 0; lnode < size; ++lnode) {
+      LO myAgg = vertex2AggId[lnode];
+      if (procWinner[lnode] == myPid)
+        sizes[myAgg] += 1;
     }
-    aggToRowMap = ArrayRCP<GO>(aggStart[numAggregates],0);
-
-    // count, how many dofs have been recorded for each aggregate so far
-    Array<LO> numDofs(numAggregates, 0); // empty array with number of Dofs for each aggregate
-
-    if (stridedblocksize_ == 1) {
-      for (LO lnode = 0; lnode < size; ++lnode) {
-        LO myAgg = vertex2AggId[lnode];
-        if (procWinner[lnode] == myPid) {
-          aggToRowMap[ aggStart[myAgg] + numDofs[myAgg] ] = ComputeGlobalDOF(nodeGlobalElts[lnode]);
-          ++(numDofs[myAgg]);
-        }
-      }
-    } else {
-      for (LO lnode = 0; lnode < size; ++lnode) {
-        LO myAgg = vertex2AggId[lnode];
-
-        if (procWinner[lnode] == myPid) {
-          GO gnodeid = nodeGlobalElts[lnode];
-          for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
-            GlobalOrdinal gDofIndex = ComputeGlobalDOF(gnodeid,k);
-            if (columnMap_->isNodeGlobalElement(gDofIndex)) {
-              aggToRowMap[ aggStart[myAgg] + numDofs[myAgg] ] = gDofIndex;
-              ++(numDofs[myAgg]);
-            }
-          }
+  } else {
+    for (LO lnode = 0; lnode < size; ++lnode) {
+      LO myAgg = vertex2AggId[lnode];
+      if (procWinner[lnode] == myPid) {
+        GO gnodeid = nodeGlobalElts[lnode];
+        for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
+          GlobalOrdinal gDofIndex = ComputeGlobalDOF(gnodeid, k);
+          if (columnMap_->isNodeGlobalElement(gDofIndex))
+            sizes[myAgg] += 1;
         }
       }
     }
-    // todo plausibility check: entry numDofs[k] == aggToRowMap[k].size()
-
-  } //UnamalgamateAggregates
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  UnamalgamateAggregatesLO(const Aggregates& aggregates,
-                           Teuchos::ArrayRCP<LO>& aggStart,
-                           Teuchos::ArrayRCP<LO>& aggToRowMap) const {
-    UnamalgamateAggregatesLO(aggregates.GetMap(),
-                             aggregates.GetProcWinner(),
-                             aggregates.GetVertex2AggId(),
-                             aggregates.GetNumAggregates(),
-                             aggStart,
-                             aggToRowMap);
   }
+  aggStart    = ArrayRCP<LO>(numAggregates + 1, 0);
+  aggStart[0] = Teuchos::ScalarTraits<LO>::zero();
+  for (GO i = 0; i < numAggregates; ++i) {
+    aggStart[i + 1] = aggStart[i] + sizes[i];
+  }
+  aggToRowMap = ArrayRCP<GO>(aggStart[numAggregates], 0);
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  UnamalgamateAggregatesLO(const Teuchos::RCP<const Map> &nodeMap,
-                           const RCP<LOVector> &procWinnerVec,
-                           const RCP<LOMultiVector> &vertex2AggIdVec,
-                           const GO numAggregates,
-                           Teuchos::ArrayRCP<LO>& aggStart,
-                           Teuchos::ArrayRCP<LO>& aggToRowMap) const {
-
-    int myPid = nodeMap->getComm()->getRank();
-    Teuchos::ArrayView<const GO> nodeGlobalElts = nodeMap->getLocalElementList();
-
-    Teuchos::ArrayRCP<LO> procWinner   = procWinnerVec  ->getDataNonConst(0);
-    Teuchos::ArrayRCP<LO> vertex2AggId = vertex2AggIdVec->getDataNonConst(0);
-
-
-    // FIXME: Do we need to compute size here? Or can we use existing?
-    const LO size = procWinner.size();
-
-    std::vector<LO> sizes(numAggregates);
-    if (stridedblocksize_ == 1) {
-      for (LO lnode = 0; lnode < size; lnode++)
-        if (procWinner[lnode] == myPid)
-          sizes[vertex2AggId[lnode]]++;
-    } else {
-      for (LO lnode = 0; lnode < size; lnode++)
-        if (procWinner[lnode] == myPid) {
-          GO nodeGID = nodeGlobalElts[lnode];
+  // count, how many dofs have been recorded for each aggregate so far
+  Array<LO> numDofs(numAggregates, 0);  // empty array with number of Dofs for each aggregate
 
-          for (LO k = 0; k < stridedblocksize_; k++) {
-            GO GID = ComputeGlobalDOF(nodeGID, k);
-            if (columnMap_->isNodeGlobalElement(GID))
-              sizes[vertex2AggId[lnode]]++;
-          }
-        }
+  if (stridedblocksize_ == 1) {
+    for (LO lnode = 0; lnode < size; ++lnode) {
+      LO myAgg = vertex2AggId[lnode];
+      if (procWinner[lnode] == myPid) {
+        aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = ComputeGlobalDOF(nodeGlobalElts[lnode]);
+        ++(numDofs[myAgg]);
+      }
     }
+  } else {
+    for (LO lnode = 0; lnode < size; ++lnode) {
+      LO myAgg = vertex2AggId[lnode];
 
-    aggStart = ArrayRCP<LO>(numAggregates+1); // FIXME: useless initialization with zeros
-    aggStart[0] = 0;
-    for (GO i = 0; i < numAggregates; i++)
-      aggStart[i+1] = aggStart[i] + sizes[i];
-
-    aggToRowMap = ArrayRCP<LO>(aggStart[numAggregates], 0);
-
-    // count, how many dofs have been recorded for each aggregate so far
-    Array<LO> numDofs(numAggregates, 0); // empty array with number of DOFs for each aggregate
-    if (stridedblocksize_ == 1) {
-      for (LO lnode = 0; lnode < size; ++lnode)
-        if (procWinner[lnode] == myPid) {
-          LO myAgg = vertex2AggId[lnode];
-          aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = lnode;
-          numDofs[myAgg]++;
-        }
-    } else {
-      for (LO lnode = 0; lnode < size; ++lnode)
-        if (procWinner[lnode] == myPid) {
-          LO myAgg = vertex2AggId[lnode];
-          GO nodeGID = nodeGlobalElts[lnode];
-
-          for (LO k = 0; k < stridedblocksize_; k++) {
-            GO GID = ComputeGlobalDOF(nodeGID, k);
-            if (columnMap_->isNodeGlobalElement(GID)) {
-              aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = lnode*stridedblocksize_ + k;
-              numDofs[myAgg]++;
-            }
+      if (procWinner[lnode] == myPid) {
+        GO gnodeid = nodeGlobalElts[lnode];
+        for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
+          GlobalOrdinal gDofIndex = ComputeGlobalDOF(gnodeid, k);
+          if (columnMap_->isNodeGlobalElement(gDofIndex)) {
+            aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = gDofIndex;
+            ++(numDofs[myAgg]);
           }
         }
+      }
     }
-    // todo plausibility check: entry numDofs[k] == aggToRowMap[k].size()
-
-  } //UnamalgamateAggregatesLO
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out,
-      const VerbLevel verbLevel) const
-  {
-    if (!(verbLevel & Debug))
-      return;
-
-    out << "AmalgamationInfo -- Striding information:"
-        << "\n  fullBlockSize = " << fullblocksize_
-        << "\n  blockID = " << blockid_
-        << "\n  stridingOffset = " << nStridedOffset_
-        << "\n  stridedBlockSize = " << stridedblocksize_
-        << "\n  indexBase = " << indexBase_
-        << std::endl;
-
-    out << "AmalgamationInfo -- DOFs to nodes mapping:\n"
-        << "  Mapping of row DOFs to row nodes:" << *rowTranslation_()
-        << "\n\n  Mapping of column DOFs to column nodes:" << *colTranslation_()
-        << std::endl;
-
-    out << "AmalgamationInfo -- row node map:" << std::endl;
-    nodeRowMap_->describe(out, Teuchos::VERB_EXTREME);
-
-    out << "AmalgamationInfo -- column node map:" << std::endl;
-    nodeColMap_->describe(out, Teuchos::VERB_EXTREME);
   }
+  // todo plausibility check: entry numDofs[k] == aggToRowMap[k].size()
 
-  /////////////////////////////////////////////////////////////////////////////
+}  // UnamalgamateAggregates
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeUnamalgamatedImportDofMap(const Aggregates& aggregates) const {
-    return ComputeUnamalgamatedImportDofMap(aggregates.GetMap());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    UnamalgamateAggregatesLO(const Aggregates &aggregates,
+                             Teuchos::ArrayRCP<LO> &aggStart,
+                             Teuchos::ArrayRCP<LO> &aggToRowMap) const {
+  UnamalgamateAggregatesLO(aggregates.GetMap(),
+                           aggregates.GetProcWinner(),
+                           aggregates.GetVertex2AggId(),
+                           aggregates.GetNumAggregates(),
+                           aggStart,
+                           aggToRowMap);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    UnamalgamateAggregatesLO(const Teuchos::RCP<const Map> &nodeMap,
+                             const RCP<LOVector> &procWinnerVec,
+                             const RCP<LOMultiVector> &vertex2AggIdVec,
+                             const GO numAggregates,
+                             Teuchos::ArrayRCP<LO> &aggStart,
+                             Teuchos::ArrayRCP<LO> &aggToRowMap) const {
+  int myPid                                   = nodeMap->getComm()->getRank();
+  Teuchos::ArrayView<const GO> nodeGlobalElts = nodeMap->getLocalElementList();
+
+  Teuchos::ArrayRCP<LO> procWinner   = procWinnerVec->getDataNonConst(0);
+  Teuchos::ArrayRCP<LO> vertex2AggId = vertex2AggIdVec->getDataNonConst(0);
+
+  // FIXME: Do we need to compute size here? Or can we use existing?
+  const LO size = procWinner.size();
+
+  std::vector<LO> sizes(numAggregates);
+  if (stridedblocksize_ == 1) {
+    for (LO lnode = 0; lnode < size; lnode++)
+      if (procWinner[lnode] == myPid)
+        sizes[vertex2AggId[lnode]]++;
+  } else {
+    for (LO lnode = 0; lnode < size; lnode++)
+      if (procWinner[lnode] == myPid) {
+        GO nodeGID = nodeGlobalElts[lnode];
+
+        for (LO k = 0; k < stridedblocksize_; k++) {
+          GO GID = ComputeGlobalDOF(nodeGID, k);
+          if (columnMap_->isNodeGlobalElement(GID))
+            sizes[vertex2AggId[lnode]]++;
+        }
+      }
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeUnamalgamatedImportDofMap(const Teuchos::RCP<const Map> &nodeMap) const {
-
-    Teuchos::RCP<std::vector<GO> > myDofGids = Teuchos::rcp(new std::vector<GO>);
-    Teuchos::ArrayView<const GO> gEltList = nodeMap->getLocalElementList();
-    LO nodeElements = Teuchos::as<LO>(nodeMap->getLocalNumElements());
-    if (stridedblocksize_ == 1) {
-      for (LO n = 0; n<nodeElements; n++) {
-        GlobalOrdinal gDofIndex = ComputeGlobalDOF(gEltList[n]);
-        myDofGids->push_back(gDofIndex);
+  aggStart    = ArrayRCP<LO>(numAggregates + 1);  // FIXME: useless initialization with zeros
+  aggStart[0] = 0;
+  for (GO i = 0; i < numAggregates; i++)
+    aggStart[i + 1] = aggStart[i] + sizes[i];
+
+  aggToRowMap = ArrayRCP<LO>(aggStart[numAggregates], 0);
+
+  // count, how many dofs have been recorded for each aggregate so far
+  Array<LO> numDofs(numAggregates, 0);  // empty array with number of DOFs for each aggregate
+  if (stridedblocksize_ == 1) {
+    for (LO lnode = 0; lnode < size; ++lnode)
+      if (procWinner[lnode] == myPid) {
+        LO myAgg                                      = vertex2AggId[lnode];
+        aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = lnode;
+        numDofs[myAgg]++;
       }
-    } else {
-      for (LO n = 0; n<nodeElements; n++) {
-        for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
-          GlobalOrdinal gDofIndex = ComputeGlobalDOF(gEltList[n],k);
-          if (columnMap_->isNodeGlobalElement(gDofIndex))
-            myDofGids->push_back(gDofIndex);
+  } else {
+    for (LO lnode = 0; lnode < size; ++lnode)
+      if (procWinner[lnode] == myPid) {
+        LO myAgg   = vertex2AggId[lnode];
+        GO nodeGID = nodeGlobalElts[lnode];
+
+        for (LO k = 0; k < stridedblocksize_; k++) {
+          GO GID = ComputeGlobalDOF(nodeGID, k);
+          if (columnMap_->isNodeGlobalElement(GID)) {
+            aggToRowMap[aggStart[myAgg] + numDofs[myAgg]] = lnode * stridedblocksize_ + k;
+            numDofs[myAgg]++;
+          }
         }
       }
-    }
-
-    Teuchos::ArrayRCP<GO> arr_myDofGids = Teuchos::arcp( myDofGids );
-    Teuchos::RCP<Map> importDofMap = MapFactory::Build(nodeMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), arr_myDofGids(), nodeMap->getIndexBase(), nodeMap->getComm());
-    return importDofMap;
-  }
-
-  /////////////////////////////////////////////////////////////////////////////
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeGlobalDOF(GlobalOrdinal const &gNodeID, LocalOrdinal const &k) const {
-    // here, the assumption is, that the node map has the same indexBase as the dof map
-    //                            this is the node map index base                    this is the dof map index base
-    GlobalOrdinal gDofIndex = offset_ + (gNodeID-indexBase_)*fullblocksize_ + nStridedOffset_ + k + indexBase_;
-    return gDofIndex;
   }
-
- template <class LocalOrdinal, class GlobalOrdinal, class Node>
- LocalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::ComputeLocalDOF(LocalOrdinal const &lNodeID, LocalOrdinal const &k) const  {
-   LocalOrdinal lDofIndex = lNodeID*fullblocksize_ + k;
-   return lDofIndex;
-  }
-
-  
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::ComputeLocalNode(LocalOrdinal const &ldofID) const {
-    return (ldofID - ldofID%fullblocksize_) / fullblocksize_;
+  // todo plausibility check: entry numDofs[k] == aggToRowMap[k].size()
+
+}  // UnamalgamateAggregatesLO
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out,
+                                                                const VerbLevel verbLevel) const {
+  if (!(verbLevel & Debug))
+    return;
+
+  out << "AmalgamationInfo -- Striding information:"
+      << "\n  fullBlockSize = " << fullblocksize_
+      << "\n  blockID = " << blockid_
+      << "\n  stridingOffset = " << nStridedOffset_
+      << "\n  stridedBlockSize = " << stridedblocksize_
+      << "\n  indexBase = " << indexBase_
+      << std::endl;
+
+  out << "AmalgamationInfo -- DOFs to nodes mapping:\n"
+      << "  Mapping of row DOFs to row nodes:" << *rowTranslation_()
+      << "\n\n  Mapping of column DOFs to column nodes:" << *colTranslation_()
+      << std::endl;
+
+  out << "AmalgamationInfo -- row node map:" << std::endl;
+  nodeRowMap_->describe(out, Teuchos::VERB_EXTREME);
+
+  out << "AmalgamationInfo -- column node map:" << std::endl;
+  nodeColMap_->describe(out, Teuchos::VERB_EXTREME);
+}
+
+/////////////////////////////////////////////////////////////////////////////
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeUnamalgamatedImportDofMap(const Aggregates &aggregates) const {
+  return ComputeUnamalgamatedImportDofMap(aggregates.GetMap());
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeUnamalgamatedImportDofMap(const Teuchos::RCP<const Map> &nodeMap) const {
+  Teuchos::RCP<std::vector<GO> > myDofGids = Teuchos::rcp(new std::vector<GO>);
+  Teuchos::ArrayView<const GO> gEltList    = nodeMap->getLocalElementList();
+  LO nodeElements                          = Teuchos::as<LO>(nodeMap->getLocalNumElements());
+  if (stridedblocksize_ == 1) {
+    for (LO n = 0; n < nodeElements; n++) {
+      GlobalOrdinal gDofIndex = ComputeGlobalDOF(gEltList[n]);
+      myDofGids->push_back(gDofIndex);
+    }
+  } else {
+    for (LO n = 0; n < nodeElements; n++) {
+      for (LocalOrdinal k = 0; k < stridedblocksize_; k++) {
+        GlobalOrdinal gDofIndex = ComputeGlobalDOF(gEltList[n], k);
+        if (columnMap_->isNodeGlobalElement(gDofIndex))
+          myDofGids->push_back(gDofIndex);
+      }
+    }
   }
 
-} //namespace
-
+  Teuchos::ArrayRCP<GO> arr_myDofGids = Teuchos::arcp(myDofGids);
+  Teuchos::RCP<Map> importDofMap      = MapFactory::Build(nodeMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), arr_myDofGids(), nodeMap->getIndexBase(), nodeMap->getComm());
+  return importDofMap;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeGlobalDOF(GlobalOrdinal const &gNodeID, LocalOrdinal const &k) const {
+  // here, the assumption is, that the node map has the same indexBase as the dof map
+  //                            this is the node map index base                    this is the dof map index base
+  GlobalOrdinal gDofIndex = offset_ + (gNodeID - indexBase_) * fullblocksize_ + nStridedOffset_ + k + indexBase_;
+  return gDofIndex;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::ComputeLocalDOF(LocalOrdinal const &lNodeID, LocalOrdinal const &k) const {
+  LocalOrdinal lDofIndex = lNodeID * fullblocksize_ + k;
+  return lDofIndex;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>::ComputeLocalNode(LocalOrdinal const &ldofID) const {
+  return (ldofID - ldofID % fullblocksize_) / fullblocksize_;
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_AMALGAMATIONINFO_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_decl.hpp
index 5e03e2fae3ef..b32aa784e053 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_decl.hpp
@@ -52,7 +52,7 @@
 #include <Xpetra_Vector_fwd.hpp>
 #include <Xpetra_ImportFactory_fwd.hpp>
 #include <Xpetra_CrsGraph_fwd.hpp>
-#include <Xpetra_CrsGraphFactory_fwd.hpp> //TODO
+#include <Xpetra_CrsGraphFactory_fwd.hpp>  //TODO
 #include <Xpetra_StridedMap_fwd.hpp>
 #include <Xpetra_Map_fwd.hpp>
 
@@ -72,116 +72,112 @@
 
 namespace MueLu {
 
-  /*!
-    @class CoalesceDropFactory
-    @brief Factory for creating a graph based on a given matrix.
+/*!
+  @class CoalesceDropFactory
+  @brief Factory for creating a graph based on a given matrix.
 
-    Factory for creating graphs from matrices with entries selectively dropped.
+  Factory for creating graphs from matrices with entries selectively dropped.
 
-    ## Code paths ##
+  ## Code paths ##
 
-    Both the classic dropping strategy as well as a coordinate-based distance laplacian method
-    is implemented. For performance reasons there are four distinctive code paths for the
-    classical method:
+  Both the classic dropping strategy as well as a coordinate-based distance laplacian method
+  is implemented. For performance reasons there are four distinctive code paths for the
+  classical method:
 
-    - one DOF per node without dropping (i.e. "aggregation: drop tol" = 0.0)
-    - one DOF per node with dropping (i.e. "aggregation: drop tol" > 0.0)
-    - number of DOFs per node > 1 withouth dropping
-    - number of DOFs per node > 1 with dropping
+  - one DOF per node without dropping (i.e. "aggregation: drop tol" = 0.0)
+  - one DOF per node with dropping (i.e. "aggregation: drop tol" > 0.0)
+  - number of DOFs per node > 1 withouth dropping
+  - number of DOFs per node > 1 with dropping
 
-    Additionally, there is a code path for the distance-laplacian mode.
+  Additionally, there is a code path for the distance-laplacian mode.
 
-    ## Input/output of CoalesceDropFactory ##
+  ## Input/output of CoalesceDropFactory ##
 
-    ### User parameters of CoalesceDropFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A              | Factory | null |   | * | * | Generating factory of the operator A
-     UnAmalgamationInfo        | Factory | null |   | * | * | Generating factory of type AmalgamationFactory which generates the variable 'UnAmalgamationInfo'. Do not change the default unless you know what you are doing.
-     Coordinates               | Factory | null |   | * | (*) | Generating factory for variable 'Coordinates'. The coordinates are only needed if "distance laplacian" is chosen for the parameter "aggregation: drop scheme"
-     "aggregation: drop scheme" | std::string | "classical" | * | * |   | Coalescing algorithm. You can choose either "classical" (=default) or "distance laplacian"
-     "aggregation: drop tol" | double | 0.0 | * | * |   | Threshold parameter for dropping small entries
-     "aggregation: Dirichlet threshold" | double | 0.0 | * | * |   | Threshold for determining whether entries are zero during Dirichlet row detection
-     "lightweight wrap" | bool | true |   |   *  |   | hidden switch between fast implementation based on MueLu::LWGraph and a failsafe slower implementation based on Xpetra::Graph (for comparison). The user should not change the default value (=true)
+  ### User parameters of CoalesceDropFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A              | Factory | null |   | * | * | Generating factory of the operator A
+   UnAmalgamationInfo        | Factory | null |   | * | * | Generating factory of type AmalgamationFactory which generates the variable 'UnAmalgamationInfo'. Do not change the default unless you know what you are doing.
+   Coordinates               | Factory | null |   | * | (*) | Generating factory for variable 'Coordinates'. The coordinates are only needed if "distance laplacian" is chosen for the parameter "aggregation: drop scheme"
+   "aggregation: drop scheme" | std::string | "classical" | * | * |   | Coalescing algorithm. You can choose either "classical" (=default) or "distance laplacian"
+   "aggregation: drop tol" | double | 0.0 | * | * |   | Threshold parameter for dropping small entries
+   "aggregation: Dirichlet threshold" | double | 0.0 | * | * |   | Threshold for determining whether entries are zero during Dirichlet row detection
+   "lightweight wrap" | bool | true |   |   *  |   | hidden switch between fast implementation based on MueLu::LWGraph and a failsafe slower implementation based on Xpetra::Graph (for comparison). The user should not change the default value (=true)
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoalesceDropFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see CoalesceDropFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoalesceDropFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see CoalesceDropFactory::DeclareInput).
 
-    ### Variables provided by UncoupledAggregationFactory ###
+  ### Variables provided by UncoupledAggregationFactory ###
 
-    After CoalesceDropFactory::Build the following data is available (if requested)
+  After CoalesceDropFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    Graph   | CoalesceDropFactory   | Graph of matrix A
-    DofsPerNode | CoalesceDropFactory | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
+  Parameter | generated by | description
+  ----------|--------------|------------
+  Graph   | CoalesceDropFactory   | Graph of matrix A
+  DofsPerNode | CoalesceDropFactory | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
 
-    ## Amalgamation process ##
+  ## Amalgamation process ##
 
-    The CoalesceDropFactory is internally using the AmalgamationFactory for amalgamating the dof-based maps to node-based maps. The AmalgamationFactory creates the "UnAmalgamationInfo" container
-    which basically stores all the necessary information for translating dof based data to node based data and vice versa. The container is used, since this way the amalgamation is only done once
-    and later reused by other factories.
+  The CoalesceDropFactory is internally using the AmalgamationFactory for amalgamating the dof-based maps to node-based maps. The AmalgamationFactory creates the "UnAmalgamationInfo" container
+  which basically stores all the necessary information for translating dof based data to node based data and vice versa. The container is used, since this way the amalgamation is only done once
+  and later reused by other factories.
 
-    Of course, often one does not need the information from the "UnAmalgamationInfo" container since the same information could be extracted of the "Graph" or the map from the "Coordinates" vector.
-    However, there are also some situations (e.g. when doing rebalancing based on HyperGraph partitioning without coordinate information) where one has not access to a "Graph" or "Coordinates" variable.
-  */
+  Of course, often one does not need the information from the "UnAmalgamationInfo" container since the same information could be extracted of the "Graph" or the map from the "Coordinates" vector.
+  However, there are also some situations (e.g. when doing rebalancing based on HyperGraph partitioning without coordinate information) where one has not access to a "Graph" or "Coordinates" variable.
+*/
 
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class CoalesceDropFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CoalesceDropFactory : public SingleLevelFactoryBase {
 #undef MUELU_COALESCEDROPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  CoalesceDropFactory();
 
-    //! Constructor
-    CoalesceDropFactory();
+  //! Destructor
+  virtual ~CoalesceDropFactory() {}
 
-    //! Destructor
-    virtual ~CoalesceDropFactory() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  /// set predrop function
+  void SetPreDropFunction(const RCP<MueLu::PreDropFunctionBaseClass<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& predrop) { predrop_ = predrop; }
 
-    /// set predrop function
-    void SetPreDropFunction(const RCP<MueLu::PreDropFunctionBaseClass<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &predrop) { predrop_ = predrop; }
+  //@}
 
-    //@}
+  void Build(Level& currentLevel) const;  // Build
 
-    void Build(Level &currentLevel) const; // Build
+ private:
+  // pre-drop function
+  mutable RCP<PreDropFunctionBaseClass> predrop_;
 
-  private:
+  //! Method to merge rows of matrix for systems of PDEs.
+  void MergeRows(const Matrix& A, const LO row, Array<LO>& cols, const Array<LO>& translation) const;
+  void MergeRowsWithDropping(const Matrix& A, const LO row, const ArrayRCP<const SC>& ghostedDiagVals, SC threshold, Array<LO>& cols, const Array<LO>& translation) const;
 
-    // pre-drop function
-    mutable
-     RCP<PreDropFunctionBaseClass> predrop_;
+  // When we want to decouple a block diagonal system (returns Teuchos::null if generate_matrix is false)
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BlockDiagonalize(Level& currentLevel, const RCP<Matrix>& A, bool generate_matrix) const;
 
-    //! Method to merge rows of matrix for systems of PDEs.
-    void MergeRows(const Matrix& A, const LO row, Array<LO>& cols, const Array<LO>& translation) const;
-    void MergeRowsWithDropping(const Matrix& A, const LO row, const ArrayRCP<const SC>& ghostedDiagVals, SC threshold, Array<LO>& cols, const Array<LO>& translation) const;
+  // When we want to decouple a block diagonal system via a *graph*
+  void BlockDiagonalizeGraph(const RCP<GraphBase>& inputGraph, const RCP<LocalOrdinalVector>& ghostedBlockNumber, RCP<GraphBase>& outputGraph, RCP<const Import>& importer) const;
 
+};  // class CoalesceDropFactory
 
-    // When we want to decouple a block diagonal system (returns Teuchos::null if generate_matrix is false)
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > BlockDiagonalize(Level & currentLevel,const RCP<Matrix> & A, bool generate_matrix) const;
-
-    // When we want to decouple a block diagonal system via a *graph*
-    void BlockDiagonalizeGraph(const RCP<GraphBase> & inputGraph, const RCP<LocalOrdinalVector> & ghostedBlockNumber, RCP<GraphBase> & outputGraph, RCP<const Import> & importer) const;
-
-  }; //class CoalesceDropFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COALESCEDROPFACTORY_SHORT
-#endif // MUELU_COALESCEDROPFACTORY_DECL_HPP
+#endif  // MUELU_COALESCEDROPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_def.hpp
index c4bf4178cbaf..34a2f3db77f4 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_def.hpp
@@ -87,1047 +87,1022 @@
 // Should be removed once we are confident that this works.
 //#define DJS_READ_ENV_VARIABLES
 
-
 namespace MueLu {
 
-  namespace Details {
-    template<class real_type, class LO>
-    struct DropTol {
-      
-      DropTol()               = default;
-      DropTol(DropTol const&) = default;
-      DropTol(DropTol &&)     = default;
-      
-      DropTol& operator=(DropTol const&) = default;
-      DropTol& operator=(DropTol &&)     = default;
-      
-      DropTol(real_type val_, real_type diag_, LO col_, bool drop_)
-        : val{val_}, diag{diag_},  col{col_}, drop{drop_}
-      {}
-      
-      real_type val  {Teuchos::ScalarTraits<real_type>::zero()};
-                  real_type diag {Teuchos::ScalarTraits<real_type>::zero()};
-      LO        col  {Teuchos::OrdinalTraits<LO>::invalid()};
-      bool      drop {true};
-
-      // CMS: Auxillary information for debugging info
-      //      real_type aux_val {Teuchos::ScalarTraits<real_type>::nan()};
-    };
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+namespace Details {
+template <class real_type, class LO>
+struct DropTol {
+  DropTol()               = default;
+  DropTol(DropTol const&) = default;
+  DropTol(DropTol&&)      = default;
 
-#define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: drop tol");
-    SET_VALID_ENTRY("aggregation: use ml scaling of drop tol");
-    SET_VALID_ENTRY("aggregation: Dirichlet threshold");
-    SET_VALID_ENTRY("aggregation: greedy Dirichlet");
-    SET_VALID_ENTRY("aggregation: row sum drop tol");
-    SET_VALID_ENTRY("aggregation: drop scheme");
-    SET_VALID_ENTRY("aggregation: block diagonal: interleaved blocksize");
-    SET_VALID_ENTRY("aggregation: distance laplacian directional weights");
-    SET_VALID_ENTRY("aggregation: dropping may create Dirichlet");
-
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      // "signed classical" is the Ruge-Stuben style (relative to max off-diagonal), "sign classical sa" is the signed version of the sa criterion (relative to the diagonal values)
-      validParamList->getEntry("aggregation: drop scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("signed classical sa","classical", "distance laplacian","signed classical","block diagonal","block diagonal classical","block diagonal distance laplacian","block diagonal signed classical","block diagonal colored signed classical"), "aggregation: drop scheme")));
-                                                                        
-    }
-    SET_VALID_ENTRY("aggregation: distance laplacian algo");
-    SET_VALID_ENTRY("aggregation: classical algo");
-    SET_VALID_ENTRY("aggregation: coloring: localize color graph");
-#undef  SET_VALID_ENTRY
-    validParamList->set< bool >                  ("lightweight wrap",           true, "Experimental option for lightweight graph access");
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory for UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory for Coordinates");
-    validParamList->set< RCP<const FactoryBase> >("BlockNumber",        Teuchos::null, "Generating factory for BlockNUmber");
-
-    return validParamList;
-  }
-    
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CoalesceDropFactory() : predrop_(Teuchos::null) { }
-  
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "UnAmalgamationInfo");
-    
-    const ParameterList& pL = GetParameterList();
-    if (pL.get<bool>("lightweight wrap") == true) {
-      std::string algo = pL.get<std::string>("aggregation: drop scheme");
-      if (algo == "distance laplacian" || algo == "block diagonal distance laplacian") {
-        Input(currentLevel, "Coordinates");
-      }
-      if(algo == "signed classical sa")
-        ;
-      else if (algo.find("block diagonal") != std::string::npos || algo.find("signed classical") != std::string::npos)  {
-        Input(currentLevel, "BlockNumber");
-      }
-    }
-    
-  }
-  
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-    typedef Xpetra::MultiVectorFactory<real_type,LO,GO,NO> RealValuedMultiVectorFactory;
-    
-    if (predrop_ != Teuchos::null)
-      GetOStream(Parameters0) << predrop_->description();
-
-    RCP<Matrix> realA = Get< RCP<Matrix> >(currentLevel, "A");
-    RCP<AmalgamationInfo> amalInfo = Get< RCP<AmalgamationInfo> >(currentLevel, "UnAmalgamationInfo");
-    const ParameterList  & pL = GetParameterList();
-    bool doExperimentalWrap = pL.get<bool>("lightweight wrap");
-
-    GetOStream(Parameters0) << "lightweight wrap = " << doExperimentalWrap << std::endl;
-    std::string algo = pL.get<std::string>("aggregation: drop scheme");
-    const bool aggregationMayCreateDirichlet = pL.get<bool>("aggregation: dropping may create Dirichlet");
-    
-    RCP<RealValuedMultiVector> Coords;
-    RCP<Matrix> A;
+  DropTol& operator=(DropTol const&) = default;
+  DropTol& operator=(DropTol&&)      = default;
 
-    bool use_block_algorithm=false;
-    LO interleaved_blocksize = as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize"));
-    bool useSignedClassicalRS = false;
-    bool useSignedClassicalSA = false;
-    bool generateColoringGraph = false;
+  DropTol(real_type val_, real_type diag_, LO col_, bool drop_)
+    : val{val_}
+    , diag{diag_}
+    , col{col_}
+    , drop{drop_} {}
 
-    // NOTE:  If we're doing blockDiagonal, we'll not want to do rowSum twice (we'll do it
-    // in the block diagonalization). So we'll clobber the rowSumTol with -1.0 in this case
-    typename STS::magnitudeType rowSumTol = as<typename STS::magnitudeType>(pL.get<double>("aggregation: row sum drop tol"));
+  real_type val{Teuchos::ScalarTraits<real_type>::zero()};
+  real_type diag{Teuchos::ScalarTraits<real_type>::zero()};
+  LO col{Teuchos::OrdinalTraits<LO>::invalid()};
+  bool drop{true};
 
+  // CMS: Auxillary information for debugging info
+  //      real_type aux_val {Teuchos::ScalarTraits<real_type>::nan()};
+};
+}  // namespace Details
 
-    RCP<LocalOrdinalVector> ghostedBlockNumber;
-    ArrayRCP<const LO> g_block_id;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    if(algo == "distance laplacian" ) {
-      // Grab the coordinates for distance laplacian
-      Coords = Get< RCP<RealValuedMultiVector > >(currentLevel, "Coordinates");
-      A = realA;
-    }
-    else if(algo == "signed classical sa") {
-      useSignedClassicalSA = true;
-      algo = "classical";
-      A = realA;
+#define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
+  SET_VALID_ENTRY("aggregation: drop tol");
+  SET_VALID_ENTRY("aggregation: use ml scaling of drop tol");
+  SET_VALID_ENTRY("aggregation: Dirichlet threshold");
+  SET_VALID_ENTRY("aggregation: greedy Dirichlet");
+  SET_VALID_ENTRY("aggregation: row sum drop tol");
+  SET_VALID_ENTRY("aggregation: drop scheme");
+  SET_VALID_ENTRY("aggregation: block diagonal: interleaved blocksize");
+  SET_VALID_ENTRY("aggregation: distance laplacian directional weights");
+  SET_VALID_ENTRY("aggregation: dropping may create Dirichlet");
+
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    // "signed classical" is the Ruge-Stuben style (relative to max off-diagonal), "sign classical sa" is the signed version of the sa criterion (relative to the diagonal values)
+    validParamList->getEntry("aggregation: drop scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("signed classical sa", "classical", "distance laplacian", "signed classical", "block diagonal", "block diagonal classical", "block diagonal distance laplacian", "block diagonal signed classical", "block diagonal colored signed classical"), "aggregation: drop scheme")));
+  }
+  SET_VALID_ENTRY("aggregation: distance laplacian algo");
+  SET_VALID_ENTRY("aggregation: classical algo");
+  SET_VALID_ENTRY("aggregation: coloring: localize color graph");
+#undef SET_VALID_ENTRY
+  validParamList->set<bool>("lightweight wrap", true, "Experimental option for lightweight graph access");
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase>>("UnAmalgamationInfo", Teuchos::null, "Generating factory for UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase>>("Coordinates", Teuchos::null, "Generating factory for Coordinates");
+  validParamList->set<RCP<const FactoryBase>>("BlockNumber", Teuchos::null, "Generating factory for BlockNUmber");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CoalesceDropFactory()
+  : predrop_(Teuchos::null) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "UnAmalgamationInfo");
+
+  const ParameterList& pL = GetParameterList();
+  if (pL.get<bool>("lightweight wrap") == true) {
+    std::string algo = pL.get<std::string>("aggregation: drop scheme");
+    if (algo == "distance laplacian" || algo == "block diagonal distance laplacian") {
+      Input(currentLevel, "Coordinates");
     }
-    else if(algo == "signed classical" || algo == "block diagonal colored signed classical" || algo == "block diagonal signed classical") {
-      useSignedClassicalRS = true;
-      //      if(realA->GetFixedBlockSize() > 1) {
-        RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector> >(currentLevel, "BlockNumber");
-        // Ghost the column block numbers if we need to
-        RCP<const Import> importer = realA->getCrsGraph()->getImporter();
-        if(!importer.is_null()) {
-          SubFactoryMonitor m1(*this, "Block Number import", currentLevel);
-          ghostedBlockNumber= Xpetra::VectorFactory<LO,LO,GO,NO>::Build(importer->getTargetMap());
-          ghostedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
-        }
-        else {
-          ghostedBlockNumber = BlockNumber;
-        }
-        g_block_id = ghostedBlockNumber->getData(0);
-        //      }
-      if(algo == "block diagonal colored signed classical")
-        generateColoringGraph=true;
-      algo = "classical";
-      A = realA;
-
+    if (algo == "signed classical sa")
+      ;
+    else if (algo.find("block diagonal") != std::string::npos || algo.find("signed classical") != std::string::npos) {
+      Input(currentLevel, "BlockNumber");
     }
-    else if(algo == "block diagonal") {
-      // Handle the "block diagonal" filtering and then leave
-      BlockDiagonalize(currentLevel,realA,false);
-      return;
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVectorFactory<real_type, LO, GO, NO> RealValuedMultiVectorFactory;
+
+  if (predrop_ != Teuchos::null)
+    GetOStream(Parameters0) << predrop_->description();
+
+  RCP<Matrix> realA              = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<AmalgamationInfo> amalInfo = Get<RCP<AmalgamationInfo>>(currentLevel, "UnAmalgamationInfo");
+  const ParameterList& pL        = GetParameterList();
+  bool doExperimentalWrap        = pL.get<bool>("lightweight wrap");
+
+  GetOStream(Parameters0) << "lightweight wrap = " << doExperimentalWrap << std::endl;
+  std::string algo                         = pL.get<std::string>("aggregation: drop scheme");
+  const bool aggregationMayCreateDirichlet = pL.get<bool>("aggregation: dropping may create Dirichlet");
+
+  RCP<RealValuedMultiVector> Coords;
+  RCP<Matrix> A;
+
+  bool use_block_algorithm   = false;
+  LO interleaved_blocksize   = as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize"));
+  bool useSignedClassicalRS  = false;
+  bool useSignedClassicalSA  = false;
+  bool generateColoringGraph = false;
+
+  // NOTE:  If we're doing blockDiagonal, we'll not want to do rowSum twice (we'll do it
+  // in the block diagonalization). So we'll clobber the rowSumTol with -1.0 in this case
+  typename STS::magnitudeType rowSumTol = as<typename STS::magnitudeType>(pL.get<double>("aggregation: row sum drop tol"));
+
+  RCP<LocalOrdinalVector> ghostedBlockNumber;
+  ArrayRCP<const LO> g_block_id;
+
+  if (algo == "distance laplacian") {
+    // Grab the coordinates for distance laplacian
+    Coords = Get<RCP<RealValuedMultiVector>>(currentLevel, "Coordinates");
+    A      = realA;
+  } else if (algo == "signed classical sa") {
+    useSignedClassicalSA = true;
+    algo                 = "classical";
+    A                    = realA;
+  } else if (algo == "signed classical" || algo == "block diagonal colored signed classical" || algo == "block diagonal signed classical") {
+    useSignedClassicalRS = true;
+    //      if(realA->GetFixedBlockSize() > 1) {
+    RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector>>(currentLevel, "BlockNumber");
+    // Ghost the column block numbers if we need to
+    RCP<const Import> importer = realA->getCrsGraph()->getImporter();
+    if (!importer.is_null()) {
+      SubFactoryMonitor m1(*this, "Block Number import", currentLevel);
+      ghostedBlockNumber = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(importer->getTargetMap());
+      ghostedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
+    } else {
+      ghostedBlockNumber = BlockNumber;
     }
-    else if (algo == "block diagonal classical" || algo == "block diagonal distance laplacian")  {
-      // Handle the "block diagonal" filtering, and then continue onward
-      use_block_algorithm = true;
-      RCP<Matrix> filteredMatrix = BlockDiagonalize(currentLevel,realA,true);
-      if(algo == "block diagonal distance laplacian") {
-        // We now need to expand the coordinates by the interleaved blocksize
-        RCP<RealValuedMultiVector> OldCoords = Get< RCP<RealValuedMultiVector > >(currentLevel, "Coordinates");
-        if (OldCoords->getLocalLength() != realA->getLocalNumRows()) {
-           LO dim = (LO) OldCoords->getNumVectors();
-           Coords = RealValuedMultiVectorFactory::Build(realA->getRowMap(),dim);
-           for(LO k=0; k<dim; k++){
-             ArrayRCP<const real_type> old_vec = OldCoords->getData(k);
-             ArrayRCP<real_type>       new_vec = Coords->getDataNonConst(k);
-             for(LO i=0; i <(LO)OldCoords->getLocalLength(); i++) {
-               LO new_base = i*dim;
-               for(LO j=0; j<interleaved_blocksize; j++)
-                new_vec[new_base + j] = old_vec[i];
-             }
+    g_block_id = ghostedBlockNumber->getData(0);
+    //      }
+    if (algo == "block diagonal colored signed classical")
+      generateColoringGraph = true;
+    algo = "classical";
+    A    = realA;
+
+  } else if (algo == "block diagonal") {
+    // Handle the "block diagonal" filtering and then leave
+    BlockDiagonalize(currentLevel, realA, false);
+    return;
+  } else if (algo == "block diagonal classical" || algo == "block diagonal distance laplacian") {
+    // Handle the "block diagonal" filtering, and then continue onward
+    use_block_algorithm        = true;
+    RCP<Matrix> filteredMatrix = BlockDiagonalize(currentLevel, realA, true);
+    if (algo == "block diagonal distance laplacian") {
+      // We now need to expand the coordinates by the interleaved blocksize
+      RCP<RealValuedMultiVector> OldCoords = Get<RCP<RealValuedMultiVector>>(currentLevel, "Coordinates");
+      if (OldCoords->getLocalLength() != realA->getLocalNumRows()) {
+        LO dim = (LO)OldCoords->getNumVectors();
+        Coords = RealValuedMultiVectorFactory::Build(realA->getRowMap(), dim);
+        for (LO k = 0; k < dim; k++) {
+          ArrayRCP<const real_type> old_vec = OldCoords->getData(k);
+          ArrayRCP<real_type> new_vec       = Coords->getDataNonConst(k);
+          for (LO i = 0; i < (LO)OldCoords->getLocalLength(); i++) {
+            LO new_base = i * dim;
+            for (LO j = 0; j < interleaved_blocksize; j++)
+              new_vec[new_base + j] = old_vec[i];
           }
         }
-        else {
-          Coords = OldCoords;
-        }
-        algo = "distance laplacian";
+      } else {
+        Coords = OldCoords;
       }
-      else if(algo == "block diagonal classical") {
-        algo = "classical";
-      }
-      // All cases
-      A = filteredMatrix;
-      rowSumTol = -1.0;
-    }
-    else {
-      A = realA;
+      algo = "distance laplacian";
+    } else if (algo == "block diagonal classical") {
+      algo = "classical";
     }
+    // All cases
+    A         = filteredMatrix;
+    rowSumTol = -1.0;
+  } else {
+    A = realA;
+  }
 
-    // Distance Laplacian weights
-    Array<double> dlap_weights = pL.get<Array<double> >("aggregation: distance laplacian directional weights");
-    enum {NO_WEIGHTS=0, SINGLE_WEIGHTS, BLOCK_WEIGHTS};
-    int use_dlap_weights = NO_WEIGHTS;
-    if(algo == "distance laplacian") {
-      LO dim = (LO) Coords->getNumVectors();
-      // If anything isn't 1.0 we need to turn on the weighting
-      bool non_unity = false;
-      for (LO i=0; !non_unity && i<(LO)dlap_weights.size(); i++) {
-        if(dlap_weights[i] != 1.0) {
-          non_unity = true;
-        }
+  // Distance Laplacian weights
+  Array<double> dlap_weights = pL.get<Array<double>>("aggregation: distance laplacian directional weights");
+  enum { NO_WEIGHTS = 0,
+         SINGLE_WEIGHTS,
+         BLOCK_WEIGHTS };
+  int use_dlap_weights = NO_WEIGHTS;
+  if (algo == "distance laplacian") {
+    LO dim = (LO)Coords->getNumVectors();
+    // If anything isn't 1.0 we need to turn on the weighting
+    bool non_unity = false;
+    for (LO i = 0; !non_unity && i < (LO)dlap_weights.size(); i++) {
+      if (dlap_weights[i] != 1.0) {
+        non_unity = true;
       }
-      if(non_unity) {
-        LO blocksize = use_block_algorithm ? as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize")) : 1;
-        if((LO)dlap_weights.size() == dim)
-          use_dlap_weights = SINGLE_WEIGHTS;
-        else if((LO)dlap_weights.size() == blocksize * dim)
-          use_dlap_weights = BLOCK_WEIGHTS;
-        else  {
-          TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::RuntimeError,
-                                     "length of 'aggregation: distance laplacian directional weights' must equal the coordinate dimension OR the coordinate dimension times the blocksize");
-        }
-        if (GetVerbLevel() & Statistics1)
-          GetOStream(Statistics1) << "Using distance laplacian weights: "<<dlap_weights<<std::endl;
+    }
+    if (non_unity) {
+      LO blocksize = use_block_algorithm ? as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize")) : 1;
+      if ((LO)dlap_weights.size() == dim)
+        use_dlap_weights = SINGLE_WEIGHTS;
+      else if ((LO)dlap_weights.size() == blocksize * dim)
+        use_dlap_weights = BLOCK_WEIGHTS;
+      else {
+        TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::RuntimeError,
+                                   "length of 'aggregation: distance laplacian directional weights' must equal the coordinate dimension OR the coordinate dimension times the blocksize");
       }
+      if (GetVerbLevel() & Statistics1)
+        GetOStream(Statistics1) << "Using distance laplacian weights: " << dlap_weights << std::endl;
     }
+  }
 
-    // decide wether to use the fast-track code path for standard maps or the somewhat slower
-    // code path for non-standard maps
-    /*bool bNonStandardMaps = false;
-    if (A->IsView("stridedMaps") == true) {
-      Teuchos::RCP<const Map> myMap = A->getRowMap("stridedMaps");
-      Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-      if (strMap->getStridedBlockId() != -1 || strMap->getOffset() > 0)
-        bNonStandardMaps = true;
-    }*/
-
-    if (doExperimentalWrap) {
-      TEUCHOS_TEST_FOR_EXCEPTION(predrop_ != null   && algo != "classical", Exceptions::RuntimeError, "Dropping function must not be provided for \"" << algo << "\" algorithm");
-      TEUCHOS_TEST_FOR_EXCEPTION(algo != "classical" && algo != "distance laplacian" && algo != "signed classical", Exceptions::RuntimeError, "\"algorithm\" must be one of (classical|distance laplacian|signed classical)");
-
-      SC threshold;
-      // If we're doing the ML-style halving of the drop tol at each level, we do that here.
-      if (pL.get<bool>("aggregation: use ml scaling of drop tol"))
-        threshold = pL.get<double>("aggregation: drop tol") / pow(2.0,currentLevel.GetLevelID());
-      else
-        threshold =  as<SC>(pL.get<double>("aggregation: drop tol"));
-
-
-      std::string distanceLaplacianAlgoStr = pL.get<std::string>("aggregation: distance laplacian algo");
-      std::string classicalAlgoStr = pL.get<std::string>("aggregation: classical algo");
-      real_type realThreshold = STS::magnitude(threshold);// CMS: Rename this to "magnitude threshold" sometime
-
-      ////////////////////////////////////////////////////
-      // Remove this bit once we are confident that cut-based dropping works.
+  // decide wether to use the fast-track code path for standard maps or the somewhat slower
+  // code path for non-standard maps
+  /*bool bNonStandardMaps = false;
+  if (A->IsView("stridedMaps") == true) {
+    Teuchos::RCP<const Map> myMap = A->getRowMap("stridedMaps");
+    Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+    if (strMap->getStridedBlockId() != -1 || strMap->getOffset() > 0)
+      bNonStandardMaps = true;
+  }*/
+
+  if (doExperimentalWrap) {
+    TEUCHOS_TEST_FOR_EXCEPTION(predrop_ != null && algo != "classical", Exceptions::RuntimeError, "Dropping function must not be provided for \"" << algo << "\" algorithm");
+    TEUCHOS_TEST_FOR_EXCEPTION(algo != "classical" && algo != "distance laplacian" && algo != "signed classical", Exceptions::RuntimeError, "\"algorithm\" must be one of (classical|distance laplacian|signed classical)");
+
+    SC threshold;
+    // If we're doing the ML-style halving of the drop tol at each level, we do that here.
+    if (pL.get<bool>("aggregation: use ml scaling of drop tol"))
+      threshold = pL.get<double>("aggregation: drop tol") / pow(2.0, currentLevel.GetLevelID());
+    else
+      threshold = as<SC>(pL.get<double>("aggregation: drop tol"));
+
+    std::string distanceLaplacianAlgoStr = pL.get<std::string>("aggregation: distance laplacian algo");
+    std::string classicalAlgoStr         = pL.get<std::string>("aggregation: classical algo");
+    real_type realThreshold              = STS::magnitude(threshold);  // CMS: Rename this to "magnitude threshold" sometime
+
+    ////////////////////////////////////////////////////
+    // Remove this bit once we are confident that cut-based dropping works.
 #ifdef HAVE_MUELU_DEBUG
-      int distanceLaplacianCutVerbose = 0;
+    int distanceLaplacianCutVerbose = 0;
 #endif
 #ifdef DJS_READ_ENV_VARIABLES
-      if (getenv("MUELU_DROP_TOLERANCE_MODE")) {
-        distanceLaplacianAlgoStr = std::string(getenv("MUELU_DROP_TOLERANCE_MODE"));
-      }
+    if (getenv("MUELU_DROP_TOLERANCE_MODE")) {
+      distanceLaplacianAlgoStr = std::string(getenv("MUELU_DROP_TOLERANCE_MODE"));
+    }
 
-      if (getenv("MUELU_DROP_TOLERANCE_THRESHOLD")) {
-        auto tmp = atoi(getenv("MUELU_DROP_TOLERANCE_THRESHOLD"));
-        realThreshold = 1e-4*tmp;
-      }
+    if (getenv("MUELU_DROP_TOLERANCE_THRESHOLD")) {
+      auto tmp      = atoi(getenv("MUELU_DROP_TOLERANCE_THRESHOLD"));
+      realThreshold = 1e-4 * tmp;
+    }
 
-# ifdef HAVE_MUELU_DEBUG
-      if (getenv("MUELU_DROP_TOLERANCE_VERBOSE")) {
-        distanceLaplacianCutVerbose = atoi(getenv("MUELU_DROP_TOLERANCE_VERBOSE"));
-      }
-# endif
+#ifdef HAVE_MUELU_DEBUG
+    if (getenv("MUELU_DROP_TOLERANCE_VERBOSE")) {
+      distanceLaplacianCutVerbose = atoi(getenv("MUELU_DROP_TOLERANCE_VERBOSE"));
+    }
 #endif
-      ////////////////////////////////////////////////////
- 
-      enum decisionAlgoType {defaultAlgo, unscaled_cut, scaled_cut, scaled_cut_symmetric};
-
-      decisionAlgoType distanceLaplacianAlgo = defaultAlgo;
-      decisionAlgoType classicalAlgo = defaultAlgo;
-      if (algo == "distance laplacian") {
-        if (distanceLaplacianAlgoStr == "default")
-          distanceLaplacianAlgo = defaultAlgo;
-        else if (distanceLaplacianAlgoStr == "unscaled cut")
-          distanceLaplacianAlgo = unscaled_cut;
-        else if (distanceLaplacianAlgoStr == "scaled cut")
-          distanceLaplacianAlgo = scaled_cut;
-        else if (distanceLaplacianAlgoStr == "scaled cut symmetric")
-          distanceLaplacianAlgo = scaled_cut_symmetric;
-        else
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "\"aggregation: distance laplacian algo\" must be one of (default|unscaled cut|scaled cut), not \"" << distanceLaplacianAlgoStr << "\"");
-        GetOStream(Runtime0) << "algorithm = \"" << algo << "\" distance laplacian algorithm = \"" << distanceLaplacianAlgoStr << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize()<< std::endl;
-      } else if (algo == "classical") {
-        if (classicalAlgoStr == "default")
-          classicalAlgo = defaultAlgo;
-        else if (classicalAlgoStr == "unscaled cut")
-          classicalAlgo = unscaled_cut;
-        else if (classicalAlgoStr == "scaled cut")
-          classicalAlgo = scaled_cut;
-        else
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "\"aggregation: classical algo\" must be one of (default|unscaled cut|scaled cut), not \"" << classicalAlgoStr << "\"");
-        GetOStream(Runtime0) << "algorithm = \"" << algo << "\" classical algorithm = \"" << classicalAlgoStr << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
-        
-      } else
-        GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
-      Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
+#endif
+    ////////////////////////////////////////////////////
+
+    enum decisionAlgoType { defaultAlgo,
+                            unscaled_cut,
+                            scaled_cut,
+                            scaled_cut_symmetric };
+
+    decisionAlgoType distanceLaplacianAlgo = defaultAlgo;
+    decisionAlgoType classicalAlgo         = defaultAlgo;
+    if (algo == "distance laplacian") {
+      if (distanceLaplacianAlgoStr == "default")
+        distanceLaplacianAlgo = defaultAlgo;
+      else if (distanceLaplacianAlgoStr == "unscaled cut")
+        distanceLaplacianAlgo = unscaled_cut;
+      else if (distanceLaplacianAlgoStr == "scaled cut")
+        distanceLaplacianAlgo = scaled_cut;
+      else if (distanceLaplacianAlgoStr == "scaled cut symmetric")
+        distanceLaplacianAlgo = scaled_cut_symmetric;
+      else
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "\"aggregation: distance laplacian algo\" must be one of (default|unscaled cut|scaled cut), not \"" << distanceLaplacianAlgoStr << "\"");
+      GetOStream(Runtime0) << "algorithm = \"" << algo << "\" distance laplacian algorithm = \"" << distanceLaplacianAlgoStr << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+    } else if (algo == "classical") {
+      if (classicalAlgoStr == "default")
+        classicalAlgo = defaultAlgo;
+      else if (classicalAlgoStr == "unscaled cut")
+        classicalAlgo = unscaled_cut;
+      else if (classicalAlgoStr == "scaled cut")
+        classicalAlgo = scaled_cut;
+      else
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "\"aggregation: classical algo\" must be one of (default|unscaled cut|scaled cut), not \"" << classicalAlgoStr << "\"");
+      GetOStream(Runtime0) << "algorithm = \"" << algo << "\" classical algorithm = \"" << classicalAlgoStr << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+
+    } else
+      GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+    Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
 
-      const typename STS::magnitudeType dirichletThreshold = STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
+    const typename STS::magnitudeType dirichletThreshold = STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
 
+    // NOTE: We don't support signed classical RS or SA with cut drop at present
+    TEUCHOS_TEST_FOR_EXCEPTION(useSignedClassicalRS && classicalAlgo != defaultAlgo, Exceptions::RuntimeError, "\"aggregation: classical algo\" != default is not supported for scalled classical aggregation");
+    TEUCHOS_TEST_FOR_EXCEPTION(useSignedClassicalSA && classicalAlgo != defaultAlgo, Exceptions::RuntimeError, "\"aggregation: classical algo\" != default is not supported for scalled classical sa aggregation");
 
-      // NOTE: We don't support signed classical RS or SA with cut drop at present
-      TEUCHOS_TEST_FOR_EXCEPTION(useSignedClassicalRS && classicalAlgo != defaultAlgo, Exceptions::RuntimeError, "\"aggregation: classical algo\" != default is not supported for scalled classical aggregation");
-      TEUCHOS_TEST_FOR_EXCEPTION(useSignedClassicalSA && classicalAlgo != defaultAlgo, Exceptions::RuntimeError, "\"aggregation: classical algo\" != default is not supported for scalled classical sa aggregation");
+    GO numDropped = 0, numTotal = 0;
+    std::string graphType = "unamalgamated";  // for description purposes only
 
-      GO numDropped = 0, numTotal = 0;
-      std::string graphType = "unamalgamated"; //for description purposes only
+    /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
+     BlockSize is the number of storage blocks that must kept together during the amalgamation process.
 
-      
-      /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
-       BlockSize is the number of storage blocks that must kept together during the amalgamation process.
+     Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
 
-       Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
+     numPDEs = BlockSize * storageblocksize.
 
-       numPDEs = BlockSize * storageblocksize.
-       
-       If numPDEs==1
-         Matrix is point storage (classical CRS storage).  storageblocksize=1 and BlockSize=1
-         No other values makes sense.
+     If numPDEs==1
+       Matrix is point storage (classical CRS storage).  storageblocksize=1 and BlockSize=1
+       No other values makes sense.
 
-       If numPDEs>1
-         If matrix uses point storage, then storageblocksize=1  and BlockSize=numPDEs.
-         If matrix uses block storage, with block size of n, then storageblocksize=n, and BlockSize=numPDEs/n.
-         Thus far, only storageblocksize=numPDEs and BlockSize=1 has been tested.
-      */
-      TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0,Exceptions::RuntimeError,"A->GetFixedBlockSize() needs to be a multiple of A->GetStorageBlockSize()");
-      const LO BlockSize = A->GetFixedBlockSize() / A->GetStorageBlockSize();
+     If numPDEs>1
+       If matrix uses point storage, then storageblocksize=1  and BlockSize=numPDEs.
+       If matrix uses block storage, with block size of n, then storageblocksize=n, and BlockSize=numPDEs/n.
+       Thus far, only storageblocksize=numPDEs and BlockSize=1 has been tested.
+    */
+    TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0, Exceptions::RuntimeError, "A->GetFixedBlockSize() needs to be a multiple of A->GetStorageBlockSize()");
+    const LO BlockSize = A->GetFixedBlockSize() / A->GetStorageBlockSize();
 
+    /************************** RS or SA-style Classical Dropping (and variants) **************************/
+    if (algo == "classical") {
+      if (predrop_ == null) {
+        // ap: this is a hack: had to declare predrop_ as mutable
+        predrop_ = rcp(new PreDropFunctionConstVal(threshold));
+      }
 
-      /************************** RS or SA-style Classical Dropping (and variants) **************************/
-      if (algo == "classical") {
-        if (predrop_ == null) {
-          // ap: this is a hack: had to declare predrop_ as mutable
-          predrop_ = rcp(new PreDropFunctionConstVal(threshold));
+      if (predrop_ != null) {
+        RCP<PreDropFunctionConstVal> predropConstVal = rcp_dynamic_cast<PreDropFunctionConstVal>(predrop_);
+        TEUCHOS_TEST_FOR_EXCEPTION(predropConstVal == Teuchos::null, Exceptions::BadCast,
+                                   "MueLu::CoalesceFactory::Build: cast to PreDropFunctionConstVal failed.");
+        // If a user provided a predrop function, it overwrites the XML threshold parameter
+        SC newt = predropConstVal->GetThreshold();
+        if (newt != threshold) {
+          GetOStream(Warnings0) << "switching threshold parameter from " << threshold << " (list) to " << newt << " (user function" << std::endl;
+          threshold = newt;
+        }
+      }
+      // At this points we either have
+      //     (predrop_ != null)
+      // Therefore, it is sufficient to check only threshold
+      if (BlockSize == 1 && threshold == STS::zero() && !useSignedClassicalRS && !useSignedClassicalSA && A->hasCrsGraph()) {
+        // Case 1:  scalar problem, no dropping => just use matrix graph
+        RCP<GraphBase> graph = rcp(new Graph(A->getCrsGraph(), "graph of A"));
+        // Detect and record rows that correspond to Dirichlet boundary conditions
+        ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+        if (rowSumTol > 0.)
+          Utilities::ApplyRowSumCriterion(*A, rowSumTol, boundaryNodes);
+
+        graph->SetBoundaryNodeMap(boundaryNodes);
+        numTotal = A->getLocalNumEntries();
+
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
+          for (LO i = 0; i < boundaryNodes.size(); ++i)
+            if (boundaryNodes[i])
+              numLocalBoundaryNodes++;
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
         }
 
-        if (predrop_ != null) {
-          RCP<PreDropFunctionConstVal> predropConstVal = rcp_dynamic_cast<PreDropFunctionConstVal>(predrop_);
-          TEUCHOS_TEST_FOR_EXCEPTION(predropConstVal == Teuchos::null, Exceptions::BadCast,
-                                     "MueLu::CoalesceFactory::Build: cast to PreDropFunctionConstVal failed.");
-          // If a user provided a predrop function, it overwrites the XML threshold parameter
-          SC newt = predropConstVal->GetThreshold();
-          if (newt != threshold) {
-            GetOStream(Warnings0) << "switching threshold parameter from " << threshold << " (list) to " << newt << " (user function" << std::endl;
-            threshold = newt;
+        Set(currentLevel, "DofsPerNode", 1);
+        Set(currentLevel, "Graph", graph);
+
+      } else if ((BlockSize == 1 && threshold != STS::zero()) ||
+                 (BlockSize == 1 && threshold == STS::zero() && !A->hasCrsGraph()) ||
+                 (BlockSize == 1 && useSignedClassicalRS) ||
+                 (BlockSize == 1 && useSignedClassicalSA)) {
+        // Case 2:  scalar problem with dropping => record the column indices of undropped entries, but still use original
+        //                                          graph's map information, e.g., whether index is local
+        // OR a matrix without a CrsGraph
+
+        // allocate space for the local graph
+        ArrayRCP<LO> rows(A->getLocalNumRows() + 1);
+        ArrayRCP<LO> columns(A->getLocalNumEntries());
+
+        using MT = typename STS::magnitudeType;
+        RCP<Vector> ghostedDiag;
+        ArrayRCP<const SC> ghostedDiagVals;
+        ArrayRCP<const MT> negMaxOffDiagonal;
+        // RS style needs the max negative off-diagonal, SA style needs the diagonal
+        if (useSignedClassicalRS) {
+          if (ghostedBlockNumber.is_null()) {
+            negMaxOffDiagonal = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixMaxMinusOffDiagonal(*A);
+            if (GetVerbLevel() & Statistics1)
+              GetOStream(Statistics1) << "Calculated max point off-diagonal" << std::endl;
+          } else {
+            negMaxOffDiagonal = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixMaxMinusOffDiagonal(*A, *ghostedBlockNumber);
+            if (GetVerbLevel() & Statistics1)
+              GetOStream(Statistics1) << "Calculating max block off-diagonal" << std::endl;
           }
+        } else {
+          ghostedDiag     = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixOverlappedDiagonal(*A);
+          ghostedDiagVals = ghostedDiag->getData(0);
         }
-        // At this points we either have
-        //     (predrop_ != null)
-        // Therefore, it is sufficient to check only threshold
-        if ( BlockSize==1 && threshold == STS::zero() && !useSignedClassicalRS && !useSignedClassicalSA && A->hasCrsGraph()) {
-          // Case 1:  scalar problem, no dropping => just use matrix graph
-          RCP<GraphBase> graph = rcp(new Graph(A->getCrsGraph(), "graph of A"));
-          // Detect and record rows that correspond to Dirichlet boundary conditions
-          ArrayRCP<bool > boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-          if (rowSumTol > 0.)
+        ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+        if (rowSumTol > 0.) {
+          if (ghostedBlockNumber.is_null()) {
+            if (GetVerbLevel() & Statistics1)
+              GetOStream(Statistics1) << "Applying point row sum criterion." << std::endl;
             Utilities::ApplyRowSumCriterion(*A, rowSumTol, boundaryNodes);
-
-          graph->SetBoundaryNodeMap(boundaryNodes);
-          numTotal = A->getLocalNumEntries();
-
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
-            for (LO i = 0; i < boundaryNodes.size(); ++i)
-              if (boundaryNodes[i])
-                numLocalBoundaryNodes++;
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-          }
-
-          Set(currentLevel, "DofsPerNode", 1);
-          Set(currentLevel, "Graph", graph);
-
-        } else if ( (BlockSize == 1 && threshold != STS::zero()) ||
-                    (BlockSize == 1 && threshold == STS::zero() && !A->hasCrsGraph()) ||
-                    (BlockSize == 1 && useSignedClassicalRS) ||
-                    (BlockSize == 1 && useSignedClassicalSA) ) {	  
-          // Case 2:  scalar problem with dropping => record the column indices of undropped entries, but still use original
-          //                                          graph's map information, e.g., whether index is local
-          // OR a matrix without a CrsGraph
-
-          // allocate space for the local graph
-          ArrayRCP<LO> rows   (A->getLocalNumRows()+1);
-          ArrayRCP<LO> columns(A->getLocalNumEntries());
-
-          using MT = typename STS::magnitudeType;
-          RCP<Vector> ghostedDiag;
-          ArrayRCP<const SC> ghostedDiagVals;
-          ArrayRCP<const MT> negMaxOffDiagonal;
-	  // RS style needs the max negative off-diagonal, SA style needs the diagonal
-          if(useSignedClassicalRS) {
-            if(ghostedBlockNumber.is_null()) {
-              negMaxOffDiagonal = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixMaxMinusOffDiagonal(*A);
-              if (GetVerbLevel() & Statistics1)
-                GetOStream(Statistics1) << "Calculated max point off-diagonal" << std::endl;
-            }
-            else {
-              negMaxOffDiagonal = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixMaxMinusOffDiagonal(*A,*ghostedBlockNumber);
-              if (GetVerbLevel() & Statistics1)
-                GetOStream(Statistics1) << "Calculating max block off-diagonal" << std::endl;
-            }
-          }
-          else {
-            ghostedDiag = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixOverlappedDiagonal(*A);
-            ghostedDiagVals = ghostedDiag->getData(0);
-          }
-          ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-          if (rowSumTol > 0.) {
-            if(ghostedBlockNumber.is_null()) {
-              if (GetVerbLevel() & Statistics1)
-                GetOStream(Statistics1) << "Applying point row sum criterion." << std::endl;
-              Utilities::ApplyRowSumCriterion(*A, rowSumTol, boundaryNodes);
-            } else {
-              if (GetVerbLevel() & Statistics1)
-                GetOStream(Statistics1) << "Applying block row sum criterion." << std::endl;
-              Utilities::ApplyRowSumCriterion(*A, *ghostedBlockNumber, rowSumTol, boundaryNodes);
-            }
+          } else {
+            if (GetVerbLevel() & Statistics1)
+              GetOStream(Statistics1) << "Applying block row sum criterion." << std::endl;
+            Utilities::ApplyRowSumCriterion(*A, *ghostedBlockNumber, rowSumTol, boundaryNodes);
           }
+        }
 
-          LO realnnz = 0;
-          rows[0] = 0;
-          for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
-            size_t nnz = A->getNumEntriesInLocalRow(row);
-            bool rowIsDirichlet = boundaryNodes[row];
-            ArrayView<const LO> indices;
-            ArrayView<const SC> vals;
-            A->getLocalRowView(row, indices, vals);
-
-            if(classicalAlgo == defaultAlgo) {
-              //FIXME the current predrop function uses the following
-              //FIXME    if(std::abs(vals[k]) > std::abs(threshold_) || grow == gcid )
-              //FIXME but the threshold doesn't take into account the rows' diagonal entries
-              //FIXME For now, hardwiring the dropping in here
-              
-              LO rownnz = 0;
-              if(useSignedClassicalRS) {
-                // Signed classical RS style
-                for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
-                  LO col = indices[colID];
-                  MT max_neg_aik = realThreshold * STS::real(negMaxOffDiagonal[row]);
-                  MT neg_aij     = - STS::real(vals[colID]);
-                  /*                  if(row==1326) printf("A(%d,%d) = %6.4e, block = (%d,%d) neg_aij = %6.4e max_neg_aik = %6.4e\n",row,col,vals[colID],
-                                       g_block_id.is_null() ? -1 :  g_block_id[row],
-                                       g_block_id.is_null() ? -1 :  g_block_id[col],
-                                       neg_aij, max_neg_aik);*/
-                  if ((!rowIsDirichlet && (g_block_id.is_null() || g_block_id[row] == g_block_id[col]) && neg_aij > max_neg_aik) || row == col) {
-                    columns[realnnz++] = col;
-                    rownnz++;
-                  } else
-                    numDropped++;
-                }
-                rows[row+1] = realnnz;
+        LO realnnz = 0;
+        rows[0]    = 0;
+        for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
+          size_t nnz          = A->getNumEntriesInLocalRow(row);
+          bool rowIsDirichlet = boundaryNodes[row];
+          ArrayView<const LO> indices;
+          ArrayView<const SC> vals;
+          A->getLocalRowView(row, indices, vals);
+
+          if (classicalAlgo == defaultAlgo) {
+            // FIXME the current predrop function uses the following
+            // FIXME    if(std::abs(vals[k]) > std::abs(threshold_) || grow == gcid )
+            // FIXME but the threshold doesn't take into account the rows' diagonal entries
+            // FIXME For now, hardwiring the dropping in here
+
+            LO rownnz = 0;
+            if (useSignedClassicalRS) {
+              // Signed classical RS style
+              for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
+                LO col         = indices[colID];
+                MT max_neg_aik = realThreshold * STS::real(negMaxOffDiagonal[row]);
+                MT neg_aij     = -STS::real(vals[colID]);
+                /*                  if(row==1326) printf("A(%d,%d) = %6.4e, block = (%d,%d) neg_aij = %6.4e max_neg_aik = %6.4e\n",row,col,vals[colID],
+                                     g_block_id.is_null() ? -1 :  g_block_id[row],
+                                     g_block_id.is_null() ? -1 :  g_block_id[col],
+                                     neg_aij, max_neg_aik);*/
+                if ((!rowIsDirichlet && (g_block_id.is_null() || g_block_id[row] == g_block_id[col]) && neg_aij > max_neg_aik) || row == col) {
+                  columns[realnnz++] = col;
+                  rownnz++;
+                } else
+                  numDropped++;
               }
-              else if(useSignedClassicalSA) {
-		// Signed classical SA style
-                for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
-                  LO col = indices[colID];
+              rows[row + 1] = realnnz;
+            } else if (useSignedClassicalSA) {
+              // Signed classical SA style
+              for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
+                LO col = indices[colID];
 
-		  bool is_nonpositive =  STS::real(vals[colID]) <= 0;
-                  MT  aiiajj = STS::magnitude(threshold*threshold * ghostedDiagVals[col]*ghostedDiagVals[row]);  // eps^2*|a_ii|*|a_jj|
-                  MT aij     = is_nonpositive ? STS::magnitude(vals[colID]*vals[colID]) : (-STS::magnitude(vals[colID]*vals[colID])); // + |a_ij|^2, if a_ij < 0, - |a_ij|^2 if a_ij >=0
-                  /*
-		  if(row==1326) printf("A(%d,%d) = %6.4e, raw_aij = %6.4e aij = %6.4e aiiajj = %6.4e\n",row,col,vals[colID],
-                                       vals[colID],aij, aiiajj);
-		  */
-
-                  if ((!rowIsDirichlet && aij > aiiajj) || row == col) {
-		    columns[realnnz++] = col;
-		    rownnz++;
-                  } else
-                    numDropped++;
-                }
-                rows[row+1] = realnnz;
+                bool is_nonpositive = STS::real(vals[colID]) <= 0;
+                MT aiiajj           = STS::magnitude(threshold * threshold * ghostedDiagVals[col] * ghostedDiagVals[row]);                        // eps^2*|a_ii|*|a_jj|
+                MT aij              = is_nonpositive ? STS::magnitude(vals[colID] * vals[colID]) : (-STS::magnitude(vals[colID] * vals[colID]));  // + |a_ij|^2, if a_ij < 0, - |a_ij|^2 if a_ij >=0
+                /*
+                if(row==1326) printf("A(%d,%d) = %6.4e, raw_aij = %6.4e aij = %6.4e aiiajj = %6.4e\n",row,col,vals[colID],
+                                     vals[colID],aij, aiiajj);
+                */
+
+                if ((!rowIsDirichlet && aij > aiiajj) || row == col) {
+                  columns[realnnz++] = col;
+                  rownnz++;
+                } else
+                  numDropped++;
               }
-              else {
-                // Standard abs classical
-                for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
-                  LO col = indices[colID];
-                  MT  aiiajj = STS::magnitude(threshold*threshold * ghostedDiagVals[col]*ghostedDiagVals[row]);  // eps^2*|a_ii|*|a_jj|
-                  MT aij    = STS::magnitude(vals[colID]*vals[colID]);                                          // |a_ij|^2
-                  
-                  if ((!rowIsDirichlet && aij > aiiajj) || row == col) {
+              rows[row + 1] = realnnz;
+            } else {
+              // Standard abs classical
+              for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
+                LO col    = indices[colID];
+                MT aiiajj = STS::magnitude(threshold * threshold * ghostedDiagVals[col] * ghostedDiagVals[row]);  // eps^2*|a_ii|*|a_jj|
+                MT aij    = STS::magnitude(vals[colID] * vals[colID]);                                            // |a_ij|^2
+
+                if ((!rowIsDirichlet && aij > aiiajj) || row == col) {
                   columns[realnnz++] = col;
                   rownnz++;
-                  } else
-                    numDropped++;
-                }
-                rows[row+1] = realnnz;
+                } else
+                  numDropped++;
               }
+              rows[row + 1] = realnnz;
             }
-            else {
-              /* Cut Algorithm */
-              //CMS
-              using DropTol = Details::DropTol<real_type,LO>;
-              std::vector<DropTol> drop_vec;
-              drop_vec.reserve(nnz);
-              const real_type zero = Teuchos::ScalarTraits<real_type>::zero();
-              const real_type one  = Teuchos::ScalarTraits<real_type>::one();
-              LO rownnz = 0;
-              // NOTE: This probably needs to be fixed for rowsum
-
-              // find magnitudes
-              for (LO colID = 0; colID < (LO)nnz; colID++) {
-                LO col = indices[colID];
-                if (row == col) {
-                  drop_vec.emplace_back( zero, one, colID, false);
-                  continue;
-                }
-                
-                // Don't aggregate boundaries
-                if(boundaryNodes[colID]) continue;
-                typename STS::magnitudeType aiiajj = STS::magnitude(threshold*threshold * ghostedDiagVals[col]*ghostedDiagVals[row]);  // eps^2*|a_ii|*|a_jj|
-                typename STS::magnitudeType aij    = STS::magnitude(vals[colID]*vals[colID]);                                          // |a_ij|^2
-                drop_vec.emplace_back(aij, aiiajj, colID, false);
+          } else {
+            /* Cut Algorithm */
+            // CMS
+            using DropTol = Details::DropTol<real_type, LO>;
+            std::vector<DropTol> drop_vec;
+            drop_vec.reserve(nnz);
+            const real_type zero = Teuchos::ScalarTraits<real_type>::zero();
+            const real_type one  = Teuchos::ScalarTraits<real_type>::one();
+            LO rownnz            = 0;
+            // NOTE: This probably needs to be fixed for rowsum
+
+            // find magnitudes
+            for (LO colID = 0; colID < (LO)nnz; colID++) {
+              LO col = indices[colID];
+              if (row == col) {
+                drop_vec.emplace_back(zero, one, colID, false);
+                continue;
               }
 
-              const size_t n = drop_vec.size();
-              
-              if (classicalAlgo == unscaled_cut) {
-                std::sort( drop_vec.begin(), drop_vec.end()
-                           , [](DropTol const& a, DropTol const& b) {
-                             return a.val > b.val;
-                           });
-                
-                bool drop = false;
-                for (size_t i=1; i<n; ++i) {
-                  if (!drop) {
-                    auto const& x = drop_vec[i-1];
-                    auto const& y = drop_vec[i];
-                    auto a = x.val;
-                    auto b = y.val;
-                    if (a > realThreshold*b) {
-                      drop = true;
+              // Don't aggregate boundaries
+              if (boundaryNodes[colID]) continue;
+              typename STS::magnitudeType aiiajj = STS::magnitude(threshold * threshold * ghostedDiagVals[col] * ghostedDiagVals[row]);  // eps^2*|a_ii|*|a_jj|
+              typename STS::magnitudeType aij    = STS::magnitude(vals[colID] * vals[colID]);                                            // |a_ij|^2
+              drop_vec.emplace_back(aij, aiiajj, colID, false);
+            }
+
+            const size_t n = drop_vec.size();
+
+            if (classicalAlgo == unscaled_cut) {
+              std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+                return a.val > b.val;
+              });
+
+              bool drop = false;
+              for (size_t i = 1; i < n; ++i) {
+                if (!drop) {
+                  auto const& x = drop_vec[i - 1];
+                  auto const& y = drop_vec[i];
+                  auto a        = x.val;
+                  auto b        = y.val;
+                  if (a > realThreshold * b) {
+                    drop = true;
 #ifdef HAVE_MUELU_DEBUG
-                      if (distanceLaplacianCutVerbose) {
-                        std::cout << "DJS: KEEP, N, ROW:  " << i+1 << ", " << n << ", " << row << std::endl;
-                      }
-#endif
+                    if (distanceLaplacianCutVerbose) {
+                      std::cout << "DJS: KEEP, N, ROW:  " << i + 1 << ", " << n << ", " << row << std::endl;
                     }
+#endif
                   }
-                  drop_vec[i].drop = drop;
                 }
-              } else if (classicalAlgo == scaled_cut) {
-                  std::sort( drop_vec.begin(), drop_vec.end()
-                           , [](DropTol const& a, DropTol const& b) {
-                               return a.val/a.diag > b.val/b.diag;
-                             });
-                  bool drop = false;
-                  //                  printf("[%d] Scaled Cut: ",(int)row);
-                  //                  printf("%3d(%4s) ",indices[drop_vec[0].col],"keep");
-                  for (size_t i=1; i<n; ++i) {
-                    if (!drop) {
-                      auto const& x = drop_vec[i-1];
-                      auto const& y = drop_vec[i];
-                      auto a = x.val/x.diag;
-                      auto b = y.val/y.diag;
-                      if (a > realThreshold*b) {
-                        drop = true;
+                drop_vec[i].drop = drop;
+              }
+            } else if (classicalAlgo == scaled_cut) {
+              std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+                return a.val / a.diag > b.val / b.diag;
+              });
+              bool drop = false;
+              //                  printf("[%d] Scaled Cut: ",(int)row);
+              //                  printf("%3d(%4s) ",indices[drop_vec[0].col],"keep");
+              for (size_t i = 1; i < n; ++i) {
+                if (!drop) {
+                  auto const& x = drop_vec[i - 1];
+                  auto const& y = drop_vec[i];
+                  auto a        = x.val / x.diag;
+                  auto b        = y.val / y.diag;
+                  if (a > realThreshold * b) {
+                    drop = true;
 
 #ifdef HAVE_MUELU_DEBUG
-                        if (distanceLaplacianCutVerbose) {
-                          std::cout << "DJS: KEEP, N, ROW:  " << i+1 << ", " << n << ", " << row << std::endl;
-                        }
-#endif
-                      }
-                      //                      printf("%3d(%4s) ",indices[drop_vec[i].col],drop?"drop":"keep");
-
+                    if (distanceLaplacianCutVerbose) {
+                      std::cout << "DJS: KEEP, N, ROW:  " << i + 1 << ", " << n << ", " << row << std::endl;
                     }
-                    drop_vec[i].drop = drop;
+#endif
                   }
-                  //                  printf("\n");
-              }
-              std::sort( drop_vec.begin(), drop_vec.end()
-                         , [](DropTol const& a, DropTol const& b) {
-                           return a.col < b.col;
-                         }
-                         );
-              
-              for (LO idxID =0; idxID<(LO)drop_vec.size(); idxID++) {
-                LO col = indices[drop_vec[idxID].col];
-                // don't drop diagonal
-                if (row == col) {
-                  columns[realnnz++] = col;
-                  rownnz++;
-                  continue;
-                }
-                
-                if (!drop_vec[idxID].drop) {
-                  columns[realnnz++] = col;
-                  rownnz++;
-                } else {
-                  numDropped++;
+                  //                      printf("%3d(%4s) ",indices[drop_vec[i].col],drop?"drop":"keep");
                 }
+                drop_vec[i].drop = drop;
               }
-              // CMS
-              rows[row+1] = realnnz;
-              
+              //                  printf("\n");
             }
-          }//end for row
-
-          columns.resize(realnnz);
-          numTotal = A->getLocalNumEntries();
+            std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+              return a.col < b.col;
+            });
+
+            for (LO idxID = 0; idxID < (LO)drop_vec.size(); idxID++) {
+              LO col = indices[drop_vec[idxID].col];
+              // don't drop diagonal
+              if (row == col) {
+                columns[realnnz++] = col;
+                rownnz++;
+                continue;
+              }
 
-          if (aggregationMayCreateDirichlet) {
-            // If the only element remaining after filtering is diagonal, mark node as boundary
-            for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
-              if (rows[row+1]- rows[row] <= 1)
-                boundaryNodes[row] = true;
+              if (!drop_vec[idxID].drop) {
+                columns[realnnz++] = col;
+                rownnz++;
+              } else {
+                numDropped++;
+              }
             }
+            // CMS
+            rows[row + 1] = realnnz;
           }
+        }  // end for row
 
-          RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, A->getRowMap(), A->getColMap(), "thresholded graph of A"));
-          graph->SetBoundaryNodeMap(boundaryNodes);
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
-            for (LO i = 0; i < boundaryNodes.size(); ++i)
-              if (boundaryNodes[i])
-                numLocalBoundaryNodes++;
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+        columns.resize(realnnz);
+        numTotal = A->getLocalNumEntries();
+
+        if (aggregationMayCreateDirichlet) {
+          // If the only element remaining after filtering is diagonal, mark node as boundary
+          for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
+            if (rows[row + 1] - rows[row] <= 1)
+              boundaryNodes[row] = true;
           }
-          Set(currentLevel, "Graph",       graph);
-          Set(currentLevel, "DofsPerNode", 1);
-
-          // If we're doing signed classical, we might want to block-diagonalize *after* the dropping
-          if(generateColoringGraph) {
-            RCP<GraphBase> colorGraph;
-            RCP<const Import> importer = A->getCrsGraph()->getImporter();
-            BlockDiagonalizeGraph(graph,ghostedBlockNumber,colorGraph,importer);
-            Set(currentLevel, "Coloring Graph",colorGraph);
-            // #define CMS_DUMP
+        }
+
+        RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, A->getRowMap(), A->getColMap(), "thresholded graph of A"));
+        graph->SetBoundaryNodeMap(boundaryNodes);
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
+          for (LO i = 0; i < boundaryNodes.size(); ++i)
+            if (boundaryNodes[i])
+              numLocalBoundaryNodes++;
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+        }
+        Set(currentLevel, "Graph", graph);
+        Set(currentLevel, "DofsPerNode", 1);
+
+        // If we're doing signed classical, we might want to block-diagonalize *after* the dropping
+        if (generateColoringGraph) {
+          RCP<GraphBase> colorGraph;
+          RCP<const Import> importer = A->getCrsGraph()->getImporter();
+          BlockDiagonalizeGraph(graph, ghostedBlockNumber, colorGraph, importer);
+          Set(currentLevel, "Coloring Graph", colorGraph);
+          // #define CMS_DUMP
 #ifdef CMS_DUMP
-           {
-             Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("m_regular_graph."+std::to_string(currentLevel.GetLevelID()), *rcp_dynamic_cast<LWGraph>(graph)->GetCrsGraph());
-             Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("m_color_graph."+std::to_string(currentLevel.GetLevelID()), *rcp_dynamic_cast<LWGraph>(colorGraph)->GetCrsGraph());
-             // int rank = graph->GetDomainMap()->getComm()->getRank();
-             // {
-             //   std::ofstream ofs(std::string("m_color_graph_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
-             //   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
-             //   colorGraph->print(*fancy,Debug);
-             // }
-             // {
-             //   std::ofstream ofs(std::string("m_regular_graph_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
-             //   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
-             //   graph->print(*fancy,Debug);
-             // }
-             
-           }
+          {
+            Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("m_regular_graph." + std::to_string(currentLevel.GetLevelID()), *rcp_dynamic_cast<LWGraph>(graph)->GetCrsGraph());
+            Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("m_color_graph." + std::to_string(currentLevel.GetLevelID()), *rcp_dynamic_cast<LWGraph>(colorGraph)->GetCrsGraph());
+            // int rank = graph->GetDomainMap()->getComm()->getRank();
+            // {
+            //   std::ofstream ofs(std::string("m_color_graph_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
+            //   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
+            //   colorGraph->print(*fancy,Debug);
+            // }
+            // {
+            //   std::ofstream ofs(std::string("m_regular_graph_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
+            //   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
+            //   graph->print(*fancy,Debug);
+            // }
+          }
 #endif
-          }//end generateColoringGraph
-        } else if (BlockSize > 1 && threshold == STS::zero()) {
-          // Case 3:  Multiple DOF/node problem without dropping
-          const RCP<const Map> rowMap = A->getRowMap();
-          const RCP<const Map> colMap = A->getColMap();
+        }  // end generateColoringGraph
+      } else if (BlockSize > 1 && threshold == STS::zero()) {
+        // Case 3:  Multiple DOF/node problem without dropping
+        const RCP<const Map> rowMap = A->getRowMap();
+        const RCP<const Map> colMap = A->getColMap();
 
-          graphType = "amalgamated";
+        graphType = "amalgamated";
 
-          // build node row map (uniqueMap) and node column map (nonUniqueMap)
-          // the arrays rowTranslation and colTranslation contain the local node id
-          // given a local dof id. The data is calculated by the AmalgamationFactory and
-          // stored in the variable container "UnAmalgamationInfo"
-          RCP<const Map> uniqueMap = amalInfo->getNodeRowMap();
-          RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
-          Array<LO> rowTranslation = *(amalInfo->getRowTranslation());
-          Array<LO> colTranslation = *(amalInfo->getColTranslation());
-
-          // get number of local nodes
-          LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
-
-          // Allocate space for the local graph
-          ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows+1);
-          ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
-
-          const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
-
-          // Detect and record rows that correspond to Dirichlet boundary conditions
-          // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
-          // TODO the array one bigger than the number of local rows, and the last entry can
-          // TODO hold the actual number of boundary nodes.  Clever, huh?
-          ArrayRCP<bool > pointBoundaryNodes;
-          pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-          if (rowSumTol > 0.)
-            Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
-
-
-          // extract striding information
-          LO blkSize = A->GetFixedBlockSize();     //< the full block size (number of dofs per node in strided map)
-          LO blkId   = -1;                         //< the block id within the strided map (or -1 if it is a full block map)
-          LO blkPartSize = A->GetFixedBlockSize(); //< stores the size of the block within the strided map
-          if (A->IsView("stridedMaps") == true) {
-            Teuchos::RCP<const Map> myMap = A->getRowMap("stridedMaps");
-            Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-            TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-            blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
-            blkId   = strMap->getStridedBlockId();
-            if (blkId > -1)
-              blkPartSize = Teuchos::as<LO>(strMap->getStridingData()[blkId]);
-          }
+        // build node row map (uniqueMap) and node column map (nonUniqueMap)
+        // the arrays rowTranslation and colTranslation contain the local node id
+        // given a local dof id. The data is calculated by the AmalgamationFactory and
+        // stored in the variable container "UnAmalgamationInfo"
+        RCP<const Map> uniqueMap    = amalInfo->getNodeRowMap();
+        RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
+        Array<LO> rowTranslation    = *(amalInfo->getRowTranslation());
+        Array<LO> colTranslation    = *(amalInfo->getColTranslation());
 
-          // loop over all local nodes
-          LO realnnz = 0;
-          rows[0] = 0;
-          Array<LO> indicesExtra;
-          for (LO row = 0; row < numRows; row++) {
-            ArrayView<const LO> indices;
-            indicesExtra.resize(0);
+        // get number of local nodes
+        LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
 
-            // The amalgamated row is marked as Dirichlet iff all point rows are Dirichlet
-            // Note, that pointBoundaryNodes lives on the dofmap (and not the node map).
-            // Therefore, looping over all dofs is fine here. We use blkPartSize as we work
-            // with local ids.
-            // TODO: Here we have different options of how to define a node to be a boundary (or Dirichlet)
-            // node.
-            bool isBoundary = false;
-            if (pL.get<bool>("aggregation: greedy Dirichlet") == true) {
-              for (LO j = 0; j < blkPartSize; j++) {
-                if (pointBoundaryNodes[row*blkPartSize+j]) {
-                  isBoundary = true;
-                  break;
-                }
+        // Allocate space for the local graph
+        ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows + 1);
+        ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
+
+        const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
+
+        // Detect and record rows that correspond to Dirichlet boundary conditions
+        // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
+        // TODO the array one bigger than the number of local rows, and the last entry can
+        // TODO hold the actual number of boundary nodes.  Clever, huh?
+        ArrayRCP<bool> pointBoundaryNodes;
+        pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+        if (rowSumTol > 0.)
+          Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
+
+        // extract striding information
+        LO blkSize     = A->GetFixedBlockSize();  //< the full block size (number of dofs per node in strided map)
+        LO blkId       = -1;                      //< the block id within the strided map (or -1 if it is a full block map)
+        LO blkPartSize = A->GetFixedBlockSize();  //< stores the size of the block within the strided map
+        if (A->IsView("stridedMaps") == true) {
+          Teuchos::RCP<const Map> myMap         = A->getRowMap("stridedMaps");
+          Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+          TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+          blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
+          blkId   = strMap->getStridedBlockId();
+          if (blkId > -1)
+            blkPartSize = Teuchos::as<LO>(strMap->getStridingData()[blkId]);
+        }
+
+        // loop over all local nodes
+        LO realnnz = 0;
+        rows[0]    = 0;
+        Array<LO> indicesExtra;
+        for (LO row = 0; row < numRows; row++) {
+          ArrayView<const LO> indices;
+          indicesExtra.resize(0);
+
+          // The amalgamated row is marked as Dirichlet iff all point rows are Dirichlet
+          // Note, that pointBoundaryNodes lives on the dofmap (and not the node map).
+          // Therefore, looping over all dofs is fine here. We use blkPartSize as we work
+          // with local ids.
+          // TODO: Here we have different options of how to define a node to be a boundary (or Dirichlet)
+          // node.
+          bool isBoundary = false;
+          if (pL.get<bool>("aggregation: greedy Dirichlet") == true) {
+            for (LO j = 0; j < blkPartSize; j++) {
+              if (pointBoundaryNodes[row * blkPartSize + j]) {
+                isBoundary = true;
+                break;
               }
-            } else {
-              isBoundary = true;
-              for (LO j = 0; j < blkPartSize; j++) {
-                if (!pointBoundaryNodes[row*blkPartSize+j]) {
-                  isBoundary = false;
-                  break;
-                }
+            }
+          } else {
+            isBoundary = true;
+            for (LO j = 0; j < blkPartSize; j++) {
+              if (!pointBoundaryNodes[row * blkPartSize + j]) {
+                isBoundary = false;
+                break;
               }
             }
+          }
 
-            // Merge rows of A
-            // The array indicesExtra contains local column node ids for the current local node "row"
-            if (!isBoundary)
-              MergeRows(*A, row, indicesExtra, colTranslation);
-            else
-              indicesExtra.push_back(row);
-            indices   = indicesExtra;
-            numTotal += indices.size();
+          // Merge rows of A
+          // The array indicesExtra contains local column node ids for the current local node "row"
+          if (!isBoundary)
+            MergeRows(*A, row, indicesExtra, colTranslation);
+          else
+            indicesExtra.push_back(row);
+          indices = indicesExtra;
+          numTotal += indices.size();
+
+          // add the local column node ids to the full columns array which
+          // contains the local column node ids for all local node rows
+          LO nnz = indices.size(), rownnz = 0;
+          for (LO colID = 0; colID < nnz; colID++) {
+            LO col             = indices[colID];
+            columns[realnnz++] = col;
+            rownnz++;
+          }
 
-            // add the local column node ids to the full columns array which
-            // contains the local column node ids for all local node rows
-            LO nnz = indices.size(), rownnz = 0;
-            for (LO colID = 0; colID < nnz; colID++) {
-              LO col = indices[colID];
-              columns[realnnz++] = col;
-              rownnz++;
-            }
+          if (rownnz == 1) {
+            // If the only element remaining after filtering is diagonal, mark node as boundary
+            // FIXME: this should really be replaced by the following
+            //    if (indices.size() == 1 && indices[0] == row)
+            //        boundaryNodes[row] = true;
+            // We do not do it this way now because there is no framework for distinguishing isolated
+            // and boundary nodes in the aggregation algorithms
+            amalgBoundaryNodes[row] = true;
+          }
+          rows[row + 1] = realnnz;
+        }  // for (LO row = 0; row < numRows; row++)
+        columns.resize(realnnz);
 
-            if (rownnz == 1) {
-              // If the only element remaining after filtering is diagonal, mark node as boundary
-              // FIXME: this should really be replaced by the following
-              //    if (indices.size() == 1 && indices[0] == row)
-              //        boundaryNodes[row] = true;
-              // We do not do it this way now because there is no framework for distinguishing isolated
-              // and boundary nodes in the aggregation algorithms
-              amalgBoundaryNodes[row] = true;
-            }
-            rows[row+1] = realnnz;
-          } //for (LO row = 0; row < numRows; row++)
-          columns.resize(realnnz);
+        RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
+        graph->SetBoundaryNodeMap(amalgBoundaryNodes);
 
-          RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
-          graph->SetBoundaryNodeMap(amalgBoundaryNodes);
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
 
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
+          for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
+            if (amalgBoundaryNodes[i])
+              numLocalBoundaryNodes++;
 
-            for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
-              if (amalgBoundaryNodes[i])
-                numLocalBoundaryNodes++;
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes
+                                  << " agglomerated Dirichlet nodes" << std::endl;
+        }
 
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes
-                                       << " agglomerated Dirichlet nodes" << std::endl;
-          }
+        Set(currentLevel, "Graph", graph);
+        Set(currentLevel, "DofsPerNode", blkSize);  // full block size
 
-          Set(currentLevel, "Graph",       graph);
-          Set(currentLevel, "DofsPerNode", blkSize); // full block size
+      } else if (BlockSize > 1 && threshold != STS::zero()) {
+        // Case 4:  Multiple DOF/node problem with dropping
+        const RCP<const Map> rowMap = A->getRowMap();
+        const RCP<const Map> colMap = A->getColMap();
+        graphType                   = "amalgamated";
 
-        } else if (BlockSize > 1 && threshold != STS::zero()) {
-          // Case 4:  Multiple DOF/node problem with dropping
-          const RCP<const Map> rowMap = A->getRowMap();
-          const RCP<const Map> colMap = A->getColMap();
-          graphType = "amalgamated";
+        // build node row map (uniqueMap) and node column map (nonUniqueMap)
+        // the arrays rowTranslation and colTranslation contain the local node id
+        // given a local dof id. The data is calculated by the AmalgamationFactory and
+        // stored in the variable container "UnAmalgamationInfo"
+        RCP<const Map> uniqueMap    = amalInfo->getNodeRowMap();
+        RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
+        Array<LO> rowTranslation    = *(amalInfo->getRowTranslation());
+        Array<LO> colTranslation    = *(amalInfo->getColTranslation());
 
-          // build node row map (uniqueMap) and node column map (nonUniqueMap)
-          // the arrays rowTranslation and colTranslation contain the local node id
-          // given a local dof id. The data is calculated by the AmalgamationFactory and
-          // stored in the variable container "UnAmalgamationInfo"
-          RCP<const Map> uniqueMap = amalInfo->getNodeRowMap();
-          RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
-          Array<LO> rowTranslation = *(amalInfo->getRowTranslation());
-          Array<LO> colTranslation = *(amalInfo->getColTranslation());
-
-          // get number of local nodes
-          LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
-
-          // Allocate space for the local graph
-          ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows+1);
-          ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
-
-          const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
-
-          // Detect and record rows that correspond to Dirichlet boundary conditions
-          // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
-          // TODO the array one bigger than the number of local rows, and the last entry can
-          // TODO hold the actual number of boundary nodes.  Clever, huh?
-          ArrayRCP<bool > pointBoundaryNodes;
-          pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-          if (rowSumTol > 0.)
-            Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
-
-
-          // extract striding information
-          LO blkSize = A->GetFixedBlockSize();     //< the full block size (number of dofs per node in strided map)
-          LO blkId   = -1;                         //< the block id within the strided map (or -1 if it is a full block map)
-          LO blkPartSize = A->GetFixedBlockSize(); //< stores the size of the block within the strided map
-          if (A->IsView("stridedMaps") == true) {
-            Teuchos::RCP<const Map> myMap = A->getRowMap("stridedMaps");
-            Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-            TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-            blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
-            blkId   = strMap->getStridedBlockId();
-            if (blkId > -1)
-              blkPartSize = Teuchos::as<LO>(strMap->getStridingData()[blkId]);
-          }
+        // get number of local nodes
+        LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
 
-          // extract diagonal data for dropping strategy
-          RCP<Vector> ghostedDiag = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixOverlappedDiagonal(*A);
-          const ArrayRCP<const SC> ghostedDiagVals = ghostedDiag->getData(0);
+        // Allocate space for the local graph
+        ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows + 1);
+        ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
 
-          // loop over all local nodes
-          LO realnnz = 0;
-          rows[0] = 0;
-          Array<LO> indicesExtra;
-          for (LO row = 0; row < numRows; row++) {
-            ArrayView<const LO> indices;
-            indicesExtra.resize(0);
+        const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
 
-            // The amalgamated row is marked as Dirichlet iff all point rows are Dirichlet
-            // Note, that pointBoundaryNodes lives on the dofmap (and not the node map).
-            // Therefore, looping over all dofs is fine here. We use blkPartSize as we work
-            // with local ids.
-            // TODO: Here we have different options of how to define a node to be a boundary (or Dirichlet)
-            // node.
-            bool isBoundary = false;
-            if (pL.get<bool>("aggregation: greedy Dirichlet") == true) {
-              for (LO j = 0; j < blkPartSize; j++) {
-                if (pointBoundaryNodes[row*blkPartSize+j]) {
-                  isBoundary = true;
-                  break;
-                }
-              }
-            } else {
-              isBoundary = true;
-              for (LO j = 0; j < blkPartSize; j++) {
-                if (!pointBoundaryNodes[row*blkPartSize+j]) {
-                  isBoundary = false;
-                  break;
-                }
-              }
-            }
+        // Detect and record rows that correspond to Dirichlet boundary conditions
+        // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
+        // TODO the array one bigger than the number of local rows, and the last entry can
+        // TODO hold the actual number of boundary nodes.  Clever, huh?
+        ArrayRCP<bool> pointBoundaryNodes;
+        pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+        if (rowSumTol > 0.)
+          Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
 
-            // Merge rows of A
-            // The array indicesExtra contains local column node ids for the current local node "row"
-            if (!isBoundary)
-              MergeRowsWithDropping(*A, row, ghostedDiagVals, threshold, indicesExtra, colTranslation);
-            else
-              indicesExtra.push_back(row);
-            indices   = indicesExtra;
-            numTotal += indices.size();
+        // extract striding information
+        LO blkSize     = A->GetFixedBlockSize();  //< the full block size (number of dofs per node in strided map)
+        LO blkId       = -1;                      //< the block id within the strided map (or -1 if it is a full block map)
+        LO blkPartSize = A->GetFixedBlockSize();  //< stores the size of the block within the strided map
+        if (A->IsView("stridedMaps") == true) {
+          Teuchos::RCP<const Map> myMap         = A->getRowMap("stridedMaps");
+          Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+          TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+          blkSize = Teuchos::as<const LO>(strMap->getFixedBlockSize());
+          blkId   = strMap->getStridedBlockId();
+          if (blkId > -1)
+            blkPartSize = Teuchos::as<LO>(strMap->getStridingData()[blkId]);
+        }
 
-            // add the local column node ids to the full columns array which
-            // contains the local column node ids for all local node rows
-            LO nnz = indices.size(), rownnz = 0;
-            for (LO colID = 0; colID < nnz; colID++) {
-              LO col = indices[colID];
-              columns[realnnz++] = col;
-              rownnz++;
-            }
+        // extract diagonal data for dropping strategy
+        RCP<Vector> ghostedDiag                  = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixOverlappedDiagonal(*A);
+        const ArrayRCP<const SC> ghostedDiagVals = ghostedDiag->getData(0);
 
-            if (rownnz == 1) {
-              // If the only element remaining after filtering is diagonal, mark node as boundary
-              // FIXME: this should really be replaced by the following
-              //    if (indices.size() == 1 && indices[0] == row)
-              //        boundaryNodes[row] = true;
-              // We do not do it this way now because there is no framework for distinguishing isolated
-              // and boundary nodes in the aggregation algorithms
-              amalgBoundaryNodes[row] = true;
+        // loop over all local nodes
+        LO realnnz = 0;
+        rows[0]    = 0;
+        Array<LO> indicesExtra;
+        for (LO row = 0; row < numRows; row++) {
+          ArrayView<const LO> indices;
+          indicesExtra.resize(0);
+
+          // The amalgamated row is marked as Dirichlet iff all point rows are Dirichlet
+          // Note, that pointBoundaryNodes lives on the dofmap (and not the node map).
+          // Therefore, looping over all dofs is fine here. We use blkPartSize as we work
+          // with local ids.
+          // TODO: Here we have different options of how to define a node to be a boundary (or Dirichlet)
+          // node.
+          bool isBoundary = false;
+          if (pL.get<bool>("aggregation: greedy Dirichlet") == true) {
+            for (LO j = 0; j < blkPartSize; j++) {
+              if (pointBoundaryNodes[row * blkPartSize + j]) {
+                isBoundary = true;
+                break;
+              }
             }
-            rows[row+1] = realnnz;
-          } //for (LO row = 0; row < numRows; row++)
-          columns.resize(realnnz);
+          } else {
+            isBoundary = true;
+            for (LO j = 0; j < blkPartSize; j++) {
+              if (!pointBoundaryNodes[row * blkPartSize + j]) {
+                isBoundary = false;
+                break;
+              }
+            }
+          }
 
-          RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
-          graph->SetBoundaryNodeMap(amalgBoundaryNodes);
+          // Merge rows of A
+          // The array indicesExtra contains local column node ids for the current local node "row"
+          if (!isBoundary)
+            MergeRowsWithDropping(*A, row, ghostedDiagVals, threshold, indicesExtra, colTranslation);
+          else
+            indicesExtra.push_back(row);
+          indices = indicesExtra;
+          numTotal += indices.size();
+
+          // add the local column node ids to the full columns array which
+          // contains the local column node ids for all local node rows
+          LO nnz = indices.size(), rownnz = 0;
+          for (LO colID = 0; colID < nnz; colID++) {
+            LO col             = indices[colID];
+            columns[realnnz++] = col;
+            rownnz++;
+          }
+
+          if (rownnz == 1) {
+            // If the only element remaining after filtering is diagonal, mark node as boundary
+            // FIXME: this should really be replaced by the following
+            //    if (indices.size() == 1 && indices[0] == row)
+            //        boundaryNodes[row] = true;
+            // We do not do it this way now because there is no framework for distinguishing isolated
+            // and boundary nodes in the aggregation algorithms
+            amalgBoundaryNodes[row] = true;
+          }
+          rows[row + 1] = realnnz;
+        }  // for (LO row = 0; row < numRows; row++)
+        columns.resize(realnnz);
 
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
+        RCP<GraphBase> graph = rcp(new LWGraph(rows, columns, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
+        graph->SetBoundaryNodeMap(amalgBoundaryNodes);
 
-            for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
-              if (amalgBoundaryNodes[i])
-                numLocalBoundaryNodes++;
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
 
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes
-                                       << " agglomerated Dirichlet nodes" << std::endl;
-          }
+          for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
+            if (amalgBoundaryNodes[i])
+              numLocalBoundaryNodes++;
 
-          Set(currentLevel, "Graph",       graph);
-          Set(currentLevel, "DofsPerNode", blkSize); // full block size
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes
+                                  << " agglomerated Dirichlet nodes" << std::endl;
         }
 
-      } else if (algo == "distance laplacian") {
-        LO blkSize   = A->GetFixedBlockSize();
-        GO indexBase = A->getRowMap()->getIndexBase();
-        // [*0*] : FIXME
-        // ap: somehow, if I move this line to [*1*], Belos throws an error
-        // I'm not sure what's going on. Do we always have to Get data, if we did
-        // DeclareInput for it?
-        //        RCP<RealValuedMultiVector> Coords = Get< RCP<RealValuedMultiVector > >(currentLevel, "Coordinates");
+        Set(currentLevel, "Graph", graph);
+        Set(currentLevel, "DofsPerNode", blkSize);  // full block size
+      }
 
-        // Detect and record rows that correspond to Dirichlet boundary conditions
-        // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
-        // TODO the array one bigger than the number of local rows, and the last entry can
-        // TODO hold the actual number of boundary nodes.  Clever, huh?
-        ArrayRCP<bool > pointBoundaryNodes;
-        pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-        if (rowSumTol > 0.)
-          Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
+    } else if (algo == "distance laplacian") {
+      LO blkSize   = A->GetFixedBlockSize();
+      GO indexBase = A->getRowMap()->getIndexBase();
+      // [*0*] : FIXME
+      // ap: somehow, if I move this line to [*1*], Belos throws an error
+      // I'm not sure what's going on. Do we always have to Get data, if we did
+      // DeclareInput for it?
+      //        RCP<RealValuedMultiVector> Coords = Get< RCP<RealValuedMultiVector > >(currentLevel, "Coordinates");
 
-        if ( (blkSize == 1) && (threshold == STS::zero()) ) {
-          // Trivial case: scalar problem, no dropping. Can return original graph
-          RCP<GraphBase> graph = rcp(new Graph(A->getCrsGraph(), "graph of A"));
-          graph->SetBoundaryNodeMap(pointBoundaryNodes);
-          graphType="unamalgamated";
-          numTotal = A->getLocalNumEntries();
-
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
-            for (LO i = 0; i < pointBoundaryNodes.size(); ++i)
-              if (pointBoundaryNodes[i])
-                numLocalBoundaryNodes++;
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-          }
+      // Detect and record rows that correspond to Dirichlet boundary conditions
+      // TODO If we use ArrayRCP<LO>, then we can record boundary nodes as usual.  Size
+      // TODO the array one bigger than the number of local rows, and the last entry can
+      // TODO hold the actual number of boundary nodes.  Clever, huh?
+      ArrayRCP<bool> pointBoundaryNodes;
+      pointBoundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+      if (rowSumTol > 0.)
+        Utilities::ApplyRowSumCriterion(*A, rowSumTol, pointBoundaryNodes);
+
+      if ((blkSize == 1) && (threshold == STS::zero())) {
+        // Trivial case: scalar problem, no dropping. Can return original graph
+        RCP<GraphBase> graph = rcp(new Graph(A->getCrsGraph(), "graph of A"));
+        graph->SetBoundaryNodeMap(pointBoundaryNodes);
+        graphType = "unamalgamated";
+        numTotal  = A->getLocalNumEntries();
+
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
+          for (LO i = 0; i < pointBoundaryNodes.size(); ++i)
+            if (pointBoundaryNodes[i])
+              numLocalBoundaryNodes++;
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+        }
 
-          Set(currentLevel, "DofsPerNode", blkSize);
-          Set(currentLevel, "Graph",       graph);
+        Set(currentLevel, "DofsPerNode", blkSize);
+        Set(currentLevel, "Graph", graph);
+
+      } else {
+        // ap: We make quite a few assumptions here; general case may be a lot different,
+        // but much much harder to implement. We assume that:
+        //  1) all maps are standard maps, not strided maps
+        //  2) global indices of dofs in A are related to dofs in coordinates in a simple arithmetic
+        //     way: rows i*blkSize, i*blkSize+1, ..., i*blkSize + (blkSize-1) correspond to node i
+        //
+        // NOTE: Potentially, some of the code below could be simplified with UnAmalgamationInfo,
+        // but as I totally don't understand that code, here is my solution
+
+        // [*1*]: see [*0*]
+
+        // Check that the number of local coordinates is consistent with the #rows in A
+        TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getLocalNumElements() / blkSize != Coords->getLocalLength(), Exceptions::Incompatible,
+                                   "Coordinate vector length (" << Coords->getLocalLength() << ") is incompatible with number of rows in A (" << A->getRowMap()->getLocalNumElements() << ") by modulo block size (" << blkSize << ").");
+
+        const RCP<const Map> colMap = A->getColMap();
+        RCP<const Map> uniqueMap, nonUniqueMap;
+        Array<LO> colTranslation;
+        if (blkSize == 1) {
+          uniqueMap    = A->getRowMap();
+          nonUniqueMap = A->getColMap();
+          graphType    = "unamalgamated";
 
         } else {
-          // ap: We make quite a few assumptions here; general case may be a lot different,
-          // but much much harder to implement. We assume that:
-          //  1) all maps are standard maps, not strided maps
-          //  2) global indices of dofs in A are related to dofs in coordinates in a simple arithmetic
-          //     way: rows i*blkSize, i*blkSize+1, ..., i*blkSize + (blkSize-1) correspond to node i
-          //
-          // NOTE: Potentially, some of the code below could be simplified with UnAmalgamationInfo,
-          // but as I totally don't understand that code, here is my solution
-
-          // [*1*]: see [*0*]
-
-          // Check that the number of local coordinates is consistent with the #rows in A
-          TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getLocalNumElements()/blkSize != Coords->getLocalLength(), Exceptions::Incompatible,
-                                     "Coordinate vector length (" << Coords->getLocalLength() << ") is incompatible with number of rows in A (" << A->getRowMap()->getLocalNumElements() << ") by modulo block size ("<< blkSize <<").");
-
-          const RCP<const Map> colMap = A->getColMap();
-          RCP<const Map> uniqueMap, nonUniqueMap;
-          Array<LO>      colTranslation;
-          if (blkSize == 1) {
-            uniqueMap    = A->getRowMap();
-            nonUniqueMap = A->getColMap();
-            graphType="unamalgamated";
-
-          } else {
-            uniqueMap    = Coords->getMap();
-            TEUCHOS_TEST_FOR_EXCEPTION(uniqueMap->getIndexBase() != indexBase, Exceptions::Incompatible,
-                                       "Different index bases for matrix and coordinates");
-
-            AmalgamationFactory::AmalgamateMap(*(A->getColMap()), *A, nonUniqueMap, colTranslation);
+          uniqueMap = Coords->getMap();
+          TEUCHOS_TEST_FOR_EXCEPTION(uniqueMap->getIndexBase() != indexBase, Exceptions::Incompatible,
+                                     "Different index bases for matrix and coordinates");
 
-            graphType = "amalgamated";
-          }
-          LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
+          AmalgamationFactory::AmalgamateMap(*(A->getColMap()), *A, nonUniqueMap, colTranslation);
 
-          RCP<RealValuedMultiVector> ghostedCoords;
-          RCP<Vector>                ghostedLaplDiag;
-          Teuchos::ArrayRCP<SC>      ghostedLaplDiagData;
-          if (threshold != STS::zero()) {
-            // Get ghost coordinates
-            RCP<const Import> importer;
-            {
-              SubFactoryMonitor m1(*this, "Import construction", currentLevel);
-              if (blkSize == 1 &&  realA->getCrsGraph()->getImporter() != Teuchos::null) {
-                GetOStream(Warnings1) << "Using existing importer from matrix graph" << std::endl;
-                importer = realA->getCrsGraph()->getImporter();
-              } else {
-                GetOStream(Warnings0) << "Constructing new importer instance" << std::endl;
-                importer = ImportFactory::Build(uniqueMap, nonUniqueMap);
-              }
-            } //subtimer
-            ghostedCoords = Xpetra::MultiVectorFactory<real_type,LO,GO,NO>::Build(nonUniqueMap, Coords->getNumVectors());
-            {
+          graphType = "amalgamated";
+        }
+        LO numRows = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
+
+        RCP<RealValuedMultiVector> ghostedCoords;
+        RCP<Vector> ghostedLaplDiag;
+        Teuchos::ArrayRCP<SC> ghostedLaplDiagData;
+        if (threshold != STS::zero()) {
+          // Get ghost coordinates
+          RCP<const Import> importer;
+          {
+            SubFactoryMonitor m1(*this, "Import construction", currentLevel);
+            if (blkSize == 1 && realA->getCrsGraph()->getImporter() != Teuchos::null) {
+              GetOStream(Warnings1) << "Using existing importer from matrix graph" << std::endl;
+              importer = realA->getCrsGraph()->getImporter();
+            } else {
+              GetOStream(Warnings0) << "Constructing new importer instance" << std::endl;
+              importer = ImportFactory::Build(uniqueMap, nonUniqueMap);
+            }
+          }  // subtimer
+          ghostedCoords = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(nonUniqueMap, Coords->getNumVectors());
+          {
             SubFactoryMonitor m1(*this, "Coordinate import", currentLevel);
             ghostedCoords->doImport(*Coords, *importer, Xpetra::INSERT);
-            } //subtimer
+          }  // subtimer
 
-            // Construct Distance Laplacian diagonal
-            RCP<Vector>  localLaplDiag     = VectorFactory::Build(uniqueMap);
-            Array<LO> indicesExtra;
-            Teuchos::Array<Teuchos::ArrayRCP<const real_type>> coordData;
-            if (threshold != STS::zero()) {
-              const size_t numVectors = ghostedCoords->getNumVectors();
-              coordData.reserve(numVectors);
-              for (size_t j = 0; j < numVectors; j++) {
-                Teuchos::ArrayRCP<const real_type> tmpData=ghostedCoords->getData(j);
-                coordData.push_back(tmpData);
-              }
+          // Construct Distance Laplacian diagonal
+          RCP<Vector> localLaplDiag = VectorFactory::Build(uniqueMap);
+          Array<LO> indicesExtra;
+          Teuchos::Array<Teuchos::ArrayRCP<const real_type>> coordData;
+          if (threshold != STS::zero()) {
+            const size_t numVectors = ghostedCoords->getNumVectors();
+            coordData.reserve(numVectors);
+            for (size_t j = 0; j < numVectors; j++) {
+              Teuchos::ArrayRCP<const real_type> tmpData = ghostedCoords->getData(j);
+              coordData.push_back(tmpData);
             }
-            {
+          }
+          {
             SubFactoryMonitor m1(*this, "Laplacian local diagonal", currentLevel);
             ArrayRCP<SC> localLaplDiagData = localLaplDiag->getDataNonConst(0);
             for (LO row = 0; row < numRows; row++) {
@@ -1144,26 +1119,24 @@ namespace MueLu {
                 indices = indicesExtra;
               }
 
-              LO nnz = indices.size();
+              LO nnz               = indices.size();
               bool haveAddedToDiag = false;
               for (LO colID = 0; colID < nnz; colID++) {
                 const LO col = indices[colID];
 
                 if (row != col) {
-                  if(use_dlap_weights == SINGLE_WEIGHTS) {
+                  if (use_dlap_weights == SINGLE_WEIGHTS) {
                     /*printf("[%d,%d] Unweighted Distance = %6.4e Weighted Distance = %6.4e\n",row,col,
                            MueLu::Utilities<real_type,LO,GO,NO>::Distance2(coordData, row, col),
                            MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(),coordData, row, col));*/
-                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(),coordData, row, col);
-                  }
-                  else if(use_dlap_weights == BLOCK_WEIGHTS)  {
-                    int block_id = row % interleaved_blocksize;
+                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(), coordData, row, col);
+                  } else if (use_dlap_weights == BLOCK_WEIGHTS) {
+                    int block_id    = row % interleaved_blocksize;
                     int block_start = block_id * interleaved_blocksize;
-                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(block_start,interleaved_blocksize),coordData, row, col);
-                  }
-                  else {
+                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(block_start, interleaved_blocksize), coordData, row, col);
+                  } else {
                     //                    printf("[%d,%d] Unweighted Distance = %6.4e\n",row,col,MueLu::Utilities<real_type,LO,GO,NO>::Distance2(coordData, row, col));
-                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(coordData, row, col);
+                    localLaplDiagData[row] += STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(coordData, row, col);
                   }
                   haveAddedToDiag = true;
                 }
@@ -1173,67 +1146,67 @@ namespace MueLu {
               if (!haveAddedToDiag)
                 localLaplDiagData[row] = STS::rmax();
             }
-            } //subtimer
-            {
+          }  // subtimer
+          {
             SubFactoryMonitor m1(*this, "Laplacian distributed diagonal", currentLevel);
             ghostedLaplDiag = VectorFactory::Build(nonUniqueMap);
             ghostedLaplDiag->doImport(*localLaplDiag, *importer, Xpetra::INSERT);
             ghostedLaplDiagData = ghostedLaplDiag->getDataNonConst(0);
-            } //subtimer
+          }  // subtimer
 
-          } else {
-            GetOStream(Runtime0) << "Skipping distance laplacian construction due to 0 threshold" << std::endl;
-          }
+        } else {
+          GetOStream(Runtime0) << "Skipping distance laplacian construction due to 0 threshold" << std::endl;
+        }
 
-          // NOTE: ghostedLaplDiagData might be zero if we don't actually calculate the laplacian
+        // NOTE: ghostedLaplDiagData might be zero if we don't actually calculate the laplacian
 
-          // allocate space for the local graph
-          ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows+1);
-          ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
+        // allocate space for the local graph
+        ArrayRCP<LO> rows    = ArrayRCP<LO>(numRows + 1);
+        ArrayRCP<LO> columns = ArrayRCP<LO>(A->getLocalNumEntries());
 
 #ifdef HAVE_MUELU_DEBUG
-	  // DEBUGGING
-	  for(LO i=0; i<(LO)columns.size(); i++) columns[i]=-666;
+        // DEBUGGING
+        for (LO i = 0; i < (LO)columns.size(); i++) columns[i] = -666;
 #endif
 
-	  // Extra array for if we're allowing symmetrization with cutting
-	  ArrayRCP<LO> rows_stop;
-	  bool use_stop_array = threshold != STS::zero() && distanceLaplacianAlgo == scaled_cut_symmetric;
-	  if(use_stop_array)
-	    rows_stop.resize(numRows);
-	 
-          const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
+        // Extra array for if we're allowing symmetrization with cutting
+        ArrayRCP<LO> rows_stop;
+        bool use_stop_array = threshold != STS::zero() && distanceLaplacianAlgo == scaled_cut_symmetric;
+        if (use_stop_array)
+          rows_stop.resize(numRows);
 
-          LO realnnz = 0;
-          rows[0] = 0;
+        const ArrayRCP<bool> amalgBoundaryNodes(numRows, false);
 
-          Array<LO> indicesExtra;
-          {
+        LO realnnz = 0;
+        rows[0]    = 0;
+
+        Array<LO> indicesExtra;
+        {
           SubFactoryMonitor m1(*this, "Laplacian dropping", currentLevel);
           Teuchos::Array<Teuchos::ArrayRCP<const real_type>> coordData;
           if (threshold != STS::zero()) {
             const size_t numVectors = ghostedCoords->getNumVectors();
             coordData.reserve(numVectors);
             for (size_t j = 0; j < numVectors; j++) {
-              Teuchos::ArrayRCP<const real_type> tmpData=ghostedCoords->getData(j);
+              Teuchos::ArrayRCP<const real_type> tmpData = ghostedCoords->getData(j);
               coordData.push_back(tmpData);
             }
           }
 
-	  ArrayView<const SC>     vals;//CMS hackery
+          ArrayView<const SC> vals;  // CMS hackery
           for (LO row = 0; row < numRows; row++) {
             ArrayView<const LO> indices;
             indicesExtra.resize(0);
-	    bool isBoundary = false;
+            bool isBoundary = false;
 
             if (blkSize == 1) {
-	      //	      ArrayView<const SC>     vals;//CMS uncomment
+              //	      ArrayView<const SC>     vals;//CMS uncomment
               A->getLocalRowView(row, indices, vals);
-	      isBoundary = pointBoundaryNodes[row];
+              isBoundary = pointBoundaryNodes[row];
             } else {
               // The amalgamated row is marked as Dirichlet iff all point rows are Dirichlet
               for (LO j = 0; j < blkSize; j++) {
-                if (!pointBoundaryNodes[row*blkSize+j]) {
+                if (!pointBoundaryNodes[row * blkSize + j]) {
                   isBoundary = false;
                   break;
                 }
@@ -1250,17 +1223,16 @@ namespace MueLu {
 
             LO nnz = indices.size(), rownnz = 0;
 
-	    if(use_stop_array) {
-	      rows[row+1] = rows[row]+nnz;
-	      realnnz = rows[row];
-	    }
+            if (use_stop_array) {
+              rows[row + 1] = rows[row] + nnz;
+              realnnz       = rows[row];
+            }
 
             if (threshold != STS::zero()) {
               // default
               if (distanceLaplacianAlgo == defaultAlgo) {
-		/* Standard Distance Laplacian */
+                /* Standard Distance Laplacian */
                 for (LO colID = 0; colID < nnz; colID++) {
-
                   LO col = indices[colID];
 
                   if (row == col) {
@@ -1269,23 +1241,21 @@ namespace MueLu {
                     continue;
                   }
 
-		  // We do not want the distance Laplacian aggregating boundary nodes
-		  if(isBoundary) continue;
+                  // We do not want the distance Laplacian aggregating boundary nodes
+                  if (isBoundary) continue;
 
                   SC laplVal;
-                  if(use_dlap_weights == SINGLE_WEIGHTS) {
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(),coordData, row, col);
-                  }
-                  else if(use_dlap_weights == BLOCK_WEIGHTS)  {
-                    int block_id = row % interleaved_blocksize;
+                  if (use_dlap_weights == SINGLE_WEIGHTS) {
+                    laplVal = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(), coordData, row, col);
+                  } else if (use_dlap_weights == BLOCK_WEIGHTS) {
+                    int block_id    = row % interleaved_blocksize;
                     int block_start = block_id * interleaved_blocksize;
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(block_start,interleaved_blocksize),coordData, row, col);
-                  }
-                  else {
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(coordData, row, col);
+                    laplVal         = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(block_start, interleaved_blocksize), coordData, row, col);
+                  } else {
+                    laplVal = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(coordData, row, col);
                   }
-                  real_type aiiajj = STS::magnitude(realThreshold*realThreshold * ghostedLaplDiagData[row]*ghostedLaplDiagData[col]);
-                  real_type aij    = STS::magnitude(laplVal*laplVal);
+                  real_type aiiajj = STS::magnitude(realThreshold * realThreshold * ghostedLaplDiagData[row] * ghostedLaplDiagData[col]);
+                  real_type aij    = STS::magnitude(laplVal * laplVal);
 
                   if (aij > aiiajj) {
                     columns[realnnz++] = col;
@@ -1295,8 +1265,8 @@ namespace MueLu {
                   }
                 }
               } else {
-		/* Cut Algorithm */
-                using DropTol = Details::DropTol<real_type,LO>;
+                /* Cut Algorithm */
+                using DropTol = Details::DropTol<real_type, LO>;
                 std::vector<DropTol> drop_vec;
                 drop_vec.reserve(nnz);
                 const real_type zero = Teuchos::ScalarTraits<real_type>::zero();
@@ -1304,31 +1274,28 @@ namespace MueLu {
 
                 // find magnitudes
                 for (LO colID = 0; colID < nnz; colID++) {
-
                   LO col = indices[colID];
 
                   if (row == col) {
-                    drop_vec.emplace_back( zero, one, colID, false);
+                    drop_vec.emplace_back(zero, one, colID, false);
                     continue;
                   }
-		  // We do not want the distance Laplacian aggregating boundary nodes
-		  if(isBoundary) continue;
+                  // We do not want the distance Laplacian aggregating boundary nodes
+                  if (isBoundary) continue;
 
                   SC laplVal;
-                  if(use_dlap_weights == SINGLE_WEIGHTS) {
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(),coordData, row, col);
-                  }
-                  else if(use_dlap_weights == BLOCK_WEIGHTS)  {
-                    int block_id = row % interleaved_blocksize;
+                  if (use_dlap_weights == SINGLE_WEIGHTS) {
+                    laplVal = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(), coordData, row, col);
+                  } else if (use_dlap_weights == BLOCK_WEIGHTS) {
+                    int block_id    = row % interleaved_blocksize;
                     int block_start = block_id * interleaved_blocksize;
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(dlap_weights(block_start,interleaved_blocksize),coordData, row, col);
-                  }
-                  else {
-                    laplVal = STS::one() / MueLu::Utilities<real_type,LO,GO,NO>::Distance2(coordData, row, col);
+                    laplVal         = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(dlap_weights(block_start, interleaved_blocksize), coordData, row, col);
+                  } else {
+                    laplVal = STS::one() / MueLu::Utilities<real_type, LO, GO, NO>::Distance2(coordData, row, col);
                   }
 
-                  real_type aiiajj = STS::magnitude(ghostedLaplDiagData[row]*ghostedLaplDiagData[col]);
-                  real_type aij    = STS::magnitude(laplVal*laplVal);
+                  real_type aiiajj = STS::magnitude(ghostedLaplDiagData[row] * ghostedLaplDiagData[col]);
+                  real_type aij    = STS::magnitude(laplVal * laplVal);
 
                   drop_vec.emplace_back(aij, aiiajj, colID, false);
                 }
@@ -1336,53 +1303,46 @@ namespace MueLu {
                 const size_t n = drop_vec.size();
 
                 if (distanceLaplacianAlgo == unscaled_cut) {
-
-                  std::sort( drop_vec.begin(), drop_vec.end()
-                           , [](DropTol const& a, DropTol const& b) {
-                               return a.val > b.val;
-                             }
-                           );
+                  std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+                    return a.val > b.val;
+                  });
 
                   bool drop = false;
-                  for (size_t i=1; i<n; ++i) {
+                  for (size_t i = 1; i < n; ++i) {
                     if (!drop) {
-                      auto const& x = drop_vec[i-1];
+                      auto const& x = drop_vec[i - 1];
                       auto const& y = drop_vec[i];
-                      auto a = x.val;
-                      auto b = y.val;
-                      if (a > realThreshold*b) {
+                      auto a        = x.val;
+                      auto b        = y.val;
+                      if (a > realThreshold * b) {
                         drop = true;
 #ifdef HAVE_MUELU_DEBUG
                         if (distanceLaplacianCutVerbose) {
-                          std::cout << "DJS: KEEP, N, ROW:  " << i+1 << ", " << n << ", " << row << std::endl;
+                          std::cout << "DJS: KEEP, N, ROW:  " << i + 1 << ", " << n << ", " << row << std::endl;
                         }
 #endif
                       }
                     }
                     drop_vec[i].drop = drop;
                   }
-                }
-                else if (distanceLaplacianAlgo == scaled_cut || distanceLaplacianAlgo == scaled_cut_symmetric) {
-
-                  std::sort( drop_vec.begin(), drop_vec.end()
-                           , [](DropTol const& a, DropTol const& b) {
-                               return a.val/a.diag > b.val/b.diag;
-                             }
-                           );
+                } else if (distanceLaplacianAlgo == scaled_cut || distanceLaplacianAlgo == scaled_cut_symmetric) {
+                  std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+                    return a.val / a.diag > b.val / b.diag;
+                  });
 
                   bool drop = false;
-                  for (size_t i=1; i<n; ++i) {
+                  for (size_t i = 1; i < n; ++i) {
                     if (!drop) {
-                      auto const& x = drop_vec[i-1];
+                      auto const& x = drop_vec[i - 1];
                       auto const& y = drop_vec[i];
-                      auto a = x.val/x.diag;
-                      auto b = y.val/y.diag;
-                      if (a > realThreshold*b) {
+                      auto a        = x.val / x.diag;
+                      auto b        = y.val / y.diag;
+                      if (a > realThreshold * b) {
                         drop = true;
 #ifdef HAVE_MUELU_DEBUG
                         if (distanceLaplacianCutVerbose) {
-                          std::cout << "DJS: KEEP, N, ROW:  " << i+1 << ", " << n << ", " << row << std::endl;
-			}
+                          std::cout << "DJS: KEEP, N, ROW:  " << i + 1 << ", " << n << ", " << row << std::endl;
+                        }
 #endif
                       }
                     }
@@ -1390,30 +1350,27 @@ namespace MueLu {
                   }
                 }
 
-                std::sort( drop_vec.begin(), drop_vec.end()
-                         , [](DropTol const& a, DropTol const& b) {
-                             return a.col < b.col;
-                           }
-                         );
+                std::sort(drop_vec.begin(), drop_vec.end(), [](DropTol const& a, DropTol const& b) {
+                  return a.col < b.col;
+                });
 
-		for (LO idxID =0; idxID<(LO)drop_vec.size(); idxID++) {
+                for (LO idxID = 0; idxID < (LO)drop_vec.size(); idxID++) {
                   LO col = indices[drop_vec[idxID].col];
 
-
                   // don't drop diagonal
                   if (row == col) {
                     columns[realnnz++] = col;
                     rownnz++;
-		    //		    printf("(%d,%d) KEEP %13s matrix = %6.4e\n",row,row,"DIAGONAL",drop_vec[idxID].aux_val);
+                    //		    printf("(%d,%d) KEEP %13s matrix = %6.4e\n",row,row,"DIAGONAL",drop_vec[idxID].aux_val);
                     continue;
                   }
 
                   if (!drop_vec[idxID].drop) {
                     columns[realnnz++] = col;
-		    //		    printf("(%d,%d) KEEP dlap = %6.4e matrix = %6.4e\n",row,col,drop_vec[idxID].val/drop_vec[idxID].diag,drop_vec[idxID].aux_val);
+                    //		    printf("(%d,%d) KEEP dlap = %6.4e matrix = %6.4e\n",row,col,drop_vec[idxID].val/drop_vec[idxID].diag,drop_vec[idxID].aux_val);
                     rownnz++;
                   } else {
-		    //		    printf("(%d,%d) DROP dlap = %6.4e matrix = %6.4e\n",row,col,drop_vec[idxID].val/drop_vec[idxID].diag,drop_vec[idxID].aux_val);
+                    //		    printf("(%d,%d) DROP dlap = %6.4e matrix = %6.4e\n",row,col,drop_vec[idxID].val/drop_vec[idxID].diag,drop_vec[idxID].aux_val);
                     numDropped++;
                   }
                 }
@@ -1421,13 +1378,13 @@ namespace MueLu {
             } else {
               // Skip laplace calculation and threshold comparison for zero threshold
               for (LO colID = 0; colID < nnz; colID++) {
-                LO col = indices[colID];
+                LO col             = indices[colID];
                 columns[realnnz++] = col;
                 rownnz++;
               }
             }
 
-            if ( rownnz == 1) {
+            if (rownnz == 1) {
               // If the only element remaining after filtering is diagonal, mark node as boundary
               // FIXME: this should really be replaced by the following
               //    if (indices.size() == 1 && indices[0] == row)
@@ -1437,584 +1394,575 @@ namespace MueLu {
               amalgBoundaryNodes[row] = true;
             }
 
-	    if(use_stop_array)
-	      rows_stop[row] = rownnz + rows[row];
-	    else
-	      rows[row+1] = realnnz;
-          } //for (LO row = 0; row < numRows; row++)
-
-          } //subtimer
-
-	  if (use_stop_array) {
-	    // Do symmetrization of the cut matrix
-	    // NOTE: We assume nested row/column maps here
-	    for (LO row = 0; row < numRows; row++) {
-	      for (LO colidx = rows[row]; colidx < rows_stop[row]; colidx++) {
-		LO col = columns[colidx];
-		if(col >= numRows) continue;
-		
-		bool found = false;
-		for(LO t_col = rows[col] ; !found && t_col  < rows_stop[col]; t_col++) {
-		  if (columns[t_col] == row)
-		    found = true;
-		}
-		// We didn't find the transpose buddy, so let's symmetrize, unless we'd be symmetrizing
-		// into a Dirichlet unknown.  In that case don't.
-		if(!found && !pointBoundaryNodes[col] && rows_stop[col] < rows[col+1]) {
-		  LO new_idx = rows_stop[col];
-		  //		  printf("(%d,%d) SYMADD entry\n",col,row);
-		  columns[new_idx] = row;
-		  rows_stop[col]++;	
-		  numDropped--;
-		}
-	      }
-	    }	      
-
-	    // Condense everything down
-	    LO current_start=0;
-	    for (LO row = 0; row < numRows; row++) {
-	      LO old_start = current_start;
-	      for (LO col = rows[row]; col < rows_stop[row]; col++) {
-		if(current_start != col) {
-		  columns[current_start] = columns[col];
-		}
-		current_start++;
-	      }
-	      rows[row] = old_start;	      	      	      	      
-	    }
-	    rows[numRows] = realnnz = current_start;
-
-	  }
-
-	  columns.resize(realnnz);
-
-          RCP<GraphBase> graph;
-          {
+            if (use_stop_array)
+              rows_stop[row] = rownnz + rows[row];
+            else
+              rows[row + 1] = realnnz;
+          }  // for (LO row = 0; row < numRows; row++)
+
+        }  // subtimer
+
+        if (use_stop_array) {
+          // Do symmetrization of the cut matrix
+          // NOTE: We assume nested row/column maps here
+          for (LO row = 0; row < numRows; row++) {
+            for (LO colidx = rows[row]; colidx < rows_stop[row]; colidx++) {
+              LO col = columns[colidx];
+              if (col >= numRows) continue;
+
+              bool found = false;
+              for (LO t_col = rows[col]; !found && t_col < rows_stop[col]; t_col++) {
+                if (columns[t_col] == row)
+                  found = true;
+              }
+              // We didn't find the transpose buddy, so let's symmetrize, unless we'd be symmetrizing
+              // into a Dirichlet unknown.  In that case don't.
+              if (!found && !pointBoundaryNodes[col] && rows_stop[col] < rows[col + 1]) {
+                LO new_idx = rows_stop[col];
+                //		  printf("(%d,%d) SYMADD entry\n",col,row);
+                columns[new_idx] = row;
+                rows_stop[col]++;
+                numDropped--;
+              }
+            }
+          }
+
+          // Condense everything down
+          LO current_start = 0;
+          for (LO row = 0; row < numRows; row++) {
+            LO old_start = current_start;
+            for (LO col = rows[row]; col < rows_stop[row]; col++) {
+              if (current_start != col) {
+                columns[current_start] = columns[col];
+              }
+              current_start++;
+            }
+            rows[row] = old_start;
+          }
+          rows[numRows] = realnnz = current_start;
+        }
+
+        columns.resize(realnnz);
+
+        RCP<GraphBase> graph;
+        {
           SubFactoryMonitor m1(*this, "Build amalgamated graph", currentLevel);
           graph = rcp(new LWGraph(rows, columns, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
           graph->SetBoundaryNodeMap(amalgBoundaryNodes);
-          } //subtimer
+        }  // subtimer
 
-          if (GetVerbLevel() & Statistics1) {
-            GO numLocalBoundaryNodes  = 0;
-            GO numGlobalBoundaryNodes = 0;
+        if (GetVerbLevel() & Statistics1) {
+          GO numLocalBoundaryNodes  = 0;
+          GO numGlobalBoundaryNodes = 0;
 
-            for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
-              if (amalgBoundaryNodes[i])
-                numLocalBoundaryNodes++;
+          for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
+            if (amalgBoundaryNodes[i])
+              numLocalBoundaryNodes++;
 
-            RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-            MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-            GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " agglomerated Dirichlet nodes"
-                                       << " using threshold " << dirichletThreshold << std::endl;
-          }
+          RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+          MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+          GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " agglomerated Dirichlet nodes"
+                                  << " using threshold " << dirichletThreshold << std::endl;
+        }
+
+        Set(currentLevel, "Graph", graph);
+        Set(currentLevel, "DofsPerNode", blkSize);
+      }
+    }
+
+    if ((GetVerbLevel() & Statistics1) && !(A->GetFixedBlockSize() > 1 && threshold != STS::zero())) {
+      RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+      GO numGlobalTotal, numGlobalDropped;
+      MueLu_sumAll(comm, numTotal, numGlobalTotal);
+      MueLu_sumAll(comm, numDropped, numGlobalDropped);
+      GetOStream(Statistics1) << "Number of dropped entries in " << graphType << " matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
+      if (numGlobalTotal != 0)
+        GetOStream(Statistics1) << " (" << 100 * Teuchos::as<double>(numGlobalDropped) / Teuchos::as<double>(numGlobalTotal) << "%)";
+      GetOStream(Statistics1) << std::endl;
+    }
 
-          Set(currentLevel, "Graph",       graph);
-          Set(currentLevel, "DofsPerNode", blkSize);
+  } else {
+    // what Tobias has implemented
+
+    SC threshold = as<SC>(pL.get<double>("aggregation: drop tol"));
+    // GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+    GetOStream(Runtime0) << "algorithm = \""
+                         << "failsafe"
+                         << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+    Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
+
+    RCP<const Map> rowMap = A->getRowMap();
+    RCP<const Map> colMap = A->getColMap();
+
+    LO blockdim  = 1;                       // block dim for fixed size blocks
+    GO indexBase = rowMap->getIndexBase();  // index base of maps
+    GO offset    = 0;
+
+    // 1) check for blocking/striding information
+    if (A->IsView("stridedMaps") &&
+        Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+      Xpetra::viewLabel_t oldView  = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+      RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
+      TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null, Exceptions::BadCast, "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
+      blockdim = strMap->getFixedBlockSize();
+      offset   = strMap->getOffset();
+      oldView  = A->SwitchToView(oldView);
+      GetOStream(Statistics1) << "CoalesceDropFactory::Build():"
+                              << " found blockdim=" << blockdim << " from strided maps. offset=" << offset << std::endl;
+    } else
+      GetOStream(Statistics1) << "CoalesceDropFactory::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
+
+    // 2) get row map for amalgamated matrix (graph of A)
+    //    with same distribution over all procs as row map of A
+    RCP<const Map> nodeMap = amalInfo->getNodeRowMap();
+    GetOStream(Statistics1) << "CoalesceDropFactory: nodeMap " << nodeMap->getLocalNumElements() << "/" << nodeMap->getGlobalNumElements() << " elements" << std::endl;
+
+    // 3) create graph of amalgamated matrix
+    RCP<CrsGraph> crsGraph = CrsGraphFactory::Build(nodeMap, A->getLocalMaxNumRowEntries() * blockdim);
+
+    LO numRows  = A->getRowMap()->getLocalNumElements();
+    LO numNodes = nodeMap->getLocalNumElements();
+    const ArrayRCP<bool> amalgBoundaryNodes(numNodes, false);
+    const ArrayRCP<int> numberDirichletRowsPerNode(numNodes, 0);  // helper array counting the number of Dirichlet nodes associated with node
+    bool bIsDiagonalEntry = false;                                // boolean flag stating that grid==gcid
+
+    // 4) do amalgamation. generate graph of amalgamated matrix
+    //    Note, this code is much more inefficient than the leightwight implementation
+    //    Most of the work has already been done in the AmalgamationFactory
+    for (LO row = 0; row < numRows; row++) {
+      // get global DOF id
+      GO grid = rowMap->getGlobalElement(row);
+
+      // reinitialize boolean helper variable
+      bIsDiagonalEntry = false;
+
+      // translate grid to nodeid
+      GO nodeId = AmalgamationFactory::DOFGid2NodeId(grid, blockdim, offset, indexBase);
+
+      size_t nnz = A->getNumEntriesInLocalRow(row);
+      Teuchos::ArrayView<const LO> indices;
+      Teuchos::ArrayView<const SC> vals;
+      A->getLocalRowView(row, indices, vals);
+
+      RCP<std::vector<GO>> cnodeIds = Teuchos::rcp(new std::vector<GO>);  // global column block ids
+      LO realnnz                    = 0;
+      for (LO col = 0; col < Teuchos::as<LO>(nnz); col++) {
+        GO gcid = colMap->getGlobalElement(indices[col]);  // global column id
+
+        if (vals[col] != STS::zero()) {
+          GO cnodeId = AmalgamationFactory::DOFGid2NodeId(gcid, blockdim, offset, indexBase);
+          cnodeIds->push_back(cnodeId);
+          realnnz++;  // increment number of nnz in matrix row
+          if (grid == gcid) bIsDiagonalEntry = true;
         }
       }
 
-      if ((GetVerbLevel() & Statistics1) && !(A->GetFixedBlockSize() > 1 && threshold != STS::zero())) {
-          RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-          GO numGlobalTotal, numGlobalDropped;
-          MueLu_sumAll(comm, numTotal,   numGlobalTotal);
-          MueLu_sumAll(comm, numDropped, numGlobalDropped);
-          GetOStream(Statistics1) << "Number of dropped entries in " << graphType << " matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
-          if (numGlobalTotal != 0)
-            GetOStream(Statistics1) << " (" << 100*Teuchos::as<double>(numGlobalDropped)/Teuchos::as<double>(numGlobalTotal) << "%)";
-          GetOStream(Statistics1) << std::endl;
+      if (realnnz == 1 && bIsDiagonalEntry == true) {
+        LO lNodeId = nodeMap->getLocalElement(nodeId);
+        numberDirichletRowsPerNode[lNodeId] += 1;             // increment Dirichlet row counter associated with lNodeId
+        if (numberDirichletRowsPerNode[lNodeId] == blockdim)  // mark full Dirichlet nodes
+          amalgBoundaryNodes[lNodeId] = true;
       }
 
-    } else {
-      //what Tobias has implemented
-
-      SC threshold = as<SC>(pL.get<double>("aggregation: drop tol"));
-      //GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
-      GetOStream(Runtime0) << "algorithm = \"" << "failsafe" << "\": threshold = " << threshold << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
-      Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
-
-      RCP<const Map> rowMap = A->getRowMap();
-      RCP<const Map> colMap = A->getColMap();
-
-      LO blockdim = 1;                          // block dim for fixed size blocks
-      GO indexBase = rowMap->getIndexBase();    // index base of maps
-      GO offset    = 0;
-
-      // 1) check for blocking/striding information
-      if(A->IsView("stridedMaps") &&
-         Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-        Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-        RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
-        TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null,Exceptions::BadCast,"MueLu::CoalesceFactory::Build: cast to strided row map failed.");
-        blockdim = strMap->getFixedBlockSize();
-        offset   = strMap->getOffset();
-        oldView = A->SwitchToView(oldView);
-        GetOStream(Statistics1) << "CoalesceDropFactory::Build():" << " found blockdim=" << blockdim << " from strided maps. offset=" << offset << std::endl;
-      } else GetOStream(Statistics1) << "CoalesceDropFactory::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
-
-      // 2) get row map for amalgamated matrix (graph of A)
-      //    with same distribution over all procs as row map of A
-      RCP<const Map> nodeMap = amalInfo->getNodeRowMap();
-      GetOStream(Statistics1) << "CoalesceDropFactory: nodeMap " << nodeMap->getLocalNumElements() << "/" << nodeMap->getGlobalNumElements() << " elements" << std::endl;
-
-      // 3) create graph of amalgamated matrix
-      RCP<CrsGraph> crsGraph = CrsGraphFactory::Build(nodeMap, A->getLocalMaxNumRowEntries()*blockdim);
-
-      LO numRows = A->getRowMap()->getLocalNumElements();
-      LO numNodes = nodeMap->getLocalNumElements();
-      const ArrayRCP<bool> amalgBoundaryNodes(numNodes, false);
-      const ArrayRCP<int>  numberDirichletRowsPerNode(numNodes, 0); // helper array counting the number of Dirichlet nodes associated with node
-      bool bIsDiagonalEntry = false;       // boolean flag stating that grid==gcid
-
-      // 4) do amalgamation. generate graph of amalgamated matrix
-      //    Note, this code is much more inefficient than the leightwight implementation
-      //    Most of the work has already been done in the AmalgamationFactory
-      for(LO row=0; row<numRows; row++) {
-        // get global DOF id
-        GO grid = rowMap->getGlobalElement(row);
-
-        // reinitialize boolean helper variable
-        bIsDiagonalEntry = false;
-
-        // translate grid to nodeid
-        GO nodeId = AmalgamationFactory::DOFGid2NodeId(grid, blockdim, offset, indexBase);
-
-        size_t nnz = A->getNumEntriesInLocalRow(row);
-        Teuchos::ArrayView<const LO> indices;
-        Teuchos::ArrayView<const SC> vals;
-        A->getLocalRowView(row, indices, vals);
-
-        RCP<std::vector<GO> > cnodeIds = Teuchos::rcp(new std::vector<GO>);  // global column block ids
-        LO realnnz = 0;
-        for(LO col=0; col<Teuchos::as<LO>(nnz); col++) {
-          GO gcid = colMap->getGlobalElement(indices[col]); // global column id
-
-          if(vals[col]!=STS::zero()) {
-            GO cnodeId = AmalgamationFactory::DOFGid2NodeId(gcid, blockdim, offset, indexBase);
-            cnodeIds->push_back(cnodeId);
-            realnnz++; // increment number of nnz in matrix row
-            if (grid == gcid) bIsDiagonalEntry = true;
-          }
-        }
+      Teuchos::ArrayRCP<GO> arr_cnodeIds = Teuchos::arcp(cnodeIds);
 
-        if(realnnz == 1 && bIsDiagonalEntry == true) {
-          LO lNodeId = nodeMap->getLocalElement(nodeId);
-          numberDirichletRowsPerNode[lNodeId] += 1;      // increment Dirichlet row counter associated with lNodeId
-          if (numberDirichletRowsPerNode[lNodeId] == blockdim) // mark full Dirichlet nodes
-            amalgBoundaryNodes[lNodeId] = true;
-        }
+      if (arr_cnodeIds.size() > 0)
+        crsGraph->insertGlobalIndices(nodeId, arr_cnodeIds());
+    }
+    // fill matrix graph
+    crsGraph->fillComplete(nodeMap, nodeMap);
 
-        Teuchos::ArrayRCP<GO> arr_cnodeIds = Teuchos::arcp( cnodeIds );
+    // 5) create MueLu Graph object
+    RCP<GraphBase> graph = rcp(new Graph(crsGraph, "amalgamated graph of A"));
 
-        if(arr_cnodeIds.size() > 0 )
-          crsGraph->insertGlobalIndices(nodeId, arr_cnodeIds());
-      }
-      // fill matrix graph
-      crsGraph->fillComplete(nodeMap,nodeMap);
+    // Detect and record rows that correspond to Dirichlet boundary conditions
+    graph->SetBoundaryNodeMap(amalgBoundaryNodes);
 
-      // 5) create MueLu Graph object
-      RCP<GraphBase> graph = rcp(new Graph(crsGraph, "amalgamated graph of A"));
+    if (GetVerbLevel() & Statistics1) {
+      GO numLocalBoundaryNodes  = 0;
+      GO numGlobalBoundaryNodes = 0;
+      for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
+        if (amalgBoundaryNodes[i])
+          numLocalBoundaryNodes++;
+      RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+      MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+      GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+    }
 
-      // Detect and record rows that correspond to Dirichlet boundary conditions
-      graph->SetBoundaryNodeMap(amalgBoundaryNodes);
-
-      if (GetVerbLevel() & Statistics1) {
-        GO numLocalBoundaryNodes  = 0;
-        GO numGlobalBoundaryNodes = 0;
-        for (LO i = 0; i < amalgBoundaryNodes.size(); ++i)
-          if (amalgBoundaryNodes[i])
-            numLocalBoundaryNodes++;
-        RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-        MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-        GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-      }
+    // 6) store results in Level
+    // graph->SetBoundaryNodeMap(gBoundaryNodeMap);
+    Set(currentLevel, "DofsPerNode", blockdim);
+    Set(currentLevel, "Graph", graph);
 
-      // 6) store results in Level
-      //graph->SetBoundaryNodeMap(gBoundaryNodeMap);
-      Set(currentLevel, "DofsPerNode", blockdim);
-      Set(currentLevel, "Graph", graph);
+  }  // if (doExperimentalWrap) ... else ...
 
-    } //if (doExperimentalWrap) ... else ...
+}  // Build
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergeRows(const Matrix& A, const LO row, Array<LO>& cols, const Array<LO>& translation) const {
+  typedef typename ArrayView<const LO>::size_type size_type;
 
-  } //Build
+  // extract striding information
+  LO blkSize = A.GetFixedBlockSize();  //< stores the size of the block within the strided map
+  if (A.IsView("stridedMaps") == true) {
+    Teuchos::RCP<const Map> myMap         = A.getRowMap("stridedMaps");
+    Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+    if (strMap->getStridedBlockId() > -1)
+      blkSize = Teuchos::as<LO>(strMap->getStridingData()[strMap->getStridedBlockId()]);
+  }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergeRows(const Matrix& A, const LO row, Array<LO>& cols, const Array<LO>& translation) const {
-    typedef typename ArrayView<const LO>::size_type size_type;
+  // count nonzero entries in all dof rows associated with node row
+  size_t nnz = 0, pos = 0;
+  for (LO j = 0; j < blkSize; j++)
+    nnz += A.getNumEntriesInLocalRow(row * blkSize + j);
 
-    // extract striding information
-    LO blkSize = A.GetFixedBlockSize();  //< stores the size of the block within the strided map
-    if (A.IsView("stridedMaps") == true) {
-      Teuchos::RCP<const Map> myMap = A.getRowMap("stridedMaps");
-      Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-      if (strMap->getStridedBlockId() > -1)
-        blkSize = Teuchos::as<LO>(strMap->getStridingData()[strMap->getStridedBlockId()]);
+  if (nnz == 0) {
+    cols.resize(0);
+    return;
+  }
+
+  cols.resize(nnz);
+
+  // loop over all local dof rows associated with local node "row"
+  ArrayView<const LO> inds;
+  ArrayView<const SC> vals;
+  for (LO j = 0; j < blkSize; j++) {
+    A.getLocalRowView(row * blkSize + j, inds, vals);
+    size_type numIndices = inds.size();
+
+    if (numIndices == 0)  // skip empty dof rows
+      continue;
+
+    // cols: stores all local node ids for current local node id "row"
+    cols[pos++] = translation[inds[0]];
+    for (size_type k = 1; k < numIndices; k++) {
+      LO nodeID = translation[inds[k]];
+      // Here we try to speed up the process by reducing the size of an array
+      // to sort. This works if the column nonzeros belonging to the same
+      // node are stored consequently.
+      if (nodeID != cols[pos - 1])
+        cols[pos++] = nodeID;
     }
+  }
+  cols.resize(pos);
+  nnz = pos;
+
+  // Sort and remove duplicates
+  std::sort(cols.begin(), cols.end());
+  pos = 0;
+  for (size_t j = 1; j < nnz; j++)
+    if (cols[j] != cols[pos])
+      cols[++pos] = cols[j];
+  cols.resize(pos + 1);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergeRowsWithDropping(const Matrix& A, const LO row, const ArrayRCP<const SC>& ghostedDiagVals, SC threshold, Array<LO>& cols, const Array<LO>& translation) const {
+  typedef typename ArrayView<const LO>::size_type size_type;
+  typedef Teuchos::ScalarTraits<SC> STS;
+
+  // extract striding information
+  LO blkSize = A.GetFixedBlockSize();  //< stores the size of the block within the strided map
+  if (A.IsView("stridedMaps") == true) {
+    Teuchos::RCP<const Map> myMap         = A.getRowMap("stridedMaps");
+    Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
+    if (strMap->getStridedBlockId() > -1)
+      blkSize = Teuchos::as<LO>(strMap->getStridingData()[strMap->getStridedBlockId()]);
+  }
 
-    // count nonzero entries in all dof rows associated with node row
-    size_t nnz = 0, pos = 0;
-    for (LO j = 0; j < blkSize; j++)
-      nnz += A.getNumEntriesInLocalRow(row*blkSize+j);
+  // count nonzero entries in all dof rows associated with node row
+  size_t nnz = 0, pos = 0;
+  for (LO j = 0; j < blkSize; j++)
+    nnz += A.getNumEntriesInLocalRow(row * blkSize + j);
 
-    if (nnz == 0) {
-      cols.resize(0);
-      return;
-    }
+  if (nnz == 0) {
+    cols.resize(0);
+    return;
+  }
 
-    cols.resize(nnz);
+  cols.resize(nnz);
 
-    // loop over all local dof rows associated with local node "row"
-    ArrayView<const LO> inds;
-    ArrayView<const SC> vals;
-    for (LO j = 0; j < blkSize; j++) {
-      A.getLocalRowView(row*blkSize+j, inds, vals);
-      size_type numIndices = inds.size();
+  // loop over all local dof rows associated with local node "row"
+  ArrayView<const LO> inds;
+  ArrayView<const SC> vals;
+  for (LO j = 0; j < blkSize; j++) {
+    A.getLocalRowView(row * blkSize + j, inds, vals);
+    size_type numIndices = inds.size();
+
+    if (numIndices == 0)  // skip empty dof rows
+      continue;
+
+    // cols: stores all local node ids for current local node id "row"
+    LO prevNodeID = -1;
+    for (size_type k = 0; k < numIndices; k++) {
+      LO dofID  = inds[k];
+      LO nodeID = translation[inds[k]];
 
-      if (numIndices == 0) // skip empty dof rows
-        continue;
+      // we avoid a square root by using squared values
+      typename STS::magnitudeType aiiajj = STS::magnitude(threshold * threshold * ghostedDiagVals[dofID] * ghostedDiagVals[row * blkSize + j]);  // eps^2 * |a_ii| * |a_jj|
+      typename STS::magnitudeType aij    = STS::magnitude(vals[k] * vals[k]);
+
+      // check dropping criterion
+      if (aij > aiiajj || (row * blkSize + j == dofID)) {
+        // accept entry in graph
 
-      // cols: stores all local node ids for current local node id "row"
-      cols[pos++] = translation[inds[0]];
-      for (size_type k = 1; k < numIndices; k++) {
-        LO nodeID = translation[inds[k]];
         // Here we try to speed up the process by reducing the size of an array
         // to sort. This works if the column nonzeros belonging to the same
         // node are stored consequently.
-        if (nodeID != cols[pos-1])
+        if (nodeID != prevNodeID) {
           cols[pos++] = nodeID;
+          prevNodeID  = nodeID;
+        }
       }
     }
-    cols.resize(pos);
-    nnz = pos;
-
-    // Sort and remove duplicates
-    std::sort(cols.begin(), cols.end());
-    pos = 0;
-    for (size_t j = 1; j < nnz; j++)
-      if (cols[j] != cols[pos])
-        cols[++pos] = cols[j];
-    cols.resize(pos+1);
+  }
+  cols.resize(pos);
+  nnz = pos;
+
+  // Sort and remove duplicates
+  std::sort(cols.begin(), cols.end());
+  pos = 0;
+  for (size_t j = 1; j < nnz; j++)
+    if (cols[j] != cols[pos])
+      cols[++pos] = cols[j];
+  cols.resize(pos + 1);
+
+  return;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockDiagonalize(Level& currentLevel, const RCP<Matrix>& A, bool generate_matrix) const {
+  typedef Teuchos::ScalarTraits<SC> STS;
+
+  const ParameterList& pL                              = GetParameterList();
+  const typename STS::magnitudeType dirichletThreshold = STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
+  const typename STS::magnitudeType rowSumTol          = as<typename STS::magnitudeType>(pL.get<double>("aggregation: row sum drop tol"));
+
+  RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector>>(currentLevel, "BlockNumber");
+  RCP<LocalOrdinalVector> ghostedBlockNumber;
+  GetOStream(Statistics1) << "Using BlockDiagonal Graph before dropping (with provided blocking)" << std::endl;
+
+  // Ghost the column block numbers if we need to
+  RCP<const Import> importer = A->getCrsGraph()->getImporter();
+  if (!importer.is_null()) {
+    SubFactoryMonitor m1(*this, "Block Number import", currentLevel);
+    ghostedBlockNumber = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(importer->getTargetMap());
+    ghostedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
+  } else {
+    ghostedBlockNumber = BlockNumber;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergeRowsWithDropping(const Matrix& A, const LO row, const ArrayRCP<const SC>& ghostedDiagVals, SC threshold, Array<LO>& cols, const Array<LO>& translation) const {
-    typedef typename ArrayView<const LO>::size_type size_type;
-    typedef Teuchos::ScalarTraits<SC> STS;
-
-    // extract striding information
-    LO blkSize = A.GetFixedBlockSize();  //< stores the size of the block within the strided map
-    if (A.IsView("stridedMaps") == true) {
-      Teuchos::RCP<const Map> myMap = A.getRowMap("stridedMaps");
-      Teuchos::RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == null, Exceptions::RuntimeError, "Map is not of type StridedMap");
-      if (strMap->getStridedBlockId() > -1)
-        blkSize = Teuchos::as<LO>(strMap->getStridingData()[strMap->getStridedBlockId()]);
-    }
-
-    // count nonzero entries in all dof rows associated with node row
-    size_t nnz = 0, pos = 0;
-    for (LO j = 0; j < blkSize; j++)
-      nnz += A.getNumEntriesInLocalRow(row*blkSize+j);
+  // Accessors for block numbers
+  Teuchos::ArrayRCP<const LO> row_block_number = BlockNumber->getData(0);
+  Teuchos::ArrayRCP<const LO> col_block_number = ghostedBlockNumber->getData(0);
+
+  // allocate space for the local graph
+  ArrayRCP<size_t> rows_mat;
+  ArrayRCP<LO> rows_graph, columns;
+  ArrayRCP<SC> values;
+  RCP<CrsMatrixWrap> crs_matrix_wrap;
+
+  if (generate_matrix) {
+    crs_matrix_wrap = rcp(new CrsMatrixWrap(A->getRowMap(), A->getColMap(), 0));
+    crs_matrix_wrap->getCrsMatrix()->allocateAllValues(A->getLocalNumEntries(), rows_mat, columns, values);
+  } else {
+    rows_graph.resize(A->getLocalNumRows() + 1);
+    columns.resize(A->getLocalNumEntries());
+    values.resize(A->getLocalNumEntries());
+  }
 
-    if (nnz == 0) {
-      cols.resize(0);
-      return;
-    }
+  LO realnnz    = 0;
+  GO numDropped = 0, numTotal = 0;
+  for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
+    LO row_block = row_block_number[row];
+    size_t nnz   = A->getNumEntriesInLocalRow(row);
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    A->getLocalRowView(row, indices, vals);
 
-    cols.resize(nnz);
+    LO rownnz = 0;
+    for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
+      LO col       = indices[colID];
+      LO col_block = col_block_number[col];
 
-    // loop over all local dof rows associated with local node "row"
-    ArrayView<const LO> inds;
-    ArrayView<const SC> vals;
-    for (LO j = 0; j < blkSize; j++) {
-      A.getLocalRowView(row*blkSize+j, inds, vals);
-      size_type numIndices = inds.size();
-
-      if (numIndices == 0) // skip empty dof rows
-        continue;
-
-      // cols: stores all local node ids for current local node id "row"
-      LO prevNodeID = -1;
-      for (size_type k = 0; k < numIndices; k++) {
-        LO dofID = inds[k];
-        LO nodeID = translation[inds[k]];
-
-        // we avoid a square root by using squared values
-        typename STS::magnitudeType aiiajj = STS::magnitude(threshold*threshold*ghostedDiagVals[dofID]*ghostedDiagVals[row*blkSize+j]); // eps^2 * |a_ii| * |a_jj|
-        typename STS::magnitudeType aij = STS::magnitude(vals[k]*vals[k]);
-
-        // check dropping criterion
-        if (aij > aiiajj || (row*blkSize+j == dofID)) {
-          // accept entry in graph
-
-          // Here we try to speed up the process by reducing the size of an array
-          // to sort. This works if the column nonzeros belonging to the same
-          // node are stored consequently.
-          if (nodeID != prevNodeID) {
-            cols[pos++] = nodeID;
-            prevNodeID = nodeID;
-          }
-        }
-      }
+      if (row_block == col_block) {
+        if (generate_matrix) values[realnnz] = vals[colID];
+        columns[realnnz++] = col;
+        rownnz++;
+      } else
+        numDropped++;
     }
-    cols.resize(pos);
-    nnz = pos;
+    if (generate_matrix)
+      rows_mat[row + 1] = realnnz;
+    else
+      rows_graph[row + 1] = realnnz;
+  }
 
-    // Sort and remove duplicates
-    std::sort(cols.begin(), cols.end());
-    pos = 0;
-    for (size_t j = 1; j < nnz; j++)
-      if (cols[j] != cols[pos])
-        cols[++pos] = cols[j];
-    cols.resize(pos+1);
+  ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC, LO, GO, NO>::DetectDirichletRows(*A, dirichletThreshold));
+  if (rowSumTol > 0.)
+    Utilities::ApplyRowSumCriterion(*A, rowSumTol, boundaryNodes);
 
-    return;
+  if (!generate_matrix) {
+    // We can't resize an Arrayrcp and pass the checks for setAllValues
+    values.resize(realnnz);
+    columns.resize(realnnz);
+  }
+  numTotal = A->getLocalNumEntries();
+
+  if (GetVerbLevel() & Statistics1) {
+    GO numLocalBoundaryNodes  = 0;
+    GO numGlobalBoundaryNodes = 0;
+    for (LO i = 0; i < boundaryNodes.size(); ++i)
+      if (boundaryNodes[i])
+        numLocalBoundaryNodes++;
+    RCP<const Teuchos::Comm<int>> comm = A->getRowMap()->getComm();
+    MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+    GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+
+    GO numGlobalTotal, numGlobalDropped;
+    MueLu_sumAll(comm, numTotal, numGlobalTotal);
+    MueLu_sumAll(comm, numDropped, numGlobalDropped);
+    GetOStream(Statistics1) << "Number of dropped entries in block-diagonalized matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
+    if (numGlobalTotal != 0)
+      GetOStream(Statistics1) << " (" << 100 * Teuchos::as<double>(numGlobalDropped) / Teuchos::as<double>(numGlobalTotal) << "%)";
+    GetOStream(Statistics1) << std::endl;
   }
 
+  Set(currentLevel, "Filtering", true);
+
+  if (generate_matrix) {
+    // NOTE: Trying to use A's Import/Export objects will cause the code to segfault back in Build() with errors on the Import
+    // if you're using Epetra.  I'm not really sure why. By using the Col==Domain and Row==Range maps, we get null Import/Export objects
+    // here, which is legit, because we never use them anyway.
+    crs_matrix_wrap->getCrsMatrix()->setAllValues(rows_mat, columns, values);
+    crs_matrix_wrap->getCrsMatrix()->expertStaticFillComplete(A->getColMap(), A->getRowMap());
+  } else {
+    RCP<GraphBase> graph = rcp(new LWGraph(rows_graph, columns, A->getRowMap(), A->getColMap(), "block-diagonalized graph of A"));
+    graph->SetBoundaryNodeMap(boundaryNodes);
+    Set(currentLevel, "Graph", graph);
+  }
 
+  Set(currentLevel, "DofsPerNode", 1);
+  return crs_matrix_wrap;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockDiagonalize(Level & currentLevel,const RCP<Matrix>& A,bool generate_matrix) const {
-    typedef Teuchos::ScalarTraits<SC> STS;
- 
-    const ParameterList  & pL = GetParameterList();
-    const typename STS::magnitudeType dirichletThreshold = STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
-    const typename STS::magnitudeType rowSumTol = as<typename STS::magnitudeType>(pL.get<double>("aggregation: row sum drop tol"));
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockDiagonalizeGraph(const RCP<GraphBase>& inputGraph, const RCP<LocalOrdinalVector>& ghostedBlockNumber, RCP<GraphBase>& outputGraph, RCP<const Import>& importer) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(ghostedBlockNumber.is_null(), Exceptions::RuntimeError, "BlockDiagonalizeGraph(): ghostedBlockNumber is null.");
+  const ParameterList& pL = GetParameterList();
 
-    RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector> >(currentLevel, "BlockNumber");
-    RCP<LocalOrdinalVector> ghostedBlockNumber;
-    GetOStream(Statistics1) << "Using BlockDiagonal Graph before dropping (with provided blocking)"<<std::endl;
+  const bool localizeColoringGraph = pL.get<bool>("aggregation: coloring: localize color graph");
 
-    // Ghost the column block numbers if we need to
-    RCP<const Import> importer = A->getCrsGraph()->getImporter();
-    if(!importer.is_null()) {
-      SubFactoryMonitor m1(*this, "Block Number import", currentLevel);
-      ghostedBlockNumber= Xpetra::VectorFactory<LO,LO,GO,NO>::Build(importer->getTargetMap());
-      ghostedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
-    }
-    else {
-      ghostedBlockNumber = BlockNumber;
-    }
+  GetOStream(Statistics1) << "Using BlockDiagonal Graph after Dropping (with provided blocking)";
+  if (localizeColoringGraph)
+    GetOStream(Statistics1) << ", with localization" << std::endl;
+  else
+    GetOStream(Statistics1) << ", without localization" << std::endl;
 
-    // Accessors for block numbers
-    Teuchos::ArrayRCP<const LO> row_block_number = BlockNumber->getData(0);
-    Teuchos::ArrayRCP<const LO> col_block_number = ghostedBlockNumber->getData(0);
-
-    // allocate space for the local graph
-    ArrayRCP<size_t> rows_mat;
-    ArrayRCP<LO> rows_graph,columns;
-    ArrayRCP<SC> values;
-    RCP<CrsMatrixWrap> crs_matrix_wrap;
-    
-    if(generate_matrix) {
-      crs_matrix_wrap = rcp(new CrsMatrixWrap(A->getRowMap(), A->getColMap(), 0));
-      crs_matrix_wrap->getCrsMatrix()->allocateAllValues(A->getLocalNumEntries(), rows_mat, columns, values);
-    }
-    else {
-      rows_graph.resize(A->getLocalNumRows()+1);
-      columns.resize(A->getLocalNumEntries());
-      values.resize(A->getLocalNumEntries());
-    }
-      
-    LO realnnz = 0;
-    GO numDropped = 0, numTotal = 0;
-    for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
-      LO row_block = row_block_number[row];
-      size_t nnz = A->getNumEntriesInLocalRow(row);
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      A->getLocalRowView(row, indices, vals);
+  // Accessors for block numbers
+  Teuchos::ArrayRCP<const LO> row_block_number = ghostedBlockNumber->getData(0);
+  Teuchos::ArrayRCP<const LO> col_block_number = ghostedBlockNumber->getData(0);
+
+  // allocate space for the local graph
+  ArrayRCP<size_t> rows_mat;
+  ArrayRCP<LO> rows_graph, columns;
+
+  rows_graph.resize(inputGraph->GetNodeNumVertices() + 1);
+  columns.resize(inputGraph->GetNodeNumEdges());
+
+  LO realnnz    = 0;
+  GO numDropped = 0, numTotal = 0;
+  const LO numRows = Teuchos::as<LO>(inputGraph->GetDomainMap()->getLocalNumElements());
+  if (localizeColoringGraph) {
+    for (LO row = 0; row < numRows; ++row) {
+      LO row_block                = row_block_number[row];
+      ArrayView<const LO> indices = inputGraph->getNeighborVertices(row);
 
       LO rownnz = 0;
-      for (LO colID = 0; colID < Teuchos::as<LO>(nnz); colID++) {
-        LO col = indices[colID];
+      for (LO colID = 0; colID < Teuchos::as<LO>(indices.size()); colID++) {
+        LO col       = indices[colID];
         LO col_block = col_block_number[col];
-        
-        if(row_block == col_block) {
-          if(generate_matrix) values[realnnz] = vals[colID];
+
+        if ((row_block == col_block) && (col < numRows)) {
           columns[realnnz++] = col;
           rownnz++;
         } else
           numDropped++;
       }
-      if(generate_matrix) rows_mat[row+1] = realnnz;
-      else rows_graph[row+1] = realnnz;
+      rows_graph[row + 1] = realnnz;
     }
-    
-    ArrayRCP<bool> boundaryNodes = Teuchos::arcp_const_cast<bool>(MueLu::Utilities<SC,LO,GO,NO>::DetectDirichletRows(*A, dirichletThreshold));
-    if (rowSumTol > 0.)
-      Utilities::ApplyRowSumCriterion(*A, rowSumTol, boundaryNodes);
-
-        
-    if(!generate_matrix) {
-      // We can't resize an Arrayrcp and pass the checks for setAllValues
-      values.resize(realnnz);
-      columns.resize(realnnz);
-    }
-    numTotal = A->getLocalNumEntries();
+  } else {
+    // ghosting of boundary node map
+    Teuchos::ArrayRCP<const bool> boundaryNodes = inputGraph->GetBoundaryNodeMap();
+    auto boundaryNodesVector                    = Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(inputGraph->GetDomainMap());
+    for (size_t i = 0; i < inputGraph->GetNodeNumVertices(); i++)
+      boundaryNodesVector->getDataNonConst(0)[i] = boundaryNodes[i];
+    // Xpetra::IO<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Write("boundary",*boundaryNodesVector);
+    auto boundaryColumnVector = Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(inputGraph->GetImportMap());
+    boundaryColumnVector->doImport(*boundaryNodesVector, *importer, Xpetra::INSERT);
+    auto boundaryColumn = boundaryColumnVector->getData(0);
+
+    for (LO row = 0; row < numRows; ++row) {
+      LO row_block                = row_block_number[row];
+      ArrayView<const LO> indices = inputGraph->getNeighborVertices(row);
 
-    if (GetVerbLevel() & Statistics1) {
-      GO numLocalBoundaryNodes  = 0;
-      GO numGlobalBoundaryNodes = 0;
-      for (LO i = 0; i < boundaryNodes.size(); ++i)
-        if (boundaryNodes[i])
-          numLocalBoundaryNodes++;
-      RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-      MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-      GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-
-      GO numGlobalTotal, numGlobalDropped;
-      MueLu_sumAll(comm, numTotal,   numGlobalTotal);
-      MueLu_sumAll(comm, numDropped, numGlobalDropped);
-      GetOStream(Statistics1) << "Number of dropped entries in block-diagonalized matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
-      if (numGlobalTotal != 0)
-        GetOStream(Statistics1) << " (" << 100*Teuchos::as<double>(numGlobalDropped)/Teuchos::as<double>(numGlobalTotal) << "%)";
-      GetOStream(Statistics1) << std::endl;
-    }
-
-    Set(currentLevel, "Filtering", true);
+      LO rownnz = 0;
+      for (LO colID = 0; colID < Teuchos::as<LO>(indices.size()); colID++) {
+        LO col       = indices[colID];
+        LO col_block = col_block_number[col];
 
-    if(generate_matrix) {
-      // NOTE: Trying to use A's Import/Export objects will cause the code to segfault back in Build() with errors on the Import
-      // if you're using Epetra.  I'm not really sure why. By using the Col==Domain and Row==Range maps, we get null Import/Export objects
-      // here, which is legit, because we never use them anyway.
-      crs_matrix_wrap->getCrsMatrix()->setAllValues(rows_mat,columns,values);
-      crs_matrix_wrap->getCrsMatrix()->expertStaticFillComplete(A->getColMap(), A->getRowMap());
-    }
-    else {
-      RCP<GraphBase> graph =  rcp(new LWGraph(rows_graph, columns, A->getRowMap(), A->getColMap(), "block-diagonalized graph of A"));
-      graph->SetBoundaryNodeMap(boundaryNodes);
-      Set(currentLevel, "Graph",       graph);
+        if ((row_block == col_block) && ((row == col) || (boundaryColumn[col] == 0))) {
+          columns[realnnz++] = col;
+          rownnz++;
+        } else
+          numDropped++;
+      }
+      rows_graph[row + 1] = realnnz;
     }
-
-
-    Set(currentLevel, "DofsPerNode", 1);
-    return crs_matrix_wrap;
   }
 
+  columns.resize(realnnz);
+  numTotal = inputGraph->GetNodeNumEdges();
+
+  if (GetVerbLevel() & Statistics1) {
+    RCP<const Teuchos::Comm<int>> comm = inputGraph->GetDomainMap()->getComm();
+    GO numGlobalTotal, numGlobalDropped;
+    MueLu_sumAll(comm, numTotal, numGlobalTotal);
+    MueLu_sumAll(comm, numDropped, numGlobalDropped);
+    GetOStream(Statistics1) << "Number of dropped entries in block-diagonalized matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
+    if (numGlobalTotal != 0)
+      GetOStream(Statistics1) << " (" << 100 * Teuchos::as<double>(numGlobalDropped) / Teuchos::as<double>(numGlobalTotal) << "%)";
+    GetOStream(Statistics1) << std::endl;
+  }
 
- template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
- void  CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockDiagonalizeGraph(const RCP<GraphBase> & inputGraph, const RCP<LocalOrdinalVector> & ghostedBlockNumber, RCP<GraphBase> & outputGraph, RCP<const Import> & importer) const {
-
-   TEUCHOS_TEST_FOR_EXCEPTION(ghostedBlockNumber.is_null(), Exceptions::RuntimeError, "BlockDiagonalizeGraph(): ghostedBlockNumber is null.");
-   const ParameterList  & pL = GetParameterList();
-
-   const bool localizeColoringGraph = pL.get<bool>("aggregation: coloring: localize color graph");
-
-   GetOStream(Statistics1) << "Using BlockDiagonal Graph after Dropping (with provided blocking)";
-   if (localizeColoringGraph)
-     GetOStream(Statistics1) << ", with localization" <<std::endl;
-   else
-     GetOStream(Statistics1) << ", without localization" <<std::endl;
-
-   // Accessors for block numbers
-   Teuchos::ArrayRCP<const LO>   row_block_number = ghostedBlockNumber->getData(0);
-   Teuchos::ArrayRCP<const LO>   col_block_number = ghostedBlockNumber->getData(0);
-
-   // allocate space for the local graph
-   ArrayRCP<size_t> rows_mat;
-   ArrayRCP<LO> rows_graph,columns;
-    
-   rows_graph.resize(inputGraph->GetNodeNumVertices()+1);
-   columns.resize(inputGraph->GetNodeNumEdges());
-
-   LO realnnz = 0;
-   GO numDropped = 0, numTotal = 0;
-   const LO numRows = Teuchos::as<LO>(inputGraph->GetDomainMap()->getLocalNumElements());
-   if (localizeColoringGraph) {
-
-     for (LO row = 0; row < numRows; ++row) {
-       LO row_block = row_block_number[row];
-       ArrayView<const LO> indices = inputGraph->getNeighborVertices(row);
-
-       LO rownnz = 0;
-       for (LO colID = 0; colID < Teuchos::as<LO>(indices.size()); colID++) {
-         LO col = indices[colID];
-         LO col_block = col_block_number[col];
-
-         if((row_block == col_block) && (col < numRows)) {
-           columns[realnnz++] = col;
-           rownnz++;
-         } else
-           numDropped++;
-       }
-       rows_graph[row+1] = realnnz;
-     }
-   } else {
-     // ghosting of boundary node map
-     Teuchos::ArrayRCP<const bool> boundaryNodes = inputGraph->GetBoundaryNodeMap();
-     auto boundaryNodesVector = Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(inputGraph->GetDomainMap());
-     for (size_t i=0; i<inputGraph->GetNodeNumVertices(); i++)
-       boundaryNodesVector->getDataNonConst(0)[i] = boundaryNodes[i];
-     // Xpetra::IO<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Write("boundary",*boundaryNodesVector);
-     auto boundaryColumnVector = Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(inputGraph->GetImportMap());
-     boundaryColumnVector->doImport(*boundaryNodesVector,*importer, Xpetra::INSERT);
-     auto boundaryColumn = boundaryColumnVector->getData(0);
-
-     for (LO row = 0; row < numRows; ++row) {
-       LO row_block = row_block_number[row];
-       ArrayView<const LO> indices = inputGraph->getNeighborVertices(row);
-
-       LO rownnz = 0;
-       for (LO colID = 0; colID < Teuchos::as<LO>(indices.size()); colID++) {
-         LO col = indices[colID];
-         LO col_block = col_block_number[col];
-
-         if((row_block == col_block) && ((row == col) || (boundaryColumn[col] == 0))) {
-           columns[realnnz++] = col;
-           rownnz++;
-         } else
-           numDropped++;
-       }
-       rows_graph[row+1] = realnnz;
-     }
-   }
-    
-   columns.resize(realnnz);
-   numTotal = inputGraph->GetNodeNumEdges();
-
-   if (GetVerbLevel() & Statistics1) {
-     RCP<const Teuchos::Comm<int> > comm = inputGraph->GetDomainMap()->getComm();
-     GO numGlobalTotal, numGlobalDropped;
-     MueLu_sumAll(comm, numTotal,   numGlobalTotal);
-     MueLu_sumAll(comm, numDropped, numGlobalDropped);
-     GetOStream(Statistics1) << "Number of dropped entries in block-diagonalized matrix graph: " << numGlobalDropped << "/" << numGlobalTotal;
-     if (numGlobalTotal != 0)
-       GetOStream(Statistics1) << " (" << 100*Teuchos::as<double>(numGlobalDropped)/Teuchos::as<double>(numGlobalTotal) << "%)";
-     GetOStream(Statistics1) << std::endl;
-   }
-
-   if (localizeColoringGraph) {
-     outputGraph =  rcp(new LWGraph(rows_graph, columns, inputGraph->GetDomainMap(), inputGraph->GetImportMap(), "block-diagonalized graph of A"));
-     outputGraph->SetBoundaryNodeMap(inputGraph->GetBoundaryNodeMap());
-   } else {
-     TEUCHOS_ASSERT(inputGraph->GetDomainMap()->lib() == Xpetra::UseTpetra);
+  if (localizeColoringGraph) {
+    outputGraph = rcp(new LWGraph(rows_graph, columns, inputGraph->GetDomainMap(), inputGraph->GetImportMap(), "block-diagonalized graph of A"));
+    outputGraph->SetBoundaryNodeMap(inputGraph->GetBoundaryNodeMap());
+  } else {
+    TEUCHOS_ASSERT(inputGraph->GetDomainMap()->lib() == Xpetra::UseTpetra);
 #ifdef HAVE_XPETRA_TPETRA
-     auto outputGraph2 =  rcp(new LWGraph(rows_graph, columns, inputGraph->GetDomainMap(), inputGraph->GetImportMap(), "block-diagonalized graph of A"));
-
-     auto tpGraph = Xpetra::toTpetra(rcp_const_cast<const CrsGraph>(outputGraph2->GetCrsGraph()));
-     auto sym = rcp(new Tpetra::CrsGraphTransposer<LocalOrdinal,GlobalOrdinal,Node>(tpGraph));
-     auto tpGraphSym = sym->symmetrize();
-
-     auto colIndsSym =        // FIXME persistingView is temporary; better fix would be change to LWGraph constructor
-          Kokkos::Compat::persistingView(tpGraphSym->getLocalIndicesHost());
-          
-     auto rowsSym = tpGraphSym->getLocalRowPtrsHost();
-     ArrayRCP<LO> rows_graphSym;
-     rows_graphSym.resize(rowsSym.size());
-     for (size_t row = 0; row < rowsSym.size(); row++)
-       rows_graphSym[row] = rowsSym[row];
-     outputGraph =  rcp(new LWGraph(rows_graphSym, colIndsSym, inputGraph->GetDomainMap(), Xpetra::toXpetra(tpGraphSym->getColMap()), "block-diagonalized graph of A"));
-     outputGraph->SetBoundaryNodeMap(inputGraph->GetBoundaryNodeMap());
+    auto outputGraph2 = rcp(new LWGraph(rows_graph, columns, inputGraph->GetDomainMap(), inputGraph->GetImportMap(), "block-diagonalized graph of A"));
+
+    auto tpGraph    = Xpetra::toTpetra(rcp_const_cast<const CrsGraph>(outputGraph2->GetCrsGraph()));
+    auto sym        = rcp(new Tpetra::CrsGraphTransposer<LocalOrdinal, GlobalOrdinal, Node>(tpGraph));
+    auto tpGraphSym = sym->symmetrize();
+
+    auto colIndsSym =  // FIXME persistingView is temporary; better fix would be change to LWGraph constructor
+        Kokkos::Compat::persistingView(tpGraphSym->getLocalIndicesHost());
+
+    auto rowsSym = tpGraphSym->getLocalRowPtrsHost();
+    ArrayRCP<LO> rows_graphSym;
+    rows_graphSym.resize(rowsSym.size());
+    for (size_t row = 0; row < rowsSym.size(); row++)
+      rows_graphSym[row] = rowsSym[row];
+    outputGraph = rcp(new LWGraph(rows_graphSym, colIndsSym, inputGraph->GetDomainMap(), Xpetra::toXpetra(tpGraphSym->getColMap()), "block-diagonalized graph of A"));
+    outputGraph->SetBoundaryNodeMap(inputGraph->GetBoundaryNodeMap());
 #endif
-   }
-
- }
-
-
+  }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_COALESCEDROPFACTORY_DEF_HPP
+#endif  // MUELU_COALESCEDROPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_decl.hpp
index 45d2601b230d..3f314a916904 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_decl.hpp
@@ -62,116 +62,114 @@
 
 namespace MueLu {
 
-  /*!
-    @class CoalesceDropFactory_kokkos
-    @brief Factory for creating a graph based on a given matrix.
-
-    Factory for creating graphs from matrices with entries selectively dropped.
-    This factory combines the functionality of CoalesceDropFactory and FilteredAFactory from the non-Kokkos
-    code path.
-
-    For an in-depth discussion, see https://github.com/trilinos/Trilinos/issues/1676.
-
-    ## Code paths ##
-
-    Both the classic dropping strategy as well as a coordinate-based distance
-    laplacian method is implemented. For performance reasons there are four
-    distinctive code paths for the classical method:
-
-    - one DOF per node without dropping (i.e. "aggregation: drop tol" = 0.0)
-    - one DOF per node with dropping (i.e. "aggregation: drop tol" > 0.0)
-    - DOFs per node > 1 withouth dropping
-    - DOFs per node > 1 with dropping
-
-    Additionally there is a code path for the distance-laplacian mode.
-
-    ## Input/output of CoalesceDropFactory_kokkos ##
-
-    ### User parameters of CoalesceDropFactory_kokkos ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A              | Factory | null |   | * | * | Generating factory of the operator A
-     UnAmalgamationInfo        | Factory | null |   | * | * | Generating factory of type AmalgamationFactory which generates the variable 'UnAmalgamationInfo'. Do not change the default unless you know what you are doing.
-     Coordinates               | Factory | null |   | * | (*) | Generating factory for variable 'Coordinates'. The coordinates are only needed if "distance laplacian" is chosen for the parameter "aggregation: drop scheme"
-     "aggregation: drop scheme" | std::string | "classical" | * | * |   | Coalescing algorithm. You can choose either "classical" (=default) or "distance laplacian"
-     "aggregation: drop tol" | double | 0.0 | * | * |   | Threshold parameter for dropping small entries
-     "aggregation: Dirichlet threshold" | double | 0.0 | * | * |   | Threshold for determining whether entries are zero during Dirichlet row detection
-     "lightweight wrap" | bool | true |   |   *  |   | hidden switch between fast implementation based on MueLu::LWGraph and a failsafe slower implementation based on Xpetra::Graph (for comparison). The user should not change the default value (=true)
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoalesceDropFactory_kokkos::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see CoalesceDropFactory_kokkos::DeclareInput).
-
-    ### Variables provided by UncoupledAggregationFactory ###
-
-    After CoalesceDropFactory_kokkos::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    Graph   | CoalesceDropFactory_kokkos   | Graph of matrix A
-    DofsPerNode | CoalesceDropFactory_kokkos | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
-
-    ## Amalgamation process ##
-
-    The CoalesceDropFactory_kokkos is internally using the AmalgamationFactory
-    for amalgamating the dof-based maps to node-based maps. The
-    AmalgamationFactory creates the "UnAmalgamationInfo" container which
-    basically stores all the necessary information for translating dof based
-    data to node based data and vice versa. The container is used, since this
-    way the amalgamation is only done once and later reused by other factories.
-
-    Of course, often one does not need the information from the
-    "UnAmalgamationInfo" container since the same information could be
-    extracted of the "Graph" or the map from the "Coordinates" vector.
-    However, there are also some situations (e.g. when doing rebalancing based
-    on HyperGraph partitioning without coordinate information) where one has
-    not access to a "Graph" or "Coordinates" variable.
-  */
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoalesceDropFactory_kokkos;
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public SingleLevelFactoryBase {
-  public:
-    using local_ordinal_type  = LocalOrdinal;
-    using global_ordinal_type = GlobalOrdinal;
-    using execution_space     = typename DeviceType::execution_space;
-    using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
-    using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
-
-  private:
-    // For compatibility
-    using Node                = node_type;
+/*!
+  @class CoalesceDropFactory_kokkos
+  @brief Factory for creating a graph based on a given matrix.
+
+  Factory for creating graphs from matrices with entries selectively dropped.
+  This factory combines the functionality of CoalesceDropFactory and FilteredAFactory from the non-Kokkos
+  code path.
+
+  For an in-depth discussion, see https://github.com/trilinos/Trilinos/issues/1676.
+
+  ## Code paths ##
+
+  Both the classic dropping strategy as well as a coordinate-based distance
+  laplacian method is implemented. For performance reasons there are four
+  distinctive code paths for the classical method:
+
+  - one DOF per node without dropping (i.e. "aggregation: drop tol" = 0.0)
+  - one DOF per node with dropping (i.e. "aggregation: drop tol" > 0.0)
+  - DOFs per node > 1 withouth dropping
+  - DOFs per node > 1 with dropping
+
+  Additionally there is a code path for the distance-laplacian mode.
+
+  ## Input/output of CoalesceDropFactory_kokkos ##
+
+  ### User parameters of CoalesceDropFactory_kokkos ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A              | Factory | null |   | * | * | Generating factory of the operator A
+   UnAmalgamationInfo        | Factory | null |   | * | * | Generating factory of type AmalgamationFactory which generates the variable 'UnAmalgamationInfo'. Do not change the default unless you know what you are doing.
+   Coordinates               | Factory | null |   | * | (*) | Generating factory for variable 'Coordinates'. The coordinates are only needed if "distance laplacian" is chosen for the parameter "aggregation: drop scheme"
+   "aggregation: drop scheme" | std::string | "classical" | * | * |   | Coalescing algorithm. You can choose either "classical" (=default) or "distance laplacian"
+   "aggregation: drop tol" | double | 0.0 | * | * |   | Threshold parameter for dropping small entries
+   "aggregation: Dirichlet threshold" | double | 0.0 | * | * |   | Threshold for determining whether entries are zero during Dirichlet row detection
+   "lightweight wrap" | bool | true |   |   *  |   | hidden switch between fast implementation based on MueLu::LWGraph and a failsafe slower implementation based on Xpetra::Graph (for comparison). The user should not change the default value (=true)
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoalesceDropFactory_kokkos::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see CoalesceDropFactory_kokkos::DeclareInput).
+
+  ### Variables provided by UncoupledAggregationFactory ###
+
+  After CoalesceDropFactory_kokkos::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  Graph   | CoalesceDropFactory_kokkos   | Graph of matrix A
+  DofsPerNode | CoalesceDropFactory_kokkos | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
+
+  ## Amalgamation process ##
+
+  The CoalesceDropFactory_kokkos is internally using the AmalgamationFactory
+  for amalgamating the dof-based maps to node-based maps. The
+  AmalgamationFactory creates the "UnAmalgamationInfo" container which
+  basically stores all the necessary information for translating dof based
+  data to node based data and vice versa. The container is used, since this
+  way the amalgamation is only done once and later reused by other factories.
+
+  Of course, often one does not need the information from the
+  "UnAmalgamationInfo" container since the same information could be
+  extracted of the "Graph" or the map from the "Coordinates" vector.
+  However, there are also some situations (e.g. when doing rebalancing based
+  on HyperGraph partitioning without coordinate information) where one has
+  not access to a "Graph" or "Coordinates" variable.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoalesceDropFactory_kokkos;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public SingleLevelFactoryBase {
+ public:
+  using local_ordinal_type  = LocalOrdinal;
+  using global_ordinal_type = GlobalOrdinal;
+  using execution_space     = typename DeviceType::execution_space;
+  using range_type          = Kokkos::RangePolicy<local_ordinal_type, execution_space>;
+  using node_type           = Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>;
+
+ private:
+  // For compatibility
+  using Node = node_type;
 #undef MUELU_COALESCEDROPFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  CoalesceDropFactory_kokkos() {}
 
-    //! Constructor
-    CoalesceDropFactory_kokkos() { }
+  //! Destructor
+  virtual ~CoalesceDropFactory_kokkos() {}
 
-    //! Destructor
-    virtual ~CoalesceDropFactory_kokkos() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //@}
+  void Build(Level& currentLevel) const;
+};
 
-    void Build(Level& currentLevel) const;
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COALESCEDROPFACTORY_KOKKOS_SHORT
-#endif // MUELU_COALESCEDROPFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_COALESCEDROPFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_def.hpp
index e7e613aa1d93..f6aa0cca78fa 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_CoalesceDropFactory_kokkos_def.hpp
@@ -63,620 +63,609 @@
 
 namespace MueLu {
 
+namespace CoalesceDrop_Kokkos_Details {  // anonymous
+
+template <class LO, class RowType>
+class ScanFunctor {
+ public:
+  ScanFunctor(RowType rows_)
+    : rows(rows_) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO i, LO& upd, const bool& final) const {
+    upd += rows(i);
+    if (final)
+      rows(i) = upd;
+  }
 
-  namespace CoalesceDrop_Kokkos_Details { // anonymous
+ private:
+  RowType rows;
+};
+
+template <class LO, class GhostedViewType>
+class ClassicalDropFunctor {
+ private:
+  typedef typename GhostedViewType::value_type SC;
+  typedef Kokkos::ArithTraits<SC> ATS;
+  typedef typename ATS::magnitudeType magnitudeType;
+
+  GhostedViewType diag;  // corresponds to overlapped diagonal multivector (2D View)
+  magnitudeType eps;
+
+ public:
+  ClassicalDropFunctor(GhostedViewType ghostedDiag, magnitudeType threshold)
+    : diag(ghostedDiag)
+    , eps(threshold) {}
+
+  // Return true if we drop, false if not
+  KOKKOS_FORCEINLINE_FUNCTION
+  bool operator()(LO row, LO col, SC val) const {
+    // We avoid square root by using squared values
+    auto aiiajj = ATS::magnitude(diag(row, 0)) * ATS::magnitude(diag(col, 0));  // |a_ii|*|a_jj|
+    auto aij2   = ATS::magnitude(val) * ATS::magnitude(val);                    // |a_ij|^2
+
+    return (aij2 <= eps * eps * aiiajj);
+  }
+};
+
+template <class LO, class CoordsType>
+class DistanceFunctor {
+ private:
+  typedef typename CoordsType::value_type SC;
+  typedef Kokkos::ArithTraits<SC> ATS;
+  typedef typename ATS::magnitudeType magnitudeType;
+
+ public:
+  typedef SC value_type;
+
+ public:
+  DistanceFunctor(CoordsType coords_)
+    : coords(coords_) {}
+
+  KOKKOS_INLINE_FUNCTION
+  magnitudeType distance2(LO row, LO col) const {
+    SC d = ATS::zero(), s;
+    for (size_t j = 0; j < coords.extent(1); j++) {
+      s = coords(row, j) - coords(col, j);
+      d += s * s;
+    }
+    return ATS::magnitude(d);
+  }
 
-    template<class LO, class RowType>
-    class ScanFunctor {
-    public:
-      ScanFunctor(RowType rows_) : rows(rows_) { }
+ private:
+  CoordsType coords;
+};
+
+template <class LO, class GhostedViewType, class DistanceFunctor>
+class DistanceLaplacianDropFunctor {
+ private:
+  typedef typename GhostedViewType::value_type SC;
+  typedef Kokkos::ArithTraits<SC> ATS;
+  typedef typename ATS::magnitudeType magnitudeType;
+
+ public:
+  DistanceLaplacianDropFunctor(GhostedViewType ghostedLaplDiag, DistanceFunctor distFunctor_, magnitudeType threshold)
+    : diag(ghostedLaplDiag)
+    , distFunctor(distFunctor_)
+    , eps(threshold) {}
+
+  // Return true if we drop, false if not
+  KOKKOS_INLINE_FUNCTION
+  bool operator()(LO row, LO col, SC /* val */) const {
+    // We avoid square root by using squared values
+
+    // We ignore incoming value of val as we operate on an auxiliary
+    // distance Laplacian matrix
+    typedef typename DistanceFunctor::value_type dSC;
+    typedef Kokkos::ArithTraits<dSC> dATS;
+    auto fval = dATS::one() / distFunctor.distance2(row, col);
+
+    auto aiiajj = ATS::magnitude(diag(row, 0)) * ATS::magnitude(diag(col, 0));  // |a_ii|*|a_jj|
+    auto aij2   = ATS::magnitude(fval) * ATS::magnitude(fval);                  // |a_ij|^2
+
+    return (aij2 <= eps * eps * aiiajj);
+  }
 
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LO i, LO& upd, const bool& final) const {
-        upd += rows(i);
-        if (final)
-          rows(i) = upd;
-      }
+ private:
+  GhostedViewType diag;  // corresponds to overlapped diagonal multivector (2D View)
+  DistanceFunctor distFunctor;
+  magnitudeType eps;
+};
+
+template <class SC, class LO, class MatrixType, class BndViewType, class DropFunctorType>
+class ScalarFunctor {
+ private:
+  typedef typename MatrixType::StaticCrsGraphType graph_type;
+  typedef typename graph_type::row_map_type rows_type;
+  typedef typename graph_type::entries_type cols_type;
+  typedef typename MatrixType::values_type vals_type;
+  typedef Kokkos::ArithTraits<SC> ATS;
+  typedef typename ATS::val_type impl_Scalar;
+  typedef Kokkos::ArithTraits<impl_Scalar> impl_ATS;
+  typedef typename ATS::magnitudeType magnitudeType;
+
+ public:
+  ScalarFunctor(MatrixType A_, BndViewType bndNodes_, DropFunctorType dropFunctor_,
+                typename rows_type::non_const_type rows_,
+                typename cols_type::non_const_type colsAux_,
+                typename vals_type::non_const_type valsAux_,
+                bool reuseGraph_, bool lumping_, SC /* threshold_ */,
+                bool aggregationMayCreateDirichlet_)
+    : A(A_)
+    , bndNodes(bndNodes_)
+    , dropFunctor(dropFunctor_)
+    , rows(rows_)
+    , colsAux(colsAux_)
+    , valsAux(valsAux_)
+    , reuseGraph(reuseGraph_)
+    , lumping(lumping_)
+    , aggregationMayCreateDirichlet(aggregationMayCreateDirichlet_) {
+    rowsA = A.graph.row_map;
+    zero  = impl_ATS::zero();
+  }
 
-    private:
-      RowType rows;
-    };
-
-    template<class LO, class GhostedViewType>
-    class ClassicalDropFunctor {
-    private:
-      typedef typename GhostedViewType::value_type      SC;
-      typedef          Kokkos::ArithTraits<SC>          ATS;
-      typedef typename ATS::magnitudeType               magnitudeType;
-
-      GhostedViewType   diag;   // corresponds to overlapped diagonal multivector (2D View)
-      magnitudeType     eps;
-
-    public:
-      ClassicalDropFunctor(GhostedViewType ghostedDiag, magnitudeType threshold) :
-          diag(ghostedDiag),
-          eps(threshold)
-      { }
-
-      // Return true if we drop, false if not
-      KOKKOS_FORCEINLINE_FUNCTION
-      bool operator()(LO row, LO col, SC val) const {
-        // We avoid square root by using squared values
-        auto aiiajj = ATS::magnitude(diag(row, 0)) * ATS::magnitude(diag(col, 0));   // |a_ii|*|a_jj|
-        auto aij2   = ATS::magnitude(val)          * ATS::magnitude(val);            // |a_ij|^2
-
-        return (aij2 <= eps*eps * aiiajj);
-      }
-    };
-
-    template<class LO, class CoordsType>
-    class DistanceFunctor {
-    private:
-      typedef typename CoordsType::value_type           SC;
-      typedef          Kokkos::ArithTraits<SC>          ATS;
-      typedef typename ATS::magnitudeType               magnitudeType;
-
-    public:
-      typedef SC value_type;
-
-    public:
-      DistanceFunctor(CoordsType coords_) : coords(coords_) { }
-
-      KOKKOS_INLINE_FUNCTION
-      magnitudeType distance2(LO row, LO col) const {
-        SC d = ATS::zero(), s;
-        for (size_t j = 0; j < coords.extent(1); j++) {
-          s = coords(row,j) - coords(col,j);
-          d += s*s;
-        }
-        return ATS::magnitude(d);
-      }
-    private:
-      CoordsType coords;
-    };
-
-    template<class LO, class GhostedViewType, class DistanceFunctor>
-    class DistanceLaplacianDropFunctor {
-    private:
-      typedef typename GhostedViewType::value_type      SC;
-      typedef          Kokkos::ArithTraits<SC>          ATS;
-      typedef typename ATS::magnitudeType               magnitudeType;
-
-    public:
-      DistanceLaplacianDropFunctor(GhostedViewType ghostedLaplDiag, DistanceFunctor distFunctor_, magnitudeType threshold) :
-          diag(ghostedLaplDiag),
-          distFunctor(distFunctor_),
-          eps(threshold)
-      { }
-
-      // Return true if we drop, false if not
-      KOKKOS_INLINE_FUNCTION
-      bool operator()(LO row, LO col, SC /* val */) const {
-        // We avoid square root by using squared values
-
-        // We ignore incoming value of val as we operate on an auxiliary
-        // distance Laplacian matrix
-        typedef typename DistanceFunctor::value_type      dSC;
-        typedef          Kokkos::ArithTraits<dSC>         dATS;
-        auto fval = dATS::one() / distFunctor.distance2(row, col);
-
-        auto aiiajj = ATS::magnitude(diag(row, 0)) * ATS::magnitude(diag(col, 0));   // |a_ii|*|a_jj|
-        auto aij2   = ATS::magnitude(fval)         * ATS::magnitude(fval);           // |a_ij|^2
-
-        return (aij2 <= eps*eps * aiiajj);
-      }
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO row, LO& nnz) const {
+    auto rowView = A.rowConst(row);
+    auto length  = rowView.length;
+    auto offset  = rowsA(row);
 
-    private:
-      GhostedViewType   diag;   // corresponds to overlapped diagonal multivector (2D View)
-      DistanceFunctor   distFunctor;
-      magnitudeType     eps;
-    };
-
-    template<class SC, class LO, class MatrixType, class BndViewType, class DropFunctorType>
-    class ScalarFunctor {
-    private:
-      typedef typename MatrixType::StaticCrsGraphType   graph_type;
-      typedef typename graph_type::row_map_type         rows_type;
-      typedef typename graph_type::entries_type         cols_type;
-      typedef typename MatrixType::values_type          vals_type;
-      typedef          Kokkos::ArithTraits<SC>          ATS;
-      typedef typename ATS::val_type                    impl_Scalar;
-      typedef Kokkos::ArithTraits<impl_Scalar>          impl_ATS;
-      typedef typename ATS::magnitudeType               magnitudeType;
-
-    public:
-      ScalarFunctor(MatrixType A_, BndViewType bndNodes_, DropFunctorType dropFunctor_,
-                    typename rows_type::non_const_type rows_,
-                    typename cols_type::non_const_type colsAux_,
-                    typename vals_type::non_const_type valsAux_,
-                    bool reuseGraph_, bool lumping_, SC /* threshold_ */,
-                    bool aggregationMayCreateDirichlet_ ) :
-          A(A_),
-          bndNodes(bndNodes_),
-          dropFunctor(dropFunctor_),
-          rows(rows_),
-          colsAux(colsAux_),
-          valsAux(valsAux_),
-          reuseGraph(reuseGraph_),
-          lumping(lumping_),
-          aggregationMayCreateDirichlet(aggregationMayCreateDirichlet_)
-      {
-        rowsA = A.graph.row_map;
-        zero = impl_ATS::zero();
-      }
+    impl_Scalar diag = zero;
+    LO rownnz        = 0;
+    LO diagID        = -1;
+    for (decltype(length) colID = 0; colID < length; colID++) {
+      LO col          = rowView.colidx(colID);
+      impl_Scalar val = rowView.value(colID);
 
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LO row, LO& nnz) const {
-        auto rowView = A.rowConst(row);
-        auto length  = rowView.length;
-        auto offset  = rowsA(row);
-
-        impl_Scalar diag = zero;
-        LO rownnz = 0;
-        LO diagID = -1;
-        for (decltype(length) colID = 0; colID < length; colID++) {
-          LO col = rowView.colidx(colID);
-          impl_Scalar val = rowView.value (colID);
-
-          if (!dropFunctor(row, col, rowView.value(colID)) || row == col) {
-            colsAux(offset+rownnz) = col;
-
-            LO valID = (reuseGraph ? colID : rownnz);
-            valsAux(offset+valID) = val;
-            if (row == col)
-              diagID = valID;
-
-            rownnz++;
-
-          } else {
-            // Rewrite with zeros (needed for reuseGraph)
-            valsAux(offset+colID) = zero;
-            diag += val;
-          }
-        }
-        // How to assert on the device?
-        // assert(diagIndex != -1);
-        rows(row+1) = rownnz;
-        // if (lumping && diagID != -1) {
-        if (lumping) {
-          // Add diag to the diagonal
-
-          // NOTE_KOKKOS: valsAux was allocated with
-          // ViewAllocateWithoutInitializing. This is not a problem here
-          // because we explicitly set this value above.
-          valsAux(offset+diagID) += diag;
-        }
+      if (!dropFunctor(row, col, rowView.value(colID)) || row == col) {
+        colsAux(offset + rownnz) = col;
 
-        // If the only element remaining after filtering is diagonal, mark node as boundary
-        // FIXME: this should really be replaced by the following
-        //    if (indices.size() == 1 && indices[0] == row)
-        //        boundaryNodes[row] = true;
-        // We do not do it this way now because there is no framework for distinguishing isolated
-        // and boundary nodes in the aggregation algorithms
-        bndNodes(row) = (rownnz == 1 && aggregationMayCreateDirichlet);
+        LO valID                = (reuseGraph ? colID : rownnz);
+        valsAux(offset + valID) = val;
+        if (row == col)
+          diagID = valID;
 
-        nnz += rownnz;
-      }
+        rownnz++;
 
-    private:
-      MatrixType                            A;
-      BndViewType                           bndNodes;
-      DropFunctorType                       dropFunctor;
-
-      rows_type                             rowsA;
-
-      typename rows_type::non_const_type    rows;
-      typename cols_type::non_const_type    colsAux;
-      typename vals_type::non_const_type    valsAux;
-
-      bool                                  reuseGraph;
-      bool                                  lumping;
-      bool                                  aggregationMayCreateDirichlet;
-      impl_Scalar                           zero;
-    };
-
-    // collect number nonzeros of blkSize rows in nnz_(row+1)
-    template<class MatrixType, class NnzType, class blkSizeType>
-    class Stage1aVectorFunctor {
-    private:
-      typedef typename MatrixType::ordinal_type LO;
-
-    public:
-      Stage1aVectorFunctor(MatrixType kokkosMatrix_, NnzType nnz_, blkSizeType blkSize_) :
-        kokkosMatrix(kokkosMatrix_),
-        nnz(nnz_),
-        blkSize(blkSize_) { }
-
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LO row, LO& totalnnz) const {
-
-        // the following code is more or less what MergeRows is doing
-        // count nonzero entries in all dof rows associated with node row
-        LO nodeRowMaxNonZeros = 0;
-        for (LO j = 0; j < blkSize; j++) {
-          auto rowView = kokkosMatrix.row(row * blkSize + j);
-          nodeRowMaxNonZeros += rowView.length;
-        }
-        nnz(row + 1) = nodeRowMaxNonZeros;
-        totalnnz += nodeRowMaxNonZeros;
+      } else {
+        // Rewrite with zeros (needed for reuseGraph)
+        valsAux(offset + colID) = zero;
+        diag += val;
       }
+    }
+    // How to assert on the device?
+    // assert(diagIndex != -1);
+    rows(row + 1) = rownnz;
+    // if (lumping && diagID != -1) {
+    if (lumping) {
+      // Add diag to the diagonal
+
+      // NOTE_KOKKOS: valsAux was allocated with
+      // ViewAllocateWithoutInitializing. This is not a problem here
+      // because we explicitly set this value above.
+      valsAux(offset + diagID) += diag;
+    }
 
+    // If the only element remaining after filtering is diagonal, mark node as boundary
+    // FIXME: this should really be replaced by the following
+    //    if (indices.size() == 1 && indices[0] == row)
+    //        boundaryNodes[row] = true;
+    // We do not do it this way now because there is no framework for distinguishing isolated
+    // and boundary nodes in the aggregation algorithms
+    bndNodes(row) = (rownnz == 1 && aggregationMayCreateDirichlet);
 
-    private:
-      MatrixType kokkosMatrix; //< local matrix part
-      NnzType nnz;             //< View containing number of nonzeros for current row
-      blkSizeType blkSize;     //< block size (or partial block size in strided maps)
-    };
-
-
-    // build the dof-based column map containing the local dof ids belonging to blkSize rows in matrix
-    // sort column ids
-    // translate them into (unique) node ids
-    // count the node column ids per node row
-    template<class MatrixType, class NnzType, class blkSizeType, class ColDofType, class Dof2NodeTranslationType, class BdryNodeTypeConst, class BdryNodeType, class boolType>
-    class Stage1bcVectorFunctor {
-    private:
-      typedef typename MatrixType::ordinal_type LO;
-
-    private:
-      MatrixType kokkosMatrix; //< local matrix part
-      NnzType coldofnnz; //< view containing start and stop indices for subviews
-      blkSizeType blkSize;     //< block size (or partial block size in strided maps)
-      ColDofType coldofs;      //< view containing the local dof ids associated with columns for the blkSize rows (not sorted)
-      Dof2NodeTranslationType dof2node; //< view containing the local node id associated with the local dof id
-      NnzType colnodennz; //< view containing number of column nodes for each node row
-      BdryNodeTypeConst  dirichletdof;  //< view containing with num dofs booleans. True if dof (not necessarily entire node) is dirichlet boundardy dof.
-      BdryNodeType  bdrynode;  //< view containing with numNodes booleans. True if node is (full) dirichlet boundardy node.
-      boolType usegreedydirichlet;    //< boolean for use of greedy Dirichlet (if any dof is Dirichlet, entire node is dirichlet) default false (need all dofs in node to be Dirichlet for node to be Dirichlet)
-
-    public:
-      Stage1bcVectorFunctor(MatrixType kokkosMatrix_,
-                           NnzType coldofnnz_,
-                           blkSizeType blkSize_,
-                           ColDofType coldofs_,
-                           Dof2NodeTranslationType dof2node_,
-                           NnzType colnodennz_,
-                           BdryNodeTypeConst dirichletdof_,
-                           BdryNodeType bdrynode_,
-                           boolType usegreedydirichlet_) :
-        kokkosMatrix(kokkosMatrix_),
-        coldofnnz(coldofnnz_),
-        blkSize(blkSize_),
-        coldofs(coldofs_),
-        dof2node(dof2node_),
-        colnodennz(colnodennz_),
-        dirichletdof(dirichletdof_),
-        bdrynode(bdrynode_),
-        usegreedydirichlet(usegreedydirichlet_) {
-      }
+    nnz += rownnz;
+  }
 
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LO rowNode, LO& nnz) const {
+ private:
+  MatrixType A;
+  BndViewType bndNodes;
+  DropFunctorType dropFunctor;
+
+  rows_type rowsA;
+
+  typename rows_type::non_const_type rows;
+  typename cols_type::non_const_type colsAux;
+  typename vals_type::non_const_type valsAux;
+
+  bool reuseGraph;
+  bool lumping;
+  bool aggregationMayCreateDirichlet;
+  impl_Scalar zero;
+};
+
+// collect number nonzeros of blkSize rows in nnz_(row+1)
+template <class MatrixType, class NnzType, class blkSizeType>
+class Stage1aVectorFunctor {
+ private:
+  typedef typename MatrixType::ordinal_type LO;
+
+ public:
+  Stage1aVectorFunctor(MatrixType kokkosMatrix_, NnzType nnz_, blkSizeType blkSize_)
+    : kokkosMatrix(kokkosMatrix_)
+    , nnz(nnz_)
+    , blkSize(blkSize_) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO row, LO& totalnnz) const {
+    // the following code is more or less what MergeRows is doing
+    // count nonzero entries in all dof rows associated with node row
+    LO nodeRowMaxNonZeros = 0;
+    for (LO j = 0; j < blkSize; j++) {
+      auto rowView = kokkosMatrix.row(row * blkSize + j);
+      nodeRowMaxNonZeros += rowView.length;
+    }
+    nnz(row + 1) = nodeRowMaxNonZeros;
+    totalnnz += nodeRowMaxNonZeros;
+  }
 
-        LO pos = coldofnnz(rowNode);
-        if( usegreedydirichlet ){
-          bdrynode(rowNode) = false;
-          for (LO j = 0; j < blkSize; j++) {
-            auto rowView = kokkosMatrix.row(rowNode * blkSize + j);
-            auto numIndices = rowView.length;
-
-            // if any dof in the node is Dirichlet
-            if( dirichletdof(rowNode * blkSize + j) )
-              bdrynode(rowNode) = true;
-
-            for (decltype(numIndices) k = 0; k < numIndices; k++) {
-              auto dofID = rowView.colidx(k);
-              coldofs(pos) = dofID;
-              pos ++;
-            }
-          }
-        }else{
+ private:
+  MatrixType kokkosMatrix;  //< local matrix part
+  NnzType nnz;              //< View containing number of nonzeros for current row
+  blkSizeType blkSize;      //< block size (or partial block size in strided maps)
+};
+
+// build the dof-based column map containing the local dof ids belonging to blkSize rows in matrix
+// sort column ids
+// translate them into (unique) node ids
+// count the node column ids per node row
+template <class MatrixType, class NnzType, class blkSizeType, class ColDofType, class Dof2NodeTranslationType, class BdryNodeTypeConst, class BdryNodeType, class boolType>
+class Stage1bcVectorFunctor {
+ private:
+  typedef typename MatrixType::ordinal_type LO;
+
+ private:
+  MatrixType kokkosMatrix;           //< local matrix part
+  NnzType coldofnnz;                 //< view containing start and stop indices for subviews
+  blkSizeType blkSize;               //< block size (or partial block size in strided maps)
+  ColDofType coldofs;                //< view containing the local dof ids associated with columns for the blkSize rows (not sorted)
+  Dof2NodeTranslationType dof2node;  //< view containing the local node id associated with the local dof id
+  NnzType colnodennz;                //< view containing number of column nodes for each node row
+  BdryNodeTypeConst dirichletdof;    //< view containing with num dofs booleans. True if dof (not necessarily entire node) is dirichlet boundardy dof.
+  BdryNodeType bdrynode;             //< view containing with numNodes booleans. True if node is (full) dirichlet boundardy node.
+  boolType usegreedydirichlet;       //< boolean for use of greedy Dirichlet (if any dof is Dirichlet, entire node is dirichlet) default false (need all dofs in node to be Dirichlet for node to be Dirichlet)
+
+ public:
+  Stage1bcVectorFunctor(MatrixType kokkosMatrix_,
+                        NnzType coldofnnz_,
+                        blkSizeType blkSize_,
+                        ColDofType coldofs_,
+                        Dof2NodeTranslationType dof2node_,
+                        NnzType colnodennz_,
+                        BdryNodeTypeConst dirichletdof_,
+                        BdryNodeType bdrynode_,
+                        boolType usegreedydirichlet_)
+    : kokkosMatrix(kokkosMatrix_)
+    , coldofnnz(coldofnnz_)
+    , blkSize(blkSize_)
+    , coldofs(coldofs_)
+    , dof2node(dof2node_)
+    , colnodennz(colnodennz_)
+    , dirichletdof(dirichletdof_)
+    , bdrynode(bdrynode_)
+    , usegreedydirichlet(usegreedydirichlet_) {
+  }
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO rowNode, LO& nnz) const {
+    LO pos = coldofnnz(rowNode);
+    if (usegreedydirichlet) {
+      bdrynode(rowNode) = false;
+      for (LO j = 0; j < blkSize; j++) {
+        auto rowView    = kokkosMatrix.row(rowNode * blkSize + j);
+        auto numIndices = rowView.length;
+
+        // if any dof in the node is Dirichlet
+        if (dirichletdof(rowNode * blkSize + j))
           bdrynode(rowNode) = true;
-          for (LO j = 0; j < blkSize; j++) {
-            auto rowView = kokkosMatrix.row(rowNode * blkSize + j);
-            auto numIndices = rowView.length;
-
-            // if any dof in the node is not Dirichlet
-            if( dirichletdof(rowNode * blkSize + j) == false )
-              bdrynode(rowNode) = false;
-
-            for (decltype(numIndices) k = 0; k < numIndices; k++) {
-              auto dofID = rowView.colidx(k);
-              coldofs(pos) = dofID;
-              pos ++;
-            }
-          }
-        }
 
-        // sort coldofs
-        LO begin = coldofnnz(rowNode);
-        LO end   = coldofnnz(rowNode+1);
-        LO n     = end - begin;
-        for (LO i = 0; i < (n-1); i++) {
-          for (LO j = 0; j < (n-i-1); j++) {
-            if (coldofs(j+begin) > coldofs(j+begin+1)) {
-              LO temp = coldofs(j+begin);
-              coldofs(j+begin) = coldofs(j+begin+1);
-              coldofs(j+begin+1) = temp;
-            }
-          }
+        for (decltype(numIndices) k = 0; k < numIndices; k++) {
+          auto dofID   = rowView.colidx(k);
+          coldofs(pos) = dofID;
+          pos++;
         }
-        size_t cnt = 0;
-        LO lastNodeID = -1;
-        for (LO i = 0; i < n; i++) {
-          LO dofID  = coldofs(begin + i);
-          LO nodeID = dof2node(dofID);
-          if(nodeID != lastNodeID) {
-            lastNodeID = nodeID;
-            coldofs(begin+cnt) = nodeID;
-            cnt++;
-          }
-        }
-        colnodennz(rowNode+1) = cnt;
-        nnz += cnt;
-      }
-
-    };
-
-    // fill column node id view
-    template<class MatrixType, class ColDofNnzType, class ColDofType, class ColNodeNnzType, class ColNodeType>
-    class Stage1dVectorFunctor {
-    private:
-      typedef typename MatrixType::ordinal_type LO;
-      typedef typename MatrixType::value_type SC;
-
-    private:
-      ColDofType     coldofs;    //< view containing mixed node and dof indices (only input)
-      ColDofNnzType  coldofnnz;  //< view containing the start and stop indices for subviews (dofs)
-      ColNodeType colnodes;      //< view containing the local node ids associated with columns
-      ColNodeNnzType colnodennz; //< view containing start and stop indices for subviews
-
-    public:
-      Stage1dVectorFunctor(ColDofType coldofs_, ColDofNnzType coldofnnz_, ColNodeType colnodes_, ColNodeNnzType colnodennz_) :
-        coldofs(coldofs_),
-        coldofnnz(coldofnnz_),
-        colnodes(colnodes_),
-        colnodennz(colnodennz_) {
       }
+    } else {
+      bdrynode(rowNode) = true;
+      for (LO j = 0; j < blkSize; j++) {
+        auto rowView    = kokkosMatrix.row(rowNode * blkSize + j);
+        auto numIndices = rowView.length;
 
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LO rowNode) const {
-        auto dofbegin  = coldofnnz(rowNode);
-        auto nodebegin = colnodennz(rowNode);
-        auto nodeend   = colnodennz(rowNode+1);
-        auto n = nodeend - nodebegin;
+        // if any dof in the node is not Dirichlet
+        if (dirichletdof(rowNode * blkSize + j) == false)
+          bdrynode(rowNode) = false;
 
-        for (decltype(nodebegin) i = 0; i < n; i++) {
-          colnodes(nodebegin + i) = coldofs(dofbegin + i);
+        for (decltype(numIndices) k = 0; k < numIndices; k++) {
+          auto dofID   = rowView.colidx(k);
+          coldofs(pos) = dofID;
+          pos++;
         }
       }
-    };
-
+    }
 
-  } // namespace
+    // sort coldofs
+    LO begin = coldofnnz(rowNode);
+    LO end   = coldofnnz(rowNode + 1);
+    LO n     = end - begin;
+    for (LO i = 0; i < (n - 1); i++) {
+      for (LO j = 0; j < (n - i - 1); j++) {
+        if (coldofs(j + begin) > coldofs(j + begin + 1)) {
+          LO temp                = coldofs(j + begin);
+          coldofs(j + begin)     = coldofs(j + begin + 1);
+          coldofs(j + begin + 1) = temp;
+        }
+      }
+    }
+    size_t cnt    = 0;
+    LO lastNodeID = -1;
+    for (LO i = 0; i < n; i++) {
+      LO dofID  = coldofs(begin + i);
+      LO nodeID = dof2node(dofID);
+      if (nodeID != lastNodeID) {
+        lastNodeID           = nodeID;
+        coldofs(begin + cnt) = nodeID;
+        cnt++;
+      }
+    }
+    colnodennz(rowNode + 1) = cnt;
+    nnz += cnt;
+  }
+};
+
+// fill column node id view
+template <class MatrixType, class ColDofNnzType, class ColDofType, class ColNodeNnzType, class ColNodeType>
+class Stage1dVectorFunctor {
+ private:
+  typedef typename MatrixType::ordinal_type LO;
+  typedef typename MatrixType::value_type SC;
+
+ private:
+  ColDofType coldofs;         //< view containing mixed node and dof indices (only input)
+  ColDofNnzType coldofnnz;    //< view containing the start and stop indices for subviews (dofs)
+  ColNodeType colnodes;       //< view containing the local node ids associated with columns
+  ColNodeNnzType colnodennz;  //< view containing start and stop indices for subviews
+
+ public:
+  Stage1dVectorFunctor(ColDofType coldofs_, ColDofNnzType coldofnnz_, ColNodeType colnodes_, ColNodeNnzType colnodennz_)
+    : coldofs(coldofs_)
+    , coldofnnz(coldofnnz_)
+    , colnodes(colnodes_)
+    , colnodennz(colnodennz_) {
+  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  RCP<const ParameterList> CoalesceDropFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO rowNode) const {
+    auto dofbegin  = coldofnnz(rowNode);
+    auto nodebegin = colnodennz(rowNode);
+    auto nodeend   = colnodennz(rowNode + 1);
+    auto n         = nodeend - nodebegin;
 
-#define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: drop tol");
-    SET_VALID_ENTRY("aggregation: Dirichlet threshold");
-    SET_VALID_ENTRY("aggregation: drop scheme");
-    SET_VALID_ENTRY("aggregation: dropping may create Dirichlet");
-    SET_VALID_ENTRY("aggregation: greedy Dirichlet");
-    SET_VALID_ENTRY("filtered matrix: use lumping");
-    SET_VALID_ENTRY("filtered matrix: reuse graph");
-    SET_VALID_ENTRY("filtered matrix: reuse eigenvalue");
-    SET_VALID_ENTRY("aggregation: use ml scaling of drop tol");
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      validParamList->getEntry("aggregation: drop scheme").setValidator(
-        rcp(new validatorType(Teuchos::tuple<std::string>("classical", "distance laplacian"), "aggregation: drop scheme")));
+    for (decltype(nodebegin) i = 0; i < n; i++) {
+      colnodes(nodebegin + i) = coldofs(dofbegin + i);
     }
-#undef  SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory for UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory for Coordinates");
-
-    return validParamList;
   }
+};
+
+}  // namespace CoalesceDrop_Kokkos_Details
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void CoalesceDropFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "UnAmalgamationInfo");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+RCP<const ParameterList> CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    const ParameterList& pL = GetParameterList();
-    if (pL.get<std::string>("aggregation: drop scheme") == "distance laplacian")
-      Input(currentLevel, "Coordinates");
+#define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
+  SET_VALID_ENTRY("aggregation: drop tol");
+  SET_VALID_ENTRY("aggregation: Dirichlet threshold");
+  SET_VALID_ENTRY("aggregation: drop scheme");
+  SET_VALID_ENTRY("aggregation: dropping may create Dirichlet");
+  SET_VALID_ENTRY("aggregation: greedy Dirichlet");
+  SET_VALID_ENTRY("filtered matrix: use lumping");
+  SET_VALID_ENTRY("filtered matrix: reuse graph");
+  SET_VALID_ENTRY("filtered matrix: reuse eigenvalue");
+  SET_VALID_ENTRY("aggregation: use ml scaling of drop tol");
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    validParamList->getEntry("aggregation: drop scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("classical", "distance laplacian"), "aggregation: drop scheme")));
   }
+#undef SET_VALID_ENTRY
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase>>("UnAmalgamationInfo", Teuchos::null, "Generating factory for UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase>>("Coordinates", Teuchos::null, "Generating factory for Coordinates");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void CoalesceDropFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
+  return validParamList;
+}
 
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType MT;
-    const MT zero = Teuchos::ScalarTraits<MT>::zero();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "UnAmalgamationInfo");
 
-    auto A         = Get< RCP<Matrix> >(currentLevel, "A");
+  const ParameterList& pL = GetParameterList();
+  if (pL.get<std::string>("aggregation: drop scheme") == "distance laplacian")
+    Input(currentLevel, "Coordinates");
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
 
-    /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
-       blkSize is the number of storage blocks that must kept together during the amalgamation process.
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType MT;
+  const MT zero = Teuchos::ScalarTraits<MT>::zero();
 
-       Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
+  auto A = Get<RCP<Matrix>>(currentLevel, "A");
 
-       numPDEs = blkSize * storageblocksize.
-       
-       If numPDEs==1
-         Matrix is point storage (classical CRS storage).  storageblocksize=1 and  blkSize=1
-         No other values makes sense.
+  /* NOTE: storageblocksize (from GetStorageBlockSize()) is the size of a block in the chosen storage scheme.
+     blkSize is the number of storage blocks that must kept together during the amalgamation process.
 
-       If numPDEs>1
-         If matrix uses point storage, then storageblocksize=1  and blkSize=numPDEs.
-         If matrix uses block storage, with block size of n, then storageblocksize=n, and blkSize=numPDEs/n.
-         Thus far, only storageblocksize=numPDEs and blkSize=1 has been tested.
-      */
- 
-    TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0,Exceptions::RuntimeError,"A->GetFixedBlockSize() needs to be a multiple of A->GetStorageBlockSize()");
-    LO   blkSize   = A->GetFixedBlockSize() / A->GetStorageBlockSize();
+     Both of these quantities may be different than numPDEs (from GetFixedBlockSize()), but the following must always hold:
 
-    auto amalInfo = Get< RCP<AmalgamationInfo> >(currentLevel, "UnAmalgamationInfo");
+     numPDEs = blkSize * storageblocksize.
 
-    const ParameterList& pL = GetParameterList();
+     If numPDEs==1
+       Matrix is point storage (classical CRS storage).  storageblocksize=1 and  blkSize=1
+       No other values makes sense.
 
-    // Sanity Checking: ML drop tol scaling is not supported in UncoupledAggregation_Kokkos
-    TEUCHOS_TEST_FOR_EXCEPTION( pL.get<bool>("aggregation: use ml scaling of drop tol"),std::invalid_argument,"Option: 'aggregation: use ml scaling of drop tol' is not supported in the Kokkos version of CoalesceDroPFactory");
+     If numPDEs>1
+       If matrix uses point storage, then storageblocksize=1  and blkSize=numPDEs.
+       If matrix uses block storage, with block size of n, then storageblocksize=n, and blkSize=numPDEs/n.
+       Thus far, only storageblocksize=numPDEs and blkSize=1 has been tested.
+    */
 
+  TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0, Exceptions::RuntimeError, "A->GetFixedBlockSize() needs to be a multiple of A->GetStorageBlockSize()");
+  LO blkSize = A->GetFixedBlockSize() / A->GetStorageBlockSize();
 
+  auto amalInfo = Get<RCP<AmalgamationInfo>>(currentLevel, "UnAmalgamationInfo");
 
-    std::string algo = pL.get<std::string>("aggregation: drop scheme");
+  const ParameterList& pL = GetParameterList();
 
-    double threshold = pL.get<double>("aggregation: drop tol");
-    GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold
-                         << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
+  // Sanity Checking: ML drop tol scaling is not supported in UncoupledAggregation_Kokkos
+  TEUCHOS_TEST_FOR_EXCEPTION(pL.get<bool>("aggregation: use ml scaling of drop tol"), std::invalid_argument, "Option: 'aggregation: use ml scaling of drop tol' is not supported in the Kokkos version of CoalesceDroPFactory");
 
-    const typename STS::magnitudeType dirichletThreshold =
-        STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
+  std::string algo = pL.get<std::string>("aggregation: drop scheme");
 
-    GO numDropped = 0, numTotal = 0;
+  double threshold = pL.get<double>("aggregation: drop tol");
+  GetOStream(Runtime0) << "algorithm = \"" << algo << "\": threshold = " << threshold
+                       << ", blocksize = " << A->GetFixedBlockSize() << std::endl;
 
-    RCP<LWGraph_kokkos> graph;
-    LO                  dofsPerNode = -1;
+  const typename STS::magnitudeType dirichletThreshold =
+      STS::magnitude(as<SC>(pL.get<double>("aggregation: Dirichlet threshold")));
 
-    typedef typename LWGraph_kokkos::boundary_nodes_type boundary_nodes_type;
-    boundary_nodes_type boundaryNodes;
+  GO numDropped = 0, numTotal = 0;
 
-    RCP<Matrix> filteredA;
-    if (blkSize == 1 && threshold == zero) {
-      // Scalar problem without dropping
+  RCP<LWGraph_kokkos> graph;
+  LO dofsPerNode = -1;
 
-      // Detect and record rows that correspond to Dirichlet boundary conditions
-      boundaryNodes = Utilities::DetectDirichletRows_kokkos(*A, dirichletThreshold);
+  typedef typename LWGraph_kokkos::boundary_nodes_type boundary_nodes_type;
+  boundary_nodes_type boundaryNodes;
 
-      // Trivial LWGraph construction
-      graph = rcp(new LWGraph_kokkos(A->getCrsGraph()->getLocalGraphDevice(), A->getRowMap(), A->getColMap(), "graph of A"));
-      graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
+  RCP<Matrix> filteredA;
+  if (blkSize == 1 && threshold == zero) {
+    // Scalar problem without dropping
 
-      numTotal = A->getLocalNumEntries();
-      dofsPerNode = 1;
+    // Detect and record rows that correspond to Dirichlet boundary conditions
+    boundaryNodes = Utilities::DetectDirichletRows_kokkos(*A, dirichletThreshold);
 
-      filteredA = A;
+    // Trivial LWGraph construction
+    graph = rcp(new LWGraph_kokkos(A->getCrsGraph()->getLocalGraphDevice(), A->getRowMap(), A->getColMap(), "graph of A"));
+    graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
 
-    } else if (blkSize == 1 && threshold != zero) {
-      // Scalar problem with dropping
+    numTotal    = A->getLocalNumEntries();
+    dofsPerNode = 1;
 
-      typedef typename Matrix::local_matrix_type          local_matrix_type;
-      typedef typename LWGraph_kokkos::local_graph_type   kokkos_graph_type;
-      typedef typename kokkos_graph_type::row_map_type::non_const_type    rows_type;
-      typedef typename kokkos_graph_type::entries_type::non_const_type    cols_type;
-      typedef typename local_matrix_type::values_type::non_const_type     vals_type;
+    filteredA = A;
 
-      LO   numRows      = A->getLocalNumRows();
-      local_matrix_type kokkosMatrix = A->getLocalMatrixDevice();
-      auto nnzA  = kokkosMatrix.nnz();
-      auto rowsA = kokkosMatrix.graph.row_map;
+  } else if (blkSize == 1 && threshold != zero) {
+    // Scalar problem with dropping
 
+    typedef typename Matrix::local_matrix_type local_matrix_type;
+    typedef typename LWGraph_kokkos::local_graph_type kokkos_graph_type;
+    typedef typename kokkos_graph_type::row_map_type::non_const_type rows_type;
+    typedef typename kokkos_graph_type::entries_type::non_const_type cols_type;
+    typedef typename local_matrix_type::values_type::non_const_type vals_type;
 
-      typedef Kokkos::ArithTraits<SC>          ATS;
-      typedef typename ATS::val_type           impl_Scalar;
-      typedef Kokkos::ArithTraits<impl_Scalar> impl_ATS;
+    LO numRows                     = A->getLocalNumRows();
+    local_matrix_type kokkosMatrix = A->getLocalMatrixDevice();
+    auto nnzA                      = kokkosMatrix.nnz();
+    auto rowsA                     = kokkosMatrix.graph.row_map;
 
-      bool reuseGraph = pL.get<bool>("filtered matrix: reuse graph");
-      bool lumping    = pL.get<bool>("filtered matrix: use lumping");
-      if (lumping)
-        GetOStream(Runtime0) << "Lumping dropped entries" << std::endl;
+    typedef Kokkos::ArithTraits<SC> ATS;
+    typedef typename ATS::val_type impl_Scalar;
+    typedef Kokkos::ArithTraits<impl_Scalar> impl_ATS;
 
-      const bool aggregationMayCreateDirichlet = pL.get<bool>("aggregation: dropping may create Dirichlet");
+    bool reuseGraph = pL.get<bool>("filtered matrix: reuse graph");
+    bool lumping    = pL.get<bool>("filtered matrix: use lumping");
+    if (lumping)
+      GetOStream(Runtime0) << "Lumping dropped entries" << std::endl;
 
-      // FIXME_KOKKOS: replace by ViewAllocateWithoutInitializing + setting a single value
-      rows_type rows   ("FA_rows",     numRows+1);
-      cols_type colsAux(Kokkos::ViewAllocateWithoutInitializing("FA_aux_cols"), nnzA);
-      vals_type valsAux;
-      if (reuseGraph) {
-        SubFactoryMonitor m2(*this, "CopyMatrix", currentLevel);
+    const bool aggregationMayCreateDirichlet = pL.get<bool>("aggregation: dropping may create Dirichlet");
 
-        // Share graph with the original matrix
-        filteredA = MatrixFactory::Build(A->getCrsGraph());
+    // FIXME_KOKKOS: replace by ViewAllocateWithoutInitializing + setting a single value
+    rows_type rows("FA_rows", numRows + 1);
+    cols_type colsAux(Kokkos::ViewAllocateWithoutInitializing("FA_aux_cols"), nnzA);
+    vals_type valsAux;
+    if (reuseGraph) {
+      SubFactoryMonitor m2(*this, "CopyMatrix", currentLevel);
 
-        // Do a no-op fill-complete
-        RCP<ParameterList> fillCompleteParams(new ParameterList);
-        fillCompleteParams->set("No Nonlocal Changes", true);
-        filteredA->fillComplete(fillCompleteParams);
+      // Share graph with the original matrix
+      filteredA = MatrixFactory::Build(A->getCrsGraph());
 
-        // No need to reuseFill, just modify in place
-        valsAux = filteredA->getLocalMatrixDevice().values;
+      // Do a no-op fill-complete
+      RCP<ParameterList> fillCompleteParams(new ParameterList);
+      fillCompleteParams->set("No Nonlocal Changes", true);
+      filteredA->fillComplete(fillCompleteParams);
 
-      } else {
-        // Need an extra array to compress
-        valsAux = vals_type(Kokkos::ViewAllocateWithoutInitializing("FA_aux_vals"), nnzA);
-      }
+      // No need to reuseFill, just modify in place
+      valsAux = filteredA->getLocalMatrixDevice().values;
 
-      typename boundary_nodes_type::non_const_type bndNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numRows);
+    } else {
+      // Need an extra array to compress
+      valsAux = vals_type(Kokkos::ViewAllocateWithoutInitializing("FA_aux_vals"), nnzA);
+    }
 
-      LO nnzFA = 0;
-      {
-        if (algo == "classical") {
-          // Construct overlapped matrix diagonal
-          RCP<Vector> ghostedDiag;
-          {
-            kokkosMatrix = local_matrix_type();
-            SubFactoryMonitor m2(*this, "Ghosted diag construction", currentLevel);
-            ghostedDiag     = Utilities::GetMatrixOverlappedDiagonal(*A);
-            kokkosMatrix=A->getLocalMatrixDevice();
-          }
+    typename boundary_nodes_type::non_const_type bndNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numRows);
 
-          // Filter out entries
-          {
-            SubFactoryMonitor m2(*this, "MainLoop", currentLevel);
+    LO nnzFA = 0;
+    {
+      if (algo == "classical") {
+        // Construct overlapped matrix diagonal
+        RCP<Vector> ghostedDiag;
+        {
+          kokkosMatrix = local_matrix_type();
+          SubFactoryMonitor m2(*this, "Ghosted diag construction", currentLevel);
+          ghostedDiag  = Utilities::GetMatrixOverlappedDiagonal(*A);
+          kokkosMatrix = A->getLocalMatrixDevice();
+        }
 
-            auto ghostedDiagView = ghostedDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
+        // Filter out entries
+        {
+          SubFactoryMonitor m2(*this, "MainLoop", currentLevel);
 
-            CoalesceDrop_Kokkos_Details::ClassicalDropFunctor<LO, decltype(ghostedDiagView)> dropFunctor(ghostedDiagView, threshold);
-            CoalesceDrop_Kokkos_Details::ScalarFunctor<typename ATS::val_type, LO, local_matrix_type, decltype(bndNodes), decltype(dropFunctor)>
+          auto ghostedDiagView = ghostedDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+          CoalesceDrop_Kokkos_Details::ClassicalDropFunctor<LO, decltype(ghostedDiagView)> dropFunctor(ghostedDiagView, threshold);
+          CoalesceDrop_Kokkos_Details::ScalarFunctor<typename ATS::val_type, LO, local_matrix_type, decltype(bndNodes), decltype(dropFunctor)>
               scalarFunctor(kokkosMatrix, bndNodes, dropFunctor, rows, colsAux, valsAux, reuseGraph, lumping, threshold, aggregationMayCreateDirichlet);
 
-            Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:main_loop", range_type(0,numRows),
-                                    scalarFunctor, nnzFA);
-          }
+          Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:main_loop", range_type(0, numRows),
+                                  scalarFunctor, nnzFA);
+        }
 
-        } else if (algo == "distance laplacian") {
-          typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> doubleMultiVector;
-          auto coords = Get<RCP<doubleMultiVector> >(currentLevel, "Coordinates");
+      } else if (algo == "distance laplacian") {
+        typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> doubleMultiVector;
+        auto coords = Get<RCP<doubleMultiVector>>(currentLevel, "Coordinates");
 
-          auto uniqueMap    = A->getRowMap();
-          auto nonUniqueMap = A->getColMap();
+        auto uniqueMap    = A->getRowMap();
+        auto nonUniqueMap = A->getColMap();
 
-          // Construct ghosted coordinates
-          RCP<const Import> importer;
-          {
-            SubFactoryMonitor m2(*this, "Coords Import construction", currentLevel);
-            importer = ImportFactory::Build(uniqueMap, nonUniqueMap);
-          }
-          RCP<doubleMultiVector> ghostedCoords;
-          {
-            SubFactoryMonitor m2(*this, "Ghosted coords construction", currentLevel);
-            ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(nonUniqueMap, coords->getNumVectors());
-            ghostedCoords->doImport(*coords, *importer, Xpetra::INSERT);
-          }
+        // Construct ghosted coordinates
+        RCP<const Import> importer;
+        {
+          SubFactoryMonitor m2(*this, "Coords Import construction", currentLevel);
+          importer = ImportFactory::Build(uniqueMap, nonUniqueMap);
+        }
+        RCP<doubleMultiVector> ghostedCoords;
+        {
+          SubFactoryMonitor m2(*this, "Ghosted coords construction", currentLevel);
+          ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(nonUniqueMap, coords->getNumVectors());
+          ghostedCoords->doImport(*coords, *importer, Xpetra::INSERT);
+        }
 
-          auto ghostedCoordsView = ghostedCoords->getDeviceLocalView(Xpetra::Access::ReadWrite);
-          CoalesceDrop_Kokkos_Details::DistanceFunctor<LO, decltype(ghostedCoordsView)> distFunctor(ghostedCoordsView);
+        auto ghostedCoordsView = ghostedCoords->getDeviceLocalView(Xpetra::Access::ReadWrite);
+        CoalesceDrop_Kokkos_Details::DistanceFunctor<LO, decltype(ghostedCoordsView)> distFunctor(ghostedCoordsView);
 
-          // Construct Laplacian diagonal
-          RCP<Vector> localLaplDiag;
-          {
-            SubFactoryMonitor m2(*this, "Local Laplacian diag construction", currentLevel);
+        // Construct Laplacian diagonal
+        RCP<Vector> localLaplDiag;
+        {
+          SubFactoryMonitor m2(*this, "Local Laplacian diag construction", currentLevel);
 
-            localLaplDiag = VectorFactory::Build(uniqueMap);
+          localLaplDiag = VectorFactory::Build(uniqueMap);
 
-            auto localLaplDiagView = localLaplDiag->getDeviceLocalView(Xpetra::Access::OverwriteAll);
-            auto kokkosGraph = kokkosMatrix.graph;
+          auto localLaplDiagView = localLaplDiag->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+          auto kokkosGraph       = kokkosMatrix.graph;
 
-            Kokkos::parallel_for("MueLu:CoalesceDropF:Build:scalar_filter:laplacian_diag", range_type(0,numRows),
+          Kokkos::parallel_for(
+              "MueLu:CoalesceDropF:Build:scalar_filter:laplacian_diag", range_type(0, numRows),
               KOKKOS_LAMBDA(const LO row) {
                 auto rowView = kokkosGraph.rowConst(row);
                 auto length  = rowView.length;
@@ -685,264 +674,267 @@ namespace MueLu {
                 for (decltype(length) colID = 0; colID < length; colID++) {
                   auto col = rowView(colID);
                   if (row != col)
-                    d += impl_ATS::one()/distFunctor.distance2(row, col);
+                    d += impl_ATS::one() / distFunctor.distance2(row, col);
                 }
-                localLaplDiagView(row,0) = d;
+                localLaplDiagView(row, 0) = d;
               });
-          }
-
-          // Construct ghosted Laplacian diagonal
-          RCP<Vector> ghostedLaplDiag;
-          {
-            SubFactoryMonitor m2(*this, "Ghosted Laplacian diag construction", currentLevel);
-            ghostedLaplDiag = VectorFactory::Build(nonUniqueMap);
-            ghostedLaplDiag->doImport(*localLaplDiag, *importer, Xpetra::INSERT);
-          }
-
-          // Filter out entries
-          {
-            SubFactoryMonitor m2(*this, "MainLoop", currentLevel);
-
-            auto ghostedLaplDiagView = ghostedLaplDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-            CoalesceDrop_Kokkos_Details::DistanceLaplacianDropFunctor<LO, decltype(ghostedLaplDiagView), decltype(distFunctor)>
-                dropFunctor(ghostedLaplDiagView, distFunctor, threshold);
-            CoalesceDrop_Kokkos_Details::ScalarFunctor<SC, LO, local_matrix_type, decltype(bndNodes), decltype(dropFunctor)>
-                scalarFunctor(kokkosMatrix, bndNodes, dropFunctor, rows, colsAux, valsAux, reuseGraph, lumping, threshold, true);
-
-            Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:main_loop", range_type(0,numRows),
-                                    scalarFunctor, nnzFA);
-          }
         }
 
+        // Construct ghosted Laplacian diagonal
+        RCP<Vector> ghostedLaplDiag;
+        {
+          SubFactoryMonitor m2(*this, "Ghosted Laplacian diag construction", currentLevel);
+          ghostedLaplDiag = VectorFactory::Build(nonUniqueMap);
+          ghostedLaplDiag->doImport(*localLaplDiag, *importer, Xpetra::INSERT);
+        }
+
+        // Filter out entries
+        {
+          SubFactoryMonitor m2(*this, "MainLoop", currentLevel);
+
+          auto ghostedLaplDiagView = ghostedLaplDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+          CoalesceDrop_Kokkos_Details::DistanceLaplacianDropFunctor<LO, decltype(ghostedLaplDiagView), decltype(distFunctor)>
+              dropFunctor(ghostedLaplDiagView, distFunctor, threshold);
+          CoalesceDrop_Kokkos_Details::ScalarFunctor<SC, LO, local_matrix_type, decltype(bndNodes), decltype(dropFunctor)>
+              scalarFunctor(kokkosMatrix, bndNodes, dropFunctor, rows, colsAux, valsAux, reuseGraph, lumping, threshold, true);
+
+          Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:main_loop", range_type(0, numRows),
+                                  scalarFunctor, nnzFA);
+        }
       }
-      numDropped = nnzA - nnzFA;
+    }
+    numDropped = nnzA - nnzFA;
 
-      boundaryNodes = bndNodes;
+    boundaryNodes = bndNodes;
 
-      {
-        SubFactoryMonitor m2(*this, "CompressRows", currentLevel);
+    {
+      SubFactoryMonitor m2(*this, "CompressRows", currentLevel);
 
-        // parallel_scan (exclusive)
-        Kokkos::parallel_scan("MueLu:CoalesceDropF:Build:scalar_filter:compress_rows", range_type(0,numRows+1),
+      // parallel_scan (exclusive)
+      Kokkos::parallel_scan(
+          "MueLu:CoalesceDropF:Build:scalar_filter:compress_rows", range_type(0, numRows + 1),
           KOKKOS_LAMBDA(const LO i, LO& update, const bool& final_pass) {
             update += rows(i);
             if (final_pass)
               rows(i) = update;
           });
-      }
+    }
 
-      // Compress cols (and optionally vals)
-      // We use a trick here: we moved all remaining elements to the beginning
-      // of the original row in the main loop, so we don't need to check for
-      // INVALID here, and just stop when achieving the new number of elements
-      // per row.
-      cols_type cols(Kokkos::ViewAllocateWithoutInitializing("FA_cols"), nnzFA);
-      vals_type vals;
-      if (reuseGraph) {
-        GetOStream(Runtime1) << "reuse matrix graph for filtering (compress matrix columns only)" << std::endl;
-        // Only compress cols
-        SubFactoryMonitor m2(*this, "CompressColsAndVals", currentLevel);
-
-        Kokkos::parallel_for("MueLu:TentativePF:Build:compress_cols", range_type(0,numRows),
+    // Compress cols (and optionally vals)
+    // We use a trick here: we moved all remaining elements to the beginning
+    // of the original row in the main loop, so we don't need to check for
+    // INVALID here, and just stop when achieving the new number of elements
+    // per row.
+    cols_type cols(Kokkos::ViewAllocateWithoutInitializing("FA_cols"), nnzFA);
+    vals_type vals;
+    if (reuseGraph) {
+      GetOStream(Runtime1) << "reuse matrix graph for filtering (compress matrix columns only)" << std::endl;
+      // Only compress cols
+      SubFactoryMonitor m2(*this, "CompressColsAndVals", currentLevel);
+
+      Kokkos::parallel_for(
+          "MueLu:TentativePF:Build:compress_cols", range_type(0, numRows),
           KOKKOS_LAMBDA(const LO i) {
             // Is there Kokkos memcpy?
-            LO rowStart  = rows(i);
-            LO rowAStart = rowsA(i);
-            size_t rownnz = rows(i+1) - rows(i);
+            LO rowStart   = rows(i);
+            LO rowAStart  = rowsA(i);
+            size_t rownnz = rows(i + 1) - rows(i);
             for (size_t j = 0; j < rownnz; j++)
-              cols(rowStart+j) = colsAux(rowAStart+j);
+              cols(rowStart + j) = colsAux(rowAStart + j);
           });
-      } else {
-        // Compress cols and vals
-        GetOStream(Runtime1) << "new matrix graph for filtering (compress matrix columns and values)" << std::endl;
-        SubFactoryMonitor m2(*this, "CompressColsAndVals", currentLevel);
+    } else {
+      // Compress cols and vals
+      GetOStream(Runtime1) << "new matrix graph for filtering (compress matrix columns and values)" << std::endl;
+      SubFactoryMonitor m2(*this, "CompressColsAndVals", currentLevel);
 
-        vals = vals_type(Kokkos::ViewAllocateWithoutInitializing("FA_vals"), nnzFA);
+      vals = vals_type(Kokkos::ViewAllocateWithoutInitializing("FA_vals"), nnzFA);
 
-        Kokkos::parallel_for("MueLu:TentativePF:Build:compress_cols", range_type(0,numRows),
+      Kokkos::parallel_for(
+          "MueLu:TentativePF:Build:compress_cols", range_type(0, numRows),
           KOKKOS_LAMBDA(const LO i) {
-            LO rowStart  = rows(i);
-            LO rowAStart = rowsA(i);
-            size_t rownnz = rows(i+1) - rows(i);
+            LO rowStart   = rows(i);
+            LO rowAStart  = rowsA(i);
+            size_t rownnz = rows(i + 1) - rows(i);
             for (size_t j = 0; j < rownnz; j++) {
-              cols(rowStart+j) = colsAux(rowAStart+j);
-              vals(rowStart+j) = valsAux(rowAStart+j);
+              cols(rowStart + j) = colsAux(rowAStart + j);
+              vals(rowStart + j) = valsAux(rowAStart + j);
             }
           });
-      }
+    }
 
-      kokkos_graph_type kokkosGraph(cols, rows);
+    kokkos_graph_type kokkosGraph(cols, rows);
 
-      {
-        SubFactoryMonitor m2(*this, "LWGraph construction", currentLevel);
+    {
+      SubFactoryMonitor m2(*this, "LWGraph construction", currentLevel);
 
-        graph = rcp(new LWGraph_kokkos(kokkosGraph, A->getRowMap(), A->getColMap(), "filtered graph of A"));
-        graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
-      }
+      graph = rcp(new LWGraph_kokkos(kokkosGraph, A->getRowMap(), A->getColMap(), "filtered graph of A"));
+      graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
+    }
 
-      numTotal = A->getLocalNumEntries();
+    numTotal = A->getLocalNumEntries();
 
-      dofsPerNode = 1;
+    dofsPerNode = 1;
 
-      if (!reuseGraph) {
-        SubFactoryMonitor m2(*this, "LocalMatrix+FillComplete", currentLevel);
+    if (!reuseGraph) {
+      SubFactoryMonitor m2(*this, "LocalMatrix+FillComplete", currentLevel);
 
-        local_matrix_type localFA = local_matrix_type("A", numRows, A->getLocalMatrixDevice().numCols(), nnzFA, vals, rows, cols);
-        auto filteredACrs = CrsMatrixFactory::Build(localFA, A->getRowMap(), A->getColMap(), A->getDomainMap(), A->getRangeMap(),
-                                                    A->getCrsGraph()->getImporter(), A->getCrsGraph()->getExporter());
-        filteredA = rcp(new CrsMatrixWrap(filteredACrs));
-      }
+      local_matrix_type localFA = local_matrix_type("A", numRows, A->getLocalMatrixDevice().numCols(), nnzFA, vals, rows, cols);
+      auto filteredACrs         = CrsMatrixFactory::Build(localFA, A->getRowMap(), A->getColMap(), A->getDomainMap(), A->getRangeMap(),
+                                                          A->getCrsGraph()->getImporter(), A->getCrsGraph()->getExporter());
+      filteredA                 = rcp(new CrsMatrixWrap(filteredACrs));
+    }
 
-      filteredA->SetFixedBlockSize(A->GetFixedBlockSize());
+    filteredA->SetFixedBlockSize(A->GetFixedBlockSize());
 
-      if (pL.get<bool>("filtered matrix: reuse eigenvalue")) {
-        // Reuse max eigenvalue from A
-        // It is unclear what eigenvalue is the best for the smoothing, but we already may have
-        // the D^{-1}A estimate in A, may as well use it.
-        // NOTE: ML does that too
-        filteredA->SetMaxEigenvalueEstimate(A->GetMaxEigenvalueEstimate());
-      } else {
-        filteredA->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
-      }
+    if (pL.get<bool>("filtered matrix: reuse eigenvalue")) {
+      // Reuse max eigenvalue from A
+      // It is unclear what eigenvalue is the best for the smoothing, but we already may have
+      // the D^{-1}A estimate in A, may as well use it.
+      // NOTE: ML does that too
+      filteredA->SetMaxEigenvalueEstimate(A->GetMaxEigenvalueEstimate());
+    } else {
+      filteredA->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
+    }
 
-    } else if (blkSize > 1 && threshold == zero) {
-      // Case 3:  block problem without filtering
-      //
-      // FIXME_KOKKOS: this code is completely unoptimized. It really should do
-      // a very simple thing: merge rows and produce nodal graph. But the code
-      // seems very complicated. Can we do better?
-
-      TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getLocalNumElements() % blkSize != 0, MueLu::Exceptions::RuntimeError, "MueLu::CoalesceDropFactory: Number of local elements is " << A->getRowMap()->getLocalNumElements() << " but should be a multiply of " << blkSize);
-
-      const RCP<const Map> rowMap = A->getRowMap();
-      const RCP<const Map> colMap = A->getColMap();
-
-      // build a node row map (uniqueMap = non-overlapping) and a node column map
-      // (nonUniqueMap = overlapping). The arrays rowTranslation and colTranslation
-      // stored in the AmalgamationInfo class container contain the local node id
-      // given a local dof id. The data is calculated in the AmalgamationFactory and
-      // stored in the variable "UnAmalgamationInfo" (which is of type AmalagamationInfo)
-      const RCP<const Map> uniqueMap = amalInfo->getNodeRowMap();
-      const RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
-      Array<LO> rowTranslationArray = *(amalInfo->getRowTranslation()); // TAW should be transform that into a View?
-      Array<LO> colTranslationArray = *(amalInfo->getColTranslation());
-
-      Kokkos::View<LO*, Kokkos::MemoryUnmanaged>
-        rowTranslationView(rowTranslationArray.getRawPtr(),rowTranslationArray.size() );
-      Kokkos::View<LO*, Kokkos::MemoryUnmanaged>
-        colTranslationView(colTranslationArray.getRawPtr(),colTranslationArray.size() );
-
-      // get number of local nodes
-      LO numNodes = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
-      typedef typename Kokkos::View<LocalOrdinal*, DeviceType> id_translation_type;
-      id_translation_type rowTranslation("dofId2nodeId",rowTranslationArray.size());
-      id_translation_type colTranslation("ov_dofId2nodeId",colTranslationArray.size());
-      Kokkos::deep_copy(rowTranslation, rowTranslationView);
-      Kokkos::deep_copy(colTranslation, colTranslationView);
-
-      // extract striding information
-      blkSize = A->GetFixedBlockSize();  //< the full block size (number of dofs per node in strided map)
-      LocalOrdinal blkId   = -1;         //< the block id within a strided map or -1 if it is a full block map
-      LocalOrdinal blkPartSize = A->GetFixedBlockSize(); //< stores block size of part blkId (or the full block size)
-      if(A->IsView("stridedMaps") == true) {
-        const RCP<const Map> myMap = A->getRowMap("stridedMaps");
-        const RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
-        TEUCHOS_TEST_FOR_EXCEPTION(strMap.is_null() == true, Exceptions::RuntimeError, "Map is not of type stridedMap");
-        blkSize = Teuchos::as<const LocalOrdinal>(strMap->getFixedBlockSize());
-        blkId   = strMap->getStridedBlockId();
-        if (blkId > -1)
-          blkPartSize = Teuchos::as<LocalOrdinal>(strMap->getStridingData()[blkId]);
-      }
-      auto kokkosMatrix = A->getLocalMatrixDevice(); // access underlying kokkos data
-
-      //
-      typedef typename LWGraph_kokkos::local_graph_type   kokkos_graph_type;
-      typedef typename kokkos_graph_type::row_map_type    row_map_type;
-      //typedef typename row_map_type::HostMirror           row_map_type_h;
-      typedef typename kokkos_graph_type::entries_type    entries_type;
-
-      // Stage 1c: get number of dof-nonzeros per blkSize node rows
-      typename row_map_type::non_const_type dofNnz("nnz_map", numNodes + 1);
-      LO numDofCols = 0;
-      CoalesceDrop_Kokkos_Details::Stage1aVectorFunctor<decltype(kokkosMatrix), decltype(dofNnz), decltype(blkPartSize)> stage1aFunctor(kokkosMatrix, dofNnz, blkPartSize);
-      Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:stage1a", range_type(0,numNodes), stage1aFunctor, numDofCols);
-      // parallel_scan (exclusive)
-      CoalesceDrop_Kokkos_Details::ScanFunctor<LO,decltype(dofNnz)> scanFunctor(dofNnz);
-      Kokkos::parallel_scan("MueLu:CoalesceDropF:Build:scalar_filter:stage1_scan", range_type(0,numNodes+1), scanFunctor);
+  } else if (blkSize > 1 && threshold == zero) {
+    // Case 3:  block problem without filtering
+    //
+    // FIXME_KOKKOS: this code is completely unoptimized. It really should do
+    // a very simple thing: merge rows and produce nodal graph. But the code
+    // seems very complicated. Can we do better?
+
+    TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getLocalNumElements() % blkSize != 0, MueLu::Exceptions::RuntimeError, "MueLu::CoalesceDropFactory: Number of local elements is " << A->getRowMap()->getLocalNumElements() << " but should be a multiply of " << blkSize);
+
+    const RCP<const Map> rowMap = A->getRowMap();
+    const RCP<const Map> colMap = A->getColMap();
+
+    // build a node row map (uniqueMap = non-overlapping) and a node column map
+    // (nonUniqueMap = overlapping). The arrays rowTranslation and colTranslation
+    // stored in the AmalgamationInfo class container contain the local node id
+    // given a local dof id. The data is calculated in the AmalgamationFactory and
+    // stored in the variable "UnAmalgamationInfo" (which is of type AmalagamationInfo)
+    const RCP<const Map> uniqueMap    = amalInfo->getNodeRowMap();
+    const RCP<const Map> nonUniqueMap = amalInfo->getNodeColMap();
+    Array<LO> rowTranslationArray     = *(amalInfo->getRowTranslation());  // TAW should be transform that into a View?
+    Array<LO> colTranslationArray     = *(amalInfo->getColTranslation());
+
+    Kokkos::View<LO*, Kokkos::MemoryUnmanaged>
+        rowTranslationView(rowTranslationArray.getRawPtr(), rowTranslationArray.size());
+    Kokkos::View<LO*, Kokkos::MemoryUnmanaged>
+        colTranslationView(colTranslationArray.getRawPtr(), colTranslationArray.size());
+
+    // get number of local nodes
+    LO numNodes = Teuchos::as<LocalOrdinal>(uniqueMap->getLocalNumElements());
+    typedef typename Kokkos::View<LocalOrdinal*, DeviceType> id_translation_type;
+    id_translation_type rowTranslation("dofId2nodeId", rowTranslationArray.size());
+    id_translation_type colTranslation("ov_dofId2nodeId", colTranslationArray.size());
+    Kokkos::deep_copy(rowTranslation, rowTranslationView);
+    Kokkos::deep_copy(colTranslation, colTranslationView);
+
+    // extract striding information
+    blkSize                  = A->GetFixedBlockSize();  //< the full block size (number of dofs per node in strided map)
+    LocalOrdinal blkId       = -1;                      //< the block id within a strided map or -1 if it is a full block map
+    LocalOrdinal blkPartSize = A->GetFixedBlockSize();  //< stores block size of part blkId (or the full block size)
+    if (A->IsView("stridedMaps") == true) {
+      const RCP<const Map> myMap         = A->getRowMap("stridedMaps");
+      const RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(myMap);
+      TEUCHOS_TEST_FOR_EXCEPTION(strMap.is_null() == true, Exceptions::RuntimeError, "Map is not of type stridedMap");
+      blkSize = Teuchos::as<const LocalOrdinal>(strMap->getFixedBlockSize());
+      blkId   = strMap->getStridedBlockId();
+      if (blkId > -1)
+        blkPartSize = Teuchos::as<LocalOrdinal>(strMap->getStridingData()[blkId]);
+    }
+    auto kokkosMatrix = A->getLocalMatrixDevice();  // access underlying kokkos data
 
-      // Detect and record dof rows that correspond to Dirichlet boundary conditions
-      boundary_nodes_type singleEntryRows = Utilities::DetectDirichletRows_kokkos(*A, dirichletThreshold);
+    //
+    typedef typename LWGraph_kokkos::local_graph_type kokkos_graph_type;
+    typedef typename kokkos_graph_type::row_map_type row_map_type;
+    // typedef typename row_map_type::HostMirror           row_map_type_h;
+    typedef typename kokkos_graph_type::entries_type entries_type;
 
-      typename entries_type::non_const_type dofcols("dofcols", numDofCols/*dofNnz(numNodes)*/); // why does dofNnz(numNodes) work? should be a parallel reduce, i guess
+    // Stage 1c: get number of dof-nonzeros per blkSize node rows
+    typename row_map_type::non_const_type dofNnz("nnz_map", numNodes + 1);
+    LO numDofCols = 0;
+    CoalesceDrop_Kokkos_Details::Stage1aVectorFunctor<decltype(kokkosMatrix), decltype(dofNnz), decltype(blkPartSize)> stage1aFunctor(kokkosMatrix, dofNnz, blkPartSize);
+    Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:stage1a", range_type(0, numNodes), stage1aFunctor, numDofCols);
+    // parallel_scan (exclusive)
+    CoalesceDrop_Kokkos_Details::ScanFunctor<LO, decltype(dofNnz)> scanFunctor(dofNnz);
+    Kokkos::parallel_scan("MueLu:CoalesceDropF:Build:scalar_filter:stage1_scan", range_type(0, numNodes + 1), scanFunctor);
 
-      // we have dofcols and dofids from Stage1dVectorFunctor
-      LO numNodeCols = 0;
-      typename row_map_type::non_const_type rows("nnz_nodemap", numNodes + 1);
-      typename boundary_nodes_type::non_const_type bndNodes("boundaryNodes", numNodes);
+    // Detect and record dof rows that correspond to Dirichlet boundary conditions
+    boundary_nodes_type singleEntryRows = Utilities::DetectDirichletRows_kokkos(*A, dirichletThreshold);
 
-      CoalesceDrop_Kokkos_Details::Stage1bcVectorFunctor <decltype(kokkosMatrix), decltype(dofNnz), decltype(blkPartSize), decltype(dofcols), decltype(colTranslation), decltype(singleEntryRows), decltype(bndNodes), bool> stage1bcFunctor(kokkosMatrix, dofNnz, blkPartSize, dofcols, colTranslation, rows, singleEntryRows, bndNodes, pL.get<bool>("aggregation: greedy Dirichlet"));
-      Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:stage1c", range_type(0,numNodes), stage1bcFunctor,numNodeCols);
+    typename entries_type::non_const_type dofcols("dofcols", numDofCols /*dofNnz(numNodes)*/);  // why does dofNnz(numNodes) work? should be a parallel reduce, i guess
 
-      // parallel_scan (exclusive)
-      CoalesceDrop_Kokkos_Details::ScanFunctor<LO,decltype(rows)> scanNodeFunctor(rows);
-      Kokkos::parallel_scan("MueLu:CoalesceDropF:Build:scalar_filter:stage1_scan", range_type(0,numNodes+1), scanNodeFunctor);
+    // we have dofcols and dofids from Stage1dVectorFunctor
+    LO numNodeCols = 0;
+    typename row_map_type::non_const_type rows("nnz_nodemap", numNodes + 1);
+    typename boundary_nodes_type::non_const_type bndNodes("boundaryNodes", numNodes);
 
-      // create column node view
-      typename entries_type::non_const_type cols("nodecols", numNodeCols);
+    CoalesceDrop_Kokkos_Details::Stage1bcVectorFunctor<decltype(kokkosMatrix), decltype(dofNnz), decltype(blkPartSize), decltype(dofcols), decltype(colTranslation), decltype(singleEntryRows), decltype(bndNodes), bool> stage1bcFunctor(kokkosMatrix, dofNnz, blkPartSize, dofcols, colTranslation, rows, singleEntryRows, bndNodes, pL.get<bool>("aggregation: greedy Dirichlet"));
+    Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:scalar_filter:stage1c", range_type(0, numNodes), stage1bcFunctor, numNodeCols);
 
+    // parallel_scan (exclusive)
+    CoalesceDrop_Kokkos_Details::ScanFunctor<LO, decltype(rows)> scanNodeFunctor(rows);
+    Kokkos::parallel_scan("MueLu:CoalesceDropF:Build:scalar_filter:stage1_scan", range_type(0, numNodes + 1), scanNodeFunctor);
 
-      CoalesceDrop_Kokkos_Details::Stage1dVectorFunctor <decltype(kokkosMatrix), decltype(dofNnz), decltype(dofcols), decltype(rows), decltype(cols)> stage1dFunctor(dofcols, dofNnz, cols, rows);
-      Kokkos::parallel_for("MueLu:CoalesceDropF:Build:scalar_filter:stage1c", range_type(0,numNodes), stage1dFunctor);
-      kokkos_graph_type kokkosGraph(cols, rows);
+    // create column node view
+    typename entries_type::non_const_type cols("nodecols", numNodeCols);
 
-      // create LW graph
-      graph = rcp(new LWGraph_kokkos(kokkosGraph, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
+    CoalesceDrop_Kokkos_Details::Stage1dVectorFunctor<decltype(kokkosMatrix), decltype(dofNnz), decltype(dofcols), decltype(rows), decltype(cols)> stage1dFunctor(dofcols, dofNnz, cols, rows);
+    Kokkos::parallel_for("MueLu:CoalesceDropF:Build:scalar_filter:stage1c", range_type(0, numNodes), stage1dFunctor);
+    kokkos_graph_type kokkosGraph(cols, rows);
 
-      boundaryNodes = bndNodes;
-      graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
-      numTotal = A->getLocalNumEntries();
+    // create LW graph
+    graph = rcp(new LWGraph_kokkos(kokkosGraph, uniqueMap, nonUniqueMap, "amalgamated graph of A"));
 
-      dofsPerNode = blkSize;
+    boundaryNodes = bndNodes;
+    graph->getLocalLWGraph().SetBoundaryNodeMap(boundaryNodes);
+    numTotal = A->getLocalNumEntries();
 
-      filteredA = A;
+    dofsPerNode = blkSize;
 
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu: CoalesceDropFactory_kokkos: Block filtering is not implemented");
-    }
+    filteredA = A;
+
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu: CoalesceDropFactory_kokkos: Block filtering is not implemented");
+  }
 
-    if (GetVerbLevel() & Statistics1) {
-      GO numLocalBoundaryNodes  = 0;
-      GO numGlobalBoundaryNodes = 0;
+  if (GetVerbLevel() & Statistics1) {
+    GO numLocalBoundaryNodes  = 0;
+    GO numGlobalBoundaryNodes = 0;
 
-      Kokkos::parallel_reduce("MueLu:CoalesceDropF:Build:bnd", range_type(0, boundaryNodes.extent(0)),
+    Kokkos::parallel_reduce(
+        "MueLu:CoalesceDropF:Build:bnd", range_type(0, boundaryNodes.extent(0)),
         KOKKOS_LAMBDA(const LO i, GO& n) {
           if (boundaryNodes(i))
             n++;
-        }, numLocalBoundaryNodes);
+        },
+        numLocalBoundaryNodes);
 
-      auto comm = A->getRowMap()->getComm();
-      MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
-      GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
-    }
+    auto comm = A->getRowMap()->getComm();
+    MueLu_sumAll(comm, numLocalBoundaryNodes, numGlobalBoundaryNodes);
+    GetOStream(Statistics1) << "Detected " << numGlobalBoundaryNodes << " Dirichlet nodes" << std::endl;
+  }
 
-    if ((GetVerbLevel() & Statistics1) && threshold != zero) {
-      auto comm = A->getRowMap()->getComm();
+  if ((GetVerbLevel() & Statistics1) && threshold != zero) {
+    auto comm = A->getRowMap()->getComm();
 
-      GO numGlobalTotal, numGlobalDropped;
-      MueLu_sumAll(comm, numTotal,   numGlobalTotal);
-      MueLu_sumAll(comm, numDropped, numGlobalDropped);
+    GO numGlobalTotal, numGlobalDropped;
+    MueLu_sumAll(comm, numTotal, numGlobalTotal);
+    MueLu_sumAll(comm, numDropped, numGlobalDropped);
 
-      if (numGlobalTotal != 0) {
-        GetOStream(Statistics1) << "Number of dropped entries: "
-            << numGlobalDropped << "/" << numGlobalTotal
-            << " (" << 100*Teuchos::as<double>(numGlobalDropped)/Teuchos::as<double>(numGlobalTotal) << "%)" << std::endl;
-      }
+    if (numGlobalTotal != 0) {
+      GetOStream(Statistics1) << "Number of dropped entries: "
+                              << numGlobalDropped << "/" << numGlobalTotal
+                              << " (" << 100 * Teuchos::as<double>(numGlobalDropped) / Teuchos::as<double>(numGlobalTotal) << "%)" << std::endl;
     }
-
-    Set(currentLevel, "DofsPerNode",  dofsPerNode);
-    Set(currentLevel, "Graph",        graph);
-    Set(currentLevel, "A",            filteredA);
   }
+
+  Set(currentLevel, "DofsPerNode", dofsPerNode);
+  Set(currentLevel, "Graph", graph);
+  Set(currentLevel, "A", filteredA);
 }
-#endif // MUELU_COALESCEDROPFACTORY_KOKKOS_DEF_HPP
+}  // namespace MueLu
+#endif  // MUELU_COALESCEDROPFACTORY_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionBaseClass_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionBaseClass_decl.hpp
index 7f94599c3ba1..a7fc2e3aad98 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionBaseClass_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionBaseClass_decl.hpp
@@ -46,35 +46,32 @@
 #ifndef MUELU_PREDROPFUNCTIONBASECLASS_DECL_HPP
 #define MUELU_PREDROPFUNCTIONBASECLASS_DECL_HPP
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_BaseClass.hpp"
 #include "MueLu_PreDropFunctionBaseClass_fwd.hpp"
 
 namespace MueLu {
 
-  /*!
-   * Base class you can derive from to allow user defined dropping
-   *
-   */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class PreDropFunctionBaseClass : public BaseClass {
+/*!
+ * Base class you can derive from to allow user defined dropping
+ *
+ */
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PreDropFunctionBaseClass : public BaseClass {
 #undef MUELU_PREDROPFUNCTIONBASECLASS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! Destructor
-    virtual ~PreDropFunctionBaseClass() { }
-
-    //! Drop
-    virtual bool Drop(size_t lrow, GlobalOrdinal grow, size_t k, LocalOrdinal lcid, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal> & indices, const Teuchos::ArrayView<const Scalar> & vals) = 0;
+ public:
+  //! Destructor
+  virtual ~PreDropFunctionBaseClass() {}
 
-  };
-}
+  //! Drop
+  virtual bool Drop(size_t lrow, GlobalOrdinal grow, size_t k, LocalOrdinal lcid, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal>& indices, const Teuchos::ArrayView<const Scalar>& vals) = 0;
+};
+}  // namespace MueLu
 
 #define MUELU_PREDROPFUNCTIONBASECLASS_SHORT
-#endif // MUELU_PREDROPFUNCTIONBASECLASS_DECL_HPP
+#endif  // MUELU_PREDROPFUNCTIONBASECLASS_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_decl.hpp
index a42b58f56b18..599ca918572b 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_decl.hpp
@@ -50,64 +50,59 @@
 #include "MueLu_PreDropFunctionBaseClass.hpp"
 #include "MueLu_PreDropFunctionConstVal_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-   * Example implementation for dropping values smaller then a constant threshold
-   *
-   */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class PreDropFunctionConstVal :
-    public MueLu::PreDropFunctionBaseClass<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+/*!
+ * Example implementation for dropping values smaller then a constant threshold
+ *
+ */
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PreDropFunctionConstVal : public MueLu::PreDropFunctionBaseClass<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! Constructor
-    explicit PreDropFunctionConstVal(const Scalar threshold = 0.0);
-
-    //! Destructor
-    virtual ~PreDropFunctionConstVal() { }
-
-    /*! Drop
-     * @param lrow (size_t): local row index (=lrowid)
-     * @param grow (GlobalOrdinal: global row id
-     * @param k    (size_t): local column iterator
-     * @param lcid (LocalOrdinal): local column id (=indices[k])
-     * @param gcid (GlobalOrdinal): global column id
-     * @param indices (ArrrayView): array of local column ids in current row (lrow)
-     * @param vals (ArrayView): array of corresponding values in current row (lrow)
-     * @return bool: false, if value in (lrow, lcid) shall be kept, true if it should be dropped
-     */
-    bool Drop(size_t lrow, GlobalOrdinal grow, size_t k, LocalOrdinal lcid, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal> & indices, const Teuchos::ArrayView<const Scalar> & vals);
-
-    //! Return threshold value.
-    Scalar GetThreshold() const;
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
+ public:
+  //! Constructor
+  explicit PreDropFunctionConstVal(const Scalar threshold = 0.0);
+
+  //! Destructor
+  virtual ~PreDropFunctionConstVal() {}
+
+  /*! Drop
+   * @param lrow (size_t): local row index (=lrowid)
+   * @param grow (GlobalOrdinal: global row id
+   * @param k    (size_t): local column iterator
+   * @param lcid (LocalOrdinal): local column id (=indices[k])
+   * @param gcid (GlobalOrdinal): global column id
+   * @param indices (ArrrayView): array of local column ids in current row (lrow)
+   * @param vals (ArrayView): array of corresponding values in current row (lrow)
+   * @return bool: false, if value in (lrow, lcid) shall be kept, true if it should be dropped
+   */
+  bool Drop(size_t lrow, GlobalOrdinal grow, size_t k, LocalOrdinal lcid, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal>& indices, const Teuchos::ArrayView<const Scalar>& vals);
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return threshold value.
+  Scalar GetThreshold() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //@}
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-  private:
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-    Scalar threshold_;
+  //@}
 
-  };
+ private:
+  Scalar threshold_;
+};
 
-}
+}  // namespace MueLu
 
 #define MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
-#endif // MUELU_PREDROPFUNCTIONCONSTVAL_DECL_HPP
+#endif  // MUELU_PREDROPFUNCTIONCONSTVAL_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_def.hpp
index 4c1577d90146..75b35c447c45 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_PreDropFunctionConstVal_def.hpp
@@ -53,40 +53,39 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PreDropFunctionConstVal(const Scalar threshold)
-    : threshold_(threshold) { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PreDropFunctionConstVal(const Scalar threshold)
+  : threshold_(threshold) {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Drop(size_t /* lrow */, GlobalOrdinal grow, size_t k, LocalOrdinal /* lcid */, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal> & /* indices */, const Teuchos::ArrayView<const Scalar> & vals) {
-    if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[k]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold_) || grow == gcid ) {
-      return false; // keep values
-    }
-    return true;    // values too small -> drop them
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Drop(size_t /* lrow */, GlobalOrdinal grow, size_t k, LocalOrdinal /* lcid */, GlobalOrdinal gcid, const Teuchos::ArrayView<const LocalOrdinal>& /* indices */, const Teuchos::ArrayView<const Scalar>& vals) {
+  if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[k]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold_) || grow == gcid) {
+    return false;  // keep values
   }
+  return true;  // values too small -> drop them
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Scalar PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThreshold() const {
-    return threshold_;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Scalar PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThreshold() const {
+  return threshold_;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << "PreDropFunctionConstVal: threshold = " << threshold_ << std::endl;
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << "PreDropFunctionConstVal: threshold = " << threshold_ << std::endl;
-    return out.str();
+/*template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+void PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+  if (verbLevel & Parameters0) {
+    out0 << "PreDropFunctionConstVal: threshold = " << threshold_ << std::endl;
   }
+}*/
 
-  /*template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-    if (verbLevel & Parameters0) {
-      out0 << "PreDropFunctionConstVal: threshold = " << threshold_ << std::endl;
-    }
-  }*/
-
-}
+}  // namespace MueLu
 
 #define MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
-#endif // MUELU_PREDROPFUNCTIONCONSTVAL_DEF_HPP
+#endif  // MUELU_PREDROPFUNCTIONCONSTVAL_DEF_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_decl.hpp
index b763c93cd955..062dd78fa71b 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_decl.hpp
@@ -65,103 +65,100 @@
 
 namespace MueLu {
 
-  /*!
-    @class SmooVecCoalesceDropFactory
-    @brief Factory for creating a graph base on a given matrix.
-
-    Factory for creating graphs from matrices with entries selectively dropped.
-
-    ## Code paths ##
-
-    Experimental dropping function based on taking a set of random vectors u, running
-    a smoother on A u = 0, and then basing the drop decisions on "how smooth" the vectors
-    are local. Neighobring regions where the vectors are smooth can be aggregated
-    together and so these are kept in the associated drop matrix. Areas that are
-    not smooth should end up in different aggregates and so the A_ij representing
-    these should be dropped.  This Factory can address both PDE systems and
-    scalar PDEs, always creating a matrix reprsenting nodal connections as opposed
-    to dof connections.
-
-     To enter this factor as opposed to the more standard CoalesceDropFactory() one
-     must set "aggregation: drop scheme" to "unsupported vector smoothing". In this
-     case some of the parameter options associated with CoalesceDropFactory (e.g.,
-     "aggregation: drop tol", "aggregation: Dirichlet threshold", "lightweight wrap")
-     will cause parameter validator errors.
-
-    ## Input/output of SmooVecCoalesceDropFactory ##
-
-    ### User parameters of SmooVecCoalesceDropFactory ###
-    Parameter                  | type      | default   | master.xml | validated | requested | description
-    ---------------------------|-----------|-----------|:----------:|:---------:|:---------:|------------
-     A                         |Factory    | null      |            | *         | *         | Generating factory of the operator A
-     "aggregation: drop scheme"|std::string|"classical"| *          | *         | *         | Must choose "unsupported vector smoothing"
-     "aggregation: number of times to pre or post smooth"|int| 10|* |           | *         | Amount of pre or post smoothing invocations
-     "aggregation: number of random vectors"|int| 10   |          * | *         | *         | Number of random vectors
-     "aggregation: penalty parameters"|Array(double)|{12.0,-.20}| * | *         | *         | Ultimately determines how much dropping is done
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SmooVecCoalesceDropFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see SmooVecCoalesceDropFactory::DeclareInput).
-
-    ### Variables provided by UncoupledAggregationFactory ###
-
-    After SmooVecCoalesceDropFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    Graph   | SmooVecCoalesceDropFactory   | Graph of matrix A
-    DofsPerNode | SmooVecCoalesceDropFactory | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
-
-  */
-
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class SmooVecCoalesceDropFactory : public SingleLevelFactoryBase {
+/*!
+  @class SmooVecCoalesceDropFactory
+  @brief Factory for creating a graph base on a given matrix.
+
+  Factory for creating graphs from matrices with entries selectively dropped.
+
+  ## Code paths ##
+
+  Experimental dropping function based on taking a set of random vectors u, running
+  a smoother on A u = 0, and then basing the drop decisions on "how smooth" the vectors
+  are local. Neighobring regions where the vectors are smooth can be aggregated
+  together and so these are kept in the associated drop matrix. Areas that are
+  not smooth should end up in different aggregates and so the A_ij representing
+  these should be dropped.  This Factory can address both PDE systems and
+  scalar PDEs, always creating a matrix reprsenting nodal connections as opposed
+  to dof connections.
+
+   To enter this factor as opposed to the more standard CoalesceDropFactory() one
+   must set "aggregation: drop scheme" to "unsupported vector smoothing". In this
+   case some of the parameter options associated with CoalesceDropFactory (e.g.,
+   "aggregation: drop tol", "aggregation: Dirichlet threshold", "lightweight wrap")
+   will cause parameter validator errors.
+
+  ## Input/output of SmooVecCoalesceDropFactory ##
+
+  ### User parameters of SmooVecCoalesceDropFactory ###
+  Parameter                  | type      | default   | master.xml | validated | requested | description
+  ---------------------------|-----------|-----------|:----------:|:---------:|:---------:|------------
+   A                         |Factory    | null      |            | *         | *         | Generating factory of the operator A
+   "aggregation: drop scheme"|std::string|"classical"| *          | *         | *         | Must choose "unsupported vector smoothing"
+   "aggregation: number of times to pre or post smooth"|int| 10|* |           | *         | Amount of pre or post smoothing invocations
+   "aggregation: number of random vectors"|int| 10   |          * | *         | *         | Number of random vectors
+   "aggregation: penalty parameters"|Array(double)|{12.0,-.20}| * | *         | *         | Ultimately determines how much dropping is done
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SmooVecCoalesceDropFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see SmooVecCoalesceDropFactory::DeclareInput).
+
+  ### Variables provided by UncoupledAggregationFactory ###
+
+  After SmooVecCoalesceDropFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  Graph   | SmooVecCoalesceDropFactory   | Graph of matrix A
+  DofsPerNode | SmooVecCoalesceDropFactory | number of DOFs per node. Note, that we assume a constant number of DOFs per node for all nodes associated with the operator A.
+
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SmooVecCoalesceDropFactory : public SingleLevelFactoryBase {
 #undef MUELU_SMOOVECCOALESCEDROPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  SmooVecCoalesceDropFactory();
 
-    //! Constructor
-    SmooVecCoalesceDropFactory();
+  //! Destructor
+  virtual ~SmooVecCoalesceDropFactory() {}
 
-    //! Destructor
-    virtual ~SmooVecCoalesceDropFactory() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  /// set predrop function
+  void SetPreDropFunction(const RCP<MueLu::PreDropFunctionBaseClass<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& predrop) { predrop_ = predrop; }
 
-    /// set predrop function
-    void SetPreDropFunction(const RCP<MueLu::PreDropFunctionBaseClass<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &predrop) { predrop_ = predrop; }
+  //@}
 
-    //@}
+  void Build(Level& currentLevel) const;  // Build
 
-    void Build(Level &currentLevel) const; // Build
+ private:
+  // pre-drop function
+  mutable RCP<PreDropFunctionBaseClass> predrop_;
 
-  private:
+  //! Methods to support compatible-relaxation style dropping
+  void badGuysCoalesceDrop(const Matrix& Amat, Teuchos::ArrayRCP<Scalar>& dropParams, LO nPDEs, const MultiVector& smoothedTVecs, const MultiVector& smoothedNull, RCP<GraphBase>& filteredGraph) const;
+  void badGuysDropfunc(LO row, const Teuchos::ArrayView<const LocalOrdinal>& indices, const Teuchos::ArrayView<const Scalar>& vals, const MultiVector& smoothedTVecs, LO nPDEs, Teuchos::ArrayRCP<Scalar>& penalties, const MultiVector& smoothedNull, Teuchos::ArrayRCP<LO>& Bcols, Teuchos::ArrayRCP<bool>& keepOrNot, LO& Nbcols, LO nLoc) const;
 
-    // pre-drop function
-    mutable
-     RCP<PreDropFunctionBaseClass> predrop_;
+};  // class SmooVecCoalesceDropFactory
 
-    //! Methods to support compatible-relaxation style dropping
-    void badGuysCoalesceDrop(const Matrix& Amat, Teuchos::ArrayRCP<Scalar> & dropParams, LO nPDEs, const MultiVector&  smoothedTVecs, const MultiVector& smoothedNull, RCP<GraphBase>& filteredGraph) const;
-    void badGuysDropfunc(LO row, const Teuchos::ArrayView<const LocalOrdinal>& indices, const Teuchos::ArrayView<const Scalar>& vals, const MultiVector&  smoothedTVecs, LO nPDEs, Teuchos::ArrayRCP<Scalar> & penalties, const MultiVector& smoothedNull, Teuchos::ArrayRCP<LO>& Bcols, Teuchos::ArrayRCP<bool>& keepOrNot, LO &Nbcols, LO nLoc) const;
-
-  }; //class SmooVecCoalesceDropFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SMOOVECCOALESCEDROPFACTORY_SHORT
-#endif // MUELU_SMOOVECCOALESCEDROPFACTORY_DECL_HPP
+#endif  // MUELU_SMOOVECCOALESCEDROPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_def.hpp
index 24a9a8334275..d6ad50c24f99 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_SmooVecCoalesceDropFactory_def.hpp
@@ -69,7 +69,6 @@
 #include "MueLu_Monitor.hpp"
 #include "MueLu_PreDropFunctionBaseClass.hpp"
 
-
 #include <Xpetra_IO.hpp>
 
 #include <algorithm>
@@ -84,160 +83,154 @@
 #include <stdlib.h>
 #include <math.h>
 
-
-#define poly0thOrderCoef   0
-#define poly1stOrderCoef   1
-#define poly2ndOrderCoef   2
-#define poly3rdOrderCoef   3
-#define poly4thOrderCoef   4
+#define poly0thOrderCoef 0
+#define poly1stOrderCoef 1
+#define poly2ndOrderCoef 2
+#define poly3rdOrderCoef 3
+#define poly4thOrderCoef 4
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: drop scheme");
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      validParamList->getEntry("aggregation: drop scheme").setValidator(
-        rcp(new validatorType(Teuchos::tuple<std::string>("unsupported vector smoothing"), "aggregation: drop scheme")));
-    }
-    SET_VALID_ENTRY("aggregation: number of random vectors");
-    SET_VALID_ENTRY("aggregation: number of times to pre or post smooth");
-    SET_VALID_ENTRY("aggregation: penalty parameters");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("PreSmoother",        Teuchos::null, "Generating factory of the PreSmoother");
-    validParamList->set< RCP<const FactoryBase> >("PostSmoother",       Teuchos::null, "Generating factory of the PostSmoother");
-
-    return validParamList;
+  SET_VALID_ENTRY("aggregation: drop scheme");
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    validParamList->getEntry("aggregation: drop scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("unsupported vector smoothing"), "aggregation: drop scheme")));
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmooVecCoalesceDropFactory() : predrop_(Teuchos::null) { }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A");
-    if (currentLevel.IsAvailable("PreSmoother")) {    // rst: totally unsure that this is legal
-      Input(currentLevel, "PreSmoother");             //      my guess is that this is not yet available
-    }                                                 //      so this always comes out false.
-    else if (currentLevel.IsAvailable("PostSmoother")) {  // perhaps we can look on the param list?
-      Input(currentLevel, "PostSmoother");
-    }
+  SET_VALID_ENTRY("aggregation: number of random vectors");
+  SET_VALID_ENTRY("aggregation: number of times to pre or post smooth");
+  SET_VALID_ENTRY("aggregation: penalty parameters");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("PreSmoother", Teuchos::null, "Generating factory of the PreSmoother");
+  validParamList->set<RCP<const FactoryBase> >("PostSmoother", Teuchos::null, "Generating factory of the PostSmoother");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmooVecCoalesceDropFactory()
+  : predrop_(Teuchos::null) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  if (currentLevel.IsAvailable("PreSmoother")) {        // rst: totally unsure that this is legal
+    Input(currentLevel, "PreSmoother");                 //      my guess is that this is not yet available
+  }                                                     //      so this always comes out false.
+  else if (currentLevel.IsAvailable("PostSmoother")) {  // perhaps we can look on the param list?
+    Input(currentLevel, "PostSmoother");
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-
-    FactoryMonitor m(*this, "Build", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
 
-    typedef Teuchos::ScalarTraits<SC> STS;
+  typedef Teuchos::ScalarTraits<SC> STS;
 
-    if (predrop_ != Teuchos::null)
-      GetOStream(Parameters0) << predrop_->description();
+  if (predrop_ != Teuchos::null)
+    GetOStream(Parameters0) << predrop_->description();
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
 
-    const ParameterList  & pL = GetParameterList();
+  const ParameterList& pL = GetParameterList();
 
-   LO nPDEs = A->GetFixedBlockSize();
+  LO nPDEs = A->GetFixedBlockSize();
 
-   RCP< MultiVector > testVecs;
-   RCP< MultiVector > nearNull;
+  RCP<MultiVector> testVecs;
+  RCP<MultiVector> nearNull;
 
 #ifdef takeOut
-   testVecs = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector("TpetraTVecs.mm", A->getRowMap());
+  testVecs = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector("TpetraTVecs.mm", A->getRowMap());
 #endif
-   size_t numRandom= as<size_t>(pL.get<int>("aggregation: number of random vectors"));
-   testVecs = MultiVectorFactory::Build(A->getRowMap(), numRandom, true);
-   // use random test vectors but should be positive in order to not get
-   // crummy results ... so take abs() of randomize().
-   testVecs->randomize();
-   for (size_t kk = 0; kk < testVecs->getNumVectors(); kk++ ) {
-     Teuchos::ArrayRCP< Scalar > curVec = testVecs->getDataNonConst(kk);
-     for (size_t ii = kk; ii < as<size_t>(A->getRowMap()->getLocalNumElements()); ii++ ) curVec[ii] = Teuchos::ScalarTraits<SC>::magnitude(curVec[ii]);
-   }
-   nearNull = MultiVectorFactory::Build(A->getRowMap(), nPDEs, true);
-
-   // initialize null space to constants
-   for (size_t kk = 0; kk < nearNull->getNumVectors(); kk++ ) {
-     Teuchos::ArrayRCP< Scalar > curVec = nearNull->getDataNonConst(kk);
-     for (size_t ii = kk; ii < as<size_t>(A->getRowMap()->getLocalNumElements()); ii += nearNull->getNumVectors() ) curVec[ii] = Teuchos::ScalarTraits<Scalar>::one();
-   }
-
-   RCP< MultiVector > zeroVec_TVecs;
-   RCP< MultiVector > zeroVec_Null;
-
-   zeroVec_TVecs = MultiVectorFactory::Build(A->getRowMap(), testVecs->getNumVectors(), true);
-   zeroVec_Null  = MultiVectorFactory::Build(A->getRowMap(), nPDEs, true);
-   zeroVec_TVecs->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
-   zeroVec_Null->putScalar( Teuchos::ScalarTraits<Scalar>::zero());
-
-   size_t nInvokeSmoother=as<size_t>(pL.get<int>("aggregation: number of times to pre or post smooth"));
-   if (currentLevel.IsAvailable("PreSmoother")) {
-     RCP<SmootherBase> preSmoo = currentLevel.Get< RCP<SmootherBase> >("PreSmoother");
-     for (size_t ii = 0; ii < nInvokeSmoother; ii++) preSmoo->Apply(*testVecs,*zeroVec_TVecs,false);
-     for (size_t ii = 0; ii < nInvokeSmoother; ii++) preSmoo->Apply(*nearNull,*zeroVec_Null,false);
-   }
-   else if (currentLevel.IsAvailable("PostSmoother")) {
-     RCP<SmootherBase> postSmoo = currentLevel.Get< RCP<SmootherBase> >("PostSmoother");
-     for (size_t ii = 0; ii < nInvokeSmoother; ii++) postSmoo->Apply(*testVecs,*zeroVec_TVecs,false);
-     for (size_t ii = 0; ii < nInvokeSmoother; ii++) postSmoo->Apply(*nearNull, *zeroVec_Null,false);
-   }
-   else
-     TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must set a smoother");
-
-   Teuchos::ArrayRCP<Scalar> penaltyPolyCoef(5);
-   Teuchos::ArrayView<const double> inputPolyCoef;
-
-   penaltyPolyCoef[poly0thOrderCoef] = 12.;
-   penaltyPolyCoef[poly1stOrderCoef] = -.2;
-   penaltyPolyCoef[poly2ndOrderCoef] = 0.0;
-   penaltyPolyCoef[poly3rdOrderCoef] = 0.0;
-   penaltyPolyCoef[poly4thOrderCoef] = 0.0;
-
-   if(pL.isParameter("aggregation: penalty parameters") && pL.get<Teuchos::Array<double> >("aggregation: penalty parameters").size() > 0) {
-      if (pL.get<Teuchos::Array<double> >("aggregation: penalty parameters").size() > penaltyPolyCoef.size())
-         TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of penalty parameters must be " << penaltyPolyCoef.size() << " or less");
-      inputPolyCoef = pL.get<Teuchos::Array<double> >("aggregation: penalty parameters")();
- 
-      for (size_t i =                                0; i < as<size_t>(inputPolyCoef.size())  ; i++) penaltyPolyCoef[i] = as<Scalar>(inputPolyCoef[i]);
-      for (size_t i = as<size_t>(inputPolyCoef.size()); i < as<size_t>(penaltyPolyCoef.size()); i++) penaltyPolyCoef[i] = Teuchos::ScalarTraits<Scalar>::zero();
-   }
-
-
-   RCP<GraphBase>   filteredGraph;
-   badGuysCoalesceDrop(*A, penaltyPolyCoef, nPDEs, *testVecs, *nearNull, filteredGraph);
+  size_t numRandom = as<size_t>(pL.get<int>("aggregation: number of random vectors"));
+  testVecs         = MultiVectorFactory::Build(A->getRowMap(), numRandom, true);
+  // use random test vectors but should be positive in order to not get
+  // crummy results ... so take abs() of randomize().
+  testVecs->randomize();
+  for (size_t kk = 0; kk < testVecs->getNumVectors(); kk++) {
+    Teuchos::ArrayRCP<Scalar> curVec = testVecs->getDataNonConst(kk);
+    for (size_t ii = kk; ii < as<size_t>(A->getRowMap()->getLocalNumElements()); ii++) curVec[ii] = Teuchos::ScalarTraits<SC>::magnitude(curVec[ii]);
+  }
+  nearNull = MultiVectorFactory::Build(A->getRowMap(), nPDEs, true);
+
+  // initialize null space to constants
+  for (size_t kk = 0; kk < nearNull->getNumVectors(); kk++) {
+    Teuchos::ArrayRCP<Scalar> curVec = nearNull->getDataNonConst(kk);
+    for (size_t ii = kk; ii < as<size_t>(A->getRowMap()->getLocalNumElements()); ii += nearNull->getNumVectors()) curVec[ii] = Teuchos::ScalarTraits<Scalar>::one();
+  }
+
+  RCP<MultiVector> zeroVec_TVecs;
+  RCP<MultiVector> zeroVec_Null;
+
+  zeroVec_TVecs = MultiVectorFactory::Build(A->getRowMap(), testVecs->getNumVectors(), true);
+  zeroVec_Null  = MultiVectorFactory::Build(A->getRowMap(), nPDEs, true);
+  zeroVec_TVecs->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
+  zeroVec_Null->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
+
+  size_t nInvokeSmoother = as<size_t>(pL.get<int>("aggregation: number of times to pre or post smooth"));
+  if (currentLevel.IsAvailable("PreSmoother")) {
+    RCP<SmootherBase> preSmoo = currentLevel.Get<RCP<SmootherBase> >("PreSmoother");
+    for (size_t ii = 0; ii < nInvokeSmoother; ii++) preSmoo->Apply(*testVecs, *zeroVec_TVecs, false);
+    for (size_t ii = 0; ii < nInvokeSmoother; ii++) preSmoo->Apply(*nearNull, *zeroVec_Null, false);
+  } else if (currentLevel.IsAvailable("PostSmoother")) {
+    RCP<SmootherBase> postSmoo = currentLevel.Get<RCP<SmootherBase> >("PostSmoother");
+    for (size_t ii = 0; ii < nInvokeSmoother; ii++) postSmoo->Apply(*testVecs, *zeroVec_TVecs, false);
+    for (size_t ii = 0; ii < nInvokeSmoother; ii++) postSmoo->Apply(*nearNull, *zeroVec_Null, false);
+  } else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must set a smoother");
+
+  Teuchos::ArrayRCP<Scalar> penaltyPolyCoef(5);
+  Teuchos::ArrayView<const double> inputPolyCoef;
+
+  penaltyPolyCoef[poly0thOrderCoef] = 12.;
+  penaltyPolyCoef[poly1stOrderCoef] = -.2;
+  penaltyPolyCoef[poly2ndOrderCoef] = 0.0;
+  penaltyPolyCoef[poly3rdOrderCoef] = 0.0;
+  penaltyPolyCoef[poly4thOrderCoef] = 0.0;
+
+  if (pL.isParameter("aggregation: penalty parameters") && pL.get<Teuchos::Array<double> >("aggregation: penalty parameters").size() > 0) {
+    if (pL.get<Teuchos::Array<double> >("aggregation: penalty parameters").size() > penaltyPolyCoef.size())
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of penalty parameters must be " << penaltyPolyCoef.size() << " or less");
+    inputPolyCoef = pL.get<Teuchos::Array<double> >("aggregation: penalty parameters")();
+
+    for (size_t i = 0; i < as<size_t>(inputPolyCoef.size()); i++) penaltyPolyCoef[i] = as<Scalar>(inputPolyCoef[i]);
+    for (size_t i = as<size_t>(inputPolyCoef.size()); i < as<size_t>(penaltyPolyCoef.size()); i++) penaltyPolyCoef[i] = Teuchos::ScalarTraits<Scalar>::zero();
+  }
 
+  RCP<GraphBase> filteredGraph;
+  badGuysCoalesceDrop(*A, penaltyPolyCoef, nPDEs, *testVecs, *nearNull, filteredGraph);
 
 #ifdef takeOut
-   /* write out graph for serial debugging purposes only. */
- 
-   FILE* fp = fopen("codeOutput","w");
-   fprintf(fp,"%d %d %d\n",(int) filteredGraph->GetNodeNumVertices(),(int) filteredGraph->GetNodeNumVertices(),
-   (int) filteredGraph->GetNodeNumEdges());
-   for (size_t i = 0; i < filteredGraph->GetNodeNumVertices(); i++) {
-       ArrayView<const LO> inds = filteredGraph->getNeighborVertices(as<LO>(i));
-       for (size_t j = 0; j < as<size_t>(inds.size()); j++) {
-         fprintf(fp,"%d %d 1.00e+00\n",(int) i+1,(int) inds[j]+1);
-       }
-   }
-   fclose(fp);
+  /* write out graph for serial debugging purposes only. */
+
+  FILE* fp = fopen("codeOutput", "w");
+  fprintf(fp, "%d %d %d\n", (int)filteredGraph->GetNodeNumVertices(), (int)filteredGraph->GetNodeNumVertices(),
+          (int)filteredGraph->GetNodeNumEdges());
+  for (size_t i = 0; i < filteredGraph->GetNodeNumVertices(); i++) {
+    ArrayView<const LO> inds = filteredGraph->getNeighborVertices(as<LO>(i));
+    for (size_t j = 0; j < as<size_t>(inds.size()); j++) {
+      fprintf(fp, "%d %d 1.00e+00\n", (int)i + 1, (int)inds[j] + 1);
+    }
+  }
+  fclose(fp);
 #endif
 
-    SC threshold = .01;
-    Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
-    Set(currentLevel, "Graph", filteredGraph);
-    Set(currentLevel, "DofsPerNode", 1);
+  SC threshold = .01;
+  Set<bool>(currentLevel, "Filtering", (threshold != STS::zero()));
+  Set(currentLevel, "Graph", filteredGraph);
+  Set(currentLevel, "DofsPerNode", 1);
 
-  } //Build
+}  // Build
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::badGuysCoalesceDrop(const Matrix& Amat, Teuchos::ArrayRCP<Scalar> & penaltyPolyCoef , LO nPDEs, const MultiVector& testVecs, const MultiVector& nearNull, RCP<GraphBase>& filteredGraph) const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::badGuysCoalesceDrop(const Matrix& Amat, Teuchos::ArrayRCP<Scalar>& penaltyPolyCoef, LO nPDEs, const MultiVector& testVecs, const MultiVector& nearNull, RCP<GraphBase>& filteredGraph) const {
   /*
    * Compute coalesce/drop graph (in filteredGraph) for A. The basic idea is to
    * balance trade-offs associated with
@@ -274,28 +267,28 @@ namespace MueLu {
    * Note: testVecs is supplied by the user, but normally is the result of
    * applying a relaxation scheme to Au = 0 where u is initial random.
    */
-  
-  GO     numMyNnz = Teuchos::as<GO>(Amat.getLocalNumEntries());
-  size_t nLoc     = Amat.getRowMap()->getLocalNumElements();
 
-  size_t nBlks    = nLoc/nPDEs;
-  if (nBlks*nPDEs != nLoc )
-     TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of local dofs not divisible by BlkSize");
+  GO numMyNnz = Teuchos::as<GO>(Amat.getLocalNumEntries());
+  size_t nLoc = Amat.getRowMap()->getLocalNumElements();
 
-  Teuchos::ArrayRCP<LO>   newRowPtr(nBlks+1); /* coalesce & drop matrix     */
-  Teuchos::ArrayRCP<LO>   newCols(numMyNnz);  /* arrays                     */
+  size_t nBlks = nLoc / nPDEs;
+  if (nBlks * nPDEs != nLoc)
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of local dofs not divisible by BlkSize");
 
-  Teuchos::ArrayRCP<LO>   bcols(nBlks);       /* returned by dropfun(j,...) */
-  Teuchos::ArrayRCP<bool> keepOrNot(nBlks);   /* gives cols for jth row and */
-                                              /* whether or not entry is    */
-                                              /* kept or dropped.           */
+  Teuchos::ArrayRCP<LO> newRowPtr(nBlks + 1); /* coalesce & drop matrix     */
+  Teuchos::ArrayRCP<LO> newCols(numMyNnz);    /* arrays                     */
 
-  LO   maxNzPerRow = 200;
+  Teuchos::ArrayRCP<LO> bcols(nBlks);       /* returned by dropfun(j,...) */
+  Teuchos::ArrayRCP<bool> keepOrNot(nBlks); /* gives cols for jth row and */
+                                            /* whether or not entry is    */
+                                            /* kept or dropped.           */
+
+  LO maxNzPerRow = 200;
   Teuchos::ArrayRCP<Scalar> penalties(maxNzPerRow); /* Penalty function */
                                                     /* described above. */
-                                    
-  Teuchos::ArrayRCP<bool> keepStatus(nBlks,true);  /* accumulated keepOrNot info */
-  Teuchos::ArrayRCP<LO>   bColList(nBlks);         /* accumulated bcols info     */
+
+  Teuchos::ArrayRCP<bool> keepStatus(nBlks, true); /* accumulated keepOrNot info */
+  Teuchos::ArrayRCP<LO> bColList(nBlks);           /* accumulated bcols info     */
                                                    /* for an entire block as     */
                                                    /* opposed to a single row    */
                                                    /* Additionally, keepOrNot[j] */
@@ -305,125 +298,123 @@ namespace MueLu {
                                                    /* whether the jth block is   */
                                                    /* kept within the block row. */
 
-  Teuchos::ArrayRCP<bool> alreadyOnBColList(nBlks,false);  /* used to avoid recording the*/
-                                                              /* same block column when     */
-                                                              /* processing different pt    */
-                                                              /* rows within a block.       */
-
-  Teuchos::ArrayRCP<bool>   boundaryNodes(nBlks,false);
+  Teuchos::ArrayRCP<bool> alreadyOnBColList(nBlks, false); /* used to avoid recording the*/
+                                                           /* same block column when     */
+                                                           /* processing different pt    */
+                                                           /* rows within a block.       */
 
+  Teuchos::ArrayRCP<bool> boundaryNodes(nBlks, false);
 
-  for (LO i = 0; i <  maxNzPerRow; i++)
+  for (LO i = 0; i < maxNzPerRow; i++)
     penalties[i] = penaltyPolyCoef[poly0thOrderCoef] +
-                   penaltyPolyCoef[poly1stOrderCoef]*(as<Scalar>(i)) +
-                   penaltyPolyCoef[poly2ndOrderCoef]*(as<Scalar>(i*i)) +
-                  (penaltyPolyCoef[poly3rdOrderCoef]*(as<Scalar>(i*i))*(as<Scalar>(i)))  +  //perhaps avoids overflow?
-                  (penaltyPolyCoef[poly4thOrderCoef]*(as<Scalar>(i*i))*(as<Scalar>(i*i)));
-
-   LO nzTotal = 0, numBCols = 0, row = -1, Nbcols, bcol;
-   newRowPtr[0] = 0;
-
-   /* proceed block by block */
-   for (LO i = 0; i < as<LO>(nBlks); i++) {
-     newRowPtr[i+1] = newRowPtr[i];
-     for (LO j = 0; j < nPDEs; j++) {
-       row = row + 1;
-
-       Teuchos::ArrayView<const LocalOrdinal> indices;
-       Teuchos::ArrayView<const Scalar> vals;
-
-       Amat.getLocalRowView(row, indices, vals);
- 
-       if (indices.size() > maxNzPerRow) {
-          LO oldSize = maxNzPerRow;
-          maxNzPerRow = indices.size() + 100;
-          penalties.resize(as<size_t>(maxNzPerRow),0.0);
-          for (LO k = oldSize; k < maxNzPerRow; k++)
-            penalties[k] = penaltyPolyCoef[poly0thOrderCoef] +
-                   penaltyPolyCoef[poly1stOrderCoef]*(as<Scalar>(i)) +
-                   penaltyPolyCoef[poly2ndOrderCoef]*(as<Scalar>(i*i)) +
-                  (penaltyPolyCoef[poly3rdOrderCoef]*(as<Scalar>(i*i))*(as<Scalar>(i)))  +
-                  (penaltyPolyCoef[poly4thOrderCoef]*(as<Scalar>(i*i))*(as<Scalar>(i*i)));
-       }
-       badGuysDropfunc(row, indices, vals, testVecs, nPDEs, penalties, nearNull, bcols,keepOrNot,Nbcols,nLoc);
-       for (LO k=0; k < Nbcols; k++) {
-         bcol = bcols[k];
-
-         /* add to bColList if not already on it */
-
-         if  (alreadyOnBColList[bcol] == false) {/* for PDE systems only record */
-           bColList[numBCols++]  = bcol;     /* neighboring block one time  */
-           alreadyOnBColList[bcol] = true;
-         }
-         /* drop if any pt row within block indicates entry should be dropped */
-
-         if (keepOrNot[k] == false) keepStatus[bcol] = false;
-
-       }  /* for (k=0; k < Nbcols; k++)   */
-     } /* for (j = 0; i < nPDEs; j++)  */
-
-     /* finished with block row. Now record block entries that we keep */
-     /* and reset keepStatus, bColList, and alreadyOnBColList.      */
-
-     if ( numBCols < 2) boundaryNodes[i] = true;
-     for (LO j=0; j < numBCols; j++) {
-       bcol = bColList[j];
-       if (keepStatus[bcol] == true) {
-         newCols[nzTotal] = bColList[j];
-         newRowPtr[i+1]++;
-         nzTotal = nzTotal + 1;
-       }
-       keepStatus[bcol] = true;
-       alreadyOnBColList[bcol] = false;
-       bColList[j] = 0;
-     }
-     numBCols = 0;
-   } /* for (i = 0; i < nBlks; i++)  */
-
-   /* create array of the correct size and copy over newCols to it */
-
-   Teuchos::ArrayRCP<LO> finalCols(nzTotal);
-   for (LO i = 0; i < nzTotal; i++) finalCols[i] = newCols[i];
-
-   // Not using column map because we do not allow for any off-proc stuff.
-   // Not sure if this is okay.  FIXME
-
-   RCP<const Map> rowMap = Amat.getRowMap(); // , colMap = Amat.getColMap();
-
-   LO nAmalgNodesOnProc = rowMap->getLocalNumElements()/nPDEs;
-   Teuchos::Array<GO> nodalGIDs(nAmalgNodesOnProc);
-   typename Teuchos::ScalarTraits<Scalar>::coordinateType temp;
-   for (size_t i = 0; i < as<size_t>(nAmalgNodesOnProc); i++ ) {
-     GO gid = rowMap->getGlobalElement(i*nPDEs);
-     temp =  ((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (gid))/((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (nPDEs));
-     nodalGIDs[i] = as<GO>(floor(temp));
-   }
-   GO nAmalgNodesGlobal = rowMap->getGlobalNumElements();
-   GO nBlkGlobal       = nAmalgNodesGlobal/nPDEs;
-   if (nBlkGlobal*nPDEs != nAmalgNodesGlobal)
-     TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of global dofs not divisible by BlkSize");
-
-   Teuchos::RCP<Map> AmalgRowMap = MapFactory::Build(rowMap->lib(), nBlkGlobal,
-                                                    nodalGIDs(),0,rowMap->getComm());
-
-   filteredGraph = rcp(new LWGraph(newRowPtr, finalCols, AmalgRowMap, AmalgRowMap, "thresholded graph of A"));
-   filteredGraph->SetBoundaryNodeMap(boundaryNodes);
+                   penaltyPolyCoef[poly1stOrderCoef] * (as<Scalar>(i)) +
+                   penaltyPolyCoef[poly2ndOrderCoef] * (as<Scalar>(i * i)) +
+                   (penaltyPolyCoef[poly3rdOrderCoef] * (as<Scalar>(i * i)) * (as<Scalar>(i))) +  // perhaps avoids overflow?
+                   (penaltyPolyCoef[poly4thOrderCoef] * (as<Scalar>(i * i)) * (as<Scalar>(i * i)));
+
+  LO nzTotal = 0, numBCols = 0, row = -1, Nbcols, bcol;
+  newRowPtr[0] = 0;
+
+  /* proceed block by block */
+  for (LO i = 0; i < as<LO>(nBlks); i++) {
+    newRowPtr[i + 1] = newRowPtr[i];
+    for (LO j = 0; j < nPDEs; j++) {
+      row = row + 1;
+
+      Teuchos::ArrayView<const LocalOrdinal> indices;
+      Teuchos::ArrayView<const Scalar> vals;
+
+      Amat.getLocalRowView(row, indices, vals);
+
+      if (indices.size() > maxNzPerRow) {
+        LO oldSize  = maxNzPerRow;
+        maxNzPerRow = indices.size() + 100;
+        penalties.resize(as<size_t>(maxNzPerRow), 0.0);
+        for (LO k = oldSize; k < maxNzPerRow; k++)
+          penalties[k] = penaltyPolyCoef[poly0thOrderCoef] +
+                         penaltyPolyCoef[poly1stOrderCoef] * (as<Scalar>(i)) +
+                         penaltyPolyCoef[poly2ndOrderCoef] * (as<Scalar>(i * i)) +
+                         (penaltyPolyCoef[poly3rdOrderCoef] * (as<Scalar>(i * i)) * (as<Scalar>(i))) +
+                         (penaltyPolyCoef[poly4thOrderCoef] * (as<Scalar>(i * i)) * (as<Scalar>(i * i)));
+      }
+      badGuysDropfunc(row, indices, vals, testVecs, nPDEs, penalties, nearNull, bcols, keepOrNot, Nbcols, nLoc);
+      for (LO k = 0; k < Nbcols; k++) {
+        bcol = bcols[k];
+
+        /* add to bColList if not already on it */
 
+        if (alreadyOnBColList[bcol] == false) { /* for PDE systems only record */
+          bColList[numBCols++]    = bcol;       /* neighboring block one time  */
+          alreadyOnBColList[bcol] = true;
+        }
+        /* drop if any pt row within block indicates entry should be dropped */
+
+        if (keepOrNot[k] == false) keepStatus[bcol] = false;
+
+      } /* for (k=0; k < Nbcols; k++)   */
+    }   /* for (j = 0; i < nPDEs; j++)  */
+
+    /* finished with block row. Now record block entries that we keep */
+    /* and reset keepStatus, bColList, and alreadyOnBColList.      */
+
+    if (numBCols < 2) boundaryNodes[i] = true;
+    for (LO j = 0; j < numBCols; j++) {
+      bcol = bColList[j];
+      if (keepStatus[bcol] == true) {
+        newCols[nzTotal] = bColList[j];
+        newRowPtr[i + 1]++;
+        nzTotal = nzTotal + 1;
+      }
+      keepStatus[bcol]        = true;
+      alreadyOnBColList[bcol] = false;
+      bColList[j]             = 0;
+    }
+    numBCols = 0;
+  } /* for (i = 0; i < nBlks; i++)  */
+
+  /* create array of the correct size and copy over newCols to it */
+
+  Teuchos::ArrayRCP<LO> finalCols(nzTotal);
+  for (LO i = 0; i < nzTotal; i++) finalCols[i] = newCols[i];
+
+  // Not using column map because we do not allow for any off-proc stuff.
+  // Not sure if this is okay.  FIXME
+
+  RCP<const Map> rowMap = Amat.getRowMap();  // , colMap = Amat.getColMap();
+
+  LO nAmalgNodesOnProc = rowMap->getLocalNumElements() / nPDEs;
+  Teuchos::Array<GO> nodalGIDs(nAmalgNodesOnProc);
+  typename Teuchos::ScalarTraits<Scalar>::coordinateType temp;
+  for (size_t i = 0; i < as<size_t>(nAmalgNodesOnProc); i++) {
+    GO gid       = rowMap->getGlobalElement(i * nPDEs);
+    temp         = ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(gid)) / ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(nPDEs));
+    nodalGIDs[i] = as<GO>(floor(temp));
   }
+  GO nAmalgNodesGlobal = rowMap->getGlobalNumElements();
+  GO nBlkGlobal        = nAmalgNodesGlobal / nPDEs;
+  if (nBlkGlobal * nPDEs != nAmalgNodesGlobal)
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Number of global dofs not divisible by BlkSize");
+
+  Teuchos::RCP<Map> AmalgRowMap = MapFactory::Build(rowMap->lib(), nBlkGlobal,
+                                                    nodalGIDs(), 0, rowMap->getComm());
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::badGuysDropfunc(LO row, const Teuchos::ArrayView<const LocalOrdinal>& cols, const Teuchos::ArrayView<const Scalar>& vals, const MultiVector&  testVecs, LO nPDEs, Teuchos::ArrayRCP<Scalar> & penalties, const MultiVector& nearNull, Teuchos::ArrayRCP<LO>& Bcols, Teuchos::ArrayRCP<bool>& keepOrNot, LO &Nbcols, LO nLoc) const {
-    using TST=Teuchos::ScalarTraits<Scalar>;
+  filteredGraph = rcp(new LWGraph(newRowPtr, finalCols, AmalgRowMap, AmalgRowMap, "thresholded graph of A"));
+  filteredGraph->SetBoundaryNodeMap(boundaryNodes);
+}
 
-  LO nLeng  = cols.size();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::badGuysDropfunc(LO row, const Teuchos::ArrayView<const LocalOrdinal>& cols, const Teuchos::ArrayView<const Scalar>& vals, const MultiVector& testVecs, LO nPDEs, Teuchos::ArrayRCP<Scalar>& penalties, const MultiVector& nearNull, Teuchos::ArrayRCP<LO>& Bcols, Teuchos::ArrayRCP<bool>& keepOrNot, LO& Nbcols, LO nLoc) const {
+  using TST = Teuchos::ScalarTraits<Scalar>;
+
+  LO nLeng = cols.size();
   typename TST::coordinateType temp;
-  temp =  ((typename TST::coordinateType) (row))/((typename TST::coordinateType) (nPDEs));
+  temp      = ((typename TST::coordinateType)(row)) / ((typename TST::coordinateType)(nPDEs));
   LO blkRow = as<LO>(floor(temp));
-  Teuchos::ArrayRCP<Scalar> badGuy( nLeng, 0.0);
-  Teuchos::ArrayRCP<Scalar> subNull(nLeng, 0.0);  /* subset of nearNull       */
-                                                  /* associated with current  */
-                                                  /* dof within node.         */
-  
+  Teuchos::ArrayRCP<Scalar> badGuy(nLeng, 0.0);
+  Teuchos::ArrayRCP<Scalar> subNull(nLeng, 0.0); /* subset of nearNull       */
+                                                 /* associated with current  */
+                                                 /* dof within node.         */
+
   /* Only consider testVecs associated with same dof & on processor. Further   */
   /* collapse testVecs to a single badGuy vector by basically taking the worst */
   /* (least smooth) values for each of the off diags. In particular, we look at*/
@@ -433,39 +424,37 @@ namespace MueLu {
   /* two guys are aggregated. So, the biggest ratio mismatch is used to choose */
   /* the testVec entry associated with each off-diagonal entry.                */
 
-
   for (LO i = 0; i < nLeng; i++) keepOrNot[i] = false;
 
-  LO diagInd = -1;
-  Nbcols     = 0;
-  LO rowDof  =  row  -  blkRow*nPDEs;
-  Teuchos::ArrayRCP< const Scalar > oneNull = nearNull.getData( as<size_t>(rowDof));
+  LO diagInd                              = -1;
+  Nbcols                                  = 0;
+  LO rowDof                               = row - blkRow * nPDEs;
+  Teuchos::ArrayRCP<const Scalar> oneNull = nearNull.getData(as<size_t>(rowDof));
 
   for (LO i = 0; i < nLeng; i++) {
-    if ((cols[i] < nLoc ) && (TST::magnitude(vals[i]) != 0.0)) {   /* on processor */
-      temp =  ((typename TST::coordinateType) (cols[i]))/((typename TST::coordinateType) (nPDEs));
-      LO colDof = cols[i] - (as<LO>(floor( temp    )))*nPDEs;
+    if ((cols[i] < nLoc) && (TST::magnitude(vals[i]) != 0.0)) { /* on processor */
+      temp      = ((typename TST::coordinateType)(cols[i])) / ((typename TST::coordinateType)(nPDEs));
+      LO colDof = cols[i] - (as<LO>(floor(temp))) * nPDEs;
       if (colDof == rowDof) { /* same dof within node as row */
-        Bcols[  Nbcols] = (cols[i] - colDof)/nPDEs;
+        Bcols[Nbcols]   = (cols[i] - colDof) / nPDEs;
         subNull[Nbcols] = oneNull[cols[i]];
 
         if (cols[i] != row) { /* not diagonal  */
-          Scalar worstRatio = -TST::one();
-          Scalar targetRatio = subNull[Nbcols]/oneNull[row];
+          Scalar worstRatio  = -TST::one();
+          Scalar targetRatio = subNull[Nbcols] / oneNull[row];
           Scalar actualRatio;
-          for (size_t kk = 0; kk < testVecs.getNumVectors(); kk++ ) {
-            Teuchos::ArrayRCP< const Scalar > curVec = testVecs.getData(kk);
-            actualRatio = curVec[cols[i]]/curVec[row];
+          for (size_t kk = 0; kk < testVecs.getNumVectors(); kk++) {
+            Teuchos::ArrayRCP<const Scalar> curVec = testVecs.getData(kk);
+            actualRatio                            = curVec[cols[i]] / curVec[row];
             if (TST::magnitude(actualRatio - targetRatio) > TST::magnitude(worstRatio)) {
-               badGuy[Nbcols] = actualRatio;
-               worstRatio = Teuchos::ScalarTraits<SC>::magnitude(actualRatio - targetRatio);
+              badGuy[Nbcols] = actualRatio;
+              worstRatio     = Teuchos::ScalarTraits<SC>::magnitude(actualRatio - targetRatio);
             }
           }
-        }
-        else {
-          badGuy[   Nbcols] = 1.;
+        } else {
+          badGuy[Nbcols]    = 1.;
           keepOrNot[Nbcols] = true;
-          diagInd            = Nbcols;
+          diagInd           = Nbcols;
         }
         (Nbcols)++;
       }
@@ -475,18 +464,18 @@ namespace MueLu {
   /* Make sure that diagonal entry is in block col list */
 
   if (diagInd == -1) {
-    Bcols[    Nbcols] = (row - rowDof)/nPDEs;
-    subNull[  Nbcols] = 1.;
-    badGuy[   Nbcols] = 1.;
+    Bcols[Nbcols]     = (row - rowDof) / nPDEs;
+    subNull[Nbcols]   = 1.;
+    badGuy[Nbcols]    = 1.;
     keepOrNot[Nbcols] = true;
-    diagInd            = Nbcols;
+    diagInd           = Nbcols;
     (Nbcols)++;
   }
 
-  Scalar currentRP          = oneNull[row]*oneNull[row];
-  Scalar currentRTimesBadGuy= oneNull[row]*badGuy[diagInd];
-  Scalar currentScore       = penalties[0]; /* (I - P inv(R*P)*R )=0 for size */
-                                     /* size 1 agg, so fit is perfect  */
+  Scalar currentRP           = oneNull[row] * oneNull[row];
+  Scalar currentRTimesBadGuy = oneNull[row] * badGuy[diagInd];
+  Scalar currentScore        = penalties[0]; /* (I - P inv(R*P)*R )=0 for size */
+                                             /* size 1 agg, so fit is perfect  */
 
   /* starting from a set that only includes the diagonal entry consider adding */
   /* one off-diagonal at a time until the fitValue exceeds the penalty term.   */
@@ -496,7 +485,7 @@ namespace MueLu {
   /* includes the diagonal, all already determined neighbors, and the potential*/
   /* new neighbor                                                              */
 
-  LO nKeep = 1, flag  = 1, minId;
+  LO nKeep = 1, flag      = 1, minId;
   Scalar minFit, minFitRP = 0., minFitRTimesBadGuy = 0.;
   Scalar newRP, newRTimesBadGuy;
 
@@ -507,45 +496,46 @@ namespace MueLu {
     minFit = 1000000.;
     minId  = -1;
 
-    for (LO i=0; i < Nbcols; i++) {
+    for (LO i = 0; i < Nbcols; i++) {
       if (keepOrNot[i] == false) {
-        keepOrNot[i] = true;  /* temporarily view i as non-dropped neighbor */
-        newRP          = currentRP           + subNull[i]*subNull[i];
-        newRTimesBadGuy= currentRTimesBadGuy + subNull[i]*badGuy[i];
-        Scalar ratio = newRTimesBadGuy/newRP;
+        keepOrNot[i]    = true; /* temporarily view i as non-dropped neighbor */
+        newRP           = currentRP + subNull[i] * subNull[i];
+        newRTimesBadGuy = currentRTimesBadGuy + subNull[i] * badGuy[i];
+        Scalar ratio    = newRTimesBadGuy / newRP;
 
         Scalar newFit = 0.0;
-        for (LO k=0; k < Nbcols; k++) {
+        for (LO k = 0; k < Nbcols; k++) {
           if (keepOrNot[k] == true) {
-             Scalar diff = badGuy[k] - ratio*subNull[k];
-             newFit = newFit + diff*diff;
+            Scalar diff = badGuy[k] - ratio * subNull[k];
+            newFit      = newFit + diff * diff;
           }
         }
         if (Teuchos::ScalarTraits<SC>::magnitude(newFit) < Teuchos::ScalarTraits<SC>::magnitude(minFit)) {
-          minId     = i;
-          minFit    = newFit;
-          minFitRP  = newRP;
-          minFitRTimesBadGuy= newRTimesBadGuy;
+          minId              = i;
+          minFit             = newFit;
+          minFitRP           = newRP;
+          minFitRTimesBadGuy = newRTimesBadGuy;
         }
         keepOrNot[i] = false;
       }
     }
-    if (minId == -1) flag = 0;
+    if (minId == -1)
+      flag = 0;
     else {
-      minFit = sqrt(minFit);
+      minFit          = sqrt(minFit);
       Scalar newScore = penalties[nKeep] + minFit;
       if (Teuchos::ScalarTraits<SC>::magnitude(newScore) < Teuchos::ScalarTraits<SC>::magnitude(currentScore)) {
-        nKeep           = nKeep + 1;
-        keepOrNot[minId]= true;
-        currentScore    = newScore;
-        currentRP       = minFitRP;
-        currentRTimesBadGuy= minFitRTimesBadGuy;
-      }
-      else flag = 0;
+        nKeep               = nKeep + 1;
+        keepOrNot[minId]    = true;
+        currentScore        = newScore;
+        currentRP           = minFitRP;
+        currentRTimesBadGuy = minFitRTimesBadGuy;
+      } else
+        flag = 0;
     }
   }
-  }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_SMOOVECCOALESCEDROPFACTORY_DEF_HPP
+#endif  // MUELU_SMOOVECCOALESCEDROPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_decl.hpp
index 80b779e26a7d..a66e5570c962 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_decl.hpp
@@ -54,85 +54,79 @@
 
 namespace MueLu {
 
-  /*!
-    @class UnsmooshFactory class.
-    @brief Factory for building "unsmooshed" transfer operators from transfer operators associated with a scalar helper problem (built by the VariableDofLaplacianFactory)
-
-    The output prolongation operator P is compatible to the input matrix A. The input prolongation matrix P is supposed to be built by the VariableDofLaplacianFactory with 1 DOF per node.
-    The DofStatus array declares for each row, whether it is a standard dof or a padded/pseudo Dirichlet dof.
-    The user has to provide the parameter "maxDofPerNode", since the coarse matrices are always padded by construction (i.e. constant number of DOFs per node). This parameter is needed for building the column map of the unsmooshed prolongation operator.
-    The parameter "fineIsPadded" allows to specify whether the input matrix A on the finest level is padded or not. In the padded case we have an artificially extended input matrix with pseudo Dirichlet rows on inactive rows. In the non-padded version
-    we have "variable" number of Dofs per node.
-
-    @ingroup MueLuGraphClasses
-
-    ## Input/output of UnsmooshFactory ##
-
-    ### User parameters of UnsmooshFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A        | Factory | null |   | * | * | Generating factory of the input matrix A with potentially variable number of DOFs. Might be padded or non-padded. Padded means, that the matrix has additional artificial rows and columns to have a constant number of DOFs per node. Needed to match the row map for the unsmooshed version of P with the row map of the unamalgamated input matrix A.
-     P        | Factory | null |   | * | * | Generating factory of the (amalgamated) prolongator P generated from a (pseudo Laplacian) with 1 Dofs per node. Will be unsmooshed to be compatible with input matrix A in this factory.
-     DofStatus |Facotry | null |   | * | * | Generating factory for dofStatus array (usually generated by the VariableDofLaplacdianFactory). It is a Teuchos::Array<char> of size number of Rows of input matrix P multiplied by the maximum possible number of Dofs per node.
-     maxDofPerNode | int | 1 |      | * |   | Maximum number of DOFs per node. Needed for generating unsmooshed P. Note, that the coarse matrices are always padded, i.e. we have constant number of DOFs per node.
-     fineIsPadded  | bool | false   | * |   | True if finest level input matrix is padded (default is false)
-
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see UnsmooshFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see UnsmooshFactory::DeclareInput).
-
-    ### Variables provided by UnsmooshFactory ###
-
-    After UnsmooshFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | UnsmooshFactory   | Unsmooshed prolongation operator
-  */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class UnsmooshFactory : public PFactory {
+/*!
+  @class UnsmooshFactory class.
+  @brief Factory for building "unsmooshed" transfer operators from transfer operators associated with a scalar helper problem (built by the VariableDofLaplacianFactory)
+
+  The output prolongation operator P is compatible to the input matrix A. The input prolongation matrix P is supposed to be built by the VariableDofLaplacianFactory with 1 DOF per node.
+  The DofStatus array declares for each row, whether it is a standard dof or a padded/pseudo Dirichlet dof.
+  The user has to provide the parameter "maxDofPerNode", since the coarse matrices are always padded by construction (i.e. constant number of DOFs per node). This parameter is needed for building the column map of the unsmooshed prolongation operator.
+  The parameter "fineIsPadded" allows to specify whether the input matrix A on the finest level is padded or not. In the padded case we have an artificially extended input matrix with pseudo Dirichlet rows on inactive rows. In the non-padded version
+  we have "variable" number of Dofs per node.
+
+  @ingroup MueLuGraphClasses
+
+  ## Input/output of UnsmooshFactory ##
+
+  ### User parameters of UnsmooshFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A        | Factory | null |   | * | * | Generating factory of the input matrix A with potentially variable number of DOFs. Might be padded or non-padded. Padded means, that the matrix has additional artificial rows and columns to have a constant number of DOFs per node. Needed to match the row map for the unsmooshed version of P with the row map of the unamalgamated input matrix A.
+   P        | Factory | null |   | * | * | Generating factory of the (amalgamated) prolongator P generated from a (pseudo Laplacian) with 1 Dofs per node. Will be unsmooshed to be compatible with input matrix A in this factory.
+   DofStatus |Facotry | null |   | * | * | Generating factory for dofStatus array (usually generated by the VariableDofLaplacdianFactory). It is a Teuchos::Array<char> of size number of Rows of input matrix P multiplied by the maximum possible number of Dofs per node.
+   maxDofPerNode | int | 1 |      | * |   | Maximum number of DOFs per node. Needed for generating unsmooshed P. Note, that the coarse matrices are always padded, i.e. we have constant number of DOFs per node.
+   fineIsPadded  | bool | false   | * |   | True if finest level input matrix is padded (default is false)
+
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see UnsmooshFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see UnsmooshFactory::DeclareInput).
+
+  ### Variables provided by UnsmooshFactory ###
+
+  After UnsmooshFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | UnsmooshFactory   | Unsmooshed prolongation operator
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class UnsmooshFactory : public PFactory {
 #undef MUELU_UNSMOOSHFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  UnsmooshFactory();
 
-    //! Constructor
-    UnsmooshFactory();
+  //! Destructor
+  virtual ~UnsmooshFactory() {}
 
-    //! Destructor
-    virtual ~UnsmooshFactory() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //@}
+  void Build(Level &fineLevel, Level &coarseLevel) const;                    // Build
+  void BuildP(Level & /* fineLevel */, Level & /* coarseLevel */) const {};  // TAW no real need for an extra BuildP routine. Just use Build
 
-    void Build (Level &fineLevel, Level &coarseLevel) const; // Build
-    void BuildP(Level &/* fineLevel */, Level &/* coarseLevel */) const {}; // TAW no real need for an extra BuildP routine. Just use Build
+ private:
+};  // class UnsmooshFactory
 
-  private:
-
-
-
-
-  }; //class UnsmooshFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UNSMOOSHFACTORY_SHORT
 
-
 #endif /* PACKAGES_MUELU_SRC_GRAPH_MUELU_UNSMOOSHFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_def.hpp
index f49f488463ea..cc87c1c8988f 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_UnsmooshFactory_def.hpp
@@ -53,198 +53,193 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::UnsmooshFactory() { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory for unamalgamated matrix. Row map of (unamalgamted) output prolongation operator should match row map of this A.");
-    validParamList->set< RCP<const FactoryBase> >("P",                  Teuchos::null, "Generating factory of the (amalgamated) prolongator P");
-    validParamList->set< RCP<const FactoryBase> >("DofStatus",          Teuchos::null, "Generating factory for dofStatus array (usually the VariableDofLaplacdianFactory)");
-
-    validParamList->set< int  >                  ("maxDofPerNode", 1,     "Maximum number of DOFs per node");
-    validParamList->set< bool >                  ("fineIsPadded" , false, "true if finest level input matrix is padded");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    //const ParameterList& pL = GetParameterList();
-    Input(fineLevel, "A");
-    Input(coarseLevel, "P");
-
-    // DofStatus only provided on the finest level (by user)
-    // On the coarser levels it is auto-generated using the DBC information from the unamalgamated matrix A
-    if(fineLevel.GetLevelID() == 0)
-      Input(fineLevel, "DofStatus");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-    typedef Teuchos::ScalarTraits<SC> STS;
-
-    const ParameterList  & pL = GetParameterList();
-
-    // extract matrices (unamalgamated A and amalgamated P)
-    RCP<Matrix> unamalgA = Get< RCP<Matrix> >(fineLevel,   "A");
-    RCP<Matrix> amalgP   = Get< RCP<Matrix> >(coarseLevel, "P");
-
-    // extract user parameters
-    int maxDofPerNode = pL.get<int> ("maxDofPerNode");
-    bool fineIsPadded = pL.get<bool>("fineIsPadded");
-
-    // get dofStatus information
-    // On the finest level it is provided by the user. On the coarser levels it is constructed
-    // using the DBC information of the matrix A
-    Teuchos::Array<char> dofStatus;
-    if(fineLevel.GetLevelID() == 0) {
-      dofStatus = Get<Teuchos::Array<char> >(fineLevel, "DofStatus");
-    } else {
-      // dof status is the dirichlet information of unsmooshed/unamalgamated A (fine level)
-      dofStatus = Teuchos::Array<char>(unamalgA->getRowMap()->getLocalNumElements() /*amalgP->getRowMap()->getLocalNumElements() * maxDofPerNode*/,'s');
-
-      bool bHasZeroDiagonal = false;
-      Teuchos::ArrayRCP<const bool> dirOrNot = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DetectDirichletRowsExt(*unamalgA,bHasZeroDiagonal,STS::magnitude(0.5));
-
-      TEUCHOS_TEST_FOR_EXCEPTION(dirOrNot.size() != dofStatus.size(), MueLu::Exceptions::RuntimeError,"MueLu::UnsmooshFactory::Build: inconsistent number of coarse DBC array and dofStatus array. dirOrNot.size() = " << dirOrNot.size() << " dofStatus.size() = " << dofStatus.size());
-      for(decltype(dirOrNot.size()) i = 0; i < dirOrNot.size(); ++i) {
-        if(dirOrNot[i] == true) dofStatus[i] = 'p';
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::UnsmooshFactory() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for unamalgamated matrix. Row map of (unamalgamted) output prolongation operator should match row map of this A.");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the (amalgamated) prolongator P");
+  validParamList->set<RCP<const FactoryBase> >("DofStatus", Teuchos::null, "Generating factory for dofStatus array (usually the VariableDofLaplacdianFactory)");
+
+  validParamList->set<int>("maxDofPerNode", 1, "Maximum number of DOFs per node");
+  validParamList->set<bool>("fineIsPadded", false, "true if finest level input matrix is padded");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  // const ParameterList& pL = GetParameterList();
+  Input(fineLevel, "A");
+  Input(coarseLevel, "P");
+
+  // DofStatus only provided on the finest level (by user)
+  // On the coarser levels it is auto-generated using the DBC information from the unamalgamated matrix A
+  if (fineLevel.GetLevelID() == 0)
+    Input(fineLevel, "DofStatus");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+  typedef Teuchos::ScalarTraits<SC> STS;
+
+  const ParameterList &pL = GetParameterList();
+
+  // extract matrices (unamalgamated A and amalgamated P)
+  RCP<Matrix> unamalgA = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<Matrix> amalgP   = Get<RCP<Matrix> >(coarseLevel, "P");
+
+  // extract user parameters
+  int maxDofPerNode = pL.get<int>("maxDofPerNode");
+  bool fineIsPadded = pL.get<bool>("fineIsPadded");
+
+  // get dofStatus information
+  // On the finest level it is provided by the user. On the coarser levels it is constructed
+  // using the DBC information of the matrix A
+  Teuchos::Array<char> dofStatus;
+  if (fineLevel.GetLevelID() == 0) {
+    dofStatus = Get<Teuchos::Array<char> >(fineLevel, "DofStatus");
+  } else {
+    // dof status is the dirichlet information of unsmooshed/unamalgamated A (fine level)
+    dofStatus = Teuchos::Array<char>(unamalgA->getRowMap()->getLocalNumElements() /*amalgP->getRowMap()->getLocalNumElements() * maxDofPerNode*/, 's');
+
+    bool bHasZeroDiagonal                  = false;
+    Teuchos::ArrayRCP<const bool> dirOrNot = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DetectDirichletRowsExt(*unamalgA, bHasZeroDiagonal, STS::magnitude(0.5));
+
+    TEUCHOS_TEST_FOR_EXCEPTION(dirOrNot.size() != dofStatus.size(), MueLu::Exceptions::RuntimeError, "MueLu::UnsmooshFactory::Build: inconsistent number of coarse DBC array and dofStatus array. dirOrNot.size() = " << dirOrNot.size() << " dofStatus.size() = " << dofStatus.size());
+    for (decltype(dirOrNot.size()) i = 0; i < dirOrNot.size(); ++i) {
+      if (dirOrNot[i] == true) dofStatus[i] = 'p';
     }
+  }
 
-    // TODO: TAW the following check is invalid for SA-AMG based input prolongators
-    //TEUCHOS_TEST_FOR_EXCEPTION(amalgP->getDomainMap()->isSameAs(*amalgP->getColMap()) == false, MueLu::Exceptions::RuntimeError,"MueLu::UnsmooshFactory::Build: only support for non-overlapping aggregates. (column map of Ptent must be the same as domain map of Ptent)");
-
-    // extract CRS information from amalgamated prolongation operator
-    Teuchos::ArrayRCP<const size_t> amalgRowPtr(amalgP->getLocalNumRows());
-    Teuchos::ArrayRCP<const LocalOrdinal> amalgCols(amalgP->getLocalNumEntries());
-    Teuchos::ArrayRCP<const Scalar> amalgVals(amalgP->getLocalNumEntries());
-    Teuchos::RCP<CrsMatrixWrap> amalgPwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(amalgP);
-    Teuchos::RCP<CrsMatrix> amalgPcrs = amalgPwrap->getCrsMatrix();
-    amalgPcrs->getAllValues(amalgRowPtr, amalgCols, amalgVals);
-
-    // calculate number of dof rows for new prolongator
-    size_t paddedNrows = amalgP->getRowMap()->getLocalNumElements() * Teuchos::as<size_t>(maxDofPerNode);
-
-    // reserve CSR arrays for new prolongation operator
-    Teuchos::ArrayRCP<size_t> newPRowPtr(paddedNrows+1);
-    Teuchos::ArrayRCP<LocalOrdinal> newPCols(amalgP->getLocalNumEntries() * maxDofPerNode);
-    Teuchos::ArrayRCP<Scalar> newPVals(amalgP->getLocalNumEntries() * maxDofPerNode);
-
-    size_t rowCount = 0; // actual number of (local) in unamalgamated prolongator
-    if(fineIsPadded == true || fineLevel.GetLevelID() > 0) {
-
-      // build prolongation operator for padded fine level matrices.
-      // Note: padded fine level dofs are transferred by injection.
-      // That is, these interpolation stencils do not take averages of
-      // coarse level variables. Further, fine level Dirichlet points
-      // also use injection.
-
-      size_t cnt = 0; // local id counter
-      for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
-        // determine number of entries in amalgamated dof row i
-        size_t rowLength = amalgRowPtr[i+1] - amalgRowPtr[i];
-
-        // loop over dofs per node (unamalgamation)
-        for(int j = 0; j < maxDofPerNode; j++) {
-          newPRowPtr[i*maxDofPerNode+j] = cnt;
-          if (dofStatus[i*maxDofPerNode+j] == 's') { // add only "standard" dofs to unamalgamated prolongator
-            // loop over column entries in amalgamated P
-            for (size_t k = 0; k < rowLength; k++) {
-              newPCols[cnt  ] = amalgCols[k+amalgRowPtr[i]] * maxDofPerNode + j;
-              newPVals[cnt++] = amalgVals[k+amalgRowPtr[i]];
-            }
-
+  // TODO: TAW the following check is invalid for SA-AMG based input prolongators
+  // TEUCHOS_TEST_FOR_EXCEPTION(amalgP->getDomainMap()->isSameAs(*amalgP->getColMap()) == false, MueLu::Exceptions::RuntimeError,"MueLu::UnsmooshFactory::Build: only support for non-overlapping aggregates. (column map of Ptent must be the same as domain map of Ptent)");
+
+  // extract CRS information from amalgamated prolongation operator
+  Teuchos::ArrayRCP<const size_t> amalgRowPtr(amalgP->getLocalNumRows());
+  Teuchos::ArrayRCP<const LocalOrdinal> amalgCols(amalgP->getLocalNumEntries());
+  Teuchos::ArrayRCP<const Scalar> amalgVals(amalgP->getLocalNumEntries());
+  Teuchos::RCP<CrsMatrixWrap> amalgPwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(amalgP);
+  Teuchos::RCP<CrsMatrix> amalgPcrs      = amalgPwrap->getCrsMatrix();
+  amalgPcrs->getAllValues(amalgRowPtr, amalgCols, amalgVals);
+
+  // calculate number of dof rows for new prolongator
+  size_t paddedNrows = amalgP->getRowMap()->getLocalNumElements() * Teuchos::as<size_t>(maxDofPerNode);
+
+  // reserve CSR arrays for new prolongation operator
+  Teuchos::ArrayRCP<size_t> newPRowPtr(paddedNrows + 1);
+  Teuchos::ArrayRCP<LocalOrdinal> newPCols(amalgP->getLocalNumEntries() * maxDofPerNode);
+  Teuchos::ArrayRCP<Scalar> newPVals(amalgP->getLocalNumEntries() * maxDofPerNode);
+
+  size_t rowCount = 0;  // actual number of (local) in unamalgamated prolongator
+  if (fineIsPadded == true || fineLevel.GetLevelID() > 0) {
+    // build prolongation operator for padded fine level matrices.
+    // Note: padded fine level dofs are transferred by injection.
+    // That is, these interpolation stencils do not take averages of
+    // coarse level variables. Further, fine level Dirichlet points
+    // also use injection.
+
+    size_t cnt = 0;  // local id counter
+    for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+      // determine number of entries in amalgamated dof row i
+      size_t rowLength = amalgRowPtr[i + 1] - amalgRowPtr[i];
+
+      // loop over dofs per node (unamalgamation)
+      for (int j = 0; j < maxDofPerNode; j++) {
+        newPRowPtr[i * maxDofPerNode + j] = cnt;
+        if (dofStatus[i * maxDofPerNode + j] == 's') {  // add only "standard" dofs to unamalgamated prolongator
+          // loop over column entries in amalgamated P
+          for (size_t k = 0; k < rowLength; k++) {
+            newPCols[cnt]   = amalgCols[k + amalgRowPtr[i]] * maxDofPerNode + j;
+            newPVals[cnt++] = amalgVals[k + amalgRowPtr[i]];
           }
         }
       }
+    }
 
-      newPRowPtr[paddedNrows] = cnt; // close row CSR array
-      rowCount = paddedNrows;
-    } else {
-      // Build prolongation operator for non-padded fine level matrices.
-      // Need to map from non-padded dofs to padded dofs. For this, look
-      // at the status array and skip padded dofs.
-
-      size_t cnt = 0; // local id counter
-
-      for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
-        // determine number of entries in amalgamated dof row i
-        size_t rowLength = amalgRowPtr[i+1] - amalgRowPtr[i];
-
-        // loop over dofs per node (unamalgamation)
-        for(int j = 0; j < maxDofPerNode; j++) {
-          // no interpolation for padded fine dofs as they do not exist
-
-          if (dofStatus[i*maxDofPerNode+j] == 's') { // add only "standard" dofs to unamalgamated prolongator
-            newPRowPtr[rowCount++] = cnt;
-            // loop over column entries in amalgamated P
-            for (size_t k = 0; k < rowLength; k++) {
-              newPCols[cnt  ] = amalgCols[k+amalgRowPtr[i]] * maxDofPerNode + j;
-              newPVals[cnt++] = amalgVals[k+amalgRowPtr[i]];
-            }
-
-          }
-          if (dofStatus[i*maxDofPerNode+j] == 'd') { // Dirichlet handling
-            newPRowPtr[rowCount++] = cnt;
+    newPRowPtr[paddedNrows] = cnt;  // close row CSR array
+    rowCount                = paddedNrows;
+  } else {
+    // Build prolongation operator for non-padded fine level matrices.
+    // Need to map from non-padded dofs to padded dofs. For this, look
+    // at the status array and skip padded dofs.
+
+    size_t cnt = 0;  // local id counter
+
+    for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+      // determine number of entries in amalgamated dof row i
+      size_t rowLength = amalgRowPtr[i + 1] - amalgRowPtr[i];
+
+      // loop over dofs per node (unamalgamation)
+      for (int j = 0; j < maxDofPerNode; j++) {
+        // no interpolation for padded fine dofs as they do not exist
+
+        if (dofStatus[i * maxDofPerNode + j] == 's') {  // add only "standard" dofs to unamalgamated prolongator
+          newPRowPtr[rowCount++] = cnt;
+          // loop over column entries in amalgamated P
+          for (size_t k = 0; k < rowLength; k++) {
+            newPCols[cnt]   = amalgCols[k + amalgRowPtr[i]] * maxDofPerNode + j;
+            newPVals[cnt++] = amalgVals[k + amalgRowPtr[i]];
           }
         }
-      }
-      newPRowPtr[rowCount] = cnt; // close row CSR array
-    } // fineIsPadded == false
-
-    // generate coarse domain map
-    // So far no support for gid offset or strided maps. This information
-    // could be gathered easily from the unamalgamated fine level operator A.
-    std::vector<size_t> stridingInfo(1, maxDofPerNode);
-
-    GlobalOrdinal nCoarseDofs = amalgP->getDomainMap()->getLocalNumElements() * maxDofPerNode;
-    GlobalOrdinal indexBase   = amalgP->getDomainMap()->getIndexBase();
-    RCP<const Map> coarseDomainMap = StridedMapFactory::Build(amalgP->getDomainMap()->lib(),
-        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-        nCoarseDofs,
-        indexBase,
-        stridingInfo,
-        amalgP->getDomainMap()->getComm(),
-        -1 /* stridedBlockId */,
-        0  /*domainGidOffset */);
-
-    size_t nColCoarseDofs = Teuchos::as<size_t>(amalgP->getColMap()->getLocalNumElements() * maxDofPerNode);
-    Teuchos::Array<GlobalOrdinal> unsmooshColMapGIDs(nColCoarseDofs);
-    for(size_t c = 0; c < amalgP->getColMap()->getLocalNumElements(); ++c) {
-      GlobalOrdinal gid = (amalgP->getColMap()->getGlobalElement(c)-indexBase) * maxDofPerNode + indexBase;
-
-      for(int i = 0; i < maxDofPerNode; ++i) {
-        unsmooshColMapGIDs[c * maxDofPerNode + i] = gid + i;
+        if (dofStatus[i * maxDofPerNode + j] == 'd') {  // Dirichlet handling
+          newPRowPtr[rowCount++] = cnt;
+        }
       }
     }
-    Teuchos::RCP<Map> coarseColMap = MapFactory::Build(amalgP->getDomainMap()->lib(),
-                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                       unsmooshColMapGIDs(), //View,
-                                                       indexBase,
-                                                       amalgP->getDomainMap()->getComm());
-
-    // Assemble unamalgamated P
-    Teuchos::RCP<CrsMatrix> unamalgPCrs = CrsMatrixFactory::Build(unamalgA->getRowMap(),
-                                                                  coarseColMap,
-                                                                  maxDofPerNode*amalgP->getLocalMaxNumRowEntries());
-    for (size_t i = 0; i < rowCount; i++) {
-      unamalgPCrs->insertLocalValues(i,
-                                     newPCols.view(newPRowPtr[i], newPRowPtr[i+1] - newPRowPtr[i]),
-                                     newPVals.view(newPRowPtr[i], newPRowPtr[i+1] - newPRowPtr[i]));
+    newPRowPtr[rowCount] = cnt;  // close row CSR array
+  }                              // fineIsPadded == false
+
+  // generate coarse domain map
+  // So far no support for gid offset or strided maps. This information
+  // could be gathered easily from the unamalgamated fine level operator A.
+  std::vector<size_t> stridingInfo(1, maxDofPerNode);
+
+  GlobalOrdinal nCoarseDofs      = amalgP->getDomainMap()->getLocalNumElements() * maxDofPerNode;
+  GlobalOrdinal indexBase        = amalgP->getDomainMap()->getIndexBase();
+  RCP<const Map> coarseDomainMap = StridedMapFactory::Build(amalgP->getDomainMap()->lib(),
+                                                            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                            nCoarseDofs,
+                                                            indexBase,
+                                                            stridingInfo,
+                                                            amalgP->getDomainMap()->getComm(),
+                                                            -1 /* stridedBlockId */,
+                                                            0 /*domainGidOffset */);
+
+  size_t nColCoarseDofs = Teuchos::as<size_t>(amalgP->getColMap()->getLocalNumElements() * maxDofPerNode);
+  Teuchos::Array<GlobalOrdinal> unsmooshColMapGIDs(nColCoarseDofs);
+  for (size_t c = 0; c < amalgP->getColMap()->getLocalNumElements(); ++c) {
+    GlobalOrdinal gid = (amalgP->getColMap()->getGlobalElement(c) - indexBase) * maxDofPerNode + indexBase;
+
+    for (int i = 0; i < maxDofPerNode; ++i) {
+      unsmooshColMapGIDs[c * maxDofPerNode + i] = gid + i;
     }
-    unamalgPCrs->fillComplete(coarseDomainMap, unamalgA->getRowMap());
-
-    Teuchos::RCP<Matrix> unamalgP = Teuchos::rcp(new CrsMatrixWrap(unamalgPCrs));
-
-    Set(coarseLevel,"P",unamalgP);
   }
+  Teuchos::RCP<Map> coarseColMap = MapFactory::Build(amalgP->getDomainMap()->lib(),
+                                                     Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                     unsmooshColMapGIDs(),  // View,
+                                                     indexBase,
+                                                     amalgP->getDomainMap()->getComm());
+
+  // Assemble unamalgamated P
+  Teuchos::RCP<CrsMatrix> unamalgPCrs = CrsMatrixFactory::Build(unamalgA->getRowMap(),
+                                                                coarseColMap,
+                                                                maxDofPerNode * amalgP->getLocalMaxNumRowEntries());
+  for (size_t i = 0; i < rowCount; i++) {
+    unamalgPCrs->insertLocalValues(i,
+                                   newPCols.view(newPRowPtr[i], newPRowPtr[i + 1] - newPRowPtr[i]),
+                                   newPVals.view(newPRowPtr[i], newPRowPtr[i + 1] - newPRowPtr[i]));
+  }
+  unamalgPCrs->fillComplete(coarseDomainMap, unamalgA->getRowMap());
 
+  Teuchos::RCP<Matrix> unamalgP = Teuchos::rcp(new CrsMatrixWrap(unamalgPCrs));
 
-} /* MueLu */
+  Set(coarseLevel, "P", unamalgP);
+}
 
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_GRAPH_MUELU_UNSMOOSHFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_decl.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_decl.hpp
index 51a333e52d81..319fbfcc56eb 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_decl.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_decl.hpp
@@ -47,7 +47,6 @@
 #ifndef PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DECL_HPP_
 #define PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DECL_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_SingleLevelFactoryBase.hpp"
 #include "MueLu_VariableDofLaplacianFactory_fwd.hpp"
@@ -57,309 +56,295 @@
 
 namespace MueLu {
 
-  /*!
-    @class VariableDofLaplacianFactory class.
-    @brief Factory for building scalar Laplace operator (that is used as fake operator for variable dof size problems)
+/*!
+  @class VariableDofLaplacianFactory class.
+  @brief Factory for building scalar Laplace operator (that is used as fake operator for variable dof size problems)
 
-    Build distance Laplacian associated with input matrix A (which might have a variable number of DOFs per node).
-    Coordinates are needed to calculate the distance laplacian values. The user-provided array "DofPresent" stores whether
-    an array is present (=1) or not (=0) in the matrix. The length of the array is number of nodes * maxDofPerNode and
-    therefore it might be larger or equal than the number of rows in the input matrix.
+  Build distance Laplacian associated with input matrix A (which might have a variable number of DOFs per node).
+  Coordinates are needed to calculate the distance laplacian values. The user-provided array "DofPresent" stores whether
+  an array is present (=1) or not (=0) in the matrix. The length of the array is number of nodes * maxDofPerNode and
+  therefore it might be larger or equal than the number of rows in the input matrix.
 
-    The factory produces the distance laplacian matrix A as output (with one dof per node) as well as the coarse version
-    of the DofStatus (needed for the next coarser level), containing information about (artificial) Dirichlet rows in the matrix.
+  The factory produces the distance laplacian matrix A as output (with one dof per node) as well as the coarse version
+  of the DofStatus (needed for the next coarser level), containing information about (artificial) Dirichlet rows in the matrix.
 
-    @ingroup MueLuGraphClasses
+  @ingroup MueLuGraphClasses
 
-    ## Input/output of VariableDofLaplacianFactory ##
+  ## Input/output of VariableDofLaplacianFactory ##
 
-    ### User parameters of VariableDofLaplacianFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A        | Factory | null |   | * | *  | Generating factory of the input matrix A with potentially variable number of DOFs. Might be padded or non-padded. Padded means, that the matrix has additional artificial rows and columns to have a constant number of DOFs per node.
-     Coordinates | Factory | null | | * | * | Generating factory for Coordinates needed for building distance laplacian.
-     DofPresent | Teuchos::ArrayRCP<LocalOrdinal> | NoFactory | | | (*) | Optional array containing information whether DOF is actually present in matrix or not.
-     Advanced Dirichlet: threshold | double | 1e-5 |   | * |   | Drop tolerance for Dirichlet detection
-     Variable DOF amalgamation: threshold | double | 1.8e-9 |  | * |  | Drop tolerance for amalgamation process
-     maxDofPerNode | int | 1 |  | * |  | Maximum number of DOFs per node
+  ### User parameters of VariableDofLaplacianFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A        | Factory | null |   | * | *  | Generating factory of the input matrix A with potentially variable number of DOFs. Might be padded or non-padded. Padded means, that the matrix has additional artificial rows and columns to have a constant number of DOFs per node.
+   Coordinates | Factory | null | | * | * | Generating factory for Coordinates needed for building distance laplacian.
+   DofPresent | Teuchos::ArrayRCP<LocalOrdinal> | NoFactory | | | (*) | Optional array containing information whether DOF is actually present in matrix or not.
+   Advanced Dirichlet: threshold | double | 1e-5 |   | * |   | Drop tolerance for Dirichlet detection
+   Variable DOF amalgamation: threshold | double | 1.8e-9 |  | * |  | Drop tolerance for amalgamation process
+   maxDofPerNode | int | 1 |  | * |  | Maximum number of DOFs per node
 
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see VariableDofLaplacianFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see VariableDofLaplacianFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see VariableDofLaplacianFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see VariableDofLaplacianFactory::DeclareInput).
 
-    ### Variables provided by VariableDofLaplacianFactory ###
+  ### Variables provided by VariableDofLaplacianFactory ###
 
-    After TentativePFactory::Build the following data is available (if requested)
+  After TentativePFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | A       | VariableDofLaplacianFactory   | Laplacian operator
-    | DofStatus | VariableDofLaplacianFactory | Status array for next coarse level
-  */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class VariableDofLaplacianFactory : public SingleLevelFactoryBase {
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | A       | VariableDofLaplacianFactory   | Laplacian operator
+  | DofStatus | VariableDofLaplacianFactory | Status array for next coarse level
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class VariableDofLaplacianFactory : public SingleLevelFactoryBase {
 #undef MUELU_VARIABLEDOFLAPLACIANFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor
-    VariableDofLaplacianFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor
-    virtual ~VariableDofLaplacianFactory() { }
+  //! Constructor
+  VariableDofLaplacianFactory();
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor
+  virtual ~VariableDofLaplacianFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    void Build(Level &currentLevel) const; // Build
+  //@}
 
-  private:
+  void Build(Level& currentLevel) const;  // Build
 
-    void buildPaddedMap(const Teuchos::ArrayRCP<const LocalOrdinal> & dofPresent, std::vector<LocalOrdinal> & map, size_t nDofs) const;
-    void assignGhostLocalNodeIds(const Teuchos::RCP<const Map> & rowDofMap, const Teuchos::RCP<const Map> & colDofMap, std::vector<LocalOrdinal> & myLocalNodeIds, const std::vector<LocalOrdinal> & dofMap, size_t maxDofPerNode, size_t& nLocalNodes, size_t& nLocalPlusGhostNodes, Teuchos::RCP< const Teuchos::Comm< int > > comm) const;
-    void squeezeOutNnzs(Teuchos::ArrayRCP<size_t> & rowPtr, Teuchos::ArrayRCP<LocalOrdinal> & cols, Teuchos::ArrayRCP<Scalar> & vals, const std::vector<bool>& keep) const;
-    void buildLaplacian(const Teuchos::ArrayRCP<size_t>& rowPtr, const Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals, const size_t& numdim, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > & ghostedCoords) const;
+ private:
+  void buildPaddedMap(const Teuchos::ArrayRCP<const LocalOrdinal>& dofPresent, std::vector<LocalOrdinal>& map, size_t nDofs) const;
+  void assignGhostLocalNodeIds(const Teuchos::RCP<const Map>& rowDofMap, const Teuchos::RCP<const Map>& colDofMap, std::vector<LocalOrdinal>& myLocalNodeIds, const std::vector<LocalOrdinal>& dofMap, size_t maxDofPerNode, size_t& nLocalNodes, size_t& nLocalPlusGhostNodes, Teuchos::RCP<const Teuchos::Comm<int> > comm) const;
+  void squeezeOutNnzs(Teuchos::ArrayRCP<size_t>& rowPtr, Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals, const std::vector<bool>& keep) const;
+  void buildLaplacian(const Teuchos::ArrayRCP<size_t>& rowPtr, const Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals, const size_t& numdim, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& ghostedCoords) const;
 
-    template <class listType>
-    void MueLu_az_sort(listType list[], size_t N, size_t list2[], Scalar list3[]) const {
-      /* local variables */
+  template <class listType>
+  void MueLu_az_sort(listType list[], size_t N, size_t list2[], Scalar list3[]) const {
+    /* local variables */
 
-      listType RR, K;
-      size_t l, r, j, i;
-      int flag;
-      size_t RR2;
-      Scalar RR3;
+    listType RR, K;
+    size_t l, r, j, i;
+    int flag;
+    size_t RR2;
+    Scalar RR3;
 
-      /*********************** execution begins ******************************/
+    /*********************** execution begins ******************************/
 
-      if (N <= 1) return;
+    if (N <= 1) return;
 
-      l   = N / 2 + 1;
-      r   = N - 1;
-      l   = l - 1;
-      RR  = list[l - 1];
-      K   = list[l - 1];
+    l  = N / 2 + 1;
+    r  = N - 1;
+    l  = l - 1;
+    RR = list[l - 1];
+    K  = list[l - 1];
 
-      if ((list2 != NULL) && (list3 != NULL)) {
-        RR2 = list2[l - 1];
-        RR3 = list3[l - 1];
-        while (r != 0) {
-          j = l;
-          flag = 1;
+    if ((list2 != NULL) && (list3 != NULL)) {
+      RR2 = list2[l - 1];
+      RR3 = list3[l - 1];
+      while (r != 0) {
+        j    = l;
+        flag = 1;
 
-          while (flag == 1) {
-            i = j;
-            j = j + j;
+        while (flag == 1) {
+          i = j;
+          j = j + j;
 
-            if (j > r + 1)
+          if (j > r + 1)
+            flag = 0;
+          else {
+            if (j < r + 1)
+              if (list[j] > list[j - 1]) j = j + 1;
+
+            if (list[j - 1] > K) {
+              list[i - 1]  = list[j - 1];
+              list2[i - 1] = list2[j - 1];
+              list3[i - 1] = list3[j - 1];
+            } else {
               flag = 0;
-            else {
-              if (j < r + 1)
-                if (list[j] > list[j - 1]) j = j + 1;
-
-              if (list[j - 1] > K) {
-                list[ i - 1] = list[ j - 1];
-                list2[i - 1] = list2[j - 1];
-                list3[i - 1] = list3[j - 1];
-              }
-              else {
-                flag = 0;
-              }
             }
           }
-
-          list[ i - 1] = RR;
-          list2[i - 1] = RR2;
-          list3[i - 1] = RR3;
-
-          if (l == 1) {
-            RR  = list [r];
-            RR2 = list2[r];
-            RR3 = list3[r];
-
-            K = list[r];
-            list[r ] = list[0];
-            list2[r] = list2[0];
-            list3[r] = list3[0];
-            r = r - 1;
-          }
-          else {
-            l   = l - 1;
-            RR  = list[ l - 1];
-            RR2 = list2[l - 1];
-            RR3 = list3[l - 1];
-            K   = list[l - 1];
-          }
         }
 
-        list[ 0] = RR;
-        list2[0] = RR2;
-        list3[0] = RR3;
+        list[i - 1]  = RR;
+        list2[i - 1] = RR2;
+        list3[i - 1] = RR3;
+
+        if (l == 1) {
+          RR  = list[r];
+          RR2 = list2[r];
+          RR3 = list3[r];
+
+          K        = list[r];
+          list[r]  = list[0];
+          list2[r] = list2[0];
+          list3[r] = list3[0];
+          r        = r - 1;
+        } else {
+          l   = l - 1;
+          RR  = list[l - 1];
+          RR2 = list2[l - 1];
+          RR3 = list3[l - 1];
+          K   = list[l - 1];
+        }
       }
-      else if (list2 != NULL) {
-        RR2 = list2[l - 1];
-        while (r != 0) {
-          j = l;
-          flag = 1;
 
-          while (flag == 1) {
-            i = j;
-            j = j + j;
+      list[0]  = RR;
+      list2[0] = RR2;
+      list3[0] = RR3;
+    } else if (list2 != NULL) {
+      RR2 = list2[l - 1];
+      while (r != 0) {
+        j    = l;
+        flag = 1;
+
+        while (flag == 1) {
+          i = j;
+          j = j + j;
+
+          if (j > r + 1)
+            flag = 0;
+          else {
+            if (j < r + 1)
+              if (list[j] > list[j - 1]) j = j + 1;
 
-            if (j > r + 1)
+            if (list[j - 1] > K) {
+              list[i - 1]  = list[j - 1];
+              list2[i - 1] = list2[j - 1];
+            } else {
               flag = 0;
-            else {
-              if (j < r + 1)
-                if (list[j] > list[j - 1]) j = j + 1;
-
-              if (list[j - 1] > K) {
-                list[ i - 1] = list[ j - 1];
-                list2[i - 1] = list2[j - 1];
-              }
-              else {
-                flag = 0;
-              }
             }
           }
-
-          list[ i - 1] = RR;
-          list2[i - 1] = RR2;
-
-          if (l == 1) {
-            RR  = list [r];
-            RR2 = list2[r];
-
-            K = list[r];
-            list[r ] = list[0];
-            list2[r] = list2[0];
-            r = r - 1;
-          }
-          else {
-            l   = l - 1;
-            RR  = list[ l - 1];
-            RR2 = list2[l - 1];
-            K   = list[l - 1];
-          }
         }
 
-        list[ 0] = RR;
-        list2[0] = RR2;
+        list[i - 1]  = RR;
+        list2[i - 1] = RR2;
+
+        if (l == 1) {
+          RR  = list[r];
+          RR2 = list2[r];
+
+          K        = list[r];
+          list[r]  = list[0];
+          list2[r] = list2[0];
+          r        = r - 1;
+        } else {
+          l   = l - 1;
+          RR  = list[l - 1];
+          RR2 = list2[l - 1];
+          K   = list[l - 1];
+        }
       }
-      else if (list3 != NULL) {
-        RR3 = list3[l - 1];
-        while (r != 0) {
-          j = l;
-          flag = 1;
 
-          while (flag == 1) {
-            i = j;
-            j = j + j;
+      list[0]  = RR;
+      list2[0] = RR2;
+    } else if (list3 != NULL) {
+      RR3 = list3[l - 1];
+      while (r != 0) {
+        j    = l;
+        flag = 1;
+
+        while (flag == 1) {
+          i = j;
+          j = j + j;
 
-            if (j > r + 1)
+          if (j > r + 1)
+            flag = 0;
+          else {
+            if (j < r + 1)
+              if (list[j] > list[j - 1]) j = j + 1;
+
+            if (list[j - 1] > K) {
+              list[i - 1]  = list[j - 1];
+              list3[i - 1] = list3[j - 1];
+            } else {
               flag = 0;
-            else {
-              if (j < r + 1)
-                if (list[j] > list[j - 1]) j = j + 1;
-
-              if (list[j - 1] > K) {
-                list[ i - 1] = list[ j - 1];
-                list3[i - 1] = list3[j - 1];
-              }
-              else {
-                flag = 0;
-              }
             }
           }
+        }
 
-          list[ i - 1] = RR;
-          list3[i - 1] = RR3;
-
-          if (l == 1) {
-            RR  = list [r];
-            RR3 = list3[r];
-
-            K = list[r];
-            list[r ] = list[0];
-            list3[r] = list3[0];
-            r = r - 1;
-          }
-          else {
-            l   = l - 1;
-            RR  = list[ l - 1];
-            RR3 = list3[l - 1];
-            K   = list[l - 1];
-          }
+        list[i - 1]  = RR;
+        list3[i - 1] = RR3;
+
+        if (l == 1) {
+          RR  = list[r];
+          RR3 = list3[r];
+
+          K        = list[r];
+          list[r]  = list[0];
+          list3[r] = list3[0];
+          r        = r - 1;
+        } else {
+          l   = l - 1;
+          RR  = list[l - 1];
+          RR3 = list3[l - 1];
+          K   = list[l - 1];
         }
+      }
 
-        list[ 0] = RR;
-        list3[0] = RR3;
+      list[0]  = RR;
+      list3[0] = RR3;
 
-      }
-      else {
-        while (r != 0) {
-          j = l;
-          flag = 1;
+    } else {
+      while (r != 0) {
+        j    = l;
+        flag = 1;
 
-          while (flag == 1) {
-            i = j;
-            j = j + j;
+        while (flag == 1) {
+          i = j;
+          j = j + j;
 
-            if (j > r + 1)
+          if (j > r + 1)
+            flag = 0;
+          else {
+            if (j < r + 1)
+              if (list[j] > list[j - 1]) j = j + 1;
+
+            if (list[j - 1] > K) {
+              list[i - 1] = list[j - 1];
+            } else {
               flag = 0;
-            else {
-              if (j < r + 1)
-                if (list[j] > list[j - 1]) j = j + 1;
-
-              if (list[j - 1] > K) {
-                list[ i - 1] = list[ j - 1];
-              }
-              else {
-                flag = 0;
-              }
             }
           }
+        }
 
-          list[ i - 1] = RR;
+        list[i - 1] = RR;
 
-          if (l == 1) {
-            RR  = list [r];
+        if (l == 1) {
+          RR = list[r];
 
-            K = list[r];
-            list[r ] = list[0];
-            r = r - 1;
-          }
-          else {
-            l   = l - 1;
-            RR  = list[ l - 1];
-            K   = list[l - 1];
-          }
+          K       = list[r];
+          list[r] = list[0];
+          r       = r - 1;
+        } else {
+          l  = l - 1;
+          RR = list[l - 1];
+          K  = list[l - 1];
         }
-
-        list[ 0] = RR;
       }
+
+      list[0] = RR;
     }
+  }
 
-  }; //class CoalesceDropFactory
+};  // class CoalesceDropFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_VARIABLEDOFLAPLACIANFACTORY_SHORT
 
-
 #endif /* PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_def.hpp b/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_def.hpp
index 21aa1c77369d..3742e119387f 100644
--- a/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_def.hpp
+++ b/packages/muelu/src/Graph/MatrixTransformation/MueLu_VariableDofLaplacianFactory_def.hpp
@@ -47,559 +47,551 @@
 #ifndef PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DEF_HPP_
 #define PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DEF_HPP_
 
-
 #include "MueLu_Monitor.hpp"
 
 #include "MueLu_VariableDofLaplacianFactory_decl.hpp"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< double >                ("Advanced Dirichlet: threshold", 1e-5, "Drop tolerance for Dirichlet detection");
-    validParamList->set< double >                ("Variable DOF amalgamation: threshold", 1.8e-9, "Drop tolerance for amalgamation process");
-    validParamList->set< int >                   ("maxDofPerNode", 1, "Maximum number of DOFs per node");
+  validParamList->set<double>("Advanced Dirichlet: threshold", 1e-5, "Drop tolerance for Dirichlet detection");
+  validParamList->set<double>("Variable DOF amalgamation: threshold", 1.8e-9, "Drop tolerance for amalgamation process");
+  validParamList->set<int>("maxDofPerNode", 1, "Maximum number of DOFs per node");
 
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory for Coordinates");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for Coordinates");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::VariableDofLaplacianFactory() { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::VariableDofLaplacianFactory() {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "Coordinates");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "Coordinates");
 
-    //if (currentLevel.GetLevelID() == 0) // TODO check for finest level (special treatment)
-    if (currentLevel.IsAvailable("DofPresent", NoFactory::get())) {
-      currentLevel.DeclareInput("DofPresent", NoFactory::get(), this);
-    }
+  // if (currentLevel.GetLevelID() == 0) // TODO check for finest level (special treatment)
+  if (currentLevel.IsAvailable("DofPresent", NoFactory::get())) {
+    currentLevel.DeclareInput("DofPresent", NoFactory::get(), this);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    typedef Teuchos::ScalarTraits<SC> STS;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  typedef Teuchos::ScalarTraits<SC> STS;
 
-    const ParameterList  & pL = GetParameterList();
+  const ParameterList& pL = GetParameterList();
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
 
-    Teuchos::RCP< const Teuchos::Comm< int > > comm = A->getRowMap()->getComm();
-    Xpetra::UnderlyingLib lib = A->getRowMap()->lib();
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+  Xpetra::UnderlyingLib lib                    = A->getRowMap()->lib();
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> dxMV;
-    RCP<dxMV> Coords = Get< RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > >(currentLevel, "Coordinates");
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> dxMV;
+  RCP<dxMV> Coords = Get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > >(currentLevel, "Coordinates");
 
-    int maxDofPerNode = pL.get<int>("maxDofPerNode");
-    Scalar dirDropTol = Teuchos::as<Scalar>(pL.get<double>("Advanced Dirichlet: threshold")); // "ML advnaced Dirichlet: threshold"
-    Scalar amalgDropTol = Teuchos::as<Scalar>(pL.get<double>("Variable DOF amalgamation: threshold")); //"variable DOF amalgamation: threshold")
+  int maxDofPerNode   = pL.get<int>("maxDofPerNode");
+  Scalar dirDropTol   = Teuchos::as<Scalar>(pL.get<double>("Advanced Dirichlet: threshold"));         // "ML advnaced Dirichlet: threshold"
+  Scalar amalgDropTol = Teuchos::as<Scalar>(pL.get<double>("Variable DOF amalgamation: threshold"));  //"variable DOF amalgamation: threshold")
 
-    bool bHasZeroDiagonal = false;
-    Teuchos::ArrayRCP<const bool> dirOrNot = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DetectDirichletRowsExt(*A,bHasZeroDiagonal,STS::magnitude(dirDropTol));
+  bool bHasZeroDiagonal                  = false;
+  Teuchos::ArrayRCP<const bool> dirOrNot = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DetectDirichletRowsExt(*A, bHasZeroDiagonal, STS::magnitude(dirDropTol));
 
-    // check availability of DofPresent array
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent;
-    if (currentLevel.IsAvailable("DofPresent", NoFactory::get())) {
-      dofPresent = currentLevel.Get< Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", NoFactory::get());
-    } else {
-      // TAW: not sure about size of array. We cannot determine the expected size in the non-padded case correctly...
-      dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(A->getRowMap()->getLocalNumElements(),1);
-    }
+  // check availability of DofPresent array
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent;
+  if (currentLevel.IsAvailable("DofPresent", NoFactory::get())) {
+    dofPresent = currentLevel.Get<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", NoFactory::get());
+  } else {
+    // TAW: not sure about size of array. We cannot determine the expected size in the non-padded case correctly...
+    dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(A->getRowMap()->getLocalNumElements(), 1);
+  }
 
-    // map[k] indicates that the kth dof in the variable dof matrix A would
-    // correspond to the map[k]th dof in the padded system. If, i.e., it is
-    // map[35] = 39 then dof no 35 in the variable dof matrix A corresponds to
-    // row map id 39 in an imaginary padded matrix Apadded.
-    // The padded system is never built but would be the associated matrix if
-    // every node had maxDofPerNode dofs.
-    std::vector<LocalOrdinal> map(A->getLocalNumRows());
-    this->buildPaddedMap(dofPresent, map, A->getLocalNumRows());
+  // map[k] indicates that the kth dof in the variable dof matrix A would
+  // correspond to the map[k]th dof in the padded system. If, i.e., it is
+  // map[35] = 39 then dof no 35 in the variable dof matrix A corresponds to
+  // row map id 39 in an imaginary padded matrix Apadded.
+  // The padded system is never built but would be the associated matrix if
+  // every node had maxDofPerNode dofs.
+  std::vector<LocalOrdinal> map(A->getLocalNumRows());
+  this->buildPaddedMap(dofPresent, map, A->getLocalNumRows());
 
-    // map of size of number of DOFs containing local node id (dof id -> node id, inclusive ghosted dofs/nodes)
-    std::vector<LocalOrdinal> myLocalNodeIds(A->getColMap()->getLocalNumElements()); // possible maximum (we need the ghost nodes, too)
+  // map of size of number of DOFs containing local node id (dof id -> node id, inclusive ghosted dofs/nodes)
+  std::vector<LocalOrdinal> myLocalNodeIds(A->getColMap()->getLocalNumElements());  // possible maximum (we need the ghost nodes, too)
 
-    // assign the local node ids for the ghosted nodes
-    size_t nLocalNodes, nLocalPlusGhostNodes;
-    this->assignGhostLocalNodeIds(A->getRowMap(), A->getColMap(), myLocalNodeIds, map, maxDofPerNode, nLocalNodes, nLocalPlusGhostNodes, comm);
+  // assign the local node ids for the ghosted nodes
+  size_t nLocalNodes, nLocalPlusGhostNodes;
+  this->assignGhostLocalNodeIds(A->getRowMap(), A->getColMap(), myLocalNodeIds, map, maxDofPerNode, nLocalNodes, nLocalPlusGhostNodes, comm);
 
-    //RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout)," ",0,false,10,false, true);
+  // RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout)," ",0,false,10,false, true);
 
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(dofPresent.size()) != Teuchos::as<size_t>(nLocalNodes * maxDofPerNode),MueLu::Exceptions::RuntimeError,"VariableDofLaplacianFactory: size of provided DofPresent array is " << dofPresent.size() << " but should be " << nLocalNodes * maxDofPerNode << " on the current processor.");
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(dofPresent.size()) != Teuchos::as<size_t>(nLocalNodes * maxDofPerNode), MueLu::Exceptions::RuntimeError, "VariableDofLaplacianFactory: size of provided DofPresent array is " << dofPresent.size() << " but should be " << nLocalNodes * maxDofPerNode << " on the current processor.");
 
-    // put content of assignGhostLocalNodeIds here...
+  // put content of assignGhostLocalNodeIds here...
 
-    // fill nodal maps
+  // fill nodal maps
 
-    Teuchos::ArrayView< const GlobalOrdinal > myGids = A->getColMap()->getLocalElementList();
+  Teuchos::ArrayView<const GlobalOrdinal> myGids = A->getColMap()->getLocalElementList();
 
-    // vector containing row/col gids of amalgamated matrix (with holes)
+  // vector containing row/col gids of amalgamated matrix (with holes)
 
-    size_t nLocalDofs = A->getRowMap()->getLocalNumElements();
-    size_t nLocalPlusGhostDofs = A->getColMap()->getLocalNumElements();
+  size_t nLocalDofs          = A->getRowMap()->getLocalNumElements();
+  size_t nLocalPlusGhostDofs = A->getColMap()->getLocalNumElements();
 
-    // myLocalNodeIds (dof -> node)
+  // myLocalNodeIds (dof -> node)
 
-    Teuchos::Array<GlobalOrdinal> amalgRowMapGIDs(nLocalNodes);
-    Teuchos::Array<GlobalOrdinal> amalgColMapGIDs(nLocalPlusGhostNodes);
+  Teuchos::Array<GlobalOrdinal> amalgRowMapGIDs(nLocalNodes);
+  Teuchos::Array<GlobalOrdinal> amalgColMapGIDs(nLocalPlusGhostNodes);
 
-    // initialize
-    size_t count = 0;
-    if (nLocalDofs > 0) {
-      amalgRowMapGIDs[count] = myGids[0];
-      amalgColMapGIDs[count] = myGids[0];
+  // initialize
+  size_t count = 0;
+  if (nLocalDofs > 0) {
+    amalgRowMapGIDs[count] = myGids[0];
+    amalgColMapGIDs[count] = myGids[0];
+    count++;
+  }
+
+  for (size_t i = 1; i < nLocalDofs; i++) {
+    if (myLocalNodeIds[i] != myLocalNodeIds[i - 1]) {
+      amalgRowMapGIDs[count] = myGids[i];
+      amalgColMapGIDs[count] = myGids[i];
       count++;
     }
+  }
 
-    for(size_t i = 1; i < nLocalDofs; i++) {
-       if (myLocalNodeIds[i] != myLocalNodeIds[i-1]) {
-          amalgRowMapGIDs[count] = myGids[i];
-          amalgColMapGIDs[count] = myGids[i];
-          count++;
-       }
-    }
+  RCP<GOVector> tempAmalgColVec = GOVectorFactory::Build(A->getDomainMap());
+  {
+    Teuchos::ArrayRCP<GlobalOrdinal> tempAmalgColVecData = tempAmalgColVec->getDataNonConst(0);
+    for (size_t i = 0; i < A->getDomainMap()->getLocalNumElements(); i++)
+      tempAmalgColVecData[i] = amalgColMapGIDs[myLocalNodeIds[i]];
+  }
 
-    RCP<GOVector> tempAmalgColVec = GOVectorFactory::Build(A->getDomainMap());
-    {
-      Teuchos::ArrayRCP<GlobalOrdinal> tempAmalgColVecData = tempAmalgColVec->getDataNonConst(0);
-      for (size_t i = 0; i < A->getDomainMap()->getLocalNumElements(); i++)
-	tempAmalgColVecData[i] = amalgColMapGIDs[ myLocalNodeIds[i]];
-    }
+  RCP<GOVector> tempAmalgColVecTarget = GOVectorFactory::Build(A->getColMap());
+  Teuchos::RCP<Import> dofImporter    = ImportFactory::Build(A->getDomainMap(), A->getColMap());
+  tempAmalgColVecTarget->doImport(*tempAmalgColVec, *dofImporter, Xpetra::INSERT);
 
-    RCP<GOVector> tempAmalgColVecTarget = GOVectorFactory::Build(A->getColMap());
-    Teuchos::RCP<Import> dofImporter = ImportFactory::Build(A->getDomainMap(), A->getColMap());
-    tempAmalgColVecTarget->doImport(*tempAmalgColVec, *dofImporter, Xpetra::INSERT);
+  {
+    Teuchos::ArrayRCP<const GlobalOrdinal> tempAmalgColVecBData = tempAmalgColVecTarget->getData(0);
+    // copy from dof vector to nodal vector
+    for (size_t i = 0; i < myLocalNodeIds.size(); i++)
+      amalgColMapGIDs[myLocalNodeIds[i]] = tempAmalgColVecBData[i];
+  }
 
-    {
-      Teuchos::ArrayRCP<const GlobalOrdinal> tempAmalgColVecBData = tempAmalgColVecTarget->getData(0);
-      // copy from dof vector to nodal vector
-      for (size_t i = 0; i < myLocalNodeIds.size(); i++)
-	amalgColMapGIDs[ myLocalNodeIds[i]] = tempAmalgColVecBData[i];
+  Teuchos::RCP<Map> amalgRowMap = MapFactory::Build(lib,
+                                                    Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),
+                                                    amalgRowMapGIDs(),  // View,
+                                                    A->getRowMap()->getIndexBase(),
+                                                    comm);
+
+  Teuchos::RCP<Map> amalgColMap = MapFactory::Build(lib,
+                                                    Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),
+                                                    amalgColMapGIDs(),  // View,
+                                                    A->getRangeMap()->getIndexBase(),
+                                                    comm);
+
+  // end fill nodal maps
+
+  // start variable dof amalgamation
+
+  Teuchos::RCP<CrsMatrixWrap> Awrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A);
+  Teuchos::RCP<CrsMatrix> Acrs      = Awrap->getCrsMatrix();
+  // Acrs->describe(*fancy, Teuchos::VERB_EXTREME);
+
+  size_t nNonZeros = 0;
+  std::vector<bool> isNonZero(nLocalPlusGhostDofs, false);
+  std::vector<size_t> nonZeroList(nLocalPlusGhostDofs);  // ???
+
+  // also used in DetectDirichletExt
+  Teuchos::RCP<Vector> diagVecUnique = VectorFactory::Build(A->getRowMap());
+  Teuchos::RCP<Vector> diagVec       = VectorFactory::Build(A->getColMap());
+  A->getLocalDiagCopy(*diagVecUnique);
+  diagVec->doImport(*diagVecUnique, *dofImporter, Xpetra::INSERT);
+  Teuchos::ArrayRCP<const Scalar> diagVecData = diagVec->getData(0);
+
+  Teuchos::ArrayRCP<const size_t> rowptr(Acrs->getLocalNumRows());
+  Teuchos::ArrayRCP<const LocalOrdinal> colind(Acrs->getLocalNumEntries());
+  Teuchos::ArrayRCP<const Scalar> values(Acrs->getLocalNumEntries());
+  Acrs->getAllValues(rowptr, colind, values);
+
+  // create arrays for amalgamated matrix
+  Teuchos::ArrayRCP<size_t> amalgRowPtr(nLocalNodes + 1);
+  Teuchos::ArrayRCP<LocalOrdinal> amalgCols(rowptr[rowptr.size() - 1]);
+
+  LocalOrdinal oldBlockRow = 0;
+  LocalOrdinal blockRow    = 0;
+  LocalOrdinal blockColumn = 0;
+
+  size_t newNzs  = 0;
+  amalgRowPtr[0] = newNzs;
+
+  bool doNotDrop = false;
+  if (amalgDropTol == Teuchos::ScalarTraits<Scalar>::zero()) doNotDrop = true;
+  if (values.size() == 0) doNotDrop = true;
+
+  for (decltype(rowptr.size()) i = 0; i < rowptr.size() - 1; i++) {
+    blockRow = std::floor<LocalOrdinal>(map[i] / maxDofPerNode);
+    if (blockRow != oldBlockRow) {
+      // zero out info recording nonzeros in oldBlockRow
+      for (size_t j = 0; j < nNonZeros; j++) isNonZero[nonZeroList[j]] = false;
+      nNonZeros             = 0;
+      amalgRowPtr[blockRow] = newNzs;  // record start of next row
     }
-
-    Teuchos::RCP<Map> amalgRowMap = MapFactory::Build(lib,
-               Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),
-               amalgRowMapGIDs(), //View,
-               A->getRowMap()->getIndexBase(),
-               comm);
-
-    Teuchos::RCP<Map> amalgColMap = MapFactory::Build(lib,
-               Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),
-               amalgColMapGIDs(), //View,
-               A->getRangeMap()->getIndexBase(),
-               comm);
-
-   // end fill nodal maps
-
-
-    // start variable dof amalgamation
-
-    Teuchos::RCP<CrsMatrixWrap> Awrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A);
-    Teuchos::RCP<CrsMatrix> Acrs = Awrap->getCrsMatrix();
-    //Acrs->describe(*fancy, Teuchos::VERB_EXTREME);
-
-    size_t nNonZeros = 0;
-    std::vector<bool> isNonZero(nLocalPlusGhostDofs,false);
-    std::vector<size_t> nonZeroList(nLocalPlusGhostDofs);  // ???
-
-    // also used in DetectDirichletExt
-    Teuchos::RCP<Vector> diagVecUnique = VectorFactory::Build(A->getRowMap());
-    Teuchos::RCP<Vector> diagVec       = VectorFactory::Build(A->getColMap());
-    A->getLocalDiagCopy(*diagVecUnique);
-    diagVec->doImport(*diagVecUnique, *dofImporter, Xpetra::INSERT);
-    Teuchos::ArrayRCP< const Scalar > diagVecData = diagVec->getData(0);
-
-    Teuchos::ArrayRCP<const size_t> rowptr(Acrs->getLocalNumRows());
-    Teuchos::ArrayRCP<const LocalOrdinal> colind(Acrs->getLocalNumEntries());
-    Teuchos::ArrayRCP<const Scalar> values(Acrs->getLocalNumEntries());
-    Acrs->getAllValues(rowptr, colind, values);
-
-
-    // create arrays for amalgamated matrix
-    Teuchos::ArrayRCP<size_t> amalgRowPtr(nLocalNodes+1);
-    Teuchos::ArrayRCP<LocalOrdinal> amalgCols(rowptr[rowptr.size()-1]);
-
-    LocalOrdinal oldBlockRow = 0;
-    LocalOrdinal blockRow = 0;
-    LocalOrdinal blockColumn = 0;
-
-    size_t newNzs = 0;
-    amalgRowPtr[0] = newNzs;
-
-    bool doNotDrop = false;
-    if (amalgDropTol == Teuchos::ScalarTraits<Scalar>::zero()) doNotDrop = true;
-    if (values.size() == 0) doNotDrop = true;
-
-    for(decltype(rowptr.size()) i = 0; i < rowptr.size()-1; i++) {
-      blockRow = std::floor<LocalOrdinal>( map[i] / maxDofPerNode);
-      if (blockRow != oldBlockRow) {
-        // zero out info recording nonzeros in oldBlockRow
-        for(size_t j = 0; j < nNonZeros; j++) isNonZero[nonZeroList[j]] = false;
-        nNonZeros = 0;
-        amalgRowPtr[blockRow] = newNzs; // record start of next row
-      }
-      for (size_t j = rowptr[i]; j < rowptr[i+1]; j++) {
-        if(doNotDrop == true ||
-           ( STS::magnitude(values[j] / STS::magnitude(sqrt(STS::magnitude(diagVecData[i]) * STS::magnitude(diagVecData[colind[j]])))   ) >= STS::magnitude(amalgDropTol) )) {
-          blockColumn = myLocalNodeIds[colind[j]];
-          if(isNonZero[blockColumn] == false) {
-            isNonZero[blockColumn] = true;
-            nonZeroList[nNonZeros++] = blockColumn;
-            amalgCols[newNzs++] = blockColumn;
-          }
+    for (size_t j = rowptr[i]; j < rowptr[i + 1]; j++) {
+      if (doNotDrop == true ||
+          (STS::magnitude(values[j] / STS::magnitude(sqrt(STS::magnitude(diagVecData[i]) * STS::magnitude(diagVecData[colind[j]])))) >= STS::magnitude(amalgDropTol))) {
+        blockColumn = myLocalNodeIds[colind[j]];
+        if (isNonZero[blockColumn] == false) {
+          isNonZero[blockColumn]   = true;
+          nonZeroList[nNonZeros++] = blockColumn;
+          amalgCols[newNzs++]      = blockColumn;
         }
       }
-      oldBlockRow = blockRow;
     }
-    amalgRowPtr[blockRow+1] = newNzs;
-
-    TEUCHOS_TEST_FOR_EXCEPTION((blockRow+1 != Teuchos::as<LO>(nLocalNodes)) && (nLocalNodes !=0), MueLu::Exceptions::RuntimeError, "VariableDofsPerNodeAmalgamation: error, computed # block rows (" << blockRow+1 <<") != nLocalNodes (" << nLocalNodes <<")");
-
-    amalgCols.resize(amalgRowPtr[nLocalNodes]);
-
-    // end variableDofAmalg
-
-    // begin rm differentDofsCrossings
-
-    // Remove matrix entries (i,j) where the ith node and the jth node have
-    // different dofs that are 'present'
-    // Specifically, on input:
-    //    dofPresent[i*maxDofPerNode+k] indicates whether or not the kth
-    //                                  dof at the ith node is present in the
-    //                                  variable dof matrix (e.g., the ith node
-    //                                  has an air pressure dof). true means
-    //                                  the dof is present while false means it
-    //                                  is not.
-    // We create a unique id for the ith node (i.e. uniqueId[i]) via
-    //    sum_{k=0 to maxDofPerNode-1} dofPresent[i*maxDofPerNode+k]*2^k
-    // and use this unique idea to remove entries (i,j) when uniqueId[i]!=uniqueId[j]
-
-    Teuchos::ArrayRCP<LocalOrdinal> uniqueId(nLocalPlusGhostNodes);       // unique id associated with DOF
-    std::vector<bool> keep(amalgRowPtr[amalgRowPtr.size()-1],true); // keep connection associated with node
-
-    size_t ii = 0; // iteration index for present dofs
-    for(decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size()-1; i++) {
-      LocalOrdinal temp = 1; // basis for dof-id
-      uniqueId[i] = 0;
-      for (decltype(maxDofPerNode) j = 0; j < maxDofPerNode; j++) {
-        if (dofPresent[ii++]) uniqueId[i] += temp; // encode dof to be present
-        temp = temp * 2; // check next dof
-      }
+    oldBlockRow = blockRow;
+  }
+  amalgRowPtr[blockRow + 1] = newNzs;
+
+  TEUCHOS_TEST_FOR_EXCEPTION((blockRow + 1 != Teuchos::as<LO>(nLocalNodes)) && (nLocalNodes != 0), MueLu::Exceptions::RuntimeError, "VariableDofsPerNodeAmalgamation: error, computed # block rows (" << blockRow + 1 << ") != nLocalNodes (" << nLocalNodes << ")");
+
+  amalgCols.resize(amalgRowPtr[nLocalNodes]);
+
+  // end variableDofAmalg
+
+  // begin rm differentDofsCrossings
+
+  // Remove matrix entries (i,j) where the ith node and the jth node have
+  // different dofs that are 'present'
+  // Specifically, on input:
+  //    dofPresent[i*maxDofPerNode+k] indicates whether or not the kth
+  //                                  dof at the ith node is present in the
+  //                                  variable dof matrix (e.g., the ith node
+  //                                  has an air pressure dof). true means
+  //                                  the dof is present while false means it
+  //                                  is not.
+  // We create a unique id for the ith node (i.e. uniqueId[i]) via
+  //    sum_{k=0 to maxDofPerNode-1} dofPresent[i*maxDofPerNode+k]*2^k
+  // and use this unique idea to remove entries (i,j) when uniqueId[i]!=uniqueId[j]
+
+  Teuchos::ArrayRCP<LocalOrdinal> uniqueId(nLocalPlusGhostNodes);     // unique id associated with DOF
+  std::vector<bool> keep(amalgRowPtr[amalgRowPtr.size() - 1], true);  // keep connection associated with node
+
+  size_t ii = 0;  // iteration index for present dofs
+  for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+    LocalOrdinal temp = 1;  // basis for dof-id
+    uniqueId[i]       = 0;
+    for (decltype(maxDofPerNode) j = 0; j < maxDofPerNode; j++) {
+      if (dofPresent[ii++]) uniqueId[i] += temp;  // encode dof to be present
+      temp = temp * 2;                            // check next dof
     }
+  }
 
-    Teuchos::RCP<Import> nodeImporter = ImportFactory::Build(amalgRowMap, amalgColMap);
+  Teuchos::RCP<Import> nodeImporter = ImportFactory::Build(amalgRowMap, amalgColMap);
 
-    RCP<LOVector> nodeIdSrc    = Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(amalgRowMap,true);
-    RCP<LOVector> nodeIdTarget = Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(amalgColMap,true);
+  RCP<LOVector> nodeIdSrc    = Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(amalgRowMap, true);
+  RCP<LOVector> nodeIdTarget = Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(amalgColMap, true);
 
-    Teuchos::ArrayRCP< LocalOrdinal > nodeIdSrcData = nodeIdSrc->getDataNonConst(0);
-    for(decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size()-1; i++) {
-      nodeIdSrcData[i] = uniqueId[i];
-    }
+  Teuchos::ArrayRCP<LocalOrdinal> nodeIdSrcData = nodeIdSrc->getDataNonConst(0);
+  for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+    nodeIdSrcData[i] = uniqueId[i];
+  }
 
-    nodeIdTarget->doImport(*nodeIdSrc, *nodeImporter, Xpetra::INSERT);
+  nodeIdTarget->doImport(*nodeIdSrc, *nodeImporter, Xpetra::INSERT);
 
-    Teuchos::ArrayRCP< const LocalOrdinal > nodeIdTargetData = nodeIdTarget->getData(0);
-    for(decltype(uniqueId.size()) i = 0; i < uniqueId.size(); i++) {
-      uniqueId[i] = nodeIdTargetData[i];
-    }
+  Teuchos::ArrayRCP<const LocalOrdinal> nodeIdTargetData = nodeIdTarget->getData(0);
+  for (decltype(uniqueId.size()) i = 0; i < uniqueId.size(); i++) {
+    uniqueId[i] = nodeIdTargetData[i];
+  }
 
-    // nodal comm uniqueId, myLocalNodeIds
+  // nodal comm uniqueId, myLocalNodeIds
 
-    // uniqueId now should contain ghosted data
+  // uniqueId now should contain ghosted data
 
-    for(decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size()-1; i++) {
-      for(size_t j = amalgRowPtr[i]; j < amalgRowPtr[i+1]; j++) {
-        if (uniqueId[i] != uniqueId[amalgCols[j]]) keep [j] = false;
-      }
+  for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+    for (size_t j = amalgRowPtr[i]; j < amalgRowPtr[i + 1]; j++) {
+      if (uniqueId[i] != uniqueId[amalgCols[j]]) keep[j] = false;
     }
+  }
 
-    // squeeze out hard-coded zeros from CSR arrays
-    Teuchos::ArrayRCP<Scalar> amalgVals;
-    this->squeezeOutNnzs(amalgRowPtr,amalgCols,amalgVals,keep);
+  // squeeze out hard-coded zeros from CSR arrays
+  Teuchos::ArrayRCP<Scalar> amalgVals;
+  this->squeezeOutNnzs(amalgRowPtr, amalgCols, amalgVals, keep);
 
-    typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> dxMVf;
-    RCP<dxMV> ghostedCoords = dxMVf::Build(amalgColMap,Coords->getNumVectors());
+  typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> dxMVf;
+  RCP<dxMV> ghostedCoords = dxMVf::Build(amalgColMap, Coords->getNumVectors());
 
-    TEUCHOS_TEST_FOR_EXCEPTION(amalgRowMap->getLocalNumElements() != Coords->getMap()->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "MueLu::VariableDofLaplacianFactory: the number of Coordinates and amalgamated nodes is inconsistent.");
+  TEUCHOS_TEST_FOR_EXCEPTION(amalgRowMap->getLocalNumElements() != Coords->getMap()->getLocalNumElements(), MueLu::Exceptions::RuntimeError, "MueLu::VariableDofLaplacianFactory: the number of Coordinates and amalgamated nodes is inconsistent.");
 
-    // Coords might live on a special nodeMap with consecutive ids (the natural numbering)
-    // The amalgRowMap might have the same number of entries, but with holes in the ids.
-    // e.g. 0,3,6,9,... as GIDs.
-    // We need the ghosted Coordinates in the buildLaplacian routine. But we access the data
-    // through getData only, i.e., the global ids are not interesting as long as we do not change
-    // the ordering of the entries
-    Coords->replaceMap(amalgRowMap);
-    ghostedCoords->doImport(*Coords, *nodeImporter, Xpetra::INSERT);
+  // Coords might live on a special nodeMap with consecutive ids (the natural numbering)
+  // The amalgRowMap might have the same number of entries, but with holes in the ids.
+  // e.g. 0,3,6,9,... as GIDs.
+  // We need the ghosted Coordinates in the buildLaplacian routine. But we access the data
+  // through getData only, i.e., the global ids are not interesting as long as we do not change
+  // the ordering of the entries
+  Coords->replaceMap(amalgRowMap);
+  ghostedCoords->doImport(*Coords, *nodeImporter, Xpetra::INSERT);
 
-    Teuchos::ArrayRCP<Scalar> lapVals(amalgRowPtr[nLocalNodes]);
-    this->buildLaplacian(amalgRowPtr, amalgCols, lapVals, Coords->getNumVectors(), ghostedCoords);
+  Teuchos::ArrayRCP<Scalar> lapVals(amalgRowPtr[nLocalNodes]);
+  this->buildLaplacian(amalgRowPtr, amalgCols, lapVals, Coords->getNumVectors(), ghostedCoords);
 
-    // sort column GIDs
-    for(decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size()-1; i++) {
-      size_t j = amalgRowPtr[i];
-      this->MueLu_az_sort<LocalOrdinal>(&(amalgCols[j]), amalgRowPtr[i+1] - j, NULL, &(lapVals[j]));
-    }
+  // sort column GIDs
+  for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+    size_t j = amalgRowPtr[i];
+    this->MueLu_az_sort<LocalOrdinal>(&(amalgCols[j]), amalgRowPtr[i + 1] - j, NULL, &(lapVals[j]));
+  }
 
-    // Caluclate status array for next level
-    Teuchos::Array<char> status(nLocalNodes * maxDofPerNode);
+  // Caluclate status array for next level
+  Teuchos::Array<char> status(nLocalNodes * maxDofPerNode);
 
-    // dir or not Teuchos::ArrayRCP<const bool> dirOrNot
-    for(decltype(status.size()) i = 0; i < status.size(); i++) status[i] = 's';
-    for(decltype(status.size()) i = 0; i < status.size(); i++) {
-      if(dofPresent[i] == false) status[i] = 'p';
-    }
-    if(dirOrNot.size() > 0) {
-      for(decltype(map.size()) i = 0; i < map.size(); i++) {
-        if(dirOrNot[i] == true){
-          status[map[i]] = 'd';
-        }
+  // dir or not Teuchos::ArrayRCP<const bool> dirOrNot
+  for (decltype(status.size()) i = 0; i < status.size(); i++) status[i] = 's';
+  for (decltype(status.size()) i = 0; i < status.size(); i++) {
+    if (dofPresent[i] == false) status[i] = 'p';
+  }
+  if (dirOrNot.size() > 0) {
+    for (decltype(map.size()) i = 0; i < map.size(); i++) {
+      if (dirOrNot[i] == true) {
+        status[map[i]] = 'd';
       }
     }
-    Set(currentLevel,"DofStatus",status);
-
-    // end status array
-
-    Teuchos::RCP<CrsMatrix> lapCrsMat = CrsMatrixFactory::Build(amalgRowMap, amalgColMap, 10); // TODO better approx for max nnz per row
+  }
+  Set(currentLevel, "DofStatus", status);
 
-    for (size_t i = 0; i < nLocalNodes; i++) {
-      lapCrsMat->insertLocalValues(i, amalgCols.view(amalgRowPtr[i],amalgRowPtr[i+1]-amalgRowPtr[i]),
-          lapVals.view(amalgRowPtr[i],amalgRowPtr[i+1]-amalgRowPtr[i]));
-    }
-    lapCrsMat->fillComplete(amalgRowMap,amalgRowMap);
+  // end status array
 
-    //lapCrsMat->describe(*fancy, Teuchos::VERB_EXTREME);
+  Teuchos::RCP<CrsMatrix> lapCrsMat = CrsMatrixFactory::Build(amalgRowMap, amalgColMap, 10);  // TODO better approx for max nnz per row
 
-    Teuchos::RCP<Matrix> lapMat = Teuchos::rcp(new CrsMatrixWrap(lapCrsMat));
-    Set(currentLevel,"A",lapMat);
+  for (size_t i = 0; i < nLocalNodes; i++) {
+    lapCrsMat->insertLocalValues(i, amalgCols.view(amalgRowPtr[i], amalgRowPtr[i + 1] - amalgRowPtr[i]),
+                                 lapVals.view(amalgRowPtr[i], amalgRowPtr[i + 1] - amalgRowPtr[i]));
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildLaplacian(const Teuchos::ArrayRCP<size_t>& rowPtr, const Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals,const size_t& numdim, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > & ghostedCoords) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(numdim != 2 && numdim !=3, MueLu::Exceptions::RuntimeError,"buildLaplacian only works for 2d or 3d examples. numdim = " << numdim);
-
-    if(numdim == 2) { // 2d
-      Teuchos::ArrayRCP< const typename Teuchos::ScalarTraits<Scalar>::magnitudeType > x = ghostedCoords->getData(0);
-      Teuchos::ArrayRCP< const typename Teuchos::ScalarTraits<Scalar>::magnitudeType > y = ghostedCoords->getData(1);
-
-      for(decltype(rowPtr.size()) i = 0; i < rowPtr.size() - 1; i++) {
-        Scalar sum = Teuchos::ScalarTraits<Scalar>::zero();
-        LocalOrdinal diag = -1;
-        for(size_t j = rowPtr[i]; j < rowPtr[i+1]; j++) {
-          if(cols[j] != Teuchos::as<LO>(i)){
-            vals[j] = std::sqrt( (x[i]-x[cols[j]]) * (x[i]-x[cols[j]]) +
-                                 (y[i]-y[cols[j]]) * (y[i]-y[cols[j]]) );
-            TEUCHOS_TEST_FOR_EXCEPTION(vals[j] == Teuchos::ScalarTraits<Scalar>::zero(), MueLu::Exceptions::RuntimeError, "buildLaplacian: error, " << i << " and " << cols[j] << " have same coordinates: " << x[i] << " and " << y[i]);
-            vals[j] = -Teuchos::ScalarTraits<SC>::one()/vals[j];
-            sum = sum - vals[j];
-          }
-          else diag = j;
-        }
-        if(sum == Teuchos::ScalarTraits<Scalar>::zero()) sum = Teuchos::ScalarTraits<Scalar>::one();
-        TEUCHOS_TEST_FOR_EXCEPTION(diag == -1, MueLu::Exceptions::RuntimeError, "buildLaplacian: error, row " << i << " has zero diagonal!");
-
-        vals[diag] = sum;
+  lapCrsMat->fillComplete(amalgRowMap, amalgRowMap);
+
+  // lapCrsMat->describe(*fancy, Teuchos::VERB_EXTREME);
+
+  Teuchos::RCP<Matrix> lapMat = Teuchos::rcp(new CrsMatrixWrap(lapCrsMat));
+  Set(currentLevel, "A", lapMat);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildLaplacian(const Teuchos::ArrayRCP<size_t>& rowPtr, const Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals, const size_t& numdim, const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& ghostedCoords) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(numdim != 2 && numdim != 3, MueLu::Exceptions::RuntimeError, "buildLaplacian only works for 2d or 3d examples. numdim = " << numdim);
+
+  if (numdim == 2) {  // 2d
+    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> x = ghostedCoords->getData(0);
+    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> y = ghostedCoords->getData(1);
+
+    for (decltype(rowPtr.size()) i = 0; i < rowPtr.size() - 1; i++) {
+      Scalar sum        = Teuchos::ScalarTraits<Scalar>::zero();
+      LocalOrdinal diag = -1;
+      for (size_t j = rowPtr[i]; j < rowPtr[i + 1]; j++) {
+        if (cols[j] != Teuchos::as<LO>(i)) {
+          vals[j] = std::sqrt((x[i] - x[cols[j]]) * (x[i] - x[cols[j]]) +
+                              (y[i] - y[cols[j]]) * (y[i] - y[cols[j]]));
+          TEUCHOS_TEST_FOR_EXCEPTION(vals[j] == Teuchos::ScalarTraits<Scalar>::zero(), MueLu::Exceptions::RuntimeError, "buildLaplacian: error, " << i << " and " << cols[j] << " have same coordinates: " << x[i] << " and " << y[i]);
+          vals[j] = -Teuchos::ScalarTraits<SC>::one() / vals[j];
+          sum     = sum - vals[j];
+        } else
+          diag = j;
       }
-    } else { // 3d
-      Teuchos::ArrayRCP< const typename Teuchos::ScalarTraits<Scalar>::magnitudeType > x = ghostedCoords->getData(0);
-      Teuchos::ArrayRCP< const typename Teuchos::ScalarTraits<Scalar>::magnitudeType > y = ghostedCoords->getData(1);
-      Teuchos::ArrayRCP< const typename Teuchos::ScalarTraits<Scalar>::magnitudeType > z = ghostedCoords->getData(2);
-
-      for(decltype(rowPtr.size()) i = 0; i < rowPtr.size() - 1; i++) {
-        Scalar sum = Teuchos::ScalarTraits<Scalar>::zero();
-        LocalOrdinal diag = -1;
-        for(size_t j = rowPtr[i]; j < rowPtr[i+1]; j++) {
-          if(cols[j] != Teuchos::as<LO>(i)){
-            vals[j] = std::sqrt( (x[i]-x[cols[j]]) * (x[i]-x[cols[j]]) +
-                                 (y[i]-y[cols[j]]) * (y[i]-y[cols[j]]) +
-                                 (z[i]-z[cols[j]]) * (z[i]-z[cols[j]]) );
-
-            TEUCHOS_TEST_FOR_EXCEPTION(vals[j] == Teuchos::ScalarTraits<Scalar>::zero(), MueLu::Exceptions::RuntimeError, "buildLaplacian: error, " << i << " and " << cols[j] << " have same coordinates: " << x[i] << " and " << y[i] << " and " << z[i]);
-
-            vals[j] = -Teuchos::ScalarTraits<SC>::one()/vals[j];
-            sum = sum - vals[j];
-          }
-          else diag = j;
-        }
-        if(sum == Teuchos::ScalarTraits<Scalar>::zero()) sum = Teuchos::ScalarTraits<Scalar>::one();
-        TEUCHOS_TEST_FOR_EXCEPTION(diag == -1, MueLu::Exceptions::RuntimeError, "buildLaplacian: error, row " << i << " has zero diagonal!");
+      if (sum == Teuchos::ScalarTraits<Scalar>::zero()) sum = Teuchos::ScalarTraits<Scalar>::one();
+      TEUCHOS_TEST_FOR_EXCEPTION(diag == -1, MueLu::Exceptions::RuntimeError, "buildLaplacian: error, row " << i << " has zero diagonal!");
 
-        vals[diag] = sum;
+      vals[diag] = sum;
+    }
+  } else {  // 3d
+    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> x = ghostedCoords->getData(0);
+    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> y = ghostedCoords->getData(1);
+    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> z = ghostedCoords->getData(2);
+
+    for (decltype(rowPtr.size()) i = 0; i < rowPtr.size() - 1; i++) {
+      Scalar sum        = Teuchos::ScalarTraits<Scalar>::zero();
+      LocalOrdinal diag = -1;
+      for (size_t j = rowPtr[i]; j < rowPtr[i + 1]; j++) {
+        if (cols[j] != Teuchos::as<LO>(i)) {
+          vals[j] = std::sqrt((x[i] - x[cols[j]]) * (x[i] - x[cols[j]]) +
+                              (y[i] - y[cols[j]]) * (y[i] - y[cols[j]]) +
+                              (z[i] - z[cols[j]]) * (z[i] - z[cols[j]]));
+
+          TEUCHOS_TEST_FOR_EXCEPTION(vals[j] == Teuchos::ScalarTraits<Scalar>::zero(), MueLu::Exceptions::RuntimeError, "buildLaplacian: error, " << i << " and " << cols[j] << " have same coordinates: " << x[i] << " and " << y[i] << " and " << z[i]);
+
+          vals[j] = -Teuchos::ScalarTraits<SC>::one() / vals[j];
+          sum     = sum - vals[j];
+        } else
+          diag = j;
       }
+      if (sum == Teuchos::ScalarTraits<Scalar>::zero()) sum = Teuchos::ScalarTraits<Scalar>::one();
+      TEUCHOS_TEST_FOR_EXCEPTION(diag == -1, MueLu::Exceptions::RuntimeError, "buildLaplacian: error, row " << i << " has zero diagonal!");
+
+      vals[diag] = sum;
     }
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::squeezeOutNnzs(Teuchos::ArrayRCP<size_t> & rowPtr, Teuchos::ArrayRCP<LocalOrdinal> & cols, Teuchos::ArrayRCP<Scalar> & vals, const std::vector<bool>& keep) const {
-    // get rid of nonzero entries that have 0's in them and properly change
-    // the row ptr array to reflect this removal (either vals == NULL or vals != NULL)
-    // Note, the arrays are squeezed. No memory is freed.
-
-    size_t count = 0;
-
-    size_t nRows = rowPtr.size()-1;
-    if(vals.size() > 0) {
-      for(size_t i = 0; i < nRows; i++) {
-        size_t newStart = count;
-        for(size_t j = rowPtr[i]; j < rowPtr[i+1]; j++) {
-          if(vals[j] != Teuchos::ScalarTraits<Scalar>::zero()) {
-            cols[count  ] = cols[j];
-            vals[count++] = vals[j];
-          }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::squeezeOutNnzs(Teuchos::ArrayRCP<size_t>& rowPtr, Teuchos::ArrayRCP<LocalOrdinal>& cols, Teuchos::ArrayRCP<Scalar>& vals, const std::vector<bool>& keep) const {
+  // get rid of nonzero entries that have 0's in them and properly change
+  // the row ptr array to reflect this removal (either vals == NULL or vals != NULL)
+  // Note, the arrays are squeezed. No memory is freed.
+
+  size_t count = 0;
+
+  size_t nRows = rowPtr.size() - 1;
+  if (vals.size() > 0) {
+    for (size_t i = 0; i < nRows; i++) {
+      size_t newStart = count;
+      for (size_t j = rowPtr[i]; j < rowPtr[i + 1]; j++) {
+        if (vals[j] != Teuchos::ScalarTraits<Scalar>::zero()) {
+          cols[count]   = cols[j];
+          vals[count++] = vals[j];
         }
-        rowPtr[i] = newStart;
       }
-    } else {
-      for (size_t i = 0; i < nRows; i++) {
-        size_t newStart = count;
-        for(size_t j = rowPtr[i]; j < rowPtr[i+1]; j++) {
-          if (keep[j] == true) {
-            cols[count++] = cols[j];
-          }
+      rowPtr[i] = newStart;
+    }
+  } else {
+    for (size_t i = 0; i < nRows; i++) {
+      size_t newStart = count;
+      for (size_t j = rowPtr[i]; j < rowPtr[i + 1]; j++) {
+        if (keep[j] == true) {
+          cols[count++] = cols[j];
         }
-        rowPtr[i] = newStart;
       }
+      rowPtr[i] = newStart;
     }
-    rowPtr[nRows] = count;
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildPaddedMap(const Teuchos::ArrayRCP<const LocalOrdinal> & dofPresent, std::vector<LocalOrdinal> & map, size_t nDofs) const {
-    size_t count = 0;
-    for (decltype(dofPresent.size()) i = 0; i < dofPresent.size(); i++)
-      if(dofPresent[i] == 1) map[count++] = Teuchos::as<LocalOrdinal>(i);
-    TEUCHOS_TEST_FOR_EXCEPTION(nDofs != count, MueLu::Exceptions::RuntimeError, "VariableDofLaplacianFactory::buildPaddedMap: #dofs in dofPresent does not match the expected value (number of rows of A): " << nDofs << " vs. " << count);
+  rowPtr[nRows] = count;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildPaddedMap(const Teuchos::ArrayRCP<const LocalOrdinal>& dofPresent, std::vector<LocalOrdinal>& map, size_t nDofs) const {
+  size_t count = 0;
+  for (decltype(dofPresent.size()) i = 0; i < dofPresent.size(); i++)
+    if (dofPresent[i] == 1) map[count++] = Teuchos::as<LocalOrdinal>(i);
+  TEUCHOS_TEST_FOR_EXCEPTION(nDofs != count, MueLu::Exceptions::RuntimeError, "VariableDofLaplacianFactory::buildPaddedMap: #dofs in dofPresent does not match the expected value (number of rows of A): " << nDofs << " vs. " << count);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::assignGhostLocalNodeIds(const Teuchos::RCP<const Map>& rowDofMap, const Teuchos::RCP<const Map>& colDofMap, std::vector<LocalOrdinal>& myLocalNodeIds, const std::vector<LocalOrdinal>& dofMap, size_t maxDofPerNode, size_t& nLocalNodes, size_t& nLocalPlusGhostNodes, Teuchos::RCP<const Teuchos::Comm<int> > comm) const {
+  size_t nLocalDofs          = rowDofMap->getLocalNumElements();
+  size_t nLocalPlusGhostDofs = colDofMap->getLocalNumElements();  // TODO remove parameters
+
+  // create importer for dof-based information
+  Teuchos::RCP<Import> importer = ImportFactory::Build(rowDofMap, colDofMap);
+
+  // create a vector living on column map of A (dof based)
+  Teuchos::RCP<LOVector> localNodeIdsTemp = LOVectorFactory::Build(rowDofMap, true);
+  Teuchos::RCP<LOVector> localNodeIds     = LOVectorFactory::Build(colDofMap, true);
+
+  // fill local dofs (padded local ids)
+  {
+    Teuchos::ArrayRCP<LocalOrdinal> localNodeIdsTempData = localNodeIdsTemp->getDataNonConst(0);
+    for (size_t i = 0; i < localNodeIdsTemp->getLocalLength(); i++)
+      localNodeIdsTempData[i] = std::floor<LocalOrdinal>(dofMap[i] / maxDofPerNode);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::assignGhostLocalNodeIds(const Teuchos::RCP<const Map> & rowDofMap, const Teuchos::RCP<const Map> & colDofMap, std::vector<LocalOrdinal> & myLocalNodeIds, const std::vector<LocalOrdinal> & dofMap, size_t maxDofPerNode, size_t& nLocalNodes, size_t& nLocalPlusGhostNodes, Teuchos::RCP< const Teuchos::Comm< int > > comm) const {
+  localNodeIds->doImport(*localNodeIdsTemp, *importer, Xpetra::INSERT);
+  Teuchos::ArrayRCP<const LocalOrdinal> localNodeIdsData = localNodeIds->getData(0);
 
-    size_t nLocalDofs = rowDofMap->getLocalNumElements();
-    size_t nLocalPlusGhostDofs = colDofMap->getLocalNumElements(); // TODO remove parameters
+  // Note: localNodeIds contains local ids for the padded version as vector values
 
-    // create importer for dof-based information
-    Teuchos::RCP<Import> importer = ImportFactory::Build(rowDofMap, colDofMap);
+  // we use Scalar instead of int as type
+  Teuchos::RCP<LOVector> myProcTemp = LOVectorFactory::Build(rowDofMap, true);
+  Teuchos::RCP<LOVector> myProc     = LOVectorFactory::Build(colDofMap, true);
 
-    // create a vector living on column map of A (dof based)
-    Teuchos::RCP<LOVector> localNodeIdsTemp = LOVectorFactory::Build(rowDofMap,true);
-    Teuchos::RCP<LOVector> localNodeIds     = LOVectorFactory::Build(colDofMap,true);
-
-    // fill local dofs (padded local ids)
-    {
-      Teuchos::ArrayRCP< LocalOrdinal > localNodeIdsTempData = localNodeIdsTemp->getDataNonConst(0);
-      for(size_t i = 0; i < localNodeIdsTemp->getLocalLength(); i++)
-	localNodeIdsTempData[i] = std::floor<LocalOrdinal>( dofMap[i] / maxDofPerNode );
-    }
-
-    localNodeIds->doImport(*localNodeIdsTemp, *importer, Xpetra::INSERT);
-    Teuchos::ArrayRCP< const LocalOrdinal > localNodeIdsData = localNodeIds->getData(0);
-
-    // Note: localNodeIds contains local ids for the padded version as vector values
-
-
-    // we use Scalar instead of int as type
-    Teuchos::RCP<LOVector> myProcTemp = LOVectorFactory::Build(rowDofMap,true);
-    Teuchos::RCP<LOVector> myProc     = LOVectorFactory::Build(colDofMap,true);
-
-    // fill local dofs (padded local ids)
-    {
-      Teuchos::ArrayRCP< LocalOrdinal > myProcTempData = myProcTemp->getDataNonConst(0);
-      for(size_t i = 0; i < myProcTemp->getLocalLength(); i++)
-	myProcTempData[i] = Teuchos::as<LocalOrdinal>(comm->getRank());
-    }
-    myProc->doImport(*myProcTemp, *importer, Xpetra::INSERT);
-    Teuchos::ArrayRCP<LocalOrdinal> myProcData = myProc->getDataNonConst(0); // we have to modify the data (therefore the non-const version)
-
-    // At this point, the ghost part of localNodeIds corresponds to the local ids
-    // associated with the current owning processor. We want to convert these to
-    // local ids associated with the processor on which these are ghosts.
-    // Thus we have to re-number them. In doing this re-numbering we must make sure
-    // that we find all ghosts with the same id & proc and assign a unique local
-    // id to this group (id&proc). To do this find, we sort all ghost entries in
-    // localNodeIds that are owned by the same processor. Then we can look for
-    // duplicates (i.e., several ghost entries corresponding to dofs with the same
-    // node id) easily and make sure these are all assigned to the same local id.
-    // To do the sorting we'll make a temporary copy of the ghosts via tempId and
-    // tempProc and sort this multiple times for each group owned by the same proc.
-
-
-    std::vector<size_t> location(nLocalPlusGhostDofs - nLocalDofs + 1);
-    std::vector<size_t> tempId  (nLocalPlusGhostDofs - nLocalDofs + 1);
-    std::vector<size_t> tempProc(nLocalPlusGhostDofs - nLocalDofs + 1);
-
-    size_t notProcessed = nLocalDofs; // iteration index over all ghosted dofs
-    size_t tempIndex = 0;
-    size_t first = tempIndex;
-    LocalOrdinal neighbor;
-
-    while (notProcessed < nLocalPlusGhostDofs) {
-      neighbor = myProcData[notProcessed]; // get processor id of not-processed element
-      first    = tempIndex;
-      location[tempIndex] = notProcessed;
-      tempId[tempIndex++] = localNodeIdsData[notProcessed];
-      myProcData[notProcessed] = -1 - neighbor;
-
-      for(size_t i = notProcessed + 1; i < nLocalPlusGhostDofs; i++) {
-        if(myProcData[i] == neighbor) {
-          location[tempIndex] = i;
-          tempId[tempIndex++] = localNodeIdsData[i];
-          myProcData[i] = -1; // mark as visited
-        }
+  // fill local dofs (padded local ids)
+  {
+    Teuchos::ArrayRCP<LocalOrdinal> myProcTempData = myProcTemp->getDataNonConst(0);
+    for (size_t i = 0; i < myProcTemp->getLocalLength(); i++)
+      myProcTempData[i] = Teuchos::as<LocalOrdinal>(comm->getRank());
+  }
+  myProc->doImport(*myProcTemp, *importer, Xpetra::INSERT);
+  Teuchos::ArrayRCP<LocalOrdinal> myProcData = myProc->getDataNonConst(0);  // we have to modify the data (therefore the non-const version)
+
+  // At this point, the ghost part of localNodeIds corresponds to the local ids
+  // associated with the current owning processor. We want to convert these to
+  // local ids associated with the processor on which these are ghosts.
+  // Thus we have to re-number them. In doing this re-numbering we must make sure
+  // that we find all ghosts with the same id & proc and assign a unique local
+  // id to this group (id&proc). To do this find, we sort all ghost entries in
+  // localNodeIds that are owned by the same processor. Then we can look for
+  // duplicates (i.e., several ghost entries corresponding to dofs with the same
+  // node id) easily and make sure these are all assigned to the same local id.
+  // To do the sorting we'll make a temporary copy of the ghosts via tempId and
+  // tempProc and sort this multiple times for each group owned by the same proc.
+
+  std::vector<size_t> location(nLocalPlusGhostDofs - nLocalDofs + 1);
+  std::vector<size_t> tempId(nLocalPlusGhostDofs - nLocalDofs + 1);
+  std::vector<size_t> tempProc(nLocalPlusGhostDofs - nLocalDofs + 1);
+
+  size_t notProcessed = nLocalDofs;  // iteration index over all ghosted dofs
+  size_t tempIndex    = 0;
+  size_t first        = tempIndex;
+  LocalOrdinal neighbor;
+
+  while (notProcessed < nLocalPlusGhostDofs) {
+    neighbor                 = myProcData[notProcessed];  // get processor id of not-processed element
+    first                    = tempIndex;
+    location[tempIndex]      = notProcessed;
+    tempId[tempIndex++]      = localNodeIdsData[notProcessed];
+    myProcData[notProcessed] = -1 - neighbor;
+
+    for (size_t i = notProcessed + 1; i < nLocalPlusGhostDofs; i++) {
+      if (myProcData[i] == neighbor) {
+        location[tempIndex] = i;
+        tempId[tempIndex++] = localNodeIdsData[i];
+        myProcData[i]       = -1;  // mark as visited
       }
-      this->MueLu_az_sort<size_t>(&(tempId[first]), tempIndex - first, &(location[first]), NULL);
-      for(size_t i = first; i < tempIndex; i++) tempProc[i] = neighbor;
-
-      // increment index. Find next notProcessed dof index corresponding to first non-visited element
-      notProcessed++;
-      while ( (notProcessed < nLocalPlusGhostDofs) && (myProcData[notProcessed] < 0))
-        notProcessed++;
     }
-    TEUCHOS_TEST_FOR_EXCEPTION(tempIndex != nLocalPlusGhostDofs-nLocalDofs, MueLu::Exceptions::RuntimeError,"Number of nonzero ghosts is inconsistent.");
-
-    // Now assign ids to all ghost nodes (giving the same id to those with the
-    // same myProc[] and the same local id on the proc that actually owns the
-    // variable associated with the ghost
-
-    nLocalNodes = 0; // initialize return value
-    if(nLocalDofs > 0) nLocalNodes = localNodeIdsData[nLocalDofs-1] + 1;
+    this->MueLu_az_sort<size_t>(&(tempId[first]), tempIndex - first, &(location[first]), NULL);
+    for (size_t i = first; i < tempIndex; i++) tempProc[i] = neighbor;
 
-    nLocalPlusGhostNodes = nLocalNodes; // initialize return value
-    if(nLocalDofs < nLocalPlusGhostDofs) nLocalPlusGhostNodes++; // 1st ghost node is unique (not accounted for). number will be increased later, if there are more ghost nodes
+    // increment index. Find next notProcessed dof index corresponding to first non-visited element
+    notProcessed++;
+    while ((notProcessed < nLocalPlusGhostDofs) && (myProcData[notProcessed] < 0))
+      notProcessed++;
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION(tempIndex != nLocalPlusGhostDofs - nLocalDofs, MueLu::Exceptions::RuntimeError, "Number of nonzero ghosts is inconsistent.");
 
-    // check if two adjacent ghost dofs correspond to different nodes. To do this,
-    // check if they are from different processors or whether they have different
-    // local node ids
+  // Now assign ids to all ghost nodes (giving the same id to those with the
+  // same myProc[] and the same local id on the proc that actually owns the
+  // variable associated with the ghost
 
-    // loop over all (remaining) ghost dofs
-    for (size_t i = nLocalDofs+1; i < nLocalPlusGhostDofs; i++) {
-      size_t lagged = nLocalPlusGhostNodes-1;
+  nLocalNodes = 0;  // initialize return value
+  if (nLocalDofs > 0) nLocalNodes = localNodeIdsData[nLocalDofs - 1] + 1;
 
-      // i is a new unique ghost node (not already accounted for)
-      if ((tempId[i-nLocalDofs] != tempId[i-1-nLocalDofs]) ||
-          (tempProc[i-nLocalDofs] != tempProc[i-1-nLocalDofs]))
-        nLocalPlusGhostNodes++; // update number of ghost nodes
-      tempId[i-1-nLocalDofs] = lagged;
-    }
-    if (nLocalPlusGhostDofs > nLocalDofs)
-      tempId[nLocalPlusGhostDofs-1-nLocalDofs] = nLocalPlusGhostNodes - 1;
+  nLocalPlusGhostNodes = nLocalNodes;                            // initialize return value
+  if (nLocalDofs < nLocalPlusGhostDofs) nLocalPlusGhostNodes++;  // 1st ghost node is unique (not accounted for). number will be increased later, if there are more ghost nodes
 
-    // fill myLocalNodeIds array. Start with local part (not ghosted)
-    for(size_t i = 0; i < nLocalDofs; i++)
-      myLocalNodeIds[i] = std::floor<LocalOrdinal>( dofMap[i] / maxDofPerNode );
+  // check if two adjacent ghost dofs correspond to different nodes. To do this,
+  // check if they are from different processors or whether they have different
+  // local node ids
 
-    // copy ghosted nodal ids into myLocalNodeIds
-    for(size_t i = nLocalDofs; i < nLocalPlusGhostDofs; i++)
-      myLocalNodeIds[location[i-nLocalDofs]] = tempId[i-nLocalDofs];
+  // loop over all (remaining) ghost dofs
+  for (size_t i = nLocalDofs + 1; i < nLocalPlusGhostDofs; i++) {
+    size_t lagged = nLocalPlusGhostNodes - 1;
 
+    // i is a new unique ghost node (not already accounted for)
+    if ((tempId[i - nLocalDofs] != tempId[i - 1 - nLocalDofs]) ||
+        (tempProc[i - nLocalDofs] != tempProc[i - 1 - nLocalDofs]))
+      nLocalPlusGhostNodes++;  // update number of ghost nodes
+    tempId[i - 1 - nLocalDofs] = lagged;
   }
+  if (nLocalPlusGhostDofs > nLocalDofs)
+    tempId[nLocalPlusGhostDofs - 1 - nLocalDofs] = nLocalPlusGhostNodes - 1;
+
+  // fill myLocalNodeIds array. Start with local part (not ghosted)
+  for (size_t i = 0; i < nLocalDofs; i++)
+    myLocalNodeIds[i] = std::floor<LocalOrdinal>(dofMap[i] / maxDofPerNode);
 
-} /* MueLu */
+  // copy ghosted nodal ids into myLocalNodeIds
+  for (size_t i = nLocalDofs; i < nLocalPlusGhostDofs; i++)
+    myLocalNodeIds[location[i - nLocalDofs]] = tempId[i - nLocalDofs];
+}
 
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_GRAPH_MUELU_VARIABLEDOFLAPLACIANFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase.hpp b/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase.hpp
index 8a78e56cd9ca..3ab43d690931 100644
--- a/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase.hpp
+++ b/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase.hpp
@@ -57,42 +57,40 @@
 
 namespace MueLu {
 
-  /*!
-       @class AggregationAlgorithmBase
-       @brief Pure virtual base class for all MueLu aggregation algorithms
+/*!
+     @class AggregationAlgorithmBase
+     @brief Pure virtual base class for all MueLu aggregation algorithms
 
-       @ingroup MueLuBaseClasses
-   */
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationAlgorithmBase : public BaseClass {
+     @ingroup MueLuBaseClasses
+ */
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationAlgorithmBase : public BaseClass {
 #undef MUELU_AGGREGATIONALGORITHMBASE_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
-    public:
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! @name Constructors/Destructors
-    //@{
+  //! Destructor.
+  virtual ~AggregationAlgorithmBase() {}
 
-    //! Destructor.
-    virtual ~AggregationAlgorithmBase() {}
+  //@}
 
-    //@}
+  //! @name Build routines
+  //@{
 
-    //! @name Build routines
-    //@{
+  //! BuildAggregates routine.
+  virtual void BuildAggregates(const Teuchos::ParameterList& params,
+                               const GraphBase& graph,
+                               Aggregates& aggregates,
+                               std::vector<unsigned>& aggStat,
+                               LO& numNonAggregatedNodes) const = 0;
+  //@}
+};
 
-    //! BuildAggregates routine.
-    virtual void BuildAggregates(const Teuchos::ParameterList& params,
-                                 const GraphBase& graph,
-                                 Aggregates& aggregates,
-                                 std::vector<unsigned>& aggStat,
-                                 LO& numNonAggregatedNodes) const = 0;
-    //@}
-
-  };
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONALGORITHMBASE_SHORT
 #endif /* MUELU_AGGREGATIONALGORITHMBASE_HPP_ */
diff --git a/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase_kokkos.hpp b/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase_kokkos.hpp
index e758bdf1c84c..b79f6cf8fd9c 100644
--- a/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase_kokkos.hpp
+++ b/packages/muelu/src/Graph/MueLu_AggregationAlgorithmBase_kokkos.hpp
@@ -59,42 +59,42 @@
 
 namespace MueLu {
 
-  /*!
-       @class AggregationAlgorithmBase
-       @brief Pure virtual base class for all MueLu aggregation algorithms
+/*!
+     @class AggregationAlgorithmBase
+     @brief Pure virtual base class for all MueLu aggregation algorithms
 
-       @ingroup MueLuBaseClasses
-   */
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationAlgorithmBase_kokkos : public BaseClass {
+     @ingroup MueLuBaseClasses
+ */
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationAlgorithmBase_kokkos : public BaseClass {
 #undef MUELU_AGGREGATIONALGORITHMBASE_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
-    public:
-    using device_type = typename LWGraph_kokkos::device_type;
+ public:
+  using device_type = typename LWGraph_kokkos::device_type;
 
-    //! @name Constructors/Destructors
-    //@{
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Destructor.
-    virtual ~AggregationAlgorithmBase_kokkos() {}
+  //! Destructor.
+  virtual ~AggregationAlgorithmBase_kokkos() {}
 
-    //@}
+  //@}
 
-    //! @name Build routines
-    //@{
+  //! @name Build routines
+  //@{
 
-    //! BuildAggregates routine.
-    virtual void BuildAggregates(const Teuchos::ParameterList& params,
-                                 const LWGraph_kokkos& graph,
-                                 Aggregates& aggregates,
-                                 Kokkos::View<unsigned*, device_type>& aggStat,
-                                 LO& numNonAggregatedNodes) const = 0;
-    //@}
-  };
+  //! BuildAggregates routine.
+  virtual void BuildAggregates(const Teuchos::ParameterList& params,
+                               const LWGraph_kokkos& graph,
+                               Aggregates& aggregates,
+                               Kokkos::View<unsigned*, device_type>& aggStat,
+                               LO& numNonAggregatedNodes) const = 0;
+  //@}
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONALGORITHMBASE_KOKKOS_SHORT
-#endif // MUELU_AGGREGATIONALGORITHMBASE_KOKKOS_HPP
+#endif  // MUELU_AGGREGATIONALGORITHMBASE_KOKKOS_HPP
diff --git a/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_decl.hpp b/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_decl.hpp
index ccdf3e3a7782..7252edd8ea03 100644
--- a/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_decl.hpp
@@ -48,11 +48,9 @@
 
 #include "MueLu_ConfigDefs.hpp"
 
-
 #include <Xpetra_Map_fwd.hpp>
 #include <Xpetra_Vector_fwd.hpp>
 
-
 #include <Xpetra_Matrix_fwd.hpp>
 
 #include "MueLu_GraphBase_fwd.hpp"
@@ -67,15 +65,15 @@
 
 namespace MueLu {
 
-template<class Scalar = DefaultScalar,
-         class LocalOrdinal = DefaultLocalOrdinal,
-         class GlobalOrdinal = DefaultGlobalOrdinal,
-         class Node = DefaultNode>
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
 class NotayAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_NOTAYAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-public:
+ public:
   //! @name typedefs
   //@{
   using local_matrix_type = typename Matrix::local_matrix_type;
@@ -86,15 +84,14 @@ class NotayAggregationFactory : public SingleLevelFactoryBase {
   using row_sum_type      = typename Kokkos::View<impl_scalar_type*, Kokkos::LayoutLeft, device_type>;
   //@}
 
-
   //! @name Constructors/Destructors.
   //@{
 
   //! Constructor.
-  NotayAggregationFactory() { };
+  NotayAggregationFactory(){};
 
   //! Destructor.
-  virtual ~NotayAggregationFactory() { }
+  virtual ~NotayAggregationFactory() {}
 
   RCP<const ParameterList> GetValidParameterList() const;
 
@@ -108,7 +105,7 @@ class NotayAggregationFactory : public SingleLevelFactoryBase {
   //! Input
   //@{
 
-  void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
   //@}
 
@@ -116,7 +113,7 @@ class NotayAggregationFactory : public SingleLevelFactoryBase {
   //@{
 
   /*! @brief Build aggregates. */
-  void Build(Level &currentLevel) const;
+  void Build(Level& currentLevel) const;
 
   /*! @brief Initial aggregation phase. */
   void BuildInitialAggregates(const Teuchos::ParameterList& params,
@@ -131,7 +128,7 @@ class NotayAggregationFactory : public SingleLevelFactoryBase {
   /*! @brief Further aggregation phase increases coarsening rate by a factor of ~2 per iteration. */
   void BuildFurtherAggregates(const Teuchos::ParameterList& params,
                               const RCP<const Matrix>& A,
-                              const Teuchos::ArrayView<const LO> & orderingVector,
+                              const Teuchos::ArrayView<const LO>& orderingVector,
                               const local_matrix_type& coarseA,
                               const magnitude_type kappa,
                               const row_sum_type& rowSum,
@@ -160,13 +157,12 @@ class NotayAggregationFactory : public SingleLevelFactoryBase {
                    const std::string matrixLabel,
                    local_matrix_type& C) const;
 
-
   //@}
 
-private:
-}; // class NotayAggregationFactory
+ private:
+};  // class NotayAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_NOTAYAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_NOTAYAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_def.hpp b/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_def.hpp
index b84022432e41..60949d3ad932 100644
--- a/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/PairwiseAggregation/MueLu_NotayAggregationFactory_def.hpp
@@ -66,880 +66,865 @@
 #include "MueLu_Types.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLu {
 
-  namespace NotayUtils {
-    template <class LocalOrdinal>
-    LocalOrdinal RandomOrdinal(LocalOrdinal min, LocalOrdinal max) {
-      return min + as<LocalOrdinal>((max-min+1) * (static_cast<double>(std::rand()) / (RAND_MAX + 1.0)));
-    }
-
-    template <class LocalOrdinal>
-    void RandomReorder(Teuchos::Array<LocalOrdinal> & list) {
-      typedef LocalOrdinal LO;
-      LO n = Teuchos::as<LO>(list.size());
-      for(LO i = 0; i < n-1; i++)
-	std::swap(list[i], list[RandomOrdinal(i,n-1)]);
-    }
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
+namespace NotayUtils {
+template <class LocalOrdinal>
+LocalOrdinal RandomOrdinal(LocalOrdinal min, LocalOrdinal max) {
+  return min + as<LocalOrdinal>((max - min + 1) * (static_cast<double>(std::rand()) / (RAND_MAX + 1.0)));
+}
+
+template <class LocalOrdinal>
+void RandomReorder(Teuchos::Array<LocalOrdinal>& list) {
+  typedef LocalOrdinal LO;
+  LO n = Teuchos::as<LO>(list.size());
+  for (LO i = 0; i < n - 1; i++)
+    std::swap(list[i], list[RandomOrdinal(i, n - 1)]);
+}
+}  // namespace NotayUtils
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: pairwise: size");
-    SET_VALID_ENTRY("aggregation: pairwise: tie threshold");
-    SET_VALID_ENTRY("aggregation: compute aggregate qualities");
-    SET_VALID_ENTRY("aggregation: Dirichlet threshold");
-    SET_VALID_ENTRY("aggregation: ordering");
+  SET_VALID_ENTRY("aggregation: pairwise: size");
+  SET_VALID_ENTRY("aggregation: pairwise: tie threshold");
+  SET_VALID_ENTRY("aggregation: compute aggregate qualities");
+  SET_VALID_ENTRY("aggregation: Dirichlet threshold");
+  SET_VALID_ENTRY("aggregation: ordering");
 #undef SET_VALID_ENTRY
 
-    // general variables needed in AggregationFactory
-    validParamList->set< RCP<const FactoryBase> >("A",           null, "Generating factory of the matrix");
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
-    validParamList->set< RCP<const FactoryBase> >("AggregateQualities", null, "Generating factory for variable \'AggregateQualities\'");
+  // general variables needed in AggregationFactory
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix");
+  validParamList->set<RCP<const FactoryBase> >("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+  validParamList->set<RCP<const FactoryBase> >("AggregateQualities", null, "Generating factory for variable \'AggregateQualities\'");
 
+  return validParamList;
+}
 
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    const ParameterList& pL = GetParameterList();
-
-    Input(currentLevel, "A");
-    Input(currentLevel, "Graph");
-    Input(currentLevel, "DofsPerNode");
-    if (pL.get<bool>("aggregation: compute aggregate qualities")) {
-        Input(currentLevel, "AggregateQualities");
-    }
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
 
+  Input(currentLevel, "A");
+  Input(currentLevel, "Graph");
+  Input(currentLevel, "DofsPerNode");
+  if (pL.get<bool>("aggregation: compute aggregate qualities")) {
+    Input(currentLevel, "AggregateQualities");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  using STS = Teuchos::ScalarTraits<Scalar>;
+  using MT  = typename STS::magnitudeType;
+
+  const MT MT_TWO = Teuchos::ScalarTraits<MT>::one() + Teuchos::ScalarTraits<MT>::one();
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
+  const ParameterList& pL = GetParameterList();
+
+  const MT kappa = static_cast<MT>(pL.get<double>("aggregation: Dirichlet threshold"));
+  TEUCHOS_TEST_FOR_EXCEPTION(kappa <= MT_TWO,
+                             Exceptions::RuntimeError,
+                             "Pairwise requires kappa > 2"
+                             " otherwise all rows are considered as Dirichlet rows.");
+
+  // Parameters
+  int maxNumIter = 3;
+  if (pL.isParameter("aggregation: pairwise: size"))
+    maxNumIter = pL.get<int>("aggregation: pairwise: size");
+  TEUCHOS_TEST_FOR_EXCEPTION(maxNumIter < 1,
+                             Exceptions::RuntimeError,
+                             "NotayAggregationFactory::Build(): \"aggregation: pairwise: size\""
+                             " must be a strictly positive integer");
+
+  RCP<const GraphBase> graph = Get<RCP<GraphBase> >(currentLevel, "Graph");
+  RCP<const Matrix> A        = Get<RCP<Matrix> >(currentLevel, "A");
+
+  // Setup aggregates & aggStat objects
+  RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("PW");
+
+  const LO numRows = graph->GetNodeNumVertices();
+
+  // construct aggStat information
+  std::vector<unsigned> aggStat(numRows, READY);
+
+  const int DofsPerNode = Get<int>(currentLevel, "DofsPerNode");
+  TEUCHOS_TEST_FOR_EXCEPTION(DofsPerNode != 1, Exceptions::RuntimeError,
+                             "Pairwise only supports one dof per node");
+
+  // This follows the paper:
+  // Notay, "Aggregation-based algebraic multigrid for convection-diffusion equations",
+  // SISC 34(3), pp. A2288-2316.
+
+  // Handle Ordering
+  std::string orderingStr = pL.get<std::string>("aggregation: ordering");
+  enum {
+    O_NATURAL,
+    O_RANDOM,
+    O_CUTHILL_MCKEE,
+  } ordering;
+
+  ordering = O_NATURAL;
+  if (orderingStr == "random")
+    ordering = O_RANDOM;
+  else if (orderingStr == "natural") {
+  } else if (orderingStr == "cuthill-mckee" || orderingStr == "cm")
+    ordering = O_CUTHILL_MCKEE;
+  else {
+    TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::RuntimeError, "Invalid ordering type");
+  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    using STS = Teuchos::ScalarTraits<Scalar>;
-    using MT  = typename STS::magnitudeType;
-
-    const MT MT_TWO = Teuchos::ScalarTraits<MT>::one() + Teuchos::ScalarTraits<MT>::one();
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    const ParameterList& pL = GetParameterList();
-
-    const MT kappa = static_cast<MT>(pL.get<double>("aggregation: Dirichlet threshold"));
-    TEUCHOS_TEST_FOR_EXCEPTION(kappa <= MT_TWO,
-                               Exceptions::RuntimeError,
-                               "Pairwise requires kappa > 2"
-                               " otherwise all rows are considered as Dirichlet rows.");
-
-    // Parameters
-    int maxNumIter = 3;
-    if (pL.isParameter("aggregation: pairwise: size"))
-      maxNumIter = pL.get<int>("aggregation: pairwise: size");
-    TEUCHOS_TEST_FOR_EXCEPTION(maxNumIter < 1,
-			       Exceptions::RuntimeError,
-			       "NotayAggregationFactory::Build(): \"aggregation: pairwise: size\""
-                               " must be a strictly positive integer");
-
-
-    RCP<const GraphBase> graph = Get< RCP<GraphBase> >(currentLevel, "Graph");
-    RCP<const Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-
-    // Setup aggregates & aggStat objects
-    RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("PW");
-
-    const LO numRows = graph->GetNodeNumVertices();
-
-    // construct aggStat information
-    std::vector<unsigned> aggStat(numRows, READY);
-
-
-    const int DofsPerNode =  Get<int>(currentLevel,"DofsPerNode");
-    TEUCHOS_TEST_FOR_EXCEPTION(DofsPerNode != 1, Exceptions::RuntimeError,
-			       "Pairwise only supports one dof per node");
+  // Get an ordering vector
+  // NOTE: The orderingVector only orders *rows* of the matrix.  Off-proc columns
+  // will get ignored in the aggregation phases, so we don't need to worry about
+  // running off the end.
+  Array<LO> orderingVector(numRows);
+  for (LO i = 0; i < numRows; i++)
+    orderingVector[i] = i;
+  if (ordering == O_RANDOM)
+    MueLu::NotayUtils::RandomReorder(orderingVector);
+  else if (ordering == O_CUTHILL_MCKEE) {
+    RCP<Xpetra::Vector<LO, LO, GO, NO> > rcmVector = MueLu::Utilities<SC, LO, GO, NO>::CuthillMcKee(*A);
+    auto localVector                               = rcmVector->getData(0);
+    for (LO i = 0; i < numRows; i++)
+      orderingVector[i] = localVector[i];
+  }
 
-    // This follows the paper:
-    // Notay, "Aggregation-based algebraic multigrid for convection-diffusion equations",
-    // SISC 34(3), pp. A2288-2316.
+  // Get the party stated
+  LO numNonAggregatedNodes = numRows, numDirichletNodes = 0;
+  BuildInitialAggregates(pL, A, orderingVector(), kappa,
+                         *aggregates, aggStat, numNonAggregatedNodes, numDirichletNodes);
+  TEUCHOS_TEST_FOR_EXCEPTION(0 < numNonAggregatedNodes, Exceptions::RuntimeError,
+                             "Initial pairwise aggregation failed to aggregate all nodes");
+  LO numLocalAggregates = aggregates->GetNumAggregates();
+  GetOStream(Statistics0) << "Init   : " << numLocalAggregates << " - "
+                          << A->getLocalNumRows() / numLocalAggregates << std::endl;
+
+  // Temporary data storage for further aggregation steps
+  local_matrix_type intermediateP;
+  local_matrix_type coarseLocalA;
+
+  // Compute the on rank part of the local matrix
+  // that the square submatrix that only contains
+  // columns corresponding to local rows.
+  LO numLocalDirichletNodes = numDirichletNodes;
+  Array<LO> localVertex2AggId(aggregates->GetVertex2AggId()->getData(0).view(0, numRows));
+  BuildOnRankLocalMatrix(A->getLocalMatrixDevice(), coarseLocalA);
+  for (LO aggregationIter = 1; aggregationIter < maxNumIter; ++aggregationIter) {
+    // Compute the intermediate prolongator
+    BuildIntermediateProlongator(coarseLocalA.numRows(), numLocalDirichletNodes, numLocalAggregates,
+                                 localVertex2AggId(), intermediateP);
+
+    // Compute the coarse local matrix and coarse row sum
+    BuildCoarseLocalMatrix(intermediateP, coarseLocalA);
+
+    // Directly compute rowsum from A, rather than coarseA
+    row_sum_type rowSum("rowSum", numLocalAggregates);
+    {
+      std::vector<std::vector<LO> > agg2vertex(numLocalAggregates);
+      auto vertex2AggId = aggregates->GetVertex2AggId()->getData(0);
+      for (LO i = 0; i < (LO)numRows; i++) {
+        if (aggStat[i] != AGGREGATED)
+          continue;
+        LO agg = vertex2AggId[i];
+        agg2vertex[agg].push_back(i);
+      }
 
-    // Handle Ordering
-    std::string orderingStr = pL.get<std::string>("aggregation: ordering");
-    enum {
-      O_NATURAL,
-      O_RANDOM,
-      O_CUTHILL_MCKEE,
-    } ordering;
+      typename row_sum_type::HostMirror rowSum_h = Kokkos::create_mirror_view(rowSum);
+      for (LO i = 0; i < numRows; i++) {
+        // If not aggregated already, skip this guy
+        if (aggStat[i] != AGGREGATED)
+          continue;
+        int agg                = vertex2AggId[i];
+        std::vector<LO>& myagg = agg2vertex[agg];
+
+        size_t nnz = A->getNumEntriesInLocalRow(i);
+        ArrayView<const LO> indices;
+        ArrayView<const SC> vals;
+        A->getLocalRowView(i, indices, vals);
+
+        SC mysum = Teuchos::ScalarTraits<Scalar>::zero();
+        for (LO colidx = 0; colidx < static_cast<LO>(nnz); colidx++) {
+          bool found = false;
+          if (indices[colidx] < numRows) {
+            for (LO j = 0; j < (LO)myagg.size(); j++)
+              if (vertex2AggId[indices[colidx]] == agg)
+                found = true;
+          }
+          if (!found) {
+            *out << "- ADDING col " << indices[colidx] << " = " << vals[colidx] << std::endl;
+            mysum += vals[colidx];
+          } else {
+            *out << "- NOT ADDING col " << indices[colidx] << " = " << vals[colidx] << std::endl;
+          }
+        }
 
-    ordering = O_NATURAL;
-    if (orderingStr == "random" ) ordering = O_RANDOM;
-    else if(orderingStr == "natural") {}
-    else if(orderingStr == "cuthill-mckee" || orderingStr == "cm") ordering = O_CUTHILL_MCKEE;
-    else {
-      TEUCHOS_TEST_FOR_EXCEPTION(1,Exceptions::RuntimeError,"Invalid ordering type");
+        rowSum_h[agg] = mysum;
+      }
+      Kokkos::deep_copy(rowSum, rowSum_h);
     }
 
-    // Get an ordering vector
-    // NOTE: The orderingVector only orders *rows* of the matrix.  Off-proc columns
-    // will get ignored in the aggregation phases, so we don't need to worry about
-    // running off the end.
-    Array<LO> orderingVector(numRows);
+    // Get local orderingVector
+    Array<LO> localOrderingVector(numRows);
     for (LO i = 0; i < numRows; i++)
-      orderingVector[i] = i;
+      localOrderingVector[i] = i;
     if (ordering == O_RANDOM)
-      MueLu::NotayUtils::RandomReorder(orderingVector);
+      MueLu::NotayUtils::RandomReorder(localOrderingVector);
     else if (ordering == O_CUTHILL_MCKEE) {
-      RCP<Xpetra::Vector<LO,LO,GO,NO> > rcmVector = MueLu::Utilities<SC,LO,GO,NO>::CuthillMcKee(*A);
-      auto localVector = rcmVector->getData(0);
+      RCP<Xpetra::Vector<LO, LO, GO, NO> > rcmVector = MueLu::Utilities<SC, LO, GO, NO>::CuthillMcKee(*A);
+      auto localVector                               = rcmVector->getData(0);
       for (LO i = 0; i < numRows; i++)
-        orderingVector[i] = localVector[i];
+        localOrderingVector[i] = localVector[i];
     }
 
-    // Get the party stated
-    LO numNonAggregatedNodes = numRows, numDirichletNodes = 0;
-    BuildInitialAggregates(pL, A, orderingVector(), kappa,
-                           *aggregates, aggStat, numNonAggregatedNodes, numDirichletNodes);
-    TEUCHOS_TEST_FOR_EXCEPTION(0 < numNonAggregatedNodes, Exceptions::RuntimeError,
-                               "Initial pairwise aggregation failed to aggregate all nodes");
-    LO numLocalAggregates = aggregates->GetNumAggregates();
-    GetOStream(Statistics0) << "Init   : " << numLocalAggregates << " - "
-                            << A->getLocalNumRows() / numLocalAggregates << std::endl;
-
-    // Temporary data storage for further aggregation steps
-    local_matrix_type intermediateP;
-    local_matrix_type coarseLocalA;
-
-    // Compute the on rank part of the local matrix
-    // that the square submatrix that only contains
-    // columns corresponding to local rows.
-    LO numLocalDirichletNodes = numDirichletNodes;
-    Array<LO> localVertex2AggId(aggregates->GetVertex2AggId()->getData(0).view(0, numRows));
-    BuildOnRankLocalMatrix(A->getLocalMatrixDevice(), coarseLocalA);
-    for(LO aggregationIter = 1; aggregationIter < maxNumIter; ++aggregationIter) {
-      // Compute the intermediate prolongator
-      BuildIntermediateProlongator(coarseLocalA.numRows(), numLocalDirichletNodes, numLocalAggregates,
-                                   localVertex2AggId(), intermediateP);
-
-      // Compute the coarse local matrix and coarse row sum
-      BuildCoarseLocalMatrix(intermediateP, coarseLocalA);
-
-      // Directly compute rowsum from A, rather than coarseA
-      row_sum_type rowSum("rowSum", numLocalAggregates);
-      {
-        std::vector<std::vector<LO> > agg2vertex(numLocalAggregates);
-        auto vertex2AggId = aggregates->GetVertex2AggId()->getData(0);
-        for(LO i=0; i<(LO)numRows; i++) {
-          if(aggStat[i] != AGGREGATED)
-            continue;
-          LO agg=vertex2AggId[i];
-          agg2vertex[agg].push_back(i);
-        }
-        
-        typename row_sum_type::HostMirror rowSum_h = Kokkos::create_mirror_view(rowSum);
-        for(LO i = 0; i < numRows; i++) {
-          // If not aggregated already, skip this guy
-          if(aggStat[i] != AGGREGATED)
-            continue;
-          int agg = vertex2AggId[i];
-          std::vector<LO> & myagg = agg2vertex[agg];
-
-          size_t nnz = A->getNumEntriesInLocalRow(i);
-          ArrayView<const LO> indices;
-          ArrayView<const SC> vals;
-          A->getLocalRowView(i, indices, vals);
-
-          SC mysum = Teuchos::ScalarTraits<Scalar>::zero();
-          for (LO colidx = 0; colidx < static_cast<LO>(nnz); colidx++) {
-            bool found = false;
-            if(indices[colidx] < numRows) {
-              for(LO j=0; j<(LO)myagg.size(); j++)
-                if (vertex2AggId[indices[colidx]] == agg)
-                  found=true;
-            }
-            if(!found) {
-              *out << "- ADDING col "<<indices[colidx]<<" = "<<vals[colidx] << std::endl;
-                mysum += vals[colidx];
-              }
-              else {
-              *out << "- NOT ADDING col "<<indices[colidx]<<" = "<<vals[colidx] << std::endl;
-              }
-          }
-          
-          rowSum_h[agg] = mysum;
-        }
-        Kokkos::deep_copy(rowSum, rowSum_h);
+    // Compute new aggregates
+    numLocalAggregates    = 0;
+    numNonAggregatedNodes = static_cast<LO>(coarseLocalA.numRows());
+    std::vector<LO> localAggStat(numNonAggregatedNodes, READY);
+    localVertex2AggId.resize(numNonAggregatedNodes, -1);
+    BuildFurtherAggregates(pL, A, localOrderingVector, coarseLocalA, kappa, rowSum,
+                           localAggStat, localVertex2AggId,
+                           numLocalAggregates, numNonAggregatedNodes);
+
+    // After the first initial pairwise aggregation
+    // the Dirichlet nodes have been removed.
+    numLocalDirichletNodes = 0;
+
+    // Update the aggregates
+    RCP<LOMultiVector> vertex2AggIdMV = aggregates->GetVertex2AggId();
+    ArrayRCP<LO> vertex2AggId         = vertex2AggIdMV->getDataNonConst(0);
+    for (size_t vertexIdx = 0; vertexIdx < A->getLocalNumRows(); ++vertexIdx) {
+      LO oldAggIdx = vertex2AggId[vertexIdx];
+      if (oldAggIdx != Teuchos::OrdinalTraits<LO>::invalid()) {
+        vertex2AggId[vertexIdx] = localVertex2AggId[oldAggIdx];
       }
-
-      // Get local orderingVector
-      Array<LO> localOrderingVector(numRows);
-      for (LO i = 0; i < numRows; i++)
-        localOrderingVector[i] = i;
-      if (ordering == O_RANDOM)
-        MueLu::NotayUtils::RandomReorder(localOrderingVector);
-      else if (ordering == O_CUTHILL_MCKEE) {
-        RCP<Xpetra::Vector<LO,LO,GO,NO> > rcmVector = MueLu::Utilities<SC,LO,GO,NO>::CuthillMcKee(*A);
-        auto localVector = rcmVector->getData(0);
-        for (LO i = 0; i < numRows; i++)
-          localOrderingVector[i] = localVector[i];
-      }
-
-      // Compute new aggregates
-      numLocalAggregates    = 0;
-      numNonAggregatedNodes = static_cast<LO>(coarseLocalA.numRows());
-      std::vector<LO> localAggStat(numNonAggregatedNodes, READY);
-      localVertex2AggId.resize(numNonAggregatedNodes, -1);
-      BuildFurtherAggregates(pL, A, localOrderingVector, coarseLocalA, kappa, rowSum,
-                             localAggStat, localVertex2AggId,
-                             numLocalAggregates, numNonAggregatedNodes);
-
-      // After the first initial pairwise aggregation
-      // the Dirichlet nodes have been removed.
-      numLocalDirichletNodes = 0;
-
-      // Update the aggregates
-      RCP<LOMultiVector> vertex2AggIdMV = aggregates->GetVertex2AggId();
-      ArrayRCP<LO>       vertex2AggId   = vertex2AggIdMV->getDataNonConst(0);
-      for(size_t vertexIdx = 0; vertexIdx < A->getLocalNumRows(); ++vertexIdx) {
-        LO oldAggIdx = vertex2AggId[vertexIdx];
-        if(oldAggIdx != Teuchos::OrdinalTraits<LO>::invalid()) {
-          vertex2AggId[vertexIdx] = localVertex2AggId[oldAggIdx];
-        }
-      }
-
-      // We could probably print some better statistics at some point
-      GetOStream(Statistics0) << "Iter " << aggregationIter << ": " << numLocalAggregates << " - "
-                              << A->getLocalNumRows() / numLocalAggregates << std::endl;
     }
-    aggregates->SetNumAggregates(numLocalAggregates);
-    aggregates->AggregatesCrossProcessors(false);
-    aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
 
-    // DO stuff
-    Set(currentLevel, "Aggregates", aggregates);
-    GetOStream(Statistics0) << aggregates->description() << std::endl;
+    // We could probably print some better statistics at some point
+    GetOStream(Statistics0) << "Iter " << aggregationIter << ": " << numLocalAggregates << " - "
+                            << A->getLocalNumRows() / numLocalAggregates << std::endl;
+  }
+  aggregates->SetNumAggregates(numLocalAggregates);
+  aggregates->AggregatesCrossProcessors(false);
+  aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
+
+  // DO stuff
+  Set(currentLevel, "Aggregates", aggregates);
+  GetOStream(Statistics0) << aggregates->description() << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildInitialAggregates(const Teuchos::ParameterList& params,
+                           const RCP<const Matrix>& A,
+                           const Teuchos::ArrayView<const LO>& orderingVector,
+                           const typename Teuchos::ScalarTraits<Scalar>::magnitudeType kappa,
+                           Aggregates& aggregates,
+                           std::vector<unsigned>& aggStat,
+                           LO& numNonAggregatedNodes,
+                           LO& numDirichletNodes) const {
+  Monitor m(*this, "BuildInitialAggregates");
+  using STS              = Teuchos::ScalarTraits<Scalar>;
+  using MT               = typename STS::magnitudeType;
+  using RealValuedVector = Xpetra::Vector<MT, LocalOrdinal, GlobalOrdinal, Node>;
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildInitialAggregates(const Teuchos::ParameterList& params,
-                         const RCP<const Matrix>& A,
-			 const Teuchos::ArrayView<const LO> & orderingVector,
-                         const typename Teuchos::ScalarTraits<Scalar>::magnitudeType kappa,
-                         Aggregates& aggregates,
-                         std::vector<unsigned>& aggStat,
-                         LO& numNonAggregatedNodes,
-                         LO& numDirichletNodes) const {
-
-    Monitor m(*this, "BuildInitialAggregates");
-    using STS = Teuchos::ScalarTraits<Scalar>;
-    using MT  = typename STS::magnitudeType;
-    using RealValuedVector = Xpetra::Vector<MT,LocalOrdinal,GlobalOrdinal,Node>;
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-
-    const SC SC_ZERO    = Teuchos::ScalarTraits<SC>::zero();
-    const MT MT_ZERO    = Teuchos::ScalarTraits<MT>::zero();
-    const MT MT_ONE     = Teuchos::ScalarTraits<MT>::one();
-    const MT MT_TWO     = MT_ONE + MT_ONE;
-    const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-    const LO LO_ZERO    = Teuchos::OrdinalTraits<LO>::zero();
-
-    const MT kappa_init = kappa / (kappa - MT_TWO);
-    const LO numRows    = aggStat.size();
-    const int myRank    = A->getMap()->getComm()->getRank();
-
-    // For finding "ties" where we fall back to the ordering.  Making this larger than
-    // hard zero substantially increases code robustness.
-    double tie_criterion = params.get<double>("aggregation: pairwise: tie threshold");
-    double tie_less = 1.0 - tie_criterion;
-    double tie_more = 1.0 + tie_criterion;
-
-    // NOTE: Assumes 1 dof per node.  This constraint is enforced in Build(),
-    // and so we're not doing again here.
-    // This should probably be fixed at some point.
-
-    // Extract diagonal, rowsums, etc
-    // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
-    RCP<Vector> ghostedDiag = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixOverlappedDiagonal(*A);
-    RCP<Vector> ghostedRowSum = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixOverlappedDeletedRowsum(*A);
-    RCP<RealValuedVector> ghostedAbsRowSum = MueLu::Utilities<SC,LO,GO,NO>::GetMatrixOverlappedAbsDeletedRowsum(*A);
-    const ArrayRCP<const SC> D     = ghostedDiag->getData(0);
-    const ArrayRCP<const SC> S     = ghostedRowSum->getData(0);
-    const ArrayRCP<const MT> AbsRs = ghostedAbsRowSum->getData(0);
-
-    // Aggregates stuff
-    ArrayRCP<LO>  vertex2AggId_rcp = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO>  procWinner_rcp   = aggregates.GetProcWinner()  ->getDataNonConst(0);
-    ArrayView<LO> vertex2AggId     = vertex2AggId_rcp();
-    ArrayView<LO> procWinner       = procWinner_rcp();
-
-    // Algorithm 4.2
-
-    // 0,1 : Initialize: Flag boundary conditions
-    // Modification: We assume symmetry here aij = aji
-    for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
-      MT aii = STS::magnitude(D[row]);
-      MT rowsum = AbsRs[row];
-
-      if(aii >= kappa_init * rowsum) {
-        *out << "Flagging index " << row << " as dirichlet "
-          "aii >= kappa*rowsum = " << aii << " >= " << kappa_init << " " << rowsum << std::endl;
-	aggStat[row] = IGNORED;
-	--numNonAggregatedNodes;
-        ++numDirichletNodes;
-      }
+  const SC SC_ZERO    = Teuchos::ScalarTraits<SC>::zero();
+  const MT MT_ZERO    = Teuchos::ScalarTraits<MT>::zero();
+  const MT MT_ONE     = Teuchos::ScalarTraits<MT>::one();
+  const MT MT_TWO     = MT_ONE + MT_ONE;
+  const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+  const LO LO_ZERO    = Teuchos::OrdinalTraits<LO>::zero();
+
+  const MT kappa_init = kappa / (kappa - MT_TWO);
+  const LO numRows    = aggStat.size();
+  const int myRank    = A->getMap()->getComm()->getRank();
+
+  // For finding "ties" where we fall back to the ordering.  Making this larger than
+  // hard zero substantially increases code robustness.
+  double tie_criterion = params.get<double>("aggregation: pairwise: tie threshold");
+  double tie_less      = 1.0 - tie_criterion;
+  double tie_more      = 1.0 + tie_criterion;
+
+  // NOTE: Assumes 1 dof per node.  This constraint is enforced in Build(),
+  // and so we're not doing again here.
+  // This should probably be fixed at some point.
+
+  // Extract diagonal, rowsums, etc
+  // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
+  RCP<Vector> ghostedDiag                = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixOverlappedDiagonal(*A);
+  RCP<Vector> ghostedRowSum              = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixOverlappedDeletedRowsum(*A);
+  RCP<RealValuedVector> ghostedAbsRowSum = MueLu::Utilities<SC, LO, GO, NO>::GetMatrixOverlappedAbsDeletedRowsum(*A);
+  const ArrayRCP<const SC> D             = ghostedDiag->getData(0);
+  const ArrayRCP<const SC> S             = ghostedRowSum->getData(0);
+  const ArrayRCP<const MT> AbsRs         = ghostedAbsRowSum->getData(0);
+
+  // Aggregates stuff
+  ArrayRCP<LO> vertex2AggId_rcp = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner_rcp   = aggregates.GetProcWinner()->getDataNonConst(0);
+  ArrayView<LO> vertex2AggId    = vertex2AggId_rcp();
+  ArrayView<LO> procWinner      = procWinner_rcp();
+
+  // Algorithm 4.2
+
+  // 0,1 : Initialize: Flag boundary conditions
+  // Modification: We assume symmetry here aij = aji
+  for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); ++row) {
+    MT aii    = STS::magnitude(D[row]);
+    MT rowsum = AbsRs[row];
+
+    if (aii >= kappa_init * rowsum) {
+      *out << "Flagging index " << row << " as dirichlet "
+                                          "aii >= kappa*rowsum = "
+           << aii << " >= " << kappa_init << " " << rowsum << std::endl;
+      aggStat[row] = IGNORED;
+      --numNonAggregatedNodes;
+      ++numDirichletNodes;
     }
+  }
 
-
-    // 2 : Iteration
-    LO aggIndex = LO_ZERO;
-    for(LO i = 0; i < numRows; i++) {
-      LO current_idx = orderingVector[i];
-      // If we're aggregated already, skip this guy
-      if(aggStat[current_idx] != READY)
+  // 2 : Iteration
+  LO aggIndex = LO_ZERO;
+  for (LO i = 0; i < numRows; i++) {
+    LO current_idx = orderingVector[i];
+    // If we're aggregated already, skip this guy
+    if (aggStat[current_idx] != READY)
+      continue;
+
+    MT best_mu  = MT_ZERO;
+    LO best_idx = LO_INVALID;
+
+    size_t nnz = A->getNumEntriesInLocalRow(current_idx);
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    A->getLocalRowView(current_idx, indices, vals);
+
+    MT aii = STS::real(D[current_idx]);
+    MT si  = STS::real(S[current_idx]);
+    for (LO colidx = 0; colidx < static_cast<LO>(nnz); colidx++) {
+      // Skip aggregated neighbors, off-rank neighbors, hard zeros and self
+      LO col = indices[colidx];
+      SC val = vals[colidx];
+      if (current_idx == col || col >= numRows || aggStat[col] != READY || val == SC_ZERO)
         continue;
 
-      MT best_mu = MT_ZERO;
-      LO best_idx = LO_INVALID;
-
-      size_t nnz = A->getNumEntriesInLocalRow(current_idx);
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      A->getLocalRowView(current_idx, indices, vals);
-
-      MT aii = STS::real(D[current_idx]);
-      MT si  = STS::real(S[current_idx]);
-      for (LO colidx = 0; colidx < static_cast<LO>(nnz); colidx++) {
-        // Skip aggregated neighbors, off-rank neighbors, hard zeros and self
-        LO col = indices[colidx];
-        SC val = vals[colidx];
-        if(current_idx == col || col >= numRows || aggStat[col] != READY  || val == SC_ZERO)
-          continue;
-
-        MT aij = STS::real(val);
-        MT ajj = STS::real(D[col]);
-        MT sj  =  - STS::real(S[col]);  // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
-        if(aii - si + ajj - sj >= MT_ZERO) {
-          // Modification: We assume symmetry here aij = aji
-          MT mu_top    = MT_TWO / ( MT_ONE / aii + MT_ONE / ajj);
-          MT mu_bottom =  - aij + MT_ONE / ( MT_ONE / (aii - si) + MT_ONE / (ajj - sj) );
-          MT mu = mu_top / mu_bottom;
-
-          // Modification: Explicitly check the tie criterion here
-          if (mu > MT_ZERO && (best_idx == LO_INVALID || mu < best_mu * tie_less ||
-                                    (mu < best_mu*tie_more && orderingVector[col] < orderingVector[best_idx]))) {
-            best_mu  = mu;
-            best_idx = col;
-            *out << "[" << current_idx << "] Column     UPDATED " << col << ": "
-                 << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-                 << " = " << aii - si + ajj - sj<< ", aij = "<<aij << ", mu = " << mu << std::endl;
-          }
-          else {
-          *out << "[" << current_idx << "] Column NOT UPDATED " << col << ": "
+      MT aij = STS::real(val);
+      MT ajj = STS::real(D[col]);
+      MT sj  = -STS::real(S[col]);  // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
+      if (aii - si + ajj - sj >= MT_ZERO) {
+        // Modification: We assume symmetry here aij = aji
+        MT mu_top    = MT_TWO / (MT_ONE / aii + MT_ONE / ajj);
+        MT mu_bottom = -aij + MT_ONE / (MT_ONE / (aii - si) + MT_ONE / (ajj - sj));
+        MT mu        = mu_top / mu_bottom;
+
+        // Modification: Explicitly check the tie criterion here
+        if (mu > MT_ZERO && (best_idx == LO_INVALID || mu < best_mu * tie_less ||
+                             (mu < best_mu * tie_more && orderingVector[col] < orderingVector[best_idx]))) {
+          best_mu  = mu;
+          best_idx = col;
+          *out << "[" << current_idx << "] Column     UPDATED " << col << ": "
                << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-               << " = " << aii - si + ajj - sj << ", aij = "<<aij<< ", mu = " << mu << std::endl;
-          }
-        }
-        else {
-          *out << "[" << current_idx << "] Column     FAILED " << col << ": "
+               << " = " << aii - si + ajj - sj << ", aij = " << aij << ", mu = " << mu << std::endl;
+        } else {
+          *out << "[" << current_idx << "] Column NOT UPDATED " << col << ": "
                << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-               << " = " << aii - si + ajj - sj << std::endl;
+               << " = " << aii - si + ajj - sj << ", aij = " << aij << ", mu = " << mu << std::endl;
         }
-      }// end column for loop
+      } else {
+        *out << "[" << current_idx << "] Column     FAILED " << col << ": "
+             << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
+             << " = " << aii - si + ajj - sj << std::endl;
+      }
+    }  // end column for loop
+
+    if (best_idx == LO_INVALID) {
+      *out << "No node buddy found for index " << current_idx
+           << " [agg " << aggIndex << "]\n"
+           << std::endl;
+      // We found no potential node-buddy, so let's just make this a singleton
+      // NOTE: The behavior of what to do if you have no un-aggregated neighbors is not specified in
+      // the paper
+
+      aggStat[current_idx]      = ONEPT;
+      vertex2AggId[current_idx] = aggIndex;
+      procWinner[current_idx]   = myRank;
+      numNonAggregatedNodes--;
+      aggIndex++;
 
-      if(best_idx == LO_INVALID) {
-        *out << "No node buddy found for index " << current_idx
-             << " [agg " << aggIndex << "]\n" << std::endl;
-        // We found no potential node-buddy, so let's just make this a singleton
-        // NOTE: The behavior of what to do if you have no un-aggregated neighbors is not specified in
-        // the paper
+    } else {
+      // We have a buddy, so aggregate, either as a singleton or as a pair, depending on mu
+      if (best_mu <= kappa) {
+        *out << "Node buddies (" << current_idx << "," << best_idx << ") [agg " << aggIndex << "]" << std::endl;
 
-        aggStat[current_idx] = ONEPT;
+        aggStat[current_idx]      = AGGREGATED;
+        aggStat[best_idx]         = AGGREGATED;
         vertex2AggId[current_idx] = aggIndex;
+        vertex2AggId[best_idx]    = aggIndex;
         procWinner[current_idx]   = myRank;
-        numNonAggregatedNodes--;
+        procWinner[best_idx]      = myRank;
+        numNonAggregatedNodes -= 2;
         aggIndex++;
 
       } else {
-        // We have a buddy, so aggregate, either as a singleton or as a pair, depending on mu
-        if(best_mu <= kappa) {
-          *out << "Node buddies (" << current_idx << "," << best_idx << ") [agg " << aggIndex << "]" << std::endl;
-
-          aggStat[current_idx] = AGGREGATED;
-          aggStat[best_idx]    = AGGREGATED;
-          vertex2AggId[current_idx] = aggIndex;
-          vertex2AggId[best_idx]    = aggIndex;
-          procWinner[current_idx]   = myRank;
-          procWinner[best_idx]      = myRank;
-          numNonAggregatedNodes-=2;
-          aggIndex++;
+        *out << "No buddy found for index " << current_idx << ","
+                                                              " but aggregating as singleton [agg "
+             << aggIndex << "]" << std::endl;
 
-        } else {
-          *out << "No buddy found for index " << current_idx << ","
-            " but aggregating as singleton [agg " << aggIndex << "]" << std::endl;
-
-          aggStat[current_idx] = ONEPT;
-          vertex2AggId[current_idx] = aggIndex;
-          procWinner[current_idx]   = myRank;
-          numNonAggregatedNodes--;
-          aggIndex++;
-        } // best_mu
-      } // best_idx
-    }// end Algorithm 4.2
-
-    *out << "vertex2aggid :";
-    for(int i = 0; i < static_cast<int>(vertex2AggId.size()); ++i) {
-      *out << i << "(" << vertex2AggId[i] << ")";
-    }
-    *out << std::endl;
-
-    // update aggregate object
-    aggregates.SetNumAggregates(aggIndex);
-  } // BuildInitialAggregates
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildFurtherAggregates(const Teuchos::ParameterList& params,
-                         const RCP<const Matrix>& A,
-                         const Teuchos::ArrayView<const LO> & orderingVector,
-                         const typename Matrix::local_matrix_type& coarseA,
-                         const typename Teuchos::ScalarTraits<Scalar>::magnitudeType kappa,
-                         const Kokkos::View<typename Kokkos::ArithTraits<Scalar>::val_type*,
-                                            Kokkos::LayoutLeft,
-                                            typename Matrix::local_matrix_type::device_type>& rowSum,
-                         std::vector<LocalOrdinal>& localAggStat,
-                         Teuchos::Array<LocalOrdinal>& localVertex2AggID,
-                         LO& numLocalAggregates,
-                         LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildFurtherAggregates");
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
+        aggStat[current_idx]      = ONEPT;
+        vertex2AggId[current_idx] = aggIndex;
+        procWinner[current_idx]   = myRank;
+        numNonAggregatedNodes--;
+        aggIndex++;
+      }  // best_mu
+    }    // best_idx
+  }      // end Algorithm 4.2
 
-    using value_type = typename local_matrix_type::value_type;
-    const value_type KAT_zero = Kokkos::ArithTraits<value_type>::zero();
-    const magnitude_type MT_zero = Teuchos::ScalarTraits<magnitude_type>::zero();
-    const magnitude_type MT_one  = Teuchos::ScalarTraits<magnitude_type>::one();
-    const magnitude_type MT_two  = MT_one + MT_one;
-    const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid() ;
-
-    // For finding "ties" where we fall back to the ordering.  Making this larger than
-    // hard zero substantially increases code robustness.
-    double tie_criterion = params.get<double>("aggregation: pairwise: tie threshold");
-    double tie_less = 1.0 - tie_criterion;
-    double tie_more = 1.0 + tie_criterion;
-
-    typename row_sum_type::HostMirror rowSum_h = Kokkos::create_mirror_view(rowSum);
-    Kokkos::deep_copy(rowSum_h, rowSum);
-
-    // Extracting the diagonal of a KokkosSparse::CrsMatrix
-    // is not currently provided in kokkos-kernels so here
-    // is an ugly way to get that done...
-    const LO numRows = static_cast<LO>(coarseA.numRows());
-    typename local_matrix_type::values_type::HostMirror diagA_h("diagA host", numRows);
-    typename local_matrix_type::row_map_type::HostMirror row_map_h
-      = Kokkos::create_mirror_view(coarseA.graph.row_map);
-    Kokkos::deep_copy(row_map_h, coarseA.graph.row_map);
-    typename local_matrix_type::index_type::HostMirror entries_h
-      = Kokkos::create_mirror_view(coarseA.graph.entries);
-    Kokkos::deep_copy(entries_h, coarseA.graph.entries);
-    typename local_matrix_type::values_type::HostMirror values_h
-      = Kokkos::create_mirror_view(coarseA.values);
-    Kokkos::deep_copy(values_h, coarseA.values);
-    for(LO rowIdx = 0; rowIdx < numRows; ++rowIdx) {
-      for(LO entryIdx = static_cast<LO>(row_map_h(rowIdx));
-          entryIdx < static_cast<LO>(row_map_h(rowIdx + 1));
-          ++entryIdx) {
-        if(rowIdx == static_cast<LO>(entries_h(entryIdx))) {
-          diagA_h(rowIdx) = values_h(entryIdx);
-        }
+  *out << "vertex2aggid :";
+  for (int i = 0; i < static_cast<int>(vertex2AggId.size()); ++i) {
+    *out << i << "(" << vertex2AggId[i] << ")";
+  }
+  *out << std::endl;
+
+  // update aggregate object
+  aggregates.SetNumAggregates(aggIndex);
+}  // BuildInitialAggregates
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildFurtherAggregates(const Teuchos::ParameterList& params,
+                           const RCP<const Matrix>& A,
+                           const Teuchos::ArrayView<const LO>& orderingVector,
+                           const typename Matrix::local_matrix_type& coarseA,
+                           const typename Teuchos::ScalarTraits<Scalar>::magnitudeType kappa,
+                           const Kokkos::View<typename Kokkos::ArithTraits<Scalar>::val_type*,
+                                              Kokkos::LayoutLeft,
+                                              typename Matrix::local_matrix_type::device_type>& rowSum,
+                           std::vector<LocalOrdinal>& localAggStat,
+                           Teuchos::Array<LocalOrdinal>& localVertex2AggID,
+                           LO& numLocalAggregates,
+                           LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildFurtherAggregates");
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  using value_type             = typename local_matrix_type::value_type;
+  const value_type KAT_zero    = Kokkos::ArithTraits<value_type>::zero();
+  const magnitude_type MT_zero = Teuchos::ScalarTraits<magnitude_type>::zero();
+  const magnitude_type MT_one  = Teuchos::ScalarTraits<magnitude_type>::one();
+  const magnitude_type MT_two  = MT_one + MT_one;
+  const LO LO_INVALID          = Teuchos::OrdinalTraits<LO>::invalid();
+
+  // For finding "ties" where we fall back to the ordering.  Making this larger than
+  // hard zero substantially increases code robustness.
+  double tie_criterion = params.get<double>("aggregation: pairwise: tie threshold");
+  double tie_less      = 1.0 - tie_criterion;
+  double tie_more      = 1.0 + tie_criterion;
+
+  typename row_sum_type::HostMirror rowSum_h = Kokkos::create_mirror_view(rowSum);
+  Kokkos::deep_copy(rowSum_h, rowSum);
+
+  // Extracting the diagonal of a KokkosSparse::CrsMatrix
+  // is not currently provided in kokkos-kernels so here
+  // is an ugly way to get that done...
+  const LO numRows = static_cast<LO>(coarseA.numRows());
+  typename local_matrix_type::values_type::HostMirror diagA_h("diagA host", numRows);
+  typename local_matrix_type::row_map_type::HostMirror row_map_h = Kokkos::create_mirror_view(coarseA.graph.row_map);
+  Kokkos::deep_copy(row_map_h, coarseA.graph.row_map);
+  typename local_matrix_type::index_type::HostMirror entries_h = Kokkos::create_mirror_view(coarseA.graph.entries);
+  Kokkos::deep_copy(entries_h, coarseA.graph.entries);
+  typename local_matrix_type::values_type::HostMirror values_h = Kokkos::create_mirror_view(coarseA.values);
+  Kokkos::deep_copy(values_h, coarseA.values);
+  for (LO rowIdx = 0; rowIdx < numRows; ++rowIdx) {
+    for (LO entryIdx = static_cast<LO>(row_map_h(rowIdx));
+         entryIdx < static_cast<LO>(row_map_h(rowIdx + 1));
+         ++entryIdx) {
+      if (rowIdx == static_cast<LO>(entries_h(entryIdx))) {
+        diagA_h(rowIdx) = values_h(entryIdx);
       }
     }
+  }
+
+  for (LO currentIdx = 0; currentIdx < numRows; ++currentIdx) {
+    if (localAggStat[currentIdx] != READY) {
+      continue;
+    }
 
-    for(LO currentIdx = 0; currentIdx < numRows; ++currentIdx) {
-      if(localAggStat[currentIdx] != READY) {
+    LO bestIdx               = Teuchos::OrdinalTraits<LO>::invalid();
+    magnitude_type best_mu   = Teuchos::ScalarTraits<magnitude_type>::zero();
+    const magnitude_type aii = Teuchos::ScalarTraits<value_type>::real(diagA_h(currentIdx));
+    const magnitude_type si  = Teuchos::ScalarTraits<value_type>::real(rowSum_h(currentIdx));
+    for (auto entryIdx = row_map_h(currentIdx); entryIdx < row_map_h(currentIdx + 1); ++entryIdx) {
+      const LO colIdx = static_cast<LO>(entries_h(entryIdx));
+      if (currentIdx == colIdx || colIdx >= numRows || localAggStat[colIdx] != READY || values_h(entryIdx) == KAT_zero) {
         continue;
       }
 
-      LO bestIdx = Teuchos::OrdinalTraits<LO>::invalid();
-      magnitude_type best_mu = Teuchos::ScalarTraits<magnitude_type>::zero();
-      const magnitude_type aii     = Teuchos::ScalarTraits<value_type>::real(diagA_h(currentIdx));
-      const magnitude_type si      = Teuchos::ScalarTraits<value_type>::real(rowSum_h(currentIdx));
-      for(auto entryIdx = row_map_h(currentIdx); entryIdx < row_map_h(currentIdx + 1); ++entryIdx) {
-        const LO colIdx = static_cast<LO>(entries_h(entryIdx));
-        if(currentIdx == colIdx || colIdx >= numRows || localAggStat[colIdx] != READY || values_h(entryIdx) == KAT_zero) {
-          continue;
-        }
-
-        const magnitude_type aij = Teuchos::ScalarTraits<value_type>::real(values_h(entryIdx));
-        const magnitude_type ajj = Teuchos::ScalarTraits<value_type>::real(diagA_h(colIdx));
-        const magnitude_type sj  = - Teuchos::ScalarTraits<value_type>::real(rowSum_h(colIdx)); // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
-        if(aii - si + ajj - sj >= MT_zero) {
-          const magnitude_type mu_top    = MT_two / ( MT_one/aii + MT_one/ajj );
-          const magnitude_type mu_bottom = -aij + MT_one / (MT_one / (aii - si) + MT_one / (ajj - sj));
-          const magnitude_type mu = mu_top / mu_bottom;
-
-          // Modification: Explicitly check the tie criterion here
-	  if (mu > MT_zero  && (bestIdx == LO_INVALID ||  mu < best_mu * tie_less ||
-                                (mu < best_mu*tie_more && orderingVector[colIdx] < orderingVector[bestIdx]))) {
-            best_mu = mu;
-            bestIdx = colIdx;
-            *out << "[" << currentIdx << "] Column     UPDATED " << colIdx << ": "
-                 << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-                 << " = " << aii - si + ajj - sj << ", aij = "<<aij<<" mu = " << mu << std::endl;
-          }
-          else {
-            *out << "[" << currentIdx << "] Column NOT UPDATED " << colIdx << ": "
-                 << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-                 << " = " << aii - si + ajj - sj << ", aij = "<<aij<<", mu = " << mu << std::endl;
-          }
+      const magnitude_type aij = Teuchos::ScalarTraits<value_type>::real(values_h(entryIdx));
+      const magnitude_type ajj = Teuchos::ScalarTraits<value_type>::real(diagA_h(colIdx));
+      const magnitude_type sj  = -Teuchos::ScalarTraits<value_type>::real(rowSum_h(colIdx));  // NOTE: The ghostedRowSum vector here has has the sign flipped from Notay's S
+      if (aii - si + ajj - sj >= MT_zero) {
+        const magnitude_type mu_top    = MT_two / (MT_one / aii + MT_one / ajj);
+        const magnitude_type mu_bottom = -aij + MT_one / (MT_one / (aii - si) + MT_one / (ajj - sj));
+        const magnitude_type mu        = mu_top / mu_bottom;
+
+        // Modification: Explicitly check the tie criterion here
+        if (mu > MT_zero && (bestIdx == LO_INVALID || mu < best_mu * tie_less ||
+                             (mu < best_mu * tie_more && orderingVector[colIdx] < orderingVector[bestIdx]))) {
+          best_mu = mu;
+          bestIdx = colIdx;
+          *out << "[" << currentIdx << "] Column     UPDATED " << colIdx << ": "
+               << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
+               << " = " << aii - si + ajj - sj << ", aij = " << aij << " mu = " << mu << std::endl;
         } else {
-            *out << "[" << currentIdx << "] Column      FAILED " << colIdx << ": "
-                 << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
-                 << " = " << aii - si + ajj - sj << std::endl;
+          *out << "[" << currentIdx << "] Column NOT UPDATED " << colIdx << ": "
+               << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
+               << " = " << aii - si + ajj - sj << ", aij = " << aij << ", mu = " << mu << std::endl;
         }
-      } // end loop over row entries
+      } else {
+        *out << "[" << currentIdx << "] Column      FAILED " << colIdx << ": "
+             << "aii - si + ajj - sj = " << aii << " - " << si << " + " << ajj << " - " << sj
+             << " = " << aii - si + ajj - sj << std::endl;
+      }
+    }  // end loop over row entries
 
-      if(bestIdx == Teuchos::OrdinalTraits<LO>::invalid()) {
-        localAggStat[currentIdx]      = ONEPT;
+    if (bestIdx == Teuchos::OrdinalTraits<LO>::invalid()) {
+      localAggStat[currentIdx]      = ONEPT;
+      localVertex2AggID[currentIdx] = numLocalAggregates;
+      --numNonAggregatedNodes;
+      ++numLocalAggregates;
+    } else {
+      if (best_mu <= kappa) {
+        *out << "Node buddies (" << currentIdx << "," << bestIdx << ") [agg " << numLocalAggregates << "]" << std::endl;
+
+        localAggStat[currentIdx]      = AGGREGATED;
         localVertex2AggID[currentIdx] = numLocalAggregates;
         --numNonAggregatedNodes;
-        ++numLocalAggregates;
-      } else {
-        if(best_mu <= kappa) {
-          *out << "Node buddies (" << currentIdx << "," << bestIdx << ") [agg " << numLocalAggregates << "]" << std::endl;
-
-          localAggStat[currentIdx]      = AGGREGATED;
-          localVertex2AggID[currentIdx] = numLocalAggregates;
-          --numNonAggregatedNodes;
 
-          localAggStat[bestIdx]         = AGGREGATED;
-          localVertex2AggID[bestIdx]    = numLocalAggregates;
-          --numNonAggregatedNodes;
+        localAggStat[bestIdx]      = AGGREGATED;
+        localVertex2AggID[bestIdx] = numLocalAggregates;
+        --numNonAggregatedNodes;
 
-          ++numLocalAggregates;
-        } else {
-          *out << "No buddy found for index " << currentIdx << ","
-            " but aggregating as singleton [agg " << numLocalAggregates << "]" << std::endl;
+        ++numLocalAggregates;
+      } else {
+        *out << "No buddy found for index " << currentIdx << ","
+                                                             " but aggregating as singleton [agg "
+             << numLocalAggregates << "]" << std::endl;
 
-          localAggStat[currentIdx]      = ONEPT;
-          localVertex2AggID[currentIdx] = numLocalAggregates;
-          --numNonAggregatedNodes;
-          ++numLocalAggregates;
-        }
+        localAggStat[currentIdx]      = ONEPT;
+        localVertex2AggID[currentIdx] = numLocalAggregates;
+        --numNonAggregatedNodes;
+        ++numLocalAggregates;
       }
-    } // end loop over matrix rows
-
-  } // BuildFurtherAggregates
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildOnRankLocalMatrix(const typename Matrix::local_matrix_type& localA,
-                         typename Matrix::local_matrix_type& onrankA) const {
-    Monitor m(*this, "BuildOnRankLocalMatrix");
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
     }
+  }  // end loop over matrix rows
+
+}  // BuildFurtherAggregates
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildOnRankLocalMatrix(const typename Matrix::local_matrix_type& localA,
+                           typename Matrix::local_matrix_type& onrankA) const {
+  Monitor m(*this, "BuildOnRankLocalMatrix");
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using values_type       = typename local_matrix_type::values_type;
-    using size_type         = typename local_graph_type::size_type;
-    using col_index_type    = typename local_graph_type::data_type;
-    using array_layout      = typename local_graph_type::array_layout;
-    using memory_traits     = typename local_graph_type::memory_traits;
-    using row_pointer_type  = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
-    using col_indices_type  = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
-    // Extract on rank part of A
-    // Simply check that the column index is less than the number of local rows
-    // otherwise remove it.
-
-    const int numRows = static_cast<int>(localA.numRows());
-    row_pointer_type rowPtr("onrankA row pointer", numRows + 1);
-    typename row_pointer_type::HostMirror rowPtr_h = Kokkos::create_mirror_view(rowPtr);
-    typename local_graph_type::row_map_type::HostMirror origRowPtr_h
-      = Kokkos::create_mirror_view(localA.graph.row_map);
-    typename local_graph_type::entries_type::HostMirror origColind_h
-      = Kokkos::create_mirror_view(localA.graph.entries);
-    typename values_type::HostMirror origValues_h
-      = Kokkos::create_mirror_view(localA.values);
-    Kokkos::deep_copy(origRowPtr_h, localA.graph.row_map);
-    Kokkos::deep_copy(origColind_h, localA.graph.entries);
-    Kokkos::deep_copy(origValues_h, localA.values);
-
-    // Compute the number of nnz entries per row
-    rowPtr_h(0) = 0;
-    for(int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
-      for(size_type entryIdx = origRowPtr_h(rowIdx); entryIdx < origRowPtr_h(rowIdx + 1); ++entryIdx) {
-        if(origColind_h(entryIdx) < numRows) {rowPtr_h(rowIdx + 1) += 1;}
+  using local_graph_type = typename local_matrix_type::staticcrsgraph_type;
+  using values_type      = typename local_matrix_type::values_type;
+  using size_type        = typename local_graph_type::size_type;
+  using col_index_type   = typename local_graph_type::data_type;
+  using array_layout     = typename local_graph_type::array_layout;
+  using memory_traits    = typename local_graph_type::memory_traits;
+  using row_pointer_type = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
+  using col_indices_type = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
+  // Extract on rank part of A
+  // Simply check that the column index is less than the number of local rows
+  // otherwise remove it.
+
+  const int numRows = static_cast<int>(localA.numRows());
+  row_pointer_type rowPtr("onrankA row pointer", numRows + 1);
+  typename row_pointer_type::HostMirror rowPtr_h                   = Kokkos::create_mirror_view(rowPtr);
+  typename local_graph_type::row_map_type::HostMirror origRowPtr_h = Kokkos::create_mirror_view(localA.graph.row_map);
+  typename local_graph_type::entries_type::HostMirror origColind_h = Kokkos::create_mirror_view(localA.graph.entries);
+  typename values_type::HostMirror origValues_h                    = Kokkos::create_mirror_view(localA.values);
+  Kokkos::deep_copy(origRowPtr_h, localA.graph.row_map);
+  Kokkos::deep_copy(origColind_h, localA.graph.entries);
+  Kokkos::deep_copy(origValues_h, localA.values);
+
+  // Compute the number of nnz entries per row
+  rowPtr_h(0) = 0;
+  for (int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
+    for (size_type entryIdx = origRowPtr_h(rowIdx); entryIdx < origRowPtr_h(rowIdx + 1); ++entryIdx) {
+      if (origColind_h(entryIdx) < numRows) {
+        rowPtr_h(rowIdx + 1) += 1;
       }
-      rowPtr_h(rowIdx + 1) = rowPtr_h(rowIdx + 1) + rowPtr_h(rowIdx);
     }
-    Kokkos::deep_copy(rowPtr, rowPtr_h);
-
-    const LO nnzOnrankA = rowPtr_h(numRows);
-
-    // Now use nnz per row to allocate matrix views and store column indices and values
-    col_indices_type colInd("onrankA column indices", rowPtr_h(numRows));
-    values_type      values("onrankA values", rowPtr_h(numRows));
-    typename col_indices_type::HostMirror colInd_h = Kokkos::create_mirror_view(colInd);
-    typename values_type::HostMirror      values_h = Kokkos::create_mirror_view(values);
-    int entriesInRow;
-    for(int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
-      entriesInRow = 0;
-      for(size_type entryIdx = origRowPtr_h(rowIdx); entryIdx < origRowPtr_h(rowIdx + 1); ++entryIdx) {
-        if(origColind_h(entryIdx) < numRows) {
-          colInd_h(rowPtr_h(rowIdx) + entriesInRow) = origColind_h(entryIdx);
-          values_h(rowPtr_h(rowIdx) + entriesInRow) = origValues_h(entryIdx);
-          ++entriesInRow;
-        }
+    rowPtr_h(rowIdx + 1) = rowPtr_h(rowIdx + 1) + rowPtr_h(rowIdx);
+  }
+  Kokkos::deep_copy(rowPtr, rowPtr_h);
+
+  const LO nnzOnrankA = rowPtr_h(numRows);
+
+  // Now use nnz per row to allocate matrix views and store column indices and values
+  col_indices_type colInd("onrankA column indices", rowPtr_h(numRows));
+  values_type values("onrankA values", rowPtr_h(numRows));
+  typename col_indices_type::HostMirror colInd_h = Kokkos::create_mirror_view(colInd);
+  typename values_type::HostMirror values_h      = Kokkos::create_mirror_view(values);
+  int entriesInRow;
+  for (int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
+    entriesInRow = 0;
+    for (size_type entryIdx = origRowPtr_h(rowIdx); entryIdx < origRowPtr_h(rowIdx + 1); ++entryIdx) {
+      if (origColind_h(entryIdx) < numRows) {
+        colInd_h(rowPtr_h(rowIdx) + entriesInRow) = origColind_h(entryIdx);
+        values_h(rowPtr_h(rowIdx) + entriesInRow) = origValues_h(entryIdx);
+        ++entriesInRow;
       }
     }
-    Kokkos::deep_copy(colInd, colInd_h);
-    Kokkos::deep_copy(values, values_h);
-
-    onrankA = local_matrix_type("onrankA", numRows, numRows,
-                                nnzOnrankA, values, rowPtr, colInd);
+  }
+  Kokkos::deep_copy(colInd, colInd_h);
+  Kokkos::deep_copy(values, values_h);
+
+  onrankA = local_matrix_type("onrankA", numRows, numRows,
+                              nnzOnrankA, values, rowPtr, colInd);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildIntermediateProlongator(const LocalOrdinal numRows,
+                                 const LocalOrdinal numDirichletNodes,
+                                 const LocalOrdinal numLocalAggregates,
+                                 const Teuchos::ArrayView<const LocalOrdinal>& localVertex2AggID,
+                                 typename Matrix::local_matrix_type& intermediateP) const {
+  Monitor m(*this, "BuildIntermediateProlongator");
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildIntermediateProlongator(const LocalOrdinal numRows,
-                               const LocalOrdinal numDirichletNodes,
-                               const LocalOrdinal numLocalAggregates,
-                               const Teuchos::ArrayView<const LocalOrdinal>& localVertex2AggID,
-                               typename Matrix::local_matrix_type& intermediateP) const {
-    Monitor m(*this, "BuildIntermediateProlongator");
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_PAIRWISEAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
+  using local_graph_type = typename local_matrix_type::staticcrsgraph_type;
+  using values_type      = typename local_matrix_type::values_type;
+  using size_type        = typename local_graph_type::size_type;
+  using col_index_type   = typename local_graph_type::data_type;
+  using array_layout     = typename local_graph_type::array_layout;
+  using memory_traits    = typename local_graph_type::memory_traits;
+  using row_pointer_type = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
+  using col_indices_type = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
+
+  const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  const int intermediatePnnz = numRows - numDirichletNodes;
+  row_pointer_type rowPtr("intermediateP row pointer", numRows + 1);
+  col_indices_type colInd("intermediateP column indices", intermediatePnnz);
+  values_type values("intermediateP values", intermediatePnnz);
+  typename row_pointer_type::HostMirror rowPtr_h = Kokkos::create_mirror_view(rowPtr);
+  typename col_indices_type::HostMirror colInd_h = Kokkos::create_mirror_view(colInd);
+
+  rowPtr_h(0) = 0;
+  for (int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
+    // Skip Dirichlet nodes
+    if (localVertex2AggID[rowIdx] == LO_INVALID) {
+      rowPtr_h(rowIdx + 1) = rowPtr_h(rowIdx);
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using values_type       = typename local_matrix_type::values_type;
-    using size_type         = typename local_graph_type::size_type;
-    using col_index_type    = typename local_graph_type::data_type;
-    using array_layout      = typename local_graph_type::array_layout;
-    using memory_traits     = typename local_graph_type::memory_traits;
-    using row_pointer_type  = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
-    using col_indices_type  = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
-
-    const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-
-    const int intermediatePnnz = numRows - numDirichletNodes;
-    row_pointer_type rowPtr("intermediateP row pointer", numRows + 1);
-    col_indices_type colInd("intermediateP column indices", intermediatePnnz);
-    values_type      values("intermediateP values", intermediatePnnz);
-    typename row_pointer_type::HostMirror rowPtr_h = Kokkos::create_mirror_view(rowPtr);
-    typename col_indices_type::HostMirror colInd_h = Kokkos::create_mirror_view(colInd);
-
-    rowPtr_h(0) = 0;
-    for(int rowIdx = 0; rowIdx < numRows; ++rowIdx) {
-      // Skip Dirichlet nodes
-      if(localVertex2AggID[rowIdx] == LO_INVALID) {
-        rowPtr_h(rowIdx + 1) = rowPtr_h(rowIdx);
-      } else {
-        rowPtr_h(rowIdx + 1) = rowPtr_h(rowIdx) + 1;
-        colInd_h(rowPtr_h(rowIdx)) = localVertex2AggID[rowIdx];
-      }
+      rowPtr_h(rowIdx + 1)       = rowPtr_h(rowIdx) + 1;
+      colInd_h(rowPtr_h(rowIdx)) = localVertex2AggID[rowIdx];
     }
+  }
 
-    Kokkos::deep_copy(rowPtr, rowPtr_h);
-    Kokkos::deep_copy(colInd, colInd_h);
-    Kokkos::deep_copy(values, Kokkos::ArithTraits<typename values_type::value_type>::one());
-
-    intermediateP = local_matrix_type("intermediateP",
-                                      numRows, numLocalAggregates, intermediatePnnz,
-                                      values, rowPtr, colInd);
-  } // BuildIntermediateProlongator
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildCoarseLocalMatrix(const typename Matrix::local_matrix_type& intermediateP,
-                         typename Matrix::local_matrix_type& coarseA) const {
-    Monitor m(*this, "BuildCoarseLocalMatrix");
-
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using values_type       = typename local_matrix_type::values_type;
-    using size_type         = typename local_graph_type::size_type;
-    using col_index_type    = typename local_graph_type::data_type;
-    using array_layout      = typename local_graph_type::array_layout;
-    using memory_traits     = typename local_graph_type::memory_traits;
-    using row_pointer_type  = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
-    using col_indices_type  = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
-
-    local_matrix_type AP;
-    localSpGEMM(coarseA, intermediateP, "AP", AP);
-
-    // Note 03/11/20, lbv: does kh need to destroy and recreate the spgemm handle
-    // I am not sure but doing it for safety in case it stashes data from the previous
-    // spgemm computation...
-
-    // Compute Ac = Pt * AP
-    // Two steps needed:
-    //   1. compute Pt
-    //   2. perform multiplication
-
-    // Step 1 compute Pt
-    // Obviously this requires the same amount of storage as P except for the rowPtr
-    row_pointer_type rowPtrPt(Kokkos::ViewAllocateWithoutInitializing("Pt row pointer"),
-                              intermediateP.numCols() + 1);
-    col_indices_type colIndPt(Kokkos::ViewAllocateWithoutInitializing("Pt column indices"),
-                              intermediateP.nnz());
-    values_type      valuesPt(Kokkos::ViewAllocateWithoutInitializing("Pt values"),
-                              intermediateP.nnz());
-
-    typename row_pointer_type::HostMirror rowPtrPt_h = Kokkos::create_mirror_view(rowPtrPt);
-    typename col_indices_type::HostMirror entries_h = Kokkos::create_mirror_view(intermediateP.graph.entries);
-    Kokkos::deep_copy(entries_h, intermediateP.graph.entries);
-    Kokkos::deep_copy(rowPtrPt_h, 0);
-    for(size_type entryIdx = 0; entryIdx < intermediateP.nnz(); ++entryIdx) {
-      rowPtrPt_h(entries_h(entryIdx) + 1) += 1;
-    }
-    for(LO rowIdx = 0; rowIdx < intermediateP.numCols(); ++rowIdx) {
-      rowPtrPt_h(rowIdx + 1) += rowPtrPt_h(rowIdx);
-    }
-    Kokkos::deep_copy(rowPtrPt, rowPtrPt_h);
-
-    typename row_pointer_type::HostMirror rowPtrP_h = Kokkos::create_mirror_view(intermediateP.graph.row_map);
-    Kokkos::deep_copy(rowPtrP_h, intermediateP.graph.row_map);
-    typename col_indices_type::HostMirror colIndP_h = Kokkos::create_mirror_view(intermediateP.graph.entries);
-    Kokkos::deep_copy(colIndP_h, intermediateP.graph.entries);
-    typename values_type::HostMirror valuesP_h  = Kokkos::create_mirror_view(intermediateP.values);
-    Kokkos::deep_copy(valuesP_h, intermediateP.values);
-    typename col_indices_type::HostMirror colIndPt_h = Kokkos::create_mirror_view(colIndPt);
-    typename values_type::HostMirror valuesPt_h = Kokkos::create_mirror_view(valuesPt);
-    const col_index_type invalidColumnIndex = KokkosSparse::OrdinalTraits<col_index_type>::invalid();
-    Kokkos::deep_copy(colIndPt_h, invalidColumnIndex);
-
-    col_index_type colIdx = 0;
-    for(LO rowIdx = 0; rowIdx < intermediateP.numRows(); ++rowIdx) {
-      for(size_type entryIdxP = rowPtrP_h(rowIdx); entryIdxP < rowPtrP_h(rowIdx + 1); ++entryIdxP) {
-        colIdx = entries_h(entryIdxP);
-        for(size_type entryIdxPt = rowPtrPt_h(colIdx); entryIdxPt < rowPtrPt_h(colIdx + 1); ++entryIdxPt) {
-          if(colIndPt_h(entryIdxPt) == invalidColumnIndex) {
-            colIndPt_h(entryIdxPt) = rowIdx;
-            valuesPt_h(entryIdxPt) = valuesP_h(entryIdxP);
-            break;
-          }
-        } // Loop over entries in row of Pt
-      } // Loop over entries in row of P
-    } // Loop over rows of P
-
-    Kokkos::deep_copy(colIndPt, colIndPt_h);
-    Kokkos::deep_copy(valuesPt, valuesPt_h);
-
-
-    local_matrix_type intermediatePt("intermediatePt",
-                                     intermediateP.numCols(),
-                                     intermediateP.numRows(),
-                                     intermediateP.nnz(),
-                                     valuesPt, rowPtrPt, colIndPt);
-
-    // Create views for coarseA matrix
-    localSpGEMM(intermediatePt, AP, "coarseA", coarseA);
-  } // BuildCoarseLocalMatrix
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  localSpGEMM(const typename Matrix::local_matrix_type& A,
-              const typename Matrix::local_matrix_type& B,
-              const std::string matrixLabel,
-              typename Matrix::local_matrix_type& C) const {
-
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using values_type       = typename local_matrix_type::values_type;
-    using size_type         = typename local_graph_type::size_type;
-    using col_index_type    = typename local_graph_type::data_type;
-    using array_layout      = typename local_graph_type::array_layout;
-    using memory_space      = typename device_type::memory_space;
-    using memory_traits     = typename local_graph_type::memory_traits;
-    using row_pointer_type  = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
-    using col_indices_type  = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
-
-    // Options
-    int team_work_size = 16;
-    std::string myalg("SPGEMM_KK_MEMORY");
-    KokkosSparse::SPGEMMAlgorithm alg_enum = KokkosSparse::StringToSPGEMMAlgorithm(myalg);
-    KokkosKernels::Experimental::KokkosKernelsHandle<typename row_pointer_type::const_value_type,
-                                                     typename col_indices_type::const_value_type,
-                                                     typename values_type::const_value_type,
-                                                     execution_space,
-                                                     memory_space,
-                                                     memory_space> kh;
-    kh.create_spgemm_handle(alg_enum);
-    kh.set_team_work_size(team_work_size);
-
-    // Create views for AP matrix
-    row_pointer_type rowPtrC(Kokkos::ViewAllocateWithoutInitializing("C row pointer"),
-                             A.numRows() + 1);
-    col_indices_type colIndC;
-    values_type      valuesC;
-
-    // Symbolic multiplication
-    KokkosSparse::Experimental::spgemm_symbolic(&kh, A.numRows(),
-                                                B.numRows(), B.numCols(),
-                                                A.graph.row_map, A.graph.entries, false,
-                                                B.graph.row_map, B.graph.entries, false,
-                                                rowPtrC);
-
-    // allocate column indices and values of AP
-    size_t nnzC = kh.get_spgemm_handle()->get_c_nnz();
-    if (nnzC) {
-      colIndC = col_indices_type(Kokkos::ViewAllocateWithoutInitializing("C column inds"), nnzC);
-      valuesC = values_type(Kokkos::ViewAllocateWithoutInitializing("C values"), nnzC);
-    }
+  Kokkos::deep_copy(rowPtr, rowPtr_h);
+  Kokkos::deep_copy(colInd, colInd_h);
+  Kokkos::deep_copy(values, Kokkos::ArithTraits<typename values_type::value_type>::one());
+
+  intermediateP = local_matrix_type("intermediateP",
+                                    numRows, numLocalAggregates, intermediatePnnz,
+                                    values, rowPtr, colInd);
+}  // BuildIntermediateProlongator
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildCoarseLocalMatrix(const typename Matrix::local_matrix_type& intermediateP,
+                           typename Matrix::local_matrix_type& coarseA) const {
+  Monitor m(*this, "BuildCoarseLocalMatrix");
+
+  using local_graph_type = typename local_matrix_type::staticcrsgraph_type;
+  using values_type      = typename local_matrix_type::values_type;
+  using size_type        = typename local_graph_type::size_type;
+  using col_index_type   = typename local_graph_type::data_type;
+  using array_layout     = typename local_graph_type::array_layout;
+  using memory_traits    = typename local_graph_type::memory_traits;
+  using row_pointer_type = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
+  using col_indices_type = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
+
+  local_matrix_type AP;
+  localSpGEMM(coarseA, intermediateP, "AP", AP);
+
+  // Note 03/11/20, lbv: does kh need to destroy and recreate the spgemm handle
+  // I am not sure but doing it for safety in case it stashes data from the previous
+  // spgemm computation...
+
+  // Compute Ac = Pt * AP
+  // Two steps needed:
+  //   1. compute Pt
+  //   2. perform multiplication
+
+  // Step 1 compute Pt
+  // Obviously this requires the same amount of storage as P except for the rowPtr
+  row_pointer_type rowPtrPt(Kokkos::ViewAllocateWithoutInitializing("Pt row pointer"),
+                            intermediateP.numCols() + 1);
+  col_indices_type colIndPt(Kokkos::ViewAllocateWithoutInitializing("Pt column indices"),
+                            intermediateP.nnz());
+  values_type valuesPt(Kokkos::ViewAllocateWithoutInitializing("Pt values"),
+                       intermediateP.nnz());
+
+  typename row_pointer_type::HostMirror rowPtrPt_h = Kokkos::create_mirror_view(rowPtrPt);
+  typename col_indices_type::HostMirror entries_h  = Kokkos::create_mirror_view(intermediateP.graph.entries);
+  Kokkos::deep_copy(entries_h, intermediateP.graph.entries);
+  Kokkos::deep_copy(rowPtrPt_h, 0);
+  for (size_type entryIdx = 0; entryIdx < intermediateP.nnz(); ++entryIdx) {
+    rowPtrPt_h(entries_h(entryIdx) + 1) += 1;
+  }
+  for (LO rowIdx = 0; rowIdx < intermediateP.numCols(); ++rowIdx) {
+    rowPtrPt_h(rowIdx + 1) += rowPtrPt_h(rowIdx);
+  }
+  Kokkos::deep_copy(rowPtrPt, rowPtrPt_h);
+
+  typename row_pointer_type::HostMirror rowPtrP_h = Kokkos::create_mirror_view(intermediateP.graph.row_map);
+  Kokkos::deep_copy(rowPtrP_h, intermediateP.graph.row_map);
+  typename col_indices_type::HostMirror colIndP_h = Kokkos::create_mirror_view(intermediateP.graph.entries);
+  Kokkos::deep_copy(colIndP_h, intermediateP.graph.entries);
+  typename values_type::HostMirror valuesP_h = Kokkos::create_mirror_view(intermediateP.values);
+  Kokkos::deep_copy(valuesP_h, intermediateP.values);
+  typename col_indices_type::HostMirror colIndPt_h = Kokkos::create_mirror_view(colIndPt);
+  typename values_type::HostMirror valuesPt_h      = Kokkos::create_mirror_view(valuesPt);
+  const col_index_type invalidColumnIndex          = KokkosSparse::OrdinalTraits<col_index_type>::invalid();
+  Kokkos::deep_copy(colIndPt_h, invalidColumnIndex);
+
+  col_index_type colIdx = 0;
+  for (LO rowIdx = 0; rowIdx < intermediateP.numRows(); ++rowIdx) {
+    for (size_type entryIdxP = rowPtrP_h(rowIdx); entryIdxP < rowPtrP_h(rowIdx + 1); ++entryIdxP) {
+      colIdx = entries_h(entryIdxP);
+      for (size_type entryIdxPt = rowPtrPt_h(colIdx); entryIdxPt < rowPtrPt_h(colIdx + 1); ++entryIdxPt) {
+        if (colIndPt_h(entryIdxPt) == invalidColumnIndex) {
+          colIndPt_h(entryIdxPt) = rowIdx;
+          valuesPt_h(entryIdxPt) = valuesP_h(entryIdxP);
+          break;
+        }
+      }  // Loop over entries in row of Pt
+    }    // Loop over entries in row of P
+  }      // Loop over rows of P
+
+  Kokkos::deep_copy(colIndPt, colIndPt_h);
+  Kokkos::deep_copy(valuesPt, valuesPt_h);
+
+  local_matrix_type intermediatePt("intermediatePt",
+                                   intermediateP.numCols(),
+                                   intermediateP.numRows(),
+                                   intermediateP.nnz(),
+                                   valuesPt, rowPtrPt, colIndPt);
+
+  // Create views for coarseA matrix
+  localSpGEMM(intermediatePt, AP, "coarseA", coarseA);
+}  // BuildCoarseLocalMatrix
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    localSpGEMM(const typename Matrix::local_matrix_type& A,
+                const typename Matrix::local_matrix_type& B,
+                const std::string matrixLabel,
+                typename Matrix::local_matrix_type& C) const {
+  using local_graph_type = typename local_matrix_type::staticcrsgraph_type;
+  using values_type      = typename local_matrix_type::values_type;
+  using size_type        = typename local_graph_type::size_type;
+  using col_index_type   = typename local_graph_type::data_type;
+  using array_layout     = typename local_graph_type::array_layout;
+  using memory_space     = typename device_type::memory_space;
+  using memory_traits    = typename local_graph_type::memory_traits;
+  using row_pointer_type = Kokkos::View<size_type*, array_layout, device_type, memory_traits>;
+  using col_indices_type = Kokkos::View<col_index_type*, array_layout, device_type, memory_traits>;
+
+  // Options
+  int team_work_size = 16;
+  std::string myalg("SPGEMM_KK_MEMORY");
+  KokkosSparse::SPGEMMAlgorithm alg_enum = KokkosSparse::StringToSPGEMMAlgorithm(myalg);
+  KokkosKernels::Experimental::KokkosKernelsHandle<typename row_pointer_type::const_value_type,
+                                                   typename col_indices_type::const_value_type,
+                                                   typename values_type::const_value_type,
+                                                   execution_space,
+                                                   memory_space,
+                                                   memory_space>
+      kh;
+  kh.create_spgemm_handle(alg_enum);
+  kh.set_team_work_size(team_work_size);
+
+  // Create views for AP matrix
+  row_pointer_type rowPtrC(Kokkos::ViewAllocateWithoutInitializing("C row pointer"),
+                           A.numRows() + 1);
+  col_indices_type colIndC;
+  values_type valuesC;
+
+  // Symbolic multiplication
+  KokkosSparse::Experimental::spgemm_symbolic(&kh, A.numRows(),
+                                              B.numRows(), B.numCols(),
+                                              A.graph.row_map, A.graph.entries, false,
+                                              B.graph.row_map, B.graph.entries, false,
+                                              rowPtrC);
+
+  // allocate column indices and values of AP
+  size_t nnzC = kh.get_spgemm_handle()->get_c_nnz();
+  if (nnzC) {
+    colIndC = col_indices_type(Kokkos::ViewAllocateWithoutInitializing("C column inds"), nnzC);
+    valuesC = values_type(Kokkos::ViewAllocateWithoutInitializing("C values"), nnzC);
+  }
 
-    // Numeric multiplication
-    KokkosSparse::Experimental::spgemm_numeric(&kh, A.numRows(),
-                                               B.numRows(), B.numCols(),
-                                               A.graph.row_map, A.graph.entries, A.values, false,
-                                               B.graph.row_map, B.graph.entries, B.values, false,
-                                               rowPtrC, colIndC, valuesC);
-    kh.destroy_spgemm_handle();
+  // Numeric multiplication
+  KokkosSparse::Experimental::spgemm_numeric(&kh, A.numRows(),
+                                             B.numRows(), B.numCols(),
+                                             A.graph.row_map, A.graph.entries, A.values, false,
+                                             B.graph.row_map, B.graph.entries, B.values, false,
+                                             rowPtrC, colIndC, valuesC);
+  kh.destroy_spgemm_handle();
 
-    C = local_matrix_type(matrixLabel, A.numRows(), B.numCols(), nnzC, valuesC, rowPtrC, colIndC);
+  C = local_matrix_type(matrixLabel, A.numRows(), B.numCols(), nnzC, valuesC, rowPtrC, colIndC);
 
-  } // localSpGEMM
+}  // localSpGEMM
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_NOTAYAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_decl.hpp
index 2b89f58299a9..2cd379401f54 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_decl.hpp
@@ -56,78 +56,74 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationStructuredAlgorithm class.
-    @brief Algorithm for coarsening a graph with structured aggregation.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Use the logical indexing of the mesh to obtain a very regular aggregation pattern and maintain
-    lines and planes of the problem as they might be useful to the smoother.
-    This algorithms is also very easy to parallelize on node due to its very regular and predictible
-    memory access patern.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: coarsen | describe the coarsening rate to be used in each direction
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationStructuredAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationStructuredAlgorithm class.
+  @brief Algorithm for coarsening a graph with structured aggregation.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Use the logical indexing of the mesh to obtain a very regular aggregation pattern and maintain
+  lines and planes of the problem as they might be useful to the smoother.
+  This algorithms is also very easy to parallelize on node due to its very regular and predictible
+  memory access patern.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: coarsen | describe the coarsening rate to be used in each direction
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationStructuredAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationStructuredAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationStructuredAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationStructuredAlgorithm() { }
+  //! Destructor.
+  virtual ~AggregationStructuredAlgorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const Teuchos::ParameterList& params, const GraphBase& graph,
+                       Aggregates& aggregates, std::vector<unsigned>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    void BuildAggregates(const Teuchos::ParameterList& params, const GraphBase& graph,
-                         Aggregates& aggregates, std::vector<unsigned>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildGraph(const GraphBase& graph, RCP<IndexManager>& geoData, const LO dofsPerNode,
+                  RCP<CrsGraph>& myGraph, RCP<const Map>& coarseCoordinatesFineMap,
+                  RCP<const Map>& coarseCoordinatesMap) const;
+  //@}
 
-    void BuildGraph(const GraphBase& graph, RCP<IndexManager>& geoData, const LO dofsPerNode,
-                    RCP<CrsGraph>& myGraph, RCP<const Map>& coarseCoordinatesFineMap,
-                    RCP<const Map>& coarseCoordinatesMap) const;
-    //@}
+  std::string description() const { return "Aggretation: structured algorithm"; }
 
-    std::string description() const { return "Aggretation: structured algorithm"; }
-
-  private:
-
-    void ComputeGraphDataConstant(const GraphBase& graph, RCP<IndexManager>& geoData,
-                                  const LO dofsPerNode, const int numInterpolationPoints,
-                                  ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
-                                  Array<LO>& colIndex) const;
-
-    void ComputeGraphDataLinear(const GraphBase& graph, RCP<IndexManager>& geoData,
+ private:
+  void ComputeGraphDataConstant(const GraphBase& graph, RCP<IndexManager>& geoData,
                                 const LO dofsPerNode, const int numInterpolationPoints,
                                 ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
                                 Array<LO>& colIndex) const;
 
-  };
+  void ComputeGraphDataLinear(const GraphBase& graph, RCP<IndexManager>& geoData,
+                              const LO dofsPerNode, const int numInterpolationPoints,
+                              ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
+                              Array<LO>& colIndex) const;
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_SHORT
 #endif /* MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_def.hpp
index 325b932fb453..96bbd070c19a 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DEF_HPP_
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DEF_HPP_
 
-
 #include <Teuchos_Comm.hpp>
 #include <Teuchos_CommHelpers.hpp>
 
@@ -65,370 +64,361 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const Teuchos::ParameterList& /* params */, const GraphBase& graph,
-                  Aggregates& aggregates, std::vector<unsigned>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const Teuchos::ParameterList& /* params */, const GraphBase& graph,
+                    Aggregates& aggregates, std::vector<unsigned>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  RCP<IndexManager> geoData    = aggregates.GetIndexManager();
+  const bool coupled           = geoData->isAggregationCoupled();
+  const bool singleCoarsePoint = geoData->isSingleCoarsePoint();
+  ArrayRCP<LO> vertex2AggId    = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner      = aggregates.GetProcWinner()->getDataNonConst(0);
+  Array<LO> ghostedCoarseNodeCoarseLIDs;
+  Array<int> ghostedCoarseNodeCoarsePIDs;
+  Array<GO> ghostedCoarseNodeCoarseGIDs;
+
+  *out << "Extract data for ghosted nodes" << std::endl;
+  geoData->getGhostedNodesData(graph.GetDomainMap(), ghostedCoarseNodeCoarseLIDs,
+                               ghostedCoarseNodeCoarsePIDs, ghostedCoarseNodeCoarseGIDs);
+
+  LO rem, rate;
+  Array<LO> ghostedIdx(3), coarseIdx(3);
+  LO ghostedCoarseNodeCoarseLID, aggId;
+  *out << "Loop over fine nodes and assign them to an aggregate and a rank" << std::endl;
+  for (LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
+    // Compute coarse ID associated with fine LID
+    geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
+
+    for (int dim = 0; dim < 3; ++dim) {
+      if (singleCoarsePoint && (geoData->getLocalFineNodesInDir(dim) - 1 < geoData->getCoarseningRate(dim))) {
+        coarseIdx[dim] = 0;
+      } else {
+        coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
+        rem            = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
+        if (ghostedIdx[dim] - geoData->getOffset(dim) < geoData->getLocalFineNodesInDir(dim) - geoData->getCoarseningEndRate(dim)) {
+          rate = geoData->getCoarseningRate(dim);
+        } else {
+          rate = geoData->getCoarseningEndRate(dim);
+        }
+        if (rem > (rate / 2)) {
+          ++coarseIdx[dim];
+        }
+        if (coupled && (geoData->getStartGhostedCoarseNode(dim) * geoData->getCoarseningRate(dim) > geoData->getStartIndex(dim))) {
+          --coarseIdx[dim];
+        }
+      }
     }
 
-    RCP<IndexManager> geoData      = aggregates.GetIndexManager();
-    const bool   coupled           = geoData->isAggregationCoupled();
-    const bool   singleCoarsePoint = geoData->isSingleCoarsePoint();
-    ArrayRCP<LO> vertex2AggId      = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner        = aggregates.GetProcWinner()  ->getDataNonConst(0);
-    Array<LO>    ghostedCoarseNodeCoarseLIDs;
-    Array<int>   ghostedCoarseNodeCoarsePIDs;
-    Array<GO>    ghostedCoarseNodeCoarseGIDs;
-
+    geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
+                                     ghostedCoarseNodeCoarseLID);
+
+    aggId                 = ghostedCoarseNodeCoarseLIDs[ghostedCoarseNodeCoarseLID];
+    vertex2AggId[nodeIdx] = aggId;
+    procWinner[nodeIdx]   = ghostedCoarseNodeCoarsePIDs[ghostedCoarseNodeCoarseLID];
+    aggStat[nodeIdx]      = AGGREGATED;
+    --numNonAggregatedNodes;
+
+  }  // Loop over fine points
+}  // BuildAggregates()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildGraph(const GraphBase& graph, RCP<IndexManager>& geoData, const LO dofsPerNode,
+               RCP<CrsGraph>& myGraph, RCP<const Map>& coarseCoordinatesFineMap,
+               RCP<const Map>& coarseCoordinatesMap) const {
+  Monitor m(*this, "BuildGraphP");
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  const bool coupled = geoData->isAggregationCoupled();
+
+  // Compute the number of coarse points needed to interpolate quantities to a fine point
+  int numInterpolationPoints = 0;
+  if (geoData->getInterpolationOrder() == 0) {
+    numInterpolationPoints = 1;
+  } else if (geoData->getInterpolationOrder() == 1) {
+    // Compute 2^numDimensions using bit logic to avoid round-off errors
+    numInterpolationPoints = 1 << geoData->getNumDimensions();
+  }
+  *out << "numInterpolationPoints=" << numInterpolationPoints << std::endl;
+
+  Array<LO> colIndex((geoData->getNumLocalCoarseNodes() + numInterpolationPoints *
+                                                              (geoData->getNumLocalFineNodes() - geoData->getNumLocalCoarseNodes())) *
+                     dofsPerNode);
+  Array<size_t> rowPtr(geoData->getNumLocalFineNodes() * dofsPerNode + 1);
+  rowPtr[0] = 0;
+  ArrayRCP<size_t> nnzOnRow(geoData->getNumLocalFineNodes() * dofsPerNode);
+
+  *out << "Compute prolongatorGraph data" << std::endl;
+  if (geoData->getInterpolationOrder() == 0) {
+    ComputeGraphDataConstant(graph, geoData, dofsPerNode, numInterpolationPoints,
+                             nnzOnRow, rowPtr, colIndex);
+  } else if (geoData->getInterpolationOrder() == 1) {
+    ComputeGraphDataLinear(graph, geoData, dofsPerNode, numInterpolationPoints,
+                           nnzOnRow, rowPtr, colIndex);
+  }
+
+  // Compute graph's rowMap, colMap and domainMap
+  RCP<Map> rowMap = MapFactory::Build(graph.GetDomainMap(), dofsPerNode);
+  RCP<Map> colMap, domainMap;
+  *out << "Compute domain and column maps of the CrsGraph" << std::endl;
+  if (coupled) {
     *out << "Extract data for ghosted nodes" << std::endl;
+    Array<LO> ghostedCoarseNodeCoarseLIDs;
+    Array<int> ghostedCoarseNodeCoarsePIDs;
+    Array<GO> ghostedCoarseNodeCoarseGIDs;
     geoData->getGhostedNodesData(graph.GetDomainMap(), ghostedCoarseNodeCoarseLIDs,
                                  ghostedCoarseNodeCoarsePIDs, ghostedCoarseNodeCoarseGIDs);
 
-    LO rem, rate;
-    Array<LO> ghostedIdx(3), coarseIdx(3);
-    LO ghostedCoarseNodeCoarseLID, aggId;
-    *out << "Loop over fine nodes and assign them to an aggregate and a rank" << std::endl;
-    for(LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
-      // Compute coarse ID associated with fine LID
-      geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
-
-      for(int dim = 0; dim < 3; ++dim) {
-        if(singleCoarsePoint
-           && (geoData->getLocalFineNodesInDir(dim) - 1 < geoData->getCoarseningRate(dim))) {
-          coarseIdx[dim] = 0;
-        } else {
-          coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
-          rem    = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
-          if(ghostedIdx[dim] - geoData->getOffset(dim)
-             < geoData->getLocalFineNodesInDir(dim) - geoData->getCoarseningEndRate(dim)) {
-            rate = geoData->getCoarseningRate(dim);
-          } else {
-            rate = geoData->getCoarseningEndRate(dim);
-          }
-          if(rem > (rate / 2)) {++coarseIdx[dim];}
-          if(coupled && (geoData->getStartGhostedCoarseNode(dim)*geoData->getCoarseningRate(dim)
-                         > geoData->getStartIndex(dim))) {--coarseIdx[dim];}
-        }
+    // In this case we specify the global number of nodes on the coarse mesh
+    // as well as the GIDs needed on rank.
+    colMap = MapFactory::Build(graph.GetDomainMap()->lib(),
+                               geoData->getNumGlobalCoarseNodes(),
+                               ghostedCoarseNodeCoarseGIDs(),
+                               graph.GetDomainMap()->getIndexBase(),
+                               graph.GetDomainMap()->getComm());
+
+    LO coarseNodeIdx = 0;
+    Array<GO> coarseNodeCoarseGIDs, coarseNodeFineGIDs;
+    geoData->getCoarseNodesData(graph.GetDomainMap(), coarseNodeCoarseGIDs, coarseNodeFineGIDs);
+    for (LO nodeIdx = 0; nodeIdx < ghostedCoarseNodeCoarseGIDs.size(); ++nodeIdx) {
+      if (ghostedCoarseNodeCoarsePIDs[nodeIdx] == colMap->getComm()->getRank()) {
+        coarseNodeCoarseGIDs[coarseNodeIdx] = ghostedCoarseNodeCoarseGIDs[nodeIdx];
+        ++coarseNodeIdx;
       }
-
-      geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
-                                       ghostedCoarseNodeCoarseLID);
-
-      aggId                 = ghostedCoarseNodeCoarseLIDs[ghostedCoarseNodeCoarseLID];
-      vertex2AggId[nodeIdx] = aggId;
-      procWinner[nodeIdx]   = ghostedCoarseNodeCoarsePIDs[ghostedCoarseNodeCoarseLID];
-      aggStat[nodeIdx]      = AGGREGATED;
-      --numNonAggregatedNodes;
-
-    } // Loop over fine points
-  } // BuildAggregates()
-
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildGraph(const GraphBase& graph, RCP<IndexManager>& geoData, const LO dofsPerNode,
-             RCP<CrsGraph>& myGraph, RCP<const Map>& coarseCoordinatesFineMap,
-             RCP<const Map>& coarseCoordinatesMap) const {
-    Monitor m(*this, "BuildGraphP");
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
     }
-
-    const bool coupled           = geoData->isAggregationCoupled();
-
-    // Compute the number of coarse points needed to interpolate quantities to a fine point
-    int numInterpolationPoints = 0;
-    if(geoData->getInterpolationOrder() == 0) {
-      numInterpolationPoints = 1;
-    } else if(geoData->getInterpolationOrder() == 1) {
-      // Compute 2^numDimensions using bit logic to avoid round-off errors
-      numInterpolationPoints = 1 << geoData->getNumDimensions();
-    }
-    *out << "numInterpolationPoints=" << numInterpolationPoints << std::endl;
-
-    Array<LO> colIndex((geoData->getNumLocalCoarseNodes() + numInterpolationPoints*
-                        (geoData->getNumLocalFineNodes() - geoData->getNumLocalCoarseNodes()))*dofsPerNode);
-    Array<size_t> rowPtr(geoData->getNumLocalFineNodes()*dofsPerNode + 1);
-    rowPtr[0] = 0;
-    ArrayRCP<size_t> nnzOnRow(geoData->getNumLocalFineNodes()*dofsPerNode);
-
-    *out << "Compute prolongatorGraph data" << std::endl;
-    if(geoData->getInterpolationOrder() == 0) {
-      ComputeGraphDataConstant(graph, geoData, dofsPerNode, numInterpolationPoints,
-                               nnzOnRow, rowPtr, colIndex);
-    } else if(geoData->getInterpolationOrder() == 1) {
-      ComputeGraphDataLinear(graph, geoData, dofsPerNode, numInterpolationPoints,
-                             nnzOnRow, rowPtr, colIndex);
+    domainMap                = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                                 geoData->getNumGlobalCoarseNodes(),
+                                                 coarseNodeCoarseGIDs(),
+                                                 graph.GetDomainMap()->getIndexBase(),
+                                                 graph.GetDomainMap()->getComm());
+    coarseCoordinatesMap     = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                                 geoData->getNumGlobalCoarseNodes(),
+                                                 coarseNodeCoarseGIDs(),
+                                                 graph.GetDomainMap()->getIndexBase(),
+                                                 graph.GetDomainMap()->getComm());
+    coarseCoordinatesFineMap = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                                 geoData->getNumGlobalCoarseNodes(),
+                                                 coarseNodeFineGIDs(),
+                                                 graph.GetDomainMap()->getIndexBase(),
+                                                 graph.GetDomainMap()->getComm());
+  } else {
+    // In this case the map will compute the global number of nodes on the coarse mesh
+    // and it will assign GIDs to the local coarse nodes.
+    colMap    = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                  Teuchos::OrdinalTraits<GO>::invalid(),
+                                  geoData->getNumLocalCoarseNodes() * dofsPerNode,
+                                  graph.GetDomainMap()->getIndexBase(),
+                                  graph.GetDomainMap()->getComm());
+    domainMap = colMap;
+
+    Array<GO> coarseNodeCoarseGIDs(geoData->getNumLocalCoarseNodes());
+    Array<GO> coarseNodeFineGIDs(geoData->getNumLocalCoarseNodes());
+    geoData->getCoarseNodesData(graph.GetDomainMap(), coarseNodeCoarseGIDs, coarseNodeFineGIDs);
+    coarseCoordinatesMap     = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                                 Teuchos::OrdinalTraits<GO>::invalid(),
+                                                 geoData->getNumLocalCoarseNodes(),
+                                                 graph.GetDomainMap()->getIndexBase(),
+                                                 graph.GetDomainMap()->getComm());
+    coarseCoordinatesFineMap = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                                 Teuchos::OrdinalTraits<GO>::invalid(),
+                                                 coarseNodeFineGIDs(),
+                                                 graph.GetDomainMap()->getIndexBase(),
+                                                 graph.GetDomainMap()->getComm());
+  }
+
+  *out << "Call constructor of CrsGraph" << std::endl;
+  myGraph = CrsGraphFactory::Build(rowMap,
+                                   colMap,
+                                   nnzOnRow);
+
+  *out << "Fill CrsGraph" << std::endl;
+  LO rowIdx = 0;
+  for (LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
+    for (LO dof = 0; dof < dofsPerNode; ++dof) {
+      rowIdx = nodeIdx * dofsPerNode + dof;
+      myGraph->insertLocalIndices(rowIdx, colIndex(rowPtr[rowIdx], nnzOnRow[rowIdx]));
     }
-
-    // Compute graph's rowMap, colMap and domainMap
-    RCP<Map> rowMap = MapFactory::Build(graph.GetDomainMap(), dofsPerNode);
-    RCP<Map> colMap, domainMap;
-    *out << "Compute domain and column maps of the CrsGraph" << std::endl;
-    if(coupled){
-      *out << "Extract data for ghosted nodes" << std::endl;
-      Array<LO>  ghostedCoarseNodeCoarseLIDs;
-      Array<int> ghostedCoarseNodeCoarsePIDs;
-      Array<GO>  ghostedCoarseNodeCoarseGIDs;
-      geoData->getGhostedNodesData(graph.GetDomainMap(), ghostedCoarseNodeCoarseLIDs,
-                                   ghostedCoarseNodeCoarsePIDs, ghostedCoarseNodeCoarseGIDs);
-
-      // In this case we specify the global number of nodes on the coarse mesh
-      // as well as the GIDs needed on rank.
-      colMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                 geoData->getNumGlobalCoarseNodes(),
-                                 ghostedCoarseNodeCoarseGIDs(),
-                                 graph.GetDomainMap()->getIndexBase(),
-                                 graph.GetDomainMap()->getComm());
-
-      LO coarseNodeIdx = 0;
-      Array<GO> coarseNodeCoarseGIDs, coarseNodeFineGIDs;
-      geoData->getCoarseNodesData(graph.GetDomainMap(), coarseNodeCoarseGIDs, coarseNodeFineGIDs);
-      for(LO nodeIdx = 0; nodeIdx < ghostedCoarseNodeCoarseGIDs.size(); ++nodeIdx) {
-        if(ghostedCoarseNodeCoarsePIDs[nodeIdx] == colMap->getComm()->getRank()) {
-          coarseNodeCoarseGIDs[coarseNodeIdx] = ghostedCoarseNodeCoarseGIDs[nodeIdx];
-          ++coarseNodeIdx;
+  }
+
+  *out << "Call fillComplete on CrsGraph" << std::endl;
+  myGraph->fillComplete(domainMap, rowMap);
+  *out << "Prolongator CrsGraph computed" << std::endl;
+
+}  // BuildGraph()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeGraphDataConstant(const GraphBase& graph, RCP<IndexManager>& geoData,
+                             const LO dofsPerNode, const int /* numInterpolationPoints */,
+                             ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
+                             Array<LO>& colIndex) const {
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  Array<LO> ghostedCoarseNodeCoarseLIDs;
+  Array<int> ghostedCoarseNodeCoarsePIDs;
+  Array<GO> ghostedCoarseNodeCoarseGIDs;
+  geoData->getGhostedNodesData(graph.GetDomainMap(), ghostedCoarseNodeCoarseLIDs,
+                               ghostedCoarseNodeCoarsePIDs, ghostedCoarseNodeCoarseGIDs);
+
+  LO ghostedCoarseNodeCoarseLID, rem, rate;
+  Array<LO> ghostedIdx(3), coarseIdx(3);
+  for (LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
+    // Compute coarse ID associated with fine LID
+    geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
+
+    for (int dim = 0; dim < 3; ++dim) {
+      if (geoData->isSingleCoarsePoint() && (geoData->getLocalFineNodesInDir(dim) - 1 < geoData->getCoarseningRate(dim))) {
+        coarseIdx[dim] = 0;
+      } else {
+        coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
+        rem            = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
+        if (ghostedIdx[dim] - geoData->getOffset(dim) < geoData->getLocalFineNodesInDir(dim) - geoData->getCoarseningEndRate(dim)) {
+          rate = geoData->getCoarseningRate(dim);
+        } else {
+          rate = geoData->getCoarseningEndRate(dim);
+        }
+        if (rem > (rate / 2)) {
+          ++coarseIdx[dim];
+        }
+        if ((geoData->getStartGhostedCoarseNode(dim) * geoData->getCoarseningRate(dim) > geoData->getStartIndex(dim)) && geoData->isAggregationCoupled()) {
+          --coarseIdx[dim];
         }
-      }
-      domainMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                    geoData->getNumGlobalCoarseNodes(),
-                                    coarseNodeCoarseGIDs(),
-                                    graph.GetDomainMap()->getIndexBase(),
-                                    graph.GetDomainMap()->getComm());
-      coarseCoordinatesMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                               geoData->getNumGlobalCoarseNodes(),
-                                               coarseNodeCoarseGIDs(),
-                                               graph.GetDomainMap()->getIndexBase(),
-                                               graph.GetDomainMap()->getComm());
-      coarseCoordinatesFineMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                                   geoData->getNumGlobalCoarseNodes(),
-                                                   coarseNodeFineGIDs(),
-                                                   graph.GetDomainMap()->getIndexBase(),
-                                                   graph.GetDomainMap()->getComm());
-    } else {
-      // In this case the map will compute the global number of nodes on the coarse mesh
-      // and it will assign GIDs to the local coarse nodes.
-      colMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                 Teuchos::OrdinalTraits<GO>::invalid(),
-                                 geoData->getNumLocalCoarseNodes()*dofsPerNode,
-                                 graph.GetDomainMap()->getIndexBase(),
-                                 graph.GetDomainMap()->getComm());
-      domainMap = colMap;
-
-      Array<GO> coarseNodeCoarseGIDs(geoData->getNumLocalCoarseNodes());
-      Array<GO> coarseNodeFineGIDs(geoData->getNumLocalCoarseNodes());
-      geoData->getCoarseNodesData(graph.GetDomainMap(), coarseNodeCoarseGIDs, coarseNodeFineGIDs);
-      coarseCoordinatesMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                               Teuchos::OrdinalTraits<GO>::invalid(),
-                                               geoData->getNumLocalCoarseNodes(),
-                                               graph.GetDomainMap()->getIndexBase(),
-                                               graph.GetDomainMap()->getComm());
-      coarseCoordinatesFineMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                                                   Teuchos::OrdinalTraits<GO>::invalid(),
-                                                   coarseNodeFineGIDs(),
-                                                   graph.GetDomainMap()->getIndexBase(),
-                                                   graph.GetDomainMap()->getComm());
-    }
-
-    *out << "Call constructor of CrsGraph" << std::endl;
-    myGraph = CrsGraphFactory::Build(rowMap,
-                                     colMap,
-                                     nnzOnRow);
-
-    *out << "Fill CrsGraph" << std::endl;
-    LO rowIdx = 0;
-    for(LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
-      for(LO dof = 0; dof < dofsPerNode; ++dof) {
-        rowIdx = nodeIdx*dofsPerNode + dof;
-        myGraph->insertLocalIndices(rowIdx, colIndex(rowPtr[rowIdx], nnzOnRow[rowIdx]) );
       }
     }
 
-    *out << "Call fillComplete on CrsGraph" << std::endl;
-    myGraph->fillComplete(domainMap, rowMap);
-    *out << "Prolongator CrsGraph computed" << std::endl;
+    geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
+                                     ghostedCoarseNodeCoarseLID);
 
-  } // BuildGraph()
+    for (LO dof = 0; dof < dofsPerNode; ++dof) {
+      nnzOnRow[nodeIdx * dofsPerNode + dof]   = 1;
+      rowPtr[nodeIdx * dofsPerNode + dof + 1] = rowPtr[nodeIdx * dofsPerNode + dof] + 1;
+      colIndex[rowPtr[nodeIdx * dofsPerNode + dof]] =
+          ghostedCoarseNodeCoarseLIDs[ghostedCoarseNodeCoarseLID] * dofsPerNode + dof;
+    }
+  }  // Loop over fine points
 
+}  // ComputeGraphDataConstant()
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeGraphDataConstant(const GraphBase& graph, RCP<IndexManager>& geoData,
-                           const LO dofsPerNode, const int /* numInterpolationPoints */,
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeGraphDataLinear(const GraphBase& /* graph */, RCP<IndexManager>& geoData,
+                           const LO dofsPerNode, const int numInterpolationPoints,
                            ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
                            Array<LO>& colIndex) const {
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    Array<LO>    ghostedCoarseNodeCoarseLIDs;
-    Array<int>   ghostedCoarseNodeCoarsePIDs;
-    Array<GO>    ghostedCoarseNodeCoarseGIDs;
-    geoData->getGhostedNodesData(graph.GetDomainMap(), ghostedCoarseNodeCoarseLIDs,
-                                 ghostedCoarseNodeCoarsePIDs, ghostedCoarseNodeCoarseGIDs);
-
-    LO ghostedCoarseNodeCoarseLID, rem, rate;
-    Array<LO> ghostedIdx(3), coarseIdx(3);
-    for(LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
-
-      // Compute coarse ID associated with fine LID
-      geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
-
-      for(int dim = 0; dim < 3; ++dim) {
-        if(geoData->isSingleCoarsePoint()
-           && (geoData->getLocalFineNodesInDir(dim) - 1 < geoData->getCoarseningRate(dim))) {
-          coarseIdx[dim] = 0;
-        } else {
-          coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
-          rem    = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
-          if(ghostedIdx[dim] - geoData->getOffset(dim)
-             < geoData->getLocalFineNodesInDir(dim) - geoData->getCoarseningEndRate(dim)) {
-            rate = geoData->getCoarseningRate(dim);
-          } else {
-            rate = geoData->getCoarseningEndRate(dim);
-          }
-          if(rem > (rate / 2)) {++coarseIdx[dim];}
-          if( (geoData->getStartGhostedCoarseNode(dim)*geoData->getCoarseningRate(dim)
-               > geoData->getStartIndex(dim)) && geoData->isAggregationCoupled() ) {
-            --coarseIdx[dim];
-          }
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  const bool coupled      = geoData->isAggregationCoupled();
+  const int numDimensions = geoData->getNumDimensions();
+  Array<LO> ghostedIdx(3, 0);
+  Array<LO> coarseIdx(3, 0);
+  Array<LO> ijkRem(3, 0);
+  const LO coarsePointOffset[8][3] = {{0, 0, 0}, {1, 0, 0}, {0, 1, 0}, {1, 1, 0}, {0, 0, 1}, {1, 0, 1}, {0, 1, 1}, {1, 1, 1}};
+
+  for (LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
+    // Compute coarse ID associated with fine LID
+    geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
+    for (int dim = 0; dim < numDimensions; dim++) {
+      coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
+      ijkRem[dim]    = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
+      if (coupled) {
+        if (geoData->getStartGhostedCoarseNode(dim) * geoData->getCoarseningRate(dim) > geoData->getStartIndex(dim)) {
+          --coarseIdx[dim];
         }
-      }
-
-      geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
-                                       ghostedCoarseNodeCoarseLID);
-
-      for(LO dof = 0; dof < dofsPerNode; ++dof) {
-        nnzOnRow[nodeIdx*dofsPerNode + dof]         = 1;
-        rowPtr[nodeIdx*dofsPerNode + dof + 1]       = rowPtr[nodeIdx*dofsPerNode + dof] + 1;
-        colIndex[rowPtr[nodeIdx*dofsPerNode + dof]] =
-          ghostedCoarseNodeCoarseLIDs[ghostedCoarseNodeCoarseLID]*dofsPerNode + dof;
-      }
-    } // Loop over fine points
-
-  } // ComputeGraphDataConstant()
-
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeGraphDataLinear(const GraphBase& /* graph */, RCP<IndexManager>& geoData,
-                         const LO dofsPerNode, const int numInterpolationPoints,
-                         ArrayRCP<size_t>& nnzOnRow, Array<size_t>& rowPtr,
-                         Array<LO>& colIndex) const {
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    const bool coupled       = geoData->isAggregationCoupled();
-    const int  numDimensions = geoData->getNumDimensions();
-    Array<LO> ghostedIdx(3,0);
-    Array<LO> coarseIdx(3,0);
-    Array<LO> ijkRem(3,0);
-    const LO coarsePointOffset[8][3] = {{0, 0, 0}, {1, 0, 0}, {0, 1, 0}, {1, 1, 0},
-                                        {0, 0, 1}, {1, 0, 1}, {0, 1, 1}, {1, 1, 1}};
-
-    for(LO nodeIdx = 0; nodeIdx < geoData->getNumLocalFineNodes(); ++nodeIdx) {
-
-      // Compute coarse ID associated with fine LID
-      geoData->getFineNodeGhostedTuple(nodeIdx, ghostedIdx[0], ghostedIdx[1], ghostedIdx[2]);
-      for(int dim=0; dim < numDimensions; dim++){
-        coarseIdx[dim] = ghostedIdx[dim] / geoData->getCoarseningRate(dim);
-        ijkRem[dim]    = ghostedIdx[dim] % geoData->getCoarseningRate(dim);
-        if(coupled) {
-          if (geoData->getStartGhostedCoarseNode(dim)*geoData->getCoarseningRate(dim)
-                       > geoData->getStartIndex(dim)) {
-            --coarseIdx[dim];
-          }
-        } else {
-          if(ghostedIdx[dim] == geoData->getLocalFineNodesInDir(dim) - 1) {
-            coarseIdx[dim] = geoData->getLocalCoarseNodesInDir(dim) - 1;
-          }
+      } else {
+        if (ghostedIdx[dim] == geoData->getLocalFineNodesInDir(dim) - 1) {
+          coarseIdx[dim] = geoData->getLocalCoarseNodesInDir(dim) - 1;
         }
       }
+    }
 
-      // Fill Graph
-      // Check if Fine node lies on Coarse Node
-      bool allCoarse = true;
-      Array<bool> isCoarse(numDimensions);
-      for(int dim = 0; dim < numDimensions; ++dim) {
-        isCoarse[dim] = false;
-        if(ijkRem[dim] == 0)
+    // Fill Graph
+    // Check if Fine node lies on Coarse Node
+    bool allCoarse = true;
+    Array<bool> isCoarse(numDimensions);
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      isCoarse[dim] = false;
+      if (ijkRem[dim] == 0)
+        isCoarse[dim] = true;
+
+      if (coupled) {
+        if (ghostedIdx[dim] - geoData->getOffset(dim) == geoData->getLocalFineNodesInDir(dim) - 1 &&
+            geoData->getMeshEdge(dim * 2 + 1))
+          isCoarse[dim] = true;
+      } else {
+        if (ghostedIdx[dim] - geoData->getOffset(dim) == geoData->getLocalFineNodesInDir(dim) - 1)
           isCoarse[dim] = true;
-
-        if(coupled){
-          if( ghostedIdx[dim]-geoData->getOffset(dim) == geoData->getLocalFineNodesInDir(dim)-1 &&
-              geoData->getMeshEdge(dim*2+1) )
-            isCoarse[dim] = true;
-        } else {
-          if( ghostedIdx[dim]-geoData->getOffset(dim) == geoData->getLocalFineNodesInDir(dim)-1)
-            isCoarse[dim] = true;
-        }
-
-        if(!isCoarse[dim])
-          allCoarse = false;
       }
 
-      LO rowIdx = 0, colIdx = 0;
-      if(allCoarse) {
-        for(LO dof = 0; dof < dofsPerNode; ++dof) {
-          rowIdx = nodeIdx*dofsPerNode + dof;
-          nnzOnRow[rowIdx] = 1;
-          rowPtr[rowIdx + 1] = rowPtr[rowIdx] + 1;
+      if (!isCoarse[dim])
+        allCoarse = false;
+    }
 
-          // Fine node lies on Coarse node, easy case, we only need the LID of the coarse node.
-          geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2], colIdx);
-          colIndex[rowPtr[rowIdx]] = colIdx*dofsPerNode + dof;
-        }
-      } else {
-        // Harder case, we need the LIDs of all the coarse nodes contributing to the interpolation
-        for(int dim = 0; dim < numDimensions; ++dim) {
-          if(coarseIdx[dim] == geoData->getGhostedNodesInDir(dim) - 1)
-            --coarseIdx[dim];
-        }
+    LO rowIdx = 0, colIdx = 0;
+    if (allCoarse) {
+      for (LO dof = 0; dof < dofsPerNode; ++dof) {
+        rowIdx             = nodeIdx * dofsPerNode + dof;
+        nnzOnRow[rowIdx]   = 1;
+        rowPtr[rowIdx + 1] = rowPtr[rowIdx] + 1;
 
-        for(LO dof = 0; dof < dofsPerNode; ++dof) {
-          // at the current node.
-          rowIdx = nodeIdx*dofsPerNode + dof;
-          nnzOnRow[rowIdx] = Teuchos::as<size_t>( numInterpolationPoints );
-          rowPtr[rowIdx + 1] = rowPtr[rowIdx] + Teuchos::as<LO>(numInterpolationPoints);
-          // Compute Coarse Node LID
-          for(LO interpIdx = 0; interpIdx < numInterpolationPoints; ++interpIdx) {
-            geoData->getCoarseNodeGhostedLID(coarseIdx[0] + coarsePointOffset[interpIdx][0],
-                                             coarseIdx[1] + coarsePointOffset[interpIdx][1],
-                                             coarseIdx[2] + coarsePointOffset[interpIdx][2],
-                                             colIdx);
-            colIndex[rowPtr[rowIdx] + interpIdx] = colIdx*dofsPerNode + dof;
-          } // Loop over numInterpolationPoints
-        } // Loop over dofsPerNode
+        // Fine node lies on Coarse node, easy case, we only need the LID of the coarse node.
+        geoData->getCoarseNodeGhostedLID(coarseIdx[0], coarseIdx[1], coarseIdx[2], colIdx);
+        colIndex[rowPtr[rowIdx]] = colIdx * dofsPerNode + dof;
+      }
+    } else {
+      // Harder case, we need the LIDs of all the coarse nodes contributing to the interpolation
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        if (coarseIdx[dim] == geoData->getGhostedNodesInDir(dim) - 1)
+          --coarseIdx[dim];
       }
-    } // Loop over fine points
-  } // ComputeGraphDataLinear()
 
-} // end namespace
+      for (LO dof = 0; dof < dofsPerNode; ++dof) {
+        // at the current node.
+        rowIdx             = nodeIdx * dofsPerNode + dof;
+        nnzOnRow[rowIdx]   = Teuchos::as<size_t>(numInterpolationPoints);
+        rowPtr[rowIdx + 1] = rowPtr[rowIdx] + Teuchos::as<LO>(numInterpolationPoints);
+        // Compute Coarse Node LID
+        for (LO interpIdx = 0; interpIdx < numInterpolationPoints; ++interpIdx) {
+          geoData->getCoarseNodeGhostedLID(coarseIdx[0] + coarsePointOffset[interpIdx][0],
+                                           coarseIdx[1] + coarsePointOffset[interpIdx][1],
+                                           coarseIdx[2] + coarsePointOffset[interpIdx][2],
+                                           colIdx);
+          colIndex[rowPtr[rowIdx] + interpIdx] = colIdx * dofsPerNode + dof;
+        }  // Loop over numInterpolationPoints
+      }    // Loop over dofsPerNode
+    }
+  }  // Loop over fine points
+}  // ComputeGraphDataLinear()
 
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_decl.hpp
index 5d83bf9a5a42..cab18f503aec 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_decl.hpp
@@ -55,169 +55,160 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationStructuredAlgorithm class.
-    @brief Algorithm for coarsening a graph with structured aggregation.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Use the logical indexing of the mesh to obtain a very regular aggregation pattern and maintain
-    lines and planes of the problem as they might be useful to the smoother.
-    This algorithms is also very easy to parallelize on node due to its very regular and predictible
-    memory access patern.
-    All the parameters needed are passed to this class by the StructuredAggregationFactory class.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationStructuredAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationStructuredAlgorithm class.
+  @brief Algorithm for coarsening a graph with structured aggregation.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Use the logical indexing of the mesh to obtain a very regular aggregation pattern and maintain
+  lines and planes of the problem as they might be useful to the smoother.
+  This algorithms is also very easy to parallelize on node due to its very regular and predictible
+  memory access patern.
+  All the parameters needed are passed to this class by the StructuredAggregationFactory class.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationStructuredAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
+ public:
+  using local_graph_type       = typename LWGraph_kokkos::local_graph_type;
+  using non_const_row_map_type = typename local_graph_type::row_map_type::non_const_type;
+  using size_type              = typename local_graph_type::size_type;
+  using entries_type           = typename local_graph_type::entries_type;
+  using device_type            = typename local_graph_type::device_type;
+  using execution_space        = typename local_graph_type::device_type::execution_space;
+  using memory_space           = typename local_graph_type::device_type::memory_space;
 
-    using local_graph_type       = typename LWGraph_kokkos::local_graph_type;
-    using non_const_row_map_type = typename local_graph_type::row_map_type::non_const_type;
-    using size_type              = typename local_graph_type::size_type;
-    using entries_type           = typename local_graph_type::entries_type;
-    using device_type            = typename local_graph_type::device_type;
-    using execution_space        = typename local_graph_type::device_type::execution_space;
-    using memory_space           = typename local_graph_type::device_type::memory_space;
+  using LOVectorView      = decltype(std::declval<LOVector>().getDeviceLocalView(Xpetra::Access::ReadWrite));
+  using constIntTupleView = typename Kokkos::View<const int[3], device_type>;
+  using constLOTupleView  = typename Kokkos::View<const LO[3], device_type>;
 
-    using LOVectorView = decltype(std::declval<LOVector>().getDeviceLocalView(Xpetra::Access::ReadWrite));
-    using constIntTupleView = typename Kokkos::View<const int[3], device_type>;
-    using constLOTupleView  = typename Kokkos::View<const LO[3],  device_type>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor.
+  AggregationStructuredAlgorithm_kokkos() {}
 
-    //! Constructor.
-    AggregationStructuredAlgorithm_kokkos() { }
+  //! Destructor.
+  virtual ~AggregationStructuredAlgorithm_kokkos() {}
 
-    //! Destructor.
-    virtual ~AggregationStructuredAlgorithm_kokkos() { }
+  //@}
 
-    //@}
+  //! @name Aggregation methods.
+  //@{
 
+  /*! @brief Build aggregates object. */
 
-    //! @name Aggregation methods.
-    //@{
+  void BuildAggregates(const Teuchos::ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    /*! @brief Build aggregates object. */
+  /*! @brief Build a CrsGraph instead of aggregates. */
 
-    void BuildAggregates(const Teuchos::ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  void BuildGraph(const LWGraph_kokkos& graph,
+                  RCP<IndexManager_kokkos>& geoData,
+                  const LO dofsPerNode,
+                  RCP<CrsGraph>& myGraph) const;
+  //@}
 
-    /*! @brief Build a CrsGraph instead of aggregates. */
+  std::string description() const { return "Aggretation: structured algorithm"; }
 
-    void BuildGraph(const LWGraph_kokkos& graph,
-                    RCP<IndexManager_kokkos>& geoData,
-                    const LO dofsPerNode,
-                    RCP<CrsGraph>& myGraph) const;
-    //@}
+  struct fillAggregatesFunctor {
+    IndexManager_kokkos geoData_;
+    const int myRank_;
+    Kokkos::View<unsigned*, device_type> aggStat_;
+    LOVectorView vertex2AggID_;
+    LOVectorView procWinner_;
 
-    std::string description() const { return "Aggretation: structured algorithm"; }
+    fillAggregatesFunctor(RCP<IndexManager_kokkos> geoData,
+                          const int myRank,
+                          Kokkos::View<unsigned*, device_type> aggStat,
+                          LOVectorView vertex2AggID,
+                          LOVectorView procWinner);
 
-    struct fillAggregatesFunctor{
+    KOKKOS_INLINE_FUNCTION
+    void operator()(const LO nodeIdx, LO& lNumAggregatedNodes) const;
 
-      IndexManager_kokkos geoData_;
-      const int myRank_;
-      Kokkos::View<unsigned*, device_type> aggStat_;
-      LOVectorView vertex2AggID_;
-      LOVectorView procWinner_;
+  };  // struct fillAggregatesFunctor
 
-      fillAggregatesFunctor(RCP<IndexManager_kokkos> geoData,
-                            const int myRank,
-                            Kokkos::View<unsigned*, device_type> aggStat,
-                            LOVectorView vertex2AggID,
-                            LOVectorView procWinner);
+  struct computeGraphDataConstantFunctor {
+    IndexManager_kokkos geoData_;
+    const int numGhostedNodes_;
+    const LO dofsPerNode_;
+    constIntTupleView coarseRate_;
+    constIntTupleView endRate_;
+    constLOTupleView lFineNodesPerDir_;
+    non_const_row_map_type rowPtr_;
+    entries_type colIndex_;
 
-      KOKKOS_INLINE_FUNCTION
-      void operator() (const LO nodeIdx, LO& lNumAggregatedNodes) const;
-
-    }; // struct fillAggregatesFunctor
-
-    struct computeGraphDataConstantFunctor {
-
-      IndexManager_kokkos geoData_;
-      const int numGhostedNodes_;
-      const LO dofsPerNode_;
-      constIntTupleView coarseRate_;
-      constIntTupleView endRate_;
-      constLOTupleView  lFineNodesPerDir_;
-      non_const_row_map_type             rowPtr_;
-      entries_type                       colIndex_;
-
-
-      computeGraphDataConstantFunctor(RCP<IndexManager_kokkos> geoData,
-                                      const LO numGhostedNodes, const LO dofsPerNode,
-                                      constIntTupleView coarseRate, constIntTupleView endRate,
-                                      constLOTupleView  lFineNodesPerDir,
-                                      non_const_row_map_type rowPtr, entries_type colIndex);
-
-      KOKKOS_INLINE_FUNCTION
-      void operator() (const LO nodeIdx) const;
-
-    }; // struct computeGraphDataConstantFunctor
-
-    struct computeGraphRowPtrFunctor {
-
-      IndexManager_kokkos geoData_;
-      const LO dofsPerNode_;
-      const int numInterpolationPoints_;
-      const LO numLocalRows_;
-      constIntTupleView coarseRate_;
-      constLOTupleView lFineNodesPerDir_;
-      non_const_row_map_type rowPtr_;
-
-      computeGraphRowPtrFunctor(RCP<IndexManager_kokkos> geoData,
-                                const LO dofsPerNode,
-                                const int numInterpolationPoints, const LO  numLocalRows,
-                                constIntTupleView coarseRate, constLOTupleView lFineNodesPerDir,
-                                non_const_row_map_type rowPtr);
-
-      KOKKOS_INLINE_FUNCTION
-      void operator() (const LO rowIdx, GO& update, const bool final) const;
-    }; // struct computeGraphRowPtrFunctor
-
-    struct computeGraphDataLinearFunctor {
-
-      IndexManager_kokkos geoData_;
-      const int numDimensions_;
-      const int numGhostedNodes_;
-      const LO dofsPerNode_;
-      const int numInterpolationPoints_;
-      constIntTupleView coarseRate_;
-      constIntTupleView endRate_;
-      constLOTupleView  lFineNodesPerDir_;
-      constLOTupleView  ghostedNodesPerDir_;
-      non_const_row_map_type                    rowPtr_;
-      entries_type                              colIndex_;
-
-
-      computeGraphDataLinearFunctor(RCP<IndexManager_kokkos> geoData,
-                                    const int numDimensions,
+    computeGraphDataConstantFunctor(RCP<IndexManager_kokkos> geoData,
                                     const LO numGhostedNodes, const LO dofsPerNode,
-                                    const int numInterpolationPoints,
                                     constIntTupleView coarseRate, constIntTupleView endRate,
                                     constLOTupleView lFineNodesPerDir,
-                                    constLOTupleView ghostedNodesPerDir,
                                     non_const_row_map_type rowPtr, entries_type colIndex);
 
-      KOKKOS_INLINE_FUNCTION
-      void operator() (const LO nodeIdx) const;
-
-    }; // struct computeGraphDataLinearFunctor
-
-  }; // class AggregationStructuredAlgorithm_kokkos
-
-} //namespace MueLu
+    KOKKOS_INLINE_FUNCTION
+    void operator()(const LO nodeIdx) const;
+
+  };  // struct computeGraphDataConstantFunctor
+
+  struct computeGraphRowPtrFunctor {
+    IndexManager_kokkos geoData_;
+    const LO dofsPerNode_;
+    const int numInterpolationPoints_;
+    const LO numLocalRows_;
+    constIntTupleView coarseRate_;
+    constLOTupleView lFineNodesPerDir_;
+    non_const_row_map_type rowPtr_;
+
+    computeGraphRowPtrFunctor(RCP<IndexManager_kokkos> geoData,
+                              const LO dofsPerNode,
+                              const int numInterpolationPoints, const LO numLocalRows,
+                              constIntTupleView coarseRate, constLOTupleView lFineNodesPerDir,
+                              non_const_row_map_type rowPtr);
+
+    KOKKOS_INLINE_FUNCTION
+    void operator()(const LO rowIdx, GO& update, const bool final) const;
+  };  // struct computeGraphRowPtrFunctor
+
+  struct computeGraphDataLinearFunctor {
+    IndexManager_kokkos geoData_;
+    const int numDimensions_;
+    const int numGhostedNodes_;
+    const LO dofsPerNode_;
+    const int numInterpolationPoints_;
+    constIntTupleView coarseRate_;
+    constIntTupleView endRate_;
+    constLOTupleView lFineNodesPerDir_;
+    constLOTupleView ghostedNodesPerDir_;
+    non_const_row_map_type rowPtr_;
+    entries_type colIndex_;
+
+    computeGraphDataLinearFunctor(RCP<IndexManager_kokkos> geoData,
+                                  const int numDimensions,
+                                  const LO numGhostedNodes, const LO dofsPerNode,
+                                  const int numInterpolationPoints,
+                                  constIntTupleView coarseRate, constIntTupleView endRate,
+                                  constLOTupleView lFineNodesPerDir,
+                                  constLOTupleView ghostedNodesPerDir,
+                                  non_const_row_map_type rowPtr, entries_type colIndex);
+
+    KOKKOS_INLINE_FUNCTION
+    void operator()(const LO nodeIdx) const;
+
+  };  // struct computeGraphDataLinearFunctor
+
+};  // class AggregationStructuredAlgorithm_kokkos
+
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_SHORT
 #endif /* MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_def.hpp
index 09ab332b27f4..d6e7bb665f33 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_AggregationStructuredAlgorithm_kokkos_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_DEF_HPP
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_DEF_HPP
 
-
 #include <Teuchos_Comm.hpp>
 #include <Teuchos_CommHelpers.hpp>
 
@@ -65,348 +64,363 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const Teuchos::ParameterList& /* params */, const LWGraph_kokkos& graph,
-                  Aggregates& aggregates,
-                  Kokkos::View<unsigned*, device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    RCP<IndexManager_kokkos> geoData = aggregates.GetIndexManagerKokkos();
-    const LO numLocalFineNodes= geoData->getNumLocalFineNodes();
-    const LO numCoarseNodes   = geoData->getNumCoarseNodes();
-    LOVectorView vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    LOVectorView procWinner   = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-    *out << "Loop over fine nodes and assign them to an aggregate and a rank" << std::endl;
-    LO numAggregatedNodes;
-    fillAggregatesFunctor fillAggregates(geoData,
-                                         graph.GetComm()->getRank(),
-                                         aggStat,
-                                         vertex2AggId,
-                                         procWinner);
-    Kokkos::parallel_reduce("StructuredAggregation: fill aggregates data",
-                            Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
-                            fillAggregates,
-                            numAggregatedNodes);
-
-    *out << "numCoarseNodes= " << numCoarseNodes
-         << ", numAggregatedNodes= " << numAggregatedNodes << std::endl;
-    numNonAggregatedNodes = numNonAggregatedNodes - numAggregatedNodes;
-
-  } // BuildAggregates()
-
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildGraph(const LWGraph_kokkos& graph, RCP<IndexManager_kokkos>& geoData, const LO dofsPerNode,
-             RCP<CrsGraph>& myGraph) const {
-    Monitor m(*this, "BuildGraphP");
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    // Compute the number of coarse points needed to interpolate quantities to a fine point
-    int numInterpolationPoints = 0;
-    if(geoData->getInterpolationOrder() == 0) {
-      numInterpolationPoints = 1;
-    } else if(geoData->getInterpolationOrder() == 1) {
-      // Compute 2^numDimensions using bit logic to avoid round-off errors from std::pow()
-      numInterpolationPoints = 1 << geoData->getNumDimensions();
-    }
-    *out << "numInterpolationPoints=" << numInterpolationPoints << std::endl;
-
-    const LO numLocalFineNodes = geoData->getNumLocalFineNodes();
-    const LO numCoarseNodes    = geoData->getNumCoarseNodes();
-    const LO numNnzEntries     = dofsPerNode*(numCoarseNodes + numInterpolationPoints
-                                              *(numLocalFineNodes - numCoarseNodes));
-
-    non_const_row_map_type rowPtr("Prolongator graph, rowPtr", dofsPerNode*(numLocalFineNodes + 1));
-    entries_type colIndex("Prolongator graph, colIndices", numNnzEntries);
-
-    *out << "Compute prolongatorGraph data" << std::endl;
-    if(geoData->getInterpolationOrder() == 0) {
-      computeGraphDataConstantFunctor computeGraphData(geoData,
-                                                       numCoarseNodes,
-                                                       dofsPerNode,
-                                                       geoData->getCoarseningRates(),
-                                                       geoData->getCoarseningEndRates(),
-                                                       geoData->getLocalFineNodesPerDir(),
-                                                       rowPtr,
-                                                       colIndex);
-      Kokkos::parallel_for("Structured Aggregation: compute loca graph data",
-                           Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
-                           computeGraphData);
-    } else if(geoData->getInterpolationOrder() == 1) {
-      // Note, lbv 2018-11-08: in the piece-wise linear case I am computing the rowPtr
-      // using a parallel scan, it might be possible to do something faster than that
-      // by including this calculation in computeGraphDataLinearFunctor but at the moment
-      // all the ideas I have include a bunch of if statements which I would like to avoid.
-      computeGraphRowPtrFunctor computeGraphRowPtr(geoData,
-                                                   dofsPerNode,
-                                                   numInterpolationPoints,
-                                                   numLocalFineNodes,
-                                                   geoData->getCoarseningRates(),
-                                                   geoData->getLocalFineNodesPerDir(),
-                                                   rowPtr);
-      Kokkos::parallel_scan("Structured Aggregation: compute rowPtr for prolongator graph",
-                            Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes + 1),
-                            computeGraphRowPtr);
-
-      computeGraphDataLinearFunctor computeGraphData(geoData,
-                                                     geoData->getNumDimensions(),
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const Teuchos::ParameterList& /* params */, const LWGraph_kokkos& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  RCP<IndexManager_kokkos> geoData = aggregates.GetIndexManagerKokkos();
+  const LO numLocalFineNodes       = geoData->getNumLocalFineNodes();
+  const LO numCoarseNodes          = geoData->getNumCoarseNodes();
+  LOVectorView vertex2AggId        = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  LOVectorView procWinner          = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+  *out << "Loop over fine nodes and assign them to an aggregate and a rank" << std::endl;
+  LO numAggregatedNodes;
+  fillAggregatesFunctor fillAggregates(geoData,
+                                       graph.GetComm()->getRank(),
+                                       aggStat,
+                                       vertex2AggId,
+                                       procWinner);
+  Kokkos::parallel_reduce("StructuredAggregation: fill aggregates data",
+                          Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
+                          fillAggregates,
+                          numAggregatedNodes);
+
+  *out << "numCoarseNodes= " << numCoarseNodes
+       << ", numAggregatedNodes= " << numAggregatedNodes << std::endl;
+  numNonAggregatedNodes = numNonAggregatedNodes - numAggregatedNodes;
+
+}  // BuildAggregates()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildGraph(const LWGraph_kokkos& graph, RCP<IndexManager_kokkos>& geoData, const LO dofsPerNode,
+               RCP<CrsGraph>& myGraph) const {
+  Monitor m(*this, "BuildGraphP");
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDALGORITHM_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  // Compute the number of coarse points needed to interpolate quantities to a fine point
+  int numInterpolationPoints = 0;
+  if (geoData->getInterpolationOrder() == 0) {
+    numInterpolationPoints = 1;
+  } else if (geoData->getInterpolationOrder() == 1) {
+    // Compute 2^numDimensions using bit logic to avoid round-off errors from std::pow()
+    numInterpolationPoints = 1 << geoData->getNumDimensions();
+  }
+  *out << "numInterpolationPoints=" << numInterpolationPoints << std::endl;
+
+  const LO numLocalFineNodes = geoData->getNumLocalFineNodes();
+  const LO numCoarseNodes    = geoData->getNumCoarseNodes();
+  const LO numNnzEntries     = dofsPerNode * (numCoarseNodes + numInterpolationPoints * (numLocalFineNodes - numCoarseNodes));
+
+  non_const_row_map_type rowPtr("Prolongator graph, rowPtr", dofsPerNode * (numLocalFineNodes + 1));
+  entries_type colIndex("Prolongator graph, colIndices", numNnzEntries);
+
+  *out << "Compute prolongatorGraph data" << std::endl;
+  if (geoData->getInterpolationOrder() == 0) {
+    computeGraphDataConstantFunctor computeGraphData(geoData,
                                                      numCoarseNodes,
                                                      dofsPerNode,
-                                                     numInterpolationPoints,
                                                      geoData->getCoarseningRates(),
                                                      geoData->getCoarseningEndRates(),
                                                      geoData->getLocalFineNodesPerDir(),
-                                                     geoData->getCoarseNodesPerDir(),
                                                      rowPtr,
                                                      colIndex);
-      Kokkos::parallel_for("Structured Aggregation: compute loca graph data",
-                           Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
-                           computeGraphData);
+    Kokkos::parallel_for("Structured Aggregation: compute loca graph data",
+                         Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
+                         computeGraphData);
+  } else if (geoData->getInterpolationOrder() == 1) {
+    // Note, lbv 2018-11-08: in the piece-wise linear case I am computing the rowPtr
+    // using a parallel scan, it might be possible to do something faster than that
+    // by including this calculation in computeGraphDataLinearFunctor but at the moment
+    // all the ideas I have include a bunch of if statements which I would like to avoid.
+    computeGraphRowPtrFunctor computeGraphRowPtr(geoData,
+                                                 dofsPerNode,
+                                                 numInterpolationPoints,
+                                                 numLocalFineNodes,
+                                                 geoData->getCoarseningRates(),
+                                                 geoData->getLocalFineNodesPerDir(),
+                                                 rowPtr);
+    Kokkos::parallel_scan("Structured Aggregation: compute rowPtr for prolongator graph",
+                          Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes + 1),
+                          computeGraphRowPtr);
+
+    computeGraphDataLinearFunctor computeGraphData(geoData,
+                                                   geoData->getNumDimensions(),
+                                                   numCoarseNodes,
+                                                   dofsPerNode,
+                                                   numInterpolationPoints,
+                                                   geoData->getCoarseningRates(),
+                                                   geoData->getCoarseningEndRates(),
+                                                   geoData->getLocalFineNodesPerDir(),
+                                                   geoData->getCoarseNodesPerDir(),
+                                                   rowPtr,
+                                                   colIndex);
+    Kokkos::parallel_for("Structured Aggregation: compute loca graph data",
+                         Kokkos::RangePolicy<execution_space>(0, numLocalFineNodes),
+                         computeGraphData);
+  }
+
+  local_graph_type myLocalGraph(colIndex, rowPtr);
+
+  // Compute graph's colMap and domainMap
+  RCP<Map> colMap, domainMap;
+  *out << "Compute domain and column maps of the CrsGraph" << std::endl;
+  colMap    = MapFactory::Build(graph.GetDomainMap()->lib(),
+                                Teuchos::OrdinalTraits<GO>::invalid(),
+                                numCoarseNodes,
+                                graph.GetDomainMap()->getIndexBase(),
+                                graph.GetDomainMap()->getComm());
+  domainMap = colMap;
+
+  myGraph = CrsGraphFactory::Build(myLocalGraph, graph.GetDomainMap(), colMap,
+                                   colMap, graph.GetDomainMap());
+
+}  // BuildGraph()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    fillAggregatesFunctor::fillAggregatesFunctor(RCP<IndexManager_kokkos> geoData,
+                                                 const int myRank,
+                                                 Kokkos::View<unsigned*, device_type> aggStat,
+                                                 LOVectorView vertex2AggID,
+                                                 LOVectorView procWinner)
+  : geoData_(*geoData)
+  , myRank_(myRank)
+  , aggStat_(aggStat)
+  , vertex2AggID_(vertex2AggID)
+  , procWinner_(procWinner) {}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+KOKKOS_INLINE_FUNCTION void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    fillAggregatesFunctor::operator()(const LO nodeIdx, LO& lNumAggregatedNodes) const {
+  // Compute coarse ID associated with fine LID
+  LO rem, rate;
+  LO coarseNodeCoarseLID;
+  LO nodeFineTuple[3], coarseIdx[3];
+  auto coarseRate       = geoData_.getCoarseningRates();
+  auto endRate          = geoData_.getCoarseningEndRates();
+  auto lFineNodesPerDir = geoData_.getLocalFineNodesPerDir();
+  // Compute coarse ID associated with fine LID
+  geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
+
+  for (int dim = 0; dim < 3; ++dim) {
+    coarseIdx[dim] = nodeFineTuple[dim] / coarseRate(dim);
+    rem            = nodeFineTuple[dim] % coarseRate(dim);
+    rate           = (nodeFineTuple[dim] < lFineNodesPerDir(dim) - endRate(dim)) ? coarseRate(dim) : endRate(dim);
+    if (rem > (rate / 2)) {
+      ++coarseIdx[dim];
     }
-
-    local_graph_type myLocalGraph(colIndex, rowPtr);
-
-    // Compute graph's colMap and domainMap
-    RCP<Map> colMap, domainMap;
-    *out << "Compute domain and column maps of the CrsGraph" << std::endl;
-    colMap = MapFactory::Build(graph.GetDomainMap()->lib(),
-                               Teuchos::OrdinalTraits<GO>::invalid(),
-                               numCoarseNodes,
-                               graph.GetDomainMap()->getIndexBase(),
-                               graph.GetDomainMap()->getComm());
-    domainMap = colMap;
-
-    myGraph = CrsGraphFactory::Build(myLocalGraph, graph.GetDomainMap(), colMap,
-                                     colMap, graph.GetDomainMap());
-
-  } // BuildGraph()
-
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  fillAggregatesFunctor::fillAggregatesFunctor(RCP<IndexManager_kokkos> geoData,
-                                               const int myRank,
-                                               Kokkos::View<unsigned*, device_type> aggStat,
-                                               LOVectorView vertex2AggID,
-                                               LOVectorView procWinner) :
-    geoData_(*geoData), myRank_(myRank), aggStat_(aggStat),
-    vertex2AggID_(vertex2AggID), procWinner_(procWinner) {}
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  KOKKOS_INLINE_FUNCTION
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  fillAggregatesFunctor::operator() (const LO nodeIdx, LO& lNumAggregatedNodes) const {
-    // Compute coarse ID associated with fine LID
-    LO rem, rate;
-    LO coarseNodeCoarseLID;
-    LO nodeFineTuple[3], coarseIdx[3];
-    auto coarseRate          = geoData_.getCoarseningRates();
-    auto endRate             = geoData_.getCoarseningEndRates();
-    auto lFineNodesPerDir    = geoData_.getLocalFineNodesPerDir();
-    // Compute coarse ID associated with fine LID
-    geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
-
-    for(int dim = 0; dim < 3; ++dim) {
-      coarseIdx[dim] = nodeFineTuple[dim] / coarseRate(dim);
-      rem  = nodeFineTuple[dim] % coarseRate(dim);
-      rate = (nodeFineTuple[dim] < lFineNodesPerDir(dim) - endRate(dim)) ? coarseRate(dim) : endRate(dim);
-      if(rem > (rate / 2)) {++coarseIdx[dim];}
+  }
+
+  geoData_.getCoarseTuple2CoarseLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
+                                    coarseNodeCoarseLID);
+
+  vertex2AggID_(nodeIdx, 0) = coarseNodeCoarseLID;
+  procWinner_(nodeIdx, 0)   = myRank_;
+  aggStat_(nodeIdx)         = AGGREGATED;
+  ++lNumAggregatedNodes;
+
+}  // fillAggregatesFunctor::operator()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphDataConstantFunctor::
+        computeGraphDataConstantFunctor(RCP<IndexManager_kokkos> geoData,
+                                        const LO NumGhostedNodes,
+                                        const LO dofsPerNode,
+                                        constIntTupleView coarseRate,
+                                        constIntTupleView endRate,
+                                        constLOTupleView lFineNodesPerDir,
+                                        non_const_row_map_type rowPtr,
+                                        entries_type colIndex)
+  : geoData_(*geoData)
+  , numGhostedNodes_(NumGhostedNodes)
+  , dofsPerNode_(dofsPerNode)
+  , coarseRate_(coarseRate)
+  , endRate_(endRate)
+  , lFineNodesPerDir_(lFineNodesPerDir)
+  , rowPtr_(rowPtr)
+  , colIndex_(colIndex) {
+}  // computeGraphDataConstantFunctor()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+KOKKOS_INLINE_FUNCTION void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphDataConstantFunctor::operator()(const LO nodeIdx) const {
+  LO nodeFineTuple[3]   = {0, 0, 0};
+  LO nodeCoarseTuple[3] = {0, 0, 0};
+
+  // Compute ghosted tuple associated with fine LID
+  geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
+
+  // Compute coarse tuple associated with fine point
+  // then overwrite it with tuple associated with aggregate
+  LO rem, rate, coarseNodeCoarseLID;
+  for (int dim = 0; dim < 3; ++dim) {
+    nodeCoarseTuple[dim] = nodeFineTuple[dim] / coarseRate_(dim);
+    rem                  = nodeFineTuple[dim] % coarseRate_(dim);
+    if (nodeFineTuple[dim] < (lFineNodesPerDir_(dim) - endRate_(dim))) {
+      rate = coarseRate_(dim);
+    } else {
+      rate = endRate_(dim);
     }
-
-    geoData_.getCoarseTuple2CoarseLID(coarseIdx[0], coarseIdx[1], coarseIdx[2],
-                                     coarseNodeCoarseLID);
-
-    vertex2AggID_(nodeIdx, 0) = coarseNodeCoarseLID;
-    procWinner_(nodeIdx, 0)   = myRank_;
-    aggStat_(nodeIdx)         = AGGREGATED;
-    ++lNumAggregatedNodes;
-
-  } // fillAggregatesFunctor::operator()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphDataConstantFunctor::
-  computeGraphDataConstantFunctor(RCP<IndexManager_kokkos> geoData,
-                                  const LO NumGhostedNodes,
-                                  const LO dofsPerNode,
-                                  constIntTupleView coarseRate,
-                                  constIntTupleView endRate,
-                                  constLOTupleView  lFineNodesPerDir,
-                                  non_const_row_map_type rowPtr,
-                                  entries_type colIndex) : geoData_(*geoData),
-    numGhostedNodes_(NumGhostedNodes), dofsPerNode_(dofsPerNode),
-    coarseRate_(coarseRate), endRate_(endRate), lFineNodesPerDir_(lFineNodesPerDir),
-    rowPtr_(rowPtr), colIndex_(colIndex) {
-
-  } // computeGraphDataConstantFunctor()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  KOKKOS_INLINE_FUNCTION
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphDataConstantFunctor::operator() (const LO nodeIdx) const {
-    LO nodeFineTuple[3]   = {0, 0, 0};
-    LO nodeCoarseTuple[3] = {0, 0, 0};
-
-    // Compute ghosted tuple associated with fine LID
-    geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
-
-    // Compute coarse tuple associated with fine point
-    // then overwrite it with tuple associated with aggregate
-    LO rem, rate, coarseNodeCoarseLID;
-    for(int dim = 0; dim < 3; ++dim) {
-      nodeCoarseTuple[dim] = nodeFineTuple[dim] / coarseRate_(dim);
-      rem            = nodeFineTuple[dim] % coarseRate_(dim);
-      if( nodeFineTuple[dim] < (lFineNodesPerDir_(dim) - endRate_(dim)) ) {
-        rate = coarseRate_(dim);
-      } else {
-        rate = endRate_(dim);
-      }
-      if(rem > (rate / 2)) {++nodeCoarseTuple[dim];}
+    if (rem > (rate / 2)) {
+      ++nodeCoarseTuple[dim];
     }
+  }
+
+  // get LID associted with aggregate
+  geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2],
+                                    coarseNodeCoarseLID);
+
+  // store data into CrsGraph taking care of multiple dofs case
+  for (LO dof = 0; dof < dofsPerNode_; ++dof) {
+    rowPtr_(nodeIdx * dofsPerNode_ + dof + 1) = nodeIdx * dofsPerNode_ + dof + 1;
+    colIndex_(nodeIdx * dofsPerNode_ + dof)   = coarseNodeCoarseLID * dofsPerNode_ + dof;
+  }
+
+}  // computeGraphDataConstantFunctor::operator()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphRowPtrFunctor::computeGraphRowPtrFunctor(RCP<IndexManager_kokkos> geoData,
+                                                         const LO dofsPerNode,
+                                                         const int numInterpolationPoints,
+                                                         const LO numLocalRows,
+                                                         constIntTupleView coarseRate,
+                                                         constLOTupleView lFineNodesPerDir,
+                                                         non_const_row_map_type rowPtr)
+  : geoData_(*geoData)
+  , dofsPerNode_(dofsPerNode)
+  , numInterpolationPoints_(numInterpolationPoints)
+  , numLocalRows_(numLocalRows)
+  , coarseRate_(coarseRate)
+  , lFineNodesPerDir_(lFineNodesPerDir)
+  , rowPtr_(rowPtr) {}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+KOKKOS_INLINE_FUNCTION void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphRowPtrFunctor::operator()(const LO rowIdx, GO& update, const bool final) const {
+  if (final) {
+    // Kokkos uses a multipass algorithm to implement scan.
+    // Only update the array on the final pass.  Updating the
+    // array before changing 'update' means that we do an
+    // exclusive scan.  Update the array after for an inclusive
+    // scan.
+    rowPtr_(rowIdx) = update;
+  }
+  if (rowIdx < numLocalRows_) {
+    LO nodeIdx          = rowIdx / dofsPerNode_;
+    bool allCoarse      = true;
+    LO nodeFineTuple[3] = {0, 0, 0};
+    geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
+    for (int dim = 0; dim < 3; ++dim) {
+      const LO rem = nodeFineTuple[dim] % coarseRate_(dim);
 
-    // get LID associted with aggregate
-    geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2],
-                                      coarseNodeCoarseLID);
-
-    // store data into CrsGraph taking care of multiple dofs case
-    for(LO dof = 0; dof < dofsPerNode_; ++dof) {
-      rowPtr_(nodeIdx*dofsPerNode_ + dof + 1) = nodeIdx*dofsPerNode_ + dof + 1;
-      colIndex_(nodeIdx*dofsPerNode_ + dof)   = coarseNodeCoarseLID*dofsPerNode_ + dof;
+      // Check if Fine node lies on Coarse Node
+      allCoarse = (allCoarse && ((rem == 0) || (nodeFineTuple[dim] == lFineNodesPerDir_(dim) - 1)));
     }
-
-  } // computeGraphDataConstantFunctor::operator()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphRowPtrFunctor::computeGraphRowPtrFunctor(RCP<IndexManager_kokkos> geoData,
-                                                       const LO dofsPerNode,
-                                                       const int numInterpolationPoints,
-                                                       const LO  numLocalRows,
-                                                       constIntTupleView coarseRate,
-                                                       constLOTupleView  lFineNodesPerDir,
-                                                       non_const_row_map_type rowPtr) :
-  geoData_(*geoData), dofsPerNode_(dofsPerNode),
-    numInterpolationPoints_(numInterpolationPoints), numLocalRows_(numLocalRows),
-    coarseRate_(coarseRate), lFineNodesPerDir_(lFineNodesPerDir), rowPtr_(rowPtr) {}
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  KOKKOS_INLINE_FUNCTION
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphRowPtrFunctor::operator() (const LO rowIdx, GO& update, const bool final) const {
-    if (final) {
-      // Kokkos uses a multipass algorithm to implement scan.
-      // Only update the array on the final pass.  Updating the
-      // array before changing 'update' means that we do an
-      // exclusive scan.  Update the array after for an inclusive
-      // scan.
-      rowPtr_(rowIdx) = update;
+    update += (allCoarse ? 1 : numInterpolationPoints_);
+  }
+}  // computeGraphRowPtrFunctor::operator()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphDataLinearFunctor::computeGraphDataLinearFunctor(RCP<IndexManager_kokkos> geoData,
+                                                                 const int numDimensions,
+                                                                 const LO numGhostedNodes,
+                                                                 const LO dofsPerNode,
+                                                                 const int numInterpolationPoints,
+                                                                 constIntTupleView coarseRate,
+                                                                 constIntTupleView endRate,
+                                                                 constLOTupleView lFineNodesPerDir,
+                                                                 constLOTupleView ghostedNodesPerDir,
+                                                                 non_const_row_map_type rowPtr,
+                                                                 entries_type colIndex)
+  : geoData_(*geoData)
+  , numDimensions_(numDimensions)
+  , numGhostedNodes_(numGhostedNodes)
+  , dofsPerNode_(dofsPerNode)
+  , numInterpolationPoints_(numInterpolationPoints)
+  , coarseRate_(coarseRate)
+  , endRate_(endRate)
+  , lFineNodesPerDir_(lFineNodesPerDir)
+  , ghostedNodesPerDir_(ghostedNodesPerDir)
+  , rowPtr_(rowPtr)
+  , colIndex_(colIndex) {
+}  // computeGraphDataLinearFunctor()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+KOKKOS_INLINE_FUNCTION void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGraphDataLinearFunctor::operator()(const LO nodeIdx) const {
+  LO nodeFineTuple[3]   = {0, 0, 0};
+  LO nodeCoarseTuple[3] = {0, 0, 0};
+
+  // Compute coarse ID associated with fine LID
+  geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
+
+  LO coarseNodeCoarseLID;
+  bool allCoarse = false;
+  for (int dim = 0; dim < 3; ++dim) {
+    nodeCoarseTuple[dim] = nodeFineTuple[dim] / coarseRate_(dim);
+  }
+  if (rowPtr_(nodeIdx + 1) == rowPtr_(nodeIdx) + 1) {
+    allCoarse = true;
+  }
+
+  geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2],
+                                    coarseNodeCoarseLID);
+
+  if (allCoarse) {
+    // Fine node lies on Coarse node, easy case, we only need the LID of the coarse node.
+    for (LO dof = 0; dof < dofsPerNode_; ++dof) {
+      colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof)) = coarseNodeCoarseLID * dofsPerNode_ + dof;
     }
-    if (rowIdx < numLocalRows_) {
-      LO nodeIdx = rowIdx / dofsPerNode_;
-      bool allCoarse = true;
-      LO nodeFineTuple[3] = {0, 0, 0};
-      geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
-      for(int dim = 0; dim < 3; ++dim) {
-        const LO rem = nodeFineTuple[dim] % coarseRate_(dim);
-
-        // Check if Fine node lies on Coarse Node
-        allCoarse = (allCoarse && ((rem == 0) || (nodeFineTuple[dim] == lFineNodesPerDir_(dim) - 1)));
+  } else {
+    for (int dim = 0; dim < numDimensions_; ++dim) {
+      if (nodeCoarseTuple[dim] == ghostedNodesPerDir_(dim) - 1) {
+        --nodeCoarseTuple[dim];
       }
-      update += (allCoarse ? 1 : numInterpolationPoints_);
     }
-  } // computeGraphRowPtrFunctor::operator()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphDataLinearFunctor::computeGraphDataLinearFunctor(RCP<IndexManager_kokkos> geoData,
-                                                               const int numDimensions,
-                                                               const LO numGhostedNodes,
-                                                               const LO dofsPerNode,
-                                                               const int numInterpolationPoints,
-                                                               constIntTupleView coarseRate,
-                                                               constIntTupleView endRate,
-                                                               constLOTupleView  lFineNodesPerDir,
-                                                               constLOTupleView  ghostedNodesPerDir,
-                                                               non_const_row_map_type rowPtr,
-                                                               entries_type colIndex) :
-  geoData_(*geoData), numDimensions_(numDimensions),
-    numGhostedNodes_(numGhostedNodes),
-    dofsPerNode_(dofsPerNode), numInterpolationPoints_(numInterpolationPoints),
-    coarseRate_(coarseRate), endRate_(endRate), lFineNodesPerDir_(lFineNodesPerDir),
-    ghostedNodesPerDir_(ghostedNodesPerDir), rowPtr_(rowPtr), colIndex_(colIndex) {
-
-  } // computeGraphDataLinearFunctor()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  KOKKOS_INLINE_FUNCTION
-  void AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGraphDataLinearFunctor::operator() (const LO nodeIdx) const {
-    LO nodeFineTuple[3]   = {0, 0, 0};
-    LO nodeCoarseTuple[3] = {0, 0, 0};
-
-    // Compute coarse ID associated with fine LID
-    geoData_.getFineLID2FineTuple(nodeIdx, nodeFineTuple);
-
-    LO coarseNodeCoarseLID;
-    bool allCoarse = false;
-    for(int dim = 0; dim < 3; ++dim) {
-      nodeCoarseTuple[dim] = nodeFineTuple[dim] / coarseRate_(dim);
-    }
-    if(rowPtr_(nodeIdx + 1) == rowPtr_(nodeIdx) + 1) {allCoarse = true;}
-
-    geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2],
-                                      coarseNodeCoarseLID);
-
-    if(allCoarse) {
-      // Fine node lies on Coarse node, easy case, we only need the LID of the coarse node.
-      for(LO dof = 0; dof < dofsPerNode_; ++dof) {
-        colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)) = coarseNodeCoarseLID*dofsPerNode_ + dof;
-      }
-    } else {
-
-      for(int dim = 0; dim < numDimensions_; ++dim) {
-        if(nodeCoarseTuple[dim] == ghostedNodesPerDir_(dim) - 1) { --nodeCoarseTuple[dim]; }
-      }
-      // Compute Coarse Node LID
-      // Note lbv 10-06-2018: it is likely benefitial to remove the two if statments and somehow
-      // find out the number of dimensions before calling the opertor() of the functor.
-      for(LO dof = 0; dof < dofsPerNode_; ++dof) {
-        geoData_.getCoarseTuple2CoarseLID(    nodeCoarseTuple[0],   nodeCoarseTuple[1],   nodeCoarseTuple[2],   colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+0));
-        geoData_.getCoarseTuple2CoarseLID(    nodeCoarseTuple[0]+1, nodeCoarseTuple[1],   nodeCoarseTuple[2],   colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+1));
-        if(numDimensions_ > 1) {
-          geoData_.getCoarseTuple2CoarseLID(  nodeCoarseTuple[0],   nodeCoarseTuple[1]+1, nodeCoarseTuple[2],   colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+2));
-          geoData_.getCoarseTuple2CoarseLID(  nodeCoarseTuple[0]+1, nodeCoarseTuple[1]+1, nodeCoarseTuple[2],   colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+3));
-          if(numDimensions_ > 2) {
-            geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0],   nodeCoarseTuple[1],   nodeCoarseTuple[2]+1, colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+4));
-            geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0]+1, nodeCoarseTuple[1],   nodeCoarseTuple[2]+1, colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+5));
-            geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0],   nodeCoarseTuple[1]+1, nodeCoarseTuple[2]+1, colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+6));
-            geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0]+1, nodeCoarseTuple[1]+1, nodeCoarseTuple[2]+1, colIndex_(rowPtr_(nodeIdx*dofsPerNode_ + dof)+7));
-          }
+    // Compute Coarse Node LID
+    // Note lbv 10-06-2018: it is likely benefitial to remove the two if statments and somehow
+    // find out the number of dimensions before calling the opertor() of the functor.
+    for (LO dof = 0; dof < dofsPerNode_; ++dof) {
+      geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2], colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 0));
+      geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0] + 1, nodeCoarseTuple[1], nodeCoarseTuple[2], colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 1));
+      if (numDimensions_ > 1) {
+        geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1] + 1, nodeCoarseTuple[2], colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 2));
+        geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0] + 1, nodeCoarseTuple[1] + 1, nodeCoarseTuple[2], colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 3));
+        if (numDimensions_ > 2) {
+          geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1], nodeCoarseTuple[2] + 1, colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 4));
+          geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0] + 1, nodeCoarseTuple[1], nodeCoarseTuple[2] + 1, colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 5));
+          geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0], nodeCoarseTuple[1] + 1, nodeCoarseTuple[2] + 1, colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 6));
+          geoData_.getCoarseTuple2CoarseLID(nodeCoarseTuple[0] + 1, nodeCoarseTuple[1] + 1, nodeCoarseTuple[2] + 1, colIndex_(rowPtr_(nodeIdx * dofsPerNode_ + dof) + 7));
         }
       }
     }
-  } // computeGraphDataLinearFunctor::operator()
-
-} // end namespace
+  }
+}  // computeGraphDataLinearFunctor::operator()
 
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONSTRUCTUREDALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_decl.hpp
index 28d26813f5d6..04d360374190 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_decl.hpp
@@ -75,178 +75,174 @@ namespace MueLu {
     and local lexicographic mesh orderings are supported.
 */
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IndexManager : public BaseClass {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IndexManager : public BaseClass {
 #undef MUELU_INDEXMANAGER_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  private:
+ private:
+ protected:
+  const RCP<const Teuchos::Comm<int> > comm_;  ///< Communicator used by uncoupled aggregation
+  const bool coupled_;                         ///< Flag for coupled vs uncoupled aggregation mode, if true aggregation is coupled.
+  const bool singleCoarsePoint_;               ///< Flag telling us if can reduce dimensions to a single layer.
+  const int numDimensions;                     ///< Number of spacial dimensions in the problem
+  const int interpolationOrder_;               ///< Interpolation order used by grid transfer operators using these aggregates.
 
-  protected:
+  Array<int> coarseRate;  ///< coarsening rate in each direction
+  Array<int> endRate;     ///< adapted coarsening rate at the edge of the mesh in each direction.
 
-    const RCP<const Teuchos::Comm<int> > comm_; ///< Communicator used by uncoupled aggregation
-    const bool coupled_;                ///< Flag for coupled vs uncoupled aggregation mode, if true aggregation is coupled.
-    const bool singleCoarsePoint_;      ///< Flag telling us if can reduce dimensions to a single layer.
-    const int numDimensions;            ///< Number of spacial dimensions in the problem
-    const int interpolationOrder_;      ///< Interpolation order used by grid transfer operators using these aggregates.
+  GO gNumFineNodes;                  ///< global number of nodes.
+  GO gNumFineNodes10;                ///< global number of nodes per 0-1 slice.
+  const Array<GO> gFineNodesPerDir;  ///< global number of nodes per direction.
 
-    Array<int> coarseRate;              ///< coarsening rate in each direction
-    Array<int> endRate;                 ///< adapted coarsening rate at the edge of the mesh in each direction.
+  LO lNumFineNodes;                  ///< local number of nodes.
+  LO lNumFineNodes10;                ///< local number of nodes per 0-1 slice.
+  const Array<LO> lFineNodesPerDir;  ///< local number of nodes per direction.
 
-    GO gNumFineNodes;                   ///< global number of nodes.
-    GO gNumFineNodes10;                 ///< global number of nodes per 0-1 slice.
-    const Array<GO> gFineNodesPerDir;   ///< global number of nodes per direction.
+  GO gNumCoarseNodes;            ///< global number of nodes remaining after coarsening.
+  GO gNumCoarseNodes10;          ///< global number of nodes per 0-1 slice remaining after coarsening.
+  Array<GO> gCoarseNodesPerDir;  ///< global number of nodes per direction remaining after coarsening.
 
-    LO lNumFineNodes;                   ///< local number of nodes.
-    LO lNumFineNodes10;                 ///< local number of nodes per 0-1 slice.
-    const Array<LO> lFineNodesPerDir;   ///< local number of nodes per direction.
+  LO lNumCoarseNodes;            ///< local number of nodes remaining after coarsening.
+  LO lNumCoarseNodes10;          ///< local number of nodes per 0-1 slice remaining after coarsening.
+  Array<LO> lCoarseNodesPerDir;  ///< local number of nodes per direction remaing after coarsening.
 
-    GO gNumCoarseNodes;                 ///< global number of nodes remaining after coarsening.
-    GO gNumCoarseNodes10;               ///< global number of nodes per 0-1 slice remaining after coarsening.
-    Array<GO> gCoarseNodesPerDir;       ///< global number of nodes per direction remaining after coarsening.
+  LO numGhostNodes;              ///< local number of ghost nodes
+  LO numGhostedNodes;            ///< local number of ghosted nodes (i.e. ghost + coarse nodes).
+  LO numGhostedNodes10;          ///< local number of ghosted nodes (i.e. ghost + coarse nodes) per 0-1 slice.
+  Array<LO> ghostedNodesPerDir;  ///< local number of ghosted nodes (i.e. ghost + coarse nodes) per direction
 
-    LO lNumCoarseNodes;                 ///< local number of nodes remaining after coarsening.
-    LO lNumCoarseNodes10;               ///< local number of nodes per 0-1 slice remaining after coarsening.
-    Array<LO> lCoarseNodesPerDir;       ///< local number of nodes per direction remaing after coarsening.
+  GO minGlobalIndex;                 ///< lowest GID of any node in the local process
+  Array<LO> offsets;                 ///< distance between lowest (resp. highest) index to the lowest (resp. highest) ghostedNodeIndex in that direction.
+  Array<LO> coarseNodeOffsets;       ///< distance between lowest (resp. highest) index to the lowest (resp. highest) coarseNodeIndex in that direction.
+  Array<GO> startIndices;            ///< lowest global tuple (i,j,k) of a node on the local process
+  Array<GO> startGhostedCoarseNode;  ///< lowest coarse global tuple (i,j,k) of a node remaing on the local process after coarsening.
 
-    LO numGhostNodes;                   ///< local number of ghost nodes
-    LO numGhostedNodes;                 ///< local number of ghosted nodes (i.e. ghost + coarse nodes).
-    LO numGhostedNodes10;               ///< local number of ghosted nodes (i.e. ghost + coarse nodes) per 0-1 slice.
-    Array<LO> ghostedNodesPerDir;       ///< local number of ghosted nodes (i.e. ghost + coarse nodes) per direction
+  bool meshEdge[6]       = {false};  ///< flags indicating if we run into the edge of the mesh in ilo, ihi, jlo, jhi, klo or khi.
+  bool ghostInterface[6] = {false};  ///< flags indicating if ghost points are needed at ilo, ihi, jlo, jhi, klo and khi boundaries.
+  bool ghostedDir[6]     = {false};  ///< flags indicating if ghost points are needed at ilo, ihi, jlo, jhi, klo and khi boundaries.
 
-    GO minGlobalIndex;                  ///< lowest GID of any node in the local process
-    Array<LO> offsets;                  ///< distance between lowest (resp. highest) index to the lowest (resp. highest) ghostedNodeIndex in that direction.
-    Array<LO> coarseNodeOffsets;        ///< distance between lowest (resp. highest) index to the lowest (resp. highest) coarseNodeIndex in that direction.
-    Array<GO> startIndices;             ///< lowest global tuple (i,j,k) of a node on the local process
-    Array<GO> startGhostedCoarseNode;   ///< lowest coarse global tuple (i,j,k) of a node remaing on the local process after coarsening.
+ public:
+  IndexManager() = default;
 
-    bool meshEdge[6] =       {false};   ///< flags indicating if we run into the edge of the mesh in ilo, ihi, jlo, jhi, klo or khi.
-    bool ghostInterface[6] = {false};   ///< flags indicating if ghost points are needed at ilo, ihi, jlo, jhi, klo and khi boundaries.
-    bool ghostedDir[6] =     {false};   ///< flags indicating if ghost points are needed at ilo, ihi, jlo, jhi, klo and khi boundaries.
+  IndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+               const bool singleCoarsePoint, const int NumDimensions,
+               const int interpolationOrder, const Array<GO> GFineNodesPerDir,
+               const Array<LO> LFineNodesPerDir);
 
-  public:
+  virtual ~IndexManager() {}
 
-    IndexManager() = default;
+  //! Sets basic parameters used to compute indices on the mesh.
+  //! This method requires you to have set this->coarseRate and this->startIndices.
+  void computeMeshParameters();
 
-    IndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                 const bool singleCoarsePoint, const int NumDimensions,
-                 const int interpolationOrder, const Array<GO> GFineNodesPerDir,
-                 const Array<LO> LFineNodesPerDir);
+  virtual void computeGlobalCoarseParameters() = 0;
 
-    virtual ~IndexManager() {}
+  virtual void getGhostedNodesData(const RCP<const Map> fineMap,
+                                   Array<LO>& ghostedNodeCoarseLIDs,
+                                   Array<int>& ghostedNodeCoarsePIDs,
+                                   Array<GO>& ghostedNodeCoarseGIDs) const = 0;
 
-    //! Sets basic parameters used to compute indices on the mesh.
-    //! This method requires you to have set this->coarseRate and this->startIndices.
-    void computeMeshParameters();
+  virtual void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                                  Array<GO>& coarseNodeCoarseGIDs,
+                                  Array<GO>& coarseNodeFineGIDs) const = 0;
 
-    virtual void computeGlobalCoarseParameters() = 0;
+  bool isAggregationCoupled() const { return coupled_; }
 
-    virtual void getGhostedNodesData(const RCP<const Map> fineMap,
-                                     Array<LO>&  ghostedNodeCoarseLIDs,
-                                     Array<int>& ghostedNodeCoarsePIDs,
-                                     Array<GO>&  ghostedNodeCoarseGIDs) const = 0;
+  bool isSingleCoarsePoint() const { return singleCoarsePoint_; }
 
-    virtual void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                                    Array<GO>& coarseNodeCoarseGIDs,
-                                    Array<GO>& coarseNodeFineGIDs) const = 0;
+  int getNumDimensions() const { return numDimensions; }
 
-    bool isAggregationCoupled() const {return coupled_;}
+  int getInterpolationOrder() const { return interpolationOrder_; }
 
-    bool isSingleCoarsePoint() const {return singleCoarsePoint_;}
+  GO getNumGlobalFineNodes() const { return gNumFineNodes; }
 
-    int getNumDimensions() const {return numDimensions;}
+  GO getNumGlobalCoarseNodes() const { return gNumCoarseNodes; }
 
-    int getInterpolationOrder() const {return interpolationOrder_;}
+  LO getNumLocalFineNodes() const { return lNumFineNodes; }
 
-    GO getNumGlobalFineNodes() const {return gNumFineNodes;}
+  LO getNumLocalCoarseNodes() const { return lNumCoarseNodes; }
 
-    GO getNumGlobalCoarseNodes() const {return gNumCoarseNodes;}
+  LO getNumLocalGhostedNodes() const { return numGhostedNodes; }
 
-    LO getNumLocalFineNodes() const {return lNumFineNodes;}
+  Array<int> getCoarseningRates() const { return coarseRate; }
 
-    LO getNumLocalCoarseNodes() const {return lNumCoarseNodes;}
+  int getCoarseningRate(const int dim) const { return coarseRate[dim]; }
 
-    LO getNumLocalGhostedNodes() const {return numGhostedNodes;}
+  Array<int> getCoarseningEndRates() const { return endRate; }
 
-    Array<int> getCoarseningRates() const {return coarseRate;}
+  int getCoarseningEndRate(const int dim) const { return endRate[dim]; }
 
-    int getCoarseningRate(const int dim) const {return coarseRate[dim];}
+  bool getMeshEdge(const int dir) const { return meshEdge[dir]; }
 
-    Array<int> getCoarseningEndRates() const {return endRate;}
+  bool getGhostInterface(const int dir) const { return ghostInterface[dir]; }
 
-    int getCoarseningEndRate(const int dim) const {return endRate[dim];}
+  Array<LO> getOffsets() const { return offsets; }
 
-    bool getMeshEdge(const int dir) const {return meshEdge[dir];}
+  LO getOffset(int const dim) const { return offsets[dim]; }
 
-    bool getGhostInterface(const int dir) const {return ghostInterface[dir];}
+  Array<LO> getCoarseNodeOffsets() const { return coarseNodeOffsets; }
 
-    Array<LO> getOffsets() const {return offsets;}
+  LO getCoarseNodeOffset(int const dim) const { return coarseNodeOffsets[dim]; }
 
-    LO getOffset(int const dim) const {return offsets[dim];}
+  Array<GO> getStartIndices() const { return startIndices; }
 
-    Array<LO> getCoarseNodeOffsets() const {return coarseNodeOffsets;}
+  GO getStartIndex(int const dim) const { return startIndices[dim]; }
 
-    LO getCoarseNodeOffset(int const dim) const {return coarseNodeOffsets[dim];}
+  Array<GO> getStartGhostedCoarseNodes() const { return startGhostedCoarseNode; }
 
-    Array<GO> getStartIndices() const {return startIndices;}
+  GO getStartGhostedCoarseNode(int const dim) const { return startGhostedCoarseNode[dim]; }
 
-    GO getStartIndex(int const dim) const {return startIndices[dim];}
+  Array<LO> getLocalFineNodesPerDir() const { return lFineNodesPerDir; }
 
-    Array<GO> getStartGhostedCoarseNodes() const {return startGhostedCoarseNode;}
+  LO getLocalFineNodesInDir(const int dim) const { return lFineNodesPerDir[dim]; }
 
-    GO getStartGhostedCoarseNode(int const dim) const {return startGhostedCoarseNode[dim];}
+  Array<GO> getGlobalFineNodesPerDir() const { return gFineNodesPerDir; }
 
-    Array<LO> getLocalFineNodesPerDir() const {return lFineNodesPerDir;}
+  GO getGlobalFineNodesInDir(const int dim) const { return gFineNodesPerDir[dim]; }
 
-    LO getLocalFineNodesInDir(const int dim) const {return lFineNodesPerDir[dim];}
+  Array<LO> getLocalCoarseNodesPerDir() const { return lCoarseNodesPerDir; }
 
-    Array<GO> getGlobalFineNodesPerDir() const {return gFineNodesPerDir;}
+  LO getLocalCoarseNodesInDir(const int dim) const { return lCoarseNodesPerDir[dim]; }
 
-    GO getGlobalFineNodesInDir(const int dim) const {return gFineNodesPerDir[dim];}
+  Array<GO> getGlobalCoarseNodesPerDir() const { return gCoarseNodesPerDir; }
 
-    Array<LO> getLocalCoarseNodesPerDir() const {return lCoarseNodesPerDir;}
+  GO getGlobalCoarseNodesInDir(const int dim) const { return gCoarseNodesPerDir[dim]; }
 
-    LO getLocalCoarseNodesInDir(const int dim) const {return lCoarseNodesPerDir[dim];}
+  Array<LO> getGhostedNodesPerDir() const { return ghostedNodesPerDir; }
 
-    Array<GO> getGlobalCoarseNodesPerDir() const {return gCoarseNodesPerDir;}
+  LO getGhostedNodesInDir(const int dim) const { return ghostedNodesPerDir[dim]; }
 
-    GO getGlobalCoarseNodesInDir(const int dim) const {return gCoarseNodesPerDir[dim];}
+  virtual std::vector<std::vector<GO> > getCoarseMeshData() const = 0;
 
-    Array<LO> getGhostedNodesPerDir() const {return ghostedNodesPerDir;}
+  virtual void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const = 0;
 
-    LO getGhostedNodesInDir(const int dim) const {return ghostedNodesPerDir[dim];}
+  virtual void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
 
-    virtual std::vector<std::vector<GO> > getCoarseMeshData() const = 0;
+  virtual void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
 
-    virtual void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const = 0;
+  virtual void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const = 0;
 
-    virtual void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
+  virtual void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
 
-    virtual void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
+  virtual void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const = 0;
 
-    virtual void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const = 0;
+  virtual void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
 
-    virtual void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
+  virtual void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const = 0;
 
-    virtual void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const = 0;
+  virtual void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
 
-    virtual void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const = 0;
+  virtual void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
 
-    virtual void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const = 0;
+  virtual void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
 
-    virtual void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
+  virtual void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
 
-    virtual void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
+  virtual void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
+};
 
-    virtual void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
-
-    virtual void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
-
-    virtual void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const = 0;
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INDEXMANAGER_SHORT
-#endif // MUELU_INDEXMANAGER_DECL_HPP
+#endif  // MUELU_INDEXMANAGER_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_def.hpp
index 894e26393906..50df1f9f9dd5 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_def.hpp
@@ -57,220 +57,233 @@
 
 namespace MueLu {
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  IndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  IndexManager(const RCP<const Teuchos::Comm<int> > comm,
-               const bool coupled,
-               const bool singleCoarsePoint,
-               const int NumDimensions,
-               const int interpolationOrder,
-               const Array<GO> GFineNodesPerDir,
-               const Array<LO> LFineNodesPerDir) :
-    comm_(comm), coupled_(coupled), singleCoarsePoint_(singleCoarsePoint),
-    numDimensions(NumDimensions), interpolationOrder_(interpolationOrder),
-    gFineNodesPerDir(GFineNodesPerDir), lFineNodesPerDir(LFineNodesPerDir) {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+IndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    IndexManager(const RCP<const Teuchos::Comm<int> > comm,
+                 const bool coupled,
+                 const bool singleCoarsePoint,
+                 const int NumDimensions,
+                 const int interpolationOrder,
+                 const Array<GO> GFineNodesPerDir,
+                 const Array<LO> LFineNodesPerDir)
+  : comm_(comm)
+  , coupled_(coupled)
+  , singleCoarsePoint_(singleCoarsePoint)
+  , numDimensions(NumDimensions)
+  , interpolationOrder_(interpolationOrder)
+  , gFineNodesPerDir(GFineNodesPerDir)
+  , lFineNodesPerDir(LFineNodesPerDir) {
+  coarseRate.resize(3);
+  endRate.resize(3);
+  gCoarseNodesPerDir.resize(3);
+  lCoarseNodesPerDir.resize(3);
+  ghostedNodesPerDir.resize(3);
 
-    coarseRate.resize(3);
-    endRate.resize(3);
-    gCoarseNodesPerDir.resize(3);
-    lCoarseNodesPerDir.resize(3);
-    ghostedNodesPerDir.resize(3);
+  offsets.resize(3);
+  coarseNodeOffsets.resize(3);
+  startIndices.resize(6);
+  startGhostedCoarseNode.resize(3);
 
-    offsets.resize(3);
-    coarseNodeOffsets.resize(3);
-    startIndices.resize(6);
-    startGhostedCoarseNode.resize(3);
+}  // Constructor
 
-  } // Constructor
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    computeMeshParameters() {
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  computeMeshParameters() {
+  if (coupled_) {
+    gNumFineNodes10 = gFineNodesPerDir[1] * gFineNodesPerDir[0];
+    gNumFineNodes   = gFineNodesPerDir[2] * gNumFineNodes10;
+  } else {
+    gNumFineNodes10 = Teuchos::OrdinalTraits<GO>::invalid();
+    gNumFineNodes   = Teuchos::OrdinalTraits<GO>::invalid();
+  }
+  lNumFineNodes10 = lFineNodesPerDir[1] * lFineNodesPerDir[0];
+  lNumFineNodes   = lFineNodesPerDir[2] * lNumFineNodes10;
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      if (coupled_) {
+        if (startIndices[dim] == 0) {
+          meshEdge[2 * dim] = true;
+        }
+        if (startIndices[dim + 3] + 1 == gFineNodesPerDir[dim]) {
+          meshEdge[2 * dim + 1] = true;
+          endRate[dim]          = startIndices[dim + 3] % coarseRate[dim];
+        }
+      } else {  // With uncoupled problem each rank might require a different endRate
+        meshEdge[2 * dim]     = true;
+        meshEdge[2 * dim + 1] = true;
+        endRate[dim]          = (lFineNodesPerDir[dim] - 1) % coarseRate[dim];
+      }
+      if (endRate[dim] == 0) {
+        endRate[dim] = coarseRate[dim];
+      }
 
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
+      // If uncoupled aggregation is used, offsets[dim] = 0, so nothing to do.
+      if (coupled_) {
+        offsets[dim] = Teuchos::as<LO>(startIndices[dim]) % coarseRate[dim];
+        if (offsets[dim] == 0) {
+          coarseNodeOffsets[dim] = 0;
+        } else if (startIndices[dim] + endRate[dim] == lFineNodesPerDir[dim]) {
+          coarseNodeOffsets[dim] = endRate[dim] - offsets[dim];
+        } else {
+          coarseNodeOffsets[dim] = coarseRate[dim] - offsets[dim];
+        }
 
-    if(coupled_) {
-      gNumFineNodes10 = gFineNodesPerDir[1]*gFineNodesPerDir[0];
-      gNumFineNodes   = gFineNodesPerDir[2]*gNumFineNodes10;
-    } else {
-      gNumFineNodes10 = Teuchos::OrdinalTraits<GO>::invalid();
-      gNumFineNodes   = Teuchos::OrdinalTraits<GO>::invalid();
-    }
-    lNumFineNodes10 = lFineNodesPerDir[1]*lFineNodesPerDir[0];
-    lNumFineNodes   = lFineNodesPerDir[2]*lNumFineNodes10;
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        if(coupled_) {
-          if(startIndices[dim] == 0) {
-            meshEdge[2*dim] = true;
+        if (interpolationOrder_ == 0) {
+          int rem = startIndices[dim] % coarseRate[dim];
+          if ((rem != 0) && (rem <= Teuchos::as<double>(coarseRate[dim]) / 2.0)) {
+            ghostInterface[2 * dim] = true;
           }
-          if(startIndices[dim + 3] + 1 == gFineNodesPerDir[dim]) {
-            meshEdge[2*dim + 1] = true;
-            endRate[dim] = startIndices[dim + 3] % coarseRate[dim];
+          rem = startIndices[dim + 3] % coarseRate[dim];
+          // uncoupled by nature does not require ghosts nodes
+          if (coupled_ && (startIndices[dim + 3] != gFineNodesPerDir[dim] - 1) &&
+              (rem > Teuchos::as<double>(coarseRate[dim]) / 2.0)) {
+            ghostInterface[2 * dim + 1] = true;
           }
-        } else { // With uncoupled problem each rank might require a different endRate
-          meshEdge[2*dim]     = true;
-          meshEdge[2*dim + 1] = true;
-          endRate[dim] = (lFineNodesPerDir[dim] - 1) % coarseRate[dim];
-        }
-        if(endRate[dim] == 0) {endRate[dim] = coarseRate[dim];}
 
-        // If uncoupled aggregation is used, offsets[dim] = 0, so nothing to do.
-        if(coupled_) {
-          offsets[dim] = Teuchos::as<LO>(startIndices[dim]) % coarseRate[dim];
-          if(offsets[dim] == 0) {
-            coarseNodeOffsets[dim] = 0;
-          } else if(startIndices[dim] + endRate[dim] == lFineNodesPerDir[dim]) {
-            coarseNodeOffsets[dim] = endRate[dim] - offsets[dim];
-          } else {
-            coarseNodeOffsets[dim] = coarseRate[dim] - offsets[dim];
+        } else if (interpolationOrder_ == 1) {
+          if (coupled_ && (startIndices[dim] % coarseRate[dim] != 0 ||
+                           startIndices[dim] == gFineNodesPerDir[dim] - 1)) {
+            ghostInterface[2 * dim] = true;
           }
-
-          if(interpolationOrder_ == 0) {
-            int rem  = startIndices[dim] % coarseRate[dim];
-            if( (rem != 0) && (rem <= Teuchos::as<double>(coarseRate[dim]) / 2.0)) {
-              ghostInterface[2*dim] = true;
-            }
-            rem  = startIndices[dim + 3] % coarseRate[dim];
-            // uncoupled by nature does not require ghosts nodes
-            if(coupled_ && (startIndices[dim + 3] != gFineNodesPerDir[dim] - 1) &&
-               (rem > Teuchos::as<double>(coarseRate[dim]) / 2.0)) {
-              ghostInterface[2*dim + 1] = true;
-            }
-
-          } else if(interpolationOrder_ == 1) {
-            if(coupled_ && (startIndices[dim] % coarseRate[dim] != 0 ||
-                            startIndices[dim] == gFineNodesPerDir[dim]-1)) {
-              ghostInterface[2*dim] = true;
-            }
-            if(coupled_ && (startIndices[dim + 3] != gFineNodesPerDir[dim] - 1) &&
-               ((lFineNodesPerDir[dim] == 1) || (startIndices[dim + 3] % coarseRate[dim] != 0))) {
-              ghostInterface[2*dim+1] = true;
-            }
+          if (coupled_ && (startIndices[dim + 3] != gFineNodesPerDir[dim] - 1) &&
+              ((lFineNodesPerDir[dim] == 1) || (startIndices[dim + 3] % coarseRate[dim] != 0))) {
+            ghostInterface[2 * dim + 1] = true;
           }
         }
-      } else { // Default value for dim >= numDimensions
-        endRate[dim] = 1;
       }
+    } else {  // Default value for dim >= numDimensions
+      endRate[dim] = 1;
     }
+  }
 
-    *out << "singleCoarsePoint? " << singleCoarsePoint_ << std::endl;
-    *out << "gFineNodesPerDir: " << gFineNodesPerDir << std::endl;
-    *out << "lFineNodesPerDir: " << lFineNodesPerDir << std::endl;
-    *out << "endRate: " << endRate << std::endl;
-    *out << "ghostInterface: {" << ghostInterface[0] << ", " << ghostInterface[1] << ", "
-              << ghostInterface[2] << ", " << ghostInterface[3] << ", " << ghostInterface[4] << ", "
-              << ghostInterface[5] << "}" << std::endl;
-    *out << "meshEdge: {" << meshEdge[0] << ", " << meshEdge[1] << ", "
-              << meshEdge[2] << ", " << meshEdge[3] << ", " << meshEdge[4] << ", "
-              << meshEdge[5] << "}" << std::endl;
-    *out << "startIndices: " << startIndices << std::endl;
-    *out << "offsets: " << offsets << std::endl;
-    *out << "coarseNodeOffsets: " << coarseNodeOffsets << std::endl;
+  *out << "singleCoarsePoint? " << singleCoarsePoint_ << std::endl;
+  *out << "gFineNodesPerDir: " << gFineNodesPerDir << std::endl;
+  *out << "lFineNodesPerDir: " << lFineNodesPerDir << std::endl;
+  *out << "endRate: " << endRate << std::endl;
+  *out << "ghostInterface: {" << ghostInterface[0] << ", " << ghostInterface[1] << ", "
+       << ghostInterface[2] << ", " << ghostInterface[3] << ", " << ghostInterface[4] << ", "
+       << ghostInterface[5] << "}" << std::endl;
+  *out << "meshEdge: {" << meshEdge[0] << ", " << meshEdge[1] << ", "
+       << meshEdge[2] << ", " << meshEdge[3] << ", " << meshEdge[4] << ", "
+       << meshEdge[5] << "}" << std::endl;
+  *out << "startIndices: " << startIndices << std::endl;
+  *out << "offsets: " << offsets << std::endl;
+  *out << "coarseNodeOffsets: " << coarseNodeOffsets << std::endl;
 
-    // Here one element can represent either the degenerate case of one node or the more general
-    // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with
-    // one node. This helps generating a 3D space from tensorial products...
-    // A good way to handle this would be to generalize the algorithm to take into account the
-    // discretization order used in each direction, at least in the FEM sense, since a 0 degree
-    // discretization will have a unique node per element. This way 1D discretization can be
-    // viewed as a 3D problem with one 0 degree element in the y direction and one 0 degre
-    // element in the z direction.
-    // !!! Operations below are aftecting both local and global values that have two         !!!
-    // different orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G.
-    // coarseRate, endRate and offsets are in the global basis, as well as all the variables
-    // starting with a g.
-    // !!! while the variables starting with an l are in the local basis.                    !!!
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Check whether the partition includes the "end" of the mesh which means that endRate
-        // will apply. Also make sure that endRate is not 0 which means that the mesh does not
-        // require a particular treatment at the boundaries.
-        if( meshEdge[2*dim + 1] ) {
-          lCoarseNodesPerDir[dim] = (lFineNodesPerDir[dim] - endRate[dim] + offsets[dim] - 1)
-            / coarseRate[dim] + 1;
-          if(offsets[dim] == 0) {++lCoarseNodesPerDir[dim];}
-          // We might want to coarsening the direction
-          // into a single layer if there are not enough
-          // points left to form two aggregates
-          if(singleCoarsePoint_ && lFineNodesPerDir[dim] - 1 < coarseRate[dim]) {
-            lCoarseNodesPerDir[dim] =1;
-          }
-        } else {
-          lCoarseNodesPerDir[dim] = (lFineNodesPerDir[dim] + offsets[dim] - 1) / coarseRate[dim];
-          if(offsets[dim] == 0) {++lCoarseNodesPerDir[dim];}
+  // Here one element can represent either the degenerate case of one node or the more general
+  // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with
+  // one node. This helps generating a 3D space from tensorial products...
+  // A good way to handle this would be to generalize the algorithm to take into account the
+  // discretization order used in each direction, at least in the FEM sense, since a 0 degree
+  // discretization will have a unique node per element. This way 1D discretization can be
+  // viewed as a 3D problem with one 0 degree element in the y direction and one 0 degre
+  // element in the z direction.
+  // !!! Operations below are aftecting both local and global values that have two         !!!
+  // different orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G.
+  // coarseRate, endRate and offsets are in the global basis, as well as all the variables
+  // starting with a g.
+  // !!! while the variables starting with an l are in the local basis.                    !!!
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Check whether the partition includes the "end" of the mesh which means that endRate
+      // will apply. Also make sure that endRate is not 0 which means that the mesh does not
+      // require a particular treatment at the boundaries.
+      if (meshEdge[2 * dim + 1]) {
+        lCoarseNodesPerDir[dim] = (lFineNodesPerDir[dim] - endRate[dim] + offsets[dim] - 1) / coarseRate[dim] + 1;
+        if (offsets[dim] == 0) {
+          ++lCoarseNodesPerDir[dim];
         }
-
-        // The first branch of this if-statement will be used if the rank contains only one layer
-        // of nodes in direction i, that layer must also coincide with the boundary of the mesh
-        // and coarseRate[i] == endRate[i]...
-        if(interpolationOrder_ == 0) {
-          startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim];
-          int rem = startIndices[dim] % coarseRate[dim];
-          if(rem > (Teuchos::as<double>(coarseRate[dim]) / 2.0) ) {
-            ++startGhostedCoarseNode[dim];
-          }
-        } else {
-          if((startIndices[dim] == gFineNodesPerDir[dim] - 1) &&
-             (startIndices[dim] % coarseRate[dim] == 0)) {
-            startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim] - 1;
-          } else {
-            startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim];
-          }
+        // We might want to coarsening the direction
+        // into a single layer if there are not enough
+        // points left to form two aggregates
+        if (singleCoarsePoint_ && lFineNodesPerDir[dim] - 1 < coarseRate[dim]) {
+          lCoarseNodesPerDir[dim] = 1;
         }
+      } else {
+        lCoarseNodesPerDir[dim] = (lFineNodesPerDir[dim] + offsets[dim] - 1) / coarseRate[dim];
+        if (offsets[dim] == 0) {
+          ++lCoarseNodesPerDir[dim];
+        }
+      }
 
-        // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
-        // level.
-        gCoarseNodesPerDir[dim] = (gFineNodesPerDir[dim] - 1) / coarseRate[dim];
-        if((gFineNodesPerDir[dim] - 1) % coarseRate[dim] == 0) {
-          ++gCoarseNodesPerDir[dim];
+      // The first branch of this if-statement will be used if the rank contains only one layer
+      // of nodes in direction i, that layer must also coincide with the boundary of the mesh
+      // and coarseRate[i] == endRate[i]...
+      if (interpolationOrder_ == 0) {
+        startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim];
+        int rem                     = startIndices[dim] % coarseRate[dim];
+        if (rem > (Teuchos::as<double>(coarseRate[dim]) / 2.0)) {
+          ++startGhostedCoarseNode[dim];
+        }
+      } else {
+        if ((startIndices[dim] == gFineNodesPerDir[dim] - 1) &&
+            (startIndices[dim] % coarseRate[dim] == 0)) {
+          startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim] - 1;
         } else {
-          gCoarseNodesPerDir[dim] += 2;
+          startGhostedCoarseNode[dim] = startIndices[dim] / coarseRate[dim];
         }
-      } else { // Default value for dim >= numDimensions
-        // endRate[dim] = 1;
-        gCoarseNodesPerDir[dim] = 1;
-        lCoarseNodesPerDir[dim] = 1;
-      } // if (dim < numDimensions)
-
-      // This would happen if the rank does not own any nodes but in that case a subcommunicator
-      // should be used so this should really not be a concern.
-      if(lFineNodesPerDir[dim] < 1) {lCoarseNodesPerDir[dim] = 0;}
-      ghostedNodesPerDir[dim] = lCoarseNodesPerDir[dim];
-      // Check whether face *low needs ghost nodes
-      if(ghostInterface[2*dim]) {ghostedNodesPerDir[dim] += 1;}
-      // Check whether face *hi needs ghost nodes
-      if(ghostInterface[2*dim + 1]) {ghostedNodesPerDir[dim] += 1;}
-    } // Loop for dim=0:3
+      }
 
-    // With uncoupled aggregation we need to communicate to compute the global number of coarse points
-    if(!coupled_) {
-      for(int dim = 0; dim < 3; ++dim) {
-        gCoarseNodesPerDir[dim] = -1;
+      // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
+      // level.
+      gCoarseNodesPerDir[dim] = (gFineNodesPerDir[dim] - 1) / coarseRate[dim];
+      if ((gFineNodesPerDir[dim] - 1) % coarseRate[dim] == 0) {
+        ++gCoarseNodesPerDir[dim];
+      } else {
+        gCoarseNodesPerDir[dim] += 2;
       }
-    }
+    } else {  // Default value for dim >= numDimensions
+      // endRate[dim] = 1;
+      gCoarseNodesPerDir[dim] = 1;
+      lCoarseNodesPerDir[dim] = 1;
+    }  // if (dim < numDimensions)
 
-    // Compute cummulative values
-    lNumCoarseNodes10 = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1];
-    lNumCoarseNodes   = lNumCoarseNodes10*lCoarseNodesPerDir[2];
-    numGhostedNodes10 = ghostedNodesPerDir[1]*ghostedNodesPerDir[0];
-    numGhostedNodes   = numGhostedNodes10*ghostedNodesPerDir[2];
-    numGhostNodes     = numGhostedNodes - lNumCoarseNodes;
+    // This would happen if the rank does not own any nodes but in that case a subcommunicator
+    // should be used so this should really not be a concern.
+    if (lFineNodesPerDir[dim] < 1) {
+      lCoarseNodesPerDir[dim] = 0;
+    }
+    ghostedNodesPerDir[dim] = lCoarseNodesPerDir[dim];
+    // Check whether face *low needs ghost nodes
+    if (ghostInterface[2 * dim]) {
+      ghostedNodesPerDir[dim] += 1;
+    }
+    // Check whether face *hi needs ghost nodes
+    if (ghostInterface[2 * dim + 1]) {
+      ghostedNodesPerDir[dim] += 1;
+    }
+  }  // Loop for dim=0:3
 
-    *out << "lCoarseNodesPerDir: " << lCoarseNodesPerDir << std::endl;
-    *out << "gCoarseNodesPerDir: " << gCoarseNodesPerDir << std::endl;
-    *out << "ghostedNodesPerDir: " << ghostedNodesPerDir << std::endl;
-    *out << "lNumCoarseNodes=" << lNumCoarseNodes << std::endl;
-    *out << "numGhostedNodes=" << numGhostedNodes << std::endl;
+  // With uncoupled aggregation we need to communicate to compute the global number of coarse points
+  if (!coupled_) {
+    for (int dim = 0; dim < 3; ++dim) {
+      gCoarseNodesPerDir[dim] = -1;
+    }
   }
 
-} //namespace MueLu
+  // Compute cummulative values
+  lNumCoarseNodes10 = lCoarseNodesPerDir[0] * lCoarseNodesPerDir[1];
+  lNumCoarseNodes   = lNumCoarseNodes10 * lCoarseNodesPerDir[2];
+  numGhostedNodes10 = ghostedNodesPerDir[1] * ghostedNodesPerDir[0];
+  numGhostedNodes   = numGhostedNodes10 * ghostedNodesPerDir[2];
+  numGhostNodes     = numGhostedNodes - lNumCoarseNodes;
+
+  *out << "lCoarseNodesPerDir: " << lCoarseNodesPerDir << std::endl;
+  *out << "gCoarseNodesPerDir: " << gCoarseNodesPerDir << std::endl;
+  *out << "ghostedNodesPerDir: " << ghostedNodesPerDir << std::endl;
+  *out << "lNumCoarseNodes=" << lNumCoarseNodes << std::endl;
+  *out << "numGhostedNodes=" << numGhostedNodes << std::endl;
+}
+
+}  // namespace MueLu
 
 #define MUELU_INDEXMANAGER_SHORT
-#endif // MUELU_INDEXMANAGER_DEF_HPP
+#endif  // MUELU_INDEXMANAGER_DEF_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_decl.hpp
index 113368ad6f50..06e72c2eb66d 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_decl.hpp
@@ -53,7 +53,6 @@
 
 #include "Teuchos_OrdinalTraits.hpp"
 
-
 #include "MueLu_BaseClass.hpp"
 #include "MueLu_IndexManager_kokkos_fwd.hpp"
 
@@ -74,112 +73,109 @@ namespace MueLu {
     spaces and it also provides utilites for coarsening.
 */
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IndexManager_kokkos : public BaseClass {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IndexManager_kokkos : public BaseClass {
 #undef MUELU_INDEXMANAGER_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using execution_space = typename Node::execution_space;
-    using memory_space    = typename Node::memory_space;
-    using device_type     = Kokkos::Device<execution_space, memory_space>;
-    using intTupleView    = typename Kokkos::View<int[3], device_type>;
-    using LOTupleView     = typename Kokkos::View<LO[3],  device_type>;
-
-  private:
-
-    const int meshLayout = UNCOUPLED;
-    int myRank     = -1;
-    int numDimensions;                 ///< Number of spacial dimensions in the problem
-    int interpolationOrder_;           ///< Interpolation order used by grid transfer operators using these aggregates.
-    intTupleView coarseRate;           ///< coarsening rate in each direction
-    intTupleView endRate;              ///< adapted coarsening rate at the edge of the mesh in each direction.
-
-    LO lNumFineNodes;                  ///< local number of nodes.
-    LO lNumFineNodes10;                ///< local number of nodes per 0-1 slice.
-    LOTupleView lFineNodesPerDir;      ///< local number of nodes per direction.
-
-    LO numCoarseNodes;                 ///< local number of nodes remaining after coarsening.
-    LO numCoarseNodes10;               ///< local number of nodes per 0-1 slice remaining after coarsening.
-    LOTupleView coarseNodesPerDir;     ///< local number of nodes per direction remaing after coarsening.
-
-  public:
-
-    //! Default constructor, return empty object
-    IndexManager_kokkos() = default;
-
-    //! Constructs for uncoupled meshes
-    IndexManager_kokkos(const int NumDimensions,
-                        const int interpolationOrder,
-                        const int MyRank,
-                        const ArrayView<const LO> LFineNodesPerDir,
-                        const ArrayView<const int> CoarseRate);
-
-    virtual ~IndexManager_kokkos() {}
-
-    //! Common setup pattern used for all the different types of undelying mesh
-    void setupIM(const int NumDimensions,
-                 const int interpolationOrder,
-                 const ArrayView<const int> coarseRate,
-                 const ArrayView<const LO> LFineNodesPerDir);
-
-    //! Sets basic parameters used to compute indices on the mesh.
-    //! This method requires you to have set this->coarseRate.
-    void computeMeshParameters();
-
-    int getNumDimensions()      const {return numDimensions;}
-
-    int getInterpolationOrder() const {return interpolationOrder_;}
-
-    LO getNumLocalFineNodes()   const {return lNumFineNodes;}
-
-    LO getNumCoarseNodes()      const {return numCoarseNodes;}
-
-    KOKKOS_INLINE_FUNCTION
-    intTupleView getCoarseningRates()      const {return coarseRate;}
-
-    KOKKOS_INLINE_FUNCTION
-    intTupleView getCoarseningEndRates()   const {return endRate;}
-
-    KOKKOS_INLINE_FUNCTION
-    LOTupleView  getLocalFineNodesPerDir() const {return lFineNodesPerDir;}
-
-    KOKKOS_INLINE_FUNCTION
-    LOTupleView  getCoarseNodesPerDir()    const {return coarseNodesPerDir;}
-
-    Array<LO> getCoarseNodesPerDirArray() const;
-
-    KOKKOS_INLINE_FUNCTION
-    void getFineLID2FineTuple(const LO myLID, LO (&tuple)[3]) const {
-      LO tmp;
-      tuple[2] = myLID / (lFineNodesPerDir(1)*lFineNodesPerDir(0));
-      tmp      = myLID % (lFineNodesPerDir(1)*lFineNodesPerDir(0));
-      tuple[1] = tmp   / lFineNodesPerDir(0);
-      tuple[0] = tmp   % lFineNodesPerDir(0);
-    } // getFineNodeLocalTuple
-
-    KOKKOS_INLINE_FUNCTION
-    void getFineTuple2FineLID(const LO tuple[3], LO& myLID) const {
-      myLID = tuple[2]*lNumFineNodes10 + tuple[1]*lFineNodesPerDir[0] + tuple[0];
-    } // getFineNodeLID
-
-    KOKKOS_INLINE_FUNCTION
-    void getCoarseLID2CoarseTuple(const LO myLID, LO (&tuple)[3]) const {
-      LO tmp;
-      tuple[2] = myLID / numCoarseNodes10;
-      tmp      = myLID % numCoarseNodes10;
-      tuple[1] = tmp   / coarseNodesPerDir[0];
-      tuple[0] = tmp   % coarseNodesPerDir[0];
-    } // getCoarseNodeLocalTuple
-
-    KOKKOS_INLINE_FUNCTION
-    void getCoarseTuple2CoarseLID(const LO i, const LO j, const LO k, LO& myLID) const {
-      myLID = k*numCoarseNodes10 + j*coarseNodesPerDir[0] + i;
-    } // getCoarseNodeLID
-
-  };
-
-} //namespace MueLu
+ public:
+  using execution_space = typename Node::execution_space;
+  using memory_space    = typename Node::memory_space;
+  using device_type     = Kokkos::Device<execution_space, memory_space>;
+  using intTupleView    = typename Kokkos::View<int[3], device_type>;
+  using LOTupleView     = typename Kokkos::View<LO[3], device_type>;
+
+ private:
+  const int meshLayout = UNCOUPLED;
+  int myRank           = -1;
+  int numDimensions;        ///< Number of spacial dimensions in the problem
+  int interpolationOrder_;  ///< Interpolation order used by grid transfer operators using these aggregates.
+  intTupleView coarseRate;  ///< coarsening rate in each direction
+  intTupleView endRate;     ///< adapted coarsening rate at the edge of the mesh in each direction.
+
+  LO lNumFineNodes;              ///< local number of nodes.
+  LO lNumFineNodes10;            ///< local number of nodes per 0-1 slice.
+  LOTupleView lFineNodesPerDir;  ///< local number of nodes per direction.
+
+  LO numCoarseNodes;              ///< local number of nodes remaining after coarsening.
+  LO numCoarseNodes10;            ///< local number of nodes per 0-1 slice remaining after coarsening.
+  LOTupleView coarseNodesPerDir;  ///< local number of nodes per direction remaing after coarsening.
+
+ public:
+  //! Default constructor, return empty object
+  IndexManager_kokkos() = default;
+
+  //! Constructs for uncoupled meshes
+  IndexManager_kokkos(const int NumDimensions,
+                      const int interpolationOrder,
+                      const int MyRank,
+                      const ArrayView<const LO> LFineNodesPerDir,
+                      const ArrayView<const int> CoarseRate);
+
+  virtual ~IndexManager_kokkos() {}
+
+  //! Common setup pattern used for all the different types of undelying mesh
+  void setupIM(const int NumDimensions,
+               const int interpolationOrder,
+               const ArrayView<const int> coarseRate,
+               const ArrayView<const LO> LFineNodesPerDir);
+
+  //! Sets basic parameters used to compute indices on the mesh.
+  //! This method requires you to have set this->coarseRate.
+  void computeMeshParameters();
+
+  int getNumDimensions() const { return numDimensions; }
+
+  int getInterpolationOrder() const { return interpolationOrder_; }
+
+  LO getNumLocalFineNodes() const { return lNumFineNodes; }
+
+  LO getNumCoarseNodes() const { return numCoarseNodes; }
+
+  KOKKOS_INLINE_FUNCTION
+  intTupleView getCoarseningRates() const { return coarseRate; }
+
+  KOKKOS_INLINE_FUNCTION
+  intTupleView getCoarseningEndRates() const { return endRate; }
+
+  KOKKOS_INLINE_FUNCTION
+  LOTupleView getLocalFineNodesPerDir() const { return lFineNodesPerDir; }
+
+  KOKKOS_INLINE_FUNCTION
+  LOTupleView getCoarseNodesPerDir() const { return coarseNodesPerDir; }
+
+  Array<LO> getCoarseNodesPerDirArray() const;
+
+  KOKKOS_INLINE_FUNCTION
+  void getFineLID2FineTuple(const LO myLID, LO (&tuple)[3]) const {
+    LO tmp;
+    tuple[2] = myLID / (lFineNodesPerDir(1) * lFineNodesPerDir(0));
+    tmp      = myLID % (lFineNodesPerDir(1) * lFineNodesPerDir(0));
+    tuple[1] = tmp / lFineNodesPerDir(0);
+    tuple[0] = tmp % lFineNodesPerDir(0);
+  }  // getFineNodeLocalTuple
+
+  KOKKOS_INLINE_FUNCTION
+  void getFineTuple2FineLID(const LO tuple[3], LO& myLID) const {
+    myLID = tuple[2] * lNumFineNodes10 + tuple[1] * lFineNodesPerDir[0] + tuple[0];
+  }  // getFineNodeLID
+
+  KOKKOS_INLINE_FUNCTION
+  void getCoarseLID2CoarseTuple(const LO myLID, LO (&tuple)[3]) const {
+    LO tmp;
+    tuple[2] = myLID / numCoarseNodes10;
+    tmp      = myLID % numCoarseNodes10;
+    tuple[1] = tmp / coarseNodesPerDir[0];
+    tuple[0] = tmp % coarseNodesPerDir[0];
+  }  // getCoarseNodeLocalTuple
+
+  KOKKOS_INLINE_FUNCTION
+  void getCoarseTuple2CoarseLID(const LO i, const LO j, const LO k, LO& myLID) const {
+    myLID = k * numCoarseNodes10 + j * coarseNodesPerDir[0] + i;
+  }  // getCoarseNodeLID
+};
+
+}  // namespace MueLu
 
 #define MUELU_INDEXMANAGER_KOKKOS_SHORT
-#endif // MUELU_INDEXMANAGER_KOKKOS_DECL_HPP
+#endif  // MUELU_INDEXMANAGER_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_def.hpp
index f31f94421d86..502f7c3667eb 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_IndexManager_kokkos_def.hpp
@@ -63,171 +63,172 @@
 
 namespace MueLu {
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  IndexManager_kokkos(const int NumDimensions,
-                      const int interpolationOrder,
-                      const int MyRank,
-                      const ArrayView<const LO> LFineNodesPerDir,
-                      const ArrayView<const int> CoarseRate) :
-    myRank(MyRank), coarseRate("coarsening rate"), endRate("endRate"),
-    lFineNodesPerDir("lFineNodesPerDir"), coarseNodesPerDir("lFineNodesPerDir") {
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    setupIM(NumDimensions, interpolationOrder, CoarseRate, LFineNodesPerDir);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    IndexManager_kokkos(const int NumDimensions,
+                        const int interpolationOrder,
+                        const int MyRank,
+                        const ArrayView<const LO> LFineNodesPerDir,
+                        const ArrayView<const int> CoarseRate)
+  : myRank(MyRank)
+  , coarseRate("coarsening rate")
+  , endRate("endRate")
+  , lFineNodesPerDir("lFineNodesPerDir")
+  , coarseNodesPerDir("lFineNodesPerDir") {
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-    *out << "Done setting up the IndexManager" << std::endl;
+  setupIM(NumDimensions, interpolationOrder, CoarseRate, LFineNodesPerDir);
 
-    computeMeshParameters();
+  *out << "Done setting up the IndexManager" << std::endl;
 
-    *out << "Computed Mesh Parameters" << std::endl;
+  computeMeshParameters();
 
-  } // IndexManager_kokkos Constructor
+  *out << "Computed Mesh Parameters" << std::endl;
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  setupIM(const int NumDimensions, const int interpolationOrder,
-          const ArrayView<const int> CoarseRate, const ArrayView<const LO> LFineNodesPerDir) {
+}  // IndexManager_kokkos Constructor
 
-    numDimensions       = NumDimensions;
-    interpolationOrder_ = interpolationOrder;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    setupIM(const int NumDimensions, const int interpolationOrder,
+            const ArrayView<const int> CoarseRate, const ArrayView<const LO> LFineNodesPerDir) {
+  numDimensions       = NumDimensions;
+  interpolationOrder_ = interpolationOrder;
 
-    TEUCHOS_TEST_FOR_EXCEPTION((LFineNodesPerDir.size() != 3)
-                               && (LFineNodesPerDir.size() != numDimensions),
-                               Exceptions::RuntimeError,
-                               "LFineNodesPerDir has to be of size 3 or of size numDimensions!");
+  TEUCHOS_TEST_FOR_EXCEPTION((LFineNodesPerDir.size() != 3) && (LFineNodesPerDir.size() != numDimensions),
+                             Exceptions::RuntimeError,
+                             "LFineNodesPerDir has to be of size 3 or of size numDimensions!");
 
-    typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h = Kokkos::create_mirror_view(lFineNodesPerDir);
-    Kokkos::deep_copy(lFineNodesPerDir_h, lFineNodesPerDir);
-    typename Kokkos::View<LO[3], device_type>::HostMirror coarseRate_h = Kokkos::create_mirror_view(coarseRate);
-    Kokkos::deep_copy(coarseRate_h, coarseRate);
+  typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h = Kokkos::create_mirror_view(lFineNodesPerDir);
+  Kokkos::deep_copy(lFineNodesPerDir_h, lFineNodesPerDir);
+  typename Kokkos::View<LO[3], device_type>::HostMirror coarseRate_h = Kokkos::create_mirror_view(coarseRate);
+  Kokkos::deep_copy(coarseRate_h, coarseRate);
 
-    // Load coarse rate, being careful about formating
-    // Also load lFineNodesPerDir
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < getNumDimensions()) {
-        lFineNodesPerDir_h(dim) = LFineNodesPerDir[dim];
-        if(CoarseRate.size() == 1) {
-          coarseRate_h(dim) = CoarseRate[0];
-        } else if(CoarseRate.size() == getNumDimensions()) {
-          coarseRate_h(dim) = CoarseRate[dim];
-        }
-      } else {
-        lFineNodesPerDir_h(dim) = 1;
-        coarseRate_h(dim) = 1;
+  // Load coarse rate, being careful about formating
+  // Also load lFineNodesPerDir
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < getNumDimensions()) {
+      lFineNodesPerDir_h(dim) = LFineNodesPerDir[dim];
+      if (CoarseRate.size() == 1) {
+        coarseRate_h(dim) = CoarseRate[0];
+      } else if (CoarseRate.size() == getNumDimensions()) {
+        coarseRate_h(dim) = CoarseRate[dim];
       }
-    }
-
-    Kokkos::deep_copy(lFineNodesPerDir, lFineNodesPerDir_h);
-    Kokkos::deep_copy(coarseRate, coarseRate_h);
-
-  } // setupIM
-
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::computeMeshParameters() {
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+      lFineNodesPerDir_h(dim) = 1;
+      coarseRate_h(dim)       = 1;
     }
+  }
 
-    typename Kokkos::View<int[3], device_type>::HostMirror coarseRate_h = Kokkos::create_mirror_view(coarseRate);
-    typename Kokkos::View<int[3], device_type>::HostMirror endRate_h = Kokkos::create_mirror_view(endRate);
-
+  Kokkos::deep_copy(lFineNodesPerDir, lFineNodesPerDir_h);
+  Kokkos::deep_copy(coarseRate, coarseRate_h);
 
-    typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h = Kokkos::create_mirror_view(lFineNodesPerDir);
-    typename Kokkos::View<LO[3], device_type>::HostMirror coarseNodesPerDir_h = Kokkos::create_mirror_view(coarseNodesPerDir);
-    Kokkos::deep_copy(lFineNodesPerDir_h, lFineNodesPerDir);
-    Kokkos::deep_copy(coarseRate_h, coarseRate);
+}  // setupIM
 
-    lNumFineNodes10 = lFineNodesPerDir_h(1)*lFineNodesPerDir_h(0);
-    lNumFineNodes   = lFineNodesPerDir_h(2)*lNumFineNodes10;
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        endRate_h(dim) = (lFineNodesPerDir_h(dim) - 1) % coarseRate_h(dim);
-        if(endRate_h(dim) == 0) {endRate_h(dim) = coarseRate_h(dim);}
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::computeMeshParameters() {
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_INDEXMANAGER_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-      } else { // Default value for dim >= numDimensions
-        endRate_h(dim) = 1;
+  typename Kokkos::View<int[3], device_type>::HostMirror coarseRate_h = Kokkos::create_mirror_view(coarseRate);
+  typename Kokkos::View<int[3], device_type>::HostMirror endRate_h    = Kokkos::create_mirror_view(endRate);
+
+  typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h  = Kokkos::create_mirror_view(lFineNodesPerDir);
+  typename Kokkos::View<LO[3], device_type>::HostMirror coarseNodesPerDir_h = Kokkos::create_mirror_view(coarseNodesPerDir);
+  Kokkos::deep_copy(lFineNodesPerDir_h, lFineNodesPerDir);
+  Kokkos::deep_copy(coarseRate_h, coarseRate);
+
+  lNumFineNodes10 = lFineNodesPerDir_h(1) * lFineNodesPerDir_h(0);
+  lNumFineNodes   = lFineNodesPerDir_h(2) * lNumFineNodes10;
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      endRate_h(dim) = (lFineNodesPerDir_h(dim) - 1) % coarseRate_h(dim);
+      if (endRate_h(dim) == 0) {
+        endRate_h(dim) = coarseRate_h(dim);
       }
-    }
 
-    *out << "lFineNodesPerDir: {" << lFineNodesPerDir_h(0) << ", " << lFineNodesPerDir_h(1) << ", "
-         << lFineNodesPerDir_h(2) << "}" << std::endl;
-    *out << "endRate: {" << endRate_h(0) << ", " << endRate_h(1) << ", "
-         << endRate_h(2) << "}"   << std::endl;
-
-    // Here one element can represent either the degenerate case of one node or the more general
-    // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with
-    // one node. This helps generating a 3D space from tensorial products...
-    // A good way to handle this would be to generalize the algorithm to take into account the
-    // discretization order used in each direction, at least in the FEM sense, since a 0 degree
-    // discretization will have a unique node per element. This way 1D discretization can be
-    // viewed as a 3D problem with one 0 degree element in the y direction and one 0 degre
-    // element in the z direction.
-    // !!! Operations below are aftecting both local and global values that have two         !!!
-    // different orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G.
-    // coarseRate, endRate and offsets are in the global basis, as well as all the variables
-    // starting with a g.
-    // !!! while the variables starting with an l are in the local basis.                    !!!
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Check whether the partition includes the "end" of the mesh which means that endRate
-        // will apply. Also make sure that endRate is not 0 which means that the mesh does not
-        // require a particular treatment at the boundaries.
-        coarseNodesPerDir_h(dim) = (lFineNodesPerDir_h(dim) - endRate_h(dim) - 1)
-          / coarseRate_h(dim) + 2;
-
-      } else { // Default value for dim >= numDimensions
-        // endRate[dim] = 1;
-        coarseNodesPerDir_h(dim) = 1;
-      } // if (dim < numDimensions)
-
-      // This would happen if the rank does not own any nodes but in that case a subcommunicator
-      // should be used so this should really not be a concern.
-      if(lFineNodesPerDir_h(dim) < 1) {coarseNodesPerDir_h(dim) = 0;}
-    } // Loop for dim=0:3
-
-    // Compute cummulative values
-    numCoarseNodes10 = coarseNodesPerDir_h(0)*coarseNodesPerDir_h(1);
-    numCoarseNodes   = numCoarseNodes10*coarseNodesPerDir_h(2);
-
-    *out << "coarseNodesPerDir: {" << coarseNodesPerDir_h(0) << ", "
-         << coarseNodesPerDir_h(1) << ", " << coarseNodesPerDir_h(2) << "}" << std::endl;
-    *out << "numCoarseNodes=" << numCoarseNodes << std::endl;
-
-    // Copy Host data to Device.
-    Kokkos::deep_copy(coarseRate, coarseRate_h);
-    Kokkos::deep_copy(endRate, endRate_h);
-    Kokkos::deep_copy(lFineNodesPerDir, lFineNodesPerDir_h);
-    Kokkos::deep_copy(coarseNodesPerDir, coarseNodesPerDir_h);
+    } else {  // Default value for dim >= numDimensions
+      endRate_h(dim) = 1;
+    }
   }
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  Array<LocalOrdinal> IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodesPerDirArray() const {
-    typename LOTupleView::HostMirror coarseNodesPerDir_h = Kokkos::create_mirror_view(coarseNodesPerDir);
-    Kokkos::deep_copy(coarseNodesPerDir_h, coarseNodesPerDir);
-    Array<LO> coarseNodesPerDirArray(3);
-
-    for(int dim = 0; dim < 3; ++dim) {
-      coarseNodesPerDirArray[dim] = coarseNodesPerDir_h(dim);
+  *out << "lFineNodesPerDir: {" << lFineNodesPerDir_h(0) << ", " << lFineNodesPerDir_h(1) << ", "
+       << lFineNodesPerDir_h(2) << "}" << std::endl;
+  *out << "endRate: {" << endRate_h(0) << ", " << endRate_h(1) << ", "
+       << endRate_h(2) << "}" << std::endl;
+
+  // Here one element can represent either the degenerate case of one node or the more general
+  // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with
+  // one node. This helps generating a 3D space from tensorial products...
+  // A good way to handle this would be to generalize the algorithm to take into account the
+  // discretization order used in each direction, at least in the FEM sense, since a 0 degree
+  // discretization will have a unique node per element. This way 1D discretization can be
+  // viewed as a 3D problem with one 0 degree element in the y direction and one 0 degre
+  // element in the z direction.
+  // !!! Operations below are aftecting both local and global values that have two         !!!
+  // different orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G.
+  // coarseRate, endRate and offsets are in the global basis, as well as all the variables
+  // starting with a g.
+  // !!! while the variables starting with an l are in the local basis.                    !!!
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Check whether the partition includes the "end" of the mesh which means that endRate
+      // will apply. Also make sure that endRate is not 0 which means that the mesh does not
+      // require a particular treatment at the boundaries.
+      coarseNodesPerDir_h(dim) = (lFineNodesPerDir_h(dim) - endRate_h(dim) - 1) / coarseRate_h(dim) + 2;
+
+    } else {  // Default value for dim >= numDimensions
+      // endRate[dim] = 1;
+      coarseNodesPerDir_h(dim) = 1;
+    }  // if (dim < numDimensions)
+
+    // This would happen if the rank does not own any nodes but in that case a subcommunicator
+    // should be used so this should really not be a concern.
+    if (lFineNodesPerDir_h(dim) < 1) {
+      coarseNodesPerDir_h(dim) = 0;
     }
+  }  // Loop for dim=0:3
+
+  // Compute cummulative values
+  numCoarseNodes10 = coarseNodesPerDir_h(0) * coarseNodesPerDir_h(1);
+  numCoarseNodes   = numCoarseNodes10 * coarseNodesPerDir_h(2);
+
+  *out << "coarseNodesPerDir: {" << coarseNodesPerDir_h(0) << ", "
+       << coarseNodesPerDir_h(1) << ", " << coarseNodesPerDir_h(2) << "}" << std::endl;
+  *out << "numCoarseNodes=" << numCoarseNodes << std::endl;
+
+  // Copy Host data to Device.
+  Kokkos::deep_copy(coarseRate, coarseRate_h);
+  Kokkos::deep_copy(endRate, endRate_h);
+  Kokkos::deep_copy(lFineNodesPerDir, lFineNodesPerDir_h);
+  Kokkos::deep_copy(coarseNodesPerDir, coarseNodesPerDir_h);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Array<LocalOrdinal> IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodesPerDirArray() const {
+  typename LOTupleView::HostMirror coarseNodesPerDir_h = Kokkos::create_mirror_view(coarseNodesPerDir);
+  Kokkos::deep_copy(coarseNodesPerDir_h, coarseNodesPerDir);
+  Array<LO> coarseNodesPerDirArray(3);
+
+  for (int dim = 0; dim < 3; ++dim) {
+    coarseNodesPerDirArray[dim] = coarseNodesPerDir_h(dim);
+  }
 
-    return coarseNodesPerDirArray;
-  } // getCoarseNodesData
+  return coarseNodesPerDirArray;
+}  // getCoarseNodesData
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INDEXMANAGER_KOKKOS_SHORT
-#endif // MUELU_INDEXMANAGER_DEF_KOKKOS_HPP
+#endif  // MUELU_INDEXMANAGER_DEF_KOKKOS_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_decl.hpp
index c43758ee8352..dadd50c145c7 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_decl.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_STRUCTUREDAGGREGATIONFACTORY_DECL_HPP
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_DECL_HPP
 
-
 // #include <Xpetra_Map_fwd.hpp>
 // #include <Xpetra_CrsGraph_fwd.hpp>
 // #include <Xpetra_CrsGraphFactory.hpp>
@@ -102,60 +101,59 @@ namespace MueLu {
     | Aggregates   | StructuredAggregationFactory   | Container class with aggregation information. See also Aggregates.
 */
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class StructuredAggregationFactory : public SingleLevelFactoryBase {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class StructuredAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_STRUCTUREDAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    StructuredAggregationFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~StructuredAggregationFactory() { }
+  //! Constructor.
+  StructuredAggregationFactory();
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~StructuredAggregationFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Set/get methods.
-    //@{
-    // set information about 1-node aggregates (map name and generating factory)
-    void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
-      SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name))); // revalidate
-      SetFactory("OnePt aggregate map factory",mapFact);
-    }
+  //@}
 
-    //@}
+  //! @name Set/get methods.
+  //@{
+  // set information about 1-node aggregates (map name and generating factory)
+  void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
+    SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name)));  // revalidate
+    SetFactory("OnePt aggregate map factory", mapFact);
+  }
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level& currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*! @brief Build aggregates. */
-    void Build(Level& currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  /*! @brief Build aggregates. */
+  void Build(Level& currentLevel) const;
 
-  private:
+  //@}
 
-    //! boolean flag: definition phase
-    //! if true, the aggregation algorithms still can be set and changed.
-    //! if false, no change in aggregation algorithms is possible any more
-    mutable bool bDefinitionPhase_;
+ private:
+  //! boolean flag: definition phase
+  //! if true, the aggregation algorithms still can be set and changed.
+  //! if false, no change in aggregation algorithms is possible any more
+  mutable bool bDefinitionPhase_;
 
-  }; // class StructuredAggregationFactory
+};  // class StructuredAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_STRUCTUREDAGGREGATIONFACTORY_DECL_HPP */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_def.hpp
index 94685aefcef6..23e537a770ae 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_def.hpp
@@ -63,173 +63,172 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  StructuredAggregationFactory() : bDefinitionPhase_(true)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    StructuredAggregationFactory()
+  : bDefinitionPhase_(true) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
-    SET_VALID_ENTRY("aggregation: allow user-specified singletons");
-    SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
-    SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
-
-    // general variables needed in StructuredAggregationFactory
-    SET_VALID_ENTRY("aggregation: mesh layout");
-    SET_VALID_ENTRY("aggregation: mode");
-    SET_VALID_ENTRY("aggregation: output type");
-    SET_VALID_ENTRY("aggregation: coarsening rate");
-    SET_VALID_ENTRY("aggregation: coarsening order");
-#undef  SET_VALID_ENTRY
-    validParamList->set<RCP<const FactoryBase> >("Graph",                   Teuchos::null,
-                                                 "Graph of the matrix after amalgamation but without dropping.");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",           Teuchos::null,
-                                                 "Number of spatial dimension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("gNodesPerDim",            Teuchos::null,
-                                                 "Global number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",            Teuchos::null,
-                                                 "Local number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("DofsPerNode",             Teuchos::null,
-                                                 "Generating factory for variable \'DofsPerNode\', usually the same as the \'Graph\' factory");
-    validParamList->set<const bool>("aggregation: single coarse point", false,
-                                                 "Allows the aggreagtion process to reduce spacial dimensions to a single layer");
-
-    return validParamList;
-  } // GetValidParameterList()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "Graph");
-    Input(currentLevel, "DofsPerNode");
-
-    ParameterList pL = GetParameterList();
-    std::string coupling = pL.get<std::string>("aggregation: mode");
-    const bool coupled = (coupling == "coupled" ? true : false);
-    if(coupled) {
-      // Request the global number of nodes per dimensions
-      if(currentLevel.GetLevelID() == 0) {
-        if(currentLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
-          currentLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
-                                     Exceptions::RuntimeError,
-                                     "gNodesPerDim was not provided by the user on level0!");
-        }
+  SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
+  SET_VALID_ENTRY("aggregation: allow user-specified singletons");
+  SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
+  SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
+
+  // general variables needed in StructuredAggregationFactory
+  SET_VALID_ENTRY("aggregation: mesh layout");
+  SET_VALID_ENTRY("aggregation: mode");
+  SET_VALID_ENTRY("aggregation: output type");
+  SET_VALID_ENTRY("aggregation: coarsening rate");
+  SET_VALID_ENTRY("aggregation: coarsening order");
+#undef SET_VALID_ENTRY
+  validParamList->set<RCP<const FactoryBase> >("Graph", Teuchos::null,
+                                               "Graph of the matrix after amalgamation but without dropping.");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spatial dimension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("gNodesPerDim", Teuchos::null,
+                                               "Global number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Local number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", Teuchos::null,
+                                               "Generating factory for variable \'DofsPerNode\', usually the same as the \'Graph\' factory");
+  validParamList->set<const bool>("aggregation: single coarse point", false,
+                                  "Allows the aggreagtion process to reduce spacial dimensions to a single layer");
+
+  return validParamList;
+}  // GetValidParameterList()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Graph");
+  Input(currentLevel, "DofsPerNode");
+
+  ParameterList pL     = GetParameterList();
+  std::string coupling = pL.get<std::string>("aggregation: mode");
+  const bool coupled   = (coupling == "coupled" ? true : false);
+  if (coupled) {
+    // Request the global number of nodes per dimensions
+    if (currentLevel.GetLevelID() == 0) {
+      if (currentLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
+        currentLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
       } else {
-        Input(currentLevel, "gNodesPerDim");
-      }
-    }
-
-    // Request the local number of nodes per dimensions
-    if(currentLevel.GetLevelID() == 0) {
-      if(currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
-        currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "numDimensions was not provided by the user on level0!");
-      }
-      if(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
                                    Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
+                                   "gNodesPerDim was not provided by the user on level0!");
       }
     } else {
-      Input(currentLevel, "numDimensions");
-      Input(currentLevel, "lNodesPerDim");
+      Input(currentLevel, "gNodesPerDim");
     }
-  } // DeclareInput()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
+  }
 
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
+  // Request the local number of nodes per dimensions
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
+      currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "numDimensions was not provided by the user on level0!");
     }
-
-    *out << "Entering structured aggregation" << std::endl;
-
-    ParameterList pL = GetParameterList();
-    bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
-
-    // General problem informations are gathered from data stored in the problem matix.
-    RCP<const GraphBase> graph  = Get< RCP<GraphBase> >(currentLevel, "Graph");
-    RCP<const Map> fineMap     = graph->GetDomainMap();
-    const int myRank           = fineMap->getComm()->getRank();
-    const int numRanks         = fineMap->getComm()->getSize();
-    const GO  minGlobalIndex   = fineMap->getMinGlobalIndex();
-    const LO  dofsPerNode      = Get<LO>(currentLevel, "DofsPerNode");
-
-    // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
-    // obtain a nodeMap.
-    const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
-    std::string meshLayout       = pL.get<std::string>("aggregation: mesh layout");
-    std::string coupling         = pL.get<std::string>("aggregation: mode");
-    const bool coupled           = (coupling == "coupled" ? true : false);
-    std::string outputType       = pL.get<std::string>("aggregation: output type");
-    const bool outputAggregates  = (outputType == "Aggregates" ? true : false);
-    const bool singleCoarsePoint = pL.get<bool>("aggregation: single coarse point");
-    int numDimensions;
-    Array<GO> gFineNodesPerDir(3);
-    Array<LO> lFineNodesPerDir(3);
-    if(currentLevel.GetLevelID() == 0) {
-      // On level 0, data is provided by applications and has no associated factory.
-      numDimensions = currentLevel.Get<int>("numDimensions", NoFactory::get());
-      lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
-      if(coupled) {
-        gFineNodesPerDir = currentLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
-      }
+    if (currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
     } else {
-      // On level > 0, data is provided directly by generating factories.
-      numDimensions = Get<int>(currentLevel, "numDimensions");
-      lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
-      if(coupled) {
-        gFineNodesPerDir = Get<Array<GO> >(currentLevel, "gNodesPerDim");
-      }
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
     }
-
-
-    // First make sure that input parameters are set logically based on dimension
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim >= numDimensions) {
-        gFineNodesPerDir[dim] = 1;
-        lFineNodesPerDir[dim] = 1;
-      }
+  } else {
+    Input(currentLevel, "numDimensions");
+    Input(currentLevel, "lNodesPerDim");
+  }
+}  // DeclareInput()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  *out << "Entering structured aggregation" << std::endl;
+
+  ParameterList pL  = GetParameterList();
+  bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
+
+  // General problem informations are gathered from data stored in the problem matix.
+  RCP<const GraphBase> graph = Get<RCP<GraphBase> >(currentLevel, "Graph");
+  RCP<const Map> fineMap     = graph->GetDomainMap();
+  const int myRank           = fineMap->getComm()->getRank();
+  const int numRanks         = fineMap->getComm()->getSize();
+  const GO minGlobalIndex    = fineMap->getMinGlobalIndex();
+  const LO dofsPerNode       = Get<LO>(currentLevel, "DofsPerNode");
+
+  // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
+  // obtain a nodeMap.
+  const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
+  std::string meshLayout       = pL.get<std::string>("aggregation: mesh layout");
+  std::string coupling         = pL.get<std::string>("aggregation: mode");
+  const bool coupled           = (coupling == "coupled" ? true : false);
+  std::string outputType       = pL.get<std::string>("aggregation: output type");
+  const bool outputAggregates  = (outputType == "Aggregates" ? true : false);
+  const bool singleCoarsePoint = pL.get<bool>("aggregation: single coarse point");
+  int numDimensions;
+  Array<GO> gFineNodesPerDir(3);
+  Array<LO> lFineNodesPerDir(3);
+  if (currentLevel.GetLevelID() == 0) {
+    // On level 0, data is provided by applications and has no associated factory.
+    numDimensions    = currentLevel.Get<int>("numDimensions", NoFactory::get());
+    lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
+    if (coupled) {
+      gFineNodesPerDir = currentLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
     }
-
-    // Get the coarsening rate
-    std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
-    Teuchos::Array<LO> coarseRate;
-    try {
-      coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
-    } catch(const Teuchos::InvalidArrayStringRepresentation& e) {
-      GetOStream(Errors,-1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
-                            << std::endl;
-      throw e;
+  } else {
+    // On level > 0, data is provided directly by generating factories.
+    numDimensions    = Get<int>(currentLevel, "numDimensions");
+    lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
+    if (coupled) {
+      gFineNodesPerDir = Get<Array<GO> >(currentLevel, "gNodesPerDim");
     }
-    TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
-                               Exceptions::RuntimeError,
-                               "\"aggregation: coarsening rate\" must have at least as many"
-                               " components as the number of spatial dimensions in the problem.");
+  }
 
-    // Now that we have extracted info from the level, create the IndexManager
-    RCP<IndexManager > geoData;
-    if(!coupled) {
-      geoData = rcp(new MueLu::UncoupledIndexManager<LO,GO,NO>(fineMap->getComm(),
+  // First make sure that input parameters are set logically based on dimension
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim >= numDimensions) {
+      gFineNodesPerDir[dim] = 1;
+      lFineNodesPerDir[dim] = 1;
+    }
+  }
+
+  // Get the coarsening rate
+  std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
+  Teuchos::Array<LO> coarseRate;
+  try {
+    coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
+  } catch (const Teuchos::InvalidArrayStringRepresentation& e) {
+    GetOStream(Errors, -1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
+                           << std::endl;
+    throw e;
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
+                             Exceptions::RuntimeError,
+                             "\"aggregation: coarsening rate\" must have at least as many"
+                             " components as the number of spatial dimensions in the problem.");
+
+  // Now that we have extracted info from the level, create the IndexManager
+  RCP<IndexManager> geoData;
+  if (!coupled) {
+    geoData = rcp(new MueLu::UncoupledIndexManager<LO, GO, NO>(fineMap->getComm(),
                                                                coupled,
                                                                numDimensions,
                                                                interpolationOrder,
@@ -239,23 +238,23 @@ namespace MueLu {
                                                                lFineNodesPerDir,
                                                                coarseRate,
                                                                singleCoarsePoint));
-    } else if(meshLayout == "Local Lexicographic") {
-      Array<GO> meshData;
-      if(currentLevel.GetLevelID() == 0) {
-        // On level 0, data is provided by applications and has no associated factory.
-        meshData = currentLevel.Get<Array<GO> >("aggregation: mesh data", NoFactory::get());
-        TEUCHOS_TEST_FOR_EXCEPTION(meshData.empty() == true, Exceptions::RuntimeError,
-                                   "The meshData array is empty, somehow the input for structured"
-                                   " aggregation are not captured correctly.");
-      } else {
-        // On level > 0, data is provided directly by generating factories.
-        meshData = Get<Array<GO> >(currentLevel, "aggregation: mesh data");
-      }
-      // Note, LBV Feb 5th 2018:
-      // I think that it might make sense to pass ghostInterface rather than interpolationOrder.
-      // For that I need to make sure that ghostInterface can be computed with minimal mesh
-      // knowledge outside of the IndexManager...
-      geoData = rcp(new MueLu::LocalLexicographicIndexManager<LO,GO,NO>(fineMap->getComm(),
+  } else if (meshLayout == "Local Lexicographic") {
+    Array<GO> meshData;
+    if (currentLevel.GetLevelID() == 0) {
+      // On level 0, data is provided by applications and has no associated factory.
+      meshData = currentLevel.Get<Array<GO> >("aggregation: mesh data", NoFactory::get());
+      TEUCHOS_TEST_FOR_EXCEPTION(meshData.empty() == true, Exceptions::RuntimeError,
+                                 "The meshData array is empty, somehow the input for structured"
+                                 " aggregation are not captured correctly.");
+    } else {
+      // On level > 0, data is provided directly by generating factories.
+      meshData = Get<Array<GO> >(currentLevel, "aggregation: mesh data");
+    }
+    // Note, LBV Feb 5th 2018:
+    // I think that it might make sense to pass ghostInterface rather than interpolationOrder.
+    // For that I need to make sure that ghostInterface can be computed with minimal mesh
+    // knowledge outside of the IndexManager...
+    geoData = rcp(new MueLu::LocalLexicographicIndexManager<LO, GO, NO>(fineMap->getComm(),
                                                                         coupled,
                                                                         numDimensions,
                                                                         interpolationOrder,
@@ -265,12 +264,12 @@ namespace MueLu {
                                                                         lFineNodesPerDir,
                                                                         coarseRate,
                                                                         meshData));
-    } else  if(meshLayout == "Global Lexicographic") {
-      // Note, LBV Feb 5th 2018:
-      // I think that it might make sense to pass ghostInterface rather than interpolationOrder.
-      // For that I need to make sure that ghostInterface can be computed with minimal mesh
-      // knowledge outside of the IndexManager...
-      geoData = rcp(new MueLu::GlobalLexicographicIndexManager<LO,GO,NO>(fineMap->getComm(),
+  } else if (meshLayout == "Global Lexicographic") {
+    // Note, LBV Feb 5th 2018:
+    // I think that it might make sense to pass ghostInterface rather than interpolationOrder.
+    // For that I need to make sure that ghostInterface can be computed with minimal mesh
+    // knowledge outside of the IndexManager...
+    geoData = rcp(new MueLu::GlobalLexicographicIndexManager<LO, GO, NO>(fineMap->getComm(),
                                                                          coupled,
                                                                          numDimensions,
                                                                          interpolationOrder,
@@ -278,75 +277,71 @@ namespace MueLu {
                                                                          lFineNodesPerDir,
                                                                          coarseRate,
                                                                          minGlobalIndex));
-    }
-
-
-    *out << "The index manager has now been built" << std::endl;
-    *out << "graph num nodes: " << fineMap->getLocalNumElements()
-         << ", structured aggregation num nodes: " << geoData->getNumLocalFineNodes() << std::endl;
-    TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements()
-                               != static_cast<size_t>(geoData->getNumLocalFineNodes()),
+  }
+
+  *out << "The index manager has now been built" << std::endl;
+  *out << "graph num nodes: " << fineMap->getLocalNumElements()
+       << ", structured aggregation num nodes: " << geoData->getNumLocalFineNodes() << std::endl;
+  TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements() != static_cast<size_t>(geoData->getNumLocalFineNodes()),
+                             Exceptions::RuntimeError,
+                             "The local number of elements in the graph's map is not equal to "
+                             "the number of nodes given by: lNodesPerDim!");
+  if (coupled) {
+    TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getGlobalNumElements() != static_cast<size_t>(geoData->getNumGlobalFineNodes()),
                                Exceptions::RuntimeError,
-                               "The local number of elements in the graph's map is not equal to "
-                               "the number of nodes given by: lNodesPerDim!");
-    if(coupled) {
-      TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getGlobalNumElements()
-                                 != static_cast<size_t>(geoData->getNumGlobalFineNodes()),
-                                 Exceptions::RuntimeError,
-                                 "The global number of elements in the graph's map is not equal to "
-                                 "the number of nodes given by: gNodesPerDim!");
-    }
-
-    *out << "Compute coarse mesh data" << std::endl;
-    std::vector<std::vector<GO> > coarseMeshData = geoData->getCoarseMeshData();
-
-    // Now we are ready for the big loop over the fine node that will assign each
-    // node on the fine grid to an aggregate and a processor.
-    RCP<const FactoryBase> graphFact = GetFactory("Graph");
-    RCP<const Map> coarseCoordinatesFineMap, coarseCoordinatesMap;
-    RCP<MueLu::AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node> >
+                               "The global number of elements in the graph's map is not equal to "
+                               "the number of nodes given by: gNodesPerDim!");
+  }
+
+  *out << "Compute coarse mesh data" << std::endl;
+  std::vector<std::vector<GO> > coarseMeshData = geoData->getCoarseMeshData();
+
+  // Now we are ready for the big loop over the fine node that will assign each
+  // node on the fine grid to an aggregate and a processor.
+  RCP<const FactoryBase> graphFact = GetFactory("Graph");
+  RCP<const Map> coarseCoordinatesFineMap, coarseCoordinatesMap;
+  RCP<MueLu::AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node> >
       myStructuredAlgorithm = rcp(new AggregationStructuredAlgorithm(graphFact));
 
-    if(interpolationOrder == 0 && outputAggregates){
-      // Create aggregates for prolongation
-      *out << "Compute Aggregates" << std::endl;
-      RCP<Aggregates> aggregates = rcp(new Aggregates(graph->GetDomainMap()));
-      aggregates->setObjectLabel("ST");
-      aggregates->SetIndexManager(geoData);
-      aggregates->AggregatesCrossProcessors(coupled);
-      aggregates->SetNumAggregates(geoData->getNumLocalCoarseNodes());
-      std::vector<unsigned> aggStat(geoData->getNumLocalFineNodes(), READY);
-      LO numNonAggregatedNodes = geoData->getNumLocalFineNodes();
-
-      myStructuredAlgorithm->BuildAggregates(pL, *graph, *aggregates, aggStat,
-                                             numNonAggregatedNodes);
-
-      TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError,
-                                 "MueLu::StructuredAggregationFactory::Build: Leftover nodes found! Error!");
-      aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
-      GetOStream(Statistics1) << aggregates->description() << std::endl;
-      Set(currentLevel, "Aggregates", aggregates);
-
-    } else {
-      // Create the graph of the prolongator
-      *out << "Compute CrsGraph" << std::endl;
-      RCP<CrsGraph> myGraph;
-      myStructuredAlgorithm->BuildGraph(*graph, geoData, dofsPerNode, myGraph,
-                                        coarseCoordinatesFineMap, coarseCoordinatesMap);
-      Set(currentLevel, "prolongatorGraph", myGraph);
-    }
-
-    if(coupled) {
-      Set(currentLevel, "gCoarseNodesPerDim", geoData->getGlobalCoarseNodesPerDir());
-    }
-    Set(currentLevel, "lCoarseNodesPerDim", geoData->getLocalCoarseNodesPerDir());
-    Set(currentLevel, "coarseCoordinatesFineMap", coarseCoordinatesFineMap);
-    Set(currentLevel, "coarseCoordinatesMap", coarseCoordinatesMap);
-    Set(currentLevel, "structuredInterpolationOrder", interpolationOrder);
-    Set(currentLevel, "numDimensions", numDimensions);
-
-  } // Build()
-} //namespace MueLu
-
+  if (interpolationOrder == 0 && outputAggregates) {
+    // Create aggregates for prolongation
+    *out << "Compute Aggregates" << std::endl;
+    RCP<Aggregates> aggregates = rcp(new Aggregates(graph->GetDomainMap()));
+    aggregates->setObjectLabel("ST");
+    aggregates->SetIndexManager(geoData);
+    aggregates->AggregatesCrossProcessors(coupled);
+    aggregates->SetNumAggregates(geoData->getNumLocalCoarseNodes());
+    std::vector<unsigned> aggStat(geoData->getNumLocalFineNodes(), READY);
+    LO numNonAggregatedNodes = geoData->getNumLocalFineNodes();
+
+    myStructuredAlgorithm->BuildAggregates(pL, *graph, *aggregates, aggStat,
+                                           numNonAggregatedNodes);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError,
+                               "MueLu::StructuredAggregationFactory::Build: Leftover nodes found! Error!");
+    aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
+    GetOStream(Statistics1) << aggregates->description() << std::endl;
+    Set(currentLevel, "Aggregates", aggregates);
+
+  } else {
+    // Create the graph of the prolongator
+    *out << "Compute CrsGraph" << std::endl;
+    RCP<CrsGraph> myGraph;
+    myStructuredAlgorithm->BuildGraph(*graph, geoData, dofsPerNode, myGraph,
+                                      coarseCoordinatesFineMap, coarseCoordinatesMap);
+    Set(currentLevel, "prolongatorGraph", myGraph);
+  }
+
+  if (coupled) {
+    Set(currentLevel, "gCoarseNodesPerDim", geoData->getGlobalCoarseNodesPerDir());
+  }
+  Set(currentLevel, "lCoarseNodesPerDim", geoData->getLocalCoarseNodesPerDir());
+  Set(currentLevel, "coarseCoordinatesFineMap", coarseCoordinatesFineMap);
+  Set(currentLevel, "coarseCoordinatesMap", coarseCoordinatesMap);
+  Set(currentLevel, "structuredInterpolationOrder", interpolationOrder);
+  Set(currentLevel, "numDimensions", numDimensions);
+
+}  // Build()
+}  // namespace MueLu
 
 #endif /* MUELU_STRUCTUREDAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_decl.hpp
index 88724397a340..b966b0043c9a 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_decl.hpp
@@ -103,62 +103,61 @@ namespace MueLu {
     | CrsGraph     | StructuredAggregationFactory_kokkos   | CrsGraph of the prolongator
 */
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class StructuredAggregationFactory_kokkos : public SingleLevelFactoryBase {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class StructuredAggregationFactory_kokkos : public SingleLevelFactoryBase {
 #undef MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    StructuredAggregationFactory_kokkos();
+  //! Constructor.
+  StructuredAggregationFactory_kokkos();
 
-    //! Destructor.
-    virtual ~StructuredAggregationFactory_kokkos() { }
+  //! Destructor.
+  virtual ~StructuredAggregationFactory_kokkos() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Set/get methods.
-    //@{
-    // set information about 1-node aggregates (map name and generating factory)
-    void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
-      SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name))); // revalidate
-      SetFactory("OnePt aggregate map factory",mapFact);
-    }
+  //! @name Set/get methods.
+  //@{
+  // set information about 1-node aggregates (map name and generating factory)
+  void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
+    SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name)));  // revalidate
+    SetFactory("OnePt aggregate map factory", mapFact);
+  }
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*! @brief Build aggregates. */
-    void Build(Level& currentLevel) const;
+  /*! @brief Build aggregates. */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
+ private:
+  //! boolean flag: definition phase
+  //! if true, the aggregation algorithms still can be set and changed.
+  //! if false, no change in aggregation algorithms is possible any more
+  mutable bool bDefinitionPhase_;
 
-    //! boolean flag: definition phase
-    //! if true, the aggregation algorithms still can be set and changed.
-    //! if false, no change in aggregation algorithms is possible any more
-    mutable bool bDefinitionPhase_;
+};  // class StructuredAggregationFactory
 
-  }; // class StructuredAggregationFactory
-
-}
+}  // namespace MueLu
 
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_SHORT
-#endif // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_def.hpp
index 1067efc3e08d..5d1911c775e8 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/MueLu_StructuredAggregationFactory_kokkos_def.hpp
@@ -65,193 +65,192 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  StructuredAggregationFactory_kokkos() : bDefinitionPhase_(true) { }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    StructuredAggregationFactory_kokkos()
+  : bDefinitionPhase_(true) {}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
-    SET_VALID_ENTRY("aggregation: allow user-specified singletons");
-    SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
-    SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
-#undef  SET_VALID_ENTRY
-
-    // general variables needed in StructuredAggregationFactory
-    validParamList->set<std::string>            ("aggregation: output type", "Aggregates",
-                                                 "Type of object holding the aggregation data: Aggregtes or CrsGraph");
-    validParamList->set<std::string>            ("aggregation: coarsening rate", "{3}",
-                                                 "Coarsening rate per spatial dimensions");
-    validParamList->set<int>                    ("aggregation: coarsening order", 0,
-                                                  "The interpolation order used to construct grid transfer operators based off these aggregates.");
-    validParamList->set<RCP<const FactoryBase> >("Graph",                   Teuchos::null,
-                                                 "Graph of the matrix after amalgamation but without dropping.");
-    validParamList->set<RCP<const FactoryBase> >("DofsPerNode",             Teuchos::null,
-                                                 "Number of degrees of freedom per mesh node, provided by the coalsce drop factory.");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",           Teuchos::null,
-                                                 "Number of spatial dimension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",            Teuchos::null,
-                                                 "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-
-    return validParamList;
-  } // GetValidParameterList()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "Graph");
-    Input(currentLevel, "DofsPerNode");
-
-    // Request the local number of nodes per dimensions
-    if(currentLevel.GetLevelID() == 0) {
-      if(currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
-        currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "numDimensions was not provided by the user on level0!");
-      }
-      if(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
-      }
+  SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
+  SET_VALID_ENTRY("aggregation: allow user-specified singletons");
+  SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
+  SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
+#undef SET_VALID_ENTRY
+
+  // general variables needed in StructuredAggregationFactory
+  validParamList->set<std::string>("aggregation: output type", "Aggregates",
+                                   "Type of object holding the aggregation data: Aggregtes or CrsGraph");
+  validParamList->set<std::string>("aggregation: coarsening rate", "{3}",
+                                   "Coarsening rate per spatial dimensions");
+  validParamList->set<int>("aggregation: coarsening order", 0,
+                           "The interpolation order used to construct grid transfer operators based off these aggregates.");
+  validParamList->set<RCP<const FactoryBase> >("Graph", Teuchos::null,
+                                               "Graph of the matrix after amalgamation but without dropping.");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", Teuchos::null,
+                                               "Number of degrees of freedom per mesh node, provided by the coalsce drop factory.");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spatial dimension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+
+  return validParamList;
+}  // GetValidParameterList()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Graph");
+  Input(currentLevel, "DofsPerNode");
+
+  // Request the local number of nodes per dimensions
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("numDimensions", NoFactory::get())) {
+      currentLevel.DeclareInput("numDimensions", NoFactory::get(), this);
     } else {
-      Input(currentLevel, "lNodesPerDim");
-      Input(currentLevel, "numDimensions");
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("numDimensions", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "numDimensions was not provided by the user on level0!");
     }
-  } // DeclareInput()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_STRUCTUREDAGGREGATION_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    using device_type     = typename LWGraph_kokkos::local_graph_type::device_type;
-    using execution_space = typename LWGraph_kokkos::local_graph_type::device_type::execution_space;
-    using memory_space    = typename LWGraph_kokkos::local_graph_type::device_type::memory_space;
-
-    *out << "Entering structured aggregation" << std::endl;
-
-    ParameterList pL = GetParameterList();
-    bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
-
-    // General problem informations are gathered from data stored in the problem matix.
-    RCP<const LWGraph_kokkos> graph = Get<RCP<LWGraph_kokkos> >(currentLevel, "Graph");
-    RCP<const Map> fineMap          = graph->GetDomainMap();
-    const int myRank                = fineMap->getComm()->getRank();
-    const LO  dofsPerNode           = Get<LO>(currentLevel, "DofsPerNode");
-
-    // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
-    // obtain a nodeMap.
-    const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
-    std::string outputType = pL.get<std::string>("aggregation: output type");
-    const bool outputAggregates = (outputType == "Aggregates" ? true : false);
-    Array<LO> lFineNodesPerDir(3);
-    int numDimensions;
-    if(currentLevel.GetLevelID() == 0) {
-      // On level 0, data is provided by applications and has no associated factory.
-      lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
-      numDimensions    = currentLevel.Get<int>("numDimensions", NoFactory::get());
+    if (currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
     } else {
-      // On level > 0, data is provided directly by generating factories.
-      lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
-      numDimensions    = Get<int>(currentLevel, "numDimensions");
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
     }
-
-
-    // First make sure that input parameters are set logically based on dimension
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim >= numDimensions) {
-        lFineNodesPerDir[dim] = 1;
-      }
+  } else {
+    Input(currentLevel, "lNodesPerDim");
+    Input(currentLevel, "numDimensions");
+  }
+}  // DeclareInput()
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_STRUCTUREDAGGREGATION_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  using device_type     = typename LWGraph_kokkos::local_graph_type::device_type;
+  using execution_space = typename LWGraph_kokkos::local_graph_type::device_type::execution_space;
+  using memory_space    = typename LWGraph_kokkos::local_graph_type::device_type::memory_space;
+
+  *out << "Entering structured aggregation" << std::endl;
+
+  ParameterList pL  = GetParameterList();
+  bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
+
+  // General problem informations are gathered from data stored in the problem matix.
+  RCP<const LWGraph_kokkos> graph = Get<RCP<LWGraph_kokkos> >(currentLevel, "Graph");
+  RCP<const Map> fineMap          = graph->GetDomainMap();
+  const int myRank                = fineMap->getComm()->getRank();
+  const LO dofsPerNode            = Get<LO>(currentLevel, "DofsPerNode");
+
+  // Since we want to operate on nodes and not dof, we need to modify the rowMap in order to
+  // obtain a nodeMap.
+  const int interpolationOrder = pL.get<int>("aggregation: coarsening order");
+  std::string outputType       = pL.get<std::string>("aggregation: output type");
+  const bool outputAggregates  = (outputType == "Aggregates" ? true : false);
+  Array<LO> lFineNodesPerDir(3);
+  int numDimensions;
+  if (currentLevel.GetLevelID() == 0) {
+    // On level 0, data is provided by applications and has no associated factory.
+    lFineNodesPerDir = currentLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
+    numDimensions    = currentLevel.Get<int>("numDimensions", NoFactory::get());
+  } else {
+    // On level > 0, data is provided directly by generating factories.
+    lFineNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
+    numDimensions    = Get<int>(currentLevel, "numDimensions");
+  }
+
+  // First make sure that input parameters are set logically based on dimension
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim >= numDimensions) {
+      lFineNodesPerDir[dim] = 1;
     }
-
-    // Get the coarsening rate
-    std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
-    Teuchos::Array<LO> coarseRate;
-    try {
-      coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
-    } catch(const Teuchos::InvalidArrayStringRepresentation& e) {
-      GetOStream(Errors,-1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
-                            << std::endl;
-      throw e;
-    }
-    TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
-                               Exceptions::RuntimeError,
-                               "\"aggregation: coarsening rate\" must have at least as many"
-                               " components as the number of spatial dimensions in the problem.");
-
-    // Now that we have extracted info from the level, create the IndexManager
-    RCP<IndexManager_kokkos> geoData = rcp(new IndexManager_kokkos(numDimensions,
-                                                                   interpolationOrder, myRank,
-                                                                   lFineNodesPerDir,
-                                                                   coarseRate));
-
-    *out << "The index manager has now been built" << std::endl;
-    TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements()
-                               != static_cast<size_t>(geoData->getNumLocalFineNodes()),
-                               Exceptions::RuntimeError,
-                               "The local number of elements in the graph's map is not equal to "
-                               "the number of nodes given by: lNodesPerDim!");
-
-    // Now we are ready for the big loop over the fine node that will assign each
-    // node on the fine grid to an aggregate and a processor.
-    RCP<AggregationStructuredAlgorithm_kokkos> myStructuredAlgorithm
-      = rcp(new AggregationStructuredAlgorithm_kokkos());
-
-    if(interpolationOrder == 0 && outputAggregates){
-      RCP<Aggregates> aggregates = rcp(new Aggregates(graph->GetDomainMap()));
-      aggregates->setObjectLabel("ST");
-      aggregates->SetIndexManagerKokkos(geoData);
-      aggregates->AggregatesCrossProcessors(false);
-      aggregates->SetNumAggregates(geoData->getNumCoarseNodes());
-
-      LO numNonAggregatedNodes = geoData->getNumLocalFineNodes();
-      Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNonAggregatedNodes);
-      Kokkos::parallel_for("StructuredAggregation: initialize aggStat",
-                           Kokkos::RangePolicy<execution_space>(0, numNonAggregatedNodes),
-                           KOKKOS_LAMBDA(const LO nodeIdx) {aggStat(nodeIdx) = READY;});
-
-      myStructuredAlgorithm->BuildAggregates(pL, *graph, *aggregates, aggStat,
-                                             numNonAggregatedNodes);
-
-      *out << "numNonAggregatedNodes: " << numNonAggregatedNodes << std::endl;
-
-      TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError,
-                                 "MueLu::StructuredAggregationFactory::Build: Leftover nodes found! Error!");
-      aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
-      GetOStream(Statistics1) << aggregates->description() << std::endl;
-      Set(currentLevel, "Aggregates", aggregates);
-
-    } else {
-      // Create Coarse Data
-      RCP<CrsGraph> myGraph;
-      myStructuredAlgorithm->BuildGraph(*graph, geoData, dofsPerNode, myGraph);
-      Set(currentLevel, "prolongatorGraph", myGraph);
-    }
-
-    Set(currentLevel, "lCoarseNodesPerDim",           geoData->getCoarseNodesPerDirArray());
-    Set(currentLevel, "indexManager",                 geoData);
-    Set(currentLevel, "structuredInterpolationOrder", interpolationOrder);
-    Set(currentLevel, "numDimensions",                numDimensions);
-
-  } // Build()
-
-} //namespace MueLu
+  }
+
+  // Get the coarsening rate
+  std::string coarseningRate = pL.get<std::string>("aggregation: coarsening rate");
+  Teuchos::Array<LO> coarseRate;
+  try {
+    coarseRate = Teuchos::fromStringToArray<LO>(coarseningRate);
+  } catch (const Teuchos::InvalidArrayStringRepresentation& e) {
+    GetOStream(Errors, -1) << " *** \"aggregation: coarsening rate\" must be a string convertible into an array! *** "
+                           << std::endl;
+    throw e;
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION((coarseRate.size() > 1) && (coarseRate.size() < numDimensions),
+                             Exceptions::RuntimeError,
+                             "\"aggregation: coarsening rate\" must have at least as many"
+                             " components as the number of spatial dimensions in the problem.");
+
+  // Now that we have extracted info from the level, create the IndexManager
+  RCP<IndexManager_kokkos> geoData = rcp(new IndexManager_kokkos(numDimensions,
+                                                                 interpolationOrder, myRank,
+                                                                 lFineNodesPerDir,
+                                                                 coarseRate));
+
+  *out << "The index manager has now been built" << std::endl;
+  TEUCHOS_TEST_FOR_EXCEPTION(fineMap->getLocalNumElements() != static_cast<size_t>(geoData->getNumLocalFineNodes()),
+                             Exceptions::RuntimeError,
+                             "The local number of elements in the graph's map is not equal to "
+                             "the number of nodes given by: lNodesPerDim!");
+
+  // Now we are ready for the big loop over the fine node that will assign each
+  // node on the fine grid to an aggregate and a processor.
+  RCP<AggregationStructuredAlgorithm_kokkos> myStructuredAlgorithm = rcp(new AggregationStructuredAlgorithm_kokkos());
+
+  if (interpolationOrder == 0 && outputAggregates) {
+    RCP<Aggregates> aggregates = rcp(new Aggregates(graph->GetDomainMap()));
+    aggregates->setObjectLabel("ST");
+    aggregates->SetIndexManagerKokkos(geoData);
+    aggregates->AggregatesCrossProcessors(false);
+    aggregates->SetNumAggregates(geoData->getNumCoarseNodes());
+
+    LO numNonAggregatedNodes = geoData->getNumLocalFineNodes();
+    Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNonAggregatedNodes);
+    Kokkos::parallel_for(
+        "StructuredAggregation: initialize aggStat",
+        Kokkos::RangePolicy<execution_space>(0, numNonAggregatedNodes),
+        KOKKOS_LAMBDA(const LO nodeIdx) { aggStat(nodeIdx) = READY; });
+
+    myStructuredAlgorithm->BuildAggregates(pL, *graph, *aggregates, aggStat,
+                                           numNonAggregatedNodes);
+
+    *out << "numNonAggregatedNodes: " << numNonAggregatedNodes << std::endl;
+
+    TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError,
+                               "MueLu::StructuredAggregationFactory::Build: Leftover nodes found! Error!");
+    aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
+    GetOStream(Statistics1) << aggregates->description() << std::endl;
+    Set(currentLevel, "Aggregates", aggregates);
+
+  } else {
+    // Create Coarse Data
+    RCP<CrsGraph> myGraph;
+    myStructuredAlgorithm->BuildGraph(*graph, geoData, dofsPerNode, myGraph);
+    Set(currentLevel, "prolongatorGraph", myGraph);
+  }
+
+  Set(currentLevel, "lCoarseNodesPerDim", geoData->getCoarseNodesPerDirArray());
+  Set(currentLevel, "indexManager", geoData);
+  Set(currentLevel, "structuredInterpolationOrder", interpolationOrder);
+  Set(currentLevel, "numDimensions", numDimensions);
+
+}  // Build()
+
+}  // namespace MueLu
 
 #endif /* MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_DEF_HPP */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_decl.hpp
index 9488bfe3dd72..3f5cda1e3962 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_decl.hpp
@@ -77,69 +77,67 @@ namespace MueLu {
     correspond to nodes. While not strictly necessary, it might be convenient.
 */
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class GlobalLexicographicIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GlobalLexicographicIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
+ public:
+  GlobalLexicographicIndexManager();
 
-    GlobalLexicographicIndexManager();
+  GlobalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                                  const int NumDimensions, const int interpolationOrder,
+                                  const Array<GO> GFineNodesPerDir,
+                                  const Array<LO> LFineNodesPerDir, const Array<LO> CoarseRate,
+                                  const GO MinGlobalIndex);
 
-    GlobalLexicographicIndexManager(const RCP< const Teuchos::Comm<int> > comm, const bool coupled,
-                                    const int NumDimensions, const int interpolationOrder,
-                                    const Array<GO> GFineNodesPerDir,
-                                    const Array<LO> LFineNodesPerDir, const Array<LO> CoarseRate,
-                                    const GO MinGlobalIndex);
+  virtual ~GlobalLexicographicIndexManager() {}
 
-    virtual ~GlobalLexicographicIndexManager() {}
+  void computeGlobalCoarseParameters();
 
-    void computeGlobalCoarseParameters();
+  void getGhostedNodesData(const RCP<const Map> fineMap,
+                           Array<LO>& ghostedNodeCoarseLIDs,
+                           Array<int>& ghostedNodeCoarsePIDs,
+                           Array<GO>& ghostedNodeCoarseGIDs) const;
 
-    void getGhostedNodesData(const RCP<const Map> fineMap,
-                            Array<LO>& ghostedNodeCoarseLIDs,
-                            Array<int>& ghostedNodeCoarsePIDs,
-                            Array<GO>& ghostedNodeCoarseGIDs) const;
+  void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                          Array<GO>& coarseNodeCoarseGIDs,
+                          Array<GO>& coarseNodeFineGIDs) const;
 
-    void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                            Array<GO>& coarseNodeCoarseGIDs,
-                            Array<GO>& coarseNodeFineGIDs) const;
+  std::vector<std::vector<GO> > getCoarseMeshData() const;
 
-    std::vector<std::vector<GO> > getCoarseMeshData() const;
+  void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
+ private:
+};
 
-  private:
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GLOBALLEXICOGRPHICINDEXMANAGER_SHORT
-#endif // MUELU_GLOBALLEXICOGRPHICINDEXMANAGER_DECL_HPP
+#endif  // MUELU_GLOBALLEXICOGRPHICINDEXMANAGER_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_def.hpp
index 8211e1a461bd..20bc4e0e7419 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_GlobalLexicographicIndexManager_def.hpp
@@ -51,341 +51,330 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  GlobalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                                  const int NumDimensions, const int interpolationOrder,
-                                  const Array<GO> GFineNodesPerDir,
-                                  const Array<LO> LFineNodesPerDir, const Array<LO> CoarseRate,
-                                  const GO MinGlobalIndex) :
-    IndexManager(comm, coupled, false, NumDimensions, interpolationOrder, GFineNodesPerDir, LFineNodesPerDir) {
-
-    // Load coarse rate, being careful about formating.
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < this->numDimensions) {
-        if(CoarseRate.size() == 1) {
-          this->coarseRate[dim] = CoarseRate[0];
-        } else if(CoarseRate.size() == this->numDimensions) {
-          this->coarseRate[dim] = CoarseRate[dim];
-        }
-      } else {
-        this->coarseRate[dim] = 1;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    GlobalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                                    const int NumDimensions, const int interpolationOrder,
+                                    const Array<GO> GFineNodesPerDir,
+                                    const Array<LO> LFineNodesPerDir, const Array<LO> CoarseRate,
+                                    const GO MinGlobalIndex)
+  : IndexManager(comm, coupled, false, NumDimensions, interpolationOrder, GFineNodesPerDir, LFineNodesPerDir) {
+  // Load coarse rate, being careful about formating.
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < this->numDimensions) {
+      if (CoarseRate.size() == 1) {
+        this->coarseRate[dim] = CoarseRate[0];
+      } else if (CoarseRate.size() == this->numDimensions) {
+        this->coarseRate[dim] = CoarseRate[dim];
       }
+    } else {
+      this->coarseRate[dim] = 1;
     }
+  }
 
-    {
-      GO tmp = 0;
-      this->startIndices[2]= MinGlobalIndex / (this->gFineNodesPerDir[1]*this->gFineNodesPerDir[0]);
-      tmp                  = MinGlobalIndex % (this->gFineNodesPerDir[1]*this->gFineNodesPerDir[0]);
-      this->startIndices[1]= tmp / this->gFineNodesPerDir[0];
-      this->startIndices[0]= tmp % this->gFineNodesPerDir[0];
+  {
+    GO tmp                = 0;
+    this->startIndices[2] = MinGlobalIndex / (this->gFineNodesPerDir[1] * this->gFineNodesPerDir[0]);
+    tmp                   = MinGlobalIndex % (this->gFineNodesPerDir[1] * this->gFineNodesPerDir[0]);
+    this->startIndices[1] = tmp / this->gFineNodesPerDir[0];
+    this->startIndices[0] = tmp % this->gFineNodesPerDir[0];
 
-      for(int dim = 0; dim < 3; ++dim) {
-        this->startIndices[dim + 3] = this->startIndices[dim] + this->lFineNodesPerDir[dim] - 1;
-      }
+    for (int dim = 0; dim < 3; ++dim) {
+      this->startIndices[dim + 3] = this->startIndices[dim] + this->lFineNodesPerDir[dim] - 1;
     }
-
-    this->computeMeshParameters();
-    computeGlobalCoarseParameters();
-
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGlobalCoarseParameters() {
-    this->gNumCoarseNodes10 = this->gCoarseNodesPerDir[0]*this->gCoarseNodesPerDir[1];
-    this->gNumCoarseNodes   = this->gNumCoarseNodes10*this->gCoarseNodesPerDir[2];
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodesData(const RCP<const Map> fineMap,
-                     Array<LO>& ghostedNodeCoarseLIDs, Array<int>& ghostedNodeCoarsePIDs, Array<GO>&ghostedNodeCoarseGIDs) const {
-
-    ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
-    ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
-    ghostedNodeCoarseGIDs.resize(this->numGhostedNodes);
-
-    // Find the GIDs, LIDs and PIDs of the coarse points on the fine mesh and coarse
-    // mesh as this data will be used to fill vertex2AggId and procWinner vectors.
-    Array<GO> lCoarseNodeCoarseGIDs(this->lNumCoarseNodes),
+  this->computeMeshParameters();
+  computeGlobalCoarseParameters();
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGlobalCoarseParameters() {
+  this->gNumCoarseNodes10 = this->gCoarseNodesPerDir[0] * this->gCoarseNodesPerDir[1];
+  this->gNumCoarseNodes   = this->gNumCoarseNodes10 * this->gCoarseNodesPerDir[2];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodesData(const RCP<const Map> fineMap,
+                        Array<LO>& ghostedNodeCoarseLIDs, Array<int>& ghostedNodeCoarsePIDs, Array<GO>& ghostedNodeCoarseGIDs) const {
+  ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
+  ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
+  ghostedNodeCoarseGIDs.resize(this->numGhostedNodes);
+
+  // Find the GIDs, LIDs and PIDs of the coarse points on the fine mesh and coarse
+  // mesh as this data will be used to fill vertex2AggId and procWinner vectors.
+  Array<GO> lCoarseNodeCoarseGIDs(this->lNumCoarseNodes),
       lCoarseNodeFineGIDs(this->lNumCoarseNodes);
-    Array<GO> ghostedCoarseNodeFineGIDs(this->numGhostedNodes);
-    Array<LO> ghostedCoarseNodeCoarseIndices(3), ghostedCoarseNodeFineIndices(3), ijk(3);
-    LO currentIndex = -1, currentCoarseIndex = -1;
-    for(ijk[2] = 0; ijk[2] < this->ghostedNodesPerDir[2]; ++ijk[2]) {
-      for(ijk[1] = 0; ijk[1] < this->ghostedNodesPerDir[1]; ++ijk[1]) {
-        for(ijk[0] = 0; ijk[0] < this->ghostedNodesPerDir[0]; ++ijk[0]) {
-          currentIndex = ijk[2]*this->numGhostedNodes10 + ijk[1]*this->ghostedNodesPerDir[0] + ijk[0];
-          ghostedCoarseNodeCoarseIndices[0] = this->startGhostedCoarseNode[0] + ijk[0];
-          ghostedCoarseNodeCoarseIndices[1] = this->startGhostedCoarseNode[1] + ijk[1];
-          ghostedCoarseNodeCoarseIndices[2] = this->startGhostedCoarseNode[2] + ijk[2];
-          GO myCoarseGID = ghostedCoarseNodeCoarseIndices[0]
-            + ghostedCoarseNodeCoarseIndices[1]*this->gCoarseNodesPerDir[0]
-            + ghostedCoarseNodeCoarseIndices[2]*this->gNumCoarseNodes10;
-          ghostedNodeCoarseGIDs[currentIndex] = myCoarseGID;
-          GO myGID = 0, factor[3] = {};
-          factor[2] = this->gNumFineNodes10;
-          factor[1] = this->gFineNodesPerDir[0];
-          factor[0] = 1;
-          for(int dim = 0; dim < 3; ++dim) {
-            if(dim < this->numDimensions) {
-              if(this->startIndices[dim] - this->offsets[dim] + ijk[dim]*this->coarseRate[dim]
-                 < this->gFineNodesPerDir[dim] - 1) {
-                myGID += (this->startIndices[dim] - this->offsets[dim]
-                          + ijk[dim]*this->coarseRate[dim])*factor[dim];
-              } else {
-                myGID += (this->startIndices[dim] - this->offsets[dim] + (ijk[dim] - 1)
-                          *this->coarseRate[dim] + this->endRate[dim])*factor[dim];
-              }
+  Array<GO> ghostedCoarseNodeFineGIDs(this->numGhostedNodes);
+  Array<LO> ghostedCoarseNodeCoarseIndices(3), ghostedCoarseNodeFineIndices(3), ijk(3);
+  LO currentIndex = -1, currentCoarseIndex = -1;
+  for (ijk[2] = 0; ijk[2] < this->ghostedNodesPerDir[2]; ++ijk[2]) {
+    for (ijk[1] = 0; ijk[1] < this->ghostedNodesPerDir[1]; ++ijk[1]) {
+      for (ijk[0] = 0; ijk[0] < this->ghostedNodesPerDir[0]; ++ijk[0]) {
+        currentIndex                        = ijk[2] * this->numGhostedNodes10 + ijk[1] * this->ghostedNodesPerDir[0] + ijk[0];
+        ghostedCoarseNodeCoarseIndices[0]   = this->startGhostedCoarseNode[0] + ijk[0];
+        ghostedCoarseNodeCoarseIndices[1]   = this->startGhostedCoarseNode[1] + ijk[1];
+        ghostedCoarseNodeCoarseIndices[2]   = this->startGhostedCoarseNode[2] + ijk[2];
+        GO myCoarseGID                      = ghostedCoarseNodeCoarseIndices[0] + ghostedCoarseNodeCoarseIndices[1] * this->gCoarseNodesPerDir[0] + ghostedCoarseNodeCoarseIndices[2] * this->gNumCoarseNodes10;
+        ghostedNodeCoarseGIDs[currentIndex] = myCoarseGID;
+        GO myGID = 0, factor[3] = {};
+        factor[2] = this->gNumFineNodes10;
+        factor[1] = this->gFineNodesPerDir[0];
+        factor[0] = 1;
+        for (int dim = 0; dim < 3; ++dim) {
+          if (dim < this->numDimensions) {
+            if (this->startIndices[dim] - this->offsets[dim] + ijk[dim] * this->coarseRate[dim] < this->gFineNodesPerDir[dim] - 1) {
+              myGID += (this->startIndices[dim] - this->offsets[dim] + ijk[dim] * this->coarseRate[dim]) * factor[dim];
+            } else {
+              myGID += (this->startIndices[dim] - this->offsets[dim] + (ijk[dim] - 1) * this->coarseRate[dim] + this->endRate[dim]) * factor[dim];
             }
           }
-          // lbv 02-08-2018:
-          // This check is simplistic and should be replaced by a condition that checks
-          // if the local tuple of the current index is wihin the range of local nodes
-          // or not in the range of ghosted nodes.
-          if((!this->ghostInterface[0] || ijk[0] != 0) &&
-             (!this->ghostInterface[2] || ijk[1] != 0) &&
-             (!this->ghostInterface[4] || ijk[2] != 0) &&
-             (!this->ghostInterface[1] || ijk[0] != this->ghostedNodesPerDir[0] - 1) &&
-             (!this->ghostInterface[3] || ijk[1] != this->ghostedNodesPerDir[1] - 1) &&
-             (!this->ghostInterface[5] || ijk[2] != this->ghostedNodesPerDir[2] - 1)) {
-
-            // this->getGhostedNodeFineLID(ijk[0], ijk[1], ijk[2], coarseNodeFineLID);
-            if(this->interpolationOrder_ == 0) {
-              currentCoarseIndex = 0;
-              if(this->ghostInterface[4]) {
-                currentCoarseIndex += (ijk[2] - 1)*this->lNumCoarseNodes10;
-              } else {
-                currentCoarseIndex += ijk[2]*this->lNumCoarseNodes10;
-              }
-              if(this->ghostInterface[2]) {
-                currentCoarseIndex += (ijk[1] - 1)*this->getLocalCoarseNodesInDir(0);
-              } else {
-                currentCoarseIndex += ijk[1]*this->getLocalCoarseNodesInDir(0);
-              }
-              if(this->ghostInterface[0]) {
-                currentCoarseIndex += ijk[0] - 1;
-              } else {
-                currentCoarseIndex += ijk[0];
-              }
+        }
+        // lbv 02-08-2018:
+        // This check is simplistic and should be replaced by a condition that checks
+        // if the local tuple of the current index is wihin the range of local nodes
+        // or not in the range of ghosted nodes.
+        if ((!this->ghostInterface[0] || ijk[0] != 0) &&
+            (!this->ghostInterface[2] || ijk[1] != 0) &&
+            (!this->ghostInterface[4] || ijk[2] != 0) &&
+            (!this->ghostInterface[1] || ijk[0] != this->ghostedNodesPerDir[0] - 1) &&
+            (!this->ghostInterface[3] || ijk[1] != this->ghostedNodesPerDir[1] - 1) &&
+            (!this->ghostInterface[5] || ijk[2] != this->ghostedNodesPerDir[2] - 1)) {
+          // this->getGhostedNodeFineLID(ijk[0], ijk[1], ijk[2], coarseNodeFineLID);
+          if (this->interpolationOrder_ == 0) {
+            currentCoarseIndex = 0;
+            if (this->ghostInterface[4]) {
+              currentCoarseIndex += (ijk[2] - 1) * this->lNumCoarseNodes10;
             } else {
-              this->getGhostedNodeCoarseLID(ijk[0], ijk[1], ijk[2], currentCoarseIndex);
+              currentCoarseIndex += ijk[2] * this->lNumCoarseNodes10;
             }
-
-            lCoarseNodeCoarseGIDs[currentCoarseIndex] = myCoarseGID;
-            lCoarseNodeFineGIDs[currentCoarseIndex]   = myGID;
+            if (this->ghostInterface[2]) {
+              currentCoarseIndex += (ijk[1] - 1) * this->getLocalCoarseNodesInDir(0);
+            } else {
+              currentCoarseIndex += ijk[1] * this->getLocalCoarseNodesInDir(0);
+            }
+            if (this->ghostInterface[0]) {
+              currentCoarseIndex += ijk[0] - 1;
+            } else {
+              currentCoarseIndex += ijk[0];
+            }
+          } else {
+            this->getGhostedNodeCoarseLID(ijk[0], ijk[1], ijk[2], currentCoarseIndex);
           }
-          ghostedCoarseNodeFineGIDs[currentIndex] = myGID;
-        }
-      }
-    }
-
-    RCP<const Map> coarseMap = Xpetra::MapFactory<LO,GO,NO>::Build (fineMap->lib(),
-                                                                    this->gNumCoarseNodes,
-                                                                    lCoarseNodeCoarseGIDs(),
-                                                                    fineMap->getIndexBase(),
-                                                                    fineMap->getComm());
-
-    coarseMap->getRemoteIndexList(ghostedNodeCoarseGIDs(),
-                                  ghostedNodeCoarsePIDs(),
-                                  ghostedNodeCoarseLIDs());
-
-  } // End getGhostedMeshData
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                     Array<GO>& coarseNodeCoarseGIDs,
-                     Array<GO>& coarseNodeFineGIDs) const {
-
-    // Allocate sufficient storage space for outputs
-    coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
-    coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
-
-    // Load all the GIDs on the fine mesh
-    ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
-
-    Array<GO> coarseStartIndices(3);
-    GO tmp;
-    for(int dim = 0; dim < 3; ++dim) {
-      coarseStartIndices[dim] = this->startIndices[dim] / this->coarseRate[dim];
-      tmp = this->startIndices[dim] % this->coarseRate[dim];
-      if(tmp > 0) {++coarseStartIndices[dim];}
-    }
 
-    // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
-    LO fineLID;
-    Array<LO> lCoarseIndices(3);
-    Array<GO> gCoarseIndices(3);
-    for(LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
-      this->getCoarseNodeLocalTuple(coarseLID,
-                                    lCoarseIndices[0],
-                                    lCoarseIndices[1],
-                                    lCoarseIndices[2]);
-      getCoarseNodeFineLID(lCoarseIndices[0], lCoarseIndices[1], lCoarseIndices[2], fineLID);
-      coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
-
-      // Get Coarse Global IJK
-      for(int dim=0; dim<3; dim++) {
-        gCoarseIndices[dim] = coarseStartIndices[dim] + lCoarseIndices[dim];
+          lCoarseNodeCoarseGIDs[currentCoarseIndex] = myCoarseGID;
+          lCoarseNodeFineGIDs[currentCoarseIndex]   = myGID;
+        }
+        ghostedCoarseNodeFineGIDs[currentIndex] = myGID;
       }
-      getCoarseNodeGID(gCoarseIndices[0],
-                       gCoarseIndices[1],
-                       gCoarseIndices[2],
-                       coarseNodeCoarseGIDs[coarseLID] );
-
     }
-
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<std::vector<GlobalOrdinal> > GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseMeshData() const {
-    std::vector<std::vector<GO> > coarseMeshData;
-    return coarseMeshData;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const {
-    GO tmp;
-    k   = myGID / this->gNumFineNodes10;
-    tmp = myGID % this->gNumFineNodes10;
-    j   = tmp / this->gFineNodesPerDir[0];
-    i   = tmp % this->gFineNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp / this->lFineNodesPerDir[0];
-    i   = tmp % this->lFineNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp / this->lFineNodesPerDir[0];
-    i   = tmp % this->lFineNodesPerDir[0];
-
-    k += this->offsets[2];
-    j += this->offsets[1];
-    i += this->offsets[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const {
-    myGID = k*this->gNumFineNodes10 + j*this->gFineNodesPerDir[0] + i;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    myLID = k*this->lNumFineNodes10 + j*this->lFineNodesPerDir[0] + i;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const {
-    GO tmp;
-    k   = myGID / this->gNumCoarseNodes10;
-    tmp = myGID % this->gNumCoarseNodes10;
-    j   = tmp / this->gCoarseNodesPerDir[0];
-    i   = tmp % this->gCoarseNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumCoarseNodes10;
-    tmp = myLID % this->lNumCoarseNodes10;
-    j   = tmp   / this->lCoarseNodesPerDir[0];
-    i   = tmp   % this->lCoarseNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const {
-    myGID = k*this->gNumCoarseNodes10 + j*this->gCoarseNodesPerDir[0] + i;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    myLID = k*this->lNumCoarseNodes10 + j*this->lCoarseNodesPerDir[0] + i;
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    myLID = k*this->numGhostedNodes10 + j*this->ghostedNodesPerDir[0] + i;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    // Assumptions: (i,j,k) is a tuple on the coarse mesh
-    //              myLID is the corresponding local ID on the fine mesh
-    const LO multiplier[3] = {1, this->lFineNodesPerDir[0], this->lNumFineNodes10};
-    const LO indices[3] = {i, j, k};
-
-    myLID = 0;
-    for(int dim = 0; dim < 3; ++dim) {
-      if((indices[dim] == this->getLocalCoarseNodesInDir(dim) - 1) && this->meshEdge[2*dim + 1]) {
-        // We are dealing with the last node on the mesh in direction dim
-        // so we can simply use the number of nodes on the fine mesh in that direction
-        myLID += (this->getLocalFineNodesInDir(dim) - 1)*multiplier[dim];
-      } else {
-        myLID += (indices[dim]*this->getCoarseningRate(dim) + this->getCoarseNodeOffset(dim))
-          *multiplier[dim];
-      }
+  RCP<const Map> coarseMap = Xpetra::MapFactory<LO, GO, NO>::Build(fineMap->lib(),
+                                                                   this->gNumCoarseNodes,
+                                                                   lCoarseNodeCoarseGIDs(),
+                                                                   fineMap->getIndexBase(),
+                                                                   fineMap->getComm());
+
+  coarseMap->getRemoteIndexList(ghostedNodeCoarseGIDs(),
+                                ghostedNodeCoarsePIDs(),
+                                ghostedNodeCoarseLIDs());
+
+}  // End getGhostedMeshData
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                       Array<GO>& coarseNodeCoarseGIDs,
+                       Array<GO>& coarseNodeFineGIDs) const {
+  // Allocate sufficient storage space for outputs
+  coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
+  coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
+
+  // Load all the GIDs on the fine mesh
+  ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
+
+  Array<GO> coarseStartIndices(3);
+  GO tmp;
+  for (int dim = 0; dim < 3; ++dim) {
+    coarseStartIndices[dim] = this->startIndices[dim] / this->coarseRate[dim];
+    tmp                     = this->startIndices[dim] % this->coarseRate[dim];
+    if (tmp > 0) {
+      ++coarseStartIndices[dim];
     }
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    LO itmp = i - (this->offsets[0] > 0 ? 1 : 0);
-    LO jtmp = j - (this->offsets[1] > 0 ? 1 : 0);
-    LO ktmp = k - (this->offsets[2] > 0 ? 1 : 0);
-    myLID = 0;
-    if(ktmp*this->coarseRate[2] < this->lFineNodesPerDir[2]) {
-      myLID += ktmp*this->coarseRate[2]*this->lNumCoarseNodes10;
-    } else {
-      myLID += (this->lFineNodesPerDir[2] - 1)*this->lNumCoarseNodes10;
-    }
-
-    if(jtmp*this->coarseRate[1] < this->lFineNodesPerDir[1]) {
-      myLID += jtmp*this->coarseRate[1]*this->lFineNodesPerDir[0];
-    } else {
-      myLID += (this->lFineNodesPerDir[1] - 1)*this->lFineNodesPerDir[1];
+  // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
+  LO fineLID;
+  Array<LO> lCoarseIndices(3);
+  Array<GO> gCoarseIndices(3);
+  for (LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
+    this->getCoarseNodeLocalTuple(coarseLID,
+                                  lCoarseIndices[0],
+                                  lCoarseIndices[1],
+                                  lCoarseIndices[2]);
+    getCoarseNodeFineLID(lCoarseIndices[0], lCoarseIndices[1], lCoarseIndices[2], fineLID);
+    coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
+
+    // Get Coarse Global IJK
+    for (int dim = 0; dim < 3; dim++) {
+      gCoarseIndices[dim] = coarseStartIndices[dim] + lCoarseIndices[dim];
     }
-
-    if(itmp*this->coarseRate[0] < this->lFineNodesPerDir[0]) {
-      myLID += itmp*this->coarseRate[0];
+    getCoarseNodeGID(gCoarseIndices[0],
+                     gCoarseIndices[1],
+                     gCoarseIndices[2],
+                     coarseNodeCoarseGIDs[coarseLID]);
+  }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<std::vector<GlobalOrdinal> > GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseMeshData() const {
+  std::vector<std::vector<GO> > coarseMeshData;
+  return coarseMeshData;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const {
+  GO tmp;
+  k   = myGID / this->gNumFineNodes10;
+  tmp = myGID % this->gNumFineNodes10;
+  j   = tmp / this->gFineNodesPerDir[0];
+  i   = tmp % this->gFineNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+
+  k += this->offsets[2];
+  j += this->offsets[1];
+  i += this->offsets[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const {
+  myGID = k * this->gNumFineNodes10 + j * this->gFineNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  myLID = k * this->lNumFineNodes10 + j * this->lFineNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const {
+  GO tmp;
+  k   = myGID / this->gNumCoarseNodes10;
+  tmp = myGID % this->gNumCoarseNodes10;
+  j   = tmp / this->gCoarseNodesPerDir[0];
+  i   = tmp % this->gCoarseNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumCoarseNodes10;
+  tmp = myLID % this->lNumCoarseNodes10;
+  j   = tmp / this->lCoarseNodesPerDir[0];
+  i   = tmp % this->lCoarseNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const {
+  myGID = k * this->gNumCoarseNodes10 + j * this->gCoarseNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  myLID = k * this->lNumCoarseNodes10 + j * this->lCoarseNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  myLID = k * this->numGhostedNodes10 + j * this->ghostedNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  // Assumptions: (i,j,k) is a tuple on the coarse mesh
+  //              myLID is the corresponding local ID on the fine mesh
+  const LO multiplier[3] = {1, this->lFineNodesPerDir[0], this->lNumFineNodes10};
+  const LO indices[3]    = {i, j, k};
+
+  myLID = 0;
+  for (int dim = 0; dim < 3; ++dim) {
+    if ((indices[dim] == this->getLocalCoarseNodesInDir(dim) - 1) && this->meshEdge[2 * dim + 1]) {
+      // We are dealing with the last node on the mesh in direction dim
+      // so we can simply use the number of nodes on the fine mesh in that direction
+      myLID += (this->getLocalFineNodesInDir(dim) - 1) * multiplier[dim];
     } else {
-      myLID += this->lFineNodesPerDir[0] - 1;
+      myLID += (indices[dim] * this->getCoarseningRate(dim) + this->getCoarseNodeOffset(dim)) * multiplier[dim];
     }
   }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  LO itmp = i - (this->offsets[0] > 0 ? 1 : 0);
+  LO jtmp = j - (this->offsets[1] > 0 ? 1 : 0);
+  LO ktmp = k - (this->offsets[2] > 0 ? 1 : 0);
+  myLID   = 0;
+  if (ktmp * this->coarseRate[2] < this->lFineNodesPerDir[2]) {
+    myLID += ktmp * this->coarseRate[2] * this->lNumCoarseNodes10;
+  } else {
+    myLID += (this->lFineNodesPerDir[2] - 1) * this->lNumCoarseNodes10;
+  }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    LO itmp = i - (this->offsets[0] > 0 ? 1 : 0);
-    LO jtmp = j - (this->offsets[1] > 0 ? 1 : 0);
-    LO ktmp = k - (this->offsets[2] > 0 ? 1 : 0);
-    myLID = ktmp*this->lNumCoarseNodes10 + jtmp*this->lCoarseNodesPerDir[0] + itmp;
+  if (jtmp * this->coarseRate[1] < this->lFineNodesPerDir[1]) {
+    myLID += jtmp * this->coarseRate[1] * this->lFineNodesPerDir[0];
+  } else {
+    myLID += (this->lFineNodesPerDir[1] - 1) * this->lFineNodesPerDir[1];
   }
 
-} //namespace MueLu
+  if (itmp * this->coarseRate[0] < this->lFineNodesPerDir[0]) {
+    myLID += itmp * this->coarseRate[0];
+  } else {
+    myLID += this->lFineNodesPerDir[0] - 1;
+  }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  LO itmp = i - (this->offsets[0] > 0 ? 1 : 0);
+  LO jtmp = j - (this->offsets[1] > 0 ? 1 : 0);
+  LO ktmp = k - (this->offsets[2] > 0 ? 1 : 0);
+  myLID   = ktmp * this->lNumCoarseNodes10 + jtmp * this->lCoarseNodesPerDir[0] + itmp;
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_decl.hpp
index a0e809aaba20..c7dab301c795 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_decl.hpp
@@ -77,93 +77,90 @@ namespace MueLu {
     correspond to nodes. While not strictly necessary, it might be convenient.
 */
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalLexicographicIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalLexicographicIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
+ public:
+  LocalLexicographicIndexManager() = default;
 
-    LocalLexicographicIndexManager() = default;
+  LocalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                                 const int NumDimensions, const int interpolationOrder,
+                                 const int MyRank, const int NumRanks,
+                                 const Array<GO> GFineNodesPerDir,
+                                 const Array<LO> LFineNodesPerDir,
+                                 const Array<LO> CoarseRate, const Array<GO> MeshData);
 
-    LocalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                                   const int NumDimensions, const int interpolationOrder,
-                                   const int MyRank, const int NumRanks,
-                                   const Array<GO> GFineNodesPerDir,
-                                   const Array<LO> LFineNodesPerDir,
-                                   const Array<LO> CoarseRate, const Array<GO> MeshData);
+  virtual ~LocalLexicographicIndexManager() {}
 
-    virtual ~LocalLexicographicIndexManager() {}
+  void computeGlobalCoarseParameters();
 
-    void computeGlobalCoarseParameters();
+  void getGhostedNodesData(const RCP<const Map> fineMap,
+                           Array<LO>& ghostedNodeCoarseLIDs,
+                           Array<int>& ghostedNodeCoarsePIDs,
+                           Array<GO>& ghostedNodeCoarseGIDs) const;
 
-    void getGhostedNodesData(const RCP<const Map> fineMap,
-                             Array<LO>& ghostedNodeCoarseLIDs,
-                             Array<int>& ghostedNodeCoarsePIDs,
-                             Array<GO>& ghostedNodeCoarseGIDs) const;
+  void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                          Array<GO>& coarseNodeCoarseGIDs,
+                          Array<GO>& coarseNodeFineGIDs) const;
 
-    void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                            Array<GO>& coarseNodeCoarseGIDs,
-                            Array<GO>& coarseNodeFineGIDs) const;
+  std::vector<std::vector<GO> > getCoarseMeshData() const;
 
-    std::vector<std::vector<GO> > getCoarseMeshData() const;
+  void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
+ private:
+  const int myRank;    ///< Local rank ID.
+  const int numRanks;  ///< Number of ranks used to decompose the problem.
 
-  private:
+  // Iterator delimiting the entries in meshData that correspond to the block that owns the local
+  // part of the mesh.
+  typename std::vector<std::vector<GO> >::iterator myBlockStart, myBlockEnd;
 
-    const int myRank;                   ///< Local rank ID.
-    const int numRanks;                 ///< Number of ranks used to decompose the problem.
+  int pi, pj, pk;  ///< Number of processors in each diretcion.
 
-    // Iterator delimiting the entries in meshData that correspond to the block that owns the local
-    // part of the mesh.
-    typename std::vector<std::vector<GO> >::iterator myBlockStart, myBlockEnd;
+  int numBlocks;  ///< Number of mesh block.
+  int myBlock;    ///< local mesh block ID.
 
-    int pi, pj, pk;              ///< Number of processors in each diretcion.
+  int myRankIndex;                               ///< local process index for record in meshData after sorting.
+  Array<int> rankIndices;                        ///< mapping between rank ID and reordered rank ID.
+  std::vector<std::vector<GO> > meshData;        ///< layout of indices accross all processes.
+  std::vector<std::vector<GO> > coarseMeshData;  ///< layout of indices accross all processes after coarsening.
 
-    int numBlocks;               ///< Number of mesh block.
-    int myBlock;                 ///< local mesh block ID.
+  void sortLocalLexicographicData();
 
-    int myRankIndex;             ///< local process index for record in meshData after sorting.
-    Array<int> rankIndices;      ///< mapping between rank ID and reordered rank ID.
-    std::vector<std::vector<GO> > meshData;          ///< layout of indices accross all processes.
-    std::vector<std::vector<GO> > coarseMeshData;    ///< layout of indices accross all processes after coarsening.
+  void computeCoarseLocalLexicographicData();
 
-    void sortLocalLexicographicData();
+  void getGIDLocalLexicographic(const LO iGhosted, const LO jGhosted, const LO kGhosted,
+                                const Array<LO> coarseNodeFineIndices, GO& myGID, LO& myPID,
+                                LO& myLID) const;
+};
 
-    void computeCoarseLocalLexicographicData();
-
-    void getGIDLocalLexicographic(const LO iGhosted, const LO jGhosted, const LO kGhosted,
-                                  const Array<LO> coarseNodeFineIndices, GO& myGID, LO& myPID,
-                                  LO& myLID) const;
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_SHORT
-#endif // MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_DECL_HPP
+#endif  // MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_def.hpp
index d3c3d8448630..ce997c4f2a80 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/coupled/MueLu_LocalLexicographicIndexManager_def.hpp
@@ -51,448 +51,418 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  LocalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                                 const int NumDimensions, const int interpolationOrder,
-                                 const int MyRank, const int NumRanks,
-                                 const Array<GO> GFineNodesPerDir, const Array<LO> LFineNodesPerDir,
-                                 const Array<LO> CoarseRate, const Array<GO> MeshData) :
-    IndexManager(comm, coupled, false, NumDimensions, interpolationOrder, GFineNodesPerDir, LFineNodesPerDir),
-  myRank(MyRank), numRanks(NumRanks) {
-
-    // Allocate data based on user input
-    meshData.resize(numRanks);
-    rankIndices.resize(numRanks);
-    coarseMeshData.resize(numRanks);
-
-    // Load coarse rate, being careful about formating
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < this->numDimensions) {
-        if(CoarseRate.size() == 1) {
-          this->coarseRate[dim] = CoarseRate[0];
-        } else if(CoarseRate.size() == this->numDimensions) {
-          this->coarseRate[dim] = CoarseRate[dim];
-        }
-      } else {
-        this->coarseRate[dim] = 1;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    LocalLexicographicIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                                   const int NumDimensions, const int interpolationOrder,
+                                   const int MyRank, const int NumRanks,
+                                   const Array<GO> GFineNodesPerDir, const Array<LO> LFineNodesPerDir,
+                                   const Array<LO> CoarseRate, const Array<GO> MeshData)
+  : IndexManager(comm, coupled, false, NumDimensions, interpolationOrder, GFineNodesPerDir, LFineNodesPerDir)
+  , myRank(MyRank)
+  , numRanks(NumRanks) {
+  // Allocate data based on user input
+  meshData.resize(numRanks);
+  rankIndices.resize(numRanks);
+  coarseMeshData.resize(numRanks);
+
+  // Load coarse rate, being careful about formating
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < this->numDimensions) {
+      if (CoarseRate.size() == 1) {
+        this->coarseRate[dim] = CoarseRate[0];
+      } else if (CoarseRate.size() == this->numDimensions) {
+        this->coarseRate[dim] = CoarseRate[dim];
       }
+    } else {
+      this->coarseRate[dim] = 1;
     }
+  }
 
-    // Load meshData for local lexicographic case
-    for(int rank = 0; rank < numRanks; ++rank) {
-      meshData[rank].resize(10);
-      for(int entry = 0; entry < 10; ++entry) {
-        meshData[rank][entry] = MeshData[10*rank + entry];
-      }
-    }
-
-    if(this->coupled_) {
-      myBlock = meshData[myRank][2];
-      sortLocalLexicographicData();
-    }
-
-    // Start simple parameter calculation
-    myRankIndex = rankIndices[myRank];
-    for(int dim = 0; dim < 3; ++dim) {
-      this->startIndices[dim]     = meshData[myRankIndex][2*dim + 3];
-      this->startIndices[dim + 3] = meshData[myRankIndex][2*dim + 4];
+  // Load meshData for local lexicographic case
+  for (int rank = 0; rank < numRanks; ++rank) {
+    meshData[rank].resize(10);
+    for (int entry = 0; entry < 10; ++entry) {
+      meshData[rank][entry] = MeshData[10 * rank + entry];
     }
+  }
 
-    this->computeMeshParameters();
-    computeGlobalCoarseParameters();
-    computeCoarseLocalLexicographicData();
-  } // Constructor
+  if (this->coupled_) {
+    myBlock = meshData[myRank][2];
+    sortLocalLexicographicData();
+  }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGlobalCoarseParameters() {
-    this->gNumCoarseNodes10 = this->gCoarseNodesPerDir[0]*this->gCoarseNodesPerDir[1];
-    this->gNumCoarseNodes   = this->gNumCoarseNodes10*this->gCoarseNodesPerDir[2];
+  // Start simple parameter calculation
+  myRankIndex = rankIndices[myRank];
+  for (int dim = 0; dim < 3; ++dim) {
+    this->startIndices[dim]     = meshData[myRankIndex][2 * dim + 3];
+    this->startIndices[dim + 3] = meshData[myRankIndex][2 * dim + 4];
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodesData(const RCP<const Map>/* fineMap */,
-                      Array<LO>& ghostedNodeCoarseLIDs,
-                      Array<int>& ghostedNodeCoarsePIDs,
-                      Array<GO>& ghostedNodeCoarseGIDs) const {
-
-    // First we allocated memory for the outputs
-    ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
-    ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
-    ghostedNodeCoarseGIDs.resize(this->numGhostedNodes);
-
-    // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
-    // This requires finding what their GID on the fine mesh is. They need to be ordered
-    // lexicographically to allow for fast sweeps through the mesh.
-
-    // We loop over all ghosted coarse nodes by increasing global lexicographic order
-    Array<LO>  ghostedCoarseNodeCoarseIndices(3), ghostedCoarseNodeFineIndices(3);
-    Array<LO>  lCoarseNodeCoarseIndices(3);
-    Array<GO>  lCoarseNodeCoarseGIDs(this->lNumCoarseNodes);
-    LO currentIndex = -1, countCoarseNodes = 0;
-    for(int k = 0; k < this->ghostedNodesPerDir[2]; ++k) {
-      for(int j = 0; j < this->ghostedNodesPerDir[1]; ++j) {
-        for(int i = 0; i < this->ghostedNodesPerDir[0]; ++i) {
-          currentIndex = k*this->numGhostedNodes10 + j*this->ghostedNodesPerDir[0] + i;
-          ghostedCoarseNodeCoarseIndices[0] = this->startGhostedCoarseNode[0] + i;
-          ghostedCoarseNodeFineIndices[0] = ghostedCoarseNodeCoarseIndices[0]*this->coarseRate[0];
-          if(ghostedCoarseNodeFineIndices[0] > this->gFineNodesPerDir[0] - 1) {
-            ghostedCoarseNodeFineIndices[0] = this->gFineNodesPerDir[0] - 1;
-          }
-          ghostedCoarseNodeCoarseIndices[1] = this->startGhostedCoarseNode[1] + j;
-          ghostedCoarseNodeFineIndices[1] = ghostedCoarseNodeCoarseIndices[1]*this->coarseRate[1];
-          if(ghostedCoarseNodeFineIndices[1] > this->gFineNodesPerDir[1] - 1) {
-            ghostedCoarseNodeFineIndices[1] = this->gFineNodesPerDir[1] - 1;
-          }
-          ghostedCoarseNodeCoarseIndices[2] = this->startGhostedCoarseNode[2] + k;
-          ghostedCoarseNodeFineIndices[2] = ghostedCoarseNodeCoarseIndices[2]*this->coarseRate[2];
-          if(ghostedCoarseNodeFineIndices[2] > this->gFineNodesPerDir[2] - 1) {
-            ghostedCoarseNodeFineIndices[2] = this->gFineNodesPerDir[2] - 1;
-          }
+  this->computeMeshParameters();
+  computeGlobalCoarseParameters();
+  computeCoarseLocalLexicographicData();
+}  // Constructor
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGlobalCoarseParameters() {
+  this->gNumCoarseNodes10 = this->gCoarseNodesPerDir[0] * this->gCoarseNodesPerDir[1];
+  this->gNumCoarseNodes   = this->gNumCoarseNodes10 * this->gCoarseNodesPerDir[2];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodesData(const RCP<const Map> /* fineMap */,
+                        Array<LO>& ghostedNodeCoarseLIDs,
+                        Array<int>& ghostedNodeCoarsePIDs,
+                        Array<GO>& ghostedNodeCoarseGIDs) const {
+  // First we allocated memory for the outputs
+  ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
+  ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
+  ghostedNodeCoarseGIDs.resize(this->numGhostedNodes);
+
+  // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
+  // This requires finding what their GID on the fine mesh is. They need to be ordered
+  // lexicographically to allow for fast sweeps through the mesh.
+
+  // We loop over all ghosted coarse nodes by increasing global lexicographic order
+  Array<LO> ghostedCoarseNodeCoarseIndices(3), ghostedCoarseNodeFineIndices(3);
+  Array<LO> lCoarseNodeCoarseIndices(3);
+  Array<GO> lCoarseNodeCoarseGIDs(this->lNumCoarseNodes);
+  LO currentIndex = -1, countCoarseNodes = 0;
+  for (int k = 0; k < this->ghostedNodesPerDir[2]; ++k) {
+    for (int j = 0; j < this->ghostedNodesPerDir[1]; ++j) {
+      for (int i = 0; i < this->ghostedNodesPerDir[0]; ++i) {
+        currentIndex                      = k * this->numGhostedNodes10 + j * this->ghostedNodesPerDir[0] + i;
+        ghostedCoarseNodeCoarseIndices[0] = this->startGhostedCoarseNode[0] + i;
+        ghostedCoarseNodeFineIndices[0]   = ghostedCoarseNodeCoarseIndices[0] * this->coarseRate[0];
+        if (ghostedCoarseNodeFineIndices[0] > this->gFineNodesPerDir[0] - 1) {
+          ghostedCoarseNodeFineIndices[0] = this->gFineNodesPerDir[0] - 1;
+        }
+        ghostedCoarseNodeCoarseIndices[1] = this->startGhostedCoarseNode[1] + j;
+        ghostedCoarseNodeFineIndices[1]   = ghostedCoarseNodeCoarseIndices[1] * this->coarseRate[1];
+        if (ghostedCoarseNodeFineIndices[1] > this->gFineNodesPerDir[1] - 1) {
+          ghostedCoarseNodeFineIndices[1] = this->gFineNodesPerDir[1] - 1;
+        }
+        ghostedCoarseNodeCoarseIndices[2] = this->startGhostedCoarseNode[2] + k;
+        ghostedCoarseNodeFineIndices[2]   = ghostedCoarseNodeCoarseIndices[2] * this->coarseRate[2];
+        if (ghostedCoarseNodeFineIndices[2] > this->gFineNodesPerDir[2] - 1) {
+          ghostedCoarseNodeFineIndices[2] = this->gFineNodesPerDir[2] - 1;
+        }
 
-          GO myGID = -1, myCoarseGID = -1;
-          LO myLID = -1, myPID = -1, myCoarseLID = -1;
-          getGIDLocalLexicographic(i, j, k, ghostedCoarseNodeFineIndices, myGID, myPID, myLID);
+        GO myGID = -1, myCoarseGID = -1;
+        LO myLID = -1, myPID = -1, myCoarseLID = -1;
+        getGIDLocalLexicographic(i, j, k, ghostedCoarseNodeFineIndices, myGID, myPID, myLID);
 
-          int rankIndex = rankIndices[myPID];
-          for(int dim = 0; dim < 3; ++dim) {
-            if(dim < this->numDimensions) {
-              lCoarseNodeCoarseIndices[dim] = ghostedCoarseNodeCoarseIndices[dim]
-                - coarseMeshData[rankIndex][3 + 2*dim];
-            }
-          }
-          LO myRankIndexCoarseNodesInDir0 = coarseMeshData[rankIndex][4]
-            - coarseMeshData[rankIndex][3] + 1;
-          LO myRankIndexCoarseNodes10 = (coarseMeshData[rankIndex][6]
-                                         - coarseMeshData[rankIndex][5] + 1)
-            *myRankIndexCoarseNodesInDir0;
-          myCoarseLID = lCoarseNodeCoarseIndices[2]*myRankIndexCoarseNodes10
-            + lCoarseNodeCoarseIndices[1]*myRankIndexCoarseNodesInDir0
-            + lCoarseNodeCoarseIndices[0];
-          myCoarseGID = myCoarseLID + coarseMeshData[rankIndex][9];
-
-          ghostedNodeCoarseLIDs[currentIndex] = myCoarseLID;
-          ghostedNodeCoarsePIDs[currentIndex] = myPID;
-          ghostedNodeCoarseGIDs[currentIndex] = myCoarseGID;
-
-          if(myPID == myRank) {
-            lCoarseNodeCoarseGIDs[countCoarseNodes] = myCoarseGID;
-            ++countCoarseNodes;
+        int rankIndex = rankIndices[myPID];
+        for (int dim = 0; dim < 3; ++dim) {
+          if (dim < this->numDimensions) {
+            lCoarseNodeCoarseIndices[dim] = ghostedCoarseNodeCoarseIndices[dim] - coarseMeshData[rankIndex][3 + 2 * dim];
           }
         }
+        LO myRankIndexCoarseNodesInDir0 = coarseMeshData[rankIndex][4] - coarseMeshData[rankIndex][3] + 1;
+        LO myRankIndexCoarseNodes10     = (coarseMeshData[rankIndex][6] - coarseMeshData[rankIndex][5] + 1) * myRankIndexCoarseNodesInDir0;
+        myCoarseLID                     = lCoarseNodeCoarseIndices[2] * myRankIndexCoarseNodes10 + lCoarseNodeCoarseIndices[1] * myRankIndexCoarseNodesInDir0 + lCoarseNodeCoarseIndices[0];
+        myCoarseGID                     = myCoarseLID + coarseMeshData[rankIndex][9];
+
+        ghostedNodeCoarseLIDs[currentIndex] = myCoarseLID;
+        ghostedNodeCoarsePIDs[currentIndex] = myPID;
+        ghostedNodeCoarseGIDs[currentIndex] = myCoarseGID;
+
+        if (myPID == myRank) {
+          lCoarseNodeCoarseGIDs[countCoarseNodes] = myCoarseGID;
+          ++countCoarseNodes;
+        }
       }
     }
   }
-
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                     Array<GO>& coarseNodeCoarseGIDs,
-                     Array<GO>& coarseNodeFineGIDs) const {
-
-    // Allocate sufficient storage space for outputs
-    coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
-    coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
-
-    // Load all the GIDs on the fine mesh
-    ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
-
-    Array<GO> coarseStartIndices(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      coarseStartIndices[dim] = this->coarseMeshData[myRankIndex][2*dim + 3];
-    }
-
-    // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
-    LO fineLID;
-    for(LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
-      Array<LO> coarseIndices(3), fineIndices(3), gCoarseIndices(3);
-      this->getCoarseNodeLocalTuple(coarseLID,
-                                    coarseIndices[0],
-                                    coarseIndices[1],
-                                    coarseIndices[2]);
-      getCoarseNodeFineLID(coarseIndices[0],coarseIndices[1],coarseIndices[2],fineLID);
-      coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
-
-      LO myRankIndexCoarseNodesInDir0 = coarseMeshData[myRankIndex][4]
-            - coarseMeshData[myRankIndex][3] + 1;
-      LO myRankIndexCoarseNodes10 = (coarseMeshData[myRankIndex][6]
-                                  - coarseMeshData[myRankIndex][5] + 1)
-                                  *myRankIndexCoarseNodesInDir0;
-      LO myCoarseLID = coarseIndices[2]*myRankIndexCoarseNodes10
-                     + coarseIndices[1]*myRankIndexCoarseNodesInDir0
-                     + coarseIndices[0];
-      GO myCoarseGID = myCoarseLID + coarseMeshData[myRankIndex][9];
-      coarseNodeCoarseGIDs[coarseLID] = myCoarseGID;
-    }
-
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                       Array<GO>& coarseNodeCoarseGIDs,
+                       Array<GO>& coarseNodeFineGIDs) const {
+  // Allocate sufficient storage space for outputs
+  coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
+  coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
+
+  // Load all the GIDs on the fine mesh
+  ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
+
+  Array<GO> coarseStartIndices(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    coarseStartIndices[dim] = this->coarseMeshData[myRankIndex][2 * dim + 3];
   }
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGIDLocalLexicographic(const LO iGhosted, const LO jGhosted, const LO kGhosted,
-                           const Array<LO> coarseNodeFineIndices,
-                           GO& myGID, LO& myPID, LO& myLID) const {
-
-    LO ni = -1, nj = -1, li = -1, lj = -1, lk = -1;
-    LO myRankGuess = myRankIndex;
-    // We try to make a logical guess as to which PID owns the current coarse node
-    if(iGhosted == 0 && this->ghostInterface[0]) {
-      --myRankGuess;
-    } else if((iGhosted == this->ghostedNodesPerDir[0] - 1) && this->ghostInterface[1]) {
-      ++myRankGuess;
-    }
-    if(jGhosted == 0 && this->ghostInterface[2]) {
-      myRankGuess -= pi;
-    } else if((jGhosted == this->ghostedNodesPerDir[1] - 1) && this->ghostInterface[3]) {
-      myRankGuess += pi;
-    }
-    if(kGhosted == 0 && this->ghostInterface[4]) {
-      myRankGuess -= pj*pi;
-    } else if((kGhosted == this->ghostedNodesPerDir[2] - 1) && this->ghostInterface[5]) {
-      myRankGuess += pj*pi;
-    }
-    if(coarseNodeFineIndices[0] >= meshData[myRankGuess][3]
-       && coarseNodeFineIndices[0] <= meshData[myRankGuess][4]
-       && coarseNodeFineIndices[1] >= meshData[myRankGuess][5]
-       && coarseNodeFineIndices[1] <= meshData[myRankGuess][6]
-       && coarseNodeFineIndices[2] >= meshData[myRankGuess][7]
-       && coarseNodeFineIndices[2] <= meshData[myRankGuess][8]
-       && myRankGuess < numRanks - 1) {
-      myPID = meshData[myRankGuess][0];
-      ni = meshData[myRankGuess][4] - meshData[myRankGuess][3] + 1;
-      nj = meshData[myRankGuess][6] - meshData[myRankGuess][5] + 1;
-      li = coarseNodeFineIndices[0] - meshData[myRankGuess][3];
-      lj = coarseNodeFineIndices[1] - meshData[myRankGuess][5];
-      lk = coarseNodeFineIndices[2] - meshData[myRankGuess][7];
-      myLID = lk*nj*ni + lj*ni + li;
-      myGID = meshData[myRankGuess][9] + myLID;
-    } else { // The guess failed, let us use the heavy artilery: std::find_if()
-      // It could be interesting to monitor how many times this branch of the code gets
-      // used as it is far more expensive than the above one...
-      auto nodeRank = std::find_if(myBlockStart, myBlockEnd,
-                                   [coarseNodeFineIndices](const std::vector<GO>& vec){
-                                     if(coarseNodeFineIndices[0] >= vec[3]
-                                        && coarseNodeFineIndices[0] <= vec[4]
-                                        && coarseNodeFineIndices[1] >= vec[5]
-                                        && coarseNodeFineIndices[1] <= vec[6]
-                                        && coarseNodeFineIndices[2] >= vec[7]
-                                        && coarseNodeFineIndices[2] <= vec[8]) {
-                                       return true;
-                                     } else {
-                                       return false;
-                                     }
-                                   });
-      myPID = (*nodeRank)[0];
-      ni = (*nodeRank)[4] - (*nodeRank)[3] + 1;
-      nj = (*nodeRank)[6] - (*nodeRank)[5] + 1;
-      li = coarseNodeFineIndices[0] - (*nodeRank)[3];
-      lj = coarseNodeFineIndices[1] - (*nodeRank)[5];
-      lk = coarseNodeFineIndices[2] - (*nodeRank)[7];
-      myLID = lk*nj*ni + lj*ni + li;
-      myGID = (*nodeRank)[9] + myLID;
-    }
+  // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
+  LO fineLID;
+  for (LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
+    Array<LO> coarseIndices(3), fineIndices(3), gCoarseIndices(3);
+    this->getCoarseNodeLocalTuple(coarseLID,
+                                  coarseIndices[0],
+                                  coarseIndices[1],
+                                  coarseIndices[2]);
+    getCoarseNodeFineLID(coarseIndices[0], coarseIndices[1], coarseIndices[2], fineLID);
+    coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
+
+    LO myRankIndexCoarseNodesInDir0 = coarseMeshData[myRankIndex][4] - coarseMeshData[myRankIndex][3] + 1;
+    LO myRankIndexCoarseNodes10     = (coarseMeshData[myRankIndex][6] - coarseMeshData[myRankIndex][5] + 1) * myRankIndexCoarseNodesInDir0;
+    LO myCoarseLID                  = coarseIndices[2] * myRankIndexCoarseNodes10 + coarseIndices[1] * myRankIndexCoarseNodesInDir0 + coarseIndices[0];
+    GO myCoarseGID                  = myCoarseLID + coarseMeshData[myRankIndex][9];
+    coarseNodeCoarseGIDs[coarseLID] = myCoarseGID;
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  sortLocalLexicographicData() {
-
-    std::sort(meshData.begin(), meshData.end(),
-              [](const std::vector<GO>& a, const std::vector<GO>& b)->bool {
-                // The below function sorts ranks by blockID, kmin, jmin and imin
-                if(a[2] < b[2]) {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGIDLocalLexicographic(const LO iGhosted, const LO jGhosted, const LO kGhosted,
+                             const Array<LO> coarseNodeFineIndices,
+                             GO& myGID, LO& myPID, LO& myLID) const {
+  LO ni = -1, nj = -1, li = -1, lj = -1, lk = -1;
+  LO myRankGuess = myRankIndex;
+  // We try to make a logical guess as to which PID owns the current coarse node
+  if (iGhosted == 0 && this->ghostInterface[0]) {
+    --myRankGuess;
+  } else if ((iGhosted == this->ghostedNodesPerDir[0] - 1) && this->ghostInterface[1]) {
+    ++myRankGuess;
+  }
+  if (jGhosted == 0 && this->ghostInterface[2]) {
+    myRankGuess -= pi;
+  } else if ((jGhosted == this->ghostedNodesPerDir[1] - 1) && this->ghostInterface[3]) {
+    myRankGuess += pi;
+  }
+  if (kGhosted == 0 && this->ghostInterface[4]) {
+    myRankGuess -= pj * pi;
+  } else if ((kGhosted == this->ghostedNodesPerDir[2] - 1) && this->ghostInterface[5]) {
+    myRankGuess += pj * pi;
+  }
+  if (coarseNodeFineIndices[0] >= meshData[myRankGuess][3] && coarseNodeFineIndices[0] <= meshData[myRankGuess][4] && coarseNodeFineIndices[1] >= meshData[myRankGuess][5] && coarseNodeFineIndices[1] <= meshData[myRankGuess][6] && coarseNodeFineIndices[2] >= meshData[myRankGuess][7] && coarseNodeFineIndices[2] <= meshData[myRankGuess][8] && myRankGuess < numRanks - 1) {
+    myPID = meshData[myRankGuess][0];
+    ni    = meshData[myRankGuess][4] - meshData[myRankGuess][3] + 1;
+    nj    = meshData[myRankGuess][6] - meshData[myRankGuess][5] + 1;
+    li    = coarseNodeFineIndices[0] - meshData[myRankGuess][3];
+    lj    = coarseNodeFineIndices[1] - meshData[myRankGuess][5];
+    lk    = coarseNodeFineIndices[2] - meshData[myRankGuess][7];
+    myLID = lk * nj * ni + lj * ni + li;
+    myGID = meshData[myRankGuess][9] + myLID;
+  } else {  // The guess failed, let us use the heavy artilery: std::find_if()
+    // It could be interesting to monitor how many times this branch of the code gets
+    // used as it is far more expensive than the above one...
+    auto nodeRank = std::find_if(myBlockStart, myBlockEnd,
+                                 [coarseNodeFineIndices](const std::vector<GO>& vec) {
+                                   if (coarseNodeFineIndices[0] >= vec[3] && coarseNodeFineIndices[0] <= vec[4] && coarseNodeFineIndices[1] >= vec[5] && coarseNodeFineIndices[1] <= vec[6] && coarseNodeFineIndices[2] >= vec[7] && coarseNodeFineIndices[2] <= vec[8]) {
+                                     return true;
+                                   } else {
+                                     return false;
+                                   }
+                                 });
+    myPID         = (*nodeRank)[0];
+    ni            = (*nodeRank)[4] - (*nodeRank)[3] + 1;
+    nj            = (*nodeRank)[6] - (*nodeRank)[5] + 1;
+    li            = coarseNodeFineIndices[0] - (*nodeRank)[3];
+    lj            = coarseNodeFineIndices[1] - (*nodeRank)[5];
+    lk            = coarseNodeFineIndices[2] - (*nodeRank)[7];
+    myLID         = lk * nj * ni + lj * ni + li;
+    myGID         = (*nodeRank)[9] + myLID;
+  }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    sortLocalLexicographicData() {
+  std::sort(meshData.begin(), meshData.end(),
+            [](const std::vector<GO>& a, const std::vector<GO>& b) -> bool {
+              // The below function sorts ranks by blockID, kmin, jmin and imin
+              if (a[2] < b[2]) {
+                return true;
+              } else if (a[2] == b[2]) {
+                if (a[7] < b[7]) {
                   return true;
-                } else if(a[2] == b[2]) {
-                  if(a[7] < b[7]) {
+                } else if (a[7] == b[7]) {
+                  if (a[5] < b[5]) {
                     return true;
-                  } else if(a[7] == b[7]) {
-                    if(a[5] < b[5]) {
+                  } else if (a[5] == b[5]) {
+                    if (a[3] < b[3]) {
                       return true;
-                    } else if(a[5] == b[5]) {
-                      if(a[3] < b[3]) {return true;}
                     }
                   }
                 }
-                return false;
-              });
-
-    numBlocks = meshData[numRanks - 1][2] + 1;
-    // Find the range of the current block
-    myBlockStart = std::lower_bound(meshData.begin(), meshData.end(), myBlock - 1,
-                                    [] (const std::vector<GO>& vec, const GO val)->bool {
-                                      return (vec[2] < val) ? true : false;
-                                    });
-    myBlockEnd = std::upper_bound(meshData.begin(), meshData.end(), myBlock,
-                                  [] (const GO val, const std::vector<GO>& vec)->bool {
-                                    return (val < vec[2]) ? true : false;
+              }
+              return false;
+            });
+
+  numBlocks = meshData[numRanks - 1][2] + 1;
+  // Find the range of the current block
+  myBlockStart = std::lower_bound(meshData.begin(), meshData.end(), myBlock - 1,
+                                  [](const std::vector<GO>& vec, const GO val) -> bool {
+                                    return (vec[2] < val) ? true : false;
                                   });
-    // Assuming that i,j,k and ranges are split in pi, pj and pk processors
-    // we search for these numbers as they will allow us to find quickly the PID of processors
-    // owning ghost nodes.
-    auto myKEnd = std::upper_bound(myBlockStart, myBlockEnd, (*myBlockStart)[3],
-                                   [] (const GO val, const std::vector<GO>& vec)->bool {
-                                     return (val < vec[7]) ? true : false;
-                                   });
-    auto myJEnd = std::upper_bound(myBlockStart, myKEnd, (*myBlockStart)[3],
-                                   [] (const GO val, const std::vector<GO>& vec)->bool {
-                                     return (val < vec[5]) ? true : false;
-                                   });
-    pi = std::distance(myBlockStart, myJEnd);
-    pj = std::distance(myBlockStart, myKEnd) / pi;
-    pk = std::distance(myBlockStart, myBlockEnd) / (pj*pi);
-
-    // We also look for the index of the local rank in the current block.
-    const int MyRank = myRank;
-    myRankIndex = std::distance(meshData.begin(),
-                                std::find_if(myBlockStart, myBlockEnd,
-                                             [MyRank] (const std::vector<GO>& vec)->bool {
-                                               return (vec[0] == MyRank) ? true : false;
-                                             })
-                                );
-    // We also construct a mapping of rank to rankIndex in the meshData vector,
-    // this will allow us to access data quickly later on.
-    for(int rankIndex = 0; rankIndex < numRanks; ++rankIndex) {
-      rankIndices[meshData[rankIndex][0]] = rankIndex;
-    }
+  myBlockEnd   = std::upper_bound(meshData.begin(), meshData.end(), myBlock,
+                                  [](const GO val, const std::vector<GO>& vec) -> bool {
+                                  return (val < vec[2]) ? true : false;
+                                });
+  // Assuming that i,j,k and ranges are split in pi, pj and pk processors
+  // we search for these numbers as they will allow us to find quickly the PID of processors
+  // owning ghost nodes.
+  auto myKEnd = std::upper_bound(myBlockStart, myBlockEnd, (*myBlockStart)[3],
+                                 [](const GO val, const std::vector<GO>& vec) -> bool {
+                                   return (val < vec[7]) ? true : false;
+                                 });
+  auto myJEnd = std::upper_bound(myBlockStart, myKEnd, (*myBlockStart)[3],
+                                 [](const GO val, const std::vector<GO>& vec) -> bool {
+                                   return (val < vec[5]) ? true : false;
+                                 });
+  pi          = std::distance(myBlockStart, myJEnd);
+  pj          = std::distance(myBlockStart, myKEnd) / pi;
+  pk          = std::distance(myBlockStart, myBlockEnd) / (pj * pi);
+
+  // We also look for the index of the local rank in the current block.
+  const int MyRank = myRank;
+  myRankIndex      = std::distance(meshData.begin(),
+                                   std::find_if(myBlockStart, myBlockEnd,
+                                                [MyRank](const std::vector<GO>& vec) -> bool {
+                                             return (vec[0] == MyRank) ? true : false;
+                                           }));
+  // We also construct a mapping of rank to rankIndex in the meshData vector,
+  // this will allow us to access data quickly later on.
+  for (int rankIndex = 0; rankIndex < numRanks; ++rankIndex) {
+    rankIndices[meshData[rankIndex][0]] = rankIndex;
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  computeCoarseLocalLexicographicData() {
-    Array<LO> rankOffset(3);
-    for(int rank = 0; rank < numRanks; ++rank) {
-      coarseMeshData[rank].resize(10);
-      coarseMeshData[rank][0] = meshData[rank][0];
-      coarseMeshData[rank][1] = meshData[rank][1];
-      coarseMeshData[rank][2] = meshData[rank][2];
-      for(int dim = 0; dim < 3; ++dim) {
-        coarseMeshData[rank][3 + 2*dim] = meshData[rank][3 + 2*dim] / this->coarseRate[dim];
-        if(meshData[rank][3 + 2*dim] % this->coarseRate[dim] > 0) {
-          ++coarseMeshData[rank][3 + 2*dim];
-        }
-        coarseMeshData[rank][3 + 2*dim + 1] = meshData[rank][3 + 2*dim + 1] / this->coarseRate[dim];
-        if(meshData[rank][3 + 2*dim + 1] == this->gFineNodesPerDir[dim] - 1 &&
-           meshData[rank][3 + 2*dim + 1] % this->coarseRate[dim] > 0) {
-           //this->endRate[dim] < this->coarseRate[dim]) {
-          ++coarseMeshData[rank][3 + 2*dim + 1];
-        }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    computeCoarseLocalLexicographicData() {
+  Array<LO> rankOffset(3);
+  for (int rank = 0; rank < numRanks; ++rank) {
+    coarseMeshData[rank].resize(10);
+    coarseMeshData[rank][0] = meshData[rank][0];
+    coarseMeshData[rank][1] = meshData[rank][1];
+    coarseMeshData[rank][2] = meshData[rank][2];
+    for (int dim = 0; dim < 3; ++dim) {
+      coarseMeshData[rank][3 + 2 * dim] = meshData[rank][3 + 2 * dim] / this->coarseRate[dim];
+      if (meshData[rank][3 + 2 * dim] % this->coarseRate[dim] > 0) {
+        ++coarseMeshData[rank][3 + 2 * dim];
       }
-      if(rank > 0) {
-        coarseMeshData[rank][9] = coarseMeshData[rank - 1][9]
-          + (coarseMeshData[rank - 1][8] - coarseMeshData[rank - 1][7] + 1)
-          * (coarseMeshData[rank - 1][6] - coarseMeshData[rank - 1][5] + 1)
-          * (coarseMeshData[rank - 1][4] - coarseMeshData[rank - 1][3] + 1);
+      coarseMeshData[rank][3 + 2 * dim + 1] = meshData[rank][3 + 2 * dim + 1] / this->coarseRate[dim];
+      if (meshData[rank][3 + 2 * dim + 1] == this->gFineNodesPerDir[dim] - 1 &&
+          meshData[rank][3 + 2 * dim + 1] % this->coarseRate[dim] > 0) {
+        // this->endRate[dim] < this->coarseRate[dim]) {
+        ++coarseMeshData[rank][3 + 2 * dim + 1];
       }
     }
+    if (rank > 0) {
+      coarseMeshData[rank][9] = coarseMeshData[rank - 1][9] + (coarseMeshData[rank - 1][8] - coarseMeshData[rank - 1][7] + 1) * (coarseMeshData[rank - 1][6] - coarseMeshData[rank - 1][5] + 1) * (coarseMeshData[rank - 1][4] - coarseMeshData[rank - 1][3] + 1);
+    }
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<std::vector<GlobalOrdinal> > LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseMeshData() const {return coarseMeshData;}
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp   / this->lFineNodesPerDir[0];
-    i   = tmp   % this->lFineNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp   / this->lFineNodesPerDir[0];
-    i   = tmp   % this->lFineNodesPerDir[0];
-
-    k += this->offsets[2];
-    j += this->offsets[1];
-    i += this->offsets[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumCoarseNodes10;
-    tmp = myLID % this->lNumCoarseNodes10;
-    j   = tmp / this->lCoarseNodesPerDir[0];
-    i   = tmp % this->lCoarseNodesPerDir[0];
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    myLID = k*this->numGhostedNodes10 + j*this->ghostedNodesPerDir[0] + i;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    // Assumptions: (i,j,k) is a tuple on the coarse mesh
-    //              myLID is the corresponding local ID on the fine mesh
-    const LO multiplier[3] = {1, this->lFineNodesPerDir[0], this->lNumFineNodes10};
-    const LO indices[3] = {i, j, k};
-
-    myLID = 0;
-    for(int dim = 0; dim < 3; ++dim) {
-      if((indices[dim] == this->getLocalCoarseNodesInDir(dim) - 1) && this->meshEdge[2*dim + 1]) {
-        // We are dealing with the last node on the mesh in direction dim
-        // so we can simply use the number of nodes on the fine mesh in that direction
-        myLID += (this->getLocalFineNodesInDir(dim) - 1)*multiplier[dim];
-      } else {
-        myLID += (indices[dim]*this->getCoarseningRate(dim) + this->getCoarseNodeOffset(dim))
-          *multiplier[dim];
-      }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<std::vector<GlobalOrdinal> > LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseMeshData() const { return coarseMeshData; }
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+
+  k += this->offsets[2];
+  j += this->offsets[1];
+  i += this->offsets[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumCoarseNodes10;
+  tmp = myLID % this->lNumCoarseNodes10;
+  j   = tmp / this->lCoarseNodesPerDir[0];
+  i   = tmp % this->lCoarseNodesPerDir[0];
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  myLID = k * this->numGhostedNodes10 + j * this->ghostedNodesPerDir[0] + i;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  // Assumptions: (i,j,k) is a tuple on the coarse mesh
+  //              myLID is the corresponding local ID on the fine mesh
+  const LO multiplier[3] = {1, this->lFineNodesPerDir[0], this->lNumFineNodes10};
+  const LO indices[3]    = {i, j, k};
+
+  myLID = 0;
+  for (int dim = 0; dim < 3; ++dim) {
+    if ((indices[dim] == this->getLocalCoarseNodesInDir(dim) - 1) && this->meshEdge[2 * dim + 1]) {
+      // We are dealing with the last node on the mesh in direction dim
+      // so we can simply use the number of nodes on the fine mesh in that direction
+      myLID += (this->getLocalFineNodesInDir(dim) - 1) * multiplier[dim];
+    } else {
+      myLID += (indices[dim] * this->getCoarseningRate(dim) + this->getCoarseNodeOffset(dim)) * multiplier[dim];
     }
   }
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeCoarseLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeCoarseLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_decl.hpp b/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_decl.hpp
index 4a8d987b423b..fbafcf783928 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_decl.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_decl.hpp
@@ -77,76 +77,73 @@ namespace MueLu {
     correspond to nodes. While not strictly necessary, it might be convenient.
 */
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UncoupledIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class UncoupledIndexManager : public IndexManager<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_UNCOUPLEDINDEXMANAGER_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
+ public:
+  // LBV: I doubt that it makes sense to have
+  //  this particular constructor since it is
+  //  not used anywhere and parameters cannot
+  //  all accessible after construction.
+  UncoupledIndexManager() = default;
 
-    //LBV: I doubt that it makes sense to have
-    // this particular constructor since it is
-    // not used anywhere and parameters cannot
-    // all accessible after construction.
-    UncoupledIndexManager() = default;
+  UncoupledIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                        const int NumDimensions, const int interpolationOrder,
+                        const int MyRank, const int NumRanks,
+                        const Array<GO> GFineNodesPerDir,
+                        const Array<LO> LFineNodesPerDir,
+                        const Array<LO> CoarseRate,
+                        const bool singleCoarsePoint);
 
-    UncoupledIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                          const int NumDimensions, const int interpolationOrder,
-                          const int MyRank, const int NumRanks,
-                          const Array<GO> GFineNodesPerDir,
-                          const Array<LO> LFineNodesPerDir,
-                          const Array<LO> CoarseRate,
-                          const bool singleCoarsePoint);
+  virtual ~UncoupledIndexManager() {}
 
-    virtual ~UncoupledIndexManager() {}
+  void computeGlobalCoarseParameters();
 
-    void computeGlobalCoarseParameters();
+  std::vector<std::vector<GO> > getCoarseMeshData() const;
 
-    std::vector<std::vector<GO> > getCoarseMeshData() const;
+  void getGhostedNodesData(const RCP<const Map> fineMap,
+                           Array<LO>& ghostedNodeCoarseLIDs,
+                           Array<int>& ghostedNodeCoarsePIDs,
+                           Array<GO>& ghostedNodeCoarseGIDs) const;
 
-    void getGhostedNodesData(const RCP<const Map> fineMap,
-                             Array<LO>& ghostedNodeCoarseLIDs,
-                             Array<int>& ghostedNodeCoarsePIDs,
-                             Array<GO>& ghostedNodeCoarseGIDs) const;
+  void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                          Array<GO>& coarseNodeCoarseGIDs,
+                          Array<GO>& coarseNodeFineGIDs) const;
 
-    void getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                            Array<GO>& coarseNodeCoarseGIDs,
-                            Array<GO>& coarseNodeFineGIDs) const;
+  void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getFineNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getFineNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getFineNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
 
-    void getCoarseNodeGlobalTuple(const GO myGID, GO& i, GO& j, GO& k) const;
+  void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
 
-    void getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const;
+  void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
 
-    void getCoarseNodeGID(const GO i, const GO j, const GO k, GO& myGID) const;
+  void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getCoarseNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeFineLID(const LO i, const LO j, const LO k, LO& myLID) const;
+  void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
 
-    void getGhostedNodeCoarseLID(const LO i, const LO j, const LO k, LO& myLID) const;
+ private:
+  const int myRank;    ///< Local rank ID.
+  const int numRanks;  ///< Number of ranks used to decompose the problem.
+};
 
-  private:
-
-    const int myRank;                   ///< Local rank ID.
-    const int numRanks;                 ///< Number of ranks used to decompose the problem.
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UNCOUPLEDINDEXMANAGER_SHORT
-#endif // MUELU_UNCOUPLEDINDEXMANAGER_DECL_HPP
+#endif  // MUELU_UNCOUPLEDINDEXMANAGER_DECL_HPP
diff --git a/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_def.hpp b/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_def.hpp
index efd12ebd68c9..a297eb60ecf8 100644
--- a/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_def.hpp
+++ b/packages/muelu/src/Graph/StructuredAggregation/uncoupled/MueLu_UncoupledIndexManager_def.hpp
@@ -52,194 +52,188 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  UncoupledIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
-                        const int NumDimensions, const int interpolationOrder,
-                        const int MyRank, const int NumRanks,
-                        const Array<GO> GFineNodesPerDir, const Array<LO> LFineNodesPerDir,
-                        const Array<LO> CoarseRate, const bool singleCoarsePoint) :
-    IndexManager(comm, coupled, singleCoarsePoint, NumDimensions, interpolationOrder,
-                 Array<GO>(3, -1), LFineNodesPerDir),
-    myRank(MyRank), numRanks(NumRanks)
-  {
-
-    // Load coarse rate, being careful about formating
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < this->numDimensions) {
-        if(CoarseRate.size() == 1) {
-          this->coarseRate[dim] = CoarseRate[0];
-        } else if(CoarseRate.size() == this->numDimensions) {
-          this->coarseRate[dim] = CoarseRate[dim];
-        }
-      } else {
-        this->coarseRate[dim] = 1;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    UncoupledIndexManager(const RCP<const Teuchos::Comm<int> > comm, const bool coupled,
+                          const int NumDimensions, const int interpolationOrder,
+                          const int MyRank, const int NumRanks,
+                          const Array<GO> GFineNodesPerDir, const Array<LO> LFineNodesPerDir,
+                          const Array<LO> CoarseRate, const bool singleCoarsePoint)
+  : IndexManager(comm, coupled, singleCoarsePoint, NumDimensions, interpolationOrder,
+                 Array<GO>(3, -1), LFineNodesPerDir)
+  , myRank(MyRank)
+  , numRanks(NumRanks) {
+  // Load coarse rate, being careful about formating
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < this->numDimensions) {
+      if (CoarseRate.size() == 1) {
+        this->coarseRate[dim] = CoarseRate[0];
+      } else if (CoarseRate.size() == this->numDimensions) {
+        this->coarseRate[dim] = CoarseRate[dim];
       }
+    } else {
+      this->coarseRate[dim] = 1;
     }
+  }
 
-    this->computeMeshParameters();
-    this->gNumCoarseNodes10 = Teuchos::OrdinalTraits<GO>::invalid();
-    this->gNumCoarseNodes   = Teuchos::OrdinalTraits<GO>::invalid();
-  } // Constructor
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  computeGlobalCoarseParameters() {
-    GO input[1] = {as<GO>(this->lNumCoarseNodes)}, output[1] = {0};
-    Teuchos::reduceAll(*(this->comm_), Teuchos::REDUCE_SUM, 1, input, output);
-    this->gNumCoarseNodes = output[0];
-  } // computeGlobalCoarseParameters
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodesData(const RCP<const Map>/* fineMap */,
-                      Array<LO>&  ghostedNodeCoarseLIDs,
-                      Array<int>& ghostedNodeCoarsePIDs,
-                      Array<GO>&  /* ghostedNodeCoarseGIDs */) const {
-
-    // First we allocate memory for the outputs
-    ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
-    ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
-    // In the uncoupled case the data required is trivial to provide!
-    for(LO idx = 0; idx < this->getNumLocalGhostedNodes(); ++idx) {
-      ghostedNodeCoarseLIDs[idx] = idx;
-      ghostedNodeCoarsePIDs[idx] = myRank;
-    }
-  } // getGhostedNodesData
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
-                     Array<GO>& coarseNodeCoarseGIDs,
-                     Array<GO>& coarseNodeFineGIDs) const {
-
-    // Allocate sufficient amount of storage in output arrays
-    coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
-    coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
-
-    // Load all the GIDs on the fine mesh
-    ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
-
-    // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
-    LO fineLID;
-    for(LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
-      Array<LO> coarseIndices(3), fineIndices(3);
-      this->getCoarseNodeLocalTuple(coarseLID,
-                                    coarseIndices[0],
-                                    coarseIndices[1],
-                                    coarseIndices[2]);
-      for(int dim = 0; dim < 3; ++dim) {
-        if(coarseIndices[dim] == this->lCoarseNodesPerDir[dim] - 1) {
-          if(this->lCoarseNodesPerDir[dim] == 1) {
-            fineIndices[dim] = 0;
-          } else {
-            fineIndices[dim] = this->lFineNodesPerDir[dim] - 1;
-          }
+  this->computeMeshParameters();
+  this->gNumCoarseNodes10 = Teuchos::OrdinalTraits<GO>::invalid();
+  this->gNumCoarseNodes   = Teuchos::OrdinalTraits<GO>::invalid();
+}  // Constructor
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    computeGlobalCoarseParameters() {
+  GO input[1] = {as<GO>(this->lNumCoarseNodes)}, output[1] = {0};
+  Teuchos::reduceAll(*(this->comm_), Teuchos::REDUCE_SUM, 1, input, output);
+  this->gNumCoarseNodes = output[0];
+}  // computeGlobalCoarseParameters
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodesData(const RCP<const Map> /* fineMap */,
+                        Array<LO>& ghostedNodeCoarseLIDs,
+                        Array<int>& ghostedNodeCoarsePIDs,
+                        Array<GO>& /* ghostedNodeCoarseGIDs */) const {
+  // First we allocate memory for the outputs
+  ghostedNodeCoarseLIDs.resize(this->getNumLocalGhostedNodes());
+  ghostedNodeCoarsePIDs.resize(this->getNumLocalGhostedNodes());
+  // In the uncoupled case the data required is trivial to provide!
+  for (LO idx = 0; idx < this->getNumLocalGhostedNodes(); ++idx) {
+    ghostedNodeCoarseLIDs[idx] = idx;
+    ghostedNodeCoarsePIDs[idx] = myRank;
+  }
+}  // getGhostedNodesData
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodesData(const RCP<const Map> fineCoordinatesMap,
+                       Array<GO>& coarseNodeCoarseGIDs,
+                       Array<GO>& coarseNodeFineGIDs) const {
+  // Allocate sufficient amount of storage in output arrays
+  coarseNodeCoarseGIDs.resize(this->getNumLocalCoarseNodes());
+  coarseNodeFineGIDs.resize(this->getNumLocalCoarseNodes());
+
+  // Load all the GIDs on the fine mesh
+  ArrayView<const GO> fineNodeGIDs = fineCoordinatesMap->getLocalElementList();
+
+  // Extract the fine LIDs of the coarse nodes and store the corresponding GIDs
+  LO fineLID;
+  for (LO coarseLID = 0; coarseLID < this->getNumLocalCoarseNodes(); ++coarseLID) {
+    Array<LO> coarseIndices(3), fineIndices(3);
+    this->getCoarseNodeLocalTuple(coarseLID,
+                                  coarseIndices[0],
+                                  coarseIndices[1],
+                                  coarseIndices[2]);
+    for (int dim = 0; dim < 3; ++dim) {
+      if (coarseIndices[dim] == this->lCoarseNodesPerDir[dim] - 1) {
+        if (this->lCoarseNodesPerDir[dim] == 1) {
+          fineIndices[dim] = 0;
         } else {
-          fineIndices[dim] = coarseIndices[dim]*this->coarseRate[dim];
+          fineIndices[dim] = this->lFineNodesPerDir[dim] - 1;
         }
+      } else {
+        fineIndices[dim] = coarseIndices[dim] * this->coarseRate[dim];
       }
-
-      fineLID = fineIndices[2]*this->lNumFineNodes10
-        + fineIndices[1]*this->lFineNodesPerDir[0]
-        + fineIndices[0];
-      coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
-
     }
-  } // getCoarseNodesData
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<std::vector<GlobalOrdinal> > UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseMeshData() const {
-    std::vector<std::vector<GO> > coarseMeshData;
-    return coarseMeshData;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp   / this->lFineNodesPerDir[0];
-    i   = tmp   % this->lFineNodesPerDir[0];
-  } // getFineNodeLocalTuple
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumFineNodes10;
-    tmp = myLID % this->lNumFineNodes10;
-    j   = tmp   / this->lFineNodesPerDir[0];
-    i   = tmp   % this->lFineNodesPerDir[0];
-
-    k += this->offsets[2];
-    j += this->offsets[1];
-    i += this->offsets[0];
-  } // getFineNodeGhostedTuple
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
-  }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getFineNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+    fineLID                       = fineIndices[2] * this->lNumFineNodes10 + fineIndices[1] * this->lFineNodesPerDir[0] + fineIndices[0];
+    coarseNodeFineGIDs[coarseLID] = fineNodeGIDs[fineLID];
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
-    LO tmp;
-    k   = myLID / this->lNumCoarseNodes10;
-    tmp = myLID % this->lNumCoarseNodes10;
-    j   = tmp   / this->lCoarseNodesPerDir[0];
-    i   = tmp   % this->lCoarseNodesPerDir[0];
-  } // getCoarseNodeLocalTuple
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
-    myLID = k*this->numGhostedNodes10 + j*this->ghostedNodesPerDir[0] + i;
-  } // getCoarseNodeGhostedLID
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getCoarseNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
-  getGhostedNodeCoarseLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
-  }
-
-} //namespace MueLu
+}  // getCoarseNodesData
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<std::vector<GlobalOrdinal> > UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseMeshData() const {
+  std::vector<std::vector<GO> > coarseMeshData;
+  return coarseMeshData;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+}  // getFineNodeLocalTuple
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGhostedTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumFineNodes10;
+  tmp = myLID % this->lNumFineNodes10;
+  j   = tmp / this->lFineNodesPerDir[0];
+  i   = tmp % this->lFineNodesPerDir[0];
+
+  k += this->offsets[2];
+  j += this->offsets[1];
+  i += this->offsets[0];
+}  // getFineNodeGhostedTuple
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getFineNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGlobalTuple(const GO /* myGID */, GO& /* i */, GO& /* j */, GO& /* k */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLocalTuple(const LO myLID, LO& i, LO& j, LO& k) const {
+  LO tmp;
+  k   = myLID / this->lNumCoarseNodes10;
+  tmp = myLID % this->lNumCoarseNodes10;
+  j   = tmp / this->lCoarseNodesPerDir[0];
+  i   = tmp % this->lCoarseNodesPerDir[0];
+}  // getCoarseNodeLocalTuple
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGID(const GO /* i */, const GO /* j */, const GO /* k */, GO& /* myGID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeGhostedLID(const LO i, const LO j, const LO k, LO& myLID) const {
+  myLID = k * this->numGhostedNodes10 + j * this->ghostedNodesPerDir[0] + i;
+}  // getCoarseNodeGhostedLID
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getCoarseNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeFineLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>::
+    getGhostedNodeCoarseLID(const LO /* i */, const LO /* j */, const LO /* k */, LO& /* myLID */) const {
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_UNCOUPLEDINDEXMANAGER_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_decl.hpp
index aacc182dc4b3..da0151d090ea 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_decl.hpp
@@ -55,75 +55,71 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase1Algorithm class.
-    @brief Algorithm for coarsening a graph with uncoupled aggregation.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Phase 1 tries to build new aggregates which fulfill the user chosen aggregation
-    criteria (i.e. minimum and maximum size of aggregates). Especially the chosen
-    ordering for the input nodes may have some influence on the final aggregates.
-    Phase 1 is the most important aggregation routine for building new aggregates.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: ordering | Ordering of graph nodes in which the nodes are processed for aggregation. The options are natural, random and graph.
-    aggregation: max selected neighbors | Maximum number of neighbor nodes which have already been added to aggregates.
-    aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
-    aggregation: max agg size | maximum allowed number of nodes in an aggregate
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED. Nodes with other states are not touched.
-  */
-
-  template <class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class AggregationPhase1Algorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase1Algorithm class.
+  @brief Algorithm for coarsening a graph with uncoupled aggregation.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Phase 1 tries to build new aggregates which fulfill the user chosen aggregation
+  criteria (i.e. minimum and maximum size of aggregates). Especially the chosen
+  ordering for the input nodes may have some influence on the final aggregates.
+  Phase 1 is the most important aggregation routine for building new aggregates.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: ordering | Ordering of graph nodes in which the nodes are processed for aggregation. The options are natural, random and graph.
+  aggregation: max selected neighbors | Maximum number of neighbor nodes which have already been added to aggregates.
+  aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
+  aggregation: max agg size | maximum allowed number of nodes in an aggregate
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED. Nodes with other states are not touched.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase1Algorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE1ALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    AggregationPhase1Algorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~AggregationPhase1Algorithm() { }
+  //! Constructor.
+  AggregationPhase1Algorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //@}
+  //! Destructor.
+  virtual ~AggregationPhase1Algorithm() {}
 
+  //@}
 
-    //! @name Aggregation methods.
-    //@{
+  //! @name Aggregation methods.
+  //@{
 
-    /*! @brief Local aggregation. */
+  /*! @brief Local aggregation. */
 
-    void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    std::string description() const { return "Phase 1 (main)"; }
+  std::string description() const { return "Phase 1 (main)"; }
 
-  private:
-
-    /*! @brief Utility to take a list of integers and reorder them randomly (by using a local permutation).
-      @param list On input, a bunch of integers. On output, the same integers in a different order
-      that is determined randomly.
-    */
-    void RandomReorder(ArrayRCP<LO> list) const;
-
-    /*! @brief Generate a random number in the range [min, max] */
-    int RandomOrdinal(int min, int max) const;
+ private:
+  /*! @brief Utility to take a list of integers and reorder them randomly (by using a local permutation).
+    @param list On input, a bunch of integers. On output, the same integers in a different order
+    that is determined randomly.
+  */
+  void RandomReorder(ArrayRCP<LO> list) const;
 
-  };
+  /*! @brief Generate a random number in the range [min, max] */
+  int RandomOrdinal(int min, int max) const;
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_SHORT
 #endif /* MUELU_AGGREGATIONPHASE1ALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_def.hpp
index 958e5b8d0930..6367fe104516 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_def.hpp
@@ -62,189 +62,188 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-
-    std::string orderingStr     = params.get<std::string>("aggregation: ordering");
-    int maxNeighAlreadySelected = params.get<int>        ("aggregation: max selected neighbors");
-    int minNodesPerAggregate    = params.get<int>        ("aggregation: min agg size");
-    int maxNodesPerAggregate    = params.get<int>        ("aggregation: max agg size");
-    bool matchMLBehavior        = params.get<bool>("aggregation: match ML phase1");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(maxNodesPerAggregate < minNodesPerAggregate, Exceptions::RuntimeError,
-      "MueLu::UncoupledAggregationAlgorithm::BuildAggregates: minNodesPerAggregate must be smaller or equal to MaxNodePerAggregate!");
-
-    enum {
-      O_NATURAL,
-      O_RANDOM,
-      O_GRAPH
-    } ordering;
-    ordering = O_NATURAL; // initialize variable (fix CID 143665)
-    if (orderingStr == "natural") ordering = O_NATURAL;
-    if (orderingStr == "random" ) ordering = O_RANDOM;
-    if (orderingStr == "graph"  ) ordering = O_GRAPH;
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()  ->getDataNonConst(0);
-
-    LO numLocalAggregates = aggregates.GetNumAggregates();
-
-    ArrayRCP<LO> randomVector;
-    if (ordering == O_RANDOM) {
-      randomVector = arcp<LO>(numRows);
-      for (LO i = 0; i < numRows; i++)
-        randomVector[i] = i;
-      RandomReorder(randomVector);
-    }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+
+  std::string orderingStr     = params.get<std::string>("aggregation: ordering");
+  int maxNeighAlreadySelected = params.get<int>("aggregation: max selected neighbors");
+  int minNodesPerAggregate    = params.get<int>("aggregation: min agg size");
+  int maxNodesPerAggregate    = params.get<int>("aggregation: max agg size");
+  bool matchMLBehavior        = params.get<bool>("aggregation: match ML phase1");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(maxNodesPerAggregate < minNodesPerAggregate, Exceptions::RuntimeError,
+                             "MueLu::UncoupledAggregationAlgorithm::BuildAggregates: minNodesPerAggregate must be smaller or equal to MaxNodePerAggregate!");
+
+  enum {
+    O_NATURAL,
+    O_RANDOM,
+    O_GRAPH
+  } ordering;
+  ordering = O_NATURAL;  // initialize variable (fix CID 143665)
+  if (orderingStr == "natural") ordering = O_NATURAL;
+  if (orderingStr == "random") ordering = O_RANDOM;
+  if (orderingStr == "graph") ordering = O_GRAPH;
+
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
+
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+
+  ArrayRCP<LO> randomVector;
+  if (ordering == O_RANDOM) {
+    randomVector = arcp<LO>(numRows);
+    for (LO i = 0; i < numRows; i++)
+      randomVector[i] = i;
+    RandomReorder(randomVector);
+  }
 
-    int              aggIndex = -1;
-    size_t           aggSize  =  0;
-    std::vector<int> aggList(graph.getLocalMaxNumRowEntries());
-
-    std::queue<LO>   graphOrderQueue;
-
-    // Main loop over all local rows of graph(A)
-    for (LO i = 0; i < numRows; i++) {
-      // Step 1: pick the next node to aggregate
-      LO rootCandidate = 0;
-      if      (ordering == O_NATURAL) rootCandidate = i;
-      else if (ordering == O_RANDOM)  rootCandidate = randomVector[i];
-      else if (ordering == O_GRAPH) {
-
-        if (graphOrderQueue.size() == 0) {
-          // Current queue is empty for "graph" ordering, populate with one READY node
-          for (LO jnode = 0; jnode < numRows; jnode++)
-            if (aggStat[jnode] == READY) {
-              graphOrderQueue.push(jnode);
-              break;
-            }
-        }
-        if (graphOrderQueue.size() == 0) {
-          // There are no more ready nodes, end the phase
-          break;
-        }
-        rootCandidate = graphOrderQueue.front();   // take next node from graph ordering queue
-        graphOrderQueue.pop();                     // delete this node in list
+  int aggIndex   = -1;
+  size_t aggSize = 0;
+  std::vector<int> aggList(graph.getLocalMaxNumRowEntries());
+
+  std::queue<LO> graphOrderQueue;
+
+  // Main loop over all local rows of graph(A)
+  for (LO i = 0; i < numRows; i++) {
+    // Step 1: pick the next node to aggregate
+    LO rootCandidate = 0;
+    if (ordering == O_NATURAL)
+      rootCandidate = i;
+    else if (ordering == O_RANDOM)
+      rootCandidate = randomVector[i];
+    else if (ordering == O_GRAPH) {
+      if (graphOrderQueue.size() == 0) {
+        // Current queue is empty for "graph" ordering, populate with one READY node
+        for (LO jnode = 0; jnode < numRows; jnode++)
+          if (aggStat[jnode] == READY) {
+            graphOrderQueue.push(jnode);
+            break;
+          }
+      }
+      if (graphOrderQueue.size() == 0) {
+        // There are no more ready nodes, end the phase
+        break;
       }
+      rootCandidate = graphOrderQueue.front();  // take next node from graph ordering queue
+      graphOrderQueue.pop();                    // delete this node in list
+    }
 
-      if (aggStat[rootCandidate] != READY)
-        continue;
+    if (aggStat[rootCandidate] != READY)
+      continue;
 
-      // Step 2: build tentative aggregate
-      aggSize = 0;
-      aggList[aggSize++] = rootCandidate;
+    // Step 2: build tentative aggregate
+    aggSize            = 0;
+    aggList[aggSize++] = rootCandidate;
 
-      ArrayView<const LO> neighOfINode = graph.getNeighborVertices(rootCandidate);
+    ArrayView<const LO> neighOfINode = graph.getNeighborVertices(rootCandidate);
 
-      // If the number of neighbors is less than the minimum number of nodes
-      // per aggregate, we know this is not going to be a valid root, and we
-      // may skip it, but only for "natural" and "random" (for "graph" we still
-      // need to fetch the list of local neighbors to continue)
-      if ((ordering == O_NATURAL || ordering == O_RANDOM) &&
-          neighOfINode.size() < minNodesPerAggregate) {
-        continue;
-      }
+    // If the number of neighbors is less than the minimum number of nodes
+    // per aggregate, we know this is not going to be a valid root, and we
+    // may skip it, but only for "natural" and "random" (for "graph" we still
+    // need to fetch the list of local neighbors to continue)
+    if ((ordering == O_NATURAL || ordering == O_RANDOM) &&
+        neighOfINode.size() < minNodesPerAggregate) {
+      continue;
+    }
 
-      LO numAggregatedNeighbours = 0;
-
-      for (int j = 0; j < neighOfINode.size(); j++) {
-        LO neigh = neighOfINode[j];
-
-        if (neigh != rootCandidate && graph.isLocalNeighborVertex(neigh)) {
-
-          if (aggStat[neigh] == READY || aggStat[neigh] == NOTSEL) {
-            // If aggregate size does not exceed max size, add node to the
-            // tentative aggregate
-            // NOTE: We do not exit the loop over all neighbours since we have
-            // still to count all aggregated neighbour nodes for the
-            // aggregation criteria
-            // NOTE: We check here for the maximum aggregation size. If we
-            // would do it below with all the other check too big aggregates
-            // would not be accepted at all.
-            if (aggSize < as<size_t>(maxNodesPerAggregate))
-              aggList[aggSize++] = neigh;
-
-          } else if(!matchMLBehavior || aggStat[neigh] != IGNORED) {
-            // NOTE: ML checks against BOUNDARY here, but boundary nodes are flagged as IGNORED by
-            // the time we get to Phase 1, so we check IGNORED instead
-            numAggregatedNeighbours++;
-          }
+    LO numAggregatedNeighbours = 0;
+
+    for (int j = 0; j < neighOfINode.size(); j++) {
+      LO neigh = neighOfINode[j];
+
+      if (neigh != rootCandidate && graph.isLocalNeighborVertex(neigh)) {
+        if (aggStat[neigh] == READY || aggStat[neigh] == NOTSEL) {
+          // If aggregate size does not exceed max size, add node to the
+          // tentative aggregate
+          // NOTE: We do not exit the loop over all neighbours since we have
+          // still to count all aggregated neighbour nodes for the
+          // aggregation criteria
+          // NOTE: We check here for the maximum aggregation size. If we
+          // would do it below with all the other check too big aggregates
+          // would not be accepted at all.
+          if (aggSize < as<size_t>(maxNodesPerAggregate))
+            aggList[aggSize++] = neigh;
+
+        } else if (!matchMLBehavior || aggStat[neigh] != IGNORED) {
+          // NOTE: ML checks against BOUNDARY here, but boundary nodes are flagged as IGNORED by
+          // the time we get to Phase 1, so we check IGNORED instead
+          numAggregatedNeighbours++;
         }
       }
+    }
 
-      // Step 3: check if tentative aggregate is acceptable
-      if ((numAggregatedNeighbours <= maxNeighAlreadySelected) &&           // too many connections to other aggregates
-          (aggSize                 >= as<size_t>(minNodesPerAggregate))) {  // too few nodes in the tentative aggregate
-        // Accept new aggregate
-        // rootCandidate becomes the root of the newly formed aggregate
-        aggregates.SetIsRoot(rootCandidate);
-        aggIndex = numLocalAggregates++;
-
-        for (size_t k = 0; k < aggSize; k++) {
-          aggStat     [aggList[k]] = AGGREGATED;
-          vertex2AggId[aggList[k]] = aggIndex;
-          procWinner  [aggList[k]] = myRank;
-        }
+    // Step 3: check if tentative aggregate is acceptable
+    if ((numAggregatedNeighbours <= maxNeighAlreadySelected) &&  // too many connections to other aggregates
+        (aggSize >= as<size_t>(minNodesPerAggregate))) {         // too few nodes in the tentative aggregate
+      // Accept new aggregate
+      // rootCandidate becomes the root of the newly formed aggregate
+      aggregates.SetIsRoot(rootCandidate);
+      aggIndex = numLocalAggregates++;
+
+      for (size_t k = 0; k < aggSize; k++) {
+        aggStat[aggList[k]]      = AGGREGATED;
+        vertex2AggId[aggList[k]] = aggIndex;
+        procWinner[aggList[k]]   = myRank;
+      }
 
-        numNonAggregatedNodes -= aggSize;
+      numNonAggregatedNodes -= aggSize;
 
-      } else {
-        // Aggregate is not accepted
-        aggStat[rootCandidate] = NOTSEL;
+    } else {
+      // Aggregate is not accepted
+      aggStat[rootCandidate] = NOTSEL;
 
-        // Need this for the "graph" ordering below
-        // The original candidate is always aggList[0]
-        aggSize = 1;
-      }
+      // Need this for the "graph" ordering below
+      // The original candidate is always aggList[0]
+      aggSize = 1;
+    }
 
-      if (ordering == O_GRAPH) {
-        // Add candidates to the list of nodes
-        // NOTE: the code have slightly different meanings depending on context:
-        //  - if aggregate was accepted, we add neighbors of neighbors of the original candidate
-        //  - if aggregate was not accepted, we add neighbors of the original candidate
-        for (size_t k = 0; k < aggSize; k++) {
-          ArrayView<const LO> neighOfJNode = graph.getNeighborVertices(aggList[k]);
+    if (ordering == O_GRAPH) {
+      // Add candidates to the list of nodes
+      // NOTE: the code have slightly different meanings depending on context:
+      //  - if aggregate was accepted, we add neighbors of neighbors of the original candidate
+      //  - if aggregate was not accepted, we add neighbors of the original candidate
+      for (size_t k = 0; k < aggSize; k++) {
+        ArrayView<const LO> neighOfJNode = graph.getNeighborVertices(aggList[k]);
 
-          for (int j = 0; j < neighOfJNode.size(); j++) {
-            LO neigh = neighOfJNode[j];
+        for (int j = 0; j < neighOfJNode.size(); j++) {
+          LO neigh = neighOfJNode[j];
 
-            if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY)
-              graphOrderQueue.push(neigh);
-          }
+          if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY)
+            graphOrderQueue.push(neigh);
         }
       }
     }
-
-    // Reset all NOTSEL vertices to READY
-    // This simplifies other algorithms
-    for (LO i = 0; i < numRows; i++)
-      if (aggStat[i] == NOTSEL)
-        aggStat[i] = READY;
-
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates);
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::RandomReorder(ArrayRCP<LO> list) const {
-    //TODO: replace int
-    int n = list.size();
-    for(int i = 0; i < n-1; i++)
-      std::swap(list[i], list[RandomOrdinal(i,n-1)]);
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  int AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::RandomOrdinal(int min, int max) const {
-    return min + as<int>((max-min+1) * (static_cast<double>(std::rand()) / (RAND_MAX + 1.0)));
-  }
-
-} // end namespace
-
+  // Reset all NOTSEL vertices to READY
+  // This simplifies other algorithms
+  for (LO i = 0; i < numRows; i++)
+    if (aggStat[i] == NOTSEL)
+      aggStat[i] = READY;
+
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::RandomReorder(ArrayRCP<LO> list) const {
+  // TODO: replace int
+  int n = list.size();
+  for (int i = 0; i < n - 1; i++)
+    std::swap(list[i], list[RandomOrdinal(i, n - 1)]);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+int AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>::RandomOrdinal(int min, int max) const {
+  return min + as<int>((max - min + 1) * (static_cast<double>(std::rand()) / (RAND_MAX + 1.0)));
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONPHASE1ALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_decl.hpp
index 295f9d927e43..dd7e9579b8b6 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_decl.hpp
@@ -58,84 +58,81 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase1Algorithm class.
-    @brief Algorithm for coarsening a graph with uncoupled aggregation.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Phase 1 tries to build new aggregates which fulfill the user chosen aggregation
-    criteria (i.e. minimum and maximum size of aggregates). Especially the chosen
-    ordering for the input nodes may have some influence on the final aggregates.
-    Phase 1 is the most important aggregation routine for building new aggregates.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: ordering | Ordering of graph nodes in which the nodes are processed for aggregation. The options are natural, random and graph.
-    aggregation: max selected neighbors | Maximum number of neighbor nodes which have already been added to aggregates.
-    aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
-    aggregation: max agg size | maximum allowed number of nodes in an aggregate
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED. Nodes with other states are not touched.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase1Algorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase1Algorithm class.
+  @brief Algorithm for coarsening a graph with uncoupled aggregation.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Phase 1 tries to build new aggregates which fulfill the user chosen aggregation
+  criteria (i.e. minimum and maximum size of aggregates). Especially the chosen
+  ordering for the input nodes may have some influence on the final aggregates.
+  Phase 1 is the most important aggregation routine for building new aggregates.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: ordering | Ordering of graph nodes in which the nodes are processed for aggregation. The options are natural, random and graph.
+  aggregation: max selected neighbors | Maximum number of neighbor nodes which have already been added to aggregates.
+  aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
+  aggregation: max agg size | maximum allowed number of nodes in an aggregate
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED. Nodes with other states are not touched.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase1Algorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type     = typename LWGraph_kokkos::device_type;
-    using execution_space = typename LWGraph_kokkos::execution_space;
-    using memory_space    = typename LWGraph_kokkos::memory_space;
+ public:
+  using device_type     = typename LWGraph_kokkos::device_type;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase1Algorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase1Algorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase1Algorithm_kokkos() { }
+  //! Destructor.
+  virtual ~AggregationPhase1Algorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const Teuchos::ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    void BuildAggregates(const Teuchos::ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  void BuildAggregatesRandom(const LO maxAggSize,
+                             const LWGraph_kokkos& graph,
+                             Aggregates& aggregates,
+                             Kokkos::View<unsigned*, device_type>& aggStat,
+                             LO& numNonAggregatedNodes) const;
 
-    void BuildAggregatesRandom(const LO maxAggSize,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const;
+  void BuildAggregatesDeterministic(const LO maxAggSize,
+                                    const LWGraph_kokkos& graph,
+                                    Aggregates& aggregates,
+                                    Kokkos::View<unsigned*, device_type>& aggStat,
+                                    LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregatesDeterministic(const LO maxAggSize,
-                                      const LWGraph_kokkos& graph,
-                                      Aggregates& aggregates,
-                                      Kokkos::View<unsigned*, device_type>& aggStat,
-                                      LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 1 (main)"; }
+};
 
-    std::string description() const { return "Phase 1 (main)"; }
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_SHORT
-#endif // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_def.hpp
index 82e837c0ef41..d69a41b93193 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase1Algorithm_kokkos_def.hpp
@@ -66,221 +66,217 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const Teuchos::ParameterList& params,
-                  const LWGraph_kokkos& graph,
-                  Aggregates& aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const Teuchos::ParameterList& params,
+                    const LWGraph_kokkos& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
+  int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
 
-    int minNodesPerAggregate    = params.get<int>        ("aggregation: min agg size");
-    int maxNodesPerAggregate    = params.get<int>        ("aggregation: max agg size");
+  TEUCHOS_TEST_FOR_EXCEPTION(maxNodesPerAggregate < minNodesPerAggregate,
+                             Exceptions::RuntimeError,
+                             "MueLu::UncoupledAggregationAlgorithm::BuildAggregates: minNodesPerAggregate must be smaller or equal to MaxNodePerAggregate!");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(maxNodesPerAggregate < minNodesPerAggregate,
-                               Exceptions::RuntimeError,
-                               "MueLu::UncoupledAggregationAlgorithm::BuildAggregates: minNodesPerAggregate must be smaller or equal to MaxNodePerAggregate!");
-
-    // Distance-2 gives less control than serial uncoupled phase 1
-    // no custom row reordering because would require making deep copy
-    // of local matrix entries and permuting it can only enforce
-    // max aggregate size
-    {
-      if(params.get<bool>("aggregation: deterministic"))
-      {
-        Monitor m(*this, "BuildAggregatesDeterministic");
-        BuildAggregatesDeterministic(maxNodesPerAggregate, graph,
-                                     aggregates, aggStat, numNonAggregatedNodes);
-      } else {
-        Monitor m(*this, "BuildAggregatesRandom");
-        BuildAggregatesRandom(maxNodesPerAggregate, graph,
-                              aggregates, aggStat, numNonAggregatedNodes);
-      }
+  // Distance-2 gives less control than serial uncoupled phase 1
+  // no custom row reordering because would require making deep copy
+  // of local matrix entries and permuting it can only enforce
+  // max aggregate size
+  {
+    if (params.get<bool>("aggregation: deterministic")) {
+      Monitor m(*this, "BuildAggregatesDeterministic");
+      BuildAggregatesDeterministic(maxNodesPerAggregate, graph,
+                                   aggregates, aggStat, numNonAggregatedNodes);
+    } else {
+      Monitor m(*this, "BuildAggregatesRandom");
+      BuildAggregatesRandom(maxNodesPerAggregate, graph,
+                            aggregates, aggStat, numNonAggregatedNodes);
     }
   }
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregatesRandom(const LO maxAggSize,
-                        const LWGraph_kokkos& graph,
-                        Aggregates& aggregates,
-                        Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                        LO& numNonAggregatedNodes) const
-  {
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    // Extract data from aggregates
-    auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner   = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors       = aggregates.GetGraphColors();
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregatesRandom(const LO maxAggSize,
+                          const LWGraph_kokkos& graph,
+                          Aggregates& aggregates,
+                          Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                          LO& numNonAggregatedNodes) const {
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
 
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    LO numAggregatedNodes = 0;
-    LO numLocalAggregates = aggregates.GetNumAggregates();
-    Kokkos::View<LO, device_type> aggCount("aggCount");
-    Kokkos::deep_copy(aggCount, numLocalAggregates);
-    Kokkos::parallel_for("Aggregation Phase 1: initial reduction over color == 1",
-                         Kokkos::RangePolicy<LO, execution_space>(0, numRows),
-                         KOKKOS_LAMBDA (const LO nodeIdx) {
-                           if(colors(nodeIdx) == 1 && aggStat(nodeIdx) == READY) {
-                             const LO aggIdx = Kokkos::atomic_fetch_add (&aggCount(), 1);
-                             vertex2AggId(nodeIdx, 0) = aggIdx;
-                             aggStat(nodeIdx) = AGGREGATED;
-                             procWinner(nodeIdx, 0) = myRank;
-                           }
-                         });
-    // Truely we wish to compute: numAggregatedNodes = aggCount - numLocalAggregates
-    // before updating the value of numLocalAggregates.
-    // But since we also do not want to create a host mirror of aggCount we do some trickery...
-    numAggregatedNodes -= numLocalAggregates;
-    Kokkos::deep_copy(numLocalAggregates, aggCount);
-    numAggregatedNodes += numLocalAggregates;
-
-    // Compute the initial size of the aggregates.
-    // Note lbv 12-21-17: I am pretty sure that the aggregates will always be of size 1
-    //                    at this point so we could simplify the code below a lot if this
-    //                    assumption is correct...
-    Kokkos::View<LO*, device_type> aggSizesView("aggSizes", numLocalAggregates);
-    {
-      // Here there is a possibility that two vertices assigned to two different threads contribute
-      // to the same aggregate if somethings happened before phase 1?
-      auto aggSizesScatterView = Kokkos::Experimental::create_scatter_view(aggSizesView);
-      Kokkos::parallel_for("Aggregation Phase 1: compute initial aggregates size",
-			   Kokkos::RangePolicy<LO, execution_space>(0, numRows),
-                           KOKKOS_LAMBDA (const LO nodeIdx) {
-                             auto aggSizesScatterViewAccess = aggSizesScatterView.access();
-                             if(vertex2AggId(nodeIdx, 0) >= 0)
-                               aggSizesScatterViewAccess(vertex2AggId(nodeIdx, 0)) += 1;
-                           });
-      Kokkos::Experimental::contribute(aggSizesView, aggSizesScatterView);
-    }
+  // Extract data from aggregates
+  auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner   = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors       = aggregates.GetGraphColors();
 
-    LO tmpNumAggregatedNodes = 0;
-    Kokkos::parallel_reduce("Aggregation Phase 1: main parallel_reduce over aggSizes",
-			    Kokkos::RangePolicy<size_t, execution_space>(0, numRows),
-                            KOKKOS_LAMBDA (const size_t nodeIdx, LO & lNumAggregatedNodes) {
-                              if(colors(nodeIdx) != 1
-                                 && (aggStat(nodeIdx) == READY || aggStat(nodeIdx) == NOTSEL)) {
-                                // Get neighbors of vertex i and look for local, aggregated,
-                                // color 1 neighbor (valid root).
-                                auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
-                                for(LO j = 0; j < neighbors.length; ++j) {
-                                  auto nei = neighbors.colidx(j);
-                                  if(lclLWGraph.isLocalNeighborVertex(nei) && colors(nei) == 1
-                                     && aggStat(nei) == AGGREGATED) {
+  auto lclLWGraph = graph.getLocalLWGraph();
 
-                                    // This atomic guarentees that any other node trying to
-                                    // join aggregate agg has the correct size.
-                                    LO agg = vertex2AggId(nei, 0);
-                                    const LO aggSize = Kokkos::atomic_fetch_add (&aggSizesView(agg),
-                                                                                 1);
-                                    if(aggSize < maxAggSize) {
-                                      //assign vertex i to aggregate with root j
-                                      vertex2AggId(nodeIdx, 0) = agg;
-                                      procWinner(nodeIdx, 0)   = myRank;
-                                      aggStat(nodeIdx)         = AGGREGATED;
-                                      ++lNumAggregatedNodes;
-                                      break;
-                                    } else {
-                                      // Decrement back the value of aggSizesView(agg)
-                                      Kokkos::atomic_decrement(&aggSizesView(agg));
-                                    }
-                                  }
-                                }
-                              }
-                              // if(aggStat(nodeIdx) != AGGREGATED) {
-                              //   lNumNonAggregatedNodes++;
-                              if(aggStat(nodeIdx) == NOTSEL) { aggStat(nodeIdx) = READY; }
-                              // }
-                            }, tmpNumAggregatedNodes);
-    numAggregatedNodes += tmpNumAggregatedNodes;
-    numNonAggregatedNodes -= numAggregatedNodes;
+  LO numAggregatedNodes = 0;
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+  Kokkos::View<LO, device_type> aggCount("aggCount");
+  Kokkos::deep_copy(aggCount, numLocalAggregates);
+  Kokkos::parallel_for(
+      "Aggregation Phase 1: initial reduction over color == 1",
+      Kokkos::RangePolicy<LO, execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const LO nodeIdx) {
+        if (colors(nodeIdx) == 1 && aggStat(nodeIdx) == READY) {
+          const LO aggIdx          = Kokkos::atomic_fetch_add(&aggCount(), 1);
+          vertex2AggId(nodeIdx, 0) = aggIdx;
+          aggStat(nodeIdx)         = AGGREGATED;
+          procWinner(nodeIdx, 0)   = myRank;
+        }
+      });
+  // Truely we wish to compute: numAggregatedNodes = aggCount - numLocalAggregates
+  // before updating the value of numLocalAggregates.
+  // But since we also do not want to create a host mirror of aggCount we do some trickery...
+  numAggregatedNodes -= numLocalAggregates;
+  Kokkos::deep_copy(numLocalAggregates, aggCount);
+  numAggregatedNodes += numLocalAggregates;
 
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates);
+  // Compute the initial size of the aggregates.
+  // Note lbv 12-21-17: I am pretty sure that the aggregates will always be of size 1
+  //                    at this point so we could simplify the code below a lot if this
+  //                    assumption is correct...
+  Kokkos::View<LO*, device_type> aggSizesView("aggSizes", numLocalAggregates);
+  {
+    // Here there is a possibility that two vertices assigned to two different threads contribute
+    // to the same aggregate if somethings happened before phase 1?
+    auto aggSizesScatterView = Kokkos::Experimental::create_scatter_view(aggSizesView);
+    Kokkos::parallel_for(
+        "Aggregation Phase 1: compute initial aggregates size",
+        Kokkos::RangePolicy<LO, execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const LO nodeIdx) {
+          auto aggSizesScatterViewAccess = aggSizesScatterView.access();
+          if (vertex2AggId(nodeIdx, 0) >= 0)
+            aggSizesScatterViewAccess(vertex2AggId(nodeIdx, 0)) += 1;
+        });
+    Kokkos::Experimental::contribute(aggSizesView, aggSizesScatterView);
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregatesDeterministic(const LO maxAggSize,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const
-  {
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
+  LO tmpNumAggregatedNodes = 0;
+  Kokkos::parallel_reduce(
+      "Aggregation Phase 1: main parallel_reduce over aggSizes",
+      Kokkos::RangePolicy<size_t, execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const size_t nodeIdx, LO& lNumAggregatedNodes) {
+        if (colors(nodeIdx) != 1 && (aggStat(nodeIdx) == READY || aggStat(nodeIdx) == NOTSEL)) {
+          // Get neighbors of vertex i and look for local, aggregated,
+          // color 1 neighbor (valid root).
+          auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
+          for (LO j = 0; j < neighbors.length; ++j) {
+            auto nei = neighbors.colidx(j);
+            if (lclLWGraph.isLocalNeighborVertex(nei) && colors(nei) == 1 && aggStat(nei) == AGGREGATED) {
+              // This atomic guarentees that any other node trying to
+              // join aggregate agg has the correct size.
+              LO agg           = vertex2AggId(nei, 0);
+              const LO aggSize = Kokkos::atomic_fetch_add(&aggSizesView(agg),
+                                                          1);
+              if (aggSize < maxAggSize) {
+                // assign vertex i to aggregate with root j
+                vertex2AggId(nodeIdx, 0) = agg;
+                procWinner(nodeIdx, 0)   = myRank;
+                aggStat(nodeIdx)         = AGGREGATED;
+                ++lNumAggregatedNodes;
+                break;
+              } else {
+                // Decrement back the value of aggSizesView(agg)
+                Kokkos::atomic_decrement(&aggSizesView(agg));
+              }
+            }
+          }
+        }
+        // if(aggStat(nodeIdx) != AGGREGATED) {
+        //   lNumNonAggregatedNodes++;
+        if (aggStat(nodeIdx) == NOTSEL) {
+          aggStat(nodeIdx) = READY;
+        }
+        // }
+      },
+      tmpNumAggregatedNodes);
+  numAggregatedNodes += tmpNumAggregatedNodes;
+  numNonAggregatedNodes -= numAggregatedNodes;
 
-    auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner   = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors       = aggregates.GetGraphColors();
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates);
+}
 
-    auto lclLWGraph = graph.getLocalLWGraph();
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregatesDeterministic(const LO maxAggSize,
+                                 const LWGraph_kokkos& graph,
+                                 Aggregates& aggregates,
+                                 Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                                 LO& numNonAggregatedNodes) const {
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
 
-    LO numLocalAggregates = aggregates.GetNumAggregates();
-    Kokkos::View<LO, device_type> numLocalAggregatesView("Num aggregates");
-    {
-      auto h_nla = Kokkos::create_mirror_view(numLocalAggregatesView);
-      h_nla() = numLocalAggregates;
-      Kokkos::deep_copy(numLocalAggregatesView, h_nla);
-    }
+  auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner   = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors       = aggregates.GetGraphColors();
 
-    Kokkos::View<LO*, device_type> newRoots("New root LIDs", numNonAggregatedNodes);
-    Kokkos::View<LO, device_type> numNewRoots("Number of new aggregates of current color");
-    auto h_numNewRoots = Kokkos::create_mirror_view(numNewRoots);
+  auto lclLWGraph = graph.getLocalLWGraph();
 
-    //first loop build the set of new roots
-    Kokkos::parallel_for("Aggregation Phase 1: building list of new roots",
-			 Kokkos::RangePolicy<execution_space>(0, numRows),
-                         KOKKOS_LAMBDA(const LO i)
-                         {
-                           if(colors(i) == 1 && aggStat(i) == READY)
-                             {
-                               //i will become a root
-                               newRoots(Kokkos::atomic_fetch_add(&numNewRoots(), 1)) = i;
-                             }
-                         });
-    Kokkos::deep_copy(h_numNewRoots, numNewRoots);
-    //sort new roots by LID to guarantee determinism in agg IDs
-    Kokkos::sort(newRoots, 0, h_numNewRoots());
-    LO numAggregated = 0;
-    Kokkos::parallel_reduce("Aggregation Phase 1: aggregating nodes",
-			    Kokkos::RangePolicy<execution_space>(0, h_numNewRoots()),
-                            KOKKOS_LAMBDA(const LO rootIndex, LO& lnumAggregated)
-                            {
-                              LO root = newRoots(rootIndex);
-                              LO aggID = numLocalAggregatesView() + rootIndex;
-                              LO aggSize = 1;
-                              vertex2AggId(root, 0) = aggID;
-                              procWinner(root, 0) = myRank;
-                              aggStat(root) = AGGREGATED;
-                              auto neighOfRoot = lclLWGraph.getNeighborVertices(root);
-                              for(LO n = 0; n < neighOfRoot.length; n++)
-                                {
-                                  LO neigh = neighOfRoot(n);
-                                  if (lclLWGraph.isLocalNeighborVertex(neigh) && aggStat(neigh) == READY)
-                                    {
-                                      //add neigh to aggregate
-                                      vertex2AggId(neigh, 0) = aggID;
-                                      procWinner(neigh, 0) = myRank;
-                                      aggStat(neigh) = AGGREGATED;
-                                      aggSize++;
-                                      if(aggSize == maxAggSize)
-                                        {
-                                          //can't add any more nodes
-                                          break;
-                                        }
-                                    }
-                                }
-                              lnumAggregated += aggSize;
-                            }, numAggregated);
-    numNonAggregatedNodes -= numAggregated;
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates + h_numNewRoots());
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+  Kokkos::View<LO, device_type> numLocalAggregatesView("Num aggregates");
+  {
+    auto h_nla = Kokkos::create_mirror_view(numLocalAggregatesView);
+    h_nla()    = numLocalAggregates;
+    Kokkos::deep_copy(numLocalAggregatesView, h_nla);
   }
 
-} // end namespace
+  Kokkos::View<LO*, device_type> newRoots("New root LIDs", numNonAggregatedNodes);
+  Kokkos::View<LO, device_type> numNewRoots("Number of new aggregates of current color");
+  auto h_numNewRoots = Kokkos::create_mirror_view(numNewRoots);
+
+  // first loop build the set of new roots
+  Kokkos::parallel_for(
+      "Aggregation Phase 1: building list of new roots",
+      Kokkos::RangePolicy<execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const LO i) {
+        if (colors(i) == 1 && aggStat(i) == READY) {
+          // i will become a root
+          newRoots(Kokkos::atomic_fetch_add(&numNewRoots(), 1)) = i;
+        }
+      });
+  Kokkos::deep_copy(h_numNewRoots, numNewRoots);
+  // sort new roots by LID to guarantee determinism in agg IDs
+  Kokkos::sort(newRoots, 0, h_numNewRoots());
+  LO numAggregated = 0;
+  Kokkos::parallel_reduce(
+      "Aggregation Phase 1: aggregating nodes",
+      Kokkos::RangePolicy<execution_space>(0, h_numNewRoots()),
+      KOKKOS_LAMBDA(const LO rootIndex, LO& lnumAggregated) {
+        LO root               = newRoots(rootIndex);
+        LO aggID              = numLocalAggregatesView() + rootIndex;
+        LO aggSize            = 1;
+        vertex2AggId(root, 0) = aggID;
+        procWinner(root, 0)   = myRank;
+        aggStat(root)         = AGGREGATED;
+        auto neighOfRoot      = lclLWGraph.getNeighborVertices(root);
+        for (LO n = 0; n < neighOfRoot.length; n++) {
+          LO neigh = neighOfRoot(n);
+          if (lclLWGraph.isLocalNeighborVertex(neigh) && aggStat(neigh) == READY) {
+            // add neigh to aggregate
+            vertex2AggId(neigh, 0) = aggID;
+            procWinner(neigh, 0)   = myRank;
+            aggStat(neigh)         = AGGREGATED;
+            aggSize++;
+            if (aggSize == maxAggSize) {
+              // can't add any more nodes
+              break;
+            }
+          }
+        }
+        lnumAggregated += aggSize;
+      },
+      numAggregated);
+  numNonAggregatedNodes -= numAggregated;
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates + h_numNewRoots());
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_decl.hpp
index 0535904d2311..735ac8cdcb4e 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_decl.hpp
@@ -56,63 +56,60 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase2aAlgorithm class.
-    @brief Among unaggregated points, see if we can make a reasonable size aggregate out of it.
-    @ingroup Aggregation
-
-    ### Idea ###
-    Among unaggregated points, see if we can make a reasonable size
-    aggregate out of it. We do this by looking at neighbors and seeing
-    how many are unaggregated and on my processor. Loosely, base the
-    number of new aggregates created on the percentage of unaggregated nodes.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
-    aggregation: max agg size | maximum allowed number of nodes in an aggregate
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase2aAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase2aAlgorithm class.
+  @brief Among unaggregated points, see if we can make a reasonable size aggregate out of it.
+  @ingroup Aggregation
+
+  ### Idea ###
+  Among unaggregated points, see if we can make a reasonable size
+  aggregate out of it. We do this by looking at neighbors and seeing
+  how many are unaggregated and on my processor. Loosely, base the
+  number of new aggregates created on the percentage of unaggregated nodes.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
+  aggregation: max agg size | maximum allowed number of nodes in an aggregate
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase2aAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE2AALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase2aAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase2aAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase2aAlgorithm() { }
+  //! Destructor.
+  virtual ~AggregationPhase2aAlgorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 2a (secondary)"; }
+};
 
-    std::string description() const { return "Phase 2a (secondary)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_SHORT
 
-
 #endif /* MUELU_AGGREGATIONPHASE2AALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_def.hpp
index fdd92afd4f8e..b9f2b394d01d 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_AGGREGATIONPHASE2AALGORITHM_DEF_HPP_
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_DEF_HPP_
 
-
 #include <Teuchos_Comm.hpp>
 #include <Teuchos_CommHelpers.hpp>
 
@@ -61,139 +60,133 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase2aAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-
-    int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
-    int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
-    bool matchMLbehavior = params.get<bool>("aggregation: match ML phase2a");
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase2aAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+
+  int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
+  int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
+  bool matchMLbehavior     = params.get<bool>("aggregation: match ML phase2a");
+
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
+
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+
+  LO numLocalNodes      = procWinner.size();
+  LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
+
+  const double aggFactor = params.get<double>("aggregation: phase2a agg factor");
+  double factor;
+
+  if (matchMLbehavior) {
+    // Note: ML uses global counts to set the factor
+    // Passing  # of nonaggregated nodes and # of nodes via aggStat
+    GO in_data[2] = {(GO)numNonAggregatedNodes, (GO)aggStat.size()};
+    GO out_data[2];
+    Teuchos::reduceAll(*graph.GetComm(), Teuchos::REDUCE_SUM, 2, in_data, out_data);
+    GO phase_one_aggregated = out_data[1] - out_data[0];
+    factor                  = as<double>(phase_one_aggregated) / (out_data[1] + 1);
+
+    LO agg_stat_unaggregated = 0;
+    LO agg_stat_aggregated   = 0;
+    LO agg_stat_bdry         = 0;
+    for (LO i = 0; i < (LO)aggStat.size(); i++) {
+      if (aggStat[i] == AGGREGATED)
+        agg_stat_aggregated++;
+      else if (aggStat[i] == BOUNDARY)
+        agg_stat_bdry++;
+      else
+        agg_stat_unaggregated++;
+    }
 
-    ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()  ->getDataNonConst(0);
+    // NOTE: ML always uses 3 as minNodesPerAggregate
+    minNodesPerAggregate = 3;
 
-    LO numLocalAggregates = aggregates.GetNumAggregates();
+  } else {
+    // MueLu defaults to using local counts to set the factor
+    factor = as<double>(numLocalAggregated) / (numLocalNodes + 1);
+  }
 
-    LO numLocalNodes      = procWinner.size();
-    LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
+  // Now apply aggFactor
+  factor = pow(factor, aggFactor);
 
-    const double aggFactor = params.get<double>("aggregation: phase2a agg factor");
-    double       factor;
+  int aggIndex   = -1;
+  size_t aggSize = 0;
+  std::vector<int> aggList(graph.getLocalMaxNumRowEntries());
 
+  for (LO rootCandidate = 0; rootCandidate < numRows; rootCandidate++) {
+    if (aggStat[rootCandidate] != READY) {
+      continue;
+    }
 
+    LO numNeighbors = 0;
+    aggSize         = 0;
     if (matchMLbehavior) {
-      // Note: ML uses global counts to set the factor
-      // Passing  # of nonaggregated nodes and # of nodes via aggStat
-      GO in_data[2] ={(GO)numNonAggregatedNodes,(GO) aggStat.size()};
-      GO out_data[2];
-      Teuchos::reduceAll(*graph.GetComm(), Teuchos::REDUCE_SUM, 2, in_data, out_data);
-      GO phase_one_aggregated = out_data[1] - out_data[0];
-      factor = as<double>(phase_one_aggregated) / (out_data[1]+1);
-      
-      LO agg_stat_unaggregated=0;
-      LO agg_stat_aggregated=0;
-      LO agg_stat_bdry=0;
-      for (LO i=0; i<(LO)aggStat.size(); i++) {
-        if (aggStat[i] == AGGREGATED)
-          agg_stat_aggregated++;
-        else if (aggStat[i] == BOUNDARY)
-          agg_stat_bdry++;
-        else
-          agg_stat_unaggregated++;
-      }
-
-      // NOTE: ML always uses 3 as minNodesPerAggregate
-      minNodesPerAggregate=3;
-      
-    }
-    else {
-      // MueLu defaults to using local counts to set the factor
-      factor = as<double>(numLocalAggregated)/(numLocalNodes+1);  
+      aggList[aggSize++] = rootCandidate;
+      numNeighbors++;
     }
 
-    // Now apply aggFactor
-    factor = pow(factor, aggFactor);
-
-    int              aggIndex = -1;
-    size_t           aggSize  =  0;
-    std::vector<int> aggList(graph.getLocalMaxNumRowEntries());
-
-    for (LO rootCandidate = 0; rootCandidate < numRows; rootCandidate++) {
-      if (aggStat[rootCandidate] != READY) {
-        continue;
-      }
+    ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(rootCandidate);
+
+    LO num_nonaggd_neighbors = 0, num_local_neighbors = 0;
+    for (int j = 0; j < neighOfINode.size(); j++) {
+      LO neigh = neighOfINode[j];
+      if (graph.isLocalNeighborVertex(neigh))
+        num_local_neighbors++;
+
+      if (neigh != rootCandidate) {
+        if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY) {
+          // If aggregate size does not exceed max size, add node to the tentative aggregate
+          // NOTE: We do not exit the loop over all neighbours since we have still
+          //       to count all aggregated neighbour nodes for the aggregation criteria
+          // NOTE: We check here for the maximum aggregation size. If we would do it below
+          //       with all the other check too big aggregates would not be accepted at all.
+          if (aggSize < as<size_t>(maxNodesPerAggregate))
+            aggList[aggSize++] = neigh;
+          num_nonaggd_neighbors++;
+        }
 
-      LO numNeighbors = 0;
-      aggSize = 0;
-      if (matchMLbehavior) {
-        aggList[aggSize++] = rootCandidate;
         numNeighbors++;
       }
+    }
 
-      ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(rootCandidate);
-
-      LO num_nonaggd_neighbors=0, num_local_neighbors=0;
-      for (int j = 0; j < neighOfINode.size(); j++) {
-        LO neigh = neighOfINode[j];
-        if (graph.isLocalNeighborVertex(neigh))
-          num_local_neighbors++;
-
+    bool accept_aggregate;
+    if (matchMLbehavior) {
+      // ML does this calculation slightly differently than MueLu does by default, specifically it
+      // uses the *local* number of neigbors, regardless of what they are.
+      // NOTE: ML does zero compression here.  Not sure if it matters
+      // NOTE: ML uses a hardcoded value 3 instead of minNodesPerAggregate.  This has been set above
+      LO rowi_N = num_local_neighbors;
+      num_nonaggd_neighbors++;  // ML counts the node itself as a nonaggd_neighbor
+      accept_aggregate = (rowi_N > as<LO>(minNodesPerAggregate)) && (num_nonaggd_neighbors > (factor * rowi_N));
+    } else {
+      accept_aggregate = (aggSize > as<size_t>(minNodesPerAggregate)) && (aggSize > factor * numNeighbors);
+    }
 
-        if (neigh != rootCandidate) {
-          if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY) {
-            // If aggregate size does not exceed max size, add node to the tentative aggregate
-            // NOTE: We do not exit the loop over all neighbours since we have still
-            //       to count all aggregated neighbour nodes for the aggregation criteria
-            // NOTE: We check here for the maximum aggregation size. If we would do it below
-            //       with all the other check too big aggregates would not be accepted at all.
-            if (aggSize < as<size_t>(maxNodesPerAggregate))
-              aggList[aggSize++] = neigh;
-            num_nonaggd_neighbors++;
-          }
+    if (accept_aggregate) {
+      // Accept new aggregate
+      // rootCandidate becomes the root of the newly formed aggregate
+      aggregates.SetIsRoot(rootCandidate);
+      aggIndex = numLocalAggregates++;
 
-          numNeighbors++;
-        }
+      for (size_t k = 0; k < aggSize; k++) {
+        aggStat[aggList[k]]      = AGGREGATED;
+        vertex2AggId[aggList[k]] = aggIndex;
+        procWinner[aggList[k]]   = myRank;
       }
 
-      
-      bool accept_aggregate;
-      if (matchMLbehavior) {
-        // ML does this calculation slightly differently than MueLu does by default, specifically it
-        // uses the *local* number of neigbors, regardless of what they are.
-        // NOTE: ML does zero compression here.  Not sure if it matters
-        // NOTE: ML uses a hardcoded value 3 instead of minNodesPerAggregate.  This has been set above
-        LO rowi_N = num_local_neighbors;
-        num_nonaggd_neighbors++; // ML counts the node itself as a nonaggd_neighbor
-        accept_aggregate = (rowi_N > as<LO>(minNodesPerAggregate)) && (num_nonaggd_neighbors > (factor*rowi_N));
-      }
-      else {
-        accept_aggregate = (aggSize > as<size_t>(minNodesPerAggregate)) &&  (aggSize > factor*numNeighbors);
-      }
-
-      
-      if (accept_aggregate) {
-        // Accept new aggregate
-        // rootCandidate becomes the root of the newly formed aggregate
-        aggregates.SetIsRoot(rootCandidate);
-        aggIndex = numLocalAggregates++;
-
-        for (size_t k = 0; k < aggSize; k++) {
-          aggStat     [aggList[k]] = AGGREGATED;
-          vertex2AggId[aggList[k]] = aggIndex;
-          procWinner  [aggList[k]] = myRank;
-        }
-
-        numNonAggregatedNodes -= aggSize;
-      }
+      numNonAggregatedNodes -= aggSize;
     }
-
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates);
   }
 
-} // end namespace
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates);
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONPHASE2AALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_decl.hpp
index d253e891cd71..b5487d78812f 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_decl.hpp
@@ -57,81 +57,79 @@
 #include "MueLu_FactoryBase_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase2aAlgorithm class.
-    @brief Among unaggregated points, see if we can make a reasonable size aggregate out of it.
-    @ingroup Aggregation
-
-    ### Idea ###
-    Among unaggregated points, see if we can make a reasonable size
-    aggregate out of it. We do this by looking at neighbors and seeing
-    how many are unaggregated and on my processor. Loosely, base the
-    number of new aggregates created on the percentage of unaggregated nodes.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
-    aggregation: max agg size | maximum allowed number of nodes in an aggregate
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase2aAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase2aAlgorithm class.
+  @brief Among unaggregated points, see if we can make a reasonable size aggregate out of it.
+  @ingroup Aggregation
+
+  ### Idea ###
+  Among unaggregated points, see if we can make a reasonable size
+  aggregate out of it. We do this by looking at neighbors and seeing
+  how many are unaggregated and on my processor. Loosely, base the
+  number of new aggregates created on the percentage of unaggregated nodes.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: min agg size | minimum number of nodes which have to be in an aggregate.
+  aggregation: max agg size | maximum allowed number of nodes in an aggregate
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase2aAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type     = typename LWGraph_kokkos::device_type;
-    using execution_space = typename LWGraph_kokkos::execution_space;
-    using memory_space    = typename LWGraph_kokkos::memory_space;
+ public:
+  using device_type     = typename LWGraph_kokkos::device_type;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase2aAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase2aAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase2aAlgorithm_kokkos() { }
+  //! Destructor.
+  virtual ~AggregationPhase2aAlgorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const Teuchos::ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    void BuildAggregates(const Teuchos::ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  void BuildAggregatesRandom(const Teuchos::ParameterList& params,
+                             const LWGraph_kokkos& graph,
+                             Aggregates& aggregates,
+                             Kokkos::View<unsigned*, device_type>& aggStat,
+                             LO& numNonAggregatedNodes) const;
 
-    void BuildAggregatesRandom(const Teuchos::ParameterList& params,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const;
+  void BuildAggregatesDeterministic(const Teuchos::ParameterList& params,
+                                    const LWGraph_kokkos& graph,
+                                    Aggregates& aggregates,
+                                    Kokkos::View<unsigned*, device_type>& aggStat,
+                                    LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregatesDeterministic(const Teuchos::ParameterList& params,
-                                      const LWGraph_kokkos& graph,
-                                      Aggregates& aggregates,
-                                      Kokkos::View<unsigned*, device_type>& aggStat,
-                                      LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 2a (secondary)"; }
+};
 
-    std::string description() const { return "Phase 2a (secondary)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_SHORT
-#endif // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_def.hpp
index 01fcb20a21b9..d016b747c7bc 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2aAlgorithm_kokkos_def.hpp
@@ -62,268 +62,266 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase2aAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const ParameterList& params,
-                  const LWGraph_kokkos& graph,
-                  Aggregates& aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-
-    if(params.get<bool>("aggregation: deterministic")) {
-      Monitor m(*this, "BuildAggregatesDeterministic");
-      BuildAggregatesDeterministic(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    } else {
-      Monitor m(*this, "BuildAggregatesRandom");
-      BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase2aAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const ParameterList& params,
+                    const LWGraph_kokkos& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  if (params.get<bool>("aggregation: deterministic")) {
+    Monitor m(*this, "BuildAggregatesDeterministic");
+    BuildAggregatesDeterministic(params, graph, aggregates, aggStat, numNonAggregatedNodes);
+  } else {
+    Monitor m(*this, "BuildAggregatesRandom");
+    BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
+  }
 
-  } // BuildAggregates
-
-  template <class LO, class GO, class Node>
-  void AggregationPhase2aAlgorithm_kokkos<LO, GO, Node>::
-  BuildAggregatesRandom(const ParameterList& params,
-                        const LWGraph_kokkos& graph,
-                        Aggregates& aggregates,
-                        Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                        LO& numNonAggregatedNodes) const
-  {
-    const int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
-    const int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
-    bool matchMLbehavior = params.get<bool>("aggregation: match ML phase2a");
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner    = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors        = aggregates.GetGraphColors();
-    const LO numColors = aggregates.GetGraphNumColors();
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    LO numLocalNodes      = numRows;
-    LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
-
-    const double aggFactor = 0.5;
-    double       factor    = static_cast<double>(numLocalAggregated)/(numLocalNodes+1);
-    factor = pow(factor, aggFactor);
-
-    // LBV on Sept 12, 2019: this looks a little heavy handed,
-    // I'm not sure a view is needed to perform atomic updates.
-    // If we can avoid this and use a simple LO that would be
-    // simpler for later maintenance.
-    Kokkos::View<LO, device_type> numLocalAggregates("numLocalAggregates");
-    typename Kokkos::View<LO, device_type>::HostMirror h_numLocalAggregates =
+}  // BuildAggregates
+
+template <class LO, class GO, class Node>
+void AggregationPhase2aAlgorithm_kokkos<LO, GO, Node>::
+    BuildAggregatesRandom(const ParameterList& params,
+                          const LWGraph_kokkos& graph,
+                          Aggregates& aggregates,
+                          Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                          LO& numNonAggregatedNodes) const {
+  const int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
+  const int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
+  bool matchMLbehavior           = params.get<bool>("aggregation: match ML phase2a");
+
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner    = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors        = aggregates.GetGraphColors();
+  const LO numColors = aggregates.GetGraphNumColors();
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  LO numLocalNodes      = numRows;
+  LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
+
+  const double aggFactor = 0.5;
+  double factor          = static_cast<double>(numLocalAggregated) / (numLocalNodes + 1);
+  factor                 = pow(factor, aggFactor);
+
+  // LBV on Sept 12, 2019: this looks a little heavy handed,
+  // I'm not sure a view is needed to perform atomic updates.
+  // If we can avoid this and use a simple LO that would be
+  // simpler for later maintenance.
+  Kokkos::View<LO, device_type> numLocalAggregates("numLocalAggregates");
+  typename Kokkos::View<LO, device_type>::HostMirror h_numLocalAggregates =
       Kokkos::create_mirror_view(numLocalAggregates);
-    h_numLocalAggregates() = aggregates.GetNumAggregates();
-    Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
+  h_numLocalAggregates() = aggregates.GetNumAggregates();
+  Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
+
+  // Now we create new aggregates using root nodes in all colors other than the first color,
+  // as the first color was already exhausted in Phase 1.
+  for (int color = 2; color < numColors + 1; ++color) {
+    LO tmpNumNonAggregatedNodes = 0;
+    Kokkos::parallel_reduce(
+        "Aggregation Phase 2a: loop over each individual color",
+        Kokkos::RangePolicy<execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const LO rootCandidate, LO& lNumNonAggregatedNodes) {
+          if (aggStat(rootCandidate) == READY &&
+              colors(rootCandidate) == color) {
+            LO numNeighbors = 0;
+            LO aggSize      = 0;
+            if (matchMLbehavior) {
+              aggSize += 1;
+              numNeighbors += 1;
+            }
+
+            auto neighbors = lclLWGraph.getNeighborVertices(rootCandidate);
+
+            // Loop over neighbors to count how many nodes could join
+            // the new aggregate
+
+            for (int j = 0; j < neighbors.length; ++j) {
+              LO neigh = neighbors(j);
+              if (neigh != rootCandidate) {
+                if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                    (aggStat(neigh) == READY) &&
+                    (aggSize < maxNodesPerAggregate)) {
+                  ++aggSize;
+                }
+                ++numNeighbors;
+              }
+            }
+
+            // If a sufficient number of nodes can join the new aggregate
+            // then we actually create the aggregate.
+            if (aggSize > minNodesPerAggregate &&
+                (aggSize > factor * numNeighbors)) {
+              // aggregates.SetIsRoot(rootCandidate);
+              LO aggIndex = Kokkos::
+                  atomic_fetch_add(&numLocalAggregates(), 1);
+
+              LO numAggregated = 0;
+
+              if (matchMLbehavior) {
+                // Add the root.
+                aggStat(rootCandidate)         = AGGREGATED;
+                vertex2AggId(rootCandidate, 0) = aggIndex;
+                procWinner(rootCandidate, 0)   = myRank;
+                ++numAggregated;
+                --lNumNonAggregatedNodes;
+              }
+
+              for (int neighIdx = 0; neighIdx < neighbors.length; ++neighIdx) {
+                LO neigh = neighbors(neighIdx);
+                if (neigh != rootCandidate) {
+                  if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                      (aggStat(neigh) == READY) &&
+                      (numAggregated < aggSize)) {
+                    aggStat(neigh)         = AGGREGATED;
+                    vertex2AggId(neigh, 0) = aggIndex;
+                    procWinner(neigh, 0)   = myRank;
+
+                    ++numAggregated;
+                    --lNumNonAggregatedNodes;
+                  }
+                }
+              }
+            }
+          }
+        },
+        tmpNumNonAggregatedNodes);
+    numNonAggregatedNodes += tmpNumNonAggregatedNodes;
+  }
 
-    // Now we create new aggregates using root nodes in all colors other than the first color,
-    // as the first color was already exhausted in Phase 1.
-    for(int color = 2; color < numColors + 1; ++color) {
-      LO tmpNumNonAggregatedNodes = 0;
-      Kokkos::parallel_reduce("Aggregation Phase 2a: loop over each individual color",
-                              Kokkos::RangePolicy<execution_space>(0, numRows),
-                              KOKKOS_LAMBDA (const LO rootCandidate, LO& lNumNonAggregatedNodes) {
-                                if(aggStat(rootCandidate) == READY &&
-                                   colors(rootCandidate) == color) {
-
-                                  LO numNeighbors = 0;
-                                  LO aggSize = 0;
-                                  if (matchMLbehavior) {
-                                    aggSize += 1;
-                                    numNeighbors +=1;
-                                  }
-
-                                  auto neighbors = lclLWGraph.getNeighborVertices(rootCandidate);
-
-                                  // Loop over neighbors to count how many nodes could join
-                                  // the new aggregate
-
-                                  for(int j = 0; j < neighbors.length; ++j) {
-                                    LO neigh = neighbors(j);
-                                    if(neigh != rootCandidate) {
-                                      if(lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                         (aggStat(neigh) == READY) &&
-                                         (aggSize < maxNodesPerAggregate)) {
-                                        ++aggSize;
-                                      }
-                                      ++numNeighbors;
-                                    }
-                                  }
-
-                                  // If a sufficient number of nodes can join the new aggregate
-                                  // then we actually create the aggregate.
-                                  if(aggSize > minNodesPerAggregate &&
-                                     (aggSize > factor*numNeighbors)) {
-
-                                    // aggregates.SetIsRoot(rootCandidate);
-                                    LO aggIndex = Kokkos::
-                                      atomic_fetch_add(&numLocalAggregates(), 1);
-
-                                    LO numAggregated = 0;
-
-                                    if (matchMLbehavior) {
-                                      // Add the root.
-                                      aggStat(rootCandidate)         = AGGREGATED;
-                                      vertex2AggId(rootCandidate, 0) = aggIndex;
-                                      procWinner(rootCandidate, 0)   = myRank;
-                                      ++numAggregated;
-                                      --lNumNonAggregatedNodes;
-                                    }
-
-                                    for(int neighIdx = 0; neighIdx < neighbors.length; ++neighIdx) {
-                                      LO neigh = neighbors(neighIdx);
-                                      if(neigh != rootCandidate) {
-                                        if(lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                           (aggStat(neigh) == READY) &&
-                                           (numAggregated < aggSize)) {
-                                          aggStat(neigh)         = AGGREGATED;
-                                          vertex2AggId(neigh, 0) = aggIndex;
-                                          procWinner(neigh, 0)   = myRank;
-
-                                          ++numAggregated;
-                                          --lNumNonAggregatedNodes;
-                                        }
-                                      }
-                                    }
-                                  }
-                                }
-                              }, tmpNumNonAggregatedNodes);
-      numNonAggregatedNodes += tmpNumNonAggregatedNodes;
-    }
+  // update aggregate object
+  Kokkos::deep_copy(h_numLocalAggregates, numLocalAggregates);
+  aggregates.SetNumAggregates(h_numLocalAggregates());
+}  // BuildAggregatesRandom
+
+template <class LO, class GO, class Node>
+void AggregationPhase2aAlgorithm_kokkos<LO, GO, Node>::
+    BuildAggregatesDeterministic(const ParameterList& params,
+                                 const LWGraph_kokkos& graph,
+                                 Aggregates& aggregates,
+                                 Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                                 LO& numNonAggregatedNodes) const {
+  const int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
+  const int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
 
-    // update aggregate object
-    Kokkos::deep_copy(h_numLocalAggregates, numLocalAggregates);
-    aggregates.SetNumAggregates(h_numLocalAggregates());
-  } // BuildAggregatesRandom
-
-  template <class LO, class GO, class Node>
-  void AggregationPhase2aAlgorithm_kokkos<LO, GO, Node>::
-  BuildAggregatesDeterministic(const ParameterList& params,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const
-  {
-    const int minNodesPerAggregate = params.get<int>("aggregation: min agg size");
-    const int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner    = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors        = aggregates.GetGraphColors();
-    const LO numColors = aggregates.GetGraphNumColors();
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    LO numLocalNodes      = procWinner.size();
-    LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
-
-    const double aggFactor = 0.5;
-    double       factor    = as<double>(numLocalAggregated)/(numLocalNodes+1);
-    factor = pow(factor, aggFactor);
-
-    Kokkos::View<LO, device_type> numLocalAggregates("numLocalAggregates");
-    typename Kokkos::View<LO, device_type>::HostMirror h_numLocalAggregates =
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner    = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors        = aggregates.GetGraphColors();
+  const LO numColors = aggregates.GetGraphNumColors();
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  LO numLocalNodes      = procWinner.size();
+  LO numLocalAggregated = numLocalNodes - numNonAggregatedNodes;
+
+  const double aggFactor = 0.5;
+  double factor          = as<double>(numLocalAggregated) / (numLocalNodes + 1);
+  factor                 = pow(factor, aggFactor);
+
+  Kokkos::View<LO, device_type> numLocalAggregates("numLocalAggregates");
+  typename Kokkos::View<LO, device_type>::HostMirror h_numLocalAggregates =
       Kokkos::create_mirror_view(numLocalAggregates);
-    h_numLocalAggregates() = aggregates.GetNumAggregates();
-    Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
-
-    // Now we create new aggregates using root nodes in all colors other than the first color,
-    // as the first color was already exhausted in Phase 1.
-    //
-    // In the deterministic version, exactly the same set of aggregates will be created
-    // (as the nondeterministic version)
-    // because no vertex V can be a neighbor of two vertices of the same color, so two root
-    // candidates can't fight over V
-    //
-    // But, the precise values in vertex2AggId need to match exactly, so just sort the new
-    // roots of each color before assigning aggregate IDs
-
-    //numNonAggregatedNodes is the best available upper bound for the number of aggregates
-    //which may be created in this phase, so use it for the size of newRoots
-    Kokkos::View<LO*, device_type> newRoots("New root LIDs", numNonAggregatedNodes);
-    Kokkos::View<LO, device_type> numNewRoots("Number of new aggregates of current color");
-    auto h_numNewRoots = Kokkos::create_mirror_view(numNewRoots);
-    for(int color = 1; color < numColors + 1; ++color) {
-      h_numNewRoots() = 0;
-      Kokkos::deep_copy(numNewRoots, h_numNewRoots);
-      Kokkos::parallel_for("Aggregation Phase 2a: determining new roots of current color",
-                           Kokkos::RangePolicy<execution_space>(0, numRows),
-                           KOKKOS_LAMBDA(const LO rootCandidate) {
-                             if(aggStat(rootCandidate) == READY &&
-                                colors(rootCandidate) == color) {
-                               LO aggSize = 0;
-                               auto neighbors = lclLWGraph.getNeighborVertices(rootCandidate);
-                               // Loop over neighbors to count how many nodes could join
-                               // the new aggregate
-                               LO numNeighbors = 0;
-                               for(int j = 0; j < neighbors.length; ++j) {
-                                 LO neigh = neighbors(j);
-                                 if(neigh != rootCandidate)
-                                   {
-                                     if(lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                        aggStat(neigh) == READY &&
-                                        aggSize < maxNodesPerAggregate)
-                                       {
-                                         ++aggSize;
-                                       }
-                                     ++numNeighbors;
-                                   }
-                               }
-                               // If a sufficient number of nodes can join the new aggregate
-                               // then we mark rootCandidate as a future root.
-                               if(aggSize > minNodesPerAggregate && aggSize > factor*numNeighbors) {
-                                 LO newRootIndex = Kokkos::atomic_fetch_add(&numNewRoots(), 1);
-                                 newRoots(newRootIndex) = rootCandidate;
-                               }
-                             }
-                           });
-      Kokkos::deep_copy(h_numNewRoots, numNewRoots);
-
-      if(h_numNewRoots() > 0) {
-        //sort the new root indices
-        Kokkos::sort(newRoots, 0, h_numNewRoots());
-        //now, loop over all new roots again and actually create the aggregates
-        LO tmpNumNonAggregatedNodes = 0;
-        //First, just find the set of color vertices which will become aggregate roots
-        Kokkos::parallel_reduce("Aggregation Phase 2a: create new aggregates",
-                                Kokkos::RangePolicy<execution_space>(0, h_numNewRoots()),
-                                KOKKOS_LAMBDA (const LO newRootIndex, LO& lNumNonAggregatedNodes) {
-                                  LO root = newRoots(newRootIndex);
-                                  LO newAggID = numLocalAggregates() + newRootIndex;
-                                  auto neighbors = lclLWGraph.getNeighborVertices(root);
-                                  // Loop over neighbors and add them to new aggregate
-                                  aggStat(root)      = AGGREGATED;
-                                  vertex2AggId(root, 0) = newAggID;
-                                  LO aggSize = 1;
-                                  for(int j = 0; j < neighbors.length; ++j) {
-                                    LO neigh = neighbors(j);
-                                    if(neigh != root) {
-                                      if(lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                         aggStat(neigh) == READY &&
-                                         aggSize < maxNodesPerAggregate) {
-                                        aggStat(neigh)      = AGGREGATED;
-                                        vertex2AggId(neigh, 0) = newAggID;
-                                        procWinner(neigh, 0)   = myRank;
-                                        aggSize++;
-                                      }
-                                    }
-                                  }
-                                  lNumNonAggregatedNodes -= aggSize;
-                                }, tmpNumNonAggregatedNodes);
-        numNonAggregatedNodes += tmpNumNonAggregatedNodes;
-        h_numLocalAggregates() += h_numNewRoots();
-        Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
-      }
+  h_numLocalAggregates() = aggregates.GetNumAggregates();
+  Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
+
+  // Now we create new aggregates using root nodes in all colors other than the first color,
+  // as the first color was already exhausted in Phase 1.
+  //
+  // In the deterministic version, exactly the same set of aggregates will be created
+  // (as the nondeterministic version)
+  // because no vertex V can be a neighbor of two vertices of the same color, so two root
+  // candidates can't fight over V
+  //
+  // But, the precise values in vertex2AggId need to match exactly, so just sort the new
+  // roots of each color before assigning aggregate IDs
+
+  // numNonAggregatedNodes is the best available upper bound for the number of aggregates
+  // which may be created in this phase, so use it for the size of newRoots
+  Kokkos::View<LO*, device_type> newRoots("New root LIDs", numNonAggregatedNodes);
+  Kokkos::View<LO, device_type> numNewRoots("Number of new aggregates of current color");
+  auto h_numNewRoots = Kokkos::create_mirror_view(numNewRoots);
+  for (int color = 1; color < numColors + 1; ++color) {
+    h_numNewRoots() = 0;
+    Kokkos::deep_copy(numNewRoots, h_numNewRoots);
+    Kokkos::parallel_for(
+        "Aggregation Phase 2a: determining new roots of current color",
+        Kokkos::RangePolicy<execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const LO rootCandidate) {
+          if (aggStat(rootCandidate) == READY &&
+              colors(rootCandidate) == color) {
+            LO aggSize     = 0;
+            auto neighbors = lclLWGraph.getNeighborVertices(rootCandidate);
+            // Loop over neighbors to count how many nodes could join
+            // the new aggregate
+            LO numNeighbors = 0;
+            for (int j = 0; j < neighbors.length; ++j) {
+              LO neigh = neighbors(j);
+              if (neigh != rootCandidate) {
+                if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                    aggStat(neigh) == READY &&
+                    aggSize < maxNodesPerAggregate) {
+                  ++aggSize;
+                }
+                ++numNeighbors;
+              }
+            }
+            // If a sufficient number of nodes can join the new aggregate
+            // then we mark rootCandidate as a future root.
+            if (aggSize > minNodesPerAggregate && aggSize > factor * numNeighbors) {
+              LO newRootIndex        = Kokkos::atomic_fetch_add(&numNewRoots(), 1);
+              newRoots(newRootIndex) = rootCandidate;
+            }
+          }
+        });
+    Kokkos::deep_copy(h_numNewRoots, numNewRoots);
+
+    if (h_numNewRoots() > 0) {
+      // sort the new root indices
+      Kokkos::sort(newRoots, 0, h_numNewRoots());
+      // now, loop over all new roots again and actually create the aggregates
+      LO tmpNumNonAggregatedNodes = 0;
+      // First, just find the set of color vertices which will become aggregate roots
+      Kokkos::parallel_reduce(
+          "Aggregation Phase 2a: create new aggregates",
+          Kokkos::RangePolicy<execution_space>(0, h_numNewRoots()),
+          KOKKOS_LAMBDA(const LO newRootIndex, LO& lNumNonAggregatedNodes) {
+            LO root        = newRoots(newRootIndex);
+            LO newAggID    = numLocalAggregates() + newRootIndex;
+            auto neighbors = lclLWGraph.getNeighborVertices(root);
+            // Loop over neighbors and add them to new aggregate
+            aggStat(root)         = AGGREGATED;
+            vertex2AggId(root, 0) = newAggID;
+            LO aggSize            = 1;
+            for (int j = 0; j < neighbors.length; ++j) {
+              LO neigh = neighbors(j);
+              if (neigh != root) {
+                if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                    aggStat(neigh) == READY &&
+                    aggSize < maxNodesPerAggregate) {
+                  aggStat(neigh)         = AGGREGATED;
+                  vertex2AggId(neigh, 0) = newAggID;
+                  procWinner(neigh, 0)   = myRank;
+                  aggSize++;
+                }
+              }
+            }
+            lNumNonAggregatedNodes -= aggSize;
+          },
+          tmpNumNonAggregatedNodes);
+      numNonAggregatedNodes += tmpNumNonAggregatedNodes;
+      h_numLocalAggregates() += h_numNewRoots();
+      Kokkos::deep_copy(numLocalAggregates, h_numLocalAggregates);
     }
-    aggregates.SetNumAggregates(h_numLocalAggregates());
   }
+  aggregates.SetNumAggregates(h_numLocalAggregates());
+}
 
-} // end namespace
+}  // namespace MueLu
 
-#endif // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_decl.hpp
index 88b156d23c14..a38f0fe0f310 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_decl.hpp
@@ -56,62 +56,59 @@
 #include "MueLu_AggregationPhase2bAlgorithm_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase2bAlgorithm class.
-    @brief Add leftovers to existing aggregates
-    @ingroup Aggregation
-
-    ### Idea ###
-    In phase 2b non-aggregated nodes are added to existing aggregates.
-    All neighbors of the unaggregated node are checked and the corresponding
-    aggregate weight is increased. The unaggregated node is added to the aggregate
-    with the best weight. A simple penalty strategy makes sure that the non-aggregated
-    nodes are added to different aggregates.
-    The routine runs twice to cover non-aggregate nodes which have a node distance
-    of two to existing aggregates. Assuming that the node distance is not greater
-    than 3 (the aggregate diameter size), running the algorithm only twice should
-    be sufficient.
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED. There are no aggregation criteria considered. Especially the aggregation: max agg size criterion is ignored.
-    This is not a problem, since after the previous aggregation phases one should not be able to build too large aggregates.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase2bAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase2bAlgorithm class.
+  @brief Add leftovers to existing aggregates
+  @ingroup Aggregation
+
+  ### Idea ###
+  In phase 2b non-aggregated nodes are added to existing aggregates.
+  All neighbors of the unaggregated node are checked and the corresponding
+  aggregate weight is increased. The unaggregated node is added to the aggregate
+  with the best weight. A simple penalty strategy makes sure that the non-aggregated
+  nodes are added to different aggregates.
+  The routine runs twice to cover non-aggregate nodes which have a node distance
+  of two to existing aggregates. Assuming that the node distance is not greater
+  than 3 (the aggregate diameter size), running the algorithm only twice should
+  be sufficient.
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED. There are no aggregation criteria considered. Especially the aggregation: max agg size criterion is ignored.
+  This is not a problem, since after the previous aggregation phases one should not be able to build too large aggregates.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase2bAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE2BALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase2bAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase2bAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase2bAlgorithm() { }
+  //! Destructor.
+  virtual ~AggregationPhase2bAlgorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 2b (expansion)"; }
+};
 
-    std::string description() const { return "Phase 2b (expansion)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_SHORT
 
-
 #endif /* MUELU_AGGREGATIONPHASE2BALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_def.hpp
index 1dc4ac52c9f1..4e402903e6de 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_def.hpp
@@ -60,91 +60,89 @@
 
 namespace MueLu {
 
-  // Try to stick unaggregated nodes into a neighboring aggregate if they are
-  // not already too big
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase2bAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params , const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-    bool matchMLbehavior = params.get<bool>("aggregation: match ML phase2b");
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()  ->getDataNonConst(0);
-
-    LO numLocalAggregates = aggregates.GetNumAggregates();
-
-    const int defaultConnectWeight = 100;
-    const int penaltyConnectWeight = 10;
-
-    std::vector<int> aggWeight    (numLocalAggregates, 0);
-    std::vector<int> connectWeight(numRows, defaultConnectWeight);
-    std::vector<int> aggPenalties (numRows, 0);
-
-    // We do this cycle twice.
-    // I don't know why, but ML does it too
-    // taw: by running the aggregation routine more than once there is a chance that also
-    // non-aggregated nodes with a node distance of two are added to existing aggregates.
-    // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
-    // should be sufficient.
-    for (int k = 0; k < 2; k++) {
-      for (LO i = 0; i < numRows; i++) {
-        if (aggStat[i] != READY)
-          continue;
-
-        ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(i);
-
-        for (int j = 0; j < neighOfINode.size(); j++) {
-          LO neigh = neighOfINode[j];
-
-          // We don't check (neigh != i), as it is covered by checking (aggStat[neigh] == AGGREGATED)
-          if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED)
-            aggWeight[vertex2AggId[neigh]] += connectWeight[neigh];
-        }
-
-        int bestScore   = -100000;
-        int bestAggId   = -1;
-        int bestConnect = -1;
-
-        for (int j = 0; j < neighOfINode.size(); j++) {
-          LO neigh = neighOfINode[j];
-          int aggId = vertex2AggId[neigh];
+// Try to stick unaggregated nodes into a neighboring aggregate if they are
+// not already too big
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase2bAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+  bool matchMLbehavior = params.get<bool>("aggregation: match ML phase2b");
+
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
+
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+
+  const int defaultConnectWeight = 100;
+  const int penaltyConnectWeight = 10;
+
+  std::vector<int> aggWeight(numLocalAggregates, 0);
+  std::vector<int> connectWeight(numRows, defaultConnectWeight);
+  std::vector<int> aggPenalties(numRows, 0);
+
+  // We do this cycle twice.
+  // I don't know why, but ML does it too
+  // taw: by running the aggregation routine more than once there is a chance that also
+  // non-aggregated nodes with a node distance of two are added to existing aggregates.
+  // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
+  // should be sufficient.
+  for (int k = 0; k < 2; k++) {
+    for (LO i = 0; i < numRows; i++) {
+      if (aggStat[i] != READY)
+        continue;
+
+      ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(i);
+
+      for (int j = 0; j < neighOfINode.size(); j++) {
+        LO neigh = neighOfINode[j];
+
+        // We don't check (neigh != i), as it is covered by checking (aggStat[neigh] == AGGREGATED)
+        if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED)
+          aggWeight[vertex2AggId[neigh]] += connectWeight[neigh];
+      }
 
-          // Note: The third condition is only relevant if the ML matching is enabled
-          if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED 
-            && (!matchMLbehavior || aggWeight[aggId] != 0) ) {
+      int bestScore   = -100000;
+      int bestAggId   = -1;
+      int bestConnect = -1;
 
-            int score = aggWeight[aggId] - aggPenalties[aggId];
+      for (int j = 0; j < neighOfINode.size(); j++) {
+        LO neigh  = neighOfINode[j];
+        int aggId = vertex2AggId[neigh];
 
-            if (score > bestScore) {
-              bestAggId   = aggId;
-              bestScore   = score;
-              bestConnect = connectWeight[neigh];
+        // Note: The third condition is only relevant if the ML matching is enabled
+        if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED && (!matchMLbehavior || aggWeight[aggId] != 0)) {
+          int score = aggWeight[aggId] - aggPenalties[aggId];
 
-            } else if (aggId == bestAggId && connectWeight[neigh] > bestConnect) {
-              bestConnect = connectWeight[neigh];
-            }
+          if (score > bestScore) {
+            bestAggId   = aggId;
+            bestScore   = score;
+            bestConnect = connectWeight[neigh];
 
-            // Reset the weights for the next loop
-            aggWeight[aggId] = 0;
+          } else if (aggId == bestAggId && connectWeight[neigh] > bestConnect) {
+            bestConnect = connectWeight[neigh];
           }
+
+          // Reset the weights for the next loop
+          aggWeight[aggId] = 0;
         }
+      }
 
-        if (bestScore >= 0) {
-          aggStat     [i] = AGGREGATED;
-          vertex2AggId[i] = bestAggId;
-          procWinner  [i] = myRank;
+      if (bestScore >= 0) {
+        aggStat[i]      = AGGREGATED;
+        vertex2AggId[i] = bestAggId;
+        procWinner[i]   = myRank;
 
-          numNonAggregatedNodes--;
+        numNonAggregatedNodes--;
 
-          aggPenalties[bestAggId]++;
-          connectWeight[i] = bestConnect - penaltyConnectWeight;
-        }
+        aggPenalties[bestAggId]++;
+        connectWeight[i] = bestConnect - penaltyConnectWeight;
       }
     }
   }
+}
 
-} // end namespace
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONPHASE2BALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_decl.hpp
index 941f732e64d0..9172487c9ca1 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_decl.hpp
@@ -58,80 +58,78 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase2bAlgorithm class.
-    @brief Add leftovers to existing aggregates
-    @ingroup Aggregation
-
-    ### Idea ###
-    In phase 2b non-aggregated nodes are added to existing aggregates.
-    All neighbors of the unaggregated node are checked and the corresponding
-    aggregate weight is increased. The unaggregated node is added to the aggregate
-    with the best weight. A simple penalty strategy makes sure that the non-aggregated
-    nodes are added to different aggregates.
-    The routine runs twice to cover non-aggregate nodes which have a node distance
-    of two to existing aggregates. Assuming that the node distance is not greater
-    than 3 (the aggregate diameter size), running the algorithm only twice should
-    be sufficient.
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED. There are no aggregation criteria considered. Especially the aggregation: max agg size criterion is ignored.
-    This is not a problem, since after the previous aggregation phases one should not be able to build too large aggregates.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase2bAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase2bAlgorithm class.
+  @brief Add leftovers to existing aggregates
+  @ingroup Aggregation
+
+  ### Idea ###
+  In phase 2b non-aggregated nodes are added to existing aggregates.
+  All neighbors of the unaggregated node are checked and the corresponding
+  aggregate weight is increased. The unaggregated node is added to the aggregate
+  with the best weight. A simple penalty strategy makes sure that the non-aggregated
+  nodes are added to different aggregates.
+  The routine runs twice to cover non-aggregate nodes which have a node distance
+  of two to existing aggregates. Assuming that the node distance is not greater
+  than 3 (the aggregate diameter size), running the algorithm only twice should
+  be sufficient.
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED. There are no aggregation criteria considered. Especially the aggregation: max agg size criterion is ignored.
+  This is not a problem, since after the previous aggregation phases one should not be able to build too large aggregates.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase2bAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type     = typename LWGraph_kokkos::device_type;
-    using execution_space = typename LWGraph_kokkos::execution_space;
-    using memory_space    = typename LWGraph_kokkos::memory_space;
+ public:
+  using device_type     = typename LWGraph_kokkos::device_type;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase2bAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase2bAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase2bAlgorithm_kokkos() { }
+  //! Destructor.
+  virtual ~AggregationPhase2bAlgorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    void BuildAggregates(const ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  void BuildAggregatesRandom(const ParameterList& params,
+                             const LWGraph_kokkos& graph,
+                             Aggregates& aggregates,
+                             Kokkos::View<unsigned*, device_type>& aggStat,
+                             LO& numNonAggregatedNodes) const;
 
-    void BuildAggregatesRandom(const ParameterList& params,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const;
+  void BuildAggregatesDeterministic(const ParameterList& params,
+                                    const LWGraph_kokkos& graph,
+                                    Aggregates& aggregates,
+                                    Kokkos::View<unsigned*, device_type>& aggStat,
+                                    LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregatesDeterministic(const ParameterList& params,
-                                      const LWGraph_kokkos& graph,
-                                      Aggregates& aggregates,
-                                      Kokkos::View<unsigned*, device_type>& aggStat,
-                                      LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 2b (expansion)"; }
+};
 
-    std::string description() const { return "Phase 2b (expansion)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_SHORT
-#endif // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_def.hpp
index bccdbf7f98ca..6168904ee0ea 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase2bAlgorithm_kokkos_def.hpp
@@ -60,179 +60,180 @@
 
 namespace MueLu {
 
-  // Try to stick unaggregated nodes into a neighboring aggregate if they are
-  // not already too big
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase2bAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const ParameterList& params,
-                  const LWGraph_kokkos& graph,
-                  Aggregates& aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-
-    if(params.get<bool>("aggregation: deterministic")) {
-      Monitor m(*this, "BuildAggregatesDeterministic");
-      BuildAggregatesDeterministic(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    } else {
-      Monitor m(*this, "BuildAggregatesRandom");
-      BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    }
+// Try to stick unaggregated nodes into a neighboring aggregate if they are
+// not already too big
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase2bAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const ParameterList& params,
+                    const LWGraph_kokkos& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  if (params.get<bool>("aggregation: deterministic")) {
+    Monitor m(*this, "BuildAggregatesDeterministic");
+    BuildAggregatesDeterministic(params, graph, aggregates, aggStat, numNonAggregatedNodes);
+  } else {
+    Monitor m(*this, "BuildAggregatesRandom");
+    BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
+  }
+
+}  // BuildAggregates
+
+template <class LO, class GO, class Node>
+void AggregationPhase2bAlgorithm_kokkos<LO, GO, Node>::
+    BuildAggregatesRandom(const ParameterList& params,
+                          const LWGraph_kokkos& graph,
+                          Aggregates& aggregates,
+                          Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                          LO& numNonAggregatedNodes) const {
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  auto vertex2AggId           = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner             = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors                 = aggregates.GetGraphColors();
+  const LO numColors          = aggregates.GetGraphNumColors();
+  const LO numLocalAggregates = aggregates.GetNumAggregates();
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  const LO defaultConnectWeight = 100;
+  const LO penaltyConnectWeight = 10;
+
+  Kokkos::View<LO*, device_type> aggWeight(Kokkos::ViewAllocateWithoutInitializing("aggWeight"), numLocalAggregates);  // This gets re-initialized at the start of each "color" loop
+  Kokkos::View<LO*, device_type> connectWeight(Kokkos::ViewAllocateWithoutInitializing("connectWeight"), numRows);
+  Kokkos::View<LO*, device_type> aggPenalties("aggPenalties", numLocalAggregates);  // This gets initialized to zero here
+
+  Kokkos::deep_copy(connectWeight, defaultConnectWeight);
+
+  // taw: by running the aggregation routine more than once there is a chance that also
+  // non-aggregated nodes with a node distance of two are added to existing aggregates.
+  // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
+  // should be sufficient.
+  // lbv: If the prior phase of aggregation where run without specifying an aggregate size,
+  // the distance 2 coloring and phase 1 aggregation actually guarantee that only one iteration
+  // is needed to reach distance 2 neighbors.
+  int maxIters             = 2;
+  int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
+  if (maxNodesPerAggregate == std::numeric_limits<int>::max()) {
+    maxIters = 1;
+  }
+  for (int iter = 0; iter < maxIters; ++iter) {
+    for (LO color = 1; color <= numColors; ++color) {
+      Kokkos::deep_copy(aggWeight, 0);
+
+      // the reduce counts how many nodes are aggregated by this phase,
+      // which will then be subtracted from numNonAggregatedNodes
+      LO numAggregated = 0;
+      Kokkos::parallel_reduce(
+          "Aggregation Phase 2b: aggregates expansion",
+          Kokkos::RangePolicy<execution_space>(0, numRows),
+          KOKKOS_LAMBDA(const LO i, LO& tmpNumAggregated) {
+            if (aggStat(i) != READY || colors(i) != color)
+              return;
+
+            auto neighOfINode = lclLWGraph.getNeighborVertices(i);
+            for (int j = 0; j < neighOfINode.length; j++) {
+              LO neigh = neighOfINode(j);
 
-  } // BuildAggregates
-
-  template <class LO, class GO, class Node>
-  void AggregationPhase2bAlgorithm_kokkos<LO, GO, Node>::
-  BuildAggregatesRandom(const ParameterList& params,
-                        const LWGraph_kokkos& graph,
-                        Aggregates& aggregates,
-                        Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                        LO& numNonAggregatedNodes) const {
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    auto vertex2AggId           = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner             = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors                 = aggregates.GetGraphColors();
-    const LO numColors          = aggregates.GetGraphNumColors();
-    const LO numLocalAggregates = aggregates.GetNumAggregates();
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    const LO defaultConnectWeight = 100;
-    const LO penaltyConnectWeight = 10;
-
-    Kokkos::View<LO*, device_type> aggWeight    (Kokkos::ViewAllocateWithoutInitializing("aggWeight"),     numLocalAggregates); // This gets re-initialized at the start of each "color" loop
-    Kokkos::View<LO*, device_type> connectWeight(Kokkos::ViewAllocateWithoutInitializing("connectWeight"), numRows);
-    Kokkos::View<LO*, device_type> aggPenalties ("aggPenalties",  numLocalAggregates);// This gets initialized to zero here
-
-    Kokkos::deep_copy(connectWeight, defaultConnectWeight);
-
-    // taw: by running the aggregation routine more than once there is a chance that also
-    // non-aggregated nodes with a node distance of two are added to existing aggregates.
-    // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
-    // should be sufficient.
-    // lbv: If the prior phase of aggregation where run without specifying an aggregate size,
-    // the distance 2 coloring and phase 1 aggregation actually guarantee that only one iteration
-    // is needed to reach distance 2 neighbors.
-    int maxIters = 2;
-    int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
-    if(maxNodesPerAggregate == std::numeric_limits<int>::max()) {maxIters = 1;}
-    for (int iter = 0; iter < maxIters; ++iter) {
-      for(LO color = 1; color <= numColors; ++color) {
-        Kokkos::deep_copy(aggWeight, 0);
-
-        //the reduce counts how many nodes are aggregated by this phase,
-        //which will then be subtracted from numNonAggregatedNodes
-        LO numAggregated = 0;
-        Kokkos::parallel_reduce("Aggregation Phase 2b: aggregates expansion",
-                                Kokkos::RangePolicy<execution_space>(0, numRows),
-                                KOKKOS_LAMBDA (const LO i, LO& tmpNumAggregated) {
-                                  if (aggStat(i) != READY || colors(i) != color)
-                                    return;
-
-                                  auto neighOfINode = lclLWGraph.getNeighborVertices(i);
-                                  for (int j = 0; j < neighOfINode.length; j++) {
-                                    LO neigh = neighOfINode(j);
-
-                                    // We don't check (neigh != i), as it is covered by checking
-                                    // (aggStat[neigh] == AGGREGATED)
-                                    if (lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                        aggStat(neigh) == AGGREGATED)
-                                      Kokkos::atomic_add(&aggWeight(vertex2AggId(neigh, 0)),
-                                                         connectWeight(neigh));
-                                  }
-
-                                  int bestScore   = -100000;
-                                  int bestAggId   = -1;
-                                  int bestConnect = -1;
-
-                                  for (int j = 0; j < neighOfINode.length; j++) {
-                                    LO neigh = neighOfINode(j);
-
-                                    if (lclLWGraph.isLocalNeighborVertex(neigh) &&
-                                        aggStat(neigh) == AGGREGATED) {
-                                      auto aggId = vertex2AggId(neigh, 0);
-                                      int score = aggWeight(aggId) - aggPenalties(aggId);
-
-                                      if (score > bestScore) {
-                                        bestAggId   = aggId;
-                                        bestScore   = score;
-                                        bestConnect = connectWeight(neigh);
-
-                                      } else if (aggId == bestAggId &&
-                                                 connectWeight(neigh) > bestConnect) {
-                                        bestConnect = connectWeight(neigh);
-                                      }
-                                    }
-                                  }
-                                  if (bestScore >= 0) {
-                                    aggStat(i)         = AGGREGATED;
-                                    vertex2AggId(i, 0) = bestAggId;
-                                    procWinner(i, 0)   = myRank;
-
-                                    Kokkos::atomic_add(&aggPenalties(bestAggId), 1);
-                                    connectWeight(i) = bestConnect - penaltyConnectWeight;
-                                    tmpNumAggregated++;
-                                  }
-                                }, numAggregated); //parallel_for
-        numNonAggregatedNodes -= numAggregated;
-      }
-    } // loop over maxIters
-
-  } // BuildAggregatesRandom
-
-
-
-  template <class LO, class GO, class Node>
-  void AggregationPhase2bAlgorithm_kokkos<LO, GO, Node>::
-  BuildAggregatesDeterministic(const ParameterList& params,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const {
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    auto vertex2AggId     = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner       = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors           = aggregates.GetGraphColors();
-    const LO numColors    = aggregates.GetGraphNumColors();
-    LO numLocalAggregates = aggregates.GetNumAggregates();
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    const int defaultConnectWeight = 100;
-    const int penaltyConnectWeight = 10;
-
-    Kokkos::View<int*, device_type> connectWeight    (Kokkos::ViewAllocateWithoutInitializing("connectWeight"),     numRows);
-    Kokkos::View<int*, device_type> aggWeight        (Kokkos::ViewAllocateWithoutInitializing("aggWeight"),         numLocalAggregates);// This gets re-initialized at the start of each "color" loop
-    Kokkos::View<int*, device_type> aggPenaltyUpdates("aggPenaltyUpdates", numLocalAggregates);
-    Kokkos::View<int*, device_type> aggPenalties     ("aggPenalties",      numLocalAggregates);
-
-    Kokkos::deep_copy(connectWeight, defaultConnectWeight);
-
-    // We do this cycle twice.
-    // I don't know why, but ML does it too
-    // taw: by running the aggregation routine more than once there is a chance that also
-    // non-aggregated nodes with a node distance of two are added to existing aggregates.
-    // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
-    // should be sufficient.
-    int maxIters = 2;
-    int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
-    if(maxNodesPerAggregate == std::numeric_limits<int>::max()) {maxIters = 1;}
-    for (int iter = 0; iter < maxIters; ++iter) {
-      for(LO color = 1; color <= numColors; color++) {
-        Kokkos::deep_copy(aggWeight, 0);
-
-        //the reduce counts how many nodes are aggregated by this phase,
-        //which will then be subtracted from numNonAggregatedNodes
-        LO numAggregated = 0;
-        Kokkos::parallel_for("Aggregation Phase 2b: updating agg weights",
+              // We don't check (neigh != i), as it is covered by checking
+              // (aggStat[neigh] == AGGREGATED)
+              if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                  aggStat(neigh) == AGGREGATED)
+                Kokkos::atomic_add(&aggWeight(vertex2AggId(neigh, 0)),
+                                   connectWeight(neigh));
+            }
+
+            int bestScore   = -100000;
+            int bestAggId   = -1;
+            int bestConnect = -1;
+
+            for (int j = 0; j < neighOfINode.length; j++) {
+              LO neigh = neighOfINode(j);
+
+              if (lclLWGraph.isLocalNeighborVertex(neigh) &&
+                  aggStat(neigh) == AGGREGATED) {
+                auto aggId = vertex2AggId(neigh, 0);
+                int score  = aggWeight(aggId) - aggPenalties(aggId);
+
+                if (score > bestScore) {
+                  bestAggId   = aggId;
+                  bestScore   = score;
+                  bestConnect = connectWeight(neigh);
+
+                } else if (aggId == bestAggId &&
+                           connectWeight(neigh) > bestConnect) {
+                  bestConnect = connectWeight(neigh);
+                }
+              }
+            }
+            if (bestScore >= 0) {
+              aggStat(i)         = AGGREGATED;
+              vertex2AggId(i, 0) = bestAggId;
+              procWinner(i, 0)   = myRank;
+
+              Kokkos::atomic_add(&aggPenalties(bestAggId), 1);
+              connectWeight(i) = bestConnect - penaltyConnectWeight;
+              tmpNumAggregated++;
+            }
+          },
+          numAggregated);  // parallel_for
+      numNonAggregatedNodes -= numAggregated;
+    }
+  }  // loop over maxIters
+
+}  // BuildAggregatesRandom
+
+template <class LO, class GO, class Node>
+void AggregationPhase2bAlgorithm_kokkos<LO, GO, Node>::
+    BuildAggregatesDeterministic(const ParameterList& params,
+                                 const LWGraph_kokkos& graph,
+                                 Aggregates& aggregates,
+                                 Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                                 LO& numNonAggregatedNodes) const {
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  auto vertex2AggId     = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner       = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors           = aggregates.GetGraphColors();
+  const LO numColors    = aggregates.GetGraphNumColors();
+  LO numLocalAggregates = aggregates.GetNumAggregates();
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  const int defaultConnectWeight = 100;
+  const int penaltyConnectWeight = 10;
+
+  Kokkos::View<int*, device_type> connectWeight(Kokkos::ViewAllocateWithoutInitializing("connectWeight"), numRows);
+  Kokkos::View<int*, device_type> aggWeight(Kokkos::ViewAllocateWithoutInitializing("aggWeight"), numLocalAggregates);  // This gets re-initialized at the start of each "color" loop
+  Kokkos::View<int*, device_type> aggPenaltyUpdates("aggPenaltyUpdates", numLocalAggregates);
+  Kokkos::View<int*, device_type> aggPenalties("aggPenalties", numLocalAggregates);
+
+  Kokkos::deep_copy(connectWeight, defaultConnectWeight);
+
+  // We do this cycle twice.
+  // I don't know why, but ML does it too
+  // taw: by running the aggregation routine more than once there is a chance that also
+  // non-aggregated nodes with a node distance of two are added to existing aggregates.
+  // Assuming that the aggregate size is 3 in each direction running the algorithm only twice
+  // should be sufficient.
+  int maxIters             = 2;
+  int maxNodesPerAggregate = params.get<int>("aggregation: max agg size");
+  if (maxNodesPerAggregate == std::numeric_limits<int>::max()) {
+    maxIters = 1;
+  }
+  for (int iter = 0; iter < maxIters; ++iter) {
+    for (LO color = 1; color <= numColors; color++) {
+      Kokkos::deep_copy(aggWeight, 0);
+
+      // the reduce counts how many nodes are aggregated by this phase,
+      // which will then be subtracted from numNonAggregatedNodes
+      LO numAggregated = 0;
+      Kokkos::parallel_for(
+          "Aggregation Phase 2b: updating agg weights",
           Kokkos::RangePolicy<execution_space>(0, numRows),
-          KOKKOS_LAMBDA (const LO i)
-          {
+          KOKKOS_LAMBDA(const LO i) {
             if (aggStat(i) != READY || colors(i) != color)
               return;
             auto neighOfINode = lclLWGraph.getNeighborVertices(i);
@@ -242,15 +243,15 @@ namespace MueLu {
               // (aggStat[neigh] == AGGREGATED)
               if (lclLWGraph.isLocalNeighborVertex(neigh) &&
                   aggStat(neigh) == AGGREGATED)
-              Kokkos::atomic_add(&aggWeight(vertex2AggId(neigh, 0)),
-                  connectWeight(neigh));
+                Kokkos::atomic_add(&aggWeight(vertex2AggId(neigh, 0)),
+                                   connectWeight(neigh));
             }
           });
 
-        Kokkos::parallel_reduce("Aggregation Phase 2b: aggregates expansion",
+      Kokkos::parallel_reduce(
+          "Aggregation Phase 2b: aggregates expansion",
           Kokkos::RangePolicy<execution_space>(0, numRows),
-          KOKKOS_LAMBDA (const LO i, LO& tmpNumAggregated)
-          {
+          KOKKOS_LAMBDA(const LO i, LO& tmpNumAggregated) {
             if (aggStat(i) != READY || colors(i) != color)
               return;
             int bestScore   = -100000;
@@ -264,7 +265,7 @@ namespace MueLu {
               if (lclLWGraph.isLocalNeighborVertex(neigh) &&
                   aggStat(neigh) == AGGREGATED) {
                 auto aggId = vertex2AggId(neigh, 0);
-                int score = aggWeight(aggId) - aggPenalties(aggId);
+                int score  = aggWeight(aggId) - aggPenalties(aggId);
 
                 if (score > bestScore) {
                   bestAggId   = aggId;
@@ -272,7 +273,7 @@ namespace MueLu {
                   bestConnect = connectWeight(neigh);
 
                 } else if (aggId == bestAggId &&
-                    connectWeight(neigh) > bestConnect) {
+                           connectWeight(neigh) > bestConnect) {
                   bestConnect = connectWeight(neigh);
                 }
               }
@@ -286,19 +287,20 @@ namespace MueLu {
               connectWeight(i) = bestConnect - penaltyConnectWeight;
               tmpNumAggregated++;
             }
-          }, numAggregated); //parallel_reduce
+          },
+          numAggregated);  // parallel_reduce
 
-        Kokkos::parallel_for("Aggregation Phase 2b: updating agg penalties",
+      Kokkos::parallel_for(
+          "Aggregation Phase 2b: updating agg penalties",
           Kokkos::RangePolicy<execution_space>(0, numLocalAggregates),
-          KOKKOS_LAMBDA (const LO agg)
-          {
+          KOKKOS_LAMBDA(const LO agg) {
             aggPenalties(agg) += aggPenaltyUpdates(agg);
             aggPenaltyUpdates(agg) = 0;
           });
-        numNonAggregatedNodes -= numAggregated;
-      }
-    } // loop over k
-  } // BuildAggregatesDeterministic
-} // end namespace
+      numNonAggregatedNodes -= numAggregated;
+    }
+  }  // loop over k
+}  // BuildAggregatesDeterministic
+}  // namespace MueLu
 
-#endif // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_decl.hpp
index 20e5fc8a7222..cfe035693b80 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_decl.hpp
@@ -55,58 +55,55 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase3Algorithm class.
-    @brief Handle leftover nodes. Try to avoid singleton nodes
-    @ingroup Aggregation
-
-    ### Idea ###
-    In phase 3 we try to stick unaggregated nodes into a neighboring aggregate.
-    We try to avoid singletons: we first try to build a new aggregate containing
-    all neighboring non-aggregated nodes. If we cannot build a new aggregate,
-    we add the non-aggregated node to the first adjacent aggregate.
-    Only if there is no adjacent aggregate, we create a singleton node aggregate.
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase3Algorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase3Algorithm class.
+  @brief Handle leftover nodes. Try to avoid singleton nodes
+  @ingroup Aggregation
+
+  ### Idea ###
+  In phase 3 we try to stick unaggregated nodes into a neighboring aggregate.
+  We try to avoid singletons: we first try to build a new aggregate containing
+  all neighboring non-aggregated nodes. If we cannot build a new aggregate,
+  we add the non-aggregated node to the first adjacent aggregate.
+  Only if there is no adjacent aggregate, we create a singleton node aggregate.
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase3Algorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE3ALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase3Algorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase3Algorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase3Algorithm() { }
+  //! Destructor.
+  virtual ~AggregationPhase3Algorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 3 (cleanup)"; }
+};
 
-    std::string description() const { return "Phase 3 (cleanup)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_SHORT
 
-
 #endif /* MUELU_AGGREGATIONPHASE3ALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_def.hpp
index ac462bcccf4b..e958226e5c4f 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_def.hpp
@@ -60,150 +60,147 @@
 
 namespace MueLu {
 
-  // Try to stick unaggregated nodes into a neighboring aggregate if they are
-  // not already too big. Otherwise, make a new aggregate
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase3Algorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
+// Try to stick unaggregated nodes into a neighboring aggregate if they are
+// not already too big. Otherwise, make a new aggregate
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase3Algorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
 
-    bool makeNonAdjAggs = false;
-    bool error_on_isolated = false;
-    if(params.isParameter("aggregation: error on nodes with no on-rank neighbors"))
-      error_on_isolated = params.get<bool>("aggregation: error on nodes with no on-rank neighbors");
-    if(params.isParameter("aggregation: phase3 avoid singletons"))
-      makeNonAdjAggs = params.get<bool>("aggregation: phase3 avoid singletons");
+  bool makeNonAdjAggs    = false;
+  bool error_on_isolated = false;
+  if (params.isParameter("aggregation: error on nodes with no on-rank neighbors"))
+    error_on_isolated = params.get<bool>("aggregation: error on nodes with no on-rank neighbors");
+  if (params.isParameter("aggregation: phase3 avoid singletons"))
+    makeNonAdjAggs = params.get<bool>("aggregation: phase3 avoid singletons");
 
-    size_t numSingletons=0;
+  size_t numSingletons = 0;
 
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
 
-    ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()  ->getDataNonConst(0);
+  ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
 
-    LO numLocalAggregates = aggregates.GetNumAggregates();
+  LO numLocalAggregates = aggregates.GetNumAggregates();
 
-    for (LO i = 0; i < numRows; i++) {
-      if (aggStat[i] == AGGREGATED || aggStat[i] == IGNORED)
-        continue;
+  for (LO i = 0; i < numRows; i++) {
+    if (aggStat[i] == AGGREGATED || aggStat[i] == IGNORED)
+      continue;
 
-      ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(i);
+    ArrayView<const LocalOrdinal> neighOfINode = graph.getNeighborVertices(i);
 
-      // We don't want a singleton. So lets see if there is an unaggregated
-      // neighbor that we can also put with this point.
-      bool isNewAggregate = false;
-      bool failedToAggregate = true;
-      for (int j = 0; j < neighOfINode.size(); j++) {
-        LO neigh = neighOfINode[j];
+    // We don't want a singleton. So lets see if there is an unaggregated
+    // neighbor that we can also put with this point.
+    bool isNewAggregate    = false;
+    bool failedToAggregate = true;
+    for (int j = 0; j < neighOfINode.size(); j++) {
+      LO neigh = neighOfINode[j];
 
-        if (neigh != i && graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY) {
-          isNewAggregate = true;
+      if (neigh != i && graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == READY) {
+        isNewAggregate = true;
 
-          aggStat     [neigh] = AGGREGATED;
-          vertex2AggId[neigh] = numLocalAggregates;
-          procWinner  [neigh] = myRank;
+        aggStat[neigh]      = AGGREGATED;
+        vertex2AggId[neigh] = numLocalAggregates;
+        procWinner[neigh]   = myRank;
 
-          numNonAggregatedNodes--;
-        }
+        numNonAggregatedNodes--;
       }
+    }
 
-      if (isNewAggregate) {
-        // Create new aggregate (not singleton)
-        aggStat     [i] = AGGREGATED;
-        procWinner  [i] = myRank;
-        numNonAggregatedNodes--;
-        aggregates.SetIsRoot(i);
-        vertex2AggId[i] = numLocalAggregates++;
+    if (isNewAggregate) {
+      // Create new aggregate (not singleton)
+      aggStat[i]    = AGGREGATED;
+      procWinner[i] = myRank;
+      numNonAggregatedNodes--;
+      aggregates.SetIsRoot(i);
+      vertex2AggId[i] = numLocalAggregates++;
+
+      failedToAggregate = false;
+    } else {
+      // We do not want a singleton, but there are no non-aggregated
+      // neighbors. Lets see if we can connect to any other aggregates
+      // NOTE: This is very similar to phase 2b, but simplier: we stop with
+      // the first found aggregate
+      int j = 0;
+      for (; j < neighOfINode.size(); j++) {
+        LO neigh = neighOfINode[j];
 
-        failedToAggregate = false;
-      } else {
-        // We do not want a singleton, but there are no non-aggregated
-        // neighbors. Lets see if we can connect to any other aggregates
-        // NOTE: This is very similar to phase 2b, but simplier: we stop with
-        // the first found aggregate
-        int j = 0;
-        for (; j < neighOfINode.size(); j++) {
-          LO neigh = neighOfINode[j];
-
-          // We don't check (neigh != rootCandidate), as it is covered by checking (aggStat[neigh] == AGGREGATED)
-          if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED)
-            break;
-        }
-
-        if (j < neighOfINode.size()) {
-          // Assign to an adjacent aggregate
-          vertex2AggId[i] = vertex2AggId[neighOfINode[j]];
-          numNonAggregatedNodes--;
-          failedToAggregate = false;
-        }
+        // We don't check (neigh != rootCandidate), as it is covered by checking (aggStat[neigh] == AGGREGATED)
+        if (graph.isLocalNeighborVertex(neigh) && aggStat[neigh] == AGGREGATED)
+          break;
       }
 
-      if (failedToAggregate && makeNonAdjAggs) {
-        //  it we are still didn't find an aggregate home for i (i.e., we have
-        //  a potential singleton), we are desperate. Basically, we seek to
-        //  group i with any other local point to form an aggregate (even if
-        //  it is not a neighbor of i. Either we find a vertex that is already
-        //  aggregated or not aggregated.
-        //    1) if found vertex is aggregated, then assign i to this aggregate
-        //    2) if found vertex is not aggregated, create new aggregate
-
-
-        for (LO ii = 0; ii < numRows; ii++) { // look for anyone else
-          if ( (ii != i) && (aggStat[ii] != IGNORED) ) {
-            failedToAggregate = false;       // found someone so start
-            aggStat[i]   = AGGREGATED;  // marking i as aggregated
-            procWinner[i]= myRank;
-
-            if (aggStat[ii] == AGGREGATED)
-              vertex2AggId[i] = vertex2AggId[ii];
-            else {
-              vertex2AggId[i]  = numLocalAggregates;
-              vertex2AggId[ii] = numLocalAggregates;
-              aggStat     [ii] = AGGREGATED;
-              procWinner  [ii] = myRank;
-              numNonAggregatedNodes--;   // acounts for ii now being aggregated
-              aggregates.SetIsRoot(i);
-              numLocalAggregates++;
-            }
-            numNonAggregatedNodes--;   // accounts for i now being aggregated
-            break;
-          } //if ( (ii != i) && (aggStat[ii] != IGNORED ...
-        } //for (LO ii = 0; ...
-      }
-      if (failedToAggregate) {
-        if (error_on_isolated) {
-          // Error on this isolated node, as the user has requested
-          std::ostringstream oss;
-          oss<<"MueLu::AggregationPhase3Algorithm::BuildAggregates: MueLu has detected a non-Dirichlet node that has no on-rank neighbors and is terminating (by user request). "<<std::endl;
-          oss<<"If this error is being generated at level 0, this is due to an initial partitioning problem in your matrix."<<std::endl;
-          oss<<"If this error is being generated at any other level, try turning on repartitioning, which may fix this problem."<<std::endl;
-          throw Exceptions::RuntimeError(oss.str());
-        } else {
-          // Create new aggregate (singleton)
-	  //          this->GetOStream(Warnings1) << "Found singleton: " << i << std::endl;
-	  numSingletons++;
-
-          aggregates.SetIsRoot(i);
-          vertex2AggId[i] = numLocalAggregates++;
-          numNonAggregatedNodes--;
-        }
+      if (j < neighOfINode.size()) {
+        // Assign to an adjacent aggregate
+        vertex2AggId[i] = vertex2AggId[neighOfINode[j]];
+        numNonAggregatedNodes--;
+        failedToAggregate = false;
       }
+    }
+
+    if (failedToAggregate && makeNonAdjAggs) {
+      //  it we are still didn't find an aggregate home for i (i.e., we have
+      //  a potential singleton), we are desperate. Basically, we seek to
+      //  group i with any other local point to form an aggregate (even if
+      //  it is not a neighbor of i. Either we find a vertex that is already
+      //  aggregated or not aggregated.
+      //    1) if found vertex is aggregated, then assign i to this aggregate
+      //    2) if found vertex is not aggregated, create new aggregate
+
+      for (LO ii = 0; ii < numRows; ii++) {  // look for anyone else
+        if ((ii != i) && (aggStat[ii] != IGNORED)) {
+          failedToAggregate = false;       // found someone so start
+          aggStat[i]        = AGGREGATED;  // marking i as aggregated
+          procWinner[i]     = myRank;
+
+          if (aggStat[ii] == AGGREGATED)
+            vertex2AggId[i] = vertex2AggId[ii];
+          else {
+            vertex2AggId[i]  = numLocalAggregates;
+            vertex2AggId[ii] = numLocalAggregates;
+            aggStat[ii]      = AGGREGATED;
+            procWinner[ii]   = myRank;
+            numNonAggregatedNodes--;  // acounts for ii now being aggregated
+            aggregates.SetIsRoot(i);
+            numLocalAggregates++;
+          }
+          numNonAggregatedNodes--;  // accounts for i now being aggregated
+          break;
+        }  // if ( (ii != i) && (aggStat[ii] != IGNORED ...
+      }    // for (LO ii = 0; ...
+    }
+    if (failedToAggregate) {
+      if (error_on_isolated) {
+        // Error on this isolated node, as the user has requested
+        std::ostringstream oss;
+        oss << "MueLu::AggregationPhase3Algorithm::BuildAggregates: MueLu has detected a non-Dirichlet node that has no on-rank neighbors and is terminating (by user request). " << std::endl;
+        oss << "If this error is being generated at level 0, this is due to an initial partitioning problem in your matrix." << std::endl;
+        oss << "If this error is being generated at any other level, try turning on repartitioning, which may fix this problem." << std::endl;
+        throw Exceptions::RuntimeError(oss.str());
+      } else {
+        // Create new aggregate (singleton)
+        //          this->GetOStream(Warnings1) << "Found singleton: " << i << std::endl;
+        numSingletons++;
 
-      // One way or another, the node is aggregated (possibly into a singleton)
-      aggStat   [i] = AGGREGATED;
-      procWinner[i] = myRank;
+        aggregates.SetIsRoot(i);
+        vertex2AggId[i] = numLocalAggregates++;
+        numNonAggregatedNodes--;
+      }
+    }
 
-    } // loop over numRows
-    
+    // One way or another, the node is aggregated (possibly into a singleton)
+    aggStat[i]    = AGGREGATED;
+    procWinner[i] = myRank;
 
-    if(numSingletons > 0)
-      this->GetOStream(Runtime0)<<"  WARNING Rank "<<myRank<<" singletons :"<<numSingletons<<" (phase)"<<std::endl;
+  }  // loop over numRows
 
+  if (numSingletons > 0)
+    this->GetOStream(Runtime0) << "  WARNING Rank " << myRank << " singletons :" << numSingletons << " (phase)" << std::endl;
 
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates);
-  }
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates);
+}
 
-} // end namespace
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONPHASE3ALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_decl.hpp
index 9911ac016c5e..5a97db952f94 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_decl.hpp
@@ -58,70 +58,68 @@
 #include "MueLu_LWGraph_kokkos_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class AggregationPhase3Algorithm class.
-    @brief Handle leftover nodes. Try to avoid singleton nodes
-    @ingroup Aggregation
-
-    ### Idea ###
-    In phase 3 we try to stick unaggregated nodes into a neighboring aggregate.
-    We try to avoid singletons: we first try to build a new aggregate containing
-    all neighboring non-aggregated nodes. If we cannot build a new aggregate,
-    we add the non-aggregated node to the first adjacent aggregate.
-    Only if there is no adjacent aggregate, we create a singleton node aggregate.
-
-    ### Comments ###
-    Only nodes with state READY are changed to AGGREGATED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class AggregationPhase3Algorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class AggregationPhase3Algorithm class.
+  @brief Handle leftover nodes. Try to avoid singleton nodes
+  @ingroup Aggregation
+
+  ### Idea ###
+  In phase 3 we try to stick unaggregated nodes into a neighboring aggregate.
+  We try to avoid singletons: we first try to build a new aggregate containing
+  all neighboring non-aggregated nodes. If we cannot build a new aggregate,
+  we add the non-aggregated node to the first adjacent aggregate.
+  Only if there is no adjacent aggregate, we create a singleton node aggregate.
+
+  ### Comments ###
+  Only nodes with state READY are changed to AGGREGATED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationPhase3Algorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type     = typename LWGraph_kokkos::device_type;
-    using execution_space = typename LWGraph_kokkos::execution_space;
-    using memory_space    = typename LWGraph_kokkos::memory_space;
+ public:
+  using device_type     = typename LWGraph_kokkos::device_type;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregationPhase3Algorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  AggregationPhase3Algorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~AggregationPhase3Algorithm_kokkos() { }
+  //! Destructor.
+  virtual ~AggregationPhase3Algorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
 
-    void BuildAggregates(const ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
+  void BuildAggregatesRandom(const ParameterList& params,
+                             const LWGraph_kokkos& graph,
+                             Aggregates& aggregates,
+                             Kokkos::View<unsigned*, device_type>& aggStat,
+                             LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregatesRandom(const ParameterList& params,
-                               const LWGraph_kokkos& graph,
-                               Aggregates& aggregates,
-                               Kokkos::View<unsigned*, device_type>& aggStat,
-                               LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase 3 (cleanup)"; }
+};
 
-    std::string description() const { return "Phase 3 (cleanup)"; }
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_SHORT
-#endif // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_def.hpp
index 4bacd24017cd..a684149e179f 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_AggregationPhase3Algorithm_kokkos_def.hpp
@@ -62,171 +62,171 @@
 
 namespace MueLu {
 
-  // Try to stick unaggregated nodes into a neighboring aggregate if they are
-  // not already too big. Otherwise, make a new aggregate
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase3Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const ParameterList& params,
-                  const LWGraph_kokkos& graph,
-                  Aggregates& aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-
-    // So far we only have the non-deterministic version of the algorithm...
-    if(params.get<bool>("aggregation: deterministic")) {
-      Monitor m(*this, "BuildAggregatesDeterministic");
-      BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    } else {
-      Monitor m(*this, "BuildAggregatesRandom");
-      BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
-    }
-
+// Try to stick unaggregated nodes into a neighboring aggregate if they are
+// not already too big. Otherwise, make a new aggregate
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase3Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const ParameterList& params,
+                    const LWGraph_kokkos& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  // So far we only have the non-deterministic version of the algorithm...
+  if (params.get<bool>("aggregation: deterministic")) {
+    Monitor m(*this, "BuildAggregatesDeterministic");
+    BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
+  } else {
+    Monitor m(*this, "BuildAggregatesRandom");
+    BuildAggregatesRandom(params, graph, aggregates, aggStat, numNonAggregatedNodes);
   }
-
-  // Try to stick unaggregated nodes into a neighboring aggregate if they are
-  // not already too big. Otherwise, make a new aggregate
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationPhase3Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregatesRandom(const ParameterList& params,
-                        const LWGraph_kokkos& graph,
-                        Aggregates& aggregates,
-                        Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                        LO& numNonAggregatedNodes) const {
-
-    bool error_on_isolated = params.get<bool>("aggregation: error on nodes with no on-rank neighbors");
-    bool makeNonAdjAggs = params.get<bool>("aggregation: phase3 avoid singletons");
-
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
-
-    auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner    = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto colors        = aggregates.GetGraphColors();
-    const LO numColors = aggregates.GetGraphNumColors();
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    Kokkos::View<LO, device_type> numAggregates("numAggregates");
-    Kokkos::deep_copy(numAggregates, aggregates.GetNumAggregates());
-
-    Kokkos::View<unsigned*, device_type> aggStatOld(Kokkos::ViewAllocateWithoutInitializing("Initial aggregation status"), aggStat.extent(0));
+}
+
+// Try to stick unaggregated nodes into a neighboring aggregate if they are
+// not already too big. Otherwise, make a new aggregate
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationPhase3Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregatesRandom(const ParameterList& params,
+                          const LWGraph_kokkos& graph,
+                          Aggregates& aggregates,
+                          Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                          LO& numNonAggregatedNodes) const {
+  bool error_on_isolated = params.get<bool>("aggregation: error on nodes with no on-rank neighbors");
+  bool makeNonAdjAggs    = params.get<bool>("aggregation: phase3 avoid singletons");
+
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
+
+  auto vertex2AggId  = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner    = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto colors        = aggregates.GetGraphColors();
+  const LO numColors = aggregates.GetGraphNumColors();
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  Kokkos::View<LO, device_type> numAggregates("numAggregates");
+  Kokkos::deep_copy(numAggregates, aggregates.GetNumAggregates());
+
+  Kokkos::View<unsigned*, device_type> aggStatOld(Kokkos::ViewAllocateWithoutInitializing("Initial aggregation status"), aggStat.extent(0));
+  Kokkos::deep_copy(aggStatOld, aggStat);
+  Kokkos::View<LO, device_type> numNonAggregated("numNonAggregated");
+  Kokkos::deep_copy(numNonAggregated, numNonAggregatedNodes);
+  for (int color = 1; color < numColors + 1; ++color) {
+    Kokkos::parallel_for(
+        "Aggregation Phase 3: aggregates clean-up",
+        Kokkos::RangePolicy<execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const LO nodeIdx) {
+          // Check if node has already been treated?
+          if ((colors(nodeIdx) != color) ||
+              (aggStatOld(nodeIdx) == AGGREGATED) ||
+              (aggStatOld(nodeIdx) == IGNORED)) {
+            return;
+          }
+
+          // Grab node neighbors
+          auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
+          LO neighIdx;
+
+          // We don't want a singleton.
+          // So lets see if any neighbors can be used to form a new aggregate?
+          bool isNewAggregate = false;
+          for (int neigh = 0; neigh < neighbors.length; ++neigh) {
+            neighIdx = neighbors(neigh);
+
+            if ((neighIdx != nodeIdx) &&
+                lclLWGraph.isLocalNeighborVertex(neighIdx) &&
+                (aggStatOld(neighIdx) == READY)) {
+              isNewAggregate = true;
+              break;
+            }
+          }
+
+          // We can form a new non singleton aggregate!
+          if (isNewAggregate) {
+            // If this is the aggregate root
+            // we need to process the nodes in the aggregate
+            const LO aggId           = Kokkos::atomic_fetch_add(&numAggregates(), 1);
+            aggStat(nodeIdx)         = AGGREGATED;
+            procWinner(nodeIdx, 0)   = myRank;
+            vertex2AggId(nodeIdx, 0) = aggId;
+            // aggregates.SetIsRoot(nodeIdx);
+            Kokkos::atomic_decrement(&numNonAggregated());
+            for (int neigh = 0; neigh < neighbors.length; ++neigh) {
+              neighIdx = neighbors(neigh);
+              if ((neighIdx != nodeIdx) &&
+                  lclLWGraph.isLocalNeighborVertex(neighIdx) &&
+                  (aggStatOld(neighIdx) == READY)) {
+                aggStat(neighIdx)         = AGGREGATED;
+                procWinner(neighIdx, 0)   = myRank;
+                vertex2AggId(neighIdx, 0) = aggId;
+                Kokkos::atomic_decrement(&numNonAggregated());
+              }
+            }
+            return;
+          }
+
+          // Getting a little desperate!
+          // Let us try to aggregate into a neighboring aggregate
+          for (int neigh = 0; neigh < neighbors.length; ++neigh) {
+            neighIdx = neighbors(neigh);
+            if (lclLWGraph.isLocalNeighborVertex(neighIdx) &&
+                (aggStatOld(neighIdx) == AGGREGATED)) {
+              aggStat(nodeIdx)         = AGGREGATED;
+              procWinner(nodeIdx, 0)   = myRank;
+              vertex2AggId(nodeIdx, 0) = vertex2AggId(neighIdx, 0);
+              Kokkos::atomic_decrement(&numNonAggregated());
+              return;
+            }
+          }
+
+          // Getting quite desperate!
+          // Let us try to make a non contiguous aggregate
+          if (makeNonAdjAggs) {
+            for (LO otherNodeIdx = 0; otherNodeIdx < numRows; ++otherNodeIdx) {
+              if ((otherNodeIdx != nodeIdx) &&
+                  (aggStatOld(otherNodeIdx) == AGGREGATED)) {
+                aggStat(nodeIdx)         = AGGREGATED;
+                procWinner(nodeIdx, 0)   = myRank;
+                vertex2AggId(nodeIdx, 0) = vertex2AggId(otherNodeIdx, 0);
+                Kokkos::atomic_decrement(&numNonAggregated());
+                return;
+              }
+            }
+          }
+
+          // Total deperation!
+          // Let us make a singleton
+          if (!error_on_isolated) {
+            const LO aggId           = Kokkos::atomic_fetch_add(&numAggregates(), 1);
+            aggStat(nodeIdx)         = AGGREGATED;
+            procWinner(nodeIdx, 0)   = myRank;
+            vertex2AggId(nodeIdx, 0) = aggId;
+            Kokkos::atomic_decrement(&numNonAggregated());
+          }
+        });
+    // LBV on 09/27/19: here we could copy numNonAggregated to host
+    // and check for it to be equal to 0 in which case we can stop
+    // looping over the different colors...
     Kokkos::deep_copy(aggStatOld, aggStat);
-    Kokkos::View<LO, device_type> numNonAggregated("numNonAggregated");
-    Kokkos::deep_copy(numNonAggregated, numNonAggregatedNodes);
-    for(int color = 1; color < numColors + 1; ++color) {
-      Kokkos::parallel_for("Aggregation Phase 3: aggregates clean-up",
-                           Kokkos::RangePolicy<execution_space>(0, numRows),
-                           KOKKOS_LAMBDA(const LO nodeIdx) {
-                             // Check if node has already been treated?
-                             if( (colors(nodeIdx) != color) ||
-                                 (aggStatOld(nodeIdx) == AGGREGATED) ||
-                                 (aggStatOld(nodeIdx) == IGNORED) ){ return; }
-
-                             // Grab node neighbors
-                             auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
-                             LO neighIdx;
-
-                             // We don't want a singleton.
-                             // So lets see if any neighbors can be used to form a new aggregate?
-                             bool isNewAggregate    = false;
-                             for(int neigh = 0; neigh < neighbors.length; ++neigh) {
-                               neighIdx = neighbors(neigh);
-
-                               if((neighIdx != nodeIdx) &&
-                                  lclLWGraph.isLocalNeighborVertex(neighIdx) &&
-                                  (aggStatOld(neighIdx) == READY)) {
-                                 isNewAggregate = true;
-                                 break;
-                               }
-                             }
-
-                             // We can form a new non singleton aggregate!
-                             if(isNewAggregate) {
-                               // If this is the aggregate root
-                               // we need to process the nodes in the aggregate
-                               const LO aggId = Kokkos::atomic_fetch_add(&numAggregates(), 1);
-                               aggStat(nodeIdx)         = AGGREGATED;
-                               procWinner(nodeIdx, 0)   = myRank;
-                               vertex2AggId(nodeIdx, 0) = aggId;
-                               // aggregates.SetIsRoot(nodeIdx);
-                               Kokkos::atomic_decrement(&numNonAggregated());
-                               for(int neigh = 0; neigh < neighbors.length; ++neigh) {
-                                 neighIdx = neighbors(neigh);
-                                 if((neighIdx != nodeIdx) &&
-                                    lclLWGraph.isLocalNeighborVertex(neighIdx) &&
-                                    (aggStatOld(neighIdx) == READY)) {
-                                   aggStat(neighIdx)         = AGGREGATED;
-                                   procWinner(neighIdx, 0)   = myRank;
-                                   vertex2AggId(neighIdx, 0) = aggId;
-                                   Kokkos::atomic_decrement(&numNonAggregated());
-                                 }
-                               }
-                               return;
-                             }
-
-                             // Getting a little desperate!
-                             // Let us try to aggregate into a neighboring aggregate
-                             for(int neigh = 0; neigh < neighbors.length; ++neigh) {
-                               neighIdx = neighbors(neigh);
-                               if (lclLWGraph.isLocalNeighborVertex(neighIdx) &&
-                                   (aggStatOld(neighIdx) == AGGREGATED)) {
-                                 aggStat(nodeIdx)         = AGGREGATED;
-                                 procWinner(nodeIdx, 0)   = myRank;
-                                 vertex2AggId(nodeIdx, 0) = vertex2AggId(neighIdx, 0);
-                                 Kokkos::atomic_decrement(&numNonAggregated());
-                                 return;
-                               }
-                             }
-
-                             // Getting quite desperate!
-                             // Let us try to make a non contiguous aggregate
-                             if(makeNonAdjAggs) {
-                               for(LO otherNodeIdx = 0; otherNodeIdx < numRows; ++otherNodeIdx) {
-                                 if((otherNodeIdx != nodeIdx) &&
-                                    (aggStatOld(otherNodeIdx) == AGGREGATED)) {
-                                   aggStat(nodeIdx)         = AGGREGATED;
-                                   procWinner(nodeIdx, 0)   = myRank;
-                                   vertex2AggId(nodeIdx, 0) = vertex2AggId(otherNodeIdx, 0);
-                                   Kokkos::atomic_decrement(&numNonAggregated());
-                                   return;
-                                 }
-                               }
-                             }
-
-                             // Total deperation!
-                             // Let us make a singleton
-                             if(!error_on_isolated) {
-                               const LO aggId = Kokkos::atomic_fetch_add(&numAggregates(), 1);
-                               aggStat(nodeIdx)         = AGGREGATED;
-                               procWinner(nodeIdx, 0)   = myRank;
-                               vertex2AggId(nodeIdx, 0) = aggId;
-                               Kokkos::atomic_decrement(&numNonAggregated());
-                             }
-                           });
-      // LBV on 09/27/19: here we could copy numNonAggregated to host
-      // and check for it to be equal to 0 in which case we can stop
-      // looping over the different colors...
-      Kokkos::deep_copy(aggStatOld, aggStat);
-    } // loop over colors
-
-    auto numNonAggregated_h = Kokkos::create_mirror_view(numNonAggregated);
-    Kokkos::deep_copy(numNonAggregated_h, numNonAggregated);
-    numNonAggregatedNodes = numNonAggregated_h();
-    if( (error_on_isolated) && (numNonAggregatedNodes > 0) ) {
-          // Error on this isolated node, as the user has requested
-          std::ostringstream oss;
-          oss<<"MueLu::AggregationPhase3Algorithm::BuildAggregates: MueLu has detected a non-Dirichlet node that has no on-rank neighbors and is terminating (by user request). "<<std::endl;
-          oss<<"If this error is being generated at level 0, this is due to an initial partitioning problem in your matrix."<<std::endl;
-          oss<<"If this error is being generated at any other level, try turning on repartitioning, which may fix this problem."<<std::endl;
-          throw Exceptions::RuntimeError(oss.str());
-    }
-
-    // update aggregate object
-    auto numAggregates_h = Kokkos::create_mirror_view(numAggregates);
-    Kokkos::deep_copy(numAggregates_h, numAggregates);
-    aggregates.SetNumAggregates(numAggregates_h());
+  }  // loop over colors
+
+  auto numNonAggregated_h = Kokkos::create_mirror_view(numNonAggregated);
+  Kokkos::deep_copy(numNonAggregated_h, numNonAggregated);
+  numNonAggregatedNodes = numNonAggregated_h();
+  if ((error_on_isolated) && (numNonAggregatedNodes > 0)) {
+    // Error on this isolated node, as the user has requested
+    std::ostringstream oss;
+    oss << "MueLu::AggregationPhase3Algorithm::BuildAggregates: MueLu has detected a non-Dirichlet node that has no on-rank neighbors and is terminating (by user request). " << std::endl;
+    oss << "If this error is being generated at level 0, this is due to an initial partitioning problem in your matrix." << std::endl;
+    oss << "If this error is being generated at any other level, try turning on repartitioning, which may fix this problem." << std::endl;
+    throw Exceptions::RuntimeError(oss.str());
   }
 
-} // end namespace
+  // update aggregate object
+  auto numAggregates_h = Kokkos::create_mirror_view(numAggregates);
+  Kokkos::deep_copy(numAggregates_h, numAggregates);
+  aggregates.SetNumAggregates(numAggregates_h());
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_decl.hpp
index fb3425d5058c..fd383eca5274 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_decl.hpp
@@ -63,54 +63,51 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class InterfaceAggregationAlgorithm class.
-    @brief Algorithm for coarsening a graph with uncoupled aggregation.
-    creates aggregates along an interface using specified root nodes.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    The user can mark some nodes as INTERFACE to build aggregates across an interface.
-    This can be very useful for certain applications. We build aggregates for nodes with
-    the state INTERFACE. Then, the state is changed to AGGREGATED.
-    The InterfaceAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class InterfaceAggregationAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
-#undef MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
-#include "MueLu_UseShortNamesOrdinal.hpp"
+/*!
+  @class InterfaceAggregationAlgorithm class.
+  @brief Algorithm for coarsening a graph with uncoupled aggregation.
+  creates aggregates along an interface using specified root nodes.
+
+  @ingroup Aggregation
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+  ### Idea ###
+  The user can mark some nodes as INTERFACE to build aggregates across an interface.
+  This can be very useful for certain applications. We build aggregates for nodes with
+  the state INTERFACE. Then, the state is changed to AGGREGATED.
+  The InterfaceAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
 
-    //! Constructor.
-    InterfaceAggregationAlgorithm(RCP<const FactoryBase> const &graphFact = Teuchos::null);
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class InterfaceAggregationAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
+#undef MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
+#include "MueLu_UseShortNamesOrdinal.hpp"
 
-    //! Destructor.
-    virtual ~InterfaceAggregationAlgorithm() { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //@}
+  //! Constructor.
+  InterfaceAggregationAlgorithm(RCP<const FactoryBase> const& graphFact = Teuchos::null);
 
+  //! Destructor.
+  virtual ~InterfaceAggregationAlgorithm() {}
 
-    //! @name Aggregation methods.
-    //@{
+  //@}
 
-    /*! @brief Local aggregation. */
+  //! @name Aggregation methods.
+  //@{
 
-    void BuildAggregates(Teuchos::ParameterList const & params, GraphBase const & graph, Aggregates & aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  /*! @brief Local aggregation. */
 
+  void BuildAggregates(Teuchos::ParameterList const& params, GraphBase const& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-  }; //class InterfaceAggregationAlgorithm
+};  // class InterfaceAggregationAlgorithm
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
 #endif /* MUELU_INTERFACEAGGREGATIONALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_def.hpp
index 22dd58e56fad..0ec5f6842c22 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_InterfaceAggregationAlgorithm_def.hpp
@@ -69,40 +69,37 @@
 namespace MueLu {
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::InterfaceAggregationAlgorithm(RCP<const FactoryBase> const &/* graphFact */)
-{
+InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::InterfaceAggregationAlgorithm(RCP<const FactoryBase> const& /* graphFact */) {
 }
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const & /* params */, GraphBase const & graph, Aggregates & aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+void InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const& /* params */, GraphBase const& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
   Monitor m(*this, "BuildAggregates");
 
   const LocalOrdinal nRows = graph.GetNodeNumVertices();
-  const int myRank = graph.GetComm()->getRank();
+  const int myRank         = graph.GetComm()->getRank();
 
   // vertex ids for output
   Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
   Teuchos::ArrayRCP<LocalOrdinal> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
 
   // some internal variables
-  LocalOrdinal numLocalAggregates = aggregates.GetNumAggregates();    // number of local aggregates on current proc
+  LocalOrdinal numLocalAggregates = aggregates.GetNumAggregates();  // number of local aggregates on current proc
 
   // main loop over all local rows of graph(A)
-  for(int iNode1 = 0; iNode1 < nRows; ++iNode1) {
-
+  for (int iNode1 = 0; iNode1 < nRows; ++iNode1) {
     if (aggStat[iNode1] == INTERFACE) {
-
-      aggregates.SetIsRoot(iNode1);    // mark iNode1 as root node for new aggregate 'agg'
+      aggregates.SetIsRoot(iNode1);  // mark iNode1 as root node for new aggregate 'agg'
       int aggIndex = numLocalAggregates;
       std::vector<int> aggList;
       aggList.push_back(iNode1);
       ArrayView<const LO> neighOfINode = graph.getNeighborVertices(iNode1);
 
-      for(int j = 0; j < neighOfINode.size(); ++j) {
+      for (int j = 0; j < neighOfINode.size(); ++j) {
         LO neigh = neighOfINode[j];
-        if(neigh != iNode1 && graph.isLocalNeighborVertex(neigh)) {
-          if(aggStat[neigh] != AGGREGATED && aggStat[neigh] != INTERFACE &&
-             aggStat[neigh] != IGNORED) {
+        if (neigh != iNode1 && graph.isLocalNeighborVertex(neigh)) {
+          if (aggStat[neigh] != AGGREGATED && aggStat[neigh] != INTERFACE &&
+              aggStat[neigh] != IGNORED) {
             aggList.push_back(neigh);
           }
         }
@@ -117,13 +114,12 @@ void InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggr
       numNonAggregatedNodes -= aggList.size();
     }
 
-  } // end for
+  }  // end for
 
   // update aggregate object
   aggregates.SetNumAggregates(numLocalAggregates);
 }
 
-} // end namespace
-
+}  // namespace MueLu
 
 #endif /* MUELU_INTERFACEAGGREGATIONALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_decl.hpp
index 47ebb8038952..b89d3e6a83a5 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_decl.hpp
@@ -57,61 +57,58 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class IsolatedNodeAggregationAlgorithm class.
-    @brief Ignores isolated nodes during aggregation. Marks the node to be "aggregated" without adding real aggregates for them.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    The isolated node aggregation algorithm loops over all non-aggregated nodes
-    (with a state different than aggregated or ignored) which have only themselves
-    as neighbor node. The state of these "isolated" nodes is then set to ignored such
-    that they are not considered in the aggregation. This aggregation algorithm should
-    run as one of the last aggregation algorithms in the aggregation method.
-
-    ### Comments ###
-    Only nodes with state different than READY or AGGREGATED are changed to IGNORED.
-    After that, all nodes should have the state AGGREGATED or IGNORED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class IsolatedNodeAggregationAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class IsolatedNodeAggregationAlgorithm class.
+  @brief Ignores isolated nodes during aggregation. Marks the node to be "aggregated" without adding real aggregates for them.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  The isolated node aggregation algorithm loops over all non-aggregated nodes
+  (with a state different than aggregated or ignored) which have only themselves
+  as neighbor node. The state of these "isolated" nodes is then set to ignored such
+  that they are not considered in the aggregation. This aggregation algorithm should
+  run as one of the last aggregation algorithms in the aggregation method.
+
+  ### Comments ###
+  Only nodes with state different than READY or AGGREGATED are changed to IGNORED.
+  After that, all nodes should have the state AGGREGATED or IGNORED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class IsolatedNodeAggregationAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    IsolatedNodeAggregationAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~IsolatedNodeAggregationAlgorithm() { }
+  //! Constructor.
+  IsolatedNodeAggregationAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //@}
+  //! Destructor.
+  virtual ~IsolatedNodeAggregationAlgorithm() {}
 
+  //@}
 
-    //! @name Aggregation methods.
-    //@{
+  //! @name Aggregation methods.
+  //@{
 
-    /*! @brief Local aggregation. */
+  /*! @brief Local aggregation. */
 
-    void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  void BuildAggregates(const ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    std::string description() const { return "Phase - (isolated)"; }
+  std::string description() const { return "Phase - (isolated)"; }
 
-  }; //class MaxLinkAggregationAlgorithm
+};  // class MaxLinkAggregationAlgorithm
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_SHORT
 
-
 #endif /* MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_def.hpp
index 77147d7990a8..73dd4eddd6c0 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_def.hpp
@@ -53,7 +53,6 @@
 #ifndef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DEF_HPP_
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DEF_HPP_
 
-
 #include <Teuchos_Comm.hpp>
 #include <Teuchos_CommHelpers.hpp>
 
@@ -68,20 +67,20 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IsolatedNodeAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& /* params */, const GraphBase& graph, Aggregates& /* aggregates */, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IsolatedNodeAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(const ParameterList& /* params */, const GraphBase& graph, Aggregates& /* aggregates */, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
 
-    const LO  numRows = graph.GetNodeNumVertices();
+  const LO numRows = graph.GetNodeNumVertices();
 
-    // Remove all isolated nodes
-    for (LO i = 0; i < numRows; i++)
-      if (aggStat[i] != AGGREGATED && aggStat[i] != IGNORED && graph.getNeighborVertices(i).size() == 1) {
-        aggStat[i] = IGNORED;
-        numNonAggregatedNodes--;
-      }
-  }
+  // Remove all isolated nodes
+  for (LO i = 0; i < numRows; i++)
+    if (aggStat[i] != AGGREGATED && aggStat[i] != IGNORED && graph.getNeighborVertices(i).size() == 1) {
+      aggStat[i] = IGNORED;
+      numNonAggregatedNodes--;
+    }
+}
 
-} // end namespace
+}  // namespace MueLu
 
 #endif /* MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_decl.hpp
index c5bbb6e9c4b9..afa979fbf9e4 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_decl.hpp
@@ -58,65 +58,63 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class IsolatedNodeAggregationAlgorithm class.
-    @brief Ignores isolated nodes during aggregation. Marks the node to be "aggregated" without adding real aggregates for them.
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    The isolated node aggregation algorithm loops over all non-aggregated nodes
-    (with a state different than aggregated or ignored) which have only themselves
-    as neighbor node. The state of these "isolated" nodes is then set to ignored such
-    that they are not considered in the aggregation. This aggregation algorithm should
-    run as one of the last aggregation algorithms in the aggregation method.
-
-    ### Comments ###
-    Only nodes with state different than READY or AGGREGATED are changed to IGNORED.
-    After that, all nodes should have the state AGGREGATED or IGNORED.
-
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class IsolatedNodeAggregationAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class IsolatedNodeAggregationAlgorithm class.
+  @brief Ignores isolated nodes during aggregation. Marks the node to be "aggregated" without adding real aggregates for them.
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  The isolated node aggregation algorithm loops over all non-aggregated nodes
+  (with a state different than aggregated or ignored) which have only themselves
+  as neighbor node. The state of these "isolated" nodes is then set to ignored such
+  that they are not considered in the aggregation. This aggregation algorithm should
+  run as one of the last aggregation algorithms in the aggregation method.
+
+  ### Comments ###
+  Only nodes with state different than READY or AGGREGATED are changed to IGNORED.
+  After that, all nodes should have the state AGGREGATED or IGNORED.
+
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class IsolatedNodeAggregationAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type  = typename LWGraph_kokkos::device_type;
-    using memory_space = typename LWGraph_kokkos::memory_space;
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    IsolatedNodeAggregationAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+ public:
+  using device_type  = typename LWGraph_kokkos::device_type;
+  using memory_space = typename LWGraph_kokkos::memory_space;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~IsolatedNodeAggregationAlgorithm_kokkos() { }
+  //! Constructor.
+  IsolatedNodeAggregationAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //@}
+  //! Destructor.
+  virtual ~IsolatedNodeAggregationAlgorithm_kokkos() {}
 
+  //@}
 
-    //! @name Aggregation methods.
-    //@{
+  //! @name Aggregation methods.
+  //@{
 
-    /*! @brief Local aggregation. */
+  /*! @brief Local aggregation. */
 
-    void BuildAggregates(const ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
-    //@}
+  void BuildAggregates(const ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
+  //@}
 
-    std::string description() const { return "Phase - (isolated)"; }
+  std::string description() const { return "Phase - (isolated)"; }
 
-  }; //class MaxLinkAggregationAlgorithm
+};  // class MaxLinkAggregationAlgorithm
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_SHORT
-#endif // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DECL_HPP
+#endif  // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_def.hpp
index 96ff102a447a..b37aa733ed03 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_def.hpp
@@ -60,41 +60,40 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IsolatedNodeAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(const ParameterList& /* params */,
-                  const LWGraph_kokkos& graph,
-                  Aggregates& /* aggregates */,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggstat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IsolatedNodeAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(const ParameterList& /* params */,
+                    const LWGraph_kokkos& graph,
+                    Aggregates& /* aggregates */,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggstat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
 
-    typename Kokkos::View<unsigned*, device_type>::HostMirror aggstatHost
-      = Kokkos::create_mirror(aggstat);
-    Kokkos::deep_copy(aggstatHost, aggstat);
-    std::vector<unsigned> aggStat;
-    aggStat.resize(aggstatHost.extent(0));
-    for(size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
-      aggStat[idx] = aggstatHost(idx);
-    }
-
-    auto lclLWGraph = graph.getLocalLWGraph();
+  typename Kokkos::View<unsigned*, device_type>::HostMirror aggstatHost = Kokkos::create_mirror(aggstat);
+  Kokkos::deep_copy(aggstatHost, aggstat);
+  std::vector<unsigned> aggStat;
+  aggStat.resize(aggstatHost.extent(0));
+  for (size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
+    aggStat[idx] = aggstatHost(idx);
+  }
 
-    const LO  numRows = graph.GetNodeNumVertices();
+  auto lclLWGraph = graph.getLocalLWGraph();
 
-    // Remove all isolated nodes
-    for (LO i = 0; i < numRows; i++)
-      if (aggStat[i] != AGGREGATED && aggStat[i] != IGNORED && lclLWGraph.getNeighborVertices(i).length == 1) {
-        aggStat[i] = IGNORED;
-        numNonAggregatedNodes--;
-      }
+  const LO numRows = graph.GetNodeNumVertices();
 
-    for(size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
-      aggstatHost(idx) = aggStat[idx];
+  // Remove all isolated nodes
+  for (LO i = 0; i < numRows; i++)
+    if (aggStat[i] != AGGREGATED && aggStat[i] != IGNORED && lclLWGraph.getNeighborVertices(i).length == 1) {
+      aggStat[i] = IGNORED;
+      numNonAggregatedNodes--;
     }
-    Kokkos::deep_copy(aggstat, aggstatHost);
+
+  for (size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
+    aggstatHost(idx) = aggStat[idx];
   }
+  Kokkos::deep_copy(aggstat, aggstatHost);
+}
 
-} // end namespace
+}  // namespace MueLu
 
-#endif // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_decl.hpp
index 6a77eb0d4a29..c7848f6106fa 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_decl.hpp
@@ -63,57 +63,54 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class OnePtAggregationAlgorithm class.
-    @brief Algorithm for coarsening a graph with uncoupled aggregation.
-    keep special marked nodes as singleton node aggregates over all multigrid levels
+/*!
+  @class OnePtAggregationAlgorithm class.
+  @brief Algorithm for coarsening a graph with uncoupled aggregation.
+  keep special marked nodes as singleton node aggregates over all multigrid levels
 
-    @ingroup Aggregation
+  @ingroup Aggregation
 
-    ### Idea ###
-    The user can mark some nodes as ONEPT to build some single node aggregates.
-    This can be very useful for certain applications. We build single node aggregates
-    for nodes with the state ONEPT. Then, the state is changed to ignored.
-    The OnePtAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
+  ### Idea ###
+  The user can mark some nodes as ONEPT to build some single node aggregates.
+  This can be very useful for certain applications. We build single node aggregates
+  for nodes with the state ONEPT. Then, the state is changed to ignored.
+  The OnePtAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
 
-    ### Comments ###
-    Only nodes with state ONEPT are changed to IGNORED.
+  ### Comments ###
+  Only nodes with state ONEPT are changed to IGNORED.
 
-  */
+*/
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class OnePtAggregationAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class OnePtAggregationAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ONEPTAGGREGATIONALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    OnePtAggregationAlgorithm(RCP<const FactoryBase> const &graphFact = Teuchos::null);
+  //! Constructor.
+  OnePtAggregationAlgorithm(RCP<const FactoryBase> const& graphFact = Teuchos::null);
 
-    //! Destructor.
-    virtual ~OnePtAggregationAlgorithm() { }
+  //! Destructor.
+  virtual ~OnePtAggregationAlgorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(Teuchos::ParameterList const& params, GraphBase const& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(Teuchos::ParameterList const & params, GraphBase const & graph, Aggregates & aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+};  // class OnePtAggregationAlgorithm
 
-
-  }; //class OnePtAggregationAlgorithm
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ONEPTAGGREGATIONALGORITHM_SHORT
 #endif /* MUELU_ONEPTAGGREGATIONALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_def.hpp
index 687778c05654..b85b3cc3a927 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_def.hpp
@@ -69,51 +69,47 @@
 namespace MueLu {
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-OnePtAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::OnePtAggregationAlgorithm(RCP<const FactoryBase> const &/* graphFact */)
-{
+OnePtAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::OnePtAggregationAlgorithm(RCP<const FactoryBase> const& /* graphFact */) {
 }
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void OnePtAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const & /* params */, GraphBase const & graph, Aggregates & aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+void OnePtAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const& /* params */, GraphBase const& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
   Monitor m(*this, "BuildAggregates");
 
   const LocalOrdinal nRows = graph.GetNodeNumVertices();
-  const int myRank = graph.GetComm()->getRank();
+  const int myRank         = graph.GetComm()->getRank();
 
   // vertex ids for output
   Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
   Teuchos::ArrayRCP<LocalOrdinal> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
 
   // some internal variables
-  LocalOrdinal nLocalAggregates = aggregates.GetNumAggregates();    // number of local aggregates on current proc
-  LocalOrdinal iNode1  = 0;        // current node
+  LocalOrdinal nLocalAggregates = aggregates.GetNumAggregates();  // number of local aggregates on current proc
+  LocalOrdinal iNode1           = 0;                              // current node
 
   // main loop over all local rows of graph(A)
   while (iNode1 < nRows) {
-
     if (aggStat[iNode1] == ONEPT) {
-
-      aggregates.SetIsRoot(iNode1);    // mark iNode1 as root node for new aggregate 'ag'
+      aggregates.SetIsRoot(iNode1);  // mark iNode1 as root node for new aggregate 'ag'
       std::vector<int> aggList;
       aggList.push_back(iNode1);
       int aggIndex = nLocalAggregates++;
 
       for (size_t k = 0; k < aggList.size(); k++) {
-        aggStat[aggList[k]] = IGNORED;
+        aggStat[aggList[k]]      = IGNORED;
         vertex2AggId[aggList[k]] = aggIndex;
-        procWinner[aggList[k]] = myRank;
+        procWinner[aggList[k]]   = myRank;
       }
       numNonAggregatedNodes -= aggList.size();
     }
 
     iNode1++;
-  } // end while
+  }  // end while
 
   // update aggregate object
   aggregates.SetNumAggregates(nLocalAggregates);
 }
 
-} // end namespace
-
+}  // namespace MueLu
 
 #endif /* MUELU_ONEPTAGGREGATIONALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_decl.hpp
index 1cae818205c2..986c2be498ca 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_decl.hpp
@@ -58,63 +58,60 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class OnePtAggregationAlgorithm class.
-    @brief Algorithm for coarsening a graph with uncoupled aggregation.
-    keep special marked nodes as singleton node aggregates over all multigrid levels
+/*!
+  @class OnePtAggregationAlgorithm class.
+  @brief Algorithm for coarsening a graph with uncoupled aggregation.
+  keep special marked nodes as singleton node aggregates over all multigrid levels
 
-    @ingroup Aggregation
+  @ingroup Aggregation
 
-    ### Idea ###
-    The user can mark some nodes as ONEPT to build some single node aggregates.
-    This can be very useful for certain applications. We build single node aggregates
-    for nodes with the state ONEPT. Then, the state is changed to ignored.
-    The OnePtAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
+  ### Idea ###
+  The user can mark some nodes as ONEPT to build some single node aggregates.
+  This can be very useful for certain applications. We build single node aggregates
+  for nodes with the state ONEPT. Then, the state is changed to ignored.
+  The OnePtAggregationAlgorithm should run before the Phase1AggregationAlgorithm.
 
-    ### Comments ###
-    Only nodes with state ONEPT are changed to IGNORED.
+  ### Comments ###
+  Only nodes with state ONEPT are changed to IGNORED.
 
-  */
+*/
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class OnePtAggregationAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class OnePtAggregationAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type  = typename LWGraph_kokkos::device_type;
-    using memory_space = typename LWGraph_kokkos::memory_space;
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  using device_type  = typename LWGraph_kokkos::device_type;
+  using memory_space = typename LWGraph_kokkos::memory_space;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    OnePtAggregationAlgorithm_kokkos(RCP<const FactoryBase> const &graphFact = Teuchos::null);
+  //! Constructor.
+  OnePtAggregationAlgorithm_kokkos(RCP<const FactoryBase> const& graphFact = Teuchos::null);
 
-    //! Destructor.
-    virtual ~OnePtAggregationAlgorithm_kokkos() { }
+  //! Destructor.
+  virtual ~OnePtAggregationAlgorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(Teuchos::ParameterList const& params,
+                       LWGraph_kokkos const& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(Teuchos::ParameterList const & params,
-                         LWGraph_kokkos const & graph,
-                         Aggregates & aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
-    //@}
+};  // class OnePtAggregationAlgorithm_kokkos
 
-
-  }; //class OnePtAggregationAlgorithm_kokkos
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_SHORT
-#endif // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_def.hpp
index 82dd4881b96c..f016c974ffe4 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_OnePtAggregationAlgorithm_kokkos_def.hpp
@@ -60,71 +60,67 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  OnePtAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::OnePtAggregationAlgorithm_kokkos(RCP<const FactoryBase> const &/* graphFact */)
-  {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+OnePtAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::OnePtAggregationAlgorithm_kokkos(RCP<const FactoryBase> const& /* graphFact */) {
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void OnePtAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(Teuchos::ParameterList const& /* params */,
+                    LWGraph_kokkos const& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggstat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+
+  typename Kokkos::View<unsigned*, device_type>::HostMirror aggstatHost = Kokkos::create_mirror(aggstat);
+  Kokkos::deep_copy(aggstatHost, aggstat);
+  std::vector<unsigned> aggStat;
+  aggStat.resize(aggstatHost.extent(0));
+  for (size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
+    aggStat[idx] = aggstatHost(idx);
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void OnePtAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(Teuchos::ParameterList const & /* params */,
-                  LWGraph_kokkos const & graph,
-                  Aggregates & aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggstat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-
-    typename Kokkos::View<unsigned*, device_type>::HostMirror aggstatHost
-      = Kokkos::create_mirror(aggstat);
-    Kokkos::deep_copy(aggstatHost, aggstat);
-    std::vector<unsigned> aggStat;
-    aggStat.resize(aggstatHost.extent(0));
-    for(size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
-      aggStat[idx] = aggstatHost(idx);
-    }
-
-    const LocalOrdinal nRows = graph.GetNodeNumVertices();
-    const int myRank = graph.GetComm()->getRank();
-
-    // vertex ids for output
-    Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    Teuchos::ArrayRCP<LocalOrdinal> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
-
-    // some internal variables
-    LocalOrdinal nLocalAggregates = aggregates.GetNumAggregates();    // number of local aggregates on current proc
-    LocalOrdinal iNode1  = 0;        // current node
-
-    // main loop over all local rows of graph(A)
-    while (iNode1 < nRows) {
-
-      if (aggStat[iNode1] == ONEPT) {
-
-        aggregates.SetIsRoot(iNode1);    // mark iNode1 as root node for new aggregate 'ag'
-        std::vector<int> aggList;
-        aggList.push_back(iNode1);
-        int aggIndex = nLocalAggregates++;
-
-        // finalize aggregate
-        for (size_t k = 0; k < aggList.size(); k++) {
-          aggStat[aggList[k]] = IGNORED;
-          vertex2AggId[aggList[k]] = aggIndex;
-          procWinner[aggList[k]] = myRank;
-        }
-        numNonAggregatedNodes -= aggList.size();
+  const LocalOrdinal nRows = graph.GetNodeNumVertices();
+  const int myRank         = graph.GetComm()->getRank();
+
+  // vertex ids for output
+  Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  Teuchos::ArrayRCP<LocalOrdinal> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
+
+  // some internal variables
+  LocalOrdinal nLocalAggregates = aggregates.GetNumAggregates();  // number of local aggregates on current proc
+  LocalOrdinal iNode1           = 0;                              // current node
+
+  // main loop over all local rows of graph(A)
+  while (iNode1 < nRows) {
+    if (aggStat[iNode1] == ONEPT) {
+      aggregates.SetIsRoot(iNode1);  // mark iNode1 as root node for new aggregate 'ag'
+      std::vector<int> aggList;
+      aggList.push_back(iNode1);
+      int aggIndex = nLocalAggregates++;
+
+      // finalize aggregate
+      for (size_t k = 0; k < aggList.size(); k++) {
+        aggStat[aggList[k]]      = IGNORED;
+        vertex2AggId[aggList[k]] = aggIndex;
+        procWinner[aggList[k]]   = myRank;
       }
-
-      iNode1++;
-    } // end while
-
-    for(size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
-      aggstatHost(idx) = aggStat[idx];
+      numNonAggregatedNodes -= aggList.size();
     }
-    Kokkos::deep_copy(aggstat, aggstatHost);
 
-    // update aggregate object
-    aggregates.SetNumAggregates(nLocalAggregates);
+    iNode1++;
+  }  // end while
+
+  for (size_t idx = 0; idx < aggstatHost.extent(0); ++idx) {
+    aggstatHost(idx) = aggStat[idx];
   }
+  Kokkos::deep_copy(aggstat, aggstatHost);
+
+  // update aggregate object
+  aggregates.SetNumAggregates(nLocalAggregates);
+}
 
-} // end namespace
+}  // namespace MueLu
 
-#endif // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_decl.hpp
index 8d1af9929a88..b53e449df3b0 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_decl.hpp
@@ -55,67 +55,63 @@
 #include "MueLu_GraphBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class PreserveDirichletAggregationAlgorithm class.
-    @brief Builds one-to-one aggregates for all Dirichlet boundary nodes. For some applications this might
-           be necessary. (default = off)
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Handles Dirichlet boundary nodes with the state Boundary.
-    Depending on the boolean parameter "aggregation: preserve Dirichlet points" one-to-one aggregates
-    with singleton nodes are built for all Dirichlet boundary nodes or the aggregates are just
-    ignored (default behavior). The state of all boundary nodes (state = Boundary)
-    is set to ignored. That means, that these nodes are not considered for further
-    aggregation in the later aggregation phases.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: preserve Dirichlet points | Boolean parameter stating whether Dirichlet boundary nodes shall be aggregated in singleton aggregates (default: false).
-
-    ### Comments ###
-    Only nodes with state BOUNDARY are changed to IGNORED. No other nodes are touched.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class PreserveDirichletAggregationAlgorithm :
-    public MueLu::AggregationAlgorithmBase<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class PreserveDirichletAggregationAlgorithm class.
+  @brief Builds one-to-one aggregates for all Dirichlet boundary nodes. For some applications this might
+         be necessary. (default = off)
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Handles Dirichlet boundary nodes with the state Boundary.
+  Depending on the boolean parameter "aggregation: preserve Dirichlet points" one-to-one aggregates
+  with singleton nodes are built for all Dirichlet boundary nodes or the aggregates are just
+  ignored (default behavior). The state of all boundary nodes (state = Boundary)
+  is set to ignored. That means, that these nodes are not considered for further
+  aggregation in the later aggregation phases.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: preserve Dirichlet points | Boolean parameter stating whether Dirichlet boundary nodes shall be aggregated in singleton aggregates (default: false).
+
+  ### Comments ###
+  Only nodes with state BOUNDARY are changed to IGNORED. No other nodes are touched.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PreserveDirichletAggregationAlgorithm : public MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    PreserveDirichletAggregationAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  PreserveDirichletAggregationAlgorithm(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~PreserveDirichletAggregationAlgorithm() { }
+  //! Destructor.
+  virtual ~PreserveDirichletAggregationAlgorithm() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const Teuchos::ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(const Teuchos::ParameterList& params, const GraphBase& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase - (Dirichlet)"; }
 
-    std::string description() const { return "Phase - (Dirichlet)"; }
+};  // class PreserveDirichletAggregationAlgorithm
 
-  }; //class PreserveDirichletAggregationAlgorithm
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_SHORT
 
-
-
 #endif /* MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_def.hpp
index 5dbd6d0dbf7f..9362342deaba 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_def.hpp
@@ -60,37 +60,37 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PreserveDirichletAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const & params, GraphBase const & graph, Aggregates & aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void PreserveDirichletAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Teuchos::ParameterList const& params, GraphBase const& graph, Aggregates& aggregates, std::vector<unsigned>& aggStat, LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
 
-    bool preserve = params.get<bool>("aggregation: preserve Dirichlet points");
+  bool preserve = params.get<bool>("aggregation: preserve Dirichlet points");
 
-    const LO  numRows = graph.GetNodeNumVertices();
-    const int myRank  = graph.GetComm()->getRank();
+  const LO numRows = graph.GetNodeNumVertices();
+  const int myRank = graph.GetComm()->getRank();
 
-    ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
-    ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()  ->getDataNonConst(0);
+  ArrayRCP<LO> vertex2AggId = aggregates.GetVertex2AggId()->getDataNonConst(0);
+  ArrayRCP<LO> procWinner   = aggregates.GetProcWinner()->getDataNonConst(0);
 
-    LO numLocalAggregates = aggregates.GetNumAggregates();
+  LO numLocalAggregates = aggregates.GetNumAggregates();
 
-    for (LO i = 0; i < numRows; i++)
-      if (aggStat[i] == BOUNDARY) {
-        aggStat[i] = IGNORED;
-        numNonAggregatedNodes--;
+  for (LO i = 0; i < numRows; i++)
+    if (aggStat[i] == BOUNDARY) {
+      aggStat[i] = IGNORED;
+      numNonAggregatedNodes--;
 
-        if (preserve) {
-          aggregates.SetIsRoot(i);
+      if (preserve) {
+        aggregates.SetIsRoot(i);
 
-          vertex2AggId[i] = numLocalAggregates++;
-          procWinner  [i] = myRank;
-        }
+        vertex2AggId[i] = numLocalAggregates++;
+        procWinner[i]   = myRank;
       }
+    }
 
-    // update aggregate object
-    aggregates.SetNumAggregates(numLocalAggregates);
-  }
+  // update aggregate object
+  aggregates.SetNumAggregates(numLocalAggregates);
+}
 
-} // end namespace
+}  // namespace MueLu
 
 #endif /* MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_decl.hpp
index 568889a49cb2..1b5b5697c09c 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_decl.hpp
@@ -58,72 +58,70 @@
 #include "MueLu_LWGraph_kokkos.hpp"
 
 namespace MueLu {
-  /*!
-    @class PreserveDirichletAggregationAlgorithm class.
-    @brief Builds one-to-one aggregates for all Dirichlet boundary nodes. For some applications this might
-           be necessary. (default = off)
-
-    @ingroup Aggregation
-
-    ### Idea ###
-    Handles Dirichlet boundary nodes with the state Boundary.
-    Depending on the boolean parameter "aggregation: preserve Dirichlet points" one-to-one aggregates
-    with singleton nodes are built for all Dirichlet boundary nodes or the aggregates are just
-    ignored (default behavior). The state of all boundary nodes (state = Boundary)
-    is set to ignored. That means, that these nodes are not considered for further
-    aggregation in the later aggregation phases.
-
-    ### Parameters ###
-    Parameter | Meaning
-    ----------|--------
-    aggregation: preserve Dirichlet points | Boolean parameter stating whether Dirichlet boundary nodes shall be aggregated in singleton aggregates (default: false).
-
-    ### Comments ###
-    Only nodes with state BOUNDARY are changed to IGNORED. No other nodes are touched.
-  */
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class PreserveDirichletAggregationAlgorithm_kokkos :
-    public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class PreserveDirichletAggregationAlgorithm class.
+  @brief Builds one-to-one aggregates for all Dirichlet boundary nodes. For some applications this might
+         be necessary. (default = off)
+
+  @ingroup Aggregation
+
+  ### Idea ###
+  Handles Dirichlet boundary nodes with the state Boundary.
+  Depending on the boolean parameter "aggregation: preserve Dirichlet points" one-to-one aggregates
+  with singleton nodes are built for all Dirichlet boundary nodes or the aggregates are just
+  ignored (default behavior). The state of all boundary nodes (state = Boundary)
+  is set to ignored. That means, that these nodes are not considered for further
+  aggregation in the later aggregation phases.
+
+  ### Parameters ###
+  Parameter | Meaning
+  ----------|--------
+  aggregation: preserve Dirichlet points | Boolean parameter stating whether Dirichlet boundary nodes shall be aggregated in singleton aggregates (default: false).
+
+  ### Comments ###
+  Only nodes with state BOUNDARY are changed to IGNORED. No other nodes are touched.
+*/
+
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PreserveDirichletAggregationAlgorithm_kokkos : public MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    using device_type     = typename LWGraph_kokkos::device_type;
-    using execution_space = typename LWGraph_kokkos::execution_space;
-    using memory_space    = typename LWGraph_kokkos::memory_space;
+ public:
+  using device_type     = typename LWGraph_kokkos::device_type;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    PreserveDirichletAggregationAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) { }
+  //! Constructor.
+  PreserveDirichletAggregationAlgorithm_kokkos(const RCP<const FactoryBase>& /* graphFact */ = Teuchos::null) {}
 
-    //! Destructor.
-    virtual ~PreserveDirichletAggregationAlgorithm_kokkos() { }
+  //! Destructor.
+  virtual ~PreserveDirichletAggregationAlgorithm_kokkos() {}
 
-    //@}
+  //@}
 
+  //! @name Aggregation methods.
+  //@{
 
-    //! @name Aggregation methods.
-    //@{
+  /*! @brief Local aggregation. */
 
-    /*! @brief Local aggregation. */
+  void BuildAggregates(const Teuchos::ParameterList& params,
+                       const LWGraph_kokkos& graph,
+                       Aggregates& aggregates,
+                       Kokkos::View<unsigned*, device_type>& aggStat,
+                       LO& numNonAggregatedNodes) const;
+  //@}
 
-    void BuildAggregates(const Teuchos::ParameterList& params,
-                         const LWGraph_kokkos& graph,
-                         Aggregates& aggregates,
-                         Kokkos::View<unsigned*, device_type>& aggStat,
-                         LO& numNonAggregatedNodes) const;
-    //@}
+  std::string description() const { return "Phase - (Dirichlet)"; }
 
-    std::string description() const { return "Phase - (Dirichlet)"; }
+};  // class PreserveDirichletAggregationAlgorithm
 
-  }; //class PreserveDirichletAggregationAlgorithm
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_SHORT
-#endif // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_DECL_HPP
+#endif  // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_def.hpp
index 498640df9c43..b28171a695a3 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_def.hpp
@@ -60,61 +60,61 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PreserveDirichletAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  BuildAggregates(Teuchos::ParameterList const & params,
-                  LWGraph_kokkos const & graph,
-                  Aggregates & aggregates,
-                  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
-                  LO& numNonAggregatedNodes) const {
-    Monitor m(*this, "BuildAggregates");
-    using local_ordinal_type = typename LWGraph_kokkos::local_ordinal_type;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void PreserveDirichletAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    BuildAggregates(Teuchos::ParameterList const& params,
+                    LWGraph_kokkos const& graph,
+                    Aggregates& aggregates,
+                    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>& aggStat,
+                    LO& numNonAggregatedNodes) const {
+  Monitor m(*this, "BuildAggregates");
+  using local_ordinal_type = typename LWGraph_kokkos::local_ordinal_type;
 
-    // Extract parameters and data from:
-    // 1) the parameter list
-    const bool preserve = params.get<bool>("aggregation: preserve Dirichlet points");
+  // Extract parameters and data from:
+  // 1) the parameter list
+  const bool preserve = params.get<bool>("aggregation: preserve Dirichlet points");
 
-    // 2) the amalgamated graph
-    const LO  numNodes  = graph.GetNodeNumVertices();
-    const int myRank    = graph.GetComm()->getRank();
+  // 2) the amalgamated graph
+  const LO numNodes = graph.GetNodeNumVertices();
+  const int myRank  = graph.GetComm()->getRank();
 
-    // 3) the aggregates
-    auto vertex2AggId   = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto procWinner     = aggregates.GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  // 3) the aggregates
+  auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto procWinner   = aggregates.GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
 
-    // A view is needed to count on the fly the current number of local aggregates
-    Kokkos::View<LO, device_type> aggCount("aggCount");
-    if(preserve) {
-      Kokkos::deep_copy(aggCount, aggregates.GetNumAggregates());
-    }
-    Kokkos::parallel_for("MueLu - PreserveDirichlet: tagging ignored nodes",
-                         Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numNodes),
-                         KOKKOS_LAMBDA(const local_ordinal_type nodeIdx) {
-                           if (aggStat(nodeIdx) == BOUNDARY) {
-                             aggStat(nodeIdx) = IGNORED;
-                             const LO aggIdx = Kokkos::atomic_fetch_add(&aggCount(), 1);
+  // A view is needed to count on the fly the current number of local aggregates
+  Kokkos::View<LO, device_type> aggCount("aggCount");
+  if (preserve) {
+    Kokkos::deep_copy(aggCount, aggregates.GetNumAggregates());
+  }
+  Kokkos::parallel_for(
+      "MueLu - PreserveDirichlet: tagging ignored nodes",
+      Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numNodes),
+      KOKKOS_LAMBDA(const local_ordinal_type nodeIdx) {
+        if (aggStat(nodeIdx) == BOUNDARY) {
+          aggStat(nodeIdx) = IGNORED;
+          const LO aggIdx  = Kokkos::atomic_fetch_add(&aggCount(), 1);
 
-                             if (preserve) {
-                               // aggregates.SetIsRoot(nodeIdx);
+          if (preserve) {
+            // aggregates.SetIsRoot(nodeIdx);
 
-                               vertex2AggId(nodeIdx, 0) = aggIdx;
-                               procWinner(nodeIdx, 0)   = myRank;
-                             }
-                           }
-                         });
-    typename Kokkos::View<LO, device_type>::HostMirror aggCount_h
-      = Kokkos::create_mirror_view(aggCount);
-    Kokkos::deep_copy(aggCount_h, aggCount);
-    // In this phase the number of new aggregates is the same
-    // as the number of newly aggregated nodes.
-    numNonAggregatedNodes -= (aggCount_h() - aggregates.GetNumAggregates());
+            vertex2AggId(nodeIdx, 0) = aggIdx;
+            procWinner(nodeIdx, 0)   = myRank;
+          }
+        }
+      });
+  typename Kokkos::View<LO, device_type>::HostMirror aggCount_h = Kokkos::create_mirror_view(aggCount);
+  Kokkos::deep_copy(aggCount_h, aggCount);
+  // In this phase the number of new aggregates is the same
+  // as the number of newly aggregated nodes.
+  numNonAggregatedNodes -= (aggCount_h() - aggregates.GetNumAggregates());
 
-    // update aggregate object
-    if(preserve) {
-      aggregates.SetNumAggregates(aggCount_h());
-    }
+  // update aggregate object
+  if (preserve) {
+    aggregates.SetNumAggregates(aggCount_h());
   }
+}
 
-} // end namespace
+}  // namespace MueLu
 
-#endif // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
+#endif  // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_decl.hpp
index 21dd4ab2a9bb..daee942895a6 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_decl.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_UNCOUPLEDAGGREGATIONFACTORY_DECL_HPP_
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_DECL_HPP_
 
-
 #include <Xpetra_Map_fwd.hpp>
 #include <Xpetra_Vector_fwd.hpp>
 
@@ -137,14 +136,14 @@ namespace MueLu {
     | Aggregates   | UncoupledAggregationFactory   | Container class with aggregation information. See also Aggregates.
 */
 
-template<class LocalOrdinal = DefaultLocalOrdinal,
-         class GlobalOrdinal = DefaultGlobalOrdinal,
-         class Node = DefaultNode>
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
 class UncoupledAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_UNCOUPLEDAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-public:
+ public:
   //! @name Constructors/Destructors.
   //@{
 
@@ -152,7 +151,7 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   UncoupledAggregationFactory();
 
   //! Destructor.
-  virtual ~UncoupledAggregationFactory() { }
+  virtual ~UncoupledAggregationFactory() {}
 
   RCP<const ParameterList> GetValidParameterList() const;
 
@@ -169,16 +168,16 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   }
   // deprecated
   void SetMaxNeighAlreadySelected(int maxNeighAlreadySelected) {
-    SetParameter("aggregation: max selected neighbors", ParameterEntry(Teuchos::as<LocalOrdinal>(maxNeighAlreadySelected))); // revalidate
+    SetParameter("aggregation: max selected neighbors", ParameterEntry(Teuchos::as<LocalOrdinal>(maxNeighAlreadySelected)));  // revalidate
   }
   // deprecated
   void SetMinNodesPerAggregate(int minNodesPerAggregate) {
-    SetParameter("aggregation: min agg size", ParameterEntry(Teuchos::as<LocalOrdinal>(minNodesPerAggregate))); // revalidate
+    SetParameter("aggregation: min agg size", ParameterEntry(Teuchos::as<LocalOrdinal>(minNodesPerAggregate)));  // revalidate
   }
   // set information about 1-node aggregates (map name and generating factory)
   void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
-    SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name))); // revalidate
-    SetFactory("OnePt aggregate map factory",mapFact);
+    SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name)));  // revalidate
+    SetFactory("OnePt aggregate map factory", mapFact);
   }
 
   // deprecated
@@ -202,7 +201,7 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   //! Input
   //@{
 
-  void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
   //@}
 
@@ -210,7 +209,7 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   //@{
 
   /*! @brief Build aggregates. */
-  void Build(Level &currentLevel) const;
+  void Build(Level& currentLevel) const;
 
   //@}
 
@@ -218,14 +217,13 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   //@{
 
   /*! @brief Append a new aggregation algorithm to list of aggregation algorithms */
-  //void Append(const RCP<MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> > & alg);
+  // void Append(const RCP<MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> > & alg);
 
   /*! @brief Remove all aggregation algorithms from list */
-  //void ClearAggregationAlgorithms() { algos_.clear(); }
+  // void ClearAggregationAlgorithms() { algos_.clear(); }
   //@}
 
-private:
-
+ private:
   //! aggregation algorithms
   // will be filled in Build routine
   mutable std::vector<RCP<MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> > > algos_;
@@ -235,9 +233,9 @@ class UncoupledAggregationFactory : public SingleLevelFactoryBase {
   //! if false, no change in aggregation algorithms is possible any more
   mutable bool bDefinitionPhase_;
 
-}; // class UncoupledAggregationFactory
+};  // class UncoupledAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_UNCOUPLEDAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_def.hpp
index d93dc6dbcdb9..9c73ad85fe25 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_def.hpp
@@ -72,239 +72,233 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::UncoupledAggregationFactory()
-  : bDefinitionPhase_(true)
-  { }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::UncoupledAggregationFactory()
+  : bDefinitionPhase_(true) {}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    // Aggregation parameters (used in aggregation algorithms)
-    // TODO introduce local member function for each aggregation algorithm such that each aggregation algorithm can define its own parameters
+  // Aggregation parameters (used in aggregation algorithms)
+  // TODO introduce local member function for each aggregation algorithm such that each aggregation algorithm can define its own parameters
 
-    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+  typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: max agg size");
-    SET_VALID_ENTRY("aggregation: min agg size");
-    SET_VALID_ENTRY("aggregation: max selected neighbors");
-    SET_VALID_ENTRY("aggregation: ordering");
-    validParamList->getEntry("aggregation: ordering").setValidator(
-      rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
-    SET_VALID_ENTRY("aggregation: enable phase 1");
-    SET_VALID_ENTRY("aggregation: enable phase 2a");
-    SET_VALID_ENTRY("aggregation: enable phase 2b");
-    SET_VALID_ENTRY("aggregation: enable phase 3");
-    SET_VALID_ENTRY("aggregation: match ML phase1");
-    SET_VALID_ENTRY("aggregation: match ML phase2a");
-    SET_VALID_ENTRY("aggregation: match ML phase2b");
-    SET_VALID_ENTRY("aggregation: phase2a agg factor");
-    SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
-    SET_VALID_ENTRY("aggregation: allow user-specified singletons");
-    SET_VALID_ENTRY("aggregation: use interface aggregation");
-    SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
-    SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
-    SET_VALID_ENTRY("aggregation: compute aggregate qualities");
-    SET_VALID_ENTRY("aggregation: phase 1 algorithm");
-#undef  SET_VALID_ENTRY
-
-    // general variables needed in AggregationFactory
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
-    validParamList->set< RCP<const FactoryBase> >("AggregateQualities", null, "Generating factory for variable \'AggregateQualities\'");
-
-    // special variables necessary for OnePtAggregationAlgorithm
-    validParamList->set< std::string >           ("OnePt aggregate map name",         "", "Name of input map for single node aggregates. (default='')");
-    validParamList->set< std::string >           ("OnePt aggregate map factory",      "", "Generating factory of (DOF) map for single node aggregates.");
-    //validParamList->set< RCP<const FactoryBase> >("OnePt aggregate map factory",    NoFactory::getRCP(), "Generating factory of (DOF) map for single node aggregates.");
-
-    // InterfaceAggregation parameters
-    //validParamList->set< bool >                  ("aggregation: use interface aggregation", "false", "Flag to trigger aggregation along an interface using specified aggregate seeds.");
-    validParamList->set< std::string >           ("Interface aggregate map name",                "", "Name of input map for interface aggregates. (default='')");
-    validParamList->set< std::string >           ("Interface aggregate map factory",             "", "Generating factory of (DOF) map for interface aggregates.");
-    validParamList->set<RCP<const FactoryBase> > ("nodeOnInterface", Teuchos::null, "Array specifying whether or not a node is on the interface (1 or 0).");
-
-    return validParamList;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "Graph");
-    Input(currentLevel, "DofsPerNode");
-
-    const ParameterList& pL = GetParameterList();
-
-    // request special data necessary for OnePtAggregationAlgorithm
-    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-    if (mapOnePtName.length() > 0) {
-      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-        currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
-      } else {
-        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-        currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
-      }
+  SET_VALID_ENTRY("aggregation: max agg size");
+  SET_VALID_ENTRY("aggregation: min agg size");
+  SET_VALID_ENTRY("aggregation: max selected neighbors");
+  SET_VALID_ENTRY("aggregation: ordering");
+  validParamList->getEntry("aggregation: ordering").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
+  SET_VALID_ENTRY("aggregation: enable phase 1");
+  SET_VALID_ENTRY("aggregation: enable phase 2a");
+  SET_VALID_ENTRY("aggregation: enable phase 2b");
+  SET_VALID_ENTRY("aggregation: enable phase 3");
+  SET_VALID_ENTRY("aggregation: match ML phase1");
+  SET_VALID_ENTRY("aggregation: match ML phase2a");
+  SET_VALID_ENTRY("aggregation: match ML phase2b");
+  SET_VALID_ENTRY("aggregation: phase2a agg factor");
+  SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
+  SET_VALID_ENTRY("aggregation: allow user-specified singletons");
+  SET_VALID_ENTRY("aggregation: use interface aggregation");
+  SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
+  SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
+  SET_VALID_ENTRY("aggregation: compute aggregate qualities");
+  SET_VALID_ENTRY("aggregation: phase 1 algorithm");
+#undef SET_VALID_ENTRY
+
+  // general variables needed in AggregationFactory
+  validParamList->set<RCP<const FactoryBase>>("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase>>("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+  validParamList->set<RCP<const FactoryBase>>("AggregateQualities", null, "Generating factory for variable \'AggregateQualities\'");
+
+  // special variables necessary for OnePtAggregationAlgorithm
+  validParamList->set<std::string>("OnePt aggregate map name", "", "Name of input map for single node aggregates. (default='')");
+  validParamList->set<std::string>("OnePt aggregate map factory", "", "Generating factory of (DOF) map for single node aggregates.");
+  // validParamList->set< RCP<const FactoryBase> >("OnePt aggregate map factory",    NoFactory::getRCP(), "Generating factory of (DOF) map for single node aggregates.");
+
+  // InterfaceAggregation parameters
+  // validParamList->set< bool >                  ("aggregation: use interface aggregation", "false", "Flag to trigger aggregation along an interface using specified aggregate seeds.");
+  validParamList->set<std::string>("Interface aggregate map name", "", "Name of input map for interface aggregates. (default='')");
+  validParamList->set<std::string>("Interface aggregate map factory", "", "Generating factory of (DOF) map for interface aggregates.");
+  validParamList->set<RCP<const FactoryBase>>("nodeOnInterface", Teuchos::null, "Array specifying whether or not a node is on the interface (1 or 0).");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Graph");
+  Input(currentLevel, "DofsPerNode");
+
+  const ParameterList& pL = GetParameterList();
+
+  // request special data necessary for OnePtAggregationAlgorithm
+  std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+  if (mapOnePtName.length() > 0) {
+    std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+    if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+      currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
+    } else {
+      RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+      currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
     }
+  }
 
-    // request special data necessary for InterfaceAggregation
-    if (pL.get<bool>("aggregation: use interface aggregation")       == true){
-      if(currentLevel.GetLevelID() == 0) {
-        if(currentLevel.IsAvailable("nodeOnInterface", NoFactory::get())) {
-          currentLevel.DeclareInput("nodeOnInterface", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("nodeOnInterface", NoFactory::get()),
-                                       Exceptions::RuntimeError,
-                                       "nodeOnInterface was not provided by the user on level0!");
-        }
+  // request special data necessary for InterfaceAggregation
+  if (pL.get<bool>("aggregation: use interface aggregation") == true) {
+    if (currentLevel.GetLevelID() == 0) {
+      if (currentLevel.IsAvailable("nodeOnInterface", NoFactory::get())) {
+        currentLevel.DeclareInput("nodeOnInterface", NoFactory::get(), this);
       } else {
-        Input(currentLevel, "nodeOnInterface");
+        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("nodeOnInterface", NoFactory::get()),
+                                   Exceptions::RuntimeError,
+                                   "nodeOnInterface was not provided by the user on level0!");
       }
-    }
-
-    if (pL.get<bool>("aggregation: compute aggregate qualities")) {
-	Input(currentLevel, "AggregateQualities");
+    } else {
+      Input(currentLevel, "nodeOnInterface");
     }
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    ParameterList pL = GetParameterList();
-    bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
-
-    if (pL.get<int>("aggregation: max agg size") == -1)
-      pL.set("aggregation: max agg size", INT_MAX);
-
-    // define aggregation algorithms
-    RCP<const FactoryBase> graphFact = GetFactory("Graph");
-
-    // TODO Can we keep different aggregation algorithms over more Build calls?
-    algos_.clear();
-    algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm(graphFact)));
-    if (pL.get<bool>("aggregation: use interface aggregation")       == true)   algos_.push_back(rcp(new InterfaceAggregationAlgorithm         (graphFact)));
-    if (pL.get<bool>("aggregation: allow user-specified singletons") == true)   algos_.push_back(rcp(new OnePtAggregationAlgorithm             (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 1" )                 == true)   algos_.push_back(rcp(new AggregationPhase1Algorithm            (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 2a")                 == true)   algos_.push_back(rcp(new AggregationPhase2aAlgorithm           (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 2b")                 == true)   algos_.push_back(rcp(new AggregationPhase2bAlgorithm           (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 3" )                 == true)   algos_.push_back(rcp(new AggregationPhase3Algorithm            (graphFact)));
-
-      // TODO: remove old aggregation mode
-      //if (pL.get<bool>("UseOnePtAggregationAlgorithm")             == true)   algos_.push_back(rcp(new OnePtAggregationAlgorithm             (graphFact)));
-      //if (pL.get<bool>("UseUncoupledAggregationAlgorithm")         == true)   algos_.push_back(rcp(new AggregationPhase1Algorithm            (graphFact)));
-      //if (pL.get<bool>("UseMaxLinkAggregationAlgorithm")           == true)   algos_.push_back(rcp(new MaxLinkAggregationAlgorithm           (graphFact)));
-      //if (pL.get<bool>("UseEmergencyAggregationAlgorithm")         == true)   algos_.push_back(rcp(new EmergencyAggregationAlgorithm         (graphFact)));
-
-    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-    RCP<Map> OnePtMap = Teuchos::null;
-    if (mapOnePtName.length()) {
-      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-        OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, NoFactory::get());
-      } else {
-        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-        OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, mapOnePtFact.get());
-      }
+  if (pL.get<bool>("aggregation: compute aggregate qualities")) {
+    Input(currentLevel, "AggregateQualities");
+  }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  ParameterList pL  = GetParameterList();
+  bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
+
+  if (pL.get<int>("aggregation: max agg size") == -1)
+    pL.set("aggregation: max agg size", INT_MAX);
+
+  // define aggregation algorithms
+  RCP<const FactoryBase> graphFact = GetFactory("Graph");
+
+  // TODO Can we keep different aggregation algorithms over more Build calls?
+  algos_.clear();
+  algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm(graphFact)));
+  if (pL.get<bool>("aggregation: use interface aggregation") == true) algos_.push_back(rcp(new InterfaceAggregationAlgorithm(graphFact)));
+  if (pL.get<bool>("aggregation: allow user-specified singletons") == true) algos_.push_back(rcp(new OnePtAggregationAlgorithm(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 1") == true) algos_.push_back(rcp(new AggregationPhase1Algorithm(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 2a") == true) algos_.push_back(rcp(new AggregationPhase2aAlgorithm(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 2b") == true) algos_.push_back(rcp(new AggregationPhase2bAlgorithm(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 3") == true) algos_.push_back(rcp(new AggregationPhase3Algorithm(graphFact)));
+
+  // TODO: remove old aggregation mode
+  // if (pL.get<bool>("UseOnePtAggregationAlgorithm")             == true)   algos_.push_back(rcp(new OnePtAggregationAlgorithm             (graphFact)));
+  // if (pL.get<bool>("UseUncoupledAggregationAlgorithm")         == true)   algos_.push_back(rcp(new AggregationPhase1Algorithm            (graphFact)));
+  // if (pL.get<bool>("UseMaxLinkAggregationAlgorithm")           == true)   algos_.push_back(rcp(new MaxLinkAggregationAlgorithm           (graphFact)));
+  // if (pL.get<bool>("UseEmergencyAggregationAlgorithm")         == true)   algos_.push_back(rcp(new EmergencyAggregationAlgorithm         (graphFact)));
+
+  std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+  RCP<Map> OnePtMap        = Teuchos::null;
+  if (mapOnePtName.length()) {
+    std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+    if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+      OnePtMap = currentLevel.Get<RCP<Map>>(mapOnePtName, NoFactory::get());
+    } else {
+      RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+      OnePtMap                            = currentLevel.Get<RCP<Map>>(mapOnePtName, mapOnePtFact.get());
     }
+  }
 
-    // Set map for interface aggregates
-    std::string mapInterfaceName = pL.get<std::string>("Interface aggregate map name");
-    RCP<Map> InterfaceMap = Teuchos::null;
+  // Set map for interface aggregates
+  std::string mapInterfaceName = pL.get<std::string>("Interface aggregate map name");
+  RCP<Map> InterfaceMap        = Teuchos::null;
 
-    RCP<const GraphBase> graph = Get< RCP<GraphBase> >(currentLevel, "Graph");
+  RCP<const GraphBase> graph = Get<RCP<GraphBase>>(currentLevel, "Graph");
 
-    // Build
-    RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("UC");
+  // Build
+  RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("UC");
 
-    const LO numRows = graph->GetNodeNumVertices();
+  const LO numRows = graph->GetNodeNumVertices();
 
-    // construct aggStat information
-    std::vector<unsigned> aggStat(numRows, READY);
+  // construct aggStat information
+  std::vector<unsigned> aggStat(numRows, READY);
 
-    // interface
-    if (pL.get<bool>("aggregation: use interface aggregation")       == true){
-      Teuchos::Array<LO> nodeOnInterface = Get<Array<LO>>(currentLevel,"nodeOnInterface");
-      for (LO i = 0; i < numRows; i++) {
-        if (nodeOnInterface[i])
-          aggStat[i] = INTERFACE;
-      }
+  // interface
+  if (pL.get<bool>("aggregation: use interface aggregation") == true) {
+    Teuchos::Array<LO> nodeOnInterface = Get<Array<LO>>(currentLevel, "nodeOnInterface");
+    for (LO i = 0; i < numRows; i++) {
+      if (nodeOnInterface[i])
+        aggStat[i] = INTERFACE;
     }
+  }
 
-    ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
-    if (dirichletBoundaryMap != Teuchos::null)
-      for (LO i = 0; i < numRows; i++)
-        if (dirichletBoundaryMap[i] == true)
-          aggStat[i] = BOUNDARY;
-
-    LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
-    GO indexBase = graph->GetDomainMap()->getIndexBase();
-    if (OnePtMap != Teuchos::null) {
-      for (LO i = 0; i < numRows; i++) {
-        // reconstruct global row id (FIXME only works for contiguous maps)
-        GO grid = (graph->GetDomainMap()->getGlobalElement(i)-indexBase) * nDofsPerNode + indexBase;
-
-        for (LO kr = 0; kr < nDofsPerNode; kr++)
-          if (OnePtMap->isNodeGlobalElement(grid + kr))
-            aggStat[i] = ONEPT;
-      }
+  ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
+  if (dirichletBoundaryMap != Teuchos::null)
+    for (LO i = 0; i < numRows; i++)
+      if (dirichletBoundaryMap[i] == true)
+        aggStat[i] = BOUNDARY;
+
+  LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
+  GO indexBase    = graph->GetDomainMap()->getIndexBase();
+  if (OnePtMap != Teuchos::null) {
+    for (LO i = 0; i < numRows; i++) {
+      // reconstruct global row id (FIXME only works for contiguous maps)
+      GO grid = (graph->GetDomainMap()->getGlobalElement(i) - indexBase) * nDofsPerNode + indexBase;
+
+      for (LO kr = 0; kr < nDofsPerNode; kr++)
+        if (OnePtMap->isNodeGlobalElement(grid + kr))
+          aggStat[i] = ONEPT;
     }
+  }
 
-
-
-    const RCP<const Teuchos::Comm<int> > comm = graph->GetComm();
-    GO numGlobalRows = 0;
-    if (IsPrint(Statistics1))
-      MueLu_sumAll(comm, as<GO>(numRows), numGlobalRows);
-
-    LO numNonAggregatedNodes = numRows;
-    GO numGlobalAggregatedPrev = 0, numGlobalAggsPrev = 0;
-    for (size_t a = 0; a < algos_.size(); a++) {
-      std::string phase = algos_[a]->description();
-      SubFactoryMonitor sfm(*this, "Algo " + phase, currentLevel);
-
-      int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
-      algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-      algos_[a]->SetProcRankVerbose(oldRank);
-
-      if (IsPrint(Statistics1)) {
-        GO numLocalAggregated = numRows - numNonAggregatedNodes, numGlobalAggregated = 0;
-        GO numLocalAggs       = aggregates->GetNumAggregates(),  numGlobalAggs = 0;
-        MueLu_sumAll(comm, numLocalAggregated, numGlobalAggregated);
-        MueLu_sumAll(comm, numLocalAggs,       numGlobalAggs);
-
-        double aggPercent = 100*as<double>(numGlobalAggregated)/as<double>(numGlobalRows);
-        if (aggPercent > 99.99 && aggPercent < 100.00) {
-          // Due to round off (for instance, for 140465733/140466897), we could
-          // get 100.00% display even if there are some remaining nodes. This
-          // is bad from the users point of view. It is much better to change
-          // it to display 99.99%.
-          aggPercent = 99.99;
-        }
-        GetOStream(Statistics1) << "  aggregated : " << (numGlobalAggregated - numGlobalAggregatedPrev) << " (phase), " << std::fixed
-                                   << std::setprecision(2) << numGlobalAggregated << "/" << numGlobalRows << " [" << aggPercent << "%] (total)\n"
-                                   << "  remaining  : " << numGlobalRows - numGlobalAggregated << "\n"
-                                   << "  aggregates : " << numGlobalAggs-numGlobalAggsPrev << " (phase), " << numGlobalAggs << " (total)" << std::endl;
-        numGlobalAggregatedPrev = numGlobalAggregated;
-        numGlobalAggsPrev       = numGlobalAggs;
+  const RCP<const Teuchos::Comm<int>> comm = graph->GetComm();
+  GO numGlobalRows                         = 0;
+  if (IsPrint(Statistics1))
+    MueLu_sumAll(comm, as<GO>(numRows), numGlobalRows);
+
+  LO numNonAggregatedNodes   = numRows;
+  GO numGlobalAggregatedPrev = 0, numGlobalAggsPrev = 0;
+  for (size_t a = 0; a < algos_.size(); a++) {
+    std::string phase = algos_[a]->description();
+    SubFactoryMonitor sfm(*this, "Algo " + phase, currentLevel);
+
+    int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
+    algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+    algos_[a]->SetProcRankVerbose(oldRank);
+
+    if (IsPrint(Statistics1)) {
+      GO numLocalAggregated = numRows - numNonAggregatedNodes, numGlobalAggregated = 0;
+      GO numLocalAggs = aggregates->GetNumAggregates(), numGlobalAggs = 0;
+      MueLu_sumAll(comm, numLocalAggregated, numGlobalAggregated);
+      MueLu_sumAll(comm, numLocalAggs, numGlobalAggs);
+
+      double aggPercent = 100 * as<double>(numGlobalAggregated) / as<double>(numGlobalRows);
+      if (aggPercent > 99.99 && aggPercent < 100.00) {
+        // Due to round off (for instance, for 140465733/140466897), we could
+        // get 100.00% display even if there are some remaining nodes. This
+        // is bad from the users point of view. It is much better to change
+        // it to display 99.99%.
+        aggPercent = 99.99;
       }
+      GetOStream(Statistics1) << "  aggregated : " << (numGlobalAggregated - numGlobalAggregatedPrev) << " (phase), " << std::fixed
+                              << std::setprecision(2) << numGlobalAggregated << "/" << numGlobalRows << " [" << aggPercent << "%] (total)\n"
+                              << "  remaining  : " << numGlobalRows - numGlobalAggregated << "\n"
+                              << "  aggregates : " << numGlobalAggs - numGlobalAggsPrev << " (phase), " << numGlobalAggs << " (total)" << std::endl;
+      numGlobalAggregatedPrev = numGlobalAggregated;
+      numGlobalAggsPrev       = numGlobalAggs;
     }
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError, "MueLu::UncoupledAggregationFactory::Build: Leftover nodes found! Error!");
+  TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError, "MueLu::UncoupledAggregationFactory::Build: Leftover nodes found! Error!");
 
-    aggregates->AggregatesCrossProcessors(false);
-    aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
+  aggregates->AggregatesCrossProcessors(false);
+  aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
 
-    Set(currentLevel, "Aggregates", aggregates);
-
-    if (pL.get<bool>("aggregation: compute aggregate qualities")) {
-      RCP<Xpetra::MultiVector<DefaultScalar,LO,GO,Node>> aggQualities = Get<RCP<Xpetra::MultiVector<DefaultScalar,LO,GO,Node>>>(currentLevel, "AggregateQualities");
-    }
+  Set(currentLevel, "Aggregates", aggregates);
 
+  if (pL.get<bool>("aggregation: compute aggregate qualities")) {
+    RCP<Xpetra::MultiVector<DefaultScalar, LO, GO, Node>> aggQualities = Get<RCP<Xpetra::MultiVector<DefaultScalar, LO, GO, Node>>>(currentLevel, "AggregateQualities");
   }
+}
 
-} //namespace MueLu
-
+}  // namespace MueLu
 
 #endif /* MUELU_UNCOUPLEDAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_decl.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_decl.hpp
index 5bf4100d95bb..4c8e70426c4d 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_decl.hpp
@@ -133,107 +133,106 @@ namespace MueLu {
     | Aggregates   | UncoupledAggregationFactory   | Container class with aggregation information. See also Aggregates.
 */
 
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class UncoupledAggregationFactory_kokkos : public SingleLevelFactoryBase {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class UncoupledAggregationFactory_kokkos : public SingleLevelFactoryBase {
 #undef MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    UncoupledAggregationFactory_kokkos();
+  //! Constructor.
+  UncoupledAggregationFactory_kokkos();
 
-    //! Destructor.
-    virtual ~UncoupledAggregationFactory_kokkos() { }
+  //! Destructor.
+  virtual ~UncoupledAggregationFactory_kokkos() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Set/get methods.
-    //@{
+  //! @name Set/get methods.
+  //@{
 
-    // Options shared by all aggregation algorithms
+  // Options shared by all aggregation algorithms
 
-    // deprecated
-    void SetOrdering(const std::string& ordering) {
-      SetParameter("aggregation: ordering", ParameterEntry(ordering));
-    }
-    // deprecated
-    void SetMaxNeighAlreadySelected(int maxNeighAlreadySelected) {
-      SetParameter("aggregation: max selected neighbors", ParameterEntry(Teuchos::as<LocalOrdinal>(maxNeighAlreadySelected))); // revalidate
-    }
-    // deprecated
-    void SetMinNodesPerAggregate(int minNodesPerAggregate) {
-      SetParameter("aggregation: min agg size", ParameterEntry(Teuchos::as<LocalOrdinal>(minNodesPerAggregate))); // revalidate
-    }
-    // set information about 1-node aggregates (map name and generating factory)
-    void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
-      SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name))); // revalidate
-      SetFactory("OnePt aggregate map factory",mapFact);
-    }
+  // deprecated
+  void SetOrdering(const std::string& ordering) {
+    SetParameter("aggregation: ordering", ParameterEntry(ordering));
+  }
+  // deprecated
+  void SetMaxNeighAlreadySelected(int maxNeighAlreadySelected) {
+    SetParameter("aggregation: max selected neighbors", ParameterEntry(Teuchos::as<LocalOrdinal>(maxNeighAlreadySelected)));  // revalidate
+  }
+  // deprecated
+  void SetMinNodesPerAggregate(int minNodesPerAggregate) {
+    SetParameter("aggregation: min agg size", ParameterEntry(Teuchos::as<LocalOrdinal>(minNodesPerAggregate)));  // revalidate
+  }
+  // set information about 1-node aggregates (map name and generating factory)
+  void SetOnePtMapName(const std::string name, Teuchos::RCP<const FactoryBase> mapFact) {
+    SetParameter("OnePt aggregate map name", ParameterEntry(std::string(name)));  // revalidate
+    SetFactory("OnePt aggregate map factory", mapFact);
+  }
 
-    // deprecated
-    const std::string& GetOrdering() const {
-      const ParameterList& pL = GetParameterList();
-      return pL.get<std::string>("aggregation: ordering");
-    }
-    // deprecated
-    int GetMaxNeighAlreadySelected() const {
-      const ParameterList& pL = GetParameterList();
-      return Teuchos::as<int>(pL.get<LocalOrdinal>("aggregation: max selected neighbors"));
-    }
-    // deprecated
-    int GetMinNodesPerAggregate() const {
-      const ParameterList& pL = GetParameterList();
-      return Teuchos::as<int>(pL.get<LocalOrdinal>("aggregation: min agg size"));
-    }
+  // deprecated
+  const std::string& GetOrdering() const {
+    const ParameterList& pL = GetParameterList();
+    return pL.get<std::string>("aggregation: ordering");
+  }
+  // deprecated
+  int GetMaxNeighAlreadySelected() const {
+    const ParameterList& pL = GetParameterList();
+    return Teuchos::as<int>(pL.get<LocalOrdinal>("aggregation: max selected neighbors"));
+  }
+  // deprecated
+  int GetMinNodesPerAggregate() const {
+    const ParameterList& pL = GetParameterList();
+    return Teuchos::as<int>(pL.get<LocalOrdinal>("aggregation: min agg size"));
+  }
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*! @brief Build aggregates. */
-    void Build(Level &currentLevel) const;
+  /*! @brief Build aggregates. */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Definition methods
-    //@{
+  //! @name Definition methods
+  //@{
 
-    /*! @brief Append a new aggregation algorithm to list of aggregation algorithms */
-    //void Append(const RCP<MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> > & alg);
+  /*! @brief Append a new aggregation algorithm to list of aggregation algorithms */
+  // void Append(const RCP<MueLu::AggregationAlgorithmBase<LocalOrdinal, GlobalOrdinal, Node> > & alg);
 
-    /*! @brief Remove all aggregation algorithms from list */
-    //void ClearAggregationAlgorithms() { algos_.clear(); }
-    //@}
+  /*! @brief Remove all aggregation algorithms from list */
+  // void ClearAggregationAlgorithms() { algos_.clear(); }
+  //@}
 
-  private:
+ private:
+  //! aggregation algorithms
+  // will be filled in Build routine
+  mutable std::vector<RCP<MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> > > algos_;
 
-    //! aggregation algorithms
-    // will be filled in Build routine
-    mutable std::vector<RCP<MueLu::AggregationAlgorithmBase_kokkos<LocalOrdinal, GlobalOrdinal, Node> > > algos_;
+  //! boolean flag: definition phase
+  //! if true, the aggregation algorithms still can be set and changed.
+  //! if false, no change in aggregation algorithms is possible any more
+  mutable bool bDefinitionPhase_;
 
-    //! boolean flag: definition phase
-    //! if true, the aggregation algorithms still can be set and changed.
-    //! if false, no change in aggregation algorithms is possible any more
-    mutable bool bDefinitionPhase_;
+};  // class UncoupledAggregationFactory_kokkos
 
-  }; // class UncoupledAggregationFactory_kokkos
-
-}
+}  // namespace MueLu
 
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_SHORT
-#endif // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_def.hpp b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_def.hpp
index 0c0daed773ac..ba7f35169768 100644
--- a/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Graph/UncoupledAggregation/MueLu_UncoupledAggregationFactory_kokkos_def.hpp
@@ -75,338 +75,328 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::UncoupledAggregationFactory_kokkos()
-  : bDefinitionPhase_(true)
-  { }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::UncoupledAggregationFactory_kokkos()
+  : bDefinitionPhase_(true) {}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    // Aggregation parameters (used in aggregation algorithms)
-    // TODO introduce local member function for each aggregation algorithm such that each aggregation algorithm can define its own parameters
+  // Aggregation parameters (used in aggregation algorithms)
+  // TODO introduce local member function for each aggregation algorithm such that each aggregation algorithm can define its own parameters
 
-    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+  typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: max agg size");
-    SET_VALID_ENTRY("aggregation: min agg size");
-    SET_VALID_ENTRY("aggregation: max selected neighbors");
-    SET_VALID_ENTRY("aggregation: ordering");
-    validParamList->getEntry("aggregation: ordering").setValidator(
-      rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
-    SET_VALID_ENTRY("aggregation: deterministic");
-    SET_VALID_ENTRY("aggregation: coloring algorithm");
-    SET_VALID_ENTRY("aggregation: enable phase 1");
-    SET_VALID_ENTRY("aggregation: enable phase 2a");
-    SET_VALID_ENTRY("aggregation: enable phase 2b");
-    SET_VALID_ENTRY("aggregation: enable phase 3");
-    SET_VALID_ENTRY("aggregation: match ML phase1");
-    SET_VALID_ENTRY("aggregation: match ML phase2a");
-    SET_VALID_ENTRY("aggregation: match ML phase2b");
-    SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
-    SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
-    SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
-    SET_VALID_ENTRY("aggregation: allow user-specified singletons");
-    SET_VALID_ENTRY("aggregation: phase 1 algorithm");
-#undef  SET_VALID_ENTRY
-
-    // general variables needed in AggregationFactory
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
-
-    // special variables necessary for OnePtAggregationAlgorithm
-    validParamList->set< std::string >           ("OnePt aggregate map name",         "", "Name of input map for single node aggregates. (default='')");
-    validParamList->set< std::string >           ("OnePt aggregate map factory",      "", "Generating factory of (DOF) map for single node aggregates.");
-    //validParamList->set< RCP<const FactoryBase> >("OnePt aggregate map factory",    NoFactory::getRCP(), "Generating factory of (DOF) map for single node aggregates.");
-
-    return validParamList;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "Graph");
-    Input(currentLevel, "DofsPerNode");
-
-    const ParameterList& pL = GetParameterList();
-
-    // request special data necessary for OnePtAggregationAlgorithm
-    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-    if (mapOnePtName.length() > 0) {
-      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-        currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
-      } else {
-        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-        currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
-      }
+  SET_VALID_ENTRY("aggregation: max agg size");
+  SET_VALID_ENTRY("aggregation: min agg size");
+  SET_VALID_ENTRY("aggregation: max selected neighbors");
+  SET_VALID_ENTRY("aggregation: ordering");
+  validParamList->getEntry("aggregation: ordering").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("natural", "graph", "random"), "aggregation: ordering")));
+  SET_VALID_ENTRY("aggregation: deterministic");
+  SET_VALID_ENTRY("aggregation: coloring algorithm");
+  SET_VALID_ENTRY("aggregation: enable phase 1");
+  SET_VALID_ENTRY("aggregation: enable phase 2a");
+  SET_VALID_ENTRY("aggregation: enable phase 2b");
+  SET_VALID_ENTRY("aggregation: enable phase 3");
+  SET_VALID_ENTRY("aggregation: match ML phase1");
+  SET_VALID_ENTRY("aggregation: match ML phase2a");
+  SET_VALID_ENTRY("aggregation: match ML phase2b");
+  SET_VALID_ENTRY("aggregation: phase3 avoid singletons");
+  SET_VALID_ENTRY("aggregation: error on nodes with no on-rank neighbors");
+  SET_VALID_ENTRY("aggregation: preserve Dirichlet points");
+  SET_VALID_ENTRY("aggregation: allow user-specified singletons");
+  SET_VALID_ENTRY("aggregation: phase 1 algorithm");
+#undef SET_VALID_ENTRY
+
+  // general variables needed in AggregationFactory
+  validParamList->set<RCP<const FactoryBase> >("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+
+  // special variables necessary for OnePtAggregationAlgorithm
+  validParamList->set<std::string>("OnePt aggregate map name", "", "Name of input map for single node aggregates. (default='')");
+  validParamList->set<std::string>("OnePt aggregate map factory", "", "Generating factory of (DOF) map for single node aggregates.");
+  // validParamList->set< RCP<const FactoryBase> >("OnePt aggregate map factory",    NoFactory::getRCP(), "Generating factory of (DOF) map for single node aggregates.");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Graph");
+  Input(currentLevel, "DofsPerNode");
+
+  const ParameterList& pL = GetParameterList();
+
+  // request special data necessary for OnePtAggregationAlgorithm
+  std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+  if (mapOnePtName.length() > 0) {
+    std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+    if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+      currentLevel.DeclareInput(mapOnePtName, NoFactory::get());
+    } else {
+      RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+      currentLevel.DeclareInput(mapOnePtName, mapOnePtFact.get());
     }
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level &currentLevel) const {
-    using execution_space    = typename LWGraph_kokkos::execution_space;
-    using memory_space       = typename LWGraph_kokkos::memory_space;
-    using local_ordinal_type = typename LWGraph_kokkos::local_ordinal_type;
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    ParameterList pL = GetParameterList();
-    bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
-
-    if (pL.get<int>("aggregation: max agg size") == -1)
-      pL.set("aggregation: max agg size", INT_MAX);
-
-    // define aggregation algorithms
-    RCP<const FactoryBase> graphFact = GetFactory("Graph");
-
-    // TODO Can we keep different aggregation algorithms over more Build calls?
-    algos_.clear();
-    algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm_kokkos(graphFact)));
-    if (pL.get<bool>("aggregation: allow user-specified singletons") == true)   algos_.push_back(rcp(new OnePtAggregationAlgorithm_kokkos             (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 1" )                 == true)   algos_.push_back(rcp(new AggregationPhase1Algorithm_kokkos            (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 2a")                 == true)   algos_.push_back(rcp(new AggregationPhase2aAlgorithm_kokkos           (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 2b")                 == true)   algos_.push_back(rcp(new AggregationPhase2bAlgorithm_kokkos           (graphFact)));
-    if (pL.get<bool>("aggregation: enable phase 3" )                 == true)   algos_.push_back(rcp(new AggregationPhase3Algorithm_kokkos            (graphFact)));
-
-    // Sanity Checking: match ML behavior is not supported in UncoupledAggregation_Kokkos in Phase 1 or Phase 2b, but is in 2a
-    TEUCHOS_TEST_FOR_EXCEPTION( pL.get<bool>("aggregation: match ML phase1"),std::invalid_argument,"Option: 'aggregation: match ML phase1' is not supported in the Kokkos version of uncoupled aggregation");
-    TEUCHOS_TEST_FOR_EXCEPTION( pL.get<bool>("aggregation: match ML phase2b"),std::invalid_argument,"Option: 'aggregation: match ML phase2b' is not supported in the Kokkos version of uncoupled aggregation");
-
-    std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
-    RCP<Map> OnePtMap = Teuchos::null;
-    if (mapOnePtName.length()) {
-      std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
-      if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
-        OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, NoFactory::get());
-      } else {
-        RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
-        OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, mapOnePtFact.get());
-      }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& currentLevel) const {
+  using execution_space    = typename LWGraph_kokkos::execution_space;
+  using memory_space       = typename LWGraph_kokkos::memory_space;
+  using local_ordinal_type = typename LWGraph_kokkos::local_ordinal_type;
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  ParameterList pL  = GetParameterList();
+  bDefinitionPhase_ = false;  // definition phase is finished, now all aggregation algorithm information is fixed
+
+  if (pL.get<int>("aggregation: max agg size") == -1)
+    pL.set("aggregation: max agg size", INT_MAX);
+
+  // define aggregation algorithms
+  RCP<const FactoryBase> graphFact = GetFactory("Graph");
+
+  // TODO Can we keep different aggregation algorithms over more Build calls?
+  algos_.clear();
+  algos_.push_back(rcp(new PreserveDirichletAggregationAlgorithm_kokkos(graphFact)));
+  if (pL.get<bool>("aggregation: allow user-specified singletons") == true) algos_.push_back(rcp(new OnePtAggregationAlgorithm_kokkos(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 1") == true) algos_.push_back(rcp(new AggregationPhase1Algorithm_kokkos(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 2a") == true) algos_.push_back(rcp(new AggregationPhase2aAlgorithm_kokkos(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 2b") == true) algos_.push_back(rcp(new AggregationPhase2bAlgorithm_kokkos(graphFact)));
+  if (pL.get<bool>("aggregation: enable phase 3") == true) algos_.push_back(rcp(new AggregationPhase3Algorithm_kokkos(graphFact)));
+
+  // Sanity Checking: match ML behavior is not supported in UncoupledAggregation_Kokkos in Phase 1 or Phase 2b, but is in 2a
+  TEUCHOS_TEST_FOR_EXCEPTION(pL.get<bool>("aggregation: match ML phase1"), std::invalid_argument, "Option: 'aggregation: match ML phase1' is not supported in the Kokkos version of uncoupled aggregation");
+  TEUCHOS_TEST_FOR_EXCEPTION(pL.get<bool>("aggregation: match ML phase2b"), std::invalid_argument, "Option: 'aggregation: match ML phase2b' is not supported in the Kokkos version of uncoupled aggregation");
+
+  std::string mapOnePtName = pL.get<std::string>("OnePt aggregate map name");
+  RCP<Map> OnePtMap        = Teuchos::null;
+  if (mapOnePtName.length()) {
+    std::string mapOnePtFactName = pL.get<std::string>("OnePt aggregate map factory");
+    if (mapOnePtFactName == "" || mapOnePtFactName == "NoFactory") {
+      OnePtMap = currentLevel.Get<RCP<Map> >(mapOnePtName, NoFactory::get());
+    } else {
+      RCP<const FactoryBase> mapOnePtFact = GetFactory(mapOnePtFactName);
+      OnePtMap                            = currentLevel.Get<RCP<Map> >(mapOnePtName, mapOnePtFact.get());
     }
+  }
 
-    RCP<const LWGraph_kokkos> graph = Get< RCP<LWGraph_kokkos> >(currentLevel, "Graph");
+  RCP<const LWGraph_kokkos> graph = Get<RCP<LWGraph_kokkos> >(currentLevel, "Graph");
+
+  // Build
+  RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("UC");
+
+  const LO numRows = graph->GetNodeNumVertices();
+
+  // construct aggStat information
+  Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type> aggStat(Kokkos::ViewAllocateWithoutInitializing("aggregation status"),
+                                                                        numRows);
+  Kokkos::deep_copy(aggStat, READY);
+
+  // LBV on Sept 06 2019: re-commenting out the dirichlet boundary map
+  // even if the map is correctly extracted from the graph, aggStat is
+  // now a Kokkos::View<unsinged*, memory_space> and filling it will
+  // require a parallel_for or to copy it to the Host which is not really
+  // good from a performance point of view.
+  // If dirichletBoundaryMap was an actual Xpetra::Map, one could call
+  // getLocalMap to have a Kokkos::View on the appropriate memory_space
+  // instead of an ArrayRCP.
+  {
+    typename LWGraph_kokkos::boundary_nodes_type dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
+    Kokkos::parallel_for(
+        "MueLu - UncoupledAggregation: tagging boundary nodes in aggStat",
+        Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const local_ordinal_type nodeIdx) {
+          if (dirichletBoundaryMap(nodeIdx) == true) {
+            aggStat(nodeIdx) = BOUNDARY;
+          }
+        });
+  }
 
-    // Build
-    RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("UC");
+  LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
+  GO indexBase    = graph->GetDomainMap()->getIndexBase();
 
-    const LO numRows = graph->GetNodeNumVertices();
+  /* FIXME: This chunk of code is still executing on the host */
+  if (OnePtMap != Teuchos::null) {
+    typename Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type>::HostMirror aggStatHost = Kokkos::create_mirror_view(aggStat);
+    Kokkos::deep_copy(aggStatHost, aggStat);
 
-    // construct aggStat information
-    Kokkos::View<unsigned*, typename LWGraph_kokkos::device_type> aggStat(Kokkos::ViewAllocateWithoutInitializing("aggregation status"),
-                                                                           numRows);
-    Kokkos::deep_copy(aggStat, READY);
+    for (LO i = 0; i < numRows; i++) {
+      // reconstruct global row id (FIXME only works for contiguous maps)
+      GO grid = (graph->GetDomainMap()->getGlobalElement(i) - indexBase) * nDofsPerNode + indexBase;
 
-    // LBV on Sept 06 2019: re-commenting out the dirichlet boundary map
-    // even if the map is correctly extracted from the graph, aggStat is
-    // now a Kokkos::View<unsinged*, memory_space> and filling it will
-    // require a parallel_for or to copy it to the Host which is not really
-    // good from a performance point of view.
-    // If dirichletBoundaryMap was an actual Xpetra::Map, one could call
-    // getLocalMap to have a Kokkos::View on the appropriate memory_space
-    // instead of an ArrayRCP.
-    {
-      typename LWGraph_kokkos::boundary_nodes_type dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
-      Kokkos::parallel_for("MueLu - UncoupledAggregation: tagging boundary nodes in aggStat",
-                           Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numRows),
-                           KOKKOS_LAMBDA(const local_ordinal_type nodeIdx) {
-                             if (dirichletBoundaryMap(nodeIdx) == true) {
-                               aggStat(nodeIdx) = BOUNDARY;
-                             }
-                           });
+      for (LO kr = 0; kr < nDofsPerNode; kr++)
+        if (OnePtMap->isNodeGlobalElement(grid + kr))
+          aggStatHost(i) = ONEPT;
     }
 
-    LO nDofsPerNode = Get<LO>(currentLevel, "DofsPerNode");
-    GO indexBase = graph->GetDomainMap()->getIndexBase();
-
-    /* FIXME: This chunk of code is still executing on the host */
-    if (OnePtMap != Teuchos::null) {
-      typename Kokkos::View<unsigned*,typename LWGraph_kokkos::device_type>::HostMirror aggStatHost
-        = Kokkos::create_mirror_view(aggStat);
-      Kokkos::deep_copy(aggStatHost, aggStat);
-
-      for (LO i = 0; i < numRows; i++) {
-        // reconstruct global row id (FIXME only works for contiguous maps)
-        GO grid = (graph->GetDomainMap()->getGlobalElement(i)-indexBase) * nDofsPerNode + indexBase;
-
-        for (LO kr = 0; kr < nDofsPerNode; kr++)
-          if (OnePtMap->isNodeGlobalElement(grid + kr))
-            aggStatHost(i) = ONEPT;
-      }
+    Kokkos::deep_copy(aggStat, aggStatHost);
+  }
 
-      Kokkos::deep_copy(aggStat, aggStatHost);
+  const RCP<const Teuchos::Comm<int> > comm = graph->GetComm();
+  GO numGlobalRows                          = 0;
+  if (IsPrint(Statistics1))
+    MueLu_sumAll(comm, as<GO>(numRows), numGlobalRows);
+
+  LO numNonAggregatedNodes = numRows;
+  std::string aggAlgo      = pL.get<std::string>("aggregation: coloring algorithm");
+  if (aggAlgo == "mis2 coarsening" || aggAlgo == "mis2 aggregation") {
+    SubFactoryMonitor sfm(*this, "Algo \"MIS2\"", currentLevel);
+    using graph_t      = typename LWGraph_kokkos::local_graph_type;
+    using device_t     = typename graph_t::device_type;
+    using exec_space   = typename device_t::execution_space;
+    using rowmap_t     = typename graph_t::row_map_type;
+    using colinds_t    = typename graph_t::entries_type;
+    using lno_t        = typename colinds_t::non_const_value_type;
+    rowmap_t aRowptrs  = graph->getLocalLWGraph().getRowPtrs();
+    colinds_t aColinds = graph->getLocalLWGraph().getEntries();
+    lno_t numAggs      = 0;
+    typename colinds_t::non_const_type labels;
+
+    if (aggAlgo == "mis2 coarsening") {
+      if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: MIS-2 coarsening" << std::endl;
+      labels = KokkosGraph::graph_mis2_coarsen<device_t, rowmap_t, colinds_t>(aRowptrs, aColinds, numAggs);
+    } else if (aggAlgo == "mis2 aggregation") {
+      if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: MIS-2 aggregation" << std::endl;
+      labels = KokkosGraph::graph_mis2_aggregate<device_t, rowmap_t, colinds_t>(aRowptrs, aColinds, numAggs);
     }
-
-    const RCP<const Teuchos::Comm<int> > comm = graph->GetComm();
-    GO numGlobalRows = 0;
-    if (IsPrint(Statistics1))
-      MueLu_sumAll(comm, as<GO>(numRows), numGlobalRows);
-
-    LO numNonAggregatedNodes = numRows;
-    std::string aggAlgo = pL.get<std::string>("aggregation: coloring algorithm");
-    if(aggAlgo == "mis2 coarsening" || aggAlgo == "mis2 aggregation")
-    {
-      SubFactoryMonitor sfm(*this, "Algo \"MIS2\"", currentLevel);
-      using graph_t = typename LWGraph_kokkos::local_graph_type;
-      using device_t = typename graph_t::device_type;
-      using exec_space = typename device_t::execution_space;
-      using rowmap_t = typename graph_t::row_map_type;
-      using colinds_t = typename graph_t::entries_type;
-      using lno_t = typename colinds_t::non_const_value_type;
-      rowmap_t aRowptrs = graph->getLocalLWGraph().getRowPtrs();
-      colinds_t aColinds = graph->getLocalLWGraph().getEntries();
-      lno_t numAggs = 0;
-      typename colinds_t::non_const_type labels;
-
-      if(aggAlgo == "mis2 coarsening")
-      {
-        if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: MIS-2 coarsening" << std::endl;
-        labels = KokkosGraph::graph_mis2_coarsen<device_t, rowmap_t, colinds_t>(aRowptrs, aColinds, numAggs);
-      }
-      else if(aggAlgo == "mis2 aggregation")
-      {
-        if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: MIS-2 aggregation" << std::endl;
-        labels = KokkosGraph::graph_mis2_aggregate<device_t, rowmap_t, colinds_t>(aRowptrs, aColinds, numAggs);
-      }
-      auto vertex2AggId  = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-      auto procWinner    = aggregates->GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::OverwriteAll);
-      int rank = comm->getRank();
-      Kokkos::parallel_for(Kokkos::RangePolicy<exec_space>(0, numRows),
-        KOKKOS_LAMBDA(lno_t i)
-        {
+    auto vertex2AggId = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+    auto procWinner   = aggregates->GetProcWinner()->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+    int rank          = comm->getRank();
+    Kokkos::parallel_for(
+        Kokkos::RangePolicy<exec_space>(0, numRows),
+        KOKKOS_LAMBDA(lno_t i) {
           procWinner(i, 0) = rank;
-          if(aggStat(i) == READY)
-          {
-            aggStat(i) = AGGREGATED;
+          if (aggStat(i) == READY) {
+            aggStat(i)         = AGGREGATED;
             vertex2AggId(i, 0) = labels(i);
           }
         });
-      numNonAggregatedNodes = 0;
-      aggregates->SetNumAggregates(numAggs);
-    }
-    else
+    numNonAggregatedNodes = 0;
+    aggregates->SetNumAggregates(numAggs);
+  } else {
     {
-      {
-        SubFactoryMonitor sfm(*this, "Algo \"Graph Coloring\"", currentLevel);
-
-        // LBV on Sept 06 2019: the note below is a little worrisome,
-        // can we guarantee that MueLu is never used on a non-symmetric
-        // graph?
-        // note: just using colinds_view in place of scalar_view_t type
-        // (it won't be used at all by symbolic SPGEMM)
-        using graph_t = typename LWGraph_kokkos::local_graph_type;
-        using KernelHandle = KokkosKernels::Experimental::
+      SubFactoryMonitor sfm(*this, "Algo \"Graph Coloring\"", currentLevel);
+
+      // LBV on Sept 06 2019: the note below is a little worrisome,
+      // can we guarantee that MueLu is never used on a non-symmetric
+      // graph?
+      // note: just using colinds_view in place of scalar_view_t type
+      // (it won't be used at all by symbolic SPGEMM)
+      using graph_t      = typename LWGraph_kokkos::local_graph_type;
+      using KernelHandle = KokkosKernels::Experimental::
           KokkosKernelsHandle<typename graph_t::row_map_type::value_type,
                               typename graph_t::entries_type::value_type,
                               typename graph_t::entries_type::value_type,
                               typename graph_t::device_type::execution_space,
                               typename graph_t::device_type::memory_space,
                               typename graph_t::device_type::memory_space>;
-        KernelHandle kh;
-        //leave gc algorithm choice as the default
-        kh.create_distance2_graph_coloring_handle();
-
-        // get the distance-2 graph coloring handle
-        auto coloringHandle = kh.get_distance2_graph_coloring_handle();
-
-        // Set the distance-2 graph coloring algorithm to use.
-        // Options:
-        //     COLORING_D2_DEFAULT        - Let the kernel handle pick the variation
-        //     COLORING_D2_SERIAL         - Use the legacy serial-only implementation
-        //     COLORING_D2_VB             - Use the parallel vertex based direct method
-        //     COLORING_D2_VB_BIT         - Same as VB but using the bitvector forbidden array
-        //     COLORING_D2_VB_BIT_EF      - Add experimental edge-filtering to VB_BIT
-        //     COLORING_D2_NB_BIT         - Net-based coloring (generally the fastest)
-        if(pL.get<bool>("aggregation: deterministic") == true) {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_SERIAL );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
-        } else if(aggAlgo == "serial") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_SERIAL );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
-        } else if(aggAlgo == "default") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_DEFAULT );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: default" << std::endl;
-        } else if(aggAlgo == "vertex based") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_VB );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based" << std::endl;
-        } else if(aggAlgo == "vertex based bit set") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_VB_BIT );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based bit set" << std::endl;
-        } else if(aggAlgo == "edge filtering") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_VB_BIT_EF );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: edge filtering" << std::endl;
-        } else if(aggAlgo == "net based bit set") {
-          coloringHandle->set_algorithm( KokkosGraph::COLORING_D2_NB_BIT );
-          if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: net based bit set" << std::endl;
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(true,std::invalid_argument,"Unrecognized distance 2 coloring algorithm, valid options are: serial, default, matrix squared, vertex based, vertex based bit set, edge filtering")
-            }
-
-        //Create device views for graph rowptrs/colinds
-        typename graph_t::row_map_type aRowptrs = graph->getLocalLWGraph().getRowPtrs();
-        typename graph_t::entries_type aColinds = graph->getLocalLWGraph().getEntries();
-
-        //run d2 graph coloring
-        //graph is symmetric so row map/entries and col map/entries are the same
-        {
-          SubFactoryMonitor sfm2(*this, "Algo \"Graph Coloring\": KokkosGraph Call", currentLevel);//CMS HACK
-          KokkosGraph::Experimental::graph_color_distance2(&kh, numRows, aRowptrs, aColinds);
-        }
+      KernelHandle kh;
+      // leave gc algorithm choice as the default
+      kh.create_distance2_graph_coloring_handle();
+
+      // get the distance-2 graph coloring handle
+      auto coloringHandle = kh.get_distance2_graph_coloring_handle();
+
+      // Set the distance-2 graph coloring algorithm to use.
+      // Options:
+      //     COLORING_D2_DEFAULT        - Let the kernel handle pick the variation
+      //     COLORING_D2_SERIAL         - Use the legacy serial-only implementation
+      //     COLORING_D2_VB             - Use the parallel vertex based direct method
+      //     COLORING_D2_VB_BIT         - Same as VB but using the bitvector forbidden array
+      //     COLORING_D2_VB_BIT_EF      - Add experimental edge-filtering to VB_BIT
+      //     COLORING_D2_NB_BIT         - Net-based coloring (generally the fastest)
+      if (pL.get<bool>("aggregation: deterministic") == true) {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_SERIAL);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
+      } else if (aggAlgo == "serial") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_SERIAL);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
+      } else if (aggAlgo == "default") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_DEFAULT);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: default" << std::endl;
+      } else if (aggAlgo == "vertex based") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_VB);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based" << std::endl;
+      } else if (aggAlgo == "vertex based bit set") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_VB_BIT);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based bit set" << std::endl;
+      } else if (aggAlgo == "edge filtering") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_VB_BIT_EF);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: edge filtering" << std::endl;
+      } else if (aggAlgo == "net based bit set") {
+        coloringHandle->set_algorithm(KokkosGraph::COLORING_D2_NB_BIT);
+        if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: net based bit set" << std::endl;
+      } else {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "Unrecognized distance 2 coloring algorithm, valid options are: serial, default, matrix squared, vertex based, vertex based bit set, edge filtering")
+      }
 
-        // extract the colors and store them in the aggregates
-        aggregates->SetGraphColors(coloringHandle->get_vertex_colors());
-        aggregates->SetGraphNumColors(static_cast<LO>(coloringHandle->get_num_colors()));
+      // Create device views for graph rowptrs/colinds
+      typename graph_t::row_map_type aRowptrs = graph->getLocalLWGraph().getRowPtrs();
+      typename graph_t::entries_type aColinds = graph->getLocalLWGraph().getEntries();
 
-        //clean up coloring handle
-        kh.destroy_distance2_graph_coloring_handle();
+      // run d2 graph coloring
+      // graph is symmetric so row map/entries and col map/entries are the same
+      {
+        SubFactoryMonitor sfm2(*this, "Algo \"Graph Coloring\": KokkosGraph Call", currentLevel);  // CMS HACK
+        KokkosGraph::Experimental::graph_color_distance2(&kh, numRows, aRowptrs, aColinds);
       }
 
+      // extract the colors and store them in the aggregates
+      aggregates->SetGraphColors(coloringHandle->get_vertex_colors());
+      aggregates->SetGraphNumColors(static_cast<LO>(coloringHandle->get_num_colors()));
+
+      // clean up coloring handle
+      kh.destroy_distance2_graph_coloring_handle();
+    }
+
+    if (IsPrint(Statistics1)) {
+      GetOStream(Statistics1) << "  num colors: " << aggregates->GetGraphNumColors() << std::endl;
+    }
+    GO numGlobalAggregatedPrev = 0, numGlobalAggsPrev = 0;
+    for (size_t a = 0; a < algos_.size(); a++) {
+      std::string phase = algos_[a]->description();
+      SubFactoryMonitor sfm2(*this, "Algo \"" + phase + "\"", currentLevel);
+
+      int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
+      algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+      algos_[a]->SetProcRankVerbose(oldRank);
+
       if (IsPrint(Statistics1)) {
-        GetOStream(Statistics1) << "  num colors: " << aggregates->GetGraphNumColors() << std::endl;
-      }
-      GO numGlobalAggregatedPrev = 0, numGlobalAggsPrev = 0;
-      for (size_t a = 0; a < algos_.size(); a++) {
-        std::string phase = algos_[a]->description();
-        SubFactoryMonitor sfm2(*this, "Algo \"" + phase + "\"", currentLevel);
-
-        int oldRank = algos_[a]->SetProcRankVerbose(this->GetProcRankVerbose());
-        algos_[a]->BuildAggregates(pL, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-        algos_[a]->SetProcRankVerbose(oldRank);
-
-        if (IsPrint(Statistics1)) {
-          GO numLocalAggregated = numRows - numNonAggregatedNodes, numGlobalAggregated = 0;
-          GO numLocalAggs       = aggregates->GetNumAggregates(),  numGlobalAggs = 0;
-          MueLu_sumAll(comm, numLocalAggregated, numGlobalAggregated);
-          MueLu_sumAll(comm, numLocalAggs,       numGlobalAggs);
-
-          double aggPercent = 100*as<double>(numGlobalAggregated)/as<double>(numGlobalRows);
-          if (aggPercent > 99.99 && aggPercent < 100.00) {
-            // Due to round off (for instance, for 140465733/140466897), we could
-            // get 100.00% display even if there are some remaining nodes. This
-            // is bad from the users point of view. It is much better to change
-            // it to display 99.99%.
-            aggPercent = 99.99;
-          }
-          GetOStream(Statistics1) << "  aggregated : " << (numGlobalAggregated - numGlobalAggregatedPrev) << " (phase), " << std::fixed
-                                  << std::setprecision(2) << numGlobalAggregated << "/" << numGlobalRows << " [" << aggPercent << "%] (total)\n"
-                                  << "  remaining  : " << numGlobalRows - numGlobalAggregated << "\n"
-                                  << "  aggregates : " << numGlobalAggs-numGlobalAggsPrev << " (phase), " << numGlobalAggs << " (total)" << std::endl;
-          numGlobalAggregatedPrev = numGlobalAggregated;
-          numGlobalAggsPrev       = numGlobalAggs;
+        GO numLocalAggregated = numRows - numNonAggregatedNodes, numGlobalAggregated = 0;
+        GO numLocalAggs = aggregates->GetNumAggregates(), numGlobalAggs = 0;
+        MueLu_sumAll(comm, numLocalAggregated, numGlobalAggregated);
+        MueLu_sumAll(comm, numLocalAggs, numGlobalAggs);
+
+        double aggPercent = 100 * as<double>(numGlobalAggregated) / as<double>(numGlobalRows);
+        if (aggPercent > 99.99 && aggPercent < 100.00) {
+          // Due to round off (for instance, for 140465733/140466897), we could
+          // get 100.00% display even if there are some remaining nodes. This
+          // is bad from the users point of view. It is much better to change
+          // it to display 99.99%.
+          aggPercent = 99.99;
         }
+        GetOStream(Statistics1) << "  aggregated : " << (numGlobalAggregated - numGlobalAggregatedPrev) << " (phase), " << std::fixed
+                                << std::setprecision(2) << numGlobalAggregated << "/" << numGlobalRows << " [" << aggPercent << "%] (total)\n"
+                                << "  remaining  : " << numGlobalRows - numGlobalAggregated << "\n"
+                                << "  aggregates : " << numGlobalAggs - numGlobalAggsPrev << " (phase), " << numGlobalAggs << " (total)" << std::endl;
+        numGlobalAggregatedPrev = numGlobalAggregated;
+        numGlobalAggsPrev       = numGlobalAggs;
       }
     }
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError, "MueLu::UncoupledAggregationFactory::Build: Leftover nodes found! Error!");
-
-    aggregates->AggregatesCrossProcessors(false);
-    aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
+  TEUCHOS_TEST_FOR_EXCEPTION(numNonAggregatedNodes, Exceptions::RuntimeError, "MueLu::UncoupledAggregationFactory::Build: Leftover nodes found! Error!");
 
-    Set(currentLevel, "Aggregates", aggregates);
+  aggregates->AggregatesCrossProcessors(false);
+  aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
 
-  }
+  Set(currentLevel, "Aggregates", aggregates);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_UNCOUPLEDAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_decl.hpp b/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_decl.hpp
index 17c7a3297fd6..9fbcbbc45951 100644
--- a/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_decl.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_USERAGGREGATIONFACTORY_DECL_HPP_
 #define MUELU_USERAGGREGATIONFACTORY_DECL_HPP_
 
-
 #include <Xpetra_Map_fwd.hpp>
 #include <Xpetra_MapFactory_fwd.hpp>
 
@@ -60,22 +59,22 @@
 
 namespace MueLu {
 
-template<class LocalOrdinal = DefaultLocalOrdinal,
-         class GlobalOrdinal = DefaultGlobalOrdinal,
-         class Node = DefaultNode>
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
 class UserAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_USERAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-public:
+ public:
   //! @name Constructors/Destructors.
   //@{
 
   //! Constructor.
-  UserAggregationFactory() { };
+  UserAggregationFactory(){};
 
   //! Destructor.
-  virtual ~UserAggregationFactory() { }
+  virtual ~UserAggregationFactory() {}
 
   RCP<const ParameterList> GetValidParameterList() const;
 
@@ -101,10 +100,10 @@ class UserAggregationFactory : public SingleLevelFactoryBase {
 
   //@}
 
-private:
-}; // class UserAggregationFactory
+ private:
+};  // class UserAggregationFactory
 
-}
+}  // namespace MueLu
 
 #define MUELU_USERAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_USERAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_def.hpp b/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_def.hpp
index bc0a418a46c7..45d7f9a229d2 100644
--- a/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_def.hpp
+++ b/packages/muelu/src/Graph/UserAggregation/MueLu_UserAggregationFactory_def.hpp
@@ -60,87 +60,87 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    // input parameters
-    validParamList->set<std::string>("filePrefix", "", "The data is read from files of this name: <filePrefix><level>_<PID>.<fileExt>");
-    validParamList->set<std::string>("fileExt",    "", "The data is read from files of this name: <filePrefix><level>_<PID>.<fileExt>");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  // input parameters
+  validParamList->set<std::string>("filePrefix", "", "The data is read from files of this name: <filePrefix><level>_<PID>.<fileExt>");
+  validParamList->set<std::string>("fileExt", "", "The data is read from files of this name: <filePrefix><level>_<PID>.<fileExt>");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* currentLevel */) const {}
+
+/**
+ * The function reads aggregate information from a file.
+ * The file structure is the following:
+ *  * line 1 : <number of aggregates>
+ *  * line 2+: <aggregate size> <node 1 (root node) GID> <node 2 GID> ... <node last GID>
+ */
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  const ParameterList& pL = GetParameterList();
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  const int myRank                    = comm->getRank();
+
+  std::string fileName = pL.get<std::string>("filePrefix") + toString(currentLevel.GetLevelID()) + "_" + toString(myRank) + "." + pL.get<std::string>("fileExt");
+  std::ifstream ifs(fileName.c_str());
+  TEUCHOS_TEST_FOR_EXCEPTION(!ifs.good(), Exceptions::RuntimeError, "Cannot read data from \"" << fileName << "\"");
+
+  LO numVertices, numAggregates;
+  ifs >> numVertices;
+  TEUCHOS_TEST_FOR_EXCEPTION(!ifs.good(), Exceptions::RuntimeError, "Cannot read data from \"" << fileName << "\"");
+  ifs >> numAggregates;
+  TEUCHOS_TEST_FOR_EXCEPTION(numVertices <= 0, Exceptions::InvalidArgument, "Number of vertices   must be > 0");
+  TEUCHOS_TEST_FOR_EXCEPTION(numAggregates <= 0, Exceptions::InvalidArgument, "Number of aggregates must be > 0");
+
+  Xpetra::UnderlyingLib lib = currentLevel.lib();
+  const int indexBase       = 0;
+  RCP<Map> map              = MapFactory::Build(lib, numVertices, indexBase, comm);
+
+  RCP<Aggregates> aggregates = rcp(new Aggregates(map));
+  aggregates->setObjectLabel("User");
+
+  aggregates->SetNumAggregates(numAggregates);
+
+  Teuchos::ArrayRCP<LO> vertex2AggId = aggregates->GetVertex2AggId()->getDataNonConst(0);
+  Teuchos::ArrayRCP<LO> procWinner   = aggregates->GetProcWinner()->getDataNonConst(0);
+
+  for (LO i = 0; i < numAggregates; i++) {
+    int aggSize = 0;
+    ifs >> aggSize;
+
+    std::vector<LO> list(aggSize);
+    for (int k = 0; k < aggSize; k++) {
+      // FIXME: File contains GIDs, we need LIDs
+      // for now, works on a single processor
+      ifs >> list[k];
+    }
 
-    return validParamList;
-  }
+    // Mark first node as root node for the aggregate
+    aggregates->SetIsRoot(list[0]);
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* currentLevel */) const { }
-
-  /**
-   * The function reads aggregate information from a file.
-   * The file structure is the following:
-   *  * line 1 : <number of aggregates>
-   *  * line 2+: <aggregate size> <node 1 (root node) GID> <node 2 GID> ... <node last GID>
-   */
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    const ParameterList& pL = GetParameterList();
-
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    const int myRank = comm->getRank();
-
-    std::string fileName = pL.get<std::string>("filePrefix") + toString(currentLevel.GetLevelID()) + "_" + toString(myRank) + "." + pL.get<std::string>("fileExt");
-    std::ifstream ifs(fileName.c_str());
-    TEUCHOS_TEST_FOR_EXCEPTION(!ifs.good(), Exceptions::RuntimeError, "Cannot read data from \"" << fileName << "\"");
-
-    LO numVertices, numAggregates;
-    ifs >> numVertices;
-    TEUCHOS_TEST_FOR_EXCEPTION(!ifs.good(), Exceptions::RuntimeError, "Cannot read data from \"" << fileName << "\"");
-    ifs >> numAggregates;
-    TEUCHOS_TEST_FOR_EXCEPTION(numVertices   <= 0, Exceptions::InvalidArgument, "Number of vertices   must be > 0");
-    TEUCHOS_TEST_FOR_EXCEPTION(numAggregates <= 0, Exceptions::InvalidArgument, "Number of aggregates must be > 0");
-
-    Xpetra::UnderlyingLib  lib       = currentLevel.lib();
-    const int              indexBase = 0;
-    RCP<Map> map = MapFactory::Build(lib, numVertices, indexBase, comm);
-
-    RCP<Aggregates> aggregates = rcp(new Aggregates(map));
-    aggregates->setObjectLabel("User");
-
-    aggregates->SetNumAggregates(numAggregates);
-
-    Teuchos::ArrayRCP<LO> vertex2AggId = aggregates->GetVertex2AggId()->getDataNonConst(0);
-    Teuchos::ArrayRCP<LO> procWinner   = aggregates->GetProcWinner()  ->getDataNonConst(0);
-
-    for (LO i = 0; i < numAggregates; i++) {
-      int aggSize = 0;
-      ifs >> aggSize;
-
-      std::vector<LO> list(aggSize);
-      for (int k = 0; k < aggSize; k++) {
-        // FIXME: File contains GIDs, we need LIDs
-        // for now, works on a single processor
-        ifs >> list[k];
-      }
-
-      // Mark first node as root node for the aggregate
-      aggregates->SetIsRoot(list[0]);
-
-      // Fill vertex2AggId and procWinner structure with information
-      for (int k = 0; k < aggSize; k++) {
-        vertex2AggId[list[k]] = i;
-        procWinner  [list[k]] = myRank;
-      }
+    // Fill vertex2AggId and procWinner structure with information
+    for (int k = 0; k < aggSize; k++) {
+      vertex2AggId[list[k]] = i;
+      procWinner[list[k]]   = myRank;
     }
+  }
 
-    // FIXME: do the proper check whether aggregates cross interprocessor boundary
-    aggregates->AggregatesCrossProcessors(false);
+  // FIXME: do the proper check whether aggregates cross interprocessor boundary
+  aggregates->AggregatesCrossProcessors(false);
 
-    Set(currentLevel, "Aggregates", aggregates);
+  Set(currentLevel, "Aggregates", aggregates);
 
-    GetOStream(Statistics0) << aggregates->description() << std::endl;
-  }
+  GetOStream(Statistics0) << aggregates->description() << std::endl;
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_USERAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Headers/MueLu_ConfigDefs.hpp b/packages/muelu/src/Headers/MueLu_ConfigDefs.hpp
index 6d65bf28102c..799914095ee3 100644
--- a/packages/muelu/src/Headers/MueLu_ConfigDefs.hpp
+++ b/packages/muelu/src/Headers/MueLu_ConfigDefs.hpp
@@ -51,7 +51,7 @@
 #include <Teuchos_ConfigDefs.hpp>
 
 // Tpetra
-#include <Tpetra_KokkosCompat_DefaultNode.hpp> // default template parameter of many MueLu classes
+#include <Tpetra_KokkosCompat_DefaultNode.hpp>  // default template parameter of many MueLu classes
 
 // Memory management
 #include <Teuchos_Array.hpp>
@@ -83,33 +83,33 @@
 //! Namespace for MueLu classes and methods
 namespace MueLu {
 
-  // import Teuchos memory management classes into MueLu
-  using Teuchos::arcp;
-  using Teuchos::arcpFromArrayView;
-  using Teuchos::arcp_reinterpret_cast;
-  using Teuchos::Array;
-  using Teuchos::ArrayRCP;
-  using Teuchos::ArrayView;
-  using Teuchos::as;
-  using Teuchos::null;
-  using Teuchos::ParameterList;
-  using Teuchos::rcp;
-  using Teuchos::RCP;
-  using Teuchos::rcp_const_cast;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcpFromRef;
-  using Teuchos::rcp_implicit_cast;
-  using Teuchos::rcp_static_cast;
+// import Teuchos memory management classes into MueLu
+using Teuchos::arcp;
+using Teuchos::arcp_reinterpret_cast;
+using Teuchos::arcpFromArrayView;
+using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::ArrayView;
+using Teuchos::as;
+using Teuchos::null;
+using Teuchos::ParameterList;
+using Teuchos::rcp;
+using Teuchos::RCP;
+using Teuchos::rcp_const_cast;
+using Teuchos::rcp_dynamic_cast;
+using Teuchos::rcp_implicit_cast;
+using Teuchos::rcp_static_cast;
+using Teuchos::rcpFromRef;
 
-  // verbose levels
-  using Teuchos::VERB_DEFAULT;
-  using Teuchos::VERB_NONE;
-  using Teuchos::VERB_LOW;
-  using Teuchos::VERB_MEDIUM;
-  using Teuchos::VERB_HIGH;
-  using Teuchos::VERB_EXTREME;
+// verbose levels
+using Teuchos::VERB_DEFAULT;
+using Teuchos::VERB_EXTREME;
+using Teuchos::VERB_HIGH;
+using Teuchos::VERB_LOW;
+using Teuchos::VERB_MEDIUM;
+using Teuchos::VERB_NONE;
 
-}
+}  // namespace MueLu
 
 // This include file defines macros to avoid warnings under CUDA.  See github issue #1133.
 #include "Teuchos_CompilerCodeTweakMacros.hpp"
diff --git a/packages/muelu/src/Headers/MueLu_Details_DefaultTypes.hpp b/packages/muelu/src/Headers/MueLu_Details_DefaultTypes.hpp
index fa9745afbe37..b1d19f73d5e8 100644
--- a/packages/muelu/src/Headers/MueLu_Details_DefaultTypes.hpp
+++ b/packages/muelu/src/Headers/MueLu_Details_DefaultTypes.hpp
@@ -52,22 +52,21 @@
 
 #include <Tpetra_Details_DefaultTypes.hpp>
 
-namespace MueLu
-{
+namespace MueLu {
 
-  typedef Tpetra::Details::DefaultTypes::scalar_type DefaultScalar;
+typedef Tpetra::Details::DefaultTypes::scalar_type DefaultScalar;
 
-  typedef int DefaultLocalOrdinal;
+typedef int DefaultLocalOrdinal;
 
-  #if defined HAVE_MUELU_DEFAULT_GO_LONG
-  typedef long DefaultGlobalOrdinal;
-  #elif defined HAVE_MUELU_DEFAULT_GO_LONGLONG
-  typedef long long DefaultGlobalOrdinal;
-  #else
-  typedef int DefaultGlobalOrdinal;
-  #endif
+#if defined HAVE_MUELU_DEFAULT_GO_LONG
+typedef long DefaultGlobalOrdinal;
+#elif defined HAVE_MUELU_DEFAULT_GO_LONGLONG
+typedef long long DefaultGlobalOrdinal;
+#else
+typedef int DefaultGlobalOrdinal;
+#endif
 
-  typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType DefaultNode;
-}
+typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType DefaultNode;
+}  // namespace MueLu
 
 #endif
diff --git a/packages/muelu/src/Headers/MueLu_Types.hpp b/packages/muelu/src/Headers/MueLu_Types.hpp
index 6ea5d31711d7..26182a928a38 100644
--- a/packages/muelu/src/Headers/MueLu_Types.hpp
+++ b/packages/muelu/src/Headers/MueLu_Types.hpp
@@ -49,55 +49,55 @@
 #include "MueLu_ConfigDefs.hpp"
 
 namespace MueLu {
-  enum CycleType {
-    VCYCLE,
-    WCYCLE
-  };
+enum CycleType {
+  VCYCLE,
+  WCYCLE
+};
 
-  enum PreOrPost {
-    PRE  = 0x1,
-    POST = 0x2,
-    BOTH = 0x3
-  };
+enum PreOrPost {
+  PRE  = 0x1,
+  POST = 0x2,
+  BOTH = 0x3
+};
 
-  // In the algorithm, aggStat[] = READY/NOTSEL/SELECTED indicates whether a node has been aggregated
-  enum NodeState {
-    READY      = 1, // indicates that a node is available to be
-                    // selected as a root node of an aggregate
+// In the algorithm, aggStat[] = READY/NOTSEL/SELECTED indicates whether a node has been aggregated
+enum NodeState {
+  READY = 1,  // indicates that a node is available to be
+              // selected as a root node of an aggregate
 
-    NOTSEL     = 2, // indicates that a node has been rejected as a root node.
-                    // This could perhaps be because if this node had been
-                    // selected a small aggregate would have resulted
-                    // This is Phase 1 specific
+  NOTSEL = 2,  // indicates that a node has been rejected as a root node.
+               // This could perhaps be because if this node had been
+               // selected a small aggregate would have resulted
+               // This is Phase 1 specific
 
-    AGGREGATED = 3, // indicates that a node has been assigned
-                    // to an aggregate
+  AGGREGATED = 3,  // indicates that a node has been assigned
+                   // to an aggregate
 
-    ONEPT      = 4, // indicates that a node shall be preserved over
-                    // all multigrid levels as 1 point aggregate
+  ONEPT = 4,  // indicates that a node shall be preserved over
+              // all multigrid levels as 1 point aggregate
 
-    IGNORED    = 5, // indicates that the node is removed from consideration,
-                    // and is not aggregated
+  IGNORED = 5,  // indicates that the node is removed from consideration,
+                // and is not aggregated
 
-    BOUNDARY   = 6, // node is a Dirichlet node
-                    // During aggregation, it is transformed either to AGGREGATED
-                    // or to IGNORED
-    INTERFACE  = 7  // node is chosen as root node on an interface where coordinated
-                    // coarsening across the interface is required.
-  };
+  BOUNDARY = 6,  // node is a Dirichlet node
+                 // During aggregation, it is transformed either to AGGREGATED
+                 // or to IGNORED
+  INTERFACE = 7  // node is chosen as root node on an interface where coordinated
+                 // coarsening across the interface is required.
+};
 
-  // This is use by the structured aggregation index manager to keep track of the underlying mesh
-  // layout.
-  enum IndexingType {
-    UNCOUPLED  = 1, // indicates that the underlying mesh is treated independently from rank to rank
+// This is use by the structured aggregation index manager to keep track of the underlying mesh
+// layout.
+enum IndexingType {
+  UNCOUPLED = 1,  // indicates that the underlying mesh is treated independently from rank to rank
 
-    LOCALLEXI  = 2, // local lexicographic indexing of the mesh, this is similar to uncoupled but
-                    // extra data is used to compute indices accross ranks
+  LOCALLEXI = 2,  // local lexicographic indexing of the mesh, this is similar to uncoupled but
+                  // extra data is used to compute indices accross ranks
 
-    GLOBALLEXI = 3  // global lexicographic indexing of the mesh means that the mesh is ordered
-                    // lexicographically accorss and subsequently split among ranks.
-  };
+  GLOBALLEXI = 3  // global lexicographic indexing of the mesh means that the mesh is ordered
+                  // lexicographically accorss and subsequently split among ranks.
+};
 
-}
+}  // namespace MueLu
 
-#endif //ifndef MUELU_TYPES_HPP
+#endif  // ifndef MUELU_TYPES_HPP
diff --git a/packages/muelu/src/Headers/MueLu_UseShortNames.hpp b/packages/muelu/src/Headers/MueLu_UseShortNames.hpp
index 7a25b031132e..4142deca5b21 100644
--- a/packages/muelu/src/Headers/MueLu_UseShortNames.hpp
+++ b/packages/muelu/src/Headers/MueLu_UseShortNames.hpp
@@ -81,4 +81,4 @@
 
 //! @file MueLu_UseShortNamesOrdinal.hpp
 
-//TODO / NOTE: This file should not be included at the global scope (to avoid name collision)
+// TODO / NOTE: This file should not be included at the global scope (to avoid name collision)
diff --git a/packages/muelu/src/Headers/MueLu_UseShortNamesOrdinal.hpp b/packages/muelu/src/Headers/MueLu_UseShortNamesOrdinal.hpp
index 9bc47bfe9b36..63fa18d9a1df 100644
--- a/packages/muelu/src/Headers/MueLu_UseShortNamesOrdinal.hpp
+++ b/packages/muelu/src/Headers/MueLu_UseShortNamesOrdinal.hpp
@@ -3,124 +3,124 @@
 #include <Xpetra_UseShortNamesOrdinal.hpp>
 
 #ifdef MUELU_AGGREGATES_SHORT
-using Aggregates [[maybe_unused]] = MueLu::Aggregates<LocalOrdinal,GlobalOrdinal,Node>;
+using Aggregates [[maybe_unused]] = MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE1ALGORITHM_SHORT
-using AggregationPhase1Algorithm [[maybe_unused]] = MueLu::AggregationPhase1Algorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase1Algorithm [[maybe_unused]] = MueLu::AggregationPhase1Algorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_SHORT
-using AggregationPhase1Algorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase1Algorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase1Algorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase1Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE2AALGORITHM_SHORT
-using AggregationPhase2aAlgorithm [[maybe_unused]] = MueLu::AggregationPhase2aAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase2aAlgorithm [[maybe_unused]] = MueLu::AggregationPhase2aAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_SHORT
-using AggregationPhase2aAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase2aAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase2aAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase2aAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE2BALGORITHM_SHORT
-using AggregationPhase2bAlgorithm [[maybe_unused]] = MueLu::AggregationPhase2bAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase2bAlgorithm [[maybe_unused]] = MueLu::AggregationPhase2bAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_SHORT
-using AggregationPhase2bAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase2bAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase2bAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase2bAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE3ALGORITHM_SHORT
-using AggregationPhase3Algorithm [[maybe_unused]] = MueLu::AggregationPhase3Algorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase3Algorithm [[maybe_unused]] = MueLu::AggregationPhase3Algorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_SHORT
-using AggregationPhase3Algorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase3Algorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationPhase3Algorithm_kokkos [[maybe_unused]] = MueLu::AggregationPhase3Algorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_SHORT
-using AggregationStructuredAlgorithm [[maybe_unused]] = MueLu::AggregationStructuredAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationStructuredAlgorithm [[maybe_unused]] = MueLu::AggregationStructuredAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_SHORT
-using AggregationStructuredAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationStructuredAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationStructuredAlgorithm_kokkos [[maybe_unused]] = MueLu::AggregationStructuredAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AMALGAMATIONINFO_SHORT
-using AmalgamationInfo [[maybe_unused]] = MueLu::AmalgamationInfo<LocalOrdinal,GlobalOrdinal,Node>;
+using AmalgamationInfo [[maybe_unused]] = MueLu::AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_SHORT
-using GlobalLexicographicIndexManager [[maybe_unused]] = MueLu::GlobalLexicographicIndexManager<LocalOrdinal,GlobalOrdinal,Node>;
+using GlobalLexicographicIndexManager [[maybe_unused]] = MueLu::GlobalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GRAPH_SHORT
-using Graph [[maybe_unused]] = MueLu::Graph<LocalOrdinal,GlobalOrdinal,Node>;
+using Graph [[maybe_unused]] = MueLu::Graph<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GRAPHBASE_SHORT
-using GraphBase [[maybe_unused]] = MueLu::GraphBase<LocalOrdinal,GlobalOrdinal,Node>;
+using GraphBase [[maybe_unused]] = MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_HYBRIDAGGREGATIONFACTORY_SHORT
-using HybridAggregationFactory [[maybe_unused]] = MueLu::HybridAggregationFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using HybridAggregationFactory [[maybe_unused]] = MueLu::HybridAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INDEXMANAGER_SHORT
-using IndexManager [[maybe_unused]] = MueLu::IndexManager<LocalOrdinal,GlobalOrdinal,Node>;
+using IndexManager [[maybe_unused]] = MueLu::IndexManager<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INDEXMANAGER_KOKKOS_SHORT
-using IndexManager_kokkos [[maybe_unused]] = MueLu::IndexManager_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using IndexManager_kokkos [[maybe_unused]] = MueLu::IndexManager_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
-using InterfaceAggregationAlgorithm [[maybe_unused]] = MueLu::InterfaceAggregationAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using InterfaceAggregationAlgorithm [[maybe_unused]] = MueLu::InterfaceAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INTERFACEMAPPINGTRANSFERFACTORY_SHORT
-using InterfaceMappingTransferFactory [[maybe_unused]] = MueLu::InterfaceMappingTransferFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using InterfaceMappingTransferFactory [[maybe_unused]] = MueLu::InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_SHORT
-using IsolatedNodeAggregationAlgorithm [[maybe_unused]] = MueLu::IsolatedNodeAggregationAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using IsolatedNodeAggregationAlgorithm [[maybe_unused]] = MueLu::IsolatedNodeAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_SHORT
-using IsolatedNodeAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::IsolatedNodeAggregationAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using IsolatedNodeAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::IsolatedNodeAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ISORROPIAINTERFACE_SHORT
-using IsorropiaInterface [[maybe_unused]] = MueLu::IsorropiaInterface<LocalOrdinal,GlobalOrdinal,Node>;
+using IsorropiaInterface [[maybe_unused]] = MueLu::IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LWGRAPH_SHORT
-using LWGraph [[maybe_unused]] = MueLu::LWGraph<LocalOrdinal,GlobalOrdinal,Node>;
+using LWGraph [[maybe_unused]] = MueLu::LWGraph<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LWGRAPH_KOKKOS_SHORT
-using LWGraph_kokkos [[maybe_unused]] = MueLu::LWGraph_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using LWGraph_kokkos [[maybe_unused]] = MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LOCALLWGRAPH_KOKKOS_SHORT
-using LocalLWGraph_kokkos [[maybe_unused]] = MueLu::LocalLWGraph_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using LocalLWGraph_kokkos [[maybe_unused]] = MueLu::LocalLWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_SHORT
-using LocalLexicographicIndexManager [[maybe_unused]] = MueLu::LocalLexicographicIndexManager<LocalOrdinal,GlobalOrdinal,Node>;
+using LocalLexicographicIndexManager [[maybe_unused]] = MueLu::LocalLexicographicIndexManager<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LOCALORDINALTRANSFERFACTORY_SHORT
-using LocalOrdinalTransferFactory [[maybe_unused]] = MueLu::LocalOrdinalTransferFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using LocalOrdinalTransferFactory [[maybe_unused]] = MueLu::LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ONEPTAGGREGATIONALGORITHM_SHORT
-using OnePtAggregationAlgorithm [[maybe_unused]] = MueLu::OnePtAggregationAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using OnePtAggregationAlgorithm [[maybe_unused]] = MueLu::OnePtAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_SHORT
-using OnePtAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::OnePtAggregationAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using OnePtAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::OnePtAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_SHORT
-using PreserveDirichletAggregationAlgorithm [[maybe_unused]] = MueLu::PreserveDirichletAggregationAlgorithm<LocalOrdinal,GlobalOrdinal,Node>;
+using PreserveDirichletAggregationAlgorithm [[maybe_unused]] = MueLu::PreserveDirichletAggregationAlgorithm<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_SHORT
-using PreserveDirichletAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::PreserveDirichletAggregationAlgorithm_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using PreserveDirichletAggregationAlgorithm_kokkos [[maybe_unused]] = MueLu::PreserveDirichletAggregationAlgorithm_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PRFACTORY_SHORT
-using PRFactory [[maybe_unused]] = MueLu::PRFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using PRFactory [[maybe_unused]] = MueLu::PRFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCEMAPFACTORY_SHORT
-using RebalanceMapFactory [[maybe_unused]] = MueLu::RebalanceMapFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceMapFactory [[maybe_unused]] = MueLu::RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REPARTITIONINTERFACE_SHORT
-using RepartitionInterface [[maybe_unused]] = MueLu::RepartitionInterface<LocalOrdinal,GlobalOrdinal,Node>;
+using RepartitionInterface [[maybe_unused]] = MueLu::RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_SHORT
-using StructuredAggregationFactory_kokkos [[maybe_unused]] = MueLu::StructuredAggregationFactory_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using StructuredAggregationFactory_kokkos [[maybe_unused]] = MueLu::StructuredAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UNCOUPLEDAGGREGATIONFACTORY_SHORT
-using UncoupledAggregationFactory [[maybe_unused]] = MueLu::UncoupledAggregationFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using UncoupledAggregationFactory [[maybe_unused]] = MueLu::UncoupledAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_SHORT
-using UncoupledAggregationFactory_kokkos [[maybe_unused]] = MueLu::UncoupledAggregationFactory_kokkos<LocalOrdinal,GlobalOrdinal,Node>;
+using UncoupledAggregationFactory_kokkos [[maybe_unused]] = MueLu::UncoupledAggregationFactory_kokkos<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UNCOUPLEDINDEXMANAGER_SHORT
-using UncoupledIndexManager [[maybe_unused]] = MueLu::UncoupledIndexManager<LocalOrdinal,GlobalOrdinal,Node>;
+using UncoupledIndexManager [[maybe_unused]] = MueLu::UncoupledIndexManager<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_USERAGGREGATIONFACTORY_SHORT
-using UserAggregationFactory [[maybe_unused]] = MueLu::UserAggregationFactory<LocalOrdinal,GlobalOrdinal,Node>;
+using UserAggregationFactory [[maybe_unused]] = MueLu::UserAggregationFactory<LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACTORY_SHORT
 using Factory [[maybe_unused]] = MueLu::Factory;
diff --git a/packages/muelu/src/Headers/MueLu_UseShortNamesScalar.hpp b/packages/muelu/src/Headers/MueLu_UseShortNamesScalar.hpp
index d326324c097f..532f051b27ef 100644
--- a/packages/muelu/src/Headers/MueLu_UseShortNamesScalar.hpp
+++ b/packages/muelu/src/Headers/MueLu_UseShortNamesScalar.hpp
@@ -3,461 +3,461 @@
 #include <Xpetra_UseShortNamesScalar.hpp>
 
 #ifdef MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_SHORT
-using AdaptiveSaMLParameterListInterpreter [[maybe_unused]] = MueLu::AdaptiveSaMLParameterListInterpreter<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AdaptiveSaMLParameterListInterpreter [[maybe_unused]] = MueLu::AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATIONEXPORTFACTORY_SHORT
-using AggregationExportFactory [[maybe_unused]] = MueLu::AggregationExportFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AggregationExportFactory [[maybe_unused]] = MueLu::AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AGGREGATEQUALITYESTIMATEFACTORY_SHORT
-using AggregateQualityEstimateFactory [[maybe_unused]] = MueLu::AggregateQualityEstimateFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AggregateQualityEstimateFactory [[maybe_unused]] = MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AMALGAMATIONFACTORY_SHORT
-using AmalgamationFactory [[maybe_unused]] = MueLu::AmalgamationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AmalgamationFactory [[maybe_unused]] = MueLu::AmalgamationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AMESOS2SMOOTHER_SHORT
-using Amesos2Smoother [[maybe_unused]] = MueLu::Amesos2Smoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Amesos2Smoother [[maybe_unused]] = MueLu::Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_AMGXOPERATOR_SHORT
-using AMGXOperator [[maybe_unused]] = MueLu::AMGXOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AMGXOperator [[maybe_unused]] = MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ALGEBRAICPERMUTATIONSTRATEGY_SHORT
-using AlgebraicPermutationStrategy [[maybe_unused]] = MueLu::AlgebraicPermutationStrategy<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using AlgebraicPermutationStrategy [[maybe_unused]] = MueLu::AlgebraicPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BELOSSMOOTHER_SHORT
-using BelosSmoother [[maybe_unused]] = MueLu::BelosSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BelosSmoother [[maybe_unused]] = MueLu::BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLACKBOXPFACTORY_SHORT
-using BlackBoxPFactory [[maybe_unused]] = MueLu::BlackBoxPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlackBoxPFactory [[maybe_unused]] = MueLu::BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDCOARSEMAPFACTORY_SHORT
-using BlockedCoarseMapFactory [[maybe_unused]] = MueLu::BlockedCoarseMapFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedCoarseMapFactory [[maybe_unused]] = MueLu::BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_SHORT
-using BlockedCoordinatesTransferFactory [[maybe_unused]] = MueLu::BlockedCoordinatesTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedCoordinatesTransferFactory [[maybe_unused]] = MueLu::BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDDIRECTSOLVER_SHORT
-using BlockedDirectSolver [[maybe_unused]] = MueLu::BlockedDirectSolver<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedDirectSolver [[maybe_unused]] = MueLu::BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_SHORT
-using BlockedGaussSeidelSmoother [[maybe_unused]] = MueLu::BlockedGaussSeidelSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedGaussSeidelSmoother [[maybe_unused]] = MueLu::BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDJACOBISMOOTHER_SHORT
-using BlockedJacobiSmoother [[maybe_unused]] = MueLu::BlockedJacobiSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedJacobiSmoother [[maybe_unused]] = MueLu::BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDPFACTORY_SHORT
-using BlockedPFactory [[maybe_unused]] = MueLu::BlockedPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedPFactory [[maybe_unused]] = MueLu::BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BLOCKEDRAPFACTORY_SHORT
-using BlockedRAPFactory [[maybe_unused]] = MueLu::BlockedRAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BlockedRAPFactory [[maybe_unused]] = MueLu::BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BRICKAGGREGATIONFACTORY_SHORT
-using BrickAggregationFactory [[maybe_unused]] = MueLu::BrickAggregationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BrickAggregationFactory [[maybe_unused]] = MueLu::BrickAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_BRAESSSARAZINSMOOTHER_SHORT
-using BraessSarazinSmoother [[maybe_unused]] = MueLu::BraessSarazinSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using BraessSarazinSmoother [[maybe_unused]] = MueLu::BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CGSOLVER_SHORT
-using CGSolver [[maybe_unused]] = MueLu::CGSolver<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CGSolver [[maybe_unused]] = MueLu::CGSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CLASSICALMAPFACTORY_SHORT
-using ClassicalMapFactory [[maybe_unused]] = MueLu::ClassicalMapFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ClassicalMapFactory [[maybe_unused]] = MueLu::ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CLASSICALPFACTORY_SHORT
-using ClassicalPFactory [[maybe_unused]] = MueLu::ClassicalPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ClassicalPFactory [[maybe_unused]] = MueLu::ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CLONEREPARTITIONINTERFACE_SHORT
-using CloneRepartitionInterface [[maybe_unused]] = MueLu::CloneRepartitionInterface<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CloneRepartitionInterface [[maybe_unused]] = MueLu::CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COALESCEDROPFACTORY_SHORT
-using CoalesceDropFactory [[maybe_unused]] = MueLu::CoalesceDropFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoalesceDropFactory [[maybe_unused]] = MueLu::CoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COALESCEDROPFACTORY_KOKKOS_SHORT
-using CoalesceDropFactory_kokkos [[maybe_unused]] = MueLu::CoalesceDropFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoalesceDropFactory_kokkos [[maybe_unused]] = MueLu::CoalesceDropFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COARSEMAPFACTORY_SHORT
-using CoarseMapFactory [[maybe_unused]] = MueLu::CoarseMapFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoarseMapFactory [[maybe_unused]] = MueLu::CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COARSENINGVISUALIZATIONFACTORY_SHORT
-using CoarseningVisualizationFactory [[maybe_unused]] = MueLu::CoarseningVisualizationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoarseningVisualizationFactory [[maybe_unused]] = MueLu::CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COMBINEPFACTORY_SHORT
-using CombinePFactory [[maybe_unused]] = MueLu::CombinePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CombinePFactory [[maybe_unused]] = MueLu::CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CONSTRAINT_SHORT
-using Constraint [[maybe_unused]] = MueLu::Constraint<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Constraint [[maybe_unused]] = MueLu::Constraint<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_CONSTRAINTFACTORY_SHORT
-using ConstraintFactory [[maybe_unused]] = MueLu::ConstraintFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ConstraintFactory [[maybe_unused]] = MueLu::ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COORDINATESTRANSFERFACTORY_SHORT
-using CoordinatesTransferFactory [[maybe_unused]] = MueLu::CoordinatesTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoordinatesTransferFactory [[maybe_unused]] = MueLu::CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_COUPLEDRBMFACTORY_SHORT
-using CoupledRBMFactory [[maybe_unused]] = MueLu::CoupledRBMFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using CoupledRBMFactory [[maybe_unused]] = MueLu::CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_DEMOFACTORY_SHORT
-using DemoFactory [[maybe_unused]] = MueLu::DemoFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using DemoFactory [[maybe_unused]] = MueLu::DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_DIRECTSOLVER_SHORT
-using DirectSolver [[maybe_unused]] = MueLu::DirectSolver<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using DirectSolver [[maybe_unused]] = MueLu::DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_DROPNEGATIVEENTRIESFACTORY_SHORT
-using DropNegativeEntriesFactory [[maybe_unused]] = MueLu::DropNegativeEntriesFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using DropNegativeEntriesFactory [[maybe_unused]] = MueLu::DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_EMINPFACTORY_SHORT
-using EminPFactory [[maybe_unused]] = MueLu::EminPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using EminPFactory [[maybe_unused]] = MueLu::EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACADEBGS2X2_SHORT
-using FacadeBGS2x2 [[maybe_unused]] = MueLu::FacadeBGS2x2<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FacadeBGS2x2 [[maybe_unused]] = MueLu::FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACADECLASSBASE_SHORT
-using FacadeClassBase [[maybe_unused]] = MueLu::FacadeClassBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FacadeClassBase [[maybe_unused]] = MueLu::FacadeClassBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACADECLASSFACTORY_SHORT
-using FacadeClassFactory [[maybe_unused]] = MueLu::FacadeClassFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FacadeClassFactory [[maybe_unused]] = MueLu::FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACADESIMPLE_SHORT
-using FacadeSimple [[maybe_unused]] = MueLu::FacadeSimple<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FacadeSimple [[maybe_unused]] = MueLu::FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACTORYFACTORY_SHORT
-using FactoryFactory [[maybe_unused]] = MueLu::FactoryFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FactoryFactory [[maybe_unused]] = MueLu::FactoryFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FACTORYMANAGER_SHORT
-using FactoryManager [[maybe_unused]] = MueLu::FactoryManager<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FactoryManager [[maybe_unused]] = MueLu::FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FAKESMOOTHERPROTOTYPE_SHORT
-using FakeSmootherPrototype [[maybe_unused]] = MueLu::FakeSmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FakeSmootherPrototype [[maybe_unused]] = MueLu::FakeSmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FILTEREDAFACTORY_SHORT
-using FilteredAFactory [[maybe_unused]] = MueLu::FilteredAFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FilteredAFactory [[maybe_unused]] = MueLu::FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_FINELEVELINPUTDATAFACTORY_SHORT
-using FineLevelInputDataFactory [[maybe_unused]] = MueLu::FineLevelInputDataFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using FineLevelInputDataFactory [[maybe_unused]] = MueLu::FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GENERALGEOMETRICPFACTORY_SHORT
-using GeneralGeometricPFactory [[maybe_unused]] = MueLu::GeneralGeometricPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using GeneralGeometricPFactory [[maybe_unused]] = MueLu::GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GENERICRFACTORY_SHORT
-using GenericRFactory [[maybe_unused]] = MueLu::GenericRFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using GenericRFactory [[maybe_unused]] = MueLu::GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GEOMETRICINTERPOLATIONPFACTORY_SHORT
-using GeometricInterpolationPFactory [[maybe_unused]] = MueLu::GeometricInterpolationPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using GeometricInterpolationPFactory [[maybe_unused]] = MueLu::GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_SHORT
-using GeometricInterpolationPFactory_kokkos [[maybe_unused]] = MueLu::GeometricInterpolationPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using GeometricInterpolationPFactory_kokkos [[maybe_unused]] = MueLu::GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_GMRESSOLVER_SHORT
-using GMRESSolver [[maybe_unused]] = MueLu::GMRESSolver<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using GMRESSolver [[maybe_unused]] = MueLu::GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_HIERARCHY_SHORT
-using Hierarchy [[maybe_unused]] = MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Hierarchy [[maybe_unused]] = MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_HIERARCHYMANAGER_SHORT
-using HierarchyManager [[maybe_unused]] = MueLu::HierarchyManager<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using HierarchyManager [[maybe_unused]] = MueLu::HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_HIERARCHYFACTORY_SHORT
-using HierarchyFactory [[maybe_unused]] = MueLu::HierarchyFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using HierarchyFactory [[maybe_unused]] = MueLu::HierarchyFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_HIERARCHYUTILS_SHORT
-using HierarchyUtils [[maybe_unused]] = MueLu::HierarchyUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using HierarchyUtils [[maybe_unused]] = MueLu::HierarchyUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INTERFACEAGGREGATIONFACTORY_SHORT
-using InterfaceAggregationFactory [[maybe_unused]] = MueLu::InterfaceAggregationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using InterfaceAggregationFactory [[maybe_unused]] = MueLu::InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_IFPACK2SMOOTHER_SHORT
-using Ifpack2Smoother [[maybe_unused]] = MueLu::Ifpack2Smoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Ifpack2Smoother [[maybe_unused]] = MueLu::Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_SHORT
-using IndefBlockedDiagonalSmoother [[maybe_unused]] = MueLu::IndefBlockedDiagonalSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using IndefBlockedDiagonalSmoother [[maybe_unused]] = MueLu::IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INITIALBLOCKNUMBERFACTORY_SHORT
-using InitialBlockNumberFactory [[maybe_unused]] = MueLu::InitialBlockNumberFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using InitialBlockNumberFactory [[maybe_unused]] = MueLu::InitialBlockNumberFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INTREPIDPCOARSENFACTORY_SHORT
-using IntrepidPCoarsenFactory [[maybe_unused]] = MueLu::IntrepidPCoarsenFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using IntrepidPCoarsenFactory [[maybe_unused]] = MueLu::IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_INVERSEAPPROXIMATIONFACTORY_SHORT
-using InverseApproximationFactory [[maybe_unused]] = MueLu::InverseApproximationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using InverseApproximationFactory [[maybe_unused]] = MueLu::InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LINEDETECTIONFACTORY_SHORT
-using LineDetectionFactory [[maybe_unused]] = MueLu::LineDetectionFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using LineDetectionFactory [[maybe_unused]] = MueLu::LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LOCALPERMUTATIONSTRATEGY_SHORT
-using LocalPermutationStrategy [[maybe_unused]] = MueLu::LocalPermutationStrategy<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using LocalPermutationStrategy [[maybe_unused]] = MueLu::LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_LOWPRECISIONFACTORY_SHORT
-using LowPrecisionFactory [[maybe_unused]] = MueLu::LowPrecisionFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using LowPrecisionFactory [[maybe_unused]] = MueLu::LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MAPTRANSFERFACTORY_SHORT
-using MapTransferFactory [[maybe_unused]] = MueLu::MapTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MapTransferFactory [[maybe_unused]] = MueLu::MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MATRIXANALYSISFACTORY_SHORT
-using MatrixAnalysisFactory [[maybe_unused]] = MueLu::MatrixAnalysisFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MatrixAnalysisFactory [[maybe_unused]] = MueLu::MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MERGEDBLOCKEDMATRIXFACTORY_SHORT
-using MergedBlockedMatrixFactory [[maybe_unused]] = MueLu::MergedBlockedMatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MergedBlockedMatrixFactory [[maybe_unused]] = MueLu::MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MERGEDSMOOTHER_SHORT
-using MergedSmoother [[maybe_unused]] = MueLu::MergedSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MergedSmoother [[maybe_unused]] = MueLu::MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MLPARAMETERLISTINTERPRETER_SHORT
-using MLParameterListInterpreter [[maybe_unused]] = MueLu::MLParameterListInterpreter<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MLParameterListInterpreter [[maybe_unused]] = MueLu::MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MULTIVECTORTRANSFERFACTORY_SHORT
-using MultiVectorTransferFactory [[maybe_unused]] = MueLu::MultiVectorTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MultiVectorTransferFactory [[maybe_unused]] = MueLu::MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_NOTAYAGGREGATIONFACTORY_SHORT
-using NotayAggregationFactory [[maybe_unused]] = MueLu::NotayAggregationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using NotayAggregationFactory [[maybe_unused]] = MueLu::NotayAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_NULLSPACEFACTORY_SHORT
-using NullspaceFactory [[maybe_unused]] = MueLu::NullspaceFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using NullspaceFactory [[maybe_unused]] = MueLu::NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_NULLSPACEFACTORY_KOKKOS_SHORT
-using NullspaceFactory_kokkos [[maybe_unused]] = MueLu::NullspaceFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using NullspaceFactory_kokkos [[maybe_unused]] = MueLu::NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_NULLSPACEPRESMOOTHFACTORY_SHORT
-using NullspacePresmoothFactory [[maybe_unused]] = MueLu::NullspacePresmoothFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using NullspacePresmoothFactory [[maybe_unused]] = MueLu::NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PARAMETERLISTINTERPRETER_SHORT
-using ParameterListInterpreter [[maybe_unused]] = MueLu::ParameterListInterpreter<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ParameterListInterpreter [[maybe_unused]] = MueLu::ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PATTERNFACTORY_SHORT
-using PatternFactory [[maybe_unused]] = MueLu::PatternFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PatternFactory [[maybe_unused]] = MueLu::PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PERFUTILS_SHORT
-using PerfUtils [[maybe_unused]] = MueLu::PerfUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PerfUtils [[maybe_unused]] = MueLu::PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PERFMODELS_SHORT
-using PerfModels [[maybe_unused]] = MueLu::PerfModels<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PerfModels [[maybe_unused]] = MueLu::PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PERMUTATIONFACTORY_SHORT
-using PermutationFactory [[maybe_unused]] = MueLu::PermutationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PermutationFactory [[maybe_unused]] = MueLu::PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PERMUTINGSMOOTHER_SHORT
-using PermutingSmoother [[maybe_unused]] = MueLu::PermutingSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PermutingSmoother [[maybe_unused]] = MueLu::PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PGPFACTORY_SHORT
-using PgPFactory [[maybe_unused]] = MueLu::PgPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PgPFactory [[maybe_unused]] = MueLu::PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PREDROPFUNCTIONBASECLASS_SHORT
-using PreDropFunctionBaseClass [[maybe_unused]] = MueLu::PreDropFunctionBaseClass<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PreDropFunctionBaseClass [[maybe_unused]] = MueLu::PreDropFunctionBaseClass<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
-using PreDropFunctionConstVal [[maybe_unused]] = MueLu::PreDropFunctionConstVal<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using PreDropFunctionConstVal [[maybe_unused]] = MueLu::PreDropFunctionConstVal<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_PROJECTORSMOOTHER_SHORT
-using ProjectorSmoother [[maybe_unused]] = MueLu::ProjectorSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ProjectorSmoother [[maybe_unused]] = MueLu::ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_RAPFACTORY_SHORT
-using RAPFactory [[maybe_unused]] = MueLu::RAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RAPFactory [[maybe_unused]] = MueLu::RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_RAPSHIFTFACTORY_SHORT
-using RAPShiftFactory [[maybe_unused]] = MueLu::RAPShiftFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RAPShiftFactory [[maybe_unused]] = MueLu::RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCEACFACTORY_SHORT
-using RebalanceAcFactory [[maybe_unused]] = MueLu::RebalanceAcFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceAcFactory [[maybe_unused]] = MueLu::RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCEBLOCKACFACTORY_SHORT
-using RebalanceBlockAcFactory [[maybe_unused]] = MueLu::RebalanceBlockAcFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceBlockAcFactory [[maybe_unused]] = MueLu::RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_SHORT
-using RebalanceBlockInterpolationFactory [[maybe_unused]] = MueLu::RebalanceBlockInterpolationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceBlockInterpolationFactory [[maybe_unused]] = MueLu::RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_SHORT
-using RebalanceBlockRestrictionFactory [[maybe_unused]] = MueLu::RebalanceBlockRestrictionFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceBlockRestrictionFactory [[maybe_unused]] = MueLu::RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REBALANCETRANSFERFACTORY_SHORT
-using RebalanceTransferFactory [[maybe_unused]] = MueLu::RebalanceTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RebalanceTransferFactory [[maybe_unused]] = MueLu::RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REFMAXWELLSMOOTHER_SHORT
-using RefMaxwellSmoother [[maybe_unused]] = MueLu::RefMaxwellSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RefMaxwellSmoother [[maybe_unused]] = MueLu::RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REGIONRFACTORY_SHORT
-using RegionRFactory [[maybe_unused]] = MueLu::RegionRFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RegionRFactory [[maybe_unused]] = MueLu::RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REGIONRFACTORY_KOKKOS_SHORT
-using RegionRFactory_kokkos [[maybe_unused]] = MueLu::RegionRFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RegionRFactory_kokkos [[maybe_unused]] = MueLu::RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REITZINGERPFACTORY_SHORT
-using ReitzingerPFactory [[maybe_unused]] = MueLu::ReitzingerPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ReitzingerPFactory [[maybe_unused]] = MueLu::ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REORDERBLOCKAFACTORY_SHORT
-using ReorderBlockAFactory [[maybe_unused]] = MueLu::ReorderBlockAFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ReorderBlockAFactory [[maybe_unused]] = MueLu::ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REPARTITIONFACTORY_SHORT
-using RepartitionFactory [[maybe_unused]] = MueLu::RepartitionFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RepartitionFactory [[maybe_unused]] = MueLu::RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REPARTITIONBLOCKDIAGONALFACTORY_SHORT
-using RepartitionBlockDiagonalFactory [[maybe_unused]] = MueLu::RepartitionBlockDiagonalFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RepartitionBlockDiagonalFactory [[maybe_unused]] = MueLu::RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REPARTITIONHEURISTICFACTORY_SHORT
-using RepartitionHeuristicFactory [[maybe_unused]] = MueLu::RepartitionHeuristicFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RepartitionHeuristicFactory [[maybe_unused]] = MueLu::RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REPLICATEPFACTORY_SHORT
-using ReplicatePFactory [[maybe_unused]] = MueLu::ReplicatePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ReplicatePFactory [[maybe_unused]] = MueLu::ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_RIGIDBODYMODEFACTORY_SHORT
-using RigidBodyModeFactory [[maybe_unused]] = MueLu::RigidBodyModeFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RigidBodyModeFactory [[maybe_unused]] = MueLu::RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SAPFACTORY_SHORT
-using SaPFactory [[maybe_unused]] = MueLu::SaPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SaPFactory [[maybe_unused]] = MueLu::SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SAPFACTORY_KOKKOS_SHORT
-using SaPFactory_kokkos [[maybe_unused]] = MueLu::SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SaPFactory_kokkos [[maybe_unused]] = MueLu::SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SCALEDNULLSPACEFACTORY_SHORT
-using ScaledNullspaceFactory [[maybe_unused]] = MueLu::ScaledNullspaceFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ScaledNullspaceFactory [[maybe_unused]] = MueLu::ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SCHURCOMPLEMENTFACTORY_SHORT
-using SchurComplementFactory [[maybe_unused]] = MueLu::SchurComplementFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SchurComplementFactory [[maybe_unused]] = MueLu::SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SEGREGATEDAFACTORY_SHORT
-using SegregatedAFactory [[maybe_unused]] = MueLu::SegregatedAFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SegregatedAFactory [[maybe_unused]] = MueLu::SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SHIFTEDLAPLACIAN_SHORT
-using ShiftedLaplacian [[maybe_unused]] = MueLu::ShiftedLaplacian<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ShiftedLaplacian [[maybe_unused]] = MueLu::ShiftedLaplacian<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SHIFTEDLAPLACIANOPERATOR_SHORT
-using ShiftedLaplacianOperator [[maybe_unused]] = MueLu::ShiftedLaplacianOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ShiftedLaplacianOperator [[maybe_unused]] = MueLu::ShiftedLaplacianOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SIMPLESMOOTHER_SHORT
-using SimpleSmoother [[maybe_unused]] = MueLu::SimpleSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SimpleSmoother [[maybe_unused]] = MueLu::SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SMOOTHER_SHORT
-using Smoother [[maybe_unused]] = MueLu::Smoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Smoother [[maybe_unused]] = MueLu::Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SMOOTHERBASE_SHORT
-using SmootherBase [[maybe_unused]] = MueLu::SmootherBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SmootherBase [[maybe_unused]] = MueLu::SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SMOOTHERFACTORY_SHORT
-using SmootherFactory [[maybe_unused]] = MueLu::SmootherFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SmootherFactory [[maybe_unused]] = MueLu::SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SMOOTHERPROTOTYPE_SHORT
-using SmootherPrototype [[maybe_unused]] = MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SmootherPrototype [[maybe_unused]] = MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SMOOVECCOALESCEDROPFACTORY_SHORT
-using SmooVecCoalesceDropFactory [[maybe_unused]] = MueLu::SmooVecCoalesceDropFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SmooVecCoalesceDropFactory [[maybe_unused]] = MueLu::SmooVecCoalesceDropFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SOLVERBASE_SHORT
-using SolverBase [[maybe_unused]] = MueLu::SolverBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SolverBase [[maybe_unused]] = MueLu::SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_STEEPESTDESCENTSOLVER_SHORT
-using SteepestDescentSolver [[maybe_unused]] = MueLu::SteepestDescentSolver<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SteepestDescentSolver [[maybe_unused]] = MueLu::SteepestDescentSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_STRATIMIKOSSMOOTHER_SHORT
-using StratimikosSmoother [[maybe_unused]] = MueLu::StratimikosSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using StratimikosSmoother [[maybe_unused]] = MueLu::StratimikosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_STRUCTUREDAGGREGATIONFACTORY_SHORT
-using StructuredAggregationFactory [[maybe_unused]] = MueLu::StructuredAggregationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using StructuredAggregationFactory [[maybe_unused]] = MueLu::StructuredAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_STRUCTUREDLINEDETECTIONFACTORY_SHORT
-using StructuredLineDetectionFactory [[maybe_unused]] = MueLu::StructuredLineDetectionFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using StructuredLineDetectionFactory [[maybe_unused]] = MueLu::StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SUBBLOCKAFACTORY_SHORT
-using SubBlockAFactory [[maybe_unused]] = MueLu::SubBlockAFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SubBlockAFactory [[maybe_unused]] = MueLu::SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TEKOSMOOTHER_SHORT
-using TekoSmoother [[maybe_unused]] = MueLu::TekoSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TekoSmoother [[maybe_unused]] = MueLu::TekoSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TENTATIVEPFACTORY_SHORT
-using TentativePFactory [[maybe_unused]] = MueLu::TentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TentativePFactory [[maybe_unused]] = MueLu::TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
-using TentativePFactory_kokkos [[maybe_unused]] = MueLu::TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TentativePFactory_kokkos [[maybe_unused]] = MueLu::TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MATRIXFREETENTATIVEP_SHORT
-using MatrixFreeTentativeP [[maybe_unused]] = MueLu::MatrixFreeTentativeP<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MatrixFreeTentativeP [[maybe_unused]] = MueLu::MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
-using MatrixFreeTentativePFactory [[maybe_unused]] = MueLu::MatrixFreeTentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MatrixFreeTentativePFactory [[maybe_unused]] = MueLu::MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_THRESHOLDAFILTERFACTORY_SHORT
-using ThresholdAFilterFactory [[maybe_unused]] = MueLu::ThresholdAFilterFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ThresholdAFilterFactory [[maybe_unused]] = MueLu::ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TOGGLECOORDINATESTRANSFERFACTORY_SHORT
-using ToggleCoordinatesTransferFactory [[maybe_unused]] = MueLu::ToggleCoordinatesTransferFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ToggleCoordinatesTransferFactory [[maybe_unused]] = MueLu::ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TOGGLEPFACTORY_SHORT
-using TogglePFactory [[maybe_unused]] = MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TogglePFactory [[maybe_unused]] = MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TOPRAPFACTORY_SHORT
-using TopRAPFactory [[maybe_unused]] = MueLu::TopRAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TopRAPFactory [[maybe_unused]] = MueLu::TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TOPSMOOTHERFACTORY_SHORT
-using TopSmootherFactory [[maybe_unused]] = MueLu::TopSmootherFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TopSmootherFactory [[maybe_unused]] = MueLu::TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TPETRAOPERATOR_SHORT
-using TpetraOperator [[maybe_unused]] = MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TpetraOperator [[maybe_unused]] = MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TRANSPFACTORY_SHORT
-using TransPFactory [[maybe_unused]] = MueLu::TransPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TransPFactory [[maybe_unused]] = MueLu::TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_RFROMP_OR_TRANSP_SHORT
-using RfromP_Or_TransP [[maybe_unused]] = MueLu::RfromP_Or_TransP<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RfromP_Or_TransP [[maybe_unused]] = MueLu::RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TRILINOSSMOOTHER_SHORT
-using TrilinosSmoother [[maybe_unused]] = MueLu::TrilinosSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using TrilinosSmoother [[maybe_unused]] = MueLu::TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UNSMOOSHFACTORY_SHORT
-using UnsmooshFactory [[maybe_unused]] = MueLu::UnsmooshFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using UnsmooshFactory [[maybe_unused]] = MueLu::UnsmooshFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_USERPFACTORY_SHORT
-using UserPFactory [[maybe_unused]] = MueLu::UserPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using UserPFactory [[maybe_unused]] = MueLu::UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UTILITIES_SHORT
-using Utilities [[maybe_unused]] = MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Utilities [[maybe_unused]] = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UTILITIESBASE_SHORT
-using UtilitiesBase [[maybe_unused]] = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using UtilitiesBase [[maybe_unused]] = MueLu::UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_VARIABLEDOFLAPLACIANFACTORY_SHORT
-using VariableDofLaplacianFactory [[maybe_unused]] = MueLu::VariableDofLaplacianFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using VariableDofLaplacianFactory [[maybe_unused]] = MueLu::VariableDofLaplacianFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SEMICOARSENPFACTORY_SHORT
-using SemiCoarsenPFactory [[maybe_unused]] = MueLu::SemiCoarsenPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SemiCoarsenPFactory [[maybe_unused]] = MueLu::SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
-using SemiCoarsenPFactory_kokkos [[maybe_unused]] = MueLu::SemiCoarsenPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using SemiCoarsenPFactory_kokkos [[maybe_unused]] = MueLu::SemiCoarsenPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_UZAWASMOOTHER_SHORT
-using UzawaSmoother [[maybe_unused]] = MueLu::UzawaSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using UzawaSmoother [[maybe_unused]] = MueLu::UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_VISUALIZATIONHELPERS_SHORT
-using VisualizationHelpers [[maybe_unused]] = MueLu::VisualizationHelpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using VisualizationHelpers [[maybe_unused]] = MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ZEROSUBBLOCKAFACTORY_SHORT
-using ZeroSubBlockAFactory [[maybe_unused]] = MueLu::ZeroSubBlockAFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ZeroSubBlockAFactory [[maybe_unused]] = MueLu::ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ZOLTANINTERFACE_SHORT
-using ZoltanInterface [[maybe_unused]] = MueLu::ZoltanInterface<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using ZoltanInterface [[maybe_unused]] = MueLu::ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_ZOLTAN2INTERFACE_SHORT
-using Zoltan2Interface [[maybe_unused]] = MueLu::Zoltan2Interface<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Zoltan2Interface [[maybe_unused]] = MueLu::Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_NODEPARTITIONINTERFACE_SHORT
-using NodePartitionInterface [[maybe_unused]] = MueLu::NodePartitionInterface<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using NodePartitionInterface [[maybe_unused]] = MueLu::NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_XPETRAOPERATOR_SHORT
-using XpetraOperator [[maybe_unused]] = MueLu::XpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using XpetraOperator [[maybe_unused]] = MueLu::XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_REFMAXWELL_SHORT
-using RefMaxwell [[maybe_unused]] = MueLu::RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using RefMaxwell [[maybe_unused]] = MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MAXWELL1_SHORT
-using Maxwell1 [[maybe_unused]] = MueLu::Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Maxwell1 [[maybe_unused]] = MueLu::Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MULTIPHYS_SHORT
-using MultiPhys [[maybe_unused]] = MueLu::MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using MultiPhys [[maybe_unused]] = MueLu::MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_MAXWELL_UTILS_SHORT
-using Maxwell_Utils [[maybe_unused]] = MueLu::Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+using Maxwell_Utils [[maybe_unused]] = MueLu::Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #endif
 #ifdef MUELU_TWOLEVELMATLABFACTORY_SHORT
-typedef MueLu::TwoLevelMatlabFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> TwoLevelMatlabFactory;
+typedef MueLu::TwoLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> TwoLevelMatlabFactory;
 #endif
 #ifdef MUELU_SINGLELEVELMATLABFACTORY_SHORT
-typedef MueLu::SingleLevelMatlabFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> SingleLevelMatlabFactory;
+typedef MueLu::SingleLevelMatlabFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> SingleLevelMatlabFactory;
 #endif
 #ifdef MUELU_MATLABSMOOTHER_SHORT
-typedef MueLu::MatlabSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node> MatlabSmoother;
+typedef MueLu::MatlabSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node> MatlabSmoother;
 #endif
diff --git a/packages/muelu/src/Headers/MueLu_Version.hpp b/packages/muelu/src/Headers/MueLu_Version.hpp
index 0fbda392c252..6eb87ab59a83 100644
--- a/packages/muelu/src/Headers/MueLu_Version.hpp
+++ b/packages/muelu/src/Headers/MueLu_Version.hpp
@@ -55,10 +55,10 @@
 
 namespace MueLu {
 
-  inline std::string const Version() {
-    return("MueLu development");
-  }
+inline std::string const Version() {
+  return ("MueLu development");
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif //ifndef MUELU_VERSION_HPP
+#endif  // ifndef MUELU_VERSION_HPP
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_decl.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_decl.hpp
index 8885d500ed3c..d201111eb8b2 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_decl.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_decl.hpp
@@ -55,37 +55,34 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FacadeBGS2x2 : public FacadeClassBase<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FacadeBGS2x2 : public FacadeClassBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Constructor.
-    FacadeBGS2x2();
+  //! Constructor.
+  FacadeBGS2x2();
 
-    //! Destructor.
-    virtual ~FacadeBGS2x2() { }
+  //! Destructor.
+  virtual ~FacadeBGS2x2() {}
 
-    //@}
+  //@}
 
-    /*! @brief Set parameter list for FacadeClass interpreter.
+  /*! @brief Set parameter list for FacadeClass interpreter.
 
-        @param[in] paramList: ParameterList containing the MueLu parameters for chosen facade class.
-    */
-    Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
-
-  private:
-
-  };
-
-} // namespace MueLu
+      @param[in] paramList: ParameterList containing the MueLu parameters for chosen facade class.
+  */
+  Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
 
+ private:
+};
 
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_BGS2x2_DECL_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_def.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_def.hpp
index f96866f4ee6e..6c38b3f54cbd 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_def.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeBGS2x2_def.hpp
@@ -50,349 +50,343 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
-
 #include "MueLu_Exceptions.hpp"
 
 #include "MueLu_FacadeBGS2x2_decl.hpp"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeBGS2x2() {
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::ParameterList> FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
-
-    // obtain ParameterList with default input parameters for this facade class
-    // Note all parameters are of type string (we use it for string replacement)
-    std::string defaultString =
-"<ParameterList name=\"Input\">"
-"<Parameter name=\"MueLu preconditioner\" type=\"string\" value=\"undefined\"/>"
-"<Parameter name=\"Block 1: dofs per node\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 2: dofs per node\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 1: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
-  "<Parameter name=\"Block 1: level-of-fill\" type=\"int\" value=\"0\"/>"
-  "<Parameter name=\"Block 1: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 1: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"Block 1: transfer smoothing\" type=\"bool\" value=\"true\"/>"
-  "<Parameter name=\"Block 2: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
-  "<Parameter name=\"Block 2: level-of-fill\" type=\"int\" value=\"0\"/>"
-  "<Parameter name=\"Block 2: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 2: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"Block 2: transfer smoothing\" type=\"bool\" value=\"true\"/>"
-  "<Parameter name=\"BGS: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"max levels\" type=\"int\" value=\"5\"/>"
-  "<Parameter name=\"coarse: max size\" type=\"int\" value=\"25000\"/>"
-  "<Parameter name=\"verbosity\" type=\"string\" value=\"High\"/>"
-  "</ParameterList>"
-;
-    Teuchos::RCP<ParameterList> defaultList = Teuchos::getParametersFromXmlString(defaultString);
-    // validate user input parameters (and set defaults if necessary)
-    Teuchos::ParameterList inputParameters = paramList;
-    inputParameters.validateParametersAndSetDefaults(*defaultList);
-    TEUCHOS_TEST_FOR_EXCEPTION(inputParameters.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeBGS2x2: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
-
-    // create copy of template string which is updated with in-place string replacements
-    // template string for preconditioner layout (factory based parameters)
-    std::string finalString =
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeBGS2x2() {
+}
 
-"<ParameterList name=\"MueLu\">"
-"  <ParameterList name=\"Factories\">"
-"    <ParameterList name=\"mySubBlockAFactory1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
-"      <Parameter name=\"block row\"                 type=\"int\"     value=\"0\"/>"
-"      <Parameter name=\"block col\"                 type=\"int\"     value=\"0\"/>"
-"      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myAggFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"UncoupledAggregationFactory\"/>"
-"      <Parameter name=\"aggregation: min agg size\" type=\"int\" value=\"5\"/>"
-"      <Parameter name=\"aggregation: max selected neighbors\" type=\"int\" value=\"1\"/>"
-"    </ParameterList>"
-""
-"    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
-"    <ParameterList name=\"myCoarseMap1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
-"      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myTentativePFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myPFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"PgPFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myRFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
-"    </ParameterList>"
-""
-"    <!-- We have to use Nullspace1 here. If \"Nullspace1\" is not set the"
-"         Factory creates the default null space containing of constant"
-"         vectors -->"
-"    <ParameterList name=\"myNspFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
-"      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace1\"/>"
-"      <Parameter name=\"Nullspace1\" type=\"string\" value=\"myTentativePFact1\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK 2 (for submatrix A_{11}) PRESSURE PART -->"
-"    <ParameterList name=\"mySubBlockAFactory2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
-"      <Parameter name=\"block row\"                 type=\"int\"     value=\"1\"/>"
-"      <Parameter name=\"block col\"                 type=\"int\"     value=\"1\"/>"
-"      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"    </ParameterList>"
-""
-"    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
-"    <ParameterList name=\"myCoarseMap2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
-"      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myTentativePFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myPFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SaPFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myRFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TransPFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myNspFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
-"      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace2\"/>"
-"      <Parameter name=\"Nullspace2\" type=\"string\" value=\"myTentativePFact2\"/>"
-"    </ParameterList>"
-""
-"    <!-- FACTORY MANAGERS -->"
-""
-"    <!-- Multigrid setup for velocity block (A_{00}) -->"
-"    <ParameterList name=\"myFirstGroup\">"
-"      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myRFact1\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
-"    </ParameterList>"
-""
-"    <!-- Multigrid setup for pressure block (A_{11}) -->"
-"    <ParameterList name=\"mySecondGroup\">"
-"      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myRFact2\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/><!-- reuse aggs from PRESSURE block! -->"
-"      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact2\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK TRANSFER operators -->"
-""
-"    <ParameterList name=\"myBlockedPFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"BlockedPFactory\"/>"
-"      <!-- factory manager for block 1 -->"
-"      <ParameterList name=\"block1\">"
-"         <Parameter name=\"group\" type=\"string\" value=\"myFirstGroup\"/>"
-"      </ParameterList>"
-"      <!-- factory manager for block 2 -->"
-"      <ParameterList name=\"block2\">"
-"        <Parameter name=\"group\" type=\"string\" value=\"mySecondGroup\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockedRFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockedRAPFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"BlockedRAPFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myBlockedRFact\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK SMOOTHERS -->"
-"    <ParameterList name=\"mySmooFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 1: relaxation: typeYYY\"/>"
-"        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 1: relaxation: sweepsYYY\"/>"
-"        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 1: relaxation: damping factorYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooILUFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 1: level-of-fillYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooDirectFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 2: relaxation: typeYYY\"/>"
-"        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 2: relaxation: sweepsYYY\"/>"
-"        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 2: relaxation: damping factorYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooILUFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 2: level-of-fillYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooDirectFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockSmoother\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"BlockedGaussSeidelSmoother\"/>"
-"      <Parameter name=\"Sweeps\" type=\"int\" value=\"1\"/>"
-"      <Parameter name=\"Damping factor\" type=\"double\" value=\"XXXBGS: damping factorYYY\"/>"
-"      <!-- factory manager for block 1 -->"
-"      <ParameterList name=\"block1\">"
-"        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother1XYZ\"/>"
-"      </ParameterList>"
-"      <!-- factory manager for block 2 -->"
-"      <ParameterList name=\"block2\">"
-"        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother2XYZ\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"  </ParameterList>"
-"  <!-- end Factories -->"
-""
-"  <!-- Definition of the multigrid preconditioner -->"
-"  <ParameterList name=\"Hierarchy\">"
-""
-"    <Parameter name=\"max levels\"          type=\"int\"      value=\"XXXmax levelsYYY\"/>"
-"    <Parameter name=\"coarse: max size\"    type=\"int\"      value=\"XXXcoarse: max sizeYYY\"/>"
-"    <Parameter name=\"verbosity\"           type=\"string\"   value=\"XXXverbosityYYY\"/>"
-""
-"    <ParameterList name=\"AllLevel\">"
-"      <Parameter name=\"startLevel\"        type=\"int\"      value=\"0\"/>"
-"      <Parameter name=\"Smoother\"          type=\"string\"   value=\"myBlockSmoother\"/>"
-"      <Parameter name=\"CoarseSolver\"      type=\"string\"   value=\"myBlockSmoother\"/>"
-"      <Parameter name=\"P\"                 type=\"string\"   value=\"myBlockedPFact\"/>"
-"      <Parameter name=\"R\"                 type=\"string\"   value=\"myBlockedRFact\"/>"
-"      <Parameter name=\"A\"                 type=\"string\"   value=\"myBlockedRAPFact\"/>"
-"    </ParameterList>"
-""
-"  </ParameterList>"
-"</ParameterList>"
-  ;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::ParameterList> FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
+  // obtain ParameterList with default input parameters for this facade class
+  // Note all parameters are of type string (we use it for string replacement)
+  std::string defaultString =
+      "<ParameterList name=\"Input\">"
+      "<Parameter name=\"MueLu preconditioner\" type=\"string\" value=\"undefined\"/>"
+      "<Parameter name=\"Block 1: dofs per node\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 2: dofs per node\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 1: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
+      "<Parameter name=\"Block 1: level-of-fill\" type=\"int\" value=\"0\"/>"
+      "<Parameter name=\"Block 1: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 1: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"Block 1: transfer smoothing\" type=\"bool\" value=\"true\"/>"
+      "<Parameter name=\"Block 2: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
+      "<Parameter name=\"Block 2: level-of-fill\" type=\"int\" value=\"0\"/>"
+      "<Parameter name=\"Block 2: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 2: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"Block 2: transfer smoothing\" type=\"bool\" value=\"true\"/>"
+      "<Parameter name=\"BGS: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"max levels\" type=\"int\" value=\"5\"/>"
+      "<Parameter name=\"coarse: max size\" type=\"int\" value=\"25000\"/>"
+      "<Parameter name=\"verbosity\" type=\"string\" value=\"High\"/>"
+      "</ParameterList>";
+  Teuchos::RCP<ParameterList> defaultList = Teuchos::getParametersFromXmlString(defaultString);
+  // validate user input parameters (and set defaults if necessary)
+  Teuchos::ParameterList inputParameters = paramList;
+  inputParameters.validateParametersAndSetDefaults(*defaultList);
+  TEUCHOS_TEST_FOR_EXCEPTION(inputParameters.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeBGS2x2: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
 
-    // logical code for more complicated distinctions
+  // create copy of template string which is updated with in-place string replacements
+  // template string for preconditioner layout (factory based parameters)
+  std::string finalString =
 
+      "<ParameterList name=\"MueLu\">"
+      "  <ParameterList name=\"Factories\">"
+      "    <ParameterList name=\"mySubBlockAFactory1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
+      "      <Parameter name=\"block row\"                 type=\"int\"     value=\"0\"/>"
+      "      <Parameter name=\"block col\"                 type=\"int\"     value=\"0\"/>"
+      "      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myAggFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"UncoupledAggregationFactory\"/>"
+      "      <Parameter name=\"aggregation: min agg size\" type=\"int\" value=\"5\"/>"
+      "      <Parameter name=\"aggregation: max selected neighbors\" type=\"int\" value=\"1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
+      "    <ParameterList name=\"myCoarseMap1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
+      "      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myTentativePFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myPFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"PgPFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myRFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- We have to use Nullspace1 here. If \"Nullspace1\" is not set the"
+      "         Factory creates the default null space containing of constant"
+      "         vectors -->"
+      "    <ParameterList name=\"myNspFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
+      "      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace1\"/>"
+      "      <Parameter name=\"Nullspace1\" type=\"string\" value=\"myTentativePFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK 2 (for submatrix A_{11}) PRESSURE PART -->"
+      "    <ParameterList name=\"mySubBlockAFactory2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
+      "      <Parameter name=\"block row\"                 type=\"int\"     value=\"1\"/>"
+      "      <Parameter name=\"block col\"                 type=\"int\"     value=\"1\"/>"
+      "      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
+      "    <ParameterList name=\"myCoarseMap2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
+      "      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myTentativePFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myPFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SaPFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myRFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TransPFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myNspFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
+      "      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace2\"/>"
+      "      <Parameter name=\"Nullspace2\" type=\"string\" value=\"myTentativePFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- FACTORY MANAGERS -->"
+      ""
+      "    <!-- Multigrid setup for velocity block (A_{00}) -->"
+      "    <ParameterList name=\"myFirstGroup\">"
+      "      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myRFact1\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- Multigrid setup for pressure block (A_{11}) -->"
+      "    <ParameterList name=\"mySecondGroup\">"
+      "      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myRFact2\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/><!-- reuse aggs from PRESSURE block! -->"
+      "      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact2\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK TRANSFER operators -->"
+      ""
+      "    <ParameterList name=\"myBlockedPFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"BlockedPFactory\"/>"
+      "      <!-- factory manager for block 1 -->"
+      "      <ParameterList name=\"block1\">"
+      "         <Parameter name=\"group\" type=\"string\" value=\"myFirstGroup\"/>"
+      "      </ParameterList>"
+      "      <!-- factory manager for block 2 -->"
+      "      <ParameterList name=\"block2\">"
+      "        <Parameter name=\"group\" type=\"string\" value=\"mySecondGroup\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockedRFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockedRAPFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"BlockedRAPFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myBlockedRFact\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK SMOOTHERS -->"
+      "    <ParameterList name=\"mySmooFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 1: relaxation: typeYYY\"/>"
+      "        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 1: relaxation: sweepsYYY\"/>"
+      "        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 1: relaxation: damping factorYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooILUFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 1: level-of-fillYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooDirectFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 2: relaxation: typeYYY\"/>"
+      "        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 2: relaxation: sweepsYYY\"/>"
+      "        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 2: relaxation: damping factorYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooILUFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 2: level-of-fillYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooDirectFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockSmoother\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"BlockedGaussSeidelSmoother\"/>"
+      "      <Parameter name=\"Sweeps\" type=\"int\" value=\"1\"/>"
+      "      <Parameter name=\"Damping factor\" type=\"double\" value=\"XXXBGS: damping factorYYY\"/>"
+      "      <!-- factory manager for block 1 -->"
+      "      <ParameterList name=\"block1\">"
+      "        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother1XYZ\"/>"
+      "      </ParameterList>"
+      "      <!-- factory manager for block 2 -->"
+      "      <ParameterList name=\"block2\">"
+      "        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother2XYZ\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "  </ParameterList>"
+      "  <!-- end Factories -->"
+      ""
+      "  <!-- Definition of the multigrid preconditioner -->"
+      "  <ParameterList name=\"Hierarchy\">"
+      ""
+      "    <Parameter name=\"max levels\"          type=\"int\"      value=\"XXXmax levelsYYY\"/>"
+      "    <Parameter name=\"coarse: max size\"    type=\"int\"      value=\"XXXcoarse: max sizeYYY\"/>"
+      "    <Parameter name=\"verbosity\"           type=\"string\"   value=\"XXXverbosityYYY\"/>"
+      ""
+      "    <ParameterList name=\"AllLevel\">"
+      "      <Parameter name=\"startLevel\"        type=\"int\"      value=\"0\"/>"
+      "      <Parameter name=\"Smoother\"          type=\"string\"   value=\"myBlockSmoother\"/>"
+      "      <Parameter name=\"CoarseSolver\"      type=\"string\"   value=\"myBlockSmoother\"/>"
+      "      <Parameter name=\"P\"                 type=\"string\"   value=\"myBlockedPFact\"/>"
+      "      <Parameter name=\"R\"                 type=\"string\"   value=\"myBlockedRFact\"/>"
+      "      <Parameter name=\"A\"                 type=\"string\"   value=\"myBlockedRAPFact\"/>"
+      "    </ParameterList>"
+      ""
+      "  </ParameterList>"
+      "</ParameterList>";
 
-    std::string smoother1 = inputParameters.get<std::string>("Block 1: smoother");
-    if(smoother1 == "ILU") {
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooILUFact1");
-    } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "SGS") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Symmetric Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "GS") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Jacobi") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Jacobi");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Direct") {
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooDirectFact1");
-    } else {
-      this->GetOStream(Errors) << "Invalid smoother type for block 1: " << smoother1 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
-    }
+  // logical code for more complicated distinctions
 
-    std::string smoother2 = inputParameters.get<std::string>("Block 2: smoother");
-    if(smoother2 == "ILU") {
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooILUFact2");
-    } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "SGS") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Symmetric Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "GS") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Jacobi") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Direct") {
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooDirectFact2");
-    } else {
-      this->GetOStream(Errors) << "Invalid smoother type for block 2: " << smoother2 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
-    }
+  std::string smoother1 = inputParameters.get<std::string>("Block 1: smoother");
+  if (smoother1 == "ILU") {
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooILUFact1");
+  } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "SGS") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Symmetric Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "GS") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Jacobi") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Jacobi");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Direct") {
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooDirectFact1");
+  } else {
+    this->GetOStream(Errors) << "Invalid smoother type for block 1: " << smoother1 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
+  }
 
-    if(inputParameters.get<bool>("Block 1: transfer smoothing") == true) {
-      this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myPFact1");
-      this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myRFact1");
-    } else {
-      this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myTentativePFact1");
-      this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myTransPFact1");
-    }
-    if(inputParameters.get<bool>("Block 2: transfer smoothing") == true) {
-      this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myPFact2");
-      this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myRFact2");
-    } else {
-      this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myTentativePFact2");
-      this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myTransPFact2");
-    }
+  std::string smoother2 = inputParameters.get<std::string>("Block 2: smoother");
+  if (smoother2 == "ILU") {
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooILUFact2");
+  } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "SGS") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Symmetric Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "GS") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Jacobi") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Direct") {
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooDirectFact2");
+  } else {
+    this->GetOStream(Errors) << "Invalid smoother type for block 2: " << smoother2 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
+  }
 
-    // end logical code
+  if (inputParameters.get<bool>("Block 1: transfer smoothing") == true) {
+    this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myPFact1");
+    this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myRFact1");
+  } else {
+    this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myTentativePFact1");
+    this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myTransPFact1");
+  }
+  if (inputParameters.get<bool>("Block 2: transfer smoothing") == true) {
+    this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myPFact2");
+    this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myRFact2");
+  } else {
+    this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myTentativePFact2");
+    this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myTransPFact2");
+  }
 
-    // loop over all input parameters
-    for(Teuchos::ParameterList::ConstIterator it = inputParameters.begin(); it != inputParameters.end(); it++) {
-      // form replacement string
-      std::string par_name = inputParameters.name(it);
-      std::stringstream ss;
-      ss << "XXX" << par_name << "YYY";
+  // end logical code
 
-      // update final string with parameters
-      Teuchos::ParameterEntry par_entry = inputParameters.entry(it);
-      this->ReplaceString(finalString,
-              ss.str(), Teuchos::toString(par_entry.getAny()));
-    }
+  // loop over all input parameters
+  for (Teuchos::ParameterList::ConstIterator it = inputParameters.begin(); it != inputParameters.end(); it++) {
+    // form replacement string
+    std::string par_name = inputParameters.name(it);
+    std::stringstream ss;
+    ss << "XXX" << par_name << "YYY";
 
-    Teuchos::RCP<ParameterList> ret = Teuchos::getParametersFromXmlString(finalString);
-    return ret;
+    // update final string with parameters
+    Teuchos::ParameterEntry par_entry = inputParameters.entry(it);
+    this->ReplaceString(finalString,
+                        ss.str(), Teuchos::toString(par_entry.getAny()));
   }
 
-} // end namespace MueLu
-#endif // PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_BGS2x2_DEF_HPP_
+  Teuchos::RCP<ParameterList> ret = Teuchos::getParametersFromXmlString(finalString);
+  return ret;
+}
+
+}  // end namespace MueLu
+#endif  // PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_BGS2x2_DEF_HPP_
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_decl.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_decl.hpp
index f9d9d8b9ff5b..ff5f1d8a40e0 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_decl.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_decl.hpp
@@ -51,52 +51,48 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FacadeClassBase
-  : public virtual BaseClass{
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FacadeClassBase
+  : public virtual BaseClass {
 #undef MUELU_FACADECLASSBASE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Constructor.
-    FacadeClassBase();
+  //! Constructor.
+  FacadeClassBase();
 
-    //! Destructor.
-    virtual ~FacadeClassBase() { }
+  //! Destructor.
+  virtual ~FacadeClassBase() {}
 
-    //@}
+  //@}
 
-    /*! @brief Set parameter list for FacadeClass (abstract member).
+  /*! @brief Set parameter list for FacadeClass (abstract member).
 
-        @param[in] paramList: ParameterList containing the MueLu parameters.
-    */
-    virtual Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList) = 0;
+      @param[in] paramList: ParameterList containing the MueLu parameters.
+  */
+  virtual Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList) = 0;
 
-  protected:
-
-    /*! @brief Replace all occurrences of search string "search" by the string in "replace" given the string "subject"
-     */
-    std::string ReplaceString(std::string& subject, const std::string& search, const std::string& replace) {
-      size_t pos = 0;
-      while ((pos = subject.find(search, pos)) != std::string::npos) {
-        subject.replace(pos, search.length(), replace);
-        pos += replace.length();
-      }
-      return subject;
+ protected:
+  /*! @brief Replace all occurrences of search string "search" by the string in "replace" given the string "subject"
+   */
+  std::string ReplaceString(std::string& subject, const std::string& search, const std::string& replace) {
+    size_t pos = 0;
+    while ((pos = subject.find(search, pos)) != std::string::npos) {
+      subject.replace(pos, search.length(), replace);
+      pos += replace.length();
     }
+    return subject;
+  }
+};
 
-  };
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACADECLASSBASE_SHORT
 
-
-
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_MUELU_FACADECLASSBASE_DECL_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_def.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_def.hpp
index fd2375c1ccc2..6ef823b40fcb 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_def.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassBase_def.hpp
@@ -50,15 +50,14 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
-
 #include "MueLu_Exceptions.hpp"
 
 #include "MueLu_FacadeClassBase_decl.hpp"
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  FacadeClassBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeClassBase() {
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+FacadeClassBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeClassBase() {
 }
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_MUELU_FACADECLASSBASE_DEF_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_decl.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_decl.hpp
index 368c8ed6bd78..e3857cc4537a 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_decl.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_decl.hpp
@@ -56,51 +56,47 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FacadeClassFactory
-  : public virtual BaseClass{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FacadeClassFactory
+  : public virtual BaseClass {
 #undef MUELU_FACADECLASSFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Constructor.
-    FacadeClassFactory();
+  //! Constructor.
+  FacadeClassFactory();
 
-    //! Destructor.
-    virtual ~FacadeClassFactory() { }
+  //! Destructor.
+  virtual ~FacadeClassFactory() {}
 
-    //@}
+  //@}
 
-    /*! @brief Set parameter list for FacadeClassFactory interpreter.
+  /*! @brief Set parameter list for FacadeClassFactory interpreter.
 
-       @param[in] paramList: ParameterList containing the MueLu parameters.
-    */
-    Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
+     @param[in] paramList: ParameterList containing the MueLu parameters.
+  */
+  Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
 
-    /*! @brief Register new facade class
-     *
-     * Register new externally provided facade class in FacadeClassFactory
-     *
-     * @param[in] name: name that is used to access Facade class
-     * @param[in] facadeclass: RCP pointer to facade class instance
-     */
-    void RegisterFacadeClass(std::string name, Teuchos::RCP<FacadeClassBase> facadeclass) {
-      facadeClasses_[name] = facadeclass;
-    }
+  /*! @brief Register new facade class
+   *
+   * Register new externally provided facade class in FacadeClassFactory
+   *
+   * @param[in] name: name that is used to access Facade class
+   * @param[in] facadeclass: RCP pointer to facade class instance
+   */
+  void RegisterFacadeClass(std::string name, Teuchos::RCP<FacadeClassBase> facadeclass) {
+    facadeClasses_[name] = facadeclass;
+  }
 
-  private:
+ private:
+  std::map<std::string, Teuchos::RCP<FacadeClassBase> > facadeClasses_;
+};
 
-    std::map<std::string, Teuchos::RCP<FacadeClassBase> > facadeClasses_;
-
-  };
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACADECLASSFACTORY_SHORT
 
-
-
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_MUELU_FACADECLASSFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_def.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_def.hpp
index 137ed1d9cfd1..787fa95ec8a6 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_def.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeClassFactory_def.hpp
@@ -50,7 +50,6 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
-
 #include "MueLu_Exceptions.hpp"
 
 #include "MueLu_FacadeClassBase.hpp"
@@ -61,34 +60,32 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeClassFactory() {
-    facadeClasses_["Simple"] = Teuchos::rcp(new FacadeSimple<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-    facadeClasses_["BGS2x2"] = Teuchos::rcp(new FacadeBGS2x2<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-  }
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeClassFactory() {
+  facadeClasses_["Simple"] = Teuchos::rcp(new FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+  facadeClasses_["BGS2x2"] = Teuchos::rcp(new FacadeBGS2x2<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::ParameterList> FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::ParameterList> FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
+  TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter("MueLu preconditioner") == false, MueLu::Exceptions::RuntimeError, "FacadeClassFactory: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
+  TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeClassFactory: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter("MueLu preconditioner") == false, MueLu::Exceptions::RuntimeError, "FacadeClassFactory: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
-    TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeClassFactory: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
+  std::string precMueLu = paramList.get<std::string>("MueLu preconditioner");
 
-    std::string precMueLu = paramList.get<std::string>("MueLu preconditioner");
-
-    // could not find requested facade class
-    if(facadeClasses_.find(precMueLu) == facadeClasses_.end()) {
-      GetOStream(Errors) << "FacadeClassFactory: Could not find facade class \"" << precMueLu << "\"!" << std::endl;
-      GetOStream(Errors) << "The available facade classes are:" << std::endl;
-      for(typename std::map<std::string, Teuchos::RCP<FacadeClassBase> >::const_iterator it =facadeClasses_.begin(); it != facadeClasses_.end(); it++){
-        GetOStream(Errors) << "   " << it->first << std::endl;
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "FacadeClassFactory: Could not find facade class \"" << precMueLu << "\".");
+  // could not find requested facade class
+  if (facadeClasses_.find(precMueLu) == facadeClasses_.end()) {
+    GetOStream(Errors) << "FacadeClassFactory: Could not find facade class \"" << precMueLu << "\"!" << std::endl;
+    GetOStream(Errors) << "The available facade classes are:" << std::endl;
+    for (typename std::map<std::string, Teuchos::RCP<FacadeClassBase> >::const_iterator it = facadeClasses_.begin(); it != facadeClasses_.end(); it++) {
+      GetOStream(Errors) << "   " << it->first << std::endl;
     }
-
-    return facadeClasses_[precMueLu]->SetParameterList(paramList);
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "FacadeClassFactory: Could not find facade class \"" << precMueLu << "\".");
   }
 
-} // end namespace MueLu
+  return facadeClasses_[precMueLu]->SetParameterList(paramList);
+}
+
+}  // end namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_MUELU_FACADECLASSFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_decl.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_decl.hpp
index c36c4a4ef283..34ea6f4de46e 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_decl.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_decl.hpp
@@ -55,37 +55,34 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FacadeSimple : public FacadeClassBase<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FacadeSimple : public FacadeClassBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Constructor.
-    FacadeSimple();
+  //! Constructor.
+  FacadeSimple();
 
-    //! Destructor.
-    virtual ~FacadeSimple() { }
+  //! Destructor.
+  virtual ~FacadeSimple() {}
 
-    //@}
+  //@}
 
-    /*! @brief Set parameter list for FacadeClass interpreter.
+  /*! @brief Set parameter list for FacadeClass interpreter.
 
-        @param[in] paramList: ParameterList containing the MueLu parameters for chosen facade class.
-    */
-    Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
-
-  private:
-
-  };
-
-} // namespace MueLu
+      @param[in] paramList: ParameterList containing the MueLu parameters for chosen facade class.
+  */
+  Teuchos::RCP<Teuchos::ParameterList> SetParameterList(const Teuchos::ParameterList& paramList);
 
+ private:
+};
 
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_Simple_DECL_HPP_ */
diff --git a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_def.hpp b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_def.hpp
index 6c49ed4bdd64..b9dba089af29 100644
--- a/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_def.hpp
+++ b/packages/muelu/src/Interface/FacadeClasses/MueLu_FacadeSimple_def.hpp
@@ -50,355 +50,349 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
-
 #include "MueLu_Exceptions.hpp"
 
 #include "MueLu_FacadeSimple_decl.hpp"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeSimple() {
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::ParameterList> FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
-
-    // obtain ParameterList with default input parameters for this facade class
-    // Note all parameters are of type string (we use it for string replacement)
-    std::string defaultString =
-"<ParameterList name=\"Input\">"
-"<Parameter name=\"MueLu preconditioner\" type=\"string\" value=\"undefined\"/>"
-"<Parameter name=\"Block 1: dofs per node\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 2: dofs per node\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 1: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
-  "<Parameter name=\"Block 1: level-of-fill\" type=\"int\" value=\"0\"/>"
-  "<Parameter name=\"Block 1: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 1: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"Block 1: transfer smoothing\" type=\"bool\" value=\"true\"/>"
-  "<Parameter name=\"Block 2: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
-  "<Parameter name=\"Block 2: level-of-fill\" type=\"int\" value=\"0\"/>"
-  "<Parameter name=\"Block 2: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
-  "<Parameter name=\"Block 2: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"Block 2: transfer smoothing\" type=\"bool\" value=\"true\"/>"
-  "<Parameter name=\"Simple: damping factor\" type=\"double\" value=\"1.0\"/>"
-  "<Parameter name=\"max levels\" type=\"int\" value=\"5\"/>"
-  "<Parameter name=\"coarse: max size\" type=\"int\" value=\"25000\"/>"
-  "<Parameter name=\"verbosity\" type=\"string\" value=\"High\"/>"
-  "</ParameterList>"
-;
-    Teuchos::RCP<ParameterList> defaultList = Teuchos::getParametersFromXmlString(defaultString);
-    // validate user input parameters (and set defaults if necessary)
-    Teuchos::ParameterList inputParameters = paramList;
-    inputParameters.validateParametersAndSetDefaults(*defaultList);
-    TEUCHOS_TEST_FOR_EXCEPTION(inputParameters.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeSimple: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
-
-    // create copy of template string which is updated with in-place string replacements
-    // template string for preconditioner layout (factory based parameters)
-    std::string finalString =
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FacadeSimple() {
+}
 
-"<ParameterList name=\"MueLu\">"
-"  <ParameterList name=\"Factories\">"
-"    <ParameterList name=\"mySubBlockAFactory1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
-"      <Parameter name=\"block row\"                 type=\"int\"     value=\"0\"/>"
-"      <Parameter name=\"block col\"                 type=\"int\"     value=\"0\"/>"
-"      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myAggFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"UncoupledAggregationFactory\"/>"
-"      <Parameter name=\"aggregation: min agg size\" type=\"int\" value=\"5\"/>"
-"      <Parameter name=\"aggregation: max selected neighbors\" type=\"int\" value=\"1\"/>"
-"    </ParameterList>"
-""
-"    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
-"    <ParameterList name=\"myCoarseMap1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
-"      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myTentativePFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myPFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"PgPFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myRFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
-"    </ParameterList>"
-""
-"    <!-- We have to use Nullspace1 here. If \"Nullspace1\" is not set the"
-"         Factory creates the default null space containing of constant"
-"         vectors -->"
-"    <ParameterList name=\"myNspFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
-"      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace1\"/>"
-"      <Parameter name=\"Nullspace1\" type=\"string\" value=\"myTentativePFact1\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK 2 (for submatrix A_{11}) PRESSURE PART -->"
-"    <ParameterList name=\"mySubBlockAFactory2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
-"      <Parameter name=\"block row\"                 type=\"int\"     value=\"1\"/>"
-"      <Parameter name=\"block col\"                 type=\"int\"     value=\"1\"/>"
-"      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"    </ParameterList>"
-""
-"    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
-"    <ParameterList name=\"myCoarseMap2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
-"      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
-"      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myTentativePFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myPFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SaPFactory\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myRFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TransPFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myNspFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
-"      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace2\"/>"
-"      <Parameter name=\"Nullspace2\" type=\"string\" value=\"myTentativePFact2\"/>"
-"    </ParameterList>"
-""
-"    <!-- FACTORY MANAGERS -->"
-""
-"    <!-- Multigrid setup for velocity block (A_{00}) -->"
-"    <ParameterList name=\"myFirstGroup\">"
-"      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myRFact1\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
-"      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact1\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
-"    </ParameterList>"
-""
-"    <!-- Multigrid setup for pressure block (A_{11}) -->"
-"    <ParameterList name=\"mySecondGroup\">"
-"      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
-"      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myRFact2\"/>"
-"      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/><!-- reuse aggs from PRESSURE block! -->"
-"      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact2\"/>"
-"      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK TRANSFER operators -->"
-""
-"    <ParameterList name=\"myBlockedPFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"BlockedPFactory\"/>"
-"      <!-- factory manager for block 1 -->"
-"      <ParameterList name=\"block1\">"
-"         <Parameter name=\"group\" type=\"string\" value=\"myFirstGroup\"/>"
-"      </ParameterList>"
-"      <!-- factory manager for block 2 -->"
-"      <ParameterList name=\"block2\">"
-"        <Parameter name=\"group\" type=\"string\" value=\"mySecondGroup\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockedRFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockedRAPFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"BlockedRAPFactory\"/>"
-"      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
-"      <Parameter name=\"R\" type=\"string\" value=\"myBlockedRFact\"/>"
-"    </ParameterList>"
-""
-"    <!-- BLOCK SMOOTHERS -->"
-"    <ParameterList name=\"mySmooFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 1: relaxation: typeYYY\"/>"
-"        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 1: relaxation: sweepsYYY\"/>"
-"        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 1: relaxation: damping factorYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooILUFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 1: level-of-fillYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooDirectFact1\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 2: relaxation: typeYYY\"/>"
-"        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 2: relaxation: sweepsYYY\"/>"
-"        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 2: relaxation: damping factorYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooILUFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
-"      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
-"      <ParameterList name=\"ParameterList\">"
-"        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 2: level-of-fillYYY\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"mySmooDirectFact2\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myNSSchurCompFact\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SchurComplementFactory\"/>"
-"      <Parameter name=\"omega\" type=\"double\" value=\"1.0\"/>"
-"      <Parameter name=\"lumping\" type=\"bool\" value=\"false\"/>"
-"    </ParameterList>"
-""
-"    <ParameterList name=\"myBlockSmoother\">"
-"      <Parameter name=\"factory\" type=\"string\" value=\"SimpleSmoother\"/>"
-"      <Parameter name=\"Sweeps\" type=\"int\" value=\"1\"/>"
-"      <Parameter name=\"Damping factor\" type=\"double\" value=\"XXXSimple: damping factorYYY\"/>"
-"      <!-- factory manager for block 1 -->"
-"      <ParameterList name=\"block1\">"
-"        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
-"        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother1XYZ\"/>"
-"      </ParameterList>"
-"      <!-- factory manager for block 2 -->"
-"      <ParameterList name=\"block2\">"
-"        <Parameter name=\"A\" type=\"string\" value=\"myNSSchurCompFact\"/>"
-"        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother2XYZ\"/>"
-"      </ParameterList>"
-"    </ParameterList>"
-""
-"  </ParameterList>"
-"  <!-- end Factories -->"
-""
-"  <!-- Definition of the multigrid preconditioner -->"
-"  <ParameterList name=\"Hierarchy\">"
-""
-"    <Parameter name=\"max levels\"          type=\"int\"      value=\"XXXmax levelsYYY\"/>"
-"    <Parameter name=\"coarse: max size\"    type=\"int\"      value=\"XXXcoarse: max sizeYYY\"/>"
-"    <Parameter name=\"verbosity\"           type=\"string\"   value=\"XXXverbosityYYY\"/>"
-""
-"    <ParameterList name=\"AllLevel\">"
-"      <Parameter name=\"startLevel\"        type=\"int\"      value=\"0\"/>"
-"      <Parameter name=\"Smoother\"          type=\"string\"   value=\"myBlockSmoother\"/>"
-"      <Parameter name=\"CoarseSolver\"      type=\"string\"   value=\"myBlockSmoother\"/>"
-"      <Parameter name=\"P\"                 type=\"string\"   value=\"myBlockedPFact\"/>"
-"      <Parameter name=\"R\"                 type=\"string\"   value=\"myBlockedRFact\"/>"
-"      <Parameter name=\"A\"                 type=\"string\"   value=\"myBlockedRAPFact\"/>"
-"    </ParameterList>"
-""
-"  </ParameterList>"
-"</ParameterList>"
-  ;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::ParameterList> FacadeSimple<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
+  // obtain ParameterList with default input parameters for this facade class
+  // Note all parameters are of type string (we use it for string replacement)
+  std::string defaultString =
+      "<ParameterList name=\"Input\">"
+      "<Parameter name=\"MueLu preconditioner\" type=\"string\" value=\"undefined\"/>"
+      "<Parameter name=\"Block 1: dofs per node\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 2: dofs per node\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 1: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
+      "<Parameter name=\"Block 1: level-of-fill\" type=\"int\" value=\"0\"/>"
+      "<Parameter name=\"Block 1: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 1: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"Block 1: transfer smoothing\" type=\"bool\" value=\"true\"/>"
+      "<Parameter name=\"Block 2: smoother\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>"
+      "<Parameter name=\"Block 2: level-of-fill\" type=\"int\" value=\"0\"/>"
+      "<Parameter name=\"Block 2: relaxation: sweeps\" type=\"int\" value=\"1\"/>"
+      "<Parameter name=\"Block 2: relaxation: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"Block 2: transfer smoothing\" type=\"bool\" value=\"true\"/>"
+      "<Parameter name=\"Simple: damping factor\" type=\"double\" value=\"1.0\"/>"
+      "<Parameter name=\"max levels\" type=\"int\" value=\"5\"/>"
+      "<Parameter name=\"coarse: max size\" type=\"int\" value=\"25000\"/>"
+      "<Parameter name=\"verbosity\" type=\"string\" value=\"High\"/>"
+      "</ParameterList>";
+  Teuchos::RCP<ParameterList> defaultList = Teuchos::getParametersFromXmlString(defaultString);
+  // validate user input parameters (and set defaults if necessary)
+  Teuchos::ParameterList inputParameters = paramList;
+  inputParameters.validateParametersAndSetDefaults(*defaultList);
+  TEUCHOS_TEST_FOR_EXCEPTION(inputParameters.get<std::string>("MueLu preconditioner") == "undefined", MueLu::Exceptions::RuntimeError, "FacadeSimple: undefined MueLu preconditioner. Set the \"MueLu preconditioner\" parameter correctly in your input file.");
 
-    // logical code for more complicated distinctions
+  // create copy of template string which is updated with in-place string replacements
+  // template string for preconditioner layout (factory based parameters)
+  std::string finalString =
 
+      "<ParameterList name=\"MueLu\">"
+      "  <ParameterList name=\"Factories\">"
+      "    <ParameterList name=\"mySubBlockAFactory1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
+      "      <Parameter name=\"block row\"                 type=\"int\"     value=\"0\"/>"
+      "      <Parameter name=\"block col\"                 type=\"int\"     value=\"0\"/>"
+      "      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myAggFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"UncoupledAggregationFactory\"/>"
+      "      <Parameter name=\"aggregation: min agg size\" type=\"int\" value=\"5\"/>"
+      "      <Parameter name=\"aggregation: max selected neighbors\" type=\"int\" value=\"1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
+      "    <ParameterList name=\"myCoarseMap1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
+      "      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 1: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myTentativePFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myPFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"PgPFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myRFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- We have to use Nullspace1 here. If \"Nullspace1\" is not set the"
+      "         Factory creates the default null space containing of constant"
+      "         vectors -->"
+      "    <ParameterList name=\"myNspFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
+      "      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace1\"/>"
+      "      <Parameter name=\"Nullspace1\" type=\"string\" value=\"myTentativePFact1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK 2 (for submatrix A_{11}) PRESSURE PART -->"
+      "    <ParameterList name=\"mySubBlockAFactory2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SubBlockAFactory\"/>"
+      "      <Parameter name=\"block row\"                 type=\"int\"     value=\"1\"/>"
+      "      <Parameter name=\"block col\"                 type=\"int\"     value=\"1\"/>"
+      "      <Parameter name=\"Range map: Striding info\"  type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Domain map: Striding info\" type=\"string\"  value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- tell the tentative prolongator that we have 2 DOFs per node on the coarse levels -->"
+      "    <ParameterList name=\"myCoarseMap2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"CoarseMapFactory\"/>"
+      "      <Parameter name=\"Striding info\" type=\"string\" value=\"{ XXXBlock 2: dofs per nodeYYY }\"/>"
+      "      <Parameter name=\"Strided block id\" type=\"int\" value=\"-1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myTentativePFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TentativePFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myPFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SaPFactory\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myTentativePFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myRFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TransPFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myNspFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"NullspaceFactory\"/>"
+      "      <Parameter name=\"Fine level nullspace\" type=\"string\" value=\"Nullspace2\"/>"
+      "      <Parameter name=\"Nullspace2\" type=\"string\" value=\"myTentativePFact2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- FACTORY MANAGERS -->"
+      ""
+      "    <!-- Multigrid setup for velocity block (A_{00}) -->"
+      "    <ParameterList name=\"myFirstGroup\">"
+      "      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact1\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myRFact1\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/>"
+      "      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact1\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap1\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- Multigrid setup for pressure block (A_{11}) -->"
+      "    <ParameterList name=\"mySecondGroup\">"
+      "      <Parameter name=\"group\" type=\"string\" value=\"FactoryManager\"/>"
+      "      <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory2\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myPFact2\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myRFact2\"/>"
+      "      <Parameter name=\"Aggregates\" type=\"string\" value=\"myAggFact1\"/><!-- reuse aggs from PRESSURE block! -->"
+      "      <Parameter name=\"Nullspace\" type=\"string\" value=\"myNspFact2\"/>"
+      "      <Parameter name=\"CoarseMap\" type=\"string\" value=\"myCoarseMap2\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK TRANSFER operators -->"
+      ""
+      "    <ParameterList name=\"myBlockedPFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"BlockedPFactory\"/>"
+      "      <!-- factory manager for block 1 -->"
+      "      <ParameterList name=\"block1\">"
+      "         <Parameter name=\"group\" type=\"string\" value=\"myFirstGroup\"/>"
+      "      </ParameterList>"
+      "      <!-- factory manager for block 2 -->"
+      "      <ParameterList name=\"block2\">"
+      "        <Parameter name=\"group\" type=\"string\" value=\"mySecondGroup\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockedRFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"GenericRFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockedRAPFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"BlockedRAPFactory\"/>"
+      "      <Parameter name=\"P\" type=\"string\" value=\"myBlockedPFact\"/>"
+      "      <Parameter name=\"R\" type=\"string\" value=\"myBlockedRFact\"/>"
+      "    </ParameterList>"
+      ""
+      "    <!-- BLOCK SMOOTHERS -->"
+      "    <ParameterList name=\"mySmooFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 1: relaxation: typeYYY\"/>"
+      "        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 1: relaxation: sweepsYYY\"/>"
+      "        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 1: relaxation: damping factorYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooILUFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 1: level-of-fillYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooDirectFact1\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"RELAXATION\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"relaxation: type\" type=\"string\" value=\"XXXBlock 2: relaxation: typeYYY\"/>"
+      "        <Parameter name=\"relaxation: sweeps\" type=\"int\"    value=\"XXXBlock 2: relaxation: sweepsYYY\"/>"
+      "        <Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"XXXBlock 2: relaxation: damping factorYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooILUFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"TrilinosSmoother\"/>"
+      "      <Parameter name=\"type\" type=\"string\" value=\"ILU\"/>"
+      "      <ParameterList name=\"ParameterList\">"
+      "        <Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"XXXBlock 2: level-of-fillYYY\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"mySmooDirectFact2\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"DirectSolver\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myNSSchurCompFact\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SchurComplementFactory\"/>"
+      "      <Parameter name=\"omega\" type=\"double\" value=\"1.0\"/>"
+      "      <Parameter name=\"lumping\" type=\"bool\" value=\"false\"/>"
+      "    </ParameterList>"
+      ""
+      "    <ParameterList name=\"myBlockSmoother\">"
+      "      <Parameter name=\"factory\" type=\"string\" value=\"SimpleSmoother\"/>"
+      "      <Parameter name=\"Sweeps\" type=\"int\" value=\"1\"/>"
+      "      <Parameter name=\"Damping factor\" type=\"double\" value=\"XXXSimple: damping factorYYY\"/>"
+      "      <!-- factory manager for block 1 -->"
+      "      <ParameterList name=\"block1\">"
+      "        <Parameter name=\"A\" type=\"string\" value=\"mySubBlockAFactory1\"/>"
+      "        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother1XYZ\"/>"
+      "      </ParameterList>"
+      "      <!-- factory manager for block 2 -->"
+      "      <ParameterList name=\"block2\">"
+      "        <Parameter name=\"A\" type=\"string\" value=\"myNSSchurCompFact\"/>"
+      "        <Parameter name=\"Smoother\" type=\"string\" value=\"XYZSmoother2XYZ\"/>"
+      "      </ParameterList>"
+      "    </ParameterList>"
+      ""
+      "  </ParameterList>"
+      "  <!-- end Factories -->"
+      ""
+      "  <!-- Definition of the multigrid preconditioner -->"
+      "  <ParameterList name=\"Hierarchy\">"
+      ""
+      "    <Parameter name=\"max levels\"          type=\"int\"      value=\"XXXmax levelsYYY\"/>"
+      "    <Parameter name=\"coarse: max size\"    type=\"int\"      value=\"XXXcoarse: max sizeYYY\"/>"
+      "    <Parameter name=\"verbosity\"           type=\"string\"   value=\"XXXverbosityYYY\"/>"
+      ""
+      "    <ParameterList name=\"AllLevel\">"
+      "      <Parameter name=\"startLevel\"        type=\"int\"      value=\"0\"/>"
+      "      <Parameter name=\"Smoother\"          type=\"string\"   value=\"myBlockSmoother\"/>"
+      "      <Parameter name=\"CoarseSolver\"      type=\"string\"   value=\"myBlockSmoother\"/>"
+      "      <Parameter name=\"P\"                 type=\"string\"   value=\"myBlockedPFact\"/>"
+      "      <Parameter name=\"R\"                 type=\"string\"   value=\"myBlockedRFact\"/>"
+      "      <Parameter name=\"A\"                 type=\"string\"   value=\"myBlockedRAPFact\"/>"
+      "    </ParameterList>"
+      ""
+      "  </ParameterList>"
+      "</ParameterList>";
 
-    std::string smoother1 = inputParameters.get<std::string>("Block 1: smoother");
-    if(smoother1 == "ILU") {
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooILUFact1");
-    } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "SGS") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Symmetric Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "GS") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Jacobi") {
-      this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Jacobi");
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
-    } else if (smoother1 == "Direct") {
-      this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooDirectFact1");
-    } else {
-      this->GetOStream(Errors) << "Invalid smoother type for block 1: " << smoother1 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
-    }
+  // logical code for more complicated distinctions
 
-    std::string smoother2 = inputParameters.get<std::string>("Block 2: smoother");
-    if(smoother2 == "ILU") {
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooILUFact2");
-    } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "SGS") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Symmetric Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "GS") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Jacobi") {
-      this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Jacobi");
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
-    } else if (smoother2 == "Direct") {
-      this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooDirectFact2");
-    } else {
-      this->GetOStream(Errors) << "Invalid smoother type for block 2: " << smoother2 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
-    }
+  std::string smoother1 = inputParameters.get<std::string>("Block 1: smoother");
+  if (smoother1 == "ILU") {
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooILUFact1");
+  } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "SGS") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Symmetric Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Symmetric Gauss-Seidel" || smoother1 == "GS") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Jacobi") {
+    this->ReplaceString(finalString, "XXXBlock 1: relaxation: typeYYY", "Jacobi");
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooFact1");
+  } else if (smoother1 == "Direct") {
+    this->ReplaceString(finalString, "XYZSmoother1XYZ", "mySmooDirectFact1");
+  } else {
+    this->GetOStream(Errors) << "Invalid smoother type for block 1: " << smoother1 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
+  }
 
-    if(inputParameters.get<bool>("Block 1: transfer smoothing") == true) {
-      this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myPFact1");
-      this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myRFact1");
-    } else {
-      this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myTentativePFact1");
-      this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myTransPFact1");
-    }
-    if(inputParameters.get<bool>("Block 2: transfer smoothing") == true) {
-      this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myPFact2");
-      this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myRFact2");
-    } else {
-      this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myTentativePFact2");
-      this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myTransPFact2");
-    }
+  std::string smoother2 = inputParameters.get<std::string>("Block 2: smoother");
+  if (smoother2 == "ILU") {
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooILUFact2");
+  } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "SGS") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Symmetric Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Symmetric Gauss-Seidel" || smoother2 == "GS") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Gauss-Seidel");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Jacobi") {
+    this->ReplaceString(finalString, "XXXBlock 2: relaxation: typeYYY", "Jacobi");
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooFact2");
+  } else if (smoother2 == "Direct") {
+    this->ReplaceString(finalString, "XYZSmoother2XYZ", "mySmooDirectFact2");
+  } else {
+    this->GetOStream(Errors) << "Invalid smoother type for block 2: " << smoother2 << ". Valid options are: \"SGS\", \"GS\", \"Jacobi\", \"ILU\" or \"Direct\"." << std::endl;
+  }
 
-    // end logical code
+  if (inputParameters.get<bool>("Block 1: transfer smoothing") == true) {
+    this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myPFact1");
+    this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myRFact1");
+  } else {
+    this->ReplaceString(finalString, "XXXBlock 1: prolongatorYYY", "myTentativePFact1");
+    this->ReplaceString(finalString, "XXXBlock 1: restrictor YYY", "myTransPFact1");
+  }
+  if (inputParameters.get<bool>("Block 2: transfer smoothing") == true) {
+    this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myPFact2");
+    this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myRFact2");
+  } else {
+    this->ReplaceString(finalString, "XXXBlock 2: prolongatorYYY", "myTentativePFact2");
+    this->ReplaceString(finalString, "XXXBlock 2: restrictor YYY", "myTransPFact2");
+  }
 
-    // loop over all input parameters
-    for(Teuchos::ParameterList::ConstIterator it = inputParameters.begin(); it != inputParameters.end(); it++) {
-      // form replacement string
-      std::string par_name = inputParameters.name(it);
-      std::stringstream ss;
-      ss << "XXX" << par_name << "YYY";
+  // end logical code
 
-      // update final string with parameters
-      Teuchos::ParameterEntry par_entry = inputParameters.entry(it);
-      this->ReplaceString(finalString,
-              ss.str(), Teuchos::toString(par_entry.getAny()));
-    }
+  // loop over all input parameters
+  for (Teuchos::ParameterList::ConstIterator it = inputParameters.begin(); it != inputParameters.end(); it++) {
+    // form replacement string
+    std::string par_name = inputParameters.name(it);
+    std::stringstream ss;
+    ss << "XXX" << par_name << "YYY";
 
-    Teuchos::RCP<ParameterList> ret = Teuchos::getParametersFromXmlString(finalString);
-    return ret;
+    // update final string with parameters
+    Teuchos::ParameterEntry par_entry = inputParameters.entry(it);
+    this->ReplaceString(finalString,
+                        ss.str(), Teuchos::toString(par_entry.getAny()));
   }
 
-} // end namespace MueLu
-#endif // PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_Simple_DEF_HPP_
+  Teuchos::RCP<ParameterList> ret = Teuchos::getParametersFromXmlString(finalString);
+  return ret;
+}
+
+}  // end namespace MueLu
+#endif  // PACKAGES_MUELU_SRC_INTERFACE_FACADECLASSES_Simple_DEF_HPP_
diff --git a/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_decl.hpp b/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_decl.hpp
index 308bb878ca53..76edf912232d 100644
--- a/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_decl.hpp
+++ b/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_decl.hpp
@@ -37,171 +37,169 @@
 
 namespace MueLu {
 
-  /*
-    Utility that from an existing Teuchos::ParameterList creates a new list, in
-    which level-specific parameters are replaced with sublists.
-
-    Currently, level-specific parameters that begin with "smoother:"
-    or "aggregation:" are placed in sublists. Coarse options are also placed
-    in a coarse list.
-
-    Example:
-    Input:
-    smoother: type (level 0) = symmetric Gauss-Seidel
-    smoother: sweeps (level 0) = 1
-    Output:
-    smoother: list (level 0) ->
-    smoother: type = symmetric Gauss-Seidel
-    smoother: sweeps = 1
-  */
-  // This function is a copy of ML_CreateSublists to avoid dependency on ML
-  // Throw exception on error instead of exit()
-  //void CreateSublists(const ParameterList &List, ParameterList &newList);
-
-
-  /*!
-    @class AdaptiveSAMLParameterListInterpreter class.
-    @brief Class that accepts ML-style parameters and builds a MueLu preconditioner.
-    This interpreter uses the same default values as ML. This allows to compare ML/MueLu results
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class AdaptiveSaMLParameterListInterpreter :
-    public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+/*
+  Utility that from an existing Teuchos::ParameterList creates a new list, in
+  which level-specific parameters are replaced with sublists.
+
+  Currently, level-specific parameters that begin with "smoother:"
+  or "aggregation:" are placed in sublists. Coarse options are also placed
+  in a coarse list.
+
+  Example:
+  Input:
+  smoother: type (level 0) = symmetric Gauss-Seidel
+  smoother: sweeps (level 0) = 1
+  Output:
+  smoother: list (level 0) ->
+  smoother: type = symmetric Gauss-Seidel
+  smoother: sweeps = 1
+*/
+// This function is a copy of ML_CreateSublists to avoid dependency on ML
+// Throw exception on error instead of exit()
+// void CreateSublists(const ParameterList &List, ParameterList &newList);
+
+/*!
+  @class AdaptiveSAMLParameterListInterpreter class.
+  @brief Class that accepts ML-style parameters and builds a MueLu preconditioner.
+  This interpreter uses the same default values as ML. This allows to compare ML/MueLu results
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AdaptiveSaMLParameterListInterpreter : public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    AdaptiveSaMLParameterListInterpreter() : nullspace_(NULL), blksize_(1) { }
-
-    //! Constructor.
-    //! @param paramList: parameter list with ML parameters
-    //! @param nspVector: MultiVector with fine-level nullspace approximation
-    //! @param factoryList: vector with RCP of FactoryBase objects
-    //!
-    //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
-    //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
-    //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    AdaptiveSaMLParameterListInterpreter(Teuchos::ParameterList & paramList,std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    //! @param xmlFileName: file name for XML file with ML parameters
-    //! @param factoryList: vector with RCP of FactoryBase objects
-    //!
-    //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
-    //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
-    //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    AdaptiveSaMLParameterListInterpreter(const std::string & xmlFileName,std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
+  //! Constructor.
+  AdaptiveSaMLParameterListInterpreter()
+    : nullspace_(NULL)
+    , blksize_(1) {}
 
-    //! Destructor.
-    virtual ~AdaptiveSaMLParameterListInterpreter() { }
+  //! Constructor.
+  //! @param paramList: parameter list with ML parameters
+  //! @param nspVector: MultiVector with fine-level nullspace approximation
+  //! @param factoryList: vector with RCP of FactoryBase objects
+  //!
+  //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
+  //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
+  //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  AdaptiveSaMLParameterListInterpreter(Teuchos::ParameterList& paramList, std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
 
-    //@}
+  //! Constructor.
+  //! @param xmlFileName: file name for XML file with ML parameters
+  //! @param factoryList: vector with RCP of FactoryBase objects
+  //!
+  //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
+  //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
+  //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  AdaptiveSaMLParameterListInterpreter(const std::string& xmlFileName, std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
 
-    //@{
+  //! Destructor.
+  virtual ~AdaptiveSaMLParameterListInterpreter() {}
 
-    void SetParameterList(const Teuchos::ParameterList & paramList);
+  //@}
 
-    //@}
+  //@{
 
-    //@{
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //! Setup Hierarchy object
-    virtual void SetupHierarchy(Hierarchy & H) const;
+  //@}
 
-    //@}
+  //@{
 
-    //@{
+  //! Setup Hierarchy object
+  virtual void SetupHierarchy(Hierarchy& H) const;
 
-    //! @name Handling of additional user-specific transfer factories
-    //@{
-    /*! @brief Add transfer factory in the end of list of transfer factories for RAPFactory.
+  //@}
 
-    This allows the user to add user-specific factories to the MueLu Hierarchy. The idea is to be able
-    to add some factories that write out some debug information etc. which are not handled by the ML
-    Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    */
-    void AddTransferFactory(const RCP<FactoryBase> & factory);
+  //@{
 
-    //! Returns number of transfer factories.
-    size_t NumTransferFactories() const;
-    //@}
+  //! @name Handling of additional user-specific transfer factories
+  //@{
+  /*! @brief Add transfer factory in the end of list of transfer factories for RAPFactory.
 
-  private:
-
-    //! build multigrid hierarchy for improving nullspace
-    //! use ML settings that are also used for the final full multigrid
-    //! hierarchy. In contrary to the final multigrid hierarchy use
-    //! only nonsmoothed transfer operators (safe time of prolongator smoothing)
-    //! and cheap level smoothers (no direct solver on coarsest level).
-    void SetupInitHierarchy(Hierarchy & H) const;
-
-    //! internal routine to add a new factory manager used for the initialization phase
-    void AddInitFactoryManager(int startLevel, int numDesiredLevel, RCP<FactoryManagerBase> manager) {
-      const int lastLevel = startLevel + numDesiredLevel - 1;
-      if (init_levelManagers_.size() < lastLevel + 1) init_levelManagers_.resize(lastLevel + 1);
-
-      for(int iLevel = startLevel; iLevel <= lastLevel; iLevel++) {
-        init_levelManagers_[iLevel] = manager;
-      }
+  This allows the user to add user-specific factories to the MueLu Hierarchy. The idea is to be able
+  to add some factories that write out some debug information etc. which are not handled by the ML
+  Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  */
+  void AddTransferFactory(const RCP<FactoryBase>& factory);
+
+  //! Returns number of transfer factories.
+  size_t NumTransferFactories() const;
+  //@}
+
+ private:
+  //! build multigrid hierarchy for improving nullspace
+  //! use ML settings that are also used for the final full multigrid
+  //! hierarchy. In contrary to the final multigrid hierarchy use
+  //! only nonsmoothed transfer operators (safe time of prolongator smoothing)
+  //! and cheap level smoothers (no direct solver on coarsest level).
+  void SetupInitHierarchy(Hierarchy& H) const;
+
+  //! internal routine to add a new factory manager used for the initialization phase
+  void AddInitFactoryManager(int startLevel, int numDesiredLevel, RCP<FactoryManagerBase> manager) {
+    const int lastLevel = startLevel + numDesiredLevel - 1;
+    if (init_levelManagers_.size() < lastLevel + 1) init_levelManagers_.resize(lastLevel + 1);
+
+    for (int iLevel = startLevel; iLevel <= lastLevel; iLevel++) {
+      init_levelManagers_[iLevel] = manager;
     }
+  }
 
-    //! Used in SetupInitHierarchy() to access levelManagers_
-    //! Inputs i=-1 and i=size() are allowed to simplify calls to hierarchy->Setup()
-    Teuchos::RCP<FactoryManagerBase> InitLvlMngr(int levelID, int lastLevelID) const {
+  //! Used in SetupInitHierarchy() to access levelManagers_
+  //! Inputs i=-1 and i=size() are allowed to simplify calls to hierarchy->Setup()
+  Teuchos::RCP<FactoryManagerBase> InitLvlMngr(int levelID, int lastLevelID) const {
+    // Please not that the order of the 'if' statements is important.
 
-      // Please not that the order of the 'if' statements is important.
+    if (levelID == -1) return Teuchos::null;               // when this routine is called with levelID == '-1', it means that we are processing the finest Level (there is no finer level)
+    if (levelID == lastLevelID + 1) return Teuchos::null;  // when this routine is called with levelID == 'lastLevelID+1', it means that we are processing the last level (ie: there is no nextLevel...)
 
-      if (levelID == -1)                    return Teuchos::null; // when this routine is called with levelID == '-1', it means that we are processing the finest Level (there is no finer level)
-      if (levelID == lastLevelID+1)         return Teuchos::null; // when this routine is called with levelID == 'lastLevelID+1', it means that we are processing the last level (ie: there is no nextLevel...)
-
-      if (0       == init_levelManagers_.size()) {                     // default factory manager.
-        // the default manager is shared across levels, initialized only if needed and deleted with the HierarchyManager.
-        static RCP<FactoryManagerBase> defaultMngr = rcp(new FactoryManager());
-        return defaultMngr;
-      }
-      if (levelID >= init_levelManagers_.size()) return init_levelManagers_[init_levelManagers_.size()-1]; // last levelManager is used for all the remaining levels.
-
-      return init_levelManagers_[levelID]; // throw exception if out of bound.
+    if (0 == init_levelManagers_.size()) {  // default factory manager.
+      // the default manager is shared across levels, initialized only if needed and deleted with the HierarchyManager.
+      static RCP<FactoryManagerBase> defaultMngr = rcp(new FactoryManager());
+      return defaultMngr;
     }
+    if (levelID >= init_levelManagers_.size()) return init_levelManagers_[init_levelManagers_.size() - 1];  // last levelManager is used for all the remaining levels.
+
+    return init_levelManagers_[levelID];  // throw exception if out of bound.
+  }
 
-    //! nullspace can be embedded in the ML parameter list
-    int     nullspaceDim_;
-    double* nullspace_;
+  //! nullspace can be embedded in the ML parameter list
+  int nullspaceDim_;
+  double* nullspace_;
 
-    //! export aggregates
-    bool    bExportAggregates_; //!< if set to true an AggregationExportFactory is used to export aggregation information (default = false)
+  //! export aggregates
+  bool bExportAggregates_;  //!< if set to true an AggregationExportFactory is used to export aggregation information (default = false)
 
-    //! list of user-defined transfer Factories
-    //! We use this vector to add some special user-given factories to the Hierarchy (RAPFactory)
-    //! This way the user can extend the standard functionality of the MLParameterListInterpreter beyond the
-    //! capabibilities of ML.
-    std::vector<RCP<FactoryBase> > TransferFacts_;
+  //! list of user-defined transfer Factories
+  //! We use this vector to add some special user-given factories to the Hierarchy (RAPFactory)
+  //! This way the user can extend the standard functionality of the MLParameterListInterpreter beyond the
+  //! capabibilities of ML.
+  std::vector<RCP<FactoryBase> > TransferFacts_;
 
-    //! list of levelManagers for adaptive smoothed aggregation
-    //! initialization phase
-    Array<RCP<FactoryManagerBase> > init_levelManagers_;
+  //! list of levelManagers for adaptive smoothed aggregation
+  //! initialization phase
+  Array<RCP<FactoryManagerBase> > init_levelManagers_;
 
-    //@{ Operator configuration
+  //@{ Operator configuration
 
-    //! Setup Operator object
-    //! overloaded from HierarchyManager to set nDofsPerNode
-    virtual void SetupOperator(Operator & Op) const;
+  //! Setup Operator object
+  //! overloaded from HierarchyManager to set nDofsPerNode
+  virtual void SetupOperator(Operator& Op) const;
 
-    //! Matrix configuration storage
-    int blksize_;
-    //@}
+  //! Matrix configuration storage
+  int blksize_;
+  //@}
 
-  }; // class AdaptiveSaMLParameterListInterpreter
+};  // class AdaptiveSaMLParameterListInterpreter
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_SHORT
 #endif /* MUELU_ADAPTIVESAMLPARAMTERLISTINTERPRETER_DECL_HPP_ */
diff --git a/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_def.hpp b/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_def.hpp
index 93d73cb53c5c..f18873a234d1 100644
--- a/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_def.hpp
+++ b/packages/muelu/src/Interface/MueLu_AdaptiveSaMLParameterListInterpreter_def.hpp
@@ -54,404 +54,404 @@
 
 // Read a parameter value from a parameter list and store it into a variable named 'varName'
 #define MUELU_READ_PARAM(paramList, paramStr, varType, defaultValue, varName) \
-  varType varName = defaultValue; if (paramList.isParameter(paramStr)) varName = paramList.get<varType>(paramStr);
+  varType varName = defaultValue;                                             \
+  if (paramList.isParameter(paramStr)) varName = paramList.get<varType>(paramStr);
 
 // Read a parameter value from a paraeter list and copy it into a new parameter list (with another parameter name)
 #define MUELU_COPY_PARAM(paramList, paramStr, varType, defaultValue, outParamList, outParamStr) \
-  if (paramList.isParameter(paramStr))                                  \
-    outParamList.set<varType>(outParamStr, paramList.get<varType>(paramStr)); \
-  else outParamList.set<varType>(outParamStr, defaultValue);            \
+  if (paramList.isParameter(paramStr))                                                          \
+    outParamList.set<varType>(outParamStr, paramList.get<varType>(paramStr));                   \
+  else                                                                                          \
+    outParamList.set<varType>(outParamStr, defaultValue);
 
 namespace MueLu {
 
-  namespace AdaptiveDetails {
-    template <class SC, class LO, class GO, class NO>
-    Teuchos::RCP<MueLu::SmootherFactory<SC, LO, GO, NO> > getSmoother(Teuchos::ParameterList & list) {
-      using SF = MueLu::SmootherFactory<SC, LO, GO, NO>;
-      ParameterListInterpreter interpreter(list);
-      return Teuchos::rcp_const_cast<SF>(Teuchos::rcp_dynamic_cast<const SF>(interpreter.GetFactoryManager(0)->GetFactory("Smoother")));
-    }
+namespace AdaptiveDetails {
+template <class SC, class LO, class GO, class NO>
+Teuchos::RCP<MueLu::SmootherFactory<SC, LO, GO, NO> > getSmoother(Teuchos::ParameterList& list) {
+  using SF = MueLu::SmootherFactory<SC, LO, GO, NO>;
+  ParameterListInterpreter interpreter(list);
+  return Teuchos::rcp_const_cast<SF>(Teuchos::rcp_dynamic_cast<const SF>(interpreter.GetFactoryManager(0)->GetFactory("Smoother")));
+}
+}  // namespace AdaptiveDetails
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AdaptiveSaMLParameterListInterpreter(Teuchos::ParameterList& paramList, std::vector<RCP<FactoryBase> > factoryList)
+  : TransferFacts_(factoryList)
+  , blksize_(1) {
+  SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AdaptiveSaMLParameterListInterpreter(const std::string& xmlFileName, std::vector<RCP<FactoryBase> > factoryList)
+  : nullspace_(NULL)
+  , TransferFacts_(factoryList)
+  , blksize_(1) {
+  Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
+  SetParameterList(*paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList_in) {
+  Teuchos::ParameterList paramList = paramList_in;
+
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));  // TODO: use internal out (GetOStream())
+
+  //
+  // Read top-level of the parameter list
+  //
+
+  // hard-coded default values == ML defaults according to the manual
+  MUELU_READ_PARAM(paramList, "ML output", int, 0, verbosityLevel);
+  MUELU_READ_PARAM(paramList, "max levels", int, 10, maxLevels);
+  MUELU_READ_PARAM(paramList, "PDE equations", int, 1, nDofsPerNode);
+
+  MUELU_READ_PARAM(paramList, "coarse: max size", int, 128, maxCoarseSize);
+
+  MUELU_READ_PARAM(paramList, "aggregation: type", std::string, "Uncoupled", agg_type);
+  // MUELU_READ_PARAM(paramList, "aggregation: threshold",                double,                 0.0,       agg_threshold);
+  MUELU_READ_PARAM(paramList, "aggregation: damping factor", double, (double)4 / (double)3, agg_damping);
+  // MUELU_READ_PARAM(paramList, "aggregation: smoothing sweeps",            int,                   1,       agg_smoothingsweeps);
+  MUELU_READ_PARAM(paramList, "aggregation: nodes per aggregate", int, 1, minPerAgg);
+
+  MUELU_READ_PARAM(paramList, "null space: type", std::string, "default vectors", nullspaceType);
+  MUELU_READ_PARAM(paramList, "null space: dimension", int, -1, nullspaceDim);      // TODO: ML default not in documentation
+  MUELU_READ_PARAM(paramList, "null space: vectors", double*, NULL, nullspaceVec);  // TODO: ML default not in documentation
+
+  MUELU_READ_PARAM(paramList, "energy minimization: enable", bool, false, bEnergyMinimization);
+
+  //
+  // Move smoothers/aggregation/coarse parameters to sublists
+  //
+
+  // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
+  // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
+  ParameterList paramListWithSubList;
+  MueLu::CreateSublists(paramList, paramListWithSubList);
+  paramList = paramListWithSubList;  // swap
+
+  // std::cout << std::endl << "Parameter list after CreateSublists" << std::endl;
+  // std::cout << paramListWithSubList << std::endl;
+
+  int maxNbrAlreadySelected = 0;
+
+  // Matrix option
+  this->blksize_ = nDofsPerNode;
+
+  // Translate verbosity parameter
+  Teuchos::EVerbosityLevel eVerbLevel = Teuchos::VERB_NONE;
+  if (verbosityLevel == 0) eVerbLevel = Teuchos::VERB_NONE;
+  if (verbosityLevel > 0) eVerbLevel = Teuchos::VERB_LOW;
+  if (verbosityLevel > 4) eVerbLevel = Teuchos::VERB_MEDIUM;
+  if (verbosityLevel > 7) eVerbLevel = Teuchos::VERB_HIGH;
+  if (verbosityLevel > 9) eVerbLevel = Teuchos::VERB_EXTREME;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(agg_type != "Uncoupled", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter::Setup(): parameter \"aggregation: type\": only 'Uncoupled' aggregation is supported.");
+
+  // Create MueLu factories
+  // RCP<NullspaceFactory>     nspFact = rcp(new NullspaceFactory());
+  RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+  // dropFact->SetVerbLevel(toMueLuVerbLevel(eVerbLevel));
+
+  // Uncoupled aggregation
+  RCP<UncoupledAggregationFactory> AggFact = rcp(new UncoupledAggregationFactory());
+  AggFact->SetMinNodesPerAggregate(minPerAgg);  // TODO should increase if run anything other than 1D
+  AggFact->SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
+  AggFact->SetOrdering("natural");
+
+  if (verbosityLevel > 3) {  // TODO fix me: Setup is a static function: we cannot use GetOStream without an object...
+    *out << "========================= Aggregate option summary =========================" << std::endl;
+    *out << "min Nodes per aggregate :               " << minPerAgg << std::endl;
+    *out << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
+    *out << "aggregate ordering :                    natural" << std::endl;
+    *out << "=============================================================================" << std::endl;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AdaptiveSaMLParameterListInterpreter(Teuchos::ParameterList & paramList, std::vector<RCP<FactoryBase> > factoryList) : TransferFacts_(factoryList), blksize_(1) {
-    SetParameterList(paramList);
+  RCP<Factory> PFact;
+  RCP<Factory> RFact;
+  RCP<Factory> PtentFact = rcp(new TentativePFactory());
+  if (agg_damping == 0.0 && bEnergyMinimization == false) {
+    // tentative prolongation operator (PA-AMG)
+    PFact = PtentFact;
+    RFact = rcp(new TransPFactory());
+  } else if (agg_damping != 0.0 && bEnergyMinimization == false) {
+    // smoothed aggregation (SA-AMG)
+    RCP<SaPFactory> SaPFact = rcp(new SaPFactory());
+    SaPFact->SetParameter("sa: damping factor", ParameterEntry(agg_damping));
+    PFact = SaPFact;
+    RFact = rcp(new TransPFactory());
+  } else if (bEnergyMinimization == true) {
+    // Petrov Galerkin PG-AMG smoothed aggregation (energy minimization in ML)
+    PFact = rcp(new PgPFactory());
+    RFact = rcp(new GenericRFactory());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AdaptiveSaMLParameterListInterpreter(const std::string & xmlFileName, std::vector<RCP<FactoryBase> > factoryList) : nullspace_(NULL), TransferFacts_(factoryList), blksize_(1) {
-    Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
-    SetParameterList(*paramList);
+  RCP<RAPFactory> AcFact = rcp(new RAPFactory());
+  for (size_t i = 0; i < TransferFacts_.size(); i++) {
+    AcFact->AddTransferFactory(TransferFacts_[i]);  // THIS WILL BE REPLACED with a call to the MLParamterListInterpreter
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList & paramList_in) {
-    Teuchos::ParameterList paramList = paramList_in;
+  //
+  // Nullspace factory
+  //
 
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout)); // TODO: use internal out (GetOStream())
+  // Set fine level nullspace
+  // extract pre-computed nullspace from ML parameter list
+  // store it in nullspace_ and nullspaceDim_
+  if (nullspaceType != "default vectors") {
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceType != "pre-computed", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (no pre-computed null space). error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceDim == -1, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace dim == -1). error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceVec == NULL, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace == NULL). You have to provide a valid fine-level nullspace in \'null space: vectors\'");
 
-    //
-    // Read top-level of the parameter list
-    //
+    nullspaceDim_ = nullspaceDim;
+    nullspace_    = nullspaceVec;
+  }
 
-    // hard-coded default values == ML defaults according to the manual
-    MUELU_READ_PARAM(paramList, "ML output",                                int,                   0,       verbosityLevel);
-    MUELU_READ_PARAM(paramList, "max levels",                               int,                  10,       maxLevels);
-    MUELU_READ_PARAM(paramList, "PDE equations",                            int,                   1,       nDofsPerNode);
+  Teuchos::RCP<NullspaceFactory> nspFact = Teuchos::rcp(new NullspaceFactory());
+  nspFact->SetFactory("Nullspace", PtentFact);
+
+  //
+  // Hierarchy + FactoryManager
+  //
+
+  // Hierarchy options
+  this->SetVerbLevel(toMueLuVerbLevel(eVerbLevel));
+  this->numDesiredLevel_ = maxLevels;
+  this->maxCoarseSize_   = maxCoarseSize;
+
+  // init smoother
+  RCP<SmootherFactory> initSmootherFact = Teuchos::null;
+  if (paramList.isSublist("init smoother")) {
+    ParameterList& initList = paramList.sublist("init smoother");  // TODO move this before for loop
+    initSmootherFact        = AdaptiveDetails::getSmoother<SC, LO, GO, NO>(initList);
+  } else {
+    std::string ifpackType = "RELAXATION";
+    Teuchos::ParameterList smootherParamList;
+    smootherParamList.set("relaxation: type", "symmetric Gauss-Seidel");
+    smootherParamList.set("smoother: sweeps", 1);
+    smootherParamList.set("smoother: damping factor", 1.0);
+    RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother(ifpackType, smootherParamList, 0));
+
+    initSmootherFact = rcp(new SmootherFactory());
+    initSmootherFact->SetSmootherPrototypes(smooProto, smooProto);
+  }
 
-    MUELU_READ_PARAM(paramList, "coarse: max size",                         int,                 128,       maxCoarseSize);
+  //
+  // Coarse Smoother
+  //
+  ParameterList& coarseList = paramList.sublist("coarse: list");
+  //    coarseList.get("smoother: type", "Amesos-KLU"); // set default
+  RCP<SmootherFactory> coarseFact = AdaptiveDetails::getSmoother<SC, LO, GO, NO>(coarseList);
 
-    MUELU_READ_PARAM(paramList, "aggregation: type",                std::string,         "Uncoupled",       agg_type);
-    //MUELU_READ_PARAM(paramList, "aggregation: threshold",                double,                 0.0,       agg_threshold);
-    MUELU_READ_PARAM(paramList, "aggregation: damping factor",           double, (double)4/(double)3,       agg_damping);
-    //MUELU_READ_PARAM(paramList, "aggregation: smoothing sweeps",            int,                   1,       agg_smoothingsweeps);
-    MUELU_READ_PARAM(paramList, "aggregation: nodes per aggregate",         int,                   1,       minPerAgg);
+  // Smoothers Top Level Parameters
 
-    MUELU_READ_PARAM(paramList, "null space: type",                 std::string,   "default vectors",       nullspaceType);
-    MUELU_READ_PARAM(paramList, "null space: dimension",                    int,                  -1,       nullspaceDim); // TODO: ML default not in documentation
-    MUELU_READ_PARAM(paramList, "null space: vectors",                   double*,               NULL,       nullspaceVec); // TODO: ML default not in documentation
+  RCP<ParameterList> topLevelSmootherParam = ExtractSetOfParameters(paramList, "smoother");
+  // std::cout << std::endl << "Top level smoother parameters:" << std::endl;
+  // std::cout << *topLevelSmootherParam << std::endl;
 
-    MUELU_READ_PARAM(paramList, "energy minimization: enable",             bool,               false,       bEnergyMinimization);
+  //
 
+  // Prepare factory managers
+  // TODO: smootherFact can be reuse accross level if same parameters/no specific parameterList
 
+  for (int levelID = 0; levelID < maxLevels; levelID++) {
     //
-    // Move smoothers/aggregation/coarse parameters to sublists
+    // Level FactoryManager
     //
 
-    // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
-    // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
-    ParameterList paramListWithSubList;
-    MueLu::CreateSublists(paramList, paramListWithSubList);
-    paramList = paramListWithSubList; // swap
-
-    // std::cout << std::endl << "Parameter list after CreateSublists" << std::endl;
-    // std::cout << paramListWithSubList << std::endl;
-
-    int maxNbrAlreadySelected = 0;
-
-    // Matrix option
-    this->blksize_ = nDofsPerNode;
-
-    // Translate verbosity parameter
-    Teuchos::EVerbosityLevel eVerbLevel = Teuchos::VERB_NONE;
-    if (verbosityLevel == 0) eVerbLevel = Teuchos::VERB_NONE;
-    if (verbosityLevel >  0) eVerbLevel = Teuchos::VERB_LOW;
-    if (verbosityLevel >  4) eVerbLevel = Teuchos::VERB_MEDIUM;
-    if (verbosityLevel >  7) eVerbLevel = Teuchos::VERB_HIGH;
-    if (verbosityLevel >  9) eVerbLevel = Teuchos::VERB_EXTREME;
-
-    TEUCHOS_TEST_FOR_EXCEPTION(agg_type != "Uncoupled", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter::Setup(): parameter \"aggregation: type\": only 'Uncoupled' aggregation is supported.");
-
-    // Create MueLu factories
-    // RCP<NullspaceFactory>     nspFact = rcp(new NullspaceFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    //dropFact->SetVerbLevel(toMueLuVerbLevel(eVerbLevel));
-
-    // Uncoupled aggregation
-    RCP<UncoupledAggregationFactory> AggFact = rcp(new UncoupledAggregationFactory());
-    AggFact->SetMinNodesPerAggregate(minPerAgg); //TODO should increase if run anything other than 1D
-    AggFact->SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
-    AggFact->SetOrdering("natural");
-
-    if (verbosityLevel > 3) { // TODO fix me: Setup is a static function: we cannot use GetOStream without an object...
-      *out << "========================= Aggregate option summary =========================" << std::endl;
-      *out << "min Nodes per aggregate :               " << minPerAgg << std::endl;
-      *out << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
-      *out << "aggregate ordering :                    natural" << std::endl;
-      *out << "=============================================================================" << std::endl;
-    }
-
-    RCP<Factory> PFact;
-    RCP<Factory> RFact;
-    RCP<Factory> PtentFact = rcp( new TentativePFactory() );
-    if (agg_damping == 0.0 && bEnergyMinimization == false) {
-      // tentative prolongation operator (PA-AMG)
-      PFact = PtentFact;
-      RFact = rcp( new TransPFactory() );
-    } else if (agg_damping != 0.0 && bEnergyMinimization == false) {
-      // smoothed aggregation (SA-AMG)
-      RCP<SaPFactory> SaPFact =  rcp( new SaPFactory() );
-      SaPFact->SetParameter("sa: damping factor", ParameterEntry(agg_damping));
-      PFact  = SaPFact;
-      RFact  = rcp( new TransPFactory() );
-    } else if (bEnergyMinimization == true) {
-      // Petrov Galerkin PG-AMG smoothed aggregation (energy minimization in ML)
-      PFact  = rcp( new PgPFactory() );
-      RFact  = rcp( new GenericRFactory() );
-    }
-
-    RCP<RAPFactory> AcFact = rcp( new RAPFactory() );
-    for (size_t i = 0; i<TransferFacts_.size(); i++) {
-      AcFact->AddTransferFactory(TransferFacts_[i]); // THIS WILL BE REPLACED with a call to the MLParamterListInterpreter
-    }
+    RCP<FactoryManager> manager     = rcp(new FactoryManager());
+    RCP<FactoryManager> initmanager = rcp(new FactoryManager());
 
     //
-    // Nullspace factory
+    // Smoothers
     //
 
-    // Set fine level nullspace
-    // extract pre-computed nullspace from ML parameter list
-    // store it in nullspace_ and nullspaceDim_
-    if (nullspaceType != "default vectors") {
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceType != "pre-computed", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (no pre-computed null space). error.");
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceDim  == -1,             Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace dim == -1). error.");
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceVec  == NULL,           Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace == NULL). You have to provide a valid fine-level nullspace in \'null space: vectors\'");
-
-      nullspaceDim_ = nullspaceDim;
-      nullspace_    = nullspaceVec;
-    }
-
-    Teuchos::RCP<NullspaceFactory> nspFact = Teuchos::rcp(new NullspaceFactory());
-    nspFact->SetFactory("Nullspace", PtentFact);
-
-    //
-    // Hierarchy + FactoryManager
-    //
-
-    // Hierarchy options
-    this->SetVerbLevel(toMueLuVerbLevel(eVerbLevel));
-    this->numDesiredLevel_ = maxLevels;
-    this->maxCoarseSize_   = maxCoarseSize;
-
-    // init smoother
-    RCP<SmootherFactory> initSmootherFact = Teuchos::null;
-    if(paramList.isSublist("init smoother")) {
-      ParameterList& initList = paramList.sublist("init smoother"); // TODO move this before for loop
-      initSmootherFact = AdaptiveDetails::getSmoother<SC,LO,GO,NO>(initList);
-    } else {
-      std::string ifpackType = "RELAXATION";
-      Teuchos::ParameterList smootherParamList;
-      smootherParamList.set("relaxation: type", "symmetric Gauss-Seidel");
-      smootherParamList.set("smoother: sweeps", 1);
-      smootherParamList.set("smoother: damping factor", 1.0);
-      RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother(ifpackType, smootherParamList, 0) );
-
-      initSmootherFact = rcp( new SmootherFactory() );
-      initSmootherFact->SetSmootherPrototypes(smooProto, smooProto);
+    {
+      // Merge level-specific parameters with global parameters. level-specific parameters takes precedence.
+      // TODO: unit-test this part alone
+
+      ParameterList levelSmootherParam = GetMLSubList(paramList, "smoother", levelID);  // copy
+      MergeParameterList(*topLevelSmootherParam, levelSmootherParam, false);            /* false = do no overwrite levelSmootherParam parameters by topLevelSmootherParam parameters */
+      // std::cout << std::endl << "Merged List for level  " << levelID << std::endl;
+      // std::cout << levelSmootherParam << std::endl;
+
+      // RCP<SmootherFactory> smootherFact = this->GetSmootherFactory(levelSmootherParam); // TODO: missing AFact input arg.
+      RCP<SmootherFactory> smootherFact = AdaptiveDetails::getSmoother<SC, LO, GO, NO>(levelSmootherParam);
+      manager->SetFactory("Smoother", smootherFact);
+      smootherFact->DisableMultipleCallCheck();
+
+      initmanager->SetFactory("Smoother", initSmootherFact);
+      initmanager->SetFactory("CoarseSolver", initSmootherFact);
+      initSmootherFact->DisableMultipleCallCheck();
     }
 
     //
-    // Coarse Smoother
-    //
-    ParameterList& coarseList = paramList.sublist("coarse: list");
-    //    coarseList.get("smoother: type", "Amesos-KLU"); // set default
-    RCP<SmootherFactory> coarseFact = AdaptiveDetails::getSmoother<SC,LO,GO,NO>(coarseList);
-
-    // Smoothers Top Level Parameters
-
-    RCP<ParameterList> topLevelSmootherParam = ExtractSetOfParameters(paramList, "smoother");
-    // std::cout << std::endl << "Top level smoother parameters:" << std::endl;
-    // std::cout << *topLevelSmootherParam << std::endl;
-
+    // Misc
     //
 
-    // Prepare factory managers
-    // TODO: smootherFact can be reuse accross level if same parameters/no specific parameterList
-
-    for (int levelID=0; levelID < maxLevels; levelID++) {
-
-      //
-      // Level FactoryManager
-      //
-
-      RCP<FactoryManager> manager = rcp(new FactoryManager());
-      RCP<FactoryManager> initmanager = rcp(new FactoryManager());
-
-      //
-      // Smoothers
-      //
-
-      {
-        // Merge level-specific parameters with global parameters. level-specific parameters takes precedence.
-        // TODO: unit-test this part alone
-
-        ParameterList levelSmootherParam = GetMLSubList(paramList, "smoother", levelID); // copy
-        MergeParameterList(*topLevelSmootherParam, levelSmootherParam, false); /* false = do no overwrite levelSmootherParam parameters by topLevelSmootherParam parameters */
-        // std::cout << std::endl << "Merged List for level  " << levelID << std::endl;
-        // std::cout << levelSmootherParam << std::endl;
-
-        //RCP<SmootherFactory> smootherFact = this->GetSmootherFactory(levelSmootherParam); // TODO: missing AFact input arg.
-        RCP<SmootherFactory> smootherFact = AdaptiveDetails::getSmoother<SC,LO,GO,NO>(levelSmootherParam);
-        manager->SetFactory("Smoother", smootherFact);
-        smootherFact->DisableMultipleCallCheck();
-
-        initmanager->SetFactory("Smoother", initSmootherFact);
-        initmanager->SetFactory("CoarseSolver", initSmootherFact);
-        initSmootherFact->DisableMultipleCallCheck();
-
-      }
-
-      //
-      // Misc
-      //
-
-      Teuchos::rcp_dynamic_cast<PFactory>(PFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<PFactory>(PtentFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(RFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(coarseFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(dropFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(AggFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(AcFact)->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(nspFact)->DisableMultipleCallCheck();
-
-      manager->SetFactory("CoarseSolver", coarseFact); // TODO: should not be done in the loop
-      manager->SetFactory("Graph", dropFact);
-      manager->SetFactory("Aggregates", AggFact);
-      manager->SetFactory("DofsPerNode", dropFact);
-      manager->SetFactory("A", AcFact);
-      manager->SetFactory("P", PFact);
-      manager->SetFactory("Ptent", PtentFact);
-      manager->SetFactory("R", RFact);
-      manager->SetFactory("Nullspace", nspFact);
-
-      //initmanager->SetFactory("CoarseSolver", coarseFact);
-      initmanager->SetFactory("Graph", dropFact);
-      initmanager->SetFactory("Aggregates", AggFact);
-      initmanager->SetFactory("DofsPerNode", dropFact);
-      initmanager->SetFactory("A", AcFact);
-      initmanager->SetFactory("P", PtentFact); // use nonsmoothed transfers
-      initmanager->SetFactory("Ptent", PtentFact);
-      initmanager->SetFactory("R", RFact);
-      initmanager->SetFactory("Nullspace", nspFact);
-
-      this->AddFactoryManager(levelID, 1, manager);
-      this->AddInitFactoryManager(levelID, 1, initmanager);
-    } // for (level loop)
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupInitHierarchy(Hierarchy & H) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(!H.GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError, "No fine level operator");
-
-    RCP<Level>    l  = H.GetLevel(0);
-    RCP<Operator> Op = l->Get<RCP<Operator> >("A");
-    SetupOperator(*Op); // use overloaded SetupMatrix routine
-    this->SetupExtra(H);
-
-    // Setup Hierarchy
-    H.SetMaxCoarseSize(this->maxCoarseSize_); // TODO
-
-    int  levelID     = 0;
-    int  lastLevelID = this->numDesiredLevel_ - 1;
-    bool isLastLevel = false;
-
-    while(!isLastLevel) {
-      bool r = H.Setup(levelID,
-          InitLvlMngr(levelID-1, lastLevelID),
-          InitLvlMngr(levelID,   lastLevelID),
-          InitLvlMngr(levelID+1, lastLevelID));
-
-      isLastLevel = r || (levelID == lastLevelID);
-      levelID++;
-    }
+    Teuchos::rcp_dynamic_cast<PFactory>(PFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<PFactory>(PtentFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(RFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(coarseFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(dropFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(AggFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(AcFact)->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<SingleLevelFactoryBase>(nspFact)->DisableMultipleCallCheck();
+
+    manager->SetFactory("CoarseSolver", coarseFact);  // TODO: should not be done in the loop
+    manager->SetFactory("Graph", dropFact);
+    manager->SetFactory("Aggregates", AggFact);
+    manager->SetFactory("DofsPerNode", dropFact);
+    manager->SetFactory("A", AcFact);
+    manager->SetFactory("P", PFact);
+    manager->SetFactory("Ptent", PtentFact);
+    manager->SetFactory("R", RFact);
+    manager->SetFactory("Nullspace", nspFact);
+
+    // initmanager->SetFactory("CoarseSolver", coarseFact);
+    initmanager->SetFactory("Graph", dropFact);
+    initmanager->SetFactory("Aggregates", AggFact);
+    initmanager->SetFactory("DofsPerNode", dropFact);
+    initmanager->SetFactory("A", AcFact);
+    initmanager->SetFactory("P", PtentFact);  // use nonsmoothed transfers
+    initmanager->SetFactory("Ptent", PtentFact);
+    initmanager->SetFactory("R", RFact);
+    initmanager->SetFactory("Nullspace", nspFact);
+
+    this->AddFactoryManager(levelID, 1, manager);
+    this->AddInitFactoryManager(levelID, 1, initmanager);
+  }  // for (level loop)
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupInitHierarchy(Hierarchy& H) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(!H.GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError, "No fine level operator");
+
+  RCP<Level> l     = H.GetLevel(0);
+  RCP<Operator> Op = l->Get<RCP<Operator> >("A");
+  SetupOperator(*Op);  // use overloaded SetupMatrix routine
+  this->SetupExtra(H);
+
+  // Setup Hierarchy
+  H.SetMaxCoarseSize(this->maxCoarseSize_);  // TODO
+
+  int levelID      = 0;
+  int lastLevelID  = this->numDesiredLevel_ - 1;
+  bool isLastLevel = false;
+
+  while (!isLastLevel) {
+    bool r = H.Setup(levelID,
+                     InitLvlMngr(levelID - 1, lastLevelID),
+                     InitLvlMngr(levelID, lastLevelID),
+                     InitLvlMngr(levelID + 1, lastLevelID));
+
+    isLastLevel = r || (levelID == lastLevelID);
+    levelID++;
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy & H) const {
-
-    // set fine level null space
-    // usually this null space is provided from outside (by the user) using
-    // the ML parameter lists.
-    if (this->nullspace_ != NULL) {
-      RCP<Level> fineLevel = H.GetLevel(0);
-      const RCP<const Map> rowMap = fineLevel->Get< RCP<Matrix> >("A")->getRowMap();
-      RCP<MultiVector> nullspace = MultiVectorFactory::Build(rowMap, nullspaceDim_, true);
-
-      for ( size_t i=0; i < Teuchos::as<size_t>(nullspaceDim_); i++) {
-        Teuchos::ArrayRCP<Scalar> nullspacei = nullspace->getDataNonConst(i);
-        const size_t              myLength   = nullspace->getLocalLength();
-
-        for (size_t j = 0; j < myLength; j++) {
-          nullspacei[j] = nullspace_[i*myLength + j];
-        }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy& H) const {
+  // set fine level null space
+  // usually this null space is provided from outside (by the user) using
+  // the ML parameter lists.
+  if (this->nullspace_ != NULL) {
+    RCP<Level> fineLevel        = H.GetLevel(0);
+    const RCP<const Map> rowMap = fineLevel->Get<RCP<Matrix> >("A")->getRowMap();
+    RCP<MultiVector> nullspace  = MultiVectorFactory::Build(rowMap, nullspaceDim_, true);
+
+    for (size_t i = 0; i < Teuchos::as<size_t>(nullspaceDim_); i++) {
+      Teuchos::ArrayRCP<Scalar> nullspacei = nullspace->getDataNonConst(i);
+      const size_t myLength                = nullspace->getLocalLength();
+
+      for (size_t j = 0; j < myLength; j++) {
+        nullspacei[j] = nullspace_[i * myLength + j];
       }
-
-      fineLevel->Set("Nullspace", nullspace);
     }
 
-    // keep aggregates
-    H.Keep("Aggregates", HierarchyManager::GetFactoryManager(0)->GetFactory("Aggregates").get());
-
-    ///////////////////////////////
-
-    // build hierarchy for initialization
-    SetupInitHierarchy(H);
+    fineLevel->Set("Nullspace", nullspace);
+  }
 
-    {
-      // do some iterations with the built hierarchy to improve the null space
-      Teuchos::RCP<MueLu::Level> Finest = H.GetLevel(0);  // get finest level,MueLu::NoFactory::get()
-      Teuchos::RCP<MultiVector> nspVector2 = Finest->Get<Teuchos::RCP<MultiVector> >("Nullspace");
+  // keep aggregates
+  H.Keep("Aggregates", HierarchyManager::GetFactoryManager(0)->GetFactory("Aggregates").get());
 
-      Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("orig_nsp.vec", *nspVector2);
+  ///////////////////////////////
 
-      RCP<Matrix> Op = Finest->Get<RCP<Matrix> >("A");
-      Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("A.mat", *Op);
+  // build hierarchy for initialization
+  SetupInitHierarchy(H);
 
+  {
+    // do some iterations with the built hierarchy to improve the null space
+    Teuchos::RCP<MueLu::Level> Finest    = H.GetLevel(0);  // get finest level,MueLu::NoFactory::get()
+    Teuchos::RCP<MultiVector> nspVector2 = Finest->Get<Teuchos::RCP<MultiVector> >("Nullspace");
 
-      Teuchos::RCP<MultiVector> homogRhsVec = MultiVectorFactory::Build(nspVector2->getMap(),nspVector2->getNumVectors(),true);
-      homogRhsVec->putScalar(0.0);
+    Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("orig_nsp.vec", *nspVector2);
 
-      // do 1 multigrid cycle for improving the null space by "solving"
-      //     A B_f = 0
-      // where A is the system matrix and B_f the fine level null space vectors
-      H.Iterate(*homogRhsVec, *nspVector2, 1, false);
+    RCP<Matrix> Op = Finest->Get<RCP<Matrix> >("A");
+    Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mat", *Op);
 
-      // store improved fine level null space
-      Finest->Set("Nullspace",nspVector2);
+    Teuchos::RCP<MultiVector> homogRhsVec = MultiVectorFactory::Build(nspVector2->getMap(), nspVector2->getNumVectors(), true);
+    homogRhsVec->putScalar(0.0);
 
-      Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write("new_nsp.vec", *nspVector2);
+    // do 1 multigrid cycle for improving the null space by "solving"
+    //     A B_f = 0
+    // where A is the system matrix and B_f the fine level null space vectors
+    H.Iterate(*homogRhsVec, *nspVector2, 1, false);
 
-      //H.Delete("CoarseSolver", init_levelManagers_[0]->GetFactory("CoarseSolver").get());
-    }
+    // store improved fine level null space
+    Finest->Set("Nullspace", nspVector2);
 
-    {
-      // do some clean up.
-      // remove all old default factories. Build new ones for the second build.
-      // this is a little bit tricky to understand
-      for(size_t k=0; k < HierarchyManager::getNumFactoryManagers(); k++) {
-        HierarchyManager::GetFactoryManager(k)->Clean();
-        //Teuchos::rcp_dynamic_cast<const SingleLevelFactoryBase>(HierarchyManager::GetFactoryManager(k)->GetFactory("Smoother"))->DisableMultipleCallCheck(); // after changing to MLParamterListInterpreter functions
-      }
-      // not sure about this. i only need it if Smoother is defined explicitely (not using default smoother)
-      // need this: otherwise RAPFactory::Build is complaining on level 0
-      //            and TentativePFactory::Build is complaining on level 1
-      Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(0)->GetFactory("A"))->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(1)->GetFactory("P"))->DisableMultipleCallCheck();
-      Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(1)->GetFactory("Ptent"))->DisableMultipleCallCheck();
-
-      HierarchyManager::SetupHierarchy(H);
-    }
+    Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("new_nsp.vec", *nspVector2);
 
+    // H.Delete("CoarseSolver", init_levelManagers_[0]->GetFactory("CoarseSolver").get());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "Transfer factory is not derived from TwoLevelFactoryBase. Since transfer factories will be handled by the RAPFactory they have to be derived from TwoLevelFactoryBase!");
-    TransferFacts_.push_back(factory);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NumTransferFactories() const {
-    return TransferFacts_.size();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator & Op) const {
-    try {
-      Matrix& A = dynamic_cast<Matrix&>(Op);
-      if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blksize_))
-        this->GetOStream(Warnings0) << "Setting matrix block size to " << blksize_ << " (value of the parameter in the list) "
-            << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl;
-
-      A.SetFixedBlockSize(blksize_);
-
-    } catch (std::bad_cast& e) {
-      this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
+  {
+    // do some clean up.
+    // remove all old default factories. Build new ones for the second build.
+    // this is a little bit tricky to understand
+    for (size_t k = 0; k < HierarchyManager::getNumFactoryManagers(); k++) {
+      HierarchyManager::GetFactoryManager(k)->Clean();
+      // Teuchos::rcp_dynamic_cast<const SingleLevelFactoryBase>(HierarchyManager::GetFactoryManager(k)->GetFactory("Smoother"))->DisableMultipleCallCheck(); // after changing to MLParamterListInterpreter functions
     }
+    // not sure about this. i only need it if Smoother is defined explicitely (not using default smoother)
+    // need this: otherwise RAPFactory::Build is complaining on level 0
+    //            and TentativePFactory::Build is complaining on level 1
+    Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(0)->GetFactory("A"))->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(1)->GetFactory("P"))->DisableMultipleCallCheck();
+    Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(HierarchyManager::GetFactoryManager(1)->GetFactory("Ptent"))->DisableMultipleCallCheck();
+
+    HierarchyManager::SetupHierarchy(H);
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "Transfer factory is not derived from TwoLevelFactoryBase. Since transfer factories will be handled by the RAPFactory they have to be derived from TwoLevelFactoryBase!");
+  TransferFacts_.push_back(factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NumTransferFactories() const {
+  return TransferFacts_.size();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AdaptiveSaMLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator& Op) const {
+  try {
+    Matrix& A = dynamic_cast<Matrix&>(Op);
+    if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blksize_))
+      this->GetOStream(Warnings0) << "Setting matrix block size to " << blksize_ << " (value of the parameter in the list) "
+                                  << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl;
+
+    A.SetFixedBlockSize(blksize_);
+
+  } catch (std::bad_cast& e) {
+    this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
+  }
+}
 
-} // namespace MueLu
-
+}  // namespace MueLu
 
 #endif /* MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_DEF_HPP_ */
diff --git a/packages/muelu/src/Interface/MueLu_FactoryFactory_decl.hpp b/packages/muelu/src/Interface/MueLu_FactoryFactory_decl.hpp
index 1832f0704030..8e60177cc07f 100644
--- a/packages/muelu/src/Interface/MueLu_FactoryFactory_decl.hpp
+++ b/packages/muelu/src/Interface/MueLu_FactoryFactory_decl.hpp
@@ -159,7 +159,6 @@
 #include "MueLu_Zoltan2Interface.hpp"
 #include "MueLu_NodePartitionInterface.hpp"
 
-
 #include "MueLu_CoalesceDropFactory_kokkos.hpp"
 #include "MueLu_GeometricInterpolationPFactory_kokkos.hpp"
 #include "MueLu_NullspaceFactory_kokkos.hpp"
@@ -187,777 +186,780 @@
 
 namespace MueLu {
 
-  /*! class FactoryFactory
+/*! class FactoryFactory
 
-  @brief Factory that can generate other factories from
+@brief Factory that can generate other factories from
 
 
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FactoryFactory : public BaseClass {
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FactoryFactory : public BaseClass {
 #undef MUELU_FACTORYFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-    typedef std::map<std::string, RCP<const FactoryBase>  > FactoryMap; // TODO: remove
-    typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
-
-  public:
-
-    /// \brief: Interpret Factory parameter list and build new factory
-    ///
-    /// \param param [in]: ParameterEntry being either the parameter list containing the "factory" parameter declaring the factory type (e.g., "TrilinosSmoother") or being a plain Parameter containing the factory type as value
-    /// \param factoryMapIn [in]: FactoryMap containing a map between factory name (e.g., "smootherFact1") and corresponding factory of all previously defined factories
-    /// \param factoryManagersIn [in]: FactoryManagerMap containing a map between group names and Factory manager objects. Needed for factories with sub-factory managers.
-    ///
-    /// Parameter List Parsing:
-    /// ---------
-    ///     <Parameter name="smootherFact0" type="string" value="TrilinosSmoother"/>
-    ///
-    /// or:
-    ///
-    ///     <ParameterList name="smootherFact1">
-    ///       <Parameter name="factory" type="string" value="TrilinosSmoother"/>
-    ///       ...
-    ///     </ParameterList>
-    ///
-    virtual RCP<const FactoryBase> BuildFactory(const Teuchos::ParameterEntry& param, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      // Find factory
-      std::string            factoryName;
-      Teuchos::ParameterList paramList;
-      if (!param.isList()) {
-        factoryName = Teuchos::getValue<std::string>(param);
-      } else {
-        paramList = Teuchos::getValue<Teuchos::ParameterList>(param);
-        factoryName = paramList.get<std::string>("factory");
-      }
+  typedef std::map<std::string, RCP<const FactoryBase> > FactoryMap;  // TODO: remove
+  typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
+
+ public:
+  /// \brief: Interpret Factory parameter list and build new factory
+  ///
+  /// \param param [in]: ParameterEntry being either the parameter list containing the "factory" parameter declaring the factory type (e.g., "TrilinosSmoother") or being a plain Parameter containing the factory type as value
+  /// \param factoryMapIn [in]: FactoryMap containing a map between factory name (e.g., "smootherFact1") and corresponding factory of all previously defined factories
+  /// \param factoryManagersIn [in]: FactoryManagerMap containing a map between group names and Factory manager objects. Needed for factories with sub-factory managers.
+  ///
+  /// Parameter List Parsing:
+  /// ---------
+  ///     <Parameter name="smootherFact0" type="string" value="TrilinosSmoother"/>
+  ///
+  /// or:
+  ///
+  ///     <ParameterList name="smootherFact1">
+  ///       <Parameter name="factory" type="string" value="TrilinosSmoother"/>
+  ///       ...
+  ///     </ParameterList>
+  ///
+  virtual RCP<const FactoryBase> BuildFactory(const Teuchos::ParameterEntry& param, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    // Find factory
+    std::string factoryName;
+    Teuchos::ParameterList paramList;
+    if (!param.isList()) {
+      factoryName = Teuchos::getValue<std::string>(param);
+    } else {
+      paramList   = Teuchos::getValue<Teuchos::ParameterList>(param);
+      factoryName = paramList.get<std::string>("factory");
+    }
 
-      // TODO: see how Teko handles this (=> register factories).
-      if (factoryName == "AggregateQualityEstimateFactory")       return Build2<AggregateQualityEstimateFactory>       (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "AggregationExportFactory")              return Build2<AggregationExportFactory>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "AmalgamationFactory")                   return Build2<AmalgamationFactory>                   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedCoarseMapFactory")               return Build2<BlockedCoarseMapFactory>               (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedRAPFactory")                     return BuildRAPFactory<BlockedRAPFactory>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BrickAggregationFactory")               return Build2<BrickAggregationFactory>               (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ClassicalMapFactory")                   return Build2<ClassicalMapFactory>                   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ClassicalPFactory")                     return Build2<ClassicalPFactory>                     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CloneRepartitionInterface")             return Build2<CloneRepartitionInterface>             (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoarseMapFactory")                      return Build2<CoarseMapFactory>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoarseningVisualizationFactory")        return Build2<CoarseningVisualizationFactory>        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoalesceDropFactory")                   return Build2<CoalesceDropFactory>                   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SmooVecCoalesceDropFactory")            return Build2<SmooVecCoalesceDropFactory>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ConstraintFactory")                     return Build2<ConstraintFactory>                     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoordinatesTransferFactory")            return Build2<CoordinatesTransferFactory>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "DirectSolver")                          return BuildDirectSolver                             (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "DropNegativeEntriesFactory")            return Build2<DropNegativeEntriesFactory>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "EminPFactory")                          return Build2<EminPFactory>                          (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "FilteredAFactory")                      return Build2<FilteredAFactory>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "FineLevelInputDataFactory")             return Build2<FineLevelInputDataFactory>             (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "GeneralGeometricPFactory")              return Build2<GeneralGeometricPFactory>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ReplicatePFactory")                     return Build2<ReplicatePFactory>                     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CombinePFactory")                       return Build2<CombinePFactory>                       (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "GenericRFactory")                       return Build2<GenericRFactory>                       (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "GeometricInterpolationPFactory")        return Build2<GeometricInterpolationPFactory>        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "HybridAggregationFactory")              return Build2<HybridAggregationFactory>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "InterfaceAggregationFactory")           return Build2<InterfaceAggregationFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "InterfaceMappingTransferFactory")       return Build2<InterfaceMappingTransferFactory>       (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "InverseApproximationFactory")           return Build2<InverseApproximationFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "InitialBlockNumberFactory")             return Build2<InitialBlockNumberFactory>             (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "LineDetectionFactory")                  return Build2<LineDetectionFactory>                  (paramList, factoryMapIn, factoryManagersIn);
-      // LocalOrdinalTransferFactory is a utility factory that can be used for multiple things, so there is no default
-      //      if (factoryName == "LocalOrdinalTransferFactory")           return Build2<LocalOrdinalTransferFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "MapTransferFactory")                    return Build2<MapTransferFactory>                    (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "MatrixAnalysisFactory")                 return Build2<MatrixAnalysisFactory>                 (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "MultiVectorTransferFactory")            return Build2<MultiVectorTransferFactory>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "NoFactory")                             return MueLu::NoFactory::getRCP();
-      if (factoryName == "NoSmoother")                            return rcp(new SmootherFactory(Teuchos::null));
-      if (factoryName == "NotayAggregationFactory")               return Build2<NotayAggregationFactory>               (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "NullspaceFactory")                      return Build2<NullspaceFactory>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "NullspacePresmoothFactory")             return Build2<NullspacePresmoothFactory>             (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "PatternFactory")                        return Build2<PatternFactory>                        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "PgPFactory")                            return Build2<PgPFactory>                            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SaPFactory")                            return Build2<SaPFactory>                            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RAPFactory")                            return BuildRAPFactory<RAPFactory>                   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RAPShiftFactory")                       return BuildRAPFactory<RAPShiftFactory>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RebalanceAcFactory")                    return Build2<RebalanceAcFactory>                    (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RebalanceTransferFactory")              return Build2<RebalanceTransferFactory>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RegionRFactory")                        return Build2<RegionRFactory>                        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RegionRFactory_kokkos")                 return Build2<RegionRFactory_kokkos>                 (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ReorderBlockAFactory")                  return Build2<ReorderBlockAFactory>                  (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RepartitionInterface")                  return Build2<RepartitionInterface>                  (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ScaledNullspaceFactory")                return Build2<ScaledNullspaceFactory>                (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SegregatedAFactory")                    return Build2<SegregatedAFactory>                    (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SemiCoarsenPFactory")                   return Build2<SemiCoarsenPFactory>                   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "StructuredAggregationFactory")          return Build2<StructuredAggregationFactory>          (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "StructuredLineDetectionFactory")        return Build2<StructuredLineDetectionFactory>        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SubBlockAFactory")                      return Build2<SubBlockAFactory>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "TentativePFactory")                     return Build2<TentativePFactory>                     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ToggleCoordinatesTransferFactory")      return BuildToggleCoordinatesTransferFactory         (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "TogglePFactory")                        return BuildTogglePFactory<TogglePFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "TransPFactory")                         return Build2<TransPFactory>                         (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RfromP_Or_TransP")                      return Build2<RfromP_Or_TransP>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "TrilinosSmoother")                      return BuildTrilinosSmoother                         (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "UncoupledAggregationFactory")           return Build2<UncoupledAggregationFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "UnsmooshFactory")                       return Build2<UnsmooshFactory>                       (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "UserAggregationFactory")                return Build2<UserAggregationFactory>                (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "UserPFactory")                          return Build2<UserPFactory>                          (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "VariableDofLaplacianFactory")           return Build2<VariableDofLaplacianFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "ZeroSubBlockAFactory")                  return Build2<ZeroSubBlockAFactory>                  (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoalesceDropFactory_kokkos")            return Build2<CoalesceDropFactory_kokkos>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "GeometricInterpolationPFactory_kokkos") return Build2<GeometricInterpolationPFactory_kokkos> (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "NullspaceFactory_kokkos")               return Build2<NullspaceFactory_kokkos>               (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SaPFactory_kokkos")                     return Build2<SaPFactory_kokkos>                     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SemiCoarsenPFactory_kokkos")            return Build2<SemiCoarsenPFactory_kokkos>            (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "StructuredAggregationFactory_kokkos")   return Build2<StructuredAggregationFactory_kokkos>   (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "TentativePFactory_kokkos")              return Build2<TentativePFactory_kokkos>              (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "MatrixFreeTentativePFactory")           return Build2<MatrixFreeTentativePFactory>           (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "UncoupledAggregationFactory_kokkos")    return Build2<UncoupledAggregationFactory_kokkos>    (paramList, factoryMapIn, factoryManagersIn);
-
-      // Handle removed Kokkos factories
-      if (factoryName == "CoarseMapFactory_kokkos")               return Build2<CoarseMapFactory>                      (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "CoordinatesTransferFactory_kokkos")     return Build2<CoordinatesTransferFactory>            (paramList, factoryMapIn, factoryManagersIn);
-
-      if (factoryName == "ZoltanInterface") {
+    // TODO: see how Teko handles this (=> register factories).
+    if (factoryName == "AggregateQualityEstimateFactory") return Build2<AggregateQualityEstimateFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "AggregationExportFactory") return Build2<AggregationExportFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "AmalgamationFactory") return Build2<AmalgamationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedCoarseMapFactory") return Build2<BlockedCoarseMapFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedRAPFactory") return BuildRAPFactory<BlockedRAPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BrickAggregationFactory") return Build2<BrickAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ClassicalMapFactory") return Build2<ClassicalMapFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ClassicalPFactory") return Build2<ClassicalPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CloneRepartitionInterface") return Build2<CloneRepartitionInterface>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoarseMapFactory") return Build2<CoarseMapFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoarseningVisualizationFactory") return Build2<CoarseningVisualizationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoalesceDropFactory") return Build2<CoalesceDropFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SmooVecCoalesceDropFactory") return Build2<SmooVecCoalesceDropFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ConstraintFactory") return Build2<ConstraintFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoordinatesTransferFactory") return Build2<CoordinatesTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "DirectSolver") return BuildDirectSolver(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "DropNegativeEntriesFactory") return Build2<DropNegativeEntriesFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "EminPFactory") return Build2<EminPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "FilteredAFactory") return Build2<FilteredAFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "FineLevelInputDataFactory") return Build2<FineLevelInputDataFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "GeneralGeometricPFactory") return Build2<GeneralGeometricPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ReplicatePFactory") return Build2<ReplicatePFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CombinePFactory") return Build2<CombinePFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "GenericRFactory") return Build2<GenericRFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "GeometricInterpolationPFactory") return Build2<GeometricInterpolationPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "HybridAggregationFactory") return Build2<HybridAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "InterfaceAggregationFactory") return Build2<InterfaceAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "InterfaceMappingTransferFactory") return Build2<InterfaceMappingTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "InverseApproximationFactory") return Build2<InverseApproximationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "InitialBlockNumberFactory") return Build2<InitialBlockNumberFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "LineDetectionFactory") return Build2<LineDetectionFactory>(paramList, factoryMapIn, factoryManagersIn);
+    // LocalOrdinalTransferFactory is a utility factory that can be used for multiple things, so there is no default
+    //      if (factoryName == "LocalOrdinalTransferFactory")           return Build2<LocalOrdinalTransferFactory>           (paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "MapTransferFactory") return Build2<MapTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "MatrixAnalysisFactory") return Build2<MatrixAnalysisFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "MultiVectorTransferFactory") return Build2<MultiVectorTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "NoFactory") return MueLu::NoFactory::getRCP();
+    if (factoryName == "NoSmoother") return rcp(new SmootherFactory(Teuchos::null));
+    if (factoryName == "NotayAggregationFactory") return Build2<NotayAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "NullspaceFactory") return Build2<NullspaceFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "NullspacePresmoothFactory") return Build2<NullspacePresmoothFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "PatternFactory") return Build2<PatternFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "PgPFactory") return Build2<PgPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SaPFactory") return Build2<SaPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RAPFactory") return BuildRAPFactory<RAPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RAPShiftFactory") return BuildRAPFactory<RAPShiftFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RebalanceAcFactory") return Build2<RebalanceAcFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RebalanceTransferFactory") return Build2<RebalanceTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RegionRFactory") return Build2<RegionRFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RegionRFactory_kokkos") return Build2<RegionRFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ReorderBlockAFactory") return Build2<ReorderBlockAFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RepartitionInterface") return Build2<RepartitionInterface>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ScaledNullspaceFactory") return Build2<ScaledNullspaceFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SegregatedAFactory") return Build2<SegregatedAFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SemiCoarsenPFactory") return Build2<SemiCoarsenPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "StructuredAggregationFactory") return Build2<StructuredAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "StructuredLineDetectionFactory") return Build2<StructuredLineDetectionFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SubBlockAFactory") return Build2<SubBlockAFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TentativePFactory") return Build2<TentativePFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ToggleCoordinatesTransferFactory") return BuildToggleCoordinatesTransferFactory(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TogglePFactory") return BuildTogglePFactory<TogglePFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TransPFactory") return Build2<TransPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RfromP_Or_TransP") return Build2<RfromP_Or_TransP>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TrilinosSmoother") return BuildTrilinosSmoother(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UncoupledAggregationFactory") return Build2<UncoupledAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UnsmooshFactory") return Build2<UnsmooshFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UserAggregationFactory") return Build2<UserAggregationFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UserPFactory") return Build2<UserPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "VariableDofLaplacianFactory") return Build2<VariableDofLaplacianFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "ZeroSubBlockAFactory") return Build2<ZeroSubBlockAFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoalesceDropFactory_kokkos") return Build2<CoalesceDropFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "GeometricInterpolationPFactory_kokkos") return Build2<GeometricInterpolationPFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "NullspaceFactory_kokkos") return Build2<NullspaceFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SaPFactory_kokkos") return Build2<SaPFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SemiCoarsenPFactory_kokkos") return Build2<SemiCoarsenPFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "StructuredAggregationFactory_kokkos") return Build2<StructuredAggregationFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TentativePFactory_kokkos") return Build2<TentativePFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "MatrixFreeTentativePFactory") return Build2<MatrixFreeTentativePFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UncoupledAggregationFactory_kokkos") return Build2<UncoupledAggregationFactory_kokkos>(paramList, factoryMapIn, factoryManagersIn);
+
+    // Handle removed Kokkos factories
+    if (factoryName == "CoarseMapFactory_kokkos") return Build2<CoarseMapFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "CoordinatesTransferFactory_kokkos") return Build2<CoordinatesTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+
+    if (factoryName == "ZoltanInterface") {
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-        return Build2<ZoltanInterface>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<ZoltanInterface>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a ZoltanInterface object: Zoltan is disabled: HAVE_MUELU_ZOLTAN && HAVE_MPI == false.");
-#endif // HAVE_MUELU_ZOLTAN && HAVE_MPI
-      }
-      if (factoryName == "Zoltan2Interface") {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a ZoltanInterface object: Zoltan is disabled: HAVE_MUELU_ZOLTAN && HAVE_MPI == false.");
+#endif  // HAVE_MUELU_ZOLTAN && HAVE_MPI
+    }
+    if (factoryName == "Zoltan2Interface") {
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
-        return Build2<Zoltan2Interface>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<Zoltan2Interface>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a Zoltan2Interface object: Zoltan2 is disabled: HAVE_MUELU_ZOLTAN2 && HAVE_MPI == false.");
-#endif // HAVE_MUELU_ZOLTAN2 && HAVE_MPI
-      }
-      if (factoryName == "IsorropiaInterface") {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a Zoltan2Interface object: Zoltan2 is disabled: HAVE_MUELU_ZOLTAN2 && HAVE_MPI == false.");
+#endif  // HAVE_MUELU_ZOLTAN2 && HAVE_MPI
+    }
+    if (factoryName == "IsorropiaInterface") {
 #if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
-        return Build2<IsorropiaInterface>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<IsorropiaInterface>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a IsorropiaInterface object: Isorropia is disabled: HAVE_MUELU_ISORROPIA && HAVE_MPI == false.");
-#endif // HAVE_MUELU_ZOLTAN2 && HAVE_MPI
-      }
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a IsorropiaInterface object: Isorropia is disabled: HAVE_MUELU_ISORROPIA && HAVE_MPI == false.");
+#endif  // HAVE_MUELU_ZOLTAN2 && HAVE_MPI
+    }
 
-      if (factoryName == "NodePartitionInterface") {
+    if (factoryName == "NodePartitionInterface") {
 #if defined(HAVE_MPI)
-        return Build2<NodePartitionInterface>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<NodePartitionInterface>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a NodePartitionInterface object: HAVE_MPI == false.");
-#endif // HAVE_MPI
-      }
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a NodePartitionInterface object: HAVE_MPI == false.");
+#endif  // HAVE_MPI
+    }
 
-      if (factoryName == "RepartitionFactory") {
+    if (factoryName == "RepartitionFactory") {
 #ifdef HAVE_MPI
-        return Build2<RepartitionFactory>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<RepartitionFactory>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a RepartitionFactory object: HAVE_MPI == false.");
-#endif // HAVE_MPI
-      }
-      if (factoryName == "RepartitionHeuristicFactory") {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a RepartitionFactory object: HAVE_MPI == false.");
+#endif  // HAVE_MPI
+    }
+    if (factoryName == "RepartitionHeuristicFactory") {
 #ifdef HAVE_MPI
-        return Build2<RepartitionHeuristicFactory>(paramList, factoryMapIn, factoryManagersIn);
+      return Build2<RepartitionHeuristicFactory>(paramList, factoryMapIn, factoryManagersIn);
 #else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a RepartitionHeuristicFactory object: HAVE_MPI == false.");
-#endif // HAVE_MPI
-      }
-      // Blocked factories
-      if (factoryName == "BlockedCoordinatesTransferFactory")     return BuildBlockedCoordFactory<BlockedCoordinatesTransferFactory>     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedDirectSolver")             return BuildBlockedDirectSolver(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedGaussSeidelSmoother")      return BuildBlockedSmoother<BlockedGaussSeidelSmoother>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedJacobiSmoother")           return BuildBlockedSmoother<BlockedJacobiSmoother>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BlockedPFactory")                 return BuildBlockedFactory<BlockedPFactory>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "BraessSarazinSmoother")           return BuildBlockedSmoother<BraessSarazinSmoother>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "IndefiniteBlockDiagonalSmoother") return BuildBlockedSmoother<IndefBlockedDiagonalSmoother>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SimpleSmoother")                  return BuildBlockedSmoother<SimpleSmoother>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SchurComplementFactory")          return Build2<SchurComplementFactory> (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RebalanceBlockRestrictionFactory")return BuildBlockedFactory<RebalanceBlockRestrictionFactory>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RebalanceBlockAcFactory")         return BuildBlockedFactory<RebalanceBlockAcFactory>(paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "RebalanceBlockInterpolationFactory") return BuildBlockedFactory<RebalanceBlockInterpolationFactory>(paramList, factoryMapIn, factoryManagersIn);
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory:BuildFactory(): Cannot create a RepartitionHeuristicFactory object: HAVE_MPI == false.");
+#endif  // HAVE_MPI
+    }
+    // Blocked factories
+    if (factoryName == "BlockedCoordinatesTransferFactory") return BuildBlockedCoordFactory<BlockedCoordinatesTransferFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedDirectSolver") return BuildBlockedDirectSolver(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedGaussSeidelSmoother") return BuildBlockedSmoother<BlockedGaussSeidelSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedJacobiSmoother") return BuildBlockedSmoother<BlockedJacobiSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BlockedPFactory") return BuildBlockedFactory<BlockedPFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "BraessSarazinSmoother") return BuildBlockedSmoother<BraessSarazinSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "IndefiniteBlockDiagonalSmoother") return BuildBlockedSmoother<IndefBlockedDiagonalSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SimpleSmoother") return BuildBlockedSmoother<SimpleSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SchurComplementFactory") return Build2<SchurComplementFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RebalanceBlockRestrictionFactory") return BuildBlockedFactory<RebalanceBlockRestrictionFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RebalanceBlockAcFactory") return BuildBlockedFactory<RebalanceBlockAcFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RebalanceBlockInterpolationFactory") return BuildBlockedFactory<RebalanceBlockInterpolationFactory>(paramList, factoryMapIn, factoryManagersIn);
 #ifdef HAVE_MPI
-      if (factoryName == "RepartitionBlockDiagonalFactory")    return Build2<RepartitionBlockDiagonalFactory>    (paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "RepartitionBlockDiagonalFactory") return Build2<RepartitionBlockDiagonalFactory>(paramList, factoryMapIn, factoryManagersIn);
 #endif
 #ifdef HAVE_MUELU_TEKO
-      if (factoryName == "TekoSmoother")                    return BuildTekoSmoother(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TekoSmoother") return BuildTekoSmoother(paramList, factoryMapIn, factoryManagersIn);
 #endif
-      if (factoryName == "UzawaSmoother")                   return BuildBlockedSmoother<UzawaSmoother>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "UzawaSmoother") return BuildBlockedSmoother<UzawaSmoother>(paramList, factoryMapIn, factoryManagersIn);
 
       // Matlab factories
 #ifdef HAVE_MUELU_MATLAB
-      if (factoryName == "TwoLevelMatlabFactory")           return Build2<TwoLevelMatlabFactory>        (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "SingleLevelMatlabFactory")        return Build2<SingleLevelMatlabFactory>     (paramList, factoryMapIn, factoryManagersIn);
-      if (factoryName == "MatlabSmoother")                  return BuildMatlabSmoother                  (paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "TwoLevelMatlabFactory") return Build2<TwoLevelMatlabFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "SingleLevelMatlabFactory") return Build2<SingleLevelMatlabFactory>(paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "MatlabSmoother") return BuildMatlabSmoother(paramList, factoryMapIn, factoryManagersIn);
 #endif
 
 #ifdef HAVE_MUELU_INTREPID2
-      if (factoryName == "IntrepidPCoarsenFactory")           return Build2<IntrepidPCoarsenFactory>        (paramList, factoryMapIn, factoryManagersIn);
+    if (factoryName == "IntrepidPCoarsenFactory") return Build2<IntrepidPCoarsenFactory>(paramList, factoryMapIn, factoryManagersIn);
 #endif
 
-      // Use a user defined factories (in <Factories> node)
-      if (factoryMapIn.find(factoryName) != factoryMapIn.end()) {
-        TEUCHOS_TEST_FOR_EXCEPTION((param.isList() && (++paramList.begin() != paramList.end())), Exceptions::RuntimeError,
-                                   "MueLu::FactoryFactory: Error during the parsing of: " << std::endl << paramList << std::endl
-                                   << "'" << factoryName << "' is not a factory name but an existing instance of a factory." << std::endl
-                                   << "Extra parameters cannot be specified after the creation of the object." << std::endl << std::endl
-                                   << "Correct syntaxes includes:" << std::endl
-                                   << " <Parameter name=\"...\" type=\"string\" value=\"" << factoryName << "\"/>" << std::endl
-                                   << "or" << std::endl
-                                   << " <ParameterList name=\"...\"><Parameter name=\"factory\" type=\"string\" value=\"" << factoryName << "\"/></ParameterList>" << std::endl
-                                   );
-
-        return factoryMapIn.find(factoryName)->second;
-      }
+    // Use a user defined factories (in <Factories> node)
+    if (factoryMapIn.find(factoryName) != factoryMapIn.end()) {
+      TEUCHOS_TEST_FOR_EXCEPTION((param.isList() && (++paramList.begin() != paramList.end())), Exceptions::RuntimeError,
+                                 "MueLu::FactoryFactory: Error during the parsing of: " << std::endl
+                                                                                        << paramList << std::endl
+                                                                                        << "'" << factoryName << "' is not a factory name but an existing instance of a factory." << std::endl
+                                                                                        << "Extra parameters cannot be specified after the creation of the object." << std::endl
+                                                                                        << std::endl
+                                                                                        << "Correct syntaxes includes:" << std::endl
+                                                                                        << " <Parameter name=\"...\" type=\"string\" value=\"" << factoryName << "\"/>" << std::endl
+                                                                                        << "or" << std::endl
+                                                                                        << " <ParameterList name=\"...\"><Parameter name=\"factory\" type=\"string\" value=\"" << factoryName << "\"/></ParameterList>" << std::endl);
+
+      return factoryMapIn.find(factoryName)->second;
+    }
 
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory: unknown factory name : " << factoryName);
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::FactoryFactory: unknown factory name : " << factoryName);
 
-      TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
-    }
+    TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+  }
 
-    //
-    //
-    //
+  //
+  //
+  //
 
-    // FOLLOWING FUNCTIONS SHOULD LIVE WITH THE CORRESPONDING CLASS
+  // FOLLOWING FUNCTIONS SHOULD LIVE WITH THE CORRESPONDING CLASS
 
-    //
-    //
-    //
+  //
+  //
+  //
 
-#define arraysize(ar)  (sizeof(ar) / sizeof(ar[0]))
+#define arraysize(ar) (sizeof(ar) / sizeof(ar[0]))
 
-    template <class T> // T must implement the Factory interface
-    RCP<T> Build(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> factory = rcp(new T());
+  template <class T>  // T must implement the Factory interface
+  RCP<T> Build(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> factory = rcp(new T());
 
-      const char* strarray[] = {"A", "P", "R", "Graph", "UnAmalgamationInfo", "Aggregates", "Nullspace", "TransferFactory", "DofsPerNode"};
-      std::vector<std::string> v(strarray, strarray + arraysize(strarray));
-      for (size_t i = 0; i < v.size(); ++i)
-        if (paramList.isParameter(v[i]))
-          factory->SetFactory(v[i], BuildFactory(paramList.getEntry(v[i]), factoryMapIn, factoryManagersIn));
+    const char* strarray[] = {"A", "P", "R", "Graph", "UnAmalgamationInfo", "Aggregates", "Nullspace", "TransferFactory", "DofsPerNode"};
+    std::vector<std::string> v(strarray, strarray + arraysize(strarray));
+    for (size_t i = 0; i < v.size(); ++i)
+      if (paramList.isParameter(v[i]))
+        factory->SetFactory(v[i], BuildFactory(paramList.getEntry(v[i]), factoryMapIn, factoryManagersIn));
 
-      return factory;
-    }
+    return factory;
+  }
 
-    template <class T> // T must implement the Factory interface
-    RCP<T> Build2(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> factory = rcp(new T());
+  template <class T>  // T must implement the Factory interface
+  RCP<T> Build2(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> factory = rcp(new T());
 
-      ParameterList paramListWithFactories;
+    ParameterList paramListWithFactories;
 
-      // Read the RCP<Factory> parameters of the class T
-      RCP<const ParameterList> validParamList = factory->GetValidParameterList(); // TODO check for Teuchos::null (no parameter list validation)
-      TEUCHOS_TEST_FOR_EXCEPTION(validParamList == Teuchos::null, Exceptions::RuntimeError, "FactoryFactory::Build2: default parameter list is null. Please fix this.");
-      for (ParameterList::ConstIterator param = validParamList->begin(); param != validParamList->end(); ++param) {
-        const std::string& pName = validParamList->name(param);
+    // Read the RCP<Factory> parameters of the class T
+    RCP<const ParameterList> validParamList = factory->GetValidParameterList();  // TODO check for Teuchos::null (no parameter list validation)
+    TEUCHOS_TEST_FOR_EXCEPTION(validParamList == Teuchos::null, Exceptions::RuntimeError, "FactoryFactory::Build2: default parameter list is null. Please fix this.");
+    for (ParameterList::ConstIterator param = validParamList->begin(); param != validParamList->end(); ++param) {
+      const std::string& pName = validParamList->name(param);
 
-        if (!paramList.isParameter(pName)) {
-          // Ignore unknown parameters
-          continue;
-        }
+      if (!paramList.isParameter(pName)) {
+        // Ignore unknown parameters
+        continue;
+      }
 
-        if (validParamList->isType< RCP<const FactoryBase> >(pName)) {
-          // Generate or get factory described by param
-          RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry(pName), factoryMapIn, factoryManagersIn);
-          paramListWithFactories.set(pName, generatingFact);
-        } else if (validParamList->isType<RCP<const ParameterList> >(pName)) {
-          if (pName == "ParameterList") {
-            // NOTE: we cannot use
-            //     subList = sublist(rcpFromRef(paramList), pName)
-            // here as that would result in sublist also being a reference to a temporary object.
-            // The resulting dereferencing in the corresponding factory would then segfault
-            RCP<const ParameterList> subList = Teuchos::sublist(rcp(new ParameterList(paramList)), pName);
-            paramListWithFactories.set(pName, subList);
-          }
-        } else {
-          paramListWithFactories.setEntry(pName, paramList.getEntry(pName));
+      if (validParamList->isType<RCP<const FactoryBase> >(pName)) {
+        // Generate or get factory described by param
+        RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry(pName), factoryMapIn, factoryManagersIn);
+        paramListWithFactories.set(pName, generatingFact);
+      } else if (validParamList->isType<RCP<const ParameterList> >(pName)) {
+        if (pName == "ParameterList") {
+          // NOTE: we cannot use
+          //     subList = sublist(rcpFromRef(paramList), pName)
+          // here as that would result in sublist also being a reference to a temporary object.
+          // The resulting dereferencing in the corresponding factory would then segfault
+          RCP<const ParameterList> subList = Teuchos::sublist(rcp(new ParameterList(paramList)), pName);
+          paramListWithFactories.set(pName, subList);
         }
+      } else {
+        paramListWithFactories.setEntry(pName, paramList.getEntry(pName));
       }
-
-      // Configure the factory
-      factory->SetParameterList(paramListWithFactories);
-
-      return factory;
     }
 
-    template <class T> // T must implement the Factory interface
-    RCP<T> BuildRAPFactory(const Teuchos::ParameterList & paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> factory;
-      if (paramList.isSublist("TransferFactories") == false) {
-        factory = Build2<T>(paramList, factoryMapIn, factoryManagersIn);
-
-      } else {
-        RCP<Teuchos::ParameterList>       paramListNonConst = rcp(new Teuchos::ParameterList(paramList));
-        RCP<const Teuchos::ParameterList> transferFactories = rcp(new Teuchos::ParameterList(*sublist(paramListNonConst, "TransferFactories")));
-
-        paramListNonConst->remove("TransferFactories");
-
-        factory = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
-
-        for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
-          RCP<const FactoryBase> p = BuildFactory(transferFactories->entry(param), factoryMapIn, factoryManagersIn);
-          factory->AddTransferFactory(p);
-        }
-      }
+    // Configure the factory
+    factory->SetParameterList(paramListWithFactories);
 
-      return factory;
-    }
+    return factory;
+  }
 
-    template <class T> // T must implement the Factory interface
-    RCP<T> BuildTogglePFactory(const Teuchos::ParameterList & paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> factory;
-      if (paramList.isSublist("TransferFactories") == false) {
-          //TODO put in an error message: the TogglePFactory needs a TransferFactories sublist!
-        factory = Build2<T>(paramList, factoryMapIn, factoryManagersIn);
+  template <class T>  // T must implement the Factory interface
+  RCP<T> BuildRAPFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> factory;
+    if (paramList.isSublist("TransferFactories") == false) {
+      factory = Build2<T>(paramList, factoryMapIn, factoryManagersIn);
 
-      } else {
-        RCP<Teuchos::ParameterList>       paramListNonConst = rcp(new Teuchos::ParameterList(paramList));
-        RCP<const Teuchos::ParameterList> transferFactories = rcp(new Teuchos::ParameterList(*sublist(paramListNonConst, "TransferFactories")));
-
-        paramListNonConst->remove("TransferFactories");
-
-        // build TogglePFactory
-        factory = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
-
-        // count how many prolongation factories and how many coarse null space factories have been declared.
-        // the numbers must match!
-        int numProlongatorFactories = 0;
-        int numPtentFactories = 0;
-        int numCoarseNspFactories   = 0;
-        for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
-          size_t foundNsp = transferFactories->name(param).find("Nullspace");
-          if (foundNsp != std::string::npos && foundNsp == 0 && transferFactories->name(param).length()==10) {
-            numCoarseNspFactories++;
-            continue;
-          }
-          size_t foundPtent   = transferFactories->name(param).find("Ptent");
-          if (foundPtent != std::string::npos && foundPtent == 0 && transferFactories->name(param).length()==6) {
-            numPtentFactories++;
-            continue;
-          }
-          size_t foundP   = transferFactories->name(param).find("P");
-          if (foundP != std::string::npos && foundP == 0 && transferFactories->name(param).length()==2) {
-            numProlongatorFactories++;
-            continue;
-          }
-        }
-        TEUCHOS_TEST_FOR_EXCEPTION(numProlongatorFactories!=numCoarseNspFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The user has to provide the same number of prolongator and coarse nullspace factories!");
-        TEUCHOS_TEST_FOR_EXCEPTION(numPtentFactories!=numCoarseNspFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The user has to provide the same number of ptent and coarse nullspace factories!");
-        TEUCHOS_TEST_FOR_EXCEPTION(numProlongatorFactories < 2, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The TogglePFactory needs at least two different prolongation operators. The factories have to be provided using the names P%i and Nullspace %i, where %i denotes a number between 1 and 9.");
-
-        // create empty vectors with data
-        std::vector<Teuchos::ParameterEntry> prolongatorFactoryNames(numProlongatorFactories);
-        std::vector<Teuchos::ParameterEntry> coarseNspFactoryNames(numProlongatorFactories);
-        std::vector<Teuchos::ParameterEntry> ptentFactoryNames(numProlongatorFactories);
-
-        for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
-          size_t foundNsp = transferFactories->name(param).find("Nullspace");
-          if (foundNsp != std::string::npos && foundNsp == 0 && transferFactories->name(param).length()==10) {
-            int number = atoi(&(transferFactories->name(param).at(9)));
-                TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numProlongatorFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format Nullspace%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
-                coarseNspFactoryNames[number-1] = transferFactories->entry(param);
-                continue;
-          }
-          size_t foundPtent   = transferFactories->name(param).find("Ptent");
-          if (foundPtent != std::string::npos && foundPtent == 0 && transferFactories->name(param).length()==6) {
-            int number = atoi(&(transferFactories->name(param).at(5)));
-                TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numPtentFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format Ptent%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
-                ptentFactoryNames[number-1] = transferFactories->entry(param);
-                continue;
-          }
-          size_t foundP   = transferFactories->name(param).find("P");
-          if (foundP != std::string::npos && foundP == 0 && transferFactories->name(param).length()==2) {
-            int number = atoi(&(transferFactories->name(param).at(1)));
-                TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numProlongatorFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format P%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
-                prolongatorFactoryNames[number-1] = transferFactories->entry(param);
-                continue;
-          }
-        }
+    } else {
+      RCP<Teuchos::ParameterList> paramListNonConst       = rcp(new Teuchos::ParameterList(paramList));
+      RCP<const Teuchos::ParameterList> transferFactories = rcp(new Teuchos::ParameterList(*sublist(paramListNonConst, "TransferFactories")));
 
-        // register all prolongation factories in TogglePFactory
-        for (std::vector<Teuchos::ParameterEntry>::const_iterator it = prolongatorFactoryNames.begin(); it != prolongatorFactoryNames.end(); ++it) {
-          RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
-          factory->AddProlongatorFactory(p);
-        }
+      paramListNonConst->remove("TransferFactories");
 
-        // register all tentative prolongation factories in TogglePFactory
-        for (std::vector<Teuchos::ParameterEntry>::const_iterator it = ptentFactoryNames.begin(); it != ptentFactoryNames.end(); ++it) {
-          RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
-          factory->AddPtentFactory(p);
-        }
+      factory = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
 
-        // register all coarse nullspace factories in TogglePFactory
-        for (std::vector<Teuchos::ParameterEntry>::const_iterator it = coarseNspFactoryNames.begin(); it != coarseNspFactoryNames.end(); ++it) {
-          RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
-          factory->AddCoarseNullspaceFactory(p);
-        }
+      for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
+        RCP<const FactoryBase> p = BuildFactory(transferFactories->entry(param), factoryMapIn, factoryManagersIn);
+        factory->AddTransferFactory(p);
       }
-      return factory;
     }
 
-    RCP<ToggleCoordinatesTransferFactory> BuildToggleCoordinatesTransferFactory(const Teuchos::ParameterList & paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<ToggleCoordinatesTransferFactory> factory;
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.isSublist("TransferFactories") == false, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransferFactory: the ToggleCoordinatesTransferFactory needs a sublist 'TransferFactories' containing information about the subfactories for coordinate transfer!");
+    return factory;
+  }
 
-      RCP<Teuchos::ParameterList>       paramListNonConst = rcp(new Teuchos::ParameterList(paramList));
+  template <class T>  // T must implement the Factory interface
+  RCP<T> BuildTogglePFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> factory;
+    if (paramList.isSublist("TransferFactories") == false) {
+      // TODO put in an error message: the TogglePFactory needs a TransferFactories sublist!
+      factory = Build2<T>(paramList, factoryMapIn, factoryManagersIn);
+
+    } else {
+      RCP<Teuchos::ParameterList> paramListNonConst       = rcp(new Teuchos::ParameterList(paramList));
       RCP<const Teuchos::ParameterList> transferFactories = rcp(new Teuchos::ParameterList(*sublist(paramListNonConst, "TransferFactories")));
+
       paramListNonConst->remove("TransferFactories");
 
-      // build CoordinatesTransferFactory
-      factory = Build2<ToggleCoordinatesTransferFactory>(*paramListNonConst, factoryMapIn, factoryManagersIn);
+      // build TogglePFactory
+      factory = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
 
-      // count how many coordinate transfer factories have been declared.
+      // count how many prolongation factories and how many coarse null space factories have been declared.
       // the numbers must match!
-      int numCoordTransferFactories = 0;
+      int numProlongatorFactories = 0;
+      int numPtentFactories       = 0;
+      int numCoarseNspFactories   = 0;
       for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
-        size_t foundCoordinates = transferFactories->name(param).find("Coordinates");
-        if (foundCoordinates != std::string::npos && foundCoordinates == 0 && transferFactories->name(param).length()==12) {
-          numCoordTransferFactories++;
+        size_t foundNsp = transferFactories->name(param).find("Nullspace");
+        if (foundNsp != std::string::npos && foundNsp == 0 && transferFactories->name(param).length() == 10) {
+          numCoarseNspFactories++;
+          continue;
+        }
+        size_t foundPtent = transferFactories->name(param).find("Ptent");
+        if (foundPtent != std::string::npos && foundPtent == 0 && transferFactories->name(param).length() == 6) {
+          numPtentFactories++;
+          continue;
+        }
+        size_t foundP = transferFactories->name(param).find("P");
+        if (foundP != std::string::npos && foundP == 0 && transferFactories->name(param).length() == 2) {
+          numProlongatorFactories++;
           continue;
         }
       }
-      TEUCHOS_TEST_FOR_EXCEPTION(numCoordTransferFactories != 2, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransfer: The ToggleCoordinatesTransferFactory needs two (different) coordinate transfer factories. The factories have to be provided using the names Coordinates%i, where %i denotes a number between 1 and 9.");
+      TEUCHOS_TEST_FOR_EXCEPTION(numProlongatorFactories != numCoarseNspFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The user has to provide the same number of prolongator and coarse nullspace factories!");
+      TEUCHOS_TEST_FOR_EXCEPTION(numPtentFactories != numCoarseNspFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The user has to provide the same number of ptent and coarse nullspace factories!");
+      TEUCHOS_TEST_FOR_EXCEPTION(numProlongatorFactories < 2, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: The TogglePFactory needs at least two different prolongation operators. The factories have to be provided using the names P%i and Nullspace %i, where %i denotes a number between 1 and 9.");
 
       // create empty vectors with data
-      std::vector<Teuchos::ParameterEntry> coarseCoordsFactoryNames(numCoordTransferFactories);
+      std::vector<Teuchos::ParameterEntry> prolongatorFactoryNames(numProlongatorFactories);
+      std::vector<Teuchos::ParameterEntry> coarseNspFactoryNames(numProlongatorFactories);
+      std::vector<Teuchos::ParameterEntry> ptentFactoryNames(numProlongatorFactories);
 
       for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
-        size_t foundCoords = transferFactories->name(param).find("Coordinates");
-        if (foundCoords != std::string::npos && foundCoords == 0 && transferFactories->name(param).length()==12) {
-          int number = atoi(&(transferFactories->name(param).at(11)));
-              TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numCoordTransferFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransfer: Please use the format Coordinates%i with %i an integer between 1 and the maximum number of coordinate transfer factories in ToggleCoordinatesTransferFactory!");
-              coarseCoordsFactoryNames[number-1] = transferFactories->entry(param);
-              continue;
+        size_t foundNsp = transferFactories->name(param).find("Nullspace");
+        if (foundNsp != std::string::npos && foundNsp == 0 && transferFactories->name(param).length() == 10) {
+          int number = atoi(&(transferFactories->name(param).at(9)));
+          TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numProlongatorFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format Nullspace%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
+          coarseNspFactoryNames[number - 1] = transferFactories->entry(param);
+          continue;
+        }
+        size_t foundPtent = transferFactories->name(param).find("Ptent");
+        if (foundPtent != std::string::npos && foundPtent == 0 && transferFactories->name(param).length() == 6) {
+          int number = atoi(&(transferFactories->name(param).at(5)));
+          TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numPtentFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format Ptent%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
+          ptentFactoryNames[number - 1] = transferFactories->entry(param);
+          continue;
+        }
+        size_t foundP = transferFactories->name(param).find("P");
+        if (foundP != std::string::npos && foundP == 0 && transferFactories->name(param).length() == 2) {
+          int number = atoi(&(transferFactories->name(param).at(1)));
+          TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numProlongatorFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleP: Please use the format P%i with %i an integer between 1 and the maximum number of prolongation operators in TogglePFactory!");
+          prolongatorFactoryNames[number - 1] = transferFactories->entry(param);
+          continue;
         }
       }
 
-      // register all coarse nullspace factories in TogglePFactory
-      for (std::vector<Teuchos::ParameterEntry>::const_iterator it = coarseCoordsFactoryNames.begin(); it != coarseCoordsFactoryNames.end(); ++it) {
+      // register all prolongation factories in TogglePFactory
+      for (std::vector<Teuchos::ParameterEntry>::const_iterator it = prolongatorFactoryNames.begin(); it != prolongatorFactoryNames.end(); ++it) {
         RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
-        factory->AddCoordTransferFactory(p);
+        factory->AddProlongatorFactory(p);
       }
 
-      return factory;
-    }
+      // register all tentative prolongation factories in TogglePFactory
+      for (std::vector<Teuchos::ParameterEntry>::const_iterator it = ptentFactoryNames.begin(); it != ptentFactoryNames.end(); ++it) {
+        RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
+        factory->AddPtentFactory(p);
+      }
 
-    //! TrilinosSmoother
-    // Parameter List Parsing:
-    //     <ParameterList name="smootherFact1">
-    //       <Parameter name="factory" type="string" value="TrilinosSmoother"/>
-    //       <Parameter name="verbose" type="string" value="Warnings"/>
-    //       <Parameter name="type" type="string" value="RELAXATION"/>
-    //       <ParameterList name="ParameterList">
-    //       ...
-    //       </ParameterList>
-    //     </ParameterList>
-    RCP<FactoryBase> BuildTrilinosSmoother(const Teuchos::ParameterList & paramList, const FactoryMap & factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      if (paramList.begin() == paramList.end())
-        return rcp(new SmootherFactory(rcp(new TrilinosSmoother())));
-
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "TrilinosSmoother", Exceptions::RuntimeError, "");
-
-      // Is it true? TEUCHOS_TEST_FOR_EXCEPTION(!paramList.isParameter("type"), Exceptions::RuntimeError, "TrilinosSmoother: parameter 'type' is mandatory");
-      // type="" is default in TrilinosSmoother, but what happen then?
-
-      std::string type="";            if(paramList.isParameter("type"))          type    = paramList.get<std::string>("type");
-      int         overlap=0;          if(paramList.isParameter("overlap"))       overlap = paramList.get<int>        ("overlap");
-      // std::string verbose;         if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
-      Teuchos::ParameterList params;  if(paramList.isParameter("ParameterList")) params  = paramList.get<Teuchos::ParameterList>("ParameterList");
-
-      // parameters from SmootherFactory
-      //bool bKeepSmootherData = false; if(paramList.isParameter("keep smoother data")) bKeepSmootherData = paramList.get<bool>("keep smoother data");
-
-      // Read in factory information for smoothers (if available...)
-      // NOTE: only a selected number of factories can be used with the Trilinos smoother
-      //       smoothers usually work with the global data available (which is A and the transfers P and R)
-
-      Teuchos::RCP<TrilinosSmoother> trilSmoo = Teuchos::rcp(new TrilinosSmoother(type, params, overlap));
-
-      if (paramList.isParameter("LineDetection_Layers")) {
-        RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("LineDetection_Layers"), factoryMapIn, factoryManagersIn);
-        trilSmoo->SetFactory("LineDetection_Layers", generatingFact);
+      // register all coarse nullspace factories in TogglePFactory
+      for (std::vector<Teuchos::ParameterEntry>::const_iterator it = coarseNspFactoryNames.begin(); it != coarseNspFactoryNames.end(); ++it) {
+        RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
+        factory->AddCoarseNullspaceFactory(p);
       }
-      if (paramList.isParameter("LineDetection_VertLineIds")) {
-        RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("LineDetection_Layers"), factoryMapIn, factoryManagersIn);
-        trilSmoo->SetFactory("LineDetection_Layers", generatingFact);
+    }
+    return factory;
+  }
+
+  RCP<ToggleCoordinatesTransferFactory> BuildToggleCoordinatesTransferFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<ToggleCoordinatesTransferFactory> factory;
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.isSublist("TransferFactories") == false, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransferFactory: the ToggleCoordinatesTransferFactory needs a sublist 'TransferFactories' containing information about the subfactories for coordinate transfer!");
+
+    RCP<Teuchos::ParameterList> paramListNonConst       = rcp(new Teuchos::ParameterList(paramList));
+    RCP<const Teuchos::ParameterList> transferFactories = rcp(new Teuchos::ParameterList(*sublist(paramListNonConst, "TransferFactories")));
+    paramListNonConst->remove("TransferFactories");
+
+    // build CoordinatesTransferFactory
+    factory = Build2<ToggleCoordinatesTransferFactory>(*paramListNonConst, factoryMapIn, factoryManagersIn);
+
+    // count how many coordinate transfer factories have been declared.
+    // the numbers must match!
+    int numCoordTransferFactories = 0;
+    for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
+      size_t foundCoordinates = transferFactories->name(param).find("Coordinates");
+      if (foundCoordinates != std::string::npos && foundCoordinates == 0 && transferFactories->name(param).length() == 12) {
+        numCoordTransferFactories++;
+        continue;
       }
-      if (paramList.isParameter("CoarseNumZLayers")) {
-        RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("CoarseNumZLayers"), factoryMapIn, factoryManagersIn);
-        trilSmoo->SetFactory("CoarseNumZLayers", generatingFact);
+    }
+    TEUCHOS_TEST_FOR_EXCEPTION(numCoordTransferFactories != 2, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransfer: The ToggleCoordinatesTransferFactory needs two (different) coordinate transfer factories. The factories have to be provided using the names Coordinates%i, where %i denotes a number between 1 and 9.");
+
+    // create empty vectors with data
+    std::vector<Teuchos::ParameterEntry> coarseCoordsFactoryNames(numCoordTransferFactories);
+
+    for (Teuchos::ParameterList::ConstIterator param = transferFactories->begin(); param != transferFactories->end(); ++param) {
+      size_t foundCoords = transferFactories->name(param).find("Coordinates");
+      if (foundCoords != std::string::npos && foundCoords == 0 && transferFactories->name(param).length() == 12) {
+        int number = atoi(&(transferFactories->name(param).at(11)));
+        TEUCHOS_TEST_FOR_EXCEPTION(number < 1 || number > numCoordTransferFactories, Exceptions::RuntimeError, "FactoryFactory::BuildToggleCoordinatesTransfer: Please use the format Coordinates%i with %i an integer between 1 and the maximum number of coordinate transfer factories in ToggleCoordinatesTransferFactory!");
+        coarseCoordsFactoryNames[number - 1] = transferFactories->entry(param);
+        continue;
       }
-
-      RCP<SmootherFactory> smooFact = rcp(new SmootherFactory(Teuchos::null));
-      Teuchos::ParameterList smooFactParams;
-      //smooFactParams.setEntry("keep smoother data", paramList.getEntry("keep smoother data"));
-      smooFact->SetParameterList(smooFactParams);
-      smooFact->SetSmootherPrototypes(trilSmoo);
-      return smooFact;
     }
 
-#ifdef HAVE_MUELU_MATLAB
-    //! MatlabSmoother
-    // Parameter List Parsing:
-    //     <ParameterList name="smootherFact1">
-    //       <Parameter name="factory" type="string" value="MatlabSmoother"/>
-    //       <Parameter name="Setup Function" type="string" value="mySmootherSetup.m"/>
-    //       <Parameter name="Solve Function" type="string" value="mySmootherSolve.m"/>
-    //       <!--A is implicitly included in this list and nothing else is needed to get diagonal-->
-    //       <Parameter name="Needs" type="string" value=""/>
-    //       <!--A,x,b are also assumed inputs to the solver: only one additional arg then (diag)-->
-    //       <Parameter name="Number of Solver Args"               type="int" value="1"/>
-    //     </ParameterList>
-    RCP<FactoryBase> BuildMatlabSmoother(const Teuchos::ParameterList & paramList, const FactoryMap & factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      if (paramList.begin() == paramList.end())
-        return rcp(new SmootherFactory(rcp(new MatlabSmoother())));
-
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "MatlabSmoother", Exceptions::RuntimeError, "");
-
-      // Read in factory information for smoothers (if available...)
-      // NOTE: only a selected number of factories can be used with the Trilinos smoother
-      //       smoothers usually work with the global data available (which is A and the transfers P and R)
-
-      Teuchos::RCP<MatlabSmoother> matSmoo = Teuchos::rcp(new MatlabSmoother(paramList));
-
-      return rcp(new SmootherFactory(matSmoo));
+    // register all coarse nullspace factories in TogglePFactory
+    for (std::vector<Teuchos::ParameterEntry>::const_iterator it = coarseCoordsFactoryNames.begin(); it != coarseCoordsFactoryNames.end(); ++it) {
+      RCP<const FactoryBase> p = BuildFactory(*it, factoryMapIn, factoryManagersIn);
+      factory->AddCoordTransferFactory(p);
     }
-#endif
-
-    RCP<FactoryBase> BuildDirectSolver(const Teuchos::ParameterList& paramList, const FactoryMap& /* factoryMapIn */, const FactoryManagerMap& /* factoryManagersIn */) const {
-      if (paramList.begin() == paramList.end())
-        return rcp(new SmootherFactory(rcp(new DirectSolver()), Teuchos::null));
-
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "DirectSolver", Exceptions::RuntimeError, "");
 
-      std::string type;              if(paramList.isParameter("type"))          type = paramList.get<std::string>("type");
-      // std::string verbose;        if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
-      Teuchos::ParameterList params; if(paramList.isParameter("ParameterList")) params  = paramList.get<Teuchos::ParameterList>("ParameterList");
-
-      return rcp(new SmootherFactory(rcp(new DirectSolver(type, params)), Teuchos::null));
+    return factory;
+  }
+
+  //! TrilinosSmoother
+  // Parameter List Parsing:
+  //     <ParameterList name="smootherFact1">
+  //       <Parameter name="factory" type="string" value="TrilinosSmoother"/>
+  //       <Parameter name="verbose" type="string" value="Warnings"/>
+  //       <Parameter name="type" type="string" value="RELAXATION"/>
+  //       <ParameterList name="ParameterList">
+  //       ...
+  //       </ParameterList>
+  //     </ParameterList>
+  RCP<FactoryBase> BuildTrilinosSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    if (paramList.begin() == paramList.end())
+      return rcp(new SmootherFactory(rcp(new TrilinosSmoother())));
+
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "TrilinosSmoother", Exceptions::RuntimeError, "");
+
+    // Is it true? TEUCHOS_TEST_FOR_EXCEPTION(!paramList.isParameter("type"), Exceptions::RuntimeError, "TrilinosSmoother: parameter 'type' is mandatory");
+    // type="" is default in TrilinosSmoother, but what happen then?
+
+    std::string type = "";
+    if (paramList.isParameter("type")) type = paramList.get<std::string>("type");
+    int overlap = 0;
+    if (paramList.isParameter("overlap")) overlap = paramList.get<int>("overlap");
+    // std::string verbose;         if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
+    Teuchos::ParameterList params;
+    if (paramList.isParameter("ParameterList")) params = paramList.get<Teuchos::ParameterList>("ParameterList");
+
+    // parameters from SmootherFactory
+    // bool bKeepSmootherData = false; if(paramList.isParameter("keep smoother data")) bKeepSmootherData = paramList.get<bool>("keep smoother data");
+
+    // Read in factory information for smoothers (if available...)
+    // NOTE: only a selected number of factories can be used with the Trilinos smoother
+    //       smoothers usually work with the global data available (which is A and the transfers P and R)
+
+    Teuchos::RCP<TrilinosSmoother> trilSmoo = Teuchos::rcp(new TrilinosSmoother(type, params, overlap));
+
+    if (paramList.isParameter("LineDetection_Layers")) {
+      RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("LineDetection_Layers"), factoryMapIn, factoryManagersIn);
+      trilSmoo->SetFactory("LineDetection_Layers", generatingFact);
+    }
+    if (paramList.isParameter("LineDetection_VertLineIds")) {
+      RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("LineDetection_Layers"), factoryMapIn, factoryManagersIn);
+      trilSmoo->SetFactory("LineDetection_Layers", generatingFact);
+    }
+    if (paramList.isParameter("CoarseNumZLayers")) {
+      RCP<const FactoryBase> generatingFact = BuildFactory(paramList.getEntry("CoarseNumZLayers"), factoryMapIn, factoryManagersIn);
+      trilSmoo->SetFactory("CoarseNumZLayers", generatingFact);
     }
 
-    template <class T> // T must implement the Factory interface
-    RCP<FactoryBase> BuildBlockedSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      // read in sub lists
-      RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
-
-      // internal vector of factory managers
-      std::vector<RCP<FactoryManager> > facManagers;
-
-      // loop over all "block%i" sublists in parameter list
-      int blockid = 1;
-      bool blockExists = true;
-      while (blockExists == true) {
-        std::stringstream ss;
-        ss << "block" << blockid;
-
-        if(paramList.isSublist(ss.str()) == true) {
-          // we either have a parameter group or we have a list of factories in here
-          RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
-
-          RCP<FactoryManager> M = Teuchos::null;
-
-          if (b->isParameter("group")) {
-            // use a factory manager
-            std::string facManagerName = b->get< std::string >("group");
-            TEUCHOS_TEST_FOR_EXCEPTION(factoryManagersIn.count(facManagerName) != 1, Exceptions::RuntimeError, "Factory manager has not been found. Please check the spelling of the factory managers in your xml file.");
-            RCP<FactoryManagerBase> Mb = factoryManagersIn.find(facManagerName)->second;
-            M = Teuchos::rcp_dynamic_cast<FactoryManager>(Mb);
-            TEUCHOS_TEST_FOR_EXCEPTION(M==Teuchos::null, Exceptions::RuntimeError, "Failed to cast FactoryManagerBase object to FactoryManager.");
-          } else {
-            // read in the list of factories
-            M = rcp(new FactoryManager());
-            for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
-              RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
-              M->SetFactory(b->name(param),p);
-            }
-          }
+    RCP<SmootherFactory> smooFact = rcp(new SmootherFactory(Teuchos::null));
+    Teuchos::ParameterList smooFactParams;
+    // smooFactParams.setEntry("keep smoother data", paramList.getEntry("keep smoother data"));
+    smooFact->SetParameterList(smooFactParams);
+    smooFact->SetSmootherPrototypes(trilSmoo);
+    return smooFact;
+  }
 
-          // add factory manager to internal vector of factory managers
-          M->SetIgnoreUserData(true);
-          facManagers.push_back(M);
-          paramListNonConst->remove(ss.str());
-          blockid++;
+#ifdef HAVE_MUELU_MATLAB
+  //! MatlabSmoother
+  // Parameter List Parsing:
+  //     <ParameterList name="smootherFact1">
+  //       <Parameter name="factory" type="string" value="MatlabSmoother"/>
+  //       <Parameter name="Setup Function" type="string" value="mySmootherSetup.m"/>
+  //       <Parameter name="Solve Function" type="string" value="mySmootherSolve.m"/>
+  //       <!--A is implicitly included in this list and nothing else is needed to get diagonal-->
+  //       <Parameter name="Needs" type="string" value=""/>
+  //       <!--A,x,b are also assumed inputs to the solver: only one additional arg then (diag)-->
+  //       <Parameter name="Number of Solver Args"               type="int" value="1"/>
+  //     </ParameterList>
+  RCP<FactoryBase> BuildMatlabSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    if (paramList.begin() == paramList.end())
+      return rcp(new SmootherFactory(rcp(new MatlabSmoother())));
+
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "MatlabSmoother", Exceptions::RuntimeError, "");
+
+    // Read in factory information for smoothers (if available...)
+    // NOTE: only a selected number of factories can be used with the Trilinos smoother
+    //       smoothers usually work with the global data available (which is A and the transfers P and R)
+
+    Teuchos::RCP<MatlabSmoother> matSmoo = Teuchos::rcp(new MatlabSmoother(paramList));
+
+    return rcp(new SmootherFactory(matSmoo));
+  }
+#endif
+
+  RCP<FactoryBase> BuildDirectSolver(const Teuchos::ParameterList& paramList, const FactoryMap& /* factoryMapIn */, const FactoryManagerMap& /* factoryManagersIn */) const {
+    if (paramList.begin() == paramList.end())
+      return rcp(new SmootherFactory(rcp(new DirectSolver()), Teuchos::null));
+
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "DirectSolver", Exceptions::RuntimeError, "");
+
+    std::string type;
+    if (paramList.isParameter("type")) type = paramList.get<std::string>("type");
+    // std::string verbose;        if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
+    Teuchos::ParameterList params;
+    if (paramList.isParameter("ParameterList")) params = paramList.get<Teuchos::ParameterList>("ParameterList");
+
+    return rcp(new SmootherFactory(rcp(new DirectSolver(type, params)), Teuchos::null));
+  }
+
+  template <class T>  // T must implement the Factory interface
+  RCP<FactoryBase> BuildBlockedSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    // read in sub lists
+    RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
+
+    // internal vector of factory managers
+    std::vector<RCP<FactoryManager> > facManagers;
+
+    // loop over all "block%i" sublists in parameter list
+    int blockid      = 1;
+    bool blockExists = true;
+    while (blockExists == true) {
+      std::stringstream ss;
+      ss << "block" << blockid;
+
+      if (paramList.isSublist(ss.str()) == true) {
+        // we either have a parameter group or we have a list of factories in here
+        RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
+
+        RCP<FactoryManager> M = Teuchos::null;
+
+        if (b->isParameter("group")) {
+          // use a factory manager
+          std::string facManagerName = b->get<std::string>("group");
+          TEUCHOS_TEST_FOR_EXCEPTION(factoryManagersIn.count(facManagerName) != 1, Exceptions::RuntimeError, "Factory manager has not been found. Please check the spelling of the factory managers in your xml file.");
+          RCP<FactoryManagerBase> Mb = factoryManagersIn.find(facManagerName)->second;
+          M                          = Teuchos::rcp_dynamic_cast<FactoryManager>(Mb);
+          TEUCHOS_TEST_FOR_EXCEPTION(M == Teuchos::null, Exceptions::RuntimeError, "Failed to cast FactoryManagerBase object to FactoryManager.");
         } else {
-          blockExists = false;
-          break;
+          // read in the list of factories
+          M = rcp(new FactoryManager());
+          for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
+            RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
+            M->SetFactory(b->name(param), p);
+          }
         }
 
+        // add factory manager to internal vector of factory managers
+        M->SetIgnoreUserData(true);
+        facManagers.push_back(M);
+        paramListNonConst->remove(ss.str());
+        blockid++;
+      } else {
+        blockExists = false;
+        break;
       }
+    }
 
-      // create a new blocked smoother
-      RCP<T> bs = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
-
-      // important: set block factory for A here!
-      // TAW: 7/6/2016: We should not need to set/hardcode the blocked operator here.
-      //                The user might want to overwrite this in the xml file, so just
-      //                use what is declared as "A"
-      //bs->SetFactory("A", MueLu::NoFactory::getRCP());
+    // create a new blocked smoother
+    RCP<T> bs = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
 
-      for (int i = 0; i<Teuchos::as<int>(facManagers.size()); i++) {
-        bs->AddFactoryManager(facManagers[i],i);
-      }
+    // important: set block factory for A here!
+    // TAW: 7/6/2016: We should not need to set/hardcode the blocked operator here.
+    //                The user might want to overwrite this in the xml file, so just
+    //                use what is declared as "A"
+    // bs->SetFactory("A", MueLu::NoFactory::getRCP());
 
-      return rcp(new SmootherFactory(bs));
+    for (int i = 0; i < Teuchos::as<int>(facManagers.size()); i++) {
+      bs->AddFactoryManager(facManagers[i], i);
     }
 
+    return rcp(new SmootherFactory(bs));
+  }
+
 #ifdef HAVE_MUELU_TEKO
-    RCP<FactoryBase> BuildTekoSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      // read in sub lists
-      RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
-      RCP<ParameterList> tekoParams = rcp(new ParameterList(paramListNonConst->sublist("Inverse Factory Library")));
-      paramListNonConst->remove("Inverse Factory Library");
+  RCP<FactoryBase> BuildTekoSmoother(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    // read in sub lists
+    RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
+    RCP<ParameterList> tekoParams        = rcp(new ParameterList(paramListNonConst->sublist("Inverse Factory Library")));
+    paramListNonConst->remove("Inverse Factory Library");
+
+    // create a new blocked smoother
+    RCP<TekoSmoother> bs = Build2<TekoSmoother>(*paramListNonConst, factoryMapIn, factoryManagersIn);
+
+    // important: set block factory for A here!
+    // TAW: 7/6/2016: We should not need to set/hardcode the blocked operator here.
+    //                The user might want to overwrite this in the xml file, so just
+    //                use what is declared as "A"
+    // bs->SetFactory("A", MueLu::NoFactory::getRCP());
+
+    // Set Teko parameters ("Inverse Factory Library")
+    bs->SetTekoParameters(tekoParams);
+
+    return rcp(new SmootherFactory(bs));
+  }
+#endif
 
-      // create a new blocked smoother
-      RCP<TekoSmoother> bs = Build2<TekoSmoother>(*paramListNonConst, factoryMapIn, factoryManagersIn);
+  RCP<FactoryBase> BuildBlockedDirectSolver(const Teuchos::ParameterList& paramList, const FactoryMap& /* factoryMapIn */, const FactoryManagerMap& /* factoryManagersIn */) const {
+    if (paramList.numParams() == 0)
+      return rcp(new SmootherFactory(rcp(new BlockedDirectSolver())));
 
-      // important: set block factory for A here!
-      // TAW: 7/6/2016: We should not need to set/hardcode the blocked operator here.
-      //                The user might want to overwrite this in the xml file, so just
-      //                use what is declared as "A"
-      //bs->SetFactory("A", MueLu::NoFactory::getRCP());
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "BlockedDirectSolver", Exceptions::RuntimeError, "FactoryFactory::BuildBlockedDirectSolver: Generating factory needs to be a BlockedDirectSolver.");
 
-      // Set Teko parameters ("Inverse Factory Library")
-      bs->SetTekoParameters(tekoParams);
+    std::string type;
+    if (paramList.isParameter("type")) type = paramList.get<std::string>("type");
+    // std::string verbose;        if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
+    Teuchos::ParameterList params;
+    if (paramList.isParameter("ParameterList")) params = paramList.get<Teuchos::ParameterList>("ParameterList");
 
-      return rcp(new SmootherFactory(bs));
-    }
-#endif
+    return rcp(new SmootherFactory(rcp(new BlockedDirectSolver(type, params))));
+  }
 
-    RCP<FactoryBase> BuildBlockedDirectSolver(const Teuchos::ParameterList& paramList, const FactoryMap& /* factoryMapIn */, const FactoryManagerMap& /* factoryManagersIn */) const {
-      if (paramList.numParams() == 0)
-        return rcp(new SmootherFactory(rcp(new BlockedDirectSolver())));
+  // RCP<FactoryBase> BuildBlockedPFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+  //   RCP<BlockedPFactory> pfac = rcp(new BlockedPFactory());
 
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.get<std::string>("factory") != "BlockedDirectSolver", Exceptions::RuntimeError, "FactoryFactory::BuildBlockedDirectSolver: Generating factory needs to be a BlockedDirectSolver.");
+  template <class T>  // T must implement the Factory interface
+  RCP<T> BuildBlockedFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> pfac = Teuchos::null;
 
-      std::string type;              if(paramList.isParameter("type"))          type = paramList.get<std::string>("type");
-      // std::string verbose;        if(paramList.isParameter("verbose"))       verbose = paramList.get<std::string>("verbose");
-      Teuchos::ParameterList params; if(paramList.isParameter("ParameterList")) params  = paramList.get<Teuchos::ParameterList>("ParameterList");
+    // read in sub lists
+    RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
 
-      return rcp(new SmootherFactory(rcp(new BlockedDirectSolver(type, params))));
-    }
+    // internal vector of factory managers
+    std::vector<RCP<FactoryManager> > facManagers;
 
-    //RCP<FactoryBase> BuildBlockedPFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-    //  RCP<BlockedPFactory> pfac = rcp(new BlockedPFactory());
-
-    template <class T> // T must implement the Factory interface
-    RCP<T> BuildBlockedFactory(const Teuchos::ParameterList & paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> pfac = Teuchos::null;
-
-      // read in sub lists
-      RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
-
-      // internal vector of factory managers
-      std::vector<RCP<FactoryManager> > facManagers;
-
-      // loop over all "block%i" sublists in parameter list
-      int blockid = 1;
-      bool blockExists = true;
-      while (blockExists == true) {
-        std::stringstream ss;
-        ss << "block" << blockid;
-
-        if(paramList.isSublist(ss.str()) == true) {
-          // we either have a parameter group or we have a list of factories in here
-          RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
-
-          RCP<FactoryManager> M = Teuchos::null;
-
-          if (b->isParameter("group")) {
-            // use a factory manager
-            std::string facManagerName = b->get< std::string >("group");
-            TEUCHOS_TEST_FOR_EXCEPTION(factoryManagersIn.count(facManagerName) != 1, Exceptions::RuntimeError, "Factory manager has not been found. Please check the spelling of the factory managers in your xml file.");
-            RCP<FactoryManagerBase> Mb = factoryManagersIn.find(facManagerName)->second;
-            M = Teuchos::rcp_dynamic_cast<FactoryManager>(Mb);
-            TEUCHOS_TEST_FOR_EXCEPTION(M==Teuchos::null, Exceptions::RuntimeError, "Failed to cast FactoryManagerBase object to FactoryManager.");
-          } else {
-            // read in the list of factories
-            M = rcp(new FactoryManager());
-            for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
-              RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
-              M->SetFactory(b->name(param),p);
-            }
-          }
+    // loop over all "block%i" sublists in parameter list
+    int blockid      = 1;
+    bool blockExists = true;
+    while (blockExists == true) {
+      std::stringstream ss;
+      ss << "block" << blockid;
+
+      if (paramList.isSublist(ss.str()) == true) {
+        // we either have a parameter group or we have a list of factories in here
+        RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
 
-          // add factory manager to internal vector of factory managers
-          M->SetIgnoreUserData(true);
-          facManagers.push_back(M);
-          paramListNonConst->remove(ss.str());
-          blockid++;
+        RCP<FactoryManager> M = Teuchos::null;
+
+        if (b->isParameter("group")) {
+          // use a factory manager
+          std::string facManagerName = b->get<std::string>("group");
+          TEUCHOS_TEST_FOR_EXCEPTION(factoryManagersIn.count(facManagerName) != 1, Exceptions::RuntimeError, "Factory manager has not been found. Please check the spelling of the factory managers in your xml file.");
+          RCP<FactoryManagerBase> Mb = factoryManagersIn.find(facManagerName)->second;
+          M                          = Teuchos::rcp_dynamic_cast<FactoryManager>(Mb);
+          TEUCHOS_TEST_FOR_EXCEPTION(M == Teuchos::null, Exceptions::RuntimeError, "Failed to cast FactoryManagerBase object to FactoryManager.");
         } else {
-          blockExists = false;
-          break;
+          // read in the list of factories
+          M = rcp(new FactoryManager());
+          for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
+            RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
+            M->SetFactory(b->name(param), p);
+          }
         }
 
+        // add factory manager to internal vector of factory managers
+        M->SetIgnoreUserData(true);
+        facManagers.push_back(M);
+        paramListNonConst->remove(ss.str());
+        blockid++;
+      } else {
+        blockExists = false;
+        break;
       }
+    }
 
-      // build BlockedPFactory (without sub block information)
-      pfac = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
-
-      // add FactoryManager objects
-      for(size_t i = 0; i<facManagers.size(); i++) {
-        pfac->AddFactoryManager(facManagers[i]); // add factory manager
-      }
+    // build BlockedPFactory (without sub block information)
+    pfac = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
 
-      return pfac;
+    // add FactoryManager objects
+    for (size_t i = 0; i < facManagers.size(); i++) {
+      pfac->AddFactoryManager(facManagers[i]);  // add factory manager
     }
 
+    return pfac;
+  }
 
-    template <class T> // T must implement the Factory interface
-    RCP<T> BuildBlockedCoordFactory(const Teuchos::ParameterList & paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
-      RCP<T> pfac = Teuchos::null;
-
-      // read in sub lists
-      RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
+  template <class T>  // T must implement the Factory interface
+  RCP<T> BuildBlockedCoordFactory(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, const FactoryManagerMap& factoryManagersIn) const {
+    RCP<T> pfac = Teuchos::null;
 
-      // internal vector of factory managers
-      std::vector<RCP<const FactoryBase> > facBase;
+    // read in sub lists
+    RCP<ParameterList> paramListNonConst = rcp(new ParameterList(paramList));
 
-      // loop over all "block%i" sublists in parameter list
-      int blockid = 1;
-      bool blockExists = true;
-      while (blockExists == true) {
-        std::stringstream ss;
-        ss << "block" << blockid;
+    // internal vector of factory managers
+    std::vector<RCP<const FactoryBase> > facBase;
 
-        if(paramList.isSublist(ss.str()) == true) {
-          // we either have a parameter group or we have a list of factories in here
-          RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
+    // loop over all "block%i" sublists in parameter list
+    int blockid      = 1;
+    bool blockExists = true;
+    while (blockExists == true) {
+      std::stringstream ss;
+      ss << "block" << blockid;
 
-            // read in the list of factories
-            for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
-              RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
-              facBase.push_back(p);
-            }
+      if (paramList.isSublist(ss.str()) == true) {
+        // we either have a parameter group or we have a list of factories in here
+        RCP<const ParameterList> b = rcp(new ParameterList(*sublist(paramListNonConst, ss.str())));
 
-          // add factory manager to internal vector of factory managers
-          paramListNonConst->remove(ss.str());
-          blockid++;
-        } else {
-          blockExists = false;
-          break;
+        // read in the list of factories
+        for (ParameterList::ConstIterator param = b->begin(); param != b->end(); ++param) {
+          RCP<const FactoryBase> p = BuildFactory(b->entry(param), factoryMapIn, factoryManagersIn);
+          facBase.push_back(p);
         }
 
+        // add factory manager to internal vector of factory managers
+        paramListNonConst->remove(ss.str());
+        blockid++;
+      } else {
+        blockExists = false;
+        break;
       }
+    }
 
-      // build BlockedPFactory (without sub block information)
-      pfac = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
-
-      // add FactoryManager objects
-      for(size_t i = 0; i<facBase.size(); i++) {
-        pfac->AddFactory(facBase[i]); // add factory manager
-      }
+    // build BlockedPFactory (without sub block information)
+    pfac = Build2<T>(*paramListNonConst, factoryMapIn, factoryManagersIn);
 
-      return pfac;
+    // add FactoryManager objects
+    for (size_t i = 0; i < facBase.size(); i++) {
+      pfac->AddFactory(facBase[i]);  // add factory manager
     }
 
-  }; // class
-} // namespace MueLu
+    return pfac;
+  }
+
+};  // class
+}  // namespace MueLu
 
 #define MUELU_FACTORYFACTORY_SHORT
-#endif // MUELU_FACTORYFACTORY_DECL_HPP
+#endif  // MUELU_FACTORYFACTORY_DECL_HPP
 
-  // TODO: handle factory parameters
-  // TODO: parameter validator
-  // TODO: static
-  // TODO: default parameters should not be duplicated here and on the Factory (ex: default for overlap (=0) is defined both here and on TrilinosSmoother constructors)
+// TODO: handle factory parameters
+// TODO: parameter validator
+// TODO: static
+// TODO: default parameters should not be duplicated here and on the Factory (ex: default for overlap (=0) is defined both here and on TrilinosSmoother constructors)
diff --git a/packages/muelu/src/Interface/MueLu_HierarchyFactory.hpp b/packages/muelu/src/Interface/MueLu_HierarchyFactory.hpp
index 03e2596daf12..a82cad42cc47 100644
--- a/packages/muelu/src/Interface/MueLu_HierarchyFactory.hpp
+++ b/packages/muelu/src/Interface/MueLu_HierarchyFactory.hpp
@@ -55,42 +55,42 @@
 
 namespace MueLu {
 
-  //!
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class HierarchyFactory : public BaseClass {
+//!
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class HierarchyFactory : public BaseClass {
 #undef MUELU_HIERARCHYFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //@{ Constructors/Destructors.
+ public:
+  //@{ Constructors/Destructors.
 
-    //! Destructor.
-    virtual ~HierarchyFactory() { }
+  //! Destructor.
+  virtual ~HierarchyFactory() {}
 
-    //@}
+  //@}
 
-    //! Create an empty Hierarchy object
-    // Note: This function is not very useful at the moment as MueLu only have on Hierarchy class.
-    //       In the future, we might have an abstract Hierarchy class and several derived Hierarchy classes.
-    //       Using this function will then be the recommended way to generate a Hierarchy.
-    //
-    // This method is called Create() instead of Build(), because it return an non-initialized
-    // object (ie: MG setup is not done).
-    // Build() function in MueLu returns initialized objects.
-    virtual RCP<Hierarchy> CreateHierarchy() const = 0;
+  //! Create an empty Hierarchy object
+  // Note: This function is not very useful at the moment as MueLu only have on Hierarchy class.
+  //       In the future, we might have an abstract Hierarchy class and several derived Hierarchy classes.
+  //       Using this function will then be the recommended way to generate a Hierarchy.
+  //
+  // This method is called Create() instead of Build(), because it return an non-initialized
+  // object (ie: MG setup is not done).
+  // Build() function in MueLu returns initialized objects.
+  virtual RCP<Hierarchy> CreateHierarchy() const = 0;
 
-    //! Create a labeled empty Hierarchy object
-    virtual RCP<Hierarchy> CreateHierarchy(const std::string& label) const = 0;
+  //! Create a labeled empty Hierarchy object
+  virtual RCP<Hierarchy> CreateHierarchy(const std::string& label) const = 0;
 
-    //! Setup Hierarchy object
-    virtual void SetupHierarchy(Hierarchy & H) const = 0;
+  //! Setup Hierarchy object
+  virtual void SetupHierarchy(Hierarchy& H) const = 0;
 
-  }; // class HierarchyFactoryBase
+};  // class HierarchyFactoryBase
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_HIERARCHYFACTORY_SHORT
-#endif //ifndef MUELU_HIERARCHYFACTORY_HPP
+#endif  // ifndef MUELU_HIERARCHYFACTORY_HPP
diff --git a/packages/muelu/src/Interface/MueLu_HierarchyManager.hpp b/packages/muelu/src/Interface/MueLu_HierarchyManager.hpp
index 04f89f47e6ae..e38e8cc1c221 100644
--- a/packages/muelu/src/Interface/MueLu_HierarchyManager.hpp
+++ b/packages/muelu/src/Interface/MueLu_HierarchyManager.hpp
@@ -70,471 +70,456 @@
 
 namespace MueLu {
 
-  // This class stores the configuration of a Hierarchy.
-  // The class also provides an algorithm to build a Hierarchy from the configuration.
-  //
-  // See also: FactoryManager
-  //
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class HierarchyManager : public HierarchyFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+// This class stores the configuration of a Hierarchy.
+// The class also provides an algorithm to build a Hierarchy from the configuration.
+//
+// See also: FactoryManager
+//
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class HierarchyManager : public HierarchyFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_HIERARCHYMANAGER_SHORT
 #include "MueLu_UseShortNames.hpp"
-    typedef std::pair<std::string, const FactoryBase*> keep_pair;
-
-  public:
-
-    //! Constructor
-    HierarchyManager(int numDesiredLevel = MasterList::getDefault<int>("max levels")) :
-        numDesiredLevel_(numDesiredLevel),
-        maxCoarseSize_(MasterList::getDefault<int>("coarse: max size")),
-        verbosity_(Medium),
-        doPRrebalance_(MasterList::getDefault<bool>("repartition: rebalance P and R")),
-        doPRViaCopyrebalance_(MasterList::getDefault<bool>("repartition: explicit via new copy rebalance P and R")),
-        implicitTranspose_(MasterList::getDefault<bool>("transpose: use implicit")),
-        fuseProlongationAndUpdate_(MasterList::getDefault<bool>("fuse prolongation and update")),
-        suppressNullspaceDimensionCheck_(MasterList::getDefault<bool>("nullspace: suppress dimension check")),
-        sizeOfMultiVectors_(MasterList::getDefault<int>("number of vectors")),
-        graphOutputLevel_(-2) { }
-
-    //! Destructor
-    virtual ~HierarchyManager() = default;
-
-    //!
-    void AddFactoryManager(int startLevel, int numDesiredLevel, RCP<FactoryManagerBase> manager) {
-      const int lastLevel = startLevel + numDesiredLevel - 1;
-      if (levelManagers_.size() < lastLevel + 1)
-        levelManagers_.resize(lastLevel + 1);
-
-      for (int iLevel = startLevel; iLevel <= lastLevel; iLevel++)
-        levelManagers_[iLevel] = manager;
-    }
-
-    //!
-    RCP<FactoryManagerBase> GetFactoryManager(int levelID) const {
-      // NOTE: last levelManager is used for all the remaining levels
-      return (levelID >= levelManagers_.size() ? levelManagers_[levelManagers_.size()-1] : levelManagers_[levelID]);
-    }
-
-    //! returns number of factory managers stored in levelManagers_ vector.
-    size_t getNumFactoryManagers() const {
-      return levelManagers_.size();
-    }
-
-    //!
-    void CheckConfig() {
-      for (int i = 0; i < levelManagers_.size(); i++)
-        TEUCHOS_TEST_FOR_EXCEPTION(levelManagers_[i] == Teuchos::null, Exceptions::RuntimeError, "MueLu:HierarchyConfig::CheckConfig(): Undefined configuration for level:");
-    }
-
-    //@{
-
-    virtual RCP<Hierarchy> CreateHierarchy() const {
-      return rcp(new Hierarchy());
-    }
-
-    virtual RCP<Hierarchy> CreateHierarchy(const std::string& label) const {
-      return rcp(new Hierarchy(label));
-    }
-
-    //! Setup Hierarchy object
-    virtual void SetupHierarchy(Hierarchy& H) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(!H.GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError, "No fine level operator");
-
-      RCP<Level> l0 = H.GetLevel(0);
-      RCP<Operator> Op = l0->Get<RCP<Operator>>("A");
-
-      // Compare nullspace dimension to NumPDEs and throw/warn based on user input
-      if (l0->IsAvailable("Nullspace")) {
-        RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(Op);
-        if (A != Teuchos::null) {
-          RCP<MultiVector> nullspace = l0->Get<RCP<MultiVector>>("Nullspace");
-
-          if (static_cast<size_t>(A->GetFixedBlockSize()) > nullspace->getNumVectors())
-          {
-            std::stringstream msg;
-            msg << "User-provided nullspace has fewer vectors ("
-                << nullspace->getNumVectors() << ") than number of PDE equations ("
-                << A->GetFixedBlockSize() << "). ";
-
-            if (suppressNullspaceDimensionCheck_)
-            {
-              msg << "It depends on the PDE, if this is a problem or not.";
-              this->GetOStream(Warnings0) << msg.str() << std::endl;
-            }
-            else
-            {
-              msg << "Add the missing nullspace vectors! (You can suppress this check. See the MueLu user guide for details.)";
-              TEUCHOS_TEST_FOR_EXCEPTION(static_cast<size_t>(A->GetFixedBlockSize()) > nullspace->getNumVectors(), Exceptions::RuntimeError, msg.str());
-            }
+  typedef std::pair<std::string, const FactoryBase*> keep_pair;
+
+ public:
+  //! Constructor
+  HierarchyManager(int numDesiredLevel = MasterList::getDefault<int>("max levels"))
+    : numDesiredLevel_(numDesiredLevel)
+    , maxCoarseSize_(MasterList::getDefault<int>("coarse: max size"))
+    , verbosity_(Medium)
+    , doPRrebalance_(MasterList::getDefault<bool>("repartition: rebalance P and R"))
+    , doPRViaCopyrebalance_(MasterList::getDefault<bool>("repartition: explicit via new copy rebalance P and R"))
+    , implicitTranspose_(MasterList::getDefault<bool>("transpose: use implicit"))
+    , fuseProlongationAndUpdate_(MasterList::getDefault<bool>("fuse prolongation and update"))
+    , suppressNullspaceDimensionCheck_(MasterList::getDefault<bool>("nullspace: suppress dimension check"))
+    , sizeOfMultiVectors_(MasterList::getDefault<int>("number of vectors"))
+    , graphOutputLevel_(-2) {}
+
+  //! Destructor
+  virtual ~HierarchyManager() = default;
+
+  //!
+  void AddFactoryManager(int startLevel, int numDesiredLevel, RCP<FactoryManagerBase> manager) {
+    const int lastLevel = startLevel + numDesiredLevel - 1;
+    if (levelManagers_.size() < lastLevel + 1)
+      levelManagers_.resize(lastLevel + 1);
+
+    for (int iLevel = startLevel; iLevel <= lastLevel; iLevel++)
+      levelManagers_[iLevel] = manager;
+  }
+
+  //!
+  RCP<FactoryManagerBase> GetFactoryManager(int levelID) const {
+    // NOTE: last levelManager is used for all the remaining levels
+    return (levelID >= levelManagers_.size() ? levelManagers_[levelManagers_.size() - 1] : levelManagers_[levelID]);
+  }
+
+  //! returns number of factory managers stored in levelManagers_ vector.
+  size_t getNumFactoryManagers() const {
+    return levelManagers_.size();
+  }
+
+  //!
+  void CheckConfig() {
+    for (int i = 0; i < levelManagers_.size(); i++)
+      TEUCHOS_TEST_FOR_EXCEPTION(levelManagers_[i] == Teuchos::null, Exceptions::RuntimeError, "MueLu:HierarchyConfig::CheckConfig(): Undefined configuration for level:");
+  }
+
+  //@{
+
+  virtual RCP<Hierarchy> CreateHierarchy() const {
+    return rcp(new Hierarchy());
+  }
+
+  virtual RCP<Hierarchy> CreateHierarchy(const std::string& label) const {
+    return rcp(new Hierarchy(label));
+  }
+
+  //! Setup Hierarchy object
+  virtual void SetupHierarchy(Hierarchy& H) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(!H.GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError, "No fine level operator");
+
+    RCP<Level> l0    = H.GetLevel(0);
+    RCP<Operator> Op = l0->Get<RCP<Operator>>("A");
+
+    // Compare nullspace dimension to NumPDEs and throw/warn based on user input
+    if (l0->IsAvailable("Nullspace")) {
+      RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(Op);
+      if (A != Teuchos::null) {
+        RCP<MultiVector> nullspace = l0->Get<RCP<MultiVector>>("Nullspace");
+
+        if (static_cast<size_t>(A->GetFixedBlockSize()) > nullspace->getNumVectors()) {
+          std::stringstream msg;
+          msg << "User-provided nullspace has fewer vectors ("
+              << nullspace->getNumVectors() << ") than number of PDE equations ("
+              << A->GetFixedBlockSize() << "). ";
+
+          if (suppressNullspaceDimensionCheck_) {
+            msg << "It depends on the PDE, if this is a problem or not.";
+            this->GetOStream(Warnings0) << msg.str() << std::endl;
+          } else {
+            msg << "Add the missing nullspace vectors! (You can suppress this check. See the MueLu user guide for details.)";
+            TEUCHOS_TEST_FOR_EXCEPTION(static_cast<size_t>(A->GetFixedBlockSize()) > nullspace->getNumVectors(), Exceptions::RuntimeError, msg.str());
           }
-        } else {
-          this->GetOStream(Warnings0) << "Skipping dimension check of user-supplied nullspace because user-supplied operator is not a matrix" << std::endl;
         }
+      } else {
+        this->GetOStream(Warnings0) << "Skipping dimension check of user-supplied nullspace because user-supplied operator is not a matrix" << std::endl;
       }
+    }
 
 #ifdef HAVE_MUELU_DEBUG
-      // Reset factories' data used for debugging
-      for (int i = 0; i < levelManagers_.size(); i++)
-        levelManagers_[i]->ResetDebugData();
+    // Reset factories' data used for debugging
+    for (int i = 0; i < levelManagers_.size(); i++)
+      levelManagers_[i]->ResetDebugData();
 
 #endif
 
-      // Setup Matrix
-      // TODO: I should certainly undo this somewhere...
+    // Setup Matrix
+    // TODO: I should certainly undo this somewhere...
 
-      Xpetra::UnderlyingLib lib = Op->getDomainMap()->lib();
-      H.setlib(lib);
+    Xpetra::UnderlyingLib lib = Op->getDomainMap()->lib();
+    H.setlib(lib);
 
-      SetupOperator(*Op);
-      SetupExtra(H);
+    SetupOperator(*Op);
+    SetupExtra(H);
 
-      // Setup Hierarchy
-      H.SetMaxCoarseSize(maxCoarseSize_);
-      VerboseObject::SetDefaultVerbLevel(verbosity_);
-      if (graphOutputLevel_ >= 0 || graphOutputLevel_ == -1)
-        H.EnableGraphDumping("dep_graph", graphOutputLevel_);
+    // Setup Hierarchy
+    H.SetMaxCoarseSize(maxCoarseSize_);
+    VerboseObject::SetDefaultVerbLevel(verbosity_);
+    if (graphOutputLevel_ >= 0 || graphOutputLevel_ == -1)
+      H.EnableGraphDumping("dep_graph", graphOutputLevel_);
 
-      if (VerboseObject::IsPrint(Statistics2)) {
-        RCP<Matrix> Amat = rcp_dynamic_cast<Matrix>(Op);
+    if (VerboseObject::IsPrint(Statistics2)) {
+      RCP<Matrix> Amat = rcp_dynamic_cast<Matrix>(Op);
 
-        if (!Amat.is_null()) {
-            RCP<ParameterList> params = rcp(new ParameterList());
-            params->set("printLoadBalancingInfo", true);
-            params->set("printCommInfo",          true);
+      if (!Amat.is_null()) {
+        RCP<ParameterList> params = rcp(new ParameterList());
+        params->set("printLoadBalancingInfo", true);
+        params->set("printCommInfo", true);
 
-            VerboseObject::GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Amat, "A0", params);
-        } else {
-            VerboseObject::GetOStream(Warnings1) << "Fine level operator is not a matrix, statistics are not available" << std::endl;
-        }
+        VerboseObject::GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Amat, "A0", params);
+      } else {
+        VerboseObject::GetOStream(Warnings1) << "Fine level operator is not a matrix, statistics are not available" << std::endl;
       }
+    }
 
-      H.SetPRrebalance(doPRrebalance_);
-      H.SetPRViaCopyrebalance(doPRViaCopyrebalance_);
-      H.SetImplicitTranspose(implicitTranspose_);
-      H.SetFuseProlongationAndUpdate(fuseProlongationAndUpdate_);
-
-      H.Clear();
-
-      // There are few issues with using Keep in the interpreter:
-      //   1. Hierarchy::Keep interface takes a name and a factory. If
-      //      factories are different on different levels, the AddNewLevel() call
-      //      in Hierarchy does not work properly, as it assume that factories are
-      //      the same.
-      //   2. FactoryManager does not have a Keep option, only Hierarchy and
-      //      Level have it
-      //   3. Interpreter constructs factory managers, but not levels. So we
-      //      cannot set up Keep flags there.
-      //
-      // The solution implemented here does the following:
-      //   1. Construct hierarchy with dummy levels. This avoids
-      //      Hierarchy::AddNewLevel() calls which will propagate wrong
-      //      inheritance.
-      //   2. Interpreter constructs keep_ array with names and factories for
-      //      that level
-      //   3. For each level, we call Keep(name, factory) for each keep_
-      for (int i = 0; i < numDesiredLevel_; i++) {
-        std::map<int, std::vector<keep_pair> >::const_iterator it = keep_.find(i);
-        if (it != keep_.end()) {
-          RCP<Level> l = H.GetLevel(i);
-          const std::vector<keep_pair>& keeps = it->second;
-          for (size_t j = 0; j < keeps.size(); j++)
-            l->Keep(keeps[j].first, keeps[j].second);
-        }
-        if (i < numDesiredLevel_-1) {
-          RCP<Level> newLevel = rcp(new Level());
-          H.AddLevel(newLevel);
-        }
+    H.SetPRrebalance(doPRrebalance_);
+    H.SetPRViaCopyrebalance(doPRViaCopyrebalance_);
+    H.SetImplicitTranspose(implicitTranspose_);
+    H.SetFuseProlongationAndUpdate(fuseProlongationAndUpdate_);
+
+    H.Clear();
+
+    // There are few issues with using Keep in the interpreter:
+    //   1. Hierarchy::Keep interface takes a name and a factory. If
+    //      factories are different on different levels, the AddNewLevel() call
+    //      in Hierarchy does not work properly, as it assume that factories are
+    //      the same.
+    //   2. FactoryManager does not have a Keep option, only Hierarchy and
+    //      Level have it
+    //   3. Interpreter constructs factory managers, but not levels. So we
+    //      cannot set up Keep flags there.
+    //
+    // The solution implemented here does the following:
+    //   1. Construct hierarchy with dummy levels. This avoids
+    //      Hierarchy::AddNewLevel() calls which will propagate wrong
+    //      inheritance.
+    //   2. Interpreter constructs keep_ array with names and factories for
+    //      that level
+    //   3. For each level, we call Keep(name, factory) for each keep_
+    for (int i = 0; i < numDesiredLevel_; i++) {
+      std::map<int, std::vector<keep_pair>>::const_iterator it = keep_.find(i);
+      if (it != keep_.end()) {
+        RCP<Level> l                        = H.GetLevel(i);
+        const std::vector<keep_pair>& keeps = it->second;
+        for (size_t j = 0; j < keeps.size(); j++)
+          l->Keep(keeps[j].first, keeps[j].second);
       }
+      if (i < numDesiredLevel_ - 1) {
+        RCP<Level> newLevel = rcp(new Level());
+        H.AddLevel(newLevel);
+      }
+    }
 
-      // Matrices to print
-      for(auto iter=matricesToPrint_.begin(); iter!=matricesToPrint_.end(); iter++)
-        ExportDataSetKeepFlags(H,iter->second,iter->first);
+    // Matrices to print
+    for (auto iter = matricesToPrint_.begin(); iter != matricesToPrint_.end(); iter++)
+      ExportDataSetKeepFlags(H, iter->second, iter->first);
 
-      // Vectors, aggregates and other things that need special case handling
-      ExportDataSetKeepFlags(H, nullspaceToPrint_,  "Nullspace");
-      ExportDataSetKeepFlags(H, coordinatesToPrint_,  "Coordinates");
-      // NOTE: Aggregates use the next level's Factory
-      ExportDataSetKeepFlagsNextLevel(H, aggregatesToPrint_,  "Aggregates");
+    // Vectors, aggregates and other things that need special case handling
+    ExportDataSetKeepFlags(H, nullspaceToPrint_, "Nullspace");
+    ExportDataSetKeepFlags(H, coordinatesToPrint_, "Coordinates");
+    // NOTE: Aggregates use the next level's Factory
+    ExportDataSetKeepFlagsNextLevel(H, aggregatesToPrint_, "Aggregates");
 #ifdef HAVE_MUELU_INTREPID2
-      ExportDataSetKeepFlags(H,elementToNodeMapsToPrint_, "pcoarsen: element to node map");
+    ExportDataSetKeepFlags(H, elementToNodeMapsToPrint_, "pcoarsen: element to node map");
 #endif
 
-      // Data to save only (these do not have a level, so we do all levels)
-      for(int i=0; i<dataToSave_.size(); i++)
-        ExportDataSetKeepFlagsAll(H,dataToSave_[i]);
-
-      int  levelID     = 0;
-      int  lastLevelID = numDesiredLevel_ - 1;
-      bool isLastLevel = false;
+    // Data to save only (these do not have a level, so we do all levels)
+    for (int i = 0; i < dataToSave_.size(); i++)
+      ExportDataSetKeepFlagsAll(H, dataToSave_[i]);
 
-      while (!isLastLevel) {
-        bool r = H.Setup(levelID,
-                         LvlMngr(levelID-1, lastLevelID),
-                         LvlMngr(levelID,   lastLevelID),
-                         LvlMngr(levelID+1, lastLevelID));
-        if (levelID < H.GetNumLevels())
-          H.GetLevel(levelID)->print(H.GetOStream(Developer), verbosity_);
-
-        isLastLevel = r || (levelID == lastLevelID);
-        levelID++;
-      }
-      if (!matvecParams_.is_null())
-        H.SetMatvecParams(matvecParams_);
-      H.AllocateLevelMultiVectors(sizeOfMultiVectors_);
-      // Set hierarchy description.
-      // This is cached, but involves and MPI_Allreduce.
-      H.description();
-      H.describe(H.GetOStream(Runtime0), verbosity_);
-
-      // When we reuse hierarchy, it is necessary that we don't
-      // change the number of levels. We also cannot make requests
-      // for coarser levels, because we don't construct all the
-      // data on previous levels. For instance, let's say our first
-      // run constructed three levels. If we try to do requests during
-      // next setup for the fourth level, it would need Aggregates
-      // which we didn't construct for level 3 because we reused P.
-      // To fix this situation, we change the number of desired levels
-      // here.
-      numDesiredLevel_ = levelID;
-
-      // Matrix prints
-      for(auto iter = matricesToPrint_.begin(); iter != matricesToPrint_.end(); iter++) {
-        WriteData<Matrix>(H,iter->second,iter->first);
-      }
+    int levelID      = 0;
+    int lastLevelID  = numDesiredLevel_ - 1;
+    bool isLastLevel = false;
 
-      // Vectors, aggregates and all things we need to print manually
-      WriteData<MultiVector>(H, nullspaceToPrint_,  "Nullspace");
-      WriteData<MultiVector>(H, coordinatesToPrint_,  "Coordinates");
-      WriteDataAggregates(H, aggregatesToPrint_,  "Aggregates");
+    while (!isLastLevel) {
+      bool r = H.Setup(levelID,
+                       LvlMngr(levelID - 1, lastLevelID),
+                       LvlMngr(levelID, lastLevelID),
+                       LvlMngr(levelID + 1, lastLevelID));
+      if (levelID < H.GetNumLevels())
+        H.GetLevel(levelID)->print(H.GetOStream(Developer), verbosity_);
 
+      isLastLevel = r || (levelID == lastLevelID);
+      levelID++;
+    }
+    if (!matvecParams_.is_null())
+      H.SetMatvecParams(matvecParams_);
+    H.AllocateLevelMultiVectors(sizeOfMultiVectors_);
+    // Set hierarchy description.
+    // This is cached, but involves and MPI_Allreduce.
+    H.description();
+    H.describe(H.GetOStream(Runtime0), verbosity_);
+
+    // When we reuse hierarchy, it is necessary that we don't
+    // change the number of levels. We also cannot make requests
+    // for coarser levels, because we don't construct all the
+    // data on previous levels. For instance, let's say our first
+    // run constructed three levels. If we try to do requests during
+    // next setup for the fourth level, it would need Aggregates
+    // which we didn't construct for level 3 because we reused P.
+    // To fix this situation, we change the number of desired levels
+    // here.
+    numDesiredLevel_ = levelID;
+
+    // Matrix prints
+    for (auto iter = matricesToPrint_.begin(); iter != matricesToPrint_.end(); iter++) {
+      WriteData<Matrix>(H, iter->second, iter->first);
+    }
 
+    // Vectors, aggregates and all things we need to print manually
+    WriteData<MultiVector>(H, nullspaceToPrint_, "Nullspace");
+    WriteData<MultiVector>(H, coordinatesToPrint_, "Coordinates");
+    WriteDataAggregates(H, aggregatesToPrint_, "Aggregates");
 
 #ifdef HAVE_MUELU_INTREPID2
-      typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-      WriteDataFC<FCi>(H,elementToNodeMapsToPrint_, "pcoarsen: element to node map","el2node");
+    typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+    WriteDataFC<FCi>(H, elementToNodeMapsToPrint_, "pcoarsen: element to node map", "el2node");
 #endif
 
+  }  // SetupHierarchy
 
-    } //SetupHierarchy
-
-    //@}
-
-    typedef std::map<std::string, RCP<const FactoryBase> > FactoryMap;
+  //@}
 
-  protected: //TODO: access function
+  typedef std::map<std::string, RCP<const FactoryBase>> FactoryMap;
 
-    //! Setup Matrix object
-    virtual void SetupOperator(Operator& /* Op */) const { }
+ protected:  // TODO: access function
+  //! Setup Matrix object
+  virtual void SetupOperator(Operator& /* Op */) const {}
 
-    //! Setup extra data
-    // TODO: merge with SetupMatrix ?
-    virtual void SetupExtra(Hierarchy& /* H */) const { }
+  //! Setup extra data
+  // TODO: merge with SetupMatrix ?
+  virtual void SetupExtra(Hierarchy& /* H */) const {}
 
-    // TODO this was private
-    // Used in SetupHierarchy() to access levelManagers_
-    // Inputs i=-1 and i=size() are allowed to simplify calls to hierarchy->Setup()
-    Teuchos::RCP<FactoryManagerBase> LvlMngr(int levelID, int lastLevelID) const {
-      // NOTE: the order of 'if' statements is important
-      if (levelID == -1)                    // levelID = -1 corresponds to the finest level
-        return Teuchos::null;
+  // TODO this was private
+  // Used in SetupHierarchy() to access levelManagers_
+  // Inputs i=-1 and i=size() are allowed to simplify calls to hierarchy->Setup()
+  Teuchos::RCP<FactoryManagerBase> LvlMngr(int levelID, int lastLevelID) const {
+    // NOTE: the order of 'if' statements is important
+    if (levelID == -1)  // levelID = -1 corresponds to the finest level
+      return Teuchos::null;
 
-      if (levelID == lastLevelID+1)         // levelID = 'lastLevelID+1' corresponds to the last level (i.e., no nextLevel)
-        return Teuchos::null;
+    if (levelID == lastLevelID + 1)  // levelID = 'lastLevelID+1' corresponds to the last level (i.e., no nextLevel)
+      return Teuchos::null;
 
-      if (levelManagers_.size() == 0) {     // default factory manager.
-        // The default manager is shared across levels, initialized only if needed and deleted with the HierarchyManager
-        static RCP<FactoryManagerBase> defaultMngr = rcp(new FactoryManager());
-        return defaultMngr;
-      }
-
-      return GetFactoryManager(levelID);
+    if (levelManagers_.size() == 0) {  // default factory manager.
+      // The default manager is shared across levels, initialized only if needed and deleted with the HierarchyManager
+      static RCP<FactoryManagerBase> defaultMngr = rcp(new FactoryManager());
+      return defaultMngr;
     }
 
-    //! @group Hierarchy parameters
-    //! @{
+    return GetFactoryManager(levelID);
+  }
 
-    mutable int numDesiredLevel_;
-    Xpetra::global_size_t maxCoarseSize_;
-    MsgType verbosity_;
+  //! @group Hierarchy parameters
+  //! @{
 
-    bool doPRrebalance_;
-    bool doPRViaCopyrebalance_;
-    bool implicitTranspose_;
-    bool fuseProlongationAndUpdate_;
+  mutable int numDesiredLevel_;
+  Xpetra::global_size_t maxCoarseSize_;
+  MsgType verbosity_;
 
-    /*! @brief Flag to indicate whether the check of the nullspace dimension is suppressed
+  bool doPRrebalance_;
+  bool doPRViaCopyrebalance_;
+  bool implicitTranspose_;
+  bool fuseProlongationAndUpdate_;
 
-    By default, we do not suppress such a check, as it acts as a safety mechanism.
-    Yet, certain scenarios deliberately use nullspaces with less nullspace vectors than NumPDEs.
-    Therefore, the user can suppress this check. Then, the error message is converted to a warning.
-    */
-    bool suppressNullspaceDimensionCheck_;
+  /*! @brief Flag to indicate whether the check of the nullspace dimension is suppressed
 
-    int sizeOfMultiVectors_;
+  By default, we do not suppress such a check, as it acts as a safety mechanism.
+  Yet, certain scenarios deliberately use nullspaces with less nullspace vectors than NumPDEs.
+  Therefore, the user can suppress this check. Then, the error message is converted to a warning.
+  */
+  bool suppressNullspaceDimensionCheck_;
 
-    //! -2 = no output, -1 = all levels
-    int graphOutputLevel_;
+  int sizeOfMultiVectors_;
 
-    //! Lists of entities to be exported (or saved)
-    // Items here get handled manually
-    Teuchos::Array<int> nullspaceToPrint_;
-    Teuchos::Array<int> coordinatesToPrint_;
-    Teuchos::Array<int> aggregatesToPrint_;
-    Teuchos::Array<int> elementToNodeMapsToPrint_;
+  //! -2 = no output, -1 = all levels
+  int graphOutputLevel_;
 
-    // Data we'll need to save, not necessarily print
-    Teuchos::Array<std::string> dataToSave_;
+  //! Lists of entities to be exported (or saved)
+  // Items here get handled manually
+  Teuchos::Array<int> nullspaceToPrint_;
+  Teuchos::Array<int> coordinatesToPrint_;
+  Teuchos::Array<int> aggregatesToPrint_;
+  Teuchos::Array<int> elementToNodeMapsToPrint_;
 
-    // Matrices we'll need to print
-    std::map<std::string,Teuchos::Array<int> > matricesToPrint_;
+  // Data we'll need to save, not necessarily print
+  Teuchos::Array<std::string> dataToSave_;
 
-    Teuchos::RCP<Teuchos::ParameterList> matvecParams_;
+  // Matrices we'll need to print
+  std::map<std::string, Teuchos::Array<int>> matricesToPrint_;
 
-    std::map<int, std::vector<keep_pair> > keep_;
-    //! @}
+  Teuchos::RCP<Teuchos::ParameterList> matvecParams_;
 
-  private:
-    // Set the keep flags for Export Data
-    void ExportDataSetKeepFlags(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
-      for (int i = 0; i < data.size(); ++i) {
-        if (data[i] < H.GetNumLevels()) {
-          RCP<Level> L = H.GetLevel(data[i]);
-      	  if(!L.is_null()  && data[i] < levelManagers_.size())
-      	    L->AddKeepFlag(name, &*levelManagers_[data[i]]->GetFactory(name));
-          }
-      }
-    }
+  std::map<int, std::vector<keep_pair>> keep_;
+  //! @}
 
-    void ExportDataSetKeepFlagsNextLevel(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
-      for (int i = 0; i < data.size(); ++i) {
-        if (data[i] < H.GetNumLevels()) {
-          RCP<Level> L = H.GetLevel(data[i]);
-      	  if(!L.is_null()  && data[i]+1 < levelManagers_.size())
-      	    L->AddKeepFlag(name, &*levelManagers_[data[i]+1]->GetFactory(name));
-        }
+ private:
+  // Set the keep flags for Export Data
+  void ExportDataSetKeepFlags(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
+    for (int i = 0; i < data.size(); ++i) {
+      if (data[i] < H.GetNumLevels()) {
+        RCP<Level> L = H.GetLevel(data[i]);
+        if (!L.is_null() && data[i] < levelManagers_.size())
+          L->AddKeepFlag(name, &*levelManagers_[data[i]]->GetFactory(name));
       }
     }
-
-     // Set the keep flags for Export Data
-    void ExportDataSetKeepFlagsAll(Hierarchy& H, const std::string& name) const {
-      for (int i=0; i < H.GetNumLevels(); i++ ) {
-          RCP<Level> L = H.GetLevel(i);
-      	  if(!L.is_null() && i < levelManagers_.size())
-      	    L->AddKeepFlag(name, &*levelManagers_[i]->GetFactory(name));
+  }
+
+  void ExportDataSetKeepFlagsNextLevel(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
+    for (int i = 0; i < data.size(); ++i) {
+      if (data[i] < H.GetNumLevels()) {
+        RCP<Level> L = H.GetLevel(data[i]);
+        if (!L.is_null() && data[i] + 1 < levelManagers_.size())
+          L->AddKeepFlag(name, &*levelManagers_[data[i] + 1]->GetFactory(name));
       }
     }
-
-
-    template<class T>
-    void WriteData(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
-      for (int i = 0; i < data.size(); ++i) {
-        std::string fileName;
-        if (H.getObjectLabel() != "")
-          fileName = H.getObjectLabel() + "_" + name + "_" + Teuchos::toString(data[i]) + ".m";
-        else
-          fileName = name + "_" + Teuchos::toString(data[i]) + ".m";
-
-        if (data[i] < H.GetNumLevels()) {
-          RCP<Level> L = H.GetLevel(data[i]);
-          if (data[i] < levelManagers_.size() && L->IsAvailable(name,&*levelManagers_[data[i]]->GetFactory(name))) {
-	    // Try generating factory
-            RCP<T> M = L->template Get< RCP<T> >(name,&*levelManagers_[data[i]]->GetFactory(name));
-            if (!M.is_null()) {
-              Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(fileName,* M);
-            }
+  }
+
+  // Set the keep flags for Export Data
+  void ExportDataSetKeepFlagsAll(Hierarchy& H, const std::string& name) const {
+    for (int i = 0; i < H.GetNumLevels(); i++) {
+      RCP<Level> L = H.GetLevel(i);
+      if (!L.is_null() && i < levelManagers_.size())
+        L->AddKeepFlag(name, &*levelManagers_[i]->GetFactory(name));
+    }
+  }
+
+  template <class T>
+  void WriteData(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
+    for (int i = 0; i < data.size(); ++i) {
+      std::string fileName;
+      if (H.getObjectLabel() != "")
+        fileName = H.getObjectLabel() + "_" + name + "_" + Teuchos::toString(data[i]) + ".m";
+      else
+        fileName = name + "_" + Teuchos::toString(data[i]) + ".m";
+
+      if (data[i] < H.GetNumLevels()) {
+        RCP<Level> L = H.GetLevel(data[i]);
+        if (data[i] < levelManagers_.size() && L->IsAvailable(name, &*levelManagers_[data[i]]->GetFactory(name))) {
+          // Try generating factory
+          RCP<T> M = L->template Get<RCP<T>>(name, &*levelManagers_[data[i]]->GetFactory(name));
+          if (!M.is_null()) {
+            Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(fileName, *M);
           }
-      	  else if (L->IsAvailable(name)) {
-      	    // Try nofactory
-            RCP<T> M = L->template Get< RCP<T> >(name);
-            if (!M.is_null()) {
-              Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(fileName,* M);
-            }
+        } else if (L->IsAvailable(name)) {
+          // Try nofactory
+          RCP<T> M = L->template Get<RCP<T>>(name);
+          if (!M.is_null()) {
+            Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(fileName, *M);
           }
         }
       }
     }
-
-    void WriteDataAggregates(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
-      for (int i = 0; i < data.size(); ++i) {
-        const std::string fileName = name + "_" + Teuchos::toString(data[i]) + ".m";
-
-        if (data[i] < H.GetNumLevels()) {
-          RCP<Level> L = H.GetLevel(data[i]);
-
-          // NOTE: Aggregates use the next level's factory
-          RCP<Aggregates> agg;
-          if(data[i]+1 < H.GetNumLevels() && L->IsAvailable(name,&*levelManagers_[data[i]+1]->GetFactory(name))) {
-            // Try generating factory
-            agg = L->template Get< RCP<Aggregates> >(name,&*levelManagers_[data[i]+1]->GetFactory(name));
-          }
-          else if (L->IsAvailable(name)) {
-            agg = L->template Get<RCP<Aggregates> >("Aggregates");
-          }
-          if(!agg.is_null()) {
-            std::ofstream ofs(fileName);
-            Teuchos::FancyOStream fofs(rcp(&ofs,false));
-            agg->print(fofs,Teuchos::VERB_EXTREME);
-          }
+  }
+
+  void WriteDataAggregates(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name) const {
+    for (int i = 0; i < data.size(); ++i) {
+      const std::string fileName = name + "_" + Teuchos::toString(data[i]) + ".m";
+
+      if (data[i] < H.GetNumLevels()) {
+        RCP<Level> L = H.GetLevel(data[i]);
+
+        // NOTE: Aggregates use the next level's factory
+        RCP<Aggregates> agg;
+        if (data[i] + 1 < H.GetNumLevels() && L->IsAvailable(name, &*levelManagers_[data[i] + 1]->GetFactory(name))) {
+          // Try generating factory
+          agg = L->template Get<RCP<Aggregates>>(name, &*levelManagers_[data[i] + 1]->GetFactory(name));
+        } else if (L->IsAvailable(name)) {
+          agg = L->template Get<RCP<Aggregates>>("Aggregates");
+        }
+        if (!agg.is_null()) {
+          std::ofstream ofs(fileName);
+          Teuchos::FancyOStream fofs(rcp(&ofs, false));
+          agg->print(fofs, Teuchos::VERB_EXTREME);
         }
       }
     }
-
-    template<class T>
-    void WriteDataFC(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name, const std::string & ofname) const {
-      for (int i = 0; i < data.size(); ++i) {
-        const std::string fileName = ofname + "_" + Teuchos::toString(data[i]) + ".m";
-
-        if (data[i] < H.GetNumLevels()) {
-          RCP<Level> L = H.GetLevel(data[i]);
-
-          if (L->IsAvailable(name)) {
-            RCP<T> M = L->template Get< RCP<T> >(name);
-            if (!M.is_null()) {
-              RCP<Matrix> A = L->template Get<RCP<Matrix> >("A");
-              RCP<const CrsGraph> AG = A->getCrsGraph();
-              WriteFieldContainer<T>(fileName,*M,*AG->getColMap());
-            }
+  }
+
+  template <class T>
+  void WriteDataFC(Hierarchy& H, const Teuchos::Array<int>& data, const std::string& name, const std::string& ofname) const {
+    for (int i = 0; i < data.size(); ++i) {
+      const std::string fileName = ofname + "_" + Teuchos::toString(data[i]) + ".m";
+
+      if (data[i] < H.GetNumLevels()) {
+        RCP<Level> L = H.GetLevel(data[i]);
+
+        if (L->IsAvailable(name)) {
+          RCP<T> M = L->template Get<RCP<T>>(name);
+          if (!M.is_null()) {
+            RCP<Matrix> A          = L->template Get<RCP<Matrix>>("A");
+            RCP<const CrsGraph> AG = A->getCrsGraph();
+            WriteFieldContainer<T>(fileName, *M, *AG->getColMap());
           }
         }
       }
     }
+  }
 
-    // For dumping an IntrepidPCoarsening element-to-node map to disk
-    template<class T>
-    void WriteFieldContainer(const std::string& fileName, T & fcont,const Map &colMap) const {
-
-      size_t num_els = (size_t) fcont.extent(0);
-      size_t num_vecs =(size_t) fcont.extent(1);
+  // For dumping an IntrepidPCoarsening element-to-node map to disk
+  template <class T>
+  void WriteFieldContainer(const std::string& fileName, T& fcont, const Map& colMap) const {
+    size_t num_els  = (size_t)fcont.extent(0);
+    size_t num_vecs = (size_t)fcont.extent(1);
 
-      // Generate rowMap
-      Teuchos::RCP<const Map> rowMap = Xpetra::MapFactory<LO,GO,NO>::Build(colMap.lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),fcont.extent(0),colMap.getIndexBase(),colMap.getComm());
+    // Generate rowMap
+    Teuchos::RCP<const Map> rowMap = Xpetra::MapFactory<LO, GO, NO>::Build(colMap.lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), fcont.extent(0), colMap.getIndexBase(), colMap.getComm());
 
-      // Fill multivector to use *petra dump routines
-      RCP<GOMultiVector> vec = Xpetra::MultiVectorFactory<GO, LO, GO, NO>::Build(rowMap,num_vecs);
+    // Fill multivector to use *petra dump routines
+    RCP<GOMultiVector> vec = Xpetra::MultiVectorFactory<GO, LO, GO, NO>::Build(rowMap, num_vecs);
 
-      for(size_t j=0; j<num_vecs; j++)  {
-        Teuchos::ArrayRCP<GO> v = vec->getDataNonConst(j);
-        for(size_t i=0; i<num_els; i++)
-          v[i] = colMap.getGlobalElement(fcont(i,j));
-      }
-
-      Xpetra::IO<GO,LO,GO,NO>::Write(fileName,*vec);
+    for (size_t j = 0; j < num_vecs; j++) {
+      Teuchos::ArrayRCP<GO> v = vec->getDataNonConst(j);
+      for (size_t i = 0; i < num_els; i++)
+        v[i] = colMap.getGlobalElement(fcont(i, j));
     }
 
+    Xpetra::IO<GO, LO, GO, NO>::Write(fileName, *vec);
+  }
 
+  // Levels
+  Array<RCP<FactoryManagerBase>> levelManagers_;  // one FactoryManager per level (the last levelManager is used for all the remaining levels)
 
-    // Levels
-    Array<RCP<FactoryManagerBase> > levelManagers_;        // one FactoryManager per level (the last levelManager is used for all the remaining levels)
-
-  }; // class HierarchyManager
+};  // class HierarchyManager
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_HIERARCHYMANAGER_SHORT
-#endif // MUELU_HIERARCHYMANAGER_HPP
+#endif  // MUELU_HIERARCHYMANAGER_HPP
 
-//TODO: split into _decl/_def
-// TODO: default value for first param (FactoryManager()) should not be duplicated (code maintainability)
+// TODO: split into _decl/_def
+//  TODO: default value for first param (FactoryManager()) should not be duplicated (code maintainability)
diff --git a/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.cpp b/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.cpp
index a2c64571b2a3..6edaeabbc366 100644
--- a/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.cpp
+++ b/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.cpp
@@ -46,452 +46,503 @@
 
 #include "MueLu_ConfigDefs.hpp"
 #if defined(HAVE_MUELU_ML)
-# include <ml_config.h>
-# if defined(HAVE_ML_EPETRA) && defined(HAVE_ML_TEUCHOS)
-#  include <ml_ValidateParameters.h>
-#  include <ml_MultiLevelPreconditioner.h> // for default values
-#  include <ml_RefMaxwell.h>
-# endif
+#include <ml_config.h>
+#if defined(HAVE_ML_EPETRA) && defined(HAVE_ML_TEUCHOS)
+#include <ml_ValidateParameters.h>
+#include <ml_MultiLevelPreconditioner.h>  // for default values
+#include <ml_RefMaxwell.h>
+#endif
 #endif
 
 #include <MueLu_ML2MueLuParameterTranslator.hpp>
 
 namespace MueLu {
 
+std::string ML2MueLuParameterTranslator::GetSmootherFactory(const Teuchos::ParameterList& paramList, Teuchos::ParameterList& adaptingParamList, const std::string& pname, const std::string& value) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pname != "coarse: type" && pname != "coarse: list" && pname != "smoother: type" && pname.find("smoother: list", 0) != 0,
+                             Exceptions::RuntimeError,
+                             "MueLu::MLParameterListInterpreter::Setup(): Only \"coarse: type\", \"smoother: type\" or \"smoother: list\" (\"coarse: list\") are "
+                             "supported as ML parameters for transformation of smoother/solver parameters to MueLu");
+
+  // string stream containing the smoother/solver xml parameters
+  std::stringstream mueluss;
+
+  // Check whether we are dealing with coarse level (solver) parameters or level smoother parameters
+  std::string mode = "smoother:";
+  bool is_coarse   = false;
+  if (pname.find("coarse:", 0) == 0) {
+    mode      = "coarse:";
+    is_coarse = true;
+  }
 
-  std::string ML2MueLuParameterTranslator::GetSmootherFactory(const Teuchos::ParameterList& paramList, Teuchos::ParameterList& adaptingParamList, const std::string& pname, const std::string& value) {
-
-    TEUCHOS_TEST_FOR_EXCEPTION(pname != "coarse: type" && pname != "coarse: list" && pname != "smoother: type" && pname.find("smoother: list",0) != 0,
-      Exceptions::RuntimeError,
-      "MueLu::MLParameterListInterpreter::Setup(): Only \"coarse: type\", \"smoother: type\" or \"smoother: list\" (\"coarse: list\") are "
-      "supported as ML parameters for transformation of smoother/solver parameters to MueLu");
+  // check whether pre and/or post smoothing
+  std::string PreOrPost = "both";
+  if (paramList.isParameter(mode + " pre or post"))
+    PreOrPost = paramList.get<std::string>(mode + " pre or post");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(mode == "coarse:" && PreOrPost != "both", Exceptions::RuntimeError,
+                             "MueLu::MLParameterListInterpreter::Setup(): The parameter \"coarse: pre or post\" is not supported by MueLu. "
+                             "It does not make sense for direct solvers. For iterative solvers you obtain the same effect by increasing, "
+                             "e.g., the number of sweeps for the coarse grid smoother. Please remove it from your parameters.");
+
+  // select smoother type
+  std::string valuestr = value;  // temporary variable
+  std::transform(valuestr.begin(), valuestr.end(), valuestr.begin(), ::tolower);
+  if (valuestr == "jacobi" || valuestr == "gauss-seidel" || valuestr == "symmetric gauss-seidel") {
+    std::string my_name;
+    if (PreOrPost == "both")
+      my_name = "\"" + pname + "\"";
+    else
+      my_name = "\"smoother: " + PreOrPost + " type\"";
+    mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"RELAXATION\"/>" << std::endl;
+
+  } else if (valuestr == "hiptmair") {
+    std::string my_name;
+    if (PreOrPost == "both")
+      my_name = "\"" + pname + "\"";
+    else
+      my_name = "\"smoother: " + PreOrPost + " type\"";
+    mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"HIPTMAIR\"/>" << std::endl;
+
+  } else if (valuestr == "ifpack") {
+    std::string my_name = "\"" + pname + "\"";
+    if (paramList.isParameter("smoother: ifpack type")) {
+      if (paramList.get<std::string>("smoother: ifpack type") == "ILU") {
+        mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"ILU\"/>" << std::endl;
+        adaptingParamList.remove("smoother: ifpack type", false);
+      }
+      if (paramList.get<std::string>("smoother: ifpack type") == "ILUT") {
+        mueluss << "<Parameter name=" << my_name << " type\" type=\"string\" value=\"ILUT\"/>" << std::endl;
+        adaptingParamList.remove("smoother: ifpack type", false);
+      }
+    }
 
-    
+  } else if ((valuestr == "chebyshev") || (valuestr == "mls")) {
+    std::string my_name = "\"" + pname + "\"";
+    mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"CHEBYSHEV\"/>" << std::endl;
 
-    // string stream containing the smoother/solver xml parameters
-    std::stringstream mueluss;
+  } else if (valuestr.length() > strlen("amesos") && valuestr.substr(0, strlen("amesos")) == "amesos") { /* catch Amesos-* */
+    std::string solverType = valuestr.substr(strlen("amesos") + 1);                                      /* ("amesos-klu" -> "klu") */
 
-    // Check whether we are dealing with coarse level (solver) parameters or level smoother parameters
-    std::string mode = "smoother:";
-    bool is_coarse = false;
-    if (pname.find("coarse:", 0) == 0) {
-      mode = "coarse:";
-      is_coarse = true;
-    }
+    bool valid                           = false;
+    const int validatorSize              = 5;
+    std::string validator[validatorSize] = {"superlu", "superludist", "klu", "umfpack", "mumps"};
+    for (int i = 0; i < validatorSize; i++)
+      if (validator[i] == solverType)
+        valid = true;
+    TEUCHOS_TEST_FOR_EXCEPTION(!valid, Exceptions::RuntimeError,
+                               "MueLu::MLParameterListInterpreter: unknown smoother type. '" << solverType << "' not supported.");
 
-    // check whether pre and/or post smoothing
-    std::string PreOrPost = "both";
-    if (paramList.isParameter(mode + " pre or post"))
-      PreOrPost = paramList.get<std::string>(mode + " pre or post");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(mode == "coarse:" && PreOrPost != "both", Exceptions::RuntimeError,
-      "MueLu::MLParameterListInterpreter::Setup(): The parameter \"coarse: pre or post\" is not supported by MueLu. "
-      "It does not make sense for direct solvers. For iterative solvers you obtain the same effect by increasing, "
-      "e.g., the number of sweeps for the coarse grid smoother. Please remove it from your parameters.");
-
-    // select smoother type
-    std::string valuestr = value; // temporary variable
-    std::transform(valuestr.begin(), valuestr.end(), valuestr.begin(), ::tolower);
-    if ( valuestr == "jacobi" || valuestr == "gauss-seidel" || valuestr == "symmetric gauss-seidel" ) {
-      std::string my_name;
-      if ( PreOrPost == "both" ) my_name = "\"" + pname + "\"";
-      else                       my_name = "\"smoother: " + PreOrPost + " type\"";
-      mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"RELAXATION\"/>" << std::endl;
-
-    } else if ( valuestr == "hiptmair" ) {
-      std::string my_name;
-      if ( PreOrPost == "both" ) my_name = "\"" + pname + "\"";
-      else                       my_name = "\"smoother: " + PreOrPost + " type\"";
-      mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"HIPTMAIR\"/>" << std::endl;
-
-    } else if ( valuestr == "ifpack" ) {
-      std::string my_name = "\"" + pname + "\"";
-      if ( paramList.isParameter("smoother: ifpack type") ) {
-        if ( paramList.get<std::string>("smoother: ifpack type") == "ILU" ) {
-          mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"ILU\"/>" << std::endl;
-          adaptingParamList.remove("smoother: ifpack type",false);
-        }
-        if ( paramList.get<std::string>("smoother: ifpack type") == "ILUT" ) {
-          mueluss << "<Parameter name=" << my_name << " type\" type=\"string\" value=\"ILUT\"/>" << std::endl;
-          adaptingParamList.remove("smoother: ifpack type",false);
-        }
-      }
+    mueluss << "<Parameter name=\"" << pname << "\" type=\"string\" value=\"" << solverType << "\"/>" << std::endl;
 
-    } else if (( valuestr == "chebyshev" ) || ( valuestr == "mls" )) {
-       std::string my_name = "\"" + pname + "\"";
-       mueluss << "<Parameter name=" << my_name << " type=\"string\" value=\"CHEBYSHEV\"/>" << std::endl;
+  } else {
+    // TODO error message
+    std::cout << "error in " << __FILE__ << ":" << __LINE__ << " could not find valid smoother/solver" << std::endl;
+  }
 
-    } else if (valuestr.length() > strlen("amesos") && valuestr.substr(0, strlen("amesos")) == "amesos") {  /* catch Amesos-* */
-      std::string solverType = valuestr.substr(strlen("amesos")+1);  /* ("amesos-klu" -> "klu") */
+  // set smoother: pre or post parameter
+  // Note that there is no "coarse: pre or post" in MueLu!
+  if (paramList.isParameter("smoother: pre or post") && mode == "smoother:") {
+    // std::cout << "paramList" << paramList << std::endl;
+    // std::string smootherPreOrPost = paramList.get<std::string>("smoother: pre or post");
+    // std::cout << "Create pre or post parameter with " << smootherPreOrPost << std::endl;
+    mueluss << "<Parameter name=\"smoother: pre or post\" type=\"string\" value=\"" << PreOrPost << "\"/>" << std::endl;
+    adaptingParamList.remove("smoother: pre or post", false);
+  }
 
-      bool valid = false;
-      const int  validatorSize = 5;
-      std::string validator[validatorSize] = {"superlu", "superludist", "klu", "umfpack", "mumps"};
-      for (int i=0; i < validatorSize; i++)
-        if (validator[i] == solverType)
-          valid = true;
-      TEUCHOS_TEST_FOR_EXCEPTION(!valid, Exceptions::RuntimeError,
-        "MueLu::MLParameterListInterpreter: unknown smoother type. '" << solverType << "' not supported.");
+  // create smoother parameter list
+  if (PreOrPost != "both") {
+    mueluss << "<ParameterList name=\"smoother: " << PreOrPost << " params\">" << std::endl;
+  } else {
+    mueluss << "<ParameterList name=\"" << mode << " params\">" << std::endl;
+  }
 
-      mueluss << "<Parameter name=\"" << pname << "\" type=\"string\" value=\"" << solverType << "\"/>" << std::endl;
+  // relaxation based smoothers:
 
-    } else {
-      // TODO error message
-      std::cout << "error in " << __FILE__ << ":" << __LINE__ << " could not find valid smoother/solver" << std::endl;
+  if (valuestr == "jacobi" || valuestr == "gauss-seidel" || valuestr == "symmetric gauss-seidel") {
+    if (valuestr == "jacobi") {
+      mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Jacobi\"/>" << std::endl;
+      adaptingParamList.remove("relaxation: type", false);
+    }
+    if (valuestr == "gauss-seidel") {
+      mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Gauss-Seidel\"/>" << std::endl;
+      adaptingParamList.remove("relaxation: type", false);
+    }
+    if (valuestr == "symmetric gauss-seidel") {
+      mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>" << std::endl;
+      adaptingParamList.remove("relaxation: type", false);
     }
 
-    // set smoother: pre or post parameter
-    // Note that there is no "coarse: pre or post" in MueLu!
-    if ( paramList.isParameter("smoother: pre or post") && mode == "smoother:") {
-      //std::cout << "paramList" << paramList << std::endl;
-      //std::string smootherPreOrPost = paramList.get<std::string>("smoother: pre or post");
-      //std::cout << "Create pre or post parameter with " << smootherPreOrPost << std::endl;
-      mueluss << "<Parameter name=\"smoother: pre or post\" type=\"string\" value=\"" << PreOrPost << "\"/>" << std::endl;
-      adaptingParamList.remove("smoother: pre or post",false);
+    if (paramList.isParameter("smoother: sweeps")) {
+      mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>("smoother: sweeps") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: sweeps", false);
+    }
+    if (paramList.isParameter("smoother: damping factor")) {
+      mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>("smoother: damping factor") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: damping factor", false);
+    }
+    if (paramList.isParameter("smoother: use l1 Gauss-Seidel")) {
+      mueluss << "<Parameter name=\"relaxation: use l1\" type=\"bool\" value=\"" << paramList.get<bool>("smoother: use l1 Gauss-Seidel") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: use l1 Gauss-Seidel", false);
     }
+  }
 
-    // create smoother parameter list
-    if (PreOrPost != "both") {
-      mueluss << "<ParameterList name=\"smoother: " << PreOrPost << " params\">" << std::endl;
+  // Chebyshev
+  if (valuestr == "chebyshev") {
+    if (paramList.isParameter("smoother: polynomial order")) {
+      mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: polynomial order") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: polynomial order", false);
     } else {
-      mueluss << "<ParameterList name=\"" << mode << " params\">" << std::endl;
+      mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"2\"/>" << std::endl;
     }
-
-    // relaxation based smoothers:
-
-    if ( valuestr == "jacobi" || valuestr == "gauss-seidel" || valuestr == "symmetric gauss-seidel" ) {
-      if ( valuestr == "jacobi" ) { mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Jacobi\"/>" << std::endl; adaptingParamList.remove("relaxation: type",false); }
-      if ( valuestr == "gauss-seidel" ) { mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Gauss-Seidel\"/>" << std::endl; adaptingParamList.remove("relaxation: type",false); }
-      if ( valuestr == "symmetric gauss-seidel" ) { mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"Symmetric Gauss-Seidel\"/>" << std::endl; adaptingParamList.remove("relaxation: type",false); }
-
-      if ( paramList.isParameter("smoother: sweeps") ) { mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>("smoother: sweeps") << "\"/>" << std::endl; adaptingParamList.remove("smoother: sweeps",false); }
-      if ( paramList.isParameter("smoother: damping factor") ) { mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>("smoother: damping factor") << "\"/>" << std::endl; adaptingParamList.remove("smoother: damping factor",false); }
-      if ( paramList.isParameter("smoother: use l1 Gauss-Seidel") ) { mueluss << "<Parameter name=\"relaxation: use l1\" type=\"bool\" value=\"" << paramList.get<bool>("smoother: use l1 Gauss-Seidel") << "\"/>" << std::endl; adaptingParamList.remove("smoother: use l1 Gauss-Seidel",false); }
+    if (paramList.isParameter("smoother: Chebyshev alpha")) {
+      mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: Chebyshev alpha") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: Chebyshev alpha", false);
+    } else {
+      mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"20\"/>" << std::endl;
+      adaptingParamList.remove("smoother: Chebyshev alpha", false);
     }
-
-    // Chebyshev
-    if ( valuestr == "chebyshev") {
-      if ( paramList.isParameter("smoother: polynomial order") ) { mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: polynomial order") << "\"/>" << std::endl; adaptingParamList.remove("smoother: polynomial order",false); }
-      else { mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"2\"/>" << std::endl; }
-      if ( paramList.isParameter("smoother: Chebyshev alpha") ) { mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: Chebyshev alpha") << "\"/>" << std::endl; adaptingParamList.remove("smoother: Chebyshev alpha",false); }
-      else { mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"20\"/>" << std::endl; adaptingParamList.remove("smoother: Chebyshev alpha",false); }
-      if ( paramList.isParameter("eigen-analysis: type") ) {  mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"" << paramList.get<std::string>("eigen-analysis: type") << "\"/>" << std::endl; adaptingParamList.remove("eigen-analysis: type",false); }
-      else { mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"cg\"/>" << std::endl; }
+    if (paramList.isParameter("eigen-analysis: type")) {
+      mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"" << paramList.get<std::string>("eigen-analysis: type") << "\"/>" << std::endl;
+      adaptingParamList.remove("eigen-analysis: type", false);
+    } else {
+      mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"cg\"/>" << std::endl;
     }
+  }
 
-    // MLS
-    if ( valuestr == "mls") {
-      if ( paramList.isParameter("smoother: MLS polynomial order") ) { mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: MLS polynomial order") << "\"/>" << std::endl; adaptingParamList.remove("smoother: MLS polynomial order",false); }
-      else if ( paramList.isParameter("smoother: polynomial order") ) { mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: polynomial order") << "\"/>" << std::endl; adaptingParamList.remove("smoother: polynomial order",false); }
-      else { mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"2\"/>" << std::endl; }
-      if ( paramList.isParameter("smoother: MLS alpha") ) { mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: MLS alpha") << "\"/>" << std::endl; adaptingParamList.remove("smoother: MLS alpha",false); }
-      else if ( paramList.isParameter("smoother: Chebyshev alpha") ) { mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: Chebyshev alpha") << "\"/>" << std::endl; adaptingParamList.remove("smoother: Chebyshev alpha",false); }
-      else { mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"20\"/>" << std::endl; }
-      if ( paramList.isParameter("eigen-analysis: type") ) {  mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"" << paramList.get<std::string>("eigen-analysis: type") << "\"/>" << std::endl; adaptingParamList.remove("eigen-analysis: type",false); }
-      else { mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"cg\"/>" << std::endl; }
+  // MLS
+  if (valuestr == "mls") {
+    if (paramList.isParameter("smoother: MLS polynomial order")) {
+      mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: MLS polynomial order") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: MLS polynomial order", false);
+    } else if (paramList.isParameter("smoother: polynomial order")) {
+      mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>("smoother: polynomial order") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: polynomial order", false);
+    } else {
+      mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"2\"/>" << std::endl;
+    }
+    if (paramList.isParameter("smoother: MLS alpha")) {
+      mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: MLS alpha") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: MLS alpha", false);
+    } else if (paramList.isParameter("smoother: Chebyshev alpha")) {
+      mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>("smoother: Chebyshev alpha") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: Chebyshev alpha", false);
+    } else {
+      mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"20\"/>" << std::endl;
+    }
+    if (paramList.isParameter("eigen-analysis: type")) {
+      mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"" << paramList.get<std::string>("eigen-analysis: type") << "\"/>" << std::endl;
+      adaptingParamList.remove("eigen-analysis: type", false);
+    } else {
+      mueluss << "<Parameter name=\"eigen-analysis: type\" type=\"string\" value=\"cg\"/>" << std::endl;
     }
+  }
 
-    if ( valuestr == "hiptmair" ) {
-      std::string subSmootherType = "Chebyshev";
-      if (!is_coarse && paramList.isParameter("subsmoother: type")) 
-        subSmootherType = paramList.get<std::string>("subsmoother: type");
-      if (is_coarse && paramList.isParameter("smoother: subsmoother type"))
-        subSmootherType = paramList.get<std::string>("smoother: subsmoother type");
-
-      std::string subSmootherIfpackType;
-      if (subSmootherType == "Chebyshev")
-        subSmootherIfpackType = "CHEBYSHEV";
-      else if (subSmootherType == "Jacobi" || subSmootherType == "Gauss-Seidel" || subSmootherType == "symmetric Gauss-Seidel") {
-        if (subSmootherType == "symmetric Gauss-Seidel") subSmootherType = "Symmetric Gauss-Seidel"; // FIXME
-        subSmootherIfpackType = "RELAXATION";
-      } else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListTranslator: unknown smoother type. '" << subSmootherType << "' not supported by MueLu.");
-
-      mueluss << "<Parameter name=\"hiptmair: smoother type 1\" type=\"string\" value=\"" << subSmootherIfpackType << "\"/>" << std::endl;
-      mueluss << "<Parameter name=\"hiptmair: smoother type 2\" type=\"string\" value=\"" << subSmootherIfpackType << "\"/>" << std::endl;
-
-      mueluss << "<ParameterList name=\"hiptmair: smoother list 1\">" << std::endl;
-      if (subSmootherType == "Chebyshev") {
-        std::string edge_sweeps = is_coarse ? "smoother: edge sweeps" : "subsmoother: edge sweeps";
-        std::string cheby_alpha = is_coarse ? "smoother: Chebyshev alpha" : "subsmoother: Chebyshev_alpha";
-
-        if (paramList.isParameter(edge_sweeps)) {
-          mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>(edge_sweeps) << "\"/>" << std::endl;
-          adaptingParamList.remove("subsmoother: edge sweeps", false);
-        }
-        if (paramList.isParameter(cheby_alpha)) {
-          mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>(cheby_alpha) << "\"/>" << std::endl;
-        }
-      } else {
-        std::string edge_sweeps = is_coarse ? "smoother: edge sweeps" : "subsmoother: edge sweeps";
-        std::string SGS_damping = is_coarse ? "smoother: SGS damping factor" : "subsmoother: SGS damping factor";
-
-        if (paramList.isParameter(edge_sweeps)) {
-          mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"" << subSmootherType << "\"/>" << std::endl;
-          mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>(edge_sweeps) << "\"/>" << std::endl;
-          adaptingParamList.remove(edge_sweeps, false);
-        }
-        if (paramList.isParameter(SGS_damping)) {
-          mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>(SGS_damping) << "\"/>" << std::endl;
-        }
+  if (valuestr == "hiptmair") {
+    std::string subSmootherType = "Chebyshev";
+    if (!is_coarse && paramList.isParameter("subsmoother: type"))
+      subSmootherType = paramList.get<std::string>("subsmoother: type");
+    if (is_coarse && paramList.isParameter("smoother: subsmoother type"))
+      subSmootherType = paramList.get<std::string>("smoother: subsmoother type");
+
+    std::string subSmootherIfpackType;
+    if (subSmootherType == "Chebyshev")
+      subSmootherIfpackType = "CHEBYSHEV";
+    else if (subSmootherType == "Jacobi" || subSmootherType == "Gauss-Seidel" || subSmootherType == "symmetric Gauss-Seidel") {
+      if (subSmootherType == "symmetric Gauss-Seidel") subSmootherType = "Symmetric Gauss-Seidel";  // FIXME
+      subSmootherIfpackType = "RELAXATION";
+    } else
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListTranslator: unknown smoother type. '" << subSmootherType << "' not supported by MueLu.");
+
+    mueluss << "<Parameter name=\"hiptmair: smoother type 1\" type=\"string\" value=\"" << subSmootherIfpackType << "\"/>" << std::endl;
+    mueluss << "<Parameter name=\"hiptmair: smoother type 2\" type=\"string\" value=\"" << subSmootherIfpackType << "\"/>" << std::endl;
+
+    mueluss << "<ParameterList name=\"hiptmair: smoother list 1\">" << std::endl;
+    if (subSmootherType == "Chebyshev") {
+      std::string edge_sweeps = is_coarse ? "smoother: edge sweeps" : "subsmoother: edge sweeps";
+      std::string cheby_alpha = is_coarse ? "smoother: Chebyshev alpha" : "subsmoother: Chebyshev_alpha";
+
+      if (paramList.isParameter(edge_sweeps)) {
+        mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>(edge_sweeps) << "\"/>" << std::endl;
+        adaptingParamList.remove("subsmoother: edge sweeps", false);
       }
-      mueluss << "</ParameterList>" << std::endl;
-
-      mueluss << "<ParameterList name=\"hiptmair: smoother list 2\">" << std::endl;
-      if (subSmootherType == "Chebyshev") {
-        std::string node_sweeps = is_coarse ? "smoother: node sweeps" : "subsmoother: node sweeps";
-        std::string cheby_alpha = is_coarse ? "smoother: Chebyshev alpha" : "subsmoother: Chebyshev_alpha";
-        if (paramList.isParameter(node_sweeps)) {
-          mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>(node_sweeps) << "\"/>" << std::endl;
-          adaptingParamList.remove("subsmoother: node sweeps", false);
-        }
-        if (paramList.isParameter(cheby_alpha)) {
-          mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>(cheby_alpha) << "\"/>" << std::endl;
-          adaptingParamList.remove("subsmoother: Chebyshev alpha", false);
-        }
-      } else {
-        std::string node_sweeps = is_coarse ? "smoother: node sweeps" : "subsmoother: node sweeps";
-        std::string SGS_damping = is_coarse ? "smoother: SGS damping factor" : "subsmoother: SGS damping factor";
-
-        if (paramList.isParameter(node_sweeps)) {
-          mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"" << subSmootherType << "\"/>" << std::endl;
-          mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>(node_sweeps) << "\"/>" << std::endl;
-          adaptingParamList.remove("subsmoother: node sweeps", false);
-        }
-        if (paramList.isParameter(SGS_damping)) {
-          mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>(SGS_damping) << "\"/>" << std::endl;
-          adaptingParamList.remove("subsmoother: SGS damping factor", false);
-        }
+      if (paramList.isParameter(cheby_alpha)) {
+        mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>(cheby_alpha) << "\"/>" << std::endl;
       }
-      mueluss << "</ParameterList>" << std::endl;
+    } else {
+      std::string edge_sweeps = is_coarse ? "smoother: edge sweeps" : "subsmoother: edge sweeps";
+      std::string SGS_damping = is_coarse ? "smoother: SGS damping factor" : "subsmoother: SGS damping factor";
 
+      if (paramList.isParameter(edge_sweeps)) {
+        mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"" << subSmootherType << "\"/>" << std::endl;
+        mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>(edge_sweeps) << "\"/>" << std::endl;
+        adaptingParamList.remove(edge_sweeps, false);
+      }
+      if (paramList.isParameter(SGS_damping)) {
+        mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>(SGS_damping) << "\"/>" << std::endl;
+      }
     }
+    mueluss << "</ParameterList>" << std::endl;
 
-    // parameters for ILU based preconditioners
-    if ( valuestr == "ifpack") {
+    mueluss << "<ParameterList name=\"hiptmair: smoother list 2\">" << std::endl;
+    if (subSmootherType == "Chebyshev") {
+      std::string node_sweeps = is_coarse ? "smoother: node sweeps" : "subsmoother: node sweeps";
+      std::string cheby_alpha = is_coarse ? "smoother: Chebyshev alpha" : "subsmoother: Chebyshev_alpha";
+      if (paramList.isParameter(node_sweeps)) {
+        mueluss << "<Parameter name=\"chebyshev: degree\" type=\"int\" value=\"" << paramList.get<int>(node_sweeps) << "\"/>" << std::endl;
+        adaptingParamList.remove("subsmoother: node sweeps", false);
+      }
+      if (paramList.isParameter(cheby_alpha)) {
+        mueluss << "<Parameter name=\"chebyshev: ratio eigenvalue\" type=\"double\" value=\"" << paramList.get<double>(cheby_alpha) << "\"/>" << std::endl;
+        adaptingParamList.remove("subsmoother: Chebyshev alpha", false);
+      }
+    } else {
+      std::string node_sweeps = is_coarse ? "smoother: node sweeps" : "subsmoother: node sweeps";
+      std::string SGS_damping = is_coarse ? "smoother: SGS damping factor" : "subsmoother: SGS damping factor";
 
-      // add Ifpack parameters
-      if ( paramList.isParameter("smoother: ifpack overlap") ) { mueluss << "<Parameter name=\"partitioner: overlap\" type=\"int\" value=\"" << paramList.get<int>("smoother: ifpack overlap") << "\"/>" << std::endl; adaptingParamList.remove("smoother: ifpack overlap",false); }
-      if ( paramList.isParameter("smoother: ifpack level-of-fill") ) { mueluss << "<Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"" << paramList.get<int>("smoother: ifpack level-of-fill") << "\"/>" << std::endl; adaptingParamList.remove("smoother: ifpack level-of-fill",false); }
-      if ( paramList.isParameter("smoother: ifpack absolute threshold") ) { mueluss << "<Parameter name=\"fact: absolute threshold\" type=\"int\" value=\"" << paramList.get<double>("smoother: ifpack absolute threshold") << "\"/>" << std::endl; adaptingParamList.remove("smoother: ifpack absolute threshold",false); }
-      if ( paramList.isParameter("smoother: ifpack relative threshold") ) { mueluss << "<Parameter name=\"fact: relative threshold\" type=\"int\" value=\"" << paramList.get<double>("smoother: ifpack relative threshold") << "\"/>" << std::endl; adaptingParamList.remove("smoother: ifpack relative threshold",false); }
+      if (paramList.isParameter(node_sweeps)) {
+        mueluss << "<Parameter name=\"relaxation: type\" type=\"string\" value=\"" << subSmootherType << "\"/>" << std::endl;
+        mueluss << "<Parameter name=\"relaxation: sweeps\" type=\"int\" value=\"" << paramList.get<int>(node_sweeps) << "\"/>" << std::endl;
+        adaptingParamList.remove("subsmoother: node sweeps", false);
+      }
+      if (paramList.isParameter(SGS_damping)) {
+        mueluss << "<Parameter name=\"relaxation: damping factor\" type=\"double\" value=\"" << paramList.get<double>(SGS_damping) << "\"/>" << std::endl;
+        adaptingParamList.remove("subsmoother: SGS damping factor", false);
+      }
     }
-
     mueluss << "</ParameterList>" << std::endl;
+  }
 
-    // max coarse level size parameter (outside of smoother parameter lists)
-    if ( paramList.isParameter("smoother: max size") ) {
-      mueluss << "<Parameter name=\"coarse: max size\" type=\"int\" value=\"" << paramList.get<int>("smoother: max size") << "\"/>" << std::endl; adaptingParamList.remove("smoother: max size",false);
+  // parameters for ILU based preconditioners
+  if (valuestr == "ifpack") {
+    // add Ifpack parameters
+    if (paramList.isParameter("smoother: ifpack overlap")) {
+      mueluss << "<Parameter name=\"partitioner: overlap\" type=\"int\" value=\"" << paramList.get<int>("smoother: ifpack overlap") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: ifpack overlap", false);
+    }
+    if (paramList.isParameter("smoother: ifpack level-of-fill")) {
+      mueluss << "<Parameter name=\"fact: level-of-fill\" type=\"int\" value=\"" << paramList.get<int>("smoother: ifpack level-of-fill") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: ifpack level-of-fill", false);
+    }
+    if (paramList.isParameter("smoother: ifpack absolute threshold")) {
+      mueluss << "<Parameter name=\"fact: absolute threshold\" type=\"int\" value=\"" << paramList.get<double>("smoother: ifpack absolute threshold") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: ifpack absolute threshold", false);
+    }
+    if (paramList.isParameter("smoother: ifpack relative threshold")) {
+      mueluss << "<Parameter name=\"fact: relative threshold\" type=\"int\" value=\"" << paramList.get<double>("smoother: ifpack relative threshold") << "\"/>" << std::endl;
+      adaptingParamList.remove("smoother: ifpack relative threshold", false);
     }
-
-    return mueluss.str();
   }
 
-  std::string ML2MueLuParameterTranslator::SetParameterList(const Teuchos::ParameterList & paramList_in, const std::string& defaultVals) {
-    Teuchos::ParameterList paramList = paramList_in;
-    
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout)); // TODO: use internal out (GetOStream())
-
-#if defined(HAVE_MUELU_ML) && defined(HAVE_ML_EPETRA) && defined(HAVE_ML_TEUCHOS)
+  mueluss << "</ParameterList>" << std::endl;
 
-    // TODO alternative with standard parameterlist from ML user guide?
+  // max coarse level size parameter (outside of smoother parameter lists)
+  if (paramList.isParameter("smoother: max size")) {
+    mueluss << "<Parameter name=\"coarse: max size\" type=\"int\" value=\"" << paramList.get<int>("smoother: max size") << "\"/>" << std::endl;
+    adaptingParamList.remove("smoother: max size", false);
+  }
 
-    if (defaultVals != "") {
-      TEUCHOS_TEST_FOR_EXCEPTION(defaultVals!="SA" && defaultVals!="NSSA" && defaultVals!="refmaxwell" && defaultVals!="Maxwell", Exceptions::RuntimeError,
-                                   "MueLu::MLParameterListInterpreter: only \"SA\", \"NSSA\", \"refmaxwell\" and \"Maxwell\" allowed as options for ML default parameters.");
-      Teuchos::ParameterList ML_defaultlist;
-      if (defaultVals == "refmaxwell")
-        ML_Epetra::SetDefaultsRefMaxwell(ML_defaultlist);
-      else
-        ML_Epetra::SetDefaults(defaultVals,ML_defaultlist);
+  return mueluss.str();
+}
 
-      // merge user parameters with default parameters
-      MueLu::MergeParameterList(paramList_in, ML_defaultlist, true);
-      paramList = ML_defaultlist;
-    }
-#else
-    if (defaultVals != "") {
-        // If no validator available: issue a warning and set parameter value to false in the output list
-        *out << "Warning: MueLu_ENABLE_ML=OFF, ML_ENABLE_Epetra=OFF or ML_ENABLE_TEUCHOS=OFF. No ML default values available." << std::endl;
-    }
-#endif // HAVE_MUELU_ML && HAVE_ML_EPETRA && HAVE_ML_TEUCHOS
+std::string ML2MueLuParameterTranslator::SetParameterList(const Teuchos::ParameterList& paramList_in, const std::string& defaultVals) {
+  Teuchos::ParameterList paramList = paramList_in;
 
-    //
-    // Move smoothers/aggregation/coarse parameters to sublists
-    //
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));  // TODO: use internal out (GetOStream())
 
-    // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
-    // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
-    ParameterList paramListWithSubList;
-    MueLu::CreateSublists(paramList, paramListWithSubList);
+#if defined(HAVE_MUELU_ML) && defined(HAVE_ML_EPETRA) && defined(HAVE_ML_TEUCHOS)
 
-    paramList = paramListWithSubList; // swap
-    Teuchos::ParameterList adaptingParamList = paramList;    // copy of paramList which is used to removed already interpreted parameters
+  // TODO alternative with standard parameterlist from ML user guide?
 
-    //
-    // Validate parameter list
-    //
-    {
-      bool validate = paramList.get("ML validate parameter list", true); /* true = default in ML */
-      if (validate && defaultVals!="refmaxwell") {
+  if (defaultVals != "") {
+    TEUCHOS_TEST_FOR_EXCEPTION(defaultVals != "SA" && defaultVals != "NSSA" && defaultVals != "refmaxwell" && defaultVals != "Maxwell", Exceptions::RuntimeError,
+                               "MueLu::MLParameterListInterpreter: only \"SA\", \"NSSA\", \"refmaxwell\" and \"Maxwell\" allowed as options for ML default parameters.");
+    Teuchos::ParameterList ML_defaultlist;
+    if (defaultVals == "refmaxwell")
+      ML_Epetra::SetDefaultsRefMaxwell(ML_defaultlist);
+    else
+      ML_Epetra::SetDefaults(defaultVals, ML_defaultlist);
 
+    // merge user parameters with default parameters
+    MueLu::MergeParameterList(paramList_in, ML_defaultlist, true);
+    paramList = ML_defaultlist;
+  }
+#else
+  if (defaultVals != "") {
+    // If no validator available: issue a warning and set parameter value to false in the output list
+    *out << "Warning: MueLu_ENABLE_ML=OFF, ML_ENABLE_Epetra=OFF or ML_ENABLE_TEUCHOS=OFF. No ML default values available." << std::endl;
+  }
+#endif  // HAVE_MUELU_ML && HAVE_ML_EPETRA && HAVE_ML_TEUCHOS
+
+  //
+  // Move smoothers/aggregation/coarse parameters to sublists
+  //
+
+  // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
+  // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
+  ParameterList paramListWithSubList;
+  MueLu::CreateSublists(paramList, paramListWithSubList);
+
+  paramList                                = paramListWithSubList;  // swap
+  Teuchos::ParameterList adaptingParamList = paramList;             // copy of paramList which is used to removed already interpreted parameters
+
+  //
+  // Validate parameter list
+  //
+  {
+    bool validate = paramList.get("ML validate parameter list", true); /* true = default in ML */
+    if (validate && defaultVals != "refmaxwell") {
 #if defined(HAVE_MUELU_ML) && defined(HAVE_ML_EPETRA) && defined(HAVE_ML_TEUCHOS)
-        // Validate parameter list using ML validator
-        int depth = paramList.get("ML validate depth", 5); /* 5 = default in ML */
-        TEUCHOS_TEST_FOR_EXCEPTION(! ML_Epetra::ValidateMLPParameters(paramList, depth), Exceptions::RuntimeError,
-                                   "ERROR: ML's Teuchos::ParameterList contains incorrect parameter!");
+      // Validate parameter list using ML validator
+      int depth = paramList.get("ML validate depth", 5); /* 5 = default in ML */
+      TEUCHOS_TEST_FOR_EXCEPTION(!ML_Epetra::ValidateMLPParameters(paramList, depth), Exceptions::RuntimeError,
+                                 "ERROR: ML's Teuchos::ParameterList contains incorrect parameter!");
 #else
-        // If no validator available: issue a warning and set parameter value to false in the output list
-        *out << "Warning: MueLu_ENABLE_ML=OFF, ML_ENABLE_Epetra=OFF or ML_ENABLE_TEUCHOS=OFF. The parameter list cannot be validated." << std::endl;
-        paramList.set("ML validate parameter list", false);
-
-#endif // HAVE_MUELU_ML && HAVE_ML_EPETRA && HAVE_ML_TEUCHOS
-      } // if(validate)
-    } // scope
-
-
-    {
-      // Special handling of ML's aux aggregation
-      //
-      // In ML, when "aggregation: aux: enable" == true, the threshold
-      // is set via "aggregation: aux: threshold" instead of
-      // "aggregation: threshold". In MueLu, we use "aggregation: drop
-      // tol" regardless of "sa: use filtering". So depending on
-      // "aggregation: aux: enable" we use either one or the other
-      // threshold to set "aggregation: drop tol".
-      if (paramListWithSubList.isParameter("aggregation: aux: enable") && paramListWithSubList.get<bool>("aggregation: aux: enable")) {
-        if (paramListWithSubList.isParameter("aggregation: aux: threshold")) {
-          paramListWithSubList.set("aggregation: threshold", paramListWithSubList.get<double>("aggregation: aux: threshold"));
-          paramListWithSubList.remove("aggregation: aux: threshold");
-          }
-      }
-    }
-
-    // stringstream for concatenating xml parameter strings.
-    std::stringstream mueluss;
+      // If no validator available: issue a warning and set parameter value to false in the output list
+      *out << "Warning: MueLu_ENABLE_ML=OFF, ML_ENABLE_Epetra=OFF or ML_ENABLE_TEUCHOS=OFF. The parameter list cannot be validated." << std::endl;
+      paramList.set("ML validate parameter list", false);
 
-    // create surrounding MueLu parameter list
-    mueluss << "<ParameterList name=\"MueLu\">" << std::endl;
+#endif  // HAVE_MUELU_ML && HAVE_ML_EPETRA && HAVE_ML_TEUCHOS
+    }   // if(validate)
+  }     // scope
 
-    // make sure that MueLu's phase1 matches ML's
-    mueluss << "<Parameter name=\"aggregation: match ML phase1\"       type=\"bool\"     value=\"true\"/>" << std::endl;
-
-    // make sure that MueLu's phase2a matches ML's
-    mueluss << "<Parameter name=\"aggregation: match ML phase2a\"      type=\"bool\"     value=\"true\"/>" << std::endl;
+  {
+    // Special handling of ML's aux aggregation
+    //
+    // In ML, when "aggregation: aux: enable" == true, the threshold
+    // is set via "aggregation: aux: threshold" instead of
+    // "aggregation: threshold". In MueLu, we use "aggregation: drop
+    // tol" regardless of "sa: use filtering". So depending on
+    // "aggregation: aux: enable" we use either one or the other
+    // threshold to set "aggregation: drop tol".
+    if (paramListWithSubList.isParameter("aggregation: aux: enable") && paramListWithSubList.get<bool>("aggregation: aux: enable")) {
+      if (paramListWithSubList.isParameter("aggregation: aux: threshold")) {
+        paramListWithSubList.set("aggregation: threshold", paramListWithSubList.get<double>("aggregation: aux: threshold"));
+        paramListWithSubList.remove("aggregation: aux: threshold");
+      }
+    }
+  }
 
-    // make sure that MueLu's phase2b matches ML's
-    mueluss << "<Parameter name=\"aggregation: match ML phase2b\"      type=\"bool\"     value=\"true\"/>" << std::endl;
+  // stringstream for concatenating xml parameter strings.
+  std::stringstream mueluss;
 
-    // make sure that MueLu's drop tol matches ML's
-    mueluss << "<Parameter name=\"aggregation: use ml scaling of drop tol\"      type=\"bool\"     value=\"true\"/>" << std::endl;
+  // create surrounding MueLu parameter list
+  mueluss << "<ParameterList name=\"MueLu\">" << std::endl;
 
+  // make sure that MueLu's phase1 matches ML's
+  mueluss << "<Parameter name=\"aggregation: match ML phase1\"       type=\"bool\"     value=\"true\"/>" << std::endl;
 
-    // loop over all ML parameters in provided parameter list
-    for (ParameterList::ConstIterator param = paramListWithSubList.begin(); param != paramListWithSubList.end(); ++param) {
+  // make sure that MueLu's phase2a matches ML's
+  mueluss << "<Parameter name=\"aggregation: match ML phase2a\"      type=\"bool\"     value=\"true\"/>" << std::endl;
 
-      // extract ML parameter name
-      const std::string & pname=paramListWithSubList.name(param);
+  // make sure that MueLu's phase2b matches ML's
+  mueluss << "<Parameter name=\"aggregation: match ML phase2b\"      type=\"bool\"     value=\"true\"/>" << std::endl;
 
-      // extract corresponding (ML) value
-      // remove ParameterList specific information from result string
-      std::stringstream valuess;
-      valuess << paramList.entry(param);
-      std::string valuestr = valuess.str();
-      replaceAll(valuestr, "[unused]", "");
-      replaceAll(valuestr, "[default]", "");
-      valuestr = trim(valuestr);
+  // make sure that MueLu's drop tol matches ML's
+  mueluss << "<Parameter name=\"aggregation: use ml scaling of drop tol\"      type=\"bool\"     value=\"true\"/>" << std::endl;
 
-      // transform ML parameter to corresponding MueLu parameter and generate XML string
-      std::string valueInterpreterStr = "\"" + valuestr + "\"";
-      std::string ret = MasterList::interpretParameterName(MasterList::ML2MueLu(pname),valueInterpreterStr);
+  // loop over all ML parameters in provided parameter list
+  for (ParameterList::ConstIterator param = paramListWithSubList.begin(); param != paramListWithSubList.end(); ++param) {
+    // extract ML parameter name
+    const std::string& pname = paramListWithSubList.name(param);
 
-      if ((pname == "aggregation: aux: enable") && (paramListWithSubList.get<bool>("aggregation: aux: enable"))) {
-        mueluss << "<Parameter name=\"aggregation: drop scheme\" type=\"string\"     value=\"" << "distance laplacian" << "\"/>" << std::endl;
-      }
+    // extract corresponding (ML) value
+    // remove ParameterList specific information from result string
+    std::stringstream valuess;
+    valuess << paramList.entry(param);
+    std::string valuestr = valuess.str();
+    replaceAll(valuestr, "[unused]", "");
+    replaceAll(valuestr, "[default]", "");
+    valuestr = trim(valuestr);
 
-      // special handling for verbosity level
-      if (pname == "ML output") {
-        // Translate verbosity parameter
-        int verbosityLevel = std::stoi(valuestr);
-        std::string eVerbLevel = "none";
-        if (verbosityLevel ==  0) eVerbLevel = "none";
-        if (verbosityLevel >=  1) eVerbLevel = "low";
-        if (verbosityLevel >=  5) eVerbLevel = "medium";
-        if (verbosityLevel >= 10) eVerbLevel = "high";
-        if (verbosityLevel >= 11) eVerbLevel = "extreme";
-        if (verbosityLevel >= 42) eVerbLevel = "test";
-        if (verbosityLevel >= 666) eVerbLevel = "interfacetest";
-        mueluss << "<Parameter name=\"verbosity\" type=\"string\"     value=\"" << eVerbLevel << "\"/>" << std::endl;
-        continue;
-      }
+    // transform ML parameter to corresponding MueLu parameter and generate XML string
+    std::string valueInterpreterStr = "\"" + valuestr + "\"";
+    std::string ret                 = MasterList::interpretParameterName(MasterList::ML2MueLu(pname), valueInterpreterStr);
 
-      // add XML string
-      if (ret != "") {
-        mueluss << ret << std::endl;
+    if ((pname == "aggregation: aux: enable") && (paramListWithSubList.get<bool>("aggregation: aux: enable"))) {
+      mueluss << "<Parameter name=\"aggregation: drop scheme\" type=\"string\"     value=\""
+              << "distance laplacian"
+              << "\"/>" << std::endl;
+    }
 
-        // remove parameter from ML parameter list
-        adaptingParamList.remove(pname,false);
-      }
+    // special handling for verbosity level
+    if (pname == "ML output") {
+      // Translate verbosity parameter
+      int verbosityLevel     = std::stoi(valuestr);
+      std::string eVerbLevel = "none";
+      if (verbosityLevel == 0) eVerbLevel = "none";
+      if (verbosityLevel >= 1) eVerbLevel = "low";
+      if (verbosityLevel >= 5) eVerbLevel = "medium";
+      if (verbosityLevel >= 10) eVerbLevel = "high";
+      if (verbosityLevel >= 11) eVerbLevel = "extreme";
+      if (verbosityLevel >= 42) eVerbLevel = "test";
+      if (verbosityLevel >= 666) eVerbLevel = "interfacetest";
+      mueluss << "<Parameter name=\"verbosity\" type=\"string\"     value=\"" << eVerbLevel << "\"/>" << std::endl;
+      continue;
+    }
 
-      // special handling for energy minimization
-      // TAW: this is not optimal for symmetric problems but at least works.
-      //      for symmetric problems the "energy minimization" parameter should not exist anyway...
-      if (pname == "energy minimization: enable") {
-        mueluss << "<Parameter name=\"problem: symmetric\"      type=\"bool\"     value=\"false\"/>" << std::endl;
-        mueluss << "<Parameter name=\"transpose: use implicit\" type=\"bool\"     value=\"false\"/>" << std::endl;
-      }
+    // add XML string
+    if (ret != "") {
+      mueluss << ret << std::endl;
 
-      // special handling for smoothers
-      if (pname == "smoother: type") {
+      // remove parameter from ML parameter list
+      adaptingParamList.remove(pname, false);
+    }
 
-        mueluss << GetSmootherFactory(paramList, adaptingParamList, pname, valuestr);
+    // special handling for energy minimization
+    // TAW: this is not optimal for symmetric problems but at least works.
+    //      for symmetric problems the "energy minimization" parameter should not exist anyway...
+    if (pname == "energy minimization: enable") {
+      mueluss << "<Parameter name=\"problem: symmetric\"      type=\"bool\"     value=\"false\"/>" << std::endl;
+      mueluss << "<Parameter name=\"transpose: use implicit\" type=\"bool\"     value=\"false\"/>" << std::endl;
+    }
 
-      }
+    // special handling for smoothers
+    if (pname == "smoother: type") {
+      mueluss << GetSmootherFactory(paramList, adaptingParamList, pname, valuestr);
+    }
 
-      // special handling for level-specific smoothers
-      if (pname.find("smoother: list (level",0) == 0) {
-        // Scan pname (ex: pname="smoother: type (level 2)")
-        std::string type, option;
-        int levelID=-1;
-        {
-          typedef Teuchos::ArrayRCP<char>::size_type size_type;
-          Teuchos::Array<char> ctype  (size_type(pname.size()+1));
-          Teuchos::Array<char> coption(size_type(pname.size()+1));
-
-          int matched = sscanf(pname.c_str(),"%s %[^(](level %d)", ctype.getRawPtr(), coption.getRawPtr(), &levelID); // use [^(] instead of %s to allow for strings with white-spaces (ex: "ifpack list")
-          type = std::string(ctype.getRawPtr());
-          option = std::string(coption.getRawPtr()); option.resize(option.size () - 1); // remove final white-space
-
-          if (matched != 3 || (type != "smoother:")) {
-            TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::CreateSublist(), Line " << __LINE__ << ". "
-                                        << "Error in creating level-specific sublists" << std::endl
-                                        << "Offending parameter: " << pname << std::endl);
-          }
-
-          mueluss << "<ParameterList name=\"level " << levelID << "\">" << std::endl;
-          mueluss << GetSmootherFactory(paramList.sublist(pname),adaptingParamList.sublist(pname), "smoother: type", paramList.sublist(pname).get<std::string>("smoother: type"));
-          mueluss << "</ParameterList>" << std::endl;
+    // special handling for level-specific smoothers
+    if (pname.find("smoother: list (level", 0) == 0) {
+      // Scan pname (ex: pname="smoother: type (level 2)")
+      std::string type, option;
+      int levelID = -1;
+      {
+        typedef Teuchos::ArrayRCP<char>::size_type size_type;
+        Teuchos::Array<char> ctype(size_type(pname.size() + 1));
+        Teuchos::Array<char> coption(size_type(pname.size() + 1));
+
+        int matched = sscanf(pname.c_str(), "%s %[^(](level %d)", ctype.getRawPtr(), coption.getRawPtr(), &levelID);  // use [^(] instead of %s to allow for strings with white-spaces (ex: "ifpack list")
+        type        = std::string(ctype.getRawPtr());
+        option      = std::string(coption.getRawPtr());
+        option.resize(option.size() - 1);  // remove final white-space
+
+        if (matched != 3 || (type != "smoother:")) {
+          TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::CreateSublist(), Line " << __LINE__ << ". "
+                                                                                                            << "Error in creating level-specific sublists" << std::endl
+                                                                                                            << "Offending parameter: " << pname << std::endl);
         }
-      }
-
-      // special handling for coarse level
-      TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter("coarse: type"), Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter::Setup(): The parameter \"coarse: type\" should not exist but being stored in \"coarse: list\" instead.");
-      if ( pname == "coarse: list" ) {
-
-        // interpret smoother/coarse solver data.
-        // Note, that we inspect the "coarse: list" sublist to define the "coarse" smoother/solver
-        // Be aware, that MueLu::CreateSublists renames the prefix of the parameters in the "coarse: list" from "coarse" to "smoother".
-        // Therefore, we have to check the values of the "smoother" parameters
-        mueluss << GetSmootherFactory(paramList.sublist("coarse: list"), adaptingParamList.sublist("coarse: list"), "coarse: type", paramList.sublist("coarse: list").get<std::string>("smoother: type"));
-
 
+        mueluss << "<ParameterList name=\"level " << levelID << "\">" << std::endl;
+        mueluss << GetSmootherFactory(paramList.sublist(pname), adaptingParamList.sublist(pname), "smoother: type", paramList.sublist(pname).get<std::string>("smoother: type"));
+        mueluss << "</ParameterList>" << std::endl;
       }
-    } // for
+    }
 
-    mueluss << "</ParameterList>" << std::endl;
+    // special handling for coarse level
+    TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter("coarse: type"), Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter::Setup(): The parameter \"coarse: type\" should not exist but being stored in \"coarse: list\" instead.");
+    if (pname == "coarse: list") {
+      // interpret smoother/coarse solver data.
+      // Note, that we inspect the "coarse: list" sublist to define the "coarse" smoother/solver
+      // Be aware, that MueLu::CreateSublists renames the prefix of the parameters in the "coarse: list" from "coarse" to "smoother".
+      // Therefore, we have to check the values of the "smoother" parameters
+      mueluss << GetSmootherFactory(paramList.sublist("coarse: list"), adaptingParamList.sublist("coarse: list"), "coarse: type", paramList.sublist("coarse: list").get<std::string>("smoother: type"));
+    }
+  }  // for
 
-    return mueluss.str();
-  }
+  mueluss << "</ParameterList>" << std::endl;
 
+  return mueluss.str();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.hpp b/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.hpp
index 34697f6fd4e9..ee67d1a075d3 100644
--- a/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.hpp
+++ b/packages/muelu/src/Interface/MueLu_ML2MueLuParameterTranslator.hpp
@@ -59,98 +59,95 @@
 
 namespace MueLu {
 
-  /*!
-    @class ML2MueLuParameterTranslator class.
-    @brief Class that accepts ML-style parameters and builds a MueLu parameter list (easy input deck)
-
-    This interpreter class is meant to make the transition from ML to MueLu easier.
-  */
-  class ML2MueLuParameterTranslator {
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    ML2MueLuParameterTranslator() { }
-
-    //! Destructor.
-    virtual ~ML2MueLuParameterTranslator() { }
-
-    //@}
-
-    //!@name Parameter translation from ML to MueLu
-    //@{
-
-    /// @brief: Translate ML parameters to MueLu parameter XML string
-    ///
-    /// @param [in] paramList_in: ML parameter list
-    /// @return std::string with MueLu XML parameters
-    static std::string translate(Teuchos::ParameterList & paramList, const std::string& defaultVals="") {
-      return SetParameterList(paramList, defaultVals);
-    }
-
-    /// @brief: Translate ML parameters to MueLu parameter XML string
-    ///
-    /// @param [in] xmlFileName: file name with ML xml parameters
-    /// @return std::string with MueLu XML parameters
-    static std::string translate(const std::string & xmlFileName, const std::string& defaultVals="") {
-      Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
-      return SetParameterList(*paramList, defaultVals);
-    }
-
-    //@}
-
-  private:
-
-    //! @name Parameter handling
-    //@{
-
-    /// @brief: Interpret parameter list
-    ///
-    /// @param [in] paramList_in: ML parameter list
-    /// @return std::string with MueLu XML parameters
-    static std::string SetParameterList(const Teuchos::ParameterList & paramList_in, const std::string& defaultVals);
-
-
-    /// @brief: Helper function which translates ML smoother/solver paramters to MueLu XML string
-    ///
-    /// @param [in] paramList: reference to Teuchos::ParameterList containing the ML smoother/solver parameters.
-    /// @param [in,out] adaptingParamList: reference to Teuchos::ParameterList containing the ML smoother/solver parameters. Note that the processed parameters are removed from the ParameterList. It can be used to detect non-interpreted ML parameters.
-    /// @param [in] pname: currently processed parameter TODO
-    /// @param [in] value: currently processed value TODO
-    static std::string GetSmootherFactory(const Teuchos::ParameterList& paramList, Teuchos::ParameterList& adaptingParamList, const std::string& pname, const std::string& value);
-
-    //@}
-
-    //
-    // helper routines
-    //
-
-    // trim from start
-    static inline std::string &ltrim(std::string &s) {
-      s.erase(s.begin(), std::find_if(s.begin(), s.end(), [](int c){return !std::isspace(c);}));
-      return s;
-    }
-
-    // trim from end
-    static inline std::string &rtrim(std::string &s) {
-      s.erase(std::find_if(s.rbegin(), s.rend(), [](int c){return !std::isspace(c);}).base(), s.end());
-      return s;
-    }
-
-    // trim from both ends
-    static inline std::string &trim(std::string &s) {
-      return ltrim(rtrim(s));
-    }
-
-    //! @name Member variables
-    //@{
-    //std::string xmlString_;   ///! string containing MueLu XML parameters corresponding to ML parameters
-    //@}
-
-  }; // class MLParameterListInterpreter
-
-
-} // end namespace MueLu
+/*!
+  @class ML2MueLuParameterTranslator class.
+  @brief Class that accepts ML-style parameters and builds a MueLu parameter list (easy input deck)
+
+  This interpreter class is meant to make the transition from ML to MueLu easier.
+*/
+class ML2MueLuParameterTranslator {
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor.
+  ML2MueLuParameterTranslator() {}
+
+  //! Destructor.
+  virtual ~ML2MueLuParameterTranslator() {}
+
+  //@}
+
+  //!@name Parameter translation from ML to MueLu
+  //@{
+
+  /// @brief: Translate ML parameters to MueLu parameter XML string
+  ///
+  /// @param [in] paramList_in: ML parameter list
+  /// @return std::string with MueLu XML parameters
+  static std::string translate(Teuchos::ParameterList& paramList, const std::string& defaultVals = "") {
+    return SetParameterList(paramList, defaultVals);
+  }
+
+  /// @brief: Translate ML parameters to MueLu parameter XML string
+  ///
+  /// @param [in] xmlFileName: file name with ML xml parameters
+  /// @return std::string with MueLu XML parameters
+  static std::string translate(const std::string& xmlFileName, const std::string& defaultVals = "") {
+    Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
+    return SetParameterList(*paramList, defaultVals);
+  }
+
+  //@}
+
+ private:
+  //! @name Parameter handling
+  //@{
+
+  /// @brief: Interpret parameter list
+  ///
+  /// @param [in] paramList_in: ML parameter list
+  /// @return std::string with MueLu XML parameters
+  static std::string SetParameterList(const Teuchos::ParameterList& paramList_in, const std::string& defaultVals);
+
+  /// @brief: Helper function which translates ML smoother/solver paramters to MueLu XML string
+  ///
+  /// @param [in] paramList: reference to Teuchos::ParameterList containing the ML smoother/solver parameters.
+  /// @param [in,out] adaptingParamList: reference to Teuchos::ParameterList containing the ML smoother/solver parameters. Note that the processed parameters are removed from the ParameterList. It can be used to detect non-interpreted ML parameters.
+  /// @param [in] pname: currently processed parameter TODO
+  /// @param [in] value: currently processed value TODO
+  static std::string GetSmootherFactory(const Teuchos::ParameterList& paramList, Teuchos::ParameterList& adaptingParamList, const std::string& pname, const std::string& value);
+
+  //@}
+
+  //
+  // helper routines
+  //
+
+  // trim from start
+  static inline std::string& ltrim(std::string& s) {
+    s.erase(s.begin(), std::find_if(s.begin(), s.end(), [](int c) { return !std::isspace(c); }));
+    return s;
+  }
+
+  // trim from end
+  static inline std::string& rtrim(std::string& s) {
+    s.erase(std::find_if(s.rbegin(), s.rend(), [](int c) { return !std::isspace(c); }).base(), s.end());
+    return s;
+  }
+
+  // trim from both ends
+  static inline std::string& trim(std::string& s) {
+    return ltrim(rtrim(s));
+  }
+
+  //! @name Member variables
+  //@{
+  // std::string xmlString_;   ///! string containing MueLu XML parameters corresponding to ML parameters
+  //@}
+
+};  // class MLParameterListInterpreter
+
+}  // end namespace MueLu
 
 #endif /* MUELU_ML2MUELUPARAMETERTRANSLATOR_HPP  */
diff --git a/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_decl.hpp b/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_decl.hpp
index de92e1a77b90..97f721807e87 100644
--- a/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_decl.hpp
+++ b/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_decl.hpp
@@ -85,162 +85,155 @@
 #include "MueLu_RebalanceMapFactory_fwd.hpp"
 #endif
 
-
 #ifdef HAVE_MUELU_DEPRECATED_CODE
 #ifndef TRILINOS_HIDE_DEPRECATED_HEADER_WARNINGS
 #warning "The header file MueLu_MLParameterListInterpreter.hpp is deprecated"
 #endif
-#else 
+#else
 #error "The header file MueLu_MLParameterListInterpreter.hpp is deprecated"
 #endif
 
-
-
-
-
 namespace MueLu {
 
-  /*
-    Utility that from an existing Teuchos::ParameterList creates a new list, in
-    which level-specific parameters are replaced with sublists.
-
-    Currently, level-specific parameters that begin with "smoother:"
-    or "aggregation:" are placed in sublists. Coarse options are also placed
-    in a coarse list.
-
-    Example:
-    Input:
-    smoother: type (level 0) = symmetric Gauss-Seidel
-    smoother: sweeps (level 0) = 1
-    Output:
-    smoother: list (level 0) ->
-    smoother: type = symmetric Gauss-Seidel
-    smoother: sweeps = 1
-  */
-  // This function is a copy of ML_CreateSublists to avoid dependency on ML
-  // Throw exception on error instead of exit()
-  void CreateSublists(const ParameterList &List, ParameterList &newList);
-
-
-  /*!
-    @class MLParameterListInterpreter class.
-    @brief Class that accepts ML-style parameters and builds a MueLu preconditioner.
-    This interpreter uses the same default values as ML. This allows to compare ML/MueLu results
-
-    The parameter list is validated only if the package ML is available and parameter "ML validate parameter list" is true.
-    TODO: A warning is issued if ML is not available
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MLParameterListInterpreter : public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+/*
+  Utility that from an existing Teuchos::ParameterList creates a new list, in
+  which level-specific parameters are replaced with sublists.
+
+  Currently, level-specific parameters that begin with "smoother:"
+  or "aggregation:" are placed in sublists. Coarse options are also placed
+  in a coarse list.
+
+  Example:
+  Input:
+  smoother: type (level 0) = symmetric Gauss-Seidel
+  smoother: sweeps (level 0) = 1
+  Output:
+  smoother: list (level 0) ->
+  smoother: type = symmetric Gauss-Seidel
+  smoother: sweeps = 1
+*/
+// This function is a copy of ML_CreateSublists to avoid dependency on ML
+// Throw exception on error instead of exit()
+void CreateSublists(const ParameterList& List, ParameterList& newList);
+
+/*!
+  @class MLParameterListInterpreter class.
+  @brief Class that accepts ML-style parameters and builds a MueLu preconditioner.
+  This interpreter uses the same default values as ML. This allows to compare ML/MueLu results
+
+  The parameter list is validated only if the package ML is available and parameter "ML validate parameter list" is true.
+  TODO: A warning is issued if ML is not available
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MLParameterListInterpreter : public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_MLPARAMETERLISTINTERPRETER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    MLParameterListInterpreter() : nullspace_(NULL), blksize_(1) { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    //! @param paramList: parameter list with ML parameters
-    //! @param[in] comm  (RCP<Teuchos::Comm<int> >): Optional RCP of a Teuchos communicator  (default: Teuchos::null)
-    //! @param factoryList: vector with RCP of FactoryBase objects
-    //!
-    //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
-    //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
-    //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    MLParameterListInterpreter(Teuchos::ParameterList & paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::null, std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
+  //! Constructor.
+  MLParameterListInterpreter()
+    : nullspace_(NULL)
+    , blksize_(1) {}
 
-    //! Constructor.
-    //! @param xmlFileName: file name for XML file with ML parameters
-    //! @param factoryList: vector with RCP of FactoryBase objects
-    //!
-    //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
-    //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
-    //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    MLParameterListInterpreter(const std::string & xmlFileName,std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
+  //! Constructor.
+  //! @param paramList: parameter list with ML parameters
+  //! @param[in] comm  (RCP<Teuchos::Comm<int> >): Optional RCP of a Teuchos communicator  (default: Teuchos::null)
+  //! @param factoryList: vector with RCP of FactoryBase objects
+  //!
+  //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
+  //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
+  //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  MLParameterListInterpreter(Teuchos::ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::null, std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
 
-    //! Destructor.
-    virtual ~MLParameterListInterpreter() { }
+  //! Constructor.
+  //! @param xmlFileName: file name for XML file with ML parameters
+  //! @param factoryList: vector with RCP of FactoryBase objects
+  //!
+  //! The factories in factoryList allow the user to add user-specific factories to the MueLu Hierarchy.
+  //! The idea is to be able to add some factories that write out some debug information etc. which are not handled by the ML
+  //! Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  MLParameterListInterpreter(const std::string& xmlFileName, std::vector<RCP<FactoryBase> > factoryList = std::vector<RCP<FactoryBase> >(0));
 
-    //@}
+  //! Destructor.
+  virtual ~MLParameterListInterpreter() {}
 
-    //@{
+  //@}
 
-    void SetParameterList(const Teuchos::ParameterList & paramList);
+  //@{
 
-    //@}
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //@{
+  //@}
 
-    //! Setup Hierarchy object
-    virtual void SetupHierarchy(Hierarchy & H) const;
+  //@{
 
-    //@}
+  //! Setup Hierarchy object
+  virtual void SetupHierarchy(Hierarchy& H) const;
 
-    //@{
+  //@}
 
-    //! @name static helper functions translating parameter list to factories
-    //! @brief static helper functions that also can be used from outside for translating ML parameters into MueLu objects
-    //@{
+  //@{
 
-    //! Read smoother options and build the corresponding smoother factory
-    // @param AFact: Factory used by smoother to find 'A'
-    static RCP<SmootherFactory> GetSmootherFactory(const Teuchos::ParameterList & paramList, const RCP<FactoryBase> & AFact = Teuchos::null);
+  //! @name static helper functions translating parameter list to factories
+  //! @brief static helper functions that also can be used from outside for translating ML parameters into MueLu objects
+  //@{
 
-    //@}
+  //! Read smoother options and build the corresponding smoother factory
+  // @param AFact: Factory used by smoother to find 'A'
+  static RCP<SmootherFactory> GetSmootherFactory(const Teuchos::ParameterList& paramList, const RCP<FactoryBase>& AFact = Teuchos::null);
 
+  //@}
 
-    //! @name Handling of additional user-specific transfer factories
-    //@{
-    /*! @brief Add transfer factory in the end of list of transfer factories for RAPFactory.
+  //! @name Handling of additional user-specific transfer factories
+  //@{
+  /*! @brief Add transfer factory in the end of list of transfer factories for RAPFactory.
 
-    This allows the user to add user-specific factories to the MueLu Hierarchy. The idea is to be able
-    to add some factories that write out some debug information etc. which are not handled by the ML
-    Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
-    */
-    void AddTransferFactory(const RCP<FactoryBase> & factory);
-
-    //! Returns number of transfer factories.
-    size_t NumTransferFactories() const;
-    //@}
-
-  private:
+  This allows the user to add user-specific factories to the MueLu Hierarchy. The idea is to be able
+  to add some factories that write out some debug information etc. which are not handled by the ML
+  Parameter List itself. See information about the RAPFactory::AddTransferFactory method, too!
+  */
+  void AddTransferFactory(const RCP<FactoryBase>& factory);
 
-    //! nullspace can be embedded in the ML parameter list
-    int     nullspaceDim_;
-    double* nullspace_; //TODO: replace by Teuchos::ArrayRCP<>
+  //! Returns number of transfer factories.
+  size_t NumTransferFactories() const;
+  //@}
 
-    //! coordinates can be embedded in the ML parameter list
-    double* xcoord_;
-    double* ycoord_;
-    double* zcoord_;
+ private:
+  //! nullspace can be embedded in the ML parameter list
+  int nullspaceDim_;
+  double* nullspace_;  // TODO: replace by Teuchos::ArrayRCP<>
 
-    //! list of user-defined transfer Factories
-    //! We use this vector to add some special user-given factories to the Hierarchy (RAPFactory)
-    //! This way the user can extend the standard functionality of the MLParameterListInterpreter beyond the
-    //! capabibilities of ML.
-    std::vector<RCP<FactoryBase> > TransferFacts_;
+  //! coordinates can be embedded in the ML parameter list
+  double* xcoord_;
+  double* ycoord_;
+  double* zcoord_;
 
-    //@{ Matrix configuration
+  //! list of user-defined transfer Factories
+  //! We use this vector to add some special user-given factories to the Hierarchy (RAPFactory)
+  //! This way the user can extend the standard functionality of the MLParameterListInterpreter beyond the
+  //! capabibilities of ML.
+  std::vector<RCP<FactoryBase> > TransferFacts_;
 
-    //! Setup Operator object
-    virtual void SetupOperator(Operator & Op) const;
+  //@{ Matrix configuration
 
-    //! Matrix configuration storage
-    int blksize_;
+  //! Setup Operator object
+  virtual void SetupOperator(Operator& Op) const;
 
-    //@}
+  //! Matrix configuration storage
+  int blksize_;
 
-  }; // class MLParameterListInterpreter
+  //@}
 
-} // namespace MueLu
+};  // class MLParameterListInterpreter
 
+}  // namespace MueLu
 
 #define MUELU_MLPARAMETERLISTINTERPRETER_SHORT
 #endif /* MUELU_MLPARAMETERLISTINTERPRETER_DECL_HPP */
diff --git a/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_def.hpp b/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_def.hpp
index 06f98e8dd2ad..c18c69acea82 100644
--- a/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_def.hpp
+++ b/packages/muelu/src/Interface/MueLu_MLParameterListInterpreter_def.hpp
@@ -100,7 +100,6 @@
 //#include "MueLu_RebalanceMapFactory.hpp"
 #endif
 
-
 // Note: do not add options that are only recognized by MueLu.
 
 // TODO: this parameter list interpreter should force MueLu to use default ML parameters
@@ -108,675 +107,668 @@
 
 // Read a parameter value from a parameter list and store it into a variable named 'varName'
 #define MUELU_READ_PARAM(paramList, paramStr, varType, defaultValue, varName) \
-  varType varName = defaultValue; if (paramList.isParameter(paramStr)) varName = paramList.get<varType>(paramStr);
+  varType varName = defaultValue;                                             \
+  if (paramList.isParameter(paramStr)) varName = paramList.get<varType>(paramStr);
 
 // Read a parameter value from a paraeter list and copy it into a new parameter list (with another parameter name)
 #define MUELU_COPY_PARAM(paramList, paramStr, varType, defaultValue, outParamList, outParamStr) \
-  if (paramList.isParameter(paramStr))                                  \
-    outParamList.set(outParamStr, paramList.get<varType>(paramStr)); \
-  else outParamList.set(outParamStr, static_cast<varType>(defaultValue));            \
+  if (paramList.isParameter(paramStr))                                                          \
+    outParamList.set(outParamStr, paramList.get<varType>(paramStr));                            \
+  else                                                                                          \
+    outParamList.set(outParamStr, static_cast<varType>(defaultValue));
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MLParameterListInterpreter(Teuchos::ParameterList & paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm, std::vector<RCP<FactoryBase> > factoryList) : nullspace_(NULL), xcoord_(NULL), ycoord_(NULL), zcoord_(NULL),TransferFacts_(factoryList), blksize_(1) {
-
-    if (paramList.isParameter("xml parameter file")){
-      std::string filename = paramList.get("xml parameter file","");
-      if (filename.length() != 0) {
-        TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), Exceptions::RuntimeError, "xml parameter file requires a valid comm");
-        Teuchos::ParameterList paramList2 = paramList;
-        Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&paramList2),*comm);
-        paramList2.remove("xml parameter file");
-        SetParameterList(paramList2);
-      }
-      else
-        SetParameterList(paramList);
-    }
-    else
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MLParameterListInterpreter(Teuchos::ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm, std::vector<RCP<FactoryBase> > factoryList)
+  : nullspace_(NULL)
+  , xcoord_(NULL)
+  , ycoord_(NULL)
+  , zcoord_(NULL)
+  , TransferFacts_(factoryList)
+  , blksize_(1) {
+  if (paramList.isParameter("xml parameter file")) {
+    std::string filename = paramList.get("xml parameter file", "");
+    if (filename.length() != 0) {
+      TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), Exceptions::RuntimeError, "xml parameter file requires a valid comm");
+      Teuchos::ParameterList paramList2 = paramList;
+      Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&paramList2), *comm);
+      paramList2.remove("xml parameter file");
+      SetParameterList(paramList2);
+    } else
       SetParameterList(paramList);
+  } else
+    SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MLParameterListInterpreter(const std::string& xmlFileName, std::vector<RCP<FactoryBase> > factoryList)
+  : nullspace_(NULL)
+  , TransferFacts_(factoryList)
+  , blksize_(1) {
+  Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
+  SetParameterList(*paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList_in) {
+  Teuchos::ParameterList paramList = paramList_in;
+
+  //
+  // Read top-level of the parameter list
+  //
+
+  // hard-coded default values == ML defaults according to the manual
+  MUELU_READ_PARAM(paramList, "ML output", int, 0, verbosityLevel);
+  MUELU_READ_PARAM(paramList, "max levels", int, 10, maxLevels);
+  MUELU_READ_PARAM(paramList, "PDE equations", int, 1, nDofsPerNode);
+
+  MUELU_READ_PARAM(paramList, "coarse: max size", int, 128, maxCoarseSize);
+
+  MUELU_READ_PARAM(paramList, "aggregation: type", std::string, "Uncoupled", agg_type);
+  // MUELU_READ_PARAM(paramList, "aggregation: threshold",                double,                 0.0,       agg_threshold);
+  MUELU_READ_PARAM(paramList, "aggregation: damping factor", double, (double)4 / (double)3, agg_damping);
+  // MUELU_READ_PARAM(paramList, "aggregation: smoothing sweeps",            int,                   1,       agg_smoothingsweeps);
+  MUELU_READ_PARAM(paramList, "aggregation: nodes per aggregate", int, 1, minPerAgg);
+  MUELU_READ_PARAM(paramList, "aggregation: keep Dirichlet bcs", bool, false, bKeepDirichletBcs);                // This is a MueLu specific extension that does not exist in ML
+  MUELU_READ_PARAM(paramList, "aggregation: max neighbours already aggregated", int, 0, maxNbrAlreadySelected);  // This is a MueLu specific extension that does not exist in M
+  MUELU_READ_PARAM(paramList, "aggregation: aux: enable", bool, false, agg_use_aux);
+  MUELU_READ_PARAM(paramList, "aggregation: aux: threshold", double, false, agg_aux_thresh);
+
+  MUELU_READ_PARAM(paramList, "null space: type", std::string, "default vectors", nullspaceType);
+  MUELU_READ_PARAM(paramList, "null space: dimension", int, -1, nullspaceDim);      // TODO: ML default not in documentation
+  MUELU_READ_PARAM(paramList, "null space: vectors", double*, NULL, nullspaceVec);  // TODO: ML default not in documentation
+
+  MUELU_READ_PARAM(paramList, "energy minimization: enable", bool, false, bEnergyMinimization);
+
+  MUELU_READ_PARAM(paramList, "RAP: fix diagonal", bool, false, bFixDiagonal);  // This is a MueLu specific extension that does not exist in ML
+
+  MUELU_READ_PARAM(paramList, "x-coordinates", double*, NULL, xcoord);
+  MUELU_READ_PARAM(paramList, "y-coordinates", double*, NULL, ycoord);
+  MUELU_READ_PARAM(paramList, "z-coordinates", double*, NULL, zcoord);
+
+  //
+  // Move smoothers/aggregation/coarse parameters to sublists
+  //
+
+  // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
+  // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
+  ParameterList paramListWithSubList;
+  MueLu::CreateSublists(paramList, paramListWithSubList);
+  paramList = paramListWithSubList;  // swap
+
+  // pull out "use kokkos refactor"
+  bool setKokkosRefactor = false;
+  bool useKokkosRefactor = !Node::is_serial;
+  if (paramList.isType<bool>("use kokkos refactor")) {
+    useKokkosRefactor = paramList.get<bool>("use kokkos refactor");
+    setKokkosRefactor = true;
+    paramList.remove("use kokkos refactor");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MLParameterListInterpreter(const std::string & xmlFileName, std::vector<RCP<FactoryBase> > factoryList) : nullspace_(NULL), TransferFacts_(factoryList), blksize_(1) {
-    Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile(xmlFileName);
-    SetParameterList(*paramList);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList & paramList_in) {
-    Teuchos::ParameterList paramList = paramList_in;
+  //
+  // Validate parameter list
+  //
 
-    //
-    // Read top-level of the parameter list
-    //
-
-    // hard-coded default values == ML defaults according to the manual
-    MUELU_READ_PARAM(paramList, "ML output",                                int,                   0,       verbosityLevel);
-    MUELU_READ_PARAM(paramList, "max levels",                               int,                  10,       maxLevels);
-    MUELU_READ_PARAM(paramList, "PDE equations",                            int,                   1,       nDofsPerNode);
-
-    MUELU_READ_PARAM(paramList, "coarse: max size",                         int,                 128,       maxCoarseSize);
-
-    MUELU_READ_PARAM(paramList, "aggregation: type",                std::string,         "Uncoupled",       agg_type);
-    //MUELU_READ_PARAM(paramList, "aggregation: threshold",                double,                 0.0,       agg_threshold);
-    MUELU_READ_PARAM(paramList, "aggregation: damping factor",           double, (double)4/(double)3,       agg_damping);
-    //MUELU_READ_PARAM(paramList, "aggregation: smoothing sweeps",            int,                   1,       agg_smoothingsweeps);
-    MUELU_READ_PARAM(paramList, "aggregation: nodes per aggregate",         int,                   1,       minPerAgg);
-    MUELU_READ_PARAM(paramList, "aggregation: keep Dirichlet bcs",         bool,               false,       bKeepDirichletBcs); // This is a MueLu specific extension that does not exist in ML
-    MUELU_READ_PARAM(paramList, "aggregation: max neighbours already aggregated", int,             0,       maxNbrAlreadySelected); // This is a MueLu specific extension that does not exist in M
-    MUELU_READ_PARAM(paramList, "aggregation: aux: enable",                bool,               false,       agg_use_aux);
-    MUELU_READ_PARAM(paramList, "aggregation: aux: threshold",           double,               false,       agg_aux_thresh);
-
-    MUELU_READ_PARAM(paramList, "null space: type",                 std::string,   "default vectors",       nullspaceType);
-    MUELU_READ_PARAM(paramList, "null space: dimension",                    int,                  -1,       nullspaceDim); // TODO: ML default not in documentation
-    MUELU_READ_PARAM(paramList, "null space: vectors",                   double*,               NULL,       nullspaceVec); // TODO: ML default not in documentation
-
-    MUELU_READ_PARAM(paramList, "energy minimization: enable",             bool,               false,       bEnergyMinimization);
+  {
+    bool validate = paramList.get("ML validate parameter list", true); /* true = default in ML */
+    if (validate) {
+#if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
+      // Validate parameter list using ML validator
+      int depth = paramList.get("ML validate depth", 5); /* 5 = default in ML */
+      TEUCHOS_TEST_FOR_EXCEPTION(!ML_Epetra::ValidateMLPParameters(paramList, depth), Exceptions::RuntimeError,
+                                 "ERROR: ML's Teuchos::ParameterList contains incorrect parameter!");
+#else
+      // If no validator available: issue a warning and set parameter value to false in the output list
+      this->GetOStream(Warnings0) << "Warning: MueLu_ENABLE_ML=OFF. The parameter list cannot be validated." << std::endl;
+      paramList.set("ML validate parameter list", false);
+
+#endif  // HAVE_MUELU_ML
+    }   // if(validate)
+  }     // scope
+
+  // Matrix option
+  blksize_ = nDofsPerNode;
+
+  // Translate verbosity parameter
+
+  // Translate verbosity parameter
+  MsgType eVerbLevel = None;
+  if (verbosityLevel == 0) eVerbLevel = None;
+  if (verbosityLevel >= 1) eVerbLevel = Low;
+  if (verbosityLevel >= 5) eVerbLevel = Medium;
+  if (verbosityLevel >= 10) eVerbLevel = High;
+  if (verbosityLevel >= 11) eVerbLevel = Extreme;
+  if (verbosityLevel >= 42) eVerbLevel = Test;
+  if (verbosityLevel >= 43) eVerbLevel = InterfaceTest;
+  this->verbosity_ = eVerbLevel;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(agg_type != "Uncoupled", Exceptions::RuntimeError,
+                             "MueLu::MLParameterListInterpreter::SetParameterList(): parameter \"aggregation: type\": only 'Uncoupled' aggregation is supported.");
+
+  // Create MueLu factories
+  RCP<Factory> dropFact;
+  if (useKokkosRefactor)
+    dropFact = rcp(new CoalesceDropFactory_kokkos());
+  else
+    dropFact = rcp(new CoalesceDropFactory());
+
+  if (agg_use_aux) {
+    dropFact->SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+    dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(agg_aux_thresh));
+  }
 
-    MUELU_READ_PARAM(paramList, "RAP: fix diagonal",                       bool,               false,       bFixDiagonal); // This is a MueLu specific extension that does not exist in ML
+  // Uncoupled aggregation
+  RCP<Factory> AggFact = Teuchos::null;
+  if (useKokkosRefactor) {
+    AggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  } else
+    AggFact = rcp(new UncoupledAggregationFactory());
+
+  AggFact->SetFactory("Graph", dropFact);
+  AggFact->SetFactory("DofsPerNode", dropFact);
+  AggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(bKeepDirichletBcs));
+  AggFact->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("natural")));
+  AggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(maxNbrAlreadySelected));
+  AggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(minPerAgg));
+
+  if (verbosityLevel > 3) {
+    std::ostringstream oss;
+    oss << "========================= Aggregate option summary  =========================" << std::endl;
+    oss << "min Nodes per aggregate :               " << minPerAgg << std::endl;
+    oss << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
+    oss << "aggregate ordering :                    natural" << std::endl;
+    oss << "=============================================================================" << std::endl;
+    this->GetOStream(Runtime1) << oss.str();
+  }
 
-    MUELU_READ_PARAM(paramList, "x-coordinates",                        double*,                NULL,       xcoord);
-    MUELU_READ_PARAM(paramList, "y-coordinates",                        double*,                NULL,       ycoord);
-    MUELU_READ_PARAM(paramList, "z-coordinates",                        double*,                NULL,       zcoord);
+  RCP<Factory> PFact;
+  RCP<Factory> RFact;
+  RCP<Factory> PtentFact;
+  if (useKokkosRefactor)
+    PtentFact = rcp(new TentativePFactory_kokkos());
+  else
+    PtentFact = rcp(new TentativePFactory());
+  if (agg_damping == 0.0 && bEnergyMinimization == false) {
+    // tentative prolongation operator (PA-AMG)
+    PFact = PtentFact;
+    RFact = rcp(new TransPFactory());
+  } else if (agg_damping != 0.0 && bEnergyMinimization == false) {
+    // smoothed aggregation (SA-AMG)
+    RCP<Factory> SaPFact;
+    if (useKokkosRefactor)
+      SaPFact = rcp(new SaPFactory_kokkos());
+    else
+      SaPFact = rcp(new SaPFactory());
+    SaPFact->SetParameter("sa: damping factor", ParameterEntry(agg_damping));
+    PFact = SaPFact;
+    RFact = rcp(new TransPFactory());
+  } else if (bEnergyMinimization == true) {
+    // Petrov Galerkin PG-AMG smoothed aggregation (energy minimization in ML)
+    PFact = rcp(new PgPFactory());
+    RFact = rcp(new GenericRFactory());
+  }
 
+  RCP<RAPFactory> AcFact = rcp(new RAPFactory());
+  AcFact->SetParameter("RepairMainDiagonal", Teuchos::ParameterEntry(bFixDiagonal));
+  for (size_t i = 0; i < TransferFacts_.size(); i++) {
+    AcFact->AddTransferFactory(TransferFacts_[i]);
+  }
 
+  //
+  // introduce rebalancing
+  //
+#if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
+  Teuchos::RCP<Factory> RebalancedPFact            = Teuchos::null;
+  Teuchos::RCP<Factory> RebalancedRFact            = Teuchos::null;
+  Teuchos::RCP<Factory> RepartitionFact            = Teuchos::null;
+  Teuchos::RCP<RebalanceAcFactory> RebalancedAFact = Teuchos::null;
+
+  MUELU_READ_PARAM(paramList, "repartition: enable", int, 0, bDoRepartition);
+  if (bDoRepartition == 1) {
+    // The Factory Manager will be configured to return the rebalanced versions of P, R, A by default.
+    // Everytime we want to use the non-rebalanced versions, we need to explicitly define the generating factory.
+    RFact->SetFactory("P", PFact);
     //
-    // Move smoothers/aggregation/coarse parameters to sublists
-    //
+    AcFact->SetFactory("P", PFact);
+    AcFact->SetFactory("R", RFact);
 
-    // ML allows to have level-specific smoothers/aggregation/coarse parameters at the top level of the list or/and defined in sublists:
-    // See also: ML Guide section 6.4.1, MueLu::CreateSublists, ML_CreateSublists
-    ParameterList paramListWithSubList;
-    MueLu::CreateSublists(paramList, paramListWithSubList);
-    paramList = paramListWithSubList; // swap
-
-    // pull out "use kokkos refactor"
-    bool setKokkosRefactor = false;
-    bool useKokkosRefactor = !Node::is_serial;
-    if (paramList.isType<bool>("use kokkos refactor")) {
-      useKokkosRefactor = paramList.get<bool>("use kokkos refactor");
-      setKokkosRefactor = true;
-      paramList.remove("use kokkos refactor");
-    }
+    // define rebalancing factory for coarse matrix
+    Teuchos::RCP<MueLu::AmalgamationFactory<SC, LO, GO, NO> > rebAmalgFact = Teuchos::rcp(new MueLu::AmalgamationFactory<SC, LO, GO, NO>());
+    rebAmalgFact->SetFactory("A", AcFact);
 
-    //
-    // Validate parameter list
-    //
+    MUELU_READ_PARAM(paramList, "repartition: max min ratio", double, 1.3, maxminratio);
+    MUELU_READ_PARAM(paramList, "repartition: min per proc", int, 512, minperproc);
 
+    // Repartitioning heuristic
+    RCP<RepartitionHeuristicFactory> RepartitionHeuristicFact = Teuchos::rcp(new RepartitionHeuristicFactory());
     {
-      bool validate = paramList.get("ML validate parameter list", true); /* true = default in ML */
-      if (validate) {
+      Teuchos::ParameterList paramListRepFact;
+      paramListRepFact.set("repartition: min rows per proc", minperproc);
+      paramListRepFact.set("repartition: max imbalance", maxminratio);
+      RepartitionHeuristicFact->SetParameterList(paramListRepFact);
+    }
+    RepartitionHeuristicFact->SetFactory("A", AcFact);
+
+    // create "Partition"
+    Teuchos::RCP<MueLu::IsorropiaInterface<LO, GO, NO> > isoInterface = Teuchos::rcp(new MueLu::IsorropiaInterface<LO, GO, NO>());
+    isoInterface->SetFactory("A", AcFact);
+    isoInterface->SetFactory("number of partitions", RepartitionHeuristicFact);
+    isoInterface->SetFactory("UnAmalgamationInfo", rebAmalgFact);
+
+    // create "Partition" by unamalgamtion
+    Teuchos::RCP<MueLu::RepartitionInterface<LO, GO, NO> > repInterface = Teuchos::rcp(new MueLu::RepartitionInterface<LO, GO, NO>());
+    repInterface->SetFactory("A", AcFact);
+    repInterface->SetFactory("number of partitions", RepartitionHeuristicFact);
+    repInterface->SetFactory("AmalgamatedPartition", isoInterface);
+    // repInterface->SetFactory("UnAmalgamationInfo", rebAmalgFact); // not necessary?
+
+    // Repartitioning (creates "Importer" from "Partition")
+    RepartitionFact = Teuchos::rcp(new RepartitionFactory());
+    RepartitionFact->SetFactory("A", AcFact);
+    RepartitionFact->SetFactory("number of partitions", RepartitionHeuristicFact);
+    RepartitionFact->SetFactory("Partition", repInterface);
+
+    // Reordering of the transfer operators
+    RebalancedPFact = Teuchos::rcp(new RebalanceTransferFactory());
+    RebalancedPFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Interpolation")));
+    RebalancedPFact->SetFactory("P", PFact);
+    RebalancedPFact->SetFactory("Nullspace", PtentFact);
+    RebalancedPFact->SetFactory("Importer", RepartitionFact);
+
+    RebalancedRFact = Teuchos::rcp(new RebalanceTransferFactory());
+    RebalancedRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
+    RebalancedRFact->SetFactory("R", RFact);
+    RebalancedRFact->SetFactory("Importer", RepartitionFact);
+
+    // Compute Ac from rebalanced P and R
+    RebalancedAFact = Teuchos::rcp(new RebalanceAcFactory());
+    RebalancedAFact->SetFactory("A", AcFact);
+  }
+#else  // #ifdef HAVE_MUELU_ISORROPIA
+  // Get rid of [-Wunused] warnings
+  //(void)
+  //
+  // ^^^ FIXME (mfh 17 Nov 2013) That definitely doesn't compile.
+#endif
 
-#if defined(HAVE_MUELU_ML) && defined(HAVE_MUELU_EPETRA)
-        // Validate parameter list using ML validator
-        int depth = paramList.get("ML validate depth", 5); /* 5 = default in ML */
-        TEUCHOS_TEST_FOR_EXCEPTION(! ML_Epetra::ValidateMLPParameters(paramList, depth), Exceptions::RuntimeError,
-                                   "ERROR: ML's Teuchos::ParameterList contains incorrect parameter!");
-#else
-        // If no validator available: issue a warning and set parameter value to false in the output list
-        this->GetOStream(Warnings0) << "Warning: MueLu_ENABLE_ML=OFF. The parameter list cannot be validated." << std::endl;
-        paramList.set("ML validate parameter list", false);
+  //
+  // Nullspace factory
+  //
 
-#endif // HAVE_MUELU_ML
-      } // if(validate)
-    } // scope
+  // Set fine level nullspace
+  // extract pre-computed nullspace from ML parameter list
+  // store it in nullspace_ and nullspaceDim_
+  if (nullspaceType != "default vectors") {
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceType != "pre-computed", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (no pre-computed null space). error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceDim == -1, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace dim == -1). error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(nullspaceVec == NULL, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace == NULL). You have to provide a valid fine-level nullspace in \'null space: vectors\'");
 
+    nullspaceDim_ = nullspaceDim;
+    nullspace_    = nullspaceVec;
+  }
 
-    // Matrix option
-    blksize_ = nDofsPerNode;
+  Teuchos::RCP<NullspaceFactory> nspFact = Teuchos::rcp(new NullspaceFactory("Nullspace"));
+  nspFact->SetFactory("Nullspace", PtentFact);
 
-    // Translate verbosity parameter
+  // Stash coordinates
+  xcoord_ = xcoord;
+  ycoord_ = ycoord;
+  zcoord_ = zcoord;
 
-    // Translate verbosity parameter
-    MsgType eVerbLevel = None;
-    if (verbosityLevel ==  0) eVerbLevel = None;
-    if (verbosityLevel >=  1) eVerbLevel = Low;
-    if (verbosityLevel >=  5) eVerbLevel = Medium;
-    if (verbosityLevel >= 10) eVerbLevel = High;
-    if (verbosityLevel >= 11) eVerbLevel = Extreme;
-    if (verbosityLevel >= 42) eVerbLevel = Test;
-    if (verbosityLevel >= 43) eVerbLevel = InterfaceTest;
-    this->verbosity_ = eVerbLevel;
+  //
+  // Hierarchy + FactoryManager
+  //
 
+  // Hierarchy options
+  this->numDesiredLevel_ = maxLevels;
+  this->maxCoarseSize_   = maxCoarseSize;
 
-    TEUCHOS_TEST_FOR_EXCEPTION(agg_type != "Uncoupled", Exceptions::RuntimeError,
-        "MueLu::MLParameterListInterpreter::SetParameterList(): parameter \"aggregation: type\": only 'Uncoupled' aggregation is supported.");
+  //
+  // Coarse Smoother
+  //
+  ParameterList& coarseList = paramList.sublist("coarse: list");
+  // check whether coarse solver is set properly. If not, set default coarse solver.
+  if (!coarseList.isParameter("smoother: type"))
+    coarseList.set("smoother: type", "Amesos-KLU");  // set default coarse solver according to ML 5.0 guide
+  RCP<SmootherFactory> coarseFact = GetSmootherFactory(coarseList, Teuchos::null);
 
-    // Create MueLu factories
-    RCP<Factory> dropFact;
-    if(useKokkosRefactor)
-      dropFact = rcp( new CoalesceDropFactory_kokkos() );
-    else
-      dropFact = rcp( new CoalesceDropFactory() );
+  // Smoothers Top Level Parameters
 
-    if (agg_use_aux) {
-      dropFact->SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-      dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(agg_aux_thresh));
-    }
+  RCP<ParameterList> topLevelSmootherParam = ExtractSetOfParameters(paramList, "smoother");
 
-    // Uncoupled aggregation
-    RCP<Factory> AggFact = Teuchos::null;
-    if(useKokkosRefactor) {
-      AggFact = rcp( new UncoupledAggregationFactory_kokkos() );
-    }
-    else
-      AggFact = rcp( new UncoupledAggregationFactory() );
-
-    AggFact->SetFactory("Graph", dropFact);
-    AggFact->SetFactory("DofsPerNode", dropFact);
-    AggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(bKeepDirichletBcs));
-    AggFact->SetParameter("aggregation: ordering",                  Teuchos::ParameterEntry(std::string("natural")));
-    AggFact->SetParameter("aggregation: max selected neighbors",    Teuchos::ParameterEntry(maxNbrAlreadySelected));
-    AggFact->SetParameter("aggregation: min agg size",              Teuchos::ParameterEntry(minPerAgg));
-
-
-    if (verbosityLevel > 3) {
-      std::ostringstream oss;
-      oss << "========================= Aggregate option summary  =========================" << std::endl;
-      oss << "min Nodes per aggregate :               " << minPerAgg << std::endl;
-      oss << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
-      oss << "aggregate ordering :                    natural" << std::endl;
-      oss << "=============================================================================" << std::endl;
-      this->GetOStream(Runtime1) << oss.str();
-    }
+  //
 
-    RCP<Factory> PFact;
-    RCP<Factory> RFact;
-    RCP<Factory> PtentFact;
-    if(useKokkosRefactor)
-      PtentFact = rcp( new TentativePFactory_kokkos() );
-    else
-      PtentFact = rcp( new TentativePFactory() );
-    if (agg_damping == 0.0 && bEnergyMinimization == false) {
-      // tentative prolongation operator (PA-AMG)
-      PFact = PtentFact;
-      RFact = rcp( new TransPFactory() );
-    } else if (agg_damping != 0.0 && bEnergyMinimization == false) {
-      // smoothed aggregation (SA-AMG)
-      RCP<Factory> SaPFact;
-      if(useKokkosRefactor)
-        SaPFact = rcp( new SaPFactory_kokkos() );
-      else
-        SaPFact = rcp( new SaPFactory() );
-      SaPFact->SetParameter("sa: damping factor", ParameterEntry(agg_damping));
-      PFact  = SaPFact;
-      RFact  = rcp( new TransPFactory() );
-    } else if (bEnergyMinimization == true) {
-      // Petrov Galerkin PG-AMG smoothed aggregation (energy minimization in ML)
-      PFact  = rcp( new PgPFactory() );
-      RFact  = rcp( new GenericRFactory() );
-    }
-
-    RCP<RAPFactory> AcFact = rcp( new RAPFactory() );
-    AcFact->SetParameter("RepairMainDiagonal", Teuchos::ParameterEntry(bFixDiagonal));
-    for (size_t i = 0; i<TransferFacts_.size(); i++) {
-      AcFact->AddTransferFactory(TransferFacts_[i]);
-    }
+  // Prepare factory managers
+  // TODO: smootherFact can be reuse accross level if same parameters/no specific parameterList
 
+  for (int levelID = 0; levelID < maxLevels; levelID++) {
     //
-    // introduce rebalancing
+    // Level FactoryManager
     //
-#if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
-    Teuchos::RCP<Factory>            RebalancedPFact = Teuchos::null;
-    Teuchos::RCP<Factory>            RebalancedRFact = Teuchos::null;
-    Teuchos::RCP<Factory>            RepartitionFact = Teuchos::null;
-    Teuchos::RCP<RebalanceAcFactory> RebalancedAFact = Teuchos::null;
 
-    MUELU_READ_PARAM(paramList, "repartition: enable",                      int,                   0,       bDoRepartition);
-    if (bDoRepartition == 1) {
-      // The Factory Manager will be configured to return the rebalanced versions of P, R, A by default.
-      // Everytime we want to use the non-rebalanced versions, we need to explicitly define the generating factory.
-      RFact->SetFactory("P", PFact);
-      //
-      AcFact->SetFactory("P", PFact);
-      AcFact->SetFactory("R", RFact);
-
-      // define rebalancing factory for coarse matrix
-      Teuchos::RCP<MueLu::AmalgamationFactory<SC, LO, GO, NO> > rebAmalgFact = Teuchos::rcp(new MueLu::AmalgamationFactory<SC, LO, GO, NO>());
-      rebAmalgFact->SetFactory("A", AcFact);
-
-      MUELU_READ_PARAM(paramList, "repartition: max min ratio",            double,                 1.3,       maxminratio);
-      MUELU_READ_PARAM(paramList, "repartition: min per proc",                int,                 512,       minperproc);
-
-      // Repartitioning heuristic
-      RCP<RepartitionHeuristicFactory> RepartitionHeuristicFact = Teuchos::rcp(new RepartitionHeuristicFactory());
-      {
-        Teuchos::ParameterList paramListRepFact;
-        paramListRepFact.set("repartition: min rows per proc", minperproc);
-        paramListRepFact.set("repartition: max imbalance", maxminratio);
-        RepartitionHeuristicFact->SetParameterList(paramListRepFact);
-      }
-      RepartitionHeuristicFact->SetFactory("A", AcFact);
-
-      // create "Partition"
-      Teuchos::RCP<MueLu::IsorropiaInterface<LO, GO, NO> > isoInterface = Teuchos::rcp(new MueLu::IsorropiaInterface<LO, GO, NO>());
-      isoInterface->SetFactory("A", AcFact);
-      isoInterface->SetFactory("number of partitions", RepartitionHeuristicFact);
-      isoInterface->SetFactory("UnAmalgamationInfo", rebAmalgFact);
-
-      // create "Partition" by unamalgamtion
-      Teuchos::RCP<MueLu::RepartitionInterface<LO, GO, NO> > repInterface = Teuchos::rcp(new MueLu::RepartitionInterface<LO, GO, NO>());
-      repInterface->SetFactory("A", AcFact);
-      repInterface->SetFactory("number of partitions", RepartitionHeuristicFact);
-      repInterface->SetFactory("AmalgamatedPartition", isoInterface);
-      //repInterface->SetFactory("UnAmalgamationInfo", rebAmalgFact); // not necessary?
-
-      // Repartitioning (creates "Importer" from "Partition")
-      RepartitionFact = Teuchos::rcp(new RepartitionFactory());
-      RepartitionFact->SetFactory("A", AcFact);
-      RepartitionFact->SetFactory("number of partitions", RepartitionHeuristicFact);
-      RepartitionFact->SetFactory("Partition", repInterface);
-
-      // Reordering of the transfer operators
-      RebalancedPFact = Teuchos::rcp(new RebalanceTransferFactory());
-      RebalancedPFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Interpolation")));
-      RebalancedPFact->SetFactory("P", PFact);
-      RebalancedPFact->SetFactory("Nullspace", PtentFact);
-      RebalancedPFact->SetFactory("Importer",    RepartitionFact);
-
-      RebalancedRFact = Teuchos::rcp(new RebalanceTransferFactory());
-      RebalancedRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
-      RebalancedRFact->SetFactory("R", RFact);
-      RebalancedRFact->SetFactory("Importer",    RepartitionFact);
-
-      // Compute Ac from rebalanced P and R
-      RebalancedAFact = Teuchos::rcp(new RebalanceAcFactory());
-      RebalancedAFact->SetFactory("A", AcFact);
-    }
-#else // #ifdef HAVE_MUELU_ISORROPIA
-    // Get rid of [-Wunused] warnings
-    //(void)
-    //
-    // ^^^ FIXME (mfh 17 Nov 2013) That definitely doesn't compile.
-#endif
+    RCP<FactoryManager> manager = rcp(new FactoryManager());
+    if (setKokkosRefactor)
+      manager->SetKokkosRefactor(useKokkosRefactor);
 
     //
-    // Nullspace factory
+    // Smoothers
     //
 
-    // Set fine level nullspace
-    // extract pre-computed nullspace from ML parameter list
-    // store it in nullspace_ and nullspaceDim_
-    if (nullspaceType != "default vectors") {
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceType != "pre-computed", Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (no pre-computed null space). error.");
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceDim  == -1,             Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace dim == -1). error.");
-      TEUCHOS_TEST_FOR_EXCEPTION(nullspaceVec  == NULL,           Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no valid nullspace (nullspace == NULL). You have to provide a valid fine-level nullspace in \'null space: vectors\'");
-
-      nullspaceDim_ = nullspaceDim;
-      nullspace_    = nullspaceVec;
-    }
-
-    Teuchos::RCP<NullspaceFactory> nspFact = Teuchos::rcp(new NullspaceFactory("Nullspace"));
-    nspFact->SetFactory("Nullspace", PtentFact);
-
-
-    // Stash coordinates
-    xcoord_ = xcoord;
-    ycoord_ = ycoord;
-    zcoord_ = zcoord;
-
+    {
+      // Merge level-specific parameters with global parameters. level-specific parameters takes precedence.
+      // TODO: unit-test this part alone
 
+      ParameterList levelSmootherParam = GetMLSubList(paramList, "smoother", levelID);  // copy
+      MergeParameterList(*topLevelSmootherParam, levelSmootherParam, false);            /* false = do no overwrite levelSmootherParam parameters by topLevelSmootherParam parameters */
+      // std::cout << std::endl << "Merged List for level  " << levelID << std::endl;
+      // std::cout << levelSmootherParam << std::endl;
 
-    //
-    // Hierarchy + FactoryManager
-    //
+      RCP<SmootherFactory> smootherFact = GetSmootherFactory(levelSmootherParam, Teuchos::null);  // TODO: missing AFact input arg.
 
-    // Hierarchy options
-    this->numDesiredLevel_ = maxLevels;
-    this->maxCoarseSize_   = maxCoarseSize;
+      manager->SetFactory("Smoother", smootherFact);
+    }
 
     //
-    // Coarse Smoother
+    // Misc
     //
-    ParameterList& coarseList = paramList.sublist("coarse: list");
-    // check whether coarse solver is set properly. If not, set default coarse solver.
-    if (!coarseList.isParameter("smoother: type"))
-      coarseList.set("smoother: type", "Amesos-KLU"); // set default coarse solver according to ML 5.0 guide
-    RCP<SmootherFactory> coarseFact = GetSmootherFactory(coarseList, Teuchos::null);
-
-    // Smoothers Top Level Parameters
-
-    RCP<ParameterList> topLevelSmootherParam = ExtractSetOfParameters(paramList, "smoother");
-
-    //
-
-    // Prepare factory managers
-    // TODO: smootherFact can be reuse accross level if same parameters/no specific parameterList
-
-    for (int levelID=0; levelID < maxLevels; levelID++) {
-
-      //
-      // Level FactoryManager
-      //
-
-      RCP<FactoryManager> manager = rcp(new FactoryManager());
-      if (setKokkosRefactor)
-        manager->SetKokkosRefactor(useKokkosRefactor);
 
-      //
-      // Smoothers
-      //
-
-      {
-        // Merge level-specific parameters with global parameters. level-specific parameters takes precedence.
-        // TODO: unit-test this part alone
-
-        ParameterList levelSmootherParam = GetMLSubList(paramList, "smoother", levelID); // copy
-        MergeParameterList(*topLevelSmootherParam, levelSmootherParam, false); /* false = do no overwrite levelSmootherParam parameters by topLevelSmootherParam parameters */
-        // std::cout << std::endl << "Merged List for level  " << levelID << std::endl;
-        // std::cout << levelSmootherParam << std::endl;
-
-        RCP<SmootherFactory> smootherFact = GetSmootherFactory(levelSmootherParam, Teuchos::null); // TODO: missing AFact input arg.
-
-        manager->SetFactory("Smoother", smootherFact);
-      }
-
-      //
-      // Misc
-      //
-
-      manager->SetFactory("CoarseSolver", coarseFact); // TODO: should not be done in the loop
-      manager->SetFactory("Graph", dropFact);
-      manager->SetFactory("Aggregates", AggFact);
-      manager->SetFactory("DofsPerNode", dropFact);
-      manager->SetFactory("Ptent", PtentFact);
+    manager->SetFactory("CoarseSolver", coarseFact);  // TODO: should not be done in the loop
+    manager->SetFactory("Graph", dropFact);
+    manager->SetFactory("Aggregates", AggFact);
+    manager->SetFactory("DofsPerNode", dropFact);
+    manager->SetFactory("Ptent", PtentFact);
 
 #if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
     if (bDoRepartition == 1) {
       manager->SetFactory("A", RebalancedAFact);
       manager->SetFactory("P", RebalancedPFact);
       manager->SetFactory("R", RebalancedRFact);
-      manager->SetFactory("Nullspace",   RebalancedPFact);
-      manager->SetFactory("Importer",    RepartitionFact);
+      manager->SetFactory("Nullspace", RebalancedPFact);
+      manager->SetFactory("Importer", RepartitionFact);
     } else {
-#endif // #ifdef HAVE_MUELU_ISORROPIA
-      manager->SetFactory("Nullspace", nspFact); // use same nullspace factory throughout all multigrid levels
-      manager->SetFactory("A", AcFact);          // same RAP factory for all levels
-      manager->SetFactory("P", PFact);           // same prolongator and restrictor factories for all levels
-      manager->SetFactory("R", RFact);           // same prolongator and restrictor factories for all levels
+#endif                                            // #ifdef HAVE_MUELU_ISORROPIA
+      manager->SetFactory("Nullspace", nspFact);  // use same nullspace factory throughout all multigrid levels
+      manager->SetFactory("A", AcFact);           // same RAP factory for all levels
+      manager->SetFactory("P", PFact);            // same prolongator and restrictor factories for all levels
+      manager->SetFactory("R", RFact);            // same prolongator and restrictor factories for all levels
 #if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
     }
 #endif
 
-      this->AddFactoryManager(levelID, 1, manager);
-    } // for (level loop)
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy & H) const {
-    // if nullspace_ has already been extracted from ML parameter list
-    // make nullspace available for MueLu
-    if (nullspace_ != NULL) {
-      RCP<Level> fineLevel = H.GetLevel(0);
-      RCP<Operator> Op = fineLevel->Get<RCP<Operator> >("A");
-      RCP<Matrix>   A  = rcp_dynamic_cast<Matrix>(Op);
-      if (!A.is_null()) {
-        const RCP<const Map> rowMap = fineLevel->Get< RCP<Matrix> >("A")->getRowMap();
-        RCP<MultiVector> nullspace = MultiVectorFactory::Build(rowMap, nullspaceDim_, true);
-
-        for ( size_t i=0; i < Teuchos::as<size_t>(nullspaceDim_); i++) {
-          Teuchos::ArrayRCP<Scalar> nullspacei = nullspace->getDataNonConst(i);
-          const size_t              myLength   = nullspace->getLocalLength();
-
-          for (size_t j = 0; j < myLength; j++) {
-            nullspacei[j] = nullspace_[i*myLength + j];
-          }
+    this->AddFactoryManager(levelID, 1, manager);
+  }  // for (level loop)
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy& H) const {
+  // if nullspace_ has already been extracted from ML parameter list
+  // make nullspace available for MueLu
+  if (nullspace_ != NULL) {
+    RCP<Level> fineLevel = H.GetLevel(0);
+    RCP<Operator> Op     = fineLevel->Get<RCP<Operator> >("A");
+    RCP<Matrix> A        = rcp_dynamic_cast<Matrix>(Op);
+    if (!A.is_null()) {
+      const RCP<const Map> rowMap = fineLevel->Get<RCP<Matrix> >("A")->getRowMap();
+      RCP<MultiVector> nullspace  = MultiVectorFactory::Build(rowMap, nullspaceDim_, true);
+
+      for (size_t i = 0; i < Teuchos::as<size_t>(nullspaceDim_); i++) {
+        Teuchos::ArrayRCP<Scalar> nullspacei = nullspace->getDataNonConst(i);
+        const size_t myLength                = nullspace->getLocalLength();
+
+        for (size_t j = 0; j < myLength; j++) {
+          nullspacei[j] = nullspace_[i * myLength + j];
         }
-
-        fineLevel->Set("Nullspace", nullspace);
       }
+
+      fineLevel->Set("Nullspace", nullspace);
     }
+  }
 
-    // Do the same for coordinates
-    size_t num_coords = 0;
-    double * coordPTR[3];
-    if (xcoord_) {
-      coordPTR[0] = xcoord_;
+  // Do the same for coordinates
+  size_t num_coords = 0;
+  double* coordPTR[3];
+  if (xcoord_) {
+    coordPTR[0] = xcoord_;
+    num_coords++;
+    if (ycoord_) {
+      coordPTR[1] = ycoord_;
       num_coords++;
-      if (ycoord_) {
-        coordPTR[1] = ycoord_;
+      if (zcoord_) {
+        coordPTR[2] = zcoord_;
         num_coords++;
-        if (zcoord_) {
-          coordPTR[2] = zcoord_;
-          num_coords++;
-        }
       }
     }
-    if (num_coords){
-      Teuchos::RCP<Level> fineLevel = H.GetLevel(0);
-      Teuchos::RCP<Operator> Op = fineLevel->Get<RCP<Operator> >("A");
-      Teuchos::RCP<Matrix>   A  = rcp_dynamic_cast<Matrix>(Op);
-      if (!A.is_null()) {
-        const Teuchos::RCP<const Map> rowMap = fineLevel->Get< RCP<Matrix> >("A")->getRowMap();
-        Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(rowMap, num_coords, true);
-
-        for ( size_t i=0; i < num_coords; i++) {
-          Teuchos::ArrayRCP<Scalar> coordsi  = coordinates->getDataNonConst(i);
-          const size_t              myLength = coordinates->getLocalLength();
-          for (size_t j = 0; j < myLength; j++) {
-            coordsi[j] = coordPTR[i][j];
-          }
+  }
+  if (num_coords) {
+    Teuchos::RCP<Level> fineLevel = H.GetLevel(0);
+    Teuchos::RCP<Operator> Op     = fineLevel->Get<RCP<Operator> >("A");
+    Teuchos::RCP<Matrix> A        = rcp_dynamic_cast<Matrix>(Op);
+    if (!A.is_null()) {
+      const Teuchos::RCP<const Map> rowMap  = fineLevel->Get<RCP<Matrix> >("A")->getRowMap();
+      Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(rowMap, num_coords, true);
+
+      for (size_t i = 0; i < num_coords; i++) {
+        Teuchos::ArrayRCP<Scalar> coordsi = coordinates->getDataNonConst(i);
+        const size_t myLength             = coordinates->getLocalLength();
+        for (size_t j = 0; j < myLength; j++) {
+          coordsi[j] = coordPTR[i][j];
         }
-        fineLevel->Set("Coordinates",coordinates);
       }
+      fineLevel->Set("Coordinates", coordinates);
     }
-
-    HierarchyManager::SetupHierarchy(H);
   }
 
-  // TODO: code factorization with MueLu_ParameterListInterpreter.
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetSmootherFactory (const Teuchos::ParameterList & paramList,
-                      const RCP<FactoryBase> & AFact)
-  {
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    SC one = STS::one();
-
-    std::string type = "symmetric Gauss-Seidel"; // default
-
-    //
-    // Get 'type'
-    //
-
-//     //TODO: fix defaults!!
-
-//     // Default coarse grid smoother
-//     std::string type;
-//     if ("smoother" == "coarse") {
-// #if (defined(HAVE_MUELU_EPETRA) && defined( HAVE_MUELU_AMESOS)) || (defined(HAVE_MUELU_AMESOS2)) // FIXME: test is wrong (ex: compiled with Epetra&&Tpetra&&Amesos2 but without Amesos => error running Epetra problem)
-//       type = ""; // use default defined by AmesosSmoother or Amesos2Smoother
-// #else
-//       type = "symmetric Gauss-Seidel"; // use a sym Gauss-Seidel (with no damping) as fallback "coarse solver" (TODO: needs Ifpack(2))
-// #endif
-//     } else {
-//       // TODO: default smoother?
-//       type = "";
-//     }
-
-
-    if (paramList.isParameter("smoother: type")) type = paramList.get<std::string>("smoother: type");
-    TEUCHOS_TEST_FOR_EXCEPTION(type.empty(), Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no \"smoother: type\" in the smoother parameter list" << std::endl << paramList);
-
-    //
-    // Create the smoother prototype
-    //
-
-    RCP<SmootherPrototype> smooProto;
-    std::string ifpackType;
-    Teuchos::ParameterList smootherParamList;
-
-    if (type == "Jacobi" || type == "Gauss-Seidel" || type == "symmetric Gauss-Seidel") {
-      if (type == "symmetric Gauss-Seidel") type = "Symmetric Gauss-Seidel"; // FIXME
-
-      ifpackType = "RELAXATION";
-      smootherParamList.set("relaxation: type", type);
-
-      MUELU_COPY_PARAM(paramList, "smoother: sweeps",            int,   2, smootherParamList, "relaxation: sweeps");
-      MUELU_COPY_PARAM(paramList, "smoother: damping factor", Scalar, one, smootherParamList, "relaxation: damping factor");
-
-      smooProto = rcp( new TrilinosSmoother(ifpackType, smootherParamList, 0) );
-      smooProto->SetFactory("A", AFact);
-
-    } else if (type == "Chebyshev" || type == "MLS") {
-
-      ifpackType = "CHEBYSHEV";
-
-      MUELU_COPY_PARAM(paramList, "smoother: sweeps",          int, 2,     smootherParamList, "chebyshev: degree");
-      if (paramList.isParameter("smoother: MLS alpha")) {
-        MUELU_COPY_PARAM(paramList, "smoother: MLS alpha", double, 20, smootherParamList, "chebyshev: ratio eigenvalue");
-      } else {
-        MUELU_COPY_PARAM(paramList, "smoother: Chebyshev alpha", double, 20, smootherParamList, "chebyshev: ratio eigenvalue");
-      }
-
-
-      smooProto = rcp( new TrilinosSmoother(ifpackType, smootherParamList, 0) );
-      smooProto->SetFactory("A", AFact);
+  HierarchyManager::SetupHierarchy(H);
+}
+
+// TODO: code factorization with MueLu_ParameterListInterpreter.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetSmootherFactory(const Teuchos::ParameterList& paramList,
+                       const RCP<FactoryBase>& AFact) {
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  SC one = STS::one();
+
+  std::string type = "symmetric Gauss-Seidel";  // default
+
+  //
+  // Get 'type'
+  //
+
+  //     //TODO: fix defaults!!
+
+  //     // Default coarse grid smoother
+  //     std::string type;
+  //     if ("smoother" == "coarse") {
+  // #if (defined(HAVE_MUELU_EPETRA) && defined( HAVE_MUELU_AMESOS)) || (defined(HAVE_MUELU_AMESOS2)) // FIXME: test is wrong (ex: compiled with Epetra&&Tpetra&&Amesos2 but without Amesos => error running Epetra problem)
+  //       type = ""; // use default defined by AmesosSmoother or Amesos2Smoother
+  // #else
+  //       type = "symmetric Gauss-Seidel"; // use a sym Gauss-Seidel (with no damping) as fallback "coarse solver" (TODO: needs Ifpack(2))
+  // #endif
+  //     } else {
+  //       // TODO: default smoother?
+  //       type = "";
+  //     }
+
+  if (paramList.isParameter("smoother: type")) type = paramList.get<std::string>("smoother: type");
+  TEUCHOS_TEST_FOR_EXCEPTION(type.empty(), Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no \"smoother: type\" in the smoother parameter list" << std::endl
+                                                                                                                                                               << paramList);
+
+  //
+  // Create the smoother prototype
+  //
+
+  RCP<SmootherPrototype> smooProto;
+  std::string ifpackType;
+  Teuchos::ParameterList smootherParamList;
+
+  if (type == "Jacobi" || type == "Gauss-Seidel" || type == "symmetric Gauss-Seidel") {
+    if (type == "symmetric Gauss-Seidel") type = "Symmetric Gauss-Seidel";  // FIXME
+
+    ifpackType = "RELAXATION";
+    smootherParamList.set("relaxation: type", type);
+
+    MUELU_COPY_PARAM(paramList, "smoother: sweeps", int, 2, smootherParamList, "relaxation: sweeps");
+    MUELU_COPY_PARAM(paramList, "smoother: damping factor", Scalar, one, smootherParamList, "relaxation: damping factor");
+
+    smooProto = rcp(new TrilinosSmoother(ifpackType, smootherParamList, 0));
+    smooProto->SetFactory("A", AFact);
+
+  } else if (type == "Chebyshev" || type == "MLS") {
+    ifpackType = "CHEBYSHEV";
+
+    MUELU_COPY_PARAM(paramList, "smoother: sweeps", int, 2, smootherParamList, "chebyshev: degree");
+    if (paramList.isParameter("smoother: MLS alpha")) {
+      MUELU_COPY_PARAM(paramList, "smoother: MLS alpha", double, 20, smootherParamList, "chebyshev: ratio eigenvalue");
+    } else {
+      MUELU_COPY_PARAM(paramList, "smoother: Chebyshev alpha", double, 20, smootherParamList, "chebyshev: ratio eigenvalue");
+    }
 
-    } else if (type == "Hiptmair") {
-      ifpackType = "HIPTMAIR";
-      std::string subSmootherType = "Chebyshev";
-      if (paramList.isParameter("subsmoother: type"))
-        subSmootherType = paramList.get<std::string>("subsmoother: type");
-      std::string subSmootherIfpackType;
-      if (subSmootherType == "Chebyshev")
-        subSmootherIfpackType = "CHEBYSHEV";
-      else if (subSmootherType == "Jacobi" || subSmootherType == "Gauss-Seidel" || subSmootherType == "symmetric Gauss-Seidel") {
-        if (subSmootherType == "symmetric Gauss-Seidel") subSmootherType = "Symmetric Gauss-Seidel"; // FIXME
-        subSmootherIfpackType = "RELAXATION";
-      } else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << subSmootherType << "' not supported by MueLu.");
-
-      smootherParamList.set("hiptmair: smoother type 1", subSmootherIfpackType);
-      smootherParamList.set("hiptmair: smoother type 2", subSmootherIfpackType);
-
-      auto smoother1ParamList = smootherParamList.sublist("hiptmair: smoother list 1");
-      auto smoother2ParamList = smootherParamList.sublist("hiptmair: smoother list 2");
-
-      if (subSmootherType == "Chebyshev") {
-        MUELU_COPY_PARAM(paramList, "subsmoother: edge sweeps", int, 2, smoother1ParamList, "chebyshev: degree");
-        MUELU_COPY_PARAM(paramList, "subsmoother: node sweeps", int, 2, smoother2ParamList, "chebyshev: degree");
-
-        MUELU_COPY_PARAM(paramList, "subsmoother: Chebyshev", double, 20, smoother1ParamList, "chebyshev: ratio eigenvalue");
-        MUELU_COPY_PARAM(paramList, "subsmoother: Chebyshev", double, 20, smoother2ParamList, "chebyshev: ratio eigenvalue");
-      } else {
-        MUELU_COPY_PARAM(paramList, "subsmoother: edge sweeps", int, 2, smoother1ParamList, "relaxation: sweeps");
-        MUELU_COPY_PARAM(paramList, "subsmoother: node sweeps", int, 2, smoother2ParamList, "relaxation: sweeps");
-
-        MUELU_COPY_PARAM(paramList, "subsmoother: SGS damping factor", double, 0.8, smoother2ParamList, "relaxation: damping factor");
-      }
+    smooProto = rcp(new TrilinosSmoother(ifpackType, smootherParamList, 0));
+    smooProto->SetFactory("A", AFact);
+
+  } else if (type == "Hiptmair") {
+    ifpackType                  = "HIPTMAIR";
+    std::string subSmootherType = "Chebyshev";
+    if (paramList.isParameter("subsmoother: type"))
+      subSmootherType = paramList.get<std::string>("subsmoother: type");
+    std::string subSmootherIfpackType;
+    if (subSmootherType == "Chebyshev")
+      subSmootherIfpackType = "CHEBYSHEV";
+    else if (subSmootherType == "Jacobi" || subSmootherType == "Gauss-Seidel" || subSmootherType == "symmetric Gauss-Seidel") {
+      if (subSmootherType == "symmetric Gauss-Seidel") subSmootherType = "Symmetric Gauss-Seidel";  // FIXME
+      subSmootherIfpackType = "RELAXATION";
+    } else
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << subSmootherType << "' not supported by MueLu.");
+
+    smootherParamList.set("hiptmair: smoother type 1", subSmootherIfpackType);
+    smootherParamList.set("hiptmair: smoother type 2", subSmootherIfpackType);
+
+    auto smoother1ParamList = smootherParamList.sublist("hiptmair: smoother list 1");
+    auto smoother2ParamList = smootherParamList.sublist("hiptmair: smoother list 2");
+
+    if (subSmootherType == "Chebyshev") {
+      MUELU_COPY_PARAM(paramList, "subsmoother: edge sweeps", int, 2, smoother1ParamList, "chebyshev: degree");
+      MUELU_COPY_PARAM(paramList, "subsmoother: node sweeps", int, 2, smoother2ParamList, "chebyshev: degree");
+
+      MUELU_COPY_PARAM(paramList, "subsmoother: Chebyshev", double, 20, smoother1ParamList, "chebyshev: ratio eigenvalue");
+      MUELU_COPY_PARAM(paramList, "subsmoother: Chebyshev", double, 20, smoother2ParamList, "chebyshev: ratio eigenvalue");
+    } else {
+      MUELU_COPY_PARAM(paramList, "subsmoother: edge sweeps", int, 2, smoother1ParamList, "relaxation: sweeps");
+      MUELU_COPY_PARAM(paramList, "subsmoother: node sweeps", int, 2, smoother2ParamList, "relaxation: sweeps");
 
+      MUELU_COPY_PARAM(paramList, "subsmoother: SGS damping factor", double, 0.8, smoother2ParamList, "relaxation: damping factor");
+    }
 
-      smooProto = rcp( new TrilinosSmoother(ifpackType, smootherParamList, 0) );
-      smooProto->SetFactory("A", AFact);
+    smooProto = rcp(new TrilinosSmoother(ifpackType, smootherParamList, 0));
+    smooProto->SetFactory("A", AFact);
 
-    } else if (type == "IFPACK") { // TODO: this option is not described in the ML Guide v5.0
+  } else if (type == "IFPACK") {  // TODO: this option is not described in the ML Guide v5.0
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
-      ifpackType = paramList.get<std::string>("smoother: ifpack type");
-
-      if (ifpackType == "ILU") {
-        // TODO fix this (type mismatch double vs. int)
-        //MUELU_COPY_PARAM(paramList, "smoother: ifpack level-of-fill", double /*int*/, 0.0 /*2*/,  smootherParamList, "fact: level-of-fill");
-        if (paramList.isParameter("smoother: ifpack level-of-fill"))
-          smootherParamList.set("fact: level-of-fill", Teuchos::as<int>(paramList.get<double>("smoother: ifpack level-of-fill")));
-        else smootherParamList.set("fact: level-of-fill", as<int>(0));
-
-        MUELU_COPY_PARAM(paramList, "smoother: ifpack overlap",       int, 2,  smootherParamList, "partitioner: overlap");
-
-        // TODO change to TrilinosSmoother as soon as Ifpack2 supports all preconditioners from Ifpack
-        smooProto =
-          MueLu::GetIfpackSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node> (ifpackType,
-                                                                               smootherParamList,
-                                                                               paramList.get<int> ("smoother: ifpack overlap"));
-        smooProto->SetFactory("A", AFact);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown ML smoother type " + type + " (IFPACK) not supported by MueLu. Only ILU is supported.");
-      }
-#else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: MueLu compiled without Ifpack support");
-#endif
+    ifpackType = paramList.get<std::string>("smoother: ifpack type");
 
-    } else if (type.length() > strlen("Amesos") && type.substr(0, strlen("Amesos")) == "Amesos") {  /* catch Amesos-* */
-      std::string solverType = type.substr(strlen("Amesos")+1);  /* ("Amesos-KLU" -> "KLU") */
-
-      // Validator: following upper/lower case is what is allowed by ML
-      bool valid = false;
-      const int  validatorSize = 5;
-      std::string validator[validatorSize] = {"Superlu", "Superludist", "KLU", "UMFPACK", "MUMPS"}; /* TODO: should "" be allowed? */
-      for (int i=0; i < validatorSize; i++) { if (validator[i] == solverType) valid = true; }
-      TEUCHOS_TEST_FOR_EXCEPTION(!valid, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << type << "' not supported.");
+    if (ifpackType == "ILU") {
+      // TODO fix this (type mismatch double vs. int)
+      // MUELU_COPY_PARAM(paramList, "smoother: ifpack level-of-fill", double /*int*/, 0.0 /*2*/,  smootherParamList, "fact: level-of-fill");
+      if (paramList.isParameter("smoother: ifpack level-of-fill"))
+        smootherParamList.set("fact: level-of-fill", Teuchos::as<int>(paramList.get<double>("smoother: ifpack level-of-fill")));
+      else
+        smootherParamList.set("fact: level-of-fill", as<int>(0));
 
-      // FIXME: MueLu should accept any Upper/Lower case. Not the case for the moment
-      std::transform(solverType.begin()+1, solverType.end(), solverType.begin()+1, ::tolower);
+      MUELU_COPY_PARAM(paramList, "smoother: ifpack overlap", int, 2, smootherParamList, "partitioner: overlap");
 
-      smooProto = Teuchos::rcp( new DirectSolver(solverType, Teuchos::ParameterList()) );
+      // TODO change to TrilinosSmoother as soon as Ifpack2 supports all preconditioners from Ifpack
+      smooProto =
+          MueLu::GetIfpackSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>(ifpackType,
+                                                                              smootherParamList,
+                                                                              paramList.get<int>("smoother: ifpack overlap"));
       smooProto->SetFactory("A", AFact);
-
     } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown ML smoother type " + type + " (IFPACK) not supported by MueLu. Only ILU is supported.");
+    }
+#else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: MueLu compiled without Ifpack support");
+#endif
 
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << type << "' not supported by MueLu.");
+  } else if (type.length() > strlen("Amesos") && type.substr(0, strlen("Amesos")) == "Amesos") { /* catch Amesos-* */
+    std::string solverType = type.substr(strlen("Amesos") + 1);                                  /* ("Amesos-KLU" -> "KLU") */
 
+    // Validator: following upper/lower case is what is allowed by ML
+    bool valid                           = false;
+    const int validatorSize              = 5;
+    std::string validator[validatorSize] = {"Superlu", "Superludist", "KLU", "UMFPACK", "MUMPS"}; /* TODO: should "" be allowed? */
+    for (int i = 0; i < validatorSize; i++) {
+      if (validator[i] == solverType) valid = true;
     }
-    TEUCHOS_TEST_FOR_EXCEPTION(smooProto == Teuchos::null, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no smoother prototype. fatal error.");
-
-    //
-    // Create the smoother factory
-    //
+    TEUCHOS_TEST_FOR_EXCEPTION(!valid, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << type << "' not supported.");
 
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory() );
+    // FIXME: MueLu should accept any Upper/Lower case. Not the case for the moment
+    std::transform(solverType.begin() + 1, solverType.end(), solverType.begin() + 1, ::tolower);
 
-    // Set parameters of the smoother factory
-    MUELU_READ_PARAM(paramList, "smoother: pre or post", std::string, "both", preOrPost);
-    if (preOrPost == "both") {
-      SmooFact->SetSmootherPrototypes(smooProto, smooProto);
-    } else if (preOrPost == "pre") {
-      SmooFact->SetSmootherPrototypes(smooProto, Teuchos::null);
-    } else if (preOrPost == "post") {
-      SmooFact->SetSmootherPrototypes(Teuchos::null, smooProto);
-    }
+    smooProto = Teuchos::rcp(new DirectSolver(solverType, Teuchos::ParameterList()));
+    smooProto->SetFactory("A", AFact);
 
-    return SmooFact;
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: unknown smoother type. '" << type << "' not supported by MueLu.");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "Transfer factory is not derived from TwoLevelFactoryBase. Since transfer factories will be handled by the RAPFactory they have to be derived from TwoLevelFactoryBase!");
-    TransferFacts_.push_back(factory);
+  TEUCHOS_TEST_FOR_EXCEPTION(smooProto == Teuchos::null, Exceptions::RuntimeError, "MueLu::MLParameterListInterpreter: no smoother prototype. fatal error.");
+
+  //
+  // Create the smoother factory
+  //
+
+  RCP<SmootherFactory> SmooFact = rcp(new SmootherFactory());
+
+  // Set parameters of the smoother factory
+  MUELU_READ_PARAM(paramList, "smoother: pre or post", std::string, "both", preOrPost);
+  if (preOrPost == "both") {
+    SmooFact->SetSmootherPrototypes(smooProto, smooProto);
+  } else if (preOrPost == "pre") {
+    SmooFact->SetSmootherPrototypes(smooProto, Teuchos::null);
+  } else if (preOrPost == "post") {
+    SmooFact->SetSmootherPrototypes(Teuchos::null, smooProto);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NumTransferFactories() const {
-    return TransferFacts_.size();
-  }
+  return SmooFact;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator & Op) const {
-    try {
-      Matrix& A = dynamic_cast<Matrix&>(Op);
-      if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blksize_))
-        this->GetOStream(Warnings0) << "Setting matrix block size to " << blksize_ << " (value of the parameter in the list) "
-            << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "Transfer factory is not derived from TwoLevelFactoryBase. Since transfer factories will be handled by the RAPFactory they have to be derived from TwoLevelFactoryBase!");
+  TransferFacts_.push_back(factory);
+}
 
-      A.SetFixedBlockSize(blksize_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NumTransferFactories() const {
+  return TransferFacts_.size();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MLParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator& Op) const {
+  try {
+    Matrix& A = dynamic_cast<Matrix&>(Op);
+    if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blksize_))
+      this->GetOStream(Warnings0) << "Setting matrix block size to " << blksize_ << " (value of the parameter in the list) "
+                                  << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl;
+
+    A.SetFixedBlockSize(blksize_);
 
 #ifdef HAVE_MUELU_DEBUG
-      MatrixUtils::checkLocalRowMapMatchesColMap(A);
-#endif // HAVE_MUELU_DEBUG
+    MatrixUtils::checkLocalRowMapMatchesColMap(A);
+#endif  // HAVE_MUELU_DEBUG
 
-    } catch (std::bad_cast&) {
-      this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
-    }
+  } catch (std::bad_cast&) {
+    this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
   }
+}
 
-} // namespace MueLu
-
+}  // namespace MueLu
 
 #define MUELU_MLPARAMETERLISTINTERPRETER_SHORT
 #endif /* MUELU_MLPARAMETERLISTINTERPRETER_DEF_HPP */
 
-//TODO: see if it can be factorized with ML interpreter (ex: generation of Ifpack param list)
+// TODO: see if it can be factorized with ML interpreter (ex: generation of Ifpack param list)
diff --git a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter.cpp b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter.cpp
index fe96d582432b..6d646b407a4d 100644
--- a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter.cpp
+++ b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter.cpp
@@ -51,44 +51,44 @@
 
 namespace MueLu {
 
-  size_t LevenshteinDistance(const char* s, size_t len_s, const char* t, size_t len_t) {
-    // degenerate cases
-    if (len_s == 0)                             return len_t;
-    if (len_t == 0)                             return len_s;
-    if (!strncmp(s, t, std::min(len_s, len_t))) return 0;
+size_t LevenshteinDistance(const char* s, size_t len_s, const char* t, size_t len_t) {
+  // degenerate cases
+  if (len_s == 0) return len_t;
+  if (len_t == 0) return len_s;
+  if (!strncmp(s, t, std::min(len_s, len_t))) return 0;
 
-    // create two work vectors of integer distances
-    size_t len = len_t + 1;
-    std::vector<size_t> v0(len);
-    std::vector<size_t> v1(len);
+  // create two work vectors of integer distances
+  size_t len = len_t + 1;
+  std::vector<size_t> v0(len);
+  std::vector<size_t> v1(len);
 
-    // initialize v0 (the previous row of distances)
-    // this row is A[0][i]: edit distance for an empty s
-    // the distance is just the number of characters to delete from t
-    for (size_t i = 0; i < len; i++)
-      v0[i] = i;
+  // initialize v0 (the previous row of distances)
+  // this row is A[0][i]: edit distance for an empty s
+  // the distance is just the number of characters to delete from t
+  for (size_t i = 0; i < len; i++)
+    v0[i] = i;
 
-    for (size_t i = 0; i < len_s; i++) {
-      // calculate v1 (current row distances) from the previous row v0
+  for (size_t i = 0; i < len_s; i++) {
+    // calculate v1 (current row distances) from the previous row v0
 
-      // first element of v1 is A[i+1][0]
-      //   edit distance is delete (i+1) chars from s to match empty t
-      v1[0] = i + 1;
+    // first element of v1 is A[i+1][0]
+    //   edit distance is delete (i+1) chars from s to match empty t
+    v1[0] = i + 1;
 
-      // use formula to fill in the rest of the row
-      for (size_t j = 0; j < len_t; j++) {
-        size_t cost = (s[i] == t[j]) ? 0 : 1;
-        v1[j+1] = std::min(v1[j] + 1,
-                  std::min(v0[j + 1] + 1,
-                           v0[j] + cost));
-      }
-
-      // copy v1 (current row) to v0 (previous row) for next iteration
-      for (size_t j = 0; j < len; j++)
-        v0[j] = v1[j];
+    // use formula to fill in the rest of the row
+    for (size_t j = 0; j < len_t; j++) {
+      size_t cost = (s[i] == t[j]) ? 0 : 1;
+      v1[j + 1]   = std::min(v1[j] + 1,
+                             std::min(v0[j + 1] + 1,
+                                      v0[j] + cost));
     }
 
-    return v1[len_t];
+    // copy v1 (current row) to v0 (previous row) for next iteration
+    for (size_t j = 0; j < len; j++)
+      v0[j] = v1[j];
   }
 
+  return v1[len_t];
 }
+
+}  // namespace MueLu
diff --git a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_decl.hpp b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_decl.hpp
index 395dc4231ffc..d4554d8016db 100644
--- a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_decl.hpp
+++ b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_decl.hpp
@@ -99,7 +99,6 @@
 #include "MueLu_SingleLevelMatlabFactory_fwd.hpp"
 #endif
 
-
 #include "MueLu_CoalesceDropFactory_kokkos_fwd.hpp"
 #include "MueLu_NullspaceFactory_kokkos_fwd.hpp"
 #include "MueLu_SaPFactory_kokkos_fwd.hpp"
@@ -113,184 +112,180 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ParameterListInterpreter :
-    public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ParameterListInterpreter : public HierarchyManager<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PARAMETERLISTINTERPRETER_SHORT
 #include "MueLu_UseShortNames.hpp"
-    typedef std::pair<std::string, const FactoryBase*> keep_pair;
+  typedef std::pair<std::string, const FactoryBase*> keep_pair;
 
-  public:
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-  protected:
-    /*! @brief Empty constructor
-     *
-     *  Constructor for derived classes
-     */
-    ParameterListInterpreter() {
-      factFact_ = Teuchos::null;
-      facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
-    }
+ protected:
+  /*! @brief Empty constructor
+   *
+   *  Constructor for derived classes
+   */
+  ParameterListInterpreter() {
+    factFact_   = Teuchos::null;
+    facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
+  }
 
-  public:
-    /*! @brief Constructor that accepts a user-provided ParameterList.
+ public:
+  /*! @brief Constructor that accepts a user-provided ParameterList.
 
-        Constructor for parameter list interpreter which directly interprets Teuchos::ParameterLists
+      Constructor for parameter list interpreter which directly interprets Teuchos::ParameterLists
 
-        @details The parameter list can be either in the easy parameter list format or in the factory driven parameter list format.
+      @details The parameter list can be either in the easy parameter list format or in the factory driven parameter list format.
 
-        @param[in] paramList (Teuchos::ParameterList): ParameterList containing the MueLu parameters
-        @param[in] comm  (RCP<Teuchos::Comm<int> >): Optional RCP of a Teuchos communicator  (default: Teuchos::null)
-        @param[in] factFact  (RCP<FactoryFactory>): Optional parameter allowing to define user-specific factory interpreters for user-specific extensions of the XML interface. (default: Teuchos::null)
-        @param[in] facadeFact (RCP<FacadeFactory>): Optional parameter containing a FacadeFactory class. The user can register its own facade classes in the FacadeFactory and provide it to the ParameterListInterpreter. (default: Teuchos::null, means, only standard FacadeClass that come with MueLu are available)
+      @param[in] paramList (Teuchos::ParameterList): ParameterList containing the MueLu parameters
+      @param[in] comm  (RCP<Teuchos::Comm<int> >): Optional RCP of a Teuchos communicator  (default: Teuchos::null)
+      @param[in] factFact  (RCP<FactoryFactory>): Optional parameter allowing to define user-specific factory interpreters for user-specific extensions of the XML interface. (default: Teuchos::null)
+      @param[in] facadeFact (RCP<FacadeFactory>): Optional parameter containing a FacadeFactory class. The user can register its own facade classes in the FacadeFactory and provide it to the ParameterListInterpreter. (default: Teuchos::null, means, only standard FacadeClass that come with MueLu are available)
 
-     */
-    ParameterListInterpreter(Teuchos::ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::null, Teuchos::RCP<FactoryFactory> factFact = Teuchos::null, Teuchos::RCP<FacadeClassFactory> facadeFact = Teuchos::null);
+   */
+  ParameterListInterpreter(Teuchos::ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::null, Teuchos::RCP<FactoryFactory> factFact = Teuchos::null, Teuchos::RCP<FacadeClassFactory> facadeFact = Teuchos::null);
 
-    /*! @brief Constructor that reads parameters from an XML file.
+  /*! @brief Constructor that reads parameters from an XML file.
 
-        XML options are converted to ParameterList entries by Teuchos.
+      XML options are converted to ParameterList entries by Teuchos.
 
-        @param[in] xmlFileName (std::string): XML file to read
-        @param[in] comm  (Teuchos::Comm<int>): Teuchos communicator
-        @param[in] factFact  (RCP<FactoryFactory>): Optional parameter allowing to define user-specific factory interpreters for user-specific extensions of the XML interface. (default: Teuchos::null)
-        @param[in] facadeFact (RCP<FacadeFactory>): Optional parameter containing a FacadeFactory class. The user can register its own facade classes in the FacadeFactory and provide it to the ParameterListInterpreter. (default: Teuchos::null, means, only standard FacadeClass that come with MueLu are available)
+      @param[in] xmlFileName (std::string): XML file to read
+      @param[in] comm  (Teuchos::Comm<int>): Teuchos communicator
+      @param[in] factFact  (RCP<FactoryFactory>): Optional parameter allowing to define user-specific factory interpreters for user-specific extensions of the XML interface. (default: Teuchos::null)
+      @param[in] facadeFact (RCP<FacadeFactory>): Optional parameter containing a FacadeFactory class. The user can register its own facade classes in the FacadeFactory and provide it to the ParameterListInterpreter. (default: Teuchos::null, means, only standard FacadeClass that come with MueLu are available)
 
-    */
-    ParameterListInterpreter(const std::string& xmlFileName, const Teuchos::Comm<int>& comm, Teuchos::RCP<FactoryFactory> factFact = Teuchos::null, Teuchos::RCP<FacadeClassFactory> facadeFact = Teuchos::null);
+  */
+  ParameterListInterpreter(const std::string& xmlFileName, const Teuchos::Comm<int>& comm, Teuchos::RCP<FactoryFactory> factFact = Teuchos::null, Teuchos::RCP<FacadeClassFactory> facadeFact = Teuchos::null);
 
-    //! Destructor.
-    virtual ~ParameterListInterpreter() { }
+  //! Destructor.
+  virtual ~ParameterListInterpreter() {}
 
-    //@}
+  //@}
 
-    /*! @brief Set parameter list for Parameter list interpreter.
+  /*! @brief Set parameter list for Parameter list interpreter.
 
-       The routine checks whether it is a parameter list in the easy parameter format or the more advanced factory-based parameter format and calls the corresponding interpreter routine.
+     The routine checks whether it is a parameter list in the easy parameter format or the more advanced factory-based parameter format and calls the corresponding interpreter routine.
 
-       When finished, the parameter list is set that will used by the hierarchy build phase.
+     When finished, the parameter list is set that will used by the hierarchy build phase.
 
-       This method includes validation and some pre-parsing of the list for:
-           - verbosity level
-           - data to export
-           - cycle type
-           - max coarse size
-           - max levels
-           - number of equations
+     This method includes validation and some pre-parsing of the list for:
+         - verbosity level
+         - data to export
+         - cycle type
+         - max coarse size
+         - max levels
+         - number of equations
 
-       @param[in] paramList: ParameterList containing the MueLu parameters.
-    */
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+     @param[in] paramList: ParameterList containing the MueLu parameters.
+  */
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //! Call the SetupHierarchy routine from the HiearchyManager object.
-    void SetupHierarchy(Hierarchy& H) const;
+  //! Call the SetupHierarchy routine from the HiearchyManager object.
+  void SetupHierarchy(Hierarchy& H) const;
 
-  private:
-    //! Setup Operator object
-    virtual void SetupOperator(Operator& A) const;
+ private:
+  //! Setup Operator object
+  virtual void SetupOperator(Operator& A) const;
 
-    int       blockSize_;     ///< block size of matrix (fixed block size)
-    CycleType Cycle_;         ///< multigrid cycle type (V-cycle or W-cycle)
-    int       WCycleStartLevel_; ///< in case of W-cycle, level on which cycle should start
-    double    scalingFactor_; ///< prolongator scaling factor
-    GlobalOrdinal dofOffset_; ///< global offset variable describing offset of DOFs in operator
+  int blockSize_;            ///< block size of matrix (fixed block size)
+  CycleType Cycle_;          ///< multigrid cycle type (V-cycle or W-cycle)
+  int WCycleStartLevel_;     ///< in case of W-cycle, level on which cycle should start
+  double scalingFactor_;     ///< prolongator scaling factor
+  GlobalOrdinal dofOffset_;  ///< global offset variable describing offset of DOFs in operator
 
-    //! Easy interpreter stuff
-    //@{
-    // These three variables are only needed to print out proper [default]
-    bool changedPRrebalance_;
-    bool changedPRViaCopyrebalance_;
-    bool changedImplicitTranspose_;
+  //! Easy interpreter stuff
+  //@{
+  // These three variables are only needed to print out proper [default]
+  bool changedPRrebalance_;
+  bool changedPRViaCopyrebalance_;
+  bool changedImplicitTranspose_;
 
-    void SetEasyParameterList(const Teuchos::ParameterList& paramList);
-    void Validate(const Teuchos::ParameterList& paramList) const;
+  void SetEasyParameterList(const Teuchos::ParameterList& paramList);
+  void Validate(const Teuchos::ParameterList& paramList) const;
 
-    void UpdateFactoryManager(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                              int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                            int levelID, std::vector<keep_pair>& keeps) const;
 
-    // "Generic components" for UpdateFactoryManager
-    void UpdateFactoryManager_Smoothers(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                        int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_CoarseSolvers(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                            int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Aggregation_TentativeP(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                                     int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Restriction(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                          int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_RAP(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                  int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Coordinates(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+  // "Generic components" for UpdateFactoryManager
+  void UpdateFactoryManager_Smoothers(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                      int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_CoarseSolvers(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
                                           int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Repartition(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                          int levelID, std::vector<keep_pair>& keeps, RCP<Factory> & nullSpaceFactory) const;
-    void UpdateFactoryManager_LowPrecision(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                                           int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Nullspace(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                        int levelID, std::vector<keep_pair>& keeps, RCP<Factory> & nullSpaceFactory) const;
-    void UpdateFactoryManager_BlockNumber(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList,
-                                          FactoryManager& manager,int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_LocalOrdinalTransfer(const std::string& VarName, const std::string& multigridAlgo, Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList,
-                                           FactoryManager& manager,int levelID, std::vector<keep_pair>& keeps) const;
-
-    // Algorithm-specific components for UpdateFactoryManager
-    void UpdateFactoryManager_SemiCoarsen(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                          int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_PCoarsen(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+  void UpdateFactoryManager_Aggregation_TentativeP(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                                   int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Restriction(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                        int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_RAP(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Coordinates(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                        int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Repartition(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                        int levelID, std::vector<keep_pair>& keeps, RCP<Factory>& nullSpaceFactory) const;
+  void UpdateFactoryManager_LowPrecision(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                         int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Nullspace(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                      int levelID, std::vector<keep_pair>& keeps, RCP<Factory>& nullSpaceFactory) const;
+  void UpdateFactoryManager_BlockNumber(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList,
+                                        FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_LocalOrdinalTransfer(const std::string& VarName, const std::string& multigridAlgo, Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList,
+                                                 FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const;
+
+  // Algorithm-specific components for UpdateFactoryManager
+  void UpdateFactoryManager_SemiCoarsen(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                        int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_PCoarsen(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                     int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_SA(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                               int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Reitzinger(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
                                        int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_SA(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+  void UpdateFactoryManager_Emin(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
                                  int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Reitzinger(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                         int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Emin(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+  void UpdateFactoryManager_PG(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                               int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Replicate(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                      int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Combine(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
+                                    int levelID, std::vector<keep_pair>& keeps) const;
+  void UpdateFactoryManager_Matlab(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
                                    int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_PG(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                 int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Replicate(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                 int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Combine(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                 int levelID, std::vector<keep_pair>& keeps) const;
-    void UpdateFactoryManager_Matlab(Teuchos::ParameterList& paramList, const Teuchos::ParameterList& defaultList, FactoryManager& manager,
-                                     int levelID, std::vector<keep_pair>& keeps) const;
-
-
-
 
-    bool useCoordinates_;
-    bool useBlockNumber_;
-    bool useKokkos_;
-    //@}
+  bool useCoordinates_;
+  bool useBlockNumber_;
+  bool useKokkos_;
+  //@}
 
-    //! Factory interpreter stuff
-    // TODO:
-    // - parameter list validator
-    // - SetParameterList
-    // - Set/Get directly Level manager
-    // - build per level
-    // - comments/docs
-    // - use FactoryManager instead of FactoryMap
-    //@{
-    void SetFactoryParameterList(const Teuchos::ParameterList& paramList);
+  //! Factory interpreter stuff
+  // TODO:
+  // - parameter list validator
+  // - SetParameterList
+  // - Set/Get directly Level manager
+  // - build per level
+  // - comments/docs
+  // - use FactoryManager instead of FactoryMap
+  //@{
+  void SetFactoryParameterList(const Teuchos::ParameterList& paramList);
 
-    typedef std::map<std::string, RCP<const FactoryBase>  > FactoryMap; //TODO: remove this line
-    typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
+  typedef std::map<std::string, RCP<const FactoryBase> > FactoryMap;  // TODO: remove this line
+  typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
 
-    void BuildFactoryMap(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, FactoryMap& factoryMapOut, FactoryManagerMap& factoryManagers) const;
+  void BuildFactoryMap(const Teuchos::ParameterList& paramList, const FactoryMap& factoryMapIn, FactoryMap& factoryMapOut, FactoryManagerMap& factoryManagers) const;
 
-    //! Internal factory for factories
-    Teuchos::RCP<FactoryFactory> factFact_;
+  //! Internal factory for factories
+  Teuchos::RCP<FactoryFactory> factFact_;
 
-    //! FacadeClass factory
-    Teuchos::RCP<MueLu::FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> > facadeFact_;
+  //! FacadeClass factory
+  Teuchos::RCP<MueLu::FacadeClassFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> > facadeFact_;
 
-    //@}
-  };
+  //@}
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PARAMETERLISTINTERPRETER_SHORT
 #endif /* MUELU_PARAMETERLISTINTERPRETER_DECL_HPP */
diff --git a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_def.hpp b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_def.hpp
index d27cfdca878c..48af3ae2fda1 100644
--- a/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_def.hpp
+++ b/packages/muelu/src/Interface/MueLu_ParameterListInterpreter_def.hpp
@@ -129,1838 +129,1789 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ParameterListInterpreter(ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm, Teuchos::RCP<FactoryFactory> factFact, Teuchos::RCP<FacadeClassFactory> facadeFact) : factFact_(factFact) {
-    RCP<Teuchos::TimeMonitor> tM = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(std::string("MueLu: ParameterListInterpreter (ParameterList)"))));
-    if(facadeFact == Teuchos::null)
-      facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
-    else
-      facadeFact_ = facadeFact;
-
-    if (paramList.isParameter("xml parameter file")) {
-      std::string filename = paramList.get("xml parameter file", "");
-      if (filename.length() != 0) {
-        TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), Exceptions::RuntimeError, "xml parameter file requires a valid comm");
-
-        ParameterList paramList2 = paramList;
-        Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&paramList2), *comm);
-        SetParameterList(paramList2);
-
-      } else {
-        SetParameterList(paramList);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ParameterListInterpreter(ParameterList& paramList, Teuchos::RCP<const Teuchos::Comm<int> > comm, Teuchos::RCP<FactoryFactory> factFact, Teuchos::RCP<FacadeClassFactory> facadeFact)
+  : factFact_(factFact) {
+  RCP<Teuchos::TimeMonitor> tM = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(std::string("MueLu: ParameterListInterpreter (ParameterList)"))));
+  if (facadeFact == Teuchos::null)
+    facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
+  else
+    facadeFact_ = facadeFact;
+
+  if (paramList.isParameter("xml parameter file")) {
+    std::string filename = paramList.get("xml parameter file", "");
+    if (filename.length() != 0) {
+      TEUCHOS_TEST_FOR_EXCEPTION(comm.is_null(), Exceptions::RuntimeError, "xml parameter file requires a valid comm");
+
+      ParameterList paramList2 = paramList;
+      Teuchos::updateParametersFromXmlFileAndBroadcast(filename, Teuchos::Ptr<Teuchos::ParameterList>(&paramList2), *comm);
+      SetParameterList(paramList2);
 
     } else {
       SetParameterList(paramList);
     }
-  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ParameterListInterpreter(const std::string& xmlFileName, const Teuchos::Comm<int>& comm, Teuchos::RCP<FactoryFactory> factFact, Teuchos::RCP<FacadeClassFactory> facadeFact) : factFact_(factFact) {
-    RCP<Teuchos::TimeMonitor> tM = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(std::string("MueLu: ParameterListInterpreter (XML)"))));
-    if(facadeFact == Teuchos::null)
-      facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
-    else
-      facadeFact_ = facadeFact;
-
-    ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), comm);
+  } else {
     SetParameterList(paramList);
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
-    Cycle_     = Hierarchy::GetDefaultCycle();
-    WCycleStartLevel_ = Hierarchy::GetDefaultCycleStartLevel();
-    scalingFactor_= Teuchos::ScalarTraits<double>::one();
-    blockSize_ = 1;
-    dofOffset_ = 0;
-
-    if (paramList.isSublist("Hierarchy")) {
-      SetFactoryParameterList(paramList);
-
-    } else if (paramList.isParameter("MueLu preconditioner") == true) {
-      this->GetOStream(Runtime0) << "Use facade class: " << paramList.get<std::string>("MueLu preconditioner")  << std::endl;
-      Teuchos::RCP<ParameterList> pp = facadeFact_->SetParameterList(paramList);
-      SetFactoryParameterList(*pp);
-
-    }  else {
-      // The validator doesn't work correctly for non-serializable data (Hint: template parameters), so strip it out
-      ParameterList serialList, nonSerialList;
-
-      ExtractNonSerializableData(paramList, serialList, nonSerialList);
-      Validate(serialList);
-      SetEasyParameterList(paramList);
-    }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ParameterListInterpreter(const std::string& xmlFileName, const Teuchos::Comm<int>& comm, Teuchos::RCP<FactoryFactory> factFact, Teuchos::RCP<FacadeClassFactory> facadeFact)
+  : factFact_(factFact) {
+  RCP<Teuchos::TimeMonitor> tM = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(std::string("MueLu: ParameterListInterpreter (XML)"))));
+  if (facadeFact == Teuchos::null)
+    facadeFact_ = Teuchos::rcp(new FacadeClassFactory());
+  else
+    facadeFact_ = facadeFact;
+
+  ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), comm);
+  SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const ParameterList& paramList) {
+  Cycle_            = Hierarchy::GetDefaultCycle();
+  WCycleStartLevel_ = Hierarchy::GetDefaultCycleStartLevel();
+  scalingFactor_    = Teuchos::ScalarTraits<double>::one();
+  blockSize_        = 1;
+  dofOffset_        = 0;
+
+  if (paramList.isSublist("Hierarchy")) {
+    SetFactoryParameterList(paramList);
+
+  } else if (paramList.isParameter("MueLu preconditioner") == true) {
+    this->GetOStream(Runtime0) << "Use facade class: " << paramList.get<std::string>("MueLu preconditioner") << std::endl;
+    Teuchos::RCP<ParameterList> pp = facadeFact_->SetParameterList(paramList);
+    SetFactoryParameterList(*pp);
+
+  } else {
+    // The validator doesn't work correctly for non-serializable data (Hint: template parameters), so strip it out
+    ParameterList serialList, nonSerialList;
+
+    ExtractNonSerializableData(paramList, serialList, nonSerialList);
+    Validate(serialList);
+    SetEasyParameterList(paramList);
   }
+}
 
-  // =====================================================================================================
-  // ====================================== EASY interpreter =============================================
-  // =====================================================================================================
-  //! Helper functions to compare two paramter lists
-  static inline bool areSame(const ParameterList& list1, const ParameterList& list2);
+// =====================================================================================================
+// ====================================== EASY interpreter =============================================
+// =====================================================================================================
+//! Helper functions to compare two paramter lists
+static inline bool areSame(const ParameterList& list1, const ParameterList& list2);
 
-  // Get value from one of the lists, or set it to default
-  // Use case: check for a parameter value in a level-specific sublist, then in a root level list;
-  // if it is absent from both, set it to default
+// Get value from one of the lists, or set it to default
+// Use case: check for a parameter value in a level-specific sublist, then in a root level list;
+// if it is absent from both, set it to default
 #define MUELU_SET_VAR_2LIST(paramList, defaultList, paramName, paramType, varName) \
-  paramType varName; \
-  if      (paramList.isParameter(paramName))   varName = paramList.get<paramType>(paramName); \
-  else if (defaultList.isParameter(paramName)) varName = defaultList.get<paramType>(paramName); \
-  else                                         varName = MasterList::getDefault<paramType>(paramName);
+  paramType varName;                                                               \
+  if (paramList.isParameter(paramName))                                            \
+    varName = paramList.get<paramType>(paramName);                                 \
+  else if (defaultList.isParameter(paramName))                                     \
+    varName = defaultList.get<paramType>(paramName);                               \
+  else                                                                             \
+    varName = MasterList::getDefault<paramType>(paramName);
 
 #define MUELU_TEST_AND_SET_VAR(paramList, paramName, paramType, varName) \
   (paramList.isParameter(paramName) ? varName = paramList.get<paramType>(paramName), true : false)
 
-  // Set parameter in a list if it is present in any of two lists
-  // User case: set factory specific parameter, first checking for a level-specific value, then cheking root level value
-#define MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, paramName, paramType, listWrite) \
-  try { \
-    if      (paramList  .isParameter(paramName)) listWrite.set(paramName, paramList  .get<paramType>(paramName)); \
-    else if (defaultList.isParameter(paramName)) listWrite.set(paramName, defaultList.get<paramType>(paramName)); \
-  } \
-  catch(Teuchos::Exceptions::InvalidParameterType&) { \
-    TEUCHOS_TEST_FOR_EXCEPTION_PURE_MSG(true, Teuchos::Exceptions::InvalidParameterType, \
+// Set parameter in a list if it is present in any of two lists
+// User case: set factory specific parameter, first checking for a level-specific value, then cheking root level value
+#define MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, paramName, paramType, listWrite)                                                     \
+  try {                                                                                                                                             \
+    if (paramList.isParameter(paramName))                                                                                                           \
+      listWrite.set(paramName, paramList.get<paramType>(paramName));                                                                                \
+    else if (defaultList.isParameter(paramName))                                                                                                    \
+      listWrite.set(paramName, defaultList.get<paramType>(paramName));                                                                              \
+  } catch (Teuchos::Exceptions::InvalidParameterType&) {                                                                                            \
+    TEUCHOS_TEST_FOR_EXCEPTION_PURE_MSG(true, Teuchos::Exceptions::InvalidParameterType,                                                            \
                                         "Error: parameter \"" << paramName << "\" must be of type " << Teuchos::TypeNameTraits<paramType>::name()); \
-  } \
+  }
 
 #define MUELU_TEST_PARAM_2LIST(paramList, defaultList, paramName, paramType, cmpValue) \
-  (cmpValue == ( \
-    paramList.isParameter(paramName)   ? paramList  .get<paramType>(paramName) : ( \
-      defaultList.isParameter(paramName) ? defaultList.get<paramType>(paramName) : \
-      MasterList::getDefault<paramType>(paramName) ) ) )
+  (cmpValue == (paramList.isParameter(paramName) ? paramList.get<paramType>(paramName) : (defaultList.isParameter(paramName) ? defaultList.get<paramType>(paramName) : MasterList::getDefault<paramType>(paramName))))
 
 #define MUELU_KOKKOS_FACTORY(varName, oldFactory, newFactory) \
-  RCP<Factory> varName; \
-  if (!useKokkos_) varName = rcp(new oldFactory()); \
-  else             varName = rcp(new newFactory());
+  RCP<Factory> varName;                                       \
+  if (!useKokkos_)                                            \
+    varName = rcp(new oldFactory());                          \
+  else                                                        \
+    varName = rcp(new newFactory());
 #define MUELU_KOKKOS_FACTORY_NO_DECL(varName, oldFactory, newFactory) \
-  if (!useKokkos_) varName = rcp(new oldFactory()); \
-  else             varName = rcp(new newFactory());
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  SetEasyParameterList(const ParameterList& constParamList) {
-    ParameterList paramList;
-
-    MUELU_SET_VAR_2LIST(constParamList, constParamList, "problem: type", std::string, problemType);
-    if (problemType != "unknown") {
-      paramList = *MasterList::GetProblemSpecificList(problemType);
-      paramList.setParameters(constParamList);
-    } else {
-      // Create a non const copy of the parameter list
-      // Working with a modifiable list is much much easier than with original one
-      paramList = constParamList;
-    }
-
-    // Check for Kokkos
-    useKokkos_ = !Node::is_serial;
-    (void)MUELU_TEST_AND_SET_VAR(paramList, "use kokkos refactor", bool, useKokkos_);
-
-    // Check for timer synchronization
-    MUELU_SET_VAR_2LIST(paramList, paramList, "synchronize factory timers", bool, syncTimers);
-    if (syncTimers)
-        Factory::EnableTimerSync();
-
-    // Translate cycle type parameter
-    if (paramList.isParameter("cycle type")) {
-      std::map<std::string, CycleType> cycleMap;
-      cycleMap["V"] = VCYCLE;
-      cycleMap["W"] = WCYCLE;
-
-      auto cycleType = paramList.get<std::string>("cycle type");
-      TEUCHOS_TEST_FOR_EXCEPTION(cycleMap.count(cycleType) == 0, Exceptions::RuntimeError,
-                                 "Invalid cycle type: \"" << cycleType << "\"");
-      Cycle_ = cycleMap[cycleType];
-    }
-
-    if (paramList.isParameter("W cycle start level")) {
-      WCycleStartLevel_ = paramList.get<int>("W cycle start level");
-    }
-
-    if (paramList.isParameter("coarse grid correction scaling factor"))
-      scalingFactor_ = paramList.get<double>("coarse grid correction scaling factor");
-
-    this->maxCoarseSize_    = paramList.get<int> ("coarse: max size",    MasterList::getDefault<int>("coarse: max size"));
-    this->numDesiredLevel_  = paramList.get<int> ("max levels",          MasterList::getDefault<int>("max levels"));
-    blockSize_              = paramList.get<int> ("number of equations", MasterList::getDefault<int>("number of equations"));
-
-
-    (void)MUELU_TEST_AND_SET_VAR(paramList, "debug: graph level", int, this->graphOutputLevel_);
-
-    // Generic data saving (this saves the data on all levels)
-    if(paramList.isParameter("save data"))
-      this->dataToSave_ = Teuchos::getArrayFromStringParameter<std::string>(paramList,"save data");
+  if (!useKokkos_)                                                    \
+    varName = rcp(new oldFactory());                                  \
+  else                                                                \
+    varName = rcp(new newFactory());
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SetEasyParameterList(const ParameterList& constParamList) {
+  ParameterList paramList;
+
+  MUELU_SET_VAR_2LIST(constParamList, constParamList, "problem: type", std::string, problemType);
+  if (problemType != "unknown") {
+    paramList = *MasterList::GetProblemSpecificList(problemType);
+    paramList.setParameters(constParamList);
+  } else {
+    // Create a non const copy of the parameter list
+    // Working with a modifiable list is much much easier than with original one
+    paramList = constParamList;
+  }
 
-    // Save level data
-    if (paramList.isSublist("export data")) {
-      ParameterList printList = paramList.sublist("export data");
+  // Check for Kokkos
+  useKokkos_ = !Node::is_serial;
+  (void)MUELU_TEST_AND_SET_VAR(paramList, "use kokkos refactor", bool, useKokkos_);
+
+  // Check for timer synchronization
+  MUELU_SET_VAR_2LIST(paramList, paramList, "synchronize factory timers", bool, syncTimers);
+  if (syncTimers)
+    Factory::EnableTimerSync();
+
+  // Translate cycle type parameter
+  if (paramList.isParameter("cycle type")) {
+    std::map<std::string, CycleType> cycleMap;
+    cycleMap["V"] = VCYCLE;
+    cycleMap["W"] = WCYCLE;
+
+    auto cycleType = paramList.get<std::string>("cycle type");
+    TEUCHOS_TEST_FOR_EXCEPTION(cycleMap.count(cycleType) == 0, Exceptions::RuntimeError,
+                               "Invalid cycle type: \"" << cycleType << "\"");
+    Cycle_ = cycleMap[cycleType];
+  }
 
-      // Vectors, aggregates and other things that need special handling
-      if (printList.isParameter("Nullspace"))
-        this->nullspaceToPrint_  = Teuchos::getArrayFromStringParameter<int>(printList, "Nullspace");
-      if (printList.isParameter("Coordinates"))
-        this->coordinatesToPrint_  = Teuchos::getArrayFromStringParameter<int>(printList, "Coordinates");
-      if (printList.isParameter("Aggregates"))
-        this->aggregatesToPrint_  = Teuchos::getArrayFromStringParameter<int>(printList, "Aggregates");
-      if (printList.isParameter("pcoarsen: element to node map"))
-        this->elementToNodeMapsToPrint_  = Teuchos::getArrayFromStringParameter<int>(printList, "pcoarsen: element to node map");
+  if (paramList.isParameter("W cycle start level")) {
+    WCycleStartLevel_ = paramList.get<int>("W cycle start level");
+  }
 
-      // If we asked for an arbitrary matrix to be printed, we do that here
-      for(auto iter = printList.begin(); iter != printList.end(); iter++) {
-        const std::string & name = printList.name(iter);
-        // Ignore the special cases
-        if(name == "Nullspace" || name == "Coordinates" || name == "Aggregates" || name == "pcoarsen: element to node map")
-          continue;
+  if (paramList.isParameter("coarse grid correction scaling factor"))
+    scalingFactor_ = paramList.get<double>("coarse grid correction scaling factor");
+
+  this->maxCoarseSize_   = paramList.get<int>("coarse: max size", MasterList::getDefault<int>("coarse: max size"));
+  this->numDesiredLevel_ = paramList.get<int>("max levels", MasterList::getDefault<int>("max levels"));
+  blockSize_             = paramList.get<int>("number of equations", MasterList::getDefault<int>("number of equations"));
+
+  (void)MUELU_TEST_AND_SET_VAR(paramList, "debug: graph level", int, this->graphOutputLevel_);
+
+  // Generic data saving (this saves the data on all levels)
+  if (paramList.isParameter("save data"))
+    this->dataToSave_ = Teuchos::getArrayFromStringParameter<std::string>(paramList, "save data");
+
+  // Save level data
+  if (paramList.isSublist("export data")) {
+    ParameterList printList = paramList.sublist("export data");
+
+    // Vectors, aggregates and other things that need special handling
+    if (printList.isParameter("Nullspace"))
+      this->nullspaceToPrint_ = Teuchos::getArrayFromStringParameter<int>(printList, "Nullspace");
+    if (printList.isParameter("Coordinates"))
+      this->coordinatesToPrint_ = Teuchos::getArrayFromStringParameter<int>(printList, "Coordinates");
+    if (printList.isParameter("Aggregates"))
+      this->aggregatesToPrint_ = Teuchos::getArrayFromStringParameter<int>(printList, "Aggregates");
+    if (printList.isParameter("pcoarsen: element to node map"))
+      this->elementToNodeMapsToPrint_ = Teuchos::getArrayFromStringParameter<int>(printList, "pcoarsen: element to node map");
+
+    // If we asked for an arbitrary matrix to be printed, we do that here
+    for (auto iter = printList.begin(); iter != printList.end(); iter++) {
+      const std::string& name = printList.name(iter);
+      // Ignore the special cases
+      if (name == "Nullspace" || name == "Coordinates" || name == "Aggregates" || name == "pcoarsen: element to node map")
+        continue;
 
-        this->matricesToPrint_[name] = Teuchos::getArrayFromStringParameter<int>(printList, name);
-      }
+      this->matricesToPrint_[name] = Teuchos::getArrayFromStringParameter<int>(printList, name);
     }
+  }
 
-    // Set verbosity parameter
-    VerbLevel oldVerbLevel = VerboseObject::GetDefaultVerbLevel();
-    {
-      MUELU_SET_VAR_2LIST(paramList, paramList, "verbosity", std::string, verbosityLevel);
-      this->verbosity_ = toVerbLevel(verbosityLevel);
-      VerboseObject::SetDefaultVerbLevel(this->verbosity_);
-    }
+  // Set verbosity parameter
+  VerbLevel oldVerbLevel = VerboseObject::GetDefaultVerbLevel();
+  {
+    MUELU_SET_VAR_2LIST(paramList, paramList, "verbosity", std::string, verbosityLevel);
+    this->verbosity_ = toVerbLevel(verbosityLevel);
+    VerboseObject::SetDefaultVerbLevel(this->verbosity_);
+  }
 
-    MUELU_SET_VAR_2LIST(paramList, paramList, "output filename", std::string, outputFilename);
-    if (outputFilename != "")
-      VerboseObject::SetMueLuOFileStream(outputFilename);
-
-    // Detect if we need to transfer coordinates to coarse levels. We do that iff
-    //  - we use "distance laplacian" dropping on some level, or
-    //  - we use a repartitioner on some level that needs coordinates
-    //  - we use brick aggregation
-    //  - we use Ifpack2 line partitioner
-    // This is not ideal, as we may have "repartition: enable" turned on by default
-    // and not present in the list, but it is better than nothing.
-    useCoordinates_ = false;
-    useBlockNumber_ = false;
-    if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "distance laplacian") ||
-        MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: type",        std::string, "brick") ||
-        MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: export visualization data", bool, true)) {
+  MUELU_SET_VAR_2LIST(paramList, paramList, "output filename", std::string, outputFilename);
+  if (outputFilename != "")
+    VerboseObject::SetMueLuOFileStream(outputFilename);
+
+  // Detect if we need to transfer coordinates to coarse levels. We do that iff
+  //  - we use "distance laplacian" dropping on some level, or
+  //  - we use a repartitioner on some level that needs coordinates
+  //  - we use brick aggregation
+  //  - we use Ifpack2 line partitioner
+  // This is not ideal, as we may have "repartition: enable" turned on by default
+  // and not present in the list, but it is better than nothing.
+  useCoordinates_ = false;
+  useBlockNumber_ = false;
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "distance laplacian") ||
+      MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: type", std::string, "brick") ||
+      MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: export visualization data", bool, true)) {
+    useCoordinates_ = true;
+  } else if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal distance laplacian")) {
+    useCoordinates_ = true;
+    useBlockNumber_ = true;
+  } else if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal") ||
+             MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal classical") ||
+             MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal signed classical") ||
+             MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal colored signed classical") ||
+             MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "signed classical")) {
+    useBlockNumber_ = true;
+  } else if (paramList.isSublist("smoother: params")) {
+    const auto smooParamList = paramList.sublist("smoother: params");
+    if (smooParamList.isParameter("partitioner: type") &&
+        (smooParamList.get<std::string>("partitioner: type") == "line")) {
       useCoordinates_ = true;
-    } else if(MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal distance laplacian")) {
-      useCoordinates_ = true;
-      useBlockNumber_ = true;
-    } else if(MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal") ||
-              MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal classical") ||
-              MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal signed classical") ||
-              MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal colored signed classical") ||
-              MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "signed classical")) {
-      useBlockNumber_ = true;
-    } else if(paramList.isSublist("smoother: params")) {
-      const auto smooParamList = paramList.sublist("smoother: params");
-      if(smooParamList.isParameter("partitioner: type") &&
-         (smooParamList.get<std::string>("partitioner: type") == "line")) {
-        useCoordinates_ = true;
-      }
-    } else {
-      for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
-        std::string levelStr = "level " + toString(levelID);
+    }
+  } else {
+    for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
+      std::string levelStr = "level " + toString(levelID);
 
-        if (paramList.isSublist(levelStr)) {
-          const ParameterList& levelList = paramList.sublist(levelStr);
+      if (paramList.isSublist(levelStr)) {
+        const ParameterList& levelList = paramList.sublist(levelStr);
 
-          if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "distance laplacian") ||
-              MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: type",        std::string, "brick") ||
-              MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: export visualization data", bool, true)) {
-            useCoordinates_ = true;
-          }
-          else if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "block diagonal distance laplacian")) {
-            useCoordinates_ = true;
-            useBlockNumber_ = true;
-          }
-          else if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "block diagonal") ||
+        if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "distance laplacian") ||
+            MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: type", std::string, "brick") ||
+            MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: export visualization data", bool, true)) {
+          useCoordinates_ = true;
+        } else if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "block diagonal distance laplacian")) {
+          useCoordinates_ = true;
+          useBlockNumber_ = true;
+        } else if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "block diagonal") ||
                    MUELU_TEST_PARAM_2LIST(levelList, paramList, "aggregation: drop scheme", std::string, "block diagonal classical") ||
                    MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal signed classical") ||
                    MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "block diagonal colored signed classical") ||
                    MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "signed classical")) {
-            useBlockNumber_ = true;
-          }
+          useBlockNumber_ = true;
         }
       }
     }
+  }
 
-    if(MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: enable",      bool,        true)) {
-      // We don't need coordinates if we're doing the in-place restriction
-      if(MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: use subcommunicators",      bool,        true) &&
-         MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: use subcommunicators in place",      bool,        true)) {
-        // do nothing --- these don't need coordinates
-      } else if (!paramList.isSublist("repartition: params")) {
-        useCoordinates_ = true;
-      } else {
-        const ParameterList& repParams = paramList.sublist("repartition: params");
-        if (repParams.isType<std::string>("algorithm")) {
-          const std::string algo = repParams.get<std::string>("algorithm");
-          if (algo == "multijagged" || algo == "rcb") {
-            useCoordinates_ = true;
-          }
-        } else {
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: enable", bool, true)) {
+    // We don't need coordinates if we're doing the in-place restriction
+    if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: use subcommunicators", bool, true) &&
+        MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: use subcommunicators in place", bool, true)) {
+      // do nothing --- these don't need coordinates
+    } else if (!paramList.isSublist("repartition: params")) {
+      useCoordinates_ = true;
+    } else {
+      const ParameterList& repParams = paramList.sublist("repartition: params");
+      if (repParams.isType<std::string>("algorithm")) {
+        const std::string algo = repParams.get<std::string>("algorithm");
+        if (algo == "multijagged" || algo == "rcb") {
           useCoordinates_ = true;
         }
+      } else {
+        useCoordinates_ = true;
       }
     }
-    for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
-      std::string levelStr = "level " + toString(levelID);
+  }
+  for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
+    std::string levelStr = "level " + toString(levelID);
 
-      if (paramList.isSublist(levelStr)) {
-        const ParameterList& levelList = paramList.sublist(levelStr);
+    if (paramList.isSublist(levelStr)) {
+      const ParameterList& levelList = paramList.sublist(levelStr);
 
-        if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "repartition: enable",      bool,        true)) {
-          if (!levelList.isSublist("repartition: params")) {
-            useCoordinates_ = true;
-            break;
-          } else {
-            const ParameterList& repParams = levelList.sublist("repartition: params");
-            if (repParams.isType<std::string>("algorithm")) {
-              const std::string algo = repParams.get<std::string>("algorithm");
-              if (algo == "multijagged" || algo == "rcb"){
-                useCoordinates_ = true;
-                break;
-              }
-            } else {
+      if (MUELU_TEST_PARAM_2LIST(levelList, paramList, "repartition: enable", bool, true)) {
+        if (!levelList.isSublist("repartition: params")) {
+          useCoordinates_ = true;
+          break;
+        } else {
+          const ParameterList& repParams = levelList.sublist("repartition: params");
+          if (repParams.isType<std::string>("algorithm")) {
+            const std::string algo = repParams.get<std::string>("algorithm");
+            if (algo == "multijagged" || algo == "rcb") {
               useCoordinates_ = true;
               break;
             }
+          } else {
+            useCoordinates_ = true;
+            break;
           }
         }
       }
     }
+  }
 
-    // Detect if we do implicit P and R rebalance
-    changedPRrebalance_ = false;
-    changedPRViaCopyrebalance_ = false;
-    if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: enable", bool, true)) {
-      changedPRrebalance_ = MUELU_TEST_AND_SET_VAR(paramList, "repartition: rebalance P and R", bool, this->doPRrebalance_);
-      changedPRViaCopyrebalance_ =  MUELU_TEST_AND_SET_VAR(paramList,"repartition: explicit via new copy rebalance P and R", bool, this->doPRViaCopyrebalance_);
-    }
-
-    // Detect if we use implicit transpose
-    changedImplicitTranspose_ = MUELU_TEST_AND_SET_VAR(paramList, "transpose: use implicit", bool, this->implicitTranspose_);
+  // Detect if we do implicit P and R rebalance
+  changedPRrebalance_        = false;
+  changedPRViaCopyrebalance_ = false;
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "repartition: enable", bool, true)) {
+    changedPRrebalance_        = MUELU_TEST_AND_SET_VAR(paramList, "repartition: rebalance P and R", bool, this->doPRrebalance_);
+    changedPRViaCopyrebalance_ = MUELU_TEST_AND_SET_VAR(paramList, "repartition: explicit via new copy rebalance P and R", bool, this->doPRViaCopyrebalance_);
+  }
 
-    // Detect if we use fuse prolongation and update
-    (void)MUELU_TEST_AND_SET_VAR(paramList, "fuse prolongation and update", bool, this->fuseProlongationAndUpdate_);
+  // Detect if we use implicit transpose
+  changedImplicitTranspose_ = MUELU_TEST_AND_SET_VAR(paramList, "transpose: use implicit", bool, this->implicitTranspose_);
 
-    // Detect if we suppress the dimension check of the user-given nullspace
-    (void)MUELU_TEST_AND_SET_VAR(paramList, "nullspace: suppress dimension check", bool, this->suppressNullspaceDimensionCheck_);
+  // Detect if we use fuse prolongation and update
+  (void)MUELU_TEST_AND_SET_VAR(paramList, "fuse prolongation and update", bool, this->fuseProlongationAndUpdate_);
 
-    if (paramList.isSublist("matvec params"))
-      this->matvecParams_ = Teuchos::parameterList(paramList.sublist("matvec params"));
+  // Detect if we suppress the dimension check of the user-given nullspace
+  (void)MUELU_TEST_AND_SET_VAR(paramList, "nullspace: suppress dimension check", bool, this->suppressNullspaceDimensionCheck_);
 
-    // Create default manager
-    // FIXME: should it be here, or higher up
-    RCP<FactoryManager> defaultManager = rcp(new FactoryManager());
-    defaultManager->SetVerbLevel(this->verbosity_);
-    defaultManager->SetKokkosRefactor(useKokkos_);
+  if (paramList.isSublist("matvec params"))
+    this->matvecParams_ = Teuchos::parameterList(paramList.sublist("matvec params"));
 
-    // We will ignore keeps0
-    std::vector<keep_pair> keeps0;
-    UpdateFactoryManager(paramList, ParameterList(), *defaultManager, 0/*levelID*/, keeps0);
+  // Create default manager
+  // FIXME: should it be here, or higher up
+  RCP<FactoryManager> defaultManager = rcp(new FactoryManager());
+  defaultManager->SetVerbLevel(this->verbosity_);
+  defaultManager->SetKokkosRefactor(useKokkos_);
 
-    //    std::cout<<"*** Default Manager ***"<<std::endl;
-    //    defaultManager->Print();
+  // We will ignore keeps0
+  std::vector<keep_pair> keeps0;
+  UpdateFactoryManager(paramList, ParameterList(), *defaultManager, 0 /*levelID*/, keeps0);
 
-    // Create level specific factory managers
-    for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
-      // Note, that originally if there were no level specific parameters, we
-      // simply copied the defaultManager However, with the introduction of
-      // levelID to UpdateFactoryManager (required for reuse), we can no longer
-      // guarantee that the kept variables are the same for each level even if
-      // dependency structure does not change.
-      RCP<FactoryManager> levelManager = rcp(new FactoryManager(*defaultManager));
-      levelManager->SetVerbLevel(defaultManager->GetVerbLevel());
-
-      std::vector<keep_pair> keeps;
-      if (paramList.isSublist("level " + toString(levelID))) {
-        // We do this so the parameters on the level get flagged correctly as "used"
-        ParameterList& levelList = paramList.sublist("level " + toString(levelID), true/*mustAlreadyExist*/);
-        UpdateFactoryManager(levelList, paramList, *levelManager, levelID, keeps);
-
-      } else {
-        ParameterList levelList;
-        UpdateFactoryManager(levelList, paramList, *levelManager, levelID, keeps);
-      }
+  //    std::cout<<"*** Default Manager ***"<<std::endl;
+  //    defaultManager->Print();
 
-      this->keep_[levelID] = keeps;
-      this->AddFactoryManager(levelID, 1, levelManager);
+  // Create level specific factory managers
+  for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
+    // Note, that originally if there were no level specific parameters, we
+    // simply copied the defaultManager However, with the introduction of
+    // levelID to UpdateFactoryManager (required for reuse), we can no longer
+    // guarantee that the kept variables are the same for each level even if
+    // dependency structure does not change.
+    RCP<FactoryManager> levelManager = rcp(new FactoryManager(*defaultManager));
+    levelManager->SetVerbLevel(defaultManager->GetVerbLevel());
 
-      //      std::cout<<"*** Level "<<levelID<<" Manager ***"<<std::endl;
-      //      levelManager->Print();
+    std::vector<keep_pair> keeps;
+    if (paramList.isSublist("level " + toString(levelID))) {
+      // We do this so the parameters on the level get flagged correctly as "used"
+      ParameterList& levelList = paramList.sublist("level " + toString(levelID), true /*mustAlreadyExist*/);
+      UpdateFactoryManager(levelList, paramList, *levelManager, levelID, keeps);
 
+    } else {
+      ParameterList levelList;
+      UpdateFactoryManager(levelList, paramList, *levelManager, levelID, keeps);
     }
 
-    // FIXME: parameters passed to packages, like Ifpack2, are not touched by us, resulting in "[unused]" flag
-    // being displayed. On the other hand, we don't want to simply iterate through them touching. I don't know
-    // what a good solution looks like
-    if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "print initial parameters", bool, true))
-      this->GetOStream(static_cast<MsgType>(Runtime1), 0) << paramList << std::endl;
+    this->keep_[levelID] = keeps;
+    this->AddFactoryManager(levelID, 1, levelManager);
 
-    if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "print unused parameters", bool, true)) {
-      // Check unused parameters
-      ParameterList unusedParamList;
+    //      std::cout<<"*** Level "<<levelID<<" Manager ***"<<std::endl;
+    //      levelManager->Print();
+  }
 
-      // Check for unused parameters that aren't lists
-      for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++) {
-        const ParameterEntry& entry = paramList.entry(it);
+  // FIXME: parameters passed to packages, like Ifpack2, are not touched by us, resulting in "[unused]" flag
+  // being displayed. On the other hand, we don't want to simply iterate through them touching. I don't know
+  // what a good solution looks like
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "print initial parameters", bool, true))
+    this->GetOStream(static_cast<MsgType>(Runtime1), 0) << paramList << std::endl;
 
-        if (!entry.isList() && !entry.isUsed())
-          unusedParamList.setEntry(paramList.name(it), entry);
-      }
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "print unused parameters", bool, true)) {
+    // Check unused parameters
+    ParameterList unusedParamList;
 
-      // Check for unused parameters in level-specific sublists
-      for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
-        std::string levelStr = "level " + toString(levelID);
+    // Check for unused parameters that aren't lists
+    for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++) {
+      const ParameterEntry& entry = paramList.entry(it);
 
-        if (paramList.isSublist(levelStr)) {
-          const ParameterList& levelList = paramList.sublist(levelStr);
+      if (!entry.isList() && !entry.isUsed())
+        unusedParamList.setEntry(paramList.name(it), entry);
+    }
 
-          for (ParameterList::ConstIterator itr = levelList.begin(); itr != levelList.end(); ++itr) {
-            const ParameterEntry& entry = levelList.entry(itr);
+    // Check for unused parameters in level-specific sublists
+    for (int levelID = 0; levelID < this->numDesiredLevel_; levelID++) {
+      std::string levelStr = "level " + toString(levelID);
 
-            if (!entry.isList() && !entry.isUsed())
-              unusedParamList.sublist(levelStr).setEntry(levelList.name(itr), entry);
-          }
-        }
-      }
+      if (paramList.isSublist(levelStr)) {
+        const ParameterList& levelList = paramList.sublist(levelStr);
 
-      if (unusedParamList.numParams() > 0) {
-        std::ostringstream unusedParamsStream;
-        int indent = 4;
-        unusedParamList.print(unusedParamsStream, indent);
+        for (ParameterList::ConstIterator itr = levelList.begin(); itr != levelList.end(); ++itr) {
+          const ParameterEntry& entry = levelList.entry(itr);
 
-        this->GetOStream(Warnings1) << "The following parameters were not used:\n" << unusedParamsStream.str() << std::endl;
+          if (!entry.isList() && !entry.isUsed())
+            unusedParamList.sublist(levelStr).setEntry(levelList.name(itr), entry);
+        }
       }
     }
 
-    VerboseObject::SetDefaultVerbLevel(oldVerbLevel);
+    if (unusedParamList.numParams() > 0) {
+      std::ostringstream unusedParamsStream;
+      int indent = 4;
+      unusedParamList.print(unusedParamsStream, indent);
 
+      this->GetOStream(Warnings1) << "The following parameters were not used:\n"
+                                  << unusedParamsStream.str() << std::endl;
+    }
   }
 
+  VerboseObject::SetDefaultVerbLevel(oldVerbLevel);
+}
 
-  // =====================================================================================================
-  // ==================================== UpdateFactoryManager ===========================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                       int levelID, std::vector<keep_pair>& keeps) const
-  {
-    // NOTE: Factory::SetParameterList must be called prior to Factory::SetFactory, as
-    // SetParameterList sets default values for non mentioned parameters, including factories
+// =====================================================================================================
+// ==================================== UpdateFactoryManager ===========================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                         int levelID, std::vector<keep_pair>& keeps) const {
+  // NOTE: Factory::SetParameterList must be called prior to Factory::SetFactory, as
+  // SetParameterList sets default values for non mentioned parameters, including factories
 
-    using strings = std::unordered_set<std::string>;
+  using strings = std::unordered_set<std::string>;
 
-    // shortcut
-    if (paramList.numParams() == 0 && defaultList.numParams() > 0)
-      paramList = ParameterList(defaultList);
+  // shortcut
+  if (paramList.numParams() == 0 && defaultList.numParams() > 0)
+    paramList = ParameterList(defaultList);
 
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-    TEUCHOS_TEST_FOR_EXCEPTION(strings({"none", "tP", "RP", "emin", "RAP", "full", "S"}).count(reuseType) == 0,
-        Exceptions::RuntimeError, "Unknown \"reuse: type\" value: \"" << reuseType << "\". Please consult User's Guide.");
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  TEUCHOS_TEST_FOR_EXCEPTION(strings({"none", "tP", "RP", "emin", "RAP", "full", "S"}).count(reuseType) == 0,
+                             Exceptions::RuntimeError, "Unknown \"reuse: type\" value: \"" << reuseType << "\". Please consult User's Guide.");
 
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
-    TEUCHOS_TEST_FOR_EXCEPTION(strings({"unsmoothed", "sa", "pg", "emin", "matlab", "pcoarsen","classical","smoothed reitzinger","unsmoothed reitzinger","replicate","combine"}).count(multigridAlgo) == 0,
-        Exceptions::RuntimeError, "Unknown \"multigrid algorithm\" value: \"" << multigridAlgo << "\". Please consult User's Guide.");
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
+  TEUCHOS_TEST_FOR_EXCEPTION(strings({"unsmoothed", "sa", "pg", "emin", "matlab", "pcoarsen", "classical", "smoothed reitzinger", "unsmoothed reitzinger", "replicate", "combine"}).count(multigridAlgo) == 0,
+                             Exceptions::RuntimeError, "Unknown \"multigrid algorithm\" value: \"" << multigridAlgo << "\". Please consult User's Guide.");
 #ifndef HAVE_MUELU_MATLAB
-    TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "matlab", Exceptions::RuntimeError,
-        "Cannot use matlab for multigrid algorithm - MueLu was not configured with MATLAB support.");
+  TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "matlab", Exceptions::RuntimeError,
+                             "Cannot use matlab for multigrid algorithm - MueLu was not configured with MATLAB support.");
 #endif
 #ifndef HAVE_MUELU_INTREPID2
-    TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "pcoarsen", Exceptions::RuntimeError,
-        "Cannot use IntrepidPCoarsen prolongator factory - MueLu was not configured with Intrepid support.");
+  TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "pcoarsen", Exceptions::RuntimeError,
+                             "Cannot use IntrepidPCoarsen prolongator factory - MueLu was not configured with Intrepid support.");
 #endif
 
-    // Only some combinations of reuse and multigrid algorithms are tested, all
-    // other are considered invalid at the moment
-    if (reuseType == "none" || reuseType == "S" || reuseType == "RP" || reuseType == "RAP") {
-      // This works for all kinds of multigrid algorithms
-
-    } else if (reuseType == "tP" && (multigridAlgo != "sa" && multigridAlgo != "unsmoothed")) {
-      reuseType = "none";
-      this->GetOStream(Warnings0) << "Ignoring \"tP\" reuse option as it is only compatible with \"sa\", "
-          "or \"unsmoothed\" multigrid algorithms" << std::endl;
-
-    } else if (reuseType == "emin" && multigridAlgo != "emin") {
-      reuseType = "none";
-      this->GetOStream(Warnings0) << "Ignoring \"emin\" reuse option it is only compatible with "
-          "\"emin\" multigrid algorithm" << std::endl;
-    }
-
-    // == Non-serializable data ===
-    // Check both the parameter and the type
-    bool have_userP = false;
-    if (paramList.isParameter("P") && !paramList.get<RCP<Matrix> >("P").is_null())
-      have_userP  = true;
+  // Only some combinations of reuse and multigrid algorithms are tested, all
+  // other are considered invalid at the moment
+  if (reuseType == "none" || reuseType == "S" || reuseType == "RP" || reuseType == "RAP") {
+    // This works for all kinds of multigrid algorithms
+
+  } else if (reuseType == "tP" && (multigridAlgo != "sa" && multigridAlgo != "unsmoothed")) {
+    reuseType = "none";
+    this->GetOStream(Warnings0) << "Ignoring \"tP\" reuse option as it is only compatible with \"sa\", "
+                                   "or \"unsmoothed\" multigrid algorithms"
+                                << std::endl;
+
+  } else if (reuseType == "emin" && multigridAlgo != "emin") {
+    reuseType = "none";
+    this->GetOStream(Warnings0) << "Ignoring \"emin\" reuse option it is only compatible with "
+                                   "\"emin\" multigrid algorithm"
+                                << std::endl;
+  }
 
-    // === Coarse solver ===
-    UpdateFactoryManager_CoarseSolvers(paramList, defaultList, manager, levelID, keeps);
+  // == Non-serializable data ===
+  // Check both the parameter and the type
+  bool have_userP = false;
+  if (paramList.isParameter("P") && !paramList.get<RCP<Matrix> >("P").is_null())
+    have_userP = true;
+
+  // === Coarse solver ===
+  UpdateFactoryManager_CoarseSolvers(paramList, defaultList, manager, levelID, keeps);
+
+  // == Smoothers ==
+  UpdateFactoryManager_Smoothers(paramList, defaultList, manager, levelID, keeps);
+
+  // === BlockNumber ===
+  if (levelID == 0)
+    UpdateFactoryManager_BlockNumber(paramList, defaultList, manager, levelID, keeps);
+
+  // === Aggregation ===
+  if (multigridAlgo == "unsmoothed reitzinger" || multigridAlgo == "smoothed reitzinger")
+    UpdateFactoryManager_Reitzinger(paramList, defaultList, manager, levelID, keeps);
+  else
+    UpdateFactoryManager_Aggregation_TentativeP(paramList, defaultList, manager, levelID, keeps);
+
+  // === Nullspace ===
+  RCP<Factory> nullSpaceFactory;  // Cache thcAN is guy for the combination of semi-coarsening & repartitioning
+  UpdateFactoryManager_Nullspace(paramList, defaultList, manager, levelID, keeps, nullSpaceFactory);
+
+  // === Prolongation ===
+  // NOTE: None of the UpdateFactoryManager routines called here check the
+  // multigridAlgo. This is intentional, to allow for reuse of components
+  // underneath. Thus, the multigridAlgo was checked in the beginning of the
+  // function.
+  if (have_userP) {
+    // User prolongator
+    manager.SetFactory("P", NoFactory::getRCP());
+
+  } else if (multigridAlgo == "unsmoothed" || multigridAlgo == "unsmoothed reitzinger") {
+    // Unsmoothed aggregation
+    manager.SetFactory("P", manager.GetFactory("Ptent"));
+
+  } else if (multigridAlgo == "classical") {
+    // Classical AMG
+    manager.SetFactory("P", manager.GetFactory("Ptent"));
+
+  } else if (multigridAlgo == "sa" || multigridAlgo == "smoothed reitzinger") {
+    // Smoothed aggregation
+    UpdateFactoryManager_SA(paramList, defaultList, manager, levelID, keeps);
 
-    // == Smoothers ==
-    UpdateFactoryManager_Smoothers(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "emin") {
+    // Energy minimization
+    UpdateFactoryManager_Emin(paramList, defaultList, manager, levelID, keeps);
 
-    // === BlockNumber ===
-    if(levelID == 0)
-      UpdateFactoryManager_BlockNumber(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "replicate") {
+    UpdateFactoryManager_Replicate(paramList, defaultList, manager, levelID, keeps);
 
-    // === Aggregation ===
-    if(multigridAlgo == "unsmoothed reitzinger" || multigridAlgo == "smoothed reitzinger")
-      UpdateFactoryManager_Reitzinger(paramList, defaultList, manager, levelID, keeps);
-    else
-      UpdateFactoryManager_Aggregation_TentativeP(paramList, defaultList, manager, levelID, keeps);
-
-    // === Nullspace ===
-    RCP<Factory> nullSpaceFactory; // Cache thcAN is guy for the combination of semi-coarsening & repartitioning
-    UpdateFactoryManager_Nullspace(paramList, defaultList, manager, levelID, keeps, nullSpaceFactory);
-
-    // === Prolongation ===
-    // NOTE: None of the UpdateFactoryManager routines called here check the
-    // multigridAlgo. This is intentional, to allow for reuse of components
-    // underneath. Thus, the multigridAlgo was checked in the beginning of the
-    // function.
-    if (have_userP) {
-      // User prolongator
-      manager.SetFactory("P", NoFactory::getRCP());
-
-    } else if (multigridAlgo == "unsmoothed" || multigridAlgo == "unsmoothed reitzinger") {
-      // Unsmoothed aggregation
-      manager.SetFactory("P", manager.GetFactory("Ptent"));
-
-    } else if (multigridAlgo == "classical") {
-      // Classical AMG
-      manager.SetFactory("P", manager.GetFactory("Ptent"));
-
-    } else if (multigridAlgo == "sa" || multigridAlgo == "smoothed reitzinger") {
-      // Smoothed aggregation
-      UpdateFactoryManager_SA(paramList, defaultList, manager, levelID, keeps);
-
-    } else if (multigridAlgo == "emin") {
-      // Energy minimization
-      UpdateFactoryManager_Emin(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "combine") {
+    UpdateFactoryManager_Combine(paramList, defaultList, manager, levelID, keeps);
 
-    } else if (multigridAlgo == "replicate") {
-      UpdateFactoryManager_Replicate(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "pg") {
+    // Petrov-Galerkin
+    UpdateFactoryManager_PG(paramList, defaultList, manager, levelID, keeps);
 
-    } else if (multigridAlgo == "combine") {
-      UpdateFactoryManager_Combine(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "matlab") {
+    // Matlab Coarsneing
+    UpdateFactoryManager_Matlab(paramList, defaultList, manager, levelID, keeps);
 
-    } else if (multigridAlgo == "pg") {
-      // Petrov-Galerkin
-      UpdateFactoryManager_PG(paramList, defaultList, manager, levelID, keeps);
+  } else if (multigridAlgo == "pcoarsen") {
+    // P-Coarsening
+    UpdateFactoryManager_PCoarsen(paramList, defaultList, manager, levelID, keeps);
+  }
 
-    } else if (multigridAlgo == "matlab") {
-      // Matlab Coarsneing
-      UpdateFactoryManager_Matlab(paramList, defaultList, manager, levelID, keeps);
+  // === Semi-coarsening ===
+  UpdateFactoryManager_SemiCoarsen(paramList, defaultList, manager, levelID, keeps);
 
-    } else if (multigridAlgo == "pcoarsen") {
-      // P-Coarsening
-      UpdateFactoryManager_PCoarsen(paramList, defaultList, manager, levelID, keeps);
-    }
+  // === Restriction ===
+  UpdateFactoryManager_Restriction(paramList, defaultList, manager, levelID, keeps);
 
-    // === Semi-coarsening ===
-    UpdateFactoryManager_SemiCoarsen(paramList, defaultList, manager, levelID, keeps);
+  // === RAP ===
+  UpdateFactoryManager_RAP(paramList, defaultList, manager, levelID, keeps);
 
-    // === Restriction ===
-    UpdateFactoryManager_Restriction(paramList, defaultList, manager, levelID, keeps);
+  // == BlockNumber Transfer ==
+  UpdateFactoryManager_LocalOrdinalTransfer("BlockNumber", multigridAlgo, paramList, defaultList, manager, levelID, keeps);
 
-    // === RAP ===
-    UpdateFactoryManager_RAP(paramList, defaultList, manager, levelID, keeps);
+  // === Coordinates ===
+  UpdateFactoryManager_Coordinates(paramList, defaultList, manager, levelID, keeps);
 
-    // == BlockNumber Transfer ==
-    UpdateFactoryManager_LocalOrdinalTransfer("BlockNumber",multigridAlgo,paramList,defaultList,manager,levelID,keeps);
+  // === Pre-Repartition Keeps for Reuse ===
+  if ((reuseType == "RP" || reuseType == "RAP" || reuseType == "full") && levelID)
+    keeps.push_back(keep_pair("Nullspace", manager.GetFactory("Nullspace").get()));
 
+  if (reuseType == "RP" && levelID) {
+    keeps.push_back(keep_pair("P", manager.GetFactory("P").get()));
+    if (!this->implicitTranspose_)
+      keeps.push_back(keep_pair("R", manager.GetFactory("R").get()));
+  }
+  if ((reuseType == "tP" || reuseType == "RP" || reuseType == "emin") && useCoordinates_ && levelID)
+    keeps.push_back(keep_pair("Coordinates", manager.GetFactory("Coordinates").get()));
 
-    // === Coordinates ===
-    UpdateFactoryManager_Coordinates(paramList, defaultList, manager, levelID, keeps);
+  // === Repartitioning ===
+  UpdateFactoryManager_Repartition(paramList, defaultList, manager, levelID, keeps, nullSpaceFactory);
 
-    // === Pre-Repartition Keeps for Reuse ===
-    if ((reuseType == "RP" || reuseType == "RAP" || reuseType == "full") && levelID)
-      keeps.push_back(keep_pair("Nullspace", manager.GetFactory("Nullspace").get()));
+  // === Lower precision transfers ===
+  UpdateFactoryManager_LowPrecision(paramList, defaultList, manager, levelID, keeps);
 
-    if (reuseType == "RP" && levelID) {
-      keeps.push_back(keep_pair("P", manager.GetFactory("P").get()));
-      if (!this->implicitTranspose_)
-        keeps.push_back(keep_pair("R", manager.GetFactory("R").get()));
-    }
-    if ((reuseType == "tP" || reuseType == "RP" || reuseType == "emin") && useCoordinates_ && levelID)
-      keeps.push_back(keep_pair("Coordinates", manager.GetFactory("Coordinates").get()));
+  // === Final Keeps for Reuse ===
+  if ((reuseType == "RAP" || reuseType == "full") && levelID) {
+    keeps.push_back(keep_pair("P", manager.GetFactory("P").get()));
+    if (!this->implicitTranspose_)
+      keeps.push_back(keep_pair("R", manager.GetFactory("R").get()));
+    keeps.push_back(keep_pair("A", manager.GetFactory("A").get()));
+  }
 
-    // === Repartitioning ===
-    UpdateFactoryManager_Repartition(paramList, defaultList, manager, levelID, keeps, nullSpaceFactory);
+  // In case you ever want to inspect the FactoryManager as it is generated for each level
+  /*std::cout<<"*** Factory Manager on level "<<levelID<<" ***"<<std::endl;
+    manager.Print(); */
+}
+
+// =====================================================================================================
+// ========================================= Smoothers =================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Smoothers(ParameterList& paramList, const ParameterList& defaultList,
+                                   FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const {
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  bool useMaxAbsDiagonalScaling = false;
+  if (defaultList.isParameter("sa: use rowsumabs diagonal scaling"))
+    useMaxAbsDiagonalScaling = defaultList.get<bool>("sa: use rowsumabs diagonal scaling");
+
+  // === Smoothing ===
+  // FIXME: should custom smoother check default list too?
+  bool isCustomSmoother =
+      paramList.isParameter("smoother: pre or post") ||
+      paramList.isParameter("smoother: type") || paramList.isParameter("smoother: pre type") || paramList.isParameter("smoother: post type") ||
+      paramList.isSublist("smoother: params") || paramList.isSublist("smoother: pre params") || paramList.isSublist("smoother: post params") ||
+      paramList.isParameter("smoother: sweeps") || paramList.isParameter("smoother: pre sweeps") || paramList.isParameter("smoother: post sweeps") ||
+      paramList.isParameter("smoother: overlap") || paramList.isParameter("smoother: pre overlap") || paramList.isParameter("smoother: post overlap");
+
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: pre or post", std::string, PreOrPost);
+  if (PreOrPost == "none") {
+    manager.SetFactory("Smoother", Teuchos::null);
+
+  } else if (isCustomSmoother) {
+    // FIXME: get default values from the factory
+    // NOTE: none of the smoothers at the moment use parameter validation framework, so we
+    // cannot get the default values from it.
+#define TEST_MUTUALLY_EXCLUSIVE(arg1, arg2)                                                \
+  TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter(#arg1) && paramList.isParameter(#arg2), \
+                             Exceptions::InvalidArgument, "You cannot specify both \"" #arg1 "\" and \"" #arg2 "\"");
+#define TEST_MUTUALLY_EXCLUSIVE_S(arg1, arg2)                                          \
+  TEUCHOS_TEST_FOR_EXCEPTION(paramList.isSublist(#arg1) && paramList.isSublist(#arg2), \
+                             Exceptions::InvalidArgument, "You cannot specify both \"" #arg1 "\" and \"" #arg2 "\"");
+
+    TEST_MUTUALLY_EXCLUSIVE("smoother: type", "smoother: pre type");
+    TEST_MUTUALLY_EXCLUSIVE("smoother: type", "smoother: post type");
+    TEST_MUTUALLY_EXCLUSIVE("smoother: sweeps", "smoother: pre sweeps");
+    TEST_MUTUALLY_EXCLUSIVE("smoother: sweeps", "smoother: post sweeps");
+    TEST_MUTUALLY_EXCLUSIVE("smoother: overlap", "smoother: pre overlap");
+    TEST_MUTUALLY_EXCLUSIVE("smoother: overlap", "smoother: post overlap");
+    TEST_MUTUALLY_EXCLUSIVE_S("smoother: params", "smoother: pre params");
+    TEST_MUTUALLY_EXCLUSIVE_S("smoother: params", "smoother: post params");
+    TEUCHOS_TEST_FOR_EXCEPTION(PreOrPost == "both" && (paramList.isParameter("smoother: pre type") != paramList.isParameter("smoother: post type")),
+                               Exceptions::InvalidArgument, "You must specify both \"smoother: pre type\" and \"smoother: post type\"");
+
+    // Default values
+    int overlap = 0;
+    ParameterList defaultSmootherParams;
+    defaultSmootherParams.set("relaxation: type", "Symmetric Gauss-Seidel");
+    defaultSmootherParams.set("relaxation: sweeps", Teuchos::OrdinalTraits<LO>::one());
+    defaultSmootherParams.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
+
+    RCP<SmootherFactory> preSmoother = Teuchos::null, postSmoother = Teuchos::null;
+    std::string preSmootherType, postSmootherType;
+    ParameterList preSmootherParams, postSmootherParams;
+
+    if (paramList.isParameter("smoother: overlap"))
+      overlap = paramList.get<int>("smoother: overlap");
+
+    if (PreOrPost == "pre" || PreOrPost == "both") {
+      if (paramList.isParameter("smoother: pre type")) {
+        preSmootherType = paramList.get<std::string>("smoother: pre type");
+      } else {
+        MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: type", std::string, preSmootherTypeTmp);
+        preSmootherType = preSmootherTypeTmp;
+      }
+      if (paramList.isParameter("smoother: pre overlap"))
+        overlap = paramList.get<int>("smoother: pre overlap");
 
-    // === Lower precision transfers ===
-    UpdateFactoryManager_LowPrecision(paramList, defaultList, manager, levelID, keeps);
+      if (paramList.isSublist("smoother: pre params"))
+        preSmootherParams = paramList.sublist("smoother: pre params");
+      else if (paramList.isSublist("smoother: params"))
+        preSmootherParams = paramList.sublist("smoother: params");
+      else if (defaultList.isSublist("smoother: params"))
+        preSmootherParams = defaultList.sublist("smoother: params");
+      else if (preSmootherType == "RELAXATION")
+        preSmootherParams = defaultSmootherParams;
 
-    // === Final Keeps for Reuse ===
-    if ((reuseType == "RAP" || reuseType == "full") && levelID) {
-      keeps.push_back(keep_pair("P", manager.GetFactory("P").get()));
-      if (!this->implicitTranspose_)
-        keeps.push_back(keep_pair("R", manager.GetFactory("R").get()));
-      keeps.push_back(keep_pair("A", manager.GetFactory("A").get()));
-    }
+      if (preSmootherType == "CHEBYSHEV" && useMaxAbsDiagonalScaling)
+        preSmootherParams.set("chebyshev: use rowsumabs diagonal scaling", true);
 
-    // In case you ever want to inspect the FactoryManager as it is generated for each level
-    /*std::cout<<"*** Factory Manager on level "<<levelID<<" ***"<<std::endl;
-      manager.Print(); */
-   }
-
-   // =====================================================================================================
-   // ========================================= Smoothers =================================================
-   // =====================================================================================================
-   template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-   void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-   UpdateFactoryManager_Smoothers(ParameterList& paramList, const ParameterList& defaultList,
-                                  FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const
-   {
-     MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
-     MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-     bool useMaxAbsDiagonalScaling = false;
-     if (defaultList.isParameter("sa: use rowsumabs diagonal scaling"))
-       useMaxAbsDiagonalScaling = defaultList.get<bool>("sa: use rowsumabs diagonal scaling");
-
-     // === Smoothing ===
-     // FIXME: should custom smoother check default list too?
-     bool isCustomSmoother =
-         paramList.isParameter("smoother: pre or post") ||
-         paramList.isParameter("smoother: type")    || paramList.isParameter("smoother: pre type")    || paramList.isParameter("smoother: post type")   ||
-         paramList.isSublist  ("smoother: params")  || paramList.isSublist  ("smoother: pre params")  || paramList.isSublist  ("smoother: post params") ||
-         paramList.isParameter("smoother: sweeps")  || paramList.isParameter("smoother: pre sweeps")  || paramList.isParameter("smoother: post sweeps") ||
-         paramList.isParameter("smoother: overlap") || paramList.isParameter("smoother: pre overlap") || paramList.isParameter("smoother: post overlap");
-
-     MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: pre or post", std::string, PreOrPost);
-     if (PreOrPost == "none") {
-       manager.SetFactory("Smoother", Teuchos::null);
-
-     } else if (isCustomSmoother) {
-       // FIXME: get default values from the factory
-       // NOTE: none of the smoothers at the moment use parameter validation framework, so we
-       // cannot get the default values from it.
- #define TEST_MUTUALLY_EXCLUSIVE(arg1,arg2) \
-       TEUCHOS_TEST_FOR_EXCEPTION(paramList.isParameter(#arg1) && paramList.isParameter(#arg2), \
-                                  Exceptions::InvalidArgument, "You cannot specify both \""#arg1"\" and \""#arg2"\"");
- #define TEST_MUTUALLY_EXCLUSIVE_S(arg1,arg2) \
-       TEUCHOS_TEST_FOR_EXCEPTION(paramList.isSublist(#arg1) && paramList.isSublist(#arg2), \
-                                  Exceptions::InvalidArgument, "You cannot specify both \""#arg1"\" and \""#arg2"\"");
-
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: type",    "smoother: pre type");
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: type",    "smoother: post type");
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: sweeps",  "smoother: pre sweeps");
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: sweeps",  "smoother: post sweeps");
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: overlap", "smoother: pre overlap");
-       TEST_MUTUALLY_EXCLUSIVE  ("smoother: overlap", "smoother: post overlap");
-       TEST_MUTUALLY_EXCLUSIVE_S("smoother: params",  "smoother: pre params");
-       TEST_MUTUALLY_EXCLUSIVE_S("smoother: params",  "smoother: post params");
-       TEUCHOS_TEST_FOR_EXCEPTION(PreOrPost == "both" && (paramList.isParameter("smoother: pre type") != paramList.isParameter("smoother: post type")),
-                                  Exceptions::InvalidArgument, "You must specify both \"smoother: pre type\" and \"smoother: post type\"");
-
-       // Default values
-       int overlap = 0;
-       ParameterList defaultSmootherParams;
-       defaultSmootherParams.set("relaxation: type",           "Symmetric Gauss-Seidel");
-       defaultSmootherParams.set("relaxation: sweeps",         Teuchos::OrdinalTraits<LO>::one());
-       defaultSmootherParams.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
-
-       RCP<SmootherFactory> preSmoother = Teuchos::null, postSmoother = Teuchos::null;
-       std::string          preSmootherType,             postSmootherType;
-       ParameterList        preSmootherParams,           postSmootherParams;
-
-       if (paramList.isParameter("smoother: overlap"))
-         overlap = paramList.get<int>("smoother: overlap");
-
-       if (PreOrPost == "pre" || PreOrPost == "both") {
-         if (paramList.isParameter("smoother: pre type")) {
-           preSmootherType = paramList.get<std::string>("smoother: pre type");
-         } else {
-           MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: type", std::string, preSmootherTypeTmp);
-           preSmootherType = preSmootherTypeTmp;
-         }
-         if (paramList.isParameter("smoother: pre overlap"))
-           overlap = paramList.get<int>("smoother: pre overlap");
-
-         if (paramList.isSublist("smoother: pre params"))
-           preSmootherParams = paramList.sublist("smoother: pre params");
-         else if (paramList.isSublist("smoother: params"))
-           preSmootherParams = paramList.sublist("smoother: params");
-         else if (defaultList.isSublist("smoother: params"))
-           preSmootherParams = defaultList.sublist("smoother: params");
-         else if (preSmootherType == "RELAXATION")
-           preSmootherParams = defaultSmootherParams;
-
-         if (preSmootherType == "CHEBYSHEV" && useMaxAbsDiagonalScaling)
-           preSmootherParams.set("chebyshev: use rowsumabs diagonal scaling",true);
-
- #ifdef HAVE_MUELU_INTREPID2
-       // Propagate P-coarsening for Topo smoothing
-       if (multigridAlgo == "pcoarsen" && preSmootherType == "TOPOLOGICAL" &&
-           defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
-         // P-Coarsening by schedule (new interface)
-         // NOTE: levelID represents the *coarse* level in this case
-         auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList, "pcoarsen: schedule");
-         auto pcoarsen_element  = defaultList.get<std::string>("pcoarsen: element");
-
-         if (levelID < (int)pcoarsen_schedule.size()) {
-           // Topo info for P-Coarsening
-           auto lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
-           preSmootherParams.set("pcoarsen: hi basis", lo);
-         }
-       }
- #endif
-
- #ifdef HAVE_MUELU_MATLAB
-         if (preSmootherType == "matlab")
-           preSmoother = rcp(new SmootherFactory(rcp(new MatlabSmoother(preSmootherParams))));
-         else
- #endif
-         preSmoother = rcp(new SmootherFactory(rcp(new TrilinosSmoother(preSmootherType, preSmootherParams, overlap))));
-       }
-
-       if (PreOrPost == "post" || PreOrPost == "both") {
-         if (paramList.isParameter("smoother: post type"))
-           postSmootherType = paramList.get<std::string>("smoother: post type");
-         else {
-           MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: type", std::string, postSmootherTypeTmp);
-           postSmootherType = postSmootherTypeTmp;
-         }
-
-         if (paramList.isSublist("smoother: post params"))
-           postSmootherParams = paramList.sublist("smoother: post params");
-         else if (paramList.isSublist("smoother: params"))
-           postSmootherParams = paramList.sublist("smoother: params");
-         else if (defaultList.isSublist("smoother: params"))
-           postSmootherParams = defaultList.sublist("smoother: params");
-         else if (postSmootherType == "RELAXATION")
-           postSmootherParams = defaultSmootherParams;
-         if (paramList.isParameter("smoother: post overlap"))
-           overlap = paramList.get<int>("smoother: post overlap");
-
-         if (postSmootherType == "CHEBYSHEV" && useMaxAbsDiagonalScaling)
-           postSmootherParams.set("chebyshev: use rowsumabs diagonal scaling",true);
-
-         if (postSmootherType == preSmootherType && areSame(preSmootherParams, postSmootherParams))
-           postSmoother = preSmoother;
-         else {
- #ifdef HAVE_MUELU_INTREPID2
-           // Propagate P-coarsening for Topo smoothing
-           if (multigridAlgo == "pcoarsen" && preSmootherType == "TOPOLOGICAL" &&
-               defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
-             // P-Coarsening by schedule (new interface)
-             // NOTE: levelID represents the *coarse* level in this case
-             auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList,"pcoarsen: schedule");
-             auto pcoarsen_element = defaultList.get<std::string>("pcoarsen: element");
-
-             if (levelID < (int)pcoarsen_schedule.size()) {
-               // Topo info for P-Coarsening
-               auto lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
-               postSmootherParams.set("pcoarsen: hi basis", lo);
-             }
-           }
- #endif
-
- #ifdef HAVE_MUELU_MATLAB
-           if (postSmootherType == "matlab")
-             postSmoother = rcp(new SmootherFactory(rcp(new MatlabSmoother(postSmootherParams))));
-           else
- #endif
-             postSmoother = rcp(new SmootherFactory(rcp(new TrilinosSmoother(postSmootherType, postSmootherParams, overlap))));
-         }
-       }
-
-       if (preSmoother == postSmoother)
-         manager.SetFactory("Smoother",     preSmoother);
-       else {
-         manager.SetFactory("PreSmoother",  preSmoother);
-         manager.SetFactory("PostSmoother", postSmoother);
-       }
-     }
-
-     // The first clause is not necessary, but it is here for clarity Smoothers
-     // are reused if smoother explicitly said to reuse them, or if any other
-     // reuse option is enabled
-     bool reuseSmoothers = (reuseType == "S" || reuseType != "none");
-     if (reuseSmoothers) {
-       auto preSmootherFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("PreSmoother")));
-
-       if (preSmootherFactory != Teuchos::null) {
-         ParameterList postSmootherFactoryParams;
-         postSmootherFactoryParams.set("keep smoother data", true);
-         preSmootherFactory->SetParameterList(postSmootherFactoryParams);
-
-         keeps.push_back(keep_pair("PreSmoother data", preSmootherFactory.get()));
-       }
-
-       auto postSmootherFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("PostSmoother")));
-       if (postSmootherFactory != Teuchos::null) {
-         ParameterList postSmootherFactoryParams;
-         postSmootherFactoryParams.set("keep smoother data", true);
-         postSmootherFactory->SetParameterList(postSmootherFactoryParams);
-
-         keeps.push_back(keep_pair("PostSmoother data", postSmootherFactory.get()));
-       }
-
-       auto coarseFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("CoarseSolver")));
-       if (coarseFactory != Teuchos::null) {
-         ParameterList coarseFactoryParams;
-         coarseFactoryParams.set("keep smoother data", true);
-         coarseFactory->SetParameterList(coarseFactoryParams);
-
-         keeps.push_back(keep_pair("PreSmoother data", coarseFactory.get()));
-       }
-     }
-
-     if ((reuseType == "RAP" && levelID) || (reuseType == "full")) {
-       // The difference between "RAP" and "full" is keeping smoothers. However,
-       // as in both cases we keep coarse matrices, we do not need to update
-       // coarse smoothers. On the other hand, if a user changes fine level
-       // matrix, "RAP" would update the fine level smoother, while "full" would
-       // not
-       keeps.push_back(keep_pair("PreSmoother",  manager.GetFactory("PreSmoother") .get()));
-       keeps.push_back(keep_pair("PostSmoother", manager.GetFactory("PostSmoother").get()));
-
-       // We do keep_pair("PreSmoother", manager.GetFactory("CoarseSolver").get())
-       // as the coarse solver factory is in fact a smoothing factory, so the
-       // only pieces of data it generates are PreSmoother and PostSmoother
-       keeps.push_back(keep_pair("PreSmoother", manager.GetFactory("CoarseSolver").get()));
-     }
-   }
-
-   // =====================================================================================================
-   // ====================================== Coarse Solvers ===============================================
-   // =====================================================================================================
-   template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-   void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-   UpdateFactoryManager_CoarseSolvers(ParameterList& paramList, const ParameterList& defaultList,
-                                      FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& /* keeps */) const
-   {
-     // FIXME: should custom coarse solver check default list too?
-     bool isCustomCoarseSolver =
-         paramList.isParameter("coarse: type")   ||
-         paramList.isParameter("coarse: params");
-     if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "coarse: type", std::string, "none")) {
-       manager.SetFactory("CoarseSolver", Teuchos::null);
-
-     } else if (isCustomCoarseSolver) {
-       // FIXME: get default values from the factory
-       // NOTE: none of the smoothers at the moment use parameter validation framework, so we
-       // cannot get the default values from it.
-       MUELU_SET_VAR_2LIST(paramList, defaultList, "coarse: type", std::string, coarseType);
-
-       int overlap = 0;
-       if (paramList.isParameter("coarse: overlap"))
-         overlap = paramList.get<int>("coarse: overlap");
-
-       ParameterList coarseParams;
-       if (paramList.isSublist("coarse: params"))
-         coarseParams = paramList.sublist("coarse: params");
-       else if (defaultList.isSublist("coarse: params"))
-         coarseParams = defaultList.sublist("coarse: params");
-
-       using strings = std::unordered_set<std::string>;
-
-       RCP<SmootherPrototype> coarseSmoother;
-       // TODO: this is not a proper place to check. If we consider direct solver to be a special
-       // case of smoother, we would like to unify Amesos and Ifpack2 smoothers in src/Smoothers, and
-       // have a single factory responsible for those. Then, this check would belong there.
-       if (strings({"RELAXATION", "CHEBYSHEV", "ILUT", "ILU", "RILUK", "SCHWARZ", "Amesos",
-         "BLOCK RELAXATION", "BLOCK_RELAXATION", "BLOCKRELAXATION" ,
-         "SPARSE BLOCK RELAXATION", "SPARSE_BLOCK_RELAXATION", "SPARSEBLOCKRELAXATION",
-         "LINESMOOTHING_BANDEDRELAXATION", "LINESMOOTHING_BANDED_RELAXATION", "LINESMOOTHING_BANDED RELAXATION",
-         "LINESMOOTHING_TRIDIRELAXATION", "LINESMOOTHING_TRIDI_RELAXATION", "LINESMOOTHING_TRIDI RELAXATION",
-         "LINESMOOTHING_TRIDIAGONALRELAXATION", "LINESMOOTHING_TRIDIAGONAL_RELAXATION", "LINESMOOTHING_TRIDIAGONAL RELAXATION",
-         "TOPOLOGICAL", "FAST_ILU", "FAST_IC", "FAST_ILDL","HIPTMAIR"}).count(coarseType)) {
-         coarseSmoother = rcp(new TrilinosSmoother(coarseType, coarseParams, overlap));
-       } else {
- #ifdef HAVE_MUELU_MATLAB
-         if (coarseType == "matlab")
-           coarseSmoother = rcp(new MatlabSmoother(coarseParams));
-         else
- #endif
-         coarseSmoother = rcp(new DirectSolver(coarseType, coarseParams));
-       }
-
-       manager.SetFactory("CoarseSolver", rcp(new SmootherFactory(coarseSmoother)));
-     }
-   }
-
-
-    // =====================================================================================================
-   // ========================================= TentativeP=================================================
-   // =====================================================================================================
-   template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-   void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-   UpdateFactoryManager_Reitzinger(ParameterList& paramList, const ParameterList& defaultList,
-                                               FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const
-   {
-     ParameterList rParams;
-     MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: enable", bool, rParams);
-     MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rParams);
-     MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: constant column sums", bool, rParams);
-     MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, rParams);
-
-     RCP<Factory> rFactory = rcp(new ReitzingerPFactory());
-     rFactory->SetParameterList(rParams);
-     
-     // These are all going to be user provided, so NoFactory
-     rFactory->SetFactory("Pnodal", NoFactory::getRCP());
-     rFactory->SetFactory("NodeAggMatrix", NoFactory::getRCP());
-     //rFactory->SetFactory("NodeMatrix", NoFactory::getRCP());
-
-     if(levelID > 1)
-       rFactory->SetFactory("D0", this->GetFactoryManager(levelID-1)->GetFactory("D0"));
-     else
-       rFactory->SetFactory("D0", NoFactory::getRCP());
-
-     manager.SetFactory("Ptent", rFactory);
-     manager.SetFactory("D0", rFactory);
-     manager.SetFactory("InPlaceMap", rFactory);
-
-   }
-
-   // =====================================================================================================
-   // ========================================= TentativeP=================================================
-   // =====================================================================================================
-   template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-   void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-   UpdateFactoryManager_Aggregation_TentativeP(ParameterList& paramList, const ParameterList& defaultList,
-                                               FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const
-   {
-     using strings = std::unordered_set<std::string>;
-
-     MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-
-     MUELU_SET_VAR_2LIST(paramList, defaultList, "aggregation: type", std::string, aggType);
-     TEUCHOS_TEST_FOR_EXCEPTION(!strings({"uncoupled", "coupled", "brick", "matlab","notay","classical"}).count(aggType),
-                                Exceptions::RuntimeError, "Unknown aggregation algorithm: \"" << aggType << "\". Please consult User's Guide.");
-
-
-     // Only doing this for classical because otherwise, the gold tests get broken badly
-     RCP<AmalgamationFactory> amalgFact;
-     if(aggType == "classical") {
-       amalgFact = rcp(new AmalgamationFactory());
-       manager.SetFactory("UnAmalgamationInfo",amalgFact);
-     }
-
-     // Aggregation graph
-     RCP<Factory> dropFactory;
-
-     if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "matlab")) {
- #ifdef HAVE_MUELU_MATLAB
-       dropFactory = rcp(new SingleLevelMatlabFactory());
-       ParameterList socParams  = paramList.sublist("strength-of-connection: params");
-       dropFactory->SetParameterList(socParams);
- #else
-       throw std::runtime_error("Cannot use MATLAB evolutionary strength-of-connection - MueLu was not configured with MATLAB support.");
- #endif
-     } else if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "unsupported vector smoothing")) {
-       dropFactory =   rcp(new MueLu::SmooVecCoalesceDropFactory<SC,LO,GO,NO>());
-       ParameterList dropParams;
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme",             std::string, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: number of random vectors", int, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: number of times to pre or post smooth", int, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: penalty parameters", Teuchos::Array<double>, dropParams);
-       dropFactory->SetParameterList(dropParams);
-     }
-     else {
-       MUELU_KOKKOS_FACTORY_NO_DECL(dropFactory, CoalesceDropFactory, CoalesceDropFactory_kokkos);
-       ParameterList dropParams;
-       if (!rcp_dynamic_cast<CoalesceDropFactory>(dropFactory).is_null())
-         dropParams.set("lightweight wrap", true);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme",             std::string, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: row sum drop tol",        double, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol",                     double, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: use ml scaling of drop tol",   bool, dropParams);
-
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: Dirichlet threshold",          double, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: greedy Dirichlet",          bool, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: distance laplacian algo", std::string, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: classical algo", std::string, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: distance laplacian directional weights",Teuchos::Array<double>,dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring: localize color graph", bool, dropParams);
-       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: dropping may create Dirichlet", bool, dropParams);
-       if (useKokkos_) {
-         MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping",      bool, dropParams);
-         MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph",      bool, dropParams);
-         MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, dropParams);
-       }
-
-       if(!amalgFact.is_null())
-         dropFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
-
-       if(dropParams.isParameter("aggregation: drop scheme")) {
-         std::string drop_scheme = dropParams.get<std::string>("aggregation: drop scheme");
-         if(drop_scheme == "block diagonal colored signed classical")
-           manager.SetFactory("Coloring Graph",dropFactory);
-         if (drop_scheme.find("block diagonal") != std::string::npos || drop_scheme == "signed classical")  {
-           if(levelID > 0)
-             dropFactory->SetFactory("BlockNumber", this->GetFactoryManager(levelID-1)->GetFactory("BlockNumber"));
-           else
-             dropFactory->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
-         }
-       }
-
-      dropFactory->SetParameterList(dropParams);
-    }
-    manager.SetFactory("Graph", dropFactory);
-
-
-    // Aggregation scheme
-    #ifndef HAVE_MUELU_MATLAB
-    if (aggType == "matlab")
-        throw std::runtime_error("Cannot use MATLAB aggregation - MueLu was not configured with MATLAB support.");
-    #endif
-    RCP<Factory> aggFactory;
-    if (aggType == "uncoupled") {
-      MUELU_KOKKOS_FACTORY_NO_DECL(aggFactory, UncoupledAggregationFactory, UncoupledAggregationFactory_kokkos);
-      ParameterList aggParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: mode",               std::string, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: ordering",           std::string, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: min agg size",               int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: max agg size",               int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: max selected neighbors",     int, aggParams);
-      if(useKokkos_) {
-        //if not using kokkos refactor Uncoupled, there is no algorithm option (always Serial)
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase 1 algorithm",  std::string, aggParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: deterministic",             bool, aggParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring algorithm", std::string, aggParams);
-      }
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 1",            bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 2a",           bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 2b",           bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 3",            bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase1",           bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase2a",          bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase2b",          bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase2a agg factor",      double, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: preserve Dirichlet points", bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: error on nodes with no on-rank neighbors", bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase3 avoid singletons", bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities",              bool, aggParams);
-      aggFactory->SetParameterList(aggParams);
-      // make sure that the aggregation factory has all necessary data
-      aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
-      aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
-      //      aggFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
-
-    } else if (aggType == "brick") {
-      aggFactory = rcp(new BrickAggregationFactory());
-      ParameterList aggParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick x size", int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick y size", int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick z size", int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick x Dirichlet", bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick y Dirichlet", bool, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick z Dirichlet", bool, aggParams);
-      aggFactory->SetParameterList(aggParams);
-
-      // Unlike other factories, BrickAggregationFactory makes the Graph/DofsPerNode itself
-      manager.SetFactory("Graph",     aggFactory);
-      manager.SetFactory("DofsPerNode",     aggFactory);
-      manager.SetFactory("Filtering",     aggFactory);
-      if (levelID > 1) {
-        // We check for levelID > 0, as in the interpreter aggFactory for
-        // levelID really corresponds to level 0. Managers are clunky, as they
-        // contain factories for two different levels
-        aggFactory->SetFactory("Coordinates", this->GetFactoryManager(levelID-1)->GetFactory("Coordinates"));
+#ifdef HAVE_MUELU_INTREPID2
+      // Propagate P-coarsening for Topo smoothing
+      if (multigridAlgo == "pcoarsen" && preSmootherType == "TOPOLOGICAL" &&
+          defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
+        // P-Coarsening by schedule (new interface)
+        // NOTE: levelID represents the *coarse* level in this case
+        auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList, "pcoarsen: schedule");
+        auto pcoarsen_element  = defaultList.get<std::string>("pcoarsen: element");
+
+        if (levelID < (int)pcoarsen_schedule.size()) {
+          // Topo info for P-Coarsening
+          auto lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
+          preSmootherParams.set("pcoarsen: hi basis", lo);
+        }
       }
+#endif
+
+#ifdef HAVE_MUELU_MATLAB
+      if (preSmootherType == "matlab")
+        preSmoother = rcp(new SmootherFactory(rcp(new MatlabSmoother(preSmootherParams))));
+      else
+#endif
+        preSmoother = rcp(new SmootherFactory(rcp(new TrilinosSmoother(preSmootherType, preSmootherParams, overlap))));
     }
-    else if (aggType == "classical") {
-      // Map and coloring
-      RCP<Factory> mapFact = rcp(new ClassicalMapFactory());
-      ParameterList mapParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: deterministic",             bool, mapParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring algorithm", std::string, mapParams);
-
-      ParameterList tempParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, tempParams);
-      std::string drop_algo = tempParams.get<std::string>("aggregation: drop scheme");
-      if(drop_algo == "block diagonal colored signed classical") {
-        mapParams.set("aggregation: coloring: use color graph",true);
-        mapFact->SetFactory("Coloring Graph", manager.GetFactory("Coloring Graph"));
 
+    if (PreOrPost == "post" || PreOrPost == "both") {
+      if (paramList.isParameter("smoother: post type"))
+        postSmootherType = paramList.get<std::string>("smoother: post type");
+      else {
+        MUELU_SET_VAR_2LIST(paramList, defaultList, "smoother: type", std::string, postSmootherTypeTmp);
+        postSmootherType = postSmootherTypeTmp;
       }
-      mapFact->SetParameterList(mapParams);
-      mapFact->SetFactory("Graph", manager.GetFactory("Graph"));
-      mapFact->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
-
-      manager.SetFactory("FC Splitting", mapFact);
-      manager.SetFactory("CoarseMap", mapFact);
-
-
-      aggFactory = rcp(new ClassicalPFactory());
-      ParameterList aggParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: classical scheme", std::string, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, aggParams);
-      aggFactory->SetParameterList(aggParams);
-      aggFactory->SetFactory("FC Splitting",manager.GetFactory("FC Splitting"));
-      aggFactory->SetFactory("CoarseMap",manager.GetFactory("CoarseMap"));
-      aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
-      aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
-
-      if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical")  {
-        if(levelID > 0)
-          aggFactory->SetFactory("BlockNumber", this->GetFactoryManager(levelID-1)->GetFactory("BlockNumber"));
-        else
-          aggFactory->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
-      }
-
-      // Now we short-circuit, because we neither need nor want TentativePFactory here
-      manager.SetFactory("Ptent",     aggFactory);
-      manager.SetFactory("P Graph",     aggFactory);
 
+      if (paramList.isSublist("smoother: post params"))
+        postSmootherParams = paramList.sublist("smoother: post params");
+      else if (paramList.isSublist("smoother: params"))
+        postSmootherParams = paramList.sublist("smoother: params");
+      else if (defaultList.isSublist("smoother: params"))
+        postSmootherParams = defaultList.sublist("smoother: params");
+      else if (postSmootherType == "RELAXATION")
+        postSmootherParams = defaultSmootherParams;
+      if (paramList.isParameter("smoother: post overlap"))
+        overlap = paramList.get<int>("smoother: post overlap");
+
+      if (postSmootherType == "CHEBYSHEV" && useMaxAbsDiagonalScaling)
+        postSmootherParams.set("chebyshev: use rowsumabs diagonal scaling", true);
+
+      if (postSmootherType == preSmootherType && areSame(preSmootherParams, postSmootherParams))
+        postSmoother = preSmoother;
+      else {
+#ifdef HAVE_MUELU_INTREPID2
+        // Propagate P-coarsening for Topo smoothing
+        if (multigridAlgo == "pcoarsen" && preSmootherType == "TOPOLOGICAL" &&
+            defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
+          // P-Coarsening by schedule (new interface)
+          // NOTE: levelID represents the *coarse* level in this case
+          auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList, "pcoarsen: schedule");
+          auto pcoarsen_element  = defaultList.get<std::string>("pcoarsen: element");
+
+          if (levelID < (int)pcoarsen_schedule.size()) {
+            // Topo info for P-Coarsening
+            auto lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
+            postSmootherParams.set("pcoarsen: hi basis", lo);
+          }
+        }
+#endif
 
-      if (reuseType == "tP" && levelID) {
-        //        keeps.push_back(keep_pair("Nullspace", Ptent.get()));
-        keeps.push_back(keep_pair("Ptent",aggFactory.get()));
-      }
-      return;
-    }
-    else if (aggType == "notay") {
-      aggFactory = rcp(new NotayAggregationFactory());
-      ParameterList aggParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: pairwise: size",             int, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: pairwise: tie threshold",    double, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: Dirichlet threshold",        double, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: ordering",                   std::string, aggParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities",bool, aggParams);
-      aggFactory->SetParameterList(aggParams);
-      aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
-      aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
-    }
 #ifdef HAVE_MUELU_MATLAB
-    else if(aggType == "matlab") {
-      ParameterList aggParams = paramList.sublist("aggregation: params");
-      aggFactory = rcp(new SingleLevelMatlabFactory());
-      aggFactory->SetParameterList(aggParams);
-    }
+        if (postSmootherType == "matlab")
+          postSmoother = rcp(new SmootherFactory(rcp(new MatlabSmoother(postSmootherParams))));
+        else
 #endif
+          postSmoother = rcp(new SmootherFactory(rcp(new TrilinosSmoother(postSmootherType, postSmootherParams, overlap))));
+      }
+    }
 
+    if (preSmoother == postSmoother)
+      manager.SetFactory("Smoother", preSmoother);
+    else {
+      manager.SetFactory("PreSmoother", preSmoother);
+      manager.SetFactory("PostSmoother", postSmoother);
+    }
+  }
 
+  // The first clause is not necessary, but it is here for clarity Smoothers
+  // are reused if smoother explicitly said to reuse them, or if any other
+  // reuse option is enabled
+  bool reuseSmoothers = (reuseType == "S" || reuseType != "none");
+  if (reuseSmoothers) {
+    auto preSmootherFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("PreSmoother")));
 
-    manager.SetFactory("Aggregates", aggFactory);
+    if (preSmootherFactory != Teuchos::null) {
+      ParameterList postSmootherFactoryParams;
+      postSmootherFactoryParams.set("keep smoother data", true);
+      preSmootherFactory->SetParameterList(postSmootherFactoryParams);
 
-    // Coarse map
-    RCP<Factory> coarseMap = rcp(new CoarseMapFactory());
-    coarseMap->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-    manager.SetFactory("CoarseMap", coarseMap);
+      keeps.push_back(keep_pair("PreSmoother data", preSmootherFactory.get()));
+    }
 
-    // Aggregate qualities
-    if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities", bool, true)) {
-        RCP<Factory> aggQualityFact = rcp(new AggregateQualityEstimateFactory());
-        ParameterList aggQualityParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: good aggregate threshold", double,      aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: file output",              bool,        aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: file base",                std::string, aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: check symmetry",           bool,        aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: algorithm",                std::string, aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: zero threshold",           double,      aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: percentiles", Teuchos::Array<double>,aggQualityParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: mode",                      std::string, aggQualityParams);
-        aggQualityFact->SetParameterList(aggQualityParams);
-        manager.SetFactory("AggregateQualities", aggQualityFact);
+    auto postSmootherFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("PostSmoother")));
+    if (postSmootherFactory != Teuchos::null) {
+      ParameterList postSmootherFactoryParams;
+      postSmootherFactoryParams.set("keep smoother data", true);
+      postSmootherFactory->SetParameterList(postSmootherFactoryParams);
 
-        assert(aggType == "uncoupled");
-        aggFactory->SetFactory("AggregateQualities", aggQualityFact);
+      keeps.push_back(keep_pair("PostSmoother data", postSmootherFactory.get()));
     }
 
+    auto coarseFactory = rcp_const_cast<Factory>(rcp_dynamic_cast<const Factory>(manager.GetFactory("CoarseSolver")));
+    if (coarseFactory != Teuchos::null) {
+      ParameterList coarseFactoryParams;
+      coarseFactoryParams.set("keep smoother data", true);
+      coarseFactory->SetParameterList(coarseFactoryParams);
 
-    // Tentative P
-    MUELU_KOKKOS_FACTORY(Ptent, TentativePFactory, TentativePFactory_kokkos);
-    ParameterList ptentParams;
-    if (paramList.isSublist("matrixmatrix: kernel params"))
-      ptentParams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
-    if (defaultList.isSublist("matrixmatrix: kernel params"))
-      ptentParams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, ptentParams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: build coarse coordinates", bool, ptentParams);
-    Ptent->SetParameterList(ptentParams);
-    Ptent->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-    Ptent->SetFactory("CoarseMap",  manager.GetFactory("CoarseMap"));
-    manager.SetFactory("Ptent",     Ptent);
-
-    if (reuseType == "tP" && levelID) {
-      keeps.push_back(keep_pair("Nullspace", Ptent.get()));
-      keeps.push_back(keep_pair("P",         Ptent.get()));
+      keeps.push_back(keep_pair("PreSmoother data", coarseFactory.get()));
     }
   }
 
-  // =====================================================================================================
-  // ============================================ RAP ====================================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_RAP(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                           int levelID, std::vector<keep_pair>& keeps) const
-  {
-    if (paramList.isParameter("A") && !paramList.get<RCP<Matrix> >("A").is_null()) {
-      // We have user matrix A
-      manager.SetFactory("A", NoFactory::getRCP());
-      return;
-    }
+  if ((reuseType == "RAP" && levelID) || (reuseType == "full")) {
+    // The difference between "RAP" and "full" is keeping smoothers. However,
+    // as in both cases we keep coarse matrices, we do not need to update
+    // coarse smoothers. On the other hand, if a user changes fine level
+    // matrix, "RAP" would update the fine level smoother, while "full" would
+    // not
+    keeps.push_back(keep_pair("PreSmoother", manager.GetFactory("PreSmoother").get()));
+    keeps.push_back(keep_pair("PostSmoother", manager.GetFactory("PostSmoother").get()));
+
+    // We do keep_pair("PreSmoother", manager.GetFactory("CoarseSolver").get())
+    // as the coarse solver factory is in fact a smoothing factory, so the
+    // only pieces of data it generates are PreSmoother and PostSmoother
+    keeps.push_back(keep_pair("PreSmoother", manager.GetFactory("CoarseSolver").get()));
+  }
+}
+
+// =====================================================================================================
+// ====================================== Coarse Solvers ===============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_CoarseSolvers(ParameterList& paramList, const ParameterList& defaultList,
+                                       FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  // FIXME: should custom coarse solver check default list too?
+  bool isCustomCoarseSolver =
+      paramList.isParameter("coarse: type") ||
+      paramList.isParameter("coarse: params");
+  if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "coarse: type", std::string, "none")) {
+    manager.SetFactory("CoarseSolver", Teuchos::null);
+
+  } else if (isCustomCoarseSolver) {
+    // FIXME: get default values from the factory
+    // NOTE: none of the smoothers at the moment use parameter validation framework, so we
+    // cannot get the default values from it.
+    MUELU_SET_VAR_2LIST(paramList, defaultList, "coarse: type", std::string, coarseType);
+
+    int overlap = 0;
+    if (paramList.isParameter("coarse: overlap"))
+      overlap = paramList.get<int>("coarse: overlap");
+
+    ParameterList coarseParams;
+    if (paramList.isSublist("coarse: params"))
+      coarseParams = paramList.sublist("coarse: params");
+    else if (defaultList.isSublist("coarse: params"))
+      coarseParams = defaultList.sublist("coarse: params");
 
-    ParameterList RAPparams;
-
-    RCP<RAPFactory>      RAP;
-    RCP<RAPShiftFactory> RAPs;
-    // Allow for Galerkin or shifted RAP
-    // FIXME: Should this not be some form of MUELU_SET_VAR_2LIST?
-    std::string alg = paramList.get("rap: algorithm", "galerkin");
-    if (alg == "shift" || alg == "non-galerkin") {
-      RAPs = rcp(new RAPShiftFactory());
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift", double, RAPparams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift diagonal M", bool, RAPparams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift low storage", bool, RAPparams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift array", Teuchos::Array<double>, RAPparams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: cfl array", Teuchos::Array<double>, RAPparams);
+    using strings = std::unordered_set<std::string>;
 
+    RCP<SmootherPrototype> coarseSmoother;
+    // TODO: this is not a proper place to check. If we consider direct solver to be a special
+    // case of smoother, we would like to unify Amesos and Ifpack2 smoothers in src/Smoothers, and
+    // have a single factory responsible for those. Then, this check would belong there.
+    if (strings({"RELAXATION", "CHEBYSHEV", "ILUT", "ILU", "RILUK", "SCHWARZ", "Amesos",
+                 "BLOCK RELAXATION", "BLOCK_RELAXATION", "BLOCKRELAXATION",
+                 "SPARSE BLOCK RELAXATION", "SPARSE_BLOCK_RELAXATION", "SPARSEBLOCKRELAXATION",
+                 "LINESMOOTHING_BANDEDRELAXATION", "LINESMOOTHING_BANDED_RELAXATION", "LINESMOOTHING_BANDED RELAXATION",
+                 "LINESMOOTHING_TRIDIRELAXATION", "LINESMOOTHING_TRIDI_RELAXATION", "LINESMOOTHING_TRIDI RELAXATION",
+                 "LINESMOOTHING_TRIDIAGONALRELAXATION", "LINESMOOTHING_TRIDIAGONAL_RELAXATION", "LINESMOOTHING_TRIDIAGONAL RELAXATION",
+                 "TOPOLOGICAL", "FAST_ILU", "FAST_IC", "FAST_ILDL", "HIPTMAIR"})
+            .count(coarseType)) {
+      coarseSmoother = rcp(new TrilinosSmoother(coarseType, coarseParams, overlap));
     } else {
-      RAP = rcp(new RAPFactory());
+#ifdef HAVE_MUELU_MATLAB
+      if (coarseType == "matlab")
+        coarseSmoother = rcp(new MatlabSmoother(coarseParams));
+      else
+#endif
+        coarseSmoother = rcp(new DirectSolver(coarseType, coarseParams));
     }
 
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: relative diagonal floor", Teuchos::Array<double>, RAPparams);
-
-    if (paramList.isSublist("matrixmatrix: kernel params"))
-      RAPparams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
-    if (defaultList.isSublist("matrixmatrix: kernel params"))
-      RAPparams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "transpose: use implicit", bool, RAPparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals", bool, RAPparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals threshold", double, RAPparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals replacement", Scalar, RAPparams);
-
-    // if "rap: triple product" has not been set and algorithm is "unsmoothed" switch triple product on
-    if (!paramList.isParameter("rap: triple product") &&
-        paramList.isType<std::string>("multigrid algorithm") &&
-        paramList.get<std::string>("multigrid algorithm") == "unsmoothed")
-      paramList.set("rap: triple product", true);
-    else
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: triple product", bool, RAPparams);
+    manager.SetFactory("CoarseSolver", rcp(new SmootherFactory(coarseSmoother)));
+  }
+}
+
+// =====================================================================================================
+// ========================================= TentativeP=================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Reitzinger(ParameterList& paramList, const ParameterList& defaultList,
+                                    FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const {
+  ParameterList rParams;
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: enable", bool, rParams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rParams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: constant column sums", bool, rParams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, rParams);
+
+  RCP<Factory> rFactory = rcp(new ReitzingerPFactory());
+  rFactory->SetParameterList(rParams);
+
+  // These are all going to be user provided, so NoFactory
+  rFactory->SetFactory("Pnodal", NoFactory::getRCP());
+  rFactory->SetFactory("NodeAggMatrix", NoFactory::getRCP());
+  // rFactory->SetFactory("NodeMatrix", NoFactory::getRCP());
+
+  if (levelID > 1)
+    rFactory->SetFactory("D0", this->GetFactoryManager(levelID - 1)->GetFactory("D0"));
+  else
+    rFactory->SetFactory("D0", NoFactory::getRCP());
+
+  manager.SetFactory("Ptent", rFactory);
+  manager.SetFactory("D0", rFactory);
+  manager.SetFactory("InPlaceMap", rFactory);
+}
+
+// =====================================================================================================
+// ========================================= TentativeP=================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Aggregation_TentativeP(ParameterList& paramList, const ParameterList& defaultList,
+                                                FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const {
+  using strings = std::unordered_set<std::string>;
+
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "aggregation: type", std::string, aggType);
+  TEUCHOS_TEST_FOR_EXCEPTION(!strings({"uncoupled", "coupled", "brick", "matlab", "notay", "classical"}).count(aggType),
+                             Exceptions::RuntimeError, "Unknown aggregation algorithm: \"" << aggType << "\". Please consult User's Guide.");
+
+  // Only doing this for classical because otherwise, the gold tests get broken badly
+  RCP<AmalgamationFactory> amalgFact;
+  if (aggType == "classical") {
+    amalgFact = rcp(new AmalgamationFactory());
+    manager.SetFactory("UnAmalgamationInfo", amalgFact);
+  }
 
-    try {
-      if (paramList.isParameter("aggregation: allow empty prolongator columns")) {
-        RAPparams.set("CheckMainDiagonal",  paramList.get<bool>("aggregation: allow empty prolongator columns"));
-        RAPparams.set("RepairMainDiagonal", paramList.get<bool>("aggregation: allow empty prolongator columns"));
-      }
-      else if (defaultList.isParameter("aggregation: allow empty prolongator columns")) {
-        RAPparams.set("CheckMainDiagonal",  defaultList.get<bool>("aggregation: allow empty prolongator columns"));
-        RAPparams.set("RepairMainDiagonal", defaultList.get<bool>("aggregation: allow empty prolongator columns"));
-      }
+  // Aggregation graph
+  RCP<Factory> dropFactory;
 
-    } catch (Teuchos::Exceptions::InvalidParameterType&) {
-      TEUCHOS_TEST_FOR_EXCEPTION_PURE_MSG(true, Teuchos::Exceptions::InvalidParameterType,
-          "Error: parameter \"aggregation: allow empty prolongator columns\" must be of type " << Teuchos::TypeNameTraits<bool>::name());
+  if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "matlab")) {
+#ifdef HAVE_MUELU_MATLAB
+    dropFactory             = rcp(new SingleLevelMatlabFactory());
+    ParameterList socParams = paramList.sublist("strength-of-connection: params");
+    dropFactory->SetParameterList(socParams);
+#else
+    throw std::runtime_error("Cannot use MATLAB evolutionary strength-of-connection - MueLu was not configured with MATLAB support.");
+#endif
+  } else if (MUELU_TEST_PARAM_2LIST(paramList, paramList, "aggregation: drop scheme", std::string, "unsupported vector smoothing")) {
+    dropFactory = rcp(new MueLu::SmooVecCoalesceDropFactory<SC, LO, GO, NO>());
+    ParameterList dropParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: number of random vectors", int, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: number of times to pre or post smooth", int, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: penalty parameters", Teuchos::Array<double>, dropParams);
+    dropFactory->SetParameterList(dropParams);
+  } else {
+    MUELU_KOKKOS_FACTORY_NO_DECL(dropFactory, CoalesceDropFactory, CoalesceDropFactory_kokkos);
+    ParameterList dropParams;
+    if (!rcp_dynamic_cast<CoalesceDropFactory>(dropFactory).is_null())
+      dropParams.set("lightweight wrap", true);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: row sum drop tol", double, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol", double, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: use ml scaling of drop tol", bool, dropParams);
+
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: Dirichlet threshold", double, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: greedy Dirichlet", bool, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: distance laplacian algo", std::string, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: classical algo", std::string, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: distance laplacian directional weights", Teuchos::Array<double>, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring: localize color graph", bool, dropParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: dropping may create Dirichlet", bool, dropParams);
+    if (useKokkos_) {
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping", bool, dropParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph", bool, dropParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, dropParams);
     }
 
-    if (!RAP.is_null()) {
-      RAP->SetParameterList(RAPparams);
-      RAP->SetFactory("P", manager.GetFactory("P"));
-    } else {
-      RAPs->SetParameterList(RAPparams);
-      RAPs->SetFactory("P", manager.GetFactory("P"));
-    }
+    if (!amalgFact.is_null())
+      dropFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
 
-    if (!this->implicitTranspose_) {
-       if (!RAP.is_null())
-         RAP->SetFactory("R", manager.GetFactory("R"));
-       else
-         RAPs->SetFactory("R", manager.GetFactory("R"));
+    if (dropParams.isParameter("aggregation: drop scheme")) {
+      std::string drop_scheme = dropParams.get<std::string>("aggregation: drop scheme");
+      if (drop_scheme == "block diagonal colored signed classical")
+        manager.SetFactory("Coloring Graph", dropFactory);
+      if (drop_scheme.find("block diagonal") != std::string::npos || drop_scheme == "signed classical") {
+        if (levelID > 0)
+          dropFactory->SetFactory("BlockNumber", this->GetFactoryManager(levelID - 1)->GetFactory("BlockNumber"));
+        else
+          dropFactory->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
+      }
     }
 
-    if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: export visualization data", bool, true)) {
-      RCP<AggregationExportFactory> aggExport = rcp(new AggregationExportFactory());
-      ParameterList aggExportParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output filename",             std::string, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: agg style",      std::string, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: iter",                   int, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: time step",              int, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: fine graph edges",      bool, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: coarse graph edges",    bool, aggExportParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: build colormap",        bool, aggExportParams);
-      aggExport->SetParameterList(aggExportParams);
-      aggExport->SetFactory("DofsPerNode", manager.GetFactory("DofsPerNode"));
+    dropFactory->SetParameterList(dropParams);
+  }
+  manager.SetFactory("Graph", dropFactory);
 
-      if (!RAP.is_null())
-        RAP->AddTransferFactory(aggExport);
+// Aggregation scheme
+#ifndef HAVE_MUELU_MATLAB
+  if (aggType == "matlab")
+    throw std::runtime_error("Cannot use MATLAB aggregation - MueLu was not configured with MATLAB support.");
+#endif
+  RCP<Factory> aggFactory;
+  if (aggType == "uncoupled") {
+    MUELU_KOKKOS_FACTORY_NO_DECL(aggFactory, UncoupledAggregationFactory, UncoupledAggregationFactory_kokkos);
+    ParameterList aggParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: mode", std::string, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: ordering", std::string, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: min agg size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: max agg size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: max selected neighbors", int, aggParams);
+    if (useKokkos_) {
+      // if not using kokkos refactor Uncoupled, there is no algorithm option (always Serial)
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase 1 algorithm", std::string, aggParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: deterministic", bool, aggParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring algorithm", std::string, aggParams);
+    }
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 1", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 2a", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 2b", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: enable phase 3", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase1", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase2a", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: match ML phase2b", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase2a agg factor", double, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: preserve Dirichlet points", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: error on nodes with no on-rank neighbors", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: phase3 avoid singletons", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities", bool, aggParams);
+    aggFactory->SetParameterList(aggParams);
+    // make sure that the aggregation factory has all necessary data
+    aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
+    aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
+    //      aggFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
+
+  } else if (aggType == "brick") {
+    aggFactory = rcp(new BrickAggregationFactory());
+    ParameterList aggParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick x size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick y size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick z size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick x Dirichlet", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick y Dirichlet", bool, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: brick z Dirichlet", bool, aggParams);
+    aggFactory->SetParameterList(aggParams);
+
+    // Unlike other factories, BrickAggregationFactory makes the Graph/DofsPerNode itself
+    manager.SetFactory("Graph", aggFactory);
+    manager.SetFactory("DofsPerNode", aggFactory);
+    manager.SetFactory("Filtering", aggFactory);
+    if (levelID > 1) {
+      // We check for levelID > 0, as in the interpreter aggFactory for
+      // levelID really corresponds to level 0. Managers are clunky, as they
+      // contain factories for two different levels
+      aggFactory->SetFactory("Coordinates", this->GetFactoryManager(levelID - 1)->GetFactory("Coordinates"));
+    }
+  } else if (aggType == "classical") {
+    // Map and coloring
+    RCP<Factory> mapFact = rcp(new ClassicalMapFactory());
+    ParameterList mapParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: deterministic", bool, mapParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: coloring algorithm", std::string, mapParams);
+
+    ParameterList tempParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, tempParams);
+    std::string drop_algo = tempParams.get<std::string>("aggregation: drop scheme");
+    if (drop_algo == "block diagonal colored signed classical") {
+      mapParams.set("aggregation: coloring: use color graph", true);
+      mapFact->SetFactory("Coloring Graph", manager.GetFactory("Coloring Graph"));
+    }
+    mapFact->SetParameterList(mapParams);
+    mapFact->SetFactory("Graph", manager.GetFactory("Graph"));
+    mapFact->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
+
+    manager.SetFactory("FC Splitting", mapFact);
+    manager.SetFactory("CoarseMap", mapFact);
+
+    aggFactory = rcp(new ClassicalPFactory());
+    ParameterList aggParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: classical scheme", std::string, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: drop scheme", std::string, aggParams);
+    aggFactory->SetParameterList(aggParams);
+    aggFactory->SetFactory("FC Splitting", manager.GetFactory("FC Splitting"));
+    aggFactory->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
+    aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
+    aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
+
+    if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical") {
+      if (levelID > 0)
+        aggFactory->SetFactory("BlockNumber", this->GetFactoryManager(levelID - 1)->GetFactory("BlockNumber"));
       else
-        RAPs->AddTransferFactory(aggExport);
+        aggFactory->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
     }
-    if (!RAP.is_null())
-      manager.SetFactory("A", RAP);
-    else
-      manager.SetFactory("A", RAPs);
 
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type",      std::string, reuseType);
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "sa: use filtered matrix", bool, useFiltering);
-    bool filteringChangesMatrix = useFiltering && !MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol", double, 0);
+    // Now we short-circuit, because we neither need nor want TentativePFactory here
+    manager.SetFactory("Ptent", aggFactory);
+    manager.SetFactory("P Graph", aggFactory);
 
-    if (reuseType == "RP" || (reuseType == "tP" && !filteringChangesMatrix)) {
-      if (!RAP.is_null()) {
-        keeps.push_back(keep_pair("AP reuse data",  RAP.get()));
-        keeps.push_back(keep_pair("RAP reuse data", RAP.get()));
-
-      } else {
-        keeps.push_back(keep_pair("AP reuse data",  RAPs.get()));
-        keeps.push_back(keep_pair("RAP reuse data", RAPs.get()));
-      }
+    if (reuseType == "tP" && levelID) {
+      //        keeps.push_back(keep_pair("Nullspace", Ptent.get()));
+      keeps.push_back(keep_pair("Ptent", aggFactory.get()));
     }
+    return;
+  } else if (aggType == "notay") {
+    aggFactory = rcp(new NotayAggregationFactory());
+    ParameterList aggParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: pairwise: size", int, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: pairwise: tie threshold", double, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: Dirichlet threshold", double, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: ordering", std::string, aggParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities", bool, aggParams);
+    aggFactory->SetParameterList(aggParams);
+    aggFactory->SetFactory("DofsPerNode", manager.GetFactory("Graph"));
+    aggFactory->SetFactory("Graph", manager.GetFactory("Graph"));
+  }
+#ifdef HAVE_MUELU_MATLAB
+  else if (aggType == "matlab") {
+    ParameterList aggParams = paramList.sublist("aggregation: params");
+    aggFactory              = rcp(new SingleLevelMatlabFactory());
+    aggFactory->SetParameterList(aggParams);
   }
+#endif
 
-  // =====================================================================================================
-  // ======================================= Coordinates =================================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Coordinates(ParameterList& paramList, const ParameterList& /* defaultList */,
-                                   FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& /* keeps */) const
-  {
-    bool have_userCO = false;
-    if (paramList.isParameter("Coordinates") && !paramList.get<RCP<MultiVector> >("Coordinates").is_null())
-      have_userCO = true;
+  manager.SetFactory("Aggregates", aggFactory);
+
+  // Coarse map
+  RCP<Factory> coarseMap = rcp(new CoarseMapFactory());
+  coarseMap->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+  manager.SetFactory("CoarseMap", coarseMap);
+
+  // Aggregate qualities
+  if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: compute aggregate qualities", bool, true)) {
+    RCP<Factory> aggQualityFact = rcp(new AggregateQualityEstimateFactory());
+    ParameterList aggQualityParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: good aggregate threshold", double, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: file output", bool, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: file base", std::string, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: check symmetry", bool, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: algorithm", std::string, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: zero threshold", double, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: percentiles", Teuchos::Array<double>, aggQualityParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregate qualities: mode", std::string, aggQualityParams);
+    aggQualityFact->SetParameterList(aggQualityParams);
+    manager.SetFactory("AggregateQualities", aggQualityFact);
+
+    assert(aggType == "uncoupled");
+    aggFactory->SetFactory("AggregateQualities", aggQualityFact);
+  }
 
-    if (useCoordinates_) {
-      if (have_userCO) {
-        manager.SetFactory("Coordinates", NoFactory::getRCP());
+  // Tentative P
+  MUELU_KOKKOS_FACTORY(Ptent, TentativePFactory, TentativePFactory_kokkos);
+  ParameterList ptentParams;
+  if (paramList.isSublist("matrixmatrix: kernel params"))
+    ptentParams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
+  if (defaultList.isSublist("matrixmatrix: kernel params"))
+    ptentParams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, ptentParams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: build coarse coordinates", bool, ptentParams);
+  Ptent->SetParameterList(ptentParams);
+  Ptent->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+  Ptent->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
+  manager.SetFactory("Ptent", Ptent);
+
+  if (reuseType == "tP" && levelID) {
+    keeps.push_back(keep_pair("Nullspace", Ptent.get()));
+    keeps.push_back(keep_pair("P", Ptent.get()));
+  }
+}
+
+// =====================================================================================================
+// ============================================ RAP ====================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_RAP(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                             int levelID, std::vector<keep_pair>& keeps) const {
+  if (paramList.isParameter("A") && !paramList.get<RCP<Matrix> >("A").is_null()) {
+    // We have user matrix A
+    manager.SetFactory("A", NoFactory::getRCP());
+    return;
+  }
 
-      } else {
-        RCP<Factory> coords = rcp(new CoordinatesTransferFactory());
-        coords->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-        coords->SetFactory("CoarseMap",  manager.GetFactory("CoarseMap"));
-        manager.SetFactory("Coordinates", coords);
-
-        auto RAP = rcp_const_cast<RAPFactory>(rcp_dynamic_cast<const RAPFactory>(manager.GetFactory("A")));
-        if (!RAP.is_null()) {
-          RAP->AddTransferFactory(manager.GetFactory("Coordinates"));
-        } else {
-          auto RAPs = rcp_const_cast<RAPShiftFactory>(rcp_dynamic_cast<const RAPShiftFactory>(manager.GetFactory("A")));
-          RAPs->AddTransferFactory(manager.GetFactory("Coordinates"));
-        }
-      }
+  ParameterList RAPparams;
+
+  RCP<RAPFactory> RAP;
+  RCP<RAPShiftFactory> RAPs;
+  // Allow for Galerkin or shifted RAP
+  // FIXME: Should this not be some form of MUELU_SET_VAR_2LIST?
+  std::string alg = paramList.get("rap: algorithm", "galerkin");
+  if (alg == "shift" || alg == "non-galerkin") {
+    RAPs = rcp(new RAPShiftFactory());
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift", double, RAPparams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift diagonal M", bool, RAPparams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift low storage", bool, RAPparams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: shift array", Teuchos::Array<double>, RAPparams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: cfl array", Teuchos::Array<double>, RAPparams);
+
+  } else {
+    RAP = rcp(new RAPFactory());
+  }
+
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: relative diagonal floor", Teuchos::Array<double>, RAPparams);
+
+  if (paramList.isSublist("matrixmatrix: kernel params"))
+    RAPparams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
+  if (defaultList.isSublist("matrixmatrix: kernel params"))
+    RAPparams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "transpose: use implicit", bool, RAPparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals", bool, RAPparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals threshold", double, RAPparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: fix zero diagonals replacement", Scalar, RAPparams);
+
+  // if "rap: triple product" has not been set and algorithm is "unsmoothed" switch triple product on
+  if (!paramList.isParameter("rap: triple product") &&
+      paramList.isType<std::string>("multigrid algorithm") &&
+      paramList.get<std::string>("multigrid algorithm") == "unsmoothed")
+    paramList.set("rap: triple product", true);
+  else
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "rap: triple product", bool, RAPparams);
+
+  try {
+    if (paramList.isParameter("aggregation: allow empty prolongator columns")) {
+      RAPparams.set("CheckMainDiagonal", paramList.get<bool>("aggregation: allow empty prolongator columns"));
+      RAPparams.set("RepairMainDiagonal", paramList.get<bool>("aggregation: allow empty prolongator columns"));
+    } else if (defaultList.isParameter("aggregation: allow empty prolongator columns")) {
+      RAPparams.set("CheckMainDiagonal", defaultList.get<bool>("aggregation: allow empty prolongator columns"));
+      RAPparams.set("RepairMainDiagonal", defaultList.get<bool>("aggregation: allow empty prolongator columns"));
     }
+
+  } catch (Teuchos::Exceptions::InvalidParameterType&) {
+    TEUCHOS_TEST_FOR_EXCEPTION_PURE_MSG(true, Teuchos::Exceptions::InvalidParameterType,
+                                        "Error: parameter \"aggregation: allow empty prolongator columns\" must be of type " << Teuchos::TypeNameTraits<bool>::name());
   }
 
-  // =====================================================================================================
-  // =================================  LocalOrdinalTransfer =============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_LocalOrdinalTransfer(const std::string & VarName, const std::string &multigridAlgo,ParameterList& paramList, const ParameterList& /* defaultList */,
-                                            FactoryManager& manager, int levelID, std::vector<keep_pair>& /* keeps */) const
-  {
-    // NOTE: You would think this would be levelID > 0, but you'd be wrong, since the FactoryManager is basically
-    // offset by a level from the things which actually do the work.
-    if (useBlockNumber_ && (levelID > 0)) {
-      auto RAP = rcp_const_cast<RAPFactory>(rcp_dynamic_cast<const RAPFactory>(manager.GetFactory("A")));
-      auto RAPs = rcp_const_cast<RAPShiftFactory>(rcp_dynamic_cast<const RAPShiftFactory>(manager.GetFactory("A")));
-      if (!RAP.is_null() || !RAPs.is_null()) {
-        RCP<Factory> fact = rcp(new LocalOrdinalTransferFactory(VarName,multigridAlgo));
-        if(multigridAlgo == "classical")
-          fact->SetFactory("P Graph", manager.GetFactory("P Graph"));
-        else
-          fact->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-        fact->SetFactory("CoarseMap",  manager.GetFactory("CoarseMap"));
+  if (!RAP.is_null()) {
+    RAP->SetParameterList(RAPparams);
+    RAP->SetFactory("P", manager.GetFactory("P"));
+  } else {
+    RAPs->SetParameterList(RAPparams);
+    RAPs->SetFactory("P", manager.GetFactory("P"));
+  }
 
-        fact->SetFactory(VarName, this->GetFactoryManager(levelID-1)->GetFactory(VarName));
+  if (!this->implicitTranspose_) {
+    if (!RAP.is_null())
+      RAP->SetFactory("R", manager.GetFactory("R"));
+    else
+      RAPs->SetFactory("R", manager.GetFactory("R"));
+  }
 
-        manager.SetFactory(VarName, fact);
+  if (MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: export visualization data", bool, true)) {
+    RCP<AggregationExportFactory> aggExport = rcp(new AggregationExportFactory());
+    ParameterList aggExportParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output filename", std::string, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: agg style", std::string, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: iter", int, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: time step", int, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: fine graph edges", bool, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: coarse graph edges", bool, aggExportParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: output file: build colormap", bool, aggExportParams);
+    aggExport->SetParameterList(aggExportParams);
+    aggExport->SetFactory("DofsPerNode", manager.GetFactory("DofsPerNode"));
 
-        if (!RAP.is_null())
-          RAP->AddTransferFactory(manager.GetFactory(VarName));
-        else
-          RAPs->AddTransferFactory(manager.GetFactory(VarName));
-      }
-    }
+    if (!RAP.is_null())
+      RAP->AddTransferFactory(aggExport);
+    else
+      RAPs->AddTransferFactory(aggExport);
   }
+  if (!RAP.is_null())
+    manager.SetFactory("A", RAP);
+  else
+    manager.SetFactory("A", RAPs);
 
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "sa: use filtered matrix", bool, useFiltering);
+  bool filteringChangesMatrix = useFiltering && !MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol", double, 0);
 
- // ======================================================================================================
-  // ======================================  BlockNumber =================================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_BlockNumber(ParameterList& paramList, const ParameterList& defaultList,
-                                   FactoryManager& manager, int levelID , std::vector<keep_pair>& keeps) const
-  {
-    if(useBlockNumber_) {
-      ParameterList myParams;
-      RCP<Factory> fact = rcp(new InitialBlockNumberFactory());
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, myParams);
-      fact->SetParameterList(myParams);
-      manager.SetFactory("BlockNumber",fact);
-    }
+  if (reuseType == "RP" || (reuseType == "tP" && !filteringChangesMatrix)) {
+    if (!RAP.is_null()) {
+      keeps.push_back(keep_pair("AP reuse data", RAP.get()));
+      keeps.push_back(keep_pair("RAP reuse data", RAP.get()));
 
+    } else {
+      keeps.push_back(keep_pair("AP reuse data", RAPs.get()));
+      keeps.push_back(keep_pair("RAP reuse data", RAPs.get()));
+    }
   }
+}
+
+// =====================================================================================================
+// ======================================= Coordinates =================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Coordinates(ParameterList& paramList, const ParameterList& /* defaultList */,
+                                     FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  bool have_userCO = false;
+  if (paramList.isParameter("Coordinates") && !paramList.get<RCP<MultiVector> >("Coordinates").is_null())
+    have_userCO = true;
+
+  if (useCoordinates_) {
+    if (have_userCO) {
+      manager.SetFactory("Coordinates", NoFactory::getRCP());
 
+    } else {
+      RCP<Factory> coords = rcp(new CoordinatesTransferFactory());
+      coords->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+      coords->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
+      manager.SetFactory("Coordinates", coords);
 
-  // =====================================================================================================
-  // =========================================== Restriction =============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Restriction(ParameterList& paramList, const ParameterList& defaultList , FactoryManager& manager,
-                                 int levelID, std::vector<keep_pair>& /* keeps */) const
-  {
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
-    bool have_userR = false;
-    if (paramList.isParameter("R") && !paramList.get<RCP<Matrix> >("R").is_null())
-      have_userR = true;
-
-    // === Restriction ===
-    RCP<Factory> R;
-    if (!this->implicitTranspose_) {
-      MUELU_SET_VAR_2LIST(paramList, defaultList, "problem: symmetric", bool, isSymmetric);
-
-      if (isSymmetric == false && (multigridAlgo == "unsmoothed" || multigridAlgo == "emin")) {
-        this->GetOStream(Warnings0) <<
-            "Switching \"problem: symmetric\" parameter to symmetric as multigrid algorithm. " <<
-            multigridAlgo << " is primarily supposed to be used for symmetric problems.\n\n" <<
-            "Please note: if you are using \"unsmoothed\" transfer operators the \"problem: symmetric\" parameter " <<
-            "has no real mathematical meaning, i.e. you can use it for non-symmetric\n" <<
-            "problems, too. With \"problem: symmetric\"=\"symmetric\" you can use implicit transpose for building " <<
-            "the restriction operators which may drastically reduce the amount of consumed memory." << std::endl;
-        isSymmetric = true;
+      auto RAP = rcp_const_cast<RAPFactory>(rcp_dynamic_cast<const RAPFactory>(manager.GetFactory("A")));
+      if (!RAP.is_null()) {
+        RAP->AddTransferFactory(manager.GetFactory("Coordinates"));
+      } else {
+        auto RAPs = rcp_const_cast<RAPShiftFactory>(rcp_dynamic_cast<const RAPShiftFactory>(manager.GetFactory("A")));
+        RAPs->AddTransferFactory(manager.GetFactory("Coordinates"));
       }
-      TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "pg" && isSymmetric == true, Exceptions::RuntimeError,
-          "Petrov-Galerkin smoothed transfer operators are only allowed for non-symmetric problems: Set \"problem: symmetric\" to false!\n" \
-          "While PG smoothed transfer operators generally would also work for symmetric problems this is an unusual use case. " \
-          "You can use the factory-based xml interface though if you need PG-AMG for symmetric problems.");
+    }
+  }
+}
+
+// =====================================================================================================
+// =================================  LocalOrdinalTransfer =============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_LocalOrdinalTransfer(const std::string& VarName, const std::string& multigridAlgo, ParameterList& paramList, const ParameterList& /* defaultList */,
+                                              FactoryManager& manager, int levelID, std::vector<keep_pair>& /* keeps */) const {
+  // NOTE: You would think this would be levelID > 0, but you'd be wrong, since the FactoryManager is basically
+  // offset by a level from the things which actually do the work.
+  if (useBlockNumber_ && (levelID > 0)) {
+    auto RAP  = rcp_const_cast<RAPFactory>(rcp_dynamic_cast<const RAPFactory>(manager.GetFactory("A")));
+    auto RAPs = rcp_const_cast<RAPShiftFactory>(rcp_dynamic_cast<const RAPShiftFactory>(manager.GetFactory("A")));
+    if (!RAP.is_null() || !RAPs.is_null()) {
+      RCP<Factory> fact = rcp(new LocalOrdinalTransferFactory(VarName, multigridAlgo));
+      if (multigridAlgo == "classical")
+        fact->SetFactory("P Graph", manager.GetFactory("P Graph"));
+      else
+        fact->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+      fact->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
 
-        if (have_userR) {
-          manager.SetFactory("R", NoFactory::getRCP());
-        } else {
-          if (isSymmetric)  R = rcp(new TransPFactory());
-          else              R = rcp(new GenericRFactory());
+      fact->SetFactory(VarName, this->GetFactoryManager(levelID - 1)->GetFactory(VarName));
 
-          R->SetFactory("P", manager.GetFactory("P"));
-          manager.SetFactory("R", R);
-        }
+      manager.SetFactory(VarName, fact);
 
-    } else {
-      manager.SetFactory("R", Teuchos::null);
+      if (!RAP.is_null())
+        RAP->AddTransferFactory(manager.GetFactory(VarName));
+      else
+        RAPs->AddTransferFactory(manager.GetFactory(VarName));
     }
-
-    // === Restriction: Nullspace Scaling ===
-    if (paramList.isParameter("restriction: scale nullspace") && paramList.get<bool>("restriction: scale nullspace")) {
-      RCP<TentativePFactory> tentPFactory = rcp(new TentativePFactory());
-      Teuchos::ParameterList tentPlist;
-      tentPlist.set("Nullspace name","Scaled Nullspace");
-      tentPFactory->SetParameterList(tentPlist);
-      tentPFactory->SetFactory("Aggregates",manager.GetFactory("Aggregates"));
-      tentPFactory->SetFactory("CoarseMap",manager.GetFactory("CoarseMap"));
-
-      if(R.is_null())   R = rcp(new TransPFactory());
-      R->SetFactory("P",tentPFactory);
+  }
+}
+
+// ======================================================================================================
+// ======================================  BlockNumber =================================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_BlockNumber(ParameterList& paramList, const ParameterList& defaultList,
+                                     FactoryManager& manager, int levelID, std::vector<keep_pair>& keeps) const {
+  if (useBlockNumber_) {
+    ParameterList myParams;
+    RCP<Factory> fact = rcp(new InitialBlockNumberFactory());
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "aggregation: block diagonal: interleaved blocksize", int, myParams);
+    fact->SetParameterList(myParams);
+    manager.SetFactory("BlockNumber", fact);
+  }
+}
+
+// =====================================================================================================
+// =========================================== Restriction =============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Restriction(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                     int levelID, std::vector<keep_pair>& /* keeps */) const {
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "multigrid algorithm", std::string, multigridAlgo);
+  bool have_userR = false;
+  if (paramList.isParameter("R") && !paramList.get<RCP<Matrix> >("R").is_null())
+    have_userR = true;
+
+  // === Restriction ===
+  RCP<Factory> R;
+  if (!this->implicitTranspose_) {
+    MUELU_SET_VAR_2LIST(paramList, defaultList, "problem: symmetric", bool, isSymmetric);
+
+    if (isSymmetric == false && (multigridAlgo == "unsmoothed" || multigridAlgo == "emin")) {
+      this->GetOStream(Warnings0) << "Switching \"problem: symmetric\" parameter to symmetric as multigrid algorithm. " << multigridAlgo << " is primarily supposed to be used for symmetric problems.\n\n"
+                                  << "Please note: if you are using \"unsmoothed\" transfer operators the \"problem: symmetric\" parameter "
+                                  << "has no real mathematical meaning, i.e. you can use it for non-symmetric\n"
+                                  << "problems, too. With \"problem: symmetric\"=\"symmetric\" you can use implicit transpose for building "
+                                  << "the restriction operators which may drastically reduce the amount of consumed memory." << std::endl;
+      isSymmetric = true;
     }
+    TEUCHOS_TEST_FOR_EXCEPTION(multigridAlgo == "pg" && isSymmetric == true, Exceptions::RuntimeError,
+                               "Petrov-Galerkin smoothed transfer operators are only allowed for non-symmetric problems: Set \"problem: symmetric\" to false!\n"
+                               "While PG smoothed transfer operators generally would also work for symmetric problems this is an unusual use case. "
+                               "You can use the factory-based xml interface though if you need PG-AMG for symmetric problems.");
 
+    if (have_userR) {
+      manager.SetFactory("R", NoFactory::getRCP());
+    } else {
+      if (isSymmetric)
+        R = rcp(new TransPFactory());
+      else
+        R = rcp(new GenericRFactory());
 
-  }
-
-  // =====================================================================================================
-  // ========================================= Repartition ===============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Repartition(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                                   int levelID, std::vector<keep_pair>& keeps, RCP<Factory> & nullSpaceFactory) const
-  {
-    // === Repartitioning ===
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: enable", bool, enableRepart);
-    if (enableRepart) {
-#if defined(HAVE_MPI) && (defined(HAVE_MUELU_ZOLTAN) || defined(HAVE_MUELU_ZOLTAN2)) // skip to the end, print warning, and turn off repartitioning if we don't have MPI and Zoltan/Zoltan2
-      MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: use subcommunicators in place", bool, enableInPlace);
-      // Short summary of the issue: RebalanceTransferFactory shares ownership
-      // of "P" with SaPFactory, and therefore, changes the stored version.
-      // That means that if SaPFactory generated P, and stored it on the level,
-      // then after rebalancing the value in that storage changed. It goes
-      // against the concept of factories (I think), that every factory is
-      // responsible for its own objects, and they are immutable outside.
-      //
-      // In reuse, this is what happens: as we reuse Importer across setups,
-      // the order of factories changes, and coupled with shared ownership
-      // leads to problems.
-      // *First setup*
-      //    SaP               builds     P [and stores it]
-      //    TransP            builds     R [and stores it]
-      //    RAP               builds     A [and stores it]
-      //    RebalanceTransfer rebalances P [and changes the P stored by SaP]   (*)
-      //    RebalanceTransfer rebalances R
-      //    RebalanceAc       rebalances A
-      // *Second setup* ("RP" reuse)
-      //    RebalanceTransfer rebalances P [which is incorrect due to (*)]
-      //    RebalanceTransfer rebalances R
-      //    RAP               builds     A [which is incorrect due to (*)]
-      //    RebalanceAc       rebalances A [which throws due to map inconsistency]
-      //    ...
-      // *Second setup* ("tP" reuse)
-      //    SaP               builds     P [and stores it]
-      //    RebalanceTransfer rebalances P [and changes the P stored by SaP]   (**)
-      //    TransP            builds     R [which is incorrect due to (**)]
-      //    RebalanceTransfer rebalances R
-      //    ...
-      //
-      // Couple solutions to this:
-      //    1. [implemented] Requre "tP" and "PR" reuse to only be used with
-      //       implicit rebalancing.
-      //    2. Do deep copy of P, and changed domain map and importer there.
-      //       Need to investigate how expensive this is.
-      TEUCHOS_TEST_FOR_EXCEPTION(this->doPRrebalance_ && (reuseType == "tP" || reuseType == "RP"), Exceptions::InvalidArgument,
-                                 "Reuse types \"tP\" and \"PR\" require \"repartition: rebalance P and R\" set to \"false\"");
-
-      // TEUCHOS_TEST_FOR_EXCEPTION(aggType == "brick", Exceptions::InvalidArgument,
-      //                            "Aggregation type \"brick\" requires \"repartition: enable\" set to \"false\"");
-
-      MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: partitioner", std::string, partName);
-      TEUCHOS_TEST_FOR_EXCEPTION(partName != "zoltan" && partName != "zoltan2", Exceptions::InvalidArgument,
-                                 "Invalid partitioner name: \"" << partName << "\". Valid options: \"zoltan\", \"zoltan2\"");
-
-# ifndef HAVE_MUELU_ZOLTAN
-      bool switched = false;
-      if (partName == "zoltan") {
-        this->GetOStream(Warnings0) << "Zoltan interface is not available, trying to switch to Zoltan2" << std::endl;
-        partName = "zoltan2";
-        switched = true;
-      }
-# else
-#  ifndef HAVE_MUELU_ZOLTAN2
-      bool switched = false;
-#  endif // HAVE_MUELU_ZOLTAN2
-# endif // HAVE_MUELU_ZOLTAN
-
-# ifndef HAVE_MUELU_ZOLTAN2
-      if (partName == "zoltan2" && !switched) {
-        this->GetOStream(Warnings0) << "Zoltan2 interface is not available, trying to switch to Zoltan" << std::endl;
-        partName = "zoltan";
-      }
-# endif // HAVE_MUELU_ZOLTAN2
-
-      MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: node repartition level",int,nodeRepartitionLevel);
-
-      // RepartitionHeuristic
-      auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
-      ParameterList repartheurParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: node repartition level", int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: start level",            int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: min rows per proc",      int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: target rows per proc",   int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: min rows per thread",    int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: target rows per thread", int,    repartheurParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: max imbalance",          double, repartheurParams);
-      repartheurFactory->SetParameterList(repartheurParams);
-      repartheurFactory->SetFactory("A",         manager.GetFactory("A"));
-      manager.SetFactory("number of partitions", repartheurFactory);
-      manager.SetFactory("repartition: heuristic target rows per process", repartheurFactory);
-
-      // Partitioner
-      RCP<Factory> partitioner;
-      if (levelID == nodeRepartitionLevel) {
-        //        partitioner = rcp(new NodePartitionInterface());
-        partitioner = rcp(new MueLu::NodePartitionInterface<SC,LO,GO,NO>());
-        ParameterList partParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: node id"               ,int,repartheurParams);
-        partitioner->SetParameterList(partParams);
-        partitioner->SetFactory("Node Comm",         manager.GetFactory("Node Comm"));
-      }
-      else if (partName == "zoltan") {
-# ifdef HAVE_MUELU_ZOLTAN
-        partitioner = rcp(new ZoltanInterface());
-        // NOTE: ZoltanInterface ("zoltan") does not support external parameters through ParameterList
-# else
-        throw Exceptions::RuntimeError("Zoltan interface is not available");
-# endif // HAVE_MUELU_ZOLTAN
-      } else if (partName == "zoltan2") {
-# ifdef HAVE_MUELU_ZOLTAN2
-        partitioner = rcp(new Zoltan2Interface());
-        ParameterList partParams;
-        RCP<const ParameterList> partpartParams = rcp(new ParameterList(paramList.sublist("repartition: params", false)));
-        partParams.set("ParameterList", partpartParams);
-        partitioner->SetParameterList(partParams);
-        partitioner->SetFactory("repartition: heuristic target rows per process",
-                                manager.GetFactory("repartition: heuristic target rows per process"));
-# else
-        throw Exceptions::RuntimeError("Zoltan2 interface is not available");
-# endif // HAVE_MUELU_ZOLTAN2
-      }
+      R->SetFactory("P", manager.GetFactory("P"));
+      manager.SetFactory("R", R);
+    }
 
-      partitioner->SetFactory("A",                    manager.GetFactory("A"));
-      partitioner->SetFactory("number of partitions", manager.GetFactory("number of partitions"));
-      if (useCoordinates_)
-        partitioner->SetFactory("Coordinates",          manager.GetFactory("Coordinates"));
-      manager.SetFactory("Partition", partitioner);
-
-      // Repartitioner
-      auto repartFactory = rcp(new RepartitionFactory());
-      ParameterList repartParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: print partition distribution", bool, repartParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: remap parts",                  bool, repartParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: remap num values",              int, repartParams);
-      repartFactory->SetParameterList(repartParams);
-      repartFactory->SetFactory("A",                    manager.GetFactory("A"));
-      repartFactory->SetFactory("number of partitions", manager.GetFactory("number of partitions"));
-      repartFactory->SetFactory("Partition",            manager.GetFactory("Partition"));
-      manager.SetFactory("Importer", repartFactory);
-      if (reuseType != "none" && reuseType != "S" && levelID)
-        keeps.push_back(keep_pair("Importer", manager.GetFactory("Importer").get()));
-
-
-      if(enableInPlace) {
-        // Rebalanced A (in place)
-        // NOTE: This is for when we want to constrain repartitioning to match some other idea of what's going on.
-        // The major application is the (1,1) hierarchy in the Maxwell1 preconditioner.
-        auto newA = rcp(new RebalanceAcFactory());
-        ParameterList rebAcParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rebAcParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators in place", bool, rebAcParams);
-        newA->SetParameterList(rebAcParams);
-        newA->SetFactory("A",          manager.GetFactory("A"));
-        newA->SetFactory("InPlaceMap", manager.GetFactory("InPlaceMap"));
-        manager.SetFactory("A",newA);
-      }
-      else {
-        // Rebalanced A
-        auto newA = rcp(new RebalanceAcFactory());
-        ParameterList rebAcParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rebAcParams);
-        newA->SetParameterList(rebAcParams);
-        newA->SetFactory("A",         manager.GetFactory("A"));
-        newA->SetFactory("Importer",  manager.GetFactory("Importer"));
-        manager.SetFactory("A", newA);
-        
-        // Rebalanced P
-        auto newP = rcp(new RebalanceTransferFactory());
-        ParameterList newPparams;
-        newPparams.set("type", "Interpolation");
-        if (changedPRrebalance_)
-          newPparams.set("repartition: rebalance P and R", this->doPRrebalance_);
-        if (changedPRViaCopyrebalance_)
-          newPparams.set("repartition: explicit via new copy rebalance P and R",true);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, newPparams);
-        newP->  SetParameterList(newPparams);
-        newP->  SetFactory("Importer",    manager.GetFactory("Importer"));
-        newP->  SetFactory("P",           manager.GetFactory("P"));
-        if (!paramList.isParameter("semicoarsen: number of levels"))
-          newP->SetFactory("Nullspace",   manager.GetFactory("Ptent"));
-        else
-          newP->SetFactory("Nullspace",   manager.GetFactory("P")); // TogglePFactory
-        if (useCoordinates_)
-          newP->  SetFactory("Coordinates", manager.GetFactory("Coordinates"));
-        manager.SetFactory("P",           newP);
-        if (useCoordinates_)
-          manager.SetFactory("Coordinates", newP);
-        if (useBlockNumber_ && (levelID > 0)) {
-          newP->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
-          manager.SetFactory("BlockNumber", newP);
-        }
-        
-        // Rebalanced R
-        auto newR = rcp(new RebalanceTransferFactory());
-        ParameterList newRparams;
-        newRparams.set("type", "Restriction");
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, newRparams);
-        if (changedPRrebalance_)
-          newRparams.set("repartition: rebalance P and R", this->doPRrebalance_);
-        if (changedPRViaCopyrebalance_)
-          newPparams.set("repartition: explicit via new copy rebalance P and R",true);
-        if (changedImplicitTranspose_)
-          newRparams.set("transpose: use implicit",        this->implicitTranspose_);
-        newR->  SetParameterList(newRparams);
-        newR->  SetFactory("Importer", manager.GetFactory("Importer"));
-        if (!this->implicitTranspose_) {
-          newR->SetFactory("R",        manager.GetFactory("R"));
-          manager.SetFactory("R",      newR);
-        }
+  } else {
+    manager.SetFactory("R", Teuchos::null);
+  }
 
-        // NOTE: the role of NullspaceFactory is to provide nullspace on the finest
-        // level if a user does not do that. For all other levels it simply passes
-        // nullspace from a real factory to whoever needs it. If we don't use
-        // repartitioning, that factory is "TentativePFactory"; if we do, it is
-        // "RebalanceTransferFactory". But we still have to have NullspaceFactory as
-        // the "Nullspace" of the manager
-        // NOTE: This really needs to be set on the *NullSpaceFactory*, not manager.get("Nullspace").
-        ParameterList newNullparams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "nullspace: calculate rotations", bool, newNullparams);
-        nullSpaceFactory->SetFactory("Nullspace", newP);
-        nullSpaceFactory->SetParameterList(newNullparams);
-      }
+  // === Restriction: Nullspace Scaling ===
+  if (paramList.isParameter("restriction: scale nullspace") && paramList.get<bool>("restriction: scale nullspace")) {
+    RCP<TentativePFactory> tentPFactory = rcp(new TentativePFactory());
+    Teuchos::ParameterList tentPlist;
+    tentPlist.set("Nullspace name", "Scaled Nullspace");
+    tentPFactory->SetParameterList(tentPlist);
+    tentPFactory->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+    tentPFactory->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
+
+    if (R.is_null()) R = rcp(new TransPFactory());
+    R->SetFactory("P", tentPFactory);
+  }
+}
+
+// =====================================================================================================
+// ========================================= Repartition ===============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Repartition(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                     int levelID, std::vector<keep_pair>& keeps, RCP<Factory>& nullSpaceFactory) const {
+  // === Repartitioning ===
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: enable", bool, enableRepart);
+  if (enableRepart) {
+#if defined(HAVE_MPI) && (defined(HAVE_MUELU_ZOLTAN) || defined(HAVE_MUELU_ZOLTAN2))  // skip to the end, print warning, and turn off repartitioning if we don't have MPI and Zoltan/Zoltan2
+    MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: use subcommunicators in place", bool, enableInPlace);
+    // Short summary of the issue: RebalanceTransferFactory shares ownership
+    // of "P" with SaPFactory, and therefore, changes the stored version.
+    // That means that if SaPFactory generated P, and stored it on the level,
+    // then after rebalancing the value in that storage changed. It goes
+    // against the concept of factories (I think), that every factory is
+    // responsible for its own objects, and they are immutable outside.
+    //
+    // In reuse, this is what happens: as we reuse Importer across setups,
+    // the order of factories changes, and coupled with shared ownership
+    // leads to problems.
+    // *First setup*
+    //    SaP               builds     P [and stores it]
+    //    TransP            builds     R [and stores it]
+    //    RAP               builds     A [and stores it]
+    //    RebalanceTransfer rebalances P [and changes the P stored by SaP]   (*)
+    //    RebalanceTransfer rebalances R
+    //    RebalanceAc       rebalances A
+    // *Second setup* ("RP" reuse)
+    //    RebalanceTransfer rebalances P [which is incorrect due to (*)]
+    //    RebalanceTransfer rebalances R
+    //    RAP               builds     A [which is incorrect due to (*)]
+    //    RebalanceAc       rebalances A [which throws due to map inconsistency]
+    //    ...
+    // *Second setup* ("tP" reuse)
+    //    SaP               builds     P [and stores it]
+    //    RebalanceTransfer rebalances P [and changes the P stored by SaP]   (**)
+    //    TransP            builds     R [which is incorrect due to (**)]
+    //    RebalanceTransfer rebalances R
+    //    ...
+    //
+    // Couple solutions to this:
+    //    1. [implemented] Requre "tP" and "PR" reuse to only be used with
+    //       implicit rebalancing.
+    //    2. Do deep copy of P, and changed domain map and importer there.
+    //       Need to investigate how expensive this is.
+    TEUCHOS_TEST_FOR_EXCEPTION(this->doPRrebalance_ && (reuseType == "tP" || reuseType == "RP"), Exceptions::InvalidArgument,
+                               "Reuse types \"tP\" and \"PR\" require \"repartition: rebalance P and R\" set to \"false\"");
+
+    // TEUCHOS_TEST_FOR_EXCEPTION(aggType == "brick", Exceptions::InvalidArgument,
+    //                            "Aggregation type \"brick\" requires \"repartition: enable\" set to \"false\"");
+
+    MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: partitioner", std::string, partName);
+    TEUCHOS_TEST_FOR_EXCEPTION(partName != "zoltan" && partName != "zoltan2", Exceptions::InvalidArgument,
+                               "Invalid partitioner name: \"" << partName << "\". Valid options: \"zoltan\", \"zoltan2\"");
+
+#ifndef HAVE_MUELU_ZOLTAN
+    bool switched = false;
+    if (partName == "zoltan") {
+      this->GetOStream(Warnings0) << "Zoltan interface is not available, trying to switch to Zoltan2" << std::endl;
+      partName = "zoltan2";
+      switched = true;
+    }
 #else
-      paramList.set("repartition: enable",false);
-# ifndef HAVE_MPI
-      this->GetOStream(Warnings0) << "No repartitioning available for a serial run\n";
-# else
-      this->GetOStream(Warnings0) << "Zoltan/Zoltan2 are unavailable for repartitioning\n";
-# endif // HAVE_MPI
-#endif // defined(HAVE_MPI) && (defined(HAVE_MUELU_ZOLTAN) || defined(HAVE_MUELU_ZOLTAN2))
+#ifndef HAVE_MUELU_ZOLTAN2
+    bool switched = false;
+#endif  // HAVE_MUELU_ZOLTAN2
+#endif  // HAVE_MUELU_ZOLTAN
+
+#ifndef HAVE_MUELU_ZOLTAN2
+    if (partName == "zoltan2" && !switched) {
+      this->GetOStream(Warnings0) << "Zoltan2 interface is not available, trying to switch to Zoltan" << std::endl;
+      partName = "zoltan";
+    }
+#endif  // HAVE_MUELU_ZOLTAN2
+
+    MUELU_SET_VAR_2LIST(paramList, defaultList, "repartition: node repartition level", int, nodeRepartitionLevel);
+
+    // RepartitionHeuristic
+    auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
+    ParameterList repartheurParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: node repartition level", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: start level", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: min rows per proc", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: target rows per proc", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: min rows per thread", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: target rows per thread", int, repartheurParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: max imbalance", double, repartheurParams);
+    repartheurFactory->SetParameterList(repartheurParams);
+    repartheurFactory->SetFactory("A", manager.GetFactory("A"));
+    manager.SetFactory("number of partitions", repartheurFactory);
+    manager.SetFactory("repartition: heuristic target rows per process", repartheurFactory);
+
+    // Partitioner
+    RCP<Factory> partitioner;
+    if (levelID == nodeRepartitionLevel) {
+      //        partitioner = rcp(new NodePartitionInterface());
+      partitioner = rcp(new MueLu::NodePartitionInterface<SC, LO, GO, NO>());
+      ParameterList partParams;
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: node id", int, repartheurParams);
+      partitioner->SetParameterList(partParams);
+      partitioner->SetFactory("Node Comm", manager.GetFactory("Node Comm"));
+    } else if (partName == "zoltan") {
+#ifdef HAVE_MUELU_ZOLTAN
+      partitioner = rcp(new ZoltanInterface());
+      // NOTE: ZoltanInterface ("zoltan") does not support external parameters through ParameterList
+#else
+      throw Exceptions::RuntimeError("Zoltan interface is not available");
+#endif  // HAVE_MUELU_ZOLTAN
+    } else if (partName == "zoltan2") {
+#ifdef HAVE_MUELU_ZOLTAN2
+      partitioner = rcp(new Zoltan2Interface());
+      ParameterList partParams;
+      RCP<const ParameterList> partpartParams = rcp(new ParameterList(paramList.sublist("repartition: params", false)));
+      partParams.set("ParameterList", partpartParams);
+      partitioner->SetParameterList(partParams);
+      partitioner->SetFactory("repartition: heuristic target rows per process",
+                              manager.GetFactory("repartition: heuristic target rows per process"));
+#else
+      throw Exceptions::RuntimeError("Zoltan2 interface is not available");
+#endif  // HAVE_MUELU_ZOLTAN2
     }
-  }
-
-  // =====================================================================================================
-  // ========================================= Low precision transfers ===================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_LowPrecision(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                                    int levelID, std::vector<keep_pair>& keeps) const
-  {
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "transfers: half precision", bool, enableLowPrecision);
 
-    if (enableLowPrecision) {
-      // Low precision P
-      auto newP = rcp(new LowPrecisionFactory());
+    partitioner->SetFactory("A", manager.GetFactory("A"));
+    partitioner->SetFactory("number of partitions", manager.GetFactory("number of partitions"));
+    if (useCoordinates_)
+      partitioner->SetFactory("Coordinates", manager.GetFactory("Coordinates"));
+    manager.SetFactory("Partition", partitioner);
+
+    // Repartitioner
+    auto repartFactory = rcp(new RepartitionFactory());
+    ParameterList repartParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: print partition distribution", bool, repartParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: remap parts", bool, repartParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: remap num values", int, repartParams);
+    repartFactory->SetParameterList(repartParams);
+    repartFactory->SetFactory("A", manager.GetFactory("A"));
+    repartFactory->SetFactory("number of partitions", manager.GetFactory("number of partitions"));
+    repartFactory->SetFactory("Partition", manager.GetFactory("Partition"));
+    manager.SetFactory("Importer", repartFactory);
+    if (reuseType != "none" && reuseType != "S" && levelID)
+      keeps.push_back(keep_pair("Importer", manager.GetFactory("Importer").get()));
+
+    if (enableInPlace) {
+      // Rebalanced A (in place)
+      // NOTE: This is for when we want to constrain repartitioning to match some other idea of what's going on.
+      // The major application is the (1,1) hierarchy in the Maxwell1 preconditioner.
+      auto newA = rcp(new RebalanceAcFactory());
+      ParameterList rebAcParams;
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rebAcParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators in place", bool, rebAcParams);
+      newA->SetParameterList(rebAcParams);
+      newA->SetFactory("A", manager.GetFactory("A"));
+      newA->SetFactory("InPlaceMap", manager.GetFactory("InPlaceMap"));
+      manager.SetFactory("A", newA);
+    } else {
+      // Rebalanced A
+      auto newA = rcp(new RebalanceAcFactory());
+      ParameterList rebAcParams;
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, rebAcParams);
+      newA->SetParameterList(rebAcParams);
+      newA->SetFactory("A", manager.GetFactory("A"));
+      newA->SetFactory("Importer", manager.GetFactory("Importer"));
+      manager.SetFactory("A", newA);
+
+      // Rebalanced P
+      auto newP = rcp(new RebalanceTransferFactory());
       ParameterList newPparams;
-      newPparams.set("matrix key", "P");
-      newP->  SetParameterList(newPparams);
-      newP->  SetFactory("P", manager.GetFactory("P"));
+      newPparams.set("type", "Interpolation");
+      if (changedPRrebalance_)
+        newPparams.set("repartition: rebalance P and R", this->doPRrebalance_);
+      if (changedPRViaCopyrebalance_)
+        newPparams.set("repartition: explicit via new copy rebalance P and R", true);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, newPparams);
+      newP->SetParameterList(newPparams);
+      newP->SetFactory("Importer", manager.GetFactory("Importer"));
+      newP->SetFactory("P", manager.GetFactory("P"));
+      if (!paramList.isParameter("semicoarsen: number of levels"))
+        newP->SetFactory("Nullspace", manager.GetFactory("Ptent"));
+      else
+        newP->SetFactory("Nullspace", manager.GetFactory("P"));  // TogglePFactory
+      if (useCoordinates_)
+        newP->SetFactory("Coordinates", manager.GetFactory("Coordinates"));
       manager.SetFactory("P", newP);
+      if (useCoordinates_)
+        manager.SetFactory("Coordinates", newP);
+      if (useBlockNumber_ && (levelID > 0)) {
+        newP->SetFactory("BlockNumber", manager.GetFactory("BlockNumber"));
+        manager.SetFactory("BlockNumber", newP);
+      }
 
+      // Rebalanced R
+      auto newR = rcp(new RebalanceTransferFactory());
+      ParameterList newRparams;
+      newRparams.set("type", "Restriction");
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "repartition: use subcommunicators", bool, newRparams);
+      if (changedPRrebalance_)
+        newRparams.set("repartition: rebalance P and R", this->doPRrebalance_);
+      if (changedPRViaCopyrebalance_)
+        newPparams.set("repartition: explicit via new copy rebalance P and R", true);
+      if (changedImplicitTranspose_)
+        newRparams.set("transpose: use implicit", this->implicitTranspose_);
+      newR->SetParameterList(newRparams);
+      newR->SetFactory("Importer", manager.GetFactory("Importer"));
       if (!this->implicitTranspose_) {
-        // Low precision R
-        auto newR = rcp(new LowPrecisionFactory());
-        ParameterList newRparams;
-        newRparams.set("matrix key", "R");
-        newR->  SetParameterList(newRparams);
-        newR->  SetFactory("R", manager.GetFactory("R"));
+        newR->SetFactory("R", manager.GetFactory("R"));
         manager.SetFactory("R", newR);
       }
-    }
-  }
-
-  // =====================================================================================================
-  // =========================================== Nullspace ===============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Nullspace(ParameterList& paramList, const ParameterList&  defaultList, FactoryManager& manager,
-                                 int /* levelID */, std::vector<keep_pair>& /* keeps */, RCP<Factory> & nullSpaceFactory) const
-  {
-    // Nullspace
-    MUELU_KOKKOS_FACTORY(nullSpace, NullspaceFactory, NullspaceFactory_kokkos);
-
-    bool have_userNS = false;
-    if (paramList.isParameter("Nullspace") && !paramList.get<RCP<MultiVector> >("Nullspace").is_null())
-      have_userNS = true;
 
-    if (!have_userNS) {
+      // NOTE: the role of NullspaceFactory is to provide nullspace on the finest
+      // level if a user does not do that. For all other levels it simply passes
+      // nullspace from a real factory to whoever needs it. If we don't use
+      // repartitioning, that factory is "TentativePFactory"; if we do, it is
+      // "RebalanceTransferFactory". But we still have to have NullspaceFactory as
+      // the "Nullspace" of the manager
+      // NOTE: This really needs to be set on the *NullSpaceFactory*, not manager.get("Nullspace").
       ParameterList newNullparams;
       MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "nullspace: calculate rotations", bool, newNullparams);
-      nullSpace->SetParameterList(newNullparams);
-      nullSpace->SetFactory("Nullspace", manager.GetFactory("Ptent"));
-      manager.SetFactory("Nullspace", nullSpace);
+      nullSpaceFactory->SetFactory("Nullspace", newP);
+      nullSpaceFactory->SetParameterList(newNullparams);
     }
-    nullSpaceFactory = nullSpace;
-
-    if (paramList.isParameter("restriction: scale nullspace") && paramList.get<bool>("restriction: scale nullspace")) {
-      RCP<ScaledNullspaceFactory> scaledNSfactory = rcp(new ScaledNullspaceFactory());
-      scaledNSfactory->SetFactory("Nullspace",nullSpaceFactory);
-      manager.SetFactory("Scaled Nullspace",scaledNSfactory);
-    }
-
+#else
+    paramList.set("repartition: enable", false);
+#ifndef HAVE_MPI
+    this->GetOStream(Warnings0) << "No repartitioning available for a serial run\n";
+#else
+    this->GetOStream(Warnings0) << "Zoltan/Zoltan2 are unavailable for repartitioning\n";
+#endif  // HAVE_MPI
+#endif  // defined(HAVE_MPI) && (defined(HAVE_MUELU_ZOLTAN) || defined(HAVE_MUELU_ZOLTAN2))
   }
+}
+
+// =====================================================================================================
+// ========================================= Low precision transfers ===================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_LowPrecision(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                      int levelID, std::vector<keep_pair>& keeps) const {
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "transfers: half precision", bool, enableLowPrecision);
+
+  if (enableLowPrecision) {
+    // Low precision P
+    auto newP = rcp(new LowPrecisionFactory());
+    ParameterList newPparams;
+    newPparams.set("matrix key", "P");
+    newP->SetParameterList(newPparams);
+    newP->SetFactory("P", manager.GetFactory("P"));
+    manager.SetFactory("P", newP);
 
-  // =====================================================================================================
-  // ================================= Algorithm: SemiCoarsening =========================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_SemiCoarsen(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                                   int /* levelID */, std::vector<keep_pair>& /* keeps */) const
-  {
-    // === Semi-coarsening ===
-    RCP<Factory>  semicoarsenFactory = Teuchos::null;
-    if (paramList.isParameter("semicoarsen: number of levels") &&
-        paramList.get<int>("semicoarsen: number of levels") > 0) {
-
-      ParameterList togglePParams;
-      ParameterList semicoarsenPParams;
-      ParameterList linedetectionParams;
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: number of levels", int,         togglePParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: coarsen rate",     int,         semicoarsenPParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: piecewise constant", bool,      semicoarsenPParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: piecewise linear", bool,      semicoarsenPParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: calculate nonsym restriction", bool, semicoarsenPParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "linedetection: orientation",    std::string, linedetectionParams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "linedetection: num layers",     int,         linedetectionParams);
-
-      MUELU_KOKKOS_FACTORY_NO_DECL(semicoarsenFactory, SemiCoarsenPFactory, SemiCoarsenPFactory_kokkos);
-      RCP<LineDetectionFactory> linedetectionFactory = rcp(new LineDetectionFactory());
-      RCP<TogglePFactory>       togglePFactory       = rcp(new TogglePFactory());
-
-      linedetectionFactory->SetParameterList(linedetectionParams);
-      semicoarsenFactory  ->SetParameterList(semicoarsenPParams);
-      togglePFactory      ->SetParameterList(togglePParams);
-
-      togglePFactory->AddCoarseNullspaceFactory (semicoarsenFactory);
-      togglePFactory->AddProlongatorFactory     (semicoarsenFactory);
-      togglePFactory->AddPtentFactory           (semicoarsenFactory);
-      togglePFactory->AddCoarseNullspaceFactory (manager.GetFactory("Ptent"));
-      togglePFactory->AddProlongatorFactory     (manager.GetFactory("P"));
-      togglePFactory->AddPtentFactory           (manager.GetFactory("Ptent"));
-
-      manager.SetFactory("CoarseNumZLayers",          linedetectionFactory);
-      manager.SetFactory("LineDetection_Layers",      linedetectionFactory);
-      manager.SetFactory("LineDetection_VertLineIds", linedetectionFactory);
-
-      manager.SetFactory("P",         togglePFactory);
-      manager.SetFactory("Ptent",     togglePFactory);
-      manager.SetFactory("Nullspace", togglePFactory);
+    if (!this->implicitTranspose_) {
+      // Low precision R
+      auto newR = rcp(new LowPrecisionFactory());
+      ParameterList newRparams;
+      newRparams.set("matrix key", "R");
+      newR->SetParameterList(newRparams);
+      newR->SetFactory("R", manager.GetFactory("R"));
+      manager.SetFactory("R", newR);
     }
+  }
+}
+
+// =====================================================================================================
+// =========================================== Nullspace ===============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Nullspace(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                   int /* levelID */, std::vector<keep_pair>& /* keeps */, RCP<Factory>& nullSpaceFactory) const {
+  // Nullspace
+  MUELU_KOKKOS_FACTORY(nullSpace, NullspaceFactory, NullspaceFactory_kokkos);
+
+  bool have_userNS = false;
+  if (paramList.isParameter("Nullspace") && !paramList.get<RCP<MultiVector> >("Nullspace").is_null())
+    have_userNS = true;
+
+  if (!have_userNS) {
+    ParameterList newNullparams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "nullspace: calculate rotations", bool, newNullparams);
+    nullSpace->SetParameterList(newNullparams);
+    nullSpace->SetFactory("Nullspace", manager.GetFactory("Ptent"));
+    manager.SetFactory("Nullspace", nullSpace);
+  }
+  nullSpaceFactory = nullSpace;
 
-    if (paramList.isParameter("semicoarsen: number of levels")) {
-      auto tf = rcp(new ToggleCoordinatesTransferFactory());
-      tf->SetFactory("Chosen P", manager.GetFactory("P"));
-      tf->AddCoordTransferFactory(semicoarsenFactory);
-
-      RCP<Factory> coords = rcp(new CoordinatesTransferFactory());
-      coords->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-      coords->SetFactory("CoarseMap",  manager.GetFactory("CoarseMap"));
-      tf->AddCoordTransferFactory(coords);
-      manager.SetFactory("Coordinates", tf);
-    }
+  if (paramList.isParameter("restriction: scale nullspace") && paramList.get<bool>("restriction: scale nullspace")) {
+    RCP<ScaledNullspaceFactory> scaledNSfactory = rcp(new ScaledNullspaceFactory());
+    scaledNSfactory->SetFactory("Nullspace", nullSpaceFactory);
+    manager.SetFactory("Scaled Nullspace", scaledNSfactory);
+  }
+}
+
+// =====================================================================================================
+// ================================= Algorithm: SemiCoarsening =========================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_SemiCoarsen(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                     int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  // === Semi-coarsening ===
+  RCP<Factory> semicoarsenFactory = Teuchos::null;
+  if (paramList.isParameter("semicoarsen: number of levels") &&
+      paramList.get<int>("semicoarsen: number of levels") > 0) {
+    ParameterList togglePParams;
+    ParameterList semicoarsenPParams;
+    ParameterList linedetectionParams;
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: number of levels", int, togglePParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: coarsen rate", int, semicoarsenPParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: piecewise constant", bool, semicoarsenPParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: piecewise linear", bool, semicoarsenPParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "semicoarsen: calculate nonsym restriction", bool, semicoarsenPParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "linedetection: orientation", std::string, linedetectionParams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "linedetection: num layers", int, linedetectionParams);
+
+    MUELU_KOKKOS_FACTORY_NO_DECL(semicoarsenFactory, SemiCoarsenPFactory, SemiCoarsenPFactory_kokkos);
+    RCP<LineDetectionFactory> linedetectionFactory = rcp(new LineDetectionFactory());
+    RCP<TogglePFactory> togglePFactory             = rcp(new TogglePFactory());
+
+    linedetectionFactory->SetParameterList(linedetectionParams);
+    semicoarsenFactory->SetParameterList(semicoarsenPParams);
+    togglePFactory->SetParameterList(togglePParams);
+
+    togglePFactory->AddCoarseNullspaceFactory(semicoarsenFactory);
+    togglePFactory->AddProlongatorFactory(semicoarsenFactory);
+    togglePFactory->AddPtentFactory(semicoarsenFactory);
+    togglePFactory->AddCoarseNullspaceFactory(manager.GetFactory("Ptent"));
+    togglePFactory->AddProlongatorFactory(manager.GetFactory("P"));
+    togglePFactory->AddPtentFactory(manager.GetFactory("Ptent"));
+
+    manager.SetFactory("CoarseNumZLayers", linedetectionFactory);
+    manager.SetFactory("LineDetection_Layers", linedetectionFactory);
+    manager.SetFactory("LineDetection_VertLineIds", linedetectionFactory);
+
+    manager.SetFactory("P", togglePFactory);
+    manager.SetFactory("Ptent", togglePFactory);
+    manager.SetFactory("Nullspace", togglePFactory);
   }
 
+  if (paramList.isParameter("semicoarsen: number of levels")) {
+    auto tf = rcp(new ToggleCoordinatesTransferFactory());
+    tf->SetFactory("Chosen P", manager.GetFactory("P"));
+    tf->AddCoordTransferFactory(semicoarsenFactory);
 
-  // =====================================================================================================
-  // ================================== Algorithm: P-Coarsening ==========================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_PCoarsen(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                                int levelID, std::vector<keep_pair>& keeps) const
-  {
+    RCP<Factory> coords = rcp(new CoordinatesTransferFactory());
+    coords->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+    coords->SetFactory("CoarseMap", manager.GetFactory("CoarseMap"));
+    tf->AddCoordTransferFactory(coords);
+    manager.SetFactory("Coordinates", tf);
+  }
+}
+
+// =====================================================================================================
+// ================================== Algorithm: P-Coarsening ==========================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_PCoarsen(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                                  int levelID, std::vector<keep_pair>& keeps) const {
 #ifdef HAVE_MUELU_INTREPID2
-    // This only makes sense to invoke from the default list.
-    if (defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
-      // P-Coarsening by schedule (new interface)
-      // NOTE: levelID represents the *coarse* level in this case
-      auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList,"pcoarsen: schedule");
-      auto pcoarsen_element  = defaultList.get<std::string>("pcoarsen: element");
-
-      if (levelID >= (int)pcoarsen_schedule.size()) {
-        // Past the p-coarsening levels, we do Smoothed Aggregation
-        // NOTE: We should probably consider allowing other options past p-coarsening
-        UpdateFactoryManager_SA(paramList, defaultList, manager, levelID, keeps);
-
-      } else {
-        // P-Coarsening
-        ParameterList Pparams;
-        auto P = rcp(new IntrepidPCoarsenFactory());
-        std::string lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
-        std::string hi = (levelID ? pcoarsen_element + std::to_string(pcoarsen_schedule[levelID-1]) : lo);
-        Pparams.set("pcoarsen: hi basis", hi);
-        Pparams.set("pcoarsen: lo basis", lo);
-        P->SetParameterList(Pparams);
-        manager.SetFactory("P", P);
-
-        // Add special nullspace handling
-        rcp_dynamic_cast<Factory>(manager.GetFactoryNonConst("Nullspace"))->SetFactory("Nullspace", manager.GetFactory("P"));
-      }
+  // This only makes sense to invoke from the default list.
+  if (defaultList.isParameter("pcoarsen: schedule") && defaultList.isParameter("pcoarsen: element")) {
+    // P-Coarsening by schedule (new interface)
+    // NOTE: levelID represents the *coarse* level in this case
+    auto pcoarsen_schedule = Teuchos::getArrayFromStringParameter<int>(defaultList, "pcoarsen: schedule");
+    auto pcoarsen_element  = defaultList.get<std::string>("pcoarsen: element");
+
+    if (levelID >= (int)pcoarsen_schedule.size()) {
+      // Past the p-coarsening levels, we do Smoothed Aggregation
+      // NOTE: We should probably consider allowing other options past p-coarsening
+      UpdateFactoryManager_SA(paramList, defaultList, manager, levelID, keeps);
 
     } else {
-      // P-Coarsening by manual specification (old interface)
+      // P-Coarsening
       ParameterList Pparams;
-      auto P = rcp(new IntrepidPCoarsenFactory());
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "pcoarsen: hi basis", std::string, Pparams);
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "pcoarsen: lo basis", std::string, Pparams);
+      auto P         = rcp(new IntrepidPCoarsenFactory());
+      std::string lo = pcoarsen_element + std::to_string(pcoarsen_schedule[levelID]);
+      std::string hi = (levelID ? pcoarsen_element + std::to_string(pcoarsen_schedule[levelID - 1]) : lo);
+      Pparams.set("pcoarsen: hi basis", hi);
+      Pparams.set("pcoarsen: lo basis", lo);
       P->SetParameterList(Pparams);
       manager.SetFactory("P", P);
 
@@ -1968,225 +1919,230 @@ namespace MueLu {
       rcp_dynamic_cast<Factory>(manager.GetFactoryNonConst("Nullspace"))->SetFactory("Nullspace", manager.GetFactory("P"));
     }
 
-#endif
-  }
-
-  // =====================================================================================================
-  // ============================== Algorithm: Smoothed Aggregation ======================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_SA(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const {
-    // Smoothed aggregation
-    MUELU_KOKKOS_FACTORY(P, SaPFactory, SaPFactory_kokkos);
+  } else {
+    // P-Coarsening by manual specification (old interface)
     ParameterList Pparams;
-    if (paramList.isSublist("matrixmatrix: kernel params"))
-      Pparams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
-    if (defaultList.isSublist("matrixmatrix: kernel params"))
-      Pparams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: damping factor", double, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: calculate eigenvalue estimate", bool, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: max eigenvalue", double, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: eigenvalue estimate num iterations", int, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: use rowsumabs diagonal scaling", bool, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs diagonal replacement tolerance", double, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs diagonal replacement value", double, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs use automatic diagonal tolerance", bool, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: enforce constraints", bool, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: eigen-analysis type", std::string, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, Pparams);
-
+    auto P = rcp(new IntrepidPCoarsenFactory());
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "pcoarsen: hi basis", std::string, Pparams);
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "pcoarsen: lo basis", std::string, Pparams);
     P->SetParameterList(Pparams);
-
-
-    // Filtering
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "sa: use filtered matrix", bool, useFiltering);
-    if (useFiltering) {
-      // NOTE: Here, non-Kokkos and Kokkos versions diverge in the way the
-      // dependency tree is setup. The Kokkos version has merged the the
-      // FilteredAFactory into the CoalesceDropFactory.
-      if (!useKokkos_) {
-        RCP<Factory> filterFactory = rcp(new FilteredAFactory());
-
-        ParameterList fParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping",      bool, fParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph",      bool, fParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use root stencil", bool, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: Dirichlet threshold", double, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use spread lumping", bool, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom growth factor", double, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom cap", double, fParams);
-        filterFactory->SetParameterList(fParams);
-        filterFactory->SetFactory("Graph",      manager.GetFactory("Graph"));
-        filterFactory->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-	filterFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
-        // I'm not sure why we need this line. See comments for DofsPerNode for UncoupledAggregation above
-        filterFactory->SetFactory("Filtering",  manager.GetFactory("Graph"));
-
-        P->SetFactory("A", filterFactory);
-
-      } else {
-        P->SetFactory("A", manager.GetFactory("Graph"));
-      }
-    }
-
-    P->SetFactory("P", manager.GetFactory("Ptent"));
     manager.SetFactory("P", P);
 
-    bool filteringChangesMatrix = useFiltering && !MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol", double, 0);
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-    if (reuseType == "tP" && !filteringChangesMatrix)
-      keeps.push_back(keep_pair("AP reuse data", P.get()));
+    // Add special nullspace handling
+    rcp_dynamic_cast<Factory>(manager.GetFactoryNonConst("Nullspace"))->SetFactory("Nullspace", manager.GetFactory("P"));
   }
 
-  // =====================================================================================================
-  // =============================== Algorithm: Energy Minimization ======================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Emin(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
-                            int /* levelID */, std::vector<keep_pair>& /* keeps */) const
-  {
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "emin: pattern", std::string, patternType);
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
-    TEUCHOS_TEST_FOR_EXCEPTION(patternType != "AkPtent", Exceptions::InvalidArgument,
-                               "Invalid pattern name: \"" << patternType << "\". Valid options: \"AkPtent\"");
-    // Pattern
-    auto patternFactory = rcp(new PatternFactory());
-    ParameterList patternParams;
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: pattern order", int, patternParams);
-    patternFactory->SetParameterList(patternParams);
-    patternFactory->SetFactory("P", manager.GetFactory("Ptent"));
-    manager.SetFactory("Ppattern", patternFactory);
-
-    // Constraint
-    auto constraintFactory = rcp(new ConstraintFactory());
-    constraintFactory->SetFactory("Ppattern",        manager.GetFactory("Ppattern"));
-    constraintFactory->SetFactory("CoarseNullspace", manager.GetFactory("Ptent"));
-    manager.SetFactory("Constraint", constraintFactory);
-
-    // Emin Factory
-    auto P = rcp(new EminPFactory());
-    // Filtering
-    MUELU_SET_VAR_2LIST(paramList, defaultList, "emin: use filtered matrix", bool, useFiltering);
-    if(useFiltering) {
-      // NOTE: Here, non-Kokkos and Kokkos versions diverge in the way the
-      // dependency tree is setup. The Kokkos version has merged the the
-      // FilteredAFactory into the CoalesceDropFactory.
-      if (!useKokkos_) {
-        RCP<Factory> filterFactory = rcp(new FilteredAFactory());
-
-        ParameterList fParams;
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping",      bool, fParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph",      bool, fParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, fParams);
-        MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use root stencil", bool, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: Dirichlet threshold", double, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use spread lumping", bool, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom growth factor", double, fParams);
-	MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom cap", double, fParams);
-        filterFactory->SetParameterList(fParams);
-        filterFactory->SetFactory("Graph",      manager.GetFactory("Graph"));
-        filterFactory->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
-        filterFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
-        // I'm not sure why we need this line. See comments for DofsPerNode for UncoupledAggregation above
-        filterFactory->SetFactory("Filtering",  manager.GetFactory("Graph"));
-
-        P->SetFactory("A", filterFactory);
-
-      } else {
-        P->SetFactory("A", manager.GetFactory("Graph"));
-      }
-    }
+#endif
+}
+
+// =====================================================================================================
+// ============================== Algorithm: Smoothed Aggregation ======================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_SA(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const {
+  // Smoothed aggregation
+  MUELU_KOKKOS_FACTORY(P, SaPFactory, SaPFactory_kokkos);
+  ParameterList Pparams;
+  if (paramList.isSublist("matrixmatrix: kernel params"))
+    Pparams.sublist("matrixmatrix: kernel params", false) = paramList.sublist("matrixmatrix: kernel params");
+  if (defaultList.isSublist("matrixmatrix: kernel params"))
+    Pparams.sublist("matrixmatrix: kernel params", false) = defaultList.sublist("matrixmatrix: kernel params");
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: damping factor", double, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: calculate eigenvalue estimate", bool, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: max eigenvalue", double, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: eigenvalue estimate num iterations", int, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: use rowsumabs diagonal scaling", bool, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs diagonal replacement tolerance", double, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs diagonal replacement value", double, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: rowsumabs use automatic diagonal tolerance", bool, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: enforce constraints", bool, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "sa: eigen-analysis type", std::string, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "tentative: calculate qr", bool, Pparams);
+
+  P->SetParameterList(Pparams);
+
+  // Filtering
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "sa: use filtered matrix", bool, useFiltering);
+  if (useFiltering) {
+    // NOTE: Here, non-Kokkos and Kokkos versions diverge in the way the
+    // dependency tree is setup. The Kokkos version has merged the the
+    // FilteredAFactory into the CoalesceDropFactory.
+    if (!useKokkos_) {
+      RCP<Factory> filterFactory = rcp(new FilteredAFactory());
+
+      ParameterList fParams;
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use root stencil", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: Dirichlet threshold", double, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use spread lumping", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom growth factor", double, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom cap", double, fParams);
+      filterFactory->SetParameterList(fParams);
+      filterFactory->SetFactory("Graph", manager.GetFactory("Graph"));
+      filterFactory->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+      filterFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
+      // I'm not sure why we need this line. See comments for DofsPerNode for UncoupledAggregation above
+      filterFactory->SetFactory("Filtering", manager.GetFactory("Graph"));
+
+      P->SetFactory("A", filterFactory);
 
-    // Energy minimization
-    ParameterList Pparams;
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: num iterations",           int, Pparams);
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: iterative method", std::string, Pparams);
-    if (reuseType == "emin") {
-      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: num reuse iterations",   int, Pparams);
-      Pparams.set("Keep P0",          true);
-      Pparams.set("Keep Constraint0", true);
+    } else {
+      P->SetFactory("A", manager.GetFactory("Graph"));
     }
-    P->SetParameterList(Pparams);
-    P->SetFactory("P",          manager.GetFactory("Ptent"));
-    P->SetFactory("Constraint", manager.GetFactory("Constraint"));
-    manager.SetFactory("P", P);
-  }
-
-  // =====================================================================================================
-  // ================================= Algorithm: Petrov-Galerkin ========================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_PG(ParameterList& /* paramList */, const ParameterList& /* defaultList */, FactoryManager& manager,
-                          int /* levelID */, std::vector<keep_pair>& /* keeps */) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(this->implicitTranspose_, Exceptions::RuntimeError,
-        "Implicit transpose not supported with Petrov-Galerkin smoothed transfer operators: Set \"transpose: use implicit\" to false!\n" \
-        "Petrov-Galerkin transfer operator smoothing for non-symmetric problems requires a separate handling of the restriction operator which " \
-        "does not allow the usage of implicit transpose easily.");
-
-    // Petrov-Galerkin
-    auto P = rcp(new PgPFactory());
-    P->SetFactory("P", manager.GetFactory("Ptent"));
-    manager.SetFactory("P", P);
   }
 
-  // =====================================================================================================
-  // ================================= Algorithm: Replicate       ========================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Replicate(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const
-  {
-    auto P = rcp(new MueLu::ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
-
-    ParameterList Pparams;
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "replicate: npdes", int, Pparams);
-
-    P->SetParameterList(Pparams);
-    manager.SetFactory("P", P);
+  P->SetFactory("P", manager.GetFactory("Ptent"));
+  manager.SetFactory("P", P);
+
+  bool filteringChangesMatrix = useFiltering && !MUELU_TEST_PARAM_2LIST(paramList, defaultList, "aggregation: drop tol", double, 0);
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  if (reuseType == "tP" && !filteringChangesMatrix)
+    keeps.push_back(keep_pair("AP reuse data", P.get()));
+}
+
+// =====================================================================================================
+// =============================== Algorithm: Energy Minimization ======================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Emin(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager,
+                              int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "emin: pattern", std::string, patternType);
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "reuse: type", std::string, reuseType);
+  TEUCHOS_TEST_FOR_EXCEPTION(patternType != "AkPtent", Exceptions::InvalidArgument,
+                             "Invalid pattern name: \"" << patternType << "\". Valid options: \"AkPtent\"");
+  // Pattern
+  auto patternFactory = rcp(new PatternFactory());
+  ParameterList patternParams;
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: pattern order", int, patternParams);
+  patternFactory->SetParameterList(patternParams);
+  patternFactory->SetFactory("P", manager.GetFactory("Ptent"));
+  manager.SetFactory("Ppattern", patternFactory);
+
+  // Constraint
+  auto constraintFactory = rcp(new ConstraintFactory());
+  constraintFactory->SetFactory("Ppattern", manager.GetFactory("Ppattern"));
+  constraintFactory->SetFactory("CoarseNullspace", manager.GetFactory("Ptent"));
+  manager.SetFactory("Constraint", constraintFactory);
+
+  // Emin Factory
+  auto P = rcp(new EminPFactory());
+  // Filtering
+  MUELU_SET_VAR_2LIST(paramList, defaultList, "emin: use filtered matrix", bool, useFiltering);
+  if (useFiltering) {
+    // NOTE: Here, non-Kokkos and Kokkos versions diverge in the way the
+    // dependency tree is setup. The Kokkos version has merged the the
+    // FilteredAFactory into the CoalesceDropFactory.
+    if (!useKokkos_) {
+      RCP<Factory> filterFactory = rcp(new FilteredAFactory());
+
+      ParameterList fParams;
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use lumping", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse graph", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: reuse eigenvalue", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use root stencil", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: Dirichlet threshold", double, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: use spread lumping", bool, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom growth factor", double, fParams);
+      MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "filtered matrix: spread lumping diag dom cap", double, fParams);
+      filterFactory->SetParameterList(fParams);
+      filterFactory->SetFactory("Graph", manager.GetFactory("Graph"));
+      filterFactory->SetFactory("Aggregates", manager.GetFactory("Aggregates"));
+      filterFactory->SetFactory("UnAmalgamationInfo", manager.GetFactory("UnAmalgamationInfo"));
+      // I'm not sure why we need this line. See comments for DofsPerNode for UncoupledAggregation above
+      filterFactory->SetFactory("Filtering", manager.GetFactory("Graph"));
+
+      P->SetFactory("A", filterFactory);
 
+    } else {
+      P->SetFactory("A", manager.GetFactory("Graph"));
+    }
   }
 
-  // =====================================================================================================
-  // ====================================== Algorithm: Combine ============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Combine(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const
-  {
-    auto P = rcp(new MueLu::CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
-
-    ParameterList Pparams;
-    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "combine: numBlks", int, Pparams);
-
-    P->SetParameterList(Pparams);
-    manager.SetFactory("P", P);
-
+  // Energy minimization
+  ParameterList Pparams;
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: num iterations", int, Pparams);
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: iterative method", std::string, Pparams);
+  if (reuseType == "emin") {
+    MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "emin: num reuse iterations", int, Pparams);
+    Pparams.set("Keep P0", true);
+    Pparams.set("Keep Constraint0", true);
   }
-
-
-  // =====================================================================================================
-  // ====================================== Algorithm: Matlab ============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  UpdateFactoryManager_Matlab(ParameterList& paramList, const ParameterList& /* defaultList */, FactoryManager& manager,
-                              int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  P->SetParameterList(Pparams);
+  P->SetFactory("P", manager.GetFactory("Ptent"));
+  P->SetFactory("Constraint", manager.GetFactory("Constraint"));
+  manager.SetFactory("P", P);
+}
+
+// =====================================================================================================
+// ================================= Algorithm: Petrov-Galerkin ========================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_PG(ParameterList& /* paramList */, const ParameterList& /* defaultList */, FactoryManager& manager,
+                            int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(this->implicitTranspose_, Exceptions::RuntimeError,
+                             "Implicit transpose not supported with Petrov-Galerkin smoothed transfer operators: Set \"transpose: use implicit\" to false!\n"
+                             "Petrov-Galerkin transfer operator smoothing for non-symmetric problems requires a separate handling of the restriction operator which "
+                             "does not allow the usage of implicit transpose easily.");
+
+  // Petrov-Galerkin
+  auto P = rcp(new PgPFactory());
+  P->SetFactory("P", manager.GetFactory("Ptent"));
+  manager.SetFactory("P", P);
+}
+
+// =====================================================================================================
+// ================================= Algorithm: Replicate       ========================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Replicate(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const {
+  auto P = rcp(new MueLu::ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+
+  ParameterList Pparams;
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "replicate: npdes", int, Pparams);
+
+  P->SetParameterList(Pparams);
+  manager.SetFactory("P", P);
+}
+
+// =====================================================================================================
+// ====================================== Algorithm: Combine ============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Combine(ParameterList& paramList, const ParameterList& defaultList, FactoryManager& manager, int /* levelID */, std::vector<keep_pair>& keeps) const {
+  auto P = rcp(new MueLu::CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+
+  ParameterList Pparams;
+  MUELU_TEST_AND_SET_PARAM_2LIST(paramList, defaultList, "combine: numBlks", int, Pparams);
+
+  P->SetParameterList(Pparams);
+  manager.SetFactory("P", P);
+}
+
+// =====================================================================================================
+// ====================================== Algorithm: Matlab ============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    UpdateFactoryManager_Matlab(ParameterList& paramList, const ParameterList& /* defaultList */, FactoryManager& manager,
+                                int /* levelID */, std::vector<keep_pair>& /* keeps */) const {
 #ifdef HAVE_MUELU_MATLAB
-    ParameterList Pparams = paramList.sublist("transfer: params");
-    auto P = rcp(new TwoLevelMatlabFactory());
-    P->SetParameterList(Pparams);
-    P->SetFactory("P", manager.GetFactory("Ptent"));
-    manager.SetFactory("P", P);
+  ParameterList Pparams = paramList.sublist("transfer: params");
+  auto P                = rcp(new TwoLevelMatlabFactory());
+  P->SetParameterList(Pparams);
+  P->SetFactory("P", manager.GetFactory("Ptent"));
+  manager.SetFactory("P", P);
 #else
-    (void)paramList;
-    (void)manager;
+  (void)paramList;
+  (void)manager;
 #endif
-  }
+}
 
 #undef MUELU_SET_VAR_2LIST
 #undef MUELU_TEST_AND_SET_VAR
@@ -2194,546 +2150,550 @@ namespace MueLu {
 #undef MUELU_TEST_PARAM_2LIST
 #undef MUELU_KOKKOS_FACTORY
 
-  size_t LevenshteinDistance(const char* s, size_t len_s, const char* t, size_t len_t);
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Validate(const ParameterList& constParamList) const {
-    ParameterList paramList = constParamList;
-    const ParameterList& validList = *MasterList::List();
-    // Validate up to maxLevels level specific parameter sublists
-    const int maxLevels = 100;
-
-    // Extract level specific list
-    std::vector<ParameterList> paramLists;
-    for (int levelID = 0; levelID < maxLevels; levelID++) {
-      std::string sublistName = "level " + toString(levelID);
-      if (paramList.isSublist(sublistName)) {
-        paramLists.push_back(paramList.sublist(sublistName));
-        // paramLists.back().setName(sublistName);
-        paramList.remove(sublistName);
-      }
+size_t LevenshteinDistance(const char* s, size_t len_s, const char* t, size_t len_t);
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Validate(const ParameterList& constParamList) const {
+  ParameterList paramList        = constParamList;
+  const ParameterList& validList = *MasterList::List();
+  // Validate up to maxLevels level specific parameter sublists
+  const int maxLevels = 100;
+
+  // Extract level specific list
+  std::vector<ParameterList> paramLists;
+  for (int levelID = 0; levelID < maxLevels; levelID++) {
+    std::string sublistName = "level " + toString(levelID);
+    if (paramList.isSublist(sublistName)) {
+      paramLists.push_back(paramList.sublist(sublistName));
+      // paramLists.back().setName(sublistName);
+      paramList.remove(sublistName);
     }
-    paramLists.push_back(paramList);
-    // paramLists.back().setName("main");
+  }
+  paramLists.push_back(paramList);
+  // paramLists.back().setName("main");
 #ifdef HAVE_MUELU_MATLAB
-    // If Muemex is supported, hide custom level variables from validator by removing them from paramList's sublists
-    for (size_t i = 0; i < paramLists.size(); i++) {
-      std::vector<std::string> customVars; // list of names (keys) to be removed from list
-
-      for(Teuchos::ParameterList::ConstIterator it = paramLists[i].begin(); it != paramLists[i].end(); it++) {
-        std::string paramName = paramLists[i].name(it);
+  // If Muemex is supported, hide custom level variables from validator by removing them from paramList's sublists
+  for (size_t i = 0; i < paramLists.size(); i++) {
+    std::vector<std::string> customVars;  // list of names (keys) to be removed from list
 
-        if (IsParamMuemexVariable(paramName))
-          customVars.push_back(paramName);
-      }
+    for (Teuchos::ParameterList::ConstIterator it = paramLists[i].begin(); it != paramLists[i].end(); it++) {
+      std::string paramName = paramLists[i].name(it);
 
-      // Remove the keys
-      for (size_t j = 0; j < customVars.size(); j++)
-        paramLists[i].remove(customVars[j], false);
+      if (IsParamMuemexVariable(paramName))
+        customVars.push_back(paramName);
     }
+
+    // Remove the keys
+    for (size_t j = 0; j < customVars.size(); j++)
+      paramLists[i].remove(customVars[j], false);
+  }
 #endif
 
-    const int maxDepth = 0;
-    for (size_t i = 0; i < paramLists.size(); i++) {
-      // validate every sublist
-      try {
-        paramLists[i].validateParameters(validList, maxDepth);
-
-      } catch (const Teuchos::Exceptions::InvalidParameterName& e) {
-        std::string eString = e.what();
-
-        // Parse name from: <Error, the parameter {name="smoothe: type",...>
-        size_t nameStart = eString.find_first_of('"') + 1;
-        size_t nameEnd   = eString.find_first_of('"', nameStart);
-        std::string name = eString.substr(nameStart, nameEnd - nameStart);
-
-        size_t bestScore = 100;
-        std::string bestName  = "";
-        for (ParameterList::ConstIterator it = validList.begin(); it != validList.end(); it++) {
-          const std::string& pName = validList.name(it);
-          this->GetOStream(Runtime1) << "| " << pName;
-          size_t score = LevenshteinDistance(name.c_str(), name.length(), pName.c_str(), pName.length());
-          this->GetOStream(Runtime1) << " -> " << score << std::endl;
-          if (score < bestScore) {
-            bestScore = score;
-            bestName  = pName;
-          }
+  const int maxDepth = 0;
+  for (size_t i = 0; i < paramLists.size(); i++) {
+    // validate every sublist
+    try {
+      paramLists[i].validateParameters(validList, maxDepth);
+
+    } catch (const Teuchos::Exceptions::InvalidParameterName& e) {
+      std::string eString = e.what();
+
+      // Parse name from: <Error, the parameter {name="smoothe: type",...>
+      size_t nameStart = eString.find_first_of('"') + 1;
+      size_t nameEnd   = eString.find_first_of('"', nameStart);
+      std::string name = eString.substr(nameStart, nameEnd - nameStart);
+
+      size_t bestScore     = 100;
+      std::string bestName = "";
+      for (ParameterList::ConstIterator it = validList.begin(); it != validList.end(); it++) {
+        const std::string& pName = validList.name(it);
+        this->GetOStream(Runtime1) << "| " << pName;
+        size_t score = LevenshteinDistance(name.c_str(), name.length(), pName.c_str(), pName.length());
+        this->GetOStream(Runtime1) << " -> " << score << std::endl;
+        if (score < bestScore) {
+          bestScore = score;
+          bestName  = pName;
         }
-        if (bestScore < 10 && bestName != "") {
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Teuchos::Exceptions::InvalidParameterName,
-            eString << "The parameter name \"" + name + "\" is not valid. Did you mean \"" + bestName << "\"?\n");
+      }
+      if (bestScore < 10 && bestName != "") {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Teuchos::Exceptions::InvalidParameterName,
+                                   eString << "The parameter name \"" + name + "\" is not valid. Did you mean \"" + bestName << "\"?\n");
 
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Teuchos::Exceptions::InvalidParameterName,
-            eString << "The parameter name \"" + name + "\" is not valid.\n");
-        }
+      } else {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Teuchos::Exceptions::InvalidParameterName,
+                                   eString << "The parameter name \"" + name + "\" is not valid.\n");
       }
     }
   }
+}
+
+// =====================================================================================================
+// ==================================== FACTORY interpreter ============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SetFactoryParameterList(const ParameterList& constParamList) {
+  // Create a non const copy of the parameter list
+  // Working with a modifiable list is much much easier than with original one
+  ParameterList paramList = constParamList;
+
+  // Parameter List Parsing:
+  // ---------
+  //   <ParameterList name="MueLu">
+  //     <ParameterList name="Matrix">
+  //   </ParameterList>
+  if (paramList.isSublist("Matrix")) {
+    blockSize_ = paramList.sublist("Matrix").get<int>("PDE equations", MasterList::getDefault<int>("number of equations"));
+    dofOffset_ = paramList.sublist("Matrix").get<GlobalOrdinal>("DOF offset", 0);  // undocumented parameter allowing to define a DOF offset of the global dofs of an operator (defaul = 0)
+  }
 
-  // =====================================================================================================
-  // ==================================== FACTORY interpreter ============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  SetFactoryParameterList(const ParameterList& constParamList) {
-    // Create a non const copy of the parameter list
-    // Working with a modifiable list is much much easier than with original one
-    ParameterList paramList = constParamList;
-
-    // Parameter List Parsing:
-    // ---------
-    //   <ParameterList name="MueLu">
-    //     <ParameterList name="Matrix">
-    //   </ParameterList>
-    if (paramList.isSublist("Matrix")) {
-      blockSize_ = paramList.sublist("Matrix").get<int>("PDE equations", MasterList::getDefault<int>("number of equations"));
-      dofOffset_ = paramList.sublist("Matrix").get<GlobalOrdinal>("DOF offset", 0); // undocumented parameter allowing to define a DOF offset of the global dofs of an operator (defaul = 0)
+  // create new FactoryFactory object if necessary
+  if (factFact_ == Teuchos::null)
+    factFact_ = Teuchos::rcp(new FactoryFactory());
+
+  // Parameter List Parsing:
+  // ---------
+  //   <ParameterList name="MueLu">
+  //     <ParameterList name="Factories"> <== call BuildFactoryMap() on this parameter list
+  //     ...
+  //     </ParameterList>
+  //   </ParameterList>
+  FactoryMap factoryMap;
+  FactoryManagerMap factoryManagers;
+  if (paramList.isSublist("Factories"))
+    this->BuildFactoryMap(paramList.sublist("Factories"), factoryMap, factoryMap, factoryManagers);
+
+  // Parameter List Parsing:
+  // ---------
+  //   <ParameterList name="MueLu">
+  //     <ParameterList name="Hierarchy">
+  //       <Parameter name="verbose"  type="string" value="Warnings"/> <== get
+  //       <Parameter name="numDesiredLevel" type="int" value="10"/>   <== get
+  //
+  //       <ParameterList name="firstLevel">                           <== parse first args and call BuildFactoryMap() on the rest of this parameter list
+  //         ...
+  //       </ParameterList>
+  //     </ParameterList>
+  //   </ParameterList>
+  if (paramList.isSublist("Hierarchy")) {
+    ParameterList hieraList = paramList.sublist("Hierarchy");  // copy because list temporally modified (remove 'id')
+
+    // Get hierarchy options
+    if (hieraList.isParameter("max levels")) {
+      this->numDesiredLevel_ = hieraList.get<int>("max levels");
+      hieraList.remove("max levels");
     }
 
-    // create new FactoryFactory object if necessary
-    if (factFact_ == Teuchos::null)
-      factFact_ = Teuchos::rcp(new FactoryFactory());
-
-    // Parameter List Parsing:
-    // ---------
-    //   <ParameterList name="MueLu">
-    //     <ParameterList name="Factories"> <== call BuildFactoryMap() on this parameter list
-    //     ...
-    //     </ParameterList>
-    //   </ParameterList>
-    FactoryMap factoryMap;
-    FactoryManagerMap factoryManagers;
-    if (paramList.isSublist("Factories"))
-      this->BuildFactoryMap(paramList.sublist("Factories"), factoryMap, factoryMap, factoryManagers);
-
-    // Parameter List Parsing:
-    // ---------
-    //   <ParameterList name="MueLu">
-    //     <ParameterList name="Hierarchy">
-    //       <Parameter name="verbose"  type="string" value="Warnings"/> <== get
-    //       <Parameter name="numDesiredLevel" type="int" value="10"/>   <== get
-    //
-    //       <ParameterList name="firstLevel">                           <== parse first args and call BuildFactoryMap() on the rest of this parameter list
-    //         ...
-    //       </ParameterList>
-    //     </ParameterList>
-    //   </ParameterList>
-    if (paramList.isSublist("Hierarchy")) {
-      ParameterList hieraList = paramList.sublist("Hierarchy"); // copy because list temporally modified (remove 'id')
-
-      // Get hierarchy options
-      if (hieraList.isParameter("max levels")) {
-        this->numDesiredLevel_ = hieraList.get<int>("max levels");
-        hieraList.remove("max levels");
-      }
+    if (hieraList.isParameter("coarse: max size")) {
+      this->maxCoarseSize_ = hieraList.get<int>("coarse: max size");
+      hieraList.remove("coarse: max size");
+    }
 
-      if (hieraList.isParameter("coarse: max size")) {
-        this->maxCoarseSize_ = hieraList.get<int>("coarse: max size");
-        hieraList.remove("coarse: max size");
-      }
+    if (hieraList.isParameter("repartition: rebalance P and R")) {
+      this->doPRrebalance_ = hieraList.get<bool>("repartition: rebalance P and R");
+      hieraList.remove("repartition: rebalance P and R");
+    }
 
-      if (hieraList.isParameter("repartition: rebalance P and R")) {
-        this->doPRrebalance_ = hieraList.get<bool>("repartition: rebalance P and R");
-        hieraList.remove("repartition: rebalance P and R");
-      }
+    if (hieraList.isParameter("transpose: use implicit")) {
+      this->implicitTranspose_ = hieraList.get<bool>("transpose: use implicit");
+      hieraList.remove("transpose: use implicit");
+    }
 
-      if (hieraList.isParameter("transpose: use implicit")) {
-        this->implicitTranspose_ = hieraList.get<bool>("transpose: use implicit");
-        hieraList.remove("transpose: use implicit");
-      }
+    if (hieraList.isParameter("fuse prolongation and update")) {
+      this->fuseProlongationAndUpdate_ = hieraList.get<bool>("fuse prolongation and update");
+      hieraList.remove("fuse prolongation and update");
+    }
 
-      if (hieraList.isParameter("fuse prolongation and update")) {
-        this->fuseProlongationAndUpdate_ = hieraList.get<bool>("fuse prolongation and update");
-        hieraList.remove("fuse prolongation and update");
-      }
+    if (hieraList.isParameter("nullspace: suppress dimension check")) {
+      this->suppressNullspaceDimensionCheck_ = hieraList.get<bool>("nullspace: suppress dimension check");
+      hieraList.remove("nullspace: suppress dimension check");
+    }
 
-      if (hieraList.isParameter("nullspace: suppress dimension check")) {
-        this->suppressNullspaceDimensionCheck_ = hieraList.get<bool>("nullspace: suppress dimension check");
-        hieraList.remove("nullspace: suppress dimension check");
-      }
+    if (hieraList.isParameter("number of vectors")) {
+      this->numDesiredLevel_ = hieraList.get<int>("number of vectors");
+      hieraList.remove("number of vectors");
+    }
 
-      if (hieraList.isParameter("number of vectors")) {
-        this->numDesiredLevel_ = hieraList.get<int>("number of vectors");
-        hieraList.remove("number of vectors");
-      }
+    if (hieraList.isSublist("matvec params"))
+      this->matvecParams_ = Teuchos::parameterList(hieraList.sublist("matvec params"));
 
-      if (hieraList.isSublist("matvec params"))
-        this->matvecParams_ = Teuchos::parameterList(hieraList.sublist("matvec params"));
+    if (hieraList.isParameter("coarse grid correction scaling factor")) {
+      this->scalingFactor_ = hieraList.get<double>("coarse grid correction scaling factor");
+      hieraList.remove("coarse grid correction scaling factor");
+    }
 
+    // Translate cycle type parameter
+    if (hieraList.isParameter("cycle type")) {
+      std::map<std::string, CycleType> cycleMap;
+      cycleMap["V"] = VCYCLE;
+      cycleMap["W"] = WCYCLE;
 
-      if (hieraList.isParameter("coarse grid correction scaling factor")) {
-        this->scalingFactor_ = hieraList.get<double>("coarse grid correction scaling factor");
-        hieraList.remove("coarse grid correction scaling factor");
-      }
+      std::string cycleType = hieraList.get<std::string>("cycle type");
+      TEUCHOS_TEST_FOR_EXCEPTION(cycleMap.count(cycleType) == 0, Exceptions::RuntimeError, "Invalid cycle type: \"" << cycleType << "\"");
+      this->Cycle_ = cycleMap[cycleType];
+    }
 
-      // Translate cycle type parameter
-      if (hieraList.isParameter("cycle type")) {
-        std::map<std::string, CycleType> cycleMap;
-        cycleMap["V"] = VCYCLE;
-        cycleMap["W"] = WCYCLE;
+    if (hieraList.isParameter("W cycle start level")) {
+      this->WCycleStartLevel_ = hieraList.get<int>("W cycle start level");
+    }
 
-        std::string cycleType = hieraList.get<std::string>("cycle type");
-        TEUCHOS_TEST_FOR_EXCEPTION(cycleMap.count(cycleType) == 0, Exceptions::RuntimeError, "Invalid cycle type: \"" << cycleType << "\"");
-        this->Cycle_ = cycleMap[cycleType];
-      }
+    if (hieraList.isParameter("verbosity")) {
+      std::string vl = hieraList.get<std::string>("verbosity");
+      hieraList.remove("verbosity");
+      this->verbosity_ = toVerbLevel(vl);
+    }
 
-      if (hieraList.isParameter("W cycle start level")) {
-        this->WCycleStartLevel_ = hieraList.get<int>("W cycle start level");
-      }
+    if (hieraList.isParameter("output filename"))
+      VerboseObject::SetMueLuOFileStream(hieraList.get<std::string>("output filename"));
+
+    if (hieraList.isParameter("dependencyOutputLevel"))
+      this->graphOutputLevel_ = hieraList.get<int>("dependencyOutputLevel");
+
+    // Check for the reuse case
+    if (hieraList.isParameter("reuse"))
+      Factory::DisableMultipleCheckGlobally();
+
+    if (hieraList.isSublist("DataToWrite")) {
+      // TODO We should be able to specify any data.  If it exists, write it.
+      // TODO This would requires something like std::set<dataName, Array<int> >
+      ParameterList foo    = hieraList.sublist("DataToWrite");
+      std::string dataName = "Matrices";
+      if (foo.isParameter(dataName))
+        this->matricesToPrint_["A"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
+      dataName = "Prolongators";
+      if (foo.isParameter(dataName))
+        this->matricesToPrint_["P"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
+      dataName = "Restrictors";
+      if (foo.isParameter(dataName))
+        this->matricesToPrint_["R"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
+      dataName = "D0";
+      if (foo.isParameter(dataName))
+        this->matricesToPrint_["D0"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
+    }
 
-      if (hieraList.isParameter("verbosity")) {
-        std::string vl = hieraList.get<std::string>("verbosity");
-        hieraList.remove("verbosity");
-        this->verbosity_ = toVerbLevel(vl);
-      }
+    // Get level configuration
+    for (ParameterList::ConstIterator param = hieraList.begin(); param != hieraList.end(); ++param) {
+      const std::string& paramName = hieraList.name(param);
 
-      if (hieraList.isParameter("output filename"))
-        VerboseObject::SetMueLuOFileStream(hieraList.get<std::string>("output filename"));
-
-      if (hieraList.isParameter("dependencyOutputLevel"))
-        this->graphOutputLevel_ = hieraList.get<int>("dependencyOutputLevel");
-
-      // Check for the reuse case
-      if (hieraList.isParameter("reuse"))
-        Factory::DisableMultipleCheckGlobally();
-
-      if (hieraList.isSublist("DataToWrite")) {
-        //TODO We should be able to specify any data.  If it exists, write it.
-        //TODO This would requires something like std::set<dataName, Array<int> >
-        ParameterList foo = hieraList.sublist("DataToWrite");
-        std::string dataName = "Matrices";
-        if (foo.isParameter(dataName))
-          this->matricesToPrint_["A"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
-        dataName = "Prolongators";
-        if (foo.isParameter(dataName))
-          this->matricesToPrint_["P"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
-        dataName = "Restrictors";
-        if (foo.isParameter(dataName))
-          this->matricesToPrint_["R"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
-        dataName = "D0";
-        if (foo.isParameter(dataName))
-          this->matricesToPrint_["D0"] = Teuchos::getArrayFromStringParameter<int>(foo, dataName);
-      }
+      if (paramName != "DataToWrite" && hieraList.isSublist(paramName)) {
+        ParameterList levelList = hieraList.sublist(paramName);  // copy because list temporally modified (remove 'id')
 
-      // Get level configuration
-      for (ParameterList::ConstIterator param = hieraList.begin(); param != hieraList.end(); ++param) {
-        const std::string & paramName  = hieraList.name(param);
-
-        if (paramName != "DataToWrite" && hieraList.isSublist(paramName)) {
-          ParameterList levelList = hieraList.sublist(paramName); // copy because list temporally modified (remove 'id')
-
-          int startLevel = 0;       if(levelList.isParameter("startLevel"))      { startLevel      = levelList.get<int>("startLevel");      levelList.remove("startLevel"); }
-          int numDesiredLevel = 1;  if(levelList.isParameter("numDesiredLevel")) { numDesiredLevel = levelList.get<int>("numDesiredLevel"); levelList.remove("numDesiredLevel"); }
-
-          // Parameter List Parsing:
-          // ---------
-          //   <ParameterList name="firstLevel">
-          //      <Parameter name="startLevel"       type="int" value="0"/>
-          //      <Parameter name="numDesiredLevel"  type="int" value="1"/>
-          //      <Parameter name="verbose"          type="string" value="Warnings"/>
-          //
-          //      [] <== call BuildFactoryMap() on the rest of the parameter list
-          //
-          //  </ParameterList>
-          FactoryMap levelFactoryMap;
-          BuildFactoryMap(levelList, factoryMap, levelFactoryMap, factoryManagers);
-
-          RCP<FactoryManager> m = rcp(new FactoryManager(levelFactoryMap));
-          if (hieraList.isParameter("use kokkos refactor"))
-            m->SetKokkosRefactor(hieraList.get<bool>("use kokkos refactor"));
-
-          if (startLevel >= 0)
-            this->AddFactoryManager(startLevel, numDesiredLevel, m);
-          else
-            TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::ParameterListInterpreter():: invalid level id");
-        } /* TODO: else { } */
-      }
+        int startLevel = 0;
+        if (levelList.isParameter("startLevel")) {
+          startLevel = levelList.get<int>("startLevel");
+          levelList.remove("startLevel");
+        }
+        int numDesiredLevel = 1;
+        if (levelList.isParameter("numDesiredLevel")) {
+          numDesiredLevel = levelList.get<int>("numDesiredLevel");
+          levelList.remove("numDesiredLevel");
+        }
+
+        // Parameter List Parsing:
+        // ---------
+        //   <ParameterList name="firstLevel">
+        //      <Parameter name="startLevel"       type="int" value="0"/>
+        //      <Parameter name="numDesiredLevel"  type="int" value="1"/>
+        //      <Parameter name="verbose"          type="string" value="Warnings"/>
+        //
+        //      [] <== call BuildFactoryMap() on the rest of the parameter list
+        //
+        //  </ParameterList>
+        FactoryMap levelFactoryMap;
+        BuildFactoryMap(levelList, factoryMap, levelFactoryMap, factoryManagers);
+
+        RCP<FactoryManager> m = rcp(new FactoryManager(levelFactoryMap));
+        if (hieraList.isParameter("use kokkos refactor"))
+          m->SetKokkosRefactor(hieraList.get<bool>("use kokkos refactor"));
+
+        if (startLevel >= 0)
+          this->AddFactoryManager(startLevel, numDesiredLevel, m);
+        else
+          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::ParameterListInterpreter():: invalid level id");
+      } /* TODO: else { } */
     }
   }
+}
+
+// TODO: static?
+/// \brief Interpret "Factories" sublist
+///
+/// \param paramList [in]: "Factories" ParameterList
+/// \param factoryMapIn [in]: FactoryMap maps variable names to factories. This factory map is used to resolve data dependencies of previously defined factories.
+/// \param factoryMapOut [out]: FactoryMap maps variable names to factories. New factory entries are added to that FactoryMap. Usually, factoryMapIn and factoryMapOut should use the same object, such that new factories are added. We have to distinguish input and output if we build sub-factory managers, though.
+/// \param factoryManagers [in/out]: FacotryManagerMap maps group names to a FactoryManager object.
+///
+/// Interpret "Factories" parameter list. For each "factory" entry, add a new entry in the factoryMapOut map or create a new FacotryManager
+///
+/// Parameter List Parsing:
+/// Create an entry in factoryMapOut for each parameter of the list paramList
+/// ---------
+///    <ParameterList name="...">
+///     <Parameter name="smootherFact0" type="string" value="TrilinosSmoother"/>
+///
+///     <ParameterList name="smootherFact1">
+///       <Parameter name="type" type="string" value="TrilinosSmoother"/>
+///       ...
+///     </ParameterList>
+///    </ParameterList>
+///
+/// ---------
+/// Group factories
+/// We can group factories using parameter sublists with the "group" parameter
+///
+///    <ParameterList name="myFirstGroup">
+///      <Parameter name="group" type="string" value="FactoryManager"/>
+///      <Parameter name="A" type="string" value="mySubBlockAFactory1"/>
+///      <Parameter name="P" type="string" value="myTentativePFact1"/>
+///      <Parameter name="Aggregates" type="string" value="myAggFact1"/>
+///      <Parameter name="Nullspace" type="string" value="myNspFact1"/>
+///      <Parameter name="CoarseMap" type="string" value="myCoarseMap1"/>
+///    </ParameterList>
+
+///    <ParameterList name="mySecondGroup">
+///      <Parameter name="group" type="string" value="FactoryManager"/>
+///      <Parameter name="A" type="string" value="mySubBlockAFactory2"/>
+///      <Parameter name="P" type="string" value="myTentativePFact2"/>
+///      <Parameter name="Aggregates" type="string" value="myAggFact1"/><!-- reuse aggs -->
+///      <Parameter name="Nullspace" type="string" value="myNspFact2"/>
+///      <Parameter name="CoarseMap" type="string" value="myCoarseMap2"/>
+///    </ParameterList>
+///
+///  These factory groups can be used with factories for blocked operators (such as the BlockedPFactory)
+///  to easily define the operations on the sub-blocks.
+///
+///    <ParameterList name="myBlockedPFact">
+///      <Parameter name="factory" type="string" value="BlockedPFactory"/>
+///      <!-- factory manager for block 1 -->
+///      <ParameterList name="block1">
+///        <Parameter name="group" type="string" value="myFirstGroup"/>
+///      </ParameterList>
+///      <!-- factory manager for block 2 -->
+///      <ParameterList name="block2">
+///        <Parameter name="group" type="string" value="mySecondGroup"/>
+///      </ParameterList>
+///    </ParameterList>
+///
+///
+///  As an alternative one can also directly specify the factories in the sublists "block1", "block2", etc..., of course.
+///  But using blocks has the advantage that one can reuse them in all blocked factories.
+///
+///    <ParameterList name="myBlockedPFact">
+///      <Parameter name="factory" type="string" value="BlockedPFactory"/>
+///      <!-- factory manager for block 1 -->
+///      <ParameterList name="block1">
+///        <Parameter name="A" type="string" value="mySubBlockAFactory1"/>
+///        <Parameter name="P" type="string" value="myTentativePFact1"/>
+///        <Parameter name="Aggregates" type="string" value="myAggFact1"/>
+///        <Parameter name="Nullspace" type="string" value="myNspFact1"/>
+///        <Parameter name="CoarseMap" type="string" value="myCoarseMap1"/>
+///      </ParameterList>
+///      <!-- factory manager for block 2 -->
+///      <ParameterList name="block2">
+///        <Parameter name="A" type="string" value="mySubBlockAFactory2"/>
+///        <Parameter name="P" type="string" value="myTentativePFact2"/>
+///        <Parameter name="Aggregates" type="string" value="myAggFact1"/><!-- reuse aggs -->
+///        <Parameter name="Nullspace" type="string" value="myNspFact2"/>
+///        <Parameter name="CoarseMap" type="string" value="myCoarseMap2"/>
+///      </ParameterList>
+///    </ParameterList>
+///
+///  As an alternative one can also directly specify the factories in the sublists "block1", "block2", etc..., of course.
+///
+///
+
+/// ---------
+/// add more dependencies (circular dependencies)
+///
+///  The NullspaceFactory needs to know which factory generates the null space on the coarse level (e.g., the TentativePFactory or the RebalancedPFactory).
+///  However, we cannot set the information in this place in the xml file, since the tentative prolongator facotry is typically defined later.
+///  We have to add that dependency later to the NullspaceFactory:
+///
+///    <ParameterList name="myNspFact">
+///      <Parameter name="factory" type="string" value="NullspaceFactory"/>
+///      <!--<Paramter name="Nullspace" type="string" value="myRebalanceProlongatorFact"/>-->
+///    </ParameterList>
+///
+///    <ParameterList name="myTentativePFact">
+///      <Parameter name="factory"   type="string" value="TentativePFactory"/>
+///      <...>
+///      <Parameter name="Nullspace" type="string" value="myNspFact"/>
+///      <Parameter name="CoarseMap" type="string" value="myCoarseMap"/>
+///    </ParameterList>
+///
+///    <ParameterList name="myRebalanceProlongatorFact">
+///      <Parameter name="factory"   type="string" value="RebalanceTransferFactory"/>
+///      <...>
+///      <Parameter name="Nullspace" type="string" value="myTentativePFact"/>
+///    </ParameterList>
+///
+///  After the definition of the generating factory for the nullspace (in this case myRebalanceProlongatorFact)
+///  we add that dependency to the NullspaceFactory instance myNspFact
+///
+///    <ParameterList name="myNspFactDeps">
+///      <Parameter name="dependency for" type="string" value="myNspFact"/>
+///      <Parameter name="Nullspace" type="string" value="myRebalanceProlongatorFact"/>
+///    </ParameterList>
+///
+///  We have to create a new block (with a different name than myNspFact). In the example we use "myNspFactDeps".
+///  It should contain a parameter "dependency for" with the name of the factory that we want the dependencies to be addded to.
+///  With above block we do not need the entry for the Nullspace in the global FactoryManager any more.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildFactoryMap(const ParameterList& paramList, const FactoryMap& factoryMapIn, FactoryMap& factoryMapOut, FactoryManagerMap& factoryManagers) const {
+  for (ParameterList::ConstIterator param = paramList.begin(); param != paramList.end(); ++param) {
+    const std::string& paramName              = paramList.name(param);   //< paramName contains the user chosen factory name (e.g., "smootherFact1")
+    const Teuchos::ParameterEntry& paramValue = paramList.entry(param);  //< for factories, paramValue should be either a list or just a MueLu Factory (e.g., TrilinosSmoother)
+
+    // TODO: do not allow name of existing MueLu classes (can be tested using FactoryFactory)
+
+    if (paramValue.isList()) {
+      ParameterList paramList1 = Teuchos::getValue<ParameterList>(paramValue);
+      if (paramList1.isParameter("factory")) {  // default: just a factory definition
+        // New Factory is a sublist with internal parameters and/or data dependencies
+        TEUCHOS_TEST_FOR_EXCEPTION(paramList1.isParameter("dependency for") == true, Exceptions::RuntimeError,
+                                   "MueLu::ParameterListInterpreter(): It seems that in the parameter lists for defining " << paramName << " there is both a 'factory' and 'dependency for' parameter. This is not allowed. Please remove the 'dependency for' parameter.");
 
+        factoryMapOut[paramName] = factFact_->BuildFactory(paramValue, factoryMapIn, factoryManagers);
 
-  //TODO: static?
-  /// \brief Interpret "Factories" sublist
-  ///
-  /// \param paramList [in]: "Factories" ParameterList
-  /// \param factoryMapIn [in]: FactoryMap maps variable names to factories. This factory map is used to resolve data dependencies of previously defined factories.
-  /// \param factoryMapOut [out]: FactoryMap maps variable names to factories. New factory entries are added to that FactoryMap. Usually, factoryMapIn and factoryMapOut should use the same object, such that new factories are added. We have to distinguish input and output if we build sub-factory managers, though.
-  /// \param factoryManagers [in/out]: FacotryManagerMap maps group names to a FactoryManager object.
-  ///
-  /// Interpret "Factories" parameter list. For each "factory" entry, add a new entry in the factoryMapOut map or create a new FacotryManager
-  ///
-  /// Parameter List Parsing:
-  /// Create an entry in factoryMapOut for each parameter of the list paramList
-  /// ---------
-  ///    <ParameterList name="...">
-  ///     <Parameter name="smootherFact0" type="string" value="TrilinosSmoother"/>
-  ///
-  ///     <ParameterList name="smootherFact1">
-  ///       <Parameter name="type" type="string" value="TrilinosSmoother"/>
-  ///       ...
-  ///     </ParameterList>
-  ///    </ParameterList>
-  ///
-  /// ---------
-  /// Group factories
-  /// We can group factories using parameter sublists with the "group" parameter
-  ///
-  ///    <ParameterList name="myFirstGroup">
-  ///      <Parameter name="group" type="string" value="FactoryManager"/>
-  ///      <Parameter name="A" type="string" value="mySubBlockAFactory1"/>
-  ///      <Parameter name="P" type="string" value="myTentativePFact1"/>
-  ///      <Parameter name="Aggregates" type="string" value="myAggFact1"/>
-  ///      <Parameter name="Nullspace" type="string" value="myNspFact1"/>
-  ///      <Parameter name="CoarseMap" type="string" value="myCoarseMap1"/>
-  ///    </ParameterList>
-
-  ///    <ParameterList name="mySecondGroup">
-  ///      <Parameter name="group" type="string" value="FactoryManager"/>
-  ///      <Parameter name="A" type="string" value="mySubBlockAFactory2"/>
-  ///      <Parameter name="P" type="string" value="myTentativePFact2"/>
-  ///      <Parameter name="Aggregates" type="string" value="myAggFact1"/><!-- reuse aggs -->
-  ///      <Parameter name="Nullspace" type="string" value="myNspFact2"/>
-  ///      <Parameter name="CoarseMap" type="string" value="myCoarseMap2"/>
-  ///    </ParameterList>
-  ///
-  ///  These factory groups can be used with factories for blocked operators (such as the BlockedPFactory)
-  ///  to easily define the operations on the sub-blocks.
-  ///
-  ///    <ParameterList name="myBlockedPFact">
-  ///      <Parameter name="factory" type="string" value="BlockedPFactory"/>
-  ///      <!-- factory manager for block 1 -->
-  ///      <ParameterList name="block1">
-  ///        <Parameter name="group" type="string" value="myFirstGroup"/>
-  ///      </ParameterList>
-  ///      <!-- factory manager for block 2 -->
-  ///      <ParameterList name="block2">
-  ///        <Parameter name="group" type="string" value="mySecondGroup"/>
-  ///      </ParameterList>
-  ///    </ParameterList>
-  ///
-  ///
-  ///  As an alternative one can also directly specify the factories in the sublists "block1", "block2", etc..., of course.
-  ///  But using blocks has the advantage that one can reuse them in all blocked factories.
-  ///
-  ///    <ParameterList name="myBlockedPFact">
-  ///      <Parameter name="factory" type="string" value="BlockedPFactory"/>
-  ///      <!-- factory manager for block 1 -->
-  ///      <ParameterList name="block1">
-  ///        <Parameter name="A" type="string" value="mySubBlockAFactory1"/>
-  ///        <Parameter name="P" type="string" value="myTentativePFact1"/>
-  ///        <Parameter name="Aggregates" type="string" value="myAggFact1"/>
-  ///        <Parameter name="Nullspace" type="string" value="myNspFact1"/>
-  ///        <Parameter name="CoarseMap" type="string" value="myCoarseMap1"/>
-  ///      </ParameterList>
-  ///      <!-- factory manager for block 2 -->
-  ///      <ParameterList name="block2">
-  ///        <Parameter name="A" type="string" value="mySubBlockAFactory2"/>
-  ///        <Parameter name="P" type="string" value="myTentativePFact2"/>
-  ///        <Parameter name="Aggregates" type="string" value="myAggFact1"/><!-- reuse aggs -->
-  ///        <Parameter name="Nullspace" type="string" value="myNspFact2"/>
-  ///        <Parameter name="CoarseMap" type="string" value="myCoarseMap2"/>
-  ///      </ParameterList>
-  ///    </ParameterList>
-  ///
-  ///  As an alternative one can also directly specify the factories in the sublists "block1", "block2", etc..., of course.
-  ///
-  ///
-
-  /// ---------
-  /// add more dependencies (circular dependencies)
-  ///
-  ///  The NullspaceFactory needs to know which factory generates the null space on the coarse level (e.g., the TentativePFactory or the RebalancedPFactory).
-  ///  However, we cannot set the information in this place in the xml file, since the tentative prolongator facotry is typically defined later.
-  ///  We have to add that dependency later to the NullspaceFactory:
-  ///
-  ///    <ParameterList name="myNspFact">
-  ///      <Parameter name="factory" type="string" value="NullspaceFactory"/>
-  ///      <!--<Paramter name="Nullspace" type="string" value="myRebalanceProlongatorFact"/>-->
-  ///    </ParameterList>
-  ///
-  ///    <ParameterList name="myTentativePFact">
-  ///      <Parameter name="factory"   type="string" value="TentativePFactory"/>
-  ///      <...>
-  ///      <Parameter name="Nullspace" type="string" value="myNspFact"/>
-  ///      <Parameter name="CoarseMap" type="string" value="myCoarseMap"/>
-  ///    </ParameterList>
-  ///
-  ///    <ParameterList name="myRebalanceProlongatorFact">
-  ///      <Parameter name="factory"   type="string" value="RebalanceTransferFactory"/>
-  ///      <...>
-  ///      <Parameter name="Nullspace" type="string" value="myTentativePFact"/>
-  ///    </ParameterList>
-  ///
-  ///  After the definition of the generating factory for the nullspace (in this case myRebalanceProlongatorFact)
-  ///  we add that dependency to the NullspaceFactory instance myNspFact
-  ///
-  ///    <ParameterList name="myNspFactDeps">
-  ///      <Parameter name="dependency for" type="string" value="myNspFact"/>
-  ///      <Parameter name="Nullspace" type="string" value="myRebalanceProlongatorFact"/>
-  ///    </ParameterList>
-  ///
-  ///  We have to create a new block (with a different name than myNspFact). In the example we use "myNspFactDeps".
-  ///  It should contain a parameter "dependency for" with the name of the factory that we want the dependencies to be addded to.
-  ///  With above block we do not need the entry for the Nullspace in the global FactoryManager any more.
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildFactoryMap(const ParameterList& paramList, const FactoryMap& factoryMapIn, FactoryMap& factoryMapOut, FactoryManagerMap& factoryManagers) const {
-    for (ParameterList::ConstIterator param = paramList.begin(); param != paramList.end(); ++param) {
-      const std::string             & paramName  = paramList.name(param);  //< paramName contains the user chosen factory name (e.g., "smootherFact1")
-      const Teuchos::ParameterEntry & paramValue = paramList.entry(param); //< for factories, paramValue should be either a list or just a MueLu Factory (e.g., TrilinosSmoother)
-
-      //TODO: do not allow name of existing MueLu classes (can be tested using FactoryFactory)
-
-      if (paramValue.isList()) {
-        ParameterList paramList1 = Teuchos::getValue<ParameterList>(paramValue);
-        if (paramList1.isParameter("factory")) { // default: just a factory definition
-          // New Factory is a sublist with internal parameters and/or data dependencies
-          TEUCHOS_TEST_FOR_EXCEPTION(paramList1.isParameter("dependency for") == true, Exceptions::RuntimeError,
-            "MueLu::ParameterListInterpreter(): It seems that in the parameter lists for defining " << paramName <<
-            " there is both a 'factory' and 'dependency for' parameter. This is not allowed. Please remove the 'dependency for' parameter.");
-
-          factoryMapOut[paramName] = factFact_->BuildFactory(paramValue, factoryMapIn, factoryManagers);
-
-        } else if (paramList1.isParameter("dependency for")) { // add more data dependencies to existing factory
-          TEUCHOS_TEST_FOR_EXCEPTION(paramList1.isParameter("factory") == true, Exceptions::RuntimeError,
-            "MueLu::ParameterListInterpreter(): It seems that in the parameter lists for defining " << paramName <<
-            " there is both a 'factory' and 'dependency for' parameter. This is not allowed.");
-
-          std::string factoryName = paramList1.get<std::string>("dependency for");
-
-          RCP<const FactoryBase> factbase = factoryMapIn.find(factoryName /*paramName*/)->second; // access previously defined factory
-          TEUCHOS_TEST_FOR_EXCEPTION(factbase.is_null() == true, Exceptions::RuntimeError,
-            "MueLu::ParameterListInterpreter(): could not find factory " + factoryName + " in factory map. Did you define it before?");
-
-          RCP<const Factory> factoryconst = Teuchos::rcp_dynamic_cast<const Factory>(factbase);
-          RCP<      Factory> factory      = Teuchos::rcp_const_cast<Factory>(factoryconst);
-
-          // Read the RCP<Factory> parameters of the class T
-          RCP<const ParameterList> validParamList = factory->GetValidParameterList();
-          for (ParameterList::ConstIterator vparam = validParamList->begin(); vparam != validParamList->end(); ++vparam) {
-            const std::string& pName = validParamList->name(vparam);
-
-            if (!paramList1.isParameter(pName)) {
-              // Ignore unknown parameters
-              continue;
-            }
+      } else if (paramList1.isParameter("dependency for")) {  // add more data dependencies to existing factory
+        TEUCHOS_TEST_FOR_EXCEPTION(paramList1.isParameter("factory") == true, Exceptions::RuntimeError,
+                                   "MueLu::ParameterListInterpreter(): It seems that in the parameter lists for defining " << paramName << " there is both a 'factory' and 'dependency for' parameter. This is not allowed.");
 
-            if (validParamList->isType< RCP<const FactoryBase> >(pName)) {
-              // Generate or get factory described by pName and set dependency
-              RCP<const FactoryBase> generatingFact = factFact_->BuildFactory(paramList1.getEntry(pName), factoryMapIn, factoryManagers);
-              factory->SetFactory(pName, generatingFact.create_weak());
-
-            } else if (validParamList->isType<RCP<const ParameterList> >(pName)) {
-              if (pName == "ParameterList") {
-                // NOTE: we cannot use
-                //     subList = sublist(rcpFromRef(paramList), pName)
-                // here as that would result in sublist also being a reference to a temporary object.
-                // The resulting dereferencing in the corresponding factory would then segfault
-                RCP<const ParameterList> subList = Teuchos::sublist(rcp(new ParameterList(paramList1)), pName);
-                factory->SetParameter(pName, ParameterEntry(subList));
-              }
-            } else {
-              factory->SetParameter(pName, paramList1.getEntry(pName));
-            }
+        std::string factoryName = paramList1.get<std::string>("dependency for");
+
+        RCP<const FactoryBase> factbase = factoryMapIn.find(factoryName /*paramName*/)->second;  // access previously defined factory
+        TEUCHOS_TEST_FOR_EXCEPTION(factbase.is_null() == true, Exceptions::RuntimeError,
+                                   "MueLu::ParameterListInterpreter(): could not find factory " + factoryName + " in factory map. Did you define it before?");
+
+        RCP<const Factory> factoryconst = Teuchos::rcp_dynamic_cast<const Factory>(factbase);
+        RCP<Factory> factory            = Teuchos::rcp_const_cast<Factory>(factoryconst);
+
+        // Read the RCP<Factory> parameters of the class T
+        RCP<const ParameterList> validParamList = factory->GetValidParameterList();
+        for (ParameterList::ConstIterator vparam = validParamList->begin(); vparam != validParamList->end(); ++vparam) {
+          const std::string& pName = validParamList->name(vparam);
+
+          if (!paramList1.isParameter(pName)) {
+            // Ignore unknown parameters
+            continue;
           }
 
-        } else if (paramList1.isParameter("group")) { // definitiion of a factory group (for a factory manager)
-          // Define a new (sub) FactoryManager
-          std::string groupType = paramList1.get<std::string>("group");
-          TEUCHOS_TEST_FOR_EXCEPTION(groupType!="FactoryManager", Exceptions::RuntimeError,
-                                     "group must be of type \"FactoryManager\".");
-
-          ParameterList groupList = paramList1; // copy because list temporally modified (remove 'id')
-          groupList.remove("group");
-
-          bool setKokkosRefactor = false;
-          bool kokkosRefactor = useKokkos_;
-          if (groupList.isParameter("use kokkos refactor")) {
-            kokkosRefactor = groupList.get<bool>("use kokkos refactor");
-            groupList.remove("use kokkos refactor");
-            setKokkosRefactor = true;
+          if (validParamList->isType<RCP<const FactoryBase> >(pName)) {
+            // Generate or get factory described by pName and set dependency
+            RCP<const FactoryBase> generatingFact = factFact_->BuildFactory(paramList1.getEntry(pName), factoryMapIn, factoryManagers);
+            factory->SetFactory(pName, generatingFact.create_weak());
+
+          } else if (validParamList->isType<RCP<const ParameterList> >(pName)) {
+            if (pName == "ParameterList") {
+              // NOTE: we cannot use
+              //     subList = sublist(rcpFromRef(paramList), pName)
+              // here as that would result in sublist also being a reference to a temporary object.
+              // The resulting dereferencing in the corresponding factory would then segfault
+              RCP<const ParameterList> subList = Teuchos::sublist(rcp(new ParameterList(paramList1)), pName);
+              factory->SetParameter(pName, ParameterEntry(subList));
+            }
+          } else {
+            factory->SetParameter(pName, paramList1.getEntry(pName));
           }
+        }
 
-          FactoryMap groupFactoryMap;
-          BuildFactoryMap(groupList, factoryMapIn, groupFactoryMap, factoryManagers);
+      } else if (paramList1.isParameter("group")) {  // definitiion of a factory group (for a factory manager)
+        // Define a new (sub) FactoryManager
+        std::string groupType = paramList1.get<std::string>("group");
+        TEUCHOS_TEST_FOR_EXCEPTION(groupType != "FactoryManager", Exceptions::RuntimeError,
+                                   "group must be of type \"FactoryManager\".");
+
+        ParameterList groupList = paramList1;  // copy because list temporally modified (remove 'id')
+        groupList.remove("group");
+
+        bool setKokkosRefactor = false;
+        bool kokkosRefactor    = useKokkos_;
+        if (groupList.isParameter("use kokkos refactor")) {
+          kokkosRefactor = groupList.get<bool>("use kokkos refactor");
+          groupList.remove("use kokkos refactor");
+          setKokkosRefactor = true;
+        }
 
-          // do not store groupFactoryMap in factoryMapOut
-          // Create a factory manager object from groupFactoryMap
-          RCP<FactoryManager> m = rcp(new FactoryManager(groupFactoryMap));
-          if (setKokkosRefactor)
-            m->SetKokkosRefactor(kokkosRefactor);
-          factoryManagers[paramName] = m;
+        FactoryMap groupFactoryMap;
+        BuildFactoryMap(groupList, factoryMapIn, groupFactoryMap, factoryManagers);
+
+        // do not store groupFactoryMap in factoryMapOut
+        // Create a factory manager object from groupFactoryMap
+        RCP<FactoryManager> m = rcp(new FactoryManager(groupFactoryMap));
+        if (setKokkosRefactor)
+          m->SetKokkosRefactor(kokkosRefactor);
+        factoryManagers[paramName] = m;
 
-        } else {
-          this->GetOStream(Warnings0) << "Could not interpret parameter list " << paramList1 << std::endl;
-          TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError,
-                                     "XML Parameter list must either be of type \"factory\" or of type \"group\".");
-        }
       } else {
-        // default: just a factory (no parameter list)
-        factoryMapOut[paramName] = factFact_->BuildFactory(paramValue, factoryMapIn, factoryManagers);
+        this->GetOStream(Warnings0) << "Could not interpret parameter list " << paramList1 << std::endl;
+        TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError,
+                                   "XML Parameter list must either be of type \"factory\" or of type \"group\".");
       }
+    } else {
+      // default: just a factory (no parameter list)
+      factoryMapOut[paramName] = factFact_->BuildFactory(paramValue, factoryMapIn, factoryManagers);
     }
   }
-
-  // =====================================================================================================
-  // ======================================= MISC functions ==============================================
-  // =====================================================================================================
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator& Op) const {
-    try {
-      Matrix& A = dynamic_cast<Matrix&>(Op);
-      if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blockSize_))
-        this->GetOStream(Warnings0) << "Setting matrix block size to " << blockSize_ << " (value of the parameter in the list) "
-            << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl
-            << "You may want to check \"number of equations\" (or \"PDE equations\" for factory style list) parameter." << std::endl;
-
-      A.SetFixedBlockSize(blockSize_, dofOffset_);
+}
+
+// =====================================================================================================
+// ======================================= MISC functions ==============================================
+// =====================================================================================================
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupOperator(Operator& Op) const {
+  try {
+    Matrix& A = dynamic_cast<Matrix&>(Op);
+    if (A.IsFixedBlockSizeSet() && (A.GetFixedBlockSize() != blockSize_))
+      this->GetOStream(Warnings0) << "Setting matrix block size to " << blockSize_ << " (value of the parameter in the list) "
+                                  << "instead of " << A.GetFixedBlockSize() << " (provided matrix)." << std::endl
+                                  << "You may want to check \"number of equations\" (or \"PDE equations\" for factory style list) parameter." << std::endl;
+
+    A.SetFixedBlockSize(blockSize_, dofOffset_);
 
 #ifdef HAVE_MUELU_DEBUG
-      MatrixUtils::checkLocalRowMapMatchesColMap(A);
-#endif // HAVE_MUELU_DEBUG
-
-    } catch (std::bad_cast&) {
-      this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
-    }
-  }
+    MatrixUtils::checkLocalRowMapMatchesColMap(A);
+#endif  // HAVE_MUELU_DEBUG
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy& H) const {
-    H.SetCycle(Cycle_);
-    H.SetCycleStartLevel(WCycleStartLevel_);
-    H.SetProlongatorScalingFactor(scalingFactor_);
-    HierarchyManager::SetupHierarchy(H);
+  } catch (std::bad_cast&) {
+    this->GetOStream(Warnings0) << "Skipping setting block size as the operator is not a matrix" << std::endl;
   }
-
-  static bool compare(const ParameterList& list1, const ParameterList& list2) {
-    // First loop through and validate the parameters at this level.
-    // In addition, we generate a list of sublists that we will search next
-    for (ParameterList::ConstIterator it = list1.begin(); it != list1.end(); it++) {
-      const std::string&             name   = it->first;
-      const Teuchos::ParameterEntry& entry1 = it->second;
-
-      const Teuchos::ParameterEntry *entry2 = list2.getEntryPtr(name);
-      if (!entry2)                                           // entry is not present in the second list
-        return false;
-      if (entry1.isList() && entry2->isList()) {             // sublist check
-        compare(Teuchos::getValue<ParameterList>(entry1), Teuchos::getValue<ParameterList>(*entry2));
-        continue;
-      }
-      if (entry1.getAny(false) != entry2->getAny(false))     // entries have different types or different values
-        return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ParameterListInterpreter<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHierarchy(Hierarchy& H) const {
+  H.SetCycle(Cycle_);
+  H.SetCycleStartLevel(WCycleStartLevel_);
+  H.SetProlongatorScalingFactor(scalingFactor_);
+  HierarchyManager::SetupHierarchy(H);
+}
+
+static bool compare(const ParameterList& list1, const ParameterList& list2) {
+  // First loop through and validate the parameters at this level.
+  // In addition, we generate a list of sublists that we will search next
+  for (ParameterList::ConstIterator it = list1.begin(); it != list1.end(); it++) {
+    const std::string& name               = it->first;
+    const Teuchos::ParameterEntry& entry1 = it->second;
+
+    const Teuchos::ParameterEntry* entry2 = list2.getEntryPtr(name);
+    if (!entry2)  // entry is not present in the second list
+      return false;
+    if (entry1.isList() && entry2->isList()) {  // sublist check
+      compare(Teuchos::getValue<ParameterList>(entry1), Teuchos::getValue<ParameterList>(*entry2));
+      continue;
     }
-
-    return true;
+    if (entry1.getAny(false) != entry2->getAny(false))  // entries have different types or different values
+      return false;
   }
 
-  static inline bool areSame(const ParameterList& list1, const ParameterList& list2) {
-    return compare(list1, list2) && compare(list2, list1);
-  }
+  return true;
+}
+
+static inline bool areSame(const ParameterList& list1, const ParameterList& list2) {
+  return compare(list1, list2) && compare(list2, list1);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PARAMETERLISTINTERPRETER_SHORT
 #endif /* MUELU_PARAMETERLISTINTERPRETER_DEF_HPP */
diff --git a/packages/muelu/src/Interface/MueLu_ParameterListUtils.cpp b/packages/muelu/src/Interface/MueLu_ParameterListUtils.cpp
index 051a2dfd622e..6a0b847ca311 100644
--- a/packages/muelu/src/Interface/MueLu_ParameterListUtils.cpp
+++ b/packages/muelu/src/Interface/MueLu_ParameterListUtils.cpp
@@ -49,137 +49,133 @@
 
 namespace MueLu {
 
-  /* See also: ML_Epetra::UpdateList */
-  //! @brief: merge two parameter lists
-  //!
-  //! @param source [in]: parameter lists with source parameters which are to be merged in into the dest parameter list
-  //! @param dest [in,out]: parameter list with, e.g., default parameters which is extended by parameters from source parameter list
-  //! @param overWrite (bool): if true, overwrite parameters in dest with entries from source
-  void MergeParameterList(const Teuchos::ParameterList &source, Teuchos::ParameterList &dest, bool overWrite){
-    for(Teuchos::ParameterList::ConstIterator param=source.begin(); param!=source.end(); ++param)
-      if (dest.isParameter(source.name(param)) == false || overWrite)
-        dest.setEntry(source.name(param),source.entry(param));
-  }
-
-  void CreateSublists(const Teuchos::ParameterList &List, Teuchos::ParameterList &newList)
-  {
-    using Teuchos::ParameterList;
-    using std::string;
+/* See also: ML_Epetra::UpdateList */
+//! @brief: merge two parameter lists
+//!
+//! @param source [in]: parameter lists with source parameters which are to be merged in into the dest parameter list
+//! @param dest [in,out]: parameter list with, e.g., default parameters which is extended by parameters from source parameter list
+//! @param overWrite (bool): if true, overwrite parameters in dest with entries from source
+void MergeParameterList(const Teuchos::ParameterList &source, Teuchos::ParameterList &dest, bool overWrite) {
+  for (Teuchos::ParameterList::ConstIterator param = source.begin(); param != source.end(); ++param)
+    if (dest.isParameter(source.name(param)) == false || overWrite)
+      dest.setEntry(source.name(param), source.entry(param));
+}
+
+void CreateSublists(const Teuchos::ParameterList &List, Teuchos::ParameterList &newList) {
+  using std::string;
+  using Teuchos::ParameterList;
+
+  newList.setName(List.name());
+
+  // Copy general (= not level-specific) options and sublists to the new list.
+  // - Coarse and level-specific parameters are not copied yet. They will be moved to sublists later.
+  // - Already existing level-specific lists are copied to the new list but the coarse list is not copied
+  //   yet because it has to be modified before copy (s/coarse/smoother/)
+  for (ParameterList::ConstIterator param = List.begin(); param != List.end(); ++param) {
+    const string &pname = List.name(param);
+
+    if ((pname.find(" (level", 0) == string::npos || pname.find("smoother: list (level", 0) == 0 || pname.find("aggregation: list (level", 0) == 0) &&
+        (pname.find("coarse: ", 0) == string::npos)) {
+      newList.setEntry(pname, List.entry(param));
+    }
+  }  // for
+
+  // Copy of the sublist "coarse: list" to the new list. Change "coarse:" to "smoother:" along the way.
+  if (List.isSublist("coarse: list")) {
+    const ParameterList &coarseList = List.sublist("coarse: list");
+    ParameterList &newCoarseList    = newList.sublist("coarse: list");
+    for (ParameterList::ConstIterator param = coarseList.begin(); param != coarseList.end(); ++param) {
+      const string &pname = coarseList.name(param);
+
+      if (pname.find("coarse:", 0) == 0) {
+        // change "coarse: " to "smoother:"
+        newCoarseList.setEntry("smoother: " + pname.substr(8), coarseList.entry(param));
+      } else {
+        newCoarseList.setEntry(pname, coarseList.entry(param));
+      }
+    }
+  }  // if
 
-    newList.setName(List.name());
+  // Copy of level-specific parameters and coarse parameters to sublist
+  for (ParameterList::ConstIterator param = List.begin(); param != List.end(); ++param) {
+    const string &pname = List.name(param);
+    if (pname.find(" (level", 0) != string::npos && pname.find("smoother: list (level", 0) != 0 && pname.find("aggregation: list (level", 0) != 0) {
+      // Copy level-specific parameters (smoother and aggregation)
 
-    // Copy general (= not level-specific) options and sublists to the new list.
-    // - Coarse and level-specific parameters are not copied yet. They will be moved to sublists later.
-    // - Already existing level-specific lists are copied to the new list but the coarse list is not copied
-    //   yet because it has to be modified before copy (s/coarse/smoother/)
-    for (ParameterList::ConstIterator param=List.begin(); param!=List.end(); ++param)
+      // Scan pname (ex: pname="smoother: type (level 2)")
+      string type, option;
+      int levelID = -1;
       {
-        const string & pname=List.name(param);
-
-        if ((pname.find(" (level",0)  == string::npos || pname.find("smoother: list (level",0) == 0 || pname.find("aggregation: list (level",0) == 0) &&
-            (pname.find("coarse: ",0) == string::npos))
-          {
-            newList.setEntry(pname,List.entry(param));
-          }
-      } // for
-
-    // Copy of the sublist "coarse: list" to the new list. Change "coarse:" to "smoother:" along the way.
-    if (List.isSublist("coarse: list")) {
-      const ParameterList &coarseList = List.sublist("coarse: list");
-      ParameterList &newCoarseList = newList.sublist("coarse: list");
-      for (ParameterList::ConstIterator param=coarseList.begin(); param!=coarseList.end() ; ++param) {
-        const string & pname=coarseList.name(param);
-
-        if (pname.find("coarse:",0) == 0) {
-          // change "coarse: " to "smoother:"
-          newCoarseList.setEntry("smoother: "+pname.substr(8),coarseList.entry(param));
-        } else {
-          newCoarseList.setEntry(pname,coarseList.entry(param));
+        typedef Teuchos::ArrayRCP<char>::size_type size_type;  // (!)
+        Teuchos::Array<char> ctype(size_type(pname.size() + 1));
+        Teuchos::Array<char> coption(size_type(pname.size() + 1));
+
+        int matched = sscanf(pname.c_str(), "%s %[^(](level %d)", ctype.getRawPtr(), coption.getRawPtr(), &levelID);  // use [^(] instead of %s to allow for strings with white-spaces (ex: "ifpack list")
+        type        = string(ctype.getRawPtr());
+        option      = string(coption.getRawPtr());
+        option.resize(option.size() - 1);  // remove final white-space
+
+        if (matched != 3 || (type != "smoother:" && type != "aggregation:")) {
+          TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::CreateSublist(), Line " << __LINE__ << ". "
+                                                                                                            << "Error in creating level-specific sublists" << std::endl
+                                                                                                            << "Offending parameter: " << pname << std::endl);
         }
       }
-    } // if
 
-    // Copy of level-specific parameters and coarse parameters to sublist
-    for (ParameterList::ConstIterator param=List.begin(); param!=List.end(); ++param)
-      {
-        const string & pname=List.name(param);
-        if (pname.find(" (level",0) != string::npos && pname.find("smoother: list (level",0) != 0 && pname.find("aggregation: list (level",0) != 0)
-          {
-            // Copy level-specific parameters (smoother and aggregation)
-
-            // Scan pname (ex: pname="smoother: type (level 2)")
-            string type, option;
-            int levelID=-1;
-            {
-              typedef Teuchos::ArrayRCP<char>::size_type size_type;    // (!)
-              Teuchos::Array<char> ctype  (size_type(pname.size()+1));
-              Teuchos::Array<char> coption(size_type(pname.size()+1));
-
-              int matched = sscanf(pname.c_str(),"%s %[^(](level %d)", ctype.getRawPtr(), coption.getRawPtr(), &levelID); // use [^(] instead of %s to allow for strings with white-spaces (ex: "ifpack list")
-              type = string(ctype.getRawPtr());
-              option = string(coption.getRawPtr()); option.resize(option.size () - 1); // remove final white-space
-
-              if (matched != 3 || (type != "smoother:" && type != "aggregation:")) {
-                TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::CreateSublist(), Line " << __LINE__ << ". "
-                                           << "Error in creating level-specific sublists" << std::endl
-                                           << "Offending parameter: " << pname << std::endl);
-              }
-            }
-
-            // Create/grab the corresponding sublist of newList
-            ParameterList &newSubList = newList.sublist(type + " list (level " + Teuchos::toString(levelID) + ")");
-            // Shove option w/o level number into sublist
-            newSubList.setEntry(type + " " + option,List.entry(param));
-
-          } else if (pname.find("coarse:",0) == 0 && pname != "coarse: list") {
-          // Copy coarse parameters
-          ParameterList &newCoarseList = newList.sublist("coarse: list"); // the coarse sublist is created only if there is at least one "coarse:" parameter
-          newCoarseList.setEntry("smoother: "+pname.substr(8),List.entry(param)); // change "coarse: " to "smoother:"
-        } // end if
-
-      } // for
-
-  } //MueLu::CreateSublist()
-
-  // Usage: GetMLSubList(paramList, "smoother", 2);
-  const Teuchos::ParameterList & GetMLSubList(const Teuchos::ParameterList & paramList, const std::string & type, int levelID) {
-    static const Teuchos::ParameterList emptyParamList;
-
-    char levelChar[11];
-    sprintf(levelChar, "(level %d)", levelID);
-    std::string levelStr(levelChar);
-
-    if (paramList.isSublist(type + ": list " + levelStr)) {
-      return paramList.sublist(type + ": list " + levelStr);
-    } else {
-      return emptyParamList;
-    }
-  }
+      // Create/grab the corresponding sublist of newList
+      ParameterList &newSubList = newList.sublist(type + " list (level " + Teuchos::toString(levelID) + ")");
+      // Shove option w/o level number into sublist
+      newSubList.setEntry(type + " " + option, List.entry(param));
 
-  // Extract all the parameters that begin with "str:" (but skip sublist)
-  Teuchos::RCP<Teuchos::ParameterList> ExtractSetOfParameters(const Teuchos::ParameterList & paramList, const std::string & str) {
-    Teuchos::RCP<Teuchos::ParameterList> subList = rcp(new Teuchos::ParameterList());
+    } else if (pname.find("coarse:", 0) == 0 && pname != "coarse: list") {
+      // Copy coarse parameters
+      ParameterList &newCoarseList = newList.sublist("coarse: list");             // the coarse sublist is created only if there is at least one "coarse:" parameter
+      newCoarseList.setEntry("smoother: " + pname.substr(8), List.entry(param));  // change "coarse: " to "smoother:"
+    }                                                                             // end if
 
-    for (Teuchos::ParameterList::ConstIterator param = paramList.begin(); param != paramList.end(); ++param) {
-      const Teuchos::ParameterEntry & entry = paramList.entry(param);
-      const std::string             & pname = paramList.name(param);
-      if (pname.find(str+":",0) == 0 && !entry.isList()) {
-        subList->setEntry(pname,entry);
-      }
-    }
+  }  // for
+
+}  // MueLu::CreateSublist()
+
+// Usage: GetMLSubList(paramList, "smoother", 2);
+const Teuchos::ParameterList &GetMLSubList(const Teuchos::ParameterList &paramList, const std::string &type, int levelID) {
+  static const Teuchos::ParameterList emptyParamList;
 
-    return subList;
+  char levelChar[11];
+  sprintf(levelChar, "(level %d)", levelID);
+  std::string levelStr(levelChar);
+
+  if (paramList.isSublist(type + ": list " + levelStr)) {
+    return paramList.sublist(type + ": list " + levelStr);
+  } else {
+    return emptyParamList;
   }
+}
+
+// Extract all the parameters that begin with "str:" (but skip sublist)
+Teuchos::RCP<Teuchos::ParameterList> ExtractSetOfParameters(const Teuchos::ParameterList &paramList, const std::string &str) {
+  Teuchos::RCP<Teuchos::ParameterList> subList = rcp(new Teuchos::ParameterList());
 
-  // replace all string occurrences "from" with "to" in "str"
-  void replaceAll(std::string& str, const std::string& from, const std::string& to) {
-    if(from.empty())
-      return;
-    size_t start_pos = 0;
-    while((start_pos = str.find(from, start_pos)) != std::string::npos) {
-      str.replace(start_pos, from.length(), to);
-      start_pos += to.length(); // In case 'to' contains 'from', like replacing 'x' with 'yx'
+  for (Teuchos::ParameterList::ConstIterator param = paramList.begin(); param != paramList.end(); ++param) {
+    const Teuchos::ParameterEntry &entry = paramList.entry(param);
+    const std::string &pname             = paramList.name(param);
+    if (pname.find(str + ":", 0) == 0 && !entry.isList()) {
+      subList->setEntry(pname, entry);
     }
   }
 
-} // namespace MueLu
+  return subList;
+}
+
+// replace all string occurrences "from" with "to" in "str"
+void replaceAll(std::string &str, const std::string &from, const std::string &to) {
+  if (from.empty())
+    return;
+  size_t start_pos = 0;
+  while ((start_pos = str.find(from, start_pos)) != std::string::npos) {
+    str.replace(start_pos, from.length(), to);
+    start_pos += to.length();  // In case 'to' contains 'from', like replacing 'x' with 'yx'
+  }
+}
+
+}  // namespace MueLu
diff --git a/packages/muelu/src/Interface/MueLu_ParameterListUtils.hpp b/packages/muelu/src/Interface/MueLu_ParameterListUtils.hpp
index 9d55b472ad02..e99129d3305d 100644
--- a/packages/muelu/src/Interface/MueLu_ParameterListUtils.hpp
+++ b/packages/muelu/src/Interface/MueLu_ParameterListUtils.hpp
@@ -56,47 +56,46 @@
 
 namespace MueLu {
 
-  /* See also: ML_Epetra::UpdateList */
-  void MergeParameterList(const Teuchos::ParameterList &source, Teuchos::ParameterList &dest, bool overWrite);
+/* See also: ML_Epetra::UpdateList */
+void MergeParameterList(const Teuchos::ParameterList& source, Teuchos::ParameterList& dest, bool overWrite);
 
-  void CreateSublists(const Teuchos::ParameterList &List, Teuchos::ParameterList &newList);
+void CreateSublists(const Teuchos::ParameterList& List, Teuchos::ParameterList& newList);
 
-  // Usage: GetMLSubList(paramList, "smoother", 2);
-  const Teuchos::ParameterList & GetMLSubList(const Teuchos::ParameterList & paramList, const std::string & type, int levelID);
+// Usage: GetMLSubList(paramList, "smoother", 2);
+const Teuchos::ParameterList& GetMLSubList(const Teuchos::ParameterList& paramList, const std::string& type, int levelID);
 
-  // Extract all the parameters that begin with "str:" (but skip sublist)
-  Teuchos::RCP<Teuchos::ParameterList> ExtractSetOfParameters(const Teuchos::ParameterList & paramList, const std::string & str);
+// Extract all the parameters that begin with "str:" (but skip sublist)
+Teuchos::RCP<Teuchos::ParameterList> ExtractSetOfParameters(const Teuchos::ParameterList& paramList, const std::string& str);
 
-  //! replace all string occurrences "from" with "to" in "str"
-  //!
-  //! @param str: input and output string
-  //! @param from: search string
-  //! @param to: replace with "to"
-  void replaceAll(std::string& str, const std::string& from, const std::string& to);
+//! replace all string occurrences "from" with "to" in "str"
+//!
+//! @param str: input and output string
+//! @param from: search string
+//! @param to: replace with "to"
+void replaceAll(std::string& str, const std::string& from, const std::string& to);
 
-  //! templated version to replace placeholder by data in "str"
-  template<typename Type>
-  bool replacePlaceholder(std::string& str, const std::string& placeholder, Type data) {
-    std::stringstream s;
-    s << data;
-    replaceAll(str, placeholder, s.str());
-    return true;
-  }
+//! templated version to replace placeholder by data in "str"
+template <typename Type>
+bool replacePlaceholder(std::string& str, const std::string& placeholder, Type data) {
+  std::stringstream s;
+  s << data;
+  replaceAll(str, placeholder, s.str());
+  return true;
+}
 
-  template<typename Type>
-  bool actionInterpretParameter(Teuchos::ParameterList& mlParams, const std::string& paramName, std::string& str) {
+template <typename Type>
+bool actionInterpretParameter(Teuchos::ParameterList& mlParams, const std::string& paramName, std::string& str) {
+  // MUELU_READ_PARAM(mlParams, paramName, int, 0, data);
 
-    //MUELU_READ_PARAM(mlParams, paramName, int, 0, data);
+  Type varName;  // = defaultValue; // extract from master list
+  if (mlParams.isParameter(paramName)) varName = mlParams.get<Type>(paramName);
 
-    Type varName; // = defaultValue; // extract from master list
-    if (mlParams.isParameter(paramName)) varName = mlParams.get<Type>(paramName);
+  std::stringstream placeholder;
+  placeholder << "$" << paramName << "$";
 
-    std::stringstream placeholder;
-    placeholder << "$" << paramName << "$";
+  return MueLu::replacePlaceholder<Type>(str, placeholder.str(), varName);
+}
 
-    return MueLu::replacePlaceholder<Type>(str, placeholder.str(), varName);
-  }
+}  // namespace MueLu
 
-} // namespace MueLu
-
-#endif // MUELU_PARAMETERLISTUTILS_HPP
+#endif  // MUELU_PARAMETERLISTUTILS_HPP
diff --git a/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_decl.hpp
index a41ee8b9ae26..a126e4f3ec08 100644
--- a/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_decl.hpp
@@ -57,103 +57,97 @@
 #include <Xpetra_Vector_fwd.hpp>
 #include <Xpetra_VectorFactory_fwd.hpp>
 
-
 #include "MueLu_Aggregates_fwd.hpp"
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLu {
 
-  /*!
-    @class AggregateQualityEstimateFactory class.
-    @brief An factory which assigns each aggregate a quality
-    estimate. Originally developed by Napov and Notay in the
-    context of plain aggregation, while this quality estimate
-    does not correspond to a robust convergence guarentee (as
-    it does for plain aggregation), we find empirically that
-    it is a good way of discovering poorly constructed aggregates
-    even in the smoothed aggregation context.
-
-    Napov, A., & Notay, Y. (2012). An algebraic multigrid method
-    with guaranteed convergence rate. SIAM journal on scientific
-    computing, 34(2), A1079-A1109.
-  */
-
-  template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class AggregateQualityEstimateFactory : public SingleLevelFactoryBase {
+/*!
+  @class AggregateQualityEstimateFactory class.
+  @brief An factory which assigns each aggregate a quality
+  estimate. Originally developed by Napov and Notay in the
+  context of plain aggregation, while this quality estimate
+  does not correspond to a robust convergence guarentee (as
+  it does for plain aggregation), we find empirically that
+  it is a good way of discovering poorly constructed aggregates
+  even in the smoothed aggregation context.
+
+  Napov, A., & Notay, Y. (2012). An algebraic multigrid method
+  with guaranteed convergence rate. SIAM journal on scientific
+  computing, 34(2), A1079-A1109.
+*/
+
+template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class AggregateQualityEstimateFactory : public SingleLevelFactoryBase {
 #undef MUELU_AGGREGATEQUALITYESTIMATEFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    AggregateQualityEstimateFactory();
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
 
-    //! Destructor.
-    virtual ~AggregateQualityEstimateFactory();
+  //! Constructor.
+  AggregateQualityEstimateFactory();
 
-    //@}
+  //! Destructor.
+  virtual ~AggregateQualityEstimateFactory();
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //! @name Input
-    //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //! @name Input
+  //@{
 
-      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-      will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
-
-    //! @name Build methods.
-    //@{
-
-    //! Build aggregate quality esimates with this factory.
-    void Build(Level & currentLevel) const;
-
-    //@}
+    If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+    will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Utility method to convert aggregate data to a convenient format.
-    //@{
+  //@}
 
-    //! Build aggregate quality esimates with this factory.
-    static void ConvertAggregatesData(RCP<const Aggregates> aggs, ArrayRCP<LO>& aggSortedVertices, ArrayRCP<LO>& aggsToIndices, ArrayRCP<LO>& aggSizes);
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build aggregate quality esimates with this factory.
+  void Build(Level& currentLevel) const;
 
-  private:
+  //@}
 
-    //! @name Internal method for computing aggregate quality.
-    //@{
+  //! @name Utility method to convert aggregate data to a convenient format.
+  //@{
 
-    void ComputeAggregateQualities(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<Xpetra::MultiVector<magnitudeType,LO,GO,Node>> agg_qualities) const;
+  //! Build aggregate quality esimates with this factory.
+  static void ConvertAggregatesData(RCP<const Aggregates> aggs, ArrayRCP<LO>& aggSortedVertices, ArrayRCP<LO>& aggsToIndices, ArrayRCP<LO>& aggSizes);
 
-    void ComputeAggregateSizes(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<LocalOrdinalVector> agg_sizes) const;
+  //@}
 
-    //@}
+ private:
+  //! @name Internal method for computing aggregate quality.
+  //@{
 
-    //! @name Internal method for outputting aggregate quality
-    //@{
+  void ComputeAggregateQualities(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<Xpetra::MultiVector<magnitudeType, LO, GO, Node>> agg_qualities) const;
 
-    void OutputAggQualities(const Level& level, RCP<const Xpetra::MultiVector<magnitudeType,LO,GO,Node>> agg_qualities) const;
+  void ComputeAggregateSizes(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<LocalOrdinalVector> agg_sizes) const;
 
-    
-    void OutputAggSizes(const Level& level, RCP<const LocalOrdinalVector> agg_sizes) const;
+  //@}
 
+  //! @name Internal method for outputting aggregate quality
+  //@{
 
+  void OutputAggQualities(const Level& level, RCP<const Xpetra::MultiVector<magnitudeType, LO, GO, Node>> agg_qualities) const;
 
-    //@}
+  void OutputAggSizes(const Level& level, RCP<const LocalOrdinalVector> agg_sizes) const;
 
+  //@}
 
-  }; // class AggregateQualityEsimateFactory();
+};  // class AggregateQualityEsimateFactory();
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AGGREGATEQUALITYESTIMATEFACTORY_SHORT
-#endif // MUELU_DEMOFACTORY_DECL_HPP
+#endif  // MUELU_DEMOFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_def.hpp b/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_def.hpp
index 44b760ea9ee0..e1117f1b6033 100644
--- a/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_AggregateQualityEstimateFactory_def.hpp
@@ -63,481 +63,448 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AggregateQualityEstimateFactory()
-  { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~AggregateQualityEstimateFactory() {}
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AggregateQualityEstimateFactory() {}
 
-    Input(currentLevel, "A");
-    Input(currentLevel, "Aggregates");
-    Input(currentLevel, "CoarseMap");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~AggregateQualityEstimateFactory() {}
 
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "Aggregates");
+  Input(currentLevel, "CoarseMap");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregate qualities: good aggregate threshold");
-    SET_VALID_ENTRY("aggregate qualities: file output");
-    SET_VALID_ENTRY("aggregate qualities: file base");
-    SET_VALID_ENTRY("aggregate qualities: check symmetry");
-    SET_VALID_ENTRY("aggregate qualities: algorithm");
-    SET_VALID_ENTRY("aggregate qualities: zero threshold");
-    SET_VALID_ENTRY("aggregate qualities: percentiles");
-    SET_VALID_ENTRY("aggregate qualities: mode");
-
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Aggregates",         Teuchos::null, "Generating factory of the aggregates");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap",          Teuchos::null, "Generating factory of the coarse map");
-
-    return validParamList;
-  }
+  SET_VALID_ENTRY("aggregate qualities: good aggregate threshold");
+  SET_VALID_ENTRY("aggregate qualities: file output");
+  SET_VALID_ENTRY("aggregate qualities: file base");
+  SET_VALID_ENTRY("aggregate qualities: check symmetry");
+  SET_VALID_ENTRY("aggregate qualities: algorithm");
+  SET_VALID_ENTRY("aggregate qualities: zero threshold");
+  SET_VALID_ENTRY("aggregate qualities: percentiles");
+  SET_VALID_ENTRY("aggregate qualities: mode");
+
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase>>("Aggregates", Teuchos::null, "Generating factory of the aggregates");
+  validParamList->set<RCP<const FactoryBase>>("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+
+  return validParamList;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & currentLevel) const {
+  RCP<Matrix> A              = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<Aggregates> aggregates = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
 
-    FactoryMonitor m(*this, "Build", currentLevel);
+  RCP<const Map> map = Get<RCP<const Map>>(currentLevel, "CoarseMap");
 
-    RCP<Matrix> A = Get<RCP<Matrix>>(currentLevel, "A");
-    RCP<Aggregates> aggregates = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
+  assert(!aggregates->AggregatesCrossProcessors());
+  ParameterList pL = GetParameterList();
+  std::string mode = pL.get<std::string>("aggregate qualities: mode");
+  GetOStream(Statistics1) << "AggregateQuality: mode " << mode << std::endl;
 
-    RCP<const Map> map = Get< RCP<const Map> >(currentLevel, "CoarseMap");
+  RCP<Xpetra::MultiVector<magnitudeType, LO, GO, NO>> aggregate_qualities;
+  if (mode == "eigenvalue" || mode == "both") {
+    aggregate_qualities = Xpetra::MultiVectorFactory<magnitudeType, LO, GO, NO>::Build(map, 1);
+    ComputeAggregateQualities(A, aggregates, aggregate_qualities);
+    OutputAggQualities(currentLevel, aggregate_qualities);
+  }
+  if (mode == "size" || mode == "both") {
+    RCP<LocalOrdinalVector> aggregate_sizes = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(map);
+    ComputeAggregateSizes(A, aggregates, aggregate_sizes);
+    Set(currentLevel, "AggregateSizes", aggregate_sizes);
+    OutputAggSizes(currentLevel, aggregate_sizes);
+  }
+  Set(currentLevel, "AggregateQualities", aggregate_qualities);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ConvertAggregatesData(RCP<const Aggregates> aggs, ArrayRCP<LO>& aggSortedVertices, ArrayRCP<LO>& aggsToIndices, ArrayRCP<LO>& aggSizes) {
+  // Reorder local aggregate information into a format amenable to computing
+  // per-aggregate quantities. Specifically, we compute a format
+  // similar to compressed sparse row format for sparse matrices in which
+  // we store all the local vertices in a single array in blocks corresponding
+  // to aggregates. (This array is aggSortedVertices.) We then store a second
+  // array (aggsToIndices) whose k-th element stores the index of the first
+  // vertex in aggregate k in the array aggSortedVertices.
 
-    assert(!aggregates->AggregatesCrossProcessors());
-    ParameterList pL = GetParameterList();
-    std::string mode = pL.get<std::string>("aggregate qualities: mode");
-    GetOStream(Statistics1) << "AggregateQuality: mode "<<mode << std::endl;
+  const LO LO_ZERO = Teuchos::OrdinalTraits<LO>::zero();
+  const LO LO_ONE  = Teuchos::OrdinalTraits<LO>::one();
 
-    RCP<Xpetra::MultiVector<magnitudeType,LO,GO,NO>> aggregate_qualities;
-    if(mode == "eigenvalue" || mode == "both") {
-      aggregate_qualities = Xpetra::MultiVectorFactory<magnitudeType,LO,GO,NO>::Build(map, 1);
-      ComputeAggregateQualities(A, aggregates, aggregate_qualities);
-      OutputAggQualities(currentLevel, aggregate_qualities);
-    }
-    if(mode == "size" || mode =="both") {
-      RCP<LocalOrdinalVector> aggregate_sizes = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(map);
-      ComputeAggregateSizes(A,aggregates,aggregate_sizes);
-      Set(currentLevel, "AggregateSizes",aggregate_sizes);
-      OutputAggSizes(currentLevel, aggregate_sizes);
-    }
-      Set(currentLevel, "AggregateQualities", aggregate_qualities);
+  LO numAggs = aggs->GetNumAggregates();
+  aggSizes   = aggs->ComputeAggregateSizesArrayRCP();
 
+  aggsToIndices = ArrayRCP<LO>(numAggs + LO_ONE, LO_ZERO);
 
+  for (LO i = 0; i < numAggs; ++i) {
+    aggsToIndices[i + LO_ONE] = aggsToIndices[i] + aggSizes[i];
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ConvertAggregatesData(RCP<const Aggregates> aggs, ArrayRCP<LO>& aggSortedVertices, ArrayRCP<LO>& aggsToIndices, ArrayRCP<LO>& aggSizes) {
+  const RCP<LOMultiVector> vertex2AggId     = aggs->GetVertex2AggId();
+  const ArrayRCP<const LO> vertex2AggIdData = vertex2AggId->getData(0);
 
-    // Reorder local aggregate information into a format amenable to computing
-    // per-aggregate quantities. Specifically, we compute a format
-    // similar to compressed sparse row format for sparse matrices in which
-    // we store all the local vertices in a single array in blocks corresponding
-    // to aggregates. (This array is aggSortedVertices.) We then store a second
-    // array (aggsToIndices) whose k-th element stores the index of the first
-    // vertex in aggregate k in the array aggSortedVertices.
+  LO numNodes       = vertex2AggId->getLocalLength();
+  aggSortedVertices = ArrayRCP<LO>(numNodes, -LO_ONE);
+  std::vector<LO> vertexInsertionIndexByAgg(numNodes, LO_ZERO);
 
-    const LO LO_ZERO = Teuchos::OrdinalTraits<LO>::zero();
-    const LO LO_ONE = Teuchos::OrdinalTraits<LO>::one();
+  for (LO i = 0; i < numNodes; ++i) {
+    LO aggId = vertex2AggIdData[i];
+    if (aggId < 0 || aggId >= numAggs) continue;
 
-    LO numAggs = aggs->GetNumAggregates();
-    aggSizes = aggs->ComputeAggregateSizesArrayRCP();
+    aggSortedVertices[aggsToIndices[aggId] + vertexInsertionIndexByAgg[aggId]] = i;
+    vertexInsertionIndexByAgg[aggId]++;
+  }
+}
 
-    aggsToIndices = ArrayRCP<LO>(numAggs+LO_ONE,LO_ZERO);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeAggregateQualities(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<Xpetra::MultiVector<magnitudeType, LO, GO, Node>> agg_qualities) const {
+  const SC SCALAR_ONE = Teuchos::ScalarTraits<SC>::one();
+  const SC SCALAR_TWO = SCALAR_ONE + SCALAR_ONE;
+
+  const LO LO_ZERO = Teuchos::OrdinalTraits<LO>::zero();
+  const LO LO_ONE  = Teuchos::OrdinalTraits<LO>::one();
+
+  using MT         = magnitudeType;
+  const MT MT_ZERO = Teuchos::ScalarTraits<MT>::zero();
+  const MT MT_ONE  = Teuchos::ScalarTraits<MT>::one();
+  ParameterList pL = GetParameterList();
+
+  RCP<const Matrix> AT = A;
+
+  // Algorithm check
+  std::string algostr = pL.get<std::string>("aggregate qualities: algorithm");
+  MT zeroThreshold    = Teuchos::as<MT>(pL.get<double>("aggregate qualities: zero threshold"));
+  enum AggAlgo { ALG_FORWARD = 0,
+                 ALG_REVERSE };
+  AggAlgo algo;
+  if (algostr == "forward") {
+    algo = ALG_FORWARD;
+    GetOStream(Statistics1) << "AggregateQuality: Using 'forward' algorithm" << std::endl;
+  } else if (algostr == "reverse") {
+    algo = ALG_REVERSE;
+    GetOStream(Statistics1) << "AggregateQuality: Using 'reverse' algorithm" << std::endl;
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::RuntimeError, "\"algorithm\" must be one of (forward|reverse)");
+  }
 
-    for (LO i=0;i<numAggs;++i) {
-      aggsToIndices[i+LO_ONE] = aggsToIndices[i] + aggSizes[i];
-    }
+  bool check_symmetry = pL.get<bool>("aggregate qualities: check symmetry");
+  if (check_symmetry) {
+    RCP<MultiVector> x = MultiVectorFactory::Build(A->getMap(), 1, false);
+    x->Xpetra_randomize();
 
-    const RCP<LOMultiVector> vertex2AggId = aggs->GetVertex2AggId();
-    const ArrayRCP<const LO> vertex2AggIdData = vertex2AggId->getData(0);
+    RCP<MultiVector> tmp = MultiVectorFactory::Build(A->getMap(), 1, false);
 
-    LO numNodes = vertex2AggId->getLocalLength();
-    aggSortedVertices = ArrayRCP<LO>(numNodes,-LO_ONE);
-    std::vector<LO> vertexInsertionIndexByAgg(numNodes,LO_ZERO);
+    A->apply(*x, *tmp, Teuchos::NO_TRANS);                        // tmp now stores A*x
+    A->apply(*x, *tmp, Teuchos::TRANS, -SCALAR_ONE, SCALAR_ONE);  // tmp now stores A*x - A^T*x
 
-    for (LO i=0;i<numNodes;++i) {
+    Array<magnitudeType> tmp_norm(1);
+    tmp->norm2(tmp_norm());
 
-      LO aggId = vertex2AggIdData[i];
-      if (aggId<0 || aggId>=numAggs) continue;
+    Array<magnitudeType> x_norm(1);
+    tmp->norm2(x_norm());
 
-      aggSortedVertices[aggsToIndices[aggId]+vertexInsertionIndexByAgg[aggId]] = i;
-      vertexInsertionIndexByAgg[aggId]++;
+    if (tmp_norm[0] > 1e-10 * x_norm[0]) {
+      std::string transpose_string = "transpose";
+      RCP<ParameterList> whatever;
+      AT = Utilities::Transpose(*rcp_const_cast<Matrix>(A), true, transpose_string, whatever);
 
+      assert(A->getMap()->isSameAs(*(AT->getMap())));
     }
-
-
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeAggregateQualities(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<Xpetra::MultiVector<magnitudeType,LO,GO,Node>> agg_qualities) const {
+  // Reorder local aggregate information into a format amenable to computing
+  // per-aggregate quantities. Specifically, we compute a format
+  // similar to compressed sparse row format for sparse matrices in which
+  // we store all the local vertices in a single array in blocks corresponding
+  // to aggregates. (This array is aggSortedVertices.) We then store a second
+  // array (aggsToIndices) whose k-th element stores the index of the first
+  // vertex in aggregate k in the array aggSortedVertices.
 
-    const SC SCALAR_ONE = Teuchos::ScalarTraits<SC>::one();
-    const SC SCALAR_TWO = SCALAR_ONE + SCALAR_ONE;
+  ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
+  ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
 
-    const LO LO_ZERO = Teuchos::OrdinalTraits<LO>::zero();
-    const LO LO_ONE = Teuchos::OrdinalTraits<LO>::one();
+  LO numAggs = aggs->GetNumAggregates();
 
-    using MT = magnitudeType;
-    const MT MT_ZERO = Teuchos::ScalarTraits<MT>::zero();
-    const MT MT_ONE = Teuchos::ScalarTraits<MT>::one();
-    ParameterList pL = GetParameterList();
+  // Compute the per-aggregate quality estimate
 
-    RCP<const Matrix> AT = A;
+  typedef Teuchos::SerialDenseMatrix<LO, MT> DenseMatrix;
+  typedef Teuchos::SerialDenseVector<LO, MT> DenseVector;
 
-    // Algorithm check
-    std::string algostr = pL.get<std::string>("aggregate qualities: algorithm");
-    MT zeroThreshold = Teuchos::as<MT>(pL.get<double>("aggregate qualities: zero threshold"));
-    enum AggAlgo {ALG_FORWARD=0, ALG_REVERSE};
-    AggAlgo algo;
-    if(algostr == "forward")      {algo = ALG_FORWARD; GetOStream(Statistics1) << "AggregateQuality: Using 'forward' algorithm" << std::endl;}
-    else if(algostr == "reverse") {algo = ALG_REVERSE; GetOStream(Statistics1) << "AggregateQuality: Using 'reverse' algorithm" << std::endl;}
-    else {
-      TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::RuntimeError, "\"algorithm\" must be one of (forward|reverse)");
-    }
+  ArrayView<const LO> rowIndices;
+  ArrayView<const SC> rowValues;
+  ArrayView<const SC> colValues;
+  Teuchos::LAPACK<LO, MT> myLapack;
 
-    bool check_symmetry = pL.get<bool>("aggregate qualities: check symmetry");
-    if (check_symmetry) {
+  // Iterate over each aggregate to compute the quality estimate
+  for (LO aggId = LO_ZERO; aggId < numAggs; ++aggId) {
+    LO aggSize = aggSizes[aggId];
+    DenseMatrix A_aggPart(aggSize, aggSize, true);
+    DenseVector offDiagonalAbsoluteSums(aggSize, true);
 
-      RCP<MultiVector> x = MultiVectorFactory::Build(A->getMap(), 1, false);
-      x->Xpetra_randomize();
-
-      RCP<MultiVector> tmp = MultiVectorFactory::Build(A->getMap(), 1, false);
-
-      A->apply(*x, *tmp, Teuchos::NO_TRANS); // tmp now stores A*x
-      A->apply(*x, *tmp, Teuchos::TRANS, -SCALAR_ONE, SCALAR_ONE); // tmp now stores A*x - A^T*x
+    // Iterate over each node in the aggregate
+    for (LO idx = LO_ZERO; idx < aggSize; ++idx) {
+      LO nodeId = aggSortedVertices[idx + aggsToIndices[aggId]];
+      A->getLocalRowView(nodeId, rowIndices, rowValues);
+      AT->getLocalRowView(nodeId, rowIndices, colValues);
+
+      // Iterate over each element in the row corresponding to the current node
+      for (LO elem = LO_ZERO; elem < rowIndices.size(); ++elem) {
+        LO nodeId2 = rowIndices[elem];
+        SC val     = (rowValues[elem] + colValues[elem]) / SCALAR_TWO;
+
+        LO idxInAgg = -LO_ONE;  // -1 if element is not in aggregate
+
+        // Check whether the element belongs in the aggregate. If it does
+        // find, its index. Otherwise, add it's value to the off diagonal
+        // sums
+        for (LO idx2 = LO_ZERO; idx2 < aggSize; ++idx2) {
+          if (aggSortedVertices[idx2 + aggsToIndices[aggId]] == nodeId2) {
+            // Element does belong to aggregate
+            idxInAgg = idx2;
+            break;
+          }
+        }
 
-      Array<magnitudeType> tmp_norm(1);
-      tmp->norm2(tmp_norm());
+        if (idxInAgg == -LO_ONE) {  // Element does not belong to aggregate
 
-      Array<magnitudeType> x_norm(1);
-      tmp->norm2(x_norm());
+          offDiagonalAbsoluteSums[idx] += Teuchos::ScalarTraits<SC>::magnitude(val);
 
-      if (tmp_norm[0] > 1e-10*x_norm[0]) {
-        std::string transpose_string = "transpose";
-        RCP<ParameterList> whatever;
-        AT = Utilities::Transpose(*rcp_const_cast<Matrix>(A), true, transpose_string, whatever);
+        } else {  // Element does belong to aggregate
 
-        assert(A->getMap()->isSameAs( *(AT->getMap()) ));
+          A_aggPart(idx, idxInAgg) = Teuchos::ScalarTraits<SC>::real(val);
+        }
       }
-
     }
 
-    // Reorder local aggregate information into a format amenable to computing
-    // per-aggregate quantities. Specifically, we compute a format
-    // similar to compressed sparse row format for sparse matrices in which
-    // we store all the local vertices in a single array in blocks corresponding
-    // to aggregates. (This array is aggSortedVertices.) We then store a second
-    // array (aggsToIndices) whose k-th element stores the index of the first
-    // vertex in aggregate k in the array aggSortedVertices.
-
-    ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
-    ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
-
-    LO numAggs = aggs->GetNumAggregates();
-
-    // Compute the per-aggregate quality estimate
-
-    typedef Teuchos::SerialDenseMatrix<LO,MT> DenseMatrix;
-    typedef Teuchos::SerialDenseVector<LO,MT> DenseVector;
-
-    ArrayView<const LO> rowIndices;
-    ArrayView<const SC> rowValues;
-    ArrayView<const SC> colValues;
-    Teuchos::LAPACK<LO,MT> myLapack;
-
-    // Iterate over each aggregate to compute the quality estimate
-    for (LO aggId=LO_ZERO; aggId<numAggs; ++aggId) {
-
-      LO aggSize = aggSizes[aggId];
-      DenseMatrix A_aggPart(aggSize, aggSize, true);
-      DenseVector offDiagonalAbsoluteSums(aggSize, true);
-
-      // Iterate over each node in the aggregate
-      for (LO idx=LO_ZERO; idx<aggSize; ++idx) {
-
-        LO nodeId = aggSortedVertices[idx+aggsToIndices[aggId]];
-        A->getLocalRowView(nodeId, rowIndices, rowValues);
-        AT->getLocalRowView(nodeId, rowIndices, colValues);
-
-        // Iterate over each element in the row corresponding to the current node
-        for (LO elem=LO_ZERO; elem<rowIndices.size();++elem) {
-
-          LO nodeId2 = rowIndices[elem];
-          SC val = (rowValues[elem]+colValues[elem])/SCALAR_TWO;
-
-          LO idxInAgg = -LO_ONE; // -1 if element is not in aggregate
-
-          // Check whether the element belongs in the aggregate. If it does
-          // find, its index. Otherwise, add it's value to the off diagonal
-          // sums
-          for (LO idx2=LO_ZERO; idx2<aggSize; ++idx2) {
-
-            if (aggSortedVertices[idx2+aggsToIndices[aggId]] == nodeId2) {
-
-              // Element does belong to aggregate
-              idxInAgg = idx2;
-              break;
-
-            }
-
-          }
+    // Construct a diagonal matrix consisting of the diagonal
+    // of A_aggPart
+    DenseMatrix A_aggPartDiagonal(aggSize, aggSize, true);
+    MT diag_sum = MT_ZERO;
+    for (int i = 0; i < aggSize; ++i) {
+      A_aggPartDiagonal(i, i) = Teuchos::ScalarTraits<SC>::real(A_aggPart(i, i));
+      diag_sum += Teuchos::ScalarTraits<SC>::real(A_aggPart(i, i));
+    }
 
-          if (idxInAgg == -LO_ONE) { // Element does not belong to aggregate
+    DenseMatrix ones(aggSize, aggSize, false);
+    ones.putScalar(MT_ONE);
 
-            offDiagonalAbsoluteSums[idx] += Teuchos::ScalarTraits<SC>::magnitude(val);
+    // Compute matrix on top of generalized Rayleigh quotient
+    // topMatrix = A_aggPartDiagonal - A_aggPartDiagonal*ones*A_aggPartDiagonal/diag_sum;
+    DenseMatrix tmp(aggSize, aggSize, false);
+    DenseMatrix topMatrix(A_aggPartDiagonal);
 
-          } else { // Element does belong to aggregate
+    tmp.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, MT_ONE, ones, A_aggPartDiagonal, MT_ZERO);
+    topMatrix.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -MT_ONE / diag_sum, A_aggPartDiagonal, tmp, MT_ONE);
 
-            A_aggPart(idx,idxInAgg) = Teuchos::ScalarTraits<SC>::real(val);
+    // Compute matrix on bottom of generalized Rayleigh quotient
+    DenseMatrix bottomMatrix(A_aggPart);
+    MT matrixNorm = A_aggPart.normInf();
 
-          }
+    // Forward mode: Include a small perturbation to the bottom matrix to make it nonsingular
+    const MT boost = (algo == ALG_FORWARD) ? (-1e4 * Teuchos::ScalarTraits<MT>::eps() * matrixNorm) : MT_ZERO;
 
-        }
+    for (int i = 0; i < aggSize; ++i) {
+      bottomMatrix(i, i) -= offDiagonalAbsoluteSums(i) + boost;
+    }
 
+    // Compute generalized eigenvalues
+    LO sdim, info;
+    DenseVector alpha_real(aggSize, false);
+    DenseVector alpha_imag(aggSize, false);
+    DenseVector beta(aggSize, false);
+
+    DenseVector workArray(14 * (aggSize + 1), false);
+
+    LO(*ptr2func)
+    (MT*, MT*, MT*);
+    ptr2func  = NULL;
+    LO* bwork = NULL;
+    MT* vl    = NULL;
+    MT* vr    = NULL;
+
+    const char compute_flag = 'N';
+    if (algo == ALG_FORWARD) {
+      // Forward: Solve the generalized eigenvalue problem as is
+      myLapack.GGES(compute_flag, compute_flag, compute_flag, ptr2func, aggSize,
+                    topMatrix.values(), aggSize, bottomMatrix.values(), aggSize, &sdim,
+                    alpha_real.values(), alpha_imag.values(), beta.values(), vl, aggSize,
+                    vr, aggSize, workArray.values(), workArray.length(), bwork,
+                    &info);
+      TEUCHOS_ASSERT(info == LO_ZERO);
+
+      MT maxEigenVal = MT_ZERO;
+      for (int i = LO_ZERO; i < aggSize; ++i) {
+        // NOTE: In theory, the eigenvalues should be nearly real
+        // TEUCHOS_ASSERT(fabs(alpha_imag[i]) <= 1e-8*fabs(alpha_real[i])); // Eigenvalues should be nearly real
+        maxEigenVal = std::max(maxEigenVal, alpha_real[i] / beta[i]);
       }
 
-      // Construct a diagonal matrix consisting of the diagonal
-      // of A_aggPart
-      DenseMatrix A_aggPartDiagonal(aggSize, aggSize, true);
-      MT diag_sum = MT_ZERO;
-      for (int i=0;i<aggSize;++i) {
-        A_aggPartDiagonal(i,i) = Teuchos::ScalarTraits<SC>::real(A_aggPart(i,i));
-        diag_sum += Teuchos::ScalarTraits<SC>::real(A_aggPart(i,i));
+      (agg_qualities->getDataNonConst(0))[aggId] = (MT_ONE + MT_ONE) * maxEigenVal;
+    } else {
+      // Reverse: Swap the top and bottom matrices for the generalized eigenvalue problem
+      // This is trickier, since we need to grab the smallest non-zero eigenvalue and invert it.
+      myLapack.GGES(compute_flag, compute_flag, compute_flag, ptr2func, aggSize,
+                    bottomMatrix.values(), aggSize, topMatrix.values(), aggSize, &sdim,
+                    alpha_real.values(), alpha_imag.values(), beta.values(), vl, aggSize,
+                    vr, aggSize, workArray.values(), workArray.length(), bwork,
+                    &info);
+
+      TEUCHOS_ASSERT(info == LO_ZERO);
+
+      MT minEigenVal = MT_ZERO;
+
+      for (int i = LO_ZERO; i < aggSize; ++i) {
+        MT ev = alpha_real[i] / beta[i];
+        if (ev > zeroThreshold) {
+          if (minEigenVal == MT_ZERO)
+            minEigenVal = ev;
+          else
+            minEigenVal = std::min(minEigenVal, ev);
+        }
       }
+      if (minEigenVal == MT_ZERO)
+        (agg_qualities->getDataNonConst(0))[aggId] = Teuchos::ScalarTraits<MT>::rmax();
+      else
+        (agg_qualities->getDataNonConst(0))[aggId] = (MT_ONE + MT_ONE) / minEigenVal;
+    }
+  }  // end aggId loop
+}
 
-      DenseMatrix ones(aggSize, aggSize, false);
-      ones.putScalar(MT_ONE);
-
-      // Compute matrix on top of generalized Rayleigh quotient
-      // topMatrix = A_aggPartDiagonal - A_aggPartDiagonal*ones*A_aggPartDiagonal/diag_sum;
-      DenseMatrix tmp(aggSize, aggSize, false);
-      DenseMatrix topMatrix(A_aggPartDiagonal);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::OutputAggQualities(const Level& level, RCP<const Xpetra::MultiVector<magnitudeType, LO, GO, Node>> agg_qualities) const {
+  ParameterList pL = GetParameterList();
 
-      tmp.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, MT_ONE, ones, A_aggPartDiagonal, MT_ZERO);
-      topMatrix.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -MT_ONE/diag_sum, A_aggPartDiagonal, tmp, MT_ONE);
+  magnitudeType good_agg_thresh = Teuchos::as<magnitudeType>(pL.get<double>("aggregate qualities: good aggregate threshold"));
+  using MT                      = magnitudeType;
 
-      // Compute matrix on bottom of generalized Rayleigh quotient
-      DenseMatrix bottomMatrix(A_aggPart);
-      MT matrixNorm = A_aggPart.normInf();
+  ArrayRCP<const MT> data = agg_qualities->getData(0);
 
-      // Forward mode: Include a small perturbation to the bottom matrix to make it nonsingular
-      const MT boost = (algo == ALG_FORWARD) ? (-1e4*Teuchos::ScalarTraits<MT>::eps()*matrixNorm) : MT_ZERO;
+  LO num_bad_aggs = 0;
+  MT worst_agg    = 0.0;
 
-      for (int i=0;i<aggSize;++i){
-        bottomMatrix(i,i) -= offDiagonalAbsoluteSums(i) + boost;
-      }
+  MT mean_bad_agg  = 0.0;
+  MT mean_good_agg = 0.0;
 
-      // Compute generalized eigenvalues
-      LO sdim, info;
-      DenseVector alpha_real(aggSize, false);
-      DenseVector alpha_imag(aggSize, false);
-      DenseVector beta(aggSize, false);
-
-      DenseVector workArray(14*(aggSize+1), false);
-
-      LO (*ptr2func)(MT*, MT*, MT*);
-      ptr2func = NULL;
-      LO* bwork = NULL;
-      MT* vl = NULL;
-      MT* vr = NULL;
-
-      const char compute_flag ='N';
-      if(algo == ALG_FORWARD) {
-        // Forward: Solve the generalized eigenvalue problem as is
-        myLapack.GGES(compute_flag,compute_flag,compute_flag,ptr2func,aggSize,
-                      topMatrix.values(),aggSize,bottomMatrix.values(),aggSize,&sdim,
-                      alpha_real.values(),alpha_imag.values(),beta.values(),vl,aggSize,
-                      vr,aggSize,workArray.values(),workArray.length(),bwork,
-                      &info);
-        TEUCHOS_ASSERT(info == LO_ZERO);
-
-        MT maxEigenVal = MT_ZERO;
-        for (int i=LO_ZERO;i<aggSize;++i) {
-          // NOTE: In theory, the eigenvalues should be nearly real
-          //TEUCHOS_ASSERT(fabs(alpha_imag[i]) <= 1e-8*fabs(alpha_real[i])); // Eigenvalues should be nearly real
-          maxEigenVal = std::max(maxEigenVal, alpha_real[i]/beta[i]);
-        }
-        
-        (agg_qualities->getDataNonConst(0))[aggId] = (MT_ONE+MT_ONE)*maxEigenVal;
-      }
-      else {
-        // Reverse: Swap the top and bottom matrices for the generalized eigenvalue problem
-        // This is trickier, since we need to grab the smallest non-zero eigenvalue and invert it.
-        myLapack.GGES(compute_flag,compute_flag,compute_flag,ptr2func,aggSize,
-                      bottomMatrix.values(),aggSize,topMatrix.values(),aggSize,&sdim,
-                      alpha_real.values(),alpha_imag.values(),beta.values(),vl,aggSize,
-                      vr,aggSize,workArray.values(),workArray.length(),bwork,
-                      &info);
-
-        TEUCHOS_ASSERT(info == LO_ZERO);
-        
-        MT minEigenVal = MT_ZERO;
-       
-        for (int i=LO_ZERO;i<aggSize;++i) {
-          MT ev = alpha_real[i] / beta[i];
-          if(ev > zeroThreshold) {
-            if (minEigenVal == MT_ZERO) minEigenVal = ev;
-            else minEigenVal = std::min(minEigenVal,ev);
-          }
-        }
-        if(minEigenVal == MT_ZERO)  (agg_qualities->getDataNonConst(0))[aggId] = Teuchos::ScalarTraits<MT>::rmax();
-        else (agg_qualities->getDataNonConst(0))[aggId] = (MT_ONE+MT_ONE) / minEigenVal;
-      }
-    }//end aggId loop
+  for (size_t i = 0; i < agg_qualities->getLocalLength(); ++i) {
+    if (data[i] > good_agg_thresh) {
+      num_bad_aggs++;
+      mean_bad_agg += data[i];
+    } else {
+      mean_good_agg += data[i];
+    }
+    worst_agg = std::max(worst_agg, data[i]);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::OutputAggQualities(const Level& level, RCP<const Xpetra::MultiVector<magnitudeType,LO,GO,Node>> agg_qualities) const {
-
-    ParameterList pL = GetParameterList();
+  if (num_bad_aggs > 0) mean_bad_agg /= num_bad_aggs;
+  mean_good_agg /= agg_qualities->getLocalLength() - num_bad_aggs;
 
-    magnitudeType good_agg_thresh = Teuchos::as<magnitudeType>(pL.get<double>("aggregate qualities: good aggregate threshold"));
-    using MT = magnitudeType;
-
-    ArrayRCP<const MT> data = agg_qualities->getData(0);
-
-    LO num_bad_aggs = 0;
-    MT worst_agg = 0.0;
+  if (num_bad_aggs == 0) {
+    GetOStream(Statistics1) << "All aggregates passed the quality measure. Worst aggregate had quality " << worst_agg << ". Mean aggregate quality " << mean_good_agg << "." << std::endl;
+  } else {
+    GetOStream(Statistics1) << num_bad_aggs << " out of " << agg_qualities->getLocalLength() << " did not pass the quality measure. Worst aggregate had quality " << worst_agg << ". "
+                            << "Mean bad aggregate quality " << mean_bad_agg << ". Mean good aggregate quality " << mean_good_agg << "." << std::endl;
+  }
 
-    MT mean_bad_agg = 0.0;
-    MT mean_good_agg  = 0.0;
+  if (pL.get<bool>("aggregate qualities: file output")) {
+    std::string filename = pL.get<std::string>("aggregate qualities: file base") + "." + std::to_string(level.GetLevelID());
+    Xpetra::IO<magnitudeType, LO, GO, Node>::Write(filename, *agg_qualities);
+  }
 
+  {
+    const auto n = size_t(agg_qualities->getLocalLength());
 
-    for (size_t i=0;i<agg_qualities->getLocalLength();++i) {
+    std::vector<MT> tmp;
+    tmp.reserve(n);
 
-      if (data[i] > good_agg_thresh) {
-        num_bad_aggs++;
-        mean_bad_agg += data[i];
-      }
-      else {
-        mean_good_agg += data[i];
-      }
-      worst_agg = std::max(worst_agg, data[i]);
+    for (size_t i = 0; i < n; ++i) {
+      tmp.push_back(data[i]);
     }
 
+    std::sort(tmp.begin(), tmp.end());
 
-    if (num_bad_aggs > 0) mean_bad_agg /= num_bad_aggs;
-    mean_good_agg /= agg_qualities->getLocalLength() - num_bad_aggs;
+    Teuchos::ArrayView<const double> percents = pL.get<Teuchos::Array<double>>("aggregate qualities: percentiles")();
 
-    if (num_bad_aggs == 0) {
-      GetOStream(Statistics1) << "All aggregates passed the quality measure. Worst aggregate had quality " << worst_agg << ". Mean aggregate quality " << mean_good_agg << "." << std::endl;
-    } else {
-      GetOStream(Statistics1) << num_bad_aggs << " out of " << agg_qualities->getLocalLength() << " did not pass the quality measure. Worst aggregate had quality " << worst_agg << ". "
-                              << "Mean bad aggregate quality " << mean_bad_agg << ". Mean good aggregate quality " << mean_good_agg << "." << std::endl;
+    GetOStream(Statistics1) << "AGG QUALITY HEADER     : | LEVEL |  TOTAL  |";
+    for (auto percent : percents) {
+      GetOStream(Statistics1) << std::fixed << std::setprecision(4) << 100.0 * percent << "% |";
     }
-
-    if (pL.get<bool>("aggregate qualities: file output")) {
-      std::string filename = pL.get<std::string>("aggregate qualities: file base")+"."+std::to_string(level.GetLevelID());
-      Xpetra::IO<magnitudeType,LO,GO,Node>::Write(filename, *agg_qualities);
-    }
-
-    {
-      const auto n = size_t(agg_qualities->getLocalLength());
-
-      std::vector<MT> tmp;
-      tmp.reserve(n);
-
-      for (size_t i=0; i<n; ++i) {
-        tmp.push_back(data[i]);
-      }
-
-      std::sort(tmp.begin(), tmp.end());
-
-      Teuchos::ArrayView<const double> percents = pL.get<Teuchos::Array<double> >("aggregate qualities: percentiles")();
-
-      GetOStream(Statistics1) << "AGG QUALITY HEADER     : | LEVEL |  TOTAL  |";
-      for (auto percent : percents) {
-        GetOStream(Statistics1) << std::fixed << std::setprecision(4) <<100.0*percent << "% |";
-      }
-      GetOStream(Statistics1) << std::endl;
-
-      GetOStream(Statistics1) << "AGG QUALITY PERCENTILES: | " << level.GetLevelID() << " | " << n << "|";
-      for (auto percent : percents) {
-        size_t i = size_t(n*percent);
-        i = i < n ? i : n-1u;
-        i = i > 0u ? i : 0u;
-        GetOStream(Statistics1) << std::fixed <<std::setprecision(4) << tmp[i] << " |";
-      }
-      GetOStream(Statistics1) << std::endl;
-
+    GetOStream(Statistics1) << std::endl;
+
+    GetOStream(Statistics1) << "AGG QUALITY PERCENTILES: | " << level.GetLevelID() << " | " << n << "|";
+    for (auto percent : percents) {
+      size_t i = size_t(n * percent);
+      i        = i < n ? i : n - 1u;
+      i        = i > 0u ? i : 0u;
+      GetOStream(Statistics1) << std::fixed << std::setprecision(4) << tmp[i] << " |";
     }
+    GetOStream(Statistics1) << std::endl;
   }
-
-
+}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeAggregateSizes(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<LocalOrdinalVector> agg_sizes) const {
-
-    ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
-    ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
-
-    // Iterate over each node in the aggregate
-    auto data = agg_sizes->getDataNonConst(0);
-    for (LO i=0; i<(LO)aggSizes.size(); i++)
-      data[i] = aggSizes[i];
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeAggregateSizes(RCP<const Matrix> A, RCP<const Aggregates> aggs, RCP<LocalOrdinalVector> agg_sizes) const {
+  ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
+  ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
+
+  // Iterate over each node in the aggregate
+  auto data = agg_sizes->getDataNonConst(0);
+  for (LO i = 0; i < (LO)aggSizes.size(); i++)
+    data[i] = aggSizes[i];
 }
 
-
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::OutputAggSizes(const Level& level, RCP<const LocalOrdinalVector> agg_sizes) const {
+void AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::OutputAggSizes(const Level& level, RCP<const LocalOrdinalVector> agg_sizes) const {
+  ParameterList pL = GetParameterList();
+  using MT         = magnitudeType;
 
-    ParameterList pL = GetParameterList();
-    using MT = magnitudeType;
+  ArrayRCP<const LO> data = agg_sizes->getData(0);
 
-    ArrayRCP<const LO> data = agg_sizes->getData(0);
-
-
-    if (pL.get<bool>("aggregate qualities: file output")) {
-      std::string filename = pL.get<std::string>("aggregate qualities: file base")+".sizes."+std::to_string(level.GetLevelID());
-      Xpetra::IO<LO,LO,GO,Node>::Write(filename, *agg_sizes);
-    }
+  if (pL.get<bool>("aggregate qualities: file output")) {
+    std::string filename = pL.get<std::string>("aggregate qualities: file base") + ".sizes." + std::to_string(level.GetLevelID());
+    Xpetra::IO<LO, LO, GO, Node>::Write(filename, *agg_sizes);
+  }
 
-    {
-      size_t n = (size_t)agg_sizes->getLocalLength();
+  {
+    size_t n = (size_t)agg_sizes->getLocalLength();
 
-      std::vector<MT> tmp;
-      tmp.reserve(n);
+    std::vector<MT> tmp;
+    tmp.reserve(n);
 
-      for (size_t i=0; i<n; ++i) {
-        tmp.push_back(Teuchos::as<MT>(data[i]));
-      }
+    for (size_t i = 0; i < n; ++i) {
+      tmp.push_back(Teuchos::as<MT>(data[i]));
+    }
 
-      std::sort(tmp.begin(), tmp.end());
+    std::sort(tmp.begin(), tmp.end());
 
-      Teuchos::ArrayView<const double> percents = pL.get<Teuchos::Array<double> >("aggregate qualities: percentiles")();
-
-      GetOStream(Statistics1) << "AGG SIZE    HEADER     : | LEVEL |  TOTAL  |";
-      for (auto percent : percents) {
-        GetOStream(Statistics1) << std::fixed << std::setprecision(4) <<100.0*percent << "% |";
-      }
-      GetOStream(Statistics1) << std::endl;
-
-      GetOStream(Statistics1) << "AGG SIZE    PERCENTILES: | " << level.GetLevelID() << " | " << n << "|";
-      for (auto percent : percents) {
-        size_t i = size_t(n*percent);
-        i = i < n ? i : n-1u;
-        i = i > 0u ? i : 0u;
-        GetOStream(Statistics1) << std::fixed <<std::setprecision(4) << tmp[i] << " |";
-      }
-      GetOStream(Statistics1) << std::endl;
+    Teuchos::ArrayView<const double> percents = pL.get<Teuchos::Array<double>>("aggregate qualities: percentiles")();
 
+    GetOStream(Statistics1) << "AGG SIZE    HEADER     : | LEVEL |  TOTAL  |";
+    for (auto percent : percents) {
+      GetOStream(Statistics1) << std::fixed << std::setprecision(4) << 100.0 * percent << "% |";
+    }
+    GetOStream(Statistics1) << std::endl;
+
+    GetOStream(Statistics1) << "AGG SIZE    PERCENTILES: | " << level.GetLevelID() << " | " << n << "|";
+    for (auto percent : percents) {
+      size_t i = size_t(n * percent);
+      i        = i < n ? i : n - 1u;
+      i        = i > 0u ? i : 0u;
+      GetOStream(Statistics1) << std::fixed << std::setprecision(4) << tmp[i] << " |";
     }
+    GetOStream(Statistics1) << std::endl;
   }
+}
 
+}  // namespace MueLu
 
-
-} // namespace MueLu
-
-#endif // MUELU_AGGREGATEQUALITYESTIMATE_DEF_HPP
+#endif  // MUELU_AGGREGATEQUALITYESTIMATE_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_decl.hpp
index 06c3a5ed6917..cb5ad7d07c98 100644
--- a/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_decl.hpp
@@ -86,73 +86,72 @@ namespace MueLu {
   ----------|--------------|------------
   | Coordinates | BlockedCoordinatesTransferFactory   | coarse level coordinates (unified)
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class BlockedCoordinatesTransferFactory : public TwoLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class BlockedCoordinatesTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
+  /*! @brief Constructor.
 
-       @param vectorName The name of the quantity to be restricted.
-       @param restrictionName The name of the restriction Matrix.
+     @param vectorName The name of the quantity to be restricted.
+     @param restrictionName The name of the restriction Matrix.
 
-       The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
-       <tt>vectorName</tt>.
-    */
-    BlockedCoordinatesTransferFactory() { }
+     The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
+     <tt>vectorName</tt>.
+  */
+  BlockedCoordinatesTransferFactory() {}
 
-    //! Destructor.
-    virtual ~BlockedCoordinatesTransferFactory() { }
+  //! Destructor.
+  virtual ~BlockedCoordinatesTransferFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &finelevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &finelevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level & fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //@{
-    /*! @brief Add (sub) coords factory in the end of list of factories in BlockedCoordinatesTransferFactory.
+  //@{
+  /*! @brief Add (sub) coords factory in the end of list of factories in BlockedCoordinatesTransferFactory.
 
-    */
-    void AddFactory(const RCP<const FactoryBase>& factory);
+  */
+  void AddFactory(const RCP<const FactoryBase> &factory);
 
+  //! Returns number of sub factories.
+  size_t NumFactories() const { return subFactories_.size(); }
 
-    //! Returns number of sub factories.
-    size_t NumFactories() const { return subFactories_.size(); }
+  //@}
+ private:
+  //! list of user-defined sub Factories
+  std::vector<RCP<const FactoryBase> > subFactories_;
 
-    //@}
-  private:
-    //! list of user-defined sub Factories
-    std::vector<RCP<const FactoryBase> > subFactories_;
+};  // class BlockedCoordinatesTransferFactory
 
-  }; // class BlockedCoordinatesTransferFactory
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_SHORT
-#endif // MUELU_BLOCKEDCOORDINATESTRANSFER_FACTORY_DECL_HPP
+#endif  // MUELU_BLOCKEDCOORDINATESTRANSFER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_def.hpp
index e83e92696def..2c8150b4a969 100644
--- a/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_BlockedCoordinatesTransferFactory_def.hpp
@@ -58,110 +58,108 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null, "Factory for coordinates generation");
-    validParamList->set<RCP<const FactoryBase> >("CoarseMap",                    Teuchos::null, "Generating factory of the coarse map");
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for coordinates generation");
+  validParamList->set<RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  Input(coarseLevel, "CoarseMap");
+
+  // Make sure the Level knows I need these sub-Factories
+  const size_t numSubFactories = NumFactories();
+  for (size_t i = 0; i < numSubFactories; i++) {
+    const RCP<const FactoryBase>& myFactory = subFactories_[i];
+    coarseLevel.DeclareInput("Coordinates", myFactory.getRawPtr(), this);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
-    Input(coarseLevel, "CoarseMap");
-
-    // Make sure the Level knows I need these sub-Factories
-    const size_t numSubFactories = NumFactories();
-    for(size_t i=0; i<numSubFactories; i++) {
-      const RCP<const FactoryBase>& myFactory = subFactories_[i];
-      coarseLevel.DeclareInput("Coordinates", myFactory.getRawPtr(), this);
-    }
-
-    // call DeclareInput of all user-given transfer factories
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = subFactories_.begin(); it != subFactories_.end(); ++it)
-      (*it)->CallDeclareInput(coarseLevel);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> dMV;
-    typedef Xpetra::BlockedMultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> dBV;
+  // call DeclareInput of all user-given transfer factories
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = subFactories_.begin(); it != subFactories_.end(); ++it)
+    (*it)->CallDeclareInput(coarseLevel);
+}
 
-    GetOStream(Runtime0) << "Transferring (blocked) coordinates" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* fineLevel */, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-    const size_t numSubFactories = NumFactories();
-    std::vector<RCP<const Map> > subBlockMaps(numSubFactories);
-    std::vector<RCP<dMV> > subBlockCoords(numSubFactories);
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> dMV;
+  typedef Xpetra::BlockedMultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> dBV;
 
-    if (coarseLevel.IsAvailable("Coordinates", this)) {
-      GetOStream(Runtime0) << "Reusing coordinates" << std::endl;
-      return;
-    }
+  GetOStream(Runtime0) << "Transferring (blocked) coordinates" << std::endl;
 
-    // Get components
-    for(size_t i=0; i<numSubFactories; i++) {
-      GetOStream(Runtime1) << "Generating Coordinates for block " << i <<"/"<<numSubFactories <<std::endl;
-      const RCP<const FactoryBase>& myFactory = subFactories_[i];
-      myFactory->CallBuild(coarseLevel);
-      subBlockCoords[i] = coarseLevel.Get<RCP<dMV> >("Coordinates", myFactory.get());
-      subBlockMaps[i]   = subBlockCoords[i]->getMap();
-    }
+  const size_t numSubFactories = NumFactories();
+  std::vector<RCP<const Map> > subBlockMaps(numSubFactories);
+  std::vector<RCP<dMV> > subBlockCoords(numSubFactories);
 
-    // Blocked Map
-    RCP<const BlockedMap> coarseCoordMapBlocked;
+  if (coarseLevel.IsAvailable("Coordinates", this)) {
+    GetOStream(Runtime0) << "Reusing coordinates" << std::endl;
+    return;
+  }
 
-    {
-      // coarseMap is being used to set up the domain map of tentative P, and therefore, the row map of Ac
-      // Therefore, if we amalgamate coarseMap, logical nodes in the coordinates vector would correspond to
-      // logical blocks in the matrix
-      RCP<const BlockedMap> coarseMap = Get< RCP<const BlockedMap> >(coarseLevel, "CoarseMap");
-      bool thyraMode = coarseMap->getThyraMode();
+  // Get components
+  for (size_t i = 0; i < numSubFactories; i++) {
+    GetOStream(Runtime1) << "Generating Coordinates for block " << i << "/" << numSubFactories << std::endl;
+    const RCP<const FactoryBase>& myFactory = subFactories_[i];
+    myFactory->CallBuild(coarseLevel);
+    subBlockCoords[i] = coarseLevel.Get<RCP<dMV> >("Coordinates", myFactory.get());
+    subBlockMaps[i]   = subBlockCoords[i]->getMap();
+  }
 
-      ArrayView<const GO> elementAList = coarseMap->getFullMap()->getLocalElementList();
+  // Blocked Map
+  RCP<const BlockedMap> coarseCoordMapBlocked;
 
-      LO blkSize = 1;
-      if (rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(0, thyraMode)) != Teuchos::null)
-        blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(0, thyraMode))->getFixedBlockSize();
+  {
+    // coarseMap is being used to set up the domain map of tentative P, and therefore, the row map of Ac
+    // Therefore, if we amalgamate coarseMap, logical nodes in the coordinates vector would correspond to
+    // logical blocks in the matrix
+    RCP<const BlockedMap> coarseMap = Get<RCP<const BlockedMap> >(coarseLevel, "CoarseMap");
+    bool thyraMode                  = coarseMap->getThyraMode();
 
-      for(size_t i=1; i<numSubFactories; i++) {
-        LO otherBlkSize = 1;
-        if (rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(i, thyraMode)) != Teuchos::null)
-          otherBlkSize = rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(i, thyraMode))->getFixedBlockSize();
-        TEUCHOS_TEST_FOR_EXCEPTION(otherBlkSize != blkSize, Exceptions::RuntimeError, "BlockedCoordinatesTransferFactory: Subblocks have different Block sizes. This is not yet supported.");
-      }
+    ArrayView<const GO> elementAList = coarseMap->getFullMap()->getLocalElementList();
 
-      GO        indexBase   = coarseMap->getFullMap()->getIndexBase();
-      size_t    numElements = elementAList.size() / blkSize;
-      Array<GO> elementList(numElements);
+    LO blkSize = 1;
+    if (rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(0, thyraMode)) != Teuchos::null)
+      blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(0, thyraMode))->getFixedBlockSize();
 
-      // Amalgamate the map
-      for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
-        elementList[i] = (elementAList[i*blkSize]-indexBase)/blkSize + indexBase;
+    for (size_t i = 1; i < numSubFactories; i++) {
+      LO otherBlkSize = 1;
+      if (rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(i, thyraMode)) != Teuchos::null)
+        otherBlkSize = rcp_dynamic_cast<const StridedMap>(coarseMap->getMap(i, thyraMode))->getFixedBlockSize();
+      TEUCHOS_TEST_FOR_EXCEPTION(otherBlkSize != blkSize, Exceptions::RuntimeError, "BlockedCoordinatesTransferFactory: Subblocks have different Block sizes. This is not yet supported.");
+    }
 
-      RCP<const Map> coarseCoordMap = MapFactory::Build(coarseMap->getFullMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, coarseMap->getFullMap()->getComm());
+    GO indexBase       = coarseMap->getFullMap()->getIndexBase();
+    size_t numElements = elementAList.size() / blkSize;
+    Array<GO> elementList(numElements);
 
-      coarseCoordMapBlocked = rcp(new BlockedMap(coarseCoordMap, subBlockMaps, thyraMode));
-    }
+    // Amalgamate the map
+    for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
+      elementList[i] = (elementAList[i * blkSize] - indexBase) / blkSize + indexBase;
 
-    // Build blocked coordinates vector
-    RCP<dBV> bcoarseCoords = rcp(new dBV(coarseCoordMapBlocked,subBlockCoords));
+    RCP<const Map> coarseCoordMap = MapFactory::Build(coarseMap->getFullMap()->lib(),
+                                                      Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, coarseMap->getFullMap()->getComm());
 
-    // Turn the blocked coordinates vector into an unblocked one
-    RCP<dMV> coarseCoords = bcoarseCoords->Merge();
-    Set<RCP<dMV> >(coarseLevel, "Coordinates", coarseCoords);
+    coarseCoordMapBlocked = rcp(new BlockedMap(coarseCoordMap, subBlockMaps, thyraMode));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactory(const RCP<const FactoryBase>& factory) {
-    subFactories_.push_back(factory);
-  }
+  // Build blocked coordinates vector
+  RCP<dBV> bcoarseCoords = rcp(new dBV(coarseCoordMapBlocked, subBlockCoords));
 
+  // Turn the blocked coordinates vector into an unblocked one
+  RCP<dMV> coarseCoords = bcoarseCoords->Merge();
+  Set<RCP<dMV> >(coarseLevel, "Coordinates", coarseCoords);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactory(const RCP<const FactoryBase>& factory) {
+  subFactories_.push_back(factory);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_BLOCKEDCOORDINATESTRANSFER_FACTORY_DEF_HPP
+#endif  // MUELU_BLOCKEDCOORDINATESTRANSFER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_decl.hpp
index 63449a12d82b..9460a32a623e 100644
--- a/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_decl.hpp
@@ -60,95 +60,94 @@
 #include "MueLu_TwoLevelFactoryBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class BlockedRAPFactory
-    @brief Factory for building coarse matrices.
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class BlockedRAPFactory : public TwoLevelFactoryBase {
+/*!
+  @class BlockedRAPFactory
+  @brief Factory for building coarse matrices.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class BlockedRAPFactory : public TwoLevelFactoryBase {
 #undef MUELU_BLOCKEDRAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    BlockedRAPFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    virtual ~BlockedRAPFactory() = default;
-    //@}
+  BlockedRAPFactory();
 
-    //! @name Input
-    //@{
+  virtual ~BlockedRAPFactory() = default;
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const override;
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    //@}
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const override;
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &fineLevel, Level &coarseLevel) const override;
-    //@}
+  //@}
 
-    //! @name Handling of user-defined transfer factories
-    //@{
+  //! @name Build methods.
+  //@{
+  void Build(Level &fineLevel, Level &coarseLevel) const override;
+  //@}
 
-    //! Indicate that zero entries on the diagonal of Ac shall be repaired (i.e. if A(i,i) == 0.0 set A(i,i) = 1.0)
-    void SetRepairZeroDiagonal(bool const &repair) {
-      repairZeroDiagonals_ = repair;
-      if(repair) checkAc_ = true; // make sure that plausibility check is performed. Otherwise SetRepairZeroDiagonal(true) has no effect.
-    }
+  //! @name Handling of user-defined transfer factories
+  //@{
 
-    //! Indicate that a simple plausibility check shall be done for Ac after building RAP
-    void SetPlausibilityCheck(bool const &check) {
-      checkAc_ = check;
-    }
+  //! Indicate that zero entries on the diagonal of Ac shall be repaired (i.e. if A(i,i) == 0.0 set A(i,i) = 1.0)
+  void SetRepairZeroDiagonal(bool const &repair) {
+    repairZeroDiagonals_ = repair;
+    if (repair) checkAc_ = true;  // make sure that plausibility check is performed. Otherwise SetRepairZeroDiagonal(true) has no effect.
+  }
 
-    //@{
-    /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
+  //! Indicate that a simple plausibility check shall be done for Ac after building RAP
+  void SetPlausibilityCheck(bool const &check) {
+    checkAc_ = check;
+  }
 
-    Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
-    the next coarser level.
-    */
-    void AddTransferFactory(const RCP<const FactoryBase>& factory);
+  //@{
+  /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
 
-    // TODO add a function to remove a specific transfer factory?
+  Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
+  the next coarser level.
+  */
+  void AddTransferFactory(const RCP<const FactoryBase> &factory);
 
-    //! Returns number of transfer factories.
-    size_t NumTransferFactories() const { return transferFacts_.size(); }
+  // TODO add a function to remove a specific transfer factory?
 
-    //@}
+  //! Returns number of transfer factories.
+  size_t NumTransferFactories() const { return transferFacts_.size(); }
 
-  private:
+  //@}
 
-    //! @name internal plausibility check methods
-    //! checks main diagonal entries of (0,0) block. Does not affect entries in (1,1) block!
-    static void CheckMainDiagonal(RCP<BlockedCrsMatrix> & bAc, bool repairZeroDiagonals = false);
+ private:
+  //! @name internal plausibility check methods
+  //! checks main diagonal entries of (0,0) block. Does not affect entries in (1,1) block!
+  static void CheckMainDiagonal(RCP<BlockedCrsMatrix> &bAc, bool repairZeroDiagonals = false);
 
-    //! If true, perform a basic plausibility check on Ac (default = false)
-    //! note, that the repairZeroDiagonals_ flag only is valid for checkAc_ == true
-    bool checkAc_;
+  //! If true, perform a basic plausibility check on Ac (default = false)
+  //! note, that the repairZeroDiagonals_ flag only is valid for checkAc_ == true
+  bool checkAc_;
 
-    //! If true, the CheckMainDiagonal routine automatically repairs zero entries on main diagonal (default = false)
-    //! i.e. if A(i,i) == 0.0 set A(i,i) = 1.0
-    //! note, that the repairZeroDiagonals_ flag only is valid for checkAc_ == true
-    bool repairZeroDiagonals_;
+  //! If true, the CheckMainDiagonal routine automatically repairs zero entries on main diagonal (default = false)
+  //! i.e. if A(i,i) == 0.0 set A(i,i) = 1.0
+  //! note, that the repairZeroDiagonals_ flag only is valid for checkAc_ == true
+  bool repairZeroDiagonals_;
 
-    //@{
+  //@{
 
-    //! list of user-defined transfer Factories
-    std::vector<RCP<const FactoryBase> > transferFacts_;
+  //! list of user-defined transfer Factories
+  std::vector<RCP<const FactoryBase> > transferFacts_;
 
-    //@}
+  //@}
 
-  }; //class BlockedRAPFactory
+};  // class BlockedRAPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDRAPFACTORY_SHORT
-#endif // MUELU_BLOCKEDRAPFACTORY_DECL_HPP
+#endif  // MUELU_BLOCKEDRAPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_def.hpp b/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_def.hpp
index 7817372296cb..400c00441afd 100644
--- a/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_BlockedRAPFactory_def.hpp
@@ -60,174 +60,169 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedRAPFactory()
-    : checkAc_(false), repairZeroDiagonals_(false)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedRAPFactory()
+  : checkAc_(false)
+  , repairZeroDiagonals_(false) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("transpose: use implicit");
-#undef  SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("P", null, "Prolongator factory");
-    validParamList->set< RCP<const FactoryBase> >("R", null, "Restrictor factory");
+  SET_VALID_ENTRY("transpose: use implicit");
+#undef SET_VALID_ENTRY
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase> >("P", null, "Prolongator factory");
+  validParamList->set<RCP<const FactoryBase> >("R", null, "Restrictor factory");
 
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    const Teuchos::ParameterList& pL = GetParameterList();
-    if (pL.get<bool>("transpose: use implicit") == false)
-      Input(coarseLevel, "R");
+  return validParamList;
+}
 
-    Input(fineLevel,   "A");
-    Input(coarseLevel, "P");
-
-    // call DeclareInput of all user-given transfer factories
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it)
-      (*it)->CallDeclareInput(coarseLevel);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  const Teuchos::ParameterList &pL = GetParameterList();
+  if (pL.get<bool>("transpose: use implicit") == false)
+    Input(coarseLevel, "R");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {  //FIXME make fineLevel const!!
-    FactoryMonitor m(*this, "Computing Ac (block)", coarseLevel);
+  Input(fineLevel, "A");
+  Input(coarseLevel, "P");
 
-    const ParameterList& pL = GetParameterList();
+  // call DeclareInput of all user-given transfer factories
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it)
+    (*it)->CallDeclareInput(coarseLevel);
+}
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel,   "A");
-    RCP<Matrix> P = Get< RCP<Matrix> >(coarseLevel, "P");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {  // FIXME make fineLevel const!!
+  FactoryMonitor m(*this, "Computing Ac (block)", coarseLevel);
 
+  const ParameterList &pL = GetParameterList();
 
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A);
-    RCP<BlockedCrsMatrix> bP = rcp_dynamic_cast<BlockedCrsMatrix>(P);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null() || bP.is_null(), Exceptions::BadCast, "Matrices A and P must be of type BlockedCrsMatrix.");
+  RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<Matrix> P = Get<RCP<Matrix> >(coarseLevel, "P");
 
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A);
+  RCP<BlockedCrsMatrix> bP = rcp_dynamic_cast<BlockedCrsMatrix>(P);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null() || bP.is_null(), Exceptions::BadCast, "Matrices A and P must be of type BlockedCrsMatrix.");
 
-    RCP<BlockedCrsMatrix> bAP;
-    RCP<BlockedCrsMatrix> bAc;
-    {
-      SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
+  RCP<BlockedCrsMatrix> bAP;
+  RCP<BlockedCrsMatrix> bAc;
+  {
+    SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
 
-      // Triple matrix product for BlockedCrsMatrixClass
-      TEUCHOS_TEST_FOR_EXCEPTION((bA->Cols() != bP->Rows()), Exceptions::BadCast,
+    // Triple matrix product for BlockedCrsMatrixClass
+    TEUCHOS_TEST_FOR_EXCEPTION((bA->Cols() != bP->Rows()), Exceptions::BadCast,
                                "Block matrix dimensions do not match: "
-                               "A is " << bA->Rows() << "x" << bA->Cols() <<
-                               "P is " << bP->Rows() << "x" << bP->Cols());
+                               "A is "
+                                   << bA->Rows() << "x" << bA->Cols() << "P is " << bP->Rows() << "x" << bP->Cols());
 
-      bAP = MatrixMatrix::TwoMatrixMultiplyBlock(*bA, false, *bP,  false, GetOStream(Statistics2), true, true);
-    }
-
-
-    // If we do not modify matrix later, allow optimization of storage.
-    // This is necessary for new faster Epetra MM kernels.
-    bool doOptimizeStorage = !checkAc_;
+    bAP = MatrixMatrix::TwoMatrixMultiplyBlock(*bA, false, *bP, false, GetOStream(Statistics2), true, true);
+  }
 
-    const bool doTranspose    = true;
-    const bool doFillComplete = true;
-    if (pL.get<bool>("transpose: use implicit") == true) {
-      SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
-      bAc = MatrixMatrix::TwoMatrixMultiplyBlock(*bP,  doTranspose, *bAP, !doTranspose, GetOStream(Statistics2), doFillComplete, doOptimizeStorage);
+  // If we do not modify matrix later, allow optimization of storage.
+  // This is necessary for new faster Epetra MM kernels.
+  bool doOptimizeStorage = !checkAc_;
 
-    } else {
-      RCP<Matrix>           R  = Get< RCP<Matrix> >(coarseLevel, "R");
-      RCP<BlockedCrsMatrix> bR = rcp_dynamic_cast<BlockedCrsMatrix>(R);
-      TEUCHOS_TEST_FOR_EXCEPTION(bR.is_null(), Exceptions::BadCast, "Matrix R must be of type BlockedCrsMatrix.");
+  const bool doTranspose    = true;
+  const bool doFillComplete = true;
+  if (pL.get<bool>("transpose: use implicit") == true) {
+    SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
+    bAc = MatrixMatrix::TwoMatrixMultiplyBlock(*bP, doTranspose, *bAP, !doTranspose, GetOStream(Statistics2), doFillComplete, doOptimizeStorage);
 
-      TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != bR->Cols(), Exceptions::BadCast,
-                                 "Block matrix dimensions do not match: "
-                                 "R is " << bR->Rows() << "x" << bR->Cols() <<
-                                 "A is " << bA->Rows() << "x" << bA->Cols());
+  } else {
+    RCP<Matrix> R            = Get<RCP<Matrix> >(coarseLevel, "R");
+    RCP<BlockedCrsMatrix> bR = rcp_dynamic_cast<BlockedCrsMatrix>(R);
+    TEUCHOS_TEST_FOR_EXCEPTION(bR.is_null(), Exceptions::BadCast, "Matrix R must be of type BlockedCrsMatrix.");
 
-      SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
-      bAc = MatrixMatrix::TwoMatrixMultiplyBlock(*bR, !doTranspose, *bAP, !doTranspose, GetOStream(Statistics2), doFillComplete, doOptimizeStorage);
-    }
+    TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != bR->Cols(), Exceptions::BadCast,
+                               "Block matrix dimensions do not match: "
+                               "R is "
+                                   << bR->Rows() << "x" << bR->Cols() << "A is " << bA->Rows() << "x" << bA->Cols());
 
+    SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
+    bAc = MatrixMatrix::TwoMatrixMultiplyBlock(*bR, !doTranspose, *bAP, !doTranspose, GetOStream(Statistics2), doFillComplete, doOptimizeStorage);
+  }
 
-    if (checkAc_)
-      CheckMainDiagonal(bAc);
+  if (checkAc_)
+    CheckMainDiagonal(bAc);
 
-    GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*bAc, "Ac (blocked)");
+  GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*bAc, "Ac (blocked)");
 
-    Set<RCP <Matrix> >(coarseLevel, "A", bAc);
+  Set<RCP<Matrix> >(coarseLevel, "A", bAc);
 
-    if (transferFacts_.begin() != transferFacts_.end()) {
-      SubFactoryMonitor m1(*this, "Projections", coarseLevel);
+  if (transferFacts_.begin() != transferFacts_.end()) {
+    SubFactoryMonitor m1(*this, "Projections", coarseLevel);
 
-      // call Build of all user-given transfer factories
-      for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
-        RCP<const FactoryBase> fac = *it;
+    // call Build of all user-given transfer factories
+    for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
+      RCP<const FactoryBase> fac = *it;
 
-        GetOStream(Runtime0) << "BlockRAPFactory: call transfer factory: " << fac->description() << std::endl;
+      GetOStream(Runtime0) << "BlockRAPFactory: call transfer factory: " << fac->description() << std::endl;
 
-        fac->CallBuild(coarseLevel);
+      fac->CallBuild(coarseLevel);
 
-        // AP (11/11/13): I am not sure exactly why we need to call Release, but we do need it to get rid
-        // of dangling data for CoordinatesTransferFactory
-        coarseLevel.Release(*fac);
-      }
+      // AP (11/11/13): I am not sure exactly why we need to call Release, but we do need it to get rid
+      // of dangling data for CoordinatesTransferFactory
+      coarseLevel.Release(*fac);
     }
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckMainDiagonal(RCP<BlockedCrsMatrix> &bAc, bool repairZeroDiagonals) {
+  RCP<Matrix> c00  = bAc->getMatrix(0, 0);
+  RCP<Matrix> Aout = MatrixFactory::Build(c00->getRowMap(), c00->getGlobalMaxNumRowEntries());
+
+  RCP<Vector> diagVec = VectorFactory::Build(c00->getRowMap());
+  c00->getLocalDiagCopy(*diagVec);
+  ArrayRCP<SC> diagVal = diagVec->getDataNonConst(0);
+
+  // loop over local rows
+  for (size_t row = 0; row < c00->getLocalNumRows(); row++) {
+    // get global row id
+    GO grid = c00->getRowMap()->getGlobalElement(row);  // global row id
+
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    c00->getLocalRowView(row, indices, vals);
+
+    // just copy all values in output
+    ArrayRCP<GO> indout(indices.size(), Teuchos::OrdinalTraits<GO>::zero());
+    ArrayRCP<SC> valout(indices.size(), Teuchos::ScalarTraits<SC>::zero());
+
+    // just copy values
+    for (size_t i = 0; i < as<size_t>(indices.size()); i++) {
+      GO gcid   = c00->getColMap()->getGlobalElement(indices[i]);  // LID -> GID (column)
+      indout[i] = gcid;
+      valout[i] = vals[i];
+    }
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckMainDiagonal(RCP<BlockedCrsMatrix> & bAc, bool repairZeroDiagonals) {
-    RCP<Matrix> c00 = bAc->getMatrix(0, 0);
-    RCP<Matrix> Aout = MatrixFactory::Build(c00->getRowMap(), c00->getGlobalMaxNumRowEntries());
-
-    RCP<Vector> diagVec = VectorFactory::Build(c00->getRowMap());
-    c00->getLocalDiagCopy(*diagVec);
-    ArrayRCP<SC> diagVal = diagVec->getDataNonConst(0);
-
-    // loop over local rows
-    for (size_t row = 0; row < c00->getLocalNumRows(); row++) {
-      // get global row id
-      GO grid = c00->getRowMap()->getGlobalElement(row); // global row id
-
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      c00->getLocalRowView(row, indices, vals);
-
-      // just copy all values in output
-      ArrayRCP<GO> indout(indices.size(), Teuchos::OrdinalTraits<GO>::zero());
-      ArrayRCP<SC> valout(indices.size(), Teuchos::ScalarTraits<SC>::zero());
-
-      // just copy values
-      for (size_t i = 0; i < as<size_t>(indices.size()); i++) {
-        GO gcid = c00->getColMap()->getGlobalElement(indices[i]); // LID -> GID (column)
-        indout [i] = gcid;
-        valout [i] = vals[i];
-      }
-
-      Aout->insertGlobalValues(grid, indout.view(0, indout.size()), valout.view(0, valout.size()));
-      if (diagVal[row] == Teuchos::ScalarTraits<SC>::zero() && repairZeroDiagonals) {
-        // always overwrite diagonal entry
-        Aout->insertGlobalValues(grid, Teuchos::tuple<GO>(grid), Teuchos::tuple<SC>(1.0));
-      }
+    Aout->insertGlobalValues(grid, indout.view(0, indout.size()), valout.view(0, valout.size()));
+    if (diagVal[row] == Teuchos::ScalarTraits<SC>::zero() && repairZeroDiagonals) {
+      // always overwrite diagonal entry
+      Aout->insertGlobalValues(grid, Teuchos::tuple<GO>(grid), Teuchos::tuple<SC>(1.0));
     }
+  }
 
-    Aout->fillComplete(c00->getDomainMap(), c00->getRangeMap());
+  Aout->fillComplete(c00->getDomainMap(), c00->getRangeMap());
 
-    bAc->setMatrix(0, 0, Aout);
-  }
+  bAc->setMatrix(0, 0, Aout);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-      "Transfer factory is not derived from TwoLevelFactoryBase. This is very strange. "
-      "(Note: you can remove this exception if there's a good reason for)");
-    transferFacts_.push_back(factory);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase> &factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "Transfer factory is not derived from TwoLevelFactoryBase. This is very strange. "
+                             "(Note: you can remove this exception if there's a good reason for)");
+  transferFacts_.push_back(factory);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDRAPFACTORY_SHORT
-#endif // MUELU_BLOCKEDRAPFACTORY_DEF_HPP
+#endif  // MUELU_BLOCKEDRAPFACTORY_DEF_HPP
 
 // TODO add plausibility check
 // TODO add CheckMainDiagonal for Blocked operator
diff --git a/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_decl.hpp
index 2ee51c482c54..d9046da451e6 100644
--- a/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_decl.hpp
@@ -94,68 +94,67 @@ namespace MueLu {
   ----------|--------------|------------
   | Coordinates | CoordinatesTransferFactory   | coarse level coordinates
 */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class CoordinatesTransferFactory : public TwoLevelFactoryBase {
-    public:
-    typedef Scalar                                              scalar_type;
-    typedef LocalOrdinal                                        local_ordinal_type;
-    typedef GlobalOrdinal                                       global_ordinal_type;
-    typedef typename Node::device_type                          DeviceType;
-    typedef typename DeviceType::execution_space                execution_space;
-
-  private:
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CoordinatesTransferFactory : public TwoLevelFactoryBase {
+ public:
+  typedef Scalar scalar_type;
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename Node::device_type DeviceType;
+  typedef typename DeviceType::execution_space execution_space;
+
+ private:
 #undef MUELU_COORDINATESTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
+  /*! @brief Constructor.
 
-       @param vectorName The name of the quantity to be restricted.
-       @param restrictionName The name of the restriction Matrix.
+     @param vectorName The name of the quantity to be restricted.
+     @param restrictionName The name of the restriction Matrix.
 
-       The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
-       <tt>vectorName</tt>.
-    */
-    CoordinatesTransferFactory() { }
+     The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
+     <tt>vectorName</tt>.
+  */
+  CoordinatesTransferFactory() {}
 
-    //! Destructor.
-    virtual ~CoordinatesTransferFactory() { }
+  //! Destructor.
+  virtual ~CoordinatesTransferFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &finelevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &finelevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level & fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
+ private:
+};  // class CoordinatesTransferFactory
 
-  }; // class CoordinatesTransferFactory
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COORDINATESTRANSFERFACTORY_SHORT
-#endif // MUELU_COORDINATESTRANSFER_FACTORY_DECL_HPP
+#endif  // MUELU_COORDINATESTRANSFER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_def.hpp
index b8f0efbbdb2f..dc6bd4c6c3ec 100644
--- a/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_CoordinatesTransferFactory_def.hpp
@@ -60,227 +60,224 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null, "Factory for coordinates generation");
-    validParamList->set<RCP<const FactoryBase> >("Aggregates",                   Teuchos::null, "Factory for coordinates generation");
-    validParamList->set<RCP<const FactoryBase> >("CoarseMap",                    Teuchos::null, "Generating factory of the coarse map");
-    validParamList->set<bool>                   ("structured aggregation",       false, "Flag specifying that the geometric data is transferred for StructuredAggregationFactory");
-    validParamList->set<bool>                   ("aggregation coupled",          false, "Flag specifying if the aggregation algorithm was used in coupled mode.");
-    validParamList->set<bool>                   ("Geometric",                    false, "Flag specifying that the coordinates are transferred for GeneralGeometricPFactory");
-    validParamList->set<RCP<const FactoryBase> >("coarseCoordinates",            Teuchos::null, "Factory for coarse coordinates generation");
-    validParamList->set<RCP<const FactoryBase> >("gCoarseNodesPerDim",           Teuchos::null, "Factory providing the global number of nodes per spatial dimensions of the mesh");
-    validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim",           Teuchos::null, "Factory providing the local number of nodes per spatial dimensions of the mesh");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions"     ,           Teuchos::null, "Factory providing the number of spatial dimensions of the mesh");
-    validParamList->set<int>                    ("write start",                  -1, "first level at which coordinates should be written to file");
-    validParamList->set<int>                    ("write end",                    -1, "last level at which coordinates should be written to file");
-    validParamList->set<bool>                   ("hybrid aggregation",           false, "Flag specifying that hybrid aggregation data is transfered for HybridAggregationFactory");
-    validParamList->set<RCP<const FactoryBase> >("aggregationRegionTypeCoarse",  Teuchos::null, "Factory indicating what aggregation type is to be used on the coarse level of the region");
-    validParamList->set<bool>                   ("interface aggregation",        false, "Flag specifying that interface aggregation data is transfered for HybridAggregationFactory");
-    validParamList->set<RCP<const FactoryBase> >("coarseInterfacesDimensions",   Teuchos::null, "Factory providing coarseInterfacesDimensions");
-    validParamList->set<RCP<const FactoryBase> >("nodeOnCoarseInterface",        Teuchos::null, "Factory providing nodeOnCoarseInterface");
-
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    static bool isAvailableCoords = false;
-
-    const ParameterList& pL = GetParameterList();
-    if(pL.get<bool>("structured aggregation") == true) {
-      if(pL.get<bool>("aggregation coupled") == true) {
-        Input(fineLevel, "gCoarseNodesPerDim");
-      }
-      Input(fineLevel, "lCoarseNodesPerDim");
-      Input(fineLevel, "numDimensions");
-    } else if(pL.get<bool>("Geometric") == true) {
-      Input(coarseLevel, "coarseCoordinates");
-      Input(coarseLevel, "gCoarseNodesPerDim");
-      Input(coarseLevel, "lCoarseNodesPerDim");
-    } else if(pL.get<bool>("hybrid aggregation") == true) {
-      Input(fineLevel, "aggregationRegionTypeCoarse");
-      Input(fineLevel, "lCoarseNodesPerDim");
-      Input(fineLevel, "numDimensions");
-      if(pL.get<bool>("interface aggregation") == true) {
-        Input(fineLevel, "coarseInterfacesDimensions");
-        Input(fineLevel, "nodeOnCoarseInterface");
-      }
-    } else {
-      if (coarseLevel.GetRequestMode() == Level::REQUEST)
-        isAvailableCoords = coarseLevel.IsAvailable("Coordinates", this);
-
-      if (isAvailableCoords == false) {
-        Input(fineLevel, "Coordinates");
-        Input(fineLevel, "Aggregates");
-        Input(fineLevel, "CoarseMap");
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for coordinates generation");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Factory for coordinates generation");
+  validParamList->set<RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+  validParamList->set<bool>("structured aggregation", false, "Flag specifying that the geometric data is transferred for StructuredAggregationFactory");
+  validParamList->set<bool>("aggregation coupled", false, "Flag specifying if the aggregation algorithm was used in coupled mode.");
+  validParamList->set<bool>("Geometric", false, "Flag specifying that the coordinates are transferred for GeneralGeometricPFactory");
+  validParamList->set<RCP<const FactoryBase> >("coarseCoordinates", Teuchos::null, "Factory for coarse coordinates generation");
+  validParamList->set<RCP<const FactoryBase> >("gCoarseNodesPerDim", Teuchos::null, "Factory providing the global number of nodes per spatial dimensions of the mesh");
+  validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim", Teuchos::null, "Factory providing the local number of nodes per spatial dimensions of the mesh");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null, "Factory providing the number of spatial dimensions of the mesh");
+  validParamList->set<int>("write start", -1, "first level at which coordinates should be written to file");
+  validParamList->set<int>("write end", -1, "last level at which coordinates should be written to file");
+  validParamList->set<bool>("hybrid aggregation", false, "Flag specifying that hybrid aggregation data is transfered for HybridAggregationFactory");
+  validParamList->set<RCP<const FactoryBase> >("aggregationRegionTypeCoarse", Teuchos::null, "Factory indicating what aggregation type is to be used on the coarse level of the region");
+  validParamList->set<bool>("interface aggregation", false, "Flag specifying that interface aggregation data is transfered for HybridAggregationFactory");
+  validParamList->set<RCP<const FactoryBase> >("coarseInterfacesDimensions", Teuchos::null, "Factory providing coarseInterfacesDimensions");
+  validParamList->set<RCP<const FactoryBase> >("nodeOnCoarseInterface", Teuchos::null, "Factory providing nodeOnCoarseInterface");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  static bool isAvailableCoords = false;
+
+  const ParameterList& pL = GetParameterList();
+  if (pL.get<bool>("structured aggregation") == true) {
+    if (pL.get<bool>("aggregation coupled") == true) {
+      Input(fineLevel, "gCoarseNodesPerDim");
+    }
+    Input(fineLevel, "lCoarseNodesPerDim");
+    Input(fineLevel, "numDimensions");
+  } else if (pL.get<bool>("Geometric") == true) {
+    Input(coarseLevel, "coarseCoordinates");
+    Input(coarseLevel, "gCoarseNodesPerDim");
+    Input(coarseLevel, "lCoarseNodesPerDim");
+  } else if (pL.get<bool>("hybrid aggregation") == true) {
+    Input(fineLevel, "aggregationRegionTypeCoarse");
+    Input(fineLevel, "lCoarseNodesPerDim");
+    Input(fineLevel, "numDimensions");
+    if (pL.get<bool>("interface aggregation") == true) {
+      Input(fineLevel, "coarseInterfacesDimensions");
+      Input(fineLevel, "nodeOnCoarseInterface");
+    }
+  } else {
+    if (coarseLevel.GetRequestMode() == Level::REQUEST)
+      isAvailableCoords = coarseLevel.IsAvailable("Coordinates", this);
+
+    if (isAvailableCoords == false) {
+      Input(fineLevel, "Coordinates");
+      Input(fineLevel, "Aggregates");
+      Input(fineLevel, "CoarseMap");
     }
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-    using xdMV = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>;
+  using xdMV = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>;
 
-    GetOStream(Runtime0) << "Transferring coordinates" << std::endl;
+  GetOStream(Runtime0) << "Transferring coordinates" << std::endl;
 
-    int numDimensions;
-    RCP<xdMV> coarseCoords;
-    RCP<xdMV> fineCoords;
-    Array<GO> gCoarseNodesPerDir;
-    Array<LO> lCoarseNodesPerDir;
+  int numDimensions;
+  RCP<xdMV> coarseCoords;
+  RCP<xdMV> fineCoords;
+  Array<GO> gCoarseNodesPerDir;
+  Array<LO> lCoarseNodesPerDir;
 
-    const ParameterList& pL = GetParameterList();
+  const ParameterList& pL = GetParameterList();
 
-    if(pL.get<bool>("hybrid aggregation") == true) {
-      std::string regionType = Get<std::string>(fineLevel,"aggregationRegionTypeCoarse");
-      numDimensions          = Get<int>(fineLevel, "numDimensions");
-      lCoarseNodesPerDir     = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
-      Set<std::string>(coarseLevel, "aggregationRegionType", regionType);
-      Set<int>        (coarseLevel, "numDimensions",         numDimensions);
-      Set<Array<LO> > (coarseLevel, "lNodesPerDim",          lCoarseNodesPerDir);
+  if (pL.get<bool>("hybrid aggregation") == true) {
+    std::string regionType = Get<std::string>(fineLevel, "aggregationRegionTypeCoarse");
+    numDimensions          = Get<int>(fineLevel, "numDimensions");
+    lCoarseNodesPerDir     = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
+    Set<std::string>(coarseLevel, "aggregationRegionType", regionType);
+    Set<int>(coarseLevel, "numDimensions", numDimensions);
+    Set<Array<LO> >(coarseLevel, "lNodesPerDim", lCoarseNodesPerDir);
 
-      if((pL.get<bool>("interface aggregation") == true) && (regionType == "uncoupled")) {
-        Array<LO> coarseInterfacesDimensions = Get<Array<LO> >(fineLevel, "coarseInterfacesDimensions");
-        Array<LO> nodeOnCoarseInterface      = Get<Array<LO> >(fineLevel, "nodeOnCoarseInterface");
-        Set<Array<LO> >(coarseLevel, "interfacesDimensions", coarseInterfacesDimensions);
-        Set<Array<LO> >(coarseLevel, "nodeOnInterface",      nodeOnCoarseInterface);
-      }
+    if ((pL.get<bool>("interface aggregation") == true) && (regionType == "uncoupled")) {
+      Array<LO> coarseInterfacesDimensions = Get<Array<LO> >(fineLevel, "coarseInterfacesDimensions");
+      Array<LO> nodeOnCoarseInterface      = Get<Array<LO> >(fineLevel, "nodeOnCoarseInterface");
+      Set<Array<LO> >(coarseLevel, "interfacesDimensions", coarseInterfacesDimensions);
+      Set<Array<LO> >(coarseLevel, "nodeOnInterface", nodeOnCoarseInterface);
+    }
 
-    } else if(pL.get<bool>("structured aggregation") == true) {
-      if(pL.get<bool>("aggregation coupled") == true) {
-        gCoarseNodesPerDir = Get<Array<GO> >(fineLevel, "gCoarseNodesPerDim");
-        Set<Array<GO> >(coarseLevel, "gNodesPerDim", gCoarseNodesPerDir);
-      }
-      lCoarseNodesPerDir = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
-      Set<Array<LO> >(coarseLevel, "lNodesPerDim", lCoarseNodesPerDir);
-      numDimensions = Get<int>(fineLevel, "numDimensions");
-      Set<int>(coarseLevel, "numDimensions", numDimensions);
-
-    } else if(pL.get<bool>("Geometric") == true) {
-      coarseCoords       = Get<RCP<xdMV> >(coarseLevel, "coarseCoordinates");
-      gCoarseNodesPerDir = Get<Array<GO> >(coarseLevel, "gCoarseNodesPerDim");
-      lCoarseNodesPerDir = Get<Array<LO> >(coarseLevel, "lCoarseNodesPerDim");
+  } else if (pL.get<bool>("structured aggregation") == true) {
+    if (pL.get<bool>("aggregation coupled") == true) {
+      gCoarseNodesPerDir = Get<Array<GO> >(fineLevel, "gCoarseNodesPerDim");
       Set<Array<GO> >(coarseLevel, "gNodesPerDim", gCoarseNodesPerDir);
-      Set<Array<LO> >(coarseLevel, "lNodesPerDim", lCoarseNodesPerDir);
+    }
+    lCoarseNodesPerDir = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
+    Set<Array<LO> >(coarseLevel, "lNodesPerDim", lCoarseNodesPerDir);
+    numDimensions = Get<int>(fineLevel, "numDimensions");
+    Set<int>(coarseLevel, "numDimensions", numDimensions);
+
+  } else if (pL.get<bool>("Geometric") == true) {
+    coarseCoords       = Get<RCP<xdMV> >(coarseLevel, "coarseCoordinates");
+    gCoarseNodesPerDir = Get<Array<GO> >(coarseLevel, "gCoarseNodesPerDim");
+    lCoarseNodesPerDir = Get<Array<LO> >(coarseLevel, "lCoarseNodesPerDim");
+    Set<Array<GO> >(coarseLevel, "gNodesPerDim", gCoarseNodesPerDir);
+    Set<Array<LO> >(coarseLevel, "lNodesPerDim", lCoarseNodesPerDir);
+
+    Set<RCP<xdMV> >(coarseLevel, "Coordinates", coarseCoords);
+
+  } else {
+    if (coarseLevel.IsAvailable("Coordinates", this)) {
+      GetOStream(Runtime0) << "Reusing coordinates" << std::endl;
+      return;
+    }
 
-      Set<RCP<xdMV> >(coarseLevel, "Coordinates", coarseCoords);
+    fineCoords               = Get<RCP<xdMV> >(fineLevel, "Coordinates");
+    RCP<const Map> coarseMap = Get<RCP<const Map> >(fineLevel, "CoarseMap");
+
+    // coarseMap is being used to set up the domain map of tentative P, and therefore, the row map of Ac
+    // Therefore, if we amalgamate coarseMap, logical nodes in the coordinates vector would correspond to
+    // logical blocks in the matrix
+    LO blkSize = 1;
+    if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null)
+      blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
+
+    RCP<const Map> coarseCoordMap;
+    RCP<const Map> uniqueMap = fineCoords->getMap();
+    if (blkSize > 1) {
+      // If the block size is greater than one, we need to create a coarse coordinate map
+      // FIXME: The amalgamation should really be done on device.
+      GO indexBase                     = coarseMap->getIndexBase();
+      ArrayView<const GO> elementAList = coarseMap->getLocalElementList();
+      size_t numElements               = elementAList.size() / blkSize;
+      Array<GO> elementList(numElements);
+
+      // Amalgamate the map
+      for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
+        elementList[i] = (elementAList[i * blkSize] - indexBase) / blkSize + indexBase;
 
-    } else {
-      if (coarseLevel.IsAvailable("Coordinates", this)) {
-        GetOStream(Runtime0) << "Reusing coordinates" << std::endl;
-        return;
-      }
-
-      fineCoords                     = Get< RCP<xdMV> >(fineLevel, "Coordinates");
-      RCP<const Map>      coarseMap  = Get< RCP<const Map> >  (fineLevel, "CoarseMap");
-
-      // coarseMap is being used to set up the domain map of tentative P, and therefore, the row map of Ac
-      // Therefore, if we amalgamate coarseMap, logical nodes in the coordinates vector would correspond to
-      // logical blocks in the matrix
-      LO                  blkSize      = 1;
-      if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null)
-        blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
-
-      RCP<const Map>   coarseCoordMap;
-      RCP<const Map>   uniqueMap      = fineCoords->getMap();
-      if(blkSize > 1) {
-        // If the block size is greater than one, we need to create a coarse coordinate map
-        // FIXME: The amalgamation should really be done on device.
-        GO                  indexBase    = coarseMap->getIndexBase();
-        ArrayView<const GO> elementAList = coarseMap->getLocalElementList();
-        size_t              numElements  = elementAList.size() / blkSize;
-        Array<GO>           elementList(numElements);
-        
-        // Amalgamate the map
-        for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
-          elementList[i] = (elementAList[i*blkSize]-indexBase)/blkSize + indexBase;
-                       
-        {
-          SubFactoryMonitor sfm(*this, "MapFactory: coarseCoordMap", fineLevel);
-          coarseCoordMap = MapFactory        ::Build(coarseMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, coarseMap->getComm());
-        }
-      }
-      else {
-        // If the block size is one, we can just use the coarse map for coordinates
-        coarseCoordMap = coarseMap;
+      {
+        SubFactoryMonitor sfm(*this, "MapFactory: coarseCoordMap", fineLevel);
+        coarseCoordMap = MapFactory ::Build(coarseMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), elementList, indexBase, coarseMap->getComm());
       }
+    } else {
+      // If the block size is one, we can just use the coarse map for coordinates
+      coarseCoordMap = coarseMap;
+    }
 
-      // Build the coarseCoords MultiVector
-      coarseCoords   = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(coarseCoordMap, fineCoords->getNumVectors());
-
+    // Build the coarseCoords MultiVector
+    coarseCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(coarseCoordMap, fineCoords->getNumVectors());
 
-      RCP<Aggregates> aggregates;
-      bool aggregatesCrossProcessors;
-      aggregates = Get<RCP<Aggregates> >(fineLevel, "Aggregates");
-      aggregatesCrossProcessors = aggregates->AggregatesCrossProcessors();
+    RCP<Aggregates> aggregates;
+    bool aggregatesCrossProcessors;
+    aggregates                = Get<RCP<Aggregates> >(fineLevel, "Aggregates");
+    aggregatesCrossProcessors = aggregates->AggregatesCrossProcessors();
 
-      // Create overlapped fine coordinates to reduce global communication
-      RCP<xdMV> ghostedCoords = fineCoords;
-      if (aggregatesCrossProcessors) {
-        RCP<const Map>    nonUniqueMap = aggregates->GetMap();
-        RCP<const Import> importer     = ImportFactory::Build(uniqueMap, nonUniqueMap);
+    // Create overlapped fine coordinates to reduce global communication
+    RCP<xdMV> ghostedCoords = fineCoords;
+    if (aggregatesCrossProcessors) {
+      RCP<const Map> nonUniqueMap = aggregates->GetMap();
+      RCP<const Import> importer  = ImportFactory::Build(uniqueMap, nonUniqueMap);
 
-        ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(nonUniqueMap, fineCoords->getNumVectors());
-        ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
-      }
+      ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(nonUniqueMap, fineCoords->getNumVectors());
+      ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
+    }
 
-      // The good news is that this graph has already been constructed for the
-      // TentativePFactory and was cached in Aggregates. So this is a no-op.
-      auto aggGraph = aggregates->GetGraph();
-      auto numAggs  = aggGraph.numRows();
+    // The good news is that this graph has already been constructed for the
+    // TentativePFactory and was cached in Aggregates. So this is a no-op.
+    auto aggGraph = aggregates->GetGraph();
+    auto numAggs  = aggGraph.numRows();
 
-      auto fineCoordsView   = ghostedCoords->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      auto coarseCoordsView = coarseCoords->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+    auto fineCoordsView   = ghostedCoords->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto coarseCoordsView = coarseCoords->getDeviceLocalView(Xpetra::Access::OverwriteAll);
 
-      // Fill in coarse coordinates
-      {
-        SubFactoryMonitor m2(*this, "AverageCoords", coarseLevel);
+    // Fill in coarse coordinates
+    {
+      SubFactoryMonitor m2(*this, "AverageCoords", coarseLevel);
 
-        const auto dim = ghostedCoords->getNumVectors();
+      const auto dim = ghostedCoords->getNumVectors();
 
-        typename AppendTrait<decltype(fineCoordsView), Kokkos::RandomAccess>::type fineCoordsRandomView = fineCoordsView;
-        for (size_t j = 0; j < dim; j++) {
-          Kokkos::parallel_for("MueLu:CoordinatesTransferF:Build:coord", Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numAggs),
-                                KOKKOS_LAMBDA(const LO i) {
-                                  // A row in this graph represents all node ids in the aggregate
-                                  // Therefore, averaging is very easy
+      typename AppendTrait<decltype(fineCoordsView), Kokkos::RandomAccess>::type fineCoordsRandomView = fineCoordsView;
+      for (size_t j = 0; j < dim; j++) {
+        Kokkos::parallel_for(
+            "MueLu:CoordinatesTransferF:Build:coord", Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numAggs),
+            KOKKOS_LAMBDA(const LO i) {
+              // A row in this graph represents all node ids in the aggregate
+              // Therefore, averaging is very easy
 
-                                  auto aggregate = aggGraph.rowConst(i);
+              auto aggregate = aggGraph.rowConst(i);
 
-                                  typename Teuchos::ScalarTraits<Scalar>::magnitudeType sum = 0.0; // do not use Scalar here (Stokhos)
-                                  for (size_t colID = 0; colID < static_cast<size_t>(aggregate.length); colID++)
-                                    sum += fineCoordsRandomView(aggregate(colID),j);
+              typename Teuchos::ScalarTraits<Scalar>::magnitudeType sum = 0.0;  // do not use Scalar here (Stokhos)
+              for (size_t colID = 0; colID < static_cast<size_t>(aggregate.length); colID++)
+                sum += fineCoordsRandomView(aggregate(colID), j);
 
-                                  coarseCoordsView(i,j) = sum / aggregate.length;
-                                });
-        }
+              coarseCoordsView(i, j) = sum / aggregate.length;
+            });
       }
-
-      Set<RCP<xdMV> >(coarseLevel, "Coordinates", coarseCoords);
-
     }
 
-    int writeStart = pL.get<int>("write start"), writeEnd = pL.get<int>("write end");
-    if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd) {
-      std::ostringstream buf;
-      buf << fineLevel.GetLevelID();
-      std::string fileName = "coordinates_before_rebalance_level_" + buf.str() + ".m";
-      Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Write(fileName,*fineCoords);
-    }
-    if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd) {
-      std::ostringstream buf;
-      buf << coarseLevel.GetLevelID();
-      std::string fileName = "coordinates_before_rebalance_level_" + buf.str() + ".m";
-      Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Write(fileName,*coarseCoords);
-    }
+    Set<RCP<xdMV> >(coarseLevel, "Coordinates", coarseCoords);
+  }
+
+  int writeStart = pL.get<int>("write start"), writeEnd = pL.get<int>("write end");
+  if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd) {
+    std::ostringstream buf;
+    buf << fineLevel.GetLevelID();
+    std::string fileName = "coordinates_before_rebalance_level_" + buf.str() + ".m";
+    Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Write(fileName, *fineCoords);
+  }
+  if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd) {
+    std::ostringstream buf;
+    buf << coarseLevel.GetLevelID();
+    std::string fileName = "coordinates_before_rebalance_level_" + buf.str() + ".m";
+    Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Write(fileName, *coarseCoords);
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_COORDINATESTRANSFER_FACTORY_DEF_HPP
+#endif  // MUELU_COORDINATESTRANSFER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_DemoFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_DemoFactory_decl.hpp
index 75931f90de75..b152ddba9d32 100644
--- a/packages/muelu/src/Misc/MueLu_DemoFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_DemoFactory_decl.hpp
@@ -52,57 +52,57 @@
 
 namespace MueLu {
 
-  /*!
-    @class DemoFactory class.
-    @brief empty factory for demonstration
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class DemoFactory : public SingleLevelFactoryBase {
+/*!
+  @class DemoFactory class.
+  @brief empty factory for demonstration
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class DemoFactory : public SingleLevelFactoryBase {
 #undef MUELU_DEMOFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    DemoFactory();
+  //! Constructor.
+  DemoFactory();
 
-    //! Destructor.
-    virtual ~DemoFactory();
+  //! Destructor.
+  virtual ~DemoFactory();
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level & currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    // TODO add member variables
+ private:
+  // TODO add member variables
 
-  }; // class DemoFactory
+};  // class DemoFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DEMOFACTORY_SHORT
-#endif // MUELU_DEMOFACTORY_DECL_HPP
+#endif  // MUELU_DEMOFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_DemoFactory_def.hpp b/packages/muelu/src/Misc/MueLu_DemoFactory_def.hpp
index b390142ca2f8..cc7a327b79f9 100644
--- a/packages/muelu/src/Misc/MueLu_DemoFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_DemoFactory_def.hpp
@@ -55,25 +55,24 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DemoFactory()
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DemoFactory() {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~DemoFactory() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~DemoFactory() {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &/* currentLevel */) const {
-    // TODO: declare input for factory
-    //Input(currentLevel, varName_);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* currentLevel */) const {
+  // TODO: declare input for factory
+  // Input(currentLevel, varName_);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* currentLevel */) const {
-    // TODO: implement factory
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DemoFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* currentLevel */) const {
+  // TODO: implement factory
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DEMOFACTORY_SHORT
-#endif // MUELU_DEMOFACTORY_DEF_HPP
+#endif  // MUELU_DEMOFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_decl.hpp
index ab99f2f9b964..82a467269b53 100644
--- a/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_decl.hpp
@@ -55,58 +55,57 @@
 
 namespace MueLu {
 
-  /*!
-    @class DropNegativeEntriesFactory class.
-    @brief Application-specific filtering for A. Can be used in context of graph coarsening and aggregation.
-
-    This factory drops all negative entries (or entries with a magnitude < 0). Only weak positive connections are kept.
-    Do not use this kind of filtering for regular PDEs unless you have very good reasons.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class DropNegativeEntriesFactory : public SingleLevelFactoryBase {
+/*!
+  @class DropNegativeEntriesFactory class.
+  @brief Application-specific filtering for A. Can be used in context of graph coarsening and aggregation.
+
+  This factory drops all negative entries (or entries with a magnitude < 0). Only weak positive connections are kept.
+  Do not use this kind of filtering for regular PDEs unless you have very good reasons.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class DropNegativeEntriesFactory : public SingleLevelFactoryBase {
 #undef MUELU_DROPNEGATIVEENTRIESFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    DropNegativeEntriesFactory() { }
+  DropNegativeEntriesFactory() {}
 
-    //! Destructor.
-    virtual ~DropNegativeEntriesFactory() { }
+  //! Destructor.
+  virtual ~DropNegativeEntriesFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Builds filtered matrix and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+    Builds filtered matrix and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  }; //class DropNegativeEntriesFactory
+};  // class DropNegativeEntriesFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DROPNEGATIVEENTRIESFACTORY_SHORT
-#endif // MUELU_DROPNEGATIVEENTRIESFACTORY_DECL_HPP
+#endif  // MUELU_DROPNEGATIVEENTRIESFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_def.hpp b/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_def.hpp
index 1bf7b0bfd4c8..4e4882ad5a0d 100644
--- a/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_DropNegativeEntriesFactory_def.hpp
@@ -60,76 +60,76 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-#undef  SET_VALID_ENTRY
+#undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used for filtering");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used for filtering");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Matrix filtering (springs)", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DropNegativeEntriesFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Matrix filtering (springs)", currentLevel);
 
-    RCP<Matrix> Ain = Get< RCP<Matrix> >(currentLevel, "A");
+  RCP<Matrix> Ain = Get<RCP<Matrix> >(currentLevel, "A");
 
-    LocalOrdinal nDofsPerNode = Ain->GetFixedBlockSize();
+  LocalOrdinal nDofsPerNode = Ain->GetFixedBlockSize();
 
-    // create new empty Operator
-    Teuchos::RCP<Matrix> Aout = MatrixFactory::Build(Ain->getRowMap(), Ain->getGlobalMaxNumRowEntries());
+  // create new empty Operator
+  Teuchos::RCP<Matrix> Aout = MatrixFactory::Build(Ain->getRowMap(), Ain->getGlobalMaxNumRowEntries());
 
-    size_t numLocalRows = Ain->getLocalNumRows();
-    for(size_t row=0; row<numLocalRows; row++) {
-      GlobalOrdinal grid = Ain->getRowMap()->getGlobalElement(row);
+  size_t numLocalRows = Ain->getLocalNumRows();
+  for (size_t row = 0; row < numLocalRows; row++) {
+    GlobalOrdinal grid = Ain->getRowMap()->getGlobalElement(row);
 
-      int rDofID = Teuchos::as<int>(grid % nDofsPerNode);
+    int rDofID = Teuchos::as<int>(grid % nDofsPerNode);
 
-      // extract row information from input matrix
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals;
-      Ain->getLocalRowView(row, indices, vals);
+    // extract row information from input matrix
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    Ain->getLocalRowView(row, indices, vals);
 
-      // just copy all values in output
-      Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(),Teuchos::ScalarTraits<GlobalOrdinal>::zero());
-      Teuchos::ArrayRCP<Scalar>        valout(indices.size(),Teuchos::ScalarTraits<Scalar>::zero());
+    // just copy all values in output
+    Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(), Teuchos::ScalarTraits<GlobalOrdinal>::zero());
+    Teuchos::ArrayRCP<Scalar> valout(indices.size(), Teuchos::ScalarTraits<Scalar>::zero());
 
-      size_t nNonzeros = 0;
-      for(size_t i=0; i<(size_t)indices.size(); i++) {
-        GlobalOrdinal gcid = Ain->getColMap()->getGlobalElement(indices[i]); // global column id
+    size_t nNonzeros = 0;
+    for (size_t i = 0; i < (size_t)indices.size(); i++) {
+      GlobalOrdinal gcid = Ain->getColMap()->getGlobalElement(indices[i]);  // global column id
 
-        int cDofID = Teuchos::as<int>(gcid % nDofsPerNode);
-        if(rDofID == cDofID && Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) >= Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero())) {
-          indout [nNonzeros] = gcid;
-          valout [nNonzeros] = vals[i];
-          nNonzeros++;
-        }
+      int cDofID = Teuchos::as<int>(gcid % nDofsPerNode);
+      if (rDofID == cDofID && Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) >= Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero())) {
+        indout[nNonzeros] = gcid;
+        valout[nNonzeros] = vals[i];
+        nNonzeros++;
       }
-      indout.resize(nNonzeros);
-      valout.resize(nNonzeros);
-
-      Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0,indout.size()), valout.view(0,valout.size()));
     }
+    indout.resize(nNonzeros);
+    valout.resize(nNonzeros);
 
-    Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
+    Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0, indout.size()), valout.view(0, valout.size()));
+  }
 
-    // copy block size information
-    Aout->SetFixedBlockSize(nDofsPerNode);
+  Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
 
-    GetOStream(Statistics0, 0) << "Nonzeros in A (input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering A: " << Aout->getGlobalNumEntries() << std::endl;
+  // copy block size information
+  Aout->SetFixedBlockSize(nDofsPerNode);
 
-    Set(currentLevel, "A", Aout);
-  }
+  GetOStream(Statistics0, 0) << "Nonzeros in A (input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering A: " << Aout->getGlobalNumEntries() << std::endl;
+
+  Set(currentLevel, "A", Aout);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_DROPNEGATIVEENTRIESFACTORY_DEF_HPP
+#endif  // MUELU_DROPNEGATIVEENTRIESFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_FilteredAFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_FilteredAFactory_decl.hpp
index 23f56137596d..bd278f9410d5 100644
--- a/packages/muelu/src/Misc/MueLu_FilteredAFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_FilteredAFactory_decl.hpp
@@ -58,60 +58,59 @@
 #include "MueLu_Aggregates_fwd.hpp"
 namespace MueLu {
 
-  /*!
-    @class FilteredAFactory class.
-    @brief Factory for building filtered matrices using filtered graphs.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FilteredAFactory : public SingleLevelFactoryBase {
+/*!
+  @class FilteredAFactory class.
+  @brief Factory for building filtered matrices using filtered graphs.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FilteredAFactory : public SingleLevelFactoryBase {
 #undef MUELU_FILTEREDAFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    FilteredAFactory() { }
+  FilteredAFactory() {}
 
-    //! Destructor.
-    virtual ~FilteredAFactory() { }
+  //! Destructor.
+  virtual ~FilteredAFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Builds filtered matrix and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+    Builds filtered matrix and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-    //@}
-  private:
-    void BuildReuse(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const;
-    void BuildNew  (const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const;
-    void BuildNewUsingRootStencil(const Matrix& A, const GraphBase& G, double dirichletThresh, Level& currentLevel, Matrix& filteredA, bool use_spread_lumping, double DdomAllowGrowthRate, double DdomCap) const;
-    void ExperimentalLumping(const Matrix& A, Matrix& filteredA, double rho, double rho2) const;
+  //@}
+ private:
+  void BuildReuse(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const;
+  void BuildNew(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const;
+  void BuildNewUsingRootStencil(const Matrix& A, const GraphBase& G, double dirichletThresh, Level& currentLevel, Matrix& filteredA, bool use_spread_lumping, double DdomAllowGrowthRate, double DdomCap) const;
+  void ExperimentalLumping(const Matrix& A, Matrix& filteredA, double rho, double rho2) const;
 
-  }; //class FilteredAFactory
+};  // class FilteredAFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FILTEREDAFACTORY_SHORT
-#endif // MUELU_FILTEREDAFACTORY_DECL_HPP
+#endif  // MUELU_FILTEREDAFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_FilteredAFactory_def.hpp b/packages/muelu/src/Misc/MueLu_FilteredAFactory_def.hpp
index 2828b3c0cb0e..702aaa763111 100644
--- a/packages/muelu/src/Misc/MueLu_FilteredAFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_FilteredAFactory_def.hpp
@@ -62,174 +62,162 @@
 // Variable to enable lots of debug output
 #define MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING 0
 
-
 namespace MueLu {
 
-  template <class T>
-  void sort_and_unique(T & array) {
-    std::sort(array.begin(),array.end());
-    std::unique(array.begin(),array.end());
-  }
-
+template <class T>
+void sort_and_unique(T& array) {
+  std::sort(array.begin(), array.end());
+  std::unique(array.begin(), array.end());
+}
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("filtered matrix: use lumping");
-    SET_VALID_ENTRY("filtered matrix: reuse graph");
-    SET_VALID_ENTRY("filtered matrix: reuse eigenvalue");
-    SET_VALID_ENTRY("filtered matrix: use root stencil");
-    SET_VALID_ENTRY("filtered matrix: use spread lumping");
-    SET_VALID_ENTRY("filtered matrix: spread lumping diag dom growth factor");
-    SET_VALID_ENTRY("filtered matrix: spread lumping diag dom cap");
-    SET_VALID_ENTRY("filtered matrix: Dirichlet threshold");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used for filtering");
-    validParamList->set< RCP<const FactoryBase> >("Graph",          Teuchos::null, "Generating factory for coalesced filtered graph");
-    validParamList->set< RCP<const FactoryBase> >("Filtering",      Teuchos::null, "Generating factory for filtering boolean");
-
-
-    // Only need these for the "use root stencil" option
-    validParamList->set< RCP<const FactoryBase> >("Aggregates",         Teuchos::null, "Generating factory of the aggregates");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    return validParamList;
+  SET_VALID_ENTRY("filtered matrix: use lumping");
+  SET_VALID_ENTRY("filtered matrix: reuse graph");
+  SET_VALID_ENTRY("filtered matrix: reuse eigenvalue");
+  SET_VALID_ENTRY("filtered matrix: use root stencil");
+  SET_VALID_ENTRY("filtered matrix: use spread lumping");
+  SET_VALID_ENTRY("filtered matrix: spread lumping diag dom growth factor");
+  SET_VALID_ENTRY("filtered matrix: spread lumping diag dom cap");
+  SET_VALID_ENTRY("filtered matrix: Dirichlet threshold");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used for filtering");
+  validParamList->set<RCP<const FactoryBase> >("Graph", Teuchos::null, "Generating factory for coalesced filtered graph");
+  validParamList->set<RCP<const FactoryBase> >("Filtering", Teuchos::null, "Generating factory for filtering boolean");
+
+  // Only need these for the "use root stencil" option
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Generating factory of the aggregates");
+  validParamList->set<RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "Filtering");
+  Input(currentLevel, "Graph");
+  const ParameterList& pL = GetParameterList();
+  if (pL.isParameter("filtered matrix: use root stencil") && pL.get<bool>("filtered matrix: use root stencil") == true) {
+    Input(currentLevel, "Aggregates");
+    Input(currentLevel, "UnAmalgamationInfo");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "Filtering");
-    Input(currentLevel, "Graph");
-    const ParameterList& pL = GetParameterList();
-    if(pL.isParameter("filtered matrix: use root stencil") && pL.get<bool>("filtered matrix: use root stencil") == true){
-      Input(currentLevel, "Aggregates");
-      Input(currentLevel, "UnAmalgamationInfo");
-    }
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Matrix filtering", currentLevel);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Matrix filtering", currentLevel);
-
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-    if (Get<bool>(currentLevel, "Filtering") == false) {
-      GetOStream(Runtime0) << "Filtered matrix is not being constructed as no filtering is being done" << std::endl;
-      Set(currentLevel, "A", A);
-      return;
-    }
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+  if (Get<bool>(currentLevel, "Filtering") == false) {
+    GetOStream(Runtime0) << "Filtered matrix is not being constructed as no filtering is being done" << std::endl;
+    Set(currentLevel, "A", A);
+    return;
+  }
 
-    const ParameterList& pL = GetParameterList();
-    bool lumping = pL.get<bool>("filtered matrix: use lumping");
-    if (lumping)
-      GetOStream(Runtime0) << "Lumping dropped entries" << std::endl;
+  const ParameterList& pL = GetParameterList();
+  bool lumping            = pL.get<bool>("filtered matrix: use lumping");
+  if (lumping)
+    GetOStream(Runtime0) << "Lumping dropped entries" << std::endl;
 
-    bool use_spread_lumping = pL.get<bool>("filtered matrix: use spread lumping");
-    if (use_spread_lumping && (!lumping) )
-      throw std::runtime_error("Must also request 'filtered matrix: use lumping' in order to use spread lumping");
+  bool use_spread_lumping = pL.get<bool>("filtered matrix: use spread lumping");
+  if (use_spread_lumping && (!lumping))
+    throw std::runtime_error("Must also request 'filtered matrix: use lumping' in order to use spread lumping");
 
-    if (use_spread_lumping) {
-      GetOStream(Runtime0) << "using spread lumping " << std::endl;
-    }
+  if (use_spread_lumping) {
+    GetOStream(Runtime0) << "using spread lumping " << std::endl;
+  }
 
-    double DdomAllowGrowthRate = 1.1;
-    double DdomCap = 2.0;
-    if (use_spread_lumping) {
-      DdomAllowGrowthRate = pL.get<double>("filtered matrix: spread lumping diag dom growth factor");
-      DdomCap             = pL.get<double>("filtered matrix: spread lumping diag dom cap");
-    }
-    bool use_root_stencil = lumping && pL.get<bool>("filtered matrix: use root stencil");
-    if (use_root_stencil)
-      GetOStream(Runtime0) << "Using root stencil for dropping" << std::endl;
-    double dirichlet_threshold = pL.get<double>("filtered matrix: Dirichlet threshold");
-    if(dirichlet_threshold >= 0.0)
-      GetOStream(Runtime0) << "Filtering Dirichlet threshold of "<<dirichlet_threshold << std::endl;
-
-    if(use_root_stencil || pL.get<bool>("filtered matrix: reuse graph"))
-      GetOStream(Runtime0) << "Reusing graph"<<std::endl;
-    else
-      GetOStream(Runtime0) << "Generating new graph"<<std::endl;
-
-
-    RCP<GraphBase> G = Get< RCP<GraphBase> >(currentLevel, "Graph");
-    if(MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING)
-    {
-      FILE * f = fopen("graph.dat","w");
-      size_t numGRows = G->GetNodeNumVertices();
-      for (size_t i = 0; i < numGRows; i++) {
-        // Set up filtering array
-        ArrayView<const LO> indsG = G->getNeighborVertices(i);
-        for(size_t j=0; j<(size_t)indsG.size(); j++) {
-          fprintf(f,"%d %d 1.0\n",(int)i,(int)indsG[j]);
-        }
+  double DdomAllowGrowthRate = 1.1;
+  double DdomCap             = 2.0;
+  if (use_spread_lumping) {
+    DdomAllowGrowthRate = pL.get<double>("filtered matrix: spread lumping diag dom growth factor");
+    DdomCap             = pL.get<double>("filtered matrix: spread lumping diag dom cap");
+  }
+  bool use_root_stencil = lumping && pL.get<bool>("filtered matrix: use root stencil");
+  if (use_root_stencil)
+    GetOStream(Runtime0) << "Using root stencil for dropping" << std::endl;
+  double dirichlet_threshold = pL.get<double>("filtered matrix: Dirichlet threshold");
+  if (dirichlet_threshold >= 0.0)
+    GetOStream(Runtime0) << "Filtering Dirichlet threshold of " << dirichlet_threshold << std::endl;
+
+  if (use_root_stencil || pL.get<bool>("filtered matrix: reuse graph"))
+    GetOStream(Runtime0) << "Reusing graph" << std::endl;
+  else
+    GetOStream(Runtime0) << "Generating new graph" << std::endl;
+
+  RCP<GraphBase> G = Get<RCP<GraphBase> >(currentLevel, "Graph");
+  if (MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING) {
+    FILE* f         = fopen("graph.dat", "w");
+    size_t numGRows = G->GetNodeNumVertices();
+    for (size_t i = 0; i < numGRows; i++) {
+      // Set up filtering array
+      ArrayView<const LO> indsG = G->getNeighborVertices(i);
+      for (size_t j = 0; j < (size_t)indsG.size(); j++) {
+        fprintf(f, "%d %d 1.0\n", (int)i, (int)indsG[j]);
       }
-      fclose(f);
     }
+    fclose(f);
+  }
 
-    RCP<ParameterList> fillCompleteParams(new ParameterList);
-    fillCompleteParams->set("No Nonlocal Changes", true);
-
-    RCP<Matrix> filteredA;
-    if(use_root_stencil) {
-      filteredA = MatrixFactory::Build(A->getCrsGraph());
-      filteredA->fillComplete(fillCompleteParams);
-      filteredA->resumeFill();
-      BuildNewUsingRootStencil(*A, *G, dirichlet_threshold, currentLevel,*filteredA, use_spread_lumping,DdomAllowGrowthRate, DdomCap);
-      filteredA->fillComplete(fillCompleteParams);
-
-    }
-    else if (pL.get<bool>("filtered matrix: reuse graph")) {
-      filteredA = MatrixFactory::Build(A->getCrsGraph());
-      filteredA->resumeFill();
-      BuildReuse(*A, *G, (lumping != use_spread_lumping), dirichlet_threshold,*filteredA);
-            // only lump inside BuildReuse if lumping is true and use_spread_lumping is false
-            // note: they use_spread_lumping cannot be true if lumping is false
-
-      if (use_spread_lumping) ExperimentalLumping(*A, *filteredA, DdomAllowGrowthRate, DdomCap);
-      filteredA->fillComplete(fillCompleteParams);
-
-    } else {
-
-      filteredA = MatrixFactory::Build(A->getRowMap(), A->getColMap(), A->getLocalMaxNumRowEntries());
-      BuildNew(*A, *G, (lumping != use_spread_lumping), dirichlet_threshold,*filteredA);
-            // only lump inside BuildNew if lumping is true and use_spread_lumping is false
-            // note: they use_spread_lumping cannot be true if lumping is false
-      if (use_spread_lumping) ExperimentalLumping(*A, *filteredA, DdomAllowGrowthRate, DdomCap);
-      filteredA->fillComplete(A->getDomainMap(), A->getRangeMap(), fillCompleteParams);
-    }
-
-
-
-      if(MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING)
-     {
-       Xpetra::IO<SC,LO,GO,NO>::Write("filteredA.dat", *filteredA);
-
-       //original filtered A  and actual A
-       Xpetra::IO<SC,LO,GO,NO>::Write("A.dat", *A);
-       RCP<Matrix> origFilteredA = MatrixFactory::Build(A->getRowMap(), A->getColMap(), A->getLocalMaxNumRowEntries());
-       BuildNew(*A, *G, lumping, dirichlet_threshold,*origFilteredA);
-       if (use_spread_lumping) ExperimentalLumping(*A, *origFilteredA, DdomAllowGrowthRate, DdomCap);
-       origFilteredA->fillComplete(A->getDomainMap(), A->getRangeMap(), fillCompleteParams);
-       Xpetra::IO<SC,LO,GO,NO>::Write("origFilteredA.dat", *origFilteredA);
-     }
+  RCP<ParameterList> fillCompleteParams(new ParameterList);
+  fillCompleteParams->set("No Nonlocal Changes", true);
+
+  RCP<Matrix> filteredA;
+  if (use_root_stencil) {
+    filteredA = MatrixFactory::Build(A->getCrsGraph());
+    filteredA->fillComplete(fillCompleteParams);
+    filteredA->resumeFill();
+    BuildNewUsingRootStencil(*A, *G, dirichlet_threshold, currentLevel, *filteredA, use_spread_lumping, DdomAllowGrowthRate, DdomCap);
+    filteredA->fillComplete(fillCompleteParams);
+
+  } else if (pL.get<bool>("filtered matrix: reuse graph")) {
+    filteredA = MatrixFactory::Build(A->getCrsGraph());
+    filteredA->resumeFill();
+    BuildReuse(*A, *G, (lumping != use_spread_lumping), dirichlet_threshold, *filteredA);
+    // only lump inside BuildReuse if lumping is true and use_spread_lumping is false
+    // note: they use_spread_lumping cannot be true if lumping is false
+
+    if (use_spread_lumping) ExperimentalLumping(*A, *filteredA, DdomAllowGrowthRate, DdomCap);
+    filteredA->fillComplete(fillCompleteParams);
+
+  } else {
+    filteredA = MatrixFactory::Build(A->getRowMap(), A->getColMap(), A->getLocalMaxNumRowEntries());
+    BuildNew(*A, *G, (lumping != use_spread_lumping), dirichlet_threshold, *filteredA);
+    // only lump inside BuildNew if lumping is true and use_spread_lumping is false
+    // note: they use_spread_lumping cannot be true if lumping is false
+    if (use_spread_lumping) ExperimentalLumping(*A, *filteredA, DdomAllowGrowthRate, DdomCap);
+    filteredA->fillComplete(A->getDomainMap(), A->getRangeMap(), fillCompleteParams);
+  }
 
+  if (MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING) {
+    Xpetra::IO<SC, LO, GO, NO>::Write("filteredA.dat", *filteredA);
 
-    filteredA->SetFixedBlockSize(A->GetFixedBlockSize());
+    // original filtered A  and actual A
+    Xpetra::IO<SC, LO, GO, NO>::Write("A.dat", *A);
+    RCP<Matrix> origFilteredA = MatrixFactory::Build(A->getRowMap(), A->getColMap(), A->getLocalMaxNumRowEntries());
+    BuildNew(*A, *G, lumping, dirichlet_threshold, *origFilteredA);
+    if (use_spread_lumping) ExperimentalLumping(*A, *origFilteredA, DdomAllowGrowthRate, DdomCap);
+    origFilteredA->fillComplete(A->getDomainMap(), A->getRangeMap(), fillCompleteParams);
+    Xpetra::IO<SC, LO, GO, NO>::Write("origFilteredA.dat", *origFilteredA);
+  }
 
-    if (pL.get<bool>("filtered matrix: reuse eigenvalue")) {
-      // Reuse max eigenvalue from A
-      // It is unclear what eigenvalue is the best for the smoothing, but we already may have
-      // the D^{-1}A estimate in A, may as well use it.
-      // NOTE: ML does that too
-      filteredA->SetMaxEigenvalueEstimate(A->GetMaxEigenvalueEstimate());
-    }
+  filteredA->SetFixedBlockSize(A->GetFixedBlockSize());
 
-    Set(currentLevel, "A", filteredA);
+  if (pL.get<bool>("filtered matrix: reuse eigenvalue")) {
+    // Reuse max eigenvalue from A
+    // It is unclear what eigenvalue is the best for the smoothing, but we already may have
+    // the D^{-1}A estimate in A, may as well use it.
+    // NOTE: ML does that too
+    filteredA->SetMaxEigenvalueEstimate(A->GetMaxEigenvalueEstimate());
   }
 
+  Set(currentLevel, "A", filteredA);
+}
+
 // Epetra's API allows direct access to row array.
 // Tpetra's API does not, providing only ArrayView<const .>
 // But in most situations we are currently interested in, it is safe to assume
@@ -238,801 +226,794 @@ namespace MueLu {
 // replaceLocalValues() call which is quite expensive due to all the searches.
 //#define ASSUME_DIRECT_ACCESS_TO_ROW // See github issue 10883#issuecomment-1256676340
 
-  // Both Epetra and Tpetra matrix-matrix multiply use the following trick:
-  // if an entry of the left matrix is zero, it does not compute or store the
-  // zero value.
-  //
-  // This trick allows us to bypass constructing a new matrix. Instead, we
-  // make a deep copy of the original one, and fill it in with zeros, which
-  // are ignored during the prolongator smoothing.
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildReuse(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const {
-    using TST = typename Teuchos::ScalarTraits<SC>;
-    SC zero = TST::zero();
-
-
-    size_t blkSize = A.GetFixedBlockSize();
-
-    ArrayView<const LO> inds;
-    ArrayView<const SC> valsA;
+// Both Epetra and Tpetra matrix-matrix multiply use the following trick:
+// if an entry of the left matrix is zero, it does not compute or store the
+// zero value.
+//
+// This trick allows us to bypass constructing a new matrix. Instead, we
+// make a deep copy of the original one, and fill it in with zeros, which
+// are ignored during the prolongator smoothing.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildReuse(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const {
+  using TST = typename Teuchos::ScalarTraits<SC>;
+  SC zero   = TST::zero();
+
+  size_t blkSize = A.GetFixedBlockSize();
+
+  ArrayView<const LO> inds;
+  ArrayView<const SC> valsA;
 #ifdef ASSUME_DIRECT_ACCESS_TO_ROW
-    ArrayView<SC>       vals;
+  ArrayView<SC> vals;
 #else
-    Array<SC>           vals;
+  Array<SC> vals;
 #endif
 
-    Array<char> filter( std::max(blkSize*G.GetImportMap()->getLocalNumElements(),
-                                 A.getColMap()->getLocalNumElements()),
-                        0);
+  Array<char> filter(std::max(blkSize * G.GetImportMap()->getLocalNumElements(),
+                              A.getColMap()->getLocalNumElements()),
+                     0);
 
-    size_t numGRows = G.GetNodeNumVertices();
-    for (size_t i = 0; i < numGRows; i++) {
-      // Set up filtering array
-      ArrayView<const LO> indsG = G.getNeighborVertices(i);
-      for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
-        for (size_t k = 0; k < blkSize; k++)
-          filter[indsG[j]*blkSize+k] = 1;
+  size_t numGRows = G.GetNodeNumVertices();
+  for (size_t i = 0; i < numGRows; i++) {
+    // Set up filtering array
+    ArrayView<const LO> indsG = G.getNeighborVertices(i);
+    for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+      for (size_t k = 0; k < blkSize; k++)
+        filter[indsG[j] * blkSize + k] = 1;
 
-      for (size_t k = 0; k < blkSize; k++) {
-        LO row = i*blkSize + k;
+    for (size_t k = 0; k < blkSize; k++) {
+      LO row = i * blkSize + k;
 
-        A.getLocalRowView(row, inds, valsA);
+      A.getLocalRowView(row, inds, valsA);
 
-        size_t nnz = inds.size();
-        if (nnz == 0)
-          continue;
+      size_t nnz = inds.size();
+      if (nnz == 0)
+        continue;
 
 #ifdef ASSUME_DIRECT_ACCESS_TO_ROW
-        // Transform ArrayView<const SC> into ArrayView<SC>
-        ArrayView<const SC> vals1;
-        filteredA.getLocalRowView(row, inds, vals1);
-        vals = ArrayView<SC>(const_cast<SC*>(vals1.getRawPtr()), nnz);
+      // Transform ArrayView<const SC> into ArrayView<SC>
+      ArrayView<const SC> vals1;
+      filteredA.getLocalRowView(row, inds, vals1);
+      vals = ArrayView<SC>(const_cast<SC*>(vals1.getRawPtr()), nnz);
 
-        memcpy(vals.getRawPtr(), valsA.getRawPtr(), nnz*sizeof(SC));
+      memcpy(vals.getRawPtr(), valsA.getRawPtr(), nnz * sizeof(SC));
 #else
-        vals = Array<SC>(valsA);
+      vals = Array<SC>(valsA);
 #endif
 
-        SC ZERO = Teuchos::ScalarTraits<SC>::zero();
-        //      SC ONE = Teuchos::ScalarTraits<SC>::one();
-        SC A_rowsum = ZERO, F_rowsum = ZERO;
-        for(LO l = 0; l < (LO)inds.size(); l++)
-          A_rowsum += valsA[l];
+      SC ZERO = Teuchos::ScalarTraits<SC>::zero();
+      //      SC ONE = Teuchos::ScalarTraits<SC>::one();
+      SC A_rowsum = ZERO, F_rowsum = ZERO;
+      for (LO l = 0; l < (LO)inds.size(); l++)
+        A_rowsum += valsA[l];
 
-        if (lumping == false) {
-          for (size_t j = 0; j < nnz; j++)
-            if (!filter[inds[j]])
-              vals[j] = zero;
+      if (lumping == false) {
+        for (size_t j = 0; j < nnz; j++)
+          if (!filter[inds[j]])
+            vals[j] = zero;
 
-        } else {
-          LO diagIndex = -1;
-          SC diagExtra = zero;
-
-          for (size_t j = 0; j < nnz; j++) {
-            if (filter[inds[j]]) {
-              if (inds[j] == row) {
-                // Remember diagonal position
-                diagIndex = j;
-              }
-              continue;
+      } else {
+        LO diagIndex = -1;
+        SC diagExtra = zero;
+
+        for (size_t j = 0; j < nnz; j++) {
+          if (filter[inds[j]]) {
+            if (inds[j] == row) {
+              // Remember diagonal position
+              diagIndex = j;
             }
+            continue;
+          }
 
-            diagExtra += vals[j];
+          diagExtra += vals[j];
 
-            vals[j] = zero;
-          }
+          vals[j] = zero;
+        }
 
-          // Lump dropped entries
-          // NOTE
-          //  * Does it make sense to lump for elasticity?
-          //  * Is it different for diffusion and elasticity?
-          //SC diagA = ZERO;
-          if (diagIndex != -1) {
-            //diagA = vals[diagIndex];
-            vals[diagIndex] += diagExtra;
-            if(dirichletThresh >= 0.0 && TST::real(vals[diagIndex]) <= dirichletThresh) {
-
-                //            printf("WARNING: row %d diag(Afiltered) = %8.2e diag(A)=%8.2e\n",row,vals[diagIndex],diagA);
-              for(LO l = 0; l < (LO)nnz; l++)
-                F_rowsum += vals[l];
-                //            printf("       : A rowsum = %8.2e F rowsum = %8.2e\n",A_rowsum,F_rowsum);
-              vals[diagIndex] = TST::one();
-            }
+        // Lump dropped entries
+        // NOTE
+        //  * Does it make sense to lump for elasticity?
+        //  * Is it different for diffusion and elasticity?
+        // SC diagA = ZERO;
+        if (diagIndex != -1) {
+          // diagA = vals[diagIndex];
+          vals[diagIndex] += diagExtra;
+          if (dirichletThresh >= 0.0 && TST::real(vals[diagIndex]) <= dirichletThresh) {
+            //            printf("WARNING: row %d diag(Afiltered) = %8.2e diag(A)=%8.2e\n",row,vals[diagIndex],diagA);
+            for (LO l = 0; l < (LO)nnz; l++)
+              F_rowsum += vals[l];
+            //            printf("       : A rowsum = %8.2e F rowsum = %8.2e\n",A_rowsum,F_rowsum);
+            vals[diagIndex] = TST::one();
           }
-
         }
+      }
 
 #ifndef ASSUME_DIRECT_ACCESS_TO_ROW
-        // Because we used a column map in the construction of the matrix
-        // we can just use insertLocalValues here instead of insertGlobalValues
-        filteredA.replaceLocalValues(row, inds, vals);
+      // Because we used a column map in the construction of the matrix
+      // we can just use insertLocalValues here instead of insertGlobalValues
+      filteredA.replaceLocalValues(row, inds, vals);
 #endif
-      }
-
-      // Reset filtering array
-      for (size_t j = 0; j < as<size_t> (indsG.size()); j++)
-        for (size_t k = 0; k < blkSize; k++)
-          filter[indsG[j]*blkSize+k] = 0;
     }
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildNew(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const {
-    using TST = typename Teuchos::ScalarTraits<SC>;
-    SC zero = Teuchos::ScalarTraits<SC>::zero();
-
-    size_t blkSize = A.GetFixedBlockSize();
 
-    ArrayView<const LO> indsA;
-    ArrayView<const SC> valsA;
-    Array<LO>           inds;
-    Array<SC>           vals;
-
-    Array<char> filter(blkSize * G.GetImportMap()->getLocalNumElements(), 0);
-
-    size_t numGRows = G.GetNodeNumVertices();
-    for (size_t i = 0; i < numGRows; i++) {
-      // Set up filtering array
-      ArrayView<const LO> indsG = G.getNeighborVertices(i);
-      for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
-        for (size_t k = 0; k < blkSize; k++)
-          filter[indsG[j]*blkSize+k] = 1;
-
-      for (size_t k = 0; k < blkSize; k++) {
-        LO row = i*blkSize + k;
-
-        A.getLocalRowView(row, indsA, valsA);
-
-        size_t nnz = indsA.size();
-        if (nnz == 0)
-          continue;
-
-        inds.resize(indsA.size());
-        vals.resize(valsA.size());
+    // Reset filtering array
+    for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+      for (size_t k = 0; k < blkSize; k++)
+        filter[indsG[j] * blkSize + k] = 0;
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildNew(const Matrix& A, const GraphBase& G, const bool lumping, double dirichletThresh, Matrix& filteredA) const {
+  using TST = typename Teuchos::ScalarTraits<SC>;
+  SC zero   = Teuchos::ScalarTraits<SC>::zero();
+
+  size_t blkSize = A.GetFixedBlockSize();
+
+  ArrayView<const LO> indsA;
+  ArrayView<const SC> valsA;
+  Array<LO> inds;
+  Array<SC> vals;
+
+  Array<char> filter(blkSize * G.GetImportMap()->getLocalNumElements(), 0);
+
+  size_t numGRows = G.GetNodeNumVertices();
+  for (size_t i = 0; i < numGRows; i++) {
+    // Set up filtering array
+    ArrayView<const LO> indsG = G.getNeighborVertices(i);
+    for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+      for (size_t k = 0; k < blkSize; k++)
+        filter[indsG[j] * blkSize + k] = 1;
+
+    for (size_t k = 0; k < blkSize; k++) {
+      LO row = i * blkSize + k;
+
+      A.getLocalRowView(row, indsA, valsA);
+
+      size_t nnz = indsA.size();
+      if (nnz == 0)
+        continue;
+
+      inds.resize(indsA.size());
+      vals.resize(valsA.size());
+
+      size_t numInds = 0;
+      if (lumping == false) {
+        for (size_t j = 0; j < nnz; j++)
+          if (filter[indsA[j]]) {
+            inds[numInds] = indsA[j];
+            vals[numInds] = valsA[j];
+            numInds++;
+          }
 
-        size_t numInds = 0;
-        if (lumping == false) {
-          for (size_t j = 0; j < nnz; j++)
-            if (filter[indsA[j]]) {
-              inds[numInds] = indsA[j];
-              vals[numInds] = valsA[j];
-              numInds++;
-            }
+      } else {
+        LO diagIndex = -1;
+        SC diagExtra = zero;
 
-        } else {
-          LO diagIndex = -1;
-          SC diagExtra = zero;
-
-          for (size_t j = 0; j < nnz; j++) {
-            if (filter[indsA[j]]) {
-              inds[numInds] = indsA[j];
-              vals[numInds] = valsA[j];
+        for (size_t j = 0; j < nnz; j++) {
+          if (filter[indsA[j]]) {
+            inds[numInds] = indsA[j];
+            vals[numInds] = valsA[j];
 
-              // Remember diagonal position
-              if (inds[numInds] == row)
-                diagIndex = numInds;
+            // Remember diagonal position
+            if (inds[numInds] == row)
+              diagIndex = numInds;
 
-              numInds++;
+            numInds++;
 
-            } else {
-              diagExtra += valsA[j];
-            }
+          } else {
+            diagExtra += valsA[j];
           }
+        }
 
-          // Lump dropped entries
-          // NOTE
-          //  * Does it make sense to lump for elasticity?
-          //  * Is it different for diffusion and elasticity?
-          if (diagIndex != -1) {
-            vals[diagIndex] += diagExtra;
-            if(dirichletThresh >= 0.0 && TST::real(vals[diagIndex]) <= dirichletThresh) {
-              //              SC A_rowsum = ZERO, F_rowsum = ZERO;
-              //              printf("WARNING: row %d diag(Afiltered) = %8.2e diag(A)=%8.2e\n",row,vals[diagIndex],diagA);
-              //              for(LO l = 0; l < (LO)nnz; l++)
-              //                F_rowsum += vals[l];
-              //              printf("       : A rowsum = %8.2e F rowsum = %8.2e\n",A_rowsum,F_rowsum);
-              vals[diagIndex] = TST::one();
-            }
+        // Lump dropped entries
+        // NOTE
+        //  * Does it make sense to lump for elasticity?
+        //  * Is it different for diffusion and elasticity?
+        if (diagIndex != -1) {
+          vals[diagIndex] += diagExtra;
+          if (dirichletThresh >= 0.0 && TST::real(vals[diagIndex]) <= dirichletThresh) {
+            //              SC A_rowsum = ZERO, F_rowsum = ZERO;
+            //              printf("WARNING: row %d diag(Afiltered) = %8.2e diag(A)=%8.2e\n",row,vals[diagIndex],diagA);
+            //              for(LO l = 0; l < (LO)nnz; l++)
+            //                F_rowsum += vals[l];
+            //              printf("       : A rowsum = %8.2e F rowsum = %8.2e\n",A_rowsum,F_rowsum);
+            vals[diagIndex] = TST::one();
           }
-
         }
-        inds.resize(numInds);
-        vals.resize(numInds);
-
-
-
-        // Because we used a column map in the construction of the matrix
-        // we can just use insertLocalValues here instead of insertGlobalValues
-        filteredA.insertLocalValues(row, inds, vals);
       }
+      inds.resize(numInds);
+      vals.resize(numInds);
 
-      // Reset filtering array
-      for (size_t j = 0; j < as<size_t> (indsG.size()); j++)
-        for (size_t k = 0; k < blkSize; k++)
-          filter[indsG[j]*blkSize+k] = 0;
+      // Because we used a column map in the construction of the matrix
+      // we can just use insertLocalValues here instead of insertGlobalValues
+      filteredA.insertLocalValues(row, inds, vals);
     }
-  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildNewUsingRootStencil(const Matrix& A, const GraphBase& G, double dirichletThresh, Level& currentLevel,  Matrix& filteredA, bool use_spread_lumping, double DdomAllowGrowthRate, double DdomCap) const {
-    using TST = typename Teuchos::ScalarTraits<SC>;
-    using Teuchos::arcp_const_cast;
-    SC ZERO = Teuchos::ScalarTraits<SC>::zero();
-    SC ONE = Teuchos::ScalarTraits<SC>::one();
-    LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-
-    size_t numNodes = G.GetNodeNumVertices();
-    size_t blkSize  = A.GetFixedBlockSize();
-    size_t numRows  = A.getMap()->getLocalNumElements();
-    ArrayView<const LO> indsA;
-    ArrayView<const SC> valsA;
-    ArrayRCP<const size_t> rowptr;
-    ArrayRCP<const LO> inds;
-    ArrayRCP<const SC> vals_const;
-    ArrayRCP<SC> vals;
-
-    // We're going to grab the vals array from filteredA and then blitz it with NAN as a placeholder for "entries that have
-    // not yey been touched."  If I see an entry in the primary loop that has a zero, then I assume it has been nuked by
-    // it's symmetric pair, so I add it to the diagonal.  If it has a NAN, process as normal.
-    RCP<CrsMatrix> filteredAcrs = dynamic_cast<const CrsMatrixWrap*>(&filteredA)->getCrsMatrix();
-    filteredAcrs->getAllValues(rowptr,inds,vals_const);
-    vals = arcp_const_cast<SC>(vals_const);
-    Array<bool> vals_dropped_indicator(vals.size(),false);
-
-    // In the badAggNeighbors loop, if the entry has any number besides NAN, I add it to the diagExtra and then zero the guy.
-    RCP<Aggregates>            aggregates     = Get< RCP<Aggregates> >           (currentLevel, "Aggregates");
-    RCP<AmalgamationInfo>      amalgInfo      = Get< RCP<AmalgamationInfo> >     (currentLevel, "UnAmalgamationInfo");
-    LO                          numAggs       = aggregates->GetNumAggregates();
-
-    // Check map nesting
-    RCP<const Map> rowMap = A.getRowMap();
-    RCP<const Map> colMap = A.getColMap();
-    bool goodMap = MueLu::Utilities<SC,LO,GO,NO>::MapsAreNested(*rowMap, *colMap);
-    TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError,"FilteredAFactory: Maps are not nested");
-
-    // Since we're going to symmetrize this
-    Array<LO> diagIndex(numRows,INVALID);
-    Array<SC> diagExtra(numRows,ZERO);
-
-    // Lists of nodes in each aggregate
-    struct {
-      // GH: For now, copy everything to host until we properly set this factory to run device code
-      // Instead, we'll copy data into HostMirrors and run the algorithms on host, saving optimization for later.
-      typename Aggregates::LO_view             ptr,   nodes,   unaggregated;
-      typename Aggregates::LO_view::HostMirror ptr_h, nodes_h, unaggregated_h;
-    } nodesInAgg;
-    aggregates->ComputeNodesInAggregate(nodesInAgg.ptr, nodesInAgg.nodes, nodesInAgg.unaggregated);
-    nodesInAgg.ptr_h =          Kokkos::create_mirror_view(nodesInAgg.ptr);
-    nodesInAgg.nodes_h =        Kokkos::create_mirror_view(nodesInAgg.nodes);
-    nodesInAgg.unaggregated_h = Kokkos::create_mirror_view(nodesInAgg.unaggregated);
-    Kokkos::deep_copy(nodesInAgg.ptr_h,          nodesInAgg.ptr);
-    Kokkos::deep_copy(nodesInAgg.nodes_h,        nodesInAgg.nodes);
-    Kokkos::deep_copy(nodesInAgg.unaggregated_h, nodesInAgg.unaggregated);
-    Teuchos::ArrayRCP<const LO> vertex2AggId  = aggregates->GetVertex2AggId()->getData(0); // GH: this is needed on device, grab the pointer after we call ComputeNodesInAggregate
-
-    LO graphNumCols = G.GetImportMap()->getLocalNumElements();
-    Array<bool> filter(graphNumCols, false);
-
-    // Loop over the unaggregated nodes. Blitz those rows. We don't want to smooth singletons.
-    for(LO i=0; i< (LO)nodesInAgg.unaggregated_h.extent(0); i++) {
-      for (LO m = 0; m < (LO)blkSize; m++) {
-        LO row = amalgInfo->ComputeLocalDOF(nodesInAgg.unaggregated_h(i),m);
-        if (row >= (LO)numRows) continue;
-        size_t index_start = rowptr[row];
-        A.getLocalRowView(row, indsA, valsA);
-        for(LO k=0; k<(LO)indsA.size(); k++) {
-          if(row == indsA[k]) {
-            vals[index_start+k] = ONE;
-            diagIndex[row] = k;
-          }
-          else
-            vals[index_start+k] = ZERO;
-        }
+    // Reset filtering array
+    for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+      for (size_t k = 0; k < blkSize; k++)
+        filter[indsG[j] * blkSize + k] = 0;
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildNewUsingRootStencil(const Matrix& A, const GraphBase& G, double dirichletThresh, Level& currentLevel, Matrix& filteredA, bool use_spread_lumping, double DdomAllowGrowthRate, double DdomCap) const {
+  using TST = typename Teuchos::ScalarTraits<SC>;
+  using Teuchos::arcp_const_cast;
+  SC ZERO    = Teuchos::ScalarTraits<SC>::zero();
+  SC ONE     = Teuchos::ScalarTraits<SC>::one();
+  LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  size_t numNodes = G.GetNodeNumVertices();
+  size_t blkSize  = A.GetFixedBlockSize();
+  size_t numRows  = A.getMap()->getLocalNumElements();
+  ArrayView<const LO> indsA;
+  ArrayView<const SC> valsA;
+  ArrayRCP<const size_t> rowptr;
+  ArrayRCP<const LO> inds;
+  ArrayRCP<const SC> vals_const;
+  ArrayRCP<SC> vals;
+
+  // We're going to grab the vals array from filteredA and then blitz it with NAN as a placeholder for "entries that have
+  // not yey been touched."  If I see an entry in the primary loop that has a zero, then I assume it has been nuked by
+  // it's symmetric pair, so I add it to the diagonal.  If it has a NAN, process as normal.
+  RCP<CrsMatrix> filteredAcrs = dynamic_cast<const CrsMatrixWrap*>(&filteredA)->getCrsMatrix();
+  filteredAcrs->getAllValues(rowptr, inds, vals_const);
+  vals = arcp_const_cast<SC>(vals_const);
+  Array<bool> vals_dropped_indicator(vals.size(), false);
+
+  // In the badAggNeighbors loop, if the entry has any number besides NAN, I add it to the diagExtra and then zero the guy.
+  RCP<Aggregates> aggregates      = Get<RCP<Aggregates> >(currentLevel, "Aggregates");
+  RCP<AmalgamationInfo> amalgInfo = Get<RCP<AmalgamationInfo> >(currentLevel, "UnAmalgamationInfo");
+  LO numAggs                      = aggregates->GetNumAggregates();
+
+  // Check map nesting
+  RCP<const Map> rowMap = A.getRowMap();
+  RCP<const Map> colMap = A.getColMap();
+  bool goodMap          = MueLu::Utilities<SC, LO, GO, NO>::MapsAreNested(*rowMap, *colMap);
+  TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError, "FilteredAFactory: Maps are not nested");
+
+  // Since we're going to symmetrize this
+  Array<LO> diagIndex(numRows, INVALID);
+  Array<SC> diagExtra(numRows, ZERO);
+
+  // Lists of nodes in each aggregate
+  struct {
+    // GH: For now, copy everything to host until we properly set this factory to run device code
+    // Instead, we'll copy data into HostMirrors and run the algorithms on host, saving optimization for later.
+    typename Aggregates::LO_view ptr, nodes, unaggregated;
+    typename Aggregates::LO_view::HostMirror ptr_h, nodes_h, unaggregated_h;
+  } nodesInAgg;
+  aggregates->ComputeNodesInAggregate(nodesInAgg.ptr, nodesInAgg.nodes, nodesInAgg.unaggregated);
+  nodesInAgg.ptr_h          = Kokkos::create_mirror_view(nodesInAgg.ptr);
+  nodesInAgg.nodes_h        = Kokkos::create_mirror_view(nodesInAgg.nodes);
+  nodesInAgg.unaggregated_h = Kokkos::create_mirror_view(nodesInAgg.unaggregated);
+  Kokkos::deep_copy(nodesInAgg.ptr_h, nodesInAgg.ptr);
+  Kokkos::deep_copy(nodesInAgg.nodes_h, nodesInAgg.nodes);
+  Kokkos::deep_copy(nodesInAgg.unaggregated_h, nodesInAgg.unaggregated);
+  Teuchos::ArrayRCP<const LO> vertex2AggId = aggregates->GetVertex2AggId()->getData(0);  // GH: this is needed on device, grab the pointer after we call ComputeNodesInAggregate
+
+  LO graphNumCols = G.GetImportMap()->getLocalNumElements();
+  Array<bool> filter(graphNumCols, false);
+
+  // Loop over the unaggregated nodes. Blitz those rows. We don't want to smooth singletons.
+  for (LO i = 0; i < (LO)nodesInAgg.unaggregated_h.extent(0); i++) {
+    for (LO m = 0; m < (LO)blkSize; m++) {
+      LO row = amalgInfo->ComputeLocalDOF(nodesInAgg.unaggregated_h(i), m);
+      if (row >= (LO)numRows) continue;
+      size_t index_start = rowptr[row];
+      A.getLocalRowView(row, indsA, valsA);
+      for (LO k = 0; k < (LO)indsA.size(); k++) {
+        if (row == indsA[k]) {
+          vals[index_start + k] = ONE;
+          diagIndex[row]        = k;
+        } else
+          vals[index_start + k] = ZERO;
       }
-    }//end nodesInAgg.unaggregated.extent(0);
-
-
-    std::vector<LO> badCount(numAggs,0);
-
-    // Find the biggest aggregate size in *nodes*
-    LO maxAggSize=0;
-    for(LO i=0; i<numAggs; i++)
-      maxAggSize = std::max(maxAggSize,nodesInAgg.ptr_h(i+1) - nodesInAgg.ptr_h(i));
-
-
-    // Loop over all the aggregates
-    std::vector<LO> goodAggNeighbors(G.getLocalMaxNumRowEntries());
-    std::vector<LO> badAggNeighbors(std::min(G.getLocalMaxNumRowEntries()*maxAggSize,numNodes));
-
-    size_t numNewDrops=0;
-    size_t numOldDrops=0;
-    size_t numFixedDiags=0;
-    size_t numSymDrops = 0;
-
-    for(LO i=0; i<numAggs; i++) {
-      LO numNodesInAggregate = nodesInAgg.ptr_h(i+1) - nodesInAgg.ptr_h(i);
-      if(numNodesInAggregate == 0) continue;
-
-      // Find the root *node*
-      LO root_node = INVALID;
-      for (LO k=nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i+1); k++) {
-        if(aggregates->IsRoot(nodesInAgg.nodes_h(k))) {
-          root_node = nodesInAgg.nodes_h(k); break;
-        }
+    }
+  }  // end nodesInAgg.unaggregated.extent(0);
+
+  std::vector<LO> badCount(numAggs, 0);
+
+  // Find the biggest aggregate size in *nodes*
+  LO maxAggSize = 0;
+  for (LO i = 0; i < numAggs; i++)
+    maxAggSize = std::max(maxAggSize, nodesInAgg.ptr_h(i + 1) - nodesInAgg.ptr_h(i));
+
+  // Loop over all the aggregates
+  std::vector<LO> goodAggNeighbors(G.getLocalMaxNumRowEntries());
+  std::vector<LO> badAggNeighbors(std::min(G.getLocalMaxNumRowEntries() * maxAggSize, numNodes));
+
+  size_t numNewDrops   = 0;
+  size_t numOldDrops   = 0;
+  size_t numFixedDiags = 0;
+  size_t numSymDrops   = 0;
+
+  for (LO i = 0; i < numAggs; i++) {
+    LO numNodesInAggregate = nodesInAgg.ptr_h(i + 1) - nodesInAgg.ptr_h(i);
+    if (numNodesInAggregate == 0) continue;
+
+    // Find the root *node*
+    LO root_node = INVALID;
+    for (LO k = nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i + 1); k++) {
+      if (aggregates->IsRoot(nodesInAgg.nodes_h(k))) {
+        root_node = nodesInAgg.nodes_h(k);
+        break;
       }
+    }
 
-      TEUCHOS_TEST_FOR_EXCEPTION(root_node == INVALID,
-                                 Exceptions::RuntimeError,"MueLu::FilteredAFactory::BuildNewUsingRootStencil: Cannot find root node");
+    TEUCHOS_TEST_FOR_EXCEPTION(root_node == INVALID,
+                               Exceptions::RuntimeError, "MueLu::FilteredAFactory::BuildNewUsingRootStencil: Cannot find root node");
 
-      // Find the list of "good" node neighbors (aka nodes which border the root node in the Graph G)
-      ArrayView<const LO> goodNodeNeighbors  = G.getNeighborVertices(root_node);
+    // Find the list of "good" node neighbors (aka nodes which border the root node in the Graph G)
+    ArrayView<const LO> goodNodeNeighbors = G.getNeighborVertices(root_node);
 
-      // Now find the list of "good" aggregate neighbors (aka the aggregates neighbor the root node in the Graph G)
-      goodAggNeighbors.resize(0);
-      for(LO k=0; k<(LO) goodNodeNeighbors.size(); k++) {
-        goodAggNeighbors.push_back(vertex2AggId[goodNodeNeighbors[k]]);
-      }
-      sort_and_unique(goodAggNeighbors);
-
-      // Now we get the list of "bad" aggregate neighbors (aka aggregates which border the
-      // root node in the original matrix A, which are not goodNodeNeighbors).  Since we
-      // don't have an amalgamated version of the original matrix, we use the matrix directly
-      badAggNeighbors.resize(0);
-      for(LO j = 0; j < (LO)blkSize; j++) {
-        LO row = amalgInfo->ComputeLocalDOF(root_node,j);
-        if (row >= (LO)numRows) continue;
-        A.getLocalRowView(row, indsA, valsA);
-        for(LO k=0; k<(LO)indsA.size(); k++) {
-          if ( (indsA[k] < (LO)numRows) && (TST::magnitude(valsA[k]) != TST::magnitude(ZERO))) {
-            LO node = amalgInfo->ComputeLocalNode(indsA[k]);
-            LO agg = vertex2AggId[node];
-            if(!std::binary_search(goodAggNeighbors.begin(),goodAggNeighbors.end(),agg))
-              badAggNeighbors.push_back(agg);
-          }
+    // Now find the list of "good" aggregate neighbors (aka the aggregates neighbor the root node in the Graph G)
+    goodAggNeighbors.resize(0);
+    for (LO k = 0; k < (LO)goodNodeNeighbors.size(); k++) {
+      goodAggNeighbors.push_back(vertex2AggId[goodNodeNeighbors[k]]);
+    }
+    sort_and_unique(goodAggNeighbors);
+
+    // Now we get the list of "bad" aggregate neighbors (aka aggregates which border the
+    // root node in the original matrix A, which are not goodNodeNeighbors).  Since we
+    // don't have an amalgamated version of the original matrix, we use the matrix directly
+    badAggNeighbors.resize(0);
+    for (LO j = 0; j < (LO)blkSize; j++) {
+      LO row = amalgInfo->ComputeLocalDOF(root_node, j);
+      if (row >= (LO)numRows) continue;
+      A.getLocalRowView(row, indsA, valsA);
+      for (LO k = 0; k < (LO)indsA.size(); k++) {
+        if ((indsA[k] < (LO)numRows) && (TST::magnitude(valsA[k]) != TST::magnitude(ZERO))) {
+          LO node = amalgInfo->ComputeLocalNode(indsA[k]);
+          LO agg  = vertex2AggId[node];
+          if (!std::binary_search(goodAggNeighbors.begin(), goodAggNeighbors.end(), agg))
+            badAggNeighbors.push_back(agg);
         }
       }
-      sort_and_unique(badAggNeighbors);
+    }
+    sort_and_unique(badAggNeighbors);
 
-      // Go through the filtered graph and count the number of connections to the badAggNeighbors
-      // if there are 2 or more of these connections, remove them from the bad list.
+    // Go through the filtered graph and count the number of connections to the badAggNeighbors
+    // if there are 2 or more of these connections, remove them from the bad list.
 
-      for (LO k=nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i+1); k++) {
-        ArrayView<const LO> nodeNeighbors  = G.getNeighborVertices(k);
-        for (LO kk=0; kk < nodeNeighbors.size(); kk++) {
-          if ( (vertex2AggId[nodeNeighbors[kk]] >= 0) && (vertex2AggId[nodeNeighbors[kk]] < numAggs))
-            (badCount[vertex2AggId[nodeNeighbors[kk]]])++;
-        }
-      }
-      std::vector<LO> reallyBadAggNeighbors(std::min(G.getLocalMaxNumRowEntries()*maxAggSize,numNodes));
-      reallyBadAggNeighbors.resize(0);
-      for (LO k=0; k < (LO) badAggNeighbors.size(); k++) {
-        if (badCount[badAggNeighbors[k]] <= 1 ) reallyBadAggNeighbors.push_back(badAggNeighbors[k]);
+    for (LO k = nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i + 1); k++) {
+      ArrayView<const LO> nodeNeighbors = G.getNeighborVertices(k);
+      for (LO kk = 0; kk < nodeNeighbors.size(); kk++) {
+        if ((vertex2AggId[nodeNeighbors[kk]] >= 0) && (vertex2AggId[nodeNeighbors[kk]] < numAggs))
+          (badCount[vertex2AggId[nodeNeighbors[kk]]])++;
       }
-      for (LO k=nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i+1); k++) {
-        ArrayView<const LO> nodeNeighbors  = G.getNeighborVertices(k);
-        for (LO kk=0; kk < nodeNeighbors.size(); kk++) {
-          if ( (vertex2AggId[nodeNeighbors[kk]] >= 0) && (vertex2AggId[nodeNeighbors[kk]] < numAggs))
-            badCount[vertex2AggId[nodeNeighbors[kk]]] = 0;
-        }
+    }
+    std::vector<LO> reallyBadAggNeighbors(std::min(G.getLocalMaxNumRowEntries() * maxAggSize, numNodes));
+    reallyBadAggNeighbors.resize(0);
+    for (LO k = 0; k < (LO)badAggNeighbors.size(); k++) {
+      if (badCount[badAggNeighbors[k]] <= 1) reallyBadAggNeighbors.push_back(badAggNeighbors[k]);
+    }
+    for (LO k = nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i + 1); k++) {
+      ArrayView<const LO> nodeNeighbors = G.getNeighborVertices(k);
+      for (LO kk = 0; kk < nodeNeighbors.size(); kk++) {
+        if ((vertex2AggId[nodeNeighbors[kk]] >= 0) && (vertex2AggId[nodeNeighbors[kk]] < numAggs))
+          badCount[vertex2AggId[nodeNeighbors[kk]]] = 0;
       }
+    }
 
-      // For each of the reallyBadAggNeighbors, we go and blitz their connections to dofs in this aggregate.
-      // We remove the INVALID marker when we do this so we don't wind up doubling this up later
-      for(LO b=0; b<(LO)reallyBadAggNeighbors.size(); b++) {
-        LO bad_agg = reallyBadAggNeighbors[b];
-        for (LO k=nodesInAgg.ptr_h(bad_agg); k < nodesInAgg.ptr_h(bad_agg+1); k++) {
-          LO bad_node = nodesInAgg.nodes_h(k);
-          for(LO j = 0; j < (LO)blkSize; j++) {
-            LO bad_row = amalgInfo->ComputeLocalDOF(bad_node,j);
-            if (bad_row >= (LO)numRows) continue;
-            size_t index_start = rowptr[bad_row];
-            A.getLocalRowView(bad_row, indsA, valsA);
-            for(LO l = 0; l < (LO)indsA.size(); l++) {
-              if(indsA[l] < (LO)numRows && vertex2AggId[amalgInfo->ComputeLocalNode(indsA[l])] == i && vals_dropped_indicator[index_start+l] == false) {
-                vals_dropped_indicator[index_start + l] = true;
-                vals[index_start + l] = ZERO;
-                diagExtra[bad_row] += valsA[l];
-                numSymDrops++;
-              }
+    // For each of the reallyBadAggNeighbors, we go and blitz their connections to dofs in this aggregate.
+    // We remove the INVALID marker when we do this so we don't wind up doubling this up later
+    for (LO b = 0; b < (LO)reallyBadAggNeighbors.size(); b++) {
+      LO bad_agg = reallyBadAggNeighbors[b];
+      for (LO k = nodesInAgg.ptr_h(bad_agg); k < nodesInAgg.ptr_h(bad_agg + 1); k++) {
+        LO bad_node = nodesInAgg.nodes_h(k);
+        for (LO j = 0; j < (LO)blkSize; j++) {
+          LO bad_row = amalgInfo->ComputeLocalDOF(bad_node, j);
+          if (bad_row >= (LO)numRows) continue;
+          size_t index_start = rowptr[bad_row];
+          A.getLocalRowView(bad_row, indsA, valsA);
+          for (LO l = 0; l < (LO)indsA.size(); l++) {
+            if (indsA[l] < (LO)numRows && vertex2AggId[amalgInfo->ComputeLocalNode(indsA[l])] == i && vals_dropped_indicator[index_start + l] == false) {
+              vals_dropped_indicator[index_start + l] = true;
+              vals[index_start + l]                   = ZERO;
+              diagExtra[bad_row] += valsA[l];
+              numSymDrops++;
             }
           }
         }
       }
+    }
 
-      // Now lets fill the rows in this aggregate and figure out the diagonal lumping
-      // We loop over each node in the aggregate and then over the neighbors of that node
-
-      for(LO k=nodesInAgg.ptr_h(i); k<nodesInAgg.ptr_h(i+1); k++) {
-        LO row_node = nodesInAgg.nodes_h(k);
-
-        // Set up filtering array
-        ArrayView<const LO> indsG = G.getNeighborVertices(row_node);
-        for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
-          filter[indsG[j]]=true;
-
-        for (LO m = 0; m < (LO)blkSize; m++) {
-          LO row = amalgInfo->ComputeLocalDOF(row_node,m);
-          if (row >= (LO)numRows) continue;
-          size_t index_start = rowptr[row];
-          A.getLocalRowView(row, indsA, valsA);
-
-          for(LO l = 0; l < (LO)indsA.size(); l++) {
-            int col_node = amalgInfo->ComputeLocalNode(indsA[l]);
-            bool is_good = filter[col_node];
-            if (indsA[l] == row) {
-              diagIndex[row] = l;
-              vals[index_start + l] = valsA[l];
-              continue;
-            }
+    // Now lets fill the rows in this aggregate and figure out the diagonal lumping
+    // We loop over each node in the aggregate and then over the neighbors of that node
 
-            // If we've already dropped this guy (from symmetry above), then continue onward
-            if(vals_dropped_indicator[index_start +l] == true) {
-              if(is_good) numOldDrops++;
-              else numNewDrops++;
-              continue;
-            }
+    for (LO k = nodesInAgg.ptr_h(i); k < nodesInAgg.ptr_h(i + 1); k++) {
+      LO row_node = nodesInAgg.nodes_h(k);
 
+      // Set up filtering array
+      ArrayView<const LO> indsG = G.getNeighborVertices(row_node);
+      for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+        filter[indsG[j]] = true;
 
-            // FIXME: I'm assuming vertex2AggId is only length of the rowmap, so
-            // we won'd do secondary dropping on off-processor neighbors
-            if(is_good && indsA[l] < (LO)numRows)  {
-              int agg = vertex2AggId[col_node];
-              if(std::binary_search(reallyBadAggNeighbors.begin(),reallyBadAggNeighbors.end(),agg))
-                is_good = false;
-            }
+      for (LO m = 0; m < (LO)blkSize; m++) {
+        LO row = amalgInfo->ComputeLocalDOF(row_node, m);
+        if (row >= (LO)numRows) continue;
+        size_t index_start = rowptr[row];
+        A.getLocalRowView(row, indsA, valsA);
 
-            if(is_good){
-              vals[index_start+l] = valsA[l];
-            }
-            else {
-              if(!filter[col_node]) numOldDrops++;
-              else numNewDrops++;
-              diagExtra[row] += valsA[l];
-              vals[index_start+l]=ZERO;
-              vals_dropped_indicator[index_start+l]=true;
-            }
-          } //end for l "indsA.size()" loop
-
-        }//end m "blkSize" loop
-
-        // Clear filtering array
-        for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
-          filter[indsG[j]]=false;
-
-      }// end k loop over number of nodes in this agg
-    }//end i loop over numAggs
-
-    if (!use_spread_lumping) {
-      // Now do the diagonal modifications in one, final pass
-      for(LO row=0; row <(LO)numRows; row++) {
-        if (diagIndex[row] != INVALID) {
-          size_t index_start = rowptr[row];
-          size_t diagIndexInMatrix =  index_start + diagIndex[row];
-          //        printf("diag_vals pre update =  %8.2e\n", vals[diagIndex] );
-          vals[diagIndexInMatrix] += diagExtra[row];
-          SC A_rowsum=ZERO, A_absrowsum = ZERO, F_rowsum = ZERO;
-
-
-          if( (dirichletThresh >= 0.0 && TST::real(vals[diagIndexInMatrix]) <= dirichletThresh) ||  TST::real(vals[diagIndexInMatrix]) == ZERO) {
-
-            if(MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING>0) {
-              A.getLocalRowView(row, indsA, valsA);
-              //            SC diagA = valsA[diagIndex[row]];
-              //            printf("WARNING: row %d (diagIndex=%d) diag(Afiltered) = %8.2e diag(A)=%8.2e numInds = %d\n",row,diagIndex[row],vals[diagIndexInMatrix],diagA,(LO)indsA.size());
-
-              for(LO l = 0; l < (LO)indsA.size(); l++) {
-                A_rowsum += valsA[l];
-                A_absrowsum+=std::abs(valsA[l]);
-              }
-              for(LO l = 0; l < (LO)indsA.size(); l++)
-                F_rowsum += vals[index_start+l];
-              //                printf("       : A rowsum = %8.2e |A| rowsum = %8.2e rowsum = %8.2e\n",A_rowsum,A_absrowsum,F_rowsum);
-              if(MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING > 1){
-                //            printf("        Avals =");
-                //            for(LO l = 0; l < (LO)indsA.size(); l++)
-                //              printf("%d(%8.2e)[%d] ",(LO)indsA[l],valsA[l],(LO)l);
-                //            printf("\n");
-                //            printf("        Fvals =");
-                //            for(LO l = 0; l < (LO)indsA.size(); l++)
-                //              if(vals[index_start+l] != ZERO)
-                //                printf("%d(%8.2e)[%d] ",(LO)indsA[l],vals[index_start+l],(LO)l);
-              }
-            }
-            // Don't know what to do, so blitz the row and dump a one on the diagonal
-            for(size_t l=rowptr[row]; l<rowptr[row+1]; l++) {
-              vals[l] = ZERO;
-            }
-            vals[diagIndexInMatrix] = TST::one();
-            numFixedDiags++;
+        for (LO l = 0; l < (LO)indsA.size(); l++) {
+          int col_node = amalgInfo->ComputeLocalNode(indsA[l]);
+          bool is_good = filter[col_node];
+          if (indsA[l] == row) {
+            diagIndex[row]        = l;
+            vals[index_start + l] = valsA[l];
+            continue;
           }
-        }
-        else {
-          GetOStream(Runtime0)<<"WARNING: Row "<<row<<" has no diagonal "<<std::endl;
-        }
-      }/*end row "numRows" loop"*/
-    }
 
-    // Copy all the goop out
-    for(LO row=0; row<(LO)numRows; row++) {
-      filteredA.replaceLocalValues(row, inds(rowptr[row],rowptr[row+1]-rowptr[row]), vals(rowptr[row],rowptr[row+1]-rowptr[row]));
-    }
-   if (use_spread_lumping) ExperimentalLumping(A, filteredA, DdomAllowGrowthRate, DdomCap);
-
-    size_t g_newDrops = 0, g_oldDrops = 0, g_fixedDiags=0;
-
-    MueLu_sumAll(A.getRowMap()->getComm(), numNewDrops, g_newDrops);
-    MueLu_sumAll(A.getRowMap()->getComm(), numOldDrops, g_oldDrops);
-    MueLu_sumAll(A.getRowMap()->getComm(), numFixedDiags, g_fixedDiags);
-    GetOStream(Runtime0)<< "Filtering out "<<g_newDrops<<" edges, in addition to the "<<g_oldDrops<<" edges dropped earlier"<<std::endl;
-    GetOStream(Runtime0)<< "Fixing "<< g_fixedDiags<<" zero diagonal values" <<std::endl;
-  }
+          // If we've already dropped this guy (from symmetry above), then continue onward
+          if (vals_dropped_indicator[index_start + l] == true) {
+            if (is_good)
+              numOldDrops++;
+            else
+              numNewDrops++;
+            continue;
+          }
 
-  // fancy lumping trying to not just move everything to the diagonal but to also consider moving
-  // some lumping to the kept off-diagonals. We basically aim to not increase the diagonal
-  // dominance in a row. In particular, the goal is that row i satisfies
-  //
-  //        lumpedDiagDomMeasure_i   <= rho2
-  // or
-  //        lumpedDiagDomMeasure <= rho*unlumpedDiagDomMeasure
-  //
-  // NOTE: THIS CODE assumes direct access to a row. See comments above concerning
-  //       ASSUME_DIRECT_ACCESS_TO_ROW
-  //
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ExperimentalLumping(const Matrix& A, Matrix& filteredA, double irho, double irho2) const {
-    using TST = typename Teuchos::ScalarTraits<SC>;
-    SC zero = TST::zero();
-    SC one  = TST::one();
-
-    ArrayView<const LO> inds;
-    ArrayView<const SC> vals;
-    ArrayView<const LO> finds;
-    ArrayView<SC> fvals;
-
-    SC PosOffSum, NegOffSum, PosOffDropSum,  NegOffDropSum;
-    SC diag, gamma, alpha;
-    LO NumPosKept, NumNegKept;
-
-    SC noLumpDdom;
-    SC numer,denom;
-    SC PosFilteredSum, NegFilteredSum;
-    SC Target;
-
-    SC rho  = as<Scalar>(irho);
-    SC rho2 = as<Scalar>(irho2);
-
-    for (LO row = 0; row < (LO) A.getRowMap()->getLocalNumElements(); row++) {
-        noLumpDdom = as<Scalar>(10000.0);  // only used if diagonal is zero
-                                           // the whole idea sort of breaks down
-                                           // when the diagonal is zero. In particular,
-                                           // the old diag dominance ratio is infinity
-                                           // ... so what do we want for the new ddom
-                                           // ratio. Do we want to allow the diagonal
-                                           // to go negative, just to have a better ddom
-                                           // ratio? This current choice essentially
-                                           // changes 'Target' to a large number
-                                           // meaning that we will allow the new
-                                           // ddom number to be fairly large (because
-                                           // the old one was infinity)
-
-        ArrayView<const SC> tvals;
-        A.getLocalRowView(row, inds, vals);
-        size_t nnz = inds.size();
-        if (nnz == 0) continue;
-        filteredA.getLocalRowView(row, finds, tvals);//assume 2 getLocalRowView()s
-                                                     // have things in same order
-        fvals = ArrayView<SC>(const_cast<SC*>(tvals.getRawPtr()), nnz);
-
-        LO diagIndex = -1, fdiagIndex = -1;
-
-        PosOffSum=zero; NegOffSum=zero; PosOffDropSum=zero; NegOffDropSum=zero;
-        diag=zero; NumPosKept=0; NumNegKept=0;
-
-        // first record diagonal, offdiagonal sums and off diag dropped sums
-        for (size_t j = 0; j < nnz; j++) {
-          if (inds[j] == row) {
-            diagIndex = j;
-            diag = vals[j];
+          // FIXME: I'm assuming vertex2AggId is only length of the rowmap, so
+          // we won'd do secondary dropping on off-processor neighbors
+          if (is_good && indsA[l] < (LO)numRows) {
+            int agg = vertex2AggId[col_node];
+            if (std::binary_search(reallyBadAggNeighbors.begin(), reallyBadAggNeighbors.end(), agg))
+              is_good = false;
           }
-          else { // offdiagonal
-            if (TST::real(vals[j]) > TST::real(zero) ) PosOffSum += vals[j];
-            else                                       NegOffSum += vals[j];
+
+          if (is_good) {
+            vals[index_start + l] = valsA[l];
+          } else {
+            if (!filter[col_node])
+              numOldDrops++;
+            else
+              numNewDrops++;
+            diagExtra[row] += valsA[l];
+            vals[index_start + l]                   = ZERO;
+            vals_dropped_indicator[index_start + l] = true;
           }
-        }
-        PosOffDropSum = PosOffSum;
-        NegOffDropSum = NegOffSum;
-        NumPosKept = 0;
-        NumNegKept = 0;
-        LO j = 0;
-        for (size_t jj = 0; jj < (size_t) finds.size(); jj++) {
-          while( inds[j] != finds[jj] ) j++;    // assumes that finds is in the same order as
-                                                // inds ... but perhaps has some entries missing
-          if (finds[jj] == row) fdiagIndex = jj;
-          else {
-            if (TST::real(vals[j]) > TST::real(zero) ) {
-              PosOffDropSum -=  fvals[jj];
-              if (TST::real(fvals[jj]) != TST::real(zero) ) NumPosKept++;
+        }  // end for l "indsA.size()" loop
+
+      }  // end m "blkSize" loop
+
+      // Clear filtering array
+      for (size_t j = 0; j < as<size_t>(indsG.size()); j++)
+        filter[indsG[j]] = false;
+
+    }  // end k loop over number of nodes in this agg
+  }    // end i loop over numAggs
+
+  if (!use_spread_lumping) {
+    // Now do the diagonal modifications in one, final pass
+    for (LO row = 0; row < (LO)numRows; row++) {
+      if (diagIndex[row] != INVALID) {
+        size_t index_start       = rowptr[row];
+        size_t diagIndexInMatrix = index_start + diagIndex[row];
+        //        printf("diag_vals pre update =  %8.2e\n", vals[diagIndex] );
+        vals[diagIndexInMatrix] += diagExtra[row];
+        SC A_rowsum = ZERO, A_absrowsum = ZERO, F_rowsum = ZERO;
+
+        if ((dirichletThresh >= 0.0 && TST::real(vals[diagIndexInMatrix]) <= dirichletThresh) || TST::real(vals[diagIndexInMatrix]) == ZERO) {
+          if (MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING > 0) {
+            A.getLocalRowView(row, indsA, valsA);
+            //            SC diagA = valsA[diagIndex[row]];
+            //            printf("WARNING: row %d (diagIndex=%d) diag(Afiltered) = %8.2e diag(A)=%8.2e numInds = %d\n",row,diagIndex[row],vals[diagIndexInMatrix],diagA,(LO)indsA.size());
+
+            for (LO l = 0; l < (LO)indsA.size(); l++) {
+              A_rowsum += valsA[l];
+              A_absrowsum += std::abs(valsA[l]);
             }
-            else {
-              NegOffDropSum -=  fvals[jj];
-              if (TST::real(fvals[jj]) != TST::real(zero) ) NumNegKept++;
+            for (LO l = 0; l < (LO)indsA.size(); l++)
+              F_rowsum += vals[index_start + l];
+            //                printf("       : A rowsum = %8.2e |A| rowsum = %8.2e rowsum = %8.2e\n",A_rowsum,A_absrowsum,F_rowsum);
+            if (MUELU_FILTEREDAFACTORY_LOTS_OF_PRINTING > 1) {
+              //            printf("        Avals =");
+              //            for(LO l = 0; l < (LO)indsA.size(); l++)
+              //              printf("%d(%8.2e)[%d] ",(LO)indsA[l],valsA[l],(LO)l);
+              //            printf("\n");
+              //            printf("        Fvals =");
+              //            for(LO l = 0; l < (LO)indsA.size(); l++)
+              //              if(vals[index_start+l] != ZERO)
+              //                printf("%d(%8.2e)[%d] ",(LO)indsA[l],vals[index_start+l],(LO)l);
             }
           }
+          // Don't know what to do, so blitz the row and dump a one on the diagonal
+          for (size_t l = rowptr[row]; l < rowptr[row + 1]; l++) {
+            vals[l] = ZERO;
+          }
+          vals[diagIndexInMatrix] = TST::one();
+          numFixedDiags++;
         }
+      } else {
+        GetOStream(Runtime0) << "WARNING: Row " << row << " has no diagonal " << std::endl;
+      }
+    } /*end row "numRows" loop"*/
+  }
 
-        // measure of diagonal dominance if no lumping is done.
-        if (TST::magnitude(diag) != TST::magnitude(zero)  )
-          noLumpDdom = (PosOffSum - NegOffSum)/diag;
-
-        // Target is an acceptable diagonal dominance ratio
-        // which should really be larger than 1
-
-        Target = rho*noLumpDdom;
-        if (TST::magnitude(Target) <= TST::magnitude(rho)) Target = rho2;
+  // Copy all the goop out
+  for (LO row = 0; row < (LO)numRows; row++) {
+    filteredA.replaceLocalValues(row, inds(rowptr[row], rowptr[row + 1] - rowptr[row]), vals(rowptr[row], rowptr[row + 1] - rowptr[row]));
+  }
+  if (use_spread_lumping) ExperimentalLumping(A, filteredA, DdomAllowGrowthRate, DdomCap);
 
-        PosFilteredSum = PosOffSum - PosOffDropSum;
-        NegFilteredSum = NegOffSum - NegOffDropSum;
-           // Note: PosNotFilterdSum is not equal to the sum of the
-           // positive entries after lumping. It just reflects the
-           // pos offdiag sum of the filtered matrix before lumping
-           // and does not account for negative dropped terms lumped
-           // to the positive kept terms.
+  size_t g_newDrops = 0, g_oldDrops = 0, g_fixedDiags = 0;
 
-        // dropped positive offdiags always go to the diagonal as these
-        // always improve diagonal dominance.
+  MueLu_sumAll(A.getRowMap()->getComm(), numNewDrops, g_newDrops);
+  MueLu_sumAll(A.getRowMap()->getComm(), numOldDrops, g_oldDrops);
+  MueLu_sumAll(A.getRowMap()->getComm(), numFixedDiags, g_fixedDiags);
+  GetOStream(Runtime0) << "Filtering out " << g_newDrops << " edges, in addition to the " << g_oldDrops << " edges dropped earlier" << std::endl;
+  GetOStream(Runtime0) << "Fixing " << g_fixedDiags << " zero diagonal values" << std::endl;
+}
 
-        diag += PosOffDropSum;
+// fancy lumping trying to not just move everything to the diagonal but to also consider moving
+// some lumping to the kept off-diagonals. We basically aim to not increase the diagonal
+// dominance in a row. In particular, the goal is that row i satisfies
+//
+//        lumpedDiagDomMeasure_i   <= rho2
+// or
+//        lumpedDiagDomMeasure <= rho*unlumpedDiagDomMeasure
+//
+// NOTE: THIS CODE assumes direct access to a row. See comments above concerning
+//       ASSUME_DIRECT_ACCESS_TO_ROW
+//
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FilteredAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ExperimentalLumping(const Matrix& A, Matrix& filteredA, double irho, double irho2) const {
+  using TST = typename Teuchos::ScalarTraits<SC>;
+  SC zero   = TST::zero();
+  SC one    = TST::one();
+
+  ArrayView<const LO> inds;
+  ArrayView<const SC> vals;
+  ArrayView<const LO> finds;
+  ArrayView<SC> fvals;
+
+  SC PosOffSum, NegOffSum, PosOffDropSum, NegOffDropSum;
+  SC diag, gamma, alpha;
+  LO NumPosKept, NumNegKept;
+
+  SC noLumpDdom;
+  SC numer, denom;
+  SC PosFilteredSum, NegFilteredSum;
+  SC Target;
+
+  SC rho  = as<Scalar>(irho);
+  SC rho2 = as<Scalar>(irho2);
+
+  for (LO row = 0; row < (LO)A.getRowMap()->getLocalNumElements(); row++) {
+    noLumpDdom = as<Scalar>(10000.0);  // only used if diagonal is zero
+                                       // the whole idea sort of breaks down
+                                       // when the diagonal is zero. In particular,
+                                       // the old diag dominance ratio is infinity
+                                       // ... so what do we want for the new ddom
+                                       // ratio. Do we want to allow the diagonal
+                                       // to go negative, just to have a better ddom
+                                       // ratio? This current choice essentially
+                                       // changes 'Target' to a large number
+                                       // meaning that we will allow the new
+                                       // ddom number to be fairly large (because
+                                       // the old one was infinity)
+
+    ArrayView<const SC> tvals;
+    A.getLocalRowView(row, inds, vals);
+    size_t nnz = inds.size();
+    if (nnz == 0) continue;
+    filteredA.getLocalRowView(row, finds, tvals);  // assume 2 getLocalRowView()s
+                                                   //  have things in same order
+    fvals = ArrayView<SC>(const_cast<SC*>(tvals.getRawPtr()), nnz);
+
+    LO diagIndex = -1, fdiagIndex = -1;
+
+    PosOffSum     = zero;
+    NegOffSum     = zero;
+    PosOffDropSum = zero;
+    NegOffDropSum = zero;
+    diag          = zero;
+    NumPosKept    = 0;
+    NumNegKept    = 0;
+
+    // first record diagonal, offdiagonal sums and off diag dropped sums
+    for (size_t j = 0; j < nnz; j++) {
+      if (inds[j] == row) {
+        diagIndex = j;
+        diag      = vals[j];
+      } else {  // offdiagonal
+        if (TST::real(vals[j]) > TST::real(zero))
+          PosOffSum += vals[j];
+        else
+          NegOffSum += vals[j];
+      }
+    }
+    PosOffDropSum = PosOffSum;
+    NegOffDropSum = NegOffSum;
+    NumPosKept    = 0;
+    NumNegKept    = 0;
+    LO j          = 0;
+    for (size_t jj = 0; jj < (size_t)finds.size(); jj++) {
+      while (inds[j] != finds[jj]) j++;  // assumes that finds is in the same order as
+                                         // inds ... but perhaps has some entries missing
+      if (finds[jj] == row)
+        fdiagIndex = jj;
+      else {
+        if (TST::real(vals[j]) > TST::real(zero)) {
+          PosOffDropSum -= fvals[jj];
+          if (TST::real(fvals[jj]) != TST::real(zero)) NumPosKept++;
+        } else {
+          NegOffDropSum -= fvals[jj];
+          if (TST::real(fvals[jj]) != TST::real(zero)) NumNegKept++;
+        }
+      }
+    }
 
-        // now lets work on lumping dropped negative offdiags
-        gamma = -NegOffDropSum - PosFilteredSum;
+    // measure of diagonal dominance if no lumping is done.
+    if (TST::magnitude(diag) != TST::magnitude(zero))
+      noLumpDdom = (PosOffSum - NegOffSum) / diag;
+
+    // Target is an acceptable diagonal dominance ratio
+    // which should really be larger than 1
+
+    Target = rho * noLumpDdom;
+    if (TST::magnitude(Target) <= TST::magnitude(rho)) Target = rho2;
+
+    PosFilteredSum = PosOffSum - PosOffDropSum;
+    NegFilteredSum = NegOffSum - NegOffDropSum;
+    // Note: PosNotFilterdSum is not equal to the sum of the
+    // positive entries after lumping. It just reflects the
+    // pos offdiag sum of the filtered matrix before lumping
+    // and does not account for negative dropped terms lumped
+    // to the positive kept terms.
+
+    // dropped positive offdiags always go to the diagonal as these
+    // always improve diagonal dominance.
+
+    diag += PosOffDropSum;
+
+    // now lets work on lumping dropped negative offdiags
+    gamma = -NegOffDropSum - PosFilteredSum;
+
+    if (TST::real(gamma) < TST::real(zero)) {
+      // the total amount of negative dropping is less than PosFilteredSum,
+      // so we can distribute this dropping to pos offdiags. After lumping
+      // the sum of the pos offdiags is just -gamma so we just assign pos
+      // offdiags proportional to vals[j]/PosFilteredSum
+      // Note: in this case the diagonal is not changed as all lumping
+      //       occurs to the pos offdiags
+
+      if (fdiagIndex != -1) fvals[fdiagIndex] = diag;
+      j = 0;
+      for (LO jj = 0; jj < (LO)finds.size(); jj++) {
+        while (inds[j] != finds[jj]) j++;  // assumes that finds is in the same order as
+                                           // inds ... but perhaps has some entries missing
+        if ((j != diagIndex) && (TST::real(vals[j]) > TST::real(zero)) && (TST::magnitude(fvals[jj]) != TST::magnitude(zero)))
+          fvals[jj] = -gamma * (vals[j] / PosFilteredSum);
+      }
+    } else {
+      // So there are more negative values that need lumping than kept
+      // positive offdiags. Meaning there is enough negative lumping to
+      // completely clear out all pos offdiags. If we lump all negs
+      // to pos off diags, we'd actually change them to negative. We
+      // only do this if we are desperate. Otherwise, we'll clear out
+      // all the positive kept offdiags and try to lump the rest
+      // somewhere else.  We defer the clearing of pos off diags
+      // to see first if we are going to be desperate.
+
+      bool flipPosOffDiagsToNeg = false;
+
+      // Even if we lumped by zeroing positive offdiags, we are still
+      // going to have more lumping to distribute to either
+      //     1) the diagonal
+      //     2) the kept negative offdiags
+      //     3) the kept positive offdiags (desperate)
+
+      // Let's first considering lumping the remaining neg offdiag stuff
+      // to the diagonal ... if this does not increase the diagonal
+      // dominance ratio too much (given  by rho).
+
+      if ((TST::real(diag) > TST::real(gamma)) &&
+          (TST::real((-NegFilteredSum) / (diag - gamma)) <= TST::real(Target))) {
+        // 1st if term above insures that resulting diagonal (=diag-gamma)
+        // is positive. . The left side of 2nd term is the diagonal dominance
+        // if we lump the remaining stuff (gamma) to the diagonal. Recall,
+        // that now there are no positive off-diags so the sum(abs(offdiags))
+        // is just the negative of NegFilteredSum
+
+        if (fdiagIndex != -1) fvals[fdiagIndex] = diag - gamma;
+      } else if (NumNegKept > 0) {
+        // need to do some lumping to neg offdiags to avoid a large
+        // increase in diagonal dominance. We first compute alpha
+        // which measures how much gamma should go to the
+        // negative offdiags. The rest will go to the diagonal
+
+        numer = -NegFilteredSum - Target * (diag - gamma);
+        denom = gamma * (Target - TST::one());
+
+        // make sure that alpha is between 0 and 1 ... and that it doesn't
+        // result in a sign flip
+        //    Note: when alpha is set to 1, then the diagonal is not modified
+        //          and the negative offdiags just get shifted from those
+        //          removed and those kept, meaning that the digaonal dominance
+        //          should be the same as before
+        //
+        //          can alpha be negative? It looks like denom should always
+        //          be positive. The 'if' statement above
+        //          Normally, diag-gamma should also be positive (but if it
+        //          is negative then numer is guaranteed to be positve).
+        //          look at the 'if' above,
+        //                if (( TST::real(diag) > TST::real(gamma)) &&
+        //                ( TST::real((-NegFilteredSum)/(diag - gamma)) <= TST::real(Target))) {
+        //
+        //          Should guarantee that numer is positive. This is obvious when
+        //          the second condition is false. When it is the first condition that
+        //          is false, it follows that the two indiviudal terms in the numer
+        //          formula must be positive.
+
+        if (TST::magnitude(denom) < TST::magnitude(numer))
+          alpha = TST::one();
+        else
+          alpha = numer / denom;
+        if (TST::real(alpha) < TST::real(zero)) alpha = zero;
+        if (TST::real(diag) < TST::real((one - alpha) * gamma)) alpha = TST::one();
+
+        // first change the diagonal
+
+        if (fdiagIndex != -1) fvals[fdiagIndex] = diag - (one - alpha) * gamma;
+
+        // after lumping the sum of neg offdiags will be NegFilteredSum
+        // + alpha*gamma. That is the remaining negative entries altered
+        // by the percent (=alpha) of stuff (=gamma) that needs to be
+        // lumped after taking into account lumping to pos offdiags
+
+        // Do this by assigning a fraction of NegFilteredSum+alpha*gamma
+        //  proportional to vals[j]/NegFilteredSum
+
+        SC temp = (NegFilteredSum + alpha * gamma) / NegFilteredSum;
+        j       = 0;
+        for (LO jj = 0; jj < (LO)finds.size(); jj++) {
+          while (inds[j] != finds[jj]) j++;  // assumes that finds is in the same order as
+                                             // inds ... but perhaps has some entries missing
+          if ((jj != fdiagIndex) && (TST::magnitude(fvals[jj]) != TST::magnitude(zero)) &&
+              (TST::real(vals[j]) < TST::real(zero)))
+            fvals[jj] = temp * vals[j];
+        }
+      } else {  // desperate case
+        // So we don't have any kept negative offdiags  ...
 
-        if (TST::real(gamma) < TST::real(zero) ) {
-          // the total amount of negative dropping is less than PosFilteredSum,
-          // so we can distribute this dropping to pos offdiags. After lumping
-          // the sum of the pos offdiags is just -gamma so we just assign pos
-          // offdiags proportional to vals[j]/PosFilteredSum
-          // Note: in this case the diagonal is not changed as all lumping
-          //       occurs to the pos offdiags
+        if (NumPosKept > 0) {
+          // luckily we can push this stuff to the pos offdiags
+          // which now makes them negative
+          flipPosOffDiagsToNeg = true;
 
-          if (fdiagIndex != -1) fvals[fdiagIndex] = diag;
           j = 0;
-          for(LO jj = 0; jj < (LO)finds.size(); jj++)  {
-            while( inds[j] != finds[jj] ) j++;   // assumes that finds is in the same order as
-                                                  // inds ... but perhaps has some entries missing
-            if ((j != diagIndex)&&(TST::real(vals[j]) > TST::real(zero) ) && (TST::magnitude(fvals[jj]) != TST::magnitude(zero)))
-              fvals[jj] =  -gamma*(vals[j]/PosFilteredSum);
-
+          for (LO jj = 0; jj < (LO)finds.size(); jj++) {
+            while (inds[j] != finds[jj]) j++;  // assumes that finds is in the same order as
+                                               // inds ... but perhaps has some entries missing
+            if ((j != diagIndex) && (TST::magnitude(fvals[jj]) != TST::magnitude(zero)) &&
+                (TST::real(vals[j]) > TST::real(zero)))
+              fvals[jj] = -gamma / ((SC)NumPosKept);
           }
         }
-        else {
-          // So there are more negative values that need lumping than kept
-          // positive offdiags. Meaning there is enough negative lumping to
-          // completely clear out all pos offdiags. If we lump all negs
-          // to pos off diags, we'd actually change them to negative. We
-          // only do this if we are desperate. Otherwise, we'll clear out
-          // all the positive kept offdiags and try to lump the rest
-          // somewhere else.  We defer the clearing of pos off diags
-          // to see first if we are going to be desperate.
-
-          bool flipPosOffDiagsToNeg = false;
-
-          // Even if we lumped by zeroing positive offdiags, we are still
-          // going to have more lumping to distribute to either
-          //     1) the diagonal
-          //     2) the kept negative offdiags
-          //     3) the kept positive offdiags (desperate)
-
-          // Let's first considering lumping the remaining neg offdiag stuff
-          // to the diagonal ... if this does not increase the diagonal
-          // dominance ratio too much (given  by rho).
-
-          if (( TST::real(diag) > TST::real(gamma)) &&
-              ( TST::real((-NegFilteredSum)/(diag - gamma)) <= TST::real(Target))) {
-            // 1st if term above insures that resulting diagonal (=diag-gamma)
-            // is positive. . The left side of 2nd term is the diagonal dominance
-            // if we lump the remaining stuff (gamma) to the diagonal. Recall,
-            // that now there are no positive off-diags so the sum(abs(offdiags))
-            // is just the negative of NegFilteredSum
-
-            if (fdiagIndex != -1) fvals[fdiagIndex] = diag - gamma;
-          }
-          else if (NumNegKept > 0) {
-            // need to do some lumping to neg offdiags to avoid a large
-            // increase in diagonal dominance. We first compute alpha
-            // which measures how much gamma should go to the
-            // negative offdiags. The rest will go to the diagonal
-
-            numer = -NegFilteredSum -  Target*(diag-gamma);
-            denom = gamma*(Target - TST::one());
-
-            // make sure that alpha is between 0 and 1 ... and that it doesn't
-            // result in a sign flip
-            //    Note: when alpha is set to 1, then the diagonal is not modified
-            //          and the negative offdiags just get shifted from those
-            //          removed and those kept, meaning that the digaonal dominance
-            //          should be the same as before
-            //
-            //          can alpha be negative? It looks like denom should always
-            //          be positive. The 'if' statement above
-            //          Normally, diag-gamma should also be positive (but if it
-            //          is negative then numer is guaranteed to be positve).
-            //          look at the 'if' above,
-            //                if (( TST::real(diag) > TST::real(gamma)) &&
-            //                ( TST::real((-NegFilteredSum)/(diag - gamma)) <= TST::real(Target))) {
-            //
-            //          Should guarantee that numer is positive. This is obvious when
-            //          the second condition is false. When it is the first condition that
-            //          is false, it follows that the two indiviudal terms in the numer
-            //          formula must be positive.
-
-            if ( TST::magnitude(denom) < TST::magnitude(numer) ) alpha = TST::one();
-            else alpha = numer/denom;
-            if ( TST::real(alpha) < TST::real(zero)) alpha = zero;
-            if ( TST::real(diag) < TST::real((one-alpha)*gamma) ) alpha =  TST::one();
-
-            // first change the diagonal
-
-            if (fdiagIndex != -1) fvals[fdiagIndex] = diag - (one-alpha)*gamma;
-
-            // after lumping the sum of neg offdiags will be NegFilteredSum
-            // + alpha*gamma. That is the remaining negative entries altered
-            // by the percent (=alpha) of stuff (=gamma) that needs to be
-            // lumped after taking into account lumping to pos offdiags
-
-            // Do this by assigning a fraction of NegFilteredSum+alpha*gamma
-            //  proportional to vals[j]/NegFilteredSum
-
-            SC temp = (NegFilteredSum+alpha*gamma)/NegFilteredSum;
-            j = 0;
-            for(LO jj = 0; jj < (LO)finds.size(); jj++)  {
-              while( inds[j] != finds[jj] ) j++;   // assumes that finds is in the same order as
-                                                    // inds ... but perhaps has some entries missing
-              if  ( (jj != fdiagIndex)&&(TST::magnitude(fvals[jj]) != TST::magnitude(zero) ) &&
-                    ( TST::real(vals[j]) < TST::real(zero) ) )
-                fvals[jj] = temp*vals[j];
-            }
-          }
-          else { // desperate case
-            // So we don't have any kept negative offdiags  ...
-
-            if (NumPosKept > 0) {
-              // luckily we can push this stuff to the pos offdiags
-              // which now makes them negative
-              flipPosOffDiagsToNeg = true;
-
-              j = 0;
-              for(LO jj = 0; jj < (LO)finds.size(); jj++)  {
-                while( inds[j] != finds[jj] ) j++;    // assumes that finds is in the same order as
-                                                      // inds ... but perhaps has some entries missing
-                if  ( (j != diagIndex)&&(TST::magnitude(fvals[jj]) != TST::magnitude(zero) ) &&
-                      (TST::real(vals[j]) > TST::real(zero) ))
-                  fvals[jj] = -gamma/( (SC) NumPosKept);
-              }
-            }
-            // else abandon rowsum preservation and do nothing
-
-          }
-          if (!flipPosOffDiagsToNeg) { // not desperate so we now zero out
-                                       // all pos terms including some
-                                       // not originally filtered
-                                       // but zeroed due to lumping
-              j = 0;
-              for(LO jj = 0; jj < (LO)finds.size(); jj++) {
-                while( inds[j] != finds[jj] ) j++;    // assumes that finds is in the same order as
-                                                      // inds ... but perhaps has some entries missing
-                if ((jj != fdiagIndex)&& (TST::real(vals[j]) > TST::real(zero))) fvals[jj] = zero;
-              }
-          }
-        } // positive gamma else
-
-    } //loop over all rows
-  }
+        // else abandon rowsum preservation and do nothing
+      }
+      if (!flipPosOffDiagsToNeg) {  // not desperate so we now zero out
+                                    // all pos terms including some
+                                    // not originally filtered
+                                    // but zeroed due to lumping
+        j = 0;
+        for (LO jj = 0; jj < (LO)finds.size(); jj++) {
+          while (inds[j] != finds[jj]) j++;  // assumes that finds is in the same order as
+                                             // inds ... but perhaps has some entries missing
+          if ((jj != fdiagIndex) && (TST::real(vals[j]) > TST::real(zero))) fvals[jj] = zero;
+        }
+      }
+    }  // positive gamma else
 
+  }  // loop over all rows
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_FILTEREDAFACTORY_DEF_HPP
+#endif  // MUELU_FILTEREDAFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_decl.hpp
index 32b5389b0e1d..a24afb8724fe 100644
--- a/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_decl.hpp
@@ -47,7 +47,6 @@
 #ifndef PACKAGES_MUELU_SRC_MISC_MUELU_FINELEVELINPUTDATAFACTORY_DECL_HPP_
 #define PACKAGES_MUELU_SRC_MISC_MUELU_FINELEVELINPUTDATAFACTORY_DECL_HPP_
 
-
 #include <string>
 
 #include "MueLu_ConfigDefs.hpp"
@@ -62,63 +61,61 @@
 #include "MueLu_Graph_fwd.hpp"
 
 namespace MueLuTests {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FineLevelInputDataFactoryTester;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FineLevelInputDataFactoryTester;
 }
 
 namespace MueLu {
 
-  /*!
-    @class FineLevelInputData class.
-    @brief Factory for piping in input data from the finest level into the MueLu data dependency system
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class FineLevelInputDataFactory : public SingleLevelFactoryBase {
-    friend class MueLuTests::FineLevelInputDataFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+/*!
+  @class FineLevelInputData class.
+  @brief Factory for piping in input data from the finest level into the MueLu data dependency system
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FineLevelInputDataFactory : public SingleLevelFactoryBase {
+  friend class MueLuTests::FineLevelInputDataFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 #undef MUELU_FINELEVELINPUTDATAFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    FineLevelInputDataFactory() { }
+  FineLevelInputDataFactory() {}
 
-    //! Destructor.
-    virtual ~FineLevelInputDataFactory() { }
+  //! Destructor.
+  virtual ~FineLevelInputDataFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
-      */
-    void Build(Level& currentLevel) const;
-
-    //@}
-  private:
+  /*!
+    @brief Build method.
+    */
+  void Build(Level& currentLevel) const;
 
-    void test() const { std::cout << "TEST" << std::endl; }
+  //@}
+ private:
+  void test() const { std::cout << "TEST" << std::endl; }
 
-  }; //class FineLevelInputDataFactory
+};  // class FineLevelInputDataFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FINELEVELINPUTDATAFACTORY_SHORT
 
diff --git a/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_def.hpp b/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_def.hpp
index a54c59a6bf1f..590b66f974ff 100644
--- a/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_FineLevelInputDataFactory_def.hpp
@@ -55,131 +55,120 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    // Variable name (e.g. A or P or Coordinates)
-    validParamList->set< std::string >("Variable", std::string("A"), "Variable name on all coarse levels (except the finest level).");
+  // Variable name (e.g. A or P or Coordinates)
+  validParamList->set<std::string>("Variable", std::string("A"), "Variable name on all coarse levels (except the finest level).");
 
-    // Names of generating factories (on finest level and coarse levels)
-    validParamList->set< RCP<const FactoryBase> >("Fine level factory", Teuchos::null, "Generating factory of the fine level variable");
-    validParamList->set< RCP<const FactoryBase> >("Coarse level factory", Teuchos::null, "Generating factory for data on all coarse levels (except the finest)");
+  // Names of generating factories (on finest level and coarse levels)
+  validParamList->set<RCP<const FactoryBase> >("Fine level factory", Teuchos::null, "Generating factory of the fine level variable");
+  validParamList->set<RCP<const FactoryBase> >("Coarse level factory", Teuchos::null, "Generating factory for data on all coarse levels (except the finest)");
 
-    // Type of variable (see source code for a complete list of all available types)
-    validParamList->set<std::string> ("Variable type", std::string("Matrix"), "Type of variable");
+  // Type of variable (see source code for a complete list of all available types)
+  validParamList->set<std::string>("Variable type", std::string("Matrix"), "Type of variable");
 
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-
-    const ParameterList & pL = GetParameterList();
-
-    std::string variableName = "";
-    if(pL.isParameter("Variable"))
-      variableName = pL.get<std::string>("Variable");
+  return validParamList;
+}
 
-    std::string factoryName  = "NoFactory";
-    if (currentLevel.GetLevelID() == 0) {
-      factoryName = "Fine level factory";
-    } else {
-      factoryName = "Coarse level factory";
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
 
-    TEUCHOS_TEST_FOR_EXCEPTION(variableName == "", MueLu::Exceptions::RuntimeError, "FineLevelInputDataFactory: no variable name provided. Please set \'Variable\' parameter in your input deck.");
+  std::string variableName = "";
+  if (pL.isParameter("Variable"))
+    variableName = pL.get<std::string>("Variable");
 
-    // data must be specified in factory! (not in factory manager)
-    RCP<const FactoryBase> fact = GetFactory(factoryName);
-    currentLevel.DeclareInput(variableName, fact.get(), this);
+  std::string factoryName = "NoFactory";
+  if (currentLevel.GetLevelID() == 0) {
+    factoryName = "Fine level factory";
+  } else {
+    factoryName = "Coarse level factory";
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "InputUserData", currentLevel);
+  TEUCHOS_TEST_FOR_EXCEPTION(variableName == "", MueLu::Exceptions::RuntimeError, "FineLevelInputDataFactory: no variable name provided. Please set \'Variable\' parameter in your input deck.");
 
-    const ParameterList& pL = GetParameterList();
+  // data must be specified in factory! (not in factory manager)
+  RCP<const FactoryBase> fact = GetFactory(factoryName);
+  currentLevel.DeclareInput(variableName, fact.get(), this);
+}
 
-    std::string variableName       = "";
-    if (pL.isParameter("Variable"))
-      variableName = pL.get<std::string>("Variable");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FineLevelInputDataFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "InputUserData", currentLevel);
 
-    std::string variableType = "";
-    if(pL.isParameter("Variable type"))
-      variableType = pL.get<std::string>("Variable type");
+  const ParameterList& pL = GetParameterList();
 
-    std::string factoryName        = "NoFactory";
-    if (currentLevel.GetLevelID() == 0) {
-      factoryName        = "Fine level factory";
-    } else {
-      factoryName        = "Coarse level factory";
-    }
-    RCP<const FactoryBase> fact = GetFactory(factoryName);
+  std::string variableName = "";
+  if (pL.isParameter("Variable"))
+    variableName = pL.get<std::string>("Variable");
 
-    GetOStream(Debug) << "Use " << variableName << " of type " << variableType << " from " << factoryName << "(" << fact.get() << ")" << std::endl;
+  std::string variableType = "";
+  if (pL.isParameter("Variable type"))
+    variableType = pL.get<std::string>("Variable type");
 
-    // check data type
-    //std::string strType = currentLevel.GetTypeName(variableName, fact.get());
-    if        (variableType == "int") {
-      int data = currentLevel.Get<int>(variableName, fact.get());
+  std::string factoryName = "NoFactory";
+  if (currentLevel.GetLevelID() == 0) {
+    factoryName = "Fine level factory";
+  } else {
+    factoryName = "Coarse level factory";
+  }
+  RCP<const FactoryBase> fact = GetFactory(factoryName);
+
+  GetOStream(Debug) << "Use " << variableName << " of type " << variableType << " from " << factoryName << "(" << fact.get() << ")" << std::endl;
+
+  // check data type
+  // std::string strType = currentLevel.GetTypeName(variableName, fact.get());
+  if (variableType == "int") {
+    int data = currentLevel.Get<int>(variableName, fact.get());
+    Set(currentLevel, variableName, data);
+  } else if (variableType == "double") {
+    double data = currentLevel.Get<double>(variableName, fact.get());
+    Set(currentLevel, variableName, data);
+  } else if (variableType == "string") {
+    std::string data = currentLevel.Get<std::string>(variableName, fact.get());
+    Set(currentLevel, variableName, data);
+  } else {
+    size_t npos = std::string::npos;
+
+    if (variableType.find("Aggregates") != npos) {
+      RCP<Aggregates> data = currentLevel.Get<RCP<Aggregates> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("Graph") != npos) {
+      RCP<Graph> data = currentLevel.Get<RCP<Graph> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("SmootherBase") != npos) {
+      RCP<SmootherBase> data = currentLevel.Get<RCP<SmootherBase> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("SmootherPrototype") != npos) {
+      RCP<SmootherPrototype> data = currentLevel.Get<RCP<SmootherPrototype> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("Export") != npos) {
+      RCP<Export> data = currentLevel.Get<RCP<Export> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("Import") != npos) {
+      RCP<Import> data = currentLevel.Get<RCP<Import> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("Map") != npos) {
+      RCP<Map> data = currentLevel.Get<RCP<Map> >(variableName, fact.get());
+      Set(currentLevel, variableName, data);
+    } else if (variableType.find("Matrix") != npos) {
+      RCP<Matrix> data = currentLevel.Get<RCP<Matrix> >(variableName, fact.get());
       Set(currentLevel, variableName, data);
-    } else if (variableType == "double") {
-      double data = currentLevel.Get<double>(variableName, fact.get());
+    } else if (variableType.find("MultiVector") != npos) {
+      RCP<MultiVector> data = currentLevel.Get<RCP<MultiVector> >(variableName, fact.get());
       Set(currentLevel, variableName, data);
-    } else if (variableType == "string") {
-      std::string data = currentLevel.Get<std::string>(variableName, fact.get());
+    } else if (variableType.find("Operator") != npos) {
+      RCP<Operator> data = currentLevel.Get<RCP<Operator> >(variableName, fact.get());
       Set(currentLevel, variableName, data);
     } else {
-      size_t npos = std::string::npos;
-
-      if      (variableType.find("Aggregates") != npos) {
-        RCP<Aggregates> data = currentLevel.Get<RCP<Aggregates> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Graph") != npos) {
-        RCP<Graph> data = currentLevel.Get<RCP<Graph> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("SmootherBase") != npos) {
-        RCP<SmootherBase> data = currentLevel.Get<RCP<SmootherBase> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("SmootherPrototype") != npos) {
-        RCP<SmootherPrototype> data = currentLevel.Get<RCP<SmootherPrototype> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Export") != npos) {
-        RCP<Export> data = currentLevel.Get<RCP<Export> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Import") != npos) {
-        RCP<Import> data = currentLevel.Get<RCP<Import> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Map") != npos) {
-        RCP<Map> data = currentLevel.Get<RCP<Map> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Matrix") != npos) {
-        RCP<Matrix> data = currentLevel.Get<RCP<Matrix> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("MultiVector") != npos) {
-        RCP<MultiVector> data = currentLevel.Get<RCP<MultiVector> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else if (variableType.find("Operator") != npos) {
-        RCP<Operator> data = currentLevel.Get<RCP<Operator> >(variableName, fact.get());
-        Set(currentLevel, variableName, data);
-      }
-      else {
-        // TAW: is this working with empty procs?
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "FineLevelInputDataFactory: cannot detect type of variable " << variableName << " generated by " << fact.get() << ". User provided type " << variableType );
-      }
+      // TAW: is this working with empty procs?
+      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "FineLevelInputDataFactory: cannot detect type of variable " << variableName << " generated by " << fact.get() << ". User provided type " << variableType);
     }
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_MISC_MUELU_FINELEVELINPUTDATAFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_decl.hpp
index 8e28f8fc1c43..55c59db3fc54 100644
--- a/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_decl.hpp
@@ -54,62 +54,60 @@
 
 #include "MueLu_InitialBlockNumberFactory_fwd.hpp"
 
-
 namespace MueLu {
 
 /*!
   @class InitialBlockNumberFactory class.
   @brief Class for generating an initial LocalOrdinal-type BlockNumber vector, based on an input paraemter for interleaved dofs.
-  
+
 
 */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class InitialBlockNumberFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class InitialBlockNumberFactory : public SingleLevelFactoryBase {
 #undef MUELU_INITIALBLOCKNUMBERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-
-    /*! @brief Constructor.
-     */
-    InitialBlockNumberFactory() { }
+ public:
+  //! @name Constructors/Destructors.
 
-    //! Destructor.
-    virtual ~InitialBlockNumberFactory() { }
+  /*! @brief Constructor.
+   */
+  InitialBlockNumberFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~InitialBlockNumberFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
+  //@}
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //! @name Input
+  //@{
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level & currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-  private:
+  //@}
 
-  }; // class InitialBlockNumberFactory
+ private:
+};  // class InitialBlockNumberFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INITIALBLOCKNUMBERFACTORY_SHORT
-#endif // MUELU_INITIALBLOCKNUMBER_FACTORY_DECL_HPP
+#endif  // MUELU_INITIALBLOCKNUMBER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_def.hpp b/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_def.hpp
index 8cfe22245925..ae774c838857 100644
--- a/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_InitialBlockNumberFactory_def.hpp
@@ -59,40 +59,40 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> InitialBlockNumberFactory<Scalar,LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> InitialBlockNumberFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: block diagonal: interleaved blocksize");
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("aggregation: block diagonal: interleaved blocksize");
+#undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void InitialBlockNumberFactory<Scalar,LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void InitialBlockNumberFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void InitialBlockNumberFactory<Scalar,LocalOrdinal, GlobalOrdinal, Node>::Build(Level & currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    const ParameterList  & pL = GetParameterList();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void InitialBlockNumberFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  const ParameterList& pL = GetParameterList();
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-    LO blocksize = as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize"));
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+  LO blocksize  = as<LO>(pL.get<int>("aggregation: block diagonal: interleaved blocksize"));
 
-    GetOStream(Statistics1) << "Generating interleaved blocking with "<<blocksize<<" equations"<<std::endl;
-    RCP<LocalOrdinalVector> BlockNumber = LocalOrdinalVectorFactory::Build(A->getRowMap(),false);
-    Teuchos::ArrayRCP<LO> bn_data = BlockNumber->getDataNonConst(0);
-    for(LO i=0; i<(LO)A->getRowMap()->getLocalNumElements();i++)
-      bn_data[i] = i % blocksize;
-    
-    Set(currentLevel,"BlockNumber",BlockNumber);
-  }
+  GetOStream(Statistics1) << "Generating interleaved blocking with " << blocksize << " equations" << std::endl;
+  RCP<LocalOrdinalVector> BlockNumber = LocalOrdinalVectorFactory::Build(A->getRowMap(), false);
+  Teuchos::ArrayRCP<LO> bn_data       = BlockNumber->getDataNonConst(0);
+  for (LO i = 0; i < (LO)A->getRowMap()->getLocalNumElements(); i++)
+    bn_data[i] = i % blocksize;
 
-} // namespace MueLu
+  Set(currentLevel, "BlockNumber", BlockNumber);
+}
 
-#endif // MUELU_INITIALBLOCKNUMBER_FACTORY_DEF_HPP
+}  // namespace MueLu
+
+#endif  // MUELU_INITIALBLOCKNUMBER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_decl.hpp
index b992723be328..3bbcf7c19fe9 100644
--- a/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_decl.hpp
@@ -48,8 +48,7 @@
 
 #include "MueLu_SingleLevelFactoryBase.hpp"
 
-namespace MueLu
-{
+namespace MueLu {
 
 /*!
   @class InterfaceAggregationFactory class.
@@ -118,23 +117,21 @@ namespace MueLu
   | CoarseDualNodeID2PrimalNodeID | InterfaceAggregationFactory | Coarsened mapping of dual node IDs two primal node IDs.
 */
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-class InterfaceAggregationFactory : public SingleLevelFactoryBase
-{
+          class Node          = DefaultNode>
+class InterfaceAggregationFactory : public SingleLevelFactoryBase {
 #undef MUELU_INTERFACEAGGREGATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-public:
-
+ public:
   //! Input
   //@{
 
   RCP<const ParameterList> GetValidParameterList() const override;
 
-  void DeclareInput(Level &currentLevel) const override;
+  void DeclareInput(Level& currentLevel) const override;
 
   //@}
 
@@ -142,11 +139,11 @@ class InterfaceAggregationFactory : public SingleLevelFactoryBase
   //@{
 
   /*! @brief Build aggregates. */
-  void Build(Level &currentLevel) const override;
+  void Build(Level& currentLevel) const override;
 
   //@}
 
-private:
+ private:
   /*! @brief Build dual aggregates based on a given dual-to-primal node mapping
    *
    * @param[in] prefix Prefix for screen output
@@ -166,10 +163,9 @@ class InterfaceAggregationFactory : public SingleLevelFactoryBase
    * @param[in/out] currentLevel Level on which the aggregation needs to be performed
    */
   void BuildBasedOnPrimalInterfaceDofMap(const std::string& prefix, Level& currentLevel) const;
-
 };
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INTERFACEAGGREGATIONFACTORY_SHORT
 #endif /* MUELU_INTERFACEAGGREGATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_def.hpp b/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_def.hpp
index 4382cca2523b..ff05a39c0b3f 100644
--- a/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_InterfaceAggregationFactory_def.hpp
@@ -58,75 +58,64 @@
 
 #include "MueLu_InterfaceAggregationFactory_decl.hpp"
 
-namespace MueLu
-{
+namespace MueLu {
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<const ParameterList> InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-{
+RCP<const ParameterList> InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
   validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of A (matrix block related to dual DOFs)");
   validParamList->set<RCP<const FactoryBase>>("Aggregates", Teuchos::null, "Generating factory of the Aggregates (for block 0,0)");
 
   validParamList->set<std::string>("Dual/primal mapping strategy", "vague",
-      "Strategy to represent mapping between dual and primal quantities [node-based, dof-based]");
+                                   "Strategy to represent mapping between dual and primal quantities [node-based, dof-based]");
 
   validParamList->set<RCP<const FactoryBase>>("DualNodeID2PrimalNodeID", Teuchos::null,
-      "Generating factory of the DualNodeID2PrimalNodeID map as input data in a Moertel-compatible std::map<LO,LO> to map local IDs of dual nodes to local IDs of primal nodes");
+                                              "Generating factory of the DualNodeID2PrimalNodeID map as input data in a Moertel-compatible std::map<LO,LO> to map local IDs of dual nodes to local IDs of primal nodes");
   validParamList->set<LocalOrdinal>("number of DOFs per dual node", -Teuchos::ScalarTraits<LocalOrdinal>::one(),
-      "Number of DOFs per dual node");
+                                    "Number of DOFs per dual node");
 
   validParamList->set<RCP<const FactoryBase>>("Primal interface DOF map", Teuchos::null,
-      "Generating factory of the primal DOF row map of slave side of the coupling surface");
+                                              "Generating factory of the primal DOF row map of slave side of the coupling surface");
 
   return validParamList;
-} // GetValidParameterList()
+}  // GetValidParameterList()
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-{
-  Input(currentLevel, "A"); // matrix block of dual variables
+void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "A");  // matrix block of dual variables
   Input(currentLevel, "Aggregates");
 
   const ParameterList &pL = GetParameterList();
-  TEUCHOS_TEST_FOR_EXCEPTION(pL.get<std::string>("Dual/primal mapping strategy")=="vague", Exceptions::InvalidArgument,
-      "Strategy for dual/primal mapping not selected. Please select one of the available strategies.")
-  if (pL.get<std::string>("Dual/primal mapping strategy") == "node-based")
-  {
-    if (currentLevel.GetLevelID() == 0)
-    {
+  TEUCHOS_TEST_FOR_EXCEPTION(pL.get<std::string>("Dual/primal mapping strategy") == "vague", Exceptions::InvalidArgument,
+                             "Strategy for dual/primal mapping not selected. Please select one of the available strategies.")
+  if (pL.get<std::string>("Dual/primal mapping strategy") == "node-based") {
+    if (currentLevel.GetLevelID() == 0) {
       TEUCHOS_TEST_FOR_EXCEPTION(!currentLevel.IsAvailable("DualNodeID2PrimalNodeID", NoFactory::get()),
-          Exceptions::RuntimeError, "DualNodeID2PrimalNodeID was not provided by the user on level 0!");
+                                 Exceptions::RuntimeError, "DualNodeID2PrimalNodeID was not provided by the user on level 0!");
 
       currentLevel.DeclareInput("DualNodeID2PrimalNodeID", NoFactory::get(), this);
-    }
-    else
-    {
+    } else {
       Input(currentLevel, "DualNodeID2PrimalNodeID");
     }
-  }
-  else if (pL.get<std::string>("Dual/primal mapping strategy") == "dof-based")
-  {
+  } else if (pL.get<std::string>("Dual/primal mapping strategy") == "dof-based") {
     if (currentLevel.GetLevelID() == 0)
       currentLevel.DeclareInput("Primal interface DOF map", NoFactory::get(), this);
     else
       Input(currentLevel, "Primal interface DOF map");
-  }
-  else
+  } else
     TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::InvalidArgument, "Unknown strategy for dual/primal mapping.")
 
-} // DeclareInput
+}  // DeclareInput
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const
-{
+void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
   const std::string prefix = "MueLu::InterfaceAggregationFactory::Build: ";
 
   FactoryMonitor m(*this, "Build", currentLevel);
 
   // Call a specialized build routine based on the format of user-given input
-  const ParameterList &pL = GetParameterList();
+  const ParameterList &pL         = GetParameterList();
   const std::string parameterName = "Dual/primal mapping strategy";
   if (pL.get<std::string>(parameterName) == "node-based")
     BuildBasedOnNodeMapping(prefix, currentLevel);
@@ -134,23 +123,22 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
     BuildBasedOnPrimalInterfaceDofMap(prefix, currentLevel);
   else
     TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::InvalidArgument,
-        "MueLu::InterfaceAggregationFactory::Builld(): Unknown strategy for dual/primal mapping. Set a valid value for the parameter \"" << parameterName << "\".")
+                               "MueLu::InterfaceAggregationFactory::Builld(): Unknown strategy for dual/primal mapping. Set a valid value for the parameter \"" << parameterName << "\".")
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildBasedOnNodeMapping(const std::string& prefix,
-    Level &currentLevel) const
-{
+void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildBasedOnNodeMapping(const std::string &prefix,
+                                                                                                     Level &currentLevel) const {
   using Dual2Primal_type = std::map<LocalOrdinal, LocalOrdinal>;
 
   const ParameterList &pL = GetParameterList();
 
-  RCP<const Matrix> A = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<const Matrix> A                   = Get<RCP<Matrix>>(currentLevel, "A");
   const LocalOrdinal numDofsPerDualNode = pL.get<LocalOrdinal>("number of DOFs per dual node");
-  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerDualNode<Teuchos::ScalarTraits<LocalOrdinal>::one(), Exceptions::InvalidArgument,
-      "Number of dual DOFs per node < 0 (default value). Specify a valid \"number of DOFs per dual node\" in the parameter list for the InterfaceAggregationFactory.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerDualNode < Teuchos::ScalarTraits<LocalOrdinal>::one(), Exceptions::InvalidArgument,
+                             "Number of dual DOFs per node < 0 (default value). Specify a valid \"number of DOFs per dual node\" in the parameter list for the InterfaceAggregationFactory.");
 
-  RCP<const Aggregates> primalAggregates = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
+  RCP<const Aggregates> primalAggregates          = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
   ArrayRCP<const LocalOrdinal> primalVertex2AggId = primalAggregates->GetVertex2AggId()->getData(0);
 
   // Get the user-prescribed mapping of dual to primal node IDs
@@ -161,21 +149,20 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
     mapNodesDualToPrimal = Get<RCP<Dual2Primal_type>>(currentLevel, "DualNodeID2PrimalNodeID");
 
   RCP<const Map> operatorRangeMap = A->getRangeMap();
-  const size_t myRank = operatorRangeMap->getComm()->getRank();
+  const size_t myRank             = operatorRangeMap->getComm()->getRank();
 
   LocalOrdinal globalNumDualNodes = operatorRangeMap->getGlobalNumElements() / numDofsPerDualNode;
-  LocalOrdinal localNumDualNodes = operatorRangeMap->getLocalNumElements() / numDofsPerDualNode;
+  LocalOrdinal localNumDualNodes  = operatorRangeMap->getLocalNumElements() / numDofsPerDualNode;
 
   TEUCHOS_TEST_FOR_EXCEPTION(localNumDualNodes != Teuchos::as<LocalOrdinal>(mapNodesDualToPrimal->size()),
-      std::runtime_error, prefix << " MueLu requires the range map and the DualNodeID2PrimalNodeID map to be compatible.");
+                             std::runtime_error, prefix << " MueLu requires the range map and the DualNodeID2PrimalNodeID map to be compatible.");
 
   RCP<const Map> dualNodeMap = Teuchos::null;
   if (numDofsPerDualNode == 1)
     dualNodeMap = operatorRangeMap;
-  else
-  {
-    GlobalOrdinal indexBase = operatorRangeMap->getIndexBase();
-    auto comm = operatorRangeMap->getComm();
+  else {
+    GlobalOrdinal indexBase                = operatorRangeMap->getIndexBase();
+    auto comm                              = operatorRangeMap->getComm();
     std::vector<GlobalOrdinal> myDualNodes = {};
 
     for (size_t i = 0; i < operatorRangeMap->getLocalNumElements(); i += numDofsPerDualNode)
@@ -184,7 +171,7 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
     dualNodeMap = MapFactory::Build(operatorRangeMap->lib(), globalNumDualNodes, myDualNodes, indexBase, comm);
   }
   TEUCHOS_TEST_FOR_EXCEPTION(localNumDualNodes != Teuchos::as<LocalOrdinal>(dualNodeMap->getLocalNumElements()),
-      std::runtime_error, prefix << " Local number of dual nodes given by user is incompatible to the dual node map.");
+                             std::runtime_error, prefix << " Local number of dual nodes given by user is incompatible to the dual node map.");
 
   RCP<Aggregates> dualAggregates = rcp(new Aggregates(dualNodeMap));
   dualAggregates->setObjectLabel("InterfaceAggregation");
@@ -193,7 +180,7 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
   dualAggregates->AggregatesCrossProcessors(primalAggregates->AggregatesCrossProcessors());
 
   ArrayRCP<LocalOrdinal> dualVertex2AggId = dualAggregates->GetVertex2AggId()->getDataNonConst(0);
-  ArrayRCP<LocalOrdinal> dualProcWinner = dualAggregates->GetProcWinner()->getDataNonConst(0);
+  ArrayRCP<LocalOrdinal> dualProcWinner   = dualAggregates->GetProcWinner()->getDataNonConst(0);
 
   RCP<Dual2Primal_type> coarseMapNodesDualToPrimal = rcp(new Dual2Primal_type());
   RCP<Dual2Primal_type> coarseMapNodesPrimalToDual = rcp(new Dual2Primal_type());
@@ -205,10 +192,9 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
    * - assign dual nodes to dual aggregates
    * - recursively coarsen the dual-to-primal node mapping
    */
-  LocalOrdinal localPrimalNodeID = - Teuchos::ScalarTraits<LocalOrdinal>::one();
-  LocalOrdinal currentPrimalAggId = - Teuchos::ScalarTraits<LocalOrdinal>::one();
-  for (LocalOrdinal localDualNodeID = 0; localDualNodeID < localNumDualNodes; ++localDualNodeID)
-  {
+  LocalOrdinal localPrimalNodeID  = -Teuchos::ScalarTraits<LocalOrdinal>::one();
+  LocalOrdinal currentPrimalAggId = -Teuchos::ScalarTraits<LocalOrdinal>::one();
+  for (LocalOrdinal localDualNodeID = 0; localDualNodeID < localNumDualNodes; ++localDualNodeID) {
     // Get local ID of the primal node associated to the current dual node
     localPrimalNodeID = (*mapNodesDualToPrimal)[localDualNodeID];
 
@@ -217,17 +203,16 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
 
     // Test if the current primal aggregate has no associated dual aggregate, yet.
     // Create new dual aggregate, if necessary.
-    if (coarseMapNodesPrimalToDual->count(currentPrimalAggId) == 0)
-    {
+    if (coarseMapNodesPrimalToDual->count(currentPrimalAggId) == 0) {
       // Associate a new dual aggregate w/ the current primal aggregate
-      (*coarseMapNodesPrimalToDual)[currentPrimalAggId] = numLocalDualAggregates;
+      (*coarseMapNodesPrimalToDual)[currentPrimalAggId]     = numLocalDualAggregates;
       (*coarseMapNodesDualToPrimal)[numLocalDualAggregates] = currentPrimalAggId;
       ++numLocalDualAggregates;
     }
 
     // Fill the dual aggregate
     dualVertex2AggId[localDualNodeID] = (*coarseMapNodesPrimalToDual)[currentPrimalAggId];
-    dualProcWinner[localDualNodeID] = myRank;
+    dualProcWinner[localDualNodeID]   = myRank;
   }
 
   // Store dual aggregeate data as well as coarsening information
@@ -235,25 +220,24 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
   Set(currentLevel, "Aggregates", dualAggregates);
   Set(currentLevel, "CoarseDualNodeID2PrimalNodeID", coarseMapNodesDualToPrimal);
   GetOStream(Statistics1) << dualAggregates->description() << std::endl;
-} // BuildBasedOnNodeMapping
+}  // BuildBasedOnNodeMapping
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildBasedOnPrimalInterfaceDofMap(
-    const std::string& prefix, Level &currentLevel) const
-{
+    const std::string &prefix, Level &currentLevel) const {
   const GlobalOrdinal GO_ZERO = Teuchos::ScalarTraits<LocalOrdinal>::zero();
-  const GlobalOrdinal GO_ONE = Teuchos::ScalarTraits<GlobalOrdinal>::one();
+  const GlobalOrdinal GO_ONE  = Teuchos::ScalarTraits<GlobalOrdinal>::one();
 
   // filled with striding information from A01
-  LocalOrdinal numDofsPerDualNode = 0;
+  LocalOrdinal numDofsPerDualNode   = 0;
   LocalOrdinal numDofsPerPrimalNode = 0;
 
   // Grab the off-diagonal block (0,1) from the global blocked operator
-  RCP<const Matrix> A01 = Get<RCP<Matrix>>(currentLevel, "A");
-  RCP<const Aggregates> primalAggregates = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
+  RCP<const Matrix> A01                           = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<const Aggregates> primalAggregates          = Get<RCP<Aggregates>>(currentLevel, "Aggregates");
   ArrayRCP<const LocalOrdinal> primalVertex2AggId = primalAggregates->GetVertex2AggId()->getData(0);
 
-  auto comm = A01->getRowMap()->getComm();
+  auto comm        = A01->getRowMap()->getComm();
   const int myRank = comm->getRank();
 
   RCP<const Map> primalInterfaceDofRowMap = Teuchos::null;
@@ -266,59 +250,59 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
   TEUCHOS_ASSERT(!primalInterfaceDofRowMap.is_null());
 
   if (A01->IsView("stridedMaps") && rcp_dynamic_cast<const StridedMap>(A01->getRowMap("stridedMaps")) != Teuchos::null) {
-    auto stridedRowMap = rcp_dynamic_cast<const StridedMap>(A01->getRowMap("stridedMaps"));
-    auto stridedColMap = rcp_dynamic_cast<const StridedMap>(A01->getColMap("stridedMaps"));
+    auto stridedRowMap   = rcp_dynamic_cast<const StridedMap>(A01->getRowMap("stridedMaps"));
+    auto stridedColMap   = rcp_dynamic_cast<const StridedMap>(A01->getColMap("stridedMaps"));
     numDofsPerPrimalNode = Teuchos::as<LocalOrdinal>(stridedRowMap->getFixedBlockSize());
-    numDofsPerDualNode = Teuchos::as<LocalOrdinal>(stridedColMap->getFixedBlockSize());
+    numDofsPerDualNode   = Teuchos::as<LocalOrdinal>(stridedColMap->getFixedBlockSize());
 
     if (numDofsPerPrimalNode != numDofsPerDualNode) {
-    GetOStream(Warnings) << "InterfaceAggregation attempts to work with "
-        << numDofsPerPrimalNode << " primal DOFs per node and " << numDofsPerDualNode << " dual DOFs per node."
-        << "Be careful! Algorithm is not well-tested, if number of primal and dual DOFs per node differ." << std::endl;
+      GetOStream(Warnings) << "InterfaceAggregation attempts to work with "
+                           << numDofsPerPrimalNode << " primal DOFs per node and " << numDofsPerDualNode << " dual DOFs per node."
+                           << "Be careful! Algorithm is not well-tested, if number of primal and dual DOFs per node differ." << std::endl;
     }
   }
 
-  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerPrimalNode==0, Exceptions::RuntimeError,
-      "InterfaceAggregationFactory could not extract the number of primal DOFs per node from striding information. At least, make sure that StridedMap information has actually been provided.");
-  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerDualNode==0, Exceptions::RuntimeError,
-      "InterfaceAggregationFactory could not extract the number of dual DOFs per node from striding information. At least, make sure that StridedMap information has actually been provided.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerPrimalNode == 0, Exceptions::RuntimeError,
+                             "InterfaceAggregationFactory could not extract the number of primal DOFs per node from striding information. At least, make sure that StridedMap information has actually been provided.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numDofsPerDualNode == 0, Exceptions::RuntimeError,
+                             "InterfaceAggregationFactory could not extract the number of dual DOFs per node from striding information. At least, make sure that StridedMap information has actually been provided.");
 
   /* Determine block information for primal block
-    *
-    * primalDofOffset: global offset of primal DOF GIDs (usually is zero (default))
-    * primalBlockDim: block dim for fixed size blocks
-    * - is 2 or 3 (for 2d or 3d problems) on the finest level (# displacement dofs per node)
-    * - is 3 or 6 (for 2d or 3d problems) on coarser levels (# nullspace vectors)
-    */
+   *
+   * primalDofOffset: global offset of primal DOF GIDs (usually is zero (default))
+   * primalBlockDim: block dim for fixed size blocks
+   * - is 2 or 3 (for 2d or 3d problems) on the finest level (# displacement dofs per node)
+   * - is 3 or 6 (for 2d or 3d problems) on coarser levels (# nullspace vectors)
+   */
   GlobalOrdinal primalDofOffset = GO_ZERO;
-  LocalOrdinal primalBlockDim = numDofsPerPrimalNode;
+  LocalOrdinal primalBlockDim   = numDofsPerPrimalNode;
 
   /* Determine block information for Lagrange multipliers
-  *
-  * dualDofOffset: usually > zero (set by domainOffset for Ptent11Fact)
-  * dualBlockDim:
-  * - is primalBlockDim (for 2d or 3d problems) on the finest level (1 Lagrange multiplier per
-  *   displacement dof)
-  * - is 2 or 3 (for 2d or 3d problems) on coarser levels (same as on finest level, whereas there
-  *   are 3 or 6 displacement dofs per node)
-  */
+   *
+   * dualDofOffset: usually > zero (set by domainOffset for Ptent11Fact)
+   * dualBlockDim:
+   * - is primalBlockDim (for 2d or 3d problems) on the finest level (1 Lagrange multiplier per
+   *   displacement dof)
+   * - is 2 or 3 (for 2d or 3d problems) on coarser levels (same as on finest level, whereas there
+   *   are 3 or 6 displacement dofs per node)
+   */
   GlobalOrdinal dualDofOffset = A01->getColMap()->getMinAllGlobalIndex();
-  LocalOrdinal dualBlockDim = numDofsPerDualNode;
+  LocalOrdinal dualBlockDim   = numDofsPerDualNode;
 
   // Generate global replicated mapping "lagrNodeId -> dispNodeId"
-  RCP<const Map> dualDofMap = A01->getDomainMap();
+  RCP<const Map> dualDofMap    = A01->getDomainMap();
   GlobalOrdinal gMaxDualNodeId = AmalgamationFactory::DOFGid2NodeId(
       dualDofMap->getMaxAllGlobalIndex(), dualBlockDim, dualDofOffset, dualDofMap->getIndexBase());
   GlobalOrdinal gMinDualNodeId = AmalgamationFactory::DOFGid2NodeId(
       dualDofMap->getMinAllGlobalIndex(), dualBlockDim, dualDofOffset, dualDofMap->getIndexBase());
 
   GetOStream(Runtime1) << "  Dual DOF map: index base = " << dualDofMap->getIndexBase()
-    << ", block dim = " << dualBlockDim
-    << ", gid offset = " << dualDofOffset
-    << std::endl;
+                       << ", block dim = " << dualBlockDim
+                       << ", gid offset = " << dualDofOffset
+                       << std::endl;
 
   GetOStream(Runtime1) << "  [primal / dual] DOFs per node = [" << numDofsPerPrimalNode
-      << "/" << numDofsPerDualNode << "]" << std::endl;
+                       << "/" << numDofsPerDualNode << "]" << std::endl;
 
   // Generate locally replicated vector for mapping dual node IDs to primal node IDs
   Array<GlobalOrdinal> dualNodeId2primalNodeId(gMaxDualNodeId - gMinDualNodeId + 1, -GO_ONE);
@@ -333,16 +317,15 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
 
   // Fill mapping of Lagrange Node IDs to displacement aggregate IDs
   const size_t numMyPrimalInterfaceDOFs = primalInterfaceDofRowMap->getLocalNumElements();
-  for (size_t r = 0; r < numMyPrimalInterfaceDOFs; r += numDofsPerPrimalNode)
-  {
+  for (size_t r = 0; r < numMyPrimalInterfaceDOFs; r += numDofsPerPrimalNode) {
     GlobalOrdinal gPrimalRowId = primalInterfaceDofRowMap->getGlobalElement(r);
 
-    if (A01->getRowMap()->isNodeGlobalElement(gPrimalRowId)) // Remove this if?
+    if (A01->getRowMap()->isNodeGlobalElement(gPrimalRowId))  // Remove this if?
     {
-      const LocalOrdinal lPrimalRowId = A01->getRowMap()->getLocalElement(gPrimalRowId);
+      const LocalOrdinal lPrimalRowId   = A01->getRowMap()->getLocalElement(gPrimalRowId);
       const GlobalOrdinal gPrimalNodeId = AmalgamationFactory::DOFGid2NodeId(gPrimalRowId, primalBlockDim, primalDofOffset, primalInterfaceDofRowMap->getIndexBase());
-      const LocalOrdinal lPrimalNodeId = lPrimalRowId / numDofsPerPrimalNode;
-      const LocalOrdinal primalAggId = primalVertex2AggId[lPrimalNodeId];
+      const LocalOrdinal lPrimalNodeId  = lPrimalRowId / numDofsPerPrimalNode;
+      const LocalOrdinal primalAggId    = primalVertex2AggId[lPrimalNodeId];
 
       const GlobalOrdinal gDualDofId = A01->getColMap()->getGlobalElement(r);
 
@@ -350,32 +333,30 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
 
       if (local_dualNodeId2primalNodeId[gDualNodeId - gMinDualNodeId] == -GO_ONE) {
         local_dualNodeId2primalNodeId[gDualNodeId - gMinDualNodeId] = gPrimalNodeId;
-        local_dualNodeId2primalAggId[gDualNodeId - gMinDualNodeId] = primalAggId;
+        local_dualNodeId2primalAggId[gDualNodeId - gMinDualNodeId]  = primalAggId;
       } else {
         GetOStream(Warnings) << "PROC: " << myRank << " gDualNodeId " << gDualNodeId << " is already connected to primal nodeId "
-            << local_dualNodeId2primalNodeId[gDualNodeId - gMinDualNodeId]
-            << ". Ignore new dispNodeId: " << gPrimalNodeId << std::endl;
+                             << local_dualNodeId2primalNodeId[gDualNodeId - gMinDualNodeId]
+                             << ". Ignore new dispNodeId: " << gPrimalNodeId << std::endl;
       }
-
     }
   }
 
   const int dualNodeId2primalNodeIdSize = Teuchos::as<int>(local_dualNodeId2primalNodeId.size());
   Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, dualNodeId2primalNodeIdSize,
-      &local_dualNodeId2primalNodeId[0], &dualNodeId2primalNodeId[0]);
+                     &local_dualNodeId2primalNodeId[0], &dualNodeId2primalNodeId[0]);
   Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, dualNodeId2primalNodeIdSize,
-      &local_dualNodeId2primalAggId[0], &dualNodeId2primalAggId[0]);
+                     &local_dualNodeId2primalAggId[0], &dualNodeId2primalAggId[0]);
 
   // build node map for dual variables
   // generate "artificial nodes" for lagrange multipliers
   // the node map is also used for defining the Aggregates for the lagrange multipliers
   std::vector<GlobalOrdinal> dualNodes;
-  for (size_t r = 0; r < A01->getDomainMap()->getLocalNumElements(); r++)
-  {
+  for (size_t r = 0; r < A01->getDomainMap()->getLocalNumElements(); r++) {
     // determine global Lagrange multiplier row Dof
     // generate a node id using the grid, lagr_blockdim and lagr_offset // todo make sure, that
     // nodeId is unique and does not interfer with the displacement nodes
-    GlobalOrdinal gDualDofId = A01->getDomainMap()->getGlobalElement(r);
+    GlobalOrdinal gDualDofId  = A01->getDomainMap()->getGlobalElement(r);
     GlobalOrdinal gDualNodeId = AmalgamationFactory::DOFGid2NodeId(gDualDofId, dualBlockDim, dualDofOffset, 0);
     dualNodes.push_back(gDualNodeId);
   }
@@ -385,7 +366,7 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
 
   // define node map for Lagrange multipliers
   Teuchos::RCP<const Map> dualNodeMap = MapFactory::Build(A01->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), dualNodes, A01->getRowMap()->getIndexBase(), comm);
+                                                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), dualNodes, A01->getRowMap()->getIndexBase(), comm);
 
   // Build aggregates using the lagrange multiplier node map
   Teuchos::RCP<Aggregates> dualAggregates = Teuchos::rcp(new Aggregates(dualNodeMap));
@@ -393,30 +374,29 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
 
   // extract aggregate data structures to fill
   Teuchos::ArrayRCP<LocalOrdinal> dualVertex2AggId = dualAggregates->GetVertex2AggId()->getDataNonConst(0);
-  Teuchos::ArrayRCP<LocalOrdinal> dualProcWinner = dualAggregates->GetProcWinner()->getDataNonConst(0);
+  Teuchos::ArrayRCP<LocalOrdinal> dualProcWinner   = dualAggregates->GetProcWinner()->getDataNonConst(0);
 
   // loop over local lagrange multiplier node ids
   LocalOrdinal nLocalAggregates = 0;
   std::map<GlobalOrdinal, LocalOrdinal> primalAggId2localDualAggId;
-  for (size_t lDualNodeID = 0; lDualNodeID < dualNodeMap->getLocalNumElements(); ++lDualNodeID)
-  {
+  for (size_t lDualNodeID = 0; lDualNodeID < dualNodeMap->getLocalNumElements(); ++lDualNodeID) {
     const GlobalOrdinal gDualNodeId = dualNodeMap->getGlobalElement(lDualNodeID);
     const GlobalOrdinal primalAggId = dualNodeId2primalAggId[gDualNodeId - gMinDualNodeId];
     if (primalAggId2localDualAggId.count(primalAggId) == 0)
       primalAggId2localDualAggId[primalAggId] = nLocalAggregates++;
     dualVertex2AggId[lDualNodeID] = primalAggId2localDualAggId[primalAggId];
-    dualProcWinner[lDualNodeID] = myRank;
+    dualProcWinner[lDualNodeID]   = myRank;
   }
 
-  const LocalOrdinal fullblocksize = numDofsPerDualNode;
-  const GlobalOrdinal offset = A01->getColMap()->getMinAllGlobalIndex();
-  const LocalOrdinal blockid = -1;
-  const LocalOrdinal nStridedOffset = 0;
+  const LocalOrdinal fullblocksize    = numDofsPerDualNode;
+  const GlobalOrdinal offset          = A01->getColMap()->getMinAllGlobalIndex();
+  const LocalOrdinal blockid          = -1;
+  const LocalOrdinal nStridedOffset   = 0;
   const LocalOrdinal stridedblocksize = fullblocksize;
 
   RCP<Array<LO>> rowTranslation = rcp(new Array<LO>());
   RCP<Array<LO>> colTranslation = rcp(new Array<LO>());
-  const size_t numMyDualNodes = dualNodeMap->getLocalNumElements();
+  const size_t numMyDualNodes   = dualNodeMap->getLocalNumElements();
   for (size_t lDualNodeID = 0; lDualNodeID < numMyDualNodes; ++lDualNodeID) {
     for (LocalOrdinal dof = 0; dof < numDofsPerDualNode; ++dof) {
       rowTranslation->push_back(lDualNodeID);
@@ -427,8 +407,8 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
   TEUCHOS_ASSERT(A01->isFillComplete());
 
   RCP<AmalgamationInfo> dualAmalgamationInfo = rcp(new AmalgamationInfo(rowTranslation, colTranslation,
-      A01->getDomainMap(), A01->getDomainMap(), A01->getDomainMap(),
-      fullblocksize, offset, blockid, nStridedOffset, stridedblocksize));
+                                                                        A01->getDomainMap(), A01->getDomainMap(), A01->getDomainMap(),
+                                                                        fullblocksize, offset, blockid, nStridedOffset, stridedblocksize));
 
   dualAggregates->SetNumAggregates(nLocalAggregates);
   dualAggregates->AggregatesCrossProcessors(primalAggregates->AggregatesCrossProcessors());
@@ -441,8 +421,8 @@ void InterfaceAggregationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Bui
   currentLevel.Set("Aggregates", dualAggregates, this);
   currentLevel.Set("UnAmalgamationInfo", dualAmalgamationInfo, this);
 
-} // BuildBasedOnPrimalInterfaceDofMap
+}  // BuildBasedOnPrimalInterfaceDofMap
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_INTERFACEAGGREGATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_decl.hpp
index 0c22adb5e443..8da95fcdcc7d 100644
--- a/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_decl.hpp
@@ -52,8 +52,7 @@
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_TwoLevelFactoryBase.hpp"
 
-namespace MueLu
-{
+namespace MueLu {
 
 /*!
   @class InterfaceMappingTransferFactory
@@ -73,14 +72,13 @@ namespace MueLu
   The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see InterfaceAggregationFactory::GetValidParameters).<br>
   The * in the @c requested column states that the data is requested as input with all dependencies (see InterfaceAggregationFactory::DeclareInput).
 */
-template <class LocalOrdinal = DefaultLocalOrdinal,
+template <class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-class InterfaceMappingTransferFactory : public TwoLevelFactoryBase
-{
+          class Node          = DefaultNode>
+class InterfaceMappingTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_INTERFACEMAPPINGTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
-public:
+ public:
   //! Constructor.
   InterfaceMappingTransferFactory() = default;
 
@@ -92,6 +90,6 @@ class InterfaceMappingTransferFactory : public TwoLevelFactoryBase
   void Build(Level &fineLevel, Level &coarseLevel) const override;
 };
 
-} // namespace MueLu
+}  // namespace MueLu
 #define MUELU_INTERFACEMAPPINGTRANSFERFACTORY_SHORT
 #endif /* MUELU_INTERFACEMAPPINGTRANSFERFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_def.hpp
index 5daab13cbdfb..a5e7f0dd0129 100644
--- a/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_InterfaceMappingTransferFactory_def.hpp
@@ -47,32 +47,28 @@
 
 #include "MueLu_InterfaceMappingTransferFactory_decl.hpp"
 
-namespace MueLu
-{
+namespace MueLu {
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<const ParameterList> InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-{
+RCP<const ParameterList> InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
   validParamList->set<RCP<const FactoryBase>>("CoarseDualNodeID2PrimalNodeID", Teuchos::null, "Generating factory of the CoarseDualNodeID2PrimalNodeID map");
   return validParamList;
 }
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const
-{
+void InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
   Input(fineLevel, "CoarseDualNodeID2PrimalNodeID");
 }
 
 template <class LocalOrdinal, class GlobalOrdinal, class Node>
-void InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const
-{
+void InterfaceMappingTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
   Monitor m(*this, "Interface Mapping transfer factory");
 
   RCP<std::map<LocalOrdinal, LocalOrdinal>> coarseLagr2Dof = Get<RCP<std::map<LocalOrdinal, LocalOrdinal>>>(fineLevel, "CoarseDualNodeID2PrimalNodeID");
   Set(coarseLevel, "DualNodeID2PrimalNodeID", coarseLagr2Dof);
 }
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_INTERFACEMAPPINGTRANSFERFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_decl.hpp
index 91b50b0d5c1f..160a2ada8c7d 100644
--- a/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_decl.hpp
@@ -60,78 +60,78 @@
 
 namespace MueLu {
 
-  /*!
-    @class InverseApproximationFactory class.
-    @brief Factory for building the approximate inverse of a matrix.
+/*!
+  @class InverseApproximationFactory class.
+  @brief Factory for building the approximate inverse of a matrix.
 
-    ## Context, assumptions, and use cases ##
+  ## Context, assumptions, and use cases ##
 
-    This factory is intended to be used for building an approximate inverse of a given matrix \A. This is for now only
-    used in the SchurComplementFactory to generate a respective \Ainv matrix.
+  This factory is intended to be used for building an approximate inverse of a given matrix \A. This is for now only
+  used in the SchurComplementFactory to generate a respective \Ainv matrix.
 
-    For blocked matrices, the InverseApproximationFactory per default generates an approximate inverse of the A_00 term.
+  For blocked matrices, the InverseApproximationFactory per default generates an approximate inverse of the A_00 term.
 
-    ## Input/output of this factory ##
+  ## Input/output of this factory ##
 
-    ### User parameters of InterfaceAggregationFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    A                           | Factory | null     |  | * | * | Generating factory of the matrix A
-    inverse: approximation type | string  | diagonal |  | * | * | Method used to approximate the inverse
-    inverse: fixing             | bool    | false    |  | * | * | Fix diagonal by replacing small entries with 1.0
+  ### User parameters of InterfaceAggregationFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  A                           | Factory | null     |  | * | * | Generating factory of the matrix A
+  inverse: approximation type | string  | diagonal |  | * | * | Method used to approximate the inverse
+  inverse: fixing             | bool    | false    |  | * | * | Fix diagonal by replacing small entries with 1.0
 
-    The * in the master.xml column denotes that the parameter is defined in the master.xml file.
-    The * in the validated column means that the parameter is declared in the list of valid input parameters (see GetValidParameters() ).
-    The * in the requested column states that the data is requested as input with all dependencies (see DeclareInput() ).
+  The * in the master.xml column denotes that the parameter is defined in the master.xml file.
+  The * in the validated column means that the parameter is declared in the list of valid input parameters (see GetValidParameters() ).
+  The * in the requested column states that the data is requested as input with all dependencies (see DeclareInput() ).
 
-    ### Variables provided by this factory ###
+  ### Variables provided by this factory ###
 
-    After InverseApproximationFactory::Build the following data is available (if requested)
+  After InverseApproximationFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Ainv | InverseApproximationFactory | The approximate inverse of a given matrix.
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Ainv | InverseApproximationFactory | The approximate inverse of a given matrix.
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class InverseApproximationFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class InverseApproximationFactory : public SingleLevelFactoryBase {
 #undef MUELU_INVERSEAPPROXIMATIONFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    InverseApproximationFactory() = default;
+  //! Constructor.
+  InverseApproximationFactory() = default;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    //! Build an object with this factory.
-    void Build(Level& currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    //! Sparse inverse calculation method.
-    RCP<Matrix> GetSparseInverse(const RCP<Matrix>& A, const RCP<const CrsGraph>& sparsityPattern) const;
+ private:
+  //! Sparse inverse calculation method.
+  RCP<Matrix> GetSparseInverse(const RCP<Matrix>& A, const RCP<const CrsGraph>& sparsityPattern) const;
 
-  }; // class InverseApproximationFactory
+};  // class InverseApproximationFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_INVERSEAPPROXIMATIONFACTORY_SHORT
 #endif /* MUELU_INVERSEAPPROXIMATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_def.hpp b/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_def.hpp
index 7ec193a94d78..30fad853d9ba 100644
--- a/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_InverseApproximationFactory_def.hpp
@@ -69,147 +69,140 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    using Magnitude = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-
-    validParamList->set<RCP<const FactoryBase> >("A", NoFactory::getRCP(), "Matrix to build the approximate inverse on.\n");
-
-    validParamList->set<std::string>            ("inverse: approximation type",  "diagonal", "Method used to approximate the inverse.");
-    validParamList->set<Magnitude>              ("inverse: drop tolerance",      0.0       , "Values below this threshold  are dropped from the matrix (or fixed if diagonal fixing is active).");
-    validParamList->set<bool>                   ("inverse: fixing",              false     , "Keep diagonal and fix small entries with 1.0");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  using Magnitude                   = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+
+  validParamList->set<RCP<const FactoryBase>>("A", NoFactory::getRCP(), "Matrix to build the approximate inverse on.\n");
+
+  validParamList->set<std::string>("inverse: approximation type", "diagonal", "Method used to approximate the inverse.");
+  validParamList->set<Magnitude>("inverse: drop tolerance", 0.0, "Values below this threshold  are dropped from the matrix (or fixed if diagonal fixing is active).");
+  validParamList->set<bool>("inverse: fixing", false, "Keep diagonal and fix small entries with 1.0");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  using STS       = Teuchos::ScalarTraits<SC>;
+  const SC one    = STS::one();
+  using Magnitude = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+
+  const ParameterList& pL = GetParameterList();
+  const bool fixing       = pL.get<bool>("inverse: fixing");
+
+  // check which approximation type to use
+  const std::string method = pL.get<std::string>("inverse: approximation type");
+  TEUCHOS_TEST_FOR_EXCEPTION(method != "diagonal" && method != "lumping" && method != "sparseapproxinverse", Exceptions::RuntimeError,
+                             "MueLu::InverseApproximationFactory::Build: Approximation type can be 'diagonal' or 'lumping' or "
+                             "'sparseapproxinverse'.");
+
+  RCP<Matrix> A            = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A);
+  const bool isBlocked     = (bA == Teuchos::null ? false : true);
+
+  // if blocked operator is used, defaults to A(0,0)
+  if (isBlocked) A = bA->getMatrix(0, 0);
+
+  const Magnitude tol = pL.get<Magnitude>("inverse: drop tolerance");
+  RCP<Matrix> Ainv    = Teuchos::null;
+
+  if (method == "diagonal") {
+    const auto diag = VectorFactory::Build(A->getRangeMap(), true);
+    A->getLocalDiagCopy(*diag);
+    const RCP<const Vector> D = (!fixing ? Utilities::GetInverse(diag) : Utilities::GetInverse(diag, tol, one));
+    Ainv                      = MatrixFactory::Build(D);
+  } else if (method == "lumping") {
+    const auto diag           = Utilities::GetLumpedMatrixDiagonal(*A);
+    const RCP<const Vector> D = (!fixing ? Utilities::GetInverse(diag) : Utilities::GetInverse(diag, tol, one));
+    Ainv                      = MatrixFactory::Build(D);
+  } else if (method == "sparseapproxinverse") {
+    RCP<CrsGraph> sparsityPattern = Utilities::GetThresholdedGraph(A, tol, A->getGlobalMaxNumRowEntries());
+    GetOStream(Statistics1) << "NNZ Graph(A): " << A->getCrsGraph()->getGlobalNumEntries() << " , NNZ Tresholded Graph(A): " << sparsityPattern->getGlobalNumEntries() << std::endl;
+    RCP<Matrix> pAinv = GetSparseInverse(A, sparsityPattern);
+    Ainv              = Utilities::GetThresholdedMatrix(pAinv, tol, fixing, pAinv->getGlobalMaxNumRowEntries());
+    GetOStream(Statistics1) << "NNZ Ainv: " << pAinv->getGlobalNumEntries() << ", NNZ Tresholded Ainv (parameter: " << tol << "): " << Ainv->getGlobalNumEntries() << std::endl;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    using STS = Teuchos::ScalarTraits<SC>;
-    const SC one = STS::one();
-    using Magnitude = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-
-    const ParameterList& pL = GetParameterList();
-    const bool fixing = pL.get<bool>("inverse: fixing");
-
-    // check which approximation type to use
-    const std::string method = pL.get<std::string>("inverse: approximation type");
-    TEUCHOS_TEST_FOR_EXCEPTION(method != "diagonal" && method != "lumping" && method != "sparseapproxinverse", Exceptions::RuntimeError,
-                               "MueLu::InverseApproximationFactory::Build: Approximation type can be 'diagonal' or 'lumping' or "
-                               "'sparseapproxinverse'.");
-
-    RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A);
-    const bool isBlocked = (bA == Teuchos::null ? false : true);
-
-    // if blocked operator is used, defaults to A(0,0)
-    if(isBlocked) A = bA->getMatrix(0,0);
-
-    const Magnitude tol = pL.get<Magnitude>("inverse: drop tolerance");
-    RCP<Matrix> Ainv = Teuchos::null;
-
-    if(method=="diagonal")
-    {
-      const auto diag = VectorFactory::Build(A->getRangeMap(), true);
-      A->getLocalDiagCopy(*diag);
-      const RCP<const Vector> D = (!fixing ? Utilities::GetInverse(diag) : Utilities::GetInverse(diag, tol, one));
-      Ainv = MatrixFactory::Build(D);
+  GetOStream(Statistics1) << "Approximate inverse calculated by: " << method << "." << std::endl;
+  GetOStream(Statistics1) << "Ainv has " << Ainv->getGlobalNumRows() << "x" << Ainv->getGlobalNumCols() << " rows and columns." << std::endl;
+
+  Set(currentLevel, "Ainv", Ainv);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSparseInverse(const RCP<Matrix>& Aorg, const RCP<const CrsGraph>& sparsityPattern) const {
+  // construct the inverse matrix with the given sparsity pattern
+  RCP<Matrix> Ainv = MatrixFactory::Build(sparsityPattern);
+  Ainv->resumeFill();
+
+  // gather missing rows from other procs to generate an overlapping map
+  RCP<Import> rowImport = ImportFactory::Build(sparsityPattern->getRowMap(), sparsityPattern->getColMap());
+  RCP<Matrix> A         = MatrixFactory::Build(Aorg, *rowImport);
+
+  // loop over all rows of the inverse sparsity pattern (this can be done in parallel)
+  for (size_t k = 0; k < sparsityPattern->getLocalNumRows(); k++) {
+    // 1. get column indices Ik of local row k
+    ArrayView<const LO> Ik;
+    sparsityPattern->getLocalRowView(k, Ik);
+
+    // 2. get all local A(Ik,:) rows
+    Array<ArrayView<const LO>> J(Ik.size());
+    Array<ArrayView<const SC>> Ak(Ik.size());
+    Array<LO> Jk;
+    for (LO i = 0; i < Ik.size(); i++) {
+      A->getLocalRowView(Ik[i], J[i], Ak[i]);
+      for (LO j = 0; j < J[i].size(); j++)
+        Jk.append(J[i][j]);
     }
-    else if(method=="lumping")
-    {
-      const auto diag = Utilities::GetLumpedMatrixDiagonal(*A);
-      const RCP<const Vector> D = (!fixing ? Utilities::GetInverse(diag) : Utilities::GetInverse(diag, tol, one));
-      Ainv = MatrixFactory::Build(D);
-    }
-    else if(method=="sparseapproxinverse")
-    {
-      RCP<CrsGraph> sparsityPattern = Utilities::GetThresholdedGraph(A, tol, A->getGlobalMaxNumRowEntries());
-      GetOStream(Statistics1) << "NNZ Graph(A): " << A->getCrsGraph()->getGlobalNumEntries() << " , NNZ Tresholded Graph(A): " << sparsityPattern->getGlobalNumEntries() << std::endl;
-      RCP<Matrix> pAinv = GetSparseInverse(A, sparsityPattern);
-      Ainv = Utilities::GetThresholdedMatrix(pAinv, tol, fixing, pAinv->getGlobalMaxNumRowEntries());
-      GetOStream(Statistics1) << "NNZ Ainv: " << pAinv->getGlobalNumEntries() << ", NNZ Tresholded Ainv (parameter: " << tol << "): " << Ainv->getGlobalNumEntries() << std::endl;
-    }
-
-    GetOStream(Statistics1) << "Approximate inverse calculated by: " << method << "." << std::endl;
-    GetOStream(Statistics1) << "Ainv has " << Ainv->getGlobalNumRows() << "x" << Ainv->getGlobalNumCols() << " rows and columns." << std::endl;
-
-    Set(currentLevel, "Ainv", Ainv);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
-  InverseApproximationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSparseInverse(const RCP<Matrix>& Aorg, const RCP<const CrsGraph>& sparsityPattern) const {
-
-    // construct the inverse matrix with the given sparsity pattern
-    RCP<Matrix> Ainv = MatrixFactory::Build(sparsityPattern);
-    Ainv->resumeFill();
-
-    // gather missing rows from other procs to generate an overlapping map
-    RCP<Import> rowImport = ImportFactory::Build(sparsityPattern->getRowMap(), sparsityPattern->getColMap());
-    RCP<Matrix> A = MatrixFactory::Build(Aorg, *rowImport);
-
-    // loop over all rows of the inverse sparsity pattern (this can be done in parallel)
-    for(size_t k=0; k<sparsityPattern->getLocalNumRows(); k++) {
-
-      // 1. get column indices Ik of local row k
-      ArrayView<const LO> Ik;
-      sparsityPattern->getLocalRowView(k, Ik);
-
-      // 2. get all local A(Ik,:) rows
-      Array<ArrayView<const LO>> J(Ik.size());
-      Array<ArrayView<const SC>> Ak(Ik.size());
-      Array<LO> Jk;
-      for (LO i = 0; i < Ik.size(); i++) {
-        A->getLocalRowView(Ik[i], J[i], Ak[i]);
-        for (LO j = 0; j < J[i].size(); j++)
-          Jk.append(J[i][j]);
+    // set of unique column indices Jk
+    std::sort(Jk.begin(), Jk.end());
+    Jk.erase(std::unique(Jk.begin(), Jk.end()), Jk.end());
+    // create map
+    std::map<LO, LO> G;
+    for (LO i = 0; i < Jk.size(); i++) G.insert(std::pair<LO, LO>(Jk[i], i));
+
+    // 3. merge rows together
+    Teuchos::SerialDenseMatrix<LO, SC> localA(Jk.size(), Ik.size(), true);
+    for (LO i = 0; i < Ik.size(); i++) {
+      for (LO j = 0; j < J[i].size(); j++) {
+        localA(G.at(J[i][j]), i) = Ak[i][j];
       }
-      // set of unique column indices Jk
-      std::sort(Jk.begin(), Jk.end());
-      Jk.erase(std::unique(Jk.begin(), Jk.end()), Jk.end());
-      // create map
-      std::map<LO, LO> G;
-      for (LO i = 0; i < Jk.size(); i++) G.insert(std::pair<LO, LO>(Jk[i], i));
-
-      // 3. merge rows together
-      Teuchos::SerialDenseMatrix<LO, SC> localA(Jk.size(), Ik.size(), true);
-      for (LO i = 0; i < Ik.size(); i++) {
-        for (LO j = 0; j < J[i].size(); j++) {
-          localA(G.at(J[i][j]), i) = Ak[i][j];
-        }
-      }
-
-      // 4. get direction-vector
-      // diagonal needs an entry!
-      Teuchos::SerialDenseVector<LO, SC> ek(Jk.size(), true);
-      ek[std::find(Jk.begin(), Jk.end(), k) - Jk.begin()] = Teuchos::ScalarTraits<Scalar>::one();;
-
-      // 5. solve linear system for x
-      Teuchos::SerialDenseVector<LO, SC> localX(Ik.size());
-      Teuchos::SerialQRDenseSolver<LO, SC> qrSolver;
-      qrSolver.setMatrix(Teuchos::rcp(&localA, false));
-      qrSolver.setVectors(Teuchos::rcp(&localX, false), Teuchos::rcp(&ek, false));
-      const int err = qrSolver.solve();
-      TEUCHOS_TEST_FOR_EXCEPTION(err != 0, Exceptions::RuntimeError,
-                                 "MueLu::InverseApproximationFactory::GetSparseInverse: Error in serial QR solve.");
-
-      // 6. set calculated row into Ainv
-      ArrayView<const SC> Mk(localX.values(), localX.length());
-      Ainv->replaceLocalValues(k, Ik, Mk);
-
     }
-    Ainv->fillComplete();
 
-    return Ainv;
+    // 4. get direction-vector
+    // diagonal needs an entry!
+    Teuchos::SerialDenseVector<LO, SC> ek(Jk.size(), true);
+    ek[std::find(Jk.begin(), Jk.end(), k) - Jk.begin()] = Teuchos::ScalarTraits<Scalar>::one();
+    ;
+
+    // 5. solve linear system for x
+    Teuchos::SerialDenseVector<LO, SC> localX(Ik.size());
+    Teuchos::SerialQRDenseSolver<LO, SC> qrSolver;
+    qrSolver.setMatrix(Teuchos::rcp(&localA, false));
+    qrSolver.setVectors(Teuchos::rcp(&localX, false), Teuchos::rcp(&ek, false));
+    const int err = qrSolver.solve();
+    TEUCHOS_TEST_FOR_EXCEPTION(err != 0, Exceptions::RuntimeError,
+                               "MueLu::InverseApproximationFactory::GetSparseInverse: Error in serial QR solve.");
+
+    // 6. set calculated row into Ainv
+    ArrayView<const SC> Mk(localX.values(), localX.length());
+    Ainv->replaceLocalValues(k, Ik, Mk);
   }
+  Ainv->fillComplete();
+
+  return Ainv;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_INVERSEAPPROXIMATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_LineDetectionFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_LineDetectionFactory_decl.hpp
index 9a8207ae430a..75f0d2539590 100644
--- a/packages/muelu/src/Misc/MueLu_LineDetectionFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_LineDetectionFactory_decl.hpp
@@ -47,9 +47,9 @@
 #define MUELU_LINEDETECTIONFACTORY_DECL_HPP
 
 // same as in SemiCoarsenPFactory (TODO rework this)
-#define  VERTICAL       1
-#define  HORIZONTAL     2
-#define  GRID_SUPPLIED -1
+#define VERTICAL 1
+#define HORIZONTAL 2
+#define GRID_SUPPLIED -1
 
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_LineDetectionFactory_fwd.hpp"
@@ -59,82 +59,82 @@
 
 namespace MueLu {
 
-  /*!
-    @class LineDetectionFactory class.
-    @brief Factory for building line detection information
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class LineDetectionFactory : public SingleLevelFactoryBase {
+/*!
+  @class LineDetectionFactory class.
+  @brief Factory for building line detection information
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class LineDetectionFactory : public SingleLevelFactoryBase {
 #undef MUELU_LINEDETECTIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  using coordinate_type       = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using CoordinateMultiVector = typename Xpetra::MultiVector<coordinate_type, LO, GO, NO>;
 
-    using coordinate_type       = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using CoordinateMultiVector = typename Xpetra::MultiVector<coordinate_type, LO, GO, NO>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  LineDetectionFactory()
+    : Zorientation_(VERTICAL) {}
 
-    LineDetectionFactory() : Zorientation_(VERTICAL) { }
+  //! Destructor.
+  virtual ~LineDetectionFactory() {}
 
-    //! Destructor.
-    virtual ~LineDetectionFactory() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
-
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Builds line detection information and stores it in currentLevel
-      */
-    void Build(Level& currentLevel) const;
+    Builds line detection information and stores it in currentLevel
+    */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    void sort_coordinates(LO numCoords, LO* OrigLoc,
-                          coordinate_type* xvals,
-                          coordinate_type* yvals,
-                          coordinate_type* zvals,
-                          coordinate_type* xtemp,
-                          coordinate_type* ytemp,
-                          coordinate_type* ztemp,
-                          bool flipXY = false) const;
+ private:
+  void sort_coordinates(LO numCoords, LO* OrigLoc,
+                        coordinate_type* xvals,
+                        coordinate_type* yvals,
+                        coordinate_type* zvals,
+                        coordinate_type* xtemp,
+                        coordinate_type* ytemp,
+                        coordinate_type* ztemp,
+                        bool flipXY = false) const;
 
-    LO ML_compute_line_info(LO LayerId[], LO VertLineId[],
-                            LO Ndof, LO DofsPerNode,
-                            LO MeshNumbering, LO NumNodesPerVertLine,
-                            coordinate_type *xvals, coordinate_type *yvals, coordinate_type *zvals,
-                            const Teuchos::Comm<int>& comm ) const ;
+  LO ML_compute_line_info(LO LayerId[], LO VertLineId[],
+                          LO Ndof, LO DofsPerNode,
+                          LO MeshNumbering, LO NumNodesPerVertLine,
+                          coordinate_type* xvals, coordinate_type* yvals, coordinate_type* zvals,
+                          const Teuchos::Comm<int>& comm) const;
 
-    void ML_az_dsort2(coordinate_type dlist[], LO N, LO list2[]) const;
+  void ML_az_dsort2(coordinate_type dlist[], LO N, LO list2[]) const;
 
-    //! internally stores line detection mode
-    //! can be either vertical, horizontal or coordinates
-    //! for the first run. On the coarser levels we automatically
-    //! switch to vertical mode
-    mutable LO Zorientation_;
+  //! internally stores line detection mode
+  //! can be either vertical, horizontal or coordinates
+  //! for the first run. On the coarser levels we automatically
+  //! switch to vertical mode
+  mutable LO Zorientation_;
 
-  }; //class LineDetectionFactory
+};  // class LineDetectionFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_LINEDETECTIONFACTORY_SHORT
-#endif // MUELU_LINEDETECTIONFACTORY_DECL_HPP
+#endif  // MUELU_LINEDETECTIONFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_LineDetectionFactory_def.hpp b/packages/muelu/src/Misc/MueLu_LineDetectionFactory_def.hpp
index f16534229ce6..d636154f66fa 100644
--- a/packages/muelu/src/Misc/MueLu_LineDetectionFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_LineDetectionFactory_def.hpp
@@ -57,452 +57,454 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("linedetection: orientation");
-    SET_VALID_ENTRY("linedetection: num layers");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",               Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",     Teuchos::null, "Generating factory for coorindates");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-
-    // The factory needs the information about the number of z-layers. While this information is
-    // provided by the user for the finest level, the factory itself is responsible to provide the
-    // corresponding information on the coarser levels. Since a factory cannot be dependent on itself
-    // we use the NoFactory class as generator class, but remove the UserData keep flag, such that
-    // "NumZLayers" is part of the request/release mechanism.
-    // Please note, that this prevents us from having several (independent) CoarsePFactory instances!
-    // TODO: allow factory to dependent on self-generated data for TwoLevelFactories -> introduce ExpertRequest/Release in Level
-    currentLevel.DeclareInput("NumZLayers", NoFactory::get(), this);
-    currentLevel.RemoveKeepFlag("NumZLayers", NoFactory::get(), MueLu::UserData);
+  SET_VALID_ENTRY("linedetection: orientation");
+  SET_VALID_ENTRY("linedetection: num layers");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for coorindates");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+
+  // The factory needs the information about the number of z-layers. While this information is
+  // provided by the user for the finest level, the factory itself is responsible to provide the
+  // corresponding information on the coarser levels. Since a factory cannot be dependent on itself
+  // we use the NoFactory class as generator class, but remove the UserData keep flag, such that
+  // "NumZLayers" is part of the request/release mechanism.
+  // Please note, that this prevents us from having several (independent) CoarsePFactory instances!
+  // TODO: allow factory to dependent on self-generated data for TwoLevelFactories -> introduce ExpertRequest/Release in Level
+  currentLevel.DeclareInput("NumZLayers", NoFactory::get(), this);
+  currentLevel.RemoveKeepFlag("NumZLayers", NoFactory::get(), MueLu::UserData);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Line detection (Ray style)", currentLevel);
+
+  LO NumZDir = 0;
+  RCP<CoordinateMultiVector> fineCoords;
+  ArrayRCP<coordinate_type> x, y, z;
+  coordinate_type *xptr = NULL, *yptr = NULL, *zptr = NULL;
+
+  // obtain general variables
+  RCP<Matrix> A         = Get<RCP<Matrix> >(currentLevel, "A");
+  LO BlkSize            = A->GetFixedBlockSize();
+  RCP<const Map> rowMap = A->getRowMap();
+  LO Ndofs              = rowMap->getLocalNumElements();
+  LO Nnodes             = Ndofs / BlkSize;
+
+  // collect information provided by user
+  const ParameterList& pL           = GetParameterList();
+  const std::string lineOrientation = pL.get<std::string>("linedetection: orientation");
+
+  // interpret "line orientation" parameter provided by the user on the finest level
+  if (currentLevel.GetLevelID() == 0) {
+    if (lineOrientation == "vertical")
+      Zorientation_ = VERTICAL;
+    else if (lineOrientation == "horizontal")
+      Zorientation_ = HORIZONTAL;
+    else if (lineOrientation == "coordinates")
+      Zorientation_ = GRID_SUPPLIED;
+    else
+      TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: The parameter 'semicoarsen: line orientation' must be either 'vertical', 'horizontal' or 'coordinates'.");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Line detection (Ray style)", currentLevel);
-
-    LO               NumZDir = 0;
-    RCP<CoordinateMultiVector> fineCoords;
-    ArrayRCP<coordinate_type>  x, y, z;
-    coordinate_type            *xptr = NULL, *yptr = NULL, *zptr = NULL;
-
-    // obtain general variables
-    RCP<Matrix> A         = Get< RCP<Matrix> >      (currentLevel, "A");
-    LO BlkSize            = A->GetFixedBlockSize();
-    RCP<const Map> rowMap = A->getRowMap();
-    LO Ndofs              = rowMap->getLocalNumElements();
-    LO Nnodes             = Ndofs/BlkSize;
-
-    // collect information provided by user
-    const ParameterList& pL = GetParameterList();
-    const std::string lineOrientation = pL.get<std::string>("linedetection: orientation");
-
-    // interpret "line orientation" parameter provided by the user on the finest level
-    if(currentLevel.GetLevelID() == 0) {
-      if(lineOrientation=="vertical")
-        Zorientation_ = VERTICAL;
-      else if (lineOrientation=="horizontal")
-        Zorientation_ = HORIZONTAL;
-      else if (lineOrientation=="coordinates")
-        Zorientation_ = GRID_SUPPLIED;
-      else
-        TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: The parameter 'semicoarsen: line orientation' must be either 'vertical', 'horizontal' or 'coordinates'.");
-    }
-
-    //TEUCHOS_TEST_FOR_EXCEPTION(Zorientation_!=VERTICAL, Exceptions::RuntimeError, "LineDetectionFactory: The 'horizontal' or 'coordinates' have not been tested!!!. Please remove this exception check and carefully test these modes!");
+  // TEUCHOS_TEST_FOR_EXCEPTION(Zorientation_!=VERTICAL, Exceptions::RuntimeError, "LineDetectionFactory: The 'horizontal' or 'coordinates' have not been tested!!!. Please remove this exception check and carefully test these modes!");
 
-    // obtain number of z layers (variable over levels)
-    // This information is user-provided on the finest level and transferred to the coarser
-    // levels by the SemiCoarsenPFactor using the internal "NumZLayers" variable.
-    if(currentLevel.GetLevelID() == 0) {
-      if(currentLevel.IsAvailable("NumZLayers", NoFactory::get())) {
-        NumZDir = currentLevel.Get<LO>("NumZLayers", NoFactory::get()); //obtain info
-        GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information from Level(0))" << std::endl;
-      } else {
-        // check whether user provides information or it can be reconstructed from coordinates
-        NumZDir = pL.get<LO>("linedetection: num layers");
-        if(NumZDir == -1) {
-          bool CoordsAvail = currentLevel.IsAvailable("Coordinates");
-
-          if (CoordsAvail == true) {
-            // try to reconstruct the number of layers from coordinates
-            fineCoords = Get< RCP<CoordinateMultiVector> > (currentLevel, "Coordinates");
-            TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
-            x = fineCoords->getDataNonConst(0);
-            y = fineCoords->getDataNonConst(1);
-            z = fineCoords->getDataNonConst(2);
-            xptr = x.getRawPtr();
-            yptr = y.getRawPtr();
-            zptr = z.getRawPtr();
-
-            LO NumCoords = Ndofs/BlkSize;
-
-            /* sort coordinates so that we can order things according to lines */
-            Teuchos::ArrayRCP<LO> TOrigLoc= Teuchos::arcp<LO>(NumCoords);   LO* OrigLoc= TOrigLoc.getRawPtr();
-            Teuchos::ArrayRCP<coordinate_type> Txtemp  = Teuchos::arcp<coordinate_type>(NumCoords);   coordinate_type* xtemp  = Txtemp.getRawPtr();
-            Teuchos::ArrayRCP<coordinate_type> Tytemp  = Teuchos::arcp<coordinate_type>(NumCoords);   coordinate_type* ytemp  = Tytemp.getRawPtr();
-            Teuchos::ArrayRCP<coordinate_type> Tztemp  = Teuchos::arcp<coordinate_type>(NumCoords);   coordinate_type* ztemp  = Tztemp.getRawPtr();
-
-            // sort coordinates in {x,y,z}vals (returned in {x,y,z}temp) so that we can order things according to lines
-            // switch x and y coordinates for semi-coarsening...
-            sort_coordinates(NumCoords, OrigLoc, xptr, yptr, zptr, xtemp, ytemp, ztemp, true);
-
-            /* go through each vertical line and populate blockIndices so all   */
-            /* dofs within a PDE within a vertical line correspond to one block.*/
-            LO NumBlocks = 0;
-            LO NumNodesPerVertLine = 0;
-            LO index = 0;
-
-            while ( index < NumCoords ) {
-              coordinate_type xfirst = xtemp[index]; coordinate_type yfirst = ytemp[index];
-              LO next = index+1;
-              while ( (next != NumCoords) && (xtemp[next] == xfirst) &&
-                      (ytemp[next] == yfirst))
-                next++;
-              if (NumBlocks == 0) {
-                NumNodesPerVertLine = next-index;
-              }
-              // the number of vertical lines must be the same on all processors
-              // TAW: Sep 14 2015: or zero as we allow "empty" processors
-              //TEUCHOS_TEST_FOR_EXCEPTION(next-index != NumNodesPerVertLine,Exceptions::RuntimeError, "Error code only works for constant block size now!!!\n");
-              NumBlocks++;
-              index = next;
+  // obtain number of z layers (variable over levels)
+  // This information is user-provided on the finest level and transferred to the coarser
+  // levels by the SemiCoarsenPFactor using the internal "NumZLayers" variable.
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("NumZLayers", NoFactory::get())) {
+      NumZDir = currentLevel.Get<LO>("NumZLayers", NoFactory::get());  // obtain info
+      GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information from Level(0))" << std::endl;
+    } else {
+      // check whether user provides information or it can be reconstructed from coordinates
+      NumZDir = pL.get<LO>("linedetection: num layers");
+      if (NumZDir == -1) {
+        bool CoordsAvail = currentLevel.IsAvailable("Coordinates");
+
+        if (CoordsAvail == true) {
+          // try to reconstruct the number of layers from coordinates
+          fineCoords = Get<RCP<CoordinateMultiVector> >(currentLevel, "Coordinates");
+          TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
+          x    = fineCoords->getDataNonConst(0);
+          y    = fineCoords->getDataNonConst(1);
+          z    = fineCoords->getDataNonConst(2);
+          xptr = x.getRawPtr();
+          yptr = y.getRawPtr();
+          zptr = z.getRawPtr();
+
+          LO NumCoords = Ndofs / BlkSize;
+
+          /* sort coordinates so that we can order things according to lines */
+          Teuchos::ArrayRCP<LO> TOrigLoc            = Teuchos::arcp<LO>(NumCoords);
+          LO* OrigLoc                               = TOrigLoc.getRawPtr();
+          Teuchos::ArrayRCP<coordinate_type> Txtemp = Teuchos::arcp<coordinate_type>(NumCoords);
+          coordinate_type* xtemp                    = Txtemp.getRawPtr();
+          Teuchos::ArrayRCP<coordinate_type> Tytemp = Teuchos::arcp<coordinate_type>(NumCoords);
+          coordinate_type* ytemp                    = Tytemp.getRawPtr();
+          Teuchos::ArrayRCP<coordinate_type> Tztemp = Teuchos::arcp<coordinate_type>(NumCoords);
+          coordinate_type* ztemp                    = Tztemp.getRawPtr();
+
+          // sort coordinates in {x,y,z}vals (returned in {x,y,z}temp) so that we can order things according to lines
+          // switch x and y coordinates for semi-coarsening...
+          sort_coordinates(NumCoords, OrigLoc, xptr, yptr, zptr, xtemp, ytemp, ztemp, true);
+
+          /* go through each vertical line and populate blockIndices so all   */
+          /* dofs within a PDE within a vertical line correspond to one block.*/
+          LO NumBlocks           = 0;
+          LO NumNodesPerVertLine = 0;
+          LO index               = 0;
+
+          while (index < NumCoords) {
+            coordinate_type xfirst = xtemp[index];
+            coordinate_type yfirst = ytemp[index];
+            LO next                = index + 1;
+            while ((next != NumCoords) && (xtemp[next] == xfirst) &&
+                   (ytemp[next] == yfirst))
+              next++;
+            if (NumBlocks == 0) {
+              NumNodesPerVertLine = next - index;
             }
-
-            NumZDir = NumNodesPerVertLine;
-            GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information reconstructed from provided node coordinates)" << std::endl;
-          } else {
-            TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: BuildP: User has to provide valid number of layers (e.g. using the 'line detection: num layers' parameter).");
+            // the number of vertical lines must be the same on all processors
+            // TAW: Sep 14 2015: or zero as we allow "empty" processors
+            // TEUCHOS_TEST_FOR_EXCEPTION(next-index != NumNodesPerVertLine,Exceptions::RuntimeError, "Error code only works for constant block size now!!!\n");
+            NumBlocks++;
+            index = next;
           }
+
+          NumZDir = NumNodesPerVertLine;
+          GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information reconstructed from provided node coordinates)" << std::endl;
         } else {
-          GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information provided by user through 'line detection: num layers')" << std::endl;
+          TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: BuildP: User has to provide valid number of layers (e.g. using the 'line detection: num layers' parameter).");
         }
-      } // end else (user provides information or can be reconstructed) on finest level
-    } else {
-      // coarse level information
-      // TODO get rid of NoFactory here and use SemiCoarsenPFactory as source of NumZLayers instead.
-      if(currentLevel.IsAvailable("NumZLayers", NoFactory::get())) {
-        NumZDir = currentLevel.Get<LO>("NumZLayers", NoFactory::get()); //obtain info
-        GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << std::endl;
       } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: BuildP: No NumZLayers variable found. This cannot be.");
+        GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << " (information provided by user through 'line detection: num layers')" << std::endl;
       }
+    }  // end else (user provides information or can be reconstructed) on finest level
+  } else {
+    // coarse level information
+    // TODO get rid of NoFactory here and use SemiCoarsenPFactory as source of NumZLayers instead.
+    if (currentLevel.IsAvailable("NumZLayers", NoFactory::get())) {
+      NumZDir = currentLevel.Get<LO>("NumZLayers", NoFactory::get());  // obtain info
+      GetOStream(Runtime1) << "Number of layers for line detection: " << NumZDir << std::endl;
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "LineDetectionFactory: BuildP: No NumZLayers variable found. This cannot be.");
     }
+  }
 
-    // plausibility check and further variable collection
-    if (Zorientation_ == GRID_SUPPLIED) { // On finest level, fetch user-provided coordinates if available...
-      bool CoordsAvail = currentLevel.IsAvailable("Coordinates");
-
-      if (CoordsAvail == false) {
-        if (currentLevel.GetLevelID() == 0)
-          throw Exceptions::RuntimeError("Coordinates must be supplied if line detection orientation not given.");
-        else
-          throw Exceptions::RuntimeError("Coordinates not generated by previous invocation of LineDetectionFactory's BuildP() method.");
-      }
-      fineCoords = Get< RCP<CoordinateMultiVector> > (currentLevel, "Coordinates");
-      TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
-      x = fineCoords->getDataNonConst(0);
-      y = fineCoords->getDataNonConst(1);
-      z = fineCoords->getDataNonConst(2);
-      xptr = x.getRawPtr();
-      yptr = y.getRawPtr();
-      zptr = z.getRawPtr();
-    }
+  // plausibility check and further variable collection
+  if (Zorientation_ == GRID_SUPPLIED) {  // On finest level, fetch user-provided coordinates if available...
+    bool CoordsAvail = currentLevel.IsAvailable("Coordinates");
 
-    // perform line detection
-    if (NumZDir > 0) {
-      LO   *LayerId, *VertLineId;
-      Teuchos::ArrayRCP<LO>     TLayerId   = Teuchos::arcp<LO>(Nnodes);  LayerId       = TLayerId.getRawPtr();
-      Teuchos::ArrayRCP<LO>     TVertLineId= Teuchos::arcp<LO>(Nnodes);  VertLineId    = TVertLineId.getRawPtr();
-
-      NumZDir = ML_compute_line_info(LayerId, VertLineId, Ndofs, BlkSize,
-                                     Zorientation_, NumZDir,xptr,yptr,zptr, *(rowMap->getComm()));
-      //it is NumZDir=NCLayers*NVertLines*DofsPerNode;
-
-      // store output data on current level
-      // The line detection data is used by the SemiCoarsenPFactory and the line smoothers in Ifpack/Ifpack2
-      Set(currentLevel, "CoarseNumZLayers", NumZDir);
-      Set(currentLevel, "LineDetection_Layers", TLayerId);
-      Set(currentLevel, "LineDetection_VertLineIds", TVertLineId);
-    } else {
-      Teuchos::ArrayRCP<LO>     TLayerId       = Teuchos::arcp<LO>(0);
-      Teuchos::ArrayRCP<LO>     TVertLineId    = Teuchos::arcp<LO>(0);
-      Teuchos::ArrayRCP<LO>     TVertLineIdSmoo= Teuchos::arcp<LO>(0);
-
-      // store output data on current level
-      // The line detection data is used by the SemiCoarsenPFactory and the line smoothers in Ifpack/Ifpack2
-      Set(currentLevel, "CoarseNumZLayers", NumZDir);
-      Set(currentLevel, "LineDetection_Layers", TLayerId);
-      Set(currentLevel, "LineDetection_VertLineIds", TVertLineId);
+    if (CoordsAvail == false) {
+      if (currentLevel.GetLevelID() == 0)
+        throw Exceptions::RuntimeError("Coordinates must be supplied if line detection orientation not given.");
+      else
+        throw Exceptions::RuntimeError("Coordinates not generated by previous invocation of LineDetectionFactory's BuildP() method.");
     }
-
-    // automatically switch to vertical mode on the coarser levels
-    if(Zorientation_ != VERTICAL)
-      Zorientation_ = VERTICAL;
+    fineCoords = Get<RCP<CoordinateMultiVector> >(currentLevel, "Coordinates");
+    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
+    x    = fineCoords->getDataNonConst(0);
+    y    = fineCoords->getDataNonConst(1);
+    z    = fineCoords->getDataNonConst(2);
+    xptr = x.getRawPtr();
+    yptr = y.getRawPtr();
+    zptr = z.getRawPtr();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalOrdinal LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ML_compute_line_info(LocalOrdinal LayerId[], LocalOrdinal VertLineId[], LocalOrdinal Ndof, LocalOrdinal DofsPerNode, LocalOrdinal MeshNumbering, LocalOrdinal NumNodesPerVertLine, typename Teuchos::ScalarTraits<Scalar>::coordinateType *xvals, typename Teuchos::ScalarTraits<Scalar>::coordinateType *yvals, typename Teuchos::ScalarTraits<Scalar>::coordinateType *zvals, const Teuchos::Comm<int>& /* comm */) const {
-
-    LO    Nnodes, NVertLines, MyNode;
-    LO    NumCoords, next; //, subindex, subnext;
-    coordinate_type xfirst, yfirst;
-    coordinate_type *xtemp, *ytemp, *ztemp;
-    LO    *OrigLoc;
-    LO    i,j,count;
-    LO    RetVal;
+  // perform line detection
+  if (NumZDir > 0) {
+    LO *LayerId, *VertLineId;
+    Teuchos::ArrayRCP<LO> TLayerId    = Teuchos::arcp<LO>(Nnodes);
+    LayerId                           = TLayerId.getRawPtr();
+    Teuchos::ArrayRCP<LO> TVertLineId = Teuchos::arcp<LO>(Nnodes);
+    VertLineId                        = TVertLineId.getRawPtr();
+
+    NumZDir = ML_compute_line_info(LayerId, VertLineId, Ndofs, BlkSize,
+                                   Zorientation_, NumZDir, xptr, yptr, zptr, *(rowMap->getComm()));
+    // it is NumZDir=NCLayers*NVertLines*DofsPerNode;
+
+    // store output data on current level
+    // The line detection data is used by the SemiCoarsenPFactory and the line smoothers in Ifpack/Ifpack2
+    Set(currentLevel, "CoarseNumZLayers", NumZDir);
+    Set(currentLevel, "LineDetection_Layers", TLayerId);
+    Set(currentLevel, "LineDetection_VertLineIds", TVertLineId);
+  } else {
+    Teuchos::ArrayRCP<LO> TLayerId        = Teuchos::arcp<LO>(0);
+    Teuchos::ArrayRCP<LO> TVertLineId     = Teuchos::arcp<LO>(0);
+    Teuchos::ArrayRCP<LO> TVertLineIdSmoo = Teuchos::arcp<LO>(0);
+
+    // store output data on current level
+    // The line detection data is used by the SemiCoarsenPFactory and the line smoothers in Ifpack/Ifpack2
+    Set(currentLevel, "CoarseNumZLayers", NumZDir);
+    Set(currentLevel, "LineDetection_Layers", TLayerId);
+    Set(currentLevel, "LineDetection_VertLineIds", TVertLineId);
+  }
 
-    RetVal = 0;
-    if ((MeshNumbering != VERTICAL) && (MeshNumbering != HORIZONTAL)) {
-      if ( (xvals == NULL) || (yvals == NULL) || (zvals == NULL)) RetVal = -1;
-    }
-    else {
-      if  (NumNodesPerVertLine == -1)                     RetVal = -4;
-      if ( ((Ndof/DofsPerNode)%NumNodesPerVertLine) != 0) RetVal = -3;
-    }
-    if ( (Ndof%DofsPerNode) != 0) RetVal = -2;
+  // automatically switch to vertical mode on the coarser levels
+  if (Zorientation_ != VERTICAL)
+    Zorientation_ = VERTICAL;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ML_compute_line_info(LocalOrdinal LayerId[], LocalOrdinal VertLineId[], LocalOrdinal Ndof, LocalOrdinal DofsPerNode, LocalOrdinal MeshNumbering, LocalOrdinal NumNodesPerVertLine, typename Teuchos::ScalarTraits<Scalar>::coordinateType* xvals, typename Teuchos::ScalarTraits<Scalar>::coordinateType* yvals, typename Teuchos::ScalarTraits<Scalar>::coordinateType* zvals, const Teuchos::Comm<int>& /* comm */) const {
+  LO Nnodes, NVertLines, MyNode;
+  LO NumCoords, next;  //, subindex, subnext;
+  coordinate_type xfirst, yfirst;
+  coordinate_type *xtemp, *ytemp, *ztemp;
+  LO* OrigLoc;
+  LO i, j, count;
+  LO RetVal;
+
+  RetVal = 0;
+  if ((MeshNumbering != VERTICAL) && (MeshNumbering != HORIZONTAL)) {
+    if ((xvals == NULL) || (yvals == NULL) || (zvals == NULL)) RetVal = -1;
+  } else {
+    if (NumNodesPerVertLine == -1) RetVal = -4;
+    if (((Ndof / DofsPerNode) % NumNodesPerVertLine) != 0) RetVal = -3;
+  }
+  if ((Ndof % DofsPerNode) != 0) RetVal = -2;
 
-    TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -1, Exceptions::RuntimeError, "Not semicoarsening as no mesh numbering information or coordinates are given\n");
-    TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -4, Exceptions::RuntimeError, "Not semicoarsening as the number of z nodes is not given.\n");
-    TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -3, Exceptions::RuntimeError, "Not semicoarsening as the total number of nodes is not evenly divisible by the number of z direction nodes .\n");
-    TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -2, Exceptions::RuntimeError, "Not semicoarsening as something is off with the number of degrees-of-freedom per node.\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -1, Exceptions::RuntimeError, "Not semicoarsening as no mesh numbering information or coordinates are given\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -4, Exceptions::RuntimeError, "Not semicoarsening as the number of z nodes is not given.\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -3, Exceptions::RuntimeError, "Not semicoarsening as the total number of nodes is not evenly divisible by the number of z direction nodes .\n");
+  TEUCHOS_TEST_FOR_EXCEPTION(RetVal == -2, Exceptions::RuntimeError, "Not semicoarsening as something is off with the number of degrees-of-freedom per node.\n");
 
-    Nnodes = Ndof/DofsPerNode;
-    for (MyNode = 0; MyNode < Nnodes;  MyNode++) VertLineId[MyNode]    = -1;
-    for (MyNode = 0; MyNode < Nnodes;  MyNode++) LayerId[MyNode]       = -1;
+  Nnodes = Ndof / DofsPerNode;
+  for (MyNode = 0; MyNode < Nnodes; MyNode++) VertLineId[MyNode] = -1;
+  for (MyNode = 0; MyNode < Nnodes; MyNode++) LayerId[MyNode] = -1;
 
-    if (MeshNumbering == VERTICAL) {
-      for (MyNode = 0; MyNode < Nnodes; MyNode++) {
-        LayerId[MyNode]= MyNode%NumNodesPerVertLine;
-        VertLineId[MyNode]= (MyNode- LayerId[MyNode])/NumNodesPerVertLine;
-      }
+  if (MeshNumbering == VERTICAL) {
+    for (MyNode = 0; MyNode < Nnodes; MyNode++) {
+      LayerId[MyNode]    = MyNode % NumNodesPerVertLine;
+      VertLineId[MyNode] = (MyNode - LayerId[MyNode]) / NumNodesPerVertLine;
     }
-    else if (MeshNumbering == HORIZONTAL) {
-      NVertLines = Nnodes/NumNodesPerVertLine;
-      for (MyNode = 0; MyNode < Nnodes; MyNode++) {
-        VertLineId[MyNode]    = MyNode%NVertLines;
-        LayerId[MyNode]   = (MyNode- VertLineId[MyNode])/NVertLines;
-      }
-    }
-    else {
-      // coordinates mode: we distinguish between vertical line numbering for semi-coarsening and line smoothing
-      NumCoords = Ndof/DofsPerNode;
-
-      // reserve temporary memory
-      Teuchos::ArrayRCP<LO> TOrigLoc= Teuchos::arcp<LO>(NumCoords);       OrigLoc= TOrigLoc.getRawPtr();
-      Teuchos::ArrayRCP<coordinate_type> Txtemp  = Teuchos::arcp<coordinate_type>(NumCoords);       xtemp  = Txtemp.getRawPtr();
-      Teuchos::ArrayRCP<coordinate_type> Tytemp  = Teuchos::arcp<coordinate_type>(NumCoords);       ytemp  = Tytemp.getRawPtr();
-      Teuchos::ArrayRCP<coordinate_type> Tztemp  = Teuchos::arcp<coordinate_type>(NumCoords);       ztemp  = Tztemp.getRawPtr();
-
-      // build vertical line info for semi-coarsening
-
-      // sort coordinates in {x,y,z}vals (returned in {x,y,z}temp) so that we can order things according to lines
-      // switch x and y coordinates for semi-coarsening...
-      sort_coordinates(NumCoords, OrigLoc, xvals, yvals, zvals, xtemp, ytemp, ztemp, /*true*/ true);
-
-      LO NumBlocks = 0;
-      LO index = 0;
-
-      while ( index < NumCoords ) {
-        xfirst = xtemp[index];  yfirst = ytemp[index];
-        next = index+1;
-        while ( (next != NumCoords) && (xtemp[next] == xfirst) &&
-                (ytemp[next] == yfirst))
-          next++;
-        if (NumBlocks == 0) {
-          NumNodesPerVertLine = next-index;
-        }
-        // The number of vertical lines must be the same on all processors
-        // TAW: Sep 14, 2015: or zero as we allow for empty processors.
-        //TEUCHOS_TEST_FOR_EXCEPTION(next-index != NumNodesPerVertLine,Exceptions::RuntimeError, "Error code only works for constant block size now!!!\n");
-        count = 0;
-        for (j= index; j < next; j++) {
-          VertLineId[OrigLoc[j]] = NumBlocks;
-          LayerId[OrigLoc[j]] = count++;
-        }
-        NumBlocks++;
-        index = next;
-      }
+  } else if (MeshNumbering == HORIZONTAL) {
+    NVertLines = Nnodes / NumNodesPerVertLine;
+    for (MyNode = 0; MyNode < Nnodes; MyNode++) {
+      VertLineId[MyNode] = MyNode % NVertLines;
+      LayerId[MyNode]    = (MyNode - VertLineId[MyNode]) / NVertLines;
     }
-
-    /* check that everyone was assigned */
-
-    for (i = 0; i < Nnodes;  i++) {
-      if (VertLineId[i] == -1) {
-        GetOStream(Warnings1) << "Warning: did not assign " << i << " to a vertical line?????\n" << std::endl;
+  } else {
+    // coordinates mode: we distinguish between vertical line numbering for semi-coarsening and line smoothing
+    NumCoords = Ndof / DofsPerNode;
+
+    // reserve temporary memory
+    Teuchos::ArrayRCP<LO> TOrigLoc            = Teuchos::arcp<LO>(NumCoords);
+    OrigLoc                                   = TOrigLoc.getRawPtr();
+    Teuchos::ArrayRCP<coordinate_type> Txtemp = Teuchos::arcp<coordinate_type>(NumCoords);
+    xtemp                                     = Txtemp.getRawPtr();
+    Teuchos::ArrayRCP<coordinate_type> Tytemp = Teuchos::arcp<coordinate_type>(NumCoords);
+    ytemp                                     = Tytemp.getRawPtr();
+    Teuchos::ArrayRCP<coordinate_type> Tztemp = Teuchos::arcp<coordinate_type>(NumCoords);
+    ztemp                                     = Tztemp.getRawPtr();
+
+    // build vertical line info for semi-coarsening
+
+    // sort coordinates in {x,y,z}vals (returned in {x,y,z}temp) so that we can order things according to lines
+    // switch x and y coordinates for semi-coarsening...
+    sort_coordinates(NumCoords, OrigLoc, xvals, yvals, zvals, xtemp, ytemp, ztemp, /*true*/ true);
+
+    LO NumBlocks = 0;
+    LO index     = 0;
+
+    while (index < NumCoords) {
+      xfirst = xtemp[index];
+      yfirst = ytemp[index];
+      next   = index + 1;
+      while ((next != NumCoords) && (xtemp[next] == xfirst) &&
+             (ytemp[next] == yfirst))
+        next++;
+      if (NumBlocks == 0) {
+        NumNodesPerVertLine = next - index;
       }
-      if (LayerId[i] == -1) {
-        GetOStream(Warnings1) << "Warning: did not assign " << i << " to a Layer?????\n" << std::endl;
+      // The number of vertical lines must be the same on all processors
+      // TAW: Sep 14, 2015: or zero as we allow for empty processors.
+      // TEUCHOS_TEST_FOR_EXCEPTION(next-index != NumNodesPerVertLine,Exceptions::RuntimeError, "Error code only works for constant block size now!!!\n");
+      count = 0;
+      for (j = index; j < next; j++) {
+        VertLineId[OrigLoc[j]] = NumBlocks;
+        LayerId[OrigLoc[j]]    = count++;
       }
+      NumBlocks++;
+      index = next;
     }
-
-    // TAW: Sep 14 2015: relax plausibility checks as we allow for empty processors
-    //MueLu_maxAll(&comm, NumNodesPerVertLine, i);
-    //if (NumNodesPerVertLine == -1)  NumNodesPerVertLine = i;
-    //TEUCHOS_TEST_FOR_EXCEPTION(NumNodesPerVertLine != i,Exceptions::RuntimeError, "Different processors have different z direction line lengths?\n");
-
-    return NumNodesPerVertLine;
   }
 
-  /* Private member function to sort coordinates in arrays. This is an expert routine. Do not use or change.*/
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sort_coordinates(LO numCoords, LO* OrigLoc,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* xvals,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* yvals,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* zvals,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* xtemp,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* ytemp,
-                                                                                         typename Teuchos::ScalarTraits<Scalar>::coordinateType* ztemp,
-                                                                                         bool flipXY) const {
-
-    if( flipXY == false ) { // for line-smoothing
-      for (LO i = 0; i < numCoords; i++) xtemp[i]= xvals[i];
-    } else { // for semi-coarsening
-      for (LO i = 0; i < numCoords; i++) xtemp[i]= yvals[i];
-    }
-    for (LO i = 0; i < numCoords; i++) OrigLoc[i]= i;
+  /* check that everyone was assigned */
 
-    ML_az_dsort2(xtemp,numCoords,OrigLoc);
-    if( flipXY == false ) { // for line-smoothing
-      for (LO i = 0; i < numCoords; i++) ytemp[i]= yvals[OrigLoc[i]];
-    } else {
-      for (LO i = 0; i < numCoords; i++) ytemp[i]= xvals[OrigLoc[i]];
+  for (i = 0; i < Nnodes; i++) {
+    if (VertLineId[i] == -1) {
+      GetOStream(Warnings1) << "Warning: did not assign " << i << " to a vertical line?????\n"
+                            << std::endl;
     }
-
-    LO index = 0;
-
-    while ( index < numCoords ) {
-      coordinate_type xfirst = xtemp[index];
-      LO next   = index+1;
-      while ( (next != numCoords) && (xtemp[next] == xfirst))
-        next++;
-      ML_az_dsort2(&(ytemp[index]),next-index,&(OrigLoc[index]));
-      for (LO i = index; i < next; i++) ztemp[i]= zvals[OrigLoc[i]];
-      /* One final sort so that the ztemps are in order */
-      LO subindex = index;
-      while (subindex != next) {
-        coordinate_type yfirst = ytemp[subindex];
-        LO subnext = subindex+1;
-        while ( (subnext != next) && (ytemp[subnext] == yfirst)) subnext++;
-        ML_az_dsort2(&(ztemp[subindex]),subnext-subindex,&(OrigLoc[subindex]));
-        subindex = subnext;
-      }
-      index = next;
+    if (LayerId[i] == -1) {
+      GetOStream(Warnings1) << "Warning: did not assign " << i << " to a Layer?????\n"
+                            << std::endl;
     }
-
   }
 
-  /* Sort coordinates and additional array accordingly (if provided). This is an expert routine borrowed from ML. Do not change.*/
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ML_az_dsort2(typename Teuchos::ScalarTraits<Scalar>::coordinateType dlist[], LocalOrdinal N, LocalOrdinal list2[]) const {
-    LO l, r, j, i, flag;
-    LO RR2;
-    coordinate_type       dRR, dK;
-
-    // note: we use that routine for sorting coordinates only. No complex coordinates are assumed...
-    typedef Teuchos::ScalarTraits<SC> STS;
-
-    if (N <= 1) return;
-
-    l    = N / 2 + 1;
-    r    = N - 1;
-    l    = l - 1;
-    dRR  = dlist[l - 1];
-    dK   = dlist[l - 1];
+  // TAW: Sep 14 2015: relax plausibility checks as we allow for empty processors
+  // MueLu_maxAll(&comm, NumNodesPerVertLine, i);
+  // if (NumNodesPerVertLine == -1)  NumNodesPerVertLine = i;
+  // TEUCHOS_TEST_FOR_EXCEPTION(NumNodesPerVertLine != i,Exceptions::RuntimeError, "Different processors have different z direction line lengths?\n");
+
+  return NumNodesPerVertLine;
+}
+
+/* Private member function to sort coordinates in arrays. This is an expert routine. Do not use or change.*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sort_coordinates(LO numCoords, LO* OrigLoc,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* xvals,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* yvals,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* zvals,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* xtemp,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* ytemp,
+                                                                                       typename Teuchos::ScalarTraits<Scalar>::coordinateType* ztemp,
+                                                                                       bool flipXY) const {
+  if (flipXY == false) {  // for line-smoothing
+    for (LO i = 0; i < numCoords; i++) xtemp[i] = xvals[i];
+  } else {  // for semi-coarsening
+    for (LO i = 0; i < numCoords; i++) xtemp[i] = yvals[i];
+  }
+  for (LO i = 0; i < numCoords; i++) OrigLoc[i] = i;
 
-    if (list2 != NULL) {
-      RR2 = list2[l - 1];
-      while (r != 0) {
-        j = l;
-        flag = 1;
+  ML_az_dsort2(xtemp, numCoords, OrigLoc);
+  if (flipXY == false) {  // for line-smoothing
+    for (LO i = 0; i < numCoords; i++) ytemp[i] = yvals[OrigLoc[i]];
+  } else {
+    for (LO i = 0; i < numCoords; i++) ytemp[i] = xvals[OrigLoc[i]];
+  }
 
-        while (flag == 1) {
-          i = j;
-          j = j + j;
+  LO index = 0;
+
+  while (index < numCoords) {
+    coordinate_type xfirst = xtemp[index];
+    LO next                = index + 1;
+    while ((next != numCoords) && (xtemp[next] == xfirst))
+      next++;
+    ML_az_dsort2(&(ytemp[index]), next - index, &(OrigLoc[index]));
+    for (LO i = index; i < next; i++) ztemp[i] = zvals[OrigLoc[i]];
+    /* One final sort so that the ztemps are in order */
+    LO subindex = index;
+    while (subindex != next) {
+      coordinate_type yfirst = ytemp[subindex];
+      LO subnext             = subindex + 1;
+      while ((subnext != next) && (ytemp[subnext] == yfirst)) subnext++;
+      ML_az_dsort2(&(ztemp[subindex]), subnext - subindex, &(OrigLoc[subindex]));
+      subindex = subnext;
+    }
+    index = next;
+  }
+}
+
+/* Sort coordinates and additional array accordingly (if provided). This is an expert routine borrowed from ML. Do not change.*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ML_az_dsort2(typename Teuchos::ScalarTraits<Scalar>::coordinateType dlist[], LocalOrdinal N, LocalOrdinal list2[]) const {
+  LO l, r, j, i, flag;
+  LO RR2;
+  coordinate_type dRR, dK;
+
+  // note: we use that routine for sorting coordinates only. No complex coordinates are assumed...
+  typedef Teuchos::ScalarTraits<SC> STS;
+
+  if (N <= 1) return;
+
+  l   = N / 2 + 1;
+  r   = N - 1;
+  l   = l - 1;
+  dRR = dlist[l - 1];
+  dK  = dlist[l - 1];
+
+  if (list2 != NULL) {
+    RR2 = list2[l - 1];
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (STS::real(dlist[j]) > STS::real(dlist[j - 1])) j = j + 1;
 
-          if (j > r + 1)
+          if (STS::real(dlist[j - 1]) > STS::real(dK)) {
+            dlist[i - 1] = dlist[j - 1];
+            list2[i - 1] = list2[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (STS::real(dlist[j]) > STS::real(dlist[j - 1])) j = j + 1;
-
-            if (STS::real(dlist[j - 1]) > STS::real(dK)) {
-              dlist[ i - 1] = dlist[ j - 1];
-              list2[i - 1] = list2[j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
-        dlist[ i - 1] = dRR;
-        list2[i - 1] = RR2;
-
-        if (l == 1) {
-          dRR  = dlist [r];
-          RR2 = list2[r];
-          dK = dlist[r];
-          dlist[r ] = dlist[0];
-          list2[r] = list2[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          dRR  = dlist[ l - 1];
-          RR2 = list2[l - 1];
-          dK   = dlist[l - 1];
-        }
       }
-      dlist[ 0] = dRR;
-      list2[0] = RR2;
+      dlist[i - 1] = dRR;
+      list2[i - 1] = RR2;
+
+      if (l == 1) {
+        dRR      = dlist[r];
+        RR2      = list2[r];
+        dK       = dlist[r];
+        dlist[r] = dlist[0];
+        list2[r] = list2[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        dRR = dlist[l - 1];
+        RR2 = list2[l - 1];
+        dK  = dlist[l - 1];
+      }
     }
-    else {
-      while (r != 0) {
-        j = l;
-        flag = 1;
-        while (flag == 1) {
-          i = j;
-          j = j + j;
-          if (j > r + 1)
+    dlist[0] = dRR;
+    list2[0] = RR2;
+  } else {
+    while (r != 0) {
+      j    = l;
+      flag = 1;
+      while (flag == 1) {
+        i = j;
+        j = j + j;
+        if (j > r + 1)
+          flag = 0;
+        else {
+          if (j < r + 1)
+            if (STS::real(dlist[j]) > STS::real(dlist[j - 1])) j = j + 1;
+          if (STS::real(dlist[j - 1]) > STS::real(dK)) {
+            dlist[i - 1] = dlist[j - 1];
+          } else {
             flag = 0;
-          else {
-            if (j < r + 1)
-              if (STS::real(dlist[j]) > STS::real(dlist[j - 1])) j = j + 1;
-            if (STS::real(dlist[j - 1]) > STS::real(dK)) {
-              dlist[ i - 1] = dlist[ j - 1];
-            }
-            else {
-              flag = 0;
-            }
           }
         }
-        dlist[ i - 1] = dRR;
-        if (l == 1) {
-          dRR  = dlist [r];
-          dK = dlist[r];
-          dlist[r ] = dlist[0];
-          r = r - 1;
-        }
-        else {
-          l   = l - 1;
-          dRR  = dlist[ l - 1];
-          dK   = dlist[l - 1];
-        }
       }
-      dlist[ 0] = dRR;
+      dlist[i - 1] = dRR;
+      if (l == 1) {
+        dRR      = dlist[r];
+        dK       = dlist[r];
+        dlist[r] = dlist[0];
+        r        = r - 1;
+      } else {
+        l   = l - 1;
+        dRR = dlist[l - 1];
+        dK  = dlist[l - 1];
+      }
     }
-
+    dlist[0] = dRR;
   }
-} //namespace MueLu
+}
+}  // namespace MueLu
 
-#endif // MUELU_LINEDETECTIONFACTORY_DEF_HPP
+#endif  // MUELU_LINEDETECTIONFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_decl.hpp
index 0fb3650ef6a2..c16d867a3444 100644
--- a/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_decl.hpp
@@ -58,7 +58,7 @@ namespace MueLu {
 /*!
   @class LocalOrdinalTransferFactory class.
   @brief Class for transferring a vector of local ordinals from a finer level to a coarser one, where each aggregate has a unique one
-  
+
 
   ## Input/output of LocalOrdinalTransferFactory ##
 
@@ -90,77 +90,77 @@ namespace MueLu {
   | TransferVec | LocalOrdinalTransferFactory   | coarse level transfervec
 */
 
-
-
-  template<class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class LocalOrdinalTransferFactory : public TwoLevelFactoryBase {
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class LocalOrdinalTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_LOCALORDINALTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNamesOrdinal.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    // Default constructor is distabled
-    LocalOrdinalTransferFactory() = delete;
+  // Default constructor is distabled
+  LocalOrdinalTransferFactory() = delete;
 
-    /*! @brief Constructor.
+  /*! @brief Constructor.
 
-       @param vectorName The name of the quantity to be restricted.
-       @param restrictionName The name of the restriction Matrix.
+     @param vectorName The name of the quantity to be restricted.
+     @param restrictionName The name of the restriction Matrix.
 
-       The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
-       <tt>vectorName</tt>.
-    */
-    LocalOrdinalTransferFactory(const std::string & TransferVecName, const std::string & mode): TransferVecName_(TransferVecName) {
-      if(mode == "classical") useAggregatesMode_ = false;
-      else useAggregatesMode_ = true;
-    }
+     The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
+     <tt>vectorName</tt>.
+  */
+  LocalOrdinalTransferFactory(const std::string &TransferVecName, const std::string &mode)
+    : TransferVecName_(TransferVecName) {
+    if (mode == "classical")
+      useAggregatesMode_ = false;
+    else
+      useAggregatesMode_ = true;
+  }
 
-    //! Destructor.
-    virtual ~LocalOrdinalTransferFactory() { }
+  //! Destructor.
+  virtual ~LocalOrdinalTransferFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &finelevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &finelevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level & fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
+ private:
+  void BuildAggregates(Level &fineLevel, Level &coarseLevel) const;
 
-    void BuildAggregates(Level & fineLevel, Level &coarseLevel) const;
+  void BuildFC(Level &fineLevel, Level &coarseLevel) const;
 
-    void BuildFC(Level & fineLevel, Level &coarseLevel) const;
-    
-    //! Use aggregates mode (as opposed to FC mode)
-    bool useAggregatesMode_;
+  //! Use aggregates mode (as opposed to FC mode)
+  bool useAggregatesMode_;
 
-    //! The name for the vector to be transfered.  This allows us to have multiple factories for different variables
-    std::string TransferVecName_;
+  //! The name for the vector to be transfered.  This allows us to have multiple factories for different variables
+  std::string TransferVecName_;
 
-  }; // class LocalOrdinalTransferFactory
+};  // class LocalOrdinalTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_LOCALORDINALTRANSFERFACTORY_SHORT
-#endif // MUELU_LOCALORDINALTRANSFER_FACTORY_DECL_HPP
+#endif  // MUELU_LOCALORDINALTRANSFER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_def.hpp
index d870306b54b6..11228f79d0b3 100644
--- a/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_LocalOrdinalTransferFactory_def.hpp
@@ -61,201 +61,196 @@
 
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set<RCP<const FactoryBase> >(TransferVecName_,               Teuchos::null, "Factory for TransferVec generation");
-    validParamList->set<RCP<const FactoryBase> >("P Graph",                      Teuchos::null, "Factory for P generation");
-    validParamList->set<RCP<const FactoryBase> >("Aggregates",                   Teuchos::null, "Factory for aggregates generation");
-    validParamList->set<RCP<const FactoryBase> >("CoarseMap",                    Teuchos::null, "Generating factory of the coarse map");
+  validParamList->set<RCP<const FactoryBase> >(TransferVecName_, Teuchos::null, "Factory for TransferVec generation");
+  validParamList->set<RCP<const FactoryBase> >("P Graph", Teuchos::null, "Factory for P generation");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Factory for aggregates generation");
+  validParamList->set<RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
 
-    return validParamList;
-  }
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  static bool isAvailableXfer = false;
+  if (coarseLevel.GetRequestMode() == Level::REQUEST) {
+    isAvailableXfer = coarseLevel.IsAvailable(TransferVecName_, this);
+    if (isAvailableXfer == false) {
+      Input(fineLevel, TransferVecName_);
+      Input(fineLevel, "CoarseMap");
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    static bool isAvailableXfer = false;
-    if (coarseLevel.GetRequestMode() == Level::REQUEST) {
-      isAvailableXfer = coarseLevel.IsAvailable(TransferVecName_, this);
-      if (isAvailableXfer == false) {
-        Input(fineLevel, TransferVecName_);
-        Input(fineLevel, "CoarseMap");
-
-        if(useAggregatesMode_)
-          Input(fineLevel, "Aggregates");
-        else {
-          Input(coarseLevel, "P Graph");
-        }
+      if (useAggregatesMode_)
+        Input(fineLevel, "Aggregates");
+      else {
+        Input(coarseLevel, "P Graph");
       }
     }
-
   }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  if (useAggregatesMode_)
+    BuildAggregates(fineLevel, coarseLevel);
+  else
+    BuildFC(fineLevel, coarseLevel);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::BuildFC(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level & fineLevel, Level &coarseLevel) const {
-    if(useAggregatesMode_) BuildAggregates(fineLevel,coarseLevel);
-    else BuildFC(fineLevel,coarseLevel);
+  GetOStream(Runtime0) << "Transferring " << TransferVecName_ << std::endl;
+  LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  if (coarseLevel.IsAvailable(TransferVecName_, this)) {
+    GetOStream(Runtime0) << "Reusing " << TransferVecName_ << std::endl;
+    return;
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::BuildFC(Level & fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+  // Get everything we need
+  RCP<const CrsGraph> P          = Get<RCP<const CrsGraph> >(coarseLevel, "P Graph");
+  RCP<LocalOrdinalVector> fineTV = Get<RCP<LocalOrdinalVector> >(fineLevel, TransferVecName_);
+  RCP<const Map> coarseMap       = Get<RCP<const Map> >(fineLevel, "CoarseMap");
+  RCP<const Map> uniqueMap       = fineTV->getMap();
+  ArrayRCP<const LO> fineData    = fineTV->getData(0);
 
-    GetOStream(Runtime0) << "Transferring " <<TransferVecName_ << std::endl;
-    LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+  // Allocate new LO Vector
+  RCP<LocalOrdinalVector> coarseTV = LocalOrdinalVectorFactory::Build(coarseMap, 1);
+  ArrayRCP<LO> coarseData          = coarseTV->getDataNonConst(0);
 
-    if (coarseLevel.IsAvailable(TransferVecName_, this)) {
-      GetOStream(Runtime0) << "Reusing "<<TransferVecName_ << std::endl;
-      return;
-    }
+  // Invalidate everything first, to check for errors
+  for (LO i = 0; i < coarseData.size(); i++)
+    coarseData[i] = LO_INVALID;
 
-    // Get everything we need
-    RCP<const CrsGraph> P                = Get< RCP<const CrsGraph> >(coarseLevel,"P Graph");
-    RCP<LocalOrdinalVector> fineTV       = Get< RCP<LocalOrdinalVector> >(fineLevel, TransferVecName_);
-    RCP<const Map>      coarseMap  = Get< RCP<const Map> >  (fineLevel, "CoarseMap");
-    RCP<const Map>      uniqueMap  = fineTV->getMap();
-    ArrayRCP<const LO> fineData    = fineTV->getData(0);
-    
-    // Allocate new LO Vector
-    RCP<LocalOrdinalVector> coarseTV   = LocalOrdinalVectorFactory::Build(coarseMap,1);
-    ArrayRCP<LO>     coarseData = coarseTV->getDataNonConst(0);
-            
-    // Invalidate everything first, to check for errors
-    for(LO i=0; i<coarseData.size(); i++)
-      coarseData[i] = LO_INVALID;
-   
-    // Fill in coarse TV
-    LO domMapNumElements = P->getDomainMap()->getLocalNumElements();
-    for (LO row=0; row<(LO)P->getLocalNumRows(); row++) {
-      LO fineNumber = fineData[row];
-      ArrayView<const LO> indices;
-      P->getLocalRowView(row,indices);
-      
-      for(LO j=0; j<(LO)indices.size(); j++) {
-        LO col = indices[j];
-        if (col >= domMapNumElements) {
-          // skip off rank entries of P
-        } else {
-          coarseData[col] = fineNumber;
-        }
+  // Fill in coarse TV
+  LO domMapNumElements = P->getDomainMap()->getLocalNumElements();
+  for (LO row = 0; row < (LO)P->getLocalNumRows(); row++) {
+    LO fineNumber = fineData[row];
+    ArrayView<const LO> indices;
+    P->getLocalRowView(row, indices);
+
+    for (LO j = 0; j < (LO)indices.size(); j++) {
+      LO col = indices[j];
+      if (col >= domMapNumElements) {
+        // skip off rank entries of P
+      } else {
+        coarseData[col] = fineNumber;
       }
     }
+  }
 
 #ifdef HAVE_MUELU_DEBUG
-    size_t error_count = 0;
-    {
-      RCP<LocalOrdinalVector> coarseTVghosted;
-      RCP<const Import> importer = P->getImporter();
-      if (!importer.is_null()) {
-        coarseTVghosted = LocalOrdinalVectorFactory::Build(P->getColMap(),1);
-        coarseTVghosted->doImport(*coarseTV, *importer, Xpetra::INSERT);
-      } else {
-        coarseTVghosted = coarseTV;
-      }
-      ArrayRCP<LO> coarseDataGhosted = coarseTVghosted->getDataNonConst(0);
-      for (LO col=0; col<(LO)P->getColMap()->getLocalNumElements(); col++) {
-        if (coarseDataGhosted[col] == LO_INVALID)
+  size_t error_count = 0;
+  {
+    RCP<LocalOrdinalVector> coarseTVghosted;
+    RCP<const Import> importer = P->getImporter();
+    if (!importer.is_null()) {
+      coarseTVghosted = LocalOrdinalVectorFactory::Build(P->getColMap(), 1);
+      coarseTVghosted->doImport(*coarseTV, *importer, Xpetra::INSERT);
+    } else {
+      coarseTVghosted = coarseTV;
+    }
+    ArrayRCP<LO> coarseDataGhosted = coarseTVghosted->getDataNonConst(0);
+    for (LO col = 0; col < (LO)P->getColMap()->getLocalNumElements(); col++) {
+      if (coarseDataGhosted[col] == LO_INVALID)
+        error_count++;
+    }
+    for (LO row = 0; row < (LO)P->getLocalNumRows(); row++) {
+      LO fineNumber = fineData[row];
+      ArrayView<const LO> indices;
+      P->getLocalRowView(row, indices);
+      for (LO j = 0; j < (LO)indices.size(); j++) {
+        if (coarseDataGhosted[indices[j]] != fineNumber)
           error_count++;
       }
-      for (LO row=0; row<(LO)P->getLocalNumRows(); row++) {
-        LO fineNumber = fineData[row];
-        ArrayView<const LO> indices;
-        P->getLocalRowView(row,indices);
-        for(LO j=0; j<(LO)indices.size(); j++) {
-          if (coarseDataGhosted[indices[j]] != fineNumber)
-            error_count++;
-        }
-      }
     }
+  }
 
-    // Error checking:  All nodes in an aggregate must share a local ordinal
-    if(error_count > 0) {
-      std::ostringstream ofs;
-      ofs << "LocalOrdinalTransferFactory("<<TransferVecName_<<"): ERROR:  Each coarse dof must have a unique LO value.  We had "<<std::to_string(error_count)<<" unknowns that did not match.";
-      throw std::runtime_error(ofs.str());
-    }
+  // Error checking:  All nodes in an aggregate must share a local ordinal
+  if (error_count > 0) {
+    std::ostringstream ofs;
+    ofs << "LocalOrdinalTransferFactory(" << TransferVecName_ << "): ERROR:  Each coarse dof must have a unique LO value.  We had " << std::to_string(error_count) << " unknowns that did not match.";
+    throw std::runtime_error(ofs.str());
+  }
 #endif
-      
-    Set<RCP<LocalOrdinalVector> >(coarseLevel, TransferVecName_, coarseTV);
 
+  Set<RCP<LocalOrdinalVector> >(coarseLevel, TransferVecName_, coarseTV);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  GetOStream(Runtime0) << "Transferring " << TransferVecName_ << std::endl;
+  RCP<LocalOrdinalVector> coarseTV;
+  RCP<LocalOrdinalVector> fineTV;
+  LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  if (coarseLevel.IsAvailable(TransferVecName_, this)) {
+    GetOStream(Runtime0) << "Reusing " << TransferVecName_ << std::endl;
+    return;
   }
-  
 
+  RCP<Aggregates> aggregates = Get<RCP<Aggregates> >(fineLevel, "Aggregates");
+  fineTV                     = Get<RCP<LocalOrdinalVector> >(fineLevel, TransferVecName_);
+  RCP<const Map> coarseMap   = Get<RCP<const Map> >(fineLevel, "CoarseMap");
+  RCP<const Map> uniqueMap   = fineTV->getMap();
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalOrdinalTransferFactory<LocalOrdinal, GlobalOrdinal, Node>::BuildAggregates(Level & fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+  ArrayView<const GO> elementAList = coarseMap->getLocalElementList();
 
-    GetOStream(Runtime0) << "Transferring " <<TransferVecName_ << std::endl;
-    RCP<LocalOrdinalVector> coarseTV;
-    RCP<LocalOrdinalVector> fineTV;
-    LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+  coarseTV = LocalOrdinalVectorFactory::Build(coarseMap, 1);
 
-    if (coarseLevel.IsAvailable(TransferVecName_, this)) {
-      GetOStream(Runtime0) << "Reusing "<<TransferVecName_ << std::endl;
-      return;
-    }
-    
-    RCP<Aggregates>     aggregates = Get< RCP<Aggregates> > (fineLevel, "Aggregates");
-    fineTV                         = Get< RCP<LocalOrdinalVector> >(fineLevel, TransferVecName_);
-    RCP<const Map>      coarseMap  = Get< RCP<const Map> >  (fineLevel, "CoarseMap");
-    RCP<const Map>      uniqueMap  = fineTV->getMap();
-
-    ArrayView<const GO> elementAList = coarseMap->getLocalElementList();
-    
-    coarseTV   = LocalOrdinalVectorFactory::Build(coarseMap,1);
-    
-    // Create overlapped fine TV to reduce global communication
-    RCP<LocalOrdinalVector> ghostedTV = fineTV;
-    if (aggregates->AggregatesCrossProcessors()) {
-
-      RCP<const Map>    nonUniqueMap = aggregates->GetMap();
-      RCP<const Import> importer     = ImportFactory::Build(uniqueMap, nonUniqueMap);
-      
-      ghostedTV = LocalOrdinalVectorFactory::Build(nonUniqueMap, 1);
-      ghostedTV->doImport(*fineTV, *importer, Xpetra::INSERT);
-    }
-    
-    // Get some info about aggregates
-    int                         myPID        = uniqueMap->getComm()->getRank();
-    ArrayRCP<LO>                aggSizes     = aggregates->ComputeAggregateSizesArrayRCP();
-    const ArrayRCP<const LO>    vertex2AggID = aggregates->GetVertex2AggId()->getData(0);
-    const ArrayRCP<const LO>    procWinner   = aggregates->GetProcWinner()->getData(0);
-    
-
-    ArrayRCP<const LO> fineData = ghostedTV->getData(0);
-    ArrayRCP<LO>     coarseData = coarseTV->getDataNonConst(0);
-    
-    // Invalidate everything first, to check for errors
-    for(LO i=0; i<coarseData.size(); i++)
-      coarseData[i] = LO_INVALID;
-   
-    // Fill in coarse TV
-    size_t error_count = 0;
-    for (LO lnode = 0; lnode < vertex2AggID.size(); lnode++) {
-      if (procWinner[lnode] == myPID &&
-          //lnode < vertex2AggID.size() &&
-          lnode < fineData.size() && // TAW do not access off-processor data
-          vertex2AggID[lnode] < coarseData.size()) {
-        if(coarseData[vertex2AggID[lnode]] == LO_INVALID)
-          coarseData[vertex2AggID[lnode]] = fineData[lnode];
-        if(coarseData[vertex2AggID[lnode]] != fineData[lnode])
-          error_count++;
-      }
-    }
+  // Create overlapped fine TV to reduce global communication
+  RCP<LocalOrdinalVector> ghostedTV = fineTV;
+  if (aggregates->AggregatesCrossProcessors()) {
+    RCP<const Map> nonUniqueMap = aggregates->GetMap();
+    RCP<const Import> importer  = ImportFactory::Build(uniqueMap, nonUniqueMap);
+
+    ghostedTV = LocalOrdinalVectorFactory::Build(nonUniqueMap, 1);
+    ghostedTV->doImport(*fineTV, *importer, Xpetra::INSERT);
+  }
+
+  // Get some info about aggregates
+  int myPID                             = uniqueMap->getComm()->getRank();
+  ArrayRCP<LO> aggSizes                 = aggregates->ComputeAggregateSizesArrayRCP();
+  const ArrayRCP<const LO> vertex2AggID = aggregates->GetVertex2AggId()->getData(0);
+  const ArrayRCP<const LO> procWinner   = aggregates->GetProcWinner()->getData(0);
 
-    // Error checking:  All nodes in an aggregate must share a local ordinal
-    if(error_count > 0) {
-      std::ostringstream ofs;
-      ofs << "LocalOrdinalTransferFactory: ERROR:  Each aggregate must have a unique LO value.  We had "<<std::to_string(error_count)<<" unknowns that did not match.";
-      throw std::runtime_error(ofs.str());
+  ArrayRCP<const LO> fineData = ghostedTV->getData(0);
+  ArrayRCP<LO> coarseData     = coarseTV->getDataNonConst(0);
+
+  // Invalidate everything first, to check for errors
+  for (LO i = 0; i < coarseData.size(); i++)
+    coarseData[i] = LO_INVALID;
+
+  // Fill in coarse TV
+  size_t error_count = 0;
+  for (LO lnode = 0; lnode < vertex2AggID.size(); lnode++) {
+    if (procWinner[lnode] == myPID &&
+        // lnode < vertex2AggID.size() &&
+        lnode < fineData.size() &&  // TAW do not access off-processor data
+        vertex2AggID[lnode] < coarseData.size()) {
+      if (coarseData[vertex2AggID[lnode]] == LO_INVALID)
+        coarseData[vertex2AggID[lnode]] = fineData[lnode];
+      if (coarseData[vertex2AggID[lnode]] != fineData[lnode])
+        error_count++;
     }
-      
-    Set<RCP<LocalOrdinalVector> >(coarseLevel, TransferVecName_, coarseTV);
+  }
 
+  // Error checking:  All nodes in an aggregate must share a local ordinal
+  if (error_count > 0) {
+    std::ostringstream ofs;
+    ofs << "LocalOrdinalTransferFactory: ERROR:  Each aggregate must have a unique LO value.  We had " << std::to_string(error_count) << " unknowns that did not match.";
+    throw std::runtime_error(ofs.str());
   }
 
-} // namespace MueLu
+  Set<RCP<LocalOrdinalVector> >(coarseLevel, TransferVecName_, coarseTV);
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_LOCALORDINALTRANSFER_FACTORY_DEF_HPP
+#endif  // MUELU_LOCALORDINALTRANSFER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_decl.hpp
index 37ff1ce0e56e..a459dfe4b886 100644
--- a/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_decl.hpp
@@ -56,141 +56,135 @@
 
 namespace MueLu {
 
-  /*!
-    @class LowPrecisionFactory class.
-    @brief Factory for converting matrices to half precision operators
-  */
+/*!
+  @class LowPrecisionFactory class.
+  @brief Factory for converting matrices to half precision operators
+*/
 
-  template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class LowPrecisionFactory : public SingleLevelFactoryBase {
+template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class LowPrecisionFactory : public SingleLevelFactoryBase {
 #undef MUELU_LOWPRECISIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    LowPrecisionFactory() { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~LowPrecisionFactory() { }
+  LowPrecisionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~LowPrecisionFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level& currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*!
-      @brief Build method.
+  //! @name Build methods.
+  //@{
 
-      Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
+    Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-  }; //class LowPrecisionFactory
+  //@}
 
+};  // class LowPrecisionFactory
 
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT)
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LowPrecisionFactory<double,LocalOrdinal,GlobalOrdinal,Node> : public SingleLevelFactoryBase {
-    typedef double Scalar;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node> : public SingleLevelFactoryBase {
+  typedef double Scalar;
 #undef MUELU_LOWPRECISIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    LowPrecisionFactory() { }
+  LowPrecisionFactory() {}
 
-    //! Destructor.
-    virtual ~LowPrecisionFactory() { }
+  //! Destructor.
+  virtual ~LowPrecisionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+    Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  }; //class LowPrecisionFactory
+};  // class LowPrecisionFactory
 #endif
 
-
 #if defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LowPrecisionFactory<std::complex<double>,LocalOrdinal,GlobalOrdinal,Node> : public SingleLevelFactoryBase {
-    typedef std::complex<double> Scalar;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node> : public SingleLevelFactoryBase {
+  typedef std::complex<double> Scalar;
 #undef MUELU_LOWPRECISIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    LowPrecisionFactory() { }
+  LowPrecisionFactory() {}
 
-    //! Destructor.
-    virtual ~LowPrecisionFactory() { }
+  //! Destructor.
+  virtual ~LowPrecisionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+    Converts a matrix to half precision operators and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  }; //class LowPrecisionFactory
+};  // class LowPrecisionFactory
 #endif
 
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_LOWPRECISIONFACTORY_SHORT
-#endif // MUELU_LOWPRECISIONFACTORY_DECL_HPP
+#endif  // MUELU_LOWPRECISIONFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_def.hpp b/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_def.hpp
index 5182d762e949..32e41c8cc7a4 100644
--- a/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_LowPrecisionFactory_def.hpp
@@ -56,142 +56,136 @@
 #include "MueLu_Level.hpp"
 #include "MueLu_Monitor.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<std::string>("matrix key", "A", "");
-    validParamList->set< RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-
-    return validParamList;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  validParamList->set<std::string>("matrix key", "A", "");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
 
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
-    Input(currentLevel, matrixKey);
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    using Teuchos::ParameterList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
+  Input(currentLevel, matrixKey);
+}
 
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  using Teuchos::ParameterList;
 
-    FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
 
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, matrixKey);
+  FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
 
-    GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
-    Set(currentLevel, matrixKey, A);
-  }
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, matrixKey);
 
+  GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
+  Set(currentLevel, matrixKey, A);
+}
 
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT)
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<std::string>("matrix key", "A", "");
-    validParamList->set< RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-
-    return validParamList;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<std::string>("matrix key", "A", "");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
+  Input(currentLevel, matrixKey);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  using Teuchos::ParameterList;
+  using HalfScalar = typename Teuchos::ScalarTraits<Scalar>::halfPrecision;
+
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
+
+  FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
+
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, matrixKey);
+
+  if ((A->getRowMap()->lib() == Xpetra::UseTpetra) && std::is_same<Scalar, double>::value) {
+    auto tpA        = rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix(), true)->getTpetra_CrsMatrix();
+    auto tpLowA     = tpA->template convert<HalfScalar>();
+    auto tpLowOpA   = rcp(new Tpetra::CrsMatrixMultiplyOp<Scalar, HalfScalar, LocalOrdinal, GlobalOrdinal, Node>(tpLowA));
+    auto xpTpLowOpA = rcp(new TpetraOperator(tpLowOpA));
+    auto xpLowOpA   = rcp_dynamic_cast<Operator>(xpTpLowOpA);
+    Set(currentLevel, matrixKey, xpLowOpA);
+    return;
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
-    Input(currentLevel, matrixKey);
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<double, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    using Teuchos::ParameterList;
-    using HalfScalar = typename Teuchos::ScalarTraits<Scalar>::halfPrecision;
-
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
-
-    FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
-
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, matrixKey);
-
-    if ((A->getRowMap()->lib() == Xpetra::UseTpetra) && std::is_same<Scalar, double>::value) {
-      auto tpA = rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix(), true)->getTpetra_CrsMatrix();
-      auto tpLowA = tpA->template convert<HalfScalar>();
-      auto tpLowOpA = rcp(new Tpetra::CrsMatrixMultiplyOp<Scalar,HalfScalar,LocalOrdinal,GlobalOrdinal,Node>(tpLowA));
-      auto xpTpLowOpA = rcp(new TpetraOperator(tpLowOpA));
-      auto xpLowOpA = rcp_dynamic_cast<Operator>(xpTpLowOpA);
-      Set(currentLevel, matrixKey, xpLowOpA);
-      return;
-    }
-
-    GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
-    Set(currentLevel, matrixKey, A);
-  }
+  GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
+  Set(currentLevel, matrixKey, A);
+}
 #endif
 
-
 #if defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<std::string>("matrix key", "A", "");
-    validParamList->set< RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
-
-    return validParamList;
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
-    Input(currentLevel, matrixKey);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<std::string>("matrix key", "A", "");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix A to be converted to lower precision");
+
+  return validParamList;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
+  Input(currentLevel, matrixKey);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  using Teuchos::ParameterList;
+  using HalfScalar = typename Teuchos::ScalarTraits<Scalar>::halfPrecision;
+
+  const ParameterList& pL = GetParameterList();
+  std::string matrixKey   = pL.get<std::string>("matrix key");
+
+  FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
+
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, matrixKey);
+
+  if ((A->getRowMap()->lib() == Xpetra::UseTpetra) && std::is_same<Scalar, std::complex<double> >::value) {
+    auto tpA        = rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix(), true)->getTpetra_CrsMatrix();
+    auto tpLowA     = tpA->template convert<HalfScalar>();
+    auto tpLowOpA   = rcp(new Tpetra::CrsMatrixMultiplyOp<Scalar, HalfScalar, LocalOrdinal, GlobalOrdinal, Node>(tpLowA));
+    auto xpTpLowOpA = rcp(new TpetraOperator(tpLowOpA));
+    auto xpLowOpA   = rcp_dynamic_cast<Operator>(xpTpLowOpA);
+    Set(currentLevel, matrixKey, xpLowOpA);
+    return;
   }
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LowPrecisionFactory<std::complex<double>, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    using Teuchos::ParameterList;
-    using HalfScalar = typename Teuchos::ScalarTraits<Scalar>::halfPrecision;
-
-    const ParameterList& pL = GetParameterList();
-    std::string matrixKey = pL.get<std::string>("matrix key");
-
-    FactoryMonitor m(*this, "Converting " + matrixKey + " to half precision", currentLevel);
-
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, matrixKey);
-
-    if ((A->getRowMap()->lib() == Xpetra::UseTpetra) && std::is_same<Scalar, std::complex<double> >::value) {
-      auto tpA = rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix(), true)->getTpetra_CrsMatrix();
-      auto tpLowA = tpA->template convert<HalfScalar>();
-      auto tpLowOpA = rcp(new Tpetra::CrsMatrixMultiplyOp<Scalar,HalfScalar,LocalOrdinal,GlobalOrdinal,Node>(tpLowA));
-      auto xpTpLowOpA = rcp(new TpetraOperator(tpLowOpA));
-      auto xpLowOpA = rcp_dynamic_cast<Operator>(xpTpLowOpA);
-      Set(currentLevel, matrixKey, xpLowOpA);
-      return;
-    }
-
-    GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
-    Set(currentLevel, matrixKey, A);
-  }
+  GetOStream(Warnings) << "Matrix not converted to half precision. This only works for Tpetra and when both Scalar and HalfScalar have been instantiated." << std::endl;
+  Set(currentLevel, matrixKey, A);
+}
 #endif
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_LOWPRECISIONFACTORY_DEF_HPP
+#endif  // MUELU_LOWPRECISIONFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_MapTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_MapTransferFactory_decl.hpp
index 5ca98e15b840..53b93c588e60 100644
--- a/packages/muelu/src/Misc/MueLu_MapTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_MapTransferFactory_decl.hpp
@@ -52,89 +52,87 @@
 
 namespace MueLu {
 
-  /*!
-    @class MapTransferFactory class.
-    @brief Factory to transfer a map from a fine to a coarse level
+/*!
+  @class MapTransferFactory class.
+  @brief Factory to transfer a map from a fine to a coarse level
 
-    Factory that transfers a map (given by a variable name and a generating factory) for building
-    a coarse version of the map. The coarse map is stored on the coarse level using the same variable name
-    and generating factory than the original fine level map.
+  Factory that transfers a map (given by a variable name and a generating factory) for building
+  a coarse version of the map. The coarse map is stored on the coarse level using the same variable name
+  and generating factory than the original fine level map.
 
-    The transfer is based on the prolongator maps. The prolongator/nullspace can also contain rotational modes,
-    that might not be of interest for the map of interest. Use the option "nullspace vectors: limit to" to exclude such modes.
+  The transfer is based on the prolongator maps. The prolongator/nullspace can also contain rotational modes,
+  that might not be of interest for the map of interest. Use the option "nullspace vectors: limit to" to exclude such modes.
 
-    ## Input/output ##
+  ## Input/output ##
 
-    ### User parameters ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    map: name                   | string  | ""     |   | * | * | Name of the map
-    map: factory                | string  | "null" |   | * | * | Name of the generating factory
-    P                           | Factory | null   |   | * | * | Generating factory of prolongator
-    nullspace vectors: limit to | string  | "all"  |   | * | * | Use only these nullspace vectors/columns of P to transfer the map (e.g. to drop rotations)
+  ### User parameters ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  map: name                   | string  | ""     |   | * | * | Name of the map
+  map: factory                | string  | "null" |   | * | * | Name of the generating factory
+  P                           | Factory | null   |   | * | * | Generating factory of prolongator
+  nullspace vectors: limit to | string  | "all"  |   | * | * | Use only these nullspace vectors/columns of P to transfer the map (e.g. to drop rotations)
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
 
-    ### Variables provided by this factory ###
+  ### Variables provided by this factory ###
 
-    After \c Build() , the following data is available (if requested):
+  After \c Build() , the following data is available (if requested):
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | map: name | MapTransferFactory | Coarse version of the input map
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | map: name | MapTransferFactory | Coarse version of the input map
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MapTransferFactory : public TwoLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MapTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_MAPTRANSFERFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
-
-  public:
+#include "MueLu_UseShortNames.hpp"
 
-    //! Input
-    //@{
+ public:
+  //! Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const override;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const override;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    //! Build an object with this factory.
-    void Build(Level& fineLevel, Level& coarseLevel) const override;
+  //! Build an object with this factory.
+  void Build(Level& fineLevel, Level& coarseLevel) const override;
 
-    //@}
+  //@}
 
-  private:
-
-    /*!
-      @brief Get the max number of entries per row of P to be considered for map transfer
+ private:
+  /*!
+    @brief Get the max number of entries per row of P to be considered for map transfer
 
-      To exclude some nullspace vectors (e.g. rotations in 2D or 3D elasticity), when doing the map transfer,
-      this routine translates the user wish to a maximal number of entries per row of P to be considered during the map transfer.
+    To exclude some nullspace vectors (e.g. rotations in 2D or 3D elasticity), when doing the map transfer,
+    this routine translates the user wish to a maximal number of entries per row of P to be considered during the map transfer.
 
-      \warning Rows of P are looped from left to right, so we rely on the usual ordering of the nullspace vectors (translations in x/y/z, then rotations around x,y,z).
+    \warning Rows of P are looped from left to right, so we rely on the usual ordering of the nullspace vectors (translations in x/y/z, then rotations around x,y,z).
 
-      @param[in] pL Parameter list with user-given configuration
-      @return Number of entries per row of the prolongator to be used for the map transfer
-    */
-    int GetLimitOfProlongatorColumns(const ParameterList& pL) const;
+    @param[in] pL Parameter list with user-given configuration
+    @return Number of entries per row of the prolongator to be used for the map transfer
+  */
+  int GetLimitOfProlongatorColumns(const ParameterList& pL) const;
 
-    //! Generating factory of input variable
-    mutable RCP<const FactoryBase> mapFact_;
+  //! Generating factory of input variable
+  mutable RCP<const FactoryBase> mapFact_;
 
-  }; // class MapTransferFactory
+};  // class MapTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MAPTRANSFERFACTORY_SHORT
 #endif /* MUELU_MAPTRANSFERFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_MapTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_MapTransferFactory_def.hpp
index b96ee324a0df..cd508ac82a50 100644
--- a/packages/muelu/src/Misc/MueLu_MapTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_MapTransferFactory_def.hpp
@@ -59,140 +59,133 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->setEntry("map: name", Teuchos::ParameterEntry(std::string("")));
-    validParamList->setEntry("map: factory", Teuchos::ParameterEntry(std::string("null")));
-
-    validParamList->set<RCP<const FactoryBase>>("P", Teuchos::null, "Tentative prolongator factory");
-    validParamList->set<std::string>("nullspace vectors: limit to", "all", "Limit the number of nullspace vectors to be used for the map transfer (especially to exclude rotational vectors).");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->setEntry("map: name", Teuchos::ParameterEntry(std::string("")));
+  validParamList->setEntry("map: factory", Teuchos::ParameterEntry(std::string("null")));
+
+  validParamList->set<RCP<const FactoryBase>>("P", Teuchos::null, "Tentative prolongator factory");
+  validParamList->set<std::string>("nullspace vectors: limit to", "all", "Limit the number of nullspace vectors to be used for the map transfer (especially to exclude rotational vectors).");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  const ParameterList& pL       = GetParameterList();
+  const std::string mapFactName = pL.get<std::string>("map: factory");
+  const std::string mapName     = pL.get<std::string>("map: name");
+
+  if (fineLevel.GetLevelID() == 0) {
+    // Not needed, if the map is provided as user data
+    fineLevel.DeclareInput(mapName, NoFactory::get(), this);
+  } else {
+    // check whether user has provided a specific name for the MapFactory
+    if (mapFactName == "" || mapFactName == "NoFactory")
+      mapFact_ = MueLu::NoFactory::getRCP();
+    else if (mapFactName != "null")
+      mapFact_ = coarseLevel.GetFactoryManager()->GetFactory(mapFactName);
+
+    // request map generated by mapFact_
+    fineLevel.DeclareInput(mapName, mapFact_.get(), this);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    const ParameterList & pL = GetParameterList();
-    const std::string mapFactName = pL.get<std::string>("map: factory");
-    const std::string mapName = pL.get<std::string>("map: name");
-
-    if (fineLevel.GetLevelID() == 0)
-    {
-      // Not needed, if the map is provided as user data
-      fineLevel.DeclareInput(mapName, NoFactory::get(), this);
-    }
-    else
-    {
-      // check whether user has provided a specific name for the MapFactory
-      if (mapFactName == "" || mapFactName == "NoFactory")
-        mapFact_ = MueLu::NoFactory::getRCP();
-      else if (mapFactName != "null")
-        mapFact_ = coarseLevel.GetFactoryManager()->GetFactory(mapFactName);
-
-      // request map generated by mapFact_
-      fineLevel.DeclareInput(mapName, mapFact_.get(), this);
-    }
-
-    // request Ptent
-    // note that "P" provided by the user (through XML file) is supposed to be of type TentativePFactory
-    Teuchos::RCP<const FactoryBase> tentPFact = GetFactory("P");
-    if (tentPFact == Teuchos::null)
-      tentPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
-    coarseLevel.DeclareInput("P", tentPFact.get(), this);
+  // request Ptent
+  // note that "P" provided by the user (through XML file) is supposed to be of type TentativePFactory
+  Teuchos::RCP<const FactoryBase> tentPFact = GetFactory("P");
+  if (tentPFact == Teuchos::null)
+    tentPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  coarseLevel.DeclareInput("P", tentPFact.get(), this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  Monitor m(*this, "MapTransferFactory");
+
+  const ParameterList& pL     = GetParameterList();
+  const std::string mapName   = pL.get<std::string>("map: name");
+  const int maxNumProlongCols = GetLimitOfProlongatorColumns(pL);
+
+  // fetch map from level
+  RCP<const Map> transferMap = Teuchos::null;
+  if (fineLevel.GetLevelID() == 0) {
+    transferMap = fineLevel.Get<RCP<const Map>>(mapName, NoFactory::get());
+  } else {
+    if (fineLevel.IsAvailable(mapName, mapFact_.get()) == false)
+      GetOStream(Runtime0) << "MapTransferFactory::Build: User provided map \"" << mapName << "\" not found in Level class on level " << fineLevel.GetLevelID() << "." << std::endl;
+    transferMap = fineLevel.Get<RCP<const Map>>(mapName, mapFact_.get());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    Monitor m(*this, "MapTransferFactory");
-
-    const ParameterList & pL = GetParameterList();
-    const std::string mapName = pL.get<std::string>("map: name");
-    const int maxNumProlongCols = GetLimitOfProlongatorColumns(pL);
-
-    // fetch map from level
-    RCP<const Map> transferMap = Teuchos::null;
-    if (fineLevel.GetLevelID() == 0) {
-      transferMap = fineLevel.Get<RCP<const Map>>(mapName, NoFactory::get());
-    } else {
-      if (fineLevel.IsAvailable(mapName, mapFact_.get()) == false)
-        GetOStream(Runtime0) << "MapTransferFactory::Build: User provided map \"" << mapName << "\" not found in Level class on level " << fineLevel.GetLevelID() << "." << std::endl;
-      transferMap = fineLevel.Get<RCP<const Map>>(mapName, mapFact_.get());
-    }
-
-    // Get default tentative prolongator factory
-    // Getting it that way ensures that the same factory instance will be used for both SaPFactory and NullspaceFactory.
-    RCP<const FactoryBase> tentPFact = GetFactory("P");
-    if (tentPFact == Teuchos::null)
-      tentPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
-    TEUCHOS_TEST_FOR_EXCEPTION(!coarseLevel.IsAvailable("P", tentPFact.get()), Exceptions::RuntimeError,
-        "MueLu::MapTransferFactory::Build(): P (generated by TentativePFactory) not available.");
-    RCP<Matrix> Ptent = coarseLevel.Get<RCP<Matrix> >("P", tentPFact.get());
-
-    // loop over local rows of Ptent and figure out the corresponding coarse GIDs
-    Array<GO > coarseMapGids;
-    RCP<const Map> prolongColMap = Ptent->getColMap();
-    GO gRowID = -1;
-    int numColEntries = 0;
-    for (size_t row = 0; row < Ptent->getLocalNumRows(); ++row) {
-      gRowID = Ptent->getRowMap()->getGlobalElement(row);
-
-      if (transferMap->isNodeGlobalElement(gRowID)) {
-        Teuchos::ArrayView<const LO> indices;
-        Teuchos::ArrayView<const SC> vals;
-        Ptent->getLocalRowView(row, indices, vals);
-
-        numColEntries = as<int>(indices.size());
-        if (maxNumProlongCols > 0)
-          numColEntries = std::min(numColEntries, maxNumProlongCols);
-
-        for (size_t col = 0; col < as<size_t>(numColEntries); ++col) {
-          // mark all (selected) columns in Ptent(gRowID,*) to be coarse Dofs of next level transferMap
-          GO gcid = prolongColMap->getGlobalElement(indices[col]);
-          coarseMapGids.push_back(gcid);
-        }
+  // Get default tentative prolongator factory
+  // Getting it that way ensures that the same factory instance will be used for both SaPFactory and NullspaceFactory.
+  RCP<const FactoryBase> tentPFact = GetFactory("P");
+  if (tentPFact == Teuchos::null)
+    tentPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  TEUCHOS_TEST_FOR_EXCEPTION(!coarseLevel.IsAvailable("P", tentPFact.get()), Exceptions::RuntimeError,
+                             "MueLu::MapTransferFactory::Build(): P (generated by TentativePFactory) not available.");
+  RCP<Matrix> Ptent = coarseLevel.Get<RCP<Matrix>>("P", tentPFact.get());
+
+  // loop over local rows of Ptent and figure out the corresponding coarse GIDs
+  Array<GO> coarseMapGids;
+  RCP<const Map> prolongColMap = Ptent->getColMap();
+  GO gRowID                    = -1;
+  int numColEntries            = 0;
+  for (size_t row = 0; row < Ptent->getLocalNumRows(); ++row) {
+    gRowID = Ptent->getRowMap()->getGlobalElement(row);
+
+    if (transferMap->isNodeGlobalElement(gRowID)) {
+      Teuchos::ArrayView<const LO> indices;
+      Teuchos::ArrayView<const SC> vals;
+      Ptent->getLocalRowView(row, indices, vals);
+
+      numColEntries = as<int>(indices.size());
+      if (maxNumProlongCols > 0)
+        numColEntries = std::min(numColEntries, maxNumProlongCols);
+
+      for (size_t col = 0; col < as<size_t>(numColEntries); ++col) {
+        // mark all (selected) columns in Ptent(gRowID,*) to be coarse Dofs of next level transferMap
+        GO gcid = prolongColMap->getGlobalElement(indices[col]);
+        coarseMapGids.push_back(gcid);
       }
     }
-
-    // build coarse version of the input map
-    const GO INVALID = Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
-    std::sort(coarseMapGids.begin(), coarseMapGids.end());
-    coarseMapGids.erase(std::unique(coarseMapGids.begin(), coarseMapGids.end()), coarseMapGids.end());
-    RCP<const Map> coarseTransferMap = MapFactory::Build(prolongColMap->lib(), INVALID, coarseMapGids(),
-        prolongColMap->getIndexBase(), prolongColMap->getComm());
-
-    // store map in coarse level
-    if (fineLevel.GetLevelID() == 0)
-    {
-      const std::string mapFactName = pL.get<std::string>("map: factory");
-      RCP<const FactoryBase> mapFact = coarseLevel.GetFactoryManager()->GetFactory(mapFactName);
-      coarseLevel.Set(mapName, coarseTransferMap, mapFact.get());
-    }
-    else
-      coarseLevel.Set(mapName, coarseTransferMap, mapFact_.get());
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLimitOfProlongatorColumns(const ParameterList& pL) const
-  {
-    const std::string useTheseNspVectors = pL.get<std::string>("nullspace vectors: limit to");
-
-    // Leave right away, if no limit is prescribed by the user
-    if (useTheseNspVectors == "all" || useTheseNspVectors == "")
-      return -1;
-
-    // Simplify? Maybe replace by boolean flag "nullspace: exclude rotations"
-    int maxNumProlongCols = -1;
-    if (useTheseNspVectors == "translations")
-      maxNumProlongCols = 1;
-    else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown subset of nullspace vectors to be used, when performing a map transfer.")
-
-    return maxNumProlongCols;
   }
 
-} // namespace MueLu
+  // build coarse version of the input map
+  const GO INVALID = Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
+  std::sort(coarseMapGids.begin(), coarseMapGids.end());
+  coarseMapGids.erase(std::unique(coarseMapGids.begin(), coarseMapGids.end()), coarseMapGids.end());
+  RCP<const Map> coarseTransferMap = MapFactory::Build(prolongColMap->lib(), INVALID, coarseMapGids(),
+                                                       prolongColMap->getIndexBase(), prolongColMap->getComm());
+
+  // store map in coarse level
+  if (fineLevel.GetLevelID() == 0) {
+    const std::string mapFactName  = pL.get<std::string>("map: factory");
+    RCP<const FactoryBase> mapFact = coarseLevel.GetFactoryManager()->GetFactory(mapFactName);
+    coarseLevel.Set(mapName, coarseTransferMap, mapFact.get());
+  } else
+    coarseLevel.Set(mapName, coarseTransferMap, mapFact_.get());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int MapTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLimitOfProlongatorColumns(const ParameterList& pL) const {
+  const std::string useTheseNspVectors = pL.get<std::string>("nullspace vectors: limit to");
+
+  // Leave right away, if no limit is prescribed by the user
+  if (useTheseNspVectors == "all" || useTheseNspVectors == "")
+    return -1;
+
+  // Simplify? Maybe replace by boolean flag "nullspace: exclude rotations"
+  int maxNumProlongCols = -1;
+  if (useTheseNspVectors == "translations")
+    maxNumProlongCols = 1;
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown subset of nullspace vectors to be used, when performing a map transfer.")
+
+  return maxNumProlongCols;
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_MAPTRANSFERFACTORY_DEF_HPP_ */
\ No newline at end of file
diff --git a/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_decl.hpp
index d5d4b459d55f..21ad1a067fb0 100644
--- a/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_decl.hpp
@@ -58,49 +58,45 @@
 #include "MueLu_FactoryBase_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class MergedBlockedMatrix
-    @brief Factory provides a merged version of a blocked matrix
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MergedBlockedMatrixFactory : public SingleLevelFactoryBase {
+/*!
+  @class MergedBlockedMatrix
+  @brief Factory provides a merged version of a blocked matrix
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MergedBlockedMatrixFactory : public SingleLevelFactoryBase {
 #undef MUELU_MERGEDBLOCKEDMATRIXFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    MergedBlockedMatrixFactory();
+  MergedBlockedMatrixFactory();
 
-    virtual ~MergedBlockedMatrixFactory() { }
-    //@}
+  virtual ~MergedBlockedMatrixFactory() {}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &currentLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &currentLevel) const;
+  //@}
 
+ private:
+};  // class MergedBlockedMatrixFactory
 
-
-  private:
-
-
-  }; //class MergedBlockedMatrixFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MERGEDBLOCKEDMATRIXFACTORY_SHORT
 
diff --git a/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_def.hpp b/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_def.hpp
index ce3a8cf46d0e..b36bed3f75de 100644
--- a/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_MergedBlockedMatrixFactory_def.hpp
@@ -56,28 +56,25 @@
 namespace MueLu {
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedBlockedMatrixFactory()
-{ }
+MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedBlockedMatrixFactory() {}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const ParameterList> MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  validParamList->set< RCP<const FactoryBase> >("A", MueLu::NoFactory::getRCP()/*Teuchos::null*/, "Generating factory of the matrix A used for building SchurComplement (must be a 2x2 blocked operator, default = MueLu::NoFactory::getRCP())");
+  validParamList->set<RCP<const FactoryBase> >("A", MueLu::NoFactory::getRCP() /*Teuchos::null*/, "Generating factory of the matrix A used for building SchurComplement (must be a 2x2 blocked operator, default = MueLu::NoFactory::getRCP())");
 
   return validParamList;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
   Input(currentLevel, "A");
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & currentLevel) const
-{
-  FactoryMonitor  m(*this, "MergedBlockedMatrix", currentLevel);
+void MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "MergedBlockedMatrix", currentLevel);
   Teuchos::RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
 
   RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A);
@@ -93,6 +90,6 @@ void MergedBlockedMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Buil
   }
 }
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_MERGEDBLOCKEDMATRIXFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_decl.hpp
index 75f8fdc123d2..a499fae0ac42 100644
--- a/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_decl.hpp
@@ -55,71 +55,70 @@
 
 namespace MueLu {
 
-  /*!
-    @class MultiVectorTransferFactory class.
-    @brief Class for restricting a MultiVector from a finer to a coarser level.
-
-    This is to be used in conjunction with Muelu::RAPFactory::AddTransferFactory().
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MultiVectorTransferFactory : public TwoLevelFactoryBase {
+/*!
+  @class MultiVectorTransferFactory class.
+  @brief Class for restricting a MultiVector from a finer to a coarser level.
+
+  This is to be used in conjunction with Muelu::RAPFactory::AddTransferFactory().
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MultiVectorTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_MULTIVECTORTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
+  /*! @brief Constructor.
 
-       @param vectorName The name of the quantity to be restricted.
-       @param restrictionName The name of the restriction Matrix.
+     @param vectorName The name of the quantity to be restricted.
+     @param restrictionName The name of the restriction Matrix.
 
-       The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
-       <tt>vectorName</tt>.
-    */
-    MultiVectorTransferFactory() { }
-
-    MultiVectorTransferFactory(std::string const & vectorName); // deprecated
+     The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
+     <tt>vectorName</tt>.
+  */
+  MultiVectorTransferFactory() {}
 
-    //! Destructor.
-    virtual ~MultiVectorTransferFactory() { }
+  MultiVectorTransferFactory(std::string const &vectorName);  // deprecated
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~MultiVectorTransferFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
+  //@}
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //! @name Input
+  //@{
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &finelevel, Level &coarseLevel) const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &finelevel, Level &coarseLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level & fineLevel, Level &coarseLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-  private:
+  //@}
 
-    static ArrayRCP<SC> expandCoordinates(ArrayRCP<SC> coord, LocalOrdinal blksize);
+ private:
+  static ArrayRCP<SC> expandCoordinates(ArrayRCP<SC> coord, LocalOrdinal blksize);
 
-  }; // class MultiVectorTransferFactory
+};  // class MultiVectorTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MULTIVECTORTRANSFERFACTORY_SHORT
-#endif // MUELU_MULTIVECTORTRANSFER_FACTORY_DECL_HPP
+#endif  // MUELU_MULTIVECTORTRANSFER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_def.hpp b/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_def.hpp
index 4ecf3bedfc09..00c1a91bd7c7 100644
--- a/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_MultiVectorTransferFactory_def.hpp
@@ -54,73 +54,73 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< std::string >           ("Vector name",      "undefined", "Name of the vector that will be transferred on the coarse grid (level key)"); // TODO: how to set a validator without default value?
-    validParamList->set< RCP<const FactoryBase> >("Vector factory", Teuchos::null, "Factory of the vector");
-    validParamList->set< RCP<const FactoryBase> >("R",              Teuchos::null, "Factory of the transfer operator (restriction)");
+  validParamList->set<std::string>("Vector name", "undefined", "Name of the vector that will be transferred on the coarse grid (level key)");  // TODO: how to set a validator without default value?
+  validParamList->set<RCP<const FactoryBase> >("Vector factory", Teuchos::null, "Factory of the vector");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Factory of the transfer operator (restriction)");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiVectorTransferFactory(std::string const & vectorName) {
-    SetParameter("Vector name", ParameterEntry(vectorName));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiVectorTransferFactory(std::string const &vectorName) {
+  SetParameter("Vector name", ParameterEntry(vectorName));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    const ParameterList & pL = GetParameterList();
-    std::string vectorName   = pL.get<std::string>("Vector name");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  const ParameterList &pL = GetParameterList();
+  std::string vectorName  = pL.get<std::string>("Vector name");
 
-    fineLevel.DeclareInput(vectorName, GetFactory("Vector factory").get(), this);
-    Input(coarseLevel, "R");
-  }
+  fineLevel.DeclareInput(vectorName, GetFactory("Vector factory").get(), this);
+  Input(coarseLevel, "R");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-    const ParameterList & pL = GetParameterList();
-    std::string vectorName   = pL.get<std::string>("Vector name");
+  const ParameterList &pL = GetParameterList();
+  std::string vectorName  = pL.get<std::string>("Vector name");
 
-    RCP<MultiVector> fineVector = fineLevel.Get< RCP<MultiVector> >(vectorName, GetFactory("Vector factory").get());
-    RCP<Matrix>      transferOp = Get<RCP<Matrix> >(coarseLevel, "R");
+  RCP<MultiVector> fineVector = fineLevel.Get<RCP<MultiVector> >(vectorName, GetFactory("Vector factory").get());
+  RCP<Matrix> transferOp      = Get<RCP<Matrix> >(coarseLevel, "R");
 
-    RCP<MultiVector> coarseVector = MultiVectorFactory::Build(transferOp->getRangeMap(), fineVector->getNumVectors());
-    GetOStream(Runtime0) << "Transferring multivector \"" << vectorName << "\"" << std::endl;
+  RCP<MultiVector> coarseVector = MultiVectorFactory::Build(transferOp->getRangeMap(), fineVector->getNumVectors());
+  GetOStream(Runtime0) << "Transferring multivector \"" << vectorName << "\"" << std::endl;
 
-    RCP<MultiVector> onesVector = MultiVectorFactory::Build(transferOp->getDomainMap(), 1);
-    onesVector->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    RCP<MultiVector> rowSumVector = MultiVectorFactory::Build(transferOp->getRangeMap(), 1);
-    transferOp->apply(*onesVector, *rowSumVector);
-    transferOp->apply(*fineVector, *coarseVector);
+  RCP<MultiVector> onesVector = MultiVectorFactory::Build(transferOp->getDomainMap(), 1);
+  onesVector->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+  RCP<MultiVector> rowSumVector = MultiVectorFactory::Build(transferOp->getRangeMap(), 1);
+  transferOp->apply(*onesVector, *rowSumVector);
+  transferOp->apply(*fineVector, *coarseVector);
 
-    if (vectorName == "Coordinates")
-      TEUCHOS_TEST_FOR_EXCEPTION(true,Exceptions::RuntimeError,"Use CoordinatesTransferFactory to transfer coordinates instead of MultiVectorTransferFactory.");
+  if (vectorName == "Coordinates")
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Use CoordinatesTransferFactory to transfer coordinates instead of MultiVectorTransferFactory.");
 
-    Set<RCP<MultiVector> >(coarseLevel, vectorName, coarseVector);
+  Set<RCP<MultiVector> >(coarseLevel, vectorName, coarseVector);
 
-  } // Build
+}  // Build
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ArrayRCP<Scalar> MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::expandCoordinates(ArrayRCP<SC> coordinates, LocalOrdinal blksize) {
-    if (blksize == 1)
-      return coordinates;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ArrayRCP<Scalar> MultiVectorTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::expandCoordinates(ArrayRCP<SC> coordinates, LocalOrdinal blksize) {
+  if (blksize == 1)
+    return coordinates;
 
-    ArrayRCP<SC> expandCoord(coordinates.size()*blksize); //TODO: how to avoid automatic initialization of the vector? using arcp()?
+  ArrayRCP<SC> expandCoord(coordinates.size() * blksize);  // TODO: how to avoid automatic initialization of the vector? using arcp()?
 
-    for(int i=0; i<coordinates.size(); i++) {
-      for(int j=0; j< blksize; j++) {
-        expandCoord[i*blksize + j] = coordinates[i];
-      }
+  for (int i = 0; i < coordinates.size(); i++) {
+    for (int j = 0; j < blksize; j++) {
+      expandCoord[i * blksize + j] = coordinates[i];
     }
-    return expandCoord;
+  }
+  return expandCoord;
 
-  } // expandCoordinates
+}  // expandCoordinates
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_MULTIVECTORTRANSFER_FACTORY_DEF_HPP
+#endif  // MUELU_MULTIVECTORTRANSFER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_RAPFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_RAPFactory_decl.hpp
index ef345754bf94..b46ad578d1c4 100644
--- a/packages/muelu/src/Misc/MueLu_RAPFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_RAPFactory_decl.hpp
@@ -63,76 +63,74 @@
 #include "MueLu_Utilities_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class RAPFactory
-    @brief Factory for building coarse matrices.
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RAPFactory : public TwoLevelFactoryBase {
+/*!
+  @class RAPFactory
+  @brief Factory for building coarse matrices.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RAPFactory : public TwoLevelFactoryBase {
 #undef MUELU_RAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    RAPFactory();
+  RAPFactory();
 
-    virtual ~RAPFactory() { }
+  virtual ~RAPFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& fineLevel, Level& coarseLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@{
-    /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
+  //@{
+  /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
 
-    Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
-    the next coarser level.
-    */
-    void AddTransferFactory(const RCP<const FactoryBase>& factory);
+  Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
+  the next coarser level.
+  */
+  void AddTransferFactory(const RCP<const FactoryBase>& factory);
 
-    // TODO add a function to remove a specific transfer factory?
+  // TODO add a function to remove a specific transfer factory?
 
-    //! Returns number of transfer factories.
-    size_t NumTransferFactories() const { return transferFacts_.size(); }
+  //! Returns number of transfer factories.
+  size_t NumTransferFactories() const { return transferFacts_.size(); }
 
-    //@}
+  //@}
 
-  private:
+ private:
+  //@{
 
-    //@{
-    
-    mutable
-    bool hasDeclaredInput_;
+  mutable bool hasDeclaredInput_;
 
-    //@}
+  //@}
 
-    //@{
+  //@{
 
-    //! list of user-defined transfer Factories
-    std::vector<RCP<const FactoryBase> > transferFacts_;
+  //! list of user-defined transfer Factories
+  std::vector<RCP<const FactoryBase> > transferFacts_;
 
-    //@}
+  //@}
 
-  }; //class RAPFactory
+};  // class RAPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RAPFACTORY_SHORT
-#endif // MUELU_RAPFACTORY_DECL_HPP
+#endif  // MUELU_RAPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_RAPFactory_def.hpp b/packages/muelu/src/Misc/MueLu_RAPFactory_def.hpp
index 5e67cc295d48..1873c5176e02 100644
--- a/packages/muelu/src/Misc/MueLu_RAPFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_RAPFactory_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_RAPFACTORY_DEF_HPP
 #define MUELU_RAPFACTORY_DEF_HPP
 
-
 #include <sstream>
 
 #include <Xpetra_Matrix.hpp>
@@ -67,342 +66,349 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RAPFactory()
-    : hasDeclaredInput_(false) { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RAPFactory()
+  : hasDeclaredInput_(false) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("transpose: use implicit");
-    SET_VALID_ENTRY("rap: triple product");
-    SET_VALID_ENTRY("rap: fix zero diagonals");
-    SET_VALID_ENTRY("rap: fix zero diagonals threshold");
-    SET_VALID_ENTRY("rap: fix zero diagonals replacement");
-    SET_VALID_ENTRY("rap: relative diagonal floor");
-#undef  SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A",                   null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("P",                   null, "Prolongator factory");
-    validParamList->set< RCP<const FactoryBase> >("R",                   null, "Restrictor factory");
-
-    validParamList->set< bool >                  ("CheckMainDiagonal",  false, "Check main diagonal for zeros");
-    validParamList->set< bool >                  ("RepairMainDiagonal", false, "Repair zeros on main diagonal");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
-  }
+  SET_VALID_ENTRY("transpose: use implicit");
+  SET_VALID_ENTRY("rap: triple product");
+  SET_VALID_ENTRY("rap: fix zero diagonals");
+  SET_VALID_ENTRY("rap: fix zero diagonals threshold");
+  SET_VALID_ENTRY("rap: fix zero diagonals replacement");
+  SET_VALID_ENTRY("rap: relative diagonal floor");
+#undef SET_VALID_ENTRY
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase> >("P", null, "Prolongator factory");
+  validParamList->set<RCP<const FactoryBase> >("R", null, "Restrictor factory");
+
+  validParamList->set<bool>("CheckMainDiagonal", false, "Check main diagonal for zeros");
+  validParamList->set<bool>("RepairMainDiagonal", false, "Repair zeros on main diagonal");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  const Teuchos::ParameterList& pL = GetParameterList();
+  if (pL.get<bool>("transpose: use implicit") == false)
+    Input(coarseLevel, "R");
+
+  Input(fineLevel, "A");
+  Input(coarseLevel, "P");
+
+  // call DeclareInput of all user-given transfer factories
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it)
+    (*it)->CallDeclareInput(coarseLevel);
+
+  hasDeclaredInput_ = true;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  const bool doTranspose       = true;
+  const bool doFillComplete    = true;
+  const bool doOptimizeStorage = true;
+  RCP<Matrix> Ac;
+  {
+    FactoryMonitor m(*this, "Computing Ac", coarseLevel);
+    std::ostringstream levelstr;
+    levelstr << coarseLevel.GetLevelID();
+    std::string labelstr = FormattingHelper::getColonLabel(coarseLevel.getObjectLabel());
+
+    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_ == false, Exceptions::RuntimeError,
+                               "MueLu::RAPFactory::Build(): CallDeclareInput has not been called before Build!");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
     const Teuchos::ParameterList& pL = GetParameterList();
-    if (pL.get<bool>("transpose: use implicit") == false)
-      Input(coarseLevel, "R");
-
-    Input(fineLevel,   "A");
-    Input(coarseLevel, "P");
-
-    // call DeclareInput of all user-given transfer factories
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it)
-      (*it)->CallDeclareInput(coarseLevel);
-
-    hasDeclaredInput_ = true;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    const bool doTranspose       = true;
-    const bool doFillComplete    = true;
-    const bool doOptimizeStorage = true;
-    RCP<Matrix> Ac;
-    {
-      FactoryMonitor m(*this, "Computing Ac", coarseLevel);
-      std::ostringstream levelstr;
-      levelstr << coarseLevel.GetLevelID();
-      std::string labelstr = FormattingHelper::getColonLabel(coarseLevel.getObjectLabel());
-
-      TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_ == false, Exceptions::RuntimeError,
-        "MueLu::RAPFactory::Build(): CallDeclareInput has not been called before Build!");
-
-      const Teuchos::ParameterList& pL = GetParameterList();
-      RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel,   "A");
-      RCP<Matrix> P = Get< RCP<Matrix> >(coarseLevel, "P"), AP;
-      // We don't have a valid P (e.g., # global aggregates = 0) so we bail.
-      // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-      if (P == Teuchos::null) {
-        Ac = Teuchos::null;
-        Set(coarseLevel, "A", Ac);
-        return;
-      }
+    RCP<Matrix> A                    = Get<RCP<Matrix> >(fineLevel, "A");
+    RCP<Matrix> P                    = Get<RCP<Matrix> >(coarseLevel, "P"), AP;
+    // We don't have a valid P (e.g., # global aggregates = 0) so we bail.
+    // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+    if (P == Teuchos::null) {
+      Ac = Teuchos::null;
+      Set(coarseLevel, "A", Ac);
+      return;
+    }
 
-      bool isEpetra = A->getRowMap()->lib() == Xpetra::UseEpetra;
-      bool isGPU =
+    bool isEpetra = A->getRowMap()->lib() == Xpetra::UseEpetra;
+    bool isGPU =
 #ifdef KOKKOS_ENABLE_CUDA
-	(typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosCudaWrapperNode).name()) ||
+        (typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosCudaWrapperNode).name()) ||
 #endif
 #ifdef KOKKOS_ENABLE_HIP
-	(typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosHIPWrapperNode).name()) ||
+        (typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosHIPWrapperNode).name()) ||
 #endif
 #ifdef KOKKOS_ENABLE_SYCL
-	(typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosSYCLWrapperNode).name()) ||
+        (typeid(Node).name() == typeid(Tpetra::KokkosCompat::KokkosSYCLWrapperNode).name()) ||
 #endif
-	false;
+        false;
 
-      if (pL.get<bool>("rap: triple product") == false || isEpetra || isGPU) {
-        if (pL.get<bool>("rap: triple product") && isEpetra)
-          GetOStream(Warnings1) << "Switching from triple product to R x (A x P) since triple product has not been implemented for Epetra.\n";
+    if (pL.get<bool>("rap: triple product") == false || isEpetra || isGPU) {
+      if (pL.get<bool>("rap: triple product") && isEpetra)
+        GetOStream(Warnings1) << "Switching from triple product to R x (A x P) since triple product has not been implemented for Epetra.\n";
 #if defined(KOKKOS_ENABLE_CUDA) || defined(KOKKOS_ENABLE_HIP) || defined(KOKKOS_ENABLE_SYCL)
-        if (pL.get<bool>("rap: triple product") && isGPU)
-          GetOStream(Warnings1) << "Switching from triple product to R x (A x P) since triple product has not been implemented for "
-				<< Node::execution_space::name() << std::endl;
+      if (pL.get<bool>("rap: triple product") && isGPU)
+        GetOStream(Warnings1) << "Switching from triple product to R x (A x P) since triple product has not been implemented for "
+                              << Node::execution_space::name() << std::endl;
 #endif
 
-        // Reuse pattern if available (multiple solve)
-        RCP<ParameterList> APparams = rcp(new ParameterList);
-        if(pL.isSublist("matrixmatrix: kernel params"))
-          APparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
-
-        // By default, we don't need global constants for A*P
-        APparams->set("compute global constants: temporaries",APparams->get("compute global constants: temporaries",false));
-        APparams->set("compute global constants",APparams->get("compute global constants",false));
-
-        if (coarseLevel.IsAvailable("AP reuse data", this)) {
-          GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
-
-          APparams = coarseLevel.Get< RCP<ParameterList> >("AP reuse data", this);
-
-          if (APparams->isParameter("graph"))
-            AP = APparams->get< RCP<Matrix> >("graph");
-        }
-
-        {
-          SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
-
-          AP = MatrixMatrix::Multiply(*A, !doTranspose, *P, !doTranspose, AP, GetOStream(Statistics2),
-                                      doFillComplete, doOptimizeStorage, labelstr+std::string("MueLu::A*P-")+levelstr.str(), APparams);
-        }
-
-        // Reuse coarse matrix memory if available (multiple solve)
-        RCP<ParameterList> RAPparams = rcp(new ParameterList);
-        if(pL.isSublist("matrixmatrix: kernel params"))
-          RAPparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
-
-        if (coarseLevel.IsAvailable("RAP reuse data", this)) {
-          GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous RAP data" << std::endl;
-
-          RAPparams = coarseLevel.Get< RCP<ParameterList> >("RAP reuse data", this);
-
-          if (RAPparams->isParameter("graph"))
-            Ac = RAPparams->get< RCP<Matrix> >("graph");
-
-          // Some eigenvalue may have been cached with the matrix in the previous run.
-          // As the matrix values will be updated, we need to reset the eigenvalue.
-          Ac->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
-        }
-
-        // We *always* need global constants for the RAP, but not for the temps
-        RAPparams->set("compute global constants: temporaries",RAPparams->get("compute global constants: temporaries",false));
-        RAPparams->set("compute global constants",true);
-
-        // Allow optimization of storage.
-        // This is necessary for new faster Epetra MM kernels.
-        // Seems to work with matrix modifications to repair diagonal entries.
-
-        if (pL.get<bool>("transpose: use implicit") == true) {
-          SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
-
-          Ac = MatrixMatrix::Multiply(*P,  doTranspose, *AP, !doTranspose, Ac, GetOStream(Statistics2),
-                                      doFillComplete, doOptimizeStorage, labelstr+std::string("MueLu::R*(AP)-implicit-")+levelstr.str(), RAPparams);
-
-        } else {
-          RCP<Matrix> R = Get< RCP<Matrix> >(coarseLevel, "R");
-
-          SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
-
-          Ac = MatrixMatrix::Multiply(*R, !doTranspose, *AP, !doTranspose, Ac, GetOStream(Statistics2),
-                                      doFillComplete, doOptimizeStorage, labelstr+std::string("MueLu::R*(AP)-explicit-")+levelstr.str(), RAPparams);
-        }
-
-        Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
-        if(relativeFloor.size() > 0) {
-          Xpetra::MatrixUtils<SC,LO,GO,NO>::RelativeDiagonalBoost(Ac, relativeFloor,GetOStream(Statistics2));
-        }
-
-        bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
-        bool checkAc             = pL.get<bool>("CheckMainDiagonal")|| pL.get<bool>("rap: fix zero diagonals"); ;
-        if (checkAc || repairZeroDiagonals) {
-          using magnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-          magnitudeType threshold;
-          if (pL.isType<magnitudeType>("rap: fix zero diagonals threshold"))
-            threshold = pL.get<magnitudeType>("rap: fix zero diagonals threshold");
-          else
-            threshold = Teuchos::as<magnitudeType>(pL.get<double>("rap: fix zero diagonals threshold"));
-          Scalar replacement = Teuchos::as<Scalar>(pL.get<double>("rap: fix zero diagonals replacement"));
-          Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1), threshold, replacement);
-        }
-
-        if (IsPrint(Statistics2)) {
-          RCP<ParameterList> params = rcp(new ParameterList());;
-          params->set("printLoadBalancingInfo", true);
-          params->set("printCommInfo",          true);
-          GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
-        }
-
-        if(!Ac.is_null()) {std::ostringstream oss; oss << "A_" << coarseLevel.GetLevelID(); Ac->setObjectLabel(oss.str());}
-        Set(coarseLevel, "A",         Ac);
-
-        if (!isGPU) {
-          APparams->set("graph", AP);
-          Set(coarseLevel, "AP reuse data",  APparams);
-        }
-        if (!isGPU) {
-          RAPparams->set("graph", Ac);
-          Set(coarseLevel, "RAP reuse data", RAPparams);
-        }
+      // Reuse pattern if available (multiple solve)
+      RCP<ParameterList> APparams = rcp(new ParameterList);
+      if (pL.isSublist("matrixmatrix: kernel params"))
+        APparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
+
+      // By default, we don't need global constants for A*P
+      APparams->set("compute global constants: temporaries", APparams->get("compute global constants: temporaries", false));
+      APparams->set("compute global constants", APparams->get("compute global constants", false));
+
+      if (coarseLevel.IsAvailable("AP reuse data", this)) {
+        GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
+
+        APparams = coarseLevel.Get<RCP<ParameterList> >("AP reuse data", this);
+
+        if (APparams->isParameter("graph"))
+          AP = APparams->get<RCP<Matrix> >("graph");
+      }
+
+      {
+        SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
+
+        AP = MatrixMatrix::Multiply(*A, !doTranspose, *P, !doTranspose, AP, GetOStream(Statistics2),
+                                    doFillComplete, doOptimizeStorage, labelstr + std::string("MueLu::A*P-") + levelstr.str(), APparams);
+      }
+
+      // Reuse coarse matrix memory if available (multiple solve)
+      RCP<ParameterList> RAPparams = rcp(new ParameterList);
+      if (pL.isSublist("matrixmatrix: kernel params"))
+        RAPparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
+
+      if (coarseLevel.IsAvailable("RAP reuse data", this)) {
+        GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous RAP data" << std::endl;
+
+        RAPparams = coarseLevel.Get<RCP<ParameterList> >("RAP reuse data", this);
+
+        if (RAPparams->isParameter("graph"))
+          Ac = RAPparams->get<RCP<Matrix> >("graph");
+
+        // Some eigenvalue may have been cached with the matrix in the previous run.
+        // As the matrix values will be updated, we need to reset the eigenvalue.
+        Ac->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
+      }
+
+      // We *always* need global constants for the RAP, but not for the temps
+      RAPparams->set("compute global constants: temporaries", RAPparams->get("compute global constants: temporaries", false));
+      RAPparams->set("compute global constants", true);
+
+      // Allow optimization of storage.
+      // This is necessary for new faster Epetra MM kernels.
+      // Seems to work with matrix modifications to repair diagonal entries.
+
+      if (pL.get<bool>("transpose: use implicit") == true) {
+        SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
+
+        Ac = MatrixMatrix::Multiply(*P, doTranspose, *AP, !doTranspose, Ac, GetOStream(Statistics2),
+                                    doFillComplete, doOptimizeStorage, labelstr + std::string("MueLu::R*(AP)-implicit-") + levelstr.str(), RAPparams);
+
       } else {
-        RCP<ParameterList> RAPparams = rcp(new ParameterList);
-        if(pL.isSublist("matrixmatrix: kernel params"))
-          RAPparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
+        RCP<Matrix> R = Get<RCP<Matrix> >(coarseLevel, "R");
 
-        if (coarseLevel.IsAvailable("RAP reuse data", this)) {
-          GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous RAP data" << std::endl;
+        SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
 
-          RAPparams = coarseLevel.Get< RCP<ParameterList> >("RAP reuse data", this);
+        Ac = MatrixMatrix::Multiply(*R, !doTranspose, *AP, !doTranspose, Ac, GetOStream(Statistics2),
+                                    doFillComplete, doOptimizeStorage, labelstr + std::string("MueLu::R*(AP)-explicit-") + levelstr.str(), RAPparams);
+      }
+
+      Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
+      if (relativeFloor.size() > 0) {
+        Xpetra::MatrixUtils<SC, LO, GO, NO>::RelativeDiagonalBoost(Ac, relativeFloor, GetOStream(Statistics2));
+      }
+
+      bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+      bool checkAc             = pL.get<bool>("CheckMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+      ;
+      if (checkAc || repairZeroDiagonals) {
+        using magnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+        magnitudeType threshold;
+        if (pL.isType<magnitudeType>("rap: fix zero diagonals threshold"))
+          threshold = pL.get<magnitudeType>("rap: fix zero diagonals threshold");
+        else
+          threshold = Teuchos::as<magnitudeType>(pL.get<double>("rap: fix zero diagonals threshold"));
+        Scalar replacement = Teuchos::as<Scalar>(pL.get<double>("rap: fix zero diagonals replacement"));
+        Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1), threshold, replacement);
+      }
+
+      if (IsPrint(Statistics2)) {
+        RCP<ParameterList> params = rcp(new ParameterList());
+        ;
+        params->set("printLoadBalancingInfo", true);
+        params->set("printCommInfo", true);
+        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
+      }
 
-          if (RAPparams->isParameter("graph"))
-            Ac = RAPparams->get< RCP<Matrix> >("graph");
+      if (!Ac.is_null()) {
+        std::ostringstream oss;
+        oss << "A_" << coarseLevel.GetLevelID();
+        Ac->setObjectLabel(oss.str());
+      }
+      Set(coarseLevel, "A", Ac);
+
+      if (!isGPU) {
+        APparams->set("graph", AP);
+        Set(coarseLevel, "AP reuse data", APparams);
+      }
+      if (!isGPU) {
+        RAPparams->set("graph", Ac);
+        Set(coarseLevel, "RAP reuse data", RAPparams);
+      }
+    } else {
+      RCP<ParameterList> RAPparams = rcp(new ParameterList);
+      if (pL.isSublist("matrixmatrix: kernel params"))
+        RAPparams->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
 
-          // Some eigenvalue may have been cached with the matrix in the previous run.
-          // As the matrix values will be updated, we need to reset the eigenvalue.
-          Ac->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
-        }
+      if (coarseLevel.IsAvailable("RAP reuse data", this)) {
+        GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous RAP data" << std::endl;
 
-        // We *always* need global constants for the RAP, but not for the temps
-        RAPparams->set("compute global constants: temporaries",RAPparams->get("compute global constants: temporaries",false));
-        RAPparams->set("compute global constants",true);
+        RAPparams = coarseLevel.Get<RCP<ParameterList> >("RAP reuse data", this);
+
+        if (RAPparams->isParameter("graph"))
+          Ac = RAPparams->get<RCP<Matrix> >("graph");
+
+        // Some eigenvalue may have been cached with the matrix in the previous run.
+        // As the matrix values will be updated, we need to reset the eigenvalue.
+        Ac->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
+      }
 
-        if (pL.get<bool>("transpose: use implicit") == true) {
+      // We *always* need global constants for the RAP, but not for the temps
+      RAPparams->set("compute global constants: temporaries", RAPparams->get("compute global constants: temporaries", false));
+      RAPparams->set("compute global constants", true);
 
-          Ac = MatrixFactory::Build(P->getDomainMap(), Teuchos::as<LO>(0));
+      if (pL.get<bool>("transpose: use implicit") == true) {
+        Ac = MatrixFactory::Build(P->getDomainMap(), Teuchos::as<LO>(0));
 
-          SubFactoryMonitor m2(*this, "MxMxM: R x A x P (implicit)", coarseLevel);
+        SubFactoryMonitor m2(*this, "MxMxM: R x A x P (implicit)", coarseLevel);
 
-          Xpetra::TripleMatrixMultiply<SC,LO,GO,NO>::
+        Xpetra::TripleMatrixMultiply<SC, LO, GO, NO>::
             MultiplyRAP(*P, doTranspose, *A, !doTranspose, *P, !doTranspose, *Ac, doFillComplete,
-                        doOptimizeStorage, labelstr+std::string("MueLu::R*A*P-implicit-")+levelstr.str(),
-                        RAPparams);         
-        } else {
-          RCP<Matrix> R = Get< RCP<Matrix> >(coarseLevel, "R");
-          Ac = MatrixFactory::Build(R->getRowMap(), Teuchos::as<LO>(0));
+                        doOptimizeStorage, labelstr + std::string("MueLu::R*A*P-implicit-") + levelstr.str(),
+                        RAPparams);
+      } else {
+        RCP<Matrix> R = Get<RCP<Matrix> >(coarseLevel, "R");
+        Ac            = MatrixFactory::Build(R->getRowMap(), Teuchos::as<LO>(0));
 
-          SubFactoryMonitor m2(*this, "MxMxM: R x A x P (explicit)", coarseLevel);
+        SubFactoryMonitor m2(*this, "MxMxM: R x A x P (explicit)", coarseLevel);
 
-          Xpetra::TripleMatrixMultiply<SC,LO,GO,NO>::
+        Xpetra::TripleMatrixMultiply<SC, LO, GO, NO>::
             MultiplyRAP(*R, !doTranspose, *A, !doTranspose, *P, !doTranspose, *Ac, doFillComplete,
-                        doOptimizeStorage, labelstr+std::string("MueLu::R*A*P-explicit-")+levelstr.str(),
+                        doOptimizeStorage, labelstr + std::string("MueLu::R*A*P-explicit-") + levelstr.str(),
                         RAPparams);
-        }
-      
-        Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
-        if(relativeFloor.size() > 0) {
-          Xpetra::MatrixUtils<SC,LO,GO,NO>::RelativeDiagonalBoost(Ac, relativeFloor,GetOStream(Statistics2));
-        }
-
-        bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
-        bool checkAc             = pL.get<bool>("CheckMainDiagonal")|| pL.get<bool>("rap: fix zero diagonals"); ;
-        if (checkAc || repairZeroDiagonals) {
-          using magnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-          magnitudeType threshold;
-          if (pL.isType<magnitudeType>("rap: fix zero diagonals threshold"))
-            threshold = pL.get<magnitudeType>("rap: fix zero diagonals threshold");
-          else
-            threshold = Teuchos::as<magnitudeType>(pL.get<double>("rap: fix zero diagonals threshold"));
-          Scalar replacement = Teuchos::as<Scalar>(pL.get<double>("rap: fix zero diagonals replacement"));
-          Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1), threshold, replacement);
-        }
-
-
-        if (IsPrint(Statistics2)) {
-          RCP<ParameterList> params = rcp(new ParameterList());;
-          params->set("printLoadBalancingInfo", true);
-          params->set("printCommInfo",          true);
-          GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
-        }
-
-        if(!Ac.is_null()) {std::ostringstream oss; oss << "A_" << coarseLevel.GetLevelID(); Ac->setObjectLabel(oss.str());}
-        Set(coarseLevel, "A",         Ac);
-
-        if (!isGPU) {
-          RAPparams->set("graph", Ac);
-          Set(coarseLevel, "RAP reuse data", RAPparams);
-        }
       }
 
+      Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
+      if (relativeFloor.size() > 0) {
+        Xpetra::MatrixUtils<SC, LO, GO, NO>::RelativeDiagonalBoost(Ac, relativeFloor, GetOStream(Statistics2));
+      }
 
-    }
+      bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+      bool checkAc             = pL.get<bool>("CheckMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+      ;
+      if (checkAc || repairZeroDiagonals) {
+        using magnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+        magnitudeType threshold;
+        if (pL.isType<magnitudeType>("rap: fix zero diagonals threshold"))
+          threshold = pL.get<magnitudeType>("rap: fix zero diagonals threshold");
+        else
+          threshold = Teuchos::as<magnitudeType>(pL.get<double>("rap: fix zero diagonals threshold"));
+        Scalar replacement = Teuchos::as<Scalar>(pL.get<double>("rap: fix zero diagonals replacement"));
+        Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1), threshold, replacement);
+      }
 
-#ifdef HAVE_MUELU_DEBUG
-    MatrixUtils::checkLocalRowMapMatchesColMap(*Ac);
-#endif // HAVE_MUELU_DEBUG
-
-    if (transferFacts_.begin() != transferFacts_.end()) {
-      SubFactoryMonitor m(*this, "Projections", coarseLevel);
-
-      // call Build of all user-given transfer factories
-      for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
-        RCP<const FactoryBase> fac = *it;
-        GetOStream(Runtime0) << "RAPFactory: call transfer factory: " << fac->description() << std::endl;
-        fac->CallBuild(coarseLevel);
-        // Coordinates transfer is marginally different from all other operations
-        // because it is *optional*, and not required. For instance, we may need
-        // coordinates only on level 4 if we start repartitioning from that level,
-        // but we don't need them on level 1,2,3. As our current Hierarchy setup
-        // assumes propagation of dependencies only through three levels, this
-        // means that we need to rely on other methods to propagate optional data.
-        //
-        // The method currently used is through RAP transfer factories, which are
-        // simply factories which are called at the end of RAP with a single goal:
-        // transfer some fine data to coarser level. Because these factories are
-        // kind of outside of the mainline factories, they behave different. In
-        // particular, we call their Build method explicitly, rather than through
-        // Get calls. This difference is significant, as the Get call is smart
-        // enough to know when to release all factory dependencies, and Build is
-        // dumb. This led to the following CoordinatesTransferFactory sequence:
-        // 1. Request level 0
-        // 2. Request level 1
-        // 3. Request level 0
-        // 4. Release level 0
-        // 5. Release level 1
-        //
-        // The problem is missing "6. Release level 0". Because it was missing,
-        // we had outstanding request on "Coordinates", "Aggregates" and
-        // "CoarseMap" on level 0.
-        //
-        // This was fixed by explicitly calling Release on transfer factories in
-        // RAPFactory. I am still unsure how exactly it works, but now we have
-        // clear data requests for all levels.
-        coarseLevel.Release(*fac);
+      if (IsPrint(Statistics2)) {
+        RCP<ParameterList> params = rcp(new ParameterList());
+        ;
+        params->set("printLoadBalancingInfo", true);
+        params->set("printCommInfo", true);
+        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
       }
-    }
 
+      if (!Ac.is_null()) {
+        std::ostringstream oss;
+        oss << "A_" << coarseLevel.GetLevelID();
+        Ac->setObjectLabel(oss.str());
+      }
+      Set(coarseLevel, "A", Ac);
+
+      if (!isGPU) {
+        RAPparams->set("graph", Ac);
+        Set(coarseLevel, "RAP reuse data", RAPparams);
+      }
+    }
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::RAPFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::RAPFactory::AddTransferFactory: Factory is being added after we have already declared input");
-    transferFacts_.push_back(factory);
+#ifdef HAVE_MUELU_DEBUG
+  MatrixUtils::checkLocalRowMapMatchesColMap(*Ac);
+#endif  // HAVE_MUELU_DEBUG
+
+  if (transferFacts_.begin() != transferFacts_.end()) {
+    SubFactoryMonitor m(*this, "Projections", coarseLevel);
+
+    // call Build of all user-given transfer factories
+    for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
+      RCP<const FactoryBase> fac = *it;
+      GetOStream(Runtime0) << "RAPFactory: call transfer factory: " << fac->description() << std::endl;
+      fac->CallBuild(coarseLevel);
+      // Coordinates transfer is marginally different from all other operations
+      // because it is *optional*, and not required. For instance, we may need
+      // coordinates only on level 4 if we start repartitioning from that level,
+      // but we don't need them on level 1,2,3. As our current Hierarchy setup
+      // assumes propagation of dependencies only through three levels, this
+      // means that we need to rely on other methods to propagate optional data.
+      //
+      // The method currently used is through RAP transfer factories, which are
+      // simply factories which are called at the end of RAP with a single goal:
+      // transfer some fine data to coarser level. Because these factories are
+      // kind of outside of the mainline factories, they behave different. In
+      // particular, we call their Build method explicitly, rather than through
+      // Get calls. This difference is significant, as the Get call is smart
+      // enough to know when to release all factory dependencies, and Build is
+      // dumb. This led to the following CoordinatesTransferFactory sequence:
+      // 1. Request level 0
+      // 2. Request level 1
+      // 3. Request level 0
+      // 4. Release level 0
+      // 5. Release level 1
+      //
+      // The problem is missing "6. Release level 0". Because it was missing,
+      // we had outstanding request on "Coordinates", "Aggregates" and
+      // "CoarseMap" on level 0.
+      //
+      // This was fixed by explicitly calling Release on transfer factories in
+      // RAPFactory. I am still unsure how exactly it works, but now we have
+      // clear data requests for all levels.
+      coarseLevel.Release(*fac);
+    }
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::RAPFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::RAPFactory::AddTransferFactory: Factory is being added after we have already declared input");
+  transferFacts_.push_back(factory);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RAPFACTORY_SHORT
-#endif // MUELU_RAPFACTORY_DEF_HPP
+#endif  // MUELU_RAPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_RAPShiftFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_RAPShiftFactory_decl.hpp
index 91cbf254d6fe..827e81695a2d 100644
--- a/packages/muelu/src/Misc/MueLu_RAPShiftFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_RAPShiftFactory_decl.hpp
@@ -62,90 +62,88 @@
 #include "MueLu_TwoLevelFactoryBase.hpp"
 
 namespace MueLu {
-  /*!
-    @class RAPShiftFactory
-    @brief Factory for building coarse grid matrices, when the matrix
-           is of the form K+a*M. Useful when you want to change the shift
-           variable ("a") at every level. Each level must store the stiffness
-           matrix K and mass matrix M separately.
-  */
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class RAPShiftFactory : public TwoLevelFactoryBase {
+/*!
+  @class RAPShiftFactory
+  @brief Factory for building coarse grid matrices, when the matrix
+         is of the form K+a*M. Useful when you want to change the shift
+         variable ("a") at every level. Each level must store the stiffness
+         matrix K and mass matrix M separately.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RAPShiftFactory : public TwoLevelFactoryBase {
 #undef MUELU_RAPSHIFTFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    RAPShiftFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    virtual ~RAPShiftFactory() { }
+  RAPShiftFactory();
 
-    //@}
+  virtual ~RAPShiftFactory() {}
 
-    //! @name Input
-    //@{
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &fineLevel, Level &coarseLevel) const;
-    //@}
+  //@}
 
-    //! @name Handling of user-defined transfer factories
-    //@{
+  //! @name Build methods.
+  //@{
+  void Build(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //! Indicate that the restriction operator action should be implicitly defined by the transpose of the prolongator.
-    void SetImplicitTranspose(bool const &implicit) {
-      implicitTranspose_ = implicit;
-    }
+  //! @name Handling of user-defined transfer factories
+  //@{
 
-    void SetShifts(std::vector<Scalar>& shifts) {
-      shifts_.clear();
-      shifts_ = shifts;
-    }
+  //! Indicate that the restriction operator action should be implicitly defined by the transpose of the prolongator.
+  void SetImplicitTranspose(bool const &implicit) {
+    implicitTranspose_ = implicit;
+  }
 
-    //@}
+  void SetShifts(std::vector<Scalar> &shifts) {
+    shifts_.clear();
+    shifts_ = shifts;
+  }
 
-    //@{
-    /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
+  //@}
 
-    Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
-    the next coarser level.
-    */
-    void AddTransferFactory(const RCP<const FactoryBase>& factory);
+  //@{
+  /*! @brief Add transfer factory in the end of list of transfer factories in RepartitionAcFactory.
 
-    // TODO add a function to remove a specific transfer factory?
-
-    //! Returns number of transfer factories.
-    size_t NumTransferFactories() const { return transferFacts_.size(); }
+  Transfer factories are derived from TwoLevelFactoryBase and project some data from the fine level to
+  the next coarser level.
+  */
+  void AddTransferFactory(const RCP<const FactoryBase> &factory);
 
-    //@}
+  // TODO add a function to remove a specific transfer factory?
 
-  private:
+  //! Returns number of transfer factories.
+  size_t NumTransferFactories() const { return transferFacts_.size(); }
 
-    //! If true, the action of the restriction operator action is implicitly defined by the transpose of the prolongator.
-    bool implicitTranspose_;
+  //@}
 
+ private:
+  //! If true, the action of the restriction operator action is implicitly defined by the transpose of the prolongator.
+  bool implicitTranspose_;
 
-    //! list of user-defined transfer Factories
-    std::vector<RCP<const FactoryBase> > transferFacts_;
+  //! list of user-defined transfer Factories
+  std::vector<RCP<const FactoryBase> > transferFacts_;
 
-    // vector of shifting terms
-    std::vector<Scalar> shifts_;
+  // vector of shifting terms
+  std::vector<Scalar> shifts_;
 
-  }; //class RAPShiftFactory
+};  // class RAPShiftFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RAPSHIFTFACTORY_SHORT
-#endif // MUELU_RAPSHIFTFACTORY_DECL_HPP
+#endif  // MUELU_RAPSHIFTFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_RAPShiftFactory_def.hpp b/packages/muelu/src/Misc/MueLu_RAPShiftFactory_def.hpp
index b3c1c0833bb4..b3670f00b788 100644
--- a/packages/muelu/src/Misc/MueLu_RAPShiftFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_RAPShiftFactory_def.hpp
@@ -54,7 +54,6 @@
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_VectorFactory.hpp>
 
-
 #include "MueLu_RAPShiftFactory_decl.hpp"
 #include "MueLu_MasterList.hpp"
 #include "MueLu_Monitor.hpp"
@@ -62,338 +61,342 @@
 
 namespace MueLu {
 
-  /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RAPShiftFactory()
-    : implicitTranspose_(false)  { }
-
+/*********************************************************************************************************/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RAPShiftFactory()
+  : implicitTranspose_(false) {}
 
-  /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+/*********************************************************************************************************/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("transpose: use implicit");
-    SET_VALID_ENTRY("rap: fix zero diagonals");
-    SET_VALID_ENTRY("rap: shift");
-    SET_VALID_ENTRY("rap: shift array");
-    SET_VALID_ENTRY("rap: cfl array");
-    SET_VALID_ENTRY("rap: shift diagonal M");
-    SET_VALID_ENTRY("rap: shift low storage");
-    SET_VALID_ENTRY("rap: relative diagonal floor");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("M",              Teuchos::null, "Generating factory of the matrix M used during the non-Galerkin RAP");
-    validParamList->set< RCP<const FactoryBase> >("Mdiag",          Teuchos::null, "Generating factory of the matrix Mdiag used during the non-Galerkin RAP");
-    validParamList->set< RCP<const FactoryBase> >("K",              Teuchos::null, "Generating factory of the matrix K used during the non-Galerkin RAP");
-    validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Prolongator factory");
-    validParamList->set< RCP<const FactoryBase> >("R",              Teuchos::null, "Restrictor factory");
-
-    validParamList->set< bool >                  ("CheckMainDiagonal",  false, "Check main diagonal for zeros");
-    validParamList->set< bool >                  ("RepairMainDiagonal", false, "Repair zeros on main diagonal");
-
-    validParamList->set<RCP<const FactoryBase> > ("deltaT", Teuchos::null, "user deltaT");
-    validParamList->set<RCP<const FactoryBase> > ("cfl", Teuchos::null, "user cfl");
-    validParamList->set<RCP<const FactoryBase> > ("cfl-based shift array", Teuchos::null, "MueLu-generated shift array for CFL-based shifting");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+  SET_VALID_ENTRY("transpose: use implicit");
+  SET_VALID_ENTRY("rap: fix zero diagonals");
+  SET_VALID_ENTRY("rap: shift");
+  SET_VALID_ENTRY("rap: shift array");
+  SET_VALID_ENTRY("rap: cfl array");
+  SET_VALID_ENTRY("rap: shift diagonal M");
+  SET_VALID_ENTRY("rap: shift low storage");
+  SET_VALID_ENTRY("rap: relative diagonal floor");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase> >("M", Teuchos::null, "Generating factory of the matrix M used during the non-Galerkin RAP");
+  validParamList->set<RCP<const FactoryBase> >("Mdiag", Teuchos::null, "Generating factory of the matrix Mdiag used during the non-Galerkin RAP");
+  validParamList->set<RCP<const FactoryBase> >("K", Teuchos::null, "Generating factory of the matrix K used during the non-Galerkin RAP");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Prolongator factory");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Restrictor factory");
+
+  validParamList->set<bool>("CheckMainDiagonal", false, "Check main diagonal for zeros");
+  validParamList->set<bool>("RepairMainDiagonal", false, "Repair zeros on main diagonal");
+
+  validParamList->set<RCP<const FactoryBase> >("deltaT", Teuchos::null, "user deltaT");
+  validParamList->set<RCP<const FactoryBase> >("cfl", Teuchos::null, "user cfl");
+  validParamList->set<RCP<const FactoryBase> >("cfl-based shift array", Teuchos::null, "MueLu-generated shift array for CFL-based shifting");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+/*********************************************************************************************************/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  const Teuchos::ParameterList &pL = GetParameterList();
+
+  bool use_mdiag = false;
+  if (pL.isParameter("rap: shift diagonal M"))
+    use_mdiag = pL.get<bool>("rap: shift diagonal M");
+
+  // The low storage version requires mdiag
+  bool use_low_storage = false;
+  if (pL.isParameter("rap: shift low storage")) {
+    use_low_storage = pL.get<bool>("rap: shift low storage");
+    use_mdiag       = use_low_storage ? true : use_mdiag;
+  }
+
+  if (implicitTranspose_ == false) {
+    Input(coarseLevel, "R");
+  }
+
+  if (!use_low_storage)
+    Input(fineLevel, "K");
+  else
+    Input(fineLevel, "A");
+  Input(coarseLevel, "P");
+
+  if (!use_mdiag)
+    Input(fineLevel, "M");
+  else
+    Input(fineLevel, "Mdiag");
+
+  // CFL array stuff
+  if (pL.isParameter("rap: cfl array") && pL.get<Teuchos::Array<double> >("rap: cfl array").size() > 0) {
+    if (fineLevel.GetLevelID() == 0) {
+      if (fineLevel.IsAvailable("deltaT", NoFactory::get())) {
+        fineLevel.DeclareInput("deltaT", NoFactory::get(), this);
+      } else {
+        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("fine deltaT", NoFactory::get()),
+                                   Exceptions::RuntimeError,
+                                   "deltaT was not provided by the user on level0!");
+      }
+
+      if (fineLevel.IsAvailable("cfl", NoFactory::get())) {
+        fineLevel.DeclareInput("cfl", NoFactory::get(), this);
+      } else {
+        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("fine cfl", NoFactory::get()),
+                                   Exceptions::RuntimeError,
+                                   "cfl was not provided by the user on level0!");
+      }
+    } else {
+      Input(fineLevel, "cfl-based shift array");
+    }
+  }
+
+  // call DeclareInput of all user-given transfer factories
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
+    (*it)->CallDeclareInput(coarseLevel);
   }
+}
 
-  /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    const Teuchos::ParameterList& pL = GetParameterList();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {  // FIXME make fineLevel const
+  {
+    FactoryMonitor m(*this, "Computing Ac", coarseLevel);
+    const Teuchos::ParameterList &pL = GetParameterList();
 
-    bool use_mdiag = false;
-    if(pL.isParameter("rap: shift diagonal M")) 
-      use_mdiag = pL.get<bool>("rap: shift diagonal M");
+    bool M_is_diagonal = false;
+    if (pL.isParameter("rap: shift diagonal M"))
+      M_is_diagonal = pL.get<bool>("rap: shift diagonal M");
 
     // The low storage version requires mdiag
     bool use_low_storage = false;
-    if(pL.isParameter("rap: shift low storage")) {
+    if (pL.isParameter("rap: shift low storage")) {
       use_low_storage = pL.get<bool>("rap: shift low storage");
-      use_mdiag = use_low_storage ? true : use_mdiag;
+      M_is_diagonal   = use_low_storage ? true : M_is_diagonal;
     }
 
-    if (implicitTranspose_ == false) {
-      Input(coarseLevel, "R");
+    Teuchos::ArrayView<const double> doubleShifts;
+    Teuchos::ArrayRCP<double> myshifts;
+    if (pL.isParameter("rap: shift array") && pL.get<Teuchos::Array<double> >("rap: shift array").size() > 0) {
+      // Do we have an array of shifts?  If so, we set doubleShifts_
+      doubleShifts = pL.get<Teuchos::Array<double> >("rap: shift array")();
     }
-
-    if(!use_low_storage) Input(fineLevel,   "K");
-    else Input(fineLevel,   "A");
-    Input(coarseLevel, "P");
-
-    if(!use_mdiag) Input(fineLevel, "M");
-    else Input(fineLevel, "Mdiag");
-
-    // CFL array stuff
-    if(pL.isParameter("rap: cfl array") && pL.get<Teuchos::Array<double> >("rap: cfl array").size() > 0) {
-      if(fineLevel.GetLevelID() == 0) {
-        if(fineLevel.IsAvailable("deltaT", NoFactory::get())) {
-          fineLevel.DeclareInput("deltaT", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("fine deltaT", NoFactory::get()),
-                                     Exceptions::RuntimeError,
-                                     "deltaT was not provided by the user on level0!");
+    if (pL.isParameter("rap: cfl array") && pL.get<Teuchos::Array<double> >("rap: cfl array").size() > 0) {
+      // Do we have an array of CFLs?  If so, we calculated the shifts from them.
+      Teuchos::ArrayView<const double> CFLs = pL.get<Teuchos::Array<double> >("rap: cfl array")();
+      if (fineLevel.GetLevelID() == 0) {
+        double dt         = Get<double>(fineLevel, "deltaT");
+        double cfl        = Get<double>(fineLevel, "cfl");
+        double ts_at_cfl1 = dt / cfl;
+        myshifts.resize(CFLs.size());
+        Teuchos::Array<double> myCFLs(CFLs.size());
+        myCFLs[0] = cfl;
+
+        // Never make the CFL bigger
+        for (int i = 1; i < (int)CFLs.size(); i++)
+          myCFLs[i] = (CFLs[i] > cfl) ? cfl : CFLs[i];
+
+        {
+          std::ostringstream ofs;
+          ofs << "RAPShiftFactory: CFL schedule = ";
+          for (int i = 0; i < (int)CFLs.size(); i++)
+            ofs << " " << myCFLs[i];
+          GetOStream(Statistics0) << ofs.str() << std::endl;
         }
-        
-        if(fineLevel.IsAvailable("cfl", NoFactory::get())) {
-          fineLevel.DeclareInput("cfl", NoFactory::get(), this);
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("fine cfl", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                     "cfl was not provided by the user on level0!");
-        }      
-      }
-      else {
-        Input(fineLevel,"cfl-based shift array");
+        GetOStream(Statistics0) << "RAPShiftFactory: Timestep at CFL=1 is " << ts_at_cfl1 << "    " << std::endl;
+
+        // The shift array needs to be 1/dt
+        for (int i = 0; i < (int)myshifts.size(); i++)
+          myshifts[i] = 1.0 / (ts_at_cfl1 * myCFLs[i]);
+        doubleShifts = myshifts();
+
+        {
+          std::ostringstream ofs;
+          ofs << "RAPShiftFactory: shift schedule = ";
+          for (int i = 0; i < (int)doubleShifts.size(); i++)
+            ofs << " " << doubleShifts[i];
+          GetOStream(Statistics0) << ofs.str() << std::endl;
+        }
+        Set(coarseLevel, "cfl-based shift array", myshifts);
+      } else {
+        myshifts     = Get<Teuchos::ArrayRCP<double> >(fineLevel, "cfl-based shift array");
+        doubleShifts = myshifts();
+        Set(coarseLevel, "cfl-based shift array", myshifts);
+        // NOTE: If we're not on level zero, then we should have a shift array
       }
     }
 
-    // call DeclareInput of all user-given transfer factories
-    for(std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it!=transferFacts_.end(); ++it) {
-      (*it)->CallDeclareInput(coarseLevel);
-    }
-  }
+    // Inputs: K, M, P
+    // Note: In the low-storage case we do not keep a separate "K", we just use A
+    RCP<Matrix> K;
+    RCP<Matrix> M;
+    RCP<Vector> Mdiag;
+
+    if (use_low_storage)
+      K = Get<RCP<Matrix> >(fineLevel, "A");
+    else
+      K = Get<RCP<Matrix> >(fineLevel, "K");
+    if (!M_is_diagonal)
+      M = Get<RCP<Matrix> >(fineLevel, "M");
+    else
+      Mdiag = Get<RCP<Vector> >(fineLevel, "Mdiag");
+
+    RCP<Matrix> P = Get<RCP<Matrix> >(coarseLevel, "P");
+
+    // Build Kc = RKP, Mc = RMP
+    RCP<Matrix> KP, MP;
+
+    // Reuse pattern if available (multiple solve)
+    // FIXME: Old style reuse doesn't work any more
+    //      if (IsAvailable(coarseLevel, "AP Pattern")) {
+    //        KP = Get< RCP<Matrix> >(coarseLevel, "AP Pattern");
+    //        MP = Get< RCP<Matrix> >(coarseLevel, "AP Pattern");
+    //      }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const { // FIXME make fineLevel const
     {
-      FactoryMonitor m(*this, "Computing Ac", coarseLevel);      
-      const Teuchos::ParameterList& pL = GetParameterList();
-
-      bool M_is_diagonal = false; 
-      if(pL.isParameter("rap: shift diagonal M"))
-         M_is_diagonal = pL.get<bool>("rap: shift diagonal M");
-
-      // The low storage version requires mdiag
-      bool use_low_storage = false;
-      if(pL.isParameter("rap: shift low storage")) {
-        use_low_storage = pL.get<bool>("rap: shift low storage");
-        M_is_diagonal = use_low_storage ? true : M_is_diagonal;
+      SubFactoryMonitor subM(*this, "MxM: K x P", coarseLevel);
+      KP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*K, false, *P, false, KP, GetOStream(Statistics2));
+      if (!M_is_diagonal) {
+        MP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*M, false, *P, false, MP, GetOStream(Statistics2));
+      } else {
+        MP = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(P);
+        MP->leftScale(*Mdiag);
       }
 
-      Teuchos::ArrayView<const double> doubleShifts;
-      Teuchos::ArrayRCP<double> myshifts;
-      if(pL.isParameter("rap: shift array") && pL.get<Teuchos::Array<double> >("rap: shift array").size() > 0 ) {
-        // Do we have an array of shifts?  If so, we set doubleShifts_
-        doubleShifts = pL.get<Teuchos::Array<double> >("rap: shift array")();
-      }
-      if(pL.isParameter("rap: cfl array") && pL.get<Teuchos::Array<double> >("rap: cfl array").size() > 0) {
-        // Do we have an array of CFLs?  If so, we calculated the shifts from them.        
-        Teuchos::ArrayView<const double> CFLs = pL.get<Teuchos::Array<double> >("rap: cfl array")();
-        if(fineLevel.GetLevelID() == 0) {
-          double dt  = Get<double>(fineLevel,"deltaT");
-          double cfl = Get<double>(fineLevel,"cfl");
-          double ts_at_cfl1 = dt / cfl;
-          myshifts.resize(CFLs.size());
-          Teuchos::Array<double> myCFLs(CFLs.size());
-          myCFLs[0] = cfl;
-
-          // Never make the CFL bigger
-          for(int i=1; i<(int)CFLs.size(); i++)
-            myCFLs[i] = (CFLs[i]> cfl) ? cfl : CFLs[i];
-          
-          {
-            std::ostringstream ofs;
-            ofs<<"RAPShiftFactory: CFL schedule = ";
-            for(int i=0; i<(int)CFLs.size(); i++)
-              ofs<<" "<<myCFLs[i];
-            GetOStream(Statistics0) <<ofs.str() << std::endl;
-          }
-          GetOStream(Statistics0)<< "RAPShiftFactory: Timestep at CFL=1 is "<< ts_at_cfl1 << "    " <<std::endl;
-
-          // The shift array needs to be 1/dt
-          for(int i=0; i<(int)myshifts.size(); i++)
-            myshifts[i] =  1.0 / (ts_at_cfl1*myCFLs[i]);
-          doubleShifts = myshifts();
-          
-          {
-            std::ostringstream ofs;
-            ofs<<"RAPShiftFactory: shift schedule = ";
-            for(int i=0; i<(int)doubleShifts.size(); i++)
-              ofs<<" "<<doubleShifts[i];
-            GetOStream(Statistics0) <<ofs.str() << std::endl;
-          }
-          Set(coarseLevel,"cfl-based shift array",myshifts);
-        }
-        else {
-          myshifts = Get<Teuchos::ArrayRCP<double> > (fineLevel,"cfl-based shift array");
-          doubleShifts = myshifts();
-          Set(coarseLevel,"cfl-based shift array",myshifts);
-          // NOTE: If we're not on level zero, then we should have a shift array
-        }
-      }
-      
-      // Inputs: K, M, P
-      // Note: In the low-storage case we do not keep a separate "K", we just use A
-      RCP<Matrix> K;
-      RCP<Matrix> M;
-      RCP<Vector> Mdiag;
-
-      if(use_low_storage)  K = Get< RCP<Matrix> >(fineLevel, "A");
-      else K = Get< RCP<Matrix> >(fineLevel, "K");
-      if(!M_is_diagonal) M = Get< RCP<Matrix> >(fineLevel, "M");
-      else Mdiag = Get< RCP<Vector> >(fineLevel, "Mdiag");
-
-      RCP<Matrix> P = Get< RCP<Matrix> >(coarseLevel, "P");
-
-      // Build Kc = RKP, Mc = RMP
-      RCP<Matrix> KP, MP;
-
-      // Reuse pattern if available (multiple solve)
-      // FIXME: Old style reuse doesn't work any more
-      //      if (IsAvailable(coarseLevel, "AP Pattern")) {
-      //        KP = Get< RCP<Matrix> >(coarseLevel, "AP Pattern");
-      //        MP = Get< RCP<Matrix> >(coarseLevel, "AP Pattern");
-      //      }
-
-      {
-        SubFactoryMonitor subM(*this, "MxM: K x P", coarseLevel);
-        KP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*K, false, *P, false, KP, GetOStream(Statistics2));
-        if(!M_is_diagonal) {
-          MP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*M, false, *P, false, MP, GetOStream(Statistics2));
-        }
-        else {
-          MP = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(P);
-          MP->leftScale(*Mdiag);
-        }
-
-        Set(coarseLevel, "AP Pattern", KP);
-      }
+      Set(coarseLevel, "AP Pattern", KP);
+    }
 
-      bool doOptimizedStorage = true;
+    bool doOptimizedStorage = true;
 
-      RCP<Matrix> Ac, Kc, Mc;
+    RCP<Matrix> Ac, Kc, Mc;
 
-      // Reuse pattern if available (multiple solve)
-      //     if (IsAvailable(coarseLevel, "RAP Pattern"))
-      // Ac = Get< RCP<Matrix> >(coarseLevel, "RAP Pattern");
+    // Reuse pattern if available (multiple solve)
+    //     if (IsAvailable(coarseLevel, "RAP Pattern"))
+    // Ac = Get< RCP<Matrix> >(coarseLevel, "RAP Pattern");
 
-      bool doFillComplete=true;
-      if (implicitTranspose_) {
-        SubFactoryMonitor m2(*this, "MxM: P' x (KP) (implicit)", coarseLevel);
-        Kc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P, true, *KP, false, Kc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
-        Mc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P, true, *MP, false, Mc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
-      }
-      else {
-        RCP<Matrix> R = Get< RCP<Matrix> >(coarseLevel, "R");
-        SubFactoryMonitor m2(*this, "MxM: R x (KP) (explicit)", coarseLevel);
-        Kc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*R, false, *KP, false, Kc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
-        Mc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*R, false, *MP, false, Mc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
-      }
+    bool doFillComplete = true;
+    if (implicitTranspose_) {
+      SubFactoryMonitor m2(*this, "MxM: P' x (KP) (implicit)", coarseLevel);
+      Kc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P, true, *KP, false, Kc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
+      Mc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P, true, *MP, false, Mc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
+    } else {
+      RCP<Matrix> R = Get<RCP<Matrix> >(coarseLevel, "R");
+      SubFactoryMonitor m2(*this, "MxM: R x (KP) (explicit)", coarseLevel);
+      Kc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*R, false, *KP, false, Kc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
+      Mc = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*R, false, *MP, false, Mc, GetOStream(Statistics2), doFillComplete, doOptimizedStorage);
+    }
 
-      // Get the shift
-      // FIXME - We should really get rid of the shifts array and drive this the same way everything else works
-      // If we're using the recursive "low storage" version, we need to shift by ( \prod_{i=1}^k shift[i] - \prod_{i=1}^{k-1} shift[i]) to
-      // get the recursive relationships correct
-      int level      = coarseLevel.GetLevelID();
-      Scalar shift   = Teuchos::ScalarTraits<Scalar>::zero();
-      if(!use_low_storage) {
-        // High Storage version
-        if(level < (int)shifts_.size()) shift = shifts_[level];
-        else shift  = Teuchos::as<Scalar>(pL.get<double>("rap: shift"));
-      }
-      else {
-        // Low Storage Version
-        if(level < (int)shifts_.size()) {
-          if(level==1) shift = shifts_[level];
-          else {
-            Scalar prod1 = Teuchos::ScalarTraits<Scalar>::one();
-            for(int i=1; i < level-1; i++) {
-              prod1 *= shifts_[i];
-            } 
-            shift = (prod1 * shifts_[level] - prod1);
-          }
-        }
-        else if(doubleShifts.size() != 0) {
-          double d_shift = 0.0;
-          if(level < doubleShifts.size())
-            d_shift = doubleShifts[level] - doubleShifts[level-1];
-
-          if(d_shift < 0.0) 
-            GetOStream(Warnings1) << "WARNING: RAPShiftFactory has detected a negative shift... This implies a less stable coarse grid."<<std::endl;
-          shift = Teuchos::as<Scalar>(d_shift);
-        }
+    // Get the shift
+    // FIXME - We should really get rid of the shifts array and drive this the same way everything else works
+    // If we're using the recursive "low storage" version, we need to shift by ( \prod_{i=1}^k shift[i] - \prod_{i=1}^{k-1} shift[i]) to
+    // get the recursive relationships correct
+    int level    = coarseLevel.GetLevelID();
+    Scalar shift = Teuchos::ScalarTraits<Scalar>::zero();
+    if (!use_low_storage) {
+      // High Storage version
+      if (level < (int)shifts_.size())
+        shift = shifts_[level];
+      else
+        shift = Teuchos::as<Scalar>(pL.get<double>("rap: shift"));
+    } else {
+      // Low Storage Version
+      if (level < (int)shifts_.size()) {
+        if (level == 1)
+          shift = shifts_[level];
         else {
-          double base_shift = pL.get<double>("rap: shift");
-          if(level == 1) shift = Teuchos::as<Scalar>(base_shift);
-          else shift = Teuchos::as<Scalar>(pow(base_shift,level) - pow(base_shift,level-1));
+          Scalar prod1 = Teuchos::ScalarTraits<Scalar>::one();
+          for (int i = 1; i < level - 1; i++) {
+            prod1 *= shifts_[i];
+          }
+          shift = (prod1 * shifts_[level] - prod1);
         }
+      } else if (doubleShifts.size() != 0) {
+        double d_shift = 0.0;
+        if (level < doubleShifts.size())
+          d_shift = doubleShifts[level] - doubleShifts[level - 1];
+
+        if (d_shift < 0.0)
+          GetOStream(Warnings1) << "WARNING: RAPShiftFactory has detected a negative shift... This implies a less stable coarse grid." << std::endl;
+        shift = Teuchos::as<Scalar>(d_shift);
+      } else {
+        double base_shift = pL.get<double>("rap: shift");
+        if (level == 1)
+          shift = Teuchos::as<Scalar>(base_shift);
+        else
+          shift = Teuchos::as<Scalar>(pow(base_shift, level) - pow(base_shift, level - 1));
       }
-      GetOStream(Runtime0) << "RAPShiftFactory: Using shift " << shift << std::endl;
-
-        
-      // recombine to get K+shift*M
-      {
-	SubFactoryMonitor m2(*this, "Add: RKP + s*RMP", coarseLevel);
-	Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Kc, false, Teuchos::ScalarTraits<Scalar>::one(), *Mc, false, shift, Ac, GetOStream(Statistics2));
-	Ac->fillComplete();
-      }
-
-      Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
-      if(relativeFloor.size() > 0) 
-        Xpetra::MatrixUtils<SC,LO,GO,NO>::RelativeDiagonalBoost(Ac, relativeFloor,GetOStream(Statistics2));
-   
-
-      bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
-      bool checkAc             = pL.get<bool>("CheckMainDiagonal")|| pL.get<bool>("rap: fix zero diagonals"); ;
-      if (checkAc || repairZeroDiagonals)
-        Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1));
-
-      RCP<ParameterList> params = rcp(new ParameterList());;
-      params->set("printLoadBalancingInfo", true);
-      GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
-
-      Set(coarseLevel, "A", Ac);
-      // We only need K in the 'high storage' mode
-      if(!use_low_storage)
-        Set(coarseLevel, "K", Kc);
-
-      if(!M_is_diagonal) {
-        Set(coarseLevel, "M", Mc);
-      }
-      else {
-        // If M is diagonal, then we only pass that part down the hierarchy
-        // NOTE: Should we be doing some kind of rowsum instead?
-        RCP<Vector> Mcv = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Mc->getRowMap(),false);
-        Mc->getLocalDiagCopy(*Mcv);
-        Set(coarseLevel, "Mdiag", Mcv);
-      }
+    }
+    GetOStream(Runtime0) << "RAPShiftFactory: Using shift " << shift << std::endl;
 
-      //      Set(coarseLevel, "RAP Pattern", Ac);
+    // recombine to get K+shift*M
+    {
+      SubFactoryMonitor m2(*this, "Add: RKP + s*RMP", coarseLevel);
+      Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Kc, false, Teuchos::ScalarTraits<Scalar>::one(), *Mc, false, shift, Ac, GetOStream(Statistics2));
+      Ac->fillComplete();
     }
 
-    if (transferFacts_.begin() != transferFacts_.end()) {
-      SubFactoryMonitor m(*this, "Projections", coarseLevel);
-
-      // call Build of all user-given transfer factories
-      for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
-        RCP<const FactoryBase> fac = *it;
-        GetOStream(Runtime0) << "RAPShiftFactory: call transfer factory: " << fac->description() << std::endl;
-        fac->CallBuild(coarseLevel);
-        // AP (11/11/13): I am not sure exactly why we need to call Release, but we do need it to get rid
-        // of dangling data for CoordinatesTransferFactory
-        coarseLevel.Release(*fac);
-      }
+    Teuchos::ArrayView<const double> relativeFloor = pL.get<Teuchos::Array<double> >("rap: relative diagonal floor")();
+    if (relativeFloor.size() > 0)
+      Xpetra::MatrixUtils<SC, LO, GO, NO>::RelativeDiagonalBoost(Ac, relativeFloor, GetOStream(Statistics2));
+
+    bool repairZeroDiagonals = pL.get<bool>("RepairMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+    bool checkAc             = pL.get<bool>("CheckMainDiagonal") || pL.get<bool>("rap: fix zero diagonals");
+    ;
+    if (checkAc || repairZeroDiagonals)
+      Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(Ac, repairZeroDiagonals, GetOStream(Warnings1));
+
+    RCP<ParameterList> params = rcp(new ParameterList());
+    ;
+    params->set("printLoadBalancingInfo", true);
+    GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*Ac, "Ac", params);
+
+    Set(coarseLevel, "A", Ac);
+    // We only need K in the 'high storage' mode
+    if (!use_low_storage)
+      Set(coarseLevel, "K", Kc);
+
+    if (!M_is_diagonal) {
+      Set(coarseLevel, "M", Mc);
+    } else {
+      // If M is diagonal, then we only pass that part down the hierarchy
+      // NOTE: Should we be doing some kind of rowsum instead?
+      RCP<Vector> Mcv = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Mc->getRowMap(), false);
+      Mc->getLocalDiagCopy(*Mcv);
+      Set(coarseLevel, "Mdiag", Mcv);
     }
+
+    //      Set(coarseLevel, "RAP Pattern", Ac);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "MueLu::RAPShiftFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    transferFacts_.push_back(factory);
+  if (transferFacts_.begin() != transferFacts_.end()) {
+    SubFactoryMonitor m(*this, "Projections", coarseLevel);
+
+    // call Build of all user-given transfer factories
+    for (std::vector<RCP<const FactoryBase> >::const_iterator it = transferFacts_.begin(); it != transferFacts_.end(); ++it) {
+      RCP<const FactoryBase> fac = *it;
+      GetOStream(Runtime0) << "RAPShiftFactory: call transfer factory: " << fac->description() << std::endl;
+      fac->CallBuild(coarseLevel);
+      // AP (11/11/13): I am not sure exactly why we need to call Release, but we do need it to get rid
+      // of dangling data for CoordinatesTransferFactory
+      coarseLevel.Release(*fac);
+    }
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RAPShiftFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddTransferFactory(const RCP<const FactoryBase> &factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast, "MueLu::RAPShiftFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  transferFacts_.push_back(factory);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RAPSHIFTFACTORY_SHORT
-#endif // MUELU_RAPSHIFTFACTORY_DEF_HPP
+#endif  // MUELU_RAPSHIFTFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_SchurComplementFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_SchurComplementFactory_decl.hpp
index e365ab23c75c..2719f96fabda 100644
--- a/packages/muelu/src/Misc/MueLu_SchurComplementFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_SchurComplementFactory_decl.hpp
@@ -60,87 +60,84 @@
 #include <Xpetra_CrsMatrix_fwd.hpp>
 #include <Xpetra_Map_fwd.hpp>
 
-
 #include "MueLu_FactoryBase_fwd.hpp"
 #include "MueLu_SingleLevelFactoryBase.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class SchurComplementFactory class.
-    @brief Factory for building the Schur Complement for a 2x2 block matrix.
+/*!
+  @class SchurComplementFactory class.
+  @brief Factory for building the Schur Complement for a 2x2 block matrix.
 
-    ## Context, assumptions, and use cases ##
+  ## Context, assumptions, and use cases ##
 
-    This factory is intended to be used for building the schur complement for respective smoothers. The calculation is
-    for now restricted to 2x2 block matrices, where the schur complement is calculated on block A_11.
+  This factory is intended to be used for building the schur complement for respective smoothers. The calculation is
+  for now restricted to 2x2 block matrices, where the schur complement is calculated on block A_11.
 
-    For a blocked matrix \A = [A_00  A_01; A_10  A_11] it computes the Schur complement S = A_11 - 1/\omega A_10 Ainv A_01,
-    where \omega is some scaling factor and \Ainv an approximation of A_00^{-1} (from InverseApproximationFactory).
+  For a blocked matrix \A = [A_00  A_01; A_10  A_11] it computes the Schur complement S = A_11 - 1/\omega A_10 Ainv A_01,
+  where \omega is some scaling factor and \Ainv an approximation of A_00^{-1} (from InverseApproximationFactory).
 
-    ## Input/output of this factory ##
+  ## Input/output of this factory ##
 
-    ### User parameters of InterfaceAggregationFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    A     | Factory | null |  | * | * | Generating factory of the matrix A
-    Ainv  | Factory | null |  | * | * | Generating factory of the approximate inverse of A (produced by the InverseApproximationFactory)
-    omega | double  | 1.0  |  | * | * | Scaling factor
+  ### User parameters of InterfaceAggregationFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  A     | Factory | null |  | * | * | Generating factory of the matrix A
+  Ainv  | Factory | null |  | * | * | Generating factory of the approximate inverse of A (produced by the InverseApproximationFactory)
+  omega | double  | 1.0  |  | * | * | Scaling factor
 
-    The * in the master.xml column denotes that the parameter is defined in the master.xml file.
-    The * in the validated column means that the parameter is declared in the list of valid input parameters (see GetValidParameters() ).
-    The * in the requested column states that the data is requested as input with all dependencies (see DeclareInput() ).
+  The * in the master.xml column denotes that the parameter is defined in the master.xml file.
+  The * in the validated column means that the parameter is declared in the list of valid input parameters (see GetValidParameters() ).
+  The * in the requested column states that the data is requested as input with all dependencies (see DeclareInput() ).
 
-    ### Variables provided by this factory ###
+  ### Variables provided by this factory ###
 
-    After SchurComplementFactory::Build the following data is available (if requested)
+  After SchurComplementFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | A | SchurComplementFactory | The schur complement of the given block matrix.
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | A | SchurComplementFactory | The schur complement of the given block matrix.
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SchurComplementFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SchurComplementFactory : public SingleLevelFactoryBase {
 #undef MUELU_SCHURCOMPLEMENTFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
-
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+#include "MueLu_UseShortNames.hpp"
 
-    //! Constructor.
-    SchurComplementFactory() = default;
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Input
-    //@{
+  //! Constructor.
+  SchurComplementFactory() = default;
 
-    void DeclareInput(Level& currentLevel) const;
+  //! Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@{
-    //! @name Build methods.
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level& currentLevel) const;
+  //@{
+  //! @name Build methods.
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
+  //@}
 
-  private:
-    //! Schur complement calculation method.
-    RCP<Matrix> ComputeSchurComplement(RCP<BlockedCrsMatrix>& bA, RCP<Matrix>& Ainv) const;
+ private:
+  //! Schur complement calculation method.
+  RCP<Matrix> ComputeSchurComplement(RCP<BlockedCrsMatrix>& bA, RCP<Matrix>& Ainv) const;
 
-  }; // class SchurComplementFactory
+};  // class SchurComplementFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SCHURCOMPLEMENTFACTORY_SHORT
 #endif /* MUELU_SCHURCOMPLEMENTFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_SchurComplementFactory_def.hpp b/packages/muelu/src/Misc/MueLu_SchurComplementFactory_def.hpp
index 66cbd74fa792..c4503eec0d17 100644
--- a/packages/muelu/src/Misc/MueLu_SchurComplementFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_SchurComplementFactory_def.hpp
@@ -62,152 +62,148 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    const SC one = Teuchos::ScalarTraits<SC>::one();
-
-    validParamList->set<RCP<const FactoryBase> >("A"    , NoFactory::getRCP(), "Generating factory of the matrix A used for building Schur complement (must be a 2x2 blocked operator)");
-    validParamList->set<RCP<const FactoryBase> >("Ainv" , Teuchos::null,       "Generating factory of the inverse matrix used in the Schur complement");
-
-    validParamList->set<SC>                     ("omega", one, "Scaling parameter in S = A(1,1) - 1/omega A(1,0) Ainv A(0,1)");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    
-    // Get default or user-given inverse approximation factory
-    RCP<const FactoryBase> AinvFact = GetFactory("Ainv");
-    currentLevel.DeclareInput("Ainv", AinvFact.get(), this);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
-                               "MueLu::SchurComplementFactory::Build: input matrix A is not of type BlockedCrsMatrix!");
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != 2 || bA->Cols() != 2, Exceptions::RuntimeError,
-                               "MueLu::SchurComplementFactory::Build: input matrix A is a " << bA->Rows() << "x" << bA->Cols() << " block matrix. We expect a 2x2 blocked operator.");
-
-    // Calculate Schur Complement
-    RCP<Matrix> Ainv = currentLevel.Get<RCP<Matrix> >("Ainv", this->GetFactory("Ainv").get());
-    RCP<Matrix> S = ComputeSchurComplement(bA, Ainv);
-
-    GetOStream(Statistics1) << "S has " << S->getGlobalNumRows() << "x" << S->getGlobalNumCols() << " rows and columns." << std::endl;
-
-    // NOTE: "A" generated by this factory is actually the Schur complement
-    // matrix, but it is required as all smoothers expect "A"
-    Set(currentLevel, "A", S);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
-  SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeSchurComplement(RCP<BlockedCrsMatrix>& bA, RCP<Matrix>& Ainv) const {
-
-    using STS = Teuchos::ScalarTraits<SC>;
-    const SC zero = STS::zero(), one = STS::one();
-
-    RCP<Matrix> A01 = bA->getMatrix(0,1);
-    RCP<Matrix> A10 = bA->getMatrix(1,0);
-    RCP<Matrix> A11 = bA->getMatrix(1,1);
-
-    RCP<BlockedCrsMatrix> bA01 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A01);
-    const bool isBlocked = (bA01 == Teuchos::null ? false : true);
-
-    const ParameterList& pL = GetParameterList();
-    const SC omega = pL.get<Scalar>("omega");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(omega == zero, Exceptions::RuntimeError,
-                               "MueLu::SchurComplementFactory::Build: Scaling parameter omega must not be zero to avoid division by zero.");
-
-    RCP<Matrix> S = Teuchos::null; // Schur complement
-    RCP<Matrix> D = Teuchos::null; // temporary result for A10*Ainv*A01
-
-    // only if the off-diagonal blocks A10 and A01 are non-zero we have to do the MM multiplication
-    if(A01.is_null() == false && A10.is_null() == false) {
-      // scale with -1/omega
-      Ainv->scale(Teuchos::as<Scalar>(-one/omega));
-
-      // build Schur complement operator
-      if (!isBlocked) {
-        RCP<ParameterList> myparams = rcp(new ParameterList);
-        myparams->set("compute global constants", true);
-
-        // -1/omega*Ainv*A01
-        TEUCHOS_TEST_FOR_EXCEPTION(A01->getRangeMap()->isSameAs(*(Ainv->getDomainMap())) == false, Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: RangeMap of A01 and domain map of Ainv are not the same.");
-        RCP<Matrix> C = MatrixMatrix::Multiply(*Ainv, false, *A01, false, GetOStream(Statistics2), true, true, std::string("SchurComplementFactory"), myparams);
-
-        // -1/omega*A10*Ainv*A01
-        TEUCHOS_TEST_FOR_EXCEPTION(A01->getRangeMap()->isSameAs(*(A10->getDomainMap())) == false, Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: RangeMap of A10 and domain map A01 are not the same.");
-        D = MatrixMatrix::Multiply(*A10, false, *C, false, GetOStream(Statistics2), true, true, std::string("SchurComplementFactory"), myparams);
-      }
-      else {
-        // nested blocking
-        auto bA10  = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A10);
-        auto bAinv = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Ainv);
-        TEUCHOS_TEST_FOR_EXCEPTION(bAinv == Teuchos::null, Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: Casting Ainv to BlockedCrsMatrix not possible.");
-
-        // -1/omega*bAinv*bA01
-        TEUCHOS_TEST_FOR_EXCEPTION(bA01->Rows() != bAinv->Cols(), Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: Block rows and cols of bA01 and bAinv are not compatible.");
-        RCP<BlockedCrsMatrix> C = MatrixMatrix::TwoMatrixMultiplyBlock(*bAinv, false, *bA01, false, GetOStream(Statistics2));
-
-        // -1/omega*A10*Ainv*A01
-        TEUCHOS_TEST_FOR_EXCEPTION(bA10->Rows() != bA01->Cols(), Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: Block rows and cols of bA10 and bA01 are not compatible.");
-        D = MatrixMatrix::TwoMatrixMultiplyBlock(*bA10, false, *C, false, GetOStream(Statistics2));
-      }
-      if (!A11.is_null()) {
-        MatrixMatrix::TwoMatrixAdd(*A11, false, one, *D, false, one, S, GetOStream(Statistics2));
-        S->fillComplete();
-
-        TEUCHOS_TEST_FOR_EXCEPTION(A11->getRangeMap()->isSameAs(*(S->getRangeMap())) == false, Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: RangeMap of A11 and S are not the same.");
-        TEUCHOS_TEST_FOR_EXCEPTION(A11->getDomainMap()->isSameAs(*(S->getDomainMap())) == false, Exceptions::RuntimeError,
-                                   "MueLu::SchurComplementFactory::Build: DomainMap of A11 and S are not the same.");
-      }
-      else {
-        S = MatrixFactory::BuildCopy(D);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  const SC one = Teuchos::ScalarTraits<SC>::one();
+
+  validParamList->set<RCP<const FactoryBase>>("A", NoFactory::getRCP(), "Generating factory of the matrix A used for building Schur complement (must be a 2x2 blocked operator)");
+  validParamList->set<RCP<const FactoryBase>>("Ainv", Teuchos::null, "Generating factory of the inverse matrix used in the Schur complement");
+
+  validParamList->set<SC>("omega", one, "Scaling parameter in S = A(1,1) - 1/omega A(1,0) Ainv A(0,1)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+
+  // Get default or user-given inverse approximation factory
+  RCP<const FactoryBase> AinvFact = GetFactory("Ainv");
+  currentLevel.DeclareInput("Ainv", AinvFact.get(), this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<Matrix> A            = Get<RCP<Matrix>>(currentLevel, "A");
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
+                             "MueLu::SchurComplementFactory::Build: input matrix A is not of type BlockedCrsMatrix!");
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != 2 || bA->Cols() != 2, Exceptions::RuntimeError,
+                             "MueLu::SchurComplementFactory::Build: input matrix A is a " << bA->Rows() << "x" << bA->Cols() << " block matrix. We expect a 2x2 blocked operator.");
+
+  // Calculate Schur Complement
+  RCP<Matrix> Ainv = currentLevel.Get<RCP<Matrix>>("Ainv", this->GetFactory("Ainv").get());
+  RCP<Matrix> S    = ComputeSchurComplement(bA, Ainv);
+
+  GetOStream(Statistics1) << "S has " << S->getGlobalNumRows() << "x" << S->getGlobalNumCols() << " rows and columns." << std::endl;
+
+  // NOTE: "A" generated by this factory is actually the Schur complement
+  // matrix, but it is required as all smoothers expect "A"
+  Set(currentLevel, "A", S);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+SchurComplementFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeSchurComplement(RCP<BlockedCrsMatrix>& bA, RCP<Matrix>& Ainv) const {
+  using STS     = Teuchos::ScalarTraits<SC>;
+  const SC zero = STS::zero(), one = STS::one();
+
+  RCP<Matrix> A01 = bA->getMatrix(0, 1);
+  RCP<Matrix> A10 = bA->getMatrix(1, 0);
+  RCP<Matrix> A11 = bA->getMatrix(1, 1);
+
+  RCP<BlockedCrsMatrix> bA01 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A01);
+  const bool isBlocked       = (bA01 == Teuchos::null ? false : true);
+
+  const ParameterList& pL = GetParameterList();
+  const SC omega          = pL.get<Scalar>("omega");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(omega == zero, Exceptions::RuntimeError,
+                             "MueLu::SchurComplementFactory::Build: Scaling parameter omega must not be zero to avoid division by zero.");
+
+  RCP<Matrix> S = Teuchos::null;  // Schur complement
+  RCP<Matrix> D = Teuchos::null;  // temporary result for A10*Ainv*A01
+
+  // only if the off-diagonal blocks A10 and A01 are non-zero we have to do the MM multiplication
+  if (A01.is_null() == false && A10.is_null() == false) {
+    // scale with -1/omega
+    Ainv->scale(Teuchos::as<Scalar>(-one / omega));
+
+    // build Schur complement operator
+    if (!isBlocked) {
+      RCP<ParameterList> myparams = rcp(new ParameterList);
+      myparams->set("compute global constants", true);
+
+      // -1/omega*Ainv*A01
+      TEUCHOS_TEST_FOR_EXCEPTION(A01->getRangeMap()->isSameAs(*(Ainv->getDomainMap())) == false, Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: RangeMap of A01 and domain map of Ainv are not the same.");
+      RCP<Matrix> C = MatrixMatrix::Multiply(*Ainv, false, *A01, false, GetOStream(Statistics2), true, true, std::string("SchurComplementFactory"), myparams);
+
+      // -1/omega*A10*Ainv*A01
+      TEUCHOS_TEST_FOR_EXCEPTION(A01->getRangeMap()->isSameAs(*(A10->getDomainMap())) == false, Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: RangeMap of A10 and domain map A01 are not the same.");
+      D = MatrixMatrix::Multiply(*A10, false, *C, false, GetOStream(Statistics2), true, true, std::string("SchurComplementFactory"), myparams);
+    } else {
+      // nested blocking
+      auto bA10  = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A10);
+      auto bAinv = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Ainv);
+      TEUCHOS_TEST_FOR_EXCEPTION(bAinv == Teuchos::null, Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: Casting Ainv to BlockedCrsMatrix not possible.");
+
+      // -1/omega*bAinv*bA01
+      TEUCHOS_TEST_FOR_EXCEPTION(bA01->Rows() != bAinv->Cols(), Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: Block rows and cols of bA01 and bAinv are not compatible.");
+      RCP<BlockedCrsMatrix> C = MatrixMatrix::TwoMatrixMultiplyBlock(*bAinv, false, *bA01, false, GetOStream(Statistics2));
+
+      // -1/omega*A10*Ainv*A01
+      TEUCHOS_TEST_FOR_EXCEPTION(bA10->Rows() != bA01->Cols(), Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: Block rows and cols of bA10 and bA01 are not compatible.");
+      D = MatrixMatrix::TwoMatrixMultiplyBlock(*bA10, false, *C, false, GetOStream(Statistics2));
     }
-    else {
-      if (!A11.is_null()) {
-        S = MatrixFactory::BuildCopy(A11);
-      } else {
-        S = MatrixFactory::Build(A11->getRowMap(), 10 /*A11->getLocalMaxNumRowEntries()*/);
-        S->fillComplete(A11->getDomainMap(),A11->getRangeMap());
-      }
+    if (!A11.is_null()) {
+      MatrixMatrix::TwoMatrixAdd(*A11, false, one, *D, false, one, S, GetOStream(Statistics2));
+      S->fillComplete();
+
+      TEUCHOS_TEST_FOR_EXCEPTION(A11->getRangeMap()->isSameAs(*(S->getRangeMap())) == false, Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: RangeMap of A11 and S are not the same.");
+      TEUCHOS_TEST_FOR_EXCEPTION(A11->getDomainMap()->isSameAs(*(S->getDomainMap())) == false, Exceptions::RuntimeError,
+                                 "MueLu::SchurComplementFactory::Build: DomainMap of A11 and S are not the same.");
+    } else {
+      S = MatrixFactory::BuildCopy(D);
     }
-
-    // Check whether Schur complement operator is a 1x1 block matrix.
-    // If so, unwrap it and return the CrsMatrix based Matrix object
-    // We need this, as single-block smoothers expect it this way.
-    // In case of Thyra GIDs we obtain a Schur complement operator in Thyra GIDs
-    // This may make some special handling in feeding the SchurComplement solver Apply routine
-    // necessary!
-    if (isBlocked) {
-      RCP<BlockedCrsMatrix> bS = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(S);
-
-      if (bS != Teuchos::null && bS->Rows() == 1 && bS->Cols() == 1) {
-        RCP<Matrix> temp = bS->getCrsMatrix();
-        S.swap(temp);
-      }
+  } else {
+    if (!A11.is_null()) {
+      S = MatrixFactory::BuildCopy(A11);
+    } else {
+      S = MatrixFactory::Build(A11->getRowMap(), 10 /*A11->getLocalMaxNumRowEntries()*/);
+      S->fillComplete(A11->getDomainMap(), A11->getRangeMap());
     }
+  }
 
-    return S;
+  // Check whether Schur complement operator is a 1x1 block matrix.
+  // If so, unwrap it and return the CrsMatrix based Matrix object
+  // We need this, as single-block smoothers expect it this way.
+  // In case of Thyra GIDs we obtain a Schur complement operator in Thyra GIDs
+  // This may make some special handling in feeding the SchurComplement solver Apply routine
+  // necessary!
+  if (isBlocked) {
+    RCP<BlockedCrsMatrix> bS = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(S);
+
+    if (bS != Teuchos::null && bS->Rows() == 1 && bS->Cols() == 1) {
+      RCP<Matrix> temp = bS->getCrsMatrix();
+      S.swap(temp);
+    }
   }
 
-} // namespace MueLu
+  return S;
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_SCHURCOMPLEMENTFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Misc/MueLu_SegregatedAFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_SegregatedAFactory_decl.hpp
index 698be7f781aa..18f87c51dd0c 100644
--- a/packages/muelu/src/Misc/MueLu_SegregatedAFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_SegregatedAFactory_decl.hpp
@@ -55,62 +55,60 @@
 
 namespace MueLu {
 
-  /*!
-    @class SegregatedAFactory class.
-    @brief Factory for building a new "segregated" A operator. Here, "segregated" means that the user
-           provides a map (containing a subset of the row gids of the input matrix A) and the factory
-           drops the off-diagonal entries (a,b) and (b,a) in A where "a" denotes a GID entry in the provided map
-           and "b" denotes a GID that is not contained in the provided map.
-
-           The idea is to use the output matrix A as input for the aggregation factory to have control over
-           the aggregates and make sure that aggregates do not cross certain areas.
-
-           Note: we have to drop the entries (i.e. not just set them to zero) as the CoalesceDropFactory
-                 does not distinguish between matrix entries which are zero and nonzero.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SegregatedAFactory : public SingleLevelFactoryBase {
+/*!
+  @class SegregatedAFactory class.
+  @brief Factory for building a new "segregated" A operator. Here, "segregated" means that the user
+         provides a map (containing a subset of the row gids of the input matrix A) and the factory
+         drops the off-diagonal entries (a,b) and (b,a) in A where "a" denotes a GID entry in the provided map
+         and "b" denotes a GID that is not contained in the provided map.
+
+         The idea is to use the output matrix A as input for the aggregation factory to have control over
+         the aggregates and make sure that aggregates do not cross certain areas.
+
+         Note: we have to drop the entries (i.e. not just set them to zero) as the CoalesceDropFactory
+               does not distinguish between matrix entries which are zero and nonzero.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SegregatedAFactory : public SingleLevelFactoryBase {
 #undef MUELU_SEGREGATEDAFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! Constructor.
-    SegregatedAFactory() = default;
+ public:
+  //! Constructor.
+  SegregatedAFactory() = default;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
-
-      Builds filtered matrix and returns it in <tt>currentLevel</tt>.
-      */
-    void Build(Level& currentLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
+    Builds filtered matrix and returns it in <tt>currentLevel</tt>.
+    */
+  void Build(Level& currentLevel) const;
 
-  private:
+  //@}
 
-    //! Generating factory of input variable
-    mutable RCP<const FactoryBase> mapFact_;
+ private:
+  //! Generating factory of input variable
+  mutable RCP<const FactoryBase> mapFact_;
 
-  }; //class SegregatedAFactory
+};  // class SegregatedAFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SEGREGATEDAFACTORY_SHORT
-#endif // MUELU_SEGREGATEDAFACTORY_DECL_HPP
+#endif  // MUELU_SEGREGATEDAFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_SegregatedAFactory_def.hpp b/packages/muelu/src/Misc/MueLu_SegregatedAFactory_def.hpp
index c2993c84253b..27d2daee78a3 100644
--- a/packages/muelu/src/Misc/MueLu_SegregatedAFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_SegregatedAFactory_def.hpp
@@ -57,112 +57,109 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-#undef  SET_VALID_ENTRY
+#undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used for filtering");
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A used for filtering");
 
-    validParamList->set< std::string >           ("map: name",    "", "Name of map (Xpetra::Map) provided by user containing the special DOFs.");
-    validParamList->set< std::string >           ("map: factory", "", "Name of generating factory for 'map: name'");
+  validParamList->set<std::string>("map: name", "", "Name of map (Xpetra::Map) provided by user containing the special DOFs.");
+  validParamList->set<std::string>("map: factory", "", "Name of generating factory for 'map: name'");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
 
-    const ParameterList& pL = GetParameterList();
-    std::string mapName                       = pL.get<std::string> ("map: name");
-    std::string mapFactName                   = pL.get<std::string> ("map: factory");
+  const ParameterList& pL = GetParameterList();
+  std::string mapName     = pL.get<std::string>("map: name");
+  std::string mapFactName = pL.get<std::string>("map: factory");
 
-    if (currentLevel.GetLevelID() == 0)
-    {
-      // Not needed, if the map is provided as user data
-      currentLevel.DeclareInput(mapName, NoFactory::get(), this);
-    }
-    else
-    {
-      // check whether user has provided a specific name for the MapFactory
-      if (mapFactName == "" || mapFactName == "NoFactory")
-        mapFact_ = MueLu::NoFactory::getRCP();
-      else if (mapFactName != "null")
-        mapFact_ = currentLevel.GetFactoryManager()->GetFactory(mapFactName);
-
-      // request map generated by mapFact_
-      currentLevel.DeclareInput(mapName, mapFact_.get(), this);
-    }
+  if (currentLevel.GetLevelID() == 0) {
+    // Not needed, if the map is provided as user data
+    currentLevel.DeclareInput(mapName, NoFactory::get(), this);
+  } else {
+    // check whether user has provided a specific name for the MapFactory
+    if (mapFactName == "" || mapFactName == "NoFactory")
+      mapFact_ = MueLu::NoFactory::getRCP();
+    else if (mapFactName != "null")
+      mapFact_ = currentLevel.GetFactoryManager()->GetFactory(mapFactName);
+
+    // request map generated by mapFact_
+    currentLevel.DeclareInput(mapName, mapFact_.get(), this);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Matrix filtering (segregation)", currentLevel);
+
+  RCP<Matrix> Ain = Get<RCP<Matrix>>(currentLevel, "A");
+
+  const ParameterList& pL = GetParameterList();
+  std::string mapName     = pL.get<std::string>("map: name");
+  std::string mapFact     = pL.get<std::string>("map: factory");
+
+  // fetch map from level
+  RCP<Map> map = Teuchos::null;
+  if (currentLevel.GetLevelID() == 0) {
+    map = currentLevel.Get<RCP<Map>>(mapName, NoFactory::get());
+    GetOStream(Statistics0) << "User provided map \"" << mapName << "\": length dimension=" << map->getGlobalNumElements() << std::endl;
+  } else {
+    if (currentLevel.IsAvailable(mapName, mapFact_.get()) == false)
+      GetOStream(Runtime0) << "User provided map \"" << mapName << "\" not found in Level class on level " << currentLevel.GetLevelID() << "." << std::endl;
+    map = currentLevel.Get<RCP<Map>>(mapName, mapFact_.get());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SegregatedAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Matrix filtering (segregation)", currentLevel);
-
-    RCP<Matrix> Ain = Get< RCP<Matrix> >(currentLevel, "A");
-
-    const ParameterList& pL = GetParameterList();
-    std::string mapName  = pL.get<std::string> ("map: name");
-    std::string mapFact  = pL.get<std::string> ("map: factory");
-
-    // fetch map from level
-    RCP<Map> map = Teuchos::null;
-    if (currentLevel.GetLevelID() == 0) {
-      map = currentLevel.Get<RCP<Map>>(mapName, NoFactory::get());
-      GetOStream(Statistics0) << "User provided map \"" << mapName << "\": length dimension=" << map->getGlobalNumElements() << std::endl;
-    } else {
-      if (currentLevel.IsAvailable(mapName, mapFact_.get()) == false)
-        GetOStream(Runtime0) << "User provided map \"" << mapName << "\" not found in Level class on level " << currentLevel.GetLevelID() << "." << std::endl;
-      map = currentLevel.Get<RCP<Map>>(mapName, mapFact_.get());
-    }
+  // create new empty Operator
+  Teuchos::RCP<Matrix> Aout = MatrixFactory::Build(Ain->getRowMap(), Ain->getGlobalMaxNumRowEntries());
 
-    // create new empty Operator
-    Teuchos::RCP<Matrix> Aout = MatrixFactory::Build(Ain->getRowMap(), Ain->getGlobalMaxNumRowEntries());
-
-    size_t numLocalRows = Ain->getLocalNumRows();
-    for(size_t row=0; row<numLocalRows; row++) { // how can i replace this by a parallel for?
-      GlobalOrdinal grid = Ain->getRowMap()->getGlobalElement(row); // global row id
-      bool isInMap = map->isNodeGlobalElement(grid);
-
-      // extract row information from input matrix
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals;
-      Ain->getLocalRowView(row, indices, vals);
-
-      // just copy all values in output
-      Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(),Teuchos::ScalarTraits<GlobalOrdinal>::zero());
-      Teuchos::ArrayRCP<Scalar>        valout(indices.size(),Teuchos::ScalarTraits<Scalar>::zero());
-
-      size_t nNonzeros = 0;
-      for(size_t i=0; i<(size_t)indices.size(); i++) { // or can be parallelize this loop?
-        GlobalOrdinal gcid = Ain->getColMap()->getGlobalElement(indices[i]); // global column id
-        bool isInMap2 = map->isNodeGlobalElement(gcid);
-
-        if (isInMap == isInMap2) {
-          indout [nNonzeros] = gcid;
-          valout [nNonzeros] = vals[i];
-          nNonzeros++;
-        }
-      }
-      indout.resize(nNonzeros);
-      valout.resize(nNonzeros);
+  size_t numLocalRows = Ain->getLocalNumRows();
+  for (size_t row = 0; row < numLocalRows; row++) {                // how can i replace this by a parallel for?
+    GlobalOrdinal grid = Ain->getRowMap()->getGlobalElement(row);  // global row id
+    bool isInMap       = map->isNodeGlobalElement(grid);
 
-      Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0,indout.size()), valout.view(0,valout.size()));
-    }
+    // extract row information from input matrix
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    Ain->getLocalRowView(row, indices, vals);
 
-    Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
+    // just copy all values in output
+    Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(), Teuchos::ScalarTraits<GlobalOrdinal>::zero());
+    Teuchos::ArrayRCP<Scalar> valout(indices.size(), Teuchos::ScalarTraits<Scalar>::zero());
 
-    // copy block size information
-    Aout->SetFixedBlockSize(Ain->GetFixedBlockSize());
+    size_t nNonzeros = 0;
+    for (size_t i = 0; i < (size_t)indices.size(); i++) {                   // or can be parallelize this loop?
+      GlobalOrdinal gcid = Ain->getColMap()->getGlobalElement(indices[i]);  // global column id
+      bool isInMap2      = map->isNodeGlobalElement(gcid);
 
-    GetOStream(Statistics0, 0) << "Nonzeros in A (input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering A: " << Aout->getGlobalNumEntries() << std::endl;
+      if (isInMap == isInMap2) {
+        indout[nNonzeros] = gcid;
+        valout[nNonzeros] = vals[i];
+        nNonzeros++;
+      }
+    }
+    indout.resize(nNonzeros);
+    valout.resize(nNonzeros);
 
-    Set(currentLevel, "A", Aout);
+    Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0, indout.size()), valout.view(0, valout.size()));
   }
 
-} //namespace MueLu
+  Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
+
+  // copy block size information
+  Aout->SetFixedBlockSize(Ain->GetFixedBlockSize());
+
+  GetOStream(Statistics0, 0) << "Nonzeros in A (input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering A: " << Aout->getGlobalNumEntries() << std::endl;
+
+  Set(currentLevel, "A", Aout);
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_SEGREGATEDAFACTORY_DEF_HPP
+#endif  // MUELU_SEGREGATEDAFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_decl.hpp
index 86491d0ad5af..cf177680a5bd 100644
--- a/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_decl.hpp
@@ -54,57 +54,55 @@
 
 namespace MueLu {
 
-  /*!
-    @class StructuredLineDetectionFactory class.
-    @brief Factory building line detection information on structured meshes
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class StructuredLineDetectionFactory : public SingleLevelFactoryBase {
+/*!
+  @class StructuredLineDetectionFactory class.
+  @brief Factory building line detection information on structured meshes
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class StructuredLineDetectionFactory : public SingleLevelFactoryBase {
 #undef MUELU_STRUCTUREDLINEDETECTIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    StructuredLineDetectionFactory() { }
+  StructuredLineDetectionFactory() {}
 
-    //! Destructor.
-    virtual ~StructuredLineDetectionFactory() { }
+  //! Destructor.
+  virtual ~StructuredLineDetectionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
-
-      Builds line detection information and stores it in currentLevel
-      */
-    void Build(Level& currentLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
+    Builds line detection information and stores it in currentLevel
+    */
+  void Build(Level& currentLevel) const;
 
-  private:
+  //@}
 
-  }; //class StructuredLineDetectionFactory
+ private:
+};  // class StructuredLineDetectionFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_STRUCTUREDLINEDETECTIONFACTORY_SHORT
-#endif // MUELU_STRUCTUREDLINEDETECTIONFACTORY_DECL_HPP
+#endif  // MUELU_STRUCTUREDLINEDETECTIONFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_def.hpp b/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_def.hpp
index 86c2e0493498..54c6dc7a5116 100644
--- a/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_StructuredLineDetectionFactory_def.hpp
@@ -54,76 +54,75 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",               Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< std::string >           ("orientation",               "Z", "Lines orientation");
-    validParamList->set< RCP<const FactoryBase> >("lNodesPerDim",    Teuchos::null, "Number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    // Request the global number of nodes per dimensions
-    if(currentLevel.GetLevelID() == 0) {
-      if(currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<std::string>("orientation", "Z", "Lines orientation");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null, "Number of nodes per spatial dimension provided by CoordinatesTransferFactory.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  // Request the global number of nodes per dimensions
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      currentLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
     } else {
-      Input(currentLevel, "lNodesPerDim");
+      TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
     }
+  } else {
+    Input(currentLevel, "lNodesPerDim");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  // The following three variables are needed by the line smoothers in Ifpack/Ifpack2
+  LO NumZDir                       = 0;
+  Teuchos::ArrayRCP<LO> VertLineId = Teuchos::arcp<LO>(0);
+
+  // collect information provided by user
+  const ParameterList& pL           = GetParameterList();
+  const std::string lineOrientation = pL.get<std::string>("orientation");
+
+  // Extract data from currentLevel
+  RCP<Matrix> A          = Get<RCP<Matrix> >(currentLevel, "A");
+  Array<LO> lNodesPerDir = Get<Array<LO> >(currentLevel, "lNodesPerDim");
+  LO numNodes            = lNodesPerDir[0] * lNodesPerDir[1] * lNodesPerDir[2];
+  VertLineId.resize(numNodes);
+  if (lineOrientation == "X") {
+    NumZDir = lNodesPerDir[0];
+  } else if (lineOrientation == "Y") {
+    NumZDir = lNodesPerDir[1];
+  } else if (lineOrientation == "Z") {
+    NumZDir = lNodesPerDir[2];
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StructuredLineDetectionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-
-    // The following three variables are needed by the line smoothers in Ifpack/Ifpack2
-    LO                        NumZDir        = 0;
-    Teuchos::ArrayRCP<LO>     VertLineId     = Teuchos::arcp<LO>(0);
-
-    // collect information provided by user
-    const ParameterList& pL = GetParameterList();
-    const std::string lineOrientation = pL.get<std::string>("orientation");
-
-    // Extract data from currentLevel
-    RCP<Matrix>      A            = Get< RCP<Matrix> >(currentLevel, "A");
-    Array<LO>        lNodesPerDir = Get<Array<LO> >   (currentLevel, "lNodesPerDim");
-    LO numNodes    = lNodesPerDir[0]*lNodesPerDir[1]*lNodesPerDir[2];
-    VertLineId.resize(numNodes);
-    if(lineOrientation == "X") {
-      NumZDir = lNodesPerDir[0];
-    } else if(lineOrientation == "Y") {
-      NumZDir = lNodesPerDir[1];
-    } else if(lineOrientation == "Z") {
-      NumZDir = lNodesPerDir[2];
-    }
-
-    for(LO k = 0; k < lNodesPerDir[2]; ++k) {
-      for(LO j = 0; j < lNodesPerDir[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDir[0]; ++i) {
-          if(lineOrientation == "X") {
-            VertLineId[k*lNodesPerDir[1]*lNodesPerDir[0] + j*lNodesPerDir[0] + i] = k*lNodesPerDir[1] + j;
-          } else if(lineOrientation == "Y") {
-            VertLineId[k*lNodesPerDir[1]*lNodesPerDir[0] + j*lNodesPerDir[0] + i] = k*lNodesPerDir[0] + i;
-          } else if(lineOrientation == "Z") {
-            VertLineId[k*lNodesPerDir[1]*lNodesPerDir[0] + j*lNodesPerDir[0] + i] = j*lNodesPerDir[0] + i;
-          }
+  for (LO k = 0; k < lNodesPerDir[2]; ++k) {
+    for (LO j = 0; j < lNodesPerDir[1]; ++j) {
+      for (LO i = 0; i < lNodesPerDir[0]; ++i) {
+        if (lineOrientation == "X") {
+          VertLineId[k * lNodesPerDir[1] * lNodesPerDir[0] + j * lNodesPerDir[0] + i] = k * lNodesPerDir[1] + j;
+        } else if (lineOrientation == "Y") {
+          VertLineId[k * lNodesPerDir[1] * lNodesPerDir[0] + j * lNodesPerDir[0] + i] = k * lNodesPerDir[0] + i;
+        } else if (lineOrientation == "Z") {
+          VertLineId[k * lNodesPerDir[1] * lNodesPerDir[0] + j * lNodesPerDir[0] + i] = j * lNodesPerDir[0] + i;
         }
       }
     }
-
-    Set(currentLevel, "CoarseNumZLayers", NumZDir);
-    Set(currentLevel, "LineDetection_VertLineIds", VertLineId);
   }
 
-} //namespace MueLu
+  Set(currentLevel, "CoarseNumZLayers", NumZDir);
+  Set(currentLevel, "LineDetection_VertLineIds", VertLineId);
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_STRUCTUREDLINEDETECTIONFACTORY_DEF_HPP
+#endif  // MUELU_STRUCTUREDLINEDETECTIONFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_decl.hpp b/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_decl.hpp
index 5f798befa2e8..5f57aac97893 100644
--- a/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_decl.hpp
+++ b/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_decl.hpp
@@ -59,52 +59,51 @@
 
 namespace MueLu {
 
-  /*!
-    @class ThresholdAFilterFactory class.
-    @brief Factory for building a thresholded operator.
+/*!
+  @class ThresholdAFilterFactory class.
+  @brief Factory for building a thresholded operator.
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ThresholdAFilterFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ThresholdAFilterFactory : public SingleLevelFactoryBase {
 #undef MUELU_THRESHOLDAFILTERFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    ThresholdAFilterFactory(const std::string& ename, const Scalar threshold, const bool keepDiagonal=true, const GlobalOrdinal expectedNNZperRow=-1);
+  //! Constructor.
+  ThresholdAFilterFactory(const std::string& ename, const Scalar threshold, const bool keepDiagonal = true, const GlobalOrdinal expectedNNZperRow = -1);
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    //! Build an object with this factory.
-    void Build(Level & currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    std::string        varName_;   ///< name of input and output variable
-    const Scalar       threshold_; ///< threshold parameter
-    const bool         keepDiagonal_;
-    const GlobalOrdinal expectedNNZperRow_;
+ private:
+  std::string varName_;     ///< name of input and output variable
+  const Scalar threshold_;  ///< threshold parameter
+  const bool keepDiagonal_;
+  const GlobalOrdinal expectedNNZperRow_;
 
+};  // class ThresholdAFilterFactory
 
-  }; // class ThresholdAFilterFactory
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_THRESHOLDAFILTERFACTORY_SHORT
-#endif // MUELU_THRESHOLDAFILTERFACTORY_DECL_HPP
+#endif  // MUELU_THRESHOLDAFILTERFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_def.hpp b/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_def.hpp
index 10da9befa53d..ad8621640bb4 100644
--- a/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_def.hpp
+++ b/packages/muelu/src/Misc/MueLu_ThresholdAFilterFactory_def.hpp
@@ -56,30 +56,31 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ThresholdAFilterFactory(const std::string& ename, const Scalar threshold, const bool keepDiagonal, const GlobalOrdinal expectedNNZperRow)
-    : varName_(ename), threshold_(threshold), keepDiagonal_(keepDiagonal), expectedNNZperRow_(expectedNNZperRow)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ThresholdAFilterFactory(const std::string& ename, const Scalar threshold, const bool keepDiagonal, const GlobalOrdinal expectedNNZperRow)
+  : varName_(ename)
+  , threshold_(threshold)
+  , keepDiagonal_(keepDiagonal)
+  , expectedNNZperRow_(expectedNNZperRow) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, varName_);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, varName_);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build (Level & currentLevel) const
-  {
-    FactoryMonitor m (*this, "A filter (thresholding)", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ThresholdAFilterFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "A filter (thresholding)", currentLevel);
 
-    RCP<Matrix> Ain = Get< RCP<Matrix> >(currentLevel, varName_);
-    RCP<CrsMatrixWrap> Aout =
-            MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThresholdedMatrix(Ain, threshold_, keepDiagonal_, expectedNNZperRow_);
+  RCP<Matrix> Ain = Get<RCP<Matrix> >(currentLevel, varName_);
+  RCP<CrsMatrixWrap> Aout =
+      MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThresholdedMatrix(Ain, threshold_, keepDiagonal_, expectedNNZperRow_);
 
-    GetOStream(Statistics0) << "Nonzeros in " << varName_ << "(input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering " << varName_ << " (parameter: " << threshold_ << "): " << Aout->getGlobalNumEntries() << std::endl;
-    currentLevel.Set(varName_, Teuchos::rcp_dynamic_cast<Matrix>(Aout), this);
-  }
+  GetOStream(Statistics0) << "Nonzeros in " << varName_ << "(input): " << Ain->getGlobalNumEntries() << ", Nonzeros after filtering " << varName_ << " (parameter: " << threshold_ << "): " << Aout->getGlobalNumEntries() << std::endl;
+  currentLevel.Set(varName_, Teuchos::rcp_dynamic_cast<Matrix>(Aout), this);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_THRESHOLDAFILTERFACTORY_DEF_HPP
+#endif  // MUELU_THRESHOLDAFILTERFACTORY_DEF_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_BaseClass.hpp b/packages/muelu/src/MueCentral/MueLu_BaseClass.hpp
index d291d05c4a9f..4b3f61292bca 100644
--- a/packages/muelu/src/MueCentral/MueLu_BaseClass.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_BaseClass.hpp
@@ -52,44 +52,42 @@
 
 namespace MueLu {
 
-  /*!
-     @class BaseClass class.
-     @brief Base class for MueLu classes
+/*!
+   @class BaseClass class.
+   @brief Base class for MueLu classes
 
-     @ingroup MueLuBaseClasses
-  */
-  class BaseClass
-    : public VerboseObject, public Describable
-  {
+   @ingroup MueLuBaseClasses
+*/
+class BaseClass
+  : public VerboseObject,
+    public Describable {
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-  public:
+  //! Destructor.
+  virtual ~BaseClass() {}
 
-    //! @name Constructors/Destructors
-    //@{
+  //@}
 
-    //! Destructor.
-    virtual ~BaseClass() {}
+};  // class BaseClass
 
-    //@}
-
-  }; // class BaseClass
-
-} // namespace MueLu
+}  // namespace MueLu
 
 //! Helper macro for implementing Describable::describe() for BaseClass objects.
 //  This macro defines ostream out0 that print only on root node. It print description() and indent the ostream.
 //  Note: Runtime1 displays basic parameter information when Parameters0 is not enabled.
-#define MUELU_DESCRIBE                                                  \
-  using std::endl;                                                      \
+#define MUELU_DESCRIBE                                                                                            \
+  using std::endl;                                                                                                \
   Teuchos::FancyOStream& out0 = (VerboseObject::GetProcRankVerbose() == 0) ? out : VerboseObject::GetBlackHole(); \
-                                                                        \
-  if ((verbLevel & Runtime1) && (!(verbLevel & Parameters0)))           \
-    out << description() << std::endl;                                  \
-  else if (verbLevel & Runtime0)                                        \
-    out << BaseClass::description() << std::endl;                       \
-                                                                        \
-  Teuchos::OSTab tab1(out);                                             \
+                                                                                                                  \
+  if ((verbLevel & Runtime1) && (!(verbLevel & Parameters0)))                                                     \
+    out << description() << std::endl;                                                                            \
+  else if (verbLevel & Runtime0)                                                                                  \
+    out << BaseClass::description() << std::endl;                                                                 \
+                                                                                                                  \
+  Teuchos::OSTab tab1(out);                                                                                       \
   //
 
 #define MUELU_BASECLASS_SHORT
-#endif // ifndef MUELU_BASECLASS_HPP
+#endif  // ifndef MUELU_BASECLASS_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Describable.cpp b/packages/muelu/src/MueCentral/MueLu_Describable.cpp
index c2bba36bc122..88d04d8023b3 100644
--- a/packages/muelu/src/MueCentral/MueLu_Describable.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_Describable.cpp
@@ -50,51 +50,50 @@
 
 namespace MueLu {
 
-    Describable::~Describable() { }
+Describable::~Describable() {}
 
-    void Describable::describe(Teuchos::FancyOStream &out_arg, const VerbLevel /* verbLevel */) const {
-      Teuchos::RCP<Teuchos::FancyOStream> out = rcp(&out_arg,false); //JG: no idea why we have to do that, but it's how Teuchos::Describable::describe() is implemented
-      Teuchos::OSTab tab(out);
-      *out << this->description() << std::endl;
-    }
+void Describable::describe(Teuchos::FancyOStream &out_arg, const VerbLevel /* verbLevel */) const {
+  Teuchos::RCP<Teuchos::FancyOStream> out = rcp(&out_arg, false);  // JG: no idea why we have to do that, but it's how Teuchos::Describable::describe() is implemented
+  Teuchos::OSTab tab(out);
+  *out << this->description() << std::endl;
+}
 
-    std::string Describable::description() const {
-      std::string str = Teuchos::Describable::description();
+std::string Describable::description() const {
+  std::string str = Teuchos::Describable::description();
 
-      // remove template parameters
-      size_t found = str.find_first_of("<");
-      if (found != std::string::npos)
-        return str.substr(0, found);
+  // remove template parameters
+  size_t found = str.find_first_of("<");
+  if (found != std::string::npos)
+    return str.substr(0, found);
 
-      return str;
-    }
+  return str;
+}
 
-    void Describable::describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel) const { describe(out, toMueLuVerbLevel(verbLevel)); }
+void Describable::describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel) const { describe(out, toMueLuVerbLevel(verbLevel)); }
 
-    std::string Describable::ShortClassName() const {
-      if ( shortClassName_.empty() )
-      {
-        std::string str = Teuchos::Describable::description();
+std::string Describable::ShortClassName() const {
+  if (shortClassName_.empty()) {
+    std::string str = Teuchos::Describable::description();
 
-        // remove template parameters
-        {
-          size_t found = str.find_first_of("<");
-          if (found != std::string::npos)
-            str = str.substr(0, found);
-        }
+    // remove template parameters
+    {
+      size_t found = str.find_first_of("<");
+      if (found != std::string::npos)
+        str = str.substr(0, found);
+    }
 
-        // remove namespace
-        {
-          size_t found = str.find_last_of(":");
-          if (found != std::string::npos)
-            str = str.substr(found+1);
-        }
-        shortClassName_ = str;
-      }
-      return shortClassName_;
+    // remove namespace
+    {
+      size_t found = str.find_last_of(":");
+      if (found != std::string::npos)
+        str = str.substr(found + 1);
     }
+    shortClassName_ = str;
+  }
+  return shortClassName_;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DESCRIBABLE_SHORT
-#endif // MUELU_DESCRIBABLE_HPP
+#endif  // MUELU_DESCRIBABLE_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Describable.hpp b/packages/muelu/src/MueCentral/MueLu_Describable.hpp
index 7dbb4dc08811..e6341b8246cc 100644
--- a/packages/muelu/src/MueCentral/MueLu_Describable.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_Describable.hpp
@@ -46,55 +46,53 @@
 #ifndef MUELU_DESCRIBABLE_DECL_HPP
 #define MUELU_DESCRIBABLE_DECL_HPP
 
-#include <string>                       // for string
-#include "Teuchos_FancyOStream.hpp"     // for FancyOStream
-#include "Teuchos_VerbosityLevel.hpp"   // for EVerbosityLevel
+#include <string>                      // for string
+#include "Teuchos_FancyOStream.hpp"    // for FancyOStream
+#include "Teuchos_VerbosityLevel.hpp"  // for EVerbosityLevel
 #include "Teuchos_Describable.hpp"
 
 #include "MueLu_VerbosityLevel.hpp"
 
 namespace MueLu {
 
-  /*!
-     @class Describable
-     @brief Base class for MueLu classes
+/*!
+   @class Describable
+   @brief Base class for MueLu classes
 
-     @ingroup MueLuBaseClasses
-  */
-  class Describable
-    : public Teuchos::Describable
-  {
-    mutable std::string shortClassName_ = ""; // cached so that we don't have to call demangleName() every time; mutable so that ShortClassName() can initialize lazily while remaining const
+   @ingroup MueLuBaseClasses
+*/
+class Describable
+  : public Teuchos::Describable {
+  mutable std::string shortClassName_ = "";  // cached so that we don't have to call demangleName() every time; mutable so that ShortClassName() can initialize lazily while remaining const
 
-  public:
+ public:
+  //! Destructor.
+  virtual ~Describable();
 
-    //! Destructor.
-    virtual ~Describable();
+  //! @name MueLu Describe
+  //@{
 
-    //! @name MueLu Describe
-    //@{
+  virtual void describe(Teuchos::FancyOStream &out_arg, const VerbLevel verbLevel = Default) const;
 
-    virtual void describe(Teuchos::FancyOStream &out_arg, const VerbLevel verbLevel = Default) const;
+  //@}
 
-    //@}
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! Return a simple one-line description of this object.
+  virtual std::string description() const;
 
-    //! Return a simple one-line description of this object.
-    virtual std::string description() const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::Describable::verbLevel_default) const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::Describable::verbLevel_default) const;
+  //@}
 
-    //@}
+  //! Return the class name of the object, without template parameters and without namespace
+  virtual std::string ShortClassName() const;
 
-    //! Return the class name of the object, without template parameters and without namespace
-    virtual std::string ShortClassName() const;
+};  // class Describable
 
-  }; // class Describable
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DESCRIBABLE_SHORT
-#endif // MUELU_DESCRIBABLE_DECL_HPP
+#endif  // MUELU_DESCRIBABLE_DECL_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Factory.cpp b/packages/muelu/src/MueCentral/MueLu_Factory.cpp
index 64c7032ce5bb..4b099c244891 100644
--- a/packages/muelu/src/MueCentral/MueLu_Factory.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_Factory.cpp
@@ -48,9 +48,9 @@
 
 namespace MueLu {
 
-  bool Factory::timerSync_ = false;
+bool Factory::timerSync_ = false;
 #ifdef HAVE_MUELU_DEBUG
-  Factory::multipleCallCheckEnum Factory::multipleCallCheckGlobal_ = ENABLED;
+Factory::multipleCallCheckEnum Factory::multipleCallCheckGlobal_ = ENABLED;
 #endif
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_Factory.hpp b/packages/muelu/src/MueCentral/MueLu_Factory.hpp
index 0be93d76bdff..ef38498019c1 100644
--- a/packages/muelu/src/MueCentral/MueLu_Factory.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_Factory.hpp
@@ -47,14 +47,14 @@
 #define MUELU_FACTORY_HPP
 
 #include <string>
-#include <deque>                        // for _Deque_iterator, operator!=
-#include <ostream>                      // for operator<<, etc
-#include "Teuchos_ENull.hpp"            // for ENull::null
+#include <deque>                         // for _Deque_iterator, operator!=
+#include <ostream>                       // for operator<<, etc
+#include "Teuchos_ENull.hpp"             // for ENull::null
 #include "Teuchos_FilteredIterator.hpp"  // for FilteredIterator, etc
-#include "Teuchos_ParameterEntry.hpp"   // for ParameterEntry
-#include "Teuchos_ParameterList.hpp"    // for ParameterList, etc
-#include "Teuchos_RCPDecl.hpp"          // for RCP
-#include "Teuchos_RCPNode.hpp"          // for operator<<
+#include "Teuchos_ParameterEntry.hpp"    // for ParameterEntry
+#include "Teuchos_ParameterList.hpp"     // for ParameterList, etc
+#include "Teuchos_RCPDecl.hpp"           // for RCP
+#include "Teuchos_RCPNode.hpp"           // for operator<<
 #include "Teuchos_StringIndexedOrderedValueObjectContainer.hpp"
 #include "Teuchos_RCP.hpp"
 
@@ -66,171 +66,170 @@
 
 namespace MueLu {
 
-  class Factory : public FactoryBase, public FactoryAcceptor, public ParameterListAcceptorImpl {
+class Factory : public FactoryBase, public FactoryAcceptor, public ParameterListAcceptorImpl {
+ public:
+  //@{ Constructors/Destructors.
 
-  public:
-    //@{ Constructors/Destructors.
-
-    //! Constructor.
-    Factory()
+  //! Constructor.
+  Factory()
 #ifdef HAVE_MUELU_DEBUG
-      : multipleCallCheck_(FIRSTCALL), lastLevelID_(-1)
+    : multipleCallCheck_(FIRSTCALL)
+    , lastLevelID_(-1)
 #endif
-    { }
-
-    //! Destructor.
-    virtual ~Factory() { }
-    //@}
-
-    //@{
-    //! Configuration
-
-    //! SetFactory is for expert users only. To change configuration of the preconditioner, use a factory manager.
-    virtual void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
-      RCP<const FactoryBase> f = factory;
-      SetParameter(varName, ParameterEntry(f)); // parameter validation done in ParameterListAcceptorImpl
+  {
+  }
+
+  //! Destructor.
+  virtual ~Factory() {}
+  //@}
+
+  //@{
+  //! Configuration
+
+  //! SetFactory is for expert users only. To change configuration of the preconditioner, use a factory manager.
+  virtual void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
+    RCP<const FactoryBase> f = factory;
+    SetParameter(varName, ParameterEntry(f));  // parameter validation done in ParameterListAcceptorImpl
+  }
+
+  //! Default implementation of FactoryAcceptor::GetFactory()
+  const RCP<const FactoryBase> GetFactory(const std::string& varName) const {
+    // Special treatment for "NoFactory"
+    if (varName == "NoFactory")
+      return MueLu::NoFactory::getRCP();
+
+    if (!GetParameterList().isParameter(varName) && GetValidParameterList() == Teuchos::null) {
+      // If the parameter is not on the list and there is not validator, the defaults values for 'varName' is not set.
+      // Failback by using directly the FactoryManager
+      // NOTE: call to GetValidParameterList() can be costly for classes that validate parameters.
+      // But it get called only (lazy '&&' operator) if the parameter 'varName' is not on the paramlist and
+      // the parameter 'varName' is always on the list when validator is present and 'varName' is valid (at least the default value is set).
+      return Teuchos::null;
     }
 
-    //! Default implementation of FactoryAcceptor::GetFactory()
-    const RCP<const FactoryBase> GetFactory(const std::string& varName) const {
-
-      // Special treatment for "NoFactory"
-      if (varName == "NoFactory")
-        return MueLu::NoFactory::getRCP();
-
-      if (!GetParameterList().isParameter(varName)&& GetValidParameterList() == Teuchos::null) {
-        // If the parameter is not on the list and there is not validator, the defaults values for 'varName' is not set.
-        // Failback by using directly the FactoryManager
-        // NOTE: call to GetValidParameterList() can be costly for classes that validate parameters.
-        // But it get called only (lazy '&&' operator) if the parameter 'varName' is not on the paramlist and
-        // the parameter 'varName' is always on the list when validator is present and 'varName' is valid (at least the default value is set).
-        return Teuchos::null;
-      }
-
-      return GetParameterList().get< RCP<const FactoryBase> >(varName);
+    return GetParameterList().get<RCP<const FactoryBase> >(varName);
+  }
+
+  RCP<ParameterList> RemoveFactoriesFromList(const ParameterList& list) const {
+    RCP<ParameterList> paramList = rcp(new ParameterList(list));
+    // Remove FactoryBase entries from the list
+    // The solution would be much more elegant if ParameterList support std::list like operations
+    // In that case, we could simply write:
+    //   for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++)
+    //     if (paramList.isType<RCP<const FactoryBase> >(it->first))
+    //       it = paramList.erase(it);
+    //     else
+    //       it++;
+    ParameterList::ConstIterator it = paramList->begin();
+    while (it != paramList->end()) {
+      it = paramList->begin();
+
+      for (; it != paramList->end(); it++)
+        if (paramList->isType<RCP<const FactoryBase> >(it->first))
+          paramList->remove(it->first);
     }
-
-    RCP<ParameterList> RemoveFactoriesFromList(const ParameterList& list) const {
-      RCP<ParameterList> paramList = rcp(new ParameterList(list));
-      // Remove FactoryBase entries from the list
-      // The solution would be much more elegant if ParameterList support std::list like operations
-      // In that case, we could simply write:
-      //   for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++)
-      //     if (paramList.isType<RCP<const FactoryBase> >(it->first))
-      //       it = paramList.erase(it);
-      //     else
-      //       it++;
-      ParameterList::ConstIterator it = paramList->begin();
-      while (it != paramList->end()) {
-        it = paramList->begin();
-
-        for (; it != paramList->end(); it++)
-          if (paramList->isType<RCP<const FactoryBase> >(it->first))
-            paramList->remove(it->first);
-      }
-      return paramList;
-    }
-
-    // SetParameterList(...);
-
-    // GetParameterList(...);
-
-    //@}
-
-    virtual RCP<const ParameterList> GetValidParameterList() const {
-      return Teuchos::null;  // Teuchos::null == GetValidParameterList() not implemented == skip validation and no default values (dangerous)
-    }
-
-  protected:
-
-    void Input(Level& level, const std::string& varName) const {
-      level.DeclareInput(varName, GetFactory(varName).get(), this);
-    }
-    // Similar to the other Input, but we have an alias (varParamName) to the generated data name (varName)
-    void Input(Level& level, const std::string& varName, const std::string& varParamName) const {
-      level.DeclareInput(varName, GetFactory(varParamName).get(), this);
-    }
-
-    template <class T>
-    T Get(Level& level, const std::string& varName) const {
-      return level.Get<T>(varName, GetFactory(varName).get());
-    }
-    // Similar to the other Get, but we have an alias (varParamName) to the generated data name (varName)
-    template <class T>
-    T Get(Level& level, const std::string& varName, const std::string& varParamName) const {
-      return level.Get<T>(varName, GetFactory(varParamName).get());
-    }
-
-    template <class T>
-    void Set(Level& level, const std::string& varName, const T& data) const {
-      return level.Set<T>(varName, data, this);
-    }
-
-    template <class T>
-    bool IsType(Level& level, const std::string& varName) const {
-      return level.IsType<T>(varName, GetFactory(varName).get());
-    }
-
-    bool IsAvailable(Level& level, const std::string& varName) const {
-      return level.IsAvailable(varName, GetFactory(varName).get());
-    }
-
-  public:
-    static void EnableTimerSync() { timerSync_ = true;  }
-    static void DisableTimerSync() { timerSync_ = false; }
-
-  protected:
-    static bool timerSync_;
+    return paramList;
+  }
+
+  // SetParameterList(...);
+
+  // GetParameterList(...);
+
+  //@}
+
+  virtual RCP<const ParameterList> GetValidParameterList() const {
+    return Teuchos::null;  // Teuchos::null == GetValidParameterList() not implemented == skip validation and no default values (dangerous)
+  }
+
+ protected:
+  void Input(Level& level, const std::string& varName) const {
+    level.DeclareInput(varName, GetFactory(varName).get(), this);
+  }
+  // Similar to the other Input, but we have an alias (varParamName) to the generated data name (varName)
+  void Input(Level& level, const std::string& varName, const std::string& varParamName) const {
+    level.DeclareInput(varName, GetFactory(varParamName).get(), this);
+  }
+
+  template <class T>
+  T Get(Level& level, const std::string& varName) const {
+    return level.Get<T>(varName, GetFactory(varName).get());
+  }
+  // Similar to the other Get, but we have an alias (varParamName) to the generated data name (varName)
+  template <class T>
+  T Get(Level& level, const std::string& varName, const std::string& varParamName) const {
+    return level.Get<T>(varName, GetFactory(varParamName).get());
+  }
+
+  template <class T>
+  void Set(Level& level, const std::string& varName, const T& data) const {
+    return level.Set<T>(varName, data, this);
+  }
+
+  template <class T>
+  bool IsType(Level& level, const std::string& varName) const {
+    return level.IsType<T>(varName, GetFactory(varName).get());
+  }
+
+  bool IsAvailable(Level& level, const std::string& varName) const {
+    return level.IsAvailable(varName, GetFactory(varName).get());
+  }
+
+ public:
+  static void EnableTimerSync() { timerSync_ = true; }
+  static void DisableTimerSync() { timerSync_ = false; }
+
+ protected:
+  static bool timerSync_;
 
 #ifdef HAVE_MUELU_DEBUG
-  public:
-    enum           multipleCallCheckEnum { ENABLED, DISABLED, FIRSTCALL };
-
-    void EnableMultipleCallCheck() const       { multipleCallCheck_       = ENABLED;  }
-    void DisableMultipleCallCheck() const      { multipleCallCheck_       = DISABLED; }
-    void ResetDebugData() const {
-      if (multipleCallCheck_ == FIRSTCALL && lastLevelID_ == -1)
-        return;
-
-      multipleCallCheck_ = FIRSTCALL;
-      lastLevelID_       = -1;
-
-      const ParameterList& paramList = GetParameterList();
-
-      // We cannot use just FactoryManager to specify which factories call ResetDebugData().
-      // The problem is that some factories are not present in the manager, but
-      // instead are only accessible through a parameter list of some factory.
-      // For instance, FilteredAFactory is only accessible from SaPFactory but
-      // nowhere else. So we miss those, and do not reset the data, resulting
-      // in problems.
-      // Therefore, for each factory we need to go through its dependent
-      // factories, and call reset on them.
-      for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++)
-        if (paramList.isType<RCP<const FactoryBase> >(it->first)) {
-          RCP<const Factory> fact = rcp_dynamic_cast<const Factory >(paramList.get<RCP<const FactoryBase> >(it->first));
-          if (fact != Teuchos::null && fact != NoFactory::getRCP())
-            fact->ResetDebugData();
-        }
-    }
+ public:
+  enum multipleCallCheckEnum{ENABLED, DISABLED, FIRSTCALL};
+
+  void EnableMultipleCallCheck() const { multipleCallCheck_ = ENABLED; }
+  void DisableMultipleCallCheck() const { multipleCallCheck_ = DISABLED; }
+  void ResetDebugData() const {
+    if (multipleCallCheck_ == FIRSTCALL && lastLevelID_ == -1)
+      return;
+
+    multipleCallCheck_ = FIRSTCALL;
+    lastLevelID_       = -1;
+
+    const ParameterList& paramList = GetParameterList();
+
+    // We cannot use just FactoryManager to specify which factories call ResetDebugData().
+    // The problem is that some factories are not present in the manager, but
+    // instead are only accessible through a parameter list of some factory.
+    // For instance, FilteredAFactory is only accessible from SaPFactory but
+    // nowhere else. So we miss those, and do not reset the data, resulting
+    // in problems.
+    // Therefore, for each factory we need to go through its dependent
+    // factories, and call reset on them.
+    for (ParameterList::ConstIterator it = paramList.begin(); it != paramList.end(); it++)
+      if (paramList.isType<RCP<const FactoryBase> >(it->first)) {
+        RCP<const Factory> fact = rcp_dynamic_cast<const Factory>(paramList.get<RCP<const FactoryBase> >(it->first));
+        if (fact != Teuchos::null && fact != NoFactory::getRCP())
+          fact->ResetDebugData();
+      }
+  }
 
-    static void EnableMultipleCheckGlobally()  { multipleCallCheckGlobal_ = ENABLED;  }
-    static void DisableMultipleCheckGlobally() { multipleCallCheckGlobal_ = DISABLED; }
+  static void EnableMultipleCheckGlobally() { multipleCallCheckGlobal_ = ENABLED; }
+  static void DisableMultipleCheckGlobally() { multipleCallCheckGlobal_ = DISABLED; }
 
-  protected:
-    mutable        multipleCallCheckEnum multipleCallCheck_;
-    static         multipleCallCheckEnum multipleCallCheckGlobal_;
-    mutable int    lastLevelID_;
+ protected:
+  mutable multipleCallCheckEnum multipleCallCheck_;
+  static multipleCallCheckEnum multipleCallCheckGlobal_;
+  mutable int lastLevelID_;
 #else
-  public:
-    void EnableMultipleCallCheck() const       { }
-    void DisableMultipleCallCheck() const      { }
-    void ResetDebugData() const                { }
-    static void EnableMultipleCheckGlobally()  { }
-    static void DisableMultipleCheckGlobally() { }
+ public:
+  void EnableMultipleCallCheck() const {}
+  void DisableMultipleCallCheck() const {}
+  void ResetDebugData() const {}
+  static void EnableMultipleCheckGlobally() {}
+  static void DisableMultipleCheckGlobally() {}
 #endif
-  }; //class Factory
+};  // class Factory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACTORY_SHORT
-#endif //ifndef MUELU_FACTORY_HPP
+#endif  // ifndef MUELU_FACTORY_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryAcceptor.hpp b/packages/muelu/src/MueCentral/MueLu_FactoryAcceptor.hpp
index efe14effc27a..4a23f4e6dce6 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryAcceptor.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryAcceptor.hpp
@@ -54,29 +54,27 @@
 
 namespace MueLu {
 
-  class FactoryAcceptor {
+class FactoryAcceptor {
+ public:
+  virtual ~FactoryAcceptor() {}
 
-  public:
+  //@{
+  //! Configuration
 
-    virtual ~FactoryAcceptor() { }
+  //! SetFactory is for expert users only. To change configuration of the preconditioner, use a factory manager.
+  virtual void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) = 0;
 
-    //@{
-    //! Configuration
+  virtual const RCP<const FactoryBase> GetFactory(const std::string& varName) const = 0;
 
-    //! SetFactory is for expert users only. To change configuration of the preconditioner, use a factory manager.
-    virtual void SetFactory(const std::string & varName, const RCP<const FactoryBase> & factory) = 0;
+  // SetParameterList(...);
 
-    virtual const RCP<const FactoryBase> GetFactory(const std::string & varName) const = 0;
+  // GetParameterList(...);
 
-    // SetParameterList(...);
+  //@}
 
-    // GetParameterList(...);
+};  // class FactoryAcceptor
 
-    //@}
-
-  }; //class FactoryAcceptor
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACTORYACCEPTOR_SHORT
-#endif //ifndef MUELU_FACTORYACCEPTOR_HPP
+#endif  // ifndef MUELU_FACTORYACCEPTOR_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryBase.cpp b/packages/muelu/src/MueCentral/MueLu_FactoryBase.cpp
index 07c413c94e8f..ab5d21912b59 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryBase.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryBase.cpp
@@ -48,9 +48,9 @@
 
 namespace MueLu {
 
-  int FactoryBase::GenerateUniqueId() {
-    static int i = 0;
-    return i++;
-  }
+int FactoryBase::GenerateUniqueId() {
+  static int i = 0;
+  return i++;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryBase.hpp b/packages/muelu/src/MueCentral/MueLu_FactoryBase.hpp
index b962d7f6f510..13e422fa3400 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryBase.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryBase.hpp
@@ -52,54 +52,51 @@
 
 namespace MueLu {
 
-  /*!
-    @class FactoryBase
-    @brief Base class for factories (e.g., R, P, and A_coarse).
-    @ingroup MueLuBaseClasses
-  */
-  class FactoryBase : public virtual BaseClass {
-
-  public:
-    //@{ Constructors/Destructors.
-
-    //! Constructor.
-    FactoryBase()
-      : id_(FactoryBase::GenerateUniqueId())
-    { }
-
-    //! Destructor.
-    virtual ~FactoryBase() { }
-    //@}
+/*!
+  @class FactoryBase
+  @brief Base class for factories (e.g., R, P, and A_coarse).
+  @ingroup MueLuBaseClasses
+*/
+class FactoryBase : public virtual BaseClass {
+ public:
+  //@{ Constructors/Destructors.
 
-    //@{
-    //! @name Build methods.
+  //! Constructor.
+  FactoryBase()
+    : id_(FactoryBase::GenerateUniqueId()) {}
 
-    virtual void CallBuild(Level & requestedLevel) const = 0;
+  //! Destructor.
+  virtual ~FactoryBase() {}
+  //@}
 
-    virtual void CallDeclareInput(Level & requestedLevel) const = 0;
-    //@}
+  //@{
+  //! @name Build methods.
+
+  virtual void CallBuild(Level& requestedLevel) const = 0;
 
-    //@{
-    //! @name Access factory properties
+  virtual void CallDeclareInput(Level& requestedLevel) const = 0;
+  //@}
 
-    /// return unique factory id
-    int GetID() const { return id_; };
+  //@{
+  //! @name Access factory properties
+
+  /// return unique factory id
+  int GetID() const { return id_; };
 
     //@}
 
 #ifdef HAVE_MUELU_DEBUG
-    virtual void ResetDebugData() const = 0;
+  virtual void ResetDebugData() const = 0;
 #endif
 
-  private:
-
-    static int GenerateUniqueId();
+ private:
+  static int GenerateUniqueId();
 
-    const int id_;
+  const int id_;
 
-  }; //class FactoryBase
+};  // class FactoryBase
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACTORYBASE_SHORT
-#endif //ifndef MUELU_FACTORYBASE_HPP
+#endif  // ifndef MUELU_FACTORYBASE_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryManagerBase.hpp b/packages/muelu/src/MueCentral/MueLu_FactoryManagerBase.hpp
index 90b2c8f86089..742cd5c7c6ee 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryManagerBase.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryManagerBase.hpp
@@ -55,57 +55,57 @@
 
 namespace MueLu {
 
-  /*!
-     @class FactoryManagerBase
-     @brief Class that provides default factories within Needs class.
-     @ingroup MueLuBaseClasses
-  */
-  class FactoryManagerBase : public BaseClass {
+/*!
+   @class FactoryManagerBase
+   @brief Class that provides default factories within Needs class.
+   @ingroup MueLuBaseClasses
+*/
+class FactoryManagerBase : public BaseClass {
+ public:
+  //@{ Constructors/Destructors.
+  FactoryManagerBase()
+    : bIgnoreUserData_(false) {}
 
-  public:
-    //@{ Constructors/Destructors.
-    FactoryManagerBase() : bIgnoreUserData_(false) { }
+  //! Destructor.
+  virtual ~FactoryManagerBase() {}
 
-    //! Destructor.
-    virtual ~FactoryManagerBase() { }
+  //@}
 
-    //@}
+  //@{ Get/Set functions.
 
-    //@{ Get/Set functions.
+  //! Get
+  // Return ref because user also give ref to the Hierarchy.
+  const virtual RCP<const FactoryBase> GetFactory(const std::string& varName) const = 0;
+  //@}
 
-    //! Get
-    // Return ref because user also give ref to the Hierarchy.
-    const virtual RCP<const FactoryBase> GetFactory(const std::string& varName) const = 0;
-    //@}
+  //! Check
+  // Return true if Factory associated with varName is registered
+  virtual bool hasFactory(const std::string& varName) const = 0;
 
-    //! Check
-    // Return true if Factory associated with varName is registered
-    virtual bool hasFactory(const std::string& varName) const = 0;
-
-    // Free temporarily hold data at the end of Hierarchy::Setup()
-    // This method is const because the clean concerns only mutable data.
-    virtual void Clean() const { } // TODO: should be used inside of MueLu::Hierarchy
+  // Free temporarily hold data at the end of Hierarchy::Setup()
+  // This method is const because the clean concerns only mutable data.
+  virtual void Clean() const {}  // TODO: should be used inside of MueLu::Hierarchy
 
 #ifdef HAVE_MUELU_DEBUG
-    virtual void ResetDebugData() const = 0;
+  virtual void ResetDebugData() const = 0;
 #endif
 
-    //! get IgnoreUserData flag
-    bool IgnoreUserData() const { return bIgnoreUserData_; }
+  //! get IgnoreUserData flag
+  bool IgnoreUserData() const { return bIgnoreUserData_; }
 
-    //! set IgnoreUserData flag
-    void SetIgnoreUserData(bool bIgnoreUserData = false) { bIgnoreUserData_ = bIgnoreUserData; }
+  //! set IgnoreUserData flag
+  void SetIgnoreUserData(bool bIgnoreUserData = false) { bIgnoreUserData_ = bIgnoreUserData; }
 
-  private:
-    //! boolean flag that controls behaviour of Level::GetFactory
-    //! if bIgnoreUserData == true,  the Level::GetFactory function always asks the Factory manager for a valid factory given a variable name
-    //! if bIgnoreUserData == false, the Level::GetFactory prefers user-provided data for a variable name if available. Otherwise the factory manager is asked for a valid factory
-    //! default: bIgnoreUserData = false;
-    bool bIgnoreUserData_;
+ private:
+  //! boolean flag that controls behaviour of Level::GetFactory
+  //! if bIgnoreUserData == true,  the Level::GetFactory function always asks the Factory manager for a valid factory given a variable name
+  //! if bIgnoreUserData == false, the Level::GetFactory prefers user-provided data for a variable name if available. Otherwise the factory manager is asked for a valid factory
+  //! default: bIgnoreUserData = false;
+  bool bIgnoreUserData_;
 
-  }; // class FactoryManagerBase
+};  // class FactoryManagerBase
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACTORYMANAGERBASE_SHORT
-#endif //ifndef MUELU_FACTORYMANAGERBASE_HPP
+#endif  // ifndef MUELU_FACTORYMANAGERBASE_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryManager_decl.hpp b/packages/muelu/src/MueCentral/MueLu_FactoryManager_decl.hpp
index 2f4618bb44c0..38348fcff50e 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryManager_decl.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryManager_decl.hpp
@@ -75,7 +75,6 @@
 #include "MueLu_InterfaceMappingTransferFactory_fwd.hpp"
 #include "MueLu_InterfaceAggregationFactory_fwd.hpp"
 
-
 #include "MueLu_CoalesceDropFactory_kokkos_fwd.hpp"
 #include "MueLu_NullspaceFactory_kokkos_fwd.hpp"
 #include "MueLu_SaPFactory_kokkos_fwd.hpp"
@@ -84,143 +83,139 @@
 
 namespace MueLu {
 
-  /*!
-    @class FactoryManager class.
-    @brief This class specifies the default factory that should generate some data on a Level if the data does not exist and
-    the generating factory has not been specified.
-
-    Consider the following example.
-
-    @code
-      RCP<SingleLevelFactory> Afact;
-      Level currentLevel;
-      RCP<Matrix> thisLevelA;
-      thisLevelA = currentLevel.Get<Matrix>("A", Afact.get());
-    @endcode
-
-    @todo If Afact is null (actually, Teuchos::null), then the FactoryManager associated with currentLevel will determine whether a default factory has
-    been specified for creating A.  If "yes", then that factory will be called, A will be stored in currentLevel, and an RCP will be returned by
-    the Get call. If "no", then the FactoryManager will <b>throw an exception indicating that it does not know how to generate A</b>.
-  */
-
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class FactoryManager : public FactoryManagerBase {
+/*!
+  @class FactoryManager class.
+  @brief This class specifies the default factory that should generate some data on a Level if the data does not exist and
+  the generating factory has not been specified.
+
+  Consider the following example.
+
+  @code
+    RCP<SingleLevelFactory> Afact;
+    Level currentLevel;
+    RCP<Matrix> thisLevelA;
+    thisLevelA = currentLevel.Get<Matrix>("A", Afact.get());
+  @endcode
+
+  @todo If Afact is null (actually, Teuchos::null), then the FactoryManager associated with currentLevel will determine whether a default factory has
+  been specified for creating A.  If "yes", then that factory will be called, A will be stored in currentLevel, and an RCP will be returned by
+  the Get call. If "no", then the FactoryManager will <b>throw an exception indicating that it does not know how to generate A</b>.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class FactoryManager : public FactoryManagerBase {
 #undef MUELU_FACTORYMANAGER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructor/Destructors
-    //@{
-
-    //! @brief Constructor.
-    FactoryManager() {
-      SetIgnoreUserData(false); // set IgnorUserData flag to false (default behaviour)
-      useKokkos_ = !Node::is_serial;
-    }
+ public:
+  //! @name Constructor/Destructors
+  //@{
 
-    //! Constructor used by HierarchyFactory (temporary, will be removed)
-    FactoryManager(const std::map<std::string, RCP<const FactoryBase> >& factoryTable) {
-      factoryTable_ = factoryTable;
-      SetIgnoreUserData(false); // set IgnorUserData flag to false (default behaviour) //TODO: use parent class constructor instead
-      useKokkos_ = !Node::is_serial;
-    }
+  //! @brief Constructor.
+  FactoryManager() {
+    SetIgnoreUserData(false);  // set IgnorUserData flag to false (default behaviour)
+    useKokkos_ = !Node::is_serial;
+  }
 
-    //! Destructor.
-    virtual ~FactoryManager() { }
+  //! Constructor used by HierarchyFactory (temporary, will be removed)
+  FactoryManager(const std::map<std::string, RCP<const FactoryBase> >& factoryTable) {
+    factoryTable_ = factoryTable;
+    SetIgnoreUserData(false);  // set IgnorUserData flag to false (default behaviour) //TODO: use parent class constructor instead
+    useKokkos_ = !Node::is_serial;
+  }
 
-    //@}
+  //! Destructor.
+  virtual ~FactoryManager() {}
 
-    //! @name Get/Set functions.
-    //@{
+  //@}
 
-    /*! @brief Set Factory
+  //! @name Get/Set functions.
+  //@{
 
-        Register the factory that should generate data if said factory is not specified in the request.
+  /*! @brief Set Factory
 
-        @param[in] name of variable
-        @param[in] factory that generates the data
-    */
-    void SetFactory(const std::string & varName, const RCP<const FactoryBase>& factory);
+      Register the factory that should generate data if said factory is not specified in the request.
 
-    /*! @brief Get factory associated with a particular data name.
+      @param[in] name of variable
+      @param[in] factory that generates the data
+  */
+  void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
 
-       @param[in] varName name of variable.
+  /*! @brief Get factory associated with a particular data name.
 
-    */
-    const RCP<const FactoryBase> GetFactory(const std::string& varName) const;
+     @param[in] varName name of variable.
 
-    /*! @brief Get factory associated with a particular data name (NONCONST version)
+  */
+  const RCP<const FactoryBase> GetFactory(const std::string& varName) const;
 
-       @param[in] varName name of variable.
+  /*! @brief Get factory associated with a particular data name (NONCONST version)
 
-    */
-    const RCP<FactoryBase> GetFactoryNonConst(const std::string& varName);
+     @param[in] varName name of variable.
 
-    //! Check
-    // Return true if Factory associated with varName is registered
-    bool hasFactory(const std::string& varName) const;
+  */
+  const RCP<FactoryBase> GetFactoryNonConst(const std::string& varName);
 
+  //! Check
+  // Return true if Factory associated with varName is registered
+  bool hasFactory(const std::string& varName) const;
 
-    //!
-    const RCP<const FactoryBase> GetDefaultFactory(const std::string& varName) const;
+  //!
+  const RCP<const FactoryBase> GetDefaultFactory(const std::string& varName) const;
 
-    void SetKokkosRefactor(const bool useKokkos) {
-      useKokkos_ = useKokkos;
-    }
+  void SetKokkosRefactor(const bool useKokkos) {
+    useKokkos_ = useKokkos;
+  }
 
-    bool GetKokkosRefactor() const { return useKokkos_; }
+  bool GetKokkosRefactor() const { return useKokkos_; }
 
-    //@}
+  //@}
 
-    void Clean() const { defaultFactoryTable_.clear(); }
+  void Clean() const { defaultFactoryTable_.clear(); }
 
 #ifdef HAVE_MUELU_DEBUG
-    void ResetDebugData() const;
+  void ResetDebugData() const;
 #endif
 
-    void Print() const;
-
-  private:
+  void Print() const;
 
-    //! @name Helper functions
-    //@{
+ private:
+  //! @name Helper functions
+  //@{
 
-    /*! Add a factory to the default factory list and return it. This helper function is used by GetDefaultFactory()
+  /*! Add a factory to the default factory list and return it. This helper function is used by GetDefaultFactory()
 
-     @todo TODO factory->setObjectLabel("Default " + varName + "Factory");
-    */
+   @todo TODO factory->setObjectLabel("Default " + varName + "Factory");
+  */
 
-    const RCP<const FactoryBase> SetAndReturnDefaultFactory(const std::string& varName, const RCP<const FactoryBase>& factory) const;
-    //@}
+  const RCP<const FactoryBase> SetAndReturnDefaultFactory(const std::string& varName, const RCP<const FactoryBase>& factory) const;
+  //@}
 
-    /*! @brief User-defined factories.
-     *
-     * User may overwrite default behaviour. The user provided factories are stored in a separate table. When we try to determine
-     * which factory generates the data, this table is searched first.
+  /*! @brief User-defined factories.
+   *
+   * User may overwrite default behaviour. The user provided factories are stored in a separate table. When we try to determine
+   * which factory generates the data, this table is searched first.
 
-      Note: we distinguish 'user defined factory' and 'default factory' to allow the deallocation of default factories separately.
-    */
-    std::map<std::string, RCP<const FactoryBase> > factoryTable_;
+    Note: we distinguish 'user defined factory' and 'default factory' to allow the deallocation of default factories separately.
+  */
+  std::map<std::string, RCP<const FactoryBase> > factoryTable_;
 
-    /*! @brief Table that holds default factories.
+  /*! @brief Table that holds default factories.
 
-      -# We distinguish 'user defined factory' and 'default factory' to allow the deallocation of default factories separately.
-      -# <tt>defaultFactoryTable_</tt> is mutable because default factories are only added to the list when they are requested
-      to avoid allocation of unused factories.
-    */
-    mutable
-    std::map<std::string, RCP<const FactoryBase> > defaultFactoryTable_;
+    -# We distinguish 'user defined factory' and 'default factory' to allow the deallocation of default factories separately.
+    -# <tt>defaultFactoryTable_</tt> is mutable because default factories are only added to the list when they are requested
+    to avoid allocation of unused factories.
+  */
+  mutable std::map<std::string, RCP<const FactoryBase> > defaultFactoryTable_;
 
-    //! Whether or not to use kokkos factories.
-    bool useKokkos_;
+  //! Whether or not to use kokkos factories.
+  bool useKokkos_;
 
-  }; // class
+};  // class
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_FACTORYMANAGER_SHORT
-#endif // MUELU_FACTORYMANAGER_DECL_HPP
+#endif  // MUELU_FACTORYMANAGER_DECL_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_FactoryManager_def.hpp b/packages/muelu/src/MueCentral/MueLu_FactoryManager_def.hpp
index f7004c043c69..6da6298d54dc 100644
--- a/packages/muelu/src/MueCentral/MueLu_FactoryManager_def.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_FactoryManager_def.hpp
@@ -85,229 +85,227 @@
 
 #include "MueLu_FactoryManager_decl.hpp"
 
-
 namespace MueLu {
 
-#define MUELU_KOKKOS_FACTORY(varName, oldFactory, newFactory)   \
-  (!useKokkos_) ? SetAndReturnDefaultFactory(varName, rcp(new oldFactory())) : \
-                  SetAndReturnDefaultFactory(varName, rcp(new newFactory()));
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
-    factoryTable_[varName] = factory;
-  }
+#define MUELU_KOKKOS_FACTORY(varName, oldFactory, newFactory) \
+  (!useKokkos_) ? SetAndReturnDefaultFactory(varName, rcp(new oldFactory())) : SetAndReturnDefaultFactory(varName, rcp(new newFactory()));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactory(const std::string& varName) const {
-    if (factoryTable_.count(varName)) {
-      // Search user provided factories
-      return factoryTable_.find(varName)->second;
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
+  factoryTable_[varName] = factory;
+}
 
-    // Search/create default factory for this name
-    return GetDefaultFactory(varName);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactory(const std::string& varName) const {
+  if (factoryTable_.count(varName)) {
+    // Search user provided factories
+    return factoryTable_.find(varName)->second;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const RCP<FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactoryNonConst(const std::string& varName) {
-    return Teuchos::rcp_const_cast<FactoryBase>(GetFactory(varName));
-  }
+  // Search/create default factory for this name
+  return GetDefaultFactory(varName);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const RCP<FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetFactoryNonConst(const std::string& varName) {
+  return Teuchos::rcp_const_cast<FactoryBase>(GetFactory(varName));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasFactory(const std::string& varName) const {
+  if (factoryTable_.count(varName)) return true;
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDefaultFactory(const std::string& varName) const {
+  if (defaultFactoryTable_.count(varName)) {
+    // The factory for this name was already created (possibly, for previous level, if we reuse factory manager)
+    return defaultFactoryTable_.find(varName)->second;
+
+  } else {
+    // No factory was created for this name, but we may know which one to create
+    if (varName == "A") return SetAndReturnDefaultFactory(varName, rcp(new RAPFactory()));
+    if (varName == "Ainv") return SetAndReturnDefaultFactory(varName, rcp(new InverseApproximationFactory()));
+    if (varName == "RAP Pattern") return GetFactory("A");
+    if (varName == "AP Pattern") return GetFactory("A");
+    if (varName == "Ptent") return MUELU_KOKKOS_FACTORY(varName, TentativePFactory, TentativePFactory_kokkos);
+    if (varName == "P") {
+      // GetFactory("Ptent"): we need to use the same factory instance for both "P" and "Nullspace"
+      RCP<Factory> factory;
+      if (useKokkos_)
+        factory = rcp(new SaPFactory_kokkos());
+      else
+        factory = rcp(new SaPFactory());
+      factory->SetFactory("P", GetFactory("Ptent"));
+      return SetAndReturnDefaultFactory(varName, factory);
+    }
+    if (varName == "Nullspace") {
+      // GetFactory("Ptent"): we need to use the same factory instance for both "P" and "Nullspace"
+      RCP<Factory> factory;
+      if (useKokkos_)
+        factory = rcp(new NullspaceFactory_kokkos());
+      else
+        factory = rcp(new NullspaceFactory());
+      factory->SetFactory("Nullspace", GetFactory("Ptent"));
+      return SetAndReturnDefaultFactory(varName, factory);
+    }
+    if (varName == "Scaled Nullspace") return SetAndReturnDefaultFactory(varName, rcp(new ScaledNullspaceFactory()));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasFactory(const std::string& varName) const {
-    if (factoryTable_.count(varName)) return true;
-    return false;
-  }
+    if (varName == "Coordinates") return GetFactory("Ptent");
+    if (varName == "Node Comm") return GetFactory("Ptent");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDefaultFactory(const std::string& varName) const {
-    if (defaultFactoryTable_.count(varName)) {
-      // The factory for this name was already created (possibly, for previous level, if we reuse factory manager)
-      return defaultFactoryTable_.find(varName)->second;
-
-    } else {
-      // No factory was created for this name, but we may know which one to create
-      if (varName == "A")                               return SetAndReturnDefaultFactory(varName, rcp(new RAPFactory()));
-      if (varName == "Ainv")                            return SetAndReturnDefaultFactory(varName, rcp(new InverseApproximationFactory()));
-      if (varName == "RAP Pattern")                     return GetFactory("A");
-      if (varName == "AP Pattern")                      return GetFactory("A");
-      if (varName == "Ptent")                           return MUELU_KOKKOS_FACTORY(varName, TentativePFactory, TentativePFactory_kokkos);
-      if (varName == "P") {
-        // GetFactory("Ptent"): we need to use the same factory instance for both "P" and "Nullspace"
-        RCP<Factory> factory;
-        if (useKokkos_)
-          factory = rcp(new SaPFactory_kokkos());
-        else
-          factory = rcp(new SaPFactory());
-        factory->SetFactory("P", GetFactory("Ptent"));
-        return SetAndReturnDefaultFactory(varName, factory);
-      }
-      if (varName == "Nullspace") {
-        // GetFactory("Ptent"): we need to use the same factory instance for both "P" and "Nullspace"
-        RCP<Factory> factory;
-        if (useKokkos_)
-          factory = rcp(new NullspaceFactory_kokkos());
-        else
-          factory = rcp(new NullspaceFactory());
-        factory->SetFactory("Nullspace", GetFactory("Ptent"));
-        return SetAndReturnDefaultFactory(varName, factory);
-      }
-      if (varName == "Scaled Nullspace")                return SetAndReturnDefaultFactory(varName, rcp(new ScaledNullspaceFactory()));
-
-      if (varName == "Coordinates")                     return GetFactory("Ptent");
-      if (varName == "Node Comm")                       return GetFactory("Ptent");
-
-      if (varName == "R")                               return SetAndReturnDefaultFactory(varName, rcp(new TransPFactory()));
-      if (varName == "RfromPfactory")                   return  GetFactory("P");
+    if (varName == "R") return SetAndReturnDefaultFactory(varName, rcp(new TransPFactory()));
+    if (varName == "RfromPfactory") return GetFactory("P");
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-      if (varName == "Partition")                       return SetAndReturnDefaultFactory(varName, rcp(new ZoltanInterface()));
-#endif //ifdef HAVE_MPI
+    if (varName == "Partition") return SetAndReturnDefaultFactory(varName, rcp(new ZoltanInterface()));
+#endif  // ifdef HAVE_MPI
 
-      if (varName == "Importer") {
+    if (varName == "Importer") {
 #ifdef HAVE_MPI
-                                                        return SetAndReturnDefaultFactory(varName, rcp(new RepartitionFactory()));
+      return SetAndReturnDefaultFactory(varName, rcp(new RepartitionFactory()));
 #else
-                                                        return SetAndReturnDefaultFactory(varName, NoFactory::getRCP());
+      return SetAndReturnDefaultFactory(varName, NoFactory::getRCP());
 #endif
-      }
-      if (varName == "number of partitions") {
+    }
+    if (varName == "number of partitions") {
 #ifdef HAVE_MPI
-                                                        return SetAndReturnDefaultFactory(varName, rcp(new RepartitionHeuristicFactory()));
+      return SetAndReturnDefaultFactory(varName, rcp(new RepartitionHeuristicFactory()));
 #else
-                                                        return SetAndReturnDefaultFactory(varName, NoFactory::getRCP());
-#endif
-      }
-      if (varName == "repartition: heuristic target rows per process") return GetFactory("number of partitions");
-
-      if (varName == "Graph")                           return MUELU_KOKKOS_FACTORY(varName, CoalesceDropFactory, CoalesceDropFactory_kokkos);
-      if (varName == "UnAmalgamationInfo")              return SetAndReturnDefaultFactory(varName, rcp(new AmalgamationFactory()));
-      if (varName == "Aggregates")                      return MUELU_KOKKOS_FACTORY(varName, UncoupledAggregationFactory, UncoupledAggregationFactory_kokkos);
-      if (varName == "AggregateQualities")              return SetAndReturnDefaultFactory(varName, rcp(new AggregateQualityEstimateFactory()));
-      if (varName == "CoarseMap")                       return SetAndReturnDefaultFactory(varName, rcp(new CoarseMapFactory()));
-      if (varName == "DofsPerNode")                     return GetFactory("Graph");
-      if (varName == "Filtering")                       return GetFactory("Graph");
-      if (varName == "BlockNumber")                     return SetAndReturnDefaultFactory(varName, rcp(new InitialBlockNumberFactory()));
-      if (varName == "LineDetection_VertLineIds")       return SetAndReturnDefaultFactory(varName, rcp(new LineDetectionFactory()));
-      if (varName == "LineDetection_Layers")            return GetFactory("LineDetection_VertLineIds");
-      if (varName == "CoarseNumZLayers")                return GetFactory("LineDetection_VertLineIds");
-      
-      // Structured
-      if (varName == "structuredInterpolationOrder")    return SetAndReturnDefaultFactory(varName, rcp(new StructuredAggregationFactory()));
-
-      // Non-Galerkin
-      if (varName == "K")                               return GetFactory("A");
-      if (varName == "M")                               return GetFactory("A");
-      if (varName == "Mdiag")                           return GetFactory("A");
-      if (varName == "cfl-based shift array")           return GetFactory("A");
-
-      // Same factory for both Pre and Post Smoother. Factory for key "Smoother" can be set by users.
-      if (varName == "PreSmoother")                     return GetFactory("Smoother");
-      if (varName == "PostSmoother")                    return GetFactory("Smoother");
-
-      if (varName == "Ppattern") {
-        RCP<PatternFactory> PpFact = rcp(new PatternFactory);
-        PpFact->SetFactory("P", GetFactory("Ptent"));
-        return SetAndReturnDefaultFactory(varName, PpFact);
-      }
-      if (varName == "Constraint")                      return SetAndReturnDefaultFactory(varName, rcp(new ConstraintFactory()));
-
-      if (varName == "Smoother") {
-        Teuchos::ParameterList smootherParamList;
-        smootherParamList.set("relaxation: type",           "Symmetric Gauss-Seidel");
-        smootherParamList.set("relaxation: sweeps",         Teuchos::OrdinalTraits<LO>::one());
-        smootherParamList.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
-        return SetAndReturnDefaultFactory(varName, rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", smootherParamList)))));
-      }
-      if (varName == "CoarseSolver")                    return SetAndReturnDefaultFactory(varName, rcp(new SmootherFactory(rcp(new DirectSolver()), Teuchos::null)));
-
-      if (varName == "DualNodeID2PrimalNodeID")         return SetAndReturnDefaultFactory(varName, rcp(new InterfaceMappingTransferFactory()));
-      if (varName == "CoarseDualNodeID2PrimalNodeID")   return SetAndReturnDefaultFactory(varName, rcp(new InterfaceAggregationFactory()));
-#ifdef HAVE_MUELU_INTREPID2
-      // If we're asking for it, find who made P
-      if (varName == "pcoarsen: element to node map")   return GetFactory("P");
+      return SetAndReturnDefaultFactory(varName, NoFactory::getRCP());
 #endif
-
-      // NOTE: These are user data, but we might want to print them, so they need a default factory
-      if (varName == "Pnodal")                        return NoFactory::getRCP();
-      if (varName == "NodeMatrix")                    return NoFactory::getRCP();
-      if (varName == "NodeAggMatrix")                 return NoFactory::getRCP();
-
-
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::FactoryManager::GetDefaultFactory(): No default factory available for building '" + varName + "'.");
     }
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetAndReturnDefaultFactory(const std::string& varName, const RCP<const FactoryBase>& factory) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(factory.is_null(), Exceptions::RuntimeError, "The default factory for building '" << varName << "' is null");
+    if (varName == "repartition: heuristic target rows per process") return GetFactory("number of partitions");
+
+    if (varName == "Graph") return MUELU_KOKKOS_FACTORY(varName, CoalesceDropFactory, CoalesceDropFactory_kokkos);
+    if (varName == "UnAmalgamationInfo") return SetAndReturnDefaultFactory(varName, rcp(new AmalgamationFactory()));
+    if (varName == "Aggregates") return MUELU_KOKKOS_FACTORY(varName, UncoupledAggregationFactory, UncoupledAggregationFactory_kokkos);
+    if (varName == "AggregateQualities") return SetAndReturnDefaultFactory(varName, rcp(new AggregateQualityEstimateFactory()));
+    if (varName == "CoarseMap") return SetAndReturnDefaultFactory(varName, rcp(new CoarseMapFactory()));
+    if (varName == "DofsPerNode") return GetFactory("Graph");
+    if (varName == "Filtering") return GetFactory("Graph");
+    if (varName == "BlockNumber") return SetAndReturnDefaultFactory(varName, rcp(new InitialBlockNumberFactory()));
+    if (varName == "LineDetection_VertLineIds") return SetAndReturnDefaultFactory(varName, rcp(new LineDetectionFactory()));
+    if (varName == "LineDetection_Layers") return GetFactory("LineDetection_VertLineIds");
+    if (varName == "CoarseNumZLayers") return GetFactory("LineDetection_VertLineIds");
+
+    // Structured
+    if (varName == "structuredInterpolationOrder") return SetAndReturnDefaultFactory(varName, rcp(new StructuredAggregationFactory()));
+
+    // Non-Galerkin
+    if (varName == "K") return GetFactory("A");
+    if (varName == "M") return GetFactory("A");
+    if (varName == "Mdiag") return GetFactory("A");
+    if (varName == "cfl-based shift array") return GetFactory("A");
+
+    // Same factory for both Pre and Post Smoother. Factory for key "Smoother" can be set by users.
+    if (varName == "PreSmoother") return GetFactory("Smoother");
+    if (varName == "PostSmoother") return GetFactory("Smoother");
+
+    if (varName == "Ppattern") {
+      RCP<PatternFactory> PpFact = rcp(new PatternFactory);
+      PpFact->SetFactory("P", GetFactory("Ptent"));
+      return SetAndReturnDefaultFactory(varName, PpFact);
+    }
+    if (varName == "Constraint") return SetAndReturnDefaultFactory(varName, rcp(new ConstraintFactory()));
+
+    if (varName == "Smoother") {
+      Teuchos::ParameterList smootherParamList;
+      smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+      smootherParamList.set("relaxation: sweeps", Teuchos::OrdinalTraits<LO>::one());
+      smootherParamList.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
+      return SetAndReturnDefaultFactory(varName, rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", smootherParamList)))));
+    }
+    if (varName == "CoarseSolver") return SetAndReturnDefaultFactory(varName, rcp(new SmootherFactory(rcp(new DirectSolver()), Teuchos::null)));
 
-    GetOStream(Runtime1) << "Using default factory (" << factory->ShortClassName() <<"["<<factory->GetID()<<"]) for building '" << varName << "'." << std::endl;
+    if (varName == "DualNodeID2PrimalNodeID") return SetAndReturnDefaultFactory(varName, rcp(new InterfaceMappingTransferFactory()));
+    if (varName == "CoarseDualNodeID2PrimalNodeID") return SetAndReturnDefaultFactory(varName, rcp(new InterfaceAggregationFactory()));
+#ifdef HAVE_MUELU_INTREPID2
+    // If we're asking for it, find who made P
+    if (varName == "pcoarsen: element to node map") return GetFactory("P");
+#endif
 
-    defaultFactoryTable_[varName] = factory;
+    // NOTE: These are user data, but we might want to print them, so they need a default factory
+    if (varName == "Pnodal") return NoFactory::getRCP();
+    if (varName == "NodeMatrix") return NoFactory::getRCP();
+    if (varName == "NodeAggMatrix") return NoFactory::getRCP();
 
-    return defaultFactoryTable_[varName];
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::FactoryManager::GetDefaultFactory(): No default factory available for building '" + varName + "'.");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Print() const {
-    std::map<std::string, RCP<const FactoryBase> >::const_iterator it;
-    Teuchos::FancyOStream& fancy = GetOStream(Debug);
-    //auto & fancy = std::cout;// For debugging
-
-
-    fancy << "Users factory table (factoryTable_):" << std::endl;
-    for (it = factoryTable_.begin(); it != factoryTable_.end(); it++) {
-      fancy << "  " << it->first << " -> ";
-      if (it->second.get() == NoFactory::get()) fancy << "NoFactory";
-      else if (!it->second.get()) fancy<< "NULL";
-      else {
-        fancy << it->second.get()->ShortClassName()<<"["<<it->second.get()->GetID()<<"]";
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const RCP<const FactoryBase> FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetAndReturnDefaultFactory(const std::string& varName, const RCP<const FactoryBase>& factory) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(factory.is_null(), Exceptions::RuntimeError, "The default factory for building '" << varName << "' is null");
+
+  GetOStream(Runtime1) << "Using default factory (" << factory->ShortClassName() << "[" << factory->GetID() << "]) for building '" << varName << "'." << std::endl;
+
+  defaultFactoryTable_[varName] = factory;
+
+  return defaultFactoryTable_[varName];
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Print() const {
+  std::map<std::string, RCP<const FactoryBase> >::const_iterator it;
+  Teuchos::FancyOStream& fancy = GetOStream(Debug);
+  // auto & fancy = std::cout;// For debugging
+
+  fancy << "Users factory table (factoryTable_):" << std::endl;
+  for (it = factoryTable_.begin(); it != factoryTable_.end(); it++) {
+    fancy << "  " << it->first << " -> ";
+    if (it->second.get() == NoFactory::get())
+      fancy << "NoFactory";
+    else if (!it->second.get())
+      fancy << "NULL";
+    else {
+      fancy << it->second.get()->ShortClassName() << "[" << it->second.get()->GetID() << "]";
 #ifdef HAVE_MUELU_DEBUG
-        fancy<<"("<<Teuchos::toString(it->second.get()) <<")";
+      fancy << "(" << Teuchos::toString(it->second.get()) << ")";
 #endif
-      }
-      fancy<< std::endl;
     }
+    fancy << std::endl;
+  }
 
-    fancy << "Default factory table (defaultFactoryTable_):" << std::endl;
-    for (it = defaultFactoryTable_.begin(); it != defaultFactoryTable_.end(); it++) {
-      fancy << "  " << it->first << " -> ";
-      if (it->second.get() == NoFactory::get()) fancy << "NoFactory";
-      else if (!it->second.get()) fancy<< "NULL";
-      else {
-        fancy << it->second.get()->ShortClassName()<<"["<<it->second.get()->GetID()<<"]";
+  fancy << "Default factory table (defaultFactoryTable_):" << std::endl;
+  for (it = defaultFactoryTable_.begin(); it != defaultFactoryTable_.end(); it++) {
+    fancy << "  " << it->first << " -> ";
+    if (it->second.get() == NoFactory::get())
+      fancy << "NoFactory";
+    else if (!it->second.get())
+      fancy << "NULL";
+    else {
+      fancy << it->second.get()->ShortClassName() << "[" << it->second.get()->GetID() << "]";
 #ifdef HAVE_MUELU_DEBUG
-        fancy<<"("<<Teuchos::toString(it->second.get()) <<")";
+      fancy << "(" << Teuchos::toString(it->second.get()) << ")";
 #endif
-      }
-      fancy<< std::endl;
     }
-
+    fancy << std::endl;
   }
+}
 
 #ifdef HAVE_MUELU_DEBUG
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ResetDebugData() const {
-    std::map<std::string, RCP<const FactoryBase> >::const_iterator it;
-
-    for (it = factoryTable_.begin(); it != factoryTable_.end(); it++)
-      if (!it->second.is_null())
-        it->second->ResetDebugData();
-
-    for (it = defaultFactoryTable_.begin(); it != defaultFactoryTable_.end(); it++)
-      if (!it->second.is_null())
-        it->second->ResetDebugData();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FactoryManager<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ResetDebugData() const {
+  std::map<std::string, RCP<const FactoryBase> >::const_iterator it;
+
+  for (it = factoryTable_.begin(); it != factoryTable_.end(); it++)
+    if (!it->second.is_null())
+      it->second->ResetDebugData();
+
+  for (it = defaultFactoryTable_.begin(); it != defaultFactoryTable_.end(); it++)
+    if (!it->second.is_null())
+      it->second->ResetDebugData();
+}
 #endif
 
-
 #undef MUELU_KOKKOS_FACTORY
 
-} // namespace MueLu
+}  // namespace MueLu
 
-//TODO: add operator[]
-//TODO: should we use a parameterList instead of a std::map? It might be useful to tag which factory have been used and report unused factory.
-//TODO: add an option 'NoDefault' to check if we are using any default factory.
-//TODO: use Teuchos::ConstNonConstObjectContainer to allow user to modify factories after a GetFactory()
+// TODO: add operator[]
+// TODO: should we use a parameterList instead of a std::map? It might be useful to tag which factory have been used and report unused factory.
+// TODO: add an option 'NoDefault' to check if we are using any default factory.
+// TODO: use Teuchos::ConstNonConstObjectContainer to allow user to modify factories after a GetFactory()
 
-#endif // MUELU_FACTORYMANAGER_DEF_HPP
+#endif  // MUELU_FACTORYMANAGER_DEF_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_decl.hpp b/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_decl.hpp
index 2bfb4b97378b..593f6f69865f 100644
--- a/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_decl.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_decl.hpp
@@ -61,84 +61,76 @@
 
 namespace MueLu {
 
-  //! An exception safe way to call the method 'Level::SetFactoryManager()'
-  class SetFactoryManager {
-
-  public:
-
-    //@{
-
-    /*!
-      @brief Constructor
-
-      Set a given factory manager on a specific level
-    */
-    SetFactoryManager(const RCP<Level> & level, const RCP<const FactoryManagerBase> & factoryManager)
-      : level_(level), prevFactoryManager_(level->GetFactoryManager())
-    {
-      // set new factory manager
-      level->SetFactoryManager(factoryManager);
-    }
-
-    //! Destructor.
-    virtual ~SetFactoryManager() {
-      // restore previous factory manager
-      level_->SetFactoryManager(prevFactoryManager_);
-    }
-
-    //@}
-
-  private:
-    //! needed to save & restore previous factoryManager
-    const RCP<Level> level_;
-    const RCP<const FactoryManagerBase> prevFactoryManager_;
-  };
-
-
-
-
-  template<class Scalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class HierarchyUtils {
+//! An exception safe way to call the method 'Level::SetFactoryManager()'
+class SetFactoryManager {
+ public:
+  //@{
+
+  /*!
+    @brief Constructor
+
+    Set a given factory manager on a specific level
+  */
+  SetFactoryManager(const RCP<Level>& level, const RCP<const FactoryManagerBase>& factoryManager)
+    : level_(level)
+    , prevFactoryManager_(level->GetFactoryManager()) {
+    // set new factory manager
+    level->SetFactoryManager(factoryManager);
+  }
+
+  //! Destructor.
+  virtual ~SetFactoryManager() {
+    // restore previous factory manager
+    level_->SetFactoryManager(prevFactoryManager_);
+  }
+
+  //@}
+
+ private:
+  //! needed to save & restore previous factoryManager
+  const RCP<Level> level_;
+  const RCP<const FactoryManagerBase> prevFactoryManager_;
+};
+
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class HierarchyUtils {
 #undef MUELU_HIERARCHYUTILS_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-    /*!
-    \brief Add non-serializable data to Hierarchy
-
-    Add non-serializable data given level-specific sublist \c nonSerialList to the Hierarchy \c H.
-    Calling \c AddLevel() along the way, if necessary.
-
-    Non-serializable data to be added:
-    - Operator "A"
-    - Prolongator "P"
-    - Restrictor "R"
-    - "M"
-    - "Mdiag"
-    - "K"
-    - Nullspace information "Nullspace"
-    - Coordinate information "Coordinates"
-    - "Node Comm"
-    - Primal-to-dual node mapping "DualNodeID2PrimalNodeID"
-    - "Primal interface DOF map"
-    - "pcoarsen: element to node map
-
-    This routine is used by the CreateXpetraPreconditioner() routine.
-
-    @param HM Hierarhcy manager
-    @param H Hierarchy, where non-serializable data needs to be added
-    @param nonSerialList Parameter list containing non-serializable data
-    */
-    static void AddNonSerializableDataToHierarchy(HierarchyManager& HM, Hierarchy& H, const ParameterList& nonSerialList);
-    static void CopyBetweenHierarchies(Hierarchy& fromHierarchy, Hierarchy& toHierarchy, const std::string fromLabel, const std::string toLabel, const std::string dataType);
-  };
-
-
-
-
-} // namespace MueLu
+ public:
+  /*!
+  \brief Add non-serializable data to Hierarchy
+
+  Add non-serializable data given level-specific sublist \c nonSerialList to the Hierarchy \c H.
+  Calling \c AddLevel() along the way, if necessary.
+
+  Non-serializable data to be added:
+  - Operator "A"
+  - Prolongator "P"
+  - Restrictor "R"
+  - "M"
+  - "Mdiag"
+  - "K"
+  - Nullspace information "Nullspace"
+  - Coordinate information "Coordinates"
+  - "Node Comm"
+  - Primal-to-dual node mapping "DualNodeID2PrimalNodeID"
+  - "Primal interface DOF map"
+  - "pcoarsen: element to node map
+
+  This routine is used by the CreateXpetraPreconditioner() routine.
+
+  @param HM Hierarhcy manager
+  @param H Hierarchy, where non-serializable data needs to be added
+  @param nonSerialList Parameter list containing non-serializable data
+  */
+  static void AddNonSerializableDataToHierarchy(HierarchyManager& HM, Hierarchy& H, const ParameterList& nonSerialList);
+  static void CopyBetweenHierarchies(Hierarchy& fromHierarchy, Hierarchy& toHierarchy, const std::string fromLabel, const std::string toLabel, const std::string dataType);
+};
+
+}  // namespace MueLu
 
 #define MUELU_HIERARCHYUTILS_SHORT
-#endif // MUELU_HIERARCHYUTILS_DECL_HPP
+#endif  // MUELU_HIERARCHYUTILS_DECL_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_def.hpp b/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_def.hpp
index 6581464aa9ad..6c4defcc4116 100644
--- a/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_def.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_HierarchyUtils_def.hpp
@@ -55,379 +55,359 @@
 #include "MueLu_HierarchyManager.hpp"
 #include "MueLu_FactoryManager.hpp"
 
-//TODO/FIXME: DeclareInput(, **this**) cannot be used here
+// TODO/FIXME: DeclareInput(, **this**) cannot be used here
 #ifdef HAVE_MUELU_INTREPID2
 #include "Kokkos_DynRankView.hpp"
 #endif
 
 namespace MueLu {
 
-  // Copy object from one hierarchy to another calling AddNewLevel as appropriate.
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void HierarchyUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CopyBetweenHierarchies(Hierarchy& fromHierarchy, Hierarchy& toHierarchy, const std::string fromLabel, const std::string toLabel, const std::string dataType) {
+// Copy object from one hierarchy to another calling AddNewLevel as appropriate.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void HierarchyUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CopyBetweenHierarchies(Hierarchy& fromHierarchy, Hierarchy& toHierarchy, const std::string fromLabel, const std::string toLabel, const std::string dataType) {
+  // add any necessary levels
+  for (int i = toHierarchy.GetNumLevels(); i < fromHierarchy.GetNumLevels(); i++)
+    toHierarchy.AddNewLevel();
 
-    // add any necessary levels
-    for (int i = toHierarchy.GetNumLevels(); i < fromHierarchy.GetNumLevels(); i++)
-      toHierarchy.AddNewLevel();
-   
-    for (int i = 0; i < fromHierarchy.GetNumLevels(); i++) {
-      RCP<Level> fromLevel = fromHierarchy.GetLevel(i);
-      RCP<Level> toLevel   = toHierarchy.GetLevel(i);
-  
-      TEUCHOS_TEST_FOR_EXCEPTION(dataType != "RCP<Matrix>"  && dataType != "RCP<const Import>"
-                                 , Exceptions::InvalidArgument,
-                                 std::string("MueLu::Utils::CopyBetweenHierarchies: unknown data type(") + dataType + ")");
-      if (fromLevel->IsAvailable(fromLabel)) {
-        if (dataType == "RCP<Matrix>" ) {
-          // Normally, we should only do 
-          //      toLevel->Set(toLabel,fromLevel->Get<RCP<Matrix> >(fromLabel));
-          // The logic below is meant to handle a special case when we 
-          // repartition a processor away, leaving behind a RCP<Operator> on 
-          // on the level instead of an RCP<Matrix>
+  for (int i = 0; i < fromHierarchy.GetNumLevels(); i++) {
+    RCP<Level> fromLevel = fromHierarchy.GetLevel(i);
+    RCP<Level> toLevel   = toHierarchy.GetLevel(i);
 
-          auto tempOp = fromLevel->Get<RCP<Operator> >(fromLabel);
-          auto tempMatrix  = rcp_dynamic_cast<Matrix>(tempOp);
-          if(!tempMatrix.is_null()) toLevel->Set(toLabel,tempMatrix);
-          else                      toLevel->Set(toLabel,tempOp);
-        }
-        if (dataType == "RCP<const Import>") {
-          toLevel->Set(toLabel,fromLevel->Get<RCP<const Import> >(fromLabel));
-        }
+    TEUCHOS_TEST_FOR_EXCEPTION(dataType != "RCP<Matrix>" && dataType != "RCP<const Import>", Exceptions::InvalidArgument,
+                               std::string("MueLu::Utils::CopyBetweenHierarchies: unknown data type(") + dataType + ")");
+    if (fromLevel->IsAvailable(fromLabel)) {
+      if (dataType == "RCP<Matrix>") {
+        // Normally, we should only do
+        //      toLevel->Set(toLabel,fromLevel->Get<RCP<Matrix> >(fromLabel));
+        // The logic below is meant to handle a special case when we
+        // repartition a processor away, leaving behind a RCP<Operator> on
+        // on the level instead of an RCP<Matrix>
+
+        auto tempOp     = fromLevel->Get<RCP<Operator>>(fromLabel);
+        auto tempMatrix = rcp_dynamic_cast<Matrix>(tempOp);
+        if (!tempMatrix.is_null())
+          toLevel->Set(toLabel, tempMatrix);
+        else
+          toLevel->Set(toLabel, tempOp);
+      }
+      if (dataType == "RCP<const Import>") {
+        toLevel->Set(toLabel, fromLevel->Get<RCP<const Import>>(fromLabel));
       }
     }
   }
+}
 
-  // Adds the following non-serializable data (A,P,R,Nullspace,Coordinates) from level-specific sublist nonSerialList,
-  // calling AddNewLevel as appropriate.
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void HierarchyUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddNonSerializableDataToHierarchy(HierarchyManager& HM, Hierarchy& H, const ParameterList& nonSerialList) {
-    typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,
-                                         LocalOrdinal, GlobalOrdinal, Node> realvaluedmultivector_type;
+// Adds the following non-serializable data (A,P,R,Nullspace,Coordinates) from level-specific sublist nonSerialList,
+// calling AddNewLevel as appropriate.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void HierarchyUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddNonSerializableDataToHierarchy(HierarchyManager& HM, Hierarchy& H, const ParameterList& nonSerialList) {
+  typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,
+                                       LocalOrdinal, GlobalOrdinal, Node>
+      realvaluedmultivector_type;
 
-    for (ParameterList::ConstIterator nonSerialEntry = nonSerialList.begin(); nonSerialEntry != nonSerialList.end(); nonSerialEntry++) {
-      const std::string& levelName = nonSerialEntry->first;
-      // Check for match of the form "level X" where X is a positive integer
-      if (nonSerialList.isSublist(levelName) && levelName.find("level ") == 0 && levelName.size() > 6) {
-        int levelID = strtol(levelName.substr(6).c_str(), 0, 0);
-        if (levelID > 0)
-        {
-          // Do enough level adding so we can be sure to add the data to the right place
-          for (int i = H.GetNumLevels(); i <= levelID; i++)
-            H.AddNewLevel();
-        }
-        RCP<Level> level = H.GetLevel(levelID);
+  for (ParameterList::ConstIterator nonSerialEntry = nonSerialList.begin(); nonSerialEntry != nonSerialList.end(); nonSerialEntry++) {
+    const std::string& levelName = nonSerialEntry->first;
+    // Check for match of the form "level X" where X is a positive integer
+    if (nonSerialList.isSublist(levelName) && levelName.find("level ") == 0 && levelName.size() > 6) {
+      int levelID = strtol(levelName.substr(6).c_str(), 0, 0);
+      if (levelID > 0) {
+        // Do enough level adding so we can be sure to add the data to the right place
+        for (int i = H.GetNumLevels(); i <= levelID; i++)
+          H.AddNewLevel();
+      }
+      RCP<Level> level = H.GetLevel(levelID);
 
-        RCP<FactoryManager> M = Teuchos::rcp_dynamic_cast<FactoryManager>(HM.GetFactoryManager(levelID));
-        TEUCHOS_TEST_FOR_EXCEPTION(M.is_null(), Exceptions::InvalidArgument, "MueLu::Utils::AddNonSerializableDataToHierarchy: cannot get FactoryManager");
+      RCP<FactoryManager> M = Teuchos::rcp_dynamic_cast<FactoryManager>(HM.GetFactoryManager(levelID));
+      TEUCHOS_TEST_FOR_EXCEPTION(M.is_null(), Exceptions::InvalidArgument, "MueLu::Utils::AddNonSerializableDataToHierarchy: cannot get FactoryManager");
 
-        // Grab the level sublist & loop over parameters
-        const ParameterList& levelList = nonSerialList.sublist(levelName);
-        for (ParameterList::ConstIterator levelListEntry = levelList.begin(); levelListEntry != levelList.end(); levelListEntry++) {
-          const std::string& name = levelListEntry->first;
-          TEUCHOS_TEST_FOR_EXCEPTION(name != "A" && name != "P" && name != "R" && name != "K"  && name != "M" && name != "Mdiag" &&
-                                     name != "D0" && name != "Dk_1" &&name != "Dk_2" &&
-                                     name != "Mk_one" && name != "Mk_1_one" && name != "M1_beta" && name != "M1_alpha" &&
-                                     name != "invMk_1_invBeta" && name != "invMk_2_invAlpha" &&
-                                     name != "M1" && name != "Ms" && name != "M0inv" &&
-                                     name != "Pnodal" && name != "NodeMatrix" && name != "NodeAggMatrix" &&
-                                     name != "Nullspace" && name != "Coordinates" && name != "pcoarsen: element to node map" &&
-                                     name != "Node Comm" && name != "DualNodeID2PrimalNodeID" && name != "Primal interface DOF map" &&
-                                     !IsParamMuemexVariable(name), Exceptions::InvalidArgument,
-                                     std::string("MueLu::Utils::AddNonSerializableDataToHierarchy: parameter list contains unknown data type(") + name + ")");
+      // Grab the level sublist & loop over parameters
+      const ParameterList& levelList = nonSerialList.sublist(levelName);
+      for (ParameterList::ConstIterator levelListEntry = levelList.begin(); levelListEntry != levelList.end(); levelListEntry++) {
+        const std::string& name = levelListEntry->first;
+        TEUCHOS_TEST_FOR_EXCEPTION(name != "A" && name != "P" && name != "R" && name != "K" && name != "M" && name != "Mdiag" &&
+                                       name != "D0" && name != "Dk_1" && name != "Dk_2" &&
+                                       name != "Mk_one" && name != "Mk_1_one" && name != "M1_beta" && name != "M1_alpha" &&
+                                       name != "invMk_1_invBeta" && name != "invMk_2_invAlpha" &&
+                                       name != "M1" && name != "Ms" && name != "M0inv" &&
+                                       name != "Pnodal" && name != "NodeMatrix" && name != "NodeAggMatrix" &&
+                                       name != "Nullspace" && name != "Coordinates" && name != "pcoarsen: element to node map" &&
+                                       name != "Node Comm" && name != "DualNodeID2PrimalNodeID" && name != "Primal interface DOF map" &&
+                                       !IsParamMuemexVariable(name),
+                                   Exceptions::InvalidArgument,
+                                   std::string("MueLu::Utils::AddNonSerializableDataToHierarchy: parameter list contains unknown data type(") + name + ")");
 
-          // Get a valid communicator and lib
-          RCP<const Teuchos::Comm<int> > comm;
-          if (!level->GetComm().is_null())
-            comm = level->GetComm();
-          else if (level->IsAvailable("A")) {
+        // Get a valid communicator and lib
+        RCP<const Teuchos::Comm<int>> comm;
+        if (!level->GetComm().is_null())
+          comm = level->GetComm();
+        else if (level->IsAvailable("A")) {
+          RCP<Matrix> mat;
+          level->Get("A", mat);
+          comm = mat->getMap()->getComm();
+        } else {
+          RCP<Level> level0 = H.GetLevel(0);
+          if (!level0->GetComm().is_null())
+            comm = level0->GetComm();
+          else {
             RCP<Matrix> mat;
-            level->Get("A", mat);
+            level0->Get("A", mat);
             comm = mat->getMap()->getComm();
-          } else {
-            RCP<Level> level0 = H.GetLevel(0);
-            if (!level0->GetComm().is_null())
-              comm = level0->GetComm();
-            else {
-              RCP<Matrix> mat;
-              level0->Get("A", mat);
-              comm = mat->getMap()->getComm();
-            }
           }
-          Xpetra::UnderlyingLib lib = level->lib();
+        }
+        Xpetra::UnderlyingLib lib = level->lib();
+
+        if (name == "A") {
+          RCP<Matrix> mat;
+          if (levelListEntry->second.isType<std::string>())
+            // We might also want to read maps here.
+            mat = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(Teuchos::getValue<std::string>(levelListEntry->second), lib, comm);
+          else
+            mat = Teuchos::getValue<RCP<Matrix>>(levelListEntry->second);
+          level->Set(name, mat, NoFactory::get());
+          M->SetFactory(name, NoFactory::getRCP());  // TAW: not sure about this: be aware that this affects all levels
+                                                     //      However, A is accessible through NoFactory anyway, so it should
+                                                     //      be fine here.
+        } else if (name == "P" || name == "R" || name == "K" || name == "M") {
+          if (levelListEntry->second.isType<RCP<Operator>>()) {
+            RCP<Operator> mat;
+            mat = Teuchos::getValue<RCP<Operator>>(levelListEntry->second);
+
+            RCP<const FactoryBase> fact = M->GetFactory(name);
+            level->AddKeepFlag(name, fact.get(), MueLu::UserData);
+            level->Set(name, mat, fact.get());
 
-          if (name == "A") {
+            level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+            level->Set(name, mat, NoFactory::get());
+          } else {
             RCP<Matrix> mat;
             if (levelListEntry->second.isType<std::string>())
               // We might also want to read maps here.
-              mat = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(Teuchos::getValue<std::string>(levelListEntry->second), lib, comm);
+              mat = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(Teuchos::getValue<std::string>(levelListEntry->second), lib, comm);
             else
-              mat = Teuchos::getValue<RCP<Matrix > > (levelListEntry->second);
-            level->Set(name, mat, NoFactory::get());
-            M->SetFactory(name, NoFactory::getRCP()); // TAW: not sure about this: be aware that this affects all levels
-                                                      //      However, A is accessible through NoFactory anyway, so it should
-                                                      //      be fine here.
-          }
-          else if(name == "P" || name == "R" || name == "K" || name == "M" ) {
-            if (levelListEntry->second.isType<RCP<Operator> >()) {
-              RCP<Operator> mat;
-              mat = Teuchos::getValue<RCP<Operator> > (levelListEntry->second);
-
-              RCP<const FactoryBase> fact = M->GetFactory(name);
-              level->AddKeepFlag(name,fact.get(),MueLu::UserData);
-              level->Set(name, mat, fact.get());
-
-              level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-              level->Set(name, mat, NoFactory::get());
-            } else {
-              RCP<Matrix> mat;
-              if (levelListEntry->second.isType<std::string>())
-                // We might also want to read maps here.
-                mat = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(Teuchos::getValue<std::string>(levelListEntry->second), lib, comm);
-              else
-                mat = Teuchos::getValue<RCP<Matrix > > (levelListEntry->second);
+              mat = Teuchos::getValue<RCP<Matrix>>(levelListEntry->second);
 
-              RCP<const FactoryBase> fact = M->GetFactory(name);
-              level->AddKeepFlag(name,fact.get(),MueLu::UserData);
-              level->Set(name, mat, fact.get());
+            RCP<const FactoryBase> fact = M->GetFactory(name);
+            level->AddKeepFlag(name, fact.get(), MueLu::UserData);
+            level->Set(name, mat, fact.get());
 
-              level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-              level->Set(name, mat, NoFactory::get());
-            }
+            level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+            level->Set(name, mat, NoFactory::get());
           }
-          else if (name == "D0" || name == "Dk_1" ||name == "Dk_2" ||
+        } else if (name == "D0" || name == "Dk_1" || name == "Dk_2" ||
                    name == "Mk_one" || name == "Mk_1_one" || name == "M1_beta" || name == "M1_alpha" ||
                    name == "invMk_1_invBeta" || name == "invMk_2_invAlpha" ||
                    name == "M1" || name == "Ms" || name == "M0inv" ||
                    name == "Pnodal" || name == "NodeMatrix" || name == "NodeAggMatrix") {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            if (levelListEntry->second.isType<RCP<Operator> >())
-              level->Set(name, Teuchos::getValue<RCP<Operator> >   (levelListEntry->second), NoFactory::get());
-            else
-              level->Set(name, Teuchos::getValue<RCP<Matrix> >     (levelListEntry->second), NoFactory::get());
-          }
-          else if (name == "Mdiag")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<Vector > >     (levelListEntry->second), NoFactory::get());
-          }
-          else if (name == "Nullspace")
-          {
-            RCP<MultiVector> vec;
-            if (levelListEntry->second.isType<std::string>()) {
-              TEUCHOS_ASSERT(level->IsAvailable("A"));
-              RCP<Matrix> mat;
-              level->Get("A", mat);
-              auto map = mat->getMap();
-              vec = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::ReadMultiVector(Teuchos::getValue<std::string>(levelListEntry->second), map);
-            } else
-              vec = Teuchos::getValue<RCP<MultiVector> > (levelListEntry->second);
-            level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
-            level->Set(name, vec, NoFactory::get());
-            //M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
-                                                        // One should do this only in very special cases
-          }
-          else if(name == "Coordinates") //Scalar of Coordinates MV is always double
-          {
-            RCP<realvaluedmultivector_type> vec;
-            if (levelListEntry->second.isType<std::string>()) {
-              TEUCHOS_ASSERT(level->IsAvailable("A"));
-              RCP<Matrix> mat;
-              level->Get("A", mat);
-              size_t blkSize = mat->GetFixedBlockSize();
-              RCP<const Map> nodeMap = mat->getRowMap();
-              if (blkSize > 1) {
-                // Create a nodal map, as coordinates have not been expanded to a DOF map yet.
-                RCP<const Map> dofMap       = mat->getRowMap();
-                GO             indexBase    = dofMap->getIndexBase();
-                size_t         numLocalDOFs = dofMap->getLocalNumElements();
-                TEUCHOS_TEST_FOR_EXCEPTION(numLocalDOFs % blkSize, Exceptions::RuntimeError,
-                                           "HierarchyUtils: block size (" << blkSize << ") is incompatible with the number of local dofs in a row map (" << numLocalDOFs);
-                ArrayView<const GO> GIDs = dofMap->getLocalElementList();
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          if (levelListEntry->second.isType<RCP<Operator>>())
+            level->Set(name, Teuchos::getValue<RCP<Operator>>(levelListEntry->second), NoFactory::get());
+          else
+            level->Set(name, Teuchos::getValue<RCP<Matrix>>(levelListEntry->second), NoFactory::get());
+        } else if (name == "Mdiag") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<Vector>>(levelListEntry->second), NoFactory::get());
+        } else if (name == "Nullspace") {
+          RCP<MultiVector> vec;
+          if (levelListEntry->second.isType<std::string>()) {
+            TEUCHOS_ASSERT(level->IsAvailable("A"));
+            RCP<Matrix> mat;
+            level->Get("A", mat);
+            auto map = mat->getMap();
+            vec      = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(Teuchos::getValue<std::string>(levelListEntry->second), map);
+          } else
+            vec = Teuchos::getValue<RCP<MultiVector>>(levelListEntry->second);
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, vec, NoFactory::get());
+          // M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
+          //  One should do this only in very special cases
+        } else if (name == "Coordinates")  // Scalar of Coordinates MV is always double
+        {
+          RCP<realvaluedmultivector_type> vec;
+          if (levelListEntry->second.isType<std::string>()) {
+            TEUCHOS_ASSERT(level->IsAvailable("A"));
+            RCP<Matrix> mat;
+            level->Get("A", mat);
+            size_t blkSize         = mat->GetFixedBlockSize();
+            RCP<const Map> nodeMap = mat->getRowMap();
+            if (blkSize > 1) {
+              // Create a nodal map, as coordinates have not been expanded to a DOF map yet.
+              RCP<const Map> dofMap = mat->getRowMap();
+              GO indexBase          = dofMap->getIndexBase();
+              size_t numLocalDOFs   = dofMap->getLocalNumElements();
+              TEUCHOS_TEST_FOR_EXCEPTION(numLocalDOFs % blkSize, Exceptions::RuntimeError,
+                                         "HierarchyUtils: block size (" << blkSize << ") is incompatible with the number of local dofs in a row map (" << numLocalDOFs);
+              ArrayView<const GO> GIDs = dofMap->getLocalElementList();
 
-                Array<GO> nodeGIDs(numLocalDOFs/blkSize);
-                for (size_t i = 0; i < numLocalDOFs; i += blkSize)
-                  nodeGIDs[i/blkSize] = (GIDs[i] - indexBase)/blkSize + indexBase;
+              Array<GO> nodeGIDs(numLocalDOFs / blkSize);
+              for (size_t i = 0; i < numLocalDOFs; i += blkSize)
+                nodeGIDs[i / blkSize] = (GIDs[i] - indexBase) / blkSize + indexBase;
 
-                Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-                nodeMap = MapFactory::Build(dofMap->lib(), INVALID, nodeGIDs(), indexBase, dofMap->getComm());
-              }
-              vec = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node>::ReadMultiVector(Teuchos::getValue<std::string>(levelListEntry->second), nodeMap);
-            } else
-              vec = Teuchos::getValue<RCP<realvaluedmultivector_type > > (levelListEntry->second);
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, vec, NoFactory::get());
-            //M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
-          }
-          else if(name == "Node Comm")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<const Teuchos::Comm<int> > >(levelListEntry->second), NoFactory::get());
-          }
-          else if(name == "DualNodeID2PrimalNodeID")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<std::map<LO, LO>>>(levelListEntry->second), NoFactory::get());
-          }
-          else if(name == "Primal interface DOF map")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<const Map>>(levelListEntry->second), NoFactory::get());
-          }
+              Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+              nodeMap                       = MapFactory::Build(dofMap->lib(), INVALID, nodeGIDs(), indexBase, dofMap->getComm());
+            }
+            vec = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(Teuchos::getValue<std::string>(levelListEntry->second), nodeMap);
+          } else
+            vec = Teuchos::getValue<RCP<realvaluedmultivector_type>>(levelListEntry->second);
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, vec, NoFactory::get());
+          // M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
+        } else if (name == "Node Comm") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<const Teuchos::Comm<int>>>(levelListEntry->second), NoFactory::get());
+        } else if (name == "DualNodeID2PrimalNodeID") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<std::map<LO, LO>>>(levelListEntry->second), NoFactory::get());
+        } else if (name == "Primal interface DOF map") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<const Map>>(levelListEntry->second), NoFactory::get());
+        }
 #ifdef HAVE_MUELU_INTREPID2
-          else if (name == "pcoarsen: element to node map")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> > >(levelListEntry->second), NoFactory::get());
-          }
+        else if (name == "pcoarsen: element to node map") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<Kokkos::DynRankView<LocalOrdinal, typename Node::device_type>>>(levelListEntry->second), NoFactory::get());
+        }
 #endif
-          else
+        else
 #ifdef HAVE_MUELU_MATLAB
-          {
-            //Custom variable for Muemex
-            size_t typeNameStart = name.find_first_not_of(' ');
-            size_t typeNameEnd = name.find(' ', typeNameStart);
-            std::string typeName = name.substr(typeNameStart, typeNameEnd - typeNameStart);
-            std::transform(typeName.begin(), typeName.end(), typeName.begin(), ::tolower);
-            level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
-            if(typeName == "matrix")
-              level->Set(name, Teuchos::getValue<RCP<Matrix> >(levelListEntry->second), NoFactory::get());
-            else if(typeName == "multivector")
-              level->Set(name, Teuchos::getValue<RCP<MultiVector> >(levelListEntry->second), NoFactory::get());
-            else if(typeName == "map")
-              level->Set(name, Teuchos::getValue<RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > >(levelListEntry->second), NoFactory::get());
-            else if(typeName == "ordinalvector")
-              level->Set(name, Teuchos::getValue<RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >(levelListEntry->second), NoFactory::get());
-            else if(typeName == "scalar")
-              level->Set(name, Teuchos::getValue<Scalar>(levelListEntry->second), NoFactory::get());
-            else if(typeName == "double")
-              level->Set(name, Teuchos::getValue<double>(levelListEntry->second), NoFactory::get());
-            else if(typeName == "complex")
-              level->Set(name, Teuchos::getValue<std::complex<double> >(levelListEntry->second), NoFactory::get());
-            else if(typeName == "int")
-              level->Set(name, Teuchos::getValue<int>(levelListEntry->second), NoFactory::get());
-            else if(typeName == "string")
-              level->Set(name, Teuchos::getValue<std::string>(levelListEntry->second), NoFactory::get());
-          }
+        {
+          // Custom variable for Muemex
+          size_t typeNameStart = name.find_first_not_of(' ');
+          size_t typeNameEnd   = name.find(' ', typeNameStart);
+          std::string typeName = name.substr(typeNameStart, typeNameEnd - typeNameStart);
+          std::transform(typeName.begin(), typeName.end(), typeName.begin(), ::tolower);
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          if (typeName == "matrix")
+            level->Set(name, Teuchos::getValue<RCP<Matrix>>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "multivector")
+            level->Set(name, Teuchos::getValue<RCP<MultiVector>>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "map")
+            level->Set(name, Teuchos::getValue<RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "ordinalvector")
+            level->Set(name, Teuchos::getValue<RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "scalar")
+            level->Set(name, Teuchos::getValue<Scalar>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "double")
+            level->Set(name, Teuchos::getValue<double>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "complex")
+            level->Set(name, Teuchos::getValue<std::complex<double>>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "int")
+            level->Set(name, Teuchos::getValue<int>(levelListEntry->second), NoFactory::get());
+          else if (typeName == "string")
+            level->Set(name, Teuchos::getValue<std::string>(levelListEntry->second), NoFactory::get());
+        }
 #else
-          {
-            throw std::runtime_error("Invalid non-serializable data on list");
-          }
-#endif
+        {
+          throw std::runtime_error("Invalid non-serializable data on list");
         }
-      } else if (nonSerialList.isSublist(levelName) && levelName.find("user data") != std::string::npos) {
-        // So far only put data on level 0
-        int levelID = 0;
-        RCP<Level> level = H.GetLevel(levelID);
+#endif
+      }
+    } else if (nonSerialList.isSublist(levelName) && levelName.find("user data") != std::string::npos) {
+      // So far only put data on level 0
+      int levelID      = 0;
+      RCP<Level> level = H.GetLevel(levelID);
 
-        RCP<FactoryManager> M = Teuchos::rcp_dynamic_cast<FactoryManager>(HM.GetFactoryManager(levelID));
-        TEUCHOS_TEST_FOR_EXCEPTION(M.is_null(), Exceptions::InvalidArgument, "MueLu::Utils::AddNonSerializableDataToHierarchy: cannot get FactoryManager");
+      RCP<FactoryManager> M = Teuchos::rcp_dynamic_cast<FactoryManager>(HM.GetFactoryManager(levelID));
+      TEUCHOS_TEST_FOR_EXCEPTION(M.is_null(), Exceptions::InvalidArgument, "MueLu::Utils::AddNonSerializableDataToHierarchy: cannot get FactoryManager");
 
-        // Grab the user data sublist & loop over parameters
-        const ParameterList& userList = nonSerialList.sublist(levelName);
-        for (ParameterList::ConstIterator userListEntry = userList.begin(); userListEntry != userList.end(); userListEntry++) {
-          const std::string& name = userListEntry->first;
-          TEUCHOS_TEST_FOR_EXCEPTION(name != "P" && name != "R"  && name != "K"  && name != "M" && name != "Mdiag" &&
-                                     name != "D0" && name != "Dk_1" &&name != "Dk_2" &&
-                                     name != "Mk_one" && name != "Mk_1_one" && name != "M1_beta" && name != "M1_alpha" &&
-                                     name != "invMk_1_invBeta" && name != "invMk_2_invAlpha" &&
-                                     name != "M1" && name != "Ms" && name != "M0inv" &&
-                                     name != "NodeMatrix" &&
-                                     name != "Nullspace" && name != "Coordinates" && name != "pcoarsen: element to node map" &&
-                                     name != "Node Comm" && name != "DualNodeID2PrimalNodeID" && name != "Primal interface DOF map" &&
-                                     name != "output stream" &&
-                                     !IsParamValidVariable(name), Exceptions::InvalidArgument,
-                                     std::string("MueLu::Utils::AddNonSerializableDataToHierarchy: user data parameter list contains unknown data type (") + name + ")");
-          if( name == "P" || name == "R" || name == "K" || name == "M" ||
-              name == "D0" || name == "Dk_1" ||name == "Dk_2" ||
-              name == "Mk_one" || name == "Mk_1_one" || name == "M1_beta" || name == "M1_alpha" ||
-              name == "invMk_1_invBeta" || name == "invMk_2_invAlpha" ||
-              name == "M1" || name == "Ms" || name == "M0inv" ||
-              name == "NodeMatrix" ) {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<Matrix > >     (userListEntry->second), NoFactory::get());
-          } else if (name == "Mdiag") {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<Vector > >(userListEntry->second), NoFactory::get());
-          } else if (name == "Nullspace") {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<MultiVector > >(userListEntry->second), NoFactory::get());
-            //M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
-                                                        // One should do this only in very special cases
-          } else if(name == "Coordinates") {//Scalar of Coordinates MV is always double
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<realvaluedmultivector_type> >(userListEntry->second), NoFactory::get());
-          }
-          else if(name == "Node Comm") {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<const Teuchos::Comm<int> > >(userListEntry->second), NoFactory::get());
-          }
-          else if(name == "DualNodeID2PrimalNodeID")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<std::map<LO, LO>>>(userListEntry->second), NoFactory::get());
-          }
-          else if(name == "Primal interface DOF map")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<const Map>>(userListEntry->second), NoFactory::get());
-          }
+      // Grab the user data sublist & loop over parameters
+      const ParameterList& userList = nonSerialList.sublist(levelName);
+      for (ParameterList::ConstIterator userListEntry = userList.begin(); userListEntry != userList.end(); userListEntry++) {
+        const std::string& name = userListEntry->first;
+        TEUCHOS_TEST_FOR_EXCEPTION(name != "P" && name != "R" && name != "K" && name != "M" && name != "Mdiag" &&
+                                       name != "D0" && name != "Dk_1" && name != "Dk_2" &&
+                                       name != "Mk_one" && name != "Mk_1_one" && name != "M1_beta" && name != "M1_alpha" &&
+                                       name != "invMk_1_invBeta" && name != "invMk_2_invAlpha" &&
+                                       name != "M1" && name != "Ms" && name != "M0inv" &&
+                                       name != "NodeMatrix" &&
+                                       name != "Nullspace" && name != "Coordinates" && name != "pcoarsen: element to node map" &&
+                                       name != "Node Comm" && name != "DualNodeID2PrimalNodeID" && name != "Primal interface DOF map" &&
+                                       name != "output stream" &&
+                                       !IsParamValidVariable(name),
+                                   Exceptions::InvalidArgument,
+                                   std::string("MueLu::Utils::AddNonSerializableDataToHierarchy: user data parameter list contains unknown data type (") + name + ")");
+        if (name == "P" || name == "R" || name == "K" || name == "M" ||
+            name == "D0" || name == "Dk_1" || name == "Dk_2" ||
+            name == "Mk_one" || name == "Mk_1_one" || name == "M1_beta" || name == "M1_alpha" ||
+            name == "invMk_1_invBeta" || name == "invMk_2_invAlpha" ||
+            name == "M1" || name == "Ms" || name == "M0inv" ||
+            name == "NodeMatrix") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<Matrix>>(userListEntry->second), NoFactory::get());
+        } else if (name == "Mdiag") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<Vector>>(userListEntry->second), NoFactory::get());
+        } else if (name == "Nullspace") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<MultiVector>>(userListEntry->second), NoFactory::get());
+          // M->SetFactory(name, NoFactory::getRCP()); // TAW: generally it is a bad idea to overwrite the factory manager data here
+          //  One should do this only in very special cases
+        } else if (name == "Coordinates") {  // Scalar of Coordinates MV is always double
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<realvaluedmultivector_type>>(userListEntry->second), NoFactory::get());
+        } else if (name == "Node Comm") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<const Teuchos::Comm<int>>>(userListEntry->second), NoFactory::get());
+        } else if (name == "DualNodeID2PrimalNodeID") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<std::map<LO, LO>>>(userListEntry->second), NoFactory::get());
+        } else if (name == "Primal interface DOF map") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<const Map>>(userListEntry->second), NoFactory::get());
+        }
 #ifdef HAVE_MUELU_INTREPID2
-          else if (name == "pcoarsen: element to node map")
-          {
-            level->AddKeepFlag(name,NoFactory::get(),MueLu::UserData);
-            level->Set(name, Teuchos::getValue<RCP<Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> > >(userListEntry->second), NoFactory::get());
-          }
+        else if (name == "pcoarsen: element to node map") {
+          level->AddKeepFlag(name, NoFactory::get(), MueLu::UserData);
+          level->Set(name, Teuchos::getValue<RCP<Kokkos::DynRankView<LocalOrdinal, typename Node::device_type>>>(userListEntry->second), NoFactory::get());
+        }
 #endif
-          else if (name == "output stream")
-          {
-            H.SetMueLuOStream(Teuchos::getValue<RCP<Teuchos::FancyOStream> >(userListEntry->second));
-          }
-          else {
-            //Custom variable
-            size_t typeNameStart = name.find_first_not_of(' ');
-            size_t typeNameEnd   = name.find(' ', typeNameStart);
-            std::string typeName = name.substr(typeNameStart, typeNameEnd - typeNameStart);
-            size_t varNameStart  = name.find_first_not_of(' ', typeNameEnd);
-            std::string varName  = name.substr(varNameStart, name.size());
-            std::transform(typeName.begin(), typeName.end(), typeName.begin(), ::tolower);
-            level->AddKeepFlag(varName, NoFactory::get(), MueLu::UserData);
-            if(typeName == "matrix")
-              level->Set(varName, Teuchos::getValue<RCP<Matrix> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "multivector")
-              level->Set(varName, Teuchos::getValue<RCP<MultiVector> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "vector")
-              level->Set(varName, Teuchos::getValue<RCP<Vector> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "map")
-              level->Set(varName, Teuchos::getValue<RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > >(userListEntry->second), NoFactory::get());
-            else if(typeName == "ordinalvector")
-              level->Set(varName, Teuchos::getValue<RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >(userListEntry->second), NoFactory::get());
-            else if(typeName == "scalar")
-              level->Set(varName, Teuchos::getValue<Scalar>(userListEntry->second), NoFactory::get());
-            else if(typeName == "double")
-              level->Set(varName, Teuchos::getValue<double>(userListEntry->second), NoFactory::get());
-            else if(typeName == "complex")
-              level->Set(varName, Teuchos::getValue<std::complex<double> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "int")
-              level->Set(varName, Teuchos::getValue<int>(userListEntry->second), NoFactory::get());
-            else if(typeName == "string")
-              level->Set(varName, Teuchos::getValue<std::string>(userListEntry->second), NoFactory::get());
-            else if(typeName == "array<go>")
-              level->Set(varName, Teuchos::getValue<Array<GlobalOrdinal> > (userListEntry->second), NoFactory::get());
-            else if(typeName == "array<lo>")
-              level->Set(varName, Teuchos::getValue<Array<LocalOrdinal> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "arrayrcp<lo>")
-              level->Set(varName, Teuchos::getValue<ArrayRCP<LocalOrdinal> >(userListEntry->second), NoFactory::get());
-            else if(typeName == "arrayrcp<go>")
-              level->Set(varName, Teuchos::getValue<ArrayRCP<GlobalOrdinal> >(userListEntry->second), NoFactory::get());
-            else
-              throw std::runtime_error("Invalid non-serializable data on list");
-          }
+        else if (name == "output stream") {
+          H.SetMueLuOStream(Teuchos::getValue<RCP<Teuchos::FancyOStream>>(userListEntry->second));
+        } else {
+          // Custom variable
+          size_t typeNameStart = name.find_first_not_of(' ');
+          size_t typeNameEnd   = name.find(' ', typeNameStart);
+          std::string typeName = name.substr(typeNameStart, typeNameEnd - typeNameStart);
+          size_t varNameStart  = name.find_first_not_of(' ', typeNameEnd);
+          std::string varName  = name.substr(varNameStart, name.size());
+          std::transform(typeName.begin(), typeName.end(), typeName.begin(), ::tolower);
+          level->AddKeepFlag(varName, NoFactory::get(), MueLu::UserData);
+          if (typeName == "matrix")
+            level->Set(varName, Teuchos::getValue<RCP<Matrix>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "multivector")
+            level->Set(varName, Teuchos::getValue<RCP<MultiVector>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "vector")
+            level->Set(varName, Teuchos::getValue<RCP<Vector>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "map")
+            level->Set(varName, Teuchos::getValue<RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "ordinalvector")
+            level->Set(varName, Teuchos::getValue<RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "scalar")
+            level->Set(varName, Teuchos::getValue<Scalar>(userListEntry->second), NoFactory::get());
+          else if (typeName == "double")
+            level->Set(varName, Teuchos::getValue<double>(userListEntry->second), NoFactory::get());
+          else if (typeName == "complex")
+            level->Set(varName, Teuchos::getValue<std::complex<double>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "int")
+            level->Set(varName, Teuchos::getValue<int>(userListEntry->second), NoFactory::get());
+          else if (typeName == "string")
+            level->Set(varName, Teuchos::getValue<std::string>(userListEntry->second), NoFactory::get());
+          else if (typeName == "array<go>")
+            level->Set(varName, Teuchos::getValue<Array<GlobalOrdinal>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "array<lo>")
+            level->Set(varName, Teuchos::getValue<Array<LocalOrdinal>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "arrayrcp<lo>")
+            level->Set(varName, Teuchos::getValue<ArrayRCP<LocalOrdinal>>(userListEntry->second), NoFactory::get());
+          else if (typeName == "arrayrcp<go>")
+            level->Set(varName, Teuchos::getValue<ArrayRCP<GlobalOrdinal>>(userListEntry->second), NoFactory::get());
+          else
+            throw std::runtime_error("Invalid non-serializable data on list");
         }
-        // level->print(std::cout, MueLu::Debug);
       }
+      // level->print(std::cout, MueLu::Debug);
     }
   }
-} // namespace MueLu
+}
+}  // namespace MueLu
 
 #define MUELU_HIERARCHY_UTILS_SHORT
-#endif // MUELU_HIERARCHYHELPERS_DEF_HPP
+#endif  // MUELU_HIERARCHYHELPERS_DEF_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Hierarchy_decl.hpp b/packages/muelu/src/MueCentral/MueLu_Hierarchy_decl.hpp
index 4c345a4e3c49..22b0f8f41843 100644
--- a/packages/muelu/src/MueCentral/MueLu_Hierarchy_decl.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_Hierarchy_decl.hpp
@@ -63,7 +63,7 @@
 #include "MueLu_Types.hpp"
 
 #include "MueLu_FactoryBase_fwd.hpp"
-#include "MueLu_FactoryManager.hpp" // no fwd declaration because constructor of FactoryManager is used as a default parameter of Setup()
+#include "MueLu_FactoryManager.hpp"  // no fwd declaration because constructor of FactoryManager is used as a default parameter of Setup()
 #include "MueLu_KeepType.hpp"
 #include "MueLu_Level_fwd.hpp"
 #include "MueLu_MasterList.hpp"
@@ -76,372 +76,377 @@
 
 namespace MueLu {
 
-  enum class ConvergenceStatus {
-    Converged,
-    Unconverged,
-    Undefined
-  };
-
-  /*!
-    @class Hierarchy
-    @brief Provides methods to build a multigrid hierarchy and apply multigrid cycles.
-
-    Allows users to manually populate operators at different levels within
-    a multigrid method and push them into the hierarchy via SetLevel()
-    and/or to supply factories for automatically generating prolongators,
-    restrictors, and coarse level discretizations.  Additionally, this class contains
-    an apply method that supports V and W cycles.
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Hierarchy : public BaseClass {
+enum class ConvergenceStatus {
+  Converged,
+  Unconverged,
+  Undefined
+};
+
+/*!
+  @class Hierarchy
+  @brief Provides methods to build a multigrid hierarchy and apply multigrid cycles.
+
+  Allows users to manually populate operators at different levels within
+  a multigrid method and push them into the hierarchy via SetLevel()
+  and/or to supply factories for automatically generating prolongators,
+  restrictors, and coarse level discretizations.  Additionally, this class contains
+  an apply method that supports V and W cycles.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Hierarchy : public BaseClass {
 #undef MUELU_HIERARCHY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType MagnitudeType;
-
-    //! Data struct for defining stopping criteria of multigrid iteration
-    struct ConvData {
-      ConvData()                              : maxIts_(1),       tol_(-STS::magnitude(STS::one())) { }
-      ConvData(LO maxIts)                     : maxIts_(maxIts),  tol_(-STS::magnitude(STS::one())) { }
-      ConvData(MagnitudeType tol)             : maxIts_(10000),   tol_(tol) { }
-      ConvData(std::pair<LO,MagnitudeType> p) : maxIts_(p.first), tol_(p.second) { }
-
-      LO            maxIts_;
-      MagnitudeType tol_;
-    };
-
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
-
-    //! Default constructor.
-    Hierarchy();
-    //! Constructor that labels the hierarchy.
-    Hierarchy(const std::string& label);
-
-    //! Constructor
-    Hierarchy(const RCP<Matrix> & A);
-
-    //! Constructor
-    Hierarchy(const RCP<Matrix> & A, const std::string& label);
-
-    //! Destructor.
-    virtual ~Hierarchy() { }
-
-    //@}
-
-    //! @name Set/Get Methods.
-    //@{
-
-    //!
-    static CycleType             GetDefaultCycle()                                     { return MasterList::getDefault<std::string>("cycle type") == "V" ? VCYCLE : WCYCLE; }
-    static int                   GetDefaultCycleStartLevel()                           { return MasterList::getDefault<int>("W cycle start level"); }
-    static bool                  GetDefaultImplicitTranspose()                         { return MasterList::getDefault<bool>("transpose: use implicit");  }
-    static bool                  GetDefaultFuseProlongationAndUpdate()                 { return MasterList::getDefault<bool>("fuse prolongation and update"); }
-    static Xpetra::global_size_t GetDefaultMaxCoarseSize()                             { return MasterList::getDefault<int>("coarse: max size");   }
-    static int                   GetDefaultMaxLevels()                                 { return MasterList::getDefault<int>("max levels");     }
-    static bool                  GetDefaultPRrebalance()                               { return MasterList::getDefault<bool>("repartition: rebalance P and R");   }
-
-    Xpetra::global_size_t        GetMaxCoarseSize() const                              { return maxCoarseSize_; }
-    bool                         GetImplicitTranspose() const                          { return implicitTranspose_; }
-    bool                         GetFuseProlongationAndUpdate() const                  { return fuseProlongationAndUpdate_; }
-
-    void                         SetMaxCoarseSize(Xpetra::global_size_t maxCoarseSize) { maxCoarseSize_ = maxCoarseSize; }
-    void                         SetPRrebalance(bool doPRrebalance)                    { doPRrebalance_ = doPRrebalance; }
-    void                         SetPRViaCopyrebalance(bool doPRViaCopyrebalance)      { doPRViaCopyrebalance_ = doPRViaCopyrebalance; }
-    void                         SetImplicitTranspose(const bool& implicit)            { implicitTranspose_ = implicit; }
-    void                         SetFuseProlongationAndUpdate(const bool& fuse)        { fuseProlongationAndUpdate_ = fuse; }
-
-    //@}
-
-    //!
-
-    template<class S2, class LO2, class GO2, class N2>
-    friend class Hierarchy;
-
-  private:
-    int  LastLevelID()      const { return Levels_.size() - 1; }
-    void DumpCurrentGraph(int level) const;
-
-  public:
-
-    //! Add a level at the end of the hierarchy
-    void AddLevel(const RCP<Level> & level);
-
-    //! Add a new level at the end of the hierarchy
-    void AddNewLevel();
-
-    //! Retrieve a certain level from hierarchy.
-    RCP<Level> & GetLevel(const int levelID = 0);
-
-    int    GetNumLevels() const;
-    int    GetGlobalNumLevels() const;
-
-    MagnitudeType GetRate() const { return rate_; }
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType MagnitudeType;
+
+  //! Data struct for defining stopping criteria of multigrid iteration
+  struct ConvData {
+    ConvData()
+      : maxIts_(1)
+      , tol_(-STS::magnitude(STS::one())) {}
+    ConvData(LO maxIts)
+      : maxIts_(maxIts)
+      , tol_(-STS::magnitude(STS::one())) {}
+    ConvData(MagnitudeType tol)
+      : maxIts_(10000)
+      , tol_(tol) {}
+    ConvData(std::pair<LO, MagnitudeType> p)
+      : maxIts_(p.first)
+      , tol_(p.second) {}
+
+    LO maxIts_;
+    MagnitudeType tol_;
+  };
 
-    // This function is global
-    double GetOperatorComplexity() const;
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    // This function is global
-    double GetSmootherComplexity() const;
+  //! Default constructor.
+  Hierarchy();
+  //! Constructor that labels the hierarchy.
+  Hierarchy(const std::string& label);
 
-    //! Helper function
-    void CheckLevel(Level& level, int levelID);
+  //! Constructor
+  Hierarchy(const RCP<Matrix>& A);
+
+  //! Constructor
+  Hierarchy(const RCP<Matrix>& A, const std::string& label);
+
+  //! Destructor.
+  virtual ~Hierarchy() {}
+
+  //@}
+
+  //! @name Set/Get Methods.
+  //@{
+
+  //!
+  static CycleType GetDefaultCycle() { return MasterList::getDefault<std::string>("cycle type") == "V" ? VCYCLE : WCYCLE; }
+  static int GetDefaultCycleStartLevel() { return MasterList::getDefault<int>("W cycle start level"); }
+  static bool GetDefaultImplicitTranspose() { return MasterList::getDefault<bool>("transpose: use implicit"); }
+  static bool GetDefaultFuseProlongationAndUpdate() { return MasterList::getDefault<bool>("fuse prolongation and update"); }
+  static Xpetra::global_size_t GetDefaultMaxCoarseSize() { return MasterList::getDefault<int>("coarse: max size"); }
+  static int GetDefaultMaxLevels() { return MasterList::getDefault<int>("max levels"); }
+  static bool GetDefaultPRrebalance() { return MasterList::getDefault<bool>("repartition: rebalance P and R"); }
+
+  Xpetra::global_size_t GetMaxCoarseSize() const { return maxCoarseSize_; }
+  bool GetImplicitTranspose() const { return implicitTranspose_; }
+  bool GetFuseProlongationAndUpdate() const { return fuseProlongationAndUpdate_; }
+
+  void SetMaxCoarseSize(Xpetra::global_size_t maxCoarseSize) { maxCoarseSize_ = maxCoarseSize; }
+  void SetPRrebalance(bool doPRrebalance) { doPRrebalance_ = doPRrebalance; }
+  void SetPRViaCopyrebalance(bool doPRViaCopyrebalance) { doPRViaCopyrebalance_ = doPRViaCopyrebalance; }
+  void SetImplicitTranspose(const bool& implicit) { implicitTranspose_ = implicit; }
+  void SetFuseProlongationAndUpdate(const bool& fuse) { fuseProlongationAndUpdate_ = fuse; }
+
+  //@}
+
+  //!
+
+  template <class S2, class LO2, class GO2, class N2>
+  friend class Hierarchy;
+
+ private:
+  int LastLevelID() const { return Levels_.size() - 1; }
+  void DumpCurrentGraph(int level) const;
+
+ public:
+  //! Add a level at the end of the hierarchy
+  void AddLevel(const RCP<Level>& level);
+
+  //! Add a new level at the end of the hierarchy
+  void AddNewLevel();
+
+  //! Retrieve a certain level from hierarchy.
+  RCP<Level>& GetLevel(const int levelID = 0);
+
+  int GetNumLevels() const;
+  int GetGlobalNumLevels() const;
+
+  MagnitudeType GetRate() const { return rate_; }
+
+  // This function is global
+  double GetOperatorComplexity() const;
+
+  // This function is global
+  double GetSmootherComplexity() const;
+
+  //! Helper function
+  void CheckLevel(Level& level, int levelID);
+
+  void SetMatvecParams(RCP<ParameterList> matvecParams);
+
+  //! Multi-level setup phase: build a new level of the hierarchy.
+  /*!  This method is aimed to be used in a loop building the hierarchy level by level. See Hierarchy::Setup(manager, startLevel, numDesiredLevels) for an example of usage.
+   *
+   *   @param coarseLevelID ID of the level to be built.
+   *   @param fineLevelManager defines how to build missing data of the fineLevel (example: aggregates)
+   *   @param coarseLevelManager defines how to build the level
+   *   @param nextLevelManager defines how the next coarse level will be built. This is used to post corresponding request before building the coarse level to keep useful data.
+
+   CoarseLevel is considered to be the last level if:
+    - input parameter isLastLevel == true
+    or
+    - Ac->getRowMap()->getGlobalNumElements() <= maxCoarseSize_
+   Method return true if CoarseLevel is the last level.
+
+   Pre-condition:
+    * FineLevel:
+       - must have kept useful data (TODO: not tested yet)
+       - must be Teuchos::null when Setup is called for finest level (Setup then automatically calls Request for "Smoother" and "CoarseSolver")
+    * CoarseLevel:
+       - already allocated (using Hierarchy::AddLevel())
+       - requests already posted
+         (exception: for finest level (=fineLevelManager==null) requests are called within setup routine)
+    * NextLevel:
+       - do not need to be allocate but could (FIXME: will be deleted if lastlevel...).
+       - should be null when Setup is called for last level
+
+   Post-condition:
+    * FineLevel:
+       - temporary data have been used and released (this condition is not tested)
+    * CoarseLevel:
+       - built, requests have been used
+       - if it is the last level (due to input parameter isLastLevel or getGlobalNumElements() <= maxCoarseSize_),
+         then the coarse solver factory of the factory manager have been used instead of the smoother factory.
+    * NextLevel:
+      If input parameter isLastLevel == false:
+       - have been allocated
+       - requests already posted.
+  */
+  bool Setup(int coarseLevelID, const RCP<const FactoryManagerBase> fineLevelManager /* = Teuchos::null */, const RCP<const FactoryManagerBase> coarseLevelManager,
+             const RCP<const FactoryManagerBase> nextLevelManager = Teuchos::null);
 
-    void SetMatvecParams(RCP<ParameterList> matvecParams);
+  //!
+  void Setup(const FactoryManagerBase& manager = FactoryManager(), int startLevel = 0, int numDesiredLevels = GetDefaultMaxLevels());
 
-    //! Multi-level setup phase: build a new level of the hierarchy.
-    /*!  This method is aimed to be used in a loop building the hierarchy level by level. See Hierarchy::Setup(manager, startLevel, numDesiredLevels) for an example of usage.
-     *
-     *   @param coarseLevelID ID of the level to be built.
-     *   @param fineLevelManager defines how to build missing data of the fineLevel (example: aggregates)
-     *   @param coarseLevelManager defines how to build the level
-     *   @param nextLevelManager defines how the next coarse level will be built. This is used to post corresponding request before building the coarse level to keep useful data.
+  void SetupRe();
 
-     CoarseLevel is considered to be the last level if:
-      - input parameter isLastLevel == true
-      or
-      - Ac->getRowMap()->getGlobalNumElements() <= maxCoarseSize_
-     Method return true if CoarseLevel is the last level.
+  //! Clear impermanent data from previous setup
+  void Clear(int startLevel = 0);
+  void ExpertClear();
 
-     Pre-condition:
-      * FineLevel:
-         - must have kept useful data (TODO: not tested yet)
-         - must be Teuchos::null when Setup is called for finest level (Setup then automatically calls Request for "Smoother" and "CoarseSolver")
-      * CoarseLevel:
-         - already allocated (using Hierarchy::AddLevel())
-         - requests already posted
-           (exception: for finest level (=fineLevelManager==null) requests are called within setup routine)
-      * NextLevel:
-         - do not need to be allocate but could (FIXME: will be deleted if lastlevel...).
-         - should be null when Setup is called for last level
+  //! Returns multigrid cycle type (supports VCYCLE and WCYCLE)
+  CycleType GetCycle() const { return Cycle_; }
 
-     Post-condition:
-      * FineLevel:
-         - temporary data have been used and released (this condition is not tested)
-      * CoarseLevel:
-         - built, requests have been used
-         - if it is the last level (due to input parameter isLastLevel or getGlobalNumElements() <= maxCoarseSize_),
-           then the coarse solver factory of the factory manager have been used instead of the smoother factory.
-      * NextLevel:
-        If input parameter isLastLevel == false:
-         - have been allocated
-         - requests already posted.
-    */
-    bool Setup(int coarseLevelID, const RCP<const FactoryManagerBase> fineLevelManager /* = Teuchos::null */, const RCP<const FactoryManagerBase> coarseLevelManager,
-               const RCP<const FactoryManagerBase> nextLevelManager = Teuchos::null);
+  //! Supports VCYCLE and WCYCLE types.
+  void SetCycle(CycleType Cycle) { Cycle_ = Cycle; }
 
-    //!
-    void Setup(const FactoryManagerBase& manager = FactoryManager(), int startLevel = 0, int numDesiredLevels = GetDefaultMaxLevels());
+  void SetCycleStartLevel(int cycleStart) { WCycleStartLevel_ = cycleStart; }
 
-    void SetupRe();
+  //! Specify damping factor alpha such that x = x + alpha*P*c, where c is the coarse grid correction.
+  void SetProlongatorScalingFactor(double scalingFactor) { scalingFactor_ = scalingFactor; }
 
-    //! Clear impermanent data from previous setup
-    void Clear(int startLevel = 0);
-    void ExpertClear();
+  /*!
+    @brief Apply the multigrid preconditioner.
 
-    //! Returns multigrid cycle type (supports VCYCLE and WCYCLE)
-    CycleType GetCycle()                 const { return Cycle_;  }
+    In theory, more general cycle types than just V- and W-cycles are possible.  However,
+    the enumerated type CycleType would have to be extended.
 
-    //! Supports VCYCLE and WCYCLE types.
-    void      SetCycle(CycleType Cycle)        { Cycle_ = Cycle; }
+    @param B right-hand side of linear problem
+    @param X initial and final (approximate) solution of linear problem
+    @param ConvData struct which stores convergence criteria (maximum number of multigrid iterations or stopping tolerance)
+    @param InitialGuessIsZero Indicates whether the initial guess is zero
+    @param startLevel index of starting level to build multigrid hierarchy (default = 0)
+  */
+  ConvergenceStatus Iterate(const MultiVector& B, MultiVector& X, ConvData conv = ConvData(),
+                            bool InitialGuessIsZero = false, LO startLevel = 0);
 
-    void      SetCycleStartLevel(int cycleStart)        { WCycleStartLevel_ = cycleStart; }
+  /*!
+    @brief Print matrices in the multigrid hierarchy to file.
 
-    //! Specify damping factor alpha such that x = x + alpha*P*c, where c is the coarse grid correction.
-    void SetProlongatorScalingFactor(double scalingFactor) { scalingFactor_ = scalingFactor; }
+    @param[in] start start level
+    @param[in] end   end level
 
-    /*!
-      @brief Apply the multigrid preconditioner.
+    Default behavior is to print system and transfer matrices from the entire hierarchy.
+    Files are named "A_0.m", "P_1.m", "R_1.m", etc, and are in matrix market coordinate format.
+  */
+  void Write(const LO& start = -1, const LO& end = -1, const std::string& suffix = "");
 
-      In theory, more general cycle types than just V- and W-cycles are possible.  However,
-      the enumerated type CycleType would have to be extended.
+  //@}
 
-      @param B right-hand side of linear problem
-      @param X initial and final (approximate) solution of linear problem
-      @param ConvData struct which stores convergence criteria (maximum number of multigrid iterations or stopping tolerance)
-      @param InitialGuessIsZero Indicates whether the initial guess is zero
-      @param startLevel index of starting level to build multigrid hierarchy (default = 0)
-    */
-    ConvergenceStatus Iterate(const MultiVector& B, MultiVector& X, ConvData conv = ConvData(),
-                       bool InitialGuessIsZero = false, LO startLevel = 0);
+  //! @name Permanent storage
+  //@{
 
-    /*!
-      @brief Print matrices in the multigrid hierarchy to file.
+  //! Call Level::Keep(ename, factory) for each level of the Hierarchy.
+  void Keep(const std::string& ename, const FactoryBase* factory = NoFactory::get());
 
-      @param[in] start start level
-      @param[in] end   end level
+  //! Call Level::Delete(ename, factory) for each level of the Hierarchy.
+  void Delete(const std::string& ename, const FactoryBase* factory = NoFactory::get());
 
-      Default behavior is to print system and transfer matrices from the entire hierarchy.
-      Files are named "A_0.m", "P_1.m", "R_1.m", etc, and are in matrix market coordinate format.
-    */
-    void Write(const LO &start=-1, const LO &end=-1, const std::string &suffix="");
+  //! Call Level::AddKeepFlag for each level of the Hierarchy.
+  void AddKeepFlag(const std::string& ename, const FactoryBase* factory = NoFactory::get(), KeepType keep = MueLu::Keep);
 
-    //@}
+  //! Call Level::RemoveKeepFlag for each level of the Hierarchy
+  void RemoveKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep = MueLu::All);
 
-    //! @name Permanent storage
-    //@{
+  //@}
 
-    //! Call Level::Keep(ename, factory) for each level of the Hierarchy.
-    void Keep(const std::string & ename, const FactoryBase* factory = NoFactory::get());
-
-    //! Call Level::Delete(ename, factory) for each level of the Hierarchy.
-    void Delete(const std::string& ename, const FactoryBase* factory = NoFactory::get());
-
-    //! Call Level::AddKeepFlag for each level of the Hierarchy.
-    void AddKeepFlag(const std::string & ename, const FactoryBase* factory = NoFactory::get(), KeepType keep = MueLu::Keep);
-
-    //! Call Level::RemoveKeepFlag for each level of the Hierarchy
-    void RemoveKeepFlag(const std::string & ename, const FactoryBase* factory, KeepType keep = MueLu::All);
-
-    //@}
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
-
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    /*! @brief Print the Hierarchy with some verbosity level to a FancyOStream object.
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-        @param[in] out The Teuchos::FancyOstream.
-        @param[in] verbLevel Controls amount of output.
-    */
-    void describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
-    void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
+  /*! @brief Print the Hierarchy with some verbosity level to a FancyOStream object.
 
-    //! Hierarchy::print is local hierarchy function, thus the statistics can be different from global ones
-    void print(std::ostream& out = std::cout, const VerbLevel verbLevel = (MueLu::Parameters | MueLu::Statistics0)) const;
+      @param[in] out The Teuchos::FancyOstream.
+      @param[in] verbLevel Controls amount of output.
+  */
+  void describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
 
-    /*! Indicate whether the multigrid method is a preconditioner or a solver.
+  //! Hierarchy::print is local hierarchy function, thus the statistics can be different from global ones
+  void print(std::ostream& out = std::cout, const VerbLevel verbLevel = (MueLu::Parameters | MueLu::Statistics0)) const;
 
-      This is used in conjunction with the verbosity level to determine whether the residuals can be printed.
-    */
-    void IsPreconditioner(const bool flag);
+  /*! Indicate whether the multigrid method is a preconditioner or a solver.
 
-    //@}
+    This is used in conjunction with the verbosity level to determine whether the residuals can be printed.
+  */
+  void IsPreconditioner(const bool flag);
 
-    void EnableGraphDumping(const std::string& filename, int levelID = 1) {
-      isDumpingEnabled_ = true;
-      dumpLevel_ = levelID;
-      dumpFile_  = filename;
-    }
+  //@}
 
-    void setlib(Xpetra::UnderlyingLib inlib) { lib_ = inlib; }
-    Xpetra::UnderlyingLib lib() { return lib_; }
+  void EnableGraphDumping(const std::string& filename, int levelID = 1) {
+    isDumpingEnabled_ = true;
+    dumpLevel_        = levelID;
+    dumpFile_         = filename;
+  }
 
-    //! force recreation of cached description_ next time description() is called:
-    void ResetDescription() {
-      description_ = "";
-    }
+  void setlib(Xpetra::UnderlyingLib inlib) { lib_ = inlib; }
+  Xpetra::UnderlyingLib lib() { return lib_; }
 
-    void AllocateLevelMultiVectors(int numvecs, bool forceMapCheck=false);
-    void DeleteLevelMultiVectors();
+  //! force recreation of cached description_ next time description() is called:
+  void ResetDescription() {
+    description_ = "";
+  }
 
-  protected:
-    const RCP<const FactoryManagerBase>& GetLevelManager(const int levelID) const {
-      return levelManagers_[levelID];
-    }
+  void AllocateLevelMultiVectors(int numvecs, bool forceMapCheck = false);
+  void DeleteLevelMultiVectors();
 
-  private:
-    //! Copy constructor is not implemented.
-    Hierarchy(const Hierarchy &h);
+ protected:
+  const RCP<const FactoryManagerBase>& GetLevelManager(const int levelID) const {
+    return levelManagers_[levelID];
+  }
 
-    //! Decide if the residual needs to be computed
-    bool IsCalculationOfResidualRequired(const LO startLevel, const ConvData& conv) const;
+ private:
+  //! Copy constructor is not implemented.
+  Hierarchy(const Hierarchy& h);
 
-    /*!
-    \brief Decide if the multigrid iteration is converged
+  //! Decide if the residual needs to be computed
+  bool IsCalculationOfResidualRequired(const LO startLevel, const ConvData& conv) const;
 
-    We judge convergence by comparing the current \c residualNorm
-    to the user given \c convergenceTolerance and then return the
-    appropriate \c ConvergenceStatus
-    */
-    ConvergenceStatus IsConverged(const Teuchos::Array<MagnitudeType>& residualNorm,
-        const MagnitudeType convergenceTolerance) const;
+  /*!
+  \brief Decide if the multigrid iteration is converged
 
-    //! Print \c residualNorm for this \c iteration to the screen
-    void PrintResidualHistory(const LO iteration,
-        const Teuchos::Array<MagnitudeType>& residualNorm) const;
+  We judge convergence by comparing the current \c residualNorm
+  to the user given \c convergenceTolerance and then return the
+  appropriate \c ConvergenceStatus
+  */
+  ConvergenceStatus IsConverged(const Teuchos::Array<MagnitudeType>& residualNorm,
+                                const MagnitudeType convergenceTolerance) const;
 
-    //! Compute the residual norm and print it depending on the verbosity level
-    ConvergenceStatus ComputeResidualAndPrintHistory(const Operator& A, const MultiVector& X,
-        const MultiVector& B, const LO iteration,
-        const LO startLevel, const ConvData& conv, MagnitudeType& previousResidualNorm);
+  //! Print \c residualNorm for this \c iteration to the screen
+  void PrintResidualHistory(const LO iteration,
+                            const Teuchos::Array<MagnitudeType>& residualNorm) const;
 
-    //! Container for Level objects
-    Array<RCP<Level> > Levels_;
+  //! Compute the residual norm and print it depending on the verbosity level
+  ConvergenceStatus ComputeResidualAndPrintHistory(const Operator& A, const MultiVector& X,
+                                                   const MultiVector& B, const LO iteration,
+                                                   const LO startLevel, const ConvData& conv, MagnitudeType& previousResidualNorm);
 
-    //! We replace coordinates GIDs to make them consistent with matrix GIDs,
-    //! even if user does not do that.  Ideally, though, we should completely
-    //! remove any notion of coordinate GIDs, and deal only with LIDs, assuming
-    //! that they are consistent with matrix block IDs
-    void ReplaceCoordinateMap(Level& level);
+  //! Container for Level objects
+  Array<RCP<Level> > Levels_;
 
-    //! Minimum size of a matrix on any level. If we fall below that, we stop
-    //! the coarsening
-    Xpetra::global_size_t maxCoarseSize_;
+  //! We replace coordinates GIDs to make them consistent with matrix GIDs,
+  //! even if user does not do that.  Ideally, though, we should completely
+  //! remove any notion of coordinate GIDs, and deal only with LIDs, assuming
+  //! that they are consistent with matrix block IDs
+  void ReplaceCoordinateMap(Level& level);
 
-    //! Potential speed up of the setup by skipping R construction, and using
-    //! transpose matrix-matrix product for RAP
-    bool implicitTranspose_;
+  //! Minimum size of a matrix on any level. If we fall below that, we stop
+  //! the coarsening
+  Xpetra::global_size_t maxCoarseSize_;
 
-    //! Potential speed up of the solve by fusing prolongation and update steps.
-    //! This can lead to more iterations to round-off error accumulation.
-    bool fuseProlongationAndUpdate_;
+  //! Potential speed up of the setup by skipping R construction, and using
+  //! transpose matrix-matrix product for RAP
+  bool implicitTranspose_;
 
-    //! Potential speed up of the setup by skipping rebalancing of P and R, and
-    //! doing extra import during solve
-    bool doPRrebalance_;
-    bool doPRViaCopyrebalance_;   // fully explicit, needed for CombinePFactory
+  //! Potential speed up of the solve by fusing prolongation and update steps.
+  //! This can lead to more iterations to round-off error accumulation.
+  bool fuseProlongationAndUpdate_;
 
-    //! Hierarchy may be used in a standalone mode, or as a preconditioner
-    bool isPreconditioner_;
+  //! Potential speed up of the setup by skipping rebalancing of P and R, and
+  //! doing extra import during solve
+  bool doPRrebalance_;
+  bool doPRViaCopyrebalance_;  // fully explicit, needed for CombinePFactory
 
-    //! V- or W-cycle
-    CycleType Cycle_;
+  //! Hierarchy may be used in a standalone mode, or as a preconditioner
+  bool isPreconditioner_;
 
-    //! Level at which to start W-cycle
-    int WCycleStartLevel_;
+  //! V- or W-cycle
+  CycleType Cycle_;
 
-    //! Scaling factor to be applied to coarse grid correction.
-    double scalingFactor_;
+  //! Level at which to start W-cycle
+  int WCycleStartLevel_;
 
-    //! Epetra/Tpetra mode
-    Xpetra::UnderlyingLib lib_;
+  //! Scaling factor to be applied to coarse grid correction.
+  double scalingFactor_;
 
-    //! cache description to avoid recreating in each call to description() - use ResetDescription() to force recreation in Setup, SetupRe, etc.
-    mutable std::string description_ = ""; // mutable so that we can lazily initialize in description(), which is declared const
+  //! Epetra/Tpetra mode
+  Xpetra::UnderlyingLib lib_;
 
-    /*!
-    @brief Graph dumping
+  //! cache description to avoid recreating in each call to description() - use ResetDescription() to force recreation in Setup, SetupRe, etc.
+  mutable std::string description_ = "";  // mutable so that we can lazily initialize in description(), which is declared const
 
-    If enabled, we dump the graph on a specified level into a specified file
-    */
-    bool isDumpingEnabled_;
-    // -1 = dump all levels, -2 = dump nothing
-    int  dumpLevel_;
-    std::string dumpFile_;
+  /*!
+  @brief Graph dumping
 
-    //! Convergece rate
-    MagnitudeType rate_;
+  If enabled, we dump the graph on a specified level into a specified file
+  */
+  bool isDumpingEnabled_;
+  // -1 = dump all levels, -2 = dump nothing
+  int dumpLevel_;
+  std::string dumpFile_;
 
-    //! Level managers used during the Setup
-    Array<RCP<const FactoryManagerBase> > levelManagers_;
+  //! Convergece rate
+  MagnitudeType rate_;
 
-    //! Caching (Multi)Vectors used in Hierarchy::Iterate()
-    int sizeOfAllocatedLevelMultiVectors_;
-    Array<RCP<MultiVector> > residual_, coarseRhs_, coarseX_, coarseImport_, coarseExport_, correction_;
+  //! Level managers used during the Setup
+  Array<RCP<const FactoryManagerBase> > levelManagers_;
 
+  //! Caching (Multi)Vectors used in Hierarchy::Iterate()
+  int sizeOfAllocatedLevelMultiVectors_;
+  Array<RCP<MultiVector> > residual_, coarseRhs_, coarseX_, coarseImport_, coarseExport_, correction_;
 
-  }; //class Hierarchy
+};  // class Hierarchy
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_HIERARCHY_SHORT
-#endif // MUELU_HIERARCHY_DECL_HPP
+#endif  // MUELU_HIERARCHY_DECL_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Hierarchy_def.hpp b/packages/muelu/src/MueCentral/MueLu_Hierarchy_def.hpp
index 13a40368f1c4..fab4ce803855 100644
--- a/packages/muelu/src/MueCentral/MueLu_Hierarchy_def.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_Hierarchy_def.hpp
@@ -72,895 +72,899 @@
 
 #include "Teuchos_TimeMonitor.hpp"
 
-
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy()
-    : maxCoarseSize_(GetDefaultMaxCoarseSize()), implicitTranspose_(GetDefaultImplicitTranspose()),
-      fuseProlongationAndUpdate_(GetDefaultFuseProlongationAndUpdate()),
-      doPRrebalance_(GetDefaultPRrebalance()), doPRViaCopyrebalance_(false), isPreconditioner_(true), Cycle_(GetDefaultCycle()), WCycleStartLevel_(0),
-      scalingFactor_(Teuchos::ScalarTraits<double>::one()), lib_(Xpetra::UseTpetra), isDumpingEnabled_(false), dumpLevel_(-2), rate_(-1),
-      sizeOfAllocatedLevelMultiVectors_(0)
-  {
-    AddLevel(rcp(new Level));
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const std::string& label)
-    : Hierarchy()
-  {
-    setObjectLabel(label);
-    Levels_[0]->setObjectLabel(label);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const RCP<Matrix>& A)
-    : maxCoarseSize_(GetDefaultMaxCoarseSize()), implicitTranspose_(GetDefaultImplicitTranspose()),
-      fuseProlongationAndUpdate_(GetDefaultFuseProlongationAndUpdate()),
-      doPRrebalance_(GetDefaultPRrebalance()), doPRViaCopyrebalance_(false), isPreconditioner_(true), Cycle_(GetDefaultCycle()), WCycleStartLevel_(0),
-      scalingFactor_(Teuchos::ScalarTraits<double>::one()), isDumpingEnabled_(false), dumpLevel_(-2), rate_(-1),
-      sizeOfAllocatedLevelMultiVectors_(0)
-  {
-    lib_ = A->getDomainMap()->lib();
-
-    RCP<Level> Finest = rcp(new Level);
-    AddLevel(Finest);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy()
+  : maxCoarseSize_(GetDefaultMaxCoarseSize())
+  , implicitTranspose_(GetDefaultImplicitTranspose())
+  , fuseProlongationAndUpdate_(GetDefaultFuseProlongationAndUpdate())
+  , doPRrebalance_(GetDefaultPRrebalance())
+  , doPRViaCopyrebalance_(false)
+  , isPreconditioner_(true)
+  , Cycle_(GetDefaultCycle())
+  , WCycleStartLevel_(0)
+  , scalingFactor_(Teuchos::ScalarTraits<double>::one())
+  , lib_(Xpetra::UseTpetra)
+  , isDumpingEnabled_(false)
+  , dumpLevel_(-2)
+  , rate_(-1)
+  , sizeOfAllocatedLevelMultiVectors_(0) {
+  AddLevel(rcp(new Level));
+}
 
-    Finest->Set("A", A);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const std::string& label)
+  : Hierarchy() {
+  setObjectLabel(label);
+  Levels_[0]->setObjectLabel(label);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const RCP<Matrix>& A, const std::string& label)
-    : Hierarchy(A)
-  {
-    setObjectLabel(label);
-    Levels_[0]->setObjectLabel(label);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const RCP<Matrix>& A)
+  : maxCoarseSize_(GetDefaultMaxCoarseSize())
+  , implicitTranspose_(GetDefaultImplicitTranspose())
+  , fuseProlongationAndUpdate_(GetDefaultFuseProlongationAndUpdate())
+  , doPRrebalance_(GetDefaultPRrebalance())
+  , doPRViaCopyrebalance_(false)
+  , isPreconditioner_(true)
+  , Cycle_(GetDefaultCycle())
+  , WCycleStartLevel_(0)
+  , scalingFactor_(Teuchos::ScalarTraits<double>::one())
+  , isDumpingEnabled_(false)
+  , dumpLevel_(-2)
+  , rate_(-1)
+  , sizeOfAllocatedLevelMultiVectors_(0) {
+  lib_ = A->getDomainMap()->lib();
+
+  RCP<Level> Finest = rcp(new Level);
+  AddLevel(Finest);
+
+  Finest->Set("A", A);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddLevel(const RCP<Level>& level) {
-    int levelID = LastLevelID() + 1; // ID of the inserted level
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Hierarchy(const RCP<Matrix>& A, const std::string& label)
+  : Hierarchy(A) {
+  setObjectLabel(label);
+  Levels_[0]->setObjectLabel(label);
+}
 
-    if (level->GetLevelID() != -1 && (level->GetLevelID() != levelID))
-      GetOStream(Warnings1) << "Hierarchy::AddLevel(): Level with ID=" << level->GetLevelID() <<
-          " have been added at the end of the hierarchy\n but its ID have been redefined" <<
-          " because last level ID of the hierarchy was " << LastLevelID() << "." << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddLevel(const RCP<Level>& level) {
+  int levelID = LastLevelID() + 1;  // ID of the inserted level
 
-    Levels_.push_back(level);
-    level->SetLevelID(levelID);
-    level->setlib(lib_);
+  if (level->GetLevelID() != -1 && (level->GetLevelID() != levelID))
+    GetOStream(Warnings1) << "Hierarchy::AddLevel(): Level with ID=" << level->GetLevelID() << " have been added at the end of the hierarchy\n but its ID have been redefined"
+                          << " because last level ID of the hierarchy was " << LastLevelID() << "." << std::endl;
 
-    level->SetPreviousLevel( (levelID == 0) ? Teuchos::null : Levels_[LastLevelID() - 1] );
-    level->setObjectLabel(this->getObjectLabel());
-  }
+  Levels_.push_back(level);
+  level->SetLevelID(levelID);
+  level->setlib(lib_);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddNewLevel() {
-    RCP<Level> newLevel = Levels_[LastLevelID()]->Build(); // new coarse level, using copy constructor
-    newLevel->setlib(lib_);
-    this->AddLevel(newLevel);                              // add to hierarchy
-  }
+  level->SetPreviousLevel((levelID == 0) ? Teuchos::null : Levels_[LastLevelID() - 1]);
+  level->setObjectLabel(this->getObjectLabel());
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Level> & Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLevel(const int levelID) {
-    TEUCHOS_TEST_FOR_EXCEPTION(levelID < 0 || levelID > LastLevelID(), Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::GetLevel(): invalid input parameter value: LevelID = " << levelID);
-    return Levels_[levelID];
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddNewLevel() {
+  RCP<Level> newLevel = Levels_[LastLevelID()]->Build();  // new coarse level, using copy constructor
+  newLevel->setlib(lib_);
+  this->AddLevel(newLevel);  // add to hierarchy
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNumLevels() const {
-    return Levels_.size();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Level>& Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLevel(const int levelID) {
+  TEUCHOS_TEST_FOR_EXCEPTION(levelID < 0 || levelID > LastLevelID(), Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::GetLevel(): invalid input parameter value: LevelID = " << levelID);
+  return Levels_[levelID];
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetGlobalNumLevels() const {
-    RCP<Operator> A = Levels_[0]->template Get<RCP<Operator> >("A");
-    RCP<const Teuchos::Comm<int> > comm = A->getDomainMap()->getComm();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNumLevels() const {
+  return Levels_.size();
+}
 
-    int numLevels = GetNumLevels();
-    int numGlobalLevels;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numLevels, Teuchos::ptr(&numGlobalLevels));
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetGlobalNumLevels() const {
+  RCP<Operator> A                     = Levels_[0]->template Get<RCP<Operator> >("A");
+  RCP<const Teuchos::Comm<int> > comm = A->getDomainMap()->getComm();
 
-    return numGlobalLevels;
-  }
+  int numLevels = GetNumLevels();
+  int numGlobalLevels;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numLevels, Teuchos::ptr(&numGlobalLevels));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetOperatorComplexity() const {
-    double totalNnz = 0, lev0Nnz = 1;
-    for (int i = 0; i < GetNumLevels(); ++i) {
-      TEUCHOS_TEST_FOR_EXCEPTION(!(Levels_[i]->IsAvailable("A")) , Exceptions::RuntimeError,
-                                 "Operator complexity cannot be calculated because A is unavailable on level " << i);
-      RCP<Operator> A = Levels_[i]->template Get<RCP<Operator> >("A");
-      if (A.is_null())
-        break;
+  return numGlobalLevels;
+}
 
-      RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
-      if (Am.is_null()) {
-        GetOStream(Warnings0) << "Some level operators are not matrices, operator complexity calculation aborted" << std::endl;
-        return 0.0;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetOperatorComplexity() const {
+  double totalNnz = 0, lev0Nnz = 1;
+  for (int i = 0; i < GetNumLevels(); ++i) {
+    TEUCHOS_TEST_FOR_EXCEPTION(!(Levels_[i]->IsAvailable("A")), Exceptions::RuntimeError,
+                               "Operator complexity cannot be calculated because A is unavailable on level " << i);
+    RCP<Operator> A = Levels_[i]->template Get<RCP<Operator> >("A");
+    if (A.is_null())
+      break;
 
-      totalNnz += as<double>(Am->getGlobalNumEntries());
-      if (i == 0)
-        lev0Nnz = totalNnz;
+    RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
+    if (Am.is_null()) {
+      GetOStream(Warnings0) << "Some level operators are not matrices, operator complexity calculation aborted" << std::endl;
+      return 0.0;
     }
-    return totalNnz / lev0Nnz;
+
+    totalNnz += as<double>(Am->getGlobalNumEntries());
+    if (i == 0)
+      lev0Nnz = totalNnz;
   }
+  return totalNnz / lev0Nnz;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSmootherComplexity() const {
-    double node_sc = 0, global_sc=0;
-    double a0_nnz =0;
-    const size_t INVALID = Teuchos::OrdinalTraits<size_t>::invalid();
-    // Get cost of fine matvec
-    if (GetNumLevels() <= 0) return -1.0;
-    if (!Levels_[0]->IsAvailable("A")) return -1.0;
-
-    RCP<Operator> A = Levels_[0]->template Get<RCP<Operator> >("A");
-    if (A.is_null()) return -1.0;
-    RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
-    if(Am.is_null()) return -1.0;
-    a0_nnz = as<double>(Am->getGlobalNumEntries());
-
-    // Get smoother complexity at each level
-    for (int i = 0; i < GetNumLevels(); ++i) {
-      size_t level_sc=0;
-      if(!Levels_[i]->IsAvailable("PreSmoother")) continue;
-      RCP<SmootherBase> S = Levels_[i]->template Get<RCP<SmootherBase> >("PreSmoother");
-      if (S.is_null()) continue;
-      level_sc = S->getNodeSmootherComplexity();
-      if(level_sc == INVALID) {global_sc=-1.0;break;}
-
-      node_sc += as<double>(level_sc);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSmootherComplexity() const {
+  double node_sc = 0, global_sc = 0;
+  double a0_nnz        = 0;
+  const size_t INVALID = Teuchos::OrdinalTraits<size_t>::invalid();
+  // Get cost of fine matvec
+  if (GetNumLevels() <= 0) return -1.0;
+  if (!Levels_[0]->IsAvailable("A")) return -1.0;
+
+  RCP<Operator> A = Levels_[0]->template Get<RCP<Operator> >("A");
+  if (A.is_null()) return -1.0;
+  RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
+  if (Am.is_null()) return -1.0;
+  a0_nnz = as<double>(Am->getGlobalNumEntries());
+
+  // Get smoother complexity at each level
+  for (int i = 0; i < GetNumLevels(); ++i) {
+    size_t level_sc = 0;
+    if (!Levels_[i]->IsAvailable("PreSmoother")) continue;
+    RCP<SmootherBase> S = Levels_[i]->template Get<RCP<SmootherBase> >("PreSmoother");
+    if (S.is_null()) continue;
+    level_sc = S->getNodeSmootherComplexity();
+    if (level_sc == INVALID) {
+      global_sc = -1.0;
+      break;
     }
 
-    double min_sc=0.0;
-    RCP<const Teuchos::Comm<int> > comm =A->getDomainMap()->getComm();
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_SUM,node_sc,Teuchos::ptr(&global_sc));
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_MIN,node_sc,Teuchos::ptr(&min_sc));
-
-    if(min_sc < 0.0) return -1.0;
-    else return global_sc / a0_nnz;
+    node_sc += as<double>(level_sc);
   }
 
+  double min_sc                       = 0.0;
+  RCP<const Teuchos::Comm<int> > comm = A->getDomainMap()->getComm();
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, node_sc, Teuchos::ptr(&global_sc));
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_MIN, node_sc, Teuchos::ptr(&min_sc));
 
+  if (min_sc < 0.0)
+    return -1.0;
+  else
+    return global_sc / a0_nnz;
+}
 
+// Coherence checks todo in Setup() (using an helper function):
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckLevel(Level& level, int levelID) {
+  TEUCHOS_TEST_FOR_EXCEPTION(level.lib() != lib_, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::CheckLevel(): wrong underlying linear algebra library.");
+  TEUCHOS_TEST_FOR_EXCEPTION(level.GetLevelID() != levelID, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::CheckLevel(): wrong level ID");
+  TEUCHOS_TEST_FOR_EXCEPTION(levelID != 0 && level.GetPreviousLevel() != Levels_[levelID - 1], Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Setup(): wrong level parent");
+}
 
-  // Coherence checks todo in Setup() (using an helper function):
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckLevel(Level& level, int levelID) {
-    TEUCHOS_TEST_FOR_EXCEPTION(level.lib() != lib_, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::CheckLevel(): wrong underlying linear algebra library.");
-    TEUCHOS_TEST_FOR_EXCEPTION(level.GetLevelID() != levelID, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::CheckLevel(): wrong level ID");
-    TEUCHOS_TEST_FOR_EXCEPTION(levelID != 0 && level.GetPreviousLevel() != Levels_[levelID-1], Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Setup(): wrong level parent");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetMatvecParams(RCP<ParameterList> matvecParams) {
-    for (int i = 0; i < GetNumLevels(); ++i) {
-      RCP<Level> level = Levels_[i];
-      if (level->IsAvailable("A")) {
-        RCP<Operator> Aop = level->Get<RCP<Operator> >("A");
-        RCP<Matrix> A = rcp_dynamic_cast<Matrix>(Aop);
-        if (!A.is_null()) {
-          RCP<const Import> xpImporter = A->getCrsGraph()->getImporter();
-          if (!xpImporter.is_null())
-            xpImporter->setDistributorParameters(matvecParams);
-          RCP<const Export> xpExporter = A->getCrsGraph()->getExporter();
-          if (!xpExporter.is_null())
-            xpExporter->setDistributorParameters(matvecParams);
-        }
-      }
-      if (level->IsAvailable("P")) {
-        RCP<Matrix> P = level->Get<RCP<Matrix> >("P");
-        RCP<const Import> xpImporter = P->getCrsGraph()->getImporter();
-        if (!xpImporter.is_null())
-          xpImporter->setDistributorParameters(matvecParams);
-        RCP<const Export> xpExporter = P->getCrsGraph()->getExporter();
-        if (!xpExporter.is_null())
-          xpExporter->setDistributorParameters(matvecParams);
-      }
-      if (level->IsAvailable("R")) {
-        RCP<Matrix> R = level->Get<RCP<Matrix> >("R");
-        RCP<const Import> xpImporter = R->getCrsGraph()->getImporter();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetMatvecParams(RCP<ParameterList> matvecParams) {
+  for (int i = 0; i < GetNumLevels(); ++i) {
+    RCP<Level> level = Levels_[i];
+    if (level->IsAvailable("A")) {
+      RCP<Operator> Aop = level->Get<RCP<Operator> >("A");
+      RCP<Matrix> A     = rcp_dynamic_cast<Matrix>(Aop);
+      if (!A.is_null()) {
+        RCP<const Import> xpImporter = A->getCrsGraph()->getImporter();
         if (!xpImporter.is_null())
           xpImporter->setDistributorParameters(matvecParams);
-        RCP<const Export> xpExporter = R->getCrsGraph()->getExporter();
+        RCP<const Export> xpExporter = A->getCrsGraph()->getExporter();
         if (!xpExporter.is_null())
           xpExporter->setDistributorParameters(matvecParams);
       }
-      if (level->IsAvailable("Importer")) {
-        RCP<const Import> xpImporter = level->Get< RCP<const Import> >("Importer");
-        if (!xpImporter.is_null())
-          xpImporter->setDistributorParameters(matvecParams);
-      }
+    }
+    if (level->IsAvailable("P")) {
+      RCP<Matrix> P                = level->Get<RCP<Matrix> >("P");
+      RCP<const Import> xpImporter = P->getCrsGraph()->getImporter();
+      if (!xpImporter.is_null())
+        xpImporter->setDistributorParameters(matvecParams);
+      RCP<const Export> xpExporter = P->getCrsGraph()->getExporter();
+      if (!xpExporter.is_null())
+        xpExporter->setDistributorParameters(matvecParams);
+    }
+    if (level->IsAvailable("R")) {
+      RCP<Matrix> R                = level->Get<RCP<Matrix> >("R");
+      RCP<const Import> xpImporter = R->getCrsGraph()->getImporter();
+      if (!xpImporter.is_null())
+        xpImporter->setDistributorParameters(matvecParams);
+      RCP<const Export> xpExporter = R->getCrsGraph()->getExporter();
+      if (!xpExporter.is_null())
+        xpExporter->setDistributorParameters(matvecParams);
+    }
+    if (level->IsAvailable("Importer")) {
+      RCP<const Import> xpImporter = level->Get<RCP<const Import> >("Importer");
+      if (!xpImporter.is_null())
+        xpImporter->setDistributorParameters(matvecParams);
     }
   }
+}
 
-  // The function uses three managers: fine, coarse and next coarse
-  // We construct the data for the coarse level, and do requests for the next coarse
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(int coarseLevelID,
-                                                                   const RCP<const FactoryManagerBase> fineLevelManager,
-                                                                   const RCP<const FactoryManagerBase> coarseLevelManager,
-                                                                   const RCP<const FactoryManagerBase> nextLevelManager) {
-    // Use PrintMonitor/TimerMonitor instead of just a FactoryMonitor to print "Level 0" instead of Hierarchy(0)
-    // Print is done after the requests for next coarse level
-
-    TEUCHOS_TEST_FOR_EXCEPTION(LastLevelID() < coarseLevelID, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy:Setup(): level " << coarseLevelID << " (specified by coarseLevelID argument) "
-                               "must be built before calling this function.");
-
-    Level& level = *Levels_[coarseLevelID];
-
-    std::string label        = FormattingHelper::getColonLabel(level.getObjectLabel());
-    TimeMonitor m1(*this, label + this->ShortClassName() + ": " + "Setup (total)");
-    TimeMonitor m2(*this, label + this->ShortClassName() + ": " + "Setup" + " (total, level=" + Teuchos::toString(coarseLevelID) + ")");
-
-    // TODO: pass coarseLevelManager by reference
-    TEUCHOS_TEST_FOR_EXCEPTION(coarseLevelManager == Teuchos::null, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Setup(): argument coarseLevelManager cannot be null");
+// The function uses three managers: fine, coarse and next coarse
+// We construct the data for the coarse level, and do requests for the next coarse
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(int coarseLevelID,
+                                                                 const RCP<const FactoryManagerBase> fineLevelManager,
+                                                                 const RCP<const FactoryManagerBase> coarseLevelManager,
+                                                                 const RCP<const FactoryManagerBase> nextLevelManager) {
+  // Use PrintMonitor/TimerMonitor instead of just a FactoryMonitor to print "Level 0" instead of Hierarchy(0)
+  // Print is done after the requests for next coarse level
 
-    typedef MueLu::TopRAPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>      TopRAPFactory;
-    typedef MueLu::TopSmootherFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> TopSmootherFactory;
+  TEUCHOS_TEST_FOR_EXCEPTION(LastLevelID() < coarseLevelID, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy:Setup(): level " << coarseLevelID << " (specified by coarseLevelID argument) "
+                                                                                    "must be built before calling this function.");
 
-    if (levelManagers_.size() < coarseLevelID+1)
-      levelManagers_.resize(coarseLevelID+1);
-    levelManagers_[coarseLevelID] = coarseLevelManager;
+  Level& level = *Levels_[coarseLevelID];
 
-    bool isFinestLevel = (fineLevelManager.is_null());
-    bool isLastLevel   = (nextLevelManager.is_null());
+  std::string label = FormattingHelper::getColonLabel(level.getObjectLabel());
+  TimeMonitor m1(*this, label + this->ShortClassName() + ": " + "Setup (total)");
+  TimeMonitor m2(*this, label + this->ShortClassName() + ": " + "Setup" + " (total, level=" + Teuchos::toString(coarseLevelID) + ")");
 
-    int oldRank = -1;
-    if (isFinestLevel) {
-      RCP<Operator>                  A         = level.Get< RCP<Operator> >("A");
-      RCP<const Map>                 domainMap = A->getDomainMap();
-      RCP<const Teuchos::Comm<int> > comm      = domainMap->getComm();
+  // TODO: pass coarseLevelManager by reference
+  TEUCHOS_TEST_FOR_EXCEPTION(coarseLevelManager == Teuchos::null, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Setup(): argument coarseLevelManager cannot be null");
 
-      // Initialize random seed for reproducibility
-      Utilities::SetRandomSeed(*comm);
+  typedef MueLu::TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> TopRAPFactory;
+  typedef MueLu::TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> TopSmootherFactory;
 
-      // Record the communicator on the level (used for timers sync)
-      level.SetComm(comm);
-      oldRank = SetProcRankVerbose(comm->getRank());
+  if (levelManagers_.size() < coarseLevelID + 1)
+    levelManagers_.resize(coarseLevelID + 1);
+  levelManagers_[coarseLevelID] = coarseLevelManager;
 
-      // Set the Hierarchy library to match that of the finest level matrix,
-      // even if it was already set
-      lib_ = domainMap->lib();
-      level.setlib(lib_);
+  bool isFinestLevel = (fineLevelManager.is_null());
+  bool isLastLevel   = (nextLevelManager.is_null());
 
-    } else {
-      // Permeate library to a coarser level
-      level.setlib(lib_);
+  int oldRank = -1;
+  if (isFinestLevel) {
+    RCP<Operator> A                     = level.Get<RCP<Operator> >("A");
+    RCP<const Map> domainMap            = A->getDomainMap();
+    RCP<const Teuchos::Comm<int> > comm = domainMap->getComm();
 
-      Level& prevLevel = *Levels_[coarseLevelID-1];
-      oldRank = SetProcRankVerbose(prevLevel.GetComm()->getRank());
-    }
+    // Initialize random seed for reproducibility
+    Utilities::SetRandomSeed(*comm);
 
-    CheckLevel(level, coarseLevelID);
+    // Record the communicator on the level (used for timers sync)
+    level.SetComm(comm);
+    oldRank = SetProcRankVerbose(comm->getRank());
 
-    // Attach FactoryManager to the fine level
-    RCP<SetFactoryManager> SFMFine;
-    if (!isFinestLevel)
-      SFMFine = rcp(new SetFactoryManager(Levels_[coarseLevelID-1], fineLevelManager));
+    // Set the Hierarchy library to match that of the finest level matrix,
+    // even if it was already set
+    lib_ = domainMap->lib();
+    level.setlib(lib_);
 
-    if (isFinestLevel && Levels_[coarseLevelID]->IsAvailable("Coordinates"))
-      ReplaceCoordinateMap(*Levels_[coarseLevelID]);
+  } else {
+    // Permeate library to a coarser level
+    level.setlib(lib_);
 
-    // Attach FactoryManager to the coarse level
-    SetFactoryManager SFMCoarse(Levels_[coarseLevelID], coarseLevelManager);
+    Level& prevLevel = *Levels_[coarseLevelID - 1];
+    oldRank          = SetProcRankVerbose(prevLevel.GetComm()->getRank());
+  }
 
-    if (isDumpingEnabled_ && (dumpLevel_ == 0 || dumpLevel_ == -1) && coarseLevelID == 1)
-      DumpCurrentGraph(0);
+  CheckLevel(level, coarseLevelID);
+
+  // Attach FactoryManager to the fine level
+  RCP<SetFactoryManager> SFMFine;
+  if (!isFinestLevel)
+    SFMFine = rcp(new SetFactoryManager(Levels_[coarseLevelID - 1], fineLevelManager));
+
+  if (isFinestLevel && Levels_[coarseLevelID]->IsAvailable("Coordinates"))
+    ReplaceCoordinateMap(*Levels_[coarseLevelID]);
+
+  // Attach FactoryManager to the coarse level
+  SetFactoryManager SFMCoarse(Levels_[coarseLevelID], coarseLevelManager);
+
+  if (isDumpingEnabled_ && (dumpLevel_ == 0 || dumpLevel_ == -1) && coarseLevelID == 1)
+    DumpCurrentGraph(0);
+
+  RCP<TopSmootherFactory> coarseFact;
+  RCP<TopSmootherFactory> smootherFact = rcp(new TopSmootherFactory(coarseLevelManager, "Smoother"));
+
+  int nextLevelID = coarseLevelID + 1;
+
+  RCP<SetFactoryManager> SFMNext;
+  if (isLastLevel == false) {
+    // We are not at the coarsest level, so there is going to be another level ("next coarse") after this one ("coarse")
+    if (nextLevelID > LastLevelID())
+      AddNewLevel();
+    CheckLevel(*Levels_[nextLevelID], nextLevelID);
+
+    // Attach FactoryManager to the next level (level after coarse)
+    SFMNext = rcp(new SetFactoryManager(Levels_[nextLevelID], nextLevelManager));
+    Levels_[nextLevelID]->Request(TopRAPFactory(coarseLevelManager, nextLevelManager));
+
+    // Do smoother requests here. We don't know whether this is going to be
+    // the coarsest level or not, but we need to DeclareInput before we call
+    // coarseRAPFactory.Build(), otherwise some stuff may be erased after
+    // level releases
+    level.Request(*smootherFact);
+
+  } else {
+    // Similar to smoother above, do the coarse solver request here. We don't
+    // know whether this is going to be the coarsest level or not, but we
+    // need to DeclareInput before we call coarseRAPFactory.Build(),
+    // otherwise some stuff may be erased after level releases. This is
+    // actually evident on ProjectorSmoother. It requires both "A" and
+    // "Nullspace". However, "Nullspace" is erased after all releases, so if
+    // we call the coarse factory request after RAP build we would not have
+    // any data, and cannot get it as we don't have previous managers. The
+    // typical trace looks like this:
+    //
+    // MueLu::Level(0)::GetFactory(Aggregates, 0): No FactoryManager
+    //   during request for data "     Aggregates" on level 0 by factory TentativePFactory
+    //   during request for data "              P" on level 1 by factory EminPFactory
+    //   during request for data "              P" on level 1 by factory TransPFactory
+    //   during request for data "              R" on level 1 by factory RAPFactory
+    //   during request for data "              A" on level 1 by factory TentativePFactory
+    //   during request for data "      Nullspace" on level 2 by factory NullspaceFactory
+    //   during request for data "      Nullspace" on level 2 by factory NullspacePresmoothFactory
+    //   during request for data "      Nullspace" on level 2 by factory ProjectorSmoother
+    //   during request for data "    PreSmoother" on level 2 by factory NoFactory
+    if (coarseFact.is_null())
+      coarseFact = rcp(new TopSmootherFactory(coarseLevelManager, "CoarseSolver"));
+    level.Request(*coarseFact);
+  }
 
-    RCP<TopSmootherFactory> coarseFact;
-    RCP<TopSmootherFactory> smootherFact = rcp(new TopSmootherFactory(coarseLevelManager, "Smoother"));
+  GetOStream(Runtime0) << std::endl;
+  PrintMonitor m0(*this, "Level " + Teuchos::toString(coarseLevelID), static_cast<MsgType>(Runtime0 | Test));
 
-    int nextLevelID = coarseLevelID + 1;
+  // Build coarse level hierarchy
+  RCP<Operator> Ac = Teuchos::null;
+  TopRAPFactory coarseRAPFactory(fineLevelManager, coarseLevelManager);
 
-    RCP<SetFactoryManager> SFMNext;
-    if (isLastLevel == false) {
-      // We are not at the coarsest level, so there is going to be another level ("next coarse") after this one ("coarse")
-      if (nextLevelID > LastLevelID())
-        AddNewLevel();
-      CheckLevel(*Levels_[nextLevelID], nextLevelID);
+  if (level.IsAvailable("A")) {
+    Ac = level.Get<RCP<Operator> >("A");
+  } else if (!isFinestLevel) {
+    // We only build here, the release is done later
+    coarseRAPFactory.Build(*level.GetPreviousLevel(), level);
+  }
 
-      // Attach FactoryManager to the next level (level after coarse)
-      SFMNext = rcp(new SetFactoryManager(Levels_[nextLevelID], nextLevelManager));
-      Levels_[nextLevelID]->Request(TopRAPFactory(coarseLevelManager, nextLevelManager));
+  bool setLastLevelviaMaxCoarseSize = false;
+  if (level.IsAvailable("A"))
+    Ac = level.Get<RCP<Operator> >("A");
+  RCP<Matrix> Acm = rcp_dynamic_cast<Matrix>(Ac);
+
+  // Record the communicator on the level
+  if (!Ac.is_null())
+    level.SetComm(Ac->getDomainMap()->getComm());
+
+  // Test if we reach the end of the hierarchy
+  bool isOrigLastLevel = isLastLevel;
+  if (isLastLevel) {
+    // Last level as we have achieved the max limit
+    isLastLevel = true;
+
+  } else if (Ac.is_null()) {
+    // Last level for this processor, as it does not belong to the next
+    // subcommunicator. Other processors may continue working on the
+    // hierarchy
+    isLastLevel = true;
+
+  } else {
+    if (!Acm.is_null() && Acm->getGlobalNumRows() <= maxCoarseSize_) {
+      // Last level as the size of the coarse matrix became too small
+      GetOStream(Runtime0) << "Max coarse size (<= " << maxCoarseSize_ << ") achieved" << std::endl;
+      isLastLevel = true;
+      if (Acm->getGlobalNumRows() != 0) setLastLevelviaMaxCoarseSize = true;
+    }
+  }
 
-      // Do smoother requests here. We don't know whether this is going to be
-      // the coarsest level or not, but we need to DeclareInput before we call
-      // coarseRAPFactory.Build(), otherwise some stuff may be erased after
-      // level releases
-      level.Request(*smootherFact);
+  if (!Ac.is_null() && !isFinestLevel) {
+    RCP<Operator> A = Levels_[coarseLevelID - 1]->template Get<RCP<Operator> >("A");
+    RCP<Matrix> Am  = rcp_dynamic_cast<Matrix>(A);
+
+    const double maxCoarse2FineRatio = 0.8;
+    if (!Acm.is_null() && !Am.is_null() && Acm->getGlobalNumRows() > maxCoarse2FineRatio * Am->getGlobalNumRows()) {
+      // We could abort here, but for now we simply notify user.
+      // Couple of additional points:
+      //   - if repartitioning is delayed until level K, but the aggregation
+      //     procedure stagnates between levels K-1 and K.  In this case,
+      //     repartitioning could enable faster coarsening once again, but the
+      //     hierarchy construction will abort due to the stagnation check.
+      //   - if the matrix is small enough, we could move it to one processor.
+      GetOStream(Warnings0) << "Aggregation stagnated. Please check your matrix and/or adjust your configuration file."
+                            << "Possible fixes:\n"
+                            << "  - reduce the maximum number of levels\n"
+                            << "  - enable repartitioning\n"
+                            << "  - increase the minimum coarse size." << std::endl;
+    }
+  }
 
-    } else {
-      // Similar to smoother above, do the coarse solver request here. We don't
-      // know whether this is going to be the coarsest level or not, but we
-      // need to DeclareInput before we call coarseRAPFactory.Build(),
-      // otherwise some stuff may be erased after level releases. This is
-      // actually evident on ProjectorSmoother. It requires both "A" and
-      // "Nullspace". However, "Nullspace" is erased after all releases, so if
-      // we call the coarse factory request after RAP build we would not have
-      // any data, and cannot get it as we don't have previous managers. The
-      // typical trace looks like this:
-      //
-      // MueLu::Level(0)::GetFactory(Aggregates, 0): No FactoryManager
-      //   during request for data "     Aggregates" on level 0 by factory TentativePFactory
-      //   during request for data "              P" on level 1 by factory EminPFactory
-      //   during request for data "              P" on level 1 by factory TransPFactory
-      //   during request for data "              R" on level 1 by factory RAPFactory
-      //   during request for data "              A" on level 1 by factory TentativePFactory
-      //   during request for data "      Nullspace" on level 2 by factory NullspaceFactory
-      //   during request for data "      Nullspace" on level 2 by factory NullspacePresmoothFactory
-      //   during request for data "      Nullspace" on level 2 by factory ProjectorSmoother
-      //   during request for data "    PreSmoother" on level 2 by factory NoFactory
+  if (isLastLevel) {
+    if (!isOrigLastLevel) {
+      // We did not expect to finish this early so we did request a smoother.
+      // We need a coarse solver instead. Do the magic.
+      level.Release(*smootherFact);
       if (coarseFact.is_null())
         coarseFact = rcp(new TopSmootherFactory(coarseLevelManager, "CoarseSolver"));
       level.Request(*coarseFact);
     }
 
-    GetOStream(Runtime0) << std::endl;
-    PrintMonitor m0(*this, "Level " +  Teuchos::toString(coarseLevelID), static_cast<MsgType>(Runtime0 | Test));
-
-    // Build coarse level hierarchy
-    RCP<Operator> Ac = Teuchos::null;
-    TopRAPFactory coarseRAPFactory(fineLevelManager, coarseLevelManager);
-
-    if (level.IsAvailable("A")) {
-      Ac = level.Get<RCP<Operator> >("A");
-    } else if (!isFinestLevel) {
-      // We only build here, the release is done later
-      coarseRAPFactory.Build(*level.GetPreviousLevel(), level);
-    }
+    // Do the actual build, if we have any data.
+    // NOTE: this is not a great check, we may want to call Build() regardless.
+    if (!Ac.is_null())
+      coarseFact->Build(level);
 
-    bool setLastLevelviaMaxCoarseSize = false;
-    if (level.IsAvailable("A"))
-      Ac = level.Get<RCP<Operator> >("A");
-    RCP<Matrix> Acm = rcp_dynamic_cast<Matrix>(Ac);
+    // Once the dirty deed is done, release stuff. The smoother has already
+    // been released.
+    level.Release(*coarseFact);
 
-    // Record the communicator on the level
+  } else {
+    // isLastLevel = false => isOrigLastLevel = false, meaning that we have
+    // requested the smoother. Now we need to build it and to release it.
+    // We don't need to worry about the coarse solver, as we didn't request it.
     if (!Ac.is_null())
-      level.SetComm(Ac->getDomainMap()->getComm());
+      smootherFact->Build(level);
 
-    // Test if we reach the end of the hierarchy
-    bool isOrigLastLevel = isLastLevel;
-    if (isLastLevel) {
-      // Last level as we have achieved the max limit
-      isLastLevel = true;
-
-    } else if (Ac.is_null()) {
-      // Last level for this processor, as it does not belong to the next
-      // subcommunicator. Other processors may continue working on the
-      // hierarchy
-      isLastLevel = true;
+    level.Release(*smootherFact);
+  }
 
-    } else {
-      if (!Acm.is_null() && Acm->getGlobalNumRows() <= maxCoarseSize_) {
-        // Last level as the size of the coarse matrix became too small
-        GetOStream(Runtime0) << "Max coarse size (<= " << maxCoarseSize_ << ") achieved" << std::endl;
-        isLastLevel = true;
-        if (Acm->getGlobalNumRows() != 0) setLastLevelviaMaxCoarseSize = true;
+  if (isLastLevel == true) {
+    int actualNumLevels = nextLevelID;
+    if (isOrigLastLevel == false) {
+      // Earlier in the function, we constructed the next coarse level, and requested data for the that level,
+      // assuming that we are not at the coarsest level. Now, we changed our mind, so we have to release those.
+      Levels_[nextLevelID]->Release(TopRAPFactory(coarseLevelManager, nextLevelManager));
+
+      // We truncate/resize the hierarchy and possibly remove the last created level if there is
+      // something wrong with it as indicated by its P not being valid. This might happen
+      // if the global number of aggregates turns out to be zero
+
+      if (!setLastLevelviaMaxCoarseSize) {
+        if (Levels_[nextLevelID - 1]->IsAvailable("P")) {
+          if (Levels_[nextLevelID - 1]->template Get<RCP<Matrix> >("P") == Teuchos::null) actualNumLevels = nextLevelID - 1;
+        } else
+          actualNumLevels = nextLevelID - 1;
       }
     }
+    if (actualNumLevels == nextLevelID - 1) {
+      // Didn't expect to finish early so we requested smoother but need coarse solver instead.
+      Levels_[nextLevelID - 2]->Release(*smootherFact);
 
-    if (!Ac.is_null() && !isFinestLevel) {
-      RCP<Operator> A  = Levels_[coarseLevelID-1]->template Get< RCP<Operator> >("A");
-      RCP<Matrix>   Am = rcp_dynamic_cast<Matrix>(A);
-
-      const double maxCoarse2FineRatio = 0.8;
-      if (!Acm.is_null() && !Am.is_null() && Acm->getGlobalNumRows() > maxCoarse2FineRatio * Am->getGlobalNumRows()) {
-        // We could abort here, but for now we simply notify user.
-        // Couple of additional points:
-        //   - if repartitioning is delayed until level K, but the aggregation
-        //     procedure stagnates between levels K-1 and K.  In this case,
-        //     repartitioning could enable faster coarsening once again, but the
-        //     hierarchy construction will abort due to the stagnation check.
-        //   - if the matrix is small enough, we could move it to one processor.
-        GetOStream(Warnings0) << "Aggregation stagnated. Please check your matrix and/or adjust your configuration file."
-            << "Possible fixes:\n"
-            << "  - reduce the maximum number of levels\n"
-            << "  - enable repartitioning\n"
-            << "  - increase the minimum coarse size." << std::endl;
-
-      }
+      if (Levels_[nextLevelID - 2]->IsAvailable("PreSmoother")) Levels_[nextLevelID - 2]->RemoveKeepFlag("PreSmoother", NoFactory::get());
+      if (Levels_[nextLevelID - 2]->IsAvailable("PostSmoother")) Levels_[nextLevelID - 2]->RemoveKeepFlag("PostSmoother", NoFactory::get());
+      if (coarseFact.is_null())
+        coarseFact = rcp(new TopSmootherFactory(coarseLevelManager, "CoarseSolver"));
+      Levels_[nextLevelID - 2]->Request(*coarseFact);
+      if (!(Levels_[nextLevelID - 2]->template Get<RCP<Matrix> >("A").is_null()))
+        coarseFact->Build(*(Levels_[nextLevelID - 2]));
+      Levels_[nextLevelID - 2]->Release(*coarseFact);
     }
+    Levels_.resize(actualNumLevels);
+  }
 
-    if (isLastLevel) {
-      if (!isOrigLastLevel) {
-        // We did not expect to finish this early so we did request a smoother.
-        // We need a coarse solver instead. Do the magic.
-        level.Release(*smootherFact);
-        if (coarseFact.is_null())
-          coarseFact = rcp(new TopSmootherFactory(coarseLevelManager, "CoarseSolver"));
-        level.Request(*coarseFact);
-      }
+  // I think this is the proper place for graph so that it shows every dependence
+  if (isDumpingEnabled_ && ((dumpLevel_ > 0 && coarseLevelID == dumpLevel_) || dumpLevel_ == -1))
+    DumpCurrentGraph(coarseLevelID);
 
-      // Do the actual build, if we have any data.
-      // NOTE: this is not a great check, we may want to call Build() regardless.
-      if (!Ac.is_null())
-        coarseFact->Build(level);
+  if (!isFinestLevel) {
+    // Release the hierarchy data
+    // We release so late to help blocked solvers, as the smoothers for them need A blocks
+    // which we construct in RAPFactory
+    level.Release(coarseRAPFactory);
+  }
 
-      // Once the dirty deed is done, release stuff. The smoother has already
-      // been released.
-      level.Release(*coarseFact);
+  if (oldRank != -1)
+    SetProcRankVerbose(oldRank);
 
-    } else {
-      // isLastLevel = false => isOrigLastLevel = false, meaning that we have
-      // requested the smoother. Now we need to build it and to release it.
-      // We don't need to worry about the coarse solver, as we didn't request it.
-      if (!Ac.is_null())
-        smootherFact->Build(level);
+  return isLastLevel;
+}
 
-      level.Release(*smootherFact);
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupRe() {
+  int numLevels = Levels_.size();
+  TEUCHOS_TEST_FOR_EXCEPTION(levelManagers_.size() != numLevels, Exceptions::RuntimeError,
+                             "Hierarchy::SetupRe: " << Levels_.size() << " levels, but  " << levelManagers_.size() << " level factory managers");
 
-    if (isLastLevel == true) {
-      int actualNumLevels = nextLevelID;
-      if (isOrigLastLevel == false) {
-        // Earlier in the function, we constructed the next coarse level, and requested data for the that level,
-        // assuming that we are not at the coarsest level. Now, we changed our mind, so we have to release those.
-        Levels_[nextLevelID]->Release(TopRAPFactory(coarseLevelManager, nextLevelManager));
+  const int startLevel = 0;
+  Clear(startLevel);
 
-        // We truncate/resize the hierarchy and possibly remove the last created level if there is
-        // something wrong with it as indicated by its P not being valid. This might happen
-        // if the global number of aggregates turns out to be zero
+#ifdef HAVE_MUELU_DEBUG
+  // Reset factories' data used for debugging
+  for (int i = 0; i < numLevels; i++)
+    levelManagers_[i]->ResetDebugData();
 
+#endif
 
-        if (!setLastLevelviaMaxCoarseSize) {
-          if   (Levels_[nextLevelID-1]->IsAvailable("P"))  {
-            if (Levels_[nextLevelID-1]->template Get<RCP<Matrix> >("P") == Teuchos::null)   actualNumLevels = nextLevelID-1;
-          }
-          else actualNumLevels = nextLevelID-1;
-        }
-      }
-       if (actualNumLevels == nextLevelID-1) {
-         // Didn't expect to finish early so we requested smoother but need coarse solver instead.
-         Levels_[nextLevelID-2]->Release(*smootherFact);
-
-         if (Levels_[nextLevelID-2]->IsAvailable("PreSmoother") ) Levels_[nextLevelID-2]->RemoveKeepFlag("PreSmoother" ,NoFactory::get());
-         if (Levels_[nextLevelID-2]->IsAvailable("PostSmoother")) Levels_[nextLevelID-2]->RemoveKeepFlag("PostSmoother",NoFactory::get());
-         if (coarseFact.is_null())
-           coarseFact = rcp(new TopSmootherFactory(coarseLevelManager, "CoarseSolver"));
-         Levels_[nextLevelID-2]->Request(*coarseFact);
-         if ( !(Levels_[nextLevelID-2]->template Get<RCP<Matrix> >("A").is_null() ))
-           coarseFact->Build( *(Levels_[nextLevelID-2]));
-         Levels_[nextLevelID-2]->Release(*coarseFact);
-       }
-       Levels_.resize(actualNumLevels);
-    }
+  int levelID;
+  for (levelID = startLevel; levelID < numLevels;) {
+    bool r = Setup(levelID,
+                   (levelID != 0 ? levelManagers_[levelID - 1] : Teuchos::null),
+                   levelManagers_[levelID],
+                   (levelID + 1 != numLevels ? levelManagers_[levelID + 1] : Teuchos::null));
+    levelID++;
+    if (r) break;
+  }
+  // We may construct fewer levels for some reason, make sure we continue
+  // doing that in the future
+  Levels_.resize(levelID);
+  levelManagers_.resize(levelID);
 
-    // I think this is the proper place for graph so that it shows every dependence
-    if (isDumpingEnabled_ && ( (dumpLevel_ > 0 && coarseLevelID == dumpLevel_) || dumpLevel_ == -1 ) )
-      DumpCurrentGraph(coarseLevelID);
+  int sizeOfVecs = sizeOfAllocatedLevelMultiVectors_;
 
-    if (!isFinestLevel) {
-      // Release the hierarchy data
-      // We release so late to help blocked solvers, as the smoothers for them need A blocks
-      // which we construct in RAPFactory
-      level.Release(coarseRAPFactory);
-    }
+  AllocateLevelMultiVectors(sizeOfVecs, true);
 
-    if (oldRank != -1)
-      SetProcRankVerbose(oldRank);
+  // since the # of levels, etc. may have changed, force re-determination of description during next call to description()
+  ResetDescription();
 
-    return isLastLevel;
-  }
+  describe(GetOStream(Statistics0), GetVerbLevel());
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupRe() {
-    int numLevels = Levels_.size();
-    TEUCHOS_TEST_FOR_EXCEPTION(levelManagers_.size() != numLevels, Exceptions::RuntimeError,
-                               "Hierarchy::SetupRe: " << Levels_.size() << " levels, but  " << levelManagers_.size() << " level factory managers");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(const FactoryManagerBase& manager, int startLevel, int numDesiredLevels) {
+  // Use MueLu::BaseClass::description() to avoid printing "{numLevels = 1}" (numLevels is increasing...)
+  PrintMonitor m0(*this, "Setup (" + this->MueLu::BaseClass::description() + ")", Runtime0);
 
-    const int startLevel = 0;
-    Clear(startLevel);
+  Clear(startLevel);
 
-#ifdef HAVE_MUELU_DEBUG
-      // Reset factories' data used for debugging
-      for (int i = 0; i < numLevels; i++)
-        levelManagers_[i]->ResetDebugData();
+  // Check Levels_[startLevel] exists.
+  TEUCHOS_TEST_FOR_EXCEPTION(Levels_.size() <= startLevel, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Setup(): fine level (" << startLevel << ") does not exist");
 
-#endif
+  TEUCHOS_TEST_FOR_EXCEPTION(numDesiredLevels <= 0, Exceptions::RuntimeError,
+                             "Constructing non-positive (" << numDesiredLevels << ") number of levels does not make sense.");
 
-    int levelID;
-    for (levelID = startLevel; levelID < numLevels;) {
-      bool r = Setup(levelID,
-                     (levelID != 0 ? levelManagers_[levelID-1] : Teuchos::null),
-                     levelManagers_[levelID],
-                     (levelID+1 != numLevels ? levelManagers_[levelID+1] : Teuchos::null));
-      levelID++;
-      if (r) break;
-    }
-    // We may construct fewer levels for some reason, make sure we continue
-    // doing that in the future
-    Levels_       .resize(levelID);
-    levelManagers_.resize(levelID);
+  // Check for fine level matrix A
+  TEUCHOS_TEST_FOR_EXCEPTION(!Levels_[startLevel]->IsAvailable("A"), Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Setup(): fine level (" << startLevel << ") has no matrix A! "
+                                                                                        "Set fine level matrix A using Level.Set()");
 
-    int sizeOfVecs = sizeOfAllocatedLevelMultiVectors_;
+  RCP<Operator> A = Levels_[startLevel]->template Get<RCP<Operator> >("A");
+  lib_            = A->getDomainMap()->lib();
 
-    AllocateLevelMultiVectors(sizeOfVecs, true);
+  if (IsPrint(Statistics2)) {
+    RCP<Matrix> Amat = rcp_dynamic_cast<Matrix>(A);
 
-    // since the # of levels, etc. may have changed, force re-determination of description during next call to description()
-    ResetDescription();
+    if (!Amat.is_null()) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
 
-    describe(GetOStream(Statistics0), GetVerbLevel());
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Amat, "A0", params);
+    } else {
+      GetOStream(Warnings1) << "Fine level operator is not a matrix, statistics are not available" << std::endl;
+    }
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(const FactoryManagerBase& manager, int startLevel, int numDesiredLevels) {
-    // Use MueLu::BaseClass::description() to avoid printing "{numLevels = 1}" (numLevels is increasing...)
-    PrintMonitor m0(*this, "Setup (" + this->MueLu::BaseClass::description() + ")", Runtime0);
+  RCP<const FactoryManagerBase> rcpmanager = rcpFromRef(manager);
+
+  const int lastLevel = startLevel + numDesiredLevels - 1;
+  GetOStream(Runtime0) << "Setup loop: startLevel = " << startLevel << ", lastLevel = " << lastLevel
+                       << " (stop if numLevels = " << numDesiredLevels << " or Ac.size() < " << maxCoarseSize_ << ")" << std::endl;
+
+  // Setup multigrid levels
+  int iLevel = 0;
+  if (numDesiredLevels == 1) {
+    iLevel = 0;
+    Setup(startLevel, Teuchos::null, rcpmanager, Teuchos::null);  // setup finest==coarsest level (first and last managers are Teuchos::null)
+
+  } else {
+    bool bIsLastLevel = Setup(startLevel, Teuchos::null, rcpmanager, rcpmanager);  // setup finest level (level 0) (first manager is Teuchos::null)
+    if (bIsLastLevel == false) {
+      for (iLevel = startLevel + 1; iLevel < lastLevel; iLevel++) {
+        bIsLastLevel = Setup(iLevel, rcpmanager, rcpmanager, rcpmanager);  // setup intermediate levels
+        if (bIsLastLevel == true)
+          break;
+      }
+      if (bIsLastLevel == false)
+        Setup(lastLevel, rcpmanager, rcpmanager, Teuchos::null);  // setup coarsest level (last manager is Teuchos::null)
+    }
+  }
 
-    Clear(startLevel);
+  // TODO: some check like this should be done at the beginning of the routine
+  TEUCHOS_TEST_FOR_EXCEPTION(iLevel != Levels_.size() - 1, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Setup(): number of level");
 
-    // Check Levels_[startLevel] exists.
-    TEUCHOS_TEST_FOR_EXCEPTION(Levels_.size() <= startLevel, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Setup(): fine level (" << startLevel << ") does not exist");
+  // TODO: this is not exception safe: manager will still hold default
+  // factories if you exit this function with an exception
+  manager.Clean();
 
-    TEUCHOS_TEST_FOR_EXCEPTION(numDesiredLevels <= 0, Exceptions::RuntimeError,
-                               "Constructing non-positive (" << numDesiredLevels << ") number of levels does not make sense.");
+  describe(GetOStream(Statistics0), GetVerbLevel());
+}
 
-    // Check for fine level matrix A
-    TEUCHOS_TEST_FOR_EXCEPTION(!Levels_[startLevel]->IsAvailable("A"), Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Setup(): fine level (" << startLevel << ") has no matrix A! "
-                               "Set fine level matrix A using Level.Set()");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Clear(int startLevel) {
+  if (startLevel < GetNumLevels())
+    GetOStream(Runtime0) << "Clearing old data (if any)" << std::endl;
 
-    RCP<Operator> A = Levels_[startLevel]->template Get<RCP<Operator> >("A");
-    lib_ = A->getDomainMap()->lib();
+  for (int iLevel = startLevel; iLevel < GetNumLevels(); iLevel++)
+    Levels_[iLevel]->Clear();
+}
 
-    if (IsPrint(Statistics2)) {
-      RCP<Matrix> Amat = rcp_dynamic_cast<Matrix>(A);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ExpertClear() {
+  GetOStream(Runtime0) << "Clearing old data (expert)" << std::endl;
+  for (int iLevel = 0; iLevel < GetNumLevels(); iLevel++)
+    Levels_[iLevel]->ExpertClear();
+}
 
-      if (!Amat.is_null()) {
-          RCP<ParameterList> params = rcp(new ParameterList());
-          params->set("printLoadBalancingInfo", true);
-          params->set("printCommInfo",          true);
+#if defined(HAVE_MUELU_EXPERIMENTAL) && defined(HAVE_MUELU_ADDITIVE_VARIANT)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const MultiVector& B, MultiVector& X, ConvData conv,
+                                                                                bool InitialGuessIsZero, LO startLevel) {
+  LO nIts           = conv.maxIts_;
+  MagnitudeType tol = conv.tol_;
 
-          GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Amat, "A0", params);
-      } else {
-          GetOStream(Warnings1) << "Fine level operator is not a matrix, statistics are not available" << std::endl;
-      }
-    }
+  std::string prefix       = this->ShortClassName() + ": ";
+  std::string levelSuffix  = " (level=" + toString(startLevel) + ")";
+  std::string levelSuffix1 = " (level=" + toString(startLevel + 1) + ")";
 
-    RCP<const FactoryManagerBase> rcpmanager = rcpFromRef(manager);
+  using namespace Teuchos;
+  RCP<Time> CompTime              = Teuchos::TimeMonitor::getNewCounter(prefix + "Computational Time (total)");
+  RCP<Time> Concurrent            = Teuchos::TimeMonitor::getNewCounter(prefix + "Concurrent portion");
+  RCP<Time> ApplyR                = Teuchos::TimeMonitor::getNewCounter(prefix + "R: Computational Time");
+  RCP<Time> ApplyPbar             = Teuchos::TimeMonitor::getNewCounter(prefix + "Pbar: Computational Time");
+  RCP<Time> CompFine              = Teuchos::TimeMonitor::getNewCounter(prefix + "Fine: Computational Time");
+  RCP<Time> CompCoarse            = Teuchos::TimeMonitor::getNewCounter(prefix + "Coarse: Computational Time");
+  RCP<Time> ApplySum              = Teuchos::TimeMonitor::getNewCounter(prefix + "Sum: Computational Time");
+  RCP<Time> Synchronize_beginning = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_beginning");
+  RCP<Time> Synchronize_center    = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_center");
+  RCP<Time> Synchronize_end       = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_end");
 
-    const int lastLevel = startLevel + numDesiredLevels - 1;
-    GetOStream(Runtime0) << "Setup loop: startLevel = " << startLevel << ", lastLevel = " << lastLevel
-        << " (stop if numLevels = " << numDesiredLevels << " or Ac.size() < " << maxCoarseSize_ << ")" << std::endl;
+  RCP<Level> Fine = Levels_[0];
+  RCP<Level> Coarse;
 
-    // Setup multigrid levels
-    int iLevel = 0;
-    if (numDesiredLevels == 1) {
-      iLevel = 0;
-      Setup(startLevel, Teuchos::null, rcpmanager, Teuchos::null);                     // setup finest==coarsest level (first and last managers are Teuchos::null)
+  RCP<Operator> A                                      = Fine->Get<RCP<Operator> >("A");
+  Teuchos::RCP<const Teuchos::Comm<int> > communicator = A->getDomainMap()->getComm();
 
-    } else {
-      bool bIsLastLevel = Setup(startLevel, Teuchos::null, rcpmanager, rcpmanager);    // setup finest level (level 0) (first manager is Teuchos::null)
-      if (bIsLastLevel == false) {
-        for (iLevel = startLevel + 1; iLevel < lastLevel; iLevel++) {
-          bIsLastLevel = Setup(iLevel, rcpmanager, rcpmanager, rcpmanager);            // setup intermediate levels
-          if (bIsLastLevel == true)
-            break;
-        }
-        if (bIsLastLevel == false)
-          Setup(lastLevel, rcpmanager, rcpmanager, Teuchos::null);                     // setup coarsest level (last manager is Teuchos::null)
-      }
-    }
+  // Synchronize_beginning->start();
+  // communicator->barrier();
+  // Synchronize_beginning->stop();
 
-    // TODO: some check like this should be done at the beginning of the routine
-    TEUCHOS_TEST_FOR_EXCEPTION(iLevel != Levels_.size() - 1, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Setup(): number of level");
+  CompTime->start();
 
-    // TODO: this is not exception safe: manager will still hold default
-    // factories if you exit this function with an exception
-    manager.Clean();
+  SC one = STS::one(), zero = STS::zero();
 
-    describe(GetOStream(Statistics0), GetVerbLevel());
-  }
+  bool zeroGuess = InitialGuessIsZero;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Clear(int startLevel) {
-    if (startLevel < GetNumLevels())
-      GetOStream(Runtime0) << "Clearing old data (if any)" << std::endl;
+  // =======   UPFRONT DEFINITION OF COARSE VARIABLES ===========
 
-    for (int iLevel = startLevel; iLevel < GetNumLevels(); iLevel++)
-      Levels_[iLevel]->Clear();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ExpertClear() {
-    GetOStream(Runtime0) << "Clearing old data (expert)" << std::endl;
-    for (int iLevel = 0; iLevel < GetNumLevels(); iLevel++)
-      Levels_[iLevel]->ExpertClear();
-  }
-
-#if defined(HAVE_MUELU_EXPERIMENTAL) && defined(HAVE_MUELU_ADDITIVE_VARIANT)
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const MultiVector& B, MultiVector& X, ConvData conv,
-                                                                           bool InitialGuessIsZero, LO startLevel) {
-    LO            nIts = conv.maxIts_;
-    MagnitudeType tol  = conv.tol_;
-
-    std::string prefix       = this->ShortClassName() + ": ";
-    std::string levelSuffix  = " (level=" + toString(startLevel) + ")";
-    std::string levelSuffix1 = " (level=" + toString(startLevel+1) + ")";
+  // RCP<const Map> origMap;
+  RCP<Operator> P;
+  RCP<Operator> Pbar;
+  RCP<Operator> R;
+  RCP<MultiVector> coarseRhs, coarseX;
+  RCP<Operator> Ac;
+  RCP<SmootherBase> preSmoo_coarse, postSmoo_coarse;
+  bool emptyCoarseSolve              = true;
+  RCP<MultiVector> coarseX_prolonged = MultiVectorFactory::Build(X.getMap(), X.getNumVectors(), true);
 
-    using namespace Teuchos;
-    RCP<Time> CompTime                 = Teuchos::TimeMonitor::getNewCounter(prefix + "Computational Time (total)");
-    RCP<Time> Concurrent               = Teuchos::TimeMonitor::getNewCounter(prefix + "Concurrent portion");
-    RCP<Time> ApplyR                   = Teuchos::TimeMonitor::getNewCounter(prefix + "R: Computational Time");
-    RCP<Time> ApplyPbar                = Teuchos::TimeMonitor::getNewCounter(prefix + "Pbar: Computational Time");
-    RCP<Time> CompFine                 = Teuchos::TimeMonitor::getNewCounter(prefix + "Fine: Computational Time");
-    RCP<Time> CompCoarse               = Teuchos::TimeMonitor::getNewCounter(prefix + "Coarse: Computational Time");
-    RCP<Time> ApplySum                 = Teuchos::TimeMonitor::getNewCounter(prefix + "Sum: Computational Time");
-    RCP<Time> Synchronize_beginning    = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_beginning");
-    RCP<Time> Synchronize_center       = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_center");
-    RCP<Time> Synchronize_end          = Teuchos::TimeMonitor::getNewCounter(prefix + "Synchronize_end");
+  RCP<const Import> importer;
 
-    RCP<Level>          Fine = Levels_[0];
-    RCP<Level>          Coarse;
+  if (Levels_.size() > 1) {
+    Coarse = Levels_[1];
+    if (Coarse->IsAvailable("Importer"))
+      importer = Coarse->Get<RCP<const Import> >("Importer");
 
-    RCP<Operator> A = Fine->Get< RCP<Operator> >("A");
-    Teuchos::RCP< const Teuchos::Comm< int > > communicator = A->getDomainMap()->getComm();
+    R = Coarse->Get<RCP<Operator> >("R");
+    P = Coarse->Get<RCP<Operator> >("P");
 
+    // if(Coarse->IsAvailable("Pbar"))
+    Pbar = Coarse->Get<RCP<Operator> >("Pbar");
 
-    //Synchronize_beginning->start();
-    //communicator->barrier();
-    //Synchronize_beginning->stop();
+    coarseRhs = MultiVectorFactory::Build(R->getRangeMap(), B.getNumVectors(), true);
 
-    CompTime->start();
+    Ac = Coarse->Get<RCP<Operator> >("A");
 
-    SC one = STS::one(), zero = STS::zero();
+    ApplyR->start();
+    R->apply(B, *coarseRhs, Teuchos::NO_TRANS, one, zero);
+    // P->apply(B, *coarseRhs, Teuchos::TRANS, one, zero);
+    ApplyR->stop();
 
-    bool zeroGuess = InitialGuessIsZero;
+    if (doPRrebalance_ || importer.is_null()) {
+      coarseX = MultiVectorFactory::Build(coarseRhs->getMap(), X.getNumVectors(), true);
 
-    // =======   UPFRONT DEFINITION OF COARSE VARIABLES ===========
-
-    //RCP<const Map> origMap;
-    RCP< Operator > P;
-    RCP< Operator > Pbar;
-    RCP< Operator > R;
-    RCP< MultiVector > coarseRhs, coarseX;
-    RCP< Operator > Ac;
-    RCP<SmootherBase> preSmoo_coarse, postSmoo_coarse;
-    bool emptyCoarseSolve = true;
-    RCP<MultiVector> coarseX_prolonged = MultiVectorFactory::Build(X.getMap(), X.getNumVectors(), true);
+    } else {
+      RCP<TimeMonitor> ITime      = rcp(new TimeMonitor(*this, prefix + "Solve : import (total)", Timings0));
+      RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : import" + levelSuffix1, Timings0));
 
-    RCP<const Import> importer;
+      // Import: range map of R --> domain map of rebalanced Ac (before subcomm replacement)
+      RCP<MultiVector> coarseTmp = MultiVectorFactory::Build(importer->getTargetMap(), coarseRhs->getNumVectors());
+      coarseTmp->doImport(*coarseRhs, *importer, Xpetra::INSERT);
+      coarseRhs.swap(coarseTmp);
 
-    if (Levels_.size() > 1) {
-      Coarse = Levels_[1];
-      if (Coarse->IsAvailable("Importer"))
-        importer = Coarse->Get< RCP<const Import> >("Importer");
+      coarseX = MultiVectorFactory::Build(importer->getTargetMap(), X.getNumVectors(), true);
+    }
 
-      R = Coarse->Get< RCP<Operator> >("R");
-      P = Coarse->Get< RCP<Operator> >("P");
+    if (Coarse->IsAvailable("PreSmoother"))
+      preSmoo_coarse = Coarse->Get<RCP<SmootherBase> >("PreSmoother");
+    if (Coarse->IsAvailable("PostSmoother"))
+      postSmoo_coarse = Coarse->Get<RCP<SmootherBase> >("PostSmoother");
+  }
 
+  // ==========================================================
 
-      //if(Coarse->IsAvailable("Pbar"))
-      Pbar = Coarse->Get< RCP<Operator> >("Pbar");
+  MagnitudeType prevNorm = STS::magnitude(STS::one()), curNorm = STS::magnitude(STS::one());
+  rate_ = 1.0;
 
-      coarseRhs = MultiVectorFactory::Build(R->getRangeMap(), B.getNumVectors(), true);
+  for (LO i = 1; i <= nIts; i++) {
+#ifdef HAVE_MUELU_DEBUG
+    if (A->getDomainMap()->isCompatible(*(X.getMap())) == false) {
+      std::ostringstream ss;
+      ss << "Level " << startLevel << ": level A's domain map is not compatible with X";
+      throw Exceptions::Incompatible(ss.str());
+    }
 
-      Ac = Coarse->Get< RCP< Operator > >("A");
+    if (A->getRangeMap()->isCompatible(*(B.getMap())) == false) {
+      std::ostringstream ss;
+      ss << "Level " << startLevel << ": level A's range map is not compatible with B";
+      throw Exceptions::Incompatible(ss.str());
+    }
+#endif
+  }
 
-      ApplyR->start();
-      R->apply(B, *coarseRhs, Teuchos::NO_TRANS, one, zero);
-      //P->apply(B, *coarseRhs, Teuchos::TRANS, one, zero);
-      ApplyR->stop();
+  bool emptyFineSolve = true;
 
-      if (doPRrebalance_ || importer.is_null()) {
-        coarseX = MultiVectorFactory::Build(coarseRhs->getMap(), X.getNumVectors(), true);
+  RCP<MultiVector> fineX;
+  fineX = MultiVectorFactory::Build(X.getMap(), X.getNumVectors(), true);
 
-      } else {
+  // Synchronize_center->start();
+  // communicator->barrier();
+  // Synchronize_center->stop();
 
-        RCP<TimeMonitor> ITime      = rcp(new TimeMonitor(*this, prefix + "Solve : import (total)"       , Timings0));
-        RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : import" + levelSuffix1, Timings0));
+  Concurrent->start();
 
-        // Import: range map of R --> domain map of rebalanced Ac (before subcomm replacement)
-        RCP<MultiVector> coarseTmp = MultiVectorFactory::Build(importer->getTargetMap(), coarseRhs->getNumVectors());
-        coarseTmp->doImport(*coarseRhs, *importer, Xpetra::INSERT);
-        coarseRhs.swap(coarseTmp);
+  // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
+  if (Fine->IsAvailable("PreSmoother")) {
+    RCP<SmootherBase> preSmoo = Fine->Get<RCP<SmootherBase> >("PreSmoother");
+    CompFine->start();
+    preSmoo->Apply(*fineX, B, zeroGuess);
+    CompFine->stop();
+    emptyFineSolve = false;
+  }
+  if (Fine->IsAvailable("PostSmoother")) {
+    RCP<SmootherBase> postSmoo = Fine->Get<RCP<SmootherBase> >("PostSmoother");
+    CompFine->start();
+    postSmoo->Apply(*fineX, B, zeroGuess);
+    CompFine->stop();
 
-        coarseX = MultiVectorFactory::Build(importer->getTargetMap(), X.getNumVectors(), true);
-      }
+    emptyFineSolve = false;
+  }
+  if (emptyFineSolve == true) {
+    GetOStream(Warnings1) << "No fine grid smoother" << std::endl;
+    // Fine grid smoother is identity
+    fineX->update(one, B, zero);
+  }
 
-      if (Coarse->IsAvailable("PreSmoother"))
-        preSmoo_coarse = Coarse->Get< RCP<SmootherBase> >("PreSmoother");
-      if (Coarse->IsAvailable("PostSmoother"))
-        postSmoo_coarse = Coarse->Get< RCP<SmootherBase> >("PostSmoother");
+  if (Levels_.size() > 1) {
+    // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
+    if (Coarse->IsAvailable("PreSmoother")) {
+      CompCoarse->start();
+      preSmoo_coarse->Apply(*coarseX, *coarseRhs, zeroGuess);
+      CompCoarse->stop();
+      emptyCoarseSolve = false;
+    }
+    if (Coarse->IsAvailable("PostSmoother")) {
+      CompCoarse->start();
+      postSmoo_coarse->Apply(*coarseX, *coarseRhs, zeroGuess);
+      CompCoarse->stop();
+      emptyCoarseSolve = false;
+    }
+    if (emptyCoarseSolve == true) {
+      GetOStream(Warnings1) << "No coarse grid solver" << std::endl;
+      // Coarse operator is identity
+      coarseX->update(one, *coarseRhs, zero);
+    }
+    Concurrent->stop();
+    // Synchronize_end->start();
+    // communicator->barrier();
+    // Synchronize_end->stop();
+
+    if (!doPRrebalance_ && !importer.is_null()) {
+      RCP<TimeMonitor> ITime      = rcp(new TimeMonitor(*this, prefix + "Solve : export (total)", Timings0));
+      RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : export" + levelSuffix1, Timings0));
+
+      // Import: range map of rebalanced Ac (before subcomm replacement) --> domain map of P
+      RCP<MultiVector> coarseTmp = MultiVectorFactory::Build(importer->getSourceMap(), coarseX->getNumVectors());
+      coarseTmp->doExport(*coarseX, *importer, Xpetra::INSERT);
+      coarseX.swap(coarseTmp);
     }
 
-    // ==========================================================
-
+    ApplyPbar->start();
+    Pbar->apply(*coarseX, *coarseX_prolonged, Teuchos::NO_TRANS, one, zero);
+    ApplyPbar->stop();
+  }
 
-    MagnitudeType prevNorm = STS::magnitude(STS::one()), curNorm = STS::magnitude(STS::one());
-    rate_ = 1.0;
+  ApplySum->start();
+  X.update(1.0, *fineX, 1.0, *coarseX_prolonged, 0.0);
+  ApplySum->stop();
 
-    for (LO i = 1; i <= nIts; i++) {
-#ifdef HAVE_MUELU_DEBUG
-      if (A->getDomainMap()->isCompatible(*(X.getMap())) == false) {
-        std::ostringstream ss;
-        ss << "Level " << startLevel << ": level A's domain map is not compatible with X";
-        throw Exceptions::Incompatible(ss.str());
-      }
+  CompTime->stop();
 
-      if (A->getRangeMap()->isCompatible(*(B.getMap())) == false) {
-        std::ostringstream ss;
-        ss << "Level " << startLevel << ": level A's range map is not compatible with B";
-        throw Exceptions::Incompatible(ss.str());
-      }
-#endif
-    }
+  // communicator->barrier();
 
-   bool emptyFineSolve = true;
-
-   RCP< MultiVector > fineX;
-   fineX = MultiVectorFactory::Build(X.getMap(), X.getNumVectors(), true);
-
-   //Synchronize_center->start();
-   //communicator->barrier();
-   //Synchronize_center->stop();
-
-   Concurrent->start();
-
-   // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
-   if (Fine->IsAvailable("PreSmoother")) {
-     RCP<SmootherBase> preSmoo = Fine->Get< RCP<SmootherBase> >("PreSmoother");
-     CompFine->start();
-     preSmoo->Apply(*fineX, B, zeroGuess);
-     CompFine->stop();
-     emptyFineSolve = false;
-   }
-   if (Fine->IsAvailable("PostSmoother")) {
-     RCP<SmootherBase> postSmoo = Fine->Get< RCP<SmootherBase> >("PostSmoother");
-     CompFine->start();
-     postSmoo->Apply(*fineX, B, zeroGuess);
-     CompFine->stop();
-
-     emptyFineSolve = false;
-   }
-   if (emptyFineSolve == true) {
-     GetOStream(Warnings1) << "No fine grid smoother" << std::endl;
-     // Fine grid smoother is identity
-     fineX->update(one, B, zero);
-   }
-
-   if (Levels_.size() > 1) {
-     // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
-     if (Coarse->IsAvailable("PreSmoother")) {
-       CompCoarse->start();
-       preSmoo_coarse->Apply(*coarseX, *coarseRhs, zeroGuess);
-       CompCoarse->stop();
-       emptyCoarseSolve = false;
-
-     }
-     if (Coarse->IsAvailable("PostSmoother")) {
-       CompCoarse->start();
-       postSmoo_coarse->Apply(*coarseX, *coarseRhs, zeroGuess);
-       CompCoarse->stop();
-       emptyCoarseSolve = false;
-
-     }
-     if (emptyCoarseSolve == true) {
-       GetOStream(Warnings1) << "No coarse grid solver" << std::endl;
-       // Coarse operator is identity
-       coarseX->update(one, *coarseRhs, zero);
-     }
-     Concurrent->stop();
-     //Synchronize_end->start();
-     //communicator->barrier();
-     //Synchronize_end->stop();
-
-     if (!doPRrebalance_ && !importer.is_null()) {
-       RCP<TimeMonitor> ITime      = rcp(new TimeMonitor(*this, prefix + "Solve : export (total)"       , Timings0));
-       RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : export" + levelSuffix1, Timings0));
-
-       // Import: range map of rebalanced Ac (before subcomm replacement) --> domain map of P
-       RCP<MultiVector> coarseTmp = MultiVectorFactory::Build(importer->getSourceMap(), coarseX->getNumVectors());
-       coarseTmp->doExport(*coarseX, *importer, Xpetra::INSERT);
-       coarseX.swap(coarseTmp);
-     }
-
-     ApplyPbar->start();
-     Pbar->apply(*coarseX, *coarseX_prolonged, Teuchos::NO_TRANS, one, zero);
-     ApplyPbar->stop();
-   }
-
-   ApplySum->start();
-   X.update(1.0, *fineX, 1.0, *coarseX_prolonged, 0.0);
-   ApplySum->stop();
-
-   CompTime->stop();
-
-   //communicator->barrier();
-
-   return (tol > 0 ? ConvergenceStatus::Unconverged : ConvergenceStatus::Undefined);
+  return (tol > 0 ? ConvergenceStatus::Unconverged : ConvergenceStatus::Undefined);
 }
 #else
-  // ---------------------------------------- Iterate -------------------------------------------------------
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const MultiVector& B, MultiVector& X, ConvData conv,
-                                                                           bool InitialGuessIsZero, LO startLevel) {
-    LO            nIts = conv.maxIts_;
-    MagnitudeType tol  = conv.tol_;
-
-    // These timers work as follows. "iterateTime" records total time spent in
-    // iterate. "levelTime" records time on a per level basis. The label is
-    // crafted to mimic the per-level messages used in Monitors. Note that a
-    // given level is timed with a TimeMonitor instead of a Monitor or
-    // SubMonitor. This is mainly because I want to time each level
-    // separately, and Monitors/SubMonitors print "(total) xx yy zz" ,
-    // "(sub,total) xx yy zz", respectively, which is subject to
-    // misinterpretation.  The per-level TimeMonitors are stopped/started
-    // manually before/after a recursive call to Iterate. A side artifact to
-    // this approach is that the counts for intermediate level timers are twice
-    // the counts for the finest and coarsest levels.
-
-    RCP<Level>    Fine = Levels_[startLevel];
-
-    std::string label        = FormattingHelper::getColonLabel(Fine->getObjectLabel());
-    std::string prefix       = label + this->ShortClassName() + ": ";
-    std::string levelSuffix  = " (level=" + toString(startLevel) + ")";
-    std::string levelSuffix1 = " (level=" + toString(startLevel+1) + ")";
-
-    bool useStackedTimer = !Teuchos::TimeMonitor::getStackedTimer().is_null();
-
-    RCP<Monitor>     iterateTime;
-    RCP<TimeMonitor> iterateTime1;
-    if (startLevel == 0)
-      iterateTime  = rcp(new Monitor(*this, "Solve", label, (nIts == 1) ? None : Runtime0, Timings0));
-    else if (!useStackedTimer)
-      iterateTime1 = rcp(new TimeMonitor(*this, prefix + "Solve (total, level=" + toString(startLevel) + ")", Timings0));
-
-    std::string iterateLevelTimeLabel = prefix + "Solve" + levelSuffix;
-    RCP<TimeMonitor> iterateLevelTime = rcp(new TimeMonitor(*this, iterateLevelTimeLabel, Timings0));
-
-    bool zeroGuess = InitialGuessIsZero;
-
-    RCP<Operator> A    = Fine->Get< RCP<Operator> >("A");
-    using namespace Teuchos;
-    RCP<Time> CompCoarse  = Teuchos::TimeMonitor::getNewCounter(prefix + "Coarse: Computational Time");
-
-    if (A.is_null()) {
-      // This processor does not have any data for this process on coarser
-      // levels. This can only happen when there are multiple processors and
-      // we use repartitioning.
-      return ConvergenceStatus::Undefined;
-    }
+// ---------------------------------------- Iterate -------------------------------------------------------
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const MultiVector& B, MultiVector& X, ConvData conv,
+                                                                                bool InitialGuessIsZero, LO startLevel) {
+  LO nIts           = conv.maxIts_;
+  MagnitudeType tol = conv.tol_;
+
+  // These timers work as follows. "iterateTime" records total time spent in
+  // iterate. "levelTime" records time on a per level basis. The label is
+  // crafted to mimic the per-level messages used in Monitors. Note that a
+  // given level is timed with a TimeMonitor instead of a Monitor or
+  // SubMonitor. This is mainly because I want to time each level
+  // separately, and Monitors/SubMonitors print "(total) xx yy zz" ,
+  // "(sub,total) xx yy zz", respectively, which is subject to
+  // misinterpretation.  The per-level TimeMonitors are stopped/started
+  // manually before/after a recursive call to Iterate. A side artifact to
+  // this approach is that the counts for intermediate level timers are twice
+  // the counts for the finest and coarsest levels.
+
+  RCP<Level> Fine = Levels_[startLevel];
+
+  std::string label        = FormattingHelper::getColonLabel(Fine->getObjectLabel());
+  std::string prefix       = label + this->ShortClassName() + ": ";
+  std::string levelSuffix  = " (level=" + toString(startLevel) + ")";
+  std::string levelSuffix1 = " (level=" + toString(startLevel + 1) + ")";
+
+  bool useStackedTimer = !Teuchos::TimeMonitor::getStackedTimer().is_null();
+
+  RCP<Monitor> iterateTime;
+  RCP<TimeMonitor> iterateTime1;
+  if (startLevel == 0)
+    iterateTime = rcp(new Monitor(*this, "Solve", label, (nIts == 1) ? None : Runtime0, Timings0));
+  else if (!useStackedTimer)
+    iterateTime1 = rcp(new TimeMonitor(*this, prefix + "Solve (total, level=" + toString(startLevel) + ")", Timings0));
+
+  std::string iterateLevelTimeLabel = prefix + "Solve" + levelSuffix;
+  RCP<TimeMonitor> iterateLevelTime = rcp(new TimeMonitor(*this, iterateLevelTimeLabel, Timings0));
+
+  bool zeroGuess = InitialGuessIsZero;
+
+  RCP<Operator> A = Fine->Get<RCP<Operator> >("A");
+  using namespace Teuchos;
+  RCP<Time> CompCoarse = Teuchos::TimeMonitor::getNewCounter(prefix + "Coarse: Computational Time");
+
+  if (A.is_null()) {
+    // This processor does not have any data for this process on coarser
+    // levels. This can only happen when there are multiple processors and
+    // we use repartitioning.
+    return ConvergenceStatus::Undefined;
+  }
 
-    // If we switched the number of vectors, we'd need to reallocate here.
-    // If the number of vectors is unchanged, this is a noop.
-    // NOTE: We need to check against B because the tests in AllocateLevelMultiVectors
-    // will fail on Stokhos Scalar types (due to the so-called 'hidden dimension')
-    const BlockedMultiVector * Bblocked = dynamic_cast<const BlockedMultiVector*>(&B);
-    if(residual_.size() > startLevel &&
-       ( ( Bblocked && !Bblocked->isSameSize(*residual_[startLevel])) ||
-         (!Bblocked && !residual_[startLevel]->isSameSize(B))))
-      DeleteLevelMultiVectors();
-    AllocateLevelMultiVectors(X.getNumVectors());
-
-    // Print residual information before iterating
-    typedef Teuchos::ScalarTraits<typename STS::magnitudeType> STM;
-    MagnitudeType prevNorm = STM::one();
-    rate_ = 1.0;
-    if (IsCalculationOfResidualRequired(startLevel, conv)) {
-      ConvergenceStatus convergenceStatus = ComputeResidualAndPrintHistory(*A, X, B, Teuchos::ScalarTraits<LO>::zero(), startLevel, conv, prevNorm);
-      if (convergenceStatus == MueLu::ConvergenceStatus::Converged)
-        return convergenceStatus;
-    }
+  // If we switched the number of vectors, we'd need to reallocate here.
+  // If the number of vectors is unchanged, this is a noop.
+  // NOTE: We need to check against B because the tests in AllocateLevelMultiVectors
+  // will fail on Stokhos Scalar types (due to the so-called 'hidden dimension')
+  const BlockedMultiVector* Bblocked = dynamic_cast<const BlockedMultiVector*>(&B);
+  if (residual_.size() > startLevel &&
+      ((Bblocked && !Bblocked->isSameSize(*residual_[startLevel])) ||
+       (!Bblocked && !residual_[startLevel]->isSameSize(B))))
+    DeleteLevelMultiVectors();
+  AllocateLevelMultiVectors(X.getNumVectors());
+
+  // Print residual information before iterating
+  typedef Teuchos::ScalarTraits<typename STS::magnitudeType> STM;
+  MagnitudeType prevNorm = STM::one();
+  rate_                  = 1.0;
+  if (IsCalculationOfResidualRequired(startLevel, conv)) {
+    ConvergenceStatus convergenceStatus = ComputeResidualAndPrintHistory(*A, X, B, Teuchos::ScalarTraits<LO>::zero(), startLevel, conv, prevNorm);
+    if (convergenceStatus == MueLu::ConvergenceStatus::Converged)
+      return convergenceStatus;
+  }
 
-    SC one = STS::one(), zero = STS::zero();
-    for (LO iteration = 1; iteration <= nIts; iteration++) {
+  SC one = STS::one(), zero = STS::zero();
+  for (LO iteration = 1; iteration <= nIts; iteration++) {
 #ifdef HAVE_MUELU_DEBUG
-#if 0 // TODO fix me
+#if 0  // TODO fix me
       if (A->getDomainMap()->isCompatible(*(X.getMap())) == false) {
         std::ostringstream ss;
         ss << "Level " << startLevel << ": level A's domain map is not compatible with X";
@@ -975,638 +979,651 @@ namespace MueLu {
 #endif
 #endif
 
-      if (startLevel == as<LO>(Levels_.size())-1) {
-        // On the coarsest level, we do either smoothing (if defined) or a direct solve.
-        RCP<TimeMonitor> CLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : coarse" + levelSuffix, Timings0));
+    if (startLevel == as<LO>(Levels_.size()) - 1) {
+      // On the coarsest level, we do either smoothing (if defined) or a direct solve.
+      RCP<TimeMonitor> CLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : coarse" + levelSuffix, Timings0));
+
+      bool emptySolve = true;
 
-        bool emptySolve = true;
+      // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
+      if (Fine->IsAvailable("PreSmoother")) {
+        RCP<SmootherBase> preSmoo = Fine->Get<RCP<SmootherBase> >("PreSmoother");
+        CompCoarse->start();
+        preSmoo->Apply(X, B, zeroGuess);
+        CompCoarse->stop();
+        zeroGuess  = false;
+        emptySolve = false;
+      }
+      if (Fine->IsAvailable("PostSmoother")) {
+        RCP<SmootherBase> postSmoo = Fine->Get<RCP<SmootherBase> >("PostSmoother");
+        CompCoarse->start();
+        postSmoo->Apply(X, B, zeroGuess);
+        CompCoarse->stop();
+        emptySolve = false;
+        zeroGuess  = false;
+      }
+      if (emptySolve == true) {
+        GetOStream(Warnings1) << "No coarse grid solver" << std::endl;
+        // Coarse operator is identity
+        X.update(one, B, zero);
+      }
+
+    } else {
+      // On intermediate levels, we do cycles
+      RCP<Level> Coarse = Levels_[startLevel + 1];
+      {
+        // ============== PRESMOOTHING ==============
+        RCP<TimeMonitor> STime;
+        if (!useStackedTimer)
+          STime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing (total)", Timings0));
+        RCP<TimeMonitor> SLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing" + levelSuffix, Timings0));
 
-        // NOTE: we need to check using IsAvailable before Get here to avoid building default smoother
         if (Fine->IsAvailable("PreSmoother")) {
-          RCP<SmootherBase> preSmoo = Fine->Get< RCP<SmootherBase> >("PreSmoother");
-          CompCoarse->start();
+          RCP<SmootherBase> preSmoo = Fine->Get<RCP<SmootherBase> >("PreSmoother");
           preSmoo->Apply(X, B, zeroGuess);
-          CompCoarse->stop();
-          zeroGuess  = false;
-          emptySolve = false;
-        }
-        if (Fine->IsAvailable("PostSmoother")) {
-          RCP<SmootherBase> postSmoo = Fine->Get< RCP<SmootherBase> >("PostSmoother");
-          CompCoarse->start();
-          postSmoo->Apply(X, B, zeroGuess);
-          CompCoarse->stop();
-          emptySolve = false;
-          zeroGuess  = false;
-        }
-        if (emptySolve == true) {
-          GetOStream(Warnings1) << "No coarse grid solver" << std::endl;
-          // Coarse operator is identity
-          X.update(one, B, zero);
+          zeroGuess = false;
         }
+      }
 
-      } else {
-        // On intermediate levels, we do cycles
-        RCP<Level> Coarse = Levels_[startLevel+1];
-        {
-          // ============== PRESMOOTHING ==============
-          RCP<TimeMonitor> STime;
-          if (!useStackedTimer)
-            STime                     = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing (total)"      , Timings0));
-          RCP<TimeMonitor> SLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing" + levelSuffix, Timings0));
-
-          if (Fine->IsAvailable("PreSmoother")) {
-            RCP<SmootherBase> preSmoo = Fine->Get< RCP<SmootherBase> >("PreSmoother");
-            preSmoo->Apply(X, B, zeroGuess);
-            zeroGuess  = false;
-          }
+      RCP<MultiVector> residual;
+      {
+        RCP<TimeMonitor> ATime;
+        if (!useStackedTimer)
+          ATime = rcp(new TimeMonitor(*this, prefix + "Solve : residual calculation (total)", Timings0));
+        RCP<TimeMonitor> ALevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : residual calculation" + levelSuffix, Timings0));
+        if (zeroGuess) {
+          // If there's a pre-smoother, then zeroGuess is false.  If there isn't and people still have zeroGuess set,
+          // then then X still has who knows what, so we need to zero it before we go to the coarse grid.
+          X.putScalar(zero);
         }
 
-        RCP<MultiVector> residual;
-        {
-          RCP<TimeMonitor> ATime;
-          if (!useStackedTimer)
-            ATime                     = rcp(new TimeMonitor(*this, prefix + "Solve : residual calculation (total)"      , Timings0));
-          RCP<TimeMonitor> ALevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : residual calculation" + levelSuffix, Timings0));
-          if (zeroGuess) {
-            // If there's a pre-smoother, then zeroGuess is false.  If there isn't and people still have zeroGuess set,
-            // then then X still has who knows what, so we need to zero it before we go to the coarse grid.
-            X.putScalar(zero);
-          }
-
-          Utilities::Residual(*A, X, B,*residual_[startLevel]);
-          residual = residual_[startLevel];
-        }
+        Utilities::Residual(*A, X, B, *residual_[startLevel]);
+        residual = residual_[startLevel];
+      }
 
-        RCP<Operator>    P = Coarse->Get< RCP<Operator> >("P");
-        if (Coarse->IsAvailable("Pbar"))
-           P = Coarse->Get< RCP<Operator> >("Pbar");
+      RCP<Operator> P = Coarse->Get<RCP<Operator> >("P");
+      if (Coarse->IsAvailable("Pbar"))
+        P = Coarse->Get<RCP<Operator> >("Pbar");
 
-        RCP<MultiVector> coarseRhs, coarseX;
-        //        const bool initializeWithZeros = true;
-        {
-          // ============== RESTRICTION ==============
-          RCP<TimeMonitor> RTime;
-          if (!useStackedTimer)
-            RTime                     = rcp(new TimeMonitor(*this, prefix + "Solve : restriction (total)"      , Timings0));
-          RCP<TimeMonitor> RLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : restriction" + levelSuffix, Timings0));
-          coarseRhs = coarseRhs_[startLevel];
-
-          if (implicitTranspose_) {
-            P->apply(*residual, *coarseRhs, Teuchos::TRANS, one, zero);
-
-          } else {
-            RCP<Operator> R = Coarse->Get< RCP<Operator> >("R");
-            R->apply(*residual, *coarseRhs, Teuchos::NO_TRANS, one, zero);
-          }
-        }
+      RCP<MultiVector> coarseRhs, coarseX;
+      //        const bool initializeWithZeros = true;
+      {
+        // ============== RESTRICTION ==============
+        RCP<TimeMonitor> RTime;
+        if (!useStackedTimer)
+          RTime = rcp(new TimeMonitor(*this, prefix + "Solve : restriction (total)", Timings0));
+        RCP<TimeMonitor> RLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : restriction" + levelSuffix, Timings0));
+        coarseRhs                   = coarseRhs_[startLevel];
+
+        if (implicitTranspose_) {
+          P->apply(*residual, *coarseRhs, Teuchos::TRANS, one, zero);
 
-        RCP<const Import> importer;
-        if (Coarse->IsAvailable("Importer"))
-          importer = Coarse->Get< RCP<const Import> >("Importer");
-
-        coarseX = coarseX_[startLevel];
-        if (!doPRrebalance_ && !importer.is_null()) {
-          RCP<TimeMonitor> ITime;
-          if (!useStackedTimer)
-            ITime                     = rcp(new TimeMonitor(*this, prefix + "Solve : import (total)"       , Timings0));
-          RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : import" + levelSuffix1, Timings0));
-
-          // Import: range map of R --> domain map of rebalanced Ac (before subcomm replacement)
-          RCP<MultiVector> coarseTmp = coarseImport_[startLevel];
-          coarseTmp->doImport(*coarseRhs, *importer, Xpetra::INSERT);
-          coarseRhs.swap(coarseTmp);
+        } else {
+          RCP<Operator> R = Coarse->Get<RCP<Operator> >("R");
+          R->apply(*residual, *coarseRhs, Teuchos::NO_TRANS, one, zero);
         }
+      }
 
-        RCP<Operator> Ac = Coarse->Get< RCP<Operator> >("A");
-        if (!Ac.is_null()) {
-          RCP<const Map> origXMap = coarseX->getMap();
-          RCP<const Map> origRhsMap = coarseRhs->getMap();
+      RCP<const Import> importer;
+      if (Coarse->IsAvailable("Importer"))
+        importer = Coarse->Get<RCP<const Import> >("Importer");
+
+      coarseX = coarseX_[startLevel];
+      if (!doPRrebalance_ && !importer.is_null()) {
+        RCP<TimeMonitor> ITime;
+        if (!useStackedTimer)
+          ITime = rcp(new TimeMonitor(*this, prefix + "Solve : import (total)", Timings0));
+        RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : import" + levelSuffix1, Timings0));
 
-          // Replace maps with maps with a subcommunicator
-          coarseRhs->replaceMap(Ac->getRangeMap());
-          coarseX  ->replaceMap(Ac->getDomainMap());
+        // Import: range map of R --> domain map of rebalanced Ac (before subcomm replacement)
+        RCP<MultiVector> coarseTmp = coarseImport_[startLevel];
+        coarseTmp->doImport(*coarseRhs, *importer, Xpetra::INSERT);
+        coarseRhs.swap(coarseTmp);
+      }
 
-          {
-            iterateLevelTime = Teuchos::null; // stop timing this level
+      RCP<Operator> Ac = Coarse->Get<RCP<Operator> >("A");
+      if (!Ac.is_null()) {
+        RCP<const Map> origXMap   = coarseX->getMap();
+        RCP<const Map> origRhsMap = coarseRhs->getMap();
 
-            Iterate(*coarseRhs, *coarseX, 1, true, startLevel+1);
-            // ^^ zero initial guess
-            if (Cycle_ == WCYCLE && WCycleStartLevel_ >= startLevel)
-              Iterate(*coarseRhs, *coarseX, 1, false, startLevel+1);
-            // ^^ nonzero initial guess
+        // Replace maps with maps with a subcommunicator
+        coarseRhs->replaceMap(Ac->getRangeMap());
+        coarseX->replaceMap(Ac->getDomainMap());
 
-            iterateLevelTime = rcp(new TimeMonitor(*this, iterateLevelTimeLabel));  // restart timing this level
-          }
-          coarseX->replaceMap(origXMap);
-          coarseRhs->replaceMap(origRhsMap);
-        }
+        {
+          iterateLevelTime = Teuchos::null;  // stop timing this level
 
-        if (!doPRrebalance_ && !importer.is_null()) {
-          RCP<TimeMonitor> ITime;
-          if (!useStackedTimer)
-            ITime                     = rcp(new TimeMonitor(*this, prefix + "Solve : export (total)"      ,  Timings0));
-          RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : export" + levelSuffix1, Timings0));
+          Iterate(*coarseRhs, *coarseX, 1, true, startLevel + 1);
+          // ^^ zero initial guess
+          if (Cycle_ == WCYCLE && WCycleStartLevel_ >= startLevel)
+            Iterate(*coarseRhs, *coarseX, 1, false, startLevel + 1);
+          // ^^ nonzero initial guess
 
-          // Import: range map of rebalanced Ac (before subcomm replacement) --> domain map of P
-          RCP<MultiVector> coarseTmp = coarseExport_[startLevel];
-          coarseTmp->doExport(*coarseX, *importer, Xpetra::INSERT);
-          coarseX.swap(coarseTmp);
+          iterateLevelTime = rcp(new TimeMonitor(*this, iterateLevelTimeLabel));  // restart timing this level
         }
+        coarseX->replaceMap(origXMap);
+        coarseRhs->replaceMap(origRhsMap);
+      }
 
-        {
-          // ============== PROLONGATION ==============
-          RCP<TimeMonitor> PTime;
-          if (!useStackedTimer)
-            PTime                     = rcp(new TimeMonitor(*this, prefix + "Solve : prolongation (total)"      , Timings0));
-          RCP<TimeMonitor> PLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : prolongation" + levelSuffix, Timings0));
-          // Update X += P * coarseX
-          // Note that due to what may be round-off error accumulation, use of the fused kernel
-          //    P->apply(*coarseX, X, Teuchos::NO_TRANS, one, one);
-          // can in some cases result in slightly higher iteration counts.
-          if (fuseProlongationAndUpdate_) {
-            P->apply(*coarseX, X, Teuchos::NO_TRANS, scalingFactor_, one);
-          } else {
-            RCP<MultiVector> correction = correction_[startLevel];
-            P->apply(*coarseX, *correction, Teuchos::NO_TRANS, one, zero);
-            X.update(scalingFactor_, *correction, one);
-          }
-        }
+      if (!doPRrebalance_ && !importer.is_null()) {
+        RCP<TimeMonitor> ITime;
+        if (!useStackedTimer)
+          ITime = rcp(new TimeMonitor(*this, prefix + "Solve : export (total)", Timings0));
+        RCP<TimeMonitor> ILevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : export" + levelSuffix1, Timings0));
 
-        {
-          // ============== POSTSMOOTHING ==============
-          RCP<TimeMonitor> STime;
-          if (!useStackedTimer)
-            STime                     = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing (total)"      , Timings0));
-          RCP<TimeMonitor> SLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing" + levelSuffix, Timings0));
-
-          if (Fine->IsAvailable("PostSmoother")) {
-            RCP<SmootherBase> postSmoo = Fine->Get< RCP<SmootherBase> >("PostSmoother");
-            postSmoo->Apply(X, B, false);
-          }
+        // Import: range map of rebalanced Ac (before subcomm replacement) --> domain map of P
+        RCP<MultiVector> coarseTmp = coarseExport_[startLevel];
+        coarseTmp->doExport(*coarseX, *importer, Xpetra::INSERT);
+        coarseX.swap(coarseTmp);
+      }
+
+      {
+        // ============== PROLONGATION ==============
+        RCP<TimeMonitor> PTime;
+        if (!useStackedTimer)
+          PTime = rcp(new TimeMonitor(*this, prefix + "Solve : prolongation (total)", Timings0));
+        RCP<TimeMonitor> PLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : prolongation" + levelSuffix, Timings0));
+        // Update X += P * coarseX
+        // Note that due to what may be round-off error accumulation, use of the fused kernel
+        //    P->apply(*coarseX, X, Teuchos::NO_TRANS, one, one);
+        // can in some cases result in slightly higher iteration counts.
+        if (fuseProlongationAndUpdate_) {
+          P->apply(*coarseX, X, Teuchos::NO_TRANS, scalingFactor_, one);
+        } else {
+          RCP<MultiVector> correction = correction_[startLevel];
+          P->apply(*coarseX, *correction, Teuchos::NO_TRANS, one, zero);
+          X.update(scalingFactor_, *correction, one);
         }
       }
-      zeroGuess = false;
 
+      {
+        // ============== POSTSMOOTHING ==============
+        RCP<TimeMonitor> STime;
+        if (!useStackedTimer)
+          STime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing (total)", Timings0));
+        RCP<TimeMonitor> SLevelTime = rcp(new TimeMonitor(*this, prefix + "Solve : smoothing" + levelSuffix, Timings0));
 
-      if (IsCalculationOfResidualRequired(startLevel, conv)) {
-        ConvergenceStatus convergenceStatus = ComputeResidualAndPrintHistory(*A, X, B, iteration, startLevel, conv, prevNorm);
-        if (convergenceStatus == MueLu::ConvergenceStatus::Converged)
-          return convergenceStatus;
+        if (Fine->IsAvailable("PostSmoother")) {
+          RCP<SmootherBase> postSmoo = Fine->Get<RCP<SmootherBase> >("PostSmoother");
+          postSmoo->Apply(X, B, false);
+        }
       }
     }
-    return (tol > 0 ? ConvergenceStatus::Unconverged : ConvergenceStatus::Undefined);
+    zeroGuess = false;
+
+    if (IsCalculationOfResidualRequired(startLevel, conv)) {
+      ConvergenceStatus convergenceStatus = ComputeResidualAndPrintHistory(*A, X, B, iteration, startLevel, conv, prevNorm);
+      if (convergenceStatus == MueLu::ConvergenceStatus::Converged)
+        return convergenceStatus;
+    }
   }
+  return (tol > 0 ? ConvergenceStatus::Unconverged : ConvergenceStatus::Undefined);
+}
 #endif
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(const LO& start, const LO& end, const std::string& suffix) {
+  LO startLevel = (start != -1 ? start : 0);
+  LO endLevel   = (end != -1 ? end : Levels_.size() - 1);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(startLevel > endLevel, Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Write : startLevel must be <= endLevel");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(startLevel < 0 || endLevel >= Levels_.size(), Exceptions::RuntimeError,
+                             "MueLu::Hierarchy::Write bad start or end level");
+
+  for (LO i = startLevel; i < endLevel + 1; i++) {
+    RCP<Matrix> A = rcp_dynamic_cast<Matrix>(Levels_[i]->template Get<RCP<Operator> >("A")), P, R;
+    if (i > 0) {
+      P = rcp_dynamic_cast<Matrix>(Levels_[i]->template Get<RCP<Operator> >("P"));
+      if (!implicitTranspose_)
+        R = rcp_dynamic_cast<Matrix>(Levels_[i]->template Get<RCP<Operator> >("R"));
+    }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(const LO& start, const LO& end, const std::string &suffix) {
-    LO startLevel = (start != -1 ? start : 0);
-    LO   endLevel = (end   != -1 ? end   : Levels_.size()-1);
+    if (!A.is_null()) Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A_" + toString(i) + suffix + ".m", *A);
+    if (!P.is_null()) {
+      Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("P_" + toString(i) + suffix + ".m", *P);
+    }
+    if (!R.is_null()) {
+      Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("R_" + toString(i) + suffix + ".m", *R);
+    }
+  }
+}
 
-    TEUCHOS_TEST_FOR_EXCEPTION(startLevel > endLevel, Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Write : startLevel must be <= endLevel");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Keep(const std::string& ename, const FactoryBase* factory) {
+  for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
+    (*it)->Keep(ename, factory);
+}
 
-    TEUCHOS_TEST_FOR_EXCEPTION(startLevel < 0 || endLevel >= Levels_.size(), Exceptions::RuntimeError,
-                               "MueLu::Hierarchy::Write bad start or end level");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Delete(const std::string& ename, const FactoryBase* factory) {
+  for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
+    (*it)->Delete(ename, factory);
+}
 
-    for (LO i = startLevel; i < endLevel + 1; i++) {
-      RCP<Matrix> A = rcp_dynamic_cast<Matrix>(Levels_[i]-> template Get< RCP< Operator> >("A")), P, R;
-      if (i > 0) {
-        P = rcp_dynamic_cast<Matrix>(Levels_[i]-> template Get< RCP< Operator> >("P"));
-        if (!implicitTranspose_)
-          R = rcp_dynamic_cast<Matrix>(Levels_[i]-> template Get< RCP< Operator> >("R"));
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
+  for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
+    (*it)->AddKeepFlag(ename, factory, keep);
+}
 
-      if (!A.is_null()) Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A_" + toString(i) + suffix + ".m", *A);
-      if (!P.is_null()) {
-        Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("P_" + toString(i) + suffix + ".m", *P);
-      }
-      if (!R.is_null()) {
-        Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("R_" + toString(i) + suffix + ".m", *R);
-      }
-    }
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RemoveKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
+  for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
+    (*it)->RemoveKeepFlag(ename, factory, keep);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Keep(const std::string & ename, const FactoryBase* factory) {
-    for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
-      (*it)->Keep(ename, factory);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  if (description_.empty()) {
+    std::ostringstream out;
+    out << BaseClass::description();
+    out << "{#levels = " << GetGlobalNumLevels() << ", complexity = " << GetOperatorComplexity() << "}";
+    description_ = out.str();
   }
+  return description_;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Delete(const std::string& ename, const FactoryBase* factory) {
-    for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
-      (*it)->Delete(ename, factory);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel tVerbLevel) const {
+  describe(out, toMueLuVerbLevel(tVerbLevel));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddKeepFlag(const std::string & ename, const FactoryBase* factory, KeepType keep) {
-    for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
-      (*it)->AddKeepFlag(ename, factory, keep);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  RCP<Operator> A0                    = Levels_[0]->template Get<RCP<Operator> >("A");
+  RCP<const Teuchos::Comm<int> > comm = A0->getDomainMap()->getComm();
+
+  int numLevels    = GetNumLevels();
+  RCP<Operator> Ac = Levels_[numLevels - 1]->template Get<RCP<Operator> >("A");
+  if (Ac.is_null()) {
+    // It may happen that we do repartition on the last level, but the matrix
+    // is small enough to satisfy "max coarse size" requirement. Then, even
+    // though we have the level, the matrix would be null on all but one processors
+    numLevels--;
   }
+  int root = comm->getRank();
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RemoveKeepFlag(const std::string & ename, const FactoryBase* factory, KeepType keep) {
-    for (Array<RCP<Level> >::iterator it = Levels_.begin(); it != Levels_.end(); ++it)
-      (*it)->RemoveKeepFlag(ename, factory, keep);
-  }
+#ifdef HAVE_MPI
+  int smartData = numLevels * comm->getSize() + comm->getRank(), maxSmartData;
+  reduceAll(*comm, Teuchos::REDUCE_MAX, smartData, Teuchos::ptr(&maxSmartData));
+  root = maxSmartData % comm->getSize();
+#endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    if ( description_.empty() )
-    {
-      std::ostringstream out;
-      out << BaseClass::description();
-      out << "{#levels = " << GetGlobalNumLevels() << ", complexity = " << GetOperatorComplexity() << "}";
-      description_ = out.str();
-    }
-    return description_;
-  }
+  // Compute smoother complexity, if needed
+  double smoother_comp = -1.0;
+  if (verbLevel & (Statistics0 | Test))
+    smoother_comp = GetSmootherComplexity();
+
+  std::string outstr;
+  if (comm->getRank() == root && verbLevel & (Statistics0 | Test)) {
+    std::vector<Xpetra::global_size_t> nnzPerLevel;
+    std::vector<Xpetra::global_size_t> rowsPerLevel;
+    std::vector<int> numProcsPerLevel;
+    bool someOpsNotMatrices             = false;
+    const Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+    for (int i = 0; i < numLevels; i++) {
+      TEUCHOS_TEST_FOR_EXCEPTION(!(Levels_[i]->IsAvailable("A")), Exceptions::RuntimeError,
+                                 "Operator A is not available on level " << i);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel tVerbLevel) const {
-    describe(out, toMueLuVerbLevel(tVerbLevel));
-  }
+      RCP<Operator> A = Levels_[i]->template Get<RCP<Operator> >("A");
+      TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::RuntimeError,
+                                 "Operator A on level " << i << " is null.");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    RCP<Operator> A0 = Levels_[0]->template Get<RCP<Operator> >("A");
-    RCP<const Teuchos::Comm<int> > comm = A0->getDomainMap()->getComm();
-
-    int numLevels = GetNumLevels();
-    RCP<Operator> Ac = Levels_[numLevels-1]->template Get<RCP<Operator> >("A");
-    if (Ac.is_null()) {
-      // It may happen that we do repartition on the last level, but the matrix
-      // is small enough to satisfy "max coarse size" requirement. Then, even
-      // though we have the level, the matrix would be null on all but one processors
-      numLevels--;
+      RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
+      if (Am.is_null()) {
+        someOpsNotMatrices = true;
+        nnzPerLevel.push_back(INVALID);
+        rowsPerLevel.push_back(A->getDomainMap()->getGlobalNumElements());
+        numProcsPerLevel.push_back(A->getDomainMap()->getComm()->getSize());
+      } else {
+        LO storageblocksize       = Am->GetStorageBlockSize();
+        Xpetra::global_size_t nnz = Am->getGlobalNumEntries() * storageblocksize * storageblocksize;
+        nnzPerLevel.push_back(nnz);
+        rowsPerLevel.push_back(Am->getGlobalNumRows() * storageblocksize);
+        numProcsPerLevel.push_back(Am->getRowMap()->getComm()->getSize());
+      }
     }
-    int root = comm->getRank();
+    if (someOpsNotMatrices)
+      GetOStream(Warnings0) << "Some level operators are not matrices, statistics calculation are incomplete" << std::endl;
 
-#ifdef HAVE_MPI
-    int smartData = numLevels*comm->getSize() + comm->getRank(), maxSmartData;
-    reduceAll(*comm, Teuchos::REDUCE_MAX, smartData, Teuchos::ptr(&maxSmartData));
-    root = maxSmartData % comm->getSize();
-#endif
+    {
+      std::string label = Levels_[0]->getObjectLabel();
+      std::ostringstream oss;
+      oss << std::setfill(' ');
+      oss << "\n--------------------------------------------------------------------------------\n";
+      oss << "---                            Multigrid Summary " << std::setw(28) << std::left << label << "---\n";
+      oss << "--------------------------------------------------------------------------------" << std::endl;
+      if (verbLevel & Parameters1)
+        oss << "Scalar              = " << Teuchos::ScalarTraits<Scalar>::name() << std::endl;
+      oss << "Number of levels    = " << numLevels << std::endl;
+      oss << "Operator complexity = " << std::setprecision(2) << std::setiosflags(std::ios::fixed);
+      if (!someOpsNotMatrices)
+        oss << GetOperatorComplexity() << std::endl;
+      else
+        oss << "not available (Some operators in hierarchy are not matrices.)" << std::endl;
+
+      if (smoother_comp != -1.0) {
+        oss << "Smoother complexity = " << std::setprecision(2) << std::setiosflags(std::ios::fixed)
+            << smoother_comp << std::endl;
+      }
 
-    // Compute smoother complexity, if needed
-    double smoother_comp =-1.0;
-    if (verbLevel & (Statistics0 | Test))
-      smoother_comp = GetSmootherComplexity();
-
-    std::string outstr;
-    if (comm->getRank() == root && verbLevel & (Statistics0 | Test)) {
-      std::vector<Xpetra::global_size_t> nnzPerLevel;
-      std::vector<Xpetra::global_size_t> rowsPerLevel;
-      std::vector<int>                   numProcsPerLevel;
-      bool someOpsNotMatrices = false;
-      const Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-      for (int i = 0; i < numLevels; i++) {
-        TEUCHOS_TEST_FOR_EXCEPTION(!(Levels_[i]->IsAvailable("A")) , Exceptions::RuntimeError,
-                                   "Operator A is not available on level " << i);
-
-        RCP<Operator> A = Levels_[i]->template Get<RCP<Operator> >("A");
-        TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::RuntimeError,
-                                   "Operator A on level " << i << " is null.");
-
-        RCP<Matrix> Am = rcp_dynamic_cast<Matrix>(A);
-        if (Am.is_null()) {
-          someOpsNotMatrices = true;
-          nnzPerLevel     .push_back(INVALID);
-          rowsPerLevel    .push_back(A->getDomainMap()->getGlobalNumElements());
-          numProcsPerLevel.push_back(A->getDomainMap()->getComm()->getSize());
+      switch (Cycle_) {
+        case VCYCLE:
+          oss << "Cycle type          = V" << std::endl;
+          break;
+        case WCYCLE:
+          oss << "Cycle type          = W" << std::endl;
+          if (WCycleStartLevel_ > 0)
+            oss << "Cycle start level   = " << WCycleStartLevel_ << std::endl;
+          break;
+        default:
+          break;
+      };
+      oss << std::endl;
+
+      Xpetra::global_size_t tt = rowsPerLevel[0];
+      int rowspacer            = 2;
+      while (tt != 0) {
+        tt /= 10;
+        rowspacer++;
+      }
+      for (size_t i = 0; i < nnzPerLevel.size(); ++i) {
+        tt = nnzPerLevel[i];
+        if (tt != INVALID)
+          break;
+        tt = 100;  // This will get used if all levels are operators.
+      }
+      int nnzspacer = 2;
+      while (tt != 0) {
+        tt /= 10;
+        nnzspacer++;
+      }
+      tt           = numProcsPerLevel[0];
+      int npspacer = 2;
+      while (tt != 0) {
+        tt /= 10;
+        npspacer++;
+      }
+      oss << "level " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz "
+          << " nnz/row" << std::setw(npspacer) << "  c ratio"
+          << "  procs" << std::endl;
+      for (size_t i = 0; i < nnzPerLevel.size(); ++i) {
+        oss << "  " << i << "  ";
+        oss << std::setw(rowspacer) << rowsPerLevel[i];
+        if (nnzPerLevel[i] != INVALID) {
+          oss << std::setw(nnzspacer) << nnzPerLevel[i];
+          oss << std::setprecision(2) << std::setiosflags(std::ios::fixed);
+          oss << std::setw(9) << as<double>(nnzPerLevel[i]) / rowsPerLevel[i];
         } else {
-          LO storageblocksize=Am->GetStorageBlockSize();
-          Xpetra::global_size_t nnz = Am->getGlobalNumEntries()*storageblocksize*storageblocksize;
-          nnzPerLevel     .push_back(nnz);
-          rowsPerLevel    .push_back(Am->getGlobalNumRows()*storageblocksize);
-          numProcsPerLevel.push_back(Am->getRowMap()->getComm()->getSize());
+          oss << std::setw(nnzspacer) << "Operator";
+          oss << std::setprecision(2) << std::setiosflags(std::ios::fixed);
+          oss << std::setw(9) << "     ";
         }
-      }
-      if (someOpsNotMatrices)
-        GetOStream(Warnings0) << "Some level operators are not matrices, statistics calculation are incomplete" << std::endl;
-
-      {
-        std::string label = Levels_[0]->getObjectLabel();
-        std::ostringstream oss;
-        oss << std::setfill(' ');
-        oss << "\n--------------------------------------------------------------------------------\n";
-        oss << "---                            Multigrid Summary "  << std::setw(28) << std::left << label << "---\n";
-        oss << "--------------------------------------------------------------------------------" << std::endl;
-        if (verbLevel & Parameters1)
-          oss << "Scalar              = " << Teuchos::ScalarTraits<Scalar>::name() << std::endl;
-        oss << "Number of levels    = " << numLevels << std::endl;
-        oss << "Operator complexity = " << std::setprecision(2) << std::setiosflags(std::ios::fixed);
-        if (!someOpsNotMatrices)
-          oss << GetOperatorComplexity() << std::endl;
+        if (i)
+          oss << std::setw(9) << as<double>(rowsPerLevel[i - 1]) / rowsPerLevel[i];
         else
-          oss << "not available (Some operators in hierarchy are not matrices.)" << std::endl;
-
-        if(smoother_comp!=-1.0) {
-          oss << "Smoother complexity = " << std::setprecision(2) << std::setiosflags(std::ios::fixed)
-              << smoother_comp << std::endl;
+          oss << std::setw(9) << "     ";
+        oss << "    " << std::setw(npspacer) << numProcsPerLevel[i] << std::endl;
+      }
+      oss << std::endl;
+      for (int i = 0; i < GetNumLevels(); ++i) {
+        RCP<SmootherBase> preSmoo, postSmoo;
+        if (Levels_[i]->IsAvailable("PreSmoother"))
+          preSmoo = Levels_[i]->template Get<RCP<SmootherBase> >("PreSmoother");
+        if (Levels_[i]->IsAvailable("PostSmoother"))
+          postSmoo = Levels_[i]->template Get<RCP<SmootherBase> >("PostSmoother");
+
+        if (preSmoo != null && preSmoo == postSmoo)
+          oss << "Smoother (level " << i << ") both : " << preSmoo->description() << std::endl;
+        else {
+          oss << "Smoother (level " << i << ") pre  : "
+              << (preSmoo != null ? preSmoo->description() : "no smoother") << std::endl;
+          oss << "Smoother (level " << i << ") post : "
+              << (postSmoo != null ? postSmoo->description() : "no smoother") << std::endl;
         }
 
-        switch (Cycle_) {
-           case VCYCLE:
-             oss << "Cycle type          = V" << std::endl;
-             break;
-           case WCYCLE:
-             oss << "Cycle type          = W" << std::endl;
-             if (WCycleStartLevel_ > 0)
-               oss << "Cycle start level   = " << WCycleStartLevel_ << std::endl;
-             break;
-           default:
-             break;
-        };
-        oss << std::endl;
-
-        Xpetra::global_size_t tt = rowsPerLevel[0];
-        int rowspacer = 2; while (tt != 0) { tt /= 10; rowspacer++; }
-        for (size_t i = 0; i < nnzPerLevel.size(); ++i) {
-          tt = nnzPerLevel[i];
-          if (tt != INVALID)
-            break;
-          tt = 100;  // This will get used if all levels are operators.
-        }
-        int nnzspacer = 2; while (tt != 0) { tt /= 10; nnzspacer++; }
-        tt = numProcsPerLevel[0];
-        int npspacer = 2;  while (tt != 0) { tt /= 10; npspacer++; }
-        oss  << "level " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << " nnz/row" << std::setw(npspacer) << "  c ratio" << "  procs" << std::endl;
-        for (size_t i = 0; i < nnzPerLevel.size(); ++i) {
-          oss << "  " << i << "  ";
-          oss << std::setw(rowspacer) << rowsPerLevel[i];
-          if (nnzPerLevel[i] != INVALID) {
-            oss << std::setw(nnzspacer) << nnzPerLevel[i];
-            oss << std::setprecision(2) << std::setiosflags(std::ios::fixed);
-            oss << std::setw(9) << as<double>(nnzPerLevel[i]) / rowsPerLevel[i];
-          } else {
-            oss << std::setw(nnzspacer) << "Operator";
-            oss << std::setprecision(2) << std::setiosflags(std::ios::fixed);
-            oss << std::setw(9) << "     ";
-          }
-          if (i) oss << std::setw(9) << as<double>(rowsPerLevel[i-1])/rowsPerLevel[i];
-          else   oss << std::setw(9) << "     ";
-          oss << "    " << std::setw(npspacer) << numProcsPerLevel[i] << std::endl;
-        }
         oss << std::endl;
-        for (int i = 0; i < GetNumLevels(); ++i) {
-          RCP<SmootherBase> preSmoo, postSmoo;
-          if (Levels_[i]->IsAvailable("PreSmoother"))
-            preSmoo = Levels_[i]->template Get< RCP<SmootherBase> >("PreSmoother");
-          if (Levels_[i]->IsAvailable("PostSmoother"))
-            postSmoo = Levels_[i]->template Get< RCP<SmootherBase> >("PostSmoother");
-
-          if (preSmoo != null && preSmoo == postSmoo)
-            oss << "Smoother (level " << i << ") both : " << preSmoo->description() << std::endl;
-          else {
-            oss << "Smoother (level " << i << ") pre  : "
-                << (preSmoo != null ?  preSmoo->description() : "no smoother") << std::endl;
-            oss << "Smoother (level " << i << ") post : "
-                << (postSmoo != null ?  postSmoo->description() : "no smoother") << std::endl;
-          }
-
-          oss << std::endl;
-        }
-
-        outstr = oss.str();
       }
+
+      outstr = oss.str();
     }
+  }
 
 #ifdef HAVE_MPI
-    RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-    MPI_Comm rawComm = (*mpiComm->getRawMpiComm())();
-
-    int strLength = outstr.size();
-    MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
-    if (comm->getRank() != root)
-      outstr.resize(strLength);
-    MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+  RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+  MPI_Comm rawComm                          = (*mpiComm->getRawMpiComm())();
+
+  int strLength = outstr.size();
+  MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
+  if (comm->getRank() != root)
+    outstr.resize(strLength);
+  MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
 #endif
 
-    out << outstr;
-  }
+  out << outstr;
+}
 
-  // NOTE: at some point this should be replaced by a friend operator <<
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(std::ostream& out, const VerbLevel verbLevel) const {
-    Teuchos::OSTab tab2(out);
-    for (int i = 0; i < GetNumLevels(); ++i)
-      Levels_[i]->print(out, verbLevel);
-  }
+// NOTE: at some point this should be replaced by a friend operator <<
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(std::ostream& out, const VerbLevel verbLevel) const {
+  Teuchos::OSTab tab2(out);
+  for (int i = 0; i < GetNumLevels(); ++i)
+    Levels_[i]->print(out, verbLevel);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsPreconditioner(const bool flag) {
-    isPreconditioner_ = flag;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsPreconditioner(const bool flag) {
+  isPreconditioner_ = flag;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DumpCurrentGraph(int currLevel) const {
-    if (GetProcRankVerbose() != 0)
-      return;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DumpCurrentGraph(int currLevel) const {
+  if (GetProcRankVerbose() != 0)
+    return;
 #if defined(HAVE_MUELU_BOOST) && defined(HAVE_MUELU_BOOST_FOR_REAL) && defined(BOOST_VERSION) && (BOOST_VERSION >= 104400)
 
-    BoostGraph      graph;
-
-    BoostProperties dp;
-    dp.property("label", boost::get(boost::vertex_name,  graph));
-    dp.property("id",    boost::get(boost::vertex_index, graph));
-    dp.property("label", boost::get(boost::edge_name,    graph));
-    dp.property("color", boost::get(boost::edge_color,   graph));
-
-    // create local maps
-    std::map<const FactoryBase*, BoostVertex>                                     vindices;
-    typedef std::map<std::pair<BoostVertex,BoostVertex>, std::string> emap; emap  edges;
-
-    static int call_id=0;
-
-    RCP<Operator> A = Levels_[0]->template Get<RCP<Operator> >("A");
-    int rank = A->getDomainMap()->getComm()->getRank();
-
-    //    printf("[%d] CMS: ----------------------\n",rank);
-    for (int i = currLevel; i <= currLevel+1 && i < GetNumLevels(); i++) {
-      edges.clear();
-      Levels_[i]->UpdateGraph(vindices, edges, dp, graph);
-
-      for (emap::const_iterator eit = edges.begin(); eit != edges.end(); eit++) {
-        std::pair<BoostEdge, bool> boost_edge = boost::add_edge(eit->first.first, eit->first.second, graph);
-        // printf("[%d] CMS:   Hierarchy, adding edge (%d->%d) %d\n",rank,(int)eit->first.first,(int)eit->first.second,(int)boost_edge.second);
-        // Because xdot.py views 'Graph' as a keyword
-        if(eit->second==std::string("Graph")) boost::put("label", dp, boost_edge.first, std::string("Graph_"));
-        else boost::put("label", dp, boost_edge.first, eit->second);
-        if (i == currLevel)
-          boost::put("color", dp, boost_edge.first, std::string("red"));
-        else
-          boost::put("color", dp, boost_edge.first, std::string("blue"));
-      }
+  BoostGraph graph;
+
+  BoostProperties dp;
+  dp.property("label", boost::get(boost::vertex_name, graph));
+  dp.property("id", boost::get(boost::vertex_index, graph));
+  dp.property("label", boost::get(boost::edge_name, graph));
+  dp.property("color", boost::get(boost::edge_color, graph));
+
+  // create local maps
+  std::map<const FactoryBase*, BoostVertex> vindices;
+  typedef std::map<std::pair<BoostVertex, BoostVertex>, std::string> emap;
+  emap edges;
+
+  static int call_id = 0;
+
+  RCP<Operator> A = Levels_[0]->template Get<RCP<Operator> >("A");
+  int rank        = A->getDomainMap()->getComm()->getRank();
+
+  //    printf("[%d] CMS: ----------------------\n",rank);
+  for (int i = currLevel; i <= currLevel + 1 && i < GetNumLevels(); i++) {
+    edges.clear();
+    Levels_[i]->UpdateGraph(vindices, edges, dp, graph);
+
+    for (emap::const_iterator eit = edges.begin(); eit != edges.end(); eit++) {
+      std::pair<BoostEdge, bool> boost_edge = boost::add_edge(eit->first.first, eit->first.second, graph);
+      // printf("[%d] CMS:   Hierarchy, adding edge (%d->%d) %d\n",rank,(int)eit->first.first,(int)eit->first.second,(int)boost_edge.second);
+      // Because xdot.py views 'Graph' as a keyword
+      if (eit->second == std::string("Graph"))
+        boost::put("label", dp, boost_edge.first, std::string("Graph_"));
+      else
+        boost::put("label", dp, boost_edge.first, eit->second);
+      if (i == currLevel)
+        boost::put("color", dp, boost_edge.first, std::string("red"));
+      else
+        boost::put("color", dp, boost_edge.first, std::string("blue"));
     }
+  }
 
-    std::ofstream out(dumpFile_.c_str()+std::string("_")+std::to_string(currLevel)+std::string("_")+std::to_string(call_id)+std::string("_")+ std::to_string(rank) + std::string(".dot"));
-    boost::write_graphviz_dp(out, graph, dp, std::string("id"));
-    out.close();
-    call_id++;
+  std::ofstream out(dumpFile_.c_str() + std::string("_") + std::to_string(currLevel) + std::string("_") + std::to_string(call_id) + std::string("_") + std::to_string(rank) + std::string(".dot"));
+  boost::write_graphviz_dp(out, graph, dp, std::string("id"));
+  out.close();
+  call_id++;
 #else
-    GetOStream(Errors) <<  "Dependency graph output requires boost and MueLu_ENABLE_Boost_for_real" << std::endl;
+  GetOStream(Errors) << "Dependency graph output requires boost and MueLu_ENABLE_Boost_for_real" << std::endl;
 #endif
+}
+
+// Enforce that coordinate vector's map is consistent with that of A
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReplaceCoordinateMap(Level& level) {
+  RCP<Operator> Ao = level.Get<RCP<Operator> >("A");
+  RCP<Matrix> A    = rcp_dynamic_cast<Matrix>(Ao);
+  if (A.is_null()) {
+    GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: operator is not a matrix, skipping..." << std::endl;
+    return;
+  }
+  if (Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A) != Teuchos::null) {
+    GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: operator is a BlockedCrsMatrix, skipping..." << std::endl;
+    return;
   }
 
-  // Enforce that coordinate vector's map is consistent with that of A
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReplaceCoordinateMap(Level& level) {
-    RCP<Operator> Ao = level.Get<RCP<Operator> >("A");
-    RCP<Matrix>   A  = rcp_dynamic_cast<Matrix>(Ao);
-    if (A.is_null()) {
-      GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: operator is not a matrix, skipping..." << std::endl;
-      return;
-    }
-    if(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A) != Teuchos::null) {
-      GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: operator is a BlockedCrsMatrix, skipping..." << std::endl;
-      return;
-    }
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> xdMV;
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> xdMV;
+  RCP<xdMV> coords = level.Get<RCP<xdMV> >("Coordinates");
 
-    RCP<xdMV> coords = level.Get<RCP<xdMV> >("Coordinates");
+  if (A->getRowMap()->isSameAs(*(coords->getMap()))) {
+    GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: matrix and coordinates maps are same, skipping..." << std::endl;
+    return;
+  }
 
-    if (A->getRowMap()->isSameAs(*(coords->getMap()))) {
-      GetOStream(Runtime1) << "Hierarchy::ReplaceCoordinateMap: matrix and coordinates maps are same, skipping..." << std::endl;
-      return;
-    }
+  if (A->IsView("stridedMaps") && rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+    RCP<const StridedMap> stridedRowMap = rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps"));
 
-    if (A->IsView("stridedMaps") && rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-      RCP<const StridedMap> stridedRowMap = rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps"));
+    // It is better to through an exceptions if maps may be inconsistent, than to ignore it and experience unfathomable breakdowns
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedRowMap->getStridedBlockId() != -1 || stridedRowMap->getOffset() != 0,
+                               Exceptions::RuntimeError, "Hierarchy::ReplaceCoordinateMap: nontrivial maps (block id = " << stridedRowMap->getStridedBlockId() << ", offset = " << stridedRowMap->getOffset() << ")");
+  }
 
-      // It is better to through an exceptions if maps may be inconsistent, than to ignore it and experience unfathomable breakdowns
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedRowMap->getStridedBlockId() != -1 || stridedRowMap->getOffset() != 0,
-                                 Exceptions::RuntimeError, "Hierarchy::ReplaceCoordinateMap: nontrivial maps (block id = " << stridedRowMap->getStridedBlockId()
-                                 << ", offset = " << stridedRowMap->getOffset() << ")");
+  GetOStream(Runtime1) << "Replacing coordinate map" << std::endl;
+  TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0, Exceptions::RuntimeError, "Hierarchy::ReplaceCoordinateMap: Storage block size does not evenly divide fixed block size");
+
+  size_t blkSize = A->GetFixedBlockSize() / A->GetStorageBlockSize();
+
+  RCP<const Map> nodeMap = A->getRowMap();
+  if (blkSize > 1) {
+    // Create a nodal map, as coordinates have not been expanded to a DOF map yet.
+    RCP<const Map> dofMap = A->getRowMap();
+    GO indexBase          = dofMap->getIndexBase();
+    size_t numLocalDOFs   = dofMap->getLocalNumElements();
+    TEUCHOS_TEST_FOR_EXCEPTION(numLocalDOFs % blkSize, Exceptions::RuntimeError,
+                               "Hierarchy::ReplaceCoordinateMap: block size (" << blkSize << ") is incompatible with the number of local dofs in a row map (" << numLocalDOFs);
+    ArrayView<const GO> GIDs = dofMap->getLocalElementList();
+
+    Array<GO> nodeGIDs(numLocalDOFs / blkSize);
+    for (size_t i = 0; i < numLocalDOFs; i += blkSize)
+      nodeGIDs[i / blkSize] = (GIDs[i] - indexBase) / blkSize + indexBase;
+
+    Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+    nodeMap                       = MapFactory::Build(dofMap->lib(), INVALID, nodeGIDs(), indexBase, dofMap->getComm());
+  } else {
+    // blkSize == 1
+    // Check whether the length of vectors fits to the size of A
+    // If yes, make sure that the maps are matching
+    // If no, throw a warning but do not touch the Coordinates
+    if (coords->getLocalLength() != A->getRowMap()->getLocalNumElements()) {
+      GetOStream(Warnings) << "Coordinate vector does not match row map of matrix A!" << std::endl;
+      return;
     }
+  }
 
-    GetOStream(Runtime1) << "Replacing coordinate map" << std::endl;
-    TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() % A->GetStorageBlockSize() != 0, Exceptions::RuntimeError, "Hierarchy::ReplaceCoordinateMap: Storage block size does not evenly divide fixed block size");
-
-    size_t blkSize = A->GetFixedBlockSize() / A->GetStorageBlockSize();
+  Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coordDataView;
+  std::vector<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coordData;
+  for (size_t i = 0; i < coords->getNumVectors(); i++) {
+    coordData.push_back(coords->getData(i));
+    coordDataView.push_back(coordData[i]());
+  }
 
-    RCP<const Map> nodeMap = A->getRowMap();
-    if (blkSize > 1) {
-      // Create a nodal map, as coordinates have not been expanded to a DOF map yet.
-      RCP<const Map> dofMap       = A->getRowMap();
-      GO             indexBase    = dofMap->getIndexBase();
-      size_t         numLocalDOFs = dofMap->getLocalNumElements();
-      TEUCHOS_TEST_FOR_EXCEPTION(numLocalDOFs % blkSize, Exceptions::RuntimeError,
-        "Hierarchy::ReplaceCoordinateMap: block size (" << blkSize << ") is incompatible with the number of local dofs in a row map (" << numLocalDOFs);
-      ArrayView<const GO> GIDs = dofMap->getLocalElementList();
+  RCP<xdMV> newCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>::Build(nodeMap, coordDataView(), coords->getNumVectors());
+  level.Set("Coordinates", newCoords);
+}
 
-      Array<GO> nodeGIDs(numLocalDOFs/blkSize);
-      for (size_t i = 0; i < numLocalDOFs; i += blkSize)
-        nodeGIDs[i/blkSize] = (GIDs[i] - indexBase)/blkSize + indexBase;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AllocateLevelMultiVectors(int numvecs, bool forceMapCheck) {
+  int N = Levels_.size();
+  if (((sizeOfAllocatedLevelMultiVectors_ == numvecs && residual_.size() == N) || numvecs <= 0) && !forceMapCheck) return;
+
+  // If, somehow, we changed the number of levels, delete everything first
+  if (residual_.size() != N) {
+    DeleteLevelMultiVectors();
+
+    residual_.resize(N);
+    coarseRhs_.resize(N);
+    coarseX_.resize(N);
+    coarseImport_.resize(N);
+    coarseExport_.resize(N);
+    correction_.resize(N);
+  }
 
-      Xpetra::global_size_t INVALID = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-      nodeMap = MapFactory::Build(dofMap->lib(), INVALID, nodeGIDs(), indexBase, dofMap->getComm());
-    } else {
-      // blkSize == 1
-      // Check whether the length of vectors fits to the size of A
-      // If yes, make sure that the maps are matching
-      // If no, throw a warning but do not touch the Coordinates
-      if(coords->getLocalLength() != A->getRowMap()->getLocalNumElements()) {
-        GetOStream(Warnings) << "Coordinate vector does not match row map of matrix A!" << std::endl;
-        return;
+  for (int i = 0; i < N; i++) {
+    RCP<Operator> A = Levels_[i]->template Get<RCP<Operator> >("A");
+    if (!A.is_null()) {
+      // This dance is because we allow A to have a BlockedMap and X/B to have (compatible) non-blocked map
+      RCP<const BlockedCrsMatrix> A_as_blocked = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
+      RCP<const Map> Arm                       = A->getRangeMap();
+      RCP<const Map> Adm                       = A->getDomainMap();
+      if (!A_as_blocked.is_null()) {
+        Adm = A_as_blocked->getFullDomainMap();
       }
-    }
 
-    Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> >      coordDataView;
-    std::vector<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coordData;
-    for (size_t i = 0; i < coords->getNumVectors(); i++) {
-      coordData.push_back(coords->getData(i));
-      coordDataView.push_back(coordData[i]());
-    }
-
-    RCP<xdMV> newCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>::Build(nodeMap, coordDataView(), coords->getNumVectors());
-    level.Set("Coordinates", newCoords);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AllocateLevelMultiVectors(int numvecs, bool forceMapCheck) {
-    int N = Levels_.size();
-    if( ( (sizeOfAllocatedLevelMultiVectors_ == numvecs && residual_.size() == N) || numvecs<=0 ) && !forceMapCheck) return;
-
-    // If, somehow, we changed the number of levels, delete everything first
-    if(residual_.size() != N) {
-      DeleteLevelMultiVectors();
-
-      residual_.resize(N);
-      coarseRhs_.resize(N);
-      coarseX_.resize(N);
-      coarseImport_.resize(N);
-      coarseExport_.resize(N);
-      correction_.resize(N);
+      if (residual_[i].is_null() || !residual_[i]->getMap()->isSameAs(*Arm))
+        // This is zero'd by default since it is filled via an operator apply
+        residual_[i] = MultiVectorFactory::Build(Arm, numvecs, true);
+      if (correction_[i].is_null() || !correction_[i]->getMap()->isSameAs(*Adm))
+        correction_[i] = MultiVectorFactory::Build(Adm, numvecs, false);
     }
 
-    for(int i=0; i<N; i++) {
-      RCP<Operator> A = Levels_[i]->template Get< RCP<Operator> >("A");
-      if(!A.is_null()) {
-        // This dance is because we allow A to have a BlockedMap and X/B to have (compatible) non-blocked map
-        RCP<const BlockedCrsMatrix> A_as_blocked = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
-        RCP<const Map> Arm = A->getRangeMap();
-        RCP<const Map> Adm = A->getDomainMap();
-        if(!A_as_blocked.is_null()) {
-          Adm = A_as_blocked->getFullDomainMap();
+    if (i + 1 < N) {
+      // This is zero'd by default since it is filled via an operator apply
+      if (implicitTranspose_) {
+        RCP<Operator> P = Levels_[i + 1]->template Get<RCP<Operator> >("P");
+        if (!P.is_null()) {
+          RCP<const Map> map = P->getDomainMap();
+          if (coarseRhs_[i].is_null() || !coarseRhs_[i]->getMap()->isSameAs(*map))
+            coarseRhs_[i] = MultiVectorFactory::Build(map, numvecs, true);
         }
-
-        if (residual_[i].is_null() || !residual_[i]->getMap()->isSameAs(*Arm))
-          // This is zero'd by default since it is filled via an operator apply
-          residual_[i] = MultiVectorFactory::Build(Arm, numvecs, true);
-        if (correction_[i].is_null() || !correction_[i]->getMap()->isSameAs(*Adm))
-          correction_[i] = MultiVectorFactory::Build(Adm, numvecs, false);
-      }
-
-      if(i+1<N) {
-        // This is zero'd by default since it is filled via an operator apply
-        if(implicitTranspose_) {
-          RCP<Operator> P = Levels_[i+1]->template Get< RCP<Operator> >("P");
-          if(!P.is_null()) {
-            RCP<const Map> map = P->getDomainMap();
-            if (coarseRhs_[i].is_null() || !coarseRhs_[i]->getMap()->isSameAs(*map))
-              coarseRhs_[i] = MultiVectorFactory::Build(map, numvecs, true);
-          }
-        } else {
-          RCP<Operator> R = Levels_[i+1]->template Get< RCP<Operator> >("R");
-          if (!R.is_null()) {
-            RCP<const Map> map = R->getRangeMap();
-            if (coarseRhs_[i].is_null() || !coarseRhs_[i]->getMap()->isSameAs(*map))
-              coarseRhs_[i] = MultiVectorFactory::Build(map, numvecs, true);
-          }
+      } else {
+        RCP<Operator> R = Levels_[i + 1]->template Get<RCP<Operator> >("R");
+        if (!R.is_null()) {
+          RCP<const Map> map = R->getRangeMap();
+          if (coarseRhs_[i].is_null() || !coarseRhs_[i]->getMap()->isSameAs(*map))
+            coarseRhs_[i] = MultiVectorFactory::Build(map, numvecs, true);
         }
+      }
 
-
-        RCP<const Import> importer;
-        if(Levels_[i+1]->IsAvailable("Importer"))
-          importer = Levels_[i+1]->template Get< RCP<const Import> >("Importer");
-        if (doPRrebalance_ || importer.is_null()) {
-          RCP<const Map> map = coarseRhs_[i]->getMap();
-          if (coarseX_[i].is_null() || !coarseX_[i]->getMap()->isSameAs(*map))
-            coarseX_[i] = MultiVectorFactory::Build(map, numvecs, true);
-        } else {
-          RCP<const Map> map;
-          map = importer->getTargetMap();
-          if (coarseImport_[i].is_null() || !coarseImport_[i]->getMap()->isSameAs(*map)) {
-            coarseImport_[i] = MultiVectorFactory::Build(map, numvecs,false);
-            coarseX_[i] = MultiVectorFactory::Build(map,numvecs,false);
-          }
-          map = importer->getSourceMap();
-          if (coarseExport_[i].is_null() || !coarseExport_[i]->getMap()->isSameAs(*map))
-            coarseExport_[i] = MultiVectorFactory::Build(map, numvecs,false);
+      RCP<const Import> importer;
+      if (Levels_[i + 1]->IsAvailable("Importer"))
+        importer = Levels_[i + 1]->template Get<RCP<const Import> >("Importer");
+      if (doPRrebalance_ || importer.is_null()) {
+        RCP<const Map> map = coarseRhs_[i]->getMap();
+        if (coarseX_[i].is_null() || !coarseX_[i]->getMap()->isSameAs(*map))
+          coarseX_[i] = MultiVectorFactory::Build(map, numvecs, true);
+      } else {
+        RCP<const Map> map;
+        map = importer->getTargetMap();
+        if (coarseImport_[i].is_null() || !coarseImport_[i]->getMap()->isSameAs(*map)) {
+          coarseImport_[i] = MultiVectorFactory::Build(map, numvecs, false);
+          coarseX_[i]      = MultiVectorFactory::Build(map, numvecs, false);
         }
+        map = importer->getSourceMap();
+        if (coarseExport_[i].is_null() || !coarseExport_[i]->getMap()->isSameAs(*map))
+          coarseExport_[i] = MultiVectorFactory::Build(map, numvecs, false);
       }
     }
-    sizeOfAllocatedLevelMultiVectors_ = numvecs;
   }
-
+  sizeOfAllocatedLevelMultiVectors_ = numvecs;
+}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeleteLevelMultiVectors() {
-  if(sizeOfAllocatedLevelMultiVectors_==0) return;
+  if (sizeOfAllocatedLevelMultiVectors_ == 0) return;
   residual_.resize(0);
   coarseRhs_.resize(0);
   coarseX_.resize(0);
@@ -1616,23 +1633,18 @@ void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeleteLevelMultiVecto
   sizeOfAllocatedLevelMultiVectors_ = 0;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 bool Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsCalculationOfResidualRequired(
-    const LO startLevel, const ConvData& conv) const
-{
+    const LO startLevel, const ConvData& conv) const {
   return (startLevel == 0 && !isPreconditioner_ && (IsPrint(Statistics1) || conv.tol_ > 0));
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsConverged(
-    const Teuchos::Array<MagnitudeType>& residualNorm, const MagnitudeType convergenceTolerance) const
-{
+    const Teuchos::Array<MagnitudeType>& residualNorm, const MagnitudeType convergenceTolerance) const {
   ConvergenceStatus convergenceStatus = ConvergenceStatus::Undefined;
 
-  if (convergenceTolerance > Teuchos::ScalarTraits<MagnitudeType>::zero())
-  {
+  if (convergenceTolerance > Teuchos::ScalarTraits<MagnitudeType>::zero()) {
     bool passed = true;
     for (LO k = 0; k < residualNorm.size(); k++)
       if (residualNorm[k] >= convergenceTolerance)
@@ -1647,30 +1659,27 @@ ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsConver
   return convergenceStatus;
 }
 
-
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintResidualHistory(
-    const LO iteration, const Teuchos::Array<MagnitudeType>& residualNorm) const
-{
+    const LO iteration, const Teuchos::Array<MagnitudeType>& residualNorm) const {
   GetOStream(Statistics1) << "iter:    "
-      << std::setiosflags(std::ios::left)
-      << std::setprecision(3) << std::setw(4) << iteration
-      << "           residual = "
-      << std::setprecision(10) << residualNorm
-      << std::endl;
+                          << std::setiosflags(std::ios::left)
+                          << std::setprecision(3) << std::setw(4) << iteration
+                          << "           residual = "
+                          << std::setprecision(10) << residualNorm
+                          << std::endl;
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeResidualAndPrintHistory(
     const Operator& A, const MultiVector& X, const MultiVector& B, const LO iteration,
-    const LO startLevel, const ConvData& conv, MagnitudeType& previousResidualNorm)
-{
+    const LO startLevel, const ConvData& conv, MagnitudeType& previousResidualNorm) {
   Teuchos::Array<MagnitudeType> residualNorm;
   residualNorm = Utilities::ResidualNorm(A, X, B, *residual_[startLevel]);
 
   const MagnitudeType currentResidualNorm = residualNorm[0];
-  rate_ = currentResidualNorm / previousResidualNorm;
-  previousResidualNorm = currentResidualNorm;
+  rate_                                   = currentResidualNorm / previousResidualNorm;
+  previousResidualNorm                    = currentResidualNorm;
 
   if (IsPrint(Statistics1))
     PrintResidualHistory(iteration, residualNorm);
@@ -1678,7 +1687,6 @@ ConvergenceStatus Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeR
   return IsConverged(residualNorm, conv.tol_);
 }
 
+}  // namespace MueLu
 
-} //namespace MueLu
-
-#endif // MUELU_HIERARCHY_DEF_HPP
+#endif  // MUELU_HIERARCHY_DEF_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_KeepType.hpp b/packages/muelu/src/MueCentral/MueLu_KeepType.hpp
index dca2403bf630..2b042e7de044 100644
--- a/packages/muelu/src/MueCentral/MueLu_KeepType.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_KeepType.hpp
@@ -48,20 +48,20 @@
 
 namespace MueLu {
 
-  //! Keep status of a variable of Level.
-  // Several keep status can be set at the same time
-  enum KeepEnum {
-    UserData  = 0x1, //!< User data are always kept. This flag is set automatically when Level::Set("data", data) is used. The keep status of the variable is not propagated to coarser level (if you use Level::Build()).
-    Keep      = 0x2, //!< Always keep data, even accross run. This flag is set by Level::Keep(). This flag is propagated to coarser level by Level::Build().
-    Final     = 0x4, //!< Keep data only for this run. Used to keep data useful for Hierarchy::Iterate(). Data will be deleted if setup phase is re-run. This flag is set by default for A, P, R, PreSmoother and PostSmoother of NoFactory by Hierarchy::Setup(). Not propagated by Level::Build().
+//! Keep status of a variable of Level.
+// Several keep status can be set at the same time
+enum KeepEnum {
+  UserData = 0x1,  //!< User data are always kept. This flag is set automatically when Level::Set("data", data) is used. The keep status of the variable is not propagated to coarser level (if you use Level::Build()).
+  Keep     = 0x2,  //!< Always keep data, even accross run. This flag is set by Level::Keep(). This flag is propagated to coarser level by Level::Build().
+  Final    = 0x4,  //!< Keep data only for this run. Used to keep data useful for Hierarchy::Iterate(). Data will be deleted if setup phase is re-run. This flag is set by default for A, P, R, PreSmoother and PostSmoother of NoFactory by Hierarchy::Setup(). Not propagated by Level::Build().
 
-    NextRun   = UserData | Keep, //!< Both UserData and Keep flags force data to be kept and reused for the next run. Do not use MueLu::NextRun in AddKeepFlag. Use it only for testing keep == UserData || keep == Keep.
-    All       = UserData | Keep | Final
-  };
+  NextRun = UserData | Keep,  //!< Both UserData and Keep flags force data to be kept and reused for the next run. Do not use MueLu::NextRun in AddKeepFlag. Use it only for testing keep == UserData || keep == Keep.
+  All     = UserData | Keep | Final
+};
 
-  //!
-  typedef short KeepType; //TODO: name it KeepFlag?
+//!
+typedef short KeepType;  // TODO: name it KeepFlag?
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_KEEPTYPE_HPP
+#endif  // MUELU_KEEPTYPE_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_Level.cpp b/packages/muelu/src/MueCentral/MueLu_Level.cpp
index 187770a6cc48..5edc957ec7fa 100644
--- a/packages/muelu/src/MueCentral/MueLu_Level.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_Level.cpp
@@ -51,482 +51,500 @@
 
 namespace MueLu {
 
-  RCP<Level> Level::Build() {
-    RCP<Level> newLevel = rcp(new Level());
+RCP<Level> Level::Build() {
+  RCP<Level> newLevel = rcp(new Level());
 
-    // Copy 'keep' status of variables
-    for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
-      const FactoryBase* factory = kt->first;
+  // Copy 'keep' status of variables
+  for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
+    const FactoryBase* factory = kt->first;
 
-      for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
-        const std::string& ename = it->first;
+    for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
+      const std::string& ename = it->first;
 
-        if (IsKept(ename, factory, MueLu::Keep)) { // MueLu::Keep is the only flag propagated
-          if (factory == NULL)                                  // TODO: Is this possible?? Throw exception. Not supposed to use the FactoryManager here.
-            newLevel->Keep(ename, NoFactory::get());
-          else
-            newLevel->Keep(ename, factory);
-        }
+      if (IsKept(ename, factory, MueLu::Keep)) {  // MueLu::Keep is the only flag propagated
+        if (factory == NULL)                      // TODO: Is this possible?? Throw exception. Not supposed to use the FactoryManager here.
+          newLevel->Keep(ename, NoFactory::get());
+        else
+          newLevel->Keep(ename, factory);
       }
     }
-
-    return newLevel;
   }
 
-  int Level::GetLevelID() const { return levelID_; }
+  return newLevel;
+}
 
-  void Level::SetLevelID(int levelID) {
-    if (levelID_ != -1 && levelID_ != levelID)
-      GetOStream(Warnings1) << "Level::SetLevelID(): Changing an already defined LevelID (previousID=" << levelID_ << ", newID=" << levelID << ")" << std::endl;
+int Level::GetLevelID() const { return levelID_; }
 
-    levelID_ = levelID;
-  }
+void Level::SetLevelID(int levelID) {
+  if (levelID_ != -1 && levelID_ != levelID)
+    GetOStream(Warnings1) << "Level::SetLevelID(): Changing an already defined LevelID (previousID=" << levelID_ << ", newID=" << levelID << ")" << std::endl;
 
-  void Level::SetPreviousLevel(const RCP<Level> & previousLevel) {
-    if (previousLevel_ != Teuchos::null && previousLevel_ != previousLevel)
-      GetOStream(Warnings1) << "Level::SetPreviousLevel(): PreviousLevel was already defined" << std::endl;
+  levelID_ = levelID;
+}
 
-    previousLevel_ = previousLevel;
-  }
+void Level::SetPreviousLevel(const RCP<Level>& previousLevel) {
+  if (previousLevel_ != Teuchos::null && previousLevel_ != previousLevel)
+    GetOStream(Warnings1) << "Level::SetPreviousLevel(): PreviousLevel was already defined" << std::endl;
 
-  void Level::SetFactoryManager(const RCP<const FactoryManagerBase> & factoryManager) {
-    factoryManager_ = factoryManager;
-  }
+  previousLevel_ = previousLevel;
+}
 
-  const RCP<const FactoryManagerBase> Level::GetFactoryManager() {
-    return factoryManager_;
-  }
+void Level::SetFactoryManager(const RCP<const FactoryManagerBase>& factoryManager) {
+  factoryManager_ = factoryManager;
+}
 
-  void Level::AddKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
-    if (!IsKey(factory, ename)) {
-      // If the entry does not exist, create it to store the keep flag
-      Teuchos::RCP<MueLu::VariableContainer> newVar = Teuchos::rcp(new MueLu::VariableContainer);
-      map_[factory][ename] = newVar;
-    }
-    // Set the flag
-    map_[factory][ename]->AddKeepFlag(keep);
-  }
+const RCP<const FactoryManagerBase> Level::GetFactoryManager() {
+  return factoryManager_;
+}
 
-  void Level::RemoveKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
-    // No entry = nothing to do
-    if (!IsKey(factory, ename))
-      return;
-
-    // Remove the flag
-    Teuchos::RCP<MueLu::VariableContainer>& v = map_[factory][ename];
-    v->RemoveKeepFlag(keep);
-
-    // Remove data if no keep flag left and counter == 0
-    if ((v->IsRequested() == false) && (v->GetKeepFlag() == 0)) {
-      v = Teuchos::null; // free data
-
-      map_[factory].erase(ename);
-      if (map_.count(factory) == 0)
-        map_.erase(factory);
-    }
+void Level::AddKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
+  if (!IsKey(factory, ename)) {
+    // If the entry does not exist, create it to store the keep flag
+    Teuchos::RCP<MueLu::VariableContainer> newVar = Teuchos::rcp(new MueLu::VariableContainer);
+    map_[factory][ename]                          = newVar;
   }
-
-  KeepType Level::GetKeepFlag(const std::string& ename, const FactoryBase* factory) const {
-    if (!IsKey(factory,ename))
-      return false;
-
-    return Get(factory, ename)->GetKeepFlag();
-  }
-
-  void Level::Request(const FactoryBase& factory) {
-    RequestMode prev = requestMode_;
-    requestMode_ = REQUEST;
-    factory.CallDeclareInput(*this);
-    requestMode_ = prev;
-  }
-
-  void Level::Release(const FactoryBase& factory) {
-    RequestMode prev = requestMode_;
-    requestMode_ = RELEASE;
-    factory.CallDeclareInput(*this);
-    requestMode_ = prev;
+  // Set the flag
+  map_[factory][ename]->AddKeepFlag(keep);
+}
+
+void Level::RemoveKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep) {
+  // No entry = nothing to do
+  if (!IsKey(factory, ename))
+    return;
+
+  // Remove the flag
+  Teuchos::RCP<MueLu::VariableContainer>& v = map_[factory][ename];
+  v->RemoveKeepFlag(keep);
+
+  // Remove data if no keep flag left and counter == 0
+  if ((v->IsRequested() == false) && (v->GetKeepFlag() == 0)) {
+    v = Teuchos::null;  // free data
+
+    map_[factory].erase(ename);
+    if (map_.count(factory) == 0)
+      map_.erase(factory);
   }
+}
+
+KeepType Level::GetKeepFlag(const std::string& ename, const FactoryBase* factory) const {
+  if (!IsKey(factory, ename))
+    return false;
+
+  return Get(factory, ename)->GetKeepFlag();
+}
+
+void Level::Request(const FactoryBase& factory) {
+  RequestMode prev = requestMode_;
+  requestMode_     = REQUEST;
+  factory.CallDeclareInput(*this);
+  requestMode_ = prev;
+}
+
+void Level::Release(const FactoryBase& factory) {
+  RequestMode prev = requestMode_;
+  requestMode_     = RELEASE;
+  factory.CallDeclareInput(*this);
+  requestMode_ = prev;
+}
+
+void Level::DeclareInput(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
+  if (requestMode_ == REQUEST) {
+    try {
+      Request(ename, factory, requestedBy);
+
+    } catch (Exceptions::DependencyError& e) {
+      std::ostringstream msg;
+      msg << requestedBy->ShortClassName() << "::DeclareInput: (" << e.what() << ") unable to find or generate requested data \""
+          << ename << "\" with generating factory \"" << ((factory != NULL) ? factory->ShortClassName() : "null") << "\" [" << factory << "]";
+      msg << "\n    during request for data \"" << std::setw(15) << ename << "\" on level " << GetLevelID()
+          << " by factory " << std::setw(25) << requestedBy->ShortClassName() << " [" << requestedBy << "]";
+      throw Exceptions::RuntimeError(msg.str());
+
+    } catch (Exceptions::RuntimeError& e) {
+      std::ostringstream msg;
+      msg << e.what() << "\n    during request for data \"" << std::setw(15) << ename << "\" on level " << GetLevelID()
+          << " by factory " << std::setw(25) << requestedBy->ShortClassName() << " [" << requestedBy << "]";
+      throw Exceptions::RuntimeError(msg.str());
+    }
 
-  void Level::DeclareInput(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
-    if (requestMode_ == REQUEST) {
-      try {
-        Request(ename, factory, requestedBy);
-
-      } catch (Exceptions::DependencyError& e) {
-        std::ostringstream msg;
-        msg << requestedBy->ShortClassName() << "::DeclareInput: (" << e.what() << ") unable to find or generate requested data \""
-            << ename << "\" with generating factory \"" << ((factory != NULL) ? factory->ShortClassName() : "null") << "\" [" << factory << "]";
-        msg << "\n    during request for data \"" << std::setw(15) << ename << "\" on level " << GetLevelID()
-            << " by factory " << std::setw(25) << requestedBy->ShortClassName() << " [" << requestedBy << "]";
-        throw Exceptions::RuntimeError(msg.str());
-
-      } catch (Exceptions::RuntimeError &e) {
-        std::ostringstream msg;
-        msg << e.what() << "\n    during request for data \"" << std::setw(15) << ename << "\" on level " << GetLevelID()
-            << " by factory " << std::setw(25) << requestedBy->ShortClassName() << " [" << requestedBy << "]";
-        throw Exceptions::RuntimeError(msg.str());
-      }
-
-    } else if (requestMode_ == RELEASE) {
-      Release(ename, factory, requestedBy);
-
-    } else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareInput(): requestMode_ undefined.");
+  } else if (requestMode_ == RELEASE) {
+    Release(ename, factory, requestedBy);
+
+  } else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareInput(): requestMode_ undefined.");
+}
+
+void Level::DeclareDependencies(const FactoryBase* factory, bool bRequestOnly, bool bReleaseOnly) {  // TODO: replace bReleaseOnly, bReleaseOnly by one RequestMode enum
+  if (bRequestOnly && bReleaseOnly)
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareDependencies(): Both bRequestOnly and bReleaseOnly set to true makes no sense.");
+
+  if (requestMode_ == REQUEST) {
+    if (bReleaseOnly == false) Request(*factory);
+
+  } else if (requestMode_ == RELEASE) {
+    if (bRequestOnly == false) Release(*factory);
+
+  } else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareDependencies(): requestMode_ undefined.");
+}
+
+void Level::Request(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
+  const FactoryBase* fac = GetFactory(ename, factory);
+  // printf("(l=%d) [%43s] requesting \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
+  // levelID_, requestedBy->description().c_str(), ename.c_str(), factory, fac, fac->description().c_str());
+
+  // We request factory only if necessary, i.e.
+  //   - factory has not been requested before, and
+  //   - we need data which is not available
+  // Let us consider an RAP factory in the reuse scenario. The factory generates "A" (coarse matrix) and
+  // "RAP graph" (pattern of A). Typically, "A" has keep flag Final, which is cleared during next Setup, but
+  // "RAP graph" has flag Keep, which is not cleared as it is part of NextRun. Therefore, during next Setup
+  // we have RAP factory with available "RAP graph" but not available "A".
+  //
+  // During the regular construction phase, we will do a single request: ("A", RAPFactory). Suppose, we used
+  //    bool test = (IsRequestedFactory(fac) == false && IsAvailable(fac) == false);
+  // This is incorrect, as IsAvailable(fac) checks whether there is *any* data generated by factory, which there is
+  // ("A"), and the dependencies of the factory would not be requested despite the need for them (we need fine A, P, and
+  // R to generate Ac even if we know sparsity pattern of Ac).
+  //
+  // On the other hand,
+  //   bool test = (IsRequestedFactory(fac) == false && IsAvailable(ename, fac) == false);
+  // is correct as ("A", fac) is not available (only ("RAP graph", fac) is), and dependent factories would be
+  // properly requested.
+  //
+  // This way, factory is requested only once (because of the IsRequested(fac) check), and only when one of the needed
+  // pieces of data is not availble.
+  bool test = (IsRequestedFactory(fac) == false && IsAvailable(ename, fac) == false);
+
+  // This request must be done before calling Request(*fac) to avoid circular dependency problems.
+  if (!IsKey(fac, ename)) {
+    Teuchos::RCP<MueLu::VariableContainer> newVar = Teuchos::rcp(new MueLu::VariableContainer);
+    map_[fac][ename]                              = newVar;
   }
 
-  void Level::DeclareDependencies(const FactoryBase* factory, bool bRequestOnly, bool bReleaseOnly) { //TODO: replace bReleaseOnly, bReleaseOnly by one RequestMode enum
-    if (bRequestOnly && bReleaseOnly)
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareDependencies(): Both bRequestOnly and bReleaseOnly set to true makes no sense.");
-
-    if (requestMode_ == REQUEST) {
-
-      if (bReleaseOnly == false) Request(*factory);
+  Teuchos::RCP<MueLu::VariableContainer>& v = map_[fac][ename];
+  v->Request(requestedBy);
 
-    } else if (requestMode_ == RELEASE) {
+  // The value of IsRequestedFactory(fac) is true, due to the above request.
+  // That is why a temporary boolean "test" is used!
+  TEUCHOS_TEST_FOR_EXCEPTION(IsRequestedFactory(fac) != true, Exceptions::RuntimeError, "Level::Request(ename, factory): internal logic error.");
 
-      if (bRequestOnly == false) Release(*factory);
-
-    } else TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::Level::DeclareDependencies(): requestMode_ undefined.");
+  if (test) {
+    // Call Request for factory dependencies.
+    // We only do that if necessary, see comments above
+    Request(*fac);
   }
-
-  void Level::Request(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
-    const FactoryBase* fac = GetFactory(ename, factory);
-    // printf("(l=%d) [%43s] requesting \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
-           // levelID_, requestedBy->description().c_str(), ename.c_str(), factory, fac, fac->description().c_str());
-
-    // We request factory only if necessary, i.e.
-    //   - factory has not been requested before, and
-    //   - we need data which is not available
-    // Let us consider an RAP factory in the reuse scenario. The factory generates "A" (coarse matrix) and
-    // "RAP graph" (pattern of A). Typically, "A" has keep flag Final, which is cleared during next Setup, but
-    // "RAP graph" has flag Keep, which is not cleared as it is part of NextRun. Therefore, during next Setup
-    // we have RAP factory with available "RAP graph" but not available "A".
-    //
-    // During the regular construction phase, we will do a single request: ("A", RAPFactory). Suppose, we used
-    //    bool test = (IsRequestedFactory(fac) == false && IsAvailable(fac) == false);
-    // This is incorrect, as IsAvailable(fac) checks whether there is *any* data generated by factory, which there is
-    // ("A"), and the dependencies of the factory would not be requested despite the need for them (we need fine A, P, and
-    // R to generate Ac even if we know sparsity pattern of Ac).
-    //
-    // On the other hand,
-    //   bool test = (IsRequestedFactory(fac) == false && IsAvailable(ename, fac) == false);
-    // is correct as ("A", fac) is not available (only ("RAP graph", fac) is), and dependent factories would be
-    // properly requested.
-    //
-    // This way, factory is requested only once (because of the IsRequested(fac) check), and only when one of the needed
-    // pieces of data is not availble.
-    bool test = (IsRequestedFactory(fac) == false && IsAvailable(ename, fac) == false);
-
-    // This request must be done before calling Request(*fac) to avoid circular dependency problems.
-    if (!IsKey(fac, ename)) {
-      Teuchos::RCP<MueLu::VariableContainer> newVar = Teuchos::rcp(new MueLu::VariableContainer);
-      map_[fac][ename] = newVar;
+}
+
+void Level::Release(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
+  const FactoryBase* fac = GetFactory(ename, factory);
+  // printf("(l=%d) [%43s] releasing  \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
+  // levelID_, requestedBy->description().c_str(), ename.c_str(), factory, fac, fac->description().c_str());
+
+  // Only a factory which has requested (fac,ename) is allowed to release it again.
+  // Do not release data if it has not been requested by the factory "requestedBy"
+  // Note: when data is released (fac,ename) depends on it often happened that some
+  //       of this data has (recursively) been released too often
+  if (IsRequestedBy(fac, ename, requestedBy)) {
+    // In general all data (fac,ename) depends on is released when calling Get in generating factory (fac) Build method
+    // Here we check the need to release the dependencies of some data that has been requested (by factory "requestedBy")
+    // but the corresponding Build function of factory "fac" has never been called. Therefore the dependencies
+    // have never been released. Do it now.
+    if (CountRequestedFactory(fac) == 1 &&   // check if factory fac is not requested by another factory
+        IsAvailableFactory(fac) == false) {  // check if Build function of factory fac has been called
+      Release(*fac);
     }
 
-    Teuchos::RCP<MueLu::VariableContainer>& v = map_[fac][ename];
-    v->Request(requestedBy);
-
-    // The value of IsRequestedFactory(fac) is true, due to the above request.
-    // That is why a temporary boolean "test" is used!
-    TEUCHOS_TEST_FOR_EXCEPTION(IsRequestedFactory(fac) != true, Exceptions::RuntimeError, "Level::Request(ename, factory): internal logic error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" not found. Do a request first.");
 
-    if (test) {
-      // Call Request for factory dependencies.
-      // We only do that if necessary, see comments above
-      Request(*fac);
-    }
-  }
-
-  void Level::Release(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy) {
-    const FactoryBase* fac = GetFactory(ename, factory);
-    // printf("(l=%d) [%43s] releasing  \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
-           // levelID_, requestedBy->description().c_str(), ename.c_str(), factory, fac, fac->description().c_str());
-
-    // Only a factory which has requested (fac,ename) is allowed to release it again.
-    // Do not release data if it has not been requested by the factory "requestedBy"
-    // Note: when data is released (fac,ename) depends on it often happened that some
-    //       of this data has (recursively) been released too often
-    if (IsRequestedBy(fac, ename, requestedBy)) {
-
-      // In general all data (fac,ename) depends on is released when calling Get in generating factory (fac) Build method
-      // Here we check the need to release the dependencies of some data that has been requested (by factory "requestedBy")
-      // but the corresponding Build function of factory "fac" has never been called. Therefore the dependencies
-      // have never been released. Do it now.
-      if (CountRequestedFactory(fac) == 1      &&     // check if factory fac is not requested by another factory
-          IsAvailableFactory(fac)    == false) {      // check if Build function of factory fac has been called
-        Release(*fac);
-      }
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac,ename), Exceptions::RuntimeError, "\"" + ename + "\" not found. Do a request first.");
-
-      Teuchos::RCP<MueLu::VariableContainer>& v = map_[fac][ename];
-      v->Release(requestedBy);
+    Teuchos::RCP<MueLu::VariableContainer>& v = map_[fac][ename];
+    v->Release(requestedBy);
 
-      // Remove data if no keep flag left and counter == 0
-      if ((v->IsRequested() == false) && (v->GetKeepFlag() == 0)) {
-        v = Teuchos::null; // free data
+    // Remove data if no keep flag left and counter == 0
+    if ((v->IsRequested() == false) && (v->GetKeepFlag() == 0)) {
+      v = Teuchos::null;  // free data
 
-        map_[fac].erase(ename);
-        if (map_.count(fac) == 0)
-          map_.erase(fac);
-      }
+      map_[fac].erase(ename);
+      if (map_.count(fac) == 0)
+        map_.erase(fac);
     }
   }
+}
+
+void Level::Clear() {
+  // TODO: needs some love, ugly as it is
+  // The ugliness is the fact that we restart both loops when we remove a single element
+  bool wasRemoved;
+  do {
+    wasRemoved = false;
+    for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
+      const FactoryBase* factory = kt->first;
 
-  void Level::Clear() {
-    // TODO: needs some love, ugly as it is
-    // The ugliness is the fact that we restart both loops when we remove a single element
-    bool wasRemoved;
-    do {
-      wasRemoved = false;
-      for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
-        const FactoryBase* factory = kt->first;
-
-        for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
-          // We really want a reference here, but because later we'll need to check whether the
-          // key was removed, we should copy the value
-          const std::string ename = it->first;
-
-          // We clear all the data that
-          //   a) has not been requested
-          //   b) is not being kept using NextRun (e.g., we clear out Final data)
-          if (!IsKept(ename, factory, MueLu::NextRun)) {
-            RemoveKeepFlag(ename, factory, MueLu::All); // will delete the data if counter == 0
-
-            // To prevent infinite looping, we need to check whether we have
-            // actually removed the data. In buggy code it may happen that we
-            // were unable to do that, for instance, if the data was outstanding
-            // request
-            if (IsKey(factory, ename)) {
-              GetOStream(Errors) << "Level::Clear found Internal data inconsistency" << std::endl;
-              print(GetOStream(Errors), VERB_EXTREME);
-
-              throw Exceptions::RuntimeError("Level::Clear found Internal data inconsistency");
-            }
-
-            wasRemoved = true;
-            break;
+      for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
+        // We really want a reference here, but because later we'll need to check whether the
+        // key was removed, we should copy the value
+        const std::string ename = it->first;
+
+        // We clear all the data that
+        //   a) has not been requested
+        //   b) is not being kept using NextRun (e.g., we clear out Final data)
+        if (!IsKept(ename, factory, MueLu::NextRun)) {
+          RemoveKeepFlag(ename, factory, MueLu::All);  // will delete the data if counter == 0
+
+          // To prevent infinite looping, we need to check whether we have
+          // actually removed the data. In buggy code it may happen that we
+          // were unable to do that, for instance, if the data was outstanding
+          // request
+          if (IsKey(factory, ename)) {
+            GetOStream(Errors) << "Level::Clear found Internal data inconsistency" << std::endl;
+            print(GetOStream(Errors), VERB_EXTREME);
+
+            throw Exceptions::RuntimeError("Level::Clear found Internal data inconsistency");
           }
-        }
 
-        if (wasRemoved)
+          wasRemoved = true;
           break;
+        }
       }
 
-    } while (wasRemoved == true);
-  }
-
-  void Level::ExpertClear() {
-    TwoKeyMap::const_iterator kt = map_.begin();
-    while (kt != map_.end()) {
-      const FactoryBase* factory = kt->first;
+      if (wasRemoved)
+        break;
+    }
 
-      SubMap::const_iterator it = kt->second.begin();
-      while ( it != kt->second.end()) {
-        const std::string& ename = it->first;
-
-        // obtain variable container
-        Teuchos::RCP<MueLu::VariableContainer>& v = map_[factory][ename];
-
-        if (v->GetKeepFlag() == 0 ||
-            v->IsKept(MueLu::UserData) == true ||
-            v->IsKept(MueLu::Final) == true) {
-          it++;
-          v = Teuchos::null; // free data
-          map_[factory].erase(ename);
-          if (map_.count(factory) == 0) {
-            break; // last occurrence for factory has been removed. proceed with next factory
-          }
+  } while (wasRemoved == true);
+}
+
+void Level::ExpertClear() {
+  TwoKeyMap::const_iterator kt = map_.begin();
+  while (kt != map_.end()) {
+    const FactoryBase* factory = kt->first;
+
+    SubMap::const_iterator it = kt->second.begin();
+    while (it != kt->second.end()) {
+      const std::string& ename = it->first;
+
+      // obtain variable container
+      Teuchos::RCP<MueLu::VariableContainer>& v = map_[factory][ename];
+
+      if (v->GetKeepFlag() == 0 ||
+          v->IsKept(MueLu::UserData) == true ||
+          v->IsKept(MueLu::Final) == true) {
+        it++;
+        v = Teuchos::null;  // free data
+        map_[factory].erase(ename);
+        if (map_.count(factory) == 0) {
+          break;  // last occurrence for factory has been removed. proceed with next factory
         }
-        else
-          it++;
-      } // end for it
+      } else
+        it++;
+    }  // end for it
 
-      if (map_.count(factory) == 0) {
-        kt++;
-        map_.erase(factory);
-      } else kt++;
-    }
-  }
-
-  std::string Level::description() const {
-    std::ostringstream out;
-    out << BaseClass::description();
-    out << "{ levelID = " << levelID_ << "}";
-    return out.str();
+    if (map_.count(factory) == 0) {
+      kt++;
+      map_.erase(factory);
+    } else
+      kt++;
   }
+}
+
+std::string Level::description() const {
+  std::ostringstream out;
+  out << BaseClass::description();
+  out << "{ levelID = " << levelID_ << "}";
+  return out.str();
+}
+
+void Level::print(std::ostream& out, const VerbLevel verbLevel) const {
+  if (!(verbLevel & Debug))
+    return;
+
+  out << "LevelID = " << GetLevelID() << std::endl;
+
+  typedef Teuchos::TabularOutputter TTO;
+  TTO outputter(out);
+  outputter.pushFieldSpec("data name", TTO::STRING, TTO::LEFT, TTO::GENERAL, 20);
+  outputter.pushFieldSpec("gen. factory addr.", TTO::STRING, TTO::LEFT, TTO::GENERAL, 40);
+  outputter.pushFieldSpec("req", TTO::INT, TTO::LEFT, TTO::GENERAL, 3);
+  outputter.pushFieldSpec("keep", TTO::STRING, TTO::LEFT, TTO::GENERAL, 5);
+  outputter.pushFieldSpec("type", TTO::STRING, TTO::LEFT, TTO::GENERAL, 18);
+  outputter.pushFieldSpec("data", TTO::STRING, TTO::LEFT, TTO::GENERAL, 14);
+  outputter.pushFieldSpec("req'd by", TTO::STRING, TTO::LEFT, TTO::GENERAL, 20);
+  outputter.outputHeader();
+
+  for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
+    const FactoryBase* factory = kt->first;
+
+    for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
+      const std::string& ename = it->first;
+
+      outputter.outputField(ename);  // variable name
+
+      // NOTE: We cannot dereference the factory pointer and call factory->description() as we do not know
+      // if the factory still exist (the factory pointer is a raw pointer by design). Instead, the level
+      // should store the factory description internally as a string for debugging purpose (and in debug mode only).
+      //         // factory name
+      //         std::stringstream ss1;
+      //         ss1 << (*kt)->description();
+      //         outputter.outputField((ss1.str()).substr(0,30));
+      if (factory == 0) {
+        outputter.outputField("Null");
+      } else if (factory == NoFactory::get()) {
+        outputter.outputField("NoFactory");
+      } else {
+        std::ostringstream oss;
+        oss << factory->ShortClassName() << "[" << factory->GetID() << "]";
+#ifdef HAVE_MUELU_DEBUG
+        oss << "(" << factory << ")";
+#endif
+        outputter.outputField(oss.str());
+      }
 
-  void Level::print(std::ostream& out, const VerbLevel verbLevel) const {
-    if (!(verbLevel & Debug))
-      return;
-
-    out << "LevelID = " << GetLevelID() << std::endl;
-
-    typedef Teuchos::TabularOutputter TTO;
-    TTO outputter(out);
-    outputter.pushFieldSpec("data name",                TTO::STRING, TTO::LEFT, TTO::GENERAL, 20);
-    outputter.pushFieldSpec("gen. factory addr.",       TTO::STRING, TTO::LEFT, TTO::GENERAL, 40);
-    outputter.pushFieldSpec("req",                      TTO::INT,    TTO::LEFT, TTO::GENERAL, 3);
-    outputter.pushFieldSpec("keep",                     TTO::STRING, TTO::LEFT, TTO::GENERAL, 5);
-    outputter.pushFieldSpec("type",                     TTO::STRING, TTO::LEFT, TTO::GENERAL, 18);
-    outputter.pushFieldSpec("data",                     TTO::STRING, TTO::LEFT, TTO::GENERAL, 14);
-    outputter.pushFieldSpec("req'd by",                 TTO::STRING, TTO::LEFT, TTO::GENERAL, 20);
-    outputter.outputHeader();
-
-    for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
-      const FactoryBase* factory = kt->first;
+      int reqcount = NumRequests(factory, ename);  // request counter
+      outputter.outputField(reqcount);
 
-      for (SubMap::const_iterator it = kt->second.begin(); it != kt->second.end(); it++) {
-        const std::string& ename = it->first;
-
-        outputter.outputField(ename);   // variable name
-
-        // NOTE: We cannot dereference the factory pointer and call factory->description() as we do not know
-        // if the factory still exist (the factory pointer is a raw pointer by design). Instead, the level
-        // should store the factory description internally as a string for debugging purpose (and in debug mode only).
-        //         // factory name
-        //         std::stringstream ss1;
-        //         ss1 << (*kt)->description();
-        //         outputter.outputField((ss1.str()).substr(0,30));
-        if(factory == 0) {
-          outputter.outputField("Null");
+      KeepType keepType = GetKeepFlag(ename, factory);
+      if (keepType != 0) {
+        std::stringstream ss;
+        if (keepType & MueLu::UserData) {
+          ss << "User";
         }
-        else if (factory == NoFactory::get()) {
-          outputter.outputField("NoFactory");
+        if (keepType & MueLu::Keep) {
+          ss << "Keep";
         }
-        else {
-          std::ostringstream oss; oss << factory->ShortClassName() << "["<<factory->GetID()<<"]";
-#ifdef HAVE_MUELU_DEBUG
-          oss<<"(" << factory <<")";
-#endif
-          outputter.outputField(oss.str());
+        if (keepType & MueLu::Final) {
+          ss << "Final";
         }
+        outputter.outputField(ss.str());
+      } else {
+        outputter.outputField("No");
+      }
 
-        int reqcount = NumRequests(factory, ename); // request counter
-        outputter.outputField(reqcount);
-
-        KeepType keepType = GetKeepFlag(ename, factory);
-        if (keepType != 0) {
-          std::stringstream ss;
-          if (keepType & MueLu::UserData) { ss << "User";  }
-          if (keepType & MueLu::Keep)     { ss << "Keep";  }
-          if (keepType & MueLu::Final)    { ss << "Final"; }
-          outputter.outputField(ss.str());
+      if (IsAvailable(ename, factory)) {
+        std::string strType = it->second->GetTypeName();
+
+        if (strType == "int") {
+          outputter.outputField(strType);
+          outputter.outputField(it->second->GetData<int>());
+        } else if (strType == "double") {
+          outputter.outputField(strType);
+          outputter.outputField(it->second->GetData<double>());
+        } else if (strType == "string") {
+          outputter.outputField(strType);
+          outputter.outputField(it->second->GetData<std::string>());
         } else {
-          outputter.outputField("No");
-        }
-
-        if (IsAvailable(ename, factory)) {
-          std::string strType = it->second->GetTypeName();
-
-          if        (strType == "int") {
-            outputter.outputField(strType);
-            outputter.outputField(it->second->GetData<int>());
-          } else if (strType == "double") {
-            outputter.outputField(strType);
-            outputter.outputField(it->second->GetData<double>());
-          } else if (strType == "string") {
+          size_t npos = std::string::npos;
+
+          if (strType.find("MueLu::Aggregates") != npos)
+            outputter.outputField("Aggregates");
+          else if (strType.find("MueLu::AmalgamationInfo") != npos)
+            outputter.outputField("AmalgamationInfo");
+          else if (strType.find("MueLu::Constraint") != npos)
+            outputter.outputField("Constraint");
+          else if (strType.find("MueLu::Graph") != npos)
+            outputter.outputField("Graph");
+          else if (strType.find("MueLu::SmootherBase") != npos)
+            outputter.outputField("SmootherBase");
+          else if (strType.find("MueLu::SmootherPrototype") != npos)
+            outputter.outputField("SmootherPrototype");
+          else if (strType.find("Xpetra::Export") != npos)
+            outputter.outputField("Export");
+          else if (strType.find("Xpetra::Import") != npos)
+            outputter.outputField("Import");
+          else if (strType.find("Xpetra::Map") != npos)
+            outputter.outputField("Map");
+          else if (strType.find("Xpetra::Matrix") != npos)
+            outputter.outputField("Matrix");
+          else if (strType.find("Xpetra::MultiVector") != npos)
+            outputter.outputField("Vector");
+          else if (strType.find("Xpetra::Operator") != npos)
+            outputter.outputField("Operator");
+          else
             outputter.outputField(strType);
-            outputter.outputField(it->second->GetData<std::string>());
-          } else {
-            size_t npos = std::string::npos;
-
-            if      (strType.find("MueLu::Aggregates")          != npos) outputter.outputField("Aggregates");
-            else if (strType.find("MueLu::AmalgamationInfo")    != npos) outputter.outputField("AmalgamationInfo");
-            else if (strType.find("MueLu::Constraint")          != npos) outputter.outputField("Constraint");
-            else if (strType.find("MueLu::Graph")               != npos) outputter.outputField("Graph");
-            else if (strType.find("MueLu::SmootherBase")        != npos) outputter.outputField("SmootherBase");
-            else if (strType.find("MueLu::SmootherPrototype")   != npos) outputter.outputField("SmootherPrototype");
-            else if (strType.find("Xpetra::Export")             != npos) outputter.outputField("Export");
-            else if (strType.find("Xpetra::Import")             != npos) outputter.outputField("Import");
-            else if (strType.find("Xpetra::Map")                != npos) outputter.outputField("Map");
-            else if (strType.find("Xpetra::Matrix")             != npos) outputter.outputField("Matrix" );
-            else if (strType.find("Xpetra::MultiVector")        != npos) outputter.outputField("Vector");
-            else if (strType.find("Xpetra::Operator")           != npos) outputter.outputField("Operator");
-            else                                                         outputter.outputField(strType);
-
-            outputter.outputField("available");
-          }
 
-        } else {
-          outputter.outputField("unknown");
-          outputter.outputField("not available");
+          outputter.outputField("available");
         }
 
-        typedef VariableContainer::request_container container_type;
-        const container_type& requestedBy = it->second->Requests();
-        std::ostringstream ss;
-        for (container_type::const_iterator ct = requestedBy.begin(); ct != requestedBy.end(); ct++) {
-          if (ct != requestedBy.begin()) ss << ",";
-          ss << ct->first->ShortClassName() <<"["<<ct->first->GetID()<<"]";
+      } else {
+        outputter.outputField("unknown");
+        outputter.outputField("not available");
+      }
+
+      typedef VariableContainer::request_container container_type;
+      const container_type& requestedBy = it->second->Requests();
+      std::ostringstream ss;
+      for (container_type::const_iterator ct = requestedBy.begin(); ct != requestedBy.end(); ct++) {
+        if (ct != requestedBy.begin()) ss << ",";
+        ss << ct->first->ShortClassName() << "[" << ct->first->GetID() << "]";
 #ifdef HAVE_MUELU_DEBUG
-          ss<<"("<<ct->first<<")";
+        ss << "(" << ct->first << ")";
 #endif
 
-          if (ct->second > 1) ss << "x" << ct->second;
-        }
-        outputter.outputField(ss.str());
-
-        outputter.nextRow();
+        if (ct->second > 1) ss << "x" << ct->second;
       }
-    } // for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
-  }
+      outputter.outputField(ss.str());
+
+      outputter.nextRow();
+    }
+  }  // for (TwoKeyMap::const_iterator kt = map_.begin(); kt != map_.end(); kt++) {
+}
 
 #if defined(HAVE_MUELU_BOOST) && defined(HAVE_MUELU_BOOST_FOR_REAL) && defined(BOOST_VERSION) && (BOOST_VERSION >= 104400)
-    void Level::UpdateGraph(std::map<const FactoryBase*, BoostVertex>&                   vindices,
-                            std::map<std::pair<BoostVertex, BoostVertex>, std::string>&  edges,
-                            BoostProperties&                                             dp,
-                            BoostGraph&                                                  graph) const {
-      size_t vind = vindices.size();
-
-      for (TwoKeyMap::const_iterator it1 = map_.begin(); it1 != map_.end(); it1++) {
-        if (vindices.find(it1->first) == vindices.end()) {
+void Level::UpdateGraph(std::map<const FactoryBase*, BoostVertex>& vindices,
+                        std::map<std::pair<BoostVertex, BoostVertex>, std::string>& edges,
+                        BoostProperties& dp,
+                        BoostGraph& graph) const {
+  size_t vind = vindices.size();
+
+  for (TwoKeyMap::const_iterator it1 = map_.begin(); it1 != map_.end(); it1++) {
+    if (vindices.find(it1->first) == vindices.end()) {
+      BoostVertex boost_vertex = boost::add_vertex(graph);
+      std::ostringstream oss;
+      oss << it1->first->ShortClassName() << "[" << it1->first->GetID() << "]";
+      boost::put("label", dp, boost_vertex, oss.str());
+      vindices[it1->first] = vind++;
+    }
+
+    for (SubMap::const_iterator it2 = it1->second.begin(); it2 != it1->second.end(); it2++) {
+      const VariableContainer::request_container& requests = it2->second->Requests();
+      for (VariableContainer::request_container::const_iterator rit = requests.begin(); rit != requests.end(); rit++) {
+        if (vindices.find(rit->first) == vindices.end()) {
+          // requested by factory which is unknown
           BoostVertex boost_vertex = boost::add_vertex(graph);
-          std::ostringstream oss; oss<<it1->first->ShortClassName() << "[" << it1->first->GetID() << "]";
+          std::ostringstream oss;
+          oss << rit->first->ShortClassName() << "[" << rit->first->GetID() << "]";
           boost::put("label", dp, boost_vertex, oss.str());
-          vindices[it1->first] = vind++;
+          vindices[rit->first] = vind++;
         }
 
-        for (SubMap::const_iterator it2 = it1->second.begin(); it2 != it1->second.end(); it2++) {
-          const VariableContainer::request_container& requests = it2->second->Requests();
-          for (VariableContainer::request_container::const_iterator rit = requests.begin(); rit != requests.end(); rit++) {
-            if (vindices.find(rit->first) == vindices.end()) {
-              // requested by factory which is unknown
-              BoostVertex boost_vertex = boost::add_vertex(graph);
-              std::ostringstream oss; oss<<rit->first->ShortClassName() << "[" << rit->first->GetID() << "]";
-	      boost::put("label", dp, boost_vertex, oss.str());
-              vindices[rit->first] = vind++;
-            }
-
-	    edges[std::pair<BoostVertex,BoostVertex>(vindices[rit->first], vindices[it1->first])] =  it2->first;
-          }
-        }
+        edges[std::pair<BoostVertex, BoostVertex>(vindices[rit->first], vindices[it1->first])] = it2->first;
       }
     }
+  }
+}
 #endif
 
-  // JG Note: should the option IgnoreUserData() moved to the Factory interface or on the specific factories that are using this option? It would simplify the level class.
-  const FactoryBase* Level::GetFactory(const std::string& ename, const FactoryBase* factory) const {
-    if (factory != NULL)
-      return factory;
-
-    // If IgnoreUserData == false and if variable "ename" generated by NoFactory is provided by the user (MueLu::UserData),
-    // use user-provided data by default without querying the FactoryManager.
-    // When FactoryManager == null, we consider that IgnoreUserData == false.
-    if ((factoryManager_ == Teuchos::null || factoryManager_->IgnoreUserData() == false) &&
-        (IsAvailable(ename, NoFactory::get()) && IsKept(ename, NoFactory::get(), MueLu::UserData))) {
-      return NoFactory::get();
-    }
-
-    // Query factory manager
-    TEUCHOS_TEST_FOR_EXCEPTION(factoryManager_ == null, Exceptions::RuntimeError, "MueLu::Level("<< levelID_ << ")::GetFactory(" << ename << ", " << factory << "): No FactoryManager");
-    const FactoryBase* fac = factoryManager_->GetFactory(ename).get();
-    TEUCHOS_TEST_FOR_EXCEPTION(fac == NULL, Exceptions::RuntimeError, "MueLu::Level("<< levelID_ << ")::GetFactory(" << ename << ", " << factory << "): Default factory returned by FactoryManager cannot be NULL");
-    return fac;
+// JG Note: should the option IgnoreUserData() moved to the Factory interface or on the specific factories that are using this option? It would simplify the level class.
+const FactoryBase* Level::GetFactory(const std::string& ename, const FactoryBase* factory) const {
+  if (factory != NULL)
+    return factory;
+
+  // If IgnoreUserData == false and if variable "ename" generated by NoFactory is provided by the user (MueLu::UserData),
+  // use user-provided data by default without querying the FactoryManager.
+  // When FactoryManager == null, we consider that IgnoreUserData == false.
+  if ((factoryManager_ == Teuchos::null || factoryManager_->IgnoreUserData() == false) &&
+      (IsAvailable(ename, NoFactory::get()) && IsKept(ename, NoFactory::get(), MueLu::UserData))) {
+    return NoFactory::get();
   }
 
-  Level::RequestMode Level::requestMode_ = UNDEF;
+  // Query factory manager
+  TEUCHOS_TEST_FOR_EXCEPTION(factoryManager_ == null, Exceptions::RuntimeError, "MueLu::Level(" << levelID_ << ")::GetFactory(" << ename << ", " << factory << "): No FactoryManager");
+  const FactoryBase* fac = factoryManager_->GetFactory(ename).get();
+  TEUCHOS_TEST_FOR_EXCEPTION(fac == NULL, Exceptions::RuntimeError, "MueLu::Level(" << levelID_ << ")::GetFactory(" << ename << ", " << factory << "): Default factory returned by FactoryManager cannot be NULL");
+  return fac;
+}
+
+Level::RequestMode Level::requestMode_ = UNDEF;
 
-} //namespace MueLu
+}  // namespace MueLu
 
-//TODO: Caps should not matter
+// TODO: Caps should not matter
diff --git a/packages/muelu/src/MueCentral/MueLu_Level.hpp b/packages/muelu/src/MueCentral/MueLu_Level.hpp
index d8718eb0e5ec..2548e8ff7d7c 100644
--- a/packages/muelu/src/MueCentral/MueLu_Level.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_Level.hpp
@@ -46,489 +46,497 @@
 #ifndef MUELU_LEVEL_HPP
 #define MUELU_LEVEL_HPP
 
-#include <algorithm>                    // for swap
-#include <map>                          // for _Rb_tree_const_iterator, etc
-#include <ostream>                      // for basic_ostream, etc
-#include <string>                       // for char_traits, string, etc
-#include <utility>                      // for pair
+#include <algorithm>  // for swap
+#include <map>        // for _Rb_tree_const_iterator, etc
+#include <ostream>    // for basic_ostream, etc
+#include <string>     // for char_traits, string, etc
+#include <utility>    // for pair
 
-#include <Teuchos_Describable.hpp>      // for operator<<
-#include <Teuchos_FancyOStream.hpp>     // for FancyOStream
-#include <Teuchos_RCPDecl.hpp>          // for RCP
-#include <Teuchos_RCP.hpp>              // for RCP::operator->, etc
-#include <Teuchos_TestForException.hpp> // for TEUCHOS_TEST_FOR_EXCEPTION
+#include <Teuchos_Describable.hpp>       // for operator<<
+#include <Teuchos_FancyOStream.hpp>      // for FancyOStream
+#include <Teuchos_RCPDecl.hpp>           // for RCP
+#include <Teuchos_RCP.hpp>               // for RCP::operator->, etc
+#include <Teuchos_TestForException.hpp>  // for TEUCHOS_TEST_FOR_EXCEPTION
 
-#include <Xpetra_Map.hpp>               // for UnderlyingLib definition
+#include <Xpetra_Map.hpp>  // for UnderlyingLib definition
 
 #include "MueLu_BoostGraphviz.hpp"
-#include "MueLu_Exceptions.hpp"         // for RuntimeError
+#include "MueLu_Exceptions.hpp"  // for RuntimeError
 #include "MueLu_FactoryManagerBase_fwd.hpp"
 #include "MueLu_KeepType.hpp"
 #include "MueLu_NoFactory.hpp"
 #include "MueLu_Utilities.hpp"
 #include "MueLu_VariableContainer.hpp"
-#include "MueLu_VerbosityLevel.hpp"     // for MsgType::Default, VerbLevel
+#include "MueLu_VerbosityLevel.hpp"  // for MsgType::Default, VerbLevel
 
 namespace MueLu {
 
-  /*!
-    @class Level
-    @brief Class that holds all level-specific information.
-
-    All data that is stored in the <tt>Level</tt> class need a variable name
-    (e.g. "A", "P", ...) and a pointer to the generating factory. Only with
-    both the variable name and the generating factory the data can be accessed.
-
-    If no pointer to the generating factory is provided (or it is NULL) then
-    the Level class uses the information from a factory manager, which stores
-    default factories for different variable names.
-
-    We use a "two key" map for storage of data, with the pointer to the
-    generating factory as primary key and the variable name as secondary key.
-    The pointer to the generating factory is only used as primary "key".  It
-    doesn't matter if the given factory pointer is valid or not. So the NULL
-    pointer can also be used.
-
-    The data itself is stored within a VariableContainer object.  A reference
-    counter keeps track of the storage and automatically frees the memory if
-    the data is not needed any more. In the standard mode, the data first has
-    to be requested by calling the Request function. Then the data can be set
-    by calling Set.  With Get the user can fetch data when needed. Release
-    decrements the reference counter for the given variable.
-  */
-  class Level : public BaseClass {
-
-  public:
-    //@{
-
-    //! @name Constructors / Destructors
-
-    Level() : lib_(Xpetra::NotSpecified), levelID_(-1) { }
-
-    Level(RCP<FactoryManagerBase>& factoryManager) : lib_(Xpetra::UseTpetra), levelID_(-1), factoryManager_(factoryManager) { }
-
-    //! Destructor
-    virtual ~Level() { }
-
-    //@}
-
-    //@{
-    //! @name Build methods
-    //! Builds a new Level object.
-    RCP<Level> Build();
-
-    //@}
-
-    //@{
-    //! @name Level handling
-
-    //! @brief Return level number.
-    int GetLevelID() const;
-
-    //! @brief Set level number.
-    void SetLevelID(int levelID);
-
-    //! Previous level
-    RCP<Level>& GetPreviousLevel() { return previousLevel_; }
-
-    //! Set previous level object
-    //! @\param[in] const RCP<Level>& previousLevel
-    void SetPreviousLevel(const RCP<Level>& previousLevel);
-    //@}
-
-    //! @name Set/Get factory manager
-    //@{
-    //! Set default factories (used internally by Hierarchy::SetLevel()).
-    // Users should not use this method.
-    void SetFactoryManager(const RCP<const FactoryManagerBase>& factoryManager);
-
-    //! returns the current factory manager
-    // Users should not use this method
-    const RCP<const FactoryManagerBase> GetFactoryManager();
-    //@}
-
-    //@{
-    //! @name Set functions
-
-    //! Store need label and its associated data. This does not increment the storage counter.
-    //! - If factory is not specified or factory == NoFactory::get(), mark data as user-defined by default
-    //! - If factory is == NULL, use defaultFactory (FactoryManager have to be available).
-    template <class T>
-    void Set(const std::string& ename, const T& entry, const FactoryBase* factory = NoFactory::get()) {
-      const FactoryBase* fac = GetFactory(ename, factory);
-
-      if (fac == NoFactory::get()) {
-        // Any data set with a NoFactory gets UserData keep flag by default
-        AddKeepFlag(ename, NoFactory::get(), MueLu::UserData);
-      }
-
-      // Store entry only if data have been requested (or any keep flag)
-      if (IsRequested(ename, factory) || GetKeepFlag(ename, factory) != 0) {
-        TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(factory, ename), Exceptions::RuntimeError, "" + ename + " not found in");
-        map_[factory][ename]->SetData(entry);
-
-      } else {
-        GetOStream(Runtime1) << "Level::Set: Not storing \"" << ename << "\" generated by factory " << factory->ShortClassName() << "["<<factory->GetID()<<"]"
-                             << " on level " << toString(GetLevelID()) << ", as it has not been requested and no keep flags were set for it" << std::endl;
-      }
-    } // Set
-
-    //@}
-
-    //! @name Get functions
-    //! @brief Get functions for accessing stored data
-
-    //@{
-    /*! @brief Get data without decrementing associated storage counter (i.e., read-only access).
-     *   Usage: Level->Get< RCP<Matrix> >("A", factory)
-     *   if factory == NULL => use default factory
-     *
-     *  @param[in] const std::string& ename
-     *  @param[in] const FactoryBase* factory
-     *  @return data (templated)
-     * */
-    template <class T>
-    T& Get(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
-      const FactoryBase* fac = GetFactory(ename, factory);
-/*      printf("(l=%d)                                               getting    \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
-        levelID_, ename.c_str(), factory, fac, fac->description().c_str());*/
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" generated by factory \"" + fac->description() + "\" not found on level " + toString(GetLevelID()) + ".");
-
-      if (!IsAvailable(ename, fac)) {
-        TEUCHOS_TEST_FOR_EXCEPTION(NumRequests(fac, ename) < 1 && GetKeepFlag(ename, fac) == 0, Exceptions::RuntimeError,
-                                   "\"" << ename << "\" has not been requested (counter = " << NumRequests(fac, ename) << ", "
-                                   "KeepFlag = " << GetKeepFlag(ename, fac) << "). " << std::endl <<
-                                   "Generating factory:" << *fac << " NoFactory = " << NoFactory::get());
-        fac->CallBuild(*this);
-        Release(*fac);
-      }
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError,
-          "MueLu::Level::Get(): factory did not produce expected output on level " << GetLevelID()
-              << ". The data \"" << ename << "\" has not been generated by " << *fac);
-
-      return map_[fac][ename]->template GetData<T>();
+/*!
+  @class Level
+  @brief Class that holds all level-specific information.
+
+  All data that is stored in the <tt>Level</tt> class need a variable name
+  (e.g. "A", "P", ...) and a pointer to the generating factory. Only with
+  both the variable name and the generating factory the data can be accessed.
+
+  If no pointer to the generating factory is provided (or it is NULL) then
+  the Level class uses the information from a factory manager, which stores
+  default factories for different variable names.
+
+  We use a "two key" map for storage of data, with the pointer to the
+  generating factory as primary key and the variable name as secondary key.
+  The pointer to the generating factory is only used as primary "key".  It
+  doesn't matter if the given factory pointer is valid or not. So the NULL
+  pointer can also be used.
+
+  The data itself is stored within a VariableContainer object.  A reference
+  counter keeps track of the storage and automatically frees the memory if
+  the data is not needed any more. In the standard mode, the data first has
+  to be requested by calling the Request function. Then the data can be set
+  by calling Set.  With Get the user can fetch data when needed. Release
+  decrements the reference counter for the given variable.
+*/
+class Level : public BaseClass {
+ public:
+  //@{
+
+  //! @name Constructors / Destructors
+
+  Level()
+    : lib_(Xpetra::NotSpecified)
+    , levelID_(-1) {}
+
+  Level(RCP<FactoryManagerBase>& factoryManager)
+    : lib_(Xpetra::UseTpetra)
+    , levelID_(-1)
+    , factoryManager_(factoryManager) {}
+
+  //! Destructor
+  virtual ~Level() {}
+
+  //@}
+
+  //@{
+  //! @name Build methods
+  //! Builds a new Level object.
+  RCP<Level> Build();
+
+  //@}
+
+  //@{
+  //! @name Level handling
+
+  //! @brief Return level number.
+  int GetLevelID() const;
+
+  //! @brief Set level number.
+  void SetLevelID(int levelID);
+
+  //! Previous level
+  RCP<Level>& GetPreviousLevel() { return previousLevel_; }
+
+  //! Set previous level object
+  //! @\param[in] const RCP<Level>& previousLevel
+  void SetPreviousLevel(const RCP<Level>& previousLevel);
+  //@}
+
+  //! @name Set/Get factory manager
+  //@{
+  //! Set default factories (used internally by Hierarchy::SetLevel()).
+  // Users should not use this method.
+  void SetFactoryManager(const RCP<const FactoryManagerBase>& factoryManager);
+
+  //! returns the current factory manager
+  // Users should not use this method
+  const RCP<const FactoryManagerBase> GetFactoryManager();
+  //@}
+
+  //@{
+  //! @name Set functions
+
+  //! Store need label and its associated data. This does not increment the storage counter.
+  //! - If factory is not specified or factory == NoFactory::get(), mark data as user-defined by default
+  //! - If factory is == NULL, use defaultFactory (FactoryManager have to be available).
+  template <class T>
+  void Set(const std::string& ename, const T& entry, const FactoryBase* factory = NoFactory::get()) {
+    const FactoryBase* fac = GetFactory(ename, factory);
+
+    if (fac == NoFactory::get()) {
+      // Any data set with a NoFactory gets UserData keep flag by default
+      AddKeepFlag(ename, NoFactory::get(), MueLu::UserData);
     }
 
-    /*! @brief Get data without decrementing associated storage counter (i.e., read-only access).*/
-    template <class T>
-    void Get(const std::string& ename, T& rValue, const FactoryBase* factory = NoFactory::get()) {
-      rValue = Get<T>(ename, factory);
-    }
-
-    template <class T>
-    bool IsType(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
-      const FactoryBase* fac = GetFactory(ename, factory);
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" generated by factory \"" + fac->description() + "\" not found on level " + toString(GetLevelID()) + ".");
+    // Store entry only if data have been requested (or any keep flag)
+    if (IsRequested(ename, factory) || GetKeepFlag(ename, factory) != 0) {
+      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(factory, ename), Exceptions::RuntimeError, "" + ename + " not found in");
+      map_[factory][ename]->SetData(entry);
 
-      if (!IsAvailable(ename, fac)) {
-        TEUCHOS_TEST_FOR_EXCEPTION(NumRequests(fac, ename) < 1 && GetKeepFlag(ename, fac) == 0, Exceptions::RuntimeError,
-                                   "\"" << ename << "\" has not been requested (counter = " << NumRequests(fac, ename) << ", "
-                                   "KeepFlag = " << GetKeepFlag(ename, fac) << "). " << std::endl <<
-                                   "Generating factory:" << *fac << " NoFactory = " << NoFactory::get());
-        fac->CallBuild(*this);
-        Release(*fac);
-      }
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError,
-          "MueLu::Level::Get(): factory did not produce expected output on level " << GetLevelID()
-              << ". The data \"" << ename << "\" has not been generated by " << *fac);
-
-      return map_[fac][ename]->template CheckType<T>();
+    } else {
+      GetOStream(Runtime1) << "Level::Set: Not storing \"" << ename << "\" generated by factory " << factory->ShortClassName() << "[" << factory->GetID() << "]"
+                           << " on level " << toString(GetLevelID()) << ", as it has not been requested and no keep flags were set for it" << std::endl;
     }
-
-    /*! @brief GetTypeName returns type string of variable stored using ename and factory
-     *
-     *  @param[in] const std::string& ename
-     *  @param[in] const FactoryBase* factory
-     *  @return std::string with type information (e.g., "int")
-     * */
-    std::string GetTypeName(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
-      const FactoryBase* fac = GetFactory(ename, factory);
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" not found");
-
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError, "MueLu::Level::GetTypeString(): Data "
-                                 "\"" << ename << "\" generated by " << *fac << " is not available.");
-
-      return map_[fac][ename]->GetTypeName();
+  }  // Set
+
+  //@}
+
+  //! @name Get functions
+  //! @brief Get functions for accessing stored data
+
+  //@{
+  /*! @brief Get data without decrementing associated storage counter (i.e., read-only access).
+   *   Usage: Level->Get< RCP<Matrix> >("A", factory)
+   *   if factory == NULL => use default factory
+   *
+   *  @param[in] const std::string& ename
+   *  @param[in] const FactoryBase* factory
+   *  @return data (templated)
+   * */
+  template <class T>
+  T& Get(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
+    const FactoryBase* fac = GetFactory(ename, factory);
+    /*      printf("(l=%d)                                               getting    \"%20s\" generated by %10p  [actually, generated by %p (%43s)]\n",
+            levelID_, ename.c_str(), factory, fac, fac->description().c_str());*/
+
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" generated by factory \"" + fac->description() + "\" not found on level " + toString(GetLevelID()) + ".");
+
+    if (!IsAvailable(ename, fac)) {
+      TEUCHOS_TEST_FOR_EXCEPTION(NumRequests(fac, ename) < 1 && GetKeepFlag(ename, fac) == 0, Exceptions::RuntimeError,
+                                 "\"" << ename << "\" has not been requested (counter = " << NumRequests(fac, ename) << ", "
+                                                                                                                        "KeepFlag = "
+                                      << GetKeepFlag(ename, fac) << "). " << std::endl
+                                      << "Generating factory:" << *fac << " NoFactory = " << NoFactory::get());
+      fac->CallBuild(*this);
+      Release(*fac);
     }
 
-    //@}
-
-    //! @name Permanent storage
-    //@{
-
-    //! Request to keep variable 'ename' generated by 'factory' after the setup phase.
-    //  This method is intented to be used by user drivers for printing, debugging or to keep some computed data for a next run of the setup phase.
-    //
-    //  This method is an alias for: AddKeepFlag(ename, factory, MueLu::Keep)
-    //  See also the description of KeepEnum for more information.
-    //
-    //  To undo a keep request, one can use:
-    //  - Delete(ename, factory) to delete the data and remove the "Keep" flag
-    //  - or RemoveKeepFlag(ename, factory, MueLu::Keep) to go back to previous condition (data are kept only if internal MueLu logic need so).
-    //
-    // Note: Level variables tagged using this methods are also keep on the levels built using this.Build().
-    //       This means that is you request to keep a specific variable on a fine level, all the coarser level that are created automatically during the setup phase will also retain the same variable.
-    void Keep(const std::string & ename, const FactoryBase* factory) { AddKeepFlag(ename, factory, MueLu::Keep); } // Note: do not add default value for input parameter 'factory'
-
-    //! Delete data that have been retained after the setup phase (using Keep(), AddKeepFlag(), or internal MueLu logic).
-    // Special cases:
-    // - If entry (ename, factory) does not exist, nothing is done.
-    // - If entry exists but counter !=, entry cannot be desallocated before counter set to 0 (using Release()) so an exeption is thrown.
-    void Delete(const std::string& ename, const FactoryBase* factory) { // Note: do not add default value for input parameter 'factory'
-      if (!IsKey(factory, ename))
-        return;
-
-      // Precondition:
-      // Delete() should only be called if counter == 0
-      // Note: It better to throw an exception rather than deleting the data if counter != 0 because users are not supposed to manipulate data with counter != 0
-      TEUCHOS_TEST_FOR_EXCEPTION(IsRequested(ename, factory) == true, Exceptions::RuntimeError, "MueLu::Level::Delete(): IsRequested() == true. Ref counter != 0. You are not allowed to delete data that are still in use.");
-      // If counter == 0 and entry exists, this means that a keep flag is set. Or there is an internal logic problem.
-      TEUCHOS_TEST_FOR_EXCEPTION(GetKeepFlag(ename, factory) == 0, Exceptions::RuntimeError, "MueLu::Level::Delete(), Keep flag == 0?");
-
-      RemoveKeepFlag(ename, factory, MueLu::All); // will delete the data if counter == 0
-
-      // Post condition: data must have been deleted
-      TEUCHOS_TEST_FOR_EXCEPTION(IsAvailable(ename, factory) == true, Exceptions::RuntimeError, "MueLu::Level::Delete(): Internal error (Post condition). Data have not been deleted.");
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError,
+                               "MueLu::Level::Get(): factory did not produce expected output on level " << GetLevelID()
+                                                                                                        << ". The data \"" << ename << "\" has not been generated by " << *fac);
+
+    return map_[fac][ename]->template GetData<T>();
+  }
+
+  /*! @brief Get data without decrementing associated storage counter (i.e., read-only access).*/
+  template <class T>
+  void Get(const std::string& ename, T& rValue, const FactoryBase* factory = NoFactory::get()) {
+    rValue = Get<T>(ename, factory);
+  }
+
+  template <class T>
+  bool IsType(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
+    const FactoryBase* fac = GetFactory(ename, factory);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" generated by factory \"" + fac->description() + "\" not found on level " + toString(GetLevelID()) + ".");
+
+    if (!IsAvailable(ename, fac)) {
+      TEUCHOS_TEST_FOR_EXCEPTION(NumRequests(fac, ename) < 1 && GetKeepFlag(ename, fac) == 0, Exceptions::RuntimeError,
+                                 "\"" << ename << "\" has not been requested (counter = " << NumRequests(fac, ename) << ", "
+                                                                                                                        "KeepFlag = "
+                                      << GetKeepFlag(ename, fac) << "). " << std::endl
+                                      << "Generating factory:" << *fac << " NoFactory = " << NoFactory::get());
+      fac->CallBuild(*this);
+      Release(*fac);
     }
 
-    //! Delete all data that have been retained after the setup phase using Final flag
-    void Clear();
-
-    //! Delete all data from level that has no Keep flag set.
-    //! This is a function for experts only
-    void ExpertClear();
-
-    //! Test if a keep flag is set for variable 'ename' generated by 'factory'
-    //! The input parameter keep can be a combination of flags. IsKept() will then return true if at least one of the flag is set.
-    //! Note: There is no default parameter for IsKept() because it might be confusing (user generally wants to test IsKept with keep=MueLu::Keep but classes Level and Needs generally use keep = All)
-    bool IsKept(const std::string& ename, const FactoryBase* factory, KeepType keep) const { return GetKeepFlag(ename, factory) & keep; }
-
-    //! Add a keep flag for variable 'ename' generated by 'factory'
-    //! A variable can cumulate several keep flags (UserData+Final for example). This function just add a flag to the current flag combination.
-    //! By default, the flag 'Keep' is added. So Keep(ename, factory) is the same as AddKeepFlag(ename, factory).
-    //! See also the description of KeepEnum for more information.
-    void AddKeepFlag(const std::string & ename, const FactoryBase* factory = NoFactory::get(), KeepType keep = MueLu::Keep); // TODO: remove default value for input parameter 'factory'?
-
-    //! Remove a keep flag for variable 'ename' generated by 'factory'
-    //! A variable can cumulate several keep flags (UserData+Final for example). This function just add a flag to the current flag combination.
-    //! By default, all the flags are removed.
-    void RemoveKeepFlag(const std::string & ename, const FactoryBase* factory, KeepType keep = MueLu::All);
-
-    //! Get the flag combination set for variable 'ename' generated by 'factory'
-    KeepType GetKeepFlag(const std::string& ename, const FactoryBase* factory) const;
-
-    //@}
-
-    //! @name Request/Release functions
-    //! @brief Request and Release for incrementing/decrementing the reference count pointer for a specific variable.
-    //@{
-
-    //! Increment the storage counter for all the inputs of a factory
-    void Request(const FactoryBase& factory);
-
-    //! Decrement the storage counter for all the inputs of a factory
-    void Release(const FactoryBase& factory);
-
-    //! Callback from FactoryBase::CallDeclareInput() and FactoryBase::DeclareInput()
-    void DeclareInput(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy = NoFactory::get() );
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError,
+                               "MueLu::Level::Get(): factory did not produce expected output on level " << GetLevelID()
+                                                                                                        << ". The data \"" << ename << "\" has not been generated by " << *fac);
+
+    return map_[fac][ename]->template CheckType<T>();
+  }
+
+  /*! @brief GetTypeName returns type string of variable stored using ename and factory
+   *
+   *  @param[in] const std::string& ename
+   *  @param[in] const FactoryBase* factory
+   *  @return std::string with type information (e.g., "int")
+   * */
+  std::string GetTypeName(const std::string& ename, const FactoryBase* factory = NoFactory::get()) {
+    const FactoryBase* fac = GetFactory(ename, factory);
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(fac, ename), Exceptions::RuntimeError, "\"" + ename + "\" not found");
+
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsAvailable(ename, fac), Exceptions::RuntimeError,
+                               "MueLu::Level::GetTypeString(): Data "
+                               "\"" << ename
+                                    << "\" generated by " << *fac << " is not available.");
+
+    return map_[fac][ename]->GetTypeName();
+  }
+
+  //@}
+
+  //! @name Permanent storage
+  //@{
+
+  //! Request to keep variable 'ename' generated by 'factory' after the setup phase.
+  //  This method is intented to be used by user drivers for printing, debugging or to keep some computed data for a next run of the setup phase.
+  //
+  //  This method is an alias for: AddKeepFlag(ename, factory, MueLu::Keep)
+  //  See also the description of KeepEnum for more information.
+  //
+  //  To undo a keep request, one can use:
+  //  - Delete(ename, factory) to delete the data and remove the "Keep" flag
+  //  - or RemoveKeepFlag(ename, factory, MueLu::Keep) to go back to previous condition (data are kept only if internal MueLu logic need so).
+  //
+  // Note: Level variables tagged using this methods are also keep on the levels built using this.Build().
+  //       This means that is you request to keep a specific variable on a fine level, all the coarser level that are created automatically during the setup phase will also retain the same variable.
+  void Keep(const std::string& ename, const FactoryBase* factory) { AddKeepFlag(ename, factory, MueLu::Keep); }  // Note: do not add default value for input parameter 'factory'
+
+  //! Delete data that have been retained after the setup phase (using Keep(), AddKeepFlag(), or internal MueLu logic).
+  // Special cases:
+  // - If entry (ename, factory) does not exist, nothing is done.
+  // - If entry exists but counter !=, entry cannot be desallocated before counter set to 0 (using Release()) so an exeption is thrown.
+  void Delete(const std::string& ename, const FactoryBase* factory) {  // Note: do not add default value for input parameter 'factory'
+    if (!IsKey(factory, ename))
+      return;
+
+    // Precondition:
+    // Delete() should only be called if counter == 0
+    // Note: It better to throw an exception rather than deleting the data if counter != 0 because users are not supposed to manipulate data with counter != 0
+    TEUCHOS_TEST_FOR_EXCEPTION(IsRequested(ename, factory) == true, Exceptions::RuntimeError, "MueLu::Level::Delete(): IsRequested() == true. Ref counter != 0. You are not allowed to delete data that are still in use.");
+    // If counter == 0 and entry exists, this means that a keep flag is set. Or there is an internal logic problem.
+    TEUCHOS_TEST_FOR_EXCEPTION(GetKeepFlag(ename, factory) == 0, Exceptions::RuntimeError, "MueLu::Level::Delete(), Keep flag == 0?");
+
+    RemoveKeepFlag(ename, factory, MueLu::All);  // will delete the data if counter == 0
+
+    // Post condition: data must have been deleted
+    TEUCHOS_TEST_FOR_EXCEPTION(IsAvailable(ename, factory) == true, Exceptions::RuntimeError, "MueLu::Level::Delete(): Internal error (Post condition). Data have not been deleted.");
+  }
+
+  //! Delete all data that have been retained after the setup phase using Final flag
+  void Clear();
+
+  //! Delete all data from level that has no Keep flag set.
+  //! This is a function for experts only
+  void ExpertClear();
+
+  //! Test if a keep flag is set for variable 'ename' generated by 'factory'
+  //! The input parameter keep can be a combination of flags. IsKept() will then return true if at least one of the flag is set.
+  //! Note: There is no default parameter for IsKept() because it might be confusing (user generally wants to test IsKept with keep=MueLu::Keep but classes Level and Needs generally use keep = All)
+  bool IsKept(const std::string& ename, const FactoryBase* factory, KeepType keep) const { return GetKeepFlag(ename, factory) & keep; }
+
+  //! Add a keep flag for variable 'ename' generated by 'factory'
+  //! A variable can cumulate several keep flags (UserData+Final for example). This function just add a flag to the current flag combination.
+  //! By default, the flag 'Keep' is added. So Keep(ename, factory) is the same as AddKeepFlag(ename, factory).
+  //! See also the description of KeepEnum for more information.
+  void AddKeepFlag(const std::string& ename, const FactoryBase* factory = NoFactory::get(), KeepType keep = MueLu::Keep);  // TODO: remove default value for input parameter 'factory'?
+
+  //! Remove a keep flag for variable 'ename' generated by 'factory'
+  //! A variable can cumulate several keep flags (UserData+Final for example). This function just add a flag to the current flag combination.
+  //! By default, all the flags are removed.
+  void RemoveKeepFlag(const std::string& ename, const FactoryBase* factory, KeepType keep = MueLu::All);
+
+  //! Get the flag combination set for variable 'ename' generated by 'factory'
+  KeepType GetKeepFlag(const std::string& ename, const FactoryBase* factory) const;
+
+  //@}
+
+  //! @name Request/Release functions
+  //! @brief Request and Release for incrementing/decrementing the reference count pointer for a specific variable.
+  //@{
+
+  //! Increment the storage counter for all the inputs of a factory
+  void Request(const FactoryBase& factory);
+
+  //! Decrement the storage counter for all the inputs of a factory
+  void Release(const FactoryBase& factory);
+
+  //! Callback from FactoryBase::CallDeclareInput() and FactoryBase::DeclareInput()
+  void DeclareInput(const std::string& ename, const FactoryBase* factory, const FactoryBase* requestedBy = NoFactory::get());
+
+  //! Callback from FactoryBase::CallDeclareInput() and FactoryBase::DeclareInput() to declare factory dependencies
+  void DeclareDependencies(const FactoryBase* factory, bool bRequestOnly = false, bool bReleaseOnly = false);
+
+  //! Indicate that an object is needed. This increments the storage counter.
+  void Request(const std::string& ename, const FactoryBase* factory = NoFactory::get(), const FactoryBase* requestedBy = NoFactory::get());
+
+  //! Decrement the storage counter.
+  void Release(const std::string& ename, const FactoryBase* factory = NoFactory::get(), const FactoryBase* requestedBy = NoFactory::get());
+
+  //@}
 
-    //! Callback from FactoryBase::CallDeclareInput() and FactoryBase::DeclareInput() to declare factory dependencies
-    void DeclareDependencies(const FactoryBase* factory, bool bRequestOnly = false, bool bReleaseOnly = false);
-
-    //! Indicate that an object is needed. This increments the storage counter.
-    void Request(const std::string& ename, const FactoryBase* factory = NoFactory::get(), const FactoryBase* requestedBy = NoFactory::get());
-
-    //! Decrement the storage counter.
-    void Release(const std::string& ename, const FactoryBase* factory = NoFactory::get(), const FactoryBase* requestedBy = NoFactory::get());
-
-    //@}
-
-    //! @name Utility functions
-    //@{
-
-    //! Test whether a need's value has been saved.
-    bool IsAvailable(const std::string& ename, const FactoryBase* factory = NoFactory::get()) const {
-      if (!IsKey(factory, ename))
-        return false;
-      try {
-        return Get(factory, ename)->IsAvailable();
-      } catch (...) {
-        return false;
-      }
+  //! @name Utility functions
+  //@{
+
+  //! Test whether a need's value has been saved.
+  bool IsAvailable(const std::string& ename, const FactoryBase* factory = NoFactory::get()) const {
+    if (!IsKey(factory, ename))
+      return false;
+    try {
+      return Get(factory, ename)->IsAvailable();
+    } catch (...) {
+      return false;
     }
+  }
 
-    //! Test whether a need has been requested.  Note: this tells nothing about whether the need's value exists.
-    bool IsRequested(const std::string& ename, const FactoryBase* factory = NoFactory::get()) const {
-      if (!IsKey(factory, ename))
-        return false;
-      try {
-        return IsRequested(Get(factory, ename));
-      } catch (...) {
-        return false;
-      }
+  //! Test whether a need has been requested.  Note: this tells nothing about whether the need's value exists.
+  bool IsRequested(const std::string& ename, const FactoryBase* factory = NoFactory::get()) const {
+    if (!IsKey(factory, ename))
+      return false;
+    try {
+      return IsRequested(Get(factory, ename));
+    } catch (...) {
+      return false;
     }
+  }
 
+  //@}
+  //! @name I/O Functions
+  //@{
 
-    //@}
-    //! @name I/O Functions
-    //@{
-
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Printing method
-    // TODO: print only shows requested variables. check if we also list kept factories with ref counter=0?
-    void print(std::ostream& out, const VerbLevel verbLevel = Default) const;
+  //! Printing method
+  // TODO: print only shows requested variables. check if we also list kept factories with ref counter=0?
+  void print(std::ostream& out, const VerbLevel verbLevel = Default) const;
 
 #if defined(HAVE_MUELU_BOOST) && defined(HAVE_MUELU_BOOST_FOR_REAL) && defined(BOOST_VERSION) && (BOOST_VERSION >= 104400)
-    void UpdateGraph(std::map<const FactoryBase*, BoostVertex>&                   vindices,
-                     std::map<std::pair<BoostVertex, BoostVertex>, std::string>&  edges,
-                     BoostProperties&                                             dp,
-                     BoostGraph&                                                  graph) const;
+  void UpdateGraph(std::map<const FactoryBase*, BoostVertex>& vindices,
+                   std::map<std::pair<BoostVertex, BoostVertex>, std::string>& edges,
+                   BoostProperties& dp,
+                   BoostGraph& graph) const;
 #endif
 
-    //@}
-
-    enum   RequestMode { REQUEST, RELEASE, UNDEF };
-    RequestMode                     GetRequestMode() const                              { return requestMode_; }
-
-    void                            setlib(Xpetra::UnderlyingLib lib2)                  { lib_ = lib2; }
-    Xpetra::UnderlyingLib           lib()                                               { return lib_; }
-
-    void                            SetComm(RCP<const Teuchos::Comm<int> > const &comm) { comm_ = comm; }
-    RCP<const Teuchos::Comm<int> >  GetComm() const                                     { return comm_; }
-
-  private:
-
-    //! Copy constructor.
-    Level(const Level& source);
-
-    //! If input factory == NULL, returns the default factory. Else, return input factory.
-    //
-    //  If factory == NULL, the default factory is defined as follow:
-    // - If user data is available, it is considered as the default and the factory manager is ignored.
-    //   => The default factory is then NoFactory.
-    // - Else, the factory manager is used to get the default factory.
-    //
-    // This strategy allows to use the same factory manager on the fine and coarse level without any trouble.
-    //     Example :
-    //
-    //     FineLevel:
-    //     ----------
-    //     A          -> User provided
-    //     Nullspace  -> User provided
-    //
-    //     CoarseLevel:
-    //     ------------
-    //     A         -> RAPFactory
-    //     NullSpace -> NullspaceFactory
-    //
-    const FactoryBase* GetFactory(const std::string& varname, const FactoryBase* factory) const;
-
-    static RequestMode                  requestMode_;
-    Xpetra::UnderlyingLib               lib_;
-    RCP<const Teuchos::Comm<int> >      comm_;
-
-    typedef const FactoryBase*          Key1;
-    typedef const std::string           Key2;
-    typedef RCP<VariableContainer>      Value;
-    typedef Teuchos::map<Key2, Value>   SubMap;     //! Sub-map container (Key2 -> Value)
-    typedef Teuchos::map<Key1, SubMap>  TwoKeyMap;  //! Map of a map (Key1 -> SubMap)
-
-    int                                 levelID_;           // id number associated with level
-    RCP<const FactoryManagerBase>       factoryManager_;
-    RCP<Level>                          previousLevel_;     // linked list of Level
-    TwoKeyMap                           map_;
-
-    //! @name Utility functions
-    //@{
-
-    //! Test whether some information about (ename, factory) are stored
-    bool IsKey(const FactoryBase* factory, const std::string& ename) const {
-      TwoKeyMap::const_iterator it = map_.find(factory);
-      return (it != map_.end()) ? (it->second).count(ename) : false;
-    }
-
-    bool IsAvailableFactory(const FactoryBase* factory) const {
-      TwoKeyMap::const_iterator it = map_.find(factory);
-      if (it == map_.end())
-        return false;
-      for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++) {
-        if (sit->second->IsAvailable())
-            return true;
-      }
+  //@}
+
+  enum RequestMode { REQUEST,
+                     RELEASE,
+                     UNDEF };
+  RequestMode GetRequestMode() const { return requestMode_; }
+
+  void setlib(Xpetra::UnderlyingLib lib2) { lib_ = lib2; }
+  Xpetra::UnderlyingLib lib() { return lib_; }
+
+  void SetComm(RCP<const Teuchos::Comm<int> > const& comm) { comm_ = comm; }
+  RCP<const Teuchos::Comm<int> > GetComm() const { return comm_; }
+
+ private:
+  //! Copy constructor.
+  Level(const Level& source);
+
+  //! If input factory == NULL, returns the default factory. Else, return input factory.
+  //
+  //  If factory == NULL, the default factory is defined as follow:
+  // - If user data is available, it is considered as the default and the factory manager is ignored.
+  //   => The default factory is then NoFactory.
+  // - Else, the factory manager is used to get the default factory.
+  //
+  // This strategy allows to use the same factory manager on the fine and coarse level without any trouble.
+  //     Example :
+  //
+  //     FineLevel:
+  //     ----------
+  //     A          -> User provided
+  //     Nullspace  -> User provided
+  //
+  //     CoarseLevel:
+  //     ------------
+  //     A         -> RAPFactory
+  //     NullSpace -> NullspaceFactory
+  //
+  const FactoryBase* GetFactory(const std::string& varname, const FactoryBase* factory) const;
+
+  static RequestMode requestMode_;
+  Xpetra::UnderlyingLib lib_;
+  RCP<const Teuchos::Comm<int> > comm_;
+
+  typedef const FactoryBase* Key1;
+  typedef const std::string Key2;
+  typedef RCP<VariableContainer> Value;
+  typedef Teuchos::map<Key2, Value> SubMap;      //! Sub-map container (Key2 -> Value)
+  typedef Teuchos::map<Key1, SubMap> TwoKeyMap;  //! Map of a map (Key1 -> SubMap)
+
+  int levelID_;  // id number associated with level
+  RCP<const FactoryManagerBase> factoryManager_;
+  RCP<Level> previousLevel_;  // linked list of Level
+  TwoKeyMap map_;
+
+  //! @name Utility functions
+  //@{
+
+  //! Test whether some information about (ename, factory) are stored
+  bool IsKey(const FactoryBase* factory, const std::string& ename) const {
+    TwoKeyMap::const_iterator it = map_.find(factory);
+    return (it != map_.end()) ? (it->second).count(ename) : false;
+  }
+
+  bool IsAvailableFactory(const FactoryBase* factory) const {
+    TwoKeyMap::const_iterator it = map_.find(factory);
+    if (it == map_.end())
       return false;
+    for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++) {
+      if (sit->second->IsAvailable())
+        return true;
     }
+    return false;
+  }
 
-    bool IsRequested(const Value& v) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
-                                 "Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
-      return v->IsRequested();
-    }
+  bool IsRequested(const Value& v) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
+                               "Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
+    return v->IsRequested();
+  }
 
-    bool IsRequestedBy(const FactoryBase* factory, const std::string& ename, const FactoryBase* requestedBy) const {
-      if (!IsKey(factory, ename))
-        return false;
+  bool IsRequestedBy(const FactoryBase* factory, const std::string& ename, const FactoryBase* requestedBy) const {
+    if (!IsKey(factory, ename))
+      return false;
 
-      return IsRequestedBy(Get(factory, ename), requestedBy);
-    }
+    return IsRequestedBy(Get(factory, ename), requestedBy);
+  }
 
-    bool IsRequestedBy(const Value& v, const FactoryBase* requestedBy) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
-                                 "Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
-      return v->IsRequested(requestedBy);
-    }
+  bool IsRequestedBy(const Value& v, const FactoryBase* requestedBy) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
+                               "Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
+    return v->IsRequested(requestedBy);
+  }
 
-    bool IsRequestedFactory(const FactoryBase* factory) const {
-      TwoKeyMap::const_iterator it = map_.find(factory);
-      if (it == map_.end())
-        return false;
-      for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++)
-        if (IsRequested(sit->second))
-            return true;
+  bool IsRequestedFactory(const FactoryBase* factory) const {
+    TwoKeyMap::const_iterator it = map_.find(factory);
+    if (it == map_.end())
       return false;
-    }
+    for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++)
+      if (IsRequested(sit->second))
+        return true;
+    return false;
+  }
 
-    const Value& Get(const FactoryBase* factory, const std::string& ename) const {
-      TwoKeyMap::const_iterator it = map_.find(factory);
-      TEUCHOS_TEST_FOR_EXCEPTION(it == map_.end(), Exceptions::RuntimeError, "Key (" << factory << ", *) does not exist.");
+  const Value& Get(const FactoryBase* factory, const std::string& ename) const {
+    TwoKeyMap::const_iterator it = map_.find(factory);
+    TEUCHOS_TEST_FOR_EXCEPTION(it == map_.end(), Exceptions::RuntimeError, "Key (" << factory << ", *) does not exist.");
 
-      SubMap::const_iterator sit = it->second.find(ename);
-      TEUCHOS_TEST_FOR_EXCEPTION(sit == it->second.end(), Exceptions::RuntimeError, "Key (" << factory << ", " << ename << ") does not exist.");
+    SubMap::const_iterator sit = it->second.find(ename);
+    TEUCHOS_TEST_FOR_EXCEPTION(sit == it->second.end(), Exceptions::RuntimeError, "Key (" << factory << ", " << ename << ") does not exist.");
 
-      return sit->second;
-    }
+    return sit->second;
+  }
 
-    int NumRequests(const FactoryBase* factory, const std::string & ename) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(factory, ename), Exceptions::RuntimeError, "\"" + ename + "\" not found. Do a request first.");
-      const Teuchos::RCP<MueLu::VariableContainer>& v = Get(factory, ename);
-      TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
-                                 "NumRequests(): Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
-      return v->NumAllRequests();
-    }
+  int NumRequests(const FactoryBase* factory, const std::string& ename) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(!IsKey(factory, ename), Exceptions::RuntimeError, "\"" + ename + "\" not found. Do a request first.");
+    const Teuchos::RCP<MueLu::VariableContainer>& v = Get(factory, ename);
+    TEUCHOS_TEST_FOR_EXCEPTION(v->NumAllRequests() == 0 && v->GetKeepFlag() == 0, Exceptions::RuntimeError,
+                               "NumRequests(): Internal logic error: if counter == 0, the entry in countTable_ should have been deleted");
+    return v->NumAllRequests();
+  }
 
-    int CountRequestedFactory(const FactoryBase* factory) const {
-      TwoKeyMap::const_iterator it = map_.find(factory);
-      if (it == map_.end())
-        return 0;
+  int CountRequestedFactory(const FactoryBase* factory) const {
+    TwoKeyMap::const_iterator it = map_.find(factory);
+    if (it == map_.end())
+      return 0;
 
-      int cnt = 0;
-      for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++)
-        cnt += sit->second->NumAllRequests();
+    int cnt = 0;
+    for (SubMap::const_iterator sit = it->second.begin(); sit != it->second.end(); sit++)
+      cnt += sit->second->NumAllRequests();
 
-      return cnt;
-    }
+    return cnt;
+  }
 
-    //@}
+  //@}
 
-  }; //class Level
+};  // class Level
 
-} //namespace MueLu
+}  // namespace MueLu
 
-//TODO: Caps should not matter
+// TODO: Caps should not matter
 
-#endif // MUELU_LEVEL_HPP
+#endif  // MUELU_LEVEL_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_MasterList.cpp b/packages/muelu/src/MueCentral/MueLu_MasterList.cpp
index 1e0b9abd3df1..9f5aa9216ccf 100644
--- a/packages/muelu/src/MueCentral/MueLu_MasterList.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_MasterList.cpp
@@ -43,6 +43,8 @@
 // ***********************************************************************
 //
 // @HEADER
+
+// clang-format off
 #include <Teuchos_XMLParameterListCoreHelpers.hpp>
 
 #include "MueLu_Exceptions.hpp"
diff --git a/packages/muelu/src/MueCentral/MueLu_MasterList.hpp b/packages/muelu/src/MueCentral/MueLu_MasterList.hpp
index a558fc316d9c..ec70e8211d31 100644
--- a/packages/muelu/src/MueCentral/MueLu_MasterList.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_MasterList.hpp
@@ -55,103 +55,107 @@
 
 namespace MueLu {
 
-  /*!
-    @class MasterList class.
-    @brief Static class that holds the complete list of valid MueLu parameters.
-
-    This class creates a ParameterList that is used to validate user-supplied ParameterLists.  This same list
-    is the source of default values if a parameter does not appear in the user ParameterList.
-    This class also provides ParameterLists for certain common problem types, such as 2D and 3D isotropic Poisson.
-    @todo Add method to register user-defined problem type.  This will need both string identifier
-     and ParameterList (or string version of parameter list).
-
-  */
-
-  template <typename T, typename U> class DefaultProblemStrings;
-
-  class MasterList {
-  public:
-    //! @brief Return a "master" list of all valid parameters and their default values.
-    static Teuchos::RCP<const Teuchos::ParameterList> List();
-    //! @brief Return default parameter settings for the specified problem type.
-    static Teuchos::RCP<Teuchos::ParameterList>       GetProblemSpecificList(std::string const & problemType);
-
-    //! @brief Returns default value on the "master" list for a parameter with the specified name and type.
-    template<typename T>
-    static const T& getDefault(const std::string& name) {
-      return List()->get<T>(name);
-    }
-
-    //! @brief Returns default entry from the "master" list corresponding to the specified name.
-    static const Teuchos::ParameterEntry& getEntry(const std::string& name) {
-      return List()->getEntry(name);
-    }
-    
-    //! @brief Create xml string for given MueLu parameter (easy xml input format)
-    //!
-    //! @note: We should check whether template type T is the same as the expected parameter type in the parameter list
-    template<typename T>
-    static std::string generateXMLParameterString(const std::string& name, T data) {
-      Teuchos::ParameterEntry entry = getEntry(name);
-      std::stringstream ss;
-      // TODO: check whether getAny().typeName() matches the typeid(T)
-      //       One could use a technique as described here: http://stackoverflow.com/questions/4484982/how-to-convert-typename-t-to-string-in-c
-      ss << "<Parameter name=\"" << name << "\" type=\"" << entry.getAny().typeName() << "\" value=\"" << data << "\"/>";
-      return ss.str();
-    }
-
-    //! @brief Translate ML parameter to corresponding MueLu parameter
-    static std::string ML2MueLu(const std::string& name) {
-      std::map<std::string,std::string>::iterator it;
-      it = ML2MueLuLists_.find(name);
-      if(it!=ML2MueLuLists_.end()) return it->second;
-      else return "";
-    }
-
-    static std::string interpretParameterName(const std::string& name, const std::string& value);
-
-  private:
-    MasterList();
-    MasterList(const MasterList&);
-    MasterList& operator=(const MasterList&);
-
-    //! @brief A ParameterList that holds all valid parameters and their default values.
-    static Teuchos::RCP<Teuchos::ParameterList>  masterList_;
-    //! @brief String equivalent of the masterList_.
-    static const std::string                     stringList_;
-    //! @brief A ParameterList that holds all valid parameters and their default values for a particular problem type.
-    static Teuchos::RCP<Teuchos::ParameterList>  problemSpecificList_;
-    //! @brief The problem type associated with the current problem-specific ParameterList.
-    static std::string                           problemType_;
-    //! @brief Map of string equivalents of the problemSpecificList_.  The first entry is the problem type, the second is the string equivalent.
-    static std::map<std::string,std::string>     DefaultProblemTypeLists_;
-    //! @brief Map of ML parameter strings to corresponding MueLu parametes
-    static std::map<std::string,std::string>     ML2MueLuLists_;
-  };
-
-  /*!
-    @class DefaultProblemStrings class.
-    @brief Helper class to initialize DefaultProblemTypeLists_ in class MasterList.
-  */
-  template <typename T, typename U>
-  class DefaultProblemStrings {
-    public:
-      DefaultProblemStrings(const T& key, const U& val) {
-        map_[key] = val;
-      }
-
-      DefaultProblemStrings<T, U>& operator()(const T& key, const U& val) {
-        map_[key] = val;
-        return *this;
-      }
-
-      operator std::map<T, U>() const {
-        return map_;
-      }
-    private:
-      std::map<T, U> map_;
-  };
-
-} // namespace MueLu
+/*!
+  @class MasterList class.
+  @brief Static class that holds the complete list of valid MueLu parameters.
+
+  This class creates a ParameterList that is used to validate user-supplied ParameterLists.  This same list
+  is the source of default values if a parameter does not appear in the user ParameterList.
+  This class also provides ParameterLists for certain common problem types, such as 2D and 3D isotropic Poisson.
+  @todo Add method to register user-defined problem type.  This will need both string identifier
+   and ParameterList (or string version of parameter list).
+
+*/
+
+template <typename T, typename U>
+class DefaultProblemStrings;
+
+class MasterList {
+ public:
+  //! @brief Return a "master" list of all valid parameters and their default values.
+  static Teuchos::RCP<const Teuchos::ParameterList> List();
+  //! @brief Return default parameter settings for the specified problem type.
+  static Teuchos::RCP<Teuchos::ParameterList> GetProblemSpecificList(std::string const& problemType);
+
+  //! @brief Returns default value on the "master" list for a parameter with the specified name and type.
+  template <typename T>
+  static const T& getDefault(const std::string& name) {
+    return List()->get<T>(name);
+  }
+
+  //! @brief Returns default entry from the "master" list corresponding to the specified name.
+  static const Teuchos::ParameterEntry& getEntry(const std::string& name) {
+    return List()->getEntry(name);
+  }
+
+  //! @brief Create xml string for given MueLu parameter (easy xml input format)
+  //!
+  //! @note: We should check whether template type T is the same as the expected parameter type in the parameter list
+  template <typename T>
+  static std::string generateXMLParameterString(const std::string& name, T data) {
+    Teuchos::ParameterEntry entry = getEntry(name);
+    std::stringstream ss;
+    // TODO: check whether getAny().typeName() matches the typeid(T)
+    //       One could use a technique as described here: http://stackoverflow.com/questions/4484982/how-to-convert-typename-t-to-string-in-c
+    ss << "<Parameter name=\"" << name << "\" type=\"" << entry.getAny().typeName() << "\" value=\"" << data << "\"/>";
+    return ss.str();
+  }
+
+  //! @brief Translate ML parameter to corresponding MueLu parameter
+  static std::string ML2MueLu(const std::string& name) {
+    std::map<std::string, std::string>::iterator it;
+    it = ML2MueLuLists_.find(name);
+    if (it != ML2MueLuLists_.end())
+      return it->second;
+    else
+      return "";
+  }
+
+  static std::string interpretParameterName(const std::string& name, const std::string& value);
+
+ private:
+  MasterList();
+  MasterList(const MasterList&);
+  MasterList& operator=(const MasterList&);
+
+  //! @brief A ParameterList that holds all valid parameters and their default values.
+  static Teuchos::RCP<Teuchos::ParameterList> masterList_;
+  //! @brief String equivalent of the masterList_.
+  static const std::string stringList_;
+  //! @brief A ParameterList that holds all valid parameters and their default values for a particular problem type.
+  static Teuchos::RCP<Teuchos::ParameterList> problemSpecificList_;
+  //! @brief The problem type associated with the current problem-specific ParameterList.
+  static std::string problemType_;
+  //! @brief Map of string equivalents of the problemSpecificList_.  The first entry is the problem type, the second is the string equivalent.
+  static std::map<std::string, std::string> DefaultProblemTypeLists_;
+  //! @brief Map of ML parameter strings to corresponding MueLu parametes
+  static std::map<std::string, std::string> ML2MueLuLists_;
+};
+
+/*!
+  @class DefaultProblemStrings class.
+  @brief Helper class to initialize DefaultProblemTypeLists_ in class MasterList.
+*/
+template <typename T, typename U>
+class DefaultProblemStrings {
+ public:
+  DefaultProblemStrings(const T& key, const U& val) {
+    map_[key] = val;
+  }
+
+  DefaultProblemStrings<T, U>& operator()(const T& key, const U& val) {
+    map_[key] = val;
+    return *this;
+  }
+
+  operator std::map<T, U>() const {
+    return map_;
+  }
+
+ private:
+  std::map<T, U> map_;
+};
+
+}  // namespace MueLu
 
 #endif
diff --git a/packages/muelu/src/MueCentral/MueLu_NoFactory.cpp b/packages/muelu/src/MueCentral/MueLu_NoFactory.cpp
index 00ca8070d764..5e1ca11fc241 100644
--- a/packages/muelu/src/MueCentral/MueLu_NoFactory.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_NoFactory.cpp
@@ -56,18 +56,19 @@
 
 namespace MueLu {
 
-  const NoFactory* NoFactory::get() {
-    return getRCP().get();
-  }
+const NoFactory* NoFactory::get() {
+  return getRCP().get();
+}
 
-  RCP<const NoFactory> NoFactory::noFactory_ = Teuchos::null;
+RCP<const NoFactory> NoFactory::noFactory_ = Teuchos::null;
 
-  void NoFactory::CallBuild(Level& requestedLevel) const {
-    GetOStream(Errors) << "MueLu::NoFactory::Build(): this method cannot be called." << std::endl << std::endl;
-    GetOStream(Errors) << "Maybe, the required variable has been generated by another factory?" << std::endl;
-    GetOStream(Errors) << "Here is a list of all available data on the current level:" << std::endl;
-    requestedLevel.print(GetOStream(Errors),MueLu::Debug);
-    throw Exceptions::RuntimeError("MueLu::NoFactory::Build(): this method cannot be called.");
-  }
+void NoFactory::CallBuild(Level& requestedLevel) const {
+  GetOStream(Errors) << "MueLu::NoFactory::Build(): this method cannot be called." << std::endl
+                     << std::endl;
+  GetOStream(Errors) << "Maybe, the required variable has been generated by another factory?" << std::endl;
+  GetOStream(Errors) << "Here is a list of all available data on the current level:" << std::endl;
+  requestedLevel.print(GetOStream(Errors), MueLu::Debug);
+  throw Exceptions::RuntimeError("MueLu::NoFactory::Build(): this method cannot be called.");
+}
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_NoFactory.hpp b/packages/muelu/src/MueCentral/MueLu_NoFactory.hpp
index f4a8a6a93c24..560fdb6aca9c 100644
--- a/packages/muelu/src/MueCentral/MueLu_NoFactory.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_NoFactory.hpp
@@ -46,67 +46,65 @@
 #ifndef MUELU_NOFACTORY_HPP
 #define MUELU_NOFACTORY_HPP
 
-#include <algorithm>                    // for swap
-#include "Teuchos_RCPDecl.hpp"          // for RCP
-#include "Teuchos_RCP.hpp"              // for RCP::RCP<T>, RCP::operator=, etc
-#include "MueLu_config.hpp"             // for HAVE_MUELU_DEBUG
+#include <algorithm>            // for swap
+#include "Teuchos_RCPDecl.hpp"  // for RCP
+#include "Teuchos_RCP.hpp"      // for RCP::RCP<T>, RCP::operator=, etc
+#include "MueLu_config.hpp"     // for HAVE_MUELU_DEBUG
 #include "MueLu_Exceptions.hpp"
 #include "MueLu_FactoryBase.hpp"
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLu {
 
-  /*!
-    @class NoFactory class.
-    @brief NoFactory that is used for data stored in level class for that no generating factory is available/necessary.
+/*!
+  @class NoFactory class.
+  @brief NoFactory that is used for data stored in level class for that no generating factory is available/necessary.
 
-    This should be used as the "generating" factory for user-data.  Uses Singleton pattern.
-  */
-  class NoFactory : public FactoryBase {
+  This should be used as the "generating" factory for user-data.  Uses Singleton pattern.
+*/
+class NoFactory : public FactoryBase {
+  //! Constructor.
+  NoFactory() {}
 
-    //! Constructor.
-    NoFactory() { }
+ public:
+  //! Destructor.
+  virtual ~NoFactory() {}
 
-  public:
+  //! Implementation of FactoryBase interface
+  //@{
 
-    //! Destructor.
-    virtual ~NoFactory() { }
+  //!
+  void CallBuild(Level& requestedLevel) const;
 
-    //! Implementation of FactoryBase interface
-    //@{
+  //!
+  void CallDeclareInput(Level& /* requestedLevel */) const {}
 
-    //!
-    void CallBuild(Level& requestedLevel) const;
+  //@}
 
-    //!
-    void CallDeclareInput(Level& /* requestedLevel */) const { }
+  //! Static Get() functions
+  //@{
 
-    //@}
+  //!
+  static const RCP<const NoFactory> getRCP() {
+    if (noFactory_.is_null())
+      noFactory_ = rcp(new NoFactory());
 
-    //! Static Get() functions
-    //@{
+    return noFactory_;
+  }
 
-    //!
-    static const RCP<const NoFactory> getRCP() {
-      if (noFactory_.is_null())
-        noFactory_ = rcp(new NoFactory());
+  //!
+  static const NoFactory* get();
 
-      return noFactory_;
-    }
-
-    //!
-    static const NoFactory* get();
-
-    //@}
+  //@}
 #ifdef HAVE_MUELU_DEBUG
-    void ResetDebugData() const { }
+  void ResetDebugData() const {}
 #endif
 
-  private:
-    static RCP<const NoFactory> noFactory_; // static NoFactory instance for user defined "factories"
+ private:
+  static RCP<const NoFactory> noFactory_;  // static NoFactory instance for user defined "factories"
 
-  }; // class NoFactory
+};  // class NoFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_NOFACTORY_HPP
+#endif  // MUELU_NOFACTORY_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.cpp b/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.cpp
index a3b59ac47d4f..ae1875cd31af 100644
--- a/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.cpp
@@ -6,86 +6,86 @@
 
 namespace MueLu {
 
-  void printParameterListOptions(std::ostream &os, const Teuchos::ParameterList & p) {
-    p.print(os, Teuchos::ParameterList::PrintOptions().showDoc(true).indent(2).showTypes(true));
-    os << std::endl;
+void printParameterListOptions(std::ostream& os, const Teuchos::ParameterList& p) {
+  p.print(os, Teuchos::ParameterList::PrintOptions().showDoc(true).indent(2).showTypes(true));
+  os << std::endl;
+}
+
+ParameterListAcceptor::ParameterListAcceptor() {}
+
+ParameterListAcceptorImpl::ParameterListAcceptorImpl() {}
+
+void ParameterListAcceptorImpl::SetParameterList(const Teuchos::ParameterList& paramList) {
+  // This call is only for cosmetic reasons.
+  // If one calls SetParameterList before GetParameterList, that would mean
+  // that paramList_ has not been initialized yet. Therefore, the parameter
+  // would be put in it in the order user provided, and not in the order of
+  // valid parameter list. We'd like to have consistency in the order, so
+  // we do this extra call, which is no-op if paramList_ has already been
+  // initialized.
+  paramList_ = GetParameterList();
+
+  paramList_.setParameters(paramList);
+
+  // Validate and add defaults parameters.
+  Teuchos::RCP<const Teuchos::ParameterList> validParamList = GetValidParameterList();
+  if (validParamList != Teuchos::null) {
+    paramList_.validateParametersAndSetDefaults(*validParamList);
+  } else {
+    // Teuchos::null means that GetValidParameterList() not implemented.
+    // As such, we skip validation and have not way to set default values,
+    // which is potentially dangerous
   }
+}
 
-  ParameterListAcceptor::ParameterListAcceptor() { }
-
-  ParameterListAcceptorImpl::ParameterListAcceptorImpl() { }
-
-  void ParameterListAcceptorImpl::SetParameterList(const Teuchos::ParameterList& paramList) {
-    // This call is only for cosmetic reasons.
-    // If one calls SetParameterList before GetParameterList, that would mean
-    // that paramList_ has not been initialized yet. Therefore, the parameter
-    // would be put in it in the order user provided, and not in the order of
-    // valid parameter list. We'd like to have consistency in the order, so
-    // we do this extra call, which is no-op if paramList_ has already been
-    // initialized.
-    paramList_ = GetParameterList();
-
-    paramList_.setParameters(paramList);
-
-    // Validate and add defaults parameters.
+const Teuchos::ParameterList& ParameterListAcceptorImpl::GetParameterList() const {
+  // The returned list always has an entry for each valid parameter.
+  // Therefore, there is not need to test if a parameter is present before getting it.
+  if (paramList_.numParams() == 0) {
+    // Set paramList_ to the default list
     Teuchos::RCP<const Teuchos::ParameterList> validParamList = GetValidParameterList();
     if (validParamList != Teuchos::null) {
+      // Instead of simply doing
+      //   paramList_ = *validParamList;
+      // we use more complicated Teuchos calls, because we would like to
+      // have [default] values in the beginning
       paramList_.validateParametersAndSetDefaults(*validParamList);
-    } else {
-      // Teuchos::null means that GetValidParameterList() not implemented.
-      // As such, we skip validation and have not way to set default values,
-      // which is potentially dangerous
-    }
-  }
-
-  const Teuchos::ParameterList& ParameterListAcceptorImpl::GetParameterList() const {
-    // The returned list always has an entry for each valid parameter.
-    // Therefore, there is not need to test if a parameter is present before getting it.
-    if (paramList_.numParams() == 0) {
-      // Set paramList_ to the default list
-      Teuchos::RCP<const Teuchos::ParameterList> validParamList = GetValidParameterList();
-      if (validParamList != Teuchos::null) {
-        // Instead of simply doing
-        //   paramList_ = *validParamList;
-        // we use more complicated Teuchos calls, because we would like to
-        // have [default] values in the beginning
-        paramList_.validateParametersAndSetDefaults(*validParamList);
-      }
-
-    } else {
-      // We are sure that the list has all the valid parameters defined
-      // because the parameter list validation process adds the default
-      // values to the user list
     }
 
-    return paramList_;
+  } else {
+    // We are sure that the list has all the valid parameters defined
+    // because the parameter list validation process adds the default
+    // values to the user list
   }
 
-  void ParameterListAcceptorImpl::SetParameter(const std::string& name, const ParameterEntry& entry) {
-    Teuchos::ParameterList paramList;
-    paramList.setEntry(name, entry);
-    SetParameterList(paramList); // This forces revalidation of the list
-  }
+  return paramList_;
+}
 
-  const ParameterEntry & ParameterListAcceptorImpl::GetParameter(const std::string &name) const {
-    return GetParameterList().getEntry(name);
-  }
+void ParameterListAcceptorImpl::SetParameter(const std::string& name, const ParameterEntry& entry) {
+  Teuchos::ParameterList paramList;
+  paramList.setEntry(name, entry);
+  SetParameterList(paramList);  // This forces revalidation of the list
+}
 
-  void ParameterListAcceptorImpl::GetDocumentation(std::ostream &os) const {
-    // default implementation
+const ParameterEntry& ParameterListAcceptorImpl::GetParameter(const std::string& name) const {
+  return GetParameterList().getEntry(name);
+}
 
-    Teuchos::RCP<const Teuchos::ParameterList> validParamList = GetValidParameterList();
-    if (validParamList == Teuchos::null) {
-      os << "## Documentation not available:" << std::endl;
-      return;
-    }
+void ParameterListAcceptorImpl::GetDocumentation(std::ostream& os) const {
+  // default implementation
 
-    os << "## Parameters:" << std::endl;
-    printParameterListOptions(os, *validParamList);
-
-    os << "## Fully described default method:" << std::endl;
-    validParamList->print(os, 2, true, false);
-    os << std::endl;
+  Teuchos::RCP<const Teuchos::ParameterList> validParamList = GetValidParameterList();
+  if (validParamList == Teuchos::null) {
+    os << "## Documentation not available:" << std::endl;
+    return;
   }
 
-} //namespace MueLu
+  os << "## Parameters:" << std::endl;
+  printParameterListOptions(os, *validParamList);
+
+  os << "## Fully described default method:" << std::endl;
+  validParamList->print(os, 2, true, false);
+  os << std::endl;
+}
+
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.hpp b/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.hpp
index 61a7b19de051..04f228846e17 100644
--- a/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_ParameterListAcceptor.hpp
@@ -12,117 +12,110 @@
 
 namespace MueLu {
 
-  using Teuchos::ParameterEntry;
+using Teuchos::ParameterEntry;
 
-  // Printing the valid parameter list only showing documentation fields
-  void printParameterListOptions(std::ostream &os, const Teuchos::ParameterList & p);
+// Printing the valid parameter list only showing documentation fields
+void printParameterListOptions(std::ostream &os, const Teuchos::ParameterList &p);
 
-  /*! @class ParameterListAcceptor
-      @brief Abstract interface of a class accepting parameter lists.
+/*! @class ParameterListAcceptor
+    @brief Abstract interface of a class accepting parameter lists.
 
-  */
-  class ParameterListAcceptor {
-
-  public:
-
-    //! @name Constructors/Destructors
-    //! @{
-    ParameterListAcceptor();
-
-    virtual ~ParameterListAcceptor() = default;
-
-    //! @}
-    //! @name Get/Set methods.
-    //! @{
-
-    /*! @brief Return a const parameter list of valid parameters that setParameterList() will accept.
-
-       Also define the default values of parameters according to the input parameter list.
-    */
-    // Questions:
-    // - Do we want this function to be static?
-    // => yes, but we also want it virtual. So static is not possible. But this method does not access any instance specific data.
-    // - Do we want to pass an optional param list as input parameter?
-    //   yes => if some parameters depends of other parameters, this would allow to return different default / valid parameters according to already set parameters.
-    //   Ex: fact: ILUT => extra param threshold; fact: ILUK => extra param level-of-fill
-    //
-    // If a parameter is unused AND default, it is printed as [default] by std::cout << paramList but printed as unused by paramList.unused(std::cout).
-    // So if we set default parameters in getValidParameters() that are not used, user will get a warning message. We don't want that for [default].
-    // One solution is to never set any unused parameter in getValidParameters().
-    // If some parameters are available only conditionnaly, do not set them by default in setValidParameters when the conditions are not met.
-    //
-    virtual Teuchos::RCP<const Teuchos::ParameterList> GetValidParameterList() const = 0;
-
-    //! @brief Set parameters from a parameter list and return with default values.
-    //
-    // Questions:
-    //  - Input: const ParameterList or not const? IE: do we want to modify user paramlist directly?
-    //    => not const: user can make a copy outside if he do not want us to modify his initial paramlist.
-    //  - Input: RCP or just reference? RCP avoid the copy of the parameter list but
-    //           I'm afraid that a user modify the list outside of the class after doing a SetParameterList
-    //           (this by-pass the validation and can create some confusion).
-    //           In another hand, if we use a copy of the list, we do not set the "[used]"/["unused"] flag
-    //           on the original list during Build(). GetParameterList has to be called to retrieve the list with the correct flags.
-    //
-    // What we really want is for the user to have a const version outside and MueLu having a non-const version inside.
-    // Is there a C++ pattern to do that?
-    virtual void SetParameterList(const Teuchos::ParameterList& paramList) = 0;
-
-    // Get the parameter list.
-    // Teuchos::ParameterListAcceptor also provides a non-const version of this but I don't see why.
-    // We don't want a user to mess with the internal parameter list.
-    // To change a parameter, one can make a copy and call SetParameterList again.
-    // No need for RCP, it's just a view
-    virtual const Teuchos::ParameterList & GetParameterList() const = 0;
-
-    //! @brief Set a parameter directly as a ParameterEntry.
-    virtual void SetParameter(const std::string &name, const ParameterEntry &entry) = 0;
-
-    //! @brief Retrieves a const entry with the name name.
-    virtual const ParameterEntry & GetParameter(const std::string &name) const = 0;
-
-    virtual void GetDocumentation(std::ostream &os) const = 0;
-
-    //! @}
-
-  }; //class ParameterListAcceptor
+*/
+class ParameterListAcceptor {
+ public:
+  //! @name Constructors/Destructors
+  //! @{
+  ParameterListAcceptor();
 
-  // Partial implementation of ParameterListAcceptor that stores the object parameters in an internal parameterList
-  class ParameterListAcceptorImpl: public ParameterListAcceptor {
+  virtual ~ParameterListAcceptor() = default;
 
-  public:
+  //! @}
+  //! @name Get/Set methods.
+  //! @{
 
-    ParameterListAcceptorImpl();
+  /*! @brief Return a const parameter list of valid parameters that setParameterList() will accept.
 
-    virtual ~ParameterListAcceptorImpl() = default;
-
-    virtual void SetParameterList(const Teuchos::ParameterList& paramList);
-
-    // The returned list always has an entry for each valid parameter.
-    // Therefore, there is not need to test if a parameter is present before getting it.
-    virtual const Teuchos::ParameterList& GetParameterList() const;
-
-    void SetParameter(const std::string& name, const ParameterEntry& entry);
-
-    const ParameterEntry & GetParameter(const std::string &name) const;
-
-    virtual void GetDocumentation(std::ostream &os) const;
-
-  private:
-
-    mutable // mutable can be avoid by changing return type of GetParameterList() to RCP but conceptually, I like the fact that GetParameterList returns ref to param list.
-    Teuchos::ParameterList paramList_; // Private: Use GetParameterList() to access this list
-                                       // This list might be empty before calling GetParameterList()
-                                       // but it is populate with automatically if needed in SetParameterList/GetParameterList/...
-                                       // Therefore, there is no need to test if a parameter exist before accessing it with pL.get("paramName")
-                                       // in sub classes.
-
-    // Note: Teuchos::ParameterListAcceptor has a getMyParamList() function to access paramList_ without going through the logic of getParameterList()
-    // Do we need that too?
-
-  };
-
-
-}
-
-#endif // MUELU_PARAMETERLISTACCEPTOR_HPP
+     Also define the default values of parameters according to the input parameter list.
+  */
+  // Questions:
+  // - Do we want this function to be static?
+  // => yes, but we also want it virtual. So static is not possible. But this method does not access any instance specific data.
+  // - Do we want to pass an optional param list as input parameter?
+  //   yes => if some parameters depends of other parameters, this would allow to return different default / valid parameters according to already set parameters.
+  //   Ex: fact: ILUT => extra param threshold; fact: ILUK => extra param level-of-fill
+  //
+  // If a parameter is unused AND default, it is printed as [default] by std::cout << paramList but printed as unused by paramList.unused(std::cout).
+  // So if we set default parameters in getValidParameters() that are not used, user will get a warning message. We don't want that for [default].
+  // One solution is to never set any unused parameter in getValidParameters().
+  // If some parameters are available only conditionnaly, do not set them by default in setValidParameters when the conditions are not met.
+  //
+  virtual Teuchos::RCP<const Teuchos::ParameterList> GetValidParameterList() const = 0;
+
+  //! @brief Set parameters from a parameter list and return with default values.
+  //
+  // Questions:
+  //  - Input: const ParameterList or not const? IE: do we want to modify user paramlist directly?
+  //    => not const: user can make a copy outside if he do not want us to modify his initial paramlist.
+  //  - Input: RCP or just reference? RCP avoid the copy of the parameter list but
+  //           I'm afraid that a user modify the list outside of the class after doing a SetParameterList
+  //           (this by-pass the validation and can create some confusion).
+  //           In another hand, if we use a copy of the list, we do not set the "[used]"/["unused"] flag
+  //           on the original list during Build(). GetParameterList has to be called to retrieve the list with the correct flags.
+  //
+  // What we really want is for the user to have a const version outside and MueLu having a non-const version inside.
+  // Is there a C++ pattern to do that?
+  virtual void SetParameterList(const Teuchos::ParameterList &paramList) = 0;
+
+  // Get the parameter list.
+  // Teuchos::ParameterListAcceptor also provides a non-const version of this but I don't see why.
+  // We don't want a user to mess with the internal parameter list.
+  // To change a parameter, one can make a copy and call SetParameterList again.
+  // No need for RCP, it's just a view
+  virtual const Teuchos::ParameterList &GetParameterList() const = 0;
+
+  //! @brief Set a parameter directly as a ParameterEntry.
+  virtual void SetParameter(const std::string &name, const ParameterEntry &entry) = 0;
+
+  //! @brief Retrieves a const entry with the name name.
+  virtual const ParameterEntry &GetParameter(const std::string &name) const = 0;
+
+  virtual void GetDocumentation(std::ostream &os) const = 0;
+
+  //! @}
+
+};  // class ParameterListAcceptor
+
+// Partial implementation of ParameterListAcceptor that stores the object parameters in an internal parameterList
+class ParameterListAcceptorImpl : public ParameterListAcceptor {
+ public:
+  ParameterListAcceptorImpl();
+
+  virtual ~ParameterListAcceptorImpl() = default;
+
+  virtual void SetParameterList(const Teuchos::ParameterList &paramList);
+
+  // The returned list always has an entry for each valid parameter.
+  // Therefore, there is not need to test if a parameter is present before getting it.
+  virtual const Teuchos::ParameterList &GetParameterList() const;
+
+  void SetParameter(const std::string &name, const ParameterEntry &entry);
+
+  const ParameterEntry &GetParameter(const std::string &name) const;
+
+  virtual void GetDocumentation(std::ostream &os) const;
+
+ private:
+  mutable                                 // mutable can be avoid by changing return type of GetParameterList() to RCP but conceptually, I like the fact that GetParameterList returns ref to param list.
+      Teuchos::ParameterList paramList_;  // Private: Use GetParameterList() to access this list
+                                          // This list might be empty before calling GetParameterList()
+                                          // but it is populate with automatically if needed in SetParameterList/GetParameterList/...
+                                          // Therefore, there is no need to test if a parameter exist before accessing it with pL.get("paramName")
+                                          // in sub classes.
+
+  // Note: Teuchos::ParameterListAcceptor has a getMyParamList() function to access paramList_ without going through the logic of getParameterList()
+  // Do we need that too?
+};
+
+}  // namespace MueLu
+
+#endif  // MUELU_PARAMETERLISTACCEPTOR_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_SingleLevelFactoryBase.hpp b/packages/muelu/src/MueCentral/MueLu_SingleLevelFactoryBase.hpp
index 93a6a08688e7..60dd6f4160ec 100644
--- a/packages/muelu/src/MueCentral/MueLu_SingleLevelFactoryBase.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_SingleLevelFactoryBase.hpp
@@ -55,108 +55,107 @@
 
 namespace MueLu {
 
-  /*!
-    @class SingleLevelFactoryBase
-    @brief Base class for factories that use one level (currentLevel).
+/*!
+  @class SingleLevelFactoryBase
+  @brief Base class for factories that use one level (currentLevel).
 
-    @ingroup MueLuBaseClasses
-  */
-  class SingleLevelFactoryBase : public Factory {
-
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+  @ingroup MueLuBaseClasses
+*/
+class SingleLevelFactoryBase : public Factory {
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    SingleLevelFactoryBase() { }
+  //! Constructor.
+  SingleLevelFactoryBase() {}
 
-    //! Destructor.
-    virtual ~SingleLevelFactoryBase() { }
+  //! Destructor.
+  virtual ~SingleLevelFactoryBase() {}
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    virtual void DeclareInput(Level &currentLevel) const = 0;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  virtual void DeclareInput(Level& currentLevel) const = 0;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    virtual void Build(Level & currentLevel) const = 0;
+  //! Build an object with this factory.
+  virtual void Build(Level& currentLevel) const = 0;
 
-    //!
-    virtual void CallBuild(Level& requestedLevel) const {
+  //!
+  virtual void CallBuild(Level& requestedLevel) const {
 #ifdef HAVE_MUELU_DEBUG
-      // We cannot call Build method twice for the same level, but we can call it multiple times for different levels
-      int levelID = requestedLevel.GetLevelID();
-      TEUCHOS_TEST_FOR_EXCEPTION((multipleCallCheck_ == ENABLED) && (multipleCallCheckGlobal_ == ENABLED) && (lastLevelID_ == levelID),
-                                 Exceptions::RuntimeError,
-                                 this->ShortClassName() << "::Build() called twice for the same level (levelID=" << levelID
-                                 << "). This is likely due to a configuration error, or calling hierarchy setup multiple times "
-                                 << "without resetting debug info through FactoryManager::ResetDebugData().");
-      if (multipleCallCheck_ == FIRSTCALL)
-        multipleCallCheck_ = ENABLED;
-
-      lastLevelID_ = levelID;
+    // We cannot call Build method twice for the same level, but we can call it multiple times for different levels
+    int levelID = requestedLevel.GetLevelID();
+    TEUCHOS_TEST_FOR_EXCEPTION((multipleCallCheck_ == ENABLED) && (multipleCallCheckGlobal_ == ENABLED) && (lastLevelID_ == levelID),
+                               Exceptions::RuntimeError,
+                               this->ShortClassName() << "::Build() called twice for the same level (levelID=" << levelID
+                                                      << "). This is likely due to a configuration error, or calling hierarchy setup multiple times "
+                                                      << "without resetting debug info through FactoryManager::ResetDebugData().");
+    if (multipleCallCheck_ == FIRSTCALL)
+      multipleCallCheck_ = ENABLED;
+
+    lastLevelID_ = levelID;
 #endif
 
-      RCP<const Teuchos::Comm<int> > comm = requestedLevel.GetComm();
-      if (comm.is_null()) {
-        // Some factories are called before we constructed Ac, and therefore,
-        // before we set the level communicator. For such factories we can get
-        // the comm from the previous level, as all processes go there
-        RCP<Level>& prevLevel = requestedLevel.GetPreviousLevel();
-        if (!prevLevel.is_null())
-          comm = prevLevel->GetComm();
-      }
-
-      int oldRank = -1;
-      if (!comm.is_null())
-        oldRank = SetProcRankVerbose(comm->getRank());
-
-      // Synchronization timer
-      std::string syncTimer = this->ShortClassName() + ": Build sync (level=" + toString(requestedLevel.GetLevelID()) + ")";
-      if (this->timerSync_ && !comm.is_null()) {
-        TimeMonitor timer(*this, syncTimer);
-        comm->barrier();
-      }
-
-      Build(requestedLevel);
-
-      if (this->timerSync_ && !comm.is_null()) {
-        TimeMonitor timer(*this, syncTimer);
-        comm->barrier();
-      }
-
-      if (IsPrint(Test1))
-        GetOStream(Test1) << *RemoveFactoriesFromList(GetParameterList()) << std::endl;
-      else
-        RemoveFactoriesFromList(GetParameterList())->print(GetOStream(Test0), Teuchos::ParameterList::PrintOptions().indent(0).showFlags(true).showDefault(false));
-
-      if (oldRank != -1)
-        SetProcRankVerbose(oldRank);
+    RCP<const Teuchos::Comm<int> > comm = requestedLevel.GetComm();
+    if (comm.is_null()) {
+      // Some factories are called before we constructed Ac, and therefore,
+      // before we set the level communicator. For such factories we can get
+      // the comm from the previous level, as all processes go there
+      RCP<Level>& prevLevel = requestedLevel.GetPreviousLevel();
+      if (!prevLevel.is_null())
+        comm = prevLevel->GetComm();
+    }
+
+    int oldRank = -1;
+    if (!comm.is_null())
+      oldRank = SetProcRankVerbose(comm->getRank());
+
+    // Synchronization timer
+    std::string syncTimer = this->ShortClassName() + ": Build sync (level=" + toString(requestedLevel.GetLevelID()) + ")";
+    if (this->timerSync_ && !comm.is_null()) {
+      TimeMonitor timer(*this, syncTimer);
+      comm->barrier();
     }
 
-    //!
-    virtual void CallDeclareInput(Level & requestedLevel) const {
-      DeclareInput(requestedLevel);
+    Build(requestedLevel);
+
+    if (this->timerSync_ && !comm.is_null()) {
+      TimeMonitor timer(*this, syncTimer);
+      comm->barrier();
     }
-    //@}
 
-  }; //class SingleLevelFactoryBase
+    if (IsPrint(Test1))
+      GetOStream(Test1) << *RemoveFactoriesFromList(GetParameterList()) << std::endl;
+    else
+      RemoveFactoriesFromList(GetParameterList())->print(GetOStream(Test0), Teuchos::ParameterList::PrintOptions().indent(0).showFlags(true).showDefault(false));
+
+    if (oldRank != -1)
+      SetProcRankVerbose(oldRank);
+  }
+
+  //!
+  virtual void CallDeclareInput(Level& requestedLevel) const {
+    DeclareInput(requestedLevel);
+  }
+  //@}
+
+};  // class SingleLevelFactoryBase
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SINGLELEVELFACTORY_SHORT
-#endif //ifndef MUELU_SINGLELEVELFACTORY_HPP
+#endif  // ifndef MUELU_SINGLELEVELFACTORY_HPP
 
-//TODO: code factorization between SingleLevelFactoryBase and TwoLevelFactoryBase
+// TODO: code factorization between SingleLevelFactoryBase and TwoLevelFactoryBase
diff --git a/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_decl.hpp b/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_decl.hpp
index 6fc6c7e3a66a..c1415c2c0010 100644
--- a/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_decl.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_decl.hpp
@@ -19,32 +19,31 @@
 
 namespace MueLu {
 
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class TopRAPFactory : public TwoLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class TopRAPFactory : public TwoLevelFactoryBase {
 #undef MUELU_TOPRAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManager);
+  TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManagerFine, RCP<const FactoryManagerBase> parentFactoryManagerCoarse);
 
-    TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManager);
-    TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManagerFine, RCP<const FactoryManagerBase> parentFactoryManagerCoarse);
+  virtual ~TopRAPFactory();
 
-    virtual ~TopRAPFactory();
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level & fineLevel, Level & coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build(Level & fineLevel, Level & coarseLevel) const;
+ private:
+  RCP<const FactoryBase> PFact_;
+  RCP<const FactoryBase> RFact_;
+  RCP<const FactoryBase> AcFact_;
+};
 
-  private:
-    RCP<const FactoryBase> PFact_;
-    RCP<const FactoryBase> RFact_;
-    RCP<const FactoryBase> AcFact_;
-  };
-
-}
+}  // namespace MueLu
 
 #define MUELU_TOPRAPFACTORY_SHORT
 #endif /* PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPRAPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_def.hpp b/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_def.hpp
index ac0f1982469b..62f2113238da 100644
--- a/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_def.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_TopRAPFactory_def.hpp
@@ -8,74 +8,74 @@
 #ifndef PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPRAPFACTORY_DEF_HPP_
 #define PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPRAPFACTORY_DEF_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 
 #include "MueLu_FactoryManagerBase.hpp"
 //#include "MueLu_HierarchyUtils_fwd.hpp"
 #include "MueLu_TopRAPFactory_decl.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManager) :
-    PFact_ (parentFactoryManager->GetFactory("P")),
-    RFact_ (parentFactoryManager->GetFactory("R")),
-    AcFact_(parentFactoryManager->GetFactory("A"))
-  { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopRAPFactory(RCP<const FactoryManagerBase> /* parentFactoryManagerFine */, RCP<const FactoryManagerBase> parentFactoryManagerCoarse) :
-    PFact_ (parentFactoryManagerCoarse->GetFactory("P")),
-    RFact_ (parentFactoryManagerCoarse->GetFactory("R")),
-    AcFact_(parentFactoryManagerCoarse->GetFactory("A"))
-  { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~TopRAPFactory() { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopRAPFactory(RCP<const FactoryManagerBase> parentFactoryManager)
+  : PFact_(parentFactoryManager->GetFactory("P"))
+  , RFact_(parentFactoryManager->GetFactory("R"))
+  , AcFact_(parentFactoryManager->GetFactory("A")) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopRAPFactory(RCP<const FactoryManagerBase> /* parentFactoryManagerFine */, RCP<const FactoryManagerBase> parentFactoryManagerCoarse)
+  : PFact_(parentFactoryManagerCoarse->GetFactory("P"))
+  , RFact_(parentFactoryManagerCoarse->GetFactory("R"))
+  , AcFact_(parentFactoryManagerCoarse->GetFactory("A")) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~TopRAPFactory() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  if (PFact_ != Teuchos::null) coarseLevel.DeclareInput("P", PFact_.get());
+  if (RFact_ != Teuchos::null) coarseLevel.DeclareInput("R", RFact_.get());
+  if ((AcFact_ != Teuchos::null) && (AcFact_ != NoFactory::getRCP())) coarseLevel.DeclareInput("A", AcFact_.get());
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & /* fineLevel */, Level & coarseLevel) const {
-    if (PFact_  != Teuchos::null)                                       coarseLevel.DeclareInput("P", PFact_.get());
-    if (RFact_  != Teuchos::null)                                       coarseLevel.DeclareInput("R", RFact_.get());
-    if ((AcFact_ != Teuchos::null) && (AcFact_ != NoFactory::getRCP())) coarseLevel.DeclareInput("A", AcFact_.get());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* fineLevel */, Level& coarseLevel) const {
+  if ((PFact_ != Teuchos::null) && (PFact_ != NoFactory::getRCP())) {
+    RCP<Operator> oP = coarseLevel.Get<RCP<Operator> >("P", PFact_.get());
+    // Don't have a valid operator (e.g., # global aggregates is 0) so we just bail out
+    // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()j
+    if (oP == Teuchos::null) return;
+    RCP<Matrix> P = rcp_dynamic_cast<Matrix>(oP);
+    if (!P.is_null())
+      coarseLevel.Set("P", P, NoFactory::get());
+    else
+      coarseLevel.Set("P", oP, NoFactory::get());
+    coarseLevel.AddKeepFlag("P", NoFactory::get(), MueLu::Final);        // FIXME2: Order of Remove/Add matter (data removed otherwise). Should do something about this
+    coarseLevel.RemoveKeepFlag("P", NoFactory::get(), MueLu::UserData);  // FIXME: This is a hack, I should change behavior of Level::Set() instead. FIXME3: Should not be removed if flag was there already
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TopRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level & coarseLevel) const {
-    if ((PFact_ != Teuchos::null) && (PFact_ != NoFactory::getRCP())) {
-      RCP<Operator> oP = coarseLevel.Get<RCP<Operator> >("P", PFact_.get());
-      // Don't have a valid operator (e.g., # global aggregates is 0) so we just bail out
-      // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()j
-      if (oP == Teuchos::null) return;
-      RCP<Matrix>    P = rcp_dynamic_cast<Matrix>(oP);
-      if (!P.is_null()) coarseLevel.Set("P",  P, NoFactory::get());
-      else              coarseLevel.Set("P", oP, NoFactory::get());
-      coarseLevel.AddKeepFlag   ("P", NoFactory::get(), MueLu::Final);    // FIXME2: Order of Remove/Add matter (data removed otherwise). Should do something about this
-      coarseLevel.RemoveKeepFlag("P", NoFactory::get(), MueLu::UserData); // FIXME: This is a hack, I should change behavior of Level::Set() instead. FIXME3: Should not be removed if flag was there already
-
-    }
-
-    if ((RFact_ != Teuchos::null) && (RFact_ != NoFactory::getRCP()) ) {
-      RCP<Operator> oR = coarseLevel.Get<RCP<Operator> >("R", RFact_.get());
-      RCP<Matrix>    R = rcp_dynamic_cast<Matrix>(oR);
-      if (!R.is_null()) coarseLevel.Set("R",  R, NoFactory::get());
-      else              coarseLevel.Set("R", oR, NoFactory::get());
-      coarseLevel.AddKeepFlag   ("R", NoFactory::get(), MueLu::Final);
-      coarseLevel.RemoveKeepFlag("R", NoFactory::get(), MueLu::UserData); // FIXME: This is a hack
-    }
+  if ((RFact_ != Teuchos::null) && (RFact_ != NoFactory::getRCP())) {
+    RCP<Operator> oR = coarseLevel.Get<RCP<Operator> >("R", RFact_.get());
+    RCP<Matrix> R    = rcp_dynamic_cast<Matrix>(oR);
+    if (!R.is_null())
+      coarseLevel.Set("R", R, NoFactory::get());
+    else
+      coarseLevel.Set("R", oR, NoFactory::get());
+    coarseLevel.AddKeepFlag("R", NoFactory::get(), MueLu::Final);
+    coarseLevel.RemoveKeepFlag("R", NoFactory::get(), MueLu::UserData);  // FIXME: This is a hack
+  }
 
-    if ((AcFact_ != Teuchos::null) && (AcFact_ != NoFactory::getRCP())) {
-      RCP<Operator> oA = coarseLevel.Get<RCP<Operator> >("A", AcFact_.get());
-      RCP<Matrix>    A = rcp_dynamic_cast<Matrix>(oA);
-      if (!A.is_null()) coarseLevel.Set("A",  A, NoFactory::get());
-      else              coarseLevel.Set("A", oA, NoFactory::get());
-      coarseLevel.AddKeepFlag   ("A", NoFactory::get(), MueLu::Final);
-      coarseLevel.RemoveKeepFlag("A", NoFactory::get(), MueLu::UserData); // FIXME: This is a hack
-    }
+  if ((AcFact_ != Teuchos::null) && (AcFact_ != NoFactory::getRCP())) {
+    RCP<Operator> oA = coarseLevel.Get<RCP<Operator> >("A", AcFact_.get());
+    RCP<Matrix> A    = rcp_dynamic_cast<Matrix>(oA);
+    if (!A.is_null())
+      coarseLevel.Set("A", A, NoFactory::get());
+    else
+      coarseLevel.Set("A", oA, NoFactory::get());
+    coarseLevel.AddKeepFlag("A", NoFactory::get(), MueLu::Final);
+    coarseLevel.RemoveKeepFlag("A", NoFactory::get(), MueLu::UserData);  // FIXME: This is a hack
   }
 }
-
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPRAPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_decl.hpp b/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_decl.hpp
index 8aee5b63c61c..3f96963e4970 100644
--- a/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_decl.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_decl.hpp
@@ -20,33 +20,31 @@
 //#include "MueLu_Hierarchy_fwd.hpp"
 //#include "MueLu_HierarchyManager_fwd.hpp"
 
-
 namespace MueLu {
 
-  template<class Scalar = DefaultScalar,
-           class LocalOrdinal = DefaultLocalOrdinal,
-           class GlobalOrdinal = DefaultGlobalOrdinal,
-           class Node = DefaultNode>
-  class TopSmootherFactory : public SingleLevelFactoryBase { //TODO: inherit from SmootherFactoryBase ?
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class TopSmootherFactory : public SingleLevelFactoryBase {  // TODO: inherit from SmootherFactoryBase ?
 #undef MUELU_TOPSMOOTHERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    TopSmootherFactory(RCP<const FactoryManagerBase> parentFactoryManager, const std::string& varName);
+ public:
+  TopSmootherFactory(RCP<const FactoryManagerBase> parentFactoryManager, const std::string& varName);
 
-    virtual ~TopSmootherFactory();
+  virtual ~TopSmootherFactory();
 
-    void DeclareInput(Level & level) const;
+  void DeclareInput(Level& level) const;
 
-    void Build(Level & level) const;
+  void Build(Level& level) const;
 
-  private:
-    RCP<const FactoryBase> preSmootherFact_;
-    RCP<const FactoryBase> postSmootherFact_;
-  };
+ private:
+  RCP<const FactoryBase> preSmootherFact_;
+  RCP<const FactoryBase> postSmootherFact_;
+};
 
-}
+}  // namespace MueLu
 
 #define MUELU_TOPSMOOTHERFACTORY_SHORT
 #endif /* PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPSMOOTHERFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_def.hpp b/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_def.hpp
index f4150e083f5a..a8fdd7407ffc 100644
--- a/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_def.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_TopSmootherFactory_def.hpp
@@ -21,111 +21,109 @@
 //#include "MueLu_HierarchyManager_fwd.hpp"
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopSmootherFactory(RCP<const FactoryManagerBase> parentFactoryManager, const std::string& varName) {
-    TEUCHOS_TEST_FOR_EXCEPTION(varName != "CoarseSolver" && varName != "Smoother", Exceptions::RuntimeError, "varName should be either \"CoarseSolver\" or \"Smoother\"");
-
-    if (varName == "CoarseSolver") {
-      // For coarsest level, we only need one smoother/solver
-      // If a user wants to do something weird there (like, solve coarsest system by using 2 forward
-      // GS and 1 backward GS), one can use MergedSmoother
-      RCP<const FactoryBase> coarseSolverFactory = parentFactoryManager->GetFactory("CoarseSolver");
-      RCP<const SmootherFactory> coarseSmootherFactory = Teuchos::rcp_dynamic_cast<const SmootherFactory>(coarseSolverFactory);
-      if (coarseSmootherFactory != Teuchos::null) {
-        RCP<SmootherPrototype> preProto;
-        RCP<SmootherPrototype> postProto;
-        coarseSmootherFactory->GetSmootherPrototypes(preProto, postProto);
-
-        if (preProto == postProto)
-          preSmootherFact_  = parentFactoryManager->GetFactory("CoarseSolver");
-        else {
-          // check whether pre- and/or post-smoothing is desired on coarsest level
-          if(preProto != Teuchos::null)
-            preSmootherFact_  = parentFactoryManager->GetFactory("CoarseSolver");
-          if(postProto != Teuchos::null)
-            postSmootherFact_  = parentFactoryManager->GetFactory("CoarseSolver");
-        }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TopSmootherFactory(RCP<const FactoryManagerBase> parentFactoryManager, const std::string& varName) {
+  TEUCHOS_TEST_FOR_EXCEPTION(varName != "CoarseSolver" && varName != "Smoother", Exceptions::RuntimeError, "varName should be either \"CoarseSolver\" or \"Smoother\"");
+
+  if (varName == "CoarseSolver") {
+    // For coarsest level, we only need one smoother/solver
+    // If a user wants to do something weird there (like, solve coarsest system by using 2 forward
+    // GS and 1 backward GS), one can use MergedSmoother
+    RCP<const FactoryBase> coarseSolverFactory       = parentFactoryManager->GetFactory("CoarseSolver");
+    RCP<const SmootherFactory> coarseSmootherFactory = Teuchos::rcp_dynamic_cast<const SmootherFactory>(coarseSolverFactory);
+    if (coarseSmootherFactory != Teuchos::null) {
+      RCP<SmootherPrototype> preProto;
+      RCP<SmootherPrototype> postProto;
+      coarseSmootherFactory->GetSmootherPrototypes(preProto, postProto);
+
+      if (preProto == postProto)
+        preSmootherFact_ = parentFactoryManager->GetFactory("CoarseSolver");
+      else {
+        // check whether pre- and/or post-smoothing is desired on coarsest level
+        if (preProto != Teuchos::null)
+          preSmootherFact_ = parentFactoryManager->GetFactory("CoarseSolver");
+        if (postProto != Teuchos::null)
+          postSmootherFact_ = parentFactoryManager->GetFactory("CoarseSolver");
       }
-      else  // default handling: get default direct solver as presmoother on coarsest level
-        preSmootherFact_  = parentFactoryManager->GetFactory("CoarseSolver");
+    } else  // default handling: get default direct solver as presmoother on coarsest level
+      preSmootherFact_ = parentFactoryManager->GetFactory("CoarseSolver");
 
-    } else {
-      preSmootherFact_  = parentFactoryManager->GetFactory("PreSmoother");
-      postSmootherFact_ = parentFactoryManager->GetFactory("PostSmoother");
-    }
+  } else {
+    preSmootherFact_  = parentFactoryManager->GetFactory("PreSmoother");
+    postSmootherFact_ = parentFactoryManager->GetFactory("PostSmoother");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~TopSmootherFactory() { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~TopSmootherFactory() {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & level) const {
-    if (preSmootherFact_  != Teuchos::null)
-      level.DeclareInput("PreSmoother",  preSmootherFact_.get());
-    if (postSmootherFact_ != Teuchos::null)
-      level.DeclareInput("PostSmoother", postSmootherFact_.get());
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& level) const {
+  if (preSmootherFact_ != Teuchos::null)
+    level.DeclareInput("PreSmoother", preSmootherFact_.get());
+  if (postSmootherFact_ != Teuchos::null)
+    level.DeclareInput("PostSmoother", postSmootherFact_.get());
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & level) const {
-    if (preSmootherFact_.is_null() && postSmootherFact_.is_null())
-      return;
-
-    // NOTE 1: We need to set at least some keep flag for the smoothers, otherwise it is going to be removed as soon as all requests are released.
-    // We choose to set the Final flag for the data. In addition, we allow this data to be retrieved by only using the name by the means
-    // of using NoFactory. However, any data set with NoFactory gets UserData flag by default. We don't really want that flag, so we remove it.
-
-    // NOTE 2: some smoother factories are tricky (see comments in MueLu::SmootherFactory
-    // Sometimes, we don't know whether the factory is able to generate "PreSmoother" or "PostSmoother"
-    // For the SmootherFactory, however, we are able to check that.
-
-    if (!preSmootherFact_.is_null()) {
-      // Checking for null is not sufficient, as SmootherFactory(null, something) does not generate "PreSmoother"
-      bool isAble = true;
-      RCP<const SmootherFactory> s = rcp_dynamic_cast<const SmootherFactory>(preSmootherFact_);
-      if (!s.is_null()) {
-        RCP<SmootherPrototype> pre, post;
-        s->GetSmootherPrototypes(pre, post);
-        if (pre.is_null())
-          isAble = false;
-      } else {
-        // We assume that  if presmoother factory is not SmootherFactory, it *is* able to generate "PreSmoother"
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
+  if (preSmootherFact_.is_null() && postSmootherFact_.is_null())
+    return;
+
+  // NOTE 1: We need to set at least some keep flag for the smoothers, otherwise it is going to be removed as soon as all requests are released.
+  // We choose to set the Final flag for the data. In addition, we allow this data to be retrieved by only using the name by the means
+  // of using NoFactory. However, any data set with NoFactory gets UserData flag by default. We don't really want that flag, so we remove it.
+
+  // NOTE 2: some smoother factories are tricky (see comments in MueLu::SmootherFactory
+  // Sometimes, we don't know whether the factory is able to generate "PreSmoother" or "PostSmoother"
+  // For the SmootherFactory, however, we are able to check that.
+
+  if (!preSmootherFact_.is_null()) {
+    // Checking for null is not sufficient, as SmootherFactory(null, something) does not generate "PreSmoother"
+    bool isAble                  = true;
+    RCP<const SmootherFactory> s = rcp_dynamic_cast<const SmootherFactory>(preSmootherFact_);
+    if (!s.is_null()) {
+      RCP<SmootherPrototype> pre, post;
+      s->GetSmootherPrototypes(pre, post);
+      if (pre.is_null())
+        isAble = false;
+    } else {
+      // We assume that  if presmoother factory is not SmootherFactory, it *is* able to generate "PreSmoother"
+    }
 
-      if (isAble) {
-        RCP<SmootherBase> Pre  = level.Get<RCP<SmootherBase> >("PreSmoother", preSmootherFact_.get());
+    if (isAble) {
+      RCP<SmootherBase> Pre = level.Get<RCP<SmootherBase> >("PreSmoother", preSmootherFact_.get());
 
-        level.Set           ("PreSmoother", Pre, NoFactory::get());
+      level.Set("PreSmoother", Pre, NoFactory::get());
 
-        level.AddKeepFlag   ("PreSmoother", NoFactory::get(), MueLu::Final);
-        level.RemoveKeepFlag("PreSmoother", NoFactory::get(), MueLu::UserData);
-      }
+      level.AddKeepFlag("PreSmoother", NoFactory::get(), MueLu::Final);
+      level.RemoveKeepFlag("PreSmoother", NoFactory::get(), MueLu::UserData);
     }
+  }
 
-    if (!postSmootherFact_.is_null()) {
-      // Checking for null is not sufficient, as SmootherFactory(something, null) does not generate "PostSmoother"
-      bool isAble = true;
-      RCP<const SmootherFactory> s = rcp_dynamic_cast<const SmootherFactory>(postSmootherFact_);
-      if (!s.is_null()) {
-        RCP<SmootherPrototype> pre, post;
-        s->GetSmootherPrototypes(pre, post);
-        if (post.is_null())
-          isAble = false;
-      } else {
-        // We assume that  if presmoother factory is not SmootherFactory, it *is* able to generate "PreSmoother"
-      }
+  if (!postSmootherFact_.is_null()) {
+    // Checking for null is not sufficient, as SmootherFactory(something, null) does not generate "PostSmoother"
+    bool isAble                  = true;
+    RCP<const SmootherFactory> s = rcp_dynamic_cast<const SmootherFactory>(postSmootherFact_);
+    if (!s.is_null()) {
+      RCP<SmootherPrototype> pre, post;
+      s->GetSmootherPrototypes(pre, post);
+      if (post.is_null())
+        isAble = false;
+    } else {
+      // We assume that  if presmoother factory is not SmootherFactory, it *is* able to generate "PreSmoother"
+    }
 
-      if (isAble) {
-        RCP<SmootherBase> Post = level.Get<RCP<SmootherBase> >("PostSmoother", postSmootherFact_.get());
+    if (isAble) {
+      RCP<SmootherBase> Post = level.Get<RCP<SmootherBase> >("PostSmoother", postSmootherFact_.get());
 
-        level.Set           ("PostSmoother", Post, NoFactory::get());
+      level.Set("PostSmoother", Post, NoFactory::get());
 
-        level.AddKeepFlag   ("PostSmoother", NoFactory::get(), MueLu::Final);
-        level.RemoveKeepFlag("PostSmoother", NoFactory::get(), MueLu::UserData);
-      }
+      level.AddKeepFlag("PostSmoother", NoFactory::get(), MueLu::Final);
+      level.RemoveKeepFlag("PostSmoother", NoFactory::get(), MueLu::UserData);
     }
   }
 }
-
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_MUECENTRAL_MUELU_TOPSMOOTHERFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/MueCentral/MueLu_TwoLevelFactoryBase.hpp b/packages/muelu/src/MueCentral/MueLu_TwoLevelFactoryBase.hpp
index d21b4d194934..e55308ddaace 100644
--- a/packages/muelu/src/MueCentral/MueLu_TwoLevelFactoryBase.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_TwoLevelFactoryBase.hpp
@@ -55,121 +55,116 @@
 
 namespace MueLu {
 
-  /*!
-    @class TwoLevelFactoryBase class.
-    @brief Base class for factories that use two levels (fineLevel and coarseLevel).
+/*!
+  @class TwoLevelFactoryBase class.
+  @brief Base class for factories that use two levels (fineLevel and coarseLevel).
 
-    Examples of such factories are R, P, and A_coarse.
+  Examples of such factories are R, P, and A_coarse.
 
-    @ingroup MueLuBaseClasses
-  */
+  @ingroup MueLuBaseClasses
+*/
 
-  class TwoLevelFactoryBase : public Factory {
+class TwoLevelFactoryBase : public Factory {
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-  public:
+  //! Constructor.
+  TwoLevelFactoryBase() {}
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Destructor.
+  virtual ~TwoLevelFactoryBase() {}
 
-    //! Constructor.
-    TwoLevelFactoryBase() { }
+  //@}
 
-    //! Destructor.
-    virtual ~TwoLevelFactoryBase()
-    { }
+  //! Input
+  //@{
 
-    //@}
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //! Input
-    //@{
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  virtual void DeclareInput(Level& fineLevel, Level& coarseLevel) const = 0;
+
+  //!
+  virtual void CallDeclareInput(Level& requestedLevel) const {
+    if (requestedLevel.GetPreviousLevel() == Teuchos::null) {
+      std::ostringstream errStr;
+      errStr << "LevelID = " << requestedLevel.GetLevelID();
+      throw Exceptions::DependencyError(errStr.str());
+    }
+    DeclareInput(*requestedLevel.GetPreviousLevel(), requestedLevel);
+  }
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //@}
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    virtual void DeclareInput(Level &fineLevel, Level &coarseLevel) const = 0;
+  //! @name Build methods.
+  //@{
 
-    //!
-    virtual void CallDeclareInput(Level & requestedLevel) const {
-      if (requestedLevel.GetPreviousLevel() == Teuchos::null) {
-        std::ostringstream errStr;
-        errStr << "LevelID = " << requestedLevel.GetLevelID();
-        throw Exceptions::DependencyError(errStr.str());
-      }
-      DeclareInput(*requestedLevel.GetPreviousLevel(), requestedLevel);
-    }
+  //! Build an object with this factory.
+  virtual void Build(Level& fineLevel, Level& coarseLevel) const = 0;
 
-    //@}
+  //!
+  virtual void CallBuild(Level& requestedLevel) const {
+    int levelID = requestedLevel.GetLevelID();
 
-    //! @name Build methods.
-    //@{
+#ifdef HAVE_MUELU_DEBUG
+    // We cannot call Build method twice for the same level, but we can call it multiple times for different levels
+    TEUCHOS_TEST_FOR_EXCEPTION((multipleCallCheck_ == ENABLED) && (multipleCallCheckGlobal_ == ENABLED) && (lastLevelID_ == levelID),
+                               Exceptions::RuntimeError,
+                               this->ShortClassName() << "::Build() called twice for the same level (levelID=" << levelID
+                                                      << "). This is likely due to a configuration error, or calling hierarchy setup multiple times "
+                                                      << "without resetting debug info through FactoryManager::ResetDebugData().");
+    if (multipleCallCheck_ == FIRSTCALL)
+      multipleCallCheck_ = ENABLED;
+
+    lastLevelID_ = levelID;
+#endif
+    TEUCHOS_TEST_FOR_EXCEPTION(requestedLevel.GetPreviousLevel() == Teuchos::null, Exceptions::RuntimeError, "LevelID = " << levelID);
+
+    RCP<const Teuchos::Comm<int> > comm = requestedLevel.GetComm();
+    if (comm.is_null()) {
+      // Some factories are called before we constructed Ac, and therefore,
+      // before we set the level communicator. For such factories we can get
+      // the comm from the previous level, as all processes go there
+      RCP<Level>& prevLevel = requestedLevel.GetPreviousLevel();
+      if (!prevLevel.is_null())
+        comm = prevLevel->GetComm();
+    }
 
+    int oldRank = -1;
+    if (!comm.is_null())
+      oldRank = SetProcRankVerbose(comm->getRank());
 
-    //! Build an object with this factory.
-    virtual void Build(Level & fineLevel, Level & coarseLevel) const = 0;
+    // Synchronization timer
+    std::string syncTimer = this->ShortClassName() + ": Build sync (level=" + toString(requestedLevel.GetLevelID()) + ")";
+    if (this->timerSync_ && !comm.is_null()) {
+      TimeMonitor timer(*this, syncTimer);
+      comm->barrier();
+    }
 
-    //!
-    virtual void CallBuild(Level& requestedLevel) const {
-      int levelID = requestedLevel.GetLevelID();
+    Build(*requestedLevel.GetPreviousLevel(), requestedLevel);
 
-#ifdef HAVE_MUELU_DEBUG
-      // We cannot call Build method twice for the same level, but we can call it multiple times for different levels
-      TEUCHOS_TEST_FOR_EXCEPTION((multipleCallCheck_ == ENABLED) && (multipleCallCheckGlobal_ == ENABLED) && (lastLevelID_ == levelID),
-                                 Exceptions::RuntimeError,
-                                 this->ShortClassName() << "::Build() called twice for the same level (levelID=" << levelID
-                                 << "). This is likely due to a configuration error, or calling hierarchy setup multiple times "
-                                 << "without resetting debug info through FactoryManager::ResetDebugData().");
-      if (multipleCallCheck_ == FIRSTCALL)
-        multipleCallCheck_ = ENABLED;
-
-      lastLevelID_ = levelID;
-#endif
-      TEUCHOS_TEST_FOR_EXCEPTION(requestedLevel.GetPreviousLevel() == Teuchos::null, Exceptions::RuntimeError, "LevelID = " << levelID);
-
-      RCP<const Teuchos::Comm<int> > comm = requestedLevel.GetComm();
-      if (comm.is_null()) {
-        // Some factories are called before we constructed Ac, and therefore,
-        // before we set the level communicator. For such factories we can get
-        // the comm from the previous level, as all processes go there
-        RCP<Level>& prevLevel = requestedLevel.GetPreviousLevel();
-        if (!prevLevel.is_null())
-          comm = prevLevel->GetComm();
-      }
-
-      int oldRank = -1;
-      if (!comm.is_null())
-        oldRank = SetProcRankVerbose(comm->getRank());
-
-      // Synchronization timer
-      std::string syncTimer = this->ShortClassName() + ": Build sync (level=" + toString(requestedLevel.GetLevelID()) + ")";
-      if (this->timerSync_ && !comm.is_null()) {
-        TimeMonitor timer(*this, syncTimer);
-        comm->barrier();
-      }
-
-      Build(*requestedLevel.GetPreviousLevel(), requestedLevel);
-
-      if (this->timerSync_ && !comm.is_null()) {
-        TimeMonitor timer(*this, syncTimer);
-        comm->barrier();
-      }
-
-      if (IsPrint(Test1))
-        GetOStream(Test1) << *RemoveFactoriesFromList(GetParameterList()) << std::endl;
-      else
-        RemoveFactoriesFromList(GetParameterList())->print(GetOStream(Test0), Teuchos::ParameterList::PrintOptions().indent(0).showFlags(true).showDefault(false));
-
-      if (oldRank != -1)
-        SetProcRankVerbose(oldRank);
+    if (this->timerSync_ && !comm.is_null()) {
+      TimeMonitor timer(*this, syncTimer);
+      comm->barrier();
     }
 
-    //@}
+    if (IsPrint(Test1))
+      GetOStream(Test1) << *RemoveFactoriesFromList(GetParameterList()) << std::endl;
+    else
+      RemoveFactoriesFromList(GetParameterList())->print(GetOStream(Test0), Teuchos::ParameterList::PrintOptions().indent(0).showFlags(true).showDefault(false));
+
+    if (oldRank != -1)
+      SetProcRankVerbose(oldRank);
+  }
 
-  }; //class TwoLevelFactoryBase
+  //@}
 
+};  // class TwoLevelFactoryBase
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TWOLEVELFACTORY_SHORT
-#endif //ifndef MUELU_TWOLEVELFACTORY_HPP
+#endif  // ifndef MUELU_TWOLEVELFACTORY_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_VariableContainer.hpp b/packages/muelu/src/MueCentral/MueLu_VariableContainer.hpp
index a0a84bbf9124..c1b5ffeb5d4c 100644
--- a/packages/muelu/src/MueCentral/MueLu_VariableContainer.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_VariableContainer.hpp
@@ -62,262 +62,275 @@
 
 namespace MueLu {
 
-  /*!
-    @class VariableContainer
-    @brief Class that stores all relevant data for a variable
-
-    Maintains all data for a variable, that is, the data itself, a boolean flag for the "Keep" status,
-    a boolean flag for the "Available" status, a reference counter for all requests and a list with
-    all requesting factories.
-  */
-  class VariableContainer : public BaseClass {
-  private:
-    // Motivated by Teuchos_any.hpp
-    class DataBase {
-    public:
-      virtual ~DataBase() {}
-      virtual const std::type_info& type() const = 0;
-      virtual std::string typeName() const = 0;
-    };
-
-    template<typename T>
-    class Data : public DataBase {
-    public:
-      Data(const T& data) : data_(data) {}
-      const std::type_info& type() const { return typeid(T); }
-      std::string typeName() const { return Teuchos::TypeNameTraits<T>::name(); }
-      T data_;
-    };
-
-    template<typename T>
-    struct Getter {
-      static T& get(DataBase* data_, DataBase*& /* datah_ */) {
-        if ((data_ == NULL) || (data_->type() != typeid(T))) // NVR added guard to avoid determining typeName unless we will use it
-        {
-          const std::string typeName = Teuchos::TypeNameTraits<T>::name(); // calls Teuchos::demangleName(), which can be expensive
-          TEUCHOS_TEST_FOR_EXCEPTION(data_ == NULL, Teuchos::bad_any_cast,
-                                     "Error, cast to type Data<" << typeName << "> failed since the content is NULL");
-        
-          TEUCHOS_TEST_FOR_EXCEPTION(data_->type() != typeid(T), Teuchos::bad_any_cast,
-                                     "Error, cast to type Data<" << typeName << "> failed since the actual underlying type is "
-                                     "\'" << data_->typeName() << "!");
-        }
-
-        Data<T>* data = dynamic_cast<Data<T>*>(data_);
-        if (!data) // NVR added guard to avoid determining typeName unless we will use it
-        {
-          const std::string typeName = Teuchos::TypeNameTraits<T>::name(); // calls Teuchos::demangleName(), which can be expensive
-          TEUCHOS_TEST_FOR_EXCEPTION(!data, std::logic_error,
-                                     "Error, cast to type Data<" << typeName << "> failed but should not have and the actual underlying type is "
-                                     "\'" << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
-        }
-        return data->data_;
-      }
-    };
-
-    template<typename T>
-    struct Checker {
-      static bool check(DataBase* data_, DataBase*& /* datah_ */) {
-        if ((data_ == NULL) || (data_->type() != typeid(T))) // NVR added guard to avoid determining typeName unless we will use it
-        {
-          return false;
-        }
-
-        Data<T>* data = dynamic_cast<Data<T>*>(data_);
-        if (!data) // NVR added guard to avoid determining typeName unless we will use it
-        {
-          return false;
-        }
-        return true;
-      }
-    };
-
-
-  public:
-    typedef std::map<const FactoryBase*,int> request_container;
-
-  private:
-    DataBase*          data_;        ///< the data itself
-    mutable
-    DataBase*          datah_;       ///< temporary data storage (need to get a reference
-                                     ///< to RCP to a base class (like Operator)
-    bool               available_;   ///< is data available?
-    KeepType           keep_;        ///< keep flag
-    int                count_;       ///< number of requests by all factories
-
-    request_container  requests_;    ///< requesting factories
-
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Default constructor.
-    VariableContainer() : data_(NULL), datah_(NULL), available_(false), keep_(false), count_(0) { }
-    ~VariableContainer() {
-      delete data_;   data_  = NULL;
-      delete datah_;  datah_ = NULL;
-    }
-
-    //@}
-
-    //! @name Data access
-    //@{
-
-    //! Store data in container class and set the "Available" status true.
-    template<typename T>
-    void SetData(const T& entry) {
-      delete data_;
-      delete datah_;
-      data_      = new Data<T>(entry);
-      datah_     = NULL;
-      available_ = true;
-    }
-
-    //! Return const reference to data stored in container
-    //! NOTE: we do not check if data is available
-    template<typename T>
-    const T& GetData() const {
-      return Getter<T>::get(data_, datah_);
-    }
-
-    //! Return reference to data stored in container
-    //! NOTE: we do not check if data is available
-    template<typename T>
-    T& GetData() {
-      return Getter<T>::get(data_, datah_);
-    }
-
-    //! Return reference to data stored in container
-    //! NOTE: we do not check if data is available
-    template<typename T>
-    bool CheckType() {
-      return Checker<T>::check(data_, datah_);
-    }
-
-    std::string GetTypeName() {
-      if (data_ == NULL)
-        return std::string("");
-      return data_->typeName();
-    }
-
-    //! Returns true if data is available, i.e.
-    //  if SetData has been called before
-    bool IsAvailable() const { return available_; }
-
-    //@}
+/*!
+  @class VariableContainer
+  @brief Class that stores all relevant data for a variable
+
+  Maintains all data for a variable, that is, the data itself, a boolean flag for the "Keep" status,
+  a boolean flag for the "Available" status, a reference counter for all requests and a list with
+  all requesting factories.
+*/
+class VariableContainer : public BaseClass {
+ private:
+  // Motivated by Teuchos_any.hpp
+  class DataBase {
+   public:
+    virtual ~DataBase() {}
+    virtual const std::type_info& type() const = 0;
+    virtual std::string typeName() const       = 0;
+  };
 
-    //! @name Request/Release
-    //@{
+  template <typename T>
+  class Data : public DataBase {
+   public:
+    Data(const T& data)
+      : data_(data) {}
+    const std::type_info& type() const { return typeid(T); }
+    std::string typeName() const { return Teuchos::TypeNameTraits<T>::name(); }
+    T data_;
+  };
 
-    //! Request data
-    void Request(const FactoryBase* reqFactory) {
-      request_container::iterator it = requests_.find(reqFactory);
-      if (it == requests_.end())
-        requests_[reqFactory] = 1;
-      else
-        (it->second)++;
-      count_++;   // increment request counter
-    }
+  template <typename T>
+  struct Getter {
+    static T& get(DataBase* data_, DataBase*& /* datah_ */) {
+      if ((data_ == NULL) || (data_->type() != typeid(T)))  // NVR added guard to avoid determining typeName unless we will use it
+      {
+        const std::string typeName = Teuchos::TypeNameTraits<T>::name();  // calls Teuchos::demangleName(), which can be expensive
+        TEUCHOS_TEST_FOR_EXCEPTION(data_ == NULL, Teuchos::bad_any_cast,
+                                   "Error, cast to type Data<" << typeName << "> failed since the content is NULL");
 
-    //! Release data
-    void Release(const FactoryBase* reqFactory) {
-      request_container::iterator it = requests_.find(reqFactory);
-      TEUCHOS_TEST_FOR_EXCEPTION(it == requests_.end(), Exceptions::RuntimeError, "MueLu::VariableContainer::Release(): "
-                                 "cannot call Release if factory has not been requested before by factory " << reqFactory);
-      if (--(it->second) == 0)
-        requests_.erase(it);
-      count_--;
-    }
+        TEUCHOS_TEST_FOR_EXCEPTION(data_->type() != typeid(T), Teuchos::bad_any_cast,
+                                   "Error, cast to type Data<" << typeName << "> failed since the actual underlying type is "
+                                                                              "\'"
+                                                               << data_->typeName() << "!");
+      }
 
-    //! Return the number of times the data has been requested by a specific factory
-    int NumRequests(const FactoryBase* reqFactory) const {
-      request_container::const_iterator it = requests_.find(reqFactory);
-      return (it != requests_.end()) ? it->second : 0;
+      Data<T>* data = dynamic_cast<Data<T>*>(data_);
+      if (!data)  // NVR added guard to avoid determining typeName unless we will use it
+      {
+        const std::string typeName = Teuchos::TypeNameTraits<T>::name();  // calls Teuchos::demangleName(), which can be expensive
+        TEUCHOS_TEST_FOR_EXCEPTION(!data, std::logic_error,
+                                   "Error, cast to type Data<" << typeName << "> failed but should not have and the actual underlying type is "
+                                                                              "\'"
+                                                               << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
+      }
+      return data->data_;
     }
-
-    //! Returns the number of times the data has been requested
-    int NumAllRequests() const                                                  { return count_; }
-
-    //! Returns true, if data is requested by reqFactory
-    bool IsRequested(const FactoryBase* reqFactory) const                       { return (NumRequests(reqFactory) > 0); }
-
-    //! Returns true, if data is requested by at least one factory
-    bool IsRequested() const                                                    { return (count_ > 0); }
-
-    const request_container& Requests() const                                   { return requests_; }
-    //@}
-
-    //! @name Keep status
-    //@{
-
-    //! Returns true if at least one keep flag is set
-    bool IsKept(KeepType keep) const                                            { return keep_ & keep; }
-
-    //! Adds a keep flag to the flag combination
-    void AddKeepFlag(KeepType keep = UserData)                                  { keep_ |= keep; }
-
-    //! Removes a keep flag to the flag combination
-    void RemoveKeepFlag(KeepType keep = UserData)                               { keep_ = keep_ & (keep_ ^ keep); }
-
-    //! Returns the keep flag combination
-    KeepType GetKeepFlag() const                                                { return keep_; }
-
-    //@}
   };
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  struct VariableContainer::Getter<Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > > {
-    typedef Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> Operator;
-    typedef Xpetra::Matrix  <Scalar,LocalOrdinal,GlobalOrdinal,Node> Matrix;
-
-    static Teuchos::RCP<Operator>& get(DataBase* data_, DataBase*& datah_) {
-      typedef Teuchos::RCP<Operator> TO;
-      typedef Teuchos::RCP<Matrix>   TM;
-
-      if (data_ == NULL) // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+  template <typename T>
+  struct Checker {
+    static bool check(DataBase* data_, DataBase*& /* datah_ */) {
+      if ((data_ == NULL) || (data_->type() != typeid(T)))  // NVR added guard to avoid determining typeName unless we will use it
       {
-        TEUCHOS_TEST_FOR_EXCEPTION(data_ == NULL, Teuchos::bad_any_cast,
-                                   "Error, cast to type Data<" << Teuchos::TypeNameTraits<TO>::name() << "> failed since the content is NULL");
-      }
-      if (data_->type() == typeid(TO)) {
-        Data<TO>* data = dynamic_cast<Data<TO>*>(data_);
-        if (!data) // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
-        {
-          TEUCHOS_TEST_FOR_EXCEPTION(!data, std::logic_error,
-                                     "Error, cast to type Data<" << Teuchos::TypeNameTraits<TO>::name() << "> failed but should not have and the actual underlying type is "
-                                     "\'" << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
-        }
-        return data->data_;
+        return false;
       }
 
-      if (data_->type() != typeid(TM)) // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+      Data<T>* data = dynamic_cast<Data<T>*>(data_);
+      if (!data)  // NVR added guard to avoid determining typeName unless we will use it
       {
-        TEUCHOS_TEST_FOR_EXCEPTION(data_->type() != typeid(TM), Teuchos::bad_any_cast,
-                                   "Error, cast to type Data<" << Teuchos::TypeNameTraits<TM>::name() << "> failed since the actual underlying type is "
-                                   "\'" << data_->typeName() << "!");
+        return false;
       }
-      Data<TM>* data = dynamic_cast<Data<TM>*>(data_);
-      if (!data) // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+      return true;
+    }
+  };
+
+ public:
+  typedef std::map<const FactoryBase*, int> request_container;
+
+ private:
+  DataBase* data_;           ///< the data itself
+  mutable DataBase* datah_;  ///< temporary data storage (need to get a reference
+                             ///< to RCP to a base class (like Operator)
+  bool available_;           ///< is data available?
+  KeepType keep_;            ///< keep flag
+  int count_;                ///< number of requests by all factories
+
+  request_container requests_;  ///< requesting factories
+
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Default constructor.
+  VariableContainer()
+    : data_(NULL)
+    , datah_(NULL)
+    , available_(false)
+    , keep_(false)
+    , count_(0) {}
+  ~VariableContainer() {
+    delete data_;
+    data_ = NULL;
+    delete datah_;
+    datah_ = NULL;
+  }
+
+  //@}
+
+  //! @name Data access
+  //@{
+
+  //! Store data in container class and set the "Available" status true.
+  template <typename T>
+  void SetData(const T& entry) {
+    delete data_;
+    delete datah_;
+    data_      = new Data<T>(entry);
+    datah_     = NULL;
+    available_ = true;
+  }
+
+  //! Return const reference to data stored in container
+  //! NOTE: we do not check if data is available
+  template <typename T>
+  const T& GetData() const {
+    return Getter<T>::get(data_, datah_);
+  }
+
+  //! Return reference to data stored in container
+  //! NOTE: we do not check if data is available
+  template <typename T>
+  T& GetData() {
+    return Getter<T>::get(data_, datah_);
+  }
+
+  //! Return reference to data stored in container
+  //! NOTE: we do not check if data is available
+  template <typename T>
+  bool CheckType() {
+    return Checker<T>::check(data_, datah_);
+  }
+
+  std::string GetTypeName() {
+    if (data_ == NULL)
+      return std::string("");
+    return data_->typeName();
+  }
+
+  //! Returns true if data is available, i.e.
+  //  if SetData has been called before
+  bool IsAvailable() const { return available_; }
+
+  //@}
+
+  //! @name Request/Release
+  //@{
+
+  //! Request data
+  void Request(const FactoryBase* reqFactory) {
+    request_container::iterator it = requests_.find(reqFactory);
+    if (it == requests_.end())
+      requests_[reqFactory] = 1;
+    else
+      (it->second)++;
+    count_++;  // increment request counter
+  }
+
+  //! Release data
+  void Release(const FactoryBase* reqFactory) {
+    request_container::iterator it = requests_.find(reqFactory);
+    TEUCHOS_TEST_FOR_EXCEPTION(it == requests_.end(), Exceptions::RuntimeError,
+                               "MueLu::VariableContainer::Release(): "
+                               "cannot call Release if factory has not been requested before by factory "
+                                   << reqFactory);
+    if (--(it->second) == 0)
+      requests_.erase(it);
+    count_--;
+  }
+
+  //! Return the number of times the data has been requested by a specific factory
+  int NumRequests(const FactoryBase* reqFactory) const {
+    request_container::const_iterator it = requests_.find(reqFactory);
+    return (it != requests_.end()) ? it->second : 0;
+  }
+
+  //! Returns the number of times the data has been requested
+  int NumAllRequests() const { return count_; }
+
+  //! Returns true, if data is requested by reqFactory
+  bool IsRequested(const FactoryBase* reqFactory) const { return (NumRequests(reqFactory) > 0); }
+
+  //! Returns true, if data is requested by at least one factory
+  bool IsRequested() const { return (count_ > 0); }
+
+  const request_container& Requests() const { return requests_; }
+  //@}
+
+  //! @name Keep status
+  //@{
+
+  //! Returns true if at least one keep flag is set
+  bool IsKept(KeepType keep) const { return keep_ & keep; }
+
+  //! Adds a keep flag to the flag combination
+  void AddKeepFlag(KeepType keep = UserData) { keep_ |= keep; }
+
+  //! Removes a keep flag to the flag combination
+  void RemoveKeepFlag(KeepType keep = UserData) { keep_ = keep_ & (keep_ ^ keep); }
+
+  //! Returns the keep flag combination
+  KeepType GetKeepFlag() const { return keep_; }
+
+  //@}
+};
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct VariableContainer::Getter<Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > > {
+  typedef Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> Operator;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+
+  static Teuchos::RCP<Operator>& get(DataBase* data_, DataBase*& datah_) {
+    typedef Teuchos::RCP<Operator> TO;
+    typedef Teuchos::RCP<Matrix> TM;
+
+    if (data_ == NULL)  // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+    {
+      TEUCHOS_TEST_FOR_EXCEPTION(data_ == NULL, Teuchos::bad_any_cast,
+                                 "Error, cast to type Data<" << Teuchos::TypeNameTraits<TO>::name() << "> failed since the content is NULL");
+    }
+    if (data_->type() == typeid(TO)) {
+      Data<TO>* data = dynamic_cast<Data<TO>*>(data_);
+      if (!data)  // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
       {
         TEUCHOS_TEST_FOR_EXCEPTION(!data, std::logic_error,
-                                   "Error, cast to type Data<" << Teuchos::TypeNameTraits<TM>::name() << "> failed but should not have and the actual underlying type is "
-                                   "\'" << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
-      }
-      if (datah_ == NULL)
-        datah_ = new Data<TO>(Teuchos::rcp_dynamic_cast<Operator>(data->data_));
-      Data<TO>* datah = dynamic_cast<Data<TO>*>(datah_);
-      
-      if (!datah) // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
-      {
-        TEUCHOS_TEST_FOR_EXCEPTION(!datah, std::logic_error,
                                    "Error, cast to type Data<" << Teuchos::TypeNameTraits<TO>::name() << "> failed but should not have and the actual underlying type is "
-                                   "\'" << datah_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
+                                                                                                         "\'"
+                                                               << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
       }
-      return datah->data_;
+      return data->data_;
     }
-  };
 
-}
+    if (data_->type() != typeid(TM))  // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+    {
+      TEUCHOS_TEST_FOR_EXCEPTION(data_->type() != typeid(TM), Teuchos::bad_any_cast,
+                                 "Error, cast to type Data<" << Teuchos::TypeNameTraits<TM>::name() << "> failed since the actual underlying type is "
+                                                                                                       "\'"
+                                                             << data_->typeName() << "!");
+    }
+    Data<TM>* data = dynamic_cast<Data<TM>*>(data_);
+    if (!data)  // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+    {
+      TEUCHOS_TEST_FOR_EXCEPTION(!data, std::logic_error,
+                                 "Error, cast to type Data<" << Teuchos::TypeNameTraits<TM>::name() << "> failed but should not have and the actual underlying type is "
+                                                                                                       "\'"
+                                                             << data_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
+    }
+    if (datah_ == NULL)
+      datah_ = new Data<TO>(Teuchos::rcp_dynamic_cast<Operator>(data->data_));
+    Data<TO>* datah = dynamic_cast<Data<TO>*>(datah_);
+
+    if (!datah)  // NVR added guard to avoid unnecessary call to TypeNameTraits<TO>::name(), which calls Teuchos::demangleName(), which can be expensive
+    {
+      TEUCHOS_TEST_FOR_EXCEPTION(!datah, std::logic_error,
+                                 "Error, cast to type Data<" << Teuchos::TypeNameTraits<TO>::name() << "> failed but should not have and the actual underlying type is "
+                                                                                                       "\'"
+                                                             << datah_->typeName() << "! The problem might be related to incompatible RTTI systems in static and shared libraries!");
+    }
+    return datah->data_;
+  }
+};
+
+}  // namespace MueLu
 
 #endif /* MUELU_VARIABLECONTAINER_HPP */
diff --git a/packages/muelu/src/MueCentral/MueLu_VerboseObject.cpp b/packages/muelu/src/MueCentral/MueLu_VerboseObject.cpp
index dce40d04eb5d..1da06b6d1e39 100644
--- a/packages/muelu/src/MueCentral/MueLu_VerboseObject.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_VerboseObject.cpp
@@ -58,110 +58,112 @@
 
 namespace MueLu {
 
-  VerboseObject::VerboseObject():
-    verbLevel_(NotSpecified), // = use global verbose level by default
-    numProcs_(0)
-  {
-    // Note: using MPI_COMM_RANK is bad idea (because a subcommunicator may be used to run MueLu)
-    // Belos have the same problem in the class BelosOutputManager.
-    //
-    // How to fix this: the FancyOStream provides setProcRankAndSize() to change the proc rank info used by setOutputToRootOnly().
-    // Adding a method FancyOStream::getProcRank() would be enough. And it makes sense to use the info that come from the stream configuration.
-    //
-    // Documentation: after the patch, it migh be nice to add to the documentation (Teuchos and MueLu)
-    //                that users of subcommunicators have to setup the output stream (of Teuchos::VerboseObject) separately.
-    procRank_ = 0;
+VerboseObject::VerboseObject()
+  : verbLevel_(NotSpecified)
+  ,  // = use global verbose level by default
+  numProcs_(0) {
+  // Note: using MPI_COMM_RANK is bad idea (because a subcommunicator may be used to run MueLu)
+  // Belos have the same problem in the class BelosOutputManager.
+  //
+  // How to fix this: the FancyOStream provides setProcRankAndSize() to change the proc rank info used by setOutputToRootOnly().
+  // Adding a method FancyOStream::getProcRank() would be enough. And it makes sense to use the info that come from the stream configuration.
+  //
+  // Documentation: after the patch, it migh be nice to add to the documentation (Teuchos and MueLu)
+  //                that users of subcommunicators have to setup the output stream (of Teuchos::VerboseObject) separately.
+  procRank_ = 0;
 #ifdef HAVE_MPI
-    int mpiStarted = 0; MPI_Initialized(&mpiStarted);
-    if (mpiStarted)     {
-      MPI_Comm_rank(MPI_COMM_WORLD, &procRank_);
-      MPI_Comm_size(MPI_COMM_WORLD, &numProcs_);
-    }
+  int mpiStarted = 0;
+  MPI_Initialized(&mpiStarted);
+  if (mpiStarted) {
+    MPI_Comm_rank(MPI_COMM_WORLD, &procRank_);
+    MPI_Comm_size(MPI_COMM_WORLD, &numProcs_);
+  }
 #endif
-    }
+}
 
-  VerboseObject::~VerboseObject() { }
+VerboseObject::~VerboseObject() {}
 
-  VerbLevel VerboseObject::GetVerbLevel() const {
-    return (verbLevel_ != NotSpecified ?  verbLevel_ : globalVerbLevel_);
-  }
+VerbLevel VerboseObject::GetVerbLevel() const {
+  return (verbLevel_ != NotSpecified ? verbLevel_ : globalVerbLevel_);
+}
 
-  void VerboseObject::SetVerbLevel(const VerbLevel verbLevel) {
-    verbLevel_ = verbLevel;
-  }
+void VerboseObject::SetVerbLevel(const VerbLevel verbLevel) {
+  verbLevel_ = verbLevel;
+}
 
-  int VerboseObject::GetProcRankVerbose() const {
-    return procRank_;
-  }
+int VerboseObject::GetProcRankVerbose() const {
+  return procRank_;
+}
 
-  int VerboseObject::SetProcRankVerbose(int procRank) const {
-    int oldRank = procRank_;
-    procRank_ = procRank;
-    return oldRank;
-  }
+int VerboseObject::SetProcRankVerbose(int procRank) const {
+  int oldRank = procRank_;
+  procRank_   = procRank;
+  return oldRank;
+}
 
-  bool VerboseObject::IsPrint(MsgType type, int thisProcRankOnly) const {
-    return ((type & GetVerbLevel()) && (thisProcRankOnly < 0 || procRank_ == thisProcRankOnly));
-  }
+bool VerboseObject::IsPrint(MsgType type, int thisProcRankOnly) const {
+  return ((type & GetVerbLevel()) && (thisProcRankOnly < 0 || procRank_ == thisProcRankOnly));
+}
 
-  Teuchos::FancyOStream & VerboseObject::GetOStream(MsgType type, int thisProcRankOnly) const {
-    if (!IsPrint(type, thisProcRankOnly))
-      return *blackHole_;
+Teuchos::FancyOStream& VerboseObject::GetOStream(MsgType type, int thisProcRankOnly) const {
+  if (!IsPrint(type, thisProcRankOnly))
+    return *blackHole_;
 
-    Teuchos::FancyOStream& os = *GetMueLuOStream();
-    if (!(type & ((Extreme | Test) ^ Warnings)))
-      os << "\n******* WARNING *******" << std::endl;
+  Teuchos::FancyOStream& os = *GetMueLuOStream();
+  if (!(type & ((Extreme | Test) ^ Warnings)))
+    os << "\n******* WARNING *******" << std::endl;
 
-    return os;
-  }
+  return os;
+}
 
-  Teuchos::FancyOStream& VerboseObject::GetBlackHole() const {
-    return *blackHole_;
-  }
+Teuchos::FancyOStream& VerboseObject::GetBlackHole() const {
+  return *blackHole_;
+}
 
-  void VerboseObject::SetDefaultVerbLevel(const VerbLevel defaultVerbLevel) {
-    TEUCHOS_TEST_FOR_EXCEPTION(defaultVerbLevel == NotSpecified, Exceptions::RuntimeError,
-                               "MueLu::VerboseObject::GetVerbLevel(): global verbose level cannot be 'NotSpecified'.");
-    globalVerbLevel_ = defaultVerbLevel;
-  }
+void VerboseObject::SetDefaultVerbLevel(const VerbLevel defaultVerbLevel) {
+  TEUCHOS_TEST_FOR_EXCEPTION(defaultVerbLevel == NotSpecified, Exceptions::RuntimeError,
+                             "MueLu::VerboseObject::GetVerbLevel(): global verbose level cannot be 'NotSpecified'.");
+  globalVerbLevel_ = defaultVerbLevel;
+}
 
-  VerbLevel VerboseObject::GetDefaultVerbLevel() {
-    return globalVerbLevel_;
-  }
+VerbLevel VerboseObject::GetDefaultVerbLevel() {
+  return globalVerbLevel_;
+}
 
-  void VerboseObject::SetMueLuOStream(const Teuchos::RCP<Teuchos::FancyOStream>& mueluOStream) {
-    mueluOStream->setOutputToRootOnly(-1);
-    mueluOutputStream_ = mueluOStream;
-  }
+void VerboseObject::SetMueLuOStream(const Teuchos::RCP<Teuchos::FancyOStream>& mueluOStream) {
+  mueluOStream->setOutputToRootOnly(-1);
+  mueluOutputStream_ = mueluOStream;
+}
 
-  void VerboseObject::SetMueLuOFileStream(const std::string& filename) {
-    std::string fn;
+void VerboseObject::SetMueLuOFileStream(const std::string& filename) {
+  std::string fn;
 #ifdef HAVE_MPI
-    int mpiStarted = 0; MPI_Initialized(&mpiStarted);
-    if (mpiStarted)     {
-      int procRank;
-      MPI_Comm_rank(MPI_COMM_WORLD, &procRank);
-      fn = filename + "." + std::to_string(procRank);
-    } else
+  int mpiStarted = 0;
+  MPI_Initialized(&mpiStarted);
+  if (mpiStarted) {
+    int procRank;
+    MPI_Comm_rank(MPI_COMM_WORLD, &procRank);
+    fn = filename + "." + std::to_string(procRank);
+  } else
 #endif
-      fn = filename;
-    RCP<std::ofstream> outFile(new std::ofstream(fn));
-    Teuchos::RCP<Teuchos::FancyOStream> fancyOutFile = Teuchos::fancyOStream(Teuchos::rcp_implicit_cast<std::ostream>(outFile));
-    SetMueLuOStream(fancyOutFile);
-  }
-
-  Teuchos::RCP<Teuchos::FancyOStream> VerboseObject::GetMueLuOStream() {
-    if (mueluOutputStream_.get()==NULL) {
-      mueluOutputStream_ = fancyOStream(rcpFromRef(std::cout));
-      mueluOutputStream_->setOutputToRootOnly(-1);
-    }
-    return mueluOutputStream_;
+    fn = filename;
+  RCP<std::ofstream> outFile(new std::ofstream(fn));
+  Teuchos::RCP<Teuchos::FancyOStream> fancyOutFile = Teuchos::fancyOStream(Teuchos::rcp_implicit_cast<std::ostream>(outFile));
+  SetMueLuOStream(fancyOutFile);
+}
+
+Teuchos::RCP<Teuchos::FancyOStream> VerboseObject::GetMueLuOStream() {
+  if (mueluOutputStream_.get() == NULL) {
+    mueluOutputStream_ = fancyOStream(rcpFromRef(std::cout));
+    mueluOutputStream_->setOutputToRootOnly(-1);
   }
+  return mueluOutputStream_;
+}
 
-  VerbLevel VerboseObject::globalVerbLevel_ = High; // Default global verbose level.
+VerbLevel VerboseObject::globalVerbLevel_ = High;  // Default global verbose level.
 
-  RCP<Teuchos::FancyOStream> VerboseObject::blackHole_ = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+RCP<Teuchos::FancyOStream> VerboseObject::blackHole_ = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
 
-  RCP<Teuchos::FancyOStream> VerboseObject::mueluOutputStream_ = Teuchos::null;
+RCP<Teuchos::FancyOStream> VerboseObject::mueluOutputStream_ = Teuchos::null;
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_VerboseObject.hpp b/packages/muelu/src/MueCentral/MueLu_VerboseObject.hpp
index 1ccc15bdd3b4..7502ab18671e 100644
--- a/packages/muelu/src/MueCentral/MueLu_VerboseObject.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_VerboseObject.hpp
@@ -46,105 +46,103 @@
 #ifndef MUELU_VERBOSEOBJECT_HPP
 #define MUELU_VERBOSEOBJECT_HPP
 
-#include "Teuchos_FancyOStream.hpp"     // for FancyOStream
-#include "Teuchos_RCPDecl.hpp"          // for RCP
+#include "Teuchos_FancyOStream.hpp"  // for FancyOStream
+#include "Teuchos_RCPDecl.hpp"       // for RCP
 #include "Teuchos_VerboseObject.hpp"
 
 #include "MueLu_VerbosityLevel.hpp"
 
 namespace MueLu {
 
-  /*!
-    @class VerboseObject class.
-    @brief Verbose class for MueLu classes
+/*!
+  @class VerboseObject class.
+  @brief Verbose class for MueLu classes
 
-    This class is inspired by BelosOutputManager but:
-      - it uses Teuchos stream
-      - it allows to print message on proc0 or every processors
-      - it is a base class
+  This class is inspired by BelosOutputManager but:
+    - it uses Teuchos stream
+    - it allows to print message on proc0 or every processors
+    - it is a base class
 
-     @todo TODO
-     - Right now, you can't change the verbosity level globally.
-       We have to take into account Teuchos::EVerbosityLevel from the base class Teuchos::VerboseObject and somehow convert it to a MsgType.
-     - Allow users to define the global verbosity level using both MueLu::MsgType and Teuchos::EVerbosityLevel?
+   @todo TODO
+   - Right now, you can't change the verbosity level globally.
+     We have to take into account Teuchos::EVerbosityLevel from the base class Teuchos::VerboseObject and somehow convert it to a MsgType.
+   - Allow users to define the global verbosity level using both MueLu::MsgType and Teuchos::EVerbosityLevel?
 
-     @todo TODO
-     -# add a function to get the object 'local' verb level (GetVerbLevel does not do this)?
-     -# interface without fancyOSstream but std::stream
-     -# add 'const' and mutable
+   @todo TODO
+   -# add a function to get the object 'local' verb level (GetVerbLevel does not do this)?
+   -# interface without fancyOSstream but std::stream
+   -# add 'const' and mutable
 
-    @ingroup MueLuBaseClasses
+  @ingroup MueLuBaseClasses
 
-  */
-  class VerboseObject : public Teuchos::VerboseObject<VerboseObject> {
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-    VerboseObject();
+*/
+class VerboseObject : public Teuchos::VerboseObject<VerboseObject> {
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+  VerboseObject();
 
-    //! Destructor.
-    virtual ~VerboseObject();
-    //@}
+  //! Destructor.
+  virtual ~VerboseObject();
+  //@}
 
-    /*! @brief Get the verbosity level
+  /*! @brief Get the verbosity level
 
-       If a verbosity level has not been specified for this object (using SetVerbLevel), this method returns the default/global verbose level.
-    */
-    VerbLevel GetVerbLevel() const;
+     If a verbosity level has not been specified for this object (using SetVerbLevel), this method returns the default/global verbose level.
+  */
+  VerbLevel GetVerbLevel() const;
 
-    //! Set the verbosity level of this object
-    void SetVerbLevel(const VerbLevel verbLevel);
+  //! Set the verbosity level of this object
+  void SetVerbLevel(const VerbLevel verbLevel);
 
-    //! Get proc rank used for printing. <b>Do not use this information for any other purpose.</b>
-    int GetProcRankVerbose() const;
+  //! Get proc rank used for printing. <b>Do not use this information for any other purpose.</b>
+  int GetProcRankVerbose() const;
 
-    //! Set proc rank used for printing.
-    int SetProcRankVerbose(int procRank) const;
+  //! Set proc rank used for printing.
+  int SetProcRankVerbose(int procRank) const;
 
-    /*! @brief Find out whether we need to print out information for a specific message type.
+  /*! @brief Find out whether we need to print out information for a specific message type.
 
-        This method is used to determine whether computations are necessary for this message type.
-    */
-    bool IsPrint(MsgType type, int thisProcRankOnly = -1) const;
+      This method is used to determine whether computations are necessary for this message type.
+  */
+  bool IsPrint(MsgType type, int thisProcRankOnly = -1) const;
 
-    //! Get an output stream for outputting the input message type.
-    Teuchos::FancyOStream & GetOStream(MsgType type, int thisProcRankOnly = 0) const;
+  //! Get an output stream for outputting the input message type.
+  Teuchos::FancyOStream& GetOStream(MsgType type, int thisProcRankOnly = 0) const;
 
-    Teuchos::FancyOStream & GetBlackHole() const;
+  Teuchos::FancyOStream& GetBlackHole() const;
 
-    static void SetMueLuOStream(const Teuchos::RCP<Teuchos::FancyOStream> &mueluOStream);
+  static void SetMueLuOStream(const Teuchos::RCP<Teuchos::FancyOStream>& mueluOStream);
 
-    static void SetMueLuOFileStream(const std::string& filename);
+  static void SetMueLuOFileStream(const std::string& filename);
 
-    static Teuchos::RCP<Teuchos::FancyOStream> GetMueLuOStream();
+  static Teuchos::RCP<Teuchos::FancyOStream> GetMueLuOStream();
 
-    //! @name Public static member functions
-    //@{
+  //! @name Public static member functions
+  //@{
 
-    //! Set the default (global) verbosity level.
-    static void SetDefaultVerbLevel(const VerbLevel defaultVerbLevel);
+  //! Set the default (global) verbosity level.
+  static void SetDefaultVerbLevel(const VerbLevel defaultVerbLevel);
 
-    //! Get the default (global) verbosity level.
-    static VerbLevel GetDefaultVerbLevel();
+  //! Get the default (global) verbosity level.
+  static VerbLevel GetDefaultVerbLevel();
 
-    //@}
+  //@}
 
-  private:
-    //! Verbose level specific to 'this'
-    VerbLevel verbLevel_;
-    mutable
-      int procRank_;
-    int numProcs_;
+ private:
+  //! Verbose level specific to 'this'
+  VerbLevel verbLevel_;
+  mutable int procRank_;
+  int numProcs_;
 
-    static Teuchos::RCP<Teuchos::FancyOStream> mueluOutputStream_;
-    static Teuchos::RCP<Teuchos::FancyOStream> blackHole_;
+  static Teuchos::RCP<Teuchos::FancyOStream> mueluOutputStream_;
+  static Teuchos::RCP<Teuchos::FancyOStream> blackHole_;
 
-    //! Global verbose level. This verbose level is used when the verbose level of the object is not specified (verbLevel_ == NotSpecified)
-    static VerbLevel globalVerbLevel_;
-  }; // class VerboseObject
+  //! Global verbose level. This verbose level is used when the verbose level of the object is not specified (verbLevel_ == NotSpecified)
+  static VerbLevel globalVerbLevel_;
+};  // class VerboseObject
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_VERBOSECLASS_SHORT
-#endif // MUELU_VERBOSEOBJECT_HPP
+#endif  // MUELU_VERBOSEOBJECT_HPP
diff --git a/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.cpp b/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.cpp
index dc1e4aadb350..6ebb25eb3f45 100644
--- a/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.cpp
+++ b/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.cpp
@@ -51,85 +51,83 @@
 
 namespace MueLu {
 
-  VerbLevel toMueLuVerbLevel(const Teuchos::EVerbosityLevel verbLevel) {
-    switch(verbLevel)
-      {
-      case Teuchos::VERB_NONE:
-        return None;
-      case Teuchos::VERB_DEFAULT:
-        return Default;
-      case Teuchos::VERB_LOW:
-        return Low;
-      case Teuchos::VERB_MEDIUM:
-        return Medium;
-      case Teuchos::VERB_HIGH:
-        return High;
-      case Teuchos::VERB_EXTREME:
-        return Extreme;
-      default:
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Unknown enum value found.");
-      }
+VerbLevel toMueLuVerbLevel(const Teuchos::EVerbosityLevel verbLevel) {
+  switch (verbLevel) {
+    case Teuchos::VERB_NONE:
+      return None;
+    case Teuchos::VERB_DEFAULT:
+      return Default;
+    case Teuchos::VERB_LOW:
+      return Low;
+    case Teuchos::VERB_MEDIUM:
+      return Medium;
+    case Teuchos::VERB_HIGH:
+      return High;
+    case Teuchos::VERB_EXTREME:
+      return Extreme;
+    default:
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Unknown enum value found.");
   }
+}
 
-  std::string
-  lowerCase (const std::string& s)
-  {
-    typedef std::string::value_type char_t;
-    typedef std::ctype<char_t> facet_type;
-    const facet_type& facet = std::use_facet<facet_type> (std::locale ());
+std::string
+lowerCase(const std::string& s) {
+  typedef std::string::value_type char_t;
+  typedef std::ctype<char_t> facet_type;
+  const facet_type& facet = std::use_facet<facet_type>(std::locale());
 
-    const std::string::size_type len = s.size ();
-    std::string s_lc (s);
-    for (std::string::size_type k = 0; k < len; ++k) {
-      s_lc[k] = facet.tolower (s[k]);
-    }
-
-    return s_lc;
+  const std::string::size_type len = s.size();
+  std::string s_lc(s);
+  for (std::string::size_type k = 0; k < len; ++k) {
+    s_lc[k] = facet.tolower(s[k]);
   }
 
-  MsgType toVerbLevel(const std::string& verbLevelStr) {
-    std::map<std::string, MsgType> verbMap;
-    //for developers
-    verbMap["errors"]         = Errors;
-    verbMap["warnings0"]      = Warnings0;
-    verbMap["warnings00"]     = Warnings00;
-    verbMap["warnings1"]      = Warnings1;
-    verbMap["perfWarnings"]   = PerfWarnings;
-    verbMap["runtime0"]       = Runtime0;
-    verbMap["runtime1"]       = Runtime1;
-    verbMap["runtimeTimings"] = RuntimeTimings;
-    verbMap["noTimeReport"]   = NoTimeReport;
-    verbMap["parameters0"]    = Parameters0;
-    verbMap["parameters1"]    = Parameters1;
-    verbMap["statistics0"]    = Statistics0;
-    verbMap["statistics1"]    = Statistics1;
-    verbMap["timings0"]       = Timings0;
-    verbMap["timings1"]       = Timings1;
-    verbMap["timingsByLevel"] = TimingsByLevel;
-    verbMap["external"]       = External;
-    verbMap["developer"]      = Developer;
-    verbMap["debug"]          = Debug;
-    verbMap["test"]           = Test;
+  return s_lc;
+}
 
-    verbMap["warnings"]       = Warnings;
-    verbMap["runtime"]        = Runtime;
-    verbMap["parameters"]     = Parameters;
-    verbMap["statistics"]     = Statistics;
-    verbMap["timings"]        = Timings;
-    verbMap["test"]           = Test;
-    verbMap["interfacetest"]  = InterfaceTest;
-    //for users and developers
-    verbMap["none"]           = None;
-    verbMap["low"]            = Low;
-    verbMap["medium"]         = Medium;
-    verbMap["high"]           = High;
-    verbMap["extreme"]        = Extreme;
+MsgType toVerbLevel(const std::string& verbLevelStr) {
+  std::map<std::string, MsgType> verbMap;
+  // for developers
+  verbMap["errors"]         = Errors;
+  verbMap["warnings0"]      = Warnings0;
+  verbMap["warnings00"]     = Warnings00;
+  verbMap["warnings1"]      = Warnings1;
+  verbMap["perfWarnings"]   = PerfWarnings;
+  verbMap["runtime0"]       = Runtime0;
+  verbMap["runtime1"]       = Runtime1;
+  verbMap["runtimeTimings"] = RuntimeTimings;
+  verbMap["noTimeReport"]   = NoTimeReport;
+  verbMap["parameters0"]    = Parameters0;
+  verbMap["parameters1"]    = Parameters1;
+  verbMap["statistics0"]    = Statistics0;
+  verbMap["statistics1"]    = Statistics1;
+  verbMap["timings0"]       = Timings0;
+  verbMap["timings1"]       = Timings1;
+  verbMap["timingsByLevel"] = TimingsByLevel;
+  verbMap["external"]       = External;
+  verbMap["developer"]      = Developer;
+  verbMap["debug"]          = Debug;
+  verbMap["test"]           = Test;
 
-    std::string lcVerb = lowerCase(verbLevelStr);
-    if (verbMap.find(lcVerb) != verbMap.end())
-      return verbMap[lcVerb];
-    else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::ParameterListInterpreter():: invalid verbosity level: " << verbLevelStr);
-  }
+  verbMap["warnings"]      = Warnings;
+  verbMap["runtime"]       = Runtime;
+  verbMap["parameters"]    = Parameters;
+  verbMap["statistics"]    = Statistics;
+  verbMap["timings"]       = Timings;
+  verbMap["test"]          = Test;
+  verbMap["interfacetest"] = InterfaceTest;
+  // for users and developers
+  verbMap["none"]    = None;
+  verbMap["low"]     = Low;
+  verbMap["medium"]  = Medium;
+  verbMap["high"]    = High;
+  verbMap["extreme"] = Extreme;
+
+  std::string lcVerb = lowerCase(verbLevelStr);
+  if (verbMap.find(lcVerb) != verbMap.end())
+    return verbMap[lcVerb];
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::ParameterListInterpreter():: invalid verbosity level: " << verbLevelStr);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.hpp b/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.hpp
index 968f7330562c..5fcfe76670c1 100644
--- a/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.hpp
+++ b/packages/muelu/src/MueCentral/MueLu_VerbosityLevel.hpp
@@ -50,74 +50,73 @@
 
 namespace MueLu {
 
-  enum MsgType
-    {
-      Errors          = 0x00000001, //!< Errors
-
-      Warnings0       = 0x00000010, //!< Important warning messages (one line)
-      Warnings00      = 0x00000020, //!< Important warning messages (more verbose)
-      Warnings1       = 0x00000040, //!< Additional warnings
-      PerfWarnings    = 0x00000080, //!< Performance warnings
-
-      Runtime0        = 0x00000100, //!< One-liner description of what is happening
-      Runtime1        = 0x00000200, //!< Description of what is happening (more verbose)
-      RuntimeTimings  = 0x00000400, //!< Timers that are enabled (using Timings0/Timings1) will be printed during the execution
-      NoTimeReport    = 0x00000800, //!< By default, enabled timers appears in the teuchos time monitor summary. Use this option if you do not want to record timing information.
-
-      Parameters0     = 0x00001000, //!< Print class parameters
-      Parameters1     = 0x00002000, //!< Print class parameters (more parameters, more verbose)
-
-      Statistics0     = 0x00010000, //!< Print statistics that do not involve significant additional computation
-      Statistics1     = 0x00020000, //!< Print more statistics
-      Statistics2     = 0x00040000, //!< Print even more statistics
-
-      Timings0        = 0x00100000, //!< High level timing information (use Teuchos::TimeMonitor::summarize() to print)
-      Timings1        = 0x00200000, //!< Detailed timing information   (use Teuchos::TimeMonitor::summarize() to print)
-      TimingsByLevel  = 0x00400000, //!< Record timing information level by level. Must be used in combinaison with Timings0/Timings1
-
-      External        = 0x01000000, //!< Print external lib objects
-      Debug           = 0x02000000, //!< Print additional debugging information
-
-      Developer       = 0x04000000, //!< Print information primarily of interest to developers
-
-      // Print skeleton for the run
-      Test0           = 0x10000000, //!< Print factory calls
-      Test1           = 0x20000000, //!< Print used parameters
-
-      // Predefined combinations of MsgType
-      // Can be used in user code or examples. Do not used as input parameters of IsPrint() or GetOStream().
-      Warnings        = Warnings0 | Warnings00 | Warnings1 | PerfWarnings, //!< Print all warning messages
-      Runtime         = Runtime0 | Runtime1,                               //!< Print description of what is going on
-      Parameters      = Parameters0 | Parameters1,                         //!< Print parameters
-      Statistics      = Statistics0 | Statistics1 | Statistics2,           //!< Print all statistics
-      Timings         = Timings0 | Timings1 | TimingsByLevel,              //!< Print all timing information
-      Test            = Test0 | Test1 | Runtime0,
-      InterfaceTest   = Test0 | Runtime0,
-
-      //
-      None    = 0,
-      Low     = Errors | Warnings0 | Statistics0,
-      Medium  = Errors | Warnings0 | Runtime0 | Parameters0 | Statistics0 | Statistics1 | Timings0,
-      High    = Errors | Warnings  | Runtime  | Parameters  | Statistics0 | Statistics1  | Timings,
+enum MsgType {
+  Errors = 0x00000001,  //!< Errors
+
+  Warnings0    = 0x00000010,  //!< Important warning messages (one line)
+  Warnings00   = 0x00000020,  //!< Important warning messages (more verbose)
+  Warnings1    = 0x00000040,  //!< Additional warnings
+  PerfWarnings = 0x00000080,  //!< Performance warnings
+
+  Runtime0       = 0x00000100,  //!< One-liner description of what is happening
+  Runtime1       = 0x00000200,  //!< Description of what is happening (more verbose)
+  RuntimeTimings = 0x00000400,  //!< Timers that are enabled (using Timings0/Timings1) will be printed during the execution
+  NoTimeReport   = 0x00000800,  //!< By default, enabled timers appears in the teuchos time monitor summary. Use this option if you do not want to record timing information.
+
+  Parameters0 = 0x00001000,  //!< Print class parameters
+  Parameters1 = 0x00002000,  //!< Print class parameters (more parameters, more verbose)
+
+  Statistics0 = 0x00010000,  //!< Print statistics that do not involve significant additional computation
+  Statistics1 = 0x00020000,  //!< Print more statistics
+  Statistics2 = 0x00040000,  //!< Print even more statistics
+
+  Timings0       = 0x00100000,  //!< High level timing information (use Teuchos::TimeMonitor::summarize() to print)
+  Timings1       = 0x00200000,  //!< Detailed timing information   (use Teuchos::TimeMonitor::summarize() to print)
+  TimingsByLevel = 0x00400000,  //!< Record timing information level by level. Must be used in combinaison with Timings0/Timings1
+
+  External = 0x01000000,  //!< Print external lib objects
+  Debug    = 0x02000000,  //!< Print additional debugging information
+
+  Developer = 0x04000000,  //!< Print information primarily of interest to developers
+
+  // Print skeleton for the run
+  Test0 = 0x10000000,  //!< Print factory calls
+  Test1 = 0x20000000,  //!< Print used parameters
+
+  // Predefined combinations of MsgType
+  // Can be used in user code or examples. Do not used as input parameters of IsPrint() or GetOStream().
+  Warnings      = Warnings0 | Warnings00 | Warnings1 | PerfWarnings,  //!< Print all warning messages
+  Runtime       = Runtime0 | Runtime1,                                //!< Print description of what is going on
+  Parameters    = Parameters0 | Parameters1,                          //!< Print parameters
+  Statistics    = Statistics0 | Statistics1 | Statistics2,            //!< Print all statistics
+  Timings       = Timings0 | Timings1 | TimingsByLevel,               //!< Print all timing information
+  Test          = Test0 | Test1 | Runtime0,
+  InterfaceTest = Test0 | Runtime0,
+
+  //
+  None   = 0,
+  Low    = Errors | Warnings0 | Statistics0,
+  Medium = Errors | Warnings0 | Runtime0 | Parameters0 | Statistics0 | Statistics1 | Timings0,
+  High   = Errors | Warnings | Runtime | Parameters | Statistics0 | Statistics1 | Timings,
 #ifdef HAVE_MUELU_DEBUG
-      Extreme = Errors | Warnings  | Runtime  | Parameters  | Statistics  | Timings | External | Developer | Debug,
+  Extreme = Errors | Warnings | Runtime | Parameters | Statistics | Timings | External | Developer | Debug,
 #else
-      Extreme = Errors | Warnings  | Runtime  | Parameters  | Statistics  | Timings | External | Developer,
+  Extreme = Errors | Warnings | Runtime | Parameters | Statistics | Timings | External | Developer,
 #endif
-      Default = High, // This is the default of print() methods. For VerboseObject, another default is set by VerboseObject::globalVerbLevel_ // TODO: move it to the VerboseObject class
+  Default = High,  // This is the default of print() methods. For VerboseObject, another default is set by VerboseObject::globalVerbLevel_ // TODO: move it to the VerboseObject class
 
-      NotSpecified = -1
-    };
+  NotSpecified = -1
+};
 
-  //!
-  typedef int VerbLevel;
+//!
+typedef int VerbLevel;
 
-  //! Translate Teuchos verbosity level to MueLu verbosity level
-  VerbLevel toMueLuVerbLevel(const Teuchos::EVerbosityLevel verbLevel);
+//! Translate Teuchos verbosity level to MueLu verbosity level
+VerbLevel toMueLuVerbLevel(const Teuchos::EVerbosityLevel verbLevel);
 
-  //!
-  MsgType toVerbLevel(const std::string& verbLevelStr);
+//!
+MsgType toVerbLevel(const std::string& verbLevelStr);
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif
diff --git a/packages/muelu/src/Operators/MueLu_CreateXpetraPreconditioner.hpp b/packages/muelu/src/Operators/MueLu_CreateXpetraPreconditioner.hpp
index 83c78993228f..36b4bfe3a023 100644
--- a/packages/muelu/src/Operators/MueLu_CreateXpetraPreconditioner.hpp
+++ b/packages/muelu/src/Operators/MueLu_CreateXpetraPreconditioner.hpp
@@ -67,187 +67,185 @@
 
 namespace MueLu {
 
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
-    @ingroup MueLuAdapters
-    Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] inParamList Parameter list
-  */
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > op,
-                             const Teuchos::ParameterList& inParamList) {
-    using SC = Scalar;
-    using LO = LocalOrdinal;
-    using GO = GlobalOrdinal;
-    using NO = Node;
-
-    using HierarchyManager = MueLu::HierarchyManager<SC, LO, GO, NO>;
-    using HierarchyUtils = MueLu::HierarchyUtils<SC, LO, GO, NO>;
-    using Hierarchy = MueLu::Hierarchy<SC, LO, GO, NO>;
-    using ParameterListInterpreter = ParameterListInterpreter<SC, LO, GO, NO>;
-
-    bool hasParamList = inParamList.numParams();
-
-    RCP<HierarchyManager> mueLuFactory;
-
-    // Rip off non-serializable data before validation
-    Teuchos::ParameterList nonSerialList,paramList;
-    MueLu::ExtractNonSerializableData(inParamList, paramList, nonSerialList);
-
-    std::string label;
-    if (hasParamList && paramList.isParameter("hierarchy label")) {
-      label = paramList.get<std::string>("hierarchy label");
-      paramList.remove("hierarchy label");
-    } else
-      label = op->getObjectLabel();
-
-    std::string timerName;
-    if (label != "")
-      timerName = "MueLu setup time (" + label + ")";
-    else
-      timerName = "MueLu setup time";
-    RCP<Teuchos::Time> tm = Teuchos::TimeMonitor::getNewTimer(timerName);
-    tm->start();
-
-    std::string syntaxStr = "parameterlist: syntax";
-    if (hasParamList && paramList.isParameter(syntaxStr) && paramList.get<std::string>(syntaxStr) == "ml") {
-      paramList.remove(syntaxStr);
-      std::string paramXML = MueLu::ML2MueLuParameterTranslator::translate(paramList, "");
-      paramList = *Teuchos::getParametersFromXmlString(paramXML);
-    }
-    mueLuFactory = rcp(new ParameterListInterpreter(paramList,op->getDomainMap()->getComm()));
-
-    // Create Hierarchy
-    RCP<Hierarchy> H = mueLuFactory->CreateHierarchy(label);
-    H->setlib(op->getDomainMap()->lib());
-
-    // Set fine level operator
-    H->GetLevel(0)->Set("A", op);
-    H->SetProcRankVerbose(op->getDomainMap()->getComm()->getRank());
-
-    // Stick the non-serializible data on the hierarchy.
-    HierarchyUtils::AddNonSerializableDataToHierarchy(*mueLuFactory,*H, nonSerialList);
-
-    mueLuFactory->SetupHierarchy(*H);
-
-    tm->stop();
-    tm->incrementNumCalls();
-
-    if (H->GetVerbLevel() & Statistics0) {
-      const bool alwaysWriteLocal = true;
-      const bool writeGlobalStats = true;
-      const bool writeZeroTimers  = false;
-      const bool ignoreZeroTimers = true;
-      const std::string filter    = timerName;
-      Teuchos::TimeMonitor::summarize(op->getRowMap()->getComm().ptr(), H->GetOStream(Statistics0), alwaysWriteLocal, writeGlobalStats,
-                                      writeZeroTimers, Teuchos::Union, filter, ignoreZeroTimers);
-    }
-
-    tm->reset();
-
-    return H;
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
+  @ingroup MueLuAdapters
+  Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] inParamList Parameter list
+*/
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> op,
+                           const Teuchos::ParameterList& inParamList) {
+  using SC = Scalar;
+  using LO = LocalOrdinal;
+  using GO = GlobalOrdinal;
+  using NO = Node;
+
+  using HierarchyManager         = MueLu::HierarchyManager<SC, LO, GO, NO>;
+  using HierarchyUtils           = MueLu::HierarchyUtils<SC, LO, GO, NO>;
+  using Hierarchy                = MueLu::Hierarchy<SC, LO, GO, NO>;
+  using ParameterListInterpreter = ParameterListInterpreter<SC, LO, GO, NO>;
+
+  bool hasParamList = inParamList.numParams();
+
+  RCP<HierarchyManager> mueLuFactory;
+
+  // Rip off non-serializable data before validation
+  Teuchos::ParameterList nonSerialList, paramList;
+  MueLu::ExtractNonSerializableData(inParamList, paramList, nonSerialList);
+
+  std::string label;
+  if (hasParamList && paramList.isParameter("hierarchy label")) {
+    label = paramList.get<std::string>("hierarchy label");
+    paramList.remove("hierarchy label");
+  } else
+    label = op->getObjectLabel();
+
+  std::string timerName;
+  if (label != "")
+    timerName = "MueLu setup time (" + label + ")";
+  else
+    timerName = "MueLu setup time";
+  RCP<Teuchos::Time> tm = Teuchos::TimeMonitor::getNewTimer(timerName);
+  tm->start();
+
+  std::string syntaxStr = "parameterlist: syntax";
+  if (hasParamList && paramList.isParameter(syntaxStr) && paramList.get<std::string>(syntaxStr) == "ml") {
+    paramList.remove(syntaxStr);
+    std::string paramXML = MueLu::ML2MueLuParameterTranslator::translate(paramList, "");
+    paramList            = *Teuchos::getParametersFromXmlString(paramXML);
   }
+  mueLuFactory = rcp(new ParameterListInterpreter(paramList, op->getDomainMap()->getComm()));
 
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
-    @ingroup MueLuAdapters
-    Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-    @param[in] xmlFileName std::string
-  */
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > op,
-                             const std::string& xmlFileName) {
-    Teuchos::ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *op->getDomainMap()->getComm());
-    return CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op, paramList);
-  }
+  // Create Hierarchy
+  RCP<Hierarchy> H = mueLuFactory->CreateHierarchy(label);
+  H->setlib(op->getDomainMap()->lib());
+
+  // Set fine level operator
+  H->GetLevel(0)->Set("A", op);
+  H->SetProcRankVerbose(op->getDomainMap()->getComm()->getRank());
+
+  // Stick the non-serializible data on the hierarchy.
+  HierarchyUtils::AddNonSerializableDataToHierarchy(*mueLuFactory, *H, nonSerialList);
+
+  mueLuFactory->SetupHierarchy(*H);
 
-  /*!
-    @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
-    @ingroup MueLuAdapters
-    Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
-    @param[in] inA Matrix
-  */
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > op) {
-    Teuchos::ParameterList paramList;
-    return CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op, paramList);
+  tm->stop();
+  tm->incrementNumCalls();
+
+  if (H->GetVerbLevel() & Statistics0) {
+    const bool alwaysWriteLocal = true;
+    const bool writeGlobalStats = true;
+    const bool writeZeroTimers  = false;
+    const bool ignoreZeroTimers = true;
+    const std::string filter    = timerName;
+    Teuchos::TimeMonitor::summarize(op->getRowMap()->getComm().ptr(), H->GetOStream(Statistics0), alwaysWriteLocal, writeGlobalStats,
+                                    writeZeroTimers, Teuchos::Union, filter, ignoreZeroTimers);
   }
 
-  /*!
-    @brief Helper function to reuse an existing MueLu preconditioner.
-    @ingroup MueLuAdapters
-
-    @param[in] inA Matrix
-    @param[in] Op  Existing MueLu preconditioner.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReuseXpetraPreconditioner(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-                                 Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& H) {
-    std::string label = H->GetLevel(0)->getObjectLabel();
-
-    std::string timerName;
-    if (label != "")
-      timerName = "MueLu setup time (" + label + ")";
-    else
-      timerName = "MueLu setup time";
-    RCP<Teuchos::Time> tm = Teuchos::TimeMonitor::getNewTimer(timerName);
-    tm->start();
-
-    typedef Scalar          SC;
-    typedef LocalOrdinal    LO;
-    typedef GlobalOrdinal   GO;
-    typedef Node            NO;
-
-    typedef Xpetra::Matrix<SC,LO,GO,NO>     Matrix;
-    typedef Xpetra::Operator<SC,LO,GO,NO>   Operator;
-
-    TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), Exceptions::RuntimeError,
-                               "MueLu::ReuseXpetraPreconditioner: Hierarchy has no levels in it");
-    TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError,
-                               "MueLu::ReuseXpetraPreconditioner: Hierarchy has no fine level operator");
-    RCP<Level> level0 = H->GetLevel(0);
-
-    RCP<Operator> O0 = level0->Get<RCP<Operator> >("A");
-    RCP<Matrix>   A0 = Teuchos::rcp_dynamic_cast<Matrix>(O0);
-
-    if (!A0.is_null()) {
-      // If a user provided a "number of equations" argument in a parameter list
-      // during the initial setup, we must honor that settings and reuse it for
-      // all consequent setups.
-      A->SetFixedBlockSize(A0->GetFixedBlockSize());
-    }
-    level0->Set("A", A);
-
-    H->SetupRe();
-
-    tm->stop();
-    tm->incrementNumCalls();
-
-    if (H->GetVerbLevel() & Statistics0) {
-      const bool alwaysWriteLocal = true;
-      const bool writeGlobalStats = true;
-      const bool writeZeroTimers  = false;
-      const bool ignoreZeroTimers = true;
-      const std::string filter    = timerName;
-      Teuchos::TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), H->GetOStream(Statistics0), alwaysWriteLocal, writeGlobalStats,
-                                      writeZeroTimers, Teuchos::Union, filter, ignoreZeroTimers);
-    }
-
-    tm->reset();
+  tm->reset();
+
+  return H;
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
+  @ingroup MueLuAdapters
+  Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+  @param[in] xmlFileName std::string
+*/
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> op,
+                           const std::string& xmlFileName) {
+  Teuchos::ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *op->getDomainMap()->getComm());
+  return CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op, paramList);
+}
+
+/*!
+  @brief Helper function to create a MueLu preconditioner that can be used by Xpetra.
+  @ingroup MueLuAdapters
+  Given an Xpetra::Matrix, this function returns a constructed MueLu preconditioner.
+  @param[in] inA Matrix
+*/
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+CreateXpetraPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> op) {
+  Teuchos::ParameterList paramList;
+  return CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(op, paramList);
+}
+
+/*!
+  @brief Helper function to reuse an existing MueLu preconditioner.
+  @ingroup MueLuAdapters
+
+  @param[in] inA Matrix
+  @param[in] Op  Existing MueLu preconditioner.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReuseXpetraPreconditioner(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                               Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& H) {
+  std::string label = H->GetLevel(0)->getObjectLabel();
+
+  std::string timerName;
+  if (label != "")
+    timerName = "MueLu setup time (" + label + ")";
+  else
+    timerName = "MueLu setup time";
+  RCP<Teuchos::Time> tm = Teuchos::TimeMonitor::getNewTimer(timerName);
+  tm->start();
+
+  typedef Scalar SC;
+  typedef LocalOrdinal LO;
+  typedef GlobalOrdinal GO;
+  typedef Node NO;
+
+  typedef Xpetra::Matrix<SC, LO, GO, NO> Matrix;
+  typedef Xpetra::Operator<SC, LO, GO, NO> Operator;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(!H->GetNumLevels(), Exceptions::RuntimeError,
+                             "MueLu::ReuseXpetraPreconditioner: Hierarchy has no levels in it");
+  TEUCHOS_TEST_FOR_EXCEPTION(!H->GetLevel(0)->IsAvailable("A"), Exceptions::RuntimeError,
+                             "MueLu::ReuseXpetraPreconditioner: Hierarchy has no fine level operator");
+  RCP<Level> level0 = H->GetLevel(0);
+
+  RCP<Operator> O0 = level0->Get<RCP<Operator>>("A");
+  RCP<Matrix> A0   = Teuchos::rcp_dynamic_cast<Matrix>(O0);
+
+  if (!A0.is_null()) {
+    // If a user provided a "number of equations" argument in a parameter list
+    // during the initial setup, we must honor that settings and reuse it for
+    // all consequent setups.
+    A->SetFixedBlockSize(A0->GetFixedBlockSize());
   }
+  level0->Set("A", A);
+
+  H->SetupRe();
 
+  tm->stop();
+  tm->incrementNumCalls();
+
+  if (H->GetVerbLevel() & Statistics0) {
+    const bool alwaysWriteLocal = true;
+    const bool writeGlobalStats = true;
+    const bool writeZeroTimers  = false;
+    const bool ignoreZeroTimers = true;
+    const std::string filter    = timerName;
+    Teuchos::TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), H->GetOStream(Statistics0), alwaysWriteLocal, writeGlobalStats,
+                                    writeZeroTimers, Teuchos::Union, filter, ignoreZeroTimers);
+  }
 
+  tm->reset();
+}
 
-} //namespace
+}  // namespace MueLu
 
 #define XPETRA_CREATEXPETRAPRECONDITIONER_SHORT
 
diff --git a/packages/muelu/src/Operators/MueLu_Maxwell1_decl.hpp b/packages/muelu/src/Operators/MueLu_Maxwell1_decl.hpp
index 20cae3f9ae75..5bc5f70e888f 100644
--- a/packages/muelu/src/Operators/MueLu_Maxwell1_decl.hpp
+++ b/packages/muelu/src/Operators/MueLu_Maxwell1_decl.hpp
@@ -67,271 +67,258 @@
 
 namespace MueLu {
 
-  /*!
-    @brief Preconditioner (wrapped as a Xpetra::Operator) for Maxwell's equations in curl-curl form.
-
- 
-    @ingroup MueLuAdapters
-  */
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class Maxwell1 : public VerboseObject, public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @brief Preconditioner (wrapped as a Xpetra::Operator) for Maxwell's equations in curl-curl form.
 
+
+  @ingroup MueLuAdapters
+*/
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class Maxwell1 : public VerboseObject, public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_MAXWELL1_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
-    typedef typename Xpetra::MultiVector<coordinateType,LO,GO,NO> RealValuedMultiVector;
-
-    //! Constructor
-    Maxwell1() :
-      Hierarchy11_(Teuchos::null),
-      Hierarchy22_(Teuchos::null),
-      HierarchyGmhd_(Teuchos::null),
-      mode_(MODE_STANDARD)
-    {
-    }
-
-    //! Constructor with Hierarchies
-    Maxwell1(Teuchos::RCP<Hierarchy> H11, Teuchos::RCP<Hierarchy> H22) :
-      Hierarchy11_(H11),
-      Hierarchy22_(H22),
-      HierarchyGmhd_(Teuchos::null),
-      mode_(MODE_STANDARD)
-    {
-    }
-
-    /** Constructor with Jacobian
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    Maxwell1(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace,
-               const Teuchos::RCP<RealValuedMultiVector> & Coords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true):
-               mode_(MODE_STANDARD)
-    {
-      RCP<Matrix> Kn_Matrix;
-      initialize(D0_Matrix,Kn_Matrix,Nullspace,Coords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Constructor with Jacobian and nodal matrix
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Kn_Matrix Nodal Laplacian
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    Maxwell1(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & Kn_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace,
-               const Teuchos::RCP<RealValuedMultiVector> & Coords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true):
-               mode_(MODE_STANDARD)
-    {
-      initialize(D0_Matrix,Kn_Matrix,Nullspace,Coords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Gmhd GMHD Constructor with Jacobian and nodal matrix AND Gmhd matrix
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Kn_Matrix Nodal Laplacian
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] GmhdA_Matrix Gmhd matrix including generalized Ohms law equations
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    Maxwell1(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & Kn_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace,
-               const Teuchos::RCP<RealValuedMultiVector> & Coords,
-               Teuchos::ParameterList& List, const Teuchos::RCP<Matrix> & GmhdA_Matrix,
-             bool ComputePrec = true):
-            mode_(MODE_GMHD_STANDARD)
-    {
-      initialize(D0_Matrix,Kn_Matrix,Nullspace,Coords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-      GmhdA_Matrix_ = GmhdA_Matrix;
-      HierarchyGmhd_ = rcp(new Hierarchy("HierarchyGmhd"));
-      GMHDSetupHierarchy(List);
-    }
-   
-    /** Constructor with parameter list
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    Maxwell1(const Teuchos::RCP<Matrix> & SM_Matrix,
-             Teuchos::ParameterList& List,
-             bool ComputePrec = true):
-             mode_(MODE_STANDARD)
-    {
-
-      RCP<MultiVector> Nullspace = List.get<RCP<MultiVector> >("Nullspace", Teuchos::null);
-      RCP<RealValuedMultiVector> Coords = List.get<RCP<RealValuedMultiVector> >("Coordinates", Teuchos::null);
-      RCP<Matrix> D0_Matrix = List.get<RCP<Matrix> >("D0");
-      RCP<Matrix> Kn_Matrix;
-      if (List.isType<RCP<Matrix> >("Kn"))
-        Kn_Matrix = List.get<RCP<Matrix> >("Kn");
-
-      initialize(D0_Matrix,Kn_Matrix,Nullspace,Coords,List);
-
-      if (SM_Matrix != Teuchos::null)
-        resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    //! Destructor.
-    virtual ~Maxwell1() {}
-
-    //! Returns the Xpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Map> getDomainMap() const;
-
-    //! Returns the Xpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Map> getRangeMap() const;
-
-    //! Returns Jacobian matrix SM
-    const Teuchos::RCP<Matrix> & getJacobian() const {
-      return SM_Matrix_;
-    }
-
-    //! Set parameters
-    void setParameters(Teuchos::ParameterList& list);
-
-    //! Setup the preconditioner
-    void compute(bool reuse=false);
-
-    //! Reset system matrix
-    void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec=true);
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
-    //! \param[out] Y - MultiVector of dimension NumVectors containing result.
-    void apply (const MultiVector& X, MultiVector& Y,
-                Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
-
-    void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const MultiVector & X,
-                  const MultiVector & B,
-                  MultiVector & R) const {
-      using STS = Teuchos::ScalarTraits<Scalar>;
-      R.update(STS::one(),B,STS::zero());
-      this->apply (X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
-    }
-    
-
-  private:
-    //! Generates the Kn matrix
-    Teuchos::RCP<Matrix> generate_kn() const;
-
-
-    //! Sets up hiearchy for GMHD matrices that include generalized Ohms law equations
-    void GMHDSetupHierarchy(Teuchos::ParameterList& List) const;
-               
-    /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
-     *
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Kn_Matrix Kn nodal matrix
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     */
-    void initialize(const Teuchos::RCP<Matrix> & D0_Matrix,
-                    const Teuchos::RCP<Matrix> & Kn_Matrix,
-                    const Teuchos::RCP<MultiVector> & Nullspace,
-                    const Teuchos::RCP<RealValuedMultiVector> & Coords,
-                    Teuchos::ParameterList& List);
-
-
-    //! apply RefMaxwell additive 2x2 style cycle
-    void applyInverseRefMaxwellAdditive(const MultiVector& RHS, MultiVector& X) const;
-
-    //! apply standard Maxwell1 cycle
-    void applyInverseStandard(const MultiVector& RHS, MultiVector& X) const;
-
-    //! allocate multivectors for solve
-    void allocateMemory(int numVectors) const;
-
-    //! dump out matrix
-    void dump(const Matrix& A, std::string name) const;
-
-    //! dump out multivector
-    void dump(const MultiVector& X, std::string name) const;
-
-    //! dump out real-valued multivector
-    void dumpCoords(const RealValuedMultiVector& X, std::string name) const;
-
-    //! dump out boolean ArrayView
-    void dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const;
-
-    //! dump out boolean Kokkos::View
-    void dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const;
-
-    //! get a (synced) timer
-    Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm=Teuchos::null) const;
-
-    //! ParameterLists
-    mutable Teuchos::ParameterList parameterList_, precList11_, precList22_;
-
-    //! Two hierarchies: one for the (1,1)-block, another for the (2,2)-block
-    Teuchos::RCP<Hierarchy> Hierarchy11_, Hierarchy22_, HierarchyGmhd_;
-
-    //! Various matrices
-    Teuchos::RCP<Matrix> SM_Matrix_, D0_Matrix_, Kn_Matrix_, GmhdA_Matrix_;
-
-    //! Vectors for BCs
-    Kokkos::View<bool*, typename Node::device_type> BCrowsKokkos_, BCcolsKokkos_, BCdomainKokkos_;
-    int BCedges_, BCnodes_;
-    Teuchos::ArrayRCP<bool> BCrows_, BCcols_, BCdomain_;
-    //! Nullspace
-    Teuchos::RCP<MultiVector> Nullspace_;
-    //! Coordinates
-    Teuchos::RCP<RealValuedMultiVector> Coords_;
-    //! Some options
-    bool useKokkos_,allEdgesBoundary_,allNodesBoundary_,dump_matrices_,enable_reuse_,syncTimers_;
-    bool applyBCsTo22_;
-    
-    //! Execution modes
-    typedef enum {MODE_STANDARD=0, MODE_REFMAXWELL, MODE_EDGE_ONLY, MODE_GMHD_STANDARD} mode_type;
-    mode_type mode_;
-
-    //! Temporary memory (cached vectors for RefMaxwell-style)
-    RCP<Matrix> P11_;
-    mutable Teuchos::RCP<MultiVector> residualFine_, residual11c_, residual22_, update11c_,update22_;
-
-  };
-
-
-
-
-} // namespace
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
+  typedef typename Xpetra::MultiVector<coordinateType, LO, GO, NO> RealValuedMultiVector;
+
+  //! Constructor
+  Maxwell1()
+    : Hierarchy11_(Teuchos::null)
+    , Hierarchy22_(Teuchos::null)
+    , HierarchyGmhd_(Teuchos::null)
+    , mode_(MODE_STANDARD) {
+  }
+
+  //! Constructor with Hierarchies
+  Maxwell1(Teuchos::RCP<Hierarchy> H11, Teuchos::RCP<Hierarchy> H22)
+    : Hierarchy11_(H11)
+    , Hierarchy22_(H22)
+    , HierarchyGmhd_(Teuchos::null)
+    , mode_(MODE_STANDARD) {
+  }
+
+  /** Constructor with Jacobian
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  Maxwell1(const Teuchos::RCP<Matrix>& SM_Matrix,
+           const Teuchos::RCP<Matrix>& D0_Matrix,
+           const Teuchos::RCP<MultiVector>& Nullspace,
+           const Teuchos::RCP<RealValuedMultiVector>& Coords,
+           Teuchos::ParameterList& List,
+           bool ComputePrec = true)
+    : mode_(MODE_STANDARD) {
+    RCP<Matrix> Kn_Matrix;
+    initialize(D0_Matrix, Kn_Matrix, Nullspace, Coords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Constructor with Jacobian and nodal matrix
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Kn_Matrix Nodal Laplacian
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  Maxwell1(const Teuchos::RCP<Matrix>& SM_Matrix,
+           const Teuchos::RCP<Matrix>& D0_Matrix,
+           const Teuchos::RCP<Matrix>& Kn_Matrix,
+           const Teuchos::RCP<MultiVector>& Nullspace,
+           const Teuchos::RCP<RealValuedMultiVector>& Coords,
+           Teuchos::ParameterList& List,
+           bool ComputePrec = true)
+    : mode_(MODE_STANDARD) {
+    initialize(D0_Matrix, Kn_Matrix, Nullspace, Coords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Gmhd GMHD Constructor with Jacobian and nodal matrix AND Gmhd matrix
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Kn_Matrix Nodal Laplacian
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] GmhdA_Matrix Gmhd matrix including generalized Ohms law equations
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  Maxwell1(const Teuchos::RCP<Matrix>& SM_Matrix,
+           const Teuchos::RCP<Matrix>& D0_Matrix,
+           const Teuchos::RCP<Matrix>& Kn_Matrix,
+           const Teuchos::RCP<MultiVector>& Nullspace,
+           const Teuchos::RCP<RealValuedMultiVector>& Coords,
+           Teuchos::ParameterList& List, const Teuchos::RCP<Matrix>& GmhdA_Matrix,
+           bool ComputePrec = true)
+    : mode_(MODE_GMHD_STANDARD) {
+    initialize(D0_Matrix, Kn_Matrix, Nullspace, Coords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+    GmhdA_Matrix_  = GmhdA_Matrix;
+    HierarchyGmhd_ = rcp(new Hierarchy("HierarchyGmhd"));
+    GMHDSetupHierarchy(List);
+  }
+
+  /** Constructor with parameter list
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  Maxwell1(const Teuchos::RCP<Matrix>& SM_Matrix,
+           Teuchos::ParameterList& List,
+           bool ComputePrec = true)
+    : mode_(MODE_STANDARD) {
+    RCP<MultiVector> Nullspace        = List.get<RCP<MultiVector> >("Nullspace", Teuchos::null);
+    RCP<RealValuedMultiVector> Coords = List.get<RCP<RealValuedMultiVector> >("Coordinates", Teuchos::null);
+    RCP<Matrix> D0_Matrix             = List.get<RCP<Matrix> >("D0");
+    RCP<Matrix> Kn_Matrix;
+    if (List.isType<RCP<Matrix> >("Kn"))
+      Kn_Matrix = List.get<RCP<Matrix> >("Kn");
+
+    initialize(D0_Matrix, Kn_Matrix, Nullspace, Coords, List);
+
+    if (SM_Matrix != Teuchos::null)
+      resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  //! Destructor.
+  virtual ~Maxwell1() {}
+
+  //! Returns the Xpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Map> getDomainMap() const;
+
+  //! Returns the Xpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Map> getRangeMap() const;
+
+  //! Returns Jacobian matrix SM
+  const Teuchos::RCP<Matrix>& getJacobian() const {
+    return SM_Matrix_;
+  }
+
+  //! Set parameters
+  void setParameters(Teuchos::ParameterList& list);
+
+  //! Setup the preconditioner
+  void compute(bool reuse = false);
+
+  //! Reset system matrix
+  void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec = true);
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
+  //! \param[out] Y - MultiVector of dimension NumVectors containing result.
+  void apply(const MultiVector& X, MultiVector& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const MultiVector& X,
+                const MultiVector& B,
+                MultiVector& R) const {
+    using STS = Teuchos::ScalarTraits<Scalar>;
+    R.update(STS::one(), B, STS::zero());
+    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+  }
+
+ private:
+  //! Generates the Kn matrix
+  Teuchos::RCP<Matrix> generate_kn() const;
+
+  //! Sets up hiearchy for GMHD matrices that include generalized Ohms law equations
+  void GMHDSetupHierarchy(Teuchos::ParameterList& List) const;
+
+  /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
+   *
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Kn_Matrix Kn nodal matrix
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   */
+  void initialize(const Teuchos::RCP<Matrix>& D0_Matrix,
+                  const Teuchos::RCP<Matrix>& Kn_Matrix,
+                  const Teuchos::RCP<MultiVector>& Nullspace,
+                  const Teuchos::RCP<RealValuedMultiVector>& Coords,
+                  Teuchos::ParameterList& List);
+
+  //! apply RefMaxwell additive 2x2 style cycle
+  void applyInverseRefMaxwellAdditive(const MultiVector& RHS, MultiVector& X) const;
+
+  //! apply standard Maxwell1 cycle
+  void applyInverseStandard(const MultiVector& RHS, MultiVector& X) const;
+
+  //! allocate multivectors for solve
+  void allocateMemory(int numVectors) const;
+
+  //! dump out matrix
+  void dump(const Matrix& A, std::string name) const;
+
+  //! dump out multivector
+  void dump(const MultiVector& X, std::string name) const;
+
+  //! dump out real-valued multivector
+  void dumpCoords(const RealValuedMultiVector& X, std::string name) const;
+
+  //! dump out boolean ArrayView
+  void dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const;
+
+  //! dump out boolean Kokkos::View
+  void dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const;
+
+  //! get a (synced) timer
+  Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm = Teuchos::null) const;
+
+  //! ParameterLists
+  mutable Teuchos::ParameterList parameterList_, precList11_, precList22_;
+
+  //! Two hierarchies: one for the (1,1)-block, another for the (2,2)-block
+  Teuchos::RCP<Hierarchy> Hierarchy11_, Hierarchy22_, HierarchyGmhd_;
+
+  //! Various matrices
+  Teuchos::RCP<Matrix> SM_Matrix_, D0_Matrix_, Kn_Matrix_, GmhdA_Matrix_;
+
+  //! Vectors for BCs
+  Kokkos::View<bool*, typename Node::device_type> BCrowsKokkos_, BCcolsKokkos_, BCdomainKokkos_;
+  int BCedges_, BCnodes_;
+  Teuchos::ArrayRCP<bool> BCrows_, BCcols_, BCdomain_;
+  //! Nullspace
+  Teuchos::RCP<MultiVector> Nullspace_;
+  //! Coordinates
+  Teuchos::RCP<RealValuedMultiVector> Coords_;
+  //! Some options
+  bool useKokkos_, allEdgesBoundary_, allNodesBoundary_, dump_matrices_, enable_reuse_, syncTimers_;
+  bool applyBCsTo22_;
+
+  //! Execution modes
+  typedef enum { MODE_STANDARD = 0,
+                 MODE_REFMAXWELL,
+                 MODE_EDGE_ONLY,
+                 MODE_GMHD_STANDARD } mode_type;
+  mode_type mode_;
+
+  //! Temporary memory (cached vectors for RefMaxwell-style)
+  RCP<Matrix> P11_;
+  mutable Teuchos::RCP<MultiVector> residualFine_, residual11c_, residual22_, update11c_, update22_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_MAXWELL1_SHORT
-#endif // MUELU_MAXWELL1_DECL_HPP
+#endif  // MUELU_MAXWELL1_DECL_HPP
diff --git a/packages/muelu/src/Operators/MueLu_Maxwell1_def.hpp b/packages/muelu/src/Operators/MueLu_Maxwell1_def.hpp
index 71b43d9d4021..6f9f47d6404f 100644
--- a/packages/muelu/src/Operators/MueLu_Maxwell1_def.hpp
+++ b/packages/muelu/src/Operators/MueLu_Maxwell1_def.hpp
@@ -76,953 +76,898 @@
 #include "cuda_profiler_api.h"
 #endif
 
-
 namespace MueLu {
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getDomainMap() const {
+  return SM_Matrix_->getDomainMap();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getDomainMap() const {
-    return SM_Matrix_->getDomainMap();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getRangeMap() const {
-    return SM_Matrix_->getRangeMap();
-  }
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRangeMap() const {
+  return SM_Matrix_->getRangeMap();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameters(Teuchos::ParameterList& list) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameters(Teuchos::ParameterList& list) {
+  if (list.isType<std::string>("parameterlist: syntax") && list.get<std::string>("parameterlist: syntax") == "ml") {
+    list.remove("parameterlist: syntax");
+    Teuchos::ParameterList newList;
 
-    if (list.isType<std::string>("parameterlist: syntax") && list.get<std::string>("parameterlist: syntax") == "ml") {
-      list.remove("parameterlist: syntax");
-      Teuchos::ParameterList newList;
+    // interpret ML list
+    newList.sublist("maxwell1: 22list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list, "Maxwell"));
 
-      // interpret ML list
-      newList.sublist("maxwell1: 22list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list,"Maxwell"));
-     
-      // Hardwiring options to ensure ML compatibility
-      newList.sublist("maxwell1: 22list").set("use kokkos refactor", false);
+    // Hardwiring options to ensure ML compatibility
+    newList.sublist("maxwell1: 22list").set("use kokkos refactor", false);
 
-      newList.sublist("maxwell1: 11list").set("use kokkos refactor", false);
-      newList.sublist("maxwell1: 11list").set("tentative: constant column sums", false);
-      newList.sublist("maxwell1: 11list").set("tentative: calculate qr", false);
+    newList.sublist("maxwell1: 11list").set("use kokkos refactor", false);
+    newList.sublist("maxwell1: 11list").set("tentative: constant column sums", false);
+    newList.sublist("maxwell1: 11list").set("tentative: calculate qr", false);
 
-      newList.sublist("maxwell1: 11list").set("aggregation: use ml scaling of drop tol",true);
-      newList.sublist("maxwell1: 22list").set("aggregation: use ml scaling of drop tol",true);
+    newList.sublist("maxwell1: 11list").set("aggregation: use ml scaling of drop tol", true);
+    newList.sublist("maxwell1: 22list").set("aggregation: use ml scaling of drop tol", true);
 
-      newList.sublist("maxwell1: 22list").set("aggregation: min agg size",3);
-      newList.sublist("maxwell1: 22list").set("aggregation: match ML phase1",true);
-      newList.sublist("maxwell1: 22list").set("aggregation: match ML phase2a",true);
-      newList.sublist("maxwell1: 22list").set("aggregation: match ML phase2b",true);
+    newList.sublist("maxwell1: 22list").set("aggregation: min agg size", 3);
+    newList.sublist("maxwell1: 22list").set("aggregation: match ML phase1", true);
+    newList.sublist("maxwell1: 22list").set("aggregation: match ML phase2a", true);
+    newList.sublist("maxwell1: 22list").set("aggregation: match ML phase2b", true);
 
-      if(list.isParameter("aggregation: damping factor") && list.get<double>("aggregation: damping factor") == 0.0)
-        newList.sublist("maxwell1: 11list").set("multigrid algorithm", "unsmoothed reitzinger");
-      else
-        newList.sublist("maxwell1: 11list").set("multigrid algorithm", "smoothed reitzinger");
-      newList.sublist("maxwell1: 11list").set("aggregation: type", "uncoupled");
+    if (list.isParameter("aggregation: damping factor") && list.get<double>("aggregation: damping factor") == 0.0)
+      newList.sublist("maxwell1: 11list").set("multigrid algorithm", "unsmoothed reitzinger");
+    else
+      newList.sublist("maxwell1: 11list").set("multigrid algorithm", "smoothed reitzinger");
+    newList.sublist("maxwell1: 11list").set("aggregation: type", "uncoupled");
 
-      newList.sublist("maxwell1: 22list").set("multigrid algorithm", "unsmoothed");
-      newList.sublist("maxwell1: 22list").set("aggregation: type", "uncoupled");
+    newList.sublist("maxwell1: 22list").set("multigrid algorithm", "unsmoothed");
+    newList.sublist("maxwell1: 22list").set("aggregation: type", "uncoupled");
 
-      if (newList.sublist("maxwell1: 22list").isType<std::string>("verbosity"))
-        newList.set("verbosity", newList.sublist("maxwell1: 22list").get<std::string>("verbosity"));
+    if (newList.sublist("maxwell1: 22list").isType<std::string>("verbosity"))
+      newList.set("verbosity", newList.sublist("maxwell1: 22list").get<std::string>("verbosity"));
 
-      // Move coarse solver and smoother stuff to 11list
-      std::vector<std::string> convert = {"coarse:", "smoother:", "smoother: pre", "smoother: post"};
-      for (auto it = convert.begin(); it != convert.end(); ++it) {
-        if (newList.sublist("maxwell1: 22list").isType<std::string>(*it + " type")) {
-          newList.sublist("maxwell1: 11list").set(*it+" type", newList.sublist("maxwell1: 22list").get<std::string>(*it+" type"));
-          newList.sublist("maxwell1: 22list").remove(*it+" type");
-        }
-        if (newList.sublist("maxwell1: 22list").isSublist(*it+" params")) {
-          newList.sublist("maxwell1: 11list").set(*it+" params", newList.sublist("maxwell1: 22list").sublist(*it+" params"));
-          newList.sublist("maxwell1: 22list").remove(*it+" params");
-        }
+    // Move coarse solver and smoother stuff to 11list
+    std::vector<std::string> convert = {"coarse:", "smoother:", "smoother: pre", "smoother: post"};
+    for (auto it = convert.begin(); it != convert.end(); ++it) {
+      if (newList.sublist("maxwell1: 22list").isType<std::string>(*it + " type")) {
+        newList.sublist("maxwell1: 11list").set(*it + " type", newList.sublist("maxwell1: 22list").get<std::string>(*it + " type"));
+        newList.sublist("maxwell1: 22list").remove(*it + " type");
       }
-
-      newList.sublist("maxwell1: 22list").set("smoother: type", "none");
-      newList.sublist("maxwell1: 22list").set("coarse: type", "none");
-   
-      newList.set("maxwell1: nodal smoother fix zero diagonal threshold",1e-10);
-      newList.sublist("maxwell1: 22list").set("rap: fix zero diagonals", true);
-      newList.sublist("maxwell1: 22list").set("rap: fix zero diagonals threshold",1e-10);
-
-
-      list = newList;
-    }
-    std::string  mode_string   = list.get("maxwell1: mode",                  MasterList::getDefault<std::string>("maxwell1: mode"));
-    applyBCsTo22_              = list.get("maxwell1: apply BCs to 22",       false);
-    dump_matrices_             = list.get("maxwell1: dump matrices",         MasterList::getDefault<bool>("maxwell1: dump matrices"));
-
-    // Default smoother.  We'll copy this around.
-    Teuchos::ParameterList defaultSmootherList;
-    defaultSmootherList.set("smoother: type", "CHEBYSHEV");
-    defaultSmootherList.sublist("smoother: params").set("chebyshev: degree",2);
-    defaultSmootherList.sublist("smoother: params").set("chebyshev: ratio eigenvalue",7.0);
-    defaultSmootherList.sublist("smoother: params").set("chebyshev: eigenvalue max iterations",30);
-
-    // Make sure verbosity gets passed to the sublists
-    std::string verbosity = list.get("verbosity","high");
-    VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosity));
-
-    // Check the validity of the run mode
-    if(mode_ != MODE_GMHD_STANDARD) {
-      if(mode_string == "standard")          mode_ = MODE_STANDARD;
-      else if(mode_string == "refmaxwell")   mode_ = MODE_REFMAXWELL;
-      else if(mode_string == "edge only")    mode_ = MODE_EDGE_ONLY;
-      else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must use mode 'standard', 'refmaxwell', 'edge only', or use the GMHD constructor.");
+      if (newList.sublist("maxwell1: 22list").isSublist(*it + " params")) {
+        newList.sublist("maxwell1: 11list").set(*it + " params", newList.sublist("maxwell1: 22list").sublist(*it + " params"));
+        newList.sublist("maxwell1: 22list").remove(*it + " params");
       }
     }
 
-    // If we're in edge only or standard modes, then the (2,2) hierarchy gets built without smoothers.
-    // Otherwise we use the user's smoothers (defaulting to Chebyshev if unspecified)
-    if(list.isSublist("maxwell1: 22list"))
-      precList22_     =  list.sublist("maxwell1: 22list");
-    else if(list.isSublist("refmaxwell: 22list"))
-      precList22_     =  list.sublist("refmaxwell: 22list");
-    if(mode_ == MODE_EDGE_ONLY || mode_ == MODE_STANDARD || mode_ == MODE_GMHD_STANDARD)
-      precList22_.set("smoother: pre or post","none");
-    else if(!precList22_.isType<std::string>("Preconditioner Type") &&
-       !precList22_.isType<std::string>("smoother: type") &&
-       !precList22_.isType<std::string>("smoother: pre type") &&
-       !precList22_.isType<std::string>("smoother: post type")) {
-      precList22_ = defaultSmootherList;
-    }
-    precList22_.set("verbosity",precList22_.get("verbosity",verbosity));
-
-    
+    newList.sublist("maxwell1: 22list").set("smoother: type", "none");
+    newList.sublist("maxwell1: 22list").set("coarse: type", "none");
 
-    // For the (1,1) hierarchy we'll use Hiptmair (STANDARD) or Chebyshev (EDGE_ONLY / REFMAXWELL) if
-    // the user doesn't specify things
-    if(list.isSublist("maxwell1: 11list"))
-      precList11_     =  list.sublist("maxwell1: 11list");
-    else if(list.isSublist("refmaxwell: 11list"))
-      precList11_     =  list.sublist("refmaxwell: 11list");
+    newList.set("maxwell1: nodal smoother fix zero diagonal threshold", 1e-10);
+    newList.sublist("maxwell1: 22list").set("rap: fix zero diagonals", true);
+    newList.sublist("maxwell1: 22list").set("rap: fix zero diagonals threshold", 1e-10);
 
-    if(mode_ == MODE_GMHD_STANDARD) {
-      precList11_.set("smoother: pre or post","none");
-      precList11_.set("smoother: type", "none");
-    }
-    if(!precList11_.isType<std::string>("Preconditioner Type") &&
-       !precList11_.isType<std::string>("smoother: type") &&
-       !precList11_.isType<std::string>("smoother: pre type") &&
-       !precList11_.isType<std::string>("smoother: post type")) {
-      if(mode_ == MODE_EDGE_ONLY || mode_ == MODE_REFMAXWELL) {
-        precList11_ = defaultSmootherList;
-      }
-      if (mode_ == MODE_STANDARD)  {
-        precList11_.set("smoother: type", "HIPTMAIR");
-        precList11_.set("hiptmair: smoother type 1","CHEBYSHEV");
-        precList11_.set("hiptmair: smoother type 2","CHEBYSHEV");
-        precList11_.sublist("hiptmair: smoother list 1") = defaultSmootherList;
-        precList11_.sublist("hiptmair: smoother list 2") = defaultSmootherList;
-      }
+    list = newList;
+  }
+  std::string mode_string = list.get("maxwell1: mode", MasterList::getDefault<std::string>("maxwell1: mode"));
+  applyBCsTo22_           = list.get("maxwell1: apply BCs to 22", false);
+  dump_matrices_          = list.get("maxwell1: dump matrices", MasterList::getDefault<bool>("maxwell1: dump matrices"));
+
+  // Default smoother.  We'll copy this around.
+  Teuchos::ParameterList defaultSmootherList;
+  defaultSmootherList.set("smoother: type", "CHEBYSHEV");
+  defaultSmootherList.sublist("smoother: params").set("chebyshev: degree", 2);
+  defaultSmootherList.sublist("smoother: params").set("chebyshev: ratio eigenvalue", 7.0);
+  defaultSmootherList.sublist("smoother: params").set("chebyshev: eigenvalue max iterations", 30);
+
+  // Make sure verbosity gets passed to the sublists
+  std::string verbosity = list.get("verbosity", "high");
+  VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosity));
+
+  // Check the validity of the run mode
+  if (mode_ != MODE_GMHD_STANDARD) {
+    if (mode_string == "standard")
+      mode_ = MODE_STANDARD;
+    else if (mode_string == "refmaxwell")
+      mode_ = MODE_REFMAXWELL;
+    else if (mode_string == "edge only")
+      mode_ = MODE_EDGE_ONLY;
+    else {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must use mode 'standard', 'refmaxwell', 'edge only', or use the GMHD constructor.");
     }
-    precList11_.set("verbosity",precList11_.get("verbosity",verbosity));
-
-    // Reuse support
-    if (enable_reuse_ &&
-        !precList11_.isType<std::string>("Preconditioner Type") &&
-        !precList11_.isParameter("reuse: type"))
-      precList11_.set("reuse: type", "full");
-    if (enable_reuse_ &&
-        !precList22_.isType<std::string>("Preconditioner Type") &&
-        !precList22_.isParameter("reuse: type"))
-      precList22_.set("reuse: type", "full");
-
-
-    // Are we using Kokkos?
-    useKokkos_ = !Node::is_serial;
-    useKokkos_ = list.get("use kokkos refactor",useKokkos_);
-
-    parameterList_ = list;
-
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GMHDSetupHierarchy(Teuchos::ParameterList& List) const {
-    Teuchos::ParameterList precListGmhd;
-
-    MueLu::HierarchyUtils<SC,LO,GO,NO>::CopyBetweenHierarchies(*Hierarchy11_,*HierarchyGmhd_, "P",  "Psubblock", "RCP<Matrix>");
-
-    HierarchyGmhd_->GetLevel(0)->Set("A", GmhdA_Matrix_);
-    GmhdA_Matrix_->setObjectLabel("GmhdA");
-
-    TEUCHOS_TEST_FOR_EXCEPTION( !List.isSublist("maxwell1: Gmhdlist"), Exceptions::RuntimeError, "Must provide maxwell1: Gmhdlist for GMHD setup");    
-    precListGmhd     =  List.sublist("maxwell1: Gmhdlist");
-    precListGmhd.set("coarse: max size",1);
-    precListGmhd.set("max levels",HierarchyGmhd_->GetNumLevels());
-    RCP<MueLu::HierarchyManager<SC,LO,GO,NO> > mueLuFactory = rcp(new MueLu::ParameterListInterpreter<SC,LO,GO,NO>(precListGmhd,GmhdA_Matrix_->getDomainMap()->getComm()));
-    HierarchyGmhd_->setlib(GmhdA_Matrix_->getDomainMap()->lib());
-    HierarchyGmhd_->SetProcRankVerbose(GmhdA_Matrix_->getDomainMap()->getComm()->getRank());
-    mueLuFactory->SetupHierarchy(*HierarchyGmhd_);
+  // If we're in edge only or standard modes, then the (2,2) hierarchy gets built without smoothers.
+  // Otherwise we use the user's smoothers (defaulting to Chebyshev if unspecified)
+  if (list.isSublist("maxwell1: 22list"))
+    precList22_ = list.sublist("maxwell1: 22list");
+  else if (list.isSublist("refmaxwell: 22list"))
+    precList22_ = list.sublist("refmaxwell: 22list");
+  if (mode_ == MODE_EDGE_ONLY || mode_ == MODE_STANDARD || mode_ == MODE_GMHD_STANDARD)
+    precList22_.set("smoother: pre or post", "none");
+  else if (!precList22_.isType<std::string>("Preconditioner Type") &&
+           !precList22_.isType<std::string>("smoother: type") &&
+           !precList22_.isType<std::string>("smoother: pre type") &&
+           !precList22_.isType<std::string>("smoother: post type")) {
+    precList22_ = defaultSmootherList;
   }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::compute(bool reuse) {
-    /* Algorithm overview for Maxwell1 construction:
-
-       1) Create a nodal auxillary hierarchy based on (a) the user's nodal matrix or (b) a matrix constructed
-       by D0^T A D0 if the user doesn't provide a nodal matrix.  We call this matrix "NodeAggMatrix."
-
-       2)  If the user provided a node matrix, we use the prolongators from the auxillary nodal hierarchy
-       to generate matrices for smoothers on all levels.  We call this "NodeMatrix."  Otherwise NodeMatrix = NodeAggMatrix
-
-       3) We stick all of the nodal P matrices and NodeMatrix objects on the final (1,1) hierarchy and then use the
-       ReitzingerPFactory to generate the edge P and A matrices.
-     */
-
+  precList22_.set("verbosity", precList22_.get("verbosity", verbosity));
+
+  // For the (1,1) hierarchy we'll use Hiptmair (STANDARD) or Chebyshev (EDGE_ONLY / REFMAXWELL) if
+  // the user doesn't specify things
+  if (list.isSublist("maxwell1: 11list"))
+    precList11_ = list.sublist("maxwell1: 11list");
+  else if (list.isSublist("refmaxwell: 11list"))
+    precList11_ = list.sublist("refmaxwell: 11list");
+
+  if (mode_ == MODE_GMHD_STANDARD) {
+    precList11_.set("smoother: pre or post", "none");
+    precList11_.set("smoother: type", "none");
+  }
+  if (!precList11_.isType<std::string>("Preconditioner Type") &&
+      !precList11_.isType<std::string>("smoother: type") &&
+      !precList11_.isType<std::string>("smoother: pre type") &&
+      !precList11_.isType<std::string>("smoother: post type")) {
+    if (mode_ == MODE_EDGE_ONLY || mode_ == MODE_REFMAXWELL) {
+      precList11_ = defaultSmootherList;
+    }
+    if (mode_ == MODE_STANDARD) {
+      precList11_.set("smoother: type", "HIPTMAIR");
+      precList11_.set("hiptmair: smoother type 1", "CHEBYSHEV");
+      precList11_.set("hiptmair: smoother type 2", "CHEBYSHEV");
+      precList11_.sublist("hiptmair: smoother list 1") = defaultSmootherList;
+      precList11_.sublist("hiptmair: smoother list 2") = defaultSmootherList;
+    }
+  }
+  precList11_.set("verbosity", precList11_.get("verbosity", verbosity));
+
+  // Reuse support
+  if (enable_reuse_ &&
+      !precList11_.isType<std::string>("Preconditioner Type") &&
+      !precList11_.isParameter("reuse: type"))
+    precList11_.set("reuse: type", "full");
+  if (enable_reuse_ &&
+      !precList22_.isType<std::string>("Preconditioner Type") &&
+      !precList22_.isParameter("reuse: type"))
+    precList22_.set("reuse: type", "full");
+
+  // Are we using Kokkos?
+  useKokkos_ = !Node::is_serial;
+  useKokkos_ = list.get("use kokkos refactor", useKokkos_);
+
+  parameterList_ = list;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GMHDSetupHierarchy(Teuchos::ParameterList& List) const {
+  Teuchos::ParameterList precListGmhd;
+
+  MueLu::HierarchyUtils<SC, LO, GO, NO>::CopyBetweenHierarchies(*Hierarchy11_, *HierarchyGmhd_, "P", "Psubblock", "RCP<Matrix>");
+
+  HierarchyGmhd_->GetLevel(0)->Set("A", GmhdA_Matrix_);
+  GmhdA_Matrix_->setObjectLabel("GmhdA");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(!List.isSublist("maxwell1: Gmhdlist"), Exceptions::RuntimeError, "Must provide maxwell1: Gmhdlist for GMHD setup");
+  precListGmhd = List.sublist("maxwell1: Gmhdlist");
+  precListGmhd.set("coarse: max size", 1);
+  precListGmhd.set("max levels", HierarchyGmhd_->GetNumLevels());
+  RCP<MueLu::HierarchyManager<SC, LO, GO, NO> > mueLuFactory = rcp(new MueLu::ParameterListInterpreter<SC, LO, GO, NO>(precListGmhd, GmhdA_Matrix_->getDomainMap()->getComm()));
+  HierarchyGmhd_->setlib(GmhdA_Matrix_->getDomainMap()->lib());
+  HierarchyGmhd_->SetProcRankVerbose(GmhdA_Matrix_->getDomainMap()->getComm()->getRank());
+  mueLuFactory->SetupHierarchy(*HierarchyGmhd_);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::compute(bool reuse) {
+  /* Algorithm overview for Maxwell1 construction:
+
+     1) Create a nodal auxillary hierarchy based on (a) the user's nodal matrix or (b) a matrix constructed
+     by D0^T A D0 if the user doesn't provide a nodal matrix.  We call this matrix "NodeAggMatrix."
+
+     2)  If the user provided a node matrix, we use the prolongators from the auxillary nodal hierarchy
+     to generate matrices for smoothers on all levels.  We call this "NodeMatrix."  Otherwise NodeMatrix = NodeAggMatrix
+
+     3) We stick all of the nodal P matrices and NodeMatrix objects on the final (1,1) hierarchy and then use the
+     ReitzingerPFactory to generate the edge P and A matrices.
+   */
 
 #ifdef HAVE_MUELU_CUDA
-    if (parameterList_.get<bool>("maxwell1: cuda profile setup", false)) cudaProfilerStart();
+  if (parameterList_.get<bool>("maxwell1: cuda profile setup", false)) cudaProfilerStart();
 #endif
 
-    std::string timerLabel;
-    if (reuse)
-      timerLabel = "MueLu Maxwell1: compute (reuse)";
+  std::string timerLabel;
+  if (reuse)
+    timerLabel = "MueLu Maxwell1: compute (reuse)";
+  else
+    timerLabel = "MueLu Maxwell1: compute";
+  RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // Generate Kn and apply BCs (if needed)
+  bool have_generated_Kn = false;
+  if (Kn_Matrix_.is_null()) {
+    // generate_kn() will do diagonal repair if requested
+    GetOStream(Runtime0) << "Maxwell1::compute(): Kn not provided.  Generating." << std::endl;
+    Kn_Matrix_        = generate_kn();
+    have_generated_Kn = true;
+  } else if (parameterList_.get<bool>("rap: fix zero diagonals", true)) {
+    magnitudeType threshold;
+    if (parameterList_.isType<magnitudeType>("rap: fix zero diagonals threshold"))
+      threshold = parameterList_.get<magnitudeType>("rap: fix zero diagonals threshold",
+                                                    Teuchos::ScalarTraits<double>::eps());
     else
-      timerLabel = "MueLu Maxwell1: compute";
-    RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Generate Kn and apply BCs (if needed)
-    bool have_generated_Kn = false;   
-    if(Kn_Matrix_.is_null()) {
-      // generate_kn() will do diagonal repair if requested
-      GetOStream(Runtime0) << "Maxwell1::compute(): Kn not provided.  Generating." << std::endl;
-      Kn_Matrix_ = generate_kn();
-      have_generated_Kn = true;
-    }
-    else if (parameterList_.get<bool>("rap: fix zero diagonals", true)) {
-      magnitudeType threshold;
-      if (parameterList_.isType<magnitudeType>("rap: fix zero diagonals threshold"))
-        threshold = parameterList_.get<magnitudeType>("rap: fix zero diagonals threshold",
-                                                      Teuchos::ScalarTraits<double>::eps());
-      else
-        threshold = Teuchos::as<magnitudeType>(parameterList_.get<double>("rap: fix zero diagonals threshold",
-                                                                          Teuchos::ScalarTraits<double>::eps()));
-      Scalar replacement = Teuchos::as<Scalar>(parameterList_.get<double>("rap: fix zero diagonals replacement",
-                                                                          MasterList::getDefault<double>("rap: fix zero diagonals replacement")));
-      Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(Kn_Matrix_, true, GetOStream(Warnings1), threshold, replacement);
-    }
-
+      threshold = Teuchos::as<magnitudeType>(parameterList_.get<double>("rap: fix zero diagonals threshold",
+                                                                        Teuchos::ScalarTraits<double>::eps()));
+    Scalar replacement = Teuchos::as<Scalar>(parameterList_.get<double>("rap: fix zero diagonals replacement",
+                                                                        MasterList::getDefault<double>("rap: fix zero diagonals replacement")));
+    Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(Kn_Matrix_, true, GetOStream(Warnings1), threshold, replacement);
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Generate the (2,2) Hierarchy
-    Kn_Matrix_->setObjectLabel("Maxwell1 (2,2)");
-    
-    /* Critical ParameterList changes */
-    if (!Coords_.is_null())
-      precList22_.sublist("user data").set("Coordinates",Coords_);
-
-    /* Repartitioning *must* be in sync between hierarchies, but the
-     only thing we need to watch here is the subcomms, since ReitzingerPFactory
-     won't look at all the other stuff */
-    if(precList22_.isParameter("repartition: enable")) {
-      bool repartition = precList22_.get<bool>("repartition: enable");
-      precList11_.set("repartition: enable",repartition);
-
-      // If we're repartitioning (2,2), we need to rebalance for (1,1) to do the right thing
-      if(repartition)
-        precList22_.set("repartition: rebalance P and R",true);
-
-      if (precList22_.isParameter("repartition: use subcommunicators")) {
-        precList11_.set("repartition: use subcommunicators", precList22_.get<bool>("repartition: use subcommunicators"));
-        
-        // We do not want (1,1) and (2,2) blocks being repartitioned seperately, so we specify the map that
-        // is going to be used (this is generated in ReitzingerPFactory)
-        if(precList11_.get<bool>("repartition: use subcommunicators")==true)
-          precList11_.set("repartition: use subcommunicators in place",true);
-      }
-      else {
-        // We'll have Maxwell1 default to using subcommunicators if you don't specify
-        precList11_.set("repartition: use subcommunicators", true);
-        precList22_.set("repartition: use subcommunicators", true);
-        
-        // We do not want (1,1) and (2,2) blocks being repartitioned seperately, so we specify the map that
-        // is going to be used (this is generated in ReitzingerPFactory)
-        precList11_.set("repartition: use subcommunicators in place",true);
-      }        
-        
+  ////////////////////////////////////////////////////////////////////////////////
+  // Generate the (2,2) Hierarchy
+  Kn_Matrix_->setObjectLabel("Maxwell1 (2,2)");
+
+  /* Critical ParameterList changes */
+  if (!Coords_.is_null())
+    precList22_.sublist("user data").set("Coordinates", Coords_);
+
+  /* Repartitioning *must* be in sync between hierarchies, but the
+   only thing we need to watch here is the subcomms, since ReitzingerPFactory
+   won't look at all the other stuff */
+  if (precList22_.isParameter("repartition: enable")) {
+    bool repartition = precList22_.get<bool>("repartition: enable");
+    precList11_.set("repartition: enable", repartition);
+
+    // If we're repartitioning (2,2), we need to rebalance for (1,1) to do the right thing
+    if (repartition)
+      precList22_.set("repartition: rebalance P and R", true);
+
+    if (precList22_.isParameter("repartition: use subcommunicators")) {
+      precList11_.set("repartition: use subcommunicators", precList22_.get<bool>("repartition: use subcommunicators"));
+
+      // We do not want (1,1) and (2,2) blocks being repartitioned seperately, so we specify the map that
+      // is going to be used (this is generated in ReitzingerPFactory)
+      if (precList11_.get<bool>("repartition: use subcommunicators") == true)
+        precList11_.set("repartition: use subcommunicators in place", true);
+    } else {
+      // We'll have Maxwell1 default to using subcommunicators if you don't specify
+      precList11_.set("repartition: use subcommunicators", true);
+      precList22_.set("repartition: use subcommunicators", true);
+
+      // We do not want (1,1) and (2,2) blocks being repartitioned seperately, so we specify the map that
+      // is going to be used (this is generated in ReitzingerPFactory)
+      precList11_.set("repartition: use subcommunicators in place", true);
     }
-    else
-      precList11_.remove("repartition: enable", false);
-   
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Remove explicit zeros from matrices
-    /*
-    Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(parameterList_,D0_Matrix_,SM_Matrix_);
+  } else
+    precList11_.remove("repartition: enable", false);
 
+  ////////////////////////////////////////////////////////////////////////////////
+  // Remove explicit zeros from matrices
+  /*
+  Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(parameterList_,D0_Matrix_,SM_Matrix_);
 
+
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());;
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo",          true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*SM_Matrix_, "SM_Matrix", params);
+  }
+  */
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // Detect Dirichlet boundary conditions
+  if (!reuse) {
+    magnitudeType rowSumTol = precList11_.get("aggregation: row sum drop tol", -1.0);
+    Maxwell_Utils<SC, LO, GO, NO>::detectBoundaryConditionsSM(SM_Matrix_, D0_Matrix_, rowSumTol,
+                                                              useKokkos_, BCrowsKokkos_, BCcolsKokkos_, BCdomainKokkos_,
+                                                              BCedges_, BCnodes_, BCrows_, BCcols_, BCdomain_,
+                                                              allEdgesBoundary_, allNodesBoundary_);
     if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());;
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*SM_Matrix_, "SM_Matrix", params);
-    }
-    */
-
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Detect Dirichlet boundary conditions
-    if (!reuse) {
-      magnitudeType rowSumTol = precList11_.get("aggregation: row sum drop tol",-1.0);
-      Maxwell_Utils<SC,LO,GO,NO>::detectBoundaryConditionsSM(SM_Matrix_,D0_Matrix_,rowSumTol,
-                                                             useKokkos_,BCrowsKokkos_,BCcolsKokkos_,BCdomainKokkos_,
-                                                             BCedges_,BCnodes_,BCrows_,BCcols_,BCdomain_,
-                                                             allEdgesBoundary_,allNodesBoundary_);
-      if (IsPrint(Statistics2)) {
-        GetOStream(Statistics2) << "MueLu::Maxwell1::compute(): Detected " << BCedges_ << " BC rows and " << BCnodes_ << " BC columns." << std::endl;
-      }
+      GetOStream(Statistics2) << "MueLu::Maxwell1::compute(): Detected " << BCedges_ << " BC rows and " << BCnodes_ << " BC columns." << std::endl;
     }
+  }
 
-    if (allEdgesBoundary_) {
-      // All edges have been detected as boundary edges.
-      // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
-      GetOStream(Warnings0) << "All edges are detected as boundary edges!" << std::endl;
-      mode_ = MODE_EDGE_ONLY;
+  if (allEdgesBoundary_) {
+    // All edges have been detected as boundary edges.
+    // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
+    GetOStream(Warnings0) << "All edges are detected as boundary edges!" << std::endl;
+    mode_ = MODE_EDGE_ONLY;
 
-      // Generate single level hierarchy for the edge
-      precList22_.set("max levels", 1);
-    }
-     
- 
-    if (allNodesBoundary_) {
-      // All Nodes have been detected as boundary nodes.
-      // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
-      GetOStream(Warnings0) << "All nodes are detected as boundary nodes!" << std::endl;
-      mode_ = MODE_EDGE_ONLY;
+    // Generate single level hierarchy for the edge
+    precList22_.set("max levels", 1);
+  }
 
-      // Generate single level hierarchy for the edge
-      precList22_.set("max levels", 1);
-    }
-                                              
-    ////////////////////////////////////////////////////////////////////////////////
-    // Build (2,2) hierarchy
-    Hierarchy22_ = MueLu::CreateXpetraPreconditioner(Kn_Matrix_, precList22_);
-
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Apply boundary conditions to D0 (if needed)
-    if(!reuse) {
-      D0_Matrix_->resumeFill();
-      Scalar replaceWith;
-      if (D0_Matrix_->getRowMap()->lib() == Xpetra::UseEpetra)
-        replaceWith= Teuchos::ScalarTraits<SC>::eps();
-      else
-        replaceWith = Teuchos::ScalarTraits<SC>::zero();
-      
-      if(applyBCsTo22_) {
-        GetOStream(Runtime0) << "Maxwell1::compute(): nuking BC rows/cols of D0" << std::endl;        
-        if (useKokkos_) {
-          Utilities::ZeroDirichletCols(D0_Matrix_,BCcolsKokkos_,replaceWith);
-        } else {
-          Utilities::ZeroDirichletCols(D0_Matrix_,BCcols_,replaceWith);
-        }
+  if (allNodesBoundary_) {
+    // All Nodes have been detected as boundary nodes.
+    // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
+    GetOStream(Warnings0) << "All nodes are detected as boundary nodes!" << std::endl;
+    mode_ = MODE_EDGE_ONLY;
+
+    // Generate single level hierarchy for the edge
+    precList22_.set("max levels", 1);
+  }
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // Build (2,2) hierarchy
+  Hierarchy22_ = MueLu::CreateXpetraPreconditioner(Kn_Matrix_, precList22_);
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // Apply boundary conditions to D0 (if needed)
+  if (!reuse) {
+    D0_Matrix_->resumeFill();
+    Scalar replaceWith;
+    if (D0_Matrix_->getRowMap()->lib() == Xpetra::UseEpetra)
+      replaceWith = Teuchos::ScalarTraits<SC>::eps();
+    else
+      replaceWith = Teuchos::ScalarTraits<SC>::zero();
+
+    if (applyBCsTo22_) {
+      GetOStream(Runtime0) << "Maxwell1::compute(): nuking BC rows/cols of D0" << std::endl;
+      if (useKokkos_) {
+        Utilities::ZeroDirichletCols(D0_Matrix_, BCcolsKokkos_, replaceWith);
+      } else {
+        Utilities::ZeroDirichletCols(D0_Matrix_, BCcols_, replaceWith);
       }
-      else {
-        GetOStream(Runtime0) << "Maxwell1::compute(): nuking BC rows of D0" << std::endl;        
-        if (useKokkos_) {
-          Utilities::ZeroDirichletRows(D0_Matrix_,BCrowsKokkos_,replaceWith);
-        } else {
-          Utilities::ZeroDirichletRows(D0_Matrix_,BCrows_,replaceWith);
-        }
+    } else {
+      GetOStream(Runtime0) << "Maxwell1::compute(): nuking BC rows of D0" << std::endl;
+      if (useKokkos_) {
+        Utilities::ZeroDirichletRows(D0_Matrix_, BCrowsKokkos_, replaceWith);
+      } else {
+        Utilities::ZeroDirichletRows(D0_Matrix_, BCrows_, replaceWith);
       }
-
-      D0_Matrix_->fillComplete(D0_Matrix_->getDomainMap(),D0_Matrix_->getRangeMap());
     }
 
+    D0_Matrix_->fillComplete(D0_Matrix_->getDomainMap(), D0_Matrix_->getRangeMap());
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // What ML does is generate nodal prolongators with an auxillary hierarchy based on the 
-    // user's (2,2) matrix.  The actual nodal matrices for smoothing are generated by the
-    // Hiptmair smoother construction.  We're not going to do that --- we'll 
-    // do as we insert them into the final (1,1) hierarchy.
-
-    // Level 0
-    RCP<Matrix> Kn_Smoother_0;
-    if(have_generated_Kn) {
-      Kn_Smoother_0 = Kn_Matrix_;
-    }
-    else {
-      Kn_Smoother_0 = generate_kn();
-    }
+  ////////////////////////////////////////////////////////////////////////////////
+  // What ML does is generate nodal prolongators with an auxillary hierarchy based on the
+  // user's (2,2) matrix.  The actual nodal matrices for smoothing are generated by the
+  // Hiptmair smoother construction.  We're not going to do that --- we'll
+  // do as we insert them into the final (1,1) hierarchy.
+
+  // Level 0
+  RCP<Matrix> Kn_Smoother_0;
+  if (have_generated_Kn) {
+    Kn_Smoother_0 = Kn_Matrix_;
+  } else {
+    Kn_Smoother_0 = generate_kn();
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Copy the relevant (2,2) data to the (1,1) hierarchy
-    Hierarchy11_ = rcp(new Hierarchy("Maxwell1 (1,1)"));
-    RCP<Matrix> OldSmootherMatrix;
-    RCP<Level>  OldEdgeLevel;
-    for(int i=0; i<Hierarchy22_->GetNumLevels(); i++) {
-      Hierarchy11_->AddNewLevel();
-      RCP<Level> NodeL = Hierarchy22_->GetLevel(i);
-      RCP<Level> EdgeL = Hierarchy11_->GetLevel(i);
-      RCP<Operator> NodeOp      = NodeL->Get<RCP<Operator> >("A");
-      RCP<Matrix> NodeAggMatrix  = rcp_dynamic_cast<Matrix>(NodeOp);
-      std::string labelstr = FormattingHelper::getColonLabel(EdgeL->getObjectLabel());      
-
-      if(i==0) {
-        EdgeL->Set("A", SM_Matrix_);
-        EdgeL->Set("D0", D0_Matrix_);
-
-        EdgeL->Set("NodeAggMatrix",NodeAggMatrix);
-        EdgeL->Set("NodeMatrix",Kn_Smoother_0);
-        OldSmootherMatrix = Kn_Smoother_0;
-        OldEdgeLevel = EdgeL;
-      }
-      else {
-        // Set the Nodal P
-        // NOTE:  ML uses normalized prolongators for the aggregation hierarchy
-        // and then prolongators of all 1's for doing the Reitzinger prolongator
-        // generation for the edge hierarchy.
-        auto NodalP = NodeL->Get<RCP<Matrix> >("P");
-        auto NodalP_ones = Utilities::ReplaceNonZerosWithOnes(NodalP);
-        TEUCHOS_TEST_FOR_EXCEPTION(NodalP_ones.is_null(), Exceptions::RuntimeError, "Applying ones to prolongator failed");        
-        EdgeL->Set("Pnodal",NodalP_ones);
-
-        // If we repartition a processor away, a RCP<Operator> is stuck
-        // on the level instead of an RCP<Matrix>
-        if(!NodeAggMatrix.is_null()) {
-          EdgeL->Set("NodeAggMatrix",NodeAggMatrix);
-          if(!have_generated_Kn) {
-            // The user gave us a Kn on the fine level, so we're using a seperate aggregation
-            // hierarchy from the smoothing hierarchy.
-
-            // ML does a *fixed* 1e-10 diagonal repair on the Nodal Smoothing Matrix
-            // This fix is applied *after* the next level is generated, but before the smoother is.
-            // We can see this behavior from ML, though it isn't 100% clear from the code *how* it happens.
-            // So, here we turn the fix off, then once we've generated the new matrix, we fix the old one.
-
-            // Generate the new matrix
-            Teuchos::ParameterList RAPlist;
-            RAPlist.set("rap: fix zero diagonals", false);
-            RCP<Matrix> NewKn = Maxwell_Utils<SC,LO,GO,NO>::PtAPWrapper(OldSmootherMatrix,NodalP_ones,RAPlist,labelstr);
-            EdgeL->Set("NodeMatrix",NewKn);
-
-            // Fix the old one
-            double thresh = parameterList_.get("maxwell1: nodal smoother fix zero diagonal threshold",1e-10);
-            if(thresh > 0.0) {
-              printf("CMS: Reparing diagonal after next level generation\n");
-              Scalar replacement = Teuchos::ScalarTraits<Scalar>::one();
-              Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(OldSmootherMatrix, true, GetOStream(Warnings1), thresh, replacement);
-            }
-            OldEdgeLevel->Set("NodeMatrix",OldSmootherMatrix);
-
-            OldSmootherMatrix = NewKn;
-          }
-          else {          
-            // The user didn't give us a Kn, so we aggregate and smooth with the same matrix
-            EdgeL->Set("NodeMatrix",NodeAggMatrix);
+  ////////////////////////////////////////////////////////////////////////////////
+  // Copy the relevant (2,2) data to the (1,1) hierarchy
+  Hierarchy11_ = rcp(new Hierarchy("Maxwell1 (1,1)"));
+  RCP<Matrix> OldSmootherMatrix;
+  RCP<Level> OldEdgeLevel;
+  for (int i = 0; i < Hierarchy22_->GetNumLevels(); i++) {
+    Hierarchy11_->AddNewLevel();
+    RCP<Level> NodeL          = Hierarchy22_->GetLevel(i);
+    RCP<Level> EdgeL          = Hierarchy11_->GetLevel(i);
+    RCP<Operator> NodeOp      = NodeL->Get<RCP<Operator> >("A");
+    RCP<Matrix> NodeAggMatrix = rcp_dynamic_cast<Matrix>(NodeOp);
+    std::string labelstr      = FormattingHelper::getColonLabel(EdgeL->getObjectLabel());
+
+    if (i == 0) {
+      EdgeL->Set("A", SM_Matrix_);
+      EdgeL->Set("D0", D0_Matrix_);
+
+      EdgeL->Set("NodeAggMatrix", NodeAggMatrix);
+      EdgeL->Set("NodeMatrix", Kn_Smoother_0);
+      OldSmootherMatrix = Kn_Smoother_0;
+      OldEdgeLevel      = EdgeL;
+    } else {
+      // Set the Nodal P
+      // NOTE:  ML uses normalized prolongators for the aggregation hierarchy
+      // and then prolongators of all 1's for doing the Reitzinger prolongator
+      // generation for the edge hierarchy.
+      auto NodalP      = NodeL->Get<RCP<Matrix> >("P");
+      auto NodalP_ones = Utilities::ReplaceNonZerosWithOnes(NodalP);
+      TEUCHOS_TEST_FOR_EXCEPTION(NodalP_ones.is_null(), Exceptions::RuntimeError, "Applying ones to prolongator failed");
+      EdgeL->Set("Pnodal", NodalP_ones);
+
+      // If we repartition a processor away, a RCP<Operator> is stuck
+      // on the level instead of an RCP<Matrix>
+      if (!NodeAggMatrix.is_null()) {
+        EdgeL->Set("NodeAggMatrix", NodeAggMatrix);
+        if (!have_generated_Kn) {
+          // The user gave us a Kn on the fine level, so we're using a seperate aggregation
+          // hierarchy from the smoothing hierarchy.
+
+          // ML does a *fixed* 1e-10 diagonal repair on the Nodal Smoothing Matrix
+          // This fix is applied *after* the next level is generated, but before the smoother is.
+          // We can see this behavior from ML, though it isn't 100% clear from the code *how* it happens.
+          // So, here we turn the fix off, then once we've generated the new matrix, we fix the old one.
+
+          // Generate the new matrix
+          Teuchos::ParameterList RAPlist;
+          RAPlist.set("rap: fix zero diagonals", false);
+          RCP<Matrix> NewKn = Maxwell_Utils<SC, LO, GO, NO>::PtAPWrapper(OldSmootherMatrix, NodalP_ones, RAPlist, labelstr);
+          EdgeL->Set("NodeMatrix", NewKn);
+
+          // Fix the old one
+          double thresh = parameterList_.get("maxwell1: nodal smoother fix zero diagonal threshold", 1e-10);
+          if (thresh > 0.0) {
+            printf("CMS: Reparing diagonal after next level generation\n");
+            Scalar replacement = Teuchos::ScalarTraits<Scalar>::one();
+            Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(OldSmootherMatrix, true, GetOStream(Warnings1), thresh, replacement);
           }
-        }
-        else {
-          // We've partitioned things away.
-          EdgeL->Set("NodeMatrix",NodeOp);
-          EdgeL->Set("NodeAggMatrix",NodeOp);
-        }
+          OldEdgeLevel->Set("NodeMatrix", OldSmootherMatrix);
 
-        OldEdgeLevel = EdgeL;            
-      }
-
-      // Get the importer if we have one (for repartitioning)
-      // This will get used in ReitzingerPFactory
-      if(NodeL->IsAvailable("Importer")) {
-        auto importer = NodeL->Get<RCP<const Import> >("Importer");
-        EdgeL->Set("NodeImporter",importer);
-      }
-    }// end Hierarchy22 loop
-
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Generating the (1,1) Hierarchy
-    std::string fine_smoother = precList11_.get<std::string>("smoother: type");
-    {
-      SM_Matrix_->setObjectLabel("A(1,1)");
-      precList11_.set("coarse: max size",1);
-      precList11_.set("max levels",Hierarchy22_->GetNumLevels());
-
-      const bool refmaxwellCoarseSolve = (precList11_.get<std::string>("coarse: type",
-                                                                       MasterList::getDefault<std::string>("coarse: type")) == "RefMaxwell");
-      if (refmaxwellCoarseSolve) {
-        GetOStream(Runtime0) << "Maxwell1::compute(): Will set up RefMaxwell coarse solver" << std::endl;
-        precList11_.set("coarse: type", "none");
-      }
-
-      // Rip off non-serializable data before validation
-      Teuchos::ParameterList nonSerialList11, processedPrecList11;
-      MueLu::ExtractNonSerializableData(precList11_, processedPrecList11, nonSerialList11);
-      RCP<HierarchyManager<SC,LO,GO,NO> > mueLuFactory = rcp(new ParameterListInterpreter<SC,LO,GO,NO>(processedPrecList11,SM_Matrix_->getDomainMap()->getComm()));
-      Hierarchy11_->setlib(SM_Matrix_->getDomainMap()->lib());
-      Hierarchy11_->SetProcRankVerbose(SM_Matrix_->getDomainMap()->getComm()->getRank());
-      // Stick the non-serializible data on the hierarchy.
-      HierarchyUtils<SC,LO,GO,NO>::AddNonSerializableDataToHierarchy(*mueLuFactory,*Hierarchy11_, nonSerialList11);
-      mueLuFactory->SetupHierarchy(*Hierarchy11_);
-
-      if (refmaxwellCoarseSolve) {
-        GetOStream(Runtime0) << "Maxwell1::compute(): Setting up RefMaxwell coarse solver" << std::endl;
-        auto coarseLvl = Hierarchy11_->GetLevel(Hierarchy11_->GetNumLevels()-1);
-        auto coarseSolver = rcp(new MueLu::RefMaxwellSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>("RefMaxwell",
-                                                                                                      precList11_.sublist("coarse: params")));
-        coarseSolver->Setup(*coarseLvl);
-        coarseLvl->Set("PreSmoother",
-                       rcp_dynamic_cast<SmootherBase>(coarseSolver, true));
-      }
-      
-      if(mode_ == MODE_REFMAXWELL) {
-        if(Hierarchy11_->GetNumLevels() > 1) {
-          RCP<Level> EdgeL = Hierarchy11_->GetLevel(1);
-          P11_ = EdgeL->Get<RCP<Matrix> >("P");
+          OldSmootherMatrix = NewKn;
+        } else {
+          // The user didn't give us a Kn, so we aggregate and smooth with the same matrix
+          EdgeL->Set("NodeMatrix", NodeAggMatrix);
         }
+      } else {
+        // We've partitioned things away.
+        EdgeL->Set("NodeMatrix", NodeOp);
+        EdgeL->Set("NodeAggMatrix", NodeOp);
       }
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Allocate temporary MultiVectors for solve (only needed for RefMaxwell style)
-    allocateMemory(1);
-
-    describe(GetOStream(Runtime0));
 
-
-#ifdef MUELU_MAXWELL1_DEBUG
-    for(int i=0; i<Hierarchy11_->GetNumLevels(); i++) {
-      RCP<Level> L = Hierarchy11_->GetLevel(i);
-      RCP<Matrix> EdgeMatrix = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("A"));
-      RCP<Matrix> NodeMatrix = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("NodeMatrix"));
-      RCP<Matrix> NodeAggMatrix = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("NodeAggMatrix"));
-      RCP<Matrix> D0         =rcp_dynamic_cast<Matrix>( L->Get<RCP<Operator> >("D0"));
-      
-      auto nrmE = EdgeMatrix->getFrobeniusNorm();
-      auto nrmN = NodeMatrix->getFrobeniusNorm();
-      auto nrmNa = NodeAggMatrix->getFrobeniusNorm();
-      auto nrmD0= D0->getFrobeniusNorm();
-
-      std::cout<<"DEBUG: Norms on Level "<<i<<" E/N/NA/D0 = "<<nrmE<<" / "<<nrmN <<" / "<<nrmNa<<" / "<< nrmD0 <<std::endl;
-      std::cout<<"DEBUG: NNZ on Level    "<<i<<" E/N/NA/D0 = "<< 
-        EdgeMatrix->getGlobalNumEntries()<<" / "<<
-        NodeMatrix->getGlobalNumEntries()<<" / "<<
-        NodeAggMatrix->getGlobalNumEntries()<<" / "<<
-        D0->getGlobalNumEntries()<<std::endl;      
+      OldEdgeLevel = EdgeL;
     }
-#endif
-
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::generate_kn() const {
-
-    // This is important, as we'll be doing diagonal repair *after* the next-level matrix is generated, not before
-    Teuchos::ParameterList RAPlist;
-    RAPlist.set("rap: fix zero diagonals", false);
-
-    std::string labelstr = "NodeMatrix (Level 0)";
-    RCP<Matrix> rv =  Maxwell_Utils<SC,LO,GO,NO>::PtAPWrapper(SM_Matrix_,D0_Matrix_,RAPlist,labelstr);
-    return rv;
-  }
 
+    // Get the importer if we have one (for repartitioning)
+    // This will get used in ReitzingerPFactory
+    if (NodeL->IsAvailable("Importer")) {
+      auto importer = NodeL->Get<RCP<const Import> >("Importer");
+      EdgeL->Set("NodeImporter", importer);
+    }
+  }  // end Hierarchy22 loop
 
+  ////////////////////////////////////////////////////////////////////////////////
+  // Generating the (1,1) Hierarchy
+  std::string fine_smoother = precList11_.get<std::string>("smoother: type");
+  {
+    SM_Matrix_->setObjectLabel("A(1,1)");
+    precList11_.set("coarse: max size", 1);
+    precList11_.set("max levels", Hierarchy22_->GetNumLevels());
+
+    const bool refmaxwellCoarseSolve = (precList11_.get<std::string>("coarse: type",
+                                                                     MasterList::getDefault<std::string>("coarse: type")) == "RefMaxwell");
+    if (refmaxwellCoarseSolve) {
+      GetOStream(Runtime0) << "Maxwell1::compute(): Will set up RefMaxwell coarse solver" << std::endl;
+      precList11_.set("coarse: type", "none");
+    }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::allocateMemory(int numVectors) const {
-    if(mode_ == MODE_REFMAXWELL) {
-      RCP<Teuchos::TimeMonitor> tmAlloc = getTimer("MueLu Maxwell1: Allocate MVs");
+    // Rip off non-serializable data before validation
+    Teuchos::ParameterList nonSerialList11, processedPrecList11;
+    MueLu::ExtractNonSerializableData(precList11_, processedPrecList11, nonSerialList11);
+    RCP<HierarchyManager<SC, LO, GO, NO> > mueLuFactory = rcp(new ParameterListInterpreter<SC, LO, GO, NO>(processedPrecList11, SM_Matrix_->getDomainMap()->getComm()));
+    Hierarchy11_->setlib(SM_Matrix_->getDomainMap()->lib());
+    Hierarchy11_->SetProcRankVerbose(SM_Matrix_->getDomainMap()->getComm()->getRank());
+    // Stick the non-serializible data on the hierarchy.
+    HierarchyUtils<SC, LO, GO, NO>::AddNonSerializableDataToHierarchy(*mueLuFactory, *Hierarchy11_, nonSerialList11);
+    mueLuFactory->SetupHierarchy(*Hierarchy11_);
+
+    if (refmaxwellCoarseSolve) {
+      GetOStream(Runtime0) << "Maxwell1::compute(): Setting up RefMaxwell coarse solver" << std::endl;
+      auto coarseLvl    = Hierarchy11_->GetLevel(Hierarchy11_->GetNumLevels() - 1);
+      auto coarseSolver = rcp(new MueLu::RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>("RefMaxwell",
+                                                                                                       precList11_.sublist("coarse: params")));
+      coarseSolver->Setup(*coarseLvl);
+      coarseLvl->Set("PreSmoother",
+                     rcp_dynamic_cast<SmootherBase>(coarseSolver, true));
+    }
 
-      residualFine_ = MultiVectorFactory::Build(SM_Matrix_->getRangeMap(), numVectors);
-     
-      if(!Hierarchy11_.is_null() && Hierarchy11_->GetNumLevels() > 1) {
+    if (mode_ == MODE_REFMAXWELL) {
+      if (Hierarchy11_->GetNumLevels() > 1) {
         RCP<Level> EdgeL = Hierarchy11_->GetLevel(1);
-        RCP<Matrix> A = EdgeL->Get<RCP<Matrix> >("A");
-        residual11c_ = MultiVectorFactory::Build(A->getRangeMap(), numVectors);
-        update11c_   = MultiVectorFactory::Build(A->getDomainMap(), numVectors);
+        P11_             = EdgeL->Get<RCP<Matrix> >("P");
       }
-      
-      if(!Hierarchy22_.is_null()) {
-        residual22_ = MultiVectorFactory::Build(D0_Matrix_->getDomainMap(), numVectors);
-        update22_   = MultiVectorFactory::Build(D0_Matrix_->getDomainMap(), numVectors);
-      }
-
     }
-
   }
 
+  ////////////////////////////////////////////////////////////////////////////////
+  // Allocate temporary MultiVectors for solve (only needed for RefMaxwell style)
+  allocateMemory(1);
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const Matrix& A, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<SC, LO, GO, NO>::Write(name, A);
-    }
-  }
-
+  describe(GetOStream(Runtime0));
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const MultiVector& X, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<SC, LO, GO, NO>::Write(name, X);
-    }
+#ifdef MUELU_MAXWELL1_DEBUG
+  for (int i = 0; i < Hierarchy11_->GetNumLevels(); i++) {
+    RCP<Level> L              = Hierarchy11_->GetLevel(i);
+    RCP<Matrix> EdgeMatrix    = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("A"));
+    RCP<Matrix> NodeMatrix    = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("NodeMatrix"));
+    RCP<Matrix> NodeAggMatrix = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("NodeAggMatrix"));
+    RCP<Matrix> D0            = rcp_dynamic_cast<Matrix>(L->Get<RCP<Operator> >("D0"));
+
+    auto nrmE  = EdgeMatrix->getFrobeniusNorm();
+    auto nrmN  = NodeMatrix->getFrobeniusNorm();
+    auto nrmNa = NodeAggMatrix->getFrobeniusNorm();
+    auto nrmD0 = D0->getFrobeniusNorm();
+
+    std::cout << "DEBUG: Norms on Level " << i << " E/N/NA/D0 = " << nrmE << " / " << nrmN << " / " << nrmNa << " / " << nrmD0 << std::endl;
+    std::cout << "DEBUG: NNZ on Level    " << i << " E/N/NA/D0 = " << EdgeMatrix->getGlobalNumEntries() << " / " << NodeMatrix->getGlobalNumEntries() << " / " << NodeAggMatrix->getGlobalNumEntries() << " / " << D0->getGlobalNumEntries() << std::endl;
   }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dumpCoords(const RealValuedMultiVector& X, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<coordinateType, LO, GO, NO>::Write(name, X);
+#endif
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::generate_kn() const {
+  // This is important, as we'll be doing diagonal repair *after* the next-level matrix is generated, not before
+  Teuchos::ParameterList RAPlist;
+  RAPlist.set("rap: fix zero diagonals", false);
+
+  std::string labelstr = "NodeMatrix (Level 0)";
+  RCP<Matrix> rv       = Maxwell_Utils<SC, LO, GO, NO>::PtAPWrapper(SM_Matrix_, D0_Matrix_, RAPlist, labelstr);
+  return rv;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::allocateMemory(int numVectors) const {
+  if (mode_ == MODE_REFMAXWELL) {
+    RCP<Teuchos::TimeMonitor> tmAlloc = getTimer("MueLu Maxwell1: Allocate MVs");
+
+    residualFine_ = MultiVectorFactory::Build(SM_Matrix_->getRangeMap(), numVectors);
+
+    if (!Hierarchy11_.is_null() && Hierarchy11_->GetNumLevels() > 1) {
+      RCP<Level> EdgeL = Hierarchy11_->GetLevel(1);
+      RCP<Matrix> A    = EdgeL->Get<RCP<Matrix> >("A");
+      residual11c_     = MultiVectorFactory::Build(A->getRangeMap(), numVectors);
+      update11c_       = MultiVectorFactory::Build(A->getDomainMap(), numVectors);
     }
-  }
-
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      std::ofstream out(name);
-      for (size_t i = 0; i < Teuchos::as<size_t>(v.size()); i++)
-        out << v[i] << "\n";
+    if (!Hierarchy22_.is_null()) {
+      residual22_ = MultiVectorFactory::Build(D0_Matrix_->getDomainMap(), numVectors);
+      update22_   = MultiVectorFactory::Build(D0_Matrix_->getDomainMap(), numVectors);
     }
   }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      std::ofstream out(name);
-      auto vH = Kokkos::create_mirror_view (v);
-          Kokkos::deep_copy(vH , v);
-          for (size_t i = 0; i < vH.size(); i++)
-            out << vH[i] << "\n";
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const Matrix& A, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<SC, LO, GO, NO>::Write(name, A);
   }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::TimeMonitor> Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm) const {
-    if (IsPrint(Timings)) {
-      if (!syncTimers_)
-        return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name)));
-      else {
-        if (comm.is_null())
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), SM_Matrix_->getRowMap()->getComm().ptr()));
-        else
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), comm.ptr()));
-      }
-    } else
-      return Teuchos::null;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const MultiVector& X, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<SC, LO, GO, NO>::Write(name, X);
   }
+}
 
-
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::resetMatrix(RCP<Matrix> SM_Matrix_new, bool ComputePrec) {
-    bool reuse = !SM_Matrix_.is_null();
-    SM_Matrix_ = SM_Matrix_new;
-    dump(*SM_Matrix_, "SM.m");
-    if (ComputePrec)
-      compute(reuse);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dumpCoords(const RealValuedMultiVector& X, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<coordinateType, LO, GO, NO>::Write(name, X);
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    std::ofstream out(name);
+    for (size_t i = 0; i < Teuchos::as<size_t>(v.size()); i++)
+      out << v[i] << "\n";
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    std::ofstream out(name);
+    auto vH = Kokkos::create_mirror_view(v);
+    Kokkos::deep_copy(vH, v);
+    for (size_t i = 0; i < vH.size(); i++)
+      out << vH[i] << "\n";
+  }
+}
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyInverseRefMaxwellAdditive(const MultiVector& RHS, MultiVector& X) const {
-    // make sure that we have enough temporary memory
-    const SC one = Teuchos::ScalarTraits<SC>::one();
-    if (!allEdgesBoundary_ && X.getNumVectors() != residualFine_->getNumVectors())
-      allocateMemory(X.getNumVectors());
-
-    TEUCHOS_TEST_FOR_EXCEPTION(Hierarchy11_.is_null() || Hierarchy11_->GetNumLevels() == 0, Exceptions::RuntimeError, "(1,1) Hiearchy is null.");
-
-    // 1) Run fine pre-smoother using Hierarchy11
-    RCP<Level> Fine = Hierarchy11_->GetLevel(0);
-    if (Fine->IsAvailable("PreSmoother")) {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: PreSmoother");
-      RCP<SmootherBase> preSmoo = Fine->Get< RCP<SmootherBase> >("PreSmoother");
-      preSmoo->Apply(X, RHS, true);
-   }
-
-    // 2) Compute residual
-    {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: residual calculation");
-      Utilities::Residual(*SM_Matrix_, X, RHS,*residualFine_);
-    }
-
-    // 3a) Restrict residual to (1,1) Hierarchy's level 1 and execute (1,1) hierarchy (use startLevel and InitialGuessIsZero)
-    if(!P11_.is_null()) {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: (1,1) correction");
-      P11_->apply(*residualFine_,*residual11c_,Teuchos::TRANS);
-      Hierarchy11_->Iterate(*residual11c_,*update11c_,true,1);
-    }
-
-    // 3b) Restrict residual to (2,2) Hierarchy's level 0 and execute (2,2) hierarchy (use InitialGuessIsZero)
-    if (!allNodesBoundary_) {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: (2,2) correction");
-      D0_Matrix_->apply(*residualFine_,*residual22_,Teuchos::TRANS);
-      Hierarchy22_->Iterate(*residual22_,*update22_,true,0);
-    }
-
-    // 4) Prolong both updates back into X-vector (Need to do both the P11 null and not null cases
-    {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: Orolongation");
-      if(!P11_.is_null())
-        P11_->apply(*update11c_,X,Teuchos::NO_TRANS,one,one);
-      if (!allNodesBoundary_)
-        D0_Matrix_->apply(*update22_,X,Teuchos::NO_TRANS,one,one);
-    }
-
-    // 5) Run fine post-smoother using Hierarchy11
-    if (Fine->IsAvailable("PostSmoother")) {
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: PostSmoother");
-      RCP<SmootherBase> postSmoo = Fine->Get< RCP<SmootherBase> >("PostSmoother");
-      postSmoo->Apply(X, RHS, false);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::TimeMonitor> Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm) const {
+  if (IsPrint(Timings)) {
+    if (!syncTimers_)
+      return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name)));
+    else {
+      if (comm.is_null())
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), SM_Matrix_->getRowMap()->getComm().ptr()));
+      else
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), comm.ptr()));
     }
-
-
+  } else
+    return Teuchos::null;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::resetMatrix(RCP<Matrix> SM_Matrix_new, bool ComputePrec) {
+  bool reuse = !SM_Matrix_.is_null();
+  SM_Matrix_ = SM_Matrix_new;
+  dump(*SM_Matrix_, "SM.m");
+  if (ComputePrec)
+    compute(reuse);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::applyInverseRefMaxwellAdditive(const MultiVector& RHS, MultiVector& X) const {
+  // make sure that we have enough temporary memory
+  const SC one = Teuchos::ScalarTraits<SC>::one();
+  if (!allEdgesBoundary_ && X.getNumVectors() != residualFine_->getNumVectors())
+    allocateMemory(X.getNumVectors());
+
+  TEUCHOS_TEST_FOR_EXCEPTION(Hierarchy11_.is_null() || Hierarchy11_->GetNumLevels() == 0, Exceptions::RuntimeError, "(1,1) Hiearchy is null.");
+
+  // 1) Run fine pre-smoother using Hierarchy11
+  RCP<Level> Fine = Hierarchy11_->GetLevel(0);
+  if (Fine->IsAvailable("PreSmoother")) {
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: PreSmoother");
+    RCP<SmootherBase> preSmoo       = Fine->Get<RCP<SmootherBase> >("PreSmoother");
+    preSmoo->Apply(X, RHS, true);
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyInverseStandard(const MultiVector& RHS, MultiVector& X) const {
-    Hierarchy11_->Iterate(RHS,X,1,true);
-  }
- 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply (const MultiVector& RHS, MultiVector& X,
-                                                                Teuchos::ETransp /* mode */,
-                                                                  Scalar /* alpha */,
-                                                                  Scalar /* beta */) const {
-    RCP<Teuchos::TimeMonitor> tm = getTimer("MueLu Maxwell1: solve");
-    if(mode_ == MODE_GMHD_STANDARD)
-      HierarchyGmhd_->Iterate(RHS,X,1,true);
-    else if(mode_ == MODE_STANDARD || mode_ == MODE_EDGE_ONLY)
-      applyInverseStandard(RHS,X);
-    else if(mode_ == MODE_REFMAXWELL)
-      applyInverseRefMaxwellAdditive(RHS,X);
-    else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must use mode 'standard', 'refmaxwell' or 'edge only' when not doing GMHD.");
+  // 2) Compute residual
+  {
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: residual calculation");
+    Utilities::Residual(*SM_Matrix_, X, RHS, *residualFine_);
   }
 
-
- 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::hasTransposeApply() const {
-    return false;
+  // 3a) Restrict residual to (1,1) Hierarchy's level 1 and execute (1,1) hierarchy (use startLevel and InitialGuessIsZero)
+  if (!P11_.is_null()) {
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: (1,1) correction");
+    P11_->apply(*residualFine_, *residual11c_, Teuchos::TRANS);
+    Hierarchy11_->Iterate(*residual11c_, *update11c_, true, 1);
   }
 
+  // 3b) Restrict residual to (2,2) Hierarchy's level 0 and execute (2,2) hierarchy (use InitialGuessIsZero)
+  if (!allNodesBoundary_) {
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: (2,2) correction");
+    D0_Matrix_->apply(*residualFine_, *residual22_, Teuchos::TRANS);
+    Hierarchy22_->Iterate(*residual22_, *update22_, true, 0);
+  }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initialize(const Teuchos::RCP<Matrix> & D0_Matrix,
-             const Teuchos::RCP<Matrix> & Kn_Matrix,
-             const Teuchos::RCP<MultiVector>  & Nullspace,
-             const Teuchos::RCP<RealValuedMultiVector>  & Coords,
-             Teuchos::ParameterList& List)
+  // 4) Prolong both updates back into X-vector (Need to do both the P11 null and not null cases
   {
-    // some pre-conditions
-    TEUCHOS_ASSERT(D0_Matrix!=Teuchos::null);
-
-#ifdef HAVE_MUELU_DEBUG
-    if(!Kn_Matrix.is_null()) {
-      TEUCHOS_ASSERT(Kn_Matrix->getDomainMap()->isSameAs(*D0_Matrix->getDomainMap()));
-      TEUCHOS_ASSERT(Kn_Matrix->getRangeMap()->isSameAs(*D0_Matrix->getDomainMap()));
-    }
-
-
-    TEUCHOS_ASSERT(D0_Matrix->getRangeMap()->isSameAs(*D0_Matrix->getRowMap()));
-#endif
-
-    Hierarchy11_   = Teuchos::null;
-    Hierarchy22_   = Teuchos::null;
-    HierarchyGmhd_  = Teuchos::null;
-    if (mode_ != MODE_GMHD_STANDARD) mode_ = MODE_STANDARD;
-
-    // Default settings
-    useKokkos_=false;
-    allEdgesBoundary_=false;
-    allNodesBoundary_=false;
-    dump_matrices_ = false;
-    enable_reuse_=false;
-    syncTimers_=false;
-    applyBCsTo22_ = false;
-
-    
-    // set parameters
-    setParameters(List);
-
-    if (D0_Matrix->getRowMap()->lib() == Xpetra::UseTpetra) {
-      // We will remove boundary conditions from D0, and potentially change maps, so we copy the input.
-      // Fortunately, D0 is quite sparse.
-      // We cannot use the Tpetra copy constructor, since it does not copy the graph.
-
-      RCP<Matrix> D0copy = MatrixFactory::Build(D0_Matrix->getRowMap(), D0_Matrix->getColMap(), 0);
-      RCP<CrsMatrix> D0copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(D0copy,true)->getCrsMatrix();
-      ArrayRCP<const size_t> D0rowptr_RCP;
-      ArrayRCP<const LO>     D0colind_RCP;
-      ArrayRCP<const SC>     D0vals_RCP;
-      rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix,true)->getCrsMatrix()->getAllValues(D0rowptr_RCP,
-                                                                                    D0colind_RCP,
-                                                                                    D0vals_RCP);
-
-      ArrayRCP<size_t> D0copyrowptr_RCP;
-      ArrayRCP<LO>     D0copycolind_RCP;
-      ArrayRCP<SC>     D0copyvals_RCP;
-      D0copyCrs->allocateAllValues(D0vals_RCP.size(),D0copyrowptr_RCP,D0copycolind_RCP,D0copyvals_RCP);
-      D0copyrowptr_RCP.deepCopy(D0rowptr_RCP());
-      D0copycolind_RCP.deepCopy(D0colind_RCP());
-      D0copyvals_RCP.deepCopy(D0vals_RCP());
-      D0copyCrs->setAllValues(D0copyrowptr_RCP,
-                              D0copycolind_RCP,
-                              D0copyvals_RCP);
-      D0copyCrs->expertStaticFillComplete(D0_Matrix->getDomainMap(), D0_Matrix->getRangeMap(),
-                                          rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix,true)->getCrsMatrix()->getCrsGraph()->getImporter(),
-                                          rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix,true)->getCrsMatrix()->getCrsGraph()->getExporter());
-      D0_Matrix_ = D0copy;
-    } else
-      D0_Matrix_ = MatrixFactory::BuildCopy(D0_Matrix);
-
-
-    Kn_Matrix_    = Kn_Matrix;
-    Coords_       = Coords;
-    Nullspace_    = Nullspace;
-
-    if(!Kn_Matrix_.is_null()) dump(*Kn_Matrix_, "Kn.m");
-    if (!Nullspace_.is_null())    dump(*Nullspace_, "nullspace.m");
-    if (!Coords_.is_null())       dumpCoords(*Coords_, "coords.m");
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: Orolongation");
+    if (!P11_.is_null())
+      P11_->apply(*update11c_, X, Teuchos::NO_TRANS, one, one);
+    if (!allNodesBoundary_)
+      D0_Matrix_->apply(*update22_, X, Teuchos::NO_TRANS, one, one);
+  }
 
+  // 5) Run fine post-smoother using Hierarchy11
+  if (Fine->IsAvailable("PostSmoother")) {
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("MueLu Maxwell1: PostSmoother");
+    RCP<SmootherBase> postSmoo      = Fine->Get<RCP<SmootherBase> >("PostSmoother");
+    postSmoo->Apply(X, RHS, false);
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::applyInverseStandard(const MultiVector& RHS, MultiVector& X) const {
+  Hierarchy11_->Iterate(RHS, X, 1, true);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const MultiVector& RHS, MultiVector& X,
+                                                                Teuchos::ETransp /* mode */,
+                                                                Scalar /* alpha */,
+                                                                Scalar /* beta */) const {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("MueLu Maxwell1: solve");
+  if (mode_ == MODE_GMHD_STANDARD)
+    HierarchyGmhd_->Iterate(RHS, X, 1, true);
+  else if (mode_ == MODE_STANDARD || mode_ == MODE_EDGE_ONLY)
+    applyInverseStandard(RHS, X);
+  else if (mode_ == MODE_REFMAXWELL)
+    applyInverseRefMaxwellAdditive(RHS, X);
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must use mode 'standard', 'refmaxwell' or 'edge only' when not doing GMHD.");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasTransposeApply() const {
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initialize(const Teuchos::RCP<Matrix>& D0_Matrix,
+               const Teuchos::RCP<Matrix>& Kn_Matrix,
+               const Teuchos::RCP<MultiVector>& Nullspace,
+               const Teuchos::RCP<RealValuedMultiVector>& Coords,
+               Teuchos::ParameterList& List) {
+  // some pre-conditions
+  TEUCHOS_ASSERT(D0_Matrix != Teuchos::null);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
+#ifdef HAVE_MUELU_DEBUG
+  if (!Kn_Matrix.is_null()) {
+    TEUCHOS_ASSERT(Kn_Matrix->getDomainMap()->isSameAs(*D0_Matrix->getDomainMap()));
+    TEUCHOS_ASSERT(Kn_Matrix->getRangeMap()->isSameAs(*D0_Matrix->getDomainMap()));
+  }
 
-    std::ostringstream oss;
+  TEUCHOS_ASSERT(D0_Matrix->getRangeMap()->isSameAs(*D0_Matrix->getRowMap()));
+#endif
 
-    RCP<const Teuchos::Comm<int> > comm = SM_Matrix_->getDomainMap()->getComm();
+  Hierarchy11_   = Teuchos::null;
+  Hierarchy22_   = Teuchos::null;
+  HierarchyGmhd_ = Teuchos::null;
+  if (mode_ != MODE_GMHD_STANDARD) mode_ = MODE_STANDARD;
+
+  // Default settings
+  useKokkos_        = false;
+  allEdgesBoundary_ = false;
+  allNodesBoundary_ = false;
+  dump_matrices_    = false;
+  enable_reuse_     = false;
+  syncTimers_       = false;
+  applyBCsTo22_     = false;
+
+  // set parameters
+  setParameters(List);
+
+  if (D0_Matrix->getRowMap()->lib() == Xpetra::UseTpetra) {
+    // We will remove boundary conditions from D0, and potentially change maps, so we copy the input.
+    // Fortunately, D0 is quite sparse.
+    // We cannot use the Tpetra copy constructor, since it does not copy the graph.
+
+    RCP<Matrix> D0copy       = MatrixFactory::Build(D0_Matrix->getRowMap(), D0_Matrix->getColMap(), 0);
+    RCP<CrsMatrix> D0copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(D0copy, true)->getCrsMatrix();
+    ArrayRCP<const size_t> D0rowptr_RCP;
+    ArrayRCP<const LO> D0colind_RCP;
+    ArrayRCP<const SC> D0vals_RCP;
+    rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix, true)->getCrsMatrix()->getAllValues(D0rowptr_RCP, D0colind_RCP, D0vals_RCP);
+
+    ArrayRCP<size_t> D0copyrowptr_RCP;
+    ArrayRCP<LO> D0copycolind_RCP;
+    ArrayRCP<SC> D0copyvals_RCP;
+    D0copyCrs->allocateAllValues(D0vals_RCP.size(), D0copyrowptr_RCP, D0copycolind_RCP, D0copyvals_RCP);
+    D0copyrowptr_RCP.deepCopy(D0rowptr_RCP());
+    D0copycolind_RCP.deepCopy(D0colind_RCP());
+    D0copyvals_RCP.deepCopy(D0vals_RCP());
+    D0copyCrs->setAllValues(D0copyrowptr_RCP,
+                            D0copycolind_RCP,
+                            D0copyvals_RCP);
+    D0copyCrs->expertStaticFillComplete(D0_Matrix->getDomainMap(), D0_Matrix->getRangeMap(),
+                                        rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix, true)->getCrsMatrix()->getCrsGraph()->getImporter(),
+                                        rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix, true)->getCrsMatrix()->getCrsGraph()->getExporter());
+    D0_Matrix_ = D0copy;
+  } else
+    D0_Matrix_ = MatrixFactory::BuildCopy(D0_Matrix);
+
+  Kn_Matrix_ = Kn_Matrix;
+  Coords_    = Coords;
+  Nullspace_ = Nullspace;
+
+  if (!Kn_Matrix_.is_null()) dump(*Kn_Matrix_, "Kn.m");
+  if (!Nullspace_.is_null()) dump(*Nullspace_, "nullspace.m");
+  if (!Coords_.is_null()) dumpCoords(*Coords_, "coords.m");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
+  std::ostringstream oss;
+
+  RCP<const Teuchos::Comm<int> > comm = SM_Matrix_->getDomainMap()->getComm();
 
 #ifdef HAVE_MPI
-    int root;
-    if (!Kn_Matrix_.is_null())
-      root = comm->getRank();
-    else
-      root = -1;
-
-    int actualRoot;
-    reduceAll(*comm, Teuchos::REDUCE_MAX, root, Teuchos::ptr(&actualRoot));
-    root = actualRoot;
+  int root;
+  if (!Kn_Matrix_.is_null())
+    root = comm->getRank();
+  else
+    root = -1;
+
+  int actualRoot;
+  reduceAll(*comm, Teuchos::REDUCE_MAX, root, Teuchos::ptr(&actualRoot));
+  root = actualRoot;
 #endif
 
+  oss << "\n--------------------------------------------------------------------------------\n"
+      << "---                            Maxwell1 Summary                            ---\n"
+         "--------------------------------------------------------------------------------"
+      << std::endl;
+  oss << std::endl;
 
-    oss << "\n--------------------------------------------------------------------------------\n" <<
-      "---                            Maxwell1 Summary                            ---\n"
-      "--------------------------------------------------------------------------------" << std::endl;
-    oss << std::endl;
-
-    GlobalOrdinal numRows;
-    GlobalOrdinal nnz;
+  GlobalOrdinal numRows;
+  GlobalOrdinal nnz;
 
-    SM_Matrix_->getRowMap()->getComm()->barrier();
+  SM_Matrix_->getRowMap()->getComm()->barrier();
 
-    numRows = SM_Matrix_->getGlobalNumRows();
-    nnz = SM_Matrix_->getGlobalNumEntries();
+  numRows = SM_Matrix_->getGlobalNumRows();
+  nnz     = SM_Matrix_->getGlobalNumEntries();
 
-    Xpetra::global_size_t tt = numRows;
-    int rowspacer = 3; while (tt != 0) { tt /= 10; rowspacer++; }
-    tt = nnz;
-    int nnzspacer = 2; while (tt != 0) { tt /= 10; nnzspacer++; }
+  Xpetra::global_size_t tt = numRows;
+  int rowspacer            = 3;
+  while (tt != 0) {
+    tt /= 10;
+    rowspacer++;
+  }
+  tt            = nnz;
+  int nnzspacer = 2;
+  while (tt != 0) {
+    tt /= 10;
+    nnzspacer++;
+  }
 
-    oss  << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
-    oss << "(1, 1)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  oss << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
+  oss << "(1, 1)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
 
-    if (!Kn_Matrix_.is_null()) {
-      // ToDo: make sure that this is printed correctly
-      numRows = Kn_Matrix_->getGlobalNumRows();
-      nnz = Kn_Matrix_->getGlobalNumEntries();
+  if (!Kn_Matrix_.is_null()) {
+    // ToDo: make sure that this is printed correctly
+    numRows = Kn_Matrix_->getGlobalNumRows();
+    nnz     = Kn_Matrix_->getGlobalNumEntries();
 
-      oss << "(2, 2)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
-    }
-
-    oss << std::endl;
+    oss << "(2, 2)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  }
 
+  oss << std::endl;
 
-    std::string outstr = oss.str();
+  std::string outstr = oss.str();
 
 #ifdef HAVE_MPI
-    RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-    MPI_Comm rawComm = (*mpiComm->getRawMpiComm())();
-
-    int strLength = outstr.size();
-    MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
-    if (comm->getRank() != root)
-      outstr.resize(strLength);
-    MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+  RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+  MPI_Comm rawComm                          = (*mpiComm->getRawMpiComm())();
+
+  int strLength = outstr.size();
+  MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
+  if (comm->getRank() != root)
+    outstr.resize(strLength);
+  MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
 #endif
 
-    out << outstr;
+  out << outstr;
 
-    if (!Hierarchy11_.is_null())
-      Hierarchy11_->describe(out, GetVerbLevel());
+  if (!Hierarchy11_.is_null())
+    Hierarchy11_->describe(out, GetVerbLevel());
 
-    if (!Hierarchy22_.is_null())
-      Hierarchy22_->describe(out, GetVerbLevel());
+  if (!Hierarchy22_.is_null())
+    Hierarchy22_->describe(out, GetVerbLevel());
 
-    if (!HierarchyGmhd_.is_null())
-      HierarchyGmhd_->describe(out, GetVerbLevel());
+  if (!HierarchyGmhd_.is_null())
+    HierarchyGmhd_->describe(out, GetVerbLevel());
 
-    if (IsPrint(Statistics2)) {
-      // Print the grid of processors
-      std::ostringstream oss2;
+  if (IsPrint(Statistics2)) {
+    // Print the grid of processors
+    std::ostringstream oss2;
 
-      oss2 << "Sub-solver distribution over ranks" << std::endl;
-      oss2 << "( (1,1) block only is indicated by '1', (2,2) block only by '2', and both blocks by 'B' and none by '.')" << std::endl;
+    oss2 << "Sub-solver distribution over ranks" << std::endl;
+    oss2 << "( (1,1) block only is indicated by '1', (2,2) block only by '2', and both blocks by 'B' and none by '.')" << std::endl;
 
-      int numProcs = comm->getSize();
+    int numProcs = comm->getSize();
 #ifdef HAVE_MPI
-      RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+    RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
 
-      RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
+    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
 #endif
 
-      char status = 0;
-      if (!Kn_Matrix_.is_null())
-        status += 1;
-      std::vector<char> states(numProcs, 0);
+    char status = 0;
+    if (!Kn_Matrix_.is_null())
+      status += 1;
+    std::vector<char> states(numProcs, 0);
 #ifdef HAVE_MPI
-      MPI_Gather(&status, 1, MPI_CHAR, &states[0], 1, MPI_CHAR, 0, *rawMpiComm);
+    MPI_Gather(&status, 1, MPI_CHAR, &states[0], 1, MPI_CHAR, 0, *rawMpiComm);
 #else
-      states.push_back(status);
+    states.push_back(status);
 #endif
 
-      int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
-      for (int proc = 0; proc < numProcs; proc += rowWidth) {
-        for (int j = 0; j < rowWidth; j++)
-          if (proc + j < numProcs)
-            if (states[proc+j] == 0)
-              oss2 << ".";
-            else if (states[proc+j] == 1)
-              oss2 << "1";
-            else if (states[proc+j] == 2)
-              oss2 << "2";
-            else
-              oss2 << "B";
+    int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
+    for (int proc = 0; proc < numProcs; proc += rowWidth) {
+      for (int j = 0; j < rowWidth; j++)
+        if (proc + j < numProcs)
+          if (states[proc + j] == 0)
+            oss2 << ".";
+          else if (states[proc + j] == 1)
+            oss2 << "1";
+          else if (states[proc + j] == 2)
+            oss2 << "2";
           else
-            oss2 << " ";
+            oss2 << "B";
+        else
+          oss2 << " ";
 
-        oss2 << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
-      }
-      oss2 << std::endl;
-      GetOStream(Statistics2) << oss2.str();
+      oss2 << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
     }
-
-
+    oss2 << std::endl;
+    GetOStream(Statistics2) << oss2.str();
   }
+}
 
-
-
-} // namespace
+}  // namespace MueLu
 
 #define MUELU_MAXWELL1_SHORT
-#endif //ifdef MUELU_MAXWELL1_DEF_HPP
+#endif  // ifdef MUELU_MAXWELL1_DEF_HPP
diff --git a/packages/muelu/src/Operators/MueLu_Maxwell_Utils_decl.hpp b/packages/muelu/src/Operators/MueLu_Maxwell_Utils_decl.hpp
index a0add4e54003..c4738a90ea64 100644
--- a/packages/muelu/src/Operators/MueLu_Maxwell_Utils_decl.hpp
+++ b/packages/muelu/src/Operators/MueLu_Maxwell_Utils_decl.hpp
@@ -58,90 +58,85 @@
 #include "MueLu_RAPFactory_fwd.hpp"
 #include "MueLu_Utilities_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @brief Utility functions for Maxwell
-    
-    @ingroup MueLuAdapters
-  */
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class Maxwell_Utils : public VerboseObject {
+/*!
+  @brief Utility functions for Maxwell
 
+  @ingroup MueLuAdapters
+*/
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class Maxwell_Utils : public VerboseObject {
 #undef MUELU_MAXWELL_UTILS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
-    //    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
-    //    typedef typename Xpetra::MultiVector<coordinateType,LO,GO,NO> RealValuedMultiVector;
-
-    //! Detect Dirichlet boundary conditions
-    static void detectBoundaryConditionsSM(RCP<Matrix> & SM_Matrix,
-                                           RCP<Matrix> & D0_Matrix,
-                                           magnitudeType rowSumTol,
-                                           bool useKokkos_,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCrowsKokkos,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCcolsKokkos,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCdomainKokkos,
-                                           int & BCedges,
-                                           int & BCnodes,
-                                           Teuchos::ArrayRCP<bool> & BCrows,
-                                           Teuchos::ArrayRCP<bool> & BCcols,
-                                           Teuchos::ArrayRCP<bool> & BCdomain,
-                                           bool & allEdgesBoundary,
-                                           bool & allNodesBoundary);
-
-    //! Detect Dirichlet boundary conditions
-    static void detectBoundaryConditionsSM(RCP<Matrix> & SM_Matrix,
-                                           RCP<Matrix> & D0_Matrix,
-                                           magnitudeType rowSumTol,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCrowsKokkos,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCcolsKokkos,
-                                           Kokkos::View<bool*, typename Node::device_type> & BCdomainKokkos,
-                                           int & BCedges,
-                                           int & BCnodes,
-                                           bool & allEdgesBoundary,
-                                           bool & allNodesBoundary);
-        
-    //! Remove explicit zeros
-    static RCP<Matrix> removeExplicitZeros(const RCP<Matrix> & A,
-                                           const magnitudeType tolerance,
-                                           const bool keepDiagonal=true,
-                                           const size_t expectedNNZperRow=0);
-
-    static void removeExplicitZeros(Teuchos::ParameterList &parameterList,
-                                    RCP<Matrix> & D0_Matrix,
-                                    RCP<Matrix> & SM_Matrix,
-                                    RCP<Matrix> & M1_Matrix,
-                                    RCP<Matrix> & Ms_Matrix);
-  
-    static void removeExplicitZeros(Teuchos::ParameterList &parameterList,
-                                    RCP<Matrix> & D0_Matrix,
-                                    RCP<Matrix> & SM_Matrix) {
-      RCP<Matrix> dummy;
-      removeExplicitZeros(parameterList,D0_Matrix,SM_Matrix,dummy,dummy);
-    }
-
-    static void thresholdedAbs(const RCP<Matrix> & A,
-                               const magnitudeType thresholded);
-
-  
-    //! Sets matvec params on a matrix
-    static void setMatvecParams(Matrix& A, RCP<ParameterList> matvecParams);
-
-
-    //! Performs an P^T AP
-    static RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-    PtAPWrapper(const RCP<Matrix>& A,const RCP<Matrix>& P, Teuchos::ParameterList &params, std::string & label);
-
-  };
-
-} // namespace
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+  //    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
+  //    typedef typename Xpetra::MultiVector<coordinateType,LO,GO,NO> RealValuedMultiVector;
+
+  //! Detect Dirichlet boundary conditions
+  static void detectBoundaryConditionsSM(RCP<Matrix>& SM_Matrix,
+                                         RCP<Matrix>& D0_Matrix,
+                                         magnitudeType rowSumTol,
+                                         bool useKokkos_,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCrowsKokkos,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCcolsKokkos,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCdomainKokkos,
+                                         int& BCedges,
+                                         int& BCnodes,
+                                         Teuchos::ArrayRCP<bool>& BCrows,
+                                         Teuchos::ArrayRCP<bool>& BCcols,
+                                         Teuchos::ArrayRCP<bool>& BCdomain,
+                                         bool& allEdgesBoundary,
+                                         bool& allNodesBoundary);
+
+  //! Detect Dirichlet boundary conditions
+  static void detectBoundaryConditionsSM(RCP<Matrix>& SM_Matrix,
+                                         RCP<Matrix>& D0_Matrix,
+                                         magnitudeType rowSumTol,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCrowsKokkos,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCcolsKokkos,
+                                         Kokkos::View<bool*, typename Node::device_type>& BCdomainKokkos,
+                                         int& BCedges,
+                                         int& BCnodes,
+                                         bool& allEdgesBoundary,
+                                         bool& allNodesBoundary);
+
+  //! Remove explicit zeros
+  static RCP<Matrix> removeExplicitZeros(const RCP<Matrix>& A,
+                                         const magnitudeType tolerance,
+                                         const bool keepDiagonal        = true,
+                                         const size_t expectedNNZperRow = 0);
+
+  static void removeExplicitZeros(Teuchos::ParameterList& parameterList,
+                                  RCP<Matrix>& D0_Matrix,
+                                  RCP<Matrix>& SM_Matrix,
+                                  RCP<Matrix>& M1_Matrix,
+                                  RCP<Matrix>& Ms_Matrix);
+
+  static void removeExplicitZeros(Teuchos::ParameterList& parameterList,
+                                  RCP<Matrix>& D0_Matrix,
+                                  RCP<Matrix>& SM_Matrix) {
+    RCP<Matrix> dummy;
+    removeExplicitZeros(parameterList, D0_Matrix, SM_Matrix, dummy, dummy);
+  }
+
+  static void thresholdedAbs(const RCP<Matrix>& A,
+                             const magnitudeType thresholded);
+
+  //! Sets matvec params on a matrix
+  static void setMatvecParams(Matrix& A, RCP<ParameterList> matvecParams);
+
+  //! Performs an P^T AP
+  static RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+  PtAPWrapper(const RCP<Matrix>& A, const RCP<Matrix>& P, Teuchos::ParameterList& params, std::string& label);
+};
+
+}  // namespace MueLu
 
 #define MUELU_MAXWELL_UTILS_SHORT
-#endif // MUELU_MAXWELL_UTILS_DECL_HPP
+#endif  // MUELU_MAXWELL_UTILS_DECL_HPP
diff --git a/packages/muelu/src/Operators/MueLu_Maxwell_Utils_def.hpp b/packages/muelu/src/Operators/MueLu_Maxwell_Utils_def.hpp
index 4193ace19e1a..204e61cd34fb 100644
--- a/packages/muelu/src/Operators/MueLu_Maxwell_Utils_def.hpp
+++ b/packages/muelu/src/Operators/MueLu_Maxwell_Utils_def.hpp
@@ -61,270 +61,261 @@
 #include "MueLu_Utilities.hpp"
 #include "MueLu_RAPFactory.hpp"
 
-
 namespace MueLu {
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::detectBoundaryConditionsSM(RCP<Matrix> & SM_Matrix_,
-                                                                                         RCP<Matrix> & D0_Matrix_,
-                                                                                         magnitudeType rowSumTol,
-                                                                                         bool useKokkos_,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCrowsKokkos_,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCcolsKokkos_,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCdomainKokkos_,
-                                                                                         int & BCedges_,
-                                                                                         int & BCnodes_,
-                                                                                         Teuchos::ArrayRCP<bool> & BCrows_,
-                                                                                         Teuchos::ArrayRCP<bool> & BCcols_,
-                                                                                         Teuchos::ArrayRCP<bool> & BCdomain_,
-                                                                                         bool & allEdgesBoundary_,
-                                                                                         bool & allNodesBoundary_) {
-    // clean rows associated with boundary conditions
-    // Find rows with only 1 or 2 nonzero entries, record them in BCrows_.
-    // BCrows_[i] is true, iff i is a boundary row
-    // BCcols_[i] is true, iff i is a boundary column
-    int BCedgesLocal = 0;
-    int BCnodesLocal = 0;
-    if (useKokkos_) {
-      BCrowsKokkos_ = Utilities::DetectDirichletRows_kokkos(*SM_Matrix_,Teuchos::ScalarTraits<magnitudeType>::eps(),/*count_twos_as_dirichlet=*/true);
-
-      if (rowSumTol > 0.)
-        Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrowsKokkos_);
-
-      BCcolsKokkos_ = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_Matrix_->getColMap()->getLocalNumElements());
-      BCdomainKokkos_ = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_Matrix_->getDomainMap()->getLocalNumElements());
-      Utilities::DetectDirichletColsAndDomains(*D0_Matrix_,BCrowsKokkos_,BCcolsKokkos_,BCdomainKokkos_);
-
-      auto BCrowsKokkos=BCrowsKokkos_;
-      Kokkos::parallel_reduce(BCrowsKokkos_.size(), KOKKOS_LAMBDA (int i, int & sum) {
-        if (BCrowsKokkos(i))
-	  ++sum;
-	}, BCedgesLocal );
-         
-      auto BCdomainKokkos = BCdomainKokkos_;
-      Kokkos::parallel_reduce(BCdomainKokkos_.size(), KOKKOS_LAMBDA (int i, int & sum) {
-        if (BCdomainKokkos(i))
-	  ++sum;
-	}, BCnodesLocal);
-    } else
-    {
-      BCrows_ = Teuchos::arcp_const_cast<bool>(Utilities::DetectDirichletRows(*SM_Matrix_,Teuchos::ScalarTraits<magnitudeType>::eps(),/*count_twos_as_dirichlet=*/true));
-
-      if (rowSumTol > 0.)
-        Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrows_);
-
-      BCcols_.resize(D0_Matrix_->getColMap()->getLocalNumElements());
-      BCdomain_.resize(D0_Matrix_->getDomainMap()->getLocalNumElements());
-      Utilities::DetectDirichletColsAndDomains(*D0_Matrix_,BCrows_,BCcols_,BCdomain_);
-
-      for (auto it = BCrows_.begin(); it != BCrows_.end(); ++it)
-        if (*it)
-          BCedgesLocal += 1;
-      for (auto it = BCdomain_.begin(); it != BCdomain_.end(); ++it)
-        if (*it)
-          BCnodesLocal += 1;
-    }
-
-    MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCedgesLocal, BCedges_);
-    MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCnodesLocal, BCnodes_);
-
-
-    allEdgesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCedges_) >= D0_Matrix_->getRangeMap()->getGlobalNumElements();
-    allNodesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCnodes_) >= D0_Matrix_->getDomainMap()->getGlobalNumElements();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::detectBoundaryConditionsSM(RCP<Matrix>& SM_Matrix_,
+                                                                                          RCP<Matrix>& D0_Matrix_,
+                                                                                          magnitudeType rowSumTol,
+                                                                                          bool useKokkos_,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCrowsKokkos_,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCcolsKokkos_,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCdomainKokkos_,
+                                                                                          int& BCedges_,
+                                                                                          int& BCnodes_,
+                                                                                          Teuchos::ArrayRCP<bool>& BCrows_,
+                                                                                          Teuchos::ArrayRCP<bool>& BCcols_,
+                                                                                          Teuchos::ArrayRCP<bool>& BCdomain_,
+                                                                                          bool& allEdgesBoundary_,
+                                                                                          bool& allNodesBoundary_) {
+  // clean rows associated with boundary conditions
+  // Find rows with only 1 or 2 nonzero entries, record them in BCrows_.
+  // BCrows_[i] is true, iff i is a boundary row
+  // BCcols_[i] is true, iff i is a boundary column
+  int BCedgesLocal = 0;
+  int BCnodesLocal = 0;
+  if (useKokkos_) {
+    BCrowsKokkos_ = Utilities::DetectDirichletRows_kokkos(*SM_Matrix_, Teuchos::ScalarTraits<magnitudeType>::eps(), /*count_twos_as_dirichlet=*/true);
+
+    if (rowSumTol > 0.)
+      Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrowsKokkos_);
+
+    BCcolsKokkos_   = Kokkos::View<bool*, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_Matrix_->getColMap()->getLocalNumElements());
+    BCdomainKokkos_ = Kokkos::View<bool*, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_Matrix_->getDomainMap()->getLocalNumElements());
+    Utilities::DetectDirichletColsAndDomains(*D0_Matrix_, BCrowsKokkos_, BCcolsKokkos_, BCdomainKokkos_);
+
+    auto BCrowsKokkos = BCrowsKokkos_;
+    Kokkos::parallel_reduce(
+        BCrowsKokkos_.size(), KOKKOS_LAMBDA(int i, int& sum) {
+          if (BCrowsKokkos(i))
+            ++sum;
+        },
+        BCedgesLocal);
+
+    auto BCdomainKokkos = BCdomainKokkos_;
+    Kokkos::parallel_reduce(
+        BCdomainKokkos_.size(), KOKKOS_LAMBDA(int i, int& sum) {
+          if (BCdomainKokkos(i))
+            ++sum;
+        },
+        BCnodesLocal);
+  } else {
+    BCrows_ = Teuchos::arcp_const_cast<bool>(Utilities::DetectDirichletRows(*SM_Matrix_, Teuchos::ScalarTraits<magnitudeType>::eps(), /*count_twos_as_dirichlet=*/true));
+
+    if (rowSumTol > 0.)
+      Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrows_);
+
+    BCcols_.resize(D0_Matrix_->getColMap()->getLocalNumElements());
+    BCdomain_.resize(D0_Matrix_->getDomainMap()->getLocalNumElements());
+    Utilities::DetectDirichletColsAndDomains(*D0_Matrix_, BCrows_, BCcols_, BCdomain_);
+
+    for (auto it = BCrows_.begin(); it != BCrows_.end(); ++it)
+      if (*it)
+        BCedgesLocal += 1;
+    for (auto it = BCdomain_.begin(); it != BCdomain_.end(); ++it)
+      if (*it)
+        BCnodesLocal += 1;
   }
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::detectBoundaryConditionsSM(RCP<Matrix> & SM_Matrix_,
-                                                                                         RCP<Matrix> & D0_Matrix_,
-                                                                                         magnitudeType rowSumTol,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCrowsKokkos_,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCcolsKokkos_,
-                                                                                         Kokkos::View<bool*, typename Node::device_type> & BCdomainKokkos_,
-                                                                                         int & BCedges_,
-                                                                                         int & BCnodes_,
-                                                                                         bool & allEdgesBoundary_,
-                                                                                         bool & allNodesBoundary_) {
-    // clean rows associated with boundary conditions
-    // Find rows with only 1 or 2 nonzero entries, record them in BCrows_.
-    // BCrows_[i] is true, iff i is a boundary row
-    // BCcols_[i] is true, iff i is a boundary column
-    int BCedgesLocal = 0;
-    int BCnodesLocal = 0;
-    {
-      BCrowsKokkos_ = Utilities::DetectDirichletRows_kokkos(*SM_Matrix_,Teuchos::ScalarTraits<magnitudeType>::eps(),/*count_twos_as_dirichlet=*/true);
-
-      if (rowSumTol > 0.)
-        Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrowsKokkos_);
-
-      BCcolsKokkos_ = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_Matrix_->getColMap()->getLocalNumElements());
-      BCdomainKokkos_ = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_Matrix_->getDomainMap()->getLocalNumElements());
-      Utilities::DetectDirichletColsAndDomains(*D0_Matrix_,BCrowsKokkos_,BCcolsKokkos_,BCdomainKokkos_);
-
-      auto BCrowsKokkos=BCrowsKokkos_;
-      Kokkos::parallel_reduce(BCrowsKokkos_.size(), KOKKOS_LAMBDA (int i, int & sum) {
-        if (BCrowsKokkos(i))
-	  ++sum;
-	}, BCedgesLocal );
-
-      auto BCdomainKokkos = BCdomainKokkos_;
-      Kokkos::parallel_reduce(BCdomainKokkos_.size(), KOKKOS_LAMBDA (int i, int & sum) {
-        if (BCdomainKokkos(i))
-	  ++sum;
-	}, BCnodesLocal);
-    }
-
-    MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCedgesLocal, BCedges_);
-    MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCnodesLocal, BCnodes_);
-
-
-    allEdgesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCedges_) >= D0_Matrix_->getRangeMap()->getGlobalNumElements();
-    allNodesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCnodes_) >= D0_Matrix_->getDomainMap()->getGlobalNumElements();
+  MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCedgesLocal, BCedges_);
+  MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCnodesLocal, BCnodes_);
+
+  allEdgesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCedges_) >= D0_Matrix_->getRangeMap()->getGlobalNumElements();
+  allNodesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCnodes_) >= D0_Matrix_->getDomainMap()->getGlobalNumElements();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::detectBoundaryConditionsSM(RCP<Matrix>& SM_Matrix_,
+                                                                                          RCP<Matrix>& D0_Matrix_,
+                                                                                          magnitudeType rowSumTol,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCrowsKokkos_,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCcolsKokkos_,
+                                                                                          Kokkos::View<bool*, typename Node::device_type>& BCdomainKokkos_,
+                                                                                          int& BCedges_,
+                                                                                          int& BCnodes_,
+                                                                                          bool& allEdgesBoundary_,
+                                                                                          bool& allNodesBoundary_) {
+  // clean rows associated with boundary conditions
+  // Find rows with only 1 or 2 nonzero entries, record them in BCrows_.
+  // BCrows_[i] is true, iff i is a boundary row
+  // BCcols_[i] is true, iff i is a boundary column
+  int BCedgesLocal = 0;
+  int BCnodesLocal = 0;
+  {
+    BCrowsKokkos_ = Utilities::DetectDirichletRows_kokkos(*SM_Matrix_, Teuchos::ScalarTraits<magnitudeType>::eps(), /*count_twos_as_dirichlet=*/true);
+
+    if (rowSumTol > 0.)
+      Utilities::ApplyRowSumCriterion(*SM_Matrix_, rowSumTol, BCrowsKokkos_);
+
+    BCcolsKokkos_   = Kokkos::View<bool*, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_Matrix_->getColMap()->getLocalNumElements());
+    BCdomainKokkos_ = Kokkos::View<bool*, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_Matrix_->getDomainMap()->getLocalNumElements());
+    Utilities::DetectDirichletColsAndDomains(*D0_Matrix_, BCrowsKokkos_, BCcolsKokkos_, BCdomainKokkos_);
+
+    auto BCrowsKokkos = BCrowsKokkos_;
+    Kokkos::parallel_reduce(
+        BCrowsKokkos_.size(), KOKKOS_LAMBDA(int i, int& sum) {
+          if (BCrowsKokkos(i))
+            ++sum;
+        },
+        BCedgesLocal);
+
+    auto BCdomainKokkos = BCdomainKokkos_;
+    Kokkos::parallel_reduce(
+        BCdomainKokkos_.size(), KOKKOS_LAMBDA(int i, int& sum) {
+          if (BCdomainKokkos(i))
+            ++sum;
+        },
+        BCnodesLocal);
   }
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  removeExplicitZeros(const RCP<Matrix> & A,
-                      const magnitudeType tolerance,
-                      const bool keepDiagonal,
-                      const size_t expectedNNZperRow) {
-    Level fineLevel;
-    fineLevel.SetFactoryManager(null);
-    fineLevel.SetLevelID(0);
-    fineLevel.Set("A",A);
-    fineLevel.setlib(A->getDomainMap()->lib());
-    RCP<ThresholdAFilterFactory> ThreshFact = rcp(new ThresholdAFilterFactory("A", tolerance, keepDiagonal, expectedNNZperRow));
-    fineLevel.Request("A",ThreshFact.get());
-    ThreshFact->Build(fineLevel);
-    return fineLevel.Get< RCP<Matrix> >("A",ThreshFact.get());
+  MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCedgesLocal, BCedges_);
+  MueLu_sumAll(SM_Matrix_->getRowMap()->getComm(), BCnodesLocal, BCnodes_);
+
+  allEdgesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCedges_) >= D0_Matrix_->getRangeMap()->getGlobalNumElements();
+  allNodesBoundary_ = Teuchos::as<Xpetra::global_size_t>(BCnodes_) >= D0_Matrix_->getDomainMap()->getGlobalNumElements();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    removeExplicitZeros(const RCP<Matrix>& A,
+                        const magnitudeType tolerance,
+                        const bool keepDiagonal,
+                        const size_t expectedNNZperRow) {
+  Level fineLevel;
+  fineLevel.SetFactoryManager(null);
+  fineLevel.SetLevelID(0);
+  fineLevel.Set("A", A);
+  fineLevel.setlib(A->getDomainMap()->lib());
+  RCP<ThresholdAFilterFactory> ThreshFact = rcp(new ThresholdAFilterFactory("A", tolerance, keepDiagonal, expectedNNZperRow));
+  fineLevel.Request("A", ThreshFact.get());
+  ThreshFact->Build(fineLevel);
+  return fineLevel.Get<RCP<Matrix> >("A", ThreshFact.get());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::removeExplicitZeros(Teuchos::ParameterList& parameterList_,
+                                                                                   RCP<Matrix>& D0_Matrix_,
+                                                                                   RCP<Matrix>& SM_Matrix_,
+                                                                                   RCP<Matrix>& M1_Matrix_,
+                                                                                   RCP<Matrix>& Ms_Matrix_) {
+  bool defaultFilter = false;
+
+  // Remove zero entries from D0 if necessary.
+  // In the construction of the prolongator we use the graph of the
+  // matrix, so zero entries mess it up.
+  if (parameterList_.get<bool>("refmaxwell: filter D0", true) && D0_Matrix_->getLocalMaxNumRowEntries() > 2) {
+    D0_Matrix_ = removeExplicitZeros(D0_Matrix_, 1e-8, false, 2);
+
+    // If D0 has too many zeros, maybe SM and M1 do as well.
+    defaultFilter = true;
   }
 
+  if (!M1_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter M1", defaultFilter)) {
+    RCP<Vector> diag = VectorFactory::Build(M1_Matrix_->getRowMap());
+    // find a reasonable absolute value threshold
+    M1_Matrix_->getLocalDiagCopy(*diag);
+    magnitudeType threshold = 1.0e-8 * diag->normInf();
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::removeExplicitZeros(Teuchos::ParameterList &parameterList_,
-                                                                                  RCP<Matrix> & D0_Matrix_,
-                                                                                  RCP<Matrix> & SM_Matrix_,
-                                                                                  RCP<Matrix> & M1_Matrix_,
-                                                                                  RCP<Matrix> & Ms_Matrix_) {
-
-    bool defaultFilter = false;
-
-    // Remove zero entries from D0 if necessary.
-    // In the construction of the prolongator we use the graph of the
-    // matrix, so zero entries mess it up.
-    if (parameterList_.get<bool>("refmaxwell: filter D0", true) && D0_Matrix_->getLocalMaxNumRowEntries()>2) {
-      D0_Matrix_ = removeExplicitZeros(D0_Matrix_, 1e-8, false, 2);
-
-      // If D0 has too many zeros, maybe SM and M1 do as well.
-      defaultFilter = true;
-    }
-
-    if (!M1_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter M1", defaultFilter)) {
-      RCP<Vector> diag = VectorFactory::Build(M1_Matrix_->getRowMap());
-      // find a reasonable absolute value threshold
-      M1_Matrix_->getLocalDiagCopy(*diag);
-      magnitudeType threshold = 1.0e-8 * diag->normInf();
-
-      M1_Matrix_ = removeExplicitZeros(M1_Matrix_, threshold, true);
-    }
-
-    if (!Ms_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter Ms", defaultFilter)) {
-      RCP<Vector> diag = VectorFactory::Build(Ms_Matrix_->getRowMap());
-      // find a reasonable absolute value threshold
-      Ms_Matrix_->getLocalDiagCopy(*diag);
-      magnitudeType threshold = 1.0e-8 * diag->normInf();
-
-      Ms_Matrix_ = removeExplicitZeros(Ms_Matrix_, threshold, true);
-    }
-
-    if (!SM_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter SM", defaultFilter)) {
-      RCP<Vector> diag = VectorFactory::Build(SM_Matrix_->getRowMap());
-      // find a reasonable absolute value threshold
-      SM_Matrix_->getLocalDiagCopy(*diag);
-      magnitudeType threshold = 1.0e-8 * diag->normInf();
-
-      SM_Matrix_ = removeExplicitZeros(SM_Matrix_, threshold, true);
-    }
-
+    M1_Matrix_ = removeExplicitZeros(M1_Matrix_, threshold, true);
   }
 
+  if (!Ms_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter Ms", defaultFilter)) {
+    RCP<Vector> diag = VectorFactory::Build(Ms_Matrix_->getRowMap());
+    // find a reasonable absolute value threshold
+    Ms_Matrix_->getLocalDiagCopy(*diag);
+    magnitudeType threshold = 1.0e-8 * diag->normInf();
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  thresholdedAbs(const RCP<Matrix> & A,
-                 const magnitudeType threshold) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_Scalar = typename ATS::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_Scalar>;
-    using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-
-    const impl_Scalar impl_SC_ONE = impl_ATS::one();
-    const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
-
-    {
-      auto lclMat = A->getLocalMatrixDevice();
-      Kokkos::parallel_for("thresholdedAbs",
-                           range_type(0,lclMat.nnz()),
-                           KOKKOS_LAMBDA(const size_t i) {
-                             if (impl_ATS::magnitude(lclMat.values(i)) >threshold)
-                               lclMat.values(i) = impl_SC_ONE;
-                             else
-                               lclMat.values(i) = impl_SC_ZERO;
-                           });
-    }
+    Ms_Matrix_ = removeExplicitZeros(Ms_Matrix_, threshold, true);
   }
 
+  if (!SM_Matrix_.is_null() && parameterList_.get<bool>("refmaxwell: filter SM", defaultFilter)) {
+    RCP<Vector> diag = VectorFactory::Build(SM_Matrix_->getRowMap());
+    // find a reasonable absolute value threshold
+    SM_Matrix_->getLocalDiagCopy(*diag);
+    magnitudeType threshold = 1.0e-8 * diag->normInf();
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  setMatvecParams(Matrix& A, RCP<ParameterList> matvecParams) {
-    RCP<const Import> xpImporter = A.getCrsGraph()->getImporter();
-    if (!xpImporter.is_null())
-      xpImporter->setDistributorParameters(matvecParams);
-    RCP<const Export> xpExporter = A.getCrsGraph()->getExporter();
-    if (!xpExporter.is_null())
-      xpExporter->setDistributorParameters(matvecParams);
+    SM_Matrix_ = removeExplicitZeros(SM_Matrix_, threshold, true);
   }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  Maxwell_Utils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  PtAPWrapper(const RCP<Matrix>& A,const RCP<Matrix>& P, ParameterList &params, std::string & label) {
-    Level fineLevel, coarseLevel;
-    fineLevel.SetFactoryManager(null);
-    coarseLevel.SetFactoryManager(null);
-    coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
-    fineLevel.SetLevelID(0);
-    coarseLevel.SetLevelID(1);
-    fineLevel.Set("A",A);
-    coarseLevel.Set("P",P);
-    coarseLevel.setlib(A->getDomainMap()->lib());
-    fineLevel.setlib(A->getDomainMap()->lib());
-    coarseLevel.setObjectLabel(label);
-    fineLevel.setObjectLabel(label);
-
-    RCP<RAPFactory> rapFact = rcp(new RAPFactory());
-    ParameterList rapList = *(rapFact->GetValidParameterList());
-    rapList.set("transpose: use implicit", true);
-    rapList.set("rap: fix zero diagonals", params.get<bool>("rap: fix zero diagonals", true));
-    rapList.set("rap: fix zero diagonals threshold", params.get<double>("rap: fix zero diagonals threshold", Teuchos::ScalarTraits<double>::eps()));
-    rapList.set("rap: triple product", params.get<bool>("rap: triple product", false));
-    rapFact->SetParameterList(rapList);
-
-    coarseLevel.Request("A", rapFact.get());
-    
-    return coarseLevel.Get< RCP<Matrix> >("A", rapFact.get());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    thresholdedAbs(const RCP<Matrix>& A,
+                   const magnitudeType threshold) {
+  using ATS         = Kokkos::ArithTraits<Scalar>;
+  using impl_Scalar = typename ATS::val_type;
+  using impl_ATS    = Kokkos::ArithTraits<impl_Scalar>;
+  using range_type  = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+
+  const impl_Scalar impl_SC_ONE  = impl_ATS::one();
+  const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
+
+  {
+    auto lclMat = A->getLocalMatrixDevice();
+    Kokkos::parallel_for(
+        "thresholdedAbs",
+        range_type(0, lclMat.nnz()),
+        KOKKOS_LAMBDA(const size_t i) {
+          if (impl_ATS::magnitude(lclMat.values(i)) > threshold)
+            lclMat.values(i) = impl_SC_ONE;
+          else
+            lclMat.values(i) = impl_SC_ZERO;
+        });
   }
-
-
-
-
-
-
-} // namespace
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    setMatvecParams(Matrix& A, RCP<ParameterList> matvecParams) {
+  RCP<const Import> xpImporter = A.getCrsGraph()->getImporter();
+  if (!xpImporter.is_null())
+    xpImporter->setDistributorParameters(matvecParams);
+  RCP<const Export> xpExporter = A.getCrsGraph()->getExporter();
+  if (!xpExporter.is_null())
+    xpExporter->setDistributorParameters(matvecParams);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+Maxwell_Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    PtAPWrapper(const RCP<Matrix>& A, const RCP<Matrix>& P, ParameterList& params, std::string& label) {
+  Level fineLevel, coarseLevel;
+  fineLevel.SetFactoryManager(null);
+  coarseLevel.SetFactoryManager(null);
+  coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
+  fineLevel.SetLevelID(0);
+  coarseLevel.SetLevelID(1);
+  fineLevel.Set("A", A);
+  coarseLevel.Set("P", P);
+  coarseLevel.setlib(A->getDomainMap()->lib());
+  fineLevel.setlib(A->getDomainMap()->lib());
+  coarseLevel.setObjectLabel(label);
+  fineLevel.setObjectLabel(label);
+
+  RCP<RAPFactory> rapFact = rcp(new RAPFactory());
+  ParameterList rapList   = *(rapFact->GetValidParameterList());
+  rapList.set("transpose: use implicit", true);
+  rapList.set("rap: fix zero diagonals", params.get<bool>("rap: fix zero diagonals", true));
+  rapList.set("rap: fix zero diagonals threshold", params.get<double>("rap: fix zero diagonals threshold", Teuchos::ScalarTraits<double>::eps()));
+  rapList.set("rap: triple product", params.get<bool>("rap: triple product", false));
+  rapFact->SetParameterList(rapList);
+
+  coarseLevel.Request("A", rapFact.get());
+
+  return coarseLevel.Get<RCP<Matrix> >("A", rapFact.get());
+}
+
+}  // namespace MueLu
 
 #define MUELU_MAXWELL_UTILS_SHORT
-#endif //ifdef MUELU_MAXWELL_UTILS_DEF_HPP
+#endif  // ifdef MUELU_MAXWELL_UTILS_DEF_HPP
diff --git a/packages/muelu/src/Operators/MueLu_MultiPhys_decl.hpp b/packages/muelu/src/Operators/MueLu_MultiPhys_decl.hpp
index 588c627df317..e1d743e9d7f8 100644
--- a/packages/muelu/src/Operators/MueLu_MultiPhys_decl.hpp
+++ b/packages/muelu/src/Operators/MueLu_MultiPhys_decl.hpp
@@ -73,147 +73,144 @@
 
 namespace MueLu {
 
-  /*!
-    @brief Preconditioner (wrapped as a Xpetra::Operator) for solving MultiPhysics PDEs.
-
- 
-    @ingroup MueLuAdapters
-  */
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class MultiPhys : public VerboseObject, public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @brief Preconditioner (wrapped as a Xpetra::Operator) for solving MultiPhysics PDEs.
 
+
+  @ingroup MueLuAdapters
+*/
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class MultiPhys : public VerboseObject, public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_MULTIPHYS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
-    typedef typename Xpetra::MultiVector<coordinateType,LO,GO,NO> RealValuedMultiVector;
-
-    //! Constructor
-    MultiPhys() :
-      AmatMultiphysics_(Teuchos::null), hierarchyMultiphysics_(Teuchos::null), nBlks_(0)
-    {
-    }
-
-    /** Constructor
-     *
-     * \param[in] AmatMultiPhysics      Multiphysics discretization matrix
-     * \param[in] arrayOfAuxMatrices    Auxiliary matrices used to generate subblock prolongators for multiphysics system
-     * \param[in] arrayOfNullspaces     Nullspace multivectors used to generate subblock prolongators for multiphysics system
-     * \param[in] arrayOfCoords         Coordinate multivectors used to generate subblock prolongators for multiphysics system
-     * \param[in] nBlks                 nBlks x nBlks gives the block dimensions of the multiphysics operator
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    MultiPhys(const Teuchos::RCP<Matrix> &                                 AmatMultiPhysics,
-              const Teuchos::ArrayRCP<RCP<Matrix>>                         arrayOfAuxMatrices,
-              const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>           arrayOfNullspaces,
-              const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
-              const int nBlks,
-              Teuchos::ParameterList& List, 
-               bool ComputePrec = true):
-      AmatMultiphysics_(AmatMultiPhysics), arrayOfAuxMatrices_(arrayOfAuxMatrices), arrayOfNullspaces_(arrayOfNullspaces),
-      arrayOfCoords_(arrayOfCoords), nBlks_(nBlks)
-    {
-      initialize(AmatMultiPhysics, arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, nBlks, List);
-      compute(false);
-    }
-
-    //! Destructor.
-    virtual ~MultiPhys() {}
-
-    //! Returns the Xpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Map> getDomainMap() const;
-
-    //! Returns the Xpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Map> getRangeMap() const;
-
-    //! Set parameters
-    void setParameters(Teuchos::ParameterList& list);
-
-    //! Setup the preconditioner
-    void compute(bool reuse=false);
-
-    //! Reset system matrix
-    void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec=true);
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
-    //! \param[out] Y - MultiVector of dimension NumVectors containing result.
-    void apply (const MultiVector& X, MultiVector& Y,
-                Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
-
-    void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const MultiVector & X,
-                  const MultiVector & B,
-                  MultiVector & R) const {
-      using STS = Teuchos::ScalarTraits<Scalar>;
-      R.update(STS::one(),B,STS::zero());
-      this->apply (X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
-    }
-    
-
-  private:
-    /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
-     *
-     * \param[in] AmatMultiPhysics      Multiphysics discretization matrix
-     * \param[in] arrayOfAuxMatrices    Array of auxiliary matrices used to generate subblock prolongators for multiphysics system
-     * \param[in] arrayOfNullspaces     Array of nullspace  multivectors used to generate subblock prolongators for multiphysics system
-     * \param[in] arrayOfCoords         Array of coordinate multivectors used to generate subblock prolongators for multiphysics system
-     * \param[in] nBlks                 nBlks x nBlks gives the block dimensions of the multiphysics operator
-     * \param[in] List Parameter list
-     */
-    void initialize(const Teuchos::RCP<Matrix> &                                 AmatMultiPhysics,
-                    const Teuchos::ArrayRCP<RCP<Matrix>>                         arrayOfAuxMatrices,
-                    const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>           arrayOfNullspaces,
-                    const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
-                    const int nBlks,
-                    Teuchos::ParameterList& List);
-
-
-    //! apply standard MultiPhys cycle
-    void applyInverse(const MultiVector& RHS, MultiVector& X) const;
-
-    //! allocate multivectors for solve
-    void allocateMemory(int numVectors) const;
-
-    //! get a (synced) timer
-    Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm=Teuchos::null) const;
-
-    //! ParameterLists
-    mutable Teuchos::ParameterList parameterList_;
-
-    //! Hierarchies: used to define P for (0,0)-block, .... (nBlks_-1,nBlks_-1) block
-
-       
-    Teuchos::RCP<Matrix> AmatMultiphysics_;    // multiphysics discretization matrix 
-    Teuchos::RCP<Teuchos::ParameterList> paramListMultiphysics_;// array of parameter lists directing MueLu's construct of subblock P operators
-    Teuchos::RCP<Hierarchy> hierarchyMultiphysics_;    // multiphysics discretization matrix 
-
-    Teuchos::ArrayRCP<Teuchos::RCP<Teuchos::ParameterList>> arrayOfParamLists_;  // array of parameter lists directing MueLu's construct of subblock P operators
-    Teuchos::ArrayRCP<Teuchos::RCP<Hierarchy>>              arrayOfHierarchies_; 
-    Teuchos::ArrayRCP<Teuchos::RCP<Matrix>>                 arrayOfAuxMatrices_; // array of discretization/auxiliary matrices used to generate subblock prolongators
-    Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>             arrayOfNullspaces_; // array of nullspaces for smoothed aggregation.
-    Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords_;       // array of coordinates for smoothed aggregation/rebalancing.
-
-    int  nBlks_;    // number of PDE sub-systems within multiphysics system
-    bool useKokkos_, enable_reuse_, syncTimers_;
-
-  };
-
-} // namespace
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
+  typedef typename Xpetra::MultiVector<coordinateType, LO, GO, NO> RealValuedMultiVector;
+
+  //! Constructor
+  MultiPhys()
+    : AmatMultiphysics_(Teuchos::null)
+    , hierarchyMultiphysics_(Teuchos::null)
+    , nBlks_(0) {
+  }
+
+  /** Constructor
+   *
+   * \param[in] AmatMultiPhysics      Multiphysics discretization matrix
+   * \param[in] arrayOfAuxMatrices    Auxiliary matrices used to generate subblock prolongators for multiphysics system
+   * \param[in] arrayOfNullspaces     Nullspace multivectors used to generate subblock prolongators for multiphysics system
+   * \param[in] arrayOfCoords         Coordinate multivectors used to generate subblock prolongators for multiphysics system
+   * \param[in] nBlks                 nBlks x nBlks gives the block dimensions of the multiphysics operator
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  MultiPhys(const Teuchos::RCP<Matrix>& AmatMultiPhysics,
+            const Teuchos::ArrayRCP<RCP<Matrix>> arrayOfAuxMatrices,
+            const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces,
+            const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
+            const int nBlks,
+            Teuchos::ParameterList& List,
+            bool ComputePrec = true)
+    : AmatMultiphysics_(AmatMultiPhysics)
+    , arrayOfAuxMatrices_(arrayOfAuxMatrices)
+    , arrayOfNullspaces_(arrayOfNullspaces)
+    , arrayOfCoords_(arrayOfCoords)
+    , nBlks_(nBlks) {
+    initialize(AmatMultiPhysics, arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, nBlks, List);
+    compute(false);
+  }
+
+  //! Destructor.
+  virtual ~MultiPhys() {}
+
+  //! Returns the Xpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Map> getDomainMap() const;
+
+  //! Returns the Xpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Map> getRangeMap() const;
+
+  //! Set parameters
+  void setParameters(Teuchos::ParameterList& list);
+
+  //! Setup the preconditioner
+  void compute(bool reuse = false);
+
+  //! Reset system matrix
+  void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec = true);
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
+  //! \param[out] Y - MultiVector of dimension NumVectors containing result.
+  void apply(const MultiVector& X, MultiVector& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
+
+  void describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const MultiVector& X,
+                const MultiVector& B,
+                MultiVector& R) const {
+    using STS = Teuchos::ScalarTraits<Scalar>;
+    R.update(STS::one(), B, STS::zero());
+    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+  }
+
+ private:
+  /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
+   *
+   * \param[in] AmatMultiPhysics      Multiphysics discretization matrix
+   * \param[in] arrayOfAuxMatrices    Array of auxiliary matrices used to generate subblock prolongators for multiphysics system
+   * \param[in] arrayOfNullspaces     Array of nullspace  multivectors used to generate subblock prolongators for multiphysics system
+   * \param[in] arrayOfCoords         Array of coordinate multivectors used to generate subblock prolongators for multiphysics system
+   * \param[in] nBlks                 nBlks x nBlks gives the block dimensions of the multiphysics operator
+   * \param[in] List Parameter list
+   */
+  void initialize(const Teuchos::RCP<Matrix>& AmatMultiPhysics,
+                  const Teuchos::ArrayRCP<RCP<Matrix>> arrayOfAuxMatrices,
+                  const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces,
+                  const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
+                  const int nBlks,
+                  Teuchos::ParameterList& List);
+
+  //! apply standard MultiPhys cycle
+  void applyInverse(const MultiVector& RHS, MultiVector& X) const;
+
+  //! allocate multivectors for solve
+  void allocateMemory(int numVectors) const;
+
+  //! get a (synced) timer
+  Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int>> comm = Teuchos::null) const;
+
+  //! ParameterLists
+  mutable Teuchos::ParameterList parameterList_;
+
+  //! Hierarchies: used to define P for (0,0)-block, .... (nBlks_-1,nBlks_-1) block
+
+  Teuchos::RCP<Matrix> AmatMultiphysics_;                       // multiphysics discretization matrix
+  Teuchos::RCP<Teuchos::ParameterList> paramListMultiphysics_;  // array of parameter lists directing MueLu's construct of subblock P operators
+  Teuchos::RCP<Hierarchy> hierarchyMultiphysics_;               // multiphysics discretization matrix
+
+  Teuchos::ArrayRCP<Teuchos::RCP<Teuchos::ParameterList>> arrayOfParamLists_;  // array of parameter lists directing MueLu's construct of subblock P operators
+  Teuchos::ArrayRCP<Teuchos::RCP<Hierarchy>> arrayOfHierarchies_;
+  Teuchos::ArrayRCP<Teuchos::RCP<Matrix>> arrayOfAuxMatrices_;            // array of discretization/auxiliary matrices used to generate subblock prolongators
+  Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces_;        // array of nullspaces for smoothed aggregation.
+  Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords_;  // array of coordinates for smoothed aggregation/rebalancing.
+
+  int nBlks_;  // number of PDE sub-systems within multiphysics system
+  bool useKokkos_, enable_reuse_, syncTimers_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_MULTIPHYS_SHORT
-#endif // MUELU_MULTIPHYS_DECL_HPP
+#endif  // MUELU_MULTIPHYS_DECL_HPP
diff --git a/packages/muelu/src/Operators/MueLu_MultiPhys_def.hpp b/packages/muelu/src/Operators/MueLu_MultiPhys_def.hpp
index 63c3a8acceff..c11ae1e5a4a2 100644
--- a/packages/muelu/src/Operators/MueLu_MultiPhys_def.hpp
+++ b/packages/muelu/src/Operators/MueLu_MultiPhys_def.hpp
@@ -65,7 +65,7 @@
 #include "MueLu_HierarchyManager.hpp"
 #include <MueLu_HierarchyUtils.hpp>
 #if defined(HAVE_MUELU_KOKKOS_REFACTOR)
-# include "MueLu_Utilities_kokkos.hpp"
+#include "MueLu_Utilities_kokkos.hpp"
 #endif
 #include "MueLu_VerbosityLevel.hpp"
 #include <MueLu_CreateXpetraPreconditioner.hpp>
@@ -77,262 +77,248 @@
 
 namespace MueLu {
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getDomainMap() const {
+  return AmatMultiphysics_->getDomainMap();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getDomainMap() const {
-    return AmatMultiphysics_->getDomainMap();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getRangeMap() const {
-    return AmatMultiphysics_->getRangeMap();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameters(Teuchos::ParameterList& list) {
-
-    // this operator only makes sense for the Combo when using TransP for R
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRangeMap() const {
+  return AmatMultiphysics_->getRangeMap();
+}
 
-    list.set("multigrid algorithm","combine");
-    list.set("combine: numBlks",nBlks_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameters(Teuchos::ParameterList& list) {
+  // this operator only makes sense for the Combo when using TransP for R
 
-    // Make sure verbosity gets passed to the sublists
-    std::string verbosity = list.get("verbosity","high");
-    VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosity));
+  list.set("multigrid algorithm", "combine");
+  list.set("combine: numBlks", nBlks_);
 
-    arrayOfParamLists_.resize(nBlks_);
-    for (int i = 0 ; i < nBlks_; i++) {
-      std::string listName = "subblockList" + Teuchos::toString(i);
-      if(list.isSublist(listName)) {
-        arrayOfParamLists_[i] = Teuchos::rcpFromRef(list.sublist(listName));
-      }
-      else
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must provide sublist " + listName);    
-
-      arrayOfParamLists_[i]->set("verbosity",arrayOfParamLists_[i]->get("verbosity",verbosity));
-      arrayOfParamLists_[i]->set("smoother: pre or post","none");
-      arrayOfParamLists_[i]->set("smoother: type", "none");
-    }
+  // Make sure verbosity gets passed to the sublists
+  std::string verbosity = list.get("verbosity", "high");
+  VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosity));
 
-    // Are we using Kokkos?
-    useKokkos_ = !Node::is_serial;
-    useKokkos_ = list.get("use kokkos refactor",useKokkos_);
+  arrayOfParamLists_.resize(nBlks_);
+  for (int i = 0; i < nBlks_; i++) {
+    std::string listName = "subblockList" + Teuchos::toString(i);
+    if (list.isSublist(listName)) {
+      arrayOfParamLists_[i] = Teuchos::rcpFromRef(list.sublist(listName));
+    } else
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Must provide sublist " + listName);
 
-    paramListMultiphysics_ = Teuchos::rcpFromRef(list); 
+    arrayOfParamLists_[i]->set("verbosity", arrayOfParamLists_[i]->get("verbosity", verbosity));
+    arrayOfParamLists_[i]->set("smoother: pre or post", "none");
+    arrayOfParamLists_[i]->set("smoother: type", "none");
   }
 
+  // Are we using Kokkos?
+  useKokkos_ = !Node::is_serial;
+  useKokkos_ = list.get("use kokkos refactor", useKokkos_);
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::compute(bool reuse) {
-    /* 
+  paramListMultiphysics_ = Teuchos::rcpFromRef(list);
+}
 
-       Create a set of AMG hierarchies whose interpolation matrices are used to build on combined 
-       AMG hierarchy for a multiphysics problem
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::compute(bool reuse) {
+  /*
 
-     */
+     Create a set of AMG hierarchies whose interpolation matrices are used to build on combined
+     AMG hierarchy for a multiphysics problem
 
-//#ifdef HAVE_MUELU_CUDA
-//   if (paramListMultiphysics_.get<bool>("multiphysics: cuda profile setup", false)) cudaProfilerStart();
-//#endif
+   */
 
-    std::string timerLabel;
-    if (reuse)
-      timerLabel = "MueLu MultiPhys: compute (reuse)";
-    else
-      timerLabel = "MueLu MultiPhys: compute";
-    RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
+  //#ifdef HAVE_MUELU_CUDA
+  //   if (paramListMultiphysics_.get<bool>("multiphysics: cuda profile setup", false)) cudaProfilerStart();
+  //#endif
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Generate the (iii,iii) Hierarchy
+  std::string timerLabel;
+  if (reuse)
+    timerLabel = "MueLu MultiPhys: compute (reuse)";
+  else
+    timerLabel = "MueLu MultiPhys: compute";
+  RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
 
-    for (int iii = 0; iii < nBlks_; iii++) {
-      arrayOfParamLists_[iii]->sublist("user data").set("Coordinates",arrayOfCoords_[iii]);
-    
-      bool wantToRepartition = false;
-      if (paramListMultiphysics_->isParameter("repartition: enable"))
-        wantToRepartition = paramListMultiphysics_->get<bool>("repartition: enable");
+  ////////////////////////////////////////////////////////////////////////////////
+  // Generate the (iii,iii) Hierarchy
 
-      arrayOfParamLists_[iii]->set("repartition: enable",wantToRepartition);
-      arrayOfParamLists_[iii]->set("repartition: rebalance P and R",true);
-      arrayOfParamLists_[iii]->set("repartition: explicit via new copy rebalance P and R",true);
+  for (int iii = 0; iii < nBlks_; iii++) {
+    arrayOfParamLists_[iii]->sublist("user data").set("Coordinates", arrayOfCoords_[iii]);
 
-      if (paramListMultiphysics_->isParameter("repartition: use subcommunicators"))
-        arrayOfParamLists_[iii]->set("repartition: use subcommunicators", paramListMultiphysics_->isParameter("repartition: use subcommunicators"));
-      else
-        arrayOfParamLists_[iii]->set("repartition: use subcommunicators", true);
-    }
-    // repartitioning should only happen when createing the individual P's , not 
-    // when combiing them 
-    
-    paramListMultiphysics_->set<bool>("repartition: enable", false);
-        
-
-    LO maxLevels = 9999;
-    for (int i = 0; i < nBlks_; i++) {
-      std::string operatorLabel = "MultiPhys (" + Teuchos::toString(i) + "," + Teuchos::toString(i) + ")"; arrayOfAuxMatrices_[i]->setObjectLabel(operatorLabel);
-      arrayOfHierarchies_[i] = MueLu::CreateXpetraPreconditioner(arrayOfAuxMatrices_[i],*arrayOfParamLists_[i]);
-      LO tempNlevels = arrayOfHierarchies_[i]->GetGlobalNumLevels();
-      if (tempNlevels < maxLevels) maxLevels = tempNlevels;
-    }
-
-    hierarchyMultiphysics_ = rcp(new Hierarchy("Combo"));
-    for (LO i = 0; i < maxLevels; i++) {
-      hierarchyMultiphysics_->AddNewLevel();
-    }
-    for (int i = 0; i < nBlks_; i++) {
-      std::string subblkName = "Psubblock" + Teuchos::toString(i);
-      MueLu::HierarchyUtils<SC,LO,GO,NO>::CopyBetweenHierarchies(*(arrayOfHierarchies_[i]),*(hierarchyMultiphysics_),"P",subblkName,"RCP<Matrix>");
-    }
-    paramListMultiphysics_->set("coarse: max size",1);
-    paramListMultiphysics_->set("max levels",maxLevels);
-
-    AmatMultiphysics_->setObjectLabel("A: block " + Teuchos::toString(nBlks_) + " x " + Teuchos::toString(nBlks_) + "multiphysics matrix");
-
-
-      // Rip off non-serializable data before validation
-      Teuchos::ParameterList nonSerialListMultiphysics, processedListMultiphysics;
-      MueLu::ExtractNonSerializableData(*paramListMultiphysics_,processedListMultiphysics, nonSerialListMultiphysics);
+    bool wantToRepartition = false;
+    if (paramListMultiphysics_->isParameter("repartition: enable"))
+      wantToRepartition = paramListMultiphysics_->get<bool>("repartition: enable");
 
-      // Rip off the subblock List stuff as we  don't need it any more and I think it messes up validator
-      
-      Teuchos::ParameterList stripped;
-      for (ParameterList::ConstIterator inListEntry = processedListMultiphysics.begin(); inListEntry != processedListMultiphysics.end(); inListEntry++) {
-        const std::string& levelName = inListEntry->first;
-        if (levelName.find("subblockList") != 0) stripped.setEntry(inListEntry->first, inListEntry->second);
-      }
-
-      RCP<HierarchyManager<SC,LO,GO,NO> > mueLuFactory = rcp(new ParameterListInterpreter<SC,LO,GO,NO>(stripped,AmatMultiphysics_->getDomainMap()->getComm()));
-      hierarchyMultiphysics_->setlib(AmatMultiphysics_->getDomainMap()->lib());
-      hierarchyMultiphysics_->SetProcRankVerbose(AmatMultiphysics_->getDomainMap()->getComm()->getRank());
-
-      // We don't need nullspace or coordinates, since we don't use them when just combining prolongators that have been already created
-      hierarchyMultiphysics_->GetLevel(0)->Set("A", AmatMultiphysics_);
-
-      // Stick the non-serializible data on the hierarchy.
-      // Not sure that we need this, since we don't use it in building the multiphysics hierarchy
-      HierarchyUtils<SC,LO,GO,NO>::AddNonSerializableDataToHierarchy(*mueLuFactory,*hierarchyMultiphysics_, nonSerialListMultiphysics);
-      mueLuFactory->SetupHierarchy(*hierarchyMultiphysics_);
-
-      describe(GetOStream(Runtime0));
-
-  }
+    arrayOfParamLists_[iii]->set("repartition: enable", wantToRepartition);
+    arrayOfParamLists_[iii]->set("repartition: rebalance P and R", true);
+    arrayOfParamLists_[iii]->set("repartition: explicit via new copy rebalance P and R", true);
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::TimeMonitor> MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm) const {
-    if (IsPrint(Timings)) {
-      if (!syncTimers_)
-        return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name)));
-      else {
-        if (comm.is_null())
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), AmatMultiphysics_->getRowMap()->getComm().ptr()));
-        else
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), comm.ptr()));
-      }
-    } else
-      return Teuchos::null;
+    if (paramListMultiphysics_->isParameter("repartition: use subcommunicators"))
+      arrayOfParamLists_[iii]->set("repartition: use subcommunicators", paramListMultiphysics_->isParameter("repartition: use subcommunicators"));
+    else
+      arrayOfParamLists_[iii]->set("repartition: use subcommunicators", true);
   }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::resetMatrix(RCP<Matrix> AmatMultiphysics_new, bool ComputePrec) {
-    bool reuse = !AmatMultiphysics_.is_null();
-    AmatMultiphysics_ = AmatMultiphysics_new;
-    if (ComputePrec) compute(reuse);
+  // repartitioning should only happen when createing the individual P's , not
+  // when combiing them
+
+  paramListMultiphysics_->set<bool>("repartition: enable", false);
+
+  LO maxLevels = 9999;
+  for (int i = 0; i < nBlks_; i++) {
+    std::string operatorLabel = "MultiPhys (" + Teuchos::toString(i) + "," + Teuchos::toString(i) + ")";
+    arrayOfAuxMatrices_[i]->setObjectLabel(operatorLabel);
+    arrayOfHierarchies_[i] = MueLu::CreateXpetraPreconditioner(arrayOfAuxMatrices_[i], *arrayOfParamLists_[i]);
+    LO tempNlevels         = arrayOfHierarchies_[i]->GetGlobalNumLevels();
+    if (tempNlevels < maxLevels) maxLevels = tempNlevels;
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyInverse(const MultiVector& RHS, MultiVector& X) const {
-    hierarchyMultiphysics_->Iterate(RHS,X,1,true);
+  hierarchyMultiphysics_ = rcp(new Hierarchy("Combo"));
+  for (LO i = 0; i < maxLevels; i++) {
+    hierarchyMultiphysics_->AddNewLevel();
   }
- 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply (const MultiVector& RHS, MultiVector& X,
-                                                                Teuchos::ETransp /* mode */,
-                                                                  Scalar /* alpha */,
-                                                                  Scalar /* beta */) const {
-    RCP<Teuchos::TimeMonitor> tm = getTimer("MueLu MultiPhys: solve");
-    hierarchyMultiphysics_->Iterate(RHS,X,1,true);
+  for (int i = 0; i < nBlks_; i++) {
+    std::string subblkName = "Psubblock" + Teuchos::toString(i);
+    MueLu::HierarchyUtils<SC, LO, GO, NO>::CopyBetweenHierarchies(*(arrayOfHierarchies_[i]), *(hierarchyMultiphysics_), "P", subblkName, "RCP<Matrix>");
   }
+  paramListMultiphysics_->set("coarse: max size", 1);
+  paramListMultiphysics_->set("max levels", maxLevels);
 
+  AmatMultiphysics_->setObjectLabel("A: block " + Teuchos::toString(nBlks_) + " x " + Teuchos::toString(nBlks_) + "multiphysics matrix");
 
- 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::hasTransposeApply() const {
-    return false;
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initialize(const Teuchos::RCP<Matrix> &                                 AmatMultiPhysics,
-             const Teuchos::ArrayRCP<RCP<Matrix>>                         arrayOfAuxMatrices,
-             const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>           arrayOfNullspaces,
-             const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
-             const int nBlks,
-             Teuchos::ParameterList& List)
-  {
-     arrayOfHierarchies_.resize(nBlks_);
-     for (int i = 0; i < nBlks_; i++) arrayOfHierarchies_[i] = Teuchos::null;
-
-
-    // Default settings
-    useKokkos_=false;
-    enable_reuse_=false;
-    syncTimers_=false;
-
-    
-    // set parameters
-    setParameters(List);
-
-  }   // initialize
-
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MultiPhys<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
-
-    std::ostringstream oss;
+  // Rip off non-serializable data before validation
+  Teuchos::ParameterList nonSerialListMultiphysics, processedListMultiphysics;
+  MueLu::ExtractNonSerializableData(*paramListMultiphysics_, processedListMultiphysics, nonSerialListMultiphysics);
 
-    RCP<const Teuchos::Comm<int> > comm = AmatMultiphysics_->getDomainMap()->getComm();
+  // Rip off the subblock List stuff as we  don't need it any more and I think it messes up validator
 
+  Teuchos::ParameterList stripped;
+  for (ParameterList::ConstIterator inListEntry = processedListMultiphysics.begin(); inListEntry != processedListMultiphysics.end(); inListEntry++) {
+    const std::string& levelName = inListEntry->first;
+    if (levelName.find("subblockList") != 0) stripped.setEntry(inListEntry->first, inListEntry->second);
+  }
 
-    oss << "\n--------------------------------------------------------------------------------\n" <<
-      "---                            MultiPhysics Summary                            ---\n"
-      "--------------------------------------------------------------------------------" << std::endl;
-    oss << std::endl;
-
-    GlobalOrdinal numRows;
-    GlobalOrdinal nnz;
-
-    AmatMultiphysics_->getRowMap()->getComm()->barrier();
-
-    for (int i = 0; i < nBlks_; i++) {
-      numRows = arrayOfAuxMatrices_[i]->getGlobalNumRows();
-      nnz = arrayOfAuxMatrices_[i]->getGlobalNumEntries();
-      Xpetra::global_size_t tt = numRows;
-      int rowspacer = 3; while (tt != 0) { tt /= 10; rowspacer++; }
-      tt = nnz;
-      int nnzspacer = 2; while (tt != 0) { tt /= 10; nnzspacer++; }
-
-      oss  << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
-      oss << "(" << Teuchos::toString(i) << ", " << Teuchos::toString(i) << ")" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  RCP<HierarchyManager<SC, LO, GO, NO>> mueLuFactory = rcp(new ParameterListInterpreter<SC, LO, GO, NO>(stripped, AmatMultiphysics_->getDomainMap()->getComm()));
+  hierarchyMultiphysics_->setlib(AmatMultiphysics_->getDomainMap()->lib());
+  hierarchyMultiphysics_->SetProcRankVerbose(AmatMultiphysics_->getDomainMap()->getComm()->getRank());
+
+  // We don't need nullspace or coordinates, since we don't use them when just combining prolongators that have been already created
+  hierarchyMultiphysics_->GetLevel(0)->Set("A", AmatMultiphysics_);
+
+  // Stick the non-serializible data on the hierarchy.
+  // Not sure that we need this, since we don't use it in building the multiphysics hierarchy
+  HierarchyUtils<SC, LO, GO, NO>::AddNonSerializableDataToHierarchy(*mueLuFactory, *hierarchyMultiphysics_, nonSerialListMultiphysics);
+  mueLuFactory->SetupHierarchy(*hierarchyMultiphysics_);
+
+  describe(GetOStream(Runtime0));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::TimeMonitor> MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int>> comm) const {
+  if (IsPrint(Timings)) {
+    if (!syncTimers_)
+      return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name)));
+    else {
+      if (comm.is_null())
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), AmatMultiphysics_->getRowMap()->getComm().ptr()));
+      else
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer(name), comm.ptr()));
     }
-    oss << std::endl;
-
-
-    for (int i = 0; i < nBlks_; i++) {
-       arrayOfHierarchies_[i]->describe(out, GetVerbLevel());
+  } else
+    return Teuchos::null;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::resetMatrix(RCP<Matrix> AmatMultiphysics_new, bool ComputePrec) {
+  bool reuse        = !AmatMultiphysics_.is_null();
+  AmatMultiphysics_ = AmatMultiphysics_new;
+  if (ComputePrec) compute(reuse);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::applyInverse(const MultiVector& RHS, MultiVector& X) const {
+  hierarchyMultiphysics_->Iterate(RHS, X, 1, true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const MultiVector& RHS, MultiVector& X,
+                                                                 Teuchos::ETransp /* mode */,
+                                                                 Scalar /* alpha */,
+                                                                 Scalar /* beta */) const {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("MueLu MultiPhys: solve");
+  hierarchyMultiphysics_->Iterate(RHS, X, 1, true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasTransposeApply() const {
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initialize(const Teuchos::RCP<Matrix>& AmatMultiPhysics,
+               const Teuchos::ArrayRCP<RCP<Matrix>> arrayOfAuxMatrices,
+               const Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces,
+               const Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords,
+               const int nBlks,
+               Teuchos::ParameterList& List) {
+  arrayOfHierarchies_.resize(nBlks_);
+  for (int i = 0; i < nBlks_; i++) arrayOfHierarchies_[i] = Teuchos::null;
+
+  // Default settings
+  useKokkos_    = false;
+  enable_reuse_ = false;
+  syncTimers_   = false;
+
+  // set parameters
+  setParameters(List);
+
+}  // initialize
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MultiPhys<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
+  std::ostringstream oss;
+
+  RCP<const Teuchos::Comm<int>> comm = AmatMultiphysics_->getDomainMap()->getComm();
+
+  oss << "\n--------------------------------------------------------------------------------\n"
+      << "---                            MultiPhysics Summary                            ---\n"
+         "--------------------------------------------------------------------------------"
+      << std::endl;
+  oss << std::endl;
+
+  GlobalOrdinal numRows;
+  GlobalOrdinal nnz;
+
+  AmatMultiphysics_->getRowMap()->getComm()->barrier();
+
+  for (int i = 0; i < nBlks_; i++) {
+    numRows                  = arrayOfAuxMatrices_[i]->getGlobalNumRows();
+    nnz                      = arrayOfAuxMatrices_[i]->getGlobalNumEntries();
+    Xpetra::global_size_t tt = numRows;
+    int rowspacer            = 3;
+    while (tt != 0) {
+      tt /= 10;
+      rowspacer++;
+    }
+    tt            = nnz;
+    int nnzspacer = 2;
+    while (tt != 0) {
+      tt /= 10;
+      nnzspacer++;
     }
 
+    oss << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
+    oss << "(" << Teuchos::toString(i) << ", " << Teuchos::toString(i) << ")" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  }
+  oss << std::endl;
 
+  for (int i = 0; i < nBlks_; i++) {
+    arrayOfHierarchies_[i]->describe(out, GetVerbLevel());
+  }
 
-  } // describe
-
-
+}  // describe
 
-} // namespace
+}  // namespace MueLu
 
 #define MUELU_MULTIPHYS_SHORT
-#endif //ifdef MUELU_MULTIPHYS_DEF_HPP
+#endif  // ifdef MUELU_MULTIPHYS_DEF_HPP
diff --git a/packages/muelu/src/Operators/MueLu_RefMaxwell_decl.hpp b/packages/muelu/src/Operators/MueLu_RefMaxwell_decl.hpp
index 536ee781eb2a..1b3aa8054ff1 100644
--- a/packages/muelu/src/Operators/MueLu_RefMaxwell_decl.hpp
+++ b/packages/muelu/src/Operators/MueLu_RefMaxwell_decl.hpp
@@ -92,620 +92,606 @@
 
 namespace MueLu {
 
-  /*!
-    @brief Preconditioner (wrapped as a Xpetra::Operator) for Maxwell's equations in curl-curl form.
+/*!
+  @brief Preconditioner (wrapped as a Xpetra::Operator) for Maxwell's equations in curl-curl form.
 
-    Let
+  Let
 
-      M_k(gamma)
+    M_k(gamma)
 
-    be the mass matrix with a weight coefficient gamma on the k-th
-    space in the deRham complex
+  be the mass matrix with a weight coefficient gamma on the k-th
+  space in the deRham complex
 
-      k=0     ---> k=1     ----> k=2    ----> k=3
+    k=0     ---> k=1     ----> k=2    ----> k=3
 
-      H(grad) ---> H(curl) ----> H(div) ----> L^2
+    H(grad) ---> H(curl) ----> H(div) ----> L^2
 
-              D_0          D_1          D_2
+            D_0          D_1          D_2
 
-              grad         curl         div
+            grad         curl         div
 
-    and let
+  and let
 
-      D_k
+    D_k
 
-    be the discretized derivative from the k-th to the (k+1)-th space.
+  be the discretized derivative from the k-th to the (k+1)-th space.
 
-    For example, in 3D, D_0 = grad, D_1 = curl and D_2 = div.
+  For example, in 3D, D_0 = grad, D_1 = curl and D_2 = div.
 
-    We want to solve the system
+  We want to solve the system
 
-      S e_k = b
+    S e_k = b
 
-    where
+  where
 
-      S = M_k(alpha) + D_k^T * M_{k+1}(beta) * D_k.
+    S = M_k(alpha) + D_k^T * M_{k+1}(beta) * D_k.
 
-    In 3D and for k=1, S corresponds to a discretization of
+  In 3D and for k=1, S corresponds to a discretization of
 
-      alpha * identity + beta * curl curl.
+    alpha * identity + beta * curl curl.
 
-    In 3D and for k=2, S corresponds to
+  In 3D and for k=2, S corresponds to
 
-      alpha * identity + beta * grad div.
+    alpha * identity + beta * grad div.
 
-    We precondition S by constructing a block diagonal preconditioner
-    for the equivalent 2x2 block system
+  We precondition S by constructing a block diagonal preconditioner
+  for the equivalent 2x2 block system
 
-      e_k = a_k + D_{k-1}*p_{k-1}
+    e_k = a_k + D_{k-1}*p_{k-1}
 
-      ( A11                      M_k(alpha) * D_{k-1} ) ( a_k     )   ( b             )
-      (                                               } (         ) = (               )
-      ( D_{k-1}^T * M_k(alpha)   A22                  ) ( p_{k-1} )   ( D_{k-1}^T * b )
+    ( A11                      M_k(alpha) * D_{k-1} ) ( a_k     )   ( b             )
+    (                                               } (         ) = (               )
+    ( D_{k-1}^T * M_k(alpha)   A22                  ) ( p_{k-1} )   ( D_{k-1}^T * b )
 
-    where
+  where
 
-      A11     = S                        + addon11
-      A22     = D_{k-1} ^T * S * D_{k-1} + addon22
-      addon11 = M_k(1)     * D_{k-1} * M_{k-1}(1/beta)^-1  * D_{k-1}^T * M_k(1)
-      addon22 = M_{k-1}(1) * D_{k-2} * M_{k-2}(1/alpha)^-1 * D_{k-2}^T * M_{k-1}(1)
+    A11     = S                        + addon11
+    A22     = D_{k-1} ^T * S * D_{k-1} + addon22
+    addon11 = M_k(1)     * D_{k-1} * M_{k-1}(1/beta)^-1  * D_{k-1}^T * M_k(1)
+    addon22 = M_{k-1}(1) * D_{k-2} * M_{k-2}(1/alpha)^-1 * D_{k-2}^T * M_{k-1}(1)
 
-    We note that due to D_k * D_{k-1} = 0 we have that
+  We note that due to D_k * D_{k-1} = 0 we have that
 
-      A22 = D_{k-1} ^T * M_k(alpha) * D_{k-1} + addon22
+    A22 = D_{k-1} ^T * M_k(alpha) * D_{k-1} + addon22
 
-    The addon terms mimic the term that would be needed to augment to
-    a Hodge Laplacian. In practice it seems that the addon terms are
-    rarely strictly required. For k=1 (Maxwell) the addon22 term (in
-    H(grad)) is always zero.
+  The addon terms mimic the term that would be needed to augment to
+  a Hodge Laplacian. In practice it seems that the addon terms are
+  rarely strictly required. For k=1 (Maxwell) the addon22 term (in
+  H(grad)) is always zero.
 
-    We cannot directly apply AMG to A11 and A22 in case they are not
-    expressed in terms of nodal bases. Let Pi_k be the projection from
-    the first order vectorial nodal finite element space into the k-th
-    space. We will apply AMG to
+  We cannot directly apply AMG to A11 and A22 in case they are not
+  expressed in terms of nodal bases. Let Pi_k be the projection from
+  the first order vectorial nodal finite element space into the k-th
+  space. We will apply AMG to
 
-      Pi_k^T     * A11 * Pi_k
-      Pi_{k-1}^T * A22 * Pi_{k-1}
+    Pi_k^T     * A11 * Pi_k
+    Pi_{k-1}^T * A22 * Pi_{k-1}
 
-    respectively. For k=1 A22 is already given in a nodal
-    discretization and this step is omitted.
+  respectively. For k=1 A22 is already given in a nodal
+  discretization and this step is omitted.
 
-    It can be beneficial to directly coarsen the projected diagonal
-    blocks and skip any smoothing on them.
+  It can be beneficial to directly coarsen the projected diagonal
+  blocks and skip any smoothing on them.
 
-    To obtain prolongators for the vectorial nodal problems, we
-    construct the auxiliary operators
+  To obtain prolongators for the vectorial nodal problems, we
+  construct the auxiliary operators
 
-      A11_nodal = D_{0}^T * M_1(beta)  * D_{0}
-      A22_nodal = D_{0}^T * M_1(alpha) * D_{0}
+    A11_nodal = D_{0}^T * M_1(beta)  * D_{0}
+    A22_nodal = D_{0}^T * M_1(alpha) * D_{0}
 
-    then constuct typical nodal scalar HGrad prolongators.
+  then constuct typical nodal scalar HGrad prolongators.
 
-    We then replicate them into prolongators for vectorial H(grad)
-    problem. Finally, we multiply from the left with the projection
-    Pi_k onto the k-th FE space.
+  We then replicate them into prolongators for vectorial H(grad)
+  problem. Finally, we multiply from the left with the projection
+  Pi_k onto the k-th FE space.
 
-    When k=1 this only needs to be done for the A11 block, as the A22
-    block is already given wrt a scalar nodal discretization.
+  When k=1 this only needs to be done for the A11 block, as the A22
+  block is already given wrt a scalar nodal discretization.
 
-    The following input matrices need to be supplied by the user:
-    - S
-    - D_{k-1}
+  The following input matrices need to be supplied by the user:
+  - S
+  - D_{k-1}
 
-    If addon11 is used we need:
-    - M_{k-1}(1/beta)^-1   - typically, this is the inverse of the lumped M_{k-1}(1/beta), not the actual inverse
-    - M_k(1)
+  If addon11 is used we need:
+  - M_{k-1}(1/beta)^-1   - typically, this is the inverse of the lumped M_{k-1}(1/beta), not the actual inverse
+  - M_k(1)
 
-    If addon22 is used we need:
-    - M_{k-2}(1/alpha)^-1  - typically, this is the inverse of the lumped M_{k-2}(1/alpha), not the actual inverse
-    - M_{k-1}(1)
+  If addon22 is used we need:
+  - M_{k-2}(1/alpha)^-1  - typically, this is the inverse of the lumped M_{k-2}(1/alpha), not the actual inverse
+  - M_{k-1}(1)
 
-    To construct the special prolongators we need
-    - D_{0}
-    - M_1(beta)
-    - M_1(alpha)
-    If these mass matrices are not given, but M_1(1) is, then that matrix can be used instead.
-    Alternatively, A11_nodal and A22_nodal can be passed directly.
+  To construct the special prolongators we need
+  - D_{0}
+  - M_1(beta)
+  - M_1(alpha)
+  If these mass matrices are not given, but M_1(1) is, then that matrix can be used instead.
+  Alternatively, A11_nodal and A22_nodal can be passed directly.
 
 
-    We are using the following variable names:
+  We are using the following variable names:
 
-      | variable                         | matrix                | note
-      |----------------------------------|-----------------------|----------------------
-      | <tt>SM               </tt>       | S                     |
-      | <tt>Dk_1             </tt>       | D_{k-1}               | same as D0 for k=1
-      | <tt>Dk_2             </tt>       | D_{k-2}               |
-      | <tt>D0               </tt>       | D_0                   | same as Dk_1 for k=1
-      | <tt>Mk_one           </tt>       | M_k(1)                |
-      | <tt>Mk_1_one         </tt>       | M_{k-1}(1)            |
-      | <tt>M1_beta          </tt>       | M_1(beta)             |
-      | <tt>M1_alpha         </tt>       | M_1(alpha)            |
-      | <tt>invMk_1_invBeta  </tt>       | M_{k-1}(1/beta)^{-1}  |
-      | <tt>invMk_2_invAlpha </tt>       | M_{k-2}(1/alpha)^{-1} |
+    | variable                         | matrix                | note
+    |----------------------------------|-----------------------|----------------------
+    | <tt>SM               </tt>       | S                     |
+    | <tt>Dk_1             </tt>       | D_{k-1}               | same as D0 for k=1
+    | <tt>Dk_2             </tt>       | D_{k-2}               |
+    | <tt>D0               </tt>       | D_0                   | same as Dk_1 for k=1
+    | <tt>Mk_one           </tt>       | M_k(1)                |
+    | <tt>Mk_1_one         </tt>       | M_{k-1}(1)            |
+    | <tt>M1_beta          </tt>       | M_1(beta)             |
+    | <tt>M1_alpha         </tt>       | M_1(alpha)            |
+    | <tt>invMk_1_invBeta  </tt>       | M_{k-1}(1/beta)^{-1}  |
+    | <tt>invMk_2_invAlpha </tt>       | M_{k-2}(1/alpha)^{-1} |
 
-    For backwards compatibility the interfaces also allow
+  For backwards compatibility the interfaces also allow
 
-      | variable          | matrix           | note
-      |-------------------|------------------|---------------------------------
-      | <tt>Ms    </tt>   | M_1(beta)        | alias for M1_beta
-      | <tt>M1    </tt>   | M_1(1)           | alias for Mk_one when k=1
-      | <tt>M0inv </tt>   | M_0(1/beta)      | alias for Mk_1_invBeta when k=1
+    | variable          | matrix           | note
+    |-------------------|------------------|---------------------------------
+    | <tt>Ms    </tt>   | M_1(beta)        | alias for M1_beta
+    | <tt>M1    </tt>   | M_1(1)           | alias for Mk_one when k=1
+    | <tt>M0inv </tt>   | M_0(1/beta)      | alias for Mk_1_invBeta when k=1
 
 
-    Reference:
-    P. Bochev, J. Hu, C. Siefert, and R. Tuminaro. "An algebraic multigrid approach based on
-    a compatible gauge reformulation of Maxwell's equations." SIAM Journal on Scientific
-    Computing, 31(1), 557-583.
-
-    Parameter list options:
-    - <tt>refmaxwell: mode</tt> - a <tt>string</tt> specifying the order of solve of the block system.
-                                  Allowed values are: "additive" (default), "121", "212", "1", "2"
-    - <tt>refmaxwell: disable addon</tt> - <tt>bool</tt> specifing whether the addon should be built for stabilization.
-                                           Default: "true"
-    - <tt>refmaxwell: use as preconditioner</tt> - <tt>bool</tt> specifing whether RefMaxwell is used as a preconditioner or as a solver.
-    - <tt>refmaxwell: dump matrices</tt> - <tt>bool</tt> specifing whether the matrices should be dumped.
-                                           Default: "false"
-    - <tt>refmaxwell: 11list</tt> and <tt>refmaxwell: 22list</tt> - parameter list for the multigrid hierarchies on 11 and 22 blocks
-    - <tt>refmaxwell: subsolves on subcommunicators</tt> - <tt>bool</tt> redistribute the two subsolves to disjoint sub-communicators (so that the additive solve can occur in parallel)
-                                           Default: "false"
-
-    @ingroup MueLuAdapters
-  */
-  template <class Scalar,
-            class LocalOrdinal,
-            class GlobalOrdinal,
-            class Node>
-  class RefMaxwell : public VerboseObject, public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-
+  Reference:
+  P. Bochev, J. Hu, C. Siefert, and R. Tuminaro. "An algebraic multigrid approach based on
+  a compatible gauge reformulation of Maxwell's equations." SIAM Journal on Scientific
+  Computing, 31(1), 557-583.
+
+  Parameter list options:
+  - <tt>refmaxwell: mode</tt> - a <tt>string</tt> specifying the order of solve of the block system.
+                                Allowed values are: "additive" (default), "121", "212", "1", "2"
+  - <tt>refmaxwell: disable addon</tt> - <tt>bool</tt> specifing whether the addon should be built for stabilization.
+                                         Default: "true"
+  - <tt>refmaxwell: use as preconditioner</tt> - <tt>bool</tt> specifing whether RefMaxwell is used as a preconditioner or as a solver.
+  - <tt>refmaxwell: dump matrices</tt> - <tt>bool</tt> specifing whether the matrices should be dumped.
+                                         Default: "false"
+  - <tt>refmaxwell: 11list</tt> and <tt>refmaxwell: 22list</tt> - parameter list for the multigrid hierarchies on 11 and 22 blocks
+  - <tt>refmaxwell: subsolves on subcommunicators</tt> - <tt>bool</tt> redistribute the two subsolves to disjoint sub-communicators (so that the additive solve can occur in parallel)
+                                         Default: "false"
+
+  @ingroup MueLuAdapters
+*/
+template <class Scalar,
+          class LocalOrdinal,
+          class GlobalOrdinal,
+          class Node>
+class RefMaxwell : public VerboseObject, public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_REFMAXWELL_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
-    typedef typename Xpetra::MultiVector<coordinateType,LO,GO,NO> RealValuedMultiVector;
-
-    //! Constructor
-    RefMaxwell() :
-      HierarchyCoarse11_(Teuchos::null),
-      Hierarchy22_(Teuchos::null),
-      disable_addon_(MasterList::getDefault<bool>("refmaxwell: disable addon")),
-      mode_(MasterList::getDefault<std::string>("refmaxwell: mode"))
-    {
-    }
-
-    //! Constructor with Hierarchies
-    RefMaxwell(Teuchos::RCP<Hierarchy> HH, Teuchos::RCP<Hierarchy> H22) :
-      HierarchyCoarse11_(HH),
-      Hierarchy22_(H22),
-      disable_addon_(MasterList::getDefault<bool>("refmaxwell: disable addon")),
-      mode_(MasterList::getDefault<std::string>("refmaxwell: mode"))
-    {
-    }
-
-    RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & Dk_1,
-               const Teuchos::RCP<Matrix> & Dk_2,
-               const Teuchos::RCP<Matrix> & D0,
-               const Teuchos::RCP<Matrix> & M1_beta,
-               const Teuchos::RCP<Matrix> & M1_alpha,
-               const Teuchos::RCP<Matrix> & Mk_one,
-               const Teuchos::RCP<Matrix> & Mk_1_one,
-               const Teuchos::RCP<Matrix> & invMk_1_invBeta,
-               const Teuchos::RCP<Matrix> & invMk_2_invAlpha,
-               const Teuchos::RCP<MultiVector> & Nullspace11,
-               const Teuchos::RCP<MultiVector> & Nullspace22,
-               const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true)
-    {
-      int spaceNumber  = List.get<int>("refmaxwell: space number", 1);
-      initialize(spaceNumber,
-                 Dk_1, Dk_2, D0,
-                 M1_beta, M1_alpha,
-                 Mk_one, Mk_1_one,
-                 invMk_1_invBeta, invMk_2_invAlpha,
-                 Nullspace11, Nullspace22, NodalCoords,
-                 List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Constructor with Jacobian (with add on)
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Ms_Matrix Edge mass matrix for the nodal aggregates
-     * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
-     * \param[in] M1_Matrix Edge mass matrix for the add on
-     * \param[in] Nullspace11 Null space (needed for periodic)
-     * \param[in] NodalCoords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & Ms_Matrix,
-               const Teuchos::RCP<Matrix> & M0inv_Matrix,
-               const Teuchos::RCP<Matrix> & M1_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace11,
-               const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true)
-    {
-      initialize(D0_Matrix,Ms_Matrix,M0inv_Matrix,M1_Matrix,Nullspace11,NodalCoords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Constructor with Jacobian (with add on)
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
-     * \param[in] M1_Matrix Edge mass matrix for the
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & M0inv_Matrix,
-               const Teuchos::RCP<Matrix> & M1_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace11,
-               const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true)
-    {
-      initialize(D0_Matrix,M1_Matrix,M0inv_Matrix,M1_Matrix,Nullspace11,NodalCoords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Constructor without Jacobian (with add on)
-     *
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
-     * \param[in] M1_Matrix Edge mass matrix for the
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & M0inv_Matrix,
-               const Teuchos::RCP<Matrix> & M1_Matrix,
-               const Teuchos::RCP<MultiVector> & Nullspace11,
-               const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-               Teuchos::ParameterList& List) : SM_Matrix_(Teuchos::null)
-    {
-      initialize(D0_Matrix,M1_Matrix,M0inv_Matrix,M1_Matrix,Nullspace11,NodalCoords,List);
-    }
-
-    /** Constructor with Jacobian (no add on)
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] M1_Matrix Edge mass matrix for the
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-               const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & M1_Matrix,
-               const Teuchos::RCP<MultiVector>  & Nullspace11,
-               const Teuchos::RCP<RealValuedMultiVector>  & NodalCoords,
-               Teuchos::ParameterList& List,
-               bool ComputePrec)
-    {
-      initialize(D0_Matrix,M1_Matrix,Teuchos::null,M1_Matrix,Nullspace11,NodalCoords,List);
-      resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-    /** Constructor without Jacobian (no add on)
-     *
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] M1_Matrix Edge mass matrix for the
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & D0_Matrix,
-               const Teuchos::RCP<Matrix> & M1_Matrix,
-               const Teuchos::RCP<MultiVector>  & Nullspace11,
-               const Teuchos::RCP<RealValuedMultiVector>  & NodalCoords,
-               Teuchos::ParameterList& List) : SM_Matrix_(Teuchos::null)
-    {
-      initialize(D0_Matrix,M1_Matrix,Teuchos::null,M1_Matrix,Nullspace11,NodalCoords,List);
-    }
-
-    /** Constructor with parameter list
-     *
-     * \param[in] SM_Matrix Jacobian
-     * \param[in] List Parameter list
-     * \param[in] ComputePrec If true, compute the preconditioner immediately
-     */
-    RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-               Teuchos::ParameterList& List,
-               bool ComputePrec = true);
-
-    //! Destructor.
-    virtual ~RefMaxwell() {}
-
-    //! Returns the Xpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Map> getDomainMap() const;
-
-    //! Returns the Xpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Map> getRangeMap() const;
-
-    //! Returns Jacobian matrix SM
-    const Teuchos::RCP<Matrix> & getJacobian() const {
-      return SM_Matrix_;
-    }
-
-    //! Set parameters
-    void setParameters(Teuchos::ParameterList& list);
-
-    //! Setup the preconditioner
-    void compute(bool reuse=false);
-
-    //! Reset system matrix
-    void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec=true);
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
-    //! \param[out] Y - MultiVector of dimension NumVectors containing result.
-    void apply (const MultiVector& X, MultiVector& Y,
-                Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(),
-                Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const;
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const;
-
-    void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const MultiVector & X,
-                  const MultiVector & B,
-                  MultiVector & R) const {
-      using STS = Teuchos::ScalarTraits<Scalar>;
-      R.update(STS::one(),B,STS::zero());
-      this->apply (X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
-    }
-    
-
-  private:
-
-    Teuchos::RCP<Teuchos::ParameterList> getValidParamterList();
-
-    /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
-     *
-     * Note: This uses old notation that only makes sense for curl-curl problems.
-     *
-     * \param[in] D0_Matrix Discrete Gradient
-     * \param[in] Ms_Matrix Edge mass matrix for nodal aggregates
-     * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
-     * \param[in] M1_Matrix Edge mass matrix for add on
-     * \param[in] Nullspace Null space (needed for periodic)
-     * \param[in] Coords Nodal coordinates
-     * \param[in] List Parameter list
-     */
-    void initialize(const Teuchos::RCP<Matrix> & D0_Matrix,
-                    const Teuchos::RCP<Matrix> & Ms_Matrix,
-                    const Teuchos::RCP<Matrix> & M0inv_Matrix,
-                    const Teuchos::RCP<Matrix> & M1_Matrix,
-                    const Teuchos::RCP<MultiVector> & Nullspace11,
-                    const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-                    Teuchos::ParameterList& List);
-
-    /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
-     *
-     * \param[in] k number of the space in the deRham sequence of the problem to be solved
-     * \param[in] Dk_1 Discrete derivative from (k-1)-th to k-th space
-     * \param[in] Dk_2 Discrete derivative from (k-2)-th to (k-1)-th space
-     * \param[in] D0 Discrete Gradient
-     * \param[in] M1_beta Mass matrix on 1-st space with weight beta for nodal aggregates
-     * \param[in] M1_alpha Mass matrix on 1-st space with weight alpha for nodal aggregates
-     * \param[in] Mk_one Mass matrix on k-th space with unit weight for addon11
-     * \param[in] Mk_1_one Mass matrix on (k-1)-th space with unit weight for addon22
-     * \param[in] invMk_1_invBeta Approximate inverse of mass matrix on (k-1)-th space with weight 1/beta (addon11 only)
-     * \param[in] invMk_2_invAlpha Approximate inverse of mass matrix on (k-2)-th space with weight 1/alpha (addon22 only)
-     * \param[in] Nullspace11 Null space (needed for periodic)
-     * \param[in] Nullspace22 Null space (needed for periodic)
-     * \param[in] NodalCoords Nodal coordinates
-     * \param[in] List Parameter list
-     */
-    void initialize(const int k,
-                    const Teuchos::RCP<Matrix> & Dk_1,
-                    const Teuchos::RCP<Matrix> & Dk_2,
-                    const Teuchos::RCP<Matrix> & D0,
-                    const Teuchos::RCP<Matrix> & M1_beta,
-                    const Teuchos::RCP<Matrix> & M1_alpha,
-                    const Teuchos::RCP<Matrix> & Mk_one,
-                    const Teuchos::RCP<Matrix> & Mk_1_one,
-                    const Teuchos::RCP<Matrix> & invMk_1_invBeta,
-                    const Teuchos::RCP<Matrix> & invMk_2_invAlpha,
-                    const Teuchos::RCP<MultiVector> & Nullspace11,
-                    const Teuchos::RCP<MultiVector> & Nullspace22,
-                    const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-                    Teuchos::ParameterList& List);
-
-    //! Determine how large the sub-communicators for the two hierarchies should be
-    void determineSubHierarchyCommSizes(bool &doRebalancing, int &rebalanceStriding, int &numProcsCoarseA11, int &numProcsA22);
-
-    //! Compute coarseA11 = P11^{T}*SM*P11 + addon efficiently
-    void buildCoarse11Matrix();
-
-    //! rebalance the coarse A11 matrix, as well as P11, CoordsCoarse11 and Addon11
-    void rebalanceCoarse11Matrix(const int rebalanceStriding, const int numProcsCoarseA11);
-
-    //! Setup A22 = D0^T SM D0 and rebalance it, as well as D0 and Coords_
-    void build22Matrix(const bool reuse, const bool doRebalancing, const int rebalanceStriding, const int numProcsA22);
-
-  public: // due to Cuda build errors otherwise
-
-    //! Builds a nullspace
-    RCP<MultiVector> buildNullspace(const int spaceNumber, const Kokkos::View<bool*, typename Node::device_type> &bcs, const bool applyBCs);
-
-    //! Builds a projection from a vector values space into a vector valued nodal space
-    Teuchos::RCP<Matrix> buildProjection(const int spaceNumber, const RCP<MultiVector> &EdgeNullspace) const;
-
-    /** Setup an auxiliary nodal prolongator
-     *
-     * \param[in]  A_nodal
-     * \param[out] P_nodal
-     * \param[out] Nullspace_nodal
-     */
-    void buildNodalProlongator(const Teuchos::RCP<Matrix> &A_nodal,
-                               Teuchos::RCP<Matrix> &P_nodal,
-                               Teuchos::RCP<MultiVector> &Nullspace_nodal,
-                               Teuchos::RCP<RealValuedMultiVector> &Coords_nodal) const;
-
-    /** Setup a vectorial nodal prolongator
-     *
-     * \param[in]  P_nodal_scalar
-     * \param[out] P_nodal_vector
-     */
-    RCP<Matrix> buildVectorNodalProlongator(const Teuchos::RCP<Matrix> &P_nodal) const;
-
-    /** Setup a special prolongator from vectorial nodal to edge or face discretization
-     *
-     * \param[in]  spaceNumber the type of target discretization
-     * \param[in]  A_nodal_Matrix used for the nodal aggregation
-     * \param[in]  EdgeNullspace edge nullspace
-     * \param[out] edgeProlongator edge prolongator
-     * \param[out] coarseEdgeNullspace coarse edge nullspace
-     * \param[out] coarseNodalCoords coarse nodal coordinates
-     */
-    void buildProlongator(const int spaceNumber,
-                          const Teuchos::RCP<Matrix> &A_nodal_Matrix,
-                          const RCP<MultiVector> &EdgeNullspace,
-                          Teuchos::RCP<Matrix> &edgeProlongator,
-                          Teuchos::RCP<MultiVector> &coarseEdgeNullspace,
-                          Teuchos::RCP<RealValuedMultiVector> &coarseNodalCoords) const;
-
-    /** Construct an addon matrix
-     *
-     * \param[in]  spaceNumber the type of target discretization
-     */
-    RCP<Matrix> buildAddon(const int spaceNumber);
-
-  private:
-
-    //! Setup a subsolve
-    void setupSubSolve(Teuchos::RCP<Hierarchy> &hierarchy,
-                       Teuchos::RCP<Operator> &thyraPrecOp,
-                       const Teuchos::RCP<Matrix> &A,
-                       const Teuchos::RCP<MultiVector> &Nullspace,
-                       const Teuchos::RCP<RealValuedMultiVector> &Coords,
-                       Teuchos::ParameterList &params,
-                       std::string &label,
-                       const bool reuse,
-                       const bool isSingular=false);
-
-    //! Set the fine level smoother
-    void setFineLevelSmoother11();
-
-    //! apply additive algorithm for 2x2 solve
-    void applyInverseAdditive(const MultiVector& RHS, MultiVector& X) const;
-
-    //! apply solve to 1-1 block only
-    void solveH(const MultiVector& RHS, MultiVector& X) const;
-
-    //! apply solve to 2-2 block only
-    void solve22(const MultiVector& RHS, MultiVector& X) const;
-
-    //! allocate multivectors for solve
-    void allocateMemory(int numVectors) const;
-
-    //! dump out matrix
-    void dump(const RCP<Matrix>& A, std::string name) const;
-
-    //! dump out multivector
-    void dump(const RCP<MultiVector>& X, std::string name) const;
-
-    //! dump out real-valued multivector
-    void dumpCoords(const RCP<RealValuedMultiVector>& X, std::string name) const;
-
-    //! dump out boolean ArrayView
-    void dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const;
-
-    //! dump out boolean Kokkos::View
-    void dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const;
-
-    //! get a (synced) timer
-    Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm=Teuchos::null) const;
-
-
-    //! Two hierarchies: one for the coarse (1,1)-block, another for the (2,2)-block
-    Teuchos::RCP<Hierarchy> HierarchyCoarse11_, Hierarchy22_;
-    Teuchos::RCP<SmootherBase> PreSmoother11_, PostSmoother11_;
-    Teuchos::RCP<SmootherPrototype> PreSmootherData11_, PostSmootherData11_;
-    RCP<Operator> thyraPrecOpH_, thyraPrecOp22_;
-    //! The number of the space in the deRham complex
-    int spaceNumber_;
-    //! The name of the solver
-    std::string solverName_;
-    //! The spatial dimension
-    size_t dim_;
-    //! The system that is getting preconditioned
-    Teuchos::RCP<Matrix> SM_Matrix_;
-    //! D_{k-1} matrix and its transpose
-    Teuchos::RCP<Matrix> Dk_1_, Dk_1_T_;
-    //! D_{k-2} matrix
-    Teuchos::RCP<Matrix> Dk_2_;
-    //! D_0 matrix
-    Teuchos::RCP<Matrix> D0_;
-    //! inverse of mass matrices on (k-1)-th and (k-2)-th space with weights 1/beta and 1/alpha respectively
-    Teuchos::RCP<Matrix> invMk_1_invBeta_, invMk_2_invAlpha_;
-    //! mass matrices with unit weight on k-th and (k-1)-th spaces
-    Teuchos::RCP<Matrix> Mk_one_, Mk_1_one_;
-    //! mass matrices on first space with weights beta and alpha respectively
-    Teuchos::RCP<Matrix> M1_beta_, M1_alpha_;
-    //! special prolongator for 11 block and its transpose
-    Teuchos::RCP<Matrix> P11_, R11_;
-    //! special prolongator for 22 block and its transpose
-    Teuchos::RCP<Matrix> P22_, R22_;
-    //! coarse 11, 22 and coarse 22 blocks
-    Teuchos::RCP<Matrix> coarseA11_, A22_, coarseA22_;
-    //! the addon for the 11 block
-    Teuchos::RCP<Matrix> Addon11_;
-    //! the addon for the 22 block
-    Teuchos::RCP<Matrix> Addon22_;
-    Teuchos::RCP<const Map> DorigDomainMap_;
-    Teuchos::RCP<const Import> DorigImporter_;
-    //! Vectors for BCs
-    Kokkos::View<bool*, typename Node::device_type> BCrows11_, BCcols22_, BCdomain22_;
-    int globalNumberBoundaryUnknowns11_, globalNumberBoundaryUnknowns22_;
-    //! Nullspace for (1.1) block
-    Teuchos::RCP<MultiVector> Nullspace11_, Nullspace22_;
-    //! Coordinates
-    Teuchos::RCP<RealValuedMultiVector> NodalCoords_, CoordsCoarse11_, Coords22_;
-    //! Nullspace for coarse (1,1) problem
-    Teuchos::RCP<MultiVector> NullspaceCoarse11_;
-    //! Nullspace for coarse (2,2) problem
-    Teuchos::RCP<MultiVector> CoarseNullspace22_;
-    //! Importer to coarse (1,1) hierarchy
-    Teuchos::RCP<const Import> ImporterCoarse11_, Importer22_;
-    bool Dk_1_T_R11_colMapsMatch_;
-    bool onlyBoundary11_, onlyBoundary22_;
-    //! Parameter lists
-    Teuchos::ParameterList parameterList_, precList11_, precList22_;
-    Teuchos::RCP<Teuchos::ParameterList> coarseA11_AP_reuse_data_, coarseA11_RAP_reuse_data_;
-    Teuchos::RCP<Teuchos::ParameterList> A22_AP_reuse_data_, A22_RAP_reuse_data_;
-    //! Some options
-    bool disable_addon_, disable_addon_22_, dump_matrices_, useKokkos_, use_as_preconditioner_, implicitTranspose_, fuseProlongationAndUpdate_, syncTimers_, enable_reuse_, skipFirst11Level_, skipFirst22Level_;
-    bool applyBCsToAnodal_, applyBCsToCoarse11_, applyBCsTo22_;
-    int numItersCoarse11_, numIters22_;
-    std::string mode_;
-
-    //! Temporary memory
-    mutable Teuchos::RCP<MultiVector> P11res_, P11x_, P11resSubComm_, P11xSubComm_, DresIntermediate_, Dres_, DxIntermediate_, Dx_, DresSubComm_, DxSubComm_, residual_, P11resTmp_, DresTmp_, DTR11Tmp_;
-  };
-
-
-} // namespace
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinateType;
+  typedef typename Xpetra::MultiVector<coordinateType, LO, GO, NO> RealValuedMultiVector;
+
+  //! Constructor
+  RefMaxwell()
+    : HierarchyCoarse11_(Teuchos::null)
+    , Hierarchy22_(Teuchos::null)
+    , disable_addon_(MasterList::getDefault<bool>("refmaxwell: disable addon"))
+    , mode_(MasterList::getDefault<std::string>("refmaxwell: mode")) {
+  }
+
+  //! Constructor with Hierarchies
+  RefMaxwell(Teuchos::RCP<Hierarchy> HH, Teuchos::RCP<Hierarchy> H22)
+    : HierarchyCoarse11_(HH)
+    , Hierarchy22_(H22)
+    , disable_addon_(MasterList::getDefault<bool>("refmaxwell: disable addon"))
+    , mode_(MasterList::getDefault<std::string>("refmaxwell: mode")) {
+  }
+
+  RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+             const Teuchos::RCP<Matrix> &Dk_1,
+             const Teuchos::RCP<Matrix> &Dk_2,
+             const Teuchos::RCP<Matrix> &D0,
+             const Teuchos::RCP<Matrix> &M1_beta,
+             const Teuchos::RCP<Matrix> &M1_alpha,
+             const Teuchos::RCP<Matrix> &Mk_one,
+             const Teuchos::RCP<Matrix> &Mk_1_one,
+             const Teuchos::RCP<Matrix> &invMk_1_invBeta,
+             const Teuchos::RCP<Matrix> &invMk_2_invAlpha,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<MultiVector> &Nullspace22,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List,
+             bool ComputePrec = true) {
+    int spaceNumber = List.get<int>("refmaxwell: space number", 1);
+    initialize(spaceNumber,
+               Dk_1, Dk_2, D0,
+               M1_beta, M1_alpha,
+               Mk_one, Mk_1_one,
+               invMk_1_invBeta, invMk_2_invAlpha,
+               Nullspace11, Nullspace22, NodalCoords,
+               List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Constructor with Jacobian (with add on)
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Ms_Matrix Edge mass matrix for the nodal aggregates
+   * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
+   * \param[in] M1_Matrix Edge mass matrix for the add on
+   * \param[in] Nullspace11 Null space (needed for periodic)
+   * \param[in] NodalCoords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+             const Teuchos::RCP<Matrix> &D0_Matrix,
+             const Teuchos::RCP<Matrix> &Ms_Matrix,
+             const Teuchos::RCP<Matrix> &M0inv_Matrix,
+             const Teuchos::RCP<Matrix> &M1_Matrix,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List,
+             bool ComputePrec = true) {
+    initialize(D0_Matrix, Ms_Matrix, M0inv_Matrix, M1_Matrix, Nullspace11, NodalCoords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Constructor with Jacobian (with add on)
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
+   * \param[in] M1_Matrix Edge mass matrix for the
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+             const Teuchos::RCP<Matrix> &D0_Matrix,
+             const Teuchos::RCP<Matrix> &M0inv_Matrix,
+             const Teuchos::RCP<Matrix> &M1_Matrix,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List,
+             bool ComputePrec = true) {
+    initialize(D0_Matrix, M1_Matrix, M0inv_Matrix, M1_Matrix, Nullspace11, NodalCoords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Constructor without Jacobian (with add on)
+   *
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
+   * \param[in] M1_Matrix Edge mass matrix for the
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &D0_Matrix,
+             const Teuchos::RCP<Matrix> &M0inv_Matrix,
+             const Teuchos::RCP<Matrix> &M1_Matrix,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List)
+    : SM_Matrix_(Teuchos::null) {
+    initialize(D0_Matrix, M1_Matrix, M0inv_Matrix, M1_Matrix, Nullspace11, NodalCoords, List);
+  }
+
+  /** Constructor with Jacobian (no add on)
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] M1_Matrix Edge mass matrix for the
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+             const Teuchos::RCP<Matrix> &D0_Matrix,
+             const Teuchos::RCP<Matrix> &M1_Matrix,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List,
+             bool ComputePrec) {
+    initialize(D0_Matrix, M1_Matrix, Teuchos::null, M1_Matrix, Nullspace11, NodalCoords, List);
+    resetMatrix(SM_Matrix, ComputePrec);
+  }
+
+  /** Constructor without Jacobian (no add on)
+   *
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] M1_Matrix Edge mass matrix for the
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &D0_Matrix,
+             const Teuchos::RCP<Matrix> &M1_Matrix,
+             const Teuchos::RCP<MultiVector> &Nullspace11,
+             const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+             Teuchos::ParameterList &List)
+    : SM_Matrix_(Teuchos::null) {
+    initialize(D0_Matrix, M1_Matrix, Teuchos::null, M1_Matrix, Nullspace11, NodalCoords, List);
+  }
+
+  /** Constructor with parameter list
+   *
+   * \param[in] SM_Matrix Jacobian
+   * \param[in] List Parameter list
+   * \param[in] ComputePrec If true, compute the preconditioner immediately
+   */
+  RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+             Teuchos::ParameterList &List,
+             bool ComputePrec = true);
+
+  //! Destructor.
+  virtual ~RefMaxwell() {}
+
+  //! Returns the Xpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Map> getDomainMap() const;
+
+  //! Returns the Xpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Map> getRangeMap() const;
+
+  //! Returns Jacobian matrix SM
+  const Teuchos::RCP<Matrix> &getJacobian() const {
+    return SM_Matrix_;
+  }
+
+  //! Set parameters
+  void setParameters(Teuchos::ParameterList &list);
+
+  //! Setup the preconditioner
+  void compute(bool reuse = false);
+
+  //! Reset system matrix
+  void resetMatrix(Teuchos::RCP<Matrix> SM_Matrix_new, bool ComputePrec = true);
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  //! \param[in]  X - MultiVector of dimension NumVectors to multiply with matrix.
+  //! \param[out] Y - MultiVector of dimension NumVectors containing result.
+  void apply(const MultiVector &X, MultiVector &Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar alpha          = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar beta           = Teuchos::ScalarTraits<Scalar>::zero()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const;
+
+  void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel = Teuchos::VERB_HIGH) const;
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const MultiVector &X,
+                const MultiVector &B,
+                MultiVector &R) const {
+    using STS = Teuchos::ScalarTraits<Scalar>;
+    R.update(STS::one(), B, STS::zero());
+    this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+  }
+
+ private:
+  Teuchos::RCP<Teuchos::ParameterList> getValidParamterList();
+
+  /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
+   *
+   * Note: This uses old notation that only makes sense for curl-curl problems.
+   *
+   * \param[in] D0_Matrix Discrete Gradient
+   * \param[in] Ms_Matrix Edge mass matrix for nodal aggregates
+   * \param[in] M0inv_Matrix Inverse of lumped nodal mass matrix (add on only)
+   * \param[in] M1_Matrix Edge mass matrix for add on
+   * \param[in] Nullspace Null space (needed for periodic)
+   * \param[in] Coords Nodal coordinates
+   * \param[in] List Parameter list
+   */
+  void initialize(const Teuchos::RCP<Matrix> &D0_Matrix,
+                  const Teuchos::RCP<Matrix> &Ms_Matrix,
+                  const Teuchos::RCP<Matrix> &M0inv_Matrix,
+                  const Teuchos::RCP<Matrix> &M1_Matrix,
+                  const Teuchos::RCP<MultiVector> &Nullspace11,
+                  const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+                  Teuchos::ParameterList &List);
+
+  /** Initialize with matrices except the Jacobian (don't compute the preconditioner)
+   *
+   * \param[in] k number of the space in the deRham sequence of the problem to be solved
+   * \param[in] Dk_1 Discrete derivative from (k-1)-th to k-th space
+   * \param[in] Dk_2 Discrete derivative from (k-2)-th to (k-1)-th space
+   * \param[in] D0 Discrete Gradient
+   * \param[in] M1_beta Mass matrix on 1-st space with weight beta for nodal aggregates
+   * \param[in] M1_alpha Mass matrix on 1-st space with weight alpha for nodal aggregates
+   * \param[in] Mk_one Mass matrix on k-th space with unit weight for addon11
+   * \param[in] Mk_1_one Mass matrix on (k-1)-th space with unit weight for addon22
+   * \param[in] invMk_1_invBeta Approximate inverse of mass matrix on (k-1)-th space with weight 1/beta (addon11 only)
+   * \param[in] invMk_2_invAlpha Approximate inverse of mass matrix on (k-2)-th space with weight 1/alpha (addon22 only)
+   * \param[in] Nullspace11 Null space (needed for periodic)
+   * \param[in] Nullspace22 Null space (needed for periodic)
+   * \param[in] NodalCoords Nodal coordinates
+   * \param[in] List Parameter list
+   */
+  void initialize(const int k,
+                  const Teuchos::RCP<Matrix> &Dk_1,
+                  const Teuchos::RCP<Matrix> &Dk_2,
+                  const Teuchos::RCP<Matrix> &D0,
+                  const Teuchos::RCP<Matrix> &M1_beta,
+                  const Teuchos::RCP<Matrix> &M1_alpha,
+                  const Teuchos::RCP<Matrix> &Mk_one,
+                  const Teuchos::RCP<Matrix> &Mk_1_one,
+                  const Teuchos::RCP<Matrix> &invMk_1_invBeta,
+                  const Teuchos::RCP<Matrix> &invMk_2_invAlpha,
+                  const Teuchos::RCP<MultiVector> &Nullspace11,
+                  const Teuchos::RCP<MultiVector> &Nullspace22,
+                  const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+                  Teuchos::ParameterList &List);
+
+  //! Determine how large the sub-communicators for the two hierarchies should be
+  void determineSubHierarchyCommSizes(bool &doRebalancing, int &rebalanceStriding, int &numProcsCoarseA11, int &numProcsA22);
+
+  //! Compute coarseA11 = P11^{T}*SM*P11 + addon efficiently
+  void buildCoarse11Matrix();
+
+  //! rebalance the coarse A11 matrix, as well as P11, CoordsCoarse11 and Addon11
+  void rebalanceCoarse11Matrix(const int rebalanceStriding, const int numProcsCoarseA11);
+
+  //! Setup A22 = D0^T SM D0 and rebalance it, as well as D0 and Coords_
+  void build22Matrix(const bool reuse, const bool doRebalancing, const int rebalanceStriding, const int numProcsA22);
+
+ public:  // due to Cuda build errors otherwise
+  //! Builds a nullspace
+  RCP<MultiVector> buildNullspace(const int spaceNumber, const Kokkos::View<bool *, typename Node::device_type> &bcs, const bool applyBCs);
+
+  //! Builds a projection from a vector values space into a vector valued nodal space
+  Teuchos::RCP<Matrix> buildProjection(const int spaceNumber, const RCP<MultiVector> &EdgeNullspace) const;
+
+  /** Setup an auxiliary nodal prolongator
+   *
+   * \param[in]  A_nodal
+   * \param[out] P_nodal
+   * \param[out] Nullspace_nodal
+   */
+  void buildNodalProlongator(const Teuchos::RCP<Matrix> &A_nodal,
+                             Teuchos::RCP<Matrix> &P_nodal,
+                             Teuchos::RCP<MultiVector> &Nullspace_nodal,
+                             Teuchos::RCP<RealValuedMultiVector> &Coords_nodal) const;
+
+  /** Setup a vectorial nodal prolongator
+   *
+   * \param[in]  P_nodal_scalar
+   * \param[out] P_nodal_vector
+   */
+  RCP<Matrix> buildVectorNodalProlongator(const Teuchos::RCP<Matrix> &P_nodal) const;
+
+  /** Setup a special prolongator from vectorial nodal to edge or face discretization
+   *
+   * \param[in]  spaceNumber the type of target discretization
+   * \param[in]  A_nodal_Matrix used for the nodal aggregation
+   * \param[in]  EdgeNullspace edge nullspace
+   * \param[out] edgeProlongator edge prolongator
+   * \param[out] coarseEdgeNullspace coarse edge nullspace
+   * \param[out] coarseNodalCoords coarse nodal coordinates
+   */
+  void buildProlongator(const int spaceNumber,
+                        const Teuchos::RCP<Matrix> &A_nodal_Matrix,
+                        const RCP<MultiVector> &EdgeNullspace,
+                        Teuchos::RCP<Matrix> &edgeProlongator,
+                        Teuchos::RCP<MultiVector> &coarseEdgeNullspace,
+                        Teuchos::RCP<RealValuedMultiVector> &coarseNodalCoords) const;
+
+  /** Construct an addon matrix
+   *
+   * \param[in]  spaceNumber the type of target discretization
+   */
+  RCP<Matrix> buildAddon(const int spaceNumber);
+
+ private:
+  //! Setup a subsolve
+  void setupSubSolve(Teuchos::RCP<Hierarchy> &hierarchy,
+                     Teuchos::RCP<Operator> &thyraPrecOp,
+                     const Teuchos::RCP<Matrix> &A,
+                     const Teuchos::RCP<MultiVector> &Nullspace,
+                     const Teuchos::RCP<RealValuedMultiVector> &Coords,
+                     Teuchos::ParameterList &params,
+                     std::string &label,
+                     const bool reuse,
+                     const bool isSingular = false);
+
+  //! Set the fine level smoother
+  void setFineLevelSmoother11();
+
+  //! apply additive algorithm for 2x2 solve
+  void applyInverseAdditive(const MultiVector &RHS, MultiVector &X) const;
+
+  //! apply solve to 1-1 block only
+  void solveH(const MultiVector &RHS, MultiVector &X) const;
+
+  //! apply solve to 2-2 block only
+  void solve22(const MultiVector &RHS, MultiVector &X) const;
+
+  //! allocate multivectors for solve
+  void allocateMemory(int numVectors) const;
+
+  //! dump out matrix
+  void dump(const RCP<Matrix> &A, std::string name) const;
+
+  //! dump out multivector
+  void dump(const RCP<MultiVector> &X, std::string name) const;
+
+  //! dump out real-valued multivector
+  void dumpCoords(const RCP<RealValuedMultiVector> &X, std::string name) const;
+
+  //! dump out boolean ArrayView
+  void dump(const Teuchos::ArrayRCP<bool> &v, std::string name) const;
+
+  //! dump out boolean Kokkos::View
+  void dump(const Kokkos::View<bool *, typename Node::device_type> &v, std::string name) const;
+
+  //! get a (synced) timer
+  Teuchos::RCP<Teuchos::TimeMonitor> getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm = Teuchos::null) const;
+
+  //! Two hierarchies: one for the coarse (1,1)-block, another for the (2,2)-block
+  Teuchos::RCP<Hierarchy> HierarchyCoarse11_, Hierarchy22_;
+  Teuchos::RCP<SmootherBase> PreSmoother11_, PostSmoother11_;
+  Teuchos::RCP<SmootherPrototype> PreSmootherData11_, PostSmootherData11_;
+  RCP<Operator> thyraPrecOpH_, thyraPrecOp22_;
+  //! The number of the space in the deRham complex
+  int spaceNumber_;
+  //! The name of the solver
+  std::string solverName_;
+  //! The spatial dimension
+  size_t dim_;
+  //! The system that is getting preconditioned
+  Teuchos::RCP<Matrix> SM_Matrix_;
+  //! D_{k-1} matrix and its transpose
+  Teuchos::RCP<Matrix> Dk_1_, Dk_1_T_;
+  //! D_{k-2} matrix
+  Teuchos::RCP<Matrix> Dk_2_;
+  //! D_0 matrix
+  Teuchos::RCP<Matrix> D0_;
+  //! inverse of mass matrices on (k-1)-th and (k-2)-th space with weights 1/beta and 1/alpha respectively
+  Teuchos::RCP<Matrix> invMk_1_invBeta_, invMk_2_invAlpha_;
+  //! mass matrices with unit weight on k-th and (k-1)-th spaces
+  Teuchos::RCP<Matrix> Mk_one_, Mk_1_one_;
+  //! mass matrices on first space with weights beta and alpha respectively
+  Teuchos::RCP<Matrix> M1_beta_, M1_alpha_;
+  //! special prolongator for 11 block and its transpose
+  Teuchos::RCP<Matrix> P11_, R11_;
+  //! special prolongator for 22 block and its transpose
+  Teuchos::RCP<Matrix> P22_, R22_;
+  //! coarse 11, 22 and coarse 22 blocks
+  Teuchos::RCP<Matrix> coarseA11_, A22_, coarseA22_;
+  //! the addon for the 11 block
+  Teuchos::RCP<Matrix> Addon11_;
+  //! the addon for the 22 block
+  Teuchos::RCP<Matrix> Addon22_;
+  Teuchos::RCP<const Map> DorigDomainMap_;
+  Teuchos::RCP<const Import> DorigImporter_;
+  //! Vectors for BCs
+  Kokkos::View<bool *, typename Node::device_type> BCrows11_, BCcols22_, BCdomain22_;
+  int globalNumberBoundaryUnknowns11_, globalNumberBoundaryUnknowns22_;
+  //! Nullspace for (1.1) block
+  Teuchos::RCP<MultiVector> Nullspace11_, Nullspace22_;
+  //! Coordinates
+  Teuchos::RCP<RealValuedMultiVector> NodalCoords_, CoordsCoarse11_, Coords22_;
+  //! Nullspace for coarse (1,1) problem
+  Teuchos::RCP<MultiVector> NullspaceCoarse11_;
+  //! Nullspace for coarse (2,2) problem
+  Teuchos::RCP<MultiVector> CoarseNullspace22_;
+  //! Importer to coarse (1,1) hierarchy
+  Teuchos::RCP<const Import> ImporterCoarse11_, Importer22_;
+  bool Dk_1_T_R11_colMapsMatch_;
+  bool onlyBoundary11_, onlyBoundary22_;
+  //! Parameter lists
+  Teuchos::ParameterList parameterList_, precList11_, precList22_;
+  Teuchos::RCP<Teuchos::ParameterList> coarseA11_AP_reuse_data_, coarseA11_RAP_reuse_data_;
+  Teuchos::RCP<Teuchos::ParameterList> A22_AP_reuse_data_, A22_RAP_reuse_data_;
+  //! Some options
+  bool disable_addon_, disable_addon_22_, dump_matrices_, useKokkos_, use_as_preconditioner_, implicitTranspose_, fuseProlongationAndUpdate_, syncTimers_, enable_reuse_, skipFirst11Level_, skipFirst22Level_;
+  bool applyBCsToAnodal_, applyBCsToCoarse11_, applyBCsTo22_;
+  int numItersCoarse11_, numIters22_;
+  std::string mode_;
+
+  //! Temporary memory
+  mutable Teuchos::RCP<MultiVector> P11res_, P11x_, P11resSubComm_, P11xSubComm_, DresIntermediate_, Dres_, DxIntermediate_, Dx_, DresSubComm_, DxSubComm_, residual_, P11resTmp_, DresTmp_, DTR11Tmp_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_REFMAXWELL_SHORT
-#endif // MUELU_REFMAXWELL_DECL_HPP
+#endif  // MUELU_REFMAXWELL_DECL_HPP
diff --git a/packages/muelu/src/Operators/MueLu_RefMaxwell_def.hpp b/packages/muelu/src/Operators/MueLu_RefMaxwell_def.hpp
index 4f671c5c42c6..029f75dbc056 100644
--- a/packages/muelu/src/Operators/MueLu_RefMaxwell_def.hpp
+++ b/packages/muelu/src/Operators/MueLu_RefMaxwell_def.hpp
@@ -100,848 +100,943 @@
 #include <Xpetra_ThyraLinearOp.hpp>
 #endif
 
-
 namespace MueLu {
 
-  template<typename T>
-  T pop(Teuchos::ParameterList& pl, std::string const& name_in)
-  {
-    T result = pl.get<T>(name_in);
-    pl.remove(name_in, true);
-    return result;
+template <typename T>
+T pop(Teuchos::ParameterList &pl, std::string const &name_in) {
+  T result = pl.get<T>(name_in);
+  pl.remove(name_in, true);
+  return result;
+}
+
+template <typename T>
+T pop(Teuchos::ParameterList &pl, std::string const &name_in, T def_value) {
+  T result = pl.get<T>(name_in, def_value);
+  pl.remove(name_in, false);
+  return result;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+Matrix2CrsMatrix(Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &matrix) {
+  return Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(matrix, true)->getCrsMatrix();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getDomainMap() const {
+  return SM_Matrix_->getDomainMap();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getRangeMap() const {
+  return SM_Matrix_->getRangeMap();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::ParameterList>
+RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getValidParamterList() {
+  bool useKokkosDefault = !Node::is_serial;
+
+  RCP<ParameterList> params = rcp(new ParameterList("RefMaxwell"));
+
+  params->set<RCP<Matrix> >("Dk_1", Teuchos::null);
+  params->set<RCP<Matrix> >("Dk_2", Teuchos::null);
+  params->set<RCP<Matrix> >("D0", Teuchos::null);
+
+  params->set<RCP<Matrix> >("M1_beta", Teuchos::null);
+  params->set<RCP<Matrix> >("M1_alpha", Teuchos::null);
+  // for backwards compatiblity
+  params->set<RCP<Matrix> >("Ms", Teuchos::null);
+
+  params->set<RCP<Matrix> >("Mk_one", Teuchos::null);
+  params->set<RCP<Matrix> >("Mk_1_one", Teuchos::null);
+  // for backwards compatiblity
+  params->set<RCP<Matrix> >("M1", Teuchos::null);
+
+  params->set<RCP<Matrix> >("invMk_1_invBeta", Teuchos::null);
+  params->set<RCP<Matrix> >("invMk_2_invAlpha", Teuchos::null);
+  // for backwards compatiblity
+  params->set<RCP<Matrix> >("M0inv", Teuchos::null);
+
+  params->set<RCP<MultiVector> >("Nullspace", Teuchos::null);
+  params->set<RCP<RealValuedMultiVector> >("Coordinates", Teuchos::null);
+
+  auto spaceValidator = rcp(new Teuchos::EnhancedNumberValidator<int>(1, 2));
+  params->set("refmaxwell: space number", 1, "", spaceValidator);
+  params->set("verbosity", MasterList::getDefault<std::string>("verbosity"));
+  params->set("use kokkos refactor", useKokkosDefault);
+  params->set("half precision", false);
+  params->set("parameterlist: syntax", MasterList::getDefault<std::string>("parameterlist: syntax"));
+  params->set("output filename", MasterList::getDefault<std::string>("output filename"));
+  params->set("print initial parameters", MasterList::getDefault<bool>("print initial parameters"));
+  params->set("refmaxwell: disable addon", MasterList::getDefault<bool>("refmaxwell: disable addon"));
+  params->set("refmaxwell: disable addon 22", true);
+  params->set("refmaxwell: mode", MasterList::getDefault<std::string>("refmaxwell: mode"));
+  params->set("refmaxwell: use as preconditioner", MasterList::getDefault<bool>("refmaxwell: use as preconditioner"));
+  params->set("refmaxwell: dump matrices", MasterList::getDefault<bool>("refmaxwell: dump matrices"));
+  params->set("refmaxwell: enable reuse", MasterList::getDefault<bool>("refmaxwell: enable reuse"));
+  params->set("refmaxwell: skip first (1,1) level", MasterList::getDefault<bool>("refmaxwell: skip first (1,1) level"));
+  params->set("refmaxwell: skip first (2,2) level", false);
+  params->set("multigrid algorithm", "Unsmoothed");
+  params->set("transpose: use implicit", MasterList::getDefault<bool>("transpose: use implicit"));
+  params->set("rap: triple product", MasterList::getDefault<bool>("rap: triple product"));
+  params->set("rap: fix zero diagonals", true);
+  params->set("rap: fix zero diagonals threshold", MasterList::getDefault<double>("rap: fix zero diagonals threshold"));
+  params->set("fuse prolongation and update", MasterList::getDefault<bool>("fuse prolongation and update"));
+  params->set("refmaxwell: subsolves on subcommunicators", MasterList::getDefault<bool>("refmaxwell: subsolves on subcommunicators"));
+  params->set("refmaxwell: subsolves striding", 1);
+  params->set("refmaxwell: row sum drop tol (1,1)", MasterList::getDefault<double>("aggregation: row sum drop tol"));
+  params->set("sync timers", false);
+  params->set("refmaxwell: num iters coarse 11", 1);
+  params->set("refmaxwell: num iters 22", 1);
+  params->set("refmaxwell: apply BCs to Anodal", false);
+  params->set("refmaxwell: apply BCs to coarse 11", true);
+  params->set("refmaxwell: apply BCs to 22", true);
+  params->set("refmaxwell: max coarse size", 1);
+
+  ParameterList &precList11 = params->sublist("refmaxwell: 11list");
+  precList11.disableRecursiveValidation();
+  ParameterList &precList22 = params->sublist("refmaxwell: 22list");
+  precList22.disableRecursiveValidation();
+
+  params->set("smoother: type", "CHEBYSHEV");
+  ParameterList &smootherList = params->sublist("smoother: params");
+  smootherList.disableRecursiveValidation();
+  params->set("smoother: pre type", "NONE");
+  ParameterList &preSmootherList = params->sublist("smoother: pre params");
+  preSmootherList.disableRecursiveValidation();
+  params->set("smoother: post type", "NONE");
+  ParameterList &postSmootherList = params->sublist("smoother: post params");
+  postSmootherList.disableRecursiveValidation();
+
+  ParameterList &matvecParams = params->sublist("matvec params");
+  matvecParams.disableRecursiveValidation();
+
+  params->set("multigrid algorithm", "unsmoothed");
+  params->set("aggregation: type", MasterList::getDefault<std::string>("aggregation: type"));
+  params->set("aggregation: drop tol", MasterList::getDefault<double>("aggregation: drop tol"));
+  params->set("aggregation: drop scheme", MasterList::getDefault<std::string>("aggregation: drop scheme"));
+  params->set("aggregation: distance laplacian algo", MasterList::getDefault<std::string>("aggregation: distance laplacian algo"));
+  params->set("aggregation: min agg size", MasterList::getDefault<int>("aggregation: min agg size"));
+  params->set("aggregation: max agg size", MasterList::getDefault<int>("aggregation: max agg size"));
+  params->set("aggregation: match ML phase1", MasterList::getDefault<bool>("aggregation: match ML phase1"));
+  params->set("aggregation: match ML phase2a", MasterList::getDefault<bool>("aggregation: match ML phase2a"));
+  params->set("aggregation: match ML phase2b", MasterList::getDefault<bool>("aggregation: match ML phase2b"));
+  params->set("aggregation: export visualization data", MasterList::getDefault<bool>("aggregation: export visualization data"));
+
+  return params;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setParameters(Teuchos::ParameterList &list) {
+  if (list.isType<std::string>("parameterlist: syntax") && list.get<std::string>("parameterlist: syntax") == "ml") {
+    Teuchos::ParameterList newList;
+    {
+      Teuchos::ParameterList newList2                = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list, "refmaxwell"));
+      RCP<Teuchos::ParameterList> validateParameters = getValidParamterList();
+      for (auto it = newList2.begin(); it != newList2.end(); ++it) {
+        const std::string &entry_name = it->first;
+        if (validateParameters->isParameter(entry_name)) {
+          ParameterEntry theEntry = newList2.getEntry(entry_name);
+          newList.setEntry(entry_name, theEntry);
+        }
+      }
+    }
+
+    if (list.isSublist("refmaxwell: 11list") && list.sublist("refmaxwell: 11list").isSublist("edge matrix free: coarse"))
+      newList.sublist("refmaxwell: 11list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list.sublist("refmaxwell: 11list").sublist("edge matrix free: coarse"), "SA"));
+    if (list.isSublist("refmaxwell: 22list"))
+      newList.sublist("refmaxwell: 22list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list.sublist("refmaxwell: 22list"), "SA"));
+    list = newList;
   }
 
+  parameterList_ = list;
+  parameterList_.validateParametersAndSetDefaults(*getValidParamterList());
+  std::string verbosityLevel = parameterList_.get<std::string>("verbosity");
+  VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosityLevel));
+  std::string outputFilename = parameterList_.get<std::string>("output filename");
+  if (outputFilename != "")
+    VerboseObject::SetMueLuOFileStream(outputFilename);
+  if (parameterList_.isType<Teuchos::RCP<Teuchos::FancyOStream> >("output stream"))
+    VerboseObject::SetMueLuOStream(parameterList_.get<Teuchos::RCP<Teuchos::FancyOStream> >("output stream"));
+
+  if (parameterList_.get<bool>("print initial parameters"))
+    GetOStream(static_cast<MsgType>(Runtime1), 0) << parameterList_ << std::endl;
+  disable_addon_             = parameterList_.get<bool>("refmaxwell: disable addon");
+  disable_addon_22_          = parameterList_.get<bool>("refmaxwell: disable addon 22");
+  mode_                      = parameterList_.get<std::string>("refmaxwell: mode");
+  use_as_preconditioner_     = parameterList_.get<bool>("refmaxwell: use as preconditioner");
+  dump_matrices_             = parameterList_.get<bool>("refmaxwell: dump matrices");
+  enable_reuse_              = parameterList_.get<bool>("refmaxwell: enable reuse");
+  implicitTranspose_         = parameterList_.get<bool>("transpose: use implicit");
+  fuseProlongationAndUpdate_ = parameterList_.get<bool>("fuse prolongation and update");
+  skipFirst11Level_          = parameterList_.get<bool>("refmaxwell: skip first (1,1) level");
+  skipFirst22Level_          = parameterList_.get<bool>("refmaxwell: skip first (2,2) level");
+  if (spaceNumber_ == 1)
+    skipFirst22Level_ = false;
+  syncTimers_         = parameterList_.get<bool>("sync timers");
+  useKokkos_          = parameterList_.get<bool>("use kokkos refactor");
+  numItersCoarse11_   = parameterList_.get<int>("refmaxwell: num iters coarse 11");
+  numIters22_         = parameterList_.get<int>("refmaxwell: num iters 22");
+  applyBCsToAnodal_   = parameterList_.get<bool>("refmaxwell: apply BCs to Anodal");
+  applyBCsToCoarse11_ = parameterList_.get<bool>("refmaxwell: apply BCs to coarse 11");
+  applyBCsTo22_       = parameterList_.get<bool>("refmaxwell: apply BCs to 22");
+
+  precList11_ = parameterList_.sublist("refmaxwell: 11list");
+  if (!precList11_.isType<std::string>("Preconditioner Type") &&
+      !precList11_.isType<std::string>("smoother: type") &&
+      !precList11_.isType<std::string>("smoother: pre type") &&
+      !precList11_.isType<std::string>("smoother: post type")) {
+    precList11_.set("smoother: type", "CHEBYSHEV");
+    precList11_.sublist("smoother: params").set("chebyshev: degree", 2);
+    precList11_.sublist("smoother: params").set("chebyshev: ratio eigenvalue", 5.4);
+    precList11_.sublist("smoother: params").set("chebyshev: eigenvalue max iterations", 30);
+  }
 
-  template<typename T>
-  T pop(Teuchos::ParameterList& pl, std::string const& name_in, T def_value)
-  {
-    T result = pl.get<T>(name_in, def_value);
-    pl.remove(name_in, false);
-    return result;
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  Matrix2CrsMatrix(Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &matrix) {
-    return Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(matrix, true)->getCrsMatrix();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getDomainMap() const {
-    return SM_Matrix_->getDomainMap();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getRangeMap() const {
-    return SM_Matrix_->getRangeMap();
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::ParameterList>
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  getValidParamterList() {
-
-    bool useKokkosDefault = !Node::is_serial;
-
-    RCP<ParameterList> params = rcp(new ParameterList("RefMaxwell"));
-
-    params->set<RCP<Matrix> >("Dk_1", Teuchos::null);
-    params->set<RCP<Matrix> >("Dk_2", Teuchos::null);
-    params->set<RCP<Matrix> >("D0", Teuchos::null);
-
-    params->set<RCP<Matrix> >("M1_beta", Teuchos::null);
-    params->set<RCP<Matrix> >("M1_alpha", Teuchos::null);
-    // for backwards compatiblity
-    params->set<RCP<Matrix> >("Ms", Teuchos::null);
-
-    params->set<RCP<Matrix> >("Mk_one", Teuchos::null);
-    params->set<RCP<Matrix> >("Mk_1_one", Teuchos::null);
-    // for backwards compatiblity
-    params->set<RCP<Matrix> >("M1", Teuchos::null);
-
-    params->set<RCP<Matrix> >("invMk_1_invBeta", Teuchos::null);
-    params->set<RCP<Matrix> >("invMk_2_invAlpha", Teuchos::null);
-    // for backwards compatiblity
-    params->set<RCP<Matrix> >("M0inv", Teuchos::null);
-
-    params->set<RCP<MultiVector> >("Nullspace", Teuchos::null);
-    params->set<RCP<RealValuedMultiVector> >("Coordinates", Teuchos::null);
-
-    auto spaceValidator = rcp(new Teuchos::EnhancedNumberValidator<int>(1, 2));
-    params->set("refmaxwell: space number", 1, "", spaceValidator);
-    params->set("verbosity", MasterList::getDefault<std::string>("verbosity"));
-    params->set("use kokkos refactor", useKokkosDefault);
-    params->set("half precision", false);
-    params->set("parameterlist: syntax", MasterList::getDefault<std::string>("parameterlist: syntax"));
-    params->set("output filename", MasterList::getDefault<std::string>("output filename"));
-    params->set("print initial parameters", MasterList::getDefault<bool>("print initial parameters"));
-    params->set("refmaxwell: disable addon", MasterList::getDefault<bool>("refmaxwell: disable addon"));
-    params->set("refmaxwell: disable addon 22", true);
-    params->set("refmaxwell: mode", MasterList::getDefault<std::string>("refmaxwell: mode"));
-    params->set("refmaxwell: use as preconditioner", MasterList::getDefault<bool>("refmaxwell: use as preconditioner"));
-    params->set("refmaxwell: dump matrices", MasterList::getDefault<bool>("refmaxwell: dump matrices"));
-    params->set("refmaxwell: enable reuse", MasterList::getDefault<bool>("refmaxwell: enable reuse"));
-    params->set("refmaxwell: skip first (1,1) level", MasterList::getDefault<bool>("refmaxwell: skip first (1,1) level"));
-    params->set("refmaxwell: skip first (2,2) level", false);
-    params->set("multigrid algorithm", "Unsmoothed");
-    params->set("transpose: use implicit", MasterList::getDefault<bool>("transpose: use implicit"));
-    params->set("rap: triple product", MasterList::getDefault<bool>("rap: triple product"));
-    params->set("rap: fix zero diagonals", true);
-    params->set("rap: fix zero diagonals threshold", MasterList::getDefault<double>("rap: fix zero diagonals threshold"));
-    params->set("fuse prolongation and update", MasterList::getDefault<bool>("fuse prolongation and update"));
-    params->set("refmaxwell: subsolves on subcommunicators", MasterList::getDefault<bool>("refmaxwell: subsolves on subcommunicators"));
-    params->set("refmaxwell: subsolves striding", 1);
-    params->set("refmaxwell: row sum drop tol (1,1)", MasterList::getDefault<double>("aggregation: row sum drop tol"));
-    params->set("sync timers", false);
-    params->set("refmaxwell: num iters coarse 11", 1);
-    params->set("refmaxwell: num iters 22", 1);
-    params->set("refmaxwell: apply BCs to Anodal",    false);
-    params->set("refmaxwell: apply BCs to coarse 11", true);
-    params->set("refmaxwell: apply BCs to 22",        true);
-    params->set("refmaxwell: max coarse size", 1);
-
-    ParameterList & precList11 = params->sublist("refmaxwell: 11list");
-    precList11.disableRecursiveValidation();
-    ParameterList & precList22 = params->sublist("refmaxwell: 22list");
-    precList22.disableRecursiveValidation();
-
-    params->set("smoother: type", "CHEBYSHEV");
-    ParameterList & smootherList = params->sublist("smoother: params");
-    smootherList.disableRecursiveValidation();
-    params->set("smoother: pre type", "NONE");
-    ParameterList & preSmootherList = params->sublist("smoother: pre params");
-    preSmootherList.disableRecursiveValidation();
-    params->set("smoother: post type", "NONE");
-    ParameterList & postSmootherList = params->sublist("smoother: post params");
-    postSmootherList.disableRecursiveValidation();
-
-    ParameterList & matvecParams = params->sublist("matvec params");
-    matvecParams.disableRecursiveValidation();
-
-    params->set("multigrid algorithm", "unsmoothed");
-    params->set("aggregation: type", MasterList::getDefault<std::string>("aggregation: type"));
-    params->set("aggregation: drop tol", MasterList::getDefault<double>("aggregation: drop tol"));
-    params->set("aggregation: drop scheme", MasterList::getDefault<std::string>("aggregation: drop scheme"));
-    params->set("aggregation: distance laplacian algo", MasterList::getDefault<std::string>("aggregation: distance laplacian algo"));
-    params->set("aggregation: min agg size", MasterList::getDefault<int>("aggregation: min agg size"));
-    params->set("aggregation: max agg size", MasterList::getDefault<int>("aggregation: max agg size"));
-    params->set("aggregation: match ML phase1", MasterList::getDefault<bool>("aggregation: match ML phase1"));
-    params->set("aggregation: match ML phase2a", MasterList::getDefault<bool>("aggregation: match ML phase2a"));
-    params->set("aggregation: match ML phase2b", MasterList::getDefault<bool>("aggregation: match ML phase2b"));
-    params->set("aggregation: export visualization data", MasterList::getDefault<bool>("aggregation: export visualization data"));
-
-    return params;
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setParameters(Teuchos::ParameterList& list) {
-
-    if (list.isType<std::string>("parameterlist: syntax") && list.get<std::string>("parameterlist: syntax") == "ml") {
-      Teuchos::ParameterList newList;
-      {
-        Teuchos::ParameterList newList2 = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list,"refmaxwell"));
-        RCP<Teuchos::ParameterList> validateParameters = getValidParamterList();
-        for (auto it = newList2.begin(); it != newList2.end(); ++it) {
-          const std::string &entry_name = it->first;
-          if (validateParameters->isParameter(entry_name)) {
-            ParameterEntry theEntry = newList2.getEntry(entry_name);
-            newList.setEntry(entry_name, theEntry);
-          }
-        }
-      }
+  precList22_ = parameterList_.sublist("refmaxwell: 22list");
+  if (!precList22_.isType<std::string>("Preconditioner Type") &&
+      !precList22_.isType<std::string>("smoother: type") &&
+      !precList22_.isType<std::string>("smoother: pre type") &&
+      !precList22_.isType<std::string>("smoother: post type")) {
+    precList22_.set("smoother: type", "CHEBYSHEV");
+    precList22_.sublist("smoother: params").set("chebyshev: degree", 2);
+    precList22_.sublist("smoother: params").set("chebyshev: ratio eigenvalue", 7.0);
+    precList22_.sublist("smoother: params").set("chebyshev: eigenvalue max iterations", 30);
+  }
 
-      if(list.isSublist("refmaxwell: 11list") && list.sublist("refmaxwell: 11list").isSublist("edge matrix free: coarse"))
-        newList.sublist("refmaxwell: 11list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list.sublist("refmaxwell: 11list").sublist("edge matrix free: coarse"),"SA"));
-      if(list.isSublist("refmaxwell: 22list"))
-        newList.sublist("refmaxwell: 22list") = *Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(list.sublist("refmaxwell: 22list"),"SA"));
-      list = newList;
-    }
-
-    parameterList_             = list;
-    parameterList_.validateParametersAndSetDefaults(*getValidParamterList());
-    std::string verbosityLevel = parameterList_.get<std::string>("verbosity");
-    VerboseObject::SetDefaultVerbLevel(toVerbLevel(verbosityLevel));
-    std::string outputFilename = parameterList_.get<std::string>("output filename");
-    if (outputFilename != "")
-      VerboseObject::SetMueLuOFileStream(outputFilename);
-    if (parameterList_.isType<Teuchos::RCP<Teuchos::FancyOStream> >("output stream"))
-      VerboseObject::SetMueLuOStream(parameterList_.get<Teuchos::RCP<Teuchos::FancyOStream> >("output stream"));
-
-    if (parameterList_.get<bool>("print initial parameters"))
-      GetOStream(static_cast<MsgType>(Runtime1), 0) << parameterList_ << std::endl;
-    disable_addon_             = parameterList_.get<bool>("refmaxwell: disable addon");
-    disable_addon_22_          = parameterList_.get<bool>("refmaxwell: disable addon 22");
-    mode_                      = parameterList_.get<std::string>("refmaxwell: mode");
-    use_as_preconditioner_     = parameterList_.get<bool>("refmaxwell: use as preconditioner");
-    dump_matrices_             = parameterList_.get<bool>("refmaxwell: dump matrices");
-    enable_reuse_              = parameterList_.get<bool>("refmaxwell: enable reuse");
-    implicitTranspose_         = parameterList_.get<bool>("transpose: use implicit");
-    fuseProlongationAndUpdate_ = parameterList_.get<bool>("fuse prolongation and update");
-    skipFirst11Level_          = parameterList_.get<bool>("refmaxwell: skip first (1,1) level");
-    skipFirst22Level_          = parameterList_.get<bool>("refmaxwell: skip first (2,2) level");
-    if (spaceNumber_ == 1)
-      skipFirst22Level_ = false;
-    syncTimers_                = parameterList_.get<bool>("sync timers");
-    useKokkos_                 = parameterList_.get<bool>("use kokkos refactor");
-    numItersCoarse11_          = parameterList_.get<int>("refmaxwell: num iters coarse 11");
-    numIters22_                = parameterList_.get<int>("refmaxwell: num iters 22");
-    applyBCsToAnodal_          = parameterList_.get<bool>("refmaxwell: apply BCs to Anodal");
-    applyBCsToCoarse11_        = parameterList_.get<bool>("refmaxwell: apply BCs to coarse 11");
-    applyBCsTo22_              = parameterList_.get<bool>("refmaxwell: apply BCs to 22");
-
-    precList11_     =  parameterList_.sublist("refmaxwell: 11list");
-    if(!precList11_.isType<std::string>("Preconditioner Type") &&
-       !precList11_.isType<std::string>("smoother: type") &&
-       !precList11_.isType<std::string>("smoother: pre type") &&
-       !precList11_.isType<std::string>("smoother: post type")) {
-      precList11_.set("smoother: type", "CHEBYSHEV");
-      precList11_.sublist("smoother: params").set("chebyshev: degree",2);
-      precList11_.sublist("smoother: params").set("chebyshev: ratio eigenvalue",5.4);
-      precList11_.sublist("smoother: params").set("chebyshev: eigenvalue max iterations",30);
-    }
-
-    precList22_     =  parameterList_.sublist("refmaxwell: 22list");
-    if(!precList22_.isType<std::string>("Preconditioner Type") &&
-       !precList22_.isType<std::string>("smoother: type") &&
-       !precList22_.isType<std::string>("smoother: pre type") &&
-       !precList22_.isType<std::string>("smoother: post type")) {
-      precList22_.set("smoother: type", "CHEBYSHEV");
-      precList22_.sublist("smoother: params").set("chebyshev: degree",2);
-      precList22_.sublist("smoother: params").set("chebyshev: ratio eigenvalue",7.0);
-      precList22_.sublist("smoother: params").set("chebyshev: eigenvalue max iterations",30);
-    }
-
-    if(!parameterList_.isType<std::string>("smoother: type") && !parameterList_.isType<std::string>("smoother: pre type") && !parameterList_.isType<std::string>("smoother: post type")) {
-      list.set("smoother: type", "CHEBYSHEV");
-      list.sublist("smoother: params").set("chebyshev: degree",2);
-      list.sublist("smoother: params").set("chebyshev: ratio eigenvalue",20.0);
-      list.sublist("smoother: params").set("chebyshev: eigenvalue max iterations",30);
-    }
-
-    if (enable_reuse_ &&
-        !precList11_.isType<std::string>("Preconditioner Type") &&
-        !precList11_.isParameter("reuse: type"))
-      precList11_.set("reuse: type", "full");
-    if (enable_reuse_ &&
-        !precList22_.isType<std::string>("Preconditioner Type") &&
-        !precList22_.isParameter("reuse: type"))
-      precList22_.set("reuse: type", "full");
-
-    // This should be taken out again as soon as
-    // CoalesceDropFactory_kokkos supports BlockSize > 1 and
-    // drop tol != 0.0
-    if (useKokkos_ && precList11_.isParameter("aggregation: drop tol") && precList11_.get<double>("aggregation: drop tol") != 0.0) {
-      GetOStream(Warnings0) << solverName_+"::compute(): Setting \"aggregation: drop tol\". to 0.0, since CoalesceDropFactory_kokkos does not "
-                            << "support BlockSize > 1 and drop tol != 0.0" << std::endl;
-      precList11_.set("aggregation: drop tol", 0.0);
-    }
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::compute(bool reuse) {
+  if (!parameterList_.isType<std::string>("smoother: type") && !parameterList_.isType<std::string>("smoother: pre type") && !parameterList_.isType<std::string>("smoother: post type")) {
+    list.set("smoother: type", "CHEBYSHEV");
+    list.sublist("smoother: params").set("chebyshev: degree", 2);
+    list.sublist("smoother: params").set("chebyshev: ratio eigenvalue", 20.0);
+    list.sublist("smoother: params").set("chebyshev: eigenvalue max iterations", 30);
+  }
+
+  if (enable_reuse_ &&
+      !precList11_.isType<std::string>("Preconditioner Type") &&
+      !precList11_.isParameter("reuse: type"))
+    precList11_.set("reuse: type", "full");
+  if (enable_reuse_ &&
+      !precList22_.isType<std::string>("Preconditioner Type") &&
+      !precList22_.isParameter("reuse: type"))
+    precList22_.set("reuse: type", "full");
+
+  // This should be taken out again as soon as
+  // CoalesceDropFactory_kokkos supports BlockSize > 1 and
+  // drop tol != 0.0
+  if (useKokkos_ && precList11_.isParameter("aggregation: drop tol") && precList11_.get<double>("aggregation: drop tol") != 0.0) {
+    GetOStream(Warnings0) << solverName_ + "::compute(): Setting \"aggregation: drop tol\". to 0.0, since CoalesceDropFactory_kokkos does not "
+                          << "support BlockSize > 1 and drop tol != 0.0" << std::endl;
+    precList11_.set("aggregation: drop tol", 0.0);
+  }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::compute(bool reuse) {
 #ifdef HAVE_MUELU_CUDA
-    if (parameterList_.get<bool>("refmaxwell: cuda profile setup", false)) cudaProfilerStart();
+  if (parameterList_.get<bool>("refmaxwell: cuda profile setup", false)) cudaProfilerStart();
 #endif
 
-    std::string timerLabel;
-    if (reuse)
-      timerLabel = "compute (reuse)";
-    else
-      timerLabel = "compute";
-    RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // COMMENTED OUT SINCE WE SHOULD NOT NEED THIS ANYMORE.
-    // Remove explicit zeros from matrices
-    // Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(parameterList_,D0_,SM_Matrix_,Mk_one_,M1_beta_);
-    // if (!Dk_1_.is_null())
-    //   Dk_1_ = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(Dk_1_, 1e-10, false);
+  std::string timerLabel;
+  if (reuse)
+    timerLabel = "compute (reuse)";
+  else
+    timerLabel = "compute";
+  RCP<Teuchos::TimeMonitor> tmCompute = getTimer(timerLabel);
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // COMMENTED OUT SINCE WE SHOULD NOT NEED THIS ANYMORE.
+  // Remove explicit zeros from matrices
+  // Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(parameterList_,D0_,SM_Matrix_,Mk_one_,M1_beta_);
+  // if (!Dk_1_.is_null())
+  //   Dk_1_ = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(Dk_1_, 1e-10, false);
+
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*SM_Matrix_, "SM_Matrix", params);
+  }
 
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*SM_Matrix_, "SM_Matrix", params);
+  ////////////////////////////////////////////////////////////////////////////////
+  // Detect Dirichlet boundary conditions
+  if (!reuse) {
+    magnitudeType rowSumTol = parameterList_.get<double>("refmaxwell: row sum drop tol (1,1)");
+    Maxwell_Utils<SC, LO, GO, NO>::detectBoundaryConditionsSM(SM_Matrix_, Dk_1_, rowSumTol,
+                                                              BCrows11_, BCcols22_, BCdomain22_,
+                                                              globalNumberBoundaryUnknowns11_,
+                                                              globalNumberBoundaryUnknowns22_,
+                                                              onlyBoundary11_, onlyBoundary22_);
+    if (spaceNumber_ == 2) {
+      Kokkos::View<bool *, typename Node::device_type> BCcolsEdge   = Kokkos::View<bool *, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), Dk_1_->getColMap()->getLocalNumElements());
+      Kokkos::View<bool *, typename Node::device_type> BCdomainEdge = Kokkos::View<bool *, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), Dk_1_->getDomainMap()->getLocalNumElements());
+      Utilities::DetectDirichletColsAndDomains(*Dk_1_, BCrows11_, BCcolsEdge, BCdomainEdge);
+
+      Kokkos::View<bool *, typename Node::device_type> BCcolsNode   = Kokkos::View<bool *, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_->getColMap()->getLocalNumElements());
+      Kokkos::View<bool *, typename Node::device_type> BCdomainNode = Kokkos::View<bool *, typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_->getDomainMap()->getLocalNumElements());
+      Utilities::DetectDirichletColsAndDomains(*D0_, BCdomainEdge, BCcolsNode, BCdomainNode);
+      BCdomain22_ = BCdomainNode;
     }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Detect Dirichlet boundary conditions
-    if (!reuse) {
-      magnitudeType rowSumTol = parameterList_.get<double>("refmaxwell: row sum drop tol (1,1)");
-      Maxwell_Utils<SC,LO,GO,NO>::detectBoundaryConditionsSM(SM_Matrix_,Dk_1_,rowSumTol,
-                                                             BCrows11_,BCcols22_,BCdomain22_,
-                                                             globalNumberBoundaryUnknowns11_,
-                                                             globalNumberBoundaryUnknowns22_,
-                                                             onlyBoundary11_,onlyBoundary22_);
-      if (spaceNumber_ == 2) {
-        Kokkos::View<bool*, typename Node::device_type> BCcolsEdge = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), Dk_1_->getColMap()->getLocalNumElements());
-        Kokkos::View<bool*, typename Node::device_type> BCdomainEdge = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), Dk_1_->getDomainMap()->getLocalNumElements());
-        Utilities::DetectDirichletColsAndDomains(*Dk_1_,BCrows11_,BCcolsEdge,BCdomainEdge);
-
-        Kokkos::View<bool*, typename Node::device_type> BCcolsNode = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), D0_->getColMap()->getLocalNumElements());
-        Kokkos::View<bool*, typename Node::device_type> BCdomainNode = Kokkos::View<bool*,typename Node::device_type>(Kokkos::ViewAllocateWithoutInitializing("dirichletDomains"), D0_->getDomainMap()->getLocalNumElements());
-        Utilities::DetectDirichletColsAndDomains(*D0_,BCdomainEdge,BCcolsNode,BCdomainNode);
-        BCdomain22_ = BCdomainNode;
-      }
-      if (IsPrint(Statistics2)) {
-        GetOStream(Statistics2) << solverName_+"::compute(): Detected " << globalNumberBoundaryUnknowns11_ << " BC rows and " << globalNumberBoundaryUnknowns22_ << " BC columns." << std::endl;
-      }
-      dump(BCrows11_, "BCrows11.m");
-      dump(BCcols22_, "BCcols22.m");
-      dump(BCdomain22_, "BCdomain22.m");
+    if (IsPrint(Statistics2)) {
+      GetOStream(Statistics2) << solverName_ + "::compute(): Detected " << globalNumberBoundaryUnknowns11_ << " BC rows and " << globalNumberBoundaryUnknowns22_ << " BC columns." << std::endl;
     }
+    dump(BCrows11_, "BCrows11.m");
+    dump(BCcols22_, "BCcols22.m");
+    dump(BCdomain22_, "BCdomain22.m");
+  }
 
-    if (onlyBoundary11_) {
-      // All unknowns of the (1,1) block have been detected as boundary unknowns.
-      // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
-      GetOStream(Warnings0) << "All unknowns of the (1,1) block have been detected as boundary unknowns!" << std::endl;
-      mode_ = "none";
-      setFineLevelSmoother11();
-      return;
-    }
+  if (onlyBoundary11_) {
+    // All unknowns of the (1,1) block have been detected as boundary unknowns.
+    // Do not attempt to construct sub-hierarchies, but just set up a single level preconditioner.
+    GetOStream(Warnings0) << "All unknowns of the (1,1) block have been detected as boundary unknowns!" << std::endl;
+    mode_ = "none";
+    setFineLevelSmoother11();
+    return;
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
+  ////////////////////////////////////////////////////////////////////////////////
 
-    dim_ = NodalCoords_->getNumVectors();
+  dim_ = NodalCoords_->getNumVectors();
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // build special prolongators
-    if (!reuse) {
+  ////////////////////////////////////////////////////////////////////////////////
+  // build special prolongators
+  if (!reuse) {
+    //////////////////////////////////////////////////////////////////////////////
+    // build nullspace for (1,1)-block (if necessary)
+    if (Nullspace11_ != null) {  // no need to do anything - nullspace is built
+      TEUCHOS_ASSERT(Nullspace11_->getMap()->isCompatible(*(SM_Matrix_->getRowMap())));
+    } else if (NodalCoords_ != null) {
+      Nullspace11_ = buildNullspace(spaceNumber_, BCrows11_, skipFirst11Level_);
+    } else {
+      GetOStream(Errors) << solverName_ + "::compute(): either the nullspace or the nodal coordinates must be provided." << std::endl;
+    }
 
-      //////////////////////////////////////////////////////////////////////////////
-      // build nullspace for (1,1)-block (if necessary)
-      if(Nullspace11_ != null) { // no need to do anything - nullspace is built
-        TEUCHOS_ASSERT(Nullspace11_->getMap()->isCompatible(*(SM_Matrix_->getRowMap())));
-      } else if (NodalCoords_ != null) {
-        Nullspace11_ = buildNullspace(spaceNumber_, BCrows11_, skipFirst11Level_);
-      } else {
-        GetOStream(Errors) << solverName_+"::compute(): either the nullspace or the nodal coordinates must be provided." << std::endl;
-      }
+    // build special prolongator for (1,1)-block
+    {
+      RCP<Matrix> A11_nodal;
+      if (skipFirst11Level_) {
+        // Form A11_nodal = D0^T * M1_beta * D0  (aka TMT_agg)
+        std::string label("D0^T*M1_beta*D0");
+        A11_nodal = Maxwell_Utils<SC, LO, GO, NO>::PtAPWrapper(M1_beta_, D0_, parameterList_, label);
 
-      // build special prolongator for (1,1)-block
-      {
-        RCP<Matrix> A11_nodal;
-        if (skipFirst11Level_) {
-          // Form A11_nodal = D0^T * M1_beta * D0  (aka TMT_agg)
-          std::string label("D0^T*M1_beta*D0");
-          A11_nodal = Maxwell_Utils<SC,LO,GO,NO>::PtAPWrapper(M1_beta_,D0_,parameterList_,label);
-
-          if (applyBCsToAnodal_) {
-            // Apply boundary conditions to A11_nodal
-            Utilities::ApplyOAZToMatrixRows(A11_nodal,BCdomain22_);
-          }
-          A11_nodal->setObjectLabel(solverName_+" (1,1) A_nodal");
-          dump(A11_nodal, "A11_nodal.m");
+        if (applyBCsToAnodal_) {
+          // Apply boundary conditions to A11_nodal
+          Utilities::ApplyOAZToMatrixRows(A11_nodal, BCdomain22_);
         }
-        // release it because we won't need it anymore
-        M1_beta_ = Teuchos::null;
-
-        // build special prolongator
-        buildProlongator(spaceNumber_, A11_nodal, Nullspace11_, P11_, NullspaceCoarse11_, CoordsCoarse11_);
-
-        dump(P11_, "P11.m");
+        A11_nodal->setObjectLabel(solverName_ + " (1,1) A_nodal");
+        dump(A11_nodal, "A11_nodal.m");
       }
+      // release it because we won't need it anymore
+      M1_beta_ = Teuchos::null;
 
-      //////////////////////////////////////////////////////////////////////////////
-      // build nullspace for (2,2)-block (if necessary)
-      if (Nullspace22_ != null) {
-        TEUCHOS_ASSERT(Nullspace22_->getMap()->isCompatible(*(Dk_1_->getDomainMap())));
-      } else if (NodalCoords_ != null)
-        Nullspace22_ = buildNullspace(spaceNumber_-1, BCdomain22_, skipFirst22Level_);
-      else {
-        GetOStream(Errors) << solverName_+"::compute(): either the nullspace or the nodal coordinates must be provided." << std::endl;
-      }
+      // build special prolongator
+      buildProlongator(spaceNumber_, A11_nodal, Nullspace11_, P11_, NullspaceCoarse11_, CoordsCoarse11_);
 
-      // build special prolongator for (2,2)-block
-      {
-        RCP<Matrix> A22_nodal;
-        if (skipFirst22Level_) {
-          // Form A22_nodal = D0^T * M1_alpha * D0
-          std::string label("D0^T*M1_alpha*D0");
-          A22_nodal = Maxwell_Utils<SC,LO,GO,NO>::PtAPWrapper(M1_alpha_,D0_,parameterList_,label);
-
-          if (applyBCsToAnodal_) {
-            // Apply boundary conditions to A22_nodal
-            Utilities::ApplyOAZToMatrixRows(A22_nodal,BCdomain22_);
-          }
-          A22_nodal->setObjectLabel(solverName_+" (2,2) A_nodal");
-          dump(A22_nodal, "A22_nodal.m");
-        }
-        // release it because we won't need it anymore
-        M1_alpha_ = Teuchos::null;
+      dump(P11_, "P11.m");
+    }
 
-        // build special prolongator
-        buildProlongator(spaceNumber_-1, A22_nodal, Nullspace22_, P22_, CoarseNullspace22_, Coords22_);
+    //////////////////////////////////////////////////////////////////////////////
+    // build nullspace for (2,2)-block (if necessary)
+    if (Nullspace22_ != null) {
+      TEUCHOS_ASSERT(Nullspace22_->getMap()->isCompatible(*(Dk_1_->getDomainMap())));
+    } else if (NodalCoords_ != null)
+      Nullspace22_ = buildNullspace(spaceNumber_ - 1, BCdomain22_, skipFirst22Level_);
+    else {
+      GetOStream(Errors) << solverName_ + "::compute(): either the nullspace or the nodal coordinates must be provided." << std::endl;
+    }
 
-        dump(P22_, "P22.m");
+    // build special prolongator for (2,2)-block
+    {
+      RCP<Matrix> A22_nodal;
+      if (skipFirst22Level_) {
+        // Form A22_nodal = D0^T * M1_alpha * D0
+        std::string label("D0^T*M1_alpha*D0");
+        A22_nodal = Maxwell_Utils<SC, LO, GO, NO>::PtAPWrapper(M1_alpha_, D0_, parameterList_, label);
+
+        if (applyBCsToAnodal_) {
+          // Apply boundary conditions to A22_nodal
+          Utilities::ApplyOAZToMatrixRows(A22_nodal, BCdomain22_);
+        }
+        A22_nodal->setObjectLabel(solverName_ + " (2,2) A_nodal");
+        dump(A22_nodal, "A22_nodal.m");
       }
+      // release it because we won't need it anymore
+      M1_alpha_ = Teuchos::null;
 
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // build coarse grid operator for (1,1)-block
-    buildCoarse11Matrix();
+      // build special prolongator
+      buildProlongator(spaceNumber_ - 1, A22_nodal, Nullspace22_, P22_, CoarseNullspace22_, Coords22_);
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // determine the communicator sizes for (1,1)- and (2,2)-blocks
-    bool doRebalancing;
-    int rebalanceStriding, numProcsCoarseA11, numProcsA22;
-    if (!reuse)
-      this->determineSubHierarchyCommSizes(doRebalancing, rebalanceStriding, numProcsCoarseA11, numProcsA22);
-    else
-      doRebalancing = false;
-
-    // rebalance the coarse A11 matrix, as well as P11, CoordsCoarse11 and Addon11
-    if (!reuse && doRebalancing)
-      rebalanceCoarse11Matrix(rebalanceStriding, numProcsCoarseA11);
-    if (!coarseA11_.is_null()) {
-      dump(coarseA11_, "coarseA11.m");
-      if (!reuse) {
-        dumpCoords(CoordsCoarse11_, "CoordsCoarse11.m");
-        dump(NullspaceCoarse11_, "NullspaceCoarse11.m");
-      }
+      dump(P22_, "P22.m");
     }
+  }
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // build coarse grid operator for (1,1)-block
+  buildCoarse11Matrix();
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // determine the communicator sizes for (1,1)- and (2,2)-blocks
+  bool doRebalancing;
+  int rebalanceStriding, numProcsCoarseA11, numProcsA22;
+  if (!reuse)
+    this->determineSubHierarchyCommSizes(doRebalancing, rebalanceStriding, numProcsCoarseA11, numProcsA22);
+  else
+    doRebalancing = false;
 
+  // rebalance the coarse A11 matrix, as well as P11, CoordsCoarse11 and Addon11
+  if (!reuse && doRebalancing)
+    rebalanceCoarse11Matrix(rebalanceStriding, numProcsCoarseA11);
+  if (!coarseA11_.is_null()) {
+    dump(coarseA11_, "coarseA11.m");
     if (!reuse) {
-      if (!implicitTranspose_) {
-        R11_ = Utilities::Transpose(*P11_);
-        dump(R11_, "R11.m");
-      }
-    }
-    ////////////////////////////////////////////////////////////////////////////////
-    // build multigrid for coarse (1,1)-block
-    if (!coarseA11_.is_null()) {
-      VerbLevel verbosityLevel = VerboseObject::GetDefaultVerbLevel();
-      std::string label("coarseA11");
-      setupSubSolve(HierarchyCoarse11_, thyraPrecOpH_, coarseA11_, NullspaceCoarse11_, CoordsCoarse11_, precList11_, label, reuse);
-      VerboseObject::SetDefaultVerbLevel(verbosityLevel);
+      dumpCoords(CoordsCoarse11_, "CoordsCoarse11.m");
+      dump(NullspaceCoarse11_, "NullspaceCoarse11.m");
     }
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Apply BCs to columns of Dk_1
-    if(!reuse && applyBCsTo22_) {
-      GetOStream(Runtime0) << solverName_+"::compute(): nuking BC columns of Dk_1" << std::endl;
-
-      Dk_1_->resumeFill();
-      Scalar replaceWith = (Dk_1_->getRowMap()->lib() == Xpetra::UseEpetra) ? Teuchos::ScalarTraits<SC>::eps() : Teuchos::ScalarTraits<SC>::zero();
-      Utilities::ZeroDirichletCols(Dk_1_,BCcols22_,replaceWith);
-      Dk_1_->fillComplete(Dk_1_->getDomainMap(),Dk_1_->getRangeMap());
+  if (!reuse) {
+    if (!implicitTranspose_) {
+      R11_ = Utilities::Transpose(*P11_);
+      dump(R11_, "R11.m");
     }
+  }
+  ////////////////////////////////////////////////////////////////////////////////
+  // build multigrid for coarse (1,1)-block
+  if (!coarseA11_.is_null()) {
+    VerbLevel verbosityLevel = VerboseObject::GetDefaultVerbLevel();
+    std::string label("coarseA11");
+    setupSubSolve(HierarchyCoarse11_, thyraPrecOpH_, coarseA11_, NullspaceCoarse11_, CoordsCoarse11_, precList11_, label, reuse);
+    VerboseObject::SetDefaultVerbLevel(verbosityLevel);
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Build A22 = Dk_1^T SM Dk_1 and hierarchy for A22
-    if (!onlyBoundary22_) {
-      GetOStream(Runtime0) << solverName_+"::compute(): building MG for (2,2)-block" << std::endl;
+  ////////////////////////////////////////////////////////////////////////////////
+  // Apply BCs to columns of Dk_1
+  if (!reuse && applyBCsTo22_) {
+    GetOStream(Runtime0) << solverName_ + "::compute(): nuking BC columns of Dk_1" << std::endl;
 
-      // Build A22 = Dk_1^T * SM * Dk_1 and rebalance it, as well as Dk_1_ and P22_ and Coords22_
-      build22Matrix(reuse, doRebalancing, rebalanceStriding, numProcsA22);
+    Dk_1_->resumeFill();
+    Scalar replaceWith = (Dk_1_->getRowMap()->lib() == Xpetra::UseEpetra) ? Teuchos::ScalarTraits<SC>::eps() : Teuchos::ScalarTraits<SC>::zero();
+    Utilities::ZeroDirichletCols(Dk_1_, BCcols22_, replaceWith);
+    Dk_1_->fillComplete(Dk_1_->getDomainMap(), Dk_1_->getRangeMap());
+  }
 
-      if (!P22_.is_null()) {
-        std::string label("P22^T*A22*P22");
-        coarseA22_ = Maxwell_Utils<SC,LO,GO,NO>::PtAPWrapper(A22_,P22_,parameterList_,label);
-        coarseA22_->SetFixedBlockSize(A22_->GetFixedBlockSize());
-        coarseA22_->setObjectLabel(solverName_+" coarse (2, 2)");
-        dump(coarseA22_, "coarseA22.m");
-      }
+  ////////////////////////////////////////////////////////////////////////////////
+  // Build A22 = Dk_1^T SM Dk_1 and hierarchy for A22
+  if (!onlyBoundary22_) {
+    GetOStream(Runtime0) << solverName_ + "::compute(): building MG for (2,2)-block" << std::endl;
 
-      if (!reuse && !implicitTranspose_) {
-        Dk_1_T_ = Utilities::Transpose(*Dk_1_);
-        if (!P22_.is_null())
-          R22_ = Utilities::Transpose(*P22_);
-      }
+    // Build A22 = Dk_1^T * SM * Dk_1 and rebalance it, as well as Dk_1_ and P22_ and Coords22_
+    build22Matrix(reuse, doRebalancing, rebalanceStriding, numProcsA22);
 
-      if (!A22_.is_null()) {
-        VerbLevel verbosityLevel = VerboseObject::GetDefaultVerbLevel();
-        std::string label("A22");
-        if (!P22_.is_null()) {
-          precList22_.sublist("level 1 user data").set("A", coarseA22_);
-          precList22_.sublist("level 1 user data").set("P", P22_);
-          if (!implicitTranspose_)
-            precList22_.sublist("level 1 user data").set("R", R22_);
-          precList22_.sublist("level 1 user data").set("Nullspace", CoarseNullspace22_);
-          precList22_.sublist("level 1 user data").set("Coordinates", Coords22_);
-          // A22 is singular, we want to coarsen at least once.
-          // So we make sure coarseA22 is not just ignored.
-          int maxCoarseSize = precList22_.get("coarse: max size", MasterList::getDefault<int>("coarse: max size"));
-          int numRows = Teuchos::as<int>(coarseA22_->getGlobalNumRows());
-          if (maxCoarseSize > numRows)
-            precList22_.set("coarse: max size", numRows);
-          int maxLevels = precList22_.get("max levels", MasterList::getDefault<int>("max levels"));
-          if (maxLevels < 2)
-            precList22_.set("max levels", 2);
-          setupSubSolve(Hierarchy22_, thyraPrecOp22_, A22_, Teuchos::null, Teuchos::null, precList22_, label, reuse, /*isSingular=*/globalNumberBoundaryUnknowns11_ == 0);
-        } else
-          setupSubSolve(Hierarchy22_, thyraPrecOp22_, A22_, CoarseNullspace22_, Coords22_, precList22_, label, reuse, /*isSingular=*/globalNumberBoundaryUnknowns11_ == 0);
+    if (!P22_.is_null()) {
+      std::string label("P22^T*A22*P22");
+      coarseA22_ = Maxwell_Utils<SC, LO, GO, NO>::PtAPWrapper(A22_, P22_, parameterList_, label);
+      coarseA22_->SetFixedBlockSize(A22_->GetFixedBlockSize());
+      coarseA22_->setObjectLabel(solverName_ + " coarse (2, 2)");
+      dump(coarseA22_, "coarseA22.m");
+    }
 
-        VerboseObject::SetDefaultVerbLevel(verbosityLevel);
-      }
+    if (!reuse && !implicitTranspose_) {
+      Dk_1_T_ = Utilities::Transpose(*Dk_1_);
+      if (!P22_.is_null())
+        R22_ = Utilities::Transpose(*P22_);
     }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Apply BCs to rows of Dk_1
-    if(!reuse && !onlyBoundary22_ && applyBCsTo22_) {
-      GetOStream(Runtime0) << solverName_+"::compute(): nuking BC rows of Dk_1" << std::endl;
+    if (!A22_.is_null()) {
+      VerbLevel verbosityLevel = VerboseObject::GetDefaultVerbLevel();
+      std::string label("A22");
+      if (!P22_.is_null()) {
+        precList22_.sublist("level 1 user data").set("A", coarseA22_);
+        precList22_.sublist("level 1 user data").set("P", P22_);
+        if (!implicitTranspose_)
+          precList22_.sublist("level 1 user data").set("R", R22_);
+        precList22_.sublist("level 1 user data").set("Nullspace", CoarseNullspace22_);
+        precList22_.sublist("level 1 user data").set("Coordinates", Coords22_);
+        // A22 is singular, we want to coarsen at least once.
+        // So we make sure coarseA22 is not just ignored.
+        int maxCoarseSize = precList22_.get("coarse: max size", MasterList::getDefault<int>("coarse: max size"));
+        int numRows       = Teuchos::as<int>(coarseA22_->getGlobalNumRows());
+        if (maxCoarseSize > numRows)
+          precList22_.set("coarse: max size", numRows);
+        int maxLevels = precList22_.get("max levels", MasterList::getDefault<int>("max levels"));
+        if (maxLevels < 2)
+          precList22_.set("max levels", 2);
+        setupSubSolve(Hierarchy22_, thyraPrecOp22_, A22_, Teuchos::null, Teuchos::null, precList22_, label, reuse, /*isSingular=*/globalNumberBoundaryUnknowns11_ == 0);
+      } else
+        setupSubSolve(Hierarchy22_, thyraPrecOp22_, A22_, CoarseNullspace22_, Coords22_, precList22_, label, reuse, /*isSingular=*/globalNumberBoundaryUnknowns11_ == 0);
 
-      Dk_1_->resumeFill();
-      Scalar replaceWith = (Dk_1_->getRowMap()->lib() == Xpetra::UseEpetra) ? Teuchos::ScalarTraits<SC>::eps() : Teuchos::ScalarTraits<SC>::zero();
-      Utilities::ZeroDirichletRows(Dk_1_,BCrows11_,replaceWith);
-      Dk_1_->fillComplete(Dk_1_->getDomainMap(),Dk_1_->getRangeMap());
-      dump(Dk_1_, "Dk_1_nuked.m");
+      VerboseObject::SetDefaultVerbLevel(verbosityLevel);
     }
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Set up the smoother on the finest level
-    setFineLevelSmoother11();
+  ////////////////////////////////////////////////////////////////////////////////
+  // Apply BCs to rows of Dk_1
+  if (!reuse && !onlyBoundary22_ && applyBCsTo22_) {
+    GetOStream(Runtime0) << solverName_ + "::compute(): nuking BC rows of Dk_1" << std::endl;
 
-    if (!reuse) {
-      if (!ImporterCoarse11_.is_null()) {
-        RCP<const Import> ImporterP11 = ImportFactory::Build(ImporterCoarse11_->getTargetMap(),P11_->getColMap());
-        rcp_dynamic_cast<CrsMatrixWrap>(P11_)->getCrsMatrix()->replaceDomainMapAndImporter(ImporterCoarse11_->getTargetMap(), ImporterP11);
-      }
+    Dk_1_->resumeFill();
+    Scalar replaceWith = (Dk_1_->getRowMap()->lib() == Xpetra::UseEpetra) ? Teuchos::ScalarTraits<SC>::eps() : Teuchos::ScalarTraits<SC>::zero();
+    Utilities::ZeroDirichletRows(Dk_1_, BCrows11_, replaceWith);
+    Dk_1_->fillComplete(Dk_1_->getDomainMap(), Dk_1_->getRangeMap());
+    dump(Dk_1_, "Dk_1_nuked.m");
+  }
 
-      if (!Importer22_.is_null()) {
-        if (enable_reuse_) {
-          DorigDomainMap_ = Dk_1_->getDomainMap();
-          DorigImporter_ = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->getCrsGraph()->getImporter();
-        }
-        RCP<const Import> ImporterD = ImportFactory::Build(Importer22_->getTargetMap(),Dk_1_->getColMap());
-        rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(Importer22_->getTargetMap(), ImporterD);
+  ////////////////////////////////////////////////////////////////////////////////
+  // Set up the smoother on the finest level
+  setFineLevelSmoother11();
+
+  if (!reuse) {
+    if (!ImporterCoarse11_.is_null()) {
+      RCP<const Import> ImporterP11 = ImportFactory::Build(ImporterCoarse11_->getTargetMap(), P11_->getColMap());
+      rcp_dynamic_cast<CrsMatrixWrap>(P11_)->getCrsMatrix()->replaceDomainMapAndImporter(ImporterCoarse11_->getTargetMap(), ImporterP11);
+    }
+
+    if (!Importer22_.is_null()) {
+      if (enable_reuse_) {
+        DorigDomainMap_ = Dk_1_->getDomainMap();
+        DorigImporter_  = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->getCrsGraph()->getImporter();
       }
+      RCP<const Import> ImporterD = ImportFactory::Build(Importer22_->getTargetMap(), Dk_1_->getColMap());
+      rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(Importer22_->getTargetMap(), ImporterD);
+    }
 
 #ifdef HAVE_MUELU_TPETRA
-      if ((!Dk_1_T_.is_null()) &&
-          (!R11_.is_null()) &&
-          (!rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_T_)->getCrsMatrix()->getCrsGraph()->getImporter().is_null()) &&
-          (!rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix()->getCrsGraph()->getImporter().is_null()) &&
-          (Dk_1_T_->getColMap()->lib() == Xpetra::UseTpetra) &&
-          (R11_->getColMap()->lib() == Xpetra::UseTpetra))
-        Dk_1_T_R11_colMapsMatch_ = Dk_1_T_->getColMap()->isSameAs(*R11_->getColMap());
-      else
+    if ((!Dk_1_T_.is_null()) &&
+        (!R11_.is_null()) &&
+        (!rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_T_)->getCrsMatrix()->getCrsGraph()->getImporter().is_null()) &&
+        (!rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix()->getCrsGraph()->getImporter().is_null()) &&
+        (Dk_1_T_->getColMap()->lib() == Xpetra::UseTpetra) &&
+        (R11_->getColMap()->lib() == Xpetra::UseTpetra))
+      Dk_1_T_R11_colMapsMatch_ = Dk_1_T_->getColMap()->isSameAs(*R11_->getColMap());
+    else
 #endif
-        Dk_1_T_R11_colMapsMatch_ = false;
-      if (Dk_1_T_R11_colMapsMatch_)
-        GetOStream(Runtime0) << solverName_+"::compute(): Dk_1_T and R11 have matching colMaps" << std::endl;
-
-      // Allocate MultiVectors for solve
-      allocateMemory(1);
-
-      // apply matvec params
-      if (parameterList_.isSublist("matvec params"))
-        {
-          RCP<ParameterList> matvecParams = rcpFromRef(parameterList_.sublist("matvec params"));
-          Maxwell_Utils<SC,LO,GO,NO>::setMatvecParams(*SM_Matrix_, matvecParams);
-          Maxwell_Utils<SC,LO,GO,NO>::setMatvecParams(*Dk_1_, matvecParams);
-          Maxwell_Utils<SC,LO,GO,NO>::setMatvecParams(*P11_, matvecParams);
-          if (!Dk_1_T_.is_null())           Maxwell_Utils<SC,LO,GO,NO>::setMatvecParams(*Dk_1_T_, matvecParams);
-          if (!R11_.is_null())              Maxwell_Utils<SC,LO,GO,NO>::setMatvecParams(*R11_, matvecParams);
-          if (!ImporterCoarse11_.is_null()) ImporterCoarse11_->setDistributorParameters(matvecParams);
-          if (!Importer22_.is_null())       Importer22_->setDistributorParameters(matvecParams);
-        }
-      if (!ImporterCoarse11_.is_null() && parameterList_.isSublist("refmaxwell: ImporterCoarse11 params")){
-        RCP<ParameterList> importerParams = rcpFromRef(parameterList_.sublist("refmaxwell: ImporterCoarse11 params"));
-        ImporterCoarse11_->setDistributorParameters(importerParams);
-      }
-      if (!Importer22_.is_null() && parameterList_.isSublist("refmaxwell: Importer22 params")){
-        RCP<ParameterList> importerParams = rcpFromRef(parameterList_.sublist("refmaxwell: Importer22 params"));
-        Importer22_->setDistributorParameters(importerParams);
-      }
+      Dk_1_T_R11_colMapsMatch_ = false;
+    if (Dk_1_T_R11_colMapsMatch_)
+      GetOStream(Runtime0) << solverName_ + "::compute(): Dk_1_T and R11 have matching colMaps" << std::endl;
+
+    // Allocate MultiVectors for solve
+    allocateMemory(1);
+
+    // apply matvec params
+    if (parameterList_.isSublist("matvec params")) {
+      RCP<ParameterList> matvecParams = rcpFromRef(parameterList_.sublist("matvec params"));
+      Maxwell_Utils<SC, LO, GO, NO>::setMatvecParams(*SM_Matrix_, matvecParams);
+      Maxwell_Utils<SC, LO, GO, NO>::setMatvecParams(*Dk_1_, matvecParams);
+      Maxwell_Utils<SC, LO, GO, NO>::setMatvecParams(*P11_, matvecParams);
+      if (!Dk_1_T_.is_null()) Maxwell_Utils<SC, LO, GO, NO>::setMatvecParams(*Dk_1_T_, matvecParams);
+      if (!R11_.is_null()) Maxwell_Utils<SC, LO, GO, NO>::setMatvecParams(*R11_, matvecParams);
+      if (!ImporterCoarse11_.is_null()) ImporterCoarse11_->setDistributorParameters(matvecParams);
+      if (!Importer22_.is_null()) Importer22_->setDistributorParameters(matvecParams);
+    }
+    if (!ImporterCoarse11_.is_null() && parameterList_.isSublist("refmaxwell: ImporterCoarse11 params")) {
+      RCP<ParameterList> importerParams = rcpFromRef(parameterList_.sublist("refmaxwell: ImporterCoarse11 params"));
+      ImporterCoarse11_->setDistributorParameters(importerParams);
+    }
+    if (!Importer22_.is_null() && parameterList_.isSublist("refmaxwell: Importer22 params")) {
+      RCP<ParameterList> importerParams = rcpFromRef(parameterList_.sublist("refmaxwell: Importer22 params"));
+      Importer22_->setDistributorParameters(importerParams);
     }
+  }
 
-    describe(GetOStream(Runtime0));
+  describe(GetOStream(Runtime0));
 
 #ifdef HAVE_MUELU_CUDA
-    if (parameterList_.get<bool>("refmaxwell: cuda profile setup", false)) cudaProfilerStop();
+  if (parameterList_.get<bool>("refmaxwell: cuda profile setup", false)) cudaProfilerStop();
 #endif
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    determineSubHierarchyCommSizes(bool &doRebalancing, int &rebalanceStriding, int &numProcsCoarseA11, int &numProcsA22) {
+  doRebalancing     = parameterList_.get<bool>("refmaxwell: subsolves on subcommunicators");
+  rebalanceStriding = parameterList_.get<int>("refmaxwell: subsolves striding", -1);
+  int numProcs      = SM_Matrix_->getDomainMap()->getComm()->getSize();
+  if (numProcs == 1) {
+    doRebalancing = false;
+    return;
   }
 
+#ifdef HAVE_MPI
+  if (doRebalancing) {
+    {
+      // decide on number of ranks for coarse (1, 1) problem
+
+      Level level;
+      level.SetFactoryManager(null);
+      level.SetLevelID(0);
+      level.Set("A", coarseA11_);
+
+      auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
+      ParameterList repartheurParams;
+      repartheurParams.set("repartition: start level", 0);
+      // Setting min == target on purpose.
+      int defaultTargetRows = 10000;
+      repartheurParams.set("repartition: min rows per proc", precList11_.get<int>("repartition: target rows per proc", defaultTargetRows));
+      repartheurParams.set("repartition: target rows per proc", precList11_.get<int>("repartition: target rows per proc", defaultTargetRows));
+      repartheurParams.set("repartition: min rows per thread", precList11_.get<int>("repartition: target rows per thread", defaultTargetRows));
+      repartheurParams.set("repartition: target rows per thread", precList11_.get<int>("repartition: target rows per thread", defaultTargetRows));
+      repartheurParams.set("repartition: max imbalance", precList11_.get<double>("repartition: max imbalance", 1.1));
+      repartheurFactory->SetParameterList(repartheurParams);
+
+      level.Request("number of partitions", repartheurFactory.get());
+      repartheurFactory->Build(level);
+      numProcsCoarseA11 = level.Get<int>("number of partitions", repartheurFactory.get());
+      numProcsCoarseA11 = std::min(numProcsCoarseA11, numProcs);
+    }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  determineSubHierarchyCommSizes(bool &doRebalancing, int &rebalanceStriding, int &numProcsCoarseA11, int &numProcsA22) {
+    {
+      // decide on number of ranks for (2, 2) problem
 
-    doRebalancing = parameterList_.get<bool>("refmaxwell: subsolves on subcommunicators");
-    rebalanceStriding = parameterList_.get<int>("refmaxwell: subsolves striding", -1);
-    int numProcs = SM_Matrix_->getDomainMap()->getComm()->getSize();
-    if (numProcs == 1) {
-      doRebalancing = false;
-      return;
-    }
+      Level level;
+      level.SetFactoryManager(null);
+      level.SetLevelID(0);
 
-#ifdef HAVE_MPI
-    if (doRebalancing) {
+      level.Set("Map", Dk_1_->getDomainMap());
 
-      {
-        // decide on number of ranks for coarse (1, 1) problem
-
-        Level level;
-        level.SetFactoryManager(null);
-        level.SetLevelID(0);
-        level.Set("A",coarseA11_);
-
-        auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
-        ParameterList repartheurParams;
-        repartheurParams.set("repartition: start level",            0);
-        // Setting min == target on purpose.
-        int defaultTargetRows = 10000;
-        repartheurParams.set("repartition: min rows per proc",      precList11_.get<int>("repartition: target rows per proc", defaultTargetRows));
-        repartheurParams.set("repartition: target rows per proc",   precList11_.get<int>("repartition: target rows per proc", defaultTargetRows));
-        repartheurParams.set("repartition: min rows per thread",    precList11_.get<int>("repartition: target rows per thread", defaultTargetRows));
-        repartheurParams.set("repartition: target rows per thread", precList11_.get<int>("repartition: target rows per thread", defaultTargetRows));
-        repartheurParams.set("repartition: max imbalance",          precList11_.get<double>("repartition: max imbalance", 1.1));
-        repartheurFactory->SetParameterList(repartheurParams);
-
-        level.Request("number of partitions", repartheurFactory.get());
-        repartheurFactory->Build(level);
-        numProcsCoarseA11 = level.Get<int>("number of partitions", repartheurFactory.get());
-        numProcsCoarseA11 = std::min(numProcsCoarseA11,numProcs);
-      }
+      auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
+      ParameterList repartheurParams;
+      repartheurParams.set("repartition: start level", 0);
+      repartheurParams.set("repartition: use map", true);
+      // Setting min == target on purpose.
+      int defaultTargetRows = 10000;
+      repartheurParams.set("repartition: min rows per proc", precList22_.get<int>("repartition: target rows per proc", defaultTargetRows));
+      repartheurParams.set("repartition: target rows per proc", precList22_.get<int>("repartition: target rows per proc", defaultTargetRows));
+      repartheurParams.set("repartition: min rows per thread", precList22_.get<int>("repartition: target rows per thread", defaultTargetRows));
+      repartheurParams.set("repartition: target rows per thread", precList22_.get<int>("repartition: target rows per thread", defaultTargetRows));
+      // repartheurParams.set("repartition: max imbalance",        precList22_.get<double>("repartition: max imbalance", 1.1));
+      repartheurFactory->SetParameterList(repartheurParams);
 
-      {
-        // decide on number of ranks for (2, 2) problem
-
-        Level level;
-        level.SetFactoryManager(null);
-        level.SetLevelID(0);
-
-        level.Set("Map",Dk_1_->getDomainMap());
-
-        auto repartheurFactory = rcp(new RepartitionHeuristicFactory());
-        ParameterList repartheurParams;
-        repartheurParams.set("repartition: start level",            0);
-        repartheurParams.set("repartition: use map",                true);
-        // Setting min == target on purpose.
-        int defaultTargetRows = 10000;
-        repartheurParams.set("repartition: min rows per proc",      precList22_.get<int>("repartition: target rows per proc", defaultTargetRows));
-        repartheurParams.set("repartition: target rows per proc",   precList22_.get<int>("repartition: target rows per proc", defaultTargetRows));
-        repartheurParams.set("repartition: min rows per thread",    precList22_.get<int>("repartition: target rows per thread", defaultTargetRows));
-        repartheurParams.set("repartition: target rows per thread", precList22_.get<int>("repartition: target rows per thread", defaultTargetRows));
-        // repartheurParams.set("repartition: max imbalance",        precList22_.get<double>("repartition: max imbalance", 1.1));
-        repartheurFactory->SetParameterList(repartheurParams);
-
-        level.Request("number of partitions", repartheurFactory.get());
-        repartheurFactory->Build(level);
-        numProcsA22 = level.Get<int>("number of partitions", repartheurFactory.get());
-        numProcsA22 = std::min(numProcsA22,numProcs);
-      }
+      level.Request("number of partitions", repartheurFactory.get());
+      repartheurFactory->Build(level);
+      numProcsA22 = level.Get<int>("number of partitions", repartheurFactory.get());
+      numProcsA22 = std::min(numProcsA22, numProcs);
+    }
 
-      if (rebalanceStriding >= 1) {
-        TEUCHOS_ASSERT(rebalanceStriding*numProcsCoarseA11<=numProcs);
-        TEUCHOS_ASSERT(rebalanceStriding*numProcsA22<=numProcs);
-        if (rebalanceStriding*(numProcsCoarseA11+numProcsA22)>numProcs) {
-          GetOStream(Warnings0) << solverName_+"::compute(): Disabling striding = " << rebalanceStriding << ", since coarseA11 needs " << numProcsCoarseA11
-                                << " procs and A22 needs " << numProcsA22 << " procs."<< std::endl;
-          rebalanceStriding = -1;
-        }
-        int lclBadMatrixDistribution = (coarseA11_->getLocalNumEntries() == 0) || (Dk_1_->getDomainMap()->getLocalNumElements() == 0);
-        int gblBadMatrixDistribution = false;
-        MueLu_maxAll(SM_Matrix_->getDomainMap()->getComm(), lclBadMatrixDistribution, gblBadMatrixDistribution);
-        if (gblBadMatrixDistribution) {
-          GetOStream(Warnings0) << solverName_+"::compute(): Disabling striding = " << rebalanceStriding << ", since coarseA11 has no entries on at least one rank or Dk_1's domain map has no entries on at least one rank."<< std::endl;
-          rebalanceStriding = -1;
-        }
+    if (rebalanceStriding >= 1) {
+      TEUCHOS_ASSERT(rebalanceStriding * numProcsCoarseA11 <= numProcs);
+      TEUCHOS_ASSERT(rebalanceStriding * numProcsA22 <= numProcs);
+      if (rebalanceStriding * (numProcsCoarseA11 + numProcsA22) > numProcs) {
+        GetOStream(Warnings0) << solverName_ + "::compute(): Disabling striding = " << rebalanceStriding << ", since coarseA11 needs " << numProcsCoarseA11
+                              << " procs and A22 needs " << numProcsA22 << " procs." << std::endl;
+        rebalanceStriding = -1;
       }
-
-      if ((numProcsCoarseA11 < 0) || (numProcsA22 < 0) || (numProcsCoarseA11 + numProcsA22 > numProcs)) {
-        GetOStream(Warnings0) << solverName_+"::compute(): Disabling rebalancing of subsolves, since partition heuristic resulted "
-                              << "in undesirable number of partitions: " << numProcsCoarseA11 << ", " << numProcsA22 << std::endl;
-        doRebalancing = false;
+      int lclBadMatrixDistribution = (coarseA11_->getLocalNumEntries() == 0) || (Dk_1_->getDomainMap()->getLocalNumElements() == 0);
+      int gblBadMatrixDistribution = false;
+      MueLu_maxAll(SM_Matrix_->getDomainMap()->getComm(), lclBadMatrixDistribution, gblBadMatrixDistribution);
+      if (gblBadMatrixDistribution) {
+        GetOStream(Warnings0) << solverName_ + "::compute(): Disabling striding = " << rebalanceStriding << ", since coarseA11 has no entries on at least one rank or Dk_1's domain map has no entries on at least one rank." << std::endl;
+        rebalanceStriding = -1;
       }
     }
-#else
-    doRebalancing = false;
-#endif // HAVE_MPI
-  }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  buildAddon(const int spaceNumber) {
-
-    if (spaceNumber == 0)
-      return Teuchos::null;
 
-    std::string timerLabel;
-    if (spaceNumber == spaceNumber_) {
-      if (skipFirst11Level_)
-        timerLabel = "Build coarse addon matrix 11";
-      else
-        timerLabel = "Build addon matrix 11";
-    } else
-      timerLabel = "Build addon matrix 22";
+    if ((numProcsCoarseA11 < 0) || (numProcsA22 < 0) || (numProcsCoarseA11 + numProcsA22 > numProcs)) {
+      GetOStream(Warnings0) << solverName_ + "::compute(): Disabling rebalancing of subsolves, since partition heuristic resulted "
+                            << "in undesirable number of partitions: " << numProcsCoarseA11 << ", " << numProcsA22 << std::endl;
+      doRebalancing = false;
+    }
+  }
+#else
+  doRebalancing = false;
+#endif  // HAVE_MPI
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    buildAddon(const int spaceNumber) {
+  if (spaceNumber == 0)
+    return Teuchos::null;
+
+  std::string timerLabel;
+  if (spaceNumber == spaceNumber_) {
+    if (skipFirst11Level_)
+      timerLabel = "Build coarse addon matrix 11";
+    else
+      timerLabel = "Build addon matrix 11";
+  } else
+    timerLabel = "Build addon matrix 22";
+
+  RCP<Teuchos::TimeMonitor> tmAddon = getTimer(timerLabel);
+
+  RCP<Matrix> addon;
+  RCP<Matrix> Z;
+  RCP<Matrix> lumpedInverse;
+  if (spaceNumber == spaceNumber_) {
+    // catch a failure
+    TEUCHOS_TEST_FOR_EXCEPTION(invMk_1_invBeta_ == Teuchos::null, std::invalid_argument,
+                               solverName_ +
+                                   "::buildCoarse11Matrix(): Inverse of "
+                                   "lumped mass matrix required for add-on (i.e. invMk_1_invBeta_ is null)");
+    lumpedInverse = invMk_1_invBeta_;
+
+    if (skipFirst11Level_) {
+      // construct Zaux = M1 P11
+      RCP<Matrix> Zaux;
+      Zaux = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Mk_one_, false, *P11_, false, Zaux, GetOStream(Runtime0), true, true);
+      // construct Z = D* M1 P11 = D^T Zaux
+      Z = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_, true, *Zaux, false, Z, GetOStream(Runtime0), true, true);
+    } else {
+      // construct Z = D* M1 P11 = D^T Zaux
+      Z = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_, true, *Mk_one_, false, Z, GetOStream(Runtime0), true, true);
+    }
 
-    RCP<Teuchos::TimeMonitor> tmAddon = getTimer(timerLabel);
+  } else if (spaceNumber == spaceNumber_ - 1) {
+    // catch a failure
+    TEUCHOS_TEST_FOR_EXCEPTION(invMk_2_invAlpha_ == Teuchos::null, std::invalid_argument,
+                               solverName_ +
+                                   "::buildCoarse11Matrix(): Inverse of "
+                                   "lumped mass matrix required for add-on (i.e. invMk_2_invAlpha_ is null)");
+    lumpedInverse = invMk_2_invAlpha_;
 
-    RCP<Matrix> addon;
-    RCP<Matrix> Z;
-    RCP<Matrix> lumpedInverse;
-    if (spaceNumber == spaceNumber_) {
-      // catch a failure
-      TEUCHOS_TEST_FOR_EXCEPTION(invMk_1_invBeta_==Teuchos::null,std::invalid_argument,
-                                 solverName_+"::buildCoarse11Matrix(): Inverse of "
-                                 "lumped mass matrix required for add-on (i.e. invMk_1_invBeta_ is null)");
-      lumpedInverse = invMk_1_invBeta_;
+    // construct Z = Dk_2^T Mk_1_one
+    Z = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_2_, true, *Mk_1_one_, false, Z, GetOStream(Runtime0), true, true);
+  }
 
-      if (skipFirst11Level_) {
-        // construct Zaux = M1 P11
-        RCP<Matrix> Zaux;
-        Zaux = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Mk_one_,false,*P11_,false,Zaux,GetOStream(Runtime0),true,true);
-        // construct Z = D* M1 P11 = D^T Zaux
-        Z = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_,true,*Zaux,false,Z,GetOStream(Runtime0),true,true);
-      } else {
-        // construct Z = D* M1 P11 = D^T Zaux
-        Z = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_,true,*Mk_one_,false,Z,GetOStream(Runtime0),true,true);
+  // construct Z^T lumpedInverse Z
+  if (lumpedInverse->getGlobalMaxNumRowEntries() <= 1) {
+    // We assume that if lumpedInverse has at most one entry per row then
+    // these are all diagonal entries.
+    RCP<Vector> diag = VectorFactory::Build(lumpedInverse->getRowMap());
+    lumpedInverse->getLocalDiagCopy(*diag);
+    {
+      ArrayRCP<Scalar> diagVals = diag->getDataNonConst(0);
+      for (size_t j = 0; j < diag->getMap()->getLocalNumElements(); j++) {
+        diagVals[j] = Teuchos::ScalarTraits<Scalar>::squareroot(diagVals[j]);
       }
-
-    } else if (spaceNumber == spaceNumber_-1) {
-      // catch a failure
-      TEUCHOS_TEST_FOR_EXCEPTION(invMk_2_invAlpha_==Teuchos::null,std::invalid_argument,
-                                 solverName_+"::buildCoarse11Matrix(): Inverse of "
-                                 "lumped mass matrix required for add-on (i.e. invMk_2_invAlpha_ is null)");
-      lumpedInverse = invMk_2_invAlpha_;
-
-      // construct Z = Dk_2^T Mk_1_one
-      Z = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_2_,true,*Mk_1_one_,false,Z,GetOStream(Runtime0),true,true);
-
     }
-
-    // construct Z^T lumpedInverse Z
-    if (lumpedInverse->getGlobalMaxNumRowEntries()<=1) {
-      // We assume that if lumpedInverse has at most one entry per row then
-      // these are all diagonal entries.
-      RCP<Vector> diag = VectorFactory::Build(lumpedInverse->getRowMap());
-      lumpedInverse->getLocalDiagCopy(*diag);
-      {
-        ArrayRCP<Scalar> diagVals = diag->getDataNonConst(0);
-        for (size_t j=0; j < diag->getMap()->getLocalNumElements(); j++) {
-          diagVals[j] = Teuchos::ScalarTraits<Scalar>::squareroot(diagVals[j]);
-        }
-      }
-      if (Z->getRowMap()->isSameAs(*(diag->getMap())))
-        Z->leftScale(*diag);
-      else {
-        RCP<Import> importer = ImportFactory::Build(diag->getMap(),Z->getRowMap());
-        RCP<Vector> diag2 = VectorFactory::Build(Z->getRowMap());
-        diag2->doImport(*diag,*importer,Xpetra::INSERT);
-        Z->leftScale(*diag2);
-      }
-      addon = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Z,true,*Z,false,addon,GetOStream(Runtime0),true,true);
-    } else if (parameterList_.get<bool>("rap: triple product", false) == false) {
-      RCP<Matrix> C2;
-      // construct C2 = lumpedInverse Z
-      C2 = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*lumpedInverse,false,*Z,false,C2,GetOStream(Runtime0),true,true);
-      // construct Matrix2 = Z* M0inv Z = Z* C2
-      addon = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Z,true,*C2,false,addon,GetOStream(Runtime0),true,true);
-    } else {
-      addon = MatrixFactory::Build(Z->getDomainMap());
-      // construct Matrix2 = Z^T lumpedInverse Z
-      Xpetra::TripleMatrixMultiply<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
+    if (Z->getRowMap()->isSameAs(*(diag->getMap())))
+      Z->leftScale(*diag);
+    else {
+      RCP<Import> importer = ImportFactory::Build(diag->getMap(), Z->getRowMap());
+      RCP<Vector> diag2    = VectorFactory::Build(Z->getRowMap());
+      diag2->doImport(*diag, *importer, Xpetra::INSERT);
+      Z->leftScale(*diag2);
+    }
+    addon = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Z, true, *Z, false, addon, GetOStream(Runtime0), true, true);
+  } else if (parameterList_.get<bool>("rap: triple product", false) == false) {
+    RCP<Matrix> C2;
+    // construct C2 = lumpedInverse Z
+    C2 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*lumpedInverse, false, *Z, false, C2, GetOStream(Runtime0), true, true);
+    // construct Matrix2 = Z* M0inv Z = Z* C2
+    addon = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Z, true, *C2, false, addon, GetOStream(Runtime0), true, true);
+  } else {
+    addon = MatrixFactory::Build(Z->getDomainMap());
+    // construct Matrix2 = Z^T lumpedInverse Z
+    Xpetra::TripleMatrixMultiply<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
         MultiplyRAP(*Z, true, *lumpedInverse, false, *Z, false, *addon, true, true);
-    }
-    return addon;
+  }
+  return addon;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildCoarse11Matrix() {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("Build coarse (1,1) matrix");
+
+  const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  // coarse matrix for P11* (M1 + D1* M2 D1) P11
+  RCP<Matrix> temp;
+  temp = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*SM_Matrix_, false, *P11_, false, temp, GetOStream(Runtime0), true, true);
+  if (ImporterCoarse11_.is_null())
+    coarseA11_ = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P11_, true, *temp, false, coarseA11_, GetOStream(Runtime0), true, true);
+  else {
+    RCP<Matrix> temp2;
+    temp2 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P11_, true, *temp, false, temp2, GetOStream(Runtime0), true, true);
+
+    RCP<const Map> map = ImporterCoarse11_->getTargetMap()->removeEmptyProcesses();
+    temp2->removeEmptyProcessesInPlace(map);
+    if (!temp2.is_null() && temp2->getRowMap().is_null())
+      temp2 = Teuchos::null;
+    coarseA11_ = temp2;
   }
 
+  if (!disable_addon_) {
+    RCP<Matrix> addon;
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::buildCoarse11Matrix() {
-    RCP<Teuchos::TimeMonitor> tm = getTimer("Build coarse (1,1) matrix");
-
-    const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    // coarse matrix for P11* (M1 + D1* M2 D1) P11
-    RCP<Matrix> temp;
-    temp = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*SM_Matrix_,false,*P11_,false,temp,GetOStream(Runtime0),true,true);
-    if (ImporterCoarse11_.is_null())
-      coarseA11_ = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*P11_,true,*temp,false,coarseA11_,GetOStream(Runtime0),true,true);
-    else {
-
-      RCP<Matrix> temp2;
-      temp2 = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*P11_,true,*temp,false,temp2,GetOStream(Runtime0),true,true);
+    if (!coarseA11_.is_null() && Addon11_.is_null()) {
+      addon = buildAddon(spaceNumber_);
+      // Should we keep the addon for next setup?
+      if (enable_reuse_)
+        Addon11_ = addon;
+    } else
+      addon = Addon11_;
 
-      RCP<const Map> map = ImporterCoarse11_->getTargetMap()->removeEmptyProcesses();
-      temp2->removeEmptyProcessesInPlace(map);
-      if (!temp2.is_null() && temp2->getRowMap().is_null())
-        temp2 = Teuchos::null;
-      coarseA11_ = temp2;
+    if (!coarseA11_.is_null()) {
+      // add matrices together
+      RCP<Matrix> newCoarseA11;
+      Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*coarseA11_, false, one, *addon, false, one, newCoarseA11, GetOStream(Runtime0));
+      newCoarseA11->fillComplete();
+      coarseA11_ = newCoarseA11;
     }
+  }
 
-    if (!disable_addon_) {
+  if (!coarseA11_.is_null() && !skipFirst11Level_) {
+    ArrayRCP<bool> coarseA11BCrows;
+    coarseA11BCrows.resize(coarseA11_->getRowMap()->getLocalNumElements());
+    for (size_t i = 0; i < BCdomain22_.size(); i++)
+      for (size_t k = 0; k < dim_; k++)
+        coarseA11BCrows[i * dim_ + k] = BCdomain22_(i);
+    magnitudeType rowSumTol = parameterList_.get<double>("refmaxwell: row sum drop tol (1,1)");
+    if (rowSumTol > 0.)
+      Utilities::ApplyRowSumCriterion(*coarseA11_, rowSumTol, coarseA11BCrows);
+    if (applyBCsToCoarse11_)
+      Utilities::ApplyOAZToMatrixRows(coarseA11_, coarseA11BCrows);
+  }
 
-      RCP<Matrix> addon;
+  if (!coarseA11_.is_null()) {
+    // If we already applied BCs to A_nodal, we likely do not need
+    // to fix up coarseA11.
+    // If we did not apply BCs to A_nodal, we now need to correct
+    // the zero diagonals of coarseA11, since we did nuke the nullspace.
+
+    bool fixZeroDiagonal = !applyBCsToAnodal_;
+    if (precList11_.isParameter("rap: fix zero diagonals"))
+      fixZeroDiagonal = precList11_.get<bool>("rap: fix zero diagonals");
+
+    if (fixZeroDiagonal) {
+      magnitudeType threshold = 1e-16;
+      Scalar replacement      = 1.0;
+      if (precList11_.isType<magnitudeType>("rap: fix zero diagonals threshold"))
+        threshold = precList11_.get<magnitudeType>("rap: fix zero diagonals threshold");
+      else if (precList11_.isType<double>("rap: fix zero diagonals threshold"))
+        threshold = Teuchos::as<magnitudeType>(precList11_.get<double>("rap: fix zero diagonals threshold"));
+      if (precList11_.isType<double>("rap: fix zero diagonals replacement"))
+        replacement = Teuchos::as<Scalar>(precList11_.get<double>("rap: fix zero diagonals replacement"));
+      Xpetra::MatrixUtils<SC, LO, GO, NO>::CheckRepairMainDiagonal(coarseA11_, true, GetOStream(Warnings1), threshold, replacement);
+    }
+
+    // Set block size
+    coarseA11_->SetFixedBlockSize(dim_);
+    if (skipFirst11Level_)
+      coarseA11_->setObjectLabel(solverName_ + " coarse (1,1)");
+    else
+      coarseA11_->setObjectLabel(solverName_ + " (1,1)");
+  }
+}
 
-      if (!coarseA11_.is_null() && Addon11_.is_null()) {
-        addon = buildAddon(spaceNumber_);
-        // Should we keep the addon for next setup?
-        if (enable_reuse_)
-          Addon11_ = addon;
-      } else
-        addon = Addon11_;
-
-      if (!coarseA11_.is_null()) {
-        // add matrices together
-        RCP<Matrix> newCoarseA11;
-        Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::TwoMatrixAdd(*coarseA11_,false,one,*addon,false,one,newCoarseA11,GetOStream(Runtime0));
-        newCoarseA11->fillComplete();
-        coarseA11_ = newCoarseA11;
-      }
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    rebalanceCoarse11Matrix(const int rebalanceStriding, const int numProcsCoarseA11) {
+#ifdef HAVE_MPI
+  // rebalance coarseA11
+  RCP<Teuchos::TimeMonitor> tm = getTimer("Rebalance coarseA11");
+
+  Level fineLevel, coarseLevel;
+  fineLevel.SetFactoryManager(null);
+  coarseLevel.SetFactoryManager(null);
+  coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
+  fineLevel.SetLevelID(0);
+  coarseLevel.SetLevelID(1);
+  coarseLevel.Set("A", coarseA11_);
+  coarseLevel.Set("P", P11_);
+  coarseLevel.Set("Coordinates", CoordsCoarse11_);
+  if (!NullspaceCoarse11_.is_null())
+    coarseLevel.Set("Nullspace", NullspaceCoarse11_);
+  coarseLevel.Set("number of partitions", numProcsCoarseA11);
+  coarseLevel.Set("repartition: heuristic target rows per process", 1000);
+
+  coarseLevel.setlib(coarseA11_->getDomainMap()->lib());
+  fineLevel.setlib(coarseA11_->getDomainMap()->lib());
+  coarseLevel.setObjectLabel(solverName_ + " coarse (1,1)");
+  fineLevel.setObjectLabel(solverName_ + " coarse (1,1)");
+
+  std::string partName = precList11_.get<std::string>("repartition: partitioner", "zoltan2");
+  RCP<Factory> partitioner;
+  if (partName == "zoltan") {
+#ifdef HAVE_MUELU_ZOLTAN
+    partitioner = rcp(new ZoltanInterface());
+    // NOTE: ZoltanInteface ("zoltan") does not support external parameters through ParameterList
+    // partitioner->SetFactory("number of partitions", repartheurFactory);
+#else
+    throw Exceptions::RuntimeError("Zoltan interface is not available");
+#endif
+  } else if (partName == "zoltan2") {
+#ifdef HAVE_MUELU_ZOLTAN2
+    partitioner = rcp(new Zoltan2Interface());
+    ParameterList partParams;
+    RCP<const ParameterList> partpartParams = rcp(new ParameterList(precList11_.sublist("repartition: params", false)));
+    partParams.set("ParameterList", partpartParams);
+    partitioner->SetParameterList(partParams);
+    // partitioner->SetFactory("number of partitions", repartheurFactory);
+#else
+    throw Exceptions::RuntimeError("Zoltan2 interface is not available");
+#endif
+  }
 
-    if (!coarseA11_.is_null() && !skipFirst11Level_) {
-      ArrayRCP<bool> coarseA11BCrows;
-      coarseA11BCrows.resize(coarseA11_->getRowMap()->getLocalNumElements());
-      for (size_t i = 0; i < BCdomain22_.size(); i++)
-        for (size_t k = 0; k < dim_; k++)
-          coarseA11BCrows[i*dim_+k] = BCdomain22_(i);
-      magnitudeType rowSumTol = parameterList_.get<double>("refmaxwell: row sum drop tol (1,1)");
-      if (rowSumTol > 0.)
-        Utilities::ApplyRowSumCriterion(*coarseA11_, rowSumTol, coarseA11BCrows);
-      if (applyBCsToCoarse11_)
-        Utilities::ApplyOAZToMatrixRows(coarseA11_, coarseA11BCrows);
-    }
+  auto repartFactory = rcp(new RepartitionFactory());
+  ParameterList repartParams;
+  repartParams.set("repartition: print partition distribution", precList11_.get<bool>("repartition: print partition distribution", false));
+  repartParams.set("repartition: remap parts", precList11_.get<bool>("repartition: remap parts", true));
+  if (rebalanceStriding >= 1) {
+    bool acceptPart = (SM_Matrix_->getDomainMap()->getComm()->getRank() % rebalanceStriding) == 0;
+    if (SM_Matrix_->getDomainMap()->getComm()->getRank() >= numProcsCoarseA11 * rebalanceStriding)
+      acceptPart = false;
+    repartParams.set("repartition: remap accept partition", acceptPart);
+  }
+  repartFactory->SetParameterList(repartParams);
+  // repartFactory->SetFactory("number of partitions", repartheurFactory);
+  repartFactory->SetFactory("Partition", partitioner);
+
+  auto newP = rcp(new RebalanceTransferFactory());
+  ParameterList newPparams;
+  newPparams.set("type", "Interpolation");
+  newPparams.set("repartition: rebalance P and R", precList11_.get<bool>("repartition: rebalance P and R", false));
+  newPparams.set("repartition: use subcommunicators", true);
+  newPparams.set("repartition: rebalance Nullspace", !NullspaceCoarse11_.is_null());
+  newP->SetFactory("Coordinates", NoFactory::getRCP());
+  if (!NullspaceCoarse11_.is_null())
+    newP->SetFactory("Nullspace", NoFactory::getRCP());
+  newP->SetParameterList(newPparams);
+  newP->SetFactory("Importer", repartFactory);
+
+  auto newA = rcp(new RebalanceAcFactory());
+  ParameterList rebAcParams;
+  rebAcParams.set("repartition: use subcommunicators", true);
+  newA->SetParameterList(rebAcParams);
+  newA->SetFactory("Importer", repartFactory);
+
+  coarseLevel.Request("P", newP.get());
+  coarseLevel.Request("Importer", repartFactory.get());
+  coarseLevel.Request("A", newA.get());
+  coarseLevel.Request("Coordinates", newP.get());
+  if (!NullspaceCoarse11_.is_null())
+    coarseLevel.Request("Nullspace", newP.get());
+  repartFactory->Build(coarseLevel);
+
+  if (!precList11_.get<bool>("repartition: rebalance P and R", false))
+    ImporterCoarse11_ = coarseLevel.Get<RCP<const Import> >("Importer", repartFactory.get());
+  P11_            = coarseLevel.Get<RCP<Matrix> >("P", newP.get());
+  coarseA11_      = coarseLevel.Get<RCP<Matrix> >("A", newA.get());
+  CoordsCoarse11_ = coarseLevel.Get<RCP<RealValuedMultiVector> >("Coordinates", newP.get());
+  if (!NullspaceCoarse11_.is_null())
+    NullspaceCoarse11_ = coarseLevel.Get<RCP<MultiVector> >("Nullspace", newP.get());
+
+  if (!coarseA11_.is_null()) {
+    // Set block size
+    coarseA11_->SetFixedBlockSize(dim_);
+    if (skipFirst11Level_)
+      coarseA11_->setObjectLabel(solverName_ + " coarse (1,1)");
+    else
+      coarseA11_->setObjectLabel(solverName_ + " (1,1)");
+  }
 
-    if (!coarseA11_.is_null()) {
-      // If we already applied BCs to A_nodal, we likely do not need
-      // to fix up coarseA11.
-      // If we did not apply BCs to A_nodal, we now need to correct
-      // the zero diagonals of coarseA11, since we did nuke the nullspace.
-
-      bool fixZeroDiagonal = !applyBCsToAnodal_;
-      if (precList11_.isParameter("rap: fix zero diagonals"))
-        fixZeroDiagonal = precList11_.get<bool>("rap: fix zero diagonals");
-
-      if (fixZeroDiagonal) {
-        magnitudeType threshold = 1e-16;
-        Scalar replacement = 1.0;
-        if (precList11_.isType<magnitudeType>("rap: fix zero diagonals threshold"))
-          threshold = precList11_.get<magnitudeType>("rap: fix zero diagonals threshold");
-        else if (precList11_.isType<double>("rap: fix zero diagonals threshold"))
-          threshold = Teuchos::as<magnitudeType>(precList11_.get<double>("rap: fix zero diagonals threshold"));
-        if (precList11_.isType<double>("rap: fix zero diagonals replacement"))
-          replacement = Teuchos::as<Scalar>(precList11_.get<double>("rap: fix zero diagonals replacement"));
-        Xpetra::MatrixUtils<SC,LO,GO,NO>::CheckRepairMainDiagonal(coarseA11_, true, GetOStream(Warnings1), threshold, replacement);
-      }
+  coarseA11_AP_reuse_data_  = Teuchos::null;
+  coarseA11_RAP_reuse_data_ = Teuchos::null;
 
-      // Set block size
-      coarseA11_->SetFixedBlockSize(dim_);
-      if (skipFirst11Level_)
-        coarseA11_->setObjectLabel(solverName_+" coarse (1,1)");
-      else
-        coarseA11_->setObjectLabel(solverName_+" (1,1)");
-    }
+  if (!disable_addon_ && enable_reuse_) {
+    // Rebalance the addon for next setup
+    RCP<const Import> ImporterCoarse11 = coarseLevel.Get<RCP<const Import> >("Importer", repartFactory.get());
+    RCP<const Map> targetMap           = ImporterCoarse11->getTargetMap();
+    ParameterList XpetraList;
+    XpetraList.set("Restrict Communicator", true);
+    Addon11_ = MatrixFactory::Build(Addon11_, *ImporterCoarse11, *ImporterCoarse11, targetMap, targetMap, rcp(&XpetraList, false));
   }
+#endif
+}
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  rebalanceCoarse11Matrix (const int rebalanceStriding, const int numProcsCoarseA11) {
-#ifdef HAVE_MPI
-    // rebalance coarseA11
-    RCP<Teuchos::TimeMonitor> tm = getTimer("Rebalance coarseA11");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::build22Matrix(const bool reuse, const bool doRebalancing, const int rebalanceStriding, const int numProcsA22) {
+  if (!reuse) {  // build fine grid operator for (2,2)-block, Dk_1^T SM Dk_1  (aka TMT)
+    RCP<Teuchos::TimeMonitor> tm = getTimer("Build A22");
 
     Level fineLevel, coarseLevel;
     fineLevel.SetFactoryManager(null);
@@ -949,1980 +1044,1807 @@ namespace MueLu {
     coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
     fineLevel.SetLevelID(0);
     coarseLevel.SetLevelID(1);
-    coarseLevel.Set("A",coarseA11_);
-    coarseLevel.Set("P",P11_);
-    coarseLevel.Set("Coordinates",CoordsCoarse11_);
-    if (!NullspaceCoarse11_.is_null())
-      coarseLevel.Set("Nullspace",NullspaceCoarse11_);
-    coarseLevel.Set("number of partitions", numProcsCoarseA11);
-    coarseLevel.Set("repartition: heuristic target rows per process", 1000);
-
-    coarseLevel.setlib(coarseA11_->getDomainMap()->lib());
-    fineLevel.setlib(coarseA11_->getDomainMap()->lib());
-    coarseLevel.setObjectLabel(solverName_+" coarse (1,1)");
-    fineLevel.setObjectLabel(solverName_+" coarse (1,1)");
-
-    std::string partName = precList11_.get<std::string>("repartition: partitioner", "zoltan2");
-    RCP<Factory> partitioner;
-    if (partName == "zoltan") {
-#ifdef HAVE_MUELU_ZOLTAN
-      partitioner = rcp(new ZoltanInterface());
-      // NOTE: ZoltanInteface ("zoltan") does not support external parameters through ParameterList
-      // partitioner->SetFactory("number of partitions", repartheurFactory);
-#else
-      throw Exceptions::RuntimeError("Zoltan interface is not available");
-#endif
-    } else if (partName == "zoltan2") {
-#ifdef HAVE_MUELU_ZOLTAN2
-      partitioner = rcp(new Zoltan2Interface());
-      ParameterList partParams;
-      RCP<const ParameterList> partpartParams = rcp(new ParameterList(precList11_.sublist("repartition: params", false)));
-      partParams.set("ParameterList", partpartParams);
-      partitioner->SetParameterList(partParams);
-      // partitioner->SetFactory("number of partitions", repartheurFactory);
-#else
-      throw Exceptions::RuntimeError("Zoltan2 interface is not available");
-#endif
-    }
+    fineLevel.Set("A", SM_Matrix_);
+    coarseLevel.Set("P", Dk_1_);
+    coarseLevel.Set("Coordinates", Coords22_);
 
-    auto repartFactory = rcp(new RepartitionFactory());
-    ParameterList repartParams;
-    repartParams.set("repartition: print partition distribution", precList11_.get<bool>("repartition: print partition distribution", false));
-    repartParams.set("repartition: remap parts", precList11_.get<bool>("repartition: remap parts", true));
-    if (rebalanceStriding >= 1) {
-      bool acceptPart = (SM_Matrix_->getDomainMap()->getComm()->getRank() % rebalanceStriding) == 0;
-      if (SM_Matrix_->getDomainMap()->getComm()->getRank() >= numProcsCoarseA11*rebalanceStriding)
-        acceptPart = false;
-      repartParams.set("repartition: remap accept partition", acceptPart);
-    }
-    repartFactory->SetParameterList(repartParams);
-    // repartFactory->SetFactory("number of partitions", repartheurFactory);
-    repartFactory->SetFactory("Partition", partitioner);
-
-    auto newP = rcp(new RebalanceTransferFactory());
-    ParameterList newPparams;
-    newPparams.set("type", "Interpolation");
-    newPparams.set("repartition: rebalance P and R", precList11_.get<bool>("repartition: rebalance P and R", false));
-    newPparams.set("repartition: use subcommunicators", true);
-    newPparams.set("repartition: rebalance Nullspace", !NullspaceCoarse11_.is_null());
-    newP->SetFactory("Coordinates", NoFactory::getRCP());
-    if (!NullspaceCoarse11_.is_null())
-      newP->SetFactory("Nullspace", NoFactory::getRCP());
-    newP->SetParameterList(newPparams);
-    newP->SetFactory("Importer", repartFactory);
-
-    auto newA = rcp(new RebalanceAcFactory());
-    ParameterList rebAcParams;
-    rebAcParams.set("repartition: use subcommunicators", true);
-    newA->SetParameterList(rebAcParams);
-    newA->SetFactory("Importer", repartFactory);
-
-    coarseLevel.Request("P", newP.get());
-    coarseLevel.Request("Importer", repartFactory.get());
-    coarseLevel.Request("A", newA.get());
-    coarseLevel.Request("Coordinates", newP.get());
-    if (!NullspaceCoarse11_.is_null())
-      coarseLevel.Request("Nullspace", newP.get());
-    repartFactory->Build(coarseLevel);
-
-    if (!precList11_.get<bool>("repartition: rebalance P and R", false))
-      ImporterCoarse11_ = coarseLevel.Get< RCP<const Import> >("Importer", repartFactory.get());
-    P11_ = coarseLevel.Get< RCP<Matrix> >("P", newP.get());
-    coarseA11_ = coarseLevel.Get< RCP<Matrix> >("A", newA.get());
-    CoordsCoarse11_ = coarseLevel.Get< RCP<RealValuedMultiVector> >("Coordinates", newP.get());
-    if (!NullspaceCoarse11_.is_null())
-      NullspaceCoarse11_ = coarseLevel.Get< RCP<MultiVector> >("Nullspace", newP.get());
+    coarseLevel.setlib(SM_Matrix_->getDomainMap()->lib());
+    fineLevel.setlib(SM_Matrix_->getDomainMap()->lib());
+    coarseLevel.setObjectLabel(solverName_ + " (2,2)");
+    fineLevel.setObjectLabel(solverName_ + " (2,2)");
 
-    if (!coarseA11_.is_null()) {
-      // Set block size
-      coarseA11_->SetFixedBlockSize(dim_);
-      if (skipFirst11Level_)
-        coarseA11_->setObjectLabel(solverName_+" coarse (1,1)");
-      else
-        coarseA11_->setObjectLabel(solverName_+" (1,1)");
-    }
+    RCP<RAPFactory> rapFact = rcp(new RAPFactory());
+    ParameterList rapList   = *(rapFact->GetValidParameterList());
+    rapList.set("transpose: use implicit", true);
+    rapList.set("rap: fix zero diagonals", parameterList_.get<bool>("rap: fix zero diagonals", true));
+    rapList.set("rap: fix zero diagonals threshold", parameterList_.get<double>("rap: fix zero diagonals threshold", Teuchos::ScalarTraits<double>::eps()));
+    rapList.set("rap: triple product", parameterList_.get<bool>("rap: triple product", false));
+    rapFact->SetParameterList(rapList);
 
-    coarseA11_AP_reuse_data_ = Teuchos::null;
-    coarseA11_RAP_reuse_data_ = Teuchos::null;
-
-    if (!disable_addon_ && enable_reuse_) {
-      // Rebalance the addon for next setup
-      RCP<const Import> ImporterCoarse11 = coarseLevel.Get< RCP<const Import> >("Importer", repartFactory.get());
-      RCP<const Map> targetMap = ImporterCoarse11->getTargetMap();
-      ParameterList XpetraList;
-      XpetraList.set("Restrict Communicator",true);
-      Addon11_ = MatrixFactory::Build(Addon11_, *ImporterCoarse11, *ImporterCoarse11, targetMap, targetMap, rcp(&XpetraList,false));
+    if (!A22_AP_reuse_data_.is_null()) {
+      coarseLevel.AddKeepFlag("AP reuse data", rapFact.get());
+      coarseLevel.Set<Teuchos::RCP<Teuchos::ParameterList> >("AP reuse data", A22_AP_reuse_data_, rapFact.get());
+    }
+    if (!A22_RAP_reuse_data_.is_null()) {
+      coarseLevel.AddKeepFlag("RAP reuse data", rapFact.get());
+      coarseLevel.Set<Teuchos::RCP<Teuchos::ParameterList> >("RAP reuse data", A22_RAP_reuse_data_, rapFact.get());
     }
-#endif
-  }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::build22Matrix (const bool reuse, const bool doRebalancing, const int rebalanceStriding, const int numProcsA22) {
-        
-    if (!reuse) { // build fine grid operator for (2,2)-block, Dk_1^T SM Dk_1  (aka TMT)
-      RCP<Teuchos::TimeMonitor> tm = getTimer("Build A22");
-
-      Level fineLevel, coarseLevel;
-      fineLevel.SetFactoryManager(null);
-      coarseLevel.SetFactoryManager(null);
-      coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
-      fineLevel.SetLevelID(0);
-      coarseLevel.SetLevelID(1);
-      fineLevel.Set("A",SM_Matrix_);
-      coarseLevel.Set("P",Dk_1_);
-      coarseLevel.Set("Coordinates",Coords22_);
-
-      coarseLevel.setlib(SM_Matrix_->getDomainMap()->lib());
-      fineLevel.setlib(SM_Matrix_->getDomainMap()->lib());
-      coarseLevel.setObjectLabel(solverName_+" (2,2)");
-      fineLevel.setObjectLabel(solverName_+" (2,2)");
-
-      RCP<RAPFactory> rapFact = rcp(new RAPFactory());
-      ParameterList rapList = *(rapFact->GetValidParameterList());
-      rapList.set("transpose: use implicit", true);
-      rapList.set("rap: fix zero diagonals", parameterList_.get<bool>("rap: fix zero diagonals", true));
-      rapList.set("rap: fix zero diagonals threshold", parameterList_.get<double>("rap: fix zero diagonals threshold", Teuchos::ScalarTraits<double>::eps()));
-      rapList.set("rap: triple product", parameterList_.get<bool>("rap: triple product", false));
-      rapFact->SetParameterList(rapList);
-
-      if (!A22_AP_reuse_data_.is_null()) {
-        coarseLevel.AddKeepFlag("AP reuse data", rapFact.get());
-        coarseLevel.Set<Teuchos::RCP<Teuchos::ParameterList> >("AP reuse data", A22_AP_reuse_data_, rapFact.get());
-      }
-      if (!A22_RAP_reuse_data_.is_null()) {
-        coarseLevel.AddKeepFlag("RAP reuse data", rapFact.get());
-        coarseLevel.Set<Teuchos::RCP<Teuchos::ParameterList> >("RAP reuse data", A22_RAP_reuse_data_, rapFact.get());
-      }
 
 #ifdef HAVE_MPI
-      if (doRebalancing) {
-
-        coarseLevel.Set("number of partitions", numProcsA22);
-        coarseLevel.Set("repartition: heuristic target rows per process", 1000);
+    if (doRebalancing) {
+      coarseLevel.Set("number of partitions", numProcsA22);
+      coarseLevel.Set("repartition: heuristic target rows per process", 1000);
 
-        std::string partName = precList22_.get<std::string>("repartition: partitioner", "zoltan2");
-        RCP<Factory> partitioner;
-        if (partName == "zoltan") {
+      std::string partName = precList22_.get<std::string>("repartition: partitioner", "zoltan2");
+      RCP<Factory> partitioner;
+      if (partName == "zoltan") {
 #ifdef HAVE_MUELU_ZOLTAN
-          partitioner = rcp(new ZoltanInterface());
-          partitioner->SetFactory("A", rapFact);
-          // partitioner->SetFactory("number of partitions", repartheurFactory);
-          // NOTE: ZoltanInteface ("zoltan") does not support external parameters through ParameterList
+        partitioner = rcp(new ZoltanInterface());
+        partitioner->SetFactory("A", rapFact);
+        // partitioner->SetFactory("number of partitions", repartheurFactory);
+        // NOTE: ZoltanInteface ("zoltan") does not support external parameters through ParameterList
 #else
-          throw Exceptions::RuntimeError("Zoltan interface is not available");
+        throw Exceptions::RuntimeError("Zoltan interface is not available");
 #endif
-        } else if (partName == "zoltan2") {
+      } else if (partName == "zoltan2") {
 #ifdef HAVE_MUELU_ZOLTAN2
-          partitioner = rcp(new Zoltan2Interface());
-          ParameterList partParams;
-          RCP<const ParameterList> partpartParams = rcp(new ParameterList(precList22_.sublist("repartition: params", false)));
-          partParams.set("ParameterList", partpartParams);
-          partitioner->SetParameterList(partParams);
-          partitioner->SetFactory("A", rapFact);
-          // partitioner->SetFactory("number of partitions", repartheurFactory);
+        partitioner = rcp(new Zoltan2Interface());
+        ParameterList partParams;
+        RCP<const ParameterList> partpartParams = rcp(new ParameterList(precList22_.sublist("repartition: params", false)));
+        partParams.set("ParameterList", partpartParams);
+        partitioner->SetParameterList(partParams);
+        partitioner->SetFactory("A", rapFact);
+        // partitioner->SetFactory("number of partitions", repartheurFactory);
 #else
-          throw Exceptions::RuntimeError("Zoltan2 interface is not available");
+        throw Exceptions::RuntimeError("Zoltan2 interface is not available");
 #endif
-        }
-
-        auto repartFactory = rcp(new RepartitionFactory());
-        ParameterList repartParams;
-        repartParams.set("repartition: print partition distribution", precList22_.get<bool>("repartition: print partition distribution", false));
-        repartParams.set("repartition: remap parts", precList22_.get<bool>("repartition: remap parts", true));
-        if (rebalanceStriding >= 1) {
-          bool acceptPart = ((SM_Matrix_->getDomainMap()->getComm()->getSize()-1-SM_Matrix_->getDomainMap()->getComm()->getRank()) % rebalanceStriding) == 0;
-          if (SM_Matrix_->getDomainMap()->getComm()->getSize()-1-SM_Matrix_->getDomainMap()->getComm()->getRank() >= numProcsA22*rebalanceStriding)
-            acceptPart = false;
-          if (acceptPart)
-            TEUCHOS_ASSERT(coarseA11_.is_null());
-          repartParams.set("repartition: remap accept partition", acceptPart);
-        } else
-          repartParams.set("repartition: remap accept partition", coarseA11_.is_null());
-        repartFactory->SetParameterList(repartParams);
-        repartFactory->SetFactory("A", rapFact);
-        // repartFactory->SetFactory("number of partitions", repartheurFactory);
-        repartFactory->SetFactory("Partition", partitioner);
-
-        auto newP = rcp(new RebalanceTransferFactory());
-        ParameterList newPparams;
-        newPparams.set("type", "Interpolation");
-        newPparams.set("repartition: rebalance P and R", precList22_.get<bool>("repartition: rebalance P and R", false));
-        newPparams.set("repartition: use subcommunicators", true);
-        newPparams.set("repartition: rebalance Nullspace", false);
-        newP->SetFactory("Coordinates", NoFactory::getRCP());
-        newP->SetParameterList(newPparams);
-        newP->SetFactory("Importer", repartFactory);
-
-        auto newA = rcp(new RebalanceAcFactory());
-        ParameterList rebAcParams;
-        rebAcParams.set("repartition: use subcommunicators", true);
-        newA->SetParameterList(rebAcParams);
-        newA->SetFactory("A", rapFact);
-        newA->SetFactory("Importer", repartFactory);
-
-        coarseLevel.Request("P", newP.get());
-        coarseLevel.Request("Importer", repartFactory.get());
-        coarseLevel.Request("A", newA.get());
-        coarseLevel.Request("Coordinates", newP.get());
-        rapFact->Build(fineLevel,coarseLevel);
-        repartFactory->Build(coarseLevel);
-
-        if (!precList22_.get<bool>("repartition: rebalance P and R", false))
-          Importer22_ = coarseLevel.Get< RCP<const Import> >("Importer", repartFactory.get());
-        Dk_1_ = coarseLevel.Get< RCP<Matrix> >("P", newP.get());
-        A22_ = coarseLevel.Get< RCP<Matrix> >("A", newA.get());
-        Coords22_ = coarseLevel.Get< RCP<RealValuedMultiVector> >("Coordinates", newP.get());
-
-        if (!P22_.is_null()) {
-          // Todo
-        }
+      }
 
+      auto repartFactory = rcp(new RepartitionFactory());
+      ParameterList repartParams;
+      repartParams.set("repartition: print partition distribution", precList22_.get<bool>("repartition: print partition distribution", false));
+      repartParams.set("repartition: remap parts", precList22_.get<bool>("repartition: remap parts", true));
+      if (rebalanceStriding >= 1) {
+        bool acceptPart = ((SM_Matrix_->getDomainMap()->getComm()->getSize() - 1 - SM_Matrix_->getDomainMap()->getComm()->getRank()) % rebalanceStriding) == 0;
+        if (SM_Matrix_->getDomainMap()->getComm()->getSize() - 1 - SM_Matrix_->getDomainMap()->getComm()->getRank() >= numProcsA22 * rebalanceStriding)
+          acceptPart = false;
+        if (acceptPart)
+          TEUCHOS_ASSERT(coarseA11_.is_null());
+        repartParams.set("repartition: remap accept partition", acceptPart);
       } else
-#endif // HAVE_MPI
-        {
-          coarseLevel.Request("A", rapFact.get());
-          if (enable_reuse_) {
-            coarseLevel.Request("AP reuse data", rapFact.get());
-            coarseLevel.Request("RAP reuse data", rapFact.get());
-          }
+        repartParams.set("repartition: remap accept partition", coarseA11_.is_null());
+      repartFactory->SetParameterList(repartParams);
+      repartFactory->SetFactory("A", rapFact);
+      // repartFactory->SetFactory("number of partitions", repartheurFactory);
+      repartFactory->SetFactory("Partition", partitioner);
+
+      auto newP = rcp(new RebalanceTransferFactory());
+      ParameterList newPparams;
+      newPparams.set("type", "Interpolation");
+      newPparams.set("repartition: rebalance P and R", precList22_.get<bool>("repartition: rebalance P and R", false));
+      newPparams.set("repartition: use subcommunicators", true);
+      newPparams.set("repartition: rebalance Nullspace", false);
+      newP->SetFactory("Coordinates", NoFactory::getRCP());
+      newP->SetParameterList(newPparams);
+      newP->SetFactory("Importer", repartFactory);
+
+      auto newA = rcp(new RebalanceAcFactory());
+      ParameterList rebAcParams;
+      rebAcParams.set("repartition: use subcommunicators", true);
+      newA->SetParameterList(rebAcParams);
+      newA->SetFactory("A", rapFact);
+      newA->SetFactory("Importer", repartFactory);
+
+      coarseLevel.Request("P", newP.get());
+      coarseLevel.Request("Importer", repartFactory.get());
+      coarseLevel.Request("A", newA.get());
+      coarseLevel.Request("Coordinates", newP.get());
+      rapFact->Build(fineLevel, coarseLevel);
+      repartFactory->Build(coarseLevel);
+
+      if (!precList22_.get<bool>("repartition: rebalance P and R", false))
+        Importer22_ = coarseLevel.Get<RCP<const Import> >("Importer", repartFactory.get());
+      Dk_1_     = coarseLevel.Get<RCP<Matrix> >("P", newP.get());
+      A22_      = coarseLevel.Get<RCP<Matrix> >("A", newA.get());
+      Coords22_ = coarseLevel.Get<RCP<RealValuedMultiVector> >("Coordinates", newP.get());
 
-          A22_ = coarseLevel.Get< RCP<Matrix> >("A", rapFact.get());
-
-          if (enable_reuse_) {
-            if (coarseLevel.IsAvailable("AP reuse data", rapFact.get()))
-              A22_AP_reuse_data_ = coarseLevel.Get< RCP<ParameterList> >("AP reuse data", rapFact.get());
-            if (coarseLevel.IsAvailable("RAP reuse data", rapFact.get()))
-              A22_RAP_reuse_data_ = coarseLevel.Get< RCP<ParameterList> >("RAP reuse data", rapFact.get());
-          }
-        }
-    } else {
-      RCP<Teuchos::TimeMonitor> tm = getTimer("Build A22");
-      if (Importer22_.is_null()) {
-        RCP<Matrix> temp;
-        temp = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*SM_Matrix_,false,*Dk_1_,false,temp,GetOStream(Runtime0),true,true);
-        if (!implicitTranspose_)
-          A22_ = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_T_,false,*temp,false,A22_,GetOStream(Runtime0),true,true);
-        else
-          A22_ = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_,true,*temp,false,A22_,GetOStream(Runtime0),true,true);
-      } else {
-        // we replaced domain map and importer on D, reverse that
-        RCP<const Import> Dimporter = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->getCrsGraph()->getImporter();
-        rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(DorigDomainMap_, DorigImporter_);
-
-        RCP<Matrix> temp, temp2;
-        temp = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*SM_Matrix_,false,*Dk_1_,false,temp,GetOStream(Runtime0),true,true);
-        if (!implicitTranspose_)
-          temp2 = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_T_,false,*temp,false,temp2,GetOStream(Runtime0),true,true);
-        else
-          temp2 = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Dk_1_,true,*temp,false,temp2,GetOStream(Runtime0),true,true);
-
-        // and back again
-        rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(Importer22_->getTargetMap(), Dimporter);
-
-        ParameterList XpetraList;
-        XpetraList.set("Restrict Communicator",true);
-        XpetraList.set("Timer Label","MueLu::RebalanceA22");
-        RCP<const Map> targetMap = Importer22_->getTargetMap();
-        A22_ = MatrixFactory::Build(temp2, *Importer22_, *Importer22_, targetMap, targetMap, rcp(&XpetraList,false));
+      if (!P22_.is_null()) {
+        // Todo
       }
-    }
-
-    if (not A22_.is_null() and not disable_addon_22_ and spaceNumber_ > 1) {
-      const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-      RCP<Matrix> addon22 = buildAddon(spaceNumber_-1);
-
-      // add matrices together
-      RCP<Matrix> newA22;
-      Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::TwoMatrixAdd(*A22_,false,one,*addon22,false,one,newA22,GetOStream(Runtime0));
-      newA22->fillComplete();
-      A22_ = newA22;
-    }
-
-    if (!A22_.is_null()) {
-      dump(A22_, "A22.m");
-      A22_->setObjectLabel(solverName_+" (2,2)");
-      // Set block size
-      if (spaceNumber_-1 == 0)
-        A22_->SetFixedBlockSize(1);
-      else
-        A22_->SetFixedBlockSize(dim_);
-    }
-  }
-
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setFineLevelSmoother11() {
-
-    Level level;
-    RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-    level.SetFactoryManager(factoryHandler);
-    level.SetLevelID(0);
-    level.setObjectLabel(solverName_+" (1,1)");
-    level.Set("A",SM_Matrix_);
-    level.setlib(SM_Matrix_->getDomainMap()->lib());
-    // For Hiptmair
-    level.Set("NodeMatrix", A22_);
-    level.Set("D0", Dk_1_);
-
-    if ((parameterList_.get<std::string>("smoother: pre type") != "NONE") && (parameterList_.get<std::string>("smoother: post type") != "NONE")) {
-      std::string preSmootherType = parameterList_.get<std::string>("smoother: pre type");
-      std::string postSmootherType = parameterList_.get<std::string>("smoother: post type");
-
-      ParameterList preSmootherList, postSmootherList;
-      if (parameterList_.isSublist("smoother: pre params"))
-        preSmootherList = parameterList_.sublist("smoother: pre params");
-      if (parameterList_.isSublist("smoother: post params"))
-        postSmootherList = parameterList_.sublist("smoother: post params");
-
-      RCP<SmootherPrototype> preSmootherPrototype = rcp(new TrilinosSmoother(preSmootherType, preSmootherList));
-      RCP<SmootherPrototype> postSmootherPrototype = rcp(new TrilinosSmoother(postSmootherType, postSmootherList));
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(preSmootherPrototype, postSmootherPrototype));
-
-      level.Request("PreSmoother",smootherFact.get());
-      level.Request("PostSmoother",smootherFact.get());
+    } else
+#endif  // HAVE_MPI
+    {
+      coarseLevel.Request("A", rapFact.get());
       if (enable_reuse_) {
-        ParameterList smootherFactoryParams;
-        smootherFactoryParams.set("keep smoother data", true);
-        smootherFact->SetParameterList(smootherFactoryParams);
-        level.Request("PreSmoother data", smootherFact.get());
-        level.Request("PostSmoother data", smootherFact.get());
-        if (!PreSmootherData11_.is_null())
-          level.Set("PreSmoother data", PreSmootherData11_, smootherFact.get());
-        if (!PostSmootherData11_.is_null())
-          level.Set("PostSmoother data", PostSmootherData11_, smootherFact.get());
+        coarseLevel.Request("AP reuse data", rapFact.get());
+        coarseLevel.Request("RAP reuse data", rapFact.get());
       }
-      smootherFact->Build(level);
-      PreSmoother11_ = level.Get<RCP<SmootherBase> >("PreSmoother",smootherFact.get());
-      PostSmoother11_ = level.Get<RCP<SmootherBase> >("PostSmoother",smootherFact.get());
+
+      A22_ = coarseLevel.Get<RCP<Matrix> >("A", rapFact.get());
+
       if (enable_reuse_) {
-        PreSmootherData11_ = level.Get<RCP<SmootherPrototype> >("PreSmoother data",smootherFact.get());
-        PostSmootherData11_ = level.Get<RCP<SmootherPrototype> >("PostSmoother data",smootherFact.get());
-      }
+        if (coarseLevel.IsAvailable("AP reuse data", rapFact.get()))
+          A22_AP_reuse_data_ = coarseLevel.Get<RCP<ParameterList> >("AP reuse data", rapFact.get());
+        if (coarseLevel.IsAvailable("RAP reuse data", rapFact.get()))
+          A22_RAP_reuse_data_ = coarseLevel.Get<RCP<ParameterList> >("RAP reuse data", rapFact.get());
+      }
+    }
+  } else {
+    RCP<Teuchos::TimeMonitor> tm = getTimer("Build A22");
+    if (Importer22_.is_null()) {
+      RCP<Matrix> temp;
+      temp = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*SM_Matrix_, false, *Dk_1_, false, temp, GetOStream(Runtime0), true, true);
+      if (!implicitTranspose_)
+        A22_ = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_T_, false, *temp, false, A22_, GetOStream(Runtime0), true, true);
+      else
+        A22_ = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_, true, *temp, false, A22_, GetOStream(Runtime0), true, true);
     } else {
-      std::string smootherType = parameterList_.get<std::string>("smoother: type");
+      // we replaced domain map and importer on D, reverse that
+      RCP<const Import> Dimporter = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->getCrsGraph()->getImporter();
+      rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(DorigDomainMap_, DorigImporter_);
+
+      RCP<Matrix> temp, temp2;
+      temp = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*SM_Matrix_, false, *Dk_1_, false, temp, GetOStream(Runtime0), true, true);
+      if (!implicitTranspose_)
+        temp2 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_T_, false, *temp, false, temp2, GetOStream(Runtime0), true, true);
+      else
+        temp2 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Dk_1_, true, *temp, false, temp2, GetOStream(Runtime0), true, true);
 
-      ParameterList smootherList;
-      if (parameterList_.isSublist("smoother: params"))
-        smootherList = parameterList_.sublist("smoother: params");
+      // and back again
+      rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_)->getCrsMatrix()->replaceDomainMapAndImporter(Importer22_->getTargetMap(), Dimporter);
 
-      RCP<SmootherPrototype> smootherPrototype = rcp(new TrilinosSmoother(smootherType, smootherList));
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
-      level.Request("PreSmoother",smootherFact.get());
-      if (enable_reuse_) {
-        ParameterList smootherFactoryParams;
-        smootherFactoryParams.set("keep smoother data", true);
-        smootherFact->SetParameterList(smootherFactoryParams);
-        level.Request("PreSmoother data", smootherFact.get());
-        if (!PreSmootherData11_.is_null())
-          level.Set("PreSmoother data", PreSmootherData11_, smootherFact.get());
-      }
-      smootherFact->Build(level);
-      PreSmoother11_ = level.Get<RCP<SmootherBase> >("PreSmoother",smootherFact.get());
-      PostSmoother11_ = PreSmoother11_;
-      if (enable_reuse_)
-        PreSmootherData11_ = level.Get<RCP<SmootherPrototype> >("PreSmoother data",smootherFact.get());
+      ParameterList XpetraList;
+      XpetraList.set("Restrict Communicator", true);
+      XpetraList.set("Timer Label", "MueLu::RebalanceA22");
+      RCP<const Map> targetMap = Importer22_->getTargetMap();
+      A22_                     = MatrixFactory::Build(temp2, *Importer22_, *Importer22_, targetMap, targetMap, rcp(&XpetraList, false));
     }
-
   }
 
+  if (not A22_.is_null() and not disable_addon_22_ and spaceNumber_ > 1) {
+    const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::allocateMemory(int numVectors) const {
+    RCP<Matrix> addon22 = buildAddon(spaceNumber_ - 1);
 
-    RCP<Teuchos::TimeMonitor> tmAlloc = getTimer("Allocate MVs");
+    // add matrices together
+    RCP<Matrix> newA22;
+    Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*A22_, false, one, *addon22, false, one, newA22, GetOStream(Runtime0));
+    newA22->fillComplete();
+    A22_ = newA22;
+  }
 
-    // 11 block
-    if (!R11_.is_null())
-      P11res_    = MultiVectorFactory::Build(R11_->getRangeMap(), numVectors);
+  if (!A22_.is_null()) {
+    dump(A22_, "A22.m");
+    A22_->setObjectLabel(solverName_ + " (2,2)");
+    // Set block size
+    if (spaceNumber_ - 1 == 0)
+      A22_->SetFixedBlockSize(1);
     else
-      P11res_    = MultiVectorFactory::Build(P11_->getDomainMap(), numVectors);
-    P11res_->setObjectLabel("P11res");
-
-    if (Dk_1_T_R11_colMapsMatch_) {
-      DTR11Tmp_ = MultiVectorFactory::Build(R11_->getColMap(), numVectors);
-      DTR11Tmp_->setObjectLabel("DTR11Tmp");
-    }
-    if (!ImporterCoarse11_.is_null()) {
-      P11resTmp_ = MultiVectorFactory::Build(ImporterCoarse11_->getTargetMap(), numVectors);
-      P11resTmp_->setObjectLabel("P11resTmp");
-      P11x_      = MultiVectorFactory::Build(ImporterCoarse11_->getTargetMap(), numVectors);
-    } else
-      P11x_      = MultiVectorFactory::Build(P11_->getDomainMap(), numVectors);
-    P11x_->setObjectLabel("P11x");
-
-    // 22 block
-    if (!Dk_1_T_.is_null())
-      Dres_     = MultiVectorFactory::Build(Dk_1_T_->getRangeMap(), numVectors);
+      A22_->SetFixedBlockSize(dim_);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setFineLevelSmoother11() {
+  Level level;
+  RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+  level.SetFactoryManager(factoryHandler);
+  level.SetLevelID(0);
+  level.setObjectLabel(solverName_ + " (1,1)");
+  level.Set("A", SM_Matrix_);
+  level.setlib(SM_Matrix_->getDomainMap()->lib());
+  // For Hiptmair
+  level.Set("NodeMatrix", A22_);
+  level.Set("D0", Dk_1_);
+
+  if ((parameterList_.get<std::string>("smoother: pre type") != "NONE") && (parameterList_.get<std::string>("smoother: post type") != "NONE")) {
+    std::string preSmootherType  = parameterList_.get<std::string>("smoother: pre type");
+    std::string postSmootherType = parameterList_.get<std::string>("smoother: post type");
+
+    ParameterList preSmootherList, postSmootherList;
+    if (parameterList_.isSublist("smoother: pre params"))
+      preSmootherList = parameterList_.sublist("smoother: pre params");
+    if (parameterList_.isSublist("smoother: post params"))
+      postSmootherList = parameterList_.sublist("smoother: post params");
+
+    RCP<SmootherPrototype> preSmootherPrototype  = rcp(new TrilinosSmoother(preSmootherType, preSmootherList));
+    RCP<SmootherPrototype> postSmootherPrototype = rcp(new TrilinosSmoother(postSmootherType, postSmootherList));
+    RCP<SmootherFactory> smootherFact            = rcp(new SmootherFactory(preSmootherPrototype, postSmootherPrototype));
+
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+    if (enable_reuse_) {
+      ParameterList smootherFactoryParams;
+      smootherFactoryParams.set("keep smoother data", true);
+      smootherFact->SetParameterList(smootherFactoryParams);
+      level.Request("PreSmoother data", smootherFact.get());
+      level.Request("PostSmoother data", smootherFact.get());
+      if (!PreSmootherData11_.is_null())
+        level.Set("PreSmoother data", PreSmootherData11_, smootherFact.get());
+      if (!PostSmootherData11_.is_null())
+        level.Set("PostSmoother data", PostSmootherData11_, smootherFact.get());
+    }
+    smootherFact->Build(level);
+    PreSmoother11_  = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    PostSmoother11_ = level.Get<RCP<SmootherBase> >("PostSmoother", smootherFact.get());
+    if (enable_reuse_) {
+      PreSmootherData11_  = level.Get<RCP<SmootherPrototype> >("PreSmoother data", smootherFact.get());
+      PostSmootherData11_ = level.Get<RCP<SmootherPrototype> >("PostSmoother data", smootherFact.get());
+    }
+  } else {
+    std::string smootherType = parameterList_.get<std::string>("smoother: type");
+
+    ParameterList smootherList;
+    if (parameterList_.isSublist("smoother: params"))
+      smootherList = parameterList_.sublist("smoother: params");
+
+    RCP<SmootherPrototype> smootherPrototype = rcp(new TrilinosSmoother(smootherType, smootherList));
+    RCP<SmootherFactory> smootherFact        = rcp(new SmootherFactory(smootherPrototype));
+    level.Request("PreSmoother", smootherFact.get());
+    if (enable_reuse_) {
+      ParameterList smootherFactoryParams;
+      smootherFactoryParams.set("keep smoother data", true);
+      smootherFact->SetParameterList(smootherFactoryParams);
+      level.Request("PreSmoother data", smootherFact.get());
+      if (!PreSmootherData11_.is_null())
+        level.Set("PreSmoother data", PreSmootherData11_, smootherFact.get());
+    }
+    smootherFact->Build(level);
+    PreSmoother11_  = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    PostSmoother11_ = PreSmoother11_;
+    if (enable_reuse_)
+      PreSmootherData11_ = level.Get<RCP<SmootherPrototype> >("PreSmoother data", smootherFact.get());
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::allocateMemory(int numVectors) const {
+  RCP<Teuchos::TimeMonitor> tmAlloc = getTimer("Allocate MVs");
+
+  // 11 block
+  if (!R11_.is_null())
+    P11res_ = MultiVectorFactory::Build(R11_->getRangeMap(), numVectors);
+  else
+    P11res_ = MultiVectorFactory::Build(P11_->getDomainMap(), numVectors);
+  P11res_->setObjectLabel("P11res");
+
+  if (Dk_1_T_R11_colMapsMatch_) {
+    DTR11Tmp_ = MultiVectorFactory::Build(R11_->getColMap(), numVectors);
+    DTR11Tmp_->setObjectLabel("DTR11Tmp");
+  }
+  if (!ImporterCoarse11_.is_null()) {
+    P11resTmp_ = MultiVectorFactory::Build(ImporterCoarse11_->getTargetMap(), numVectors);
+    P11resTmp_->setObjectLabel("P11resTmp");
+    P11x_ = MultiVectorFactory::Build(ImporterCoarse11_->getTargetMap(), numVectors);
+  } else
+    P11x_ = MultiVectorFactory::Build(P11_->getDomainMap(), numVectors);
+  P11x_->setObjectLabel("P11x");
+
+  // 22 block
+  if (!Dk_1_T_.is_null())
+    Dres_ = MultiVectorFactory::Build(Dk_1_T_->getRangeMap(), numVectors);
+  else
+    Dres_ = MultiVectorFactory::Build(Dk_1_->getDomainMap(), numVectors);
+  Dres_->setObjectLabel("Dres");
+
+  if (!Importer22_.is_null()) {
+    DresTmp_ = MultiVectorFactory::Build(Importer22_->getTargetMap(), numVectors);
+    DresTmp_->setObjectLabel("DresTmp");
+    Dx_ = MultiVectorFactory::Build(Importer22_->getTargetMap(), numVectors);
+  } else if (!onlyBoundary22_)
+    Dx_ = MultiVectorFactory::Build(A22_->getDomainMap(), numVectors);
+  if (!Dx_.is_null())
+    Dx_->setObjectLabel("Dx");
+
+  if (!coarseA11_.is_null()) {
+    if (!ImporterCoarse11_.is_null() && !implicitTranspose_)
+      P11resSubComm_ = MultiVectorFactory::Build(P11resTmp_, Teuchos::View);
     else
-      Dres_     = MultiVectorFactory::Build(Dk_1_->getDomainMap(), numVectors);
-    Dres_->setObjectLabel("Dres");
-
-    if (!Importer22_.is_null()) {
-      DresTmp_ = MultiVectorFactory::Build(Importer22_->getTargetMap(), numVectors);
-      DresTmp_->setObjectLabel("DresTmp");
-      Dx_      = MultiVectorFactory::Build(Importer22_->getTargetMap(), numVectors);
-    } else if (!onlyBoundary22_)
-      Dx_      = MultiVectorFactory::Build(A22_->getDomainMap(), numVectors);
-    if (!Dx_.is_null())
-      Dx_->setObjectLabel("Dx");
-
-    if (!coarseA11_.is_null()) {
-      if (!ImporterCoarse11_.is_null() && !implicitTranspose_)
-        P11resSubComm_ = MultiVectorFactory::Build(P11resTmp_, Teuchos::View);
-      else
-        P11resSubComm_ = MultiVectorFactory::Build(P11res_, Teuchos::View);
-      P11resSubComm_->replaceMap(coarseA11_->getRangeMap());
-      P11resSubComm_->setObjectLabel("P11resSubComm");
-
-      P11xSubComm_ = MultiVectorFactory::Build(P11x_, Teuchos::View);
-      P11xSubComm_->replaceMap(coarseA11_->getDomainMap());
-      P11xSubComm_->setObjectLabel("P11xSubComm");
-    }
+      P11resSubComm_ = MultiVectorFactory::Build(P11res_, Teuchos::View);
+    P11resSubComm_->replaceMap(coarseA11_->getRangeMap());
+    P11resSubComm_->setObjectLabel("P11resSubComm");
 
-    if (!A22_.is_null()) {
-      if (!Importer22_.is_null() && !implicitTranspose_)
-        DresSubComm_ = MultiVectorFactory::Build(DresTmp_, Teuchos::View);
-      else
-        DresSubComm_ = MultiVectorFactory::Build(Dres_, Teuchos::View);
-      DresSubComm_->replaceMap(A22_->getRangeMap());
-      DresSubComm_->setObjectLabel("DresSubComm");
-
-      DxSubComm_ = MultiVectorFactory::Build(Dx_, Teuchos::View);
-      DxSubComm_->replaceMap(A22_->getDomainMap());
-      DxSubComm_->setObjectLabel("DxSubComm");
-    }
-
-    residual_  = MultiVectorFactory::Build(SM_Matrix_->getDomainMap(), numVectors);
-    residual_->setObjectLabel("residual");
+    P11xSubComm_ = MultiVectorFactory::Build(P11x_, Teuchos::View);
+    P11xSubComm_->replaceMap(coarseA11_->getDomainMap());
+    P11xSubComm_->setObjectLabel("P11xSubComm");
   }
 
+  if (!A22_.is_null()) {
+    if (!Importer22_.is_null() && !implicitTranspose_)
+      DresSubComm_ = MultiVectorFactory::Build(DresTmp_, Teuchos::View);
+    else
+      DresSubComm_ = MultiVectorFactory::Build(Dres_, Teuchos::View);
+    DresSubComm_->replaceMap(A22_->getRangeMap());
+    DresSubComm_->setObjectLabel("DresSubComm");
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const RCP<Matrix>& A, std::string name) const {
-    if (dump_matrices_ && !A.is_null()) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<SC, LO, GO, NO>::Write(name, *A);
-    }
+    DxSubComm_ = MultiVectorFactory::Build(Dx_, Teuchos::View);
+    DxSubComm_->replaceMap(A22_->getDomainMap());
+    DxSubComm_->setObjectLabel("DxSubComm");
   }
 
+  residual_ = MultiVectorFactory::Build(SM_Matrix_->getDomainMap(), numVectors);
+  residual_->setObjectLabel("residual");
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const RCP<MultiVector>& X, std::string name) const {
-    if (dump_matrices_ && !X.is_null()) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<SC, LO, GO, NO>::Write(name, *X);
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const RCP<Matrix> &A, std::string name) const {
+  if (dump_matrices_ && !A.is_null()) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<SC, LO, GO, NO>::Write(name, *A);
   }
+}
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dumpCoords(const RCP<RealValuedMultiVector>& X, std::string name) const {
-    if (dump_matrices_ && !X.is_null()) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      Xpetra::IO<coordinateType, LO, GO, NO>::Write(name, *X);
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const RCP<MultiVector> &X, std::string name) const {
+  if (dump_matrices_ && !X.is_null()) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<SC, LO, GO, NO>::Write(name, *X);
   }
+}
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const Teuchos::ArrayRCP<bool>& v, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      std::ofstream out(name);
-      for (size_t i = 0; i < Teuchos::as<size_t>(v.size()); i++)
-        out << v[i] << "\n";
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dumpCoords(const RCP<RealValuedMultiVector> &X, std::string name) const {
+  if (dump_matrices_ && !X.is_null()) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    Xpetra::IO<coordinateType, LO, GO, NO>::Write(name, *X);
   }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dump(const Kokkos::View<bool*, typename Node::device_type>& v, std::string name) const {
-    if (dump_matrices_) {
-      GetOStream(Runtime0) << "Dumping to " << name << std::endl;
-      std::ofstream out(name);
-      auto vH = Kokkos::create_mirror_view (v);
-      Kokkos::deep_copy(vH , v);
-      out << "%%MatrixMarket matrix array real general\n" << vH.extent(0) << " 1\n";
-      for (size_t i = 0; i < vH.size(); i++)
-        out << vH[i] << "\n";
-    }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const Teuchos::ArrayRCP<bool> &v, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    std::ofstream out(name);
+    for (size_t i = 0; i < Teuchos::as<size_t>(v.size()); i++)
+      out << v[i] << "\n";
   }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Teuchos::TimeMonitor> RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm) const {
-    if (IsPrint(Timings)) {
-      if (!syncTimers_)
-        return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu "+solverName_+": "+name)));
-      else {
-        if (comm.is_null())
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu "+solverName_+": "+name), SM_Matrix_->getRowMap()->getComm().ptr()));
-        else
-          return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu "+solverName_+": "+name), comm.ptr()));
-      }
-    } else
-      return Teuchos::null;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dump(const Kokkos::View<bool *, typename Node::device_type> &v, std::string name) const {
+  if (dump_matrices_) {
+    GetOStream(Runtime0) << "Dumping to " << name << std::endl;
+    std::ofstream out(name);
+    auto vH = Kokkos::create_mirror_view(v);
+    Kokkos::deep_copy(vH, v);
+    out << "%%MatrixMarket matrix array real general\n"
+        << vH.extent(0) << " 1\n";
+    for (size_t i = 0; i < vH.size(); i++)
+      out << vH[i] << "\n";
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Teuchos::TimeMonitor> RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getTimer(std::string name, RCP<const Teuchos::Comm<int> > comm) const {
+  if (IsPrint(Timings)) {
+    if (!syncTimers_)
+      return Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu " + solverName_ + ": " + name)));
+    else {
+      if (comm.is_null())
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu " + solverName_ + ": " + name), SM_Matrix_->getRowMap()->getComm().ptr()));
+      else
+        return Teuchos::rcp(new Teuchos::SyncTimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MueLu " + solverName_ + ": " + name), comm.ptr()));
+    }
+  } else
+    return Teuchos::null;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    buildNullspace(const int spaceNumber, const Kokkos::View<bool *, typename Node::device_type> &bcs, const bool applyBCs) {
+  std::string spaceLabel;
+  if (spaceNumber == 0)
+    spaceLabel = "nodal";
+  else if (spaceNumber == 1)
+    spaceLabel = "edge";
+  else if (spaceNumber == 2)
+    spaceLabel = "face";
+  else
+    TEUCHOS_ASSERT(false);
+
+  RCP<Teuchos::TimeMonitor> tm;
+  if (spaceNumber > 0) {
+    tm = getTimer("nullspace " + spaceLabel);
+    GetOStream(Runtime0) << solverName_ + "::compute(): building " + spaceLabel + " nullspace" << std::endl;
+  }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  buildNullspace(const int spaceNumber, const Kokkos::View<bool*, typename Node::device_type> &bcs, const bool applyBCs) {
-
-    std::string spaceLabel;
-    if (spaceNumber == 0)
-      spaceLabel = "nodal";
-    else if (spaceNumber == 1)
-      spaceLabel = "edge";
-    else if (spaceNumber == 2)
-      spaceLabel = "face";
-    else
-      TEUCHOS_ASSERT(false);
-
-    RCP<Teuchos::TimeMonitor> tm;
-    if (spaceNumber > 0) {
-      tm = getTimer("nullspace "+spaceLabel);
-      GetOStream(Runtime0) << solverName_+"::compute(): building "+spaceLabel+" nullspace" << std::endl;
-    }
-
-    if (spaceNumber == 0) {
-
-      return Teuchos::null;
-
-    } else if (spaceNumber == 1) {
-
-      RCP<MultiVector> CoordsSC;
-      CoordsSC = Utilities::RealValuedToScalarMultiVector(NodalCoords_);
-      RCP<MultiVector> Nullspace = MultiVectorFactory::Build(D0_->getRowMap(),NodalCoords_->getNumVectors());
-      D0_->apply(*CoordsSC,*Nullspace);
-
-      bool normalize = parameterList_.get<bool>("refmaxwell: normalize nullspace", MasterList::getDefault<bool>("refmaxwell: normalize nullspace"));
-
-      coordinateType minLen, maxLen, meanLen;
-      if (IsPrint(Statistics2) || normalize){
-        // compute edge lengths
-        ArrayRCP<ArrayRCP<const Scalar> > localNullspace(dim_);
+  if (spaceNumber == 0) {
+    return Teuchos::null;
+
+  } else if (spaceNumber == 1) {
+    RCP<MultiVector> CoordsSC;
+    CoordsSC                   = Utilities::RealValuedToScalarMultiVector(NodalCoords_);
+    RCP<MultiVector> Nullspace = MultiVectorFactory::Build(D0_->getRowMap(), NodalCoords_->getNumVectors());
+    D0_->apply(*CoordsSC, *Nullspace);
+
+    bool normalize = parameterList_.get<bool>("refmaxwell: normalize nullspace", MasterList::getDefault<bool>("refmaxwell: normalize nullspace"));
+
+    coordinateType minLen, maxLen, meanLen;
+    if (IsPrint(Statistics2) || normalize) {
+      // compute edge lengths
+      ArrayRCP<ArrayRCP<const Scalar> > localNullspace(dim_);
+      for (size_t i = 0; i < dim_; i++)
+        localNullspace[i] = Nullspace->getData(i);
+      coordinateType localMinLen  = Teuchos::ScalarTraits<coordinateType>::rmax();
+      coordinateType localMeanLen = Teuchos::ScalarTraits<coordinateType>::zero();
+      coordinateType localMaxLen  = Teuchos::ScalarTraits<coordinateType>::zero();
+      for (size_t j = 0; j < Nullspace->getMap()->getLocalNumElements(); j++) {
+        Scalar lenSC = Teuchos::ScalarTraits<Scalar>::zero();
         for (size_t i = 0; i < dim_; i++)
-          localNullspace[i] = Nullspace->getData(i);
-        coordinateType localMinLen = Teuchos::ScalarTraits<coordinateType>::rmax();
-        coordinateType localMeanLen = Teuchos::ScalarTraits<coordinateType>::zero();
-        coordinateType localMaxLen = Teuchos::ScalarTraits<coordinateType>::zero();
-        for (size_t j=0; j < Nullspace->getMap()->getLocalNumElements(); j++) {
-          Scalar lenSC = Teuchos::ScalarTraits<Scalar>::zero();
-          for (size_t i=0; i < dim_; i++)
-            lenSC += localNullspace[i][j]*localNullspace[i][j];
-          coordinateType len = Teuchos::as<coordinateType>(Teuchos::ScalarTraits<Scalar>::real(Teuchos::ScalarTraits<Scalar>::squareroot(lenSC)));
-          localMinLen = std::min(localMinLen, len);
-          localMaxLen = std::max(localMaxLen, len);
-          localMeanLen += len;
-        }
-
-        RCP<const Teuchos::Comm<int> > comm = Nullspace->getMap()->getComm();
-        MueLu_minAll(comm, localMinLen,  minLen);
-        MueLu_sumAll(comm, localMeanLen, meanLen);
-        MueLu_maxAll(comm, localMaxLen,  maxLen);
-        meanLen /= Nullspace->getMap()->getGlobalNumElements();
+          lenSC += localNullspace[i][j] * localNullspace[i][j];
+        coordinateType len = Teuchos::as<coordinateType>(Teuchos::ScalarTraits<Scalar>::real(Teuchos::ScalarTraits<Scalar>::squareroot(lenSC)));
+        localMinLen        = std::min(localMinLen, len);
+        localMaxLen        = std::max(localMaxLen, len);
+        localMeanLen += len;
       }
 
-      if (IsPrint(Statistics2)) {
-        GetOStream(Statistics2) << "Edge length (min/mean/max): " << minLen << " / " << meanLen << " / " << maxLen << std::endl;
-      }
+      RCP<const Teuchos::Comm<int> > comm = Nullspace->getMap()->getComm();
+      MueLu_minAll(comm, localMinLen, minLen);
+      MueLu_sumAll(comm, localMeanLen, meanLen);
+      MueLu_maxAll(comm, localMaxLen, maxLen);
+      meanLen /= Nullspace->getMap()->getGlobalNumElements();
+    }
 
-      if (normalize) {
-        // normalize the nullspace
-        GetOStream(Runtime0) << solverName_+"::compute(): normalizing nullspace" << std::endl;
+    if (IsPrint(Statistics2)) {
+      GetOStream(Statistics2) << "Edge length (min/mean/max): " << minLen << " / " << meanLen << " / " << maxLen << std::endl;
+    }
 
-        const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+    if (normalize) {
+      // normalize the nullspace
+      GetOStream(Runtime0) << solverName_ + "::compute(): normalizing nullspace" << std::endl;
 
-        Array<Scalar> normsSC(NodalCoords_->getNumVectors(), one / Teuchos::as<Scalar>(meanLen));
-        Nullspace->scale(normsSC());
-      }
+      const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
 
-      if (applyBCs) {
-        // Nuke the BC edges in nullspace
-        Utilities::ZeroDirichletRows(Nullspace,bcs);
-      }
-      dump(Nullspace, "nullspaceEdge.m");
+      Array<Scalar> normsSC(NodalCoords_->getNumVectors(), one / Teuchos::as<Scalar>(meanLen));
+      Nullspace->scale(normsSC());
+    }
 
-      return Nullspace;
+    if (applyBCs) {
+      // Nuke the BC edges in nullspace
+      Utilities::ZeroDirichletRows(Nullspace, bcs);
+    }
+    dump(Nullspace, "nullspaceEdge.m");
 
-    } else if (spaceNumber == 2) {
+    return Nullspace;
 
-      using ATS         = Kokkos::ArithTraits<Scalar>;
-      using impl_Scalar = typename ATS::val_type;
-      using impl_ATS    = Kokkos::ArithTraits<impl_Scalar>;
-      using range_type  = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+  } else if (spaceNumber == 2) {
+    using ATS         = Kokkos::ArithTraits<Scalar>;
+    using impl_Scalar = typename ATS::val_type;
+    using impl_ATS    = Kokkos::ArithTraits<impl_Scalar>;
+    using range_type  = Kokkos::RangePolicy<LO, typename NO::execution_space>;
 
-      RCP<Matrix> facesToNodes;
-      {
-        RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(D0_);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(edgesToNodes, 1e-3);
+    RCP<Matrix> facesToNodes;
+    {
+      RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(D0_);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(edgesToNodes, 1e-3);
 
-        // dump(edgesToNodes, "edgesToNodes.m");
+      // dump(edgesToNodes, "edgesToNodes.m");
 
-        RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(Dk_1_);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToEdges, 1e-3);
-        facesToEdges = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToEdges, 1e-3, false);
+      RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(Dk_1_);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(facesToEdges, 1e-3);
+      facesToEdges = Maxwell_Utils<SC, LO, GO, NO>::removeExplicitZeros(facesToEdges, 1e-3, false);
 
-        // dump(facesToEdges, "facesToEdges.m");
+      // dump(facesToEdges, "facesToEdges.m");
 
-        facesToNodes = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToEdges,false,*edgesToNodes,false,facesToNodes,GetOStream(Runtime0),true,true);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToNodes, 1e-3);
-        facesToNodes = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToNodes, 1e-3, false);
-      }
+      facesToNodes = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*facesToEdges, false, *edgesToNodes, false, facesToNodes, GetOStream(Runtime0), true, true);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(facesToNodes, 1e-3);
+      facesToNodes = Maxwell_Utils<SC, LO, GO, NO>::removeExplicitZeros(facesToNodes, 1e-3, false);
+    }
 
-      // dump(facesToNodes, "facesToNodes.m");
+    // dump(facesToNodes, "facesToNodes.m");
 
-      RCP<RealValuedMultiVector> ghostedNodalCoordinates;
-      auto importer = facesToNodes->getCrsGraph()->getImporter();
-      if (!importer.is_null()) {
-        ghostedNodalCoordinates = Xpetra::MultiVectorFactory<coordinateType,LocalOrdinal,GlobalOrdinal,Node>::Build(importer->getTargetMap(), dim_);
-        ghostedNodalCoordinates->doImport(*NodalCoords_, *importer, Xpetra::INSERT);
-      } else
-        ghostedNodalCoordinates = NodalCoords_;
+    RCP<RealValuedMultiVector> ghostedNodalCoordinates;
+    auto importer = facesToNodes->getCrsGraph()->getImporter();
+    if (!importer.is_null()) {
+      ghostedNodalCoordinates = Xpetra::MultiVectorFactory<coordinateType, LocalOrdinal, GlobalOrdinal, Node>::Build(importer->getTargetMap(), dim_);
+      ghostedNodalCoordinates->doImport(*NodalCoords_, *importer, Xpetra::INSERT);
+    } else
+      ghostedNodalCoordinates = NodalCoords_;
 
-      RCP<MultiVector> Nullspace = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(facesToNodes->getRangeMap(), dim_);
-      {
-        auto facesToNodesLocal = facesToNodes->getLocalMatrixDevice();
-        auto localNodalCoordinates = ghostedNodalCoordinates->getDeviceLocalView(Xpetra::Access::ReadOnly);
-        auto localFaceNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-        // enter values
-        Kokkos::parallel_for(solverName_+"::buildFaceProjection_nullspace",
-                             range_type(0,Nullspace->getMap()->getLocalNumElements()),
-                             KOKKOS_LAMBDA(const size_t f) {
-
-                               size_t n0 = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f));
-                               size_t n1 = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f)+1);
-                               size_t n2 = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f)+2);
-                               impl_Scalar elementNullspace00 = localNodalCoordinates(n1, 0)-localNodalCoordinates(n0, 0);
-                               impl_Scalar elementNullspace10 = localNodalCoordinates(n2, 0)-localNodalCoordinates(n0, 0);
-                               impl_Scalar elementNullspace01 = localNodalCoordinates(n1, 1)-localNodalCoordinates(n0, 1);
-                               impl_Scalar elementNullspace11 = localNodalCoordinates(n2, 1)-localNodalCoordinates(n0, 1);
-                               impl_Scalar elementNullspace02 = localNodalCoordinates(n1, 2)-localNodalCoordinates(n0, 2);
-                               impl_Scalar elementNullspace12 = localNodalCoordinates(n2, 2)-localNodalCoordinates(n0, 2);
-
-                               localFaceNullspace(f, 0) = impl_ATS::magnitude(elementNullspace01*elementNullspace12-elementNullspace02*elementNullspace11) / 6.0;
-                               localFaceNullspace(f, 1) = impl_ATS::magnitude(elementNullspace02*elementNullspace10-elementNullspace00*elementNullspace12) / 6.0;
-                               localFaceNullspace(f, 2) = impl_ATS::magnitude(elementNullspace00*elementNullspace11-elementNullspace01*elementNullspace10) / 6.0;
-                             });
-      }
+    RCP<MultiVector> Nullspace = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(facesToNodes->getRangeMap(), dim_);
+    {
+      auto facesToNodesLocal     = facesToNodes->getLocalMatrixDevice();
+      auto localNodalCoordinates = ghostedNodalCoordinates->getDeviceLocalView(Xpetra::Access::ReadOnly);
+      auto localFaceNullspace    = Nullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+      // enter values
+      Kokkos::parallel_for(
+          solverName_ + "::buildFaceProjection_nullspace",
+          range_type(0, Nullspace->getMap()->getLocalNumElements()),
+          KOKKOS_LAMBDA(const size_t f) {
+            size_t n0                      = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f));
+            size_t n1                      = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f) + 1);
+            size_t n2                      = facesToNodesLocal.graph.entries(facesToNodesLocal.graph.row_map(f) + 2);
+            impl_Scalar elementNullspace00 = localNodalCoordinates(n1, 0) - localNodalCoordinates(n0, 0);
+            impl_Scalar elementNullspace10 = localNodalCoordinates(n2, 0) - localNodalCoordinates(n0, 0);
+            impl_Scalar elementNullspace01 = localNodalCoordinates(n1, 1) - localNodalCoordinates(n0, 1);
+            impl_Scalar elementNullspace11 = localNodalCoordinates(n2, 1) - localNodalCoordinates(n0, 1);
+            impl_Scalar elementNullspace02 = localNodalCoordinates(n1, 2) - localNodalCoordinates(n0, 2);
+            impl_Scalar elementNullspace12 = localNodalCoordinates(n2, 2) - localNodalCoordinates(n0, 2);
+
+            localFaceNullspace(f, 0) = impl_ATS::magnitude(elementNullspace01 * elementNullspace12 - elementNullspace02 * elementNullspace11) / 6.0;
+            localFaceNullspace(f, 1) = impl_ATS::magnitude(elementNullspace02 * elementNullspace10 - elementNullspace00 * elementNullspace12) / 6.0;
+            localFaceNullspace(f, 2) = impl_ATS::magnitude(elementNullspace00 * elementNullspace11 - elementNullspace01 * elementNullspace10) / 6.0;
+          });
+    }
+
+    if (applyBCs) {
+      // Nuke the BC faces in nullspace
+      Utilities::ZeroDirichletRows(Nullspace, bcs);
+    }
+
+    dump(Nullspace, "nullspaceFace.m");
+
+    return Nullspace;
+
+  } else
+    TEUCHOS_ASSERT(false);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildProjection(const int spaceNumber, const RCP<MultiVector> &Nullspace) const {
+  using ATS         = Kokkos::ArithTraits<Scalar>;
+  using impl_Scalar = typename ATS::val_type;
+  using impl_ATS    = Kokkos::ArithTraits<impl_Scalar>;
+  using range_type  = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+
+  typedef typename Matrix::local_matrix_type KCRS;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
+  typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+  typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
+
+  const impl_Scalar impl_SC_ONE  = impl_ATS::one();
+  const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
+  const impl_Scalar impl_half    = impl_SC_ONE / (impl_SC_ONE + impl_SC_ONE);
+
+  std::string spaceLabel;
+  if (spaceNumber == 0)
+    spaceLabel = "nodal";
+  else if (spaceNumber == 1)
+    spaceLabel = "edge";
+  else if (spaceNumber == 2)
+    spaceLabel = "face";
+  else
+    TEUCHOS_ASSERT(false);
+
+  RCP<Teuchos::TimeMonitor> tm;
+  if (spaceNumber > 0) {
+    tm = getTimer("projection " + spaceLabel);
+    GetOStream(Runtime0) << solverName_ + "::compute(): building " + spaceLabel + " projection" << std::endl;
+  }
 
-      if (applyBCs) {
-        // Nuke the BC faces in nullspace
-        Utilities::ZeroDirichletRows(Nullspace,bcs);
-      }
+  RCP<Matrix> incidence;
+  if (spaceNumber == 0) {
+    // identity projection
+    return Teuchos::null;
+
+  } else if (spaceNumber == 1) {
+    // D0 is incidence from nodes to edges
+    incidence = D0_;
 
-      dump(Nullspace, "nullspaceFace.m");
+  } else if (spaceNumber == 2) {
+    // get incidence from nodes to faces by multiplying D0 and D1
 
-      return Nullspace;
+    TEUCHOS_ASSERT(spaceNumber_ == 2);
 
-    } else
-      TEUCHOS_ASSERT(false);
-  }
+    RCP<Matrix> facesToNodes;
+    {
+      RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(D0_);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(edgesToNodes, 1e-10);
 
+      dump(edgesToNodes, "edgesToNodes.m");
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::buildProjection(const int spaceNumber, const RCP<MultiVector> &Nullspace) const {
+      RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(Dk_1_);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(facesToEdges, 1e-10);
+      // facesToEdges = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToEdges, 1e-2, false);
 
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_Scalar = typename ATS::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_Scalar>;
-    using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-
-    typedef typename Matrix::local_matrix_type KCRS;
-    typedef typename KCRS::StaticCrsGraphType graph_t;
-    typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-    typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type scalar_view_t;
-
-    const impl_Scalar impl_SC_ONE  = impl_ATS::one();
-    const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
-    const impl_Scalar impl_half    = impl_SC_ONE / (impl_SC_ONE + impl_SC_ONE);
-
-    std::string spaceLabel;
-    if (spaceNumber == 0)
-      spaceLabel = "nodal";
-    else if (spaceNumber == 1)
-      spaceLabel = "edge";
-    else if (spaceNumber == 2)
-      spaceLabel = "face";
-    else
-      TEUCHOS_ASSERT(false);
+      dump(facesToEdges, "facesToEdges.m");
 
-    RCP<Teuchos::TimeMonitor> tm;
-    if (spaceNumber > 0) {
-      tm = getTimer("projection "+spaceLabel);
-      GetOStream(Runtime0) << solverName_+"::compute(): building "+spaceLabel+" projection" << std::endl;
+      facesToNodes = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*facesToEdges, false, *edgesToNodes, false, facesToNodes, GetOStream(Runtime0), true, true);
+      Maxwell_Utils<SC, LO, GO, NO>::thresholdedAbs(facesToNodes, 1e-10);
+      facesToNodes = Maxwell_Utils<SC, LO, GO, NO>::removeExplicitZeros(facesToNodes, 1e-2, false);
     }
 
-    RCP<Matrix> incidence;
-    if (spaceNumber == 0) {
+    dump(facesToNodes, "facesToNodes.m");
 
-      // identity projection
-      return Teuchos::null;
+    incidence = facesToNodes;
 
-    } else if (spaceNumber == 1) {
+  } else
+    TEUCHOS_ASSERT(false);
 
-      // D0 is incidence from nodes to edges
-      incidence = D0_;
+  size_t dim = dim_;
 
-    } else if (spaceNumber == 2) {
+  // Create maps
+  RCP<const Map> rowMap         = incidence->getRowMap();
+  RCP<const Map> blockColMap    = MapFactory::Build(incidence->getColMap(), dim);
+  RCP<const Map> blockDomainMap = MapFactory::Build(incidence->getDomainMap(), dim);
 
-      // get incidence from nodes to faces by multiplying D0 and D1
+  auto localIncidence    = incidence->getLocalMatrixDevice();
+  size_t numLocalRows    = rowMap->getLocalNumElements();
+  size_t numLocalColumns = dim * incidence->getColMap()->getLocalNumElements();
+  size_t nnzEstimate     = dim * localIncidence.graph.entries.size();
+  lno_view_t rowptr(Kokkos::ViewAllocateWithoutInitializing("projection_rowptr_" + spaceLabel), numLocalRows + 1);
+  lno_nnz_view_t colind(Kokkos::ViewAllocateWithoutInitializing("projection_colind_" + spaceLabel), nnzEstimate);
+  scalar_view_t vals("projection_vals_" + spaceLabel, nnzEstimate);
 
-      TEUCHOS_ASSERT(spaceNumber_ == 2);
+  // set rowpointer
+  Kokkos::parallel_for(
+      solverName_ + "::buildProjection_adjustRowptr_" + spaceLabel,
+      range_type(0, numLocalRows + 1),
+      KOKKOS_LAMBDA(const size_t i) {
+        rowptr(i) = dim * localIncidence.graph.row_map(i);
+      });
 
-      RCP<Matrix> facesToNodes;
-      {
-        RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(D0_);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(edgesToNodes, 1e-10);
+  auto localNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
 
-        dump(edgesToNodes, "edgesToNodes.m");
+  // set column indices and values
+  magnitudeType tol = 1e-5;
+  Kokkos::parallel_for(
+      solverName_ + "::buildProjection_enterValues_" + spaceLabel,
+      range_type(0, numLocalRows),
+      KOKKOS_LAMBDA(const size_t f) {
+        for (size_t jj = localIncidence.graph.row_map(f); jj < localIncidence.graph.row_map(f + 1); jj++) {
+          for (size_t k = 0; k < dim; k++) {
+            colind(dim * jj + k) = dim * localIncidence.graph.entries(jj) + k;
+            if (impl_ATS::magnitude(localIncidence.values(jj)) > tol)
+              vals(dim * jj + k) = impl_half * localNullspace(f, k);
+            else
+              vals(dim * jj + k) = impl_SC_ZERO;
+          }
+        }
+      });
 
-        RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(Dk_1_);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToEdges, 1e-10);
-        // facesToEdges = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToEdges, 1e-2, false);
+  // Create matrix
+  typename CrsMatrix::local_matrix_type lclProjection("local projection " + spaceLabel,
+                                                      numLocalRows, numLocalColumns, nnzEstimate,
+                                                      vals, rowptr, colind);
+  RCP<Matrix> projection = MatrixFactory::Build(lclProjection,
+                                                rowMap, blockColMap,
+                                                blockDomainMap, rowMap);
 
-        dump(facesToEdges, "facesToEdges.m");
+  return projection;
+}
 
-        facesToNodes = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToEdges,false,*edgesToNodes,false,facesToNodes,GetOStream(Runtime0),true,true);
-        Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToNodes, 1e-10);
-        facesToNodes = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToNodes, 1e-2, false);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildNodalProlongator(const Teuchos::RCP<Matrix> &A_nodal,
+                                                                                  Teuchos::RCP<Matrix> &P_nodal,
+                                                                                  Teuchos::RCP<MultiVector> &Nullspace_nodal,
+                                                                                  Teuchos::RCP<RealValuedMultiVector> &CoarseCoords_nodal) const {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("nodal prolongator");
+  GetOStream(Runtime0) << solverName_ + "::compute(): building nodal prolongator" << std::endl;
 
-      dump(facesToNodes, "facesToNodes.m");
+  // build prolongator: algorithm 1 in the reference paper
+  // First, build nodal unsmoothed prolongator using the matrix A_nodal
 
-      incidence = facesToNodes;
+  const SC SC_ONE = Teuchos::ScalarTraits<SC>::one();
 
+  {
+    Level fineLevel, coarseLevel;
+    fineLevel.SetFactoryManager(null);
+    coarseLevel.SetFactoryManager(null);
+    coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
+    fineLevel.SetLevelID(0);
+    coarseLevel.SetLevelID(1);
+    fineLevel.Set("A", A_nodal);
+    fineLevel.Set("Coordinates", NodalCoords_);
+    fineLevel.Set("DofsPerNode", 1);
+    coarseLevel.setlib(A_nodal->getDomainMap()->lib());
+    fineLevel.setlib(A_nodal->getDomainMap()->lib());
+    coarseLevel.setObjectLabel(A_nodal->getObjectLabel());
+    fineLevel.setObjectLabel(A_nodal->getObjectLabel());
+
+    LocalOrdinal NSdim         = 1;
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A_nodal->getRowMap(), NSdim);
+    nullSpace->putScalar(SC_ONE);
+    fineLevel.Set("Nullspace", nullSpace);
+
+    std::string algo = parameterList_.get<std::string>("multigrid algorithm");
+
+    RCP<Factory> amalgFact, dropFact, UncoupledAggFact, coarseMapFact, TentativePFact, Tfact, SaPFact;
+    amalgFact     = rcp(new AmalgamationFactory());
+    coarseMapFact = rcp(new CoarseMapFactory());
+    Tfact         = rcp(new CoordinatesTransferFactory());
+    if (useKokkos_) {
+      dropFact         = rcp(new CoalesceDropFactory_kokkos());
+      UncoupledAggFact = rcp(new UncoupledAggregationFactory_kokkos());
+      TentativePFact   = rcp(new TentativePFactory_kokkos());
+      if (algo == "sa")
+        SaPFact = rcp(new SaPFactory_kokkos());
+    } else {
+      dropFact         = rcp(new CoalesceDropFactory());
+      UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+      TentativePFact   = rcp(new TentativePFactory());
+      if (algo == "sa")
+        SaPFact = rcp(new SaPFactory());
+    }
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+    double dropTol           = parameterList_.get<double>("aggregation: drop tol");
+    std::string dropScheme   = parameterList_.get<std::string>("aggregation: drop scheme");
+    std::string distLaplAlgo = parameterList_.get<std::string>("aggregation: distance laplacian algo");
+    dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+    dropFact->SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(dropScheme));
+    if (!useKokkos_)
+      dropFact->SetParameter("aggregation: distance laplacian algo", Teuchos::ParameterEntry(distLaplAlgo));
+
+    UncoupledAggFact->SetFactory("Graph", dropFact);
+    int minAggSize = parameterList_.get<int>("aggregation: min agg size");
+    UncoupledAggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(minAggSize));
+    int maxAggSize = parameterList_.get<int>("aggregation: max agg size");
+    UncoupledAggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(maxAggSize));
+    bool matchMLbehavior1 = parameterList_.get<bool>("aggregation: match ML phase1");
+    UncoupledAggFact->SetParameter("aggregation: match ML phase1", Teuchos::ParameterEntry(matchMLbehavior1));
+    bool matchMLbehavior2a = parameterList_.get<bool>("aggregation: match ML phase2a");
+    UncoupledAggFact->SetParameter("aggregation: match ML phase2a", Teuchos::ParameterEntry(matchMLbehavior2a));
+    bool matchMLbehavior2b = parameterList_.get<bool>("aggregation: match ML phase2b");
+    UncoupledAggFact->SetParameter("aggregation: match ML phase2b", Teuchos::ParameterEntry(matchMLbehavior2b));
+
+    coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
+
+    TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
+    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+    Tfact->SetFactory("Aggregates", UncoupledAggFact);
+    Tfact->SetFactory("CoarseMap", coarseMapFact);
+
+    if (algo == "sa") {
+      SaPFact->SetFactory("P", TentativePFact);
+      coarseLevel.Request("P", SaPFact.get());
     } else
-      TEUCHOS_ASSERT(false);
+      coarseLevel.Request("P", TentativePFact.get());
+    coarseLevel.Request("Nullspace", TentativePFact.get());
+    coarseLevel.Request("Coordinates", Tfact.get());
+
+    RCP<AggregationExportFactory> aggExport;
+    bool exportVizData = parameterList_.get<bool>("aggregation: export visualization data");
+    if (exportVizData) {
+      aggExport = rcp(new AggregationExportFactory());
+      ParameterList aggExportParams;
+      aggExportParams.set("aggregation: output filename", "aggs.vtk");
+      aggExportParams.set("aggregation: output file: agg style", "Jacks");
+      aggExport->SetParameterList(aggExportParams);
+
+      aggExport->SetFactory("Aggregates", UncoupledAggFact);
+      aggExport->SetFactory("UnAmalgamationInfo", amalgFact);
+      fineLevel.Request("Aggregates", UncoupledAggFact.get());
+      fineLevel.Request("UnAmalgamationInfo", amalgFact.get());
+    }
+
+    if (algo == "sa")
+      coarseLevel.Get("P", P_nodal, SaPFact.get());
+    else
+      coarseLevel.Get("P", P_nodal, TentativePFact.get());
+    coarseLevel.Get("Nullspace", Nullspace_nodal, TentativePFact.get());
+    coarseLevel.Get("Coordinates", CoarseCoords_nodal, Tfact.get());
+
+    if (exportVizData)
+      aggExport->Build(fineLevel, coarseLevel);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildVectorNodalProlongator(const Teuchos::RCP<Matrix> &P_nodal) const {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("vectorial nodal prolongator");
+  GetOStream(Runtime0) << solverName_ + "::compute(): building vectorial nodal prolongator" << std::endl;
+
+  using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+
+  typedef typename Matrix::local_matrix_type KCRS;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
+  typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+  typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
+
+  size_t dim = dim_;
+
+  // Create the matrix object
+  RCP<Map> blockRowMap    = MapFactory::Build(P_nodal->getRowMap(), dim);
+  RCP<Map> blockColMap    = MapFactory::Build(P_nodal->getColMap(), dim);
+  RCP<Map> blockDomainMap = MapFactory::Build(P_nodal->getDomainMap(), dim);
+
+  // Get data out of P_nodal.
+  auto localP_nodal = P_nodal->getLocalMatrixDevice();
+
+  size_t numLocalRows    = blockRowMap->getLocalNumElements();
+  size_t numLocalColumns = blockColMap->getLocalNumElements();
+  size_t nnzEstimate     = dim * localP_nodal.graph.entries.size();
+  lno_view_t rowptr(Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_rowptr"), numLocalRows + 1);
+  lno_nnz_view_t colind(Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_colind"), nnzEstimate);
+  scalar_view_t vals(Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_vals"), nnzEstimate);
+
+  // fill rowpointer
+  Kokkos::parallel_for(
+      solverName_ + "::buildVectorNodalProlongator_adjustRowptr",
+      range_type(0, localP_nodal.numRows() + 1),
+      KOKKOS_LAMBDA(const LocalOrdinal i) {
+        if (i < localP_nodal.numRows()) {
+          for (size_t k = 0; k < dim; k++) {
+            rowptr(dim * i + k) = dim * localP_nodal.graph.row_map(i) + k;
+          }
+        } else
+          rowptr(dim * localP_nodal.numRows()) = dim * localP_nodal.graph.row_map(i);
+      });
+
+  // fill column indices and values
+  Kokkos::parallel_for(
+      solverName_ + "::buildVectorNodalProlongator_adjustColind",
+      range_type(0, localP_nodal.graph.entries.size()),
+      KOKKOS_LAMBDA(const size_t jj) {
+        for (size_t k = 0; k < dim; k++) {
+          colind(dim * jj + k) = dim * localP_nodal.graph.entries(jj) + k;
+          // vals(dim*jj+k) = localP_nodal.values(jj);
+          vals(dim * jj + k) = 1.;
+        }
+      });
 
-    size_t dim = dim_;
-
-    // Create maps
-    RCP<const Map> rowMap         = incidence->getRowMap();
-    RCP<const Map> blockColMap    = MapFactory::Build(incidence->getColMap(), dim);
-    RCP<const Map> blockDomainMap = MapFactory::Build(incidence->getDomainMap(), dim);
-
-    auto localIncidence = incidence->getLocalMatrixDevice();
-    size_t numLocalRows = rowMap->getLocalNumElements();
-    size_t numLocalColumns = dim*incidence->getColMap()->getLocalNumElements();
-    size_t nnzEstimate = dim*localIncidence.graph.entries.size();
-    lno_view_t rowptr(Kokkos::ViewAllocateWithoutInitializing("projection_rowptr_"+spaceLabel), numLocalRows+1);
-    lno_nnz_view_t colind(Kokkos::ViewAllocateWithoutInitializing("projection_colind_"+spaceLabel), nnzEstimate);
-    scalar_view_t vals("projection_vals_"+spaceLabel, nnzEstimate);
-
-    // set rowpointer
-    Kokkos::parallel_for(solverName_+"::buildProjection_adjustRowptr_"+spaceLabel,
-                         range_type(0, numLocalRows+1),
-                         KOKKOS_LAMBDA(const size_t i) {
-                           rowptr(i) = dim*localIncidence.graph.row_map(i);
-                         });
-
-    auto localNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
-
-    // set column indices and values
-    magnitudeType tol = 1e-5;
-    Kokkos::parallel_for(solverName_+"::buildProjection_enterValues_"+spaceLabel,
-                         range_type(0,numLocalRows),
-                         KOKKOS_LAMBDA(const size_t f) {
-                           for (size_t jj = localIncidence.graph.row_map(f); jj < localIncidence.graph.row_map(f+1); jj++) {
-                             for (size_t k = 0; k < dim; k++) {
-                               colind(dim*jj+k) = dim*localIncidence.graph.entries(jj)+k;
-                               if (impl_ATS::magnitude(localIncidence.values(jj)) > tol)
-                                 vals(dim*jj+k) = impl_half * localNullspace(f, k);
-                               else
-                                 vals(dim*jj+k) = impl_SC_ZERO;
-                             }
-                           }
-                         });
-
-    // Create matrix
-    typename CrsMatrix::local_matrix_type lclProjection("local projection "+spaceLabel,
+  typename CrsMatrix::local_matrix_type lclVectorNodalP("local vector nodal prolongator",
                                                         numLocalRows, numLocalColumns, nnzEstimate,
                                                         vals, rowptr, colind);
-    RCP<Matrix> projection = MatrixFactory::Build(lclProjection,
-                                                  rowMap, blockColMap,
-                                                  blockDomainMap, rowMap);
+  RCP<Matrix> vectorNodalP = MatrixFactory::Build(lclVectorNodalP,
+                                                  blockRowMap, blockColMap,
+                                                  blockDomainMap, blockRowMap);
+
+  return vectorNodalP;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    buildProlongator(const int spaceNumber,
+                     const Teuchos::RCP<Matrix> &A_nodal,
+                     const Teuchos::RCP<MultiVector> &Nullspace,
+                     Teuchos::RCP<Matrix> &Prolongator,
+                     Teuchos::RCP<MultiVector> &coarseNullspace,
+                     Teuchos::RCP<RealValuedMultiVector> &coarseNodalCoords) const {
+  using ATS         = Kokkos::ArithTraits<Scalar>;
+  using impl_Scalar = typename ATS::val_type;
+  using range_type  = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+
+  std::string typeStr;
+  switch (spaceNumber) {
+    case 0:
+      typeStr = "node";
+      TEUCHOS_ASSERT(A_nodal.is_null());
+      break;
+    case 1:
+      typeStr = "edge";
+      break;
+    case 2:
+      typeStr = "face";
+      break;
+    default:
+      TEUCHOS_ASSERT(false);
+  }
 
-    return projection;
+  const bool skipFirstLevel = !A_nodal.is_null();
 
+  RCP<Teuchos::TimeMonitor> tm;
+  if (spaceNumber > 0) {
+    tm = getTimer("special prolongator " + typeStr);
+    GetOStream(Runtime0) << solverName_ + "::compute(): building special " + typeStr + " prolongator" << std::endl;
   }
 
+  RCP<Matrix> projection = buildProjection(spaceNumber, Nullspace);
+  dump(projection, typeStr + "Projection.m");
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::buildNodalProlongator(const Teuchos::RCP<Matrix> &A_nodal,
-                                                                                 Teuchos::RCP<Matrix> &P_nodal,
-                                                                                 Teuchos::RCP<MultiVector> &Nullspace_nodal,
-                                                                                 Teuchos::RCP<RealValuedMultiVector> &CoarseCoords_nodal) const {
+  if (skipFirstLevel) {
+    RCP<Matrix> P_nodal;
+    RCP<MultiVector> coarseNodalNullspace;
 
-    RCP<Teuchos::TimeMonitor> tm = getTimer("nodal prolongator");
-    GetOStream(Runtime0) << solverName_+"::compute(): building nodal prolongator" << std::endl;
+    buildNodalProlongator(A_nodal, P_nodal, coarseNodalNullspace, coarseNodalCoords);
 
-    // build prolongator: algorithm 1 in the reference paper
-    // First, build nodal unsmoothed prolongator using the matrix A_nodal
+    dump(P_nodal, "P_nodal_" + typeStr + ".m");
+    dump(coarseNodalNullspace, "coarseNullspace_nodal_" + typeStr + ".m");
 
-    const SC SC_ONE = Teuchos::ScalarTraits<SC>::one();
+    RCP<Matrix> vectorP_nodal = buildVectorNodalProlongator(P_nodal);
 
-    {
-      Level fineLevel, coarseLevel;
-      fineLevel.SetFactoryManager(null);
-      coarseLevel.SetFactoryManager(null);
-      coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
-      fineLevel.SetLevelID(0);
-      coarseLevel.SetLevelID(1);
-      fineLevel.Set("A",A_nodal);
-      fineLevel.Set("Coordinates",NodalCoords_);
-      fineLevel.Set("DofsPerNode",1);
-      coarseLevel.setlib(A_nodal->getDomainMap()->lib());
-      fineLevel.setlib(A_nodal->getDomainMap()->lib());
-      coarseLevel.setObjectLabel(A_nodal->getObjectLabel());
-      fineLevel.setObjectLabel(A_nodal->getObjectLabel());
-
-      LocalOrdinal NSdim = 1;
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A_nodal->getRowMap(),NSdim);
-      nullSpace->putScalar(SC_ONE);
-      fineLevel.Set("Nullspace",nullSpace);
-
-      std::string algo = parameterList_.get<std::string>("multigrid algorithm");
-
-      RCP<Factory> amalgFact, dropFact, UncoupledAggFact, coarseMapFact, TentativePFact, Tfact, SaPFact;
-      amalgFact = rcp(new AmalgamationFactory());
-      coarseMapFact = rcp(new CoarseMapFactory());
-      Tfact = rcp(new CoordinatesTransferFactory());
-      if (useKokkos_) {
-        dropFact = rcp(new CoalesceDropFactory_kokkos());
-        UncoupledAggFact = rcp(new UncoupledAggregationFactory_kokkos());
-        TentativePFact = rcp(new TentativePFactory_kokkos());
-        if (algo == "sa")
-          SaPFact = rcp(new SaPFactory_kokkos());
-      } else {
-        dropFact = rcp(new CoalesceDropFactory());
-        UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-        TentativePFact = rcp(new TentativePFactory());
-        if (algo == "sa")
-          SaPFact = rcp(new SaPFactory());
-      }
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      double dropTol = parameterList_.get<double>("aggregation: drop tol");
-      std::string dropScheme = parameterList_.get<std::string>("aggregation: drop scheme");
-      std::string distLaplAlgo = parameterList_.get<std::string>("aggregation: distance laplacian algo");
-      dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-      dropFact->SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(dropScheme));
-      if (!useKokkos_)
-        dropFact->SetParameter("aggregation: distance laplacian algo",Teuchos::ParameterEntry(distLaplAlgo));
-
-      UncoupledAggFact->SetFactory("Graph", dropFact);
-      int minAggSize = parameterList_.get<int>("aggregation: min agg size");
-      UncoupledAggFact->SetParameter("aggregation: min agg size",Teuchos::ParameterEntry(minAggSize));
-      int maxAggSize = parameterList_.get<int>("aggregation: max agg size");
-      UncoupledAggFact->SetParameter("aggregation: max agg size",Teuchos::ParameterEntry(maxAggSize));
-      bool matchMLbehavior1 = parameterList_.get<bool>("aggregation: match ML phase1");
-      UncoupledAggFact->SetParameter("aggregation: match ML phase1",Teuchos::ParameterEntry(matchMLbehavior1));
-      bool matchMLbehavior2a = parameterList_.get<bool>("aggregation: match ML phase2a");
-      UncoupledAggFact->SetParameter("aggregation: match ML phase2a",Teuchos::ParameterEntry(matchMLbehavior2a));
-      bool matchMLbehavior2b = parameterList_.get<bool>("aggregation: match ML phase2b");
-      UncoupledAggFact->SetParameter("aggregation: match ML phase2b",Teuchos::ParameterEntry(matchMLbehavior2b));
-
-      coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
-
-      TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
-      TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-      Tfact->SetFactory("Aggregates", UncoupledAggFact);
-      Tfact->SetFactory("CoarseMap", coarseMapFact);
-
-      if (algo == "sa") {
-        SaPFact->SetFactory("P", TentativePFact);
-        coarseLevel.Request("P", SaPFact.get());
-      } else
-        coarseLevel.Request("P",TentativePFact.get());
-      coarseLevel.Request("Nullspace",TentativePFact.get());
-      coarseLevel.Request("Coordinates",Tfact.get());
-
-      RCP<AggregationExportFactory> aggExport;
-      bool exportVizData = parameterList_.get<bool>("aggregation: export visualization data");
-      if (exportVizData) {
-        aggExport = rcp(new AggregationExportFactory());
-        ParameterList aggExportParams;
-        aggExportParams.set("aggregation: output filename", "aggs.vtk");
-        aggExportParams.set("aggregation: output file: agg style", "Jacks");
-        aggExport->SetParameterList(aggExportParams);
-
-        aggExport->SetFactory("Aggregates", UncoupledAggFact);
-        aggExport->SetFactory("UnAmalgamationInfo", amalgFact);
-        fineLevel.Request("Aggregates",UncoupledAggFact.get());
-        fineLevel.Request("UnAmalgamationInfo",amalgFact.get());
-      }
+    dump(vectorP_nodal, "vectorP_nodal_" + typeStr + ".m");
 
-      if (algo == "sa")
-        coarseLevel.Get("P",P_nodal,SaPFact.get());
-      else
-        coarseLevel.Get("P",P_nodal,TentativePFact.get());
-      coarseLevel.Get("Nullspace",Nullspace_nodal,TentativePFact.get());
-      coarseLevel.Get("Coordinates",CoarseCoords_nodal,Tfact.get());
-
-
-      if (exportVizData)
-        aggExport->Build(fineLevel, coarseLevel);
-    }
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal, GlobalOrdinal, Node> >
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::buildVectorNodalProlongator(const Teuchos::RCP<Matrix> &P_nodal) const {
-
-    RCP<Teuchos::TimeMonitor> tm = getTimer("vectorial nodal prolongator");
-    GetOStream(Runtime0) << solverName_+"::compute(): building vectorial nodal prolongator" << std::endl;
-
-    using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-
-    typedef typename Matrix::local_matrix_type KCRS;
-    typedef typename KCRS::StaticCrsGraphType graph_t;
-    typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-    typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type scalar_view_t;
-
-    size_t dim = dim_;
-
-    // Create the matrix object
-    RCP<Map> blockRowMap    = MapFactory::Build(P_nodal->getRowMap(),    dim);
-    RCP<Map> blockColMap    = MapFactory::Build(P_nodal->getColMap(),    dim);
-    RCP<Map> blockDomainMap = MapFactory::Build(P_nodal->getDomainMap(), dim);
-
-    // Get data out of P_nodal.
-    auto localP_nodal = P_nodal->getLocalMatrixDevice();
-
-    size_t numLocalRows = blockRowMap->getLocalNumElements();
-    size_t numLocalColumns = blockColMap->getLocalNumElements();
-    size_t nnzEstimate  = dim*localP_nodal.graph.entries.size();
-    lno_view_t     rowptr(Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_rowptr"), numLocalRows+1);
-    lno_nnz_view_t colind(Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_colind"), nnzEstimate);
-    scalar_view_t  vals  (Kokkos::ViewAllocateWithoutInitializing("vectorPNodal_vals"),   nnzEstimate);
-
-    // fill rowpointer
-    Kokkos::parallel_for(solverName_+"::buildVectorNodalProlongator_adjustRowptr",
-                         range_type(0,localP_nodal.numRows()+1),
-                         KOKKOS_LAMBDA(const LocalOrdinal i) {
-                           if (i < localP_nodal.numRows()) {
-                             for (size_t k = 0; k < dim; k++) {
-                               rowptr(dim*i+k) = dim*localP_nodal.graph.row_map(i)+k;
-                             }
-                           } else
-                             rowptr(dim*localP_nodal.numRows()) = dim * localP_nodal.graph.row_map(i);
-                         });
-
-    // fill column indices and values
-    Kokkos::parallel_for(solverName_+"::buildVectorNodalProlongator_adjustColind",
-                         range_type(0,localP_nodal.graph.entries.size()),
-                         KOKKOS_LAMBDA(const size_t jj) {
-                           for (size_t k = 0; k < dim; k++) {
-                             colind(dim*jj+k) = dim*localP_nodal.graph.entries(jj)+k;
-                             // vals(dim*jj+k) = localP_nodal.values(jj);
-                             vals(dim*jj+k) = 1.;
-                           }
-                         });
-
-    typename CrsMatrix::local_matrix_type lclVectorNodalP("local vector nodal prolongator",
-                                                          numLocalRows, numLocalColumns, nnzEstimate,
-                                                          vals, rowptr, colind);
-    RCP<Matrix> vectorNodalP = MatrixFactory::Build(lclVectorNodalP,
-                                                    blockRowMap, blockColMap,
-                                                    blockDomainMap, blockRowMap);
-
-    return vectorNodalP;
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  buildProlongator(const int spaceNumber,
-                   const Teuchos::RCP<Matrix> &A_nodal,
-                   const Teuchos::RCP<MultiVector> &Nullspace,
-                   Teuchos::RCP<Matrix> &Prolongator,
-                   Teuchos::RCP<MultiVector> &coarseNullspace,
-                   Teuchos::RCP<RealValuedMultiVector> &coarseNodalCoords) const {
-
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_Scalar = typename ATS::val_type;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    std::string typeStr;
-    switch (spaceNumber) {
-      case 0:
-        typeStr = "node";
-        TEUCHOS_ASSERT(A_nodal.is_null());
-        break;
-      case 1:
-        typeStr = "edge";
-        break;
-      case 2:
-        typeStr = "face";
-        break;
-      default:
-        TEUCHOS_ASSERT(false);
-    }
-
-    const bool skipFirstLevel = !A_nodal.is_null();
-
-    RCP<Teuchos::TimeMonitor> tm;
-    if (spaceNumber > 0) {
-      tm = getTimer("special prolongator "+typeStr);
-      GetOStream(Runtime0) << solverName_+"::compute(): building special "+typeStr+" prolongator" << std::endl;
-    }
-
-    RCP<Matrix> projection = buildProjection(spaceNumber, Nullspace);
-    dump(projection, typeStr+"Projection.m");
-
-    if (skipFirstLevel) {
-      RCP<Matrix> P_nodal;
-      RCP<MultiVector> coarseNodalNullspace;
-
-      buildNodalProlongator(A_nodal, P_nodal, coarseNodalNullspace, coarseNodalCoords);
-
-      dump(P_nodal, "P_nodal_"+typeStr+".m");
-      dump(coarseNodalNullspace, "coarseNullspace_nodal_"+typeStr+".m");
-
-      RCP<Matrix> vectorP_nodal = buildVectorNodalProlongator(P_nodal);
-
-      dump(vectorP_nodal, "vectorP_nodal_"+typeStr+".m");
-
-      Prolongator = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*projection,false,*vectorP_nodal,false,Prolongator,GetOStream(Runtime0),true,true);
-
-      // This is how ML computes P22 for Darcy.
-      // The difference is the scaling by nonzeros. I don't think that that is actually needed.
-      //
-      // if (spaceNumber==2) {
-
-      //   RCP<Matrix> facesToNodes, aggsToFaces;
-      //   {
-      //     RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(D0_);
-      //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(edgesToNodes, 1e-10);
-
-      //     dump(edgesToNodes, "edgesToNodes.m");
-
-      //     RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(Dk_1_);
-      //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToEdges, 1e-10);
-      //     // facesToEdges = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToEdges, 1e-2, false);
-
-      //     dump(facesToEdges, "facesToEdges.m");
-
-      //     facesToNodes = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToEdges,false,*edgesToNodes,false,facesToNodes,GetOStream(Runtime0),true,true);
-      //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToNodes, 1e-10);
-      //     facesToNodes = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToNodes, 1e-2, false);
-      //   }
-      //   aggsToFaces = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToNodes,false,*P_nodal,false,aggsToFaces,GetOStream(Runtime0),true,true);
-
-      //   auto localP = Prolongator->getLocalMatrixDevice();
-      //   auto localAggsToFaces = aggsToFaces->getLocalMatrixDevice();
-      //   auto localNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
-
-      //   size_t dim = dim_;
-      //   Kokkos::parallel_for(solverName_+"::buildVectorNodalProlongator_adjustRowptr",
-      //                        range_type(0,localP.numRows()),
-      //                        KOKKOS_LAMBDA(const LocalOrdinal i) {
-      //                          LocalOrdinal nonzeros = localAggsToFaces.graph.row_map(i+1)-localAggsToFaces.graph.row_map(i);
-      //                          for (LocalOrdinal jj = localAggsToFaces.graph.row_map(i); jj < localAggsToFaces.graph.row_map(i+1); jj++ ) {
-      //                            LocalOrdinal j = localAggsToFaces.graph.entries(jj);
-      //                            for (LocalOrdinal k = 0; k<dim; k++)
-      //                              for (LocalOrdinal kk = localP.graph.row_map(i); kk < localP.graph.row_map(i+1); kk++)
-      //                                if (localP.graph.entries(kk) == (dim * j+k)) {
-      //                                  localP.values(kk) = localNullspace(i, k) / nonzeros;
-      //                                  break;
-      //                                }
-      //                          }
-      //                        });
-      // }
-      //
-
-      size_t dim = dim_;
-      coarseNullspace = MultiVectorFactory::Build(vectorP_nodal->getDomainMap(), dim);
-
-      auto localNullspace_nodal = coarseNodalNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      auto localNullspace_coarse = coarseNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
-      Kokkos::parallel_for(solverName_+"::buildProlongator_nullspace_"+typeStr,
-                           range_type(0,coarseNodalNullspace->getLocalLength()),
-                           KOKKOS_LAMBDA(const size_t i) {
-                             impl_Scalar val = localNullspace_nodal(i,0);
-                             for (size_t j = 0; j < dim; j++)
-                               localNullspace_coarse(dim*i+j, j) = val;
-                           });
+    Prolongator = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*projection, false, *vectorP_nodal, false, Prolongator, GetOStream(Runtime0), true, true);
 
-    } else {
-      Prolongator = projection;
-      coarseNodalCoords = NodalCoords_;
-
-      if (spaceNumber == 0) {
-        // nothing, just use the default constant vector
-      } else if (spaceNumber >= 1) {
-        size_t dim = dim_;
-        coarseNullspace = MultiVectorFactory::Build(projection->getDomainMap(), dim);
-        auto localNullspace_coarse = coarseNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
-        Kokkos::parallel_for(solverName_+"::buildProlongator_nullspace_"+typeStr,
-                             range_type(0,coarseNullspace->getLocalLength() / dim),
-                             KOKKOS_LAMBDA(const size_t i) {
-                               for (size_t j = 0; j < dim; j++)
-                                 localNullspace_coarse(dim*i+j, j) = 1.0;
-                             });
-      }
-    }
+    // This is how ML computes P22 for Darcy.
+    // The difference is the scaling by nonzeros. I don't think that that is actually needed.
+    //
+    // if (spaceNumber==2) {
+
+    //   RCP<Matrix> facesToNodes, aggsToFaces;
+    //   {
+    //     RCP<Matrix> edgesToNodes = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(D0_);
+    //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(edgesToNodes, 1e-10);
+
+    //     dump(edgesToNodes, "edgesToNodes.m");
+
+    //     RCP<Matrix> facesToEdges = Xpetra::MatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildCopy(Dk_1_);
+    //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToEdges, 1e-10);
+    //     // facesToEdges = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToEdges, 1e-2, false);
+
+    //     dump(facesToEdges, "facesToEdges.m");
+
+    //     facesToNodes = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToEdges,false,*edgesToNodes,false,facesToNodes,GetOStream(Runtime0),true,true);
+    //     Maxwell_Utils<SC,LO,GO,NO>::thresholdedAbs(facesToNodes, 1e-10);
+    //     facesToNodes = Maxwell_Utils<SC,LO,GO,NO>::removeExplicitZeros(facesToNodes, 1e-2, false);
+    //   }
+    //   aggsToFaces = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*facesToNodes,false,*P_nodal,false,aggsToFaces,GetOStream(Runtime0),true,true);
+
+    //   auto localP = Prolongator->getLocalMatrixDevice();
+    //   auto localAggsToFaces = aggsToFaces->getLocalMatrixDevice();
+    //   auto localNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
+
+    //   size_t dim = dim_;
+    //   Kokkos::parallel_for(solverName_+"::buildVectorNodalProlongator_adjustRowptr",
+    //                        range_type(0,localP.numRows()),
+    //                        KOKKOS_LAMBDA(const LocalOrdinal i) {
+    //                          LocalOrdinal nonzeros = localAggsToFaces.graph.row_map(i+1)-localAggsToFaces.graph.row_map(i);
+    //                          for (LocalOrdinal jj = localAggsToFaces.graph.row_map(i); jj < localAggsToFaces.graph.row_map(i+1); jj++ ) {
+    //                            LocalOrdinal j = localAggsToFaces.graph.entries(jj);
+    //                            for (LocalOrdinal k = 0; k<dim; k++)
+    //                              for (LocalOrdinal kk = localP.graph.row_map(i); kk < localP.graph.row_map(i+1); kk++)
+    //                                if (localP.graph.entries(kk) == (dim * j+k)) {
+    //                                  localP.values(kk) = localNullspace(i, k) / nonzeros;
+    //                                  break;
+    //                                }
+    //                          }
+    //                        });
+    // }
+    //
 
-  }
+    size_t dim      = dim_;
+    coarseNullspace = MultiVectorFactory::Build(vectorP_nodal->getDomainMap(), dim);
 
+    auto localNullspace_nodal  = coarseNodalNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto localNullspace_coarse = coarseNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
+    Kokkos::parallel_for(
+        solverName_ + "::buildProlongator_nullspace_" + typeStr,
+        range_type(0, coarseNodalNullspace->getLocalLength()),
+        KOKKOS_LAMBDA(const size_t i) {
+          impl_Scalar val = localNullspace_nodal(i, 0);
+          for (size_t j = 0; j < dim; j++)
+            localNullspace_coarse(dim * i + j, j) = val;
+        });
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::setupSubSolve(Teuchos::RCP<Hierarchy> &hierarchy,
-                                                                    Teuchos::RCP<Operator> &thyraPrecOp,
-                                                                    const Teuchos::RCP<Matrix> &A,
-                                                                    const Teuchos::RCP<MultiVector> &Nullspace,
-                                                                    const Teuchos::RCP<RealValuedMultiVector> &Coords,
-                                                                    Teuchos::ParameterList &params,
-                                                                    std::string &label,
-                                                                    const bool reuse,
-                                                                    const bool isSingular) {
-    int oldRank = SetProcRankVerbose(A->getDomainMap()->getComm()->getRank());
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> pl = rcp(new ParameterList());
-      pl->set("printLoadBalancingInfo", true);
-      pl->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*A, label, pl);
+  } else {
+    Prolongator       = projection;
+    coarseNodalCoords = NodalCoords_;
+
+    if (spaceNumber == 0) {
+      // nothing, just use the default constant vector
+    } else if (spaceNumber >= 1) {
+      size_t dim                 = dim_;
+      coarseNullspace            = MultiVectorFactory::Build(projection->getDomainMap(), dim);
+      auto localNullspace_coarse = coarseNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
+      Kokkos::parallel_for(
+          solverName_ + "::buildProlongator_nullspace_" + typeStr,
+          range_type(0, coarseNullspace->getLocalLength() / dim),
+          KOKKOS_LAMBDA(const size_t i) {
+            for (size_t j = 0; j < dim; j++)
+              localNullspace_coarse(dim * i + j, j) = 1.0;
+          });
     }
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setupSubSolve(Teuchos::RCP<Hierarchy> &hierarchy,
+                                                                          Teuchos::RCP<Operator> &thyraPrecOp,
+                                                                          const Teuchos::RCP<Matrix> &A,
+                                                                          const Teuchos::RCP<MultiVector> &Nullspace,
+                                                                          const Teuchos::RCP<RealValuedMultiVector> &Coords,
+                                                                          Teuchos::ParameterList &params,
+                                                                          std::string &label,
+                                                                          const bool reuse,
+                                                                          const bool isSingular) {
+  int oldRank = SetProcRankVerbose(A->getDomainMap()->getComm()->getRank());
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> pl = rcp(new ParameterList());
+    pl->set("printLoadBalancingInfo", true);
+    pl->set("printCommInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*A, label, pl);
+  }
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    if (params.isType<std::string>("Preconditioner Type")) {
-      TEUCHOS_ASSERT(!reuse);
-      // build a Stratimikos preconditioner
-      if (params.get<std::string>("Preconditioner Type") == "MueLu") {
-        ParameterList& userParamList = params.sublist("Preconditioner Types").sublist("MueLu").sublist("user data");
-        if (!Nullspace.is_null())
-          userParamList.set<RCP<MultiVector> >("Nullspace", Nullspace);
-        userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", Coords);
-      }
-      thyraPrecOp = rcp(new XpetraThyraLinearOp<Scalar,LocalOrdinal,GlobalOrdinal,Node>(coarseA11_, rcp(&params, false)));
-    } else
+  if (params.isType<std::string>("Preconditioner Type")) {
+    TEUCHOS_ASSERT(!reuse);
+    // build a Stratimikos preconditioner
+    if (params.get<std::string>("Preconditioner Type") == "MueLu") {
+      ParameterList &userParamList = params.sublist("Preconditioner Types").sublist("MueLu").sublist("user data");
+      if (!Nullspace.is_null())
+        userParamList.set<RCP<MultiVector> >("Nullspace", Nullspace);
+      userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", Coords);
+    }
+    thyraPrecOp = rcp(new XpetraThyraLinearOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(coarseA11_, rcp(&params, false)));
+  } else
 #endif
-      {
-        // build a MueLu hierarchy
-
-        if (!reuse) {
-          ParameterList& userParamList = params.sublist("user data");
-          if (!Coords.is_null())
-            userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", Coords);
-          if (!Nullspace.is_null())
-            userParamList.set<RCP<MultiVector> >("Nullspace", Nullspace);
-
-          if (isSingular) {
-            std::string coarseType = "";
-            if (params.isParameter("coarse: type")) {
-              coarseType = params.get<std::string>("coarse: type");
-              // Transform string to "Abcde" notation
-              std::transform(coarseType.begin(),   coarseType.end(),   coarseType.begin(), ::tolower);
-              std::transform(coarseType.begin(), ++coarseType.begin(), coarseType.begin(), ::toupper);
-            }
-            if ((coarseType == "" ||
-                 coarseType == "Klu" ||
-                 coarseType == "Klu2") &&
-                (!params.isSublist("coarse: params") ||
-                 !params.sublist("coarse: params").isParameter("fix nullspace")))
-              params.sublist("coarse: params").set("fix nullspace",true);
-          }
+  {
+    // build a MueLu hierarchy
 
-          hierarchy = MueLu::CreateXpetraPreconditioner(A, params);
-        } else {
-          RCP<MueLu::Level> level0 = hierarchy->GetLevel(0);
-          level0->Set("A", A);
-          hierarchy->SetupRe();
+    if (!reuse) {
+      ParameterList &userParamList = params.sublist("user data");
+      if (!Coords.is_null())
+        userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", Coords);
+      if (!Nullspace.is_null())
+        userParamList.set<RCP<MultiVector> >("Nullspace", Nullspace);
+
+      if (isSingular) {
+        std::string coarseType = "";
+        if (params.isParameter("coarse: type")) {
+          coarseType = params.get<std::string>("coarse: type");
+          // Transform string to "Abcde" notation
+          std::transform(coarseType.begin(), coarseType.end(), coarseType.begin(), ::tolower);
+          std::transform(coarseType.begin(), ++coarseType.begin(), coarseType.begin(), ::toupper);
         }
+        if ((coarseType == "" ||
+             coarseType == "Klu" ||
+             coarseType == "Klu2") &&
+            (!params.isSublist("coarse: params") ||
+             !params.sublist("coarse: params").isParameter("fix nullspace")))
+          params.sublist("coarse: params").set("fix nullspace", true);
       }
-    SetProcRankVerbose(oldRank);
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::resetMatrix(RCP<Matrix> SM_Matrix_new, bool ComputePrec) {
-    bool reuse = !SM_Matrix_.is_null();
-    SM_Matrix_ = SM_Matrix_new;
-    dump(SM_Matrix_, "SM.m");
-    if (ComputePrec)
-      compute(reuse);
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::applyInverseAdditive(const MultiVector& RHS, MultiVector& X) const {
-
-    // residual(SM_Matrix_, X, RHS, residual_)
-    //
-    // P11res_ = P11_^T*residual_ or P11res_ = R11_*residual_
-    //
-    // Dres_ = Dk_1_^T*residual or Dres_ = Dk_1_T_*residual
-    //
-    // if ImporterCoarse11_ is not null
-    //     ImporterCoarse11: P11res_ -> P11resTmp_
-    // if Importer22_ is not null
-    //     Importer22: Dres_ -> DresTmp_
-    //
-    // if coarseA11 is not null
-    //
-    //     Hierarchy11(P11resSubComm, P11xSubComm)   P11resSubComm aliases P11res or P11resTmp
-    //                                               P11xSubComm aliases P11x
-    //
-    // if A22 is not null
-    //
-    //     Hierarchy22(DresSubComm, DxSubComm)       DresSubComm aliases Dres or DresTmp
-    //                                               DxSubComm aliases Dx
-    //
-    // if ImporterCoarse11_ is not null
-    //     ImporterCoarse11: P11xTmp_ -> P11x
-    // if Importer22_ is not null
-    //     Importer22: DxTmp_ -> Dx_
-    //
-    // if fuse
-    //     X += P11*P11x
-    //     X += P11*Dx
-    // else
-    //     residual = P11*P11x
-    //     residual += Dk_1*Dx
-    //     X += residual
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
 
-    { // compute residual
-
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
-      Utilities::Residual(*SM_Matrix_, X, RHS, *residual_);
+      hierarchy = MueLu::CreateXpetraPreconditioner(A, params);
+    } else {
+      RCP<MueLu::Level> level0 = hierarchy->GetLevel(0);
+      level0->Set("A", A);
+      hierarchy->SetupRe();
     }
+  }
+  SetProcRankVerbose(oldRank);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::resetMatrix(RCP<Matrix> SM_Matrix_new, bool ComputePrec) {
+  bool reuse = !SM_Matrix_.is_null();
+  SM_Matrix_ = SM_Matrix_new;
+  dump(SM_Matrix_, "SM.m");
+  if (ComputePrec)
+    compute(reuse);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::applyInverseAdditive(const MultiVector &RHS, MultiVector &X) const {
+  // residual(SM_Matrix_, X, RHS, residual_)
+  //
+  // P11res_ = P11_^T*residual_ or P11res_ = R11_*residual_
+  //
+  // Dres_ = Dk_1_^T*residual or Dres_ = Dk_1_T_*residual
+  //
+  // if ImporterCoarse11_ is not null
+  //     ImporterCoarse11: P11res_ -> P11resTmp_
+  // if Importer22_ is not null
+  //     Importer22: Dres_ -> DresTmp_
+  //
+  // if coarseA11 is not null
+  //
+  //     Hierarchy11(P11resSubComm, P11xSubComm)   P11resSubComm aliases P11res or P11resTmp
+  //                                               P11xSubComm aliases P11x
+  //
+  // if A22 is not null
+  //
+  //     Hierarchy22(DresSubComm, DxSubComm)       DresSubComm aliases Dres or DresTmp
+  //                                               DxSubComm aliases Dx
+  //
+  // if ImporterCoarse11_ is not null
+  //     ImporterCoarse11: P11xTmp_ -> P11x
+  // if Importer22_ is not null
+  //     Importer22: DxTmp_ -> Dx_
+  //
+  // if fuse
+  //     X += P11*P11x
+  //     X += P11*Dx
+  // else
+  //     residual = P11*P11x
+  //     residual += Dk_1*Dx
+  //     X += residual
+
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  {  // compute residual
+
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
+    Utilities::Residual(*SM_Matrix_, X, RHS, *residual_);
+  }
 
-    { // restrict residual to sub-hierarchies
+  {  // restrict residual to sub-hierarchies
 
-      if (implicitTranspose_) {
+    if (implicitTranspose_) {
+      {
+        RCP<Teuchos::TimeMonitor> tmRes = getTimer("restriction coarse (1,1) (implicit)");
+        P11_->apply(*residual_, *P11res_, Teuchos::TRANS);
+      }
+      if (!onlyBoundary22_) {
+        RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (implicit)");
+        Dk_1_->apply(*residual_, *Dres_, Teuchos::TRANS);
+      }
+    } else {
+      if (Dk_1_T_R11_colMapsMatch_) {
+        // Column maps of D_T and R11 match, and we're running Tpetra
         {
-          RCP<Teuchos::TimeMonitor> tmRes = getTimer("restriction coarse (1,1) (implicit)");
-          P11_->apply(*residual_,*P11res_,Teuchos::TRANS);
+          RCP<Teuchos::TimeMonitor> tmD = getTimer("restrictions import");
+          DTR11Tmp_->doImport(*residual_, *rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix()->getCrsGraph()->getImporter(), Xpetra::INSERT);
         }
         if (!onlyBoundary22_) {
-          RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (implicit)");
-          Dk_1_->apply(*residual_,*Dres_,Teuchos::TRANS);
+          RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (explicit)");
+          rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_T_)->getCrsMatrix())->getTpetra_CrsMatrix()->localApply(toTpetra(*DTR11Tmp_), toTpetra(*Dres_), Teuchos::NO_TRANS);
+        }
+        {
+          RCP<Teuchos::TimeMonitor> tmP11 = getTimer("restriction coarse (1,1) (explicit)");
+          rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix())->getTpetra_CrsMatrix()->localApply(toTpetra(*DTR11Tmp_), toTpetra(*P11res_), Teuchos::NO_TRANS);
         }
       } else {
-        if (Dk_1_T_R11_colMapsMatch_) {
-          // Column maps of D_T and R11 match, and we're running Tpetra
-          {
-            RCP<Teuchos::TimeMonitor> tmD = getTimer("restrictions import");
-            DTR11Tmp_->doImport(*residual_, *rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix()->getCrsGraph()->getImporter(), Xpetra::INSERT);
-          }
-          if (!onlyBoundary22_) {
-            RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (explicit)");
-            rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(Dk_1_T_)->getCrsMatrix())->getTpetra_CrsMatrix()->localApply(toTpetra(*DTR11Tmp_),toTpetra(*Dres_),Teuchos::NO_TRANS);
-          }
-          {
-            RCP<Teuchos::TimeMonitor> tmP11 = getTimer("restriction coarse (1,1) (explicit)");
-            rcp_dynamic_cast<TpetraCrsMatrix>(rcp_dynamic_cast<CrsMatrixWrap>(R11_)->getCrsMatrix())->getTpetra_CrsMatrix()->localApply(toTpetra(*DTR11Tmp_),toTpetra(*P11res_),Teuchos::NO_TRANS);
-          }
-        } else {
-          {
-            RCP<Teuchos::TimeMonitor> tmP11 = getTimer("restriction coarse (1,1) (explicit)");
-            R11_->apply(*residual_,*P11res_,Teuchos::NO_TRANS);
-          }
-          if (!onlyBoundary22_) {
-            RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (explicit)");
-            Dk_1_T_->apply(*residual_,*Dres_,Teuchos::NO_TRANS);
-          }
+        {
+          RCP<Teuchos::TimeMonitor> tmP11 = getTimer("restriction coarse (1,1) (explicit)");
+          R11_->apply(*residual_, *P11res_, Teuchos::NO_TRANS);
+        }
+        if (!onlyBoundary22_) {
+          RCP<Teuchos::TimeMonitor> tmD = getTimer("restriction (2,2) (explicit)");
+          Dk_1_T_->apply(*residual_, *Dres_, Teuchos::NO_TRANS);
         }
       }
     }
+  }
 
-    {
-      RCP<Teuchos::TimeMonitor> tmSubSolves = getTimer("subsolves");
+  {
+    RCP<Teuchos::TimeMonitor> tmSubSolves = getTimer("subsolves");
 
-      // block diagonal preconditioner on 2x2 (V-cycle for diagonal blocks)
+    // block diagonal preconditioner on 2x2 (V-cycle for diagonal blocks)
 
-      if (!ImporterCoarse11_.is_null() && !implicitTranspose_) {
-        RCP<Teuchos::TimeMonitor> tmH = getTimer("import coarse (1,1)");
-        P11resTmp_->beginImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
-      }
-      if (!onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_) {
-        RCP<Teuchos::TimeMonitor> tm22 = getTimer("import (2,2)");
-        DresTmp_->beginImport(*Dres_, *Importer22_, Xpetra::INSERT);
-      }
+    if (!ImporterCoarse11_.is_null() && !implicitTranspose_) {
+      RCP<Teuchos::TimeMonitor> tmH = getTimer("import coarse (1,1)");
+      P11resTmp_->beginImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
+    }
+    if (!onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_) {
+      RCP<Teuchos::TimeMonitor> tm22 = getTimer("import (2,2)");
+      DresTmp_->beginImport(*Dres_, *Importer22_, Xpetra::INSERT);
+    }
 
-      // iterate on coarse (1, 1) block
-      if (!coarseA11_.is_null()) {
-        if (!ImporterCoarse11_.is_null() && !implicitTranspose_)
-          P11resTmp_->endImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
+    // iterate on coarse (1, 1) block
+    if (!coarseA11_.is_null()) {
+      if (!ImporterCoarse11_.is_null() && !implicitTranspose_)
+        P11resTmp_->endImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
 
-        RCP<Teuchos::TimeMonitor> tmH = getTimer("solve coarse (1,1)", coarseA11_->getRowMap()->getComm());
+      RCP<Teuchos::TimeMonitor> tmH = getTimer("solve coarse (1,1)", coarseA11_->getRowMap()->getComm());
 
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-        if (!thyraPrecOpH_.is_null()) {
-          Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-          thyraPrecOpH_->apply(*P11resSubComm_, *P11xSubComm_, Teuchos::NO_TRANS, one, zero);
-        }
-        else
+      if (!thyraPrecOpH_.is_null()) {
+        Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+        thyraPrecOpH_->apply(*P11resSubComm_, *P11xSubComm_, Teuchos::NO_TRANS, one, zero);
+      } else
 #endif
-           HierarchyCoarse11_->Iterate(*P11resSubComm_, *P11xSubComm_, numItersCoarse11_, true);
-      }
-
-      // iterate on (2, 2) block
-      if (!A22_.is_null()) {
+        HierarchyCoarse11_->Iterate(*P11resSubComm_, *P11xSubComm_, numItersCoarse11_, true);
+    }
 
-        if (!onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_)
-          DresTmp_->endImport(*Dres_, *Importer22_, Xpetra::INSERT);
+    // iterate on (2, 2) block
+    if (!A22_.is_null()) {
+      if (!onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_)
+        DresTmp_->endImport(*Dres_, *Importer22_, Xpetra::INSERT);
 
-        RCP<Teuchos::TimeMonitor> tm22 = getTimer("solve (2,2)", A22_->getRowMap()->getComm());
+      RCP<Teuchos::TimeMonitor> tm22 = getTimer("solve (2,2)", A22_->getRowMap()->getComm());
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-        if (!thyraPrecOp22_.is_null()) {
-          Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-          thyraPrecOp22_->apply(*DresSubComm_, *DxSubComm_, Teuchos::NO_TRANS, one, zero);
-        }
-        else
+      if (!thyraPrecOp22_.is_null()) {
+        Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+        thyraPrecOp22_->apply(*DresSubComm_, *DxSubComm_, Teuchos::NO_TRANS, one, zero);
+      } else
 #endif
-          Hierarchy22_->Iterate(*DresSubComm_, *DxSubComm_, numIters22_, true);
-      }
-
-      if (coarseA11_.is_null() && !ImporterCoarse11_.is_null() && !implicitTranspose_)
-        P11resTmp_->endImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
-      if (A22_.is_null() && !onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_)
-        DresTmp_->endImport(*Dres_, *Importer22_, Xpetra::INSERT);
-    }
-
-    if (fuseProlongationAndUpdate_) {
-      { // prolongate (1,1) block
-        RCP<Teuchos::TimeMonitor> tmP11 = getTimer("prolongation coarse (1,1) (fused)");
-        P11_->apply(*P11x_,X,Teuchos::NO_TRANS,one,one);
-      }
-
-      if (!onlyBoundary22_) { // prolongate (2,2) block
-        RCP<Teuchos::TimeMonitor> tmD = getTimer("prolongation (2,2) (fused)");
-        Dk_1_->apply(*Dx_,X,Teuchos::NO_TRANS,one,one);
-      }
-    } else {
-      { // prolongate (1,1) block
-        RCP<Teuchos::TimeMonitor> tmP11 = getTimer("prolongation coarse (1,1) (unfused)");
-        P11_->apply(*P11x_,*residual_,Teuchos::NO_TRANS);
-      }
-
-      if (!onlyBoundary22_) { // prolongate (2,2) block
-        RCP<Teuchos::TimeMonitor> tmD = getTimer("prolongation (2,2) (unfused)");
-        Dk_1_->apply(*Dx_,*residual_,Teuchos::NO_TRANS,one,one);
-      }
-
-      { // update current solution
-        RCP<Teuchos::TimeMonitor> tmUpdate = getTimer("update");
-        X.update(one, *residual_, one);
-      }
+        Hierarchy22_->Iterate(*DresSubComm_, *DxSubComm_, numIters22_, true);
     }
 
+    if (coarseA11_.is_null() && !ImporterCoarse11_.is_null() && !implicitTranspose_)
+      P11resTmp_->endImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
+    if (A22_.is_null() && !onlyBoundary22_ && !Importer22_.is_null() && !implicitTranspose_)
+      DresTmp_->endImport(*Dres_, *Importer22_, Xpetra::INSERT);
   }
 
+  if (fuseProlongationAndUpdate_) {
+    {  // prolongate (1,1) block
+      RCP<Teuchos::TimeMonitor> tmP11 = getTimer("prolongation coarse (1,1) (fused)");
+      P11_->apply(*P11x_, X, Teuchos::NO_TRANS, one, one);
+    }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::solveH(const MultiVector& RHS, MultiVector& X) const {
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    { // compute residual
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
-      Utilities::Residual(*SM_Matrix_, X, RHS,*residual_);
-      if (implicitTranspose_)
-        P11_->apply(*residual_,*P11res_,Teuchos::TRANS);
-      else
-        R11_->apply(*residual_,*P11res_,Teuchos::NO_TRANS);
+    if (!onlyBoundary22_) {  // prolongate (2,2) block
+      RCP<Teuchos::TimeMonitor> tmD = getTimer("prolongation (2,2) (fused)");
+      Dk_1_->apply(*Dx_, X, Teuchos::NO_TRANS, one, one);
+    }
+  } else {
+    {  // prolongate (1,1) block
+      RCP<Teuchos::TimeMonitor> tmP11 = getTimer("prolongation coarse (1,1) (unfused)");
+      P11_->apply(*P11x_, *residual_, Teuchos::NO_TRANS);
     }
 
-    { // solve coarse (1,1) block
-      if (!ImporterCoarse11_.is_null() && !implicitTranspose_) {
-        RCP<Teuchos::TimeMonitor> tmH = getTimer("import coarse (1,1)");
-        P11resTmp_->doImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
-      }
-      if (!coarseA11_.is_null()) {
-        RCP<Teuchos::TimeMonitor> tmH = getTimer("solve coarse (1,1)", coarseA11_->getRowMap()->getComm());
-        HierarchyCoarse11_->Iterate(*P11resSubComm_, *P11xSubComm_, numItersCoarse11_, true);
-      }
+    if (!onlyBoundary22_) {  // prolongate (2,2) block
+      RCP<Teuchos::TimeMonitor> tmD = getTimer("prolongation (2,2) (unfused)");
+      Dk_1_->apply(*Dx_, *residual_, Teuchos::NO_TRANS, one, one);
     }
 
-    { // update current solution
-      RCP<Teuchos::TimeMonitor> tmUp = getTimer("update");
-      P11_->apply(*P11x_,*residual_,Teuchos::NO_TRANS);
+    {  // update current solution
+      RCP<Teuchos::TimeMonitor> tmUpdate = getTimer("update");
       X.update(one, *residual_, one);
     }
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::solveH(const MultiVector &RHS, MultiVector &X) const {
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
 
+  {  // compute residual
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
+    Utilities::Residual(*SM_Matrix_, X, RHS, *residual_);
+    if (implicitTranspose_)
+      P11_->apply(*residual_, *P11res_, Teuchos::TRANS);
+    else
+      R11_->apply(*residual_, *P11res_, Teuchos::NO_TRANS);
   }
 
+  {  // solve coarse (1,1) block
+    if (!ImporterCoarse11_.is_null() && !implicitTranspose_) {
+      RCP<Teuchos::TimeMonitor> tmH = getTimer("import coarse (1,1)");
+      P11resTmp_->doImport(*P11res_, *ImporterCoarse11_, Xpetra::INSERT);
+    }
+    if (!coarseA11_.is_null()) {
+      RCP<Teuchos::TimeMonitor> tmH = getTimer("solve coarse (1,1)", coarseA11_->getRowMap()->getComm());
+      HierarchyCoarse11_->Iterate(*P11resSubComm_, *P11xSubComm_, numItersCoarse11_, true);
+    }
+  }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::solve22(const MultiVector& RHS, MultiVector& X) const {
+  {  // update current solution
+    RCP<Teuchos::TimeMonitor> tmUp = getTimer("update");
+    P11_->apply(*P11x_, *residual_, Teuchos::NO_TRANS);
+    X.update(one, *residual_, one);
+  }
+}
 
-    if (onlyBoundary22_)
-      return;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::solve22(const MultiVector &RHS, MultiVector &X) const {
+  if (onlyBoundary22_)
+    return;
 
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
 
-    { // compute residual
-      RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
-      Utilities::Residual(*SM_Matrix_, X, RHS, *residual_);
-      if (implicitTranspose_)
-        Dk_1_->apply(*residual_,*Dres_,Teuchos::TRANS);
-      else
-        Dk_1_T_->apply(*residual_,*Dres_,Teuchos::NO_TRANS);
-    }
+  {  // compute residual
+    RCP<Teuchos::TimeMonitor> tmRes = getTimer("residual calculation");
+    Utilities::Residual(*SM_Matrix_, X, RHS, *residual_);
+    if (implicitTranspose_)
+      Dk_1_->apply(*residual_, *Dres_, Teuchos::TRANS);
+    else
+      Dk_1_T_->apply(*residual_, *Dres_, Teuchos::NO_TRANS);
+  }
 
-    { // solve (2,2) block
-      if (!Importer22_.is_null() && !implicitTranspose_) {
-        RCP<Teuchos::TimeMonitor> tm22 = getTimer("import (2,2)");
-        DresTmp_->doImport(*Dres_, *Importer22_, Xpetra::INSERT);
-      }
-      if (!A22_.is_null()) {
-        RCP<Teuchos::TimeMonitor> tm22 = getTimer("solve (2,2)", A22_->getRowMap()->getComm());
-        Hierarchy22_->Iterate(*DresSubComm_, *DxSubComm_, numIters22_, true);
-      }
+  {  // solve (2,2) block
+    if (!Importer22_.is_null() && !implicitTranspose_) {
+      RCP<Teuchos::TimeMonitor> tm22 = getTimer("import (2,2)");
+      DresTmp_->doImport(*Dres_, *Importer22_, Xpetra::INSERT);
     }
-
-    { //update current solution
-      RCP<Teuchos::TimeMonitor> tmUp = getTimer("update");
-      Dk_1_->apply(*Dx_,*residual_,Teuchos::NO_TRANS);
-      X.update(one, *residual_, one);
+    if (!A22_.is_null()) {
+      RCP<Teuchos::TimeMonitor> tm22 = getTimer("solve (2,2)", A22_->getRowMap()->getComm());
+      Hierarchy22_->Iterate(*DresSubComm_, *DxSubComm_, numIters22_, true);
     }
-
   }
 
+  {  // update current solution
+    RCP<Teuchos::TimeMonitor> tmUp = getTimer("update");
+    Dk_1_->apply(*Dx_, *residual_, Teuchos::NO_TRANS);
+    X.update(one, *residual_, one);
+  }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::apply (const MultiVector& RHS, MultiVector& X,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::apply(const MultiVector &RHS, MultiVector &X,
                                                                   Teuchos::ETransp /* mode */,
                                                                   Scalar /* alpha */,
                                                                   Scalar /* beta */) const {
+  RCP<Teuchos::TimeMonitor> tm = getTimer("solve");
 
-    RCP<Teuchos::TimeMonitor> tm = getTimer("solve");
+  // make sure that we have enough temporary memory
+  if (!onlyBoundary11_ && X.getNumVectors() != P11res_->getNumVectors())
+    allocateMemory(X.getNumVectors());
 
-    // make sure that we have enough temporary memory
-    if (!onlyBoundary11_ && X.getNumVectors() != P11res_->getNumVectors())
-      allocateMemory(X.getNumVectors());
+  {  // apply pre-smoothing
 
-    { // apply pre-smoothing
+    RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
 
-      RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
+    PreSmoother11_->Apply(X, RHS, use_as_preconditioner_);
+  }
 
-      PreSmoother11_->Apply(X, RHS, use_as_preconditioner_);
-    }
-
-    // do solve for the 2x2 block system
-    if(mode_=="additive")
-      applyInverseAdditive(RHS,X);
-    else if(mode_=="121") {
-      solveH(RHS,X);
-      solve22(RHS,X);
-      solveH(RHS,X);
-    } else if(mode_=="212") {
-      solve22(RHS,X);
-      solveH(RHS,X);
-      solve22(RHS,X);
-    } else if(mode_=="1")
-      solveH(RHS,X);
-    else if(mode_=="2")
-      solve22(RHS,X);
-    else if(mode_=="7") {
-      solveH(RHS,X);
-      { // apply pre-smoothing
-
-        RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
-
-        PreSmoother11_->Apply(X, RHS, false);
-      }
-      solve22(RHS,X);
-      { // apply post-smoothing
+  // do solve for the 2x2 block system
+  if (mode_ == "additive")
+    applyInverseAdditive(RHS, X);
+  else if (mode_ == "121") {
+    solveH(RHS, X);
+    solve22(RHS, X);
+    solveH(RHS, X);
+  } else if (mode_ == "212") {
+    solve22(RHS, X);
+    solveH(RHS, X);
+    solve22(RHS, X);
+  } else if (mode_ == "1")
+    solveH(RHS, X);
+  else if (mode_ == "2")
+    solve22(RHS, X);
+  else if (mode_ == "7") {
+    solveH(RHS, X);
+    {  // apply pre-smoothing
 
-        RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
+      RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
 
-        PostSmoother11_->Apply(X, RHS, false);
-      }
-      solveH(RHS,X);
-    } else if(mode_=="none") {
-      // do nothing
+      PreSmoother11_->Apply(X, RHS, false);
     }
-    else
-      applyInverseAdditive(RHS,X);
-
-    { // apply post-smoothing
+    solve22(RHS, X);
+    {  // apply post-smoothing
 
       RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
 
       PostSmoother11_->Apply(X, RHS, false);
     }
+    solveH(RHS, X);
+  } else if (mode_ == "none") {
+    // do nothing
+  } else
+    applyInverseAdditive(RHS, X);
 
-  }
-
-    
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::hasTransposeApply() const {
-    return false;
-  }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  RefMaxwell(const Teuchos::RCP<Matrix> & SM_Matrix,
-             Teuchos::ParameterList& List,
-             bool ComputePrec)
-    {
-
-      int spaceNumber  = List.get<int>("refmaxwell: space number", 1);
+  {  // apply post-smoothing
 
-      RCP<Matrix> Dk_1, Dk_2, D0;
-      RCP<Matrix> M1_beta, M1_alpha;
-      RCP<Matrix> Mk_one, Mk_1_one;
-      RCP<Matrix> invMk_1_invBeta, invMk_2_invAlpha;
-      RCP<MultiVector> Nullspace11, Nullspace22;
-      RCP<RealValuedMultiVector> NodalCoords;
+    RCP<Teuchos::TimeMonitor> tmSm = getTimer("smoothing");
 
+    PostSmoother11_->Apply(X, RHS, false);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::hasTransposeApply() const {
+  return false;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    RefMaxwell(const Teuchos::RCP<Matrix> &SM_Matrix,
+               Teuchos::ParameterList &List,
+               bool ComputePrec) {
+  int spaceNumber = List.get<int>("refmaxwell: space number", 1);
+
+  RCP<Matrix> Dk_1, Dk_2, D0;
+  RCP<Matrix> M1_beta, M1_alpha;
+  RCP<Matrix> Mk_one, Mk_1_one;
+  RCP<Matrix> invMk_1_invBeta, invMk_2_invAlpha;
+  RCP<MultiVector> Nullspace11, Nullspace22;
+  RCP<RealValuedMultiVector> NodalCoords;
+
+  Dk_1 = pop(List, "Dk_1", Dk_1);
+  Dk_2 = pop<RCP<Matrix> >(List, "Dk_2", Dk_2);
+  D0   = pop<RCP<Matrix> >(List, "D0", D0);
+
+  M1_beta  = pop<RCP<Matrix> >(List, "M1_beta", M1_beta);
+  M1_alpha = pop<RCP<Matrix> >(List, "M1_alpha", M1_alpha);
+
+  Mk_one   = pop<RCP<Matrix> >(List, "Mk_one", Mk_one);
+  Mk_1_one = pop<RCP<Matrix> >(List, "Mk_1_one", Mk_1_one);
+
+  invMk_1_invBeta  = pop<RCP<Matrix> >(List, "invMk_1_invBeta", invMk_1_invBeta);
+  invMk_2_invAlpha = pop<RCP<Matrix> >(List, "invMk_2_invAlpha", invMk_2_invAlpha);
+
+  Nullspace11 = pop<RCP<MultiVector> >(List, "Nullspace11", Nullspace11);
+  Nullspace22 = pop<RCP<MultiVector> >(List, "Nullspace22", Nullspace22);
+  NodalCoords = pop<RCP<RealValuedMultiVector> >(List, "Coordinates", NodalCoords);
+
+  // old parameter names
+  if (List.isType<RCP<Matrix> >("Ms")) {
+    if (M1_beta.is_null())
+      M1_beta = pop<RCP<Matrix> >(List, "Ms");
+    else
+      TEUCHOS_ASSERT(false);
+  }
+  if (List.isType<RCP<Matrix> >("M1")) {
+    if (Mk_one.is_null())
+      Mk_one = pop<RCP<Matrix> >(List, "M1");
+    else
+      TEUCHOS_ASSERT(false);
+  }
+  if (List.isType<RCP<Matrix> >("M0inv")) {
+    if (invMk_1_invBeta.is_null())
+      invMk_1_invBeta = pop<RCP<Matrix> >(List, "M0inv");
+    else
+      TEUCHOS_ASSERT(false);
+  }
+  if (List.isType<RCP<MultiVector> >("Nullspace")) {
+    if (Nullspace11.is_null())
+      Nullspace11 = pop<RCP<MultiVector> >(List, "Nullspace");
+    else
+      TEUCHOS_ASSERT(false);
+  }
 
-      Dk_1 = pop(List, "Dk_1", Dk_1);
-      Dk_2 = pop<RCP<Matrix> >(List, "Dk_2", Dk_2);
-      D0 = pop<RCP<Matrix> >(List, "D0", D0);
-
-      M1_beta = pop<RCP<Matrix> >(List, "M1_beta", M1_beta);
-      M1_alpha = pop<RCP<Matrix> >(List, "M1_alpha", M1_alpha);
-
-      Mk_one = pop<RCP<Matrix> >(List, "Mk_one", Mk_one);
-      Mk_1_one = pop<RCP<Matrix> >(List, "Mk_1_one", Mk_1_one);
-
-      invMk_1_invBeta = pop<RCP<Matrix> >(List, "invMk_1_invBeta", invMk_1_invBeta);
-      invMk_2_invAlpha = pop<RCP<Matrix> >(List, "invMk_2_invAlpha", invMk_2_invAlpha);
-
-      Nullspace11 = pop<RCP<MultiVector> >(List, "Nullspace11", Nullspace11);
-      Nullspace22 = pop<RCP<MultiVector> >(List, "Nullspace22", Nullspace22);
-      NodalCoords = pop<RCP<RealValuedMultiVector> >(List, "Coordinates", NodalCoords);
-
-      // old parameter names
-      if (List.isType<RCP<Matrix> >("Ms")) {
-        if (M1_beta.is_null())
-          M1_beta = pop<RCP<Matrix> >(List, "Ms");
-        else
-          TEUCHOS_ASSERT(false);
-      }
-      if (List.isType<RCP<Matrix> >("M1")) {
-        if (Mk_one.is_null())
-          Mk_one = pop<RCP<Matrix> >(List, "M1");
-        else
-          TEUCHOS_ASSERT(false);
-      }
-      if (List.isType<RCP<Matrix> >("M0inv")) {
-        if (invMk_1_invBeta.is_null())
-          invMk_1_invBeta = pop<RCP<Matrix> >(List, "M0inv");
-        else
-          TEUCHOS_ASSERT(false);
-      }
-      if (List.isType<RCP<MultiVector> >("Nullspace")) {
-        if (Nullspace11.is_null())
-          Nullspace11 = pop<RCP<MultiVector> >(List, "Nullspace");
-        else
-          TEUCHOS_ASSERT(false);
-      }
+  if (spaceNumber == 1) {
+    if (Dk_1.is_null())
+      Dk_1 = D0;
+    else if (D0.is_null())
+      D0 = Dk_1;
+    if (M1_beta.is_null())
+      M1_beta = Mk_one;
+  } else if (spaceNumber == 2) {
+    if (Dk_2.is_null())
+      Dk_2 = D0;
+    else if (D0.is_null())
+      D0 = Dk_2;
+  }
 
-      if (spaceNumber == 1) {
-        if (Dk_1.is_null())
-          Dk_1 = D0;
-        else if (D0.is_null())
-          D0 = Dk_1;
-        if (M1_beta.is_null())
-          M1_beta = Mk_one;
-      } else if (spaceNumber == 2) {
-        if (Dk_2.is_null())
-          Dk_2 = D0;
-        else if (D0.is_null())
-          D0 = Dk_2;
-      }
+  initialize(spaceNumber,
+             Dk_1, Dk_2, D0,
+             M1_beta, M1_alpha,
+             Mk_one, Mk_1_one,
+             invMk_1_invBeta, invMk_2_invAlpha,
+             Nullspace11, Nullspace22,
+             NodalCoords,
+             List);
+
+  if (SM_Matrix != Teuchos::null)
+    resetMatrix(SM_Matrix, ComputePrec);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initialize(const Teuchos::RCP<Matrix> &D0_Matrix,
+               const Teuchos::RCP<Matrix> &Ms_Matrix,
+               const Teuchos::RCP<Matrix> &M0inv_Matrix,
+               const Teuchos::RCP<Matrix> &M1_Matrix,
+               const Teuchos::RCP<MultiVector> &Nullspace11,
+               const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+               Teuchos::ParameterList &List) {
+  initialize(1,
+             D0_Matrix, Teuchos::null, D0_Matrix,
+             Ms_Matrix, Teuchos::null,
+             M1_Matrix, Teuchos::null,
+             M0inv_Matrix, Teuchos::null,
+             Nullspace11, Teuchos::null,
+             NodalCoords,
+             List);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    initialize(const int k,
+               const Teuchos::RCP<Matrix> &Dk_1,
+               const Teuchos::RCP<Matrix> &Dk_2,
+               const Teuchos::RCP<Matrix> &D0,
+               const Teuchos::RCP<Matrix> &M1_beta,
+               const Teuchos::RCP<Matrix> &M1_alpha,
+               const Teuchos::RCP<Matrix> &Mk_one,
+               const Teuchos::RCP<Matrix> &Mk_1_one,
+               const Teuchos::RCP<Matrix> &invMk_1_invBeta,
+               const Teuchos::RCP<Matrix> &invMk_2_invAlpha,
+               const Teuchos::RCP<MultiVector> &Nullspace11,
+               const Teuchos::RCP<MultiVector> &Nullspace22,
+               const Teuchos::RCP<RealValuedMultiVector> &NodalCoords,
+               Teuchos::ParameterList &List) {
+  spaceNumber_ = k;
+  if (spaceNumber_ == 1)
+    solverName_ = "RefMaxwell";
+  else if (spaceNumber_ == 2)
+    solverName_ = "RefDarcy";
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument,
+                               "spaceNumber needs to be 1 (HCurl) or 2 (HDiv)");
+  HierarchyCoarse11_ = Teuchos::null;
+  Hierarchy22_       = Teuchos::null;
+  PreSmoother11_     = Teuchos::null;
+  PostSmoother11_    = Teuchos::null;
+  disable_addon_     = false;
+  disable_addon_22_  = true;
+  mode_              = "additive";
+
+  // set parameters
+  setParameters(List);
+
+  // some pre-conditions
+  TEUCHOS_ASSERT((k == 1) || (k == 2));
+  // Need Dk_1
+  TEUCHOS_ASSERT(Dk_1 != Teuchos::null);
+  // Need D0 for aggregation
+  TEUCHOS_ASSERT(D0 != Teuchos::null);
+
+  // Need M1_beta for aggregation
+  TEUCHOS_ASSERT(M1_beta != Teuchos::null);
+  // Need M1_alpha for aggregation if k>=1
+  if (k >= 2)
+    TEUCHOS_ASSERT(M1_alpha != Teuchos::null);
+
+  if (!disable_addon_) {
+    // Need Mk_one and invMk_1_invBeta for addon11
+    TEUCHOS_ASSERT(Mk_one != Teuchos::null);
+    TEUCHOS_ASSERT(invMk_1_invBeta != Teuchos::null);
+  }
 
-      initialize(spaceNumber,
-                 Dk_1, Dk_2, D0,
-                 M1_beta, M1_alpha,
-                 Mk_one, Mk_1_one,
-                 invMk_1_invBeta, invMk_2_invAlpha,
-                 Nullspace11, Nullspace22,
-                 NodalCoords,
-                 List);
-
-      if (SM_Matrix != Teuchos::null)
-        resetMatrix(SM_Matrix,ComputePrec);
-    }
-
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initialize(const Teuchos::RCP<Matrix> & D0_Matrix,
-             const Teuchos::RCP<Matrix> & Ms_Matrix,
-             const Teuchos::RCP<Matrix> & M0inv_Matrix,
-             const Teuchos::RCP<Matrix> & M1_Matrix,
-             const Teuchos::RCP<MultiVector>  & Nullspace11,
-             const Teuchos::RCP<RealValuedMultiVector>  & NodalCoords,
-             Teuchos::ParameterList& List)
-  {
-    initialize(1,
-               D0_Matrix, Teuchos::null, D0_Matrix,
-               Ms_Matrix, Teuchos::null,
-               M1_Matrix, Teuchos::null,
-               M0inv_Matrix, Teuchos::null,
-               Nullspace11, Teuchos::null,
-               NodalCoords,
-               List);
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  initialize(const int k,
-             const Teuchos::RCP<Matrix> & Dk_1,
-             const Teuchos::RCP<Matrix> & Dk_2,
-             const Teuchos::RCP<Matrix> & D0,
-             const Teuchos::RCP<Matrix> & M1_beta,
-             const Teuchos::RCP<Matrix> & M1_alpha,
-             const Teuchos::RCP<Matrix> & Mk_one,
-             const Teuchos::RCP<Matrix> & Mk_1_one,
-             const Teuchos::RCP<Matrix> & invMk_1_invBeta,
-             const Teuchos::RCP<Matrix> & invMk_2_invAlpha,
-             const Teuchos::RCP<MultiVector> & Nullspace11,
-             const Teuchos::RCP<MultiVector> & Nullspace22,
-             const Teuchos::RCP<RealValuedMultiVector> & NodalCoords,
-             Teuchos::ParameterList& List) {
-
-    spaceNumber_ = k;
-    if (spaceNumber_ == 1)
-      solverName_ = "RefMaxwell";
-    else if (spaceNumber_ == 2)
-      solverName_ = "RefDarcy";
-    else
-      TEUCHOS_TEST_FOR_EXCEPTION(true,std::invalid_argument,
-                                 "spaceNumber needs to be 1 (HCurl) or 2 (HDiv)");
-    HierarchyCoarse11_    = Teuchos::null;
-    Hierarchy22_          = Teuchos::null;
-    PreSmoother11_        = Teuchos::null;
-    PostSmoother11_       = Teuchos::null;
-    disable_addon_        = false;
-    disable_addon_22_     = true;
-    mode_                 = "additive";
-
-    // set parameters
-    setParameters(List);
-
-    // some pre-conditions
-    TEUCHOS_ASSERT((k == 1) || (k == 2));
-    // Need Dk_1
-    TEUCHOS_ASSERT(Dk_1!=Teuchos::null);
-    // Need D0 for aggregation
-    TEUCHOS_ASSERT(D0!=Teuchos::null);
-
-    // Need M1_beta for aggregation
-    TEUCHOS_ASSERT(M1_beta!=Teuchos::null);
-    // Need M1_alpha for aggregation if k>=1
-    if (k >= 2)
-      TEUCHOS_ASSERT(M1_alpha!=Teuchos::null);
-
-    if (!disable_addon_) {
-      // Need Mk_one and invMk_1_invBeta for addon11
-      TEUCHOS_ASSERT(Mk_one!=Teuchos::null);
-      TEUCHOS_ASSERT(invMk_1_invBeta!=Teuchos::null);
-    }
-
-    if ((k >= 2) && !disable_addon_22_) {
-      // Need Dk_2, Mk_1_one and invMk_2_invAlpha for addon22
-      TEUCHOS_ASSERT(Dk_2!=Teuchos::null);
-      TEUCHOS_ASSERT(Mk_1_one!=Teuchos::null);
-      TEUCHOS_ASSERT(invMk_2_invAlpha!=Teuchos::null);
-    }
+  if ((k >= 2) && !disable_addon_22_) {
+    // Need Dk_2, Mk_1_one and invMk_2_invAlpha for addon22
+    TEUCHOS_ASSERT(Dk_2 != Teuchos::null);
+    TEUCHOS_ASSERT(Mk_1_one != Teuchos::null);
+    TEUCHOS_ASSERT(invMk_2_invAlpha != Teuchos::null);
+  }
 
 #ifdef HAVE_MUELU_DEBUG
 
-    TEUCHOS_ASSERT(D0->getRangeMap()->isSameAs(*D0->getRowMap()));
+  TEUCHOS_ASSERT(D0->getRangeMap()->isSameAs(*D0->getRowMap()));
 
-    // M1_beta is square
-    TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*M1_beta->getRangeMap()));
-    TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*M1_beta->getRowMap()));
+  // M1_beta is square
+  TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*M1_beta->getRangeMap()));
+  TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*M1_beta->getRowMap()));
 
-    // M1_beta is consistent with D0
-    TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*D0->getRangeMap()));
+  // M1_beta is consistent with D0
+  TEUCHOS_ASSERT(M1_beta->getDomainMap()->isSameAs(*D0->getRangeMap()));
 
-    if (k >= 2) {
-      // M1_alpha is square
-      TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*M1_alpha->getRangeMap()));
-      TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*M1_alpha->getRowMap()));
+  if (k >= 2) {
+    // M1_alpha is square
+    TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*M1_alpha->getRangeMap()));
+    TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*M1_alpha->getRowMap()));
 
-      // M1_alpha is consistent with D0
-      TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*D0->getRangeMap()))
-    }
+    // M1_alpha is consistent with D0
+    TEUCHOS_ASSERT(M1_alpha->getDomainMap()->isSameAs(*D0->getRangeMap()))
+  }
 
-    if (!disable_addon_) {
-      // Mk_one is square
-      TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Mk_one->getRangeMap()));
-      TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Mk_one->getRowMap()));
+  if (!disable_addon_) {
+    // Mk_one is square
+    TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Mk_one->getRangeMap()));
+    TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Mk_one->getRowMap()));
 
-      // Mk_one is consistent with Dk_1
-      TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Dk_1->getRangeMap()));
+    // Mk_one is consistent with Dk_1
+    TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Dk_1->getRangeMap()));
 
-      // invMk_1_invBeta is square
-      TEUCHOS_ASSERT(invMk_1_invBeta->getDomainMap()->isSameAs(*invMk_1_invBeta->getRangeMap()));
-      TEUCHOS_ASSERT(invMk_1_invBeta->getDomainMap()->isSameAs(*invMk_1_invBeta->getRowMap()));
+    // invMk_1_invBeta is square
+    TEUCHOS_ASSERT(invMk_1_invBeta->getDomainMap()->isSameAs(*invMk_1_invBeta->getRangeMap()));
+    TEUCHOS_ASSERT(invMk_1_invBeta->getDomainMap()->isSameAs(*invMk_1_invBeta->getRowMap()));
 
-      // invMk_1_invBeta is consistent with Dk_1
-      TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Dk_1->getRangeMap()));
-    }
+    // invMk_1_invBeta is consistent with Dk_1
+    TEUCHOS_ASSERT(Mk_one->getDomainMap()->isSameAs(*Dk_1->getRangeMap()));
+  }
 
-    if ((k >= 2) && !disable_addon_22_) {
-      // Mk_1_one is square
-      TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Mk_1_one->getRangeMap()));
-      TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Mk_1_one->getRowMap()));
+  if ((k >= 2) && !disable_addon_22_) {
+    // Mk_1_one is square
+    TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Mk_1_one->getRangeMap()));
+    TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Mk_1_one->getRowMap()));
 
-      // Mk_1_one is consistent with Dk_1
-      TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Dk_1->getDomainMap()));
+    // Mk_1_one is consistent with Dk_1
+    TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Dk_1->getDomainMap()));
 
-      // Mk_1_one is consistent with Dk_2
-      TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Dk_2->getRangeMap()));
+    // Mk_1_one is consistent with Dk_2
+    TEUCHOS_ASSERT(Mk_1_one->getDomainMap()->isSameAs(*Dk_2->getRangeMap()));
 
-      // invMk_2_invAlpha is square
-      TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*invMk_2_invAlpha->getRangeMap()));
-      TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*invMk_2_invAlpha->getRowMap()));
+    // invMk_2_invAlpha is square
+    TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*invMk_2_invAlpha->getRangeMap()));
+    TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*invMk_2_invAlpha->getRowMap()));
 
-      // invMk_2_invAlpha is consistent with Dk_2
-      TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*Dk_2->getDomainMap()));
-    }
+    // invMk_2_invAlpha is consistent with Dk_2
+    TEUCHOS_ASSERT(invMk_2_invAlpha->getDomainMap()->isSameAs(*Dk_2->getDomainMap()));
+  }
 #endif
 
-    D0_ = D0;
-    if (Dk_1->getRowMap()->lib() == Xpetra::UseTpetra) {
-      // We will remove boundary conditions from Dk_1, and potentially change maps, so we copy the input.
-      // Fortunately, Dk_1 is quite sparse.
-      // We cannot use the Tpetra copy constructor, since it does not copy the graph.
-
-      RCP<Matrix> Dk_1copy = MatrixFactory::Build(Dk_1->getRowMap(), Dk_1->getColMap(), 0);
-      RCP<CrsMatrix> Dk_1copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1copy,true)->getCrsMatrix();
-      ArrayRCP<const size_t> Dk_1rowptr_RCP;
-      ArrayRCP<const LO>     Dk_1colind_RCP;
-      ArrayRCP<const SC>     Dk_1vals_RCP;
-      rcp_dynamic_cast<CrsMatrixWrap>(Dk_1,true)->getCrsMatrix()->getAllValues(Dk_1rowptr_RCP,
-                                                                               Dk_1colind_RCP,
-                                                                               Dk_1vals_RCP);
-
-      ArrayRCP<size_t> Dk_1copyrowptr_RCP;
-      ArrayRCP<LO>     Dk_1copycolind_RCP;
-      ArrayRCP<SC>     Dk_1copyvals_RCP;
-      Dk_1copyCrs->allocateAllValues(Dk_1vals_RCP.size(),Dk_1copyrowptr_RCP,Dk_1copycolind_RCP,Dk_1copyvals_RCP);
-      Dk_1copyrowptr_RCP.deepCopy(Dk_1rowptr_RCP());
-      Dk_1copycolind_RCP.deepCopy(Dk_1colind_RCP());
-      Dk_1copyvals_RCP.deepCopy(Dk_1vals_RCP());
-      Dk_1copyCrs->setAllValues(Dk_1copyrowptr_RCP,
+  D0_ = D0;
+  if (Dk_1->getRowMap()->lib() == Xpetra::UseTpetra) {
+    // We will remove boundary conditions from Dk_1, and potentially change maps, so we copy the input.
+    // Fortunately, Dk_1 is quite sparse.
+    // We cannot use the Tpetra copy constructor, since it does not copy the graph.
+
+    RCP<Matrix> Dk_1copy       = MatrixFactory::Build(Dk_1->getRowMap(), Dk_1->getColMap(), 0);
+    RCP<CrsMatrix> Dk_1copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(Dk_1copy, true)->getCrsMatrix();
+    ArrayRCP<const size_t> Dk_1rowptr_RCP;
+    ArrayRCP<const LO> Dk_1colind_RCP;
+    ArrayRCP<const SC> Dk_1vals_RCP;
+    rcp_dynamic_cast<CrsMatrixWrap>(Dk_1, true)->getCrsMatrix()->getAllValues(Dk_1rowptr_RCP, Dk_1colind_RCP, Dk_1vals_RCP);
+
+    ArrayRCP<size_t> Dk_1copyrowptr_RCP;
+    ArrayRCP<LO> Dk_1copycolind_RCP;
+    ArrayRCP<SC> Dk_1copyvals_RCP;
+    Dk_1copyCrs->allocateAllValues(Dk_1vals_RCP.size(), Dk_1copyrowptr_RCP, Dk_1copycolind_RCP, Dk_1copyvals_RCP);
+    Dk_1copyrowptr_RCP.deepCopy(Dk_1rowptr_RCP());
+    Dk_1copycolind_RCP.deepCopy(Dk_1colind_RCP());
+    Dk_1copyvals_RCP.deepCopy(Dk_1vals_RCP());
+    Dk_1copyCrs->setAllValues(Dk_1copyrowptr_RCP,
                               Dk_1copycolind_RCP,
                               Dk_1copyvals_RCP);
-      Dk_1copyCrs->expertStaticFillComplete(Dk_1->getDomainMap(), Dk_1->getRangeMap(),
-                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_1,true)->getCrsMatrix()->getCrsGraph()->getImporter(),
-                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_1,true)->getCrsMatrix()->getCrsGraph()->getExporter());
-      Dk_1_ = Dk_1copy;
-    } else
-      Dk_1_ = MatrixFactory::BuildCopy(Dk_1);
-
-    if ((!Dk_2.is_null()) && (Dk_2->getRowMap()->lib() == Xpetra::UseTpetra)) {
-      // We will remove boundary conditions from Dk_2, and potentially change maps, so we copy the input.
-      // Fortunately, Dk_2 is quite sparse.
-      // We cannot use the Tpetra copy constructor, since it does not copy the graph.
-
-      RCP<Matrix> Dk_2copy = MatrixFactory::Build(Dk_2->getRowMap(), Dk_2->getColMap(), 0);
-      RCP<CrsMatrix> Dk_2copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(Dk_2copy,true)->getCrsMatrix();
-      ArrayRCP<const size_t> Dk_2rowptr_RCP;
-      ArrayRCP<const LO>     Dk_2colind_RCP;
-      ArrayRCP<const SC>     Dk_2vals_RCP;
-      rcp_dynamic_cast<CrsMatrixWrap>(Dk_2,true)->getCrsMatrix()->getAllValues(Dk_2rowptr_RCP,
-                                                                               Dk_2colind_RCP,
-                                                                               Dk_2vals_RCP);
-
-      ArrayRCP<size_t> Dk_2copyrowptr_RCP;
-      ArrayRCP<LO>     Dk_2copycolind_RCP;
-      ArrayRCP<SC>     Dk_2copyvals_RCP;
-      Dk_2copyCrs->allocateAllValues(Dk_2vals_RCP.size(),Dk_2copyrowptr_RCP,Dk_2copycolind_RCP,Dk_2copyvals_RCP);
-      Dk_2copyrowptr_RCP.deepCopy(Dk_2rowptr_RCP());
-      Dk_2copycolind_RCP.deepCopy(Dk_2colind_RCP());
-      Dk_2copyvals_RCP.deepCopy(Dk_2vals_RCP());
-      Dk_2copyCrs->setAllValues(Dk_2copyrowptr_RCP,
-                                Dk_2copycolind_RCP,
-                                Dk_2copyvals_RCP);
-      Dk_2copyCrs->expertStaticFillComplete(Dk_2->getDomainMap(), Dk_2->getRangeMap(),
-                                            rcp_dynamic_cast<CrsMatrixWrap>(Dk_2,true)->getCrsMatrix()->getCrsGraph()->getImporter(),
-                                            rcp_dynamic_cast<CrsMatrixWrap>(Dk_2,true)->getCrsMatrix()->getCrsGraph()->getExporter());
-      Dk_2_ = Dk_2copy;
-    } else if (!Dk_2.is_null())
-      Dk_2_ = MatrixFactory::BuildCopy(Dk_2);
-
-    M1_beta_  = M1_beta;
-    M1_alpha_ = M1_alpha;
-
-    Mk_one_   = Mk_one;
-    Mk_1_one_ = Mk_1_one;
-
-    invMk_1_invBeta_  = invMk_1_invBeta;
-    invMk_2_invAlpha_ = invMk_2_invAlpha;
-
-    NodalCoords_ = NodalCoords;
-    Nullspace11_ = Nullspace11;
-    Nullspace22_ = Nullspace22;
-
-    dump(D0_,               "D0.m");
-    dump(Dk_1_,             "Dk_1_clean.m");
-    dump(Dk_2_,             "Dk_2_clean.m");
-
-    dump(M1_beta_,          "M1_beta.m");
-    dump(M1_alpha_,         "M1_alpha.m");
-
-    dump(Mk_one_,           "Mk_one.m");
-    dump(Mk_1_one_,         "Mk_1_one.m");
-
-    dump(invMk_1_invBeta_,  "invMk_1_invBeta.m");
-    dump(invMk_2_invAlpha_, "invMk_2_invAlpha.m");
-
-    dumpCoords(NodalCoords_, "coords.m");
-
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  describe(Teuchos::FancyOStream& out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
-
-    std::ostringstream oss;
-
-    RCP<const Teuchos::Comm<int> > comm = SM_Matrix_->getDomainMap()->getComm();
+    Dk_1copyCrs->expertStaticFillComplete(Dk_1->getDomainMap(), Dk_1->getRangeMap(),
+                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_1, true)->getCrsMatrix()->getCrsGraph()->getImporter(),
+                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_1, true)->getCrsMatrix()->getCrsGraph()->getExporter());
+    Dk_1_ = Dk_1copy;
+  } else
+    Dk_1_ = MatrixFactory::BuildCopy(Dk_1);
+
+  if ((!Dk_2.is_null()) && (Dk_2->getRowMap()->lib() == Xpetra::UseTpetra)) {
+    // We will remove boundary conditions from Dk_2, and potentially change maps, so we copy the input.
+    // Fortunately, Dk_2 is quite sparse.
+    // We cannot use the Tpetra copy constructor, since it does not copy the graph.
+
+    RCP<Matrix> Dk_2copy       = MatrixFactory::Build(Dk_2->getRowMap(), Dk_2->getColMap(), 0);
+    RCP<CrsMatrix> Dk_2copyCrs = rcp_dynamic_cast<CrsMatrixWrap>(Dk_2copy, true)->getCrsMatrix();
+    ArrayRCP<const size_t> Dk_2rowptr_RCP;
+    ArrayRCP<const LO> Dk_2colind_RCP;
+    ArrayRCP<const SC> Dk_2vals_RCP;
+    rcp_dynamic_cast<CrsMatrixWrap>(Dk_2, true)->getCrsMatrix()->getAllValues(Dk_2rowptr_RCP, Dk_2colind_RCP, Dk_2vals_RCP);
+
+    ArrayRCP<size_t> Dk_2copyrowptr_RCP;
+    ArrayRCP<LO> Dk_2copycolind_RCP;
+    ArrayRCP<SC> Dk_2copyvals_RCP;
+    Dk_2copyCrs->allocateAllValues(Dk_2vals_RCP.size(), Dk_2copyrowptr_RCP, Dk_2copycolind_RCP, Dk_2copyvals_RCP);
+    Dk_2copyrowptr_RCP.deepCopy(Dk_2rowptr_RCP());
+    Dk_2copycolind_RCP.deepCopy(Dk_2colind_RCP());
+    Dk_2copyvals_RCP.deepCopy(Dk_2vals_RCP());
+    Dk_2copyCrs->setAllValues(Dk_2copyrowptr_RCP,
+                              Dk_2copycolind_RCP,
+                              Dk_2copyvals_RCP);
+    Dk_2copyCrs->expertStaticFillComplete(Dk_2->getDomainMap(), Dk_2->getRangeMap(),
+                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_2, true)->getCrsMatrix()->getCrsGraph()->getImporter(),
+                                          rcp_dynamic_cast<CrsMatrixWrap>(Dk_2, true)->getCrsMatrix()->getCrsGraph()->getExporter());
+    Dk_2_ = Dk_2copy;
+  } else if (!Dk_2.is_null())
+    Dk_2_ = MatrixFactory::BuildCopy(Dk_2);
+
+  M1_beta_  = M1_beta;
+  M1_alpha_ = M1_alpha;
+
+  Mk_one_   = Mk_one;
+  Mk_1_one_ = Mk_1_one;
+
+  invMk_1_invBeta_  = invMk_1_invBeta;
+  invMk_2_invAlpha_ = invMk_2_invAlpha;
+
+  NodalCoords_ = NodalCoords;
+  Nullspace11_ = Nullspace11;
+  Nullspace22_ = Nullspace22;
+
+  dump(D0_, "D0.m");
+  dump(Dk_1_, "Dk_1_clean.m");
+  dump(Dk_2_, "Dk_2_clean.m");
+
+  dump(M1_beta_, "M1_beta.m");
+  dump(M1_alpha_, "M1_alpha.m");
+
+  dump(Mk_one_, "Mk_one.m");
+  dump(Mk_1_one_, "Mk_1_one.m");
+
+  dump(invMk_1_invBeta_, "invMk_1_invBeta.m");
+  dump(invMk_2_invAlpha_, "invMk_2_invAlpha.m");
+
+  dumpCoords(NodalCoords_, "coords.m");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel /* verbLevel */) const {
+  std::ostringstream oss;
+
+  RCP<const Teuchos::Comm<int> > comm = SM_Matrix_->getDomainMap()->getComm();
 
 #ifdef HAVE_MPI
-    int root;
-    if (!coarseA11_.is_null())
-      root = comm->getRank();
-    else
-      root = -1;
-
-    int actualRoot;
-    reduceAll(*comm, Teuchos::REDUCE_MAX, root, Teuchos::ptr(&actualRoot));
-    root = actualRoot;
+  int root;
+  if (!coarseA11_.is_null())
+    root = comm->getRank();
+  else
+    root = -1;
+
+  int actualRoot;
+  reduceAll(*comm, Teuchos::REDUCE_MAX, root, Teuchos::ptr(&actualRoot));
+  root = actualRoot;
 #endif
 
+  oss << "\n--------------------------------------------------------------------------------\n"
+      << "---                            " + solverName_ +
+             " Summary                            ---\n"
+             "--------------------------------------------------------------------------------"
+      << std::endl;
+  oss << std::endl;
 
-    oss << "\n--------------------------------------------------------------------------------\n" <<
-      "---                            "+solverName_+" Summary                            ---\n"
-      "--------------------------------------------------------------------------------" << std::endl;
-    oss << std::endl;
-
-    GlobalOrdinal numRows;
-    GlobalOrdinal nnz;
+  GlobalOrdinal numRows;
+  GlobalOrdinal nnz;
 
-    SM_Matrix_->getRowMap()->getComm()->barrier();
+  SM_Matrix_->getRowMap()->getComm()->barrier();
 
-    numRows = SM_Matrix_->getGlobalNumRows();
-    nnz = SM_Matrix_->getGlobalNumEntries();
+  numRows = SM_Matrix_->getGlobalNumRows();
+  nnz     = SM_Matrix_->getGlobalNumEntries();
 
-    Xpetra::global_size_t tt = numRows;
-    int rowspacer = 3; while (tt != 0) { tt /= 10; rowspacer++; }
-    tt = nnz;
-    int nnzspacer = 2; while (tt != 0) { tt /= 10; nnzspacer++; }
+  Xpetra::global_size_t tt = numRows;
+  int rowspacer            = 3;
+  while (tt != 0) {
+    tt /= 10;
+    rowspacer++;
+  }
+  tt            = nnz;
+  int nnzspacer = 2;
+  while (tt != 0) {
+    tt /= 10;
+    nnzspacer++;
+  }
 
-    oss  << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
-    oss << "(1, 1)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  oss << "block " << std::setw(rowspacer) << " rows " << std::setw(nnzspacer) << " nnz " << std::setw(9) << " nnz/row" << std::endl;
+  oss << "(1, 1)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
 
-    if (!A22_.is_null()) {
-      numRows = A22_->getGlobalNumRows();
-      nnz = A22_->getGlobalNumEntries();
+  if (!A22_.is_null()) {
+    numRows = A22_->getGlobalNumRows();
+    nnz     = A22_->getGlobalNumEntries();
 
-      oss << "(2, 2)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
-    }
+    oss << "(2, 2)" << std::setw(rowspacer) << numRows << std::setw(nnzspacer) << nnz << std::setw(9) << as<double>(nnz) / numRows << std::endl;
+  }
 
-    oss << std::endl;
+  oss << std::endl;
 
-    {
-      if (PreSmoother11_ != null && PreSmoother11_ == PostSmoother11_)
-        oss << "Smoother 11 both : " << PreSmoother11_->description() << std::endl;
-      else {
-        oss << "Smoother 11 pre  : "
-            << (PreSmoother11_ != null ?  PreSmoother11_->description() : "no smoother") << std::endl;
-        oss << "Smoother 11 post : "
-            << (PostSmoother11_ != null ?  PostSmoother11_->description() : "no smoother") << std::endl;
-      }
+  {
+    if (PreSmoother11_ != null && PreSmoother11_ == PostSmoother11_)
+      oss << "Smoother 11 both : " << PreSmoother11_->description() << std::endl;
+    else {
+      oss << "Smoother 11 pre  : "
+          << (PreSmoother11_ != null ? PreSmoother11_->description() : "no smoother") << std::endl;
+      oss << "Smoother 11 post : "
+          << (PostSmoother11_ != null ? PostSmoother11_->description() : "no smoother") << std::endl;
     }
-    oss << std::endl;
+  }
+  oss << std::endl;
 
-    std::string outstr = oss.str();
+  std::string outstr = oss.str();
 
 #ifdef HAVE_MPI
-    RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-    MPI_Comm rawComm = (*mpiComm->getRawMpiComm())();
-
-    int strLength = outstr.size();
-    MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
-    if (comm->getRank() != root)
-      outstr.resize(strLength);
-    MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+  RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+  MPI_Comm rawComm                          = (*mpiComm->getRawMpiComm())();
+
+  int strLength = outstr.size();
+  MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
+  if (comm->getRank() != root)
+    outstr.resize(strLength);
+  MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
 #endif
 
-    out << outstr;
+  out << outstr;
 
-    if (!HierarchyCoarse11_.is_null())
-      HierarchyCoarse11_->describe(out, GetVerbLevel());
+  if (!HierarchyCoarse11_.is_null())
+    HierarchyCoarse11_->describe(out, GetVerbLevel());
 
-    if (!Hierarchy22_.is_null())
-      Hierarchy22_->describe(out, GetVerbLevel());
+  if (!Hierarchy22_.is_null())
+    Hierarchy22_->describe(out, GetVerbLevel());
 
-    if (IsPrint(Statistics2)) {
-      // Print the grid of processors
-      std::ostringstream oss2;
+  if (IsPrint(Statistics2)) {
+    // Print the grid of processors
+    std::ostringstream oss2;
 
-      oss2 << "Sub-solver distribution over ranks" << std::endl;
-      oss2 << "( (1,1) block only is indicated by '1', (2,2) block only by '2', and both blocks by 'B' and none by '.')" << std::endl;
+    oss2 << "Sub-solver distribution over ranks" << std::endl;
+    oss2 << "( (1,1) block only is indicated by '1', (2,2) block only by '2', and both blocks by 'B' and none by '.')" << std::endl;
 
-      int numProcs = comm->getSize();
+    int numProcs = comm->getSize();
 #ifdef HAVE_MPI
-      RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-      TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
-      RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
+    RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+    TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
+    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
 #endif
 
-      char status = 0;
-      if (!coarseA11_.is_null())
-        status += 1;
-      if (!A22_.is_null())
-        status += 2;
-      std::vector<char> states(numProcs, 0);
+    char status = 0;
+    if (!coarseA11_.is_null())
+      status += 1;
+    if (!A22_.is_null())
+      status += 2;
+    std::vector<char> states(numProcs, 0);
 #ifdef HAVE_MPI
-      MPI_Gather(&status, 1, MPI_CHAR, &states[0], 1, MPI_CHAR, 0, *rawMpiComm);
+    MPI_Gather(&status, 1, MPI_CHAR, &states[0], 1, MPI_CHAR, 0, *rawMpiComm);
 #else
-      states.push_back(status);
+    states.push_back(status);
 #endif
 
-      int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
-      for (int proc = 0; proc < numProcs; proc += rowWidth) {
-        for (int j = 0; j < rowWidth; j++)
-          if (proc + j < numProcs)
-            if (states[proc+j] == 0)
-              oss2 << ".";
-            else if (states[proc+j] == 1)
-              oss2 << "1";
-            else if (states[proc+j] == 2)
-              oss2 << "2";
-            else
-              oss2 << "B";
+    int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
+    for (int proc = 0; proc < numProcs; proc += rowWidth) {
+      for (int j = 0; j < rowWidth; j++)
+        if (proc + j < numProcs)
+          if (states[proc + j] == 0)
+            oss2 << ".";
+          else if (states[proc + j] == 1)
+            oss2 << "1";
+          else if (states[proc + j] == 2)
+            oss2 << "2";
           else
-            oss2 << " ";
+            oss2 << "B";
+        else
+          oss2 << " ";
 
-        oss2 << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
-      }
-      oss2 << std::endl;
-      GetOStream(Statistics2) << oss2.str();
+      oss2 << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
     }
-
-
+    oss2 << std::endl;
+    GetOStream(Statistics2) << oss2.str();
   }
+}
 
-
-} // namespace
+}  // namespace MueLu
 
 #define MUELU_REFMAXWELL_SHORT
-#endif //ifdef MUELU_REFMAXWELL_DEF_HPP
+#endif  // ifdef MUELU_REFMAXWELL_DEF_HPP
diff --git a/packages/muelu/src/Operators/MueLu_XpetraOperator_decl.hpp b/packages/muelu/src/Operators/MueLu_XpetraOperator_decl.hpp
index ee1f0b756aaf..3d0e2a9b2265 100644
--- a/packages/muelu/src/Operators/MueLu_XpetraOperator_decl.hpp
+++ b/packages/muelu/src/Operators/MueLu_XpetraOperator_decl.hpp
@@ -56,64 +56,65 @@
 namespace MueLu {
 
 /*!  @brief Wraps an existing MueLu::Hierarchy as a Xpetra::Operator.
-*/
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class XpetraOperator : public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
-  protected:
-    XpetraOperator() { }
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-
-    //! Constructor
-    XpetraOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H) : Hierarchy_(H) { }
-
-    //! Destructor.
-    virtual ~XpetraOperator() { }
-
-    //@}
-
-    //! Returns the Tpetra::Map object associated with the domain of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getDomainMap() const;
-
-    //! Returns the Tpetra::Map object associated with the range of this operator.
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > getRangeMap() const;
-
-    //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
-    /*!
-      \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
-      \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
-    */
-    void apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-                                         Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-                                         Teuchos::ETransp mode = Teuchos::NO_TRANS,
-                                         Scalar /* alpha */ = Teuchos::ScalarTraits<Scalar>::one(),
-                                         Scalar /* beta */  = Teuchos::ScalarTraits<Scalar>::one()) const;
-
-    //! Indicates whether this operator supports applying the adjoint operator.
-    bool hasTransposeApply() const { return false; }
-
-    //! Compute a residual R = B - (*this) * X
-    void residual(const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & X,
-                  const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & B,
-                  Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & R) const;
-    
-    //! @name MueLu specific
-    //@{
-
-    //! Direct access to the underlying MueLu::Hierarchy.
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetHierarchy() const { return Hierarchy_; }
-
-    //@}
-
-  private:
-    RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
-  };
-
-} // namespace
-
-#endif // MUELU_XPETRAOPERATOR_DECL_HPP
+ */
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class XpetraOperator : public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ protected:
+  XpetraOperator() {}
+
+ public:
+  //! @name Constructor/Destructor
+  //@{
+
+  //! Constructor
+  XpetraOperator(const RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& H)
+    : Hierarchy_(H) {}
+
+  //! Destructor.
+  virtual ~XpetraOperator() {}
+
+  //@}
+
+  //! Returns the Tpetra::Map object associated with the domain of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getDomainMap() const;
+
+  //! Returns the Tpetra::Map object associated with the range of this operator.
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > getRangeMap() const;
+
+  //! Returns in Y the result of a Xpetra::Operator applied to a Xpetra::MultiVector X.
+  /*!
+    \param[in]  X - Xpetra::MultiVector of dimension NumVectors to multiply with matrix.
+    \param[out] Y - Xpetra::MultiVector of dimension NumVectors containing result.
+  */
+  void apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+             Teuchos::ETransp mode = Teuchos::NO_TRANS,
+             Scalar /* alpha */    = Teuchos::ScalarTraits<Scalar>::one(),
+             Scalar /* beta */     = Teuchos::ScalarTraits<Scalar>::one()) const;
+
+  //! Indicates whether this operator supports applying the adjoint operator.
+  bool hasTransposeApply() const { return false; }
+
+  //! Compute a residual R = B - (*this) * X
+  void residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+                const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+                Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const;
+
+  //! @name MueLu specific
+  //@{
+
+  //! Direct access to the underlying MueLu::Hierarchy.
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GetHierarchy() const { return Hierarchy_; }
+
+  //@}
+
+ private:
+  RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Hierarchy_;
+};
+
+}  // namespace MueLu
+
+#endif  // MUELU_XPETRAOPERATOR_DECL_HPP
diff --git a/packages/muelu/src/Operators/MueLu_XpetraOperator_def.hpp b/packages/muelu/src/Operators/MueLu_XpetraOperator_def.hpp
index 7b0f3c5c448f..b9540afe0069 100644
--- a/packages/muelu/src/Operators/MueLu_XpetraOperator_def.hpp
+++ b/packages/muelu/src/Operators/MueLu_XpetraOperator_def.hpp
@@ -48,74 +48,69 @@
 
 #include "MueLu_XpetraOperator_decl.hpp"
 
-
-
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >
-  XpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  getDomainMap() const {
-    typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >
+XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getDomainMap() const {
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
 
-    RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
-    return A->getDomainMap();
-  }
+  RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
+  return A->getDomainMap();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >
-  XpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  getRangeMap() const {
-    typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >
+XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getRangeMap() const {
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
 
-    RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
-    return A->getRangeMap();
-  }
+  RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
+  return A->getRangeMap();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  XpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  apply(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& X,
-             Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Y,
-             Teuchos::ETransp mode,
-             Scalar /* alpha */,
-             Scalar /* beta */ ) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(mode!=Teuchos::NO_TRANS,std::logic_error,"MueLu::XpetraOperator does not support applying the adjoint operator");
-    try {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    apply(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+          Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Y,
+          Teuchos::ETransp mode,
+          Scalar /* alpha */,
+          Scalar /* beta */) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(mode != Teuchos::NO_TRANS, std::logic_error, "MueLu::XpetraOperator does not support applying the adjoint operator");
+  try {
 #ifdef HAVE_MUELU_DEBUG
-      typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
-      RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
+    typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+    RCP<Matrix> A = Hierarchy_->GetLevel(0)->template Get<RCP<Matrix> >("A");
 
-      // X is supposed to live in the range map of the operator (const rhs = B)
-      RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Xop =
-        Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A->getRangeMap(),X.getNumVectors());
-      RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Yop =
-        Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A->getDomainMap(),Y.getNumVectors());
-      TEUCHOS_TEST_FOR_EXCEPTION(A->getRangeMap()->isSameAs(*(Xop->getMap())) == false, std::logic_error,
-                                 "MueLu::XpetraOperator::apply: map of X is incompatible with range map of A");
-      TEUCHOS_TEST_FOR_EXCEPTION(A->getDomainMap()->isSameAs(*(Yop->getMap())) == false, std::logic_error,
-                                 "MueLu::XpetraOperator::apply: map of Y is incompatible with domain map of A");
+    // X is supposed to live in the range map of the operator (const rhs = B)
+    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Xop =
+        Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A->getRangeMap(), X.getNumVectors());
+    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Yop =
+        Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A->getDomainMap(), Y.getNumVectors());
+    TEUCHOS_TEST_FOR_EXCEPTION(A->getRangeMap()->isSameAs(*(Xop->getMap())) == false, std::logic_error,
+                               "MueLu::XpetraOperator::apply: map of X is incompatible with range map of A");
+    TEUCHOS_TEST_FOR_EXCEPTION(A->getDomainMap()->isSameAs(*(Yop->getMap())) == false, std::logic_error,
+                               "MueLu::XpetraOperator::apply: map of Y is incompatible with domain map of A");
 #endif
-      Hierarchy_->Iterate(X, Y, 1, true);
-    } catch (std::exception& e) {
-      //FIXME add message and rethrow
-      std::cerr << "Caught an exception in MueLu::XpetraOperator::apply():" << std::endl
-                << e.what() << std::endl;
-    }
-  }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  XpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  residual(const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & X,
-                const Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & B,
-                Xpetra::MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node > & R) const {
-    using STS = Teuchos::ScalarTraits<Scalar>;
-    R.update(STS::one(),B,STS::zero());
-    this->apply (X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+    Hierarchy_->Iterate(X, Y, 1, true);
+  } catch (std::exception& e) {
+    // FIXME add message and rethrow
+    std::cerr << "Caught an exception in MueLu::XpetraOperator::apply():" << std::endl
+              << e.what() << std::endl;
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void XpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    residual(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+             const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B,
+             Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& R) const {
+  using STS = Teuchos::ScalarTraits<Scalar>;
+  R.update(STS::one(), B, STS::zero());
+  this->apply(X, R, Teuchos::NO_TRANS, -STS::one(), STS::one());
+}
 
-} // namespace
+}  // namespace MueLu
 
-#endif // MUELU_XPETRAOPERATOR_DEF_HPP
+#endif  // MUELU_XPETRAOPERATOR_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_decl.hpp
index dfa2834b9607..a26bdecc31f6 100644
--- a/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_decl.hpp
@@ -46,7 +46,6 @@
 #ifndef PACKAGES_MUELU_SRC_REBALANCING_MUELU_CLONEREPARTITIONINTERFACE_DECL_HPP_
 #define PACKAGES_MUELU_SRC_REBALANCING_MUELU_CLONEREPARTITIONINTERFACE_DECL_HPP_
 
-
 #include <Xpetra_Map_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 #include <Xpetra_VectorFactory_fwd.hpp>
@@ -58,83 +57,75 @@
 
 namespace MueLu {
 
-  /*!
-    @class CloneRepartitionInterface
-    @brief Helper class which transforms an "Partition" array generated from a block in a blocked operator to a new "Partition" vector for another compatible block in the blocked operator.
-
-    This class is meant to be used with another {Zoltan,Zoltan2,Isorropia,...}Interface class which generates a "Partition" vector for a specific block to be rebalanced.
-    This class reuses the rebalancing and creates a compatible "Partition" array for another block.
+/*!
+  @class CloneRepartitionInterface
+  @brief Helper class which transforms an "Partition" array generated from a block in a blocked operator to a new "Partition" vector for another compatible block in the blocked operator.
 
-    @note This class is used for blocked operators only.
+  This class is meant to be used with another {Zoltan,Zoltan2,Isorropia,...}Interface class which generates a "Partition" vector for a specific block to be rebalanced.
+  This class reuses the rebalancing and creates a compatible "Partition" array for another block.
 
-    ## Input/output of CloneRepartitionInterface ##
+  @note This class is used for blocked operators only.
 
-    ### User parameters of CloneRepartitionInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process (must be of type SubBlockAFactory representing the corresponding diagonal block in the blocked operator) |
-    | number of partitions                   | GO      | - |  | * | * | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (not used in this factory)
-    | Partition | null |  | * | * | Factory generating the Partition array (e.g. ZoltanInterface) associated with the sub block in the blocked operator which partitioning information shall to be cloned
+  ## Input/output of CloneRepartitionInterface ##
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CloneRepartitionInterface::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see CloneRepartitionInterface::DeclareInput).
+  ### User parameters of CloneRepartitionInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process (must be of type SubBlockAFactory representing the corresponding diagonal block in the blocked operator) |
+  | number of partitions                   | GO      | - |  | * | * | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (not used in this factory)
+  | Partition | null |  | * | * | Factory generating the Partition array (e.g. ZoltanInterface) associated with the sub block in the blocked operator which partitioning information shall to be cloned
 
-    ### Variables provided by CloneRepartitionInterface ###
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CloneRepartitionInterface::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see CloneRepartitionInterface::DeclareInput).
 
-    After CloneRepartitionInterface::Build the following data is available (if requested)
+  ### Variables provided by CloneRepartitionInterface ###
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Partition       | CloneRepartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
+  After CloneRepartitionInterface::Build the following data is available (if requested)
 
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Partition       | CloneRepartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class CloneRepartitionInterface : public SingleLevelFactoryBase {
+*/
 
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CloneRepartitionInterface : public SingleLevelFactoryBase {
 #undef MUELU_CLONEREPARTITIONINTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
-
-    //! Constructor
-    CloneRepartitionInterface() { }
-
-    //! Destructor
-    virtual ~CloneRepartitionInterface() { }
-    //@}
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Constructor
+  CloneRepartitionInterface() {}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level & level) const;
-    //@}
+  //! Destructor
+  virtual ~CloneRepartitionInterface() {}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &level) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
+  //@}
 
-  private:
+ private:
+};  // class CloneRepartitionInterface
 
-
-
-  };  //class CloneRepartitionInterface
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CLONEREPARTITIONINTERFACE_SHORT
 
-
 #endif /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_CLONEREPARTITIONINTERFACE_DECL_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_def.hpp
index c82dfdc72979..c018d959a214 100644
--- a/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_CloneRepartitionInterface_def.hpp
@@ -55,106 +55,102 @@
 #include "MueLu_Exceptions.hpp"
 #include "MueLu_Monitor.hpp"
 
-
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- Teuchos::RCP<const ParameterList> CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    Teuchos::RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< Teuchos::RCP<const FactoryBase> >("A",         Teuchos::null, "Factory of the matrix A");
-    validParamList->set< Teuchos::RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< Teuchos::RCP<const FactoryBase> >("Partition", Teuchos::null, "Factory generating the Partition array (e.g. ZoltanInterface)");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const ParameterList> CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  Teuchos::RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<Teuchos::RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<Teuchos::RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<Teuchos::RCP<const FactoryBase> >("Partition", Teuchos::null, "Factory generating the Partition array (e.g. ZoltanInterface)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "Partition");
+}  // DeclareInput()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  currentLevel.print(GetOStream(Statistics0, 0));
+  // extract blocked operator A from current level
+  Teuchos::RCP<Matrix> A                       = Get<Teuchos::RCP<Matrix> >(currentLevel, "A");
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+
+  // number of Partitions only used for a shortcut.
+  if (currentLevel.IsAvailable("number of partitions")) {
+    GetOStream(Warnings0) << "Using user-provided \"number of partitions\", the performance is unknown" << std::endl;
   }
 
+  // ======================================================================================================
+  // Construct decomposition vector
+  // ======================================================================================================
+  RCP<GOVector> decomposition = Teuchos::null;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "Partition");
-  } //DeclareInput()
+  // extract decomposition vector
+  decomposition                    = Get<RCP<GOVector> >(currentLevel, "Partition");
+  ArrayRCP<const GO> decompEntries = decomposition->getData(0);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CloneRepartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    currentLevel.print(GetOStream(Statistics0,0));
-    // extract blocked operator A from current level
-    Teuchos::RCP<Matrix> A = Get< Teuchos::RCP<Matrix> >     (currentLevel, "A");
-    Teuchos::RCP<const Teuchos::Comm< int > > comm = A->getRowMap()->getComm();
-
-    // number of Partitions only used for a shortcut.
-    if (currentLevel.IsAvailable("number of partitions")) {
-      GetOStream(Warnings0) << "Using user-provided \"number of partitions\", the performance is unknown" << std::endl;
+  if (decomposition.is_null()) {
+    GetOStream(Warnings0) << "No repartitioning necessary: partitions were left unchanged by the repartitioner" << std::endl;
+    Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
+    return;
+  }
 
+  // create new decomposition vector
+  Teuchos::RCP<GOVector> ret    = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(A->getRowMap(), false);
+  ArrayRCP<GO> retDecompEntries = ret->getDataNonConst(0);
+
+  // block size of output vector
+  LocalOrdinal blkSize = 1;
+
+  // check for blocking/striding information
+  if (A->IsView("stridedMaps") &&
+      Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+    Xpetra::viewLabel_t oldView  = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+    RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null, Exceptions::BadCast, "MueLu::CloneRepartitionInterface::Build: cast to strided row map failed.");
+    LocalOrdinal stridedBlock = strMap->getStridedBlockId();
+    if (stridedBlock == -1)
+      blkSize = strMap->getFixedBlockSize();
+    else {
+      std::vector<size_t> strInfo = strMap->getStridingData();
+      blkSize                     = strInfo[stridedBlock];
     }
+    oldView = A->SwitchToView(oldView);
+    GetOStream(Statistics1) << "CloneRepartitionInterface::Build():"
+                            << " found blockdim=" << blkSize << " from strided maps." << std::endl;
+  } else {
+    GetOStream(Statistics1) << "CloneRepartitionInterface::Build(): no striding information available. Use blockdim=" << blkSize << " (DofsPerNode)." << std::endl;
+    blkSize = A->GetFixedBlockSize();
+  }
 
-    // ======================================================================================================
-    // Construct decomposition vector
-    // ======================================================================================================
-    RCP<GOVector> decomposition = Teuchos::null;
+  // plausibility check!
+  size_t inLocalLength  = decomposition->getLocalLength();
+  size_t outLocalLength = A->getRowMap()->getLocalNumElements();
 
-    // extract decomposition vector
-    decomposition = Get<RCP<GOVector> >(currentLevel, "Partition");
-    ArrayRCP<const GO> decompEntries = decomposition->getData(0);
+  // only for non-strided maps
+  size_t numLocalNodes = outLocalLength / blkSize;
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(outLocalLength % blkSize) != 0, MueLu::Exceptions::RuntimeError, "CloneRepartitionInterface: inconsistent number of local DOFs (" << outLocalLength << ") and degrees of freedoms (" << blkSize << ")");
 
-    if (decomposition.is_null()) {
-      GetOStream(Warnings0) << "No repartitioning necessary: partitions were left unchanged by the repartitioner" << std::endl;
-      Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
-      return;
-    }
+  if (numLocalNodes > 0) {
+    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(inLocalLength % numLocalNodes) != 0, MueLu::Exceptions::RuntimeError, "CloneRepartitionInterface: inconsistent number of local DOFs (" << inLocalLength << ") and number of local nodes (" << numLocalNodes << ")");
+    LocalOrdinal inBlkSize = Teuchos::as<LocalOrdinal>(inLocalLength / numLocalNodes);
+    // TEUCHOS_TEST_FOR_EXCEPTION(blkSize != inBlkSize, MueLu::Exceptions::RuntimeError,"CloneRepartitionInterface: input block size = " << inBlkSize << " outpub block size = " << blkSize << ". They should be the same.");
 
-    // create new decomposition vector
-    Teuchos::RCP<GOVector> ret = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(A->getRowMap(), false);
-    ArrayRCP<GO> retDecompEntries = ret->getDataNonConst(0);
-
-    // block size of output vector
-    LocalOrdinal blkSize = 1;
-
-    // check for blocking/striding information
-    if(A->IsView("stridedMaps") &&
-       Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null,Exceptions::BadCast,"MueLu::CloneRepartitionInterface::Build: cast to strided row map failed.");
-      LocalOrdinal stridedBlock = strMap->getStridedBlockId();
-      if (stridedBlock == -1)
-        blkSize = strMap->getFixedBlockSize();
-      else {
-        std::vector<size_t> strInfo = strMap->getStridingData();
-        blkSize = strInfo[stridedBlock];
+    for (LO i = 0; i < Teuchos::as<LO>(numLocalNodes); i++) {
+      for (LO j = 0; j < blkSize; j++) {
+        retDecompEntries[i * blkSize + j] = Teuchos::as<GO>(decompEntries[i * inBlkSize]);
       }
-      oldView = A->SwitchToView(oldView);
-      GetOStream(Statistics1) << "CloneRepartitionInterface::Build():" << " found blockdim=" << blkSize << " from strided maps."<< std::endl;
-    } else {
-      GetOStream(Statistics1) << "CloneRepartitionInterface::Build(): no striding information available. Use blockdim=" << blkSize << " (DofsPerNode)." << std::endl;
-      blkSize = A->GetFixedBlockSize();
     }
+  }  // end if numLocalNodes > 0
+  Set(currentLevel, "Partition", ret);
+}  // Build()
 
-    // plausibility check!
-    size_t inLocalLength  = decomposition->getLocalLength();
-    size_t outLocalLength = A->getRowMap()->getLocalNumElements();
-
-    // only for non-strided maps
-    size_t numLocalNodes = outLocalLength / blkSize;
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(outLocalLength  % blkSize) != 0, MueLu::Exceptions::RuntimeError,"CloneRepartitionInterface: inconsistent number of local DOFs (" << outLocalLength << ") and degrees of freedoms (" << blkSize <<")");
-
-    if (numLocalNodes > 0) {
-      TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(inLocalLength  % numLocalNodes) != 0, MueLu::Exceptions::RuntimeError,"CloneRepartitionInterface: inconsistent number of local DOFs (" << inLocalLength << ") and number of local nodes (" << numLocalNodes << ")");
-      LocalOrdinal inBlkSize = Teuchos::as<LocalOrdinal>(inLocalLength / numLocalNodes);
-      //TEUCHOS_TEST_FOR_EXCEPTION(blkSize != inBlkSize, MueLu::Exceptions::RuntimeError,"CloneRepartitionInterface: input block size = " << inBlkSize << " outpub block size = " << blkSize << ". They should be the same.");
-
-      for(LO i = 0; i<Teuchos::as<LO>(numLocalNodes); i++) {
-        for(LO j = 0; j < blkSize; j++) {
-          retDecompEntries[i*blkSize + j] = Teuchos::as<GO>(decompEntries[i*inBlkSize]);
-        }
-      }
-    } // end if numLocalNodes > 0
-    Set(currentLevel, "Partition", ret);
-  } //Build()
-
-
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_CLONEREPARTITIONINTERFACE_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_decl.hpp
index 58a230c5fcf2..a339027e78a5 100644
--- a/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_decl.hpp
@@ -14,7 +14,7 @@
 
 #include <Xpetra_Matrix_fwd.hpp>
 #include <Xpetra_VectorFactory_fwd.hpp>
-#include <Xpetra_CrsGraphFactory_fwd.hpp> //TODO
+#include <Xpetra_CrsGraphFactory_fwd.hpp>  //TODO
 
 #ifdef HAVE_MUELU_EPETRA
 #include <Xpetra_EpetraCrsGraph_fwd.hpp>
@@ -66,53 +66,45 @@ namespace MueLu {
 
 */
 
-  //FIXME: this class should not be templated
-  template <class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class IsorropiaInterface : public SingleLevelFactoryBase {
-
-    typedef double Scalar; // FIXME This class only works with the Epetra stack, i.e., Scalar = double
+// FIXME: this class should not be templated
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class IsorropiaInterface : public SingleLevelFactoryBase {
+  typedef double Scalar;  // FIXME This class only works with the Epetra stack, i.e., Scalar = double
 #undef MUELU_ISORROPIAINTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
-
-    //! Constructor
-    IsorropiaInterface() { }
-
-    //! Destructor
-    virtual ~IsorropiaInterface() { }
-    //@}
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Constructor
+  IsorropiaInterface() {}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level & level) const;
-    //@}
+  //! Destructor
+  virtual ~IsorropiaInterface() {}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &level) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
+  //@}
 
-  private:
+ private:
+};  // class IsorropiaInterface
 
-
-
-  };  //class IsorropiaInterface
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ISORROPIAINTERFACE_SHORT
 //#endif //if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
 
-
 #endif /* MUELU_ISORROPIAINTERFACE_DECL_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_def.hpp
index bcf7361dfc14..1ff5a0861627 100644
--- a/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_IsorropiaInterface_def.hpp
@@ -22,7 +22,6 @@
 #ifdef HAVE_MUELU_ISORROPIA
 #include <Isorropia_Exception.hpp>
 
-
 #ifdef HAVE_MUELU_EPETRA
 #include <Xpetra_EpetraCrsGraph.hpp>
 #include <Epetra_CrsGraph.h>
@@ -30,7 +29,7 @@
 #endif
 
 #include <Xpetra_TpetraCrsGraph.hpp>
-#endif // ENDIF HAVE_MUELU_ISORROPIA
+#endif  // ENDIF HAVE_MUELU_ISORROPIA
 
 #include "MueLu_Level.hpp"
 #include "MueLu_Exceptions.hpp"
@@ -40,204 +39,196 @@
 
 namespace MueLu {
 
- template <class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",                    Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo",   Teuchos::null, "Generating factory of UnAmalgamationInfo");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    return validParamList;
-  }
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
 
+  return validParamList;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    Input(currentLevel, "UnAmalgamationInfo");
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  Input(currentLevel, "UnAmalgamationInfo");
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
-    FactoryMonitor m(*this, "Build", level);
-    typedef Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>  BlockMap;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void IsorropiaInterface<LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
+  FactoryMonitor m(*this, "Build", level);
+  typedef Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node> BlockMap;
 
-    RCP<Matrix> A                  = Get< RCP<Matrix> >(level, "A");
-    RCP<AmalgamationInfo> amalInfo = Get< RCP<AmalgamationInfo> >(level, "UnAmalgamationInfo");
-    GO          numParts           = Get< int >(level, "number of partitions");
+  RCP<Matrix> A                  = Get<RCP<Matrix> >(level, "A");
+  RCP<AmalgamationInfo> amalInfo = Get<RCP<AmalgamationInfo> >(level, "UnAmalgamationInfo");
+  GO numParts                    = Get<int>(level, "number of partitions");
 
-    RCP<const Map> rowMap = A->getRowMap();
-    RCP<const Map> colMap = A->getColMap();
+  RCP<const Map> rowMap = A->getRowMap();
+  RCP<const Map> colMap = A->getColMap();
 
-    if (numParts == 1 || numParts == -1) {
-      // Running on one processor, so decomposition is the trivial one, all zeros.
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      Set(level, "AmalgamatedPartition", decomposition);
-      return;
-    }
+  if (numParts == 1 || numParts == -1) {
+    // Running on one processor, so decomposition is the trivial one, all zeros.
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+    Set(level, "AmalgamatedPartition", decomposition);
+    return;
+  }
 
-    // ok, reconstruct graph information of matrix A
-    // Note, this is the non-rebalanced matrix A and we need the graph
-    // of the non-rebalanced matrix A. We cannot make use of the "Graph"
-    // that is/will be built for the aggregates later for several reasons
-    // 1) the "Graph" for the aggregates is meant to be based on the rebalanced matrix A
-    // 2) we cannot call a CoalesceDropFactory::Build here since this is very complicated and
-    //    completely messes up the whole factory chain
-    // 3) CoalesceDropFactory is meant to provide some minimal Graph information for the aggregation
-    //    (LWGraph), but here we need the full CrsGraph for Isorropia as input
-
-    // That is, why this code is somewhat redundant to CoalesceDropFactory
-
-    LO blockdim = 1;                          // block dim for fixed size blocks
-    GO indexBase = rowMap->getIndexBase();    // index base of maps
-    GO offset    = 0;
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    //LO stridedblocksize = blockdim; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
-
-    // 1) check for blocking/striding information
-    //    fill above variables
-    if(A->IsView("stridedMaps") &&
-       Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null,Exceptions::BadCast,"MueLu::IsorropiaInterface::Build: cast to strided row map failed.");
-      blockdim = strMap->getFixedBlockSize();
-      offset   = strMap->getOffset();
-      blockid  = strMap->getStridedBlockId();
-      if (blockid > -1) {
-        std::vector<size_t> stridingInfo = strMap->getStridingData();
-        for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
-          nStridedOffset += stridingInfo[j];
-        //stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
-
-      }// else {
-      //  stridedblocksize = blockdim;
-      //}
-      oldView = A->SwitchToView(oldView);
-      //GetOStream(Statistics0) << "IsorropiaInterface::Build():" << " found blockdim=" << blockdim << " from strided maps (blockid=" << blockid << ", strided block size=" << stridedblocksize << "). offset=" << offset << std::endl;
-    } else GetOStream(Statistics0) << "IsorropiaInterface::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
-
-    // 2) get row map for amalgamated matrix (graph of A)
-    //    with same distribution over all procs as row map of A
-    RCP<const Map> nodeMap= amalInfo->getNodeRowMap();
-    RCP<const BlockedMap> bnodeMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(nodeMap);
-    if(!bnodeMap.is_null()) nodeMap=bnodeMap->getMap();
-
-    GetOStream(Statistics0) << "IsorropiaInterface:Build(): nodeMap " << nodeMap->getLocalNumElements() << "/" << nodeMap->getGlobalNumElements() << " elements" << std::endl;
-    
-
-    // 3) create graph of amalgamated matrix
-    RCP<CrsGraph> crsGraph = CrsGraphFactory::Build(nodeMap, A->getLocalMaxNumRowEntries()*blockdim);
-
-    // 4) do amalgamation. generate graph of amalgamated matrix
-    for(LO row=0; row<Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); row++) {
-      // get global DOF id
-      GO grid = rowMap->getGlobalElement(row);
-
-      // translate grid to nodeid
-      // JHU 2019-20-May this is identical to AmalgamationFactory::DOFGid2NodeId(), and is done
-      // to break a circular dependence when libraries are built statically
-      GO nodeId = (grid - offset - indexBase) / blockdim + indexBase;
-
-      size_t nnz = A->getNumEntriesInLocalRow(row);
-      Teuchos::ArrayView<const LO> indices;
-      Teuchos::ArrayView<const SC> vals;
-      A->getLocalRowView(row, indices, vals);
-
-      RCP<std::vector<GO> > cnodeIds = Teuchos::rcp(new std::vector<GO>);  // global column block ids
-      LO realnnz = 0;
-      for(LO col=0; col<Teuchos::as<LO>(nnz); col++) {
-        GO gcid = colMap->getGlobalElement(indices[col]); // global column id
-
-        if(vals[col]!=0.0) {
-          // JHU 2019-20-May this is identical to AmalgamationFactory::DOFGid2NodeId(), and is done
-          // to break a circular dependence when libraries are built statically
-          GO cnodeId = (gcid - offset - indexBase) / blockdim + indexBase;
-          cnodeIds->push_back(cnodeId);
-          realnnz++; // increment number of nnz in matrix row
-        }
+  // ok, reconstruct graph information of matrix A
+  // Note, this is the non-rebalanced matrix A and we need the graph
+  // of the non-rebalanced matrix A. We cannot make use of the "Graph"
+  // that is/will be built for the aggregates later for several reasons
+  // 1) the "Graph" for the aggregates is meant to be based on the rebalanced matrix A
+  // 2) we cannot call a CoalesceDropFactory::Build here since this is very complicated and
+  //    completely messes up the whole factory chain
+  // 3) CoalesceDropFactory is meant to provide some minimal Graph information for the aggregation
+  //    (LWGraph), but here we need the full CrsGraph for Isorropia as input
+
+  // That is, why this code is somewhat redundant to CoalesceDropFactory
+
+  LO blockdim       = 1;                       // block dim for fixed size blocks
+  GO indexBase      = rowMap->getIndexBase();  // index base of maps
+  GO offset         = 0;
+  LO blockid        = -1;  // block id in strided map
+  LO nStridedOffset = 0;   // DOF offset for strided block id "blockid" (default = 0)
+  // LO stridedblocksize = blockdim; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+
+  // 1) check for blocking/striding information
+  //    fill above variables
+  if (A->IsView("stridedMaps") &&
+      Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+    Xpetra::viewLabel_t oldView  = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+    RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null, Exceptions::BadCast, "MueLu::IsorropiaInterface::Build: cast to strided row map failed.");
+    blockdim = strMap->getFixedBlockSize();
+    offset   = strMap->getOffset();
+    blockid  = strMap->getStridedBlockId();
+    if (blockid > -1) {
+      std::vector<size_t> stridingInfo = strMap->getStridingData();
+      for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
+        nStridedOffset += stridingInfo[j];
+      // stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+
+    }  // else {
+    //  stridedblocksize = blockdim;
+    //}
+    oldView = A->SwitchToView(oldView);
+    // GetOStream(Statistics0) << "IsorropiaInterface::Build():" << " found blockdim=" << blockdim << " from strided maps (blockid=" << blockid << ", strided block size=" << stridedblocksize << "). offset=" << offset << std::endl;
+  } else
+    GetOStream(Statistics0) << "IsorropiaInterface::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
+
+  // 2) get row map for amalgamated matrix (graph of A)
+  //    with same distribution over all procs as row map of A
+  RCP<const Map> nodeMap         = amalInfo->getNodeRowMap();
+  RCP<const BlockedMap> bnodeMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(nodeMap);
+  if (!bnodeMap.is_null()) nodeMap = bnodeMap->getMap();
+
+  GetOStream(Statistics0) << "IsorropiaInterface:Build(): nodeMap " << nodeMap->getLocalNumElements() << "/" << nodeMap->getGlobalNumElements() << " elements" << std::endl;
+
+  // 3) create graph of amalgamated matrix
+  RCP<CrsGraph> crsGraph = CrsGraphFactory::Build(nodeMap, A->getLocalMaxNumRowEntries() * blockdim);
+
+  // 4) do amalgamation. generate graph of amalgamated matrix
+  for (LO row = 0; row < Teuchos::as<LO>(A->getRowMap()->getLocalNumElements()); row++) {
+    // get global DOF id
+    GO grid = rowMap->getGlobalElement(row);
+
+    // translate grid to nodeid
+    // JHU 2019-20-May this is identical to AmalgamationFactory::DOFGid2NodeId(), and is done
+    // to break a circular dependence when libraries are built statically
+    GO nodeId = (grid - offset - indexBase) / blockdim + indexBase;
+
+    size_t nnz = A->getNumEntriesInLocalRow(row);
+    Teuchos::ArrayView<const LO> indices;
+    Teuchos::ArrayView<const SC> vals;
+    A->getLocalRowView(row, indices, vals);
+
+    RCP<std::vector<GO> > cnodeIds = Teuchos::rcp(new std::vector<GO>);  // global column block ids
+    LO realnnz                     = 0;
+    for (LO col = 0; col < Teuchos::as<LO>(nnz); col++) {
+      GO gcid = colMap->getGlobalElement(indices[col]);  // global column id
+
+      if (vals[col] != 0.0) {
+        // JHU 2019-20-May this is identical to AmalgamationFactory::DOFGid2NodeId(), and is done
+        // to break a circular dependence when libraries are built statically
+        GO cnodeId = (gcid - offset - indexBase) / blockdim + indexBase;
+        cnodeIds->push_back(cnodeId);
+        realnnz++;  // increment number of nnz in matrix row
       }
+    }
 
-      Teuchos::ArrayRCP<GO> arr_cnodeIds = Teuchos::arcp( cnodeIds );
+    Teuchos::ArrayRCP<GO> arr_cnodeIds = Teuchos::arcp(cnodeIds);
 
-      if(arr_cnodeIds.size() > 0 )
-        crsGraph->insertGlobalIndices(nodeId, arr_cnodeIds());
-    }
-    // fill matrix graph
-    crsGraph->fillComplete(nodeMap,nodeMap);
+    if (arr_cnodeIds.size() > 0)
+      crsGraph->insertGlobalIndices(nodeId, arr_cnodeIds());
+  }
+  // fill matrix graph
+  crsGraph->fillComplete(nodeMap, nodeMap);
 
 #ifdef HAVE_MPI
 #ifdef HAVE_MUELU_ISORROPIA
 
-    // prepare parameter list for Isorropia
-    Teuchos::ParameterList paramlist;
-    paramlist.set("NUM PARTS", toString(numParts));
-
-    /*STRUCTURALLY SYMMETRIC [NO/yes] (is symmetrization required?)
-    PARTITIONING METHOD [block/cyclic/random/rcb/rib/hsfc/graph/HYPERGRAPH]
-    NUM PARTS [int k] (global number of parts)
-    IMBALANCE TOL [float tol] (1.0 is perfect balance)
-    BALANCE OBJECTIVE [ROWS/nonzeros]
-    */
-    Teuchos::ParameterList& sublist = paramlist.sublist("Zoltan");
-    sublist.set("LB_APPROACH", "PARTITION");
-
+  // prepare parameter list for Isorropia
+  Teuchos::ParameterList paramlist;
+  paramlist.set("NUM PARTS", toString(numParts));
 
+  /*STRUCTURALLY SYMMETRIC [NO/yes] (is symmetrization required?)
+  PARTITIONING METHOD [block/cyclic/random/rcb/rib/hsfc/graph/HYPERGRAPH]
+  NUM PARTS [int k] (global number of parts)
+  IMBALANCE TOL [float tol] (1.0 is perfect balance)
+  BALANCE OBJECTIVE [ROWS/nonzeros]
+  */
+  Teuchos::ParameterList& sublist = paramlist.sublist("Zoltan");
+  sublist.set("LB_APPROACH", "PARTITION");
 
 #ifdef HAVE_MUELU_EPETRA
-    RCP< Xpetra::EpetraCrsGraphT<GO, Node> > epCrsGraph = Teuchos::rcp_dynamic_cast<Xpetra::EpetraCrsGraphT<GO, Node> >(crsGraph);
-    if(epCrsGraph != Teuchos::null) {
-      RCP< const Epetra_CrsGraph> epetraCrsGraph = epCrsGraph->getEpetra_CrsGraph();
+  RCP<Xpetra::EpetraCrsGraphT<GO, Node> > epCrsGraph = Teuchos::rcp_dynamic_cast<Xpetra::EpetraCrsGraphT<GO, Node> >(crsGraph);
+  if (epCrsGraph != Teuchos::null) {
+    RCP<const Epetra_CrsGraph> epetraCrsGraph = epCrsGraph->getEpetra_CrsGraph();
 
-      RCP<Isorropia::Epetra::Partitioner> isoPart = Teuchos::rcp(new Isorropia::Epetra::Partitioner(epetraCrsGraph,paramlist));
+    RCP<Isorropia::Epetra::Partitioner> isoPart = Teuchos::rcp(new Isorropia::Epetra::Partitioner(epetraCrsGraph, paramlist));
 
-      int size = 0;
-      const int* array = NULL;
-      isoPart->extractPartsView(size,array);
+    int size         = 0;
+    const int* array = NULL;
+    isoPart->extractPartsView(size, array);
 
-      TEUCHOS_TEST_FOR_EXCEPTION(size != Teuchos::as<int>(nodeMap->getLocalNumElements()), Exceptions::RuntimeError, "length of array returned from extractPartsView does not match local length of rowMap");
+    TEUCHOS_TEST_FOR_EXCEPTION(size != Teuchos::as<int>(nodeMap->getLocalNumElements()), Exceptions::RuntimeError, "length of array returned from extractPartsView does not match local length of rowMap");
 
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(nodeMap, false);
-      ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-
-      // fill vector with amalgamated information about partitioning
-      for(int i = 0; i<size; i++) {
-        decompEntries[i] = Teuchos::as<GO>(array[i]);
-      }
-
-      Set(level, "AmalgamatedPartition", decomposition);
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(nodeMap, false);
+    ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
 
+    // fill vector with amalgamated information about partitioning
+    for (int i = 0; i < size; i++) {
+      decompEntries[i] = Teuchos::as<GO>(array[i]);
     }
-#endif // ENDIF HAVE_MUELU_EPETRA
-
-#ifdef HAVE_MUELU_INST_DOUBLE_INT_INT
-    RCP< Xpetra::TpetraCrsGraph<LO, GO, Node> > tpCrsGraph = Teuchos::rcp_dynamic_cast<Xpetra::TpetraCrsGraph<LO, GO, Node> >(crsGraph);
-    TEUCHOS_TEST_FOR_EXCEPTION(tpCrsGraph != Teuchos::null, Exceptions::RuntimeError, "Tpetra is not supported with Isorropia.");
-#else
-    TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "Isorropia is an interface to Zoltan which only has support for LO=GO=int and SC=double.");
-#endif // ENDIF HAVE_MUELU_INST_DOUBLE_INT_INT
-#endif // HAVE_MUELU_ISORROPIA
-#else  // if we don't have MPI
-
 
-    // Running on one processor, so decomposition is the trivial one, all zeros.
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
     Set(level, "AmalgamatedPartition", decomposition);
+  }
+#endif  // ENDIF HAVE_MUELU_EPETRA
 
-#endif // HAVE_MPI
-    // throw a more helpful error message if something failed
-    //TEUCHOS_TEST_FOR_EXCEPTION(!level.IsAvailable("AmalgamatedPartition"), Exceptions::RuntimeError, "IsorropiaInterface::Build : no \'Partition\' vector available on level. Isorropia failed to build a partition of the non-repartitioned graph of A. Please make sure, that Isorropia is correctly compiled (Epetra/Tpetra).");
+#ifdef HAVE_MUELU_INST_DOUBLE_INT_INT
+  RCP<Xpetra::TpetraCrsGraph<LO, GO, Node> > tpCrsGraph = Teuchos::rcp_dynamic_cast<Xpetra::TpetraCrsGraph<LO, GO, Node> >(crsGraph);
+  TEUCHOS_TEST_FOR_EXCEPTION(tpCrsGraph != Teuchos::null, Exceptions::RuntimeError, "Tpetra is not supported with Isorropia.");
+#else
+  TEUCHOS_TEST_FOR_EXCEPTION(false, Exceptions::RuntimeError, "Isorropia is an interface to Zoltan which only has support for LO=GO=int and SC=double.");
+#endif  // ENDIF HAVE_MUELU_INST_DOUBLE_INT_INT
+#endif  // HAVE_MUELU_ISORROPIA
+#else   // if we don't have MPI
 
-  } //Build()
+  // Running on one processor, so decomposition is the trivial one, all zeros.
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+  Set(level, "AmalgamatedPartition", decomposition);
 
+#endif  // HAVE_MPI
+  // throw a more helpful error message if something failed
+  // TEUCHOS_TEST_FOR_EXCEPTION(!level.IsAvailable("AmalgamatedPartition"), Exceptions::RuntimeError, "IsorropiaInterface::Build : no \'Partition\' vector available on level. Isorropia failed to build a partition of the non-repartitioned graph of A. Please make sure, that Isorropia is correctly compiled (Epetra/Tpetra).");
 
+}  // Build()
 
-} //namespace MueLu
+}  // namespace MueLu
 
 //#endif //if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
 
-
 #endif /* MUELU_ISORROPIAINTERFACE_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_decl.hpp
index 3a99be3a28fe..955682a2676f 100644
--- a/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_decl.hpp
@@ -61,90 +61,89 @@
 
 namespace MueLu {
 
-  /*!
-    @class NodePartitionInterface
-    @brief Partitioning within a node only
-    @ingroup Rebalancing
-
-    This interface provides partitioning within a node
-
-    ## Input/output of NodePartitionInterface ##
-
-    ### User parameters of NodePartitionInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | Coordinates                            | Factory | null  |   | * | (*) | Factory generating coordinates vector used for rebalancing. The coordinates are only needed when the chosen algorithm is 'multijagged' or 'rcb'.
-    | ParameterList                          | ParamterList | null |  | * |  | NodePartition parameters
-    | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
-    | NodeComm                               | Factory | null  |   | * | (*) | Factory generating NodeComm.
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NodePartitionInterface::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see NodePartitionInterface::DeclareInput).
-
-    ### Variables provided by NodePartitionInterface ###
-
-    After NodePartitionInterface::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Partition       | NodePartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
-
-    The "Partition" vector is used as input for the RepartitionFactory class.
-    If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
-    by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
-
-    The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
-    It stores the result in the "number of partitions" variable on the current level (type = GO).
-    If it is "number of partitions=1" we skip the NodePartition call and just create an dummy "Partition" vector containing zeros only.
-    If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
-    If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
-
-  */
-
-  //FIXME: this class should not be templated
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class NodePartitionInterface : public SingleLevelFactoryBase {
+/*!
+  @class NodePartitionInterface
+  @brief Partitioning within a node only
+  @ingroup Rebalancing
+
+  This interface provides partitioning within a node
+
+  ## Input/output of NodePartitionInterface ##
+
+  ### User parameters of NodePartitionInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | Coordinates                            | Factory | null  |   | * | (*) | Factory generating coordinates vector used for rebalancing. The coordinates are only needed when the chosen algorithm is 'multijagged' or 'rcb'.
+  | ParameterList                          | ParamterList | null |  | * |  | NodePartition parameters
+  | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
+  | NodeComm                               | Factory | null  |   | * | (*) | Factory generating NodeComm.
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NodePartitionInterface::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see NodePartitionInterface::DeclareInput).
+
+  ### Variables provided by NodePartitionInterface ###
+
+  After NodePartitionInterface::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Partition       | NodePartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
+
+  The "Partition" vector is used as input for the RepartitionFactory class.
+  If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
+  by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
+
+  The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
+  It stores the result in the "number of partitions" variable on the current level (type = GO).
+  If it is "number of partitions=1" we skip the NodePartition call and just create an dummy "Partition" vector containing zeros only.
+  If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
+  If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
+
+*/
+
+// FIXME: this class should not be templated
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class NodePartitionInterface : public SingleLevelFactoryBase {
 #undef MUELU_NODEPARTITIONINTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! @name Constructors/Destructors
-    //@{
+  //! Constructor
+  NodePartitionInterface();
 
-    //! Constructor
-    NodePartitionInterface();
+  //! Destructor
+  virtual ~NodePartitionInterface() {}
+  //@}
 
-    //! Destructor
-    virtual ~NodePartitionInterface() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level& currentLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& currentLevel) const;
+  //@}
 
-    //@}
+ private:
+  RCP<ParameterList> defaultNodePartitionParams;
 
-  private:
-    RCP<ParameterList> defaultNodePartitionParams;
+};  // class NodePartitionInterface
 
-  };  //class NodePartitionInterface
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_NODEPARTITIONINTERFACE_SHORT
-#endif //if defined(HAVE_MPI)
+#endif  // if defined(HAVE_MPI)
 
-#endif // MUELU_NODEPARTITIONINTERFACE_DECL_HPP
+#endif  // MUELU_NODEPARTITIONINTERFACE_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_def.hpp
index 52c70bf1e51d..0b42897d2cf2 100644
--- a/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_NodePartitionInterface_def.hpp
@@ -62,67 +62,65 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodePartitionInterface() { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::NodePartitionInterface() {}
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: node id");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >   ("A",             Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >   ("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >   ("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
-
-    // We don't really need this, but we might need it on coarser levels
-    validParamList->set< RCP<const FactoryBase> >   ("Coordinates",   Teuchos::null, "Factory of the coordinates");
-    return validParamList;
+  SET_VALID_ENTRY("repartition: node id");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
+
+  // We don't really need this, but we might need it on coarser levels
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory of the coordinates");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  Input(currentLevel, "Node Comm");
+  Input(currentLevel, "Coordinates");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
+  FactoryMonitor m(*this, "Build", level);
+  RCP<Matrix> A         = Get<RCP<Matrix> >(level, "A");
+  RCP<const Map> rowMap = A->getRowMap();
+
+  int numParts = Get<int>(level, "number of partitions");
+  if (numParts == 1 || numParts == -1) {
+    // Single processor, decomposition is trivial: all zeros
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+    Set(level, "Partition", decomposition);
+    return;
   }
 
+  // Let us repartition nodally
+  RCP<const Teuchos::Comm<int> > NodeComm = Get<RCP<const Teuchos::Comm<int> > >(level, "Node Comm");
+  TEUCHOS_TEST_FOR_EXCEPTION(NodeComm.is_null(), Exceptions::RuntimeError, "MueLu::NodePartitionInterface::Build(): NodeComm is null.");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    Input(currentLevel, "Node Comm");
-    Input(currentLevel, "Coordinates");
-  }
+  // Get the rank (in current comm) of rank 0 in my NodeComm
+  int nodeZeroRank = A->getMap()->getComm()->getRank();
+  Teuchos::broadcast<int, int>(*NodeComm, 0, Teuchos::inOutArg(nodeZeroRank));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NodePartitionInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
-    FactoryMonitor m(*this, "Build", level);
-    RCP<Matrix>    A      = Get<RCP<Matrix> >(level, "A");
-    RCP<const Map> rowMap = A->getRowMap();
-
-    int numParts = Get<int>(level, "number of partitions");
-    if (numParts == 1 || numParts == -1) {
-      // Single processor, decomposition is trivial: all zeros
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      Set(level, "Partition", decomposition);
-      return;
-    }
-
-    // Let us repartition nodally
-    RCP<const Teuchos::Comm<int> > NodeComm = Get< RCP<const Teuchos::Comm<int> > >(level, "Node Comm");
-    TEUCHOS_TEST_FOR_EXCEPTION(NodeComm.is_null(), Exceptions::RuntimeError, "MueLu::NodePartitionInterface::Build(): NodeComm is null.");
-
-    // Get the rank (in current comm) of rank 0 in my NodeComm
-    int nodeZeroRank =A->getMap()->getComm()->getRank();
-    Teuchos::broadcast<int,int>(*NodeComm,0,Teuchos::inOutArg(nodeZeroRank));
-
-    // A "Partition" from a *Interface is supposed to be is a vector of length # of my rows with the partition number to which the unknown is assigned
-    // BUT, since we're bypassing remap for NodePartition, we'll return a *rank* of the guy who gets each unknown (which is what remap outputs).
-    RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO,LO,GO,NO>::Build(rowMap, false);
-    decomposition->putScalar(Teuchos::as<GO>(nodeZeroRank));
+  // A "Partition" from a *Interface is supposed to be is a vector of length # of my rows with the partition number to which the unknown is assigned
+  // BUT, since we're bypassing remap for NodePartition, we'll return a *rank* of the guy who gets each unknown (which is what remap outputs).
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);
+  decomposition->putScalar(Teuchos::as<GO>(nodeZeroRank));
 
-    Set(level, "Partition", decomposition);
-
-  }
+  Set(level, "Partition", decomposition);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif //if defined(HAVE_MPI)
+#endif  // if defined(HAVE_MPI)
 
-#endif // MUELU_NODEPARTITIONINTERFACE_DEF_HPP
+#endif  // MUELU_NODEPARTITIONINTERFACE_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_decl.hpp
index 91cb53545f56..1639b2c943e6 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_decl.hpp
@@ -59,61 +59,61 @@
 #include "MueLu_PerfUtils_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class RebalanceAcFactory
-    @brief Factory for building coarse matrices.
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceAcFactory : public TwoLevelFactoryBase {
+/*!
+  @class RebalanceAcFactory
+  @brief Factory for building coarse matrices.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceAcFactory : public TwoLevelFactoryBase {
 #undef MUELU_REBALANCEACFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    RebalanceAcFactory() { }
+  RebalanceAcFactory() {}
 
-    virtual ~RebalanceAcFactory() { }
+  virtual ~RebalanceAcFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &fineLevel, Level &coarseLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //@{
-    /*! @brief Add rebalancing factory in the end of list of rebalancing factories in RebalanceAcFactory.
+  //@{
+  /*! @brief Add rebalancing factory in the end of list of rebalancing factories in RebalanceAcFactory.
 
-    Rebalancing factories are derived from SingleLevelFactoryBase and rebalance the underlaying object
-    (e.g. map, vector,...) to fit to the rebalanced maps.
-    */
-    void AddRebalanceFactory(const RCP<const FactoryBase>& factory);
+  Rebalancing factories are derived from SingleLevelFactoryBase and rebalance the underlaying object
+  (e.g. map, vector,...) to fit to the rebalanced maps.
+  */
+  void AddRebalanceFactory(const RCP<const FactoryBase> &factory);
 
-    //! Returns number of transfer factories.
-    size_t NumRebalanceFactories() const { return rebalanceFacts_.size(); }
+  //! Returns number of transfer factories.
+  size_t NumRebalanceFactories() const { return rebalanceFacts_.size(); }
 
-    //@}
+  //@}
 
-  private:
-    //! list of user-defined rebalancing Factories
-    std::vector<RCP<const FactoryBase> > rebalanceFacts_;
+ private:
+  //! list of user-defined rebalancing Factories
+  std::vector<RCP<const FactoryBase> > rebalanceFacts_;
 
-  }; //class RebalanceAcFactory
+};  // class RebalanceAcFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCEACFACTORY_SHORT
-#endif // MUELU_REBALANCEACFACTORY_DECL_HPP
+#endif  // MUELU_REBALANCEACFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_def.hpp
index 81f8f5eb0130..31d4003a4edc 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceAcFactory_def.hpp
@@ -60,122 +60,120 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: use subcommunicators");
-    SET_VALID_ENTRY("repartition: use subcommunicators in place");
+  SET_VALID_ENTRY("repartition: use subcommunicators");
+  SET_VALID_ENTRY("repartition: use subcommunicators in place");
 #undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",        Teuchos::null, "Generating factory of the matrix A for rebalancing");
-    validParamList->set< RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the importer");
-    validParamList->set< RCP<const FactoryBase> >("InPlaceMap", Teuchos::null, "Generating factory of the InPlaceMap");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A for rebalancing");
+  validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the importer");
+  validParamList->set<RCP<const FactoryBase> >("InPlaceMap", Teuchos::null, "Generating factory of the InPlaceMap");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &/* fineLevel */, Level &coarseLevel) const {
-    Input(coarseLevel, "A");
-    const Teuchos::ParameterList & pL = GetParameterList();
-    if(pL.isParameter("repartition: use subcommunicators in place") && pL.get<bool>("repartition: use subcommunicators in place")==true) {
-      Input(coarseLevel,"InPlaceMap");
-    }
-    else
-      Input(coarseLevel, "Importer");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & /* fineLevel */, Level &coarseLevel) const {
+  Input(coarseLevel, "A");
+  const Teuchos::ParameterList &pL = GetParameterList();
+  if (pL.isParameter("repartition: use subcommunicators in place") && pL.get<bool>("repartition: use subcommunicators in place") == true) {
+    Input(coarseLevel, "InPlaceMap");
+  } else
+    Input(coarseLevel, "Importer");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &/* fineLevel */, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Computing Ac", coarseLevel);
-    
-    const Teuchos::ParameterList & pL = GetParameterList();
-    RCP<Matrix> originalAc = Get< RCP<Matrix> >(coarseLevel, "A");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Computing Ac", coarseLevel);
 
-    // This is a short-circuit for if we want to leave A where it is, but restrict its communicator
-    // to something corresponding to a given map.  Maxwell1 is the prime customer for this.
-    bool inPlace = pL.get<bool>("repartition: use subcommunicators in place");
-    if(inPlace) {
-      SubFactoryMonitor subM(*this, "Rebalancing existing Ac in-place", coarseLevel);
-      RCP<const Map> newMap = Get< RCP<const Map> >(coarseLevel, "InPlaceMap");
+  const Teuchos::ParameterList &pL = GetParameterList();
+  RCP<Matrix> originalAc           = Get<RCP<Matrix> >(coarseLevel, "A");
 
-      originalAc->removeEmptyProcessesInPlace(newMap);
+  // This is a short-circuit for if we want to leave A where it is, but restrict its communicator
+  // to something corresponding to a given map.  Maxwell1 is the prime customer for this.
+  bool inPlace = pL.get<bool>("repartition: use subcommunicators in place");
+  if (inPlace) {
+    SubFactoryMonitor subM(*this, "Rebalancing existing Ac in-place", coarseLevel);
+    RCP<const Map> newMap = Get<RCP<const Map> >(coarseLevel, "InPlaceMap");
 
-      // The "in place" still leaves a dummy matrix here.  That needs to go
-      if(newMap.is_null()) originalAc = Teuchos::null;
+    originalAc->removeEmptyProcessesInPlace(newMap);
 
-      Set(coarseLevel, "A", originalAc);
-      return;
-    }
+    // The "in place" still leaves a dummy matrix here.  That needs to go
+    if (newMap.is_null()) originalAc = Teuchos::null;
 
+    Set(coarseLevel, "A", originalAc);
+    return;
+  }
+
+  RCP<const Import> rebalanceImporter = Get<RCP<const Import> >(coarseLevel, "Importer");
 
-    RCP<const Import> rebalanceImporter = Get< RCP<const Import> >(coarseLevel, "Importer");
+  if (rebalanceImporter != Teuchos::null) {
+    RCP<Matrix> rebalancedAc;
+    {
+      SubFactoryMonitor subM(*this, "Rebalancing existing Ac", coarseLevel);
+      RCP<const Map> targetMap = rebalanceImporter->getTargetMap();
 
-    if (rebalanceImporter != Teuchos::null) {
-      RCP<Matrix> rebalancedAc;
+      ParameterList XpetraList;
+      if (pL.get<bool>("repartition: use subcommunicators") == true) {
+        GetOStream(Runtime0) << "Replacing maps with a subcommunicator" << std::endl;
+        XpetraList.set("Restrict Communicator", true);
+      }
+      // NOTE: If the communicator is restricted away, Build returns Teuchos::null.
+      XpetraList.set("Timer Label", "MueLu::RebalanceAc-" + Teuchos::toString(coarseLevel.GetLevelID()));
       {
-        SubFactoryMonitor subM(*this, "Rebalancing existing Ac", coarseLevel);
-        RCP<const Map> targetMap = rebalanceImporter->getTargetMap();
-
-        ParameterList XpetraList;
-        if (pL.get<bool>("repartition: use subcommunicators") == true) {
-          GetOStream(Runtime0) << "Replacing maps with a subcommunicator" << std::endl;
-          XpetraList.set("Restrict Communicator",true);
-        }
-        // NOTE: If the communicator is restricted away, Build returns Teuchos::null.
-        XpetraList.set("Timer Label","MueLu::RebalanceAc-" + Teuchos::toString(coarseLevel.GetLevelID()));
-        {
-          SubFactoryMonitor subM2(*this, "Rebalancing existing Ac: MatrixFactory::Build", coarseLevel);
-          rebalancedAc = MatrixFactory::Build(originalAc, *rebalanceImporter, *rebalanceImporter, targetMap, targetMap, rcp(&XpetraList,false));
-        }
-
-        if (!rebalancedAc.is_null()) {
-          rebalancedAc->SetFixedBlockSize(originalAc->GetFixedBlockSize());
-          std::ostringstream oss; oss << "A_" << coarseLevel.GetLevelID(); rebalancedAc->setObjectLabel(oss.str());
-        }
-        Set(coarseLevel, "A", rebalancedAc);
+        SubFactoryMonitor subM2(*this, "Rebalancing existing Ac: MatrixFactory::Build", coarseLevel);
+        rebalancedAc = MatrixFactory::Build(originalAc, *rebalanceImporter, *rebalanceImporter, targetMap, targetMap, rcp(&XpetraList, false));
       }
-      if (!rebalancedAc.is_null() && IsPrint(Statistics2)) {
-        int oldRank = SetProcRankVerbose(rebalancedAc->getRowMap()->getComm()->getRank());
 
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedAc, "Ac (rebalanced)", params);
-
-        SetProcRankVerbose(oldRank);
+      if (!rebalancedAc.is_null()) {
+        rebalancedAc->SetFixedBlockSize(originalAc->GetFixedBlockSize());
+        std::ostringstream oss;
+        oss << "A_" << coarseLevel.GetLevelID();
+        rebalancedAc->setObjectLabel(oss.str());
       }
-
-    } else {
-      // Ac already built by the load balancing process and no load balancing needed
-      GetOStream(Runtime1) << "No rebalancing" << std::endl;
-      Set(coarseLevel, "A", originalAc);
+      Set(coarseLevel, "A", rebalancedAc);
     }
+    if (!rebalancedAc.is_null() && IsPrint(Statistics2)) {
+      int oldRank = SetProcRankVerbose(rebalancedAc->getRowMap()->getComm()->getRank());
 
-    if (rebalanceFacts_.begin() != rebalanceFacts_.end()) {
-      SubFactoryMonitor m2(*this, "Rebalance additional data", coarseLevel);
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedAc, "Ac (rebalanced)", params);
 
-      // call Build of all user-given transfer factories
-      for (std::vector<RCP<const FactoryBase> >::const_iterator it = rebalanceFacts_.begin(); it != rebalanceFacts_.end(); ++it) {
-        GetOStream(Runtime0) << "RebalanceAc: call rebalance factory " << (*it).get() << ": " << (*it)->description() << std::endl;
-        (*it)->CallBuild(coarseLevel);
-      }
+      SetProcRankVerbose(oldRank);
     }
-  } //Build()
 
+  } else {
+    // Ac already built by the load balancing process and no load balancing needed
+    GetOStream(Runtime1) << "No rebalancing" << std::endl;
+    Set(coarseLevel, "A", originalAc);
+  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddRebalanceFactory(const RCP<const FactoryBase>& factory) {
+  if (rebalanceFacts_.begin() != rebalanceFacts_.end()) {
+    SubFactoryMonitor m2(*this, "Rebalance additional data", coarseLevel);
+
+    // call Build of all user-given transfer factories
+    for (std::vector<RCP<const FactoryBase> >::const_iterator it = rebalanceFacts_.begin(); it != rebalanceFacts_.end(); ++it) {
+      GetOStream(Runtime0) << "RebalanceAc: call rebalance factory " << (*it).get() << ": " << (*it)->description() << std::endl;
+      (*it)->CallBuild(coarseLevel);
+    }
+  }
+}  // Build()
 
-    /*TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::RAPFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::RAPFactory::AddTransferFactory: Factory is being added after we have already declared input");*/
-    rebalanceFacts_.push_back(factory);
-  } //AddRebalanceFactory()
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddRebalanceFactory(const RCP<const FactoryBase> &factory) {
+  /*TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::RAPFactory::AddTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::RAPFactory::AddTransferFactory: Factory is being added after we have already declared input");*/
+  rebalanceFacts_.push_back(factory);
+}  // AddRebalanceFactory()
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_REBALANCEACFACTORY_DEF_HPP
+#endif  // MUELU_REBALANCEACFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_decl.hpp
index 3b72ea98b564..a872c21b8133 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_decl.hpp
@@ -66,81 +66,81 @@
 #include "MueLu_Types.hpp"
 
 namespace MueLu {
-  /*!
-    @class RebalanceAcFactory
-    @brief Factory to perform rebalancing of coarse matrices.
-
-    Factory that performes the actual rebalancing of coarse maps based on pre-computed rebalancing data
-    (generated by other factories).
-
-    ## Input/output ##
-
-    ### User parameters ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    A             | string | null |   | * | * | Generating factory of the un-rebalanced coarse matrix
-    Importer      | string | null |   | * | * | Generating factory of the importer used in rebalancing
-    SubImporters  | string | null |   | * | * | ??
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
-
-    ### Variables provided by this factory ###
-
-    After \c Build() , the following data is available (if requested):
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | A | RebalanceBlockAcFactory | Rebalanced coarse matrix
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceBlockAcFactory : public TwoLevelFactoryBase {
+/*!
+  @class RebalanceAcFactory
+  @brief Factory to perform rebalancing of coarse matrices.
+
+  Factory that performes the actual rebalancing of coarse maps based on pre-computed rebalancing data
+  (generated by other factories).
+
+  ## Input/output ##
+
+  ### User parameters ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  A             | string | null |   | * | * | Generating factory of the un-rebalanced coarse matrix
+  Importer      | string | null |   | * | * | Generating factory of the importer used in rebalancing
+  SubImporters  | string | null |   | * | * | ??
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
+
+  ### Variables provided by this factory ###
+
+  After \c Build() , the following data is available (if requested):
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | A | RebalanceBlockAcFactory | Rebalanced coarse matrix
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceBlockAcFactory : public TwoLevelFactoryBase {
 #undef MUELU_REBALANCEBLOCKACFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-    //! @name Input
-    //@{
+ public:
+  //! @name Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //! Add a factory manager
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
-    //@}
+  //! Add a factory manager
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &fineLevel, Level &coarseLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //@{
-    /*! @brief Add rebalancing factory in the end of list of rebalancing factories in RebalanceAcFactory.
+  //@{
+  /*! @brief Add rebalancing factory in the end of list of rebalancing factories in RebalanceAcFactory.
 
-    Rebalancing factories are derived from SingleLevelFactoryBase and rebalance the underlaying object
-    (e.g. map, vector,...) to fit to the rebalanced maps.
-    */
-    void AddRebalanceFactory(const RCP<const FactoryBase>& factory);
+  Rebalancing factories are derived from SingleLevelFactoryBase and rebalance the underlaying object
+  (e.g. map, vector,...) to fit to the rebalanced maps.
+  */
+  void AddRebalanceFactory(const RCP<const FactoryBase> &factory);
 
-    //! Returns number of transfer factories.
-    size_t NumRebalanceFactories() const { return rebalanceFacts_.size(); }
+  //! Returns number of transfer factories.
+  size_t NumRebalanceFactories() const { return rebalanceFacts_.size(); }
 
-    //@}
+  //@}
 
-  private:
-    //! list of user-defined rebalancing Factories
-    std::vector<RCP<const FactoryBase> > rebalanceFacts_;
+ private:
+  //! list of user-defined rebalancing Factories
+  std::vector<RCP<const FactoryBase> > rebalanceFacts_;
 
-    //! Input factories
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+  //! Input factories
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-  }; //class RebalanceAcFactory
+};  // class RebalanceAcFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCEBLOCKACFACTORY_SHORT
 
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_def.hpp
index 15868ad3baf7..81a5d75aa1c1 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockAcFactory_def.hpp
@@ -69,357 +69,356 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: use subcommunicators");
+  SET_VALID_ENTRY("repartition: use subcommunicators");
 #undef SET_VALID_ENTRY
 
-    validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A for rebalancing");
-    validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the matrix Importer for rebalancing");
-    validParamList->set<RCP<const FactoryBase> >("SubImporters", Teuchos::null, "Generating factory of the matrix sub-Importers for rebalancing");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A for rebalancing");
+  validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the matrix Importer for rebalancing");
+  validParamList->set<RCP<const FactoryBase> >("SubImporters", Teuchos::null, "Generating factory of the matrix sub-Importers for rebalancing");
 
-    return validParamList;
-  }
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager) {
+  FactManager_.push_back(FactManager);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  Input(coarseLevel, "A");
+
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
+    SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager) {
-    FactManager_.push_back(FactManager);
+    coarseLevel.DeclareInput("Importer", (*it)->GetFactory("Importer").get(), this);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    Input(coarseLevel, "A");
+  // Use the non-manager path if the maps / importers are generated in one place
+  if (FactManager_.size() == 0) {
+    Input(coarseLevel, "SubImporters");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Computing blocked Ac", coarseLevel);
+
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+
+  RCP<Matrix> originalAc = Get<RCP<Matrix> >(coarseLevel, "A");
+
+  RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(originalAc);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockAcFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != bA->Cols(), Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Blocked operator has " << bA->Rows() << " and " << bA->Cols() << ". We only support square matrices (with same number of blocks and columns).");
+
+  // Variables to set up map extractors for blocked operators
+  std::vector<GO> fullRangeMapVector;
+  std::vector<GO> fullDomainMapVector;
+  std::vector<RCP<const Map> > subBlockARangeMaps;
+  std::vector<RCP<const Map> > subBlockADomainMaps;
+  subBlockARangeMaps.reserve(bA->Rows());
+  subBlockADomainMaps.reserve(bA->Cols());
+
+  // store map extractors
+  RCP<const MapExtractor> rangeMapExtractor  = bA->getRangeMapExtractor();
+  RCP<const MapExtractor> domainMapExtractor = bA->getDomainMapExtractor();
+
+  // check if GIDs for full maps have to be sorted:
+  // For the Thyra mode ordering they do not have to be sorted since the GIDs are
+  // numbered as 0...n1,0...,n2 (starting with zero for each subblock). The MapExtractor
+  // generates unique GIDs during the construction.
+  // For Xpetra style, the GIDs have to be reordered. Such that one obtains a ordered
+  // list of GIDs in an increasing ordering. In Xpetra, the GIDs are all unique through
+  // out all submaps.
+  bool bThyraRangeGIDs  = rangeMapExtractor->getThyraMode();
+  bool bThyraDomainGIDs = domainMapExtractor->getThyraMode();
+
+  // vector containing rebalanced blocks (for final output)
+  std::vector<RCP<Matrix> > subBlockRebA =
+      std::vector<RCP<Matrix> >(bA->Cols() * bA->Rows(), Teuchos::null);
+
+  // vector with Import objects from the different
+  // RepartitionFactory instances
+  std::vector<RCP<const Import> > importers = std::vector<RCP<const Import> >(bA->Rows(), Teuchos::null);
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  size_t idx = 0;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
+    SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
+
+    RCP<const Import> rebalanceImporter = coarseLevel.Get<RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
+    importers[idx]                      = rebalanceImporter;
+    idx++;
+  }
+  if (FactManager_.size() == 0) {
+    importers = Get<std::vector<RCP<const Import> > >(coarseLevel, "SubImporters");
+  }
 
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
-      SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
+  // restrict communicator?
+  bool bRestrictComm      = false;
+  const ParameterList &pL = GetParameterList();
+  if (pL.get<bool>("repartition: use subcommunicators") == true)
+    bRestrictComm = true;
+
+  RCP<ParameterList> XpetraList = Teuchos::rcp(new ParameterList());
+  if (bRestrictComm)
+    XpetraList->set("Restrict Communicator", true);
+  else
+    XpetraList->set("Restrict Communicator", false);
+
+  // communicator for final (rebalanced) operator.
+  // If communicator is not restricted it should be identical to the communicator in bA
+  RCP<const Teuchos::Comm<int> > rebalancedComm = Teuchos::null;
+
+  // loop through all blocks and rebalance blocks
+  // Note: so far we do not support rebalancing of nested operators
+  //       TODO add a check for this
+  for (size_t i = 0; i < bA->Rows(); i++) {
+    for (size_t j = 0; j < bA->Cols(); j++) {
+      // extract matrix block
+      RCP<Matrix> Aij = bA->getMatrix(i, j);
+
+      std::stringstream ss;
+      ss << "Rebalancing matrix block A(" << i << "," << j << ")";
+      SubFactoryMonitor subM(*this, ss.str(), coarseLevel);
+
+      RCP<Matrix> rebAij = Teuchos::null;
+      // General rebalancing
+      if (importers[i] != Teuchos::null &&
+          importers[j] != Teuchos::null &&
+          Aij != Teuchos::null) {
+        RCP<const Map> targetRangeMap  = importers[i]->getTargetMap();
+        RCP<const Map> targetDomainMap = importers[j]->getTargetMap();
+
+        // Copy the block Aij
+        // TAW: Do we need a copy or can we do in-place rebalancing?
+        //      If we do in-place rebalancing the original distribution is lost
+        //      We don't really need it any more, though.
+        // RCP<Matrix> cAij = MatrixFactory::BuildCopy(Aij);
+        RCP<Matrix> cAij = Aij;  // do not copy the matrix data (just an rcp pointer)
+
+        // create a new importer for column map needed for rebalanced Aij
+        Teuchos::RCP<const Import> rebAijImport = ImportFactory::Build(importers[j]->getTargetMap(), cAij->getColMap());
+        TEUCHOS_TEST_FOR_EXCEPTION(rebAijImport.is_null() == true, Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Importer associated with block " << j << " is null.");
+
+        Teuchos::RCP<const CrsMatrixWrap> cAwij = Teuchos::rcp_dynamic_cast<const CrsMatrixWrap>(cAij);
+        TEUCHOS_TEST_FOR_EXCEPTION(cAwij.is_null() == true, Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Block (" << i << "," << j << ") is not of type CrsMatrix. We cannot rebalanced (nested) operators.");
+        Teuchos::RCP<CrsMatrix> cAmij = cAwij->getCrsMatrix();
+
+        // change domain map to rebalanced domain map (in-place). Update the importer to represent the column map
+        // cAmij->replaceDomainMapAndImporter(importers[j]->getTargetMap(),rebAijImport);
+
+        // rebalance rows of matrix block. Don't change the domain map (-> Teuchos::null)
+        // NOTE: If the communicator is restricted away, Build returns Teuchos::null.
+        rebAij = MatrixFactory::Build(cAij, *(importers[i]), *(importers[j]), targetDomainMap, targetRangeMap, XpetraList);
+      }  // rebalance matrix block A(i,i)
+      else {
+        rebAij = Aij;  // no rebalancing or empty block!
+      }
 
-      coarseLevel.DeclareInput("Importer",(*it)->GetFactory("Importer").get(), this);
-    }
+      // store new block in output
+      subBlockRebA[i * bA->Cols() + j] = rebAij;
 
-    // Use the non-manager path if the maps / importers are generated in one place
-    if(FactManager_.size() == 0) {
-      Input(coarseLevel,"SubImporters");
-    }
+      if (!rebAij.is_null()) {
+        // store communicator
+        if (rebalancedComm.is_null()) rebalancedComm = rebAij->getRowMap()->getComm();
 
-  }
+        // printout rebalancing information
+        RCP<ParameterList> params = rcp(new ParameterList());
+        params->set("printLoadBalancingInfo", true);
+        std::stringstream ss2;
+        ss2 << "A(" << i << "," << j << ") rebalanced:";
+        GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*rebAij, ss2.str(), params);
+      }
+    }  // loop over columns j
+
+    // fix striding information of diagonal blocks
+    // Note: we do not care about the off-diagonal blocks. We just make sure, that the
+    // diagonal blocks have the corresponding striding information from the map extractors
+    // Note: the diagonal block never should be zero.
+    // TODO what if a diagonal block is Teuchos::null?
+    if (subBlockRebA[i * bA->Cols() + i].is_null() == false) {
+      RCP<Matrix> rebAii                               = subBlockRebA[i * bA->Cols() + i];
+      Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(i, rangeMapExtractor->getThyraMode()));
+      Teuchos::RCP<const Map> stridedRgMap             = Teuchos::null;
+      if (orig_stridedRgMap != Teuchos::null) {
+        std::vector<size_t> stridingData                       = orig_stridedRgMap->getStridingData();
+        Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMapii = rebAii->getRangeMap()->getLocalElementList();
+        stridedRgMap                                           = StridedMapFactory::Build(
+                                                      bA->getRangeMap()->lib(),
+                                                      Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                      nodeRangeMapii,
+                                                      rebAii->getRangeMap()->getIndexBase(),
+                                                      stridingData,
+                                                      rebalancedComm,
+                                                      orig_stridedRgMap->getStridedBlockId(),
+                                                      orig_stridedRgMap->getOffset());
+      }
+      Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(i, domainMapExtractor->getThyraMode()));
+      Teuchos::RCP<const Map> stridedDoMap             = Teuchos::null;
+      if (orig_stridedDoMap != Teuchos::null) {
+        std::vector<size_t> stridingData                        = orig_stridedDoMap->getStridingData();
+        Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMapii = rebAii->getDomainMap()->getLocalElementList();
+        stridedDoMap                                            = StridedMapFactory::Build(
+                                                       bA->getDomainMap()->lib(),
+                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                       nodeDomainMapii,
+                                                       rebAii->getDomainMap()->getIndexBase(),
+                                                       stridingData,
+                                                       rebalancedComm,
+                                                       orig_stridedDoMap->getStridedBlockId(),
+                                                       orig_stridedDoMap->getOffset());
+      }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Computing blocked Ac", coarseLevel);
-
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-
-    RCP<Matrix> originalAc = Get< RCP<Matrix> >(coarseLevel, "A");
-
-    RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(originalAc);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA==Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockAcFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != bA->Cols(), Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Blocked operator has " << bA->Rows() << " and " << bA->Cols() << ". We only support square matrices (with same number of blocks and columns).");
-
-    // Variables to set up map extractors for blocked operators
-    std::vector<GO> fullRangeMapVector;
-    std::vector<GO> fullDomainMapVector;
-    std::vector<RCP<const Map> > subBlockARangeMaps;
-    std::vector<RCP<const Map> > subBlockADomainMaps;
-    subBlockARangeMaps.reserve(bA->Rows());
-    subBlockADomainMaps.reserve(bA->Cols());
-
-    // store map extractors
-    RCP<const MapExtractor> rangeMapExtractor = bA->getRangeMapExtractor();
-    RCP<const MapExtractor> domainMapExtractor = bA->getDomainMapExtractor();
-
-    // check if GIDs for full maps have to be sorted:
-    // For the Thyra mode ordering they do not have to be sorted since the GIDs are
-    // numbered as 0...n1,0...,n2 (starting with zero for each subblock). The MapExtractor
-    // generates unique GIDs during the construction.
-    // For Xpetra style, the GIDs have to be reordered. Such that one obtains a ordered
-    // list of GIDs in an increasing ordering. In Xpetra, the GIDs are all unique through
-    // out all submaps.
-    bool bThyraRangeGIDs  = rangeMapExtractor->getThyraMode();
-    bool bThyraDomainGIDs = domainMapExtractor->getThyraMode();
-
-    // vector containing rebalanced blocks (for final output)
-    std::vector<RCP<Matrix> > subBlockRebA =
-        std::vector<RCP<Matrix> >(bA->Cols() * bA->Rows(), Teuchos::null);
-
-    // vector with Import objects from the different
-    // RepartitionFactory instances
-    std::vector<RCP<const Import> > importers = std::vector<RCP<const Import> >(bA->Rows(), Teuchos::null);
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    size_t idx = 0;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
-      SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
-
-      RCP<const Import> rebalanceImporter = coarseLevel.Get<RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
-      importers[idx] = rebalanceImporter;
-      idx++;
-    }
-    if(FactManager_.size() == 0) {
-      importers = Get<std::vector<RCP<const Import> > >(coarseLevel,"SubImporters");
-    }
+      if (bRestrictComm) {
+        stridedRgMap->removeEmptyProcesses();
+        stridedDoMap->removeEmptyProcesses();
+      }
 
+      TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: failed to generate striding information. error.");
+      TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: failed to generate striding information. error.");
+
+      // replace stridedMaps view in diagonal sub block
+      if (rebAii->IsView("stridedMaps")) rebAii->RemoveView("stridedMaps");
+      rebAii->CreateView("stridedMaps", stridedRgMap, stridedDoMap);
+      // collect Xpetra-based global row ids for map extractors
+      subBlockARangeMaps.push_back(rebAii->getRowMap("stridedMaps"));
+      Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMap = rebAii->getRangeMap()->getLocalElementList();
+      // append the GIDs in the end. Do not sort if we have Thyra style GIDs
+      fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMap.begin(), nodeRangeMap.end());
+      if (bThyraRangeGIDs == false)
+        sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
+
+      subBlockADomainMaps.push_back(rebAii->getColMap("stridedMaps"));
+      Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMap = rebAii->getDomainMap()->getLocalElementList();
+      // append the GIDs in the end. Do not sort if we have Thyra style GIDs
+      fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMap.begin(), nodeDomainMap.end());
+      if (bThyraDomainGIDs == false)
+        sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
+    }  // end if rebAii != Teuchos::null
+  }    // loop over rows i
+
+  // all sub blocks are rebalanced (if available)
+
+  // Short cut if this processor is not in the list of active processors
+  if (rebalancedComm == Teuchos::null) {
+    GetOStream(Debug, -1) << "RebalanceBlockedAc: deactivate proc " << originalAc->getRowMap()->getComm()->getRank() << std::endl;
+    // TAW: it is important that we create a dummy object of type BlockedCrsMatrix (even if we set it to Teuchos::null)
+    Teuchos::RCP<BlockedCrsMatrix> reb_bA = Teuchos::null;
+    coarseLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(reb_bA), this);
+    return;
+  }
 
-    // restrict communicator?
-    bool bRestrictComm = false;
-    const ParameterList& pL = GetParameterList();
-    if (pL.get<bool>("repartition: use subcommunicators") == true)
-      bRestrictComm = true;
-
-    RCP<ParameterList> XpetraList = Teuchos::rcp(new ParameterList());
-    if(bRestrictComm)
-      XpetraList->set("Restrict Communicator",true);
-    else
-      XpetraList->set("Restrict Communicator",false);
-
-    // communicator for final (rebalanced) operator.
-    // If communicator is not restricted it should be identical to the communicator in bA
-    RCP<const Teuchos::Comm<int> > rebalancedComm = Teuchos::null;
-
-    // loop through all blocks and rebalance blocks
-    // Note: so far we do not support rebalancing of nested operators
-    //       TODO add a check for this
-    for(size_t i=0; i<bA->Rows(); i++) {
-      for(size_t j=0; j<bA->Cols(); j++) {
-        // extract matrix block
-        RCP<Matrix> Aij = bA->getMatrix(i, j);
-
-        std::stringstream ss; ss << "Rebalancing matrix block A(" << i << "," << j << ")";
-        SubFactoryMonitor subM(*this, ss.str(), coarseLevel);
-
-        RCP<Matrix> rebAij = Teuchos::null;
-        // General rebalancing
-        if( importers[i] != Teuchos::null &&
-            importers[j] != Teuchos::null &&
-            Aij != Teuchos::null) {
-          RCP<const Map> targetRangeMap  = importers[i]->getTargetMap();
-          RCP<const Map> targetDomainMap = importers[j]->getTargetMap();
-
-          // Copy the block Aij
-          // TAW: Do we need a copy or can we do in-place rebalancing?
-          //      If we do in-place rebalancing the original distribution is lost
-          //      We don't really need it any more, though.
-          //RCP<Matrix> cAij = MatrixFactory::BuildCopy(Aij);
-          RCP<Matrix> cAij = Aij; // do not copy the matrix data (just an rcp pointer)
-
-          // create a new importer for column map needed for rebalanced Aij
-          Teuchos::RCP<const Import> rebAijImport = ImportFactory::Build(importers[j]->getTargetMap(),cAij->getColMap());
-          TEUCHOS_TEST_FOR_EXCEPTION(rebAijImport.is_null() == true,Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Importer associated with block " << j << " is null.");
-
-          Teuchos::RCP<const CrsMatrixWrap> cAwij = Teuchos::rcp_dynamic_cast<const CrsMatrixWrap>(cAij);
-          TEUCHOS_TEST_FOR_EXCEPTION(cAwij.is_null() == true,Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: Block (" << i << "," << j << ") is not of type CrsMatrix. We cannot rebalanced (nested) operators.");
-          Teuchos::RCP<CrsMatrix> cAmij = cAwij->getCrsMatrix();
-
-          // change domain map to rebalanced domain map (in-place). Update the importer to represent the column map
-          //cAmij->replaceDomainMapAndImporter(importers[j]->getTargetMap(),rebAijImport);
-
-          // rebalance rows of matrix block. Don't change the domain map (-> Teuchos::null)
-          // NOTE: If the communicator is restricted away, Build returns Teuchos::null.
-          rebAij = MatrixFactory::Build(cAij, *(importers[i]), *(importers[j]), targetDomainMap, targetRangeMap, XpetraList);
-        }  // rebalance matrix block A(i,i)
-        else {
-          rebAij = Aij; // no rebalancing or empty block!
-        }
-
-        // store new block in output
-        subBlockRebA[i*bA->Cols() + j] = rebAij;
-
-        if (!rebAij.is_null()) {
-          // store communicator
-          if(rebalancedComm.is_null()) rebalancedComm = rebAij->getRowMap()->getComm();
-
-          // printout rebalancing information
-          RCP<ParameterList> params = rcp(new ParameterList());
-          params->set("printLoadBalancingInfo", true);
-          std::stringstream ss2; ss2 << "A(" << i << "," << j << ") rebalanced:";
-          GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*rebAij, ss2.str(), params);
-        }
-      } // loop over columns j
-
-      // fix striding information of diagonal blocks
-      // Note: we do not care about the off-diagonal blocks. We just make sure, that the
-      // diagonal blocks have the corresponding striding information from the map extractors
-      // Note: the diagonal block never should be zero.
-      // TODO what if a diagonal block is Teuchos::null?
-      if ( subBlockRebA[i*bA->Cols() + i].is_null() == false ) {
-        RCP<Matrix> rebAii = subBlockRebA[i*bA->Cols() + i];
-        Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(i,rangeMapExtractor->getThyraMode()));
-        Teuchos::RCP<const Map> stridedRgMap = Teuchos::null;
-        if(orig_stridedRgMap != Teuchos::null) {
-          std::vector<size_t> stridingData = orig_stridedRgMap->getStridingData();
-          Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMapii = rebAii->getRangeMap()->getLocalElementList();
-          stridedRgMap = StridedMapFactory::Build(
-              bA->getRangeMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              nodeRangeMapii,
-              rebAii->getRangeMap()->getIndexBase(),
-              stridingData,
-              rebalancedComm,
-              orig_stridedRgMap->getStridedBlockId(),
-              orig_stridedRgMap->getOffset());
-        }
-        Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(i,domainMapExtractor->getThyraMode()));
-        Teuchos::RCP<const Map> stridedDoMap = Teuchos::null;
-        if(orig_stridedDoMap != Teuchos::null) {
-          std::vector<size_t> stridingData = orig_stridedDoMap->getStridingData();
-          Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMapii = rebAii->getDomainMap()->getLocalElementList();
-          stridedDoMap = StridedMapFactory::Build(
-              bA->getDomainMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              nodeDomainMapii,
-              rebAii->getDomainMap()->getIndexBase(),
-              stridingData,
-              rebalancedComm,
-              orig_stridedDoMap->getStridedBlockId(),
-              orig_stridedDoMap->getOffset());
-        }
-
-        if(bRestrictComm) {
-          stridedRgMap->removeEmptyProcesses();
-          stridedDoMap->removeEmptyProcesses();
-        }
-
-        TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: failed to generate striding information. error.");
-        TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockAcFactory::Build: failed to generate striding information. error.");
-
-        // replace stridedMaps view in diagonal sub block
-        if(rebAii->IsView("stridedMaps")) rebAii->RemoveView("stridedMaps");
-        rebAii->CreateView("stridedMaps", stridedRgMap, stridedDoMap);
-        // collect Xpetra-based global row ids for map extractors
-        subBlockARangeMaps.push_back(rebAii->getRowMap("stridedMaps"));
-        Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMap = rebAii->getRangeMap()->getLocalElementList();
-        // append the GIDs in the end. Do not sort if we have Thyra style GIDs
-        fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMap.begin(), nodeRangeMap.end());
-        if(bThyraRangeGIDs == false)
-          sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
-
-        subBlockADomainMaps.push_back(rebAii->getColMap("stridedMaps"));
-        Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMap = rebAii->getDomainMap()->getLocalElementList();
-        // append the GIDs in the end. Do not sort if we have Thyra style GIDs
-        fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMap.begin(), nodeDomainMap.end());
-        if(bThyraDomainGIDs == false)
-          sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
-      } // end if rebAii != Teuchos::null
-    } // loop over rows i
-
-    // all sub blocks are rebalanced (if available)
-
-    // Short cut if this processor is not in the list of active processors
-    if (rebalancedComm == Teuchos::null) {
-      GetOStream(Debug,-1) << "RebalanceBlockedAc: deactivate proc " << originalAc->getRowMap()->getComm()->getRank() << std::endl;
-      // TAW: it is important that we create a dummy object of type BlockedCrsMatrix (even if we set it to Teuchos::null)
-      Teuchos::RCP<BlockedCrsMatrix> reb_bA = Teuchos::null;
-      coarseLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(reb_bA), this);
-      return;
-    }
+  // now, subBlockRebA contains all rebalanced matrix blocks
+  // extract map index base from maps of blocked A
+  GO rangeIndexBase  = bA->getRangeMap()->getIndexBase();
+  GO domainIndexBase = bA->getDomainMap()->getIndexBase();
+
+  Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0, fullRangeMapVector.size());
+  Teuchos::RCP<const StridedMap> stridedRgFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getFullMap());
+  Teuchos::RCP<const Map> fullRangeMap            = Teuchos::null;
+  if (stridedRgFullMap != Teuchos::null) {
+    std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
+    fullRangeMap =
+        StridedMapFactory::Build(
+            bA->getRangeMap()->lib(),
+            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+            fullRangeMapGIDs,
+            rangeIndexBase,
+            stridedData,
+            rebalancedComm,
+            stridedRgFullMap->getStridedBlockId(),
+            stridedRgFullMap->getOffset());
+  } else {
+    fullRangeMap =
+        MapFactory::Build(
+            bA->getRangeMap()->lib(),
+            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+            fullRangeMapGIDs,
+            rangeIndexBase,
+            rebalancedComm);
+  }
+  Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0, fullDomainMapVector.size());
+
+  Teuchos::RCP<const StridedMap> stridedDoFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getFullMap());
+  Teuchos::RCP<const Map> fullDomainMap           = Teuchos::null;
+  if (stridedDoFullMap != Teuchos::null) {
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap == Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockedAc::Build: full map in domain map extractor has no striding information! error.");
+    std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
+    fullDomainMap =
+        StridedMapFactory::Build(
+            bA->getDomainMap()->lib(),
+            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+            fullDomainMapGIDs,
+            domainIndexBase,
+            stridedData2,
+            rebalancedComm,
+            stridedDoFullMap->getStridedBlockId(),
+            stridedDoFullMap->getOffset());
+  } else {
+    fullDomainMap =
+        MapFactory::Build(
+            bA->getDomainMap()->lib(),
+            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+            fullDomainMapGIDs,
+            domainIndexBase,
+            rebalancedComm);
+  }
 
-    // now, subBlockRebA contains all rebalanced matrix blocks
-    // extract map index base from maps of blocked A
-    GO rangeIndexBase  = bA->getRangeMap()->getIndexBase();
-    GO domainIndexBase = bA->getDomainMap()->getIndexBase();
-
-    Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0,fullRangeMapVector.size());
-    Teuchos::RCP<const StridedMap> stridedRgFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getFullMap());
-    Teuchos::RCP<const Map > fullRangeMap = Teuchos::null;
-    if(stridedRgFullMap != Teuchos::null) {
-      std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
-      fullRangeMap =
-          StridedMapFactory::Build(
-              bA->getRangeMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              fullRangeMapGIDs,
-              rangeIndexBase,
-              stridedData,
-              rebalancedComm,
-              stridedRgFullMap->getStridedBlockId(),
-              stridedRgFullMap->getOffset());
-    } else {
-      fullRangeMap =
-          MapFactory::Build(
-              bA->getRangeMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              fullRangeMapGIDs,
-              rangeIndexBase,
-              rebalancedComm);
-    }
-    Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0,fullDomainMapVector.size());
-
-    Teuchos::RCP<const StridedMap> stridedDoFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getFullMap());
-    Teuchos::RCP<const Map > fullDomainMap = Teuchos::null;
-    if(stridedDoFullMap != Teuchos::null) {
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap==Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockedAc::Build: full map in domain map extractor has no striding information! error.");
-      std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
-      fullDomainMap =
-          StridedMapFactory::Build(
-              bA->getDomainMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              fullDomainMapGIDs,
-              domainIndexBase,
-              stridedData2,
-              rebalancedComm,
-              stridedDoFullMap->getStridedBlockId(),
-              stridedDoFullMap->getOffset());
-    } else {
-
-      fullDomainMap =
-          MapFactory::Build(
-              bA->getDomainMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              fullDomainMapGIDs,
-              domainIndexBase,
-              rebalancedComm);
-    }
+  if (bRestrictComm) {
+    fullRangeMap->removeEmptyProcesses();
+    fullDomainMap->removeEmptyProcesses();
+  }
 
-    if(bRestrictComm) {
-      fullRangeMap->removeEmptyProcesses();
-      fullDomainMap->removeEmptyProcesses();
+  // build map extractors
+  RCP<const MapExtractor> rebRangeMapExtractor  = MapExtractorFactory::Build(fullRangeMap, subBlockARangeMaps, bThyraRangeGIDs);
+  RCP<const MapExtractor> rebDomainMapExtractor = MapExtractorFactory::Build(fullDomainMap, subBlockADomainMaps, bThyraDomainGIDs);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(rangeMapExtractor->NumMaps() != rebRangeMapExtractor->NumMaps(), Exceptions::RuntimeError,
+                             "MueLu::RebalanceBlockedAc::Build: Rebalanced RangeMapExtractor has " << rebRangeMapExtractor->NumMaps()
+                                                                                                   << " sub maps. Original RangeMapExtractor has " << rangeMapExtractor->NumMaps() << ". They must match!");
+  TEUCHOS_TEST_FOR_EXCEPTION(domainMapExtractor->NumMaps() != rebDomainMapExtractor->NumMaps(), Exceptions::RuntimeError,
+                             "MueLu::RebalanceBlockedAc::Build: Rebalanced DomainMapExtractor has " << rebDomainMapExtractor->NumMaps()
+                                                                                                    << " sub maps. Original DomainMapExtractor has " << domainMapExtractor->NumMaps() << ". They must match!");
+
+  Teuchos::RCP<BlockedCrsMatrix> reb_bA = Teuchos::rcp(new BlockedCrsMatrix(rebRangeMapExtractor, rebDomainMapExtractor, 10));
+  for (size_t i = 0; i < bA->Rows(); i++) {
+    for (size_t j = 0; j < bA->Cols(); j++) {
+      reb_bA->setMatrix(i, j, subBlockRebA[i * bA->Cols() + j]);
     }
+  }
 
-    // build map extractors
-    RCP<const MapExtractor> rebRangeMapExtractor = MapExtractorFactory::Build(fullRangeMap, subBlockARangeMaps, bThyraRangeGIDs);
-    RCP<const MapExtractor> rebDomainMapExtractor = MapExtractorFactory::Build(fullDomainMap, subBlockADomainMaps, bThyraDomainGIDs);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(rangeMapExtractor->NumMaps() != rebRangeMapExtractor->NumMaps(), Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockedAc::Build: Rebalanced RangeMapExtractor has " << rebRangeMapExtractor->NumMaps()
-        << " sub maps. Original RangeMapExtractor has " << rangeMapExtractor->NumMaps() << ". They must match!");
-    TEUCHOS_TEST_FOR_EXCEPTION(domainMapExtractor->NumMaps() != rebDomainMapExtractor->NumMaps(), Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockedAc::Build: Rebalanced DomainMapExtractor has " << rebDomainMapExtractor->NumMaps()
-        << " sub maps. Original DomainMapExtractor has " << domainMapExtractor->NumMaps() << ". They must match!");
-
-    Teuchos::RCP<BlockedCrsMatrix> reb_bA = Teuchos::rcp(new BlockedCrsMatrix(rebRangeMapExtractor,rebDomainMapExtractor,10));
-    for(size_t i=0; i<bA->Rows(); i++) {
-      for(size_t j=0; j<bA->Cols(); j++) {
-       reb_bA->setMatrix(i,j,subBlockRebA[i*bA->Cols() + j]);
-      }
-    }
+  reb_bA->fillComplete();
 
-    reb_bA->fillComplete();
+  coarseLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(reb_bA), this);
+  // rebalance additional data:
+  // be aware, that we just call the rebalance factories without switching to local
+  // factory managers, i.e. the rebalance factories have to be defined with the appropriate
+  // factories by the user!
+  if (rebalanceFacts_.begin() != rebalanceFacts_.end()) {
+    SubFactoryMonitor m2(*this, "Rebalance additional data", coarseLevel);
 
-    coarseLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(reb_bA), this);
-    // rebalance additional data:
-    // be aware, that we just call the rebalance factories without switching to local
-    // factory managers, i.e. the rebalance factories have to be defined with the appropriate
-    // factories by the user!
-    if (rebalanceFacts_.begin() != rebalanceFacts_.end()) {
-      SubFactoryMonitor m2(*this, "Rebalance additional data", coarseLevel);
-
-      // call Build of all user-given transfer factories
-      for (std::vector<RCP<const FactoryBase> >::const_iterator it2 = rebalanceFacts_.begin(); it2 != rebalanceFacts_.end(); ++it2) {
-        GetOStream(Runtime0) << "RebalanceBlockedAc: call rebalance factory " << (*it2).get() << ": " << (*it2)->description() << std::endl;
-        (*it2)->CallBuild(coarseLevel);
-      }
+    // call Build of all user-given transfer factories
+    for (std::vector<RCP<const FactoryBase> >::const_iterator it2 = rebalanceFacts_.begin(); it2 != rebalanceFacts_.end(); ++it2) {
+      GetOStream(Runtime0) << "RebalanceBlockedAc: call rebalance factory " << (*it2).get() << ": " << (*it2)->description() << std::endl;
+      (*it2)->CallBuild(coarseLevel);
     }
-  } //Build()
+  }
+}  // Build()
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddRebalanceFactory(const RCP<const FactoryBase>& factory) {
-    rebalanceFacts_.push_back(factory);
-  } //AddRebalanceFactory()
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceBlockAcFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddRebalanceFactory(const RCP<const FactoryBase> &factory) {
+  rebalanceFacts_.push_back(factory);
+}  // AddRebalanceFactory()
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_REBALANCEBLOCKACFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_decl.hpp
index 5309f6b4114f..2a38058ebf42 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_decl.hpp
@@ -64,64 +64,64 @@
 
 namespace MueLu {
 
-  /*!
-    @class RebalanceBlockInterpolationFactory class.
-    @brief Applies permutation to prolongation operators.
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceBlockInterpolationFactory : public TwoLevelFactoryBase {
+/*!
+  @class RebalanceBlockInterpolationFactory class.
+  @brief Applies permutation to prolongation operators.
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceBlockInterpolationFactory : public TwoLevelFactoryBase {
 #undef MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RebalanceBlockInterpolationFactory() { }
+  //! Constructor.
+  RebalanceBlockInterpolationFactory() {}
 
-    //! Destructor.
-    virtual ~RebalanceBlockInterpolationFactory() { }
+  //! Destructor.
+  virtual ~RebalanceBlockInterpolationFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //! Add a factory manager
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
+  //! Add a factory manager
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    //! Input factories
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+ private:
+  //! Input factories
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-  }; // class RebalanceBlockTransferFactory
+};  // class RebalanceBlockTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_SHORT
 
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_def.hpp
index eb1f31f06d7b..b0b08ff16938 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockInterpolationFactory_def.hpp
@@ -75,10 +75,10 @@ template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const ParameterList> RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
-  validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Factory for generating the non-rebalanced coarse level A. We need this to make sure the non-rebalanced coarse A is calculated first before rebalancing takes place.");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory for generating the non-rebalanced coarse level A. We need this to make sure the non-rebalanced coarse A is calculated first before rebalancing takes place.");
 
-  validParamList->set< RCP<const FactoryBase> >("Coordinates",    Teuchos::null, "Factory for generating the non-rebalanced Coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for generating the non-rebalanced Coordinates.");
   validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the matrix Importer for rebalancing");
   validParamList->set<RCP<const FactoryBase> >("SubImporters", Teuchos::null, "Generating factory of the matrix sub-Importers for rebalancing");
 
@@ -93,7 +93,7 @@ RCP<const ParameterList> RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal
   return validParamList;
 }
 
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager) {
   FactManager_.push_back(FactManager);
 }
@@ -101,49 +101,47 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
   Input(coarseLevel, "P");
-  Input(coarseLevel, "A"); // we request the non-rebalanced coarse level A since we have to make sure it is calculated before rebalancing starts!
+  Input(coarseLevel, "A");  // we request the non-rebalanced coarse level A since we have to make sure it is calculated before rebalancing starts!
 
   std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-  for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-    SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
     SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
 
     // Request Importer and Coordinates (if defined in xml file)
     // Note, that we have to use the Level::DeclareInput routine in order to use the FactoryManager *it (rather than the main factory manager)
-    coarseLevel.DeclareInput("Importer",(*it)->GetFactory("Importer").get(), this);
-    if((*it)->hasFactory("Coordinates") == true)
-      coarseLevel.DeclareInput("Coordinates",(*it)->GetFactory("Coordinates").get(), this);
+    coarseLevel.DeclareInput("Importer", (*it)->GetFactory("Importer").get(), this);
+    if ((*it)->hasFactory("Coordinates") == true)
+      coarseLevel.DeclareInput("Coordinates", (*it)->GetFactory("Coordinates").get(), this);
   }
 
   // Use the non-manager path if the maps / importers are generated in one place
-  if(FactManager_.size() == 0) {
-    Input(coarseLevel,"Importer");
-    Input(coarseLevel,"SubImporters");
-    Input(coarseLevel,"Coordinates");
+  if (FactManager_.size() == 0) {
+    Input(coarseLevel, "Importer");
+    Input(coarseLevel, "SubImporters");
+    Input(coarseLevel, "Coordinates");
   }
-
-
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
   FactoryMonitor m(*this, "Build", coarseLevel);
   typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
-  typedef Xpetra::BlockedMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> xdBV;
+  typedef Xpetra::BlockedMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdBV;
 
   bool UseSingleSource = FactManager_.size() == 0;
-  //RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  // RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
-  Teuchos::RCP<Matrix> nonrebCoarseA = Get< RCP<Matrix> >(coarseLevel, "A");
+  Teuchos::RCP<Matrix> nonrebCoarseA = Get<RCP<Matrix> >(coarseLevel, "A");
 
   Teuchos::RCP<Matrix> originalTransferOp = Teuchos::null;
-  originalTransferOp = Get< RCP<Matrix> >(coarseLevel, "P");
+  originalTransferOp                      = Get<RCP<Matrix> >(coarseLevel, "P");
 
   RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOriginalTransferOp =
-    Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > (originalTransferOp);
-  TEUCHOS_TEST_FOR_EXCEPTION(bOriginalTransferOp==Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: input matrix P or R is not of type BlockedCrsMatrix! error.");
+      Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(originalTransferOp);
+  TEUCHOS_TEST_FOR_EXCEPTION(bOriginalTransferOp == Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: input matrix P or R is not of type BlockedCrsMatrix! error.");
 
-  RCP<const MapExtractor> rangeMapExtractor = bOriginalTransferOp->getRangeMapExtractor();
+  RCP<const MapExtractor> rangeMapExtractor  = bOriginalTransferOp->getRangeMapExtractor();
   RCP<const MapExtractor> domainMapExtractor = bOriginalTransferOp->getDomainMapExtractor();
 
   // check if GIDs for full maps have to be sorted:
@@ -157,12 +155,12 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
   bool bThyraDomainGIDs = domainMapExtractor->getThyraMode();
 
   // declare variables for maps of blocked rebalanced prolongation operator
-  std::vector<GO> fullRangeMapVector;                             // contains all range GIDs on current processor
-  std::vector<GO> fullDomainMapVector;                            // contains all domain GIDs on current processor
+  std::vector<GO> fullRangeMapVector;   // contains all range GIDs on current processor
+  std::vector<GO> fullDomainMapVector;  // contains all domain GIDs on current processor
   std::vector<RCP<const Map> > subBlockPRangeMaps;
   std::vector<RCP<const Map> > subBlockPDomainMaps;
-  subBlockPRangeMaps.reserve(bOriginalTransferOp->Rows());        // reserve size for block P operators
-  subBlockPDomainMaps.reserve(bOriginalTransferOp->Cols());       // reserve size for block P operators
+  subBlockPRangeMaps.reserve(bOriginalTransferOp->Rows());   // reserve size for block P operators
+  subBlockPDomainMaps.reserve(bOriginalTransferOp->Cols());  // reserve size for block P operators
 
   std::vector<Teuchos::RCP<Matrix> > subBlockRebP;
   subBlockRebP.reserve(bOriginalTransferOp->Rows());
@@ -171,38 +169,41 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
   std::vector<RCP<const Import> > importers = std::vector<RCP<const Import> >(bOriginalTransferOp->Rows(), Teuchos::null);
   std::vector<RCP<xdMV> > newCoordinates(bOriginalTransferOp->Rows());
   RCP<xdBV> oldCoordinates;
-  if(UseSingleSource) {
-    importers = Get<std::vector<RCP<const Import> > >(coarseLevel,"SubImporters");
-    oldCoordinates = Get<RCP<xdBV> >(coarseLevel,"Coordinates");
+  if (UseSingleSource) {
+    importers      = Get<std::vector<RCP<const Import> > >(coarseLevel, "SubImporters");
+    oldCoordinates = Get<RCP<xdBV> >(coarseLevel, "Coordinates");
   }
 
-  int curBlockId = 0;
+  int curBlockId                               = 0;
   Teuchos::RCP<const Import> rebalanceImporter = Teuchos::null;
   std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
   for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
     // begin SubFactoryManager environment
-    SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
     SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
 
     // TAW: use the Level::Get routine in order to access the data declared in (*it) factory manager (rather than the main factory manager)
-    if(UseSingleSource) rebalanceImporter = importers[curBlockId];
-    else rebalanceImporter = coarseLevel.Get<Teuchos::RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
+    if (UseSingleSource)
+      rebalanceImporter = importers[curBlockId];
+    else
+      rebalanceImporter = coarseLevel.Get<Teuchos::RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
 
     // extract diagonal matrix block
-    Teuchos::RCP<Matrix> Pii = bOriginalTransferOp->getMatrix(curBlockId, curBlockId);
+    Teuchos::RCP<Matrix> Pii         = bOriginalTransferOp->getMatrix(curBlockId, curBlockId);
     Teuchos::RCP<CrsMatrixWrap> Pwii = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(Pii);
-    TEUCHOS_TEST_FOR_EXCEPTION(Pwii == Teuchos::null,Xpetra::Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: block " << curBlockId << " is not of type CrsMatrixWrap. We need an underlying CsrMatrix to replace domain map and importer!");
+    TEUCHOS_TEST_FOR_EXCEPTION(Pwii == Teuchos::null, Xpetra::Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: block " << curBlockId << " is not of type CrsMatrixWrap. We need an underlying CsrMatrix to replace domain map and importer!");
 
     MUELU_TEST_FOR_EXCEPTION(bThyraRangeGIDs == true && Pii->getRowMap()->getMinAllGlobalIndex() != 0,
-        Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockInterpolationFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block range " << curBlockId << " start with " << Pii->getRowMap()->getMinAllGlobalIndex() << " but should start with 0");
+                             Exceptions::RuntimeError,
+                             "MueLu::RebalanceBlockInterpolationFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block range " << curBlockId << " start with " << Pii->getRowMap()->getMinAllGlobalIndex() << " but should start with 0");
     MUELU_TEST_FOR_EXCEPTION(bThyraDomainGIDs == true && Pii->getColMap()->getMinAllGlobalIndex() != 0,
-        Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockInterpolationFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block domain " << curBlockId << " start with " << Pii->getColMap()->getMinAllGlobalIndex() << " but should start with 0");
+                             Exceptions::RuntimeError,
+                             "MueLu::RebalanceBlockInterpolationFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block domain " << curBlockId << " start with " << Pii->getColMap()->getMinAllGlobalIndex() << " but should start with 0");
 
     // rebalance P11
-    if(rebalanceImporter != Teuchos::null) {
-      std::stringstream ss; ss << "Rebalancing prolongator block P(" << curBlockId << "," << curBlockId << ")";
+    if (rebalanceImporter != Teuchos::null) {
+      std::stringstream ss;
+      ss << "Rebalancing prolongator block P(" << curBlockId << "," << curBlockId << ")";
       SubFactoryMonitor m1(*this, ss.str(), coarseLevel);
 
       // P is the transfer operator from the coarse grid to the fine grid.
@@ -225,7 +226,8 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
 
       RCP<ParameterList> params = rcp(new ParameterList());
       params->set("printLoadBalancingInfo", true);
-      std::stringstream ss2; ss2 << "P(" << curBlockId << "," << curBlockId << ") rebalanced:";
+      std::stringstream ss2;
+      ss2 << "P(" << curBlockId << "," << curBlockId << ") rebalanced:";
       GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*Pii, ss2.str(), params);
 
       // store rebalanced P block
@@ -234,19 +236,18 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
       // rebalance coordinates
       // TAW: Note, that each sub-block manager overwrites the Coordinates. So far we only support one set of Coordinates
       //      for a multiphysics problem (i.e., we only support volume coupled problems with the same mesh)
-      if(UseSingleSource)  {
-	RCP<xdMV> localCoords = oldCoordinates->getMultiVector(curBlockId);
+      if (UseSingleSource) {
+        RCP<xdMV> localCoords = oldCoordinates->getMultiVector(curBlockId);
 
-	// FIXME: This should be extended to work with blocking
-	RCP<const Import> coordImporter = rebalanceImporter;
+        // FIXME: This should be extended to work with blocking
+        RCP<const Import> coordImporter = rebalanceImporter;
 
-	RCP<xdMV> permutedLocalCoords  = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(coordImporter->getTargetMap(), localCoords->getNumVectors());
+        RCP<xdMV> permutedLocalCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(coordImporter->getTargetMap(), localCoords->getNumVectors());
         permutedLocalCoords->doImport(*localCoords, *coordImporter, Xpetra::INSERT);
 
-	newCoordinates[curBlockId] = permutedLocalCoords;
-      }
-      else if ( (*it)->hasFactory("Coordinates") == true && coarseLevel.IsAvailable("Coordinates",(*it)->GetFactory("Coordinates").get()) == true) {
-        RCP<xdMV> coords = coarseLevel.Get< RCP<xdMV> >("Coordinates",(*it)->GetFactory("Coordinates").get());
+        newCoordinates[curBlockId] = permutedLocalCoords;
+      } else if ((*it)->hasFactory("Coordinates") == true && coarseLevel.IsAvailable("Coordinates", (*it)->GetFactory("Coordinates").get()) == true) {
+        RCP<xdMV> coords = coarseLevel.Get<RCP<xdMV> >("Coordinates", (*it)->GetFactory("Coordinates").get());
 
         // This line must be after the Get call
         SubFactoryMonitor subM(*this, "Rebalancing coordinates", coarseLevel);
@@ -258,11 +259,11 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
 
         if (nodeNumElts > 0) {
           MUELU_TEST_FOR_EXCEPTION(rebalanceImporter->getSourceMap()->getLocalNumElements() % nodeNumElts != 0,
-                          Exceptions::RuntimeError,
-                          "MueLu::RebalanceBlockInterpolationFactory::Build: block size. " << rebalanceImporter->getSourceMap()->getLocalNumElements() << " not divisable by " << nodeNumElts);
+                                   Exceptions::RuntimeError,
+                                   "MueLu::RebalanceBlockInterpolationFactory::Build: block size. " << rebalanceImporter->getSourceMap()->getLocalNumElements() << " not divisable by " << nodeNumElts);
           myBlkSize = rebalanceImporter->getSourceMap()->getLocalNumElements() / nodeNumElts;
         }
-      
+
         MueLu_maxAll(coords->getMap()->getComm(), myBlkSize, blkSize);
 
         RCP<const Import> coordImporter = Teuchos::null;
@@ -272,33 +273,34 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
           // NOTE: there is an implicit assumption here: we assume that dof any node are enumerated consequently
           // Proper fix would require using decomposition similar to how we construct importer in the
           // RepartitionFactory
-          RCP<const Map> origMap   = coords->getMap();
-          GO             indexBase = origMap->getIndexBase();
+          RCP<const Map> origMap = coords->getMap();
+          GO indexBase           = origMap->getIndexBase();
 
-          ArrayView<const GO> OEntries   = rebalanceImporter->getTargetMap()->getLocalElementList();
-          LO                  numEntries = OEntries.size()/blkSize;
+          ArrayView<const GO> OEntries = rebalanceImporter->getTargetMap()->getLocalElementList();
+          LO numEntries                = OEntries.size() / blkSize;
           ArrayRCP<GO> Entries(numEntries);
           for (LO i = 0; i < numEntries; i++)
-            Entries[i] = (OEntries[i*blkSize]-indexBase)/blkSize + indexBase;
+            Entries[i] = (OEntries[i * blkSize] - indexBase) / blkSize + indexBase;
 
           RCP<const Map> targetMap = MapFactory::Build(origMap->lib(), origMap->getGlobalNumElements(), Entries(), indexBase, origMap->getComm());
-          coordImporter = ImportFactory::Build(origMap, targetMap);
+          coordImporter            = ImportFactory::Build(origMap, targetMap);
         }
-      
-        RCP<xdMV> permutedCoords  = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(coordImporter->getTargetMap(), coords->getNumVectors());
+
+        RCP<xdMV> permutedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(coordImporter->getTargetMap(), coords->getNumVectors());
         permutedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
 
-        const ParameterList& pL = GetParameterList();
+        const ParameterList &pL = GetParameterList();
         if (pL.isParameter("repartition: use subcommunicators") == true && pL.get<bool>("repartition: use subcommunicators") == true)
           permutedCoords->replaceMap(permutedCoords->getMap()->removeEmptyProcesses());
 
         Set(coarseLevel, "Coordinates", permutedCoords);
       }
-    } // end rebalance P(1,1)
+    }  // end rebalance P(1,1)
     else {
       RCP<ParameterList> params = rcp(new ParameterList());
       params->set("printLoadBalancingInfo", true);
-      std::stringstream ss; ss << "P(" << curBlockId << "," << curBlockId << ") not rebalanced:";
+      std::stringstream ss;
+      ss << "P(" << curBlockId << "," << curBlockId << ") not rebalanced:";
       GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*Pii, ss.str(), params);
       // store rebalanced P block
       subBlockRebP.push_back(Pii);
@@ -306,84 +308,86 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
       // Store Coordinates on coarse level (generated by this)
       // TAW: Note, that each sub-block manager overwrites the Coordinates. So far we only support one set of Coordinates
       //      for a multiphysics problem (i.e., we only support volume coupled problems with the same mesh)
-      if((*it)->hasFactory("Coordinates") == true && coarseLevel.IsAvailable("Coordinates",(*it)->GetFactory("Coordinates").get()) == true) {
-        coarseLevel.Set("Coordinates", coarseLevel.Get< RCP<xdMV> >("Coordinates",(*it)->GetFactory("Coordinates").get()),this);
+      if ((*it)->hasFactory("Coordinates") == true && coarseLevel.IsAvailable("Coordinates", (*it)->GetFactory("Coordinates").get()) == true) {
+        coarseLevel.Set("Coordinates", coarseLevel.Get<RCP<xdMV> >("Coordinates", (*it)->GetFactory("Coordinates").get()), this);
       }
     }
-  
+
     // fix striding information for rebalanced diagonal block Pii
-    //RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rgPMapExtractor = bOriginalTransferOp->getRangeMapExtractor(); // original map extractor
-    Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(Teuchos::as<size_t>(curBlockId),rangeMapExtractor->getThyraMode()));
-    Teuchos::RCP<const Map> stridedRgMap = Teuchos::null;
-    if(orig_stridedRgMap != Teuchos::null) {
-      std::vector<size_t> stridingData = orig_stridedRgMap->getStridingData();
-      Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMapii = Pii->getRangeMap()->getLocalElementList();
-      stridedRgMap = StridedMapFactory::Build(
-          originalTransferOp->getRangeMap()->lib(),
-          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-          nodeRangeMapii,
-          Pii->getRangeMap()->getIndexBase(),
-          stridingData,
-          originalTransferOp->getRangeMap()->getComm(),
-          orig_stridedRgMap->getStridedBlockId(),
-          orig_stridedRgMap->getOffset());
-    } else stridedRgMap = Pii->getRangeMap();
-
-    //RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > doPMapExtractor = bOriginalTransferOp->getDomainMapExtractor(); // original map extractor
-    Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(Teuchos::as<size_t>(curBlockId),domainMapExtractor->getThyraMode()));
+    // RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rgPMapExtractor = bOriginalTransferOp->getRangeMapExtractor(); // original map extractor
+    Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(Teuchos::as<size_t>(curBlockId), rangeMapExtractor->getThyraMode()));
+    Teuchos::RCP<const Map> stridedRgMap             = Teuchos::null;
+    if (orig_stridedRgMap != Teuchos::null) {
+      std::vector<size_t> stridingData                       = orig_stridedRgMap->getStridingData();
+      Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMapii = Pii->getRangeMap()->getLocalElementList();
+      stridedRgMap                                           = StridedMapFactory::Build(
+                                                    originalTransferOp->getRangeMap()->lib(),
+                                                    Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                    nodeRangeMapii,
+                                                    Pii->getRangeMap()->getIndexBase(),
+                                                    stridingData,
+                                                    originalTransferOp->getRangeMap()->getComm(),
+                                                    orig_stridedRgMap->getStridedBlockId(),
+                                                    orig_stridedRgMap->getOffset());
+    } else
+      stridedRgMap = Pii->getRangeMap();
+
+    // RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > doPMapExtractor = bOriginalTransferOp->getDomainMapExtractor(); // original map extractor
+    Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(Teuchos::as<size_t>(curBlockId), domainMapExtractor->getThyraMode()));
 
     Teuchos::RCP<const Map> stridedDoMap = Teuchos::null;
-    if(orig_stridedDoMap != Teuchos::null) {
-      std::vector<size_t> stridingData = orig_stridedDoMap->getStridingData();
-      Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMapii = Pii->getDomainMap()->getLocalElementList();
-      stridedDoMap = StridedMapFactory::Build(originalTransferOp->getDomainMap()->lib(),
-        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-        nodeDomainMapii,
-        Pii->getDomainMap()->getIndexBase(),
-        stridingData,
-        originalTransferOp->getDomainMap()->getComm(),
-        orig_stridedDoMap->getStridedBlockId(),
-        orig_stridedDoMap->getOffset());
-    } else stridedDoMap = Pii->getDomainMap();
-
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockInterpolationFactory::Build: failed to generate striding information. error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockInterpolationFactory::Build: failed to generate striding information. error.");
+    if (orig_stridedDoMap != Teuchos::null) {
+      std::vector<size_t> stridingData                        = orig_stridedDoMap->getStridingData();
+      Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMapii = Pii->getDomainMap()->getLocalElementList();
+      stridedDoMap                                            = StridedMapFactory::Build(originalTransferOp->getDomainMap()->lib(),
+                                                                                         Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                                                         nodeDomainMapii,
+                                                                                         Pii->getDomainMap()->getIndexBase(),
+                                                                                         stridingData,
+                                                                                         originalTransferOp->getDomainMap()->getComm(),
+                                                                                         orig_stridedDoMap->getStridedBlockId(),
+                                                                                         orig_stridedDoMap->getOffset());
+    } else
+      stridedDoMap = Pii->getDomainMap();
+
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockInterpolationFactory::Build: failed to generate striding information. error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockInterpolationFactory::Build: failed to generate striding information. error.");
 
     // replace stridedMaps view in diagonal sub block
-    if(Pii->IsView("stridedMaps")) Pii->RemoveView("stridedMaps");
-      Pii->CreateView("stridedMaps", stridedRgMap, stridedDoMap);
+    if (Pii->IsView("stridedMaps")) Pii->RemoveView("stridedMaps");
+    Pii->CreateView("stridedMaps", stridedRgMap, stridedDoMap);
 
     // append strided row map (= range map) to list of range maps.
-    Teuchos::RCP<const Map> rangeMapii = Pii->getRowMap("stridedMaps"); // strided range map
+    Teuchos::RCP<const Map> rangeMapii = Pii->getRowMap("stridedMaps");  // strided range map
     subBlockPRangeMaps.push_back(rangeMapii);
-    Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMapii = Pii->getRangeMap()->getLocalElementList();
+    Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMapii = Pii->getRangeMap()->getLocalElementList();
     // append the GIDs in the end. Do not sort if we have Thyra style GIDs
     fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMapii.begin(), nodeRangeMapii.end());
-    if(bThyraRangeGIDs == false)
+    if (bThyraRangeGIDs == false)
       sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
 
     // append strided col map (= domain map) to list of range maps.
-    Teuchos::RCP<const Map> domainMapii = Pii->getColMap("stridedMaps"); // strided domain map
+    Teuchos::RCP<const Map> domainMapii = Pii->getColMap("stridedMaps");  // strided domain map
     subBlockPDomainMaps.push_back(domainMapii);
-    Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMapii = Pii->getDomainMap()->getLocalElementList();
+    Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMapii = Pii->getDomainMap()->getLocalElementList();
     // append the GIDs in the end. Do not sort if we have Thyra style GIDs
     fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMapii.begin(), nodeDomainMapii.end());
-    if(bThyraDomainGIDs == false)
+    if (bThyraDomainGIDs == false)
       sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
 
-    curBlockId++; // increase block id index
+    curBlockId++;  // increase block id index
 
-  } // end SubFactoryManager environment
+  }  // end SubFactoryManager environment
 
   // extract map index base from maps of blocked P
-  GO rangeIndexBase = originalTransferOp->getRangeMap()->getIndexBase();
-  GO domainIndexBase= originalTransferOp->getDomainMap()->getIndexBase();
+  GO rangeIndexBase  = originalTransferOp->getRangeMap()->getIndexBase();
+  GO domainIndexBase = originalTransferOp->getDomainMap()->getIndexBase();
 
-  RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rangePMapExtractor = bOriginalTransferOp->getRangeMapExtractor(); // original map extractor
-  Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0,fullRangeMapVector.size());
+  RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rangePMapExtractor = bOriginalTransferOp->getRangeMapExtractor();  // original map extractor
+  Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0, fullRangeMapVector.size());
   Teuchos::RCP<const StridedMap> stridedRgFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangePMapExtractor->getFullMap());
-  Teuchos::RCP<const Map > fullRangeMap = Teuchos::null;
-  if(stridedRgFullMap != Teuchos::null) {
+  Teuchos::RCP<const Map> fullRangeMap            = Teuchos::null;
+  if (stridedRgFullMap != Teuchos::null) {
     std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
     fullRangeMap =
         StridedMapFactory::Build(
@@ -406,11 +410,11 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
   }
 
   RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > domainAMapExtractor = bOriginalTransferOp->getDomainMapExtractor();
-  Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0,fullDomainMapVector.size());
+  Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0, fullDomainMapVector.size());
   Teuchos::RCP<const StridedMap> stridedDoFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainAMapExtractor->getFullMap());
-  Teuchos::RCP<const Map > fullDomainMap = Teuchos::null;
-  if(stridedDoFullMap != Teuchos::null) {
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap==Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information! error.");
+  Teuchos::RCP<const Map> fullDomainMap           = Teuchos::null;
+  if (stridedDoFullMap != Teuchos::null) {
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap == Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information! error.");
     std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
     fullDomainMap =
         StridedMapFactory::Build(
@@ -423,7 +427,6 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
             stridedDoFullMap->getStridedBlockId(),
             stridedDoFullMap->getOffset());
   } else {
-
     fullDomainMap =
         MapFactory::Build(
             originalTransferOp->getDomainMap()->lib(),
@@ -434,31 +437,29 @@ void RebalanceBlockInterpolationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Nod
   }
 
   // build map extractors
-  Teuchos::RCP<const MapExtractor> rebrangeMapExtractor  =
-      MapExtractorFactory::Build(fullRangeMap,  subBlockPRangeMaps, bThyraRangeGIDs);
+  Teuchos::RCP<const MapExtractor> rebrangeMapExtractor =
+      MapExtractorFactory::Build(fullRangeMap, subBlockPRangeMaps, bThyraRangeGIDs);
   Teuchos::RCP<const MapExtractor> rebdomainMapExtractor =
       MapExtractorFactory::Build(fullDomainMap, subBlockPDomainMaps, bThyraDomainGIDs);
 
-  Teuchos::RCP<BlockedCrsMatrix> bRebP = Teuchos::rcp(new BlockedCrsMatrix(rebrangeMapExtractor,rebdomainMapExtractor,10));
-  for(size_t i = 0; i<subBlockPRangeMaps.size(); i++) {
-     Teuchos::RCP<CrsMatrixWrap> crsOpii = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(subBlockRebP[i]);
-     TEUCHOS_TEST_FOR_EXCEPTION(crsOpii == Teuchos::null,Xpetra::Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: block P" << i << " is not of type CrsMatrixWrap.");
-     bRebP->setMatrix(i,i,crsOpii);
-   }
+  Teuchos::RCP<BlockedCrsMatrix> bRebP = Teuchos::rcp(new BlockedCrsMatrix(rebrangeMapExtractor, rebdomainMapExtractor, 10));
+  for (size_t i = 0; i < subBlockPRangeMaps.size(); i++) {
+    Teuchos::RCP<CrsMatrixWrap> crsOpii = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(subBlockRebP[i]);
+    TEUCHOS_TEST_FOR_EXCEPTION(crsOpii == Teuchos::null, Xpetra::Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: block P" << i << " is not of type CrsMatrixWrap.");
+    bRebP->setMatrix(i, i, crsOpii);
+  }
   bRebP->fillComplete();
 
   Set(coarseLevel, "P", Teuchos::rcp_dynamic_cast<Matrix>(bRebP));
 
-
   // Finish up the coordinates (single source only)
-  if(UseSingleSource)  {
-    RCP<xdBV> bcoarseCoords = rcp(new xdBV(rebrangeMapExtractor->getBlockedMap(),newCoordinates));
-    Set(coarseLevel,"Coordinates",bcoarseCoords);
+  if (UseSingleSource) {
+    RCP<xdBV> bcoarseCoords = rcp(new xdBV(rebrangeMapExtractor->getBlockedMap(), newCoordinates));
+    Set(coarseLevel, "Coordinates", bcoarseCoords);
   }
-  
 
-} // Build
+}  // Build
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_decl.hpp
index 3be167b8df6c..052348e53337 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_decl.hpp
@@ -65,70 +65,69 @@
 
 namespace MueLu {
 
-  /*!
-    @class RebalanceBlockRestrictionFactory class.
-    @brief Applies permutation to restriction operators.
-    @ingroup MueLuTransferClasses
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceBlockRestrictionFactory : public TwoLevelFactoryBase {
+/*!
+  @class RebalanceBlockRestrictionFactory class.
+  @brief Applies permutation to restriction operators.
+  @ingroup MueLuTransferClasses
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceBlockRestrictionFactory : public TwoLevelFactoryBase {
 #undef MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RebalanceBlockRestrictionFactory():UseSingleSourceImporters_(false) { }
+  //! Constructor.
+  RebalanceBlockRestrictionFactory()
+    : UseSingleSourceImporters_(false) {}
 
-    //! Destructor.
-    virtual ~RebalanceBlockRestrictionFactory() { }
+  //! Destructor.
+  virtual ~RebalanceBlockRestrictionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //! Add a factory manager
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
+  //! Add a factory manager
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
 
-    //! Get the importers from a single source
-    void SetSingleSourceImporters(bool value){UseSingleSourceImporters_=value;}
+  //! Get the importers from a single source
+  void SetSingleSourceImporters(bool value) { UseSingleSourceImporters_ = value; }
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  
-  private:
+ private:
+  //! Input factories
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-    //! Input factories
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
-    
-    bool UseSingleSourceImporters_;
+  bool UseSingleSourceImporters_;
 
-  }; // class RebalanceBlockTransferFactory
+};  // class RebalanceBlockTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_SHORT
 #endif /* MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_def.hpp
index bcb12f6fea54..0be0670f42c5 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceBlockRestrictionFactory_def.hpp
@@ -77,12 +77,12 @@ RCP<const ParameterList> RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal,
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: use subcommunicators");
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("repartition: use subcommunicators");
+#undef SET_VALID_ENTRY
 
-  //validParamList->set< RCP<const FactoryBase> >("repartition: use subcommunicators", Teuchos::null, "test");
+  // validParamList->set< RCP<const FactoryBase> >("repartition: use subcommunicators", Teuchos::null, "test");
 
-  validParamList->set< RCP<const FactoryBase> >("R", Teuchos::null, "Factory of the restriction operator that need to be rebalanced (only used if type=Restriction)");
+  validParamList->set<RCP<const FactoryBase> >("R", Teuchos::null, "Factory of the restriction operator that need to be rebalanced (only used if type=Restriction)");
   validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Generating factory of the matrix Importer for rebalancing");
   validParamList->set<RCP<const FactoryBase> >("SubImporters", Teuchos::null, "Generating factory of the matrix sub-Importers for rebalancing");
   validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the Nullspace operator");
@@ -90,7 +90,7 @@ RCP<const ParameterList> RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal,
   return validParamList;
 }
 
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager) {
   FactManager_.push_back(FactManager);
 }
@@ -100,43 +100,39 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
   Input(coarseLevel, "R");
 
   std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-  for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-    SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
     SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
 
-    if(!UseSingleSourceImporters_) coarseLevel.DeclareInput("Importer",(*it)->GetFactory("Importer").get(), this);
-    coarseLevel.DeclareInput("Nullspace",(*it)->GetFactory("Nullspace").get(), this);
+    if (!UseSingleSourceImporters_) coarseLevel.DeclareInput("Importer", (*it)->GetFactory("Importer").get(), this);
+    coarseLevel.DeclareInput("Nullspace", (*it)->GetFactory("Nullspace").get(), this);
   }
 
   // Use the non-manager path if the maps / importers are generated in one place
-  if(UseSingleSourceImporters_) {
-    Input(coarseLevel,"SubImporters");
+  if (UseSingleSourceImporters_) {
+    Input(coarseLevel, "SubImporters");
   }
-
-
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
   FactoryMonitor m(*this, "Build", coarseLevel);
 
-
-
   RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
   Teuchos::RCP<Matrix> originalTransferOp = Teuchos::null;
-  originalTransferOp = Get< RCP<Matrix> >(coarseLevel, "R");
+  originalTransferOp                      = Get<RCP<Matrix> >(coarseLevel, "R");
 
   RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOriginalTransferOp =
-    Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(originalTransferOp);
-  TEUCHOS_TEST_FOR_EXCEPTION(bOriginalTransferOp==Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: input matrix P or R is not of type BlockedCrsMatrix! error.");
+      Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(originalTransferOp);
+  TEUCHOS_TEST_FOR_EXCEPTION(bOriginalTransferOp == Teuchos::null, Exceptions::BadCast, "MueLu::RebalanceBlockTransferFactory::Build: input matrix P or R is not of type BlockedCrsMatrix! error.");
 
-  RCP<const MapExtractor> rangeMapExtractor = bOriginalTransferOp->getRangeMapExtractor();
+  RCP<const MapExtractor> rangeMapExtractor  = bOriginalTransferOp->getRangeMapExtractor();
   RCP<const MapExtractor> domainMapExtractor = bOriginalTransferOp->getDomainMapExtractor();
 
   // restrict communicator?
-  bool bRestrictComm = false;
-  const ParameterList& pL = GetParameterList();
+  bool bRestrictComm      = false;
+  const ParameterList &pL = GetParameterList();
   if (pL.get<bool>("repartition: use subcommunicators") == true)
     bRestrictComm = true;
 
@@ -155,15 +151,15 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
   std::vector<GO> fullDomainMapVector;
   std::vector<RCP<const Map> > subBlockRRangeMaps;
   std::vector<RCP<const Map> > subBlockRDomainMaps;
-  subBlockRRangeMaps.reserve(bOriginalTransferOp->Rows());       // reserve size for block P operators
-  subBlockRDomainMaps.reserve(bOriginalTransferOp->Cols());       // reserve size for block P operators
+  subBlockRRangeMaps.reserve(bOriginalTransferOp->Rows());   // reserve size for block P operators
+  subBlockRDomainMaps.reserve(bOriginalTransferOp->Cols());  // reserve size for block P operators
 
   std::vector<Teuchos::RCP<Matrix> > subBlockRebR;
   subBlockRebR.reserve(bOriginalTransferOp->Cols());
 
   std::vector<RCP<const Import> > importers = std::vector<RCP<const Import> >(bOriginalTransferOp->Rows(), Teuchos::null);
-  if(UseSingleSourceImporters_) {
-    importers = Get<std::vector<RCP<const Import> > >(coarseLevel,"SubImporters");
+  if (UseSingleSourceImporters_) {
+    importers = Get<std::vector<RCP<const Import> > >(coarseLevel, "SubImporters");
   }
 
   int curBlockId = 0;
@@ -171,90 +167,97 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
   std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
   for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
     // begin SubFactoryManager environment
-    SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   *it);
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), *it);
     SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), *it);
 
-    if(UseSingleSourceImporters_) rebalanceImporter = importers[curBlockId];
-    else rebalanceImporter = coarseLevel.Get<Teuchos::RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
+    if (UseSingleSourceImporters_)
+      rebalanceImporter = importers[curBlockId];
+    else
+      rebalanceImporter = coarseLevel.Get<Teuchos::RCP<const Import> >("Importer", (*it)->GetFactory("Importer").get());
 
     // extract matrix block
     Teuchos::RCP<Matrix> Rii = bOriginalTransferOp->getMatrix(curBlockId, curBlockId);
 
     // TODO run this only in the debug version
     TEUCHOS_TEST_FOR_EXCEPTION(bThyraRangeGIDs == true && Rii->getRowMap()->getMinAllGlobalIndex() != 0,
-        Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockRestrictionFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block range " << curBlockId << " start with " << Rii->getRowMap()->getMinAllGlobalIndex() << " but should start with 0");
+                               Exceptions::RuntimeError,
+                               "MueLu::RebalanceBlockRestrictionFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block range " << curBlockId << " start with " << Rii->getRowMap()->getMinAllGlobalIndex() << " but should start with 0");
     TEUCHOS_TEST_FOR_EXCEPTION(bThyraDomainGIDs == true && Rii->getColMap()->getMinAllGlobalIndex() != 0,
-        Exceptions::RuntimeError,
-        "MueLu::RebalanceBlockRestrictionFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block domain " << curBlockId << " start with " << Rii->getColMap()->getMinAllGlobalIndex() << " but should start with 0");
+                               Exceptions::RuntimeError,
+                               "MueLu::RebalanceBlockRestrictionFactory::Build: inconsistent Thyra GIDs. Thyra global ids for block domain " << curBlockId << " start with " << Rii->getColMap()->getMinAllGlobalIndex() << " but should start with 0");
 
     Teuchos::RCP<Matrix> rebRii;
-    if(rebalanceImporter != Teuchos::null) {
-      std::stringstream ss; ss << "Rebalancing restriction block R(" << curBlockId << "," << curBlockId << ")";
+    if (rebalanceImporter != Teuchos::null) {
+      std::stringstream ss;
+      ss << "Rebalancing restriction block R(" << curBlockId << "," << curBlockId << ")";
       SubFactoryMonitor m1(*this, ss.str(), coarseLevel);
       {
         SubFactoryMonitor subM(*this, "Rebalancing restriction -- fusedImport", coarseLevel);
         // Note: The 3rd argument says to use originalR's domain map.
 
         RCP<Map> dummy;
-        rebRii = MatrixFactory::Build(Rii,*rebalanceImporter,dummy,rebalanceImporter->getTargetMap());
+        rebRii = MatrixFactory::Build(Rii, *rebalanceImporter, dummy, rebalanceImporter->getTargetMap());
       }
 
       RCP<ParameterList> params = rcp(new ParameterList());
       params->set("printLoadBalancingInfo", true);
-      std::stringstream ss2; ss2 << "R(" << curBlockId << "," << curBlockId << ") rebalanced:";
+      std::stringstream ss2;
+      ss2 << "R(" << curBlockId << "," << curBlockId << ") rebalanced:";
       GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*rebRii, ss2.str(), params);
     } else {
-      rebRii = Rii;
+      rebRii                    = Rii;
       RCP<ParameterList> params = rcp(new ParameterList());
       params->set("printLoadBalancingInfo", true);
-      std::stringstream ss2; ss2 << "R(" << curBlockId << "," << curBlockId << ") not rebalanced:";
+      std::stringstream ss2;
+      ss2 << "R(" << curBlockId << "," << curBlockId << ") not rebalanced:";
       GetOStream(Statistics0) << PerfUtils::PrintMatrixInfo(*rebRii, ss2.str(), params);
     }
 
     // fix striding information for rebalanced diagonal block rebRii
-    Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(Teuchos::as<size_t>(curBlockId),rangeMapExtractor->getThyraMode()));
-    Teuchos::RCP<const Map> stridedRgMap = Teuchos::null;
-    if(orig_stridedRgMap != Teuchos::null) {
-      std::vector<size_t> stridingData = orig_stridedRgMap->getStridingData();
-      Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMapii = rebRii->getRangeMap()->getLocalElementList();
-      stridedRgMap = StridedMapFactory::Build(
-          originalTransferOp->getRangeMap()->lib(),
-          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-          nodeRangeMapii,
-          rebRii->getRangeMap()->getIndexBase(),
-          stridingData,
-          originalTransferOp->getRangeMap()->getComm(),
-          orig_stridedRgMap->getStridedBlockId(),
-          orig_stridedRgMap->getOffset());
-    } else stridedRgMap = Rii->getRangeMap();
-
-    Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(Teuchos::as<size_t>(curBlockId),domainMapExtractor->getThyraMode()));
-    Teuchos::RCP<const Map> stridedDoMap = Teuchos::null;
-    if(orig_stridedDoMap != Teuchos::null) {
-      std::vector<size_t> stridingData = orig_stridedDoMap->getStridingData();
-      Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMapii = rebRii->getDomainMap()->getLocalElementList();
-      stridedDoMap = StridedMapFactory::Build(
-          originalTransferOp->getDomainMap()->lib(),
-          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-          nodeDomainMapii,
-          rebRii->getDomainMap()->getIndexBase(),
-          stridingData,
-          originalTransferOp->getDomainMap()->getComm(),
-          orig_stridedDoMap->getStridedBlockId(),
-          orig_stridedDoMap->getOffset());
-    } else stridedDoMap = Rii->getDomainMap();
-
-    if(bRestrictComm) {
+    Teuchos::RCP<const StridedMap> orig_stridedRgMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(Teuchos::as<size_t>(curBlockId), rangeMapExtractor->getThyraMode()));
+    Teuchos::RCP<const Map> stridedRgMap             = Teuchos::null;
+    if (orig_stridedRgMap != Teuchos::null) {
+      std::vector<size_t> stridingData                       = orig_stridedRgMap->getStridingData();
+      Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMapii = rebRii->getRangeMap()->getLocalElementList();
+      stridedRgMap                                           = StridedMapFactory::Build(
+                                                    originalTransferOp->getRangeMap()->lib(),
+                                                    Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                    nodeRangeMapii,
+                                                    rebRii->getRangeMap()->getIndexBase(),
+                                                    stridingData,
+                                                    originalTransferOp->getRangeMap()->getComm(),
+                                                    orig_stridedRgMap->getStridedBlockId(),
+                                                    orig_stridedRgMap->getOffset());
+    } else
+      stridedRgMap = Rii->getRangeMap();
+
+    Teuchos::RCP<const StridedMap> orig_stridedDoMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(Teuchos::as<size_t>(curBlockId), domainMapExtractor->getThyraMode()));
+    Teuchos::RCP<const Map> stridedDoMap             = Teuchos::null;
+    if (orig_stridedDoMap != Teuchos::null) {
+      std::vector<size_t> stridingData                        = orig_stridedDoMap->getStridingData();
+      Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMapii = rebRii->getDomainMap()->getLocalElementList();
+      stridedDoMap                                            = StridedMapFactory::Build(
+                                                     originalTransferOp->getDomainMap()->lib(),
+                                                     Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                     nodeDomainMapii,
+                                                     rebRii->getDomainMap()->getIndexBase(),
+                                                     stridingData,
+                                                     originalTransferOp->getDomainMap()->getComm(),
+                                                     orig_stridedDoMap->getStridedBlockId(),
+                                                     orig_stridedDoMap->getOffset());
+    } else
+      stridedDoMap = Rii->getDomainMap();
+
+    if (bRestrictComm) {
       stridedRgMap->removeEmptyProcesses();
       stridedDoMap->removeEmptyProcesses();
     }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockRestrictionFactory::Build: failed to generate striding information. error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null,Exceptions::RuntimeError, "MueLu::RebalanceBlockRestrictionFactory::Build: failed to generate striding information. error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedRgMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockRestrictionFactory::Build: failed to generate striding information. error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoMap == Teuchos::null, Exceptions::RuntimeError, "MueLu::RebalanceBlockRestrictionFactory::Build: failed to generate striding information. error.");
 
     // replace stridedMaps view in diagonal sub block
-    if(rebRii->IsView("stridedMaps")) rebRii->RemoveView("stridedMaps");
+    if (rebRii->IsView("stridedMaps")) rebRii->RemoveView("stridedMaps");
     rebRii->CreateView("stridedMaps", stridedRgMap, stridedDoMap);
 
     // store rebalanced subblock
@@ -263,19 +266,19 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
     // append strided row map (= range map) to list of range maps.
     Teuchos::RCP<const Map> rangeMapii = rebRii->getRowMap("stridedMaps");
     subBlockRRangeMaps.push_back(rangeMapii);
-    Teuchos::ArrayView< const GlobalOrdinal > nodeRangeMapii = rebRii->getRangeMap()->getLocalElementList();
+    Teuchos::ArrayView<const GlobalOrdinal> nodeRangeMapii = rebRii->getRangeMap()->getLocalElementList();
     // append the GIDs in the end. Do not sort if we have Thyra style GIDs
     fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMapii.begin(), nodeRangeMapii.end());
-    if(bThyraRangeGIDs == false)
+    if (bThyraRangeGIDs == false)
       sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
 
     // append strided col map (= domain map) to list of range maps.
     Teuchos::RCP<const Map> domainMapii = rebRii->getColMap("stridedMaps");
     subBlockRDomainMaps.push_back(domainMapii);
-    Teuchos::ArrayView< const GlobalOrdinal > nodeDomainMapii = rebRii->getDomainMap()->getLocalElementList();
+    Teuchos::ArrayView<const GlobalOrdinal> nodeDomainMapii = rebRii->getDomainMap()->getLocalElementList();
     // append the GIDs in the end. Do not sort if we have Thyra style GIDs
     fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMapii.begin(), nodeDomainMapii.end());
-    if(bThyraDomainGIDs == false)
+    if (bThyraDomainGIDs == false)
       sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
 
     ////////////////////////////////////////////////////////////
@@ -283,12 +286,12 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
     // rebalance null space
     // This rebalances the null space partial vector associated with the current block (generated by the NullspaceFactory
     // associated with the block)
-    if(rebalanceImporter != Teuchos::null)
-    { // rebalance null space
-      std::stringstream ss2; ss2 << "Rebalancing nullspace block(" << curBlockId << "," << curBlockId << ")";
+    if (rebalanceImporter != Teuchos::null) {  // rebalance null space
+      std::stringstream ss2;
+      ss2 << "Rebalancing nullspace block(" << curBlockId << "," << curBlockId << ")";
       SubFactoryMonitor subM(*this, ss2.str(), coarseLevel);
 
-      RCP<MultiVector> nullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", (*it)->GetFactory("Nullspace").get());
+      RCP<MultiVector> nullspace         = coarseLevel.Get<RCP<MultiVector> >("Nullspace", (*it)->GetFactory("Nullspace").get());
       RCP<MultiVector> permutedNullspace = MultiVectorFactory::Build(rebalanceImporter->getTargetMap(), nullspace->getNumVectors());
       permutedNullspace->doImport(*nullspace, *rebalanceImporter, Xpetra::INSERT);
 
@@ -298,8 +301,8 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
 
       coarseLevel.Set<RCP<MultiVector> >("Nullspace", permutedNullspace, (*it)->GetFactory("Nullspace").get());
 
-    } // end rebalance null space
-    else { // do nothing
+    }       // end rebalance null space
+    else {  // do nothing
       RCP<MultiVector> nullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", (*it)->GetFactory("Nullspace").get());
       coarseLevel.Set<RCP<MultiVector> >("Nullspace", nullspace, (*it)->GetFactory("Nullspace").get());
     }
@@ -307,17 +310,17 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
     ////////////////////////////////////////////////////////////
 
     curBlockId++;
-  } // end for loop
+  }  // end for loop
 
   // extract map index base from maps of blocked P
-  GO rangeIndexBase = originalTransferOp->getRangeMap()->getIndexBase();
-  GO domainIndexBase= originalTransferOp->getDomainMap()->getIndexBase();
+  GO rangeIndexBase  = originalTransferOp->getRangeMap()->getIndexBase();
+  GO domainIndexBase = originalTransferOp->getDomainMap()->getIndexBase();
 
   // check this
-  Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0,fullRangeMapVector.size());
+  Teuchos::ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0, fullRangeMapVector.size());
   Teuchos::RCP<const StridedMap> stridedRgFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getFullMap());
-  Teuchos::RCP<const Map > fullRangeMap = Teuchos::null;
-  if(stridedRgFullMap != Teuchos::null) {
+  Teuchos::RCP<const Map> fullRangeMap            = Teuchos::null;
+  if (stridedRgFullMap != Teuchos::null) {
     std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
     fullRangeMap =
         StridedMapFactory::Build(
@@ -339,11 +342,11 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
             originalTransferOp->getRangeMap()->getComm());
   }
 
-  Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0,fullDomainMapVector.size());
+  Teuchos::ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0, fullDomainMapVector.size());
   Teuchos::RCP<const StridedMap> stridedDoFullMap = Teuchos::rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getFullMap());
-  Teuchos::RCP<const Map > fullDomainMap = Teuchos::null;
-  if(stridedDoFullMap != Teuchos::null) {
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap==Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information! error.");
+  Teuchos::RCP<const Map> fullDomainMap           = Teuchos::null;
+  if (stridedDoFullMap != Teuchos::null) {
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap == Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information! error.");
     std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
     fullDomainMap =
         StridedMapFactory::Build(
@@ -356,7 +359,6 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
             stridedDoFullMap->getStridedBlockId(),
             stridedDoFullMap->getOffset());
   } else {
-
     fullDomainMap =
         MapFactory::Build(
             originalTransferOp->getDomainMap()->lib(),
@@ -366,29 +368,29 @@ void RebalanceBlockRestrictionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>
             originalTransferOp->getDomainMap()->getComm());
   }
 
-  if(bRestrictComm) {
+  if (bRestrictComm) {
     fullRangeMap->removeEmptyProcesses();
     fullDomainMap->removeEmptyProcesses();
   }
 
   // build map extractors
-  Teuchos::RCP<const MapExtractor> rebrangeMapExtractor  =
-      MapExtractorFactory::Build(fullRangeMap,  subBlockRRangeMaps, bThyraRangeGIDs);
+  Teuchos::RCP<const MapExtractor> rebrangeMapExtractor =
+      MapExtractorFactory::Build(fullRangeMap, subBlockRRangeMaps, bThyraRangeGIDs);
   Teuchos::RCP<const MapExtractor> rebdomainMapExtractor =
       MapExtractorFactory::Build(fullDomainMap, subBlockRDomainMaps, bThyraDomainGIDs);
 
-  Teuchos::RCP<BlockedCrsMatrix> bRebR = Teuchos::rcp(new BlockedCrsMatrix(rebrangeMapExtractor,rebdomainMapExtractor,10));
-  for(size_t i = 0; i<subBlockRRangeMaps.size(); i++) {
+  Teuchos::RCP<BlockedCrsMatrix> bRebR = Teuchos::rcp(new BlockedCrsMatrix(rebrangeMapExtractor, rebdomainMapExtractor, 10));
+  for (size_t i = 0; i < subBlockRRangeMaps.size(); i++) {
     Teuchos::RCP<CrsMatrixWrap> crsOpii = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(subBlockRebR[i]);
-    bRebR->setMatrix(i,i,crsOpii);
+    bRebR->setMatrix(i, i, crsOpii);
   }
 
   bRebR->fillComplete();
 
-  Set(coarseLevel, "R", Teuchos::rcp_dynamic_cast<Matrix>(bRebR)); // do nothing  // TODO remove this!
+  Set(coarseLevel, "R", Teuchos::rcp_dynamic_cast<Matrix>(bRebR));  // do nothing  // TODO remove this!
 
-} // Build
+}  // Build
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_decl.hpp
index dad4765aae73..4c63dcff6844 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_decl.hpp
@@ -59,47 +59,46 @@
 
 namespace MueLu {
 
-  /*!
-    @class RebalanceMapFactory
-    @brief Factory which rebalances a map on current level using the Importer object generated by the RepartitionFactory
-  */
+/*!
+  @class RebalanceMapFactory
+  @brief Factory which rebalances a map on current level using the Importer object generated by the RepartitionFactory
+*/
 
-  template <class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceMapFactory : public SingleLevelFactoryBase {
-    typedef DefaultScalar Scalar; // FIXME this is just a dummy scalar for being able to access Xpetra::Matrix
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceMapFactory : public SingleLevelFactoryBase {
+  typedef DefaultScalar Scalar;  // FIXME this is just a dummy scalar for being able to access Xpetra::Matrix
 
 #undef MUELU_REBALANCEMAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! @name Constructors/Destructors
-    //@{
+  //! Constructor
+  RebalanceMapFactory() {}
 
-    //! Constructor
-    RebalanceMapFactory() {}
+  //! Destructor
+  virtual ~RebalanceMapFactory() {}
+  //@}
 
-    //! Destructor
-    virtual ~RebalanceMapFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level & level) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &level) const;
+  //@}
+};
 
-    //@}
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCEMAPFACTORY_SHORT
 
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_def.hpp
index 9adb2a0f77cf..945f94a943d7 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceMapFactory_def.hpp
@@ -57,94 +57,93 @@
 
 namespace MueLu {
 
- template <class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: use subcommunicators");
+  SET_VALID_ENTRY("repartition: use subcommunicators");
 #undef SET_VALID_ENTRY
 
-    // Information about map that is to be rebalanced
-    validParamList->set< std::string >           ("Map name"   ,                         "", "Name of map to rebalanced.");
-    validParamList->set< RCP<const FactoryBase> >("Map factory", MueLu::NoFactory::getRCP(), "Generating factory of map to be rebalanced.");
+  // Information about map that is to be rebalanced
+  validParamList->set<std::string>("Map name", "", "Name of map to rebalanced.");
+  validParamList->set<RCP<const FactoryBase> >("Map factory", MueLu::NoFactory::getRCP(), "Generating factory of map to be rebalanced.");
 
-    // Importer object with rebalancing information
-    validParamList->set< RCP<const FactoryBase> >("Importer",                 Teuchos::null, "Factory of the importer object used for the rebalancing");
-
-    return validParamList;
-  }
+  // Importer object with rebalancing information
+  validParamList->set<RCP<const FactoryBase> >("Importer", Teuchos::null, "Factory of the importer object used for the rebalancing");
 
+  return validParamList;
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & currentLevel) const {
-    const Teuchos::ParameterList & pL = GetParameterList();
-    std::string mapName                        = pL.get<std::string> ("Map name");
-    Teuchos::RCP<const FactoryBase> mapFactory = GetFactory          ("Map factory");
-    currentLevel.DeclareInput(mapName,mapFactory.get(),this);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  const Teuchos::ParameterList &pL           = GetParameterList();
+  std::string mapName                        = pL.get<std::string>("Map name");
+  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory("Map factory");
+  currentLevel.DeclareInput(mapName, mapFactory.get(), this);
 
-    Input(currentLevel, "Importer");
-  } //DeclareInput()
+  Input(currentLevel, "Importer");
+}  // DeclareInput()
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &level) const {
-    FactoryMonitor m(*this, "Build", level);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &level) const {
+  FactoryMonitor m(*this, "Build", level);
 
-    //Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
+  // Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
 
-    // extract data from Level object
-    const Teuchos::ParameterList & pL = GetParameterList();
-    std::string mapName = pL.get<std::string> ("Map name");
-    Teuchos::RCP<const FactoryBase> mapFactory = GetFactory ("Map factory");
+  // extract data from Level object
+  const Teuchos::ParameterList &pL           = GetParameterList();
+  std::string mapName                        = pL.get<std::string>("Map name");
+  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory("Map factory");
 
-    RCP<const Import> rebalanceImporter = Get<RCP<const Import> >(level, "Importer");
+  RCP<const Import> rebalanceImporter = Get<RCP<const Import> >(level, "Importer");
 
-    if(rebalanceImporter != Teuchos::null) {
-      // input map (not rebalanced)
-      RCP<const Map> map = level.Get< RCP<const Map> >(mapName,mapFactory.get());
+  if (rebalanceImporter != Teuchos::null) {
+    // input map (not rebalanced)
+    RCP<const Map> map = level.Get<RCP<const Map> >(mapName, mapFactory.get());
 
-      // create vector based on input map
-      // Note, that the map can be a part only of the full map stored in rebalanceImporter.getSourceMap()
-      RCP<Vector> v = VectorFactory::Build(map);
-      v->putScalar(1.0);
+    // create vector based on input map
+    // Note, that the map can be a part only of the full map stored in rebalanceImporter.getSourceMap()
+    RCP<Vector> v = VectorFactory::Build(map);
+    v->putScalar(1.0);
 
-      // create a new vector based on the full rebalanceImporter.getSourceMap()
-      // import the partial map information to the full source map
-      RCP<const Import> blowUpImporter = ImportFactory::Build(map, rebalanceImporter->getSourceMap());
-      RCP<Vector> pv = VectorFactory::Build(rebalanceImporter->getSourceMap());
-      pv->doImport(*v,*blowUpImporter,Xpetra::INSERT);
+    // create a new vector based on the full rebalanceImporter.getSourceMap()
+    // import the partial map information to the full source map
+    RCP<const Import> blowUpImporter = ImportFactory::Build(map, rebalanceImporter->getSourceMap());
+    RCP<Vector> pv                   = VectorFactory::Build(rebalanceImporter->getSourceMap());
+    pv->doImport(*v, *blowUpImporter, Xpetra::INSERT);
 
-      // do rebalancing using rebalanceImporter
-      RCP<Vector> ptv = VectorFactory::Build(rebalanceImporter->getTargetMap());
-      ptv->doImport(*pv,*rebalanceImporter,Xpetra::INSERT);
+    // do rebalancing using rebalanceImporter
+    RCP<Vector> ptv = VectorFactory::Build(rebalanceImporter->getTargetMap());
+    ptv->doImport(*pv, *rebalanceImporter, Xpetra::INSERT);
 
-      if (pL.get<bool>("repartition: use subcommunicators") == true)
-        ptv->replaceMap(ptv->getMap()->removeEmptyProcesses());
+    if (pL.get<bool>("repartition: use subcommunicators") == true)
+      ptv->replaceMap(ptv->getMap()->removeEmptyProcesses());
 
-      // reconstruct rebalanced partial map
-      Teuchos::ArrayRCP< const Scalar > ptvData = ptv->getData(0);
-      std::vector<GlobalOrdinal> localGIDs;  // vector with GIDs that are stored on current proc
+    // reconstruct rebalanced partial map
+    Teuchos::ArrayRCP<const Scalar> ptvData = ptv->getData(0);
+    std::vector<GlobalOrdinal> localGIDs;  // vector with GIDs that are stored on current proc
 
-      for (size_t k = 0; k < ptv->getLocalLength(); k++) {
-        if(ptvData[k] == 1.0) {
-          localGIDs.push_back(ptv->getMap()->getGlobalElement(k));
-        }
+    for (size_t k = 0; k < ptv->getLocalLength(); k++) {
+      if (ptvData[k] == 1.0) {
+        localGIDs.push_back(ptv->getMap()->getGlobalElement(k));
       }
+    }
 
-      const Teuchos::ArrayView<const GlobalOrdinal> localGIDs_view(&localGIDs[0],localGIDs.size());
+    const Teuchos::ArrayView<const GlobalOrdinal> localGIDs_view(&localGIDs[0], localGIDs.size());
 
-      Teuchos::RCP<const Map> localGIDsMap = MapFactory::Build(
-          map->lib(),
-          Teuchos::OrdinalTraits<int>::invalid(),
-          localGIDs_view,
-          0, ptv->getMap()->getComm());  // use correct communicator here!
+    Teuchos::RCP<const Map> localGIDsMap = MapFactory::Build(
+        map->lib(),
+        Teuchos::OrdinalTraits<int>::invalid(),
+        localGIDs_view,
+        0, ptv->getMap()->getComm());  // use correct communicator here!
 
-      // store rebalanced partial map using the same name and generating factory as the original map
-      // in the level class
-      level.Set(mapName, localGIDsMap, mapFactory.get());
-    }
-  } //Build()
+    // store rebalanced partial map using the same name and generating factory as the original map
+    // in the level class
+    level.Set(mapName, localGIDsMap, mapFactory.get());
+  }
+}  // Build()
 
-} // end namespace MueLu
+}  // end namespace MueLu
 
 #endif /* MUELU_REBALANCEMAPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_decl.hpp
index dddce0d6e132..01bb2e4aedb5 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_decl.hpp
@@ -65,57 +65,57 @@
 
 namespace MueLu {
 
-  /*!
-    @class RebalanceTransferFactory class.
-    @brief Applies permutation to grid transfer operators.
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RebalanceTransferFactory : public TwoLevelFactoryBase {
+/*!
+  @class RebalanceTransferFactory class.
+  @brief Applies permutation to grid transfer operators.
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RebalanceTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_REBALANCETRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RebalanceTransferFactory() { }
+  //! Constructor.
+  RebalanceTransferFactory() {}
 
-    //! Destructor.
-    virtual ~RebalanceTransferFactory() { }
+  //! Destructor.
+  virtual ~RebalanceTransferFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level& fineLevel, Level& coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; // class RebalanceTransferFactory
+};  // class RebalanceTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REBALANCETRANSFERFACTORY_SHORT
-#endif // MUELU_REBALANCETRANSFERFACTORY_DECL_HPP
+#endif  // MUELU_REBALANCETRANSFERFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_def.hpp
index c7d8b97cee0b..3b002a6cbb91 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RebalanceTransferFactory_def.hpp
@@ -69,341 +69,348 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: rebalance P and R");
-    SET_VALID_ENTRY("repartition: explicit via new copy rebalance P and R");
-    SET_VALID_ENTRY("repartition: rebalance Nullspace");
-    SET_VALID_ENTRY("transpose: use implicit");
-    SET_VALID_ENTRY("repartition: use subcommunicators");
-#undef  SET_VALID_ENTRY
-
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      RCP<validatorType> typeValidator = rcp (new validatorType(Teuchos::tuple<std::string>("Interpolation", "Restriction"), "type"));
-      validParamList->set("type", "Interpolation", "Type of the transfer operator that need to be rebalanced (Interpolation or Restriction)", typeValidator);
-    }
-
-    validParamList->set< RCP<const FactoryBase> >("P",                   null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
-    validParamList->set< RCP<const FactoryBase> >("R",                   null, "Factory of the restriction operator that need to be rebalanced (only used if type=Restriction)");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",           null, "Factory of the nullspace that need to be rebalanced (only used if type=Interpolation)");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",         null, "Factory of the coordinates that need to be rebalanced (only used if type=Interpolation)");
-    validParamList->set< RCP<const FactoryBase> >("BlockNumber",         null, "Factory of the block ids that need to be rebalanced (only used if type=Interpolation)");
-    validParamList->set< RCP<const FactoryBase> >("Importer",            null, "Factory of the importer object used for the rebalancing");
-    validParamList->set< int >                   ("write start",           -1, "First level at which coordinates should be written to file");
-    validParamList->set< int >                   ("write end",             -1, "Last level at which coordinates should be written to file");
-
-    // TODO validation: "P" parameter valid only for type="Interpolation" and "R" valid only for type="Restriction". Like so:
-    // if (paramList.isEntry("type") && paramList.get("type) == "Interpolation) {
-    //     validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
+  SET_VALID_ENTRY("repartition: rebalance P and R");
+  SET_VALID_ENTRY("repartition: explicit via new copy rebalance P and R");
+  SET_VALID_ENTRY("repartition: rebalance Nullspace");
+  SET_VALID_ENTRY("transpose: use implicit");
+  SET_VALID_ENTRY("repartition: use subcommunicators");
+#undef SET_VALID_ENTRY
+
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    RCP<validatorType> typeValidator = rcp(new validatorType(Teuchos::tuple<std::string>("Interpolation", "Restriction"), "type"));
+    validParamList->set("type", "Interpolation", "Type of the transfer operator that need to be rebalanced (Interpolation or Restriction)", typeValidator);
+  }
 
-    return validParamList;
+  validParamList->set<RCP<const FactoryBase> >("P", null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
+  validParamList->set<RCP<const FactoryBase> >("R", null, "Factory of the restriction operator that need to be rebalanced (only used if type=Restriction)");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", null, "Factory of the nullspace that need to be rebalanced (only used if type=Interpolation)");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", null, "Factory of the coordinates that need to be rebalanced (only used if type=Interpolation)");
+  validParamList->set<RCP<const FactoryBase> >("BlockNumber", null, "Factory of the block ids that need to be rebalanced (only used if type=Interpolation)");
+  validParamList->set<RCP<const FactoryBase> >("Importer", null, "Factory of the importer object used for the rebalancing");
+  validParamList->set<int>("write start", -1, "First level at which coordinates should be written to file");
+  validParamList->set<int>("write end", -1, "Last level at which coordinates should be written to file");
+
+  // TODO validation: "P" parameter valid only for type="Interpolation" and "R" valid only for type="Restriction". Like so:
+  // if (paramList.isEntry("type") && paramList.get("type) == "Interpolation) {
+  //     validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Factory of the prolongation operator that need to be rebalanced (only used if type=Interpolation)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  const ParameterList& pL = GetParameterList();
+
+  if (pL.get<std::string>("type") == "Interpolation") {
+    Input(coarseLevel, "P");
+    if (pL.get<bool>("repartition: rebalance Nullspace"))
+      Input(coarseLevel, "Nullspace");
+    if (pL.get<RCP<const FactoryBase> >("Coordinates") != Teuchos::null)
+      Input(coarseLevel, "Coordinates");
+    if (pL.get<RCP<const FactoryBase> >("BlockNumber") != Teuchos::null)
+      Input(coarseLevel, "BlockNumber");
+
+  } else {
+    if (pL.get<bool>("transpose: use implicit") == false)
+      Input(coarseLevel, "R");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
-    const ParameterList& pL = GetParameterList();
-
-    if (pL.get<std::string>("type") == "Interpolation") {
-      Input(coarseLevel, "P");
-      if (pL.get<bool>("repartition: rebalance Nullspace"))
-        Input(coarseLevel, "Nullspace");
-      if (pL.get< RCP<const FactoryBase> >("Coordinates") != Teuchos::null)
-        Input(coarseLevel, "Coordinates");
-      if (pL.get< RCP<const FactoryBase> >("BlockNumber") != Teuchos::null)
-        Input(coarseLevel, "BlockNumber");
-
-    } else {
-      if (pL.get<bool>("transpose: use implicit") == false)
-        Input(coarseLevel, "R");
-    }
+  Input(coarseLevel, "Importer");
+}
 
-    Input(coarseLevel, "Importer");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RebalanceTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
+  const ParameterList& pL = GetParameterList();
 
-    const ParameterList& pL = GetParameterList();
+  RCP<Matrix> originalP = Get<RCP<Matrix> >(coarseLevel, "P");
+  // If we don't have a valid P (e.g., # global aggregates is 0), skip this rebalancing. This level will
+  // ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (originalP == Teuchos::null) {
+    Set(coarseLevel, "P", originalP);
+    return;
+  }
+  int implicit       = !pL.get<bool>("repartition: rebalance P and R");
+  int reallyExplicit = pL.get<bool>("repartition: explicit via new copy rebalance P and R");
+  int writeStart     = pL.get<int>("write start");
+  int writeEnd       = pL.get<int>("write end");
+
+  if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd && IsAvailable(fineLevel, "Coordinates")) {
+    std::string fileName = "coordinates_level_0.m";
+    RCP<xdMV> fineCoords = fineLevel.Get<RCP<xdMV> >("Coordinates");
+    if (fineCoords != Teuchos::null)
+      Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Write(fileName, *fineCoords);
+  }
 
-    RCP<Matrix> originalP = Get< RCP<Matrix> >(coarseLevel, "P");
-    // If we don't have a valid P (e.g., # global aggregates is 0), skip this rebalancing. This level will
-    // ultimately be removed in MueLu_Hierarchy_defs.h via a resize() 
-    if (originalP == Teuchos::null) {
-      Set(coarseLevel, "P", originalP);
-      return;
-    }
-    int implicit   = !pL.get<bool>("repartition: rebalance P and R");
-    int reallyExplicit = pL.get<bool>("repartition: explicit via new copy rebalance P and R");
-    int writeStart = pL.get<int> ("write start");
-    int writeEnd   = pL.get<int> ("write end");
-
-    if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd && IsAvailable(fineLevel, "Coordinates")) {
-      std::string fileName = "coordinates_level_0.m";
-      RCP<xdMV> fineCoords = fineLevel.Get< RCP<xdMV> >("Coordinates");
-      if (fineCoords != Teuchos::null)
-        Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Write(fileName, *fineCoords);
-    }
+  if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd && IsAvailable(fineLevel, "BlockNumber")) {
+    std::string fileName                    = "BlockNumber_level_0.m";
+    RCP<LocalOrdinalVector> fineBlockNumber = fineLevel.Get<RCP<LocalOrdinalVector> >("BlockNumber");
+    if (fineBlockNumber != Teuchos::null)
+      Xpetra::IO<LO, LO, GO, NO>::Write(fileName, *fineBlockNumber);
+  }
 
-    if (writeStart == 0 && fineLevel.GetLevelID() == 0 && writeStart <= writeEnd && IsAvailable(fineLevel, "BlockNumber")) {
-      std::string fileName = "BlockNumber_level_0.m";
-      RCP<LocalOrdinalVector> fineBlockNumber = fineLevel.Get< RCP<LocalOrdinalVector> >("BlockNumber");
-      if (fineBlockNumber != Teuchos::null)
-        Xpetra::IO<LO,LO,GO,NO>::Write(fileName, *fineBlockNumber);
-    }
+  RCP<const Import> importer = Get<RCP<const Import> >(coarseLevel, "Importer");
+  if (implicit) {
+    // Save the importer, we'll need it for solve
+    coarseLevel.Set("Importer", importer, NoFactory::get());
+  }
 
-    RCP<const Import> importer = Get<RCP<const Import> >(coarseLevel, "Importer");
-    if (implicit) {
-      // Save the importer, we'll need it for solve
-      coarseLevel.Set("Importer", importer, NoFactory::get());
-    }
+  RCP<ParameterList> params = rcp(new ParameterList());
+  if (IsPrint(Statistics2)) {
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+  }
 
-    RCP<ParameterList> params = rcp(new ParameterList());
-    if (IsPrint(Statistics2)) {
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-    }
+  std::string transferType = pL.get<std::string>("type");
+  if (transferType == "Interpolation") {
+    originalP = Get<RCP<Matrix> >(coarseLevel, "P");
 
-    std::string transferType = pL.get<std::string>("type");
-    if (transferType == "Interpolation") {
-      originalP = Get< RCP<Matrix> >(coarseLevel, "P");
+    {
+      // This line must be after the Get call
+      SubFactoryMonitor m1(*this, "Rebalancing prolongator", coarseLevel);
+
+      if (implicit || importer.is_null()) {
+        GetOStream(Runtime0) << "Using original prolongator" << std::endl;
+        Set(coarseLevel, "P", originalP);
+
+      } else {
+        // There are two version of an explicit rebalanced P and R.
+        // The !reallyExplicit way, is sufficient for all MueLu purposes
+        // with the exception of the CombinePFactory that needs true domain
+        // and column maps.
+        //   !reallyExplicit:
+        //     Rather than calling fillComplete (which would entail creating a new
+        //     column map), it's sufficient to replace the domain map and importer.
+        //     Note that this potentially violates the assumption that in the
+        //     column map, local IDs appear before any off-rank IDs.
+        //
+        //  reallyExplicit:
+        //     P transfers from coarse grid to the fine grid. Here, we change
+        //     the domain map (coarse) of Paccording to the new partition. The
+        //     range map (fine) is kept unchanged.
+        //
+        //     The domain map of P must match the range map of R. To change the
+        //     domain map of P, P needs to be fillCompleted again with the new
+        //     domain map.  To achieve this, P is copied into a new matrix that
+        //     is not fill-completed.  The doImport() operation is just used
+        //     here to make a copy of P: the importer is trivial and there is
+        //     no data movement involved.  The reordering actually happens during
+        //     fillComplete() with domainMap == importer->getTargetMap().
+
+        RCP<Matrix> rebalancedP;
+        if (reallyExplicit) {
+          size_t totalMaxPerRow = 0;
+          ArrayRCP<size_t> nnzPerRow(originalP->getRowMap()->getLocalNumElements(), 0);
+          for (size_t i = 0; i < originalP->getRowMap()->getLocalNumElements(); ++i) {
+            nnzPerRow[i] = originalP->getNumEntriesInLocalRow(i);
+            if (nnzPerRow[i] > totalMaxPerRow) totalMaxPerRow = nnzPerRow[i];
+          }
 
-      {
-        // This line must be after the Get call
-        SubFactoryMonitor m1(*this, "Rebalancing prolongator", coarseLevel);
+          rebalancedP = MatrixFactory::Build(originalP->getRowMap(), totalMaxPerRow);
 
-        if (implicit || importer.is_null()) {
-          GetOStream(Runtime0) << "Using original prolongator" << std::endl;
-          Set(coarseLevel, "P", originalP);
+          {
+            RCP<Import> trivialImporter = ImportFactory::Build(originalP->getRowMap(), originalP->getRowMap());
+            SubFactoryMonitor m2(*this, "Rebalancing prolongator -- import only", coarseLevel);
+            rebalancedP->doImport(*originalP, *trivialImporter, Xpetra::INSERT);
+          }
+          rebalancedP->fillComplete(importer->getTargetMap(), originalP->getRangeMap());
 
         } else {
-           // There are two version of an explicit rebalanced P and R.
-           // The !reallyExplicit way, is sufficient for all MueLu purposes
-           // with the exception of the CombinePFactory that needs true domain
-           // and column maps.
-           //   !reallyExplicit:
-           //     Rather than calling fillComplete (which would entail creating a new
-           //     column map), it's sufficient to replace the domain map and importer.
-           //     Note that this potentially violates the assumption that in the
-           //     column map, local IDs appear before any off-rank IDs.
-           //
-           //  reallyExplicit:
-           //     P transfers from coarse grid to the fine grid. Here, we change
-           //     the domain map (coarse) of Paccording to the new partition. The
-           //     range map (fine) is kept unchanged.
-           //
-           //     The domain map of P must match the range map of R. To change the
-           //     domain map of P, P needs to be fillCompleted again with the new
-           //     domain map.  To achieve this, P is copied into a new matrix that
-           //     is not fill-completed.  The doImport() operation is just used 
-           //     here to make a copy of P: the importer is trivial and there is 
-           //     no data movement involved.  The reordering actually happens during
-           //     fillComplete() with domainMap == importer->getTargetMap().
-
-          RCP<Matrix> rebalancedP;
-          if (reallyExplicit) {
-            size_t totalMaxPerRow = 0;
-            ArrayRCP<size_t> nnzPerRow(originalP->getRowMap()->getLocalNumElements(), 0);
-            for (size_t i=0; i<originalP->getRowMap()->getLocalNumElements();  ++i) {
-              nnzPerRow[i] = originalP->getNumEntriesInLocalRow(i);
-              if (nnzPerRow[i] > totalMaxPerRow) totalMaxPerRow = nnzPerRow[i];
-            }
+          rebalancedP                    = originalP;
+          RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(originalP);
+          TEUCHOS_TEST_FOR_EXCEPTION(crsOp == Teuchos::null, Exceptions::BadCast, "Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
 
-            rebalancedP = MatrixFactory::Build(originalP->getRowMap(), totalMaxPerRow);
+          RCP<CrsMatrix> rebalancedP2 = crsOp->getCrsMatrix();
+          TEUCHOS_TEST_FOR_EXCEPTION(rebalancedP2 == Teuchos::null, std::runtime_error, "Xpetra::CrsMatrixWrap doesn't have a CrsMatrix");
 
-           {
-             RCP<Import> trivialImporter = ImportFactory::Build(originalP->getRowMap(), originalP->getRowMap());
-             SubFactoryMonitor m2(*this, "Rebalancing prolongator -- import only", coarseLevel);
-             rebalancedP->doImport(*originalP, *trivialImporter, Xpetra::INSERT);
-           }
-          rebalancedP->fillComplete(importer->getTargetMap(), originalP->getRangeMap() );
+          {
+            SubFactoryMonitor subM(*this, "Rebalancing prolongator -- fast map replacement", coarseLevel);
 
-          }
-          else  {
-            rebalancedP = originalP;
-            RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(originalP);
-            TEUCHOS_TEST_FOR_EXCEPTION(crsOp == Teuchos::null, Exceptions::BadCast, "Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-  
-            RCP<CrsMatrix> rebalancedP2 = crsOp->getCrsMatrix();
-            TEUCHOS_TEST_FOR_EXCEPTION(rebalancedP2 == Teuchos::null, std::runtime_error, "Xpetra::CrsMatrixWrap doesn't have a CrsMatrix");
-  
+            RCP<const Import> newImporter;
             {
-              SubFactoryMonitor subM(*this, "Rebalancing prolongator -- fast map replacement", coarseLevel);
-  
-              RCP<const Import> newImporter;
-              {
-                SubFactoryMonitor(*this, "Import construction", coarseLevel);
-                newImporter = ImportFactory::Build(importer->getTargetMap(), rebalancedP->getColMap());
-              }
-              rebalancedP2->replaceDomainMapAndImporter(importer->getTargetMap(), newImporter);
+              SubFactoryMonitor(*this, "Import construction", coarseLevel);
+              newImporter = ImportFactory::Build(importer->getTargetMap(), rebalancedP->getColMap());
             }
+            rebalancedP2->replaceDomainMapAndImporter(importer->getTargetMap(), newImporter);
           }
-          ///////////////////////// EXPERIMENTAL
-          // TODO FIXME somehow we have to transfer the striding information of the permuted domain/range maps.
-          // That is probably something for an external permutation factory
-          //   if (originalP->IsView("stridedMaps"))
-          //     rebalancedP->CreateView("stridedMaps", originalP);
-          ///////////////////////// EXPERIMENTAL
-          if(!rebalancedP.is_null()) {std::ostringstream oss; oss << "P_" << coarseLevel.GetLevelID(); rebalancedP->setObjectLabel(oss.str());}
-          Set(coarseLevel, "P", rebalancedP);
-
-          if (IsPrint(Statistics2))
-            GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedP, "P (rebalanced)", params);
         }
-      }
-
-      if (importer.is_null()) {
-        if (IsAvailable(coarseLevel, "Nullspace"))
-          Set(coarseLevel, "Nullspace", Get<RCP<MultiVector> >(coarseLevel, "Nullspace"));
-
-        if (pL.isParameter("Coordinates") && pL.get< RCP<const FactoryBase> >("Coordinates") != Teuchos::null)
-          if (IsAvailable(coarseLevel, "Coordinates"))
-            Set(coarseLevel, "Coordinates", Get< RCP<xdMV> >(coarseLevel, "Coordinates"));
-
-        if (pL.isParameter("BlockNumber") && pL.get< RCP<const FactoryBase> >("BlockNumber") != Teuchos::null)
-          if (IsAvailable(coarseLevel, "BlockNumber"))
-            Set(coarseLevel, "BlockNumber", Get< RCP<LocalOrdinalVector> >(coarseLevel, "BlockNumber"));
+        ///////////////////////// EXPERIMENTAL
+        // TODO FIXME somehow we have to transfer the striding information of the permuted domain/range maps.
+        // That is probably something for an external permutation factory
+        //   if (originalP->IsView("stridedMaps"))
+        //     rebalancedP->CreateView("stridedMaps", originalP);
+        ///////////////////////// EXPERIMENTAL
+        if (!rebalancedP.is_null()) {
+          std::ostringstream oss;
+          oss << "P_" << coarseLevel.GetLevelID();
+          rebalancedP->setObjectLabel(oss.str());
+        }
+        Set(coarseLevel, "P", rebalancedP);
 
-        return;
+        if (IsPrint(Statistics2))
+          GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedP, "P (rebalanced)", params);
       }
+    }
 
-      if (pL.isParameter("Coordinates") &&
-          pL.get< RCP<const FactoryBase> >("Coordinates") != Teuchos::null &&
-          IsAvailable(coarseLevel, "Coordinates")) {
-        RCP<xdMV> coords = Get<RCP<xdMV> >(coarseLevel, "Coordinates");
-
-        // This line must be after the Get call
-        SubFactoryMonitor subM(*this, "Rebalancing coordinates", coarseLevel);
+    if (importer.is_null()) {
+      if (IsAvailable(coarseLevel, "Nullspace"))
+        Set(coarseLevel, "Nullspace", Get<RCP<MultiVector> >(coarseLevel, "Nullspace"));
 
-        LO nodeNumElts = coords->getMap()->getLocalNumElements();
+      if (pL.isParameter("Coordinates") && pL.get<RCP<const FactoryBase> >("Coordinates") != Teuchos::null)
+        if (IsAvailable(coarseLevel, "Coordinates"))
+          Set(coarseLevel, "Coordinates", Get<RCP<xdMV> >(coarseLevel, "Coordinates"));
 
-        // If a process has no matrix rows, then we can't calculate blocksize using the formula below.
-        LO myBlkSize = 0, blkSize = 0;
-        if (nodeNumElts > 0)
-          myBlkSize = importer->getSourceMap()->getLocalNumElements() / nodeNumElts;
-        MueLu_maxAll(coords->getMap()->getComm(), myBlkSize, blkSize);
+      if (pL.isParameter("BlockNumber") && pL.get<RCP<const FactoryBase> >("BlockNumber") != Teuchos::null)
+        if (IsAvailable(coarseLevel, "BlockNumber"))
+          Set(coarseLevel, "BlockNumber", Get<RCP<LocalOrdinalVector> >(coarseLevel, "BlockNumber"));
 
-        RCP<const Import> coordImporter;
-        if (blkSize == 1) {
-          coordImporter = importer;
+      return;
+    }
 
-        } else {
-          // NOTE: there is an implicit assumption here: we assume that dof any node are enumerated consequently
-          // Proper fix would require using decomposition similar to how we construct importer in the
-          // RepartitionFactory
-          RCP<const Map> origMap   = coords->getMap();
-          GO             indexBase = origMap->getIndexBase();
-
-          ArrayView<const GO> OEntries   = importer->getTargetMap()->getLocalElementList();
-          LO                  numEntries = OEntries.size()/blkSize;
-          ArrayRCP<GO> Entries(numEntries);
-          for (LO i = 0; i < numEntries; i++)
-            Entries[i] = (OEntries[i*blkSize]-indexBase)/blkSize + indexBase;
-
-          RCP<const Map> targetMap = MapFactory::Build(origMap->lib(), origMap->getGlobalNumElements(), Entries(), indexBase, origMap->getComm());
-          coordImporter = ImportFactory::Build(origMap, targetMap);
-        }
+    if (pL.isParameter("Coordinates") &&
+        pL.get<RCP<const FactoryBase> >("Coordinates") != Teuchos::null &&
+        IsAvailable(coarseLevel, "Coordinates")) {
+      RCP<xdMV> coords = Get<RCP<xdMV> >(coarseLevel, "Coordinates");
+
+      // This line must be after the Get call
+      SubFactoryMonitor subM(*this, "Rebalancing coordinates", coarseLevel);
+
+      LO nodeNumElts = coords->getMap()->getLocalNumElements();
+
+      // If a process has no matrix rows, then we can't calculate blocksize using the formula below.
+      LO myBlkSize = 0, blkSize = 0;
+      if (nodeNumElts > 0)
+        myBlkSize = importer->getSourceMap()->getLocalNumElements() / nodeNumElts;
+      MueLu_maxAll(coords->getMap()->getComm(), myBlkSize, blkSize);
+
+      RCP<const Import> coordImporter;
+      if (blkSize == 1) {
+        coordImporter = importer;
+
+      } else {
+        // NOTE: there is an implicit assumption here: we assume that dof any node are enumerated consequently
+        // Proper fix would require using decomposition similar to how we construct importer in the
+        // RepartitionFactory
+        RCP<const Map> origMap = coords->getMap();
+        GO indexBase           = origMap->getIndexBase();
+
+        ArrayView<const GO> OEntries = importer->getTargetMap()->getLocalElementList();
+        LO numEntries                = OEntries.size() / blkSize;
+        ArrayRCP<GO> Entries(numEntries);
+        for (LO i = 0; i < numEntries; i++)
+          Entries[i] = (OEntries[i * blkSize] - indexBase) / blkSize + indexBase;
+
+        RCP<const Map> targetMap = MapFactory::Build(origMap->lib(), origMap->getGlobalNumElements(), Entries(), indexBase, origMap->getComm());
+        coordImporter            = ImportFactory::Build(origMap, targetMap);
+      }
 
-        RCP<xdMV> permutedCoords  = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(coordImporter->getTargetMap(), coords->getNumVectors());
-        permutedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
+      RCP<xdMV> permutedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(coordImporter->getTargetMap(), coords->getNumVectors());
+      permutedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
 
-        if (pL.isParameter("repartition: use subcommunicators") == true && pL.get<bool>("repartition: use subcommunicators") == true)
-          permutedCoords->replaceMap(permutedCoords->getMap()->removeEmptyProcesses());
+      if (pL.isParameter("repartition: use subcommunicators") == true && pL.get<bool>("repartition: use subcommunicators") == true)
+        permutedCoords->replaceMap(permutedCoords->getMap()->removeEmptyProcesses());
 
-        if (permutedCoords->getMap() == Teuchos::null)
-          permutedCoords = Teuchos::null;
+      if (permutedCoords->getMap() == Teuchos::null)
+        permutedCoords = Teuchos::null;
 
-        Set(coarseLevel, "Coordinates", permutedCoords);
+      Set(coarseLevel, "Coordinates", permutedCoords);
 
-        std::string fileName = "rebalanced_coordinates_level_" + toString(coarseLevel.GetLevelID()) + ".m";
-        if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd && permutedCoords->getMap() != Teuchos::null)
-          Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Write(fileName, *permutedCoords);
-      }
+      std::string fileName = "rebalanced_coordinates_level_" + toString(coarseLevel.GetLevelID()) + ".m";
+      if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd && permutedCoords->getMap() != Teuchos::null)
+        Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Write(fileName, *permutedCoords);
+    }
 
-      if (pL.isParameter("BlockNumber") &&
-          pL.get< RCP<const FactoryBase> >("BlockNumber") != Teuchos::null &&
-          IsAvailable(coarseLevel, "BlockNumber")) {
-        RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector> >(coarseLevel, "BlockNumber");
+    if (pL.isParameter("BlockNumber") &&
+        pL.get<RCP<const FactoryBase> >("BlockNumber") != Teuchos::null &&
+        IsAvailable(coarseLevel, "BlockNumber")) {
+      RCP<LocalOrdinalVector> BlockNumber = Get<RCP<LocalOrdinalVector> >(coarseLevel, "BlockNumber");
 
-        // This line must be after the Get call
-        SubFactoryMonitor subM(*this, "Rebalancing BlockNumber", coarseLevel);
+      // This line must be after the Get call
+      SubFactoryMonitor subM(*this, "Rebalancing BlockNumber", coarseLevel);
 
-        RCP<LocalOrdinalVector> permutedBlockNumber  = LocalOrdinalVectorFactory::Build(importer->getTargetMap(), false);
-        permutedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
+      RCP<LocalOrdinalVector> permutedBlockNumber = LocalOrdinalVectorFactory::Build(importer->getTargetMap(), false);
+      permutedBlockNumber->doImport(*BlockNumber, *importer, Xpetra::INSERT);
 
-        if (pL.isParameter("repartition: use subcommunicators") == true && pL.get<bool>("repartition: use subcommunicators") == true)
-          permutedBlockNumber->replaceMap(permutedBlockNumber->getMap()->removeEmptyProcesses());
+      if (pL.isParameter("repartition: use subcommunicators") == true && pL.get<bool>("repartition: use subcommunicators") == true)
+        permutedBlockNumber->replaceMap(permutedBlockNumber->getMap()->removeEmptyProcesses());
 
-        if (permutedBlockNumber->getMap() == Teuchos::null)
-          permutedBlockNumber = Teuchos::null;
+      if (permutedBlockNumber->getMap() == Teuchos::null)
+        permutedBlockNumber = Teuchos::null;
 
-        Set(coarseLevel, "BlockNumber", permutedBlockNumber);
+      Set(coarseLevel, "BlockNumber", permutedBlockNumber);
 
-        std::string fileName = "rebalanced_BlockNumber_level_" + toString(coarseLevel.GetLevelID()) + ".m";
-        if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd && permutedBlockNumber->getMap() != Teuchos::null)
-          Xpetra::IO<LO,LO,GO,NO>::Write(fileName, *permutedBlockNumber);
-      }
+      std::string fileName = "rebalanced_BlockNumber_level_" + toString(coarseLevel.GetLevelID()) + ".m";
+      if (writeStart <= coarseLevel.GetLevelID() && coarseLevel.GetLevelID() <= writeEnd && permutedBlockNumber->getMap() != Teuchos::null)
+        Xpetra::IO<LO, LO, GO, NO>::Write(fileName, *permutedBlockNumber);
+    }
 
-      if (IsAvailable(coarseLevel, "Nullspace")) {
-        RCP<MultiVector> nullspace = Get< RCP<MultiVector> >(coarseLevel, "Nullspace");
+    if (IsAvailable(coarseLevel, "Nullspace")) {
+      RCP<MultiVector> nullspace = Get<RCP<MultiVector> >(coarseLevel, "Nullspace");
 
-        // This line must be after the Get call
-        SubFactoryMonitor subM(*this, "Rebalancing nullspace", coarseLevel);
+      // This line must be after the Get call
+      SubFactoryMonitor subM(*this, "Rebalancing nullspace", coarseLevel);
 
-        RCP<MultiVector> permutedNullspace = MultiVectorFactory::Build(importer->getTargetMap(), nullspace->getNumVectors());
-        permutedNullspace->doImport(*nullspace, *importer, Xpetra::INSERT);
+      RCP<MultiVector> permutedNullspace = MultiVectorFactory::Build(importer->getTargetMap(), nullspace->getNumVectors());
+      permutedNullspace->doImport(*nullspace, *importer, Xpetra::INSERT);
 
-        if (pL.get<bool>("repartition: use subcommunicators") == true)
-          permutedNullspace->replaceMap(permutedNullspace->getMap()->removeEmptyProcesses());
+      if (pL.get<bool>("repartition: use subcommunicators") == true)
+        permutedNullspace->replaceMap(permutedNullspace->getMap()->removeEmptyProcesses());
 
-        if (permutedNullspace->getMap() == Teuchos::null)
-          permutedNullspace = Teuchos::null;
+      if (permutedNullspace->getMap() == Teuchos::null)
+        permutedNullspace = Teuchos::null;
 
-        Set(coarseLevel, "Nullspace", permutedNullspace);
-      }
+      Set(coarseLevel, "Nullspace", permutedNullspace);
+    }
 
-    } else {
-      if (pL.get<bool>("transpose: use implicit") == false) {
-        RCP<Matrix> originalR = Get< RCP<Matrix> >(coarseLevel, "R");
+  } else {
+    if (pL.get<bool>("transpose: use implicit") == false) {
+      RCP<Matrix> originalR = Get<RCP<Matrix> >(coarseLevel, "R");
 
-        SubFactoryMonitor m2(*this, "Rebalancing restrictor", coarseLevel);
+      SubFactoryMonitor m2(*this, "Rebalancing restrictor", coarseLevel);
 
-        if (implicit || importer.is_null()) {
-          GetOStream(Runtime0) << "Using original restrictor" << std::endl;
-          Set(coarseLevel, "R", originalR);
+      if (implicit || importer.is_null()) {
+        GetOStream(Runtime0) << "Using original restrictor" << std::endl;
+        Set(coarseLevel, "R", originalR);
 
-        } else {
-          RCP<Matrix> rebalancedR;
-          {
-            SubFactoryMonitor subM(*this, "Rebalancing restriction -- fusedImport", coarseLevel);
+      } else {
+        RCP<Matrix> rebalancedR;
+        {
+          SubFactoryMonitor subM(*this, "Rebalancing restriction -- fusedImport", coarseLevel);
 
-            RCP<Map> dummy;         // meaning: use originalR's domain map.
-            Teuchos::ParameterList listLabel;
-            listLabel.set("Timer Label","MueLu::RebalanceR-" + Teuchos::toString(coarseLevel.GetLevelID()));
-            rebalancedR = MatrixFactory::Build(originalR, *importer, dummy, importer->getTargetMap(),Teuchos::rcp(&listLabel,false));
-          }
-          if(!rebalancedR.is_null()) {std::ostringstream oss; oss << "R_" << coarseLevel.GetLevelID(); rebalancedR->setObjectLabel(oss.str());}
-          Set(coarseLevel, "R", rebalancedR);
-
-          ///////////////////////// EXPERIMENTAL
-          // TODO FIXME somehow we have to transfer the striding information of the permuted domain/range maps.
-          // That is probably something for an external permutation factory
-          // if (originalR->IsView("stridedMaps"))
-          //   rebalancedR->CreateView("stridedMaps", originalR);
-          ///////////////////////// EXPERIMENTAL
-
-          if (IsPrint(Statistics2))
-            GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedR, "R (rebalanced)", params);
+          RCP<Map> dummy;  // meaning: use originalR's domain map.
+          Teuchos::ParameterList listLabel;
+          listLabel.set("Timer Label", "MueLu::RebalanceR-" + Teuchos::toString(coarseLevel.GetLevelID()));
+          rebalancedR = MatrixFactory::Build(originalR, *importer, dummy, importer->getTargetMap(), Teuchos::rcp(&listLabel, false));
         }
+        if (!rebalancedR.is_null()) {
+          std::ostringstream oss;
+          oss << "R_" << coarseLevel.GetLevelID();
+          rebalancedR->setObjectLabel(oss.str());
+        }
+        Set(coarseLevel, "R", rebalancedR);
+
+        ///////////////////////// EXPERIMENTAL
+        // TODO FIXME somehow we have to transfer the striding information of the permuted domain/range maps.
+        // That is probably something for an external permutation factory
+        // if (originalR->IsView("stridedMaps"))
+        //   rebalancedR->CreateView("stridedMaps", originalR);
+        ///////////////////////// EXPERIMENTAL
+
+        if (IsPrint(Statistics2))
+          GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*rebalancedR, "R (rebalanced)", params);
       }
     }
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_REBALANCETRANSFERFACTORY_DEF_HPP
+#endif  // MUELU_REBALANCETRANSFERFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_decl.hpp
index 6e3ad77d059d..ba943cfcfa9b 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_decl.hpp
@@ -56,46 +56,45 @@
 
 namespace MueLu {
 
-  /*!
-    @class RepartitionBlockDiagonalFactory
-    @brief Factory which rebalances a map on current level using the Importer object generated by the RepartitionFactory
-  */
-
-  template <class Scalar = DefaultScalar,
-	     class LocalOrdinal = DefaultLocalOrdinal,
-	     class GlobalOrdinal = DefaultGlobalOrdinal,
-	     class Node = DefaultNode>
-  class RepartitionBlockDiagonalFactory : public SingleLevelFactoryBase {
+/*!
+  @class RepartitionBlockDiagonalFactory
+  @brief Factory which rebalances a map on current level using the Importer object generated by the RepartitionFactory
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RepartitionBlockDiagonalFactory : public SingleLevelFactoryBase {
 #undef MUELU_REPARTITIONBLOCKDIAGONALFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! @name Constructors/Destructors
-    //@{
+  //! Constructor
+  RepartitionBlockDiagonalFactory() {}
 
-    //! Constructor
-    RepartitionBlockDiagonalFactory() {}
+  //! Destructor
+  virtual ~RepartitionBlockDiagonalFactory() {}
+  //@}
 
-    //! Destructor
-    virtual ~RepartitionBlockDiagonalFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level & level) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &level) const;
+  //@}
+};
 
-    //@}
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPARTITIONBLOCKDIAGONALFACTORY_SHORT
 
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_def.hpp
index a539c2c824cb..0ea725b4a553 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionBlockDiagonalFactory_def.hpp
@@ -58,41 +58,39 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >   ("A",             Teuchos::null, "Factory of the matrix A");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "A");
+}  // DeclareInput()
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & currentLevel) const {
-    Input(currentLevel, "A");
-  } //DeclareInput()
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockCrs;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionBlockDiagonalFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>  BlockCrs;
+  RCP<Matrix> originalA = Get<RCP<Matrix> >(currentLevel, "A");
+  RCP<BlockCrs> A       = Teuchos::rcp_dynamic_cast<BlockCrs>(originalA);
 
-    RCP<Matrix> originalA = Get< RCP<Matrix> >(currentLevel, "A");
-    RCP<BlockCrs> A = Teuchos::rcp_dynamic_cast<BlockCrs>(originalA);
+  //    RCP<BlockCrs> A = Get< RCP<BlockCrs> >(currentLevel, "A");
+  TEUCHOS_TEST_FOR_EXCEPTION(A == Teuchos::null, Exceptions::BadCast, "MueLu::RepartitionBlockDiagonalFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
 
-    //    RCP<BlockCrs> A = Get< RCP<BlockCrs> >(currentLevel, "A");
-    TEUCHOS_TEST_FOR_EXCEPTION(A==Teuchos::null, Exceptions::BadCast, "MueLu::RepartitionBlockDiagonalFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
+  // Build the block diagonal
+  RCP<BlockCrs> DiagonalMatrix = Teuchos::rcp(new BlockCrs(A->getBlockedRangeMap(), A->getBlockedDomainMap(), 0));
+  for (size_t i = 0; i < A->Rows(); i++)
+    DiagonalMatrix->setMatrix(i, i, A->getMatrix(i, i));
 
-    // Build the block diagonal
-    RCP<BlockCrs> DiagonalMatrix = Teuchos::rcp(new BlockCrs(A->getBlockedRangeMap(),A->getBlockedDomainMap(),0));
-    for(size_t i=0; i< A->Rows(); i++)
-      DiagonalMatrix->setMatrix(i,i,A->getMatrix(i,i));
+  Set(currentLevel, "A", Teuchos::rcp_dynamic_cast<Matrix>(DiagonalMatrix));
 
-    Set(currentLevel,"A",Teuchos::rcp_dynamic_cast<Matrix>(DiagonalMatrix));
+}  // Build()
 
-    
-  } //Build()
-
-} // end namespace MueLu
+}  // end namespace MueLu
 
 #endif /* MUELU_REPARTITIONBLOCKDIAGONALFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_decl.hpp
index 81e3ea3d1cd9..d65ccd92ef6c 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_decl.hpp
@@ -68,106 +68,106 @@
 
 namespace MueLu {
 
-  /*!
-    @class RepartitionFactory class.
-    @brief Factory for building permutation matrix that can be be used to shuffle data (matrices, vectors) among processes
+/*!
+  @class RepartitionFactory class.
+  @brief Factory for building permutation matrix that can be be used to shuffle data (matrices, vectors) among processes
 
-    This factory acts on both the number of partitions and a vector (usually created by Zoltan) that indicates to which partitions
-    the current level's system matrix's DOFS belong.
+  This factory acts on both the number of partitions and a vector (usually created by Zoltan) that indicates to which partitions
+  the current level's system matrix's DOFS belong.
 
-    We always call the Interface routines in ZoltanInterface, Zoltan2Interface or IsorropiaInterface, but they have special short-cuts
-    if no rebalancing is necessary (then return "Partition = null") or all data is moved to one processor.
+  We always call the Interface routines in ZoltanInterface, Zoltan2Interface or IsorropiaInterface, but they have special short-cuts
+  if no rebalancing is necessary (then return "Partition = null") or all data is moved to one processor.
 
-    ## Input/output of RepartitionInterface ##
+  ## Input/output of RepartitionInterface ##
 
-    ### User parameters of RepartitionInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | number of partitions                   | Factory | null  |   | * | * | Factory calculating the "number of partitions" for rebalancing (usually an instance of RepartitionHeuristicFactory)
-    | Partition                              | Factory | null  |   | * | * | Factory generating the "Partition" variable containing the distribution of the DOFs over the reduced number of processors (given by "number of partitions")
-    | repartition: print partition distribution | bool | false | * | * |   | Partition distribution printout.
-    | repartition: remap parts                  | bool | false | * | * |   | Postprocessing for partitioning to reduce data migration.
-    | repartition: remap num values             | int  |  4    | * | * |   | Number of maximum components from each processor used to construct partial bipartite graph.
+  ### User parameters of RepartitionInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | number of partitions                   | Factory | null  |   | * | * | Factory calculating the "number of partitions" for rebalancing (usually an instance of RepartitionHeuristicFactory)
+  | Partition                              | Factory | null  |   | * | * | Factory generating the "Partition" variable containing the distribution of the DOFs over the reduced number of processors (given by "number of partitions")
+  | repartition: print partition distribution | bool | false | * | * |   | Partition distribution printout.
+  | repartition: remap parts                  | bool | false | * | * |   | Postprocessing for partitioning to reduce data migration.
+  | repartition: remap num values             | int  |  4    | * | * |   | Number of maximum components from each processor used to construct partial bipartite graph.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionFactory::DeclareInput).
 
-    ### Variables provided by RepartitionInterface ###
+  ### Variables provided by RepartitionInterface ###
 
-    After RepartitionFactory::Build the following data is available (if requested)
+  After RepartitionFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Importer | RepartitionFactory   | Importer used for rebalancing
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Importer | RepartitionFactory   | Importer used for rebalancing
 
-    Importer contains the Xpetra::Import object to rebalance the data (matrix or vectors). It is Teuchos::null if no rebalancing is necessary.
+  Importer contains the Xpetra::Import object to rebalance the data (matrix or vectors). It is Teuchos::null if no rebalancing is necessary.
 
-    @note For blocked systems you need one RepartionFactory for each sub block together with a separate Instance of an Interface class.
+  @note For blocked systems you need one RepartionFactory for each sub block together with a separate Instance of an Interface class.
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RepartitionFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RepartitionFactory : public SingleLevelFactoryBase {
 #undef MUELU_REPARTITIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RepartitionFactory() { }
+  //! Constructor.
+  RepartitionFactory() {}
 
-    //! Destructor.
-    virtual ~RepartitionFactory() { }
+  //! Destructor.
+  virtual ~RepartitionFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Determines the data that RepartitionFactory needs, and the factories that generate that data.
+  /*! @brief Determines the data that RepartitionFactory needs, and the factories that generate that data.
 
-        If this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level& currentLevel) const;
+      If this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level& currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Helper methods.
-    //@{
+  //! @name Helper methods.
+  //@{
 
-    /*!
-      @brief Determine which process should own each partition.
+  /*!
+    @brief Determine which process should own each partition.
 
-      Partitions are assigned to processes in order to minimize data movement.  The basic idea is that a good choice for partition
-      owner is to choose the pid that already has the greatest number of nonzeros for a particular partition. If willAcceptPartition=false
-      is set on a rank, no partition will be placed there.
+    Partitions are assigned to processes in order to minimize data movement.  The basic idea is that a good choice for partition
+    owner is to choose the pid that already has the greatest number of nonzeros for a particular partition. If willAcceptPartition=false
+    is set on a rank, no partition will be placed there.
 
-    */
-    void DeterminePartitionPlacement(const Matrix& A, GOVector& decomposition, GO numPartitions, bool willAcceptPartition=true, bool allSubdomainsAcceptPartitions=true) const;
+  */
+  void DeterminePartitionPlacement(const Matrix& A, GOVector& decomposition, GO numPartitions, bool willAcceptPartition = true, bool allSubdomainsAcceptPartitions = true) const;
 
-  }; // class RepartitionFactory
+};  // class RepartitionFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPARTITIONFACTORY_SHORT
 
-#endif //ifdef HAVE_MPI
-#endif // MUELU_REPARTITIONFACTORY_DECL_HPP
+#endif  // ifdef HAVE_MPI
+#endif  // MUELU_REPARTITIONFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_def.hpp
index 73b1b5291f1e..8276a93f95c5 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionFactory_def.hpp
@@ -50,7 +50,7 @@
 #include <iostream>
 #include <sstream>
 
-#include "MueLu_RepartitionFactory_decl.hpp" // TMP JG NOTE: before other includes, otherwise I cannot test the fwd declaration in _def
+#include "MueLu_RepartitionFactory_decl.hpp"  // TMP JG NOTE: before other includes, otherwise I cannot test the fwd declaration in _def
 
 #ifdef HAVE_MPI
 #include <Teuchos_DefaultMpiComm.hpp>
@@ -79,504 +79,502 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: print partition distribution");
-    SET_VALID_ENTRY("repartition: remap parts");
-    SET_VALID_ENTRY("repartition: remap num values");
-    SET_VALID_ENTRY("repartition: remap accept partition");
-    SET_VALID_ENTRY("repartition: node repartition level");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",                    Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >("Partition",            Teuchos::null, "Factory of the partition");
-
-    return validParamList;
+  SET_VALID_ENTRY("repartition: print partition distribution");
+  SET_VALID_ENTRY("repartition: remap parts");
+  SET_VALID_ENTRY("repartition: remap num values");
+  SET_VALID_ENTRY("repartition: remap accept partition");
+  SET_VALID_ENTRY("repartition: node repartition level");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("Partition", Teuchos::null, "Factory of the partition");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  Input(currentLevel, "Partition");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+  // Access parameters here to make sure that we set the parameter entry flag to "used" even in case of short-circuit evaluation.
+  // TODO (JG): I don't really know if we want to do this.
+  bool remapPartitions = pL.get<bool>("repartition: remap parts");
+
+  // TODO: We only need a CrsGraph. This class does not have to be templated on Scalar types.
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+  if (A == Teuchos::null) {
+    Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
+    return;
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    Input(currentLevel, "Partition");
+  RCP<const Map> rowMap     = A->getRowMap();
+  GO indexBase              = rowMap->getIndexBase();
+  Xpetra::UnderlyingLib lib = rowMap->lib();
+
+  RCP<const Teuchos::Comm<int> > origComm = rowMap->getComm();
+  RCP<const Teuchos::Comm<int> > comm     = origComm;
+
+  int myRank   = comm->getRank();
+  int numProcs = comm->getSize();
+
+  RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+  TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
+  RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
+
+  /////
+  int numPartitions = Get<int>(currentLevel, "number of partitions");
+
+  // ======================================================================================================
+  // Construct decomposition vector
+  // ======================================================================================================
+  RCP<GOVector> decomposition = Get<RCP<GOVector> >(currentLevel, "Partition");
+
+  // check which factory provides "Partition"
+  if (remapPartitions == true && Teuchos::rcp_dynamic_cast<const CloneRepartitionInterface>(GetFactory("Partition")) != Teuchos::null) {
+    // if "Partition" is provided by a CloneRepartitionInterface class we have to switch of remapPartitions
+    // as we can assume the processor Ids in Partition to be the expected ones. If we would do remapping we
+    // would get different processors for the different blocks which screws up matrix-matrix multiplication.
+    remapPartitions = false;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    const Teuchos::ParameterList & pL = GetParameterList();
-    // Access parameters here to make sure that we set the parameter entry flag to "used" even in case of short-circuit evaluation.
-    // TODO (JG): I don't really know if we want to do this.
-    bool   remapPartitions     = pL.get<bool>  ("repartition: remap parts");
-
-    // TODO: We only need a CrsGraph. This class does not have to be templated on Scalar types.
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-    if (A == Teuchos::null) {
-        Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
-        return;
-    }
-    RCP<const Map>            rowMap = A->getRowMap();
-    GO                     indexBase = rowMap->getIndexBase();
-    Xpetra::UnderlyingLib  lib       = rowMap->lib();
-
-    RCP<const Teuchos::Comm<int> > origComm = rowMap->getComm();
-    RCP<const Teuchos::Comm<int> > comm     = origComm;
-
-    int myRank   = comm->getRank();
-    int numProcs = comm->getSize();
-
-    RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-    TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
-    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
-
-    /////
-    int numPartitions = Get<int>(currentLevel, "number of partitions");
-
-    // ======================================================================================================
-    // Construct decomposition vector
-    // ======================================================================================================
-    RCP<GOVector> decomposition = Get<RCP<GOVector> >(currentLevel, "Partition");
-
-    // check which factory provides "Partition"
-    if(remapPartitions == true && Teuchos::rcp_dynamic_cast<const CloneRepartitionInterface>(GetFactory("Partition")) != Teuchos::null) {
-      // if "Partition" is provided by a CloneRepartitionInterface class we have to switch of remapPartitions
-      // as we can assume the processor Ids in Partition to be the expected ones. If we would do remapping we
-      // would get different processors for the different blocks which screws up matrix-matrix multiplication.
-      remapPartitions = false;
-    }
+  // check special cases
+  if (numPartitions == 1) {
+    // Trivial case: decomposition is the trivial one, all zeros. We skip the call to Zoltan_Interface
+    // (this is mostly done to avoid extra output messages, as even if we didn't skip there is a shortcut
+    // in Zoltan[12]Interface).
+    // TODO: We can probably skip more work in this case (like building all extra data structures)
+    GetOStream(Runtime0) << "Only one partition: Skip call to the repartitioner." << std::endl;
+  } else if (numPartitions == -1) {
+    // No repartitioning necessary: decomposition should be Teuchos::null
+    GetOStream(Runtime0) << "No repartitioning necessary: partitions were left unchanged by the repartitioner" << std::endl;
+    Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
+    return;
+  }
 
-    // check special cases
-    if (numPartitions == 1) {
-      // Trivial case: decomposition is the trivial one, all zeros. We skip the call to Zoltan_Interface
-      // (this is mostly done to avoid extra output messages, as even if we didn't skip there is a shortcut
-      // in Zoltan[12]Interface).
-      // TODO: We can probably skip more work in this case (like building all extra data structures)
-      GetOStream(Runtime0) << "Only one partition: Skip call to the repartitioner." << std::endl;
-    } else if (numPartitions == -1) {
-      // No repartitioning necessary: decomposition should be Teuchos::null
-      GetOStream(Runtime0) << "No repartitioning necessary: partitions were left unchanged by the repartitioner" << std::endl;
-      Set<RCP<const Import> >(currentLevel, "Importer", Teuchos::null);
-      return;
-    }
+  // If we're doing node away, we need to be sure to get the mapping to the NodeComm's rank 0.
+  const int nodeRepartLevel = pL.get<int>("repartition: node repartition level");
+  if (currentLevel.GetLevelID() == nodeRepartLevel) {
+    // NodePartitionInterface returns the *ranks* of the guy who gets the info, not the *partition number*
+    // In a sense, we've already done remap here.
 
-    // If we're doing node away, we need to be sure to get the mapping to the NodeComm's rank 0.
-    const int    nodeRepartLevel      = pL.get<int>   ("repartition: node repartition level");
-    if(currentLevel.GetLevelID() == nodeRepartLevel) {
-      // NodePartitionInterface returns the *ranks* of the guy who gets the info, not the *partition number*
-      // In a sense, we've already done remap here.
+    // FIXME: We need a low-comm import construction
+    remapPartitions = false;
+  }
 
-      // FIXME: We need a low-comm import construction
-      remapPartitions = false;
+  // ======================================================================================================
+  // Remap if necessary
+  // ======================================================================================================
+  // From a user perspective, we want user to not care about remapping, thinking of it as only a performance feature.
+  // There are two problems, however.
+  // (1) Next level aggregation depends on the order of GIDs in the vector, if one uses "natural" or "random" orderings.
+  //     This also means that remapping affects next level aggregation, despite the fact that the _set_ of GIDs for
+  //     each partition is the same.
+  // (2) Even with the fixed order of GIDs, the remapping may influence the aggregation for the next-next level.
+  //     Let us consider the following example. Lets assume that when we don't do remapping, processor 0 would have
+  //     GIDs {0,1,2}, and processor 1 GIDs {3,4,5}, and if we do remapping processor 0 would contain {3,4,5} and
+  //     processor 1 {0,1,2}. Now, when we run repartitioning algorithm on the next level (say Zoltan1 RCB), it may
+  //     be dependent on whether whether it is [{0,1,2}, {3,4,5}] or [{3,4,5}, {0,1,2}]. Specifically, the tie-breaking
+  //     algorithm can resolve these differently. For instance, running
+  //         mpirun -np 5 ./MueLu_ScalingTestParamList.exe --xml=easy_sa.xml --nx=12 --ny=12 --nz=12
+  //     with
+  //         <ParameterList name="MueLu">
+  //           <Parameter name="coarse: max size"                type="int"      value="1"/>
+  //           <Parameter name="repartition: enable"             type="bool"     value="true"/>
+  //           <Parameter name="repartition: min rows per proc"  type="int"      value="2"/>
+  //           <ParameterList name="level 1">
+  //             <Parameter name="repartition: remap parts"      type="bool"     value="false/true"/>
+  //           </ParameterList>
+  //         </ParameterList>
+  //     produces different repartitioning for level 2.
+  //     This different repartitioning may then escalate into different aggregation for the next level.
+  //
+  // We fix (1) by fixing the order of GIDs in a vector by sorting the resulting vector.
+  // Fixing (2) is more complicated.
+  // FIXME: Fixing (2) in Zoltan may not be enough, as we may use some arbitration in MueLu,
+  // for instance with CoupledAggregation. What we really need to do is to use the same order of processors containing
+  // the same order of GIDs. To achieve that, the newly created subcommunicator must be conforming with the order. For
+  // instance, if we have [{0,1,2}, {3,4,5}], we create a subcommunicator where processor 0 gets rank 0, and processor 1
+  // gets rank 1. If, on the other hand, we have [{3,4,5}, {0,1,2}], we assign rank 1 to processor 0, and rank 0 to processor 1.
+  // This rank permutation requires help from Epetra/Tpetra, both of which have no such API in place.
+  // One should also be concerned that if we had such API in place, rank 0 in subcommunicator may no longer be rank 0 in
+  // MPI_COMM_WORLD, which may lead to issues for logging.
+  if (remapPartitions) {
+    SubFactoryMonitor m1(*this, "DeterminePartitionPlacement", currentLevel);
+
+    bool acceptPartition = pL.get<bool>("repartition: remap accept partition");
+    bool allSubdomainsAcceptPartitions;
+    int localNumAcceptPartition = acceptPartition;
+    int globalNumAcceptPartition;
+    MueLu_sumAll(comm, localNumAcceptPartition, globalNumAcceptPartition);
+    GetOStream(Statistics2) << "Number of ranks that accept partitions: " << globalNumAcceptPartition << std::endl;
+    if (globalNumAcceptPartition < numPartitions) {
+      GetOStream(Warnings0) << "Not enough ranks are willing to accept a partition, allowing partitions on all ranks." << std::endl;
+      acceptPartition               = true;
+      allSubdomainsAcceptPartitions = true;
+    } else if (numPartitions > numProcs) {
+      // We are trying to repartition to a larger communicator.
+      allSubdomainsAcceptPartitions = true;
+    } else {
+      allSubdomainsAcceptPartitions = false;
     }
 
-    // ======================================================================================================
-    // Remap if necessary
-    // ======================================================================================================
-    // From a user perspective, we want user to not care about remapping, thinking of it as only a performance feature.
-    // There are two problems, however.
-    // (1) Next level aggregation depends on the order of GIDs in the vector, if one uses "natural" or "random" orderings.
-    //     This also means that remapping affects next level aggregation, despite the fact that the _set_ of GIDs for
-    //     each partition is the same.
-    // (2) Even with the fixed order of GIDs, the remapping may influence the aggregation for the next-next level.
-    //     Let us consider the following example. Lets assume that when we don't do remapping, processor 0 would have
-    //     GIDs {0,1,2}, and processor 1 GIDs {3,4,5}, and if we do remapping processor 0 would contain {3,4,5} and
-    //     processor 1 {0,1,2}. Now, when we run repartitioning algorithm on the next level (say Zoltan1 RCB), it may
-    //     be dependent on whether whether it is [{0,1,2}, {3,4,5}] or [{3,4,5}, {0,1,2}]. Specifically, the tie-breaking
-    //     algorithm can resolve these differently. For instance, running
-    //         mpirun -np 5 ./MueLu_ScalingTestParamList.exe --xml=easy_sa.xml --nx=12 --ny=12 --nz=12
-    //     with
-    //         <ParameterList name="MueLu">
-    //           <Parameter name="coarse: max size"                type="int"      value="1"/>
-    //           <Parameter name="repartition: enable"             type="bool"     value="true"/>
-    //           <Parameter name="repartition: min rows per proc"  type="int"      value="2"/>
-    //           <ParameterList name="level 1">
-    //             <Parameter name="repartition: remap parts"      type="bool"     value="false/true"/>
-    //           </ParameterList>
-    //         </ParameterList>
-    //     produces different repartitioning for level 2.
-    //     This different repartitioning may then escalate into different aggregation for the next level.
-    //
-    // We fix (1) by fixing the order of GIDs in a vector by sorting the resulting vector.
-    // Fixing (2) is more complicated.
-    // FIXME: Fixing (2) in Zoltan may not be enough, as we may use some arbitration in MueLu,
-    // for instance with CoupledAggregation. What we really need to do is to use the same order of processors containing
-    // the same order of GIDs. To achieve that, the newly created subcommunicator must be conforming with the order. For
-    // instance, if we have [{0,1,2}, {3,4,5}], we create a subcommunicator where processor 0 gets rank 0, and processor 1
-    // gets rank 1. If, on the other hand, we have [{3,4,5}, {0,1,2}], we assign rank 1 to processor 0, and rank 0 to processor 1.
-    // This rank permutation requires help from Epetra/Tpetra, both of which have no such API in place.
-    // One should also be concerned that if we had such API in place, rank 0 in subcommunicator may no longer be rank 0 in
-    // MPI_COMM_WORLD, which may lead to issues for logging.
-    if (remapPartitions) {
-      SubFactoryMonitor m1(*this, "DeterminePartitionPlacement", currentLevel);
-
-      bool acceptPartition = pL.get<bool>("repartition: remap accept partition");
-      bool allSubdomainsAcceptPartitions;
-      int localNumAcceptPartition = acceptPartition;
-      int globalNumAcceptPartition;
-      MueLu_sumAll(comm, localNumAcceptPartition, globalNumAcceptPartition);
-      GetOStream(Statistics2) << "Number of ranks that accept partitions: " << globalNumAcceptPartition << std::endl;
-      if (globalNumAcceptPartition < numPartitions) {
-        GetOStream(Warnings0) << "Not enough ranks are willing to accept a partition, allowing partitions on all ranks." << std::endl;
-        acceptPartition = true;
-        allSubdomainsAcceptPartitions = true;
-      } else if (numPartitions > numProcs) {
-        // We are trying to repartition to a larger communicator.
-        allSubdomainsAcceptPartitions = true;
-      } else {
-        allSubdomainsAcceptPartitions = false;
-      }
-
-      DeterminePartitionPlacement(*A, *decomposition, numPartitions, acceptPartition, allSubdomainsAcceptPartitions);
-    }
+    DeterminePartitionPlacement(*A, *decomposition, numPartitions, acceptPartition, allSubdomainsAcceptPartitions);
+  }
 
-    // ======================================================================================================
-    // Construct importer
-    // ======================================================================================================
-    // At this point, the following is true:
-    //  * Each processors owns 0 or 1 partitions
-    //  * If a processor owns a partition, that partition number is equal to the processor rank
-    //  * The decomposition vector contains the partitions ids that the corresponding GID belongs to
+  // ======================================================================================================
+  // Construct importer
+  // ======================================================================================================
+  // At this point, the following is true:
+  //  * Each processors owns 0 or 1 partitions
+  //  * If a processor owns a partition, that partition number is equal to the processor rank
+  //  * The decomposition vector contains the partitions ids that the corresponding GID belongs to
 
-    ArrayRCP<const GO> decompEntries;
-    if (decomposition->getLocalLength() > 0)
-      decompEntries = decomposition->getData(0);
+  ArrayRCP<const GO> decompEntries;
+  if (decomposition->getLocalLength() > 0)
+    decompEntries = decomposition->getData(0);
 
 #ifdef HAVE_MUELU_DEBUG
-    // Test range of partition ids
-    int incorrectRank = -1;
-    for (int i = 0; i < decompEntries.size(); i++)
-      if (decompEntries[i] >= numProcs || decompEntries[i] < 0) {
-        incorrectRank = myRank;
-        break;
-      }
+  // Test range of partition ids
+  int incorrectRank = -1;
+  for (int i = 0; i < decompEntries.size(); i++)
+    if (decompEntries[i] >= numProcs || decompEntries[i] < 0) {
+      incorrectRank = myRank;
+      break;
+    }
 
-    int incorrectGlobalRank = -1;
-    MueLu_maxAll(comm, incorrectRank, incorrectGlobalRank);
-    TEUCHOS_TEST_FOR_EXCEPTION(incorrectGlobalRank >- 1, Exceptions::RuntimeError, "pid " + Teuchos::toString(incorrectGlobalRank) + " encountered a partition number is that out-of-range");
+  int incorrectGlobalRank = -1;
+  MueLu_maxAll(comm, incorrectRank, incorrectGlobalRank);
+  TEUCHOS_TEST_FOR_EXCEPTION(incorrectGlobalRank > -1, Exceptions::RuntimeError, "pid " + Teuchos::toString(incorrectGlobalRank) + " encountered a partition number is that out-of-range");
 #endif
 
-    Array<GO> myGIDs;
-    myGIDs.reserve(decomposition->getLocalLength());
-
-    // Step 0: Construct mapping
-    //    part number -> GIDs I own which belong to this part
-    // NOTE: my own part GIDs are not part of the map
-    typedef std::map<GO, Array<GO> > map_type;
-    map_type sendMap;
-    for (LO i = 0; i < decompEntries.size(); i++) {
-      GO id  = decompEntries[i];
-      GO GID = rowMap->getGlobalElement(i);
-
-      if (id == myRank)
-        myGIDs     .push_back(GID);
-      else
-        sendMap[id].push_back(GID);
-    }
-    decompEntries = Teuchos::null;
+  Array<GO> myGIDs;
+  myGIDs.reserve(decomposition->getLocalLength());
+
+  // Step 0: Construct mapping
+  //    part number -> GIDs I own which belong to this part
+  // NOTE: my own part GIDs are not part of the map
+  typedef std::map<GO, Array<GO> > map_type;
+  map_type sendMap;
+  for (LO i = 0; i < decompEntries.size(); i++) {
+    GO id  = decompEntries[i];
+    GO GID = rowMap->getGlobalElement(i);
+
+    if (id == myRank)
+      myGIDs.push_back(GID);
+    else
+      sendMap[id].push_back(GID);
+  }
+  decompEntries = Teuchos::null;
 
-    if (IsPrint(Statistics2)) {
-      GO numLocalKept = myGIDs.size(), numGlobalKept, numGlobalRows = A->getGlobalNumRows();
-      MueLu_sumAll(comm,numLocalKept, numGlobalKept);
-      GetOStream(Statistics2) << "Unmoved rows: " << numGlobalKept << " / " << numGlobalRows << " (" << 100*Teuchos::as<double>(numGlobalKept)/numGlobalRows << "%)" << std::endl;
-    }
+  if (IsPrint(Statistics2)) {
+    GO numLocalKept = myGIDs.size(), numGlobalKept, numGlobalRows = A->getGlobalNumRows();
+    MueLu_sumAll(comm, numLocalKept, numGlobalKept);
+    GetOStream(Statistics2) << "Unmoved rows: " << numGlobalKept << " / " << numGlobalRows << " (" << 100 * Teuchos::as<double>(numGlobalKept) / numGlobalRows << "%)" << std::endl;
+  }
 
-    int numSend = sendMap.size(), numRecv;
-
-    // Arrayify map keys
-    Array<GO> myParts(numSend), myPart(1);
-    int cnt = 0;
-    myPart[0] = myRank;
-    for (typename map_type::const_iterator it = sendMap.begin(); it != sendMap.end(); it++)
-      myParts[cnt++] = it->first;
-
-    // Step 1: Find out how many processors send me data
-    // partsIndexBase starts from zero, as the processors ids start from zero
-    {
-      SubFactoryMonitor m1(*this, "Mapping Step 1", currentLevel);
-      GO partsIndexBase = 0;
-      RCP<Map>    partsIHave  = MapFactory   ::Build(lib, Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), myParts(), partsIndexBase, comm);
-      RCP<Map>    partsIOwn   = MapFactory   ::Build(lib,                                                 numProcs,  myPart(), partsIndexBase, comm);
-      RCP<Export> partsExport = ExportFactory::Build(partsIHave, partsIOwn);
-      
-      RCP<GOVector> partsISend    = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(partsIHave);
-      RCP<GOVector> numPartsIRecv = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(partsIOwn);
-      if (numSend) {
-        ArrayRCP<GO> partsISendData = partsISend->getDataNonConst(0);
-        for (int i = 0; i < numSend; i++)
-          partsISendData[i] = 1;
-      }
-      (numPartsIRecv->getDataNonConst(0))[0] = 0;
-      
-      numPartsIRecv->doExport(*partsISend, *partsExport, Xpetra::ADD);
-      numRecv = (numPartsIRecv->getData(0))[0];
+  int numSend = sendMap.size(), numRecv;
+
+  // Arrayify map keys
+  Array<GO> myParts(numSend), myPart(1);
+  int cnt   = 0;
+  myPart[0] = myRank;
+  for (typename map_type::const_iterator it = sendMap.begin(); it != sendMap.end(); it++)
+    myParts[cnt++] = it->first;
+
+  // Step 1: Find out how many processors send me data
+  // partsIndexBase starts from zero, as the processors ids start from zero
+  {
+    SubFactoryMonitor m1(*this, "Mapping Step 1", currentLevel);
+    GO partsIndexBase       = 0;
+    RCP<Map> partsIHave     = MapFactory ::Build(lib, Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), myParts(), partsIndexBase, comm);
+    RCP<Map> partsIOwn      = MapFactory ::Build(lib, numProcs, myPart(), partsIndexBase, comm);
+    RCP<Export> partsExport = ExportFactory::Build(partsIHave, partsIOwn);
+
+    RCP<GOVector> partsISend    = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(partsIHave);
+    RCP<GOVector> numPartsIRecv = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(partsIOwn);
+    if (numSend) {
+      ArrayRCP<GO> partsISendData = partsISend->getDataNonConst(0);
+      for (int i = 0; i < numSend; i++)
+        partsISendData[i] = 1;
     }
+    (numPartsIRecv->getDataNonConst(0))[0] = 0;
 
+    numPartsIRecv->doExport(*partsISend, *partsExport, Xpetra::ADD);
+    numRecv = (numPartsIRecv->getData(0))[0];
+  }
 
-    // Step 2: Get my GIDs from everybody else
-    MPI_Datatype MpiType = Teuchos::Details::MpiTypeTraits<GO>::getType();
-    int msgTag = 12345;  // TODO: use Comm::dup for all internal messaging
+  // Step 2: Get my GIDs from everybody else
+  MPI_Datatype MpiType = Teuchos::Details::MpiTypeTraits<GO>::getType();
+  int msgTag           = 12345;  // TODO: use Comm::dup for all internal messaging
 
-    // Post sends
-    Array<MPI_Request> sendReqs(numSend);
-    cnt = 0;
-    for (typename map_type::iterator it = sendMap.begin(); it != sendMap.end(); it++)
-      MPI_Isend(static_cast<void*>(it->second.getRawPtr()), it->second.size(), MpiType, Teuchos::as<GO>(it->first), msgTag, *rawMpiComm, &sendReqs[cnt++]);
+  // Post sends
+  Array<MPI_Request> sendReqs(numSend);
+  cnt = 0;
+  for (typename map_type::iterator it = sendMap.begin(); it != sendMap.end(); it++)
+    MPI_Isend(static_cast<void*>(it->second.getRawPtr()), it->second.size(), MpiType, Teuchos::as<GO>(it->first), msgTag, *rawMpiComm, &sendReqs[cnt++]);
 
-    map_type recvMap;
-    size_t totalGIDs = myGIDs.size();
-    for (int i = 0; i < numRecv; i++) {
-      MPI_Status status;
-      MPI_Probe(MPI_ANY_SOURCE, msgTag, *rawMpiComm, &status);
+  map_type recvMap;
+  size_t totalGIDs = myGIDs.size();
+  for (int i = 0; i < numRecv; i++) {
+    MPI_Status status;
+    MPI_Probe(MPI_ANY_SOURCE, msgTag, *rawMpiComm, &status);
 
-      // Get rank and number of elements from status
-      int fromRank = status.MPI_SOURCE, count;
-      MPI_Get_count(&status, MpiType, &count);
+    // Get rank and number of elements from status
+    int fromRank = status.MPI_SOURCE, count;
+    MPI_Get_count(&status, MpiType, &count);
 
-      recvMap[fromRank].resize(count);
-      MPI_Recv(static_cast<void*>(recvMap[fromRank].getRawPtr()), count, MpiType, fromRank, msgTag, *rawMpiComm, &status);
+    recvMap[fromRank].resize(count);
+    MPI_Recv(static_cast<void*>(recvMap[fromRank].getRawPtr()), count, MpiType, fromRank, msgTag, *rawMpiComm, &status);
 
-      totalGIDs += count;
-    }
+    totalGIDs += count;
+  }
 
-    // Do waits on send requests
-    if (numSend) {
-      Array<MPI_Status> sendStatuses(numSend);
-      MPI_Waitall(numSend, sendReqs.getRawPtr(), sendStatuses.getRawPtr());
-    }
+  // Do waits on send requests
+  if (numSend) {
+    Array<MPI_Status> sendStatuses(numSend);
+    MPI_Waitall(numSend, sendReqs.getRawPtr(), sendStatuses.getRawPtr());
+  }
 
-    // Merge GIDs
-    myGIDs.reserve(totalGIDs);
-    for (typename map_type::const_iterator it = recvMap.begin(); it != recvMap.end(); it++) {
-      int offset = myGIDs.size(), len = it->second.size();
-      if (len) {
-        myGIDs.resize(offset + len);
-        memcpy(myGIDs.getRawPtr() + offset, it->second.getRawPtr(), len*sizeof(GO));
-      }
-    }
-    // NOTE 2: The general sorting algorithm could be sped up by using the knowledge that original myGIDs and all received chunks
-    // (i.e. it->second) are sorted. Therefore, a merge sort would work well in this situation.
-    std::sort(myGIDs.begin(), myGIDs.end());
-
-    // Step 3: Construct importer
-    RCP<Map>          newRowMap;
-    {
-      SubFactoryMonitor m1(*this, "Map construction", currentLevel);
-      newRowMap = MapFactory   ::Build(lib, rowMap->getGlobalNumElements(), myGIDs(), indexBase, origComm);
+  // Merge GIDs
+  myGIDs.reserve(totalGIDs);
+  for (typename map_type::const_iterator it = recvMap.begin(); it != recvMap.end(); it++) {
+    int offset = myGIDs.size(), len = it->second.size();
+    if (len) {
+      myGIDs.resize(offset + len);
+      memcpy(myGIDs.getRawPtr() + offset, it->second.getRawPtr(), len * sizeof(GO));
     }
+  }
+  // NOTE 2: The general sorting algorithm could be sped up by using the knowledge that original myGIDs and all received chunks
+  // (i.e. it->second) are sorted. Therefore, a merge sort would work well in this situation.
+  std::sort(myGIDs.begin(), myGIDs.end());
+
+  // Step 3: Construct importer
+  RCP<Map> newRowMap;
+  {
+    SubFactoryMonitor m1(*this, "Map construction", currentLevel);
+    newRowMap = MapFactory ::Build(lib, rowMap->getGlobalNumElements(), myGIDs(), indexBase, origComm);
+  }
 
-    RCP<const Import> rowMapImporter;
+  RCP<const Import> rowMapImporter;
 
-    RCP<const BlockedMap> blockedRowMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
+  RCP<const BlockedMap> blockedRowMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
 
-    {
-      SubFactoryMonitor m1(*this, "Import construction", currentLevel);
-      // Generate a nonblocked rowmap if we need one
-      if(blockedRowMap.is_null())
-	rowMapImporter = ImportFactory::Build(rowMap, newRowMap);
-      else {
-	rowMapImporter = ImportFactory::Build(blockedRowMap->getMap(), newRowMap);
-      }
+  {
+    SubFactoryMonitor m1(*this, "Import construction", currentLevel);
+    // Generate a nonblocked rowmap if we need one
+    if (blockedRowMap.is_null())
+      rowMapImporter = ImportFactory::Build(rowMap, newRowMap);
+    else {
+      rowMapImporter = ImportFactory::Build(blockedRowMap->getMap(), newRowMap);
     }
+  }
 
-    // If we're running BlockedCrs we should chop up the newRowMap into a newBlockedRowMap here (and do likewise for importers)
-    if(!blockedRowMap.is_null()) {
-      SubFactoryMonitor m1(*this, "Blocking newRowMap and Importer", currentLevel);
-      RCP<const BlockedMap> blockedTargetMap = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GeneratedBlockedTargetMap(*blockedRowMap,*rowMapImporter);
+  // If we're running BlockedCrs we should chop up the newRowMap into a newBlockedRowMap here (and do likewise for importers)
+  if (!blockedRowMap.is_null()) {
+    SubFactoryMonitor m1(*this, "Blocking newRowMap and Importer", currentLevel);
+    RCP<const BlockedMap> blockedTargetMap = MueLu::UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GeneratedBlockedTargetMap(*blockedRowMap, *rowMapImporter);
 
-      // NOTE: This code qualifies as "correct but not particularly performant"  If this needs to be sped up, we can probably read data from the existing importer to
-      // build sub-importers rather than generating new ones ex nihilo
-      size_t numBlocks = blockedRowMap->getNumMaps();
-      std::vector<RCP<const Import> > subImports(numBlocks);
+    // NOTE: This code qualifies as "correct but not particularly performant"  If this needs to be sped up, we can probably read data from the existing importer to
+    // build sub-importers rather than generating new ones ex nihilo
+    size_t numBlocks = blockedRowMap->getNumMaps();
+    std::vector<RCP<const Import> > subImports(numBlocks);
 
-      for(size_t i=0; i<numBlocks; i++) {
-	RCP<const Map> source = blockedRowMap->getMap(i);
-	RCP<const Map> target = blockedTargetMap->getMap(i);
-	subImports[i] = ImportFactory::Build(source,target);
-      }
-      Set(currentLevel,"SubImporters",subImports);	
+    for (size_t i = 0; i < numBlocks; i++) {
+      RCP<const Map> source = blockedRowMap->getMap(i);
+      RCP<const Map> target = blockedTargetMap->getMap(i);
+      subImports[i]         = ImportFactory::Build(source, target);
     }
+    Set(currentLevel, "SubImporters", subImports);
+  }
 
+  Set(currentLevel, "Importer", rowMapImporter);
 
-    Set(currentLevel, "Importer", rowMapImporter);
-
-    // ======================================================================================================
-    // Print some data
-    // ======================================================================================================
-    if (!rowMapImporter.is_null() && IsPrint(Statistics2)) {
-      // int oldRank = SetProcRankVerbose(rebalancedAc->getRowMap()->getComm()->getRank());
-      GetOStream(Statistics2) << PerfUtils::PrintImporterInfo(rowMapImporter, "Importer for rebalancing");
-      // SetProcRankVerbose(oldRank);
-    }
-    if (pL.get<bool>("repartition: print partition distribution") && IsPrint(Statistics2)) {
-      // Print the grid of processors
-      GetOStream(Statistics2) << "Partition distribution over cores (ownership is indicated by '+')" << std::endl;
-
-      char amActive = (myGIDs.size() ? 1 : 0);
-      std::vector<char> areActive(numProcs, 0);
-      MPI_Gather(&amActive, 1, MPI_CHAR, &areActive[0], 1, MPI_CHAR, 0, *rawMpiComm);
-
-      int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
-      for (int proc = 0; proc < numProcs; proc += rowWidth) {
-        for (int j = 0; j < rowWidth; j++)
-          if (proc + j < numProcs)
-            GetOStream(Statistics2) << (areActive[proc + j] ? "+" : ".");
-          else
+  // ======================================================================================================
+  // Print some data
+  // ======================================================================================================
+  if (!rowMapImporter.is_null() && IsPrint(Statistics2)) {
+    // int oldRank = SetProcRankVerbose(rebalancedAc->getRowMap()->getComm()->getRank());
+    GetOStream(Statistics2) << PerfUtils::PrintImporterInfo(rowMapImporter, "Importer for rebalancing");
+    // SetProcRankVerbose(oldRank);
+  }
+  if (pL.get<bool>("repartition: print partition distribution") && IsPrint(Statistics2)) {
+    // Print the grid of processors
+    GetOStream(Statistics2) << "Partition distribution over cores (ownership is indicated by '+')" << std::endl;
+
+    char amActive = (myGIDs.size() ? 1 : 0);
+    std::vector<char> areActive(numProcs, 0);
+    MPI_Gather(&amActive, 1, MPI_CHAR, &areActive[0], 1, MPI_CHAR, 0, *rawMpiComm);
+
+    int rowWidth = std::min(Teuchos::as<int>(ceil(sqrt(numProcs))), 100);
+    for (int proc = 0; proc < numProcs; proc += rowWidth) {
+      for (int j = 0; j < rowWidth; j++)
+        if (proc + j < numProcs)
+          GetOStream(Statistics2) << (areActive[proc + j] ? "+" : ".");
+        else
           GetOStream(Statistics2) << " ";
 
-        GetOStream(Statistics2) << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
-      }
+      GetOStream(Statistics2) << "      " << proc << ":" << std::min(proc + rowWidth, numProcs) - 1 << std::endl;
     }
+  }
 
-  } // Build
-
-  //----------------------------------------------------------------------
-  template<typename T, typename W>
-  struct Triplet {
-    T    i, j;
-    W    v;
-  };
-  template<typename T, typename W>
-  static bool compareTriplets(const Triplet<T,W>& a, const Triplet<T,W>& b) {
-    return (a.v > b.v); // descending order
+}  // Build
+
+//----------------------------------------------------------------------
+template <typename T, typename W>
+struct Triplet {
+  T i, j;
+  W v;
+};
+template <typename T, typename W>
+static bool compareTriplets(const Triplet<T, W>& a, const Triplet<T, W>& b) {
+  return (a.v > b.v);  // descending order
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeterminePartitionPlacement(const Matrix& A, GOVector& decomposition, GO numPartitions, bool willAcceptPartition, bool allSubdomainsAcceptPartitions) const {
+  RCP<const Map> rowMap = A.getRowMap();
+
+  RCP<const Teuchos::Comm<int> > comm = rowMap->getComm()->duplicate();
+  int numProcs                        = comm->getSize();
+
+  RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
+  TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
+  RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  // maxLocal is a constant which determins the number of largest edges which are being exchanged
+  // The idea is that we do not want to construct the full bipartite graph, but simply a subset of
+  // it, which requires less communication. By selecting largest local edges we hope to achieve
+  // similar results but at a lower cost.
+  const int maxLocal = pL.get<int>("repartition: remap num values");
+  const int dataSize = 2 * maxLocal;
+
+  ArrayRCP<GO> decompEntries;
+  if (decomposition.getLocalLength() > 0)
+    decompEntries = decomposition.getDataNonConst(0);
+
+  // Step 1: Sort local edges by weight
+  // Each edge of a bipartite graph corresponds to a triplet (i, j, v) where
+  //   i: processor id that has some piece of part with part_id = j
+  //   j: part id
+  //   v: weight of the edge
+  // We set edge weights to be the total number of nonzeros in rows on this processor which
+  // correspond to this part_id. The idea is that when we redistribute matrix, this weight
+  // is a good approximation of the amount of data to move.
+  // We use two maps, original which maps a partition id of an edge to the corresponding weight,
+  // and a reverse one, which is necessary to sort by edges.
+  std::map<GO, GO> lEdges;
+  if (willAcceptPartition)
+    for (LO i = 0; i < decompEntries.size(); i++)
+      lEdges[decompEntries[i]] += A.getNumEntriesInLocalRow(i);
+
+  // Reverse map, so that edges are sorted by weight.
+  // This results in multimap, as we may have edges with the same weight
+  std::multimap<GO, GO> revlEdges;
+  for (typename std::map<GO, GO>::const_iterator it = lEdges.begin(); it != lEdges.end(); it++)
+    revlEdges.insert(std::make_pair(it->second, it->first));
+
+  // Both lData and gData are arrays of data which we communicate. The data is stored
+  // in pairs, so that data[2*i+0] is the part index, and data[2*i+1] is the corresponding edge weight.
+  // We do not store processor id in data, as we can compute that by looking on the offset in the gData.
+  Array<GO> lData(dataSize, -1), gData(numProcs * dataSize);
+  int numEdges = 0;
+  for (typename std::multimap<GO, GO>::reverse_iterator rit = revlEdges.rbegin(); rit != revlEdges.rend() && numEdges < maxLocal; rit++) {
+    lData[2 * numEdges + 0] = rit->second;  // part id
+    lData[2 * numEdges + 1] = rit->first;   // edge weight
+    numEdges++;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeterminePartitionPlacement(const Matrix& A, GOVector& decomposition, GO numPartitions, bool willAcceptPartition, bool allSubdomainsAcceptPartitions) const {
-    RCP<const Map> rowMap = A.getRowMap();
-
-    RCP<const Teuchos::Comm<int> > comm = rowMap->getComm()->duplicate();
-    int numProcs = comm->getSize();
-
-    RCP<const Teuchos::MpiComm<int> > tmpic = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(comm);
-    TEUCHOS_TEST_FOR_EXCEPTION(tmpic == Teuchos::null, Exceptions::RuntimeError, "Cannot cast base Teuchos::Comm to Teuchos::MpiComm object.");
-    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > rawMpiComm = tmpic->getRawMpiComm();
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-
-    // maxLocal is a constant which determins the number of largest edges which are being exchanged
-    // The idea is that we do not want to construct the full bipartite graph, but simply a subset of
-    // it, which requires less communication. By selecting largest local edges we hope to achieve
-    // similar results but at a lower cost.
-    const int maxLocal = pL.get<int>("repartition: remap num values");
-    const int dataSize = 2*maxLocal;
-
-    ArrayRCP<GO> decompEntries;
-    if (decomposition.getLocalLength() > 0)
-      decompEntries = decomposition.getDataNonConst(0);
-
-    // Step 1: Sort local edges by weight
-    // Each edge of a bipartite graph corresponds to a triplet (i, j, v) where
-    //   i: processor id that has some piece of part with part_id = j
-    //   j: part id
-    //   v: weight of the edge
-    // We set edge weights to be the total number of nonzeros in rows on this processor which
-    // correspond to this part_id. The idea is that when we redistribute matrix, this weight
-    // is a good approximation of the amount of data to move.
-    // We use two maps, original which maps a partition id of an edge to the corresponding weight,
-    // and a reverse one, which is necessary to sort by edges.
-    std::map<GO,GO> lEdges;
-    if (willAcceptPartition)
-      for (LO i = 0; i < decompEntries.size(); i++)
-        lEdges[decompEntries[i]] += A.getNumEntriesInLocalRow(i);
-
-    // Reverse map, so that edges are sorted by weight.
-    // This results in multimap, as we may have edges with the same weight
-    std::multimap<GO,GO> revlEdges;
-    for (typename std::map<GO,GO>::const_iterator it = lEdges.begin(); it != lEdges.end(); it++)
-      revlEdges.insert(std::make_pair(it->second, it->first));
-
-    // Both lData and gData are arrays of data which we communicate. The data is stored
-    // in pairs, so that data[2*i+0] is the part index, and data[2*i+1] is the corresponding edge weight.
-    // We do not store processor id in data, as we can compute that by looking on the offset in the gData.
-    Array<GO> lData(dataSize, -1), gData(numProcs * dataSize);
-    int numEdges = 0;
-    for (typename std::multimap<GO,GO>::reverse_iterator rit = revlEdges.rbegin(); rit != revlEdges.rend() && numEdges < maxLocal; rit++) {
-      lData[2*numEdges+0] = rit->second; // part id
-      lData[2*numEdges+1] = rit->first;  // edge weight
-      numEdges++;
+  // Step 2: Gather most edges
+  // Each processors contributes maxLocal edges by providing maxLocal pairs <part id, weight>, which is of size dataSize
+  MPI_Datatype MpiType = Teuchos::Details::MpiTypeTraits<GO>::getType();
+  MPI_Allgather(static_cast<void*>(lData.getRawPtr()), dataSize, MpiType, static_cast<void*>(gData.getRawPtr()), dataSize, MpiType, *rawMpiComm);
+
+  // Step 3: Construct mapping
+
+  // Construct the set of triplets
+  Teuchos::Array<Triplet<int, int> > gEdges(numProcs * maxLocal);
+  Teuchos::Array<bool> procWillAcceptPartition(numProcs, allSubdomainsAcceptPartitions);
+  size_t k = 0;
+  for (LO i = 0; i < gData.size(); i += 2) {
+    int procNo = i / dataSize;  // determine the processor by its offset (since every processor sends the same amount)
+    GO part    = gData[i + 0];
+    GO weight  = gData[i + 1];
+    if (part != -1) {  // skip nonexistent edges
+      gEdges[k].i                     = procNo;
+      gEdges[k].j                     = part;
+      gEdges[k].v                     = weight;
+      procWillAcceptPartition[procNo] = true;
+      k++;
     }
-
-    // Step 2: Gather most edges
-    // Each processors contributes maxLocal edges by providing maxLocal pairs <part id, weight>, which is of size dataSize
-    MPI_Datatype MpiType = Teuchos::Details::MpiTypeTraits<GO>::getType();
-    MPI_Allgather(static_cast<void*>(lData.getRawPtr()), dataSize, MpiType, static_cast<void*>(gData.getRawPtr()), dataSize, MpiType, *rawMpiComm);
-
-    // Step 3: Construct mapping
-
-    // Construct the set of triplets
-    Teuchos::Array<Triplet<int,int> > gEdges(numProcs * maxLocal);
-    Teuchos::Array<bool> procWillAcceptPartition(numProcs, allSubdomainsAcceptPartitions);
-    size_t k = 0;
-    for (LO i = 0; i < gData.size(); i += 2) {
-      int procNo = i/dataSize;              // determine the processor by its offset (since every processor sends the same amount)
-      GO part   = gData[i+0];
-      GO weight = gData[i+1];
-      if (part != -1) {                     // skip nonexistent edges
-        gEdges[k].i = procNo;
-        gEdges[k].j = part;
-        gEdges[k].v = weight;
-        procWillAcceptPartition[procNo] = true;
-        k++;
-      }
+  }
+  gEdges.resize(k);
+
+  // Sort edges by weight
+  // NOTE: compareTriplets is actually a reverse sort, so the edges weight is in decreasing order
+  std::sort(gEdges.begin(), gEdges.end(), compareTriplets<int, int>);
+
+  // Do matching
+  std::map<int, int> match;
+  Teuchos::Array<char> matchedRanks(numProcs, 0), matchedParts(numPartitions, 0);
+  int numMatched = 0;
+  for (typename Teuchos::Array<Triplet<int, int> >::const_iterator it = gEdges.begin(); it != gEdges.end(); it++) {
+    GO rank = it->i;
+    GO part = it->j;
+    if (matchedRanks[rank] == 0 && matchedParts[part] == 0) {
+      matchedRanks[rank] = 1;
+      matchedParts[part] = 1;
+      match[part]        = rank;
+      numMatched++;
     }
-    gEdges.resize(k);
-
-    // Sort edges by weight
-    // NOTE: compareTriplets is actually a reverse sort, so the edges weight is in decreasing order
-    std::sort(gEdges.begin(), gEdges.end(), compareTriplets<int,int>);
-
-    // Do matching
-    std::map<int,int> match;
-    Teuchos::Array<char> matchedRanks(numProcs, 0), matchedParts(numPartitions, 0);
-    int numMatched = 0;
-    for (typename Teuchos::Array<Triplet<int,int> >::const_iterator it = gEdges.begin(); it != gEdges.end(); it++) {
-      GO rank = it->i;
-      GO part = it->j;
-      if (matchedRanks[rank] == 0 && matchedParts[part] == 0) {
-        matchedRanks[rank] = 1;
-        matchedParts[part] = 1;
-        match[part] = rank;
+  }
+  GetOStream(Statistics1) << "Number of unassigned partitions before cleanup stage: " << (numPartitions - numMatched) << " / " << numPartitions << std::endl;
+
+  // Step 4: Assign unassigned partitions if necessary.
+  // We do that through desperate matching for remaining partitions:
+  // We select the lowest rank that can still take a partition.
+  // The reason it is done this way is that we don't need any extra communication, as we don't
+  // need to know which parts are valid.
+  if (numPartitions - numMatched > 0) {
+    Teuchos::Array<char> partitionCounts(numPartitions, 0);
+    for (typename std::map<int, int>::const_iterator it = match.begin(); it != match.end(); it++)
+      partitionCounts[it->first] += 1;
+    for (int part = 0, matcher = 0; part < numPartitions; part++) {
+      if (partitionCounts[part] == 0) {
+        // Find first non-matched rank that accepts partitions
+        while (matchedRanks[matcher] || !procWillAcceptPartition[matcher])
+          matcher++;
+
+        match[part] = matcher++;
         numMatched++;
       }
     }
-    GetOStream(Statistics1) << "Number of unassigned partitions before cleanup stage: " << (numPartitions - numMatched) << " / " << numPartitions << std::endl;
-
-    // Step 4: Assign unassigned partitions if necessary.
-    // We do that through desperate matching for remaining partitions:
-    // We select the lowest rank that can still take a partition.
-    // The reason it is done this way is that we don't need any extra communication, as we don't
-    // need to know which parts are valid.
-    if (numPartitions - numMatched > 0) {
-      Teuchos::Array<char> partitionCounts(numPartitions, 0);
-      for (typename std::map<int,int>::const_iterator it = match.begin(); it != match.end(); it++)
-        partitionCounts[it->first] += 1;
-      for (int part = 0, matcher = 0; part < numPartitions; part++) {
-        if (partitionCounts[part] == 0) {
-          // Find first non-matched rank that accepts partitions
-          while (matchedRanks[matcher] || !procWillAcceptPartition[matcher])
-            matcher++;
-
-          match[part] = matcher++;
-          numMatched++;
-        }
-      }
-    }
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(numMatched != numPartitions, Exceptions::RuntimeError, "MueLu::RepartitionFactory::DeterminePartitionPlacement: Only " << numMatched << " partitions out of " << numPartitions << " got assigned to ranks.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numMatched != numPartitions, Exceptions::RuntimeError, "MueLu::RepartitionFactory::DeterminePartitionPlacement: Only " << numMatched << " partitions out of " << numPartitions << " got assigned to ranks.");
 
-    // Step 5: Permute entries in the decomposition vector
-    for (LO i = 0; i < decompEntries.size(); i++)
-      decompEntries[i] = match[decompEntries[i]];
-  }
+  // Step 5: Permute entries in the decomposition vector
+  for (LO i = 0; i < decompEntries.size(); i++)
+    decompEntries[i] = match[decompEntries[i]];
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif //ifdef HAVE_MPI
+#endif  // ifdef HAVE_MPI
 
-#endif // MUELU_REPARTITIONFACTORY_DEF_HPP
+#endif  // MUELU_REPARTITIONFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_decl.hpp
index 0fbea3e48d04..148f257e9bdf 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_decl.hpp
@@ -63,96 +63,96 @@
 
 namespace MueLu {
 
-  /*!
-    @class RepartitionHeuristicFactory class.
-    @brief Factory for determing the number of partitions for rebalancing
+/*!
+  @class RepartitionHeuristicFactory class.
+  @brief Factory for determing the number of partitions for rebalancing
 
-    Given the unrebalanced (coarse) matrix A and some user parameters we determine the number
-    of partitions (processes) the (rebalanced) matrix should use. The information is used as input
-    for the Interface classes which perform the actual Zoltan call to determine the partition placement
-    on the processors.
+  Given the unrebalanced (coarse) matrix A and some user parameters we determine the number
+  of partitions (processes) the (rebalanced) matrix should use. The information is used as input
+  for the Interface classes which perform the actual Zoltan call to determine the partition placement
+  on the processors.
 
-    @note A must be a "non-rebalanced" matrix generated by RAPFactory, SubBlockAFactory or BlockedRAPFactory
+  @note A must be a "non-rebalanced" matrix generated by RAPFactory, SubBlockAFactory or BlockedRAPFactory
 
-    @note For blocked operators you have only one RepartitionHeuristicInterface instance. But you need an
-    Interface class for each sub block on the diagonal of the input matrix.
+  @note For blocked operators you have only one RepartitionHeuristicInterface instance. But you need an
+  Interface class for each sub block on the diagonal of the input matrix.
 
-    ## Input/output of RepartitionHeuristicInterface ##
+  ## Input/output of RepartitionHeuristicInterface ##
 
-    ### User parameters of RepartitionInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | repartition: start level | int | 2 | * | * | | Minimum level to run partitioner. MueLu does not rebalance levels finer than this one.
-    | repartition: min rows per proc | LO      | 800 | * | * |  | Minimum number of rows per processor. If the actual number if smaller, then rebalancing occurs.
-    | repartition: max imbalance | double      | 1.2 | * | * |  | Maximum nonzero imbalance ratio. If the actual number is larger, the rebalancing occurs.
+  ### User parameters of RepartitionInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | repartition: start level | int | 2 | * | * | | Minimum level to run partitioner. MueLu does not rebalance levels finer than this one.
+  | repartition: min rows per proc | LO      | 800 | * | * |  | Minimum number of rows per processor. If the actual number if smaller, then rebalancing occurs.
+  | repartition: max imbalance | double      | 1.2 | * | * |  | Maximum nonzero imbalance ratio. If the actual number is larger, the rebalancing occurs.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionHeuristicFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionHeuristicFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionHeuristicFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionHeuristicFactory::DeclareInput).
 
-    ### Variables provided by RepartitionHeuristicFactory ###
+  ### Variables provided by RepartitionHeuristicFactory ###
 
-    After RepartitionHeuristicFactory::Build the following data is available (if requested)
+  After RepartitionHeuristicFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | number of partitions | RepartitionHeuristicFactory   | Integer containing the number of partitions to be built. See below for details
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | number of partitions | RepartitionHeuristicFactory   | Integer containing the number of partitions to be built. See below for details
 
-    The "number of partitions" variable contains the requested number of partitions for rebalancing. It is needed as input by the Interface classes
-    (see, e.g., ZoltanInterface, Zoltan2Interface, IsorropiaInterface) and the RepartitionFactory. The Interface classes use it as input for the Zoltan
-    call which considers the information when finding the rebalanced partition.
+  The "number of partitions" variable contains the requested number of partitions for rebalancing. It is needed as input by the Interface classes
+  (see, e.g., ZoltanInterface, Zoltan2Interface, IsorropiaInterface) and the RepartitionFactory. The Interface classes use it as input for the Zoltan
+  call which considers the information when finding the rebalanced partition.
 
-    If "number of partitions" is 1 (i.e., move all data to one processor), we can skip the call to the Interface classes (see RepartitionFactory).
-    If "number of partitions" is -1 we don't perform rebalancing (just keep the partition unchanged).
+  If "number of partitions" is 1 (i.e., move all data to one processor), we can skip the call to the Interface classes (see RepartitionFactory).
+  If "number of partitions" is -1 we don't perform rebalancing (just keep the partition unchanged).
 
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RepartitionHeuristicFactory : public SingleLevelFactoryBase {
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RepartitionHeuristicFactory : public SingleLevelFactoryBase {
 #undef MUELU_REPARTITIONHEURISTICFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RepartitionHeuristicFactory() { }
+  //! Constructor.
+  RepartitionHeuristicFactory() {}
 
-    //! Destructor.
-    virtual ~RepartitionHeuristicFactory() { }
+  //! Destructor.
+  virtual ~RepartitionHeuristicFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Determines the data that RepartitionHeuristicFactory needs, and the factories that generate that data.
+  /*! @brief Determines the data that RepartitionHeuristicFactory needs, and the factories that generate that data.
 
-        If this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level& currentLevel) const;
+      If this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level& currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-    //@}
-  }; // class RepartitionHeuristicFactory
+  //@}
+};  // class RepartitionHeuristicFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPARTITIONHEURISTICFACTORY_SHORT
 
-#endif //ifdef HAVE_MPI
-#endif /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_REPARTITIONHEURISTICFACTORY_DECL_HPP_ */
+#endif  // ifdef HAVE_MPI
+#endif  /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_REPARTITIONHEURISTICFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_def.hpp
index 6a95deca39c2..e3fffda905f2 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionHeuristicFactory_def.hpp
@@ -58,7 +58,6 @@
 //#include <Xpetra_Map.hpp>
 #include <Xpetra_Matrix.hpp>
 
-
 #include "MueLu_RAPFactory.hpp"
 #include "MueLu_BlockedRAPFactory.hpp"
 #include "MueLu_SubBlockAFactory.hpp"
@@ -70,250 +69,248 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: start level");
-    SET_VALID_ENTRY("repartition: use map");
-    SET_VALID_ENTRY("repartition: node repartition level");
-    SET_VALID_ENTRY("repartition: min rows per proc");
-    SET_VALID_ENTRY("repartition: target rows per proc");
-    SET_VALID_ENTRY("repartition: min rows per thread");
-    SET_VALID_ENTRY("repartition: target rows per thread");
-    SET_VALID_ENTRY("repartition: max imbalance");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Map", Teuchos::null, "Factory of the map Map");
-    validParamList->set< RCP<const FactoryBase> >("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    const Teuchos::ParameterList & pL = GetParameterList();
-    if(pL.isParameter("repartition: use map")) {
-      const bool useMap = pL.get<bool>("repartition: use map");
-      if (useMap)
-        Input(currentLevel, "Map");
-      else
-        Input(currentLevel, "A");
-    } else
+  SET_VALID_ENTRY("repartition: start level");
+  SET_VALID_ENTRY("repartition: use map");
+  SET_VALID_ENTRY("repartition: node repartition level");
+  SET_VALID_ENTRY("repartition: min rows per proc");
+  SET_VALID_ENTRY("repartition: target rows per proc");
+  SET_VALID_ENTRY("repartition: min rows per thread");
+  SET_VALID_ENTRY("repartition: target rows per thread");
+  SET_VALID_ENTRY("repartition: max imbalance");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Map", Teuchos::null, "Factory of the map Map");
+  validParamList->set<RCP<const FactoryBase> >("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const Teuchos::ParameterList& pL = GetParameterList();
+  if (pL.isParameter("repartition: use map")) {
+    const bool useMap = pL.get<bool>("repartition: use map");
+    if (useMap)
+      Input(currentLevel, "Map");
+    else
       Input(currentLevel, "A");
-    if(pL.isParameter("repartition: node repartition level")) {
-      const int nodeRepartLevel = pL.get<int>("repartition: node repartition level");
-      if(currentLevel.GetLevelID() == nodeRepartLevel) {
-        Input(currentLevel,"Node Comm");
-      }
+  } else
+    Input(currentLevel, "A");
+  if (pL.isParameter("repartition: node repartition level")) {
+    const int nodeRepartLevel = pL.get<int>("repartition: node repartition level");
+    if (currentLevel.GetLevelID() == nodeRepartLevel) {
+      Input(currentLevel, "Node Comm");
     }
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    const Teuchos::ParameterList & pL = GetParameterList();
-    // Access parameters here to make sure that we set the parameter entry flag to "used" even in case of short-circuit evaluation.
-    // TODO (JG): I don't really know if we want to do this.
-    const int    startLevel           = pL.get<int>   ("repartition: start level");
-    const int    nodeRepartLevel      = pL.get<int>   ("repartition: node repartition level");
-          LO     minRowsPerProcess    = pL.get<LO>    ("repartition: min rows per proc");
-          LO     targetRowsPerProcess = pL.get<LO>    ("repartition: target rows per proc");
-          LO     minRowsPerThread     = pL.get<LO>    ("repartition: min rows per thread");
-          LO     targetRowsPerThread  = pL.get<LO>    ("repartition: target rows per thread");
-    const double nonzeroImbalance     = pL.get<double>("repartition: max imbalance");
-    const bool   useMap               = pL.get<bool>  ("repartition: use map");
-
-    int thread_per_mpi_rank = 1;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionHeuristicFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+  // Access parameters here to make sure that we set the parameter entry flag to "used" even in case of short-circuit evaluation.
+  // TODO (JG): I don't really know if we want to do this.
+  const int startLevel          = pL.get<int>("repartition: start level");
+  const int nodeRepartLevel     = pL.get<int>("repartition: node repartition level");
+  LO minRowsPerProcess          = pL.get<LO>("repartition: min rows per proc");
+  LO targetRowsPerProcess       = pL.get<LO>("repartition: target rows per proc");
+  LO minRowsPerThread           = pL.get<LO>("repartition: min rows per thread");
+  LO targetRowsPerThread        = pL.get<LO>("repartition: target rows per thread");
+  const double nonzeroImbalance = pL.get<double>("repartition: max imbalance");
+  const bool useMap             = pL.get<bool>("repartition: use map");
+
+  int thread_per_mpi_rank = 1;
 #if defined(KOKKOS_ENABLE_OPENMP)
-    using execution_space = typename Node::device_type::execution_space;
-    if (std::is_same<execution_space, Kokkos::OpenMP>::value)
-      thread_per_mpi_rank = execution_space().concurrency();
+  using execution_space = typename Node::device_type::execution_space;
+  if (std::is_same<execution_space, Kokkos::OpenMP>::value)
+    thread_per_mpi_rank = execution_space().concurrency();
 #endif
 
-    if (minRowsPerThread > 0)
-      // We ignore the value given by minRowsPerProcess and repartition based on threads instead
-      minRowsPerProcess = minRowsPerThread*thread_per_mpi_rank;
-
-    if (targetRowsPerThread == 0)
-      targetRowsPerThread = minRowsPerThread;
-
-    if (targetRowsPerThread > 0)
-      // We ignore the value given by targetRowsPerProcess and repartition based on threads instead
-      targetRowsPerProcess = targetRowsPerThread*thread_per_mpi_rank;
-
-    if (targetRowsPerProcess == 0)
-      targetRowsPerProcess = minRowsPerProcess;
-
-    // Stick this on the level so Zoltan2Interface can use this later
-    Set<LO>(currentLevel,"repartition: heuristic target rows per process",targetRowsPerProcess);
-
-    // Check for validity of the node repartition option
-    TEUCHOS_TEST_FOR_EXCEPTION(nodeRepartLevel >= startLevel, Exceptions::RuntimeError, "MueLu::RepartitionHeuristicFactory::Build(): If 'repartition: node repartition level' is set, it must be less than or equal to 'repartition: start level'");
-
-
-    RCP<Matrix> A;
-    RCP<const FactoryBase> Afact;
-    RCP<const Map> map;
-    if (!useMap) {
-      Afact = GetFactory("A");
-      if(!Afact.is_null() && Teuchos::rcp_dynamic_cast<const RAPFactory>(Afact) == Teuchos::null &&
-         Teuchos::rcp_dynamic_cast<const BlockedRAPFactory>(Afact) == Teuchos::null &&
-         Teuchos::rcp_dynamic_cast<const SubBlockAFactory>(Afact) == Teuchos::null) {
-        GetOStream(Warnings) <<
-          "MueLu::RepartitionHeuristicFactory::Build: The generation factory for A must " \
-          "be a RAPFactory or a SubBlockAFactory providing the non-rebalanced matrix information! " \
-          "It specifically must not be of type Rebalance(Blocked)AcFactory or similar. " \
-          "Please check the input. Make also sure that \"number of partitions\" is provided to " \
-          "the Interface class and the RepartitionFactory instance.  Instead, we have a "<<Afact->description() << std::endl;
-      }
-      // TODO: We only need a CrsGraph. This class does not have to be templated on Scalar types.
-      A = Get< RCP<Matrix> >(currentLevel, "A");
-      map = A->getRowMap();
-    } else
-      map = Get< RCP<const Map> >(currentLevel, "Map");
-
-    // ======================================================================================================
-    // Determine whether partitioning is needed
-    // ======================================================================================================
-    // NOTE: most tests include some global communication, which is why we currently only do tests until we make
-    // a decision on whether to repartition. However, there is value in knowing how "close" we are to having to
-    // rebalance an operator. So, it would probably be beneficial to do and report *all* tests.
-
-    // Test0: Should we do node repartitioning?
-    if (currentLevel.GetLevelID() == nodeRepartLevel && map->getComm()->getSize() > 1) {
-      RCP<const Teuchos::Comm<int> > NodeComm = Get< RCP<const Teuchos::Comm<int> > >(currentLevel, "Node Comm");
-      TEUCHOS_TEST_FOR_EXCEPTION(NodeComm.is_null(), Exceptions::RuntimeError, "MueLu::RepartitionHeuristicFactory::Build(): NodeComm is null.");
-
-      // If we only have one node, then we don't want to pop down to one rank
-      if(NodeComm()->getSize() != map->getComm()->getSize()) {
-        GetOStream(Statistics1) << "Repartitioning?  YES: \n  Within node only"<<std::endl;
-        int nodeRank = NodeComm->getRank();
-        
-        // Do a reduction to get the total number of nodes
-        int isZero = (nodeRank == 0);
-        int numNodes=0;
-        Teuchos::reduceAll(*map->getComm(), Teuchos::REDUCE_SUM, isZero, Teuchos::outArg(numNodes));
-        Set(currentLevel, "number of partitions", numNodes);
-        return;
-      }
+  if (minRowsPerThread > 0)
+    // We ignore the value given by minRowsPerProcess and repartition based on threads instead
+    minRowsPerProcess = minRowsPerThread * thread_per_mpi_rank;
+
+  if (targetRowsPerThread == 0)
+    targetRowsPerThread = minRowsPerThread;
+
+  if (targetRowsPerThread > 0)
+    // We ignore the value given by targetRowsPerProcess and repartition based on threads instead
+    targetRowsPerProcess = targetRowsPerThread * thread_per_mpi_rank;
+
+  if (targetRowsPerProcess == 0)
+    targetRowsPerProcess = minRowsPerProcess;
+
+  // Stick this on the level so Zoltan2Interface can use this later
+  Set<LO>(currentLevel, "repartition: heuristic target rows per process", targetRowsPerProcess);
+
+  // Check for validity of the node repartition option
+  TEUCHOS_TEST_FOR_EXCEPTION(nodeRepartLevel >= startLevel, Exceptions::RuntimeError, "MueLu::RepartitionHeuristicFactory::Build(): If 'repartition: node repartition level' is set, it must be less than or equal to 'repartition: start level'");
+
+  RCP<Matrix> A;
+  RCP<const FactoryBase> Afact;
+  RCP<const Map> map;
+  if (!useMap) {
+    Afact = GetFactory("A");
+    if (!Afact.is_null() && Teuchos::rcp_dynamic_cast<const RAPFactory>(Afact) == Teuchos::null &&
+        Teuchos::rcp_dynamic_cast<const BlockedRAPFactory>(Afact) == Teuchos::null &&
+        Teuchos::rcp_dynamic_cast<const SubBlockAFactory>(Afact) == Teuchos::null) {
+      GetOStream(Warnings) << "MueLu::RepartitionHeuristicFactory::Build: The generation factory for A must "
+                              "be a RAPFactory or a SubBlockAFactory providing the non-rebalanced matrix information! "
+                              "It specifically must not be of type Rebalance(Blocked)AcFactory or similar. "
+                              "Please check the input. Make also sure that \"number of partitions\" is provided to "
+                              "the Interface class and the RepartitionFactory instance.  Instead, we have a "
+                           << Afact->description() << std::endl;
     }
+    // TODO: We only need a CrsGraph. This class does not have to be templated on Scalar types.
+    A   = Get<RCP<Matrix> >(currentLevel, "A");
+    map = A->getRowMap();
+  } else
+    map = Get<RCP<const Map> >(currentLevel, "Map");
+
+  // ======================================================================================================
+  // Determine whether partitioning is needed
+  // ======================================================================================================
+  // NOTE: most tests include some global communication, which is why we currently only do tests until we make
+  // a decision on whether to repartition. However, there is value in knowing how "close" we are to having to
+  // rebalance an operator. So, it would probably be beneficial to do and report *all* tests.
+
+  // Test0: Should we do node repartitioning?
+  if (currentLevel.GetLevelID() == nodeRepartLevel && map->getComm()->getSize() > 1) {
+    RCP<const Teuchos::Comm<int> > NodeComm = Get<RCP<const Teuchos::Comm<int> > >(currentLevel, "Node Comm");
+    TEUCHOS_TEST_FOR_EXCEPTION(NodeComm.is_null(), Exceptions::RuntimeError, "MueLu::RepartitionHeuristicFactory::Build(): NodeComm is null.");
+
+    // If we only have one node, then we don't want to pop down to one rank
+    if (NodeComm()->getSize() != map->getComm()->getSize()) {
+      GetOStream(Statistics1) << "Repartitioning?  YES: \n  Within node only" << std::endl;
+      int nodeRank = NodeComm->getRank();
+
+      // Do a reduction to get the total number of nodes
+      int isZero   = (nodeRank == 0);
+      int numNodes = 0;
+      Teuchos::reduceAll(*map->getComm(), Teuchos::REDUCE_SUM, isZero, Teuchos::outArg(numNodes));
+      Set(currentLevel, "number of partitions", numNodes);
+      return;
+    }
+  }
 
-    // Test1: skip repartitioning if current level is less than the specified minimum level for repartitioning
-    if (currentLevel.GetLevelID() < startLevel) {
-      GetOStream(Statistics1) << "Repartitioning?  NO:" <<
-          "\n  current level = " << Teuchos::toString(currentLevel.GetLevelID()) <<
-          ", first level where repartitioning can happen is " + Teuchos::toString(startLevel) << std::endl;
+  // Test1: skip repartitioning if current level is less than the specified minimum level for repartitioning
+  if (currentLevel.GetLevelID() < startLevel) {
+    GetOStream(Statistics1) << "Repartitioning?  NO:"
+                            << "\n  current level = " << Teuchos::toString(currentLevel.GetLevelID()) << ", first level where repartitioning can happen is " + Teuchos::toString(startLevel) << std::endl;
 
-      // a negative number of processors means: no repartitioning
-      Set(currentLevel, "number of partitions", -1);
+    // a negative number of processors means: no repartitioning
+    Set(currentLevel, "number of partitions", -1);
 
-      return;
-    }
+    return;
+  }
 
-    RCP<const Teuchos::Comm<int> > origComm = map->getComm();
-    RCP<const Teuchos::Comm<int> > comm     = origComm;
+  RCP<const Teuchos::Comm<int> > origComm = map->getComm();
+  RCP<const Teuchos::Comm<int> > comm     = origComm;
 
-    // Test 2: check whether A is actually distributed, i.e. more than one processor owns part of A
-    // TODO: this global communication can be avoided if we store the information with the matrix (it is known when matrix is created)
-    // TODO: further improvements could be achieved when we use subcommunicator for the active set. Then we only need to check its size
+  // Test 2: check whether A is actually distributed, i.e. more than one processor owns part of A
+  // TODO: this global communication can be avoided if we store the information with the matrix (it is known when matrix is created)
+  // TODO: further improvements could be achieved when we use subcommunicator for the active set. Then we only need to check its size
 
-    // TODO: The block transfer factories do not check correctly whether or not repartitioning actually took place.
-    {
-      if (comm->getSize() == 1 && Teuchos::rcp_dynamic_cast<const RAPFactory>(Afact) != Teuchos::null) {
-        GetOStream(Statistics1) << "Repartitioning?  NO:" <<
-            "\n  comm size = 1" << std::endl;
+  // TODO: The block transfer factories do not check correctly whether or not repartitioning actually took place.
+  {
+    if (comm->getSize() == 1 && Teuchos::rcp_dynamic_cast<const RAPFactory>(Afact) != Teuchos::null) {
+      GetOStream(Statistics1) << "Repartitioning?  NO:"
+                              << "\n  comm size = 1" << std::endl;
 
-        Set(currentLevel, "number of partitions", -1);
-        return;
-      }
+      Set(currentLevel, "number of partitions", -1);
+      return;
+    }
 
-      int numActiveProcesses = 0;
-      MueLu_sumAll(comm, Teuchos::as<int>((map->getLocalNumElements() > 0) ? 1 : 0), numActiveProcesses);
+    int numActiveProcesses = 0;
+    MueLu_sumAll(comm, Teuchos::as<int>((map->getLocalNumElements() > 0) ? 1 : 0), numActiveProcesses);
 
-      if (numActiveProcesses == 1) {
-        GetOStream(Statistics1) << "Repartitioning?  NO:" <<
-            "\n  # processes with rows = " << Teuchos::toString(numActiveProcesses) << std::endl;
+    if (numActiveProcesses == 1) {
+      GetOStream(Statistics1) << "Repartitioning?  NO:"
+                              << "\n  # processes with rows = " << Teuchos::toString(numActiveProcesses) << std::endl;
 
-        Set(currentLevel, "number of partitions", 1);
-        return;
-      }
+      Set(currentLevel, "number of partitions", 1);
+      return;
     }
+  }
 
-    bool test3 = false, test4 = false;
-    std::string msg3, msg4;
+  bool test3 = false, test4 = false;
+  std::string msg3, msg4;
 
-    // Test3: check whether number of rows on any processor satisfies the minimum number of rows requirement
-    // NOTE: Test2 ensures that repartitionning is not done when there is only one processor (it may or may not satisfy Test3)
-    if (minRowsPerProcess > 0) {
-      LO numMyRows = Teuchos::as<LO>(map->getLocalNumElements()), minNumRows, LOMAX = Teuchos::OrdinalTraits<LO>::max();
-      LO haveFewRows = (numMyRows < minRowsPerProcess ? 1 : 0), numWithFewRows = 0;
-      MueLu_sumAll(comm, haveFewRows, numWithFewRows);
-      MueLu_minAll(comm, (numMyRows > 0 ? numMyRows : LOMAX), minNumRows);
+  // Test3: check whether number of rows on any processor satisfies the minimum number of rows requirement
+  // NOTE: Test2 ensures that repartitionning is not done when there is only one processor (it may or may not satisfy Test3)
+  if (minRowsPerProcess > 0) {
+    LO numMyRows = Teuchos::as<LO>(map->getLocalNumElements()), minNumRows, LOMAX = Teuchos::OrdinalTraits<LO>::max();
+    LO haveFewRows = (numMyRows < minRowsPerProcess ? 1 : 0), numWithFewRows = 0;
+    MueLu_sumAll(comm, haveFewRows, numWithFewRows);
+    MueLu_minAll(comm, (numMyRows > 0 ? numMyRows : LOMAX), minNumRows);
 
-      // TODO: we could change it to repartition only if the number of processors with numRows < minNumRows is larger than some
-      // percentage of the total number. This way, we won't repartition if 2 out of 1000 processors don't have enough elements.
-      // I'm thinking maybe 20% threshold. To implement, simply add " && numWithFewRows < .2*numProcs" to the if statement.
-      if (numWithFewRows > 0)
-        test3 = true;
+    // TODO: we could change it to repartition only if the number of processors with numRows < minNumRows is larger than some
+    // percentage of the total number. This way, we won't repartition if 2 out of 1000 processors don't have enough elements.
+    // I'm thinking maybe 20% threshold. To implement, simply add " && numWithFewRows < .2*numProcs" to the if statement.
+    if (numWithFewRows > 0)
+      test3 = true;
 
-      msg3 = "\n  min # rows per proc = " + Teuchos::toString(minNumRows) + ", min allowable = " + Teuchos::toString(minRowsPerProcess);
-    }
+    msg3 = "\n  min # rows per proc = " + Teuchos::toString(minNumRows) + ", min allowable = " + Teuchos::toString(minRowsPerProcess);
+  }
 
-    // Test4: check whether the balance in the number of nonzeros per processor is greater than threshold
-    if (!test3) {
-      if (useMap)
-        msg4 = "";
-      else {
-        GO minNnz, maxNnz, numMyNnz = Teuchos::as<GO>(A->getLocalNumEntries());
-        MueLu_maxAll(comm, numMyNnz,                           maxNnz);
-        MueLu_minAll(comm, (numMyNnz > 0 ? numMyNnz : maxNnz), minNnz); // min nnz over all active processors
-        double imbalance = Teuchos::as<double>(maxNnz)/minNnz;
-
-        if (imbalance > nonzeroImbalance)
-          test4 = true;
-
-        msg4 = "\n  nonzero imbalance = " + Teuchos::toString(imbalance) + ", max allowable = " + Teuchos::toString(nonzeroImbalance);
-      }
-    }
+  // Test4: check whether the balance in the number of nonzeros per processor is greater than threshold
+  if (!test3) {
+    if (useMap)
+      msg4 = "";
+    else {
+      GO minNnz, maxNnz, numMyNnz = Teuchos::as<GO>(A->getLocalNumEntries());
+      MueLu_maxAll(comm, numMyNnz, maxNnz);
+      MueLu_minAll(comm, (numMyNnz > 0 ? numMyNnz : maxNnz), minNnz);  // min nnz over all active processors
+      double imbalance = Teuchos::as<double>(maxNnz) / minNnz;
 
-    if (!test3 && !test4) {
-      GetOStream(Statistics1) << "Repartitioning?  NO:" << msg3 + msg4 << std::endl;
+      if (imbalance > nonzeroImbalance)
+        test4 = true;
 
-      // A negative number of partitions means: no repartitioning
-      Set(currentLevel, "number of partitions", -1);
-      return;
+      msg4 = "\n  nonzero imbalance = " + Teuchos::toString(imbalance) + ", max allowable = " + Teuchos::toString(nonzeroImbalance);
     }
+  }
 
-    GetOStream(Statistics1) << "Repartitioning? YES:" << msg3 + msg4 << std::endl;
-
-    // ======================================================================================================
-    // Calculate number of partitions
-    // ======================================================================================================
-    // FIXME Quick way to figure out how many partitions there should be (same algorithm as ML)
-    // FIXME Should take into account nnz? Perhaps only when user is using min #nnz per row threshold.
-
-    // The number of partitions is calculated by the RepartitionFactory and stored in "number of partitions" variable on
-    // the current level. If this variable is already set (e.g., by another instance of RepartitionFactory) then this number
-    // is used. The "number of partitions" variable serves as basic communication between the RepartitionFactory (which
-    // requests a certain number of partitions) and the *Interface classes which call the underlying partitioning algorithms
-    // and produce the "Partition" array with the requested number of partitions.
-    const auto globalNumRows = Teuchos::as<GO>(map->getGlobalNumElements());
-    int numPartitions = 1;
-    if (globalNumRows >= targetRowsPerProcess) {
-      // Make sure that each CPU thread has approximately targetRowsPerProcess
-      numPartitions = std::max(Teuchos::as<int>(globalNumRows / targetRowsPerProcess), 1);
-    }
-    numPartitions = std::min(numPartitions, comm->getSize());
+  if (!test3 && !test4) {
+    GetOStream(Statistics1) << "Repartitioning?  NO:" << msg3 + msg4 << std::endl;
+
+    // A negative number of partitions means: no repartitioning
+    Set(currentLevel, "number of partitions", -1);
+    return;
+  }
+
+  GetOStream(Statistics1) << "Repartitioning? YES:" << msg3 + msg4 << std::endl;
+
+  // ======================================================================================================
+  // Calculate number of partitions
+  // ======================================================================================================
+  // FIXME Quick way to figure out how many partitions there should be (same algorithm as ML)
+  // FIXME Should take into account nnz? Perhaps only when user is using min #nnz per row threshold.
+
+  // The number of partitions is calculated by the RepartitionFactory and stored in "number of partitions" variable on
+  // the current level. If this variable is already set (e.g., by another instance of RepartitionFactory) then this number
+  // is used. The "number of partitions" variable serves as basic communication between the RepartitionFactory (which
+  // requests a certain number of partitions) and the *Interface classes which call the underlying partitioning algorithms
+  // and produce the "Partition" array with the requested number of partitions.
+  const auto globalNumRows = Teuchos::as<GO>(map->getGlobalNumElements());
+  int numPartitions        = 1;
+  if (globalNumRows >= targetRowsPerProcess) {
+    // Make sure that each CPU thread has approximately targetRowsPerProcess
+    numPartitions = std::max(Teuchos::as<int>(globalNumRows / targetRowsPerProcess), 1);
+  }
+  numPartitions = std::min(numPartitions, comm->getSize());
 
-    Set(currentLevel, "number of partitions", numPartitions);
+  Set(currentLevel, "number of partitions", numPartitions);
 
-    GetOStream(Statistics1) << "Number of partitions to use = " << numPartitions << std::endl;
-  } // Build
-} // namespace MueLu
+  GetOStream(Statistics1) << "Number of partitions to use = " << numPartitions << std::endl;
+}  // Build
+}  // namespace MueLu
 
-#endif //ifdef HAVE_MPI
-#endif /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_REPARTITIONHEURISTICFACTORY_DEF_HPP_ */
+#endif  // ifdef HAVE_MPI
+#endif  /* PACKAGES_MUELU_SRC_REBALANCING_MUELU_REPARTITIONHEURISTICFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_decl.hpp
index 3e0804ff8819..0eaf9ac14e51 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_decl.hpp
@@ -17,98 +17,90 @@
 #include "MueLu_Level_fwd.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class RepartitionInterface
-    @brief Helper class which transforms an "AmalgamatedPartition" array to an unamalgamated "Partition".
-    @ingroup Rebalancing
-
-    This is a general class that allows to translate node-based rebalancing information (given by "AmalgamatedPartition") to
-    DOF-based rebalancing information (stored as output in the "Partition" variable).
-    It is meant to be used together with the IsorropiaInterface class which provides the node-based rebalancing information
-    in the "AmalgamatedPartition" variable. It uses the striding information of "A" to transform the amalgamated rebalaning info
-    into DOF-based rebalancing information that can be processed by the RepartitionFactory class.
-
-    @note: We assume a constant number of DOFs per node
-
-    ## Input/output of RepartitionInterface ##
-
-    ### User parameters of RepartitionInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | AmalgamatedPartition | Factory | null |  | * | * | Factory generating the AmalgamatedPartition (e.g. an IsorropiaInterface)
-    | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionInterface::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionInterface::DeclareInput).
-
-    ### Variables provided by RepartitionInterface ###
-
-    After RepartitionInterface::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Partition | RepartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
-
-    The "Partition" vector is used as input for the RepartitionFactory class.
-    If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
-    by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
-
-    The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
-    It stores the result in the "number of partitions" variable on the current level (type = GO).
-    If it is "number of partitions=1" we skip the underlying Zoltan call and just create an dummy "Partition" vector containing zeros only.
-    If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
-    If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
-  */
-
-  //FIXME: this class should not be templated
-  template <class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RepartitionInterface : public SingleLevelFactoryBase {
-
-    typedef double Scalar; // FIXME
+/*!
+  @class RepartitionInterface
+  @brief Helper class which transforms an "AmalgamatedPartition" array to an unamalgamated "Partition".
+  @ingroup Rebalancing
+
+  This is a general class that allows to translate node-based rebalancing information (given by "AmalgamatedPartition") to
+  DOF-based rebalancing information (stored as output in the "Partition" variable).
+  It is meant to be used together with the IsorropiaInterface class which provides the node-based rebalancing information
+  in the "AmalgamatedPartition" variable. It uses the striding information of "A" to transform the amalgamated rebalaning info
+  into DOF-based rebalancing information that can be processed by the RepartitionFactory class.
+
+  @note: We assume a constant number of DOFs per node
+
+  ## Input/output of RepartitionInterface ##
+
+  ### User parameters of RepartitionInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | AmalgamatedPartition | Factory | null |  | * | * | Factory generating the AmalgamatedPartition (e.g. an IsorropiaInterface)
+  | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see RepartitionInterface::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see RepartitionInterface::DeclareInput).
+
+  ### Variables provided by RepartitionInterface ###
+
+  After RepartitionInterface::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Partition | RepartitionInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
+
+  The "Partition" vector is used as input for the RepartitionFactory class.
+  If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
+  by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
+
+  The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
+  It stores the result in the "number of partitions" variable on the current level (type = GO).
+  If it is "number of partitions=1" we skip the underlying Zoltan call and just create an dummy "Partition" vector containing zeros only.
+  If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
+  If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
+*/
+
+// FIXME: this class should not be templated
+template <class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RepartitionInterface : public SingleLevelFactoryBase {
+  typedef double Scalar;  // FIXME
 #undef MUELU_REPARTITIONINTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
-
-    //! Constructor
-    RepartitionInterface() { }
-
-    //! Destructor
-    virtual ~RepartitionInterface() { }
-    //@}
-
-    RCP<const ParameterList> GetValidParameterList() const;
-
-    //! @name Input
-    //@{
-    void DeclareInput(Level & level) const;
-    //@}
-
-    //! @name Build methods.
-    //@{
-    void Build(Level &level) const;
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //@}
+  //! Constructor
+  RepartitionInterface() {}
 
+  //! Destructor
+  virtual ~RepartitionInterface() {}
+  //@}
 
+  RCP<const ParameterList> GetValidParameterList() const;
 
-  private:
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
+  //@}
 
-  };  //class RepartitionInterface
+ private:
+};  // class RepartitionInterface
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPARTITIONINTERFACE_SHORT
 #endif /* MUELU_REPARTITIONINTERFACE_DECL_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_def.hpp
index b7b3b922ebd7..c7b274e2f7b3 100644
--- a/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_RepartitionInterface_def.hpp
@@ -14,127 +14,122 @@
 #include "MueLu_Exceptions.hpp"
 #include "MueLu_Monitor.hpp"
 
-
 namespace MueLu {
 
- template <class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("A",                    Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >("AmalgamatedPartition", Teuchos::null, "(advanced) Factory generating the AmalgamatedPartition (e.g. an IsorropiaInterface)");
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("AmalgamatedPartition", Teuchos::null, "(advanced) Factory generating the AmalgamatedPartition (e.g. an IsorropiaInterface)");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  Input(currentLevel, "AmalgamatedPartition");
+}  // DeclareInput()
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    Input(currentLevel, "AmalgamatedPartition");
-  } //DeclareInput()
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &level) const {
-    FactoryMonitor m(*this, "Build", level);
-
-    RCP<Matrix>      A                                  = Get< RCP<Matrix> >     (level, "A");
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > amalgPartition = Get< RCP<Xpetra::Vector<GO, LO, GO, NO> > >(level, "AmalgamatedPartition");
-    int numParts                                        = Get<int>(level, "number of partitions");
-
-    RCP<const Map> rowMap        = A->getRowMap();
-
-    // standard case: use matrix info and amalgamated rebalancing info to create "Partition" vector
-    RCP<const Teuchos::Comm< int > > comm = A->getRowMap()->getComm();
-
-    // Short cut: if we only need one partition, then create a dummy partition vector
-    if (numParts == 1 || numParts == -1) {
-      // Single processor, decomposition is trivial: all zeros
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      Set(level, "Partition", decomposition);
-      return;
-    }/* else if (numParts == -1) {
-      // No repartitioning
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Teuchos::null; //Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      //decomposition->putScalar(Teuchos::as<Scalar>(comm->getRank()));
-      Set(level, "Partition", decomposition);
-      return;
-    }*/
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void RepartitionInterface<LocalOrdinal, GlobalOrdinal, Node>::Build(Level &level) const {
+  FactoryMonitor m(*this, "Build", level);
 
-    ArrayRCP<GO> amalgPartitionData = amalgPartition->getDataNonConst(0);
-    RCP<const Map> nodeMap       = amalgPartition->getMap();
-
-    // extract amalgamation information from matrix A
-    LO blockdim = 1;                          // block dim for fixed size blocks
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    LO stridedblocksize = blockdim; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
-
-    // 1) check for blocking/striding information
-    //    fill above variables
-    if(A->IsView("stridedMaps") &&
-       Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
-      TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null,Exceptions::BadCast,"MueLu::RepartitionInterface::Build: cast to strided row map failed.");
-      blockdim = strMap->getFixedBlockSize();
-      blockid  = strMap->getStridedBlockId();
-      if (blockid > -1) {
-        std::vector<size_t> stridingInfo = strMap->getStridingData();
-        for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
-          nStridedOffset += stridingInfo[j];
-        stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
-
-      } else {
-        stridedblocksize = blockdim;
-      }
-      oldView = A->SwitchToView(oldView);
-      //GetOStream(Statistics0, -1) << "RepartitionInterface::Build():" << " found blockdim=" << blockdim << " from strided maps (blockid=" << blockid << ", strided block size=" << stridedblocksize << "). offset=" << offset << std::endl;
-    } else GetOStream(Statistics0, -1) << "RepartitionInterface::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
-
-    // vector which stores final (unamalgamated) repartitioning
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);
-    ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<int>(nodeMap->getLocalNumElements())*stridedblocksize != Teuchos::as<int>(rowMap->getLocalNumElements()), Exceptions::RuntimeError, "Inconsistency between nodeMap and dofMap: we are supporting block maps only. No support for general strided maps, yet!");
-
-    //RCP<std::map<GO,std::vector<GO> > > nodegid2dofgids = amalgInfo->GetGlobalAmalgamationParams();
-
-    // fill vector with information about partitioning
-    // TODO: we assume simple block maps here
-    // TODO: adapt this to usage of nodegid2dofgids
-    for(size_t i = 0; i < nodeMap->getLocalNumElements(); i++) {
-      // not fully sure about this. We're filling local ids in the decomposition vector with
-      // the results stored in array. The decomposition vector is created using the rowMap of A
-
-      // transform local node id to global node id.
-      //GO gNodeId = nodeMap->getGlobalElement(i);
-
-      // extract global DOF ids that belong to gNodeId
-      /*std::vector<GlobalOrdinal> DOFs = (*nodegid2dofgids)[gNodeId];
-      for(size_t j=0; j<stridedblocksize; j++) {
-        decompEntries[i*stridedblocksize + j] = myRank;
-      }*/
-      for (LO j = 0; j < stridedblocksize/*DOFs.size()*/; j++) {
-        // transform global DOF ids to local DOF ids using rowMap
-        // note: The vector decomposition is based on rowMap
-        //LO lDofId = rowMap->getLocalElement(DOFs[j]);     // -> i doubt that we need this!
-
-        // put the same domain id to all DOFs of the same node
-        decompEntries[i*stridedblocksize + j] = amalgPartitionData[i];
-        //decompEntries[lDofId] = amalgPartitionData[i];
-      }
+  RCP<Matrix> A                                       = Get<RCP<Matrix> >(level, "A");
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > amalgPartition = Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >(level, "AmalgamatedPartition");
+  int numParts                                        = Get<int>(level, "number of partitions");
 
-    }
+  RCP<const Map> rowMap = A->getRowMap();
 
+  // standard case: use matrix info and amalgamated rebalancing info to create "Partition" vector
+  RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+
+  // Short cut: if we only need one partition, then create a dummy partition vector
+  if (numParts == 1 || numParts == -1) {
+    // Single processor, decomposition is trivial: all zeros
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
     Set(level, "Partition", decomposition);
+    return;
+  } /* else if (numParts == -1) {
+     // No repartitioning
+     RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Teuchos::null; //Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+     //decomposition->putScalar(Teuchos::as<Scalar>(comm->getRank()));
+     Set(level, "Partition", decomposition);
+     return;
+   }*/
+
+  ArrayRCP<GO> amalgPartitionData = amalgPartition->getDataNonConst(0);
+  RCP<const Map> nodeMap          = amalgPartition->getMap();
+
+  // extract amalgamation information from matrix A
+  LO blockdim         = 1;         // block dim for fixed size blocks
+  LO blockid          = -1;        // block id in strided map
+  LO nStridedOffset   = 0;         // DOF offset for strided block id "blockid" (default = 0)
+  LO stridedblocksize = blockdim;  // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+
+  // 1) check for blocking/striding information
+  //    fill above variables
+  if (A->IsView("stridedMaps") &&
+      Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
+    Xpetra::viewLabel_t oldView  = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+    RCP<const StridedMap> strMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap());
+    TEUCHOS_TEST_FOR_EXCEPTION(strMap == Teuchos::null, Exceptions::BadCast, "MueLu::RepartitionInterface::Build: cast to strided row map failed.");
+    blockdim = strMap->getFixedBlockSize();
+    blockid  = strMap->getStridedBlockId();
+    if (blockid > -1) {
+      std::vector<size_t> stridingInfo = strMap->getStridingData();
+      for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
+        nStridedOffset += stridingInfo[j];
+      stridedblocksize = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+
+    } else {
+      stridedblocksize = blockdim;
+    }
+    oldView = A->SwitchToView(oldView);
+    // GetOStream(Statistics0, -1) << "RepartitionInterface::Build():" << " found blockdim=" << blockdim << " from strided maps (blockid=" << blockid << ", strided block size=" << stridedblocksize << "). offset=" << offset << std::endl;
+  } else
+    GetOStream(Statistics0, -1) << "RepartitionInterface::Build(): no striding information available. Use blockdim=1 with offset=0" << std::endl;
+
+  // vector which stores final (unamalgamated) repartitioning
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);
+  ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
 
-  } //Build()
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<int>(nodeMap->getLocalNumElements()) * stridedblocksize != Teuchos::as<int>(rowMap->getLocalNumElements()), Exceptions::RuntimeError, "Inconsistency between nodeMap and dofMap: we are supporting block maps only. No support for general strided maps, yet!");
 
+  // RCP<std::map<GO,std::vector<GO> > > nodegid2dofgids = amalgInfo->GetGlobalAmalgamationParams();
+
+  // fill vector with information about partitioning
+  // TODO: we assume simple block maps here
+  // TODO: adapt this to usage of nodegid2dofgids
+  for (size_t i = 0; i < nodeMap->getLocalNumElements(); i++) {
+    // not fully sure about this. We're filling local ids in the decomposition vector with
+    // the results stored in array. The decomposition vector is created using the rowMap of A
+
+    // transform local node id to global node id.
+    // GO gNodeId = nodeMap->getGlobalElement(i);
+
+    // extract global DOF ids that belong to gNodeId
+    /*std::vector<GlobalOrdinal> DOFs = (*nodegid2dofgids)[gNodeId];
+    for(size_t j=0; j<stridedblocksize; j++) {
+      decompEntries[i*stridedblocksize + j] = myRank;
+    }*/
+    for (LO j = 0; j < stridedblocksize /*DOFs.size()*/; j++) {
+      // transform global DOF ids to local DOF ids using rowMap
+      // note: The vector decomposition is based on rowMap
+      // LO lDofId = rowMap->getLocalElement(DOFs[j]);     // -> i doubt that we need this!
+
+      // put the same domain id to all DOFs of the same node
+      decompEntries[i * stridedblocksize + j] = amalgPartitionData[i];
+      // decompEntries[lDofId] = amalgPartitionData[i];
+    }
+  }
 
+  Set(level, "Partition", decomposition);
 
-} //namespace MueLu
+}  // Build()
 
+}  // namespace MueLu
 
 #endif /* MUELU_REPARTITIONINTERFACE_DEF_HPP_ */
diff --git a/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_decl.hpp
index f349c12cb6be..6fb098af8bf9 100644
--- a/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_decl.hpp
@@ -65,184 +65,183 @@
 
 namespace MueLu {
 
-
-  /*!
-    @class Zoltan2Interface
-    @brief Interface to Zoltan2 library.
-    @ingroup Rebalancing
-
-    This interface provides access to partitioning methods in Zoltan2.
-    Currently, it supports RCB and multijagged as well as all graph partitioning algorithms from Zoltan2.
-
-    ## Input/output of Zoltan2Interface ##
-
-    ### User parameters of Zoltan2Interface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | Coordinates                            | Factory | null  |   | * | (*) | Factory generating coordinates vector used for rebalancing. The coordinates are only needed when the chosen algorithm is 'multijagged' or 'rcb'.
-    | ParameterList                          | ParamterList | null |  | * |  | Zoltan2 parameters
-    | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see Zoltan2Interface::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see Zoltan2Interface::DeclareInput).
-
-    ### Variables provided by Zoltan2Interface ###
-
-    After Zoltan2Interface::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Partition       | Zoltan2Interface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
-
-    The "Partition" vector is used as input for the RepartitionFactory class.
-    If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
-    by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
-
-    The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
-    It stores the result in the "number of partitions" variable on the current level (type = GO).
-    If it is "number of partitions=1" we skip the Zoltan2 call and just create an dummy "Partition" vector containing zeros only.
-    If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
-    If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
-
-  */
-
-  //FIXME: this class should not be templated
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Zoltan2Interface : public SingleLevelFactoryBase {
+/*!
+  @class Zoltan2Interface
+  @brief Interface to Zoltan2 library.
+  @ingroup Rebalancing
+
+  This interface provides access to partitioning methods in Zoltan2.
+  Currently, it supports RCB and multijagged as well as all graph partitioning algorithms from Zoltan2.
+
+  ## Input/output of Zoltan2Interface ##
+
+  ### User parameters of Zoltan2Interface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | Coordinates                            | Factory | null  |   | * | (*) | Factory generating coordinates vector used for rebalancing. The coordinates are only needed when the chosen algorithm is 'multijagged' or 'rcb'.
+  | ParameterList                          | ParamterList | null |  | * |  | Zoltan2 parameters
+  | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see Zoltan2Interface::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see Zoltan2Interface::DeclareInput).
+
+  ### Variables provided by Zoltan2Interface ###
+
+  After Zoltan2Interface::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Partition       | Zoltan2Interface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
+
+  The "Partition" vector is used as input for the RepartitionFactory class.
+  If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
+  by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
+
+  The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
+  It stores the result in the "number of partitions" variable on the current level (type = GO).
+  If it is "number of partitions=1" we skip the Zoltan2 call and just create an dummy "Partition" vector containing zeros only.
+  If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
+  If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
+
+*/
+
+// FIXME: this class should not be templated
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Zoltan2Interface : public SingleLevelFactoryBase {
 #undef MUELU_ZOLTAN2INTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    //! Constructor
-    Zoltan2Interface();
+  //! Constructor
+  Zoltan2Interface();
 
-    //! Destructor
-    virtual ~Zoltan2Interface() { }
-    //@}
+  //! Destructor
+  virtual ~Zoltan2Interface() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level& currentLevel) const;
-    //@}
+  //! @name Input
+  //@{
+  void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& currentLevel) const;
+  //! @name Build methods.
+  //@{
+  void Build(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    RCP<ParameterList> defaultZoltan2Params;
+ private:
+  RCP<ParameterList> defaultZoltan2Params;
 
-  };  //class Zoltan2Interface
+};  // class Zoltan2Interface
 
 #ifdef HAVE_MUELU_EPETRA
 
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
 
 #if defined(HAVE_MUELU_ZOLTAN)
-  // Stub partial specialization of Zoltan2Interface for EpetraNode
-  // Use ZoltanInterface instead of Zoltan2Interface
-  template<>
-  class Zoltan2Interface<double,int,int,Xpetra::EpetraNode> : public SingleLevelFactoryBase {
-    typedef double              Scalar;
-    typedef int                 LocalOrdinal;
-    typedef int                 GlobalOrdinal;
-    typedef Xpetra::EpetraNode  Node;
+// Stub partial specialization of Zoltan2Interface for EpetraNode
+// Use ZoltanInterface instead of Zoltan2Interface
+template <>
+class Zoltan2Interface<double, int, int, Xpetra::EpetraNode> : public SingleLevelFactoryBase {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
 #undef MUELU_ZOLTAN2INTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    Zoltan2Interface() {
-      level_           = rcp(new Level());
-      zoltanInterface_ = rcp(new ZoltanInterface());
-
-      level_->SetLevelID(1);
-    }
-    virtual ~Zoltan2Interface() {
-      zoltanInterface_ = Teuchos::null;
-      level_           = Teuchos::null;
-    }
-    RCP<const ParameterList> GetValidParameterList() const {
-      RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-      validParamList->set< RCP<const FactoryBase> >   ("A",             Teuchos::null, "Factory of the matrix A");
-      validParamList->set< RCP<const FactoryBase> >   ("Coordinates",   Teuchos::null, "Factory of the coordinates");
-      validParamList->set< RCP<const FactoryBase> >   ("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-      validParamList->set< RCP<const ParameterList> > ("ParameterList", Teuchos::null, "Zoltan2 parameters");
-
-      return validParamList;
-    };
-    void DeclareInput(Level& currentLevel) const {
-      Input(currentLevel, "A");
-      Input(currentLevel, "number of partitions");
-      const ParameterList& pL = GetParameterList();
-      // We do this dance, because we don't want "ParameterList" to be marked as used.
-      // Is there a better way?
-      Teuchos::ParameterEntry entry = pL.getEntry("ParameterList");
-      RCP<const Teuchos::ParameterList> providedList = Teuchos::any_cast<RCP<const Teuchos::ParameterList> >(entry.getAny(false));
-      if (providedList != Teuchos::null && providedList->isType<std::string>("algorithm")) {
-        const std::string algo = providedList->get<std::string>("algorithm");
-        if (algo == "multijagged" || algo == "rcb")
-          Input(currentLevel, "Coordinates");
-      } else
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  Zoltan2Interface() {
+    level_           = rcp(new Level());
+    zoltanInterface_ = rcp(new ZoltanInterface());
+
+    level_->SetLevelID(1);
+  }
+  virtual ~Zoltan2Interface() {
+    zoltanInterface_ = Teuchos::null;
+    level_           = Teuchos::null;
+  }
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+    validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+    validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory of the coordinates");
+    validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+    validParamList->set<RCP<const ParameterList> >("ParameterList", Teuchos::null, "Zoltan2 parameters");
+
+    return validParamList;
+  };
+  void DeclareInput(Level& currentLevel) const {
+    Input(currentLevel, "A");
+    Input(currentLevel, "number of partitions");
+    const ParameterList& pL = GetParameterList();
+    // We do this dance, because we don't want "ParameterList" to be marked as used.
+    // Is there a better way?
+    Teuchos::ParameterEntry entry                  = pL.getEntry("ParameterList");
+    RCP<const Teuchos::ParameterList> providedList = Teuchos::any_cast<RCP<const Teuchos::ParameterList> >(entry.getAny(false));
+    if (providedList != Teuchos::null && providedList->isType<std::string>("algorithm")) {
+      const std::string algo = providedList->get<std::string>("algorithm");
+      if (algo == "multijagged" || algo == "rcb")
         Input(currentLevel, "Coordinates");
-    };
-    void Build(Level& currentLevel) const {
-      this->GetOStream(Warnings0) << "Tpetra does not support <double,int,int,EpetraNode> instantiation, "
-          "switching Zoltan2Interface to ZoltanInterface" << std::endl;
-
-      // Put the data into a fake level
-      level_->Set("A",                    Get<RCP<Matrix> >      (currentLevel, "A"));
-      level_->Set("Coordinates",          Get<RCP<RealValuedMultiVector> >(currentLevel, "Coordinates"));
-      level_->Set("number of partitions", currentLevel.Get<GO>("number of partitions"));
-
-      level_->Request("Partition", zoltanInterface_.get());
-      zoltanInterface_->Build(*level_);
-
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition;
-      level_->Get("Partition", decomposition, zoltanInterface_.get());
-      Set(currentLevel, "Partition", decomposition);
-    };
-
-  private:
-    RCP<Level>           level_;            // fake Level
-    RCP<ZoltanInterface> zoltanInterface_;
+    } else
+      Input(currentLevel, "Coordinates");
   };
-#else
-  // Stub partial specialization of Zoltan2Interface for EpetraNode
-  template<>
-  class Zoltan2Interface<double,int,int,Xpetra::EpetraNode> : public SingleLevelFactoryBase {
-  public:
-    Zoltan2Interface() { throw Exceptions::RuntimeError("Tpetra does not support <double,int,int,EpetraNode> instantiation"); }
-    virtual ~Zoltan2Interface() { }
-    RCP<const ParameterList> GetValidParameterList() const { return Teuchos::null; };
-    void DeclareInput(Level& level) const {};
-    void Build(Level &level) const {};
+  void Build(Level& currentLevel) const {
+    this->GetOStream(Warnings0) << "Tpetra does not support <double,int,int,EpetraNode> instantiation, "
+                                   "switching Zoltan2Interface to ZoltanInterface"
+                                << std::endl;
+
+    // Put the data into a fake level
+    level_->Set("A", Get<RCP<Matrix> >(currentLevel, "A"));
+    level_->Set("Coordinates", Get<RCP<RealValuedMultiVector> >(currentLevel, "Coordinates"));
+    level_->Set("number of partitions", currentLevel.Get<GO>("number of partitions"));
+
+    level_->Request("Partition", zoltanInterface_.get());
+    zoltanInterface_->Build(*level_);
+
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition;
+    level_->Get("Partition", decomposition, zoltanInterface_.get());
+    Set(currentLevel, "Partition", decomposition);
   };
-#endif // HAVE_MUELU_ZOLTAN
+
+ private:
+  RCP<Level> level_;  // fake Level
+  RCP<ZoltanInterface> zoltanInterface_;
+};
+#else
+// Stub partial specialization of Zoltan2Interface for EpetraNode
+template <>
+class Zoltan2Interface<double, int, int, Xpetra::EpetraNode> : public SingleLevelFactoryBase {
+ public:
+  Zoltan2Interface() { throw Exceptions::RuntimeError("Tpetra does not support <double,int,int,EpetraNode> instantiation"); }
+  virtual ~Zoltan2Interface() {}
+  RCP<const ParameterList> GetValidParameterList() const { return Teuchos::null; };
+  void DeclareInput(Level& level) const {};
+  void Build(Level& level) const {};
+};
+#endif  // HAVE_MUELU_ZOLTAN
 
 #endif
 
-#endif // HAVE_MUELU_EPETRA
+#endif  // HAVE_MUELU_EPETRA
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ZOLTAN2INTERFACE_SHORT
-#endif //if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
+#endif  // if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
 
-#endif // MUELU_ZOLTAN2INTERFACE_DECL_HPP
+#endif  // MUELU_ZOLTAN2INTERFACE_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_def.hpp
index 581ff8588e5a..cbe29792a00f 100644
--- a/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_Zoltan2Interface_def.hpp
@@ -66,209 +66,205 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Zoltan2Interface() {
-    defaultZoltan2Params = rcp(new ParameterList());
-    defaultZoltan2Params->set("algorithm",             "multijagged");
-    defaultZoltan2Params->set("partitioning_approach", "partition");
-
-    // Improve scaling for communication bound algorithms by premigrating
-    // coordinates to a subset of processors.
-    // For more information, see Github issue #1538
-    defaultZoltan2Params->set("mj_premigration_option", 1);
- }
-
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >   ("A",             Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >   ("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >   ("Coordinates",   Teuchos::null, "Factory of the coordinates");
-    validParamList->set< RCP<const ParameterList> > ("ParameterList", Teuchos::null, "Zoltan2 parameters");
-    validParamList->set< RCP<const FactoryBase> >   ("repartition: heuristic target rows per process", Teuchos::null, "Factory for number of rows per process to use with MultiJagged");
-
-    return validParamList;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    const ParameterList& pL = GetParameterList();
-    // We do this dance, because we don't want "ParameterList" to be marked as used.
-    // Is there a better way?
-    Teuchos::ParameterEntry entry = pL.getEntry("ParameterList");
-    RCP<const Teuchos::ParameterList> providedList = Teuchos::any_cast<RCP<const Teuchos::ParameterList> >(entry.getAny(false));
-    if (providedList != Teuchos::null && providedList->isType<std::string>("algorithm")) {
-      const std::string algo = providedList->get<std::string>("algorithm");
-      if (algo == "multijagged") {
-        Input(currentLevel, "Coordinates");
-        Input(currentLevel, "repartition: heuristic target rows per process");
-      } else if (algo == "rcb") {
-        Input(currentLevel, "Coordinates");
-      }
-    } else {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Zoltan2Interface() {
+  defaultZoltan2Params = rcp(new ParameterList());
+  defaultZoltan2Params->set("algorithm", "multijagged");
+  defaultZoltan2Params->set("partitioning_approach", "partition");
+
+  // Improve scaling for communication bound algorithms by premigrating
+  // coordinates to a subset of processors.
+  // For more information, see Github issue #1538
+  defaultZoltan2Params->set("mj_premigration_option", 1);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory of the coordinates");
+  validParamList->set<RCP<const ParameterList> >("ParameterList", Teuchos::null, "Zoltan2 parameters");
+  validParamList->set<RCP<const FactoryBase> >("repartition: heuristic target rows per process", Teuchos::null, "Factory for number of rows per process to use with MultiJagged");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  const ParameterList& pL = GetParameterList();
+  // We do this dance, because we don't want "ParameterList" to be marked as used.
+  // Is there a better way?
+  Teuchos::ParameterEntry entry                  = pL.getEntry("ParameterList");
+  RCP<const Teuchos::ParameterList> providedList = Teuchos::any_cast<RCP<const Teuchos::ParameterList> >(entry.getAny(false));
+  if (providedList != Teuchos::null && providedList->isType<std::string>("algorithm")) {
+    const std::string algo = providedList->get<std::string>("algorithm");
+    if (algo == "multijagged") {
+      Input(currentLevel, "Coordinates");
       Input(currentLevel, "repartition: heuristic target rows per process");
+    } else if (algo == "rcb") {
       Input(currentLevel, "Coordinates");
     }
+  } else {
+    Input(currentLevel, "repartition: heuristic target rows per process");
+    Input(currentLevel, "Coordinates");
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
+  FactoryMonitor m(*this, "Build", level);
+
+  typedef typename Teuchos::ScalarTraits<SC>::coordinateType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<Matrix> A         = Get<RCP<Matrix> >(level, "A");
+  RCP<const Map> rowMap = A->getRowMap();
+  LO blkSize            = A->GetFixedBlockSize();
+
+  int numParts = Get<int>(level, "number of partitions");
+  if (numParts == 1 || numParts == -1) {
+    // Single processor, decomposition is trivial: all zeros
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+    Set(level, "Partition", decomposition);
+    return;
+  } /* else if (numParts == -1) {
+     // No repartitioning
+     RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Teuchos::null; //Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+     //decomposition->putScalar(Teuchos::as<Scalar>(rowMap->getComm()->getRank()));
+     Set(level, "Partition", decomposition);
+     return;
+   }*/
+
+  const ParameterList& pL = GetParameterList();
+
+  RCP<const ParameterList> providedList = pL.get<RCP<const ParameterList> >("ParameterList");
+  ParameterList Zoltan2Params;
+  if (providedList != Teuchos::null)
+    Zoltan2Params = *providedList;
+
+  // Merge defalt Zoltan2 parameters with user provided
+  // If default and user parameters contain the same parameter name, user one is always preferred
+  for (ParameterList::ConstIterator param = defaultZoltan2Params->begin(); param != defaultZoltan2Params->end(); param++) {
+    const std::string& pName = defaultZoltan2Params->name(param);
+    if (!Zoltan2Params.isParameter(pName))
+      Zoltan2Params.setEntry(pName, defaultZoltan2Params->getEntry(pName));
+  }
+  Zoltan2Params.set("num_global_parts", Teuchos::as<int>(numParts));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Zoltan2Interface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
-    FactoryMonitor m(*this, "Build", level);
-
-    typedef typename Teuchos::ScalarTraits<SC>::coordinateType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<Matrix>    A      = Get<RCP<Matrix> >(level, "A");
-    RCP<const Map> rowMap = A->getRowMap();
-    LO blkSize  = A->GetFixedBlockSize();
-
-    int numParts = Get<int>(level, "number of partitions");
-    if (numParts == 1 || numParts == -1) {
-      // Single processor, decomposition is trivial: all zeros
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      Set(level, "Partition", decomposition);
-      return;
-    }/* else if (numParts == -1) {
-      // No repartitioning
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Teuchos::null; //Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      //decomposition->putScalar(Teuchos::as<Scalar>(rowMap->getComm()->getRank()));
-      Set(level, "Partition", decomposition);
-      return;
-    }*/
-
-    const ParameterList& pL = GetParameterList();
-
-    RCP<const ParameterList> providedList = pL.get<RCP<const ParameterList> >("ParameterList");
-    ParameterList Zoltan2Params;
-    if (providedList != Teuchos::null)
-      Zoltan2Params = *providedList;
-
-    // Merge defalt Zoltan2 parameters with user provided
-    // If default and user parameters contain the same parameter name, user one is always preferred
-    for (ParameterList::ConstIterator param = defaultZoltan2Params->begin(); param != defaultZoltan2Params->end(); param++) {
-      const std::string& pName = defaultZoltan2Params->name(param);
-      if (!Zoltan2Params.isParameter(pName))
-        Zoltan2Params.setEntry(pName, defaultZoltan2Params->getEntry(pName));
-    }
-    Zoltan2Params.set("num_global_parts", Teuchos::as<int>(numParts));
-
-    GetOStream(Runtime0) << "Zoltan2 parameters:\n----------\n" << Zoltan2Params << "----------" << std::endl;
-
-    const std::string& algo = Zoltan2Params.get<std::string>("algorithm");
+  GetOStream(Runtime0) << "Zoltan2 parameters:\n----------\n"
+                       << Zoltan2Params << "----------" << std::endl;
 
-    if (algo == "multijagged" || algo == "rcb") {
+  const std::string& algo = Zoltan2Params.get<std::string>("algorithm");
 
-      RCP<RealValuedMultiVector> coords      = Get<RCP<RealValuedMultiVector> >(level, "Coordinates");
-      RCP<const Map>    map         = coords->getMap();
-      GO                numElements = map->getLocalNumElements();
+  if (algo == "multijagged" || algo == "rcb") {
+    RCP<RealValuedMultiVector> coords = Get<RCP<RealValuedMultiVector> >(level, "Coordinates");
+    RCP<const Map> map                = coords->getMap();
+    GO numElements                    = map->getLocalNumElements();
 
-      // Check that the number of local coordinates is consistent with the #rows in A
-      TEUCHOS_TEST_FOR_EXCEPTION(rowMap->getLocalNumElements()/blkSize != coords->getLocalLength(), Exceptions::Incompatible,
-                                 "Coordinate vector length (" + toString(coords->getLocalLength()) << " is incompatible with number of block rows in A ("
-                                 + toString(rowMap->getLocalNumElements()/blkSize) + "The vector length should be the same as the number of mesh points.");
+    // Check that the number of local coordinates is consistent with the #rows in A
+    TEUCHOS_TEST_FOR_EXCEPTION(rowMap->getLocalNumElements() / blkSize != coords->getLocalLength(), Exceptions::Incompatible,
+                               "Coordinate vector length (" + toString(coords->getLocalLength()) << " is incompatible with number of block rows in A (" + toString(rowMap->getLocalNumElements() / blkSize) + "The vector length should be the same as the number of mesh points.");
 #ifdef HAVE_MUELU_DEBUG
-      GO indexBase = rowMap->getIndexBase();
-      GetOStream(Runtime0) << "Checking consistence of row and coordinates maps" << std::endl;
-      // Make sure that logical blocks in row map coincide with logical nodes in coordinates map
-      ArrayView<const GO> rowElements    = rowMap->getLocalElementList();
-      ArrayView<const GO> coordsElements = map   ->getLocalElementList();
-      for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
-        TEUCHOS_TEST_FOR_EXCEPTION((coordsElements[i]-indexBase)*blkSize + indexBase != rowElements[i*blkSize],
-                                   Exceptions::RuntimeError, "i = " << i << ", coords GID = " << coordsElements[i]
-                                   << ", row GID = " << rowElements[i*blkSize] << ", blkSize = " << blkSize << std::endl);
+    GO indexBase = rowMap->getIndexBase();
+    GetOStream(Runtime0) << "Checking consistence of row and coordinates maps" << std::endl;
+    // Make sure that logical blocks in row map coincide with logical nodes in coordinates map
+    ArrayView<const GO> rowElements    = rowMap->getLocalElementList();
+    ArrayView<const GO> coordsElements = map->getLocalElementList();
+    for (LO i = 0; i < Teuchos::as<LO>(numElements); i++)
+      TEUCHOS_TEST_FOR_EXCEPTION((coordsElements[i] - indexBase) * blkSize + indexBase != rowElements[i * blkSize],
+                                 Exceptions::RuntimeError, "i = " << i << ", coords GID = " << coordsElements[i] << ", row GID = " << rowElements[i * blkSize] << ", blkSize = " << blkSize << std::endl);
 #endif
 
-      typedef Zoltan2::XpetraMultiVectorAdapter<RealValuedMultiVector>  InputAdapterType;
-      typedef Zoltan2::PartitioningProblem<InputAdapterType>   ProblemType;
+    typedef Zoltan2::XpetraMultiVectorAdapter<RealValuedMultiVector> InputAdapterType;
+    typedef Zoltan2::PartitioningProblem<InputAdapterType> ProblemType;
 
-      Array<real_type> weightsPerRow(numElements);
-      for (LO i = 0; i < numElements; i++) {
-        weightsPerRow[i] = 0.0;
+    Array<real_type> weightsPerRow(numElements);
+    for (LO i = 0; i < numElements; i++) {
+      weightsPerRow[i] = 0.0;
 
-        for (LO j = 0; j < blkSize; j++) {
-          weightsPerRow[i] += A->getNumEntriesInLocalRow(i*blkSize+j);
-        }
-      }
-     
-      // MultiJagged: Grab the target rows per process from the Heuristic to use unless the Zoltan2 list says otherwise
-      if(algo == "multijagged" && !Zoltan2Params.isParameter("mj_premigration_coordinate_count")) {
-        LO heuristicTargetRowsPerProcess = Get<LO>(level,"repartition: heuristic target rows per process");
-        Zoltan2Params.set("mj_premigration_coordinate_count", heuristicTargetRowsPerProcess);
+      for (LO j = 0; j < blkSize; j++) {
+        weightsPerRow[i] += A->getNumEntriesInLocalRow(i * blkSize + j);
       }
-      const bool writeZoltan2DebuggingFiles = Zoltan2Params.get("mj_debug",false);
-      Zoltan2Params.remove("mj_debug");
+    }
 
-      std::vector<int>           strides;
-      std::vector<const real_type*> weights(1, weightsPerRow.getRawPtr());
+    // MultiJagged: Grab the target rows per process from the Heuristic to use unless the Zoltan2 list says otherwise
+    if (algo == "multijagged" && !Zoltan2Params.isParameter("mj_premigration_coordinate_count")) {
+      LO heuristicTargetRowsPerProcess = Get<LO>(level, "repartition: heuristic target rows per process");
+      Zoltan2Params.set("mj_premigration_coordinate_count", heuristicTargetRowsPerProcess);
+    }
+    const bool writeZoltan2DebuggingFiles = Zoltan2Params.get("mj_debug", false);
+    Zoltan2Params.remove("mj_debug");
 
-      RCP<const Teuchos::MpiComm<int> >            dupMpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
-      RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
+    std::vector<int> strides;
+    std::vector<const real_type*> weights(1, weightsPerRow.getRawPtr());
 
-      InputAdapterType adapter(coords, weights, strides);
-      RCP<ProblemType> problem(new ProblemType(&adapter, &Zoltan2Params, (*zoltanComm)()));
+    RCP<const Teuchos::MpiComm<int> > dupMpiComm            = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
+    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
 
-      {
-        SubFactoryMonitor m1(*this, "Zoltan2 " + toString(algo), level);
-        if (writeZoltan2DebuggingFiles)
-          adapter.generateFiles(("mj_debug.lvl_"+std::to_string(level.GetLevelID())).c_str(), *(rowMap->getComm()));
-        problem->solve();
-      }
+    InputAdapterType adapter(coords, weights, strides);
+    RCP<ProblemType> problem(new ProblemType(&adapter, &Zoltan2Params, (*zoltanComm)()));
 
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO,LO,GO,NO>::Build(rowMap, false);
-      ArrayRCP<GO>                      decompEntries = decomposition->getDataNonConst(0);
+    {
+      SubFactoryMonitor m1(*this, "Zoltan2 " + toString(algo), level);
+      if (writeZoltan2DebuggingFiles)
+        adapter.generateFiles(("mj_debug.lvl_" + std::to_string(level.GetLevelID())).c_str(), *(rowMap->getComm()));
+      problem->solve();
+    }
 
-      const typename InputAdapterType::part_t * parts = problem->getSolution().getPartListView();
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);
+    ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
 
-      for (GO i = 0; i < numElements; i++) {
-        int partNum = parts[i];
+    const typename InputAdapterType::part_t* parts = problem->getSolution().getPartListView();
 
-        for (LO j = 0; j < blkSize; j++)
-          decompEntries[i*blkSize + j] = partNum;
-      }
+    for (GO i = 0; i < numElements; i++) {
+      int partNum = parts[i];
 
-      Set(level, "Partition", decomposition);
+      for (LO j = 0; j < blkSize; j++)
+        decompEntries[i * blkSize + j] = partNum;
+    }
 
-    } else {
+    Set(level, "Partition", decomposition);
 
-      GO numElements = rowMap->getLocalNumElements();
+  } else {
+    GO numElements = rowMap->getLocalNumElements();
 
-      typedef Zoltan2::XpetraCrsGraphAdapter<CrsGraph>  InputAdapterType;
-      typedef Zoltan2::PartitioningProblem<InputAdapterType>   ProblemType;
+    typedef Zoltan2::XpetraCrsGraphAdapter<CrsGraph> InputAdapterType;
+    typedef Zoltan2::PartitioningProblem<InputAdapterType> ProblemType;
 
-      RCP<const Teuchos::MpiComm<int> >            dupMpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
-      RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
+    RCP<const Teuchos::MpiComm<int> > dupMpiComm            = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
+    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
 
-      InputAdapterType adapter(A->getCrsGraph());
-      RCP<ProblemType> problem(new ProblemType(&adapter, &Zoltan2Params, (*zoltanComm)()));
+    InputAdapterType adapter(A->getCrsGraph());
+    RCP<ProblemType> problem(new ProblemType(&adapter, &Zoltan2Params, (*zoltanComm)()));
 
-      {
-        SubFactoryMonitor m1(*this, "Zoltan2 " + toString(algo), level);
-        problem->solve();
-      }
+    {
+      SubFactoryMonitor m1(*this, "Zoltan2 " + toString(algo), level);
+      problem->solve();
+    }
 
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Xpetra::VectorFactory<GO,LO,GO,NO>::Build(rowMap, false);
-      ArrayRCP<GO>                      decompEntries = decomposition->getDataNonConst(0);
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);
+    ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
 
-      const typename InputAdapterType::part_t * parts = problem->getSolution().getPartListView();
+    const typename InputAdapterType::part_t* parts = problem->getSolution().getPartListView();
 
-      // For blkSize > 1, ignore solution for every row but the first ones in a block.
-      for (GO i = 0; i < numElements/blkSize; i++) {
-	int partNum = parts[i*blkSize];
-	
-	for (LO j = 0; j < blkSize; j++)
-	  decompEntries[i*blkSize + j] = partNum;
-      }
+    // For blkSize > 1, ignore solution for every row but the first ones in a block.
+    for (GO i = 0; i < numElements / blkSize; i++) {
+      int partNum = parts[i * blkSize];
 
-      Set(level, "Partition", decomposition);
+      for (LO j = 0; j < blkSize; j++)
+        decompEntries[i * blkSize + j] = partNum;
     }
+
+    Set(level, "Partition", decomposition);
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif //if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
+#endif  // if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
 
-#endif // MUELU_ZOLTAN2INTERFACE_DEF_HPP
+#endif  // MUELU_ZOLTAN2INTERFACE_DEF_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_decl.hpp b/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_decl.hpp
index a672cfbb09ac..c29b79855b63 100644
--- a/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_decl.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_decl.hpp
@@ -63,135 +63,132 @@
 
 namespace MueLu {
 
-  /*!
-    @class ZoltanInterface
-    @brief Interface to Zoltan library.
-    @ingroup Rebalancing
-
-    This interface provides access to partitioning methods in Zoltan.
-    Currently, it supports the RCB algorithm only.
-
-    ## Input/output of ZoltanInterface ##
-
-    ### User parameters of ZoltanInterface ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | Coordinates                            | Factory | null  |   | * | * | Factory generating coordinates vector used for rebalancing (RCB algorithm)
-    | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ZoltanInterface::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see ZoltanInterface::DeclareInput).
-
-    ### Variables provided by ZoltanInterface ###
-
-    After ZoltanInterface::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Partition       | ZoltanInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
-
-    The "Partition" vector is used as input for the RepartitionFactory class.
-    If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
-    by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
-
-    The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
-    It stores the result in the "number of partitions" variable on the current level (type = GO).
-    If it is "number of partitions=1" we skip the Zoltan call and just create an dummy "Partition" vector containing zeros only.
-    If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
-    If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
-
-    @note: The ZoltanInterface class does not have all the features availabe in the Zoltan2Interface class
-  */
-
-  //FIXME: this class should not be templated
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ZoltanInterface : public SingleLevelFactoryBase {
+/*!
+  @class ZoltanInterface
+  @brief Interface to Zoltan library.
+  @ingroup Rebalancing
+
+  This interface provides access to partitioning methods in Zoltan.
+  Currently, it supports the RCB algorithm only.
+
+  ## Input/output of ZoltanInterface ##
+
+  ### User parameters of ZoltanInterface ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | Coordinates                            | Factory | null  |   | * | * | Factory generating coordinates vector used for rebalancing (RCB algorithm)
+  | number of partitions                   | GO      | - |  |  |  | Short-cut parameter set by RepartitionFactory. Avoid repartitioning algorithms if only one partition is necessary (see details below)
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ZoltanInterface::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see ZoltanInterface::DeclareInput).
+
+  ### Variables provided by ZoltanInterface ###
+
+  After ZoltanInterface::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Partition       | ZoltanInterface   | GOVector based on the Row map of A (DOF-based) containing the process id the DOF should be living in after rebalancing/repartitioning
+
+  The "Partition" vector is used as input for the RepartitionFactory class.
+  If Re-partitioning/rebalancing is necessary it uses the "Partition" variable to create the corresponding Xpetra::Import object which then is used
+  by the RebalanceFactory classes (e.g., RebalanceAcFactory, RebalanceTransferFactory,...) to rebalance the coarse level operators.
+
+  The RepartitionHeuristicFactory calculates how many partitions are to be built when performing rebalancing.
+  It stores the result in the "number of partitions" variable on the current level (type = GO).
+  If it is "number of partitions=1" we skip the Zoltan call and just create an dummy "Partition" vector containing zeros only.
+  If no repartitioning is necessary (i.e., just keep the current partitioning) we return "Partition = Teuchos::null".
+  If "number of partitions" > 1, the algorithm tries to find the requested number of partitions.
+
+  @note: The ZoltanInterface class does not have all the features availabe in the Zoltan2Interface class
+*/
+
+// FIXME: this class should not be templated
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ZoltanInterface : public SingleLevelFactoryBase {
 #undef MUELU_ZOLTANINTERFACE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors
-    //@{
-
-    //! Constructor
-    ZoltanInterface() {}
-
-    //! Destructor
-    virtual ~ZoltanInterface() { }
-    //@}
+ public:
+  //! @name Constructors/Destructors
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Constructor
+  ZoltanInterface() {}
 
-    //! @name Input
-    //@{
-    void DeclareInput(Level& level) const;
-    //@}
+  //! Destructor
+  virtual ~ZoltanInterface() {}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level& level) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //! @name Input
+  //@{
+  void DeclareInput(Level &level) const;
+  //@}
 
-    //! @name Query methods (really functions) required by Zoltan.
-    //@{
+  //! @name Build methods.
+  //@{
+  void Build(Level &level) const;
 
-    /*! Callback function that returns the local number of objects. Required by Zoltan.
+  //@}
 
-    In this case, the number of objects is the number of local rows.
+  //! @name Query methods (really functions) required by Zoltan.
+  //@{
 
-    @param data (in) void pointer to an Xpetra::Matrix.
-    @param ierr (out) error code.
-    */
-    static int GetLocalNumberOfRows(void *data, int *ierr);
+  /*! Callback function that returns the local number of objects. Required by Zoltan.
 
-    /*! Callback function that returns the local number of nonzeros in the matrix. Required by Zoltan.
+  In this case, the number of objects is the number of local rows.
 
-    FIXME: Note that this will not work properly for non-point matrices.
+  @param data (in) void pointer to an Xpetra::Matrix.
+  @param ierr (out) error code.
+  */
+  static int GetLocalNumberOfRows(void *data, int *ierr);
 
-    @param data (in) void pointer to an Xpetra::Matrix
-    @param weights (out) array whose <tt>i</tt><sup>th</sup> entry is the number of nonzeros in local row \c i.
-    @param ierr (out) error code
-    */
-    static void GetLocalNumberOfNonzeros(void *data, int NumGidEntries, int NumLidEntries, ZOLTAN_ID_PTR gids,
-                                         ZOLTAN_ID_PTR lids, int wgtDim, float *weights, int *ierr);
+  /*! Callback function that returns the local number of nonzeros in the matrix. Required by Zoltan.
 
-    /*! Callback function that returns the problem dimension. Required by Zoltan.
+  FIXME: Note that this will not work properly for non-point matrices.
 
-    @param data (in) void pointer to integer dimension
-    @param ierr (out) error code
-    */
-    static int GetProblemDimension(void *data, int *ierr);
+  @param data (in) void pointer to an Xpetra::Matrix
+  @param weights (out) array whose <tt>i</tt><sup>th</sup> entry is the number of nonzeros in local row \c i.
+  @param ierr (out) error code
+  */
+  static void GetLocalNumberOfNonzeros(void *data, int NumGidEntries, int NumLidEntries, ZOLTAN_ID_PTR gids,
+                                       ZOLTAN_ID_PTR lids, int wgtDim, float *weights, int *ierr);
 
+  /*! Callback function that returns the problem dimension. Required by Zoltan.
 
-    /*! Callback function that returns the problem dimension. Required by Zoltan.
+  @param data (in) void pointer to integer dimension
+  @param ierr (out) error code
+  */
+  static int GetProblemDimension(void *data, int *ierr);
 
-    @param data (in) void pointer to Xpetra::MultiVector.
-    @param coordinates (out) array of double coordinates, arranged like so: [x1 y1 z1 x2 y2 z2 ...].
-    @param ierr (out) error code
+  /*! Callback function that returns the problem dimension. Required by Zoltan.
 
-    TODO -- should I return a view of the coordinates instead of copying them?
-    */
-    static void GetProblemGeometry(void *data, int numGIDEntries, int numLIDEntries, int numObjectIDs,
-                                   ZOLTAN_ID_PTR gids, ZOLTAN_ID_PTR lids, int dim, double *coordinates, int *ierr);
+  @param data (in) void pointer to Xpetra::MultiVector.
+  @param coordinates (out) array of double coordinates, arranged like so: [x1 y1 z1 x2 y2 z2 ...].
+  @param ierr (out) error code
 
-    //@}
+  TODO -- should I return a view of the coordinates instead of copying them?
+  */
+  static void GetProblemGeometry(void *data, int numGIDEntries, int numLIDEntries, int numObjectIDs,
+                                 ZOLTAN_ID_PTR gids, ZOLTAN_ID_PTR lids, int dim, double *coordinates, int *ierr);
 
-  private:
+  //@}
 
-    static ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coalesceCoordinates(ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coord, LocalOrdinal blksize);
+ private:
+  static ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coalesceCoordinates(ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coord, LocalOrdinal blksize);
 
-  };  //class ZoltanInterface
+};  // class ZoltanInterface
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ZOLTANINTERFACE_SHORT
-#endif //if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
+#endif  // if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
 
-#endif // MUELU_ZOLTANINTERFACE_DECL_HPP
+#endif  // MUELU_ZOLTANINTERFACE_DECL_HPP
diff --git a/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_def.hpp b/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_def.hpp
index 1eb440057a63..97a14ce5ea40 100644
--- a/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_def.hpp
+++ b/packages/muelu/src/Rebalancing/MueLu_ZoltanInterface_def.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
 
 #include <Teuchos_Utils.hpp>
-#include <Teuchos_DefaultMpiComm.hpp> //TODO: fwd decl.
-#include <Teuchos_OpaqueWrapper.hpp>  //TODO: fwd decl.
+#include <Teuchos_DefaultMpiComm.hpp>  //TODO: fwd decl.
+#include <Teuchos_OpaqueWrapper.hpp>   //TODO: fwd decl.
 
 #include "MueLu_Level.hpp"
 #include "MueLu_Exceptions.hpp"
@@ -59,260 +59,258 @@
 
 namespace MueLu {
 
- template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
- RCP<const ParameterList> ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory of the coordinates");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "number of partitions");
+  Input(currentLevel, "Coordinates");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &level) const {
+  FactoryMonitor m(*this, "Build", level);
+
+  RCP<Matrix> A            = Get<RCP<Matrix> >(level, "A");
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A);
+  RCP<const Map> rowMap;
+  if (bA != Teuchos::null) {
+    // Extracting the full the row map here...
+    RCP<const Map> bArowMap       = bA->getRowMap();
+    RCP<const BlockedMap> bRowMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(bArowMap);
+    rowMap                        = bRowMap->getFullMap();
+  } else {
+    rowMap = A->getRowMap();
+  }
 
-    validParamList->set< RCP<const FactoryBase> >("A",           Teuchos::null, "Factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("number of partitions", Teuchos::null, "Instance of RepartitionHeuristicFactory.");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory of the coordinates");
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
+  RCP<double_multivector_type> Coords = Get<RCP<double_multivector_type> >(level, "Coordinates");
+  size_t dim                          = Coords->getNumVectors();
+  int numParts                        = Get<int>(level, "number of partitions");
 
-    return validParamList;
+  if (numParts == 1 || numParts == -1) {
+    // Running on one processor, so decomposition is the trivial one, all zeros.
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
+    Set(level, "Partition", decomposition);
+    return;
+  } else if (numParts == -1) {
+    // No repartitioning
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Teuchos::null;
+    Set(level, "Partition", decomposition);
+    return;
   }
 
+  float zoltanVersion_;
+  Zoltan_Initialize(0, NULL, &zoltanVersion_);
+
+  RCP<const Teuchos::MpiComm<int> > dupMpiComm            = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
+  RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
+
+  RCP<Zoltan> zoltanObj_ = rcp(new Zoltan((*zoltanComm)()));  // extract the underlying MPI_Comm handle and create a Zoltan object
+  if (zoltanObj_ == Teuchos::null)
+    throw Exceptions::RuntimeError("MueLu::Zoltan : Unable to create Zoltan data structure");
+
+  // Tell Zoltan what kind of local/global IDs we will use.
+  // In our case, each GID is two ints and there are no local ids.
+  // One can skip this step if the IDs are just single ints.
+  int rv;
+  if ((rv = zoltanObj_->Set_Param("num_gid_entries", "1")) != ZOLTAN_OK)
+    throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'num_gid_entries' returned error code " + Teuchos::toString(rv));
+  if ((rv = zoltanObj_->Set_Param("num_lid_entries", "0")) != ZOLTAN_OK)
+    throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'num_lid_entries' returned error code " + Teuchos::toString(rv));
+  if ((rv = zoltanObj_->Set_Param("obj_weight_dim", "1")) != ZOLTAN_OK)
+    throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'obj_weight_dim' returned error code " + Teuchos::toString(rv));
+
+  if (GetVerbLevel() & Statistics1)
+    zoltanObj_->Set_Param("debug_level", "1");
+  else
+    zoltanObj_->Set_Param("debug_level", "0");
+
+  zoltanObj_->Set_Param("num_global_partitions", toString(numParts));
+
+  zoltanObj_->Set_Num_Obj_Fn(GetLocalNumberOfRows, (void *)A.getRawPtr());
+  zoltanObj_->Set_Obj_List_Fn(GetLocalNumberOfNonzeros, (void *)A.getRawPtr());
+  zoltanObj_->Set_Num_Geom_Fn(GetProblemDimension, (void *)&dim);
+  zoltanObj_->Set_Geom_Multi_Fn(GetProblemGeometry, (void *)Coords.get());
+
+  // Data pointers that Zoltan requires.
+  ZOLTAN_ID_PTR import_gids = NULL;  // Global nums of objs to be imported
+  ZOLTAN_ID_PTR import_lids = NULL;  // Local indices to objs to be imported
+  int *import_procs         = NULL;  // Proc IDs of procs owning objs to be imported.
+  int *import_to_part       = NULL;  // Partition #s to which imported objs should be assigned.
+  ZOLTAN_ID_PTR export_gids = NULL;  // Global nums of objs to be exported
+  ZOLTAN_ID_PTR export_lids = NULL;  // local indices to objs to be exported
+  int *export_procs         = NULL;  // Proc IDs of destination procs for objs to be exported.
+  int *export_to_part       = NULL;  // Partition #s for objs to be exported.
+  int num_imported;                  // Number of objs to be imported.
+  int num_exported;                  // Number of objs to be exported.
+  int newDecomp;                     // Flag indicating whether the decomposition has changed
+  int num_gid_entries;               // Number of array entries in a global ID.
+  int num_lid_entries;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-    Input(currentLevel, "number of partitions");
-    Input(currentLevel, "Coordinates");
+  {
+    SubFactoryMonitor m1(*this, "Zoltan RCB", level);
+    rv = zoltanObj_->LB_Partition(newDecomp, num_gid_entries, num_lid_entries,
+                                  num_imported, import_gids, import_lids, import_procs, import_to_part,
+                                  num_exported, export_gids, export_lids, export_procs, export_to_part);
+    if (rv == ZOLTAN_FATAL)
+      throw Exceptions::RuntimeError("Zoltan::LB_Partition() returned error code");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& level) const {
-    FactoryMonitor m(*this, "Build", level);
-
-    RCP<Matrix>      A        = Get< RCP<Matrix> >     (level, "A");
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A);
-    RCP<const Map>   rowMap;
-    if(bA != Teuchos::null) {
-      // Extracting the full the row map here...
-      RCP<const Map> bArowMap = bA->getRowMap();
-      RCP<const BlockedMap> bRowMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(bArowMap);
-      rowMap = bRowMap->getFullMap();
-    } else {
-      rowMap = A->getRowMap();
-    }
+  // TODO check that A's row map is 1-1.  Zoltan requires this.
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
-    RCP<double_multivector_type> Coords   = Get< RCP<double_multivector_type> >(level, "Coordinates");
-    size_t           dim      = Coords->getNumVectors();
-    int numParts = Get<int>(level, "number of partitions");
-
-    if (numParts == 1 || numParts == -1) {
-      // Running on one processor, so decomposition is the trivial one, all zeros.
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, true);
-      Set(level, "Partition", decomposition);
-      return;
-    } else if (numParts == -1) {
-      // No repartitioning
-      RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = Teuchos::null;
-      Set(level, "Partition", decomposition);
-      return;
-    }
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition;
+  if (newDecomp) {
+    decomposition              = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false);  // Don't initialize, will be overwritten
+    ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
 
-    float zoltanVersion_;
-    Zoltan_Initialize(0, NULL, &zoltanVersion_);
-
-    RCP<const Teuchos::MpiComm<int> >            dupMpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(rowMap->getComm()->duplicate());
-    RCP<const Teuchos::OpaqueWrapper<MPI_Comm> > zoltanComm = dupMpiComm->getRawMpiComm();
-
-    RCP<Zoltan> zoltanObj_ = rcp(new Zoltan((*zoltanComm)()));  //extract the underlying MPI_Comm handle and create a Zoltan object
-    if (zoltanObj_ == Teuchos::null)
-      throw Exceptions::RuntimeError("MueLu::Zoltan : Unable to create Zoltan data structure");
-
-    // Tell Zoltan what kind of local/global IDs we will use.
-    // In our case, each GID is two ints and there are no local ids.
-    // One can skip this step if the IDs are just single ints.
-    int rv;
-    if ((rv = zoltanObj_->Set_Param("num_gid_entries", "1")) != ZOLTAN_OK)
-      throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'num_gid_entries' returned error code " + Teuchos::toString(rv));
-    if ((rv = zoltanObj_->Set_Param("num_lid_entries", "0") ) != ZOLTAN_OK)
-      throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'num_lid_entries' returned error code " + Teuchos::toString(rv));
-    if ((rv = zoltanObj_->Set_Param("obj_weight_dim", "1") ) != ZOLTAN_OK)
-      throw Exceptions::RuntimeError("MueLu::Zoltan::Setup : setting parameter 'obj_weight_dim' returned error code "  + Teuchos::toString(rv));
-
-    if (GetVerbLevel() & Statistics1) zoltanObj_->Set_Param("debug_level", "1");
-    else                              zoltanObj_->Set_Param("debug_level", "0");
-
-    zoltanObj_->Set_Param("num_global_partitions", toString(numParts));
-
-    zoltanObj_->Set_Num_Obj_Fn(GetLocalNumberOfRows,      (void *) A.getRawPtr());
-    zoltanObj_->Set_Obj_List_Fn(GetLocalNumberOfNonzeros, (void *) A.getRawPtr());
-    zoltanObj_->Set_Num_Geom_Fn(GetProblemDimension,      (void *) &dim);
-    zoltanObj_->Set_Geom_Multi_Fn(GetProblemGeometry,     (void *) Coords.get());
-
-    // Data pointers that Zoltan requires.
-    ZOLTAN_ID_PTR import_gids = NULL;  // Global nums of objs to be imported
-    ZOLTAN_ID_PTR import_lids = NULL;  // Local indices to objs to be imported
-    int   *import_procs       = NULL;  // Proc IDs of procs owning objs to be imported.
-    int   *import_to_part     = NULL;  // Partition #s to which imported objs should be assigned.
-    ZOLTAN_ID_PTR export_gids = NULL;  // Global nums of objs to be exported
-    ZOLTAN_ID_PTR export_lids = NULL;  // local indices to objs to be exported
-    int   *export_procs       = NULL;  // Proc IDs of destination procs for objs to be exported.
-    int   *export_to_part     = NULL;  // Partition #s for objs to be exported.
-    int   num_imported;                // Number of objs to be imported.
-    int   num_exported;                // Number of objs to be exported.
-    int   newDecomp;                   // Flag indicating whether the decomposition has changed
-    int   num_gid_entries;             // Number of array entries in a global ID.
-    int   num_lid_entries;
-
-    {
-      SubFactoryMonitor m1(*this, "Zoltan RCB", level);
-      rv = zoltanObj_->LB_Partition(newDecomp, num_gid_entries, num_lid_entries,
-                                    num_imported, import_gids, import_lids, import_procs, import_to_part,
-                                    num_exported, export_gids, export_lids, export_procs, export_to_part);
-      if (rv == ZOLTAN_FATAL)
-        throw Exceptions::RuntimeError("Zoltan::LB_Partition() returned error code");
-    }
+    int mypid = rowMap->getComm()->getRank();
+    for (typename ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
+      *i = mypid;
 
-    // TODO check that A's row map is 1-1.  Zoltan requires this.
-
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition;
-    if (newDecomp) {
-      decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(rowMap, false); // Don't initialize, will be overwritten
-      ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-
-      int mypid = rowMap->getComm()->getRank();
-      for (typename ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
-        *i = mypid;
-
-      LO blockSize = A->GetFixedBlockSize();
-      for (int i = 0; i < num_exported; ++i) {
-        // We have assigned Zoltan gids to first row GID in the block
-        // NOTE: Zoltan GIDs are different from GIDs in the Coordinates vector
-        LO  localEl = rowMap->getLocalElement(export_gids[i]);
-        int partNum = export_to_part[i];
-        for (LO j = 0; j < blockSize; ++j)
-          decompEntries[localEl + j] = partNum;
-      }
+    LO blockSize = A->GetFixedBlockSize();
+    for (int i = 0; i < num_exported; ++i) {
+      // We have assigned Zoltan gids to first row GID in the block
+      // NOTE: Zoltan GIDs are different from GIDs in the Coordinates vector
+      LO localEl  = rowMap->getLocalElement(export_gids[i]);
+      int partNum = export_to_part[i];
+      for (LO j = 0; j < blockSize; ++j)
+        decompEntries[localEl + j] = partNum;
     }
+  }
 
-    Set(level, "Partition", decomposition);
+  Set(level, "Partition", decomposition);
 
-    zoltanObj_->LB_Free_Part(&import_gids, &import_lids, &import_procs, &import_to_part);
-    zoltanObj_->LB_Free_Part(&export_gids, &export_lids, &export_procs, &export_to_part);
+  zoltanObj_->LB_Free_Part(&import_gids, &import_lids, &import_procs, &import_to_part);
+  zoltanObj_->LB_Free_Part(&export_gids, &export_lids, &export_procs, &export_to_part);
 
-  } //Build()
+}  // Build()
 
-  //-------------------------------------------------------------------------------------------------------------
-  // GetLocalNumberOfRows
-  //-------------------------------------------------------------------------------------------------------------
+//-------------------------------------------------------------------------------------------------------------
+// GetLocalNumberOfRows
+//-------------------------------------------------------------------------------------------------------------
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLocalNumberOfRows(void *data, int *ierr) {
-    if (data == NULL) {
-      *ierr = ZOLTAN_FATAL;
-      return -1;
-    }
-    Matrix *A = (Matrix*) data;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetLocalNumberOfRows(void *data, int *ierr) {
+  if (data == NULL) {
+    *ierr = ZOLTAN_FATAL;
+    return -1;
+  }
+  Matrix *A = (Matrix *)data;
+  *ierr     = ZOLTAN_OK;
+
+  LO blockSize = A->GetFixedBlockSize();
+  TEUCHOS_TEST_FOR_EXCEPTION(blockSize == 0, Exceptions::RuntimeError, "MueLu::Zoltan : Matrix has block size 0.");
+
+  return A->getRowMap()->getLocalNumElements() / blockSize;
+}  // GetLocalNumberOfRows()
+
+//-------------------------------------------------------------------------------------------------------------
+// GetLocalNumberOfNonzeros
+//-------------------------------------------------------------------------------------------------------------
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetLocalNumberOfNonzeros(void *data, int NumGidEntries, int /* NumLidEntries */, ZOLTAN_ID_PTR gids,
+                             ZOLTAN_ID_PTR /* lids */, int /* wgtDim */, float *weights, int *ierr) {
+  if (data == NULL || NumGidEntries < 1) {
+    *ierr = ZOLTAN_FATAL;
+    return;
+  } else {
     *ierr = ZOLTAN_OK;
-
-    LO blockSize = A->GetFixedBlockSize();
-    TEUCHOS_TEST_FOR_EXCEPTION(blockSize == 0, Exceptions::RuntimeError, "MueLu::Zoltan : Matrix has block size 0.");
-
-    return A->getRowMap()->getLocalNumElements() / blockSize;
-  } //GetLocalNumberOfRows()
-
-  //-------------------------------------------------------------------------------------------------------------
-  // GetLocalNumberOfNonzeros
-  //-------------------------------------------------------------------------------------------------------------
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetLocalNumberOfNonzeros(void *data, int NumGidEntries, int /* NumLidEntries */, ZOLTAN_ID_PTR gids,
-                           ZOLTAN_ID_PTR /* lids */, int /* wgtDim */, float *weights, int *ierr) {
-    if (data == NULL || NumGidEntries < 1) {
-      *ierr = ZOLTAN_FATAL;
-      return;
-    } else {
-      *ierr = ZOLTAN_OK;
-    }
-
-    Matrix         *A  = (Matrix*) data;
-    RCP<const Map> map = A->getRowMap();
-
-    LO blockSize = A->GetFixedBlockSize();
-    TEUCHOS_TEST_FOR_EXCEPTION(blockSize == 0, Exceptions::RuntimeError, "MueLu::Zoltan : Matrix has block size 0.");
-
-    size_t              numElements = map->getLocalNumElements();
-    ArrayView<const GO> mapGIDs     = map->getLocalElementList();
-
-    if (blockSize == 1) {
-      for (size_t i = 0; i < numElements; i++) {
-        gids[i]    = as<ZOLTAN_ID_TYPE>(mapGIDs[i]);
-        weights[i] = A->getNumEntriesInLocalRow(i);
-      }
-
-    } else {
-      LO numBlockElements = numElements / blockSize;
-
-      for (LO i = 0; i < numBlockElements; i++) {
-        // Assign zoltan GID to the first row GID in the block
-        // NOTE: Zoltan GIDs are different from GIDs in the Coordinates vector
-        gids[i]    = as<ZOLTAN_ID_TYPE>(mapGIDs[i*blockSize]);
-        weights[i] = 0.0;
-        for (LO j = 0; j < blockSize; j++)
-          weights[i] += A->getNumEntriesInLocalRow(i*blockSize+j);
-      }
-    }
-
   }
 
-  //-------------------------------------------------------------------------------------------------------------
-  // GetProblemDimension
-  //-------------------------------------------------------------------------------------------------------------
+  Matrix *A          = (Matrix *)data;
+  RCP<const Map> map = A->getRowMap();
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetProblemDimension(void *data, int *ierr)
-  {
-    int dim = *((int*)data);
-    *ierr = ZOLTAN_OK;
+  LO blockSize = A->GetFixedBlockSize();
+  TEUCHOS_TEST_FOR_EXCEPTION(blockSize == 0, Exceptions::RuntimeError, "MueLu::Zoltan : Matrix has block size 0.");
 
-    return dim;
-  } //GetProblemDimension
+  size_t numElements          = map->getLocalNumElements();
+  ArrayView<const GO> mapGIDs = map->getLocalElementList();
 
-  //-------------------------------------------------------------------------------------------------------------
-  // GetProblemGeometry
-  //-------------------------------------------------------------------------------------------------------------
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetProblemGeometry(void *data, int /* numGIDEntries */, int /* numLIDEntries */, int numObjectIDs,
-                     ZOLTAN_ID_PTR /* gids */, ZOLTAN_ID_PTR /* lids */, int dim, double *coordinates, int *ierr)
-  {
-    if (data == NULL) {
-      *ierr = ZOLTAN_FATAL;
-      return;
+  if (blockSize == 1) {
+    for (size_t i = 0; i < numElements; i++) {
+      gids[i]    = as<ZOLTAN_ID_TYPE>(mapGIDs[i]);
+      weights[i] = A->getNumEntriesInLocalRow(i);
     }
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
-    double_multivector_type *Coords = (double_multivector_type*) data;
+  } else {
+    LO numBlockElements = numElements / blockSize;
 
-    if (dim != Teuchos::as<int>(Coords->getNumVectors())) {
-      //FIXME I'm assuming dim should be 1, 2, or 3 coming in?!
-      *ierr = ZOLTAN_FATAL;
-      return;
+    for (LO i = 0; i < numBlockElements; i++) {
+      // Assign zoltan GID to the first row GID in the block
+      // NOTE: Zoltan GIDs are different from GIDs in the Coordinates vector
+      gids[i]    = as<ZOLTAN_ID_TYPE>(mapGIDs[i * blockSize]);
+      weights[i] = 0.0;
+      for (LO j = 0; j < blockSize; j++)
+        weights[i] += A->getNumEntriesInLocalRow(i * blockSize + j);
     }
+  }
+}
+
+//-------------------------------------------------------------------------------------------------------------
+// GetProblemDimension
+//-------------------------------------------------------------------------------------------------------------
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetProblemDimension(void *data, int *ierr) {
+  int dim = *((int *)data);
+  *ierr   = ZOLTAN_OK;
+
+  return dim;
+}  // GetProblemDimension
+
+//-------------------------------------------------------------------------------------------------------------
+// GetProblemGeometry
+//-------------------------------------------------------------------------------------------------------------
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZoltanInterface<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetProblemGeometry(void *data, int /* numGIDEntries */, int /* numLIDEntries */, int numObjectIDs,
+                       ZOLTAN_ID_PTR /* gids */, ZOLTAN_ID_PTR /* lids */, int dim, double *coordinates, int *ierr) {
+  if (data == NULL) {
+    *ierr = ZOLTAN_FATAL;
+    return;
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(numObjectIDs != Teuchos::as<int>(Coords->getLocalLength()), Exceptions::Incompatible, "Length of coordinates must be the same as the number of objects");
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
+  double_multivector_type *Coords = (double_multivector_type *)data;
 
-    ArrayRCP<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > CoordsData(dim);
-    for (int j = 0; j < dim; ++j)
-      CoordsData[j] = Coords->getData(j);
+  if (dim != Teuchos::as<int>(Coords->getNumVectors())) {
+    // FIXME I'm assuming dim should be 1, 2, or 3 coming in?!
+    *ierr = ZOLTAN_FATAL;
+    return;
+  }
 
-    size_t numElements = Coords->getLocalLength();
-    for (size_t i = 0; i < numElements; ++i)
-      for (int j = 0; j < dim; ++j)
-        coordinates[i*dim+j] = (double) CoordsData[j][i];
+  TEUCHOS_TEST_FOR_EXCEPTION(numObjectIDs != Teuchos::as<int>(Coords->getLocalLength()), Exceptions::Incompatible, "Length of coordinates must be the same as the number of objects");
 
-    *ierr = ZOLTAN_OK;
+  ArrayRCP<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > CoordsData(dim);
+  for (int j = 0; j < dim; ++j)
+    CoordsData[j] = Coords->getData(j);
+
+  size_t numElements = Coords->getLocalLength();
+  for (size_t i = 0; i < numElements; ++i)
+    for (int j = 0; j < dim; ++j)
+      coordinates[i * dim + j] = (double)CoordsData[j][i];
+
+  *ierr = ZOLTAN_OK;
 
-  } //GetProblemGeometry
+}  // GetProblemGeometry
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif //if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
+#endif  // if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
 
-#endif // MUELU_ZOLTANINTERFACE_DEF_HPP
+#endif  // MUELU_ZOLTANINTERFACE_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_decl.hpp
index bd85df304524..4a4b6f702b37 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_decl.hpp
@@ -67,96 +67,93 @@
 
 namespace MueLu {
 
-  /*!
-    @class BlockedDirectSolver
-    @brief direct solver for nxn blocked matrices
-
-    The nxn block matrix A as input is automatically merged and then solved
-    by a direct solver.
-      */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BlockedDirectSolver : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+/*!
+  @class BlockedDirectSolver
+  @brief direct solver for nxn blocked matrices
 
-#undef MUELU_BLOCKEDDIRECTSOLVER_SHORT
-#include "MueLu_UseShortNames.hpp"
+  The nxn block matrix A as input is automatically merged and then solved
+  by a direct solver.
+    */
 
-  public:
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BlockedDirectSolver : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
-    //! @name Constructors / destructors
-    //@{
+#undef MUELU_BLOCKEDDIRECTSOLVER_SHORT
+#include "MueLu_UseShortNames.hpp"
 
-    /*! @brief Constructor
-    */
-    BlockedDirectSolver(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! Destructor
-    virtual ~BlockedDirectSolver() { }
-    //@}
+  /*! @brief Constructor
+   */
+  BlockedDirectSolver(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor
+  virtual ~BlockedDirectSolver() {}
+  //@}
 
-    void DeclareInput(Level &currentLevel) const;
-    //@}
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Setup and Apply methods.
-    //@{
+  void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    /*! @brief Setup routine
-     * Call the underlaying Setup routine of the nested direct solver once the input block matrix has been merged
-    */
-    void Setup(Level &currentLevel);
+  //! @name Setup and Apply methods.
+  //@{
 
-    /*! @brief Apply the direct solver.
-    Solves the linear system <tt>AX=B</tt> using the constructed solver.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
+  /*! @brief Setup routine
+   * Call the underlaying Setup routine of the nested direct solver once the input block matrix has been merged
+   */
+  void Setup(Level& currentLevel);
 
-    RCP<SmootherPrototype> Copy() const;
+  /*! @brief Apply the direct solver.
+  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+  //@}
 
-    //! smoother type
-    std::string type_;
+ private:
+  //! smoother type
+  std::string type_;
 
-    //! Factory to generate merged block matrix
-    RCP<MergedBlockedMatrixFactory> MergedAFact_;
+  //! Factory to generate merged block matrix
+  RCP<MergedBlockedMatrixFactory> MergedAFact_;
 
-    //! Direct solver
-    RCP<DirectSolver> s_; // solver object
+  //! Direct solver
+  RCP<DirectSolver> s_;  // solver object
 
-    //! block operator
-    RCP<Matrix> A_;                  // < ! internal blocked operator "A" generated by AFact_
+  //! block operator
+  RCP<Matrix> A_;  // < ! internal blocked operator "A" generated by AFact_
 
-  }; // class BlockedDirectSolver
+};  // class BlockedDirectSolver
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDDIRECTSOLVER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_def.hpp
index c9199c353b23..be03156533c4 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedDirectSolver_def.hpp
@@ -70,140 +70,136 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedDirectSolver(const std::string& type, const Teuchos::ParameterList& paramList)
-  {
-    MergedAFact_ = Teuchos::rcp(new MergedBlockedMatrixFactory());
-    s_ = Teuchos::rcp(new DirectSolver(type, paramList));
-    type_ = "blocked direct solver (" + type + ")";
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    // Note that we have a nested smoother/solver object (of type DirectSolver), so we have to declare the dependencies by hand
-    // call DeclareInput by hand, since this->Input(currentLevel, "A") would not properly free A in the release mode (dependencies)
-    // We need the blocked version of A as input for the MergedAFact_
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
-
-    // syncronize input factory for "A" and nested representation for "A"
-    MergedAFact_->SetFactory("A", this->GetFactory("A"));
-
-    // declare input factories for nested direct solver
-    s_->SetFactory("A",MergedAFact_);
-    s_->DeclareInput(currentLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-
-    FactoryMonitor m(*this, "Setup BlockedDirectSolver", currentLevel);
-    if (this->IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::BlockedDirectSolver::Setup(): Setup() has already been called";
-
-    // extract blocked operator A from current level
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A"); // A needed for extracting map extractors
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
-                               "MueLu::BlockedDirectSolver::Build: input matrix A is not of type BlockedCrsMatrix.");
-
-    s_->Setup(currentLevel);
-
-    this->IsSetup(true);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(this->IsSetup() == false, Exceptions::RuntimeError,
-                               "MueLu::BlockedDirectSolver::Apply(): Setup() has not been called");
-
-    RCP<MultiVector> rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedDirectSolver(const std::string &type, const Teuchos::ParameterList &paramList) {
+  MergedAFact_ = Teuchos::rcp(new MergedBlockedMatrixFactory());
+  s_           = Teuchos::rcp(new DirectSolver(type, paramList));
+  type_        = "blocked direct solver (" + type + ")";
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  // Note that we have a nested smoother/solver object (of type DirectSolver), so we have to declare the dependencies by hand
+  // call DeclareInput by hand, since this->Input(currentLevel, "A") would not properly free A in the release mode (dependencies)
+  // We need the blocked version of A as input for the MergedAFact_
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
+
+  // syncronize input factory for "A" and nested representation for "A"
+  MergedAFact_->SetFactory("A", this->GetFactory("A"));
+
+  // declare input factories for nested direct solver
+  s_->SetFactory("A", MergedAFact_);
+  s_->DeclareInput(currentLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+
+  FactoryMonitor m(*this, "Setup BlockedDirectSolver", currentLevel);
+  if (this->IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::BlockedDirectSolver::Setup(): Setup() has already been called";
+
+  // extract blocked operator A from current level
+  A_                       = Factory::Get<RCP<Matrix> >(currentLevel, "A");  // A needed for extracting map extractors
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
+                             "MueLu::BlockedDirectSolver::Build: input matrix A is not of type BlockedCrsMatrix.");
+
+  s_->Setup(currentLevel);
+
+  this->IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(this->IsSetup() == false, Exceptions::RuntimeError,
+                             "MueLu::BlockedDirectSolver::Apply(): Setup() has not been called");
+
+  RCP<MultiVector> rcpX            = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB      = Teuchos::rcpFromRef(B);
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
 #ifdef HAVE_MUELU_DEBUG
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    if(bB.is_null() == false) {
-      //TEUCHOS_TEST_FOR_EXCEPTION(A_->getRangeMap()->isSameAs(*(B.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of RHS vector B is not the same as range map of the blocked operator A. Please check the map of B and A.");
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullRangeMap()->isSameAs(*(B.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of RHS vector B is not the same as range map of the blocked operator A. Please check the map of B and A.");
-    }
-    if(bX.is_null() == false) {
-      //TEUCHOS_TEST_FOR_EXCEPTION(A_->getDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
-    }
-#endif
-
-    if (bB.is_null() == true && bX.is_null() == true) {
-      // standard case (neither B nor X are blocked)
-      s_->Apply(X, B, InitialGuessIsZero);
-    } else if(bB.is_null() == false && bX.is_null() == false) {
-      // both B and X are blocked
-      RCP<MultiVector> mergedX = bX->Merge();
-      RCP<const MultiVector> mergedB = bB->Merge();
-      s_->Apply(*mergedX, *mergedB, InitialGuessIsZero);
-      RCP<MultiVector> xx = Teuchos::rcp(new BlockedMultiVector(bX->getBlockedMap(),mergedX));
-      SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-      X.update(one,*xx,zero);
-    } else if (bB.is_null() == true && bX.is_null() == false) {
-      // the solution vector is blocked
-      RCP<MultiVector> mergedX = bX->Merge();
-      s_->Apply(*mergedX, B, InitialGuessIsZero);
-      RCP<MultiVector> xx = Teuchos::rcp(new BlockedMultiVector(bX->getBlockedMap(),mergedX));
-      SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-      X.update(one,*xx,zero);
-    } else if (bB.is_null() == false && bX.is_null() == true) {
-      // only the RHS vector is blocked
-      RCP<const MultiVector> mergedB = bB->Merge();
-      s_->Apply(X, *mergedB, InitialGuessIsZero);
-    }
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  if (bB.is_null() == false) {
+    // TEUCHOS_TEST_FOR_EXCEPTION(A_->getRangeMap()->isSameAs(*(B.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of RHS vector B is not the same as range map of the blocked operator A. Please check the map of B and A.");
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullRangeMap()->isSameAs(*(B.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of RHS vector B is not the same as range map of the blocked operator A. Please check the map of B and A.");
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp (new BlockedDirectSolver (*this));
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
-
-    if (verbLevel & Debug)
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  if (bX.is_null() == false) {
+    // TEUCHOS_TEST_FOR_EXCEPTION(A_->getDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedDirectSolver::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
   }
+#endif
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (bB.is_null() == true && bX.is_null() == true) {
+    // standard case (neither B nor X are blocked)
+    s_->Apply(X, B, InitialGuessIsZero);
+  } else if (bB.is_null() == false && bX.is_null() == false) {
+    // both B and X are blocked
+    RCP<MultiVector> mergedX       = bX->Merge();
+    RCP<const MultiVector> mergedB = bB->Merge();
+    s_->Apply(*mergedX, *mergedB, InitialGuessIsZero);
+    RCP<MultiVector> xx = Teuchos::rcp(new BlockedMultiVector(bX->getBlockedMap(), mergedX));
+    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+    X.update(one, *xx, zero);
+  } else if (bB.is_null() == true && bX.is_null() == false) {
+    // the solution vector is blocked
+    RCP<MultiVector> mergedX = bX->Merge();
+    s_->Apply(*mergedX, B, InitialGuessIsZero);
+    RCP<MultiVector> xx = Teuchos::rcp(new BlockedMultiVector(bX->getBlockedMap(), mergedX));
+    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+    X.update(one, *xx, zero);
+  } else if (bB.is_null() == false && bX.is_null() == true) {
+    // only the RHS vector is blocked
+    RCP<const MultiVector> mergedB = bB->Merge();
+    s_->Apply(X, *mergedB, InitialGuessIsZero);
   }
-  
-
-} // namespace MueLu
-
-
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new BlockedDirectSolver(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
+
+  if (verbLevel & Debug)
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t BlockedDirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_BLOCKEDDIRECTSOLVER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_decl.hpp
index bac8d4a06c82..8d1790bb4273 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_decl.hpp
@@ -66,154 +66,149 @@
 
 namespace MueLu {
 
-  /*!
-    @class BlockedGaussSeidelSmoother
-    @brief block Gauss-Seidel method for blocked matrices
-
-    Implementation of a block Gauss-Seidel methods for blocked matrices
-    @param LocalOrdinal sweeps = 1: number of BGS sweeps
-    @param Scalar omega = 1.0: damping parameter
-    @param RCP<FactoryBase> AFact = Teuchos::null: factory for blocked "A" operator
-
-    Use the AddFactoryManager routine to declare the subsmoothers/subsolvers for the block Gauss-Seidel method for
-    the block rows. The corresponding factory manager has to provide a variable "A" (pointing to the subblock of the blocked A operator)
-    and a smoother object (variable: "PreSmoother").
-
-    Example
-    \code
-
-    // prototypes for direct solvers for blocks 1 and 2
-    RCP<SmootherPrototype> smoProto11 = rcp(new DirectSolver("", Teuchos::ParameterList(), A11Fact));
-    RCP<SmootherPrototype> smoProto22 = rcp(new DirectSolver("", Teuchos::ParameterList(), A22Fact));
-    RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
-    RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
-
-    // define factory manager objects for sublocks
-    RCP<FactoryManager> M11 = rcp(new FactoryManager());
-    M11->SetFactory("A", A11Fact);
-    M11->SetFactory("Smoother", Smoo11Fact);
-
-    RCP<FactoryManager> M22 = rcp(new FactoryManager());
-    M22->SetFactory("A", A22Fact);
-    M22->SetFactory("Smoother", Smoo22Fact);
-
-    // create blocked Gauss-Seidel smoother for 2x2 blocked matrix
-    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother(2,1.0));
-    smootherPrototype->AddFactoryManager(M11);
-    smootherPrototype->AddFactoryManager(M22);
-    RCP<SmootherFactory> smootherFact = rcp( new SmootherFactory(smootherPrototype) );
-
-    // use smootherFact in main-factory manager
-    \endcode
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BlockedGaussSeidelSmoother :
-    public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+/*!
+  @class BlockedGaussSeidelSmoother
+  @brief block Gauss-Seidel method for blocked matrices
+
+  Implementation of a block Gauss-Seidel methods for blocked matrices
+  @param LocalOrdinal sweeps = 1: number of BGS sweeps
+  @param Scalar omega = 1.0: damping parameter
+  @param RCP<FactoryBase> AFact = Teuchos::null: factory for blocked "A" operator
+
+  Use the AddFactoryManager routine to declare the subsmoothers/subsolvers for the block Gauss-Seidel method for
+  the block rows. The corresponding factory manager has to provide a variable "A" (pointing to the subblock of the blocked A operator)
+  and a smoother object (variable: "PreSmoother").
+
+  Example
+  \code
+
+  // prototypes for direct solvers for blocks 1 and 2
+  RCP<SmootherPrototype> smoProto11 = rcp(new DirectSolver("", Teuchos::ParameterList(), A11Fact));
+  RCP<SmootherPrototype> smoProto22 = rcp(new DirectSolver("", Teuchos::ParameterList(), A22Fact));
+  RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
+  RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
+
+  // define factory manager objects for sublocks
+  RCP<FactoryManager> M11 = rcp(new FactoryManager());
+  M11->SetFactory("A", A11Fact);
+  M11->SetFactory("Smoother", Smoo11Fact);
+
+  RCP<FactoryManager> M22 = rcp(new FactoryManager());
+  M22->SetFactory("A", A22Fact);
+  M22->SetFactory("Smoother", Smoo22Fact);
+
+  // create blocked Gauss-Seidel smoother for 2x2 blocked matrix
+  RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother(2,1.0));
+  smootherPrototype->AddFactoryManager(M11);
+  smootherPrototype->AddFactoryManager(M22);
+  RCP<SmootherFactory> smootherFact = rcp( new SmootherFactory(smootherPrototype) );
+
+  // use smootherFact in main-factory manager
+  \endcode
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BlockedGaussSeidelSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! @brief Constructor
-    */
-    BlockedGaussSeidelSmoother();
+  /*! @brief Constructor
+   */
+  BlockedGaussSeidelSmoother();
 
-    //! Destructor
-    virtual ~BlockedGaussSeidelSmoother();
-    //@}
+  //! Destructor
+  virtual ~BlockedGaussSeidelSmoother();
+  //@}
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //! Add a factory manager
-    //void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
+  //! Add a factory manager
+  // void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
 
-    //! Add a factory manager at a specific position
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
+  //! Add a factory manager at a specific position
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
 
+  //@}
 
-    //@}
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! @name Setup and Apply methods.
-    //@{
+  /*! @brief Setup routine
+   * In the Setup method the Inverse_ vector is filled with the corresponding
+   * SmootherBase objects. Without the Inverse_ vector being filled we cannot call
+   * BlockedGaussSeidelSmoother::Apply.
+   */
+  void Setup(Level &currentLevel);
 
-    /*! @brief Setup routine
-     * In the Setup method the Inverse_ vector is filled with the corresponding
-     * SmootherBase objects. Without the Inverse_ vector being filled we cannot call
-     * BlockedGaussSeidelSmoother::Apply.
-    */
-    void Setup(Level &currentLevel);
-
-    /*! @brief Apply the direct solver.
-    Solves the linear system <tt>AX=B</tt> using the constructed solver.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
-
-    RCP<SmootherPrototype> Copy() const;
+  /*! @brief Apply the direct solver.
+  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    /*! \brief Print the object with some verbosity level \c verbLevel to an FancyOStream object \c out
-     *
-     * - use MueLu::Describable::describe;
-     * - overloading, not hiding
-     */
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  /*! \brief Print the object with some verbosity level \c verbLevel to an FancyOStream object \c out
+   *
+   * - use MueLu::Describable::describe;
+   * - overloading, not hiding
+   */
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+  //@}
 
-    //! smoother type
-    std::string type_;
+ private:
+  //! smoother type
+  std::string type_;
 
-    //! vector of factory managers
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+  //! vector of factory managers
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-    //! vector of smoother/solver factories
-    std::vector<Teuchos::RCP<const SmootherBase> > Inverse_;
+  //! vector of smoother/solver factories
+  std::vector<Teuchos::RCP<const SmootherBase> > Inverse_;
 
-    //! vector storing whether sub-block is a blocked operator (needed for nested blocked smoothers using Thyra GIDs)
-    std::vector<bool> bIsBlockedOperator_;
+  //! vector storing whether sub-block is a blocked operator (needed for nested blocked smoothers using Thyra GIDs)
+  std::vector<bool> bIsBlockedOperator_;
 
-    //! A Factory
-    RCP<FactoryBase> AFact_;
+  //! A Factory
+  RCP<FactoryBase> AFact_;
 
-    //! internal blocked operator "A" generated by AFact_
-    RCP<Matrix> A_;
+  //! internal blocked operator "A" generated by AFact_
+  RCP<Matrix> A_;
 
-    //! range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> rangeMapExtractor_;
+  //! range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> rangeMapExtractor_;
 
-    //! domain map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> domainMapExtractor_;
-  }; // class Amesos2Smoother
+  //! domain map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> domainMapExtractor_;
+};  // class Amesos2Smoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_def.hpp
index c0b689f1d72e..157d7db7a944 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedGaussSeidelSmoother_def.hpp
@@ -67,111 +67,109 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedGaussSeidelSmoother()
-    : type_("blocked GaussSeidel"), A_(Teuchos::null)
-  {
-    FactManager_.reserve(10); // TODO fix me!
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~BlockedGaussSeidelSmoother() {}
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< Scalar >                ("Damping factor",     1.0, "Damping/Scaling factor in BGS");
-    validParamList->set< LocalOrdinal >          ("Sweeps",             1, "Number of BGS sweeps (default = 1)");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedGaussSeidelSmoother()
+  : type_("blocked GaussSeidel")
+  , A_(Teuchos::null) {
+  FactManager_.reserve(10);  // TODO fix me!
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~BlockedGaussSeidelSmoother() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in BGS");
+  validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of BGS sweeps (default = 1)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+
+  size_t myPos = Teuchos::as<size_t>(pos);
+
+  if (myPos < FactManager_.size()) {
+    // replace existing entries in FactManager_ vector
+    FactManager_.at(myPos) = FactManager;
+  } else if (myPos == FactManager_.size()) {
+    // append new Factory manager at the end of the vector
+    FactManager_.push_back(FactManager);
+  } else {  // if(myPos > FactManager_.size())
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
 
-    return validParamList;
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  // this->Input(currentLevel, "A");
+  //  TODO: check me: why is this->Input not freeing properly A in release mode?
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
 
-    size_t myPos = Teuchos::as<size_t>(pos);
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
-    if (myPos < FactManager_.size()) {
-      // replace existing entries in FactManager_ vector
-      FactManager_.at(myPos) = FactManager;
-    } else if(myPos == FactManager_.size()) {
-      // append new Factory manager at the end of the vector
-      FactManager_.push_back(FactManager);
-    } else { // if(myPos > FactManager_.size())
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", (*it)->GetFactory("Smoother").get());
 
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    }
+    // request "A" for current subblock row (only needed for Thyra mode)
+    currentLevel.DeclareInput("A", (*it)->GetFactory("A").get());
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    //this->Input(currentLevel, "A");
-    // TODO: check me: why is this->Input not freeing properly A in release mode?
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel),   *it);
+  FactoryMonitor m(*this, "Setup blocked Gauss-Seidel Smoother", currentLevel);
+  if (SmootherPrototype::IsSetup() == true) this->GetOStream(Warnings0) << "MueLu::BlockedGaussSeidelSmoother::Setup(): Setup() has already been called";
 
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother",(*it)->GetFactory("Smoother").get());
+  // extract blocked operator A from current level
+  A_                       = Factory::Get<RCP<Matrix> >(currentLevel, "A");  // A needed for extracting map extractors
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
 
-      // request "A" for current subblock row (only needed for Thyra mode)
-      currentLevel.DeclareInput("A",(*it)->GetFactory("A").get());
-    }
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  // plausibility check
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Setup: number of block rows of A is " << bA->Rows() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Cols() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Setup: number of block cols of A is " << bA->Cols() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
 
-    FactoryMonitor m(*this, "Setup blocked Gauss-Seidel Smoother", currentLevel);
-    if (SmootherPrototype::IsSetup() == true) this->GetOStream(Warnings0) << "MueLu::BlockedGaussSeidelSmoother::Setup(): Setup() has already been called";
+  // store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
 
-    // extract blocked operator A from current level
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A"); // A needed for extracting map extractors
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA==Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
-    // plausibility check
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Setup: number of block rows of A is " << bA->Rows() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Cols() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Setup: number of block cols of A is " << bA->Cols() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
+    // extract Smoother for current block row (BGS ordering)
+    RCP<const SmootherBase> Smoo = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", (*it)->GetFactory("Smoother").get());
+    Inverse_.push_back(Smoo);
 
-    // store map extractors
-    rangeMapExtractor_  = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
-
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), *it);
-
-      // extract Smoother for current block row (BGS ordering)
-      RCP<const SmootherBase> Smoo = currentLevel.Get< RCP<SmootherBase> >("PreSmoother",(*it)->GetFactory("Smoother").get());
-      Inverse_.push_back(Smoo);
-
-      // store whether subblock matrix is blocked or not!
-      RCP<Matrix> Aii = currentLevel.Get< RCP<Matrix> >("A",(*it)->GetFactory("A").get());
-      bIsBlockedOperator_.push_back(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Aii)!=Teuchos::null);
-    }
-
-    SmootherPrototype::IsSetup(true);
+    // store whether subblock matrix is blocked or not!
+    RCP<Matrix> Aii = currentLevel.Get<RCP<Matrix> >("A", (*it)->GetFactory("A").get());
+    bIsBlockedOperator_.push_back(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Aii) != Teuchos::null);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector& B, bool InitialGuessIsZero) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): Setup() has not been called");
+  SmootherPrototype::IsSetup(true);
+}
 
-#if 0 // def HAVE_MUELU_DEBUG
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): Setup() has not been called");
+
+#if 0  // def HAVE_MUELU_DEBUG
     // TODO simplify this debug check
     RCP<MultiVector> rcpDebugX = Teuchos::rcpFromRef(X);
     RCP<const MultiVector> rcpDebugB = Teuchos::rcpFromRef(B);
@@ -193,167 +191,161 @@ namespace MueLu {
       //TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
     }
 #endif
-    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-
-    // Input variables used for the rest of the algorithm
-    RCP<MultiVector> rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
-
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
-    }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
-
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+  SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+
+  // Input variables used for the rest of the algorithm
+  RCP<MultiVector> rcpX       = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
+
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(bA);
+
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
     }
 
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
-
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
-
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
-
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
     }
-
-    // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
-
-    RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-    RCP<MultiVector> tempres = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
-
-
-    // Clear solution from previos V cycles in case it is still stored
-    if( InitialGuessIsZero==true )
-      rcpX->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
-
-
-    // outer Richardson loop
-    for (LocalOrdinal run = 0; run < nSweeps; ++run) {
-      // one BGS sweep
-      // loop over all block rows
-      for(size_t i = 0; i<Inverse_.size(); i++) {
-
-        // calculate block residual r = B-A*X
-        // note: A_ is the full blocked operator
-        residual->update(1.0,*rcpB,0.0); // r = B
-        if(InitialGuessIsZero == false || i > 0 || run > 0)
-          bA->bgs_apply(*rcpX, *residual, i, Teuchos::NO_TRANS, -1.0, 1.0);
-
-        // extract corresponding subvectors from X and residual
-        bool bRangeThyraMode =  rangeMapExtractor_->getThyraMode();
-        bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
-        Teuchos::RCP<MultiVector> Xi = domainMapExtractor_->ExtractVector(rcpX, i, bDomainThyraMode);
-        Teuchos::RCP<MultiVector> ri = rangeMapExtractor_->ExtractVector(residual, i, bRangeThyraMode);
-        Teuchos::RCP<MultiVector> tXi = domainMapExtractor_->getVector(i, X.getNumVectors(), bDomainThyraMode);
-
-        // apply solver/smoother
-        Inverse_.at(i)->Apply(*tXi, *ri, false);
-
-        // update vector
-        Xi->update(omega,*tXi,1.0);  // X_{i+1} = X_i + omega \Delta X_i
-
-      }
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
     }
 
-    if (bCopyResultX == true) {
-      RCP<MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
     }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp( new BlockedGaussSeidelSmoother(*this) );
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
-  }
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
+    }
 
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << " Sweeps: " << nSweeps << " damping: " << omega << std::endl;
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
     }
+  }
 
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
+
+  RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  RCP<MultiVector> tempres  = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
+
+  // Clear solution from previos V cycles in case it is still stored
+  if (InitialGuessIsZero == true)
+    rcpX->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
+
+  // outer Richardson loop
+  for (LocalOrdinal run = 0; run < nSweeps; ++run) {
+    // one BGS sweep
+    // loop over all block rows
+    for (size_t i = 0; i < Inverse_.size(); i++) {
+      // calculate block residual r = B-A*X
+      // note: A_ is the full blocked operator
+      residual->update(1.0, *rcpB, 0.0);  // r = B
+      if (InitialGuessIsZero == false || i > 0 || run > 0)
+        bA->bgs_apply(*rcpX, *residual, i, Teuchos::NO_TRANS, -1.0, 1.0);
+
+      // extract corresponding subvectors from X and residual
+      bool bRangeThyraMode          = rangeMapExtractor_->getThyraMode();
+      bool bDomainThyraMode         = domainMapExtractor_->getThyraMode();
+      Teuchos::RCP<MultiVector> Xi  = domainMapExtractor_->ExtractVector(rcpX, i, bDomainThyraMode);
+      Teuchos::RCP<MultiVector> ri  = rangeMapExtractor_->ExtractVector(residual, i, bRangeThyraMode);
+      Teuchos::RCP<MultiVector> tXi = domainMapExtractor_->getVector(i, X.getNumVectors(), bDomainThyraMode);
+
+      // apply solver/smoother
+      Inverse_.at(i)->Apply(*tXi, *ri, false);
+
+      // update vector
+      Xi->update(omega, *tXi, 1.0);  // X_{i+1} = X_i + omega \Delta X_i
     }
   }
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+
+  if (bCopyResultX == true) {
+    RCP<MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new BlockedGaussSeidelSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << " Sweeps: " << nSweeps << " damping: " << omega << std::endl;
   }
 
-} // namespace MueLu
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  }
+}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t BlockedGaussSeidelSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_decl.hpp
index 9bbd1775a969..9b7a87381d09 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_decl.hpp
@@ -66,156 +66,151 @@
 
 namespace MueLu {
 
-  /*!
-    @class BlockedJacobiSmoother
-    @brief block Jacobi method for blocked matrices
-
-    Implementation of a block Jacobi methods for blocked matrices
-    @param LocalOrdinal sweeps = 1: number of sweeps
-    @param Scalar omega = 1.0: damping parameter
-    @param RCP<FactoryBase> AFact = Teuchos::null: factory for blocked "A" operator
-
-    Use the AddFactoryManager routine to declare the subsmoothers/subsolvers for the block Jacobi method for
-    the block rows. The corresponding factory manager has to provide a variable "A" (pointing to the subblock of the blocked A operator)
-    and a smoother object (variable: "PreSmoother").
-
-    Example
-    \code
-
-    // prototypes for direct solvers for blocks 1 and 2
-    RCP<SmootherPrototype> smoProto11 = rcp( new DirectSolver("", Teuchos::ParameterList(), A11Fact) );
-    RCP<SmootherPrototype> smoProto22 = rcp( new DirectSolver("", Teuchos::ParameterList(), A22Fact) );
-    RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
-    RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
-
-    // define factory manager objects for sublocks
-    RCP<FactoryManager> M11 = rcp(new FactoryManager());
-    M11->SetFactory("A", A11Fact);
-    M11->SetFactory("Smoother", Smoo11Fact);
-
-    RCP<FactoryManager> M22 = rcp(new FactoryManager());
-    M22->SetFactory("A", A22Fact);
-    M22->SetFactory("Smoother", Smoo22Fact);
-
-    // create blocked Jacobi smoother for 2x2 blocked matrix
-    RCP<BlockedJacobiSmoother> smootherPrototype = rcp(new BlockedJacobiSmoother(2,1.0));
-    smootherPrototype->AddFactoryManager(M11);
-    smootherPrototype->AddFactoryManager(M22);
-    RCP<SmootherFactory> smootherFact = rcp( new SmootherFactory(smootherPrototype) );
-
-    // use smootherFact in main-factory manager
-    \endcode
-
-    \author mayr.mt \date 06/2017
-  */
-  // ToDo (mayr) Inherit from common base class with BlockedGaussSeidelSmoother
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BlockedJacobiSmoother :
-    public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+/*!
+  @class BlockedJacobiSmoother
+  @brief block Jacobi method for blocked matrices
+
+  Implementation of a block Jacobi methods for blocked matrices
+  @param LocalOrdinal sweeps = 1: number of sweeps
+  @param Scalar omega = 1.0: damping parameter
+  @param RCP<FactoryBase> AFact = Teuchos::null: factory for blocked "A" operator
+
+  Use the AddFactoryManager routine to declare the subsmoothers/subsolvers for the block Jacobi method for
+  the block rows. The corresponding factory manager has to provide a variable "A" (pointing to the subblock of the blocked A operator)
+  and a smoother object (variable: "PreSmoother").
+
+  Example
+  \code
+
+  // prototypes for direct solvers for blocks 1 and 2
+  RCP<SmootherPrototype> smoProto11 = rcp( new DirectSolver("", Teuchos::ParameterList(), A11Fact) );
+  RCP<SmootherPrototype> smoProto22 = rcp( new DirectSolver("", Teuchos::ParameterList(), A22Fact) );
+  RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
+  RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
+
+  // define factory manager objects for sublocks
+  RCP<FactoryManager> M11 = rcp(new FactoryManager());
+  M11->SetFactory("A", A11Fact);
+  M11->SetFactory("Smoother", Smoo11Fact);
+
+  RCP<FactoryManager> M22 = rcp(new FactoryManager());
+  M22->SetFactory("A", A22Fact);
+  M22->SetFactory("Smoother", Smoo22Fact);
+
+  // create blocked Jacobi smoother for 2x2 blocked matrix
+  RCP<BlockedJacobiSmoother> smootherPrototype = rcp(new BlockedJacobiSmoother(2,1.0));
+  smootherPrototype->AddFactoryManager(M11);
+  smootherPrototype->AddFactoryManager(M22);
+  RCP<SmootherFactory> smootherFact = rcp( new SmootherFactory(smootherPrototype) );
+
+  // use smootherFact in main-factory manager
+  \endcode
+
+  \author mayr.mt \date 06/2017
+*/
+// ToDo (mayr) Inherit from common base class with BlockedGaussSeidelSmoother
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BlockedJacobiSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_BLOCKEDJACOBISMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! @name Constructors / destructors
-    //@{
+  /*! @brief Constructor
+   */
+  BlockedJacobiSmoother();
 
-    /*! @brief Constructor
-    */
-    BlockedJacobiSmoother();
+  //! Destructor
+  virtual ~BlockedJacobiSmoother();
+  //@}
 
-    //! Destructor
-    virtual ~BlockedJacobiSmoother();
-    //@}
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  void DeclareInput(Level &currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Add a factory manager
+  // void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
 
-    //! Add a factory manager
-    //void AddFactoryManager(RCP<const FactoryManagerBase> FactManager);
+  //! Add a factory manager at a specific position
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
 
-    //! Add a factory manager at a specific position
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
+  //@}
 
+  //! @name Setup and Apply methods.
+  //@{
 
-    //@}
+  /*! @brief Setup routine
+   * In the Setup method the Inverse_ vector is filled with the corresponding
+   * SmootherBase objects. Without the Inverse_ vector being filled we cannot call
+   * BlockedJacobiSmoother::Apply.
+   */
+  void Setup(Level &currentLevel);
 
-    //! @name Setup and Apply methods.
-    //@{
+  /*! @brief Apply the direct solver.
+   *
+   *  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+   *  @param X initial guess
+   *  @param B right-hand side
+   *  @param InitialGuessIsZero This option has no effect.
+   */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    /*! @brief Setup routine
-     * In the Setup method the Inverse_ vector is filled with the corresponding
-     * SmootherBase objects. Without the Inverse_ vector being filled we cannot call
-     * BlockedJacobiSmoother::Apply.
-    */
-    void Setup(Level &currentLevel);
+  RCP<SmootherPrototype> Copy() const;
 
-    /*! @brief Apply the direct solver.
-     *
-     *  Solves the linear system <tt>AX=B</tt> using the constructed solver.
-     *  @param X initial guess
-     *  @param B right-hand side
-     *  @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    RCP<SmootherPrototype> Copy() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  /*! \brief Print the object with some verbosity level to an FancyOStream object.
+   *
+   *  Using MueLu::Describable::describe; // overloading, not hiding
+   */
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    /*! \brief Print the object with some verbosity level to an FancyOStream object.
-     *
-     *  Using MueLu::Describable::describe; // overloading, not hiding
-     */
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //@}
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+ private:
+  //! smoother type
+  std::string type_;
 
-    //@}
+  //! vector of factory managers
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-  private:
+  //! vector of smoother/solver factories
+  std::vector<Teuchos::RCP<const SmootherBase> > Inverse_;
 
-    //! smoother type
-    std::string type_;
+  //! vector storing whether sub-block is a blocked operator (needed for nested blocked smoothers using Thyra GIDs)
+  std::vector<bool> bIsBlockedOperator_;
 
-    //! vector of factory managers
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+  //! A Factory
+  RCP<FactoryBase> AFact_;
 
-    //! vector of smoother/solver factories
-    std::vector<Teuchos::RCP<const SmootherBase> > Inverse_;
+  //! internal blocked operator "A" generated by AFact_
+  RCP<Matrix> A_;
 
-    //! vector storing whether sub-block is a blocked operator (needed for nested blocked smoothers using Thyra GIDs)
-    std::vector<bool> bIsBlockedOperator_;
+  //!< range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> rangeMapExtractor_;
 
-    //! A Factory
-    RCP<FactoryBase> AFact_;
+  //!< domain map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> domainMapExtractor_;
+};  // class BlockedJacobiSmoother
 
-    //! internal blocked operator "A" generated by AFact_
-    RCP<Matrix> A_;
-
-    //!< range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> rangeMapExtractor_;
-
-    //!< domain map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> domainMapExtractor_;
-  }; // class BlockedJacobiSmoother
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDJACOBISMOOTHER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_def.hpp
index e1e1a908987e..7abb9f70a504 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BlockedJacobiSmoother_def.hpp
@@ -67,111 +67,109 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedJacobiSmoother()
-    : type_("blocked Jacobi"), A_(Teuchos::null)
-  {
-    FactManager_.reserve(10); // TODO fix me!
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~BlockedJacobiSmoother() {}
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in Jacobi");
-    validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of sweeps (default = 1)");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BlockedJacobiSmoother()
+  : type_("blocked Jacobi")
+  , A_(Teuchos::null) {
+  FactManager_.reserve(10);  // TODO fix me!
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~BlockedJacobiSmoother() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in Jacobi");
+  validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of sweeps (default = 1)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+
+  size_t myPos = Teuchos::as<size_t>(pos);
+
+  if (myPos < FactManager_.size()) {
+    // replace existing entris in FactManager_ vector
+    FactManager_.at(myPos) = FactManager;
+  } else if (myPos == FactManager_.size()) {
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
+  } else {  // if(myPos > FactManager_.size())
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
 
-    return validParamList;
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  // this->Input(currentLevel, "A");
+  //  TODO: check me: why is this->Input not freeing properly A in release mode?
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
 
-    size_t myPos = Teuchos::as<size_t>(pos);
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
-    if (myPos < FactManager_.size()) {
-      // replace existing entris in FactManager_ vector
-      FactManager_.at(myPos) = FactManager;
-    } else if( myPos == FactManager_.size()) {
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    } else { // if(myPos > FactManager_.size())
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", (*it)->GetFactory("Smoother").get());
 
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    }
+    // request "A" for current subblock row (only needed for Thyra mode)
+    currentLevel.DeclareInput("A", (*it)->GetFactory("A").get());
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    //this->Input(currentLevel, "A");
-    // TODO: check me: why is this->Input not freeing properly A in release mode?
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel),   *it);
+  FactoryMonitor m(*this, "Setup blocked Jacobi Smoother", currentLevel);
+  if (SmootherPrototype::IsSetup() == true) this->GetOStream(Warnings0) << "MueLu::BlockedJacobiSmoother::Setup(): Setup() has already been called";
 
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother",(*it)->GetFactory("Smoother").get());
+  // extract blocked operator A from current level
+  A_                       = Factory::Get<RCP<Matrix> >(currentLevel, "A");  // A needed for extracting map extractors
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
 
-      // request "A" for current subblock row (only needed for Thyra mode)
-      currentLevel.DeclareInput("A",(*it)->GetFactory("A").get());
-    }
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  // plausibility check
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Setup: number of block rows of A is " << bA->Rows() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
+  TEUCHOS_TEST_FOR_EXCEPTION(bA->Cols() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Setup: number of block cols of A is " << bA->Cols() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
 
-    FactoryMonitor m(*this, "Setup blocked Jacobi Smoother", currentLevel);
-    if (SmootherPrototype::IsSetup() == true) this->GetOStream(Warnings0) << "MueLu::BlockedJacobiSmoother::Setup(): Setup() has already been called";
+  // store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
 
-    // extract blocked operator A from current level
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A"); // A needed for extracting map extractors
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA==Teuchos::null, Exceptions::BadCast, "MueLu::BlockedPFactory::Build: input matrix A is not of type BlockedCrsMatrix! error.");
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
-    // plausibility check
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Rows() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Setup: number of block rows of A is " << bA->Rows() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(bA->Cols() != FactManager_.size(), Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Setup: number of block cols of A is " << bA->Cols() << " and does not match number of SubFactoryManagers " << FactManager_.size() << ". error.");
+    // extract Smoother for current block row
+    RCP<const SmootherBase> Smoo = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", (*it)->GetFactory("Smoother").get());
+    Inverse_.push_back(Smoo);
 
-    // store map extractors
-    rangeMapExtractor_ = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
-
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), *it);
-
-      // extract Smoother for current block row
-      RCP<const SmootherBase> Smoo = currentLevel.Get< RCP<SmootherBase> >("PreSmoother",(*it)->GetFactory("Smoother").get());
-      Inverse_.push_back(Smoo);
-
-      // store whether subblock matrix is blocked or not!
-      RCP<Matrix> Aii = currentLevel.Get< RCP<Matrix> >("A",(*it)->GetFactory("A").get());
-      bIsBlockedOperator_.push_back(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Aii)!=Teuchos::null);
-    }
-
-    SmootherPrototype::IsSetup(true);
+    // store whether subblock matrix is blocked or not!
+    RCP<Matrix> Aii = currentLevel.Get<RCP<Matrix> >("A", (*it)->GetFactory("A").get());
+    bIsBlockedOperator_.push_back(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(Aii) != Teuchos::null);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector& B, bool InitialGuessIsZero) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Apply(): Setup() has not been called");
+  SmootherPrototype::IsSetup(true);
+}
 
-#if 0 // def HAVE_MUELU_DEBUG
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Apply(): Setup() has not been called");
+
+#if 0  // def HAVE_MUELU_DEBUG
     // TODO simplify this debug check
     RCP<MultiVector> rcpDebugX = Teuchos::rcpFromRef(X);
     RCP<const MultiVector> rcpDebugB = Teuchos::rcpFromRef(B);
@@ -193,165 +191,160 @@ namespace MueLu {
       //TEUCHOS_TEST_FOR_EXCEPTION(bA->getFullDomainMap()->isSameAs(*(X.getMap())) == false, Exceptions::RuntimeError, "MueLu::BlockedJacobiSmoother::Apply(): The map of the solution vector X is not the same as domain map of the blocked operator A. Please check the map of X and A.");
     }
 
-
 #endif
-    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-
-    // Input variables used for the rest of the algorithm
-    RCP<MultiVector> rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
-
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
+  SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+
+  // Input variables used for the rest of the algorithm
+  RCP<MultiVector> rcpX       = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
+
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(bA);
+
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
     }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
 
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
+    }
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
     }
 
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+    }
+  }
 
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
-
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
     }
 
-    // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
+    }
+  }
 
-    RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-    RCP<MultiVector> tempres = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
 
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
+  RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  RCP<MultiVector> tempres  = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
 
-    // outer Richardson loop
-    for (LocalOrdinal run = 0; run < nSweeps; ++run) {
-      residual->update(1.0,*rcpB,0.0); // r = B
-      if(InitialGuessIsZero == false || run > 0){
-        bA->apply(*rcpX, *residual, Teuchos::NO_TRANS, -1.0, 1.0);
-      }
-      // one sweep of Jacobi: loop over all block rows
-      for(size_t i = 0; i<Inverse_.size(); i++) {
-
-        // extract corresponding subvectors from X and residual
-        bool bRangeThyraMode =  rangeMapExtractor_->getThyraMode();
-        bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
-        Teuchos::RCP<MultiVector> Xi = domainMapExtractor_->ExtractVector(rcpX, i, bDomainThyraMode);
-        Teuchos::RCP<MultiVector> ri = rangeMapExtractor_->ExtractVector(residual, i, bRangeThyraMode);
-        Teuchos::RCP<MultiVector> tXi = domainMapExtractor_->getVector(i, X.getNumVectors(), bDomainThyraMode);
-
-        // apply solver/smoother
-        Inverse_.at(i)->Apply(*tXi, *ri, false);
-
-        // update vector
-        if( InitialGuessIsZero && run == 0 ){
-          Xi->update(omega,*tXi,0.0); // X_{i+1} = X_i + omega \Delta X_i
-        } else {
-          Xi->update(omega,*tXi,1.0); // X_{i+1} = X_i + omega \Delta X_i
-        }
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
 
-      }
+  // outer Richardson loop
+  for (LocalOrdinal run = 0; run < nSweeps; ++run) {
+    residual->update(1.0, *rcpB, 0.0);  // r = B
+    if (InitialGuessIsZero == false || run > 0) {
+      bA->apply(*rcpX, *residual, Teuchos::NO_TRANS, -1.0, 1.0);
     }
-
-    if (bCopyResultX == true) {
-      RCP<MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
+    // one sweep of Jacobi: loop over all block rows
+    for (size_t i = 0; i < Inverse_.size(); i++) {
+      // extract corresponding subvectors from X and residual
+      bool bRangeThyraMode          = rangeMapExtractor_->getThyraMode();
+      bool bDomainThyraMode         = domainMapExtractor_->getThyraMode();
+      Teuchos::RCP<MultiVector> Xi  = domainMapExtractor_->ExtractVector(rcpX, i, bDomainThyraMode);
+      Teuchos::RCP<MultiVector> ri  = rangeMapExtractor_->ExtractVector(residual, i, bRangeThyraMode);
+      Teuchos::RCP<MultiVector> tXi = domainMapExtractor_->getVector(i, X.getNumVectors(), bDomainThyraMode);
+
+      // apply solver/smoother
+      Inverse_.at(i)->Apply(*tXi, *ri, false);
+
+      // update vector
+      if (InitialGuessIsZero && run == 0) {
+        Xi->update(omega, *tXi, 0.0);  // X_{i+1} = X_i + omega \Delta X_i
+      } else {
+        Xi->update(omega, *tXi, 1.0);  // X_{i+1} = X_i + omega \Delta X_i
+      }
     }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp( new BlockedJacobiSmoother(*this) );
+  if (bCopyResultX == true) {
+    RCP<MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new BlockedJacobiSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << " Sweeps: " << nSweeps << " damping: " << omega << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
-
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << " Sweeps: " << nSweeps << " damping: " << omega << std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t BlockedJacobiSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_BLOCKEDJACOBISMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_decl.hpp
index 59e378d56eda..0286c6039629 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_decl.hpp
@@ -50,12 +50,12 @@
 
 #include <Teuchos_ParameterList.hpp>
 
-//Xpetra
+// Xpetra
 #include <Xpetra_MapExtractor_fwd.hpp>
 #include <Xpetra_MultiVectorFactory_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 
-//MueLu
+// MueLu
 #include "MueLu_BraessSarazinSmoother_fwd.hpp"
 #include "MueLu_SmootherPrototype.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
@@ -67,103 +67,100 @@
 
 namespace MueLu {
 
-  /*!
-    @class BraessSarazinSmoother
-    @brief BraessSarazin smoother for 2x2 block matrices
+/*!
+  @class BraessSarazinSmoother
+  @brief BraessSarazin smoother for 2x2 block matrices
 
-  */
+*/
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BraessSarazinSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BraessSarazinSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_BRAESSSARAZINSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! Input
-    //@{
-
-    RCP<const ParameterList> GetValidParameterList() const;
+ public:
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Add a factory manager for BraessSarazin internal SchurComplement handling
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos = 0);
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //! Add a factory manager for BraessSarazin internal SchurComplement handling
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos = 0);
 
-    //! @name Setup and Apply methods.
-    //@{
+  //@}
 
-    /*! @brief Setup routine
+  //! @name Setup and Apply methods.
+  //@{
 
-    Setup routine can be summarized in 4 steps:
-    - set the map extractors
-    - set the blocks
-    - create and set the inverse of the diagonal of F
-    - set the smoother for the Schur Complement
-    */
-    void Setup(Level &currentLevel);
+  /*! @brief Setup routine
 
-    /*! @brief Apply the Braess Sarazin smoother.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero TODO This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
+  Setup routine can be summarized in 4 steps:
+  - set the map extractors
+  - set the blocks
+  - create and set the inverse of the diagonal of F
+  - set the smoother for the Schur Complement
+  */
+  void Setup(Level &currentLevel);
 
-    RCP<SmootherPrototype> Copy() const;
+  /*! @brief Apply the Braess Sarazin smoother.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero TODO This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    /*!
-    @brief Print the object with some verbosity level to an FancyOStream object
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    Using MueLu::Describable::describe; overloading, not hiding
-    */
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  /*!
+  @brief Print the object with some verbosity level to an FancyOStream object
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  Using MueLu::Describable::describe; overloading, not hiding
+  */
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+  //@}
 
-    //! smoother type
-    std::string type_ = "Braess Sarazin";
+ private:
+  //! smoother type
+  std::string type_ = "Braess Sarazin";
 
-    RCP<const FactoryBase>              AFact_;               //!< A Factory
-    RCP<const FactoryManagerBase>       FactManager_;         //!< Factory manager for creating the Schur Complement
+  RCP<const FactoryBase> AFact_;               //!< A Factory
+  RCP<const FactoryManagerBase> FactManager_;  //!< Factory manager for creating the Schur Complement
 
-    //! block operator
-    RCP<Matrix>                         A_ = Teuchos::null;   // < ! internal blocked operator "A" generated by AFact_
+  //! block operator
+  RCP<Matrix> A_ = Teuchos::null;  // < ! internal blocked operator "A" generated by AFact_
 
-    RCP<const MapExtractor>             rangeMapExtractor_;   //!< range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractor>             domainMapExtractor_;  //!< domain map extractor (from A_ generated by AFact)
+  RCP<const MapExtractor> rangeMapExtractor_;   //!< range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractor> domainMapExtractor_;  //!< domain map extractor (from A_ generated by AFact)
 
-    //! matrices
-    RCP<Matrix>                         A00_;                 //!< Block (0,0) [typically, fluid operator]
-    RCP<Matrix>                         A01_;                 //!< Block (0,1) [typically, pressure gradient operator]
-    RCP<Matrix>                         A10_;                 //!< Block (1,0) [typically, divergence operator]
-    RCP<Matrix>                         A11_;                 //!< Block (1,1) [typically, pressure stabilization term or null block]
-    RCP<Matrix>                         S_;                   //!< Schur complement
-    RCP<Vector>                         D_;                   //!< Inverse to approximation to block (0,0). Here, D_ = omega*inv(diag(A(0,0)))
+  //! matrices
+  RCP<Matrix> A00_;  //!< Block (0,0) [typically, fluid operator]
+  RCP<Matrix> A01_;  //!< Block (0,1) [typically, pressure gradient operator]
+  RCP<Matrix> A10_;  //!< Block (1,0) [typically, divergence operator]
+  RCP<Matrix> A11_;  //!< Block (1,1) [typically, pressure stabilization term or null block]
+  RCP<Matrix> S_;    //!< Schur complement
+  RCP<Vector> D_;    //!< Inverse to approximation to block (0,0). Here, D_ = omega*inv(diag(A(0,0)))
 
-    Teuchos::RCP<SmootherBase>          smoo_;                //!< Smoother for SchurComplement equation
+  Teuchos::RCP<SmootherBase> smoo_;  //!< Smoother for SchurComplement equation
 
-  }; // class Amesos2Smoother
+};  // class Amesos2Smoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BRAESSSARAZINSMOOTHER_SHORT
 #endif /* MUELU_BRAESSSARAZINSMOOTHER_DECL_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_def.hpp
index 4e8b2c8ab29d..181c276e6b0f 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_BraessSarazinSmoother_def.hpp
@@ -69,308 +69,304 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos != 0, Exceptions::RuntimeError, "MueLu::BraessSarazinSmoother::AddFactoryManager: parameter \'pos\' must be zero! error.");
-    FactManager_ = FactManager;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos != 0, Exceptions::RuntimeError, "MueLu::BraessSarazinSmoother::AddFactoryManager: parameter \'pos\' must be zero! error.");
+  FactManager_ = FactManager;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    SC one = Teuchos::ScalarTraits<SC>::one();
+  SC one = Teuchos::ScalarTraits<SC>::one();
 
-    validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
 
-    validParamList->set<bool>("lumping", false, "Use lumping to construct diag(A(0,0)), "
-        "i.e. use row sum of the abs values on the diagonal as approximation of A00 (and A00^{-1})");
-    validParamList->set<SC>("Damping factor", one, "Damping/Scaling factor in BraessSarazin (usually has to be chosen > 1, default = 1.0)");
-    validParamList->set<LO>("Sweeps", 1, "Number of BraessSarazin sweeps (default = 1)");
-    validParamList->set<bool>("q2q1 mode", false, "Run in the mode matching Q2Q1 matlab code");
+  validParamList->set<bool>("lumping", false,
+                            "Use lumping to construct diag(A(0,0)), "
+                            "i.e. use row sum of the abs values on the diagonal as approximation of A00 (and A00^{-1})");
+  validParamList->set<SC>("Damping factor", one, "Damping/Scaling factor in BraessSarazin (usually has to be chosen > 1, default = 1.0)");
+  validParamList->set<LO>("Sweeps", 1, "Number of BraessSarazin sweeps (default = 1)");
+  validParamList->set<bool>("q2q1 mode", false, "Run in the mode matching Q2Q1 matlab code");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    this->Input(currentLevel, "A");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.is_null(), Exceptions::RuntimeError,
-                               "MueLu::BraessSarazinSmoother::DeclareInput: FactManager_ must not be null! "
-                               "Introduce a FactoryManager for the SchurComplement equation.");
+  TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.is_null(), Exceptions::RuntimeError,
+                             "MueLu::BraessSarazinSmoother::DeclareInput: FactManager_ must not be null! "
+                             "Introduce a FactoryManager for the SchurComplement equation.");
 
-    // carefully call DeclareInput after switching to internal FactoryManager
-    {
-      SetFactoryManager currentSFM(rcpFromRef(currentLevel), FactManager_);
+  // carefully call DeclareInput after switching to internal FactoryManager
+  {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), FactManager_);
 
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother", FactManager_->GetFactory("Smoother").get());
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", FactManager_->GetFactory("Smoother").get());
 
-      // request Schur matrix just in case
-      currentLevel.DeclareInput("A", FactManager_->GetFactory("A").get());
-    }
+    // request Schur matrix just in case
+    currentLevel.DeclareInput("A", FactManager_->GetFactory("A").get());
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::BreaessSarazinSmoother::Setup(): Setup() has already been called";
-
-    // Extract blocked operator A from current level
-    A_ = Factory::Get<RCP<Matrix> > (currentLevel, "A");
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
-                               "MueLu::BraessSarazinSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
-
-    // Store map extractors
-    rangeMapExtractor_  = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
-
-    // Store the blocks in local member variables
-    A00_ = bA->getMatrix(0,0);
-    A01_ = bA->getMatrix(0,1);
-    A10_ = bA->getMatrix(1,0);
-    A11_ = bA->getMatrix(1,1);
-
-    const ParameterList& pL = Factory::GetParameterList();
-    SC omega = pL.get<SC>("Damping factor");
-
-    // Create the inverse of the diagonal of F
-    // TODO add safety check for zeros on diagonal of F!
-    RCP<Vector> diagFVector = VectorFactory::Build(A00_->getRowMap());
-    if (pL.get<bool>("lumping") == false) {
-      A00_->getLocalDiagCopy(*diagFVector);       // extract diagonal of F
-    } else {
-      diagFVector = Utilities::GetLumpedMatrixDiagonal(*A00_);
-    }
-    diagFVector->scale(omega);
-    D_ = Utilities::GetInverse(diagFVector);
-
-    // check whether D_ is a blocked vector with only 1 block
-    RCP<BlockedVector> bD = Teuchos::rcp_dynamic_cast<BlockedVector>(D_);
-    if(bD.is_null() == false && bD->getBlockedMap()->getNumMaps() == 1) {
-      RCP<Vector> nestedVec = bD->getMultiVector(0,bD->getBlockedMap()->getThyraMode())->getVectorNonConst(0);
-      D_.swap(nestedVec);
-    }
-
-    // Set the Smoother
-    // carefully switch to the SubFactoryManagers (defined by the users)
-    {
-      SetFactoryManager currentSFM(rcpFromRef(currentLevel), FactManager_);
-      smoo_ = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", FactManager_->GetFactory("Smoother").get());
-      S_    = currentLevel.Get<RCP<Matrix> >      ("A",           FactManager_->GetFactory("A").get());
-    }
-
-    SmootherPrototype::IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::BreaessSarazinSmoother::Setup(): Setup() has already been called";
+
+  // Extract blocked operator A from current level
+  A_                       = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
+                             "MueLu::BraessSarazinSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
+
+  // Store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
+
+  // Store the blocks in local member variables
+  A00_ = bA->getMatrix(0, 0);
+  A01_ = bA->getMatrix(0, 1);
+  A10_ = bA->getMatrix(1, 0);
+  A11_ = bA->getMatrix(1, 1);
+
+  const ParameterList& pL = Factory::GetParameterList();
+  SC omega                = pL.get<SC>("Damping factor");
+
+  // Create the inverse of the diagonal of F
+  // TODO add safety check for zeros on diagonal of F!
+  RCP<Vector> diagFVector = VectorFactory::Build(A00_->getRowMap());
+  if (pL.get<bool>("lumping") == false) {
+    A00_->getLocalDiagCopy(*diagFVector);  // extract diagonal of F
+  } else {
+    diagFVector = Utilities::GetLumpedMatrixDiagonal(*A00_);
+  }
+  diagFVector->scale(omega);
+  D_ = Utilities::GetInverse(diagFVector);
+
+  // check whether D_ is a blocked vector with only 1 block
+  RCP<BlockedVector> bD = Teuchos::rcp_dynamic_cast<BlockedVector>(D_);
+  if (bD.is_null() == false && bD->getBlockedMap()->getNumMaps() == 1) {
+    RCP<Vector> nestedVec = bD->getMultiVector(0, bD->getBlockedMap()->getThyraMode())->getVectorNonConst(0);
+    D_.swap(nestedVec);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
-                               "MueLu::BraessSarazinSmoother::Apply(): Setup() has not been called");
-
-    RCP<MultiVector> rcpX       = rcpFromRef(X);
-    RCP<const MultiVector> rcpB = rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
+  // Set the Smoother
+  // carefully switch to the SubFactoryManagers (defined by the users)
+  {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), FactManager_);
+    smoo_ = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", FactManager_->GetFactory("Smoother").get());
+    S_    = currentLevel.Get<RCP<Matrix> >("A", FactManager_->GetFactory("A").get());
+  }
 
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
+  SmootherPrototype::IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
+                             "MueLu::BraessSarazinSmoother::Apply(): Setup() has not been called");
+
+  RCP<MultiVector> rcpX       = rcpFromRef(X);
+  RCP<const MultiVector> rcpB = rcpFromRef(B);
+
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(bA);
+
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
     }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
 
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
+    }
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
     }
 
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+    }
+  }
 
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
+    }
 
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
     }
+  }
 
-    // use the GIDs of the sub blocks
-    // This is valid as the subblocks actually represent the GIDs (either Thyra, Xpetra or pseudo Xpetra)
+  // use the GIDs of the sub blocks
+  // This is valid as the subblocks actually represent the GIDs (either Thyra, Xpetra or pseudo Xpetra)
 
-    bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
-    bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
+  bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
+  bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
 
-    RCP<MultiVector> deltaX     = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
-    RCP<BlockedMultiVector> bdeltaX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(deltaX);
-    RCP<MultiVector> deltaX0 = bdeltaX->getMultiVector(0,bDomainThyraMode);
-    RCP<MultiVector> deltaX1 = bdeltaX->getMultiVector(1,bDomainThyraMode);
+  RCP<MultiVector> deltaX         = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
+  RCP<BlockedMultiVector> bdeltaX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(deltaX);
+  RCP<MultiVector> deltaX0        = bdeltaX->getMultiVector(0, bDomainThyraMode);
+  RCP<MultiVector> deltaX1        = bdeltaX->getMultiVector(1, bDomainThyraMode);
 
-    RCP<MultiVector> Rtmp    = rangeMapExtractor_->getVector(1, rcpB->getNumVectors(), bRangeThyraMode);
+  RCP<MultiVector> Rtmp = rangeMapExtractor_->getVector(1, rcpB->getNumVectors(), bRangeThyraMode);
 
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC one = STS::one(), zero = STS::zero();
 
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC one = STS::one(), zero = STS::zero();
+  // extract parameters from internal parameter list
+  const ParameterList& pL = Factory::GetParameterList();
+  LO nSweeps              = pL.get<LO>("Sweeps");
 
-    // extract parameters from internal parameter list
-    const ParameterList& pL = Factory::GetParameterList();
-    LO nSweeps = pL.get<LO>("Sweeps");
+  RCP<MultiVector> R;  // = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());;
+  if (InitialGuessIsZero) {
+    R = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+    R->update(one, *rcpB, zero);
+  } else {
+    R = Utilities::Residual(*A_, *rcpX, *rcpB);
+  }
 
-    RCP<MultiVector> R;// = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());;
-    if (InitialGuessIsZero)  {
-      R = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-      R->update(one, *rcpB, zero);
+  // extract diagonal of Schur complement operator
+  RCP<Vector> diagSVector = VectorFactory::Build(S_->getRowMap());
+  S_->getLocalDiagCopy(*diagSVector);
+
+  for (LO run = 0; run < nSweeps; ++run) {
+    // Extract corresponding subvectors from X and R
+    // Automatically detect whether we use Thyra or Xpetra GIDs
+    // The GIDs should be always compatible with the GIDs of A00, A01, etc...
+    RCP<MultiVector> R0 = rangeMapExtractor_->ExtractVector(R, 0, bRangeThyraMode);
+    RCP<MultiVector> R1 = rangeMapExtractor_->ExtractVector(R, 1, bRangeThyraMode);
+
+    // Calculate Rtmp = R1 - D * deltaX0 (equation 8.14)
+    deltaX0->putScalar(zero);
+    deltaX0->elementWiseMultiply(one, *D_, *R0, zero);  // deltaX0 = D * R0 (equation 8.13)
+    A10_->apply(*deltaX0, *Rtmp);                       // Rtmp    = A10*D*deltaX0 (intermediate step)
+    Rtmp->update(one, *R1, -one);                       // Rtmp    = R1 - A10*D*deltaX0
+
+    if (!pL.get<bool>("q2q1 mode")) {
+      deltaX1->putScalar(zero);
     } else {
-      R = Utilities::Residual(*A_, *rcpX, *rcpB);
-    }
-
-    // extract diagonal of Schur complement operator
-    RCP<Vector> diagSVector = VectorFactory::Build(S_->getRowMap());
-    S_->getLocalDiagCopy(*diagSVector);
-
-    for (LO run = 0; run < nSweeps; ++run) {
-      // Extract corresponding subvectors from X and R
-      // Automatically detect whether we use Thyra or Xpetra GIDs
-      // The GIDs should be always compatible with the GIDs of A00, A01, etc...
-      RCP<MultiVector> R0 = rangeMapExtractor_ ->ExtractVector(R, 0, bRangeThyraMode);
-      RCP<MultiVector> R1 = rangeMapExtractor_ ->ExtractVector(R, 1, bRangeThyraMode);
-
-      // Calculate Rtmp = R1 - D * deltaX0 (equation 8.14)
-      deltaX0->putScalar(zero);
-      deltaX0->elementWiseMultiply(one, *D_, *R0, zero);    // deltaX0 = D * R0 (equation 8.13)
-      A10_->apply(*deltaX0, *Rtmp);                         // Rtmp    = A10*D*deltaX0 (intermediate step)
-      Rtmp->update(one, *R1, -one);                         // Rtmp    = R1 - A10*D*deltaX0
-
-      if (!pL.get<bool>("q2q1 mode")) {
-        deltaX1->putScalar(zero);
+      // special code for q2q1
+      if (Teuchos::rcp_dynamic_cast<BlockedVector>(diagSVector) == Teuchos::null) {
+        ArrayRCP<SC> Sdiag       = diagSVector->getDataNonConst(0);
+        ArrayRCP<SC> deltaX1data = deltaX1->getDataNonConst(0);
+        ArrayRCP<SC> Rtmpdata    = Rtmp->getDataNonConst(0);
+        for (GO row = 0; row < deltaX1data.size(); row++)
+          deltaX1data[row] = Teuchos::as<SC>(1.1) * Rtmpdata[row] / Sdiag[row];
       } else {
-        // special code for q2q1
-        if(Teuchos::rcp_dynamic_cast<BlockedVector>(diagSVector) == Teuchos::null) {
-          ArrayRCP<SC> Sdiag = diagSVector->getDataNonConst(0);
-          ArrayRCP<SC> deltaX1data = deltaX1->getDataNonConst(0);
-          ArrayRCP<SC> Rtmpdata    = Rtmp->getDataNonConst(0);
-          for (GO row = 0; row < deltaX1data.size(); row++)
-            deltaX1data[row] = Teuchos::as<SC>(1.1)*Rtmpdata[row] / Sdiag[row];
-        } else {
-          TEUCHOS_TEST_FOR_EXCEPTION(true,MueLu::Exceptions::RuntimeError,"MueLu::BraessSarazinSmoother: q2q1 mode only supported for non-blocked operators.")
-        }
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "MueLu::BraessSarazinSmoother: q2q1 mode only supported for non-blocked operators.")
       }
+    }
 
-      smoo_->Apply(*deltaX1,*Rtmp);
-
-      // Compute deltaX0
-      deltaX0->putScalar(zero);                             // just for safety
-      A01_->apply(*deltaX1, *deltaX0);                      // deltaX0 = A01*deltaX1
-      deltaX0->update(one, *R0, -one);                      // deltaX0 = R0 - A01*deltaX1
-      R0.swap(deltaX0);
-      deltaX0->elementWiseMultiply(one, *D_, *R0, zero);    // deltaX0 = D*(R0 - A01*deltaX1)
-
-      RCP<MultiVector> X0 = domainMapExtractor_->ExtractVector(rcpX, 0, bDomainThyraMode);
-      RCP<MultiVector> X1 = domainMapExtractor_->ExtractVector(rcpX, 1, bDomainThyraMode);
+    smoo_->Apply(*deltaX1, *Rtmp);
 
-      // Update solution
-      X0->update(one, *deltaX0, one);
-      X1->update(one, *deltaX1, one);
+    // Compute deltaX0
+    deltaX0->putScalar(zero);         // just for safety
+    A01_->apply(*deltaX1, *deltaX0);  // deltaX0 = A01*deltaX1
+    deltaX0->update(one, *R0, -one);  // deltaX0 = R0 - A01*deltaX1
+    R0.swap(deltaX0);
+    deltaX0->elementWiseMultiply(one, *D_, *R0, zero);  // deltaX0 = D*(R0 - A01*deltaX1)
 
-      domainMapExtractor_->InsertVector(X0, 0, rcpX, bDomainThyraMode);
-      domainMapExtractor_->InsertVector(X1, 1, rcpX, bDomainThyraMode);
+    RCP<MultiVector> X0 = domainMapExtractor_->ExtractVector(rcpX, 0, bDomainThyraMode);
+    RCP<MultiVector> X1 = domainMapExtractor_->ExtractVector(rcpX, 1, bDomainThyraMode);
 
-      if (run < nSweeps-1) {
-        R = Utilities::Residual(*A_, *rcpX, *rcpB);
-      }
+    // Update solution
+    X0->update(one, *deltaX0, one);
+    X1->update(one, *deltaX1, one);
 
-    }
+    domainMapExtractor_->InsertVector(X0, 0, rcpX, bDomainThyraMode);
+    domainMapExtractor_->InsertVector(X1, 1, rcpX, bDomainThyraMode);
 
-    if (bCopyResultX == true) {
-      RCP<MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
+    if (run < nSweeps - 1) {
+      R = Utilities::Residual(*A_, *rcpX, *rcpB);
     }
-
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy () const {
-    return rcp (new BraessSarazinSmoother (*this));
+  if (bCopyResultX == true) {
+    RCP<MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  description () const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new BraessSarazinSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t BraessSarazinSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_BRAESSSARAZINSMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_decl.hpp
index 06d08f912235..c25ce542ecd3 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_decl.hpp
@@ -53,17 +53,16 @@
 #ifndef MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_DECL_HPP_
 #define MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_DECL_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 
 #include <Teuchos_ParameterList.hpp>
 
-//Xpetra
+// Xpetra
 #include <Xpetra_MapExtractor_fwd.hpp>
 #include <Xpetra_MultiVectorFactory_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 
-//MueLu
+// MueLu
 #include "MueLu_IndefBlockedDiagonalSmoother_fwd.hpp"
 #include "MueLu_SmootherPrototype.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
@@ -75,114 +74,110 @@
 
 namespace MueLu {
 
-  /*!
-    @class IndefBlockedDiagonalSmoother
-    @brief Cheap Blocked diagonal smoother for indefinite 2x2 block matrices
+/*!
+  @class IndefBlockedDiagonalSmoother
+  @brief Cheap Blocked diagonal smoother for indefinite 2x2 block matrices
 
-    Uses the original upper left block and the Schur Complement block on the diagonal blocks.
-    Instead of solving the block equations exactly, we apply some sweeps with cheap iterative smoothers
-    (e.g. Gauss-Seidel iterations)
+  Uses the original upper left block and the Schur Complement block on the diagonal blocks.
+  Instead of solving the block equations exactly, we apply some sweeps with cheap iterative smoothers
+  (e.g. Gauss-Seidel iterations)
 
-  */
+*/
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class IndefBlockedDiagonalSmoother :
-    public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class IndefBlockedDiagonalSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! @brief Constructor
-    */
-    IndefBlockedDiagonalSmoother();
+  /*! @brief Constructor
+   */
+  IndefBlockedDiagonalSmoother();
 
-    //! Destructor
-    virtual ~IndefBlockedDiagonalSmoother();
-    //@}
+  //! Destructor
+  virtual ~IndefBlockedDiagonalSmoother();
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //! Add a factory manager for BraessSarazin internal SchurComplement handling
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos = 0);
+  //! Add a factory manager for BraessSarazin internal SchurComplement handling
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos = 0);
 
-    //@}
+  //@}
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @name Setup and Apply methods.
+  //@{
 
-    /*! @brief Setup routine
-     */
-    void Setup(Level &currentLevel);
+  /*! @brief Setup routine
+   */
+  void Setup(Level &currentLevel);
 
-    /*! @brief Apply the Braess Sarazin smoother.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero TODO This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
-
-    RCP<SmootherPrototype> Copy() const;
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  /*! @brief Apply the Braess Sarazin smoother.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero TODO This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //@}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-  private:
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    //! smoother type
-    std::string type_;
+  //@}
 
-    RCP<const FactoryBase>                AFact_;           //!< A Factory
+ private:
+  //! smoother type
+  std::string type_;
 
-    //! block operator
-    RCP<Matrix>                           A_;               // < ! internal blocked operator "A" generated by AFact_
-    RCP<Matrix>                           F_;               //!< fluid operator
-    RCP<Matrix>                           Z_;               //!< pressure stabilization term or null block
+  RCP<const FactoryBase> AFact_;  //!< A Factory
 
-    RCP<const MapExtractorClass>          rangeMapExtractor_;  //!< range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass>          domainMapExtractor_; //!< domain map extractor (from A_ generated by AFact)
+  //! block operator
+  RCP<Matrix> A_;  // < ! internal blocked operator "A" generated by AFact_
+  RCP<Matrix> F_;  //!< fluid operator
+  RCP<Matrix> Z_;  //!< pressure stabilization term or null block
 
-    //! Block smoothers
-    Teuchos::RCP<SmootherBase>            velPredictSmoo_;              //!< smoother for velocity prediction
-    Teuchos::RCP<SmootherBase>            schurCompSmoo_;               //!< smoother for SchurComplement equation
+  RCP<const MapExtractorClass> rangeMapExtractor_;   //!< range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> domainMapExtractor_;  //!< domain map extractor (from A_ generated by AFact)
 
-    //! vector of factory managers
-    /*!
-     * vector of internal factory managers
-     * FactManager_[0] holds the factory manager for the predicting the primary variable
-     * FactManager_[1] stores the factory manager used for the SchurComplement correction step.
-     */
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
-  };
+  //! Block smoothers
+  Teuchos::RCP<SmootherBase> velPredictSmoo_;  //!< smoother for velocity prediction
+  Teuchos::RCP<SmootherBase> schurCompSmoo_;   //!< smoother for SchurComplement equation
 
-} // namespace MueLu
+  //! vector of factory managers
+  /*!
+   * vector of internal factory managers
+   * FactManager_[0] holds the factory manager for the predicting the primary variable
+   * FactManager_[1] stores the factory manager used for the SchurComplement correction step.
+   */
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+};
+
+}  // namespace MueLu
 
 #define MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_def.hpp
index 37c5eb238afa..4793bea398b9 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_IndefBlockedDiagonalSmoother_def.hpp
@@ -77,316 +77,310 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IndefBlockedDiagonalSmoother()
-    : type_("IndefiniteBlockDiagonalSmoother"), A_(Teuchos::null)
-  {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IndefBlockedDiagonalSmoother()
+  : type_("IndefiniteBlockDiagonalSmoother")
+  , A_(Teuchos::null) {
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~IndefBlockedDiagonalSmoother() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A (must be a 2x2 block matrix)");
+  validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor");
+  validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of SIMPLE sweeps (default = 1)");
+  // validParamList->set< bool >                  ("UseSIMPLEC",         false, "Use SIMPLEC instead of SIMPLE (default = false)");
+
+  return validParamList;
+}
+
+//! Add a factory manager at a specific position
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::IndefBlockedDiagonalSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+
+  size_t myPos = Teuchos::as<size_t>(pos);
+
+  if (myPos < FactManager_.size()) {
+    // replace existing entris in FactManager_ vector
+    FactManager_.at(myPos) = FactManager;
+  } else if (myPos == FactManager_.size()) {
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
+  } else {  // if(myPos > FactManager_.size())
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~IndefBlockedDiagonalSmoother() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+  TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError, "MueLu::IndefBlockedDiagonalSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\"!");
 
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A (must be a 2x2 block matrix)");
-    validParamList->set< Scalar >                ("Damping factor",     1.0, "Damping/Scaling factor");
-    validParamList->set< LocalOrdinal >          ("Sweeps",             1, "Number of SIMPLE sweeps (default = 1)");
-    //validParamList->set< bool >                  ("UseSIMPLEC",         false, "Use SIMPLEC instead of SIMPLE (default = false)");
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
-    return validParamList;
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", (*it)->GetFactory("Smoother").get());
   }
-
-  //! Add a factory manager at a specific position
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::IndefBlockedDiagonalSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
-
-    size_t myPos = Teuchos::as<size_t>(pos);
-
-    if (myPos < FactManager_.size()) {
-      // replace existing entris in FactManager_ vector
-      FactManager_.at(myPos) = FactManager;
-    } else if( myPos == FactManager_.size()) {
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    } else { // if(myPos > FactManager_.size())
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
-
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  FactoryMonitor m(*this, "Setup for indefinite blocked diagonal smoother", currentLevel);
+
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::IndefBlockedDiagonalSmoother::Setup(): Setup() has already been called";
+
+  // extract blocked operator A from current level
+  A_ = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::IndefBlockedDiagonalSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
+
+  // store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
+
+  // Store the blocks in local member variables
+  Teuchos::RCP<Matrix> A00 = bA->getMatrix(0, 0);
+  Teuchos::RCP<Matrix> A01 = bA->getMatrix(0, 1);
+  Teuchos::RCP<Matrix> A10 = bA->getMatrix(1, 0);
+  Teuchos::RCP<Matrix> A11 = bA->getMatrix(1, 1);
+
+  F_ = A00;
+  Z_ = A11;
+
+  /*const ParameterList & pL = Factory::GetParameterList();
+  bool bSIMPLEC = pL.get<bool>("UseSIMPLEC");
+
+  // Create the inverse of the diagonal of F
+  RCP<Vector> diagFVector = VectorFactory::Build(F_->getRowMap());
+  if(!bSIMPLEC) {
+    F_->getLocalDiagCopy(*diagFVector);       // extract diagonal of F
+    diagFVector->reciprocal(*diagFVector);    // build reciprocal
+  } else {
+    const RCP<const Map> rowmap = F_->getRowMap();
+    size_t locSize = rowmap->getLocalNumElements();
+    Teuchos::ArrayRCP<SC> diag = diagFVector->getDataNonConst(0);
+    Teuchos::ArrayView<const LO> cols;
+    Teuchos::ArrayView<const SC> vals;
+    for (size_t i=0; i<locSize; ++i) { // loop over rows
+      F_->getLocalRowView(i,cols,vals);
+      Scalar absRowSum = Teuchos::ScalarTraits<Scalar>::zero();
+      for (LO j=0; j<cols.size(); ++j) { // loop over cols
+        absRowSum += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+      }
+      diag[i] = absRowSum;
     }
+    diagFVector->reciprocal(*diagFVector);    // build reciprocal
   }
+  diagFinv_ = diagFVector;*/
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
-
-    TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError,"MueLu::IndefBlockedDiagonalSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\"!");
-
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel),   *it);
-
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother",(*it)->GetFactory("Smoother").get());
-    }
-
+  // Set the Smoother
+  // carefully switch to the SubFactoryManagers (defined by the users)
+  {
+    RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), velpredictFactManager);
+    velPredictSmoo_ = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", velpredictFactManager->GetFactory("Smoother").get());
   }
+  {
+    RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), schurFactManager);
+    schurCompSmoo_ = currentLevel.Get<RCP<SmootherBase> >("PreSmoother", schurFactManager->GetFactory("Smoother").get());
+  }
+  SmootherPrototype::IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool /* InitialGuessIsZero */) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::IndefBlockedDiagonalSmoother::Apply(): Setup() has not been called");
+
+  Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
+
+  SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
+
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
+
+  bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
+  bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
+
+  // wrap current solution vector in RCP
+  RCP<MultiVector> rcpX       = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
+
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(bA);
+
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
+    }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-    FactoryMonitor m(*this, "Setup for indefinite blocked diagonal smoother", currentLevel);
-
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::IndefBlockedDiagonalSmoother::Setup(): Setup() has already been called";
-
-    // extract blocked operator A from current level
-    A_ = Factory::Get<RCP<Matrix> > (currentLevel, "A");
-
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::IndefBlockedDiagonalSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
-
-    // store map extractors
-    rangeMapExtractor_ = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
-
-    // Store the blocks in local member variables
-    Teuchos::RCP<Matrix> A00 = bA->getMatrix(0, 0);
-    Teuchos::RCP<Matrix> A01 = bA->getMatrix(0, 1);
-    Teuchos::RCP<Matrix> A10 = bA->getMatrix(1, 0);
-    Teuchos::RCP<Matrix> A11 = bA->getMatrix(1, 1);
-
-    F_ = A00;
-    Z_ = A11;
-
-    /*const ParameterList & pL = Factory::GetParameterList();
-    bool bSIMPLEC = pL.get<bool>("UseSIMPLEC");
-
-    // Create the inverse of the diagonal of F
-    RCP<Vector> diagFVector = VectorFactory::Build(F_->getRowMap());
-    if(!bSIMPLEC) {
-      F_->getLocalDiagCopy(*diagFVector);       // extract diagonal of F
-      diagFVector->reciprocal(*diagFVector);    // build reciprocal
-    } else {
-      const RCP<const Map> rowmap = F_->getRowMap();
-      size_t locSize = rowmap->getLocalNumElements();
-      Teuchos::ArrayRCP<SC> diag = diagFVector->getDataNonConst(0);
-      Teuchos::ArrayView<const LO> cols;
-      Teuchos::ArrayView<const SC> vals;
-      for (size_t i=0; i<locSize; ++i) { // loop over rows
-        F_->getLocalRowView(i,cols,vals);
-        Scalar absRowSum = Teuchos::ScalarTraits<Scalar>::zero();
-        for (LO j=0; j<cols.size(); ++j) { // loop over cols
-          absRowSum += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-        }
-        diag[i] = absRowSum;
-      }
-      diagFVector->reciprocal(*diagFVector);    // build reciprocal
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
     }
-    diagFinv_ = diagFVector;*/
-
-    // Set the Smoother
-    // carefully switch to the SubFactoryManagers (defined by the users)
-    {
-      RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), velpredictFactManager);
-      velPredictSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother",velpredictFactManager->GetFactory("Smoother").get());
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
     }
-    {
-      RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), schurFactManager);
-      schurCompSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother", schurFactManager->GetFactory("Smoother").get());
+
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
     }
-    SmootherPrototype::IsSetup(true);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::IndefBlockedDiagonalSmoother::Apply(): Setup() has not been called");
-
-    Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
-
-    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
-
-    bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
-    bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
-
-    // wrap current solution vector in RCP
-    RCP<MultiVector>       rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
-    }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
     }
 
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
-
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
-
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
-
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
     }
+  }
 
-    // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
-
-    // create residual vector
-    // contains current residual of current solution X with rhs B
-    RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-    RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
-    Teuchos::RCP<MultiVector> r1 = bresidual->getMultiVector(0,bRangeThyraMode);
-    Teuchos::RCP<MultiVector> r2 = bresidual->getMultiVector(1,bRangeThyraMode);
-
-    // helper vector 1
-    RCP<MultiVector> xtilde     = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
-    RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
-    RCP<MultiVector> xtilde1 = bxtilde->getMultiVector(0,bDomainThyraMode);
-    RCP<MultiVector> xtilde2 = bxtilde->getMultiVector(1,bDomainThyraMode);
-
-    // incrementally improve solution vector X
-    for (LocalOrdinal run = 0; run < nSweeps; ++run) {
-      // 1) calculate current residual
-      residual->update(one,*rcpB,zero); // residual = B
-      A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
-
-      // split residual vector
-      //Teuchos::RCP<MultiVector> r1 = rangeMapExtractor_->ExtractVector(residual, 0, bRangeThyraMode);
-      //Teuchos::RCP<MultiVector> r2 = rangeMapExtractor_->ExtractVector(residual, 1, bRangeThyraMode);
-
-      // 2) solve F * \Delta \tilde{x}_1 = r_1
-      //    start with zero guess \Delta \tilde{x}_1
-      //RCP<MultiVector> xtilde1 = MultiVectorFactory::Build(r1->getMap(),rcpX->getNumVectors(),true);
-      //RCP<MultiVector> xtilde2 = MultiVectorFactory::Build(r2->getMap(),rcpX->getNumVectors(),true);
-      bxtilde->putScalar(zero);
-      velPredictSmoo_->Apply(*xtilde1,*r1);
-
-      // 3) solve SchurComp equation
-      //    start with zero guess \Delta \tilde{x}_2
-      schurCompSmoo_->Apply(*xtilde2,*r2);
+  // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
+
+  // create residual vector
+  // contains current residual of current solution X with rhs B
+  RCP<MultiVector> residual         = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
+  Teuchos::RCP<MultiVector> r1      = bresidual->getMultiVector(0, bRangeThyraMode);
+  Teuchos::RCP<MultiVector> r2      = bresidual->getMultiVector(1, bRangeThyraMode);
+
+  // helper vector 1
+  RCP<MultiVector> xtilde         = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
+  RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
+  RCP<MultiVector> xtilde1        = bxtilde->getMultiVector(0, bDomainThyraMode);
+  RCP<MultiVector> xtilde2        = bxtilde->getMultiVector(1, bDomainThyraMode);
+
+  // incrementally improve solution vector X
+  for (LocalOrdinal run = 0; run < nSweeps; ++run) {
+    // 1) calculate current residual
+    residual->update(one, *rcpB, zero);  // residual = B
+    A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
+
+    // split residual vector
+    // Teuchos::RCP<MultiVector> r1 = rangeMapExtractor_->ExtractVector(residual, 0, bRangeThyraMode);
+    // Teuchos::RCP<MultiVector> r2 = rangeMapExtractor_->ExtractVector(residual, 1, bRangeThyraMode);
+
+    // 2) solve F * \Delta \tilde{x}_1 = r_1
+    //    start with zero guess \Delta \tilde{x}_1
+    // RCP<MultiVector> xtilde1 = MultiVectorFactory::Build(r1->getMap(),rcpX->getNumVectors(),true);
+    // RCP<MultiVector> xtilde2 = MultiVectorFactory::Build(r2->getMap(),rcpX->getNumVectors(),true);
+    bxtilde->putScalar(zero);
+    velPredictSmoo_->Apply(*xtilde1, *r1);
+
+    // 3) solve SchurComp equation
+    //    start with zero guess \Delta \tilde{x}_2
+    schurCompSmoo_->Apply(*xtilde2, *r2);
 #if 1
-      // 4) update solution vector
-      rcpX->update(omega,*bxtilde,one);
+    // 4) update solution vector
+    rcpX->update(omega, *bxtilde, one);
 #else
-      Teuchos::RCP<MultiVector> x1 = domainMapExtractor_->ExtractVector(rcpX, 0, bDomainThyraMode);
-      Teuchos::RCP<MultiVector> x2 = domainMapExtractor_->ExtractVector(rcpX, 1, bDomainThyraMode);
+    Teuchos::RCP<MultiVector> x1 = domainMapExtractor_->ExtractVector(rcpX, 0, bDomainThyraMode);
+    Teuchos::RCP<MultiVector> x2 = domainMapExtractor_->ExtractVector(rcpX, 1, bDomainThyraMode);
 
-      // 5) update solution vector with increments xhat1 and xhat2
-      //    rescale increment for x2 with omega_
-      x1->update(omega,*xtilde1,one); // x1 = x1_old + omega xtilde1
-      x2->update(omega,*xtilde2,one); // x2 = x2_old + omega xtilde2
+    // 5) update solution vector with increments xhat1 and xhat2
+    //    rescale increment for x2 with omega_
+    x1->update(omega, *xtilde1, one);  // x1 = x1_old + omega xtilde1
+    x2->update(omega, *xtilde2, one);  // x2 = x2_old + omega xtilde2
 
-      // write back solution in global vector X
-      domainMapExtractor_->InsertVector(x1, 0, rcpX, bDomainThyraMode);
-      domainMapExtractor_->InsertVector(x2, 1, rcpX, bDomainThyraMode);
+    // write back solution in global vector X
+    domainMapExtractor_->InsertVector(x1, 0, rcpX, bDomainThyraMode);
+    domainMapExtractor_->InsertVector(x2, 1, rcpX, bDomainThyraMode);
 #endif
-    }
-
-    if (bCopyResultX == true) {
-      RCP<MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
-    }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp( new IndefBlockedDiagonalSmoother(*this) );
+  if (bCopyResultX == true) {
+    RCP<MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new IndefBlockedDiagonalSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
-  }
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
   }
-  
-
-
-} // namespace MueLu
+}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t IndefBlockedDiagonalSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_decl.hpp
index 79509b72ba79..2646601c0988 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_decl.hpp
@@ -47,17 +47,16 @@
 #ifndef MUELU_SIMPLESMOOTHER_DECL_HPP_
 #define MUELU_SIMPLESMOOTHER_DECL_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 
 #include <Teuchos_ParameterList.hpp>
 
-//Xpetra
+// Xpetra
 #include <Xpetra_MapExtractor_fwd.hpp>
 #include <Xpetra_MultiVectorFactory_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 
-//MueLu
+// MueLu
 #include "MueLu_SimpleSmoother_fwd.hpp"
 #include "MueLu_SmootherPrototype.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
@@ -70,148 +69,144 @@
 
 namespace MueLu {
 
-  /*!
-    @class SimpleSmoother
-    @brief SIMPLE smoother for 2x2 block matrices
+/*!
+  @class SimpleSmoother
+  @brief SIMPLE smoother for 2x2 block matrices
 
-  */
+*/
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class SimpleSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class SimpleSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_SIMPLESMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-
-    //! Constructor
-    SimpleSmoother();
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! Destructor
-    virtual ~SimpleSmoother();
+  //! Constructor
+  SimpleSmoother();
 
-    //@}
+  //! Destructor
+  virtual ~SimpleSmoother();
 
-    //! Input
-    //@{
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Set factory manager for internal velocity prediction
-    //! This routine is outdated. Use AddFactoryManager instead
-    void SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager);
+  void DeclareInput(Level &currentLevel) const;
 
-    //! Set factory manager for internal SchurComplement handling
-    //! This routine is outdated. Use AddFactoryManager instead
-    void SetSchurCompFactoryManager(RCP<FactoryManager> FactManager);
+  //! Set factory manager for internal velocity prediction
+  //! This routine is outdated. Use AddFactoryManager instead
+  void SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager);
 
-    //! Add a factory manager at a specific position
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
+  //! Set factory manager for internal SchurComplement handling
+  //! This routine is outdated. Use AddFactoryManager instead
+  void SetSchurCompFactoryManager(RCP<FactoryManager> FactManager);
 
-    //@}
+  //! Add a factory manager at a specific position
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
 
-    //! @name Setup and Apply methods.
-    //@{
+  //@}
 
-    /*!
-      @brief Setup routine
+  //! @name Setup and Apply methods.
+  //@{
 
-      The Setup() routine can be summarized in 4 steps:
-      1. Set the map extractors
-      2. Set the blocks
-      3. Create and set the inverse of the diagonal of F
-      4. Set the smoother for the Schur Complement
-     */
-    void Setup(Level &currentLevel);
-
-    /*! @brief Apply the Braess Sarazin smoother.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero TODO This option has no effect.
-    */
-    void Apply(MultiVector &X, MultiVector const &B, bool InitialGuessIsZero = false) const;
-    //@}
-
-    RCP<SmootherPrototype> Copy() const;
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  /*!
+    @brief Setup routine
+
+    The Setup() routine can be summarized in 4 steps:
+    1. Set the map extractors
+    2. Set the blocks
+    3. Create and set the inverse of the diagonal of F
+    4. Set the smoother for the Schur Complement
+   */
+  void Setup(Level &currentLevel);
+
+  /*! @brief Apply the Braess Sarazin smoother.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero TODO This option has no effect.
+  */
+  void Apply(MultiVector &X, MultiVector const &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //@}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-  private:
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    //! smoother type
-    std::string type_;
+  //@}
 
-    //!< Generating factory for operator "A"
-    RCP<const FactoryBase> AFact_;
+ private:
+  //! smoother type
+  std::string type_;
 
-    //! Internal blocked operator "A" generated by #AFact_
-    RCP<Matrix> A_;
+  //!< Generating factory for operator "A"
+  RCP<const FactoryBase> AFact_;
 
-    //! Range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> rangeMapExtractor_;
+  //! Internal blocked operator "A" generated by #AFact_
+  RCP<Matrix> A_;
 
-    //! Domain map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass> domainMapExtractor_;
+  //! Range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> rangeMapExtractor_;
 
-    //! @group Matrices
-    //! @{
+  //! Domain map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> domainMapExtractor_;
 
-    //! Inverse diagonal of fluid operator (vector).
-    Teuchos::RCP<Vector> diagFinv_;
+  //! @group Matrices
+  //! @{
 
-    //! Fluid operator
-    Teuchos::RCP<Matrix> F_;
+  //! Inverse diagonal of fluid operator (vector).
+  Teuchos::RCP<Vector> diagFinv_;
 
-    //! Pressure gradient operator
-    Teuchos::RCP<Matrix> G_;
+  //! Fluid operator
+  Teuchos::RCP<Matrix> F_;
 
-    //! Divergence operator
-    Teuchos::RCP<Matrix> D_;
+  //! Pressure gradient operator
+  Teuchos::RCP<Matrix> G_;
 
-    //! Pressure stabilization term or null block
-    Teuchos::RCP<Matrix> Z_;
+  //! Divergence operator
+  Teuchos::RCP<Matrix> D_;
 
-    //! @}
+  //! Pressure stabilization term or null block
+  Teuchos::RCP<Matrix> Z_;
 
-    //! Smoother for velocity prediction
-    Teuchos::RCP<SmootherBase> velPredictSmoo_;
+  //! @}
 
-    //! Smoother for SchurComplement equation
-    Teuchos::RCP<SmootherBase> schurCompSmoo_;
+  //! Smoother for velocity prediction
+  Teuchos::RCP<SmootherBase> velPredictSmoo_;
 
-    /*! @brief Vector of internal factory managers
-     *
-     * - FactManager_[0] holds the factory manager for the prediction of the primary variable
-     * - FactManager_[1] stores the factory manager used for the SchurComplement correction step.
-     */
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+  //! Smoother for SchurComplement equation
+  Teuchos::RCP<SmootherBase> schurCompSmoo_;
 
-  };
+  /*! @brief Vector of internal factory managers
+   *
+   * - FactManager_[0] holds the factory manager for the prediction of the primary variable
+   * - FactManager_[1] stores the factory manager used for the SchurComplement correction step.
+   */
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SIMPLESMOOTHER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_def.hpp
index 6e892869de0f..d671cbc918d8 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_SimpleSmoother_def.hpp
@@ -72,154 +72,151 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SimpleSmoother()
-    : type_("SIMPLE"), A_(Teuchos::null) {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SimpleSmoother()
+  : type_("SIMPLE")
+  , A_(Teuchos::null) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~SimpleSmoother() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in SIMPLE");
+  validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of SIMPLE sweeps (default = 1)");
+  validParamList->set<bool>("UseSIMPLEC", false, "Use SIMPLEC instead of SIMPLE (default = false)");
+
+  return validParamList;
+}
+
+//! Add a factory manager at a specific position
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::SimpleSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+
+  size_t myPos = Teuchos::as<size_t>(pos);
+
+  if (myPos < FactManager_.size()) {
+    // replace existing entris in FactManager_ vector
+    FactManager_.at(myPos) = FactManager;
+  } else if (myPos == FactManager_.size()) {
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
+  } else {  // if(myPos > FactManager_.size())
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
+  }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~SimpleSmoother() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager) {
+  AddFactoryManager(FactManager, 0);  // overwrite factory manager for predicting the primary variable
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSchurCompFactoryManager(RCP<FactoryManager> FactManager) {
+  AddFactoryManager(FactManager, 1);  // overwrite factory manager for SchurComplement
+}
 
-    validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in SIMPLE");
-    validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of SIMPLE sweeps (default = 1)");
-    validParamList->set<bool>("UseSIMPLEC", false, "Use SIMPLEC instead of SIMPLE (default = false)");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
 
-    return validParamList;
-  }
+  TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError, "MueLu::SimpleSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\". make sure that you use the same proper damping factors for omega both in the SchurComplementFactory and in the SIMPLE smoother!");
 
-  //! Add a factory manager at a specific position
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::SimpleSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
-
-    size_t myPos = Teuchos::as<size_t>(pos);
-
-    if (myPos < FactManager_.size()) {
-      // replace existing entris in FactManager_ vector
-      FactManager_.at(myPos) = FactManager;
-    } else if( myPos == FactManager_.size()) {
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    } else { // if(myPos > FactManager_.size())
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
-
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    }
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase>>::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", (*it)->GetFactory("Smoother").get());
   }
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager) {
-    AddFactoryManager(FactManager, 0); // overwrite factory manager for predicting the primary variable
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  FactoryMonitor m(*this, "Setup blocked SIMPLE Smoother", currentLevel);
+
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::SimpleSmoother::Setup(): Setup() has already been called";
+
+  // extract blocked operator A from current level
+  A_ = Factory::Get<RCP<Matrix>>(currentLevel, "A");
+
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::SimpleSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
+
+  // store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
+
+  // Store the blocks in local member variables
+  F_ = bA->getMatrix(0, 0);
+  G_ = bA->getMatrix(0, 1);
+  D_ = bA->getMatrix(1, 0);
+  Z_ = bA->getMatrix(1, 1);
+
+  const ParameterList &pL = Factory::GetParameterList();
+  bool bSIMPLEC           = pL.get<bool>("UseSIMPLEC");
+
+  // Create the inverse of the diagonal of F
+  // TODO add safety check for zeros on diagonal of F!
+  RCP<Vector> diagFVector = VectorFactory::Build(F_->getRowMap());
+  if (!bSIMPLEC) {
+    F_->getLocalDiagCopy(*diagFVector);  // extract diagonal of F
+  } else {
+    /*const RCP<const Map> rowmap = F_->getRowMap();
+    size_t locSize = rowmap->getLocalNumElements();
+    Teuchos::ArrayRCP<SC> diag = diagFVector->getDataNonConst(0);
+    Teuchos::ArrayView<const LO> cols;
+    Teuchos::ArrayView<const SC> vals;
+    for (size_t i=0; i<locSize; ++i) { // loop over rows
+      F_->getLocalRowView(i,cols,vals);
+      Scalar absRowSum = Teuchos::ScalarTraits<Scalar>::zero();
+      for (LO j=0; j<cols.size(); ++j) { // loop over cols
+        absRowSum += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+      }
+      diag[i] = absRowSum;
+    }*/
+    // TODO this does not work if F_ is nested!
+    diagFVector = Utilities::GetLumpedMatrixDiagonal(*F_);
   }
+  diagFinv_ = Utilities::GetInverse(diagFVector);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSchurCompFactoryManager(RCP<FactoryManager> FactManager) {
-    AddFactoryManager(FactManager, 1); // overwrite factory manager for SchurComplement
+  // check whether diagFinv_ is a blocked vector with only 1 block
+  RCP<BlockedVector> bdiagFinv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagFinv_);
+  if (bdiagFinv.is_null() == false && bdiagFinv->getBlockedMap()->getNumMaps() == 1) {
+    RCP<Vector> nestedVec = bdiagFinv->getMultiVector(0, bdiagFinv->getBlockedMap()->getThyraMode())->getVectorNonConst(0);
+    diagFinv_.swap(nestedVec);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
-
-    TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError,"MueLu::SimpleSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\". make sure that you use the same proper damping factors for omega both in the SchurComplementFactory and in the SIMPLE smoother!");
-
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel),   *it);
-
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother",(*it)->GetFactory("Smoother").get());
-    }
+  // Set the Smoother
+  // carefully switch to the SubFactoryManagers (defined by the users)
+  {
+    RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), velpredictFactManager);
+    velPredictSmoo_ = currentLevel.Get<RCP<SmootherBase>>("PreSmoother", velpredictFactManager->GetFactory("Smoother").get());
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-
-    FactoryMonitor m(*this, "Setup blocked SIMPLE Smoother", currentLevel);
-
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::SimpleSmoother::Setup(): Setup() has already been called";
-
-    // extract blocked operator A from current level
-    A_ = Factory::Get<RCP<Matrix> > (currentLevel, "A");
-
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::SimpleSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
-
-    // store map extractors
-    rangeMapExtractor_ = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
-
-    // Store the blocks in local member variables
-    F_ = bA->getMatrix(0, 0);
-    G_ = bA->getMatrix(0, 1);
-    D_ = bA->getMatrix(1, 0);
-    Z_ = bA->getMatrix(1, 1);
-
-    const ParameterList & pL = Factory::GetParameterList();
-    bool bSIMPLEC = pL.get<bool>("UseSIMPLEC");
-
-    // Create the inverse of the diagonal of F
-    // TODO add safety check for zeros on diagonal of F!
-    RCP<Vector> diagFVector = VectorFactory::Build(F_->getRowMap());
-    if(!bSIMPLEC) {
-      F_->getLocalDiagCopy(*diagFVector);       // extract diagonal of F
-    } else {
-      /*const RCP<const Map> rowmap = F_->getRowMap();
-      size_t locSize = rowmap->getLocalNumElements();
-      Teuchos::ArrayRCP<SC> diag = diagFVector->getDataNonConst(0);
-      Teuchos::ArrayView<const LO> cols;
-      Teuchos::ArrayView<const SC> vals;
-      for (size_t i=0; i<locSize; ++i) { // loop over rows
-        F_->getLocalRowView(i,cols,vals);
-        Scalar absRowSum = Teuchos::ScalarTraits<Scalar>::zero();
-        for (LO j=0; j<cols.size(); ++j) { // loop over cols
-          absRowSum += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-        }
-        diag[i] = absRowSum;
-      }*/
-      // TODO this does not work if F_ is nested!
-      diagFVector = Utilities::GetLumpedMatrixDiagonal(*F_);
-    }
-    diagFinv_ = Utilities::GetInverse(diagFVector);
-
-    // check whether diagFinv_ is a blocked vector with only 1 block
-    RCP<BlockedVector> bdiagFinv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagFinv_);
-    if(bdiagFinv.is_null() == false && bdiagFinv->getBlockedMap()->getNumMaps() == 1) {
-      RCP<Vector> nestedVec = bdiagFinv->getMultiVector(0,bdiagFinv->getBlockedMap()->getThyraMode())->getVectorNonConst(0);
-      diagFinv_.swap(nestedVec);
-    }
-
-    // Set the Smoother
-    // carefully switch to the SubFactoryManagers (defined by the users)
-    {
-      RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), velpredictFactManager);
-      velPredictSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother",velpredictFactManager->GetFactory("Smoother").get());
-    }
-    {
-      RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), schurFactManager);
-      schurCompSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother", schurFactManager->GetFactory("Smoother").get());
-    }
-
-    SmootherPrototype::IsSetup(true);
+  {
+    RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), schurFactManager);
+    schurCompSmoo_ = currentLevel.Get<RCP<SmootherBase>>("PreSmoother", schurFactManager->GetFactory("Smoother").get());
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
-        "MueLu::SimpleSmoother::Apply(): Setup() has not been called");
+  SmootherPrototype::IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError,
+                             "MueLu::SimpleSmoother::Apply(): Setup() has not been called");
 #if 0
     // TODO simplify this debug check
     RCP<MultiVector> rcpDebugX = Teuchos::rcpFromRef(X);
@@ -238,195 +235,189 @@ namespace MueLu {
     }
 #endif
 
-    const SC zero = Teuchos::ScalarTraits<SC>::zero();
-    const SC one = Teuchos::ScalarTraits<SC>::one();
-
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
-
-    // The boolean flags check whether we use Thyra or Xpetra style GIDs
-    bool bRangeThyraMode = rangeMapExtractor_->getThyraMode(); //  && (Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(F_) == Teuchos::null);
-    bool bDomainThyraMode = domainMapExtractor_->getThyraMode(); // && (Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(F_) == Teuchos::null);
-
-    //RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-
-    // wrap current solution vector in RCP
-    RCP<MultiVector> rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
-
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
+  const SC zero = Teuchos::ScalarTraits<SC>::zero();
+  const SC one  = Teuchos::ScalarTraits<SC>::one();
+
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
+
+  // The boolean flags check whether we use Thyra or Xpetra style GIDs
+  bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();   //  && (Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(F_) == Teuchos::null);
+  bool bDomainThyraMode = domainMapExtractor_->getThyraMode();  // && (Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(F_) == Teuchos::null);
+
+  // RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+
+  // wrap current solution vector in RCP
+  RCP<MultiVector> rcpX       = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
+
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(bA);
+
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
     }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
 
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
     }
-
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
-
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
-
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
-
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
     }
 
-    // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
-
-    // create residual vector
-    // contains current residual of current solution X with rhs B
-    RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-    RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
-    RCP<MultiVector> r1 = bresidual->getMultiVector(0,bRangeThyraMode);
-    RCP<MultiVector> r2 = bresidual->getMultiVector(1,bRangeThyraMode);
-
-    // helper vector 1
-    RCP<MultiVector> xtilde = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
-    RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
-    RCP<MultiVector> xtilde1 = bxtilde->getMultiVector(0, bDomainThyraMode);
-    RCP<MultiVector> xtilde2 = bxtilde->getMultiVector(1, bDomainThyraMode);
-
-    // helper vector 2
-    RCP<MultiVector> xhat = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
-    RCP<BlockedMultiVector> bxhat = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xhat);
-    RCP<MultiVector> xhat1 = bxhat->getMultiVector(0, bDomainThyraMode);
-    RCP<MultiVector> xhat2 = bxhat->getMultiVector(1, bDomainThyraMode);
-
-
-    // incrementally improve solution vector X
-    for (LocalOrdinal run = 0; run < nSweeps; ++run) {
-      // 1) calculate current residual
-      residual->update(one, *rcpB, zero); // residual = B
-      if(InitialGuessIsZero == false || run > 0)
-        A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
-
-      // 2) solve F * \Delta \tilde{x}_1 = r_1
-      //    start with zero guess \Delta \tilde{x}_1
-      xtilde1->putScalar(zero);
-      xtilde2->putScalar(zero);
-      velPredictSmoo_->Apply(*xtilde1, *r1);
-
-      // 3) calculate rhs for SchurComp equation
-      //    r_2 - D \Delta \tilde{x}_1
-      RCP<MultiVector> schurCompRHS = rangeMapExtractor_->getVector(1, rcpB->getNumVectors(), bRangeThyraMode);
-      D_->apply(*xtilde1, *schurCompRHS);
-
-      schurCompRHS->update(one, *r2, -one);
-
-      // 4) solve SchurComp equation
-      //    start with zero guess \Delta \tilde{x}_2
-      schurCompSmoo_->Apply(*xtilde2, *schurCompRHS);
-
-      // 5) scale xtilde2 with omega
-      //    store this in xhat2
-      xhat2->update(omega, *xtilde2, zero);
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+    }
+  }
 
-      // 6) calculate xhat1
-      RCP<MultiVector> xhat1_temp = domainMapExtractor_->getVector(0, rcpX->getNumVectors(), bDomainThyraMode);
-      G_->apply(*xhat2, *xhat1_temp); // store result temporarely in xtilde1_temp
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-      xhat1->elementWiseMultiply(one/*/omega*/, *diagFinv_, *xhat1_temp, zero);
-      xhat1->update(one, *xtilde1, -one);
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-      rcpX->update(one, *bxhat, one);
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
     }
 
-    if (bCopyResultX == true) {
-      RCP<const MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
     }
-
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy () const {
-    return rcp( new SimpleSmoother(*this) );
+  // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
+
+  // create residual vector
+  // contains current residual of current solution X with rhs B
+  RCP<MultiVector> residual         = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
+  RCP<MultiVector> r1               = bresidual->getMultiVector(0, bRangeThyraMode);
+  RCP<MultiVector> r2               = bresidual->getMultiVector(1, bRangeThyraMode);
+
+  // helper vector 1
+  RCP<MultiVector> xtilde         = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
+  RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
+  RCP<MultiVector> xtilde1        = bxtilde->getMultiVector(0, bDomainThyraMode);
+  RCP<MultiVector> xtilde2        = bxtilde->getMultiVector(1, bDomainThyraMode);
+
+  // helper vector 2
+  RCP<MultiVector> xhat         = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
+  RCP<BlockedMultiVector> bxhat = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xhat);
+  RCP<MultiVector> xhat1        = bxhat->getMultiVector(0, bDomainThyraMode);
+  RCP<MultiVector> xhat2        = bxhat->getMultiVector(1, bDomainThyraMode);
+
+  // incrementally improve solution vector X
+  for (LocalOrdinal run = 0; run < nSweeps; ++run) {
+    // 1) calculate current residual
+    residual->update(one, *rcpB, zero);  // residual = B
+    if (InitialGuessIsZero == false || run > 0)
+      A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
+
+    // 2) solve F * \Delta \tilde{x}_1 = r_1
+    //    start with zero guess \Delta \tilde{x}_1
+    xtilde1->putScalar(zero);
+    xtilde2->putScalar(zero);
+    velPredictSmoo_->Apply(*xtilde1, *r1);
+
+    // 3) calculate rhs for SchurComp equation
+    //    r_2 - D \Delta \tilde{x}_1
+    RCP<MultiVector> schurCompRHS = rangeMapExtractor_->getVector(1, rcpB->getNumVectors(), bRangeThyraMode);
+    D_->apply(*xtilde1, *schurCompRHS);
+
+    schurCompRHS->update(one, *r2, -one);
+
+    // 4) solve SchurComp equation
+    //    start with zero guess \Delta \tilde{x}_2
+    schurCompSmoo_->Apply(*xtilde2, *schurCompRHS);
+
+    // 5) scale xtilde2 with omega
+    //    store this in xhat2
+    xhat2->update(omega, *xtilde2, zero);
+
+    // 6) calculate xhat1
+    RCP<MultiVector> xhat1_temp = domainMapExtractor_->getVector(0, rcpX->getNumVectors(), bDomainThyraMode);
+    G_->apply(*xhat2, *xhat1_temp);  // store result temporarely in xtilde1_temp
+
+    xhat1->elementWiseMultiply(one /*/omega*/, *diagFinv_, *xhat1_temp, zero);
+    xhat1->update(one, *xtilde1, -one);
+
+    rcpX->update(one, *bxhat, one);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
+  if (bCopyResultX == true) {
+    RCP<const MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new SimpleSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t SimpleSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
-
+}  // namespace MueLu
 
 #endif /* MUELU_SIMPLESMOOTHER_DEF_HPP_ */
\ No newline at end of file
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_decl.hpp
index 8420cfd7e25f..392f69c41d59 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_decl.hpp
@@ -68,309 +68,312 @@
 
 namespace MueLu {
 
-  /*!
-    @class TekoSmoother
-    @brief Interface to block smoothers in Teko package
-
-    This is a dummy implementation for arbitrary scalars (different than SC=double).
-    We use Teko and the Teko::LinearOp which is declared as
-    Teuchos::RCP<const Thyra::LinearOpBase<ST> > with ST=double. See e.g. Teko_ConfigDefs.hpp.
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class TekoSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+/*!
+  @class TekoSmoother
+  @brief Interface to block smoothers in Teko package
+
+  This is a dummy implementation for arbitrary scalars (different than SC=double).
+  We use Teko and the Teko::LinearOp which is declared as
+  Teuchos::RCP<const Thyra::LinearOpBase<ST> > with ST=double. See e.g. Teko_ConfigDefs.hpp.
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class TekoSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_TEKOSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! @brief Constructor
-    */
-    TekoSmoother() : type_("Teko smoother") {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::TekoSmoother: Teko can only be used with SC=double. For more information refer to the doxygen documentation of TekoSmoother.");
-    };
+  /*! @brief Constructor
+   */
+  TekoSmoother()
+    : type_("Teko smoother") {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::TekoSmoother: Teko can only be used with SC=double. For more information refer to the doxygen documentation of TekoSmoother.");
+  };
 
-    //! Destructor
-    virtual ~TekoSmoother() { }
-    //@}
+  //! Destructor
+  virtual ~TekoSmoother() {}
+  //@}
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const {
-      RCP<ParameterList> validParamList = rcp(new ParameterList());
-      return validParamList;
-    }
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
+    return validParamList;
+  }
 
-    void DeclareInput(Level &currentLevel) const { }
+  void DeclareInput(Level &currentLevel) const {}
 
-    void SetTekoParameters(RCP<ParameterList> tekoParams) { };
-    //@}
+  void SetTekoParameters(RCP<ParameterList> tekoParams){};
+  //@}
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @name Setup and Apply methods.
+  //@{
 
-    /*! @brief Setup routine
-    */
-    void Setup(Level &currentLevel) { }
+  /*! @brief Setup routine
+   */
+  void Setup(Level &currentLevel) {}
 
-    /*! @brief Apply the Teko smoother.
+  /*! @brief Apply the Teko smoother.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const { }
-    //@}
-
-    RCP<SmootherPrototype> Copy() const { return Teuchos::null; }
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const {}
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const {
-      std::ostringstream out;
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-      return out.str();
-    }
+  RCP<SmootherPrototype> Copy() const { return Teuchos::null; }
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
-      MUELU_DESCRIBE;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-      if (verbLevel & Parameters0)
-        out0 << "Prec. type: " << type_ << std::endl;
+  //! Return a simple one-line description of this object.
+  std::string description() const {
+    std::ostringstream out;
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+    return out.str();
+  }
 
-      if (verbLevel & Debug)
-        out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
+    MUELU_DESCRIBE;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+    if (verbLevel & Parameters0)
+      out0 << "Prec. type: " << type_ << std::endl;
 
-    //@}
+    if (verbLevel & Debug)
+      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  }
 
-  private:
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    //! smoother type
-    std::string type_;
-  }; // class TekoSmoother
+  //@}
 
+ private:
+  //! smoother type
+  std::string type_;
+};  // class TekoSmoother
 
-  /*!
-    @class TekoSmoother
-    @brief Interface to block smoothers in Teko package
+/*!
+  @class TekoSmoother
+  @brief Interface to block smoothers in Teko package
 
-    This is the specialization for SC=double and LO=int, which actually implements the Teko smoother interface.
-    We use Teko and the Teko::LinearOp which is declared as
-    Teuchos::RCP<const Thyra::LinearOpBase<ST> > with ST=double. See e.g. Teko_ConfigDefs.hpp.
+  This is the specialization for SC=double and LO=int, which actually implements the Teko smoother interface.
+  We use Teko and the Teko::LinearOp which is declared as
+  Teuchos::RCP<const Thyra::LinearOpBase<ST> > with ST=double. See e.g. Teko_ConfigDefs.hpp.
 
-    In Teko SC=double and LO=int are hard-coded. The node is Tpetra::Map<>::node_type.
-    The global ordinal GO is chosen as a conservative choice from what Tpetra supports
-    (see Teko_ConfigDefs.hpp for more details)
+  In Teko SC=double and LO=int are hard-coded. The node is Tpetra::Map<>::node_type.
+  The global ordinal GO is chosen as a conservative choice from what Tpetra supports
+  (see Teko_ConfigDefs.hpp for more details)
 
-    Note, that the underlying Thyra package is only templated on the Scalar (not the node type).
-  */
-  template <class GlobalOrdinal,
-            class Node>
-  class TekoSmoother<double,int,GlobalOrdinal,Node> : public SmootherPrototype<double,int,GlobalOrdinal,Node>
-  {
-    typedef int LocalOrdinal;
-    typedef double Scalar;
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+  Note, that the underlying Thyra package is only templated on the Scalar (not the node type).
+*/
+template <class GlobalOrdinal,
+          class Node>
+class TekoSmoother<double, int, GlobalOrdinal, Node> : public SmootherPrototype<double, int, GlobalOrdinal, Node> {
+  typedef int LocalOrdinal;
+  typedef double Scalar;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_TEKOSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-
-    /*! @brief Constructor
-    */
-    TekoSmoother() : type_("Teko smoother"), A_(Teuchos::null), bA_(Teuchos::null), bThyOp_(Teuchos::null), tekoParams_(Teuchos::null), inverseOp_(Teuchos::null) { };
-
-    //! Destructor
-    virtual ~TekoSmoother() { }
-    //@}
-
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const {
-      RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-      validParamList->set< RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A");
-      validParamList->set< std::string >           ("Inverse Type", "", "Name of parameter list within 'Teko parameters' containing the Teko smoother parameters.");
-
-      return validParamList;
-    }
-
-    void DeclareInput(Level &currentLevel) const {
-      this->Input(currentLevel, "A");
-    }
-
-    void SetTekoParameters(RCP<ParameterList> tekoParams) { tekoParams_ = tekoParams; };
-    //@}
-
-    //! @name Setup and Apply methods.
-    //@{
-
-    /*! @brief Setup routine
-    */
-    void Setup(Level &currentLevel) {
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-
-      FactoryMonitor m(*this, "Setup TekoSmoother", currentLevel);
-      if (this->IsSetup() == true)
-        this->GetOStream(Warnings0) << "MueLu::TekoSmoother::Setup(): Setup() has already been called";
-
-      // extract blocked operator A from current level
-      A_  = Factory::Get< RCP<Matrix> >(currentLevel, "A"); // A needed for extracting map extractors
-      bA_ = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-      TEUCHOS_TEST_FOR_EXCEPTION(bA_.is_null(), Exceptions::BadCast,
-                                 "MueLu::TekoSmoother::Build: input matrix A is not of type BlockedCrsMatrix.");
-
-      bThyOp_ = bA_->getThyraOperator();
-      TEUCHOS_TEST_FOR_EXCEPTION(bThyOp_.is_null(), Exceptions::BadCast,
-                                 "MueLu::TekoSmoother::Build: Could not extract thyra operator from BlockedCrsMatrix.");
-
-      Teuchos::RCP<const Thyra::LinearOpBase<Scalar> > thyOp = Teuchos::rcp_dynamic_cast<const Thyra::LinearOpBase<Scalar> >(bThyOp_);
-      TEUCHOS_TEST_FOR_EXCEPTION(thyOp.is_null(), Exceptions::BadCast,
-                                 "MueLu::TekoSmoother::Build: Downcast of Thyra::BlockedLinearOpBase to Teko::LinearOp failed.");
-
-      // parameter list contains TekoSmoother parameters but does not handle the Teko parameters itself!
-      const ParameterList& pL = Factory::GetParameterList();
-      std::string smootherType = pL.get<std::string>("Inverse Type");
-      TEUCHOS_TEST_FOR_EXCEPTION(smootherType.empty(), Exceptions::RuntimeError,
-          "MueLu::TekoSmoother::Build: You must provide a 'Smoother Type' name that is defined in the 'Teko parameters' sublist.");
-      type_ = smootherType;
-
-      TEUCHOS_TEST_FOR_EXCEPTION(tekoParams_.is_null(), Exceptions::BadCast,
-                                 "MueLu::TekoSmoother::Build: No Teko parameters have been set.");
-
-      Teuchos::RCP<Teko::InverseLibrary> invLib = Teko::InverseLibrary::buildFromParameterList(*tekoParams_);
-      Teuchos::RCP<Teko::InverseFactory> inverse = invLib->getInverseFactory(smootherType);
-
-      inverseOp_ = Teko::buildInverse(*inverse, thyOp);
-      TEUCHOS_TEST_FOR_EXCEPTION(inverseOp_.is_null(), Exceptions::BadCast,
-                                 "MueLu::TekoSmoother::Build: Failed to build Teko inverse operator. Probably a problem with the Teko parameters.");
-
-      this->IsSetup(true);
-    }
-
-    /*! @brief Apply the Teko smoother.
-
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */ = false) const {
-      TEUCHOS_TEST_FOR_EXCEPTION(this->IsSetup() == false, Exceptions::RuntimeError,
-                                 "MueLu::TekoSmoother::Apply(): Setup() has not been called");
+ public:
+  //! @name Constructors / destructors
+  //@{
+
+  /*! @brief Constructor
+   */
+  TekoSmoother()
+    : type_("Teko smoother")
+    , A_(Teuchos::null)
+    , bA_(Teuchos::null)
+    , bThyOp_(Teuchos::null)
+    , tekoParams_(Teuchos::null)
+    , inverseOp_(Teuchos::null){};
+
+  //! Destructor
+  virtual ~TekoSmoother() {}
+  //@}
+
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+    validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A");
+    validParamList->set<std::string>("Inverse Type", "", "Name of parameter list within 'Teko parameters' containing the Teko smoother parameters.");
+
+    return validParamList;
+  }
+
+  void DeclareInput(Level &currentLevel) const {
+    this->Input(currentLevel, "A");
+  }
+
+  void SetTekoParameters(RCP<ParameterList> tekoParams) { tekoParams_ = tekoParams; };
+  //@}
+
+  //! @name Setup and Apply methods.
+  //@{
+
+  /*! @brief Setup routine
+   */
+  void Setup(Level &currentLevel) {
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+
+    FactoryMonitor m(*this, "Setup TekoSmoother", currentLevel);
+    if (this->IsSetup() == true)
+      this->GetOStream(Warnings0) << "MueLu::TekoSmoother::Setup(): Setup() has already been called";
+
+    // extract blocked operator A from current level
+    A_  = Factory::Get<RCP<Matrix> >(currentLevel, "A");  // A needed for extracting map extractors
+    bA_ = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+    TEUCHOS_TEST_FOR_EXCEPTION(bA_.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Build: input matrix A is not of type BlockedCrsMatrix.");
+
+    bThyOp_ = bA_->getThyraOperator();
+    TEUCHOS_TEST_FOR_EXCEPTION(bThyOp_.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Build: Could not extract thyra operator from BlockedCrsMatrix.");
+
+    Teuchos::RCP<const Thyra::LinearOpBase<Scalar> > thyOp = Teuchos::rcp_dynamic_cast<const Thyra::LinearOpBase<Scalar> >(bThyOp_);
+    TEUCHOS_TEST_FOR_EXCEPTION(thyOp.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Build: Downcast of Thyra::BlockedLinearOpBase to Teko::LinearOp failed.");
+
+    // parameter list contains TekoSmoother parameters but does not handle the Teko parameters itself!
+    const ParameterList &pL  = Factory::GetParameterList();
+    std::string smootherType = pL.get<std::string>("Inverse Type");
+    TEUCHOS_TEST_FOR_EXCEPTION(smootherType.empty(), Exceptions::RuntimeError,
+                               "MueLu::TekoSmoother::Build: You must provide a 'Smoother Type' name that is defined in the 'Teko parameters' sublist.");
+    type_ = smootherType;
+
+    TEUCHOS_TEST_FOR_EXCEPTION(tekoParams_.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Build: No Teko parameters have been set.");
+
+    Teuchos::RCP<Teko::InverseLibrary> invLib  = Teko::InverseLibrary::buildFromParameterList(*tekoParams_);
+    Teuchos::RCP<Teko::InverseFactory> inverse = invLib->getInverseFactory(smootherType);
+
+    inverseOp_ = Teko::buildInverse(*inverse, thyOp);
+    TEUCHOS_TEST_FOR_EXCEPTION(inverseOp_.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Build: Failed to build Teko inverse operator. Probably a problem with the Teko parameters.");
+
+    this->IsSetup(true);
+  }
+
+  /*! @brief Apply the Teko smoother.
+
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool /* InitialGuessIsZero */ = false) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(this->IsSetup() == false, Exceptions::RuntimeError,
+                               "MueLu::TekoSmoother::Apply(): Setup() has not been called");
 
-      Teuchos::RCP<const Teuchos::Comm<int> > comm = X.getMap()->getComm();
+    Teuchos::RCP<const Teuchos::Comm<int> > comm = X.getMap()->getComm();
 
-      Teuchos::RCP<const MapExtractor> rgMapExtractor = bA_->getRangeMapExtractor();
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(rgMapExtractor));
+    Teuchos::RCP<const MapExtractor> rgMapExtractor = bA_->getRangeMapExtractor();
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(rgMapExtractor));
 
-      // copy initial solution vector X to Ptr<Thyra::MultiVectorBase> YY
+    // copy initial solution vector X to Ptr<Thyra::MultiVectorBase> YY
 
-      // create a Thyra RHS vector
-      Teuchos::RCP<Thyra::MultiVectorBase<Scalar> > thyB = Thyra::createMembers(Teuchos::rcp_dynamic_cast<const Thyra::VectorSpaceBase<Scalar> >(bThyOp_->productRange()),Teuchos::as<int>(B.getNumVectors()));
-      Teuchos::RCP<Thyra::ProductMultiVectorBase<Scalar> > thyProdB =
-          Teuchos::rcp_dynamic_cast<Thyra::ProductMultiVectorBase<Scalar> >(thyB);
-      TEUCHOS_TEST_FOR_EXCEPTION(thyProdB.is_null(), Exceptions::BadCast,
-          "MueLu::TekoSmoother::Apply: Failed to cast range space to product range space.");
+    // create a Thyra RHS vector
+    Teuchos::RCP<Thyra::MultiVectorBase<Scalar> > thyB = Thyra::createMembers(Teuchos::rcp_dynamic_cast<const Thyra::VectorSpaceBase<Scalar> >(bThyOp_->productRange()), Teuchos::as<int>(B.getNumVectors()));
+    Teuchos::RCP<Thyra::ProductMultiVectorBase<Scalar> > thyProdB =
+        Teuchos::rcp_dynamic_cast<Thyra::ProductMultiVectorBase<Scalar> >(thyB);
+    TEUCHOS_TEST_FOR_EXCEPTION(thyProdB.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Apply: Failed to cast range space to product range space.");
 
-      // copy RHS vector B to Thyra::MultiVectorBase thyProdB
-      Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::updateThyra(Teuchos::rcpFromRef(B), rgMapExtractor, thyProdB);
+    // copy RHS vector B to Thyra::MultiVectorBase thyProdB
+    Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::updateThyra(Teuchos::rcpFromRef(B), rgMapExtractor, thyProdB);
 
-      // create a Thyra SOL vector
-      Teuchos::RCP<Thyra::MultiVectorBase<Scalar> > thyX = Thyra::createMembers(Teuchos::rcp_dynamic_cast<const Thyra::VectorSpaceBase<Scalar> >(bThyOp_->productDomain()),Teuchos::as<int>(X.getNumVectors()));
-      Teuchos::RCP<Thyra::ProductMultiVectorBase<Scalar> > thyProdX =
-          Teuchos::rcp_dynamic_cast<Thyra::ProductMultiVectorBase<Scalar> >(thyX);
-      TEUCHOS_TEST_FOR_EXCEPTION(thyProdX.is_null(), Exceptions::BadCast,
-          "MueLu::TekoSmoother::Apply: Failed to cast domain space to product domain space.");
+    // create a Thyra SOL vector
+    Teuchos::RCP<Thyra::MultiVectorBase<Scalar> > thyX = Thyra::createMembers(Teuchos::rcp_dynamic_cast<const Thyra::VectorSpaceBase<Scalar> >(bThyOp_->productDomain()), Teuchos::as<int>(X.getNumVectors()));
+    Teuchos::RCP<Thyra::ProductMultiVectorBase<Scalar> > thyProdX =
+        Teuchos::rcp_dynamic_cast<Thyra::ProductMultiVectorBase<Scalar> >(thyX);
+    TEUCHOS_TEST_FOR_EXCEPTION(thyProdX.is_null(), Exceptions::BadCast,
+                               "MueLu::TekoSmoother::Apply: Failed to cast domain space to product domain space.");
 
-      // copy RHS vector X to Thyra::MultiVectorBase thyProdX
-      Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::updateThyra(Teuchos::rcpFromRef(X), rgMapExtractor, thyProdX);
+    // copy RHS vector X to Thyra::MultiVectorBase thyProdX
+    Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::updateThyra(Teuchos::rcpFromRef(X), rgMapExtractor, thyProdX);
 
-      inverseOp_->apply(
+    inverseOp_->apply(
         Thyra::NOTRANS,
-        *thyB, //const MultiVectorBase<Scalar> &X,
-        thyX.ptr(), //const Ptr<MultiVectorBase<Scalar> > &Y,
+        *thyB,       // const MultiVectorBase<Scalar> &X,
+        thyX.ptr(),  // const Ptr<MultiVectorBase<Scalar> > &Y,
         1.0,
         0.0);
 
-      // copy back content of Ptr<Thyra::MultiVectorBase> thyX into X
-      Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > XX =
-            Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(thyX, comm);
-
-      X.update(Teuchos::ScalarTraits<Scalar>::one(), *XX, Teuchos::ScalarTraits<Scalar>::zero());
-    }
-    //@}
+    // copy back content of Ptr<Thyra::MultiVectorBase> thyX into X
+    Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > XX =
+        Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(thyX, comm);
 
-    RCP<SmootherPrototype> Copy() const { return Teuchos::rcp (new MueLu::TekoSmoother<double,int,GlobalOrdinal,Node> (*this)); }
+    X.update(Teuchos::ScalarTraits<Scalar>::one(), *XX, Teuchos::ScalarTraits<Scalar>::zero());
+  }
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const { return Teuchos::rcp(new MueLu::TekoSmoother<double, int, GlobalOrdinal, Node>(*this)); }
 
-    //! Return a simple one-line description of this object.
-    std::string description() const {
-      std::ostringstream out;
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-      return out.str();
-    }
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
-      MUELU_DESCRIBE;
+  //! Return a simple one-line description of this object.
+  std::string description() const {
+    std::ostringstream out;
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+    return out.str();
+  }
 
-      if (verbLevel & Parameters0)
-        out0 << "Prec. type: " << type_ << std::endl;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
+    MUELU_DESCRIBE;
 
-      if (verbLevel & Debug)
-        out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+    if (verbLevel & Parameters0)
+      out0 << "Prec. type: " << type_ << std::endl;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {size_t cplx=0; return cplx;}
+    if (verbLevel & Debug)
+      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+  }
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const {
+    size_t cplx = 0;
+    return cplx;
+  }
 
-  private:
+  //@}
 
-    //! smoother type
-    std::string type_;
+ private:
+  //! smoother type
+  std::string type_;
 
-    //! block operator
-    RCP<Matrix> A_;                  // < ! internal blocked operator "A" generated by AFact_
-    RCP<BlockedCrsMatrix> bA_;
-    RCP<const Thyra::BlockedLinearOpBase<Scalar> > bThyOp_;
+  //! block operator
+  RCP<Matrix> A_;  // < ! internal blocked operator "A" generated by AFact_
+  RCP<BlockedCrsMatrix> bA_;
+  RCP<const Thyra::BlockedLinearOpBase<Scalar> > bThyOp_;
 
-    //! Teko parameters
-    RCP<ParameterList> tekoParams_;  // < ! parameter list containing Teko parameters. These parameters are not administrated by the factory and not validated.
+  //! Teko parameters
+  RCP<ParameterList> tekoParams_;  // < ! parameter list containing Teko parameters. These parameters are not administrated by the factory and not validated.
 
-    Teko::LinearOp inverseOp_;       // < ! Teko inverse operator
-  }; // class TekoSmoother (specialization on SC=double)
-} // namespace MueLu
+  Teko::LinearOp inverseOp_;  // < ! Teko inverse operator
+};                            // class TekoSmoother (specialization on SC=double)
+}  // namespace MueLu
 
 #define MUELU_TEKOSMOOTHER_SHORT
 
-#endif // HAVE_MUELU_TEKO
+#endif  // HAVE_MUELU_TEKO
 
 #endif /* MUELU_TEKOSMOOTHER_DECL_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_def.hpp
index e2e3bb204fef..fa1bdede86fe 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_TekoSmoother_def.hpp
@@ -52,15 +52,14 @@
 #include "MueLu_TekoSmoother_decl.hpp"
 
 namespace MueLu {
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t TekoSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
-  
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t TekoSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // HAVE_MUELU_TEKO
+#endif  // HAVE_MUELU_TEKO
 
 #endif /* MUELU_TEKOSMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_decl.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_decl.hpp
index 171eb73b441a..40dec830290e 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_decl.hpp
@@ -47,17 +47,16 @@
 #ifndef MUELU_UZAWASMOOTHER_DECL_HPP_
 #define MUELU_UZAWASMOOTHER_DECL_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 
 #include <Teuchos_ParameterList.hpp>
 
-//Xpetra
+// Xpetra
 #include <Xpetra_MapExtractor_fwd.hpp>
 #include <Xpetra_MultiVectorFactory_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 
-//MueLu
+// MueLu
 #include "MueLu_UzawaSmoother_fwd.hpp"
 #include "MueLu_SmootherPrototype.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
@@ -69,118 +68,115 @@
 
 namespace MueLu {
 
-  /*!
-    @class UzawaSmoother
-    @brief Block triangle Uzawa smoother for 2x2 block matrices
+/*!
+  @class UzawaSmoother
+  @brief Block triangle Uzawa smoother for 2x2 block matrices
 
-  */
+*/
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class UzawaSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
-    typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class UzawaSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractorClass;
 
 #undef MUELU_UZAWASMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  void DeclareInput(Level &currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  /*! @brief Set factory manager for internal velocity prediction
 
-    /*! @brief Set factory manager for internal velocity prediction
-
-    \note This routine is outdated. Use AddFactoryManager instead
-    */
-    void SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager);
-
-    /*! @brief Set factory manager for internal SchurComplement handling
+  \note This routine is outdated. Use AddFactoryManager instead
+  */
+  void SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager);
 
-    \note This routine is outdated. Use AddFactoryManager instead
-    */
-    void SetSchurCompFactoryManager(RCP<FactoryManager> FactManager);
+  /*! @brief Set factory manager for internal SchurComplement handling
 
-    /*! @brief Add a factory manager at a specific position
+  \note This routine is outdated. Use AddFactoryManager instead
+  */
+  void SetSchurCompFactoryManager(RCP<FactoryManager> FactManager);
 
-    See #FactManager_ to explanation of positions.
-    */
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
+  /*! @brief Add a factory manager at a specific position
 
-    //@}
+  See #FactManager_ to explanation of positions.
+  */
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos);
 
-    //! @name Setup and Apply methods.
-    //@{
+  //@}
 
-    /*! @brief Setup routine
-     */
-    void Setup(Level &currentLevel);
+  //! @name Setup and Apply methods.
+  //@{
 
-    /*! @brief Apply the Braess Sarazin smoother.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero TODO This option has no effect.
-    */
-    void Apply(MultiVector &X, MultiVector const &B, bool InitialGuessIsZero = false) const;
-    //@}
+  /*! @brief Setup routine
+   */
+  void Setup(Level &currentLevel);
 
-    RCP<SmootherPrototype> Copy() const;
+  /*! @brief Apply the Braess Sarazin smoother.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero TODO This option has no effect.
+  */
+  void Apply(MultiVector &X, MultiVector const &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+  //@}
 
-    //! smoother type
-    std::string type_ = "Uzawa";
+ private:
+  //! smoother type
+  std::string type_ = "Uzawa";
 
-    RCP<const FactoryBase>                AFact_;              //!< A Factory
+  RCP<const FactoryBase> AFact_;  //!< A Factory
 
-    //! block operator
-    RCP<Matrix>                           A_ = Teuchos::null;  //<! internal blocked operator "A" generated by AFact_
+  //! block operator
+  RCP<Matrix> A_ = Teuchos::null;  //<! internal blocked operator "A" generated by AFact_
 
-    RCP<const MapExtractorClass>          rangeMapExtractor_;  //!< range  map extractor (from A_ generated by AFact)
-    RCP<const MapExtractorClass>          domainMapExtractor_; //!< domain map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> rangeMapExtractor_;   //!< range  map extractor (from A_ generated by AFact)
+  RCP<const MapExtractorClass> domainMapExtractor_;  //!< domain map extractor (from A_ generated by AFact)
 
-    //! matrices
-    Teuchos::RCP<Vector>                  diagFinv_;           //!< inverse diagonal of fluid operator (vector).
-    Teuchos::RCP<Matrix>                  F_;                  //!< fluid operator
-    Teuchos::RCP<Matrix>                  G_;                  //!< pressure gradient operator
-    Teuchos::RCP<Matrix>                  D_;                  //!< divergence operator
-    Teuchos::RCP<Matrix>                  Z_;                  //!< pressure stabilization term or null block
+  //! matrices
+  Teuchos::RCP<Vector> diagFinv_;  //!< inverse diagonal of fluid operator (vector).
+  Teuchos::RCP<Matrix> F_;         //!< fluid operator
+  Teuchos::RCP<Matrix> G_;         //!< pressure gradient operator
+  Teuchos::RCP<Matrix> D_;         //!< divergence operator
+  Teuchos::RCP<Matrix> Z_;         //!< pressure stabilization term or null block
 
-    Teuchos::RCP<SmootherBase>            velPredictSmoo_;     //!< smoother for velocity prediction
-    Teuchos::RCP<SmootherBase>            schurCompSmoo_;      //!< smoother for SchurComplement equation
+  Teuchos::RCP<SmootherBase> velPredictSmoo_;  //!< smoother for velocity prediction
+  Teuchos::RCP<SmootherBase> schurCompSmoo_;   //!< smoother for SchurComplement equation
 
-    /*!
-    @brief Vector of internal factory managers
+  /*!
+  @brief Vector of internal factory managers
 
-    - FactManager_[0] holds the factory manager for the predicting the primary variable
-    - FactManager_[1] stores the factory manager used for the SchurComplement correction step.
-    */
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+  - FactManager_[0] holds the factory manager for the predicting the primary variable
+  - FactManager_[1] stores the factory manager used for the SchurComplement correction step.
+  */
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-  }; // class UzawaSmoother
+};  // class UzawaSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UZAWASMOOTHER_SHORT
 
diff --git a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_def.hpp b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_def.hpp
index a9d1d98bca89..f02d1cf1efca 100644
--- a/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/BlockedSmoothers/MueLu_UzawaSmoother_def.hpp
@@ -68,275 +68,268 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set<Scalar> ("Damping factor", 1.0, "Damping/Scaling factor in SIMPLE");
-    validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of SIMPLE sweeps (default = 1)");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<Scalar>("Damping factor", 1.0, "Damping/Scaling factor in SIMPLE");
+  validParamList->set<LocalOrdinal>("Sweeps", 1, "Number of SIMPLE sweeps (default = 1)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
+  TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::UzawaSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+
+  size_t myPos = Teuchos::as<size_t>(pos);
+
+  if (myPos < FactManager_.size()) {
+    // replace existing entris in FactManager_ vector
+    FactManager_.at(myPos) = FactManager;
+  } else if (myPos == FactManager_.size()) {
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
+  } else {  // if(myPos > FactManager_.size())
+    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+
+    // add new Factory manager in the end of the vector
+    FactManager_.push_back(FactManager);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddFactoryManager(RCP<const FactoryManagerBase> FactManager, int pos) {
-    TEUCHOS_TEST_FOR_EXCEPTION(pos < 0, Exceptions::RuntimeError, "MueLu::UzawaSmoother::AddFactoryManager: parameter \'pos\' must not be negative! error.");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager) {
+  AddFactoryManager(FactManager, 0);  // overwrite factory manager for predicting the primary variable
+}
 
-    size_t myPos = Teuchos::as<size_t>(pos);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSchurCompFactoryManager(RCP<FactoryManager> FactManager) {
+  AddFactoryManager(FactManager, 1);  // overwrite factory manager for SchurComplement
+}
 
-    if (myPos < FactManager_.size()) {
-      // replace existing entris in FactManager_ vector
-      FactManager_.at(myPos) = FactManager;
-    } else if( myPos == FactManager_.size()) {
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    } else { // if(myPos > FactManager_.size())
-      RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      *out << "Warning: cannot add new FactoryManager at proper position " << pos << ". The FactoryManager is just appended to the end. Check this!" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  currentLevel.DeclareInput("A", this->GetFactory("A").get());
 
-      // add new Factory manager in the end of the vector
-      FactManager_.push_back(FactManager);
-    }
+  TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError, "MueLu::UzawaSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\". make sure that you use the same proper damping factors for omega both in the SchurComplementFactory and in the SIMPLE smoother!");
+
+  // loop over all factory managers for the subblocks of blocked operator A
+  std::vector<Teuchos::RCP<const FactoryManagerBase>>::const_iterator it;
+  for (it = FactManager_.begin(); it != FactManager_.end(); ++it) {
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), *it);
 
+    // request "Smoother" for current subblock row.
+    currentLevel.DeclareInput("PreSmoother", (*it)->GetFactory("Smoother").get());
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  FactoryMonitor m(*this, "Setup blocked Uzawa Smoother", currentLevel);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetVelocityPredictionFactoryManager(RCP<FactoryManager> FactManager) {
-    AddFactoryManager(FactManager, 0); // overwrite factory manager for predicting the primary variable
-  }
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::UzawaSmoother::Setup(): Setup() has already been called";
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSchurCompFactoryManager(RCP<FactoryManager> FactManager) {
-    AddFactoryManager(FactManager, 1); // overwrite factory manager for SchurComplement
-  }
+  // extract blocked operator A from current level
+  A_ = Factory::Get<RCP<Matrix>>(currentLevel, "A");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    currentLevel.DeclareInput("A",this->GetFactory("A").get());
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::UzawaSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(FactManager_.size() != 2, Exceptions::RuntimeError,"MueLu::UzawaSmoother::DeclareInput: You have to declare two FactoryManagers with a \"Smoother\" object: One for predicting the primary variable and one for the SchurComplement system. The smoother for the SchurComplement system needs a SchurComplementFactory as input for variable \"A\". make sure that you use the same proper damping factors for omega both in the SchurComplementFactory and in the SIMPLE smoother!");
+  // store map extractors
+  rangeMapExtractor_  = bA->getRangeMapExtractor();
+  domainMapExtractor_ = bA->getDomainMapExtractor();
 
-    // loop over all factory managers for the subblocks of blocked operator A
-    std::vector<Teuchos::RCP<const FactoryManagerBase> >::const_iterator it;
-    for(it = FactManager_.begin(); it!=FactManager_.end(); ++it) {
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel),   *it);
+  // Store the blocks in local member variables
+  F_ = bA->getMatrix(0, 0);
+  G_ = bA->getMatrix(0, 1);
+  D_ = bA->getMatrix(1, 0);
+  Z_ = bA->getMatrix(1, 1);
 
-      // request "Smoother" for current subblock row.
-      currentLevel.DeclareInput("PreSmoother",(*it)->GetFactory("Smoother").get());
-    }
+  // Set the Smoother
+  // carefully switch to the SubFactoryManagers (defined by the users)
+  {
+    RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), velpredictFactManager);
+    velPredictSmoo_ = currentLevel.Get<RCP<SmootherBase>>("PreSmoother", velpredictFactManager->GetFactory("Smoother").get());
+  }
+  {
+    RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
+    SetFactoryManager currentSFM(rcpFromRef(currentLevel), schurFactManager);
+    schurCompSmoo_ = currentLevel.Get<RCP<SmootherBase>>("PreSmoother", schurFactManager->GetFactory("Smoother").get());
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  SmootherPrototype::IsSetup(true);
+}
 
-    FactoryMonitor m(*this, "Setup blocked Uzawa Smoother", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool /* InitialGuessIsZero */) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::UzawaSmoother::Apply(): Setup() has not been called");
 
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::UzawaSmoother::Setup(): Setup() has already been called";
+  Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
 
-    // extract blocked operator A from current level
-    A_ = Factory::Get<RCP<Matrix> > (currentLevel, "A");
+  SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
 
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    TEUCHOS_TEST_FOR_EXCEPTION(bA == Teuchos::null, Exceptions::BadCast, "MueLu::UzawaSmoother::Setup: input matrix A is not of type BlockedCrsMatrix! error.");
+  bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
+  bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
 
-    // store map extractors
-    rangeMapExtractor_ = bA->getRangeMapExtractor();
-    domainMapExtractor_ = bA->getDomainMapExtractor();
+  // extract parameters from internal parameter list
+  const ParameterList &pL = Factory::GetParameterList();
+  LocalOrdinal nSweeps    = pL.get<LocalOrdinal>("Sweeps");
+  Scalar omega            = pL.get<Scalar>("Damping factor");
 
-    // Store the blocks in local member variables
-    F_ = bA->getMatrix(0, 0);
-    G_ = bA->getMatrix(0, 1);
-    D_ = bA->getMatrix(1, 0);
-    Z_ = bA->getMatrix(1, 1);
+  // wrap current solution vector in RCP
+  RCP<MultiVector> rcpX       = Teuchos::rcpFromRef(X);
+  RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
 
-    // Set the Smoother
-    // carefully switch to the SubFactoryManagers (defined by the users)
-    {
-      RCP<const FactoryManagerBase> velpredictFactManager = FactManager_.at(0);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), velpredictFactManager);
-      velPredictSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother",velpredictFactManager->GetFactory("Smoother").get());
-    }
-    {
-      RCP<const FactoryManagerBase> schurFactManager = FactManager_.at(1);
-      SetFactoryManager currentSFM  (rcpFromRef(currentLevel), schurFactManager);
-      schurCompSmoo_ = currentLevel.Get< RCP<SmootherBase> >("PreSmoother", schurFactManager->GetFactory("Smoother").get());
-    }
+  // make sure that both rcpX and rcpB are BlockedMultiVector objects
+  bool bCopyResultX        = false;
+  bool bReorderX           = false;
+  bool bReorderB           = false;
+  RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
+  RCP<BlockedMultiVector> bX       = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-    SmootherPrototype::IsSetup(true);
-  }
+  // check the type of operator
+  RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> rbA =
+      Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(bA);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */) const
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::UzawaSmoother::Apply(): Setup() has not been called");
-
-    Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
-
-    SC zero = Teuchos::ScalarTraits<SC>::zero(), one = Teuchos::ScalarTraits<SC>::one();
-
-    bool bRangeThyraMode  = rangeMapExtractor_->getThyraMode();
-    bool bDomainThyraMode = domainMapExtractor_->getThyraMode();
-
-    // extract parameters from internal parameter list
-    const ParameterList & pL = Factory::GetParameterList();
-    LocalOrdinal nSweeps = pL.get<LocalOrdinal>("Sweeps");
-    Scalar omega = pL.get<Scalar>("Damping factor");
-
-    // wrap current solution vector in RCP
-    RCP<MultiVector>       rcpX = Teuchos::rcpFromRef(X);
-    RCP<const MultiVector> rcpB = Teuchos::rcpFromRef(B);
-
-    // make sure that both rcpX and rcpB are BlockedMultiVector objects
-    bool bCopyResultX = false;
-    bool bReorderX = false;
-    bool bReorderB = false;
-    RCP<BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    MUELU_TEST_FOR_EXCEPTION(bA.is_null() == true, Exceptions::RuntimeError, "MueLu::BlockedGaussSeidelSmoother::Apply(): A_ must be a BlockedCrsMatrix");
-    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    RCP<const BlockedMultiVector> bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
-
-    // check the type of operator
-    RCP<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > rbA =
-            Teuchos::rcp_dynamic_cast<Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA);
-
-    if(rbA.is_null() == false) {
-      // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
-      // map for the construction of the blocked multivectors
-
-      // check type of X vector
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-        bReorderX = true;
-      }
-
-      // check type of B vector
-      if (bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-        bReorderB = true;
-      }
-    }
-    else {
-      // A is a BlockedCrsMatrix and uses a plain blocked map
-      if (bX.is_null() == true) {
-        RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
-        rcpX.swap(vectorWithBlockedMap);
-        bCopyResultX = true;
-      }
-
-      if(bB.is_null() == true) {
-        RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(),rcpB));
-        rcpB.swap(vectorWithBlockedMap);
-      }
+  if (rbA.is_null() == false) {
+    // A is a ReorderedBlockedCrsMatrix and thus uses nested maps, retrieve BlockedCrsMatrix and use plain blocked
+    // map for the construction of the blocked multivectors
+
+    // check type of X vector
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+      bReorderX    = true;
     }
 
-    // we now can guarantee that X and B are blocked multi vectors
-    bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
-    bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
+    // check type of B vector
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(rbA->getBlockedCrsMatrix()->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+      bReorderB = true;
+    }
+  } else {
+    // A is a BlockedCrsMatrix and uses a plain blocked map
+    if (bX.is_null() == true) {
+      RCP<MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedDomainMap(), rcpX));
+      rcpX.swap(vectorWithBlockedMap);
+      bCopyResultX = true;
+    }
 
-    // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
-    if(rbA.is_null() == false) {
+    if (bB.is_null() == true) {
+      RCP<const MultiVector> vectorWithBlockedMap = Teuchos::rcp(new BlockedMultiVector(bA->getBlockedRangeMap(), rcpB));
+      rcpB.swap(vectorWithBlockedMap);
+    }
+  }
 
-      Teuchos::RCP<const Xpetra::BlockReorderManager > brm = rbA->getBlockReorderManager();
+  // we now can guarantee that X and B are blocked multi vectors
+  bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(rcpX);
+  bB = Teuchos::rcp_dynamic_cast<const BlockedMultiVector>(rcpB);
 
-      // check type of X vector
-      if(bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
-        // X is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
-        rcpX.swap(reorderedBlockedVector);
-      }
+  // Finally we can do a reordering of the blocked multivectors if A is a ReorderedBlockedCrsMatrix
+  if (rbA.is_null() == false) {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = rbA->getBlockReorderManager();
 
-      if(bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
-        // B is a blocked multi vector but incompatible to the reordered blocked operator A
-        Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
-        rcpB.swap(reorderedBlockedVector);
-      }
+    // check type of X vector
+    if (bX->getBlockedMap()->getNumMaps() != bA->getDomainMapExtractor()->NumMaps() || bReorderX) {
+      // X is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bX);
+      rcpX.swap(reorderedBlockedVector);
     }
 
-    // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
-
-    // create residual vector
-    // contains current residual of current solution X with rhs B
-    RCP<MultiVector> residual = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
-    RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
-    Teuchos::RCP<MultiVector> r1 = bresidual->getMultiVector(0,bRangeThyraMode);
-    Teuchos::RCP<MultiVector> r2 = bresidual->getMultiVector(1,bRangeThyraMode);
-
-    // helper vector 1
-    RCP<MultiVector> xtilde     = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
-    RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
-    RCP<MultiVector> xtilde1 = bxtilde->getMultiVector(0,bDomainThyraMode);
-    RCP<MultiVector> xtilde2 = bxtilde->getMultiVector(1,bDomainThyraMode);
-
-    // incrementally improve solution vector X
-    for (LocalOrdinal run = 0; run < nSweeps; ++run) {
-      // 1) calculate current residual
-      residual->update(one,*rcpB,zero); // residual = B
-      A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
-
-      // 2) solve F * \Delta \tilde{x}_1 = r_1
-      //    start with zero guess \Delta \tilde{x}_1
-      bxtilde->putScalar(zero);
-      velPredictSmoo_->Apply(*xtilde1,*r1);
-
-      // 3) calculate rhs for SchurComp equation
-      //    r_2 - D \Delta \tilde{x}_1
-      RCP<MultiVector> schurCompRHS = MultiVectorFactory::Build(r2->getMap(),rcpB->getNumVectors());
-      D_->apply(*xtilde1,*schurCompRHS);
-      schurCompRHS->update(one,*r2,-one);
-
-      // 4) solve SchurComp equation
-      //    start with zero guess \Delta \tilde{x}_2
-      schurCompSmoo_->Apply(*xtilde2,*schurCompRHS);
-
-      rcpX->update(omega,*bxtilde,one);
-    }
-
-    if (bCopyResultX == true) {
-      RCP<MultiVector> Xmerged = bX->Merge();
-      X.update(one, *Xmerged, zero);
+    if (bB->getBlockedMap()->getNumMaps() != bA->getRangeMapExtractor()->NumMaps() || bReorderB) {
+      // B is a blocked multi vector but incompatible to the reordered blocked operator A
+      Teuchos::RCP<const MultiVector> reorderedBlockedVector = buildReorderedBlockedMultiVector(brm, bB);
+      rcpB.swap(reorderedBlockedVector);
     }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp( new UzawaSmoother(*this) );
+  // Throughout the rest of the algorithm rcpX and rcpB are used for solution vector and RHS
+
+  // create residual vector
+  // contains current residual of current solution X with rhs B
+  RCP<MultiVector> residual         = MultiVectorFactory::Build(rcpB->getMap(), rcpB->getNumVectors());
+  RCP<BlockedMultiVector> bresidual = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(residual);
+  Teuchos::RCP<MultiVector> r1      = bresidual->getMultiVector(0, bRangeThyraMode);
+  Teuchos::RCP<MultiVector> r2      = bresidual->getMultiVector(1, bRangeThyraMode);
+
+  // helper vector 1
+  RCP<MultiVector> xtilde         = MultiVectorFactory::Build(rcpX->getMap(), rcpX->getNumVectors());
+  RCP<BlockedMultiVector> bxtilde = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(xtilde);
+  RCP<MultiVector> xtilde1        = bxtilde->getMultiVector(0, bDomainThyraMode);
+  RCP<MultiVector> xtilde2        = bxtilde->getMultiVector(1, bDomainThyraMode);
+
+  // incrementally improve solution vector X
+  for (LocalOrdinal run = 0; run < nSweeps; ++run) {
+    // 1) calculate current residual
+    residual->update(one, *rcpB, zero);  // residual = B
+    A_->apply(*rcpX, *residual, Teuchos::NO_TRANS, -one, one);
+
+    // 2) solve F * \Delta \tilde{x}_1 = r_1
+    //    start with zero guess \Delta \tilde{x}_1
+    bxtilde->putScalar(zero);
+    velPredictSmoo_->Apply(*xtilde1, *r1);
+
+    // 3) calculate rhs for SchurComp equation
+    //    r_2 - D \Delta \tilde{x}_1
+    RCP<MultiVector> schurCompRHS = MultiVectorFactory::Build(r2->getMap(), rcpB->getNumVectors());
+    D_->apply(*xtilde1, *schurCompRHS);
+    schurCompRHS->update(one, *r2, -one);
+
+    // 4) solve SchurComp equation
+    //    start with zero guess \Delta \tilde{x}_2
+    schurCompSmoo_->Apply(*xtilde2, *schurCompRHS);
+
+    rcpX->update(omega, *bxtilde, one);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
+  if (bCopyResultX == true) {
+    RCP<MultiVector> Xmerged = bX->Merge();
+    X.update(one, *Xmerged, zero);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-    }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new UzawaSmoother(*this));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0) {
+    out0 << "Prec. type: " << type_ << /*" Sweeps: " << nSweeps_ << " damping: " << omega_ <<*/ std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
   }
+}
 
-} // namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t UzawaSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
+}  // namespace MueLu
 
 #endif /* MUELU_UZAWASMOOTHER_DEF_HPP_ */
diff --git a/packages/muelu/src/Smoothers/MueLu_AdvSmootherPrototype.hpp b/packages/muelu/src/Smoothers/MueLu_AdvSmootherPrototype.hpp
index c6840ea46ccd..c6c991559754 100644
--- a/packages/muelu/src/Smoothers/MueLu_AdvSmootherPrototype.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_AdvSmootherPrototype.hpp
@@ -51,63 +51,62 @@
 
 namespace MueLu {
 
-  class Level;
-
-  /*!
-    @class AdvSmootherPrototype
-
-    'Advanced Smoother prototypes' can be fully copied using the Copy() method.
-    They can also copy the parameters of
-    another smoother object of the same type (CopyParameters()). Both
-    capabilities are used by the SmootherFactory.
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class AdvSmootherPrototype : public SmootherPrototypex<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+class Level;
+
+/*!
+  @class AdvSmootherPrototype
+
+  'Advanced Smoother prototypes' can be fully copied using the Copy() method.
+  They can also copy the parameters of
+  another smoother object of the same type (CopyParameters()). Both
+  capabilities are used by the SmootherFactory.
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class AdvSmootherPrototype : public SmootherPrototypex<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_ADVSMOOTHERPROTOTYPE_HPP
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //!@nameConstructors/Destructors.
-    //@{
-    AdvSmootherPrototype()
-      : type_("undefined")
-    {}
+ public:
+  //!@nameConstructors/Destructors.
+  //@{
+  AdvSmootherPrototype()
+    : type_("undefined") {}
 
-    virtual ~AdvSmootherPrototype() {}
-    //@}
+  virtual ~AdvSmootherPrototype() {}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    virtual void CopyParameters(const AdvSmootherPrototype& smootherPrototype) = 0;
+  virtual void CopyParameters(const AdvSmootherPrototype& smootherPrototype) = 0;
 
-    //@}
+  //@}
 
-    //! @name Get/Set methods.
-    //@{
+  //! @name Get/Set methods.
+  //@{
 
-    //! Get the smoother type.
-    std::string GetType() const { return type_; }
+  //! Get the smoother type.
+  std::string GetType() const { return type_; }
 
-    /*! @brief Set the smoother type.
-      This method must be called by constructors of derived classes.
-    */
-    //TODO: remove, type_ should be const
-    void SetType(std::string & type) { type_ = type; }
+  /*! @brief Set the smoother type.
+    This method must be called by constructors of derived classes.
+  */
+  // TODO: remove, type_ should be const
+  void SetType(std::string& type) { type_ = type; }
 
-    //@}
+  //@}
 
-  private:
-    std::string type_;
+ private:
+  std::string type_;
 
-  }; //class AdvSmootherPrototype
+};  // class AdvSmootherPrototype
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ADVSMOOTHERPROTOTYPE_SHORT
 
-#endif //ifndef MUELU_ADVSMOOTHERPROTOTYPE_HPP
+#endif  // ifndef MUELU_ADVSMOOTHERPROTOTYPE_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_decl.hpp
index 720d24d4654a..addb498ffa39 100644
--- a/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_decl.hpp
@@ -59,157 +59,158 @@
 #include "MueLu_FactoryBase_fwd.hpp"
 #include "MueLu_Utilities_fwd.hpp"
 
-namespace Amesos2 { template<class OP, class MV> class Solver; }
+namespace Amesos2 {
+template <class OP, class MV>
+class Solver;
+}
 
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Projection {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Projection {
+ public:
+  Projection(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& Nullspace);
 
-  public:
+  void projectOut(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X);
 
-    Projection(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &Nullspace);
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Nullspace_;
 
-    void projectOut(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &X);
+ private:
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> localMap_;
+};
 
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Nullspace_;
+/*!
+  @class Amesos2Smoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Amesos2 direct solvers.
 
-  private:
+  This class creates an Amesos2 preconditioner factory.  The factory is capable of generating direct solvers
+  based on the type and ParameterList passed into the constructor.  See the constructor for more information.
+*/
 
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > localMap_;
-
-  };
-
-  /*!
-    @class Amesos2Smoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Amesos2 direct solvers.
-
-    This class creates an Amesos2 preconditioner factory.  The factory is capable of generating direct solvers
-    based on the type and ParameterList passed into the constructor.  See the constructor for more information.
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class Amesos2Smoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class Amesos2Smoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AMESOS2SMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! @name Constructors / destructors
-    //@{
+  /*! @brief Constructor
+    Creates a MueLu interface to the direct solvers in the Amesos2 package.
+    If you are using type=="", then either SuperLU or KLU2 are used by default.
+  */
+  Amesos2Smoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    /*! @brief Constructor
-      Creates a MueLu interface to the direct solvers in the Amesos2 package.
-      If you are using type=="", then either SuperLU or KLU2 are used by default.
-    */
-    Amesos2Smoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
+  //! Destructor
+  virtual ~Amesos2Smoother();
 
-    //! Destructor
-    virtual ~Amesos2Smoother();
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //@}
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! @name Setup and Apply methods.
-    //@{
+  /*! @brief Set up the direct solver.
+    This creates the underlying Amesos2 solver object according to the parameter list options passed into the
+    Amesos2Smoother constructor.  This includes doing a numeric factorization of the matrix.
+  */
+  void Setup(Level& currentLevel);
 
-    /*! @brief Set up the direct solver.
-      This creates the underlying Amesos2 solver object according to the parameter list options passed into the
-      Amesos2Smoother constructor.  This includes doing a numeric factorization of the matrix.
-    */
-    void Setup(Level& currentLevel);
+  /*! @brief Apply the direct solver.
+  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    /*! @brief Apply the direct solver.
-    Solves the linear system <tt>AX=B</tt> using the constructed solver.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
+  RCP<SmootherPrototype> Copy() const;
 
-    RCP<SmootherPrototype> Copy() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
-   
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    //@}
+  //@}
 
-  private:
-    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_CrsMatrix;
-    typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_MultiVector;
+ private:
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_CrsMatrix;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> Tpetra_MultiVector;
 
-    //! amesos2-specific key phrase that denote smoother type
-    std::string type_;
+  //! amesos2-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! pointer to Amesos2 solver object
-    RCP<Amesos2::Solver<Tpetra_CrsMatrix, Tpetra_MultiVector> > prec_;
+  //! pointer to Amesos2 solver object
+  RCP<Amesos2::Solver<Tpetra_CrsMatrix, Tpetra_MultiVector>> prec_;
 
-    bool useTransformation_;
-    RCP<MultiVector> X_, B_;
+  bool useTransformation_;
+  RCP<MultiVector> X_, B_;
 
-    RCP<Projection<Scalar,LocalOrdinal,GlobalOrdinal,Node>> projection_;
+  RCP<Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>> projection_;
 
-  }; // class Amesos2Smoother
+};  // class Amesos2Smoother
 
 #ifdef HAVE_MUELU_EPETRA
 
-# if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  // Stub specialization for missing Epetra template args
-  template<>
-  class Amesos2Smoother<double,int,int,Xpetra::EpetraNode> : public SmootherPrototype<double,int,int,Xpetra::EpetraNode> {
-    typedef double              Scalar;
-    typedef int                 LocalOrdinal;
-    typedef int                 GlobalOrdinal;
-    typedef Xpetra::EpetraNode  Node;
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+// Stub specialization for missing Epetra template args
+template <>
+class Amesos2Smoother<double, int, int, Xpetra::EpetraNode> : public SmootherPrototype<double, int, int, Xpetra::EpetraNode> {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
 #undef MUELU_AMESOS2SMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    Amesos2Smoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList()) {
-      MUELU_TPETRA_ETI_EXCEPTION("Amesos2Smoother<double,int,int,EpetraNode>","Amesos2Smoother<double,int,int,EpetraNode>","int");
-    }
-    virtual ~Amesos2Smoother() {};
-    void DeclareInput(Level& currentLevel) const {};
-    void Setup(Level& currentLevel) {};
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {};
-
-    RCP<SmootherPrototype> Copy() const { return Teuchos::null; };
-
-    std::string description() const { return std::string(""); };
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const {};
-   
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {size_t cplx=0; return cplx;};
+ public:
+  Amesos2Smoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList()) {
+    MUELU_TPETRA_ETI_EXCEPTION("Amesos2Smoother<double,int,int,EpetraNode>", "Amesos2Smoother<double,int,int,EpetraNode>", "int");
+  }
+  virtual ~Amesos2Smoother(){};
+  void DeclareInput(Level& currentLevel) const {};
+  void Setup(Level& currentLevel){};
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {};
+
+  RCP<SmootherPrototype> Copy() const { return Teuchos::null; };
+
+  std::string description() const { return std::string(""); };
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const {};
+
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const {
+    size_t cplx = 0;
+    return cplx;
   };
-# endif
-#endif // HAVE_MUELU_EPETRA
+};
+#endif
+#endif  // HAVE_MUELU_EPETRA
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AMESOS2SMOOTHER_SHORT
-#endif // HAVE_MUELU_AMESOS2
-#endif // MUELU_AMESOS2SMOOTHER_DECL_HPP
+#endif  // HAVE_MUELU_AMESOS2
+#endif  // MUELU_AMESOS2SMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_def.hpp
index 2dc53c5ed253..9fc5b02538bb 100644
--- a/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_Amesos2Smoother_def.hpp
@@ -63,394 +63,391 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Projection(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &Nullspace) {
-
-    localMap_ = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(Nullspace->getMap()->lib(),
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Projection(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Nullspace) {
+  localMap_ = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(Nullspace->getMap()->lib(),
                                                                            Nullspace->getNumVectors(),
-                                                                           Nullspace->getMap()->getIndexBase (),
-                                                                           Nullspace->getMap()->getComm (),
+                                                                           Nullspace->getMap()->getIndexBase(),
+                                                                           Nullspace->getMap()->getComm(),
                                                                            Xpetra::LocallyReplicated);
 
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tempMV = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(localMap_, Nullspace->getNumVectors());
-    const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    tempMV->multiply(Teuchos::CONJ_TRANS, Teuchos::NO_TRANS, ONE, *Nullspace, *Nullspace, ZERO);
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tempMV = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(localMap_, Nullspace->getNumVectors());
+  const Scalar ONE                                                                     = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar ZERO                                                                    = Teuchos::ScalarTraits<Scalar>::zero();
+  tempMV->multiply(Teuchos::CONJ_TRANS, Teuchos::NO_TRANS, ONE, *Nullspace, *Nullspace, ZERO);
 
-    Kokkos::View<Scalar**, Kokkos::LayoutLeft, Kokkos::HostSpace> Q("Q", Nullspace->getNumVectors(), Nullspace->getNumVectors());
-    int LDQ;
-    {
-      auto dots = tempMV->getHostLocalView(Xpetra::Access::ReadOnly);
-      Kokkos::deep_copy(Q, dots);
-      int strides[2];
-      Q.stride(strides);
-      LDQ = strides[1];
-    }
+  Kokkos::View<Scalar**, Kokkos::LayoutLeft, Kokkos::HostSpace> Q("Q", Nullspace->getNumVectors(), Nullspace->getNumVectors());
+  int LDQ;
+  {
+    auto dots = tempMV->getHostLocalView(Xpetra::Access::ReadOnly);
+    Kokkos::deep_copy(Q, dots);
+    int strides[2];
+    Q.stride(strides);
+    LDQ = strides[1];
+  }
 
-    Teuchos::LAPACK<LocalOrdinal,Scalar> lapack;
-    int info = 0;
-    lapack.POTRF('L', Nullspace->getNumVectors(), Q.data(), LDQ, &info);
-    TEUCHOS_ASSERT(info == 0);
-    lapack.TRTRI('L', 'N', Nullspace->getNumVectors(), Q.data(), LDQ, &info);
-    TEUCHOS_ASSERT(info == 0);
+  Teuchos::LAPACK<LocalOrdinal, Scalar> lapack;
+  int info = 0;
+  lapack.POTRF('L', Nullspace->getNumVectors(), Q.data(), LDQ, &info);
+  TEUCHOS_ASSERT(info == 0);
+  lapack.TRTRI('L', 'N', Nullspace->getNumVectors(), Q.data(), LDQ, &info);
+  TEUCHOS_ASSERT(info == 0);
 
-    Nullspace_ = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(Nullspace->getMap(), Nullspace->getNumVectors());
+  Nullspace_ = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Nullspace->getMap(), Nullspace->getNumVectors());
 
-    for (size_t i = 0; i<Nullspace->getNumVectors(); i++) {
-      for (size_t j = 0; j <= i; j++) {
-        Nullspace_->getVectorNonConst(i)->update(Q(i, j), *Nullspace->getVector(j), ONE);
-      }
+  for (size_t i = 0; i < Nullspace->getNumVectors(); i++) {
+    for (size_t j = 0; j <= i; j++) {
+      Nullspace_->getVectorNonConst(i)->update(Q(i, j), *Nullspace->getVector(j), ONE);
     }
   }
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  projectOut(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &X) {
-    const Scalar ONE = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-
-    // Project X onto orthonormal nullspace
-    // Nullspace_ ^T * X
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tempMV = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(localMap_, X.getNumVectors());
-    tempMV->multiply(Teuchos::CONJ_TRANS, Teuchos::NO_TRANS, ONE, *Nullspace_, X, ZERO);
-    auto dots = tempMV->getHostLocalView(Xpetra::Access::ReadOnly);
-    bool doProject = true;
-    for (size_t i = 0; i<X.getNumVectors(); i++) {
-      for (size_t j = 0; j<Nullspace_->getNumVectors(); j++) {
-        doProject = doProject || (Teuchos::ScalarTraits<Scalar>::magnitude(dots(j, i)) > 100*Teuchos::ScalarTraits<Scalar>::eps());
-      }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    projectOut(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X) {
+  const Scalar ONE  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
+
+  // Project X onto orthonormal nullspace
+  // Nullspace_ ^T * X
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tempMV = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(localMap_, X.getNumVectors());
+  tempMV->multiply(Teuchos::CONJ_TRANS, Teuchos::NO_TRANS, ONE, *Nullspace_, X, ZERO);
+  auto dots      = tempMV->getHostLocalView(Xpetra::Access::ReadOnly);
+  bool doProject = true;
+  for (size_t i = 0; i < X.getNumVectors(); i++) {
+    for (size_t j = 0; j < Nullspace_->getNumVectors(); j++) {
+      doProject = doProject || (Teuchos::ScalarTraits<Scalar>::magnitude(dots(j, i)) > 100 * Teuchos::ScalarTraits<Scalar>::eps());
     }
-    if (doProject) {
-      for (size_t i = 0; i<X.getNumVectors(); i++) {
-        for (size_t j = 0; j<Nullspace_->getNumVectors(); j++) {
-          X.getVectorNonConst(i)->update(-dots(j, i), *Nullspace_->getVector(j), ONE);
-        }
+  }
+  if (doProject) {
+    for (size_t i = 0; i < X.getNumVectors(); i++) {
+      for (size_t j = 0; j < Nullspace_->getNumVectors(); j++) {
+        X.getVectorNonConst(i)->update(-dots(j, i), *Nullspace_->getVector(j), ONE);
       }
     }
   }
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Amesos2Smoother(const std::string& type, const Teuchos::ParameterList& paramList)
-    : type_(type), useTransformation_(false) {
-    this->SetParameterList(paramList);
-
-    if (!type_.empty()) {
-      // Transform string to "Abcde" notation
-      std::transform(type_.begin(),   type_.end(),   type_.begin(), ::tolower);
-      std::transform(type_.begin(), ++type_.begin(), type_.begin(), ::toupper);
-    }
-    if (type_ == "Superlu_dist")
-      type_ = "Superludist";
-
-    // Try to come up with something availble
-    // Order corresponds to our preference
-    // TODO: It would be great is Amesos2 provides directly this kind of logic for us
-    if (type_ == "" || Amesos2::query(type_) == false) {
-      std::string oldtype = type_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Amesos2Smoother(const std::string& type, const Teuchos::ParameterList& paramList)
+  : type_(type)
+  , useTransformation_(false) {
+  this->SetParameterList(paramList);
+
+  if (!type_.empty()) {
+    // Transform string to "Abcde" notation
+    std::transform(type_.begin(), type_.end(), type_.begin(), ::tolower);
+    std::transform(type_.begin(), ++type_.begin(), type_.begin(), ::toupper);
+  }
+  if (type_ == "Superlu_dist")
+    type_ = "Superludist";
+
+  // Try to come up with something availble
+  // Order corresponds to our preference
+  // TODO: It would be great is Amesos2 provides directly this kind of logic for us
+  if (type_ == "" || Amesos2::query(type_) == false) {
+    std::string oldtype = type_;
 #if defined(HAVE_AMESOS2_SUPERLU)
-      type_ = "Superlu";
+    type_ = "Superlu";
 #elif defined(HAVE_AMESOS2_KLU2)
-      type_ = "Klu";
+    type_ = "Klu";
 #elif defined(HAVE_AMESOS2_SUPERLUDIST)
-      type_ = "Superludist";
+    type_ = "Superludist";
 #elif defined(HAVE_AMESOS2_BASKER)
-      type_ = "Basker";
+    type_ = "Basker";
 #else
-      this->declareConstructionOutcome(true, std::string("Amesos2 has been compiled without SuperLU_DIST, SuperLU, Klu, or Basker. By default, MueLu tries") +
-                                       "to use one of these libraries. Amesos2 must be compiled with one of these solvers, " +
-                                       "or a valid Amesos2 solver has to be specified explicitly.");
-      return;
+    this->declareConstructionOutcome(true, std::string("Amesos2 has been compiled without SuperLU_DIST, SuperLU, Klu, or Basker. By default, MueLu tries") +
+                                               "to use one of these libraries. Amesos2 must be compiled with one of these solvers, " +
+                                               "or a valid Amesos2 solver has to be specified explicitly.");
+    return;
 #endif
-      if (oldtype != "")
-        this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother: \"" << oldtype << "\" is not available. Using \"" << type_ << "\" instead" << std::endl;
-      else
-        this->GetOStream(Runtime1) << "MueLu::Amesos2Smoother: using \"" << type_ << "\"" << std::endl;
-    }
-
-    // Check the validity of the solver type parameter
-    this->declareConstructionOutcome(Amesos2::query(type_) == false, "The Amesos2 library reported that the solver '" + type_ + "' is not available. " +
-                                     "Amesos2 has been compiled without the support of this solver, or the solver name is misspelled.");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~Amesos2Smoother() { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("A",         null, "Factory of the coarse matrix");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace", null, "Factory of the nullspace");
-    validParamList->set<bool>("fix nullspace", false, "Remove zero eigenvalue by adding rank one correction.");
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("Amesos2", norecurse, "Parameters that are passed to Amesos2");
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    ParameterList pL = this->GetParameterList();
-
-    this->Input(currentLevel, "A");
-    if (pL.get<bool>("fix nullspace"))
-      this->Input(currentLevel, "Nullspace");
+    if (oldtype != "")
+      this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother: \"" << oldtype << "\" is not available. Using \"" << type_ << "\" instead" << std::endl;
+    else
+      this->GetOStream(Runtime1) << "MueLu::Amesos2Smoother: using \"" << type_ << "\"" << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+  // Check the validity of the solver type parameter
+  this->declareConstructionOutcome(Amesos2::query(type_) == false, "The Amesos2 library reported that the solver '" + type_ + "' is not available. " +
+                                                                       "Amesos2 has been compiled without the support of this solver, or the solver name is misspelled.");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~Amesos2Smoother() {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Factory of the coarse matrix");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", null, "Factory of the nullspace");
+  validParamList->set<bool>("fix nullspace", false, "Remove zero eigenvalue by adding rank one correction.");
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("Amesos2", norecurse, "Parameters that are passed to Amesos2");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  ParameterList pL = this->GetParameterList();
+
+  this->Input(currentLevel, "A");
+  if (pL.get<bool>("fix nullspace"))
+    this->Input(currentLevel, "Nullspace");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother::Setup(): Setup() has already been called" << std::endl;
+
+  RCP<Matrix> A = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+
+  // Do a quick check if we need to modify the matrix
+  RCP<const Map> rowMap = A->getRowMap();
+  RCP<Matrix> factorA;
+  Teuchos::ParameterList pL = this->GetParameterList();
+
+  if (pL.get<bool>("fix nullspace")) {
+    this->GetOStream(Runtime1) << "MueLu::Amesos2Smoother::Setup(): fixing nullspace" << std::endl;
+
+    rowMap            = A->getRowMap();
+    size_t gblNumCols = rowMap->getGlobalNumElements();
+
+    RCP<MultiVector> NullspaceOrig = Factory::Get<RCP<MultiVector> >(currentLevel, "Nullspace");
+
+    projection_                = rcp(new Projection<Scalar, LocalOrdinal, GlobalOrdinal, Node>(NullspaceOrig));
+    RCP<MultiVector> Nullspace = projection_->Nullspace_;
+
+    RCP<MultiVector> ghostedNullspace;
+    RCP<const Map> colMap;
+    RCP<const Import> importer;
+    if (rowMap->getComm()->getSize() > 1) {
+      this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother::Setup(): Applying nullspace fix on distributed matrix. Try rebalancing to single rank!" << std::endl;
+      ArrayRCP<GO> elements_RCP;
+      elements_RCP.resize(gblNumCols);
+      ArrayView<GO> elements = elements_RCP();
+      for (size_t k = 0; k < gblNumCols; k++)
+        elements[k] = Teuchos::as<GO>(k);
+      colMap           = MapFactory::Build(rowMap->lib(), gblNumCols * rowMap->getComm()->getSize(), elements, Teuchos::ScalarTraits<GO>::zero(), rowMap->getComm());
+      importer         = ImportFactory::Build(rowMap, colMap);
+      ghostedNullspace = MultiVectorFactory::Build(colMap, Nullspace->getNumVectors());
+      ghostedNullspace->doImport(*Nullspace, *importer, Xpetra::INSERT);
+    } else {
+      ghostedNullspace = Nullspace;
+      colMap           = rowMap;
+    }
 
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother::Setup(): Setup() has already been called" << std::endl;
+    using ATS         = Kokkos::ArithTraits<SC>;
+    using impl_Scalar = typename ATS::val_type;
+    using impl_ATS    = Kokkos::ArithTraits<impl_Scalar>;
+    using range_type  = Kokkos::RangePolicy<LO, typename NO::execution_space>;
 
-    RCP<Matrix> A = Factory::Get< RCP<Matrix> >(currentLevel, "A");
+    typedef typename Matrix::local_matrix_type KCRS;
+    typedef typename KCRS::StaticCrsGraphType graph_t;
+    typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+    typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+    typedef typename KCRS::values_type::non_const_type scalar_view_t;
 
-    // Do a quick check if we need to modify the matrix
-    RCP<const Map> rowMap = A->getRowMap();
-    RCP<Matrix> factorA;
-    Teuchos::ParameterList pL = this->GetParameterList();
+    const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
 
-    if (pL.get<bool>("fix nullspace")) {
-      this->GetOStream(Runtime1) << "MueLu::Amesos2Smoother::Setup(): fixing nullspace" << std::endl;
-
-      rowMap = A->getRowMap();
-      size_t gblNumCols = rowMap->getGlobalNumElements();
-
-      RCP<MultiVector> NullspaceOrig = Factory::Get< RCP<MultiVector> >(currentLevel, "Nullspace");
-
-      projection_ = rcp(new Projection<Scalar,LocalOrdinal,GlobalOrdinal,Node>(NullspaceOrig));
-      RCP<MultiVector> Nullspace = projection_->Nullspace_;
-
-      RCP<MultiVector> ghostedNullspace;
-      RCP<const Map> colMap;
-      RCP<const Import> importer;
-      if (rowMap->getComm()->getSize() > 1) {
-        this->GetOStream(Warnings0) << "MueLu::Amesos2Smoother::Setup(): Applying nullspace fix on distributed matrix. Try rebalancing to single rank!" << std::endl;
-        ArrayRCP<GO> elements_RCP;
-        elements_RCP.resize(gblNumCols);
-        ArrayView<GO> elements = elements_RCP();
-        for (size_t k = 0; k<gblNumCols; k++)
-          elements[k] = Teuchos::as<GO>(k);
-        colMap = MapFactory::Build(rowMap->lib(),gblNumCols*rowMap->getComm()->getSize(),elements,Teuchos::ScalarTraits<GO>::zero(),rowMap->getComm());
-        importer = ImportFactory::Build(rowMap,colMap);
-        ghostedNullspace = MultiVectorFactory::Build(colMap, Nullspace->getNumVectors());
-        ghostedNullspace->doImport(*Nullspace,*importer,Xpetra::INSERT);
-      } else {
-        ghostedNullspace = Nullspace;
-        colMap = rowMap;
-      }
+    size_t lclNumRows       = rowMap->getLocalNumElements();
+    LocalOrdinal lclNumCols = Teuchos::as<LocalOrdinal>(gblNumCols);
+    lno_view_t newRowPointers("newRowPointers", lclNumRows + 1);
+    lno_nnz_view_t newColIndices("newColIndices", lclNumRows * gblNumCols);
+    scalar_view_t newValues("newValues", lclNumRows * gblNumCols);
 
-      using ATS        = Kokkos::ArithTraits<SC>;
-      using impl_Scalar = typename ATS::val_type;
-      using impl_ATS = Kokkos::ArithTraits<impl_Scalar>;
-      using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-
-      typedef typename Matrix::local_matrix_type KCRS;
-      typedef typename KCRS::StaticCrsGraphType graph_t;
-      typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-      typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-      typedef typename KCRS::values_type::non_const_type scalar_view_t;
-
-      const impl_Scalar impl_SC_ZERO = impl_ATS::zero();
-
-      size_t         lclNumRows = rowMap->getLocalNumElements();
-      LocalOrdinal   lclNumCols = Teuchos::as<LocalOrdinal>(gblNumCols);
-      lno_view_t     newRowPointers("newRowPointers", lclNumRows+1);
-      lno_nnz_view_t newColIndices("newColIndices", lclNumRows*gblNumCols);
-      scalar_view_t  newValues("newValues", lclNumRows*gblNumCols);
-
-      impl_Scalar shift;
-      {
-        RCP<Vector> diag = VectorFactory::Build(A->getRowMap());
-        A->getLocalDiagCopy(*diag);
-        shift = diag->normInf();
-      }
+    impl_Scalar shift;
+    {
+      RCP<Vector> diag = VectorFactory::Build(A->getRowMap());
+      A->getLocalDiagCopy(*diag);
+      shift = diag->normInf();
+    }
 
-      // form normalization * nullspace * nullspace^T
-      {
-        auto lclNullspace = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
-        auto lclGhostedNullspace = ghostedNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
-        Kokkos::parallel_for("MueLu:Amesos2Smoother::fixNullspace_1", range_type(0, lclNumRows+1),
-                             KOKKOS_LAMBDA(const size_t i) {
-                               if (i<lclNumRows) {
-                                 newRowPointers(i) = i*gblNumCols;
-                                 for (LocalOrdinal j = 0; j < lclNumCols; j++) {
-                                   newColIndices(i*gblNumCols+j) = j;
-                                   newValues(i*gblNumCols+j) = impl_SC_ZERO;
-                                   for (size_t I = 0; I < lclNullspace.extent(1); I++)
-                                     for (size_t J = 0; J < lclGhostedNullspace.extent(1); J++)
-                                       newValues(i*gblNumCols+j) += shift * lclNullspace(i, I)*impl_ATS::conjugate(lclGhostedNullspace(j, J));
-                                 }
-                               } else
-                                 newRowPointers(lclNumRows) = lclNumRows*gblNumCols;
-                             });
-      }
+    // form normalization * nullspace * nullspace^T
+    {
+      auto lclNullspace        = Nullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
+      auto lclGhostedNullspace = ghostedNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
+      Kokkos::parallel_for(
+          "MueLu:Amesos2Smoother::fixNullspace_1", range_type(0, lclNumRows + 1),
+          KOKKOS_LAMBDA(const size_t i) {
+            if (i < lclNumRows) {
+              newRowPointers(i) = i * gblNumCols;
+              for (LocalOrdinal j = 0; j < lclNumCols; j++) {
+                newColIndices(i * gblNumCols + j) = j;
+                newValues(i * gblNumCols + j)     = impl_SC_ZERO;
+                for (size_t I = 0; I < lclNullspace.extent(1); I++)
+                  for (size_t J = 0; J < lclGhostedNullspace.extent(1); J++)
+                    newValues(i * gblNumCols + j) += shift * lclNullspace(i, I) * impl_ATS::conjugate(lclGhostedNullspace(j, J));
+              }
+            } else
+              newRowPointers(lclNumRows) = lclNumRows * gblNumCols;
+          });
+    }
 
-      // add A
-      if (colMap->lib() == Xpetra::UseTpetra) {
-        auto lclA = A->getLocalMatrixDevice();
-        auto lclColMapA = A->getColMap()->getLocalMap();
-        auto lclColMapANew = colMap->getLocalMap();
-        Kokkos::parallel_for("MueLu:Amesos2Smoother::fixNullspace_2", range_type(0, lclNumRows),
-                             KOKKOS_LAMBDA(const size_t i) {
-                               for (size_t jj = lclA.graph.row_map(i); jj < lclA.graph.row_map(i+1); jj++) {
-                                 LO j = lclColMapANew.getLocalElement(lclColMapA.getGlobalElement(lclA.graph.entries(jj)));
-                                 impl_Scalar v = lclA.values(jj);
-                                 newValues(i*gblNumCols+j) += v;
-                               }
-                             });
-      } else {
-        auto lclA = A->getLocalMatrixHost();
-        for (size_t i = 0; i<lclNumRows; i++) {
-          for (size_t jj = lclA.graph.row_map(i); jj < lclA.graph.row_map(i+1); jj++) {
-            LO j = colMap->getLocalElement(A->getColMap()->getGlobalElement(lclA.graph.entries(jj)));
-            SC v = lclA.values(jj);
-            newValues(i*gblNumCols+j) += v;
-          }
+    // add A
+    if (colMap->lib() == Xpetra::UseTpetra) {
+      auto lclA          = A->getLocalMatrixDevice();
+      auto lclColMapA    = A->getColMap()->getLocalMap();
+      auto lclColMapANew = colMap->getLocalMap();
+      Kokkos::parallel_for(
+          "MueLu:Amesos2Smoother::fixNullspace_2", range_type(0, lclNumRows),
+          KOKKOS_LAMBDA(const size_t i) {
+            for (size_t jj = lclA.graph.row_map(i); jj < lclA.graph.row_map(i + 1); jj++) {
+              LO j          = lclColMapANew.getLocalElement(lclColMapA.getGlobalElement(lclA.graph.entries(jj)));
+              impl_Scalar v = lclA.values(jj);
+              newValues(i * gblNumCols + j) += v;
+            }
+          });
+    } else {
+      auto lclA = A->getLocalMatrixHost();
+      for (size_t i = 0; i < lclNumRows; i++) {
+        for (size_t jj = lclA.graph.row_map(i); jj < lclA.graph.row_map(i + 1); jj++) {
+          LO j = colMap->getLocalElement(A->getColMap()->getGlobalElement(lclA.graph.entries(jj)));
+          SC v = lclA.values(jj);
+          newValues(i * gblNumCols + j) += v;
         }
       }
-
-      RCP<Matrix>    newA    = rcp(new CrsMatrixWrap(rowMap, colMap, 0));
-      RCP<CrsMatrix> newAcrs = rcp_dynamic_cast<CrsMatrixWrap>(newA)->getCrsMatrix();
-      newAcrs->setAllValues(newRowPointers, newColIndices, newValues);
-      newAcrs->expertStaticFillComplete(A->getDomainMap(), A->getRangeMap(),
-                                        importer, A->getCrsGraph()->getExporter());
-
-      factorA = newA;
-      rowMap = factorA->getRowMap();
-    } else {
-      factorA = A;
     }
 
-    RCP<Tpetra_CrsMatrix> tA = Utilities::Op2NonConstTpetraCrs(factorA);
-
-    prec_ = Amesos2::create<Tpetra_CrsMatrix,Tpetra_MultiVector>(type_, tA);
-    TEUCHOS_TEST_FOR_EXCEPTION(prec_ == Teuchos::null, Exceptions::RuntimeError, "Amesos2::create returns Teuchos::null");
-    RCP<Teuchos::ParameterList> amesos2_params = Teuchos::rcpFromRef(pL.sublist("Amesos2"));
-    amesos2_params->setName("Amesos2");
-    if ((rowMap->getGlobalNumElements() != as<size_t>((rowMap->getMaxAllGlobalIndex() - rowMap->getMinAllGlobalIndex())+1)) ||
-        (!rowMap->isContiguous() && (rowMap->getComm()->getSize() == 1))) {
-      if (!(amesos2_params->sublist(prec_->name()).template isType<bool>("IsContiguous")))
-        amesos2_params->sublist(prec_->name()).set("IsContiguous", false, "Are GIDs Contiguous");
-    }
-    prec_->setParameters(amesos2_params);
+    RCP<Matrix> newA       = rcp(new CrsMatrixWrap(rowMap, colMap, 0));
+    RCP<CrsMatrix> newAcrs = rcp_dynamic_cast<CrsMatrixWrap>(newA)->getCrsMatrix();
+    newAcrs->setAllValues(newRowPointers, newColIndices, newValues);
+    newAcrs->expertStaticFillComplete(A->getDomainMap(), A->getRangeMap(),
+                                      importer, A->getCrsGraph()->getExporter());
 
-    SmootherPrototype::IsSetup(true);
+    factorA = newA;
+    rowMap  = factorA->getRowMap();
+  } else {
+    factorA = A;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::Amesos2Smoother::Apply(): Setup() has not been called");
-
-    RCP<Tpetra_MultiVector> tX, tB;
-    if (!useTransformation_) {
-      tX = Utilities::MV2NonConstTpetraMV2(X);
-      tB = Utilities::MV2NonConstTpetraMV2(const_cast<MultiVector&>(B));
-    } else {
-      // Copy data of the original vectors into the transformed ones
-      size_t numVectors = X.getNumVectors();
-      size_t length     = X.getLocalLength();
-
-      TEUCHOS_TEST_FOR_EXCEPTION(numVectors > 1, Exceptions::RuntimeError,
-        "MueLu::Amesos2Smoother::Apply: Fixing coarse matrix for Amesos2 for multivectors has not been implemented yet.");
-      ArrayRCP<const SC> Xdata  = X.  getData(0),         Bdata  = B.  getData(0);
-      ArrayRCP<SC>       X_data = X_->getDataNonConst(0), B_data = B_->getDataNonConst(0);
-
-      for (size_t i = 0; i < length; i++) {
-        X_data[i] = Xdata[i];
-        B_data[i] = Bdata[i];
-      }
+  RCP<Tpetra_CrsMatrix> tA = Utilities::Op2NonConstTpetraCrs(factorA);
 
-      tX = Utilities::MV2NonConstTpetraMV2(*X_);
-      tB = Utilities::MV2NonConstTpetraMV2(*B_);
+  prec_ = Amesos2::create<Tpetra_CrsMatrix, Tpetra_MultiVector>(type_, tA);
+  TEUCHOS_TEST_FOR_EXCEPTION(prec_ == Teuchos::null, Exceptions::RuntimeError, "Amesos2::create returns Teuchos::null");
+  RCP<Teuchos::ParameterList> amesos2_params = Teuchos::rcpFromRef(pL.sublist("Amesos2"));
+  amesos2_params->setName("Amesos2");
+  if ((rowMap->getGlobalNumElements() != as<size_t>((rowMap->getMaxAllGlobalIndex() - rowMap->getMinAllGlobalIndex()) + 1)) ||
+      (!rowMap->isContiguous() && (rowMap->getComm()->getSize() == 1))) {
+    if (!(amesos2_params->sublist(prec_->name()).template isType<bool>("IsContiguous")))
+      amesos2_params->sublist(prec_->name()).set("IsContiguous", false, "Are GIDs Contiguous");
+  }
+  prec_->setParameters(amesos2_params);
+
+  SmootherPrototype::IsSetup(true);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::Amesos2Smoother::Apply(): Setup() has not been called");
+
+  RCP<Tpetra_MultiVector> tX, tB;
+  if (!useTransformation_) {
+    tX = Utilities::MV2NonConstTpetraMV2(X);
+    tB = Utilities::MV2NonConstTpetraMV2(const_cast<MultiVector&>(B));
+  } else {
+    // Copy data of the original vectors into the transformed ones
+    size_t numVectors = X.getNumVectors();
+    size_t length     = X.getLocalLength();
+
+    TEUCHOS_TEST_FOR_EXCEPTION(numVectors > 1, Exceptions::RuntimeError,
+                               "MueLu::Amesos2Smoother::Apply: Fixing coarse matrix for Amesos2 for multivectors has not been implemented yet.");
+    ArrayRCP<const SC> Xdata = X.getData(0), Bdata = B.getData(0);
+    ArrayRCP<SC> X_data = X_->getDataNonConst(0), B_data = B_->getDataNonConst(0);
+
+    for (size_t i = 0; i < length; i++) {
+      X_data[i] = Xdata[i];
+      B_data[i] = Bdata[i];
     }
 
-    prec_->setX(tX);
-    prec_->setB(tB);
+    tX = Utilities::MV2NonConstTpetraMV2(*X_);
+    tB = Utilities::MV2NonConstTpetraMV2(*B_);
+  }
 
-    prec_->solve();
+  prec_->setX(tX);
+  prec_->setB(tB);
 
-    prec_->setX(Teuchos::null);
-    prec_->setB(Teuchos::null);
+  prec_->solve();
 
-    if (useTransformation_) {
-      // Copy data from the transformed vectors into the original ones
-      size_t length     = X.getLocalLength();
+  prec_->setX(Teuchos::null);
+  prec_->setB(Teuchos::null);
 
-      ArrayRCP<SC>       Xdata  = X.  getDataNonConst(0);
-      ArrayRCP<const SC> X_data = X_->getData(0);
+  if (useTransformation_) {
+    // Copy data from the transformed vectors into the original ones
+    size_t length = X.getLocalLength();
 
-      for (size_t i = 0; i < length; i++)
-        Xdata[i] = X_data[i];
-    }
+    ArrayRCP<SC> Xdata        = X.getDataNonConst(0);
+    ArrayRCP<const SC> X_data = X_->getData(0);
 
-    {
-      Teuchos::ParameterList pL = this->GetParameterList();
-      if (pL.get<bool>("fix nullspace")) {
-        projection_->projectOut(X);
-      }
-    }
+    for (size_t i = 0; i < length; i++)
+      Xdata[i] = X_data[i];
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Copy() const
   {
-    return rcp (new Amesos2Smoother (*this));
+    Teuchos::ParameterList pL = this->GetParameterList();
+    if (pL.get<bool>("fix nullspace")) {
+      projection_->projectOut(X);
+    }
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-
-    if (SmootherPrototype::IsSetup() == true) {
-      out << prec_->description();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Copy() const {
+  return rcp(new Amesos2Smoother(*this));
+}
 
-    } else {
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-    }
-    return out.str();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+  if (SmootherPrototype::IsSetup() == true) {
+    out << prec_->description();
 
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
+  } else {
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+  }
+  return out.str();
+}
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    if ((verbLevel & External) && prec_ != Teuchos::null) {
-      Teuchos::OSTab tab2(out);
-      out << *prec_ << std::endl;
-    }
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
 
-    if (verbLevel & Debug)
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "RCP<prec_>: " << prec_ << std::endl;
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    if(!prec_.is_null())
-      return prec_->getStatus().getNnzLU();
-    else
-      return 0.0;
-
+  if ((verbLevel & External) && prec_ != Teuchos::null) {
+    Teuchos::OSTab tab2(out);
+    out << *prec_ << std::endl;
   }
-} // namespace MueLu
 
-#endif // HAVE_MUELU_AMESOS2
-#endif // MUELU_AMESOS2SMOOTHER_DEF_HPP
+  if (verbLevel & Debug)
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "RCP<prec_>: " << prec_ << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t Amesos2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  if (!prec_.is_null())
+    return prec_->getStatus().getNnzLU();
+  else
+    return 0.0;
+}
+}  // namespace MueLu
+
+#endif  // HAVE_MUELU_AMESOS2
+#endif  // MUELU_AMESOS2SMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.cpp b/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.cpp
index 1afe832916bf..576a4dca6365 100644
--- a/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.cpp
+++ b/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.cpp
@@ -63,161 +63,158 @@
 
 namespace MueLu {
 
-  template <class Node>
-  AmesosSmoother<Node>::AmesosSmoother(const std::string& type, const Teuchos::ParameterList& paramList)
-    : type_(type) {
-    this->SetParameterList(paramList);
-
-    if (!type_.empty()) {
-      // Transform string to "Abcde" notation
-      std::transform(type_.begin(),   type_.end(),   type_.begin(), ::tolower);
-      std::transform(type_.begin(), ++type_.begin(), type_.begin(), ::toupper);
-    }
-    if (type_ == "Amesos_klu")      type_ = "Klu";
-    if (type_ == "Klu2")            type_ = "Klu";
-    if (type_ == "Amesos_umfpack")  type_ = "Umfpack";
-    if (type_ == "Superlu_dist")    type_ = "Superludist";
-    if (type_ == "Amesos_mumps")    type_ = "Mumps";
-
-    // Try to come up with something availble
-    // Order corresponds to our preference
-    // TODO: It would be great is Amesos provides directly this kind of logic for us
-    std::string oldtype = type_;
-    if (type_ == "" || Amesos().Query(type_) == false) {
+template <class Node>
+AmesosSmoother<Node>::AmesosSmoother(const std::string& type, const Teuchos::ParameterList& paramList)
+  : type_(type) {
+  this->SetParameterList(paramList);
+
+  if (!type_.empty()) {
+    // Transform string to "Abcde" notation
+    std::transform(type_.begin(), type_.end(), type_.begin(), ::tolower);
+    std::transform(type_.begin(), ++type_.begin(), type_.begin(), ::toupper);
+  }
+  if (type_ == "Amesos_klu") type_ = "Klu";
+  if (type_ == "Klu2") type_ = "Klu";
+  if (type_ == "Amesos_umfpack") type_ = "Umfpack";
+  if (type_ == "Superlu_dist") type_ = "Superludist";
+  if (type_ == "Amesos_mumps") type_ = "Mumps";
+
+  // Try to come up with something availble
+  // Order corresponds to our preference
+  // TODO: It would be great is Amesos provides directly this kind of logic for us
+  std::string oldtype = type_;
+  if (type_ == "" || Amesos().Query(type_) == false) {
 #if defined(HAVE_AMESOS_SUPERLU)
-      type_ = "Superlu";
+    type_ = "Superlu";
 #elif defined(HAVE_AMESOS_KLU)
-      type_ = "Klu";
+    type_ = "Klu";
 #elif defined(HAVE_AMESOS_SUPERLUDIST)
-      type_ = "Superludist";
+    type_ = "Superludist";
 #elif defined(HAVE_AMESOS_UMFPACK)
-      type_ = "Umfpack";
+    type_ = "Umfpack";
 #else
-      this->declareConstructionOutcome(true, "Amesos has been compiled without SuperLU_DIST, SuperLU, Umfpack or Klu. By default, MueLu tries" +
-                                       "to use one of these libraries. Amesos must be compiled with one of these solvers,  " +
-                                       "or a valid Amesos solver has to be specified explicitly.");
-      return;
+    this->declareConstructionOutcome(true, "Amesos has been compiled without SuperLU_DIST, SuperLU, Umfpack or Klu. By default, MueLu tries" +
+                                               "to use one of these libraries. Amesos must be compiled with one of these solvers,  " +
+                                               "or a valid Amesos solver has to be specified explicitly.");
+    return;
 #endif
-      if (oldtype != "")
-        this->GetOStream(Warnings0) << "MueLu::AmesosSmoother: \"" << oldtype << "\" is not available. Using \"" << type_ << "\" instead" << std::endl;
-      else
-        this->GetOStream(Runtime1) << "MueLu::AmesosSmoother: using \"" << type_ << "\"" << std::endl;
-    }
-    this->declareConstructionOutcome(false, "");
+    if (oldtype != "")
+      this->GetOStream(Warnings0) << "MueLu::AmesosSmoother: \"" << oldtype << "\" is not available. Using \"" << type_ << "\" instead" << std::endl;
+    else
+      this->GetOStream(Runtime1) << "MueLu::AmesosSmoother: using \"" << type_ << "\"" << std::endl;
   }
+  this->declareConstructionOutcome(false, "");
+}
 
-  template <class Node>
-  void AmesosSmoother<Node>::DeclareInput(Level &currentLevel) const {
-    this->Input(currentLevel, "A");
-  }
+template <class Node>
+void AmesosSmoother<Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
+}
 
-  template <class Node>
-  void AmesosSmoother<Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+template <class Node>
+void AmesosSmoother<Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
 
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::AmesosSmoother::Setup(): Setup() has already been called" << std::endl;
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::AmesosSmoother::Setup(): Setup() has already been called" << std::endl;
 
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A");
+  A_ = Factory::Get<RCP<Matrix> >(currentLevel, "A");
 
-    RCP<Epetra_CrsMatrix> epA = Utilities::Op2NonConstEpetraCrs(A_);
-    linearProblem_ = rcp( new Epetra_LinearProblem() );
-    linearProblem_->SetOperator(epA.get());
+  RCP<Epetra_CrsMatrix> epA = Utilities::Op2NonConstEpetraCrs(A_);
+  linearProblem_            = rcp(new Epetra_LinearProblem());
+  linearProblem_->SetOperator(epA.get());
 
-    Amesos factory;
-    prec_ = rcp(factory.Create(type_, *linearProblem_));
-    TEUCHOS_TEST_FOR_EXCEPTION(prec_ == Teuchos::null, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Setup(): Solver '" + type_ + "' not supported by Amesos");
+  Amesos factory;
+  prec_ = rcp(factory.Create(type_, *linearProblem_));
+  TEUCHOS_TEST_FOR_EXCEPTION(prec_ == Teuchos::null, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Setup(): Solver '" + type_ + "' not supported by Amesos");
 
-    // set Reindex flag, if A is distributed with non-contiguous maps
-    // unfortunately there is no reindex for Amesos2, yet. So, this only works for Epetra based problems
-    if (A_->getRowMap()->isDistributed() == true && A_->getRowMap()->isContiguous() == false)
-      const_cast<ParameterList&>(this->GetParameterList()).set("Reindex", true);
+  // set Reindex flag, if A is distributed with non-contiguous maps
+  // unfortunately there is no reindex for Amesos2, yet. So, this only works for Epetra based problems
+  if (A_->getRowMap()->isDistributed() == true && A_->getRowMap()->isContiguous() == false)
+    const_cast<ParameterList&>(this->GetParameterList()).set("Reindex", true);
 
-    const ParameterList& paramList = this->GetParameterList();
-    RCP<ParameterList> precList = this->RemoveFactoriesFromList(paramList);
+  const ParameterList& paramList = this->GetParameterList();
+  RCP<ParameterList> precList    = this->RemoveFactoriesFromList(paramList);
 
-    prec_->SetParameters(*precList);
+  prec_->SetParameters(*precList);
 
-    const_cast<ParameterList&>(paramList).setParameters(*precList);
+  const_cast<ParameterList&>(paramList).setParameters(*precList);
 
-    int r = prec_->NumericFactorization();
-    TEUCHOS_TEST_FOR_EXCEPTION(r != 0, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Setup(): Amesos solver returns value of " +
-                               Teuchos::toString(r) + " during NumericFactorization()");
+  int r = prec_->NumericFactorization();
+  TEUCHOS_TEST_FOR_EXCEPTION(r != 0, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Setup(): Amesos solver returns value of " + Teuchos::toString(r) + " during NumericFactorization()");
 
-    SmootherPrototype::IsSetup(true);
-  }
+  SmootherPrototype::IsSetup(true);
+}
 
-  template <class Node>
-  void AmesosSmoother<Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Apply(): Setup() has not been called");
+template <class Node>
+void AmesosSmoother<Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Apply(): Setup() has not been called");
 
-    Epetra_MultiVector &epX = Utilities::MV2NonConstEpetraMV(X);
-    Epetra_MultiVector const &epB = Utilities::MV2EpetraMV(B);
-    //Epetra_LinearProblem takes the right-hand side as a non-const pointer.
-    //I think this const_cast is safe because Amesos won't modify the rhs.
-    Epetra_MultiVector &nonconstB = const_cast<Epetra_MultiVector&>(epB);
+  Epetra_MultiVector& epX       = Utilities::MV2NonConstEpetraMV(X);
+  Epetra_MultiVector const& epB = Utilities::MV2EpetraMV(B);
+  // Epetra_LinearProblem takes the right-hand side as a non-const pointer.
+  // I think this const_cast is safe because Amesos won't modify the rhs.
+  Epetra_MultiVector& nonconstB = const_cast<Epetra_MultiVector&>(epB);
 
-    linearProblem_->SetLHS(&epX);
-    linearProblem_->SetRHS(&nonconstB);
+  linearProblem_->SetLHS(&epX);
+  linearProblem_->SetRHS(&nonconstB);
 
-    prec_->Solve();
+  prec_->Solve();
 
-    // Don't keep pointers to our vectors in the Epetra_LinearProblem.
-    linearProblem_->SetLHS(0);
-    linearProblem_->SetRHS(0);
-  }
+  // Don't keep pointers to our vectors in the Epetra_LinearProblem.
+  linearProblem_->SetLHS(0);
+  linearProblem_->SetRHS(0);
+}
 
-  template <class Node>
-  RCP<MueLu::SmootherPrototype<double,int,int,Node> > AmesosSmoother<Node>::Copy() const {
-    return rcp( new AmesosSmoother<Node>(*this) );
-  }
+template <class Node>
+RCP<MueLu::SmootherPrototype<double, int, int, Node> > AmesosSmoother<Node>::Copy() const {
+  return rcp(new AmesosSmoother<Node>(*this));
+}
 
-  template <class Node>
-  std::string AmesosSmoother<Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    out << "{type = " << type_ << "}";
-    return out.str();
-  }
+template <class Node>
+std::string AmesosSmoother<Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  out << "{type = " << type_ << "}";
+  return out.str();
+}
 
-  //using MueLu::Describable::describe; // overloading, not hiding
-  template <class Node>
-  void AmesosSmoother<Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+// using MueLu::Describable::describe; // overloading, not hiding
+template <class Node>
+void AmesosSmoother<Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-    }
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
+  }
 
-    if (verbLevel & External)
-      if (prec_ != Teuchos::null) {
-        prec_->PrintStatus();
-        prec_->PrintTiming();
-      }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "RCP<A_>: " << A_ << std::endl
-           << "RCP<linearProblem__>: " << linearProblem_ << std::endl
-           << "RCP<prec_>: " << prec_ << std::endl;
+  if (verbLevel & External)
+    if (prec_ != Teuchos::null) {
+      prec_->PrintStatus();
+      prec_->PrintTiming();
     }
-  }
 
-  template <class Node>
-  size_t AmesosSmoother<Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "RCP<A_>: " << A_ << std::endl
+         << "RCP<linearProblem__>: " << linearProblem_ << std::endl
+         << "RCP<prec_>: " << prec_ << std::endl;
   }
+}
 
+template <class Node>
+size_t AmesosSmoother<Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-
-} // namespace MueLu
+}  // namespace MueLu
 
 // The AmesosSmoother is only templated on the Node, since it is an Epetra only object
 // Therefore we do not need the full ETI instantiations as we do for the other MueLu
@@ -226,5 +223,4 @@ namespace MueLu {
 template class MueLu::AmesosSmoother<Xpetra::EpetraNode>;
 #endif
 
-
-#endif // HAVE_MUELU_EPETRA && HAVE_MUELU_AMESOS
+#endif  // HAVE_MUELU_EPETRA && HAVE_MUELU_AMESOS
diff --git a/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.hpp b/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.hpp
index 19ae2d96c278..cceb51015e3a 100644
--- a/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_AmesosSmoother.hpp
@@ -64,155 +64,151 @@ class Amesos_BaseSolver;
 
 namespace MueLu {
 
-  /*!
-    @class AmesosSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Amesos direct solvers.
-
-    This class creates an Amesos preconditioner factory.  The factory is capable of generating direct solvers
-    based on the type and ParameterList passed into the constructor.  See the constructor for more information.
-  */
-  template <class Node = typename SmootherPrototype<double,int,int>::node_type>
-  class AmesosSmoother : public SmootherPrototype<double, int, int, Node>
-  {
-    typedef double Scalar;
-    typedef int    LocalOrdinal;
-    typedef int    GlobalOrdinal;
+/*!
+  @class AmesosSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Amesos direct solvers.
+
+  This class creates an Amesos preconditioner factory.  The factory is capable of generating direct solvers
+  based on the type and ParameterList passed into the constructor.  See the constructor for more information.
+*/
+template <class Node = typename SmootherPrototype<double, int, int>::node_type>
+class AmesosSmoother : public SmootherPrototype<double, int, int, Node> {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
 
 #undef MUELU_AMESOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-
-    /*! @brief Constructor
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-        Creates a MueLu interface to the direct solvers in the Amesos package.  The options are those specified in
-        the Amesos user's manual.
+  /*! @brief Constructor
 
-        @param type solver type
-        @param list options for the particular solver type
+      Creates a MueLu interface to the direct solvers in the Amesos package.  The options are those specified in
+      the Amesos user's manual.
 
-        Here is how to select the more commonly used direct solvers:
+      @param type solver type
+      @param list options for the particular solver type
 
-        - KLU (serial sparse direct solver)
-            - <tt>type</tt> = <tt>Amesos-KLU</tt>
-            - parameter list options
-                - none required
+      Here is how to select the more commonly used direct solvers:
 
-        - SuperLU (serial sparse super-nodal direct solver)
-            - <tt>type</tt> = <tt>Amesos-SuperLU</tt>
-            - parameter list options
-                - none required
+      - KLU (serial sparse direct solver)
+          - <tt>type</tt> = <tt>Amesos-KLU</tt>
+          - parameter list options
+              - none required
 
-        If you are using type=="", then either SuperLU or KLU are used by default.
+      - SuperLU (serial sparse super-nodal direct solver)
+          - <tt>type</tt> = <tt>Amesos-SuperLU</tt>
+          - parameter list options
+              - none required
 
-        See also Amesos_Klu and Amesos_Superlu.
+      If you are using type=="", then either SuperLU or KLU are used by default.
 
-    */
+      See also Amesos_Klu and Amesos_Superlu.
 
-    AmesosSmoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
-
-    //! Destructor
-    virtual ~AmesosSmoother() { };
+  */
 
-    //@}
+  AmesosSmoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    //! Input
-    //@{
+  //! Destructor
+  virtual ~AmesosSmoother(){};
 
-    void DeclareInput(Level& currentLevel) const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! @name Setup and Apply methods.
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    /*! @brief Set up the direct solver.
-      This creates the underlying Amesos solver object according to the parameter list options passed into the
-      AmesosSmoother constructor.  This includes doing a numeric factorization of the matrix.
-    */
-    void Setup(Level& currentLevel);
+  //@}
 
-    /*! @brief Apply the direct solver.
+  //! @name Setup and Apply methods.
+  //@{
 
-        Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  /*! @brief Set up the direct solver.
+    This creates the underlying Amesos solver object according to the parameter list options passed into the
+    AmesosSmoother constructor.  This includes doing a numeric factorization of the matrix.
+  */
+  void Setup(Level& currentLevel);
 
-        @param X initial guess
-        @param B right-hand side
-        @param InitialGuessIsZero This option has no effect with this smoother
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */ = false) const;
+  /*! @brief Apply the direct solver.
 
-    //@}
+      Solves the linear system <tt>AX=B</tt> using the constructed solver.
 
-    RCP<SmootherPrototype> Copy() const;
+      @param X initial guess
+      @param B right-hand side
+      @param InitialGuessIsZero This option has no effect with this smoother
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool /* InitialGuessIsZero */ = false) const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+  //@}
 
-    // Important note:
-    // linearProblem_ must be destroyed before A_, because destructor of linearProblem_ is using A_.
-    // In C++, destructor of member objects are called in the reverse order they appear within the declaration for the class.
-    // ==18029== Invalid read of size 8
-    // ==18029==    at 0xC0780A: Epetra_LinearProblem::GetMatrix() const (Epetra_LinearProblem.h:173)
-    // ==18029==    by 0xC5EC27: Amesos_Superlu::PrintTiming() const (Amesos_Superlu.cpp:664)
-    // ==18029==    by 0xC628C6: Amesos_Superlu::~Amesos_Superlu() (Amesos_Superlu.cpp:108)
+ private:
+  // Important note:
+  // linearProblem_ must be destroyed before A_, because destructor of linearProblem_ is using A_.
+  // In C++, destructor of member objects are called in the reverse order they appear within the declaration for the class.
+  // ==18029== Invalid read of size 8
+  // ==18029==    at 0xC0780A: Epetra_LinearProblem::GetMatrix() const (Epetra_LinearProblem.h:173)
+  // ==18029==    by 0xC5EC27: Amesos_Superlu::PrintTiming() const (Amesos_Superlu.cpp:664)
+  // ==18029==    by 0xC628C6: Amesos_Superlu::~Amesos_Superlu() (Amesos_Superlu.cpp:108)
 
-    //! amesos-specific key phrase that denote smoother type
-    std::string type_;
+  //! amesos-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! Matrix. Not used directly, but held inside of linearProblem_. So we have to keep an RCP pointer to it!
-    RCP<Matrix> A_;
+  //! Matrix. Not used directly, but held inside of linearProblem_. So we have to keep an RCP pointer to it!
+  RCP<Matrix> A_;
 
-    //! Problem that Amesos uses internally.
-    RCP<Epetra_LinearProblem> linearProblem_;
+  //! Problem that Amesos uses internally.
+  RCP<Epetra_LinearProblem> linearProblem_;
 
-    //! pointer to Amesos solver object
-    RCP<Amesos_BaseSolver> prec_;
+  //! pointer to Amesos solver object
+  RCP<Amesos_BaseSolver> prec_;
 
-  }; // class AmesosSmoother
+};  // class AmesosSmoother
 
-  //! Non-member templated function GetAmesosSmoother() returns a new AmesosSmoother object
-  // when <Scalar, LocalOrdinal, GlobalOrdinal> == <double, int, int>. Otherwise, an exception is thrown.
-  //! This function simplifies the usage of AmesosSmoother objects inside of templates as templates do not have to be specialized for <double, int, int> (see DirectSolver for an example).
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  GetAmesosSmoother (const std::string& /* type */ = "", const Teuchos::ParameterList& /* paramList */ = Teuchos::ParameterList ()) {
-    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                               "AmesosSmoother cannot be used with Scalar != double, LocalOrdinal != int, GlobalOrdinal != int");
-    TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
-  }
+//! Non-member templated function GetAmesosSmoother() returns a new AmesosSmoother object
+// when <Scalar, LocalOrdinal, GlobalOrdinal> == <double, int, int>. Otherwise, an exception is thrown.
+//! This function simplifies the usage of AmesosSmoother objects inside of templates as templates do not have to be specialized for <double, int, int> (see DirectSolver for an example).
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GetAmesosSmoother(const std::string& /* type */ = "", const Teuchos::ParameterList& /* paramList */ = Teuchos::ParameterList()) {
+  TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                             "AmesosSmoother cannot be used with Scalar != double, LocalOrdinal != int, GlobalOrdinal != int");
+  TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+}
 
-  // specialization for Epetra
+// specialization for Epetra
 #if defined(HAVE_MUELU_SERIAL)
-  template <>
-  inline RCP<MueLu::SmootherPrototype<double, int, int, Xpetra::EpetraNode> >
-  GetAmesosSmoother<double, int, int, Xpetra::EpetraNode> (const std::string& type, const Teuchos::ParameterList& paramList) {
-    return rcp (new MueLu::AmesosSmoother<Xpetra::EpetraNode>(type, paramList));
-  }
-#endif // HAVE_MUELU_SERIAL
+template <>
+inline RCP<MueLu::SmootherPrototype<double, int, int, Xpetra::EpetraNode> >
+GetAmesosSmoother<double, int, int, Xpetra::EpetraNode>(const std::string& type, const Teuchos::ParameterList& paramList) {
+  return rcp(new MueLu::AmesosSmoother<Xpetra::EpetraNode>(type, paramList));
+}
+#endif  // HAVE_MUELU_SERIAL
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_AMESOSSMOOTHER_SHORT
 
-#endif // HAVE_MUELU_AMESOS
-#endif // MUELU_AMESOSSMOOTHER_HPP
+#endif  // HAVE_MUELU_AMESOS
+#endif  // MUELU_AMESOSSMOOTHER_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_BelosSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_BelosSmoother_decl.hpp
index 2fca3c6651f7..2603386b3534 100644
--- a/packages/muelu/src/Smoothers/MueLu_BelosSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_BelosSmoother_decl.hpp
@@ -69,176 +69,173 @@
 
 #include <BelosSolverManager.hpp>
 
-
 namespace MueLu {
 
-  /*!
-    @class BelosSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Belos smoothers.
-
-    This class creates an Belos preconditioner factory. The factory creates a smoother based
-    on the type and ParameterList passed into the constructor. See the constructor for more
-    information.
-    */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class BelosSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class BelosSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Belos smoothers.
+
+  This class creates an Belos preconditioner factory. The factory creates a smoother based
+  on the type and ParameterList passed into the constructor. See the constructor for more
+  information.
+  */
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class BelosSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_BELOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-    //TODO: update doc for Belos.
-    /*! @brief Constructor
+ public:
+  //! @name Constructors / destructors
+  //@{
+  // TODO: update doc for Belos.
+  /*! @brief Constructor
 
-    The options passed into BelosSmoother are those given in the Belos user's manual.
+  The options passed into BelosSmoother are those given in the Belos user's manual.
 
-    @param type smoother type
-    @param list options for the particular smoother (e.g., fill factor or damping parameter)
+  @param type smoother type
+  @param list options for the particular smoother (e.g., fill factor or damping parameter)
 
-    */
+  */
 
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class BelosSmoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class BelosSmoother;
 #endif
 
-    BelosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
-
-    //! Destructor
-    virtual ~BelosSmoother() = default;
-
-    //@}
+  BelosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  //! Destructor
+  virtual ~BelosSmoother() = default;
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level &currentLevel) const;
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //@}
+  //! Input
+  //@{
 
-    //! @name Computational methods.
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    /*! @brief Set up the smoother.
+  //@}
 
-    This creates the underlying Belos smoother object, copies any parameter list options
-    supplied to the constructor to the Belos object, and computes the preconditioner.
-    */
-    void Setup(Level &currentLevel);
+  //! @name Computational methods.
+  //@{
 
-    /*! @brief Apply the preconditioner.
+  /*! @brief Set up the smoother.
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  This creates the underlying Belos smoother object, copies any parameter list options
+  supplied to the constructor to the Belos object, and computes the preconditioner.
+  */
+  void Setup(Level& currentLevel);
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Belos implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  /*! @brief Apply the preconditioner.
 
-    //@}
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    //! @name Utilities
-    //@{
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Belos implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    RCP<SmootherPrototype> Copy() const;
+  //@}
 
-    //@}
+  //! @name Utilities
+  //@{
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //@}
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //@}
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! For diagnostic purposes
-    // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //@}
 
-  private:
-    void SetupBelos           (Level& currentLevel);
+  //! For diagnostic purposes
+  // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
 
-  private:
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    std::string                         type_;
+ private:
+  void SetupBelos(Level& currentLevel);
 
-    typedef Tpetra::MultiVector<SC, LO, GO, NO> tMV;
-    typedef Tpetra::Operator<SC, LO, GO, NO>    tOP;
-    RCP<Belos::LinearProblem<Scalar, tMV, tOP> > tBelosProblem_;
-    RCP<Belos::SolverManager<Scalar, tMV, tOP> > tSolver_;
+ private:
+  std::string type_;
 
-    //! matrix, used in apply if solving residual equation
-    RCP<Matrix> A_;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> tMV;
+  typedef Tpetra::Operator<SC, LO, GO, NO> tOP;
+  RCP<Belos::LinearProblem<Scalar, tMV, tOP> > tBelosProblem_;
+  RCP<Belos::SolverManager<Scalar, tMV, tOP> > tSolver_;
 
-  }; // class BelosSmoother
+  //! matrix, used in apply if solving residual equation
+  RCP<Matrix> A_;
 
+};  // class BelosSmoother
 
-  #ifdef HAVE_MUELU_EPETRA
+#ifdef HAVE_MUELU_EPETRA
 
-# if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  // Stub specialization for missing Epetra template args
-  template<>
-  class BelosSmoother<double,int,int,Xpetra::EpetraNode> : public SmootherPrototype<double,int,int,Xpetra::EpetraNode> {
-    typedef double              Scalar;
-    typedef int                 LocalOrdinal;
-    typedef int                 GlobalOrdinal;
-    typedef Xpetra::EpetraNode  Node;
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+// Stub specialization for missing Epetra template args
+template <>
+class BelosSmoother<double, int, int, Xpetra::EpetraNode> : public SmootherPrototype<double, int, int, Xpetra::EpetraNode> {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
 #undef MUELU_BELOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class BelosSmoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class BelosSmoother;
 #endif
 
-    BelosSmoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LocalOrdinal& overlap = 0) {
-      MUELU_TPETRA_ETI_EXCEPTION("BelosSmoother<double,int,int,EpetraNode>","BelosSmoother<double,int,int,EpetraNode>","int");
-    };
-
-    virtual ~BelosSmoother() { }
+  BelosSmoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LocalOrdinal& overlap = 0) {
+    MUELU_TPETRA_ETI_EXCEPTION("BelosSmoother<double,int,int,EpetraNode>", "BelosSmoother<double,int,int,EpetraNode>", "int");
+  };
 
-    void SetParameterList(const Teuchos::ParameterList& paramList) {}
-    void DeclareInput(Level &currentLevel) const {}
-    void Setup(Level &currentLevel) {}
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {}
-    RCP<SmootherPrototype> Copy() const { return Teuchos::null;}
+  virtual ~BelosSmoother() {}
 
-    std::string description() const { return std::string(""); }
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {}
+  void SetParameterList(const Teuchos::ParameterList& paramList) {}
+  void DeclareInput(Level& currentLevel) const {}
+  void Setup(Level& currentLevel) {}
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {}
+  RCP<SmootherPrototype> Copy() const { return Teuchos::null; }
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {size_t cplx=0; return cplx;};
+  std::string description() const { return std::string(""); }
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const {}
 
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const {
+    size_t cplx = 0;
+    return cplx;
   };
-# endif
-
-#endif // HAVE_MUELU_EPETRA
+};
+#endif
 
+#endif  // HAVE_MUELU_EPETRA
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BELOSSMOOTHER_SHORT
-#endif // HAVE_MUELU_BELOS
-#endif // MUELU_BELOSSMOOTHER_DECL_HPP
+#endif  // HAVE_MUELU_BELOS
+#endif  // MUELU_BELOSSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_BelosSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_BelosSmoother_def.hpp
index 177e2e023aec..62f245b64f34 100644
--- a/packages/muelu/src/Smoothers/MueLu_BelosSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_BelosSmoother_def.hpp
@@ -67,166 +67,157 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosSolverFactory.hpp>
 
-
-
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BelosSmoother(const std::string type, const Teuchos::ParameterList& paramList)
-  : type_(type)
-  {
-    bool solverSupported = false;
-    Belos::SolverFactory<Scalar,tMV,tOP> tFactory;
-    solverSupported = solverSupported || tFactory.isSupported(type);
-    // if (!solverSupported) {
-    //   Teuchos::Array<std::string> supportedSolverNames = factory.supportedSolverNames();
-
-    //   std::ostringstream outString;
-    //   outString << "[";
-    //   for (auto iter = supportedSolverNames.begin(); iter != supportedSolverNames.end(); ++iter) {
-    //     outString << "\"" << *iter << "\"";
-    //     if (iter + 1 != supportedSolverNames.end()) {
-    //       outString << ", ";
-    //     }
-    //   }
-    //   outString << "]";
-
-    //   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Belos does not provide the solver '" << type_ << "'.\nSupported Solvers: " << outString.str());
-    // }
-    this->declareConstructionOutcome(!solverSupported, "Belos does not provide the smoother '" + type_ + "'.");
-    if (solverSupported)
-      SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
-    Factory::SetParameterList(paramList);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BelosSmoother(const std::string type, const Teuchos::ParameterList& paramList)
+  : type_(type) {
+  bool solverSupported = false;
+  Belos::SolverFactory<Scalar, tMV, tOP> tFactory;
+  solverSupported = solverSupported || tFactory.isSupported(type);
+  // if (!solverSupported) {
+  //   Teuchos::Array<std::string> supportedSolverNames = factory.supportedSolverNames();
+
+  //   std::ostringstream outString;
+  //   outString << "[";
+  //   for (auto iter = supportedSolverNames.begin(); iter != supportedSolverNames.end(); ++iter) {
+  //     outString << "\"" << *iter << "\"";
+  //     if (iter + 1 != supportedSolverNames.end()) {
+  //       outString << ", ";
+  //     }
+  //   }
+  //   outString << "]";
+
+  //   TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Belos does not provide the solver '" << type_ << "'.\nSupported Solvers: " << outString.str());
+  // }
+  this->declareConstructionOutcome(!solverSupported, "Belos does not provide the smoother '" + type_ + "'.");
+  if (solverSupported)
+    SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+
+  A_ = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+  SetupBelos(currentLevel);
+  SmootherPrototype::IsSetup(true);
+  this->GetOStream(Statistics1) << description() << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupBelos(Level& currentLevel) {
+  bool useTpetra = A_->getRowMap()->lib() == Xpetra::UseTpetra;
+
+  if (useTpetra) {
+    tBelosProblem_ = rcp(new Belos::LinearProblem<Scalar, tMV, tOP>());
+    RCP<tOP> tA    = Utilities::Op2NonConstTpetraCrs(A_);
+    tBelosProblem_->setOperator(tA);
+
+    Belos::SolverFactory<SC, tMV, tOP> solverFactory;
+    tSolver_ = solverFactory.create(type_, rcpFromRef(const_cast<ParameterList&>(this->GetParameterList())));
+    tSolver_->setProblem(tBelosProblem_);
+  } else {
+    TEUCHOS_ASSERT(false);
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BelosSmoother::Apply(): Setup() has not been called");
 
-    this->Input(currentLevel, "A");
+  if (A_->getRowMap()->lib() == Xpetra::UseTpetra) {
+    if (InitialGuessIsZero) {
+      X.putScalar(0.0);
 
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A");
-    SetupBelos(currentLevel);
-    SmootherPrototype::IsSetup(true);
-    this->GetOStream(Statistics1) << description() << std::endl;
-  }
+      RCP<Tpetra::MultiVector<SC, LO, GO, NO> > tpX       = rcpFromRef(Utilities::MV2NonConstTpetraMV(X));
+      RCP<const Tpetra::MultiVector<SC, LO, GO, NO> > tpB = rcpFromRef(Utilities::MV2TpetraMV(B));
 
+      tBelosProblem_->setInitResVec(tpB);
+      tBelosProblem_->setProblem(tpX, tpB);
+      tSolver_->solve();
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupBelos(Level& currentLevel) {
+    } else {
+      typedef Teuchos::ScalarTraits<Scalar> TST;
+      RCP<MultiVector> Residual   = Utilities::Residual(*A_, X, B);
+      RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
 
-    bool useTpetra = A_->getRowMap()->lib() == Xpetra::UseTpetra;
+      RCP<Tpetra::MultiVector<SC, LO, GO, NO> > tpX       = rcpFromRef(Utilities::MV2NonConstTpetraMV(*Correction));
+      RCP<const Tpetra::MultiVector<SC, LO, GO, NO> > tpB = rcpFromRef(Utilities::MV2TpetraMV(*Residual));
 
-    if (useTpetra) {
-      tBelosProblem_ = rcp(new Belos::LinearProblem<Scalar, tMV, tOP>());
-      RCP<tOP> tA = Utilities::Op2NonConstTpetraCrs(A_);
-      tBelosProblem_->setOperator(tA);
+      tBelosProblem_->setInitResVec(tpB);
+      tBelosProblem_->setProblem(tpX, tpB);
+      tSolver_->solve();
 
-      Belos::SolverFactory<SC, tMV, tOP> solverFactory;
-      tSolver_ = solverFactory.create(type_, rcpFromRef(const_cast<ParameterList&>(this->GetParameterList())));
-      tSolver_->setProblem(tBelosProblem_);
-    } else {
-      TEUCHOS_ASSERT(false);
+      X.update(TST::one(), *Correction, TST::one());
     }
+  } else {
+    TEUCHOS_ASSERT(false);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::BelosSmoother::Apply(): Setup() has not been called");
-
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<BelosSmoother> smoother = rcp(new BelosSmoother(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return smoother;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (SmootherPrototype::IsSetup()) {
     if (A_->getRowMap()->lib() == Xpetra::UseTpetra) {
-      if (InitialGuessIsZero) {
-        X.putScalar(0.0);
-
-        RCP<Tpetra::MultiVector<SC,LO,GO,NO> >       tpX = rcpFromRef(Utilities::MV2NonConstTpetraMV(X));
-        RCP<const Tpetra::MultiVector<SC,LO,GO,NO> > tpB = rcpFromRef(Utilities::MV2TpetraMV(B));
-
-        tBelosProblem_->setInitResVec(tpB);
-        tBelosProblem_->setProblem(tpX, tpB);
-        tSolver_->solve();
-
-      } else {
-        typedef Teuchos::ScalarTraits<Scalar> TST;
-        RCP<MultiVector> Residual   = Utilities::Residual(*A_, X, B);
-        RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
-
-        RCP<Tpetra::MultiVector<SC,LO,GO,NO> >       tpX = rcpFromRef(Utilities::MV2NonConstTpetraMV(*Correction));
-        RCP<const Tpetra::MultiVector<SC,LO,GO,NO> > tpB = rcpFromRef(Utilities::MV2TpetraMV(*Residual));
-
-        tBelosProblem_->setInitResVec(tpB);
-        tBelosProblem_->setProblem(tpX, tpB);
-        tSolver_->solve();
-
-        X.update(TST::one(), *Correction, TST::one());
-      }
-    } else {
-      TEUCHOS_ASSERT(false);
+      out << tSolver_->description();
     }
-
+  } else {
+    out << "BELOS {type = " << type_ << "}";
   }
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<BelosSmoother> smoother = rcp(new BelosSmoother(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return smoother;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (SmootherPrototype::IsSetup()) {
-      if (A_->getRowMap()->lib() == Xpetra::UseTpetra) {
-        out << tSolver_->description();
-      }
-    } else {
-      out << "BELOS {type = " << type_ << "}";
-    }
-    return out.str();
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
+  if (verbLevel & External) {
+    if (tSolver_ != Teuchos::null) {
       Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-    }
-
-    if (verbLevel & External) {
-      if (tSolver_ != Teuchos::null) {
-        Teuchos::OSTab tab2(out);
-        out << *tSolver_ << std::endl;
-      }
-    }
-
-    if (verbLevel & Debug) {
-      if (A_->getRowMap()->lib() == Xpetra::UseTpetra) {
-        out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-             << "-" << std::endl
-             << "RCP<solver_>: " << tSolver_ << std::endl;
-      }
+      out << *tSolver_ << std::endl;
     }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    if (A_->getRowMap()->lib() == Xpetra::UseTpetra) {
+      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+           << "-" << std::endl
+           << "RCP<solver_>: " << tSolver_ << std::endl;
+    }
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t BelosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // HAVE_MUELU_BELOS
-#endif // MUELU_BELOSSMOOTHER_DEF_HPP
+#endif  // HAVE_MUELU_BELOS
+#endif  // MUELU_BELOSSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_DirectSolver_decl.hpp b/packages/muelu/src/Smoothers/MueLu_DirectSolver_decl.hpp
index 360b6f787752..812f2d120ea2 100644
--- a/packages/muelu/src/Smoothers/MueLu_DirectSolver_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_DirectSolver_decl.hpp
@@ -48,7 +48,6 @@
 
 #include <Teuchos_ParameterList.hpp>
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_DirectSolver_fwd.hpp"
 
@@ -67,100 +66,98 @@
 
 namespace MueLu {
 
-  /*!
-    @class DirectSolver
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates direct solvers. Autoselection of AmesosSmoother or Amesos2Smoother according to the compile time configuration of Trilinos
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class DirectSolver : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class DirectSolver
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates direct solvers. Autoselection of AmesosSmoother or Amesos2Smoother according to the compile time configuration of Trilinos
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class DirectSolver : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_DIRECTSOLVER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-
-    //! @brief Constructor
-    //! Note: only parameters shared by Amesos and Amesos2 should be used for type and paramList (example: type= "Klu", "Superlu", paramList = <empty>) .
-    DirectSolver(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! Destructor
-    virtual ~DirectSolver() { }
+  //! @brief Constructor
+  //! Note: only parameters shared by Amesos and Amesos2 should be used for type and paramList (example: type= "Klu", "Superlu", paramList = <empty>) .
+  DirectSolver(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    //@}
+  //! Destructor
+  virtual ~DirectSolver() {}
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level& currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Setup and Apply methods.
-    //@{
+  //@}
 
-    //! DirectSolver cannot be turned into a smoother using Setup(). Setup() always returns a RuntimeError exception.
-    void Setup(Level& currentLevel);
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! DirectSolver cannot be applied. Apply() always returns a RuntimeError exception.
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  //! DirectSolver cannot be turned into a smoother using Setup(). Setup() always returns a RuntimeError exception.
+  void Setup(Level& currentLevel);
 
-    //@}
+  //! DirectSolver cannot be applied. Apply() always returns a RuntimeError exception.
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //! Custom SetFactory
-    void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
+  //@}
 
-    //! When this prototype is cloned using Copy(), the clone is an Amesos or an Amesos2 smoother.
-    RCP<SmootherPrototype> Copy() const;
+  //! Custom SetFactory
+  void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! When this prototype is cloned using Copy(), the clone is an Amesos or an Amesos2 smoother.
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {
-        if(!s_.is_null())
-          return s_->getNodeSmootherComplexity();
-        else return 0.0;
-    }
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const {
+    if (!s_.is_null())
+      return s_->getNodeSmootherComplexity();
+    else
+      return 0.0;
+  }
 
-  private:
+  //@}
 
-    //
-    // Parameters
-    //
+ private:
+  //
+  // Parameters
+  //
 
-    //! amesos1/2-specific key phrase that denote smoother type
-    std::string type_;
+  //! amesos1/2-specific key phrase that denote smoother type
+  std::string type_;
 
-    //
-    // Underlying Smoother
-    //
+  //
+  // Underlying Smoother
+  //
 
-    //! Smoother
-    RCP<SmootherPrototype> sEpetra_, sTpetra_, sBelos_, sStratimikos_, sRefMaxwell_;
-    mutable
-      RCP<SmootherPrototype> s_;
+  //! Smoother
+  RCP<SmootherPrototype> sEpetra_, sTpetra_, sBelos_, sStratimikos_, sRefMaxwell_;
+  mutable RCP<SmootherPrototype> s_;
 
-    // Records for the case if something goes wrong
-    bool triedEpetra_, triedTpetra_, triedBelos_, triedStratimikos_, triedRefMaxwell_;
-    std::string errorEpetra_, errorTpetra_, errorBelos_, errorStratimikos_, errorRefMaxwell_;
+  // Records for the case if something goes wrong
+  bool triedEpetra_, triedTpetra_, triedBelos_, triedStratimikos_, triedRefMaxwell_;
+  std::string errorEpetra_, errorTpetra_, errorBelos_, errorStratimikos_, errorRefMaxwell_;
 
-  }; // class DirectSolver
+};  // class DirectSolver
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_DIRECTSOLVER_SHORT
-#endif // MUELU_DIRECTSOLVER_DECL_HPP
+#endif  // MUELU_DIRECTSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_DirectSolver_def.hpp b/packages/muelu/src/Smoothers/MueLu_DirectSolver_def.hpp
index 65a59e0c7206..5e67d045ec36 100644
--- a/packages/muelu/src/Smoothers/MueLu_DirectSolver_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_DirectSolver_def.hpp
@@ -62,267 +62,271 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DirectSolver(const std::string& type, const Teuchos::ParameterList& paramListIn)
-    : type_(type) {
-    // The original idea behind all smoothers was to use prototype pattern. However, it does not fully work of the dependencies
-    // calculation. Particularly, we need to propagate DeclareInput to proper prototypes. Therefore, both TrilinosSmoother and
-    // DirectSolver do not follow the pattern exactly.
-    // The difference is that in order to propagate the calculation of dependencies, we need to call a DeclareInput on a
-    // constructed object (or objects, as we have two different code branches for Epetra and Tpetra). The only place where we
-    // could construct these objects is the constructor. Thus, we need to store RCPs, and both TrilinosSmoother and DirectSolver
-    // obtain a state: they contain RCP to smoother prototypes.
-    sEpetra_ = Teuchos::null;
-    sTpetra_ = Teuchos::null;
-    sBelos_ = Teuchos::null;
-
-    ParameterList paramList = paramListIn;
-
-    // We want DirectSolver to be able to work with both Epetra and Tpetra objects, therefore we try to construct both
-    // Amesos and Amesos2 solver prototypes. The construction really depends on configuration options.
-    triedEpetra_ = triedTpetra_ = false;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DirectSolver(const std::string& type, const Teuchos::ParameterList& paramListIn)
+  : type_(type) {
+  // The original idea behind all smoothers was to use prototype pattern. However, it does not fully work of the dependencies
+  // calculation. Particularly, we need to propagate DeclareInput to proper prototypes. Therefore, both TrilinosSmoother and
+  // DirectSolver do not follow the pattern exactly.
+  // The difference is that in order to propagate the calculation of dependencies, we need to call a DeclareInput on a
+  // constructed object (or objects, as we have two different code branches for Epetra and Tpetra). The only place where we
+  // could construct these objects is the constructor. Thus, we need to store RCPs, and both TrilinosSmoother and DirectSolver
+  // obtain a state: they contain RCP to smoother prototypes.
+  sEpetra_ = Teuchos::null;
+  sTpetra_ = Teuchos::null;
+  sBelos_  = Teuchos::null;
+
+  ParameterList paramList = paramListIn;
+
+  // We want DirectSolver to be able to work with both Epetra and Tpetra objects, therefore we try to construct both
+  // Amesos and Amesos2 solver prototypes. The construction really depends on configuration options.
+  triedEpetra_ = triedTpetra_ = false;
 #if defined(HAVE_MUELU_AMESOS2)
-    try {
-      sTpetra_ = rcp(new Amesos2Smoother(type_, paramList));
-      if (sTpetra_.is_null())
-        errorTpetra_ = "Unable to construct Amesos2 direct solver";
-      else if (!sTpetra_->constructionSuccessful()) {
-        errorTpetra_ = sTpetra_->constructionErrorMsg();
-        sTpetra_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e) {
-      errorTpetra_ = e.what();
-    } catch (Exceptions::BadCast& e) {
-      errorTpetra_ = e.what();
-    } catch (Teuchos::Exceptions::InvalidParameterName& e) {
-      errorTpetra_ = e.what();
+  try {
+    sTpetra_ = rcp(new Amesos2Smoother(type_, paramList));
+    if (sTpetra_.is_null())
+      errorTpetra_ = "Unable to construct Amesos2 direct solver";
+    else if (!sTpetra_->constructionSuccessful()) {
+      errorTpetra_ = sTpetra_->constructionErrorMsg();
+      sTpetra_     = Teuchos::null;
     }
-    triedTpetra_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    errorTpetra_ = e.what();
+  } catch (Exceptions::BadCast& e) {
+    errorTpetra_ = e.what();
+  } catch (Teuchos::Exceptions::InvalidParameterName& e) {
+    errorTpetra_ = e.what();
+  }
+  triedTpetra_ = true;
 #endif
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
-    try {
-      // GetAmesosSmoother masks the template argument matching, and simply throws if template arguments are incompatible with Epetra
-      sEpetra_ = GetAmesosSmoother<SC,LO,GO,NO>(type_, paramList);
-      if (sEpetra_.is_null())
-        errorEpetra_ = "Unable to construct Amesos direct solver";
-      else if (!sEpetra_->constructionSuccessful()) {
-        errorEpetra_ = sEpetra_->constructionErrorMsg();
-        sEpetra_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e) {
-      // AmesosSmoother throws if Scalar != double, LocalOrdinal != int, GlobalOrdinal != int
-      errorEpetra_ = e.what();
+  try {
+    // GetAmesosSmoother masks the template argument matching, and simply throws if template arguments are incompatible with Epetra
+    sEpetra_ = GetAmesosSmoother<SC, LO, GO, NO>(type_, paramList);
+    if (sEpetra_.is_null())
+      errorEpetra_ = "Unable to construct Amesos direct solver";
+    else if (!sEpetra_->constructionSuccessful()) {
+      errorEpetra_ = sEpetra_->constructionErrorMsg();
+      sEpetra_     = Teuchos::null;
     }
-    triedEpetra_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    // AmesosSmoother throws if Scalar != double, LocalOrdinal != int, GlobalOrdinal != int
+    errorEpetra_ = e.what();
+  }
+  triedEpetra_ = true;
 #endif
 #if defined(HAVE_MUELU_BELOS)
-    try {
-      sBelos_ = rcp(new BelosSmoother(type_, paramList));
-      if (sBelos_.is_null())
-        errorBelos_ = "Unable to construct Belos solver";
-      else if (!sBelos_->constructionSuccessful()) {
-        errorBelos_ = sBelos_->constructionErrorMsg();
-        sBelos_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e) {
-      errorBelos_ = e.what();
-    } catch (Exceptions::BadCast& e) {
-      errorBelos_ = e.what();
+  try {
+    sBelos_ = rcp(new BelosSmoother(type_, paramList));
+    if (sBelos_.is_null())
+      errorBelos_ = "Unable to construct Belos solver";
+    else if (!sBelos_->constructionSuccessful()) {
+      errorBelos_ = sBelos_->constructionErrorMsg();
+      sBelos_     = Teuchos::null;
     }
-    triedBelos_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    errorBelos_ = e.what();
+  } catch (Exceptions::BadCast& e) {
+    errorBelos_ = e.what();
+  }
+  triedBelos_ = true;
 #endif
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    try {
-      sStratimikos_ = rcp(new StratimikosSmoother(type_, paramList));
-      if (sStratimikos_.is_null())
-        errorStratimikos_ = "Unable to construct Stratimikos smoother";
-      else if (!sStratimikos_->constructionSuccessful()) {
-        errorStratimikos_ = sStratimikos_->constructionErrorMsg();
-        sStratimikos_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e){
-      errorStratimikos_ = e.what();
+  try {
+    sStratimikos_ = rcp(new StratimikosSmoother(type_, paramList));
+    if (sStratimikos_.is_null())
+      errorStratimikos_ = "Unable to construct Stratimikos smoother";
+    else if (!sStratimikos_->constructionSuccessful()) {
+      errorStratimikos_ = sStratimikos_->constructionErrorMsg();
+      sStratimikos_     = Teuchos::null;
     }
-    triedStratimikos_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    errorStratimikos_ = e.what();
+  }
+  triedStratimikos_ = true;
 #endif
-    try {
-      sRefMaxwell_ = rcp(new RefMaxwellSmoother(type_, paramList));
-      if (sRefMaxwell_.is_null())
-        errorRefMaxwell_ = "Unable to construct RefMaxwell smoother";
-      else if (!sRefMaxwell_->constructionSuccessful()) {
-        errorRefMaxwell_ = sRefMaxwell_->constructionErrorMsg();
-        sRefMaxwell_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e){
-      errorRefMaxwell_ = e.what();
+  try {
+    sRefMaxwell_ = rcp(new RefMaxwellSmoother(type_, paramList));
+    if (sRefMaxwell_.is_null())
+      errorRefMaxwell_ = "Unable to construct RefMaxwell smoother";
+    else if (!sRefMaxwell_->constructionSuccessful()) {
+      errorRefMaxwell_ = sRefMaxwell_->constructionErrorMsg();
+      sRefMaxwell_     = Teuchos::null;
     }
-    triedRefMaxwell_ = true;
-
-    // Check if we were able to construct at least one solver. In many cases that's all we need, for instance if a user
-    // simply wants to use Tpetra only stack, never enables Amesos, and always runs Tpetra objects.
-    TEUCHOS_TEST_FOR_EXCEPTION(!triedEpetra_ && !triedTpetra_ && !triedBelos_ && !triedStratimikos_ && !triedRefMaxwell_, Exceptions::RuntimeError, "Unable to construct any direct solver."
-                               "Plase enable (TPETRA and AMESOS2) or (EPETRA and AMESOS) or (BELOS) or (STRATIMIKOS)");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(sEpetra_.is_null() && sTpetra_.is_null() && sBelos_.is_null() && sStratimikos_.is_null() && sRefMaxwell_.is_null(), Exceptions::RuntimeError,
-        "Could not enable any direct solver:\n"
-        << (triedEpetra_ ? "Epetra mode was disabled due to an error:\n" : "")
-        << (triedEpetra_ ? errorEpetra_ : "")
-        << (triedTpetra_ ? "Tpetra mode was disabled due to an error:\n" : "")
-        << (triedTpetra_ ? errorTpetra_ : "")
-        << (triedBelos_ ? "Belos was disabled due to an error:\n" : "")
-        << (triedBelos_ ? errorBelos_ : "")
-        << (triedStratimikos_ ? "Stratimikos was disabled due to an error:\n" : "")
-        << (triedStratimikos_ ? errorStratimikos_ : "")
-        << (triedRefMaxwell_ ? "RefMaxwell was disabled due to an error:\n" : "")
-        << (triedRefMaxwell_ ? errorRefMaxwell_ : ""));;
-
-    this->SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
-    // We need to propagate SetFactory to proper place
-    if (!sEpetra_.is_null()) sEpetra_->SetFactory(varName, factory);
-    if (!sTpetra_.is_null()) sTpetra_->SetFactory(varName, factory);
-    if (!sBelos_.is_null())  sBelos_->SetFactory(varName, factory);
-    if (!sStratimikos_.is_null()) sStratimikos_->SetFactory(varName, factory);
-    if (!sRefMaxwell_.is_null()) sRefMaxwell_->SetFactory(varName, factory);
+  } catch (Exceptions::RuntimeError& e) {
+    errorRefMaxwell_ = e.what();
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    if (!sBelos_.is_null())
-      s_ = sBelos_;
-    else if (!sStratimikos_.is_null())
-      s_ = sStratimikos_;
-    else if (!sRefMaxwell_.is_null())
-      s_ = sRefMaxwell_;
-    else {
-      // Decide whether we are running in Epetra or Tpetra mode
-      //
-      // Theoretically, we could make this decision in the constructor, and create only
-      // one of the smoothers. But we want to be able to reuse, so one can imagine a scenario
-      // where one first runs hierarchy with tpetra matrix, and then with epetra.
-      bool useTpetra = (currentLevel.lib() == Xpetra::UseTpetra);
-      s_ = (useTpetra ? sTpetra_ : sEpetra_);
-      if (s_.is_null()) {
-        if (useTpetra) {
+  triedRefMaxwell_ = true;
+
+  // Check if we were able to construct at least one solver. In many cases that's all we need, for instance if a user
+  // simply wants to use Tpetra only stack, never enables Amesos, and always runs Tpetra objects.
+  TEUCHOS_TEST_FOR_EXCEPTION(!triedEpetra_ && !triedTpetra_ && !triedBelos_ && !triedStratimikos_ && !triedRefMaxwell_, Exceptions::RuntimeError,
+                             "Unable to construct any direct solver."
+                             "Plase enable (TPETRA and AMESOS2) or (EPETRA and AMESOS) or (BELOS) or (STRATIMIKOS)");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(sEpetra_.is_null() && sTpetra_.is_null() && sBelos_.is_null() && sStratimikos_.is_null() && sRefMaxwell_.is_null(), Exceptions::RuntimeError,
+                             "Could not enable any direct solver:\n"
+                                 << (triedEpetra_ ? "Epetra mode was disabled due to an error:\n" : "")
+                                 << (triedEpetra_ ? errorEpetra_ : "")
+                                 << (triedTpetra_ ? "Tpetra mode was disabled due to an error:\n" : "")
+                                 << (triedTpetra_ ? errorTpetra_ : "")
+                                 << (triedBelos_ ? "Belos was disabled due to an error:\n" : "")
+                                 << (triedBelos_ ? errorBelos_ : "")
+                                 << (triedStratimikos_ ? "Stratimikos was disabled due to an error:\n" : "")
+                                 << (triedStratimikos_ ? errorStratimikos_ : "")
+                                 << (triedRefMaxwell_ ? "RefMaxwell was disabled due to an error:\n" : "")
+                                 << (triedRefMaxwell_ ? errorRefMaxwell_ : ""));
+  ;
+
+  this->SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
+  // We need to propagate SetFactory to proper place
+  if (!sEpetra_.is_null()) sEpetra_->SetFactory(varName, factory);
+  if (!sTpetra_.is_null()) sTpetra_->SetFactory(varName, factory);
+  if (!sBelos_.is_null()) sBelos_->SetFactory(varName, factory);
+  if (!sStratimikos_.is_null()) sStratimikos_->SetFactory(varName, factory);
+  if (!sRefMaxwell_.is_null()) sRefMaxwell_->SetFactory(varName, factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  if (!sBelos_.is_null())
+    s_ = sBelos_;
+  else if (!sStratimikos_.is_null())
+    s_ = sStratimikos_;
+  else if (!sRefMaxwell_.is_null())
+    s_ = sRefMaxwell_;
+  else {
+    // Decide whether we are running in Epetra or Tpetra mode
+    //
+    // Theoretically, we could make this decision in the constructor, and create only
+    // one of the smoothers. But we want to be able to reuse, so one can imagine a scenario
+    // where one first runs hierarchy with tpetra matrix, and then with epetra.
+    bool useTpetra = (currentLevel.lib() == Xpetra::UseTpetra);
+    s_             = (useTpetra ? sTpetra_ : sEpetra_);
+    if (s_.is_null()) {
+      if (useTpetra) {
 #if not defined(HAVE_MUELU_AMESOS2)
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                     "Error: running in Tpetra mode, but MueLu with Amesos2 was disabled during the configure stage.\n"
-                                     "Please make sure that:\n"
-                                     "  - Amesos2 is enabled (Trilinos_ENABLE_Amesos2=ON),\n"
-                                     "  - Amesos2 is available for MueLu to use (MueLu_ENABLE_Amesos2=ON)\n");
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                   "Error: running in Tpetra mode, but MueLu with Amesos2 was disabled during the configure stage.\n"
+                                   "Please make sure that:\n"
+                                   "  - Amesos2 is enabled (Trilinos_ENABLE_Amesos2=ON),\n"
+                                   "  - Amesos2 is available for MueLu to use (MueLu_ENABLE_Amesos2=ON)\n");
 #else
-          if (triedTpetra_)
-            this->GetOStream(Errors) << "Tpetra mode was disabled due to an error:\n" << errorTpetra_ << std::endl;
+        if (triedTpetra_)
+          this->GetOStream(Errors) << "Tpetra mode was disabled due to an error:\n"
+                                   << errorTpetra_ << std::endl;
 #endif
-        }
-        if (!useTpetra) {
+      }
+      if (!useTpetra) {
 #if not defined(HAVE_MUELU_AMESOS)
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                     "Error: running in Epetra mode, but MueLu with Amesos was disabled during the configure stage.\n"
-                                     "Please make sure that:\n"
-                                     "  - Amesos is enabled (you can do that with Trilinos_ENABLE_Amesos=ON),\n"
-                                     "  - Amesos is available for MueLu to use (MueLu_ENABLE_Amesos=ON)\n");
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                   "Error: running in Epetra mode, but MueLu with Amesos was disabled during the configure stage.\n"
+                                   "Please make sure that:\n"
+                                   "  - Amesos is enabled (you can do that with Trilinos_ENABLE_Amesos=ON),\n"
+                                   "  - Amesos is available for MueLu to use (MueLu_ENABLE_Amesos=ON)\n");
 #else
-          if (triedEpetra_)
-            this->GetOStream(Errors) << "Epetra mode was disabled due to an error:\n" << errorEpetra_ << std::endl;
+        if (triedEpetra_)
+          this->GetOStream(Errors) << "Epetra mode was disabled due to an error:\n"
+                                   << errorEpetra_ << std::endl;
 #endif
-        }
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                   "Direct solver for " << (useTpetra ? "Tpetra" : "Epetra") << " was not constructed");
       }
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                 "Direct solver for " << (useTpetra ? "Tpetra" : "Epetra") << " was not constructed");
     }
-
-    s_->DeclareInput(currentLevel);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::DirectSolver::Setup(): Setup() has already been called" << std::endl;
-
-    int oldRank = s_->SetProcRankVerbose(this->GetProcRankVerbose());
-
-    s_->Setup(currentLevel);
-
-    s_->SetProcRankVerbose(oldRank);
-
-    SmootherPrototype::IsSetup(true);
-
-    this->SetParameterList(s_->GetParameterList());
+  s_->DeclareInput(currentLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::DirectSolver::Setup(): Setup() has already been called" << std::endl;
+
+  int oldRank = s_->SetProcRankVerbose(this->GetProcRankVerbose());
+
+  s_->Setup(currentLevel);
+
+  s_->SetProcRankVerbose(oldRank);
+
+  SmootherPrototype::IsSetup(true);
+
+  this->SetParameterList(s_->GetParameterList());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Apply(): Setup() has not been called");
+
+  s_->Apply(X, B, InitialGuessIsZero);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<DirectSolver> newSmoo = rcp(new DirectSolver(*this));
+
+  // We need to be quite careful with Copy
+  // We still want DirectSolver to follow Prototype Pattern, so we need to hide the fact that we do have some state
+  if (!sEpetra_.is_null())
+    newSmoo->sEpetra_ = sEpetra_->Copy();
+  if (!sTpetra_.is_null())
+    newSmoo->sTpetra_ = sTpetra_->Copy();
+  if (!sBelos_.is_null())
+    newSmoo->sBelos_ = sBelos_->Copy();
+  if (!sStratimikos_.is_null())
+    newSmoo->sStratimikos_ = sStratimikos_->Copy();
+  if (!sRefMaxwell_.is_null())
+    newSmoo->sRefMaxwell_ = sRefMaxwell_->Copy();
+
+  // Copy the default mode
+  if (s_.get() == sBelos_.get())
+    newSmoo->s_ = newSmoo->sBelos_;
+  else if (s_.get() == sStratimikos_.get())
+    newSmoo->s_ = newSmoo->sStratimikos_;
+  else if (s_.get() == sRefMaxwell_.get())
+    newSmoo->s_ = newSmoo->sRefMaxwell_;
+  else if (s_.get() == sTpetra_.get())
+    newSmoo->s_ = newSmoo->sTpetra_;
+  else
+    newSmoo->s_ = newSmoo->sEpetra_;
+  newSmoo->SetParameterList(this->GetParameterList());
+
+  return newSmoo;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (s_ != Teuchos::null) {
+    out << s_->description();
+  } else {
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
   }
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::AmesosSmoother::Apply(): Setup() has not been called");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    s_->Apply(X, B, InitialGuessIsZero);
-  }
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<DirectSolver> newSmoo =  rcp(new DirectSolver(*this));
-
-    // We need to be quite careful with Copy
-    // We still want DirectSolver to follow Prototype Pattern, so we need to hide the fact that we do have some state
-    if (!sEpetra_.is_null())
-      newSmoo->sEpetra_ = sEpetra_->Copy();
-    if (!sTpetra_.is_null())
-      newSmoo->sTpetra_ = sTpetra_->Copy();
-    if (!sBelos_.is_null())
-      newSmoo->sBelos_ = sBelos_->Copy();
-    if (!sStratimikos_.is_null())
-      newSmoo->sStratimikos_ = sStratimikos_->Copy();
-    if (!sRefMaxwell_.is_null())
-      newSmoo->sRefMaxwell_ = sRefMaxwell_->Copy();
-
-    // Copy the default mode
-    if (s_.get() == sBelos_.get())
-      newSmoo->s_ = newSmoo->sBelos_;
-    else if (s_.get() == sStratimikos_.get())
-      newSmoo->s_ = newSmoo->sStratimikos_;
-    else if (s_.get() == sRefMaxwell_.get())
-      newSmoo->s_ = newSmoo->sRefMaxwell_;
-    else if (s_.get() == sTpetra_.get())
-      newSmoo->s_ = newSmoo->sTpetra_;
-    else
-      newSmoo->s_ = newSmoo->sEpetra_;
-    newSmoo->SetParameterList(this->GetParameterList());
-
-    return newSmoo;
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab3(out);
+    out << this->GetParameterList();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (s_ != Teuchos::null) {
-      out << s_->description();
-    } else {
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-    }
-    return out.str();
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void DirectSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
-
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab3(out);
-      out << this->GetParameterList();
-    }
-
-    if (verbLevel & Debug)
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
-  }
+  if (verbLevel & Debug)
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_DIRECTSOLVER_DEF_HPP
+#endif  // MUELU_DIRECTSOLVER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_decl.hpp
index 67a06832067a..83d9ae3060b0 100644
--- a/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_decl.hpp
@@ -52,7 +52,7 @@
 #include <Xpetra_CrsMatrixWrap_fwd.hpp>
 #include <Xpetra_Matrix_fwd.hpp>
 #include <Xpetra_MultiVectorFactory_fwd.hpp>
-#ifdef HAVE_XPETRA_TPETRA // needed for clone()
+#ifdef HAVE_XPETRA_TPETRA  // needed for clone()
 #include <Xpetra_TpetraCrsMatrix_fwd.hpp>
 #endif
 
@@ -74,215 +74,217 @@
 
 namespace MueLu {
 
-  /*!
-    @class Ifpack2Smoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Ifpack2 smoothers.
-
-    This class creates an Ifpack2 preconditioner factory. The factory creates a smoother based
-    on the type and ParameterList passed into the constructor. See the constructor for more
-    information.
-    */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class Ifpack2Smoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class Ifpack2Smoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Ifpack2 smoothers.
+
+  This class creates an Ifpack2 preconditioner factory. The factory creates a smoother based
+  on the type and ParameterList passed into the constructor. See the constructor for more
+  information.
+  */
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class Ifpack2Smoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_IFPACK2SMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-    //TODO: update doc for Ifpack2. Right now, it's a copy of the doc of IfpackSmoother
-    /*! @brief Constructor
-
-    The options passed into Ifpack2Smoother are those given in the Ifpack2 user's manual.
-
-    @param type smoother type
-    @param list options for the particular smoother (e.g., fill factor or damping parameter)
-
-    Here is how to select some of the most common smoothers.
-
-    - Gauss-Seidel
-      - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
-      - parameter list options
-        - <tt>relaxation: type</tt> = <tt>Gauss-Seidel</tt>
-        - <tt>relaxation: damping factor</tt>
-    - symmetric Gauss-Seidel
-      - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
-      - parameter list options
-        - <tt>relaxation: type</tt> = <tt>symmetric Gauss-Seidel</tt>
-        - <tt>relaxation: damping factor</tt>
-    - Chebyshev
-      - <tt>type</tt> = <tt>Chebyshev</tt>
-      - parameter list options
-        - <tt>chebyshev: ratio eigenvalue</tt>
-        - <tt>chebyshev: min eigenvalue</tt>
-        - <tt>chebyshev: max eigenvalue</tt>
-        - <tt>chebyshev: degree</tt>
-        - <tt>chebyshev: zero starting solution</tt> (defaults to <tt>true</tt>)
-    - ILU
-      - <tt>type</tt> = <tt>ILU</tt>
-      - parameter list options
-        - <tt>fact: level-of-fill</tt>
-
-    See also Ifpack2::Relaxation, Ifpack2::Chebyshev, or Ifpack2::ILUT.
-    */
+ public:
+  //! @name Constructors / destructors
+  //@{
+  // TODO: update doc for Ifpack2. Right now, it's a copy of the doc of IfpackSmoother
+  /*! @brief Constructor
+
+  The options passed into Ifpack2Smoother are those given in the Ifpack2 user's manual.
+
+  @param type smoother type
+  @param list options for the particular smoother (e.g., fill factor or damping parameter)
+
+  Here is how to select some of the most common smoothers.
+
+  - Gauss-Seidel
+    - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
+    - parameter list options
+      - <tt>relaxation: type</tt> = <tt>Gauss-Seidel</tt>
+      - <tt>relaxation: damping factor</tt>
+  - symmetric Gauss-Seidel
+    - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
+    - parameter list options
+      - <tt>relaxation: type</tt> = <tt>symmetric Gauss-Seidel</tt>
+      - <tt>relaxation: damping factor</tt>
+  - Chebyshev
+    - <tt>type</tt> = <tt>Chebyshev</tt>
+    - parameter list options
+      - <tt>chebyshev: ratio eigenvalue</tt>
+      - <tt>chebyshev: min eigenvalue</tt>
+      - <tt>chebyshev: max eigenvalue</tt>
+      - <tt>chebyshev: degree</tt>
+      - <tt>chebyshev: zero starting solution</tt> (defaults to <tt>true</tt>)
+  - ILU
+    - <tt>type</tt> = <tt>ILU</tt>
+    - parameter list options
+      - <tt>fact: level-of-fill</tt>
+
+  See also Ifpack2::Relaxation, Ifpack2::Chebyshev, or Ifpack2::ILUT.
+  */
 
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class Ifpack2Smoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class Ifpack2Smoother;
 #endif
 
-    Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LO& overlap = 0); //TODO: empty paramList valid for Ifpack??
+  Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LO& overlap = 0);  // TODO: empty paramList valid for Ifpack??
 
-    //! Destructor
-    virtual ~Ifpack2Smoother() { }
+  //! Destructor
+  virtual ~Ifpack2Smoother() {}
 
-    //@}
+  //@}
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Computational methods.
-    //@{
+  //! @name Computational methods.
+  //@{
 
-    /*! @brief Set up the smoother.
+  /*! @brief Set up the smoother.
 
-    This creates the underlying Ifpack2 smoother object, copies any parameter list options
-    supplied to the constructor to the Ifpack2 object, and computes the preconditioner.
+  This creates the underlying Ifpack2 smoother object, copies any parameter list options
+  supplied to the constructor to the Ifpack2 object, and computes the preconditioner.
 
-    TODO The eigenvalue estimate should come from A_, not the Ifpack2 parameter list.
-    */
-    void Setup(Level &currentLevel);
+  TODO The eigenvalue estimate should come from A_, not the Ifpack2 parameter list.
+  */
+  void Setup(Level& currentLevel);
 
-    /*! @brief Apply the preconditioner.
+  /*! @brief Apply the preconditioner.
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Ifpack2 implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Ifpack2 implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //@}
+  //@}
 
-    //! @name Utilities
-    //@{
+  //! @name Utilities
+  //@{
 
-    RCP<SmootherPrototype> Copy() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //@}
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //@}
 
-    //! For diagnostic purposes
-    RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
+  //! For diagnostic purposes
+  RCP<Ifpack2::Preconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node> > getPreconditioner() { return prec_; }
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
-    void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const;
+ private:
+  void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const;
 
-    void SetupSchwarz         (Level& currentLevel);
-    void SetupChebyshev       (Level& currentLevel);
-    void SetupLineSmoothing   (Level& currentLevel);
-    void SetupBlockRelaxation (Level& currentLevel);
-    void SetupTopological     (Level& currentLevel);
-    void SetupAggregate       (Level& currentLevel);
-    void SetupHiptmair        (Level& currentLevel);
-    void SetupGeneric         (Level& currentLevel);
-    Scalar SetupChebyshevEigenvalues(Level & currentLevel, const std::string & matrixName, const std::string & label, Teuchos::ParameterList & paramList) const;
-  private:
+  void SetupSchwarz(Level& currentLevel);
+  void SetupChebyshev(Level& currentLevel);
+  void SetupLineSmoothing(Level& currentLevel);
+  void SetupBlockRelaxation(Level& currentLevel);
+  void SetupTopological(Level& currentLevel);
+  void SetupAggregate(Level& currentLevel);
+  void SetupHiptmair(Level& currentLevel);
+  void SetupGeneric(Level& currentLevel);
+  Scalar SetupChebyshevEigenvalues(Level& currentLevel, const std::string& matrixName, const std::string& label, Teuchos::ParameterList& paramList) const;
 
-    //!
-    bool constructionSuccessful_;
+ private:
+  //!
+  bool constructionSuccessful_;
 
-    //! ifpack2-specific key phrase that denote smoother type
-    std::string type_;
+  //! ifpack2-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! overlap when using the smoother in additive Schwarz mode
-    LO overlap_;
+  //! overlap when using the smoother in additive Schwarz mode
+  LO overlap_;
 
-    //! pointer to Ifpack2 preconditioner object
-    RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > prec_;
+  //! pointer to Ifpack2 preconditioner object
+  RCP<Ifpack2::Preconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node> > prec_;
 
-    //! matrix, used in apply if solving residual equation
-    RCP<Operator> A_;
+  //! matrix, used in apply if solving residual equation
+  RCP<Operator> A_;
 
-  }; // class Ifpack2Smoother
+};  // class Ifpack2Smoother
 
 #ifdef HAVE_MUELU_EPETRA
 
-# if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  // Stub specialization for missing Epetra template args
-  template<>
-  class Ifpack2Smoother<double,int,int,Xpetra::EpetraNode> : public SmootherPrototype<double,int,int,Xpetra::EpetraNode> {
-    typedef double              Scalar;
-    typedef int                 LocalOrdinal;
-    typedef int                 GlobalOrdinal;
-    typedef Xpetra::EpetraNode  Node;
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+// Stub specialization for missing Epetra template args
+template <>
+class Ifpack2Smoother<double, int, int, Xpetra::EpetraNode> : public SmootherPrototype<double, int, int, Xpetra::EpetraNode> {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
 #undef MUELU_AMESOS2SMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class Ifpack2Smoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class Ifpack2Smoother;
 #endif
 
-    Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LocalOrdinal& overlap = 0) {
-      MUELU_TPETRA_ETI_EXCEPTION("Ifpack2Smoother<double,int,int,EpetraNode>","Ifpack2Smoother<double,int,int,EpetraNode>","int");
-    };
-
-    virtual ~Ifpack2Smoother() { }
+  Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LocalOrdinal& overlap = 0) {
+    MUELU_TPETRA_ETI_EXCEPTION("Ifpack2Smoother<double,int,int,EpetraNode>", "Ifpack2Smoother<double,int,int,EpetraNode>", "int");
+  };
 
-    void SetParameterList(const Teuchos::ParameterList& paramList) {}
-    void DeclareInput(Level &currentLevel) const {}
-    void Setup(Level &currentLevel) {}
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {}
-    RCP<SmootherPrototype> Copy() const { return Teuchos::null;}
+  virtual ~Ifpack2Smoother() {}
 
-    std::string description() const { return std::string(""); }
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {}
+  void SetParameterList(const Teuchos::ParameterList& paramList) {}
+  void DeclareInput(Level& currentLevel) const {}
+  void Setup(Level& currentLevel) {}
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {}
+  RCP<SmootherPrototype> Copy() const { return Teuchos::null; }
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {size_t cplx=0; return cplx;};
+  std::string description() const { return std::string(""); }
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const {}
 
-  private:
-    void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const {}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const {
+    size_t cplx = 0;
+    return cplx;
   };
-# endif
 
-#endif // HAVE_MUELU_EPETRA
+ private:
+  void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const {}
+};
+#endif
+
+#endif  // HAVE_MUELU_EPETRA
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_IFPACK2SMOOTHER_SHORT
-#endif // HAVE_MUELU_IFPACK2
-#endif // MUELU_IFPACK2SMOOTHER_DECL_HPP
+#endif  // HAVE_MUELU_IFPACK2
+#endif  // MUELU_IFPACK2SMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_def.hpp
index 4e077c993075..55f5a0c8cfb7 100644
--- a/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_Ifpack2Smoother_def.hpp
@@ -80,7 +80,6 @@
 #include "MueLu_Monitor.hpp"
 #include "MueLu_Aggregates.hpp"
 
-
 #ifdef HAVE_MUELU_INTREPID2
 #include "MueLu_IntrepidPCoarsenFactory_decl.hpp"
 #include "MueLu_IntrepidPCoarsenFactory_def.hpp"
@@ -88,94 +87,92 @@
 #include "Kokkos_DynRankView.hpp"
 #endif
 
-
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList, const LO& overlap)
-    : type_(type), overlap_(overlap)
-  {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
-    bool isSupported = Ifpack2::Factory::isSupported<tRowMatrix>(type_) || (type_ == "LINESMOOTHING_TRIDI_RELAXATION"        ||
-                                                                            type_ == "LINESMOOTHING_TRIDI RELAXATION"        ||
-                                                                            type_ == "LINESMOOTHING_TRIDIRELAXATION"         ||
-                                                                            type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION"  ||
-                                                                            type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION"  ||
-                                                                            type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION"   ||
-                                                                            type_ == "LINESMOOTHING_BANDED_RELAXATION"       ||
-                                                                            type_ == "LINESMOOTHING_BANDED RELAXATION"       ||
-                                                                            type_ == "LINESMOOTHING_BANDEDRELAXATION"        ||
-                                                                            type_ == "LINESMOOTHING_BLOCK_RELAXATION"        ||
-                                                                            type_ == "LINESMOOTHING_BLOCK RELAXATION"        ||
-                                                                            type_ == "LINESMOOTHING_BLOCKRELAXATION"         ||
-                                                                            type_ == "TOPOLOGICAL"                           ||
-                                                                            type_ == "AGGREGATE");
-    this->declareConstructionOutcome(!isSupported, "Ifpack2 does not provide the smoother '" + type_ + "'.");
-    if (isSupported)
-      SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
-    Factory::SetParameterList(paramList);
-
-    if (SmootherPrototype::IsSetup()) {
-      // It might be invalid to change parameters after the setup, but it depends entirely on Ifpack implementation.
-      // TODO: I don't know if Ifpack returns an error code or exception or ignore parameters modification in this case...
-      SetPrecParameters();
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2Smoother(const std::string& type, const Teuchos::ParameterList& paramList, const LO& overlap)
+  : type_(type)
+  , overlap_(overlap) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
+  bool isSupported = Ifpack2::Factory::isSupported<tRowMatrix>(type_) || (type_ == "LINESMOOTHING_TRIDI_RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_TRIDI RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_TRIDIRELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BANDEDRELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BLOCK_RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BLOCK RELAXATION" ||
+                                                                          type_ == "LINESMOOTHING_BLOCKRELAXATION" ||
+                                                                          type_ == "TOPOLOGICAL" ||
+                                                                          type_ == "AGGREGATE");
+  this->declareConstructionOutcome(!isSupported, "Ifpack2 does not provide the smoother '" + type_ + "'.");
+  if (isSupported)
+    SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+
+  if (SmootherPrototype::IsSetup()) {
+    // It might be invalid to change parameters after the setup, but it depends entirely on Ifpack implementation.
+    // TODO: I don't know if Ifpack returns an error code or exception or ignore parameters modification in this case...
+    SetPrecParameters();
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetPrecParameters(const Teuchos::ParameterList& list) const {
-    std::string prefix = this->ShortClassName() + ": SetPrecParameters";
-    RCP<TimeMonitor> tM = rcp(new TimeMonitor(*this, prefix, Timings0));
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetPrecParameters(const Teuchos::ParameterList& list) const {
+  std::string prefix       = this->ShortClassName() + ": SetPrecParameters";
+  RCP<TimeMonitor> tM      = rcp(new TimeMonitor(*this, prefix, Timings0));
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
 
-    paramList.setParameters(list);
+  paramList.setParameters(list);
 
-    RCP<ParameterList> precList = this->RemoveFactoriesFromList(this->GetParameterList());
+  RCP<ParameterList> precList = this->RemoveFactoriesFromList(this->GetParameterList());
 
-    // Do we want an Ifpack2 apply timer?
-    precList->set("timer for apply", this->IsPrint(Timings));
+  // Do we want an Ifpack2 apply timer?
+  precList->set("timer for apply", this->IsPrint(Timings));
 
-    if(!precList.is_null() && precList->isParameter("partitioner: type") && precList->get<std::string>("partitioner: type") == "linear" &&
-       !precList->isParameter("partitioner: local parts")) {
-      LO matrixBlockSize = 1;
-      int lclSize = A_->getRangeMap()->getLocalNumElements();
-      RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-      if (!matA.is_null()) {
-        lclSize = matA->getLocalNumRows();
-        matrixBlockSize = matA->GetFixedBlockSize();
-      }
-      precList->set("partitioner: local parts", lclSize / matrixBlockSize);
+  if (!precList.is_null() && precList->isParameter("partitioner: type") && precList->get<std::string>("partitioner: type") == "linear" &&
+      !precList->isParameter("partitioner: local parts")) {
+    LO matrixBlockSize = 1;
+    int lclSize        = A_->getRangeMap()->getLocalNumElements();
+    RCP<Matrix> matA   = rcp_dynamic_cast<Matrix>(A_);
+    if (!matA.is_null()) {
+      lclSize         = matA->getLocalNumRows();
+      matrixBlockSize = matA->GetFixedBlockSize();
     }
-
-    prec_->setParameters(*precList);
-
-    paramList.setParameters(*precList); // what about that??
+    precList->set("partitioner: local parts", lclSize / matrixBlockSize);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    this->Input(currentLevel, "A");
-
-    if (type_ == "LINESMOOTHING_TRIDI_RELAXATION"        ||
-        type_ == "LINESMOOTHING_TRIDI RELAXATION"        ||
-        type_ == "LINESMOOTHING_TRIDIRELAXATION"         ||
-        type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION"  ||
-        type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION"  ||
-        type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION"   ||
-        type_ == "LINESMOOTHING_BANDED_RELAXATION"       ||
-        type_ == "LINESMOOTHING_BANDED RELAXATION"       ||
-        type_ == "LINESMOOTHING_BANDEDRELAXATION"        ||
-        type_ == "LINESMOOTHING_BLOCK_RELAXATION"        ||
-        type_ == "LINESMOOTHING_BLOCK RELAXATION"        ||
-        type_ == "LINESMOOTHING_BLOCKRELAXATION") {
-      this->Input(currentLevel, "CoarseNumZLayers");            // necessary for fallback criterion
-      this->Input(currentLevel, "LineDetection_VertLineIds");   // necessary to feed block smoother
-    }
-    else if (type_ == "BLOCK RELAXATION" ||
+  prec_->setParameters(*precList);
+
+  paramList.setParameters(*precList);  // what about that??
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
+
+  if (type_ == "LINESMOOTHING_TRIDI_RELAXATION" ||
+      type_ == "LINESMOOTHING_TRIDI RELAXATION" ||
+      type_ == "LINESMOOTHING_TRIDIRELAXATION" ||
+      type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
+      type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
+      type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION" ||
+      type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDEDRELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK_RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCKRELAXATION") {
+    this->Input(currentLevel, "CoarseNumZLayers");           // necessary for fallback criterion
+    this->Input(currentLevel, "LineDetection_VertLineIds");  // necessary to feed block smoother
+  } else if (type_ == "BLOCK RELAXATION" ||
              type_ == "BLOCK_RELAXATION" ||
              type_ == "BLOCKRELAXATION" ||
              // Banded
@@ -188,929 +185,906 @@ namespace MueLu {
              type_ == "TRIDIRELAXATION" ||
              type_ == "TRIDIAGONAL_RELAXATION" ||
              type_ == "TRIDIAGONAL RELAXATION" ||
-             type_ == "TRIDIAGONALRELAXATION")
-    {
-      //We need to check for the "partitioner type" = "line"
-      ParameterList precList = this->GetParameterList();
-      if(precList.isParameter("partitioner: type") &&
-         precList.get<std::string>("partitioner: type") == "line") {
-        this->Input(currentLevel, "Coordinates");
-      }
-    }
-    else if (type_ == "TOPOLOGICAL")
-    {
-      // for the topological smoother, we require an element to node map:
-      this->Input(currentLevel, "pcoarsen: element to node map");
-    }
-    else if (type_ == "AGGREGATE")
-    {
-      // Aggregate smoothing needs aggregates
-      this->Input(currentLevel,"Aggregates");
-    }
-    else if (type_ == "HIPTMAIR") {
-      // Hiptmair needs D0 and NodeMatrix
-      this->Input(currentLevel,"NodeMatrix");
-      this->Input(currentLevel,"D0");
+             type_ == "TRIDIAGONALRELAXATION") {
+    // We need to check for the "partitioner type" = "line"
+    ParameterList precList = this->GetParameterList();
+    if (precList.isParameter("partitioner: type") &&
+        precList.get<std::string>("partitioner: type") == "line") {
+      this->Input(currentLevel, "Coordinates");
     }
-
+  } else if (type_ == "TOPOLOGICAL") {
+    // for the topological smoother, we require an element to node map:
+    this->Input(currentLevel, "pcoarsen: element to node map");
+  } else if (type_ == "AGGREGATE") {
+    // Aggregate smoothing needs aggregates
+    this->Input(currentLevel, "Aggregates");
+  } else if (type_ == "HIPTMAIR") {
+    // Hiptmair needs D0 and NodeMatrix
+    this->Input(currentLevel, "NodeMatrix");
+    this->Input(currentLevel, "D0");
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-    A_ = Factory::Get< RCP<Operator> >(currentLevel, "A");
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+  A_                       = Factory::Get<RCP<Operator>>(currentLevel, "A");
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
 
-    // If the user asked us to convert the matrix into BlockCrsMatrix form, we do that now
+  // If the user asked us to convert the matrix into BlockCrsMatrix form, we do that now
 
-    if(paramList.isParameter("smoother: use blockcrsmatrix storage") && paramList.get<bool>("smoother: use blockcrsmatrix storage")) {
-      int blocksize = 1;
-      RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-      if (!matA.is_null())
-        blocksize = matA->GetFixedBlockSize();
-      if (blocksize) {
-        // NOTE: Don't think you can move this out of the if block.  You can't. The test MueLu_MeshTyingBlocked_SimpleSmoother_2dof_medium_MPI_1 will fail
-
-        RCP<CrsMatrixWrap> AcrsWrap = rcp_dynamic_cast<CrsMatrixWrap>(A_);
-        if(AcrsWrap.is_null())
-          throw std::runtime_error("Ifpack2Smoother: Cannot convert matrix A to CrsMatrixWrap object.");
-        RCP<CrsMatrix> Acrs =  AcrsWrap->getCrsMatrix();
-        if(Acrs.is_null())
-          throw std::runtime_error("Ifpack2Smoother: Cannot extract CrsMatrix from matrix A.");
-        RCP<TpetraCrsMatrix> At = rcp_dynamic_cast<TpetraCrsMatrix>(Acrs);
-        if(At.is_null()) {
-          if(!Xpetra::Helpers<Scalar,LO,GO,Node>::isTpetraBlockCrs(matA))
-            throw std::runtime_error("Ifpack2Smoother: Cannot extract CrsMatrix or BlockCrsMatrix from matrix A.");
-          this->GetOStream(Statistics0) << "Ifpack2Smoother: Using (native) BlockCrsMatrix storage with blocksize "<<blocksize<<std::endl;
-          paramList.remove("smoother: use blockcrsmatrix storage");
-        }
-        else {
-          RCP<Tpetra::BlockCrsMatrix<Scalar, LO, GO, Node> > blockCrs = Tpetra::convertToBlockCrsMatrix(*At->getTpetra_CrsMatrix(),blocksize);
-          RCP<CrsMatrix> blockCrs_as_crs  = rcp(new TpetraBlockCrsMatrix(blockCrs));
-          RCP<CrsMatrixWrap> blockWrap = rcp(new CrsMatrixWrap(blockCrs_as_crs));
-          A_ = blockWrap;
-          this->GetOStream(Statistics0) << "Ifpack2Smoother: Using BlockCrsMatrix storage with blocksize "<<blocksize<<std::endl;
-          
-          paramList.remove("smoother: use blockcrsmatrix storage");
-        }
+  if (paramList.isParameter("smoother: use blockcrsmatrix storage") && paramList.get<bool>("smoother: use blockcrsmatrix storage")) {
+    int blocksize    = 1;
+    RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
+    if (!matA.is_null())
+      blocksize = matA->GetFixedBlockSize();
+    if (blocksize) {
+      // NOTE: Don't think you can move this out of the if block.  You can't. The test MueLu_MeshTyingBlocked_SimpleSmoother_2dof_medium_MPI_1 will fail
+
+      RCP<CrsMatrixWrap> AcrsWrap = rcp_dynamic_cast<CrsMatrixWrap>(A_);
+      if (AcrsWrap.is_null())
+        throw std::runtime_error("Ifpack2Smoother: Cannot convert matrix A to CrsMatrixWrap object.");
+      RCP<CrsMatrix> Acrs = AcrsWrap->getCrsMatrix();
+      if (Acrs.is_null())
+        throw std::runtime_error("Ifpack2Smoother: Cannot extract CrsMatrix from matrix A.");
+      RCP<TpetraCrsMatrix> At = rcp_dynamic_cast<TpetraCrsMatrix>(Acrs);
+      if (At.is_null()) {
+        if (!Xpetra::Helpers<Scalar, LO, GO, Node>::isTpetraBlockCrs(matA))
+          throw std::runtime_error("Ifpack2Smoother: Cannot extract CrsMatrix or BlockCrsMatrix from matrix A.");
+        this->GetOStream(Statistics0) << "Ifpack2Smoother: Using (native) BlockCrsMatrix storage with blocksize " << blocksize << std::endl;
+        paramList.remove("smoother: use blockcrsmatrix storage");
+      } else {
+        RCP<Tpetra::BlockCrsMatrix<Scalar, LO, GO, Node>> blockCrs = Tpetra::convertToBlockCrsMatrix(*At->getTpetra_CrsMatrix(), blocksize);
+        RCP<CrsMatrix> blockCrs_as_crs                             = rcp(new TpetraBlockCrsMatrix(blockCrs));
+        RCP<CrsMatrixWrap> blockWrap                               = rcp(new CrsMatrixWrap(blockCrs_as_crs));
+        A_                                                         = blockWrap;
+        this->GetOStream(Statistics0) << "Ifpack2Smoother: Using BlockCrsMatrix storage with blocksize " << blocksize << std::endl;
+
+        paramList.remove("smoother: use blockcrsmatrix storage");
       }
     }
+  }
 
-    if      (type_ == "SCHWARZ")
-      SetupSchwarz(currentLevel);
-
-    else if (type_ == "LINESMOOTHING_TRIDI_RELAXATION"       ||
-             type_ == "LINESMOOTHING_TRIDI RELAXATION"       ||
-             type_ == "LINESMOOTHING_TRIDIRELAXATION"        ||
-             type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
-             type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
-             type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION"  ||
-             type_ == "LINESMOOTHING_BANDED_RELAXATION"      ||
-             type_ == "LINESMOOTHING_BANDED RELAXATION"      ||
-             type_ == "LINESMOOTHING_BANDEDRELAXATION"       ||
-             type_ == "LINESMOOTHING_BLOCK_RELAXATION"       ||
-             type_ == "LINESMOOTHING_BLOCK RELAXATION"       ||
-             type_ == "LINESMOOTHING_BLOCKRELAXATION")
-      SetupLineSmoothing(currentLevel);
-
-    else if (type_ == "BLOCK_RELAXATION" ||
-             type_ == "BLOCK RELAXATION" ||
-             type_ == "BLOCKRELAXATION" ||
-             // Banded
-             type_ == "BANDED_RELAXATION" ||
-             type_ == "BANDED RELAXATION" ||
-             type_ == "BANDEDRELAXATION" ||
-             // Tridiagonal
-             type_ == "TRIDI_RELAXATION" ||
-             type_ == "TRIDI RELAXATION" ||
-             type_ == "TRIDIRELAXATION" ||
-             type_ == "TRIDIAGONAL_RELAXATION" ||
-             type_ == "TRIDIAGONAL RELAXATION" ||
-             type_ == "TRIDIAGONALRELAXATION")
-      SetupBlockRelaxation(currentLevel);
-
-    else if (type_ == "CHEBYSHEV")
-      SetupChebyshev(currentLevel);
-
-    else if (type_ == "TOPOLOGICAL")
-    {
+  if (type_ == "SCHWARZ")
+    SetupSchwarz(currentLevel);
+
+  else if (type_ == "LINESMOOTHING_TRIDI_RELAXATION" ||
+           type_ == "LINESMOOTHING_TRIDI RELAXATION" ||
+           type_ == "LINESMOOTHING_TRIDIRELAXATION" ||
+           type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
+           type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
+           type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION" ||
+           type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+           type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+           type_ == "LINESMOOTHING_BANDEDRELAXATION" ||
+           type_ == "LINESMOOTHING_BLOCK_RELAXATION" ||
+           type_ == "LINESMOOTHING_BLOCK RELAXATION" ||
+           type_ == "LINESMOOTHING_BLOCKRELAXATION")
+    SetupLineSmoothing(currentLevel);
+
+  else if (type_ == "BLOCK_RELAXATION" ||
+           type_ == "BLOCK RELAXATION" ||
+           type_ == "BLOCKRELAXATION" ||
+           // Banded
+           type_ == "BANDED_RELAXATION" ||
+           type_ == "BANDED RELAXATION" ||
+           type_ == "BANDEDRELAXATION" ||
+           // Tridiagonal
+           type_ == "TRIDI_RELAXATION" ||
+           type_ == "TRIDI RELAXATION" ||
+           type_ == "TRIDIRELAXATION" ||
+           type_ == "TRIDIAGONAL_RELAXATION" ||
+           type_ == "TRIDIAGONAL RELAXATION" ||
+           type_ == "TRIDIAGONALRELAXATION")
+    SetupBlockRelaxation(currentLevel);
+
+  else if (type_ == "CHEBYSHEV")
+    SetupChebyshev(currentLevel);
+
+  else if (type_ == "TOPOLOGICAL") {
 #ifdef HAVE_MUELU_INTREPID2
-      SetupTopological(currentLevel);
+    SetupTopological(currentLevel);
 #else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "'TOPOLOGICAL' smoother choice requires Intrepid2");
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "'TOPOLOGICAL' smoother choice requires Intrepid2");
 #endif
-    }
-    else if (type_ == "AGGREGATE")
-      SetupAggregate(currentLevel);
+  } else if (type_ == "AGGREGATE")
+    SetupAggregate(currentLevel);
 
-    else if (type_ == "HIPTMAIR")
-      SetupHiptmair(currentLevel);
+  else if (type_ == "HIPTMAIR")
+    SetupHiptmair(currentLevel);
 
-    else
-    {
-      SetupGeneric(currentLevel);
-    }
-
-    SmootherPrototype::IsSetup(true);
-
-    this->GetOStream(Statistics1) << description() << std::endl;
+  else {
+    SetupGeneric(currentLevel);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupSchwarz(Level& /* currentLevel */) {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
-
-    bool reusePreconditioner = false;
-    if (this->IsSetup() == true) {
-      // Reuse the constructed preconditioner
-      this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupSchwarz(): Setup() has already been called, assuming reuse" << std::endl;
-
-      bool isTRowMatrix = true;
-      RCP<const tRowMatrix> tA;
-      try {
-        tA = Utilities::Op2NonConstTpetraRow(A_);
-      } catch (Exceptions::BadCast&) {
-        isTRowMatrix = false;
-      }
+  SmootherPrototype::IsSetup(true);
 
-      RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix> > prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix> >(prec_);
-      if (!prec.is_null() && isTRowMatrix) {
-        prec->setMatrix(tA);
-        reusePreconditioner = true;
-      } else {
-        this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupSchwarz(): reuse of this type is not available "
-            "(either failed cast to CanChangeMatrix, or to Tpetra Row Matrix), reverting to full construction" << std::endl;
-      }
-    }
+  this->GetOStream(Statistics1) << description() << std::endl;
+}
 
-    if (!reusePreconditioner) {
-      ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-
-      bool isBlockedMatrix = false;
-      RCP<Matrix> merged2Mat;
-
-      std::string sublistName = "subdomain solver parameters";
-      if (paramList.isSublist(sublistName)) {
-        // If we are doing "user" partitioning, we assume that what the user
-        // really wants to do is make tiny little subdomains with one row
-        // assigned to each subdomain. The rows used for these little
-        // subdomains correspond to those in the 2nd block row. Then,
-        // if we overlap these mini-subdomains, we will do something that
-        // looks like Vanka (grabbing all velocities associated with each
-        // each pressure unknown). In addition, we put all Dirichlet points
-        // as a little mini-domain.
-        ParameterList& subList = paramList.sublist(sublistName);
-
-        std::string partName = "partitioner: type";
-        // Pretty sure no one has been using this. Unfortunately, old if
-        // statement (which checked for equality with "user") prevented
-        // anyone from employing other types of Ifpack2 user partition
-        // options. Leaving this and switching if to "vanka user" just
-        // in case some day someone might want to use this.
-        if (subList.isParameter(partName) && subList.get<std::string>(partName) == "vanka user") {
-          isBlockedMatrix = true;
-
-          RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-          TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
-                                     "Matrix A must be of type BlockedCrsMatrix.");
-
-          size_t numVels = bA->getMatrix(0,0)->getLocalNumRows();
-          size_t numPres = bA->getMatrix(1,0)->getLocalNumRows();
-          size_t numRows = rcp_dynamic_cast<Matrix>(A_, true)->getLocalNumRows();
-
-          ArrayRCP<LocalOrdinal> blockSeeds(numRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
-
-          size_t numBlocks = 0;
-          for (size_t rowOfB = numVels; rowOfB < numVels+numPres; ++rowOfB)
-            blockSeeds[rowOfB] = numBlocks++;
-
-          RCP<BlockedCrsMatrix> bA2 = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-          TEUCHOS_TEST_FOR_EXCEPTION(bA2.is_null(), Exceptions::BadCast,
-                                     "Matrix A must be of type BlockedCrsMatrix.");
-
-          merged2Mat = bA2->Merge();
-
-          // Add Dirichlet rows to the list of seeds
-          ArrayRCP<const bool> boundaryNodes = Utilities::DetectDirichletRows(*merged2Mat, 0.0);
-          bool haveBoundary = false;
-          for (LO i = 0; i < boundaryNodes.size(); i++)
-            if (boundaryNodes[i]) {
-              // FIXME:
-              // 1. would not this [] overlap with some in the previos blockSeed loop?
-              // 2. do we need to distinguish between pressure and velocity Dirichlet b.c.
-              blockSeeds[i] = numBlocks;
-              haveBoundary = true;
-            }
-          if (haveBoundary)
-            numBlocks++;
-
-          subList.set("partitioner: type","user");
-          subList.set("partitioner: map",         blockSeeds);
-          subList.set("partitioner: local parts", as<int>(numBlocks));
-
-        } else {
-          RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-          if (!bA.is_null()) {
-            isBlockedMatrix = true;
-            merged2Mat = bA->Merge();
-          }
-        }
-      }
-
-      RCP<const tRowMatrix> tA;
-      if (isBlockedMatrix == true) tA = Utilities::Op2NonConstTpetraRow(merged2Mat);
-      else                         tA = Utilities::Op2NonConstTpetraRow(A_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupSchwarz(Level& /* currentLevel */) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
 
-      prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-      SetPrecParameters();
+  bool reusePreconditioner = false;
+  if (this->IsSetup() == true) {
+    // Reuse the constructed preconditioner
+    this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupSchwarz(): Setup() has already been called, assuming reuse" << std::endl;
 
-      prec_->initialize();
+    bool isTRowMatrix = true;
+    RCP<const tRowMatrix> tA;
+    try {
+      tA = Utilities::Op2NonConstTpetraRow(A_);
+    } catch (Exceptions::BadCast&) {
+      isTRowMatrix = false;
     }
 
-    prec_->compute();
+    RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix>> prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix>>(prec_);
+    if (!prec.is_null() && isTRowMatrix) {
+      prec->setMatrix(tA);
+      reusePreconditioner = true;
+    } else {
+      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupSchwarz(): reuse of this type is not available "
+                                     "(either failed cast to CanChangeMatrix, or to Tpetra Row Matrix), reverting to full construction"
+                                  << std::endl;
+    }
   }
 
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupAggregate(Level& currentLevel) {
-
+  if (!reusePreconditioner) {
     ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
 
-    if (this->IsSetup() == true) {
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupAggregate(): Setup() has already been called" << std::endl;
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupAggregate(): reuse of this type is not available, reverting to full construction" << std::endl;
-    }
-
-    this->GetOStream(Statistics0) << "Ifpack2Smoother: Using Aggregate Smoothing"<<std::endl;
-
-    RCP<Aggregates> aggregates = Factory::Get<RCP<Aggregates> >(currentLevel,"Aggregates");
+    bool isBlockedMatrix = false;
+    RCP<Matrix> merged2Mat;
+
+    std::string sublistName = "subdomain solver parameters";
+    if (paramList.isSublist(sublistName)) {
+      // If we are doing "user" partitioning, we assume that what the user
+      // really wants to do is make tiny little subdomains with one row
+      // assigned to each subdomain. The rows used for these little
+      // subdomains correspond to those in the 2nd block row. Then,
+      // if we overlap these mini-subdomains, we will do something that
+      // looks like Vanka (grabbing all velocities associated with each
+      // each pressure unknown). In addition, we put all Dirichlet points
+      // as a little mini-domain.
+      ParameterList& subList = paramList.sublist(sublistName);
+
+      std::string partName = "partitioner: type";
+      // Pretty sure no one has been using this. Unfortunately, old if
+      // statement (which checked for equality with "user") prevented
+      // anyone from employing other types of Ifpack2 user partition
+      // options. Leaving this and switching if to "vanka user" just
+      // in case some day someone might want to use this.
+      if (subList.isParameter(partName) && subList.get<std::string>(partName) == "vanka user") {
+        isBlockedMatrix = true;
+
+        RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+        TEUCHOS_TEST_FOR_EXCEPTION(bA.is_null(), Exceptions::BadCast,
+                                   "Matrix A must be of type BlockedCrsMatrix.");
+
+        size_t numVels = bA->getMatrix(0, 0)->getLocalNumRows();
+        size_t numPres = bA->getMatrix(1, 0)->getLocalNumRows();
+        size_t numRows = rcp_dynamic_cast<Matrix>(A_, true)->getLocalNumRows();
+
+        ArrayRCP<LocalOrdinal> blockSeeds(numRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
+
+        size_t numBlocks = 0;
+        for (size_t rowOfB = numVels; rowOfB < numVels + numPres; ++rowOfB)
+          blockSeeds[rowOfB] = numBlocks++;
+
+        RCP<BlockedCrsMatrix> bA2 = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+        TEUCHOS_TEST_FOR_EXCEPTION(bA2.is_null(), Exceptions::BadCast,
+                                   "Matrix A must be of type BlockedCrsMatrix.");
+
+        merged2Mat = bA2->Merge();
+
+        // Add Dirichlet rows to the list of seeds
+        ArrayRCP<const bool> boundaryNodes = Utilities::DetectDirichletRows(*merged2Mat, 0.0);
+        bool haveBoundary                  = false;
+        for (LO i = 0; i < boundaryNodes.size(); i++)
+          if (boundaryNodes[i]) {
+            // FIXME:
+            // 1. would not this [] overlap with some in the previos blockSeed loop?
+            // 2. do we need to distinguish between pressure and velocity Dirichlet b.c.
+            blockSeeds[i] = numBlocks;
+            haveBoundary  = true;
+          }
+        if (haveBoundary)
+          numBlocks++;
 
-    RCP<const LOMultiVector> vertex2AggId = aggregates->GetVertex2AggId();
-    ArrayRCP<LO> aggregate_ids = rcp_const_cast<LOMultiVector>(vertex2AggId)->getDataNonConst(0);
-    ArrayRCP<LO> dof_ids;
+        subList.set("partitioner: type", "user");
+        subList.set("partitioner: map", blockSeeds);
+        subList.set("partitioner: local parts", as<int>(numBlocks));
 
-    // We need to unamalgamate, if the FixedBlockSize > 1
-    LO blocksize = 1;
-    RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-    if (!matA.is_null())
-      blocksize = matA->GetFixedBlockSize();
-    if(blocksize > 1) {
-      dof_ids.resize(aggregate_ids.size() * blocksize);
-      for(LO i=0; i<(LO)aggregate_ids.size(); i++) {
-        for(LO j=0; j<(LO)blocksize; j++)
-          dof_ids[i*blocksize+j] = aggregate_ids[i];
+      } else {
+        RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+        if (!bA.is_null()) {
+          isBlockedMatrix = true;
+          merged2Mat      = bA->Merge();
+        }
       }
     }
-    else {
-      dof_ids = aggregate_ids;
-    }
-        
-    
-    paramList.set("partitioner: map", dof_ids);
-    paramList.set("partitioner: type", "user");
-    paramList.set("partitioner: overlap", 0);
-    paramList.set("partitioner: local parts", (int)aggregates->GetNumAggregates());
-
-    RCP<const Tpetra::RowMatrix<SC, LO, GO, NO> > tA = Utilities::Op2NonConstTpetraRow(A_);
-    
-    type_ = "BLOCKRELAXATION";
-    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-    SetPrecParameters();
-    prec_->initialize();
-    prec_->compute();
-  }
 
-#ifdef HAVE_MUELU_INTREPID2
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupTopological(Level& currentLevel) {
-    /*
-     
-     basic notion:
-     
-     Look for user input indicating topo dimension, something like "topological domain type: {node|edge|face}"
-     Call something like what you can find in Poisson example line 1180 to set seeds for a smoother
-     
-     */
-    if (this->IsSetup() == true) {
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupTopological(): Setup() has already been called" << std::endl;
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupTopological(): reuse of this type is not available, reverting to full construction" << std::endl;
-    }
-    
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-    
-    typedef typename Node::device_type::execution_space ES;
-    
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCO; //
-    
-    LocalOrdinal  lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
-    
-    using namespace std;
-    
-    const Teuchos::RCP<FCO> elemToNode = Factory::Get<Teuchos::RCP<FCO> >(currentLevel,"pcoarsen: element to node map");
-    
-    string basisString = paramList.get<string>("pcoarsen: hi basis");
-    int degree;
-    // NOTE: To make sure Stokhos works we only instantiate these guys with double.  There's a lot
-    // of stuff in the guts of Intrepid2 that doesn't play well with Stokhos as of yet.  Here, we only
-    // care about the assignment of basis ordinals to topological entities, so this code is actually
-    // independent of the Scalar type--hard-coding double here won't hurt us.
-    auto basis = MueLuIntrepid::BasisFactory<double,ES>(basisString, degree);
-    
-    string topologyTypeString = paramList.get<string>("smoother: neighborhood type");
-    int dimension;
-    if (topologyTypeString == "node")
-      dimension = 0;
-    else if (topologyTypeString == "edge")
-      dimension = 1;
-    else if (topologyTypeString == "face")
-      dimension = 2;
-    else if (topologyTypeString == "cell")
-      dimension = basis->getBaseCellTopology().getDimension();
+    RCP<const tRowMatrix> tA;
+    if (isBlockedMatrix == true)
+      tA = Utilities::Op2NonConstTpetraRow(merged2Mat);
     else
-      TEUCHOS_TEST_FOR_EXCEPTION(true,std::invalid_argument,"Unrecognized smoother neighborhood type.  Supported types are node, edge, face.");
-    RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_, true);
-    vector<vector<LocalOrdinal>> seeds;
-    MueLuIntrepid::FindGeometricSeedOrdinals(basis, *elemToNode, seeds, *matA->getRowMap(), *matA->getColMap());
-    
-    // Ifpack2 wants the seeds in an array of the same length as the number of local elements,
-    // with local partition #s marked for the ones that are seeds, and invalid for the rest
-    int myNodeCount = matA->getRowMap()->getLocalNumElements();
-    ArrayRCP<LocalOrdinal> nodeSeeds(myNodeCount,lo_invalid);
-    int localPartitionNumber = 0;
-    for (LocalOrdinal seed : seeds[dimension])
-    {
-      nodeSeeds[seed] = localPartitionNumber++;
-    }
-    
-    paramList.remove("smoother: neighborhood type");
-    paramList.remove("pcoarsen: hi basis");
-    
-    paramList.set("partitioner: map", nodeSeeds);
-    paramList.set("partitioner: type", "user");
-    paramList.set("partitioner: overlap", 1);
-    paramList.set("partitioner: local parts", int(seeds[dimension].size()));
-
-    RCP<const Tpetra::RowMatrix<SC, LO, GO, NO> > tA = Utilities::Op2NonConstTpetraRow(A_);
-    
-    type_ = "BLOCKRELAXATION";
+      tA = Utilities::Op2NonConstTpetraRow(A_);
+
     prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
     SetPrecParameters();
+
     prec_->initialize();
-    prec_->compute();
   }
-#endif
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupLineSmoothing(Level& currentLevel) {
-    if (this->IsSetup() == true) {
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupLineSmoothing(): Setup() has already been called" << std::endl;
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupLineSmoothing(): reuse of this type is not available, reverting to full construction" << std::endl;
-    }
-
-    ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
+  prec_->compute();
+}
 
-    LO CoarseNumZLayers = Factory::Get<LO>(currentLevel,"CoarseNumZLayers");
-    if (CoarseNumZLayers > 0) {
-      Teuchos::ArrayRCP<LO> TVertLineIdSmoo = Factory::Get< Teuchos::ArrayRCP<LO> >(currentLevel, "LineDetection_VertLineIds");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupAggregate(Level& currentLevel) {
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
 
-      // determine number of local parts
-      LO maxPart = 0;
-      for(size_t k = 0; k < Teuchos::as<size_t>(TVertLineIdSmoo.size()); k++) {
-        if(maxPart < TVertLineIdSmoo[k]) maxPart = TVertLineIdSmoo[k];
-      }
-      RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_, true);
-      size_t numLocalRows = matA->getLocalNumRows();
-
-      TEUCHOS_TEST_FOR_EXCEPTION(numLocalRows % TVertLineIdSmoo.size() != 0, Exceptions::RuntimeError,
-        "MueLu::Ifpack2Smoother::Setup(): the number of local nodes is incompatible with the TVertLineIdsSmoo.");
-
-      //actualDofsPerNode is the actual number of matrix rows per mesh element.
-      //It is encoded in either the MueLu Level, or in the Xpetra matrix block size.
-      //This value is needed by Ifpack2 to do decoupled block relaxation.
-      int actualDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
-      LO matrixBlockSize = matA->GetFixedBlockSize();
-      if(matrixBlockSize > 1 && actualDofsPerNode > 1)
-      {
-        TEUCHOS_TEST_FOR_EXCEPTION(actualDofsPerNode != matrixBlockSize, Exceptions::RuntimeError,
-            "MueLu::Ifpack2Smoother::Setup(): A is a block matrix but its block size and DOFs/node from partitioner disagree");
-      }
-      else if(matrixBlockSize > 1)
-      {
-        actualDofsPerNode = matrixBlockSize;
-      }
-      myparamList.set("partitioner: PDE equations", actualDofsPerNode);
+  if (this->IsSetup() == true) {
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupAggregate(): Setup() has already been called" << std::endl;
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupAggregate(): reuse of this type is not available, reverting to full construction" << std::endl;
+  }
 
-      if (numLocalRows == Teuchos::as<size_t>(TVertLineIdSmoo.size())) {
-        myparamList.set("partitioner: type","user");
-        myparamList.set("partitioner: map",TVertLineIdSmoo);
-        myparamList.set("partitioner: local parts",maxPart+1);
-      } else {
-        // we assume a constant number of DOFs per node
-        size_t numDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
-
-        // Create a new Teuchos::ArrayRCP<LO> of size numLocalRows and fill it with the corresponding information
-        Teuchos::ArrayRCP<LO> partitionerMap(numLocalRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
-        for (size_t blockRow = 0; blockRow < Teuchos::as<size_t>(TVertLineIdSmoo.size()); ++blockRow)
-          for (size_t dof = 0; dof < numDofsPerNode; dof++)
-            partitionerMap[blockRow * numDofsPerNode + dof] = TVertLineIdSmoo[blockRow];
-        myparamList.set("partitioner: type","user");
-        myparamList.set("partitioner: map",partitionerMap);
-        myparamList.set("partitioner: local parts",maxPart + 1);
-      }
+  this->GetOStream(Statistics0) << "Ifpack2Smoother: Using Aggregate Smoothing" << std::endl;
 
-      if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
-          type_ == "LINESMOOTHING_BANDED RELAXATION" ||
-          type_ == "LINESMOOTHING_BANDEDRELAXATION")
-        type_ = "BANDEDRELAXATION";
-      else if (type_ == "LINESMOOTHING_TRIDI_RELAXATION"       ||
-               type_ == "LINESMOOTHING_TRIDI RELAXATION"       ||
-               type_ == "LINESMOOTHING_TRIDIRELAXATION"        ||
-               type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
-               type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
-               type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION")
-        type_ = "TRIDIAGONALRELAXATION";
-      else
-        type_ = "BLOCKRELAXATION";
-    } else {
-      // line detection failed -> fallback to point-wise relaxation
-      this->GetOStream(Runtime0) << "Line detection failed: fall back to point-wise relaxation" << std::endl;
-      myparamList.remove("partitioner: type",false);
-      myparamList.remove("partitioner: map", false);
-      myparamList.remove("partitioner: local parts",false);
-      type_ = "RELAXATION";
-    }
+  RCP<Aggregates> aggregates = Factory::Get<RCP<Aggregates>>(currentLevel, "Aggregates");
 
-    RCP<const Tpetra::RowMatrix<SC, LO, GO, NO> > tA = Utilities::Op2NonConstTpetraRow(A_);
+  RCP<const LOMultiVector> vertex2AggId = aggregates->GetVertex2AggId();
+  ArrayRCP<LO> aggregate_ids            = rcp_const_cast<LOMultiVector>(vertex2AggId)->getDataNonConst(0);
+  ArrayRCP<LO> dof_ids;
 
-    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-    SetPrecParameters();
-    prec_->initialize();
-    prec_->compute();
+  // We need to unamalgamate, if the FixedBlockSize > 1
+  LO blocksize     = 1;
+  RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
+  if (!matA.is_null())
+    blocksize = matA->GetFixedBlockSize();
+  if (blocksize > 1) {
+    dof_ids.resize(aggregate_ids.size() * blocksize);
+    for (LO i = 0; i < (LO)aggregate_ids.size(); i++) {
+      for (LO j = 0; j < (LO)blocksize; j++)
+        dof_ids[i * blocksize + j] = aggregate_ids[i];
+    }
+  } else {
+    dof_ids = aggregate_ids;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupBlockRelaxation(Level& currentLevel) {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
+  paramList.set("partitioner: map", dof_ids);
+  paramList.set("partitioner: type", "user");
+  paramList.set("partitioner: overlap", 0);
+  paramList.set("partitioner: local parts", (int)aggregates->GetNumAggregates());
 
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    if (!bA.is_null())
-      A_ = bA->Merge();
+  RCP<const Tpetra::RowMatrix<SC, LO, GO, NO>> tA = Utilities::Op2NonConstTpetraRow(A_);
 
-    RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
+  type_ = "BLOCKRELAXATION";
+  prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+  SetPrecParameters();
+  prec_->initialize();
+  prec_->compute();
+}
 
-    bool reusePreconditioner = false;
-    if (this->IsSetup() == true) {
-      // Reuse the constructed preconditioner
-      this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupBlockRelaxation(): Setup() has already been called, assuming reuse" << std::endl;
+#ifdef HAVE_MUELU_INTREPID2
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupTopological(Level& currentLevel) {
+  /*
 
-      RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix> > prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix> >(prec_);
-      if (!prec.is_null()) {
-        prec->setMatrix(tA);
-        reusePreconditioner = true;
-      } else {
-        this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupBlockRelaxation(): reuse of this type is not available (failed cast to CanChangeMatrix), "
-            "reverting to full construction" << std::endl;
-      }
-    }
+   basic notion:
 
-    if (!reusePreconditioner) {
-      ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
-      myparamList.print();
-      if(myparamList.isParameter("partitioner: type") &&
-         myparamList.get<std::string>("partitioner: type") == "line") {
-        Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > xCoordinates =
-          Factory::Get<Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > >(currentLevel, "Coordinates");
-        Teuchos::RCP<Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates = Teuchos::rcpFromRef(Xpetra::toTpetra<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>(*xCoordinates));
+   Look for user input indicating topo dimension, something like "topological domain type: {node|edge|face}"
+   Call something like what you can find in Poisson example line 1180 to set seeds for a smoother
 
-        RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-        size_t lclSize = A_->getRangeMap()->getLocalNumElements();
-        if (!matA.is_null())
-          lclSize = matA->getLocalNumRows();
-        size_t numDofsPerNode = lclSize / xCoordinates->getMap()->getLocalNumElements();
-        myparamList.set("partitioner: coordinates", coordinates);
-        myparamList.set("partitioner: PDE equations", (int) numDofsPerNode);
-      }
-
-      prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-      SetPrecParameters();
-      prec_->initialize();
-    }
-
-    prec_->compute();
+   */
+  if (this->IsSetup() == true) {
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupTopological(): Setup() has already been called" << std::endl;
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupTopological(): reuse of this type is not available, reverting to full construction" << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupChebyshev(Level& currentLevel) {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
-    using STS = Teuchos::ScalarTraits<SC>;
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    if (!bA.is_null())
-      A_ = bA->Merge();
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+
+  typedef typename Node::device_type::execution_space ES;
+
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCO;  //
+
+  LocalOrdinal lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
+
+  using namespace std;
+
+  const Teuchos::RCP<FCO> elemToNode = Factory::Get<Teuchos::RCP<FCO>>(currentLevel, "pcoarsen: element to node map");
+
+  string basisString = paramList.get<string>("pcoarsen: hi basis");
+  int degree;
+  // NOTE: To make sure Stokhos works we only instantiate these guys with double.  There's a lot
+  // of stuff in the guts of Intrepid2 that doesn't play well with Stokhos as of yet.  Here, we only
+  // care about the assignment of basis ordinals to topological entities, so this code is actually
+  // independent of the Scalar type--hard-coding double here won't hurt us.
+  auto basis = MueLuIntrepid::BasisFactory<double, ES>(basisString, degree);
+
+  string topologyTypeString = paramList.get<string>("smoother: neighborhood type");
+  int dimension;
+  if (topologyTypeString == "node")
+    dimension = 0;
+  else if (topologyTypeString == "edge")
+    dimension = 1;
+  else if (topologyTypeString == "face")
+    dimension = 2;
+  else if (topologyTypeString == "cell")
+    dimension = basis->getBaseCellTopology().getDimension();
+  else
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "Unrecognized smoother neighborhood type.  Supported types are node, edge, face.");
+  RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_, true);
+  vector<vector<LocalOrdinal>> seeds;
+  MueLuIntrepid::FindGeometricSeedOrdinals(basis, *elemToNode, seeds, *matA->getRowMap(), *matA->getColMap());
+
+  // Ifpack2 wants the seeds in an array of the same length as the number of local elements,
+  // with local partition #s marked for the ones that are seeds, and invalid for the rest
+  int myNodeCount = matA->getRowMap()->getLocalNumElements();
+  ArrayRCP<LocalOrdinal> nodeSeeds(myNodeCount, lo_invalid);
+  int localPartitionNumber = 0;
+  for (LocalOrdinal seed : seeds[dimension]) {
+    nodeSeeds[seed] = localPartitionNumber++;
+  }
 
-    RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
+  paramList.remove("smoother: neighborhood type");
+  paramList.remove("pcoarsen: hi basis");
 
-    bool reusePreconditioner = false;
+  paramList.set("partitioner: map", nodeSeeds);
+  paramList.set("partitioner: type", "user");
+  paramList.set("partitioner: overlap", 1);
+  paramList.set("partitioner: local parts", int(seeds[dimension].size()));
 
-    if (this->IsSetup() == true) {
-      // Reuse the constructed preconditioner
-      this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupChebyshev(): Setup() has already been called, assuming reuse" << std::endl;
+  RCP<const Tpetra::RowMatrix<SC, LO, GO, NO>> tA = Utilities::Op2NonConstTpetraRow(A_);
 
-      RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix> > prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix> >(prec_);
-      if (!prec.is_null()) {
-        prec->setMatrix(tA);
-        reusePreconditioner = true;
-      } else {
-        this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupChebyshev(): reuse of this type is not available (failed cast to CanChangeMatrix), "
-            "reverting to full construction" << std::endl;
-      }
-    }
+  type_ = "BLOCKRELAXATION";
+  prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+  SetPrecParameters();
+  prec_->initialize();
+  prec_->compute();
+}
+#endif
 
-    // Take care of the eigenvalues
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-    SC negone = -STS::one();
-    SC lambdaMax = SetupChebyshevEigenvalues(currentLevel,"A","",paramList);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupLineSmoothing(Level& currentLevel) {
+  if (this->IsSetup() == true) {
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupLineSmoothing(): Setup() has already been called" << std::endl;
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupLineSmoothing(): reuse of this type is not available, reverting to full construction" << std::endl;
+  }
 
+  ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
 
-    if (!reusePreconditioner) {
-      prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-      SetPrecParameters();
-      {
-        SubFactoryMonitor(*this, "Preconditioner init", currentLevel);
-        prec_->initialize();
-      }
-    } else
-      SetPrecParameters();
+  LO CoarseNumZLayers = Factory::Get<LO>(currentLevel, "CoarseNumZLayers");
+  if (CoarseNumZLayers > 0) {
+    Teuchos::ArrayRCP<LO> TVertLineIdSmoo = Factory::Get<Teuchos::ArrayRCP<LO>>(currentLevel, "LineDetection_VertLineIds");
 
-    {
-      SubFactoryMonitor(*this, "Preconditioner compute", currentLevel);
-      prec_->compute();
+    // determine number of local parts
+    LO maxPart = 0;
+    for (size_t k = 0; k < Teuchos::as<size_t>(TVertLineIdSmoo.size()); k++) {
+      if (maxPart < TVertLineIdSmoo[k]) maxPart = TVertLineIdSmoo[k];
     }
+    RCP<Matrix> matA    = rcp_dynamic_cast<Matrix>(A_, true);
+    size_t numLocalRows = matA->getLocalNumRows();
+
+    TEUCHOS_TEST_FOR_EXCEPTION(numLocalRows % TVertLineIdSmoo.size() != 0, Exceptions::RuntimeError,
+                               "MueLu::Ifpack2Smoother::Setup(): the number of local nodes is incompatible with the TVertLineIdsSmoo.");
+
+    // actualDofsPerNode is the actual number of matrix rows per mesh element.
+    // It is encoded in either the MueLu Level, or in the Xpetra matrix block size.
+    // This value is needed by Ifpack2 to do decoupled block relaxation.
+    int actualDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
+    LO matrixBlockSize    = matA->GetFixedBlockSize();
+    if (matrixBlockSize > 1 && actualDofsPerNode > 1) {
+      TEUCHOS_TEST_FOR_EXCEPTION(actualDofsPerNode != matrixBlockSize, Exceptions::RuntimeError,
+                                 "MueLu::Ifpack2Smoother::Setup(): A is a block matrix but its block size and DOFs/node from partitioner disagree");
+    } else if (matrixBlockSize > 1) {
+      actualDofsPerNode = matrixBlockSize;
+    }
+    myparamList.set("partitioner: PDE equations", actualDofsPerNode);
 
-    if (lambdaMax == negone) {
-      typedef Tpetra::RowMatrix<SC, LO, GO, NO> MatrixType;
-
-      Teuchos::RCP<Ifpack2::Chebyshev<MatrixType> > chebyPrec = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType> >(prec_);
-      if (chebyPrec != Teuchos::null) {
-        lambdaMax = chebyPrec->getLambdaMaxForApply();
-        RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-        if (!matA.is_null())
-          matA->SetMaxEigenvalueEstimate(lambdaMax);
-        this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack2)" << " = " << lambdaMax << std::endl;
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax == negone, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): no maximum eigenvalue estimate");
+    if (numLocalRows == Teuchos::as<size_t>(TVertLineIdSmoo.size())) {
+      myparamList.set("partitioner: type", "user");
+      myparamList.set("partitioner: map", TVertLineIdSmoo);
+      myparamList.set("partitioner: local parts", maxPart + 1);
+    } else {
+      // we assume a constant number of DOFs per node
+      size_t numDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
+
+      // Create a new Teuchos::ArrayRCP<LO> of size numLocalRows and fill it with the corresponding information
+      Teuchos::ArrayRCP<LO> partitionerMap(numLocalRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
+      for (size_t blockRow = 0; blockRow < Teuchos::as<size_t>(TVertLineIdSmoo.size()); ++blockRow)
+        for (size_t dof = 0; dof < numDofsPerNode; dof++)
+          partitionerMap[blockRow * numDofsPerNode + dof] = TVertLineIdSmoo[blockRow];
+      myparamList.set("partitioner: type", "user");
+      myparamList.set("partitioner: map", partitionerMap);
+      myparamList.set("partitioner: local parts", maxPart + 1);
     }
+
+    if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+        type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+        type_ == "LINESMOOTHING_BANDEDRELAXATION")
+      type_ = "BANDEDRELAXATION";
+    else if (type_ == "LINESMOOTHING_TRIDI_RELAXATION" ||
+             type_ == "LINESMOOTHING_TRIDI RELAXATION" ||
+             type_ == "LINESMOOTHING_TRIDIRELAXATION" ||
+             type_ == "LINESMOOTHING_TRIDIAGONAL_RELAXATION" ||
+             type_ == "LINESMOOTHING_TRIDIAGONAL RELAXATION" ||
+             type_ == "LINESMOOTHING_TRIDIAGONALRELAXATION")
+      type_ = "TRIDIAGONALRELAXATION";
+    else
+      type_ = "BLOCKRELAXATION";
+  } else {
+    // line detection failed -> fallback to point-wise relaxation
+    this->GetOStream(Runtime0) << "Line detection failed: fall back to point-wise relaxation" << std::endl;
+    myparamList.remove("partitioner: type", false);
+    myparamList.remove("partitioner: map", false);
+    myparamList.remove("partitioner: local parts", false);
+    type_ = "RELAXATION";
   }
 
+  RCP<const Tpetra::RowMatrix<SC, LO, GO, NO>> tA = Utilities::Op2NonConstTpetraRow(A_);
 
- template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHiptmair(Level&  currentLevel ) {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    if (!bA.is_null())
-      A_ = bA->Merge();
+  prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+  SetPrecParameters();
+  prec_->initialize();
+  prec_->compute();
+}
 
-    RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupBlockRelaxation(Level& currentLevel) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
 
-    bool reusePreconditioner = false;
-    if (this->IsSetup() == true) {
-      // Reuse the constructed preconditioner
-      this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupHiptmair(): Setup() has already been called, assuming reuse" << std::endl;
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  if (!bA.is_null())
+    A_ = bA->Merge();
 
-      RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix> > prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix> >(prec_);
-      if (!prec.is_null()) {
-        prec->setMatrix(tA);
-        reusePreconditioner = true;
-      } else {
-        this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupHiptmair(): reuse of this type is not available (failed cast to CanChangeMatrix), "
-            "reverting to full construction" << std::endl;
-      }
-    }
+  RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
 
-    // If we're doing Chebyshev subsmoothing, we'll need to get the eigenvalues
-    SC negone = -Teuchos::ScalarTraits<Scalar>::one();
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-    std::string smoother1  = paramList.get("hiptmair: smoother type 1","CHEBYSHEV");
-    std::string smoother2  = paramList.get("hiptmair: smoother type 2","CHEBYSHEV");
-    SC lambdaMax11 = negone;
+  bool reusePreconditioner = false;
+  if (this->IsSetup() == true) {
+    // Reuse the constructed preconditioner
+    this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupBlockRelaxation(): Setup() has already been called, assuming reuse" << std::endl;
 
-    if(smoother1 == "CHEBYSHEV") {
-      ParameterList & list1 = paramList.sublist("hiptmair: smoother list 1");
-      lambdaMax11 = SetupChebyshevEigenvalues(currentLevel,"A","EdgeMatrix ",list1);
+    RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix>> prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix>>(prec_);
+    if (!prec.is_null()) {
+      prec->setMatrix(tA);
+      reusePreconditioner = true;
+    } else {
+      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupBlockRelaxation(): reuse of this type is not available (failed cast to CanChangeMatrix), "
+                                     "reverting to full construction"
+                                  << std::endl;
     }
-    if(smoother2 == "CHEBYSHEV") {
-      ParameterList & list2 = paramList.sublist("hiptmair: smoother list 2");
-      SetupChebyshevEigenvalues(currentLevel,"A","EdgeMatrix ",list2);
+  }
+
+  if (!reusePreconditioner) {
+    ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
+    myparamList.print();
+    if (myparamList.isParameter("partitioner: type") &&
+        myparamList.get<std::string>("partitioner: type") == "line") {
+      Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>> xCoordinates =
+          Factory::Get<Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>>>(currentLevel, "Coordinates");
+      Teuchos::RCP<Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>> coordinates = Teuchos::rcpFromRef(Xpetra::toTpetra<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>(*xCoordinates));
+
+      RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
+      size_t lclSize   = A_->getRangeMap()->getLocalNumElements();
+      if (!matA.is_null())
+        lclSize = matA->getLocalNumRows();
+      size_t numDofsPerNode = lclSize / xCoordinates->getMap()->getLocalNumElements();
+      myparamList.set("partitioner: coordinates", coordinates);
+      myparamList.set("partitioner: PDE equations", (int)numDofsPerNode);
     }
 
-    // FIXME: Should really add some checks to make sure the eigenvalue calcs worked like in
-    // the regular SetupChebyshev
+    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+    SetPrecParameters();
+    prec_->initialize();
+  }
 
-    // Grab the auxillary matrices and stick them on the list
-    RCP<Operator> NodeMatrix = currentLevel.Get<RCP<Operator> >("NodeMatrix");
-    RCP<Operator> D0         = currentLevel.Get<RCP<Operator> >("D0");
+  prec_->compute();
+}
 
-    RCP<tRowMatrix> tNodeMatrix = Utilities::Op2NonConstTpetraRow(NodeMatrix);
-    RCP<tRowMatrix> tD0         = Utilities::Op2NonConstTpetraRow(D0);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupChebyshev(Level& currentLevel) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
+  using STS                = Teuchos::ScalarTraits<SC>;
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  if (!bA.is_null())
+    A_ = bA->Merge();
 
+  RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
 
-    Teuchos::ParameterList newlist;
-    newlist.set("P",tD0);
-    newlist.set("PtAP",tNodeMatrix);
-    if (!reusePreconditioner) {
-      prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-      SetPrecParameters(newlist);
-      prec_->initialize();
-    }
+  bool reusePreconditioner = false;
 
-    prec_->compute();
+  if (this->IsSetup() == true) {
+    // Reuse the constructed preconditioner
+    this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupChebyshev(): Setup() has already been called, assuming reuse" << std::endl;
 
-    // Post-processing the (1,1) eigenvalue, if we have one
-    if(smoother1 == "CHEBYSHEV" && lambdaMax11 == negone) {
-      using Teuchos::rcp_dynamic_cast;
-      typedef Tpetra::RowMatrix<SC, LO, GO, NO> MatrixType;
-      auto hiptmairPrec = rcp_dynamic_cast<Ifpack2::Hiptmair<MatrixType> >(prec_);
-      if(hiptmairPrec != Teuchos::null) {
-        auto chebyPrec = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType> >(hiptmairPrec->getPrec1());
-        if (chebyPrec != Teuchos::null) {
-          lambdaMax11 = chebyPrec->getLambdaMaxForApply();
-          RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
-          if (!matA.is_null())
-            matA->SetMaxEigenvalueEstimate(lambdaMax11);
-          this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack2)" << " = " << lambdaMax11 << std::endl;
-        }
-        TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax11 == negone, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): no maximum eigenvalue estimate");
-      }
+    RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix>> prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix>>(prec_);
+    if (!prec.is_null()) {
+      prec->setMatrix(tA);
+      reusePreconditioner = true;
+    } else {
+      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupChebyshev(): reuse of this type is not available (failed cast to CanChangeMatrix), "
+                                     "reverting to full construction"
+                                  << std::endl;
     }
+  }
 
+  // Take care of the eigenvalues
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+  SC negone                = -STS::one();
+  SC lambdaMax             = SetupChebyshevEigenvalues(currentLevel, "A", "", paramList);
 
-  }
+  if (!reusePreconditioner) {
+    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+    SetPrecParameters();
+    {
+      SubFactoryMonitor(*this, "Preconditioner init", currentLevel);
+      prec_->initialize();
+    }
+  } else
+    SetPrecParameters();
 
+  {
+    SubFactoryMonitor(*this, "Preconditioner compute", currentLevel);
+    prec_->compute();
+  }
 
+  if (lambdaMax == negone) {
+    typedef Tpetra::RowMatrix<SC, LO, GO, NO> MatrixType;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Scalar Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupChebyshevEigenvalues(Level & currentLevel, const std::string & matrixName, const std::string & label, ParameterList & paramList) const {
-    // Helper: This gets used for smoothers that want to set up Chebyhev
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC negone = -STS::one();
-    RCP<const Matrix> currentA = currentLevel.Get<RCP<Matrix> >(matrixName);
-    SC lambdaMax = negone;
-
-    std::string maxEigString   = "chebyshev: max eigenvalue";
-    std::string eigRatioString = "chebyshev: ratio eigenvalue";
-    
-    // Get/calculate the maximum eigenvalue
-    if (paramList.isParameter(maxEigString)) {
-      if (paramList.isType<double>(maxEigString))
-        lambdaMax = paramList.get<double>(maxEigString);
-      else
-        lambdaMax = paramList.get<SC>(maxEigString);
-      this->GetOStream(Statistics1) << label << maxEigString << " (cached with smoother parameter list) = " << lambdaMax << std::endl;
+    Teuchos::RCP<Ifpack2::Chebyshev<MatrixType>> chebyPrec = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType>>(prec_);
+    if (chebyPrec != Teuchos::null) {
+      lambdaMax        = chebyPrec->getLambdaMaxForApply();
       RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
       if (!matA.is_null())
         matA->SetMaxEigenvalueEstimate(lambdaMax);
-      
-    } else {
-      lambdaMax = currentA->GetMaxEigenvalueEstimate();
-      if (lambdaMax != negone) {
-        this->GetOStream(Statistics1) << label << maxEigString << " (cached with matrix) = " << lambdaMax << std::endl;
-        paramList.set(maxEigString, lambdaMax);
-      }
-    }
-    
-    // Calculate the eigenvalue ratio
-    const SC defaultEigRatio = 20;
-    
-    SC ratio = defaultEigRatio;
-    if (paramList.isParameter(eigRatioString)) {
-      if (paramList.isType<double>(eigRatioString))
-        ratio = paramList.get<double>(eigRatioString);
-      else
-          ratio = paramList.get<SC>(eigRatioString);
+      this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack2)"
+                                    << " = " << lambdaMax << std::endl;
     }
-    if (currentLevel.GetLevelID()) {
-      // Update ratio to be
-      //   ratio = max(number of fine DOFs / number of coarse DOFs, defaultValue)
-      //
-      // NOTE: We don't need to request previous level matrix as we know for sure it was constructed
-      RCP<const Matrix> fineA = currentLevel.GetPreviousLevel()->Get<RCP<Matrix> >(matrixName);
-      size_t nRowsFine   = fineA->getGlobalNumRows();
-      size_t nRowsCoarse = currentA->getGlobalNumRows();
-      
-      SC levelRatio = as<SC>(as<float>(nRowsFine)/nRowsCoarse);
-      if (STS::magnitude(levelRatio) > STS::magnitude(ratio))
-        ratio = levelRatio;
-    }
-    
-    this->GetOStream(Statistics1) << label << eigRatioString << " (computed) = " << ratio << std::endl;
-    paramList.set(eigRatioString, ratio);
-
-    if (paramList.isParameter("chebyshev: use rowsumabs diagonal scaling")) {
-      this->GetOStream(Runtime1) << "chebyshev: using rowsumabs diagonal scaling" << std::endl;
-      bool doScale = false;
-      doScale = paramList.get<bool>("chebyshev: use rowsumabs diagonal scaling");
-      paramList.remove("chebyshev: use rowsumabs diagonal scaling");
-      double chebyReplaceTol = Teuchos::ScalarTraits<Scalar>::eps()*100;
-      std::string paramName = "chebyshev: rowsumabs diagonal replacement tolerance";
-      if (paramList.isParameter(paramName)) {
-        chebyReplaceTol = paramList.get<double>(paramName);
-        paramList.remove(paramName);
-      }
-      double chebyReplaceVal = Teuchos::ScalarTraits<double>::zero();
-      paramName = "chebyshev: rowsumabs diagonal replacement value";
-      if (paramList.isParameter(paramName)) {
-        chebyReplaceVal = paramList.get<double>(paramName);
-        paramList.remove(paramName);
-      }
-      bool chebyReplaceSingleEntryRowWithZero = false;
-      paramName = "chebyshev: rowsumabs replace single entry row with zero";
-      if (paramList.isParameter(paramName)) {
-        chebyReplaceSingleEntryRowWithZero = paramList.get<bool>(paramName);
-        paramList.remove(paramName);
-      }
-      bool useAverageAbsDiagVal = false;
-      paramName = "chebyshev: rowsumabs use automatic diagonal tolerance";
-      if (paramList.isParameter(paramName)) {
-        useAverageAbsDiagVal = paramList.get<bool>(paramName);
-        paramList.remove(paramName);
-      }
-      if (doScale) {
-        const bool doReciprocal = true;
-        RCP<Vector> lumpedDiagonal = Utilities::GetLumpedMatrixDiagonal(*currentA, doReciprocal, chebyReplaceTol, chebyReplaceVal, chebyReplaceSingleEntryRowWithZero, useAverageAbsDiagVal);
-        const Xpetra::TpetraVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*lumpedDiagonal);
-        paramList.set("chebyshev: operator inv diagonal",tmpVec.getTpetra_Vector());
-      }
+    TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax == negone, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): no maximum eigenvalue estimate");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupHiptmair(Level& currentLevel) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  if (!bA.is_null())
+    A_ = bA->Merge();
+
+  RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
+
+  bool reusePreconditioner = false;
+  if (this->IsSetup() == true) {
+    // Reuse the constructed preconditioner
+    this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupHiptmair(): Setup() has already been called, assuming reuse" << std::endl;
+
+    RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix>> prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix>>(prec_);
+    if (!prec.is_null()) {
+      prec->setMatrix(tA);
+      reusePreconditioner = true;
+    } else {
+      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupHiptmair(): reuse of this type is not available (failed cast to CanChangeMatrix), "
+                                     "reverting to full construction"
+                                  << std::endl;
     }
-
-    return lambdaMax;
   }
 
+  // If we're doing Chebyshev subsmoothing, we'll need to get the eigenvalues
+  SC negone                = -Teuchos::ScalarTraits<Scalar>::one();
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+  std::string smoother1    = paramList.get("hiptmair: smoother type 1", "CHEBYSHEV");
+  std::string smoother2    = paramList.get("hiptmair: smoother type 2", "CHEBYSHEV");
+  SC lambdaMax11           = negone;
 
+  if (smoother1 == "CHEBYSHEV") {
+    ParameterList& list1 = paramList.sublist("hiptmair: smoother list 1");
+    lambdaMax11          = SetupChebyshevEigenvalues(currentLevel, "A", "EdgeMatrix ", list1);
+  }
+  if (smoother2 == "CHEBYSHEV") {
+    ParameterList& list2 = paramList.sublist("hiptmair: smoother list 2");
+    SetupChebyshevEigenvalues(currentLevel, "A", "EdgeMatrix ", list2);
+  }
+
+  // FIXME: Should really add some checks to make sure the eigenvalue calcs worked like in
+  // the regular SetupChebyshev
 
+  // Grab the auxillary matrices and stick them on the list
+  RCP<Operator> NodeMatrix = currentLevel.Get<RCP<Operator>>("NodeMatrix");
+  RCP<Operator> D0         = currentLevel.Get<RCP<Operator>>("D0");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupGeneric(Level& /* currentLevel */) {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> tRowMatrix;
-    RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
-    if (!bA.is_null())
-      A_ = bA->Merge();
+  RCP<tRowMatrix> tNodeMatrix = Utilities::Op2NonConstTpetraRow(NodeMatrix);
+  RCP<tRowMatrix> tD0         = Utilities::Op2NonConstTpetraRow(D0);
 
-    RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
+  Teuchos::ParameterList newlist;
+  newlist.set("P", tD0);
+  newlist.set("PtAP", tNodeMatrix);
+  if (!reusePreconditioner) {
+    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+    SetPrecParameters(newlist);
+    prec_->initialize();
+  }
 
-    bool reusePreconditioner = false;
-    if (this->IsSetup() == true) {
-      // Reuse the constructed preconditioner
-      this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupGeneric(): Setup() has already been called, assuming reuse" << std::endl;
+  prec_->compute();
 
-      RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix> > prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix> >(prec_);
-      if (!prec.is_null()) {
-        prec->setMatrix(tA);
-        reusePreconditioner = true;
-      } else {
-        this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupGeneric(): reuse of this type is not available (failed cast to CanChangeMatrix), "
-            "reverting to full construction" << std::endl;
+  // Post-processing the (1,1) eigenvalue, if we have one
+  if (smoother1 == "CHEBYSHEV" && lambdaMax11 == negone) {
+    using Teuchos::rcp_dynamic_cast;
+    typedef Tpetra::RowMatrix<SC, LO, GO, NO> MatrixType;
+    auto hiptmairPrec = rcp_dynamic_cast<Ifpack2::Hiptmair<MatrixType>>(prec_);
+    if (hiptmairPrec != Teuchos::null) {
+      auto chebyPrec = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType>>(hiptmairPrec->getPrec1());
+      if (chebyPrec != Teuchos::null) {
+        lambdaMax11      = chebyPrec->getLambdaMaxForApply();
+        RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
+        if (!matA.is_null())
+          matA->SetMaxEigenvalueEstimate(lambdaMax11);
+        this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack2)"
+                                      << " = " << lambdaMax11 << std::endl;
       }
+      TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax11 == negone, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): no maximum eigenvalue estimate");
     }
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Scalar Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupChebyshevEigenvalues(Level& currentLevel, const std::string& matrixName, const std::string& label, ParameterList& paramList) const {
+  // Helper: This gets used for smoothers that want to set up Chebyhev
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC negone                  = -STS::one();
+  RCP<const Matrix> currentA = currentLevel.Get<RCP<Matrix>>(matrixName);
+  SC lambdaMax               = negone;
+
+  std::string maxEigString   = "chebyshev: max eigenvalue";
+  std::string eigRatioString = "chebyshev: ratio eigenvalue";
+
+  // Get/calculate the maximum eigenvalue
+  if (paramList.isParameter(maxEigString)) {
+    if (paramList.isType<double>(maxEigString))
+      lambdaMax = paramList.get<double>(maxEigString);
+    else
+      lambdaMax = paramList.get<SC>(maxEigString);
+    this->GetOStream(Statistics1) << label << maxEigString << " (cached with smoother parameter list) = " << lambdaMax << std::endl;
+    RCP<Matrix> matA = rcp_dynamic_cast<Matrix>(A_);
+    if (!matA.is_null())
+      matA->SetMaxEigenvalueEstimate(lambdaMax);
 
-    if (!reusePreconditioner) {
-      prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
-      SetPrecParameters();
-      prec_->initialize();
+  } else {
+    lambdaMax = currentA->GetMaxEigenvalueEstimate();
+    if (lambdaMax != negone) {
+      this->GetOStream(Statistics1) << label << maxEigString << " (cached with matrix) = " << lambdaMax << std::endl;
+      paramList.set(maxEigString, lambdaMax);
     }
+  }
 
-    prec_->compute();
+  // Calculate the eigenvalue ratio
+  const SC defaultEigRatio = 20;
+
+  SC ratio = defaultEigRatio;
+  if (paramList.isParameter(eigRatioString)) {
+    if (paramList.isType<double>(eigRatioString))
+      ratio = paramList.get<double>(eigRatioString);
+    else
+      ratio = paramList.get<SC>(eigRatioString);
+  }
+  if (currentLevel.GetLevelID()) {
+    // Update ratio to be
+    //   ratio = max(number of fine DOFs / number of coarse DOFs, defaultValue)
+    //
+    // NOTE: We don't need to request previous level matrix as we know for sure it was constructed
+    RCP<const Matrix> fineA = currentLevel.GetPreviousLevel()->Get<RCP<Matrix>>(matrixName);
+    size_t nRowsFine        = fineA->getGlobalNumRows();
+    size_t nRowsCoarse      = currentA->getGlobalNumRows();
+
+    SC levelRatio = as<SC>(as<float>(nRowsFine) / nRowsCoarse);
+    if (STS::magnitude(levelRatio) > STS::magnitude(ratio))
+      ratio = levelRatio;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Apply(): Setup() has not been called");
-
-    // Forward the InitialGuessIsZero option to Ifpack2
-    // TODO:  It might be nice to switch back the internal
-    //        "zero starting solution" option of the ifpack2 object prec_ to his
-    //        initial value at the end but there is no way right now to get
-    //        the current value of the "zero starting solution" in ifpack2.
-    //        It's not really an issue, as prec_  can only be used by this method.
-    Teuchos::ParameterList paramList;
-    bool supportInitialGuess = false;
-    const Teuchos::ParameterList params = this->GetParameterList();
-
-    if (prec_->supportsZeroStartingSolution()) {
-      prec_->setZeroStartingSolution(InitialGuessIsZero);
-      supportInitialGuess = true;
-    } else if (type_ == "SCHWARZ") {
-      paramList.set("schwarz: zero starting solution", InitialGuessIsZero);
-      //Because additive Schwarz has "delta" semantics, it's sufficient to
-      //toggle only the zero initial guess flag, and not pass in already
-      //set parameters.  If we call SetPrecParameters, the subdomain solver
-      //will be destroyed.
-      prec_->setParameters(paramList);
-      supportInitialGuess = true;
+  this->GetOStream(Statistics1) << label << eigRatioString << " (computed) = " << ratio << std::endl;
+  paramList.set(eigRatioString, ratio);
+
+  if (paramList.isParameter("chebyshev: use rowsumabs diagonal scaling")) {
+    this->GetOStream(Runtime1) << "chebyshev: using rowsumabs diagonal scaling" << std::endl;
+    bool doScale = false;
+    doScale      = paramList.get<bool>("chebyshev: use rowsumabs diagonal scaling");
+    paramList.remove("chebyshev: use rowsumabs diagonal scaling");
+    double chebyReplaceTol = Teuchos::ScalarTraits<Scalar>::eps() * 100;
+    std::string paramName  = "chebyshev: rowsumabs diagonal replacement tolerance";
+    if (paramList.isParameter(paramName)) {
+      chebyReplaceTol = paramList.get<double>(paramName);
+      paramList.remove(paramName);
+    }
+    double chebyReplaceVal = Teuchos::ScalarTraits<double>::zero();
+    paramName              = "chebyshev: rowsumabs diagonal replacement value";
+    if (paramList.isParameter(paramName)) {
+      chebyReplaceVal = paramList.get<double>(paramName);
+      paramList.remove(paramName);
+    }
+    bool chebyReplaceSingleEntryRowWithZero = false;
+    paramName                               = "chebyshev: rowsumabs replace single entry row with zero";
+    if (paramList.isParameter(paramName)) {
+      chebyReplaceSingleEntryRowWithZero = paramList.get<bool>(paramName);
+      paramList.remove(paramName);
+    }
+    bool useAverageAbsDiagVal = false;
+    paramName                 = "chebyshev: rowsumabs use automatic diagonal tolerance";
+    if (paramList.isParameter(paramName)) {
+      useAverageAbsDiagVal = paramList.get<bool>(paramName);
+      paramList.remove(paramName);
     }
+    if (doScale) {
+      const bool doReciprocal                                                       = true;
+      RCP<Vector> lumpedDiagonal                                                    = Utilities::GetLumpedMatrixDiagonal(*currentA, doReciprocal, chebyReplaceTol, chebyReplaceVal, chebyReplaceSingleEntryRowWithZero, useAverageAbsDiagVal);
+      const Xpetra::TpetraVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*lumpedDiagonal);
+      paramList.set("chebyshev: operator inv diagonal", tmpVec.getTpetra_Vector());
+    }
+  }
 
-    //TODO JJH 30Apr2014  Calling SetPrecParameters(paramList) when the smoother
-    //is Ifpack2::AdditiveSchwarz::setParameterList() will destroy the subdomain
-    //(aka inner) solver.  This behavior is documented but a departure from what
-    //it previously did, and what other Ifpack2 solvers currently do.  So I have
-    //moved SetPrecParameters(paramList) into the if-else block above.
-
-    // Apply
-    if (InitialGuessIsZero || supportInitialGuess) {
-      Tpetra::MultiVector<SC,LO,GO,NO>&       tpX = Utilities::MV2NonConstTpetraMV(X);
-      const Tpetra::MultiVector<SC,LO,GO,NO>& tpB = Utilities::MV2TpetraMV(B);
-      prec_->apply(tpB, tpX);
-    } else {
-      typedef Teuchos::ScalarTraits<Scalar> TST;
+  return lambdaMax;
+}
 
-      RCP<MultiVector> Residual;
-      {
-        std::string prefix = this->ShortClassName() + ": Apply: ";
-        RCP<TimeMonitor> tM = rcp(new TimeMonitor(*this, prefix + "residual calculation", Timings0));
-        Residual = Utilities::Residual(*A_, X, B);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupGeneric(Level& /* currentLevel */) {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> tRowMatrix;
+  RCP<BlockedCrsMatrix> bA = rcp_dynamic_cast<BlockedCrsMatrix>(A_);
+  if (!bA.is_null())
+    A_ = bA->Merge();
 
-      RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
+  RCP<const tRowMatrix> tA = Utilities::Op2NonConstTpetraRow(A_);
 
-      Tpetra::MultiVector<SC,LO,GO,NO>&       tpX = Utilities::MV2NonConstTpetraMV(*Correction);
-      const Tpetra::MultiVector<SC,LO,GO,NO>& tpB = Utilities::MV2TpetraMV(*Residual);
+  bool reusePreconditioner = false;
+  if (this->IsSetup() == true) {
+    // Reuse the constructed preconditioner
+    this->GetOStream(Runtime1) << "MueLu::Ifpack2Smoother::SetupGeneric(): Setup() has already been called, assuming reuse" << std::endl;
 
-      prec_->apply(tpB, tpX);
-
-      X.update(TST::one(), *Correction, TST::one());
+    RCP<Ifpack2::Details::CanChangeMatrix<tRowMatrix>> prec = rcp_dynamic_cast<Ifpack2::Details::CanChangeMatrix<tRowMatrix>>(prec_);
+    if (!prec.is_null()) {
+      prec->setMatrix(tA);
+      reusePreconditioner = true;
+    } else {
+      this->GetOStream(Warnings0) << "MueLu::Ifpack2Smoother::SetupGeneric(): reuse of this type is not available (failed cast to CanChangeMatrix), "
+                                     "reverting to full construction"
+                                  << std::endl;
     }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<Ifpack2Smoother> smoother = rcp(new Ifpack2Smoother(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return smoother;
+  if (!reusePreconditioner) {
+    prec_ = Ifpack2::Factory::create(type_, tA, overlap_);
+    SetPrecParameters();
+    prec_->initialize();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (SmootherPrototype::IsSetup()) {
-      out << prec_->description();
-    } else {
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-    }
-    return out.str();
+  prec_->compute();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Apply(): Setup() has not been called");
+
+  // Forward the InitialGuessIsZero option to Ifpack2
+  // TODO:  It might be nice to switch back the internal
+  //        "zero starting solution" option of the ifpack2 object prec_ to his
+  //        initial value at the end but there is no way right now to get
+  //        the current value of the "zero starting solution" in ifpack2.
+  //        It's not really an issue, as prec_  can only be used by this method.
+  Teuchos::ParameterList paramList;
+  bool supportInitialGuess            = false;
+  const Teuchos::ParameterList params = this->GetParameterList();
+
+  if (prec_->supportsZeroStartingSolution()) {
+    prec_->setZeroStartingSolution(InitialGuessIsZero);
+    supportInitialGuess = true;
+  } else if (type_ == "SCHWARZ") {
+    paramList.set("schwarz: zero starting solution", InitialGuessIsZero);
+    // Because additive Schwarz has "delta" semantics, it's sufficient to
+    // toggle only the zero initial guess flag, and not pass in already
+    // set parameters.  If we call SetPrecParameters, the subdomain solver
+    // will be destroyed.
+    prec_->setParameters(paramList);
+    supportInitialGuess = true;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+  // TODO JJH 30Apr2014  Calling SetPrecParameters(paramList) when the smoother
+  // is Ifpack2::AdditiveSchwarz::setParameterList() will destroy the subdomain
+  //(aka inner) solver.  This behavior is documented but a departure from what
+  // it previously did, and what other Ifpack2 solvers currently do.  So I have
+  // moved SetPrecParameters(paramList) into the if-else block above.
+
+  // Apply
+  if (InitialGuessIsZero || supportInitialGuess) {
+    Tpetra::MultiVector<SC, LO, GO, NO>& tpX       = Utilities::MV2NonConstTpetraMV(X);
+    const Tpetra::MultiVector<SC, LO, GO, NO>& tpB = Utilities::MV2TpetraMV(B);
+    prec_->apply(tpB, tpX);
+  } else {
+    typedef Teuchos::ScalarTraits<Scalar> TST;
+
+    RCP<MultiVector> Residual;
+    {
+      std::string prefix  = this->ShortClassName() + ": Apply: ";
+      RCP<TimeMonitor> tM = rcp(new TimeMonitor(*this, prefix + "residual calculation", Timings0));
+      Residual            = Utilities::Residual(*A_, X, B);
+    }
 
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
+    RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-      out0 << "Overlap: "        << overlap_ << std::endl;
-    }
+    Tpetra::MultiVector<SC, LO, GO, NO>& tpX       = Utilities::MV2NonConstTpetraMV(*Correction);
+    const Tpetra::MultiVector<SC, LO, GO, NO>& tpB = Utilities::MV2TpetraMV(*Residual);
 
-    if (verbLevel & External)
-      if (prec_ != Teuchos::null) {
-        Teuchos::OSTab tab2(out);
-        out << *prec_ << std::endl;
-      }
+    prec_->apply(tpB, tpX);
 
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "RCP<prec_>: " << prec_ << std::endl;
-    }
+    X.update(TST::one(), *Correction, TST::one());
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<Ifpack2Smoother> smoother = rcp(new Ifpack2Smoother(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return smoother;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (SmootherPrototype::IsSetup()) {
+    out << prec_->description();
+  } else {
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+  }
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    typedef Tpetra::RowMatrix<SC,LO,GO,NO> MatrixType;
-    // NOTE: Only works for a subset of Ifpack2's smoothers
-    RCP<Ifpack2::Relaxation<MatrixType> > pr     = rcp_dynamic_cast<Ifpack2::Relaxation<MatrixType> >(prec_);
-    if(!pr.is_null()) return pr->getNodeSmootherComplexity();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    RCP<Ifpack2::Chebyshev<MatrixType> > pc       = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType> >(prec_);
-    if(!pc.is_null()) return pc->getNodeSmootherComplexity();
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
 
-    RCP<Ifpack2::BlockRelaxation<MatrixType> > pb = rcp_dynamic_cast<Ifpack2::BlockRelaxation<MatrixType> >(prec_);
-    if(!pb.is_null()) return pb->getNodeSmootherComplexity();
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
+    out0 << "Overlap: " << overlap_ << std::endl;
+  }
 
-    RCP<Ifpack2::ILUT<MatrixType> > pi            = rcp_dynamic_cast<Ifpack2::ILUT<MatrixType> >(prec_);
-    if(!pi.is_null()) return pi->getNodeSmootherComplexity();
+  if (verbLevel & External)
+    if (prec_ != Teuchos::null) {
+      Teuchos::OSTab tab2(out);
+      out << *prec_ << std::endl;
+    }
 
-    RCP<Ifpack2::RILUK<MatrixType> > pk            = rcp_dynamic_cast<Ifpack2::RILUK<MatrixType> >(prec_);
-    if(!pk.is_null()) return pk->getNodeSmootherComplexity();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "RCP<prec_>: " << prec_ << std::endl;
+  }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t Ifpack2Smoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  typedef Tpetra::RowMatrix<SC, LO, GO, NO> MatrixType;
+  // NOTE: Only works for a subset of Ifpack2's smoothers
+  RCP<Ifpack2::Relaxation<MatrixType>> pr = rcp_dynamic_cast<Ifpack2::Relaxation<MatrixType>>(prec_);
+  if (!pr.is_null()) return pr->getNodeSmootherComplexity();
 
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
+  RCP<Ifpack2::Chebyshev<MatrixType>> pc = rcp_dynamic_cast<Ifpack2::Chebyshev<MatrixType>>(prec_);
+  if (!pc.is_null()) return pc->getNodeSmootherComplexity();
+
+  RCP<Ifpack2::BlockRelaxation<MatrixType>> pb = rcp_dynamic_cast<Ifpack2::BlockRelaxation<MatrixType>>(prec_);
+  if (!pb.is_null()) return pb->getNodeSmootherComplexity();
+
+  RCP<Ifpack2::ILUT<MatrixType>> pi = rcp_dynamic_cast<Ifpack2::ILUT<MatrixType>>(prec_);
+  if (!pi.is_null()) return pi->getNodeSmootherComplexity();
+
+  RCP<Ifpack2::RILUK<MatrixType>> pk = rcp_dynamic_cast<Ifpack2::RILUK<MatrixType>>(prec_);
+  if (!pk.is_null()) return pk->getNodeSmootherComplexity();
 
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // HAVE_MUELU_IFPACK2
-#endif // MUELU_IFPACK2SMOOTHER_DEF_HPP
+#endif  // HAVE_MUELU_IFPACK2
+#endif  // MUELU_IFPACK2SMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.cpp b/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.cpp
index 1801ab0f6784..6755fdf9286e 100644
--- a/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.cpp
+++ b/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.cpp
@@ -57,380 +57,370 @@
 #include "MueLu_Monitor.hpp"
 #include "MueLu_Aggregates.hpp"
 
-
 namespace MueLu {
 
-  template <class Node>
-  IfpackSmoother<Node>::IfpackSmoother(std::string const & type, Teuchos::ParameterList const & paramList, LO const &overlap)
-    : type_(type), overlap_(overlap)
-  {
-    this->declareConstructionOutcome(false, "");
-    SetParameterList(paramList);
+template <class Node>
+IfpackSmoother<Node>::IfpackSmoother(std::string const& type, Teuchos::ParameterList const& paramList, LO const& overlap)
+  : type_(type)
+  , overlap_(overlap) {
+  this->declareConstructionOutcome(false, "");
+  SetParameterList(paramList);
+}
+
+template <class Node>
+void IfpackSmoother<Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+
+  if (SmootherPrototype::IsSetup()) {
+    // It might be invalid to change parameters after the setup, but it depends entirely on Ifpack implementation.
+    // TODO: I don't know if Ifpack returns an error code or exception or ignore parameters modification in this case...
+    prec_->SetParameters(const_cast<ParameterList&>(this->GetParameterList()));
   }
-
-  template <class Node>
-  void IfpackSmoother<Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
-    Factory::SetParameterList(paramList);
-
-    if (SmootherPrototype::IsSetup()) {
-      // It might be invalid to change parameters after the setup, but it depends entirely on Ifpack implementation.
-      // TODO: I don't know if Ifpack returns an error code or exception or ignore parameters modification in this case...
-      prec_->SetParameters(const_cast<ParameterList&>(this->GetParameterList()));
-    }
+}
+
+template <class Node>
+void IfpackSmoother<Node>::SetPrecParameters(const Teuchos::ParameterList& list) const {
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+  paramList.setParameters(list);
+
+  RCP<ParameterList> precList = this->RemoveFactoriesFromList(this->GetParameterList());
+
+  prec_->SetParameters(*precList);
+
+  // We would like to have the following line here:
+  //      paramList.setParameters(*precList);
+  // For instance, if Ifpack sets somem parameters internally, we would like to have
+  // them listed when we call this->GetParameterList()
+  // But because of the way Ifpack handles the list, we cannot do that.
+  // The bad scenario goes like this:
+  //   * SmootherFactory calls Setup
+  //   * Setup calls SetPrecParameters
+  //   * We call prec_->SetParameters(*precList)
+  //     This actually updates the internal parameter list  with default prec_ parameters
+  //     This means that we get a parameter ("chebyshev: max eigenvalue", -1) in the list
+  //   * Setup calls prec_->Compute()
+  //     Here we may compute the max eigenvalue, but we get no indication of this. If we
+  //     do compute it, our parameter list becomes outdated
+  //   * SmootherFactory calls Apply
+  //   * Apply constructs a list with a list with an entry "chebyshev: zero starting solution"
+  //   * We call prec_->SetParameters(*precList)
+  // The last call is the problem. At this point, we have a list with an outdated entry
+  // "chebyshev: max eigenvalue", but prec_ uses this entry and replaces the computed max
+  // eigenvalue with the one from the list, resulting in -1.0 eigenvalue.
+  //
+  // Ifpack2 does not have this problem, as it does not populate the list with new entries
+}
+
+template <class Node>
+void IfpackSmoother<Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
+
+  if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDEDRELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK_RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCKRELAXATION") {
+    this->Input(currentLevel, "CoarseNumZLayers");           // necessary for fallback criterion
+    this->Input(currentLevel, "LineDetection_VertLineIds");  // necessary to feed block smoother
+  }                                                          // if (type_ == "LINESMOOTHING_BANDEDRELAXATION")
+  else if (type_ == "AGGREGATE") {
+    // Aggregate smoothing needs aggregates
+    this->Input(currentLevel, "Aggregates");
   }
+}
 
-  template <class Node>
-  void IfpackSmoother<Node>::SetPrecParameters(const Teuchos::ParameterList& list) const {
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-    paramList.setParameters(list);
-
-    RCP<ParameterList> precList = this->RemoveFactoriesFromList(this->GetParameterList());
-
-    prec_->SetParameters(*precList);
-
-    // We would like to have the following line here:
-    //      paramList.setParameters(*precList);
-    // For instance, if Ifpack sets somem parameters internally, we would like to have
-    // them listed when we call this->GetParameterList()
-    // But because of the way Ifpack handles the list, we cannot do that.
-    // The bad scenario goes like this:
-    //   * SmootherFactory calls Setup
-    //   * Setup calls SetPrecParameters
-    //   * We call prec_->SetParameters(*precList)
-    //     This actually updates the internal parameter list  with default prec_ parameters
-    //     This means that we get a parameter ("chebyshev: max eigenvalue", -1) in the list
-    //   * Setup calls prec_->Compute()
-    //     Here we may compute the max eigenvalue, but we get no indication of this. If we
-    //     do compute it, our parameter list becomes outdated
-    //   * SmootherFactory calls Apply
-    //   * Apply constructs a list with a list with an entry "chebyshev: zero starting solution"
-    //   * We call prec_->SetParameters(*precList)
-    // The last call is the problem. At this point, we have a list with an outdated entry
-    // "chebyshev: max eigenvalue", but prec_ uses this entry and replaces the computed max
-    // eigenvalue with the one from the list, resulting in -1.0 eigenvalue.
-    //
-    // Ifpack2 does not have this problem, as it does not populate the list with new entries
-  }
+template <class Node>
+void IfpackSmoother<Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::IfpackSmoother::Setup(): Setup() has already been called" << std::endl;
 
-  template <class Node>
-  void IfpackSmoother<Node>::DeclareInput(Level &currentLevel) const {
-    this->Input(currentLevel, "A");
-
-    if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
-        type_ == "LINESMOOTHING_BANDED RELAXATION" ||
-        type_ == "LINESMOOTHING_BANDEDRELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCK_RELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCK RELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCKRELAXATION") {
-      this->Input(currentLevel, "CoarseNumZLayers");              // necessary for fallback criterion
-      this->Input(currentLevel, "LineDetection_VertLineIds"); // necessary to feed block smoother
-    } // if (type_ == "LINESMOOTHING_BANDEDRELAXATION")
-    else if (type_ == "AGGREGATE")
-    {
-      // Aggregate smoothing needs aggregates
-      this->Input(currentLevel,"Aggregates");
-    }
+  A_ = Factory::Get<RCP<Matrix> >(currentLevel, "A");
 
-  }
+  double lambdaMax = -1.0;
+  if (type_ == "Chebyshev") {
+    std::string maxEigString   = "chebyshev: max eigenvalue";
+    std::string eigRatioString = "chebyshev: ratio eigenvalue";
 
-  template <class Node>
-  void IfpackSmoother<Node>::Setup(Level &currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::IfpackSmoother::Setup(): Setup() has already been called" << std::endl;
+    try {
+      lambdaMax = Teuchos::getValue<Scalar>(this->GetParameter(maxEigString));
+      this->GetOStream(Statistics1) << maxEigString << " (cached with smoother parameter list) = " << lambdaMax << std::endl;
 
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A");
+    } catch (Teuchos::Exceptions::InvalidParameterName&) {
+      lambdaMax = A_->GetMaxEigenvalueEstimate();
 
-    double lambdaMax = -1.0;
-    if (type_ == "Chebyshev") {
-      std::string maxEigString   = "chebyshev: max eigenvalue";
-      std::string eigRatioString = "chebyshev: ratio eigenvalue";
+      if (lambdaMax != -1.0) {
+        this->GetOStream(Statistics1) << maxEigString << " (cached with matrix) = " << lambdaMax << std::endl;
+        this->SetParameter(maxEigString, ParameterEntry(lambdaMax));
+      }
+    }
 
-      try {
-        lambdaMax = Teuchos::getValue<Scalar>(this->GetParameter(maxEigString));
-        this->GetOStream(Statistics1) << maxEigString << " (cached with smoother parameter list) = " << lambdaMax << std::endl;
+    // Calculate the eigenvalue ratio
+    const Scalar defaultEigRatio = 20;
 
-      } catch (Teuchos::Exceptions::InvalidParameterName&) {
-        lambdaMax = A_->GetMaxEigenvalueEstimate();
+    Scalar ratio = defaultEigRatio;
+    try {
+      ratio = Teuchos::getValue<Scalar>(this->GetParameter(eigRatioString));
 
-        if (lambdaMax != -1.0) {
-          this->GetOStream(Statistics1) << maxEigString << " (cached with matrix) = " << lambdaMax << std::endl;
-          this->SetParameter(maxEigString, ParameterEntry(lambdaMax));
-        }
-      }
+    } catch (Teuchos::Exceptions::InvalidParameterName&) {
+      this->SetParameter(eigRatioString, ParameterEntry(ratio));
+    }
 
-      // Calculate the eigenvalue ratio
-      const Scalar defaultEigRatio = 20;
+    if (currentLevel.GetLevelID()) {
+      // Update ratio to be
+      //   ratio = max(number of fine DOFs / number of coarse DOFs, defaultValue)
+      //
+      // NOTE: We don't need to request previous level matrix as we know for sure it was constructed
+      RCP<const Matrix> fineA = currentLevel.GetPreviousLevel()->Get<RCP<Matrix> >("A");
+      size_t nRowsFine        = fineA->getGlobalNumRows();
+      size_t nRowsCoarse      = A_->getGlobalNumRows();
 
-      Scalar ratio = defaultEigRatio;
-      try {
-        ratio = Teuchos::getValue<Scalar>(this->GetParameter(eigRatioString));
+      ratio = std::max(ratio, as<Scalar>(nRowsFine) / nRowsCoarse);
 
-      } catch (Teuchos::Exceptions::InvalidParameterName&) {
-        this->SetParameter(eigRatioString, ParameterEntry(ratio));
+      this->GetOStream(Statistics1) << eigRatioString << " (computed) = " << ratio << std::endl;
+      this->SetParameter(eigRatioString, ParameterEntry(ratio));
+    }
+  }  // if (type_ == "Chebyshev")
+
+  if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+      type_ == "LINESMOOTHING_BANDEDRELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK_RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCK RELAXATION" ||
+      type_ == "LINESMOOTHING_BLOCKRELAXATION") {
+    ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
+
+    LO CoarseNumZLayers = currentLevel.Get<LO>("CoarseNumZLayers", Factory::GetFactory("CoarseNumZLayers").get());
+    if (CoarseNumZLayers > 0) {
+      Teuchos::ArrayRCP<LO> TVertLineIdSmoo = currentLevel.Get<Teuchos::ArrayRCP<LO> >("LineDetection_VertLineIds", Factory::GetFactory("LineDetection_VertLineIds").get());
+
+      // determine number of local parts
+      LO maxPart = 0;
+      for (size_t k = 0; k < Teuchos::as<size_t>(TVertLineIdSmoo.size()); k++) {
+        if (maxPart < TVertLineIdSmoo[k]) maxPart = TVertLineIdSmoo[k];
       }
 
-      if (currentLevel.GetLevelID()) {
-        // Update ratio to be
-        //   ratio = max(number of fine DOFs / number of coarse DOFs, defaultValue)
-        //
-        // NOTE: We don't need to request previous level matrix as we know for sure it was constructed
-        RCP<const Matrix> fineA = currentLevel.GetPreviousLevel()->Get<RCP<Matrix> >("A");
-        size_t nRowsFine   = fineA->getGlobalNumRows();
-        size_t nRowsCoarse = A_->getGlobalNumRows();
+      size_t numLocalRows = A_->getLocalNumRows();
+      TEUCHOS_TEST_FOR_EXCEPTION(numLocalRows % TVertLineIdSmoo.size() != 0, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): the number of local nodes is incompatible with the TVertLineIdsSmoo.");
 
-        ratio = std::max(ratio, as<Scalar>(nRowsFine)/nRowsCoarse);
-
-        this->GetOStream(Statistics1) << eigRatioString << " (computed) = " << ratio << std::endl;
-        this->SetParameter(eigRatioString, ParameterEntry(ratio));
-      }
-    } // if (type_ == "Chebyshev")
-
-    if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
-        type_ == "LINESMOOTHING_BANDED RELAXATION" ||
-        type_ == "LINESMOOTHING_BANDEDRELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCK_RELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCK RELAXATION"  ||
-        type_ == "LINESMOOTHING_BLOCKRELAXATION" ) {
-      ParameterList& myparamList = const_cast<ParameterList&>(this->GetParameterList());
-
-      LO CoarseNumZLayers = currentLevel.Get<LO>("CoarseNumZLayers",Factory::GetFactory("CoarseNumZLayers").get());
-      if (CoarseNumZLayers > 0) {
-        Teuchos::ArrayRCP<LO> TVertLineIdSmoo = currentLevel.Get< Teuchos::ArrayRCP<LO> >("LineDetection_VertLineIds", Factory::GetFactory("LineDetection_VertLineIds").get());
-
-        // determine number of local parts
-        LO maxPart = 0;
-        for(size_t k = 0; k < Teuchos::as<size_t>(TVertLineIdSmoo.size()); k++) {
-          if(maxPart < TVertLineIdSmoo[k]) maxPart = TVertLineIdSmoo[k];
-        }
-
-        size_t numLocalRows = A_->getLocalNumRows();
-        TEUCHOS_TEST_FOR_EXCEPTION(numLocalRows % TVertLineIdSmoo.size() != 0, Exceptions::RuntimeError, "MueLu::Ifpack2Smoother::Setup(): the number of local nodes is incompatible with the TVertLineIdsSmoo.");
-
-        if (numLocalRows == Teuchos::as<size_t>(TVertLineIdSmoo.size())) {
-          myparamList.set("partitioner: type","user");
-          myparamList.set("partitioner: map",&(TVertLineIdSmoo[0]));
-          myparamList.set("partitioner: local parts",maxPart+1);
-        } else {
-          // we assume a constant number of DOFs per node
-          size_t numDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
-
-          // Create a new Teuchos::ArrayRCP<LO> of size numLocalRows and fill it with the corresponding information
-          Teuchos::ArrayRCP<LO> partitionerMap(numLocalRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
-          for (size_t blockRow = 0; blockRow < Teuchos::as<size_t>(TVertLineIdSmoo.size()); ++blockRow)
-            for (size_t dof = 0; dof < numDofsPerNode; dof++)
-              partitionerMap[blockRow * numDofsPerNode + dof] = TVertLineIdSmoo[blockRow];
-          myparamList.set("partitioner: type","user");
-          myparamList.set("partitioner: map",&(partitionerMap[0]));
-          myparamList.set("partitioner: local parts",maxPart + 1);
-        }
-
-        if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
-            type_ == "LINESMOOTHING_BANDED RELAXATION" ||
-            type_ == "LINESMOOTHING_BANDEDRELAXATION")
-          type_ = "block relaxation";
-        else
-          type_ = "block relaxation";
+      if (numLocalRows == Teuchos::as<size_t>(TVertLineIdSmoo.size())) {
+        myparamList.set("partitioner: type", "user");
+        myparamList.set("partitioner: map", &(TVertLineIdSmoo[0]));
+        myparamList.set("partitioner: local parts", maxPart + 1);
       } else {
-        // line detection failed -> fallback to point-wise relaxation
-        this->GetOStream(Runtime0) << "Line detection failed: fall back to point-wise relaxation" << std::endl;
-        myparamList.remove("partitioner: type",false);
-        myparamList.remove("partitioner: map", false);
-        myparamList.remove("partitioner: local parts",false);
-        type_ = "point relaxation stand-alone";
+        // we assume a constant number of DOFs per node
+        size_t numDofsPerNode = numLocalRows / TVertLineIdSmoo.size();
+
+        // Create a new Teuchos::ArrayRCP<LO> of size numLocalRows and fill it with the corresponding information
+        Teuchos::ArrayRCP<LO> partitionerMap(numLocalRows, Teuchos::OrdinalTraits<LocalOrdinal>::invalid());
+        for (size_t blockRow = 0; blockRow < Teuchos::as<size_t>(TVertLineIdSmoo.size()); ++blockRow)
+          for (size_t dof = 0; dof < numDofsPerNode; dof++)
+            partitionerMap[blockRow * numDofsPerNode + dof] = TVertLineIdSmoo[blockRow];
+        myparamList.set("partitioner: type", "user");
+        myparamList.set("partitioner: map", &(partitionerMap[0]));
+        myparamList.set("partitioner: local parts", maxPart + 1);
       }
 
-    } // if (type_ == "LINESMOOTHING_BANDEDRELAXATION")
-
-    if(type_ == "AGGREGATE") {
-      SetupAggregate(currentLevel);
-    }
-
-    else {
-      // If we're using a linear partitioner and haven't set the # local parts, set it to match the operator's block size
-      ParameterList precList = this->GetParameterList();
-      if(precList.isParameter("partitioner: type") && precList.get<std::string>("partitioner: type") == "linear" &&
-         !precList.isParameter("partitioner: local parts")) {
-        precList.set("partitioner: local parts", (int)A_->getLocalNumRows() / A_->GetFixedBlockSize());
-      }
-      
-      
-      RCP<Epetra_CrsMatrix> epA = Utilities::Op2NonConstEpetraCrs(A_);
-      
-      Ifpack factory;
-      prec_ = rcp(factory.Create(type_, &(*epA), overlap_));
-      TEUCHOS_TEST_FOR_EXCEPTION(prec_.is_null(), Exceptions::RuntimeError, "Could not create an Ifpack preconditioner with type = \"" << type_ << "\"");
-      SetPrecParameters();
-      prec_->Compute();
+      if (type_ == "LINESMOOTHING_BANDED_RELAXATION" ||
+          type_ == "LINESMOOTHING_BANDED RELAXATION" ||
+          type_ == "LINESMOOTHING_BANDEDRELAXATION")
+        type_ = "block relaxation";
+      else
+        type_ = "block relaxation";
+    } else {
+      // line detection failed -> fallback to point-wise relaxation
+      this->GetOStream(Runtime0) << "Line detection failed: fall back to point-wise relaxation" << std::endl;
+      myparamList.remove("partitioner: type", false);
+      myparamList.remove("partitioner: map", false);
+      myparamList.remove("partitioner: local parts", false);
+      type_ = "point relaxation stand-alone";
     }
 
-    SmootherPrototype::IsSetup(true);
-
-    if (type_ == "Chebyshev" && lambdaMax == -1.0) {
-      Teuchos::RCP<Ifpack_Chebyshev> chebyPrec = rcp_dynamic_cast<Ifpack_Chebyshev>(prec_);
-      if (chebyPrec != Teuchos::null) {
-        lambdaMax = chebyPrec->GetLambdaMax();
-        A_->SetMaxEigenvalueEstimate(lambdaMax);
-        this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack)" << " = " << lambdaMax << std::endl;
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax == -1.0, Exceptions::RuntimeError, "MueLu::IfpackSmoother::Setup(): no maximum eigenvalue estimate");
-    }
+  }  // if (type_ == "LINESMOOTHING_BANDEDRELAXATION")
 
-    this->GetOStream(Statistics1) << description() << std::endl;
+  if (type_ == "AGGREGATE") {
+    SetupAggregate(currentLevel);
   }
 
-
-  template <class Node>
-  void IfpackSmoother<Node>::SetupAggregate(Level& currentLevel) {
-
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-
-    if (this->IsSetup() == true) {
-      this->GetOStream(Warnings0) << "MueLu::Ifpack2moother::SetupAggregate(): Setup() has already been called" << std::endl;
-      this->GetOStream(Warnings0) << "MueLu::IfpackSmoother::SetupAggregate(): reuse of this type is not available, reverting to full construction" << std::endl;
-    }
-
-    this->GetOStream(Statistics0) << "IfpackSmoother: Using Aggregate Smoothing"<<std::endl;
-
-    RCP<Aggregates> aggregates = Factory::Get<RCP<Aggregates> >(currentLevel,"Aggregates");
-    RCP<const LOMultiVector> vertex2AggId = aggregates->GetVertex2AggId();
-    ArrayRCP<LO> aggregate_ids = rcp_const_cast<LOMultiVector>(vertex2AggId)->getDataNonConst(0);
-    ArrayRCP<LO> dof_ids;
-
-    // We need to unamalgamate, if the FixedBlockSize > 1
-    if(A_->GetFixedBlockSize() > 1) {
-      // NOTE: We're basically going to have to leave a deallocated pointer hanging out
-      // in the paramList object (and inside the partitioner).  This never gets
-      // use again after Compute() gets called, so this is OK, but I'm still leaving
-      // this note here in case it bites us again later.
-      LO blocksize = (LO) A_->GetFixedBlockSize();
-      dof_ids.resize(aggregate_ids.size() * blocksize);
-      for(LO i=0; i<(LO)aggregate_ids.size(); i++) {
-        for(LO j=0; j<(LO)blocksize; j++)
-          dof_ids[i*blocksize+j] = aggregate_ids[i];
-      }
-    }
-    else {
-      dof_ids = aggregate_ids;
+  else {
+    // If we're using a linear partitioner and haven't set the # local parts, set it to match the operator's block size
+    ParameterList precList = this->GetParameterList();
+    if (precList.isParameter("partitioner: type") && precList.get<std::string>("partitioner: type") == "linear" &&
+        !precList.isParameter("partitioner: local parts")) {
+      precList.set("partitioner: local parts", (int)A_->getLocalNumRows() / A_->GetFixedBlockSize());
     }
-        
-    paramList.set("partitioner: map", dof_ids.getRawPtr());
-    paramList.set("partitioner: type", "user");
-    paramList.set("partitioner: overlap", 0);
-    paramList.set("partitioner: local parts", (int)aggregates->GetNumAggregates());
-    // In case of Dirichlet nodes
-    paramList.set("partitioner: keep singletons",true);
 
-    RCP<Epetra_CrsMatrix> A = Utilities::Op2NonConstEpetraCrs(A_);
-    type_ = "block relaxation stand-alone";
+    RCP<Epetra_CrsMatrix> epA = Utilities::Op2NonConstEpetraCrs(A_);
 
     Ifpack factory;
-    prec_ = rcp(factory.Create(type_, &(*A), overlap_));
+    prec_ = rcp(factory.Create(type_, &(*epA), overlap_));
     TEUCHOS_TEST_FOR_EXCEPTION(prec_.is_null(), Exceptions::RuntimeError, "Could not create an Ifpack preconditioner with type = \"" << type_ << "\"");
     SetPrecParameters();
+    prec_->Compute();
+  }
 
-    int rv = prec_->Compute();
-    TEUCHOS_TEST_FOR_EXCEPTION(rv, Exceptions::RuntimeError, "Ifpack preconditioner with type = \"" << type_ << "\" Compute() call failed.");
+  SmootherPrototype::IsSetup(true);
 
+  if (type_ == "Chebyshev" && lambdaMax == -1.0) {
+    Teuchos::RCP<Ifpack_Chebyshev> chebyPrec = rcp_dynamic_cast<Ifpack_Chebyshev>(prec_);
+    if (chebyPrec != Teuchos::null) {
+      lambdaMax = chebyPrec->GetLambdaMax();
+      A_->SetMaxEigenvalueEstimate(lambdaMax);
+      this->GetOStream(Statistics1) << "chebyshev: max eigenvalue (calculated by Ifpack)"
+                                    << " = " << lambdaMax << std::endl;
+    }
+    TEUCHOS_TEST_FOR_EXCEPTION(lambdaMax == -1.0, Exceptions::RuntimeError, "MueLu::IfpackSmoother::Setup(): no maximum eigenvalue estimate");
   }
 
+  this->GetOStream(Statistics1) << description() << std::endl;
+}
 
-  template <class Node>
-  void IfpackSmoother<Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::IfpackSmoother::Apply(): Setup() has not been called");
-
+template <class Node>
+void IfpackSmoother<Node>::SetupAggregate(Level& currentLevel) {
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
 
-    // Forward the InitialGuessIsZero option to Ifpack
-    Teuchos::ParameterList paramList;
-    bool supportInitialGuess = false;
-    if (type_ == "Chebyshev") {
-      paramList.set("chebyshev: zero starting solution",  InitialGuessIsZero);
-      supportInitialGuess = true;
+  if (this->IsSetup() == true) {
+    this->GetOStream(Warnings0) << "MueLu::Ifpack2moother::SetupAggregate(): Setup() has already been called" << std::endl;
+    this->GetOStream(Warnings0) << "MueLu::IfpackSmoother::SetupAggregate(): reuse of this type is not available, reverting to full construction" << std::endl;
+  }
 
-    } else if (type_ == "point relaxation stand-alone") {
-      paramList.set("relaxation: zero starting solution", InitialGuessIsZero);
-      supportInitialGuess = true;
+  this->GetOStream(Statistics0) << "IfpackSmoother: Using Aggregate Smoothing" << std::endl;
+
+  RCP<Aggregates> aggregates            = Factory::Get<RCP<Aggregates> >(currentLevel, "Aggregates");
+  RCP<const LOMultiVector> vertex2AggId = aggregates->GetVertex2AggId();
+  ArrayRCP<LO> aggregate_ids            = rcp_const_cast<LOMultiVector>(vertex2AggId)->getDataNonConst(0);
+  ArrayRCP<LO> dof_ids;
+
+  // We need to unamalgamate, if the FixedBlockSize > 1
+  if (A_->GetFixedBlockSize() > 1) {
+    // NOTE: We're basically going to have to leave a deallocated pointer hanging out
+    // in the paramList object (and inside the partitioner).  This never gets
+    // use again after Compute() gets called, so this is OK, but I'm still leaving
+    // this note here in case it bites us again later.
+    LO blocksize = (LO)A_->GetFixedBlockSize();
+    dof_ids.resize(aggregate_ids.size() * blocksize);
+    for (LO i = 0; i < (LO)aggregate_ids.size(); i++) {
+      for (LO j = 0; j < (LO)blocksize; j++)
+        dof_ids[i * blocksize + j] = aggregate_ids[i];
     }
+  } else {
+    dof_ids = aggregate_ids;
+  }
 
-    SetPrecParameters(paramList);
+  paramList.set("partitioner: map", dof_ids.getRawPtr());
+  paramList.set("partitioner: type", "user");
+  paramList.set("partitioner: overlap", 0);
+  paramList.set("partitioner: local parts", (int)aggregates->GetNumAggregates());
+  // In case of Dirichlet nodes
+  paramList.set("partitioner: keep singletons", true);
+
+  RCP<Epetra_CrsMatrix> A = Utilities::Op2NonConstEpetraCrs(A_);
+  type_                   = "block relaxation stand-alone";
+
+  Ifpack factory;
+  prec_ = rcp(factory.Create(type_, &(*A), overlap_));
+  TEUCHOS_TEST_FOR_EXCEPTION(prec_.is_null(), Exceptions::RuntimeError, "Could not create an Ifpack preconditioner with type = \"" << type_ << "\"");
+  SetPrecParameters();
+
+  int rv = prec_->Compute();
+  TEUCHOS_TEST_FOR_EXCEPTION(rv, Exceptions::RuntimeError, "Ifpack preconditioner with type = \"" << type_ << "\" Compute() call failed.");
+}
+
+template <class Node>
+void IfpackSmoother<Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::IfpackSmoother::Apply(): Setup() has not been called");
+
+  // Forward the InitialGuessIsZero option to Ifpack
+  Teuchos::ParameterList paramList;
+  bool supportInitialGuess = false;
+  if (type_ == "Chebyshev") {
+    paramList.set("chebyshev: zero starting solution", InitialGuessIsZero);
+    supportInitialGuess = true;
+
+  } else if (type_ == "point relaxation stand-alone") {
+    paramList.set("relaxation: zero starting solution", InitialGuessIsZero);
+    supportInitialGuess = true;
+  }
 
-    // Apply
-    if (InitialGuessIsZero || supportInitialGuess) {
-      Epetra_MultiVector&       epX = Utilities::MV2NonConstEpetraMV(X);
-      const Epetra_MultiVector& epB = Utilities::MV2EpetraMV(B);
+  SetPrecParameters(paramList);
 
-      prec_->ApplyInverse(epB, epX);
+  // Apply
+  if (InitialGuessIsZero || supportInitialGuess) {
+    Epetra_MultiVector& epX       = Utilities::MV2NonConstEpetraMV(X);
+    const Epetra_MultiVector& epB = Utilities::MV2EpetraMV(B);
 
-    } else {
-      RCP<MultiVector> Residual   = Utilities::Residual(*A_, X, B);
-      RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
+    prec_->ApplyInverse(epB, epX);
 
-      Epetra_MultiVector&       epX = Utilities::MV2NonConstEpetraMV(*Correction);
-      const Epetra_MultiVector& epB = Utilities::MV2EpetraMV(*Residual);
+  } else {
+    RCP<MultiVector> Residual   = Utilities::Residual(*A_, X, B);
+    RCP<MultiVector> Correction = MultiVectorFactory::Build(A_->getDomainMap(), X.getNumVectors());
 
-      prec_->ApplyInverse(epB, epX);
+    Epetra_MultiVector& epX       = Utilities::MV2NonConstEpetraMV(*Correction);
+    const Epetra_MultiVector& epB = Utilities::MV2EpetraMV(*Residual);
 
-      X.update(1.0, *Correction, 1.0);
-    }
-  }
+    prec_->ApplyInverse(epB, epX);
 
-  template <class Node>
-  RCP<MueLu::SmootherPrototype<double, int, int, Node> > IfpackSmoother<Node>::Copy() const {
-    RCP<IfpackSmoother<Node> > smoother = rcp(new IfpackSmoother<Node>(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return Teuchos::rcp_dynamic_cast<MueLu::SmootherPrototype<double, int, int, Node> >(smoother);
+    X.update(1.0, *Correction, 1.0);
   }
-
-  template <class Node>
-  std::string IfpackSmoother<Node>::description() const {
-    std::ostringstream out;
-    // The check "GetVerbLevel() == Test" is to avoid
-    // failures in the EasyInterface test.
-    if (prec_ == Teuchos::null || this->GetVerbLevel() == InterfaceTest) {
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-    } else {
-      out << prec_->Label();
-    }
-    return out.str();
+}
+
+template <class Node>
+RCP<MueLu::SmootherPrototype<double, int, int, Node> > IfpackSmoother<Node>::Copy() const {
+  RCP<IfpackSmoother<Node> > smoother = rcp(new IfpackSmoother<Node>(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return Teuchos::rcp_dynamic_cast<MueLu::SmootherPrototype<double, int, int, Node> >(smoother);
+}
+
+template <class Node>
+std::string IfpackSmoother<Node>::description() const {
+  std::ostringstream out;
+  // The check "GetVerbLevel() == Test" is to avoid
+  // failures in the EasyInterface test.
+  if (prec_ == Teuchos::null || this->GetVerbLevel() == InterfaceTest) {
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+  } else {
+    out << prec_->Label();
   }
+  return out.str();
+}
 
-  template <class Node>
-  void IfpackSmoother<Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+template <class Node>
+void IfpackSmoother<Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
+    out0 << "Overlap: " << overlap_ << std::endl;
+  }
+
+  if (verbLevel & External)
+    if (prec_ != Teuchos::null) {
       Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-      out0 << "Overlap: "        << overlap_ << std::endl;
+      out << *prec_ << std::endl;
     }
 
-    if (verbLevel & External)
-      if (prec_ != Teuchos::null) {
-        Teuchos::OSTab tab2(out);
-        out << *prec_ << std::endl;
-      }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "RCP<A_>: " << A_ << std::endl
-           << "RCP<prec_>: " << prec_ << std::endl;
-    }
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "RCP<A_>: " << A_ << std::endl
+         << "RCP<prec_>: " << prec_ << std::endl;
   }
+}
 
-  template <class Node>
-  size_t IfpackSmoother<Node>::getNodeSmootherComplexity() const {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
-  
+template <class Node>
+size_t IfpackSmoother<Node>::getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 // The IfpackSmoother is only templated on the Node, since it is an Epetra only object
 // Therefore we do not need the full ETI instantiations as we do for the other MueLu
diff --git a/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.hpp b/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.hpp
index b66b758e5e4e..9e908092c252 100644
--- a/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_IfpackSmoother.hpp
@@ -65,167 +65,164 @@ class Ifpack_Preconditioner;
 
 namespace MueLu {
 
-  /*!
-    @class IfpackSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Ifpack smoothers.
-
-    This class creates an Ifpack preconditioner factory. The factory creates a smoother based on the
-    type and ParameterList passed into the constructor. See the constructor for more information.
-  */
-  template <class Node = typename SmootherPrototype<double,int,int>::node_type>
-  class IfpackSmoother : public MueLu::SmootherPrototype<double,int,int,Node> {
-    typedef double Scalar;
-    typedef int LocalOrdinal;
-    typedef int GlobalOrdinal;
+/*!
+  @class IfpackSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Ifpack smoothers.
+
+  This class creates an Ifpack preconditioner factory. The factory creates a smoother based on the
+  type and ParameterList passed into the constructor. See the constructor for more information.
+*/
+template <class Node = typename SmootherPrototype<double, int, int>::node_type>
+class IfpackSmoother : public MueLu::SmootherPrototype<double, int, int, Node> {
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
 #undef MUELU_IFPACKSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-
-    /*! @brief Constructor
-
-    The options passed into IfpackSmoother are those given in the Ifpack user's manual.
-
-    @param type smoother type
-    @param list options for the particular smoother (e.g., fill factor or damping parameter)
-
-    Here is how to select some of the most common smoothers.
-
-    - Gauss-Seidel
-    - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
-    - parameter list options
-    - <tt>relaxation: type</tt> = <tt>Gauss-Seidel</tt>
-    - <tt>relaxation: damping factor</tt>
-    - symmetric Gauss-Seidel
-    - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
-    - parameter list options
-    - <tt>relaxation: type</tt> = <tt>symmetric Gauss-Seidel</tt>
-    - <tt>relaxation: damping factor</tt>
-    - Chebyshev
-    - <tt>type</tt> = <tt>Chebyshev</tt>
-    - parameter list options
-    - <tt>chebyshev: ratio eigenvalue</tt>
-    - <tt>chebyshev: min eigenvalue</tt>
-    - <tt>chebyshev: max eigenvalue</tt>
-    - <tt>chebyshev: degree</tt>
-    - <tt>chebyshev: zero starting solution</tt> (defaults to <tt>true</tt>)
-    - ILU
-    - <tt>type</tt> = <tt>ILU</tt>
-    - parameter list options
-    - <tt>fact: level-of-fill</tt>
-
-    See also Ifpack_PointRelaxation, Ifpack_Chebyshev, Ifpack_ILU.
-    */
-    IfpackSmoother(std::string const & type, Teuchos::ParameterList const & paramList = Teuchos::ParameterList(), LO const &overlap=0); //TODO: empty paramList valid for Ifpack??
-
-    //! Destructor
-    virtual ~IfpackSmoother() { }
+ public:
+  //! @name Constructors / destructors
+  //@{
+
+  /*! @brief Constructor
+
+  The options passed into IfpackSmoother are those given in the Ifpack user's manual.
+
+  @param type smoother type
+  @param list options for the particular smoother (e.g., fill factor or damping parameter)
+
+  Here is how to select some of the most common smoothers.
+
+  - Gauss-Seidel
+  - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
+  - parameter list options
+  - <tt>relaxation: type</tt> = <tt>Gauss-Seidel</tt>
+  - <tt>relaxation: damping factor</tt>
+  - symmetric Gauss-Seidel
+  - <tt>type</tt> = <tt>point relaxation stand-alone</tt>
+  - parameter list options
+  - <tt>relaxation: type</tt> = <tt>symmetric Gauss-Seidel</tt>
+  - <tt>relaxation: damping factor</tt>
+  - Chebyshev
+  - <tt>type</tt> = <tt>Chebyshev</tt>
+  - parameter list options
+  - <tt>chebyshev: ratio eigenvalue</tt>
+  - <tt>chebyshev: min eigenvalue</tt>
+  - <tt>chebyshev: max eigenvalue</tt>
+  - <tt>chebyshev: degree</tt>
+  - <tt>chebyshev: zero starting solution</tt> (defaults to <tt>true</tt>)
+  - ILU
+  - <tt>type</tt> = <tt>ILU</tt>
+  - parameter list options
+  - <tt>fact: level-of-fill</tt>
+
+  See also Ifpack_PointRelaxation, Ifpack_Chebyshev, Ifpack_ILU.
+  */
+  IfpackSmoother(std::string const& type, Teuchos::ParameterList const& paramList = Teuchos::ParameterList(), LO const& overlap = 0);  // TODO: empty paramList valid for Ifpack??
 
-    //@}
+  //! Destructor
+  virtual ~IfpackSmoother() {}
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  //@}
 
-    //! Input
-    //@{
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Computational methods.
-    //@{
+  //@}
 
-    /*! @brief Set up the smoother.
+  //! @name Computational methods.
+  //@{
 
-    This creates the underlying Ifpack smoother object, copies any parameter list options
-    supplied to the constructor to the Ifpack object, and computes the preconditioner.
-    */
-    void Setup(Level &currentLevel);
+  /*! @brief Set up the smoother.
 
-    /*! @brief Apply the preconditioner.
+  This creates the underlying Ifpack smoother object, copies any parameter list options
+  supplied to the constructor to the Ifpack object, and computes the preconditioner.
+  */
+  void Setup(Level& currentLevel);
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  /*! @brief Apply the preconditioner.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided.  Whether this actually saves any work depends on the underlying Ifpack implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    //@}
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided.  Whether this actually saves any work depends on the underlying Ifpack implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //! @name Utilities
-    //@{
+  //@}
 
-    RCP<SmootherPrototype> Copy() const;
+  //! @name Utilities
+  //@{
 
-    //@}
+  RCP<SmootherPrototype> Copy() const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
-    void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const;
+  //@}
 
-    void SetupAggregate(Level& currentLevel);
+ private:
+  void SetPrecParameters(const Teuchos::ParameterList& list = Teuchos::ParameterList()) const;
 
-  private:
+  void SetupAggregate(Level& currentLevel);
 
-    //! ifpack-specific key phrase that denote smoother type
-    std::string type_;
+ private:
+  //! ifpack-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! overlap when using the smoother in additive Schwarz mode
-    LO overlap_;
+  //! overlap when using the smoother in additive Schwarz mode
+  LO overlap_;
 
-    //! Matrix. Not used directly, but held inside of prec_. So we have to keep an RCP pointer to it!
-    RCP<Matrix> A_;
+  //! Matrix. Not used directly, but held inside of prec_. So we have to keep an RCP pointer to it!
+  RCP<Matrix> A_;
 
-    //! pointer to Ifpack solver object
-    // Note: prec_ must be destroyed before A_, so declaration of prec_ appears after declaration of A_
-    RCP<Ifpack_Preconditioner> prec_;
+  //! pointer to Ifpack solver object
+  // Note: prec_ must be destroyed before A_, so declaration of prec_ appears after declaration of A_
+  RCP<Ifpack_Preconditioner> prec_;
 
-  }; // class IfpackSmoother
+};  // class IfpackSmoother
 
-  //! Non-member templated function GetIfpackSmoother() returns a new IfpackSmoother object when <Scalar, LocalOrdinal, GlobalOrdinal> == <double, int, int>. Otherwise, an exception is thrown.
-  //! This function simplifies the usage of IfpackSmoother objects inside of templates as templates do not have to be specialized for <double, int, int> (see DirectSolver for an example).
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  GetIfpackSmoother (const std::string& /* type */ = "",
-                     const Teuchos::ParameterList& /* paramList */ = Teuchos::ParameterList (),
-                     const LocalOrdinal& /* overlap */ = 0)
-  {
-    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "IfpackSmoother cannot be used with Scalar != double, LocalOrdinal != int, GlobalOrdinal != int");
-    TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
-  }
+//! Non-member templated function GetIfpackSmoother() returns a new IfpackSmoother object when <Scalar, LocalOrdinal, GlobalOrdinal> == <double, int, int>. Otherwise, an exception is thrown.
+//! This function simplifies the usage of IfpackSmoother objects inside of templates as templates do not have to be specialized for <double, int, int> (see DirectSolver for an example).
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GetIfpackSmoother(const std::string& /* type */                 = "",
+                  const Teuchos::ParameterList& /* paramList */ = Teuchos::ParameterList(),
+                  const LocalOrdinal& /* overlap */             = 0) {
+  TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "IfpackSmoother cannot be used with Scalar != double, LocalOrdinal != int, GlobalOrdinal != int");
+  TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+}
 
-  // Specialization for serial node (used for Epetra)
+// Specialization for serial node (used for Epetra)
 #if defined(HAVE_MUELU_EPETRA)
-  template <>
-  inline RCP<MueLu::SmootherPrototype<double, int, int, Xpetra::EpetraNode> >
-  GetIfpackSmoother<double, int, int, Xpetra::EpetraNode> (const std::string& type, const Teuchos::ParameterList& paramList, const int& overlap) {
-    return rcp(new MueLu::IfpackSmoother<Xpetra::EpetraNode>(type, paramList, overlap));
-  }
+template <>
+inline RCP<MueLu::SmootherPrototype<double, int, int, Xpetra::EpetraNode> >
+GetIfpackSmoother<double, int, int, Xpetra::EpetraNode>(const std::string& type, const Teuchos::ParameterList& paramList, const int& overlap) {
+  return rcp(new MueLu::IfpackSmoother<Xpetra::EpetraNode>(type, paramList, overlap));
+}
 #endif
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_IFPACKSMOOTHER_SHORT
-#endif // HAVE_MUELU_EPETRA && HAVE_MUELU_IFPACK
-#endif // MUELU_IFPACKSMOOTHER_HPP
+#endif  // HAVE_MUELU_EPETRA && HAVE_MUELU_IFPACK
+#endif  // MUELU_IFPACKSMOOTHER_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_MergedSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_MergedSmoother_decl.hpp
index 71b5b7c6d0fa..2adcdc4fa8e1 100644
--- a/packages/muelu/src/Smoothers/MueLu_MergedSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_MergedSmoother_decl.hpp
@@ -56,96 +56,96 @@
 
 namespace MueLu {
 
-  class Level;
+class Level;
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class MergedSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class MergedSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_MERGEDSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! Constructor
-    MergedSmoother(ArrayRCP<RCP<SmootherPrototype> >& smootherList, bool verbose = false);
+  //! Constructor
+  MergedSmoother(ArrayRCP<RCP<SmootherPrototype> >& smootherList, bool verbose = false);
 
-    //! Copy constructor (performs a deep copy of input object)
-    MergedSmoother(const MergedSmoother& src);
+  //! Copy constructor (performs a deep copy of input object)
+  MergedSmoother(const MergedSmoother& src);
 
-    //! Copy method (performs a deep copy of input object)
-    RCP<SmootherPrototype> Copy() const;
+  //! Copy method (performs a deep copy of input object)
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Destructor
-    virtual ~MergedSmoother() { }
-    //@}
+  //! Destructor
+  virtual ~MergedSmoother() {}
+  //@}
 
-    //! @name Set/Get methods
-    //@{
+  //! @name Set/Get methods
+  //@{
 
-    void StandardOrder() { reverseOrder_ = false; }
-    void ReverseOrder()  { reverseOrder_ = true;  }
+  void StandardOrder() { reverseOrder_ = false; }
+  void ReverseOrder() { reverseOrder_ = true; }
 
-    // TODO: GetOrder() is a better name (+ enum to define order)
-    bool GetReverseOrder() const { return reverseOrder_; }
+  // TODO: GetOrder() is a better name (+ enum to define order)
+  bool GetReverseOrder() const { return reverseOrder_; }
 
-    //  TODO  const ArrayRCP<const RCP<const SmootherPrototype> > & GetSmootherList() const;
-    const ArrayRCP<const RCP<SmootherPrototype> > GetSmootherList() const { return smootherList_; }
+  //  TODO  const ArrayRCP<const RCP<const SmootherPrototype> > & GetSmootherList() const;
+  const ArrayRCP<const RCP<SmootherPrototype> > GetSmootherList() const { return smootherList_; }
 
-    //@}
+  //@}
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @name Setup and Apply methods.
+  //@{
 
-    /*! @brief Set up. */
-    void Setup(Level& level);
+  /*! @brief Set up. */
+  void Setup(Level& level);
 
-    /*! @brief Apply
+  /*! @brief Apply
 
-    Solves the linear system <tt>AX=B</tt> using the smoothers of the list.
+  Solves the linear system <tt>AX=B</tt> using the smoothers of the list.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //@}
+  //@}
 
-    //! Custom SetFactory
-    void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
+  //! Custom SetFactory
+  void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
 
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    void CopyParameters(RCP<SmootherPrototype> src); // TODO: wrong prototype. We do not need an RCP here.
+  void CopyParameters(RCP<SmootherPrototype> src);  // TODO: wrong prototype. We do not need an RCP here.
 
-    ArrayRCP<RCP<SmootherPrototype> > SmootherListDeepCopy(const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList);
+  ArrayRCP<RCP<SmootherPrototype> > SmootherListDeepCopy(const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList);
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-    //@}
+  //@}
 
-  private:
-    // List of smoothers. It is an ArrayRCP of RCP because:
-    //  1) I need a vector of pointers (to avoid slicing problems)
-    //  2) I can use an std::vector insead of an ArrayRCP but then the constructor will do a copy of user input
-    ArrayRCP<RCP<SmootherPrototype> > smootherList_;
+ private:
+  // List of smoothers. It is an ArrayRCP of RCP because:
+  //  1) I need a vector of pointers (to avoid slicing problems)
+  //  2) I can use an std::vector insead of an ArrayRCP but then the constructor will do a copy of user input
+  ArrayRCP<RCP<SmootherPrototype> > smootherList_;
 
-    //
-    bool reverseOrder_;
+  //
+  bool reverseOrder_;
 
-    // tmp, for debug
-    bool verbose_;
+  // tmp, for debug
+  bool verbose_;
 
-  }; //class MergedSmoother
+};  // class MergedSmoother
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MERGEDSMOOTHER_SHORT
-#endif // MUELU_MERGEDSMOOTHER_DECL_HPP
+#endif  // MUELU_MERGEDSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_MergedSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_MergedSmoother_def.hpp
index 05d27789784f..1b42cd533241 100644
--- a/packages/muelu/src/Smoothers/MueLu_MergedSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_MergedSmoother_def.hpp
@@ -51,175 +51,172 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedSmoother(ArrayRCP<RCP<SmootherPrototype> > & smootherList, bool verbose)
-    : smootherList_(smootherList), reverseOrder_(false), verbose_(verbose) {
-    // TODO: check that on each method TEUCHOS_TEST_FOR_EXCEPTION(smootherList == Teuchos::null, MueLu::Exceptions::RuntimeError, "");
-
-    SmootherPrototype::IsSetup(false);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedSmoother(const MergedSmoother& src)
-    : reverseOrder_(src.reverseOrder_), verbose_(src.verbose_) {
-    // Deep copy of src.smootherList_
-    smootherList_ = SmootherListDeepCopy(src.GetSmootherList());
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
-    // We need to propagate SetFactory to proper place
-    for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); it++)
-      (*it)->SetFactory(varName, factory);
-  }
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); ++it)
-      (*it)->DeclareInput(currentLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& level) {
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::MergedSmoother::Setup(): Setup() has already been called";
-
-    for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); ++it) {
-      try {
-        (*it)->Setup(level);
-
-      } catch(MueLu::Exceptions::RuntimeError& e) {
-        std::string msg = "MueLu::MergedSmoother<>::Setup(): Runtime Error.\n One of the underlying smoother throwed the following exception: \n";
-        msg += e.what();
-        throw MueLu::Exceptions::RuntimeError(msg);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedSmoother(ArrayRCP<RCP<SmootherPrototype> >& smootherList, bool verbose)
+  : smootherList_(smootherList)
+  , reverseOrder_(false)
+  , verbose_(verbose) {
+  // TODO: check that on each method TEUCHOS_TEST_FOR_EXCEPTION(smootherList == Teuchos::null, MueLu::Exceptions::RuntimeError, "");
+
+  SmootherPrototype::IsSetup(false);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MergedSmoother(const MergedSmoother& src)
+  : reverseOrder_(src.reverseOrder_)
+  , verbose_(src.verbose_) {
+  // Deep copy of src.smootherList_
+  smootherList_ = SmootherListDeepCopy(src.GetSmootherList());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
+  // We need to propagate SetFactory to proper place
+  for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); it++)
+    (*it)->SetFactory(varName, factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); ++it)
+    (*it)->DeclareInput(currentLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& level) {
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::MergedSmoother::Setup(): Setup() has already been called";
+
+  for (typename ArrayView<RCP<SmootherPrototype> >::iterator it = smootherList_.begin(); it != smootherList_.end(); ++it) {
+    try {
+      (*it)->Setup(level);
+
+    } catch (MueLu::Exceptions::RuntimeError& e) {
+      std::string msg = "MueLu::MergedSmoother<>::Setup(): Runtime Error.\n One of the underlying smoother throwed the following exception: \n";
+      msg += e.what();
+      throw MueLu::Exceptions::RuntimeError(msg);
     }
-
-    SmootherPrototype::IsSetup(true);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, MueLu::Exceptions::RuntimeError, "MueLu::MergedSmoother<>:Apply(): Setup() has not been called");
+  SmootherPrototype::IsSetup(true);
+}
 
-    typedef typename ArrayRCP<RCP<SmootherPrototype> >::size_type sz_t;
-    sz_t n = smootherList_.size(), c = (reverseOrder_ ? n-1 : 0);
-    char d = (reverseOrder_ ? -1 : 1);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, MueLu::Exceptions::RuntimeError, "MueLu::MergedSmoother<>:Apply(): Setup() has not been called");
 
-    for (sz_t i = 0; i < n; i++)  // loop unifying both forward and reverse order
-      try {
-        // Be careful with nonnegative numbers
-        smootherList_[c + d*Teuchos::as<char>(i)]->Apply(X, B, InitialGuessIsZero);
+  typedef typename ArrayRCP<RCP<SmootherPrototype> >::size_type sz_t;
+  sz_t n = smootherList_.size(), c = (reverseOrder_ ? n - 1 : 0);
+  char d = (reverseOrder_ ? -1 : 1);
 
-        // For second and later iterations, initial guess = previous result
-        InitialGuessIsZero = false;
+  for (sz_t i = 0; i < n; i++)  // loop unifying both forward and reverse order
+    try {
+      // Be careful with nonnegative numbers
+      smootherList_[c + d * Teuchos::as<char>(i)]->Apply(X, B, InitialGuessIsZero);
 
-      } catch(MueLu::Exceptions::RuntimeError& e) {
-        std::string msg = "MueLu::MergedSmoother<>::Apply(): Runtime Error. One of the underlying smoothers throws the following exception: \n";
-        msg += e.what();
-        throw MueLu::Exceptions::RuntimeError(msg);
-      }
-  }
+      // For second and later iterations, initial guess = previous result
+      InitialGuessIsZero = false;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& /* out */, const VerbLevel /* verbLevel */) const {
-    throw Exceptions::NotImplemented("MueLu::MergedSmoother<>::Print() is not implemented");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CopyParameters(RCP<SmootherPrototype> src) { // TODO: wrong prototype. We do not need an RCP here.
-    RCP<MergedSmoother> srcMergedSmoother = rcp_dynamic_cast<MergedSmoother>(src); // TODO: check if dynamic cast fails
-
-    reverseOrder_ = srcMergedSmoother->GetReverseOrder();
-
-    {
-      const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList  = srcMergedSmoother->GetSmootherList();
-      const ArrayRCP<const RCP<SmootherPrototype> >& thisSmootherList = smootherList_;
-      TEUCHOS_TEST_FOR_EXCEPTION(srcSmootherList == Teuchos::null, MueLu::Exceptions::RuntimeError,             // might be allowed later if useful
-                                 "MueLu::MergedSmoother<>:CopyParameters(): thisSmootherList == Teuchos::null");
-
-      // If the smootherList of 'this' and 'src' contains the same type of smoothers,
-      // we can transfert parameters from src to 'this' in order to tentatively reuse
-      // the current setup information of each smoothers. Note that the reuse of the
-      // setup phase of the MergedSmoother 'src' can be implemented for a larger set
-      // of cases (and more complicated cases), but it does not seems useful for now.
+    } catch (MueLu::Exceptions::RuntimeError& e) {
+      std::string msg = "MueLu::MergedSmoother<>::Apply(): Runtime Error. One of the underlying smoothers throws the following exception: \n";
+      msg += e.what();
+      throw MueLu::Exceptions::RuntimeError(msg);
+    }
+}
 
-      bool reuse = true; // true == can we transfert parameters of smoothers one by one or do we have to copy the whole list of src?
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& /* out */, const VerbLevel /* verbLevel */) const {
+  throw Exceptions::NotImplemented("MueLu::MergedSmoother<>::Print() is not implemented");
+}
 
-      // test 1: same list size
-      reuse = reuse && (thisSmootherList.size() == srcSmootherList.size());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CopyParameters(RCP<SmootherPrototype> src) {  // TODO: wrong prototype. We do not need an RCP here.
+  RCP<MergedSmoother> srcMergedSmoother = rcp_dynamic_cast<MergedSmoother>(src);                              // TODO: check if dynamic cast fails
 
-      if (reuse) {
-        //  test 2: one-by-one comparison of smoother types
-        for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++) {
-          // The following test should never throw in our use cases because 'src' is a prototype and
-          // 'this' is a real smoother so they don't share any data. We may allow such case later if useful.
-          TEUCHOS_TEST_FOR_EXCEPTION((thisSmootherList[i] == srcSmootherList[i]) && (thisSmootherList[i] != Teuchos::null), MueLu::Exceptions::RuntimeError,
-                                     "MueLu::MergedSmoother<>:CopyParameters(): internal logic error");
+  reverseOrder_ = srcMergedSmoother->GetReverseOrder();
 
-          //TODO
-          // reuse = reuse && ((thisSmootherList[i] == Teuchos::null && srcSmootherList[i] == Teuchos::null) ||
-          //                   thisSmootherList[i]->GetType() == srcSmootherList[i]->GetType());
-        }
+  {
+    const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList  = srcMergedSmoother->GetSmootherList();
+    const ArrayRCP<const RCP<SmootherPrototype> >& thisSmootherList = smootherList_;
+    TEUCHOS_TEST_FOR_EXCEPTION(srcSmootherList == Teuchos::null, MueLu::Exceptions::RuntimeError,  // might be allowed later if useful
+                               "MueLu::MergedSmoother<>:CopyParameters(): thisSmootherList == Teuchos::null");
+
+    // If the smootherList of 'this' and 'src' contains the same type of smoothers,
+    // we can transfert parameters from src to 'this' in order to tentatively reuse
+    // the current setup information of each smoothers. Note that the reuse of the
+    // setup phase of the MergedSmoother 'src' can be implemented for a larger set
+    // of cases (and more complicated cases), but it does not seems useful for now.
+
+    bool reuse = true;  // true == can we transfert parameters of smoothers one by one or do we have to copy the whole list of src?
+
+    // test 1: same list size
+    reuse = reuse && (thisSmootherList.size() == srcSmootherList.size());
+
+    if (reuse) {
+      //  test 2: one-by-one comparison of smoother types
+      for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++) {
+        // The following test should never throw in our use cases because 'src' is a prototype and
+        // 'this' is a real smoother so they don't share any data. We may allow such case later if useful.
+        TEUCHOS_TEST_FOR_EXCEPTION((thisSmootherList[i] == srcSmootherList[i]) && (thisSmootherList[i] != Teuchos::null), MueLu::Exceptions::RuntimeError,
+                                   "MueLu::MergedSmoother<>:CopyParameters(): internal logic error");
+
+        // TODO
+        //  reuse = reuse && ((thisSmootherList[i] == Teuchos::null && srcSmootherList[i] == Teuchos::null) ||
+        //                    thisSmootherList[i]->GetType() == srcSmootherList[i]->GetType());
       }
+    }
 
-      reuse = false; //TODO: temporary disactivated.
+    reuse = false;  // TODO: temporary disactivated.
 
-      if (reuse) {
-        bool isSetup = true;
+    if (reuse) {
+      bool isSetup = true;
 
-        // Call CopyParameters for each smoothers and update IsSetup status of the MergedSmoother
-        for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++) {
-          if (srcSmootherList[i] != Teuchos::null) {
-            TEUCHOS_TEST_FOR_EXCEPTION(srcSmootherList[i] == Teuchos::null, MueLu::Exceptions::RuntimeError, "MueLu::MergedSmoother<>:CopyParameters(): internal logic error");
+      // Call CopyParameters for each smoothers and update IsSetup status of the MergedSmoother
+      for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++) {
+        if (srcSmootherList[i] != Teuchos::null) {
+          TEUCHOS_TEST_FOR_EXCEPTION(srcSmootherList[i] == Teuchos::null, MueLu::Exceptions::RuntimeError, "MueLu::MergedSmoother<>:CopyParameters(): internal logic error");
 
-            //TODO              thisSmootherList[i]->CopyParameters(srcSmootherList[i]);
-            isSetup = isSetup && thisSmootherList[i]->IsSetup();
-          }
-          SmootherPrototype::IsSetup(isSetup);
+          // TODO              thisSmootherList[i]->CopyParameters(srcSmootherList[i]);
+          isSetup = isSetup && thisSmootherList[i]->IsSetup();
         }
-
-      } else {
-        // No reuse: copy srcSmootherList.
-        smootherList_ = Teuchos::null;
-        smootherList_ = SmootherListDeepCopy(srcSmootherList);
-        SmootherPrototype::IsSetup(false);
+        SmootherPrototype::IsSetup(isSetup);
       }
 
+    } else {
+      // No reuse: copy srcSmootherList.
+      smootherList_ = Teuchos::null;
+      smootherList_ = SmootherListDeepCopy(srcSmootherList);
+      SmootherPrototype::IsSetup(false);
     }
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Copy () const
-  {
-    return rcp(new MergedSmoother(*this));
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  ArrayRCP<RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >
-  MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  SmootherListDeepCopy (const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList)
-  {
-    ArrayRCP<RCP<SmootherPrototype> > newSmootherList(srcSmootherList.size());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Copy() const {
+  return rcp(new MergedSmoother(*this));
+}
 
-    for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++)
-      newSmootherList[i] = srcSmootherList[i]->Copy();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ArrayRCP<RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >
+MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SmootherListDeepCopy(const ArrayRCP<const RCP<SmootherPrototype> >& srcSmootherList) {
+  ArrayRCP<RCP<SmootherPrototype> > newSmootherList(srcSmootherList.size());
 
-    return newSmootherList;
-  }
+  for (typename ArrayRCP<RCP<SmootherPrototype> >::size_type i = 0; i < srcSmootherList.size(); i++)
+    newSmootherList[i] = srcSmootherList[i]->Copy();
 
+  return newSmootherList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t MergedSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>::
-  getNodeSmootherComplexity() const
-  {
-    // FIXME: This is a placeholder
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t MergedSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    getNodeSmootherComplexity() const {
+  // FIXME: This is a placeholder
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_MERGEDSMOOTHER_DEF_HPP
+#endif  // MUELU_MERGEDSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_decl.hpp
index ba1f63d85e37..d0787b0577f9 100644
--- a/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_decl.hpp
@@ -69,108 +69,106 @@
 
 namespace MueLu {
 
-  /*!
-    @class PermutingSmoother
-    @ingroup MueLuSmootherClasses
-    @brief This class first calculates row- and column permutation operators and applies a smoother to the permuted linear system.
-
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class PermutingSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
+/*!
+  @class PermutingSmoother
+  @ingroup MueLuSmootherClasses
+  @brief This class first calculates row- and column permutation operators and applies a smoother to the permuted linear system.
+
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class PermutingSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PERMUTINGSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! @name Constructors / destructors
-    //@{
+  //! @brief Constructor
+  //!    @param[in] mapName   Name of map object in level class, which rows/cols can be permuted
+  //!    @param[in] mapFact   generating factory of map with name mapName
+  //!    @param[in] type      string that contains type of smoother (e.g. "RELAXATION" or "ILU")
+  //!    @param[in] paramList parameter list with parameters for smoother (default: empty)
+  //!    @param[in] overlap   LocalOrdinal with overlap inforation (default: 0)
+  //!    @param[in] permFact  factory, generating permutation and scaling matrices (default: Teuchos::null -> use internal PermutationFactory instance)
+  PermutingSmoother(std::string const& mapName, const RCP<const FactoryBase>& mapFact, std::string const& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), LO const& overlap = 0, RCP<FactoryBase> permFact = Teuchos::null);
 
-    //! @brief Constructor
-    //!    @param[in] mapName   Name of map object in level class, which rows/cols can be permuted
-    //!    @param[in] mapFact   generating factory of map with name mapName
-    //!    @param[in] type      string that contains type of smoother (e.g. "RELAXATION" or "ILU")
-    //!    @param[in] paramList parameter list with parameters for smoother (default: empty)
-    //!    @param[in] overlap   LocalOrdinal with overlap inforation (default: 0)
-    //!    @param[in] permFact  factory, generating permutation and scaling matrices (default: Teuchos::null -> use internal PermutationFactory instance)
-    PermutingSmoother(std::string const & mapName, const RCP<const FactoryBase> & mapFact, std::string const & type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), LO const & overlap = 0, RCP<FactoryBase> permFact = Teuchos::null);
+  //! Destructor
+  virtual ~PermutingSmoother();
+  //@}
 
-    //! Destructor
-    virtual ~PermutingSmoother();
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  //@}
 
-    //@}
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @brief Set up the direct solver.
+  void Setup(Level& currentLevel);
 
-    //! @brief Set up the direct solver.
-    void Setup(Level &currentLevel);
-
-    /*! @brief Apply the direct solver.
-    Solves the linear system <tt>AX=B</tt> using the constructed solver.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
+  /*! @brief Apply the direct solver.
+  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    RCP<SmootherPrototype> Copy() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {return s_->getNodeSmootherComplexity();}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const { return s_->getNodeSmootherComplexity(); }
 
-    //@}
+  //@}
 
-  private:
-    //! ifpack1/2-specific key phrase that denote smoother type
-    std::string type_;
+ private:
+  //! ifpack1/2-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! overlap when using the smoother in additive Schwarz mode
-    LO overlap_;
+  //! overlap when using the smoother in additive Schwarz mode
+  LO overlap_;
 
-    //! Permutation Factory
-    RCP<FactoryBase> permFact_;
+  //! Permutation Factory
+  RCP<FactoryBase> permFact_;
 
-    //! permQT matrix object
-    RCP<Matrix> permQT_;
+  //! permQT matrix object
+  RCP<Matrix> permQT_;
 
-    //! permP matrix object
-    RCP<Matrix> permP_;
+  //! permP matrix object
+  RCP<Matrix> permP_;
 
-    //! scaling matrix object
-    Teuchos::RCP<Matrix> diagScalingOp_;
+  //! scaling matrix object
+  Teuchos::RCP<Matrix> diagScalingOp_;
 
-    //
-    // Underlying Smoother
-    //
+  //
+  // Underlying Smoother
+  //
 
-    //! Smoother
-    RCP<SmootherPrototype> s_; // TrilinosSmoother object
+  //! Smoother
+  RCP<SmootherPrototype> s_;  // TrilinosSmoother object
 
-  }; // class PermutingSmoother
+};  // class PermutingSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PERMUTINGSMOOTHER_SHORT
 #endif /* MUELU_PERMUTINGSMOOTHER_DECL_HPP */
diff --git a/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_def.hpp
index 241d550d5b0b..01163376f77d 100644
--- a/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_PermutingSmoother_def.hpp
@@ -70,107 +70,109 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PermutingSmoother(const std::string& mapName, const RCP<const FactoryBase>& mapFact, const std::string& type, const Teuchos::ParameterList& paramList, const LO& overlap, RCP<FactoryBase> permFact)
-  : type_(type), overlap_(overlap), permQT_(Teuchos::null), permP_(Teuchos::null), diagScalingOp_(Teuchos::null)
-  {
-    this->SetParameterList(paramList);
-
-    permFact_ = permFact;
-    if (permFact_ == Teuchos::null) {
-      RCP<PermutationFactory> newPermFact = Teuchos::rcp(new PermutationFactory());
-      newPermFact->SetParameter("PermutationRowMapName",    Teuchos::ParameterEntry(mapName));
-      newPermFact->SetFactory  ("PermutationRowMapFactory", mapFact);
-      permFact_ = newPermFact;
-    }
-
-    // create internal smoother
-    if (type_ == "ILU") {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PermutingSmoother(const std::string& mapName, const RCP<const FactoryBase>& mapFact, const std::string& type, const Teuchos::ParameterList& paramList, const LO& overlap, RCP<FactoryBase> permFact)
+  : type_(type)
+  , overlap_(overlap)
+  , permQT_(Teuchos::null)
+  , permP_(Teuchos::null)
+  , diagScalingOp_(Teuchos::null) {
+  this->SetParameterList(paramList);
+
+  permFact_ = permFact;
+  if (permFact_ == Teuchos::null) {
+    RCP<PermutationFactory> newPermFact = Teuchos::rcp(new PermutationFactory());
+    newPermFact->SetParameter("PermutationRowMapName", Teuchos::ParameterEntry(mapName));
+    newPermFact->SetFactory("PermutationRowMapFactory", mapFact);
+    permFact_ = newPermFact;
+  }
+
+  // create internal smoother
+  if (type_ == "ILU") {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
-      s_ = MueLu::GetIfpackSmoother<Scalar,LocalOrdinal,GlobalOrdinal,Node>(type_, this->GetParameterList(), overlap_);
+    s_ = MueLu::GetIfpackSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>(type_, this->GetParameterList(), overlap_);
 #else
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PermutingSmoother requires Epetra and Ifpack.");
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PermutingSmoother requires Epetra and Ifpack.");
 #endif
-    } else {
-      s_ = Teuchos::rcp(new TrilinosSmoother(type_, this->GetParameterList(), overlap_));
-    }
-    TEUCHOS_TEST_FOR_EXCEPTION(s_ == Teuchos::null, Exceptions::RuntimeError, "");
-
-    // Use permuted matrix A
-    s_->SetFactory("A", permFact_);
+  } else {
+    s_ = Teuchos::rcp(new TrilinosSmoother(type_, this->GetParameterList(), overlap_));
   }
+  TEUCHOS_TEST_FOR_EXCEPTION(s_ == Teuchos::null, Exceptions::RuntimeError, "");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~PermutingSmoother() { }
+  // Use permuted matrix A
+  s_->SetFactory("A", permFact_);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    currentLevel.DeclareInput("permP",       permFact_.get());
-    currentLevel.DeclareInput("permQT",      permFact_.get());
-    currentLevel.DeclareInput("permScaling", permFact_.get());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~PermutingSmoother() {}
 
-    s_->DeclareInput(currentLevel);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  currentLevel.DeclareInput("permP", permFact_.get());
+  currentLevel.DeclareInput("permQT", permFact_.get());
+  currentLevel.DeclareInput("permScaling", permFact_.get());
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor monitor(*this, "Permuting Smoother", currentLevel);
+  s_->DeclareInput(currentLevel);
+}
 
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::PermutingSmoother::Setup(): Setup() has already been called" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor monitor(*this, "Permuting Smoother", currentLevel);
 
-    // extract information from level class
-    permP_         = currentLevel.Get< RCP<Matrix> > ("permP",       permFact_.get());
-    permQT_        = currentLevel.Get< RCP<Matrix> > ("permQT",      permFact_.get());
-    diagScalingOp_ = currentLevel.Get< RCP<Matrix> > ("permScaling", permFact_.get());
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::PermutingSmoother::Setup(): Setup() has already been called" << std::endl;
 
-    s_->Setup(currentLevel);
+  // extract information from level class
+  permP_         = currentLevel.Get<RCP<Matrix> >("permP", permFact_.get());
+  permQT_        = currentLevel.Get<RCP<Matrix> >("permQT", permFact_.get());
+  diagScalingOp_ = currentLevel.Get<RCP<Matrix> >("permScaling", permFact_.get());
 
-    SmootherPrototype::IsSetup(true);
-  }
+  s_->Setup(currentLevel);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::PermutingSmoother::Apply(): Setup() has not been called");
+  SmootherPrototype::IsSetup(true);
+}
 
-    typedef Teuchos::ScalarTraits<Scalar> STS;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::PermutingSmoother::Apply(): Setup() has not been called");
 
-    Teuchos::RCP<MultiVector> Xtemp = MultiVectorFactory::Build(X.getMap(), 1, true);
-    Xtemp->update(STS::one(), X, STS::zero());
+  typedef Teuchos::ScalarTraits<Scalar> STS;
 
-    // TODO: unify scaling and left permutation operator
-    Teuchos::RCP<MultiVector> Btemp  = MultiVectorFactory::Build(B.getMap(), 1, true);
-    Teuchos::RCP<MultiVector> Btemp2 = MultiVectorFactory::Build(B.getMap(), 1, true);
-    permP_->apply(B, *Btemp, Teuchos::NO_TRANS);                // apply permutation operator to rhs
-    diagScalingOp_->apply(*Btemp, *Btemp2, Teuchos::NO_TRANS);  // apply scaling operator to rhs
+  Teuchos::RCP<MultiVector> Xtemp = MultiVectorFactory::Build(X.getMap(), 1, true);
+  Xtemp->update(STS::one(), X, STS::zero());
 
-    // apply smoother to permuted linear system
-    s_->Apply(*Xtemp, *Btemp2, InitialGuessIsZero);
+  // TODO: unify scaling and left permutation operator
+  Teuchos::RCP<MultiVector> Btemp  = MultiVectorFactory::Build(B.getMap(), 1, true);
+  Teuchos::RCP<MultiVector> Btemp2 = MultiVectorFactory::Build(B.getMap(), 1, true);
+  permP_->apply(B, *Btemp, Teuchos::NO_TRANS);                // apply permutation operator to rhs
+  diagScalingOp_->apply(*Btemp, *Btemp2, Teuchos::NO_TRANS);  // apply scaling operator to rhs
 
-    // retransform smooth solution
-    permQT_->apply(*Xtemp, X, Teuchos::NO_TRANS);
-  }
+  // apply smoother to permuted linear system
+  s_->Apply(*Xtemp, *Btemp2, InitialGuessIsZero);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp(new PermutingSmoother(*this));
-  }
+  // retransform smooth solution
+  permQT_->apply(*Xtemp, X, Teuchos::NO_TRANS);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    return out.str();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new PermutingSmoother(*this));
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-    out0 << ""; // avoid warning
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  return out.str();
+}
 
-} // namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PermutingSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+  out0 << "";  // avoid warning
+}
 
+}  // namespace MueLu
 
 #endif /* MUELU_PERMUTINGSMOOTHER_DEF_HPP */
diff --git a/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_decl.hpp
index e7096fc390ed..2eaf0950f77c 100644
--- a/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_decl.hpp
@@ -56,90 +56,87 @@
 
 namespace MueLu {
 
-  /*!
-    @class ProjectorSmoother
-    @ingroup MueLuSmootherClasses
-    @brief This class enables the elimination of the nullspace component of the solution through the use of projection
-
-    The solution of the coarsest level system may have a significant nullspace component. We can try to eliminate it through the use of nullspaces.
-    Due to our construction algorithms, we always have some nullspace vectors on the coarsest level. However, we do not know if they are true
-    nullspace vectors or are only approximations to them.
-
-    We will use the Rayleigh quotiont to separate the true nullspace vectors. If Rayleigh quotient is less than specified epsilon, we consider the
-    vector to be a true nullspace vector.
-
-    After we separate true nullspace vectors, we project our solution to the space orthogonal to them. To do that in a computationally efficient manner,
-    in the Setup stage we orthonormalize the selected nullspace components.
-  */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class ProjectorSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>
-  {
+/*!
+  @class ProjectorSmoother
+  @ingroup MueLuSmootherClasses
+  @brief This class enables the elimination of the nullspace component of the solution through the use of projection
+
+  The solution of the coarsest level system may have a significant nullspace component. We can try to eliminate it through the use of nullspaces.
+  Due to our construction algorithms, we always have some nullspace vectors on the coarsest level. However, we do not know if they are true
+  nullspace vectors or are only approximations to them.
+
+  We will use the Rayleigh quotiont to separate the true nullspace vectors. If Rayleigh quotient is less than specified epsilon, we consider the
+  vector to be a true nullspace vector.
+
+  After we separate true nullspace vectors, we project our solution to the space orthogonal to them. To do that in a computationally efficient manner,
+  in the Setup stage we orthonormalize the selected nullspace components.
+*/
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class ProjectorSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_PROJECTORSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! @brief Constructor
-    ProjectorSmoother(RCP<SmootherPrototype> coarseSolver);
+  //! @brief Constructor
+  ProjectorSmoother(RCP<SmootherPrototype> coarseSolver);
 
-    //! Destructor
-    virtual ~ProjectorSmoother();
-    //@}
+  //! Destructor
+  virtual ~ProjectorSmoother();
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! @brief Set up the direct solver.
-    void Setup(Level &currentLevel);
+  //! @brief Set up the direct solver.
+  void Setup(Level &currentLevel);
 
-    /*! @brief Apply the direct solver.
-    Solves the linear system <tt>AX=B</tt> using the constructed solver.
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero This option has no effect.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
-    //@}
-
-    RCP<SmootherPrototype> Copy() const;
+  /*! @brief Apply the direct solver.
+  Solves the linear system <tt>AX=B</tt> using the constructed solver.
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero This option has no effect.
+  */
+  void Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero = false) const;
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  RCP<SmootherPrototype> Copy() const;
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {return coarseSolver_->getNodeSmootherComplexity();}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
 
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const { return coarseSolver_->getNodeSmootherComplexity(); }
 
-    //@}
+  //@}
 
-  private:
-    RCP<MultiVector>             Borth_;
-    RCP<SmootherPrototype>       coarseSolver_;
+ private:
+  RCP<MultiVector> Borth_;
+  RCP<SmootherPrototype> coarseSolver_;
 
-  }; // class ProjectorSmoother
+};  // class ProjectorSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PROJECTORSMOOTHER_SHORT
-#endif // MUELU_PROJECTORSMOOTHER_DECL_HPP
+#endif  // MUELU_PROJECTORSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_def.hpp
index 072f2245377f..ccb4effa5d97 100644
--- a/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_ProjectorSmoother_def.hpp
@@ -58,123 +58,122 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ProjectorSmoother(RCP<SmootherPrototype> coarseSolver)
-    : coarseSolver_(coarseSolver)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ProjectorSmoother(RCP<SmootherPrototype> coarseSolver)
+  : coarseSolver_(coarseSolver) {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~ProjectorSmoother() { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~ProjectorSmoother() {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Factory::Input(currentLevel, "A");
-    Factory::Input(currentLevel, "Nullspace");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Factory::Input(currentLevel, "A");
+  Factory::Input(currentLevel, "Nullspace");
 
-    coarseSolver_->DeclareInput(currentLevel);
-  }
+  coarseSolver_->DeclareInput(currentLevel);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
-    FactoryMonitor monitor(*this, "Projector Smoother", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level &currentLevel) {
+  FactoryMonitor monitor(*this, "Projector Smoother", currentLevel);
 
-    coarseSolver_->Setup(currentLevel);
+  coarseSolver_->Setup(currentLevel);
 
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::ProjectorSmoother::Setup(): Setup() has already been called" << std::endl;
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::ProjectorSmoother::Setup(): Setup() has already been called" << std::endl;
 
-    RCP<Matrix>      A = Factory::Get< RCP<Matrix> >     (currentLevel, "A");
-    RCP<MultiVector> B = Factory::Get< RCP<MultiVector> >(currentLevel, "Nullspace");
+  RCP<Matrix> A      = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+  RCP<MultiVector> B = Factory::Get<RCP<MultiVector> >(currentLevel, "Nullspace");
 
-    int m = B->getNumVectors();
+  int m = B->getNumVectors();
 
-    // Find which vectors we want to keep
-    Array<Scalar> br(m), bb(m);
-    RCP<MultiVector> R = MultiVectorFactory::Build(B->getMap(), m);
-    A->apply(*B, *R);
-    B->  dot(*R, br);
-    B->  dot(*B, bb);
+  // Find which vectors we want to keep
+  Array<Scalar> br(m), bb(m);
+  RCP<MultiVector> R = MultiVectorFactory::Build(B->getMap(), m);
+  A->apply(*B, *R);
+  B->dot(*R, br);
+  B->dot(*B, bb);
 
-    Array<size_t> selectedIndices;
-    for (int i = 0; i < m; i++) {
-      Scalar rayleigh = br[i] / bb[i];
+  Array<size_t> selectedIndices;
+  for (int i = 0; i < m; i++) {
+    Scalar rayleigh = br[i] / bb[i];
 
-      if (Teuchos::ScalarTraits<Scalar>::magnitude(rayleigh) < 1e-12)
-        selectedIndices.push_back(i);
-    }
-    this->GetOStream(Runtime0) << "Coarse level orth indices: " << selectedIndices << std::endl;
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(rayleigh) < 1e-12)
+      selectedIndices.push_back(i);
+  }
+  this->GetOStream(Runtime0) << "Coarse level orth indices: " << selectedIndices << std::endl;
 
 #if defined(HAVE_XPETRA_TPETRA)
 #ifdef HAVE_MUELU_TPETRA_INST_INT_INT
-    // Orthonormalize
-    RCP<const Tpetra::MultiVector<SC,LO,GO,NO> > B_ = Utilities::MV2TpetraMV(B);
-    // TAW: Oct 16 2015: subCopy is not part of Xpetra. One should either add it to Xpetra (with an emulator for Epetra)
-    //                   or replace this call by a local loop. I'm not motivated to do this now...
-    RCP<Tpetra::MultiVector<SC,LO,GO,NO> > Borth = B_->subCopy(selectedIndices);               // copy
-    for (int i = 0; i < selectedIndices.size(); i++) {
-      RCP<Tpetra::Vector<SC,LO,GO,NO> > Bi = Borth->getVectorNonConst(i);
-
-      Scalar dot;
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType norm;
-      for (int k = 0; k < i; k++) {                                     // orthogonalize
-        RCP<const Tpetra::Vector<SC,LO,GO,NO> > Bk = Borth->getVector(k);
-
-        dot = Bi->dot(*Bk);
-        Bi->update(-dot, *Bk, Teuchos::ScalarTraits<Scalar>::one());
-      }
-
-      norm = Bi->norm2();
-      Bi->scale(Teuchos::ScalarTraits<Scalar>::one()/norm);           // normalize
+  // Orthonormalize
+  RCP<const Tpetra::MultiVector<SC, LO, GO, NO> > B_ = Utilities::MV2TpetraMV(B);
+  // TAW: Oct 16 2015: subCopy is not part of Xpetra. One should either add it to Xpetra (with an emulator for Epetra)
+  //                   or replace this call by a local loop. I'm not motivated to do this now...
+  RCP<Tpetra::MultiVector<SC, LO, GO, NO> > Borth = B_->subCopy(selectedIndices);  // copy
+  for (int i = 0; i < selectedIndices.size(); i++) {
+    RCP<Tpetra::Vector<SC, LO, GO, NO> > Bi = Borth->getVectorNonConst(i);
+
+    Scalar dot;
+    typename Teuchos::ScalarTraits<Scalar>::magnitudeType norm;
+    for (int k = 0; k < i; k++) {  // orthogonalize
+      RCP<const Tpetra::Vector<SC, LO, GO, NO> > Bk = Borth->getVector(k);
+
+      dot = Bi->dot(*Bk);
+      Bi->update(-dot, *Bk, Teuchos::ScalarTraits<Scalar>::one());
     }
 
-    Borth_ = rcp(static_cast<MultiVector*>(new TpetraMultiVector(Borth)));
+    norm = Bi->norm2();
+    Bi->scale(Teuchos::ScalarTraits<Scalar>::one() / norm);  // normalize
+  }
+
+  Borth_ = rcp(static_cast<MultiVector *>(new TpetraMultiVector(Borth)));
 #else
-    TEUCHOS_TEST_FOR_EXCEPTION(true,Exceptions::RuntimeError,"Tpetra with GO=int not available. The code in ProjectorSmoother should be rewritten!");
+  TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Tpetra with GO=int not available. The code in ProjectorSmoother should be rewritten!");
 #endif
 #endif
 
-    SmootherPrototype::IsSetup(true);
-  }
+  SmootherPrototype::IsSetup(true);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    coarseSolver_->Apply(X, B, InitialGuessIsZero);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector &X, const MultiVector &B, bool InitialGuessIsZero) const {
+  coarseSolver_->Apply(X, B, InitialGuessIsZero);
 
-    int m = Borth_->getNumVectors();
-    int n = X.getNumVectors();
+  int m = Borth_->getNumVectors();
+  int n = X.getNumVectors();
 
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > X_ = Teuchos::rcpFromRef(X);
-    for (int i = 0; i < n; i++) {
-      RCP<Xpetra::Vector<SC,LO,GO,NO> > Xi = X_->getVectorNonConst(i);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > X_ = Teuchos::rcpFromRef(X);
+  for (int i = 0; i < n; i++) {
+    RCP<Xpetra::Vector<SC, LO, GO, NO> > Xi = X_->getVectorNonConst(i);
 
-      Array<Scalar> dot(1);
-      for (int k = 0; k < m; k++) {                                     // orthogonalize
-        RCP<const Xpetra::Vector<SC,LO,GO,NO> > Bk = Borth_->getVector(k);
+    Array<Scalar> dot(1);
+    for (int k = 0; k < m; k++) {  // orthogonalize
+      RCP<const Xpetra::Vector<SC, LO, GO, NO> > Bk = Borth_->getVector(k);
 
-        Xi->dot(*Bk, dot());
-        Xi->update(-dot[0], *Bk, Teuchos::ScalarTraits<Scalar>::one());
-      }
+      Xi->dot(*Bk, dot());
+      Xi->update(-dot[0], *Bk, Teuchos::ScalarTraits<Scalar>::one());
     }
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    return rcp( new ProjectorSmoother(*this) );
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  return rcp(new ProjectorSmoother(*this));
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherPrototype::description();
-    return out.str();
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherPrototype::description();
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-    out0 << "";
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ProjectorSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+  out0 << "";
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_PROJECTORSMOOTHER_DEF_HPP
+#endif  // MUELU_PROJECTORSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_decl.hpp
index 07556d669390..5d12e50f5789 100644
--- a/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_decl.hpp
@@ -58,110 +58,107 @@
 #include "MueLu_SmootherPrototype.hpp"
 #include "MueLu_RefMaxwell.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class RefMaxwellSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Operator smoothers.
+/*!
+  @class RefMaxwellSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Operator smoothers.
 
-    */
+  */
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class RefMaxwellSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class RefMaxwellSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_REFMAXWELLSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-    /*! @brief Constructor
-
-    @param type smoother type
-    @param list options for the particular smoother (e.g., fill factor or damping parameter)
+ public:
+  //! @name Constructors / destructors
+  //@{
+  /*! @brief Constructor
 
-    */
+  @param type smoother type
+  @param list options for the particular smoother (e.g., fill factor or damping parameter)
 
-    RefMaxwellSmoother(const std::string type, const Teuchos::ParameterList& paramList);
+  */
 
-    //! Destructor
-    virtual ~RefMaxwellSmoother() = default;
+  RefMaxwellSmoother(const std::string type, const Teuchos::ParameterList& paramList);
 
-    //@}
+  //! Destructor
+  virtual ~RefMaxwellSmoother() = default;
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  //@}
 
-    //! Input
-    //@{
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    void DeclareInput(Level &currentLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Computational methods.
-    //@{
+  //@}
 
-    /*! @brief Set up the smoother.
+  //! @name Computational methods.
+  //@{
 
-    This creates the underlying Operator smoother object, copies any parameter list options
-    supplied to the constructor to the Operator object, and computes the preconditioner.
-    */
-    void Setup(Level &currentLevel);
+  /*! @brief Set up the smoother.
 
-    /*! @brief Apply the preconditioner.
+  This creates the underlying Operator smoother object, copies any parameter list options
+  supplied to the constructor to the Operator object, and computes the preconditioner.
+  */
+  void Setup(Level& currentLevel);
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  /*! @brief Apply the preconditioner.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Operator implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    //@}
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Operator implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //! @name Utilities
-    //@{
+  //@}
 
-    RCP<SmootherPrototype> Copy() const;
+  //! @name Utilities
+  //@{
 
-    //@}
+  RCP<SmootherPrototype> Copy() const;
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //@}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //@}
 
-  private:
-    void SetupRefMaxwell(Level& currentLevel);
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
+ private:
+  void SetupRefMaxwell(Level& currentLevel);
 
-    std::string                         type_;
-    std::string                         cachedDescription_;
+ private:
+  std::string type_;
+  std::string cachedDescription_;
 
-    //! matrix, used in apply if solving residual equation
-    RCP<Operator> op_;
+  //! matrix, used in apply if solving residual equation
+  RCP<Operator> op_;
 
-  }; // class RefMaxwellSmoother
+};  // class RefMaxwellSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REFMAXWELLSMOOTHER_SHORT
-#endif // MUELU_REFMAXWELLSMOOTHER_DECL_HPP
+#endif  // MUELU_REFMAXWELLSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_def.hpp
index 525ec9adc9b0..5410392cb1d6 100644
--- a/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_RefMaxwellSmoother_def.hpp
@@ -58,160 +58,156 @@
 #include "MueLu_Level.hpp"
 #include "MueLu_Monitor.hpp"
 
-
-
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RefMaxwellSmoother(const std::string type, const Teuchos::ParameterList& paramList)
-  : type_(type)
-  {
-    const bool solverSupported = (type_ == "RefMaxwell");
-    this->declareConstructionOutcome(!solverSupported, "RefMaxwellSmoother does not provide the smoother '" + type_ + "'.");
-    if (solverSupported)
-      SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
-    Factory::SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    const ParameterList& pL = this->GetParameterList();
-
-    int spaceNumber = 1;
-    if (pL.isType<int>("refmaxwell: space number"))
-      spaceNumber = pL.get<int>("refmaxwell: space number");
-    this->Input(currentLevel, "A");
-    this->Input(currentLevel, "Dk_1");
-    if (spaceNumber > 1)
-      this->Input(currentLevel, "Dk_2");
-    this->Input(currentLevel, "D0");
-    this->Input(currentLevel, "Mk_one");
-    if (spaceNumber > 1)
-      this->Input(currentLevel, "Mk_1_one");
-    this->Input(currentLevel, "M1_beta");
-    if (spaceNumber > 1)
-      this->Input(currentLevel, "M1_alpha");
-    this->Input(currentLevel, "invMk_1_invBeta");
-    if (spaceNumber > 1)
-      this->Input(currentLevel, "invMk_2_invAlpha");
-    this->Input(currentLevel, "Coordinates");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
-
-    SetupRefMaxwell(currentLevel);
-    SmootherPrototype::IsSetup(true);
-    this->GetOStream(Statistics1) << description() << std::endl;
-  }
-
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupRefMaxwell(Level& currentLevel) {
-
-    using coordinateType = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<coordinateType,LO,GO,NO>;
-
-    ParameterList params           = this->GetParameterList();
-    int spaceNumber = 1;
-    if (params.isType<int>("refmaxwell: space number"))
-      spaceNumber = params.get<int>("refmaxwell: space number");
-
-    RCP<Matrix> SM, Dk_1, Dk_2, D0, Mk_one, Mk_1_one, M1_beta, M1_alpha, invMk_1_invBeta, invMk_2_invAlpha;
-
-    SM               = currentLevel.Get<RCP<Matrix> >("A");
-    Dk_1             = currentLevel.Get<RCP<Matrix> >("Dk_1");
-    if (spaceNumber > 1)
-      Dk_2             = currentLevel.Get<RCP<Matrix> >("Dk_2");
-    D0               = currentLevel.Get<RCP<Matrix> >("D0");
-    Mk_one           = currentLevel.Get<RCP<Matrix> >("Mk_one");
-    if (spaceNumber > 1)
-      Mk_1_one         = currentLevel.Get<RCP<Matrix> >("Mk_1_one");
-    M1_beta          = currentLevel.Get<RCP<Matrix> >("M1_beta");
-    if (spaceNumber > 1)
-      M1_alpha         = currentLevel.Get<RCP<Matrix> >("M1_alpha");
-    invMk_1_invBeta  = currentLevel.Get<RCP<Matrix> >("invMk_1_invBeta");
-    if (spaceNumber > 1)
-      invMk_2_invAlpha = currentLevel.Get<RCP<Matrix> >("invMk_2_invAlpha");
-    auto coords           = currentLevel.Get<RCP<RealValuedMultiVector> >("Coordinates");
-
-    op_ = rcp(new MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>(SM,
-                                                                               Dk_1, Dk_2, D0,
-                                                                               M1_beta, M1_alpha,
-                                                                               Mk_one, Mk_1_one,
-                                                                               invMk_1_invBeta, invMk_2_invAlpha,
-                                                                               Teuchos::null, Teuchos::null, coords,
-                                                                               params));
-    std::ostringstream oss;
-    op_->describe(*Teuchos::fancyOStream(Teuchos::rcpFromRef(oss)));
-    cachedDescription_ = oss.str();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RefMaxwellSmoother(const std::string type, const Teuchos::ParameterList& paramList)
+  : type_(type) {
+  const bool solverSupported = (type_ == "RefMaxwell");
+  this->declareConstructionOutcome(!solverSupported, "RefMaxwellSmoother does not provide the smoother '" + type_ + "'.");
+  if (solverSupported)
+    SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = this->GetParameterList();
+
+  int spaceNumber = 1;
+  if (pL.isType<int>("refmaxwell: space number"))
+    spaceNumber = pL.get<int>("refmaxwell: space number");
+  this->Input(currentLevel, "A");
+  this->Input(currentLevel, "Dk_1");
+  if (spaceNumber > 1)
+    this->Input(currentLevel, "Dk_2");
+  this->Input(currentLevel, "D0");
+  this->Input(currentLevel, "Mk_one");
+  if (spaceNumber > 1)
+    this->Input(currentLevel, "Mk_1_one");
+  this->Input(currentLevel, "M1_beta");
+  if (spaceNumber > 1)
+    this->Input(currentLevel, "M1_alpha");
+  this->Input(currentLevel, "invMk_1_invBeta");
+  if (spaceNumber > 1)
+    this->Input(currentLevel, "invMk_2_invAlpha");
+  this->Input(currentLevel, "Coordinates");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+
+  SetupRefMaxwell(currentLevel);
+  SmootherPrototype::IsSetup(true);
+  this->GetOStream(Statistics1) << description() << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetupRefMaxwell(Level& currentLevel) {
+  using coordinateType        = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<coordinateType, LO, GO, NO>;
+
+  ParameterList params = this->GetParameterList();
+  int spaceNumber      = 1;
+  if (params.isType<int>("refmaxwell: space number"))
+    spaceNumber = params.get<int>("refmaxwell: space number");
+
+  RCP<Matrix> SM, Dk_1, Dk_2, D0, Mk_one, Mk_1_one, M1_beta, M1_alpha, invMk_1_invBeta, invMk_2_invAlpha;
+
+  SM   = currentLevel.Get<RCP<Matrix> >("A");
+  Dk_1 = currentLevel.Get<RCP<Matrix> >("Dk_1");
+  if (spaceNumber > 1)
+    Dk_2 = currentLevel.Get<RCP<Matrix> >("Dk_2");
+  D0     = currentLevel.Get<RCP<Matrix> >("D0");
+  Mk_one = currentLevel.Get<RCP<Matrix> >("Mk_one");
+  if (spaceNumber > 1)
+    Mk_1_one = currentLevel.Get<RCP<Matrix> >("Mk_1_one");
+  M1_beta = currentLevel.Get<RCP<Matrix> >("M1_beta");
+  if (spaceNumber > 1)
+    M1_alpha = currentLevel.Get<RCP<Matrix> >("M1_alpha");
+  invMk_1_invBeta = currentLevel.Get<RCP<Matrix> >("invMk_1_invBeta");
+  if (spaceNumber > 1)
+    invMk_2_invAlpha = currentLevel.Get<RCP<Matrix> >("invMk_2_invAlpha");
+  auto coords = currentLevel.Get<RCP<RealValuedMultiVector> >("Coordinates");
+
+  op_ = rcp(new MueLu::RefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>(SM,
+                                                                             Dk_1, Dk_2, D0,
+                                                                             M1_beta, M1_alpha,
+                                                                             Mk_one, Mk_1_one,
+                                                                             invMk_1_invBeta, invMk_2_invAlpha,
+                                                                             Teuchos::null, Teuchos::null, coords,
+                                                                             params));
+  std::ostringstream oss;
+  op_->describe(*Teuchos::fancyOStream(Teuchos::rcpFromRef(oss)));
+  cachedDescription_ = oss.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::RefMaxwellSmoother::Apply(): Setup() has not been called");
+
+  if (InitialGuessIsZero) {
+    X.putScalar(0.0);
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::RefMaxwellSmoother::Apply(): Setup() has not been called");
-
-    if (InitialGuessIsZero) {
-      X.putScalar(0.0);
-    }
-    op_->apply(B, X);
+  op_->apply(B, X);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<RefMaxwellSmoother> smoother = rcp(new RefMaxwellSmoother(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return smoother;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (SmootherPrototype::IsSetup()) {
+    out << op_->description();
+    out << std::endl
+        << std::endl;
+    // out << cachedDescription_;
+  } else {
+    out << "RefMaxwell";
   }
+  return out.str();
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> > RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<RefMaxwellSmoother> smoother = rcp(new RefMaxwellSmoother(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return smoother;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (SmootherPrototype::IsSetup()) {
-      out << op_->description();
-      out << std::endl << std::endl;
-      // out << cachedDescription_;
-    } else {
-      out << "RefMaxwell";
-    }
-    return out.str();
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
+  if (verbLevel & External) {
+    if (op_ != Teuchos::null) {
       Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-    }
-
-    if (verbLevel & External) {
-      if (op_ != Teuchos::null) {
-        Teuchos::OSTab tab2(out);
-        out << *op_ << std::endl << std::endl;
-      }
+      out << *op_ << std::endl
+          << std::endl;
     }
-
-    // if (verbLevel & Debug) {
-    //   out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-    //        << "-" << std::endl
-    //        << "RCP<solver_>: " << tSolver_ << std::endl;
-    // }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    return Teuchos::OrdinalTraits<size_t>::invalid();
-  }
+  // if (verbLevel & Debug) {
+  //   out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+  //        << "-" << std::endl
+  //        << "RCP<solver_>: " << tSolver_ << std::endl;
+  // }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t RefMaxwellSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_REFMAXWELLSMOOTHER_DEF_HPP
+#endif  // MUELU_REFMAXWELLSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherBase.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherBase.hpp
index f2bc6c9a66b2..cb81323f15a7 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherBase.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherBase.hpp
@@ -52,69 +52,68 @@
 #include "MueLu_BaseClass.hpp"
 
 namespace MueLu {
-  /*!
-    @class SmootherBase
-    @ingroup MueLuSmootherClasses
-    @brief Base class for smoothers
-
-    This has the signature for the required Apply function and contains data that is generic across all
-    smoothers.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SmootherBase : public virtual BaseClass {
-  public:
-    typedef Scalar scalar_type;
-    typedef LocalOrdinal local_ordinal_type;
-    typedef GlobalOrdinal global_ordinal_type;
-    typedef Node node_type;
-
-  private:
-
+/*!
+  @class SmootherBase
+  @ingroup MueLuSmootherClasses
+  @brief Base class for smoothers
+
+  This has the signature for the required Apply function and contains data that is generic across all
+  smoothers.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SmootherBase : public virtual BaseClass {
+ public:
+  typedef Scalar scalar_type;
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef Node node_type;
+
+ private:
 #undef MUELU_SMOOTHERBASE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //@{ Constructors/Destructors.
-    SmootherBase() {}
+ public:
+  //@{ Constructors/Destructors.
+  SmootherBase() {}
 
-    virtual ~SmootherBase() {}
-    //@}
+  virtual ~SmootherBase() {}
+  //@}
 
-    //! @name Apply methods.
-    //@{
+  //! @name Apply methods.
+  //@{
 
-    //! Apply smoother.
-    virtual void Apply(MultiVector& x, const MultiVector& rhs, bool InitialGuessIsZero = false) const = 0;
+  //! Apply smoother.
+  virtual void Apply(MultiVector& x, const MultiVector& rhs, bool InitialGuessIsZero = false) const = 0;
 
-    //! Compute a rough estimate of the cost to apply this smoother on this MPI rank.  Return Teuchos::OrdinalTraits<size_t>::invalid() if such an estimate cannot be computed.
-    virtual size_t getNodeSmootherComplexity() const= 0;
+  //! Compute a rough estimate of the cost to apply this smoother on this MPI rank.  Return Teuchos::OrdinalTraits<size_t>::invalid() if such an estimate cannot be computed.
+  virtual size_t getNodeSmootherComplexity() const = 0;
 
-    void declareConstructionOutcome(bool fail, std::string msg) {
-      constructionSuccessful_ = !fail;
-      if (!fail)
-        constructionErrorMsg_ = "";
-      else
-        constructionErrorMsg_ = msg;
-    };
-    bool constructionSuccessful() { return constructionSuccessful_; }
-    std::string constructionErrorMsg() { return constructionErrorMsg_; }
+  void declareConstructionOutcome(bool fail, std::string msg) {
+    constructionSuccessful_ = !fail;
+    if (!fail)
+      constructionErrorMsg_ = "";
+    else
+      constructionErrorMsg_ = msg;
+  };
+  bool constructionSuccessful() { return constructionSuccessful_; }
+  std::string constructionErrorMsg() { return constructionErrorMsg_; }
 
-    //@}
+  //@}
 
-  private:
-    bool constructionSuccessful_;
-    std::string constructionErrorMsg_;
+ private:
+  bool constructionSuccessful_;
+  std::string constructionErrorMsg_;
 
-  }; //class SmootherBase
-} //namespace MueLu
+};  // class SmootherBase
+}  // namespace MueLu
 
 #define MUELU_SMOOTHERBASE_SHORT
 
-#endif //ifndef MUELU_SMOOTHERBASE_HPP
+#endif  // ifndef MUELU_SMOOTHERBASE_HPP
 
 // SmootherBase = Interface used by Hierarchy.Iterate(). Minimal condition to be used as smoother.
 // SmootherPrototype = minimal interface used by the generic SmootherFactory.
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherCloner.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherCloner.hpp
index 6535da0881f5..b204c8cebb28 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherCloner.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherCloner.hpp
@@ -59,8 +59,7 @@
 
 namespace MueLu {
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SMOOTHERBASECLONER_SHORT
-#endif //ifndef MUELU_SMOOTHERBASECLONER_HPP
-
+#endif  // ifndef MUELU_SMOOTHERBASECLONER_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherFactoryBase.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherFactoryBase.hpp
index d34e7bcc7174..a48ff1bb5b3d 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherFactoryBase.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherFactoryBase.hpp
@@ -52,41 +52,40 @@
 #include "MueLu_Types.hpp"
 
 namespace MueLu {
-  class Level;
+class Level;
 
-  /*!
-    @class Smoother factory base class.
-    @ingroup MueLuSmootherClasses
-    @brief Base class for smoother factories.
+/*!
+  @class Smoother factory base class.
+  @ingroup MueLuSmootherClasses
+  @brief Base class for smoother factories.
 
-    This interface inherits from SingleLevelFactoryBase and defines an additional Build method (BuildSmoother)
-    This new build method is used by Hierarchy for the CoarsestLevel.
-  */
+  This interface inherits from SingleLevelFactoryBase and defines an additional Build method (BuildSmoother)
+  This new build method is used by Hierarchy for the CoarsestLevel.
+*/
 
-  class SmootherFactoryBase : public SingleLevelFactoryBase {
+class SmootherFactoryBase : public SingleLevelFactoryBase {
+ public:
+  //@{ Constructors/Destructors.
+  SmootherFactoryBase() {}
 
-  public:
-    //@{ Constructors/Destructors.
-    SmootherFactoryBase() {}
+  virtual ~SmootherFactoryBase() {}
+  //@}
 
-    virtual ~SmootherFactoryBase() {}
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  //! Build pre-smoother and/or post-smoother
+  virtual void Build(Level& currentLevel) const = 0;
 
-    //! Build pre-smoother and/or post-smoother
-    virtual void Build(Level & currentLevel) const = 0;
+  virtual void BuildSmoother(Level& currentLevel, PreOrPost const preOrPost = BOTH) const = 0;
+  //@}
 
-    virtual void BuildSmoother(Level & currentLevel, PreOrPost const preOrPost = BOTH) const = 0;
-    //@}
+};  // class SmootherFactoryBase
 
-  }; //class SmootherFactoryBase
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SMOOTHERFACTORYBASE_SHORT
 
-#endif //ifndef MUELU_SMOOTHERFACTORYBASE_HPP
+#endif  // ifndef MUELU_SMOOTHERFACTORYBASE_HPP
 
-//TODO: remove this interface?
+// TODO: remove this interface?
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherFactory_decl.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherFactory_decl.hpp
index b97844138575..316610beb5a1 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherFactory_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherFactory_decl.hpp
@@ -57,169 +57,168 @@
 
 namespace MueLu {
 
-  /*!
-    @class SmootherFactory
-    @ingroup MueLuSmootherClasses
-    @brief Generic Smoother Factory for generating the smoothers of the MG hierarchy
-
-    This factory is generic and can produce any kind of Smoother.
-
-    The design of the Smoother factory is based on the prototype design
-    pattern.  The smoother factory uses prototypical instances of
-    smoothers, and prototypes are cloned to produce the smoothers of the MG
-    hierarchy.
-
-    See also:
-    - http://en.wikipedia.org/wiki/Factory_method_pattern
-    - http://en.wikipedia.org/wiki/Prototype_pattern
-    - http://www.oodesign.com/prototype-pattern.html
-
-    There is one prototype for the pre-smoother and one for the
-    post-smoother. Thus, this factory can produce two different smoothers
-    for pre and post-smoothing. Prototypes are stored in this factory.
-
-    The type of smoother to create is determined by the prototypical
-    instances. Prototypes store their own parameters (nits, omega), and the
-    factory doesn't have any knowledge about that.
-  */
-
-  template <class Scalar = double,
-            class LocalOrdinal = int,
-            class GlobalOrdinal = LocalOrdinal,
-            class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class SmootherFactory : public SmootherFactoryBase {
-  public:
-    typedef Scalar scalar_type;
-    typedef LocalOrdinal local_ordinal_type;
-    typedef GlobalOrdinal global_ordinal_type;
-    typedef Node node_type;
-
-  private:
+/*!
+  @class SmootherFactory
+  @ingroup MueLuSmootherClasses
+  @brief Generic Smoother Factory for generating the smoothers of the MG hierarchy
+
+  This factory is generic and can produce any kind of Smoother.
+
+  The design of the Smoother factory is based on the prototype design
+  pattern.  The smoother factory uses prototypical instances of
+  smoothers, and prototypes are cloned to produce the smoothers of the MG
+  hierarchy.
+
+  See also:
+  - http://en.wikipedia.org/wiki/Factory_method_pattern
+  - http://en.wikipedia.org/wiki/Prototype_pattern
+  - http://www.oodesign.com/prototype-pattern.html
+
+  There is one prototype for the pre-smoother and one for the
+  post-smoother. Thus, this factory can produce two different smoothers
+  for pre and post-smoothing. Prototypes are stored in this factory.
+
+  The type of smoother to create is determined by the prototypical
+  instances. Prototypes store their own parameters (nits, omega), and the
+  factory doesn't have any knowledge about that.
+*/
+
+template <class Scalar        = double,
+          class LocalOrdinal  = int,
+          class GlobalOrdinal = LocalOrdinal,
+          class Node          = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class SmootherFactory : public SmootherFactoryBase {
+ public:
+  typedef Scalar scalar_type;
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef Node node_type;
+
+ private:
 #undef MUELU_SMOOTHERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    /*!
-      @brief Constructor
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-      TODO update the documentation from Matlab syntax
+  /*!
+    @brief Constructor
 
-      The constructor takes as arguments pre-configured smoother
-      object(s). They are used as prototypes by the factory.
+    TODO update the documentation from Matlab syntax
 
-      SYNTAX
+    The constructor takes as arguments pre-configured smoother
+    object(s). They are used as prototypes by the factory.
 
-      SmootherFactory(preSmootherPrototype_, postSmootherPrototype_);
+    SYNTAX
 
-      preSmootherPrototype_  - prototype for pre-smoothers  (SmootherPrototype)
-      postSmootherPrototype_ - prototype for post-smoothers
-      (SmootherPrototype,optional,default=preSmootherPrototype_)
+    SmootherFactory(preSmootherPrototype_, postSmootherPrototype_);
 
-      EXAMPLES:
+    preSmootherPrototype_  - prototype for pre-smoothers  (SmootherPrototype)
+    postSmootherPrototype_ - prototype for post-smoothers
+    (SmootherPrototype,optional,default=preSmootherPrototype_)
 
-      SmootherFactory = SmootherFactory(ChebySmoother(5,1/30))
+    EXAMPLES:
 
-      or
+    SmootherFactory = SmootherFactory(ChebySmoother(5,1/30))
 
-      nIts        = 5;
-      lambdaRatio = 1/30;
-      Smoother        = ChebySmoother()
-      Smoother        = Smoother.SetIts(nIts);
-      Smoother        = Smoother.SetLambdaRatio(1/30);
-      SmootherFactory = SmootherFactory(Smoother);
+    or
 
-      To use different smoothers for pre and post smoothing, two prototypes can be passed in as argument:
-      PreSmoother     = ChebySmoother(2, 1/30)
-      PostSmoother    = ChebySmoother(10,1/30)
-      SmootherFactory = SmootherFactory(PreSmoother, PostSmoother);
+    nIts        = 5;
+    lambdaRatio = 1/30;
+    Smoother        = ChebySmoother()
+    Smoother        = Smoother.SetIts(nIts);
+    Smoother        = Smoother.SetLambdaRatio(1/30);
+    SmootherFactory = SmootherFactory(Smoother);
 
-      [] is also a valid smoother which do nothing:
-      PostSmoother    = ChebySmoother(10,1/30)
-      SmootherFactory = SmootherFactory([], PostSmoother);
-    */
-    //Note: Teuchos::null as parameter allowed (= no smoother)
-    //Note: precondition: input smoother must not be Setup()
-    SmootherFactory(RCP<SmootherPrototype> preAndPostSmootherPrototype = Teuchos::null);
+    To use different smoothers for pre and post smoothing, two prototypes can be passed in as argument:
+    PreSmoother     = ChebySmoother(2, 1/30)
+    PostSmoother    = ChebySmoother(10,1/30)
+    SmootherFactory = SmootherFactory(PreSmoother, PostSmoother);
 
-    SmootherFactory(RCP<SmootherPrototype> preSmootherPrototype, RCP<SmootherPrototype> postSmootherPrototype);
+    [] is also a valid smoother which do nothing:
+    PostSmoother    = ChebySmoother(10,1/30)
+    SmootherFactory = SmootherFactory([], PostSmoother);
+  */
+  // Note: Teuchos::null as parameter allowed (= no smoother)
+  // Note: precondition: input smoother must not be Setup()
+  SmootherFactory(RCP<SmootherPrototype> preAndPostSmootherPrototype = Teuchos::null);
 
-    virtual ~SmootherFactory() { }
-    //@}
+  SmootherFactory(RCP<SmootherPrototype> preSmootherPrototype, RCP<SmootherPrototype> postSmootherPrototype);
 
-    //! @name Set/Get methods.
-    //@{
+  virtual ~SmootherFactory() {}
+  //@}
 
-    //! Set smoother prototypes.
-    void SetSmootherPrototypes(RCP<SmootherPrototype> preAndPostSmootherPrototype);
+  //! @name Set/Get methods.
+  //@{
 
-    //! Set smoother prototypes.
-    void SetSmootherPrototypes(RCP<SmootherPrototype> preSmootherPrototype, RCP<SmootherPrototype> postSmootherPrototype);
+  //! Set smoother prototypes.
+  void SetSmootherPrototypes(RCP<SmootherPrototype> preAndPostSmootherPrototype);
 
-    //! Get smoother prototypes.
-    void GetSmootherPrototypes(RCP<SmootherPrototype>& preSmootherPrototype, RCP<SmootherPrototype>& postSmootherPrototype) const;
+  //! Set smoother prototypes.
+  void SetSmootherPrototypes(RCP<SmootherPrototype> preSmootherPrototype, RCP<SmootherPrototype> postSmootherPrototype);
 
-    //! Input
-    //@{
+  //! Get smoother prototypes.
+  void GetSmootherPrototypes(RCP<SmootherPrototype>& preSmootherPrototype, RCP<SmootherPrototype>& postSmootherPrototype) const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*! @brief Creates pre and post smoothers.
+  //! @name Build methods.
+  //@{
 
-    Factory.Build() clones its prototypes and calls Setup() on the
-    new objects to create fully functional smoothers for the current
-    hierarchy level. Note that smoothers do their own setup
-    (ie: ILUSmoother.Setup() computes the ILU factorization).
+  /*! @brief Creates pre and post smoothers.
 
-    If pre and post smoother are identical, the Setup() phase
-    is done only once: to create the post-smoother, the
-    pre-smoother is duplicated and its parameters are changed
-    according to the parameters of the post-smoother prototype.
+  Factory.Build() clones its prototypes and calls Setup() on the
+  new objects to create fully functional smoothers for the current
+  hierarchy level. Note that smoothers do their own setup
+  (ie: ILUSmoother.Setup() computes the ILU factorization).
 
-    If the parameters of pre and post smoothers are not the same,
-    the Setup() phase is also done only once when parameters
-    don't change the result of the setup computation.
-    */
-    void Build(Level& currentLevel) const;
+  If pre and post smoother are identical, the Setup() phase
+  is done only once: to create the post-smoother, the
+  pre-smoother is duplicated and its parameters are changed
+  according to the parameters of the post-smoother prototype.
 
-    void BuildSmoother(Level& currentLevel, const PreOrPost preOrPost = BOTH) const; // Build()
+  If the parameters of pre and post smoothers are not the same,
+  the Setup() phase is also done only once when parameters
+  don't change the result of the setup computation.
+  */
+  void Build(Level& currentLevel) const;
 
-    //@}
+  void BuildSmoother(Level& currentLevel, const PreOrPost preOrPost = BOTH) const;  // Build()
 
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    using MueLu::Describable::describe; // overloading, not hiding
-    void describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  using MueLu::Describable::describe;  // overloading, not hiding
+  void describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //@}
 
-  private:
-    RCP<SmootherPrototype> preSmootherPrototype_;
-    RCP<SmootherPrototype> postSmootherPrototype_;
+ private:
+  RCP<SmootherPrototype> preSmootherPrototype_;
+  RCP<SmootherPrototype> postSmootherPrototype_;
 
-    void CheckPrototypes() const;
+  void CheckPrototypes() const;
 
-    //@}
+  //@}
 
-  }; // class SmootherFactory
+};  // class SmootherFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
-//TODO: doc: setup done twice if PostSmoother object != PreSmoother object and no adv. reused capability
+// TODO: doc: setup done twice if PostSmoother object != PreSmoother object and no adv. reused capability
 #define MUELU_SMOOTHERFACTORY_SHORT
-#endif // MUELU_SMOOTHERFACTORY_DECL_HPP
+#endif  // MUELU_SMOOTHERFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherFactory_def.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherFactory_def.hpp
index 84d5a7986e1b..32c61a5831d7 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherFactory_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherFactory_def.hpp
@@ -52,244 +52,248 @@
 #include "MueLu_Exceptions.hpp"
 #include "MueLu_SmootherPrototype.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherFactory(RCP<SmootherPrototype> preAndPostSmootherPrototype) {
-    SetSmootherPrototypes(preAndPostSmootherPrototype);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherFactory(RCP<SmootherPrototype> preSmootherPrototype,
-                                                                              RCP<SmootherPrototype> postSmootherPrototype) {
-    SetSmootherPrototypes(preSmootherPrototype, postSmootherPrototype);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmootherPrototypes(RCP<SmootherPrototype> preAndPostSmootherPrototype) {
-    preSmootherPrototype_  = postSmootherPrototype_ = preAndPostSmootherPrototype;
-    CheckPrototypes();
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmootherPrototypes(RCP<SmootherPrototype> preSmootherPrototype,
-                                                                                         RCP<SmootherPrototype> postSmootherPrototype) {
-    preSmootherPrototype_  = preSmootherPrototype;
-    postSmootherPrototype_ = postSmootherPrototype;
-    CheckPrototypes();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<bool>("keep smoother data", false, "Keep constructed smoothers for later reuse");
-
-    validParamList->set< RCP<SmootherPrototype> >("PreSmoother data",   null, "Pre-smoother data for reuse");
-    validParamList->set< RCP<SmootherPrototype> >("PostSmoother data",  null, "Post-smoother data for reuse");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckPrototypes() const {
-    TEUCHOS_TEST_FOR_EXCEPTION(preSmootherPrototype_  != Teuchos::null && preSmootherPrototype_->IsSetup()  == true,
-                               Exceptions::RuntimeError, "preSmoother prototype is not a smoother prototype (IsSetup() == true)");
-    TEUCHOS_TEST_FOR_EXCEPTION(postSmootherPrototype_ != Teuchos::null && postSmootherPrototype_->IsSetup() == true,
-                               Exceptions::RuntimeError, "postSmoother prototype is not a smoother prototype (IsSetup() == true)");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherFactory(RCP<SmootherPrototype> preAndPostSmootherPrototype) {
+  SetSmootherPrototypes(preAndPostSmootherPrototype);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherFactory(RCP<SmootherPrototype> preSmootherPrototype,
+                                                                            RCP<SmootherPrototype> postSmootherPrototype) {
+  SetSmootherPrototypes(preSmootherPrototype, postSmootherPrototype);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmootherPrototypes(RCP<SmootherPrototype> preAndPostSmootherPrototype) {
+  preSmootherPrototype_ = postSmootherPrototype_ = preAndPostSmootherPrototype;
+  CheckPrototypes();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetSmootherPrototypes(RCP<SmootherPrototype> preSmootherPrototype,
+                                                                                       RCP<SmootherPrototype> postSmootherPrototype) {
+  preSmootherPrototype_  = preSmootherPrototype;
+  postSmootherPrototype_ = postSmootherPrototype;
+  CheckPrototypes();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<bool>("keep smoother data", false, "Keep constructed smoothers for later reuse");
+
+  validParamList->set<RCP<SmootherPrototype> >("PreSmoother data", null, "Pre-smoother data for reuse");
+  validParamList->set<RCP<SmootherPrototype> >("PostSmoother data", null, "Post-smoother data for reuse");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckPrototypes() const {
+  TEUCHOS_TEST_FOR_EXCEPTION(preSmootherPrototype_ != Teuchos::null && preSmootherPrototype_->IsSetup() == true,
+                             Exceptions::RuntimeError, "preSmoother prototype is not a smoother prototype (IsSetup() == true)");
+  TEUCHOS_TEST_FOR_EXCEPTION(postSmootherPrototype_ != Teuchos::null && postSmootherPrototype_->IsSetup() == true,
+                             Exceptions::RuntimeError, "postSmoother prototype is not a smoother prototype (IsSetup() == true)");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSmootherPrototypes(RCP<SmootherPrototype>& preSmootherPrototype,
+                                                                                       RCP<SmootherPrototype>& postSmootherPrototype) const {
+  preSmootherPrototype  = preSmootherPrototype_;
+  postSmootherPrototype = postSmootherPrototype_;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  if (preSmootherPrototype_ != Teuchos::null)
+    preSmootherPrototype_->DeclareInput(currentLevel);
+
+  if ((postSmootherPrototype_ != Teuchos::null) && (preSmootherPrototype_ != postSmootherPrototype_))
+    postSmootherPrototype_->DeclareInput(currentLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  return BuildSmoother(currentLevel, BOTH);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildSmoother(Level& currentLevel, PreOrPost const preOrPost) const {
+  // SmootherFactory is quite tricky because of the fact that one of the smoother prototypes may be zero.
+  // The challenge is that we have no way of knowing how user uses this factory. For instance, lets say
+  // user wants to use s1 prototype as a presmoother, and s2 as a postsmoother. They could do:
+  //   (a) create SmootherFactory(s1, s2), or
+  //   (b) create SmootherFactory(s1, null) and SmootherFactory(null, s2)
+  // It may also happen that somewhere somebody set presmoother factory = postsmoother factory = (a)
+  // How do you do DeclareInput in this case? It could easily introduce a bug if a user does not check
+  // whether presmoother = postsmoother. A buggy code could look like that:
+  //   RCP<SmootherFactory> s = rcp(new SmootherFactory(s1,s2));
+  //   level.Request("PreSmoother",  s.get());
+  //   level.Request("PostSmoother", s.get());
+  //   Get<RCP<SmootherBase> > pre  = Get<RCP<SmootherBase> >("PreSmoother",  s.get());
+  //   Get<RCP<SmootherBase> > post = Get<RCP<SmootherBase> >("PostSmoother", s.get());
+  // This code would call DeclareInput in request mode twice, but as the Build method generates both Pre and Post
+  // smoothers, it would call DelcareInput in release mode only once, leaving requests.
+  // This code has another problem if s2 = Teuchos::null. In that case, despite the request for PostSmoother, the factory
+  // would not generate one, and second Get would throw. The real issue here is that given a Factory pointer
+  // there is no way to be sure that this factory would generate any of "PreSmoother" or "PostSmoother", unless you are
+  // able to cast it to SmootherFactory, do GetPrototypes and to check whether any of those is Teuchos::null.
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  RCP<SmootherPrototype> preSmoother, postSmoother;
+  ParameterList preSmootherParams, postSmootherParams;
+
+  if ((preOrPost & PRE) && !preSmootherPrototype_.is_null()) {
+    if (currentLevel.IsAvailable("PreSmoother data", this))
+      preSmoother = currentLevel.Get<RCP<SmootherPrototype> >("PreSmoother data", this);
+    else
+      preSmoother = preSmootherPrototype_->Copy();
+
+    int oldRank = -1;
+    if (!currentLevel.GetComm().is_null())
+      oldRank = preSmoother->SetProcRankVerbose(currentLevel.GetComm()->getRank());
+
+    preSmoother->Setup(currentLevel);
+    preSmootherParams = preSmoother->GetParameterList();
+
+    if (oldRank != -1)
+      preSmoother->SetProcRankVerbose(oldRank);
+
+    currentLevel.Set<RCP<SmootherBase> >("PreSmoother", preSmoother, this);
+
+    if (pL.get<bool>("keep smoother data"))
+      Set(currentLevel, "PreSmoother data", preSmoother);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetSmootherPrototypes(RCP<SmootherPrototype>& preSmootherPrototype,
-                                                                                         RCP<SmootherPrototype>& postSmootherPrototype) const {
-    preSmootherPrototype  = preSmootherPrototype_;
-    postSmootherPrototype = postSmootherPrototype_;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    if (preSmootherPrototype_ != Teuchos::null)
-      preSmootherPrototype_->DeclareInput(currentLevel);
+  if ((preOrPost & POST) && !postSmootherPrototype_.is_null()) {
+    if (preOrPost == BOTH && preSmootherPrototype_ == postSmootherPrototype_) {
+      // Simple reuse
+      // Same prototypes for pre- and post-smoothers mean that we only need to call Setup only once
+      postSmoother = preSmoother;
+
+      //               else if (preOrPost == BOTH &&
+      //                        preSmootherPrototype_ != Teuchos::null &&
+      //                        preSmootherPrototype_->GetType() == postSmootherPrototype_->GetType()) {
+
+      //               // More complex reuse case: need implementation of CopyParameters() and a smoothers smart enough to know when parameters affect the setup phase.
+
+      //               // YES: post-smoother == pre-smoother
+      //               // => copy the pre-smoother to avoid the setup phase of the post-smoother.
+      //               postSmoother = preSmoother->Copy();
+      //               // If the post-smoother parameters are different from
+      //               // pre-smoother, the parameters stored in the post-smoother
+      //               // prototype are copied in the new post-smoother object.
+      //               postSmoother->CopyParameters(postSmootherPrototype_);
+      //               // If parameters don't influence the Setup phase (it is the case
+      //               // for Jacobi, Chebyshev...), PostSmoother is already setup. Nothing
+      //               // more to do. In the case of ILU, parameters of the smoother
+      //               // are in fact the parameters of the Setup phase. The call to
+      //               // CopyParameters resets the smoother (only if parameters are
+      //               // different) and we must call Setup() again.
+      //               postSmoother->Setup(currentLevel);
+      //               }
+
+      //               // TODO: if CopyParameters do not exist, do setup twice.
 
-    if ((postSmootherPrototype_ != Teuchos::null) && (preSmootherPrototype_ != postSmootherPrototype_))
-      postSmootherPrototype_->DeclareInput(currentLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    return BuildSmoother(currentLevel, BOTH);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildSmoother(Level& currentLevel, PreOrPost const preOrPost) const {
-    // SmootherFactory is quite tricky because of the fact that one of the smoother prototypes may be zero.
-    // The challenge is that we have no way of knowing how user uses this factory. For instance, lets say
-    // user wants to use s1 prototype as a presmoother, and s2 as a postsmoother. They could do:
-    //   (a) create SmootherFactory(s1, s2), or
-    //   (b) create SmootherFactory(s1, null) and SmootherFactory(null, s2)
-    // It may also happen that somewhere somebody set presmoother factory = postsmoother factory = (a)
-    // How do you do DeclareInput in this case? It could easily introduce a bug if a user does not check
-    // whether presmoother = postsmoother. A buggy code could look like that:
-    //   RCP<SmootherFactory> s = rcp(new SmootherFactory(s1,s2));
-    //   level.Request("PreSmoother",  s.get());
-    //   level.Request("PostSmoother", s.get());
-    //   Get<RCP<SmootherBase> > pre  = Get<RCP<SmootherBase> >("PreSmoother",  s.get());
-    //   Get<RCP<SmootherBase> > post = Get<RCP<SmootherBase> >("PostSmoother", s.get());
-    // This code would call DeclareInput in request mode twice, but as the Build method generates both Pre and Post
-    // smoothers, it would call DelcareInput in release mode only once, leaving requests.
-    // This code has another problem if s2 = Teuchos::null. In that case, despite the request for PostSmoother, the factory
-    // would not generate one, and second Get would throw. The real issue here is that given a Factory pointer
-    // there is no way to be sure that this factory would generate any of "PreSmoother" or "PostSmoother", unless you are
-    // able to cast it to SmootherFactory, do GetPrototypes and to check whether any of those is Teuchos::null.
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-
-    RCP<SmootherPrototype> preSmoother, postSmoother;
-    ParameterList preSmootherParams, postSmootherParams;
-
-    if ((preOrPost & PRE) && !preSmootherPrototype_.is_null()) {
-
-      if (currentLevel.IsAvailable("PreSmoother data", this))
-        preSmoother = currentLevel.Get<RCP<SmootherPrototype> >("PreSmoother data", this);
-      else
-        preSmoother = preSmootherPrototype_->Copy();
+    } else {
+      if (currentLevel.IsAvailable("PostSmoother data", this)) {
+        postSmoother = currentLevel.Get<RCP<SmootherPrototype> >("PostSmoother data", this);
+      } else {
+        // No reuse:
+        //  - either we only do postsmoothing without any presmoothing
+        //  - or our postsmoother is different from presmoother
+        postSmoother = postSmootherPrototype_->Copy();
+      }
 
       int oldRank = -1;
       if (!currentLevel.GetComm().is_null())
-        oldRank = preSmoother->SetProcRankVerbose(currentLevel.GetComm()->getRank());
+        oldRank = postSmoother->SetProcRankVerbose(GetProcRankVerbose());
 
-      preSmoother->Setup(currentLevel);
-      preSmootherParams = preSmoother->GetParameterList();
+      postSmoother->Setup(currentLevel);
+      postSmootherParams = postSmoother->GetParameterList();
 
       if (oldRank != -1)
-        preSmoother->SetProcRankVerbose(oldRank);
-
-      currentLevel.Set<RCP<SmootherBase> >("PreSmoother", preSmoother, this);
-
-      if (pL.get<bool>("keep smoother data"))
-        Set(currentLevel, "PreSmoother data", preSmoother);
+        postSmoother->SetProcRankVerbose(oldRank);
     }
 
-    if ((preOrPost & POST) && !postSmootherPrototype_.is_null()) {
-      if (preOrPost == BOTH && preSmootherPrototype_ == postSmootherPrototype_) {
-        // Simple reuse
-        // Same prototypes for pre- and post-smoothers mean that we only need to call Setup only once
-        postSmoother = preSmoother;
-
-        //               else if (preOrPost == BOTH &&
-        //                        preSmootherPrototype_ != Teuchos::null &&
-        //                        preSmootherPrototype_->GetType() == postSmootherPrototype_->GetType()) {
-
-        //               // More complex reuse case: need implementation of CopyParameters() and a smoothers smart enough to know when parameters affect the setup phase.
-
-        //               // YES: post-smoother == pre-smoother
-        //               // => copy the pre-smoother to avoid the setup phase of the post-smoother.
-        //               postSmoother = preSmoother->Copy();
-        //               // If the post-smoother parameters are different from
-        //               // pre-smoother, the parameters stored in the post-smoother
-        //               // prototype are copied in the new post-smoother object.
-        //               postSmoother->CopyParameters(postSmootherPrototype_);
-        //               // If parameters don't influence the Setup phase (it is the case
-        //               // for Jacobi, Chebyshev...), PostSmoother is already setup. Nothing
-        //               // more to do. In the case of ILU, parameters of the smoother
-        //               // are in fact the parameters of the Setup phase. The call to
-        //               // CopyParameters resets the smoother (only if parameters are
-        //               // different) and we must call Setup() again.
-        //               postSmoother->Setup(currentLevel);
-        //               }
-
-        //               // TODO: if CopyParameters do not exist, do setup twice.
-
-      } else {
-
-        if (currentLevel.IsAvailable("PostSmoother data", this)) {
-          postSmoother = currentLevel.Get<RCP<SmootherPrototype> >("PostSmoother data", this);
-        } else {
-          // No reuse:
-          //  - either we only do postsmoothing without any presmoothing
-          //  - or our postsmoother is different from presmoother
-          postSmoother = postSmootherPrototype_->Copy();
-        }
+    currentLevel.Set<RCP<SmootherBase> >("PostSmoother", postSmoother, this);
 
-        int oldRank = -1;
-        if (!currentLevel.GetComm().is_null())
-          oldRank = postSmoother->SetProcRankVerbose(GetProcRankVerbose());
+    if (pL.get<bool>("keep smoother data"))
+      Set(currentLevel, "PostSmoother data", postSmoother);
+  }
 
-        postSmoother->Setup(currentLevel);
-        postSmootherParams = postSmoother->GetParameterList();
+  ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
+  if (postSmoother == preSmoother && !preSmoother.is_null()) {
+    paramList.sublist("smoother", false) = preSmoother->GetParameterList();
 
-        if (oldRank != -1)
-          postSmoother->SetProcRankVerbose(oldRank);
-      }
+  } else {
+    if (!preSmoother.is_null())
+      paramList.sublist("presmoother", false) = preSmootherParams;
 
-      currentLevel.Set<RCP<SmootherBase> >("PostSmoother", postSmoother, this);
+    if (!postSmoother.is_null())
+      paramList.sublist("postsmoother", false) = postSmootherParams;
+  }
 
-      if (pL.get<bool>("keep smoother data"))
-        Set(currentLevel, "PostSmoother data", postSmoother);
+}  // Build()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  out << SmootherFactoryBase::description();
+  std::string preStr  = (preSmootherPrototype_ == Teuchos::null) ? "null" : preSmootherPrototype_->description();
+  std::string postStr = (preSmootherPrototype_ == postSmootherPrototype_) ? "pre" : ((postSmootherPrototype_ == Teuchos::null) ? "null" : postSmootherPrototype_->description());
+  out << "{pre = " << preStr << ", post = " << postStr << "}";
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0) {
+    out0 << "PreSmoother : ";
+    if (preSmootherPrototype_.is_null()) {
+      out0 << "null" << std::endl;
+    } else {
+      Teuchos::OSTab tab2(out);
+      preSmootherPrototype_->describe(out, verbLevel);
     }
 
-    ParameterList& paramList = const_cast<ParameterList&>(this->GetParameterList());
-    if (postSmoother == preSmoother && !preSmoother.is_null()) {
-      paramList.sublist("smoother", false) = preSmoother->GetParameterList();
-
+    out0 << "PostSmoother: ";
+    if (postSmootherPrototype_ == preSmootherPrototype_) {
+      out0 << "same as PreSmoother" << std::endl;
+    } else if (postSmootherPrototype_ == Teuchos::null) {
+      out0 << "null" << std::endl;
     } else {
-      if (!preSmoother.is_null())
-        paramList.sublist("presmoother", false) = preSmootherParams;
-
-      if (!postSmoother.is_null())
-        paramList.sublist("postsmoother", false) = postSmootherParams;
+      Teuchos::OSTab tab2(out);
+      postSmootherPrototype_->describe(out, verbLevel);
+      out0 << "PostSmoother is different than PreSmoother (not the same object)" << std::endl;
     }
-
-  } // Build()
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    out << SmootherFactoryBase::description();
-    std::string preStr  = (preSmootherPrototype_ == Teuchos::null)          ? "null" : preSmootherPrototype_->description();
-    std::string postStr = (preSmootherPrototype_ == postSmootherPrototype_) ? "pre"  : ( (postSmootherPrototype_ == Teuchos::null) ? "null" : postSmootherPrototype_->description() );
-    out << "{pre = "  << preStr << ", post = "<< postStr << "}";
-    return out.str();
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::describe(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0) {
-      out0 << "PreSmoother : ";
-      if (preSmootherPrototype_.is_null()) {
-        out0 << "null" << std::endl;
-      } else {
-        Teuchos::OSTab tab2(out);
-        preSmootherPrototype_->describe(out, verbLevel);
-      }
-
-      out0 << "PostSmoother: ";
-      if      (postSmootherPrototype_ == preSmootherPrototype_) { out0 << "same as PreSmoother" << std::endl; }
-      else if (postSmootherPrototype_ == Teuchos::null)         { out0 << "null" << std::endl; }
-      else {
-        Teuchos::OSTab tab2(out);
-        postSmootherPrototype_->describe(out, verbLevel);
-        out0 << "PostSmoother is different than PreSmoother (not the same object)" << std::endl;
-      }
+  if (verbLevel & Debug) {
+    if (preSmootherPrototype_ != Teuchos::null || postSmootherPrototype_ != Teuchos::null) {
+      out0 << "-" << std::endl;
     }
-
-    if (verbLevel & Debug) {
-      if (preSmootherPrototype_  != Teuchos::null || postSmootherPrototype_ != Teuchos::null) { out0 << "-" << std::endl; }
-      if (preSmootherPrototype_  != Teuchos::null) { out0 << "RCP<preSmootherPrototype_> : " << preSmootherPrototype_  << std::endl; }
-      if (postSmootherPrototype_ != Teuchos::null) { out0 << "RCP<postSmootherPrototype_>: " << postSmootherPrototype_ << std::endl; }
+    if (preSmootherPrototype_ != Teuchos::null) {
+      out0 << "RCP<preSmootherPrototype_> : " << preSmootherPrototype_ << std::endl;
+    }
+    if (postSmootherPrototype_ != Teuchos::null) {
+      out0 << "RCP<postSmootherPrototype_>: " << postSmootherPrototype_ << std::endl;
     }
   }
+}
 
+}  // namespace MueLu
 
-} // namespace MueLu
-
-//TODO: doc: setup done twice if PostSmoother object != PreSmoother object and no adv. reused capability
+// TODO: doc: setup done twice if PostSmoother object != PreSmoother object and no adv. reused capability
 
 // TODO ReUse:   If only one smoother is missing, SmootherFactory can be smart and build only the missing smoother.
 // TODO (optim): we can also reuse if preOrPost = post and preSmoother available in Level
 //               we can also reuse if preOrPost = pre  and postSmoother available in Level
 
-#endif // MUELU_SMOOTHERFACTORY_DEF_HPP
+#endif  // MUELU_SMOOTHERFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_decl.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_decl.hpp
index 1be1b53f3075..f8732d93742b 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_decl.hpp
@@ -53,98 +53,96 @@
 
 namespace MueLu {
 
-  class Level;
-
-  /*!
-    @class SmootherPrototype
-    @ingroup MueLuSmootherClasses
-    @brief Base class for smoother prototypes
-
-    A smoother prototype is a smoother which can be in two states:
-    - ready to be duplicated (parameters defined)
-    - ready to be used (setup phase completed)
-
-    'Smoother prototypes' can be fully copied using the Copy() method.
-  */
-
-  template <class Scalar = SmootherBase<>::scalar_type,
-            class LocalOrdinal = typename SmootherBase<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherBase<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class SmootherPrototype :
-    public SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>,
-    public Factory
-  {
-  public:
-    typedef Scalar scalar_type;
-    typedef LocalOrdinal local_ordinal_type;
-    typedef GlobalOrdinal global_ordinal_type;
-    typedef Node node_type;
-
-  private:
+class Level;
+
+/*!
+  @class SmootherPrototype
+  @ingroup MueLuSmootherClasses
+  @brief Base class for smoother prototypes
+
+  A smoother prototype is a smoother which can be in two states:
+  - ready to be duplicated (parameters defined)
+  - ready to be used (setup phase completed)
+
+  'Smoother prototypes' can be fully copied using the Copy() method.
+*/
+
+template <class Scalar        = SmootherBase<>::scalar_type,
+          class LocalOrdinal  = typename SmootherBase<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherBase<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class SmootherPrototype : public SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>,
+                          public Factory {
+ public:
+  typedef Scalar scalar_type;
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef Node node_type;
+
+ private:
 #undef MUELU_SMOOTHERPROTOTYPE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //!@nameConstructors/Destructors.
-    //@{
+ public:
+  //!@nameConstructors/Destructors.
+  //@{
 
-    SmootherPrototype();
+  SmootherPrototype();
 
-    virtual ~SmootherPrototype();
+  virtual ~SmootherPrototype();
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    virtual void DeclareInput(Level &currentLevel) const = 0;
+  virtual void DeclareInput(Level &currentLevel) const = 0;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    virtual void Setup(Level &) = 0;
+  virtual void Setup(Level &) = 0;
 
-    virtual RCP<SmootherPrototype> Copy() const = 0;
+  virtual RCP<SmootherPrototype> Copy() const = 0;
 
-    //@}
+  //@}
 
-    //! @name Get/Set methods.
-    //@{
+  //! @name Get/Set methods.
+  //@{
 
-    //! Get the state of a smoother prototype.
-    bool IsSetup() const;
+  //! Get the state of a smoother prototype.
+  bool IsSetup() const;
 
-    //! Set the state of a smoother prototype.
-    // Developpers: this method must be called by your Setup() method.
-    void IsSetup(bool const &ToF);
+  //! Set the state of a smoother prototype.
+  // Developpers: this method must be called by your Setup() method.
+  void IsSetup(bool const &ToF);
 
-    //@}
+  //@}
 
-    //! @name Implements FactoryBase interface
-    //@{
-    virtual void CallBuild(Level & /* requestedLevel */) const {
-      TEUCHOS_TEST_FOR_EXCEPT(true);
-    }
+  //! @name Implements FactoryBase interface
+  //@{
+  virtual void CallBuild(Level & /* requestedLevel */) const {
+    TEUCHOS_TEST_FOR_EXCEPT(true);
+  }
 
-    //!
-    virtual void CallDeclareInput(Level & requestedLevel) const {
-      DeclareInput(requestedLevel);
-    }
+  //!
+  virtual void CallDeclareInput(Level &requestedLevel) const {
+    DeclareInput(requestedLevel);
+  }
 
-    //@}
+  //@}
 
-  private:
-    bool isSetup_;
+ private:
+  bool isSetup_;
 
-  }; // class SmootherPrototype
+};  // class SmootherPrototype
 
-} // namespace MueLu
+}  // namespace MueLu
 
-//TODO: private copy constructor
-//TODO: update comments
+// TODO: private copy constructor
+// TODO: update comments
 
 #define MUELU_SMOOTHERPROTOTYPE_SHORT
-#endif // MUELU_SMOOTHERPROTOTYPE_DECL_HPP
+#endif  // MUELU_SMOOTHERPROTOTYPE_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_def.hpp b/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_def.hpp
index 376a4a3c8ef2..ce87ca4f5339 100644
--- a/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_SmootherPrototype_def.hpp
@@ -50,20 +50,21 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherPrototype() : isSetup_(false) {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SmootherPrototype()
+  : isSetup_(false) {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~SmootherPrototype() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~SmootherPrototype() {}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsSetup() const { return isSetup_; }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsSetup() const { return isSetup_; }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsSetup(bool const &ToF) { isSetup_ = ToF; }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>::IsSetup(bool const &ToF) { isSetup_ = ToF; }
 
-} // namespace MueLu
+}  // namespace MueLu
 
-//TODO: private copy constructor
-//TODO: update comments
-#endif // MUELU_SMOOTHERPROTOTYPE_DEF_HPP
+// TODO: private copy constructor
+// TODO: update comments
+#endif  // MUELU_SMOOTHERPROTOTYPE_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_decl.hpp
index bf2cd4b6f604..55e2c10cb42f 100644
--- a/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_decl.hpp
@@ -65,210 +65,206 @@
 
 #include <Thyra_LinearOpWithSolveBase.hpp>
 
-
 namespace MueLu {
 
-  /*!
-    @class StratimikosSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates Stratimikos smoothers.
-
-    This class creates an Stratimikos preconditioner factory. The factory creates a smoother based
-    on the type and ParameterList passed into the constructor. See the constructor for more
-    information.
-    */
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class StratimikosSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+/*!
+  @class StratimikosSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates Stratimikos smoothers.
+
+  This class creates an Stratimikos preconditioner factory. The factory creates a smoother based
+  on the type and ParameterList passed into the constructor. See the constructor for more
+  information.
+  */
+
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class StratimikosSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_STRATIMIKOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors / destructors
-    //@{
-    //TODO: update doc for Stratimikos.
-    /*! @brief Constructor
+ public:
+  //! @name Constructors / destructors
+  //@{
+  // TODO: update doc for Stratimikos.
+  /*! @brief Constructor
 
-    The options passed into StratimikosSmoother are those given in the Stratimikos user's manual.
+  The options passed into StratimikosSmoother are those given in the Stratimikos user's manual.
 
-    @param type smoother type
-    @param list options for the particular smoother (e.g., fill factor or damping parameter)
+  @param type smoother type
+  @param list options for the particular smoother (e.g., fill factor or damping parameter)
 
-    */
+  */
 
 #ifndef _MSC_VER
-    // Avoid error C3772: invalid friend template declaration
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class StratimikosSmoother;
+  // Avoid error C3772: invalid friend template declaration
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class StratimikosSmoother;
 #endif
 
-    StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList()) {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Stratimikos only works for Scalar=double.")
-    };
+  StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList()){
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Stratimikos only works for Scalar=double.")};
 
-    //! Destructor
-    virtual ~StratimikosSmoother() = default;
+  //! Destructor
+  virtual ~StratimikosSmoother() = default;
 
-    //@}
+  //@}
 
-    void SetParameterList(const Teuchos::ParameterList& paramList) {};
+  void SetParameterList(const Teuchos::ParameterList& paramList){};
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const {};
+  void DeclareInput(Level& currentLevel) const {};
 
-    //@}
+  //@}
 
-    //! @name Computational methods.
-    //@{
+  //! @name Computational methods.
+  //@{
 
-    /*! @brief Set up the smoother.
+  /*! @brief Set up the smoother.
 
-    This creates the underlying Stratimikos smoother object, copies any parameter list options
-    supplied to the constructor to the Stratimikos object, and computes the preconditioner.
-    */
-    void Setup(Level &currentLevel) {};
-
-    /*! @brief Apply the preconditioner.
+  This creates the underlying Stratimikos smoother object, copies any parameter list options
+  supplied to the constructor to the Stratimikos object, and computes the preconditioner.
+  */
+  void Setup(Level& currentLevel){};
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  /*! @brief Apply the preconditioner.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Stratimikos implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {};
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    //@}
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Stratimikos implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const {};
 
-    //! @name Utilities
-    //@{
+  //@}
 
-    RCP<SmootherPrototype> Copy() const { return Teuchos::null; };
+  //! @name Utilities
+  //@{
 
-    //@}
+  RCP<SmootherPrototype> Copy() const { return Teuchos::null; };
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //@}
 
-    //! Return a simple one-line description of this object.
-    std::string description() const { return std::string(""); };
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {};
+  //! Return a simple one-line description of this object.
+  std::string description() const { return std::string(""); };
 
-    //@}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const {};
 
-    //! For diagnostic purposes
-    // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
+  //@}
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const { return Teuchos::OrdinalTraits<size_t>::invalid(); };
+  //! For diagnostic purposes
+  // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
 
-  }; // class StratimikosSmoother
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const { return Teuchos::OrdinalTraits<size_t>::invalid(); };
 
+};  // class StratimikosSmoother
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node>
-  struct StratimikosSmoother<double,LocalOrdinal,GlobalOrdinal,Node> : public SmootherPrototype<double,LocalOrdinal,GlobalOrdinal,Node>  {
-    typedef double Scalar;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+struct StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node> : public SmootherPrototype<double, LocalOrdinal, GlobalOrdinal, Node> {
+  typedef double Scalar;
 #undef MUELU_STRATIMIKOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-    StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
+  StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList = Teuchos::ParameterList());
 
-    //! Destructor
-    virtual ~StratimikosSmoother() = default;
+  //! Destructor
+  virtual ~StratimikosSmoother() = default;
 
-    //@}
+  //@}
 
-    void SetParameterList(const Teuchos::ParameterList& paramList);
+  void SetParameterList(const Teuchos::ParameterList& paramList);
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Computational methods.
-    //@{
+  //! @name Computational methods.
+  //@{
 
-    /*! @brief Set up the smoother.
+  /*! @brief Set up the smoother.
 
-    This creates the underlying Stratimikos smoother object, copies any parameter list options
-    supplied to the constructor to the Stratimikos object, and computes the preconditioner.
-    */
-    void Setup(Level &currentLevel);
+  This creates the underlying Stratimikos smoother object, copies any parameter list options
+  supplied to the constructor to the Stratimikos object, and computes the preconditioner.
+  */
+  void Setup(Level& currentLevel);
 
-    /*! @brief Apply the preconditioner.
+  /*! @brief Apply the preconditioner.
 
-    Solves the linear system <tt>AX=B</tt> using the constructed smoother.
+  Solves the linear system <tt>AX=B</tt> using the constructed smoother.
 
-    @param X initial guess
-    @param B right-hand side
-    @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Stratimikos implementation.
-    */
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  @param X initial guess
+  @param B right-hand side
+  @param InitialGuessIsZero (optional) If false, some work can be avoided. Whether this actually saves any work depends on the underlying Stratimikos implementation.
+  */
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //@}
+  //@}
 
-    //! @name Utilities
-    //@{
+  //! @name Utilities
+  //@{
 
-    RCP<SmootherPrototype> Copy() const;
+  RCP<SmootherPrototype> Copy() const;
 
-    //@}
+  //@}
 
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
-    void print(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const;
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //@}
+  //@}
 
-    //! For diagnostic purposes
-    // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
+  //! For diagnostic purposes
+  // RCP<Ifpack2::Preconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node> > getPreconditioner(){return prec_;}
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const;
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const;
 
-  private:
-    void SetupStratimikos           (Level& currentLevel);
-    /*!
-      @brief Filter out any matrix connections that corresponds to coupling along a vertical grid line
+ private:
+  void SetupStratimikos(Level& currentLevel);
+  /*!
+    @brief Filter out any matrix connections that corresponds to coupling along a vertical grid line
 
-      Builds a special filtered matrix for meshes that are structured in at least one direction (the z direction). Specifically
-      the new matrix has all the vertical connections removed.
-      */
-    void ExperimentalDropVertConnections(RCP<Matrix> & filteredA, Level& currentLevel);
+    Builds a special filtered matrix for meshes that are structured in at least one direction (the z direction). Specifically
+    the new matrix has all the vertical connections removed.
+    */
+  void ExperimentalDropVertConnections(RCP<Matrix>& filteredA, Level& currentLevel);
 
-    //@}
+  //@}
 
-    std::string                         type_;
+  std::string type_;
 
-    Teuchos::RCP<Thyra::LinearOpWithSolveBase<Scalar> > solver_;
+  Teuchos::RCP<Thyra::LinearOpWithSolveBase<Scalar> > solver_;
 
-    //! matrix, used in apply if solving residual equation
-    RCP<Matrix> A_;
+  //! matrix, used in apply if solving residual equation
+  RCP<Matrix> A_;
 
-    bool recurMgOnFilteredA_ = false;
-  }; // class StratimikosSmoother
+  bool recurMgOnFilteredA_ = false;
+};  // class StratimikosSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_STRATIMIKOSSMOOTHER_SHORT
-#endif // HAVE_MUELU_STRATIMIKOS
-#endif // MUELU_STRATIMIKOSSMOOTHER_DECL_HPP
+#endif  // HAVE_MUELU_STRATIMIKOS
+#endif  // MUELU_STRATIMIKOSSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_def.hpp
index a5ae1f8ae30d..dc7c13adc53b 100644
--- a/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_StratimikosSmoother_def.hpp
@@ -69,257 +69,251 @@
 #include <Teuchos_ParameterList.hpp>
 #include <unordered_map>
 
-
 namespace MueLu {
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList)
-  : type_(type)
-  {
-    std::transform(type_.begin(), type_.end(), type_.begin(), ::toupper);
-    ParameterList& pList = const_cast<ParameterList&>(paramList);
-
-    if (pList.isParameter("smoother: recurMgOnFilteredA")) {
-      recurMgOnFilteredA_ = true;
-      pList.remove("smoother: recurMgOnFilteredA");
-    }
-    bool isSupported = type_ == "STRATIMIKOS";
-    this->declareConstructionOutcome(!isSupported, "Stratimikos does not provide the smoother '" + type_ + "'.");
-    if (isSupported)
-      SetParameterList(paramList);
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
-    Factory::SetParameterList(paramList);
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    this->Input(currentLevel, "A");
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::StratimikosSmoother(const std::string type, const Teuchos::ParameterList& paramList)
+  : type_(type) {
+  std::transform(type_.begin(), type_.end(), type_.begin(), ::toupper);
+  ParameterList& pList = const_cast<ParameterList&>(paramList);
 
-    A_ = Factory::Get< RCP<Matrix> >(currentLevel, "A");
-    SetupStratimikos(currentLevel);
-    SmootherPrototype::IsSetup(true);
-    this->GetOStream(Statistics1) << description() << std::endl;
+  if (pList.isParameter("smoother: recurMgOnFilteredA")) {
+    recurMgOnFilteredA_ = true;
+    pList.remove("smoother: recurMgOnFilteredA");
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::SetupStratimikos(Level& currentLevel) {
-
-    RCP<const Thyra::LinearOpBase<Scalar> > thyraA;
-    if (recurMgOnFilteredA_) {
-      RCP<Matrix> filteredA;
-      ExperimentalDropVertConnections(filteredA, currentLevel);
-      thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(filteredA)->getCrsMatrix());
-    }
-    else thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A_)->getCrsMatrix());
-
-    // Build Stratimikos solver
-    Stratimikos::DefaultLinearSolverBuilder linearSolverBuilder;
-    if (recurMgOnFilteredA_) {
+  bool isSupported = type_ == "STRATIMIKOS";
+  this->declareConstructionOutcome(!isSupported, "Stratimikos does not provide the smoother '" + type_ + "'.");
+  if (isSupported)
+    SetParameterList(paramList);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::SetParameterList(const Teuchos::ParameterList& paramList) {
+  Factory::SetParameterList(paramList);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  this->Input(currentLevel, "A");
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  FactoryMonitor m(*this, "Setup Smoother", currentLevel);
+
+  A_ = Factory::Get<RCP<Matrix> >(currentLevel, "A");
+  SetupStratimikos(currentLevel);
+  SmootherPrototype::IsSetup(true);
+  this->GetOStream(Statistics1) << description() << std::endl;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::SetupStratimikos(Level& currentLevel) {
+  RCP<const Thyra::LinearOpBase<Scalar> > thyraA;
+  if (recurMgOnFilteredA_) {
+    RCP<Matrix> filteredA;
+    ExperimentalDropVertConnections(filteredA, currentLevel);
+    thyraA = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(filteredA)->getCrsMatrix());
+  } else
+    thyraA = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(A_)->getCrsMatrix());
+
+  // Build Stratimikos solver
+  Stratimikos::DefaultLinearSolverBuilder linearSolverBuilder;
+  if (recurMgOnFilteredA_) {
 #ifdef MUELU_RECURMG
-       Stratimikos::enableMueLu<LocalOrdinal,GlobalOrdinal,Node>(linearSolverBuilder);
+    Stratimikos::enableMueLu<LocalOrdinal, GlobalOrdinal, Node>(linearSolverBuilder);
 #else
-    TEUCHOS_TEST_FOR_EXCEPTION( true , Exceptions::RuntimeError, "MueLu::StratimikosSmoother:: must compile with MUELU_RECURMG defined. Unfortunately, cmake does not always produce a proper link.txt file (which sometimes requires libmuelu.a before and after libmuelu-interface.a). After configuring, run script muelu/utils/misc/patchLinkForRecurMG to change link.txt files manually. If you want to create test example, add -DMUELU_RECURMG=ON to cmake arguments.");
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::StratimikosSmoother:: must compile with MUELU_RECURMG defined. Unfortunately, cmake does not always produce a proper link.txt file (which sometimes requires libmuelu.a before and after libmuelu-interface.a). After configuring, run script muelu/utils/misc/patchLinkForRecurMG to change link.txt files manually. If you want to create test example, add -DMUELU_RECURMG=ON to cmake arguments.");
 #endif
-    }
-
-    linearSolverBuilder.setParameterList(rcpFromRef(const_cast<ParameterList&>(this->GetParameterList())));
+  }
 
+  linearSolverBuilder.setParameterList(rcpFromRef(const_cast<ParameterList&>(this->GetParameterList())));
 
-    // Build a new "solver factory" according to the previously specified parameter list.
-    RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
-    solver_ = Thyra::linearOpWithSolve(*solverFactory, thyraA);
+  // Build a new "solver factory" according to the previously specified parameter list.
+  RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
+  solver_                                                         = Thyra::linearOpWithSolve(*solverFactory, thyraA);
 #ifdef dumpOutRecurMGDebug
-char mystring[120];
-sprintf(mystring,"for i in A_[0123456789].m P_[0123456789].m;  do T=Xecho $i | sed Xs/.m$/%d.m/XX; mv $i $T; done", (int) currentLevel.GetLevelID()); fflush(stdout);
-mystring[50]='`'; mystring[65]='"'; mystring[76]='"'; mystring[77]='`';
-system(mystring);
+  char mystring[120];
+  sprintf(mystring, "for i in A_[0123456789].m P_[0123456789].m;  do T=Xecho $i | sed Xs/.m$/%d.m/XX; mv $i $T; done", (int)currentLevel.GetLevelID());
+  fflush(stdout);
+  mystring[50] = '`';
+  mystring[65] = '"';
+  mystring[76] = '"';
+  mystring[77] = '`';
+  system(mystring);
 #endif
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::ExperimentalDropVertConnections(RCP<Matrix> & filteredA, Level& currentLevel) {
-
-    // strip out the veritcal connections.
-    //
-    // There is some code, which is currently turned off (via sumDropped). That
-    // makes things more complicated. I want to keep it for now, and so it is
-    // explained here. The basic idea is to try and maintain the character
-    // of the horizontal stencil by summing dropped entries appropriately.
-    // However, this does not correspond to plane relexation, and so I am
-    // not sure it is really justified. Anyway, the picture below
-    // gives a situation with a 15-pt stencil
-    //
-    //                              a
-    //                             /
-    //                            /
-    //                     b ----c----- d
-    //                          /
-    //                         /
-    //                        e
-    //                              f         dropped a & l summed into f
-    //                             /          dropped b & m summed into g
-    //                            /           dropped c & n summed into i
-    //                     g ----i----- j     dropped d & o summed into j
-    //                          /             dropped e & p summed into k
-    //                         /
-    //                        k
-    //                              l
-    //                             /
-    //                            /
-    //                     m ----n----- o
-    //                          /
-    //                         /
-    //                        p
-    // To do this, we use umap to record locations within the middle layer associated
-    // with each line ID (e.g. g corresponds to the 13th line). Then, when dropping (in a 2nd pass) we
-    // use lineId and umap to find where dropped entries should be summed (e.g., b corresponds to the
-    // 13th line and umap[13] points at location for g).
-    // using TST = typename Teuchos::ScalarTraits<SC>;
-
-    bool sumDropped = false;
-
-    LO dofsPerNode = A_->GetFixedBlockSize();
-
-
-    RCP<ParameterList> fillCompleteParams(new ParameterList);    // This code taken from Build method
-    fillCompleteParams->set("No Nonlocal Changes", true);        // within MueLu_FilteredAFactory_def
-    filteredA = MatrixFactory::Build(A_->getCrsGraph());
-    filteredA->resumeFill();
-
-    ArrayView<const LocalOrdinal> inds;
-    ArrayView<const Scalar>       valsA;
-    ArrayView<Scalar>             vals;
-    Teuchos::ArrayRCP<LocalOrdinal> TVertLineId = Factory::Get< Teuchos::ArrayRCP<LocalOrdinal> > (currentLevel, "LineDetection_VertLineIds");
-    Teuchos::ArrayRCP<LocalOrdinal> TLayerId    = Factory::Get< Teuchos::ArrayRCP<LocalOrdinal> > (currentLevel, "LineDetection_Layers");
-    LocalOrdinal* VertLineId = TVertLineId.getRawPtr();
-    LocalOrdinal* LayerId    = TLayerId.getRawPtr();
-    TEUCHOS_TEST_FOR_EXCEPTION( (LayerId == NULL) || (VertLineId == NULL), Exceptions::RuntimeError, "MueLu::StratimikosSmoother:: no line information found on this level. Cannot use recurMgOnFilteredA on this level.");
-
-    Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    for (size_t i = 0; i < A_->getRowMap()->getLocalNumElements(); i++) {
-      A_->getLocalRowView(i, inds, valsA);
-      size_t nnz = inds.size();
-      ArrayView<const Scalar> vals1;
-      filteredA->getLocalRowView(i, inds, vals1);
-      vals = ArrayView<Scalar>(const_cast<Scalar*>(vals1.getRawPtr()), nnz);
-      memcpy(vals.getRawPtr(), valsA.getRawPtr(), nnz*sizeof(Scalar));
-      size_t inode, jdof, jnode, jdof_offset;
-      inode = i/dofsPerNode;
-
-      std::unordered_map<LocalOrdinal, LocalOrdinal> umap; // umap[k] indicates where the dropped entry
-                                                           // corresponding to kth line should be added
-                                                           // within the row. See comments above.
-      if (sumDropped) {
-        for (size_t j = 0; j < nnz; j++) {
-          jdof  = inds[j];
-          jnode = jdof/dofsPerNode;
-          jdof_offset = jdof - jnode*dofsPerNode;
-          if ( LayerId[jnode]  == LayerId[inode] ) umap[dofsPerNode*VertLineId[jnode]+jdof_offset] = j;
-        }
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::ExperimentalDropVertConnections(RCP<Matrix>& filteredA, Level& currentLevel) {
+  // strip out the veritcal connections.
+  //
+  // There is some code, which is currently turned off (via sumDropped). That
+  // makes things more complicated. I want to keep it for now, and so it is
+  // explained here. The basic idea is to try and maintain the character
+  // of the horizontal stencil by summing dropped entries appropriately.
+  // However, this does not correspond to plane relexation, and so I am
+  // not sure it is really justified. Anyway, the picture below
+  // gives a situation with a 15-pt stencil
+  //
+  //                              a
+  //                             /
+  //                            /
+  //                     b ----c----- d
+  //                          /
+  //                         /
+  //                        e
+  //                              f         dropped a & l summed into f
+  //                             /          dropped b & m summed into g
+  //                            /           dropped c & n summed into i
+  //                     g ----i----- j     dropped d & o summed into j
+  //                          /             dropped e & p summed into k
+  //                         /
+  //                        k
+  //                              l
+  //                             /
+  //                            /
+  //                     m ----n----- o
+  //                          /
+  //                         /
+  //                        p
+  // To do this, we use umap to record locations within the middle layer associated
+  // with each line ID (e.g. g corresponds to the 13th line). Then, when dropping (in a 2nd pass) we
+  // use lineId and umap to find where dropped entries should be summed (e.g., b corresponds to the
+  // 13th line and umap[13] points at location for g).
+  // using TST = typename Teuchos::ScalarTraits<SC>;
+
+  bool sumDropped = false;
+
+  LO dofsPerNode = A_->GetFixedBlockSize();
+
+  RCP<ParameterList> fillCompleteParams(new ParameterList);  // This code taken from Build method
+  fillCompleteParams->set("No Nonlocal Changes", true);      // within MueLu_FilteredAFactory_def
+  filteredA = MatrixFactory::Build(A_->getCrsGraph());
+  filteredA->resumeFill();
+
+  ArrayView<const LocalOrdinal> inds;
+  ArrayView<const Scalar> valsA;
+  ArrayView<Scalar> vals;
+  Teuchos::ArrayRCP<LocalOrdinal> TVertLineId = Factory::Get<Teuchos::ArrayRCP<LocalOrdinal> >(currentLevel, "LineDetection_VertLineIds");
+  Teuchos::ArrayRCP<LocalOrdinal> TLayerId    = Factory::Get<Teuchos::ArrayRCP<LocalOrdinal> >(currentLevel, "LineDetection_Layers");
+  LocalOrdinal* VertLineId                    = TVertLineId.getRawPtr();
+  LocalOrdinal* LayerId                       = TLayerId.getRawPtr();
+  TEUCHOS_TEST_FOR_EXCEPTION((LayerId == NULL) || (VertLineId == NULL), Exceptions::RuntimeError, "MueLu::StratimikosSmoother:: no line information found on this level. Cannot use recurMgOnFilteredA on this level.");
+
+  Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
+  for (size_t i = 0; i < A_->getRowMap()->getLocalNumElements(); i++) {
+    A_->getLocalRowView(i, inds, valsA);
+    size_t nnz = inds.size();
+    ArrayView<const Scalar> vals1;
+    filteredA->getLocalRowView(i, inds, vals1);
+    vals = ArrayView<Scalar>(const_cast<Scalar*>(vals1.getRawPtr()), nnz);
+    memcpy(vals.getRawPtr(), valsA.getRawPtr(), nnz * sizeof(Scalar));
+    size_t inode, jdof, jnode, jdof_offset;
+    inode = i / dofsPerNode;
+
+    std::unordered_map<LocalOrdinal, LocalOrdinal> umap;  // umap[k] indicates where the dropped entry
+                                                          // corresponding to kth line should be added
+                                                          // within the row. See comments above.
+    if (sumDropped) {
+      for (size_t j = 0; j < nnz; j++) {
+        jdof        = inds[j];
+        jnode       = jdof / dofsPerNode;
+        jdof_offset = jdof - jnode * dofsPerNode;
+        if (LayerId[jnode] == LayerId[inode]) umap[dofsPerNode * VertLineId[jnode] + jdof_offset] = j;
       }
+    }
 
-      // drop non-middle layer entries. When sumDropped=true,  sum dropped entries to corresponding mid-layer entry
-      for (size_t j = 0; j < nnz; j++) {
-        jdof  = inds[j];
-        jnode = jdof/dofsPerNode;
-        jdof_offset = jdof - jnode*dofsPerNode;
-        if ( LayerId[jnode]  != LayerId[inode] ) {
-          if (sumDropped) {
-            if (umap.find(dofsPerNode*VertLineId[jnode + jdof_offset]) != umap.end())
-              vals[umap[dofsPerNode*VertLineId[jnode + jdof_offset]]] +=  vals[j];
-          }
-          vals[j] = ZERO;
+    // drop non-middle layer entries. When sumDropped=true,  sum dropped entries to corresponding mid-layer entry
+    for (size_t j = 0; j < nnz; j++) {
+      jdof        = inds[j];
+      jnode       = jdof / dofsPerNode;
+      jdof_offset = jdof - jnode * dofsPerNode;
+      if (LayerId[jnode] != LayerId[inode]) {
+        if (sumDropped) {
+          if (umap.find(dofsPerNode * VertLineId[jnode + jdof_offset]) != umap.end())
+            vals[umap[dofsPerNode * VertLineId[jnode + jdof_offset]]] += vals[j];
         }
+        vals[j] = ZERO;
       }
     }
-    filteredA->fillComplete(fillCompleteParams);
-
   }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::StratimikosSmoother::Apply(): Setup() has not been called");
-
-    // Apply
-    if (InitialGuessIsZero) {
-      X.putScalar(0.0);
-      RCP<      Thyra::MultiVectorBase<Scalar> > thyraX = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(rcpFromRef(X)));
-      RCP<const Thyra::MultiVectorBase<Scalar> > thyraB = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(rcpFromRef(B));
-      Thyra::SolveStatus<Scalar> status = Thyra::solve<Scalar>(*solver_, Thyra::NOTRANS, *thyraB, thyraX.ptr());
-      RCP<MultiVector> thyXpX = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(thyraX, X.getMap()->getComm());
-      X = *thyXpX;
-
-    } else {
-      typedef Teuchos::ScalarTraits<Scalar> TST;
-      RCP<MultiVector> Residual   = Utilities::Residual(*A_, X, B);
-
-      RCP<MultiVector> Cor = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(X.getMap(),X.getNumVectors(), true);
-      RCP<      Thyra::MultiVectorBase<Scalar> > thyraCor = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(Cor));
-      RCP<const Thyra::MultiVectorBase<Scalar> > thyraRes = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraMultiVector(Residual);
-      Thyra::SolveStatus<Scalar> status = Thyra::solve<Scalar>(*solver_, Thyra::NOTRANS, *thyraRes, thyraCor.ptr());
-      RCP<MultiVector> thyXpCor = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toXpetra(thyraCor, X.getMap()->getComm());
-      X.update(TST::one(), *thyXpCor, TST::one());
-    }
-
+  filteredA->fillComplete(fillCompleteParams);
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::StratimikosSmoother::Apply(): Setup() has not been called");
+
+  // Apply
+  if (InitialGuessIsZero) {
+    X.putScalar(0.0);
+    RCP<Thyra::MultiVectorBase<Scalar> > thyraX       = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(rcpFromRef(X)));
+    RCP<const Thyra::MultiVectorBase<Scalar> > thyraB = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(rcpFromRef(B));
+    Thyra::SolveStatus<Scalar> status                 = Thyra::solve<Scalar>(*solver_, Thyra::NOTRANS, *thyraB, thyraX.ptr());
+    RCP<MultiVector> thyXpX                           = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(thyraX, X.getMap()->getComm());
+    X                                                 = *thyXpX;
+
+  } else {
+    typedef Teuchos::ScalarTraits<Scalar> TST;
+    RCP<MultiVector> Residual = Utilities::Residual(*A_, X, B);
+
+    RCP<MultiVector> Cor                                = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(X.getMap(), X.getNumVectors(), true);
+    RCP<Thyra::MultiVectorBase<Scalar> > thyraCor       = Teuchos::rcp_const_cast<Thyra::MultiVectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(Cor));
+    RCP<const Thyra::MultiVectorBase<Scalar> > thyraRes = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraMultiVector(Residual);
+    Thyra::SolveStatus<Scalar> status                   = Thyra::solve<Scalar>(*solver_, Thyra::NOTRANS, *thyraRes, thyraCor.ptr());
+    RCP<MultiVector> thyXpCor                           = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toXpetra(thyraCor, X.getMap()->getComm());
+    X.update(TST::one(), *thyXpCor, TST::one());
   }
-
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<double, LocalOrdinal, GlobalOrdinal, Node> > StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<StratimikosSmoother> smoother = rcp(new StratimikosSmoother(*this) );
-    smoother->SetParameterList(this->GetParameterList());
-    return smoother;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<double, LocalOrdinal, GlobalOrdinal, Node> > StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<StratimikosSmoother> smoother = rcp(new StratimikosSmoother(*this));
+  smoother->SetParameterList(this->GetParameterList());
+  return smoother;
+}
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (SmootherPrototype::IsSetup()) {
+    out << solver_->description();
+  } else {
+    out << "STRATIMIKOS {type = " << type_ << "}";
   }
+  return out.str();
+}
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (SmootherPrototype::IsSetup()) {
-      out << solver_->description();
-    } else {
-      out << "STRATIMIKOS {type = " << type_ << "}";
-    }
-    return out.str();
-  }
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream &out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
+  if (verbLevel & Parameters1) {
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
+  }
 
-    if (verbLevel & Parameters1) {
-      out0 << "Parameter list: " << std::endl;
+  if (verbLevel & External)
+    if (solver_ != Teuchos::null) {
       Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
+      out << *solver_ << std::endl;
     }
 
-    if (verbLevel & External)
-      if (solver_ != Teuchos::null) {
-        Teuchos::OSTab tab2(out);
-        out << *solver_ << std::endl;
-      }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "RCP<solver_>: " << solver_ << std::endl;
-    }
-  }
-
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  size_t StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
-    return Teuchos::OrdinalTraits<size_t>::invalid();
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "RCP<solver_>: " << solver_ << std::endl;
   }
+}
 
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+size_t StratimikosSmoother<double, LocalOrdinal, GlobalOrdinal, Node>::getNodeSmootherComplexity() const {
+  return Teuchos::OrdinalTraits<size_t>::invalid();
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // HAVE_MUELU_STRATIMIKOS
-#endif // MUELU_STRATIMIKOSSMOOTHER_DEF_HPP
+#endif  // HAVE_MUELU_STRATIMIKOS
+#endif  // MUELU_STRATIMIKOSSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_decl.hpp b/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_decl.hpp
index 45203e00a546..7391f345d697 100644
--- a/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_decl.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_decl.hpp
@@ -48,7 +48,6 @@
 
 #include <Teuchos_ParameterList.hpp>
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_TrilinosSmoother_fwd.hpp"
 #include "MueLu_SmootherPrototype.hpp"
@@ -75,139 +74,136 @@
 
 namespace MueLu {
 
-  /*!
-    @class TrilinosSmoother
-    @ingroup MueLuSmootherClasses
-    @brief Class that encapsulates external library smoothers.
+/*!
+  @class TrilinosSmoother
+  @ingroup MueLuSmootherClasses
+  @brief Class that encapsulates external library smoothers.
 
-    Autoselection of Ifpack or Ifpack2 according to the underlying linear algebra library.
-  */
+  Autoselection of Ifpack or Ifpack2 according to the underlying linear algebra library.
+*/
 
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-            class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-            class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-            class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-  class TrilinosSmoother : public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class TrilinosSmoother : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_TRILINOSSMOOTHER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    //! @name Constructors / destructors
-    //@{
+  /*! @brief Constructor
 
-    /*! @brief Constructor
+       @param[in] type Smoother type.  Can currently be
+         - RELAXATION for (symmetric) Gauss-Seidel or Jacobi smoothing
+         - CHEBYSHEV for Chebyshev polynomial smoothing
+         - ILU for incomplete LU smoothing
 
-         @param[in] type Smoother type.  Can currently be
-           - RELAXATION for (symmetric) Gauss-Seidel or Jacobi smoothing
-           - CHEBYSHEV for Chebyshev polynomial smoothing
-           - ILU for incomplete LU smoothing
-
-         @param[in] paramList  A list holding parameters understood by either Ifpack or Ifpack2 to guide the smoother
-         setup.  See MueLu::Ifpack2Smoother for summary of the most commonly used parameters. See the Ifpack/Ifpack2
-         documentation for the full set.
-    */
+       @param[in] paramList  A list holding parameters understood by either Ifpack or Ifpack2 to guide the smoother
+       setup.  See MueLu::Ifpack2Smoother for summary of the most commonly used parameters. See the Ifpack/Ifpack2
+       documentation for the full set.
+  */
 
-    TrilinosSmoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LO& overlap = 0);
+  TrilinosSmoother(const std::string& type = "", const Teuchos::ParameterList& paramList = Teuchos::ParameterList(), const LO& overlap = 0);
 
-    //! Destructor
-    virtual ~TrilinosSmoother() { }
+  //! Destructor
+  virtual ~TrilinosSmoother() {}
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& currentLevel) const;
+  void DeclareInput(Level& currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Setup and Apply methods.
-    //@{
+  //! @name Setup and Apply methods.
+  //@{
 
-    //! TrilinosSmoother cannot be turned into a smoother using Setup(). Setup() always returns a RuntimeError exception.
-    void Setup(Level& currentLevel);
+  //! TrilinosSmoother cannot be turned into a smoother using Setup(). Setup() always returns a RuntimeError exception.
+  void Setup(Level& currentLevel);
 
-    //! TrilinosSmoother cannot be applied. Apply() always returns a RuntimeError exception.
-    void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
+  //! TrilinosSmoother cannot be applied. Apply() always returns a RuntimeError exception.
+  void Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero = false) const;
 
-    //@}
+  //@}
 
-    //! Custom SetFactory
-    void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
+  //! Custom SetFactory
+  void SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory);
 
-    //! When this prototype is cloned using Copy(), the clone is an Ifpack or an Ifpack2 smoother.
-    RCP<SmootherPrototype> Copy() const;
+  //! When this prototype is cloned using Copy(), the clone is an Ifpack or an Ifpack2 smoother.
+  RCP<SmootherPrototype> Copy() const;
 
 #ifndef _MSC_VER
-    //! Friend declaration required for clone() functionality
-    template<class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
-    friend class TrilinosSmoother;
+  //! Friend declaration required for clone() functionality
+  template <class Scalar2, class LocalOrdinal2, class GlobalOrdinal2, class Node2>
+  friend class TrilinosSmoother;
 #endif
 
-    //! Convert an Ifpack2 preconditioner name to Ifpack
-    // As a temporary solution.
-    // See https://software.sandia.gov/bugzilla/show_bug.cgi?id=5283#c5 for what I proposed to do
-    static std::string Ifpack2ToIfpack1Type(const std::string& type);
-
-    //! Convert an Ifpack2 parameter list to Ifpack
-    // As a temporary solution.
-    // See https://software.sandia.gov/bugzilla/show_bug.cgi?id=5283#c5 for what I proposed to do
-    /* TODO:
-      ifpackList.set("type", "Chebyshev");
-      ifpackList.set("chebyshev: degree", (int) 1);
-      ifpackList.set("chebyshev: max eigenvalue", (double) 2.0);
-      ifpackList.set("chebyshev: min eigenvalue", (double) 1.0);
-      ifpackList.set("chebyshev: zero starting solution", false);
-    */
-    static Teuchos::ParameterList Ifpack2ToIfpack1Param(const Teuchos::ParameterList& ifpack2List);
-
-    //! @name Overridden from Teuchos::Describable
-    //@{
+  //! Convert an Ifpack2 preconditioner name to Ifpack
+  // As a temporary solution.
+  // See https://software.sandia.gov/bugzilla/show_bug.cgi?id=5283#c5 for what I proposed to do
+  static std::string Ifpack2ToIfpack1Type(const std::string& type);
+
+  //! Convert an Ifpack2 parameter list to Ifpack
+  // As a temporary solution.
+  // See https://software.sandia.gov/bugzilla/show_bug.cgi?id=5283#c5 for what I proposed to do
+  /* TODO:
+    ifpackList.set("type", "Chebyshev");
+    ifpackList.set("chebyshev: degree", (int) 1);
+    ifpackList.set("chebyshev: max eigenvalue", (double) 2.0);
+    ifpackList.set("chebyshev: min eigenvalue", (double) 1.0);
+    ifpackList.set("chebyshev: zero starting solution", false);
+  */
+  static Teuchos::ParameterList Ifpack2ToIfpack1Param(const Teuchos::ParameterList& ifpack2List);
 
-    //! Return a simple one-line description of this object.
-    std::string description() const;
+  //! @name Overridden from Teuchos::Describable
+  //@{
 
-    //! Print the object with some verbosity level to an FancyOStream object.
-    //using MueLu::Describable::describe; // overloading, not hiding
-    //void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
-    void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
+  //! Return a simple one-line description of this object.
+  std::string description() const;
 
-    //! For diagnostic purposes
-    RCP<SmootherPrototype> getSmoother() {return s_;}
+  //! Print the object with some verbosity level to an FancyOStream object.
+  // using MueLu::Describable::describe; // overloading, not hiding
+  // void describe(Teuchos::FancyOStream &out, const VerbLevel verbLevel = Default) const {
+  void print(Teuchos::FancyOStream& out, const VerbLevel verbLevel = Default) const;
 
-    //! Get a rough estimate of cost per iteration
-    size_t getNodeSmootherComplexity() const {return s_->getNodeSmootherComplexity();}
+  //! For diagnostic purposes
+  RCP<SmootherPrototype> getSmoother() { return s_; }
 
-    //@}
+  //! Get a rough estimate of cost per iteration
+  size_t getNodeSmootherComplexity() const { return s_->getNodeSmootherComplexity(); }
 
-  private:
+  //@}
 
-    //! ifpack1/2-specific key phrase that denote smoother type
-    std::string type_;
+ private:
+  //! ifpack1/2-specific key phrase that denote smoother type
+  std::string type_;
 
-    //! overlap when using the smoother in additive Schwarz mode
-    LO overlap_;
+  //! overlap when using the smoother in additive Schwarz mode
+  LO overlap_;
 
-    //! A Factory
-    RCP<FactoryBase> AFact_;
+  //! A Factory
+  RCP<FactoryBase> AFact_;
 
-    //
-    // Underlying Smoother
-    //
+  //
+  // Underlying Smoother
+  //
 
-    //! Smoother
-    RCP<SmootherPrototype> sEpetra_, sTpetra_, sBelos_, sStratimikos_;
-    mutable
-      RCP<SmootherPrototype> s_;
+  //! Smoother
+  RCP<SmootherPrototype> sEpetra_, sTpetra_, sBelos_, sStratimikos_;
+  mutable RCP<SmootherPrototype> s_;
 
-    // Records for the case if something goes wrong
-    bool        triedEpetra_, triedTpetra_, triedBelos_, triedStratimikos_;
-    std::string errorEpetra_, errorTpetra_, errorBelos_, errorStratimikos_;
+  // Records for the case if something goes wrong
+  bool triedEpetra_, triedTpetra_, triedBelos_, triedStratimikos_;
+  std::string errorEpetra_, errorTpetra_, errorBelos_, errorStratimikos_;
 
-  }; // class TrilinosSmoother
+};  // class TrilinosSmoother
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TRILINOSSMOOTHER_SHORT
-#endif // MUELU_TRILINOSSMOOTHER_DECL_HPP
+#endif  // MUELU_TRILINOSSMOOTHER_DECL_HPP
diff --git a/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_def.hpp b/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_def.hpp
index 7b57a3d21082..ceca2b0dd5d3 100644
--- a/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_def.hpp
+++ b/packages/muelu/src/Smoothers/MueLu_TrilinosSmoother_def.hpp
@@ -60,323 +60,364 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TrilinosSmoother(const std::string& type, const Teuchos::ParameterList& paramListIn, const LO& overlap)
-    : type_(type), overlap_(overlap)
-  {
-    // The original idea behind all smoothers was to use prototype pattern. However, it does not fully work of the dependencies
-    // calculation. Particularly, we need to propagate DeclareInput to proper prototypes. Therefore, both TrilinosSmoother and
-    // DirectSolver do not follow the pattern exactly.
-    // The difference is that in order to propagate the calculation of dependencies, we need to call a DeclareInput on a
-    // constructed object (or objects, as we have two different code branches for Epetra and Tpetra). The only place where we
-    // could construct these objects is the constructor. Thus, we need to store RCPs, and both TrilinosSmoother and DirectSolver
-    // obtain a state: they contain RCP to smoother prototypes.
-    sEpetra_ = Teuchos::null;
-    sTpetra_ = Teuchos::null;
-    sBelos_  = Teuchos::null;
-    sStratimikos_ = Teuchos::null;
-
-    TEUCHOS_TEST_FOR_EXCEPTION(overlap_ < 0, Exceptions::RuntimeError, "Overlap parameter is negative (" << overlap << ")");
-
-    // paramList contains only parameters which are understood by Ifpack/Ifpack2
-    ParameterList paramList = paramListIn;
-
-
-    // We want TrilinosSmoother to be able to work with both Epetra and Tpetra objects, therefore we try to construct both
-    // Ifpack and Ifpack2 smoother prototypes. The construction really depends on configuration options.
-    triedEpetra_ = triedTpetra_ = triedBelos_ = triedStratimikos_ = false;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TrilinosSmoother(const std::string& type, const Teuchos::ParameterList& paramListIn, const LO& overlap)
+  : type_(type)
+  , overlap_(overlap) {
+  // The original idea behind all smoothers was to use prototype pattern. However, it does not fully work of the dependencies
+  // calculation. Particularly, we need to propagate DeclareInput to proper prototypes. Therefore, both TrilinosSmoother and
+  // DirectSolver do not follow the pattern exactly.
+  // The difference is that in order to propagate the calculation of dependencies, we need to call a DeclareInput on a
+  // constructed object (or objects, as we have two different code branches for Epetra and Tpetra). The only place where we
+  // could construct these objects is the constructor. Thus, we need to store RCPs, and both TrilinosSmoother and DirectSolver
+  // obtain a state: they contain RCP to smoother prototypes.
+  sEpetra_      = Teuchos::null;
+  sTpetra_      = Teuchos::null;
+  sBelos_       = Teuchos::null;
+  sStratimikos_ = Teuchos::null;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(overlap_ < 0, Exceptions::RuntimeError, "Overlap parameter is negative (" << overlap << ")");
+
+  // paramList contains only parameters which are understood by Ifpack/Ifpack2
+  ParameterList paramList = paramListIn;
+
+  // We want TrilinosSmoother to be able to work with both Epetra and Tpetra objects, therefore we try to construct both
+  // Ifpack and Ifpack2 smoother prototypes. The construction really depends on configuration options.
+  triedEpetra_ = triedTpetra_ = triedBelos_ = triedStratimikos_ = false;
 #if defined(HAVE_MUELU_IFPACK2)
-    try {
-      sTpetra_ = rcp(new Ifpack2Smoother(type_, paramList, overlap_));
-      if (sTpetra_.is_null())
-        errorTpetra_ = "Unable to construct Ifpack2 smoother";
-      else if (!sTpetra_->constructionSuccessful()) {
-        errorTpetra_ = sTpetra_->constructionErrorMsg();
-        sTpetra_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e) {
-      errorTpetra_ = e.what();
-    } catch (Exceptions::BadCast& e) {
-      errorTpetra_ = e.what();
-    } catch (Teuchos::Exceptions::InvalidParameterName& e) {
-      errorTpetra_ = e.what();
+  try {
+    sTpetra_ = rcp(new Ifpack2Smoother(type_, paramList, overlap_));
+    if (sTpetra_.is_null())
+      errorTpetra_ = "Unable to construct Ifpack2 smoother";
+    else if (!sTpetra_->constructionSuccessful()) {
+      errorTpetra_ = sTpetra_->constructionErrorMsg();
+      sTpetra_     = Teuchos::null;
     }
-    triedTpetra_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    errorTpetra_ = e.what();
+  } catch (Exceptions::BadCast& e) {
+    errorTpetra_ = e.what();
+  } catch (Teuchos::Exceptions::InvalidParameterName& e) {
+    errorTpetra_ = e.what();
+  }
+  triedTpetra_ = true;
 #endif
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
-    try {
-      // GetIfpackSmoother masks the template argument matching, and simply throws if template arguments are incompatible with Epetra
-      sEpetra_ = GetIfpackSmoother<SC,LO,GO,NO>(TrilinosSmoother::Ifpack2ToIfpack1Type(type_), TrilinosSmoother::Ifpack2ToIfpack1Param(paramList), overlap_);
-      if (sEpetra_.is_null())
-        errorEpetra_ = "Unable to construct Ifpack smoother";
-      else if (!sEpetra_->constructionSuccessful()) {
-        errorEpetra_ = sEpetra_->constructionErrorMsg();
-        sEpetra_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e) {
-      // IfpackSmoother throws if Scalar != double, LocalOrdinal != int, GlobalOrdinal != int
-      errorEpetra_ = e.what();
+  try {
+    // GetIfpackSmoother masks the template argument matching, and simply throws if template arguments are incompatible with Epetra
+    sEpetra_ = GetIfpackSmoother<SC, LO, GO, NO>(TrilinosSmoother::Ifpack2ToIfpack1Type(type_), TrilinosSmoother::Ifpack2ToIfpack1Param(paramList), overlap_);
+    if (sEpetra_.is_null())
+      errorEpetra_ = "Unable to construct Ifpack smoother";
+    else if (!sEpetra_->constructionSuccessful()) {
+      errorEpetra_ = sEpetra_->constructionErrorMsg();
+      sEpetra_     = Teuchos::null;
     }
-    triedEpetra_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    // IfpackSmoother throws if Scalar != double, LocalOrdinal != int, GlobalOrdinal != int
+    errorEpetra_ = e.what();
+  }
+  triedEpetra_ = true;
 #endif
 #if defined(HAVE_MUELU_BELOS)
-    try {
-      sBelos_ = rcp(new BelosSmoother(type_, paramList));
-      if (sBelos_.is_null())
-        errorBelos_ = "Unable to construct Belos smoother";
-      else if (!sBelos_->constructionSuccessful()) {
-        errorBelos_ = sBelos_->constructionErrorMsg();
-        sBelos_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e){
-      errorBelos_ = e.what();
-    } catch (Exceptions::BadCast& e) {
-      errorBelos_ = e.what();
+  try {
+    sBelos_ = rcp(new BelosSmoother(type_, paramList));
+    if (sBelos_.is_null())
+      errorBelos_ = "Unable to construct Belos smoother";
+    else if (!sBelos_->constructionSuccessful()) {
+      errorBelos_ = sBelos_->constructionErrorMsg();
+      sBelos_     = Teuchos::null;
     }
-    triedBelos_ = true;
+  } catch (Exceptions::RuntimeError& e) {
+    errorBelos_ = e.what();
+  } catch (Exceptions::BadCast& e) {
+    errorBelos_ = e.what();
+  }
+  triedBelos_ = true;
 #endif
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
-    try {
-      sStratimikos_ = rcp(new StratimikosSmoother(type_, paramList));
-      if (sStratimikos_.is_null())
-        errorStratimikos_ = "Unable to construct Stratimikos smoother";
-      else if (!sStratimikos_->constructionSuccessful()) {
-        errorStratimikos_ = sStratimikos_->constructionErrorMsg();
-        sStratimikos_ = Teuchos::null;
-      }
-    } catch (Exceptions::RuntimeError& e){
-      errorStratimikos_ = e.what();
+  try {
+    sStratimikos_ = rcp(new StratimikosSmoother(type_, paramList));
+    if (sStratimikos_.is_null())
+      errorStratimikos_ = "Unable to construct Stratimikos smoother";
+    else if (!sStratimikos_->constructionSuccessful()) {
+      errorStratimikos_ = sStratimikos_->constructionErrorMsg();
+      sStratimikos_     = Teuchos::null;
     }
-    triedStratimikos_ = true;
-#endif
-
-    // Check if we were able to construct at least one smoother. In many cases that's all we need, for instance if a user
-    // simply wants to use Tpetra only stack, never enables Ifpack, and always runs Tpetra objects.
-    TEUCHOS_TEST_FOR_EXCEPTION(!triedEpetra_ && !triedTpetra_ && !triedBelos_ && !triedStratimikos_, Exceptions::RuntimeError, "Unable to construct any smoother."
-                               "Please enable (TPETRA and IFPACK2) or (EPETRA and IFPACK) or (BELOS) or (STRATIMIKOS)");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(sEpetra_.is_null() && sTpetra_.is_null() && sBelos_.is_null() && sStratimikos_.is_null(), Exceptions::RuntimeError,
-        "Could not construct any smoother:\n"
-        << (triedEpetra_ ? "=> Failed to build an Epetra smoother due to the following exception:\n" : "=> Epetra and/or Ifpack are not enabled.\n")
-        << (triedEpetra_ ? errorEpetra_ + "\n" : "")
-        << (triedTpetra_ ? "=> Failed to build a Tpetra smoother due to the following exception:\n" : "=> Tpetra and/or Ifpack2 are not enabled.\n")
-        << (triedTpetra_ ? errorTpetra_ + "\n" : "")
-        << (triedBelos_ ? "=> Failed to build a Belos smoother due to the following exception:\n" : "=> Belos not enabled.\n")
-        << (triedBelos_ ? errorBelos_ + "\n" : "")
-        << (triedStratimikos_ ? "=> Failed to build a Stratimikos smoother due to the following exception:\n" : "=> Stratimikos not enabled.\n")
-        << (triedStratimikos_ ? errorStratimikos_ + "\n" : ""));
-
-    this->SetParameterList(paramList);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
-    // We need to propagate SetFactory to proper place
-    if (!sEpetra_.is_null()) sEpetra_->SetFactory(varName, factory);
-    if (!sTpetra_.is_null()) sTpetra_->SetFactory(varName, factory);
-    if (!sBelos_.is_null())  sBelos_->SetFactory(varName, factory);
-    if (!sStratimikos_.is_null()) sStratimikos_->SetFactory(varName, factory);
+  } catch (Exceptions::RuntimeError& e) {
+    errorStratimikos_ = e.what();
   }
+  triedStratimikos_ = true;
+#endif
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    if (!sBelos_.is_null())
-      s_ = sBelos_;
-    else if (!sStratimikos_.is_null())
-      s_ = sStratimikos_;
-    else {
-      // Decide whether we are running in Epetra or Tpetra mode
-      //
-      // Theoretically, we could make this decision in the constructor, and create only
-      // one of the smoothers. But we want to be able to reuse, so one can imagine a scenario
-      // where one first runs hierarchy with tpetra matrix, and then with epetra.
-
-      bool useTpetra = (currentLevel.lib() == Xpetra::UseTpetra);
-      s_ = (useTpetra ? sTpetra_ : sEpetra_);
-      if (s_.is_null()) {
-        if (useTpetra) {
+  // Check if we were able to construct at least one smoother. In many cases that's all we need, for instance if a user
+  // simply wants to use Tpetra only stack, never enables Ifpack, and always runs Tpetra objects.
+  TEUCHOS_TEST_FOR_EXCEPTION(!triedEpetra_ && !triedTpetra_ && !triedBelos_ && !triedStratimikos_, Exceptions::RuntimeError,
+                             "Unable to construct any smoother."
+                             "Please enable (TPETRA and IFPACK2) or (EPETRA and IFPACK) or (BELOS) or (STRATIMIKOS)");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(sEpetra_.is_null() && sTpetra_.is_null() && sBelos_.is_null() && sStratimikos_.is_null(), Exceptions::RuntimeError,
+                             "Could not construct any smoother:\n"
+                                 << (triedEpetra_ ? "=> Failed to build an Epetra smoother due to the following exception:\n" : "=> Epetra and/or Ifpack are not enabled.\n")
+                                 << (triedEpetra_ ? errorEpetra_ + "\n" : "")
+                                 << (triedTpetra_ ? "=> Failed to build a Tpetra smoother due to the following exception:\n" : "=> Tpetra and/or Ifpack2 are not enabled.\n")
+                                 << (triedTpetra_ ? errorTpetra_ + "\n" : "")
+                                 << (triedBelos_ ? "=> Failed to build a Belos smoother due to the following exception:\n" : "=> Belos not enabled.\n")
+                                 << (triedBelos_ ? errorBelos_ + "\n" : "")
+                                 << (triedStratimikos_ ? "=> Failed to build a Stratimikos smoother due to the following exception:\n" : "=> Stratimikos not enabled.\n")
+                                 << (triedStratimikos_ ? errorStratimikos_ + "\n" : ""));
+
+  this->SetParameterList(paramList);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetFactory(const std::string& varName, const RCP<const FactoryBase>& factory) {
+  // We need to propagate SetFactory to proper place
+  if (!sEpetra_.is_null()) sEpetra_->SetFactory(varName, factory);
+  if (!sTpetra_.is_null()) sTpetra_->SetFactory(varName, factory);
+  if (!sBelos_.is_null()) sBelos_->SetFactory(varName, factory);
+  if (!sStratimikos_.is_null()) sStratimikos_->SetFactory(varName, factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  if (!sBelos_.is_null())
+    s_ = sBelos_;
+  else if (!sStratimikos_.is_null())
+    s_ = sStratimikos_;
+  else {
+    // Decide whether we are running in Epetra or Tpetra mode
+    //
+    // Theoretically, we could make this decision in the constructor, and create only
+    // one of the smoothers. But we want to be able to reuse, so one can imagine a scenario
+    // where one first runs hierarchy with tpetra matrix, and then with epetra.
+
+    bool useTpetra = (currentLevel.lib() == Xpetra::UseTpetra);
+    s_             = (useTpetra ? sTpetra_ : sEpetra_);
+    if (s_.is_null()) {
+      if (useTpetra) {
 #if not defined(HAVE_MUELU_IFPACK2)
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                     "Error: running in Tpetra mode, but MueLu with Ifpack2 was disabled during the configure stage.\n"
-                                     "Please make sure that:\n"
-                                     "  - Ifpack2 is enabled (Trilinos_ENABLE_Ifpack2=ON),\n"
-                                     "  - Ifpack2 is available for MueLu to use (MueLu_ENABLE_Ifpack2=ON)\n");
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                   "Error: running in Tpetra mode, but MueLu with Ifpack2 was disabled during the configure stage.\n"
+                                   "Please make sure that:\n"
+                                   "  - Ifpack2 is enabled (Trilinos_ENABLE_Ifpack2=ON),\n"
+                                   "  - Ifpack2 is available for MueLu to use (MueLu_ENABLE_Ifpack2=ON)\n");
 #else
-          if (triedTpetra_)
-            this->GetOStream(Errors) << "Tpetra mode was disabled due to an error:\n" << errorTpetra_ << std::endl;
+        if (triedTpetra_)
+          this->GetOStream(Errors) << "Tpetra mode was disabled due to an error:\n"
+                                   << errorTpetra_ << std::endl;
 #endif
-        }
-        if (!useTpetra) {
+      }
+      if (!useTpetra) {
 #if not defined(HAVE_MUELU_IFPACK)
-          TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                     "Error: running in Epetra mode, but MueLu with Ifpack was disabled during the configure stage.\n"
-                                     "Please make sure that:\n"
-                                     "  - Ifpack is enabled (you can do that with Trilinos_ENABLE_Ifpack=ON),\n"
-                                     "  - Ifpack is available for MueLu to use (MueLu_ENABLE_Ifpack=ON)\n");
+        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                   "Error: running in Epetra mode, but MueLu with Ifpack was disabled during the configure stage.\n"
+                                   "Please make sure that:\n"
+                                   "  - Ifpack is enabled (you can do that with Trilinos_ENABLE_Ifpack=ON),\n"
+                                   "  - Ifpack is available for MueLu to use (MueLu_ENABLE_Ifpack=ON)\n");
 #else
-          if (triedEpetra_)
-            this->GetOStream(Errors) << "Epetra mode was disabled due to an error:\n" << errorEpetra_ << std::endl;
+        if (triedEpetra_)
+          this->GetOStream(Errors) << "Epetra mode was disabled due to an error:\n"
+                                   << errorEpetra_ << std::endl;
 #endif
-        }
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                                   "Smoother for " << (useTpetra ? "Tpetra" : "Epetra") << " was not constructed");
       }
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                 "Smoother for " << (useTpetra ? "Tpetra" : "Epetra") << " was not constructed");
     }
-
-    s_->DeclareInput(currentLevel);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
-    if (SmootherPrototype::IsSetup() == true)
-      this->GetOStream(Warnings0) << "MueLu::TrilinosSmoother::Setup(): Setup() has already been called" << std::endl;
-
-    int oldRank = s_->SetProcRankVerbose(this->GetProcRankVerbose());
-
-    s_->Setup(currentLevel);
-
-    s_->SetProcRankVerbose(oldRank);
-
-    SmootherPrototype::IsSetup(true);
-
-    this->SetParameterList(s_->GetParameterList());
+  s_->DeclareInput(currentLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(Level& currentLevel) {
+  if (SmootherPrototype::IsSetup() == true)
+    this->GetOStream(Warnings0) << "MueLu::TrilinosSmoother::Setup(): Setup() has already been called" << std::endl;
+
+  int oldRank = s_->SetProcRankVerbose(this->GetProcRankVerbose());
+
+  s_->Setup(currentLevel);
+
+  s_->SetProcRankVerbose(oldRank);
+
+  SmootherPrototype::IsSetup(true);
+
+  this->SetParameterList(s_->GetParameterList());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::TrilinosSmoother::Apply(): Setup() has not been called");
+
+  s_->Apply(X, B, InitialGuessIsZero);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
+  RCP<TrilinosSmoother> newSmoo = rcp(new TrilinosSmoother(type_, this->GetParameterList(), overlap_));
+
+  // We need to be quite careful with Copy
+  // We still want TrilinosSmoother to follow Prototype Pattern, so we need to hide the fact that we do have some state
+  if (!sEpetra_.is_null())
+    newSmoo->sEpetra_ = sEpetra_->Copy();
+  if (!sTpetra_.is_null())
+    newSmoo->sTpetra_ = sTpetra_->Copy();
+  if (!sBelos_.is_null())
+    newSmoo->sBelos_ = sBelos_->Copy();
+  if (!sStratimikos_.is_null())
+    newSmoo->sStratimikos_ = sStratimikos_->Copy();
+
+  // Copy the default mode
+  if (s_.get() == sBelos_.get())
+    newSmoo->s_ = newSmoo->sBelos_;
+  else if (s_.get() == sStratimikos_.get())
+    newSmoo->s_ = newSmoo->sStratimikos_;
+  else if (s_.get() == sTpetra_.get())
+    newSmoo->s_ = newSmoo->sTpetra_;
+  else
+    newSmoo->s_ = newSmoo->sEpetra_;
+  newSmoo->SetParameterList(this->GetParameterList());
+
+  return newSmoo;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2ToIfpack1Type(const std::string& type) {
+  if (type == "RELAXATION") {
+    return "point relaxation stand-alone";
+  }
+  if (type == "CHEBYSHEV") {
+    return "Chebyshev";
+  }
+  if (type == "ILUT") {
+    return "ILUT";
+  }
+  if (type == "RILUK") {
+    return "ILU";
+  }
+  if (type == "ILU") {
+    return "ILU";
+  }
+  if (type == "Amesos") {
+    return "Amesos";
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(MultiVector& X, const MultiVector& B, bool InitialGuessIsZero) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == false, Exceptions::RuntimeError, "MueLu::TrilinosSmoother::Apply(): Setup() has not been called");
+  // special types
 
-    s_->Apply(X, B, InitialGuessIsZero);
+  // Note: for Ifpack there is no distinction between block and banded relaxation as there is no
+  // BandedContainer or TridiagonalContainer.
+  if (type == "LINESMOOTHING_BLOCKRELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Copy() const {
-    RCP<TrilinosSmoother> newSmoo = rcp(new TrilinosSmoother(type_, this->GetParameterList(), overlap_));
-
-    // We need to be quite careful with Copy
-    // We still want TrilinosSmoother to follow Prototype Pattern, so we need to hide the fact that we do have some state
-    if (!sEpetra_.is_null())
-      newSmoo->sEpetra_ = sEpetra_->Copy();
-    if (!sTpetra_.is_null())
-      newSmoo->sTpetra_ = sTpetra_->Copy();
-    if (!sBelos_.is_null())
-      newSmoo->sBelos_ = sBelos_->Copy();
-    if (!sStratimikos_.is_null())
-      newSmoo->sStratimikos_ = sStratimikos_->Copy();
-
-    // Copy the default mode
-    if (s_.get() == sBelos_.get())
-      newSmoo->s_ = newSmoo->sBelos_;
-    else if (s_.get() == sStratimikos_.get())
-      newSmoo->s_ = newSmoo->sStratimikos_;
-    else if (s_.get() == sTpetra_.get())
-      newSmoo->s_ = newSmoo->sTpetra_;
-    else
-      newSmoo->s_ = newSmoo->sEpetra_;
-    newSmoo->SetParameterList(this->GetParameterList());
-
-    return newSmoo;
+  if (type == "LINESMOOTHING_BLOCK RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2ToIfpack1Type(const std::string& type) {
-    if (type == "RELAXATION") { return "point relaxation stand-alone"; }
-    if (type == "CHEBYSHEV")  { return "Chebyshev";                    }
-    if (type == "ILUT")       { return "ILUT";                         }
-    if (type == "RILUK")      { return "ILU";                          }
-    if (type == "ILU")        { return "ILU";                          }
-    if (type == "Amesos")     { return "Amesos";                       }
-
-    // special types
-
-    // Note: for Ifpack there is no distinction between block and banded relaxation as there is no
-    // BandedContainer or TridiagonalContainer.
-    if (type == "LINESMOOTHING_BLOCKRELAXATION")         { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_BLOCK RELAXATION")        { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_BLOCK_RELAXATION")        { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_BANDEDRELAXATION")        { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_BANDED RELAXATION")       { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_BANDED_RELAXATION")       { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDIRELAXATION")         { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDI RELAXATION")        { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDI_RELAXATION")        { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDIAGONALRELAXATION")   { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDIAGONAL RELAXATION")  { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "LINESMOOTHING_TRIDIAGONAL_RELAXATION")  { return "LINESMOOTHING_BLOCKRELAXATION"; }
-    if (type == "AGGREGATE")                             { return "AGGREGATE";                     }
-    if(type == "BLOCK_RELAXATION" ||
-       type == "BLOCK RELAXATION" ||
-       type == "BLOCKRELAXATION" ||
-       // Banded
-       type == "BANDED_RELAXATION" ||
-       type == "BANDED RELAXATION" ||
-       type == "BANDEDRELAXATION" ||
-       // Tridiagonal
-       type == "TRIDI_RELAXATION" ||
-       type == "TRIDI RELAXATION" ||
-       type == "TRIDIRELAXATION" ||
-       type == "TRIDIAGONAL_RELAXATION" ||
-       type == "TRIDIAGONAL RELAXATION" ||
-       type == "TRIDIAGONALRELAXATION") {
-      return "block relaxation stand-alone";
-    }
-    
-    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Cannot convert Ifpack2 preconditioner name to Ifpack: unknown type: \"" + type + "\"");
+  if (type == "LINESMOOTHING_BLOCK_RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_BANDEDRELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_BANDED RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_BANDED_RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDIRELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDI RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDI_RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDIAGONALRELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDIAGONAL RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "LINESMOOTHING_TRIDIAGONAL_RELAXATION") {
+    return "LINESMOOTHING_BLOCKRELAXATION";
+  }
+  if (type == "AGGREGATE") {
+    return "AGGREGATE";
+  }
+  if (type == "BLOCK_RELAXATION" ||
+      type == "BLOCK RELAXATION" ||
+      type == "BLOCKRELAXATION" ||
+      // Banded
+      type == "BANDED_RELAXATION" ||
+      type == "BANDED RELAXATION" ||
+      type == "BANDEDRELAXATION" ||
+      // Tridiagonal
+      type == "TRIDI_RELAXATION" ||
+      type == "TRIDI RELAXATION" ||
+      type == "TRIDIRELAXATION" ||
+      type == "TRIDIAGONAL_RELAXATION" ||
+      type == "TRIDIAGONAL RELAXATION" ||
+      type == "TRIDIAGONALRELAXATION") {
+    return "block relaxation stand-alone";
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ParameterList TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2ToIfpack1Param(const Teuchos::ParameterList& ifpack2List) {
-    Teuchos::ParameterList ifpack1List = ifpack2List;
+  TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "Cannot convert Ifpack2 preconditioner name to Ifpack: unknown type: \"" + type + "\"");
+}
 
-    if (ifpack2List.isParameter("relaxation: type") && ifpack2List.get<std::string>("relaxation: type") == "Symmetric Gauss-Seidel")
-      ifpack1List.set("relaxation: type", "symmetric Gauss-Seidel");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ParameterList TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Ifpack2ToIfpack1Param(const Teuchos::ParameterList& ifpack2List) {
+  Teuchos::ParameterList ifpack1List = ifpack2List;
 
-    if (ifpack2List.isParameter("fact: iluk level-of-fill")) {
-      ifpack1List.remove("fact: iluk level-of-fill");
-      ifpack1List.set("fact: level-of-fill", ifpack2List.get<int>("fact: iluk level-of-fill"));
-    }
+  if (ifpack2List.isParameter("relaxation: type") && ifpack2List.get<std::string>("relaxation: type") == "Symmetric Gauss-Seidel")
+    ifpack1List.set("relaxation: type", "symmetric Gauss-Seidel");
 
-    return ifpack1List;
+  if (ifpack2List.isParameter("fact: iluk level-of-fill")) {
+    ifpack1List.remove("fact: iluk level-of-fill");
+    ifpack1List.set("fact: level-of-fill", ifpack2List.get<int>("fact: iluk level-of-fill"));
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
-    std::ostringstream out;
-    if (s_ != Teuchos::null) {
-      out << s_->description();
-    } else {
-      out << SmootherPrototype::description();
-      out << "{type = " << type_ << "}";
-    }
-    return out.str();
+  return ifpack1List;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::description() const {
+  std::ostringstream out;
+  if (s_ != Teuchos::null) {
+    out << s_->description();
+  } else {
+    out << SmootherPrototype::description();
+    out << "{type = " << type_ << "}";
+  }
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
+  MUELU_DESCRIBE;
+
+  if (verbLevel & Parameters0)
+    out0 << "Prec. type: " << type_ << std::endl;
+
+  if (verbLevel & Parameters1) {
+    out0 << "PrecType: " << type_ << std::endl;
+    out0 << "Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << this->GetParameterList();
+    out0 << "Overlap: " << overlap_ << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TrilinosSmoother<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print(Teuchos::FancyOStream& out, const VerbLevel verbLevel) const {
-    MUELU_DESCRIBE;
-
-    if (verbLevel & Parameters0)
-      out0 << "Prec. type: " << type_ << std::endl;
-
-    if (verbLevel & Parameters1) {
-      out0 << "PrecType: " << type_ << std::endl;
-      out0 << "Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << this->GetParameterList();
-      out0 << "Overlap: " << overlap_ << std::endl;
-    }
-
-    if (verbLevel & Debug) {
-      out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
-           << "-" << std::endl
-           << "Epetra PrecType: " << Ifpack2ToIfpack1Type(type_) << std::endl
-           << "Epetra Parameter list: " << std::endl;
-      Teuchos::OSTab tab2(out);
-      out << Ifpack2ToIfpack1Param(this->GetParameterList());;
-    }
+  if (verbLevel & Debug) {
+    out0 << "IsSetup: " << Teuchos::toString(SmootherPrototype::IsSetup()) << std::endl
+         << "-" << std::endl
+         << "Epetra PrecType: " << Ifpack2ToIfpack1Type(type_) << std::endl
+         << "Epetra Parameter list: " << std::endl;
+    Teuchos::OSTab tab2(out);
+    out << Ifpack2ToIfpack1Param(this->GetParameterList());
+    ;
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_TRILINOSSMOOTHER_DEF_HPP
+#endif  // MUELU_TRILINOSSMOOTHER_DEF_HPP
diff --git a/packages/muelu/src/Transfers/BaseClass/MueLu_PFactory_decl.hpp b/packages/muelu/src/Transfers/BaseClass/MueLu_PFactory_decl.hpp
index 9159d5d48688..e7e9e029f65a 100644
--- a/packages/muelu/src/Transfers/BaseClass/MueLu_PFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BaseClass/MueLu_PFactory_decl.hpp
@@ -54,57 +54,55 @@
 
 namespace MueLu {
 
-  /*!
-    @class PFactory
-    @brief Factory that provides an interface for a concrete implementation of a prolongation operator.
-
-    For a concrete implementation the user has to overwrite the virtual Build method.
-  */
-
-  class PFactory : public TwoLevelFactoryBase {
+/*!
+  @class PFactory
+  @brief Factory that provides an interface for a concrete implementation of a prolongation operator.
 
-  protected:
+  For a concrete implementation the user has to overwrite the virtual Build method.
+*/
 
-    bool restrictionMode_;  //< true, if PFactory is used for generating the restriction operator
+class PFactory : public TwoLevelFactoryBase {
+ protected:
+  bool restrictionMode_;  //< true, if PFactory is used for generating the restriction operator
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    PFactory() :  restrictionMode_(false) { }
+  //! Constructor.
+  PFactory()
+    : restrictionMode_(false) {}
 
-    //! Destructor.
-    virtual ~PFactory() {}
-    //@}
+  //! Destructor.
+  virtual ~PFactory() {}
+  //@}
 
-    //! @name Build methods.
-    //@{
-
-    /*!
-      @brief Abstract Build method.
-    */
-    virtual void BuildP(Level &fineLevel, Level &coarseLevel) const = 0;
-    //@}
+  //! @name Build methods.
+  //@{
 
+  /*!
+    @brief Abstract Build method.
+  */
+  virtual void BuildP(Level &fineLevel, Level &coarseLevel) const = 0;
+  //@}
 
-    //! @name Restriction mode
-    //@{
+  //! @name Restriction mode
+  //@{
 
-    /// switch prolongator factory to restriction mode
-    /// if set to true, the prolongation factory generates a restriciton operator instead of a prolongation operator
-    void setRestrictionMode(bool bRestrictionMode = false) {
-      restrictionMode_ = bRestrictionMode;
-    }
+  /// switch prolongator factory to restriction mode
+  /// if set to true, the prolongation factory generates a restriciton operator instead of a prolongation operator
+  void setRestrictionMode(bool bRestrictionMode = false) {
+    restrictionMode_ = bRestrictionMode;
+  }
 
-    /// returns restrictionMode flag
-    bool isRestrictionModeSet() { return restrictionMode_; }
+  /// returns restrictionMode flag
+  bool isRestrictionModeSet() { return restrictionMode_; }
 
-    //@}
+  //@}
 
-  }; //class PFactory
+};  // class PFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PFACTORY_SHORT
-#endif // MUELU_PFACTORY_DEF_HPP
+#endif  // MUELU_PFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_decl.hpp b/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_decl.hpp
index f649b8759458..51edbfbf61f3 100644
--- a/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_decl.hpp
@@ -58,10 +58,10 @@
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLuTests {
-  // Forward declaration of friend tester class used to UnitTest BlackBoxPFactory
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlackBoxPFactoryTester;
-}
+// Forward declaration of friend tester class used to UnitTest BlackBoxPFactory
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlackBoxPFactoryTester;
+}  // namespace MueLuTests
 
 namespace MueLu {
 
@@ -113,110 +113,108 @@ namespace MueLu {
   |                   |                          | coordinates so that on coarser levels coordinates are available in case another factory needs them.              |
 
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class BlackBoxPFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class BlackBoxPFactory : public PFactory {
 #undef MUELU_BLACKBOXPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  friend class MueLuTests::BlackBoxPFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    friend class MueLuTests::BlackBoxPFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  BlackBoxPFactory() {}
 
-    //! Constructor
-    BlackBoxPFactory() { }
+  //! Destructor.
+  virtual ~BlackBoxPFactory() {}
+  //@}
 
-    //! Destructor.
-    virtual ~BlackBoxPFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+ private:
+  struct NodesIDs {
+    // This small struct just carries basic data associated with coarse nodes that is needed
+    // to compute colMapP and to fillComplete P,
 
-  private:
+    Array<GO> GIDs, coarseGIDs;
+    Array<int> PIDs;
+    Array<LO> LIDs;
+    std::vector<GO> colInds;
+  };
 
-    struct NodesIDs {
-      // This small struct just carries basic data associated with coarse nodes that is needed
-      // to compute colMapP and to fillComplete P,
+  struct NodeID {
+    // This small struct is similar to the one above but only for one node.
+    // It is used to create a vector of NodeID that can easily be sorted
 
-      Array<GO>  GIDs, coarseGIDs;
-      Array<int> PIDs;
-      Array<LO>  LIDs;
-      std::vector<GO> colInds;
-    };
+    GO GID;
+    int PID;
+    LO LID, lexiInd;
+  };
 
-    struct NodeID {
-      // This small struct is similar to the one above but only for one node.
-      // It is used to create a vector of NodeID that can easily be sorted
+  void GetGeometricData(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coordinates,
+                        const Array<LO> coarseRate, const Array<GO> gFineNodesPerDir,
+                        const Array<LO> lFineNodesPerDir, const LO BlkSize, Array<GO>& gIndices,
+                        Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
+                        Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir,
+                        Array<LO>& glCoarseNodesPerDir, Array<GO>& ghostGIDs,
+                        Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs, GO& gNumCoarseNodes,
+                        LO& lNumCoarseNodes, ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes,
+                        Array<int>& boundaryFlags, RCP<NodesIDs> ghostedCoarseNodes) const;
 
-      GO  GID;
-      int PID;
-      LO  LID, lexiInd;
-    };
+  void ComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
+                           const Array<LO> endRate, const LO BlkSize, const Array<LO> elemInds,
+                           const Array<LO> lCoarseElementsPerDir,
+                           const LO numDimensions, const Array<LO> lFineNodesPerDir,
+                           const Array<GO> gFineNodesPerDir, const Array<GO> gIndices,
+                           const Array<LO> lCoarseNodesPerDir, const Array<bool> ghostInterface,
+                           const Array<int> elementFlags, const std::string stencilType,
+                           const std::string blockStrategy, const Array<LO> elementNodesPerDir,
+                           const LO numNodesInElement, const Array<GO> colGIDs,
+                           Teuchos::SerialDenseMatrix<LO, SC>& Pi,
+                           Teuchos::SerialDenseMatrix<LO, SC>& Pf,
+                           Teuchos::SerialDenseMatrix<LO, SC>& Pe,
+                           Array<LO>& dofType, Array<LO>& lDofInd) const;
 
-    void GetGeometricData(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> >& coordinates,
-                          const Array<LO> coarseRate, const Array<GO> gFineNodesPerDir,
-                          const Array<LO> lFineNodesPerDir, const LO BlkSize, Array<GO>& gIndices,
-                          Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
-                          Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir,
-                          Array<LO>& glCoarseNodesPerDir, Array<GO>& ghostGIDs,
-                          Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs, GO& gNumCoarseNodes,
-                          LO& lNumCoarseNodes, ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes,
-                          Array<int>& boundaryFlags, RCP<NodesIDs> ghostedCoarseNodes) const;
+  void CollapseStencil(const int type, const int orientation, const int collapseFlags[3],
+                       Array<SC>& stencil) const;
 
-    void ComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
-                             const Array<LO> endRate, const LO BlkSize, const Array<LO> elemInds,
-                             const Array<LO> lCoarseElementsPerDir,
-                             const LO numDimensions, const Array<LO> lFineNodesPerDir,
-                             const Array<GO> gFineNodesPerDir, const Array<GO> gIndices,
-                             const Array<LO> lCoarseNodesPerDir, const Array<bool> ghostInterface,
-                             const Array<int> elementFlags, const std::string stencilType,
-                             const std::string blockStrategy, const Array<LO> elementNodesPerDir,
-                             const LO numNodesInElement, const Array<GO> colGIDs,
-                             Teuchos::SerialDenseMatrix<LO,SC>& Pi,
-                             Teuchos::SerialDenseMatrix<LO,SC>& Pf,
-                             Teuchos::SerialDenseMatrix<LO,SC>& Pe,
-                             Array<LO>& dofType, Array<LO>& lDofInd) const;
+  void FormatStencil(const LO BlkSize, const Array<bool> ghostInterface, const LO ie,
+                     const LO je, const LO ke, const ArrayView<const SC> rowValues,
+                     const Array<LO> elementNodesPerDir, const int collapseFlags[3],
+                     const std::string stencilType, Array<SC>& stencil) const;
 
-    void CollapseStencil(const int type, const int orientation, const int collapseFlags[3],
-                         Array<SC>& stencil) const ;
+  void GetNodeInfo(const LO ie, const LO je, const LO ke, const Array<LO> elementNodesPerDir,
+                   int* type, LO& ind, int* orientation) const;
 
-    void FormatStencil(const LO BlkSize, const Array<bool> ghostInterface, const LO ie,
-                       const LO je, const LO ke, const ArrayView<const SC> rowValues,
-                       const Array<LO> elementNodesPerDir, const int collapseFlags[3],
-                       const std::string stencilType, Array<SC>& stencil) const;
+  void sh_sort_permute(
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
 
-    void GetNodeInfo(const LO ie, const LO je, const LO ke, const Array<LO> elementNodesPerDir,
-                     int* type, LO& ind, int* orientation) const;
+};  // class BlackBoxPFactory
 
-    void sh_sort_permute(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
-
-  }; //class BlackBoxPFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLACKBOXPFACTORY_SHORT
-#endif // MUELU_BLACKBOXPFACTORY_DECL_HPP
+#endif  // MUELU_BLACKBOXPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_def.hpp b/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_def.hpp
index ada59b09a552..5bc141fc40bb 100644
--- a/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlackBox/MueLu_BlackBoxPFactory_def.hpp
@@ -49,7 +49,6 @@
 #include <stdlib.h>
 #include <iomanip>
 
-
 // #include <Teuchos_LAPACK.hpp>
 #include <Teuchos_SerialDenseMatrix.hpp>
 #include <Teuchos_SerialDenseVector.hpp>
@@ -71,1901 +70,1852 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    // Coarsen can come in two forms, either a single char that will be interpreted as an integer
-    // which is used as the coarsening rate in every spatial dimentions,
-    // or it can be a longer string that will then be interpreted as an array of integers.
-    // By default coarsen is set as "{2}", hence a coarsening rate of 2 in every spatial dimension
-    // is the default setting!
-    validParamList->set<std::string >           ("Coarsen",                 "{3}", "Coarsening rate in all spatial dimensions");
-    validParamList->set<RCP<const FactoryBase> >("A",                         Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",                 Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",               Teuchos::null, "Generating factory for coorindates");
-    validParamList->set<RCP<const FactoryBase> >("gNodesPerDim",              Teuchos::null, "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",              Teuchos::null, "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-    validParamList->set<std::string>            ("stencil type",            "full", "You can use two type of stencils: full and reduced, that correspond to 27 and 7 points stencils respectively in 3D.");
-    validParamList->set<std::string>            ("block strategy",          "coupled", "The strategy used to handle systems of PDEs can be: coupled or uncoupled.");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel,
-                                                                                 Level& /* coarseLevel */)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  // Coarsen can come in two forms, either a single char that will be interpreted as an integer
+  // which is used as the coarsening rate in every spatial dimentions,
+  // or it can be a longer string that will then be interpreted as an array of integers.
+  // By default coarsen is set as "{2}", hence a coarsening rate of 2 in every spatial dimension
+  // is the default setting!
+  validParamList->set<std::string>("Coarsen", "{3}", "Coarsening rate in all spatial dimensions");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for coorindates");
+  validParamList->set<RCP<const FactoryBase> >("gNodesPerDim", Teuchos::null, "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null, "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+  validParamList->set<std::string>("stencil type", "full", "You can use two type of stencils: full and reduced, that correspond to 27 and 7 points stencils respectively in 3D.");
+  validParamList->set<std::string>("block strategy", "coupled", "The strategy used to handle systems of PDEs can be: coupled or uncoupled.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel,
+                                                                               Level& /* coarseLevel */)
     const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "Coordinates");
-    // Request the global number of nodes per dimensions
-    if(fineLevel.GetLevelID() == 0) {
-      if(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
-        fineLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "gNodesPerDim was not provided by the user on level0!");
-      }
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "Coordinates");
+  // Request the global number of nodes per dimensions
+  if (fineLevel.GetLevelID() == 0) {
+    if (fineLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
+      fineLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
     } else {
-      Input(fineLevel, "gNodesPerDim");
+      TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "gNodesPerDim was not provided by the user on level0!");
     }
+  } else {
+    Input(fineLevel, "gNodesPerDim");
+  }
 
-    // Request the local number of nodes per dimensions
-    if(fineLevel.GetLevelID() == 0) {
-      if(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        fineLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
-      }
+  // Request the local number of nodes per dimensions
+  if (fineLevel.GetLevelID() == 0) {
+    if (fineLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      fineLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
     } else {
-      Input(fineLevel, "lNodesPerDim");
+      TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
     }
+  } else {
+    Input(fineLevel, "lNodesPerDim");
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
-                                                                          Level& coarseLevel) const{
-    return BuildP(fineLevel, coarseLevel);
-
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
+                                                                        Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
+                                                                         Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  // Get parameter list
+  const ParameterList& pL = GetParameterList();
+
+  // obtain general variables
+  RCP<Matrix> A                  = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<MultiVector> fineNullspace = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates =
+      Get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > >(fineLevel, "Coordinates");
+  LO numDimensions = coordinates->getNumVectors();
+  LO BlkSize       = A->GetFixedBlockSize();
+
+  // Get fine level geometric data: g(l)FineNodesPerDir and g(l)NumFineNodes
+  Array<GO> gFineNodesPerDir(3);
+  Array<LO> lFineNodesPerDir(3);
+  // Get the number of points in each direction
+  if (fineLevel.GetLevelID() == 0) {
+    gFineNodesPerDir = fineLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
+    lFineNodesPerDir = fineLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
+  } else {
+    // Loading global number of nodes per diretions
+    gFineNodesPerDir = Get<Array<GO> >(fineLevel, "gNodesPerDim");
+
+    // Loading local number of nodes per diretions
+    lFineNodesPerDir = Get<Array<LO> >(fineLevel, "lNodesPerDim");
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
-                                                                           Level& coarseLevel)const{
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    // Get parameter list
-    const ParameterList& pL = GetParameterList();
-
-    // obtain general variables
-    RCP<Matrix>      A             = Get< RCP<Matrix> >      (fineLevel, "A");
-    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates =
-      Get< RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > >(fineLevel, "Coordinates");
-    LO numDimensions  = coordinates->getNumVectors();
-    LO BlkSize = A->GetFixedBlockSize();
-
-    // Get fine level geometric data: g(l)FineNodesPerDir and g(l)NumFineNodes
-    Array<GO> gFineNodesPerDir(3);
-    Array<LO> lFineNodesPerDir(3);
-    // Get the number of points in each direction
-    if(fineLevel.GetLevelID() == 0) {
-      gFineNodesPerDir = fineLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
-      lFineNodesPerDir = fineLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
-    } else {
-      // Loading global number of nodes per diretions
-      gFineNodesPerDir = Get<Array<GO> >(fineLevel, "gNodesPerDim");
-
-      // Loading local number of nodes per diretions
-      lFineNodesPerDir = Get<Array<LO> >(fineLevel, "lNodesPerDim");
+  for (LO i = 0; i < 3; ++i) {
+    if (gFineNodesPerDir[i] == 0) {
+      GetOStream(Runtime0) << "gNodesPerDim in direction " << i << " is set to 1 from 0"
+                           << std::endl;
+      gFineNodesPerDir[i] = 1;
     }
-    for(LO i = 0; i < 3; ++i) {
-      if(gFineNodesPerDir[i] == 0) {
-        GetOStream(Runtime0) << "gNodesPerDim in direction " << i << " is set to 1 from 0"
-                             << std::endl;
-        gFineNodesPerDir[i] = 1;
-      }
-      if(lFineNodesPerDir[i] == 0) {
-        GetOStream(Runtime0) << "lNodesPerDim in direction " << i << " is set to 1 from 0"
+    if (lFineNodesPerDir[i] == 0) {
+      GetOStream(Runtime0) << "lNodesPerDim in direction " << i << " is set to 1 from 0"
+                           << std::endl;
+      lFineNodesPerDir[i] = 1;
+    }
+  }
+  GO gNumFineNodes = gFineNodesPerDir[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+  LO lNumFineNodes = lFineNodesPerDir[2] * lFineNodesPerDir[1] * lFineNodesPerDir[0];
+
+  // Get the coarsening rate
+  std::string coarsenRate = pL.get<std::string>("Coarsen");
+  Array<LO> coarseRate(3);
+  {
+    Teuchos::Array<LO> crates;
+    try {
+      crates = Teuchos::fromStringToArray<LO>(coarsenRate);
+    } catch (const Teuchos::InvalidArrayStringRepresentation& e) {
+      GetOStream(Errors, -1) << " *** Coarsen must be a string convertible into an array! *** "
                              << std::endl;
-        lFineNodesPerDir[i] = 1;
-      }
+      throw e;
     }
-    GO gNumFineNodes = gFineNodesPerDir[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-    LO lNumFineNodes = lFineNodesPerDir[2]*lFineNodesPerDir[1]*lFineNodesPerDir[0];
-
-    // Get the coarsening rate
-    std::string coarsenRate = pL.get<std::string>("Coarsen");
-    Array<LO> coarseRate(3);
-    {
-      Teuchos::Array<LO> crates;
-      try {
-        crates = Teuchos::fromStringToArray<LO>(coarsenRate);
-      } catch(const Teuchos::InvalidArrayStringRepresentation& e) {
-        GetOStream(Errors,-1) << " *** Coarsen must be a string convertible into an array! *** "
-                              << std::endl;
-        throw e;
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION((crates.size() > 1) && (crates.size() < numDimensions),
-                                 Exceptions::RuntimeError,
-                                 "Coarsen must have at least as many components as the number of"
-                                 " spatial dimensions in the problem.");
-      for(LO i = 0; i < 3; ++i) {
-        if(i < numDimensions) {
-          if(crates.size() == 1) {
-            coarseRate[i] = crates[0];
-          } else if(i < crates.size()) {
-            coarseRate[i] = crates[i];
-          } else {
-            GetOStream(Errors,-1) << " *** Coarsen must be at least as long as the number of"
-              " spatial dimensions! *** " << std::endl;
-            throw Exceptions::RuntimeError(" *** Coarsen must be at least as long as the number of"
-                                           " spatial dimensions! *** \n");
-          }
+    TEUCHOS_TEST_FOR_EXCEPTION((crates.size() > 1) && (crates.size() < numDimensions),
+                               Exceptions::RuntimeError,
+                               "Coarsen must have at least as many components as the number of"
+                               " spatial dimensions in the problem.");
+    for (LO i = 0; i < 3; ++i) {
+      if (i < numDimensions) {
+        if (crates.size() == 1) {
+          coarseRate[i] = crates[0];
+        } else if (i < crates.size()) {
+          coarseRate[i] = crates[i];
         } else {
-          coarseRate[i] = 1;
+          GetOStream(Errors, -1) << " *** Coarsen must be at least as long as the number of"
+                                    " spatial dimensions! *** "
+                                 << std::endl;
+          throw Exceptions::RuntimeError(
+              " *** Coarsen must be at least as long as the number of"
+              " spatial dimensions! *** \n");
         }
+      } else {
+        coarseRate[i] = 1;
       }
-    } // End of scope for crates
-
-    // Get the stencil type used for discretization
-    const std::string stencilType = pL.get<std::string>("stencil type");
-    if(stencilType != "full" && stencilType != "reduced") {
-      GetOStream(Errors,-1) << " *** stencil type must be set to: full or reduced, any other value "
-              "is trated as an error! *** " << std::endl;
-      throw Exceptions::RuntimeError(" *** stencil type is neither full, nor reduced! *** \n");
     }
+  }  // End of scope for crates
+
+  // Get the stencil type used for discretization
+  const std::string stencilType = pL.get<std::string>("stencil type");
+  if (stencilType != "full" && stencilType != "reduced") {
+    GetOStream(Errors, -1) << " *** stencil type must be set to: full or reduced, any other value "
+                              "is trated as an error! *** "
+                           << std::endl;
+    throw Exceptions::RuntimeError(" *** stencil type is neither full, nor reduced! *** \n");
+  }
 
-    // Get the strategy for PDE systems
-    const std::string blockStrategy = pL.get<std::string>("block strategy");
-    if(blockStrategy != "coupled" && blockStrategy != "uncoupled") {
-      GetOStream(Errors,-1) << " *** block strategy must be set to: coupled or uncoupled, any other "
-              "value is trated as an error! *** " << std::endl;
-      throw Exceptions::RuntimeError(" *** block strategy is neither coupled, nor uncoupled! *** \n");
-    }
+  // Get the strategy for PDE systems
+  const std::string blockStrategy = pL.get<std::string>("block strategy");
+  if (blockStrategy != "coupled" && blockStrategy != "uncoupled") {
+    GetOStream(Errors, -1) << " *** block strategy must be set to: coupled or uncoupled, any other "
+                              "value is trated as an error! *** "
+                           << std::endl;
+    throw Exceptions::RuntimeError(" *** block strategy is neither coupled, nor uncoupled! *** \n");
+  }
 
-    GO gNumCoarseNodes = 0;
-    LO lNumCoarseNodes = 0;
-    Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
-    Array<LO> myOffset(3), lCoarseNodesPerDir(3), glCoarseNodesPerDir(3), endRate(3);
-    Array<bool> ghostInterface(6);
-    Array<int> boundaryFlags(3);
-    ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
-    Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
-    for(LO dim = 0; dim < numDimensions; ++dim) {fineNodes[dim] = coordinates->getData(dim)();}
-
-    // This struct stores PIDs, LIDs and GIDs on the fine mesh and GIDs on the coarse mesh.
-    RCP<NodesIDs> ghostedCoarseNodes = rcp(new NodesIDs{});
-
-    GetGeometricData(coordinates, coarseRate, gFineNodesPerDir, lFineNodesPerDir, BlkSize,// inputs
-                     gIndices, myOffset, ghostInterface, endRate, gCoarseNodesPerDir,     // outputs
-                     lCoarseNodesPerDir, glCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs, colGIDs,
-                     gNumCoarseNodes, lNumCoarseNodes, coarseNodes, boundaryFlags,
-                     ghostedCoarseNodes);
-
-    // Create the MultiVector of coarse coordinates
-    Xpetra::UnderlyingLib lib = coordinates->getMap()->lib();
-    RCP<const Map> coarseCoordsMap = MapFactory::Build (lib,
-                                                        gNumCoarseNodes,
-                                                        coarseNodesGIDs.view(0, lNumCoarseNodes),
-                                                        coordinates->getMap()->getIndexBase(),
-                                                        coordinates->getMap()->getComm());
-    Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseCoords(numDimensions);
-    for(LO dim = 0; dim < numDimensions; ++dim) {
-      coarseCoords[dim] = coarseNodes[dim]();
-    }
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coarseCoordinates =
-      Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(coarseCoordsMap, coarseCoords(),
-                                                         numDimensions);
-
-    // Now create a new matrix: Aghost that contains all the data
-    // locally needed to compute the local part of the prolongator.
-    // Here assuming that all the coarse nodes o and fine nodes +
-    // are local then all the data associated with the coarse
-    // nodes O and the fine nodes * needs to be imported.
-    //
-    //                  *--*--*--*--*--*--*--*
-    //                  |  |  |  |  |  |  |  |
-    //                  o--+--+--o--+--+--O--*
-    //                  |  |  |  |  |  |  |  |
-    //                  +--+--+--+--+--+--*--*
-    //                  |  |  |  |  |  |  |  |
-    //                  +--+--+--+--+--+--*--*
-    //                  |  |  |  |  |  |  |  |
-    //                  o--+--+--o--+--+--O--*
-    //                  |  |  |  |  |  |  |  |
-    //                  +--+--+--+--+--+--*--*
-    //                  |  |  |  |  |  |  |  |
-    //                  *--*--*--*--*--*--*--*
-    //                  |  |  |  |  |  |  |  |
-    //                  O--*--*--O--*--*--O--*
-    //
-    // Creating that local matrix is easy enough using proper range
-    // and domain maps to import data from A. Note that with this
-    // approach we reorder the local entries using the domain map and
-    // can subsequently compute the prolongator without reordering.
-    // As usual we need to be careful about any coarsening rate
-    // change at the boundary!
-
-    // The ingredients needed are an importer, a range map and a domain map
-    Array<GO> ghostRowGIDs, ghostColGIDs, nodeSteps(3);
-    nodeSteps[0] = 1;
-    nodeSteps[1] = gFineNodesPerDir[0];
-    nodeSteps[2] = gFineNodesPerDir[0]*gFineNodesPerDir[1];
-    Array<LO> glFineNodesPerDir(3);
-    GO startingGID = A->getRowMap()->getMinGlobalIndex();
-    for(LO dim = 0; dim < 3; ++dim) {
-      LO numCoarseNodes = 0;
-      if(dim < numDimensions) {
-        startingGID -= myOffset[dim]*nodeSteps[dim];
-        numCoarseNodes = lCoarseNodesPerDir[dim];
-        if(ghostInterface[2*dim]) {++numCoarseNodes;}
-        if(ghostInterface[2*dim+1]) {++numCoarseNodes;}
-        if(gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim]) {
-          glFineNodesPerDir[dim] = (numCoarseNodes - 2)*coarseRate[dim] + endRate[dim] + 1;
-        } else {
-          glFineNodesPerDir[dim] = (numCoarseNodes - 1)*coarseRate[dim] + 1;
-        }
+  GO gNumCoarseNodes = 0;
+  LO lNumCoarseNodes = 0;
+  Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
+  Array<LO> myOffset(3), lCoarseNodesPerDir(3), glCoarseNodesPerDir(3), endRate(3);
+  Array<bool> ghostInterface(6);
+  Array<int> boundaryFlags(3);
+  ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
+  Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
+  for (LO dim = 0; dim < numDimensions; ++dim) {
+    fineNodes[dim] = coordinates->getData(dim)();
+  }
+
+  // This struct stores PIDs, LIDs and GIDs on the fine mesh and GIDs on the coarse mesh.
+  RCP<NodesIDs> ghostedCoarseNodes = rcp(new NodesIDs{});
+
+  GetGeometricData(coordinates, coarseRate, gFineNodesPerDir, lFineNodesPerDir, BlkSize,  // inputs
+                   gIndices, myOffset, ghostInterface, endRate, gCoarseNodesPerDir,       // outputs
+                   lCoarseNodesPerDir, glCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs, colGIDs,
+                   gNumCoarseNodes, lNumCoarseNodes, coarseNodes, boundaryFlags,
+                   ghostedCoarseNodes);
+
+  // Create the MultiVector of coarse coordinates
+  Xpetra::UnderlyingLib lib      = coordinates->getMap()->lib();
+  RCP<const Map> coarseCoordsMap = MapFactory::Build(lib,
+                                                     gNumCoarseNodes,
+                                                     coarseNodesGIDs.view(0, lNumCoarseNodes),
+                                                     coordinates->getMap()->getIndexBase(),
+                                                     coordinates->getMap()->getComm());
+  Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseCoords(numDimensions);
+  for (LO dim = 0; dim < numDimensions; ++dim) {
+    coarseCoords[dim] = coarseNodes[dim]();
+  }
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coarseCoordinates =
+      Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(coarseCoordsMap, coarseCoords(),
+                                                                                                           numDimensions);
+
+  // Now create a new matrix: Aghost that contains all the data
+  // locally needed to compute the local part of the prolongator.
+  // Here assuming that all the coarse nodes o and fine nodes +
+  // are local then all the data associated with the coarse
+  // nodes O and the fine nodes * needs to be imported.
+  //
+  //                  *--*--*--*--*--*--*--*
+  //                  |  |  |  |  |  |  |  |
+  //                  o--+--+--o--+--+--O--*
+  //                  |  |  |  |  |  |  |  |
+  //                  +--+--+--+--+--+--*--*
+  //                  |  |  |  |  |  |  |  |
+  //                  +--+--+--+--+--+--*--*
+  //                  |  |  |  |  |  |  |  |
+  //                  o--+--+--o--+--+--O--*
+  //                  |  |  |  |  |  |  |  |
+  //                  +--+--+--+--+--+--*--*
+  //                  |  |  |  |  |  |  |  |
+  //                  *--*--*--*--*--*--*--*
+  //                  |  |  |  |  |  |  |  |
+  //                  O--*--*--O--*--*--O--*
+  //
+  // Creating that local matrix is easy enough using proper range
+  // and domain maps to import data from A. Note that with this
+  // approach we reorder the local entries using the domain map and
+  // can subsequently compute the prolongator without reordering.
+  // As usual we need to be careful about any coarsening rate
+  // change at the boundary!
+
+  // The ingredients needed are an importer, a range map and a domain map
+  Array<GO> ghostRowGIDs, ghostColGIDs, nodeSteps(3);
+  nodeSteps[0] = 1;
+  nodeSteps[1] = gFineNodesPerDir[0];
+  nodeSteps[2] = gFineNodesPerDir[0] * gFineNodesPerDir[1];
+  Array<LO> glFineNodesPerDir(3);
+  GO startingGID = A->getRowMap()->getMinGlobalIndex();
+  for (LO dim = 0; dim < 3; ++dim) {
+    LO numCoarseNodes = 0;
+    if (dim < numDimensions) {
+      startingGID -= myOffset[dim] * nodeSteps[dim];
+      numCoarseNodes = lCoarseNodesPerDir[dim];
+      if (ghostInterface[2 * dim]) {
+        ++numCoarseNodes;
+      }
+      if (ghostInterface[2 * dim + 1]) {
+        ++numCoarseNodes;
+      }
+      if (gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim]) {
+        glFineNodesPerDir[dim] = (numCoarseNodes - 2) * coarseRate[dim] + endRate[dim] + 1;
       } else {
-        glFineNodesPerDir[dim] = 1;
+        glFineNodesPerDir[dim] = (numCoarseNodes - 1) * coarseRate[dim] + 1;
       }
+    } else {
+      glFineNodesPerDir[dim] = 1;
     }
-    ghostRowGIDs.resize(glFineNodesPerDir[0]*glFineNodesPerDir[1]*glFineNodesPerDir[2]*BlkSize);
-    for(LO k = 0; k < glFineNodesPerDir[2]; ++k) {
-      for(LO j = 0; j < glFineNodesPerDir[1]; ++j) {
-        for(LO i = 0; i < glFineNodesPerDir[0]; ++i) {
-          for(LO l = 0; l < BlkSize; ++l) {
-            ghostRowGIDs[(k*glFineNodesPerDir[1]*glFineNodesPerDir[0]
-                          + j*glFineNodesPerDir[0] + i)*BlkSize + l] = startingGID
-              + (k*gFineNodesPerDir[1]*gFineNodesPerDir[0] + j*gFineNodesPerDir[0] + i)*BlkSize + l;
-          }
+  }
+  ghostRowGIDs.resize(glFineNodesPerDir[0] * glFineNodesPerDir[1] * glFineNodesPerDir[2] * BlkSize);
+  for (LO k = 0; k < glFineNodesPerDir[2]; ++k) {
+    for (LO j = 0; j < glFineNodesPerDir[1]; ++j) {
+      for (LO i = 0; i < glFineNodesPerDir[0]; ++i) {
+        for (LO l = 0; l < BlkSize; ++l) {
+          ghostRowGIDs[(k * glFineNodesPerDir[1] * glFineNodesPerDir[0] + j * glFineNodesPerDir[0] + i) * BlkSize + l] = startingGID + (k * gFineNodesPerDir[1] * gFineNodesPerDir[0] + j * gFineNodesPerDir[0] + i) * BlkSize + l;
         }
       }
     }
+  }
 
-    // Looking at the above loops it is easy to find startingGID for the ghostColGIDs
-    Array<GO> startingGlobalIndices(numDimensions), dimStride(numDimensions);
-    Array<GO> startingColIndices(numDimensions), finishingColIndices(numDimensions);
-    GO colMinGID = 0;
-    Array<LO> colRange(numDimensions);
-    dimStride[0] = 1;
-    for(int dim = 1; dim < numDimensions; ++dim) {
-      dimStride[dim] = dimStride[dim - 1]*gFineNodesPerDir[dim - 1];
-    }
-    {
-      GO tmp = startingGID;
-      for(int dim = numDimensions; dim > 0; --dim) {
-        startingGlobalIndices[dim - 1] = tmp / dimStride[dim - 1];
-        tmp = tmp % dimStride[dim - 1];
-
-        if(startingGlobalIndices[dim - 1] > 0) {
-          startingColIndices[dim - 1] = startingGlobalIndices[dim - 1] - 1;
-        }
-        if(startingGlobalIndices[dim - 1] + glFineNodesPerDir[dim - 1] < gFineNodesPerDir[dim - 1]){
-          finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1]
-            + glFineNodesPerDir[dim - 1];
-        } else {
-          finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1]
-            + glFineNodesPerDir[dim - 1] - 1;
-        }
-        colRange[dim - 1] = finishingColIndices[dim - 1] - startingColIndices[dim - 1] + 1;
-        colMinGID += startingColIndices[dim - 1]*dimStride[dim - 1];
+  // Looking at the above loops it is easy to find startingGID for the ghostColGIDs
+  Array<GO> startingGlobalIndices(numDimensions), dimStride(numDimensions);
+  Array<GO> startingColIndices(numDimensions), finishingColIndices(numDimensions);
+  GO colMinGID = 0;
+  Array<LO> colRange(numDimensions);
+  dimStride[0] = 1;
+  for (int dim = 1; dim < numDimensions; ++dim) {
+    dimStride[dim] = dimStride[dim - 1] * gFineNodesPerDir[dim - 1];
+  }
+  {
+    GO tmp = startingGID;
+    for (int dim = numDimensions; dim > 0; --dim) {
+      startingGlobalIndices[dim - 1] = tmp / dimStride[dim - 1];
+      tmp                            = tmp % dimStride[dim - 1];
+
+      if (startingGlobalIndices[dim - 1] > 0) {
+        startingColIndices[dim - 1] = startingGlobalIndices[dim - 1] - 1;
+      }
+      if (startingGlobalIndices[dim - 1] + glFineNodesPerDir[dim - 1] < gFineNodesPerDir[dim - 1]) {
+        finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + glFineNodesPerDir[dim - 1];
+      } else {
+        finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + glFineNodesPerDir[dim - 1] - 1;
       }
+      colRange[dim - 1] = finishingColIndices[dim - 1] - startingColIndices[dim - 1] + 1;
+      colMinGID += startingColIndices[dim - 1] * dimStride[dim - 1];
     }
-    ghostColGIDs.resize(colRange[0]*colRange[1]*colRange[2]*BlkSize);
-    for(LO k = 0; k < colRange[2]; ++k) {
-      for(LO j = 0; j < colRange[1]; ++j) {
-        for(LO i = 0; i < colRange[0]; ++i) {
-          for(LO l = 0; l < BlkSize; ++l) {
-            ghostColGIDs[(k*colRange[1]*colRange[0] + j*colRange[0] + i)*BlkSize + l] = colMinGID
-              + (k*gFineNodesPerDir[1]*gFineNodesPerDir[0] + j*gFineNodesPerDir[0] + i)*BlkSize + l;
-          }
+  }
+  ghostColGIDs.resize(colRange[0] * colRange[1] * colRange[2] * BlkSize);
+  for (LO k = 0; k < colRange[2]; ++k) {
+    for (LO j = 0; j < colRange[1]; ++j) {
+      for (LO i = 0; i < colRange[0]; ++i) {
+        for (LO l = 0; l < BlkSize; ++l) {
+          ghostColGIDs[(k * colRange[1] * colRange[0] + j * colRange[0] + i) * BlkSize + l] = colMinGID + (k * gFineNodesPerDir[1] * gFineNodesPerDir[0] + j * gFineNodesPerDir[0] + i) * BlkSize + l;
         }
       }
     }
+  }
 
-    RCP<const Map> ghostedRowMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getRowMap()->lib(),
-                                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                           ghostRowGIDs(),
-                                           A->getRowMap()->getIndexBase(),
-                                           A->getRowMap()->getComm());
-    RCP<const Map> ghostedColMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getRowMap()->lib(),
-                                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                           ghostColGIDs(),
-                                           A->getRowMap()->getIndexBase(),
-                                           A->getRowMap()->getComm());
-    RCP<const Import> ghostImporter = Xpetra::ImportFactory<LO,GO,NO>::Build(A->getRowMap(),
+  RCP<const Map> ghostedRowMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getRowMap()->lib(),
+                                                                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                                          ghostRowGIDs(),
+                                                                          A->getRowMap()->getIndexBase(),
+                                                                          A->getRowMap()->getComm());
+  RCP<const Map> ghostedColMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getRowMap()->lib(),
+                                                                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                                          ghostColGIDs(),
+                                                                          A->getRowMap()->getIndexBase(),
+                                                                          A->getRowMap()->getComm());
+  RCP<const Import> ghostImporter = Xpetra::ImportFactory<LO, GO, NO>::Build(A->getRowMap(),
                                                                              ghostedRowMap);
-    RCP<const Matrix> Aghost        = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(A, *ghostImporter,
-                                                                                ghostedRowMap,
-                                                                                ghostedRowMap);
-
-    // Create the maps and data structures for the projection matrix
-    RCP<const Map> rowMapP    = A->getDomainMap();
-
-    RCP<const Map> domainMapP;
-
-    RCP<const Map> colMapP;
-
-    // At this point we need to create the column map which is a delicate operation.
-    // The entries in that map need to be ordered as follows:
-    //         1) first owned entries ordered by LID
-    //         2) second order the remaining entries by PID
-    //         3) entries with the same remote PID are ordered by GID.
-    // One piece of good news: lNumCoarseNodes is the number of ownedNodes and lNumGhostNodes
-    // is the number of remote nodes. The sorting can be limited to remote nodes
-    // as the owned ones are alreaded ordered by LID!
-
-    LO lNumGhostedNodes = ghostedCoarseNodes->GIDs.size();
-    {
-      std::vector<NodeID> colMapOrdering(lNumGhostedNodes);
-      for(LO ind = 0; ind < lNumGhostedNodes; ++ind) {
-        colMapOrdering[ind].GID = ghostedCoarseNodes->GIDs[ind];
-        if(ghostedCoarseNodes->PIDs[ind] == rowMapP->getComm()->getRank()) {
-          colMapOrdering[ind].PID = -1;
-        } else {
-          colMapOrdering[ind].PID = ghostedCoarseNodes->PIDs[ind];
-        }
-        colMapOrdering[ind].LID = ghostedCoarseNodes->LIDs[ind];
-        colMapOrdering[ind].lexiInd = ind;
+  RCP<const Matrix> Aghost        = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(A, *ghostImporter,
+                                                                                 ghostedRowMap,
+                                                                                 ghostedRowMap);
+
+  // Create the maps and data structures for the projection matrix
+  RCP<const Map> rowMapP = A->getDomainMap();
+
+  RCP<const Map> domainMapP;
+
+  RCP<const Map> colMapP;
+
+  // At this point we need to create the column map which is a delicate operation.
+  // The entries in that map need to be ordered as follows:
+  //         1) first owned entries ordered by LID
+  //         2) second order the remaining entries by PID
+  //         3) entries with the same remote PID are ordered by GID.
+  // One piece of good news: lNumCoarseNodes is the number of ownedNodes and lNumGhostNodes
+  // is the number of remote nodes. The sorting can be limited to remote nodes
+  // as the owned ones are alreaded ordered by LID!
+
+  LO lNumGhostedNodes = ghostedCoarseNodes->GIDs.size();
+  {
+    std::vector<NodeID> colMapOrdering(lNumGhostedNodes);
+    for (LO ind = 0; ind < lNumGhostedNodes; ++ind) {
+      colMapOrdering[ind].GID = ghostedCoarseNodes->GIDs[ind];
+      if (ghostedCoarseNodes->PIDs[ind] == rowMapP->getComm()->getRank()) {
+        colMapOrdering[ind].PID = -1;
+      } else {
+        colMapOrdering[ind].PID = ghostedCoarseNodes->PIDs[ind];
       }
-      std::sort(colMapOrdering.begin(), colMapOrdering.end(),
-                [](NodeID a, NodeID b)->bool{
-                  return ( (a.PID < b.PID) || ((a.PID == b.PID) && (a.GID < b.GID)) );
-                });
-
-      colGIDs.resize(BlkSize*lNumGhostedNodes);
-      for(LO ind = 0; ind < lNumGhostedNodes; ++ind) {
-        // Save the permutation calculated to go from Lexicographic indexing to column map indexing
-        ghostedCoarseNodes->colInds[colMapOrdering[ind].lexiInd] = ind;
-        for(LO dof = 0; dof < BlkSize; ++dof) {
-          colGIDs[BlkSize*ind + dof] = BlkSize*colMapOrdering[ind].GID + dof;
-        }
+      colMapOrdering[ind].LID     = ghostedCoarseNodes->LIDs[ind];
+      colMapOrdering[ind].lexiInd = ind;
+    }
+    std::sort(colMapOrdering.begin(), colMapOrdering.end(),
+              [](NodeID a, NodeID b) -> bool {
+                return ((a.PID < b.PID) || ((a.PID == b.PID) && (a.GID < b.GID)));
+              });
+
+    colGIDs.resize(BlkSize * lNumGhostedNodes);
+    for (LO ind = 0; ind < lNumGhostedNodes; ++ind) {
+      // Save the permutation calculated to go from Lexicographic indexing to column map indexing
+      ghostedCoarseNodes->colInds[colMapOrdering[ind].lexiInd] = ind;
+      for (LO dof = 0; dof < BlkSize; ++dof) {
+        colGIDs[BlkSize * ind + dof] = BlkSize * colMapOrdering[ind].GID + dof;
       }
-      domainMapP = Xpetra::MapFactory<LO,GO,NO>::Build(rowMapP->lib(),
-                                                       BlkSize*gNumCoarseNodes,
-                                                       colGIDs.view(0,BlkSize*lNumCoarseNodes),
+    }
+    domainMapP = Xpetra::MapFactory<LO, GO, NO>::Build(rowMapP->lib(),
+                                                       BlkSize * gNumCoarseNodes,
+                                                       colGIDs.view(0, BlkSize * lNumCoarseNodes),
+                                                       rowMapP->getIndexBase(),
+                                                       rowMapP->getComm());
+    colMapP    = Xpetra::MapFactory<LO, GO, NO>::Build(rowMapP->lib(),
+                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                       colGIDs.view(0, colGIDs.size()),
                                                        rowMapP->getIndexBase(),
                                                        rowMapP->getComm());
-      colMapP = Xpetra::MapFactory<LO,GO,NO>::Build(rowMapP->lib(),
-                                                    Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                    colGIDs.view(0, colGIDs.size()),
-                                                    rowMapP->getIndexBase(),
-                                                    rowMapP->getComm());
-    } // End of scope for colMapOrdering and colGIDs
-
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0] = BlkSize;
-    RCP<const Map> stridedDomainMapP = Xpetra::StridedMapFactory<LO,GO,NO>::Build(domainMapP,
+  }  // End of scope for colMapOrdering and colGIDs
+
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0]                    = BlkSize;
+  RCP<const Map> stridedDomainMapP = Xpetra::StridedMapFactory<LO, GO, NO>::Build(domainMapP,
                                                                                   strideInfo);
 
-    GO gnnzP = 0;
-    LO lnnzP = 0;
-    // coarse points have one nnz per row
-    gnnzP += gNumCoarseNodes;
-    lnnzP += lNumCoarseNodes;
-    // add all other points multiplying by 2^numDimensions
-    gnnzP += (gNumFineNodes - gNumCoarseNodes)*std::pow(2, numDimensions);
-    lnnzP += (lNumFineNodes - lNumCoarseNodes)*std::pow(2, numDimensions);
-    // finally multiply by the number of dofs per node
-    gnnzP = gnnzP*BlkSize;
-    lnnzP = lnnzP*BlkSize;
-
-    // Create the matrix itself using the above maps
-    RCP<Matrix> P;
-    P = rcp(new CrsMatrixWrap(rowMapP, colMapP, 0));
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    ArrayRCP<size_t>  iaP;
-    ArrayRCP<LO>      jaP;
-    ArrayRCP<SC>     valP;
-
-    PCrs->allocateAllValues(lnnzP, iaP, jaP, valP);
-
-    ArrayView<size_t> ia  = iaP();
-    ArrayView<LO>     ja  = jaP();
-    ArrayView<SC>     val = valP();
-    ia[0] = 0;
-
-
-    LO numCoarseElements = 1;
-    Array<LO> lCoarseElementsPerDir(3);
-    for(LO dim = 0; dim < numDimensions; ++dim) {
-      lCoarseElementsPerDir[dim] = lCoarseNodesPerDir[dim];
-      if(ghostInterface[2*dim]) {++lCoarseElementsPerDir[dim];}
-      if(!ghostInterface[2*dim+1]) {--lCoarseElementsPerDir[dim];}
-      numCoarseElements = numCoarseElements*lCoarseElementsPerDir[dim];
+  GO gnnzP = 0;
+  LO lnnzP = 0;
+  // coarse points have one nnz per row
+  gnnzP += gNumCoarseNodes;
+  lnnzP += lNumCoarseNodes;
+  // add all other points multiplying by 2^numDimensions
+  gnnzP += (gNumFineNodes - gNumCoarseNodes) * std::pow(2, numDimensions);
+  lnnzP += (lNumFineNodes - lNumCoarseNodes) * std::pow(2, numDimensions);
+  // finally multiply by the number of dofs per node
+  gnnzP = gnnzP * BlkSize;
+  lnnzP = lnnzP * BlkSize;
+
+  // Create the matrix itself using the above maps
+  RCP<Matrix> P;
+  P                   = rcp(new CrsMatrixWrap(rowMapP, colMapP, 0));
+  RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  ArrayRCP<size_t> iaP;
+  ArrayRCP<LO> jaP;
+  ArrayRCP<SC> valP;
+
+  PCrs->allocateAllValues(lnnzP, iaP, jaP, valP);
+
+  ArrayView<size_t> ia = iaP();
+  ArrayView<LO> ja     = jaP();
+  ArrayView<SC> val    = valP();
+  ia[0]                = 0;
+
+  LO numCoarseElements = 1;
+  Array<LO> lCoarseElementsPerDir(3);
+  for (LO dim = 0; dim < numDimensions; ++dim) {
+    lCoarseElementsPerDir[dim] = lCoarseNodesPerDir[dim];
+    if (ghostInterface[2 * dim]) {
+      ++lCoarseElementsPerDir[dim];
     }
-
-    for(LO dim = numDimensions; dim < 3; ++dim) {
-      lCoarseElementsPerDir[dim] = 1;
+    if (!ghostInterface[2 * dim + 1]) {
+      --lCoarseElementsPerDir[dim];
     }
+    numCoarseElements = numCoarseElements * lCoarseElementsPerDir[dim];
+  }
 
-    // Loop over the coarse elements
-    Array<int> elementFlags(3);
-    Array<LO> elemInds(3), elementNodesPerDir(3), glElementRefTuple(3);
-    Array<LO> glElementRefTupleCG(3), glElementCoarseNodeCG(8);
-    const int numCoarseNodesInElement = std::pow(2, numDimensions);
-    const int nnzPerCoarseNode = (blockStrategy == "coupled") ? BlkSize : 1;
-    const int numRowsPerPoint = BlkSize;
-    for(elemInds[2] = 0; elemInds[2] < lCoarseElementsPerDir[2]; ++elemInds[2]) {
-      for(elemInds[1] = 0; elemInds[1] < lCoarseElementsPerDir[1]; ++elemInds[1]) {
-        for(elemInds[0] = 0; elemInds[0] < lCoarseElementsPerDir[0]; ++elemInds[0]) {
-          elementFlags[0] = 0; elementFlags[1] = 0; elementFlags[2] = 0;
-          for(int dim = 0; dim < 3; ++dim) {
-            // Detect boundary conditions on the element and set corresponding flags.
-            if(elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
-              elementFlags[dim] = boundaryFlags[dim];
-            } else if(elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
-              elementFlags[dim] += 1;
-            } else if((elemInds[dim] == lCoarseElementsPerDir[dim] - 1)
-                      && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
-              elementFlags[dim] += 2;
-            } else {
-              elementFlags[dim] = 0;
-            }
+  for (LO dim = numDimensions; dim < 3; ++dim) {
+    lCoarseElementsPerDir[dim] = 1;
+  }
 
-            // Compute the number of nodes in the current element.
-            if(dim < numDimensions) {
-              if((elemInds[dim] == lCoarseElementsPerDir[dim])
-                 && (gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim])) {
-                elementNodesPerDir[dim] = endRate[dim] + 1;
-              } else {
-                elementNodesPerDir[dim] = coarseRate[dim] + 1;
-              }
+  // Loop over the coarse elements
+  Array<int> elementFlags(3);
+  Array<LO> elemInds(3), elementNodesPerDir(3), glElementRefTuple(3);
+  Array<LO> glElementRefTupleCG(3), glElementCoarseNodeCG(8);
+  const int numCoarseNodesInElement = std::pow(2, numDimensions);
+  const int nnzPerCoarseNode        = (blockStrategy == "coupled") ? BlkSize : 1;
+  const int numRowsPerPoint         = BlkSize;
+  for (elemInds[2] = 0; elemInds[2] < lCoarseElementsPerDir[2]; ++elemInds[2]) {
+    for (elemInds[1] = 0; elemInds[1] < lCoarseElementsPerDir[1]; ++elemInds[1]) {
+      for (elemInds[0] = 0; elemInds[0] < lCoarseElementsPerDir[0]; ++elemInds[0]) {
+        elementFlags[0] = 0;
+        elementFlags[1] = 0;
+        elementFlags[2] = 0;
+        for (int dim = 0; dim < 3; ++dim) {
+          // Detect boundary conditions on the element and set corresponding flags.
+          if (elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
+            elementFlags[dim] = boundaryFlags[dim];
+          } else if (elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
+            elementFlags[dim] += 1;
+          } else if ((elemInds[dim] == lCoarseElementsPerDir[dim] - 1) && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
+            elementFlags[dim] += 2;
+          } else {
+            elementFlags[dim] = 0;
+          }
+
+          // Compute the number of nodes in the current element.
+          if (dim < numDimensions) {
+            if ((elemInds[dim] == lCoarseElementsPerDir[dim]) && (gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim])) {
+              elementNodesPerDir[dim] = endRate[dim] + 1;
             } else {
-              elementNodesPerDir[dim] = 1;
+              elementNodesPerDir[dim] = coarseRate[dim] + 1;
             }
-
-            // Get the lowest tuple of the element using the ghosted local coordinate system
-            glElementRefTuple[dim]   = elemInds[dim]*coarseRate[dim];
-            glElementRefTupleCG[dim] = elemInds[dim];
+          } else {
+            elementNodesPerDir[dim] = 1;
           }
 
-          // Now get the column map indices corresponding to the dofs associated with the current
-          // element's coarse nodes.
-          for(typename Array<LO>::size_type elem = 0; elem < glElementCoarseNodeCG.size(); ++elem) {
-            glElementCoarseNodeCG[elem]
-              = glElementRefTupleCG[2]*glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0]
-                 + glElementRefTupleCG[1]*glCoarseNodesPerDir[0] + glElementRefTupleCG[0];
-          }
-          glElementCoarseNodeCG[4] += glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[5] += glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[6] += glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[7] += glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0];
-
-          glElementCoarseNodeCG[2] += glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[3] += glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[6] += glCoarseNodesPerDir[0];
-          glElementCoarseNodeCG[7] += glCoarseNodesPerDir[0];
-
-          glElementCoarseNodeCG[1] += 1;
-          glElementCoarseNodeCG[3] += 1;
-          glElementCoarseNodeCG[5] += 1;
-          glElementCoarseNodeCG[7] += 1;
-
-          LO numNodesInElement = elementNodesPerDir[0]*elementNodesPerDir[1]*elementNodesPerDir[2];
-          // LO elementOffset = elemInds[2]*coarseRate[2]*glFineNodesPerDir[1]*glFineNodesPerDir[0]
-          //   + elemInds[1]*coarseRate[1]*glFineNodesPerDir[0] + elemInds[0]*coarseRate[0];
-
-          // Compute the element prolongator
-          Teuchos::SerialDenseMatrix<LO,SC> Pi, Pf, Pe;
-          Array<LO> dofType(numNodesInElement*BlkSize), lDofInd(numNodesInElement*BlkSize);
-          ComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds, lCoarseElementsPerDir,
-                              numDimensions, glFineNodesPerDir, gFineNodesPerDir, gIndices,
-                              lCoarseNodesPerDir, ghostInterface, elementFlags, stencilType,
-                              blockStrategy, elementNodesPerDir, numNodesInElement, colGIDs,
-                              Pi, Pf, Pe, dofType, lDofInd);
-
-          // Find ghosted LID associated with nodes in the element and eventually which of these
-          // nodes are ghosts, this information is used to fill the local prolongator.
-          Array<LO> lNodeLIDs(numNodesInElement);
-          {
-            Array<LO> lNodeTuple(3), nodeInd(3);
-            for(nodeInd[2] = 0; nodeInd[2] < elementNodesPerDir[2]; ++nodeInd[2]) {
-              for(nodeInd[1] = 0; nodeInd[1] < elementNodesPerDir[1]; ++nodeInd[1]) {
-                for(nodeInd[0] = 0; nodeInd[0] < elementNodesPerDir[0]; ++nodeInd[0]) {
-                  int stencilLength = 0;
-                  if((nodeInd[0] == 0 || nodeInd[0] == elementNodesPerDir[0] - 1) &&
-                     (nodeInd[1] == 0 || nodeInd[1] == elementNodesPerDir[1] - 1) &&
-                     (nodeInd[2] == 0 || nodeInd[2] == elementNodesPerDir[2] - 1)) {
-                    stencilLength = 1;
-                  } else {
-                    stencilLength = std::pow(2, numDimensions);
-                  }
-                  LO nodeElementInd = nodeInd[2]*elementNodesPerDir[1]*elementNodesPerDir[1]
-                    + nodeInd[1]*elementNodesPerDir[0] + nodeInd[0];
-                  for(int dim = 0; dim < 3; ++dim) {
-                    lNodeTuple[dim] = glElementRefTuple[dim] - myOffset[dim] + nodeInd[dim];
-                  }
-                  if(lNodeTuple[0] < 0 || lNodeTuple[1] < 0 || lNodeTuple[2] < 0 ||
-                     lNodeTuple[0] > lFineNodesPerDir[0] - 1 ||
-                     lNodeTuple[1] > lFineNodesPerDir[1] - 1 ||
-                     lNodeTuple[2] > lFineNodesPerDir[2] - 1) {
-                    // This flags the ghosts nodes used for prolongator calculation but for which
-                    // we do not own the row, hence we won't fill these values on this rank.
-                    lNodeLIDs[nodeElementInd] = -1;
-                  } else if ((nodeInd[0] == 0 && elemInds[0] !=0) ||
-                             (nodeInd[1] == 0 && elemInds[1] !=0) ||
-                             (nodeInd[2] == 0 && elemInds[2] !=0)) {
-                    // This flags nodes that are owned but common to two coarse elements and that
-                    // were already filled by another element, we don't want to fill them twice so
-                    // we skip them
-                    lNodeLIDs[nodeElementInd] = -2;
-                  } else {
-                    // The remaining nodes are locally owned and we need to fill the coresponding
-                    // rows of the prolongator
-
-                    // First we need to find which row needs to be filled
-                    lNodeLIDs[nodeElementInd] = lNodeTuple[2]*lFineNodesPerDir[1]*lFineNodesPerDir[0]
-                      + lNodeTuple[1]*lFineNodesPerDir[0] + lNodeTuple[0];
-
-                    // We also compute the row offset using arithmetic to ensure that we can loop
-                    // easily over the nodes in a macro-element as well as facilitate on-node
-                    // parallelization. The node serial code could be rewritten with two loops over
-                    // the local part of the mesh to avoid the costly integer divisions...
-                    Array<LO> refCoarsePointTuple(3);
-                    for(int dim = 2; dim > -1; --dim) {
-                      if(dim == 0) {
-                        refCoarsePointTuple[dim] = (lNodeTuple[dim] + myOffset[dim]) / coarseRate[dim];
-                        if(myOffset[dim] == 0) {
-                          ++refCoarsePointTuple[dim];
-                        }
-                      } else {
-                        // Note:  no need for magnitudeType here, just use double because these things are LO's
-                        refCoarsePointTuple[dim] =
-                          std::ceil(static_cast<double>(lNodeTuple[dim] + myOffset[dim])
-                                    / coarseRate[dim]);
+          // Get the lowest tuple of the element using the ghosted local coordinate system
+          glElementRefTuple[dim]   = elemInds[dim] * coarseRate[dim];
+          glElementRefTupleCG[dim] = elemInds[dim];
+        }
+
+        // Now get the column map indices corresponding to the dofs associated with the current
+        // element's coarse nodes.
+        for (typename Array<LO>::size_type elem = 0; elem < glElementCoarseNodeCG.size(); ++elem) {
+          glElementCoarseNodeCG[elem] = glElementRefTupleCG[2] * glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0] + glElementRefTupleCG[1] * glCoarseNodesPerDir[0] + glElementRefTupleCG[0];
+        }
+        glElementCoarseNodeCG[4] += glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[5] += glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[6] += glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[7] += glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0];
+
+        glElementCoarseNodeCG[2] += glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[3] += glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[6] += glCoarseNodesPerDir[0];
+        glElementCoarseNodeCG[7] += glCoarseNodesPerDir[0];
+
+        glElementCoarseNodeCG[1] += 1;
+        glElementCoarseNodeCG[3] += 1;
+        glElementCoarseNodeCG[5] += 1;
+        glElementCoarseNodeCG[7] += 1;
+
+        LO numNodesInElement = elementNodesPerDir[0] * elementNodesPerDir[1] * elementNodesPerDir[2];
+        // LO elementOffset = elemInds[2]*coarseRate[2]*glFineNodesPerDir[1]*glFineNodesPerDir[0]
+        //   + elemInds[1]*coarseRate[1]*glFineNodesPerDir[0] + elemInds[0]*coarseRate[0];
+
+        // Compute the element prolongator
+        Teuchos::SerialDenseMatrix<LO, SC> Pi, Pf, Pe;
+        Array<LO> dofType(numNodesInElement * BlkSize), lDofInd(numNodesInElement * BlkSize);
+        ComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds, lCoarseElementsPerDir,
+                            numDimensions, glFineNodesPerDir, gFineNodesPerDir, gIndices,
+                            lCoarseNodesPerDir, ghostInterface, elementFlags, stencilType,
+                            blockStrategy, elementNodesPerDir, numNodesInElement, colGIDs,
+                            Pi, Pf, Pe, dofType, lDofInd);
+
+        // Find ghosted LID associated with nodes in the element and eventually which of these
+        // nodes are ghosts, this information is used to fill the local prolongator.
+        Array<LO> lNodeLIDs(numNodesInElement);
+        {
+          Array<LO> lNodeTuple(3), nodeInd(3);
+          for (nodeInd[2] = 0; nodeInd[2] < elementNodesPerDir[2]; ++nodeInd[2]) {
+            for (nodeInd[1] = 0; nodeInd[1] < elementNodesPerDir[1]; ++nodeInd[1]) {
+              for (nodeInd[0] = 0; nodeInd[0] < elementNodesPerDir[0]; ++nodeInd[0]) {
+                int stencilLength = 0;
+                if ((nodeInd[0] == 0 || nodeInd[0] == elementNodesPerDir[0] - 1) &&
+                    (nodeInd[1] == 0 || nodeInd[1] == elementNodesPerDir[1] - 1) &&
+                    (nodeInd[2] == 0 || nodeInd[2] == elementNodesPerDir[2] - 1)) {
+                  stencilLength = 1;
+                } else {
+                  stencilLength = std::pow(2, numDimensions);
+                }
+                LO nodeElementInd = nodeInd[2] * elementNodesPerDir[1] * elementNodesPerDir[1] + nodeInd[1] * elementNodesPerDir[0] + nodeInd[0];
+                for (int dim = 0; dim < 3; ++dim) {
+                  lNodeTuple[dim] = glElementRefTuple[dim] - myOffset[dim] + nodeInd[dim];
+                }
+                if (lNodeTuple[0] < 0 || lNodeTuple[1] < 0 || lNodeTuple[2] < 0 ||
+                    lNodeTuple[0] > lFineNodesPerDir[0] - 1 ||
+                    lNodeTuple[1] > lFineNodesPerDir[1] - 1 ||
+                    lNodeTuple[2] > lFineNodesPerDir[2] - 1) {
+                  // This flags the ghosts nodes used for prolongator calculation but for which
+                  // we do not own the row, hence we won't fill these values on this rank.
+                  lNodeLIDs[nodeElementInd] = -1;
+                } else if ((nodeInd[0] == 0 && elemInds[0] != 0) ||
+                           (nodeInd[1] == 0 && elemInds[1] != 0) ||
+                           (nodeInd[2] == 0 && elemInds[2] != 0)) {
+                  // This flags nodes that are owned but common to two coarse elements and that
+                  // were already filled by another element, we don't want to fill them twice so
+                  // we skip them
+                  lNodeLIDs[nodeElementInd] = -2;
+                } else {
+                  // The remaining nodes are locally owned and we need to fill the coresponding
+                  // rows of the prolongator
+
+                  // First we need to find which row needs to be filled
+                  lNodeLIDs[nodeElementInd] = lNodeTuple[2] * lFineNodesPerDir[1] * lFineNodesPerDir[0] + lNodeTuple[1] * lFineNodesPerDir[0] + lNodeTuple[0];
+
+                  // We also compute the row offset using arithmetic to ensure that we can loop
+                  // easily over the nodes in a macro-element as well as facilitate on-node
+                  // parallelization. The node serial code could be rewritten with two loops over
+                  // the local part of the mesh to avoid the costly integer divisions...
+                  Array<LO> refCoarsePointTuple(3);
+                  for (int dim = 2; dim > -1; --dim) {
+                    if (dim == 0) {
+                      refCoarsePointTuple[dim] = (lNodeTuple[dim] + myOffset[dim]) / coarseRate[dim];
+                      if (myOffset[dim] == 0) {
+                        ++refCoarsePointTuple[dim];
                       }
-                      if((lNodeTuple[dim] + myOffset[dim]) % coarseRate[dim] > 0) {break;}
-                    }
-                    const LO numCoarsePoints = refCoarsePointTuple[0]
-                      + refCoarsePointTuple[1]*lCoarseNodesPerDir[0]
-                      + refCoarsePointTuple[2]*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0];
-                    const LO numFinePoints = lNodeLIDs[nodeElementInd] + 1;
-
-                    // The below formula computes the rowPtr for row 'n+BlkSize' and we are about to
-                    // fill row 'n' to 'n+BlkSize'.
-                    size_t rowPtr = (numFinePoints - numCoarsePoints)*numRowsPerPoint
-                      *numCoarseNodesInElement*nnzPerCoarseNode + numCoarsePoints*numRowsPerPoint;
-                    if(dofType[nodeElementInd*BlkSize] == 0) {
-                      // Check if current node is a coarse point
-                      rowPtr = rowPtr - numRowsPerPoint;
                     } else {
-                      rowPtr = rowPtr - numRowsPerPoint*numCoarseNodesInElement*nnzPerCoarseNode;
+                      // Note:  no need for magnitudeType here, just use double because these things are LO's
+                      refCoarsePointTuple[dim] =
+                          std::ceil(static_cast<double>(lNodeTuple[dim] + myOffset[dim]) / coarseRate[dim]);
+                    }
+                    if ((lNodeTuple[dim] + myOffset[dim]) % coarseRate[dim] > 0) {
+                      break;
                     }
+                  }
+                  const LO numCoarsePoints = refCoarsePointTuple[0] + refCoarsePointTuple[1] * lCoarseNodesPerDir[0] + refCoarsePointTuple[2] * lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0];
+                  const LO numFinePoints   = lNodeLIDs[nodeElementInd] + 1;
+
+                  // The below formula computes the rowPtr for row 'n+BlkSize' and we are about to
+                  // fill row 'n' to 'n+BlkSize'.
+                  size_t rowPtr = (numFinePoints - numCoarsePoints) * numRowsPerPoint * numCoarseNodesInElement * nnzPerCoarseNode + numCoarsePoints * numRowsPerPoint;
+                  if (dofType[nodeElementInd * BlkSize] == 0) {
+                    // Check if current node is a coarse point
+                    rowPtr = rowPtr - numRowsPerPoint;
+                  } else {
+                    rowPtr = rowPtr - numRowsPerPoint * numCoarseNodesInElement * nnzPerCoarseNode;
+                  }
 
-                    for(int dof = 0; dof < BlkSize; ++dof) {
-
-                      // Now we loop over the stencil and fill the column indices and row values
-                      int cornerInd = 0;
-                      switch(dofType[nodeElementInd*BlkSize + dof]) {
-                      case 0: // Corner node
-                        if(nodeInd[2] == elementNodesPerDir[2] - 1) {cornerInd += 4;}
-                        if(nodeInd[1] == elementNodesPerDir[1] - 1) {cornerInd += 2;}
-                        if(nodeInd[0] == elementNodesPerDir[0] - 1) {cornerInd += 1;}
-                        ia[lNodeLIDs[nodeElementInd]*BlkSize + dof + 1] = rowPtr + dof + 1;
-                        ja[rowPtr + dof] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[cornerInd]]*BlkSize + dof;
-                        val[rowPtr + dof] = 1.0;
+                  for (int dof = 0; dof < BlkSize; ++dof) {
+                    // Now we loop over the stencil and fill the column indices and row values
+                    int cornerInd = 0;
+                    switch (dofType[nodeElementInd * BlkSize + dof]) {
+                      case 0:  // Corner node
+                        if (nodeInd[2] == elementNodesPerDir[2] - 1) {
+                          cornerInd += 4;
+                        }
+                        if (nodeInd[1] == elementNodesPerDir[1] - 1) {
+                          cornerInd += 2;
+                        }
+                        if (nodeInd[0] == elementNodesPerDir[0] - 1) {
+                          cornerInd += 1;
+                        }
+                        ia[lNodeLIDs[nodeElementInd] * BlkSize + dof + 1] = rowPtr + dof + 1;
+                        ja[rowPtr + dof]                                  = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[cornerInd]] * BlkSize + dof;
+                        val[rowPtr + dof]                                 = 1.0;
                         break;
 
-                      case 1: // Edge node
-                        ia[lNodeLIDs[nodeElementInd]*BlkSize + dof + 1]
-                          = rowPtr + (dof + 1)*numCoarseNodesInElement*nnzPerCoarseNode;
-                        for(int ind1 = 0; ind1 < stencilLength; ++ind1) {
-                          if(blockStrategy == "coupled") {
-                            for(int ind2 = 0; ind2 < BlkSize; ++ind2) {
-                              size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                                + ind1*BlkSize + ind2;
-                              ja[lRowPtr]  = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + ind2;
-                              val[lRowPtr] = Pe(lDofInd[nodeElementInd*BlkSize + dof],
-                                                ind1*BlkSize + ind2);
+                      case 1:  // Edge node
+                        ia[lNodeLIDs[nodeElementInd] * BlkSize + dof + 1] = rowPtr + (dof + 1) * numCoarseNodesInElement * nnzPerCoarseNode;
+                        for (int ind1 = 0; ind1 < stencilLength; ++ind1) {
+                          if (blockStrategy == "coupled") {
+                            for (int ind2 = 0; ind2 < BlkSize; ++ind2) {
+                              size_t lRowPtr = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1 * BlkSize + ind2;
+                              ja[lRowPtr]    = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + ind2;
+                              val[lRowPtr]   = Pe(lDofInd[nodeElementInd * BlkSize + dof],
+                                                  ind1 * BlkSize + ind2);
                             }
-                          } else if(blockStrategy == "uncoupled") {
-                            size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                              + ind1;
-                            ja[rowPtr + dof] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + dof;
-                            val[lRowPtr] = Pe(lDofInd[nodeElementInd*BlkSize + dof],
-                                              ind1*BlkSize + dof);
+                          } else if (blockStrategy == "uncoupled") {
+                            size_t lRowPtr   = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1;
+                            ja[rowPtr + dof] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + dof;
+                            val[lRowPtr]     = Pe(lDofInd[nodeElementInd * BlkSize + dof],
+                                                  ind1 * BlkSize + dof);
                           }
                         }
                         break;
 
-                      case 2: // Face node
-                        ia[lNodeLIDs[nodeElementInd]*BlkSize + dof + 1]
-                          = rowPtr + (dof + 1)*numCoarseNodesInElement*nnzPerCoarseNode;
-                        for(int ind1 = 0; ind1 < stencilLength; ++ind1) {
-                          if(blockStrategy == "coupled") {
-                            for(int ind2 = 0; ind2 < BlkSize; ++ind2) {
-                              size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                                + ind1*BlkSize + ind2;
-                              ja [lRowPtr] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + ind2;
-                              val[lRowPtr] = Pf(lDofInd[nodeElementInd*BlkSize + dof],
-                                                ind1*BlkSize + ind2);
+                      case 2:  // Face node
+                        ia[lNodeLIDs[nodeElementInd] * BlkSize + dof + 1] = rowPtr + (dof + 1) * numCoarseNodesInElement * nnzPerCoarseNode;
+                        for (int ind1 = 0; ind1 < stencilLength; ++ind1) {
+                          if (blockStrategy == "coupled") {
+                            for (int ind2 = 0; ind2 < BlkSize; ++ind2) {
+                              size_t lRowPtr = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1 * BlkSize + ind2;
+                              ja[lRowPtr]    = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + ind2;
+                              val[lRowPtr]   = Pf(lDofInd[nodeElementInd * BlkSize + dof],
+                                                  ind1 * BlkSize + ind2);
                             }
-                          } else if(blockStrategy == "uncoupled") {
-                            size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                              + ind1;
+                          } else if (blockStrategy == "uncoupled") {
+                            size_t lRowPtr = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1;
                             // ja [lRowPtr] = colGIDs[glElementCoarseNodeCG[ind1]*BlkSize + dof];
-                            ja [lRowPtr] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + dof;
-                            val[lRowPtr] = Pf(lDofInd[nodeElementInd*BlkSize + dof],
-                                              ind1*BlkSize + dof);
+                            ja[lRowPtr]  = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + dof;
+                            val[lRowPtr] = Pf(lDofInd[nodeElementInd * BlkSize + dof],
+                                              ind1 * BlkSize + dof);
                           }
                         }
                         break;
 
-                      case 3: // Interior node
-                        ia[lNodeLIDs[nodeElementInd]*BlkSize + dof + 1]
-                          = rowPtr + (dof + 1)*numCoarseNodesInElement*nnzPerCoarseNode;
-                        for(int ind1 = 0; ind1 < stencilLength; ++ind1) {
-                          if(blockStrategy == "coupled") {
-                            for(int ind2 = 0; ind2 < BlkSize; ++ind2) {
-                              size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                                + ind1*BlkSize + ind2;
-                              ja [lRowPtr] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + ind2;
-                              val[lRowPtr] = Pi(lDofInd[nodeElementInd*BlkSize + dof],
-                                                ind1*BlkSize + ind2);
+                      case 3:  // Interior node
+                        ia[lNodeLIDs[nodeElementInd] * BlkSize + dof + 1] = rowPtr + (dof + 1) * numCoarseNodesInElement * nnzPerCoarseNode;
+                        for (int ind1 = 0; ind1 < stencilLength; ++ind1) {
+                          if (blockStrategy == "coupled") {
+                            for (int ind2 = 0; ind2 < BlkSize; ++ind2) {
+                              size_t lRowPtr = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1 * BlkSize + ind2;
+                              ja[lRowPtr]    = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + ind2;
+                              val[lRowPtr]   = Pi(lDofInd[nodeElementInd * BlkSize + dof],
+                                                  ind1 * BlkSize + ind2);
                             }
-                          } else if(blockStrategy == "uncoupled") {
-                            size_t lRowPtr = rowPtr + dof*numCoarseNodesInElement*nnzPerCoarseNode
-                              + ind1;
-                            ja [lRowPtr] = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]]*BlkSize + dof;
-                            val[lRowPtr] = Pi(lDofInd[nodeElementInd*BlkSize + dof],
-                                              ind1*BlkSize + dof);
+                          } else if (blockStrategy == "uncoupled") {
+                            size_t lRowPtr = rowPtr + dof * numCoarseNodesInElement * nnzPerCoarseNode + ind1;
+                            ja[lRowPtr]    = ghostedCoarseNodes->colInds[glElementCoarseNodeCG[ind1]] * BlkSize + dof;
+                            val[lRowPtr]   = Pi(lDofInd[nodeElementInd * BlkSize + dof],
+                                                ind1 * BlkSize + dof);
                           }
                         }
                         break;
-                      }
                     }
                   }
                 }
               }
             }
-          } // End of scopt for lNodeTuple and nodeInd
-        }
+          }
+        }  // End of scopt for lNodeTuple and nodeInd
       }
     }
+  }
 
-    // Sort all row's column indicies and entries by LID
-    Xpetra::CrsMatrixUtils<SC,LO,GO,NO>::sortCrsEntries(ia, ja, val, rowMapP->lib());
-
-    // Set the values of the prolongation operators into the CrsMatrix P and call FillComplete
-    PCrs->setAllValues(iaP, jaP, valP);
-    PCrs->expertStaticFillComplete(domainMapP,rowMapP);
+  // Sort all row's column indicies and entries by LID
+  Xpetra::CrsMatrixUtils<SC, LO, GO, NO>::sortCrsEntries(ia, ja, val, rowMapP->lib());
 
-    // set StridingInformation of P
-    if (A->IsView("stridedMaps") == true) {
-      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMapP);
-    } else {
-      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMapP);
-    }
+  // Set the values of the prolongation operators into the CrsMatrix P and call FillComplete
+  PCrs->setAllValues(iaP, jaP, valP);
+  PCrs->expertStaticFillComplete(domainMapP, rowMapP);
 
-    // store the transfer operator and node coordinates on coarse level
-    Set(coarseLevel, "P", P);
-    Set(coarseLevel, "coarseCoordinates", coarseCoordinates);
-    Set<Array<GO> >(coarseLevel, "gCoarseNodesPerDim", gCoarseNodesPerDir);
-    Set<Array<LO> >(coarseLevel, "lCoarseNodesPerDim", lCoarseNodesPerDir);
+  // set StridingInformation of P
+  if (A->IsView("stridedMaps") == true) {
+    P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMapP);
+  } else {
+    P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMapP);
+  }
 
+  // store the transfer operator and node coordinates on coarse level
+  Set(coarseLevel, "P", P);
+  Set(coarseLevel, "coarseCoordinates", coarseCoordinates);
+  Set<Array<GO> >(coarseLevel, "gCoarseNodesPerDim", gCoarseNodesPerDir);
+  Set<Array<LO> >(coarseLevel, "lCoarseNodesPerDim", lCoarseNodesPerDir);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetGeometricData(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coordinates,
+                     const Array<LO> coarseRate, const Array<GO> gFineNodesPerDir,
+                     const Array<LO> lFineNodesPerDir, const LO BlkSize, Array<GO>& gIndices,
+                     Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
+                     Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir,
+                     Array<LO>& glCoarseNodesPerDir, Array<GO>& ghostGIDs, Array<GO>& coarseNodesGIDs,
+                     Array<GO>& colGIDs, GO& gNumCoarseNodes, LO& lNumCoarseNodes,
+                     ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes, Array<int>& boundaryFlags,
+                     RCP<NodesIDs> ghostedCoarseNodes) const {
+  // This function is extracting the geometric information from the coordinates
+  // and creates the necessary data/formatting to perform locally the calculation
+  // of the pronlongator.
+  //
+  // inputs:
+
+  RCP<const Map> coordinatesMap = coordinates->getMap();
+  LO numDimensions              = coordinates->getNumVectors();
+
+  // Using the coarsening rate and the fine level data,
+  // compute coarse level data
+
+  //                              Phase 1                               //
+  // ------------------------------------------------------------------ //
+  // We first start by finding small informations on the mesh such as   //
+  // the number of coarse nodes (local and global) and the number of    //
+  // ghost nodes / the end rate of coarsening.                          //
+  // ------------------------------------------------------------------ //
+  GO minGlobalIndex = coordinatesMap->getMinGlobalIndex();
+  {
+    GO tmp;
+    gIndices[2] = minGlobalIndex / (gFineNodesPerDir[1] * gFineNodesPerDir[0]);
+    tmp         = minGlobalIndex % (gFineNodesPerDir[1] * gFineNodesPerDir[0]);
+    gIndices[1] = tmp / gFineNodesPerDir[0];
+    gIndices[0] = tmp % gFineNodesPerDir[0];
+
+    myOffset[2] = gIndices[2] % coarseRate[2];
+    myOffset[1] = gIndices[1] % coarseRate[1];
+    myOffset[0] = gIndices[0] % coarseRate[0];
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetGeometricData(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> >& coordinates,
-                   const Array<LO> coarseRate, const Array<GO> gFineNodesPerDir,
-                   const Array<LO> lFineNodesPerDir, const LO BlkSize, Array<GO>& gIndices,
-                   Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
-                   Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir,
-                   Array<LO>& glCoarseNodesPerDir, Array<GO>& ghostGIDs, Array<GO>& coarseNodesGIDs,
-                   Array<GO>& colGIDs, GO& gNumCoarseNodes, LO& lNumCoarseNodes,
-                   ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes, Array<int>& boundaryFlags,
-                   RCP<NodesIDs> ghostedCoarseNodes) const {
-    // This function is extracting the geometric information from the coordinates
-    // and creates the necessary data/formatting to perform locally the calculation
-    // of the pronlongator.
-    //
-    // inputs:
-
-    RCP<const Map> coordinatesMap = coordinates->getMap();
-    LO numDimensions  = coordinates->getNumVectors();
-
-    // Using the coarsening rate and the fine level data,
-    // compute coarse level data
-
-    //                              Phase 1                               //
-    // ------------------------------------------------------------------ //
-    // We first start by finding small informations on the mesh such as   //
-    // the number of coarse nodes (local and global) and the number of    //
-    // ghost nodes / the end rate of coarsening.                          //
-    // ------------------------------------------------------------------ //
-    GO minGlobalIndex = coordinatesMap->getMinGlobalIndex();
-    {
-      GO tmp;
-      gIndices[2] = minGlobalIndex / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-      tmp         = minGlobalIndex % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-      gIndices[1] = tmp / gFineNodesPerDir[0];
-      gIndices[0] = tmp % gFineNodesPerDir[0];
-
-      myOffset[2] = gIndices[2] % coarseRate[2];
-      myOffset[1] = gIndices[1] % coarseRate[1];
-      myOffset[0] = gIndices[0] % coarseRate[0];
+  for (int dim = 0; dim < 3; ++dim) {
+    if (gIndices[dim] == 0) {
+      boundaryFlags[dim] += 1;
     }
+    if (gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim]) {
+      boundaryFlags[dim] += 2;
+    }
+  }
 
-    for(int dim = 0; dim < 3; ++dim) {
-      if(gIndices[dim] == 0) {
-        boundaryFlags[dim] += 1;
-      }
-      if(gIndices[dim] + lFineNodesPerDir[dim] == gFineNodesPerDir[dim]) {
-        boundaryFlags[dim] += 2;
-      }
+  // Check whether ghost nodes are needed in each direction
+  for (LO i = 0; i < numDimensions; ++i) {
+    if ((gIndices[i] != 0) && (gIndices[i] % coarseRate[i] > 0)) {
+      ghostInterface[2 * i] = true;
+    }
+    if (((gIndices[i] + lFineNodesPerDir[i]) != gFineNodesPerDir[i]) && ((gIndices[i] + lFineNodesPerDir[i] - 1) % coarseRate[i] > 0)) {
+      ghostInterface[2 * i + 1] = true;
     }
+  }
 
-    // Check whether ghost nodes are needed in each direction
-    for(LO i=0; i < numDimensions; ++i) {
-      if((gIndices[i] != 0) && (gIndices[i] % coarseRate[i] > 0)) {
-        ghostInterface[2*i] = true;
+  for (LO i = 0; i < 3; ++i) {
+    if (i < numDimensions) {
+      lCoarseNodesPerDir[i] = (lFineNodesPerDir[i] + myOffset[i] - 1) / coarseRate[i];
+      if (myOffset[i] == 0) {
+        ++lCoarseNodesPerDir[i];
       }
-      if(((gIndices[i] + lFineNodesPerDir[i]) != gFineNodesPerDir[i])
-         && ((gIndices[i] + lFineNodesPerDir[i] - 1) % coarseRate[i] > 0)) {
-        ghostInterface[2*i + 1] = true;
+      gCoarseNodesPerDir[i] = (gFineNodesPerDir[i] - 1) / coarseRate[i];
+      endRate[i]            = (gFineNodesPerDir[i] - 1) % coarseRate[i];
+      if (endRate[i] == 0) {
+        ++gCoarseNodesPerDir[i];
+        endRate[i] = coarseRate[i];
       }
+    } else {
+      // Most quantities need to be set to 1 for extra dimensions
+      // this is rather logical, an x-y plane is like a single layer
+      // of nodes in the z direction...
+      gCoarseNodesPerDir[i] = 1;
+      lCoarseNodesPerDir[i] = 1;
+      endRate[i]            = 1;
     }
+  }
 
-    for(LO i = 0; i < 3; ++i) {
-      if(i < numDimensions) {
-        lCoarseNodesPerDir[i] = (lFineNodesPerDir[i] + myOffset[i] - 1) / coarseRate[i];
-        if(myOffset[i] == 0) { ++lCoarseNodesPerDir[i]; }
-        gCoarseNodesPerDir[i] = (gFineNodesPerDir[i] - 1) / coarseRate[i];
-        endRate[i]            = (gFineNodesPerDir[i] - 1) % coarseRate[i];
-        if(endRate[i] == 0) {
-          ++gCoarseNodesPerDir[i];
-          endRate[i] = coarseRate[i];
-        }
+  gNumCoarseNodes = gCoarseNodesPerDir[0] * gCoarseNodesPerDir[1] * gCoarseNodesPerDir[2];
+  lNumCoarseNodes = lCoarseNodesPerDir[0] * lCoarseNodesPerDir[1] * lCoarseNodesPerDir[2];
+
+  // For each direction, determine how many ghost points are required.
+  LO lNumGhostNodes = 0;
+  Array<GO> startGhostedCoarseNode(3);
+  {
+    const int complementaryIndices[3][2] = {{1, 2}, {0, 2}, {0, 1}};
+    LO tmp                               = 0;
+    for (LO i = 0; i < 3; ++i) {
+      // The first branch of this if-statement will be used if the rank contains only one layer
+      // of nodes in direction i, that layer must also coincide with the boundary of the mesh
+      // and coarseRate[i] == endRate[i]...
+      if ((gIndices[i] == gFineNodesPerDir[i] - 1) && (gIndices[i] % coarseRate[i] == 0)) {
+        startGhostedCoarseNode[i] = gIndices[i] / coarseRate[i] - 1;
       } else {
-        // Most quantities need to be set to 1 for extra dimensions
-        // this is rather logical, an x-y plane is like a single layer
-        // of nodes in the z direction...
-        gCoarseNodesPerDir[i] = 1;
-        lCoarseNodesPerDir[i] = 1;
-        endRate[i]            = 1;
+        startGhostedCoarseNode[i] = gIndices[i] / coarseRate[i];
       }
-    }
-
-    gNumCoarseNodes = gCoarseNodesPerDir[0]*gCoarseNodesPerDir[1]*gCoarseNodesPerDir[2];
-    lNumCoarseNodes = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];
-
-    // For each direction, determine how many ghost points are required.
-    LO lNumGhostNodes = 0;
-    Array<GO> startGhostedCoarseNode(3);
-    {
-      const int complementaryIndices[3][2] = {{1,2}, {0,2}, {0,1}};
-      LO tmp = 0;
-      for(LO i = 0; i < 3; ++i) {
-        // The first branch of this if-statement will be used if the rank contains only one layer
-        // of nodes in direction i, that layer must also coincide with the boundary of the mesh
-        // and coarseRate[i] == endRate[i]...
-        if((gIndices[i] == gFineNodesPerDir[i] - 1) && (gIndices[i] % coarseRate[i] == 0)) {
-          startGhostedCoarseNode[i] = gIndices[i] / coarseRate[i] - 1;
-        } else {
-          startGhostedCoarseNode[i] = gIndices[i] / coarseRate[i];
+      // First we load the number of locally owned coarse nodes
+      glCoarseNodesPerDir[i] = lCoarseNodesPerDir[i];
+
+      // Check whether a face in direction i needs ghost nodes
+      if (ghostInterface[2 * i] || ghostInterface[2 * i + 1]) {
+        ++glCoarseNodesPerDir[i];
+        if (i == 0) {
+          tmp = lCoarseNodesPerDir[1] * lCoarseNodesPerDir[2];
         }
-        // First we load the number of locally owned coarse nodes
-        glCoarseNodesPerDir[i] = lCoarseNodesPerDir[i];
-
-        // Check whether a face in direction i needs ghost nodes
-        if(ghostInterface[2*i] || ghostInterface[2*i+1]) {
-          ++glCoarseNodesPerDir[i];
-          if(i == 0) {tmp = lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];}
-          if(i == 1) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[2];}
-          if(i == 2) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1];}
+        if (i == 1) {
+          tmp = lCoarseNodesPerDir[0] * lCoarseNodesPerDir[2];
         }
-        // If both faces in direction i need nodes, double the number of ghost nodes
-        if(ghostInterface[2*i] && ghostInterface[2*i+1]) {
-          ++glCoarseNodesPerDir[i];
-          tmp = 2*tmp;
+        if (i == 2) {
+          tmp = lCoarseNodesPerDir[0] * lCoarseNodesPerDir[1];
         }
-        lNumGhostNodes += tmp;
-
-        // The corners and edges need to be checked in 2D / 3D to add more ghosts nodes
-        for(LO j = 0; j < 2; ++j) {
-          for(LO k = 0; k < 2; ++k) {
-            // Check if two adjoining faces need ghost nodes and then add their common edge
-            if(ghostInterface[2*complementaryIndices[i][0]+j]
-               && ghostInterface[2*complementaryIndices[i][1]+k]) {
-              lNumGhostNodes += lCoarseNodesPerDir[i];
-              // Add corners if three adjoining faces need ghost nodes,
-              // but add them only once! Hence when i == 0.
-              if(ghostInterface[2*i] && (i == 0)) { lNumGhostNodes += 1; }
-              if(ghostInterface[2*i+1] && (i == 0)) { lNumGhostNodes += 1; }
+      }
+      // If both faces in direction i need nodes, double the number of ghost nodes
+      if (ghostInterface[2 * i] && ghostInterface[2 * i + 1]) {
+        ++glCoarseNodesPerDir[i];
+        tmp = 2 * tmp;
+      }
+      lNumGhostNodes += tmp;
+
+      // The corners and edges need to be checked in 2D / 3D to add more ghosts nodes
+      for (LO j = 0; j < 2; ++j) {
+        for (LO k = 0; k < 2; ++k) {
+          // Check if two adjoining faces need ghost nodes and then add their common edge
+          if (ghostInterface[2 * complementaryIndices[i][0] + j] && ghostInterface[2 * complementaryIndices[i][1] + k]) {
+            lNumGhostNodes += lCoarseNodesPerDir[i];
+            // Add corners if three adjoining faces need ghost nodes,
+            // but add them only once! Hence when i == 0.
+            if (ghostInterface[2 * i] && (i == 0)) {
+              lNumGhostNodes += 1;
+            }
+            if (ghostInterface[2 * i + 1] && (i == 0)) {
+              lNumGhostNodes += 1;
             }
           }
         }
-        tmp = 0;
       }
-    } // end of scope for tmp and complementaryIndices
-
-    const LO lNumGhostedNodes = glCoarseNodesPerDir[0]*glCoarseNodesPerDir[1]
-      *glCoarseNodesPerDir[2];
-    ghostedCoarseNodes->PIDs.resize(lNumGhostedNodes);
-    ghostedCoarseNodes->LIDs.resize(lNumGhostedNodes);
-    ghostedCoarseNodes->GIDs.resize(lNumGhostedNodes);
-    ghostedCoarseNodes->coarseGIDs.resize(lNumGhostedNodes);
-    ghostedCoarseNodes->colInds.resize(lNumGhostedNodes);
-
-    // We loop over all ghosted coarse nodes by increasing global lexicographic order
-    Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3), ijk(3);
-    LO currentIndex = -1;
-    for(ijk[2] = 0; ijk[2] < glCoarseNodesPerDir[2]; ++ijk[2]) {
-      for(ijk[1] = 0; ijk[1] < glCoarseNodesPerDir[1]; ++ijk[1]) {
-        for(ijk[0] = 0; ijk[0] < glCoarseNodesPerDir[0]; ++ijk[0]) {
-          currentIndex = ijk[2]*glCoarseNodesPerDir[1]*glCoarseNodesPerDir[0]
-            + ijk[1]*glCoarseNodesPerDir[0] + ijk[0];
-          coarseNodeCoarseIndices[0] = startGhostedCoarseNode[0] + ijk[0];
-          coarseNodeFineIndices[0] = coarseNodeCoarseIndices[0]*coarseRate[0];
-          if(coarseNodeFineIndices[0] > gFineNodesPerDir[0] - 1) {
-            coarseNodeFineIndices[0] = gFineNodesPerDir[0] - 1;
-          }
-          coarseNodeCoarseIndices[1] = startGhostedCoarseNode[1] + ijk[1];
-          coarseNodeFineIndices[1] = coarseNodeCoarseIndices[1]*coarseRate[1];
-          if(coarseNodeFineIndices[1] > gFineNodesPerDir[1] - 1) {
-            coarseNodeFineIndices[1] = gFineNodesPerDir[1] - 1;
-          }
-          coarseNodeCoarseIndices[2] = startGhostedCoarseNode[2] + ijk[2];
-          coarseNodeFineIndices[2] = coarseNodeCoarseIndices[2]*coarseRate[2];
-          if(coarseNodeFineIndices[2] > gFineNodesPerDir[2] - 1) {
-            coarseNodeFineIndices[2] = gFineNodesPerDir[2] - 1;
-          }
-          GO myGID = 0, myCoarseGID = -1;
-          GO factor[3] = {};
-          factor[2] = gFineNodesPerDir[1]*gFineNodesPerDir[0];
-          factor[1] = gFineNodesPerDir[0];
-          factor[0] = 1;
-          for(int dim = 0; dim < 3; ++dim) {
-            if(dim < numDimensions) {
-              if(gIndices[dim] - myOffset[dim] + ijk[dim]*coarseRate[dim]
-                 < gFineNodesPerDir[dim] - 1) {
-                myGID += (gIndices[dim] - myOffset[dim]
-                          + ijk[dim]*coarseRate[dim])*factor[dim];
-              } else {
-                myGID += (gIndices[dim] - myOffset[dim]
-                          + (ijk[dim] - 1)*coarseRate[dim] + endRate[dim])*factor[dim];
-              }
+      tmp = 0;
+    }
+  }  // end of scope for tmp and complementaryIndices
+
+  const LO lNumGhostedNodes = glCoarseNodesPerDir[0] * glCoarseNodesPerDir[1] * glCoarseNodesPerDir[2];
+  ghostedCoarseNodes->PIDs.resize(lNumGhostedNodes);
+  ghostedCoarseNodes->LIDs.resize(lNumGhostedNodes);
+  ghostedCoarseNodes->GIDs.resize(lNumGhostedNodes);
+  ghostedCoarseNodes->coarseGIDs.resize(lNumGhostedNodes);
+  ghostedCoarseNodes->colInds.resize(lNumGhostedNodes);
+
+  // We loop over all ghosted coarse nodes by increasing global lexicographic order
+  Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3), ijk(3);
+  LO currentIndex = -1;
+  for (ijk[2] = 0; ijk[2] < glCoarseNodesPerDir[2]; ++ijk[2]) {
+    for (ijk[1] = 0; ijk[1] < glCoarseNodesPerDir[1]; ++ijk[1]) {
+      for (ijk[0] = 0; ijk[0] < glCoarseNodesPerDir[0]; ++ijk[0]) {
+        currentIndex               = ijk[2] * glCoarseNodesPerDir[1] * glCoarseNodesPerDir[0] + ijk[1] * glCoarseNodesPerDir[0] + ijk[0];
+        coarseNodeCoarseIndices[0] = startGhostedCoarseNode[0] + ijk[0];
+        coarseNodeFineIndices[0]   = coarseNodeCoarseIndices[0] * coarseRate[0];
+        if (coarseNodeFineIndices[0] > gFineNodesPerDir[0] - 1) {
+          coarseNodeFineIndices[0] = gFineNodesPerDir[0] - 1;
+        }
+        coarseNodeCoarseIndices[1] = startGhostedCoarseNode[1] + ijk[1];
+        coarseNodeFineIndices[1]   = coarseNodeCoarseIndices[1] * coarseRate[1];
+        if (coarseNodeFineIndices[1] > gFineNodesPerDir[1] - 1) {
+          coarseNodeFineIndices[1] = gFineNodesPerDir[1] - 1;
+        }
+        coarseNodeCoarseIndices[2] = startGhostedCoarseNode[2] + ijk[2];
+        coarseNodeFineIndices[2]   = coarseNodeCoarseIndices[2] * coarseRate[2];
+        if (coarseNodeFineIndices[2] > gFineNodesPerDir[2] - 1) {
+          coarseNodeFineIndices[2] = gFineNodesPerDir[2] - 1;
+        }
+        GO myGID = 0, myCoarseGID = -1;
+        GO factor[3] = {};
+        factor[2]    = gFineNodesPerDir[1] * gFineNodesPerDir[0];
+        factor[1]    = gFineNodesPerDir[0];
+        factor[0]    = 1;
+        for (int dim = 0; dim < 3; ++dim) {
+          if (dim < numDimensions) {
+            if (gIndices[dim] - myOffset[dim] + ijk[dim] * coarseRate[dim] < gFineNodesPerDir[dim] - 1) {
+              myGID += (gIndices[dim] - myOffset[dim] + ijk[dim] * coarseRate[dim]) * factor[dim];
+            } else {
+              myGID += (gIndices[dim] - myOffset[dim] + (ijk[dim] - 1) * coarseRate[dim] + endRate[dim]) * factor[dim];
             }
           }
-          myCoarseGID = coarseNodeCoarseIndices[0]
-            + coarseNodeCoarseIndices[1]*gCoarseNodesPerDir[0]
-            + coarseNodeCoarseIndices[2]*gCoarseNodesPerDir[1]*gCoarseNodesPerDir[0];
-          ghostedCoarseNodes->GIDs[currentIndex] = myGID;
-          ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
         }
+        myCoarseGID                                  = coarseNodeCoarseIndices[0] + coarseNodeCoarseIndices[1] * gCoarseNodesPerDir[0] + coarseNodeCoarseIndices[2] * gCoarseNodesPerDir[1] * gCoarseNodesPerDir[0];
+        ghostedCoarseNodes->GIDs[currentIndex]       = myGID;
+        ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
       }
     }
-    coordinatesMap->getRemoteIndexList(ghostedCoarseNodes->GIDs(),
-                                       ghostedCoarseNodes->PIDs(),
-                                       ghostedCoarseNodes->LIDs());
-
-    //                              Phase 2                               //
-    // ------------------------------------------------------------------ //
-    // Build a list of GIDs to import the required ghost nodes.           //
-    // The ordering of the ghosts nodes will be as natural as possible,   //
-    // i.e. it should follow the GID ordering of the mesh.                //
-    // ------------------------------------------------------------------ //
-
-    // Saddly we have to more or less redo what was just done to figure out the value of
-    // lNumGhostNodes, there might be some optimization possibility here...
-    ghostGIDs.resize(lNumGhostNodes);
-    LO countGhosts = 0;
-    // Get the GID of the first point on the current partition.
-    GO startingGID = minGlobalIndex;
-    Array<GO> startingIndices(3);
-    // We still want ghost nodes even if have with a 0 offset,
-    // except when we are on a boundary
-    if(ghostInterface[4] && (myOffset[2] == 0)) {
-      if(gIndices[2] + coarseRate[2] > gFineNodesPerDir[2]) {
-        startingGID -= endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-      } else {
-        startingGID -= coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-      }
+  }
+  coordinatesMap->getRemoteIndexList(ghostedCoarseNodes->GIDs(),
+                                     ghostedCoarseNodes->PIDs(),
+                                     ghostedCoarseNodes->LIDs());
+
+  //                              Phase 2                               //
+  // ------------------------------------------------------------------ //
+  // Build a list of GIDs to import the required ghost nodes.           //
+  // The ordering of the ghosts nodes will be as natural as possible,   //
+  // i.e. it should follow the GID ordering of the mesh.                //
+  // ------------------------------------------------------------------ //
+
+  // Saddly we have to more or less redo what was just done to figure out the value of
+  // lNumGhostNodes, there might be some optimization possibility here...
+  ghostGIDs.resize(lNumGhostNodes);
+  LO countGhosts = 0;
+  // Get the GID of the first point on the current partition.
+  GO startingGID = minGlobalIndex;
+  Array<GO> startingIndices(3);
+  // We still want ghost nodes even if have with a 0 offset,
+  // except when we are on a boundary
+  if (ghostInterface[4] && (myOffset[2] == 0)) {
+    if (gIndices[2] + coarseRate[2] > gFineNodesPerDir[2]) {
+      startingGID -= endRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
     } else {
-      startingGID -= myOffset[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+      startingGID -= coarseRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
     }
-    if(ghostInterface[2] && (myOffset[1] == 0)) {
-      if(gIndices[1] + coarseRate[1] > gFineNodesPerDir[1]) {
-        startingGID -= endRate[1]*gFineNodesPerDir[0];
-      } else {
-        startingGID -= coarseRate[1]*gFineNodesPerDir[0];
-      }
+  } else {
+    startingGID -= myOffset[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+  }
+  if (ghostInterface[2] && (myOffset[1] == 0)) {
+    if (gIndices[1] + coarseRate[1] > gFineNodesPerDir[1]) {
+      startingGID -= endRate[1] * gFineNodesPerDir[0];
     } else {
-      startingGID -= myOffset[1]*gFineNodesPerDir[0];
+      startingGID -= coarseRate[1] * gFineNodesPerDir[0];
     }
-    if(ghostInterface[0] && (myOffset[0] == 0)) {
-      if(gIndices[0] + coarseRate[0] > gFineNodesPerDir[0]) {
-        startingGID -= endRate[0];
-      } else {
-        startingGID -= coarseRate[0];
-      }
+  } else {
+    startingGID -= myOffset[1] * gFineNodesPerDir[0];
+  }
+  if (ghostInterface[0] && (myOffset[0] == 0)) {
+    if (gIndices[0] + coarseRate[0] > gFineNodesPerDir[0]) {
+      startingGID -= endRate[0];
     } else {
-      startingGID -= myOffset[0];
+      startingGID -= coarseRate[0];
     }
+  } else {
+    startingGID -= myOffset[0];
+  }
 
-    { // scope for tmp
-      GO tmp;
-      startingIndices[2] = startingGID / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-      tmp = startingGID % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-      startingIndices[1] = tmp / gFineNodesPerDir[0];
-      startingIndices[0] = tmp % gFineNodesPerDir[0];
-    }
+  {  // scope for tmp
+    GO tmp;
+    startingIndices[2] = startingGID / (gFineNodesPerDir[1] * gFineNodesPerDir[0]);
+    tmp                = startingGID % (gFineNodesPerDir[1] * gFineNodesPerDir[0]);
+    startingIndices[1] = tmp / gFineNodesPerDir[0];
+    startingIndices[0] = tmp % gFineNodesPerDir[0];
+  }
 
-    GO ghostOffset[3] = {0, 0, 0};
-    LO lengthZero  = lCoarseNodesPerDir[0];
-    LO lengthOne   = lCoarseNodesPerDir[1];
-    LO lengthTwo   = lCoarseNodesPerDir[2];
-    if(ghostInterface[0]) {++lengthZero;}
-    if(ghostInterface[1]) {++lengthZero;}
-    if(ghostInterface[2]) {++lengthOne;}
-    if(ghostInterface[3]) {++lengthOne;}
-    if(ghostInterface[4]) {++lengthTwo;}
-    if(ghostInterface[5]) {++lengthTwo;}
-
-    // First check the bottom face as it will have the lowest GIDs
-    if(ghostInterface[4]) {
-      ghostOffset[2] = startingGID;
-      for(LO j = 0; j < lengthOne; ++j) {
-        if( (j == lengthOne-1)
-            && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-          ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
+  GO ghostOffset[3] = {0, 0, 0};
+  LO lengthZero     = lCoarseNodesPerDir[0];
+  LO lengthOne      = lCoarseNodesPerDir[1];
+  LO lengthTwo      = lCoarseNodesPerDir[2];
+  if (ghostInterface[0]) {
+    ++lengthZero;
+  }
+  if (ghostInterface[1]) {
+    ++lengthZero;
+  }
+  if (ghostInterface[2]) {
+    ++lengthOne;
+  }
+  if (ghostInterface[3]) {
+    ++lengthOne;
+  }
+  if (ghostInterface[4]) {
+    ++lengthTwo;
+  }
+  if (ghostInterface[5]) {
+    ++lengthTwo;
+  }
+
+  // First check the bottom face as it will have the lowest GIDs
+  if (ghostInterface[4]) {
+    ghostOffset[2] = startingGID;
+    for (LO j = 0; j < lengthOne; ++j) {
+      if ((j == lengthOne - 1) && (startingIndices[1] + j * coarseRate[1] + 1 > gFineNodesPerDir[1])) {
+        ghostOffset[1] = ((j - 1) * coarseRate[1] + endRate[1]) * gFineNodesPerDir[0];
+      } else {
+        ghostOffset[1] = j * coarseRate[1] * gFineNodesPerDir[0];
+      }
+      for (LO k = 0; k < lengthZero; ++k) {
+        if ((k == lengthZero - 1) && (startingIndices[0] + k * coarseRate[0] + 1 > gFineNodesPerDir[0])) {
+          ghostOffset[0] = (k - 1) * coarseRate[0] + endRate[0];
         } else {
-          ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-        }
-        for(LO k = 0; k < lengthZero; ++k) {
-          if( (k == lengthZero-1)
-              && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-            ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-          } else {
-            ghostOffset[0] = k*coarseRate[0];
-          }
-          // If the partition includes a changed rate at the edge, ghost nodes need to be picked
-          // carefully.
-          // This if statement is repeated each time a ghost node is selected.
-          ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-          ++countGhosts;
+          ghostOffset[0] = k * coarseRate[0];
         }
+        // If the partition includes a changed rate at the edge, ghost nodes need to be picked
+        // carefully.
+        // This if statement is repeated each time a ghost node is selected.
+        ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+        ++countGhosts;
       }
     }
+  }
 
-    // Sweep over the lCoarseNodesPerDir[2] coarse layers in direction 2 and gather necessary ghost
-    // nodes located on these layers.
-    for(LO i = 0; i < lengthTwo; ++i) {
-      // Exclude the cases where ghost nodes exists on the faces in directions 2, these faces are
-      // swept seperatly for efficiency.
-      if( !((i == lengthTwo-1) && ghostInterface[5]) && !((i == 0) && ghostInterface[4]) ) {
-        // Set the ghostOffset in direction 2 taking into account a possible endRate different
-        // from the regular coarseRate.
-        if( (i == lengthTwo-1) && !ghostInterface[5] ) {
-          ghostOffset[2] = startingGID + ((i-1)*coarseRate[2] + endRate[2])
-            *gFineNodesPerDir[1]*gFineNodesPerDir[0];
-        } else {
-          ghostOffset[2] = startingGID + i*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-        }
-        for(LO j = 0; j < lengthOne; ++j) {
-          if( (j == 0) && ghostInterface[2] ) {
-            for(LO k = 0; k < lengthZero; ++k) {
-              if( (k == lengthZero-1)
-                  && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-                if(k == 0) {
-                  ghostOffset[0] = endRate[0];
-                } else {
-                  ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-                }
+  // Sweep over the lCoarseNodesPerDir[2] coarse layers in direction 2 and gather necessary ghost
+  // nodes located on these layers.
+  for (LO i = 0; i < lengthTwo; ++i) {
+    // Exclude the cases where ghost nodes exists on the faces in directions 2, these faces are
+    // swept seperatly for efficiency.
+    if (!((i == lengthTwo - 1) && ghostInterface[5]) && !((i == 0) && ghostInterface[4])) {
+      // Set the ghostOffset in direction 2 taking into account a possible endRate different
+      // from the regular coarseRate.
+      if ((i == lengthTwo - 1) && !ghostInterface[5]) {
+        ghostOffset[2] = startingGID + ((i - 1) * coarseRate[2] + endRate[2]) * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+      } else {
+        ghostOffset[2] = startingGID + i * coarseRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+      }
+      for (LO j = 0; j < lengthOne; ++j) {
+        if ((j == 0) && ghostInterface[2]) {
+          for (LO k = 0; k < lengthZero; ++k) {
+            if ((k == lengthZero - 1) && (startingIndices[0] + k * coarseRate[0] + 1 > gFineNodesPerDir[0])) {
+              if (k == 0) {
+                ghostOffset[0] = endRate[0];
               } else {
-                ghostOffset[0] = k*coarseRate[0];
+                ghostOffset[0] = (k - 1) * coarseRate[0] + endRate[0];
               }
-              // In this case j == 0 so ghostOffset[1] is zero and can be ignored
-              ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[0];
-              ++countGhosts;
-            }
-          } else if( (j == lengthOne-1) && ghostInterface[3] ) {
-            // Set the ghostOffset in direction 1 taking into account a possible endRate different
-            // from the regular coarseRate.
-            if( (j == lengthOne-1)
-                && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-              ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
             } else {
-              ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-            }
-            for(LO k = 0; k < lengthZero; ++k) {
-              if( (k == lengthZero-1)
-                  && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-                ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-              } else {
-                ghostOffset[0] = k*coarseRate[0];
-              }
-              ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-              ++countGhosts;
+              ghostOffset[0] = k * coarseRate[0];
             }
+            // In this case j == 0 so ghostOffset[1] is zero and can be ignored
+            ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[0];
+            ++countGhosts;
+          }
+        } else if ((j == lengthOne - 1) && ghostInterface[3]) {
+          // Set the ghostOffset in direction 1 taking into account a possible endRate different
+          // from the regular coarseRate.
+          if ((j == lengthOne - 1) && (startingIndices[1] + j * coarseRate[1] + 1 > gFineNodesPerDir[1])) {
+            ghostOffset[1] = ((j - 1) * coarseRate[1] + endRate[1]) * gFineNodesPerDir[0];
           } else {
-            // Set ghostOffset[1] for side faces sweep
-            if( (j == lengthOne-1)
-                && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-              ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
+            ghostOffset[1] = j * coarseRate[1] * gFineNodesPerDir[0];
+          }
+          for (LO k = 0; k < lengthZero; ++k) {
+            if ((k == lengthZero - 1) && (startingIndices[0] + k * coarseRate[0] + 1 > gFineNodesPerDir[0])) {
+              ghostOffset[0] = (k - 1) * coarseRate[0] + endRate[0];
             } else {
-              ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
+              ghostOffset[0] = k * coarseRate[0];
             }
+            ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+            ++countGhosts;
+          }
+        } else {
+          // Set ghostOffset[1] for side faces sweep
+          if ((j == lengthOne - 1) && (startingIndices[1] + j * coarseRate[1] + 1 > gFineNodesPerDir[1])) {
+            ghostOffset[1] = ((j - 1) * coarseRate[1] + endRate[1]) * gFineNodesPerDir[0];
+          } else {
+            ghostOffset[1] = j * coarseRate[1] * gFineNodesPerDir[0];
+          }
 
-            // Set ghostOffset[0], ghostGIDs and countGhosts
-            if(ghostInterface[0]) { // In that case ghostOffset[0]==0, so we can ignore it
-              ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1];
-              ++countGhosts;
-            }
-            if(ghostInterface[1]) { // Grab ghost point at the end of direction 0.
-              if( (startingIndices[0] + (lengthZero-1)*coarseRate[0]) > gFineNodesPerDir[0] - 1 ) {
-                if(lengthZero > 2) {
-                  ghostOffset[0] = (lengthZero-2)*coarseRate[0] + endRate[0];
-                } else {
-                  ghostOffset[0] = endRate[0];
-                }
+          // Set ghostOffset[0], ghostGIDs and countGhosts
+          if (ghostInterface[0]) {  // In that case ghostOffset[0]==0, so we can ignore it
+            ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1];
+            ++countGhosts;
+          }
+          if (ghostInterface[1]) {  // Grab ghost point at the end of direction 0.
+            if ((startingIndices[0] + (lengthZero - 1) * coarseRate[0]) > gFineNodesPerDir[0] - 1) {
+              if (lengthZero > 2) {
+                ghostOffset[0] = (lengthZero - 2) * coarseRate[0] + endRate[0];
               } else {
-                ghostOffset[0] = (lengthZero-1)*coarseRate[0];
+                ghostOffset[0] = endRate[0];
               }
-              ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-              ++countGhosts;
+            } else {
+              ghostOffset[0] = (lengthZero - 1) * coarseRate[0];
             }
+            ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+            ++countGhosts;
           }
         }
       }
     }
+  }
 
-    // Finally check the top face
-    if(ghostInterface[5]) {
-      if( startingIndices[2] + (lengthTwo-1)*coarseRate[2] + 1 > gFineNodesPerDir[2] ) {
-        ghostOffset[2] = startingGID
-          + ((lengthTwo-2)*coarseRate[2] + endRate[2])*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+  // Finally check the top face
+  if (ghostInterface[5]) {
+    if (startingIndices[2] + (lengthTwo - 1) * coarseRate[2] + 1 > gFineNodesPerDir[2]) {
+      ghostOffset[2] = startingGID + ((lengthTwo - 2) * coarseRate[2] + endRate[2]) * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+    } else {
+      ghostOffset[2] = startingGID + (lengthTwo - 1) * coarseRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+    }
+    for (LO j = 0; j < lengthOne; ++j) {
+      if ((j == lengthOne - 1) && (startingIndices[1] + j * coarseRate[1] + 1 > gFineNodesPerDir[1])) {
+        ghostOffset[1] = ((j - 1) * coarseRate[1] + endRate[1]) * gFineNodesPerDir[0];
       } else {
-        ghostOffset[2] = startingGID
-          + (lengthTwo-1)*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+        ghostOffset[1] = j * coarseRate[1] * gFineNodesPerDir[0];
       }
-      for(LO j = 0; j < lengthOne; ++j) {
-        if( (j == lengthOne-1)
-            && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-          ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
+      for (LO k = 0; k < lengthZero; ++k) {
+        if ((k == lengthZero - 1) && (startingIndices[0] + k * coarseRate[0] + 1 > gFineNodesPerDir[0])) {
+          ghostOffset[0] = (k - 1) * coarseRate[0] + endRate[0];
         } else {
-          ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-        }
-        for(LO k = 0; k < lengthZero; ++k) {
-          if( (k == lengthZero-1)
-              && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-            ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-          } else {
-            ghostOffset[0] = k*coarseRate[0];
-          }
-          ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-          ++countGhosts;
+          ghostOffset[0] = k * coarseRate[0];
         }
+        ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+        ++countGhosts;
+      }
+    }
+  }
+
+  //                              Phase 3                               //
+  // ------------------------------------------------------------------ //
+  // Final phase of this function, lists are being built to create the  //
+  // column and domain maps of the projection as well as the map and    //
+  // arrays for the coarse coordinates multivector.                     //
+  // ------------------------------------------------------------------ //
+
+  Array<GO> gCoarseNodesGIDs(lNumCoarseNodes);
+  LO currentNode, offset2, offset1, offset0;
+  // Find the GIDs of the coarse nodes on the partition.
+  for (LO ind2 = 0; ind2 < lCoarseNodesPerDir[2]; ++ind2) {
+    if (myOffset[2] == 0) {
+      offset2 = startingIndices[2] + myOffset[2];
+    } else {
+      if (startingIndices[2] + endRate[2] == gFineNodesPerDir[2] - 1) {
+        offset2 = startingIndices[2] + endRate[2];
+      } else {
+        offset2 = startingIndices[2] + coarseRate[2];
       }
     }
+    if (offset2 + ind2 * coarseRate[2] > gFineNodesPerDir[2] - 1) {
+      offset2 += (ind2 - 1) * coarseRate[2] + endRate[2];
+    } else {
+      offset2 += ind2 * coarseRate[2];
+    }
+    offset2 = offset2 * gFineNodesPerDir[1] * gFineNodesPerDir[0];
 
-    //                              Phase 3                               //
-    // ------------------------------------------------------------------ //
-    // Final phase of this function, lists are being built to create the  //
-    // column and domain maps of the projection as well as the map and    //
-    // arrays for the coarse coordinates multivector.                     //
-    // ------------------------------------------------------------------ //
-
-    Array<GO> gCoarseNodesGIDs(lNumCoarseNodes);
-    LO currentNode, offset2, offset1, offset0;
-    // Find the GIDs of the coarse nodes on the partition.
-    for(LO ind2 = 0; ind2 < lCoarseNodesPerDir[2]; ++ind2) {
-      if(myOffset[2] == 0) {
-        offset2 = startingIndices[2] + myOffset[2];
+    for (LO ind1 = 0; ind1 < lCoarseNodesPerDir[1]; ++ind1) {
+      if (myOffset[1] == 0) {
+        offset1 = startingIndices[1] + myOffset[1];
       } else {
-        if(startingIndices[2] + endRate[2] == gFineNodesPerDir[2] - 1) {
-          offset2 = startingIndices[2] + endRate[2];
+        if (startingIndices[1] + endRate[1] == gFineNodesPerDir[1] - 1) {
+          offset1 = startingIndices[1] + endRate[1];
         } else {
-          offset2 = startingIndices[2] + coarseRate[2];
+          offset1 = startingIndices[1] + coarseRate[1];
         }
       }
-      if(offset2 + ind2*coarseRate[2] > gFineNodesPerDir[2] - 1) {
-        offset2 += (ind2 - 1)*coarseRate[2] + endRate[2];
+      if (offset1 + ind1 * coarseRate[1] > gFineNodesPerDir[1] - 1) {
+        offset1 += (ind1 - 1) * coarseRate[1] + endRate[1];
       } else {
-        offset2 += ind2*coarseRate[2];
+        offset1 += ind1 * coarseRate[1];
       }
-      offset2 = offset2*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-
-      for(LO ind1 = 0; ind1 < lCoarseNodesPerDir[1]; ++ind1) {
-        if(myOffset[1] == 0) {
-          offset1 = startingIndices[1] + myOffset[1];
+      offset1 = offset1 * gFineNodesPerDir[0];
+      for (LO ind0 = 0; ind0 < lCoarseNodesPerDir[0]; ++ind0) {
+        offset0 = startingIndices[0];
+        if (myOffset[0] == 0) {
+          offset0 += ind0 * coarseRate[0];
         } else {
-          if(startingIndices[1] + endRate[1] == gFineNodesPerDir[1] - 1) {
-            offset1 = startingIndices[1] + endRate[1];
-          } else {
-            offset1 = startingIndices[1] + coarseRate[1];
-          }
+          offset0 += (ind0 + 1) * coarseRate[0];
         }
-        if(offset1 + ind1*coarseRate[1] > gFineNodesPerDir[1] - 1) {
-          offset1 += (ind1 - 1)*coarseRate[1] + endRate[1];
-        } else {
-          offset1 += ind1*coarseRate[1];
+        if (offset0 > gFineNodesPerDir[0] - 1) {
+          offset0 += endRate[0] - coarseRate[0];
         }
-        offset1 = offset1*gFineNodesPerDir[0];
-        for(LO ind0 = 0; ind0 < lCoarseNodesPerDir[0]; ++ind0) {
-          offset0 = startingIndices[0];
-          if(myOffset[0] == 0) {
-            offset0 += ind0*coarseRate[0];
-          } else {
-            offset0 += (ind0 + 1)*coarseRate[0];
-          }
-          if(offset0 > gFineNodesPerDir[0] - 1) {offset0 += endRate[0] - coarseRate[0];}
 
-          currentNode = ind2*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]
-                      + ind1*lCoarseNodesPerDir[0]
-                      + ind0;
-          gCoarseNodesGIDs[currentNode] = offset2 + offset1 + offset0;
-        }
+        currentNode                   = ind2 * lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0] + ind1 * lCoarseNodesPerDir[0] + ind0;
+        gCoarseNodesGIDs[currentNode] = offset2 + offset1 + offset0;
       }
     }
+  }
 
-    // Actual loop over all the coarse/ghost nodes to find their index on the coarse mesh
-    // and the corresponding dofs that will need to be added to colMapP.
-    colGIDs.resize(BlkSize*(lNumCoarseNodes+lNumGhostNodes));
-    coarseNodesGIDs.resize(lNumCoarseNodes);
-    for(LO i = 0; i < numDimensions; ++i) {coarseNodes[i].resize(lNumCoarseNodes);}
-    GO fineNodesPerCoarseSlab    = coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-    GO fineNodesEndCoarseSlab    = endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-    GO fineNodesPerCoarsePlane   = coarseRate[1]*gFineNodesPerDir[0];
-    GO fineNodesEndCoarsePlane   = endRate[1]*gFineNodesPerDir[0];
-    GO coarseNodesPerCoarseLayer = gCoarseNodesPerDir[1]*gCoarseNodesPerDir[0];
-    GO gCoarseNodeOnCoarseGridGID;
-    LO gInd[3], lCol;
-    Array<int> ghostPIDs  (lNumGhostNodes);
-    Array<LO>  ghostLIDs  (lNumGhostNodes);
-    Array<LO>  ghostPermut(lNumGhostNodes);
-    for(LO k = 0; k < lNumGhostNodes; ++k) {ghostPermut[k] = k;}
-    coordinatesMap->getRemoteIndexList(ghostGIDs, ghostPIDs, ghostLIDs);
-    sh_sort_permute(ghostPIDs.begin(),ghostPIDs.end(), ghostPermut.begin(),ghostPermut.end());
-
-    { // scope for tmpInds, tmpVars and tmp.
-      GO tmpInds[3], tmpVars[2];
-      LO tmp;
-      // Loop over the coarse nodes of the partition and add them to colGIDs
-      // that will be used to construct the column and domain maps of P as well
-      // as to construct the coarse coordinates map.
-      // for(LO col = 0; col < lNumCoarseNodes; ++col) { // This should most likely be replaced
-      // by loops of lCoarseNodesPerDir[] to simplify arithmetics
-      LO col = 0;
-      LO firstCoarseNodeInds[3], currentCoarseNode;
-      for(LO dim = 0; dim < 3; ++dim) {
-        if(myOffset[dim] == 0) {
-          firstCoarseNodeInds[dim] = 0;
-        } else {
-          firstCoarseNodeInds[dim] = coarseRate[dim] - myOffset[dim];
-        }
+  // Actual loop over all the coarse/ghost nodes to find their index on the coarse mesh
+  // and the corresponding dofs that will need to be added to colMapP.
+  colGIDs.resize(BlkSize * (lNumCoarseNodes + lNumGhostNodes));
+  coarseNodesGIDs.resize(lNumCoarseNodes);
+  for (LO i = 0; i < numDimensions; ++i) {
+    coarseNodes[i].resize(lNumCoarseNodes);
+  }
+  GO fineNodesPerCoarseSlab    = coarseRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+  GO fineNodesEndCoarseSlab    = endRate[2] * gFineNodesPerDir[1] * gFineNodesPerDir[0];
+  GO fineNodesPerCoarsePlane   = coarseRate[1] * gFineNodesPerDir[0];
+  GO fineNodesEndCoarsePlane   = endRate[1] * gFineNodesPerDir[0];
+  GO coarseNodesPerCoarseLayer = gCoarseNodesPerDir[1] * gCoarseNodesPerDir[0];
+  GO gCoarseNodeOnCoarseGridGID;
+  LO gInd[3], lCol;
+  Array<int> ghostPIDs(lNumGhostNodes);
+  Array<LO> ghostLIDs(lNumGhostNodes);
+  Array<LO> ghostPermut(lNumGhostNodes);
+  for (LO k = 0; k < lNumGhostNodes; ++k) {
+    ghostPermut[k] = k;
+  }
+  coordinatesMap->getRemoteIndexList(ghostGIDs, ghostPIDs, ghostLIDs);
+  sh_sort_permute(ghostPIDs.begin(), ghostPIDs.end(), ghostPermut.begin(), ghostPermut.end());
+
+  {  // scope for tmpInds, tmpVars and tmp.
+    GO tmpInds[3], tmpVars[2];
+    LO tmp;
+    // Loop over the coarse nodes of the partition and add them to colGIDs
+    // that will be used to construct the column and domain maps of P as well
+    // as to construct the coarse coordinates map.
+    // for(LO col = 0; col < lNumCoarseNodes; ++col) { // This should most likely be replaced
+    // by loops of lCoarseNodesPerDir[] to simplify arithmetics
+    LO col = 0;
+    LO firstCoarseNodeInds[3], currentCoarseNode;
+    for (LO dim = 0; dim < 3; ++dim) {
+      if (myOffset[dim] == 0) {
+        firstCoarseNodeInds[dim] = 0;
+      } else {
+        firstCoarseNodeInds[dim] = coarseRate[dim] - myOffset[dim];
       }
-      Array<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
-      for(LO dim = 0; dim < numDimensions; ++dim) {fineNodes[dim] = coordinates->getData(dim);}
-      for(LO k = 0; k < lCoarseNodesPerDir[2]; ++k) {
-        for(LO j = 0; j < lCoarseNodesPerDir[1]; ++j) {
-          for(LO i = 0; i < lCoarseNodesPerDir[0]; ++i) {
-            col = k*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0] + j*lCoarseNodesPerDir[0] + i;
-
-            // Check for endRate
-            currentCoarseNode = 0;
-            if(firstCoarseNodeInds[0] + i*coarseRate[0] > lFineNodesPerDir[0] - 1) {
-              currentCoarseNode += firstCoarseNodeInds[0] + (i-1)*coarseRate[0] + endRate[0];
-            } else {
-              currentCoarseNode += firstCoarseNodeInds[0] + i*coarseRate[0];
-            }
-            if(firstCoarseNodeInds[1] + j*coarseRate[1] > lFineNodesPerDir[1] - 1) {
-              currentCoarseNode += (firstCoarseNodeInds[1] + (j-1)*coarseRate[1] + endRate[1])
-                *lFineNodesPerDir[0];
-            } else {
-              currentCoarseNode += (firstCoarseNodeInds[1] + j*coarseRate[1])*lFineNodesPerDir[0];
-            }
-            if(firstCoarseNodeInds[2] + k*coarseRate[2] > lFineNodesPerDir[2] - 1) {
-              currentCoarseNode += (firstCoarseNodeInds[2] + (k-1)*coarseRate[2] + endRate[2])
-                *lFineNodesPerDir[1]*lFineNodesPerDir[0];
-            } else {
-              currentCoarseNode += (firstCoarseNodeInds[2] + k*coarseRate[2])
-                *lFineNodesPerDir[1]*lFineNodesPerDir[0];
-            }
-            // Load coordinates
-            for(LO dim = 0; dim < numDimensions; ++dim) {
-              coarseNodes[dim][col] = fineNodes[dim][currentCoarseNode];
-            }
+    }
+    Array<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
+    for (LO dim = 0; dim < numDimensions; ++dim) {
+      fineNodes[dim] = coordinates->getData(dim);
+    }
+    for (LO k = 0; k < lCoarseNodesPerDir[2]; ++k) {
+      for (LO j = 0; j < lCoarseNodesPerDir[1]; ++j) {
+        for (LO i = 0; i < lCoarseNodesPerDir[0]; ++i) {
+          col = k * lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0] + j * lCoarseNodesPerDir[0] + i;
+
+          // Check for endRate
+          currentCoarseNode = 0;
+          if (firstCoarseNodeInds[0] + i * coarseRate[0] > lFineNodesPerDir[0] - 1) {
+            currentCoarseNode += firstCoarseNodeInds[0] + (i - 1) * coarseRate[0] + endRate[0];
+          } else {
+            currentCoarseNode += firstCoarseNodeInds[0] + i * coarseRate[0];
+          }
+          if (firstCoarseNodeInds[1] + j * coarseRate[1] > lFineNodesPerDir[1] - 1) {
+            currentCoarseNode += (firstCoarseNodeInds[1] + (j - 1) * coarseRate[1] + endRate[1]) * lFineNodesPerDir[0];
+          } else {
+            currentCoarseNode += (firstCoarseNodeInds[1] + j * coarseRate[1]) * lFineNodesPerDir[0];
+          }
+          if (firstCoarseNodeInds[2] + k * coarseRate[2] > lFineNodesPerDir[2] - 1) {
+            currentCoarseNode += (firstCoarseNodeInds[2] + (k - 1) * coarseRate[2] + endRate[2]) * lFineNodesPerDir[1] * lFineNodesPerDir[0];
+          } else {
+            currentCoarseNode += (firstCoarseNodeInds[2] + k * coarseRate[2]) * lFineNodesPerDir[1] * lFineNodesPerDir[0];
+          }
+          // Load coordinates
+          for (LO dim = 0; dim < numDimensions; ++dim) {
+            coarseNodes[dim][col] = fineNodes[dim][currentCoarseNode];
+          }
 
-            if((endRate[2] != coarseRate[2])
-               && (gCoarseNodesGIDs[col] > (gCoarseNodesPerDir[2] - 2)*fineNodesPerCoarseSlab
-                   + fineNodesEndCoarseSlab - 1)) {
-              tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab + 1;
-              tmpVars[0] = gCoarseNodesGIDs[col] - (tmpInds[2] - 1)*fineNodesPerCoarseSlab
-                - fineNodesEndCoarseSlab;
-            } else {
-              tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab;
-              tmpVars[0] = gCoarseNodesGIDs[col] % fineNodesPerCoarseSlab;
-            }
-            if((endRate[1] != coarseRate[1])
-               && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane
-                   + fineNodesEndCoarsePlane - 1)) {
-              tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
-              tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane
-                - fineNodesEndCoarsePlane;
-            } else {
-              tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
-              tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
-            }
-            if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
-              tmpInds[0] = gCoarseNodesPerDir[0] - 1;
-            } else {
-              tmpInds[0] = tmpVars[1] / coarseRate[0];
-            }
-            gInd[2] = col / (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
-            tmp     = col % (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
-            gInd[1] = tmp / lCoarseNodesPerDir[0];
-            gInd[0] = tmp % lCoarseNodesPerDir[0];
-            lCol = gInd[2]*(lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0])
-              + gInd[1]*lCoarseNodesPerDir[0] + gInd[0];
-            gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer
-              + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
-            coarseNodesGIDs[lCol] = gCoarseNodeOnCoarseGridGID;
-            for(LO dof = 0; dof < BlkSize; ++dof) {
-              colGIDs[BlkSize*lCol + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
-            }
+          if ((endRate[2] != coarseRate[2]) && (gCoarseNodesGIDs[col] > (gCoarseNodesPerDir[2] - 2) * fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
+            tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab + 1;
+            tmpVars[0] = gCoarseNodesGIDs[col] - (tmpInds[2] - 1) * fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
+          } else {
+            tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab;
+            tmpVars[0] = gCoarseNodesGIDs[col] % fineNodesPerCoarseSlab;
+          }
+          if ((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2) * fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
+            tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
+            tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1) * fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
+          } else {
+            tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
+            tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
+          }
+          if (tmpVars[1] == gFineNodesPerDir[0] - 1) {
+            tmpInds[0] = gCoarseNodesPerDir[0] - 1;
+          } else {
+            tmpInds[0] = tmpVars[1] / coarseRate[0];
+          }
+          gInd[2]                    = col / (lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0]);
+          tmp                        = col % (lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0]);
+          gInd[1]                    = tmp / lCoarseNodesPerDir[0];
+          gInd[0]                    = tmp % lCoarseNodesPerDir[0];
+          lCol                       = gInd[2] * (lCoarseNodesPerDir[1] * lCoarseNodesPerDir[0]) + gInd[1] * lCoarseNodesPerDir[0] + gInd[0];
+          gCoarseNodeOnCoarseGridGID = tmpInds[2] * coarseNodesPerCoarseLayer + tmpInds[1] * gCoarseNodesPerDir[0] + tmpInds[0];
+          coarseNodesGIDs[lCol]      = gCoarseNodeOnCoarseGridGID;
+          for (LO dof = 0; dof < BlkSize; ++dof) {
+            colGIDs[BlkSize * lCol + dof] = BlkSize * gCoarseNodeOnCoarseGridGID + dof;
           }
         }
       }
-      // Now loop over the ghost nodes of the partition to add them to colGIDs
-      // since they will need to be included in the column map of P
-      for(col = lNumCoarseNodes; col < lNumCoarseNodes + lNumGhostNodes; ++col) {
-        if((endRate[2] != coarseRate[2])
-           && (ghostGIDs[ghostPermut[col - lNumCoarseNodes]] > (gCoarseNodesPerDir[2] - 2)
-               *fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
-          tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab + 1;
-          tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]]
-            - (tmpInds[2] - 1)*fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
-        } else {
-          tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab;
-          tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] % fineNodesPerCoarseSlab;
-        }
-        if((endRate[1] != coarseRate[1])
-           && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane
-               + fineNodesEndCoarsePlane - 1)) {
-          tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
-          tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane
-            - fineNodesEndCoarsePlane;
-        } else {
-          tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
-          tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
-        }
-        if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
-          tmpInds[0] = gCoarseNodesPerDir[0] - 1;
-        } else {
-          tmpInds[0] = tmpVars[1] / coarseRate[0];
-        }
-        gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer
-          + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
-        for(LO dof = 0; dof < BlkSize; ++dof) {
-          colGIDs[BlkSize*col + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
-        }
+    }
+    // Now loop over the ghost nodes of the partition to add them to colGIDs
+    // since they will need to be included in the column map of P
+    for (col = lNumCoarseNodes; col < lNumCoarseNodes + lNumGhostNodes; ++col) {
+      if ((endRate[2] != coarseRate[2]) && (ghostGIDs[ghostPermut[col - lNumCoarseNodes]] > (gCoarseNodesPerDir[2] - 2) * fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
+        tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab + 1;
+        tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] - (tmpInds[2] - 1) * fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
+      } else {
+        tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab;
+        tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] % fineNodesPerCoarseSlab;
       }
-    } // End of scope for tmpInds, tmpVars and tmp
-
-  } // GetGeometricData()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
-                      const Array<LO> /* endRate */, const LO BlkSize, const Array<LO> elemInds,
-                      const Array<LO> /* lCoarseElementsPerDir */, const LO numDimensions,
-                      const Array<LO> lFineNodesPerDir, const Array<GO> /* gFineNodesPerDir */,
-                      const Array<GO> /* gIndices */, const Array<LO> /* lCoarseNodesPerDir */,
-                      const Array<bool> ghostInterface, const Array<int> elementFlags,
-                      const std::string stencilType, const std::string /* blockStrategy */,
-                      const Array<LO> elementNodesPerDir, const LO numNodesInElement,
-                      const Array<GO> /* colGIDs */,
-                      Teuchos::SerialDenseMatrix<LO,SC>& Pi, Teuchos::SerialDenseMatrix<LO,SC>& Pf,
-                      Teuchos::SerialDenseMatrix<LO,SC>& Pe, Array<LO>& dofType,
-                      Array<LO>& lDofInd) const {
-
-    // First extract data from Aghost and move it to the corresponding dense matrix
-    // This step requires to compute the number of nodes (resp dofs) in the current
-    // coarse element, then allocate storage for local dense matrices, finally loop
-    // over the dofs and extract the corresponding row in Aghost before dispactching
-    // its data to the dense matrices.
-    // At the same time we want to compute a mapping from the element indexing to
-    // group indexing. The groups are the following: corner, edge, face and interior
-    // nodes. We uses these groups to operate on the dense matrices but need to
-    // store the nodes in their original order after groupd operations are completed.
-    LO countInterior=0, countFace=0, countEdge=0, countCorner =0;
-    Array<LO> collapseDir(numNodesInElement*BlkSize);
-    for(LO ke = 0; ke < elementNodesPerDir[2]; ++ke) {
-      for(LO je = 0; je < elementNodesPerDir[1]; ++je) {
-        for(LO ie = 0; ie < elementNodesPerDir[0]; ++ie) {
-          // Check for corner node
-          if((ke == 0 || ke == elementNodesPerDir[2]-1)
-             && (je == 0 || je == elementNodesPerDir[1]-1)
-             && (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-            for(LO dof = 0; dof < BlkSize; ++dof) {
-              dofType[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = 0;
-              lDofInd[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = BlkSize*countCorner + dof;
-            }
-            ++countCorner;
+      if ((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2) * fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
+        tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
+        tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1) * fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
+      } else {
+        tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
+        tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
+      }
+      if (tmpVars[1] == gFineNodesPerDir[0] - 1) {
+        tmpInds[0] = gCoarseNodesPerDir[0] - 1;
+      } else {
+        tmpInds[0] = tmpVars[1] / coarseRate[0];
+      }
+      gCoarseNodeOnCoarseGridGID = tmpInds[2] * coarseNodesPerCoarseLayer + tmpInds[1] * gCoarseNodesPerDir[0] + tmpInds[0];
+      for (LO dof = 0; dof < BlkSize; ++dof) {
+        colGIDs[BlkSize * col + dof] = BlkSize * gCoarseNodeOnCoarseGridGID + dof;
+      }
+    }
+  }  // End of scope for tmpInds, tmpVars and tmp
+
+}  // GetGeometricData()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
+                        const Array<LO> /* endRate */, const LO BlkSize, const Array<LO> elemInds,
+                        const Array<LO> /* lCoarseElementsPerDir */, const LO numDimensions,
+                        const Array<LO> lFineNodesPerDir, const Array<GO> /* gFineNodesPerDir */,
+                        const Array<GO> /* gIndices */, const Array<LO> /* lCoarseNodesPerDir */,
+                        const Array<bool> ghostInterface, const Array<int> elementFlags,
+                        const std::string stencilType, const std::string /* blockStrategy */,
+                        const Array<LO> elementNodesPerDir, const LO numNodesInElement,
+                        const Array<GO> /* colGIDs */,
+                        Teuchos::SerialDenseMatrix<LO, SC>& Pi, Teuchos::SerialDenseMatrix<LO, SC>& Pf,
+                        Teuchos::SerialDenseMatrix<LO, SC>& Pe, Array<LO>& dofType,
+                        Array<LO>& lDofInd) const {
+  // First extract data from Aghost and move it to the corresponding dense matrix
+  // This step requires to compute the number of nodes (resp dofs) in the current
+  // coarse element, then allocate storage for local dense matrices, finally loop
+  // over the dofs and extract the corresponding row in Aghost before dispactching
+  // its data to the dense matrices.
+  // At the same time we want to compute a mapping from the element indexing to
+  // group indexing. The groups are the following: corner, edge, face and interior
+  // nodes. We uses these groups to operate on the dense matrices but need to
+  // store the nodes in their original order after groupd operations are completed.
+  LO countInterior = 0, countFace = 0, countEdge = 0, countCorner = 0;
+  Array<LO> collapseDir(numNodesInElement * BlkSize);
+  for (LO ke = 0; ke < elementNodesPerDir[2]; ++ke) {
+    for (LO je = 0; je < elementNodesPerDir[1]; ++je) {
+      for (LO ie = 0; ie < elementNodesPerDir[0]; ++ie) {
+        // Check for corner node
+        if ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (je == 0 || je == elementNodesPerDir[1] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+          for (LO dof = 0; dof < BlkSize; ++dof) {
+            dofType[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 0;
+            lDofInd[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = BlkSize * countCorner + dof;
+          }
+          ++countCorner;
 
           // Check for edge node
-          } else if (((ke == 0 || ke == elementNodesPerDir[2]-1)
-                      && (je == 0 || je == elementNodesPerDir[1]-1))
-                     || ((ke == 0 || ke == elementNodesPerDir[2]-1)
-                         && (ie == 0 || ie == elementNodesPerDir[0]-1))
-                     || ((je == 0 || je == elementNodesPerDir[1]-1)
-                         && (ie == 0 || ie == elementNodesPerDir[0]-1))) {
-            for(LO dof = 0; dof < BlkSize; ++dof) {
-              dofType[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = 1;
-              lDofInd[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = BlkSize*countEdge + dof;
-              if((ke == 0 || ke == elementNodesPerDir[2]-1)
-                 && (je == 0 || je == elementNodesPerDir[1]-1)) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 0;
-              } else if((ke == 0 || ke == elementNodesPerDir[2]-1)
-                        && (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 1;
-              } else if((je == 0 || je == elementNodesPerDir[1]-1
-                         ) && (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 2;
-              }
+        } else if (((ke == 0 || ke == elementNodesPerDir[2] - 1) && (je == 0 || je == elementNodesPerDir[1] - 1)) || ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) || ((je == 0 || je == elementNodesPerDir[1] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1))) {
+          for (LO dof = 0; dof < BlkSize; ++dof) {
+            dofType[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 1;
+            lDofInd[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = BlkSize * countEdge + dof;
+            if ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (je == 0 || je == elementNodesPerDir[1] - 1)) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 0;
+            } else if ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 1;
+            } else if ((je == 0 || je == elementNodesPerDir[1] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 2;
             }
-            ++countEdge;
+          }
+          ++countEdge;
 
           // Check for face node
-          } else if ((ke == 0 || ke == elementNodesPerDir[2]-1)
-                     || (je == 0 || je == elementNodesPerDir[1]-1)
-                     || (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-            for(LO dof = 0; dof < BlkSize; ++dof) {
-              dofType[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = 2;
-              lDofInd[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = BlkSize*countFace + dof;
-              if(ke == 0 || ke == elementNodesPerDir[2]-1) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 2;
-              } else if(je == 0 || je == elementNodesPerDir[1]-1) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 1;
-              } else if(ie == 0 || ie == elementNodesPerDir[0]-1) {
-                collapseDir[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                                     + je*elementNodesPerDir[0] + ie) + dof] = 0;
-              }
+        } else if ((ke == 0 || ke == elementNodesPerDir[2] - 1) || (je == 0 || je == elementNodesPerDir[1] - 1) || (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+          for (LO dof = 0; dof < BlkSize; ++dof) {
+            dofType[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 2;
+            lDofInd[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = BlkSize * countFace + dof;
+            if (ke == 0 || ke == elementNodesPerDir[2] - 1) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 2;
+            } else if (je == 0 || je == elementNodesPerDir[1] - 1) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 1;
+            } else if (ie == 0 || ie == elementNodesPerDir[0] - 1) {
+              collapseDir[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 0;
             }
-            ++countFace;
+          }
+          ++countFace;
 
           // Otherwise it is an interior node
-          } else {
-            for(LO dof = 0; dof < BlkSize; ++dof) {
-              dofType[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = 3;
-              lDofInd[BlkSize*(ke*elementNodesPerDir[1]*elementNodesPerDir[0]
-                               + je*elementNodesPerDir[0] + ie) + dof] = BlkSize*countInterior +dof;
-            }
-            ++countInterior;
+        } else {
+          for (LO dof = 0; dof < BlkSize; ++dof) {
+            dofType[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = 3;
+            lDofInd[BlkSize * (ke * elementNodesPerDir[1] * elementNodesPerDir[0] + je * elementNodesPerDir[0] + ie) + dof] = BlkSize * countInterior + dof;
           }
+          ++countInterior;
         }
       }
     }
+  }
 
-    LO numInteriorNodes = 0, numFaceNodes = 0, numEdgeNodes = 0, numCornerNodes = 8;
-    numInteriorNodes = (elementNodesPerDir[0] - 2)*(elementNodesPerDir[1] - 2)
-      *(elementNodesPerDir[2] - 2);
-    numFaceNodes = 2*(elementNodesPerDir[0] - 2)*(elementNodesPerDir[1] - 2)
-      + 2*(elementNodesPerDir[0] - 2)*(elementNodesPerDir[2] - 2)
-      + 2*(elementNodesPerDir[1] - 2)*(elementNodesPerDir[2] - 2);
-    numEdgeNodes = 4*(elementNodesPerDir[0] - 2) + 4*(elementNodesPerDir[1] - 2)
-      + 4*(elementNodesPerDir[2] - 2);
-    // Diagonal blocks
-    Teuchos::SerialDenseMatrix<LO,SC> Aii(BlkSize*numInteriorNodes, BlkSize*numInteriorNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Aff(BlkSize*numFaceNodes,     BlkSize*numFaceNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Aee(BlkSize*numEdgeNodes,     BlkSize*numEdgeNodes);
-    // Upper triangular blocks
-    Teuchos::SerialDenseMatrix<LO,SC> Aif(BlkSize*numInteriorNodes, BlkSize*numFaceNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Aie(BlkSize*numInteriorNodes, BlkSize*numEdgeNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Afe(BlkSize*numFaceNodes,     BlkSize*numEdgeNodes);
-    // Coarse nodes "right hand sides" and local prolongators
-    Teuchos::SerialDenseMatrix<LO,SC> Aic(BlkSize*numInteriorNodes, BlkSize*numCornerNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Afc(BlkSize*numFaceNodes,     BlkSize*numCornerNodes);
-    Teuchos::SerialDenseMatrix<LO,SC> Aec(BlkSize*numEdgeNodes,     BlkSize*numCornerNodes);
-    Pi.shape( BlkSize*numInteriorNodes, BlkSize*numCornerNodes);
-    Pf.shape( BlkSize*numFaceNodes,     BlkSize*numCornerNodes);
-    Pe.shape( BlkSize*numEdgeNodes,     BlkSize*numCornerNodes);
-
-    ArrayView<const LO> rowIndices;
-    ArrayView<const SC> rowValues;
-    LO idof, iInd, jInd;
-    int iType = 0, jType = 0;
-    int orientation = -1;
-    int collapseFlags[3] = {};
-    Array<SC> stencil((std::pow(3,numDimensions))*BlkSize);
-    // LBV Note: we could skip the extraction of rows corresponding to coarse nodes
-    //           we might want to fuse the first three loops and do integer arithmetic
-    //           to have more optimization during compilation...
-    for(LO ke = 0; ke < elementNodesPerDir[2]; ++ke) {
-      for(LO je = 0; je < elementNodesPerDir[1]; ++je) {
-        for(LO ie = 0; ie < elementNodesPerDir[0]; ++ie) {
-          collapseFlags[0] = 0; collapseFlags[1] = 0; collapseFlags[2] = 0;
-          if((elementFlags[0] == 1 || elementFlags[0] == 3) && ie == 0) {
-            collapseFlags[0] += 1;
-          }
-          if((elementFlags[0] == 2 || elementFlags[0] == 3) && ie == elementNodesPerDir[0] - 1) {
-            collapseFlags[0] += 2;
-          }
-          if((elementFlags[1] == 1 || elementFlags[1] == 3) && je == 0) {
-            collapseFlags[1] += 1;
-          }
-          if((elementFlags[1] == 2 || elementFlags[1] == 3) && je == elementNodesPerDir[1] - 1) {
-            collapseFlags[1] += 2;
-          }
-          if((elementFlags[2] == 1 || elementFlags[2] == 3) && ke == 0) {
-            collapseFlags[2] += 1;
-          }
-          if((elementFlags[2] == 2 || elementFlags[2] == 3) && ke == elementNodesPerDir[2] - 1) {
-            collapseFlags[2] += 2;
-          }
+  LO numInteriorNodes = 0, numFaceNodes = 0, numEdgeNodes = 0, numCornerNodes = 8;
+  numInteriorNodes = (elementNodesPerDir[0] - 2) * (elementNodesPerDir[1] - 2) * (elementNodesPerDir[2] - 2);
+  numFaceNodes     = 2 * (elementNodesPerDir[0] - 2) * (elementNodesPerDir[1] - 2) + 2 * (elementNodesPerDir[0] - 2) * (elementNodesPerDir[2] - 2) + 2 * (elementNodesPerDir[1] - 2) * (elementNodesPerDir[2] - 2);
+  numEdgeNodes     = 4 * (elementNodesPerDir[0] - 2) + 4 * (elementNodesPerDir[1] - 2) + 4 * (elementNodesPerDir[2] - 2);
+  // Diagonal blocks
+  Teuchos::SerialDenseMatrix<LO, SC> Aii(BlkSize * numInteriorNodes, BlkSize * numInteriorNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Aff(BlkSize * numFaceNodes, BlkSize * numFaceNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Aee(BlkSize * numEdgeNodes, BlkSize * numEdgeNodes);
+  // Upper triangular blocks
+  Teuchos::SerialDenseMatrix<LO, SC> Aif(BlkSize * numInteriorNodes, BlkSize * numFaceNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Aie(BlkSize * numInteriorNodes, BlkSize * numEdgeNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Afe(BlkSize * numFaceNodes, BlkSize * numEdgeNodes);
+  // Coarse nodes "right hand sides" and local prolongators
+  Teuchos::SerialDenseMatrix<LO, SC> Aic(BlkSize * numInteriorNodes, BlkSize * numCornerNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Afc(BlkSize * numFaceNodes, BlkSize * numCornerNodes);
+  Teuchos::SerialDenseMatrix<LO, SC> Aec(BlkSize * numEdgeNodes, BlkSize * numCornerNodes);
+  Pi.shape(BlkSize * numInteriorNodes, BlkSize * numCornerNodes);
+  Pf.shape(BlkSize * numFaceNodes, BlkSize * numCornerNodes);
+  Pe.shape(BlkSize * numEdgeNodes, BlkSize * numCornerNodes);
+
+  ArrayView<const LO> rowIndices;
+  ArrayView<const SC> rowValues;
+  LO idof, iInd, jInd;
+  int iType = 0, jType = 0;
+  int orientation      = -1;
+  int collapseFlags[3] = {};
+  Array<SC> stencil((std::pow(3, numDimensions)) * BlkSize);
+  // LBV Note: we could skip the extraction of rows corresponding to coarse nodes
+  //           we might want to fuse the first three loops and do integer arithmetic
+  //           to have more optimization during compilation...
+  for (LO ke = 0; ke < elementNodesPerDir[2]; ++ke) {
+    for (LO je = 0; je < elementNodesPerDir[1]; ++je) {
+      for (LO ie = 0; ie < elementNodesPerDir[0]; ++ie) {
+        collapseFlags[0] = 0;
+        collapseFlags[1] = 0;
+        collapseFlags[2] = 0;
+        if ((elementFlags[0] == 1 || elementFlags[0] == 3) && ie == 0) {
+          collapseFlags[0] += 1;
+        }
+        if ((elementFlags[0] == 2 || elementFlags[0] == 3) && ie == elementNodesPerDir[0] - 1) {
+          collapseFlags[0] += 2;
+        }
+        if ((elementFlags[1] == 1 || elementFlags[1] == 3) && je == 0) {
+          collapseFlags[1] += 1;
+        }
+        if ((elementFlags[1] == 2 || elementFlags[1] == 3) && je == elementNodesPerDir[1] - 1) {
+          collapseFlags[1] += 2;
+        }
+        if ((elementFlags[2] == 1 || elementFlags[2] == 3) && ke == 0) {
+          collapseFlags[2] += 1;
+        }
+        if ((elementFlags[2] == 2 || elementFlags[2] == 3) && ke == elementNodesPerDir[2] - 1) {
+          collapseFlags[2] += 2;
+        }
 
-          // Based on (ie, je, ke) we detect the type of node we are dealing with
-          GetNodeInfo(ie, je, ke, elementNodesPerDir, &iType, iInd, &orientation);
-          for(LO dof0 = 0; dof0 < BlkSize; ++dof0) {
-            // Compute the dof index for each dof at point (ie, je, ke)
-            idof = ((elemInds[2]*coarseRate[2] + ke)*lFineNodesPerDir[1]*lFineNodesPerDir[0]
-                    + (elemInds[1]*coarseRate[1] + je)*lFineNodesPerDir[0]
-                    + elemInds[0]*coarseRate[0] + ie)*BlkSize + dof0;
-            Aghost->getLocalRowView(idof, rowIndices, rowValues);
-            FormatStencil(BlkSize, ghostInterface, ie, je, ke, rowValues,
-                          elementNodesPerDir, collapseFlags, stencilType, stencil);
-            LO io, jo, ko;
-            if(iType == 3) {// interior node, no stencil collapse needed
-              for(LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
-                // Find (if, jf, kf) the indices associated with the interacting node
-                ko = ke + (interactingNode / 9 - 1);
-                {
-                  LO tmp = interactingNode % 9;
-                  jo = je + (tmp / 3 - 1);
-                  io = ie + (tmp % 3 - 1);
-                  int dummy;
-                  GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
-                }
-                if((io > -1 && io < elementNodesPerDir[0]) &&
-                   (jo > -1 && jo < elementNodesPerDir[1]) &&
-                   (ko > -1 && ko < elementNodesPerDir[2])) {
-                  for(LO dof1 = 0; dof1 < BlkSize; ++dof1) {
-                    if (jType == 3) {
-                      Aii(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 2) {
-                      Aif(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 1) {
-                      Aie(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 0) {
-                      Aic(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        -stencil[interactingNode*BlkSize + dof1];
-                    }
+        // Based on (ie, je, ke) we detect the type of node we are dealing with
+        GetNodeInfo(ie, je, ke, elementNodesPerDir, &iType, iInd, &orientation);
+        for (LO dof0 = 0; dof0 < BlkSize; ++dof0) {
+          // Compute the dof index for each dof at point (ie, je, ke)
+          idof = ((elemInds[2] * coarseRate[2] + ke) * lFineNodesPerDir[1] * lFineNodesPerDir[0] + (elemInds[1] * coarseRate[1] + je) * lFineNodesPerDir[0] + elemInds[0] * coarseRate[0] + ie) * BlkSize + dof0;
+          Aghost->getLocalRowView(idof, rowIndices, rowValues);
+          FormatStencil(BlkSize, ghostInterface, ie, je, ke, rowValues,
+                        elementNodesPerDir, collapseFlags, stencilType, stencil);
+          LO io, jo, ko;
+          if (iType == 3) {  // interior node, no stencil collapse needed
+            for (LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
+              // Find (if, jf, kf) the indices associated with the interacting node
+              ko = ke + (interactingNode / 9 - 1);
+              {
+                LO tmp = interactingNode % 9;
+                jo     = je + (tmp / 3 - 1);
+                io     = ie + (tmp % 3 - 1);
+                int dummy;
+                GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
+              }
+              if ((io > -1 && io < elementNodesPerDir[0]) &&
+                  (jo > -1 && jo < elementNodesPerDir[1]) &&
+                  (ko > -1 && ko < elementNodesPerDir[2])) {
+                for (LO dof1 = 0; dof1 < BlkSize; ++dof1) {
+                  if (jType == 3) {
+                    Aii(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 2) {
+                    Aif(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 1) {
+                    Aie(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 0) {
+                    Aic(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        -stencil[interactingNode * BlkSize + dof1];
                   }
                 }
               }
-            } else if(iType == 2) {// Face node, collapse stencil along face normal (*orientation)
-              CollapseStencil(2, orientation, collapseFlags, stencil);
-              for(LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
-                // Find (if, jf, kf) the indices associated with the interacting node
-                ko = ke + (interactingNode / 9 - 1);
-                {
-                  LO tmp = interactingNode % 9;
-                  jo = je + (tmp / 3 - 1);
-                  io = ie + (tmp % 3 - 1);
-                  int dummy;
-                  GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
-                }
-                if((io > -1 && io < elementNodesPerDir[0]) &&
-                   (jo > -1 && jo < elementNodesPerDir[1]) &&
-                   (ko > -1 && ko < elementNodesPerDir[2])) {
-                  for(LO dof1 = 0; dof1 < BlkSize; ++dof1) {
-                    if(jType == 2) {
-                      Aff(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 1) {
-                      Afe(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 0) {
-                      Afc(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        -stencil[interactingNode*BlkSize + dof1];
-                    }
+            }
+          } else if (iType == 2) {  // Face node, collapse stencil along face normal (*orientation)
+            CollapseStencil(2, orientation, collapseFlags, stencil);
+            for (LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
+              // Find (if, jf, kf) the indices associated with the interacting node
+              ko = ke + (interactingNode / 9 - 1);
+              {
+                LO tmp = interactingNode % 9;
+                jo     = je + (tmp / 3 - 1);
+                io     = ie + (tmp % 3 - 1);
+                int dummy;
+                GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
+              }
+              if ((io > -1 && io < elementNodesPerDir[0]) &&
+                  (jo > -1 && jo < elementNodesPerDir[1]) &&
+                  (ko > -1 && ko < elementNodesPerDir[2])) {
+                for (LO dof1 = 0; dof1 < BlkSize; ++dof1) {
+                  if (jType == 2) {
+                    Aff(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 1) {
+                    Afe(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 0) {
+                    Afc(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        -stencil[interactingNode * BlkSize + dof1];
                   }
                 }
               }
-            } else if(iType == 1) {// Edge node, collapse stencil perpendicular to edge direction
-              CollapseStencil(1, orientation, collapseFlags, stencil);
-              for(LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
-                // Find (if, jf, kf) the indices associated with the interacting node
-                ko = ke + (interactingNode / 9 - 1);
-                {
-                  LO tmp = interactingNode % 9;
-                  jo = je + (tmp / 3 - 1);
-                  io = ie + (tmp % 3 - 1);
-                  int dummy;
-                  GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
-                }
-                if((io > -1 && io < elementNodesPerDir[0]) &&
-                   (jo > -1 && jo < elementNodesPerDir[1]) &&
-                   (ko > -1 && ko < elementNodesPerDir[2])) {
-                  for(LO dof1 = 0; dof1 < BlkSize; ++dof1) {
-                    if(jType == 1) {
-                      Aee(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        stencil[interactingNode*BlkSize + dof1];
-                    } else if(jType == 0) {
-                      Aec(iInd*BlkSize + dof0, jInd*BlkSize + dof1) =
-                        -stencil[interactingNode*BlkSize + dof1];
-                    }
-                  } // Check if interacting node is in element or not
-                } // Pc is the identity so no need to treat iType == 0
-              } // Loop over interacting nodes within a row
-            } // Check for current node type
-          } // Loop over degrees of freedom associated to a given node
-        } // Loop over i
-      } // Loop over j
-    } // Loop over k
-
-    // Compute the projection operators for edge and interior nodes
-    //
-    //         [P_i] = [A_{ii} & A_{if} & A_{ie}]^{-1} [A_{ic}]
-    //         [P_f] = [  0    & A_{ff} & A_{fe}]      [A_{fc}]
-    //         [P_e] = [  0    &   0    & A_{ee}]      [A_{ec}]
-    //         [P_c] =  I_c
-    //
-    { // Compute P_e
-      // We need to solve for P_e in: A_{ee}*P_e = A_{fc}
-      // So we simple do P_e = A_{ee}^{-1)*A_{ec}
-      Teuchos::SerialDenseSolver<LO,SC> problem;
-      problem.setMatrix(Teuchos::rcp(&Aee, false));
-      problem.setVectors(Teuchos::rcp(&Pe, false), Teuchos::rcp(&Aec, false));
-      problem.factorWithEquilibration(true);
-      problem.solve();
-      problem.unequilibrateLHS();
-    }
-
-    { // Compute P_f
-      // We need to solve for P_f in: A_{ff}*P_f + A_{fe}*P_e = A_{fc}
-      // Step one: A_{fc} = 1.0*A_{fc} + (-1.0)*A_{fe}*P_e
-      Afc.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Afe, Pe, 1.0);
-      // Step two: P_f = A_{ff}^{-1}*A_{fc}
-      Teuchos::SerialDenseSolver<LO,SC> problem;
-      problem.setMatrix(Teuchos::rcp(&Aff, false));
-      problem.setVectors(Teuchos::rcp(&Pf, false), Teuchos::rcp(&Afc, false));
-      problem.factorWithEquilibration(true);
-      problem.solve();
-      problem.unequilibrateLHS();
-    }
-
-    { // Compute P_i
-      // We need to solve for P_i in: A_{ii}*P_i + A_{if}*P_f + A_{ie}*P_e = A_{ic}
-      // Step one: A_{ic} = 1.0*A_{ic} + (-1.0)*A_{ie}*P_e
-      Aic.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Aie, Pe, 1.0);
-      // Step one: A_{ic} = 1.0*A_{ic} + (-1.0)*A_{if}*P_f
-      Aic.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Aif, Pf, 1.0);
-      // Step two: P_i = A_{ii}^{-1}*A_{ic}
-      Teuchos::SerialDenseSolver<LO,SC> problem;
-      problem.setMatrix(Teuchos::rcp(&Aii, false));
-      problem.setVectors(Teuchos::rcp(&Pi, false), Teuchos::rcp(&Aic, false));
-      problem.factorWithEquilibration(true);
-      problem.solve();
-      problem.unequilibrateLHS();
-    }
+            }
+          } else if (iType == 1) {  // Edge node, collapse stencil perpendicular to edge direction
+            CollapseStencil(1, orientation, collapseFlags, stencil);
+            for (LO interactingNode = 0; interactingNode < 27; ++interactingNode) {
+              // Find (if, jf, kf) the indices associated with the interacting node
+              ko = ke + (interactingNode / 9 - 1);
+              {
+                LO tmp = interactingNode % 9;
+                jo     = je + (tmp / 3 - 1);
+                io     = ie + (tmp % 3 - 1);
+                int dummy;
+                GetNodeInfo(io, jo, ko, elementNodesPerDir, &jType, jInd, &dummy);
+              }
+              if ((io > -1 && io < elementNodesPerDir[0]) &&
+                  (jo > -1 && jo < elementNodesPerDir[1]) &&
+                  (ko > -1 && ko < elementNodesPerDir[2])) {
+                for (LO dof1 = 0; dof1 < BlkSize; ++dof1) {
+                  if (jType == 1) {
+                    Aee(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        stencil[interactingNode * BlkSize + dof1];
+                  } else if (jType == 0) {
+                    Aec(iInd * BlkSize + dof0, jInd * BlkSize + dof1) =
+                        -stencil[interactingNode * BlkSize + dof1];
+                  }
+                }  // Check if interacting node is in element or not
+              }    // Pc is the identity so no need to treat iType == 0
+            }      // Loop over interacting nodes within a row
+          }        // Check for current node type
+        }          // Loop over degrees of freedom associated to a given node
+      }            // Loop over i
+    }              // Loop over j
+  }                // Loop over k
+
+  // Compute the projection operators for edge and interior nodes
+  //
+  //         [P_i] = [A_{ii} & A_{if} & A_{ie}]^{-1} [A_{ic}]
+  //         [P_f] = [  0    & A_{ff} & A_{fe}]      [A_{fc}]
+  //         [P_e] = [  0    &   0    & A_{ee}]      [A_{ec}]
+  //         [P_c] =  I_c
+  //
+  {  // Compute P_e
+    // We need to solve for P_e in: A_{ee}*P_e = A_{fc}
+    // So we simple do P_e = A_{ee}^{-1)*A_{ec}
+    Teuchos::SerialDenseSolver<LO, SC> problem;
+    problem.setMatrix(Teuchos::rcp(&Aee, false));
+    problem.setVectors(Teuchos::rcp(&Pe, false), Teuchos::rcp(&Aec, false));
+    problem.factorWithEquilibration(true);
+    problem.solve();
+    problem.unequilibrateLHS();
+  }
 
-  } // ComputeLocalEntries()
+  {  // Compute P_f
+    // We need to solve for P_f in: A_{ff}*P_f + A_{fe}*P_e = A_{fc}
+    // Step one: A_{fc} = 1.0*A_{fc} + (-1.0)*A_{fe}*P_e
+    Afc.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Afe, Pe, 1.0);
+    // Step two: P_f = A_{ff}^{-1}*A_{fc}
+    Teuchos::SerialDenseSolver<LO, SC> problem;
+    problem.setMatrix(Teuchos::rcp(&Aff, false));
+    problem.setVectors(Teuchos::rcp(&Pf, false), Teuchos::rcp(&Afc, false));
+    problem.factorWithEquilibration(true);
+    problem.solve();
+    problem.unequilibrateLHS();
+  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CollapseStencil(const int type, const int orientation, const int /* collapseFlags */[3],
-                  Array<SC>& stencil) const {
+  {  // Compute P_i
+    // We need to solve for P_i in: A_{ii}*P_i + A_{if}*P_f + A_{ie}*P_e = A_{ic}
+    // Step one: A_{ic} = 1.0*A_{ic} + (-1.0)*A_{ie}*P_e
+    Aic.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Aie, Pe, 1.0);
+    // Step one: A_{ic} = 1.0*A_{ic} + (-1.0)*A_{if}*P_f
+    Aic.multiply(Teuchos::NO_TRANS, Teuchos::NO_TRANS, -1.0, Aif, Pf, 1.0);
+    // Step two: P_i = A_{ii}^{-1}*A_{ic}
+    Teuchos::SerialDenseSolver<LO, SC> problem;
+    problem.setMatrix(Teuchos::rcp(&Aii, false));
+    problem.setVectors(Teuchos::rcp(&Pi, false), Teuchos::rcp(&Aic, false));
+    problem.factorWithEquilibration(true);
+    problem.solve();
+    problem.unequilibrateLHS();
+  }
 
-    if(type == 2) {// Face stencil collapse
-      // Let's hope things vectorize well inside this if statement
-      if(orientation == 0) {
-        for(LO i = 0; i < 9; ++i) {
-          stencil[3*i + 1] = stencil[3*i] + stencil[3*i + 1] + stencil[3*i + 2];
-          stencil[3*i]     = 0;
-          stencil[3*i + 2] = 0;
+}  // ComputeLocalEntries()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CollapseStencil(const int type, const int orientation, const int /* collapseFlags */[3],
+                    Array<SC>& stencil) const {
+  if (type == 2) {  // Face stencil collapse
+    // Let's hope things vectorize well inside this if statement
+    if (orientation == 0) {
+      for (LO i = 0; i < 9; ++i) {
+        stencil[3 * i + 1] = stencil[3 * i] + stencil[3 * i + 1] + stencil[3 * i + 2];
+        stencil[3 * i]     = 0;
+        stencil[3 * i + 2] = 0;
+      }
+    } else if (orientation == 1) {
+      for (LO i = 0; i < 3; ++i) {
+        stencil[9 * i + 3] = stencil[9 * i + 0] + stencil[9 * i + 3] + stencil[9 * i + 6];
+        stencil[9 * i + 0] = 0;
+        stencil[9 * i + 6] = 0;
+        stencil[9 * i + 4] = stencil[9 * i + 1] + stencil[9 * i + 4] + stencil[9 * i + 7];
+        stencil[9 * i + 1] = 0;
+        stencil[9 * i + 7] = 0;
+        stencil[9 * i + 5] = stencil[9 * i + 2] + stencil[9 * i + 5] + stencil[9 * i + 8];
+        stencil[9 * i + 2] = 0;
+        stencil[9 * i + 8] = 0;
+      }
+    } else if (orientation == 2) {
+      for (LO i = 0; i < 9; ++i) {
+        stencil[i + 9]  = stencil[i] + stencil[i + 9] + stencil[i + 18];
+        stencil[i]      = 0;
+        stencil[i + 18] = 0;
+      }
+    }
+  } else if (type == 1) {  // Edge stencil collapse
+    SC tmp1 = 0, tmp2 = 0, tmp3 = 0;
+    if (orientation == 0) {
+      for (LO i = 0; i < 9; ++i) {
+        tmp1 += stencil[0 + 3 * i];
+        stencil[0 + 3 * i] = 0;
+        tmp2 += stencil[1 + 3 * i];
+        stencil[1 + 3 * i] = 0;
+        tmp3 += stencil[2 + 3 * i];
+        stencil[2 + 3 * i] = 0;
+      }
+      stencil[12] = tmp1;
+      stencil[13] = tmp2;
+      stencil[14] = tmp3;
+    } else if (orientation == 1) {
+      for (LO i = 0; i < 3; ++i) {
+        tmp1 += stencil[0 + i] + stencil[9 + i] + stencil[18 + i];
+        stencil[0 + i]  = 0;
+        stencil[9 + i]  = 0;
+        stencil[18 + i] = 0;
+        tmp2 += stencil[3 + i] + stencil[12 + i] + stencil[21 + i];
+        stencil[3 + i]  = 0;
+        stencil[12 + i] = 0;
+        stencil[21 + i] = 0;
+        tmp3 += stencil[6 + i] + stencil[15 + i] + stencil[24 + i];
+        stencil[6 + i]  = 0;
+        stencil[15 + i] = 0;
+        stencil[24 + i] = 0;
+      }
+      stencil[10] = tmp1;
+      stencil[13] = tmp2;
+      stencil[16] = tmp3;
+    } else if (orientation == 2) {
+      for (LO i = 0; i < 9; ++i) {
+        tmp1 += stencil[i];
+        stencil[i] = 0;
+        tmp2 += stencil[i + 9];
+        stencil[i + 9] = 0;
+        tmp3 += stencil[i + 18];
+        stencil[i + 18] = 0;
+      }
+      stencil[4]  = tmp1;
+      stencil[13] = tmp2;
+      stencil[22] = tmp3;
+    }
+  }
+}  // CollapseStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FormatStencil(const LO BlkSize, const Array<bool> /* ghostInterface */, const LO /* ie */, const LO /* je */,
+                  const LO /* ke */, const ArrayView<const SC> rowValues, const Array<LO> /* elementNodesPerDir */,
+                  const int collapseFlags[3], const std::string stencilType, Array<SC>& stencil)
+        const {
+  if (stencilType == "reduced") {
+    Array<int> fullStencilInds(7);
+    fullStencilInds[0] = 4;
+    fullStencilInds[1] = 10;
+    fullStencilInds[2] = 12;
+    fullStencilInds[3] = 13;
+    fullStencilInds[4] = 14;
+    fullStencilInds[5] = 16;
+    fullStencilInds[6] = 22;
+
+    // Create a mask array to automate mesh boundary processing
+    Array<int> mask(7);
+    int stencilSize = rowValues.size();
+    if (collapseFlags[0] + collapseFlags[1] + collapseFlags[2] > 0) {
+      if (stencilSize == 1) {
+        // The assumption is made that if only one non-zero exists in the row
+        // then this represent a Dirichlet BC being enforced.
+        // One might want to zero out the corresponding row in the prolongator.
+        mask[0] = 1;
+        mask[1] = 1;
+        mask[2] = 1;
+        mask[4] = 1;
+        mask[5] = 1;
+        mask[6] = 1;
+      } else {
+        // Here we are looking at Neumann like BC where only a flux is prescribed
+        // and less non-zeros will be present in the corresponding row.
+        if (collapseFlags[2] == 1 || collapseFlags[2] == 3) {
+          mask[0] = 1;
         }
-      } else if(orientation == 1) {
-        for(LO i = 0; i < 3; ++i) {
-          stencil[9*i + 3] = stencil[9*i + 0] + stencil[9*i + 3] + stencil[9*i + 6];
-          stencil[9*i + 0] = 0;
-          stencil[9*i + 6] = 0;
-          stencil[9*i + 4] = stencil[9*i + 1] + stencil[9*i + 4] + stencil[9*i + 7];
-          stencil[9*i + 1] = 0;
-          stencil[9*i + 7] = 0;
-          stencil[9*i + 5] = stencil[9*i + 2] + stencil[9*i + 5] + stencil[9*i + 8];
-          stencil[9*i + 2] = 0;
-          stencil[9*i + 8] = 0;
+        if (collapseFlags[2] == 2 || collapseFlags[2] == 3) {
+          mask[6] = 1;
         }
-      } else if(orientation == 2) {
-        for(LO i = 0; i < 9; ++i) {
-          stencil[i + 9]  = stencil[i] + stencil[i + 9] + stencil[i + 18];
-          stencil[i]      = 0;
-          stencil[i + 18] = 0;
+        if (collapseFlags[1] == 1 || collapseFlags[1] == 3) {
+          mask[1] = 1;
         }
-      }
-    } else if(type == 1) {// Edge stencil collapse
-      SC tmp1 = 0, tmp2 = 0, tmp3 = 0;
-      if(orientation == 0) {
-        for(LO i = 0; i < 9; ++i) {
-          tmp1 += stencil[0 + 3*i];
-          stencil[0 + 3*i] = 0;
-          tmp2 += stencil[1 + 3*i];
-          stencil[1 + 3*i] = 0;
-          tmp3 += stencil[2 + 3*i];
-          stencil[2 + 3*i] = 0;
+        if (collapseFlags[1] == 2 || collapseFlags[1] == 3) {
+          mask[5] = 1;
         }
-        stencil[12] = tmp1;
-        stencil[13] = tmp2;
-        stencil[14] = tmp3;
-      } else if(orientation == 1) {
-        for(LO i = 0; i < 3; ++i) {
-          tmp1 += stencil[0 + i] + stencil[9 + i] + stencil[18 + i];
-          stencil[ 0 + i] = 0;
-          stencil[ 9 + i] = 0;
-          stencil[18 + i] = 0;
-          tmp2 += stencil[3 + i] + stencil[12 + i] + stencil[21 + i];
-          stencil[ 3 + i] = 0;
-          stencil[12 + i] = 0;
-          stencil[21 + i] = 0;
-          tmp3 += stencil[6 + i] + stencil[15 + i] + stencil[24 + i];
-          stencil[ 6 + i] = 0;
-          stencil[15 + i] = 0;
-          stencil[24 + i] = 0;
+        if (collapseFlags[0] == 1 || collapseFlags[0] == 3) {
+          mask[2] = 1;
         }
-        stencil[10] = tmp1;
-        stencil[13] = tmp2;
-        stencil[16] = tmp3;
-      } else if(orientation == 2) {
-        for(LO i = 0; i < 9; ++i) {
-          tmp1 += stencil[i];
-          stencil[i] = 0;
-          tmp2 += stencil[i +  9];
-          stencil[i +  9] = 0;
-          tmp3 += stencil[i + 18];
-          stencil[i + 18] = 0;
+        if (collapseFlags[0] == 2 || collapseFlags[0] == 3) {
+          mask[4] = 1;
         }
-        stencil[ 4] = tmp1;
-        stencil[13] = tmp2;
-        stencil[22] = tmp3;
       }
     }
-  } // CollapseStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FormatStencil(const LO BlkSize, const Array<bool> /* ghostInterface */, const LO /* ie */, const LO /* je */,
-                const LO /* ke */, const ArrayView<const SC> rowValues,const Array<LO> /* elementNodesPerDir */,
-                const int collapseFlags[3], const std::string stencilType, Array<SC>& stencil)
-    const {
 
-    if(stencilType == "reduced") {
-      Array<int> fullStencilInds(7);
-      fullStencilInds[0] =  4; fullStencilInds[1] = 10; fullStencilInds[2] = 12;
-      fullStencilInds[3] = 13; fullStencilInds[4] = 14; fullStencilInds[5] = 16;
-      fullStencilInds[6] = 22;
-
-      // Create a mask array to automate mesh boundary processing
-      Array<int> mask(7);
-      int stencilSize = rowValues.size();
-      if(collapseFlags[0] + collapseFlags[1] + collapseFlags[2] > 0) {
-        if(stencilSize == 1) {
-          // The assumption is made that if only one non-zero exists in the row
-          // then this represent a Dirichlet BC being enforced.
-          // One might want to zero out the corresponding row in the prolongator.
-          mask[0] = 1; mask[1] = 1; mask[2] = 1;
-          mask[4] = 1; mask[5] = 1; mask[6] = 1;
-        } else {
-          // Here we are looking at Neumann like BC where only a flux is prescribed
-          // and less non-zeros will be present in the corresponding row.
-          if(collapseFlags[2] == 1 || collapseFlags[2] == 3) {
-            mask[0] = 1;
-          }
-          if(collapseFlags[2] == 2 || collapseFlags[2] == 3) {
-            mask[6] = 1;
-          }
-          if(collapseFlags[1] == 1 || collapseFlags[1] == 3) {
-            mask[1] = 1;
-          }
-          if(collapseFlags[1] == 2 || collapseFlags[1] == 3) {
-            mask[5] = 1;
-          }
-          if(collapseFlags[0] == 1 || collapseFlags[0] == 3) {
-            mask[2] = 1;
-          }
-          if(collapseFlags[0] == 2 || collapseFlags[0] == 3) {
-            mask[4] = 1;
-          }
+    int offset = 0;
+    for (int ind = 0; ind < 7; ++ind) {
+      if (mask[ind] == 1) {
+        for (int dof = 0; dof < BlkSize; ++dof) {
+          stencil[BlkSize * fullStencilInds[ind] + dof] = 0.0;
         }
-      }
-
-      int offset = 0;
-      for(int ind = 0; ind < 7; ++ind) {
-        if(mask[ind] == 1) {
-          for(int dof = 0; dof < BlkSize; ++dof) {
-            stencil[BlkSize*fullStencilInds[ind] + dof] = 0.0;
-          }
-          ++offset; // The offset is used to stay within rowValues bounds
-        } else {
-          for(int dof = 0; dof < BlkSize; ++dof) {
-            stencil[BlkSize*fullStencilInds[ind] + dof] = rowValues[BlkSize*(ind - offset) + dof];
-          }
+        ++offset;  // The offset is used to stay within rowValues bounds
+      } else {
+        for (int dof = 0; dof < BlkSize; ++dof) {
+          stencil[BlkSize * fullStencilInds[ind] + dof] = rowValues[BlkSize * (ind - offset) + dof];
         }
       }
-    } else if(stencilType == "full") {
-      // Create a mask array to automate mesh boundary processing
-      Array<int> mask(27);
-      if(collapseFlags[2] == 1 || collapseFlags[2] == 3) {
-        for(int ind = 0; ind < 9; ++ind) {
-          mask[ind] = 1;
-        }
+    }
+  } else if (stencilType == "full") {
+    // Create a mask array to automate mesh boundary processing
+    Array<int> mask(27);
+    if (collapseFlags[2] == 1 || collapseFlags[2] == 3) {
+      for (int ind = 0; ind < 9; ++ind) {
+        mask[ind] = 1;
       }
-      if(collapseFlags[2] == 2 || collapseFlags[2] == 3) {
-        for(int ind = 0; ind < 9; ++ind) {
-          mask[18 + ind] = 1;
-        }
+    }
+    if (collapseFlags[2] == 2 || collapseFlags[2] == 3) {
+      for (int ind = 0; ind < 9; ++ind) {
+        mask[18 + ind] = 1;
       }
-      if(collapseFlags[1] == 1 || collapseFlags[1] == 3) {
-        for(int ind1 = 0; ind1 < 3; ++ind1) {
-          for(int ind2 = 0; ind2 < 3; ++ind2) {
-            mask[ind1*9 + ind2] = 1;
-          }
+    }
+    if (collapseFlags[1] == 1 || collapseFlags[1] == 3) {
+      for (int ind1 = 0; ind1 < 3; ++ind1) {
+        for (int ind2 = 0; ind2 < 3; ++ind2) {
+          mask[ind1 * 9 + ind2] = 1;
         }
       }
-      if(collapseFlags[1] == 2 || collapseFlags[1] == 3) {
-        for(int ind1 = 0; ind1 < 3; ++ind1) {
-          for(int ind2 = 0; ind2 < 3; ++ind2) {
-            mask[ind1*9 + ind2 + 6] = 1;
-          }
+    }
+    if (collapseFlags[1] == 2 || collapseFlags[1] == 3) {
+      for (int ind1 = 0; ind1 < 3; ++ind1) {
+        for (int ind2 = 0; ind2 < 3; ++ind2) {
+          mask[ind1 * 9 + ind2 + 6] = 1;
         }
       }
-      if(collapseFlags[0] == 1 || collapseFlags[0] == 3) {
-        for(int ind = 0; ind < 9; ++ind) {
-          mask[3*ind] = 1;
-        }
+    }
+    if (collapseFlags[0] == 1 || collapseFlags[0] == 3) {
+      for (int ind = 0; ind < 9; ++ind) {
+        mask[3 * ind] = 1;
       }
-      if(collapseFlags[0] == 2 || collapseFlags[0] == 3) {
-        for(int ind = 0; ind < 9; ++ind) {
-          mask[3*ind + 2] = 1;
-        }
+    }
+    if (collapseFlags[0] == 2 || collapseFlags[0] == 3) {
+      for (int ind = 0; ind < 9; ++ind) {
+        mask[3 * ind + 2] = 1;
       }
+    }
 
-      // Now the stencil is extracted and formated
-      int offset = 0;
-      for(LO ind = 0; ind < 27; ++ind) {
-        if(mask[ind] == 0) {
-          for(int dof = 0; dof < BlkSize; ++dof) {
-            stencil[BlkSize*ind + dof] = 0.0;
-          }
-          ++offset; // The offset is used to stay within rowValues bounds
-        } else {
-          for(int dof = 0; dof < BlkSize; ++dof) {
-            stencil[BlkSize*ind + dof] = rowValues[BlkSize*(ind - offset) + dof];
-          }
+    // Now the stencil is extracted and formated
+    int offset = 0;
+    for (LO ind = 0; ind < 27; ++ind) {
+      if (mask[ind] == 0) {
+        for (int dof = 0; dof < BlkSize; ++dof) {
+          stencil[BlkSize * ind + dof] = 0.0;
         }
-      }
-    } // stencilTpye
-  } // FormatStencil()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNodeInfo(
-                        const LO ie, const LO je, const LO ke,
-                        const Array<LO> elementNodesPerDir,
-                        int* type, LO& ind, int* orientation) const {
-
-    // Initialize flags with -1 to be able to catch issues easily
-    *type = -1, ind = 0, *orientation = -1;
-    if((ke == 0 || ke == elementNodesPerDir[2]-1)
-             && (je == 0 || je == elementNodesPerDir[1]-1)
-             && (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-      // Corner node
-      *type = 0;
-      ind  = 4*ke / (elementNodesPerDir[2]-1) + 2*je / (elementNodesPerDir[1]-1)
-        + ie / (elementNodesPerDir[0]-1);
-    } else if(((ke == 0 || ke == elementNodesPerDir[2]-1)
-               && (je == 0 || je == elementNodesPerDir[1]-1))
-              || ((ke == 0 || ke == elementNodesPerDir[2]-1)
-                  && (ie == 0 || ie == elementNodesPerDir[0]-1))
-              || ((je == 0 || je == elementNodesPerDir[1]-1)
-                  && (ie == 0 || ie == elementNodesPerDir[0]-1))) {
-      // Edge node
-      *type = 1;
-      if(ke > 0) {ind += 2*(elementNodesPerDir[0] - 2 + elementNodesPerDir[1] - 2);}
-      if(ke == elementNodesPerDir[2] - 1) {ind += 4*(elementNodesPerDir[2] - 2);}
-      if((ke == 0) || (ke == elementNodesPerDir[2] - 1)) { // je or ie edges
-        if(je == 0) {// jlo ie edge
-          *orientation = 0;
-          ind += ie - 1;
-        } else if(je == elementNodesPerDir[1] - 1) {// jhi ie edge
-          *orientation = 0;
-          ind += 2*(elementNodesPerDir[1] - 2) + elementNodesPerDir[0] - 2 + ie - 1;
-        } else {// ilo or ihi je edge
-          *orientation = 1;
-          ind += elementNodesPerDir[0] - 2 + 2*(je - 1) + ie / (elementNodesPerDir[0] - 1);
+        ++offset;  // The offset is used to stay within rowValues bounds
+      } else {
+        for (int dof = 0; dof < BlkSize; ++dof) {
+          stencil[BlkSize * ind + dof] = rowValues[BlkSize * (ind - offset) + dof];
         }
-      } else {// ke edges
-        *orientation = 2;
-        ind += 4*(ke - 1) + 2*(je/(elementNodesPerDir[1] - 1)) + ie / (elementNodesPerDir[0] - 1);
       }
-    } else if ((ke == 0 || ke == elementNodesPerDir[2]-1)
-                     || (je == 0 || je == elementNodesPerDir[1]-1)
-                     || (ie == 0 || ie == elementNodesPerDir[0]-1)) {
-      // Face node
-      *type = 2;
-      if(ke == 0) {// current node is on "bottom" face
+    }
+  }  // stencilTpye
+}  // FormatStencil()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetNodeInfo(
+    const LO ie, const LO je, const LO ke,
+    const Array<LO> elementNodesPerDir,
+    int* type, LO& ind, int* orientation) const {
+  // Initialize flags with -1 to be able to catch issues easily
+  *type = -1, ind = 0, *orientation = -1;
+  if ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (je == 0 || je == elementNodesPerDir[1] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+    // Corner node
+    *type = 0;
+    ind   = 4 * ke / (elementNodesPerDir[2] - 1) + 2 * je / (elementNodesPerDir[1] - 1) + ie / (elementNodesPerDir[0] - 1);
+  } else if (((ke == 0 || ke == elementNodesPerDir[2] - 1) && (je == 0 || je == elementNodesPerDir[1] - 1)) || ((ke == 0 || ke == elementNodesPerDir[2] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1)) || ((je == 0 || je == elementNodesPerDir[1] - 1) && (ie == 0 || ie == elementNodesPerDir[0] - 1))) {
+    // Edge node
+    *type = 1;
+    if (ke > 0) {
+      ind += 2 * (elementNodesPerDir[0] - 2 + elementNodesPerDir[1] - 2);
+    }
+    if (ke == elementNodesPerDir[2] - 1) {
+      ind += 4 * (elementNodesPerDir[2] - 2);
+    }
+    if ((ke == 0) || (ke == elementNodesPerDir[2] - 1)) {  // je or ie edges
+      if (je == 0) {                                       // jlo ie edge
+        *orientation = 0;
+        ind += ie - 1;
+      } else if (je == elementNodesPerDir[1] - 1) {  // jhi ie edge
+        *orientation = 0;
+        ind += 2 * (elementNodesPerDir[1] - 2) + elementNodesPerDir[0] - 2 + ie - 1;
+      } else {  // ilo or ihi je edge
+        *orientation = 1;
+        ind += elementNodesPerDir[0] - 2 + 2 * (je - 1) + ie / (elementNodesPerDir[0] - 1);
+      }
+    } else {  // ke edges
+      *orientation = 2;
+      ind += 4 * (ke - 1) + 2 * (je / (elementNodesPerDir[1] - 1)) + ie / (elementNodesPerDir[0] - 1);
+    }
+  } else if ((ke == 0 || ke == elementNodesPerDir[2] - 1) || (je == 0 || je == elementNodesPerDir[1] - 1) || (ie == 0 || ie == elementNodesPerDir[0] - 1)) {
+    // Face node
+    *type = 2;
+    if (ke == 0) {  // current node is on "bottom" face
+      *orientation = 2;
+      ind          = (je - 1) * (elementNodesPerDir[0] - 2) + ie - 1;
+    } else {
+      // add nodes from "bottom" face
+      ind += (elementNodesPerDir[1] - 2) * (elementNodesPerDir[0] - 2);
+      // add nodes from side faces
+      ind += 2 * (ke - 1) * (elementNodesPerDir[1] - 2 + elementNodesPerDir[0] - 2);
+      if (ke == elementNodesPerDir[2] - 1) {  // current node is on "top" face
         *orientation = 2;
-        ind = (je - 1)*(elementNodesPerDir[0] - 2) + ie - 1;
-      } else {
-        // add nodes from "bottom" face
-        ind += (elementNodesPerDir[1] - 2)*(elementNodesPerDir[0] - 2);
-        // add nodes from side faces
-        ind += 2*(ke - 1)*(elementNodesPerDir[1] - 2 + elementNodesPerDir[0] - 2);
-        if(ke == elementNodesPerDir[2]-1) {// current node is on "top" face
-          *orientation = 2;
-          ind += (je - 1)*(elementNodesPerDir[0] - 2) + ie - 1;
-        } else {// current node is on a side face
-          if(je == 0) {
-            *orientation = 1;
-            ind += ie - 1;
-          } else if(je == elementNodesPerDir[1] - 1) {
-            *orientation = 1;
-            ind += 2*(elementNodesPerDir[1] - 2) + elementNodesPerDir[0] - 2 + ie - 1;
-          } else {
-            *orientation = 0;
-            ind += elementNodesPerDir[0] - 2 + 2*(je - 1) + ie / (elementNodesPerDir[0]-1);
-          }
+        ind += (je - 1) * (elementNodesPerDir[0] - 2) + ie - 1;
+      } else {  // current node is on a side face
+        if (je == 0) {
+          *orientation = 1;
+          ind += ie - 1;
+        } else if (je == elementNodesPerDir[1] - 1) {
+          *orientation = 1;
+          ind += 2 * (elementNodesPerDir[1] - 2) + elementNodesPerDir[0] - 2 + ie - 1;
+        } else {
+          *orientation = 0;
+          ind += elementNodesPerDir[0] - 2 + 2 * (je - 1) + ie / (elementNodesPerDir[0] - 1);
         }
       }
-    } else {
-      // Interior node
-      *type = 3;
-      ind  = (ke - 1)*(elementNodesPerDir[1] - 2)*(elementNodesPerDir[0] - 2)
-        + (je - 1)*(elementNodesPerDir[0] - 2) + ie - 1;
     }
-  } // GetNodeInfo()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort_permute(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const
-  {
-    typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
-    DT n = last1 - first1;
-    DT m = n / 2;
-    DT z = Teuchos::OrdinalTraits<DT>::zero();
-    while (m > z)
-      {
-        DT max = n - m;
-        for (DT j = 0; j < max; j++)
-          {
-            for (DT k = j; k >= 0; k-=m)
-              {
-                if (first1[first2[k+m]] >= first1[first2[k]])
-                  break;
-                std::swap(first2[k+m], first2[k]);
-              }
-          }
-        m = m/2;
+  } else {
+    // Interior node
+    *type = 3;
+    ind   = (ke - 1) * (elementNodesPerDir[1] - 2) * (elementNodesPerDir[0] - 2) + (je - 1) * (elementNodesPerDir[0] - 2) + ie - 1;
+  }
+}  // GetNodeInfo()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort_permute(
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const {
+  typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
+  DT n = last1 - first1;
+  DT m = n / 2;
+  DT z = Teuchos::OrdinalTraits<DT>::zero();
+  while (m > z) {
+    DT max = n - m;
+    for (DT j = 0; j < max; j++) {
+      for (DT k = j; k >= 0; k -= m) {
+        if (first1[first2[k + m]] >= first1[first2[k]])
+          break;
+        std::swap(first2[k + m], first2[k]);
       }
+    }
+    m = m / 2;
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLACKBOXPFACTORY_SHORT
-#endif // MUELU_BLACKBOXPFACTORY_DEF_HPP
+#endif  // MUELU_BLACKBOXPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_decl.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_decl.hpp
index 2f6a593d4179..c0934105cc66 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_decl.hpp
@@ -89,16 +89,15 @@ namespace MueLu {
   | CoarseMap | BlockedCoarseMapFactory | Map containing the coarse map used as domain map in the tentative prolongation operator
 */
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-class BlockedCoarseMapFactory : public MueLu::CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> { //SingleLevelFactoryBase {
+          class Node          = DefaultNode>
+class BlockedCoarseMapFactory : public MueLu::CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> {  // SingleLevelFactoryBase {
 #undef MUELU_BLOCKEDCOARSEMAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-public:
-
+ public:
   //! @name Input
   //@{
 
@@ -121,14 +120,13 @@ class BlockedCoarseMapFactory : public MueLu::CoarseMapFactory<Scalar, LocalOrdi
 
   //@}
 
-protected:
-
+ protected:
   //! Compute domain GID offset from another CoarseMap
-  virtual GlobalOrdinal GetDomainGIDOffset(Level& currentLevel) const final;
+  virtual GlobalOrdinal GetDomainGIDOffset(Level &currentLevel) const final;
 
-}; //class BlockedCoarseMapFactory
+};  // class BlockedCoarseMapFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDCOARSEMAPFACTORY_SHORT
 #endif /* MUELU_BLOCKEDCOARSEMAPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_def.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_def.hpp
index 88dcdf6e7779..b0c16e4957e8 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedCoarseMapFactory_def.hpp
@@ -60,54 +60,50 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-  {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Generating factory for aggregates.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory for null space.");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of previous coarse map. (must be set by user!).");
-
-    // do we need this?
-    validParamList->set<std::string>("Striding info", "{}", "Striding information");
-    validParamList->set<LocalOrdinal>("Strided block id", -1, "Strided block id");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-  {
-    this->Input(currentLevel, "Aggregates");
-    this->Input(currentLevel, "Nullspace");
-
-    // Get CoarseMap from previously defined block
-    RCP<const FactoryBase> prevCoarseMapFact = this->GetFactory("CoarseMap");
-    TEUCHOS_TEST_FOR_EXCEPTION(prevCoarseMapFact==Teuchos::null, Exceptions::RuntimeError, "MueLu::BlockedCoarseMapFactory::getDomainMapOffset: user did not specify CoarseMap of previous block. Do not forget to set the CoarseMap factory.");
-    currentLevel.DeclareInput("CoarseMap", prevCoarseMapFact.get(), this);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const
-  {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    GlobalOrdinal domainGIDOffset = GetDomainGIDOffset(currentLevel);
-    CoarseMapFactory::BuildCoarseMap(currentLevel, domainGIDOffset);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDomainGIDOffset(
-    Level& currentLevel) const
-  {
-    RCP<const FactoryBase> prevCoarseMapFact = this->GetFactory("CoarseMap");
-    RCP<const Map> subPDomainMap = currentLevel.Get<RCP<const Map>>("CoarseMap", prevCoarseMapFact.get());
-    GlobalOrdinal maxGlobalIndex = subPDomainMap->getMaxAllGlobalIndex();
-
-    return maxGlobalIndex + Teuchos::ScalarTraits<GlobalOrdinal>::one();
-  }
-
-} //namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("Aggregates", Teuchos::null, "Generating factory for aggregates.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Generating factory for null space.");
+  validParamList->set<RCP<const FactoryBase>>("CoarseMap", Teuchos::null, "Generating factory of previous coarse map. (must be set by user!).");
+
+  // do we need this?
+  validParamList->set<std::string>("Striding info", "{}", "Striding information");
+  validParamList->set<LocalOrdinal>("Strided block id", -1, "Strided block id");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  this->Input(currentLevel, "Aggregates");
+  this->Input(currentLevel, "Nullspace");
+
+  // Get CoarseMap from previously defined block
+  RCP<const FactoryBase> prevCoarseMapFact = this->GetFactory("CoarseMap");
+  TEUCHOS_TEST_FOR_EXCEPTION(prevCoarseMapFact == Teuchos::null, Exceptions::RuntimeError, "MueLu::BlockedCoarseMapFactory::getDomainMapOffset: user did not specify CoarseMap of previous block. Do not forget to set the CoarseMap factory.");
+  currentLevel.DeclareInput("CoarseMap", prevCoarseMapFact.get(), this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  GlobalOrdinal domainGIDOffset = GetDomainGIDOffset(currentLevel);
+  CoarseMapFactory::BuildCoarseMap(currentLevel, domainGIDOffset);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal BlockedCoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDomainGIDOffset(
+    Level &currentLevel) const {
+  RCP<const FactoryBase> prevCoarseMapFact = this->GetFactory("CoarseMap");
+  RCP<const Map> subPDomainMap             = currentLevel.Get<RCP<const Map>>("CoarseMap", prevCoarseMapFact.get());
+  GlobalOrdinal maxGlobalIndex             = subPDomainMap->getMaxAllGlobalIndex();
+
+  return maxGlobalIndex + Teuchos::ScalarTraits<GlobalOrdinal>::one();
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_BLOCKEDCOARSEMAPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_decl.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_decl.hpp
index 71fa06b007b1..03cca3339fcd 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_decl.hpp
@@ -58,144 +58,142 @@
 
 namespace MueLu {
 
-
-  /*!
-    @class BlockedPFactory class.
-    @brief Factory for building blocked, segregated prolongation operators.
-
-    Factory for building blocked, segregated prolongation operators of the form
-    \f$ P=\diag(P_{11},P_{22},\ldots)\f$, where \f$ P_{ii}\f$ are prolongation operators
-    for the corresponding subblocks \f$A_{ii}\f$ in the blocked operator \f$ A \f$.
-
-    @param RCP<FactoryBase> AFact = Teuchos::null: factory for generating blocked operator \f$ A\f$.
-
-    The blocked operator \f$ A \f$ is needed for accessing the underlaying blocked maps (MapExtractors).
-    Use the AddFactoryManager function to define block rows in the blocked operator. For each block row you have to define
-    a special factory manager that handles the prolongation and restriction operator for the corresponding blocks. The
-    SubBlockAFactory class provides a factory interface for accessing specific blocks in the blocked operator \f$ A \f$.
-
-    \code{.cpp}
-    // define SubBlockAFactory objects for accessing the diagonal blocks in the blocked operator A
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 0, 0));
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 1, 1));
-
-    // define transfer operators for first block row
-    RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
-    RCP<TransPFactory> R11Fact = rcp(new TransPFactory(P11Fact));
-
-    // define transfer operators for second block row
-    RCP<TentativePFactory> P22TentFact = rcp(new TentativePFactory());
-    RCP<PgPFactory> P22Fact = rcp(new PgPFactory(P22TentFact));
-    RCP<GenericRFactory> R22Fact = rcp(new GenericRFactory(P22Fact));
-
-    // setup facatory managers for first and second block row
-    RCP<FactoryManager> M11 = rcp(new FactoryManager());
-    M11->SetFactory("A", A11Fact);
-    M11->SetFactory("P", P11Fact);
-    M11->SetFactory("R", R11Fact);
-
-    RCP<FactoryManager> M22 = rcp(new FactoryManager());
-    M22->SetFactory("A", A22Fact);
-    M22->SetFactory("P", P22Fact);
-    M22->SetFactory("R", R22Fact);
-
-    // setup blocked transfer operator
-    // use Teuchos::null as AFactory -> standard "A" variable (in this case the blocked operator A)
-    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
-    PFact->SetFactory("A",Teuchos::null); // not necessary (use blocked operator A)
-    PFact->AddFactoryManager(M11); // add factory manager for first block row
-    PFact->AddFactoryManager(M22); // add factory manager for second block row
-
-    // define restriction operator
-    // Note: always use GenericRFactory here.
-    // The simple TransPFactory class cannot handle blocked operators, yet.
-    RCP<GenericRFactory> RFact = rcp(new GenericRFactory(PFact));
-
-    // RAP factory
-    RCP<RAPFactory> AcFact = rcp(new RAPFactory(PFact, RFact));
-    \endcode
-
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class BlockedPFactory : public PFactory {
+/*!
+  @class BlockedPFactory class.
+  @brief Factory for building blocked, segregated prolongation operators.
+
+  Factory for building blocked, segregated prolongation operators of the form
+  \f$ P=\diag(P_{11},P_{22},\ldots)\f$, where \f$ P_{ii}\f$ are prolongation operators
+  for the corresponding subblocks \f$A_{ii}\f$ in the blocked operator \f$ A \f$.
+
+  @param RCP<FactoryBase> AFact = Teuchos::null: factory for generating blocked operator \f$ A\f$.
+
+  The blocked operator \f$ A \f$ is needed for accessing the underlaying blocked maps (MapExtractors).
+  Use the AddFactoryManager function to define block rows in the blocked operator. For each block row you have to define
+  a special factory manager that handles the prolongation and restriction operator for the corresponding blocks. The
+  SubBlockAFactory class provides a factory interface for accessing specific blocks in the blocked operator \f$ A \f$.
+
+  \code{.cpp}
+  // define SubBlockAFactory objects for accessing the diagonal blocks in the blocked operator A
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 0, 0));
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 1, 1));
+
+  // define transfer operators for first block row
+  RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
+  RCP<TransPFactory> R11Fact = rcp(new TransPFactory(P11Fact));
+
+  // define transfer operators for second block row
+  RCP<TentativePFactory> P22TentFact = rcp(new TentativePFactory());
+  RCP<PgPFactory> P22Fact = rcp(new PgPFactory(P22TentFact));
+  RCP<GenericRFactory> R22Fact = rcp(new GenericRFactory(P22Fact));
+
+  // setup facatory managers for first and second block row
+  RCP<FactoryManager> M11 = rcp(new FactoryManager());
+  M11->SetFactory("A", A11Fact);
+  M11->SetFactory("P", P11Fact);
+  M11->SetFactory("R", R11Fact);
+
+  RCP<FactoryManager> M22 = rcp(new FactoryManager());
+  M22->SetFactory("A", A22Fact);
+  M22->SetFactory("P", P22Fact);
+  M22->SetFactory("R", R22Fact);
+
+  // setup blocked transfer operator
+  // use Teuchos::null as AFactory -> standard "A" variable (in this case the blocked operator A)
+  RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
+  PFact->SetFactory("A",Teuchos::null); // not necessary (use blocked operator A)
+  PFact->AddFactoryManager(M11); // add factory manager for first block row
+  PFact->AddFactoryManager(M22); // add factory manager for second block row
+
+  // define restriction operator
+  // Note: always use GenericRFactory here.
+  // The simple TransPFactory class cannot handle blocked operators, yet.
+  RCP<GenericRFactory> RFact = rcp(new GenericRFactory(PFact));
+
+  // RAP factory
+  RCP<RAPFactory> AcFact = rcp(new RAPFactory(PFact, RFact));
+  \endcode
+
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class BlockedPFactory : public PFactory {
 #undef MUELU_BLOCKEDPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
-
-    /*! @brief Constructor.
-      User can supply a factory for generating the tentative prolongator.
-    */
-    BlockedPFactory(/*RCP<FactoryBase> AFact = Teuchos::null*/): diagonalView_("current") { }
-
-    //! Destructor.
-    virtual ~BlockedPFactory() { }
+  /*! @brief Constructor.
+    User can supply a factory for generating the tentative prolongator.
+  */
+  BlockedPFactory(/*RCP<FactoryBase> AFact = Teuchos::null*/)
+    : diagonalView_("current") {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~BlockedPFactory() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Set methods.
-    //@{
+  //@}
 
-    //! Change view of diagonal.
-    void SetDiagonalView(std::string const& diagView) { diagonalView_ = diagView; }
+  //! @name Set methods.
+  //@{
 
-    //! Add a factory manager
-    void AddFactoryManager(RCP<const FactoryManagerBase> FactManager) { FactManager_.push_back(FactManager); }
+  //! Change view of diagonal.
+  void SetDiagonalView(std::string const &diagView) { diagonalView_ = diagView; }
 
-    //@}
+  //! Add a factory manager
+  void AddFactoryManager(RCP<const FactoryManagerBase> FactManager) { FactManager_.push_back(FactManager); }
 
-    //! @name Get methods.
-    //@{
+  //@}
 
-    //! Returns current view of diagonal.
-    std::string GetDiagonalView() { return diagonalView_; }
+  //! @name Get methods.
+  //@{
 
-    //@}
+  //! Returns current view of diagonal.
+  std::string GetDiagonalView() { return diagonalView_; }
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
-    //@}
+  //! Input
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    /*!
-      @brief Build method.
+  //! @name Build methods.
+  //@{
 
-      Build segregated blocked prolongator by assembly from prolongators of each subblock in A
-      and return it in \c coarseLevel .
-    */
-    void Build(Level& fineLevel, Level &coarseLevel) const;
+  /*!
+    @brief Build method.
 
-    void BuildP(Level &fineLevel, Level &coarseLevel) const;
+    Build segregated blocked prolongator by assembly from prolongators of each subblock in A
+    and return it in \c coarseLevel .
+  */
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  void BuildP(Level &fineLevel, Level &coarseLevel) const;
 
-  private:
-    bool areGidsUnique(const std::vector<GO>& X) const {
-      std::set<GO> Y(X.begin(), X.end());
-      return X.size() == Y.size();
-    }
+  //@}
 
-    //! Input factories
-    std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
+ private:
+  bool areGidsUnique(const std::vector<GO> &X) const {
+    std::set<GO> Y(X.begin(), X.end());
+    return X.size() == Y.size();
+  }
 
-    //! Factory parameters
-    std::string diagonalView_;
+  //! Input factories
+  std::vector<Teuchos::RCP<const FactoryManagerBase> > FactManager_;
 
-  };
+  //! Factory parameters
+  std::string diagonalView_;
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_BLOCKEDPFACTORY_SHORT
 #endif /* MUELU_BLOCKEDPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_def.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_def.hpp
index 2bd92b673397..1309078337fb 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_BlockedPFactory_def.hpp
@@ -67,243 +67,245 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("A",          null, "Generating factory of the matrix A (block matrix)");
-    validParamList->set< bool >                  ("backwards", false, "Forward/backward order");
+  validParamList->set<RCP<const FactoryBase> >("A", null, "Generating factory of the matrix A (block matrix)");
+  validParamList->set<bool>("backwards", false, "Forward/backward order");
 
-    return validParamList;
-  }
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+
+  const ParameterList& pL = GetParameterList();
+  const bool backwards    = pL.get<bool>("backwards");
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    Input(fineLevel, "A");
+  const int numFactManagers = FactManager_.size();
+  for (int k = 0; k < numFactManagers; k++) {
+    int i                                            = (backwards ? numFactManagers - 1 - k : k);
+    const RCP<const FactoryManagerBase>& factManager = FactManager_[i];
 
-    const ParameterList& pL = GetParameterList();
-    const bool backwards = pL.get<bool>("backwards");
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), factManager);
+    SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), factManager);
+
+    if (!restrictionMode_)
+      coarseLevel.DeclareInput("P", factManager->GetFactory("P").get(), this);
+    else
+      coarseLevel.DeclareInput("R", factManager->GetFactory("R").get(), this);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& /* fineLevel */, Level& /* coarseLevel */) const {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  RCP<Matrix> Ain = Get<RCP<Matrix> >(fineLevel, "A");
+
+  RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
+  TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
+
+  const int numFactManagers = FactManager_.size();
+
+  // Plausibility check
+  TEUCHOS_TEST_FOR_EXCEPTION(A->Rows() != as<size_t>(numFactManagers), Exceptions::RuntimeError,
+                             "Number of block rows [" << A->Rows() << "] does not match the number of SubFactorManagers [" << numFactManagers << "]");
+  TEUCHOS_TEST_FOR_EXCEPTION(A->Cols() != as<size_t>(numFactManagers), Exceptions::RuntimeError,
+                             "Number of block cols [" << A->Cols() << "] does not match the number of SubFactorManagers [" << numFactManagers << "]");
+
+  // Build blocked prolongator
+  std::vector<RCP<Matrix> > subBlockP(numFactManagers);
+  std::vector<RCP<const Map> > subBlockPRangeMaps(numFactManagers);
+  std::vector<RCP<const Map> > subBlockPDomainMaps(numFactManagers);
+
+  std::vector<GO> fullRangeMapVector;
+  std::vector<GO> fullDomainMapVector;
+
+  RCP<const MapExtractor> rangeAMapExtractor  = A->getRangeMapExtractor();
+  RCP<const MapExtractor> domainAMapExtractor = A->getDomainMapExtractor();
+
+  const ParameterList& pL = GetParameterList();
+  const bool backwards    = pL.get<bool>("backwards");
+
+  // Build and store the subblocks and the corresponding range and domain
+  // maps.  Since we put together the full range and domain map from the
+  // submaps, we do not have to use the maps from blocked A
+  for (int k = 0; k < numFactManagers; k++) {
+    int i = (backwards ? numFactManagers - 1 - k : k);
+    if (restrictionMode_)
+      GetOStream(Runtime1) << "Generating R for block " << k << "/" << numFactManagers << std::endl;
+    else
+      GetOStream(Runtime1) << "Generating P for block " << k << "/" << numFactManagers << std::endl;
+
+    const RCP<const FactoryManagerBase>& factManager = FactManager_[i];
+
+    SetFactoryManager fineSFM(rcpFromRef(fineLevel), factManager);
+    SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), factManager);
+
+    if (!restrictionMode_)
+      subBlockP[i] = coarseLevel.Get<RCP<Matrix> >("P", factManager->GetFactory("P").get());
+    else
+      subBlockP[i] = coarseLevel.Get<RCP<Matrix> >("R", factManager->GetFactory("R").get());
+
+    // Check if prolongator/restrictor operators have strided maps
+    TEUCHOS_TEST_FOR_EXCEPTION(subBlockP[i]->IsView("stridedMaps") == false, Exceptions::BadCast,
+                               "subBlock P operator [" << i << "] has no strided map information.");
+
+    // Append strided row map (= range map) to list of range maps.
+    // TAW the row map GIDs extracted here represent the actual row map GIDs.
+    // No support for nested operators! (at least if Thyra style gids are used)
+    RCP<const StridedMap> strPartialMap = Teuchos::rcp_dynamic_cast<const StridedMap>(subBlockP[i]->getRowMap("stridedMaps"));
+    std::vector<size_t> stridedRgData   = strPartialMap->getStridingData();
+    subBlockPRangeMaps[i]               = StridedMapFactory::Build(
+                      subBlockP[i]->getRangeMap(),  // actual global IDs (Thyra or Xpetra)
+                      stridedRgData,
+                      strPartialMap->getStridedBlockId(),
+                      strPartialMap->getOffset());
+    // subBlockPRangeMaps[i] = subBlockP[i]->getRowMap("stridedMaps");
+
+    // Use plain range map to determine the DOF ids
+    ArrayView<const GO> nodeRangeMap = subBlockPRangeMaps[i]->getLocalElementList();
+    fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMap.begin(), nodeRangeMap.end());
+
+    // Append strided col map (= domain map) to list of range maps.
+    RCP<const StridedMap> strPartialMap2 = Teuchos::rcp_dynamic_cast<const StridedMap>(subBlockP[i]->getColMap("stridedMaps"));
+    std::vector<size_t> stridedRgData2   = strPartialMap2->getStridingData();
+    subBlockPDomainMaps[i]               = StridedMapFactory::Build(
+                      subBlockP[i]->getDomainMap(),  // actual global IDs (Thyra or Xpetra)
+                      stridedRgData2,
+                      strPartialMap2->getStridedBlockId(),
+                      strPartialMap2->getOffset());
+    // subBlockPDomainMaps[i] = subBlockP[i]->getColMap("stridedMaps");
+
+    // Use plain domain map to determine the DOF ids
+    ArrayView<const GO> nodeDomainMap = subBlockPDomainMaps[i]->getLocalElementList();
+    fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMap.begin(), nodeDomainMap.end());
+  }
 
-    const int numFactManagers = FactManager_.size();
-    for (int k = 0; k < numFactManagers; k++) {
-      int i = (backwards ? numFactManagers-1 - k : k);
-      const RCP<const FactoryManagerBase>& factManager = FactManager_[i];
+  // check if GIDs for full maps have to be sorted:
+  // For the Thyra mode ordering they do not have to be sorted since the GIDs are
+  // numbered as 0...n1,0...,n2 (starting with zero for each subblock). The MapExtractor
+  // generates unique GIDs during the construction.
+  // For Xpetra style, the GIDs have to be reordered. Such that one obtains a ordered
+  // list of GIDs in an increasing ordering. In Xpetra, the GIDs are all unique through
+  // out all submaps.
+  bool bDoSortRangeGIDs  = rangeAMapExtractor->getThyraMode() ? false : true;
+  bool bDoSortDomainGIDs = domainAMapExtractor->getThyraMode() ? false : true;
+
+  if (bDoSortRangeGIDs) {
+    sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
+    fullRangeMapVector.erase(std::unique(fullRangeMapVector.begin(), fullRangeMapVector.end()), fullRangeMapVector.end());
+  }
+  if (bDoSortDomainGIDs) {
+    sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
+    fullDomainMapVector.erase(std::unique(fullDomainMapVector.begin(), fullDomainMapVector.end()), fullDomainMapVector.end());
+  }
 
-      SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   factManager);
-      SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), factManager);
+  // extract map index base from maps of blocked A
+  GO rangeIndexBase  = 0;
+  GO domainIndexBase = 0;
+  if (!restrictionMode_) {
+    // Prolongation mode: just use index base of range and domain map of A
+    rangeIndexBase  = A->getRangeMap()->getIndexBase();
+    domainIndexBase = A->getDomainMap()->getIndexBase();
+
+  } else {
+    // Restriction mode: switch range and domain map for blocked restriction operator
+    rangeIndexBase  = A->getDomainMap()->getIndexBase();
+    domainIndexBase = A->getRangeMap()->getIndexBase();
+  }
 
-      if (!restrictionMode_)
-        coarseLevel.DeclareInput("P", factManager->GetFactory("P").get(), this);
-      else
-        coarseLevel.DeclareInput("R", factManager->GetFactory("R").get(), this);
-    }
+  // Build full range map.
+  // If original range map has striding information, then transfer it to the
+  // new range map
+  RCP<const StridedMap> stridedRgFullMap = rcp_dynamic_cast<const StridedMap>(rangeAMapExtractor->getFullMap());
+  RCP<const Map> fullRangeMap            = Teuchos::null;
+
+  ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0, fullRangeMapVector.size());
+  if (stridedRgFullMap != Teuchos::null) {
+    std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
+    fullRangeMap                    = StridedMapFactory::Build(
+                           A->getRangeMap()->lib(),
+                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                           fullRangeMapGIDs,
+                           rangeIndexBase,
+                           stridedData,
+                           A->getRangeMap()->getComm(),
+                           -1, /* the full map vector should always have strided block id -1! */
+                           stridedRgFullMap->getOffset());
+  } else {
+    fullRangeMap = MapFactory::Build(
+        A->getRangeMap()->lib(),
+        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+        fullRangeMapGIDs,
+        rangeIndexBase,
+        A->getRangeMap()->getComm());
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& /* fineLevel */, Level& /* coarseLevel */) const
-  { }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void BlockedPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    RCP<Matrix> Ain = Get< RCP<Matrix> >(fineLevel, "A");
-
-    RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
-    TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
-
-    const int numFactManagers = FactManager_.size();
-
-    // Plausibility check
-    TEUCHOS_TEST_FOR_EXCEPTION(A->Rows() != as<size_t>(numFactManagers), Exceptions::RuntimeError,
-                               "Number of block rows [" << A->Rows() << "] does not match the number of SubFactorManagers [" << numFactManagers << "]");
-    TEUCHOS_TEST_FOR_EXCEPTION(A->Cols() != as<size_t>(numFactManagers), Exceptions::RuntimeError,
-                               "Number of block cols [" << A->Cols() << "] does not match the number of SubFactorManagers [" << numFactManagers << "]");
-
-    // Build blocked prolongator
-    std::vector<RCP<Matrix> >     subBlockP          (numFactManagers);
-    std::vector<RCP<const Map> >  subBlockPRangeMaps (numFactManagers);
-    std::vector<RCP<const Map> >  subBlockPDomainMaps(numFactManagers);
-
-    std::vector<GO> fullRangeMapVector;
-    std::vector<GO> fullDomainMapVector;
-
-    RCP<const MapExtractor> rangeAMapExtractor = A->getRangeMapExtractor();
-    RCP<const MapExtractor> domainAMapExtractor = A->getDomainMapExtractor();
-
-    const ParameterList& pL = GetParameterList();
-    const bool backwards = pL.get<bool>("backwards");
-
-    // Build and store the subblocks and the corresponding range and domain
-    // maps.  Since we put together the full range and domain map from the
-    // submaps, we do not have to use the maps from blocked A
-    for (int k = 0; k < numFactManagers; k++) {
-      int i = (backwards ? numFactManagers-1 - k : k);
-       if (restrictionMode_) GetOStream(Runtime1) << "Generating R for block " << k <<"/"<<numFactManagers <<std::endl;
-       else GetOStream(Runtime1) << "Generating P for block " << k <<"/"<<numFactManagers <<std::endl;
-
-      const RCP<const FactoryManagerBase>& factManager = FactManager_[i];
-
-      SetFactoryManager fineSFM  (rcpFromRef(fineLevel),   factManager);
-      SetFactoryManager coarseSFM(rcpFromRef(coarseLevel), factManager);
-
-      if (!restrictionMode_) subBlockP[i] = coarseLevel.Get<RCP<Matrix> >("P", factManager->GetFactory("P").get());
-      else                   subBlockP[i] = coarseLevel.Get<RCP<Matrix> >("R", factManager->GetFactory("R").get());
-
-      // Check if prolongator/restrictor operators have strided maps
-      TEUCHOS_TEST_FOR_EXCEPTION(subBlockP[i]->IsView("stridedMaps") == false, Exceptions::BadCast,
-                                 "subBlock P operator [" << i << "] has no strided map information.");
-
-      // Append strided row map (= range map) to list of range maps.
-      // TAW the row map GIDs extracted here represent the actual row map GIDs.
-      // No support for nested operators! (at least if Thyra style gids are used)
-      RCP<const StridedMap> strPartialMap = Teuchos::rcp_dynamic_cast<const StridedMap>(subBlockP[i]->getRowMap("stridedMaps"));
-      std::vector<size_t> stridedRgData = strPartialMap->getStridingData();
-      subBlockPRangeMaps[i] = StridedMapFactory::Build(
-          subBlockP[i]->getRangeMap(), // actual global IDs (Thyra or Xpetra)
-          stridedRgData,
-          strPartialMap->getStridedBlockId(),
-          strPartialMap->getOffset());
-      //subBlockPRangeMaps[i] = subBlockP[i]->getRowMap("stridedMaps");
-
-      // Use plain range map to determine the DOF ids
-      ArrayView<const GO> nodeRangeMap = subBlockPRangeMaps[i]->getLocalElementList();
-      fullRangeMapVector.insert(fullRangeMapVector.end(), nodeRangeMap.begin(), nodeRangeMap.end());
-
-      // Append strided col map (= domain map) to list of range maps.
-      RCP<const StridedMap> strPartialMap2 = Teuchos::rcp_dynamic_cast<const StridedMap>(subBlockP[i]->getColMap("stridedMaps"));
-      std::vector<size_t> stridedRgData2 = strPartialMap2->getStridingData();
-      subBlockPDomainMaps[i] = StridedMapFactory::Build(
-          subBlockP[i]->getDomainMap(), // actual global IDs (Thyra or Xpetra)
-          stridedRgData2,
-          strPartialMap2->getStridedBlockId(),
-          strPartialMap2->getOffset());
-      //subBlockPDomainMaps[i] = subBlockP[i]->getColMap("stridedMaps");
-
-      // Use plain domain map to determine the DOF ids
-      ArrayView<const GO> nodeDomainMap = subBlockPDomainMaps[i]->getLocalElementList();
-      fullDomainMapVector.insert(fullDomainMapVector.end(), nodeDomainMap.begin(), nodeDomainMap.end());
-    }
-
-    // check if GIDs for full maps have to be sorted:
-    // For the Thyra mode ordering they do not have to be sorted since the GIDs are
-    // numbered as 0...n1,0...,n2 (starting with zero for each subblock). The MapExtractor
-    // generates unique GIDs during the construction.
-    // For Xpetra style, the GIDs have to be reordered. Such that one obtains a ordered
-    // list of GIDs in an increasing ordering. In Xpetra, the GIDs are all unique through
-    // out all submaps.
-    bool bDoSortRangeGIDs  = rangeAMapExtractor->getThyraMode()  ? false : true;
-    bool bDoSortDomainGIDs = domainAMapExtractor->getThyraMode() ? false : true;
-
-    if (bDoSortRangeGIDs) {
-      sort(fullRangeMapVector.begin(), fullRangeMapVector.end());
-      fullRangeMapVector.erase(std::unique(fullRangeMapVector.begin(), fullRangeMapVector.end()), fullRangeMapVector.end());
-    }
-    if (bDoSortDomainGIDs) {
-      sort(fullDomainMapVector.begin(), fullDomainMapVector.end());
-      fullDomainMapVector.erase(std::unique(fullDomainMapVector.begin(), fullDomainMapVector.end()), fullDomainMapVector.end());
-    }
-
-    // extract map index base from maps of blocked A
-    GO rangeIndexBase  = 0;
-    GO domainIndexBase = 0;
-    if (!restrictionMode_) {
-      // Prolongation mode: just use index base of range and domain map of A
-      rangeIndexBase  = A->getRangeMap() ->getIndexBase();
-      domainIndexBase = A->getDomainMap()->getIndexBase();
-
-    } else {
-      // Restriction mode: switch range and domain map for blocked restriction operator
-      rangeIndexBase  = A->getDomainMap()->getIndexBase();
-      domainIndexBase = A->getRangeMap()->getIndexBase();
-    }
-
-    // Build full range map.
-    // If original range map has striding information, then transfer it to the
-    // new range map
-    RCP<const StridedMap>   stridedRgFullMap = rcp_dynamic_cast<const StridedMap>(rangeAMapExtractor->getFullMap());
-    RCP<const Map >         fullRangeMap = Teuchos::null;
-
-    ArrayView<GO> fullRangeMapGIDs(fullRangeMapVector.size() ? &fullRangeMapVector[0] : 0, fullRangeMapVector.size());
-    if (stridedRgFullMap != Teuchos::null) {
-      std::vector<size_t> stridedData = stridedRgFullMap->getStridingData();
-      fullRangeMap = StridedMapFactory::Build(
-                                   A->getRangeMap()->lib(),
-                                   Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                   fullRangeMapGIDs,
-                                   rangeIndexBase,
-                                   stridedData,
-                                   A->getRangeMap()->getComm(),
-                                   -1, /* the full map vector should always have strided block id -1! */
-                                   stridedRgFullMap->getOffset());
-    } else {
-      fullRangeMap = MapFactory::Build(
-                            A->getRangeMap()->lib(),
-                            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                            fullRangeMapGIDs,
-                            rangeIndexBase,
-                            A->getRangeMap()->getComm());
-    }
-
-    ArrayView<GO>         fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0,fullDomainMapVector.size());
-    RCP<const StridedMap> stridedDoFullMap = rcp_dynamic_cast<const StridedMap>(domainAMapExtractor->getFullMap());
-    RCP<const Map >       fullDomainMap    = null;
-    if (stridedDoFullMap != null) {
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap == Teuchos::null, Exceptions::BadCast,
-        "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information!");
-
-      std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
-      fullDomainMap = StridedMapFactory::Build(
-                                   A->getDomainMap()->lib(),
-                                   Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                   fullDomainMapGIDs,
-                                   domainIndexBase,
-                                   stridedData2,
-                                   A->getDomainMap()->getComm(),
-                                   -1, /* the full map vector should always have strided block id -1! */
-                                   stridedDoFullMap->getOffset());
-    } else {
-      fullDomainMap = MapFactory::Build(
-                            A->getDomainMap()->lib(),
-                            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                            fullDomainMapGIDs,
-                            domainIndexBase,
-                            A->getDomainMap()->getComm());
-    }
-
-    // Build map extractors
-    RCP<const MapExtractor> rangeMapExtractor  = MapExtractorFactory::Build(fullRangeMap,  subBlockPRangeMaps, rangeAMapExtractor->getThyraMode());
-    RCP<const MapExtractor> domainMapExtractor = MapExtractorFactory::Build(fullDomainMap, subBlockPDomainMaps, domainAMapExtractor->getThyraMode());
-
-    RCP<BlockedCrsMatrix> P = rcp(new BlockedCrsMatrix(rangeMapExtractor, domainMapExtractor, 10));
-    for (size_t i = 0; i < subBlockPRangeMaps.size(); i++)
-      for (size_t j = 0; j < subBlockPRangeMaps.size(); j++)
-        if (i == j) {
-          RCP<CrsMatrixWrap> crsOpii  = rcp_dynamic_cast<CrsMatrixWrap>(subBlockP[i]);
-          TEUCHOS_TEST_FOR_EXCEPTION(crsOpii == Teuchos::null, Xpetra::Exceptions::BadCast,
-                                     "Block [" << i << ","<< j << "] must be of type CrsMatrixWrap.");
-          P->setMatrix(i, i, crsOpii);
-        } else {
-          P->setMatrix(i, j, Teuchos::null);
-        }
-
-    P->fillComplete();
-
-    // Level Set
-    if (!restrictionMode_) {
-      // Prolongation mode
-      coarseLevel.Set("P", rcp_dynamic_cast<Matrix>(P), this);
-      // Stick the CoarseMap on the level if somebody wants it (useful for repartitioning)
-      coarseLevel.Set("CoarseMap",P->getBlockedDomainMap(),this);
-    } else {
-      // Restriction mode
-      // We do not have to transpose the blocked R operator since the subblocks
-      // on the diagonal are already valid R subblocks
-      coarseLevel.Set("R", Teuchos::rcp_dynamic_cast<Matrix>(P), this);
-    }
+  ArrayView<GO> fullDomainMapGIDs(fullDomainMapVector.size() ? &fullDomainMapVector[0] : 0, fullDomainMapVector.size());
+  RCP<const StridedMap> stridedDoFullMap = rcp_dynamic_cast<const StridedMap>(domainAMapExtractor->getFullMap());
+  RCP<const Map> fullDomainMap           = null;
+  if (stridedDoFullMap != null) {
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedDoFullMap == Teuchos::null, Exceptions::BadCast,
+                               "MueLu::BlockedPFactory::Build: full map in domain map extractor has no striding information!");
+
+    std::vector<size_t> stridedData2 = stridedDoFullMap->getStridingData();
+    fullDomainMap                    = StridedMapFactory::Build(
+                           A->getDomainMap()->lib(),
+                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                           fullDomainMapGIDs,
+                           domainIndexBase,
+                           stridedData2,
+                           A->getDomainMap()->getComm(),
+                           -1, /* the full map vector should always have strided block id -1! */
+                           stridedDoFullMap->getOffset());
+  } else {
+    fullDomainMap = MapFactory::Build(
+        A->getDomainMap()->lib(),
+        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+        fullDomainMapGIDs,
+        domainIndexBase,
+        A->getDomainMap()->getComm());
+  }
 
+  // Build map extractors
+  RCP<const MapExtractor> rangeMapExtractor  = MapExtractorFactory::Build(fullRangeMap, subBlockPRangeMaps, rangeAMapExtractor->getThyraMode());
+  RCP<const MapExtractor> domainMapExtractor = MapExtractorFactory::Build(fullDomainMap, subBlockPDomainMaps, domainAMapExtractor->getThyraMode());
+
+  RCP<BlockedCrsMatrix> P = rcp(new BlockedCrsMatrix(rangeMapExtractor, domainMapExtractor, 10));
+  for (size_t i = 0; i < subBlockPRangeMaps.size(); i++)
+    for (size_t j = 0; j < subBlockPRangeMaps.size(); j++)
+      if (i == j) {
+        RCP<CrsMatrixWrap> crsOpii = rcp_dynamic_cast<CrsMatrixWrap>(subBlockP[i]);
+        TEUCHOS_TEST_FOR_EXCEPTION(crsOpii == Teuchos::null, Xpetra::Exceptions::BadCast,
+                                   "Block [" << i << "," << j << "] must be of type CrsMatrixWrap.");
+        P->setMatrix(i, i, crsOpii);
+      } else {
+        P->setMatrix(i, j, Teuchos::null);
+      }
+
+  P->fillComplete();
+
+  // Level Set
+  if (!restrictionMode_) {
+    // Prolongation mode
+    coarseLevel.Set("P", rcp_dynamic_cast<Matrix>(P), this);
+    // Stick the CoarseMap on the level if somebody wants it (useful for repartitioning)
+    coarseLevel.Set("CoarseMap", P->getBlockedDomainMap(), this);
+  } else {
+    // Restriction mode
+    // We do not have to transpose the blocked R operator since the subblocks
+    // on the diagonal are already valid R subblocks
+    coarseLevel.Set("R", Teuchos::rcp_dynamic_cast<Matrix>(P), this);
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_BLOCKEDPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_decl.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_decl.hpp
index cd0d2378c6f7..7760b2535901 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_decl.hpp
@@ -57,49 +57,47 @@
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_SingleLevelFactoryBase.hpp"
 
-
 namespace MueLu {
 
+/*!
+  @class ReorderBlockAFactory class.
+  @brief Factory for building a reordered (nested) block operator
 
-  /*!
-    @class ReorderBlockAFactory class.
-    @brief Factory for building a reordered (nested) block operator
-
-    Example
-    \code
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = ...;
-    bOp->fillComplete();
+  Example
+  \code
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = ...;
+  bOp->fillComplete();
 
-    TODO
-    \endcode
-  */
+  TODO
+  \endcode
+*/
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ReorderBlockAFactory : public SingleLevelFactoryBase {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ReorderBlockAFactory : public SingleLevelFactoryBase {
 #undef MUELU_REORDERBLOCKAFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! Input
-    //@{
+ public:
+  //! Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    /*! @brief Build an object with this factory.
-     */
-    void Build(Level & currentLevel) const;
+  /*! @brief Build an object with this factory.
+   */
+  void Build(Level &currentLevel) const;
 
-    //@}
-  }; // class ReorderBlockAFactory
+  //@}
+};  // class ReorderBlockAFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REORDERBLOCKAFACTORY_SHORT
 #endif /* MUELU_REORDERBLOCKAFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_def.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_def.hpp
index 8877dda77fb9..106f7df9a2ed 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ReorderBlockAFactory_def.hpp
@@ -48,7 +48,6 @@
 #ifndef MUELU_REORDERBLOCKAFACTORY_DEF_HPP_
 #define MUELU_REORDERBLOCKAFACTORY_DEF_HPP_
 
-
 #include "MueLu_ReorderBlockAFactory_decl.hpp"
 
 #include <Xpetra_BlockReorderManager.hpp>
@@ -64,121 +63,116 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("A",          MueLu::NoFactory::getRCP(), "Generating factory for A.");
+  validParamList->set<RCP<const FactoryBase> >("A", MueLu::NoFactory::getRCP(), "Generating factory for A.");
 
-    validParamList->set< std::string  >          ("Reorder Type",    "", "String describing the reordering of blocks");
+  validParamList->set<std::string>("Reorder Type", "", "String describing the reordering of blocks");
 
-    // TODO not very elegant.
-    validParamList->set< RCP<const FactoryBase> >("Map1", Teuchos::null, "Generating factory of the fine level map associated with the (1,1) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map2", Teuchos::null, "Generating factory of the fine level map associated with the (2,2) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map3", Teuchos::null, "Generating factory of the fine level map associated with the (3,3) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map4", Teuchos::null, "Generating factory of the fine level map associated with the (4,4) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map5", Teuchos::null, "Generating factory of the fine level map associated with the (5,5) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map6", Teuchos::null, "Generating factory of the fine level map associated with the (6,6) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map7", Teuchos::null, "Generating factory of the fine level map associated with the (7,7) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map8", Teuchos::null, "Generating factory of the fine level map associated with the (8,8) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Map9", Teuchos::null, "Generating factory of the fine level map associated with the (9,9) block in your n x n block matrix.");
+  // TODO not very elegant.
+  validParamList->set<RCP<const FactoryBase> >("Map1", Teuchos::null, "Generating factory of the fine level map associated with the (1,1) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map2", Teuchos::null, "Generating factory of the fine level map associated with the (2,2) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map3", Teuchos::null, "Generating factory of the fine level map associated with the (3,3) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map4", Teuchos::null, "Generating factory of the fine level map associated with the (4,4) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map5", Teuchos::null, "Generating factory of the fine level map associated with the (5,5) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map6", Teuchos::null, "Generating factory of the fine level map associated with the (6,6) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map7", Teuchos::null, "Generating factory of the fine level map associated with the (7,7) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map8", Teuchos::null, "Generating factory of the fine level map associated with the (8,8) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Map9", Teuchos::null, "Generating factory of the fine level map associated with the (9,9) block in your n x n block matrix.");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "ReorderBlockA factory", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReorderBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "ReorderBlockA factory", currentLevel);
 
-    const ParameterList& pL = GetParameterList();
-    std::string reorderStr = pL.get<std::string>("Reorder Type");
+  const ParameterList& pL = GetParameterList();
+  std::string reorderStr  = pL.get<std::string>("Reorder Type");
 
-    RCP<Matrix>           Ain = Get<RCP<Matrix> >(currentLevel, "A");
+  RCP<Matrix> Ain = Get<RCP<Matrix> >(currentLevel, "A");
 
-    RCP<BlockedCrsMatrix> A   = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
+  RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
 
-    // special case: we get a single block CrsMatrix object on the finest level and
-    // split it into a nxn blocked operator
-    if (A == Teuchos::null && currentLevel.GetLevelID() == 0)
-    {
-      GetOStream(Warnings0) << "Split input matrix (Warning: this is a rather expensive operation)" << std::endl;
+  // special case: we get a single block CrsMatrix object on the finest level and
+  // split it into a nxn blocked operator
+  if (A == Teuchos::null && currentLevel.GetLevelID() == 0) {
+    GetOStream(Warnings0) << "Split input matrix (Warning: this is a rather expensive operation)" << std::endl;
 
-      std::vector<Teuchos::RCP<const Map> > xmaps;
+    std::vector<Teuchos::RCP<const Map> > xmaps;
 
-      for(int it = 1; it < 10; it++) {
-        std::stringstream ss;
-        ss << "Map" << it;
-        if (currentLevel.IsAvailable(ss.str(), NoFactory::get())) {
-          RCP<const Map> submap = currentLevel.Get< RCP<const Map> >(ss.str(), NoFactory::get());
-          GetOStream(Runtime1) << "Use user-given submap #" << it << ": length dimension=" << submap->getGlobalNumElements() << std::endl;
-          xmaps.push_back(submap);
-        }
+    for (int it = 1; it < 10; it++) {
+      std::stringstream ss;
+      ss << "Map" << it;
+      if (currentLevel.IsAvailable(ss.str(), NoFactory::get())) {
+        RCP<const Map> submap = currentLevel.Get<RCP<const Map> >(ss.str(), NoFactory::get());
+        GetOStream(Runtime1) << "Use user-given submap #" << it << ": length dimension=" << submap->getGlobalNumElements() << std::endl;
+        xmaps.push_back(submap);
       }
+    }
 
-      bool bThyraMode = false; // no support for Thyra mode (yet)
-      RCP<const MapExtractor> map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Ain->getRowMap(),xmaps,bThyraMode);
-
-      // split null space vectors
-      // TODO: if he matrix blocks have different striding, this could be quite complicated
-      //RCP<MultiVector> nullspace1 = map_extractor->ExtractVector(nullspace,0);
-      //RCP<MultiVector> nullspace2 = map_extractor->ExtractVector(nullspace,1);
-
-      Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bOp =
-          Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SplitMatrix(*Ain,map_extractor,map_extractor,Teuchos::null,bThyraMode);
+    bool bThyraMode                       = false;  // no support for Thyra mode (yet)
+    RCP<const MapExtractor> map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Ain->getRowMap(), xmaps, bThyraMode);
 
-      TEUCHOS_TEST_FOR_EXCEPTION(Ain->getGlobalNumRows()    != bOp->getGlobalNumRows(),    Exceptions::RuntimeError,      "Split operator not consistent with input operator (different number of rows).");
-      TEUCHOS_TEST_FOR_EXCEPTION(Ain->getLocalNumRows()      != bOp->getLocalNumRows(),      Exceptions::RuntimeError,      "Split operator not consistent with input operator (different number of node rows).");
-      TEUCHOS_TEST_FOR_EXCEPTION(Ain->getLocalNumEntries()   != bOp->getLocalNumEntries(),   Exceptions::RuntimeError,      "Split operator not consistent with input operator (different number of local entries).");
-      TEUCHOS_TEST_FOR_EXCEPTION(Ain->getGlobalNumEntries() != bOp->getGlobalNumEntries(), Exceptions::RuntimeError,      "Split operator not consistent with input operator (different number of global entries).");
+    // split null space vectors
+    // TODO: if he matrix blocks have different striding, this could be quite complicated
+    // RCP<MultiVector> nullspace1 = map_extractor->ExtractVector(nullspace,0);
+    // RCP<MultiVector> nullspace2 = map_extractor->ExtractVector(nullspace,1);
 
-      A = bOp;
-    }
+    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOp =
+        Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SplitMatrix(*Ain, map_extractor, map_extractor, Teuchos::null, bThyraMode);
 
-    // we have a blocked operator as input
-    TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(),     Exceptions::BadCast,      "Input matrix A is not a BlockedCrsMatrix.");
-    GetOStream(Statistics1) << "Got a " << A->Rows() << "x" << A->Cols() << " blocked operator as input" << std::endl;
+    TEUCHOS_TEST_FOR_EXCEPTION(Ain->getGlobalNumRows() != bOp->getGlobalNumRows(), Exceptions::RuntimeError, "Split operator not consistent with input operator (different number of rows).");
+    TEUCHOS_TEST_FOR_EXCEPTION(Ain->getLocalNumRows() != bOp->getLocalNumRows(), Exceptions::RuntimeError, "Split operator not consistent with input operator (different number of node rows).");
+    TEUCHOS_TEST_FOR_EXCEPTION(Ain->getLocalNumEntries() != bOp->getLocalNumEntries(), Exceptions::RuntimeError, "Split operator not consistent with input operator (different number of local entries).");
+    TEUCHOS_TEST_FOR_EXCEPTION(Ain->getGlobalNumEntries() != bOp->getGlobalNumEntries(), Exceptions::RuntimeError, "Split operator not consistent with input operator (different number of global entries).");
 
-    // if we have a blocked operator and a reordering string, create a nested blocked operator, if not skip the process
-    if(reorderStr.empty())
-    {
-      GetOStream(Statistics1) << "No reordering information provided. Skipping reordering of A." << std::endl;
-    }
-    else
-    {
-      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString(reorderStr);
-      GetOStream(Debug) << "Reordering A using " << brm->toString() << std::endl;
-
-      Teuchos::RCP<const ReorderedBlockedCrsMatrix> brop =
-              Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(
-                      Xpetra::buildReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(brm, A));
-
-      TEUCHOS_TEST_FOR_EXCEPTION(brop.is_null(), Exceptions::RuntimeError,
-                                 "Block reordering of " << A->Rows() << "x" << A->Cols()
-                                                        << " blocked operator failed.");
-
-      GetOStream(Statistics1) << "Reordering A using " << brm->toString() << " block gives a " << brop->Rows() << "x"
-                              << brop->Cols() << " blocked operator" << std::endl;
-      GetOStream(Debug) << "Reordered operator has " << brop->getRangeMap()->getGlobalNumElements() << " rows and "
-                        << brop->getDomainMap()->getGlobalNumElements() << " columns" << std::endl;
-      GetOStream(Debug) << "Reordered operator: Use of Thyra style gids = "
-                        << brop->getRangeMapExtractor()->getThyraMode() << std::endl;
-
-      // get rid of const (we expect non-const operators stored in Level)
-      Teuchos::RCP<ReorderedBlockedCrsMatrix> bret =
-              Teuchos::rcp_const_cast<ReorderedBlockedCrsMatrix>(brop);
-
-      A = bret;
-    }
+    A = bOp;
+  }
 
-    currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A), this);
+  // we have a blocked operator as input
+  TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
+  GetOStream(Statistics1) << "Got a " << A->Rows() << "x" << A->Cols() << " blocked operator as input" << std::endl;
+
+  // if we have a blocked operator and a reordering string, create a nested blocked operator, if not skip the process
+  if (reorderStr.empty()) {
+    GetOStream(Statistics1) << "No reordering information provided. Skipping reordering of A." << std::endl;
+  } else {
+    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString(reorderStr);
+    GetOStream(Debug) << "Reordering A using " << brm->toString() << std::endl;
+
+    Teuchos::RCP<const ReorderedBlockedCrsMatrix> brop =
+        Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(
+            Xpetra::buildReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(brm, A));
+
+    TEUCHOS_TEST_FOR_EXCEPTION(brop.is_null(), Exceptions::RuntimeError,
+                               "Block reordering of " << A->Rows() << "x" << A->Cols()
+                                                      << " blocked operator failed.");
+
+    GetOStream(Statistics1) << "Reordering A using " << brm->toString() << " block gives a " << brop->Rows() << "x"
+                            << brop->Cols() << " blocked operator" << std::endl;
+    GetOStream(Debug) << "Reordered operator has " << brop->getRangeMap()->getGlobalNumElements() << " rows and "
+                      << brop->getDomainMap()->getGlobalNumElements() << " columns" << std::endl;
+    GetOStream(Debug) << "Reordered operator: Use of Thyra style gids = "
+                      << brop->getRangeMapExtractor()->getThyraMode() << std::endl;
+
+    // get rid of const (we expect non-const operators stored in Level)
+    Teuchos::RCP<ReorderedBlockedCrsMatrix> bret =
+        Teuchos::rcp_const_cast<ReorderedBlockedCrsMatrix>(brop);
+
+    A = bret;
   }
 
-} // namespace MueLu
+  currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A), this);
+}
 
-#endif /* MUELU_REORDERBLOCKAFACTORY_DEF_HPP_ */
+}  // namespace MueLu
 
+#endif /* MUELU_REORDERBLOCKAFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_decl.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_decl.hpp
index da326cd81964..ca97b54b856e 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_decl.hpp
@@ -56,95 +56,94 @@
 #include "MueLu_SingleLevelFactoryBase.hpp"
 #include "MueLu_SubBlockAFactory_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class SubBlockAFactory class.
-    @brief Factory for building a thresholded operator.
-
-    This is a very simple class to access a single matrix block in a blocked operator A.
-
-    Example
-    \code{.cpp}
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO>(mapExtractor,mapExtractor,10));
-    // ... let bOp be a 2x2 blocked operator ...
-    bOp->fillComplete();
-
-    // define factory for accessing block (0,0) in blocked operator A (assuming that the blocked operator is stored in Level class with NoFactory as generating factory)
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row", 0);
-    A11Fact->SetParameter("block col", 0);
-
-    // define factory for accessing block (1,1) in blocked operator A
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row", 1);
-    A22Fact->SetParameter("block col", 1);
-
-    RCP<Matrix> A11 = level.Get<RCP<Matrix> >("A", A11Fact); // extract (0,0) block from blocked operator A
-    RCP<Matrix> A22 = level.Get<RCP<Matrix> >("A", A22Fact); // extract (1,1) block from blocked operator A
-    \endcode
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SubBlockAFactory : public SingleLevelFactoryBase {
+/*!
+  @class SubBlockAFactory class.
+  @brief Factory for building a thresholded operator.
+
+  This is a very simple class to access a single matrix block in a blocked operator A.
+
+  Example
+  \code{.cpp}
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO>(mapExtractor,mapExtractor,10));
+  // ... let bOp be a 2x2 blocked operator ...
+  bOp->fillComplete();
+
+  // define factory for accessing block (0,0) in blocked operator A (assuming that the blocked operator is stored in Level class with NoFactory as generating factory)
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", 0);
+  A11Fact->SetParameter("block col", 0);
+
+  // define factory for accessing block (1,1) in blocked operator A
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", 1);
+  A22Fact->SetParameter("block col", 1);
+
+  RCP<Matrix> A11 = level.Get<RCP<Matrix> >("A", A11Fact); // extract (0,0) block from blocked operator A
+  RCP<Matrix> A22 = level.Get<RCP<Matrix> >("A", A22Fact); // extract (1,1) block from blocked operator A
+  \endcode
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SubBlockAFactory : public SingleLevelFactoryBase {
 #undef MUELU_SUBBLOCKAFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    SubBlockAFactory() = default;
+  //! Constructor.
+  SubBlockAFactory() = default;
 
-    //! Destructor.
-    virtual ~SubBlockAFactory() = default;
-    //@}
+  //! Destructor.
+  virtual ~SubBlockAFactory() = default;
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    void DeclareInput(Level &currentLevel) const override;
+  void DeclareInput(Level& currentLevel) const override;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    /*! @brief Build an object with this factory.
-     *
-     * Extract sub block matrix from a given blocked crs operator.
-     * Strided or block information is extracted in the following way:
-     *   1) first check whether the corresponding sub maps are strided
-     *      If yes, use the fixed block size and strided block id
-     *   2) If no, get the full map of the map extractors.
-     *      Check whether the full map is strided.
-     *      If yes, use the strided information of the full map and build
-     *      partial (strided) maps with it
-     *      If no, throw an exception
-     *
-     * For blocked operators with block maps one should use the striding
-     * information from the sub maps. For strided operators, the striding
-     * information of the full map is the best choice.
-     */
-    void Build(Level & currentLevel) const override;
+  /*! @brief Build an object with this factory.
+   *
+   * Extract sub block matrix from a given blocked crs operator.
+   * Strided or block information is extracted in the following way:
+   *   1) first check whether the corresponding sub maps are strided
+   *      If yes, use the fixed block size and strided block id
+   *   2) If no, get the full map of the map extractors.
+   *      Check whether the full map is strided.
+   *      If yes, use the strided information of the full map and build
+   *      partial (strided) maps with it
+   *      If no, throw an exception
+   *
+   * For blocked operators with block maps one should use the striding
+   * information from the sub maps. For strided operators, the striding
+   * information of the full map is the best choice.
+   */
+  void Build(Level& currentLevel) const override;
 
-    //@}
+  //@}
 
-  private:
-    bool CheckForUserSpecifiedBlockInfo(bool bRange, std::vector<size_t>& stridingInfo, LocalOrdinal& stridedBlockId) const;
+ private:
+  bool CheckForUserSpecifiedBlockInfo(bool bRange, std::vector<size_t>& stridingInfo, LocalOrdinal& stridedBlockId) const;
 
-  }; // class SubBlockAFactory
+};  // class SubBlockAFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SUBBLOCKAFACTORY_SHORT
 #endif /* MUELU_SUBBLOCKAFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_def.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_def.hpp
index 0c1ebf235264..18dd726a8d10 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_SubBlockAFactory_def.hpp
@@ -47,7 +47,6 @@
 #ifndef MUELU_SUBBLOCKAFACTORY_DEF_HPP_
 #define MUELU_SUBBLOCKAFACTORY_DEF_HPP_
 
-
 #include "MueLu_SubBlockAFactory_decl.hpp"
 
 #include <Xpetra_BlockedCrsMatrix.hpp>
@@ -60,193 +59,199 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase>>("A", MueLu::NoFactory::getRCP(), "Generating factory for A.");
-
-    validParamList->set<int>("block row", 0, "Block row of subblock matrix A");
-    validParamList->set<int>("block col", 0, "Block column of subblock matrix A");
-
-    validParamList->set<std::string >("Range map: Striding info", "{}", "Striding information for range map");
-    validParamList->set<LocalOrdinal>("Range map: Strided block id", -1, "Strided block id for range map");
-    validParamList->set<std::string >("Domain map: Striding info", "{}", "Striding information for domain map");
-    validParamList->set<LocalOrdinal>("Domain map: Strided block id", -1, "Strided block id for domain map");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    const ParameterList& pL = GetParameterList();
-    size_t row = Teuchos::as<size_t>(pL.get<int>("block row"));
-    size_t col = Teuchos::as<size_t>(pL.get<int>("block col"));
-
-    RCP<Matrix> Ain = currentLevel.Get< RCP<Matrix> >("A",this->GetFactory("A").get());
-    RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast,  "Input matrix A is not a BlockedCrsMatrix.");
-    TEUCHOS_TEST_FOR_EXCEPTION(row > A->Rows(), Exceptions::RuntimeError, "row [" << row << "] > A.Rows() [" << A->Rows() << "].");
-    TEUCHOS_TEST_FOR_EXCEPTION(col > A->Cols(), Exceptions::RuntimeError, "col [" << col << "] > A.Cols() [" << A->Cols() << "].");
-
-    // get sub-matrix
-    RCP<Matrix> Op = A->getMatrix(row, col);
-
-    // Check if it is a BlockedCrsMatrix object
-    // If it is a BlockedCrsMatrix object (most likely a ReorderedBlockedCrsMatrix)
-    // we have to distinguish whether it is a 1x1 leaf block in the ReorderedBlockedCrsMatrix
-    // or a nxm block. If it is a 1x1 leaf block, we "unpack" it and return the underlying
-    // CrsMatrixWrap object.
-    RCP<BlockedCrsMatrix> bOp = rcp_dynamic_cast<BlockedCrsMatrix>(Op);
-    if (bOp != Teuchos::null) {
-      // check if it is a 1x1 leaf block
-      if (bOp->Rows() == 1 && bOp->Cols() == 1) {
-        // return the unwrapped CrsMatrixWrap object underneath
-        Op = bOp->getCrsMatrix();
-        TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<CrsMatrixWrap>(Op) == Teuchos::null, Exceptions::BadCast,
-            "SubBlockAFactory::Build: sub block A[" << row << "," << col << "] must be a single block CrsMatrixWrap object!");
-      } else {
-        // If it is a regular nxm blocked operator, just return it.
-        // We do not set any kind of striding or blocking information as this
-        // usually would not make sense for general blocked operators
-        GetOStream(Statistics1) << "A(" << row << "," << col << ") is a " << bOp->Rows() << "x" << bOp->Cols() << " block matrix" << std::endl;
-        GetOStream(Statistics2) << "with altogether " << bOp->getGlobalNumRows() << "x" << bOp->getGlobalNumCols() << " rows and columns." << std::endl;
-        currentLevel.Set("A", Op, this);
-        return;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("A", MueLu::NoFactory::getRCP(), "Generating factory for A.");
+
+  validParamList->set<int>("block row", 0, "Block row of subblock matrix A");
+  validParamList->set<int>("block col", 0, "Block column of subblock matrix A");
+
+  validParamList->set<std::string>("Range map: Striding info", "{}", "Striding information for range map");
+  validParamList->set<LocalOrdinal>("Range map: Strided block id", -1, "Strided block id for range map");
+  validParamList->set<std::string>("Domain map: Striding info", "{}", "Striding information for domain map");
+  validParamList->set<LocalOrdinal>("Domain map: Strided block id", -1, "Strided block id for domain map");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  const ParameterList& pL = GetParameterList();
+  size_t row              = Teuchos::as<size_t>(pL.get<int>("block row"));
+  size_t col              = Teuchos::as<size_t>(pL.get<int>("block col"));
+
+  RCP<Matrix> Ain         = currentLevel.Get<RCP<Matrix>>("A", this->GetFactory("A").get());
+  RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
+  TEUCHOS_TEST_FOR_EXCEPTION(row > A->Rows(), Exceptions::RuntimeError, "row [" << row << "] > A.Rows() [" << A->Rows() << "].");
+  TEUCHOS_TEST_FOR_EXCEPTION(col > A->Cols(), Exceptions::RuntimeError, "col [" << col << "] > A.Cols() [" << A->Cols() << "].");
+
+  // get sub-matrix
+  RCP<Matrix> Op = A->getMatrix(row, col);
+
+  // Check if it is a BlockedCrsMatrix object
+  // If it is a BlockedCrsMatrix object (most likely a ReorderedBlockedCrsMatrix)
+  // we have to distinguish whether it is a 1x1 leaf block in the ReorderedBlockedCrsMatrix
+  // or a nxm block. If it is a 1x1 leaf block, we "unpack" it and return the underlying
+  // CrsMatrixWrap object.
+  RCP<BlockedCrsMatrix> bOp = rcp_dynamic_cast<BlockedCrsMatrix>(Op);
+  if (bOp != Teuchos::null) {
+    // check if it is a 1x1 leaf block
+    if (bOp->Rows() == 1 && bOp->Cols() == 1) {
+      // return the unwrapped CrsMatrixWrap object underneath
+      Op = bOp->getCrsMatrix();
+      TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<CrsMatrixWrap>(Op) == Teuchos::null, Exceptions::BadCast,
+                                 "SubBlockAFactory::Build: sub block A[" << row << "," << col << "] must be a single block CrsMatrixWrap object!");
+    } else {
+      // If it is a regular nxm blocked operator, just return it.
+      // We do not set any kind of striding or blocking information as this
+      // usually would not make sense for general blocked operators
+      GetOStream(Statistics1) << "A(" << row << "," << col << ") is a " << bOp->Rows() << "x" << bOp->Cols() << " block matrix" << std::endl;
+      GetOStream(Statistics2) << "with altogether " << bOp->getGlobalNumRows() << "x" << bOp->getGlobalNumCols() << " rows and columns." << std::endl;
+      currentLevel.Set("A", Op, this);
+      return;
     }
+  }
 
-    // The sub-block is not a BlockedCrsMatrix object, that is, we expect
-    // it to be of type CrsMatrixWrap allowing direct access to the corresponding
-    // data. For a single block CrsMatrixWrap type matrix we can/should set the
-    // corresponding striding/blocking information for the algorithms to work
-    // properly.
-    //
-    // TAW: In fact, a 1x1 BlockedCrsMatrix object also allows to access the data
-    //      directly, but this feature is nowhere really used in the algorithms.
-    //      So let's keep checking for the CrsMatrixWrap class to avoid screwing
-    //      things up
-    //
-    TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<CrsMatrixWrap>(Op) == Teuchos::null, Exceptions::BadCast,
-        "SubBlockAFactory::Build: sub block A[" << row << "," << col << "] is NOT a BlockedCrsMatrix, but also NOT a CrsMatrixWrap object. This cannot be.");
-
-    // strided maps for range and domain map of sub matrix
-    RCP<const StridedMap> stridedRangeMap = Teuchos::null;
-    RCP<const StridedMap> stridedDomainMap = Teuchos::null;
-
-    // check for user-specified striding information from XML file
-    std::vector<size_t> rangeStridingInfo;
-    std::vector<size_t> domainStridingInfo;
-    LocalOrdinal rangeStridedBlockId = 0;
-    LocalOrdinal domainStridedBlockId = 0;
-    bool bRangeUserSpecified = CheckForUserSpecifiedBlockInfo(true, rangeStridingInfo, rangeStridedBlockId);
-    bool bDomainUserSpecified = CheckForUserSpecifiedBlockInfo(false, domainStridingInfo, domainStridedBlockId);
-    TEUCHOS_TEST_FOR_EXCEPTION(bRangeUserSpecified != bDomainUserSpecified, Exceptions::RuntimeError,
-        "MueLu::SubBlockAFactory[" << row << "," << col << "]: the user has to specify either both domain and range map or none.");
-
-    // extract map information from MapExtractor
-    RCP<const MapExtractor> rangeMapExtractor = A->getRangeMapExtractor();
-    RCP<const MapExtractor> domainMapExtractor = A->getDomainMapExtractor();
-
-    RCP<const Map> rangeMap = rangeMapExtractor ->getMap(row);
-    RCP<const Map> domainMap = domainMapExtractor->getMap(col);
-
-    // use user-specified striding information if available. Otherwise try to use internal striding information from the submaps!
-    if(bRangeUserSpecified) stridedRangeMap = rcp(new StridedMap(rangeMap, rangeStridingInfo, rangeMap->getIndexBase(), rangeStridedBlockId, 0));
-    else stridedRangeMap = rcp_dynamic_cast<const StridedMap>(rangeMap);
-
-    if(bDomainUserSpecified) stridedDomainMap = rcp(new StridedMap(domainMap, domainStridingInfo, domainMap->getIndexBase(), domainStridedBlockId, 0));
-    else stridedDomainMap = rcp_dynamic_cast<const StridedMap>(domainMap);
-
-    // In case that both user-specified and internal striding information from the submaps
-    // does not contain valid striding information, try to extract it from the global maps
-    // in the map extractor.
-    if (stridedRangeMap.is_null()) {
-      RCP<const Map> fullRangeMap = rangeMapExtractor->getFullMap();
-      RCP<const StridedMap> stridedFullRangeMap = rcp_dynamic_cast<const StridedMap>(fullRangeMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedFullRangeMap.is_null(), Exceptions::BadCast, "Full rangeMap is not a strided map.");
-
-      std::vector<size_t> stridedData = stridedFullRangeMap->getStridingData();
-      if (stridedData.size() == 1 && row > 0) {
-        // We have block matrices. use striding block information 0
-        stridedRangeMap = StridedMapFactory::Build(rangeMap, stridedData, 0, stridedFullRangeMap->getOffset());
-
-      } else {
-        // We have strided matrices. use striding information of the corresponding block
-        stridedRangeMap = StridedMapFactory::Build(rangeMap, stridedData, row, stridedFullRangeMap->getOffset());
-      }
+  // The sub-block is not a BlockedCrsMatrix object, that is, we expect
+  // it to be of type CrsMatrixWrap allowing direct access to the corresponding
+  // data. For a single block CrsMatrixWrap type matrix we can/should set the
+  // corresponding striding/blocking information for the algorithms to work
+  // properly.
+  //
+  // TAW: In fact, a 1x1 BlockedCrsMatrix object also allows to access the data
+  //      directly, but this feature is nowhere really used in the algorithms.
+  //      So let's keep checking for the CrsMatrixWrap class to avoid screwing
+  //      things up
+  //
+  TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<CrsMatrixWrap>(Op) == Teuchos::null, Exceptions::BadCast,
+                             "SubBlockAFactory::Build: sub block A[" << row << "," << col << "] is NOT a BlockedCrsMatrix, but also NOT a CrsMatrixWrap object. This cannot be.");
+
+  // strided maps for range and domain map of sub matrix
+  RCP<const StridedMap> stridedRangeMap  = Teuchos::null;
+  RCP<const StridedMap> stridedDomainMap = Teuchos::null;
+
+  // check for user-specified striding information from XML file
+  std::vector<size_t> rangeStridingInfo;
+  std::vector<size_t> domainStridingInfo;
+  LocalOrdinal rangeStridedBlockId  = 0;
+  LocalOrdinal domainStridedBlockId = 0;
+  bool bRangeUserSpecified          = CheckForUserSpecifiedBlockInfo(true, rangeStridingInfo, rangeStridedBlockId);
+  bool bDomainUserSpecified         = CheckForUserSpecifiedBlockInfo(false, domainStridingInfo, domainStridedBlockId);
+  TEUCHOS_TEST_FOR_EXCEPTION(bRangeUserSpecified != bDomainUserSpecified, Exceptions::RuntimeError,
+                             "MueLu::SubBlockAFactory[" << row << "," << col << "]: the user has to specify either both domain and range map or none.");
+
+  // extract map information from MapExtractor
+  RCP<const MapExtractor> rangeMapExtractor  = A->getRangeMapExtractor();
+  RCP<const MapExtractor> domainMapExtractor = A->getDomainMapExtractor();
+
+  RCP<const Map> rangeMap  = rangeMapExtractor->getMap(row);
+  RCP<const Map> domainMap = domainMapExtractor->getMap(col);
+
+  // use user-specified striding information if available. Otherwise try to use internal striding information from the submaps!
+  if (bRangeUserSpecified)
+    stridedRangeMap = rcp(new StridedMap(rangeMap, rangeStridingInfo, rangeMap->getIndexBase(), rangeStridedBlockId, 0));
+  else
+    stridedRangeMap = rcp_dynamic_cast<const StridedMap>(rangeMap);
+
+  if (bDomainUserSpecified)
+    stridedDomainMap = rcp(new StridedMap(domainMap, domainStridingInfo, domainMap->getIndexBase(), domainStridedBlockId, 0));
+  else
+    stridedDomainMap = rcp_dynamic_cast<const StridedMap>(domainMap);
+
+  // In case that both user-specified and internal striding information from the submaps
+  // does not contain valid striding information, try to extract it from the global maps
+  // in the map extractor.
+  if (stridedRangeMap.is_null()) {
+    RCP<const Map> fullRangeMap               = rangeMapExtractor->getFullMap();
+    RCP<const StridedMap> stridedFullRangeMap = rcp_dynamic_cast<const StridedMap>(fullRangeMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedFullRangeMap.is_null(), Exceptions::BadCast, "Full rangeMap is not a strided map.");
+
+    std::vector<size_t> stridedData = stridedFullRangeMap->getStridingData();
+    if (stridedData.size() == 1 && row > 0) {
+      // We have block matrices. use striding block information 0
+      stridedRangeMap = StridedMapFactory::Build(rangeMap, stridedData, 0, stridedFullRangeMap->getOffset());
+
+    } else {
+      // We have strided matrices. use striding information of the corresponding block
+      stridedRangeMap = StridedMapFactory::Build(rangeMap, stridedData, row, stridedFullRangeMap->getOffset());
     }
+  }
 
-    if (stridedDomainMap.is_null()) {
-      RCP<const Map> fullDomainMap = domainMapExtractor->getFullMap();
-      RCP<const StridedMap> stridedFullDomainMap = rcp_dynamic_cast<const StridedMap>(fullDomainMap);
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedFullDomainMap.is_null(), Exceptions::BadCast, "Full domainMap is not a strided map");
+  if (stridedDomainMap.is_null()) {
+    RCP<const Map> fullDomainMap               = domainMapExtractor->getFullMap();
+    RCP<const StridedMap> stridedFullDomainMap = rcp_dynamic_cast<const StridedMap>(fullDomainMap);
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedFullDomainMap.is_null(), Exceptions::BadCast, "Full domainMap is not a strided map");
 
-      std::vector<size_t> stridedData = stridedFullDomainMap->getStridingData();
-      if (stridedData.size() == 1 && col > 0) {
-        // We have block matrices. use striding block information 0
-        stridedDomainMap = StridedMapFactory::Build(domainMap, stridedData, 0, stridedFullDomainMap->getOffset());
+    std::vector<size_t> stridedData = stridedFullDomainMap->getStridingData();
+    if (stridedData.size() == 1 && col > 0) {
+      // We have block matrices. use striding block information 0
+      stridedDomainMap = StridedMapFactory::Build(domainMap, stridedData, 0, stridedFullDomainMap->getOffset());
 
-      } else {
-        // We have strided matrices. use striding information of the corresponding block
-        stridedDomainMap = StridedMapFactory::Build(domainMap, stridedData, col, stridedFullDomainMap->getOffset());
-      }
+    } else {
+      // We have strided matrices. use striding information of the corresponding block
+      stridedDomainMap = StridedMapFactory::Build(domainMap, stridedData, col, stridedFullDomainMap->getOffset());
     }
-
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedRangeMap.is_null(), Exceptions::BadCast, "rangeMap " << row << " is not a strided map.");
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDomainMap.is_null(), Exceptions::BadCast, "domainMap " << col << " is not a strided map.");
-
-    GetOStream(Statistics1) << "A(" << row << "," << col << ") is a single block and has strided maps:"
-        << "\n  range  map fixed block size = " << stridedRangeMap ->getFixedBlockSize() << ", strided block id = " << stridedRangeMap ->getStridedBlockId()
-        << "\n  domain map fixed block size = " << stridedDomainMap->getFixedBlockSize() << ", strided block id = " << stridedDomainMap->getStridedBlockId() << std::endl;
-    GetOStream(Statistics2) << "A(" << row << "," << col << ") has " << Op->getGlobalNumRows() << "x" << Op->getGlobalNumCols() << " rows and columns." << std::endl;
-
-    // TODO do we really need that? we moved the code to getMatrix...
-    if (Op->IsView("stridedMaps") == true)
-      Op->RemoveView("stridedMaps");
-    Op->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(Op->IsView("stridedMaps") == false, Exceptions::RuntimeError, "Failed to set \"stridedMaps\" view.");
-
-    currentLevel.Set("A", Op, this);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CheckForUserSpecifiedBlockInfo(bool bRange, std::vector<size_t>& stridingInfo, LocalOrdinal& stridedBlockId) const {
-    const ParameterList & pL = GetParameterList();
-
-    if(bRange == true)
-      stridedBlockId = pL.get<LocalOrdinal>("Range map: Strided block id");
-    else
-      stridedBlockId = pL.get<LocalOrdinal>("Domain map: Strided block id");
-
-    // read in stridingInfo from parameter list and fill the internal member variable
-    // read the data only if the parameter "Striding info" exists and is non-empty
-    std::string str;
-    if(bRange == true) str = std::string("Range map: Striding info");
-    else str = std::string("Domain map: Striding info");
-
-    if(pL.isParameter(str)) {
-      std::string strStridingInfo = pL.get<std::string>(str);
-      if(strStridingInfo.empty() == false) {
-        Array<size_t> arrayVal = Teuchos::fromStringToArray<size_t>(strStridingInfo);
-        stridingInfo = Teuchos::createVector(arrayVal);
-      }
+  TEUCHOS_TEST_FOR_EXCEPTION(stridedRangeMap.is_null(), Exceptions::BadCast, "rangeMap " << row << " is not a strided map.");
+  TEUCHOS_TEST_FOR_EXCEPTION(stridedDomainMap.is_null(), Exceptions::BadCast, "domainMap " << col << " is not a strided map.");
+
+  GetOStream(Statistics1) << "A(" << row << "," << col << ") is a single block and has strided maps:"
+                          << "\n  range  map fixed block size = " << stridedRangeMap->getFixedBlockSize() << ", strided block id = " << stridedRangeMap->getStridedBlockId()
+                          << "\n  domain map fixed block size = " << stridedDomainMap->getFixedBlockSize() << ", strided block id = " << stridedDomainMap->getStridedBlockId() << std::endl;
+  GetOStream(Statistics2) << "A(" << row << "," << col << ") has " << Op->getGlobalNumRows() << "x" << Op->getGlobalNumCols() << " rows and columns." << std::endl;
+
+  // TODO do we really need that? we moved the code to getMatrix...
+  if (Op->IsView("stridedMaps") == true)
+    Op->RemoveView("stridedMaps");
+  Op->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(Op->IsView("stridedMaps") == false, Exceptions::RuntimeError, "Failed to set \"stridedMaps\" view.");
+
+  currentLevel.Set("A", Op, this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool SubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CheckForUserSpecifiedBlockInfo(bool bRange, std::vector<size_t>& stridingInfo, LocalOrdinal& stridedBlockId) const {
+  const ParameterList& pL = GetParameterList();
+
+  if (bRange == true)
+    stridedBlockId = pL.get<LocalOrdinal>("Range map: Strided block id");
+  else
+    stridedBlockId = pL.get<LocalOrdinal>("Domain map: Strided block id");
+
+  // read in stridingInfo from parameter list and fill the internal member variable
+  // read the data only if the parameter "Striding info" exists and is non-empty
+  std::string str;
+  if (bRange == true)
+    str = std::string("Range map: Striding info");
+  else
+    str = std::string("Domain map: Striding info");
+
+  if (pL.isParameter(str)) {
+    std::string strStridingInfo = pL.get<std::string>(str);
+    if (strStridingInfo.empty() == false) {
+      Array<size_t> arrayVal = Teuchos::fromStringToArray<size_t>(strStridingInfo);
+      stridingInfo           = Teuchos::createVector(arrayVal);
     }
-
-    if(stridingInfo.size() > 0) return true;
-    return false;
   }
 
-} // namespace MueLu
+  if (stridingInfo.size() > 0) return true;
+  return false;
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_SUBBLOCKAFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_decl.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_decl.hpp
index ba8cbc0bdf68..6024c5a322bb 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_decl.hpp
@@ -53,65 +53,63 @@
 #include "MueLu_SingleLevelFactoryBase.hpp"
 #include "MueLu_ZeroSubBlockAFactory_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class ZeroSubBlockAFactory class.
-    @brief Factory for extracting a zero block from a BlockedCrsMatrix.
+/*!
+  @class ZeroSubBlockAFactory class.
+  @brief Factory for extracting a zero block from a BlockedCrsMatrix.
 
-    This is a very simple class to access a single matrix block in a blocked operator A,
-    where the matrix block of interest is an actual zero block (as e.g. in saddle point systems).
+  This is a very simple class to access a single matrix block in a blocked operator A,
+  where the matrix block of interest is an actual zero block (as e.g. in saddle point systems).
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ZeroSubBlockAFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ZeroSubBlockAFactory : public SingleLevelFactoryBase {
 #undef MUELU_ZEROSUBBLOCKAFACTORY_SHORT
-    #include "MueLu_UseShortNames.hpp"
-
-  public:
-
-    //! Input
-    //@{
+#include "MueLu_UseShortNames.hpp"
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+ public:
+  //! Input
+  //@{
 
-    void DeclareInput(Level &currentLevel) const override;
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    //@}
+  void DeclareInput(Level &currentLevel) const override;
 
-    //@{
-    //! @name Build methods
+  //@}
 
-    /*!
-      @brief Build a zero sub-block object with this factory
+  //@{
+  //! @name Build methods
 
-      Create a zero sub block matrix, that fits into a given blocked crs operator.
-
-      Strided or block information is extracted in the following way:
-        1) first check whether the corresponding sub maps are strided
-           If yes, use the fixed block size and strided block id
-        2) If no, get the full map of the map extractors.
-           Check whether the full map is strided.
-           If yes, use the strided information of the full map and build
-           partial (strided) maps with it
-           If no, throw an exception
-
-      For blocked operators with block maps one should use the striding
-      information from the sub maps. For strided operators, the striding
-      information of the full map is the best choice.
-    */
-    void Build(Level & currentLevel) const override;
+  /*!
+    @brief Build a zero sub-block object with this factory
+
+    Create a zero sub block matrix, that fits into a given blocked crs operator.
+
+    Strided or block information is extracted in the following way:
+      1) first check whether the corresponding sub maps are strided
+         If yes, use the fixed block size and strided block id
+      2) If no, get the full map of the map extractors.
+         Check whether the full map is strided.
+         If yes, use the strided information of the full map and build
+         partial (strided) maps with it
+         If no, throw an exception
+
+    For blocked operators with block maps one should use the striding
+    information from the sub maps. For strided operators, the striding
+    information of the full map is the best choice.
+  */
+  void Build(Level &currentLevel) const override;
 
-    //@}
+  //@}
 
-  }; // class ZeroSubBlockAFactory
+};  // class ZeroSubBlockAFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ZEROSUBBLOCKAFACTORY_SHORT
 #endif /* MUELU_ZEROSUBBLOCKAFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_def.hpp b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_def.hpp
index da249fdff73e..6485c70cdee3 100644
--- a/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_def.hpp
+++ b/packages/muelu/src/Transfers/BlockedTransfers/MueLu_ZeroSubBlockAFactory_def.hpp
@@ -47,7 +47,6 @@
 #ifndef MUELU_ZEROSUBBLOCKAFACTORY_DEF_HPP_
 #define MUELU_ZEROSUBBLOCKAFACTORY_DEF_HPP_
 
-
 #include "MueLu_ZeroSubBlockAFactory_decl.hpp"
 
 #include <Xpetra_BlockedCrsMatrix.hpp>
@@ -61,84 +60,84 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set<RCP<const FactoryBase>>("A", MueLu::NoFactory::getRCP(), "Generating factory for A.");
+  validParamList->set<RCP<const FactoryBase>>("A", MueLu::NoFactory::getRCP(), "Generating factory for A.");
 
-    validParamList->set<int>("block row", 0, "Block row of subblock in block matrix A");
-    validParamList->set<int>("block col", 0, "Block column of subblock in block matrix A");
+  validParamList->set<int>("block row", 0, "Block row of subblock in block matrix A");
+  validParamList->set<int>("block col", 0, "Block column of subblock in block matrix A");
 
-    return validParamList;
-  }
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  const ParameterList& pL = GetParameterList();
+  const size_t row        = Teuchos::as<size_t>(pL.get<int>("block row"));
+  const size_t col        = Teuchos::as<size_t>(pL.get<int>("block col"));
+
+  // Get the blocked input matrix
+  RCP<Matrix> Ain         = currentLevel.Get<RCP<Matrix>>("A", this->GetFactory("A").get());
+  RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
+
+  // Perform some basic sanity checks
+  TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
+  TEUCHOS_TEST_FOR_EXCEPTION(row > A->Rows(), Exceptions::RuntimeError, "row [" << row << "] > A.Rows() [" << A->Rows() << "].");
+  TEUCHOS_TEST_FOR_EXCEPTION(col > A->Cols(), Exceptions::RuntimeError, "col [" << col << "] > A.Cols() [" << A->Cols() << "].");
+
+  // strided maps for range and domain map of sub matrix
+  RCP<const StridedMap> stridedRangeMap  = Teuchos::null;
+  RCP<const StridedMap> stridedDomainMap = Teuchos::null;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
-    Input(currentLevel, "A");
+  // extract map information from MapExtractor
+  RCP<const MapExtractor> rangeMapExtractor  = A->getRangeMapExtractor();
+  RCP<const MapExtractor> domainMapExtractor = A->getDomainMapExtractor();
+
+  // In case that both user-specified and internal striding information from the submaps
+  // does not contain valid striding information, try to extract it from the global maps
+  // in the map extractor.
+  if (stridedRangeMap.is_null()) {
+    TEUCHOS_TEST_FOR_EXCEPTION(rangeMapExtractor->getMap(row).is_null(), Exceptions::BadCast,
+                               "Range map extractor contains non-strided maps in block row " << row << ". This should not be.");
+    stridedRangeMap = rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(row));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ZeroSubBlockAFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    const ParameterList& pL = GetParameterList();
-    const size_t row = Teuchos::as<size_t>(pL.get<int>("block row"));
-    const size_t col = Teuchos::as<size_t>(pL.get<int>("block col"));
-
-    // Get the blocked input matrix
-    RCP<Matrix> Ain = currentLevel.Get<RCP<Matrix>>("A", this->GetFactory("A").get());
-    RCP<BlockedCrsMatrix> A = rcp_dynamic_cast<BlockedCrsMatrix>(Ain);
-
-    // Perform some basic sanity checks
-    TEUCHOS_TEST_FOR_EXCEPTION(A.is_null(), Exceptions::BadCast, "Input matrix A is not a BlockedCrsMatrix.");
-    TEUCHOS_TEST_FOR_EXCEPTION(row > A->Rows(), Exceptions::RuntimeError, "row [" << row << "] > A.Rows() [" << A->Rows() << "].");
-    TEUCHOS_TEST_FOR_EXCEPTION(col > A->Cols(), Exceptions::RuntimeError, "col [" << col << "] > A.Cols() [" << A->Cols() << "].");
-
-    // strided maps for range and domain map of sub matrix
-    RCP<const StridedMap> stridedRangeMap = Teuchos::null;
-    RCP<const StridedMap> stridedDomainMap = Teuchos::null;
-
-    // extract map information from MapExtractor
-    RCP<const MapExtractor> rangeMapExtractor = A->getRangeMapExtractor();
-    RCP<const MapExtractor> domainMapExtractor = A->getDomainMapExtractor();
-
-    // In case that both user-specified and internal striding information from the submaps
-    // does not contain valid striding information, try to extract it from the global maps
-    // in the map extractor.
-    if (stridedRangeMap.is_null()) {
-      TEUCHOS_TEST_FOR_EXCEPTION(rangeMapExtractor->getMap(row).is_null(), Exceptions::BadCast,
-          "Range map extractor contains non-strided maps in block row " << row << ". This should not be.");
-      stridedRangeMap = rcp_dynamic_cast<const StridedMap>(rangeMapExtractor->getMap(row));
-    }
-
-    if (stridedDomainMap.is_null()) {
-      TEUCHOS_TEST_FOR_EXCEPTION(domainMapExtractor->getMap(row).is_null(), Exceptions::BadCast,
-          "Domain map extractor contains non-strided maps in block row " << row << ". This should not be.");
-      stridedDomainMap = rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(col));
-    }
-
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedRangeMap.is_null(), Exceptions::BadCast, "rangeMap " << row << " is not a strided map.");
-    TEUCHOS_TEST_FOR_EXCEPTION(stridedDomainMap.is_null(), Exceptions::BadCast, "domainMap " << col << " is not a strided map.");
-
-    RCP<Matrix> Op = MatrixFactory::Build(stridedRangeMap, stridedDomainMap, static_cast<size_t>(0));
-    TEUCHOS_ASSERT(!Op.is_null());
-
-    Op->fillComplete(stridedDomainMap, stridedRangeMap);
-    TEUCHOS_ASSERT(Op->isFillComplete());
-
-    GetOStream(Statistics1) << "A(" << row << "," << col << ") is a single block and has strided maps:"
-        << "\n  range  map fixed block size = " << stridedRangeMap ->getFixedBlockSize() << ", strided block id = " << stridedRangeMap ->getStridedBlockId()
-        << "\n  domain map fixed block size = " << stridedDomainMap->getFixedBlockSize() << ", strided block id = " << stridedDomainMap->getStridedBlockId() << std::endl;
-    GetOStream(Statistics2) << "A(" << row << "," << col << ") has " << Op->getGlobalNumRows() << "x" << Op->getGlobalNumCols() << " rows and columns." << std::endl;
-
-    if (Op->IsView("stridedMaps") == true)
-      Op->RemoveView("stridedMaps");
-    Op->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    currentLevel.Set("A", Op, this);
+  if (stridedDomainMap.is_null()) {
+    TEUCHOS_TEST_FOR_EXCEPTION(domainMapExtractor->getMap(row).is_null(), Exceptions::BadCast,
+                               "Domain map extractor contains non-strided maps in block row " << row << ". This should not be.");
+    stridedDomainMap = rcp_dynamic_cast<const StridedMap>(domainMapExtractor->getMap(col));
   }
 
-} // namespace MueLu
+  TEUCHOS_TEST_FOR_EXCEPTION(stridedRangeMap.is_null(), Exceptions::BadCast, "rangeMap " << row << " is not a strided map.");
+  TEUCHOS_TEST_FOR_EXCEPTION(stridedDomainMap.is_null(), Exceptions::BadCast, "domainMap " << col << " is not a strided map.");
+
+  RCP<Matrix> Op = MatrixFactory::Build(stridedRangeMap, stridedDomainMap, static_cast<size_t>(0));
+  TEUCHOS_ASSERT(!Op.is_null());
+
+  Op->fillComplete(stridedDomainMap, stridedRangeMap);
+  TEUCHOS_ASSERT(Op->isFillComplete());
+
+  GetOStream(Statistics1) << "A(" << row << "," << col << ") is a single block and has strided maps:"
+                          << "\n  range  map fixed block size = " << stridedRangeMap->getFixedBlockSize() << ", strided block id = " << stridedRangeMap->getStridedBlockId()
+                          << "\n  domain map fixed block size = " << stridedDomainMap->getFixedBlockSize() << ", strided block id = " << stridedDomainMap->getStridedBlockId() << std::endl;
+  GetOStream(Statistics2) << "A(" << row << "," << col << ") has " << Op->getGlobalNumRows() << "x" << Op->getGlobalNumCols() << " rows and columns." << std::endl;
+
+  if (Op->IsView("stridedMaps") == true)
+    Op->RemoveView("stridedMaps");
+  Op->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  currentLevel.Set("A", Op, this);
+}
+
+}  // namespace MueLu
 
 #endif /* MUELU_ZEROSUBBLOCKAFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_decl.hpp b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_decl.hpp
index f1a80a4422eb..6ca5817af373 100644
--- a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_decl.hpp
@@ -64,82 +64,84 @@
 
 namespace MueLu {
 
-  /*!
-    @class ClassicalMapFactory class.
-    @brief Factory for generating F/C-splitting and a coarse level map. Used by ClassicalPFactory.
+/*!
+  @class ClassicalMapFactory class.
+  @brief Factory for generating F/C-splitting and a coarse level map. Used by ClassicalPFactory.
 
-    @ingroup MueLuTransferClasses
+  @ingroup MueLuTransferClasses
 
-    ## Input/output ##
+  ## Input/output ##
 
-    ### User parameters of this factory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    Graph | Factory | null | | * | * | Generating factory for graph.
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
+  ### User parameters of this factory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  Graph | Factory | null | | * | * | Generating factory for graph.
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
 
 
-    ### Variables provided by this factory ###
+  ### Variables provided by this factory ###
 
-    After @c Build() the following data is available (if requested)
+  After @c Build() the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Colors | ClassicalMapFactory | ArrayRCP<LO> of colors
-    | CoarseMap | CoarseMapFactory | Map containing the coarse map used as domain map in the classical prolongator
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Colors | ClassicalMapFactory | ArrayRCP<LO> of colors
+  | CoarseMap | CoarseMapFactory | Map containing the coarse map used as domain map in the classical prolongator
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class ClassicalMapFactory : public SingleLevelFactoryBase {
+          class Node          = DefaultNode>
+class ClassicalMapFactory : public SingleLevelFactoryBase {
 #undef MUELU_CLASSICALMAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! F/C/Dirichlet point type
-    typedef enum {F_PT=-1, UNASSIGNED=0, C_PT=1, DIRICHLET_PT=2} point_type;
-
-    //! @name Input
-    //@{
+ public:
+  //! F/C/Dirichlet point type
+  typedef enum { F_PT         = -1,
+                 UNASSIGNED   = 0,
+                 C_PT         = 1,
+                 DIRICHLET_PT = 2 } point_type;
 
-    RCP<const ParameterList> GetValidParameterList() const override;
+  //! @name Input
+  //@{
 
-    /*!
-      @brief Specifies the data that this class needs, and the factories that generate that data.
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput,
-      then this class will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const override;
+  /*!
+    @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
+    If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput,
+    then this class will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &currentLevel) const override;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const override;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const override;
 
+  //@}
 
-  protected:
-    virtual void GenerateCoarseMap(const Map & fineMap, LO num_c_points, Teuchos::RCP<const Map> & coarseMap) const;
+ protected:
+  virtual void GenerateCoarseMap(const Map &fineMap, LO num_c_points, Teuchos::RCP<const Map> &coarseMap) const;
 
-    virtual void DoGraphColoring(const GraphBase & graph, Teuchos::ArrayRCP<LO> & myColors, LO & numColors) const;
+  virtual void DoGraphColoring(const GraphBase &graph, Teuchos::ArrayRCP<LO> &myColors, LO &numColors) const;
 
-    virtual void DoMISNaive(const GraphBase & graph, Teuchos::ArrayRCP<LO> & myColors, LO & numColors) const;
+  virtual void DoMISNaive(const GraphBase &graph, Teuchos::ArrayRCP<LO> &myColors, LO &numColors) const;
 
-    virtual void DoDistributedGraphColoring(RCP<const GraphBase> & graph, Teuchos::ArrayRCP<LO> & myColors, LO & numColors) const;
+  virtual void DoDistributedGraphColoring(RCP<const GraphBase> &graph, Teuchos::ArrayRCP<LO> &myColors, LO &numColors) const;
 
-  }; //class ClassicalMapFactory
+};  // class ClassicalMapFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CLASSICALMAPFACTORY_SHORT
 #endif /* MUELU_CLASSICALMAPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_def.hpp b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_def.hpp
index 57a327ce3e10..defe38a5a5ba 100644
--- a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalMapFactory_def.hpp
@@ -47,12 +47,9 @@
 #ifndef MUELU_CLASSICALMAPFACTORY_DEF_HPP_
 #define MUELU_CLASSICALMAPFACTORY_DEF_HPP_
 
-
-
 #include <Teuchos_Array.hpp>
 #include <Teuchos_ArrayRCP.hpp>
 
-
 #ifdef HAVE_MPI
 #include <Teuchos_DefaultMpiComm.hpp>
 #endif
@@ -85,440 +82,420 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-  {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: deterministic");
-    SET_VALID_ENTRY("aggregation: coloring algorithm");
-    SET_VALID_ENTRY("aggregation: coloring: use color graph");
+  SET_VALID_ENTRY("aggregation: deterministic");
+  SET_VALID_ENTRY("aggregation: coloring algorithm");
+  SET_VALID_ENTRY("aggregation: coloring: use color graph");
 #undef SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("Coloring Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
- 
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-  {
-    Input(currentLevel, "A");
-    Input(currentLevel, "UnAmalgamationInfo");
-    Input(currentLevel, "Graph");
-    
-    const ParameterList& pL = GetParameterList();
-    bool use_color_graph = pL.get<bool>("aggregation: coloring: use color graph");
-    if(use_color_graph)
-      Input(currentLevel, "Coloring Graph");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const
-  {
-    FactoryMonitor m(*this, "Build", currentLevel);
-    const ParameterList& pL = GetParameterList();
-    RCP<const Matrix> A = Get<RCP<Matrix> >(currentLevel,"A");
-
-    RCP<const GraphBase> graph;
-    bool use_color_graph = pL.get<bool>("aggregation: coloring: use color graph");
-    if(use_color_graph) graph = Get<RCP<GraphBase> >(currentLevel,"Coloring Graph");
-    else graph = Get<RCP<GraphBase> >(currentLevel,"Graph");
-
-
-
-    /* ============================================================= */
-    /* Phase 1 : Compute an initial MIS                              */
-    /* ============================================================= */
-    ArrayRCP<LO> myColors;
-    LO numColors=0;
-
-    RCP<LocalOrdinalVector> fc_splitting;
-    std::string coloringAlgo = pL.get<std::string>("aggregation: coloring algorithm");
-
-    // Switch to Zoltan2 if we're parallel and Tpetra (and not file)
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase> >("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("Coloring Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "A");
+  Input(currentLevel, "UnAmalgamationInfo");
+  Input(currentLevel, "Graph");
+
+  const ParameterList& pL = GetParameterList();
+  bool use_color_graph    = pL.get<bool>("aggregation: coloring: use color graph");
+  if (use_color_graph)
+    Input(currentLevel, "Coloring Graph");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+  const ParameterList& pL = GetParameterList();
+  RCP<const Matrix> A     = Get<RCP<Matrix> >(currentLevel, "A");
+
+  RCP<const GraphBase> graph;
+  bool use_color_graph = pL.get<bool>("aggregation: coloring: use color graph");
+  if (use_color_graph)
+    graph = Get<RCP<GraphBase> >(currentLevel, "Coloring Graph");
+  else
+    graph = Get<RCP<GraphBase> >(currentLevel, "Graph");
+
+  /* ============================================================= */
+  /* Phase 1 : Compute an initial MIS                              */
+  /* ============================================================= */
+  ArrayRCP<LO> myColors;
+  LO numColors = 0;
+
+  RCP<LocalOrdinalVector> fc_splitting;
+  std::string coloringAlgo = pL.get<std::string>("aggregation: coloring algorithm");
+
+  // Switch to Zoltan2 if we're parallel and Tpetra (and not file)
 #ifdef HAVE_MUELU_ZOLTAN2
-    int numProcs = A->getRowMap()->getComm()->getSize();
-    if(coloringAlgo!="file" && numProcs>1 && graph->GetDomainMap()->lib() == Xpetra::UseTpetra)
-      coloringAlgo="Zoltan2";
+  int numProcs = A->getRowMap()->getComm()->getSize();
+  if (coloringAlgo != "file" && numProcs > 1 && graph->GetDomainMap()->lib() == Xpetra::UseTpetra)
+    coloringAlgo = "Zoltan2";
 #endif
 
     //#define CMS_DUMP
 #ifdef CMS_DUMP
-    {
-      int rank = graph->GetDomainMap()->getComm()->getRank();
+  {
+    int rank = graph->GetDomainMap()->getComm()->getRank();
 
-      printf("[%d,%d] graph local size = %dx%d\n",rank,currentLevel.GetLevelID(),(int)graph->GetDomainMap()->getLocalNumElements(),(int)graph->GetImportMap()->getLocalNumElements());
+    printf("[%d,%d] graph local size = %dx%d\n", rank, currentLevel.GetLevelID(), (int)graph->GetDomainMap()->getLocalNumElements(), (int)graph->GetImportMap()->getLocalNumElements());
 
-      std::ofstream ofs(std::string("m_dropped_graph_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
-      RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
-      graph->print(*fancy,Debug);
-    }
-    {
-      A->getRowMap()->getComm()->barrier();
-    }
+    std::ofstream ofs(std::string("m_dropped_graph_") + std::to_string(currentLevel.GetLevelID()) + std::string("_") + std::to_string(rank) + std::string(".dat"), std::ofstream::out);
+    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
+    graph->print(*fancy, Debug);
+  }
+  {
+    A->getRowMap()->getComm()->barrier();
+  }
 
 #endif
 
+  // Switch to MIS if we're in Epetra (and not file)
+  if (coloringAlgo != "file" && graph->GetDomainMap()->lib() == Xpetra::UseEpetra)
+    coloringAlgo = "MIS";
 
+  if (coloringAlgo == "file") {
+    // Read the CF splitting from disk
+    // NOTE: For interoperability reasons, this is dependent on the point_type enum not changing
+    std::string map_file   = std::string("map_fcsplitting_") + std::to_string(currentLevel.GetLevelID()) + std::string(".m");
+    std::string color_file = std::string("fcsplitting_") + std::to_string(currentLevel.GetLevelID()) + std::string(".m");
 
-    // Switch to MIS if we're in Epetra (and not file)
-    if(coloringAlgo!="file" && graph->GetDomainMap()->lib() == Xpetra::UseEpetra)
-      coloringAlgo="MIS";
-
-
-    if(coloringAlgo == "file") {
-      // Read the CF splitting from disk
-      // NOTE: For interoperability reasons, this is dependent on the point_type enum not changing
-      std::string map_file   = std::string("map_fcsplitting_") + std::to_string(currentLevel.GetLevelID()) + std::string(".m");
-      std::string color_file = std::string("fcsplitting_")     + std::to_string(currentLevel.GetLevelID()) + std::string(".m");
-
-      FILE * mapfile = fopen(map_file.c_str(),"r");
-      using real_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-      using RealValuedMultiVector = typename Xpetra::MultiVector<real_type,LO,GO,NO>;
-      RCP<RealValuedMultiVector> mv;
-           
-
-      GetOStream(Statistics1) << "Reading FC splitting from " << color_file << ", using map file " << map_file << ". On rank " << A->getRowMap()->getComm()->getRank() << " local size is " << A->getRowMap()->getLocalNumElements() << std::endl;
-      if(mapfile) {
-        fclose(mapfile);
-        RCP<const Map> colorMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(map_file, A->getRowMap()->lib(), A->getRowMap()->getComm());
-        TEUCHOS_TEST_FOR_EXCEPTION(!colorMap->isCompatible(*A->getRowMap()),std::invalid_argument,"Coloring on disk has incompatible map with A");
-
-        mv = Xpetra::IO<real_type, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(color_file,colorMap);
-      }
-      else {
-        // Use A's rowmap and hope it matches
-        mv = Xpetra::IO<real_type, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(color_file,A->getRowMap());
-      }
-      TEUCHOS_TEST_FOR_EXCEPTION(mv.is_null(),std::invalid_argument,"Coloring on disk cannot be read");
-      fc_splitting = LocalOrdinalVectorFactory::Build(A->getRowMap());
-      TEUCHOS_TEST_FOR_EXCEPTION(mv->getLocalLength() != fc_splitting->getLocalLength(),std::invalid_argument,"Coloring map mismatch");
-
-      // Overlay the Dirichlet Points (and copy out the rest)
-      auto boundaryNodes = graph->GetBoundaryNodeMap();
-      ArrayRCP<const real_type> mv_data= mv->getData(0);
-      ArrayRCP<LO> fc_data= fc_splitting->getDataNonConst(0);
-      for(LO i=0; i<(LO)fc_data.size(); i++) {
-        if(boundaryNodes[i])
-          fc_data[i] = DIRICHLET_PT;
-        else
-          fc_data[i] = Teuchos::as<LO>(mv_data[i]);
-      }
+    FILE* mapfile               = fopen(map_file.c_str(), "r");
+    using real_type             = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+    using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+    RCP<RealValuedMultiVector> mv;
+
+    GetOStream(Statistics1) << "Reading FC splitting from " << color_file << ", using map file " << map_file << ". On rank " << A->getRowMap()->getComm()->getRank() << " local size is " << A->getRowMap()->getLocalNumElements() << std::endl;
+    if (mapfile) {
+      fclose(mapfile);
+      RCP<const Map> colorMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(map_file, A->getRowMap()->lib(), A->getRowMap()->getComm());
+      TEUCHOS_TEST_FOR_EXCEPTION(!colorMap->isCompatible(*A->getRowMap()), std::invalid_argument, "Coloring on disk has incompatible map with A");
+
+      mv = Xpetra::IO<real_type, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(color_file, colorMap);
+    } else {
+      // Use A's rowmap and hope it matches
+      mv = Xpetra::IO<real_type, LocalOrdinal, GlobalOrdinal, Node>::ReadMultiVector(color_file, A->getRowMap());
     }
-#ifdef HAVE_MUELU_ZOLTAN2
-    else if(coloringAlgo.find("Zoltan2")!=std::string::npos && graph->GetDomainMap()->lib() == Xpetra::UseTpetra) {
-      SubFactoryMonitor sfm(*this,"DistributedGraphColoring",currentLevel);
-      DoDistributedGraphColoring(graph,myColors,numColors);
+    TEUCHOS_TEST_FOR_EXCEPTION(mv.is_null(), std::invalid_argument, "Coloring on disk cannot be read");
+    fc_splitting = LocalOrdinalVectorFactory::Build(A->getRowMap());
+    TEUCHOS_TEST_FOR_EXCEPTION(mv->getLocalLength() != fc_splitting->getLocalLength(), std::invalid_argument, "Coloring map mismatch");
+
+    // Overlay the Dirichlet Points (and copy out the rest)
+    auto boundaryNodes                = graph->GetBoundaryNodeMap();
+    ArrayRCP<const real_type> mv_data = mv->getData(0);
+    ArrayRCP<LO> fc_data              = fc_splitting->getDataNonConst(0);
+    for (LO i = 0; i < (LO)fc_data.size(); i++) {
+      if (boundaryNodes[i])
+        fc_data[i] = DIRICHLET_PT;
+      else
+        fc_data[i] = Teuchos::as<LO>(mv_data[i]);
     }
+  }
+#ifdef HAVE_MUELU_ZOLTAN2
+  else if (coloringAlgo.find("Zoltan2") != std::string::npos && graph->GetDomainMap()->lib() == Xpetra::UseTpetra) {
+    SubFactoryMonitor sfm(*this, "DistributedGraphColoring", currentLevel);
+    DoDistributedGraphColoring(graph, myColors, numColors);
+  }
 #endif
-    else if(coloringAlgo == "MIS" || graph->GetDomainMap()->lib() == Xpetra::UseTpetra) {
-      SubFactoryMonitor sfm(*this,"MIS",currentLevel)
-;      TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getComm()->getSize() != 1, std::invalid_argument,"MIS on more than 1 MPI rank is not supported");
-      DoMISNaive(*graph,myColors,numColors);
-    }
-    else {
-      SubFactoryMonitor sfm(*this,"GraphColoring",currentLevel);
-      TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getComm()->getSize() != 1, std::invalid_argument,"KokkosKernels graph coloring on more than 1 MPI rank is not supported");
-      DoGraphColoring(*graph,myColors,numColors);
-    }
+  else if (coloringAlgo == "MIS" || graph->GetDomainMap()->lib() == Xpetra::UseTpetra) {
+    SubFactoryMonitor sfm(*this, "MIS", currentLevel);
+    TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getComm()->getSize() != 1, std::invalid_argument, "MIS on more than 1 MPI rank is not supported");
+    DoMISNaive(*graph, myColors, numColors);
+  } else {
+    SubFactoryMonitor sfm(*this, "GraphColoring", currentLevel);
+    TEUCHOS_TEST_FOR_EXCEPTION(A->getRowMap()->getComm()->getSize() != 1, std::invalid_argument, "KokkosKernels graph coloring on more than 1 MPI rank is not supported");
+    DoGraphColoring(*graph, myColors, numColors);
+  }
 
 #ifdef CMS_DUMP
-    {
-      int rank = graph->GetDomainMap()->getComm()->getRank();
+  {
+    int rank = graph->GetDomainMap()->getComm()->getRank();
 
-      printf("[%d,%d] num colors %d\n",rank,currentLevel.GetLevelID(),numColors);
+    printf("[%d,%d] num colors %d\n", rank, currentLevel.GetLevelID(), numColors);
 
-      std::ofstream ofs(std::string("m_colors_") + std::to_string(currentLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
-      RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
-      *fancy << myColors();
-    }
-    {
-      A->getRowMap()->getComm()->barrier();
-    }
+    std::ofstream ofs(std::string("m_colors_") + std::to_string(currentLevel.GetLevelID()) + std::string("_") + std::to_string(rank) + std::string(".dat"), std::ofstream::out);
+    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
+    *fancy << myColors();
+  }
+  {
+    A->getRowMap()->getComm()->barrier();
+  }
 
 #endif
 
-    /* ============================================================= */
-    /* Phase 2 : Mark the C-Points                                   */
-    /* ============================================================= */
-    LO num_c_points = 0, num_d_points=0, num_f_points = 0;
-    if(fc_splitting.is_null()) {
-      // We just have a coloring, so we need to generate a splitting
-      auto boundaryNodes = graph->GetBoundaryNodeMap();
-      fc_splitting = LocalOrdinalVectorFactory::Build(A->getRowMap());
-      ArrayRCP<LO> myPointType = fc_splitting->getDataNonConst(0);
-      for(LO i=0; i<(LO)myColors.size(); i++) {
-        if(boundaryNodes[i]) {
-          myPointType[i] = DIRICHLET_PT;
-          num_d_points++;
-        }
-        else if ((LO)myColors[i] == 1) {
-          myPointType[i] = C_PT;
-          num_c_points++;
-        }
-        else
-          myPointType[i] = F_PT;
-      }
-      num_f_points = (LO)myColors.size() - num_d_points - num_c_points;
+  /* ============================================================= */
+  /* Phase 2 : Mark the C-Points                                   */
+  /* ============================================================= */
+  LO num_c_points = 0, num_d_points = 0, num_f_points = 0;
+  if (fc_splitting.is_null()) {
+    // We just have a coloring, so we need to generate a splitting
+    auto boundaryNodes       = graph->GetBoundaryNodeMap();
+    fc_splitting             = LocalOrdinalVectorFactory::Build(A->getRowMap());
+    ArrayRCP<LO> myPointType = fc_splitting->getDataNonConst(0);
+    for (LO i = 0; i < (LO)myColors.size(); i++) {
+      if (boundaryNodes[i]) {
+        myPointType[i] = DIRICHLET_PT;
+        num_d_points++;
+      } else if ((LO)myColors[i] == 1) {
+        myPointType[i] = C_PT;
+        num_c_points++;
+      } else
+        myPointType[i] = F_PT;
     }
-    else {
-      // If we read the splitting off disk, we just need to count
-      ArrayRCP<LO> myPointType = fc_splitting->getDataNonConst(0);
-
-      for(LO i=0; i<(LO)myPointType.size(); i++) {
-        if(myPointType[i] == DIRICHLET_PT)
-          num_d_points++;
-        else if (myPointType[i] == C_PT)
-          num_c_points++;
-      }
-      num_f_points = (LO)myPointType.size() - num_d_points - num_c_points;
+    num_f_points = (LO)myColors.size() - num_d_points - num_c_points;
+  } else {
+    // If we read the splitting off disk, we just need to count
+    ArrayRCP<LO> myPointType = fc_splitting->getDataNonConst(0);
+
+    for (LO i = 0; i < (LO)myPointType.size(); i++) {
+      if (myPointType[i] == DIRICHLET_PT)
+        num_d_points++;
+      else if (myPointType[i] == C_PT)
+        num_c_points++;
     }
+    num_f_points = (LO)myPointType.size() - num_d_points - num_c_points;
+  }
 
-    /* Output statistics on c/f/d points */
-    if (GetVerbLevel() & Statistics1) {
-      // NOTE: We batch the communication here
-      GO l_counts[] = {(GO)num_c_points, (GO) num_f_points, (GO) num_d_points};
-      GO g_counts[3];
-
-      RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 3, l_counts, g_counts);
-      GetOStream(Statistics1) << "ClassicalMapFactory("<<coloringAlgo<<"): C/F/D = "<<g_counts[0]<<"/"<<g_counts[1]<<"/"<<g_counts[2]<<std::endl;
-    }
+  /* Output statistics on c/f/d points */
+  if (GetVerbLevel() & Statistics1) {
+    // NOTE: We batch the communication here
+    GO l_counts[] = {(GO)num_c_points, (GO)num_f_points, (GO)num_d_points};
+    GO g_counts[3];
 
+    RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 3, l_counts, g_counts);
+    GetOStream(Statistics1) << "ClassicalMapFactory(" << coloringAlgo << "): C/F/D = " << g_counts[0] << "/" << g_counts[1] << "/" << g_counts[2] << std::endl;
+  }
 
-    /* Generate the Coarse map */
-    RCP<const Map> coarseMap;
-    {
-      SubFactoryMonitor sfm(*this,"Coarse Map",currentLevel);
-      GenerateCoarseMap(*A->getRowMap(),num_c_points,coarseMap);
-    }
-        
-    Set(currentLevel, "FC Splitting",fc_splitting);
-    Set(currentLevel, "CoarseMap", coarseMap);
-   
+  /* Generate the Coarse map */
+  RCP<const Map> coarseMap;
+  {
+    SubFactoryMonitor sfm(*this, "Coarse Map", currentLevel);
+    GenerateCoarseMap(*A->getRowMap(), num_c_points, coarseMap);
   }
 
+  Set(currentLevel, "FC Splitting", fc_splitting);
+  Set(currentLevel, "CoarseMap", coarseMap);
+}
+
 /* ************************************************************************* */
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GenerateCoarseMap(const Map & fineMap, LO num_c_points, RCP<const Map> & coarseMap) const {
-
-    // FIXME: Assumes scalar PDE
-    std::vector<size_t> stridingInfo_(1);
-    stridingInfo_[0]=1;
-    GO domainGIDOffset = 0;
-
-    coarseMap = StridedMapFactory::Build(fineMap.lib(),
-                                         Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                         num_c_points,
-                                         fineMap.getIndexBase(),
-                                         stridingInfo_,
-                                         fineMap.getComm(),
-                                         domainGIDOffset);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GenerateCoarseMap(const Map& fineMap, LO num_c_points, RCP<const Map>& coarseMap) const {
+  // FIXME: Assumes scalar PDE
+  std::vector<size_t> stridingInfo_(1);
+  stridingInfo_[0]   = 1;
+  GO domainGIDOffset = 0;
+
+  coarseMap = StridedMapFactory::Build(fineMap.lib(),
+                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                       num_c_points,
+                                       fineMap.getIndexBase(),
+                                       stridingInfo_,
+                                       fineMap.getComm(),
+                                       domainGIDOffset);
+}
 
 /* ************************************************************************* */
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DoGraphColoring(const GraphBase & graph, ArrayRCP<LO> & myColors_out, LO & numColors) const {
-    const ParameterList& pL = GetParameterList();
-    using graph_t = typename LWGraph_kokkos::local_graph_type;
-    using KernelHandle = KokkosKernels::Experimental::
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DoGraphColoring(const GraphBase& graph, ArrayRCP<LO>& myColors_out, LO& numColors) const {
+  const ParameterList& pL = GetParameterList();
+  using graph_t           = typename LWGraph_kokkos::local_graph_type;
+  using KernelHandle      = KokkosKernels::Experimental::
       KokkosKernelsHandle<typename graph_t::row_map_type::value_type,
                           typename graph_t::entries_type::value_type,
                           typename graph_t::entries_type::value_type,
                           typename graph_t::device_type::execution_space,
                           typename graph_t::device_type::memory_space,
                           typename graph_t::device_type::memory_space>;
-    KernelHandle kh;
-
-    // Leave gc algorithm choice as the default
-    kh.create_graph_coloring_handle();
-    
-    // Get the distance-1 graph coloring handle
-    auto coloringHandle = kh.get_graph_coloring_handle();
-    
-    // Set the distance-1 coloring algorithm to use
-    if(pL.get<bool>("aggregation: deterministic") == true) {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_SERIAL );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "serial") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_SERIAL );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "vertex based") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_VB );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "vertex based bit array") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_VBBIT );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based bit array" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "vertex based color set") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_VBCS );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based color set" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "vertex based deterministic") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_VBD );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based deterministic" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "vertex based deterministic bit array") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_VBDBIT );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based deterministic bit array" << std::endl;
-    } else if(pL.get<std::string>("aggregation: coloring algorithm") == "edge based") {
-      coloringHandle->set_algorithm( KokkosGraph::COLORING_EB );
-      if(IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: edge based" << std::endl;
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true,std::invalid_argument,"Unrecognized distance 1 coloring algorithm");
-    }
-    
-    // Create device views for graph rowptrs/colinds
-    size_t numRows = graph.GetNodeNumVertices();
-    auto graphLWK = dynamic_cast<const LWGraph_kokkos*>(&graph);
-    auto graphLW  = dynamic_cast<const LWGraph*>(&graph);
-    auto graphG   = dynamic_cast<const Graph*>(&graph);
-    TEUCHOS_TEST_FOR_EXCEPTION(!graphLW && !graphLWK && !graphG,std::invalid_argument,"Graph is not a LWGraph or LWGraph_kokkos object");
-      // Run d1 graph coloring
-      // Assume that the graph is symmetric so row map/entries and col map/entries are the same
-
-    if(graphLWK) {
-      KokkosGraph::Experimental::graph_color(&kh,
-                                             numRows,
-                                             numRows, // FIXME: This should be the number of columns
-                                             graphLWK->getLocalLWGraph().getRowPtrs(),
-                                             graphLWK->getLocalLWGraph().getEntries(),
-                                             true);
-    }
-    else if(graphLW) {
-      auto rowptrs = graphLW->getRowPtrs();
-      auto entries = graphLW->getEntries();
-      // Copy rowptrs to a size_t, because kokkos-kernels doesn't like rowptrs as LO's
-      Teuchos::Array<size_t> rowptrs_s(rowptrs.size());
-      std::copy(rowptrs.begin(),rowptrs.end(),rowptrs_s.begin());
-      Kokkos::View<const size_t*,Kokkos::LayoutLeft,Kokkos::HostSpace> rowptrs_v(rowptrs_s.data(),(size_t)rowptrs.size());
-      Kokkos::View<const LO*,Kokkos::LayoutLeft,Kokkos::HostSpace> entries_v(entries.getRawPtr(),(size_t)entries.size());
-      KokkosGraph::Experimental::graph_color(&kh,
-                                             numRows,
-                                             numRows, // FIXME: This should be the number of columns
-                                             rowptrs_v,
-                                             entries_v,
-                                             true);
-    }
-    else if(graphG) {
-      // FIXME:  This is a terrible, terrible hack, based on 0-based local indexing.
-      RCP<const CrsGraph> graphC = graphG->GetGraph();
-      size_t numEntries = graphC->getLocalNumEntries();
-      ArrayView<const LO> indices;
-      graphC->getLocalRowView(0,indices);
-      Kokkos::View<size_t*,Kokkos::LayoutLeft,Kokkos::HostSpace> rowptrs_v("rowptrs_v",graphC->getLocalNumRows()+1);
-      rowptrs_v[0]=0;
-      for(LO i=0; i<(LO)graphC->getLocalNumRows()+1; i++)
-        rowptrs_v[i+1] = rowptrs_v[i] + graphC->getNumEntriesInLocalRow(i);
-      Kokkos::View<const LO*,Kokkos::LayoutLeft,Kokkos::HostSpace> entries_v(&indices[0],numEntries);
-      KokkosGraph::Experimental::graph_color(&kh,
-                                             numRows,
-                                             numRows, // FIXME: This should be the number of columns
-                                             rowptrs_v,
-                                             entries_v,
-                                             true);
-    }
+  KernelHandle kh;
+
+  // Leave gc algorithm choice as the default
+  kh.create_graph_coloring_handle();
+
+  // Get the distance-1 graph coloring handle
+  auto coloringHandle = kh.get_graph_coloring_handle();
+
+  // Set the distance-1 coloring algorithm to use
+  if (pL.get<bool>("aggregation: deterministic") == true) {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_SERIAL);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "serial") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_SERIAL);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: serial" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "vertex based") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_VB);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "vertex based bit array") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_VBBIT);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based bit array" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "vertex based color set") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_VBCS);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based color set" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "vertex based deterministic") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_VBD);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based deterministic" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "vertex based deterministic bit array") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_VBDBIT);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: vertex based deterministic bit array" << std::endl;
+  } else if (pL.get<std::string>("aggregation: coloring algorithm") == "edge based") {
+    coloringHandle->set_algorithm(KokkosGraph::COLORING_EB);
+    if (IsPrint(Statistics1)) GetOStream(Statistics1) << "  algorithm: edge based" << std::endl;
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "Unrecognized distance 1 coloring algorithm");
+  }
+
+  // Create device views for graph rowptrs/colinds
+  size_t numRows = graph.GetNodeNumVertices();
+  auto graphLWK  = dynamic_cast<const LWGraph_kokkos*>(&graph);
+  auto graphLW   = dynamic_cast<const LWGraph*>(&graph);
+  auto graphG    = dynamic_cast<const Graph*>(&graph);
+  TEUCHOS_TEST_FOR_EXCEPTION(!graphLW && !graphLWK && !graphG, std::invalid_argument, "Graph is not a LWGraph or LWGraph_kokkos object");
+  // Run d1 graph coloring
+  // Assume that the graph is symmetric so row map/entries and col map/entries are the same
+
+  if (graphLWK) {
+    KokkosGraph::Experimental::graph_color(&kh,
+                                           numRows,
+                                           numRows,  // FIXME: This should be the number of columns
+                                           graphLWK->getLocalLWGraph().getRowPtrs(),
+                                           graphLWK->getLocalLWGraph().getEntries(),
+                                           true);
+  } else if (graphLW) {
+    auto rowptrs = graphLW->getRowPtrs();
+    auto entries = graphLW->getEntries();
+    // Copy rowptrs to a size_t, because kokkos-kernels doesn't like rowptrs as LO's
+    Teuchos::Array<size_t> rowptrs_s(rowptrs.size());
+    std::copy(rowptrs.begin(), rowptrs.end(), rowptrs_s.begin());
+    Kokkos::View<const size_t*, Kokkos::LayoutLeft, Kokkos::HostSpace> rowptrs_v(rowptrs_s.data(), (size_t)rowptrs.size());
+    Kokkos::View<const LO*, Kokkos::LayoutLeft, Kokkos::HostSpace> entries_v(entries.getRawPtr(), (size_t)entries.size());
+    KokkosGraph::Experimental::graph_color(&kh,
+                                           numRows,
+                                           numRows,  // FIXME: This should be the number of columns
+                                           rowptrs_v,
+                                           entries_v,
+                                           true);
+  } else if (graphG) {
+    // FIXME:  This is a terrible, terrible hack, based on 0-based local indexing.
+    RCP<const CrsGraph> graphC = graphG->GetGraph();
+    size_t numEntries          = graphC->getLocalNumEntries();
+    ArrayView<const LO> indices;
+    graphC->getLocalRowView(0, indices);
+    Kokkos::View<size_t*, Kokkos::LayoutLeft, Kokkos::HostSpace> rowptrs_v("rowptrs_v", graphC->getLocalNumRows() + 1);
+    rowptrs_v[0] = 0;
+    for (LO i = 0; i < (LO)graphC->getLocalNumRows() + 1; i++)
+      rowptrs_v[i + 1] = rowptrs_v[i] + graphC->getNumEntriesInLocalRow(i);
+    Kokkos::View<const LO*, Kokkos::LayoutLeft, Kokkos::HostSpace> entries_v(&indices[0], numEntries);
+    KokkosGraph::Experimental::graph_color(&kh,
+                                           numRows,
+                                           numRows,  // FIXME: This should be the number of columns
+                                           rowptrs_v,
+                                           entries_v,
+                                           true);
+  }
+
+  // Extract the colors and store them in the aggregates
+  auto myColors_d = coloringHandle->get_vertex_colors();
+  numColors       = static_cast<LO>(coloringHandle->get_num_colors());
 
-    
-    // Extract the colors and store them in the aggregates
-    auto myColors_d = coloringHandle->get_vertex_colors();
-    numColors = static_cast<LO>(coloringHandle->get_num_colors());
+  // Copy back to host
+  auto myColors_h = Kokkos::create_mirror_view(myColors_d);
+  myColors_out.resize(myColors_h.size());
+  Kokkos::View<LO*, Kokkos::LayoutLeft, Kokkos::HostSpace> myColors_v(&myColors_out[0], myColors_h.size());
+  Kokkos::deep_copy(myColors_v, myColors_h);
 
-    // Copy back to host
-    auto myColors_h = Kokkos::create_mirror_view(myColors_d);
-    myColors_out.resize(myColors_h.size());
-    Kokkos::View<LO*,Kokkos::LayoutLeft,Kokkos::HostSpace> myColors_v(&myColors_out[0],myColors_h.size());
-    Kokkos::deep_copy(myColors_v,myColors_h);
-    
-    //clean up coloring handle
-    kh.destroy_graph_coloring_handle();
+  // clean up coloring handle
+  kh.destroy_graph_coloring_handle();
 
-  }// end DoGraphColoring
-    
+}  // end DoGraphColoring
 
 /* ************************************************************************* */
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DoMISNaive(const GraphBase & graph, ArrayRCP<LO> & myColors, LO & numColors) const {
-    // This is a fall-back routine for when we don't have Kokkos or when it isn't initialized
-    // We just do greedy MIS because this is easy to write.
-
-    LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-    LO MIS = Teuchos::ScalarTraits<LO>::one();
-
-    //FIXME: Not efficient
-    myColors.resize(0);
-    myColors.resize(graph.GetNodeNumVertices(),LO_INVALID);
-    auto boundaryNodes = graph.GetBoundaryNodeMap();
-    LO Nrows = (LO)graph.GetNodeNumVertices();
-
-    
-    for(LO row=0; row < Nrows; row++) {
-      if(boundaryNodes[row])
-        continue;
-      ArrayView<const LO> indices = graph.getNeighborVertices(row);
-      bool has_colored_neighbor=false;
-      for(LO j=0; !has_colored_neighbor && j<(LO)indices.size(); j++) {
-        // FIXME: This does not handle ghosting correctly
-        if(myColors[indices[j]] == MIS)
-          has_colored_neighbor=true;
-      }
-      if(!has_colored_neighbor)
-        myColors[row] = MIS;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DoMISNaive(const GraphBase& graph, ArrayRCP<LO>& myColors, LO& numColors) const {
+  // This is a fall-back routine for when we don't have Kokkos or when it isn't initialized
+  // We just do greedy MIS because this is easy to write.
+
+  LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+  LO MIS        = Teuchos::ScalarTraits<LO>::one();
+
+  // FIXME: Not efficient
+  myColors.resize(0);
+  myColors.resize(graph.GetNodeNumVertices(), LO_INVALID);
+  auto boundaryNodes = graph.GetBoundaryNodeMap();
+  LO Nrows           = (LO)graph.GetNodeNumVertices();
+
+  for (LO row = 0; row < Nrows; row++) {
+    if (boundaryNodes[row])
+      continue;
+    ArrayView<const LO> indices = graph.getNeighborVertices(row);
+    bool has_colored_neighbor   = false;
+    for (LO j = 0; !has_colored_neighbor && j < (LO)indices.size(); j++) {
+      // FIXME: This does not handle ghosting correctly
+      if (myColors[indices[j]] == MIS)
+        has_colored_neighbor = true;
     }
-    numColors=1;
+    if (!has_colored_neighbor)
+      myColors[row] = MIS;
   }
+  numColors = 1;
+}
 
 /* ************************************************************************* */
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DoDistributedGraphColoring(RCP<const GraphBase> & graph, ArrayRCP<LO> & myColors_out, LO & numColors) const {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DoDistributedGraphColoring(RCP<const GraphBase>& graph, ArrayRCP<LO>& myColors_out, LO& numColors) const {
 #ifdef HAVE_MUELU_ZOLTAN2
-    //  const ParameterList& pL = GetParameterList();
-    Teuchos::ParameterList params;
-    params.set("color_choice","FirstFit");
-    params.set("color_method","D1");
-    //  params.set("color_choice", colorMethod);
-    //  params.set("color_method", colorAlg);
-    //  params.set("verbose", verbose);
-    //  params.set("serial_threshold",serialThreshold);
-    //params.set("recolor_degrees",recolorDegrees);
-
-    // Do the coloring via Zoltan2
-    using GraphAdapter = MueLuGraphBaseAdapter<GraphBase>;
-    GraphAdapter z_adapter(graph);
-
-    // We need to provide the MPI Comm, or else we wind up using the default (eep!)
-    Zoltan2::ColoringProblem<GraphAdapter> problem(&z_adapter,&params,graph->GetDomainMap()->getComm());
-    problem.solve();
-    Zoltan2::ColoringSolution<GraphAdapter> * soln = problem.getSolution();
-    ArrayRCP<int> colors = soln->getColorsRCP();
-    numColors = (LO)soln->getNumColors();
-
-    // Assign the Array RCP or Copy Out
-    // FIXME:  This probably won't work if LO!=int
-    if(std::is_same<LO,int>::value)
-      myColors_out = colors;
-    else {
-      myColors_out.resize(colors.size());
-      for(LO i=0; i<(LO)myColors_out.size(); i++)
-        myColors_out[i] = (LO) colors[i];
-    }
+  //  const ParameterList& pL = GetParameterList();
+  Teuchos::ParameterList params;
+  params.set("color_choice", "FirstFit");
+  params.set("color_method", "D1");
+  //  params.set("color_choice", colorMethod);
+  //  params.set("color_method", colorAlg);
+  //  params.set("verbose", verbose);
+  //  params.set("serial_threshold",serialThreshold);
+  // params.set("recolor_degrees",recolorDegrees);
+
+  // Do the coloring via Zoltan2
+  using GraphAdapter = MueLuGraphBaseAdapter<GraphBase>;
+  GraphAdapter z_adapter(graph);
+
+  // We need to provide the MPI Comm, or else we wind up using the default (eep!)
+  Zoltan2::ColoringProblem<GraphAdapter> problem(&z_adapter, &params, graph->GetDomainMap()->getComm());
+  problem.solve();
+  Zoltan2::ColoringSolution<GraphAdapter>* soln = problem.getSolution();
+  ArrayRCP<int> colors                          = soln->getColorsRCP();
+  numColors                                     = (LO)soln->getNumColors();
+
+  // Assign the Array RCP or Copy Out
+  // FIXME:  This probably won't work if LO!=int
+  if (std::is_same<LO, int>::value)
+    myColors_out = colors;
+  else {
+    myColors_out.resize(colors.size());
+    for (LO i = 0; i < (LO)myColors_out.size(); i++)
+      myColors_out[i] = (LO)colors[i];
+  }
 
-    /*
+  /*
 
-    printf("CMS: numColors = %d\ncolors = ",numColors);
-    for(int i=0;i<colors.size(); i++)
-      printf("%d ",colors[i]);
-    printf("\n");
+  printf("CMS: numColors = %d\ncolors = ",numColors);
+  for(int i=0;i<colors.size(); i++)
+    printf("%d ",colors[i]);
+  printf("\n");
 
-    */
-#endif //ifdef HAVE_MUELU_ZOLTAN2
-  }
+  */
+#endif  // ifdef HAVE_MUELU_ZOLTAN2
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_CLASSICALMAPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_decl.hpp b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_decl.hpp
index 3f9a44e1cf7d..11dc50bd1a02 100644
--- a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_decl.hpp
@@ -65,61 +65,60 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ClassicalPFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ClassicalPFactory : public PFactory {
 #undef MUELU_CLASSICALPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    // Defining types that require the short names included above
-    using point_type = typename ClassicalMapFactory::point_type;
+ public:
+  // Defining types that require the short names included above
+  using point_type = typename ClassicalMapFactory::point_type;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    ClassicalPFactory() { }
+  //! Constructor
+  ClassicalPFactory() {}
 
-    //! Destructor.
-    virtual ~ClassicalPFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~ClassicalPFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
-    
-  private:
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    // Utility algorithms
-    void GenerateStrengthFlags(const Matrix & A,const GraphBase & graph, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong) const;
+ private:
+  // Utility algorithms
+  void GenerateStrengthFlags(const Matrix& A, const GraphBase& graph, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong) const;
 
-    // Ghosting Algorithms
-    void GhostCoarseMap(const Matrix &A,const Import & Importer, const ArrayRCP<const LO> myPointType,const RCP<const Map> & coarseMap, RCP<const Map> & coarseColMap) const;
+  // Ghosting Algorithms
+  void GhostCoarseMap(const Matrix& A, const Import& Importer, const ArrayRCP<const LO> myPointType, const RCP<const Map>& coarseMap, RCP<const Map>& coarseColMap) const;
 
-    // Coarsening algorithms
-    void Coarsen_ClassicalModified(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph, RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<const Import> remoteOnlyImporter, RCP<Matrix> & P) const;
-    void Coarsen_Direct(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph, RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<Matrix> & P) const;
-    void Coarsen_Ext_Plus_I(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph, RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<Matrix> & P) const;
+  // Coarsening algorithms
+  void Coarsen_ClassicalModified(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<const Import> remoteOnlyImporter, RCP<Matrix>& P) const;
+  void Coarsen_Direct(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<Matrix>& P) const;
+  void Coarsen_Ext_Plus_I(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<Matrix>& P) const;
 
-    //@}
+  //@}
 
-  }; //class ClassicalPFactory
+};  // class ClassicalPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CLASSICALPFACTORY_SHORT
-#endif // MUELU_CLASSICALPFACTORY_DECL_HPP
+#endif  // MUELU_CLASSICALPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_def.hpp b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_def.hpp
index d04f76faaa7c..4af210f3352c 100644
--- a/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Classical/MueLu_ClassicalPFactory_def.hpp
@@ -68,1090 +68,1057 @@
 #include "MueLu_AmalgamationInfo.hpp"
 #include "MueLu_GraphBase.hpp"
 
-
 //#define CMS_DEBUG
 //#define CMS_DUMP
 
 namespace {
 
-template<class SC>
+template <class SC>
 int Sign(SC val) {
-  using STS = typename Teuchos::ScalarTraits<SC>;
+  using STS                           = typename Teuchos::ScalarTraits<SC>;
   typename STS::magnitudeType MT_ZERO = Teuchos::ScalarTraits<typename STS::magnitudeType>::zero();
-  if(STS::real(val) > MT_ZERO) return 1;
-  else if(STS::real(val) < MT_ZERO) return -1;
-  else return 0;
+  if (STS::real(val) > MT_ZERO)
+    return 1;
+  else if (STS::real(val) < MT_ZERO)
+    return -1;
+  else
+    return 0;
 }
 
-}// anonymous namepsace
+}  // namespace
 
 namespace MueLu {
 
-
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("aggregation: deterministic");
-    SET_VALID_ENTRY("aggregation: coloring algorithm");
-    SET_VALID_ENTRY("aggregation: classical scheme");
-
-    // To know if we need BlockNumber
-    SET_VALID_ENTRY("aggregation: drop scheme");
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      validParamList->getEntry("aggregation: classical scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("direct","ext+i","classical modified"), "aggregation: classical scheme")));
-                                                                        
-    }
+  SET_VALID_ENTRY("aggregation: deterministic");
+  SET_VALID_ENTRY("aggregation: coloring algorithm");
+  SET_VALID_ENTRY("aggregation: classical scheme");
+
+  // To know if we need BlockNumber
+  SET_VALID_ENTRY("aggregation: drop scheme");
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    validParamList->getEntry("aggregation: classical scheme").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("direct", "ext+i", "classical modified"), "aggregation: classical scheme")));
+  }
 
 #undef SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A");
-    //    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("Graph",       null, "Generating factory of the graph");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap",         Teuchos::null, "Generating factory of the CoarseMap");
-    validParamList->set< RCP<const FactoryBase> >("FC Splitting",         Teuchos::null, "Generating factory of the FC Splitting");
-    validParamList->set< RCP<const FactoryBase> >("BlockNumber",        Teuchos::null, "Generating factory for Block Number");
-    //    validParamList->set< RCP<const FactoryBase> >("Nullspace",      Teuchos::null, "Generating factory of the nullspace");
-
-    return validParamList;
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  //    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase> >("Graph", null, "Generating factory of the graph");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", null, "Generating factory for variable \'DofsPerNode\', usually the same as for \'Graph\'");
+  validParamList->set<RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of the CoarseMap");
+  validParamList->set<RCP<const FactoryBase> >("FC Splitting", Teuchos::null, "Generating factory of the FC Splitting");
+  validParamList->set<RCP<const FactoryBase> >("BlockNumber", Teuchos::null, "Generating factory for Block Number");
+  //    validParamList->set< RCP<const FactoryBase> >("Nullspace",      Teuchos::null, "Generating factory of the nullspace");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "Graph");
+  Input(fineLevel, "DofsPerNode");
+  //    Input(fineLevel, "UnAmalgamationInfo");
+  Input(fineLevel, "CoarseMap");
+  Input(fineLevel, "FC Splitting");
+
+  const ParameterList& pL = GetParameterList();
+  std::string drop_algo   = pL.get<std::string>("aggregation: drop scheme");
+  if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical") {
+    Input(fineLevel, "BlockNumber");
   }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "Graph");
-    Input(fineLevel, "DofsPerNode");
-    //    Input(fineLevel, "UnAmalgamationInfo");
-    Input(fineLevel, "CoarseMap");
-    Input(fineLevel, "FC Splitting");
-    
-    const ParameterList& pL = GetParameterList();
-    std::string drop_algo = pL.get<std::string>("aggregation: drop scheme");
-    if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical") {
-      Input(fineLevel, "BlockNumber");
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+  // using STS = Teuchos::ScalarTraits<SC>;
+
+  // We start by assuming that someone did a reasonable strength of connection
+  // algorithm before we start to get our Graph, DofsPerNode and UnAmalgamationInfo
+
+  // We begin by getting a MIS (from a graph coloring) and then at that point we need
+  // to start generating entries for the prolongator.
+  RCP<const Matrix> A                              = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<const Map> ownedCoarseMap                    = Get<RCP<const Map> >(fineLevel, "CoarseMap");
+  RCP<const LocalOrdinalVector> owned_fc_splitting = Get<RCP<LocalOrdinalVector> >(fineLevel, "FC Splitting");
+  RCP<const GraphBase> graph                       = Get<RCP<GraphBase> >(fineLevel, "Graph");
+  //    LO nDofsPerNode                 = Get<LO>(fineLevel, "DofsPerNode");
+  //    RCP<AmalgamationInfo> amalgInfo = Get< RCP<AmalgamationInfo> >     (fineLevel, "UnAmalgamationInfo");
+  RCP<const Import> Importer = A->getCrsGraph()->getImporter();
+  Xpetra::UnderlyingLib lib  = ownedCoarseMap->lib();
+
+  //    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
+  RCP<Matrix> P;
+  //    SC SC_ZERO = STS::zero();
+  LO LO_INVALID           = Teuchos::OrdinalTraits<LO>::invalid();
+  const point_type C_PT   = ClassicalMapFactory::C_PT;
+  const point_type F_PT   = ClassicalMapFactory::F_PT;
+  const ParameterList& pL = GetParameterList();
+
+  // FIXME: This guy doesn't work right now for NumPDEs != 1
+  TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() != 1, Exceptions::RuntimeError, "ClassicalPFactory: Multiple PDEs per node not supported yet");
+
+  // NOTE: Let's hope we never need to deal with this case
+  TEUCHOS_TEST_FOR_EXCEPTION(!A->getRowMap()->isSameAs(*A->getDomainMap()), Exceptions::RuntimeError, "ClassicalPFactory: MPI Ranks > 1 not supported yet");
+
+  // Do we need ghosts rows of A and myPointType?
+  std::string scheme   = pL.get<std::string>("aggregation: classical scheme");
+  bool need_ghost_rows = false;
+  if (scheme == "ext+i")
+    need_ghost_rows = true;
+  else if (scheme == "direct")
+    need_ghost_rows = false;
+  else if (scheme == "classical modified")
+    need_ghost_rows = true;
+  // NOTE: ParameterList validator will check this guy so we don't really need an "else" here
+
+  if (GetVerbLevel() & Statistics1) {
+    GetOStream(Statistics1) << "ClassicalPFactory: scheme = " << scheme << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  // Ghost the FC splitting and grab the data (if needed)
+  RCP<const LocalOrdinalVector> fc_splitting;
+  ArrayRCP<const LO> myPointType;
+  if (Importer.is_null()) {
+    fc_splitting = owned_fc_splitting;
+  } else {
+    RCP<LocalOrdinalVector> fc_splitting_nonconst = LocalOrdinalVectorFactory::Build(A->getCrsGraph()->getColMap());
+    fc_splitting_nonconst->doImport(*owned_fc_splitting, *Importer, Xpetra::INSERT);
+    fc_splitting = fc_splitting_nonconst;
+  }
+  myPointType = fc_splitting->getData(0);
+
+  /* Ghost A (if needed) */
+  RCP<const Matrix> Aghost;
+  RCP<const LocalOrdinalVector> fc_splitting_ghost;
+  ArrayRCP<const LO> myPointType_ghost;
+  RCP<const Import> remoteOnlyImporter;
+  if (need_ghost_rows && !Importer.is_null()) {
+    ArrayView<const LO> remoteLIDs = Importer->getRemoteLIDs();
+    size_t numRemote               = Importer->getNumRemoteIDs();
+    Array<GO> remoteRows(numRemote);
+    for (size_t i = 0; i < numRemote; i++)
+      remoteRows[i] = Importer->getTargetMap()->getGlobalElement(remoteLIDs[i]);
+
+    RCP<const Map> remoteRowMap = MapFactory::Build(lib, Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), remoteRows(),
+                                                    A->getDomainMap()->getIndexBase(), A->getDomainMap()->getComm());
+
+    remoteOnlyImporter        = Importer->createRemoteOnlyImport(remoteRowMap);
+    RCP<const CrsMatrix> Acrs = rcp_dynamic_cast<const CrsMatrixWrap>(A)->getCrsMatrix();
+    RCP<CrsMatrix> Aghost_crs = CrsMatrixFactory::Build(Acrs, *remoteOnlyImporter, A->getDomainMap(), remoteOnlyImporter->getTargetMap());
+    Aghost                    = rcp(new CrsMatrixWrap(Aghost_crs));
+    // CAG: It seems that this isn't actually needed?
+    // We also may need need to ghost myPointType for Aghost
+    // RCP<const Import> Importer2 = Aghost->getCrsGraph()->getImporter();
+    // if(Importer2.is_null()) {
+    //   RCP<LocalOrdinalVector> fc_splitting_ghost_nonconst = LocalOrdinalVectorFactory::Build(Aghost->getColMap());
+    //   fc_splitting_ghost_nonconst->doImport(*owned_fc_splitting,*Importer,Xpetra::INSERT);
+    //   fc_splitting_ghost = fc_splitting_ghost_nonconst;
+    //   myPointType_ghost  = fc_splitting_ghost->getData(0);
+    // }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-    // using STS = Teuchos::ScalarTraits<SC>;
-
-    // We start by assuming that someone did a reasonable strength of connection
-    // algorithm before we start to get our Graph, DofsPerNode and UnAmalgamationInfo
-
-    // We begin by getting a MIS (from a graph coloring) and then at that point we need
-    // to start generating entries for the prolongator.
-    RCP<const Matrix>      A        = Get< RCP<Matrix> >(fineLevel, "A");
-    RCP<const Map> ownedCoarseMap   = Get<RCP<const Map> >(fineLevel,"CoarseMap");
-    RCP<const LocalOrdinalVector> owned_fc_splitting = Get<RCP<LocalOrdinalVector> >(fineLevel,"FC Splitting");
-    RCP<const GraphBase> graph      = Get< RCP<GraphBase> >(fineLevel, "Graph");
-    //    LO nDofsPerNode                 = Get<LO>(fineLevel, "DofsPerNode");
-    //    RCP<AmalgamationInfo> amalgInfo = Get< RCP<AmalgamationInfo> >     (fineLevel, "UnAmalgamationInfo");
-    RCP<const Import>    Importer   = A->getCrsGraph()->getImporter();
-    Xpetra::UnderlyingLib lib = ownedCoarseMap->lib();
-
-    //    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-    RCP<Matrix> P;
-    //    SC SC_ZERO = STS::zero();
-    LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-    const point_type C_PT = ClassicalMapFactory::C_PT;
-    const point_type F_PT = ClassicalMapFactory::F_PT;
-    const ParameterList& pL = GetParameterList();
-  
-    // FIXME: This guy doesn't work right now for NumPDEs != 1
-    TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() != 1, Exceptions::RuntimeError,"ClassicalPFactory: Multiple PDEs per node not supported yet");
- 
-    // NOTE: Let's hope we never need to deal with this case
-    TEUCHOS_TEST_FOR_EXCEPTION(!A->getRowMap()->isSameAs(*A->getDomainMap()),Exceptions::RuntimeError,"ClassicalPFactory: MPI Ranks > 1 not supported yet");
-
-    // Do we need ghosts rows of A and myPointType?
-    std::string scheme = pL.get<std::string>("aggregation: classical scheme");
-    bool need_ghost_rows =false;
-    if(scheme == "ext+i")
-      need_ghost_rows=true;
-    else if(scheme == "direct")
-      need_ghost_rows=false;
-    else if(scheme == "classical modified")
-      need_ghost_rows=true;
-    // NOTE: ParameterList validator will check this guy so we don't really need an "else" here
-
-    if (GetVerbLevel() & Statistics1) {
-      GetOStream(Statistics1) << "ClassicalPFactory: scheme = "<<scheme<<std::endl;
-    }
-
-    // Ghost the FC splitting and grab the data (if needed)
-    RCP<const LocalOrdinalVector> fc_splitting;
-    ArrayRCP<const LO> myPointType;
-    if(Importer.is_null()) {
-      fc_splitting = owned_fc_splitting;
-    }
-    else {
-      RCP<LocalOrdinalVector> fc_splitting_nonconst = LocalOrdinalVectorFactory::Build(A->getCrsGraph()->getColMap());
-      fc_splitting_nonconst->doImport(*owned_fc_splitting,*Importer,Xpetra::INSERT);
-      fc_splitting = fc_splitting_nonconst;
-    }
-    myPointType = fc_splitting->getData(0);
-
-
-    /* Ghost A (if needed) */
-    RCP<const Matrix> Aghost;
-    RCP<const LocalOrdinalVector> fc_splitting_ghost;
-    ArrayRCP<const LO> myPointType_ghost;
-    RCP<const Import> remoteOnlyImporter;
-    if(need_ghost_rows && !Importer.is_null()){
-      ArrayView<const LO> remoteLIDs = Importer->getRemoteLIDs();
-      size_t numRemote = Importer->getNumRemoteIDs();
-      Array<GO> remoteRows(numRemote);
-      for (size_t i = 0; i < numRemote; i++)
-        remoteRows[i] = Importer->getTargetMap()->getGlobalElement(remoteLIDs[i]);
-
-      RCP<const Map> remoteRowMap = MapFactory::Build(lib,Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), remoteRows(),
-                                                      A->getDomainMap()->getIndexBase(), A->getDomainMap()->getComm());
-
-      remoteOnlyImporter = Importer->createRemoteOnlyImport(remoteRowMap);
-      RCP<const CrsMatrix> Acrs = rcp_dynamic_cast<const CrsMatrixWrap>(A)->getCrsMatrix();
-      RCP<CrsMatrix> Aghost_crs = CrsMatrixFactory::Build(Acrs,*remoteOnlyImporter,A->getDomainMap(),remoteOnlyImporter->getTargetMap());
-      Aghost = rcp(new CrsMatrixWrap(Aghost_crs));
-      // CAG: It seems that this isn't actually needed?
-      // We also may need need to ghost myPointType for Aghost
-      // RCP<const Import> Importer2 = Aghost->getCrsGraph()->getImporter();
-      // if(Importer2.is_null()) {
-      //   RCP<LocalOrdinalVector> fc_splitting_ghost_nonconst = LocalOrdinalVectorFactory::Build(Aghost->getColMap());
-      //   fc_splitting_ghost_nonconst->doImport(*owned_fc_splitting,*Importer,Xpetra::INSERT);
-      //   fc_splitting_ghost = fc_splitting_ghost_nonconst;
-      //   myPointType_ghost  = fc_splitting_ghost->getData(0);
-      // }
-    }
+  /* Generate the ghosted Coarse map using the "Tuminaro maneuver" (if needed)*/
+  RCP<const Map> coarseMap;
+  if (Importer.is_null())
+    coarseMap = ownedCoarseMap;
+  else {
+    // Generate a domain vector with the coarse ID's as entries for C points
+    GhostCoarseMap(*A, *Importer, myPointType, ownedCoarseMap, coarseMap);
+  }
 
-    /* Generate the ghosted Coarse map using the "Tuminaro maneuver" (if needed)*/
-    RCP<const Map> coarseMap;
-    if(Importer.is_null())
-      coarseMap = ownedCoarseMap;
+  /* Get the block number, if we need it (and ghost it) */
+  RCP<LocalOrdinalVector> BlockNumber;
+  std::string drop_algo = pL.get<std::string>("aggregation: drop scheme");
+  if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical") {
+    RCP<LocalOrdinalVector> OwnedBlockNumber;
+    OwnedBlockNumber = Get<RCP<LocalOrdinalVector> >(fineLevel, "BlockNumber");
+    if (Importer.is_null())
+      BlockNumber = OwnedBlockNumber;
     else {
-      // Generate a domain vector with the coarse ID's as entries for C points
-      GhostCoarseMap(*A,*Importer,myPointType,ownedCoarseMap,coarseMap);
-    }
-
-    /* Get the block number, if we need it (and ghost it) */
-    RCP<LocalOrdinalVector>  BlockNumber;
-    std::string drop_algo = pL.get<std::string>("aggregation: drop scheme");
-    if (drop_algo.find("block diagonal") != std::string::npos || drop_algo == "signed classical") {
-      RCP<LocalOrdinalVector> OwnedBlockNumber;
-      OwnedBlockNumber = Get<RCP<LocalOrdinalVector> >(fineLevel, "BlockNumber");
-      if(Importer.is_null())
-        BlockNumber = OwnedBlockNumber;
-      else{
-        BlockNumber = LocalOrdinalVectorFactory::Build(A->getColMap());
-        BlockNumber->doImport(*OwnedBlockNumber,*Importer,Xpetra::INSERT);
-      }
+      BlockNumber = LocalOrdinalVectorFactory::Build(A->getColMap());
+      BlockNumber->doImport(*OwnedBlockNumber, *Importer, Xpetra::INSERT);
     }
+  }
 
 #if defined(CMS_DEBUG) || defined(CMS_DUMP)
-    {
-      RCP<const CrsMatrix> Acrs = rcp_dynamic_cast<const CrsMatrixWrap>(A)->getCrsMatrix();
-      int rank = A->getRowMap()->getComm()->getRank();
-      printf("[%d] A local size = %dx%d\n",rank,(int)Acrs->getRowMap()->getLocalNumElements(),(int)Acrs->getColMap()->getLocalNumElements());
-
-      printf("[%d] graph local size = %dx%d\n",rank,(int)graph->GetDomainMap()->getLocalNumElements(),(int)graph->GetImportMap()->getLocalNumElements());
-
-      std::ofstream ofs(std::string("dropped_graph_") + std::to_string(fineLevel.GetLevelID())+std::string("_") + std::to_string(rank) + std::string(".dat"),std::ofstream::out);
-      RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
-      graph->print(*fancy,Debug);
-      std::string out_fc = std::string("fc_splitting_") + std::to_string(fineLevel.GetLevelID()) +std::string("_") + std::to_string(rank)+ std::string(".dat");
-      std::string out_block = std::string("block_number_") + std::to_string(fineLevel.GetLevelID()) +std::string("_") + std::to_string(rank)+ std::string(".dat");
-
-      // We don't support writing LO vectors in Xpetra (boo!) so....
-      using real_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-      using RealValuedMultiVector = typename Xpetra::MultiVector<real_type,LO,GO,NO>;
-      typedef Xpetra::MultiVectorFactory<real_type,LO,GO,NO> RealValuedMultiVectorFactory;
- 
-      RCP<RealValuedMultiVector> mv = RealValuedMultiVectorFactory::Build(fc_splitting->getMap(),1);
-      ArrayRCP<real_type> mv_data= mv->getDataNonConst(0);
-
-      // FC Splitting
-      ArrayRCP<const LO> fc_data= fc_splitting->getData(0);
-      for(LO i=0; i<(LO)fc_data.size(); i++)
-        mv_data[i] = Teuchos::as<real_type>(fc_data[i]);
-      Xpetra::IO<real_type,LO,GO,NO>::Write(out_fc,*mv);
-
-      // Block Number
-      if(!BlockNumber.is_null()) {
-        RCP<RealValuedMultiVector> mv2 = RealValuedMultiVectorFactory::Build(BlockNumber->getMap(),1);
-        ArrayRCP<real_type> mv_data2= mv2->getDataNonConst(0);
-        ArrayRCP<const LO> b_data= BlockNumber->getData(0);
-        for(LO i=0; i<(LO)b_data.size(); i++) {
-          mv_data2[i] = Teuchos::as<real_type>(b_data[i]);
-        }
-        Xpetra::IO<real_type,LO,GO,NO>::Write(out_block,*mv2);
+  {
+    RCP<const CrsMatrix> Acrs = rcp_dynamic_cast<const CrsMatrixWrap>(A)->getCrsMatrix();
+    int rank                  = A->getRowMap()->getComm()->getRank();
+    printf("[%d] A local size = %dx%d\n", rank, (int)Acrs->getRowMap()->getLocalNumElements(), (int)Acrs->getColMap()->getLocalNumElements());
+
+    printf("[%d] graph local size = %dx%d\n", rank, (int)graph->GetDomainMap()->getLocalNumElements(), (int)graph->GetImportMap()->getLocalNumElements());
+
+    std::ofstream ofs(std::string("dropped_graph_") + std::to_string(fineLevel.GetLevelID()) + std::string("_") + std::to_string(rank) + std::string(".dat"), std::ofstream::out);
+    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(ofs));
+    graph->print(*fancy, Debug);
+    std::string out_fc    = std::string("fc_splitting_") + std::to_string(fineLevel.GetLevelID()) + std::string("_") + std::to_string(rank) + std::string(".dat");
+    std::string out_block = std::string("block_number_") + std::to_string(fineLevel.GetLevelID()) + std::string("_") + std::to_string(rank) + std::string(".dat");
+
+    // We don't support writing LO vectors in Xpetra (boo!) so....
+    using real_type             = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+    using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+    typedef Xpetra::MultiVectorFactory<real_type, LO, GO, NO> RealValuedMultiVectorFactory;
+
+    RCP<RealValuedMultiVector> mv = RealValuedMultiVectorFactory::Build(fc_splitting->getMap(), 1);
+    ArrayRCP<real_type> mv_data   = mv->getDataNonConst(0);
+
+    // FC Splitting
+    ArrayRCP<const LO> fc_data = fc_splitting->getData(0);
+    for (LO i = 0; i < (LO)fc_data.size(); i++)
+      mv_data[i] = Teuchos::as<real_type>(fc_data[i]);
+    Xpetra::IO<real_type, LO, GO, NO>::Write(out_fc, *mv);
+
+    // Block Number
+    if (!BlockNumber.is_null()) {
+      RCP<RealValuedMultiVector> mv2 = RealValuedMultiVectorFactory::Build(BlockNumber->getMap(), 1);
+      ArrayRCP<real_type> mv_data2   = mv2->getDataNonConst(0);
+      ArrayRCP<const LO> b_data      = BlockNumber->getData(0);
+      for (LO i = 0; i < (LO)b_data.size(); i++) {
+        mv_data2[i] = Teuchos::as<real_type>(b_data[i]);
       }
+      Xpetra::IO<real_type, LO, GO, NO>::Write(out_block, *mv2);
     }
-
+  }
 
 #endif
 
-  
-    /* Generate reindexing arrays */
-    // Note: cpoint2pcol is ghosted if myPointType is
-    // NOTE: Since the ghosted coarse column map follows the ordering of
-    // the fine column map, this *should* work, because it is in local indices.
-    // pcol2cpoint - Takes a LCID for P and turns in into an LCID for A.
-    // cpoint2pcol - Takes a LCID for A --- if it is a C Point --- and turns in into an LCID for P.
-    Array<LO> cpoint2pcol(myPointType.size(),LO_INVALID);
-    Array<LO> pcol2cpoint(coarseMap->getLocalNumElements(),LO_INVALID);
-    LO num_c_points = 0;
-    LO num_f_points =0;
-    for(LO i=0; i<(LO) myPointType.size(); i++) {
-        if(myPointType[i] == C_PT) {
-        cpoint2pcol[i] = num_c_points;
-        num_c_points++;
-      }
-      else if (myPointType[i] == F_PT)
-        num_f_points++;
-    }
-    for(LO i=0; i<(LO)cpoint2pcol.size(); i++) {
-      if(cpoint2pcol[i] != LO_INVALID)
-        pcol2cpoint[cpoint2pcol[i]] =i;
-    }
+  /* Generate reindexing arrays */
+  // Note: cpoint2pcol is ghosted if myPointType is
+  // NOTE: Since the ghosted coarse column map follows the ordering of
+  // the fine column map, this *should* work, because it is in local indices.
+  // pcol2cpoint - Takes a LCID for P and turns in into an LCID for A.
+  // cpoint2pcol - Takes a LCID for A --- if it is a C Point --- and turns in into an LCID for P.
+  Array<LO> cpoint2pcol(myPointType.size(), LO_INVALID);
+  Array<LO> pcol2cpoint(coarseMap->getLocalNumElements(), LO_INVALID);
+  LO num_c_points = 0;
+  LO num_f_points = 0;
+  for (LO i = 0; i < (LO)myPointType.size(); i++) {
+    if (myPointType[i] == C_PT) {
+      cpoint2pcol[i] = num_c_points;
+      num_c_points++;
+    } else if (myPointType[i] == F_PT)
+      num_f_points++;
+  }
+  for (LO i = 0; i < (LO)cpoint2pcol.size(); i++) {
+    if (cpoint2pcol[i] != LO_INVALID)
+      pcol2cpoint[cpoint2pcol[i]] = i;
+  }
 
-    // Generate edge strength flags (this will make everything easier later)
-    // These do *not* need to be ghosted (unlike A)
-    Teuchos::Array<size_t> eis_rowptr;
-    Teuchos::Array<bool> edgeIsStrong;
-    {
-      SubFactoryMonitor sfm(*this,"Strength Flags",coarseLevel);
-      GenerateStrengthFlags(*A,*graph,eis_rowptr,edgeIsStrong);
-    }
+  // Generate edge strength flags (this will make everything easier later)
+  // These do *not* need to be ghosted (unlike A)
+  Teuchos::Array<size_t> eis_rowptr;
+  Teuchos::Array<bool> edgeIsStrong;
+  {
+    SubFactoryMonitor sfm(*this, "Strength Flags", coarseLevel);
+    GenerateStrengthFlags(*A, *graph, eis_rowptr, edgeIsStrong);
+  }
 
-    // Phase 3: Generate the P matrix
-    RCP<const Map> coarseColMap = coarseMap;
-    RCP<const Map> coarseDomainMap = ownedCoarseMap;
-    if(scheme == "ext+i") {
-      SubFactoryMonitor sfm(*this,"Ext+i Interpolation",coarseLevel);
-      Coarsen_Ext_Plus_I(*A,Aghost,*graph,coarseColMap,coarseDomainMap,num_c_points,num_f_points,myPointType(),myPointType_ghost(),cpoint2pcol,pcol2cpoint,eis_rowptr,edgeIsStrong,BlockNumber,P);
-    }
-    else if(scheme == "direct") {
-      SubFactoryMonitor sfm(*this,"Direct Interpolation",coarseLevel);
-      Coarsen_Direct(*A,Aghost,*graph,coarseColMap,coarseDomainMap,num_c_points,num_f_points,myPointType(),myPointType_ghost(),cpoint2pcol,pcol2cpoint,eis_rowptr,edgeIsStrong,BlockNumber,P);
-    }
-    else if(scheme == "classical modified") {
-      SubFactoryMonitor sfm(*this,"Classical Modified Interpolation",coarseLevel);
-      Coarsen_ClassicalModified(*A,Aghost,*graph,coarseColMap,coarseDomainMap,num_c_points,num_f_points,myPointType(),myPointType_ghost(),cpoint2pcol,pcol2cpoint,eis_rowptr,edgeIsStrong,BlockNumber,remoteOnlyImporter,P);
-    }
-    // NOTE: ParameterList validator will check this guy so we don't really need an "else" here
+  // Phase 3: Generate the P matrix
+  RCP<const Map> coarseColMap    = coarseMap;
+  RCP<const Map> coarseDomainMap = ownedCoarseMap;
+  if (scheme == "ext+i") {
+    SubFactoryMonitor sfm(*this, "Ext+i Interpolation", coarseLevel);
+    Coarsen_Ext_Plus_I(*A, Aghost, *graph, coarseColMap, coarseDomainMap, num_c_points, num_f_points, myPointType(), myPointType_ghost(), cpoint2pcol, pcol2cpoint, eis_rowptr, edgeIsStrong, BlockNumber, P);
+  } else if (scheme == "direct") {
+    SubFactoryMonitor sfm(*this, "Direct Interpolation", coarseLevel);
+    Coarsen_Direct(*A, Aghost, *graph, coarseColMap, coarseDomainMap, num_c_points, num_f_points, myPointType(), myPointType_ghost(), cpoint2pcol, pcol2cpoint, eis_rowptr, edgeIsStrong, BlockNumber, P);
+  } else if (scheme == "classical modified") {
+    SubFactoryMonitor sfm(*this, "Classical Modified Interpolation", coarseLevel);
+    Coarsen_ClassicalModified(*A, Aghost, *graph, coarseColMap, coarseDomainMap, num_c_points, num_f_points, myPointType(), myPointType_ghost(), cpoint2pcol, pcol2cpoint, eis_rowptr, edgeIsStrong, BlockNumber, remoteOnlyImporter, P);
+  }
+  // NOTE: ParameterList validator will check this guy so we don't really need an "else" here
 
 #ifdef CMS_DEBUG
-    Xpetra::IO<SC,LO,GO,NO>::Write("classical_p.mat."+std::to_string(coarseLevel.GetLevelID()), *P);
-    // Xpetra::IO<SC,LO,GO,NO>::WriteLocal("classical_p.mat."+std::to_string(coarseLevel.GetLevelID()), *P);
+  Xpetra::IO<SC, LO, GO, NO>::Write("classical_p.mat." + std::to_string(coarseLevel.GetLevelID()), *P);
+  // Xpetra::IO<SC,LO,GO,NO>::WriteLocal("classical_p.mat."+std::to_string(coarseLevel.GetLevelID()), *P);
 #endif
 
-    // Save output
-    Set(coarseLevel,"P",P);
-    RCP<const CrsGraph> pg = P->getCrsGraph();
-    Set(coarseLevel,"P Graph",pg);
+  // Save output
+  Set(coarseLevel, "P", P);
+  RCP<const CrsGraph> pg = P->getCrsGraph();
+  Set(coarseLevel, "P Graph", pg);
 
-    //RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
-    //    P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(), Teuchos::ScalarTraits<SC>::zero());
-    //    Set(coarseLevel, "Nullspace", coarseNullspace);
+  // RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
+  //     P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(), Teuchos::ScalarTraits<SC>::zero());
+  //     Set(coarseLevel, "Nullspace", coarseNullspace);
 
-    if (IsPrint(Statistics1)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
-    }
+  if (IsPrint(Statistics1)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
   }
+}
 /* ************************************************************************* */
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-Coarsen_ClassicalModified(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph,  RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<const Import> remoteOnlyImporter,RCP<Matrix> & P) const {
-    /* ============================================================= */
-    /* Phase 3 : Classical Modified Interpolation                    */
-    /* De Sterck, Falgout, Nolting and Yang. "Distance-two           */
-    /* interpolation for parallel algebraic multigrid", NLAA 2008    */
-    /* 15:115-139                                                    */
-    /* ============================================================= */
-    /* Definitions:                                                        */
-    /* F = F-points                                                        */
-    /* C = C-points                                                        */
-    /* N_i = non-zero neighbors of node i                                  */
-    /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
-    /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
-    /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
-
-    /* N_i^w = N_i\ (F_i^s \cup C_i^s) [weak neighbors of i]               */
-    /*         This guy has a typo.  The paper had a \cap instead of \cup  */
-    /*         I would note that this set can contain both F-points and    */
-    /*         C-points.  They're just weak neighbors of this guy.         */
-    /*         Note that N_i^w \cup F_i^s \cup C_i^s = N_i by construction */
-
-
-    /* \bar{a}_ij = {    0, if sign(a_ij) == sign(a_ii)                    */
-    /*              { a_ij, otherwise                                      */
-
-    /* F_i^s\star = {k\in N_i | C_i^s \cap C_k^s = \emptyset}              */
-    /*              [set of F-neighbors of i that do not share a strong    */
-    /*               C-neighbor with i]                                    */
-
-
-    /* Rewritten Equation (9) on p. 120                                    */
-    /* \tilde{a}_ii =  (a_ij + \sum_{k\in{N_i^w \cup F_i^s\star}} a_ik     */
-    /*                                                                     */
-    /* f_ij = \sum_{k\in{F_i^s\setminusF_i^s*}} \frac{a_ik \bar{a}_kj}{\sum_{m\inC_i^s \bar{a}_km}}    */
-    /*                                                                     */
-    /* w_ij = \frac{1}{\tilde{a}_ii} ( a_ij + f_ij)  for all j in C_i^s    */
-  
-    //    const point_type F_PT = ClassicalMapFactory::F_PT;
-    const point_type C_PT = ClassicalMapFactory::C_PT;
-    const point_type DIRICHLET_PT = ClassicalMapFactory::DIRICHLET_PT;
-    using STS = typename Teuchos::ScalarTraits<SC>;
-    LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-    //    size_t ST_INVALID = Teuchos::OrdinalTraits<size_t>::invalid();
-    SC SC_ZERO = STS::zero();
+    Coarsen_ClassicalModified(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<const Import> remoteOnlyImporter, RCP<Matrix>& P) const {
+  /* ============================================================= */
+  /* Phase 3 : Classical Modified Interpolation                    */
+  /* De Sterck, Falgout, Nolting and Yang. "Distance-two           */
+  /* interpolation for parallel algebraic multigrid", NLAA 2008    */
+  /* 15:115-139                                                    */
+  /* ============================================================= */
+  /* Definitions:                                                        */
+  /* F = F-points                                                        */
+  /* C = C-points                                                        */
+  /* N_i = non-zero neighbors of node i                                  */
+  /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
+  /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
+  /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
+
+  /* N_i^w = N_i\ (F_i^s \cup C_i^s) [weak neighbors of i]               */
+  /*         This guy has a typo.  The paper had a \cap instead of \cup  */
+  /*         I would note that this set can contain both F-points and    */
+  /*         C-points.  They're just weak neighbors of this guy.         */
+  /*         Note that N_i^w \cup F_i^s \cup C_i^s = N_i by construction */
+
+  /* \bar{a}_ij = {    0, if sign(a_ij) == sign(a_ii)                    */
+  /*              { a_ij, otherwise                                      */
+
+  /* F_i^s\star = {k\in N_i | C_i^s \cap C_k^s = \emptyset}              */
+  /*              [set of F-neighbors of i that do not share a strong    */
+  /*               C-neighbor with i]                                    */
+
+  /* Rewritten Equation (9) on p. 120                                    */
+  /* \tilde{a}_ii =  (a_ij + \sum_{k\in{N_i^w \cup F_i^s\star}} a_ik     */
+  /*                                                                     */
+  /* f_ij = \sum_{k\in{F_i^s\setminusF_i^s*}} \frac{a_ik \bar{a}_kj}{\sum_{m\inC_i^s \bar{a}_km}}    */
+  /*                                                                     */
+  /* w_ij = \frac{1}{\tilde{a}_ii} ( a_ij + f_ij)  for all j in C_i^s    */
+
+  //    const point_type F_PT = ClassicalMapFactory::F_PT;
+  const point_type C_PT         = ClassicalMapFactory::C_PT;
+  const point_type DIRICHLET_PT = ClassicalMapFactory::DIRICHLET_PT;
+  using STS                     = typename Teuchos::ScalarTraits<SC>;
+  LO LO_INVALID                 = Teuchos::OrdinalTraits<LO>::invalid();
+  //    size_t ST_INVALID = Teuchos::OrdinalTraits<size_t>::invalid();
+  SC SC_ZERO = STS::zero();
 #ifdef CMS_DEBUG
-    int rank = A.getRowMap()->getComm()->getRank();
+  int rank = A.getRowMap()->getComm()->getRank();
 #endif
 
-    // Get the block number if we have it.
-    ArrayRCP<const LO> block_id;
-    if(!BlockNumber.is_null())
-      block_id = BlockNumber->getData(0);
-
-    // Initial (estimated) allocation
-    // NOTE: If we only used Tpetra, then we could use these guys as is, but because Epetra, we can't, so there
-    // needs to be a copy below.
-    size_t Nrows = A.getLocalNumRows();
-    double c_point_density = (double)num_c_points / (num_c_points+num_f_points);
-    double mean_strong_neighbors_per_row = (double) graph.GetNodeNumEdges() / graph.GetNodeNumVertices();
-    //    double mean_neighbors_per_row = (double)A.getLocalNumEntries() / Nrows;
-    double nnz_per_row_est = c_point_density*mean_strong_neighbors_per_row;
-
-    size_t nnz_est = std::max(Nrows,std::min((size_t)A.getLocalNumEntries(),(size_t)(nnz_per_row_est*Nrows)));
-    Array<size_t> tmp_rowptr(Nrows+1);
-    Array<LO> tmp_colind(nnz_est);
-
-    // Algorithm (count+realloc)
-    // For each row, i,
-    // - Count the number of elements in \hat{C}_j, aka [C-neighbors and C-neighbors of strong F-neighbors of i]
-    size_t ct=0;
-    for(LO row=0; row < (LO) Nrows; row++) {
-      size_t row_start = eis_rowptr[row];
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      std::set<LO> C_hat;
-      if(myPointType[row] == DIRICHLET_PT) {
-        // Dirichlet points get ignored completely
-      }
-      else if(myPointType[row] == C_PT) {
-        // C-Points get a single 1 in their row
-        C_hat.insert(cpoint2pcol[row]);
-      }
-      else {
-        // F-Points have a more complicated interpolation
-
-        // C-neighbors of row
-        A.getLocalRowView(row, indices, vals);
-        for(LO j=0; j<indices.size(); j++)
-          if(myPointType[indices[j]] == C_PT && edgeIsStrong[row_start + j])
-            C_hat.insert(cpoint2pcol[indices[j]]);
-      }// end else
-      
-      // Realloc if needed
-      if(ct + (size_t)C_hat.size() > (size_t)tmp_colind.size()) {
-        tmp_colind.resize(std::max(ct+(size_t)C_hat.size(),(size_t)2*tmp_colind.size()));
-      }
-      
-      // Copy
-      std::copy(C_hat.begin(), C_hat.end(),tmp_colind.begin()+ct);
-      ct+=C_hat.size();
-      tmp_rowptr[row+1] = tmp_rowptr[row] + C_hat.size();
+  // Get the block number if we have it.
+  ArrayRCP<const LO> block_id;
+  if (!BlockNumber.is_null())
+    block_id = BlockNumber->getData(0);
+
+  // Initial (estimated) allocation
+  // NOTE: If we only used Tpetra, then we could use these guys as is, but because Epetra, we can't, so there
+  // needs to be a copy below.
+  size_t Nrows                         = A.getLocalNumRows();
+  double c_point_density               = (double)num_c_points / (num_c_points + num_f_points);
+  double mean_strong_neighbors_per_row = (double)graph.GetNodeNumEdges() / graph.GetNodeNumVertices();
+  //    double mean_neighbors_per_row = (double)A.getLocalNumEntries() / Nrows;
+  double nnz_per_row_est = c_point_density * mean_strong_neighbors_per_row;
+
+  size_t nnz_est = std::max(Nrows, std::min((size_t)A.getLocalNumEntries(), (size_t)(nnz_per_row_est * Nrows)));
+  Array<size_t> tmp_rowptr(Nrows + 1);
+  Array<LO> tmp_colind(nnz_est);
+
+  // Algorithm (count+realloc)
+  // For each row, i,
+  // - Count the number of elements in \hat{C}_j, aka [C-neighbors and C-neighbors of strong F-neighbors of i]
+  size_t ct = 0;
+  for (LO row = 0; row < (LO)Nrows; row++) {
+    size_t row_start = eis_rowptr[row];
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    std::set<LO> C_hat;
+    if (myPointType[row] == DIRICHLET_PT) {
+      // Dirichlet points get ignored completely
+    } else if (myPointType[row] == C_PT) {
+      // C-Points get a single 1 in their row
+      C_hat.insert(cpoint2pcol[row]);
+    } else {
+      // F-Points have a more complicated interpolation
+
+      // C-neighbors of row
+      A.getLocalRowView(row, indices, vals);
+      for (LO j = 0; j < indices.size(); j++)
+        if (myPointType[indices[j]] == C_PT && edgeIsStrong[row_start + j])
+          C_hat.insert(cpoint2pcol[indices[j]]);
+    }  // end else
+
+    // Realloc if needed
+    if (ct + (size_t)C_hat.size() > (size_t)tmp_colind.size()) {
+      tmp_colind.resize(std::max(ct + (size_t)C_hat.size(), (size_t)2 * tmp_colind.size()));
     }
-    // Resize down
-    tmp_colind.resize(tmp_rowptr[Nrows]);
 
-    RCP<CrsMatrix> Pcrs = CrsMatrixFactory::Build(A.getRowMap(),coarseColMap,0);
-    ArrayRCP<size_t>  P_rowptr;
-    ArrayRCP<LO>      P_colind;
-    ArrayRCP<SC>      P_values;
+    // Copy
+    std::copy(C_hat.begin(), C_hat.end(), tmp_colind.begin() + ct);
+    ct += C_hat.size();
+    tmp_rowptr[row + 1] = tmp_rowptr[row] + C_hat.size();
+  }
+  // Resize down
+  tmp_colind.resize(tmp_rowptr[Nrows]);
+
+  RCP<CrsMatrix> Pcrs = CrsMatrixFactory::Build(A.getRowMap(), coarseColMap, 0);
+  ArrayRCP<size_t> P_rowptr;
+  ArrayRCP<LO> P_colind;
+  ArrayRCP<SC> P_values;
+
+  if (A.getRowMap()->lib() == Xpetra::UseTpetra) {
+    P_rowptr = Teuchos::arcpFromArray(tmp_rowptr);
+    P_colind = Teuchos::arcpFromArray(tmp_colind);
+    P_values.resize(P_rowptr[Nrows]);
+  } else {
+    // Make the matrix and then get the graph out of it (necessary for Epetra)
+    // NOTE: The lack of an Epetra_CrsGraph::ExpertStaticFillComplete makes this
+    // impossible to do the obvious way
+    Pcrs->allocateAllValues(tmp_rowptr[Nrows], P_rowptr, P_colind, P_values);
+    std::copy(tmp_rowptr.begin(), tmp_rowptr.end(), P_rowptr.begin());
+    std::copy(tmp_colind.begin(), tmp_colind.end(), P_colind.begin());
+    Pcrs->setAllValues(P_rowptr, P_colind, P_values);
+    Pcrs->expertStaticFillComplete(/*domain*/ coarseDomainMap, /*range*/ A.getDomainMap());
+  }
 
+  // Generate a remote-ghosted version of the graph (if we're in parallel)
+  RCP<const CrsGraph> Pgraph;
+  RCP<const CrsGraph> Pghost;
+  // TODO: We might want to be more efficient here and actually use
+  // Pgraph in the matrix constructor.
+  ArrayRCP<const size_t> Pghost_rowptr;
+  ArrayRCP<const LO> Pghost_colind;
+  if (!remoteOnlyImporter.is_null()) {
     if (A.getRowMap()->lib() == Xpetra::UseTpetra) {
-      P_rowptr = Teuchos::arcpFromArray(tmp_rowptr);
-      P_colind = Teuchos::arcpFromArray(tmp_colind);
-      P_values.resize(P_rowptr[Nrows]);
+      RCP<CrsGraph> tempPgraph = CrsGraphFactory::Build(A.getRowMap(), coarseColMap, P_rowptr, P_colind);
+      tempPgraph->fillComplete(coarseDomainMap, A.getDomainMap());
+      Pgraph = tempPgraph;
     } else {
-      // Make the matrix and then get the graph out of it (necessary for Epetra)
-      // NOTE: The lack of an Epetra_CrsGraph::ExpertStaticFillComplete makes this
-      // impossible to do the obvious way
-      Pcrs->allocateAllValues(tmp_rowptr[Nrows],P_rowptr,P_colind,P_values);
-      std::copy(tmp_rowptr.begin(),tmp_rowptr.end(), P_rowptr.begin());
-      std::copy(tmp_colind.begin(),tmp_colind.end(), P_colind.begin());
-      Pcrs->setAllValues(P_rowptr,P_colind,P_values);
-      Pcrs->expertStaticFillComplete(/*domain*/coarseDomainMap, /*range*/A.getDomainMap());
-    }
-
-    // Generate a remote-ghosted version of the graph (if we're in parallel)
-    RCP<const CrsGraph> Pgraph;
-    RCP<const CrsGraph> Pghost;
-    // TODO: We might want to be more efficient here and actually use
-    // Pgraph in the matrix constructor.
-    ArrayRCP<const size_t> Pghost_rowptr;
-    ArrayRCP<const LO> Pghost_colind;
-    if(!remoteOnlyImporter.is_null()) {
-      if (A.getRowMap()->lib() == Xpetra::UseTpetra) {
-        RCP<CrsGraph> tempPgraph = CrsGraphFactory::Build(A.getRowMap(),coarseColMap,P_rowptr,P_colind);
-        tempPgraph->fillComplete(coarseDomainMap,A.getDomainMap());
-        Pgraph = tempPgraph;
-      } else {
-        // Epetra does not have a graph constructor that uses rowptr and colind.
-        Pgraph = Pcrs->getCrsGraph();
-      }
-      TEUCHOS_ASSERT(!Pgraph.is_null());
-      Pghost = CrsGraphFactory::Build(Pgraph,*remoteOnlyImporter,Pgraph->getDomainMap(),remoteOnlyImporter->getTargetMap());
-      Pghost->getAllIndices(Pghost_rowptr,Pghost_colind);
+      // Epetra does not have a graph constructor that uses rowptr and colind.
+      Pgraph = Pcrs->getCrsGraph();
     }
+    TEUCHOS_ASSERT(!Pgraph.is_null());
+    Pghost = CrsGraphFactory::Build(Pgraph, *remoteOnlyImporter, Pgraph->getDomainMap(), remoteOnlyImporter->getTargetMap());
+    Pghost->getAllIndices(Pghost_rowptr, Pghost_colind);
+  }
 
-    // Gustavson-style perfect hashing
-    ArrayRCP<LO> Acol_to_Pcol(A.getColMap()->getLocalNumElements(),LO_INVALID);
-
-    // Get a quick reindexing array from Pghost LCIDs to P LCIDs
-    ArrayRCP<LO> Pghostcol_to_Pcol;
-    if(!Pghost.is_null()) {
-      Pghostcol_to_Pcol.resize(Pghost->getColMap()->getLocalNumElements(),LO_INVALID);
-      for(LO i=0; i<(LO) Pghost->getColMap()->getLocalNumElements(); i++)
-        Pghostcol_to_Pcol[i] = Pgraph->getColMap()->getLocalElement(Pghost->getColMap()->getGlobalElement(i));
-    }//end Pghost
-
-    // Get a quick reindexing array from Aghost LCIDs to A LCIDs
-    ArrayRCP<LO> Aghostcol_to_Acol;
-    if(!Aghost.is_null()) {
-      Aghostcol_to_Acol.resize(Aghost->getColMap()->getLocalNumElements(),LO_INVALID);
-      for(LO i=0; i<(LO)Aghost->getColMap()->getLocalNumElements(); i++)
-        Aghostcol_to_Acol[i] = A.getColMap()->getLocalElement(Aghost->getColMap()->getGlobalElement(i));
-    }//end Aghost
-
-
-    // Algorithm (numeric)
-    for(LO i=0; i < (LO)Nrows; i++) {
-      if(myPointType[i] == DIRICHLET_PT) {
-        // Dirichlet points get ignored completely
+  // Gustavson-style perfect hashing
+  ArrayRCP<LO> Acol_to_Pcol(A.getColMap()->getLocalNumElements(), LO_INVALID);
+
+  // Get a quick reindexing array from Pghost LCIDs to P LCIDs
+  ArrayRCP<LO> Pghostcol_to_Pcol;
+  if (!Pghost.is_null()) {
+    Pghostcol_to_Pcol.resize(Pghost->getColMap()->getLocalNumElements(), LO_INVALID);
+    for (LO i = 0; i < (LO)Pghost->getColMap()->getLocalNumElements(); i++)
+      Pghostcol_to_Pcol[i] = Pgraph->getColMap()->getLocalElement(Pghost->getColMap()->getGlobalElement(i));
+  }  // end Pghost
+
+  // Get a quick reindexing array from Aghost LCIDs to A LCIDs
+  ArrayRCP<LO> Aghostcol_to_Acol;
+  if (!Aghost.is_null()) {
+    Aghostcol_to_Acol.resize(Aghost->getColMap()->getLocalNumElements(), LO_INVALID);
+    for (LO i = 0; i < (LO)Aghost->getColMap()->getLocalNumElements(); i++)
+      Aghostcol_to_Acol[i] = A.getColMap()->getLocalElement(Aghost->getColMap()->getGlobalElement(i));
+  }  // end Aghost
+
+  // Algorithm (numeric)
+  for (LO i = 0; i < (LO)Nrows; i++) {
+    if (myPointType[i] == DIRICHLET_PT) {
+      // Dirichlet points get ignored completely
 #ifdef CMS_DEBUG
-        // DEBUG
-        printf("[%d] ** A(%d,:) is a Dirichlet-Point.\n",rank,i);
+      // DEBUG
+      printf("[%d] ** A(%d,:) is a Dirichlet-Point.\n", rank, i);
 #endif
-      }
-      else if (myPointType[i] == C_PT) {
-        // C Points get a single 1 in their row
-        P_values[P_rowptr[i]] = Teuchos::ScalarTraits<SC>::one();
+    } else if (myPointType[i] == C_PT) {
+      // C Points get a single 1 in their row
+      P_values[P_rowptr[i]] = Teuchos::ScalarTraits<SC>::one();
 #ifdef CMS_DEBUG
-        // DEBUG
-        printf("[%d] ** A(%d,:) is a C-Point.\n",rank,i);
+      // DEBUG
+      printf("[%d] ** A(%d,:) is a C-Point.\n", rank, i);
 #endif
-      }
-      else {
-        // F Points get all of the fancy stuff
+    } else {
+      // F Points get all of the fancy stuff
 #ifdef CMS_DEBUG
-        // DEBUG
-        printf("[%d] ** A(%d,:) is a F-Point.\n",rank,i);
+      // DEBUG
+      printf("[%d] ** A(%d,:) is a F-Point.\n", rank, i);
 #endif
-        
-        // Get all of the relevant information about this row
-        ArrayView<const LO> A_indices_i, A_indices_k;
-        ArrayView<const SC> A_vals_i, A_vals_k;
-        A.getLocalRowView(i, A_indices_i, A_vals_i);
-        size_t row_start = eis_rowptr[i];
-        
-        ArrayView<const LO> P_indices_i  = P_colind.view(P_rowptr[i],P_rowptr[i+1] - P_rowptr[i]);
-        ArrayView<SC> P_vals_i     = P_values.view(P_rowptr[i],P_rowptr[i+1] - P_rowptr[i]);
-        
-        // FIXME: Do we need this?
-        for(LO j=0; j<(LO)P_vals_i.size(); j++)
-          P_vals_i[j] = SC_ZERO;
-
-        // Stash the hash:  Flag any strong C-points with their index into P_colind
-        // NOTE:  We'll consider any points that are LO_INVALID or less than P_rowptr[i] as not strong C-points
-        for(LO j=0; j<(LO)P_indices_i.size(); j++)  {
-          Acol_to_Pcol[pcol2cpoint[P_indices_i[j]]] = P_rowptr[i] + j;
-        }
 
-        // Loop over the entries in the row
-        SC first_denominator  = SC_ZERO;
+      // Get all of the relevant information about this row
+      ArrayView<const LO> A_indices_i, A_indices_k;
+      ArrayView<const SC> A_vals_i, A_vals_k;
+      A.getLocalRowView(i, A_indices_i, A_vals_i);
+      size_t row_start = eis_rowptr[i];
+
+      ArrayView<const LO> P_indices_i = P_colind.view(P_rowptr[i], P_rowptr[i + 1] - P_rowptr[i]);
+      ArrayView<SC> P_vals_i          = P_values.view(P_rowptr[i], P_rowptr[i + 1] - P_rowptr[i]);
+
+      // FIXME: Do we need this?
+      for (LO j = 0; j < (LO)P_vals_i.size(); j++)
+        P_vals_i[j] = SC_ZERO;
+
+      // Stash the hash:  Flag any strong C-points with their index into P_colind
+      // NOTE:  We'll consider any points that are LO_INVALID or less than P_rowptr[i] as not strong C-points
+      for (LO j = 0; j < (LO)P_indices_i.size(); j++) {
+        Acol_to_Pcol[pcol2cpoint[P_indices_i[j]]] = P_rowptr[i] + j;
+      }
+
+      // Loop over the entries in the row
+      SC first_denominator = SC_ZERO;
 #ifdef CMS_DEBUG
-        SC a_ii = SC_ZERO;
+      SC a_ii = SC_ZERO;
 #endif
-        for(LO k0=0; k0<(LO)A_indices_i.size(); k0++) {
-          LO k      = A_indices_i[k0];
-          SC a_ik   = A_vals_i[k0];
-          LO pcol_k = Acol_to_Pcol[k];
-
-          if(k == i) {
-            // Case A: Diagonal value (add to first denominator)
-            // FIXME:  Add BlockNumber matching here
-            first_denominator += a_ik;
+      for (LO k0 = 0; k0 < (LO)A_indices_i.size(); k0++) {
+        LO k      = A_indices_i[k0];
+        SC a_ik   = A_vals_i[k0];
+        LO pcol_k = Acol_to_Pcol[k];
+
+        if (k == i) {
+          // Case A: Diagonal value (add to first denominator)
+          // FIXME:  Add BlockNumber matching here
+          first_denominator += a_ik;
 #ifdef CMS_DEBUG
-            a_ii = a_ik;
-            printf("- A(%d,%d) is the diagonal\n",i,k);
+          a_ii = a_ik;
+          printf("- A(%d,%d) is the diagonal\n", i, k);
 #endif
 
-          }
-          else if(myPointType[k] == DIRICHLET_PT) {
-            // Case B: Ignore dirichlet connections completely
+        } else if (myPointType[k] == DIRICHLET_PT) {
+          // Case B: Ignore dirichlet connections completely
 #ifdef CMS_DEBUG
-            printf("- A(%d,%d) is a Dirichlet point\n",i,k);
+          printf("- A(%d,%d) is a Dirichlet point\n", i, k);
 #endif
-            
-          }
-          else if(pcol_k != LO_INVALID && pcol_k >= (LO)P_rowptr[i]) {
-            // Case C: a_ik is strong C-Point (goes directly into the weight)
-            P_values[pcol_k] += a_ik;
+
+        } else if (pcol_k != LO_INVALID && pcol_k >= (LO)P_rowptr[i]) {
+          // Case C: a_ik is strong C-Point (goes directly into the weight)
+          P_values[pcol_k] += a_ik;
 #ifdef CMS_DEBUG
-            printf("- A(%d,%d) is a strong C-Point\n",i,k);
+          printf("- A(%d,%d) is a strong C-Point\n", i, k);
 #endif
-          }
-          else if (!edgeIsStrong[row_start + k0]) {
-            // Case D: Weak non-Dirichlet neighbor (add to first denominator)
-            if(block_id.size() == 0 || block_id[i] == block_id[k])  {
-              first_denominator += a_ik;
+        } else if (!edgeIsStrong[row_start + k0]) {
+          // Case D: Weak non-Dirichlet neighbor (add to first denominator)
+          if (block_id.size() == 0 || block_id[i] == block_id[k]) {
+            first_denominator += a_ik;
 #ifdef CMS_DEBUG
-              printf("- A(%d,%d) is weak adding to diagonal(%d,%d) (%6.4e)\n",i,k,block_id[i],block_id[k],a_ik);
-            }
-            else {
-              printf("- A(%d,%d) is weak but does not match blocks (%d,%d), discarding\n",i,k,block_id[i],block_id[k]);
+            printf("- A(%d,%d) is weak adding to diagonal(%d,%d) (%6.4e)\n", i, k, block_id[i], block_id[k], a_ik);
+          } else {
+            printf("- A(%d,%d) is weak but does not match blocks (%d,%d), discarding\n", i, k, block_id[i], block_id[k]);
 #endif
-            }
-            
           }
-          else {//Case E
-            // Case E: Strong F-Point (adds to the first denominator if we don't share a
-            // a strong C-Point with i; adds to the second denominator otherwise)
+
+        } else {  // Case E
+                  // Case E: Strong F-Point (adds to the first denominator if we don't share a
+                  // a strong C-Point with i; adds to the second denominator otherwise)
 #ifdef CMS_DEBUG
-            printf("- A(%d,%d) is a strong F-Point\n",i,k);
+          printf("- A(%d,%d) is a strong F-Point\n", i, k);
 #endif
 
-            SC a_kk = SC_ZERO;
-            SC second_denominator = SC_ZERO;
-            int sign_akk = 0;
-            
-            if(k < (LO)Nrows) {
-              // Grab the diagonal a_kk
-              A.getLocalRowView(k, A_indices_k, A_vals_k);
-              for(LO m0=0; m0<(LO)A_indices_k.size(); m0++){
-                LO m    = A_indices_k[m0];
-                if(k == m) {
-                  a_kk = A_vals_k[m0];
-                  break;
-                }
-              }//end for A_indices_k
-              
-              // Compute the second denominator term
-              sign_akk = Sign(a_kk);
-              for(LO m0=0; m0<(LO)A_indices_k.size(); m0++){
-                LO m    = A_indices_k[m0];
-                if(m != k && Acol_to_Pcol[A_indices_k[m0]] >=  (LO)P_rowptr[i]) {
-                  SC a_km = A_vals_k[m0];
-                  second_denominator+= (Sign(a_km) == sign_akk ? SC_ZERO : a_km);
-                }
-              }//end for A_indices_k
-
-              // Now we have the second denominator, for this particular strong F point.
-              // So we can now add the sum to the w_ij components for the P values
-              if(second_denominator != SC_ZERO) {
-                for(LO j0=0; j0<(LO)A_indices_k.size(); j0++) {
-                  LO j    = A_indices_k[j0];
-                  // NOTE: Row k should be in fis_star, so I should have to check for diagonals here
-                  //                  printf("Acol_to_Pcol[%d] = %d P_values.size() = %d\n",j,Acol_to_Pcol[j],(int)P_values.size());
-                  if(Acol_to_Pcol[j] >=  (LO)P_rowptr[i]) {
-                    SC a_kj = A_vals_k[j0];
-                    SC sign_akj_val = sign_akk == Sign(a_kj) ? SC_ZERO : a_kj;
-                    P_values[Acol_to_Pcol[j]] += a_ik * sign_akj_val / second_denominator;
+          SC a_kk               = SC_ZERO;
+          SC second_denominator = SC_ZERO;
+          int sign_akk          = 0;
+
+          if (k < (LO)Nrows) {
+            // Grab the diagonal a_kk
+            A.getLocalRowView(k, A_indices_k, A_vals_k);
+            for (LO m0 = 0; m0 < (LO)A_indices_k.size(); m0++) {
+              LO m = A_indices_k[m0];
+              if (k == m) {
+                a_kk = A_vals_k[m0];
+                break;
+              }
+            }  // end for A_indices_k
+
+            // Compute the second denominator term
+            sign_akk = Sign(a_kk);
+            for (LO m0 = 0; m0 < (LO)A_indices_k.size(); m0++) {
+              LO m = A_indices_k[m0];
+              if (m != k && Acol_to_Pcol[A_indices_k[m0]] >= (LO)P_rowptr[i]) {
+                SC a_km = A_vals_k[m0];
+                second_denominator += (Sign(a_km) == sign_akk ? SC_ZERO : a_km);
+              }
+            }  // end for A_indices_k
+
+            // Now we have the second denominator, for this particular strong F point.
+            // So we can now add the sum to the w_ij components for the P values
+            if (second_denominator != SC_ZERO) {
+              for (LO j0 = 0; j0 < (LO)A_indices_k.size(); j0++) {
+                LO j = A_indices_k[j0];
+                // NOTE: Row k should be in fis_star, so I should have to check for diagonals here
+                //                  printf("Acol_to_Pcol[%d] = %d P_values.size() = %d\n",j,Acol_to_Pcol[j],(int)P_values.size());
+                if (Acol_to_Pcol[j] >= (LO)P_rowptr[i]) {
+                  SC a_kj         = A_vals_k[j0];
+                  SC sign_akj_val = sign_akk == Sign(a_kj) ? SC_ZERO : a_kj;
+                  P_values[Acol_to_Pcol[j]] += a_ik * sign_akj_val / second_denominator;
 #ifdef CMS_DEBUG
-                    printf("- - Unscaled P(%d,A-%d) += %6.4e = %5.4e\n",i,j,a_ik * sign_akj_val / second_denominator,P_values[Acol_to_Pcol[j]]);
+                  printf("- - Unscaled P(%d,A-%d) += %6.4e = %5.4e\n", i, j, a_ik * sign_akj_val / second_denominator, P_values[Acol_to_Pcol[j]]);
 #endif
-                  }
-                }//end for A_indices_k
-              }//end if second_denominator != 0
-              else {
+                }
+              }  // end for A_indices_k
+            }    // end if second_denominator != 0
+            else {
 #ifdef CMS_DEBUG
-                printf("- - A(%d,%d) second denominator is zero\n",i,k);
+              printf("- - A(%d,%d) second denominator is zero\n", i, k);
 #endif
-                if(block_id.size() == 0 || block_id[i] == block_id[k])
-                  first_denominator += a_ik;
-              }//end else second_denominator != 0
-            }// end if k < Nrows
-            else {
-              // Ghost row
-              LO kless = k-Nrows;
-              // Grab the diagonal a_kk
-              // NOTE: ColMap is not locally fitted to the RowMap
-              // so we need to check GIDs here
-              Aghost->getLocalRowView(kless, A_indices_k, A_vals_k);
-              GO k_g = Aghost->getRowMap()->getGlobalElement(kless);
-              for(LO m0=0; m0<(LO)A_indices_k.size(); m0++){
-                GO m_g    = Aghost->getColMap()->getGlobalElement(A_indices_k[m0]);
-                if(k_g == m_g) {
-                  a_kk = A_vals_k[m0];
-                  break;
-                }
-              }//end for A_indices_k
-
-              // Compute the second denominator term
-              sign_akk = Sign(a_kk);
-              for(LO m0=0; m0<(LO)A_indices_k.size(); m0++){
-                GO m_g    = Aghost->getColMap()->getGlobalElement(A_indices_k[m0]);
-                LO mghost = A_indices_k[m0];//Aghost LCID
-                LO m = Aghostcol_to_Acol[mghost]; //A's LID (could be LO_INVALID)
-                if(m_g != k_g && m != LO_INVALID && Acol_to_Pcol[m] >=  (LO)P_rowptr[i]) {
-                  SC a_km = A_vals_k[m0];
-                  second_denominator+= (Sign(a_km) == sign_akk ? SC_ZERO : a_km);
-                }
-              }//end for A_indices_k
-              
-              
-              // Now we have the second denominator, for this particular strong F point.
-              // So we can now add the sum to the w_ij components for the P values
-              if(second_denominator != SC_ZERO) {
-                for(LO j0=0; j0<(LO)A_indices_k.size(); j0++){
-                  LO jghost = A_indices_k[j0];//Aghost LCID
-                  LO j = Aghostcol_to_Acol[jghost]; //A's LID (could be LO_INVALID)
-                  // NOTE: Row k should be in fis_star, so I should have to check for diagonals here
-                  if((j != LO_INVALID) && (Acol_to_Pcol[j] >=  (LO)P_rowptr[i])) {
-                    SC a_kj = A_vals_k[j0];
-                    SC sign_akj_val = sign_akk == Sign(a_kj) ? SC_ZERO : a_kj;
-                    P_values[Acol_to_Pcol[j]] += a_ik * sign_akj_val / second_denominator;
+              if (block_id.size() == 0 || block_id[i] == block_id[k])
+                first_denominator += a_ik;
+            }  // end else second_denominator != 0
+          }    // end if k < Nrows
+          else {
+            // Ghost row
+            LO kless = k - Nrows;
+            // Grab the diagonal a_kk
+            // NOTE: ColMap is not locally fitted to the RowMap
+            // so we need to check GIDs here
+            Aghost->getLocalRowView(kless, A_indices_k, A_vals_k);
+            GO k_g = Aghost->getRowMap()->getGlobalElement(kless);
+            for (LO m0 = 0; m0 < (LO)A_indices_k.size(); m0++) {
+              GO m_g = Aghost->getColMap()->getGlobalElement(A_indices_k[m0]);
+              if (k_g == m_g) {
+                a_kk = A_vals_k[m0];
+                break;
+              }
+            }  // end for A_indices_k
+
+            // Compute the second denominator term
+            sign_akk = Sign(a_kk);
+            for (LO m0 = 0; m0 < (LO)A_indices_k.size(); m0++) {
+              GO m_g    = Aghost->getColMap()->getGlobalElement(A_indices_k[m0]);
+              LO mghost = A_indices_k[m0];            // Aghost LCID
+              LO m      = Aghostcol_to_Acol[mghost];  // A's LID (could be LO_INVALID)
+              if (m_g != k_g && m != LO_INVALID && Acol_to_Pcol[m] >= (LO)P_rowptr[i]) {
+                SC a_km = A_vals_k[m0];
+                second_denominator += (Sign(a_km) == sign_akk ? SC_ZERO : a_km);
+              }
+            }  // end for A_indices_k
+
+            // Now we have the second denominator, for this particular strong F point.
+            // So we can now add the sum to the w_ij components for the P values
+            if (second_denominator != SC_ZERO) {
+              for (LO j0 = 0; j0 < (LO)A_indices_k.size(); j0++) {
+                LO jghost = A_indices_k[j0];            // Aghost LCID
+                LO j      = Aghostcol_to_Acol[jghost];  // A's LID (could be LO_INVALID)
+                // NOTE: Row k should be in fis_star, so I should have to check for diagonals here
+                if ((j != LO_INVALID) && (Acol_to_Pcol[j] >= (LO)P_rowptr[i])) {
+                  SC a_kj         = A_vals_k[j0];
+                  SC sign_akj_val = sign_akk == Sign(a_kj) ? SC_ZERO : a_kj;
+                  P_values[Acol_to_Pcol[j]] += a_ik * sign_akj_val / second_denominator;
 #ifdef CMS_DEBUG
-                    printf("- - Unscaled P(%d,A-%d) += %6.4e\n",i,j,a_ik * sign_akj_val / second_denominator);
+                  printf("- - Unscaled P(%d,A-%d) += %6.4e\n", i, j, a_ik * sign_akj_val / second_denominator);
 #endif
-                  }
+                }
 
-                }//end for A_indices_k
-              }//end if second_denominator != 0
-              else {
+              }  // end for A_indices_k
+            }    // end if second_denominator != 0
+            else {
 #ifdef CMS_DEBUG
-                printf("- - A(%d,%d) second denominator is zero\n",i,k);
+              printf("- - A(%d,%d) second denominator is zero\n", i, k);
 #endif
-                if(block_id.size() == 0 || block_id[i] == block_id[k])
-                  first_denominator += a_ik;
-              }//end else second_denominator != 0
-            }//end else k < Nrows
-          }//end else Case A,...,E
-                           
-        }//end for A_indices_i
-
-        // Now, downscale by the first_denominator
-        if(first_denominator != SC_ZERO) {
-          for(LO j0=0; j0<(LO)P_indices_i.size(); j0++)  {
+              if (block_id.size() == 0 || block_id[i] == block_id[k])
+                first_denominator += a_ik;
+            }  // end else second_denominator != 0
+          }    // end else k < Nrows
+        }      // end else Case A,...,E
+
+      }  // end for A_indices_i
+
+      // Now, downscale by the first_denominator
+      if (first_denominator != SC_ZERO) {
+        for (LO j0 = 0; j0 < (LO)P_indices_i.size(); j0++) {
 #ifdef CMS_DEBUG
-            SC old_pij = P_vals_i[j0];
-            P_vals_i[j0] /= -first_denominator;
-            printf("P(%d,%d) = %6.4e = %6.4e / (%6.4e + %6.4e)\n",i,P_indices_i[j0],P_vals_i[j0],old_pij,a_ii,first_denominator - a_ii);
+          SC old_pij = P_vals_i[j0];
+          P_vals_i[j0] /= -first_denominator;
+          printf("P(%d,%d) = %6.4e = %6.4e / (%6.4e + %6.4e)\n", i, P_indices_i[j0], P_vals_i[j0], old_pij, a_ii, first_denominator - a_ii);
 #else
-            P_vals_i[j0] /= -first_denominator;
+          P_vals_i[j0] /= -first_denominator;
 #endif
-          }//end for P_indices_i
-        }//end if first_denominator != 0
- 
-      }//end else C-Point
+        }  // end for P_indices_i
+      }    // end if first_denominator != 0
 
-    }// end if i < Nrows
+    }  // end else C-Point
 
-    // Finish up
-    Pcrs->setAllValues(P_rowptr,P_colind,P_values);
-    Pcrs->expertStaticFillComplete(/*domain*/coarseDomainMap, /*range*/A.getDomainMap());
-    // Wrap from CrsMatrix to Matrix
-    P = rcp(new CrsMatrixWrap(Pcrs));
+  }  // end if i < Nrows
 
-}//end Coarsen_ClassicalModified
+  // Finish up
+  Pcrs->setAllValues(P_rowptr, P_colind, P_values);
+  Pcrs->expertStaticFillComplete(/*domain*/ coarseDomainMap, /*range*/ A.getDomainMap());
+  // Wrap from CrsMatrix to Matrix
+  P = rcp(new CrsMatrixWrap(Pcrs));
 
+}  // end Coarsen_ClassicalModified
 
 /* ************************************************************************* */
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-Coarsen_Direct(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph,  RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<Matrix> & P) const {
-    /* ============================================================= */
-    /* Phase 3 : Direct Interpolation                                */
-    /* We do not use De Sterck, Falgout, Nolting and Yang (2008)     */
-    /* here.  Instead we follow:                                     */
-    /* Trottenberg, Oosterlee and Schueller, Multigrid, 2001.        */
-    /* with some modifications inspirted by PyAMG                    */
-    /* ============================================================= */
-    /* Definitions:                                                        */
-    /* F = F-points                                                        */
-    /* C = C-points                                                        */
-    /* N_i = non-zero neighbors of node i                                  */
-    /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
-    /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
-    /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
-    /* P_i = Set of interpolatory variables for row i [here = C_i^s]       */
-
-    /* (A.2.17) from p. 426                                                */
-    /* a_ij^- = {  a_ij,  if a_ij < 0                                      */
-    /*          {     0,  otherwise                                        */
-    /* a_ij^+ = {  a_ij,  if a_ij > 0                                      */
-    /*          {     0,  otherwise                                        */
-    /* P_i^- =  P_i \cap {k | a_ij^- != 0 and a_ij^- = a_ij}               */
-    /*          [strong C-neighbors with negative edges]                   */
-    /* P_i^+ =  P_i \cap {k | a_ij^+ != 0 and a_ij^+ = a_ij}               */
-    /*          [strong C-neighbors with positive edges]                   */
-
-
-    /* de Sterck et al., gives us this:                                                      */
-    /* Rewritten Equation (6) on p. 119                                                      */
-    /* w_ij = - a_ji / a_ii \frac{\sum_{k\in N_i} a_ik} {\sum k\inC_i^s} a_ik},   j\in C_i^s */
-
-    /* Trottenberg et al. (A.7.6) and (A.7.7) on p. 479 gives this:                          */
-    /* alpha_i = \frac{ \sum_{j\in N_i} a_ij^- }{ \sum_{k\in P_i} a_ik^- }                   */
-    /* beta_i  = \frac{ \sum_{j\in N_i} a_ij^+ }{ \sum_{k\in P_i} a_ik^+ }                   */
-    /* w_ik    = { - alpha_i (a_ik / a_ii),   if k\in P_i^-                                  */
-    /*           { -  beta_i (a_ik / a_ii),   if k\in P_i^+                                  */
-    /* NOTE: The text says to modify, if  P_i^+ is zero but it isn't entirely clear how that */
-    /* works.  We'll follow the PyAMG implementation in a few important ways.                */
-     
-    const point_type C_PT = ClassicalMapFactory::C_PT;
-    const point_type DIRICHLET_PT = ClassicalMapFactory::DIRICHLET_PT;
-   
-    // Initial (estimated) allocation
-    // NOTE: If we only used Tpetra, then we could use these guys as is, but because Epetra, we can't, so there
-    // needs to be a copy below.
-    using STS = typename Teuchos::ScalarTraits<SC>;
-    using MT  = typename STS::magnitudeType;
-    using MTS = typename Teuchos::ScalarTraits<MT>;
-    size_t Nrows = A.getLocalNumRows();
-    double c_point_density = (double)num_c_points / (num_c_points+num_f_points);
-    double mean_strong_neighbors_per_row = (double) graph.GetNodeNumEdges() / graph.GetNodeNumVertices();
-    //    double mean_neighbors_per_row = (double)A.getLocalNumEntries() / Nrows;
-    double nnz_per_row_est = c_point_density*mean_strong_neighbors_per_row;
-
-    size_t nnz_est = std::max(Nrows,std::min((size_t)A.getLocalNumEntries(),(size_t)(nnz_per_row_est*Nrows)));
-    SC SC_ZERO = STS::zero();
-    MT MT_ZERO = MTS::zero();
-    Array<size_t> tmp_rowptr(Nrows+1);
-    Array<LO> tmp_colind(nnz_est);
-
-    // Algorithm (count+realloc)
-    // For each row, i,
-    // - Count the number of elements in \hat{C}_j, aka [C-neighbors and C-neighbors of strong F-neighbors of i]
-    size_t ct=0;
-    for(LO row=0; row < (LO) Nrows; row++) {
-      size_t row_start = eis_rowptr[row];
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      std::set<LO> C_hat;
-      if(myPointType[row] == DIRICHLET_PT) {
-        // Dirichlet points get ignored completely
-      }
-      else if(myPointType[row] == C_PT) {
-        // C-Points get a single 1 in their row
-        C_hat.insert(cpoint2pcol[row]);
-      }
-      else {
-        // F-Points have a more complicated interpolation
-
-        // C-neighbors of row
-        A.getLocalRowView(row, indices, vals);
-        for(LO j=0; j<indices.size(); j++)
-          if(myPointType[indices[j]] == C_PT && edgeIsStrong[row_start + j])
-            C_hat.insert(cpoint2pcol[indices[j]]);
-      }// end else
-      
-      // Realloc if needed
-      if(ct + (size_t)C_hat.size() > (size_t)tmp_colind.size()) {
-        tmp_colind.resize(std::max(ct+(size_t)C_hat.size(),(size_t)2*tmp_colind.size()));
-      }
-      
-      // Copy
-      std::copy(C_hat.begin(), C_hat.end(),tmp_colind.begin()+ct);
-      ct+=C_hat.size();
-      tmp_rowptr[row+1] = tmp_rowptr[row] + C_hat.size();
+    Coarsen_Direct(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<Matrix>& P) const {
+  /* ============================================================= */
+  /* Phase 3 : Direct Interpolation                                */
+  /* We do not use De Sterck, Falgout, Nolting and Yang (2008)     */
+  /* here.  Instead we follow:                                     */
+  /* Trottenberg, Oosterlee and Schueller, Multigrid, 2001.        */
+  /* with some modifications inspirted by PyAMG                    */
+  /* ============================================================= */
+  /* Definitions:                                                        */
+  /* F = F-points                                                        */
+  /* C = C-points                                                        */
+  /* N_i = non-zero neighbors of node i                                  */
+  /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
+  /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
+  /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
+  /* P_i = Set of interpolatory variables for row i [here = C_i^s]       */
+
+  /* (A.2.17) from p. 426                                                */
+  /* a_ij^- = {  a_ij,  if a_ij < 0                                      */
+  /*          {     0,  otherwise                                        */
+  /* a_ij^+ = {  a_ij,  if a_ij > 0                                      */
+  /*          {     0,  otherwise                                        */
+  /* P_i^- =  P_i \cap {k | a_ij^- != 0 and a_ij^- = a_ij}               */
+  /*          [strong C-neighbors with negative edges]                   */
+  /* P_i^+ =  P_i \cap {k | a_ij^+ != 0 and a_ij^+ = a_ij}               */
+  /*          [strong C-neighbors with positive edges]                   */
+
+  /* de Sterck et al., gives us this:                                                      */
+  /* Rewritten Equation (6) on p. 119                                                      */
+  /* w_ij = - a_ji / a_ii \frac{\sum_{k\in N_i} a_ik} {\sum k\inC_i^s} a_ik},   j\in C_i^s */
+
+  /* Trottenberg et al. (A.7.6) and (A.7.7) on p. 479 gives this:                          */
+  /* alpha_i = \frac{ \sum_{j\in N_i} a_ij^- }{ \sum_{k\in P_i} a_ik^- }                   */
+  /* beta_i  = \frac{ \sum_{j\in N_i} a_ij^+ }{ \sum_{k\in P_i} a_ik^+ }                   */
+  /* w_ik    = { - alpha_i (a_ik / a_ii),   if k\in P_i^-                                  */
+  /*           { -  beta_i (a_ik / a_ii),   if k\in P_i^+                                  */
+  /* NOTE: The text says to modify, if  P_i^+ is zero but it isn't entirely clear how that */
+  /* works.  We'll follow the PyAMG implementation in a few important ways.                */
+
+  const point_type C_PT         = ClassicalMapFactory::C_PT;
+  const point_type DIRICHLET_PT = ClassicalMapFactory::DIRICHLET_PT;
+
+  // Initial (estimated) allocation
+  // NOTE: If we only used Tpetra, then we could use these guys as is, but because Epetra, we can't, so there
+  // needs to be a copy below.
+  using STS                            = typename Teuchos::ScalarTraits<SC>;
+  using MT                             = typename STS::magnitudeType;
+  using MTS                            = typename Teuchos::ScalarTraits<MT>;
+  size_t Nrows                         = A.getLocalNumRows();
+  double c_point_density               = (double)num_c_points / (num_c_points + num_f_points);
+  double mean_strong_neighbors_per_row = (double)graph.GetNodeNumEdges() / graph.GetNodeNumVertices();
+  //    double mean_neighbors_per_row = (double)A.getLocalNumEntries() / Nrows;
+  double nnz_per_row_est = c_point_density * mean_strong_neighbors_per_row;
+
+  size_t nnz_est = std::max(Nrows, std::min((size_t)A.getLocalNumEntries(), (size_t)(nnz_per_row_est * Nrows)));
+  SC SC_ZERO     = STS::zero();
+  MT MT_ZERO     = MTS::zero();
+  Array<size_t> tmp_rowptr(Nrows + 1);
+  Array<LO> tmp_colind(nnz_est);
+
+  // Algorithm (count+realloc)
+  // For each row, i,
+  // - Count the number of elements in \hat{C}_j, aka [C-neighbors and C-neighbors of strong F-neighbors of i]
+  size_t ct = 0;
+  for (LO row = 0; row < (LO)Nrows; row++) {
+    size_t row_start = eis_rowptr[row];
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    std::set<LO> C_hat;
+    if (myPointType[row] == DIRICHLET_PT) {
+      // Dirichlet points get ignored completely
+    } else if (myPointType[row] == C_PT) {
+      // C-Points get a single 1 in their row
+      C_hat.insert(cpoint2pcol[row]);
+    } else {
+      // F-Points have a more complicated interpolation
+
+      // C-neighbors of row
+      A.getLocalRowView(row, indices, vals);
+      for (LO j = 0; j < indices.size(); j++)
+        if (myPointType[indices[j]] == C_PT && edgeIsStrong[row_start + j])
+          C_hat.insert(cpoint2pcol[indices[j]]);
+    }  // end else
+
+    // Realloc if needed
+    if (ct + (size_t)C_hat.size() > (size_t)tmp_colind.size()) {
+      tmp_colind.resize(std::max(ct + (size_t)C_hat.size(), (size_t)2 * tmp_colind.size()));
     }
-    // Resize down
-    tmp_colind.resize(tmp_rowptr[Nrows]);
 
-    // Allocate memory & copy
-    P = rcp(new CrsMatrixWrap(A.getRowMap(), coarseColMap, 0));
-    RCP<CrsMatrix> PCrs   = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-    ArrayRCP<size_t>  P_rowptr;
-    ArrayRCP<LO>      P_colind;
-    ArrayRCP<SC>      P_values;
+    // Copy
+    std::copy(C_hat.begin(), C_hat.end(), tmp_colind.begin() + ct);
+    ct += C_hat.size();
+    tmp_rowptr[row + 1] = tmp_rowptr[row] + C_hat.size();
+  }
+  // Resize down
+  tmp_colind.resize(tmp_rowptr[Nrows]);
+
+  // Allocate memory & copy
+  P                   = rcp(new CrsMatrixWrap(A.getRowMap(), coarseColMap, 0));
+  RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  ArrayRCP<size_t> P_rowptr;
+  ArrayRCP<LO> P_colind;
+  ArrayRCP<SC> P_values;
 
 #ifdef CMS_DEBUG
-printf("CMS: Allocating P w/ %d nonzeros\n",(int)tmp_rowptr[Nrows]);
+  printf("CMS: Allocating P w/ %d nonzeros\n", (int)tmp_rowptr[Nrows]);
 #endif
-    PCrs->allocateAllValues(tmp_rowptr[Nrows], P_rowptr, P_colind, P_values);
-    TEUCHOS_TEST_FOR_EXCEPTION(tmp_rowptr.size() !=P_rowptr.size(), Exceptions::RuntimeError,"ClassicalPFactory: Allocation size error (rowptr)");
-    TEUCHOS_TEST_FOR_EXCEPTION(tmp_colind.size() !=P_colind.size(), Exceptions::RuntimeError,"ClassicalPFactory: Allocation size error (colind)");
-    // FIXME:  This can be short-circuited for Tpetra, if we decide we care
-    for(LO i=0; i<(LO)Nrows+1; i++)
-      P_rowptr[i] = tmp_rowptr[i];
-    for(LO i=0; i<(LO)tmp_rowptr[Nrows]; i++)
-      P_colind[i] = tmp_colind[i];
-
-
-    // Algorithm (numeric)
-    for(LO i=0; i < (LO)Nrows; i++) {
-      if(myPointType[i] == DIRICHLET_PT) {
-        // Dirichlet points get ignored completely
+  PCrs->allocateAllValues(tmp_rowptr[Nrows], P_rowptr, P_colind, P_values);
+  TEUCHOS_TEST_FOR_EXCEPTION(tmp_rowptr.size() != P_rowptr.size(), Exceptions::RuntimeError, "ClassicalPFactory: Allocation size error (rowptr)");
+  TEUCHOS_TEST_FOR_EXCEPTION(tmp_colind.size() != P_colind.size(), Exceptions::RuntimeError, "ClassicalPFactory: Allocation size error (colind)");
+  // FIXME:  This can be short-circuited for Tpetra, if we decide we care
+  for (LO i = 0; i < (LO)Nrows + 1; i++)
+    P_rowptr[i] = tmp_rowptr[i];
+  for (LO i = 0; i < (LO)tmp_rowptr[Nrows]; i++)
+    P_colind[i] = tmp_colind[i];
+
+  // Algorithm (numeric)
+  for (LO i = 0; i < (LO)Nrows; i++) {
+    if (myPointType[i] == DIRICHLET_PT) {
+      // Dirichlet points get ignored completely
 #ifdef CMS_DEBUG
-        // DEBUG
-        printf("** A(%d,:) is a Dirichlet-Point.\n",i);
+      // DEBUG
+      printf("** A(%d,:) is a Dirichlet-Point.\n", i);
 #endif
-      }
-      else if (myPointType[i] == C_PT) {
-        // C Points get a single 1 in their row
-        P_values[P_rowptr[i]] = Teuchos::ScalarTraits<SC>::one();
+    } else if (myPointType[i] == C_PT) {
+      // C Points get a single 1 in their row
+      P_values[P_rowptr[i]] = Teuchos::ScalarTraits<SC>::one();
 #ifdef CMS_DEBUG
-        // DEBUG
-        printf("** A(%d,:) is a C-Point.\n",i);
+      // DEBUG
+      printf("** A(%d,:) is a C-Point.\n", i);
 #endif
-      }
-      else {
-        /* Trottenberg et al. (A.7.6) and (A.7.7) on p. 479 gives this:                          */
-        /* alpha_i = \frac{ \sum_{j\in N_i} a_ij^- }{ \sum_{k\in P_i} a_ik^- }                   */
-        /* beta_i  = \frac{ \sum_{j\in N_i} a_ij^+ }{ \sum_{k\in P_i} a_ik^+ }                   */
-        /* w_ik    = { - alpha_i (a_ik / a_ii),   if k\in P_i^-                                  */
-        /*           { -  beta_i (a_ik / a_ii),   if k\in P_i^+                                  */
-        ArrayView<const LO> A_indices_i, A_incides_k;
-        ArrayView<const SC> A_vals_i, A_indices_k;
-        A.getLocalRowView(i, A_indices_i, A_vals_i);
-        size_t row_start = eis_rowptr[i];
-        
-        ArrayView<LO> P_indices_i  = P_colind.view(P_rowptr[i],P_rowptr[i+1] - P_rowptr[i]);
-        ArrayView<SC> P_vals_i     = P_values.view(P_rowptr[i],P_rowptr[i+1] - P_rowptr[i]);
-        
+    } else {
+      /* Trottenberg et al. (A.7.6) and (A.7.7) on p. 479 gives this:                          */
+      /* alpha_i = \frac{ \sum_{j\in N_i} a_ij^- }{ \sum_{k\in P_i} a_ik^- }                   */
+      /* beta_i  = \frac{ \sum_{j\in N_i} a_ij^+ }{ \sum_{k\in P_i} a_ik^+ }                   */
+      /* w_ik    = { - alpha_i (a_ik / a_ii),   if k\in P_i^-                                  */
+      /*           { -  beta_i (a_ik / a_ii),   if k\in P_i^+                                  */
+      ArrayView<const LO> A_indices_i, A_incides_k;
+      ArrayView<const SC> A_vals_i, A_indices_k;
+      A.getLocalRowView(i, A_indices_i, A_vals_i);
+      size_t row_start = eis_rowptr[i];
+
+      ArrayView<LO> P_indices_i = P_colind.view(P_rowptr[i], P_rowptr[i + 1] - P_rowptr[i]);
+      ArrayView<SC> P_vals_i    = P_values.view(P_rowptr[i], P_rowptr[i + 1] - P_rowptr[i]);
+
 #ifdef CMS_DEBUG
-        // DEBUG
-        {
-          char mylabel[5]="FUCD";
-          char sw[3]="ws";
-          printf("** A(%d,:) = ",i);
-          for(LO j=0; j<(LO)A_indices_i.size(); j++){
-            printf("%6.4e(%d-%c%c) ",A_vals_i[j],A_indices_i[j],mylabel[1+myPointType[A_indices_i[j]]],sw[(int)edgeIsStrong[row_start+j]]);
-          }
-          printf("\n");
+      // DEBUG
+      {
+        char mylabel[5] = "FUCD";
+        char sw[3]      = "ws";
+        printf("** A(%d,:) = ", i);
+        for (LO j = 0; j < (LO)A_indices_i.size(); j++) {
+          printf("%6.4e(%d-%c%c) ", A_vals_i[j], A_indices_i[j], mylabel[1 + myPointType[A_indices_i[j]]], sw[(int)edgeIsStrong[row_start + j]]);
         }
+        printf("\n");
+      }
 #endif
-                      
-        SC a_ii            = SC_ZERO;
-        SC pos_numerator   = SC_ZERO, neg_numerator   = SC_ZERO;
-        SC pos_denominator = SC_ZERO, neg_denominator = SC_ZERO;
-        // Find the diagonal and compute the sum ratio
-        for(LO j=0; j<(LO)A_indices_i.size(); j++) {
-          SC a_ik = A_vals_i[j];
-          LO k = A_indices_i[j];
-          
-          // Diagonal
-          if(i == k) {
-            a_ii = a_ik;
-          }
-          // Only strong C-neighbors are in the denomintor
-          if(myPointType[k] == C_PT && edgeIsStrong[row_start + j]) {
-            if(STS::real(a_ik) > MT_ZERO) pos_denominator += a_ik;
-            else neg_denominator += a_ik;
-          }
-          
-          // All neighbors are in the numerator
-          // NOTE: As per PyAMG, this does not include the diagonal
-          if(i != k) {
-            if(STS::real(a_ik) > MT_ZERO) pos_numerator += a_ik;
-            else neg_numerator += a_ik;
-          }
+
+      SC a_ii          = SC_ZERO;
+      SC pos_numerator = SC_ZERO, neg_numerator = SC_ZERO;
+      SC pos_denominator = SC_ZERO, neg_denominator = SC_ZERO;
+      // Find the diagonal and compute the sum ratio
+      for (LO j = 0; j < (LO)A_indices_i.size(); j++) {
+        SC a_ik = A_vals_i[j];
+        LO k    = A_indices_i[j];
+
+        // Diagonal
+        if (i == k) {
+          a_ii = a_ik;
+        }
+        // Only strong C-neighbors are in the denomintor
+        if (myPointType[k] == C_PT && edgeIsStrong[row_start + j]) {
+          if (STS::real(a_ik) > MT_ZERO)
+            pos_denominator += a_ik;
+          else
+            neg_denominator += a_ik;
+        }
+
+        // All neighbors are in the numerator
+        // NOTE: As per PyAMG, this does not include the diagonal
+        if (i != k) {
+          if (STS::real(a_ik) > MT_ZERO)
+            pos_numerator += a_ik;
+          else
+            neg_numerator += a_ik;
         }
-        SC alpha = (neg_denominator == MT_ZERO) ? SC_ZERO : (neg_numerator / neg_denominator);
-        SC beta  = (pos_denominator == MT_ZERO) ? SC_ZERO : (pos_numerator / pos_denominator);
-        alpha /= -a_ii;
-        beta  /= -a_ii;
-
-        // Loop over the entries
-        for(LO p_j=0; p_j<(LO)P_indices_i.size(); p_j++){
-          LO P_col = pcol2cpoint[P_indices_i[p_j]];
-          SC a_ij = SC_ZERO;
-          
-          // Find A_ij (if it is there)
-          // FIXME: We can optimize this if we assume sorting
-          for(LO a_j =0; a_j<(LO)A_indices_i.size(); a_j++) {
-            if(A_indices_i[a_j] == P_col) {
-              a_ij = A_vals_i[a_j];
-              break;
-            }
+      }
+      SC alpha = (neg_denominator == MT_ZERO) ? SC_ZERO : (neg_numerator / neg_denominator);
+      SC beta  = (pos_denominator == MT_ZERO) ? SC_ZERO : (pos_numerator / pos_denominator);
+      alpha /= -a_ii;
+      beta /= -a_ii;
+
+      // Loop over the entries
+      for (LO p_j = 0; p_j < (LO)P_indices_i.size(); p_j++) {
+        LO P_col = pcol2cpoint[P_indices_i[p_j]];
+        SC a_ij  = SC_ZERO;
+
+        // Find A_ij (if it is there)
+        // FIXME: We can optimize this if we assume sorting
+        for (LO a_j = 0; a_j < (LO)A_indices_i.size(); a_j++) {
+          if (A_indices_i[a_j] == P_col) {
+            a_ij = A_vals_i[a_j];
+            break;
           }
-          SC w_ij = (STS::real(a_ij) < 0 ) ? (alpha * a_ij) : (beta * a_ij);
+        }
+        SC w_ij = (STS::real(a_ij) < 0) ? (alpha * a_ij) : (beta * a_ij);
 #ifdef CMS_DEBUG
-          SC alpha_or_beta = (STS::real(a_ij) < 0 ) ? alpha : beta;
-          printf("P(%d,%d/%d) =  - %6.4e  * %6.4e  = %6.4e\n",i,P_indices_i[p_j],pcol2cpoint[P_indices_i[p_j]],alpha_or_beta,a_ij,w_ij);
+        SC alpha_or_beta = (STS::real(a_ij) < 0) ? alpha : beta;
+        printf("P(%d,%d/%d) =  - %6.4e  * %6.4e  = %6.4e\n", i, P_indices_i[p_j], pcol2cpoint[P_indices_i[p_j]], alpha_or_beta, a_ij, w_ij);
 #endif
-          P_vals_i[p_j] = w_ij;
-        }//end for A_indices_i
-      }//end else C_PT
-    }//end for Numrows
-
-    // Finish up
-    PCrs->setAllValues(P_rowptr, P_colind, P_values);
-    PCrs->expertStaticFillComplete(/*domain*/coarseDomainMap, /*range*/A.getDomainMap());
+        P_vals_i[p_j] = w_ij;
+      }  // end for A_indices_i
+    }    // end else C_PT
+  }      // end for Numrows
+
+  // Finish up
+  PCrs->setAllValues(P_rowptr, P_colind, P_values);
+  PCrs->expertStaticFillComplete(/*domain*/ coarseDomainMap, /*range*/ A.getDomainMap());
 }
 
-
 /* ************************************************************************* */
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-Coarsen_Ext_Plus_I(const Matrix & A,const RCP<const Matrix> & Aghost, const GraphBase & graph,  RCP<const Map> & coarseColMap, RCP<const Map> & coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO> & myPointType, const Teuchos::ArrayView<const LO> & myPointType_ghost, const Teuchos::Array<LO> & cpoint2pcol, const Teuchos::Array<LO> & pcol2cpoint, Teuchos::Array<size_t> & eis_rowptr, Teuchos::Array<bool> & edgeIsStrong, RCP<LocalOrdinalVector> & BlockNumber, RCP<Matrix> & P) const {
-
-    /* ============================================================= */
-    /* Phase 3 : Extended+i Interpolation                            */
-    /* De Sterck, Falgout, Nolting and Yang. "Distance-two           */
-    /* interpolation for parallel algebraic multigrid", NLAA 2008    */
-    /* 15:115-139                                                    */
-    /* ============================================================= */
-    /* Definitions:                                                        */
-    /* F = F-points                                                        */
-    /* C = C-points                                                        */
-    /* N_i = non-zero neighbors of node i                                  */
-    /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
-    /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
-    /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
-    /* N_i^w = N_i\ (F_i^s \cup C_i^s) [weak neighbors of i]               */
-    /*         This guy has a typo.  The paper had a \cap instead of \cup  */
-    /*         I would note that this set can contain both F-points and    */
-    /*         C-points.  They're just weak neighbors of this guy.         */
-    /*         Note that N_i^w \cup F_i^s \cup C_i^s = N_i by construction */
-
-    /* \hat{C}_i = C_i \cup (\bigcup_{j\inF_i^s} C_j)                      */
-    /*         [C-neighbors and C-neighbors of strong F-neighbors of i]    */
-    /*                                                                     */
-
-    /* \bar{a}_ij = {    0, if sign(a_ij) == sign(a_ii)                    */
-    /*              { a_ij, otherwise                                      */
-
-
-    /* Rewritten Equation (19) on p. 123                                   */
-    /* f_ik = \frac{\bar{a}_kj}{\sum{l\in \hat{C}_i\cup {i}} \bar{a}_kl    */
-    /* w_ij = -\tilde{a}_ii^{-1} (a_ij + \sum_{k\inF_i^s} a_ik f_ik        */
-    /*         for j in \hat{C}_i                                          */
-    
-    /* Rewritten Equation (20) on p. 124 [for the lumped diagonal]                                  */
-    /* g_ik = \frac{\bar{a}_ki}{\sum{l\in \hat{C}_i\cup {i}} \bar{a}_kl                             */
-    /* \tilde{a}_ii = a_ii + \sum_{n\inN_i^w\setminus \hat{C}_i} a_in + \sum_{k\inF_i^s} a_ik g_ik  */
-    TEUCHOS_TEST_FOR_EXCEPTION(1,std::runtime_error,"ClassicalPFactory: Ext+i not implemented");
-
+    Coarsen_Ext_Plus_I(const Matrix& A, const RCP<const Matrix>& Aghost, const GraphBase& graph, RCP<const Map>& coarseColMap, RCP<const Map>& coarseDomainMap, LO num_c_points, LO num_f_points, const Teuchos::ArrayView<const LO>& myPointType, const Teuchos::ArrayView<const LO>& myPointType_ghost, const Teuchos::Array<LO>& cpoint2pcol, const Teuchos::Array<LO>& pcol2cpoint, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong, RCP<LocalOrdinalVector>& BlockNumber, RCP<Matrix>& P) const {
+  /* ============================================================= */
+  /* Phase 3 : Extended+i Interpolation                            */
+  /* De Sterck, Falgout, Nolting and Yang. "Distance-two           */
+  /* interpolation for parallel algebraic multigrid", NLAA 2008    */
+  /* 15:115-139                                                    */
+  /* ============================================================= */
+  /* Definitions:                                                        */
+  /* F = F-points                                                        */
+  /* C = C-points                                                        */
+  /* N_i = non-zero neighbors of node i                                  */
+  /* S_i = {j\in N_i | j strongly influences i } [strong neighbors of i] */
+  /* F_i^s = F \cap S_i [strong F-neighbors of i]                        */
+  /* C_i^s = C \cap S_i [strong C-neighbors of i]                        */
+  /* N_i^w = N_i\ (F_i^s \cup C_i^s) [weak neighbors of i]               */
+  /*         This guy has a typo.  The paper had a \cap instead of \cup  */
+  /*         I would note that this set can contain both F-points and    */
+  /*         C-points.  They're just weak neighbors of this guy.         */
+  /*         Note that N_i^w \cup F_i^s \cup C_i^s = N_i by construction */
+
+  /* \hat{C}_i = C_i \cup (\bigcup_{j\inF_i^s} C_j)                      */
+  /*         [C-neighbors and C-neighbors of strong F-neighbors of i]    */
+  /*                                                                     */
+
+  /* \bar{a}_ij = {    0, if sign(a_ij) == sign(a_ii)                    */
+  /*              { a_ij, otherwise                                      */
+
+  /* Rewritten Equation (19) on p. 123                                   */
+  /* f_ik = \frac{\bar{a}_kj}{\sum{l\in \hat{C}_i\cup {i}} \bar{a}_kl    */
+  /* w_ij = -\tilde{a}_ii^{-1} (a_ij + \sum_{k\inF_i^s} a_ik f_ik        */
+  /*         for j in \hat{C}_i                                          */
+
+  /* Rewritten Equation (20) on p. 124 [for the lumped diagonal]                                  */
+  /* g_ik = \frac{\bar{a}_ki}{\sum{l\in \hat{C}_i\cup {i}} \bar{a}_kl                             */
+  /* \tilde{a}_ii = a_ii + \sum_{n\inN_i^w\setminus \hat{C}_i} a_in + \sum_{k\inF_i^s} a_ik g_ik  */
+  TEUCHOS_TEST_FOR_EXCEPTION(1, std::runtime_error, "ClassicalPFactory: Ext+i not implemented");
 }
 
-
-
-
 /* ************************************************************************* */
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-GenerateStrengthFlags(const Matrix & A,const GraphBase & graph, Teuchos::Array<size_t> & eis_rowptr,Teuchos::Array<bool> & edgeIsStrong) const {
+    GenerateStrengthFlags(const Matrix& A, const GraphBase& graph, Teuchos::Array<size_t>& eis_rowptr, Teuchos::Array<bool>& edgeIsStrong) const {
   // To make this easier, we'll create a bool array equal to the nnz in the matrix
   // so we know whether each edge is strong or not.  This will save us a bunch of
   // trying to match the graph and matrix later
   size_t Nrows = A.getLocalNumRows();
-  eis_rowptr.resize(Nrows+1);
+  eis_rowptr.resize(Nrows + 1);
 
-  if(edgeIsStrong.size() == 0) {
+  if (edgeIsStrong.size() == 0) {
     // Preferred
-    edgeIsStrong.resize(A.getLocalNumEntries(),false);
+    edgeIsStrong.resize(A.getLocalNumEntries(), false);
+  } else {
+    edgeIsStrong.resize(A.getLocalNumEntries(), false);
+    edgeIsStrong.assign(A.getLocalNumEntries(), false);
   }
-  else {
-    edgeIsStrong.resize(A.getLocalNumEntries(),false);
-    edgeIsStrong.assign(A.getLocalNumEntries(),false);
-  }
-  
+
   eis_rowptr[0] = 0;
-  for (LO i=0; i<(LO)Nrows; i++) {
+  for (LO i = 0; i < (LO)Nrows; i++) {
     LO rowstart = eis_rowptr[i];
     ArrayView<const LO> A_indices;
     ArrayView<const SC> A_values;
     A.getLocalRowView(i, A_indices, A_values);
-    LO A_size = (LO) A_indices.size();
+    LO A_size = (LO)A_indices.size();
 
     ArrayView<const LO> G_indices = graph.getNeighborVertices(i);
-    LO G_size = (LO) G_indices.size();
-    
+    LO G_size                     = (LO)G_indices.size();
+
     // Both of these guys should be in the same (sorted) order, but let's check
-    bool is_ok=true;
-    for(LO j=0; j<A_size-1; j++)
-      if (A_indices[j] >= A_indices[j+1]) { is_ok=false; break;}
-    for(LO j=0; j<G_size-1; j++)
-      if (G_indices[j] >= G_indices[j+1]) { is_ok=false; break;}
-    TEUCHOS_TEST_FOR_EXCEPTION(!is_ok, Exceptions::RuntimeError,"ClassicalPFactory: Exected A and Graph to be sorted");
-    
+    bool is_ok = true;
+    for (LO j = 0; j < A_size - 1; j++)
+      if (A_indices[j] >= A_indices[j + 1]) {
+        is_ok = false;
+        break;
+      }
+    for (LO j = 0; j < G_size - 1; j++)
+      if (G_indices[j] >= G_indices[j + 1]) {
+        is_ok = false;
+        break;
+      }
+    TEUCHOS_TEST_FOR_EXCEPTION(!is_ok, Exceptions::RuntimeError, "ClassicalPFactory: Exected A and Graph to be sorted");
+
     // Now cycle through and set the flags - if the edge is in G it is strong
-    for(LO g_idx=0, a_idx=0; g_idx < G_size; g_idx++) {
+    for (LO g_idx = 0, a_idx = 0; g_idx < G_size; g_idx++) {
       LO col = G_indices[g_idx];
       while (A_indices[a_idx] != col && a_idx < A_size) a_idx++;
-      if(a_idx == A_size) {is_ok=false;break;}
-      edgeIsStrong[rowstart+a_idx] = true;
+      if (a_idx == A_size) {
+        is_ok = false;
+        break;
+      }
+      edgeIsStrong[rowstart + a_idx] = true;
     }
 
-    eis_rowptr[i+1] = eis_rowptr[i] + A_size;
+    eis_rowptr[i + 1] = eis_rowptr[i] + A_size;
   }
 }
-   
 
 /* ************************************************************************* */
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void ClassicalPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-GhostCoarseMap(const Matrix &A, const Import & Importer, const ArrayRCP<const LO> myPointType, const RCP<const Map> & coarseMap, RCP<const Map> & coarseColMap) const {
-  const point_type C_PT = ClassicalMapFactory::C_PT;
-  const GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
+    GhostCoarseMap(const Matrix& A, const Import& Importer, const ArrayRCP<const LO> myPointType, const RCP<const Map>& coarseMap, RCP<const Map>& coarseColMap) const {
+  const point_type C_PT                = ClassicalMapFactory::C_PT;
+  const GO GO_INVALID                  = Teuchos::OrdinalTraits<GO>::invalid();
   RCP<GlobalOrdinalVector> d_coarseIds = GlobalOrdinalVectorFactory::Build(A.getRowMap());
-  ArrayRCP<GO> d_data = d_coarseIds->getDataNonConst(0);
-  LO ct=0;
-      
-  for(LO i=0; i<(LO)d_data.size(); i++) {
-    if(myPointType[i] == C_PT) {
+  ArrayRCP<GO> d_data                  = d_coarseIds->getDataNonConst(0);
+  LO ct                                = 0;
+
+  for (LO i = 0; i < (LO)d_data.size(); i++) {
+    if (myPointType[i] == C_PT) {
       d_data[i] = coarseMap->getGlobalElement(ct);
       ct++;
-    }
-    else
+    } else
       d_data[i] = GO_INVALID;
   }
-  
+
   // Ghost this guy
   RCP<GlobalOrdinalVector> c_coarseIds = GlobalOrdinalVectorFactory::Build(A.getColMap());
-  c_coarseIds->doImport(*d_coarseIds,Importer,Xpetra::INSERT);
-  
+  c_coarseIds->doImport(*d_coarseIds, Importer, Xpetra::INSERT);
+
   // If we assume that A is in Aztec ordering, then any subset of A's unknowns will
   // be in Aztec ordering as well, which means we can just condense these guys down
   // Overallocate, count and view
   ArrayRCP<GO> c_data = c_coarseIds->getDataNonConst(0);
-  
+
   Array<GO> c_gids(c_data.size());
-  LO count=0;
-  
-  for(LO i=0; i<(LO)c_data.size(); i++) {
-    if(c_data[i] != GO_INVALID) {
+  LO count = 0;
+
+  for (LO i = 0; i < (LO)c_data.size(); i++) {
+    if (c_data[i] != GO_INVALID) {
       c_gids[count] = c_data[i];
       count++;
     }
   }
   // FIXME: Assumes scalar PDE
   std::vector<size_t> stridingInfo_(1);
-  stridingInfo_[0]=1;
+  stridingInfo_[0]   = 1;
   GO domainGIDOffset = 0;
-  
+
   coarseColMap = StridedMapFactory::Build(coarseMap->lib(),
                                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                          c_gids.view(0,count),
+                                          c_gids.view(0, count),
                                           coarseMap->getIndexBase(),
                                           stridingInfo_,
                                           coarseMap->getComm(),
                                           domainGIDOffset);
-  
 }
 
-
-} //namespace MueLu
-
-
+}  // namespace MueLu
 
 #define MUELU_CLASSICALPFACTORY_SHORT
-#endif // MUELU_CLASSICALPFACTORY_DEF_HPP
-
- 
+#endif  // MUELU_CLASSICALPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_decl.hpp
index a979de92ee88..10ae4452fafc 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_decl.hpp
@@ -53,62 +53,60 @@
 
 namespace MueLu {
 
-  /*!
-    @class ConstraintFactory class.
-    @brief Factory for building the constraint operator
+/*!
+  @class ConstraintFactory class.
+  @brief Factory for building the constraint operator
 
-    Factory for creating the constraint operator.
+  Factory for creating the constraint operator.
 
-    @ingroup MueLuTransferClasses
-  */
+  @ingroup MueLuTransferClasses
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ConstraintFactory : public TwoLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ConstraintFactory : public TwoLevelFactoryBase {
 #undef MUELU_CONSTRAINTFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
-      User can supply a factory for generating the nonzero pattern. The nullspace vectors (both fine and coarse) will
-      be taken from the corresponding level factories
-      */
-    ConstraintFactory() { }
+  /*! @brief Constructor.
+    User can supply a factory for generating the nonzero pattern. The nullspace vectors (both fine and coarse) will
+    be taken from the corresponding level factories
+    */
+  ConstraintFactory() {}
 
-    //! Destructor.
-    virtual ~ConstraintFactory() { }
+  //! Destructor.
+  virtual ~ConstraintFactory() {}
 
-    //@}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
-    //! @name Build methods.
-    //@{
+  //@}
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
-
-      Builds Constraint and returns it in <tt>coarseLevel</tt>.
-      */
-    void Build(Level & fineLevel, Level& coarseLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
-  }; // class ConstraintFactory
+    Builds Constraint and returns it in <tt>coarseLevel</tt>.
+    */
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
+  //@}
+};  // class ConstraintFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CONSTRAINTFACTORY_SHORT
-#endif // MUELU_CONSTRAINTFACTORY_DECL_HPP
+#endif  // MUELU_CONSTRAINTFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_def.hpp
index 61da86b64f8a..757a958b447c 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_ConstraintFactory_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_CONSTRAINTFACTORY_DEF_HPP
 #define MUELU_CONSTRAINTFACTORY_DEF_HPP
 
-
 #include "MueLu_ConstraintFactory_decl.hpp"
 
 #include "MueLu_Constraint.hpp"
@@ -54,39 +53,38 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("FineNullspace",    Teuchos::null, "Generating factory for the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("CoarseNullspace",  Teuchos::null, "Generating factory for the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("Ppattern",         Teuchos::null, "Generating factory for the nonzero pattern");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    return validParamList;
-  }
+  validParamList->set<RCP<const FactoryBase> >("FineNullspace", Teuchos::null, "Generating factory for the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("CoarseNullspace", Teuchos::null, "Generating factory for the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("Ppattern", Teuchos::null, "Generating factory for the nonzero pattern");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level& coarseLevel) const {
-    Input(fineLevel,   "Nullspace", "FineNullspace");
-    Input(coarseLevel, "Nullspace", "CoarseNullspace");
-    Input(coarseLevel, "Ppattern");
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Constraint", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "Nullspace", "FineNullspace");
+  Input(coarseLevel, "Nullspace", "CoarseNullspace");
+  Input(coarseLevel, "Ppattern");
+}
 
-    RCP<MultiVector> fineNullspace   = Get< RCP<MultiVector> >(fineLevel,   "Nullspace", "FineNullspace");
-    RCP<MultiVector> coarseNullspace = Get< RCP<MultiVector> >(coarseLevel, "Nullspace", "CoarseNullspace");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ConstraintFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Constraint", coarseLevel);
 
-    RCP<Constraint> constraint(new Constraint);
-    constraint->Setup(*fineNullspace, *coarseNullspace,
-                       Get< RCP<const CrsGraph> >(coarseLevel, "Ppattern"));
+  RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace", "FineNullspace");
+  RCP<MultiVector> coarseNullspace = Get<RCP<MultiVector> >(coarseLevel, "Nullspace", "CoarseNullspace");
 
-    Set(coarseLevel, "Constraint", constraint);
-  }
+  RCP<Constraint> constraint(new Constraint);
+  constraint->Setup(*fineNullspace, *coarseNullspace,
+                    Get<RCP<const CrsGraph> >(coarseLevel, "Ppattern"));
 
+  Set(coarseLevel, "Constraint", constraint);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_CONSTRAINTFACTORY_DEF_HPP
+#endif  // MUELU_CONSTRAINTFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_decl.hpp
index 66759ad498ff..431ff5cc0579 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_decl.hpp
@@ -57,92 +57,91 @@
 
 namespace MueLu {
 
-  //! @brief Constraint space information for the potential prolongator
-  /*!
-    This class implements an idea of the constrained space.  In energy
-    minimization, constrained space is used simultaneously with the iterative
-    method to construct the final prolongator. The space has two different
-    constraints.
-
-    \section pattern_constraint Nonzero pattern constraint
-
-    Nonzero pattern constraint means that the final prolongator must have the
-    provided nonzero pattern. This is achieved on each step of the iterative
-    method by restricting the graph of the temporary prolongator to the desired
-    pattern. It is implemented in the Apply function.  \note We do not update
-    the graph of the provided temporary prolongator as this is a very expensive
-    procedure. Rather, we extract its values and replace some of the values of
-    the matrix with the correct graph.
-
-    \section space_constraint Coarse space approximation constraint
-
-    Generally, the coarse space constraint is presented by some matrix (X or Q)
-    (see, for instance, the article by Mandel, Brezina and Vanek '99. It is
-    well known that this matrix can be permuted to have a block diagonal form,
-    where each block corresponds to a row in the prolongator. Specifically, let
-    P be the prolongator, and Q be the constraint matrix. Then the constraint
-    is generally written as \f$Q P = B,\f$ where B is the fine nullspace
-    multivector. Q is a block diagonal matrix, \f$Q = diag(Q_1, ..., Q_n)\f$, where n
-    is the number of rows in P. Each block Q_i is of size NSDim x nnz_i, where
-    NSDim is the number of fine nullspace vectors, and nnz_i is the number of
-    nonzero elements in the i-th row of P.
-
-    To constrain the potential prolongator (with correct sparsity pattern, i.e.
-    after the application of the nonzero pattern constraint), one updates its
-    values as
-
-                \f[P = P - Q^H(QQ^H)^{-1}QP.\f]
-
-    Because of the block diagonal form of Q, this can be done row-by-row.
-    \note For efficiency reasons, we store \f[(QQ^H)^{-1}\f] in the XXtInv_
-    array. These matrices are dense, but have small size (NSDim x NSDim).
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Constraint : public BaseClass {
+//! @brief Constraint space information for the potential prolongator
+/*!
+  This class implements an idea of the constrained space.  In energy
+  minimization, constrained space is used simultaneously with the iterative
+  method to construct the final prolongator. The space has two different
+  constraints.
+
+  \section pattern_constraint Nonzero pattern constraint
+
+  Nonzero pattern constraint means that the final prolongator must have the
+  provided nonzero pattern. This is achieved on each step of the iterative
+  method by restricting the graph of the temporary prolongator to the desired
+  pattern. It is implemented in the Apply function.  \note We do not update
+  the graph of the provided temporary prolongator as this is a very expensive
+  procedure. Rather, we extract its values and replace some of the values of
+  the matrix with the correct graph.
+
+  \section space_constraint Coarse space approximation constraint
+
+  Generally, the coarse space constraint is presented by some matrix (X or Q)
+  (see, for instance, the article by Mandel, Brezina and Vanek '99. It is
+  well known that this matrix can be permuted to have a block diagonal form,
+  where each block corresponds to a row in the prolongator. Specifically, let
+  P be the prolongator, and Q be the constraint matrix. Then the constraint
+  is generally written as \f$Q P = B,\f$ where B is the fine nullspace
+  multivector. Q is a block diagonal matrix, \f$Q = diag(Q_1, ..., Q_n)\f$, where n
+  is the number of rows in P. Each block Q_i is of size NSDim x nnz_i, where
+  NSDim is the number of fine nullspace vectors, and nnz_i is the number of
+  nonzero elements in the i-th row of P.
+
+  To constrain the potential prolongator (with correct sparsity pattern, i.e.
+  after the application of the nonzero pattern constraint), one updates its
+  values as
+
+              \f[P = P - Q^H(QQ^H)^{-1}QP.\f]
+
+  Because of the block diagonal form of Q, this can be done row-by-row.
+  \note For efficiency reasons, we store \f[(QQ^H)^{-1}\f] in the XXtInv_
+  array. These matrices are dense, but have small size (NSDim x NSDim).
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Constraint : public BaseClass {
 #undef MUELU_CONSTRAINT_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-
+ public:
   /*!
     @class Constraint class.
     @brief Class which contains the constraint space details
     */
 
-    //! @name Setup methods.
-    //@{
+  //! @name Setup methods.
+  //@{
 
-    /*! Setup constraint.
-        \param B -- Fine nullspace vectors
-        \param Bc -- Coarse nullspace vectors
-        \param Ppattern -- Nonzero sparsity pattern for the prolongator
-     */
-    void Setup(const MultiVector& B, const MultiVector& Bc, RCP<const CrsGraph> Ppattern);
+  /*! Setup constraint.
+      \param B -- Fine nullspace vectors
+      \param Bc -- Coarse nullspace vectors
+      \param Ppattern -- Nonzero sparsity pattern for the prolongator
+   */
+  void Setup(const MultiVector& B, const MultiVector& Bc, RCP<const CrsGraph> Ppattern);
 
-    //@}
+  //@}
 
-    //! @name Apply methods.
-    //@{
+  //! @name Apply methods.
+  //@{
 
-    //! Apply constraint.
-    void Apply(const Matrix& P, Matrix& Projected) const;
+  //! Apply constraint.
+  void Apply(const Matrix& P, Matrix& Projected) const;
 
-    //@}
+  //@}
 
-    RCP<const CrsGraph> GetPattern() const {
-      return Ppattern_;
-    }
+  RCP<const CrsGraph> GetPattern() const {
+    return Ppattern_;
+  }
 
-  private:
-    RCP<MultiVector>    X_;                                     //!< Overlapped coarse nullspace
-    RCP<const CrsGraph> Ppattern_;                              //!< Nonzero sparsity pattern
-    ArrayRCP<Teuchos::SerialDenseMatrix<LO,SC> > XXtInv_;       //!< Array storing \f$(Q_i Q_i^H)^{-1}\f$
-  };
+ private:
+  RCP<MultiVector> X_;                                    //!< Overlapped coarse nullspace
+  RCP<const CrsGraph> Ppattern_;                          //!< Nonzero sparsity pattern
+  ArrayRCP<Teuchos::SerialDenseMatrix<LO, SC> > XXtInv_;  //!< Array storing \f$(Q_i Q_i^H)^{-1}\f$
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CONSTRAINT_SHORT
-#endif // MUELU_CONSTRAINT_DECL_HPP
+#endif  // MUELU_CONSTRAINT_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_def.hpp
index 802098b9feb2..17bfaecb4d8c 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_Constraint_def.hpp
@@ -63,170 +63,169 @@
 
 #include "MueLu_Constraint_decl.hpp"
 
-
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Constraint<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(const MultiVector& /* B */, const MultiVector& Bc, RCP<const CrsGraph> Ppattern) {
-    const size_t NSDim = Bc.getNumVectors();
-
-    Ppattern_ = Ppattern;
-
-    size_t numRows = Ppattern_->getLocalNumRows();
-    XXtInv_.resize(numRows);
-
-    RCP<const Import> importer = Ppattern_->getImporter();
-
-    X_ = MultiVectorFactory::Build(Ppattern_->getColMap(), NSDim);
-    if (!importer.is_null())
-      X_->doImport(Bc, *importer, Xpetra::INSERT);
-    else
-      *X_ = Bc;
-
-    std::vector<const SC*> Xval(NSDim);
-    for (size_t j = 0; j < NSDim; j++)
-      Xval[j] = X_->getData(j).get();
-
-    SC zero = Teuchos::ScalarTraits<SC>::zero();
-    SC one  = Teuchos::ScalarTraits<SC>::one();
-
-    Teuchos::BLAS  <LO,SC> blas;
-    Teuchos::LAPACK<LO,SC> lapack;
-    LO lwork = 3*NSDim;
-    ArrayRCP<LO> IPIV(NSDim);
-    ArrayRCP<SC> WORK(lwork);
-
-    for (size_t i = 0; i < numRows; i++) {
-      Teuchos::ArrayView<const LO> indices;
-      Ppattern_->getLocalRowView(i, indices);
-
-      size_t nnz = indices.size();
-
-      XXtInv_[i] = Teuchos::SerialDenseMatrix<LO,SC>(NSDim, NSDim, false/*zeroOut*/);
-      Teuchos::SerialDenseMatrix<LO,SC>& XXtInv = XXtInv_[i];
-
-      if (NSDim == 1) {
-        SC d = zero;
-        for (size_t j = 0; j < nnz; j++)
-          d += Xval[0][indices[j]] * Xval[0][indices[j]];
-        XXtInv(0,0) = one/d;
-
-      } else {
-        Teuchos::SerialDenseMatrix<LO,SC> locX(NSDim, nnz, false/*zeroOut*/);
-        for (size_t j = 0; j < nnz; j++)
-          for (size_t k = 0; k < NSDim; k++)
-            locX(k,j) = Xval[k][indices[j]];
-
-        // XXtInv_ = (locX*locX^T)^{-1}
-        blas.GEMM(Teuchos::NO_TRANS, Teuchos::CONJ_TRANS, NSDim, NSDim, nnz,
-                   one,   locX.values(),   locX.stride(),
-                          locX.values(),   locX.stride(),
-                  zero, XXtInv.values(), XXtInv.stride());
-
-        LO info;
-        // Compute LU factorization using partial pivoting with row exchanges
-        lapack.GETRF(NSDim, NSDim, XXtInv.values(), XXtInv.stride(), IPIV.get(), &info);
-        // Use the computed factorization to compute the inverse
-        lapack.GETRI(NSDim, XXtInv.values(), XXtInv.stride(), IPIV.get(), WORK.get(), lwork, &info);
-      }
-    }
-  }
-
-  //! \note We assume that the graph of Projected is the same as Ppattern_
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Constraint<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(const Matrix& P, Matrix& Projected) const {
-    // We check only row maps. Column may be different.
-    TEUCHOS_TEST_FOR_EXCEPTION(!P.getRowMap()->isSameAs(*Projected.getRowMap()), Exceptions::Incompatible,
-                               "Row maps are incompatible");
-    const size_t NSDim   = X_->getNumVectors();
-    const size_t numRows = P.getLocalNumRows();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Constraint<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Setup(const MultiVector& /* B */, const MultiVector& Bc, RCP<const CrsGraph> Ppattern) {
+  const size_t NSDim = Bc.getNumVectors();
 
-    const Map& colMap  = *P.getColMap();
-    const Map& PColMap = *Projected.getColMap();
+  Ppattern_ = Ppattern;
 
-    Projected.resumeFill();
+  size_t numRows = Ppattern_->getLocalNumRows();
+  XXtInv_.resize(numRows);
 
-    Teuchos::ArrayView<const LO> indices, pindices;
-    Teuchos::ArrayView<const SC> values,  pvalues;
-    Teuchos::Array<SC> valuesAll(colMap.getLocalNumElements()), newValues;
+  RCP<const Import> importer = Ppattern_->getImporter();
 
-    LO invalid = Teuchos::OrdinalTraits<LO>::invalid();
-    LO oneLO   = Teuchos::OrdinalTraits<LO>::one();
-    SC zero    = Teuchos::ScalarTraits<SC> ::zero();
-    SC one     = Teuchos::ScalarTraits<SC> ::one();
+  X_ = MultiVectorFactory::Build(Ppattern_->getColMap(), NSDim);
+  if (!importer.is_null())
+    X_->doImport(Bc, *importer, Xpetra::INSERT);
+  else
+    *X_ = Bc;
 
-    std::vector<const SC*> Xval(NSDim);
-    for (size_t j = 0; j < NSDim; j++)
-      Xval[j] = X_->getData(j).get();
+  std::vector<const SC*> Xval(NSDim);
+  for (size_t j = 0; j < NSDim; j++)
+    Xval[j] = X_->getData(j).get();
 
-    for (size_t i = 0; i < numRows; i++) {
-      P        .getLocalRowView(i,  indices,  values);
-      Projected.getLocalRowView(i, pindices, pvalues);
+  SC zero = Teuchos::ScalarTraits<SC>::zero();
+  SC one  = Teuchos::ScalarTraits<SC>::one();
 
-      size_t nnz  = indices.size();     // number of nonzeros in the supplied matrix
-      size_t pnnz = pindices.size();    // number of nonzeros in the constrained matrix
+  Teuchos::BLAS<LO, SC> blas;
+  Teuchos::LAPACK<LO, SC> lapack;
+  LO lwork = 3 * NSDim;
+  ArrayRCP<LO> IPIV(NSDim);
+  ArrayRCP<SC> WORK(lwork);
 
-      newValues.resize(pnnz);
+  for (size_t i = 0; i < numRows; i++) {
+    Teuchos::ArrayView<const LO> indices;
+    Ppattern_->getLocalRowView(i, indices);
 
-      // Step 1: fix stencil
-      // Projected *must* already have the correct stencil
+    size_t nnz = indices.size();
 
-      // Step 2: copy correct stencil values
-      // The algorithm is very similar to the one used in the calculation of
-      // Frobenius dot product, see src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp
+    XXtInv_[i]                                 = Teuchos::SerialDenseMatrix<LO, SC>(NSDim, NSDim, false /*zeroOut*/);
+    Teuchos::SerialDenseMatrix<LO, SC>& XXtInv = XXtInv_[i];
 
-      // NOTE: using extra array allows us to skip the search among indices
+    if (NSDim == 1) {
+      SC d = zero;
       for (size_t j = 0; j < nnz; j++)
-        valuesAll[indices[j]] = values[j];
-      for (size_t j = 0; j < pnnz; j++) {
-        LO ind = colMap.getLocalElement(PColMap.getGlobalElement(pindices[j])); // FIXME: we could do that before the full loop just once
-        if (ind != invalid)
-          // index indices[j] is part of template, copy corresponding value
-          newValues[j] = valuesAll[ind];
-        else
-          newValues[j] = zero;
-      }
-      for (size_t j = 0; j < nnz; j++)
-        valuesAll[indices[j]] = zero;
-
-      // Step 3: project to the space
-      Teuchos::SerialDenseMatrix<LO,SC>& XXtInv = XXtInv_[i];
+        d += Xval[0][indices[j]] * Xval[0][indices[j]];
+      XXtInv(0, 0) = one / d;
 
-      Teuchos::SerialDenseMatrix<LO,SC> locX(NSDim, pnnz, false);
-      for (size_t j = 0; j < pnnz; j++)
+    } else {
+      Teuchos::SerialDenseMatrix<LO, SC> locX(NSDim, nnz, false /*zeroOut*/);
+      for (size_t j = 0; j < nnz; j++)
         for (size_t k = 0; k < NSDim; k++)
-          locX(k,j) = Xval[k][pindices[j]];
-
-      Teuchos::SerialDenseVector<LO,SC> val(pnnz, false), val1(NSDim, false), val2(NSDim, false);
-      for (size_t j = 0; j < pnnz; j++)
-        val[j] = newValues[j];
+          locX(k, j) = Xval[k][indices[j]];
 
-      Teuchos::BLAS<LO,SC> blas;
-      // val1 = locX * val;
-      blas.GEMV(Teuchos::NO_TRANS, NSDim, pnnz,
+      // XXtInv_ = (locX*locX^T)^{-1}
+      blas.GEMM(Teuchos::NO_TRANS, Teuchos::CONJ_TRANS, NSDim, NSDim, nnz,
                 one, locX.values(), locX.stride(),
-                val.values(), oneLO,
-                zero, val1.values(), oneLO);
-      // val2 = XXtInv * val1
-      blas.GEMV(Teuchos::NO_TRANS, NSDim, NSDim,
-                one, XXtInv.values(), XXtInv.stride(),
-                val1.values(), oneLO,
-                zero,   val2.values(), oneLO);
-      // val = X^T * val2
-      blas.GEMV(Teuchos::CONJ_TRANS, NSDim, pnnz,
-                one, locX.values(), locX.stride(),
-                val2.values(), oneLO,
-                zero,  val.values(), oneLO);
-
-      for (size_t j = 0; j < pnnz; j++)
-        newValues[j] -= val[j];
-
-      Projected.replaceLocalValues(i, pindices, newValues);
+                locX.values(), locX.stride(),
+                zero, XXtInv.values(), XXtInv.stride());
+
+      LO info;
+      // Compute LU factorization using partial pivoting with row exchanges
+      lapack.GETRF(NSDim, NSDim, XXtInv.values(), XXtInv.stride(), IPIV.get(), &info);
+      // Use the computed factorization to compute the inverse
+      lapack.GETRI(NSDim, XXtInv.values(), XXtInv.stride(), IPIV.get(), WORK.get(), lwork, &info);
     }
-
-    Projected.fillComplete(Projected.getDomainMap(), Projected.getRangeMap()); //FIXME: maps needed?
   }
+}
+
+//! \note We assume that the graph of Projected is the same as Ppattern_
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Constraint<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Apply(const Matrix& P, Matrix& Projected) const {
+  // We check only row maps. Column may be different.
+  TEUCHOS_TEST_FOR_EXCEPTION(!P.getRowMap()->isSameAs(*Projected.getRowMap()), Exceptions::Incompatible,
+                             "Row maps are incompatible");
+  const size_t NSDim   = X_->getNumVectors();
+  const size_t numRows = P.getLocalNumRows();
+
+  const Map& colMap  = *P.getColMap();
+  const Map& PColMap = *Projected.getColMap();
+
+  Projected.resumeFill();
+
+  Teuchos::ArrayView<const LO> indices, pindices;
+  Teuchos::ArrayView<const SC> values, pvalues;
+  Teuchos::Array<SC> valuesAll(colMap.getLocalNumElements()), newValues;
+
+  LO invalid = Teuchos::OrdinalTraits<LO>::invalid();
+  LO oneLO   = Teuchos::OrdinalTraits<LO>::one();
+  SC zero    = Teuchos::ScalarTraits<SC>::zero();
+  SC one     = Teuchos::ScalarTraits<SC>::one();
+
+  std::vector<const SC*> Xval(NSDim);
+  for (size_t j = 0; j < NSDim; j++)
+    Xval[j] = X_->getData(j).get();
+
+  for (size_t i = 0; i < numRows; i++) {
+    P.getLocalRowView(i, indices, values);
+    Projected.getLocalRowView(i, pindices, pvalues);
+
+    size_t nnz  = indices.size();   // number of nonzeros in the supplied matrix
+    size_t pnnz = pindices.size();  // number of nonzeros in the constrained matrix
+
+    newValues.resize(pnnz);
+
+    // Step 1: fix stencil
+    // Projected *must* already have the correct stencil
+
+    // Step 2: copy correct stencil values
+    // The algorithm is very similar to the one used in the calculation of
+    // Frobenius dot product, see src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp
+
+    // NOTE: using extra array allows us to skip the search among indices
+    for (size_t j = 0; j < nnz; j++)
+      valuesAll[indices[j]] = values[j];
+    for (size_t j = 0; j < pnnz; j++) {
+      LO ind = colMap.getLocalElement(PColMap.getGlobalElement(pindices[j]));  // FIXME: we could do that before the full loop just once
+      if (ind != invalid)
+        // index indices[j] is part of template, copy corresponding value
+        newValues[j] = valuesAll[ind];
+      else
+        newValues[j] = zero;
+    }
+    for (size_t j = 0; j < nnz; j++)
+      valuesAll[indices[j]] = zero;
+
+    // Step 3: project to the space
+    Teuchos::SerialDenseMatrix<LO, SC>& XXtInv = XXtInv_[i];
+
+    Teuchos::SerialDenseMatrix<LO, SC> locX(NSDim, pnnz, false);
+    for (size_t j = 0; j < pnnz; j++)
+      for (size_t k = 0; k < NSDim; k++)
+        locX(k, j) = Xval[k][pindices[j]];
+
+    Teuchos::SerialDenseVector<LO, SC> val(pnnz, false), val1(NSDim, false), val2(NSDim, false);
+    for (size_t j = 0; j < pnnz; j++)
+      val[j] = newValues[j];
+
+    Teuchos::BLAS<LO, SC> blas;
+    // val1 = locX * val;
+    blas.GEMV(Teuchos::NO_TRANS, NSDim, pnnz,
+              one, locX.values(), locX.stride(),
+              val.values(), oneLO,
+              zero, val1.values(), oneLO);
+    // val2 = XXtInv * val1
+    blas.GEMV(Teuchos::NO_TRANS, NSDim, NSDim,
+              one, XXtInv.values(), XXtInv.stride(),
+              val1.values(), oneLO,
+              zero, val2.values(), oneLO);
+    // val = X^T * val2
+    blas.GEMV(Teuchos::CONJ_TRANS, NSDim, pnnz,
+              one, locX.values(), locX.stride(),
+              val2.values(), oneLO,
+              zero, val.values(), oneLO);
+
+    for (size_t j = 0; j < pnnz; j++)
+      newValues[j] -= val[j];
+
+    Projected.replaceLocalValues(i, pindices, newValues);
+  }
+
+  Projected.fillComplete(Projected.getDomainMap(), Projected.getRangeMap());  // FIXME: maps needed?
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif //ifndef MUELU_CONSTRAINT_DEF_HPP
+#endif  // ifndef MUELU_CONSTRAINT_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_decl.hpp
index 62a2b21bb889..7c369a893066 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_decl.hpp
@@ -64,58 +64,57 @@
 
 namespace MueLu {
 
-  /*!
-    @class EminPFactory class.
-    @brief Factory for building Energy Minimization prolongators.
-    @ingroup MueLuTransferClasses
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class EminPFactory : public PFactory {
+/*!
+  @class EminPFactory class.
+  @brief Factory for building Energy Minimization prolongators.
+  @ingroup MueLuTransferClasses
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class EminPFactory : public PFactory {
 #undef MUELU_EMINPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @brief Constructor.
+  EminPFactory() {}
 
-    //! @brief Constructor.
-    EminPFactory() { }
+  //! Destructor.
+  virtual ~EminPFactory() {}
 
-    //! Destructor.
-    virtual ~EminPFactory() { }
+  //@}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
 
-    //! @name Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
-
-    /*!
-      @brief Build method.
+  /*!
+    @brief Build method.
 
-      Builds energy minimization prolongator and returns it in <tt>coarseLevel</tt>.
-      */
-    void Build(Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+    Builds energy minimization prolongator and returns it in <tt>coarseLevel</tt>.
+    */
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; // class EminPFactory
+};  // class EminPFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_EMINPFACTORY_SHORT
-#endif // MUELU_EMINPFACTORY_DECL_HPP
+#endif  // MUELU_EMINPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_def.hpp
index f5f6f6edd49a..aad7bfb621f7 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_EminPFactory_def.hpp
@@ -17,193 +17,192 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("emin: num iterations");
-    SET_VALID_ENTRY("emin: num reuse iterations");
-    SET_VALID_ENTRY("emin: iterative method");
-    {
-      typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
-      validParamList->getEntry("emin: iterative method").setValidator(
-        rcp(new validatorType(Teuchos::tuple<std::string>("cg", "sd", "gmres"), "emin: iterative method")));
-    }
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("emin: num iterations");
+  SET_VALID_ENTRY("emin: num reuse iterations");
+  SET_VALID_ENTRY("emin: iterative method");
+  {
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validatorType;
+    validParamList->getEntry("emin: iterative method").setValidator(rcp(new validatorType(Teuchos::tuple<std::string>("cg", "sd", "gmres"), "emin: iterative method")));
+  }
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for the matrix A used during internal iterations");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory for the initial guess");
+  validParamList->set<RCP<const FactoryBase> >("Constraint", Teuchos::null, "Generating factory for constraints");
+
+  validParamList->set<RCP<Matrix> >("P0", Teuchos::null, "Initial guess at P");
+  validParamList->set<bool>("Keep P0", false, "Keep an initial P0 (for reuse)");
+
+  validParamList->set<RCP<Constraint> >("Constraint0", Teuchos::null, "Initial Constraint");
+  validParamList->set<bool>("Keep Constraint0", false, "Keep an initial Constraint (for reuse)");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+
+  static bool isAvailableP0          = false;
+  static bool isAvailableConstraint0 = false;
+
+  // Here is a tricky little piece of code
+  // We don't want to request (aka call Input) when we reuse and P0 is available
+  // However, we cannot run something simple like this:
+  //   if (!coarseLevel.IsAvailable("P0", this))
+  //     Input(coarseLevel, "P");
+  // The reason is that it works fine during the request stage, but fails
+  // in the release stage as we _construct_ P0 during Build process. Therefore,
+  // we need to understand whether we are in Request or Release mode
+  // NOTE: This is a very unique situation, please try not to propagate the
+  // mode check any further
+
+  if (coarseLevel.GetRequestMode() == Level::REQUEST) {
+    isAvailableP0          = coarseLevel.IsAvailable("P0", this);
+    isAvailableConstraint0 = coarseLevel.IsAvailable("Constraint0", this);
+  }
 
-    validParamList->set< RCP<const FactoryBase> >("A",                 Teuchos::null, "Generating factory for the matrix A used during internal iterations");
-    validParamList->set< RCP<const FactoryBase> >("P",                 Teuchos::null, "Generating factory for the initial guess");
-    validParamList->set< RCP<const FactoryBase> >("Constraint",        Teuchos::null, "Generating factory for constraints");
+  if (isAvailableP0 == false)
+    Input(coarseLevel, "P");
 
-    validParamList->set< RCP<Matrix> >           ("P0",                Teuchos::null, "Initial guess at P");
-    validParamList->set< bool >                  ("Keep P0",                   false, "Keep an initial P0 (for reuse)");
+  if (isAvailableConstraint0 == false)
+    Input(coarseLevel, "Constraint");
+}
 
-    validParamList->set< RCP<Constraint> >       ("Constraint0",       Teuchos::null, "Initial Constraint");
-    validParamList->set< bool >                  ("Keep Constraint0",          false, "Keep an initial Constraint (for reuse)");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  BuildP(fineLevel, coarseLevel);
+}
 
-    return validParamList;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Prolongator minimization", coarseLevel);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(fineLevel, "A");
-
-    static bool isAvailableP0          = false;
-    static bool isAvailableConstraint0 = false;
-
-    // Here is a tricky little piece of code
-    // We don't want to request (aka call Input) when we reuse and P0 is available
-    // However, we cannot run something simple like this:
-    //   if (!coarseLevel.IsAvailable("P0", this))
-    //     Input(coarseLevel, "P");
-    // The reason is that it works fine during the request stage, but fails
-    // in the release stage as we _construct_ P0 during Build process. Therefore,
-    // we need to understand whether we are in Request or Release mode
-    // NOTE: This is a very unique situation, please try not to propagate the
-    // mode check any further
-
-    if (coarseLevel.GetRequestMode() == Level::REQUEST) {
-      isAvailableP0          = coarseLevel.IsAvailable("P0",          this);
-      isAvailableConstraint0 = coarseLevel.IsAvailable("Constraint0", this);
-    }
+  const ParameterList& pL = GetParameterList();
 
-    if (isAvailableP0 == false)
-      Input(coarseLevel, "P");
+  // Get the matrix
+  RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
 
-    if (isAvailableConstraint0 == false)
-      Input(coarseLevel, "Constraint");
-  }
+  if (restrictionMode_) {
+    SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    BuildP(fineLevel, coarseLevel);
+    A = Utilities::Transpose(*A, true);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void EminPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Prolongator minimization", coarseLevel);
-
-    const ParameterList& pL = GetParameterList();
-
-    // Get the matrix
-    RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel, "A");
-
-    if (restrictionMode_) {
-      SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
+  // Get/make initial guess
+  RCP<Matrix> P0;
+  int numIts;
+  if (coarseLevel.IsAvailable("P0", this)) {
+    // Reuse data
+    P0     = coarseLevel.Get<RCP<Matrix> >("P0", this);
+    numIts = pL.get<int>("emin: num reuse iterations");
+    GetOStream(Runtime0) << "Reusing P0" << std::endl;
+
+  } else {
+    // Construct data
+    P0     = Get<RCP<Matrix> >(coarseLevel, "P");
+    numIts = pL.get<int>("emin: num iterations");
+  }
+  // NOTE: the main assumption here that P0 satisfies both constraints:
+  //   - nonzero pattern
+  //   - nullspace preservation
+
+  // Get/make constraint operator
+  RCP<Constraint> X;
+  if (coarseLevel.IsAvailable("Constraint0", this)) {
+    // Reuse data
+    X = coarseLevel.Get<RCP<Constraint> >("Constraint0", this);
+    GetOStream(Runtime0) << "Reusing Constraint0" << std::endl;
+
+  } else {
+    // Construct data
+    X = Get<RCP<Constraint> >(coarseLevel, "Constraint");
+  }
+  GetOStream(Runtime0) << "Number of emin iterations = " << numIts << std::endl;
+
+  std::string solverType = pL.get<std::string>("emin: iterative method");
+  RCP<SolverBase> solver;
+  if (solverType == "cg")
+    solver = rcp(new CGSolver(numIts));
+  else if (solverType == "sd")
+    solver = rcp(new SteepestDescentSolver(numIts));
+  else if (solverType == "gmres")
+    solver = rcp(new GMRESSolver(numIts));
+
+  RCP<Matrix> P;
+  solver->Iterate(*A, *X, *P0, P);
+
+  // NOTE: EXPERIMENTAL and FRAGILE
+  if (!P->IsView("stridedMaps")) {
+    if (A->IsView("stridedMaps") == true) {
+      GetOStream(Runtime1) << "Using A to fillComplete P" << std::endl;
+
+      // FIXME: X->GetPattern() actually returns a CrsGraph.
+      // CrsGraph has no knowledge of Xpetra's sup/Matrix views. As such,
+      // it has no idea about strided maps. We create one, which is
+      // most likely incorrect for many use cases.
+      std::vector<size_t> stridingInfo(1, 1);
+      RCP<const StridedMap> dMap = StridedMapFactory::Build(X->GetPattern()->getDomainMap(), stridingInfo);
+
+      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), dMap);
 
-      A = Utilities::Transpose(*A, true);
+    } else {
+      P->CreateView("stridedMaps", P->getRangeMap(), P->getDomainMap());
     }
+  }
 
-    // Get/make initial guess
-    RCP<Matrix> P0;
-    int         numIts;
-    if (coarseLevel.IsAvailable("P0", this)) {
-      // Reuse data
-      P0     = coarseLevel.Get<RCP<Matrix> >("P0", this);
-      numIts = pL.get<int>("emin: num reuse iterations");
-      GetOStream(Runtime0) << "Reusing P0" << std::endl;
+  // Level Set
+  if (!restrictionMode_) {
+    // The factory is in prolongation mode
+    Set(coarseLevel, "P", P);
+
+    if (pL.get<bool>("Keep P0")) {
+      // NOTE: we must do Keep _before_ set as the Needs class only sets if
+      //  a) data has been requested (which is not the case here), or
+      //  b) data has some keep flag
+      coarseLevel.Keep("P0", this);
+      Set(coarseLevel, "P0", P);
+    }
+    if (pL.get<bool>("Keep Constraint0")) {
+      // NOTE: we must do Keep _before_ set as the Needs class only sets if
+      //  a) data has been requested (which is not the case here), or
+      //  b) data has some keep flag
+      coarseLevel.Keep("Constraint0", this);
+      Set(coarseLevel, "Constraint0", X);
+    }
 
-    } else {
-      // Construct data
-      P0     = Get< RCP<Matrix> >(coarseLevel, "P");
-      numIts = pL.get<int>("emin: num iterations");
+    if (IsPrint(Statistics2)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*P, "P", params);
     }
-    // NOTE: the main assumption here that P0 satisfies both constraints:
-    //   - nonzero pattern
-    //   - nullspace preservation
 
-    // Get/make constraint operator
-    RCP<Constraint> X;
-    if (coarseLevel.IsAvailable("Constraint0", this)) {
-      // Reuse data
-      X = coarseLevel.Get<RCP<Constraint> >("Constraint0", this);
-      GetOStream(Runtime0) << "Reusing Constraint0" << std::endl;
+  } else {
+    // The factory is in restriction mode
+    RCP<Matrix> R;
+    {
+      SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
 
-    } else {
-      // Construct data
-      X = Get< RCP<Constraint> >(coarseLevel, "Constraint");
-    }
-    GetOStream(Runtime0) << "Number of emin iterations = " << numIts << std::endl;
-
-    std::string solverType = pL.get<std::string>("emin: iterative method");
-    RCP<SolverBase> solver;
-    if (solverType == "cg")
-      solver = rcp(new CGSolver(numIts));
-    else if (solverType == "sd")
-      solver = rcp(new SteepestDescentSolver(numIts));
-    else if (solverType == "gmres")
-      solver = rcp(new GMRESSolver(numIts));
-
-    RCP<Matrix> P;
-    solver->Iterate(*A, *X, *P0, P);
-
-    // NOTE: EXPERIMENTAL and FRAGILE
-    if (!P->IsView("stridedMaps")) {
-      if (A->IsView("stridedMaps") == true) {
-        GetOStream(Runtime1) << "Using A to fillComplete P" << std::endl;
-
-        // FIXME: X->GetPattern() actually returns a CrsGraph.
-        // CrsGraph has no knowledge of Xpetra's sup/Matrix views. As such,
-        // it has no idea about strided maps. We create one, which is
-        // most likely incorrect for many use cases.
-        std::vector<size_t> stridingInfo(1, 1);
-        RCP<const StridedMap> dMap = StridedMapFactory::Build(X->GetPattern()->getDomainMap(), stridingInfo);
-
-        P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), dMap);
-
-      } else {
-        P->CreateView("stridedMaps", P->getRangeMap(), P->getDomainMap());
-      }
+      R = Utilities::Transpose(*P, true);
     }
 
-    // Level Set
-    if (!restrictionMode_) {
-      // The factory is in prolongation mode
-      Set(coarseLevel, "P", P);
-
-      if (pL.get<bool>("Keep P0")) {
-        // NOTE: we must do Keep _before_ set as the Needs class only sets if
-        //  a) data has been requested (which is not the case here), or
-        //  b) data has some keep flag
-        coarseLevel.Keep("P0", this);
-        Set(coarseLevel, "P0", P);
-      }
-      if (pL.get<bool>("Keep Constraint0")) {
-        // NOTE: we must do Keep _before_ set as the Needs class only sets if
-        //  a) data has been requested (which is not the case here), or
-        //  b) data has some keep flag
-        coarseLevel.Keep("Constraint0", this);
-        Set(coarseLevel, "Constraint0", X);
-      }
-
-      if (IsPrint(Statistics2)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*P, "P", params);
-      }
+    Set(coarseLevel, "R", R);
 
-    } else {
-      // The factory is in restriction mode
-      RCP<Matrix> R;
-      {
-        SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
-
-        R = Utilities::Transpose(*P, true);
-      }
-
-      Set(coarseLevel, "R", R);
-
-      if (IsPrint(Statistics2)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
-      }
+    if (IsPrint(Statistics2)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
     }
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_EMINPFACTORY_DEF_HPP
+#endif  // MUELU_EMINPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_decl.hpp
index 546b49ee08b0..9f688e697792 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_decl.hpp
@@ -52,54 +52,52 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class NullspacePresmoothFactory : public SingleLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class NullspacePresmoothFactory : public SingleLevelFactoryBase {
 #undef MUELU_NULLSPACEPRESMOOTHFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  NullspacePresmoothFactory() {}
 
-    //! Constructor
-    NullspacePresmoothFactory() { }
+  //! Destructor
+  virtual ~NullspacePresmoothFactory() {}
 
-    //! Destructor
-    virtual ~NullspacePresmoothFactory() { }
+  //@}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! @name Input
+  //@{
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //! @name Input
-    //@{
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
+  void DeclareInput(Level &currentLevel) const;
 
-    void DeclareInput(Level &currentLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const;
+  //@}
 
-    //@}
+ private:
+};  // class NullspacePresmoothFactory
 
-  private:
-
-  }; //class NullspacePresmoothFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_NULLSPACEPRESMOOTHFACTORY_SHORT
-#endif // MUELU_NULLSPACEPRESMOOTHFACTORY_DECL_HPP
+#endif  // MUELU_NULLSPACEPRESMOOTHFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_def.hpp
index bcb66d36af3f..25bd92a7bfdd 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_NullspacePresmoothFactory_def.hpp
@@ -57,59 +57,59 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("A"        , Teuchos::null, "Generating factory for A");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory for the nonsmoothed nullspace");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for A");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory for the nonsmoothed nullspace");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    Input(currentLevel, "Nullspace");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  Input(currentLevel, "Nullspace");
 
-    if (currentLevel.GetLevelID() == 0)
-      Input(currentLevel, "A");
-  }
+  if (currentLevel.GetLevelID() == 0)
+    Input(currentLevel, "A");
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Nullspace presmooth factory", currentLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NullspacePresmoothFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Nullspace presmooth factory", currentLevel);
 
-    RCP<MultiVector> newB;
-    if (currentLevel.GetLevelID() == 0) {
-      RCP<Matrix>      A = Get< RCP<Matrix> >     (currentLevel, "A");
-      RCP<MultiVector> B = Get< RCP<MultiVector> >(currentLevel, "Nullspace");
-      newB = MultiVectorFactory::Build(B->getMap(), B->getNumVectors());
+  RCP<MultiVector> newB;
+  if (currentLevel.GetLevelID() == 0) {
+    RCP<Matrix> A      = Get<RCP<Matrix> >(currentLevel, "A");
+    RCP<MultiVector> B = Get<RCP<MultiVector> >(currentLevel, "Nullspace");
+    newB               = MultiVectorFactory::Build(B->getMap(), B->getNumVectors());
 
-      Teuchos::ArrayRCP<SC> D = Utilities::GetMatrixDiagonal_arcp(*A);
+    Teuchos::ArrayRCP<SC> D = Utilities::GetMatrixDiagonal_arcp(*A);
 
-      SC damping = 4./3;
-      damping /= Utilities::PowerMethod(*A, true, (LO) 10, 1e-4);
+    SC damping = 4. / 3;
+    damping /= Utilities::PowerMethod(*A, true, (LO)10, 1e-4);
 
-      A->apply(*B, *newB, Teuchos::NO_TRANS);
+    A->apply(*B, *newB, Teuchos::NO_TRANS);
 
-      size_t numVec  = newB->getNumVectors();
-      LO numElements = newB->getLocalLength();
-      for (size_t j = 0; j < numVec; j++) {
-        Teuchos::ArrayRCP<const SC> Bj    = B->getData(j);
-        Teuchos::ArrayRCP<SC>       newBj = newB->getDataNonConst(j);
+    size_t numVec  = newB->getNumVectors();
+    LO numElements = newB->getLocalLength();
+    for (size_t j = 0; j < numVec; j++) {
+      Teuchos::ArrayRCP<const SC> Bj = B->getData(j);
+      Teuchos::ArrayRCP<SC> newBj    = newB->getDataNonConst(j);
 
-        for (LO i = 0; i < numElements; i++)
-          newBj[i] = Bj[i] - damping*newBj[i]/D[i];
-      }
-    } else {
-      newB = Get< RCP<MultiVector> >(currentLevel, "Nullspace");
+      for (LO i = 0; i < numElements; i++)
+        newBj[i] = Bj[i] - damping * newBj[i] / D[i];
     }
+  } else {
+    newB = Get<RCP<MultiVector> >(currentLevel, "Nullspace");
+  }
 
-    // provide "Nullspace" variable on current level
-    Set(currentLevel, "Nullspace", newB);
+  // provide "Nullspace" variable on current level
+  Set(currentLevel, "Nullspace", newB);
 
-  } // Build
+}  // Build
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_NULLSPACEPRESMOOTHFACTORY_DEF_HPP
+#endif  // MUELU_NULLSPACEPRESMOOTHFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_decl.hpp
index 46fb5b1aed0b..fc6bfa742d85 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_decl.hpp
@@ -54,58 +54,56 @@
 
 namespace MueLu {
 
-  /*!
-    @class PatternFactory class.
-    @brief Factory for building nonzero patterns for energy minimization.
-    @ingroup MueLuTransferClasses
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class PatternFactory : public TwoLevelFactoryBase {
+/*!
+  @class PatternFactory class.
+  @brief Factory for building nonzero patterns for energy minimization.
+  @ingroup MueLuTransferClasses
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PatternFactory : public TwoLevelFactoryBase {
 #undef MUELU_PATTERNFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @brief Constructor.
-    PatternFactory() { }
+  //! @brief Constructor.
+  PatternFactory() {}
 
-    //! Destructor.
-    virtual ~PatternFactory() { }
+  //! Destructor.
+  virtual ~PatternFactory() {}
 
-    //@}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    /*!
-      @brief Build method.
-
-      Builds nonzero pattern (graph) and returns it in <tt>coarseLevel</tt>.
-      */
-    void Build(Level& fineLevel, Level& coarseLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
+    Builds nonzero pattern (graph) and returns it in <tt>coarseLevel</tt>.
+    */
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-  }; // class PatternFactory
+  //@}
 
+};  // class PatternFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PATTERNFACTORY_SHORT
-#endif // MUELU_PATTERNFACTORY_DECL_HPP
+#endif  // MUELU_PATTERNFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_def.hpp
index a37737d4493c..7d48c6ad241f 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/MueLu_PatternFactory_def.hpp
@@ -57,55 +57,54 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("emin: pattern order");
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("emin: pattern order");
+#undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for the matrix");
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory for the matrix providing nonzero graph");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory for the matrix");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory for the matrix providing nonzero graph");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(coarseLevel, "P");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(coarseLevel, "P");
 
-    const ParameterList& pL = GetParameterList();
-    if (pL.get<int>("emin: pattern order") > 0)
-      Input(fineLevel, "A");
-  }
+  const ParameterList& pL = GetParameterList();
+  if (pL.get<int>("emin: pattern order") > 0)
+    Input(fineLevel, "A");
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Ppattern", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PatternFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Ppattern", coarseLevel);
 
-    RCP<Matrix> P = Get< RCP<Matrix> >(coarseLevel, "P");
+  RCP<Matrix> P = Get<RCP<Matrix> >(coarseLevel, "P");
 
-    const ParameterList& pL = GetParameterList();
-    int k = pL.get<int>("emin: pattern order");
+  const ParameterList& pL = GetParameterList();
+  int k                   = pL.get<int>("emin: pattern order");
 
-    if (k > 0) {
-      RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel, "A");
-      RCP<Matrix> AP;
+  if (k > 0) {
+    RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
+    RCP<Matrix> AP;
 
-      bool doFillComplete  = true;
-      bool optimizeStorage = true;
+    bool doFillComplete  = true;
+    bool optimizeStorage = true;
 
-      for (int i = 0; i < k; i++) {
-        AP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
-        P.swap(AP);
-      }
+    for (int i = 0; i < k; i++) {
+      AP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
+      P.swap(AP);
     }
-
-    Set(coarseLevel, "Ppattern", P->getCrsGraph());
   }
 
+  Set(coarseLevel, "Ppattern", P->getCrsGraph());
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_PATTERNFACTORY_DEF_HPP
+#endif  // MUELU_PATTERNFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_decl.hpp
index 24feba352a59..108239249afe 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_decl.hpp
@@ -57,53 +57,53 @@
 
 namespace MueLu {
 
-  /*!
-    @class CGSolver class.
-    @brief Implements conjugate gradient algorithm for energy-minimization
-
-    This is CG applied to the problem
-    \f[ \min_P  \frac12 \sum_i (p_i)^T A p_i \f]
-    subject to
-    \f[ P(i,j) = 0 \quad \mbox{if} \; (i,j) \mbox{ is not in SparsityPattern} \f]
-    and
-    \f[ P cnull =  fnull. \f]
-
-    \note
-        For now we assume, that matrices have real values. The case of complex values is not explored.
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class CGSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+/*!
+  @class CGSolver class.
+  @brief Implements conjugate gradient algorithm for energy-minimization
+
+  This is CG applied to the problem
+  \f[ \min_P  \frac12 \sum_i (p_i)^T A p_i \f]
+  subject to
+  \f[ P(i,j) = 0 \quad \mbox{if} \; (i,j) \mbox{ is not in SparsityPattern} \f]
+  and
+  \f[ P cnull =  fnull. \f]
+
+  \note
+      For now we assume, that matrices have real values. The case of complex values is not explored.
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CGSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_CGSOLVER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! Constructor
-      \param Its -- Number of performed iterations
-      */
-    CGSolver(size_t Its);
+  /*! Constructor
+    \param Its -- Number of performed iterations
+    */
+  CGSolver(size_t Its);
 
-    //@}
+  //@}
 
-    //! @name Iterate methods.
-    //@{
+  //! @name Iterate methods.
+  //@{
 
-    //! Iterate
-    void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
+  //! Iterate
+  void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
 
-    //@}
+  //@}
 
-  private:
-    size_t nIts_;           //!< Number of performed iterations
-  };
+ private:
+  size_t nIts_;  //!< Number of performed iterations
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_CGSOLVER_SHORT
-#endif // MUELU_CGSOLVER_DECL_HPP
+#endif  // MUELU_CGSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp
index e04f6052b64f..a7094f5c1465 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_CGSolver_def.hpp
@@ -53,146 +53,142 @@
 #include "MueLu_Constraint.hpp"
 #include "MueLu_Monitor.hpp"
 
-
 #include "MueLu_CGSolver_decl.hpp"
 
-
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  CGSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CGSolver(size_t Its)
-  : nIts_(Its)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+CGSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CGSolver(size_t Its)
+  : nIts_(Its) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CGSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& finalP) const {
-    // Note: this function matrix notations follow Saad's "Iterative methods", ed. 2, pg. 246
-    // So, X is the unknown prolongator, P's are conjugate directions, Z's are preconditioned P's
-    PrintMonitor m(*this, "CG iterations");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CGSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& finalP) const {
+  // Note: this function matrix notations follow Saad's "Iterative methods", ed. 2, pg. 246
+  // So, X is the unknown prolongator, P's are conjugate directions, Z's are preconditioned P's
+  PrintMonitor m(*this, "CG iterations");
 
-    if (nIts_ == 0) {
-      finalP = MatrixFactory2::BuildCopy(rcpFromRef(P0));
-      return;
-    }
+  if (nIts_ == 0) {
+    finalP = MatrixFactory2::BuildCopy(rcpFromRef(P0));
+    return;
+  }
 
-    RCP<const Matrix>  A         = rcpFromRef(Aref);
-    ArrayRCP<const SC> D         = Utilities::GetMatrixDiagonal_arcp(*A);
-    bool               useTpetra = (A->getRowMap()->lib() == Xpetra::UseTpetra);
+  RCP<const Matrix> A  = rcpFromRef(Aref);
+  ArrayRCP<const SC> D = Utilities::GetMatrixDiagonal_arcp(*A);
+  bool useTpetra       = (A->getRowMap()->lib() == Xpetra::UseTpetra);
 
-    Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
+  Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
 
-    SC one = Teuchos::ScalarTraits<SC>::one();
+  SC one = Teuchos::ScalarTraits<SC>::one();
 
-    RCP<Matrix> X, P, R, Z, AP;
-    RCP<Matrix> newX, tmpAP;
+  RCP<Matrix> X, P, R, Z, AP;
+  RCP<Matrix> newX, tmpAP;
 #ifndef TWO_ARG_MATRIX_ADD
-    RCP<Matrix> newR, newP;
+  RCP<Matrix> newR, newP;
 #endif
 
-    SC oldRZ, newRZ, alpha, beta, app;
+  SC oldRZ, newRZ, alpha, beta, app;
 
-    // T is used only for projecting onto
-    RCP<CrsMatrix> T_ = CrsMatrixFactory::Build(C.GetPattern());
-    T_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-    RCP<Matrix>    T = rcp(new CrsMatrixWrap(T_));
+  // T is used only for projecting onto
+  RCP<CrsMatrix> T_ = CrsMatrixFactory::Build(C.GetPattern());
+  T_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+  RCP<Matrix> T = rcp(new CrsMatrixWrap(T_));
 
-    // Initial P0 would only be used for multiplication
-    X = rcp_const_cast<Matrix>(rcpFromRef(P0));
+  // Initial P0 would only be used for multiplication
+  X = rcp_const_cast<Matrix>(rcpFromRef(P0));
 
-    tmpAP = MatrixMatrix::Multiply(*A, false, *X, false, mmfancy, true/*doFillComplete*/, true/*optimizeStorage*/);
-    C.Apply(*tmpAP, *T);
+  tmpAP = MatrixMatrix::Multiply(*A, false, *X, false, mmfancy, true /*doFillComplete*/, true /*optimizeStorage*/);
+  C.Apply(*tmpAP, *T);
 
-    // R_0 = -A*X_0
-    R = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(T);
+  // R_0 = -A*X_0
+  R = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(T);
 
-    R->scale(-one);
+  R->scale(-one);
 
-    // Z_0 = M^{-1}R_0
-    Z = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(R);
-    Utilities::MyOldScaleMatrix(*Z, D, true, true, false);
+  // Z_0 = M^{-1}R_0
+  Z = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(R);
+  Utilities::MyOldScaleMatrix(*Z, D, true, true, false);
+
+  // P_0 = Z_0
+  P = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(Z);
+
+  oldRZ = Utilities::Frobenius(*R, *Z);
 
-    // P_0 = Z_0
-    P = Xpetra::MatrixFactory2<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCopy(Z);
-
-    oldRZ = Utilities::Frobenius(*R, *Z);
-
-    for (size_t i = 0; i < nIts_; i++) {
-      // AP = constrain(A*P)
-      if (i == 0 || useTpetra) {
-        // Construct the MxM pattern from scratch
-        // This is done by default for Tpetra as the three argument version requires tmpAP
-        // to *not* be locally indexed which defeats the purpose
-        // TODO: need a three argument Tpetra version which allows reuse of already fill-completed matrix
-        tmpAP = MatrixMatrix::Multiply(*A, false, *P, false,        mmfancy, true/*doFillComplete*/, true/*optimizeStorage*/);
-      } else {
-        // Reuse the MxM pattern
-        tmpAP = MatrixMatrix::Multiply(*A, false, *P, false, tmpAP, mmfancy, true/*doFillComplete*/, true/*optimizeStorage*/);
-      }
-      C.Apply(*tmpAP, *T);
-      AP = T;
-
-      app = Utilities::Frobenius(*AP, *P);
-      if (Teuchos::ScalarTraits<SC>::magnitude(app) < Teuchos::ScalarTraits<SC>::sfmin()) {
-        // It happens, for instance, if P = 0
-        // For example, if we use TentativePFactory for both nonzero pattern and initial guess
-        // I think it might also happen because of numerical breakdown, but we don't test for that yet
-        if (i == 0)
-          X = MatrixFactory2::BuildCopy(rcpFromRef(P0));
-        break;
-      }
-
-      // alpha = (R_i, Z_i)/(A*P_i, P_i)
-      alpha = oldRZ / app;
-      this->GetOStream(Runtime1,1) << "alpha = " << alpha << std::endl;
-
-      // X_{i+1} = X_i + alpha*P_i
+  for (size_t i = 0; i < nIts_; i++) {
+    // AP = constrain(A*P)
+    if (i == 0 || useTpetra) {
+      // Construct the MxM pattern from scratch
+      // This is done by default for Tpetra as the three argument version requires tmpAP
+      // to *not* be locally indexed which defeats the purpose
+      // TODO: need a three argument Tpetra version which allows reuse of already fill-completed matrix
+      tmpAP = MatrixMatrix::Multiply(*A, false, *P, false, mmfancy, true /*doFillComplete*/, true /*optimizeStorage*/);
+    } else {
+      // Reuse the MxM pattern
+      tmpAP = MatrixMatrix::Multiply(*A, false, *P, false, tmpAP, mmfancy, true /*doFillComplete*/, true /*optimizeStorage*/);
+    }
+    C.Apply(*tmpAP, *T);
+    AP = T;
+
+    app = Utilities::Frobenius(*AP, *P);
+    if (Teuchos::ScalarTraits<SC>::magnitude(app) < Teuchos::ScalarTraits<SC>::sfmin()) {
+      // It happens, for instance, if P = 0
+      // For example, if we use TentativePFactory for both nonzero pattern and initial guess
+      // I think it might also happen because of numerical breakdown, but we don't test for that yet
+      if (i == 0)
+        X = MatrixFactory2::BuildCopy(rcpFromRef(P0));
+      break;
+    }
+
+    // alpha = (R_i, Z_i)/(A*P_i, P_i)
+    alpha = oldRZ / app;
+    this->GetOStream(Runtime1, 1) << "alpha = " << alpha << std::endl;
+
+    // X_{i+1} = X_i + alpha*P_i
 #ifndef TWO_ARG_MATRIX_ADD
-      newX = Teuchos::null;
-      MatrixMatrix::TwoMatrixAdd(*P, false, alpha, *X, false, one, newX, mmfancy);
-      newX->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-      X.swap(newX);
+    newX = Teuchos::null;
+    MatrixMatrix::TwoMatrixAdd(*P, false, alpha, *X, false, one, newX, mmfancy);
+    newX->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+    X.swap(newX);
 #else
-      MatrixMatrix::TwoMatrixAdd(*P, false, alpha, *X, one);
+    MatrixMatrix::TwoMatrixAdd(*P, false, alpha, *X, one);
 #endif
 
-      if (i == nIts_ - 1)
-        break;
+    if (i == nIts_ - 1)
+      break;
 
       // R_{i+1} = R_i - alpha*A*P_i
 #ifndef TWO_ARG_MATRIX_ADD
-      newR = Teuchos::null;
-      MatrixMatrix::TwoMatrixAdd(*AP, false, -alpha, *R, false, one, newR, mmfancy);
-      newR->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-      R.swap(newR);
+    newR = Teuchos::null;
+    MatrixMatrix::TwoMatrixAdd(*AP, false, -alpha, *R, false, one, newR, mmfancy);
+    newR->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+    R.swap(newR);
 #else
-      MatrixMatrix::TwoMatrixAdd(*AP, false, -alpha, *R, one);
+    MatrixMatrix::TwoMatrixAdd(*AP, false, -alpha, *R, one);
 #endif
 
-      // Z_{i+1} = M^{-1} R_{i+1}
-      Z = MatrixFactory2::BuildCopy(R);
-      Utilities::MyOldScaleMatrix(*Z, D, true, true, false);
+    // Z_{i+1} = M^{-1} R_{i+1}
+    Z = MatrixFactory2::BuildCopy(R);
+    Utilities::MyOldScaleMatrix(*Z, D, true, true, false);
 
-      // beta = (R_{i+1}, Z_{i+1})/(R_i, Z_i)
-      newRZ = Utilities::Frobenius(*R, *Z);
-      beta = newRZ / oldRZ;
+    // beta = (R_{i+1}, Z_{i+1})/(R_i, Z_i)
+    newRZ = Utilities::Frobenius(*R, *Z);
+    beta  = newRZ / oldRZ;
 
-      // P_{i+1} = Z_{i+1} + beta*P_i
+    // P_{i+1} = Z_{i+1} + beta*P_i
 #ifndef TWO_ARG_MATRIX_ADD
-      newP = Teuchos::null;
-      MatrixMatrix::TwoMatrixAdd(*P, false, beta, *Z, false, one, newP, mmfancy);
-      newP->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-      P.swap(newP);
+    newP = Teuchos::null;
+    MatrixMatrix::TwoMatrixAdd(*P, false, beta, *Z, false, one, newP, mmfancy);
+    newP->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+    P.swap(newP);
 #else
-      MatrixMatrix::TwoMatrixAdd(*Z, false, one, *P, beta);
+    MatrixMatrix::TwoMatrixAdd(*Z, false, one, *P, beta);
 #endif
 
-      oldRZ = newRZ;
-    }
-
-    finalP = X;
+    oldRZ = newRZ;
   }
 
-} // namespace MueLu
+  finalP = X;
+}
+
+}  // namespace MueLu
 
-#endif //ifndef MUELU_CGSOLVER_DECL_HPP
+#endif  // ifndef MUELU_CGSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_decl.hpp
index 0471266a58df..6f5af37521f7 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_decl.hpp
@@ -57,56 +57,56 @@
 
 namespace MueLu {
 
-  /*!
-    @class GMRESSolver class.
-    @brief Implements conjugate gradient algorithm for energy-minimization
-
-    This is GMRES applied to the problem
-    \f[ \min_P  \frac12 \sum_i (p_i)^T A p_i \f]
-    subject to
-    \f[ P(i,j) = 0 \quad \mbox{if} \; (i,j) \mbox{ is not in SparsityPattern} \f]
-    and
-    \f[ P cnull =  fnull. \f]
-
-    \note
-        For now we assume, that matrices have real values. The case of complex values is not explored.
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class GMRESSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+/*!
+  @class GMRESSolver class.
+  @brief Implements conjugate gradient algorithm for energy-minimization
+
+  This is GMRES applied to the problem
+  \f[ \min_P  \frac12 \sum_i (p_i)^T A p_i \f]
+  subject to
+  \f[ P(i,j) = 0 \quad \mbox{if} \; (i,j) \mbox{ is not in SparsityPattern} \f]
+  and
+  \f[ P cnull =  fnull. \f]
+
+  \note
+      For now we assume, that matrices have real values. The case of complex values is not explored.
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GMRESSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_GMRESSOLVER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! Constructor
-      \param Its -- Number of performed iterations
-      */
-    GMRESSolver(size_t Its);
+  /*! Constructor
+    \param Its -- Number of performed iterations
+    */
+  GMRESSolver(size_t Its);
 
-    //@}
+  //@}
 
-    //! @name Iterate methods.
-    //@{
+  //! @name Iterate methods.
+  //@{
 
-    //! Iterate
-    void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
+  //! Iterate
+  void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
 
-    //@}
+  //@}
 
-  private:
-    void givapp(SC* c, SC* s, SC* v, int k) const;
+ private:
+  void givapp(SC* c, SC* s, SC* v, int k) const;
 
-  private:
-    size_t nIts_;           //!< Number of performed iterations
-  };
+ private:
+  size_t nIts_;  //!< Number of performed iterations
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GMRESSOLVER_SHORT
-#endif // MUELU_GMRESSOLVER_DECL_HPP
+#endif  // MUELU_GMRESSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_def.hpp
index cc079ffc38c3..cb71ef7c22b2 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_GMRESSolver_def.hpp
@@ -58,166 +58,164 @@
 #include "MueLu_Monitor.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GMRESSolver(size_t Its)
-  : nIts_(Its)
-  { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::givapp(Scalar* c, Scalar* s, Scalar* v, int k) const {
-    for (int i = 0; i < k; i++) {
-      SC w1 = c[i]*v[i] - s[i]*v[i+1];
-      SC w2 = s[i]*v[i] + c[i]*v[i+1];
-      v[i]   = w1;
-      v[i+1] = w2;
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GMRESSolver(size_t Its)
+  : nIts_(Its) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::givapp(Scalar* c, Scalar* s, Scalar* v, int k) const {
+  for (int i = 0; i < k; i++) {
+    SC w1    = c[i] * v[i] - s[i] * v[i + 1];
+    SC w2    = s[i] * v[i] + c[i] * v[i + 1];
+    v[i]     = w1;
+    v[i + 1] = w2;
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& finalP) const {
-    PrintMonitor m(*this, "GMRES iterations");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GMRESSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& finalP) const {
+  PrintMonitor m(*this, "GMRES iterations");
 
-    finalP = MatrixFactory2::BuildCopy(rcpFromRef(P0));
-    if (nIts_ == 0)
-      return;
+  finalP = MatrixFactory2::BuildCopy(rcpFromRef(P0));
+  if (nIts_ == 0)
+    return;
 
-    TEUCHOS_TEST_FOR_EXCEPTION(nIts_ > 1, Exceptions::RuntimeError,
-        "For now, solving Hessenberg system works only for a single iteration");
+  TEUCHOS_TEST_FOR_EXCEPTION(nIts_ > 1, Exceptions::RuntimeError,
+                             "For now, solving Hessenberg system works only for a single iteration");
 
-    SC one = Teuchos::ScalarTraits<SC>::one(), zero = Teuchos::ScalarTraits<SC>::zero();
+  SC one = Teuchos::ScalarTraits<SC>::one(), zero = Teuchos::ScalarTraits<SC>::zero();
 
-    RCP<const Matrix>  A         = rcpFromRef(Aref);
-    //bool               useTpetra = (A->getRowMap()->lib() == Xpetra::UseTpetra);
+  RCP<const Matrix> A = rcpFromRef(Aref);
+  // bool               useTpetra = (A->getRowMap()->lib() == Xpetra::UseTpetra);
 
-    // FIXME: Don't know why, but in the MATLAB code we have D = I. Follow that for now.
+  // FIXME: Don't know why, but in the MATLAB code we have D = I. Follow that for now.
 #if 0
     ArrayRCP<const SC> D         = Utilities::GetMatrixDiagonal_arcp(*A);
 #else
-    ArrayRCP<const SC> D(A->getLocalNumRows(), one);
+  ArrayRCP<const SC> D(A->getLocalNumRows(), one);
 #endif
 
-    Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
+  Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
 
-    // Initial P0 would only be used for multiplication
-    RCP<Matrix>               X = rcp_const_cast<Matrix>(rcpFromRef(P0)), tmpAP, newV;
-    std::vector<RCP<Matrix> > V(nIts_+1);
+  // Initial P0 would only be used for multiplication
+  RCP<Matrix> X = rcp_const_cast<Matrix>(rcpFromRef(P0)), tmpAP, newV;
+  std::vector<RCP<Matrix> > V(nIts_ + 1);
 
-    // T is used only for projecting onto
-    RCP<CrsMatrix> T_ = CrsMatrixFactory::Build(C.GetPattern());
-    T_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-    RCP<Matrix>    T = rcp(new CrsMatrixWrap(T_));
+  // T is used only for projecting onto
+  RCP<CrsMatrix> T_ = CrsMatrixFactory::Build(C.GetPattern());
+  T_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+  RCP<Matrix> T = rcp(new CrsMatrixWrap(T_));
 
-    SC rho;
-    {
-      // R_0 = -D^{-1}*A*X_0
-      // V_0 = R_0 / ||R_0||_F
-      tmpAP = MatrixMatrix::Multiply(*A, false, *X, false, mmfancy, true/*doFillComplete*/, true/*optimizeStorage*/);
-      C.Apply(*tmpAP, *T);
+  SC rho;
+  {
+    // R_0 = -D^{-1}*A*X_0
+    // V_0 = R_0 / ||R_0||_F
+    tmpAP = MatrixMatrix::Multiply(*A, false, *X, false, mmfancy, true /*doFillComplete*/, true /*optimizeStorage*/);
+    C.Apply(*tmpAP, *T);
 
-      V[0] = MatrixFactory2::BuildCopy(T);
-      Utilities::MyOldScaleMatrix(*V[0], D, true/*doInverse*/, true/*doFillComplete*/, false/*doOptimizeStorage*/);
+    V[0] = MatrixFactory2::BuildCopy(T);
+    Utilities::MyOldScaleMatrix(*V[0], D, true /*doInverse*/, true /*doFillComplete*/, false /*doOptimizeStorage*/);
 
-      rho = sqrt(Utilities::Frobenius(*V[0], *V[0]));
+    rho = sqrt(Utilities::Frobenius(*V[0], *V[0]));
 
-      V[0]->scale(-one/rho);
-    }
+    V[0]->scale(-one / rho);
+  }
 
-    std::vector<SC> h((nIts_+1) * (nIts_+1));
-    std::vector<SC> c(nIts_+1, 0.0);
-    std::vector<SC> s(nIts_+1, 0.0);
-    std::vector<SC> g(nIts_+1, 0.0);
-    g[0] = rho;
+  std::vector<SC> h((nIts_ + 1) * (nIts_ + 1));
+  std::vector<SC> c(nIts_ + 1, 0.0);
+  std::vector<SC> s(nIts_ + 1, 0.0);
+  std::vector<SC> g(nIts_ + 1, 0.0);
+  g[0] = rho;
 
-#define I(i,j) ((i) + (j)*(nIts_+1)) // column ordering
-    for (size_t i = 0; i < nIts_; i++) {
-      // V_{i+1} = D^{-1}*A*V_i
-      tmpAP = MatrixMatrix::Multiply(*A, false, *V[i], false, mmfancy, true/*doFillComplete*/, true/*optimizeStorage*/);
-      C.Apply(*tmpAP, *T);
+#define I(i, j) ((i) + (j) * (nIts_ + 1))  // column ordering
+  for (size_t i = 0; i < nIts_; i++) {
+    // V_{i+1} = D^{-1}*A*V_i
+    tmpAP = MatrixMatrix::Multiply(*A, false, *V[i], false, mmfancy, true /*doFillComplete*/, true /*optimizeStorage*/);
+    C.Apply(*tmpAP, *T);
 
-      V[i+1] = MatrixFactory2::BuildCopy(T);
-      Utilities::MyOldScaleMatrix(*V[i+1], D, true/*doInverse*/, true/*doFillComplete*/, false/*doOptimizeStorage*/);
+    V[i + 1] = MatrixFactory2::BuildCopy(T);
+    Utilities::MyOldScaleMatrix(*V[i + 1], D, true /*doInverse*/, true /*doFillComplete*/, false /*doOptimizeStorage*/);
 
-      // Update Hessenberg matrix
-      for (size_t j = 0; j <= i; j++) {
-        h[I(j,i)] = Utilities::Frobenius(*V[i+1], *V[j]);
+    // Update Hessenberg matrix
+    for (size_t j = 0; j <= i; j++) {
+      h[I(j, i)] = Utilities::Frobenius(*V[i + 1], *V[j]);
 
-        // V_{i+1} = V_{i+1} - h(j,i+1)*V_j
+      // V_{i+1} = V_{i+1} - h(j,i+1)*V_j
 #ifndef TWO_ARG_MATRIX_ADD
-        newV = Teuchos::null;
-        MatrixMatrix::TwoMatrixAdd(*V[j], false, -h[I(j,i)], *V[i+1], false, one, newV, mmfancy);
-        newV->fillComplete(V[i+1]->getDomainMap(), V[i+1]->getRangeMap());
-        V[i+1].swap(newV);
+      newV = Teuchos::null;
+      MatrixMatrix::TwoMatrixAdd(*V[j], false, -h[I(j, i)], *V[i + 1], false, one, newV, mmfancy);
+      newV->fillComplete(V[i + 1]->getDomainMap(), V[i + 1]->getRangeMap());
+      V[i + 1].swap(newV);
 #else
-        // FIXME: this does not work now. Fails with the following exception:
-        //   what():  ../../packages/tpetra/core/ext/TpetraExt_MatrixMatrix_def.hpp:408:
-        //
-        //   Throw number = 1
-        //
-        //   Throw test that evaluated to true: B.isLocallyIndexed()
-        //
-        //   TpetraExt::MatrixMatrix::Add(): ERROR, input matrix B must not be locally indexed
-        MatrixMatrix::TwoMatrixAdd(*V[j], false, -h[I(j,i)], *V[i+1], one);
+      // FIXME: this does not work now. Fails with the following exception:
+      //   what():  ../../packages/tpetra/core/ext/TpetraExt_MatrixMatrix_def.hpp:408:
+      //
+      //   Throw number = 1
+      //
+      //   Throw test that evaluated to true: B.isLocallyIndexed()
+      //
+      //   TpetraExt::MatrixMatrix::Add(): ERROR, input matrix B must not be locally indexed
+      MatrixMatrix::TwoMatrixAdd(*V[j], false, -h[I(j, i)], *V[i + 1], one);
 #endif
-      }
-      h[I(i+1,i)] = sqrt(Utilities::Frobenius(*V[i+1], *V[i+1]));
-
-      // NOTE: potentially we'll need some reorthogonalization code here
-      // The matching MATLAB code is
-      //    normav  = norm(v.num(k+1).matrix, 'fro');
-      //    normav2 = h(k+1,k);
-      //    if  (reorth == -1 && normav + .001*normav2 == normav)
-      //        for j = 1:k
-      //            hr       = v(:,j)'*v(:,k+1);    % hr=v(:,k+1)'*v(:,j);
-      //            h(j,k)   = h(j,k)+hr;
-      //            v(:,k+1) = v(:,k+1)-hr*v(:,j);
-      //        end
-      //        h(k+1,k) = norm(v(:,k+1));
-      //    end
-
-      // Check for nonsymmetric case
-      if (h[I(i+1,i)] != zero) {
-        // Normalize V_i
-        V[i+1]->scale(one/h[I(i+1,i)]);
-      }
-
-      if (i > 0)
-        givapp(&c[0], &s[0], &h[I(0,i)], i); // Due to column ordering &h[...] is a column
-
-      SC nu = sqrt(h[I(i,i)]*h[I(i,i)] + h[I(i+1,i)]*h[I(i+1,i)]);
-      if (nu != zero) {
-        c[i]        =  h[I(i,  i)] / nu;
-        s[i]        = -h[I(i+1,i)] / nu;
-        h[I(i,i)]   =  c[i] * h[I(i,i)] - s[i] * h[I(i+1,i)];
-        h[I(i+1,i)] =  zero;
-
-        givapp(&c[i], &s[i], &g[i], 1);
-      }
     }
+    h[I(i + 1, i)] = sqrt(Utilities::Frobenius(*V[i + 1], *V[i + 1]));
+
+    // NOTE: potentially we'll need some reorthogonalization code here
+    // The matching MATLAB code is
+    //    normav  = norm(v.num(k+1).matrix, 'fro');
+    //    normav2 = h(k+1,k);
+    //    if  (reorth == -1 && normav + .001*normav2 == normav)
+    //        for j = 1:k
+    //            hr       = v(:,j)'*v(:,k+1);    % hr=v(:,k+1)'*v(:,j);
+    //            h(j,k)   = h(j,k)+hr;
+    //            v(:,k+1) = v(:,k+1)-hr*v(:,j);
+    //        end
+    //        h(k+1,k) = norm(v(:,k+1));
+    //    end
+
+    // Check for nonsymmetric case
+    if (h[I(i + 1, i)] != zero) {
+      // Normalize V_i
+      V[i + 1]->scale(one / h[I(i + 1, i)]);
+    }
+
+    if (i > 0)
+      givapp(&c[0], &s[0], &h[I(0, i)], i);  // Due to column ordering &h[...] is a column
 
-    // Solve Hessenberg system
-    //   y = solve(H, \rho e_1)
-    std::vector<SC> y(nIts_);
-    if (nIts_ == 1) {
-      y[0] = g[0] / h[I(0,0)];
+    SC nu = sqrt(h[I(i, i)] * h[I(i, i)] + h[I(i + 1, i)] * h[I(i + 1, i)]);
+    if (nu != zero) {
+      c[i]           = h[I(i, i)] / nu;
+      s[i]           = -h[I(i + 1, i)] / nu;
+      h[I(i, i)]     = c[i] * h[I(i, i)] - s[i] * h[I(i + 1, i)];
+      h[I(i + 1, i)] = zero;
+
+      givapp(&c[i], &s[i], &g[i], 1);
     }
+  }
+
+  // Solve Hessenberg system
+  //   y = solve(H, \rho e_1)
+  std::vector<SC> y(nIts_);
+  if (nIts_ == 1) {
+    y[0] = g[0] / h[I(0, 0)];
+  }
 #undef I
 
-    // Compute final
-    for (size_t i = 0; i < nIts_; i++) {
+  // Compute final
+  for (size_t i = 0; i < nIts_; i++) {
 #ifndef TWO_ARG_MATRIX_ADD
-      newV = Teuchos::null;
-      MatrixMatrix::TwoMatrixAdd(*V[i], false, y[i], *finalP, false, one, newV, mmfancy);
-      newV->fillComplete(finalP->getDomainMap(), finalP->getRangeMap());
-      finalP.swap(newV);
+    newV = Teuchos::null;
+    MatrixMatrix::TwoMatrixAdd(*V[i], false, y[i], *finalP, false, one, newV, mmfancy);
+    newV->fillComplete(finalP->getDomainMap(), finalP->getRangeMap());
+    finalP.swap(newV);
 #else
-      MatrixMatrix::TwoMatrixAdd(*V[i], false, y[i], *finalP, one);
+    MatrixMatrix::TwoMatrixAdd(*V[i], false, y[i], *finalP, one);
 #endif
-    }
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif //ifndef MUELU_GMRESSOLVER_DECL_HPP
+#endif  // ifndef MUELU_GMRESSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SolverBase.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SolverBase.hpp
index 0447bda0f587..4796631a4d16 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SolverBase.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SolverBase.hpp
@@ -55,44 +55,44 @@
 
 namespace MueLu {
 
-  /*!
-    @class SolverBase class.
-    @brief Base class for energy-minimization iterative solvers
-    */
+/*!
+  @class SolverBase class.
+  @brief Base class for energy-minimization iterative solvers
+  */
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SolverBase : public BaseClass {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SolverBase : public BaseClass {
 #undef MUELU_SOLVERBASE_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //@{ Constructors/Destructors.
-    SolverBase() {}
+ public:
+  //@{ Constructors/Destructors.
+  SolverBase() {}
 
-    virtual ~SolverBase() {}
-    //@}
+  virtual ~SolverBase() {}
+  //@}
 
-    //! @name Iterate methods.
-    //@{
+  //! @name Iterate methods.
+  //@{
 
-    /*! Iterate
-     \param A -- Energy matrix
-     \param C -- Constraints
-     \param P0 -- Initial guess
-     \param P -- Resulting prolongator
-     \note We assume that P0 has correct sparsity pattern
-     */
-    virtual void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const = 0;
+  /*! Iterate
+   \param A -- Energy matrix
+   \param C -- Constraints
+   \param P0 -- Initial guess
+   \param P -- Resulting prolongator
+   \note We assume that P0 has correct sparsity pattern
+   */
+  virtual void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const = 0;
 
-    //@}
+  //@}
 
-  }; //class SolverBase
+};  // class SolverBase
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SOLVERBASE_SHORT
 
-#endif //ifndef MUELU_SOLVERBASE_HPP
+#endif  // ifndef MUELU_SOLVERBASE_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_decl.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_decl.hpp
index 9cf3e5916fc3..5b4e93a39f71 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_decl.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_decl.hpp
@@ -50,54 +50,52 @@
 #include <Xpetra_CrsMatrixWrap_fwd.hpp>
 #include <Xpetra_CrsMatrixFactory_fwd.hpp>
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_Utilities_fwd.hpp"
 #include "MueLu_SolverBase.hpp"
 #include "MueLu_Constraint_fwd.hpp"
 
-
 namespace MueLu {
 
-  /*!
-    @class SteepestDescentSolver class.
-    @brief Implements steepest descent algorithm for energy-minimization
-    */
+/*!
+  @class SteepestDescentSolver class.
+  @brief Implements steepest descent algorithm for energy-minimization
+  */
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SteepestDescentSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SteepestDescentSolver : public SolverBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_STEEPESTDESCENTSOLVER_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors / destructors
-    //@{
+ public:
+  //! @name Constructors / destructors
+  //@{
 
-    /*! Constructor
-      \param Its -- Number of performed iterations
-      \param StepLength -- The modifier of the step length
-      */
-    SteepestDescentSolver(size_t Its, SC StepLength = Teuchos::ScalarTraits<Scalar>::one());
+  /*! Constructor
+    \param Its -- Number of performed iterations
+    \param StepLength -- The modifier of the step length
+    */
+  SteepestDescentSolver(size_t Its, SC StepLength = Teuchos::ScalarTraits<Scalar>::one());
 
-    //@}
+  //@}
 
-    //! @name Iterate methods.
-    //@{
+  //! @name Iterate methods.
+  //@{
 
-    //! Iterate
-    void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
+  //! Iterate
+  void Iterate(const Matrix& A, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const;
 
-    //@}
+  //@}
 
-  private:
-    size_t nIts_;           //!< Number of performed iterations
-    SC     stepLength_;     //!< Modifier of the step length
-  };
+ private:
+  size_t nIts_;    //!< Number of performed iterations
+  SC stepLength_;  //!< Modifier of the step length
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_STEEPESTDESCENTSOLVER_SHORT
-#endif // MUELU_STEEPESTDESCENTSOLVER_DECL_HPP
+#endif  // MUELU_STEEPESTDESCENTSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_def.hpp b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_def.hpp
index d5e339a2ab35..13b555727ba4 100644
--- a/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_def.hpp
+++ b/packages/muelu/src/Transfers/Energy-Minimization/Solvers/MueLu_SteepestDescentSolver_def.hpp
@@ -58,51 +58,51 @@
 
 namespace MueLu {
 
-  using Teuchos::rcp_const_cast;
+using Teuchos::rcp_const_cast;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  SteepestDescentSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SteepestDescentSolver(size_t Its, SC StepLength)
-  : nIts_(Its), stepLength_(StepLength)
-  { }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+SteepestDescentSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SteepestDescentSolver(size_t Its, SC StepLength)
+  : nIts_(Its)
+  , stepLength_(StepLength) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SteepestDescentSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const {
-    PrintMonitor m(*this, "SD iterations");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SteepestDescentSolver<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Iterate(const Matrix& Aref, const Constraint& C, const Matrix& P0, RCP<Matrix>& P) const {
+  PrintMonitor m(*this, "SD iterations");
 
-    RCP<const Matrix> A = rcpFromRef(Aref);
-    RCP<Matrix> AP, G;
+  RCP<const Matrix> A = rcpFromRef(Aref);
+  RCP<Matrix> AP, G;
 
-    Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
+  Teuchos::FancyOStream& mmfancy = this->GetOStream(Statistics2);
 
-    Teuchos::ArrayRCP<const SC> D = Utilities::GetMatrixDiagonal_arcp(*A);
+  Teuchos::ArrayRCP<const SC> D = Utilities::GetMatrixDiagonal_arcp(*A);
 
-    RCP<CrsMatrix> Ptmp_ = CrsMatrixFactory::Build(C.GetPattern());
-    Ptmp_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
-    RCP<Matrix>    Ptmp  = rcp(new CrsMatrixWrap(Ptmp_));
+  RCP<CrsMatrix> Ptmp_ = CrsMatrixFactory::Build(C.GetPattern());
+  Ptmp_->fillComplete(P0.getDomainMap(), P0.getRangeMap());
+  RCP<Matrix> Ptmp = rcp(new CrsMatrixWrap(Ptmp_));
 
-    // Initial P0 would only be used for multiplication
-    P = rcp_const_cast<Matrix>(rcpFromRef(P0));
+  // Initial P0 would only be used for multiplication
+  P = rcp_const_cast<Matrix>(rcpFromRef(P0));
 
-    for (size_t k = 0; k < nIts_; k++) {
-      AP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false, mmfancy, true, true);
+  for (size_t k = 0; k < nIts_; k++) {
+    AP = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false, mmfancy, true, true);
 #if 0
       // gradient = -2 A^T * A * P
       SC stepLength = 2*stepLength_;
       G = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, true, *AP, false, true, true);
       C.Apply(*G, *Ptmp);
 #else
-      // gradient = - A * P
-      SC stepLength = stepLength_;
-      Utilities::MyOldScaleMatrix(*AP, D, true, true, false);
-      C.Apply(*AP, *Ptmp);
+    // gradient = - A * P
+    SC stepLength = stepLength_;
+    Utilities::MyOldScaleMatrix(*AP, D, true, true, false);
+    C.Apply(*AP, *Ptmp);
 #endif
 
-      RCP<Matrix> newP;
-      Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Ptmp, false, -stepLength, *P, false, Teuchos::ScalarTraits<Scalar>::one(), newP, mmfancy);
-      newP->fillComplete(P->getDomainMap(), P->getRangeMap() );
-      P = newP;
-    }
+    RCP<Matrix> newP;
+    Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Ptmp, false, -stepLength, *P, false, Teuchos::ScalarTraits<Scalar>::one(), newP, mmfancy);
+    newP->fillComplete(P->getDomainMap(), P->getRangeMap());
+    P = newP;
   }
 }
+}  // namespace MueLu
 
-#endif //ifndef MUELU_STEEPESTDESCENTSOLVER_DECL_HPP
+#endif  // ifndef MUELU_STEEPESTDESCENTSOLVER_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_decl.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_decl.hpp
index 5e11741fcbe1..aebfe35e0b24 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_decl.hpp
@@ -58,10 +58,10 @@
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLuTests {
-  // Forward declaration of friend tester class used to UnitTest GeneralGeometricPFactory
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeneralGeometricPFactoryTester;
-}
+// Forward declaration of friend tester class used to UnitTest GeneralGeometricPFactory
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeneralGeometricPFactoryTester;
+}  // namespace MueLuTests
 
 namespace MueLu {
 
@@ -112,153 +112,152 @@ namespace MueLu {
   | coarseCoordinates | NoFactory                | Coarse coordinates that will be used on the next fine level to compute the coarsening stencils                   |
 
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class GeneralGeometricPFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GeneralGeometricPFactory : public PFactory {
 #undef MUELU_GENERALGEOMETRICPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    friend class MueLuTests::GeneralGeometricPFactoryTester<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor
-    GeneralGeometricPFactory() { }
-
-    //! Destructor.
-    virtual ~GeneralGeometricPFactory() { }
-    //@}
-
-    RCP<const ParameterList> GetValidParameterList() const;
-
-    //! Input
-    //@{
-
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
-
-    //@}
-
-    //! @name Build methods.
-    //@{
-
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
-
-    //@}
-
-  private:
-    struct GeometricData {
-      // Geometric algorithm require a copious amount of data to be passed around so this struct
-      // will reduce the amount of input/output parameters of methods in the class. Additionally
-      // the struct can be rewritten to accomodate constraints of Kokkos/CUDA data types
-
-      std::string meshLayout = "Global Lexicographic";
-      int numDimensions;
-      LO lNumFineNodes = -1, lNumCoarseNodes = -1, lNumGhostNodes = -1,lNumGhostedNodes  = -1;
-      LO myBlock = -1, numBlocks = -1, lNumFineNodes10 = -1;
-      GO gNumFineNodes = -1, gNumCoarseNodes = -1, gNumFineNodes10 = -1, minGlobalIndex = -1;
-      Array<int> coarseRate, endRate;
-      Array<LO> lFineNodesPerDir, lCoarseNodesPerDir, offsets, ghostedCoarseNodesPerDir;
-      Array<GO> startIndices, gFineNodesPerDir, gCoarseNodesPerDir, startGhostedCoarseNode;
-      std::vector<std::vector<GO> > meshData; // These are sorted later so they are in std::vector
-      bool ghostInterface[6] = {false};
-
-      GeometricData() {
-        coarseRate.resize(3);
-        endRate.resize(3);
-        lFineNodesPerDir.resize(3);
-        lCoarseNodesPerDir.resize(3);
-        offsets.resize(6);
-        ghostedCoarseNodesPerDir.resize(3);
-        startIndices.resize(6);
-        gFineNodesPerDir.resize(3);
-        gCoarseNodesPerDir.resize(3);
-        startGhostedCoarseNode.resize(3);
-      }
-    };
-
-    struct NodesIDs {
-      // This small struct just carries basic data associated with coarse nodes that is needed
-      // to compute colMapP and to fillComplete P,
-
-      Array<GO>  GIDs, coarseGIDs;
-      Array<int> PIDs;
-      Array<LO>  LIDs;
-      std::vector<GO> colInds;
-    };
-
-    struct NodeID {
-      // This small struct is similar to the one above but only for one node.
-      // It is used to create a vector of NodeID that can easily be sorted
-
-      GO  GID;
-      int PID;
-      LO  LID, lexiInd;
-    };
-
-    void MeshLayoutInterface(const int interpolationOrder, const LO blkSize,
-                             RCP<const Map> fineCoordsMap, RCP<GeometricData> myGeometry,
-                             RCP<NodesIDs> ghostedCoarseNodes,
-                             Array<Array<GO> >& lCoarseNodesGIDs) const;
-
-    void GetCoarsePoints(const int interpolationOrder, const LO blkSize,
-                         RCP<const Map> fineCoordsMap, RCP<GeometricData> myGeometry,
-                         RCP<NodesIDs> ghostedCoarseNodes,
-                         Array<Array<GO> >& lCoarseNodesGIDs) const;
-
-    void MakeGeneralGeometricP(RCP<GeometricData> myGeo,
-                               const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> >& fCoords,
-                               const LO nnzP, const LO dofsPerNode,
-                               RCP<const Map>& stridedDomainMapP,
-                               RCP<Matrix> & Amat, RCP<Matrix>& P,
-                               RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> >& cCoords,
-                               RCP<NodesIDs> ghostedCoarseNodes, Array<Array<GO> > coarseNodesGIDs,
-                               int interpolationOrder) const;
-
-    void ComputeStencil(const LO numDimension, const Array<GO> currentNodeIndices,
-                        const Array<GO> coarseNodeIndices, const LO rate[3],
-                        const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord, const int interpolationOrder,
-                        std::vector<double>& stencil) const;
-
-    void ComputeConstantInterpolationStencil(const LO numDimension,
-                                             const Array<GO> currentNodeIndices,
-                                             const Array<GO> coarseNodeIndices,
-                                             const LO rate[3], std::vector<double>& stencil) const;
-
-    void ComputeLinearInterpolationStencil(const LO numDimension, const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord,
-                                           std::vector<double>& stencil) const;
-    void GetInterpolationFunctions(const LO numDimension,
-                                   const Teuchos::SerialDenseVector<LO,double> parameters,
-                                   double functions[4][8]) const;
-
-    void sh_sort_permute(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
-
-    void sh_sort2(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
-
-    void GetGIDLocalLexicographic(const GO i, const GO j, const GO k,
-                                  const Array<LO> coarseNodeFineIndices,
-                                  const RCP<GeometricData> myGeo, const LO myRankIndex, const LO pi,
-                                  const LO pj, const LO pk,
-                                  const typename std::vector<std::vector<GO> >::iterator blockStart,
-                                  const typename std::vector<std::vector<GO> >::iterator blockEnd,
-                                  GO& myGID, LO& myPID, LO& myLID) const;
-
-  }; //class GeneralGeometricPFactory
-
-} //namespace MueLu
+ public:
+  friend class MueLuTests::GeneralGeometricPFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor
+  GeneralGeometricPFactory() {}
+
+  //! Destructor.
+  virtual ~GeneralGeometricPFactory() {}
+  //@}
+
+  RCP<const ParameterList> GetValidParameterList() const;
+
+  //! Input
+  //@{
+
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+
+  //@}
+
+  //! @name Build methods.
+  //@{
+
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
+
+  //@}
+
+ private:
+  struct GeometricData {
+    // Geometric algorithm require a copious amount of data to be passed around so this struct
+    // will reduce the amount of input/output parameters of methods in the class. Additionally
+    // the struct can be rewritten to accomodate constraints of Kokkos/CUDA data types
+
+    std::string meshLayout = "Global Lexicographic";
+    int numDimensions;
+    LO lNumFineNodes = -1, lNumCoarseNodes = -1, lNumGhostNodes = -1, lNumGhostedNodes = -1;
+    LO myBlock = -1, numBlocks = -1, lNumFineNodes10 = -1;
+    GO gNumFineNodes = -1, gNumCoarseNodes = -1, gNumFineNodes10 = -1, minGlobalIndex = -1;
+    Array<int> coarseRate, endRate;
+    Array<LO> lFineNodesPerDir, lCoarseNodesPerDir, offsets, ghostedCoarseNodesPerDir;
+    Array<GO> startIndices, gFineNodesPerDir, gCoarseNodesPerDir, startGhostedCoarseNode;
+    std::vector<std::vector<GO> > meshData;  // These are sorted later so they are in std::vector
+    bool ghostInterface[6] = {false};
+
+    GeometricData() {
+      coarseRate.resize(3);
+      endRate.resize(3);
+      lFineNodesPerDir.resize(3);
+      lCoarseNodesPerDir.resize(3);
+      offsets.resize(6);
+      ghostedCoarseNodesPerDir.resize(3);
+      startIndices.resize(6);
+      gFineNodesPerDir.resize(3);
+      gCoarseNodesPerDir.resize(3);
+      startGhostedCoarseNode.resize(3);
+    }
+  };
+
+  struct NodesIDs {
+    // This small struct just carries basic data associated with coarse nodes that is needed
+    // to compute colMapP and to fillComplete P,
+
+    Array<GO> GIDs, coarseGIDs;
+    Array<int> PIDs;
+    Array<LO> LIDs;
+    std::vector<GO> colInds;
+  };
+
+  struct NodeID {
+    // This small struct is similar to the one above but only for one node.
+    // It is used to create a vector of NodeID that can easily be sorted
+
+    GO GID;
+    int PID;
+    LO LID, lexiInd;
+  };
+
+  void MeshLayoutInterface(const int interpolationOrder, const LO blkSize,
+                           RCP<const Map> fineCoordsMap, RCP<GeometricData> myGeometry,
+                           RCP<NodesIDs> ghostedCoarseNodes,
+                           Array<Array<GO> >& lCoarseNodesGIDs) const;
+
+  void GetCoarsePoints(const int interpolationOrder, const LO blkSize,
+                       RCP<const Map> fineCoordsMap, RCP<GeometricData> myGeometry,
+                       RCP<NodesIDs> ghostedCoarseNodes,
+                       Array<Array<GO> >& lCoarseNodesGIDs) const;
+
+  void MakeGeneralGeometricP(RCP<GeometricData> myGeo,
+                             const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> >& fCoords,
+                             const LO nnzP, const LO dofsPerNode,
+                             RCP<const Map>& stridedDomainMapP,
+                             RCP<Matrix>& Amat, RCP<Matrix>& P,
+                             RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> >& cCoords,
+                             RCP<NodesIDs> ghostedCoarseNodes, Array<Array<GO> > coarseNodesGIDs,
+                             int interpolationOrder) const;
+
+  void ComputeStencil(const LO numDimension, const Array<GO> currentNodeIndices,
+                      const Array<GO> coarseNodeIndices, const LO rate[3],
+                      const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord, const int interpolationOrder,
+                      std::vector<double>& stencil) const;
+
+  void ComputeConstantInterpolationStencil(const LO numDimension,
+                                           const Array<GO> currentNodeIndices,
+                                           const Array<GO> coarseNodeIndices,
+                                           const LO rate[3], std::vector<double>& stencil) const;
+
+  void ComputeLinearInterpolationStencil(const LO numDimension, const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord,
+                                         std::vector<double>& stencil) const;
+  void GetInterpolationFunctions(const LO numDimension,
+                                 const Teuchos::SerialDenseVector<LO, double> parameters,
+                                 double functions[4][8]) const;
+
+  void sh_sort_permute(
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
+
+  void sh_sort2(
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+      const typename Teuchos::Array<LocalOrdinal>::iterator& last2) const;
+
+  void GetGIDLocalLexicographic(const GO i, const GO j, const GO k,
+                                const Array<LO> coarseNodeFineIndices,
+                                const RCP<GeometricData> myGeo, const LO myRankIndex, const LO pi,
+                                const LO pj, const LO pk,
+                                const typename std::vector<std::vector<GO> >::iterator blockStart,
+                                const typename std::vector<std::vector<GO> >::iterator blockEnd,
+                                GO& myGID, LO& myPID, LO& myLID) const;
+
+};  // class GeneralGeometricPFactory
+
+}  // namespace MueLu
 
 #define MUELU_GENERALGEOMETRICPFACTORY_SHORT
-#endif // MUELU_GENERALGEOMETRICPFACTORY_DECL_HPP
+#endif  // MUELU_GENERALGEOMETRICPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_def.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_def.hpp
index 1c1a564abda5..5c1c57c66ade 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeneralGeometricPFactory_def.hpp
@@ -49,7 +49,6 @@
 #include <stdlib.h>
 #include <iomanip>
 
-
 // #include <Teuchos_LAPACK.hpp>
 #include <Teuchos_SerialDenseMatrix.hpp>
 #include <Teuchos_SerialDenseVector.hpp>
@@ -70,1848 +69,1792 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    // Coarsen can come in two forms, either a single char that will be interpreted as an integer
-    // which is used as the coarsening rate in every spatial dimentions, or it can be a longer
-    // string that will then be interpreted as an array of integers.
-    // By default coarsen is set as "{2}", hence a coarsening rate of 2 in every spatial dimension
-    // is the default setting!
-    validParamList->set<std::string >           ("Coarsen",                 "{2}",
-                                                 "Coarsening rate in all spatial dimensions");
-    validParamList->set<int>                    ("order",                   1,
-                                                 "Order of the interpolation scheme used");
-    validParamList->set<RCP<const FactoryBase> >("A",                       Teuchos::null,
-                                                 "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",               Teuchos::null,
-                                                 "Generating factory of the nullspace");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",             Teuchos::null,
-                                                 "Generating factory for coorindates");
-    validParamList->set<RCP<const FactoryBase> >("gNodesPerDim",            Teuchos::null,
-                                                 "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",            Teuchos::null,
-                                                 "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
-    validParamList->set<std::string >           ("meshLayout",              "Global Lexicographic",
-                                                 "Type of mesh ordering");
-    validParamList->set<RCP<const FactoryBase> >("meshData",                Teuchos::null,
-                                                 "Mesh ordering associated data");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "Coordinates");
-    // Request the global number of nodes per dimensions
-    if(fineLevel.GetLevelID() == 0) {
-      if(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
-        fineLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "gNodesPerDim was not provided by the user on level0!");
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  // Coarsen can come in two forms, either a single char that will be interpreted as an integer
+  // which is used as the coarsening rate in every spatial dimentions, or it can be a longer
+  // string that will then be interpreted as an array of integers.
+  // By default coarsen is set as "{2}", hence a coarsening rate of 2 in every spatial dimension
+  // is the default setting!
+  validParamList->set<std::string>("Coarsen", "{2}",
+                                   "Coarsening rate in all spatial dimensions");
+  validParamList->set<int>("order", 1,
+                           "Order of the interpolation scheme used");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null,
+                                               "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null,
+                                               "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null,
+                                               "Generating factory for coorindates");
+  validParamList->set<RCP<const FactoryBase> >("gNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimmension provided by CoordinatesTransferFactory.");
+  validParamList->set<std::string>("meshLayout", "Global Lexicographic",
+                                   "Type of mesh ordering");
+  validParamList->set<RCP<const FactoryBase> >("meshData", Teuchos::null,
+                                               "Mesh ordering associated data");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "Coordinates");
+  // Request the global number of nodes per dimensions
+  if (fineLevel.GetLevelID() == 0) {
+    if (fineLevel.IsAvailable("gNodesPerDim", NoFactory::get())) {
+      fineLevel.DeclareInput("gNodesPerDim", NoFactory::get(), this);
     } else {
-      Input(fineLevel, "gNodesPerDim");
+      TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("gNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "gNodesPerDim was not provided by the user on level0!");
     }
+  } else {
+    Input(fineLevel, "gNodesPerDim");
+  }
 
-    // Request the local number of nodes per dimensions
-    if(fineLevel.GetLevelID() == 0) {
-      if(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
-        fineLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
-                                   Exceptions::RuntimeError,
-                                   "lNodesPerDim was not provided by the user on level0!");
-      }
+  // Request the local number of nodes per dimensions
+  if (fineLevel.GetLevelID() == 0) {
+    if (fineLevel.IsAvailable("lNodesPerDim", NoFactory::get())) {
+      fineLevel.DeclareInput("lNodesPerDim", NoFactory::get(), this);
     } else {
-      Input(fineLevel, "lNodesPerDim");
+      TEUCHOS_TEST_FOR_EXCEPTION(fineLevel.IsAvailable("lNodesPerDim", NoFactory::get()),
+                                 Exceptions::RuntimeError,
+                                 "lNodesPerDim was not provided by the user on level0!");
     }
+  } else {
+    Input(fineLevel, "lNodesPerDim");
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
-                                Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
-                                Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    // Obtain general variables
-    using xdMV = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>;
-    RCP<Matrix>      A             = Get< RCP<Matrix> >      (fineLevel, "A");
-    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-    RCP<xdMV>        fineCoords    = Get< RCP<xdMV> >(fineLevel, "Coordinates");
-    RCP<xdMV>        coarseCoords;
-
-    // Get user-provided coarsening rate parameter (constant over all levels)
-    const ParameterList& pL = GetParameterList();
-
-    // collect general input data
-    const LO blkSize      = A->GetFixedBlockSize();
-    RCP<const Map> rowMap = A->getRowMap();
-    RCP<GeometricData> myGeometry = rcp(new GeometricData{});
-
-    // Load the mesh layout type and the associated mesh data
-    myGeometry->meshLayout = pL.get<std::string>("meshLayout");
-    if(fineLevel.GetLevelID() == 0) {
-      if(myGeometry->meshLayout == "Local Lexicographic") {
-        Array<GO> meshData;
-        meshData = fineLevel.Get<Array<GO> >("meshData", NoFactory::get());
-        TEUCHOS_TEST_FOR_EXCEPTION(meshData.empty() == true, Exceptions::RuntimeError,
-                                   "The meshData array is empty, somehow the input for geometric"
-                                   " multigrid are not captured correctly.");
-        myGeometry->meshData.resize(rowMap->getComm()->getSize());
-        for(int i = 0; i < rowMap->getComm()->getSize(); ++i) {
-          myGeometry->meshData[i].resize(10);
-          for(int j = 0; j < 10; ++j) {
-            myGeometry->meshData[i][j] = meshData[10*i + j];
-          }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
+                                                                                Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
+                                                                                 Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  // Obtain general variables
+  using xdMV                     = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>;
+  RCP<Matrix> A                  = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<MultiVector> fineNullspace = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+  RCP<xdMV> fineCoords           = Get<RCP<xdMV> >(fineLevel, "Coordinates");
+  RCP<xdMV> coarseCoords;
+
+  // Get user-provided coarsening rate parameter (constant over all levels)
+  const ParameterList& pL = GetParameterList();
+
+  // collect general input data
+  const LO blkSize              = A->GetFixedBlockSize();
+  RCP<const Map> rowMap         = A->getRowMap();
+  RCP<GeometricData> myGeometry = rcp(new GeometricData{});
+
+  // Load the mesh layout type and the associated mesh data
+  myGeometry->meshLayout = pL.get<std::string>("meshLayout");
+  if (fineLevel.GetLevelID() == 0) {
+    if (myGeometry->meshLayout == "Local Lexicographic") {
+      Array<GO> meshData;
+      meshData = fineLevel.Get<Array<GO> >("meshData", NoFactory::get());
+      TEUCHOS_TEST_FOR_EXCEPTION(meshData.empty() == true, Exceptions::RuntimeError,
+                                 "The meshData array is empty, somehow the input for geometric"
+                                 " multigrid are not captured correctly.");
+      myGeometry->meshData.resize(rowMap->getComm()->getSize());
+      for (int i = 0; i < rowMap->getComm()->getSize(); ++i) {
+        myGeometry->meshData[i].resize(10);
+        for (int j = 0; j < 10; ++j) {
+          myGeometry->meshData[i][j] = meshData[10 * i + j];
         }
       }
     }
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords == Teuchos::null, Exceptions::RuntimeError,
-                               "Coordinates cannot be accessed from fine level!");
-    myGeometry->numDimensions = fineCoords->getNumVectors();
+  TEUCHOS_TEST_FOR_EXCEPTION(fineCoords == Teuchos::null, Exceptions::RuntimeError,
+                             "Coordinates cannot be accessed from fine level!");
+  myGeometry->numDimensions = fineCoords->getNumVectors();
 
-    // Get the number of points in each direction
-    if(fineLevel.GetLevelID() == 0) {
-      myGeometry->gFineNodesPerDir = fineLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
-      myGeometry->lFineNodesPerDir = fineLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
-    } else {
-      // Loading global number of nodes per diretions
-      myGeometry->gFineNodesPerDir = Get<Array<GO> >(fineLevel, "gNodesPerDim");
+  // Get the number of points in each direction
+  if (fineLevel.GetLevelID() == 0) {
+    myGeometry->gFineNodesPerDir = fineLevel.Get<Array<GO> >("gNodesPerDim", NoFactory::get());
+    myGeometry->lFineNodesPerDir = fineLevel.Get<Array<LO> >("lNodesPerDim", NoFactory::get());
+  } else {
+    // Loading global number of nodes per diretions
+    myGeometry->gFineNodesPerDir = Get<Array<GO> >(fineLevel, "gNodesPerDim");
 
-      // Loading local number of nodes per diretions
-      myGeometry->lFineNodesPerDir = Get<Array<LO> >(fineLevel, "lNodesPerDim");
-    }
-    myGeometry->gNumFineNodes10 = myGeometry->gFineNodesPerDir[1]*myGeometry->gFineNodesPerDir[0];
-    myGeometry->gNumFineNodes   = myGeometry->gFineNodesPerDir[2]*myGeometry->gNumFineNodes10;
-    myGeometry->lNumFineNodes10 = myGeometry->lFineNodesPerDir[1]*myGeometry->lFineNodesPerDir[0];
-    myGeometry->lNumFineNodes   = myGeometry->lFineNodesPerDir[2]*myGeometry->lNumFineNodes10;
-
-    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getLocalLength()
-                               != static_cast<size_t>(myGeometry->lNumFineNodes),
-                               Exceptions::RuntimeError,
-                               "The local number of elements in Coordinates is not equal to the"
-                               " number of nodes given by: lNodesPerDim!");
-    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getGlobalLength()
-                               != static_cast<size_t>(myGeometry->gNumFineNodes),
-                               Exceptions::RuntimeError,
-                               "The global number of elements in Coordinates is not equal to the"
-                               " number of nodes given by: gNodesPerDim!");
-
-    // Get the coarsening rate
-    std::string coarsenRate = pL.get<std::string>("Coarsen");
-    Teuchos::Array<LO> crates;
-    try {
-      crates = Teuchos::fromStringToArray<LO>(coarsenRate);
-    } catch(const Teuchos::InvalidArrayStringRepresentation& e) {
-      GetOStream(Errors,-1) << " *** Coarsen must be a string convertible into an array! *** "
-                            << std::endl;
-      throw e;
-    }
-    TEUCHOS_TEST_FOR_EXCEPTION((crates.size() > 1) && (crates.size() < myGeometry->numDimensions),
-                               Exceptions::RuntimeError,
-                               "Coarsen must have at least as many components as the number of"
-                               " spatial dimensions in the problem.");
-
-    for(LO i = 0; i < 3; ++i) {
-      if(i < myGeometry->numDimensions) {
-        if(crates.size()==1) {
-          myGeometry->coarseRate[i] = crates[0];
-        } else if(crates.size() == myGeometry->numDimensions) {
-          myGeometry->coarseRate[i] = crates[i];
-        }
-      } else {
-        myGeometry->coarseRate[i] = 1;
+    // Loading local number of nodes per diretions
+    myGeometry->lFineNodesPerDir = Get<Array<LO> >(fineLevel, "lNodesPerDim");
+  }
+  myGeometry->gNumFineNodes10 = myGeometry->gFineNodesPerDir[1] * myGeometry->gFineNodesPerDir[0];
+  myGeometry->gNumFineNodes   = myGeometry->gFineNodesPerDir[2] * myGeometry->gNumFineNodes10;
+  myGeometry->lNumFineNodes10 = myGeometry->lFineNodesPerDir[1] * myGeometry->lFineNodesPerDir[0];
+  myGeometry->lNumFineNodes   = myGeometry->lFineNodesPerDir[2] * myGeometry->lNumFineNodes10;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getLocalLength() != static_cast<size_t>(myGeometry->lNumFineNodes),
+                             Exceptions::RuntimeError,
+                             "The local number of elements in Coordinates is not equal to the"
+                             " number of nodes given by: lNodesPerDim!");
+  TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getGlobalLength() != static_cast<size_t>(myGeometry->gNumFineNodes),
+                             Exceptions::RuntimeError,
+                             "The global number of elements in Coordinates is not equal to the"
+                             " number of nodes given by: gNodesPerDim!");
+
+  // Get the coarsening rate
+  std::string coarsenRate = pL.get<std::string>("Coarsen");
+  Teuchos::Array<LO> crates;
+  try {
+    crates = Teuchos::fromStringToArray<LO>(coarsenRate);
+  } catch (const Teuchos::InvalidArrayStringRepresentation& e) {
+    GetOStream(Errors, -1) << " *** Coarsen must be a string convertible into an array! *** "
+                           << std::endl;
+    throw e;
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION((crates.size() > 1) && (crates.size() < myGeometry->numDimensions),
+                             Exceptions::RuntimeError,
+                             "Coarsen must have at least as many components as the number of"
+                             " spatial dimensions in the problem.");
+
+  for (LO i = 0; i < 3; ++i) {
+    if (i < myGeometry->numDimensions) {
+      if (crates.size() == 1) {
+        myGeometry->coarseRate[i] = crates[0];
+      } else if (crates.size() == myGeometry->numDimensions) {
+        myGeometry->coarseRate[i] = crates[i];
       }
-    }
-
-    int interpolationOrder = pL.get<int>("order");
-    TEUCHOS_TEST_FOR_EXCEPTION((interpolationOrder < 0) || (interpolationOrder > 1),
-                               Exceptions::RuntimeError,
-                               "The interpolation order can only be set to 0 or 1.");
-
-    // Get the axis permutation from Global axis to Local axis
-    Array<LO> mapDirG2L(3), mapDirL2G(3);
-    for(LO i = 0; i < myGeometry->numDimensions; ++i) {
-      mapDirG2L[i] = i;
-    }
-    for(LO i = 0; i < myGeometry->numDimensions; ++i) {
-      TEUCHOS_TEST_FOR_EXCEPTION(mapDirG2L[i] > myGeometry->numDimensions,
-                                 Exceptions::RuntimeError,
-                                 "axis permutation values must all be less than"
-                                 " the number of spatial dimensions.");
-      mapDirL2G[mapDirG2L[i]] = i;
-    }
-    RCP<const Map> fineCoordsMap = fineCoords->getMap();
-
-    // This struct stores PIDs, LIDs and GIDs on the fine mesh and GIDs on the coarse mesh.
-    RCP<NodesIDs> ghostedCoarseNodes = rcp(new NodesIDs{});
-    Array<Array<GO> > lCoarseNodesGIDs(2);
-    if((fineLevel.GetLevelID() == 0) && (myGeometry->meshLayout != "Global Lexicographic")) {
-      MeshLayoutInterface(interpolationOrder, blkSize, fineCoordsMap, myGeometry,
-                          ghostedCoarseNodes, lCoarseNodesGIDs);
     } else {
-      // This function expects perfect global lexicographic ordering of nodes and will not process
-      // data correctly otherwise. These restrictions allow for the simplest and most efficient
-      // processing of the levels (hopefully at least).
-      GetCoarsePoints(interpolationOrder, blkSize, fineCoordsMap, myGeometry, ghostedCoarseNodes,
-                      lCoarseNodesGIDs);
-    }
-
-    // All that is left to do is loop over NCpts and:
-    //   - extract coarse points coordiate for coarseCoords
-    //   - get coordinates for current stencil computation
-    //   - compute current stencil
-    //   - compute row and column indices for stencil entries
-    RCP<const Map> stridedDomainMapP;
-    RCP<Matrix>    P;
-    // Fancy formula for the number of non-zero terms
-    // All coarse points are injected, other points are using polynomial interpolation
-    // and have contribution from (interpolationOrder + 1)^numDimensions
-    // Noticebly this leads to 1 when the order is zero, hence fancy MatMatMatMult can be used.
-    GO lnnzP = Teuchos::as<LO>(std::pow(interpolationOrder + 1, myGeometry->numDimensions))
-      *(myGeometry->lNumFineNodes - myGeometry->lNumCoarseNodes) + myGeometry->lNumCoarseNodes;
-    lnnzP = lnnzP*blkSize;
-    GO gnnzP = Teuchos::as<LO>(std::pow(interpolationOrder + 1, myGeometry->numDimensions))
-      *(myGeometry->gNumFineNodes - myGeometry->gNumCoarseNodes) + myGeometry->gNumCoarseNodes;
-    gnnzP = gnnzP*blkSize;
-
-    GetOStream(Runtime1) << "P size = " << blkSize*myGeometry->gNumFineNodes
-                         << " x " << blkSize*myGeometry->gNumCoarseNodes << std::endl;
-    GetOStream(Runtime1) << "P Fine   grid : " << myGeometry->gFineNodesPerDir[0] << " -- "
-                         << myGeometry->gFineNodesPerDir[1] << " -- "
-                         << myGeometry->gFineNodesPerDir[2] << std::endl;
-    GetOStream(Runtime1) << "P Coarse grid : " << myGeometry->gCoarseNodesPerDir[0] << " -- "
-                         << myGeometry->gCoarseNodesPerDir[1] << " -- "
-                         << myGeometry->gCoarseNodesPerDir[2] << std::endl;
-    GetOStream(Runtime1) << "P nnz estimate: " << gnnzP << std::endl;
-
-    MakeGeneralGeometricP(myGeometry, fineCoords, lnnzP, blkSize, stridedDomainMapP,
-                           A, P, coarseCoords, ghostedCoarseNodes, lCoarseNodesGIDs,
-                           interpolationOrder);
-
-    // set StridingInformation of P
-    if (A->IsView("stridedMaps") == true) {
-      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMapP);
-    } else {
-      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMapP);
+      myGeometry->coarseRate[i] = 1;
     }
+  }
 
-    // store the transfer operator and node coordinates on coarse level
-    Set(coarseLevel, "P", P);
-    Set(coarseLevel, "coarseCoordinates", coarseCoords);
-    Set<Array<GO> >(coarseLevel, "gCoarseNodesPerDim", myGeometry->gCoarseNodesPerDir);
-    Set<Array<LO> >(coarseLevel, "lCoarseNodesPerDim", myGeometry->lCoarseNodesPerDir);
+  int interpolationOrder = pL.get<int>("order");
+  TEUCHOS_TEST_FOR_EXCEPTION((interpolationOrder < 0) || (interpolationOrder > 1),
+                             Exceptions::RuntimeError,
+                             "The interpolation order can only be set to 0 or 1.");
 
-    // rst: null space might get scaled here ... do we care. We could just inject at the cpoints,
-    // but I don't feel that this is needed.
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
-                                                                 fineNullspace->getNumVectors());
-    P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
-             Teuchos::ScalarTraits<SC>::zero());
-    Set(coarseLevel, "Nullspace", coarseNullspace);
+  // Get the axis permutation from Global axis to Local axis
+  Array<LO> mapDirG2L(3), mapDirL2G(3);
+  for (LO i = 0; i < myGeometry->numDimensions; ++i) {
+    mapDirG2L[i] = i;
+  }
+  for (LO i = 0; i < myGeometry->numDimensions; ++i) {
+    TEUCHOS_TEST_FOR_EXCEPTION(mapDirG2L[i] > myGeometry->numDimensions,
+                               Exceptions::RuntimeError,
+                               "axis permutation values must all be less than"
+                               " the number of spatial dimensions.");
+    mapDirL2G[mapDirG2L[i]] = i;
+  }
+  RCP<const Map> fineCoordsMap = fineCoords->getMap();
+
+  // This struct stores PIDs, LIDs and GIDs on the fine mesh and GIDs on the coarse mesh.
+  RCP<NodesIDs> ghostedCoarseNodes = rcp(new NodesIDs{});
+  Array<Array<GO> > lCoarseNodesGIDs(2);
+  if ((fineLevel.GetLevelID() == 0) && (myGeometry->meshLayout != "Global Lexicographic")) {
+    MeshLayoutInterface(interpolationOrder, blkSize, fineCoordsMap, myGeometry,
+                        ghostedCoarseNodes, lCoarseNodesGIDs);
+  } else {
+    // This function expects perfect global lexicographic ordering of nodes and will not process
+    // data correctly otherwise. These restrictions allow for the simplest and most efficient
+    // processing of the levels (hopefully at least).
+    GetCoarsePoints(interpolationOrder, blkSize, fineCoordsMap, myGeometry, ghostedCoarseNodes,
+                    lCoarseNodesGIDs);
+  }
 
+  // All that is left to do is loop over NCpts and:
+  //   - extract coarse points coordiate for coarseCoords
+  //   - get coordinates for current stencil computation
+  //   - compute current stencil
+  //   - compute row and column indices for stencil entries
+  RCP<const Map> stridedDomainMapP;
+  RCP<Matrix> P;
+  // Fancy formula for the number of non-zero terms
+  // All coarse points are injected, other points are using polynomial interpolation
+  // and have contribution from (interpolationOrder + 1)^numDimensions
+  // Noticebly this leads to 1 when the order is zero, hence fancy MatMatMatMult can be used.
+  GO lnnzP = Teuchos::as<LO>(std::pow(interpolationOrder + 1, myGeometry->numDimensions)) * (myGeometry->lNumFineNodes - myGeometry->lNumCoarseNodes) + myGeometry->lNumCoarseNodes;
+  lnnzP    = lnnzP * blkSize;
+  GO gnnzP = Teuchos::as<LO>(std::pow(interpolationOrder + 1, myGeometry->numDimensions)) * (myGeometry->gNumFineNodes - myGeometry->gNumCoarseNodes) + myGeometry->gNumCoarseNodes;
+  gnnzP    = gnnzP * blkSize;
+
+  GetOStream(Runtime1) << "P size = " << blkSize * myGeometry->gNumFineNodes
+                       << " x " << blkSize * myGeometry->gNumCoarseNodes << std::endl;
+  GetOStream(Runtime1) << "P Fine   grid : " << myGeometry->gFineNodesPerDir[0] << " -- "
+                       << myGeometry->gFineNodesPerDir[1] << " -- "
+                       << myGeometry->gFineNodesPerDir[2] << std::endl;
+  GetOStream(Runtime1) << "P Coarse grid : " << myGeometry->gCoarseNodesPerDir[0] << " -- "
+                       << myGeometry->gCoarseNodesPerDir[1] << " -- "
+                       << myGeometry->gCoarseNodesPerDir[2] << std::endl;
+  GetOStream(Runtime1) << "P nnz estimate: " << gnnzP << std::endl;
+
+  MakeGeneralGeometricP(myGeometry, fineCoords, lnnzP, blkSize, stridedDomainMapP,
+                        A, P, coarseCoords, ghostedCoarseNodes, lCoarseNodesGIDs,
+                        interpolationOrder);
+
+  // set StridingInformation of P
+  if (A->IsView("stridedMaps") == true) {
+    P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMapP);
+  } else {
+    P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMapP);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  MeshLayoutInterface(const int /* interpolationOrder */, const LO /* blkSize */, RCP<const Map> fineCoordsMap,
-                      RCP<GeometricData> myGeo, RCP<NodesIDs> ghostedCoarseNodes,
-                      Array<Array<GO> >& lCoarseNodesGIDs) const{
-    // The goal here is to produce maps that globally labels the mesh lexicographically.
-    // These maps will replace the current maps of A, the coordinate vector and the nullspace.
-    // Ideally if the user provides the necessary maps then nothing needs to be done, otherwise
-    // it could be advantageous to allow the user to register a re-labeling function. Ultimately
-    // for common ordering schemes, some re-labeling can be directly implemented here.
-
-    int numRanks = fineCoordsMap->getComm()->getSize();
-    int myRank   = fineCoordsMap->getComm()->getRank();
-
-    myGeo->myBlock = myGeo->meshData[myRank][2];
-    myGeo->startIndices[0] = myGeo->meshData[myRank][3];
-    myGeo->startIndices[1] = myGeo->meshData[myRank][5];
-    myGeo->startIndices[2] = myGeo->meshData[myRank][7];
-    myGeo->startIndices[3] = myGeo->meshData[myRank][4];
-    myGeo->startIndices[4] = myGeo->meshData[myRank][6];
-    myGeo->startIndices[5] = myGeo->meshData[myRank][8];
-    std::sort(myGeo->meshData.begin(), myGeo->meshData.end(),
-              [](const std::vector<GO>& a, const std::vector<GO>& b)->bool {
-                // The below function sorts ranks by blockID, kmin, jmin and imin
-                if(a[2] < b[2]) {
+  // store the transfer operator and node coordinates on coarse level
+  Set(coarseLevel, "P", P);
+  Set(coarseLevel, "coarseCoordinates", coarseCoords);
+  Set<Array<GO> >(coarseLevel, "gCoarseNodesPerDim", myGeometry->gCoarseNodesPerDir);
+  Set<Array<LO> >(coarseLevel, "lCoarseNodesPerDim", myGeometry->lCoarseNodesPerDir);
+
+  // rst: null space might get scaled here ... do we care. We could just inject at the cpoints,
+  // but I don't feel that this is needed.
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
+                                                               fineNullspace->getNumVectors());
+  P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
+           Teuchos::ScalarTraits<SC>::zero());
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    MeshLayoutInterface(const int /* interpolationOrder */, const LO /* blkSize */, RCP<const Map> fineCoordsMap,
+                        RCP<GeometricData> myGeo, RCP<NodesIDs> ghostedCoarseNodes,
+                        Array<Array<GO> >& lCoarseNodesGIDs) const {
+  // The goal here is to produce maps that globally labels the mesh lexicographically.
+  // These maps will replace the current maps of A, the coordinate vector and the nullspace.
+  // Ideally if the user provides the necessary maps then nothing needs to be done, otherwise
+  // it could be advantageous to allow the user to register a re-labeling function. Ultimately
+  // for common ordering schemes, some re-labeling can be directly implemented here.
+
+  int numRanks = fineCoordsMap->getComm()->getSize();
+  int myRank   = fineCoordsMap->getComm()->getRank();
+
+  myGeo->myBlock         = myGeo->meshData[myRank][2];
+  myGeo->startIndices[0] = myGeo->meshData[myRank][3];
+  myGeo->startIndices[1] = myGeo->meshData[myRank][5];
+  myGeo->startIndices[2] = myGeo->meshData[myRank][7];
+  myGeo->startIndices[3] = myGeo->meshData[myRank][4];
+  myGeo->startIndices[4] = myGeo->meshData[myRank][6];
+  myGeo->startIndices[5] = myGeo->meshData[myRank][8];
+  std::sort(myGeo->meshData.begin(), myGeo->meshData.end(),
+            [](const std::vector<GO>& a, const std::vector<GO>& b) -> bool {
+              // The below function sorts ranks by blockID, kmin, jmin and imin
+              if (a[2] < b[2]) {
+                return true;
+              } else if (a[2] == b[2]) {
+                if (a[7] < b[7]) {
                   return true;
-                } else if(a[2] == b[2]) {
-                  if(a[7] < b[7]) {
+                } else if (a[7] == b[7]) {
+                  if (a[5] < b[5]) {
                     return true;
-                  } else if(a[7] == b[7]) {
-                    if(a[5] < b[5]) {
+                  } else if (a[5] == b[5]) {
+                    if (a[3] < b[3]) {
                       return true;
-                    } else if(a[5] == b[5]) {
-                      if(a[3] < b[3]) {return true;}
                     }
                   }
                 }
-                return false;
-              });
-
-    myGeo->numBlocks = myGeo->meshData[numRanks - 1][2] + 1;
-    // Find the range of the current block
-    auto myBlockStart = std::lower_bound(myGeo->meshData.begin(), myGeo->meshData.end(),
-                                         myGeo->myBlock - 1,
-                                         [](const std::vector<GO>& vec, const GO val)->bool{
-                                           return (vec[2] < val) ? true : false;
-                                         });
-    auto myBlockEnd = std::upper_bound(myGeo->meshData.begin(), myGeo->meshData.end(),
-                                       myGeo->myBlock,
-                                       [](const GO val, const std::vector<GO>& vec)->bool{
-                                         return (val < vec[2]) ? true : false;
+              }
+              return false;
+            });
+
+  myGeo->numBlocks = myGeo->meshData[numRanks - 1][2] + 1;
+  // Find the range of the current block
+  auto myBlockStart = std::lower_bound(myGeo->meshData.begin(), myGeo->meshData.end(),
+                                       myGeo->myBlock - 1,
+                                       [](const std::vector<GO>& vec, const GO val) -> bool {
+                                         return (vec[2] < val) ? true : false;
                                        });
-    // Assuming that i,j,k and ranges are split in pi, pj and pk processors
-    // we search for these numbers as they will allow us to find quickly the PID of processors
-    // owning ghost nodes.
-    auto myKEnd = std::upper_bound(myBlockStart, myBlockEnd, (*myBlockStart)[3],
-                                   [](const GO val, const std::vector<GO>& vec)->bool{
-                                     return (val < vec[7]) ? true : false;
-                                   });
-    auto myJEnd = std::upper_bound(myBlockStart, myKEnd, (*myBlockStart)[3],
-                                   [](const GO val, const std::vector<GO>& vec)->bool{
-                                     return (val < vec[5]) ? true : false;
-                                   });
-    LO pi = std::distance(myBlockStart, myJEnd);
-    LO pj = std::distance(myBlockStart, myKEnd) / pi;
-    LO pk = std::distance(myBlockStart, myBlockEnd) / (pj*pi);
-
-    // We also look for the index of the local rank in the current block.
-    LO myRankIndex = std::distance(myGeo->meshData.begin(),
-                                   std::find_if(myBlockStart, myBlockEnd,
-                                                [myRank](const std::vector<GO>& vec)->bool{
-                                                  return (vec[0] == myRank) ? true : false;
-                                                })
-                                   );
-
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < myGeo->numDimensions) {
-        myGeo->offsets[dim]= Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
-        myGeo->offsets[dim + 3]= Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
-      }
+  auto myBlockEnd   = std::upper_bound(myGeo->meshData.begin(), myGeo->meshData.end(),
+                                       myGeo->myBlock,
+                                       [](const GO val, const std::vector<GO>& vec) -> bool {
+                                       return (val < vec[2]) ? true : false;
+                                     });
+  // Assuming that i,j,k and ranges are split in pi, pj and pk processors
+  // we search for these numbers as they will allow us to find quickly the PID of processors
+  // owning ghost nodes.
+  auto myKEnd = std::upper_bound(myBlockStart, myBlockEnd, (*myBlockStart)[3],
+                                 [](const GO val, const std::vector<GO>& vec) -> bool {
+                                   return (val < vec[7]) ? true : false;
+                                 });
+  auto myJEnd = std::upper_bound(myBlockStart, myKEnd, (*myBlockStart)[3],
+                                 [](const GO val, const std::vector<GO>& vec) -> bool {
+                                   return (val < vec[5]) ? true : false;
+                                 });
+  LO pi       = std::distance(myBlockStart, myJEnd);
+  LO pj       = std::distance(myBlockStart, myKEnd) / pi;
+  LO pk       = std::distance(myBlockStart, myBlockEnd) / (pj * pi);
+
+  // We also look for the index of the local rank in the current block.
+  LO myRankIndex = std::distance(myGeo->meshData.begin(),
+                                 std::find_if(myBlockStart, myBlockEnd,
+                                              [myRank](const std::vector<GO>& vec) -> bool {
+                                                return (vec[0] == myRank) ? true : false;
+                                              }));
+
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < myGeo->numDimensions) {
+      myGeo->offsets[dim]     = Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
+      myGeo->offsets[dim + 3] = Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
     }
+  }
 
-    // Check if the partition contains nodes on a boundary, if so that boundary (face, line or
-    // point) will not require ghost nodes.
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < myGeo->numDimensions &&  (myGeo->startIndices[dim] % myGeo->coarseRate[dim] != 0 ||
-                                         myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim]-1)) {
-        myGeo->ghostInterface[2*dim] = true;
-      }
-      if(dim < myGeo->numDimensions
-         && myGeo->startIndices[dim + 3] != myGeo->gFineNodesPerDir[dim] - 1
-         && (myGeo->lFineNodesPerDir[dim] == 1 ||
-             myGeo->startIndices[dim + 3] % myGeo->coarseRate[dim] != 0)) {
-        myGeo->ghostInterface[2*dim+1] = true;
-      }
+  // Check if the partition contains nodes on a boundary, if so that boundary (face, line or
+  // point) will not require ghost nodes.
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < myGeo->numDimensions && (myGeo->startIndices[dim] % myGeo->coarseRate[dim] != 0 ||
+                                       myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1)) {
+      myGeo->ghostInterface[2 * dim] = true;
+    }
+    if (dim < myGeo->numDimensions && myGeo->startIndices[dim + 3] != myGeo->gFineNodesPerDir[dim] - 1 && (myGeo->lFineNodesPerDir[dim] == 1 || myGeo->startIndices[dim + 3] % myGeo->coarseRate[dim] != 0)) {
+      myGeo->ghostInterface[2 * dim + 1] = true;
     }
+  }
 
-    // Here one element can represent either the degenerate case of one node or the more general
-    // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with one
-    // node. This helps generating a 3D space from tensorial products...
-    // A good way to handle this would be to generalize the algorithm to take into account the
-    // discretization order used in each direction, at least in the FEM sense, since a 0 degree
-    // discretization will have a unique node per element. This way 1D discretization can be viewed
-    // as a 3D problem with one 0 degree element in the y direction and one 0 degre element in the z
-    // direction.
-    // !!! Operations below are aftecting both local and global values that have two different   !!!
-    // orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G. coarseRate,
-    // endRate and offsets are in the global basis, as well as all the variables starting with a g,
-    // !!! while the variables starting with an l are in the local basis.                        !!!
-    for(int i = 0; i < 3; ++i) {
-      if(i < myGeo->numDimensions) {
-        // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
-        // level.
-        myGeo->gCoarseNodesPerDir[i] = (myGeo->gFineNodesPerDir[i] - 1) / myGeo->coarseRate[i];
-        myGeo->endRate[i] = Teuchos::as<LO>((myGeo->gFineNodesPerDir[i] - 1) %myGeo->coarseRate[i]);
-        if(myGeo->endRate[i] == 0) {
-          myGeo->endRate[i] = myGeo->coarseRate[i];
-          ++myGeo->gCoarseNodesPerDir[i];
-        } else {
-          myGeo->gCoarseNodesPerDir[i] += 2;
-        }
+  // Here one element can represent either the degenerate case of one node or the more general
+  // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with one
+  // node. This helps generating a 3D space from tensorial products...
+  // A good way to handle this would be to generalize the algorithm to take into account the
+  // discretization order used in each direction, at least in the FEM sense, since a 0 degree
+  // discretization will have a unique node per element. This way 1D discretization can be viewed
+  // as a 3D problem with one 0 degree element in the y direction and one 0 degre element in the z
+  // direction.
+  // !!! Operations below are aftecting both local and global values that have two different   !!!
+  // orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G. coarseRate,
+  // endRate and offsets are in the global basis, as well as all the variables starting with a g,
+  // !!! while the variables starting with an l are in the local basis.                        !!!
+  for (int i = 0; i < 3; ++i) {
+    if (i < myGeo->numDimensions) {
+      // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
+      // level.
+      myGeo->gCoarseNodesPerDir[i] = (myGeo->gFineNodesPerDir[i] - 1) / myGeo->coarseRate[i];
+      myGeo->endRate[i]            = Teuchos::as<LO>((myGeo->gFineNodesPerDir[i] - 1) % myGeo->coarseRate[i]);
+      if (myGeo->endRate[i] == 0) {
+        myGeo->endRate[i] = myGeo->coarseRate[i];
+        ++myGeo->gCoarseNodesPerDir[i];
       } else {
-        myGeo->endRate[i] = 1;
-        myGeo->gCoarseNodesPerDir[i] = 1;
+        myGeo->gCoarseNodesPerDir[i] += 2;
       }
+    } else {
+      myGeo->endRate[i]            = 1;
+      myGeo->gCoarseNodesPerDir[i] = 1;
     }
+  }
 
-    myGeo->gNumCoarseNodes = myGeo->gCoarseNodesPerDir[0]*myGeo->gCoarseNodesPerDir[1]
-      *myGeo->gCoarseNodesPerDir[2];
-
-    for(LO i = 0; i < 3; ++i) {
-      if(i < myGeo->numDimensions) {
-        // Check whether the partition includes the "end" of the mesh which means that endRate will
-        // apply. Also make sure that endRate is not 0 which means that the mesh does not require a
-        // particular treatment at the boundaries.
-        if( (myGeo->startIndices[i] + myGeo->lFineNodesPerDir[i]) == myGeo->gFineNodesPerDir[i] ) {
-          myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] - myGeo->endRate[i]
-                                   + myGeo->offsets[i] - 1) / myGeo->coarseRate[i] + 1;
-          if(myGeo->offsets[i] == 0) {++myGeo->lCoarseNodesPerDir[i];}
-        } else {
-          myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] + myGeo->offsets[i] - 1)
-            / myGeo->coarseRate[i];
-          if(myGeo->offsets[i] == 0) {++myGeo->lCoarseNodesPerDir[i];}
+  myGeo->gNumCoarseNodes = myGeo->gCoarseNodesPerDir[0] * myGeo->gCoarseNodesPerDir[1] * myGeo->gCoarseNodesPerDir[2];
+
+  for (LO i = 0; i < 3; ++i) {
+    if (i < myGeo->numDimensions) {
+      // Check whether the partition includes the "end" of the mesh which means that endRate will
+      // apply. Also make sure that endRate is not 0 which means that the mesh does not require a
+      // particular treatment at the boundaries.
+      if ((myGeo->startIndices[i] + myGeo->lFineNodesPerDir[i]) == myGeo->gFineNodesPerDir[i]) {
+        myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] - myGeo->endRate[i] + myGeo->offsets[i] - 1) / myGeo->coarseRate[i] + 1;
+        if (myGeo->offsets[i] == 0) {
+          ++myGeo->lCoarseNodesPerDir[i];
         }
       } else {
-        myGeo->lCoarseNodesPerDir[i] = 1;
+        myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] + myGeo->offsets[i] - 1) / myGeo->coarseRate[i];
+        if (myGeo->offsets[i] == 0) {
+          ++myGeo->lCoarseNodesPerDir[i];
+        }
       }
-      // This would happen if the rank does not own any nodes but in that case a subcommunicator
-      // should be used so this should really not be a concern.
-      if(myGeo->lFineNodesPerDir[i] < 1) {myGeo->lCoarseNodesPerDir[i] = 0;}
+    } else {
+      myGeo->lCoarseNodesPerDir[i] = 1;
+    }
+    // This would happen if the rank does not own any nodes but in that case a subcommunicator
+    // should be used so this should really not be a concern.
+    if (myGeo->lFineNodesPerDir[i] < 1) {
+      myGeo->lCoarseNodesPerDir[i] = 0;
     }
+  }
 
-    // Assuming linear interpolation, each fine point has contribution from 8 coarse points
-    // and each coarse point value gets injected.
-    // For systems of PDEs we assume that all dofs have the same P operator.
-    myGeo->lNumCoarseNodes = myGeo->lCoarseNodesPerDir[0]*myGeo->lCoarseNodesPerDir[1]
-      *myGeo->lCoarseNodesPerDir[2];
-
-    // For each direction, determine how many points (including ghosts) are required.
-    for(int dim = 0; dim < 3; ++dim) {
-      // The first branch of this if-statement will be used if the rank contains only one layer
-      // of nodes in direction i, that layer must also coincide with the boundary of the mesh
-      // and coarseRate[i] == endRate[i]...
-      if(myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1 &&
-         myGeo->startIndices[dim] % myGeo->coarseRate[dim] == 0) {
-        myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim] - 1;
-      } else {
-        myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim];
-      }
-      myGeo->ghostedCoarseNodesPerDir[dim] = myGeo->lCoarseNodesPerDir[dim];
-      // Check whether face *low needs ghost nodes
-      if(myGeo->ghostInterface[2*dim]) {myGeo->ghostedCoarseNodesPerDir[dim] += 1;}
-      // Check whether face *hi needs ghost nodes
-      if(myGeo->ghostInterface[2*dim + 1]) {myGeo->ghostedCoarseNodesPerDir[dim] += 1;}
+  // Assuming linear interpolation, each fine point has contribution from 8 coarse points
+  // and each coarse point value gets injected.
+  // For systems of PDEs we assume that all dofs have the same P operator.
+  myGeo->lNumCoarseNodes = myGeo->lCoarseNodesPerDir[0] * myGeo->lCoarseNodesPerDir[1] * myGeo->lCoarseNodesPerDir[2];
+
+  // For each direction, determine how many points (including ghosts) are required.
+  for (int dim = 0; dim < 3; ++dim) {
+    // The first branch of this if-statement will be used if the rank contains only one layer
+    // of nodes in direction i, that layer must also coincide with the boundary of the mesh
+    // and coarseRate[i] == endRate[i]...
+    if (myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1 &&
+        myGeo->startIndices[dim] % myGeo->coarseRate[dim] == 0) {
+      myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim] - 1;
+    } else {
+      myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim];
     }
-    myGeo->lNumGhostedNodes = myGeo->ghostedCoarseNodesPerDir[2]*myGeo->ghostedCoarseNodesPerDir[1]
-      *myGeo->ghostedCoarseNodesPerDir[0];
-    myGeo->lNumGhostNodes = myGeo->lNumGhostedNodes - myGeo->lNumCoarseNodes;
-    ghostedCoarseNodes->PIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->LIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->GIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->coarseGIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->colInds.resize(myGeo->lNumGhostedNodes);
-    lCoarseNodesGIDs[0].resize(myGeo->lNumCoarseNodes);
-    lCoarseNodesGIDs[1].resize(myGeo->lNumCoarseNodes);
-
-    // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
-    // This requires finding what their GID on the fine mesh is. They need to be ordered
-    // lexicographically to allow for fast sweeps through the mesh.
-
-    // We loop over all ghosted coarse nodes by increasing global lexicographic order
-    Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3);
-    LO currentIndex = -1, countCoarseNodes = 0;
-    for(int k = 0; k < myGeo->ghostedCoarseNodesPerDir[2]; ++k) {
-      for(int j = 0; j < myGeo->ghostedCoarseNodesPerDir[1]; ++j) {
-        for(int i = 0; i < myGeo->ghostedCoarseNodesPerDir[0]; ++i) {
-          currentIndex = k*myGeo->ghostedCoarseNodesPerDir[1]*myGeo->ghostedCoarseNodesPerDir[0]
-            + j*myGeo->ghostedCoarseNodesPerDir[0] + i;
-          coarseNodeCoarseIndices[0] = myGeo->startGhostedCoarseNode[0] + i;
-          coarseNodeFineIndices[0] = coarseNodeCoarseIndices[0]*myGeo->coarseRate[0];
-          if(coarseNodeFineIndices[0] > myGeo->gFineNodesPerDir[0] - 1) {
-            coarseNodeFineIndices[0] = myGeo->gFineNodesPerDir[0] - 1;
-          }
-          coarseNodeCoarseIndices[1] = myGeo->startGhostedCoarseNode[1] + j;
-          coarseNodeFineIndices[1] = coarseNodeCoarseIndices[1]*myGeo->coarseRate[1];
-          if(coarseNodeFineIndices[1] > myGeo->gFineNodesPerDir[1] - 1) {
-            coarseNodeFineIndices[1] = myGeo->gFineNodesPerDir[1] - 1;
-          }
-          coarseNodeCoarseIndices[2] = myGeo->startGhostedCoarseNode[2] + k;
-          coarseNodeFineIndices[2] = coarseNodeCoarseIndices[2]*myGeo->coarseRate[2];
-          if(coarseNodeFineIndices[2] > myGeo->gFineNodesPerDir[2] - 1) {
-            coarseNodeFineIndices[2] = myGeo->gFineNodesPerDir[2] - 1;
-          }
-          GO myGID = -1, myCoarseGID = -1;
-          LO myLID = -1, myPID = -1;
-          GetGIDLocalLexicographic(i, j, k, coarseNodeFineIndices, myGeo, myRankIndex, pi, pj, pk,
-                                   myBlockStart, myBlockEnd, myGID, myPID, myLID);
-          myCoarseGID = coarseNodeCoarseIndices[0]
-            + coarseNodeCoarseIndices[1]*myGeo->gCoarseNodesPerDir[0]
-            + coarseNodeCoarseIndices[2]*myGeo->gCoarseNodesPerDir[1]*myGeo->gCoarseNodesPerDir[0];
-          ghostedCoarseNodes->PIDs[currentIndex] = myPID;
-          ghostedCoarseNodes->LIDs[currentIndex] = myLID;
-          ghostedCoarseNodes->GIDs[currentIndex] = myGID;
-          ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
-          if(myPID == myRank){
-            lCoarseNodesGIDs[0][countCoarseNodes] = myCoarseGID;
-            lCoarseNodesGIDs[1][countCoarseNodes] = myGID;
-            ++countCoarseNodes;
-          }
+    myGeo->ghostedCoarseNodesPerDir[dim] = myGeo->lCoarseNodesPerDir[dim];
+    // Check whether face *low needs ghost nodes
+    if (myGeo->ghostInterface[2 * dim]) {
+      myGeo->ghostedCoarseNodesPerDir[dim] += 1;
+    }
+    // Check whether face *hi needs ghost nodes
+    if (myGeo->ghostInterface[2 * dim + 1]) {
+      myGeo->ghostedCoarseNodesPerDir[dim] += 1;
+    }
+  }
+  myGeo->lNumGhostedNodes = myGeo->ghostedCoarseNodesPerDir[2] * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0];
+  myGeo->lNumGhostNodes   = myGeo->lNumGhostedNodes - myGeo->lNumCoarseNodes;
+  ghostedCoarseNodes->PIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->LIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->GIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->coarseGIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->colInds.resize(myGeo->lNumGhostedNodes);
+  lCoarseNodesGIDs[0].resize(myGeo->lNumCoarseNodes);
+  lCoarseNodesGIDs[1].resize(myGeo->lNumCoarseNodes);
+
+  // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
+  // This requires finding what their GID on the fine mesh is. They need to be ordered
+  // lexicographically to allow for fast sweeps through the mesh.
+
+  // We loop over all ghosted coarse nodes by increasing global lexicographic order
+  Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3);
+  LO currentIndex = -1, countCoarseNodes = 0;
+  for (int k = 0; k < myGeo->ghostedCoarseNodesPerDir[2]; ++k) {
+    for (int j = 0; j < myGeo->ghostedCoarseNodesPerDir[1]; ++j) {
+      for (int i = 0; i < myGeo->ghostedCoarseNodesPerDir[0]; ++i) {
+        currentIndex               = k * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0] + j * myGeo->ghostedCoarseNodesPerDir[0] + i;
+        coarseNodeCoarseIndices[0] = myGeo->startGhostedCoarseNode[0] + i;
+        coarseNodeFineIndices[0]   = coarseNodeCoarseIndices[0] * myGeo->coarseRate[0];
+        if (coarseNodeFineIndices[0] > myGeo->gFineNodesPerDir[0] - 1) {
+          coarseNodeFineIndices[0] = myGeo->gFineNodesPerDir[0] - 1;
+        }
+        coarseNodeCoarseIndices[1] = myGeo->startGhostedCoarseNode[1] + j;
+        coarseNodeFineIndices[1]   = coarseNodeCoarseIndices[1] * myGeo->coarseRate[1];
+        if (coarseNodeFineIndices[1] > myGeo->gFineNodesPerDir[1] - 1) {
+          coarseNodeFineIndices[1] = myGeo->gFineNodesPerDir[1] - 1;
+        }
+        coarseNodeCoarseIndices[2] = myGeo->startGhostedCoarseNode[2] + k;
+        coarseNodeFineIndices[2]   = coarseNodeCoarseIndices[2] * myGeo->coarseRate[2];
+        if (coarseNodeFineIndices[2] > myGeo->gFineNodesPerDir[2] - 1) {
+          coarseNodeFineIndices[2] = myGeo->gFineNodesPerDir[2] - 1;
+        }
+        GO myGID = -1, myCoarseGID = -1;
+        LO myLID = -1, myPID = -1;
+        GetGIDLocalLexicographic(i, j, k, coarseNodeFineIndices, myGeo, myRankIndex, pi, pj, pk,
+                                 myBlockStart, myBlockEnd, myGID, myPID, myLID);
+        myCoarseGID                                  = coarseNodeCoarseIndices[0] + coarseNodeCoarseIndices[1] * myGeo->gCoarseNodesPerDir[0] + coarseNodeCoarseIndices[2] * myGeo->gCoarseNodesPerDir[1] * myGeo->gCoarseNodesPerDir[0];
+        ghostedCoarseNodes->PIDs[currentIndex]       = myPID;
+        ghostedCoarseNodes->LIDs[currentIndex]       = myLID;
+        ghostedCoarseNodes->GIDs[currentIndex]       = myGID;
+        ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
+        if (myPID == myRank) {
+          lCoarseNodesGIDs[0][countCoarseNodes] = myCoarseGID;
+          lCoarseNodesGIDs[1][countCoarseNodes] = myGID;
+          ++countCoarseNodes;
         }
       }
     }
+  }
 
-  } // End MeshLayoutInterface
+}  // End MeshLayoutInterface
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetCoarsePoints(const int /* interpolationOrder */, const LO /* blkSize */, RCP<const Map> fineCoordsMap,
-                  RCP<GeometricData> myGeo, RCP<NodesIDs> ghostedCoarseNodes,
-                  Array<Array<GO> >& lCoarseNodesGIDs) const{
-    // Assuming perfect global lexicographic ordering of the mesh, produce two arrays:
-    //   1) lGhostNodesIDs that stores PID, LID, GID and coarseGID associated with the coarse nodes
-    //      need to compute the local part of the prolongator.
-    //   2) lCoarseNodesGIDs that stores the GIDs associated with the local nodes needed to create
-    //      the map of the MultiVector of coarse node coordinates.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetCoarsePoints(const int /* interpolationOrder */, const LO /* blkSize */, RCP<const Map> fineCoordsMap,
+                    RCP<GeometricData> myGeo, RCP<NodesIDs> ghostedCoarseNodes,
+                    Array<Array<GO> >& lCoarseNodesGIDs) const {
+  // Assuming perfect global lexicographic ordering of the mesh, produce two arrays:
+  //   1) lGhostNodesIDs that stores PID, LID, GID and coarseGID associated with the coarse nodes
+  //      need to compute the local part of the prolongator.
+  //   2) lCoarseNodesGIDs that stores the GIDs associated with the local nodes needed to create
+  //      the map of the MultiVector of coarse node coordinates.
 
-    {
-      GO tmp = 0;
-      myGeo->startIndices[2] = fineCoordsMap->getMinGlobalIndex()
-        / (myGeo->gFineNodesPerDir[1]*myGeo->gFineNodesPerDir[0]);
-      tmp = fineCoordsMap->getMinGlobalIndex()
-        % (myGeo->gFineNodesPerDir[1]*myGeo->gFineNodesPerDir[0]);
-      myGeo->startIndices[1] = tmp / myGeo->gFineNodesPerDir[0];
-      myGeo->startIndices[0] = tmp % myGeo->gFineNodesPerDir[0];
-    } // End of scope for tmp
-    for(int dim = 0; dim < 3; ++dim) {
-      myGeo->startIndices[dim + 3] = myGeo->startIndices[dim] + myGeo->lFineNodesPerDir[dim] - 1;
-    }
+  {
+    GO tmp                 = 0;
+    myGeo->startIndices[2] = fineCoordsMap->getMinGlobalIndex() / (myGeo->gFineNodesPerDir[1] * myGeo->gFineNodesPerDir[0]);
+    tmp                    = fineCoordsMap->getMinGlobalIndex() % (myGeo->gFineNodesPerDir[1] * myGeo->gFineNodesPerDir[0]);
+    myGeo->startIndices[1] = tmp / myGeo->gFineNodesPerDir[0];
+    myGeo->startIndices[0] = tmp % myGeo->gFineNodesPerDir[0];
+  }  // End of scope for tmp
+  for (int dim = 0; dim < 3; ++dim) {
+    myGeo->startIndices[dim + 3] = myGeo->startIndices[dim] + myGeo->lFineNodesPerDir[dim] - 1;
+  }
 
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < myGeo->numDimensions) {
-        myGeo->offsets[dim]= Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
-        myGeo->offsets[dim + 3]= Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
-      }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < myGeo->numDimensions) {
+      myGeo->offsets[dim]     = Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
+      myGeo->offsets[dim + 3] = Teuchos::as<LO>(myGeo->startIndices[dim]) % myGeo->coarseRate[dim];
     }
+  }
 
-    // Check if the partition contains nodes on a boundary, if so that boundary (face, line or
-    // point) will not require ghost nodes, unless there is only one node in that direction.
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < myGeo->numDimensions && (myGeo->startIndices[dim] % myGeo->coarseRate[dim] != 0 ||
-                                        myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim]-1)) {
-        myGeo->ghostInterface[2*dim] = true;
-      }
-      if(dim < myGeo->numDimensions
-         && myGeo->startIndices[dim + 3] != myGeo->gFineNodesPerDir[dim] - 1
-         && (myGeo->lFineNodesPerDir[dim] == 1 ||
-             myGeo->startIndices[dim + 3] % myGeo->coarseRate[dim] != 0)) {
-        myGeo->ghostInterface[2*dim + 1] = true;
-      }
+  // Check if the partition contains nodes on a boundary, if so that boundary (face, line or
+  // point) will not require ghost nodes, unless there is only one node in that direction.
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < myGeo->numDimensions && (myGeo->startIndices[dim] % myGeo->coarseRate[dim] != 0 ||
+                                       myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1)) {
+      myGeo->ghostInterface[2 * dim] = true;
     }
+    if (dim < myGeo->numDimensions && myGeo->startIndices[dim + 3] != myGeo->gFineNodesPerDir[dim] - 1 && (myGeo->lFineNodesPerDir[dim] == 1 || myGeo->startIndices[dim + 3] % myGeo->coarseRate[dim] != 0)) {
+      myGeo->ghostInterface[2 * dim + 1] = true;
+    }
+  }
 
-    // Here one element can represent either the degenerate case of one node or the more general
-    // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with one
-    // node. This helps generating a 3D space from tensorial products...
-    // A good way to handle this would be to generalize the algorithm to take into account the
-    // discretization order used in each direction, at least in the FEM sense, since a 0 degree
-    // discretization will have a unique node per element. This way 1D discretization can be viewed
-    // as a 3D problem with one 0 degree element in the y direction and one 0 degre element in the z
-    // direction.
-    // !!! Operations below are aftecting both local and global values that have two different   !!!
-    // orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G. coarseRate,
-    // endRate and offsets are in the global basis, as well as all the variables starting with a g,
-    // !!! while the variables starting with an l are in the local basis.                        !!!
-    for(int i = 0; i < 3; ++i) {
-      if(i < myGeo->numDimensions) {
-        // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
-        // level.
-        myGeo->gCoarseNodesPerDir[i] = (myGeo->gFineNodesPerDir[i] - 1) / myGeo->coarseRate[i];
-        myGeo->endRate[i] = Teuchos::as<LO>((myGeo->gFineNodesPerDir[i] - 1) %myGeo->coarseRate[i]);
-        if(myGeo->endRate[i] == 0) {
-          myGeo->endRate[i] = myGeo->coarseRate[i];
-          ++myGeo->gCoarseNodesPerDir[i];
-        } else {
-          myGeo->gCoarseNodesPerDir[i] += 2;
-        }
+  // Here one element can represent either the degenerate case of one node or the more general
+  // case of two nodes, i.e. x---x is a 1D element with two nodes and x is a 1D element with one
+  // node. This helps generating a 3D space from tensorial products...
+  // A good way to handle this would be to generalize the algorithm to take into account the
+  // discretization order used in each direction, at least in the FEM sense, since a 0 degree
+  // discretization will have a unique node per element. This way 1D discretization can be viewed
+  // as a 3D problem with one 0 degree element in the y direction and one 0 degre element in the z
+  // direction.
+  // !!! Operations below are aftecting both local and global values that have two different   !!!
+  // orientations. Orientations can be interchanged using mapDirG2L and mapDirL2G. coarseRate,
+  // endRate and offsets are in the global basis, as well as all the variables starting with a g,
+  // !!! while the variables starting with an l are in the local basis.                        !!!
+  for (int i = 0; i < 3; ++i) {
+    if (i < myGeo->numDimensions) {
+      // This array is passed to the RAPFactory and eventually becomes gFineNodePerDir on the next
+      // level.
+      myGeo->gCoarseNodesPerDir[i] = (myGeo->gFineNodesPerDir[i] - 1) / myGeo->coarseRate[i];
+      myGeo->endRate[i]            = Teuchos::as<LO>((myGeo->gFineNodesPerDir[i] - 1) % myGeo->coarseRate[i]);
+      if (myGeo->endRate[i] == 0) {
+        myGeo->endRate[i] = myGeo->coarseRate[i];
+        ++myGeo->gCoarseNodesPerDir[i];
       } else {
-        myGeo->endRate[i] = 1;
-        myGeo->gCoarseNodesPerDir[i] = 1;
+        myGeo->gCoarseNodesPerDir[i] += 2;
       }
+    } else {
+      myGeo->endRate[i]            = 1;
+      myGeo->gCoarseNodesPerDir[i] = 1;
     }
+  }
 
-    myGeo->gNumCoarseNodes = myGeo->gCoarseNodesPerDir[0]*myGeo->gCoarseNodesPerDir[1]
-      *myGeo->gCoarseNodesPerDir[2];
-
-    for(LO i = 0; i < 3; ++i) {
-      if(i < myGeo->numDimensions) {
-        // Check whether the partition includes the "end" of the mesh which means that endRate will
-        // apply. Also make sure that endRate is not 0 which means that the mesh does not require a
-        // particular treatment at the boundaries.
-        if( (myGeo->startIndices[i] + myGeo->lFineNodesPerDir[i]) == myGeo->gFineNodesPerDir[i] ) {
-          myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] - myGeo->endRate[i]
-                                   + myGeo->offsets[i] - 1) / myGeo->coarseRate[i] + 1;
-          if(myGeo->offsets[i] == 0) {++myGeo->lCoarseNodesPerDir[i];}
-        } else {
-          myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] + myGeo->offsets[i] - 1)
-            / myGeo->coarseRate[i];
-          if(myGeo->offsets[i] == 0) {++myGeo->lCoarseNodesPerDir[i];}
+  myGeo->gNumCoarseNodes = myGeo->gCoarseNodesPerDir[0] * myGeo->gCoarseNodesPerDir[1] * myGeo->gCoarseNodesPerDir[2];
+
+  for (LO i = 0; i < 3; ++i) {
+    if (i < myGeo->numDimensions) {
+      // Check whether the partition includes the "end" of the mesh which means that endRate will
+      // apply. Also make sure that endRate is not 0 which means that the mesh does not require a
+      // particular treatment at the boundaries.
+      if ((myGeo->startIndices[i] + myGeo->lFineNodesPerDir[i]) == myGeo->gFineNodesPerDir[i]) {
+        myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] - myGeo->endRate[i] + myGeo->offsets[i] - 1) / myGeo->coarseRate[i] + 1;
+        if (myGeo->offsets[i] == 0) {
+          ++myGeo->lCoarseNodesPerDir[i];
         }
       } else {
-        myGeo->lCoarseNodesPerDir[i] = 1;
+        myGeo->lCoarseNodesPerDir[i] = (myGeo->lFineNodesPerDir[i] + myGeo->offsets[i] - 1) / myGeo->coarseRate[i];
+        if (myGeo->offsets[i] == 0) {
+          ++myGeo->lCoarseNodesPerDir[i];
+        }
       }
-      // This would happen if the rank does not own any nodes but in that case a subcommunicator
-      // should be used so this should really not be a concern.
-      if(myGeo->lFineNodesPerDir[i] < 1) {myGeo->lCoarseNodesPerDir[i] = 0;}
+    } else {
+      myGeo->lCoarseNodesPerDir[i] = 1;
+    }
+    // This would happen if the rank does not own any nodes but in that case a subcommunicator
+    // should be used so this should really not be a concern.
+    if (myGeo->lFineNodesPerDir[i] < 1) {
+      myGeo->lCoarseNodesPerDir[i] = 0;
     }
+  }
 
-    // Assuming linear interpolation, each fine point has contribution from 8 coarse points
-    // and each coarse point value gets injected.
-    // For systems of PDEs we assume that all dofs have the same P operator.
-    myGeo->lNumCoarseNodes = myGeo->lCoarseNodesPerDir[0]*myGeo->lCoarseNodesPerDir[1]
-      *myGeo->lCoarseNodesPerDir[2];
-
-    // For each direction, determine how many points (including ghosts) are required.
-    bool ghostedDir[6] = {false};
-    for(int dim = 0; dim < 3; ++dim) {
-      // The first branch of this if-statement will be used if the rank contains only one layer
-      // of nodes in direction i, that layer must also coincide with the boundary of the mesh
-      // and coarseRate[i] == endRate[i]...
-      if(myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1 &&
-         myGeo->startIndices[dim] % myGeo->coarseRate[dim] == 0) {
-        myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim] - 1;
-      } else {
-        myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim];
-      }
-      myGeo->ghostedCoarseNodesPerDir[dim] = myGeo->lCoarseNodesPerDir[dim];
-      // Check whether face *low needs ghost nodes
-      if(myGeo->ghostInterface[2*dim]) {
-        myGeo->ghostedCoarseNodesPerDir[dim] += 1;
-        ghostedDir[2*dim] = true;
-      }
-      // Check whether face *hi needs ghost nodes
-      if(myGeo->ghostInterface[2*dim + 1]) {
-        myGeo->ghostedCoarseNodesPerDir[dim] += 1;
-        ghostedDir[2*dim + 1] = true;
-      }
+  // Assuming linear interpolation, each fine point has contribution from 8 coarse points
+  // and each coarse point value gets injected.
+  // For systems of PDEs we assume that all dofs have the same P operator.
+  myGeo->lNumCoarseNodes = myGeo->lCoarseNodesPerDir[0] * myGeo->lCoarseNodesPerDir[1] * myGeo->lCoarseNodesPerDir[2];
+
+  // For each direction, determine how many points (including ghosts) are required.
+  bool ghostedDir[6] = {false};
+  for (int dim = 0; dim < 3; ++dim) {
+    // The first branch of this if-statement will be used if the rank contains only one layer
+    // of nodes in direction i, that layer must also coincide with the boundary of the mesh
+    // and coarseRate[i] == endRate[i]...
+    if (myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1 &&
+        myGeo->startIndices[dim] % myGeo->coarseRate[dim] == 0) {
+      myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim] - 1;
+    } else {
+      myGeo->startGhostedCoarseNode[dim] = myGeo->startIndices[dim] / myGeo->coarseRate[dim];
     }
-    myGeo->lNumGhostedNodes = myGeo->ghostedCoarseNodesPerDir[2]*myGeo->ghostedCoarseNodesPerDir[1]
-      *myGeo->ghostedCoarseNodesPerDir[0];
-    myGeo->lNumGhostNodes = myGeo->lNumGhostedNodes - myGeo->lNumCoarseNodes;
-    ghostedCoarseNodes->PIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->LIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->GIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->coarseGIDs.resize(myGeo->lNumGhostedNodes);
-    ghostedCoarseNodes->colInds.resize(myGeo->lNumGhostedNodes);
-    lCoarseNodesGIDs[0].resize(myGeo->lNumCoarseNodes);
-    lCoarseNodesGIDs[1].resize(myGeo->lNumCoarseNodes);
-
-    // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
-    // This requires finding what their GID on the fine mesh is. They need to be ordered
-    // lexicographically to allow for fast sweeps through the mesh.
-
-    // We loop over all ghosted coarse nodes by increasing global lexicographic order
-    Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3), ijk(3);
-    LO currentIndex = -1, countCoarseNodes = 0;
-    for(ijk[2] = 0; ijk[2] < myGeo->ghostedCoarseNodesPerDir[2]; ++ijk[2]) {
-      for(ijk[1] = 0; ijk[1] < myGeo->ghostedCoarseNodesPerDir[1]; ++ijk[1]) {
-        for(ijk[0] = 0; ijk[0] < myGeo->ghostedCoarseNodesPerDir[0]; ++ijk[0]) {
-          currentIndex = ijk[2]*myGeo->ghostedCoarseNodesPerDir[1]*myGeo->ghostedCoarseNodesPerDir[0]
-            + ijk[1]*myGeo->ghostedCoarseNodesPerDir[0] + ijk[0];
-          coarseNodeCoarseIndices[0] = myGeo->startGhostedCoarseNode[0] + ijk[0];
-          coarseNodeFineIndices[0] = coarseNodeCoarseIndices[0]*myGeo->coarseRate[0];
-          if(coarseNodeFineIndices[0] > myGeo->gFineNodesPerDir[0] - 1) {
-            coarseNodeFineIndices[0] = myGeo->gFineNodesPerDir[0] - 1;
-          }
-          coarseNodeCoarseIndices[1] = myGeo->startGhostedCoarseNode[1] + ijk[1];
-          coarseNodeFineIndices[1] = coarseNodeCoarseIndices[1]*myGeo->coarseRate[1];
-          if(coarseNodeFineIndices[1] > myGeo->gFineNodesPerDir[1] - 1) {
-            coarseNodeFineIndices[1] = myGeo->gFineNodesPerDir[1] - 1;
-          }
-          coarseNodeCoarseIndices[2] = myGeo->startGhostedCoarseNode[2] + ijk[2];
-          coarseNodeFineIndices[2] = coarseNodeCoarseIndices[2]*myGeo->coarseRate[2];
-          if(coarseNodeFineIndices[2] > myGeo->gFineNodesPerDir[2] - 1) {
-            coarseNodeFineIndices[2] = myGeo->gFineNodesPerDir[2] - 1;
-          }
-          GO myGID = 0, myCoarseGID = -1;
-          GO factor[3] = {};
-          factor[2] = myGeo->gNumFineNodes10;
-          factor[1] = myGeo->gFineNodesPerDir[0];
-          factor[0] = 1;
-          for(int dim = 0; dim < 3; ++dim) {
-            if(dim < myGeo->numDimensions) {
-              if(myGeo->startIndices[dim] - myGeo->offsets[dim] + ijk[dim]*myGeo->coarseRate[dim]
-                 < myGeo->gFineNodesPerDir[dim] - 1) {
-                myGID += (myGeo->startIndices[dim] - myGeo->offsets[dim]
-                          + ijk[dim]*myGeo->coarseRate[dim])*factor[dim];
-              } else {
-                myGID += (myGeo->startIndices[dim] - myGeo->offsets[dim]
-                          + (ijk[dim] - 1)*myGeo->coarseRate[dim] + myGeo->endRate[dim])*factor[dim];
-              }
+    myGeo->ghostedCoarseNodesPerDir[dim] = myGeo->lCoarseNodesPerDir[dim];
+    // Check whether face *low needs ghost nodes
+    if (myGeo->ghostInterface[2 * dim]) {
+      myGeo->ghostedCoarseNodesPerDir[dim] += 1;
+      ghostedDir[2 * dim] = true;
+    }
+    // Check whether face *hi needs ghost nodes
+    if (myGeo->ghostInterface[2 * dim + 1]) {
+      myGeo->ghostedCoarseNodesPerDir[dim] += 1;
+      ghostedDir[2 * dim + 1] = true;
+    }
+  }
+  myGeo->lNumGhostedNodes = myGeo->ghostedCoarseNodesPerDir[2] * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0];
+  myGeo->lNumGhostNodes   = myGeo->lNumGhostedNodes - myGeo->lNumCoarseNodes;
+  ghostedCoarseNodes->PIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->LIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->GIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->coarseGIDs.resize(myGeo->lNumGhostedNodes);
+  ghostedCoarseNodes->colInds.resize(myGeo->lNumGhostedNodes);
+  lCoarseNodesGIDs[0].resize(myGeo->lNumCoarseNodes);
+  lCoarseNodesGIDs[1].resize(myGeo->lNumCoarseNodes);
+
+  // Now the tricky part starts, the coarse nodes / ghosted coarse nodes need to be imported.
+  // This requires finding what their GID on the fine mesh is. They need to be ordered
+  // lexicographically to allow for fast sweeps through the mesh.
+
+  // We loop over all ghosted coarse nodes by increasing global lexicographic order
+  Array<LO> coarseNodeCoarseIndices(3), coarseNodeFineIndices(3), ijk(3);
+  LO currentIndex = -1, countCoarseNodes = 0;
+  for (ijk[2] = 0; ijk[2] < myGeo->ghostedCoarseNodesPerDir[2]; ++ijk[2]) {
+    for (ijk[1] = 0; ijk[1] < myGeo->ghostedCoarseNodesPerDir[1]; ++ijk[1]) {
+      for (ijk[0] = 0; ijk[0] < myGeo->ghostedCoarseNodesPerDir[0]; ++ijk[0]) {
+        currentIndex               = ijk[2] * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0] + ijk[1] * myGeo->ghostedCoarseNodesPerDir[0] + ijk[0];
+        coarseNodeCoarseIndices[0] = myGeo->startGhostedCoarseNode[0] + ijk[0];
+        coarseNodeFineIndices[0]   = coarseNodeCoarseIndices[0] * myGeo->coarseRate[0];
+        if (coarseNodeFineIndices[0] > myGeo->gFineNodesPerDir[0] - 1) {
+          coarseNodeFineIndices[0] = myGeo->gFineNodesPerDir[0] - 1;
+        }
+        coarseNodeCoarseIndices[1] = myGeo->startGhostedCoarseNode[1] + ijk[1];
+        coarseNodeFineIndices[1]   = coarseNodeCoarseIndices[1] * myGeo->coarseRate[1];
+        if (coarseNodeFineIndices[1] > myGeo->gFineNodesPerDir[1] - 1) {
+          coarseNodeFineIndices[1] = myGeo->gFineNodesPerDir[1] - 1;
+        }
+        coarseNodeCoarseIndices[2] = myGeo->startGhostedCoarseNode[2] + ijk[2];
+        coarseNodeFineIndices[2]   = coarseNodeCoarseIndices[2] * myGeo->coarseRate[2];
+        if (coarseNodeFineIndices[2] > myGeo->gFineNodesPerDir[2] - 1) {
+          coarseNodeFineIndices[2] = myGeo->gFineNodesPerDir[2] - 1;
+        }
+        GO myGID = 0, myCoarseGID = -1;
+        GO factor[3] = {};
+        factor[2]    = myGeo->gNumFineNodes10;
+        factor[1]    = myGeo->gFineNodesPerDir[0];
+        factor[0]    = 1;
+        for (int dim = 0; dim < 3; ++dim) {
+          if (dim < myGeo->numDimensions) {
+            if (myGeo->startIndices[dim] - myGeo->offsets[dim] + ijk[dim] * myGeo->coarseRate[dim] < myGeo->gFineNodesPerDir[dim] - 1) {
+              myGID += (myGeo->startIndices[dim] - myGeo->offsets[dim] + ijk[dim] * myGeo->coarseRate[dim]) * factor[dim];
+            } else {
+              myGID += (myGeo->startIndices[dim] - myGeo->offsets[dim] + (ijk[dim] - 1) * myGeo->coarseRate[dim] + myGeo->endRate[dim]) * factor[dim];
             }
           }
-          myCoarseGID = coarseNodeCoarseIndices[0]
-            + coarseNodeCoarseIndices[1]*myGeo->gCoarseNodesPerDir[0]
-            + coarseNodeCoarseIndices[2]*myGeo->gCoarseNodesPerDir[1]*myGeo->gCoarseNodesPerDir[0];
-          ghostedCoarseNodes->GIDs[currentIndex] = myGID;
-          ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
-          if((!ghostedDir[0] || ijk[0] != 0)
-             && (!ghostedDir[2] || ijk[1] != 0)
-             && (!ghostedDir[4] || ijk[2] != 0)
-             && (!ghostedDir[1] || ijk[0] != myGeo->ghostedCoarseNodesPerDir[0] - 1)
-             && (!ghostedDir[3] || ijk[1] != myGeo->ghostedCoarseNodesPerDir[1] - 1)
-             && (!ghostedDir[5] || ijk[2] != myGeo->ghostedCoarseNodesPerDir[2] - 1)){
-            lCoarseNodesGIDs[0][countCoarseNodes] = myCoarseGID;
-            lCoarseNodesGIDs[1][countCoarseNodes] = myGID;
-            ++countCoarseNodes;
-          }
+        }
+        myCoarseGID                                  = coarseNodeCoarseIndices[0] + coarseNodeCoarseIndices[1] * myGeo->gCoarseNodesPerDir[0] + coarseNodeCoarseIndices[2] * myGeo->gCoarseNodesPerDir[1] * myGeo->gCoarseNodesPerDir[0];
+        ghostedCoarseNodes->GIDs[currentIndex]       = myGID;
+        ghostedCoarseNodes->coarseGIDs[currentIndex] = myCoarseGID;
+        if ((!ghostedDir[0] || ijk[0] != 0) && (!ghostedDir[2] || ijk[1] != 0) && (!ghostedDir[4] || ijk[2] != 0) && (!ghostedDir[1] || ijk[0] != myGeo->ghostedCoarseNodesPerDir[0] - 1) && (!ghostedDir[3] || ijk[1] != myGeo->ghostedCoarseNodesPerDir[1] - 1) && (!ghostedDir[5] || ijk[2] != myGeo->ghostedCoarseNodesPerDir[2] - 1)) {
+          lCoarseNodesGIDs[0][countCoarseNodes] = myCoarseGID;
+          lCoarseNodesGIDs[1][countCoarseNodes] = myGID;
+          ++countCoarseNodes;
         }
       }
     }
-    Array<int> ghostsPIDs(myGeo->lNumGhostedNodes);
-    Array<LO>  ghostsLIDs(myGeo->lNumGhostedNodes);
-    fineCoordsMap->getRemoteIndexList(ghostedCoarseNodes->GIDs(),
-                                      ghostedCoarseNodes->PIDs(),
-                                      ghostedCoarseNodes->LIDs());
-  } // End GetCoarsePoint
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  MakeGeneralGeometricP(RCP<GeometricData> myGeo,
-                        const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,Node> >& fineCoords,
-                        const LO nnzP, const LO dofsPerNode,
-                        RCP<const Map>& stridedDomainMapP,RCP<Matrix> & Amat, RCP<Matrix>& P,
-                        RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,Node> >& coarseCoords,
-                        RCP<NodesIDs> ghostedCoarseNodes, Array<Array<GO> > coarseNodesGIDs,
-                        int interpolationOrder) const {
-
-    /* On termination, return the number of local prolongator columns owned by
-     * this processor.
-     *
-     * Input
-     * =====
-     *    nNodes       Number of fine level Blk Rows owned by this processor
-     *    coarseRate   Rate of coarsening in each spatial direction.
-     *    endRate      Rate of coarsening in each spatial direction for the last
-     *                 nodes in the mesh where an adaptive coarsening rate is
-     *                 required.
-     *    nTerms       Number of nonzero entries in the prolongation matrix.
-     *    dofsPerNode  Number of degrees-of-freedom per mesh node.
-     *
-     * Output
-     * =====
-     *    So far nothing...
-     */
-
-    using xdMV                 = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>;
-    Xpetra::global_size_t OTI  = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-
-    LO myRank          = Amat->getRowMap()->getComm()->getRank();
-    GO numGloCols      = dofsPerNode*myGeo->gNumCoarseNodes;
-
-    // Build maps necessary to create and fill complete the prolongator
-    // note: rowMapP == rangeMapP and colMapP != domainMapP
-    RCP<const Map> rowMapP = Amat->getDomainMap();
-
-    RCP<const Map> domainMapP;
-
-    RCP<const Map> colMapP;
-
-    // At this point we need to create the column map which is a delicate operation.
-    // The entries in that map need to be ordered as follows:
-    //         1) first owned entries ordered by LID
-    //         2) second order the remaining entries by PID
-    //         3) entries with the same remote PID are ordered by GID.
-    // One piece of good news: myGeo->lNumCoarseNodes is the number of ownedNodes and
-    // myGeo->lNumGhostNodes is the number of remote nodes. The sorting can be limited to remote
-    // nodes as the owned ones are alreaded ordered by LID!
+  }
+  Array<int> ghostsPIDs(myGeo->lNumGhostedNodes);
+  Array<LO> ghostsLIDs(myGeo->lNumGhostedNodes);
+  fineCoordsMap->getRemoteIndexList(ghostedCoarseNodes->GIDs(),
+                                    ghostedCoarseNodes->PIDs(),
+                                    ghostedCoarseNodes->LIDs());
+}  // End GetCoarsePoint
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    MakeGeneralGeometricP(RCP<GeometricData> myGeo,
+                          const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, Node> >& fineCoords,
+                          const LO nnzP, const LO dofsPerNode,
+                          RCP<const Map>& stridedDomainMapP, RCP<Matrix>& Amat, RCP<Matrix>& P,
+                          RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, Node> >& coarseCoords,
+                          RCP<NodesIDs> ghostedCoarseNodes, Array<Array<GO> > coarseNodesGIDs,
+                          int interpolationOrder) const {
+  /* On termination, return the number of local prolongator columns owned by
+   * this processor.
+   *
+   * Input
+   * =====
+   *    nNodes       Number of fine level Blk Rows owned by this processor
+   *    coarseRate   Rate of coarsening in each spatial direction.
+   *    endRate      Rate of coarsening in each spatial direction for the last
+   *                 nodes in the mesh where an adaptive coarsening rate is
+   *                 required.
+   *    nTerms       Number of nonzero entries in the prolongation matrix.
+   *    dofsPerNode  Number of degrees-of-freedom per mesh node.
+   *
+   * Output
+   * =====
+   *    So far nothing...
+   */
+
+  using xdMV                = Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>;
+  Xpetra::global_size_t OTI = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+
+  LO myRank     = Amat->getRowMap()->getComm()->getRank();
+  GO numGloCols = dofsPerNode * myGeo->gNumCoarseNodes;
+
+  // Build maps necessary to create and fill complete the prolongator
+  // note: rowMapP == rangeMapP and colMapP != domainMapP
+  RCP<const Map> rowMapP = Amat->getDomainMap();
+
+  RCP<const Map> domainMapP;
+
+  RCP<const Map> colMapP;
+
+  // At this point we need to create the column map which is a delicate operation.
+  // The entries in that map need to be ordered as follows:
+  //         1) first owned entries ordered by LID
+  //         2) second order the remaining entries by PID
+  //         3) entries with the same remote PID are ordered by GID.
+  // One piece of good news: myGeo->lNumCoarseNodes is the number of ownedNodes and
+  // myGeo->lNumGhostNodes is the number of remote nodes. The sorting can be limited to remote
+  // nodes as the owned ones are alreaded ordered by LID!
 
-    {
-      std::vector<NodeID> colMapOrdering(myGeo->lNumGhostedNodes);
-      for(LO ind = 0; ind < myGeo->lNumGhostedNodes; ++ind) {
-        colMapOrdering[ind].GID = ghostedCoarseNodes->GIDs[ind];
-        if(ghostedCoarseNodes->PIDs[ind] == myRank) {
-          colMapOrdering[ind].PID = -1;
-        } else {
-          colMapOrdering[ind].PID = ghostedCoarseNodes->PIDs[ind];
-        }
-        colMapOrdering[ind].LID = ghostedCoarseNodes->LIDs[ind];
-        colMapOrdering[ind].lexiInd = ind;
+  {
+    std::vector<NodeID> colMapOrdering(myGeo->lNumGhostedNodes);
+    for (LO ind = 0; ind < myGeo->lNumGhostedNodes; ++ind) {
+      colMapOrdering[ind].GID = ghostedCoarseNodes->GIDs[ind];
+      if (ghostedCoarseNodes->PIDs[ind] == myRank) {
+        colMapOrdering[ind].PID = -1;
+      } else {
+        colMapOrdering[ind].PID = ghostedCoarseNodes->PIDs[ind];
       }
-      std::sort(colMapOrdering.begin(), colMapOrdering.end(),
-                [](NodeID a, NodeID b)->bool{
-                  return ( (a.PID < b.PID) || ((a.PID == b.PID) && (a.GID < b.GID)) );
-                });
-
-      Array<GO> colGIDs(dofsPerNode*myGeo->lNumGhostedNodes);
-      for(LO ind = 0; ind < myGeo->lNumGhostedNodes; ++ind) {
-        // Save the permutation calculated to go from Lexicographic indexing to column map indexing
-        ghostedCoarseNodes->colInds[colMapOrdering[ind].lexiInd] = ind;
-        for(LO dof = 0; dof < dofsPerNode; ++dof) {
-          colGIDs[dofsPerNode*ind + dof] = dofsPerNode*colMapOrdering[ind].GID + dof;
-        }
+      colMapOrdering[ind].LID     = ghostedCoarseNodes->LIDs[ind];
+      colMapOrdering[ind].lexiInd = ind;
+    }
+    std::sort(colMapOrdering.begin(), colMapOrdering.end(),
+              [](NodeID a, NodeID b) -> bool {
+                return ((a.PID < b.PID) || ((a.PID == b.PID) && (a.GID < b.GID)));
+              });
+
+    Array<GO> colGIDs(dofsPerNode * myGeo->lNumGhostedNodes);
+    for (LO ind = 0; ind < myGeo->lNumGhostedNodes; ++ind) {
+      // Save the permutation calculated to go from Lexicographic indexing to column map indexing
+      ghostedCoarseNodes->colInds[colMapOrdering[ind].lexiInd] = ind;
+      for (LO dof = 0; dof < dofsPerNode; ++dof) {
+        colGIDs[dofsPerNode * ind + dof] = dofsPerNode * colMapOrdering[ind].GID + dof;
       }
-      domainMapP = Xpetra::MapFactory<LO,GO,NO>::Build(rowMapP->lib(),
+    }
+    domainMapP = Xpetra::MapFactory<LO, GO, NO>::Build(rowMapP->lib(),
                                                        numGloCols,
-                                                       colGIDs.view(0,dofsPerNode*
-                                                                    myGeo->lNumCoarseNodes),
+                                                       colGIDs.view(0, dofsPerNode *
+                                                                           myGeo->lNumCoarseNodes),
                                                        rowMapP->getIndexBase(),
                                                        rowMapP->getComm());
-      colMapP = Xpetra::MapFactory<LO,GO,NO>::Build(rowMapP->lib(),
-                                                    OTI,
-                                                    colGIDs.view(0, colGIDs.size()),
-                                                    rowMapP->getIndexBase(),
-                                                    rowMapP->getComm());
-    } // End of scope for colMapOrdering and colGIDs
-
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0] = dofsPerNode;
-    stridedDomainMapP = Xpetra::StridedMapFactory<LO,GO,NO>::Build(domainMapP, strideInfo);
-
-    // Build the map for the coarse level coordinates, create the associated MultiVector and fill it
-    // with an import from the fine coordinates MultiVector. As data is local this should not create
-    // communications during the importer creation.
-    RCP<const Map> coarseCoordsMap = MapFactory::Build (fineCoords->getMap()->lib(),
-                                                        myGeo->gNumCoarseNodes,
-                                                        coarseNodesGIDs[0](),
-                                                        fineCoords->getMap()->getIndexBase(),
-                                                        fineCoords->getMap()->getComm());
-    RCP<const Map> coarseCoordsFineMap = MapFactory::Build (fineCoords->getMap()->lib(),
-                                                            myGeo->gNumCoarseNodes,
-                                                            coarseNodesGIDs[1](),
-                                                            fineCoords->getMap()->getIndexBase(),
-                                                            fineCoords->getMap()->getComm());
-
-    RCP<const Import> coarseImporter = ImportFactory::Build(fineCoords->getMap(),
-                                                            coarseCoordsFineMap);
-    coarseCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>::Build(coarseCoordsFineMap,
-                                                                      myGeo->numDimensions);
-    coarseCoords->doImport(*fineCoords, *coarseImporter, Xpetra::INSERT);
-    coarseCoords->replaceMap(coarseCoordsMap);
-
-    // Do the actual import using the fineCoords->getMap()
-    RCP<const Map> ghostMap = Xpetra::MapFactory<LO,GO,NO>::Build(fineCoords->getMap()->lib(),
-                                                               OTI,
-                                                               ghostedCoarseNodes->GIDs(),
-                                                               fineCoords->getMap()->getIndexBase(),
-                                                               fineCoords->getMap()->getComm());
-    RCP<const Import> ghostImporter = ImportFactory::Build(fineCoords->getMap(), ghostMap);
-    RCP<xdMV> ghostCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>::Build(ghostMap,
-                                                                              myGeo->numDimensions);
-    ghostCoords->doImport(*fineCoords, *ghostImporter, Xpetra::INSERT);
-
-    P = rcp(new CrsMatrixWrap(rowMapP, colMapP, 0));
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    ArrayRCP<size_t>  iaP;
-    ArrayRCP<LO>      jaP;
-    ArrayRCP<SC>     valP;
-
-    PCrs->allocateAllValues(nnzP, iaP, jaP, valP);
-
-    ArrayView<size_t> ia  = iaP();
-    ArrayView<LO>     ja  = jaP();
-    ArrayView<SC>     val = valP();
-    ia[0] = 0;
-
-    Array<ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > ghostedCoords(3);
-    {
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> tmp(ghostCoords->getLocalLength(),  0.0);
-      for(int dim = 0; dim < 3; ++dim) {
-        if(dim < myGeo->numDimensions) {
-          ghostedCoords[dim] = ghostCoords->getDataNonConst(dim);
-        } else {
-          ghostedCoords[dim] = tmp;
-        }
+    colMapP    = Xpetra::MapFactory<LO, GO, NO>::Build(rowMapP->lib(),
+                                                       OTI,
+                                                       colGIDs.view(0, colGIDs.size()),
+                                                       rowMapP->getIndexBase(),
+                                                       rowMapP->getComm());
+  }  // End of scope for colMapOrdering and colGIDs
+
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0]     = dofsPerNode;
+  stridedDomainMapP = Xpetra::StridedMapFactory<LO, GO, NO>::Build(domainMapP, strideInfo);
+
+  // Build the map for the coarse level coordinates, create the associated MultiVector and fill it
+  // with an import from the fine coordinates MultiVector. As data is local this should not create
+  // communications during the importer creation.
+  RCP<const Map> coarseCoordsMap     = MapFactory::Build(fineCoords->getMap()->lib(),
+                                                         myGeo->gNumCoarseNodes,
+                                                         coarseNodesGIDs[0](),
+                                                         fineCoords->getMap()->getIndexBase(),
+                                                         fineCoords->getMap()->getComm());
+  RCP<const Map> coarseCoordsFineMap = MapFactory::Build(fineCoords->getMap()->lib(),
+                                                         myGeo->gNumCoarseNodes,
+                                                         coarseNodesGIDs[1](),
+                                                         fineCoords->getMap()->getIndexBase(),
+                                                         fineCoords->getMap()->getComm());
+
+  RCP<const Import> coarseImporter = ImportFactory::Build(fineCoords->getMap(),
+                                                          coarseCoordsFineMap);
+  coarseCoords                     = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>::Build(coarseCoordsFineMap,
+                                                                                                                                           myGeo->numDimensions);
+  coarseCoords->doImport(*fineCoords, *coarseImporter, Xpetra::INSERT);
+  coarseCoords->replaceMap(coarseCoordsMap);
+
+  // Do the actual import using the fineCoords->getMap()
+  RCP<const Map> ghostMap         = Xpetra::MapFactory<LO, GO, NO>::Build(fineCoords->getMap()->lib(),
+                                                                          OTI,
+                                                                          ghostedCoarseNodes->GIDs(),
+                                                                          fineCoords->getMap()->getIndexBase(),
+                                                                          fineCoords->getMap()->getComm());
+  RCP<const Import> ghostImporter = ImportFactory::Build(fineCoords->getMap(), ghostMap);
+  RCP<xdMV> ghostCoords           = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>::Build(ghostMap,
+                                                                                                                                          myGeo->numDimensions);
+  ghostCoords->doImport(*fineCoords, *ghostImporter, Xpetra::INSERT);
+
+  P                   = rcp(new CrsMatrixWrap(rowMapP, colMapP, 0));
+  RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  ArrayRCP<size_t> iaP;
+  ArrayRCP<LO> jaP;
+  ArrayRCP<SC> valP;
+
+  PCrs->allocateAllValues(nnzP, iaP, jaP, valP);
+
+  ArrayView<size_t> ia = iaP();
+  ArrayView<LO> ja     = jaP();
+  ArrayView<SC> val    = valP();
+  ia[0]                = 0;
+
+  Array<ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > ghostedCoords(3);
+  {
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> tmp(ghostCoords->getLocalLength(), 0.0);
+    for (int dim = 0; dim < 3; ++dim) {
+      if (dim < myGeo->numDimensions) {
+        ghostedCoords[dim] = ghostCoords->getDataNonConst(dim);
+      } else {
+        ghostedCoords[dim] = tmp;
       }
     }
+  }
 
-    // Declaration and assignment of fineCoords which holds the coordinates of the fine nodes in 3D.
-    // To do so we pull the nD coordinates from fineCoords and pad the rest with zero vectors...
-    RCP<Xpetra::Vector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> > zeros
-      = Xpetra::VectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>::Build(fineCoords->getMap(), true);
-    ArrayRCP< ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > lFineCoords(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < myGeo->numDimensions) {
-        lFineCoords[dim] = fineCoords->getDataNonConst(dim);
-      } else {
-        lFineCoords[dim] = zeros->getDataNonConst(0);
-      }
+  // Declaration and assignment of fineCoords which holds the coordinates of the fine nodes in 3D.
+  // To do so we pull the nD coordinates from fineCoords and pad the rest with zero vectors...
+  RCP<Xpetra::Vector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> > zeros = Xpetra::VectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>::Build(fineCoords->getMap(), true);
+  ArrayRCP<ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > lFineCoords(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < myGeo->numDimensions) {
+      lFineCoords[dim] = fineCoords->getDataNonConst(dim);
+    } else {
+      lFineCoords[dim] = zeros->getDataNonConst(0);
     }
+  }
 
-    GO tStencil = 0;
-    for(int currentIndex = 0; currentIndex < myGeo->lNumFineNodes; ++currentIndex) {
-      Array<GO> ghostedIndices(3), firstInterpolationIndices(3);
-      Array<LO> interpolationPIDs(8), interpolationLIDs(8), interpolationGIDs(8);
-      Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > interpolationCoords(9);
-      interpolationCoords[0].resize(3);
-      GO firstInterpolationNodeIndex;
-      int nStencil = 0;
-      for(int dim = 0; dim < 3; ++dim) {
-        interpolationCoords[0][dim] = lFineCoords[dim][currentIndex];
-      }
+  GO tStencil = 0;
+  for (int currentIndex = 0; currentIndex < myGeo->lNumFineNodes; ++currentIndex) {
+    Array<GO> ghostedIndices(3), firstInterpolationIndices(3);
+    Array<LO> interpolationPIDs(8), interpolationLIDs(8), interpolationGIDs(8);
+    Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > interpolationCoords(9);
+    interpolationCoords[0].resize(3);
+    GO firstInterpolationNodeIndex;
+    int nStencil = 0;
+    for (int dim = 0; dim < 3; ++dim) {
+      interpolationCoords[0][dim] = lFineCoords[dim][currentIndex];
+    }
 
-      // Compute the ghosted (i,j,k) of the current node, that assumes (I,J,K)=(0,0,0) to be
-      // associated with the first node in ghostCoords
-      {// Scope for tmp
-        ghostedIndices[2] = currentIndex / (myGeo->lFineNodesPerDir[1]*myGeo->lFineNodesPerDir[0]);
-        LO tmp = currentIndex % (myGeo->lFineNodesPerDir[1]*myGeo->lFineNodesPerDir[0]);
-        ghostedIndices[1] = tmp / myGeo->lFineNodesPerDir[0];
-        ghostedIndices[0] = tmp % myGeo->lFineNodesPerDir[0];
-        for(int dim = 0; dim < 3; ++dim) {
-          ghostedIndices[dim] += myGeo->offsets[dim];
-        }
-        // A special case appears when the mesh is really coarse: it is possible for a rank to
-        // have a single coarse node in a given direction. If this happens on the highest i, j or k
-        // we need to "grab" a coarse node with a lower i, j, or k which leads us to add to the
-        // value of ghostedIndices
+    // Compute the ghosted (i,j,k) of the current node, that assumes (I,J,K)=(0,0,0) to be
+    // associated with the first node in ghostCoords
+    {  // Scope for tmp
+      ghostedIndices[2] = currentIndex / (myGeo->lFineNodesPerDir[1] * myGeo->lFineNodesPerDir[0]);
+      LO tmp            = currentIndex % (myGeo->lFineNodesPerDir[1] * myGeo->lFineNodesPerDir[0]);
+      ghostedIndices[1] = tmp / myGeo->lFineNodesPerDir[0];
+      ghostedIndices[0] = tmp % myGeo->lFineNodesPerDir[0];
+      for (int dim = 0; dim < 3; ++dim) {
+        ghostedIndices[dim] += myGeo->offsets[dim];
       }
-      // No we find the indices of the coarse nodes used for interpolation simply by integer
-      // division.
-      for(int dim = 0; dim < 3; ++dim) {
-        firstInterpolationIndices[dim] = ghostedIndices[dim] / myGeo->coarseRate[dim];
-        // If you are on the edge of the local domain go back one coarse node, unless there is only
-        // one node on the local domain...
-        if(firstInterpolationIndices[dim] == myGeo->ghostedCoarseNodesPerDir[dim] - 1
-           && myGeo->ghostedCoarseNodesPerDir[dim] > 1) {
-          firstInterpolationIndices[dim] -= 1;
-        }
+      // A special case appears when the mesh is really coarse: it is possible for a rank to
+      // have a single coarse node in a given direction. If this happens on the highest i, j or k
+      // we need to "grab" a coarse node with a lower i, j, or k which leads us to add to the
+      // value of ghostedIndices
+    }
+    // No we find the indices of the coarse nodes used for interpolation simply by integer
+    // division.
+    for (int dim = 0; dim < 3; ++dim) {
+      firstInterpolationIndices[dim] = ghostedIndices[dim] / myGeo->coarseRate[dim];
+      // If you are on the edge of the local domain go back one coarse node, unless there is only
+      // one node on the local domain...
+      if (firstInterpolationIndices[dim] == myGeo->ghostedCoarseNodesPerDir[dim] - 1 && myGeo->ghostedCoarseNodesPerDir[dim] > 1) {
+        firstInterpolationIndices[dim] -= 1;
       }
-      firstInterpolationNodeIndex =
-        firstInterpolationIndices[2]*myGeo->ghostedCoarseNodesPerDir[1]*myGeo->ghostedCoarseNodesPerDir[0]
-        + firstInterpolationIndices[1]*myGeo->ghostedCoarseNodesPerDir[0]
-        + firstInterpolationIndices[0];
-
-      // We extract the coordinates and PIDs associated with each coarse node used during
-      // inteprolation in order to fill the prolongator correctly
-      {
-        LO ind = -1;
-        for(int k = 0; k < 2; ++k) {
-          for(int j = 0; j < 2; ++j) {
-            for(int i = 0; i < 2; ++i) {
-              ind = k*4 + j*2 + i;
-              interpolationLIDs[ind] = firstInterpolationNodeIndex
-                + k*myGeo->ghostedCoarseNodesPerDir[1]*myGeo->ghostedCoarseNodesPerDir[0]
-                + j*myGeo->ghostedCoarseNodesPerDir[0] + i;
-              if(ghostedCoarseNodes->PIDs[interpolationLIDs[ind]] == rowMapP->getComm()->getRank()){
-                interpolationPIDs[ind] = -1;
-              } else {
-                interpolationPIDs[ind] = ghostedCoarseNodes->PIDs[interpolationLIDs[ind]];
-              }
-              interpolationGIDs[ind] = ghostedCoarseNodes->coarseGIDs[interpolationLIDs[ind]];
+    }
+    firstInterpolationNodeIndex =
+        firstInterpolationIndices[2] * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0] + firstInterpolationIndices[1] * myGeo->ghostedCoarseNodesPerDir[0] + firstInterpolationIndices[0];
 
-              interpolationCoords[ind + 1].resize(3);
-              for(int dim = 0; dim < 3; ++dim) {
-                interpolationCoords[ind + 1][dim]
-                  = ghostedCoords[dim][interpolationLIDs[ind]];
-              }
+    // We extract the coordinates and PIDs associated with each coarse node used during
+    // inteprolation in order to fill the prolongator correctly
+    {
+      LO ind = -1;
+      for (int k = 0; k < 2; ++k) {
+        for (int j = 0; j < 2; ++j) {
+          for (int i = 0; i < 2; ++i) {
+            ind                    = k * 4 + j * 2 + i;
+            interpolationLIDs[ind] = firstInterpolationNodeIndex + k * myGeo->ghostedCoarseNodesPerDir[1] * myGeo->ghostedCoarseNodesPerDir[0] + j * myGeo->ghostedCoarseNodesPerDir[0] + i;
+            if (ghostedCoarseNodes->PIDs[interpolationLIDs[ind]] == rowMapP->getComm()->getRank()) {
+              interpolationPIDs[ind] = -1;
+            } else {
+              interpolationPIDs[ind] = ghostedCoarseNodes->PIDs[interpolationLIDs[ind]];
+            }
+            interpolationGIDs[ind] = ghostedCoarseNodes->coarseGIDs[interpolationLIDs[ind]];
+
+            interpolationCoords[ind + 1].resize(3);
+            for (int dim = 0; dim < 3; ++dim) {
+              interpolationCoords[ind + 1][dim] = ghostedCoords[dim][interpolationLIDs[ind]];
             }
           }
         }
-      } // End of ind scope
-
-      // Compute the actual geometric interpolation stencil
-      // LO stencilLength = static_cast<LO>(std::pow(interpolationOrder + 1, 3));
-      std::vector<double> stencil(8);
-      Array<GO> firstCoarseNodeFineIndices(3);
-      int rate[3] = {};
-      for(int dim = 0; dim < 3; ++dim) {
-        firstCoarseNodeFineIndices[dim] = firstInterpolationIndices[dim]*myGeo->coarseRate[dim];
-        if((myGeo->startIndices[dim + 3] == myGeo->gFineNodesPerDir[dim] - 1)
-           && (ghostedIndices[dim] >=
-               (myGeo->ghostedCoarseNodesPerDir[dim] - 2)*myGeo->coarseRate[dim])) {
-          rate[dim] = myGeo->endRate[dim];
-        } else {
-          rate[dim] = myGeo->coarseRate[dim];
-        }
       }
-      Array<GO> trueGhostedIndices(3);
-      // This handles the case of a rank having a single node that also happens to be the last node
-      // in any direction. It might be more efficient to re-write the algo so that this is
-      // incorporated in the definition of ghostedIndices directly...
-      for(int dim = 0; dim < 3; ++dim) {
-        if (myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1) {
-          trueGhostedIndices[dim] = ghostedIndices[dim] + rate[dim];
-        } else {
-          trueGhostedIndices[dim] = ghostedIndices[dim];
-        }
+    }  // End of ind scope
+
+    // Compute the actual geometric interpolation stencil
+    // LO stencilLength = static_cast<LO>(std::pow(interpolationOrder + 1, 3));
+    std::vector<double> stencil(8);
+    Array<GO> firstCoarseNodeFineIndices(3);
+    int rate[3] = {};
+    for (int dim = 0; dim < 3; ++dim) {
+      firstCoarseNodeFineIndices[dim] = firstInterpolationIndices[dim] * myGeo->coarseRate[dim];
+      if ((myGeo->startIndices[dim + 3] == myGeo->gFineNodesPerDir[dim] - 1) && (ghostedIndices[dim] >=
+                                                                                 (myGeo->ghostedCoarseNodesPerDir[dim] - 2) * myGeo->coarseRate[dim])) {
+        rate[dim] = myGeo->endRate[dim];
+      } else {
+        rate[dim] = myGeo->coarseRate[dim];
+      }
+    }
+    Array<GO> trueGhostedIndices(3);
+    // This handles the case of a rank having a single node that also happens to be the last node
+    // in any direction. It might be more efficient to re-write the algo so that this is
+    // incorporated in the definition of ghostedIndices directly...
+    for (int dim = 0; dim < 3; ++dim) {
+      if (myGeo->startIndices[dim] == myGeo->gFineNodesPerDir[dim] - 1) {
+        trueGhostedIndices[dim] = ghostedIndices[dim] + rate[dim];
+      } else {
+        trueGhostedIndices[dim] = ghostedIndices[dim];
+      }
+    }
+    ComputeStencil(myGeo->numDimensions, trueGhostedIndices, firstCoarseNodeFineIndices, rate,
+                   interpolationCoords, interpolationOrder, stencil);
+
+    // Finally check whether the fine node is on a coarse: node, edge or face
+    // and select accordingly the non-zero values from the stencil and the
+    // corresponding column indices
+    Array<LO> nzIndStencil(8), permutation(8);
+    for (LO k = 0; k < 8; ++k) {
+      permutation[k] = k;
+    }
+    if (interpolationOrder == 0) {
+      nStencil = 1;
+      for (LO k = 0; k < 8; ++k) {
+        nzIndStencil[k] = static_cast<LO>(stencil[0]);
+      }
+      stencil[0]               = 0.0;
+      stencil[nzIndStencil[0]] = 1.0;
+    } else if (interpolationOrder == 1) {
+      Array<GO> currentNodeGlobalFineIndices(3);
+      for (int dim = 0; dim < 3; ++dim) {
+        currentNodeGlobalFineIndices[dim] = ghostedIndices[dim] - myGeo->offsets[dim] + myGeo->startIndices[dim];
       }
-      ComputeStencil(myGeo->numDimensions, trueGhostedIndices, firstCoarseNodeFineIndices, rate,
-                     interpolationCoords, interpolationOrder, stencil);
-
-      // Finally check whether the fine node is on a coarse: node, edge or face
-      // and select accordingly the non-zero values from the stencil and the
-      // corresponding column indices
-      Array<LO> nzIndStencil(8), permutation(8);
-      for(LO k = 0; k < 8; ++k) {permutation[k] = k;}
-      if(interpolationOrder == 0) {
+      if (((ghostedIndices[0] % myGeo->coarseRate[0] == 0) || currentNodeGlobalFineIndices[0] == myGeo->gFineNodesPerDir[0] - 1) && ((ghostedIndices[1] % myGeo->coarseRate[1] == 0) || currentNodeGlobalFineIndices[1] == myGeo->gFineNodesPerDir[1] - 1) && ((ghostedIndices[2] % myGeo->coarseRate[2] == 0) || currentNodeGlobalFineIndices[2] == myGeo->gFineNodesPerDir[2] - 1)) {
+        if ((currentNodeGlobalFineIndices[0] == myGeo->gFineNodesPerDir[0] - 1) ||
+            (ghostedIndices[0] / myGeo->coarseRate[0] == myGeo->ghostedCoarseNodesPerDir[0] - 1)) {
+          nzIndStencil[0] += 1;
+        }
+        if (((currentNodeGlobalFineIndices[1] == myGeo->gFineNodesPerDir[1] - 1) ||
+             (ghostedIndices[1] / myGeo->coarseRate[1] == myGeo->ghostedCoarseNodesPerDir[1] - 1)) &&
+            (myGeo->numDimensions > 1)) {
+          nzIndStencil[0] += 2;
+        }
+        if (((currentNodeGlobalFineIndices[2] == myGeo->gFineNodesPerDir[2] - 1) ||
+             (ghostedIndices[2] / myGeo->coarseRate[2] == myGeo->ghostedCoarseNodesPerDir[2] - 1)) &&
+            myGeo->numDimensions > 2) {
+          nzIndStencil[0] += 4;
+        }
         nStencil = 1;
-        for(LO k = 0; k < 8; ++k) {
-          nzIndStencil[k] = static_cast<LO>(stencil[0]);
+        for (LO k = 0; k < 8; ++k) {
+          nzIndStencil[k] = nzIndStencil[0];
         }
-        stencil[0] = 0.0;
-        stencil[nzIndStencil[0]] = 1.0;
-      } else if(interpolationOrder == 1) {
-        Array<GO> currentNodeGlobalFineIndices(3);
-        for(int dim = 0; dim < 3; ++dim) {
-          currentNodeGlobalFineIndices[dim] = ghostedIndices[dim] - myGeo->offsets[dim]
-            + myGeo->startIndices[dim];
-      }
-        if( ((ghostedIndices[0] % myGeo->coarseRate[0] == 0)
-             || currentNodeGlobalFineIndices[0] == myGeo->gFineNodesPerDir[0] - 1)
-            && ((ghostedIndices[1] % myGeo->coarseRate[1] == 0)
-                || currentNodeGlobalFineIndices[1] == myGeo->gFineNodesPerDir[1] - 1)
-            && ((ghostedIndices[2] % myGeo->coarseRate[2] == 0)
-                || currentNodeGlobalFineIndices[2] == myGeo->gFineNodesPerDir[2] - 1) ) {
-          if((currentNodeGlobalFineIndices[0] == myGeo->gFineNodesPerDir[0] - 1) ||
-             (ghostedIndices[0] / myGeo->coarseRate[0] == myGeo->ghostedCoarseNodesPerDir[0] - 1)) {
-            nzIndStencil[0] += 1;
-          }
-          if(((currentNodeGlobalFineIndices[1] == myGeo->gFineNodesPerDir[1] - 1) ||
-              (ghostedIndices[1] / myGeo->coarseRate[1] == myGeo->ghostedCoarseNodesPerDir[1] - 1))
-             && (myGeo->numDimensions > 1)){
-            nzIndStencil[0] += 2;
-          }
-          if(((currentNodeGlobalFineIndices[2] == myGeo->gFineNodesPerDir[2] - 1) ||
-              (ghostedIndices[2] / myGeo->coarseRate[2] == myGeo->ghostedCoarseNodesPerDir[2] - 1))
-             && myGeo->numDimensions > 2) {
-            nzIndStencil[0] += 4;
-          }
-          nStencil = 1;
-          for(LO k = 0; k < 8; ++k) {
-            nzIndStencil[k] = nzIndStencil[0];
-          }
-        } else {
-          nStencil = 8;
-          for(LO k = 0; k < 8; ++k) {
-            nzIndStencil[k] = k;
-          }
+      } else {
+        nStencil = 8;
+        for (LO k = 0; k < 8; ++k) {
+          nzIndStencil[k] = k;
         }
       }
+    }
 
-      // Here the values are filled in the Crs matrix arrays
-      // This is basically the only place these variables are modified
-      // Hopefully this makes handling system of PDEs easy!
-
-      // Loop on dofsPerNode and process each row for the current Node
+    // Here the values are filled in the Crs matrix arrays
+    // This is basically the only place these variables are modified
+    // Hopefully this makes handling system of PDEs easy!
 
+    // Loop on dofsPerNode and process each row for the current Node
 
-      // Sort nodes by PIDs using stable sort to keep sublist ordered by LIDs and GIDs
-      sh_sort2(interpolationPIDs.begin(),interpolationPIDs.end(),
-               permutation.begin(), permutation.end());
+    // Sort nodes by PIDs using stable sort to keep sublist ordered by LIDs and GIDs
+    sh_sort2(interpolationPIDs.begin(), interpolationPIDs.end(),
+             permutation.begin(), permutation.end());
 
-      GO colInd;
-      for(LO j = 0; j < dofsPerNode; ++j) {
-        ia[currentIndex*dofsPerNode + j + 1] = ia[currentIndex*dofsPerNode + j] + nStencil;
-        for(LO k = 0; k < nStencil; ++k) {
-          colInd = ghostedCoarseNodes->colInds[interpolationLIDs[nzIndStencil[permutation[k]]]];
-          ja[ia[currentIndex*dofsPerNode + j] + k] = colInd*dofsPerNode + j;
-          val[ia[currentIndex*dofsPerNode + j] + k] = stencil[nzIndStencil[permutation[k]]];
-        }
-        // Add the stencil for each degree of freedom.
-        tStencil += nStencil;
+    GO colInd;
+    for (LO j = 0; j < dofsPerNode; ++j) {
+      ia[currentIndex * dofsPerNode + j + 1] = ia[currentIndex * dofsPerNode + j] + nStencil;
+      for (LO k = 0; k < nStencil; ++k) {
+        colInd                                      = ghostedCoarseNodes->colInds[interpolationLIDs[nzIndStencil[permutation[k]]]];
+        ja[ia[currentIndex * dofsPerNode + j] + k]  = colInd * dofsPerNode + j;
+        val[ia[currentIndex * dofsPerNode + j] + k] = stencil[nzIndStencil[permutation[k]]];
       }
-    } // End loop over fine nodes
-
-    if (rowMapP->lib() == Xpetra::UseTpetra) {
-      // - Cannot resize for Epetra, as it checks for same pointers
-      // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
-      // NOTE: these invalidate ja and val views
-      jaP .resize(tStencil);
-      valP.resize(tStencil);
+      // Add the stencil for each degree of freedom.
+      tStencil += nStencil;
     }
+  }  // End loop over fine nodes
+
+  if (rowMapP->lib() == Xpetra::UseTpetra) {
+    // - Cannot resize for Epetra, as it checks for same pointers
+    // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
+    // NOTE: these invalidate ja and val views
+    jaP.resize(tStencil);
+    valP.resize(tStencil);
+  }
 
-    // Set the values of the prolongation operators into the CrsMatrix P and call FillComplete
-    PCrs->setAllValues(iaP, jaP, valP);
-    PCrs->expertStaticFillComplete(domainMapP,rowMapP);
-  } // End MakeGeneralGeometricP
-
-  // template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  // void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetGeometricData(
-  //                       RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> >& coordinates, const Array<LO> coarseRate,
-  //                       const Array<GO> gFineNodesPerDir, const Array<LO> lFineNodesPerDir, const LO BlkSize,
-  //                       Array<GO>& gIndices, Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
-  //                       Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir, Array<GO>& ghostGIDs,
-  //                       Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs, GO& gNumCoarseNodes, LO& lNumCoarseNodes,
-  //                       ArrayRCP<Array<double> > coarseNodes) const {
-
-  //   RCP<const Map> coordinatesMap = coordinates->getMap();
-  //   LO numDimensions  = coordinates->getNumVectors();
-
-  //   // Using the coarsening rate and the fine level data,
-  //   // compute coarse level data
-
-  //   //                              Phase 1                               //
-  //   // ------------------------------------------------------------------ //
-  //   // We first start by finding small informations on the mesh such as   //
-  //   // the number of coarse nodes (local and global) and the number of    //
-  //   // ghost nodes / the end rate of coarsening.                          //
-  //   // ------------------------------------------------------------------ //
-  //   GO minGlobalIndex = coordinatesMap->getMinGlobalIndex();
-  //   {
-  //     GO tmp;
-  //     gIndices[2] = minGlobalIndex / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-  //     tmp         = minGlobalIndex % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-  //     gIndices[1] = tmp / gFineNodesPerDir[0];
-  //     gIndices[0] = tmp % gFineNodesPerDir[0];
-
-  //     myOffset[2] = gIndices[2] % coarseRate[2];
-  //     myOffset[1] = gIndices[1] % coarseRate[1];
-  //     myOffset[0] = gIndices[0] % coarseRate[0];
-  //   }
-
-  //   // Check whether ghost nodes are needed in each direction
-  //   for(LO i=0; i < numDimensions; ++i) {
-  //     if((gIndices[i] != 0) && (gIndices[i] % coarseRate[i] > 0)) {
-  //       ghostInterface[2*i] = true;
-  //     }
-  //     if(((gIndices[i] + lFineNodesPerDir[i]) != gFineNodesPerDir[i]) && ((gIndices[i] + lFineNodesPerDir[i] - 1) % coarseRate[i] > 0)) {
-  //       ghostInterface[2*i + 1] = true;
-  //     }
-  //   }
-
-  //   for(LO i = 0; i < 3; ++i) {
-  //     if(i < numDimensions) {
-  //       lCoarseNodesPerDir[i] = (lFineNodesPerDir[i] + myOffset[i] - 1) / coarseRate[i];
-  //       if(myOffset[i] == 0) { ++lCoarseNodesPerDir[i]; }
-  //       gCoarseNodesPerDir[i] = (gFineNodesPerDir[i] - 1) / coarseRate[i];
-  //       endRate[i]            = (gFineNodesPerDir[i] - 1) % coarseRate[i];
-  //       if(endRate[i] == 0) {
-  //         ++gCoarseNodesPerDir[i];
-  //         endRate[i] = coarseRate[i];
-  //       }
-  //     } else {
-  //       // Most quantities need to be set to 1 for extra dimensions
-  //       // this is rather logical, an x-y plane is like a single layer
-  //       // of nodes in the z direction...
-  //       gCoarseNodesPerDir[i] = 1;
-  //       lCoarseNodesPerDir[i] = 1;
-  //       endRate[i]            = 1;
-  //     }
-  //   }
-
-  //   gNumCoarseNodes = gCoarseNodesPerDir[0]*gCoarseNodesPerDir[1]*gCoarseNodesPerDir[2];
-  //   lNumCoarseNodes = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];
-
-  //   // For each direction, determine how many ghost points are required.
-  //   LO lNumGhostNodes = 0;
-  //   {
-  //     const int complementaryIndices[3][2] = {{1,2}, {0,2}, {0,1}};
-  //     LO tmp = 0;
-  //     for(LO i = 0; i < 3; ++i) {
-  //       // Check whether a face in direction i needs ghost nodes
-  //       if(ghostInterface[2*i] || ghostInterface[2*i+1]) {
-  //         if(i == 0) {tmp = lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];}
-  //         if(i == 1) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[2];}
-  //         if(i == 2) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1];}
-  //       }
-  //       // If both faces in direction i need nodes, double the number of ghost nodes
-  //       if(ghostInterface[2*i] && ghostInterface[2*i+1]) {tmp = 2*tmp;}
-  //       lNumGhostNodes += tmp;
-
-  //       // The corners and edges need to be checked in 2D / 3D to add more ghosts nodes
-  //       for(LO j = 0; j < 2; ++j) {
-  //         for(LO k = 0; k < 2; ++k) {
-  //           // Check if two adjoining faces need ghost nodes and then add their common edge
-  //           if(ghostInterface[2*complementaryIndices[i][0]+j] && ghostInterface[2*complementaryIndices[i][1]+k]) {
-  //             lNumGhostNodes += lCoarseNodesPerDir[i];
-  //             // Add corners if three adjoining faces need ghost nodes,
-  //             // but add them only once! Hence when i == 0.
-  //             if(ghostInterface[2*i] && (i == 0)) { lNumGhostNodes += 1; }
-  //             if(ghostInterface[2*i+1] && (i == 0)) { lNumGhostNodes += 1; }
-  //           }
-  //         }
-  //       }
-  //       tmp = 0;
-  //     }
-  //   } // end of scope for tmp and complementaryIndices
-
-  //   //                              Phase 2                               //
-  //   // ------------------------------------------------------------------ //
-  //   // Build a list of GIDs to import the required ghost nodes.           //
-  //   // The ordering of the ghosts nodes will be as natural as possible,   //
-  //   // i.e. it should follow the GID ordering of the mesh.                //
-  //   // ------------------------------------------------------------------ //
-
-  //   // Saddly we have to more or less redo what was just done to figure out the value of lNumGhostNodes,
-  //   // there might be some optimization possibility here...
-  //   ghostGIDs.resize(lNumGhostNodes);
-  //   LO countGhosts = 0;
-  //   // Get the GID of the first point on the current partition.
-  //   GO startingGID = minGlobalIndex;
-  //   Array<GO> startingIndices(3);
-  //   // We still want ghost nodes even if have with a 0 offset,
-  //   // except when we are on a boundary
-  //   if(ghostInterface[4] && (myOffset[2] == 0)) {
-  //     if(gIndices[2] + coarseRate[2] > gFineNodesPerDir[2]) {
-  //       startingGID -= endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //     } else {
-  //       startingGID -= coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //     }
-  //   } else {
-  //     startingGID -= myOffset[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //   }
-  //   if(ghostInterface[2] && (myOffset[1] == 0)) {
-  //     if(gIndices[1] + coarseRate[1] > gFineNodesPerDir[1]) {
-  //       startingGID -= endRate[1]*gFineNodesPerDir[0];
-  //     } else {
-  //       startingGID -= coarseRate[1]*gFineNodesPerDir[0];
-  //     }
-  //   } else {
-  //     startingGID -= myOffset[1]*gFineNodesPerDir[0];
-  //   }
-  //   if(ghostInterface[0] && (myOffset[0] == 0)) {
-  //     if(gIndices[0] + coarseRate[0] > gFineNodesPerDir[0]) {
-  //       startingGID -= endRate[0];
-  //     } else {
-  //       startingGID -= coarseRate[0];
-  //     }
-  //   } else {
-  //     startingGID -= myOffset[0];
-  //   }
-
-  //   { // scope for tmp
-  //     GO tmp;
-  //     startingIndices[2] = startingGID / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-  //     tmp = startingGID % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
-  //     startingIndices[1] = tmp / gFineNodesPerDir[0];
-  //     startingIndices[0] = tmp % gFineNodesPerDir[0];
-  //   }
-
-  //   GO ghostOffset[3] = {0, 0, 0};
-  //   LO lengthZero  = lCoarseNodesPerDir[0];
-  //   LO lengthOne   = lCoarseNodesPerDir[1];
-  //   LO lengthTwo   = lCoarseNodesPerDir[2];
-  //   if(ghostInterface[0]) {++lengthZero;}
-  //   if(ghostInterface[1]) {++lengthZero;}
-  //   if(ghostInterface[2]) {++lengthOne;}
-  //   if(ghostInterface[3]) {++lengthOne;}
-  //   if(ghostInterface[4]) {++lengthTwo;}
-  //   if(ghostInterface[5]) {++lengthTwo;}
-
-  //   // First check the bottom face as it will have the lowest GIDs
-  //   if(ghostInterface[4]) {
-  //     ghostOffset[2] = startingGID;
-  //     for(LO j = 0; j < lengthOne; ++j) {
-  //       if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-  //         ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
-  //       } else {
-  //         ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-  //       }
-  //       for(LO k = 0; k < lengthZero; ++k) {
-  //         if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-  //           ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-  //         } else {
-  //           ghostOffset[0] = k*coarseRate[0];
-  //         }
-  //         // If the partition includes a changed rate at the edge, ghost nodes need to be picked carefully.
-  //         // This if statement is repeated each time a ghost node is selected.
-  //         ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-  //         ++countGhosts;
-  //       }
-  //     }
-  //   }
-
-  //   // Sweep over the lCoarseNodesPerDir[2] coarse layers in direction 2 and gather necessary ghost nodes
-  //   // located on these layers.
-  //   for(LO i = 0; i < lengthTwo; ++i) {
-  //     // Exclude the cases where ghost nodes exists on the faces in directions 2, these faces are swept
-  //     // seperatly for efficiency.
-  //     if( !((i == lengthTwo-1) && ghostInterface[5]) && !((i == 0) && ghostInterface[4]) ) {
-  //       // Set the ghostOffset in direction 2 taking into account a possible endRate different
-  //       // from the regular coarseRate.
-  //       if( (i == lengthTwo-1) && !ghostInterface[5] ) {
-  //         ghostOffset[2] = startingGID + ((i-1)*coarseRate[2] + endRate[2])*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //       } else {
-  //         ghostOffset[2] = startingGID + i*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //       }
-  //       for(LO j = 0; j < lengthOne; ++j) {
-  //         if( (j == 0) && ghostInterface[2] ) {
-  //           for(LO k = 0; k < lengthZero; ++k) {
-  //             if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-  //               if(k == 0) {
-  //                 ghostOffset[0] = endRate[0];
-  //               } else {
-  //                 ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-  //               }
-  //             } else {
-  //               ghostOffset[0] = k*coarseRate[0];
-  //             }
-  //             // In this case j == 0 so ghostOffset[1] is zero and can be ignored
-  //             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[0];
-  //             ++countGhosts;
-  //           }
-  //         } else if( (j == lengthOne-1) && ghostInterface[3] ) {
-  //           // Set the ghostOffset in direction 1 taking into account a possible endRate different
-  //           // from the regular coarseRate.
-  //           if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-  //             ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
-  //           } else {
-  //             ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-  //           }
-  //           for(LO k = 0; k < lengthZero; ++k) {
-  //             if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-  //               ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-  //             } else {
-  //               ghostOffset[0] = k*coarseRate[0];
-  //             }
-  //             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-  //             ++countGhosts;
-  //           }
-  //         } else {
-  //           // Set ghostOffset[1] for side faces sweep
-  //           if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
-  //             ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
-  //           } else {
-  //             ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-  //           }
-
-  //           // Set ghostOffset[0], ghostGIDs and countGhosts
-  //           if(ghostInterface[0]) { // In that case ghostOffset[0]==0, so we can ignore it
-  //             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1];
-  //             ++countGhosts;
-  //           }
-  //           if(ghostInterface[1]) { // Grab ghost point at the end of direction 0.
-  //             if( (startingIndices[0] + (lengthZero-1)*coarseRate[0]) > gFineNodesPerDir[0] - 1 ) {
-  //               if(lengthZero > 2) {
-  //                 ghostOffset[0] = (lengthZero-2)*coarseRate[0] + endRate[0];
-  //               } else {
-  //                 ghostOffset[0] = endRate[0];
-  //               }
-  //             } else {
-  //               ghostOffset[0] = (lengthZero-1)*coarseRate[0];
-  //             }
-  //             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-  //             ++countGhosts;
-  //           }
-  //         }
-  //       }
-  //     }
-  //   }
-
-  //   // Finally check the top face
-  //   if(ghostInterface[5]) {
-  //     if( startingIndices[2] + (lengthTwo-1)*coarseRate[2] + 1 > gFineNodesPerDir[2] ) {
-  //       ghostOffset[2] = startingGID + ((lengthTwo-2)*coarseRate[2] + endRate[2])*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //     } else {
-  //       ghostOffset[2] = startingGID + (lengthTwo-1)*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //     }
-  //     for(LO j = 0; j < lengthOne; ++j) {
-  //       if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) { // && !ghostInterface[3] ) {
-  //         ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
-  //       } else {
-  //         ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
-  //       }
-  //       for(LO k = 0; k < lengthZero; ++k) {
-  //         if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
-  //           ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
-  //         } else {
-  //           ghostOffset[0] = k*coarseRate[0];
-  //         }
-  //         ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
-  //         ++countGhosts;
-  //       }
-  //     }
-  //   }
-
-  //   //                              Phase 3                               //
-  //   // ------------------------------------------------------------------ //
-  //   // Final phase of this function, lists are being built to create the  //
-  //   // column and domain maps of the projection as well as the map and    //
-  //   // arrays for the coarse coordinates multivector.                     //
-  //   // ------------------------------------------------------------------ //
-
-  //   Array<GO> gCoarseNodesGIDs(lNumCoarseNodes);
-  //   LO currentNode, offset2, offset1, offset0;
-  //   // Find the GIDs of the coarse nodes on the partition.
-  //   for(LO ind2 = 0; ind2 < lCoarseNodesPerDir[2]; ++ind2) {
-  //     if(myOffset[2] == 0) {
-  //       offset2 = startingIndices[2] + myOffset[2];
-  //     } else {
-  //       if(startingIndices[2] + endRate[2] == gFineNodesPerDir[2] - 1) {
-  //         offset2 = startingIndices[2] + endRate[2];
-  //       } else {
-  //         offset2 = startingIndices[2] + coarseRate[2];
-  //       }
-  //     }
-  //     if(offset2 + ind2*coarseRate[2] > gFineNodesPerDir[2] - 1) {
-  //       offset2 += (ind2 - 1)*coarseRate[2] + endRate[2];
-  //     } else {
-  //       offset2 += ind2*coarseRate[2];
-  //     }
-  //     offset2 = offset2*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-
-  //     for(LO ind1 = 0; ind1 < lCoarseNodesPerDir[1]; ++ind1) {
-  //       if(myOffset[1] == 0) {
-  //         offset1 = startingIndices[1] + myOffset[1];
-  //       } else {
-  //         if(startingIndices[1] + endRate[1] == gFineNodesPerDir[1] - 1) {
-  //           offset1 = startingIndices[1] + endRate[1];
-  //         } else {
-  //           offset1 = startingIndices[1] + coarseRate[1];
-  //         }
-  //       }
-  //       if(offset1 + ind1*coarseRate[1] > gFineNodesPerDir[1] - 1) {
-  //         offset1 += (ind1 - 1)*coarseRate[1] + endRate[1];
-  //       } else {
-  //         offset1 += ind1*coarseRate[1];
-  //       }
-  //       offset1 = offset1*gFineNodesPerDir[0];
-  //       for(LO ind0 = 0; ind0 < lCoarseNodesPerDir[0]; ++ind0) {
-  //         offset0 = startingIndices[0];
-  //         if(myOffset[0] == 0) {
-  //           offset0 += ind0*coarseRate[0];
-  //         } else {
-  //           offset0 += (ind0 + 1)*coarseRate[0];
-  //         }
-  //         if(offset0 > gFineNodesPerDir[0] - 1) {offset0 += endRate[0] - coarseRate[0];}
-
-  //         currentNode = ind2*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]
-  //                     + ind1*lCoarseNodesPerDir[0]
-  //                     + ind0;
-  //         gCoarseNodesGIDs[currentNode] = offset2 + offset1 + offset0;
-  //       }
-  //     }
-  //   }
-
-  //   // Actual loop over all the coarse/ghost nodes to find their index on the coarse mesh
-  //   // and the corresponding dofs that will need to be added to colMapP.
-  //   colGIDs.resize(BlkSize*(lNumCoarseNodes+lNumGhostNodes));
-  //   coarseNodesGIDs.resize(lNumCoarseNodes);
-  //   for(LO i = 0; i < numDimensions; ++i) {coarseNodes[i].resize(lNumCoarseNodes);}
-  //   GO fineNodesPerCoarseSlab    = coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //   GO fineNodesEndCoarseSlab    = endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
-  //   GO fineNodesPerCoarsePlane   = coarseRate[1]*gFineNodesPerDir[0];
-  //   GO fineNodesEndCoarsePlane   = endRate[1]*gFineNodesPerDir[0];
-  //   GO coarseNodesPerCoarseLayer = gCoarseNodesPerDir[1]*gCoarseNodesPerDir[0];
-  //   GO gCoarseNodeOnCoarseGridGID;
-  //   LO gInd[3], lCol;
-  //   Array<int> ghostPIDs  (lNumGhostNodes);
-  //   Array<LO>  ghostLIDs  (lNumGhostNodes);
-  //   Array<LO>  ghostPermut(lNumGhostNodes);
-  //   for(LO k = 0; k < lNumGhostNodes; ++k) {ghostPermut[k] = k;}
-  //   coordinatesMap->getRemoteIndexList(ghostGIDs, ghostPIDs, ghostLIDs);
-  //   sh_sort_permute(ghostPIDs.begin(),ghostPIDs.end(), ghostPermut.begin(),ghostPermut.end());
-
-  //   { // scope for tmpInds, tmpVars and tmp.
-  //     GO tmpInds[3], tmpVars[2];
-  //     LO tmp;
-  //     // Loop over the coarse nodes of the partition and add them to colGIDs
-  //     // that will be used to construct the column and domain maps of P as well
-  //     // as to construct the coarse coordinates map.
-  //     // for(LO col = 0; col < lNumCoarseNodes; ++col) { // This should most likely be replaced by loops of lCoarseNodesPerDir[] to simplify arithmetics
-  //     LO col = 0;
-  //     LO firstCoarseNodeInds[3], currentCoarseNode;
-  //     for(LO dim = 0; dim < 3; ++dim) {
-  //       if(myOffset[dim] == 0) {
-  //         firstCoarseNodeInds[dim] = 0;
-  //       } else {
-  //         firstCoarseNodeInds[dim] = coarseRate[dim] - myOffset[dim];
-  //       }
-  //     }
-  //     Array<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > fineNodes(numDimensions);
-  //     for(LO dim = 0; dim < numDimensions; ++dim) {fineNodes[dim] = coordinates->getData(dim);}
-  //     for(LO k = 0; k < lCoarseNodesPerDir[2]; ++k) {
-  //       for(LO j = 0; j < lCoarseNodesPerDir[1]; ++j) {
-  //         for(LO i = 0; i < lCoarseNodesPerDir[0]; ++i) {
-  //           col = k*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0] + j*lCoarseNodesPerDir[0] + i;
-
-  //           // Check for endRate
-  //           currentCoarseNode = 0;
-  //           if(firstCoarseNodeInds[0] + i*coarseRate[0] > lFineNodesPerDir[0] - 1) {
-  //             currentCoarseNode += firstCoarseNodeInds[0] + (i-1)*coarseRate[0] + endRate[0];
-  //           } else {
-  //             currentCoarseNode += firstCoarseNodeInds[0] + i*coarseRate[0];
-  //           }
-  //           if(firstCoarseNodeInds[1] + j*coarseRate[1] > lFineNodesPerDir[1] - 1) {
-  //             currentCoarseNode += (firstCoarseNodeInds[1] + (j-1)*coarseRate[1] + endRate[1])*lFineNodesPerDir[0];
-  //           } else {
-  //             currentCoarseNode += (firstCoarseNodeInds[1] + j*coarseRate[1])*lFineNodesPerDir[0];
-  //           }
-  //           if(firstCoarseNodeInds[2] + k*coarseRate[2] > lFineNodesPerDir[2] - 1) {
-  //             currentCoarseNode += (firstCoarseNodeInds[2] + (k-1)*coarseRate[2] + endRate[2])*lFineNodesPerDir[1]*lFineNodesPerDir[0];
-  //           } else {
-  //             currentCoarseNode += (firstCoarseNodeInds[2] + k*coarseRate[2])*lFineNodesPerDir[1]*lFineNodesPerDir[0];
-  //           }
-  //           // Load coordinates
-  //           for(LO dim = 0; dim < numDimensions; ++dim) {
-  //             coarseNodes[dim][col] = fineNodes[dim][currentCoarseNode];
-  //           }
-
-  //           if((endRate[2] != coarseRate[2]) && (gCoarseNodesGIDs[col] > (gCoarseNodesPerDir[2] - 2)*fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
-  //             tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab + 1;
-  //             tmpVars[0] = gCoarseNodesGIDs[col] - (tmpInds[2] - 1)*fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
-  //           } else {
-  //             tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab;
-  //             tmpVars[0] = gCoarseNodesGIDs[col] % fineNodesPerCoarseSlab;
-  //           }
-  //           if((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
-  //             tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
-  //             tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
-  //           } else {
-  //             tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
-  //             tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
-  //           }
-  //           if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
-  //             tmpInds[0] = gCoarseNodesPerDir[0] - 1;
-  //           } else {
-  //             tmpInds[0] = tmpVars[1] / coarseRate[0];
-  //           }
-  //           gInd[2] = col / (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
-  //           tmp     = col % (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
-  //           gInd[1] = tmp / lCoarseNodesPerDir[0];
-  //           gInd[0] = tmp % lCoarseNodesPerDir[0];
-  //           lCol = gInd[2]*(lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]) + gInd[1]*lCoarseNodesPerDir[0] + gInd[0];
-  //           gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
-  //           coarseNodesGIDs[lCol] = gCoarseNodeOnCoarseGridGID;
-  //           for(LO dof = 0; dof < BlkSize; ++dof) {
-  //             colGIDs[BlkSize*lCol + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
-  //           }
-  //         }
-  //       }
-  //     }
-  //     // Now loop over the ghost nodes of the partition to add them to colGIDs
-  //     // since they will need to be included in the column map of P
-  //     for(col = lNumCoarseNodes; col < lNumCoarseNodes + lNumGhostNodes; ++col) {
-  //       if((endRate[2] != coarseRate[2]) && (ghostGIDs[ghostPermut[col - lNumCoarseNodes]] > (gCoarseNodesPerDir[2] - 2)*fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
-  //         tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab + 1;
-  //         tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] - (tmpInds[2] - 1)*fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
-  //       } else {
-  //         tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab;
-  //         tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] % fineNodesPerCoarseSlab;
-  //       }
-  //       if((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
-  //         tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
-  //         tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
-  //       } else {
-  //         tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
-  //         tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
-  //       }
-  //       if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
-  //         tmpInds[0] = gCoarseNodesPerDir[0] - 1;
-  //       } else {
-  //         tmpInds[0] = tmpVars[1] / coarseRate[0];
-  //       }
-  //       gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
-  //       for(LO dof = 0; dof < BlkSize; ++dof) {
-  //         colGIDs[BlkSize*col + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
-  //       }
-  //     }
-  //   } // End of scope for tmpInds, tmpVars and tmp
-
-  // } // GetGeometricData()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeStencil(
-                                const LO numDimensions, const Array<GO> currentNodeIndices,
-                                const Array<GO> coarseNodeIndices, const LO rate[3],
-                                const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord, const int interpolationOrder,
-                                std::vector<double>& stencil) const {
-
-    TEUCHOS_TEST_FOR_EXCEPTION((interpolationOrder > 1) || (interpolationOrder < 0),
-                               Exceptions::RuntimeError,
-                               "The interpolation order can be set to 0 or 1 only.");
+  // Set the values of the prolongation operators into the CrsMatrix P and call FillComplete
+  PCrs->setAllValues(iaP, jaP, valP);
+  PCrs->expertStaticFillComplete(domainMapP, rowMapP);
+}  // End MakeGeneralGeometricP
+
+// template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// void BlackBoxPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetGeometricData(
+//                       RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> >& coordinates, const Array<LO> coarseRate,
+//                       const Array<GO> gFineNodesPerDir, const Array<LO> lFineNodesPerDir, const LO BlkSize,
+//                       Array<GO>& gIndices, Array<LO>& myOffset, Array<bool>& ghostInterface, Array<LO>& endRate,
+//                       Array<GO>& gCoarseNodesPerDir, Array<LO>& lCoarseNodesPerDir, Array<GO>& ghostGIDs,
+//                       Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs, GO& gNumCoarseNodes, LO& lNumCoarseNodes,
+//                       ArrayRCP<Array<double> > coarseNodes) const {
+
+//   RCP<const Map> coordinatesMap = coordinates->getMap();
+//   LO numDimensions  = coordinates->getNumVectors();
+
+//   // Using the coarsening rate and the fine level data,
+//   // compute coarse level data
+
+//   //                              Phase 1                               //
+//   // ------------------------------------------------------------------ //
+//   // We first start by finding small informations on the mesh such as   //
+//   // the number of coarse nodes (local and global) and the number of    //
+//   // ghost nodes / the end rate of coarsening.                          //
+//   // ------------------------------------------------------------------ //
+//   GO minGlobalIndex = coordinatesMap->getMinGlobalIndex();
+//   {
+//     GO tmp;
+//     gIndices[2] = minGlobalIndex / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
+//     tmp         = minGlobalIndex % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
+//     gIndices[1] = tmp / gFineNodesPerDir[0];
+//     gIndices[0] = tmp % gFineNodesPerDir[0];
+
+//     myOffset[2] = gIndices[2] % coarseRate[2];
+//     myOffset[1] = gIndices[1] % coarseRate[1];
+//     myOffset[0] = gIndices[0] % coarseRate[0];
+//   }
+
+//   // Check whether ghost nodes are needed in each direction
+//   for(LO i=0; i < numDimensions; ++i) {
+//     if((gIndices[i] != 0) && (gIndices[i] % coarseRate[i] > 0)) {
+//       ghostInterface[2*i] = true;
+//     }
+//     if(((gIndices[i] + lFineNodesPerDir[i]) != gFineNodesPerDir[i]) && ((gIndices[i] + lFineNodesPerDir[i] - 1) % coarseRate[i] > 0)) {
+//       ghostInterface[2*i + 1] = true;
+//     }
+//   }
+
+//   for(LO i = 0; i < 3; ++i) {
+//     if(i < numDimensions) {
+//       lCoarseNodesPerDir[i] = (lFineNodesPerDir[i] + myOffset[i] - 1) / coarseRate[i];
+//       if(myOffset[i] == 0) { ++lCoarseNodesPerDir[i]; }
+//       gCoarseNodesPerDir[i] = (gFineNodesPerDir[i] - 1) / coarseRate[i];
+//       endRate[i]            = (gFineNodesPerDir[i] - 1) % coarseRate[i];
+//       if(endRate[i] == 0) {
+//         ++gCoarseNodesPerDir[i];
+//         endRate[i] = coarseRate[i];
+//       }
+//     } else {
+//       // Most quantities need to be set to 1 for extra dimensions
+//       // this is rather logical, an x-y plane is like a single layer
+//       // of nodes in the z direction...
+//       gCoarseNodesPerDir[i] = 1;
+//       lCoarseNodesPerDir[i] = 1;
+//       endRate[i]            = 1;
+//     }
+//   }
+
+//   gNumCoarseNodes = gCoarseNodesPerDir[0]*gCoarseNodesPerDir[1]*gCoarseNodesPerDir[2];
+//   lNumCoarseNodes = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];
+
+//   // For each direction, determine how many ghost points are required.
+//   LO lNumGhostNodes = 0;
+//   {
+//     const int complementaryIndices[3][2] = {{1,2}, {0,2}, {0,1}};
+//     LO tmp = 0;
+//     for(LO i = 0; i < 3; ++i) {
+//       // Check whether a face in direction i needs ghost nodes
+//       if(ghostInterface[2*i] || ghostInterface[2*i+1]) {
+//         if(i == 0) {tmp = lCoarseNodesPerDir[1]*lCoarseNodesPerDir[2];}
+//         if(i == 1) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[2];}
+//         if(i == 2) {tmp = lCoarseNodesPerDir[0]*lCoarseNodesPerDir[1];}
+//       }
+//       // If both faces in direction i need nodes, double the number of ghost nodes
+//       if(ghostInterface[2*i] && ghostInterface[2*i+1]) {tmp = 2*tmp;}
+//       lNumGhostNodes += tmp;
+
+//       // The corners and edges need to be checked in 2D / 3D to add more ghosts nodes
+//       for(LO j = 0; j < 2; ++j) {
+//         for(LO k = 0; k < 2; ++k) {
+//           // Check if two adjoining faces need ghost nodes and then add their common edge
+//           if(ghostInterface[2*complementaryIndices[i][0]+j] && ghostInterface[2*complementaryIndices[i][1]+k]) {
+//             lNumGhostNodes += lCoarseNodesPerDir[i];
+//             // Add corners if three adjoining faces need ghost nodes,
+//             // but add them only once! Hence when i == 0.
+//             if(ghostInterface[2*i] && (i == 0)) { lNumGhostNodes += 1; }
+//             if(ghostInterface[2*i+1] && (i == 0)) { lNumGhostNodes += 1; }
+//           }
+//         }
+//       }
+//       tmp = 0;
+//     }
+//   } // end of scope for tmp and complementaryIndices
+
+//   //                              Phase 2                               //
+//   // ------------------------------------------------------------------ //
+//   // Build a list of GIDs to import the required ghost nodes.           //
+//   // The ordering of the ghosts nodes will be as natural as possible,   //
+//   // i.e. it should follow the GID ordering of the mesh.                //
+//   // ------------------------------------------------------------------ //
+
+//   // Saddly we have to more or less redo what was just done to figure out the value of lNumGhostNodes,
+//   // there might be some optimization possibility here...
+//   ghostGIDs.resize(lNumGhostNodes);
+//   LO countGhosts = 0;
+//   // Get the GID of the first point on the current partition.
+//   GO startingGID = minGlobalIndex;
+//   Array<GO> startingIndices(3);
+//   // We still want ghost nodes even if have with a 0 offset,
+//   // except when we are on a boundary
+//   if(ghostInterface[4] && (myOffset[2] == 0)) {
+//     if(gIndices[2] + coarseRate[2] > gFineNodesPerDir[2]) {
+//       startingGID -= endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//     } else {
+//       startingGID -= coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//     }
+//   } else {
+//     startingGID -= myOffset[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//   }
+//   if(ghostInterface[2] && (myOffset[1] == 0)) {
+//     if(gIndices[1] + coarseRate[1] > gFineNodesPerDir[1]) {
+//       startingGID -= endRate[1]*gFineNodesPerDir[0];
+//     } else {
+//       startingGID -= coarseRate[1]*gFineNodesPerDir[0];
+//     }
+//   } else {
+//     startingGID -= myOffset[1]*gFineNodesPerDir[0];
+//   }
+//   if(ghostInterface[0] && (myOffset[0] == 0)) {
+//     if(gIndices[0] + coarseRate[0] > gFineNodesPerDir[0]) {
+//       startingGID -= endRate[0];
+//     } else {
+//       startingGID -= coarseRate[0];
+//     }
+//   } else {
+//     startingGID -= myOffset[0];
+//   }
+
+//   { // scope for tmp
+//     GO tmp;
+//     startingIndices[2] = startingGID / (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
+//     tmp = startingGID % (gFineNodesPerDir[1]*gFineNodesPerDir[0]);
+//     startingIndices[1] = tmp / gFineNodesPerDir[0];
+//     startingIndices[0] = tmp % gFineNodesPerDir[0];
+//   }
+
+//   GO ghostOffset[3] = {0, 0, 0};
+//   LO lengthZero  = lCoarseNodesPerDir[0];
+//   LO lengthOne   = lCoarseNodesPerDir[1];
+//   LO lengthTwo   = lCoarseNodesPerDir[2];
+//   if(ghostInterface[0]) {++lengthZero;}
+//   if(ghostInterface[1]) {++lengthZero;}
+//   if(ghostInterface[2]) {++lengthOne;}
+//   if(ghostInterface[3]) {++lengthOne;}
+//   if(ghostInterface[4]) {++lengthTwo;}
+//   if(ghostInterface[5]) {++lengthTwo;}
+
+//   // First check the bottom face as it will have the lowest GIDs
+//   if(ghostInterface[4]) {
+//     ghostOffset[2] = startingGID;
+//     for(LO j = 0; j < lengthOne; ++j) {
+//       if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
+//         ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
+//       } else {
+//         ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
+//       }
+//       for(LO k = 0; k < lengthZero; ++k) {
+//         if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
+//           ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
+//         } else {
+//           ghostOffset[0] = k*coarseRate[0];
+//         }
+//         // If the partition includes a changed rate at the edge, ghost nodes need to be picked carefully.
+//         // This if statement is repeated each time a ghost node is selected.
+//         ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+//         ++countGhosts;
+//       }
+//     }
+//   }
+
+//   // Sweep over the lCoarseNodesPerDir[2] coarse layers in direction 2 and gather necessary ghost nodes
+//   // located on these layers.
+//   for(LO i = 0; i < lengthTwo; ++i) {
+//     // Exclude the cases where ghost nodes exists on the faces in directions 2, these faces are swept
+//     // seperatly for efficiency.
+//     if( !((i == lengthTwo-1) && ghostInterface[5]) && !((i == 0) && ghostInterface[4]) ) {
+//       // Set the ghostOffset in direction 2 taking into account a possible endRate different
+//       // from the regular coarseRate.
+//       if( (i == lengthTwo-1) && !ghostInterface[5] ) {
+//         ghostOffset[2] = startingGID + ((i-1)*coarseRate[2] + endRate[2])*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//       } else {
+//         ghostOffset[2] = startingGID + i*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//       }
+//       for(LO j = 0; j < lengthOne; ++j) {
+//         if( (j == 0) && ghostInterface[2] ) {
+//           for(LO k = 0; k < lengthZero; ++k) {
+//             if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
+//               if(k == 0) {
+//                 ghostOffset[0] = endRate[0];
+//               } else {
+//                 ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
+//               }
+//             } else {
+//               ghostOffset[0] = k*coarseRate[0];
+//             }
+//             // In this case j == 0 so ghostOffset[1] is zero and can be ignored
+//             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[0];
+//             ++countGhosts;
+//           }
+//         } else if( (j == lengthOne-1) && ghostInterface[3] ) {
+//           // Set the ghostOffset in direction 1 taking into account a possible endRate different
+//           // from the regular coarseRate.
+//           if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
+//             ghostOffset[1] = ((j-1)*coarseRate[1] + endRate[1])*gFineNodesPerDir[0];
+//           } else {
+//             ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
+//           }
+//           for(LO k = 0; k < lengthZero; ++k) {
+//             if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
+//               ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
+//             } else {
+//               ghostOffset[0] = k*coarseRate[0];
+//             }
+//             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+//             ++countGhosts;
+//           }
+//         } else {
+//           // Set ghostOffset[1] for side faces sweep
+//           if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) {
+//             ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
+//           } else {
+//             ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
+//           }
+
+//           // Set ghostOffset[0], ghostGIDs and countGhosts
+//           if(ghostInterface[0]) { // In that case ghostOffset[0]==0, so we can ignore it
+//             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1];
+//             ++countGhosts;
+//           }
+//           if(ghostInterface[1]) { // Grab ghost point at the end of direction 0.
+//             if( (startingIndices[0] + (lengthZero-1)*coarseRate[0]) > gFineNodesPerDir[0] - 1 ) {
+//               if(lengthZero > 2) {
+//                 ghostOffset[0] = (lengthZero-2)*coarseRate[0] + endRate[0];
+//               } else {
+//                 ghostOffset[0] = endRate[0];
+//               }
+//             } else {
+//               ghostOffset[0] = (lengthZero-1)*coarseRate[0];
+//             }
+//             ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+//             ++countGhosts;
+//           }
+//         }
+//       }
+//     }
+//   }
+
+//   // Finally check the top face
+//   if(ghostInterface[5]) {
+//     if( startingIndices[2] + (lengthTwo-1)*coarseRate[2] + 1 > gFineNodesPerDir[2] ) {
+//       ghostOffset[2] = startingGID + ((lengthTwo-2)*coarseRate[2] + endRate[2])*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//     } else {
+//       ghostOffset[2] = startingGID + (lengthTwo-1)*coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//     }
+//     for(LO j = 0; j < lengthOne; ++j) {
+//       if( (j == lengthOne-1) && (startingIndices[1] + j*coarseRate[1] + 1 > gFineNodesPerDir[1]) ) { // && !ghostInterface[3] ) {
+//         ghostOffset[1] = ( (j-1)*coarseRate[1] + endRate[1] )*gFineNodesPerDir[0];
+//       } else {
+//         ghostOffset[1] = j*coarseRate[1]*gFineNodesPerDir[0];
+//       }
+//       for(LO k = 0; k < lengthZero; ++k) {
+//         if( (k == lengthZero-1) && (startingIndices[0] + k*coarseRate[0] + 1 > gFineNodesPerDir[0]) ) {
+//           ghostOffset[0] = (k-1)*coarseRate[0] + endRate[0];
+//         } else {
+//           ghostOffset[0] = k*coarseRate[0];
+//         }
+//         ghostGIDs[countGhosts] = ghostOffset[2] + ghostOffset[1] + ghostOffset[0];
+//         ++countGhosts;
+//       }
+//     }
+//   }
+
+//   //                              Phase 3                               //
+//   // ------------------------------------------------------------------ //
+//   // Final phase of this function, lists are being built to create the  //
+//   // column and domain maps of the projection as well as the map and    //
+//   // arrays for the coarse coordinates multivector.                     //
+//   // ------------------------------------------------------------------ //
+
+//   Array<GO> gCoarseNodesGIDs(lNumCoarseNodes);
+//   LO currentNode, offset2, offset1, offset0;
+//   // Find the GIDs of the coarse nodes on the partition.
+//   for(LO ind2 = 0; ind2 < lCoarseNodesPerDir[2]; ++ind2) {
+//     if(myOffset[2] == 0) {
+//       offset2 = startingIndices[2] + myOffset[2];
+//     } else {
+//       if(startingIndices[2] + endRate[2] == gFineNodesPerDir[2] - 1) {
+//         offset2 = startingIndices[2] + endRate[2];
+//       } else {
+//         offset2 = startingIndices[2] + coarseRate[2];
+//       }
+//     }
+//     if(offset2 + ind2*coarseRate[2] > gFineNodesPerDir[2] - 1) {
+//       offset2 += (ind2 - 1)*coarseRate[2] + endRate[2];
+//     } else {
+//       offset2 += ind2*coarseRate[2];
+//     }
+//     offset2 = offset2*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+
+//     for(LO ind1 = 0; ind1 < lCoarseNodesPerDir[1]; ++ind1) {
+//       if(myOffset[1] == 0) {
+//         offset1 = startingIndices[1] + myOffset[1];
+//       } else {
+//         if(startingIndices[1] + endRate[1] == gFineNodesPerDir[1] - 1) {
+//           offset1 = startingIndices[1] + endRate[1];
+//         } else {
+//           offset1 = startingIndices[1] + coarseRate[1];
+//         }
+//       }
+//       if(offset1 + ind1*coarseRate[1] > gFineNodesPerDir[1] - 1) {
+//         offset1 += (ind1 - 1)*coarseRate[1] + endRate[1];
+//       } else {
+//         offset1 += ind1*coarseRate[1];
+//       }
+//       offset1 = offset1*gFineNodesPerDir[0];
+//       for(LO ind0 = 0; ind0 < lCoarseNodesPerDir[0]; ++ind0) {
+//         offset0 = startingIndices[0];
+//         if(myOffset[0] == 0) {
+//           offset0 += ind0*coarseRate[0];
+//         } else {
+//           offset0 += (ind0 + 1)*coarseRate[0];
+//         }
+//         if(offset0 > gFineNodesPerDir[0] - 1) {offset0 += endRate[0] - coarseRate[0];}
+
+//         currentNode = ind2*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]
+//                     + ind1*lCoarseNodesPerDir[0]
+//                     + ind0;
+//         gCoarseNodesGIDs[currentNode] = offset2 + offset1 + offset0;
+//       }
+//     }
+//   }
+
+//   // Actual loop over all the coarse/ghost nodes to find their index on the coarse mesh
+//   // and the corresponding dofs that will need to be added to colMapP.
+//   colGIDs.resize(BlkSize*(lNumCoarseNodes+lNumGhostNodes));
+//   coarseNodesGIDs.resize(lNumCoarseNodes);
+//   for(LO i = 0; i < numDimensions; ++i) {coarseNodes[i].resize(lNumCoarseNodes);}
+//   GO fineNodesPerCoarseSlab    = coarseRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//   GO fineNodesEndCoarseSlab    = endRate[2]*gFineNodesPerDir[1]*gFineNodesPerDir[0];
+//   GO fineNodesPerCoarsePlane   = coarseRate[1]*gFineNodesPerDir[0];
+//   GO fineNodesEndCoarsePlane   = endRate[1]*gFineNodesPerDir[0];
+//   GO coarseNodesPerCoarseLayer = gCoarseNodesPerDir[1]*gCoarseNodesPerDir[0];
+//   GO gCoarseNodeOnCoarseGridGID;
+//   LO gInd[3], lCol;
+//   Array<int> ghostPIDs  (lNumGhostNodes);
+//   Array<LO>  ghostLIDs  (lNumGhostNodes);
+//   Array<LO>  ghostPermut(lNumGhostNodes);
+//   for(LO k = 0; k < lNumGhostNodes; ++k) {ghostPermut[k] = k;}
+//   coordinatesMap->getRemoteIndexList(ghostGIDs, ghostPIDs, ghostLIDs);
+//   sh_sort_permute(ghostPIDs.begin(),ghostPIDs.end(), ghostPermut.begin(),ghostPermut.end());
+
+//   { // scope for tmpInds, tmpVars and tmp.
+//     GO tmpInds[3], tmpVars[2];
+//     LO tmp;
+//     // Loop over the coarse nodes of the partition and add them to colGIDs
+//     // that will be used to construct the column and domain maps of P as well
+//     // as to construct the coarse coordinates map.
+//     // for(LO col = 0; col < lNumCoarseNodes; ++col) { // This should most likely be replaced by loops of lCoarseNodesPerDir[] to simplify arithmetics
+//     LO col = 0;
+//     LO firstCoarseNodeInds[3], currentCoarseNode;
+//     for(LO dim = 0; dim < 3; ++dim) {
+//       if(myOffset[dim] == 0) {
+//         firstCoarseNodeInds[dim] = 0;
+//       } else {
+//         firstCoarseNodeInds[dim] = coarseRate[dim] - myOffset[dim];
+//       }
+//     }
+//     Array<ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> > fineNodes(numDimensions);
+//     for(LO dim = 0; dim < numDimensions; ++dim) {fineNodes[dim] = coordinates->getData(dim);}
+//     for(LO k = 0; k < lCoarseNodesPerDir[2]; ++k) {
+//       for(LO j = 0; j < lCoarseNodesPerDir[1]; ++j) {
+//         for(LO i = 0; i < lCoarseNodesPerDir[0]; ++i) {
+//           col = k*lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0] + j*lCoarseNodesPerDir[0] + i;
+
+//           // Check for endRate
+//           currentCoarseNode = 0;
+//           if(firstCoarseNodeInds[0] + i*coarseRate[0] > lFineNodesPerDir[0] - 1) {
+//             currentCoarseNode += firstCoarseNodeInds[0] + (i-1)*coarseRate[0] + endRate[0];
+//           } else {
+//             currentCoarseNode += firstCoarseNodeInds[0] + i*coarseRate[0];
+//           }
+//           if(firstCoarseNodeInds[1] + j*coarseRate[1] > lFineNodesPerDir[1] - 1) {
+//             currentCoarseNode += (firstCoarseNodeInds[1] + (j-1)*coarseRate[1] + endRate[1])*lFineNodesPerDir[0];
+//           } else {
+//             currentCoarseNode += (firstCoarseNodeInds[1] + j*coarseRate[1])*lFineNodesPerDir[0];
+//           }
+//           if(firstCoarseNodeInds[2] + k*coarseRate[2] > lFineNodesPerDir[2] - 1) {
+//             currentCoarseNode += (firstCoarseNodeInds[2] + (k-1)*coarseRate[2] + endRate[2])*lFineNodesPerDir[1]*lFineNodesPerDir[0];
+//           } else {
+//             currentCoarseNode += (firstCoarseNodeInds[2] + k*coarseRate[2])*lFineNodesPerDir[1]*lFineNodesPerDir[0];
+//           }
+//           // Load coordinates
+//           for(LO dim = 0; dim < numDimensions; ++dim) {
+//             coarseNodes[dim][col] = fineNodes[dim][currentCoarseNode];
+//           }
+
+//           if((endRate[2] != coarseRate[2]) && (gCoarseNodesGIDs[col] > (gCoarseNodesPerDir[2] - 2)*fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
+//             tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab + 1;
+//             tmpVars[0] = gCoarseNodesGIDs[col] - (tmpInds[2] - 1)*fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
+//           } else {
+//             tmpInds[2] = gCoarseNodesGIDs[col] / fineNodesPerCoarseSlab;
+//             tmpVars[0] = gCoarseNodesGIDs[col] % fineNodesPerCoarseSlab;
+//           }
+//           if((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
+//             tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
+//             tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
+//           } else {
+//             tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
+//             tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
+//           }
+//           if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
+//             tmpInds[0] = gCoarseNodesPerDir[0] - 1;
+//           } else {
+//             tmpInds[0] = tmpVars[1] / coarseRate[0];
+//           }
+//           gInd[2] = col / (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
+//           tmp     = col % (lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]);
+//           gInd[1] = tmp / lCoarseNodesPerDir[0];
+//           gInd[0] = tmp % lCoarseNodesPerDir[0];
+//           lCol = gInd[2]*(lCoarseNodesPerDir[1]*lCoarseNodesPerDir[0]) + gInd[1]*lCoarseNodesPerDir[0] + gInd[0];
+//           gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
+//           coarseNodesGIDs[lCol] = gCoarseNodeOnCoarseGridGID;
+//           for(LO dof = 0; dof < BlkSize; ++dof) {
+//             colGIDs[BlkSize*lCol + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
+//           }
+//         }
+//       }
+//     }
+//     // Now loop over the ghost nodes of the partition to add them to colGIDs
+//     // since they will need to be included in the column map of P
+//     for(col = lNumCoarseNodes; col < lNumCoarseNodes + lNumGhostNodes; ++col) {
+//       if((endRate[2] != coarseRate[2]) && (ghostGIDs[ghostPermut[col - lNumCoarseNodes]] > (gCoarseNodesPerDir[2] - 2)*fineNodesPerCoarseSlab + fineNodesEndCoarseSlab - 1)) {
+//         tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab + 1;
+//         tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] - (tmpInds[2] - 1)*fineNodesPerCoarseSlab - fineNodesEndCoarseSlab;
+//       } else {
+//         tmpInds[2] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] / fineNodesPerCoarseSlab;
+//         tmpVars[0] = ghostGIDs[ghostPermut[col - lNumCoarseNodes]] % fineNodesPerCoarseSlab;
+//       }
+//       if((endRate[1] != coarseRate[1]) && (tmpVars[0] > (gCoarseNodesPerDir[1] - 2)*fineNodesPerCoarsePlane + fineNodesEndCoarsePlane - 1)) {
+//         tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane + 1;
+//         tmpVars[1] = tmpVars[0] - (tmpInds[1] - 1)*fineNodesPerCoarsePlane - fineNodesEndCoarsePlane;
+//       } else {
+//         tmpInds[1] = tmpVars[0] / fineNodesPerCoarsePlane;
+//         tmpVars[1] = tmpVars[0] % fineNodesPerCoarsePlane;
+//       }
+//       if(tmpVars[1] == gFineNodesPerDir[0] - 1) {
+//         tmpInds[0] = gCoarseNodesPerDir[0] - 1;
+//       } else {
+//         tmpInds[0] = tmpVars[1] / coarseRate[0];
+//       }
+//       gCoarseNodeOnCoarseGridGID = tmpInds[2]*coarseNodesPerCoarseLayer + tmpInds[1]*gCoarseNodesPerDir[0] + tmpInds[0];
+//       for(LO dof = 0; dof < BlkSize; ++dof) {
+//         colGIDs[BlkSize*col + dof] = BlkSize*gCoarseNodeOnCoarseGridGID + dof;
+//       }
+//     }
+//   } // End of scope for tmpInds, tmpVars and tmp
+
+// } // GetGeometricData()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeStencil(
+    const LO numDimensions, const Array<GO> currentNodeIndices,
+    const Array<GO> coarseNodeIndices, const LO rate[3],
+    const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord, const int interpolationOrder,
+    std::vector<double>& stencil) const {
+  TEUCHOS_TEST_FOR_EXCEPTION((interpolationOrder > 1) || (interpolationOrder < 0),
+                             Exceptions::RuntimeError,
+                             "The interpolation order can be set to 0 or 1 only.");
+
+  if (interpolationOrder == 0) {
+    ComputeConstantInterpolationStencil(numDimensions, currentNodeIndices, coarseNodeIndices,
+                                        rate, stencil);
+  } else if (interpolationOrder == 1) {
+    ComputeLinearInterpolationStencil(numDimensions, coord, stencil);
+  }
 
-    if(interpolationOrder == 0) {
-      ComputeConstantInterpolationStencil(numDimensions, currentNodeIndices, coarseNodeIndices,
-                                          rate, stencil);
-    } else if(interpolationOrder == 1) {
-      ComputeLinearInterpolationStencil(numDimensions, coord, stencil);
+}  // End ComputeStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeConstantInterpolationStencil(const LO numDimensions, const Array<GO> currentNodeIndices,
+                                        const Array<GO> coarseNodeIndices, const LO rate[3],
+                                        std::vector<double>& stencil) const {
+  LO coarseNode = 0;
+  if (numDimensions > 2) {
+    if ((currentNodeIndices[2] - coarseNodeIndices[2]) > (rate[2] / 2)) {
+      coarseNode += 4;
     }
-
-  } // End ComputeStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeConstantInterpolationStencil(const LO numDimensions, const Array<GO> currentNodeIndices,
-                                      const Array<GO> coarseNodeIndices, const LO rate[3],
-                                      std::vector<double>& stencil) const {
-
-    LO coarseNode = 0;
-    if(numDimensions > 2) {
-      if((currentNodeIndices[2] - coarseNodeIndices[2]) > (rate[2] / 2)) {
-        coarseNode += 4;
-      }
+  }
+  if (numDimensions > 1) {
+    if ((currentNodeIndices[1] - coarseNodeIndices[1]) > (rate[1] / 2)) {
+      coarseNode += 2;
     }
-    if(numDimensions > 1) {
-      if((currentNodeIndices[1] - coarseNodeIndices[1]) > (rate[1] / 2)) {
-        coarseNode += 2;
+  }
+  if ((currentNodeIndices[0] - coarseNodeIndices[0]) > (rate[0] / 2)) {
+    coarseNode += 1;
+  }
+  stencil[0] = coarseNode;
+
+}  // ComputeConstantInterpolationStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeLinearInterpolationStencil(const LO numDimensions, const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord,
+                                      std::vector<double>& stencil)
+        const {
+  //                7         8                Find xi, eta and zeta such that
+  //                x---------x
+  //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
+  //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
+  //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
+  //             |  | *P   |  |
+  //             |  x------|--x          We can do this with a Newton solver:
+  //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
+  //             |/        |/            We compute the Jacobian and iterate until convergence...
+  //  z  y       x---------x
+  //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
+  //  |/                                 give us the weights for the interpolation stencil!
+  //  o---x
+  //
+
+  Teuchos::SerialDenseMatrix<LO, double> Jacobian(numDimensions, numDimensions);
+  Teuchos::SerialDenseVector<LO, double> residual(numDimensions);
+  Teuchos::SerialDenseVector<LO, double> solutionDirection(numDimensions);
+  Teuchos::SerialDenseVector<LO, double> paramCoords(numDimensions);
+  Teuchos::SerialDenseSolver<LO, double> problem;
+  LO numTerms = std::pow(2, numDimensions), iter = 0, max_iter = 5;
+  double functions[4][8], norm_ref = 1, norm2 = 1, tol = 1e-5;
+  paramCoords.size(numDimensions);
+
+  while ((iter < max_iter) && (norm2 > tol * norm_ref)) {
+    ++iter;
+    norm2 = 0;
+    solutionDirection.size(numDimensions);
+    residual.size(numDimensions);
+    Jacobian = 0.0;
+
+    // Compute Jacobian and Residual
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      residual(i) = coord[0][i];  // Add coordinates from point of interest
+      for (LO k = 0; k < numTerms; ++k) {
+        residual(i) -= functions[0][k] * coord[k + 1][i];  // Remove contribution from all coarse points
       }
-    }
-    if((currentNodeIndices[0] - coarseNodeIndices[0]) > (rate[0] / 2)) {
-      coarseNode += 1;
-    }
-    stencil[0] = coarseNode;
-
-  } // ComputeConstantInterpolationStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeLinearInterpolationStencil(const LO numDimensions, const Array<Array<typename Teuchos::ScalarTraits<Scalar>::coordinateType> > coord,
-                                    std::vector<double>& stencil)
-    const {
-
-    //                7         8                Find xi, eta and zeta such that
-    //                x---------x
-    //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
-    //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
-    //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
-    //             |  | *P   |  |
-    //             |  x------|--x          We can do this with a Newton solver:
-    //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
-    //             |/        |/            We compute the Jacobian and iterate until convergence...
-    //  z  y       x---------x
-    //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
-    //  |/                                 give us the weights for the interpolation stencil!
-    //  o---x
-    //
-
-    Teuchos::SerialDenseMatrix<LO,double> Jacobian(numDimensions, numDimensions);
-    Teuchos::SerialDenseVector<LO,double> residual(numDimensions);
-    Teuchos::SerialDenseVector<LO,double> solutionDirection(numDimensions);
-    Teuchos::SerialDenseVector<LO,double> paramCoords(numDimensions);
-    Teuchos::SerialDenseSolver<LO,double> problem;
-    LO numTerms = std::pow(2,numDimensions), iter = 0, max_iter = 5;
-    double functions[4][8], norm_ref = 1, norm2 = 1, tol = 1e-5;
-    paramCoords.size(numDimensions);
-
-    while( (iter < max_iter) && (norm2 > tol*norm_ref) ) {
-      ++iter;
-      norm2 = 0;
-      solutionDirection.size(numDimensions);
-      residual.size(numDimensions);
-      Jacobian = 0.0;
-
-      // Compute Jacobian and Residual
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        residual(i) = coord[0][i];                 // Add coordinates from point of interest
-        for(LO k = 0; k < numTerms; ++k) {
-          residual(i) -= functions[0][k]*coord[k+1][i]; //Remove contribution from all coarse points
-        }
-        if(iter == 1) {
-          norm_ref += residual(i)*residual(i);
-          if(i == numDimensions - 1) {
-            norm_ref = std::sqrt(norm_ref);
-          }
+      if (iter == 1) {
+        norm_ref += residual(i) * residual(i);
+        if (i == numDimensions - 1) {
+          norm_ref = std::sqrt(norm_ref);
         }
+      }
 
-        for(LO j = 0; j < numDimensions; ++j) {
-          for(LO k = 0; k < numTerms; ++k) {
-            Jacobian(i,j) += functions[j+1][k]*coord[k+1][i];
-          }
+      for (LO j = 0; j < numDimensions; ++j) {
+        for (LO k = 0; k < numTerms; ++k) {
+          Jacobian(i, j) += functions[j + 1][k] * coord[k + 1][i];
         }
       }
+    }
 
-      // Set Jacobian, Vectors and solve problem
-      problem.setMatrix(Teuchos::rcp(&Jacobian, false));
-      problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
-      problem.factorWithEquilibration(true);
-      problem.solve();
-      problem.unequilibrateLHS();
+    // Set Jacobian, Vectors and solve problem
+    problem.setMatrix(Teuchos::rcp(&Jacobian, false));
+    problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
+    problem.factorWithEquilibration(true);
+    problem.solve();
+    problem.unequilibrateLHS();
 
-      for(LO i = 0; i < numDimensions; ++i) {
-        paramCoords(i) = paramCoords(i) + solutionDirection(i);
-      }
+    for (LO i = 0; i < numDimensions; ++i) {
+      paramCoords(i) = paramCoords(i) + solutionDirection(i);
+    }
 
-      // Recompute Residual norm
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        double tmp = coord[0][i];
-        for(LO k = 0; k < numTerms; ++k) {
-          tmp -= functions[0][k]*coord[k+1][i];
-        }
-        norm2 += tmp*tmp;
-        tmp = 0;
+    // Recompute Residual norm
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      double tmp = coord[0][i];
+      for (LO k = 0; k < numTerms; ++k) {
+        tmp -= functions[0][k] * coord[k + 1][i];
       }
-      norm2 = std::sqrt(norm2);
+      norm2 += tmp * tmp;
+      tmp = 0;
     }
+    norm2 = std::sqrt(norm2);
+  }
 
-    // Load the interpolation values onto the stencil.
-    for(LO i = 0; i < 8; ++i) {
-      stencil[i] = functions[0][i];
-    }
+  // Load the interpolation values onto the stencil.
+  for (LO i = 0; i < 8; ++i) {
+    stencil[i] = functions[0][i];
+  }
 
-  } // End ComputeLinearInterpolationStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetInterpolationFunctions(const LO numDimensions,
-                            const Teuchos::SerialDenseVector<LO,double> parameters,
-                            double functions[4][8]) const {
-    double xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
-    if(numDimensions == 1) {
-      xi = parameters[0];
-      denominator = 2.0;
-    } else if(numDimensions == 2) {
-      xi  = parameters[0];
-      eta = parameters[1];
-      denominator = 4.0;
-    } else if(numDimensions == 3) {
-      xi   = parameters[0];
-      eta  = parameters[1];
-      zeta = parameters[2];
-      denominator = 8.0;
-    }
+}  // End ComputeLinearInterpolationStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetInterpolationFunctions(const LO numDimensions,
+                              const Teuchos::SerialDenseVector<LO, double> parameters,
+                              double functions[4][8]) const {
+  double xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
+  if (numDimensions == 1) {
+    xi          = parameters[0];
+    denominator = 2.0;
+  } else if (numDimensions == 2) {
+    xi          = parameters[0];
+    eta         = parameters[1];
+    denominator = 4.0;
+  } else if (numDimensions == 3) {
+    xi          = parameters[0];
+    eta         = parameters[1];
+    zeta        = parameters[2];
+    denominator = 8.0;
+  }
 
-    functions[0][0] = (1.0 - xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][1] = (1.0 + xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][2] = (1.0 - xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][3] = (1.0 + xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][4] = (1.0 - xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][5] = (1.0 + xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][6] = (1.0 - xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[0][7] = (1.0 + xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[1][0] = -(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][1] =  (1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][2] = -(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][3] =  (1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][4] = -(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][5] =  (1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][6] = -(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[1][7] =  (1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[2][0] = -(1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][1] = -(1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][2] =  (1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][3] =  (1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][4] = -(1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][5] = -(1.0 + xi)*(1.0 + zeta) / denominator;
-    functions[2][6] =  (1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][7] =  (1.0 + xi)*(1.0 + zeta) / denominator;
-
-    functions[3][0] = -(1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][1] = -(1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][2] = -(1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][3] = -(1.0 + xi)*(1.0 + eta) / denominator;
-    functions[3][4] =  (1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][5] =  (1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][6] =  (1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][7] =  (1.0 + xi)*(1.0 + eta) / denominator;
-
-  } // End GetInterpolationFunctions
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort_permute(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const
-  {
-    typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
-    DT n = last1 - first1;
-    DT m = n / 2;
-    DT z = Teuchos::OrdinalTraits<DT>::zero();
-    while (m > z)
-      {
-        DT max = n - m;
-        for (DT j = 0; j < max; j++)
-          {
-            for (DT k = j; k >= 0; k-=m)
-              {
-                if (first1[first2[k+m]] >= first1[first2[k]])
-                  break;
-                std::swap(first2[k+m], first2[k]);
-              }
-          }
-        m = m/2;
+  functions[0][0] = (1.0 - xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][1] = (1.0 + xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][2] = (1.0 - xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][3] = (1.0 + xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][4] = (1.0 - xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][5] = (1.0 + xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][6] = (1.0 - xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[0][7] = (1.0 + xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[1][0] = -(1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][1] = (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][2] = -(1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][3] = (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][4] = -(1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][5] = (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][6] = -(1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[1][7] = (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[2][0] = -(1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][1] = -(1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][2] = (1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][3] = (1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][4] = -(1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][5] = -(1.0 + xi) * (1.0 + zeta) / denominator;
+  functions[2][6] = (1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][7] = (1.0 + xi) * (1.0 + zeta) / denominator;
+
+  functions[3][0] = -(1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][1] = -(1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][2] = -(1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][3] = -(1.0 + xi) * (1.0 + eta) / denominator;
+  functions[3][4] = (1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][5] = (1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][6] = (1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][7] = (1.0 + xi) * (1.0 + eta) / denominator;
+
+}  // End GetInterpolationFunctions
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort_permute(
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const {
+  typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
+  DT n = last1 - first1;
+  DT m = n / 2;
+  DT z = Teuchos::OrdinalTraits<DT>::zero();
+  while (m > z) {
+    DT max = n - m;
+    for (DT j = 0; j < max; j++) {
+      for (DT k = j; k >= 0; k -= m) {
+        if (first1[first2[k + m]] >= first1[first2[k]])
+          break;
+        std::swap(first2[k + m], first2[k]);
       }
-  } // End sh_sort_permute
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort2(
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
-                const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const
-  {
-    typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
-    DT n = last1 - first1;
-    DT m = n / 2;
-    DT z = Teuchos::OrdinalTraits<DT>::zero();
-    while (m > z)
-      {
-        DT max = n - m;
-        for (DT j = 0; j < max; j++)
-          {
-            for (DT k = j; k >= 0; k-=m)
-              {
-                if (first1[k+m] >= first1[k])
-                  break;
-                std::swap(first1[k+m], first1[k]);
-                std::swap(first2[k+m], first2[k]);
-              }
-          }
-        m = m/2;
-      }
-  } // End sh_sort2
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetGIDLocalLexicographic(const GO i, const GO j, const GO k,
-                           const Array<LO> coarseNodeFineIndices, const RCP<GeometricData> myGeo,
-                           const LO myRankIndex, const LO pi, const LO pj, const LO /* pk */,
-                           const typename std::vector<std::vector<GO> >::iterator blockStart,
-                           const typename std::vector<std::vector<GO> >::iterator blockEnd,
-                           GO& myGID, LO& myPID, LO& myLID) const {
-
-    LO ni = -1, nj = -1, li = -1, lj = -1, lk = -1;
-    LO myRankGuess = myRankIndex;
-    // We try to make a logical guess as to which PID owns the current coarse node
-    if(i == 0 && myGeo->ghostInterface[0]) {
-      --myRankGuess;
-    } else if((i == myGeo->ghostedCoarseNodesPerDir[0] - 1) && myGeo->ghostInterface[1]) {
-      ++myRankGuess;
     }
-    if(j == 0 && myGeo->ghostInterface[2]) {
-      myRankGuess -= pi;
-    } else if((j == myGeo->ghostedCoarseNodesPerDir[1] - 1) && myGeo->ghostInterface[3]) {
-      myRankGuess += pi;
-    }
-    if(k == 0 && myGeo->ghostInterface[4]) {
-      myRankGuess -= pj*pi;
-    } else if((k == myGeo->ghostedCoarseNodesPerDir[2] - 1) && myGeo->ghostInterface[5]) {
-      myRankGuess += pj*pi;
-    }
-    if(coarseNodeFineIndices[0] >= myGeo->meshData[myRankGuess][3]
-       && coarseNodeFineIndices[0] <= myGeo->meshData[myRankGuess][4]
-       && coarseNodeFineIndices[1] >= myGeo->meshData[myRankGuess][5]
-       && coarseNodeFineIndices[1] <= myGeo->meshData[myRankGuess][6]
-       && coarseNodeFineIndices[2] >= myGeo->meshData[myRankGuess][7]
-       && coarseNodeFineIndices[2] <= myGeo->meshData[myRankGuess][8]) {
-      myPID = myGeo->meshData[myRankGuess][0];
-      ni = myGeo->meshData[myRankGuess][4] - myGeo->meshData[myRankGuess][3] + 1;
-      nj = myGeo->meshData[myRankGuess][6] - myGeo->meshData[myRankGuess][5] + 1;
-      li = coarseNodeFineIndices[0] - myGeo->meshData[myRankGuess][3];
-      lj = coarseNodeFineIndices[1] - myGeo->meshData[myRankGuess][5];
-      lk = coarseNodeFineIndices[2] - myGeo->meshData[myRankGuess][7];
-      myLID = lk*nj*ni + lj*ni + li;
-      myGID = myGeo->meshData[myRankGuess][9] + myLID;
-    } else { // The guess failed, let us use the heavy artilery: std::find_if()
-      // It could be interesting to monitor how many times this branch of the code gets
-      // used as it is far more expensive than the above one...
-      auto nodeRank = std::find_if(blockStart, blockEnd,
-                                   [coarseNodeFineIndices](const std::vector<GO>& vec){
-                                     if(coarseNodeFineIndices[0] >= vec[3]
-                                        && coarseNodeFineIndices[0] <= vec[4]
-                                        && coarseNodeFineIndices[1] >= vec[5]
-                                        && coarseNodeFineIndices[1] <= vec[6]
-                                        && coarseNodeFineIndices[2] >= vec[7]
-                                        && coarseNodeFineIndices[2] <= vec[8]) {
-                                       return true;
-                                     } else {
-                                       return false;
-                                     }
-                                   });
-      myPID = (*nodeRank)[0];
-      ni = (*nodeRank)[4] - (*nodeRank)[3] + 1;
-      nj = (*nodeRank)[6] - (*nodeRank)[5] + 1;
-      li = coarseNodeFineIndices[0] - (*nodeRank)[3];
-      lj = coarseNodeFineIndices[1] - (*nodeRank)[5];
-      lk = coarseNodeFineIndices[2] - (*nodeRank)[7];
-      myLID = lk*nj*ni + lj*ni + li;
-      myGID = (*nodeRank)[9] + myLID;
+    m = m / 2;
+  }
+}  // End sh_sort_permute
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::sh_sort2(
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& last1,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& first2,
+    const typename Teuchos::Array<LocalOrdinal>::iterator& /* last2 */) const {
+  typedef typename std::iterator_traits<typename Teuchos::Array<LocalOrdinal>::iterator>::difference_type DT;
+  DT n = last1 - first1;
+  DT m = n / 2;
+  DT z = Teuchos::OrdinalTraits<DT>::zero();
+  while (m > z) {
+    DT max = n - m;
+    for (DT j = 0; j < max; j++) {
+      for (DT k = j; k >= 0; k -= m) {
+        if (first1[k + m] >= first1[k])
+          break;
+        std::swap(first1[k + m], first1[k]);
+        std::swap(first2[k + m], first2[k]);
+      }
     }
-  } // End GetGIDLocalLexicographic
+    m = m / 2;
+  }
+}  // End sh_sort2
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeneralGeometricPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetGIDLocalLexicographic(const GO i, const GO j, const GO k,
+                             const Array<LO> coarseNodeFineIndices, const RCP<GeometricData> myGeo,
+                             const LO myRankIndex, const LO pi, const LO pj, const LO /* pk */,
+                             const typename std::vector<std::vector<GO> >::iterator blockStart,
+                             const typename std::vector<std::vector<GO> >::iterator blockEnd,
+                             GO& myGID, LO& myPID, LO& myLID) const {
+  LO ni = -1, nj = -1, li = -1, lj = -1, lk = -1;
+  LO myRankGuess = myRankIndex;
+  // We try to make a logical guess as to which PID owns the current coarse node
+  if (i == 0 && myGeo->ghostInterface[0]) {
+    --myRankGuess;
+  } else if ((i == myGeo->ghostedCoarseNodesPerDir[0] - 1) && myGeo->ghostInterface[1]) {
+    ++myRankGuess;
+  }
+  if (j == 0 && myGeo->ghostInterface[2]) {
+    myRankGuess -= pi;
+  } else if ((j == myGeo->ghostedCoarseNodesPerDir[1] - 1) && myGeo->ghostInterface[3]) {
+    myRankGuess += pi;
+  }
+  if (k == 0 && myGeo->ghostInterface[4]) {
+    myRankGuess -= pj * pi;
+  } else if ((k == myGeo->ghostedCoarseNodesPerDir[2] - 1) && myGeo->ghostInterface[5]) {
+    myRankGuess += pj * pi;
+  }
+  if (coarseNodeFineIndices[0] >= myGeo->meshData[myRankGuess][3] && coarseNodeFineIndices[0] <= myGeo->meshData[myRankGuess][4] && coarseNodeFineIndices[1] >= myGeo->meshData[myRankGuess][5] && coarseNodeFineIndices[1] <= myGeo->meshData[myRankGuess][6] && coarseNodeFineIndices[2] >= myGeo->meshData[myRankGuess][7] && coarseNodeFineIndices[2] <= myGeo->meshData[myRankGuess][8]) {
+    myPID = myGeo->meshData[myRankGuess][0];
+    ni    = myGeo->meshData[myRankGuess][4] - myGeo->meshData[myRankGuess][3] + 1;
+    nj    = myGeo->meshData[myRankGuess][6] - myGeo->meshData[myRankGuess][5] + 1;
+    li    = coarseNodeFineIndices[0] - myGeo->meshData[myRankGuess][3];
+    lj    = coarseNodeFineIndices[1] - myGeo->meshData[myRankGuess][5];
+    lk    = coarseNodeFineIndices[2] - myGeo->meshData[myRankGuess][7];
+    myLID = lk * nj * ni + lj * ni + li;
+    myGID = myGeo->meshData[myRankGuess][9] + myLID;
+  } else {  // The guess failed, let us use the heavy artilery: std::find_if()
+    // It could be interesting to monitor how many times this branch of the code gets
+    // used as it is far more expensive than the above one...
+    auto nodeRank = std::find_if(blockStart, blockEnd,
+                                 [coarseNodeFineIndices](const std::vector<GO>& vec) {
+                                   if (coarseNodeFineIndices[0] >= vec[3] && coarseNodeFineIndices[0] <= vec[4] && coarseNodeFineIndices[1] >= vec[5] && coarseNodeFineIndices[1] <= vec[6] && coarseNodeFineIndices[2] >= vec[7] && coarseNodeFineIndices[2] <= vec[8]) {
+                                     return true;
+                                   } else {
+                                     return false;
+                                   }
+                                 });
+    myPID         = (*nodeRank)[0];
+    ni            = (*nodeRank)[4] - (*nodeRank)[3] + 1;
+    nj            = (*nodeRank)[6] - (*nodeRank)[5] + 1;
+    li            = coarseNodeFineIndices[0] - (*nodeRank)[3];
+    lj            = coarseNodeFineIndices[1] - (*nodeRank)[5];
+    lk            = coarseNodeFineIndices[2] - (*nodeRank)[7];
+    myLID         = lk * nj * ni + lj * ni + li;
+    myGID         = (*nodeRank)[9] + myLID;
+  }
+}  // End GetGIDLocalLexicographic
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GENERALGEOMETRICPFACTORY_SHORT
-#endif // MUELU_GENERALGEOMETRICPFACTORY_DEF_HPP
+#endif  // MUELU_GENERALGEOMETRICPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_decl.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_decl.hpp
index 877ff11718dd..e590b77d97e4 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_decl.hpp
@@ -57,64 +57,63 @@
 #include "MueLu_Level_fwd.hpp"
 #include "MueLu_Aggregates_fwd.hpp"
 
-namespace MueLu{
+namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeometricInterpolationPFactory : public PFactory {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeometricInterpolationPFactory : public PFactory {
 #undef MUELU_GEOMETRICINTERPOLATIONPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  // Declare useful types
+  using real_type                  = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using realvaluedmultivector_type = Xpetra::MultiVector<real_type, LO, GO, Node>;
 
-    // Declare useful types
-    using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using realvaluedmultivector_type = Xpetra::MultiVector<real_type,LO,GO,Node>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  GeometricInterpolationPFactory() {}
 
-    //! Constructor
-    GeometricInterpolationPFactory() { }
+  //! Destructor.
+  virtual ~GeometricInterpolationPFactory() {}
+  //@}
 
-    //! Destructor.
-    virtual ~GeometricInterpolationPFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+ private:
+  void BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const;
+  void BuildLinearP(Level& coarseLevel,
+                    RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
+                    RCP<realvaluedmultivector_type>& fineCoordinates,
+                    RCP<realvaluedmultivector_type>& ghostCoordinates,
+                    const int numDimensions, const bool keepD2,
+                    RCP<Matrix>& P) const;
+  void ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
+                                         const Array<Array<real_type> > coord,
+                                         Array<real_type>& stencil) const;
+  void GetInterpolationFunctions(const LO numDimensions,
+                                 const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
+                                 real_type functions[4][8]) const;
 
-  private:
-    void BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const;
-    void BuildLinearP(Level& coarseLevel,
-                      RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
-                      RCP<realvaluedmultivector_type>& fineCoordinates,
-                      RCP<realvaluedmultivector_type>& ghostCoordinates,
-                      const int numDimensions, const bool keepD2,
-                      RCP<Matrix>& P) const;
-    void ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
-                                           const Array<Array<real_type> > coord,
-                                           Array<real_type>& stencil) const;
-    void GetInterpolationFunctions(const LO numDimensions,
-                                   const Teuchos::SerialDenseVector<LO,real_type> parametricCoordinates,
-                                   real_type functions[4][8]) const;
+};  // class GeometricInterpolationPFactory
 
-  }; // class GeometricInterpolationPFactory
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_SHORT
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_DECL_HPP
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_def.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_def.hpp
index cf6e507277e2..63e4bb9fdb7c 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_def.hpp
@@ -59,605 +59,599 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("interp: build coarse coordinates");
-#undef  SET_VALID_ENTRY
-
-    // general variables needed in GeometricInterpolationPFactory
-    validParamList->set<RCP<const FactoryBase> >("A",                       	 Teuchos::null,
-                                                 "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("Aggregates",                   Teuchos::null,
-                                                 "Aggregates generated by StructuredAggregationFactory used to construct a piece-constant prolongator.");
-    validParamList->set<RCP<const FactoryBase> >("prolongatorGraph",             Teuchos::null,
-                                                 "Graph generated by StructuredAggregationFactory used to construct a piece-linear prolongator.");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null,
-                                                 "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
-    validParamList->set<RCP<const FactoryBase> >("coarseCoordinatesFineMap",     Teuchos::null,
-                                                 "map of the coarse coordinates' GIDs as indexed on the fine mesh.");
-    validParamList->set<RCP<const FactoryBase> >("coarseCoordinatesMap",         Teuchos::null,
-                                                 "map of the coarse coordinates' GIDs as indexed on the coarse mesh.");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",                    Teuchos::null,
-                                                 "Fine level nullspace used to construct the coarse level nullspace.");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",                Teuchos::null,
-                                                 "Number of spacial dimensions in the problem.");
-    validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim",           Teuchos::null,
-                                                 "Number of nodes per spatial dimension on the coarse grid.");
-    validParamList->set<RCP<const FactoryBase> >("structuredInterpolationOrder", Teuchos::null,
-    						 "Interpolation order for constructing the prolongator.");
-    validParamList->set<bool>                   ("keep coarse coords",           false, "Flag to keep coordinates for special coarse grid solve");
-    validParamList->set<bool>                   ("interp: remove small entries", true, "Remove small interpolation coeficient from prolongator to reduce fill-in on coarse level");
-
-    return validParamList;
+  SET_VALID_ENTRY("interp: build coarse coordinates");
+#undef SET_VALID_ENTRY
+
+  // general variables needed in GeometricInterpolationPFactory
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null,
+                                               "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null,
+                                               "Aggregates generated by StructuredAggregationFactory used to construct a piece-constant prolongator.");
+  validParamList->set<RCP<const FactoryBase> >("prolongatorGraph", Teuchos::null,
+                                               "Graph generated by StructuredAggregationFactory used to construct a piece-linear prolongator.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null,
+                                               "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("coarseCoordinatesFineMap", Teuchos::null,
+                                               "map of the coarse coordinates' GIDs as indexed on the fine mesh.");
+  validParamList->set<RCP<const FactoryBase> >("coarseCoordinatesMap", Teuchos::null,
+                                               "map of the coarse coordinates' GIDs as indexed on the coarse mesh.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null,
+                                               "Fine level nullspace used to construct the coarse level nullspace.");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spacial dimensions in the problem.");
+  validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimension on the coarse grid.");
+  validParamList->set<RCP<const FactoryBase> >("structuredInterpolationOrder", Teuchos::null,
+                                               "Interpolation order for constructing the prolongator.");
+  validParamList->set<bool>("keep coarse coords", false, "Flag to keep coordinates for special coarse grid solve");
+  validParamList->set<bool>("interp: remove small entries", true, "Remove small interpolation coeficient from prolongator to reduce fill-in on coarse level");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  const ParameterList& pL = GetParameterList();
+
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "numDimensions");
+  Input(fineLevel, "prolongatorGraph");
+  Input(fineLevel, "lCoarseNodesPerDim");
+  Input(fineLevel, "structuredInterpolationOrder");
+
+  if (pL.get<bool>("interp: build coarse coordinates") ||
+      Get<int>(fineLevel, "structuredInterpolationOrder") == 1) {
+    Input(fineLevel, "Coordinates");
+    Input(fineLevel, "coarseCoordinatesFineMap");
+    Input(fineLevel, "coarseCoordinatesMap");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "BuildP", coarseLevel);
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-    const ParameterList& pL = GetParameterList();
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "numDimensions");
-    Input(fineLevel, "prolongatorGraph");
-    Input(fineLevel, "lCoarseNodesPerDim");
-    Input(fineLevel, "structuredInterpolationOrder");
-
-    if( pL.get<bool>("interp: build coarse coordinates") ||
-	Get<int>(fineLevel, "structuredInterpolationOrder") == 1) {
-      Input(fineLevel, "Coordinates");
-      Input(fineLevel, "coarseCoordinatesFineMap");
-      Input(fineLevel, "coarseCoordinatesMap");
+  *out << "Starting GeometricInterpolationPFactory::BuildP." << std::endl;
+
+  // Get inputs from the parameter list
+  const ParameterList& pL           = GetParameterList();
+  const bool removeSmallEntries     = pL.get<bool>("interp: remove small entries");
+  const bool buildCoarseCoordinates = pL.get<bool>("interp: build coarse coordinates");
+  const int interpolationOrder      = Get<int>(fineLevel, "structuredInterpolationOrder");
+  const int numDimensions           = Get<int>(fineLevel, "numDimensions");
+
+  // Declared main input/outputs to be retrieved and placed on the fine resp. coarse level
+  RCP<Matrix> A                        = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<const CrsGraph> prolongatorGraph = Get<RCP<CrsGraph> >(fineLevel, "prolongatorGraph");
+  RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
+  RCP<Matrix> P;
+
+  // Check if we need to build coarse coordinates as they are used if we construct
+  // a linear interpolation prolongator
+  if (buildCoarseCoordinates || (interpolationOrder == 1)) {
+    SubFactoryMonitor sfm(*this, "BuildCoordinates", coarseLevel);
+    RCP<const Map> coarseCoordsFineMap = Get<RCP<const Map> >(fineLevel, "coarseCoordinatesFineMap");
+    RCP<const Map> coarseCoordsMap     = Get<RCP<const Map> >(fineLevel, "coarseCoordinatesMap");
+
+    fineCoordinates                  = Get<RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
+    coarseCoordinates                = Xpetra::MultiVectorFactory<real_type, LO, GO, Node>::Build(coarseCoordsFineMap,
+                                                                                                  fineCoordinates->getNumVectors());
+    RCP<const Import> coordsImporter = ImportFactory::Build(fineCoordinates->getMap(),
+                                                            coarseCoordsFineMap);
+    coarseCoordinates->doImport(*fineCoordinates, *coordsImporter, Xpetra::INSERT);
+    coarseCoordinates->replaceMap(coarseCoordsMap);
+
+    Set(coarseLevel, "Coordinates", coarseCoordinates);
+
+    if (pL.get<bool>("keep coarse coords")) {
+      coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
     }
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level& fineLevel, Level &coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  *out << "Fine and coarse coordinates have been loaded from the fine level and set on the coarse level." << std::endl;
+
+  if (interpolationOrder == 0) {
+    SubFactoryMonitor sfm(*this, "BuildConstantP", coarseLevel);
+    // Compute the prolongator using piece-wise constant interpolation
+    BuildConstantP(P, prolongatorGraph, A);
+  } else if (interpolationOrder == 1) {
+    // Compute the prolongator using piece-wise linear interpolation
+    // First get all the required coordinates to compute the local part of P
+    RCP<const Map> prolongatorColMap = prolongatorGraph->getColMap();
+
+    const size_t dofsPerNode   = static_cast<size_t>(A->GetFixedBlockSize());
+    const size_t numColIndices = prolongatorColMap->getLocalNumElements();
+    TEUCHOS_TEST_FOR_EXCEPTION((numColIndices % dofsPerNode) != 0,
+                               Exceptions::RuntimeError,
+                               "Something went wrong, the number of columns in the prolongator is not a multiple of dofsPerNode!");
+    const size_t numGhostCoords = numColIndices / dofsPerNode;
+    const GO indexBase          = prolongatorColMap->getIndexBase();
+    const GO coordIndexBase     = fineCoordinates->getMap()->getIndexBase();
+
+    ArrayView<const GO> prolongatorColIndices = prolongatorColMap->getLocalElementList();
+    Array<GO> ghostCoordIndices(numGhostCoords);
+    for (size_t ghostCoordIdx = 0; ghostCoordIdx < numGhostCoords; ++ghostCoordIdx) {
+      ghostCoordIndices[ghostCoordIdx] = (prolongatorColIndices[ghostCoordIdx * dofsPerNode] - indexBase) / dofsPerNode + coordIndexBase;
+    }
+    RCP<Map> ghostCoordMap = MapFactory::Build(fineCoordinates->getMap()->lib(),
+                                               prolongatorColMap->getGlobalNumElements() / dofsPerNode,
+                                               ghostCoordIndices(),
+                                               coordIndexBase,
+                                               fineCoordinates->getMap()->getComm());
+
+    RCP<realvaluedmultivector_type> ghostCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(ghostCoordMap,
+                                                                                                                fineCoordinates->getNumVectors());
+    RCP<const Import> ghostImporter                  = ImportFactory::Build(coarseCoordinates->getMap(),
+                                                                            ghostCoordMap);
+    ghostCoordinates->doImport(*coarseCoordinates, *ghostImporter, Xpetra::INSERT);
+
+    BuildLinearP(coarseLevel, A, prolongatorGraph, fineCoordinates,
+                 ghostCoordinates, numDimensions, removeSmallEntries, P);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildP(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "BuildP", coarseLevel);
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
+  *out << "The prolongator matrix has been built." << std::endl;
+
+  {
+    SubFactoryMonitor sfm(*this, "BuildNullspace", coarseLevel);
+    // Build the coarse nullspace
+    RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
+                                                                 fineNullspace->getNumVectors());
+
+    using helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    if (helpers::isTpetraBlockCrs(A)) {
+      // FIXME: BlockCrs doesn't currently support transpose apply, so we have to do this the hard way
+      RCP<Matrix> Ptrans = Utilities::Transpose(*P);
+      Ptrans->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+                    Teuchos::ScalarTraits<SC>::zero());
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    *out << "Starting GeometricInterpolationPFactory::BuildP." << std::endl;
-
-    // Get inputs from the parameter list
-    const ParameterList& pL = GetParameterList();
-    const bool removeSmallEntries     = pL.get<bool>("interp: remove small entries");
-    const bool buildCoarseCoordinates = pL.get<bool>("interp: build coarse coordinates");
-    const int interpolationOrder      = Get<int>(fineLevel, "structuredInterpolationOrder");
-    const int numDimensions           = Get<int>(fineLevel, "numDimensions");
-
-    // Declared main input/outputs to be retrieved and placed on the fine resp. coarse level
-    RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
-    RCP<const CrsGraph> prolongatorGraph = Get<RCP<CrsGraph> >(fineLevel, "prolongatorGraph");
-    RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
-    RCP<Matrix> P;
-
-    // Check if we need to build coarse coordinates as they are used if we construct
-    // a linear interpolation prolongator
-    if(buildCoarseCoordinates || (interpolationOrder == 1)) {
-      SubFactoryMonitor sfm(*this, "BuildCoordinates", coarseLevel);
-      RCP<const Map> coarseCoordsFineMap = Get< RCP<const Map> >(fineLevel, "coarseCoordinatesFineMap");
-      RCP<const Map> coarseCoordsMap = Get< RCP<const Map> >(fineLevel, "coarseCoordinatesMap");
-
-      fineCoordinates   = Get< RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
-      coarseCoordinates = Xpetra::MultiVectorFactory<real_type,LO,GO,Node>::Build(coarseCoordsFineMap,
-                                                                                  fineCoordinates->getNumVectors());
-      RCP<const Import> coordsImporter = ImportFactory::Build(fineCoordinates->getMap(),
-                                                              coarseCoordsFineMap);
-      coarseCoordinates->doImport(*fineCoordinates, *coordsImporter, Xpetra::INSERT);
-      coarseCoordinates->replaceMap(coarseCoordsMap);
-
-      Set(coarseLevel, "Coordinates", coarseCoordinates);
-
-      if(pL.get<bool>("keep coarse coords")) {
-        coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
-      }
+      P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
+               Teuchos::ScalarTraits<SC>::zero());
     }
 
-    *out << "Fine and coarse coordinates have been loaded from the fine level and set on the coarse level." << std::endl;
-
-
-    if(interpolationOrder == 0) {
-      SubFactoryMonitor sfm(*this, "BuildConstantP", coarseLevel);
-      // Compute the prolongator using piece-wise constant interpolation
-      BuildConstantP(P, prolongatorGraph, A);
-    } else if(interpolationOrder == 1) {
-      // Compute the prolongator using piece-wise linear interpolation
-      // First get all the required coordinates to compute the local part of P
-      RCP<const Map> prolongatorColMap = prolongatorGraph->getColMap();
-
-      const size_t dofsPerNode = static_cast<size_t>(A->GetFixedBlockSize());
-      const size_t numColIndices = prolongatorColMap->getLocalNumElements();
-      TEUCHOS_TEST_FOR_EXCEPTION((numColIndices % dofsPerNode) != 0,
-                                 Exceptions::RuntimeError,
-                                 "Something went wrong, the number of columns in the prolongator is not a multiple of dofsPerNode!");
-      const size_t numGhostCoords = numColIndices / dofsPerNode;
-      const GO indexBase = prolongatorColMap->getIndexBase();
-      const GO coordIndexBase = fineCoordinates->getMap()->getIndexBase();
-
-      ArrayView<const GO> prolongatorColIndices = prolongatorColMap->getLocalElementList();
-      Array<GO> ghostCoordIndices(numGhostCoords);
-      for(size_t ghostCoordIdx = 0; ghostCoordIdx < numGhostCoords; ++ghostCoordIdx) {
-        ghostCoordIndices[ghostCoordIdx]
-          = (prolongatorColIndices[ghostCoordIdx*dofsPerNode] - indexBase) / dofsPerNode
-          + coordIndexBase;
-      }
-      RCP<Map> ghostCoordMap = MapFactory::Build(fineCoordinates->getMap()->lib(),
-                                                 prolongatorColMap->getGlobalNumElements() / dofsPerNode,
-                                                 ghostCoordIndices(),
-                                                 coordIndexBase,
-                                                 fineCoordinates->getMap()->getComm());
-
-      RCP<realvaluedmultivector_type> ghostCoordinates
-        = Xpetra::MultiVectorFactory<real_type,LO,GO,NO>::Build(ghostCoordMap,
-                                                                fineCoordinates->getNumVectors());
-      RCP<const Import> ghostImporter = ImportFactory::Build(coarseCoordinates->getMap(),
-                                                             ghostCoordMap);
-      ghostCoordinates->doImport(*coarseCoordinates, *ghostImporter, Xpetra::INSERT);
-
-      BuildLinearP(coarseLevel, A, prolongatorGraph, fineCoordinates,
-                   ghostCoordinates, numDimensions, removeSmallEntries, P);
-    }
-
-    *out << "The prolongator matrix has been built." << std::endl;
-
-    {
-      SubFactoryMonitor sfm(*this, "BuildNullspace", coarseLevel);
-      // Build the coarse nullspace
-      RCP<MultiVector> fineNullspace   = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-      RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
-                                                                   fineNullspace->getNumVectors());
-      
-      using helpers=Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      if(helpers::isTpetraBlockCrs(A)) {
-        // FIXME: BlockCrs doesn't currently support transpose apply, so we have to do this the hard way
-        RCP<Matrix> Ptrans = Utilities::Transpose(*P);
-        Ptrans->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                 Teuchos::ScalarTraits<SC>::zero());
-      }
-      else {
-        P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
-                 Teuchos::ScalarTraits<SC>::zero());
-      }
-
-      Set(coarseLevel, "Nullspace", coarseNullspace);
-    }
-
-    *out << "The coarse nullspace is constructed and set on the coarse level." << std::endl;
+    Set(coarseLevel, "Nullspace", coarseNullspace);
+  }
 
-    Set(coarseLevel, "P", P);
+  *out << "The coarse nullspace is constructed and set on the coarse level." << std::endl;
 
-    *out << "GeometricInterpolationPFactory::BuildP has completed." << std::endl;
+  Set(coarseLevel, "P", P);
 
-  } // BuildP
+  *out << "GeometricInterpolationPFactory::BuildP has completed." << std::endl;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const {
+}  // BuildP
 
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-    *out << "BuildConstantP" << std::endl;
+  *out << "BuildConstantP" << std::endl;
 
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0]    = A->GetFixedBlockSize();
-    RCP<const StridedMap> stridedDomainMap =
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0] = A->GetFixedBlockSize();
+  RCP<const StridedMap> stridedDomainMap =
       StridedMapFactory::Build(prolongatorGraph->getDomainMap(), strideInfo);
 
-    *out << "Call prolongator constructor" << std::endl;
-
-    using helpers=Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    if(helpers::isTpetraBlockCrs(A)) {
-      SC one  = Teuchos::ScalarTraits<SC>::one();
-      SC zero = Teuchos::ScalarTraits<SC>::zero();
-      LO NSDim = A->GetStorageBlockSize();
-
-      // Build the exploded Map
-      RCP<const Map> BlockMap = prolongatorGraph->getDomainMap();
-      Teuchos::ArrayView<const GO> block_dofs = BlockMap->getLocalElementList();
-      Teuchos::Array<GO> point_dofs(block_dofs.size()*NSDim);
-      for(LO i=0, ct=0; i<block_dofs.size(); i++) {
-        for(LO j=0; j<NSDim; j++) {
-          point_dofs[ct] = block_dofs[i]*NSDim + j;
-          ct++;
-        }
-      }
-      
-      RCP<const Map> PointMap = MapFactory::Build(BlockMap->lib(),
-                                                  BlockMap->getGlobalNumElements() *NSDim,
-                                                  point_dofs(),
-                                                  BlockMap->getIndexBase(),
-                                                  BlockMap->getComm());
-      strideInfo[0]    = A->GetFixedBlockSize();
-      RCP<const StridedMap> stridedPointMap =  StridedMapFactory::Build(PointMap, strideInfo);
-
-      RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > P_xpetra = Xpetra::CrsMatrixFactory<SC,LO,GO,NO>::BuildBlock(prolongatorGraph, PointMap, A->getRangeMap(),NSDim);
-      RCP<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> >(P_xpetra);
-      if(P_tpetra.is_null()) throw std::runtime_error("BuildConstantP Matrix factory did not return a Tpetra::BlockCrsMatrix");
-      RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
-
-      // NOTE: Assumes block-diagonal prolongation
-      Teuchos::Array<LO> temp(1);
-      Teuchos::ArrayView<const LO> indices;
-      Teuchos::Array<Scalar> block(NSDim*NSDim, zero);
-      for(LO i=0; i<NSDim; i++)
-        block[i*NSDim+i] = one;
-      for(LO i=0; i<(int)prolongatorGraph->getLocalNumRows(); i++) {
-        prolongatorGraph->getLocalRowView(i,indices);
-        for(LO j=0; j<(LO)indices.size();j++) {
-          temp[0] = indices[j];
-          P_tpetra->replaceLocalValues(i,temp(),block());
-        }
+  *out << "Call prolongator constructor" << std::endl;
+
+  using helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  if (helpers::isTpetraBlockCrs(A)) {
+    SC one   = Teuchos::ScalarTraits<SC>::one();
+    SC zero  = Teuchos::ScalarTraits<SC>::zero();
+    LO NSDim = A->GetStorageBlockSize();
+
+    // Build the exploded Map
+    RCP<const Map> BlockMap                 = prolongatorGraph->getDomainMap();
+    Teuchos::ArrayView<const GO> block_dofs = BlockMap->getLocalElementList();
+    Teuchos::Array<GO> point_dofs(block_dofs.size() * NSDim);
+    for (LO i = 0, ct = 0; i < block_dofs.size(); i++) {
+      for (LO j = 0; j < NSDim; j++) {
+        point_dofs[ct] = block_dofs[i] * NSDim + j;
+        ct++;
       }
+    }
 
-      P = P_wrap;
-      if (A->IsView("stridedMaps") == true) {
-        P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedPointMap);
+    RCP<const Map> PointMap               = MapFactory::Build(BlockMap->lib(),
+                                                              BlockMap->getGlobalNumElements() * NSDim,
+                                                              point_dofs(),
+                                                              BlockMap->getIndexBase(),
+                                                              BlockMap->getComm());
+    strideInfo[0]                         = A->GetFixedBlockSize();
+    RCP<const StridedMap> stridedPointMap = StridedMapFactory::Build(PointMap, strideInfo);
+
+    RCP<Xpetra::CrsMatrix<SC, LO, GO, NO> > P_xpetra            = Xpetra::CrsMatrixFactory<SC, LO, GO, NO>::BuildBlock(prolongatorGraph, PointMap, A->getRangeMap(), NSDim);
+    RCP<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> >(P_xpetra);
+    if (P_tpetra.is_null()) throw std::runtime_error("BuildConstantP Matrix factory did not return a Tpetra::BlockCrsMatrix");
+    RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
+
+    // NOTE: Assumes block-diagonal prolongation
+    Teuchos::Array<LO> temp(1);
+    Teuchos::ArrayView<const LO> indices;
+    Teuchos::Array<Scalar> block(NSDim * NSDim, zero);
+    for (LO i = 0; i < NSDim; i++)
+      block[i * NSDim + i] = one;
+    for (LO i = 0; i < (int)prolongatorGraph->getLocalNumRows(); i++) {
+      prolongatorGraph->getLocalRowView(i, indices);
+      for (LO j = 0; j < (LO)indices.size(); j++) {
+        temp[0] = indices[j];
+        P_tpetra->replaceLocalValues(i, temp(), block());
       }
-      else {
-        P->CreateView("stridedMaps", P->getRangeMap(),   PointMap);
-      }
-    }
-    else {
-      // Create the prolongator matrix and its associated objects
-      RCP<ParameterList> dummyList = rcp(new ParameterList());
-      P = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
-      RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-      PCrs->setAllToScalar(1.0);
-      PCrs->fillComplete();
-
-      // set StridingInformation of P
-      if (A->IsView("stridedMaps") == true)
-        P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
-      else
-        P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
     }
 
-  } // BuildConstantP
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildLinearP(Level& coarseLevel,
-               RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
-               RCP<realvaluedmultivector_type>& fineCoordinates,
-               RCP<realvaluedmultivector_type>& ghostCoordinates,
-               const int numDimensions, const bool removeSmallEntries,
-               RCP<Matrix>& P) const {
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
+    P = P_wrap;
+    if (A->IsView("stridedMaps") == true) {
+      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedPointMap);
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+      P->CreateView("stridedMaps", P->getRangeMap(), PointMap);
     }
+  } else {
+    // Create the prolongator matrix and its associated objects
+    RCP<ParameterList> dummyList = rcp(new ParameterList());
+    P                            = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
+    RCP<CrsMatrix> PCrs          = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+    PCrs->setAllToScalar(1.0);
+    PCrs->fillComplete();
 
-    // Compute 2^numDimensions using bit logic to avoid round-off errors
-    const int numInterpolationPoints = 1 << numDimensions;
-    const int dofsPerNode = A->GetFixedBlockSize()/ A->GetStorageBlockSize();;
+    // set StridingInformation of P
+    if (A->IsView("stridedMaps") == true)
+      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
+    else
+      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
+  }
 
-    RCP<ParameterList> dummyList = rcp(new ParameterList());
-    P = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-    PCrs->resumeFill(); // The Epetra matrix is considered filled at this point.
-
-    {
-      *out << "Entering BuildLinearP" << std::endl;
-      SubFactoryMonitor sfm(*this, "BuildLinearP", coarseLevel);
-
-      // Extract coordinates for interpolation stencil calculations
-      const LO numFineNodes  = fineCoordinates->getLocalLength();
-      const LO numGhostNodes = ghostCoordinates->getLocalLength();
-      Array<ArrayRCP<const real_type> > fineCoords(3);
-      Array<ArrayRCP<const real_type> > ghostCoords(3);
-      const real_type realZero = Teuchos::as<real_type>(0.0);
-      ArrayRCP<real_type> fineZero(numFineNodes, realZero);
-      ArrayRCP<real_type> ghostZero(numGhostNodes, realZero);
-      for(int dim = 0; dim < 3; ++dim) {
-        if(dim < numDimensions) {
-          fineCoords[dim]  = fineCoordinates->getData(dim);
-          ghostCoords[dim] = ghostCoordinates->getData(dim);
-        } else {
-          fineCoords[dim]  = fineZero;
-          ghostCoords[dim] = ghostZero;
-        }
+}  // BuildConstantP
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildLinearP(Level& coarseLevel,
+                 RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
+                 RCP<realvaluedmultivector_type>& fineCoordinates,
+                 RCP<realvaluedmultivector_type>& ghostCoordinates,
+                 const int numDimensions, const bool removeSmallEntries,
+                 RCP<Matrix>& P) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  // Compute 2^numDimensions using bit logic to avoid round-off errors
+  const int numInterpolationPoints = 1 << numDimensions;
+  const int dofsPerNode            = A->GetFixedBlockSize() / A->GetStorageBlockSize();
+  ;
+
+  RCP<ParameterList> dummyList = rcp(new ParameterList());
+  P                            = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
+  RCP<CrsMatrix> PCrs          = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  PCrs->resumeFill();  // The Epetra matrix is considered filled at this point.
+
+  {
+    *out << "Entering BuildLinearP" << std::endl;
+    SubFactoryMonitor sfm(*this, "BuildLinearP", coarseLevel);
+
+    // Extract coordinates for interpolation stencil calculations
+    const LO numFineNodes  = fineCoordinates->getLocalLength();
+    const LO numGhostNodes = ghostCoordinates->getLocalLength();
+    Array<ArrayRCP<const real_type> > fineCoords(3);
+    Array<ArrayRCP<const real_type> > ghostCoords(3);
+    const real_type realZero = Teuchos::as<real_type>(0.0);
+    ArrayRCP<real_type> fineZero(numFineNodes, realZero);
+    ArrayRCP<real_type> ghostZero(numGhostNodes, realZero);
+    for (int dim = 0; dim < 3; ++dim) {
+      if (dim < numDimensions) {
+        fineCoords[dim]  = fineCoordinates->getData(dim);
+        ghostCoords[dim] = ghostCoordinates->getData(dim);
+      } else {
+        fineCoords[dim]  = fineZero;
+        ghostCoords[dim] = ghostZero;
       }
+    }
 
-      *out << "Coordinates extracted from the multivectors!" << std::endl;
-
-      { // Construct the linear interpolation prolongator
-        LO interpolationNodeIdx = 0, rowIdx = 0;
-        ArrayView<const LO> colIndices;
-        Array<SC> values;
-        Array<Array<real_type> > coords(numInterpolationPoints + 1);
-        Array<real_type> stencil(numInterpolationPoints);
-        for(LO nodeIdx = 0; nodeIdx < numFineNodes; ++nodeIdx) {
-          if(PCrs->getNumEntriesInLocalRow(nodeIdx*dofsPerNode) == 1) {
-            values.resize(1);
-            values[0] = 1.0;
-            for(LO dof = 0; dof < dofsPerNode; ++dof) {
-              rowIdx = nodeIdx*dofsPerNode + dof;
-              prolongatorGraph->getLocalRowView(rowIdx, colIndices);
-              PCrs->replaceLocalValues(rowIdx, colIndices, values());
-            }
-          } else {
-            // Extract the coordinates associated with the current node
-            // and the neighboring coarse nodes
-            coords[0].resize(3);
-            for(int dim = 0; dim < 3; ++dim) {
-              coords[0][dim] = fineCoords[dim][nodeIdx];
-            }
-            prolongatorGraph->getLocalRowView(nodeIdx*dofsPerNode, colIndices);
-            for(int interpolationIdx=0; interpolationIdx < numInterpolationPoints; ++interpolationIdx) {
-              coords[interpolationIdx + 1].resize(3);
-              interpolationNodeIdx = colIndices[interpolationIdx] / dofsPerNode;
-              for(int dim = 0; dim < 3; ++dim) {
-                coords[interpolationIdx + 1][dim] = ghostCoords[dim][interpolationNodeIdx];
-              }
-            }
-            RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Compute Linear Interpolation")));
-            ComputeLinearInterpolationStencil(numDimensions, numInterpolationPoints, coords, stencil);
-            tm = Teuchos::null;
-            values.resize(numInterpolationPoints);
-            for(LO valueIdx = 0; valueIdx < numInterpolationPoints; ++valueIdx) {
-              values[valueIdx] = Teuchos::as<SC>(stencil[valueIdx]);
+    *out << "Coordinates extracted from the multivectors!" << std::endl;
+
+    {  // Construct the linear interpolation prolongator
+      LO interpolationNodeIdx = 0, rowIdx = 0;
+      ArrayView<const LO> colIndices;
+      Array<SC> values;
+      Array<Array<real_type> > coords(numInterpolationPoints + 1);
+      Array<real_type> stencil(numInterpolationPoints);
+      for (LO nodeIdx = 0; nodeIdx < numFineNodes; ++nodeIdx) {
+        if (PCrs->getNumEntriesInLocalRow(nodeIdx * dofsPerNode) == 1) {
+          values.resize(1);
+          values[0] = 1.0;
+          for (LO dof = 0; dof < dofsPerNode; ++dof) {
+            rowIdx = nodeIdx * dofsPerNode + dof;
+            prolongatorGraph->getLocalRowView(rowIdx, colIndices);
+            PCrs->replaceLocalValues(rowIdx, colIndices, values());
+          }
+        } else {
+          // Extract the coordinates associated with the current node
+          // and the neighboring coarse nodes
+          coords[0].resize(3);
+          for (int dim = 0; dim < 3; ++dim) {
+            coords[0][dim] = fineCoords[dim][nodeIdx];
+          }
+          prolongatorGraph->getLocalRowView(nodeIdx * dofsPerNode, colIndices);
+          for (int interpolationIdx = 0; interpolationIdx < numInterpolationPoints; ++interpolationIdx) {
+            coords[interpolationIdx + 1].resize(3);
+            interpolationNodeIdx = colIndices[interpolationIdx] / dofsPerNode;
+            for (int dim = 0; dim < 3; ++dim) {
+              coords[interpolationIdx + 1][dim] = ghostCoords[dim][interpolationNodeIdx];
             }
+          }
+          RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Compute Linear Interpolation")));
+          ComputeLinearInterpolationStencil(numDimensions, numInterpolationPoints, coords, stencil);
+          tm = Teuchos::null;
+          values.resize(numInterpolationPoints);
+          for (LO valueIdx = 0; valueIdx < numInterpolationPoints; ++valueIdx) {
+            values[valueIdx] = Teuchos::as<SC>(stencil[valueIdx]);
+          }
 
-            // Set values in all the rows corresponding to nodeIdx
-            for(LO dof = 0; dof < dofsPerNode; ++dof) {
-              rowIdx = nodeIdx*dofsPerNode + dof;
-              prolongatorGraph->getLocalRowView(rowIdx, colIndices);
-              PCrs->replaceLocalValues(rowIdx, colIndices, values());
-            }
-          } // Check for Coarse vs. Fine point
-        } // Loop over fine nodes
-      } // Construct the linear interpolation prolongator
+          // Set values in all the rows corresponding to nodeIdx
+          for (LO dof = 0; dof < dofsPerNode; ++dof) {
+            rowIdx = nodeIdx * dofsPerNode + dof;
+            prolongatorGraph->getLocalRowView(rowIdx, colIndices);
+            PCrs->replaceLocalValues(rowIdx, colIndices, values());
+          }
+        }  // Check for Coarse vs. Fine point
+      }    // Loop over fine nodes
+    }      // Construct the linear interpolation prolongator
 
-      *out << "The calculation of the interpolation stencils has completed." << std::endl;
+    *out << "The calculation of the interpolation stencils has completed." << std::endl;
 
-      PCrs->fillComplete();
-    }
+    PCrs->fillComplete();
+  }
 
-    *out << "All values in P have been set and fillComplete has been performed." << std::endl;
-
-    // Note lbv Jan 29 2019: this should be handle at aggregation level
-    // if the user really does not want potential d2 neighbors on coarse grid
-    // that way we would avoid a new graph construction...
-
-    // Check if we want to remove small entries from P
-    // to reduce stencil growth on next level.
-    if(removeSmallEntries) {
-      *out << "Entering remove small entries" << std::endl;
-      SubFactoryMonitor sfm(*this, "remove small entries", coarseLevel);
-
-      ArrayRCP<const size_t> rowptrOrig;
-      ArrayRCP<const LO>     colindOrig;
-      ArrayRCP<const Scalar> valuesOrig;
-      PCrs->getAllValues(rowptrOrig, colindOrig, valuesOrig);
-
-      const size_t numRows = static_cast<size_t>(rowptrOrig.size() - 1);
-      ArrayRCP<size_t> rowPtr(numRows + 1);
-      ArrayRCP<size_t> nnzOnRows(numRows);
-      rowPtr[0] = 0;
-      size_t countRemovedEntries = 0;
-      for(size_t rowIdx = 0; rowIdx < numRows; ++rowIdx) {
-        for(size_t entryIdx = rowptrOrig[rowIdx]; entryIdx < rowptrOrig[rowIdx + 1]; ++entryIdx) {
-          if(Teuchos::ScalarTraits<Scalar>::magnitude(valuesOrig[entryIdx]) < 1e-6) {++countRemovedEntries;}
+  *out << "All values in P have been set and fillComplete has been performed." << std::endl;
+
+  // Note lbv Jan 29 2019: this should be handle at aggregation level
+  // if the user really does not want potential d2 neighbors on coarse grid
+  // that way we would avoid a new graph construction...
+
+  // Check if we want to remove small entries from P
+  // to reduce stencil growth on next level.
+  if (removeSmallEntries) {
+    *out << "Entering remove small entries" << std::endl;
+    SubFactoryMonitor sfm(*this, "remove small entries", coarseLevel);
+
+    ArrayRCP<const size_t> rowptrOrig;
+    ArrayRCP<const LO> colindOrig;
+    ArrayRCP<const Scalar> valuesOrig;
+    PCrs->getAllValues(rowptrOrig, colindOrig, valuesOrig);
+
+    const size_t numRows = static_cast<size_t>(rowptrOrig.size() - 1);
+    ArrayRCP<size_t> rowPtr(numRows + 1);
+    ArrayRCP<size_t> nnzOnRows(numRows);
+    rowPtr[0]                  = 0;
+    size_t countRemovedEntries = 0;
+    for (size_t rowIdx = 0; rowIdx < numRows; ++rowIdx) {
+      for (size_t entryIdx = rowptrOrig[rowIdx]; entryIdx < rowptrOrig[rowIdx + 1]; ++entryIdx) {
+        if (Teuchos::ScalarTraits<Scalar>::magnitude(valuesOrig[entryIdx]) < 1e-6) {
+          ++countRemovedEntries;
         }
-        rowPtr[rowIdx + 1] = rowptrOrig[rowIdx + 1] - countRemovedEntries;
-        nnzOnRows[rowIdx] = rowPtr[rowIdx + 1] - rowPtr[rowIdx];
       }
-      GetOStream(Statistics1) << "interp: number of small entries removed= " << countRemovedEntries << " / " << rowptrOrig[numRows] << std::endl;
-
-      size_t countKeptEntries = 0;
-      ArrayRCP<LO> colInd(rowPtr[numRows]);
-      ArrayRCP<SC> values(rowPtr[numRows]);
-      for(size_t entryIdx = 0; entryIdx < rowptrOrig[numRows]; ++entryIdx) {
-        if(Teuchos::ScalarTraits<Scalar>::magnitude(valuesOrig[entryIdx]) > 1e-6) {
-          colInd[countKeptEntries] = colindOrig[entryIdx];
-          values[countKeptEntries] = valuesOrig[entryIdx];
-          ++countKeptEntries;
-        }
+      rowPtr[rowIdx + 1] = rowptrOrig[rowIdx + 1] - countRemovedEntries;
+      nnzOnRows[rowIdx]  = rowPtr[rowIdx + 1] - rowPtr[rowIdx];
+    }
+    GetOStream(Statistics1) << "interp: number of small entries removed= " << countRemovedEntries << " / " << rowptrOrig[numRows] << std::endl;
+
+    size_t countKeptEntries = 0;
+    ArrayRCP<LO> colInd(rowPtr[numRows]);
+    ArrayRCP<SC> values(rowPtr[numRows]);
+    for (size_t entryIdx = 0; entryIdx < rowptrOrig[numRows]; ++entryIdx) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(valuesOrig[entryIdx]) > 1e-6) {
+        colInd[countKeptEntries] = colindOrig[entryIdx];
+        values[countKeptEntries] = valuesOrig[entryIdx];
+        ++countKeptEntries;
       }
-
-      P = rcp(new CrsMatrixWrap(prolongatorGraph->getRowMap(),
-                                prolongatorGraph->getColMap(),
-                                nnzOnRows));
-      RCP<CrsMatrix> PCrsSqueezed = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-      PCrsSqueezed->resumeFill(); // The Epetra matrix is considered filled at this point.
-      PCrsSqueezed->setAllValues(rowPtr, colInd, values);
-      PCrsSqueezed->expertStaticFillComplete(prolongatorGraph->getDomainMap(),
-                                             prolongatorGraph->getRangeMap());
     }
 
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0] = dofsPerNode;
-    RCP<const StridedMap> stridedDomainMap =
+    P                           = rcp(new CrsMatrixWrap(prolongatorGraph->getRowMap(),
+                                                        prolongatorGraph->getColMap(),
+                                                        nnzOnRows));
+    RCP<CrsMatrix> PCrsSqueezed = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+    PCrsSqueezed->resumeFill();  // The Epetra matrix is considered filled at this point.
+    PCrsSqueezed->setAllValues(rowPtr, colInd, values);
+    PCrsSqueezed->expertStaticFillComplete(prolongatorGraph->getDomainMap(),
+                                           prolongatorGraph->getRangeMap());
+  }
+
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0] = dofsPerNode;
+  RCP<const StridedMap> stridedDomainMap =
       StridedMapFactory::Build(prolongatorGraph->getDomainMap(), strideInfo);
 
-    *out << "The strided maps of P have been computed" << std::endl;
+  *out << "The strided maps of P have been computed" << std::endl;
 
-    // set StridingInformation of P
-    if (A->IsView("stridedMaps") == true) {
-      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
-    } else {
-      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
-    }
+  // set StridingInformation of P
+  if (A->IsView("stridedMaps") == true) {
+    P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
+  } else {
+    P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
+  }
 
-  } // BuildLinearP
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
-                                    const Array<Array<real_type> > coord,
-                                    Array<real_type>& stencil) const {
-
-    //                7         8                Find xi, eta and zeta such that
-    //                x---------x
-    //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
-    //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
-    //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
-    //             |  | *P   |  |
-    //             |  x------|--x          We can do this with a Newton solver:
-    //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
-    //             |/        |/            We compute the Jacobian and iterate until convergence...
-    //  z  y       x---------x
-    //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
-    //  |/                                 give us the weights for the interpolation stencil!
-    //  o---x
-    //
-
-    Teuchos::SerialDenseMatrix<LO,real_type> Jacobian(numDimensions, numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> residual(numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> solutionDirection(numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> paramCoords(numDimensions);
-    Teuchos::SerialDenseSolver<LO,real_type> problem;
-    int iter = 0, max_iter = 5;
-    real_type functions[4][8], norm_ref = 1.0, norm2 = 1.0, tol = 1.0e-5;
-    paramCoords.size(numDimensions);
-
-    while( (iter < max_iter) && (norm2 > tol*norm_ref) ) {
-      ++iter;
-      norm2 = 0.0;
-      solutionDirection.size(numDimensions);
-      residual.size(numDimensions);
-      Jacobian = 0.0;
-
-      // Compute Jacobian and Residual
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        residual(i) = coord[0][i];                 // Add coordinates from point of interest
-        for(LO k = 0; k < numInterpolationPoints; ++k) {
-          residual(i) -= functions[0][k]*coord[k+1][i]; //Remove contribution from all coarse points
-        }
-        if(iter == 1) {
-          norm_ref += residual(i)*residual(i);
-          if(i == numDimensions - 1) {
-            norm_ref = std::sqrt(norm_ref);
-          }
+}  // BuildLinearP
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
+                                      const Array<Array<real_type> > coord,
+                                      Array<real_type>& stencil) const {
+  //                7         8                Find xi, eta and zeta such that
+  //                x---------x
+  //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
+  //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
+  //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
+  //             |  | *P   |  |
+  //             |  x------|--x          We can do this with a Newton solver:
+  //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
+  //             |/        |/            We compute the Jacobian and iterate until convergence...
+  //  z  y       x---------x
+  //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
+  //  |/                                 give us the weights for the interpolation stencil!
+  //  o---x
+  //
+
+  Teuchos::SerialDenseMatrix<LO, real_type> Jacobian(numDimensions, numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> residual(numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> solutionDirection(numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> paramCoords(numDimensions);
+  Teuchos::SerialDenseSolver<LO, real_type> problem;
+  int iter = 0, max_iter              = 5;
+  real_type functions[4][8], norm_ref = 1.0, norm2 = 1.0, tol = 1.0e-5;
+  paramCoords.size(numDimensions);
+
+  while ((iter < max_iter) && (norm2 > tol * norm_ref)) {
+    ++iter;
+    norm2 = 0.0;
+    solutionDirection.size(numDimensions);
+    residual.size(numDimensions);
+    Jacobian = 0.0;
+
+    // Compute Jacobian and Residual
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      residual(i) = coord[0][i];  // Add coordinates from point of interest
+      for (LO k = 0; k < numInterpolationPoints; ++k) {
+        residual(i) -= functions[0][k] * coord[k + 1][i];  // Remove contribution from all coarse points
+      }
+      if (iter == 1) {
+        norm_ref += residual(i) * residual(i);
+        if (i == numDimensions - 1) {
+          norm_ref = std::sqrt(norm_ref);
         }
+      }
 
-        for(LO j = 0; j < numDimensions; ++j) {
-          for(LO k = 0; k < numInterpolationPoints; ++k) {
-            Jacobian(i,j) += functions[j+1][k]*coord[k+1][i];
-          }
+      for (LO j = 0; j < numDimensions; ++j) {
+        for (LO k = 0; k < numInterpolationPoints; ++k) {
+          Jacobian(i, j) += functions[j + 1][k] * coord[k + 1][i];
         }
       }
+    }
 
-      // Set Jacobian, Vectors and solve problem
-      problem.setMatrix(Teuchos::rcp(&Jacobian, false));
-      problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
-      if(problem.shouldEquilibrate()) {problem.factorWithEquilibration(true);}
-      problem.solve();
+    // Set Jacobian, Vectors and solve problem
+    problem.setMatrix(Teuchos::rcp(&Jacobian, false));
+    problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
+    if (problem.shouldEquilibrate()) {
+      problem.factorWithEquilibration(true);
+    }
+    problem.solve();
 
-      for(LO i = 0; i < numDimensions; ++i) {
-        paramCoords(i) = paramCoords(i) + solutionDirection(i);
-      }
+    for (LO i = 0; i < numDimensions; ++i) {
+      paramCoords(i) = paramCoords(i) + solutionDirection(i);
+    }
 
-      // Recompute Residual norm
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        real_type tmp = coord[0][i];
-        for(LO k = 0; k < numInterpolationPoints; ++k) {
-          tmp -= functions[0][k]*coord[k+1][i];
-        }
-        norm2 += tmp*tmp;
-        tmp = 0.0;
+    // Recompute Residual norm
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      real_type tmp = coord[0][i];
+      for (LO k = 0; k < numInterpolationPoints; ++k) {
+        tmp -= functions[0][k] * coord[k + 1][i];
       }
-      norm2 = std::sqrt(norm2);
+      norm2 += tmp * tmp;
+      tmp = 0.0;
     }
+    norm2 = std::sqrt(norm2);
+  }
 
-    // Load the interpolation values onto the stencil.
-    for(LO i = 0; i < numInterpolationPoints; ++i) {
-      stencil[i] = functions[0][i];
-    }
+  // Load the interpolation values onto the stencil.
+  for (LO i = 0; i < numInterpolationPoints; ++i) {
+    stencil[i] = functions[0][i];
+  }
 
-  } // End ComputeLinearInterpolationStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetInterpolationFunctions(const LO numDimensions,
-                            const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
-                            real_type functions[4][8]) const {
-    real_type xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
-    if(numDimensions == 1) {
-      xi = parametricCoordinates[0];
-      denominator = 2.0;
-    } else if(numDimensions == 2) {
-      xi  = parametricCoordinates[0];
-      eta = parametricCoordinates[1];
-      denominator = 4.0;
-    } else if(numDimensions == 3) {
-      xi   = parametricCoordinates[0];
-      eta  = parametricCoordinates[1];
-      zeta = parametricCoordinates[2];
-      denominator = 8.0;
-    }
+}  // End ComputeLinearInterpolationStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetInterpolationFunctions(const LO numDimensions,
+                              const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
+                              real_type functions[4][8]) const {
+  real_type xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
+  if (numDimensions == 1) {
+    xi          = parametricCoordinates[0];
+    denominator = 2.0;
+  } else if (numDimensions == 2) {
+    xi          = parametricCoordinates[0];
+    eta         = parametricCoordinates[1];
+    denominator = 4.0;
+  } else if (numDimensions == 3) {
+    xi          = parametricCoordinates[0];
+    eta         = parametricCoordinates[1];
+    zeta        = parametricCoordinates[2];
+    denominator = 8.0;
+  }
 
-    functions[0][0] = (1.0 - xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][1] = (1.0 + xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][2] = (1.0 - xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][3] = (1.0 + xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][4] = (1.0 - xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][5] = (1.0 + xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][6] = (1.0 - xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[0][7] = (1.0 + xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[1][0] = -(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][1] =  (1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][2] = -(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][3] =  (1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][4] = -(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][5] =  (1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][6] = -(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[1][7] =  (1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[2][0] = -(1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][1] = -(1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][2] =  (1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][3] =  (1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][4] = -(1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][5] = -(1.0 + xi)*(1.0 + zeta) / denominator;
-    functions[2][6] =  (1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][7] =  (1.0 + xi)*(1.0 + zeta) / denominator;
-
-    functions[3][0] = -(1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][1] = -(1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][2] = -(1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][3] = -(1.0 + xi)*(1.0 + eta) / denominator;
-    functions[3][4] =  (1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][5] =  (1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][6] =  (1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][7] =  (1.0 + xi)*(1.0 + eta) / denominator;
-
-  } // End GetInterpolationFunctions
-
-} // namespace MueLu
-
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEF_HPP
+  functions[0][0] = (1.0 - xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][1] = (1.0 + xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][2] = (1.0 - xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][3] = (1.0 + xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][4] = (1.0 - xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][5] = (1.0 + xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][6] = (1.0 - xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[0][7] = (1.0 + xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[1][0] = -(1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][1] = (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][2] = -(1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][3] = (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][4] = -(1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][5] = (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][6] = -(1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[1][7] = (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[2][0] = -(1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][1] = -(1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][2] = (1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][3] = (1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][4] = -(1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][5] = -(1.0 + xi) * (1.0 + zeta) / denominator;
+  functions[2][6] = (1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][7] = (1.0 + xi) * (1.0 + zeta) / denominator;
+
+  functions[3][0] = -(1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][1] = -(1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][2] = -(1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][3] = -(1.0 + xi) * (1.0 + eta) / denominator;
+  functions[3][4] = (1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][5] = (1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][6] = (1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][7] = (1.0 + xi) * (1.0 + eta) / denominator;
+
+}  // End GetInterpolationFunctions
+
+}  // namespace MueLu
+
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_decl.hpp
index 4fc7abb7fbb5..9e70b77c7d5b 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_decl.hpp
@@ -57,88 +57,87 @@
 #include "MueLu_Level_fwd.hpp"
 #include "MueLu_IndexManager_kokkos_fwd.hpp"
 
-namespace MueLu{
+namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeometricInterpolationPFactory_kokkos : public PFactory {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeometricInterpolationPFactory_kokkos : public PFactory {
 #undef MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  // Declare useful types
+  using real_type                  = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using realvaluedmultivector_type = Xpetra::MultiVector<real_type, LO, GO, Node>;
+  using device_type                = typename Node::device_type;
+  using execution_space            = typename Node::execution_space;
+  using impl_scalar_type           = typename Kokkos::ArithTraits<real_type>::val_type;
+  using coord_view_type            = typename Kokkos::View<impl_scalar_type**,
+                                                Kokkos::LayoutLeft,
+                                                device_type>;
 
-    // Declare useful types
-    using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using realvaluedmultivector_type = Xpetra::MultiVector<real_type,LO,GO,Node>;
-    using device_type      = typename Node::device_type;
-    using execution_space  = typename Node::execution_space;
-    using impl_scalar_type = typename Kokkos::ArithTraits<real_type>::val_type;
-    using coord_view_type  = typename Kokkos::View<impl_scalar_type**,
-                                                   Kokkos::LayoutLeft,
-                                                   device_type>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  GeometricInterpolationPFactory_kokkos() {}
 
-    //! Constructor
-    GeometricInterpolationPFactory_kokkos() { }
+  //! Destructor.
+  virtual ~GeometricInterpolationPFactory_kokkos() {}
+  //@}
 
-    //! Destructor.
-    virtual ~GeometricInterpolationPFactory_kokkos() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Public functors
+  //@{
 
-    //! @name Public functors
-    //@{
+  struct coarseCoordinatesBuilderFunctor {
+    IndexManager_kokkos geoData_;
+    coord_view_type fineCoordView_;
+    coord_view_type coarseCoordView_;
 
-    struct coarseCoordinatesBuilderFunctor{
+    coarseCoordinatesBuilderFunctor(RCP<IndexManager_kokkos> geoData,
+                                    coord_view_type fineCoordView,
+                                    coord_view_type coarseCoordView);
 
-      IndexManager_kokkos geoData_;
-      coord_view_type fineCoordView_;
-      coord_view_type coarseCoordView_;
+    KOKKOS_INLINE_FUNCTION
+    void operator()(const LO nodeIdx) const;
 
-      coarseCoordinatesBuilderFunctor(RCP<IndexManager_kokkos> geoData,
-                                      coord_view_type fineCoordView,
-                                      coord_view_type coarseCoordView);
+  };  // struct coarseCoordinatesBuilderFunctor
 
-      KOKKOS_INLINE_FUNCTION
-      void operator() (const LO nodeIdx) const;
+  //@}
 
-    }; // struct coarseCoordinatesBuilderFunctor
+  void BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const;
 
-    //@}
+ private:
+  void BuildLinearP(RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
+                    RCP<realvaluedmultivector_type>& fineCoordinates,
+                    RCP<realvaluedmultivector_type>& ghostCoordinates,
+                    const int numDimensions, RCP<Matrix>& P) const;
+  void ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
+                                         const Array<Array<real_type> > coord,
+                                         Array<real_type>& stencil) const;
+  void GetInterpolationFunctions(const LO numDimensions,
+                                 const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
+                                 real_type functions[4][8]) const;
 
-    void BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const;
-  private:
-    void BuildLinearP(RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
-                      RCP<realvaluedmultivector_type>& fineCoordinates,
-                      RCP<realvaluedmultivector_type>& ghostCoordinates,
-                      const int numDimensions, RCP<Matrix>& P) const;
-    void ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
-                                           const Array<Array<real_type> > coord,
-                                           Array<real_type>& stencil) const;
-    void GetInterpolationFunctions(const LO numDimensions,
-                                   const Teuchos::SerialDenseVector<LO,real_type> parametricCoordinates,
-                                   real_type functions[4][8]) const;
+};  // class GeometricInterpolationPFactory_kokkos
 
-  }; // class GeometricInterpolationPFactory_kokkos
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_SHORT
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_def.hpp
index eca95fe94b8e..5911eed2785c 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_GeometricInterpolationPFactory_kokkos_def.hpp
@@ -55,564 +55,556 @@
 
 #include "Xpetra_TpetraCrsMatrix.hpp"
 
-
 // Including this one last ensure that the short names of the above headers are defined properly
 #include "MueLu_GeometricInterpolationPFactory_kokkos_decl.hpp"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("interp: build coarse coordinates");
-#undef  SET_VALID_ENTRY
-
-    // general variables needed in GeometricInterpolationPFactory_kokkos
-    validParamList->set<RCP<const FactoryBase> >("A",                       Teuchos::null,
-                                                 "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("prolongatorGraph",             Teuchos::null,
-                                                 "Graph generated by StructuredAggregationFactory used to construct a piece-linear prolongator.");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null,
-                                                 "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",                    Teuchos::null,
-                                                 "Fine level nullspace used to construct the coarse level nullspace.");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",                Teuchos::null,
-                                                 "Number of spacial dimensions in the problem.");
-    validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim",           Teuchos::null,
-                                                 "Number of nodes per spatial dimension on the coarse grid.");
-    validParamList->set<RCP<const FactoryBase> >("indexManager",                 Teuchos::null,
-                                                 "The index manager associated with the local mesh.");
-    validParamList->set<RCP<const FactoryBase> >("structuredInterpolationOrder", Teuchos::null,
-    						 "Interpolation order for constructing the prolongator.");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    const ParameterList& pL = GetParameterList();
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "numDimensions");
-    Input(fineLevel, "prolongatorGraph");
-    Input(fineLevel, "lCoarseNodesPerDim");
-    Input(fineLevel, "structuredInterpolationOrder");
-
-    if( pL.get<bool>("interp: build coarse coordinates") ||
-        Get<int>(fineLevel, "structuredInterpolationOrder") == 1) {
-      Input(fineLevel, "Coordinates");
-      Input(fineLevel, "indexManager");
-    }
-
+  SET_VALID_ENTRY("interp: build coarse coordinates");
+#undef SET_VALID_ENTRY
+
+  // general variables needed in GeometricInterpolationPFactory_kokkos
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null,
+                                               "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("prolongatorGraph", Teuchos::null,
+                                               "Graph generated by StructuredAggregationFactory used to construct a piece-linear prolongator.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null,
+                                               "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null,
+                                               "Fine level nullspace used to construct the coarse level nullspace.");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spacial dimensions in the problem.");
+  validParamList->set<RCP<const FactoryBase> >("lCoarseNodesPerDim", Teuchos::null,
+                                               "Number of nodes per spatial dimension on the coarse grid.");
+  validParamList->set<RCP<const FactoryBase> >("indexManager", Teuchos::null,
+                                               "The index manager associated with the local mesh.");
+  validParamList->set<RCP<const FactoryBase> >("structuredInterpolationOrder", Teuchos::null,
+                                               "Interpolation order for constructing the prolongator.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  const ParameterList& pL = GetParameterList();
+
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "numDimensions");
+  Input(fineLevel, "prolongatorGraph");
+  Input(fineLevel, "lCoarseNodesPerDim");
+  Input(fineLevel, "structuredInterpolationOrder");
+
+  if (pL.get<bool>("interp: build coarse coordinates") ||
+      Get<int>(fineLevel, "structuredInterpolationOrder") == 1) {
+    Input(fineLevel, "Coordinates");
+    Input(fineLevel, "indexManager");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level& fineLevel, Level &coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "BuildP", coarseLevel);
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildP(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "BuildP", coarseLevel);
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
-
-    *out << "Starting GeometricInterpolationPFactory_kokkos::BuildP." << std::endl;
-
-    // Get inputs from the parameter list
-    const ParameterList& pL = GetParameterList();
-    const bool buildCoarseCoordinates = pL.get<bool>("interp: build coarse coordinates");
-    const int interpolationOrder      = Get<int>(fineLevel, "structuredInterpolationOrder");
-    const int numDimensions           = Get<int>(fineLevel, "numDimensions");
-
-    // Declared main input/outputs to be retrieved and placed on the fine resp. coarse level
-    RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
-    RCP<const CrsGraph> prolongatorGraph = Get<RCP<CrsGraph> >(fineLevel, "prolongatorGraph");
-    RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
-    RCP<Matrix> P;
-
-    // Check if we need to build coarse coordinates as they are used if we construct
-    // a linear interpolation prolongator
-    if(buildCoarseCoordinates || (interpolationOrder == 1)) {
-      SubFactoryMonitor sfm(*this, "BuildCoordinates", coarseLevel);
-
-      // Extract data from fine level
-      RCP<IndexManager_kokkos> geoData = Get<RCP<IndexManager_kokkos> >(fineLevel, "indexManager");
-      fineCoordinates = Get< RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
-
-      // Build coarse coordinates map/multivector
-      RCP<const Map> coarseCoordsMap = MapFactory::Build(fineCoordinates->getMap()->lib(),
-                                                         Teuchos::OrdinalTraits<GO>::invalid(),
-                                                         geoData->getNumCoarseNodes(),
-                                                         fineCoordinates->getMap()->getIndexBase(),
-                                                         fineCoordinates->getMap()->getComm());
-      coarseCoordinates = Xpetra::MultiVectorFactory<real_type,LO,GO,Node>::
+  *out << "Starting GeometricInterpolationPFactory_kokkos::BuildP." << std::endl;
+
+  // Get inputs from the parameter list
+  const ParameterList& pL           = GetParameterList();
+  const bool buildCoarseCoordinates = pL.get<bool>("interp: build coarse coordinates");
+  const int interpolationOrder      = Get<int>(fineLevel, "structuredInterpolationOrder");
+  const int numDimensions           = Get<int>(fineLevel, "numDimensions");
+
+  // Declared main input/outputs to be retrieved and placed on the fine resp. coarse level
+  RCP<Matrix> A                        = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<const CrsGraph> prolongatorGraph = Get<RCP<CrsGraph> >(fineLevel, "prolongatorGraph");
+  RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
+  RCP<Matrix> P;
+
+  // Check if we need to build coarse coordinates as they are used if we construct
+  // a linear interpolation prolongator
+  if (buildCoarseCoordinates || (interpolationOrder == 1)) {
+    SubFactoryMonitor sfm(*this, "BuildCoordinates", coarseLevel);
+
+    // Extract data from fine level
+    RCP<IndexManager_kokkos> geoData = Get<RCP<IndexManager_kokkos> >(fineLevel, "indexManager");
+    fineCoordinates                  = Get<RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
+
+    // Build coarse coordinates map/multivector
+    RCP<const Map> coarseCoordsMap = MapFactory::Build(fineCoordinates->getMap()->lib(),
+                                                       Teuchos::OrdinalTraits<GO>::invalid(),
+                                                       geoData->getNumCoarseNodes(),
+                                                       fineCoordinates->getMap()->getIndexBase(),
+                                                       fineCoordinates->getMap()->getComm());
+    coarseCoordinates              = Xpetra::MultiVectorFactory<real_type, LO, GO, Node>::
         Build(coarseCoordsMap, fineCoordinates->getNumVectors());
 
-      // Construct and launch functor to fill coarse coordinates values
-      // function should take a const view really
-      coarseCoordinatesBuilderFunctor myCoarseCoordinatesBuilder(geoData,
-                                                                 fineCoordinates->  getDeviceLocalView(Xpetra::Access::ReadWrite),
-                                                                 coarseCoordinates->getDeviceLocalView(Xpetra::Access::OverwriteAll));
-      Kokkos::parallel_for("GeometricInterpolation: build coarse coordinates",
-                           Kokkos::RangePolicy<execution_space>(0, geoData->getNumCoarseNodes()),
-                           myCoarseCoordinatesBuilder);
+    // Construct and launch functor to fill coarse coordinates values
+    // function should take a const view really
+    coarseCoordinatesBuilderFunctor myCoarseCoordinatesBuilder(geoData,
+                                                               fineCoordinates->getDeviceLocalView(Xpetra::Access::ReadWrite),
+                                                               coarseCoordinates->getDeviceLocalView(Xpetra::Access::OverwriteAll));
+    Kokkos::parallel_for("GeometricInterpolation: build coarse coordinates",
+                         Kokkos::RangePolicy<execution_space>(0, geoData->getNumCoarseNodes()),
+                         myCoarseCoordinatesBuilder);
 
-      Set(coarseLevel, "Coordinates", coarseCoordinates);
-    }
-
-    *out << "Fine and coarse coordinates have been loaded from the fine level and set on the coarse level." << std::endl;
-
-    if(interpolationOrder == 0) {
-      SubFactoryMonitor sfm(*this, "BuildConstantP", coarseLevel);
-      // Compute the prolongator using piece-wise constant interpolation
-      BuildConstantP(P, prolongatorGraph, A);
-    } else if(interpolationOrder == 1) {
-      // Compute the prolongator using piece-wise linear interpolation
-      // First get all the required coordinates to compute the local part of P
-      RCP<realvaluedmultivector_type> ghostCoordinates
-        = Xpetra::MultiVectorFactory<real_type,LO,GO,NO>::Build(prolongatorGraph->getColMap(),
-                                                                fineCoordinates->getNumVectors());
-      RCP<const Import> ghostImporter = ImportFactory::Build(coarseCoordinates->getMap(),
-                                                             prolongatorGraph->getColMap());
-      ghostCoordinates->doImport(*coarseCoordinates, *ghostImporter, Xpetra::INSERT);
-
-      SubFactoryMonitor sfm(*this, "BuildLinearP", coarseLevel);
-      BuildLinearP(A, prolongatorGraph, fineCoordinates, ghostCoordinates, numDimensions, P);
-    }
+    Set(coarseLevel, "Coordinates", coarseCoordinates);
+  }
 
-    *out << "The prolongator matrix has been built." << std::endl;
-
-    {
-      SubFactoryMonitor sfm(*this, "BuildNullspace", coarseLevel);
-      // Build the coarse nullspace
-      RCP<MultiVector> fineNullspace   = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-      RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
-                                                                   fineNullspace->getNumVectors());
-      P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
-               Teuchos::ScalarTraits<SC>::zero());
-      Set(coarseLevel, "Nullspace", coarseNullspace);
-    }
+  *out << "Fine and coarse coordinates have been loaded from the fine level and set on the coarse level." << std::endl;
+
+  if (interpolationOrder == 0) {
+    SubFactoryMonitor sfm(*this, "BuildConstantP", coarseLevel);
+    // Compute the prolongator using piece-wise constant interpolation
+    BuildConstantP(P, prolongatorGraph, A);
+  } else if (interpolationOrder == 1) {
+    // Compute the prolongator using piece-wise linear interpolation
+    // First get all the required coordinates to compute the local part of P
+    RCP<realvaluedmultivector_type> ghostCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(prolongatorGraph->getColMap(),
+                                                                                                                fineCoordinates->getNumVectors());
+    RCP<const Import> ghostImporter                  = ImportFactory::Build(coarseCoordinates->getMap(),
+                                                                            prolongatorGraph->getColMap());
+    ghostCoordinates->doImport(*coarseCoordinates, *ghostImporter, Xpetra::INSERT);
+
+    SubFactoryMonitor sfm(*this, "BuildLinearP", coarseLevel);
+    BuildLinearP(A, prolongatorGraph, fineCoordinates, ghostCoordinates, numDimensions, P);
+  }
 
-    *out << "The coarse nullspace is constructed and set on the coarse level." << std::endl;
+  *out << "The prolongator matrix has been built." << std::endl;
+
+  {
+    SubFactoryMonitor sfm(*this, "BuildNullspace", coarseLevel);
+    // Build the coarse nullspace
+    RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P->getDomainMap(),
+                                                                 fineNullspace->getNumVectors());
+    P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(),
+             Teuchos::ScalarTraits<SC>::zero());
+    Set(coarseLevel, "Nullspace", coarseNullspace);
+  }
 
-    Array<LO> lNodesPerDir = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
-    Set(coarseLevel, "numDimensions", numDimensions);
-    Set(coarseLevel, "lNodesPerDim", lNodesPerDir);
-    Set(coarseLevel, "P", P);
+  *out << "The coarse nullspace is constructed and set on the coarse level." << std::endl;
 
-    *out << "GeometricInterpolationPFactory_kokkos::BuildP has completed." << std::endl;
+  Array<LO> lNodesPerDir = Get<Array<LO> >(fineLevel, "lCoarseNodesPerDim");
+  Set(coarseLevel, "numDimensions", numDimensions);
+  Set(coarseLevel, "lNodesPerDim", lNodesPerDir);
+  Set(coarseLevel, "P", P);
 
-  } // BuildP
+  *out << "GeometricInterpolationPFactory_kokkos::BuildP has completed." << std::endl;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const {
+}  // BuildP
 
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildConstantP(RCP<Matrix>& P, RCP<const CrsGraph>& prolongatorGraph, RCP<Matrix>& A) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
 
-    *out << "BuildConstantP" << std::endl;
+  *out << "BuildConstantP" << std::endl;
 
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0]    = A->GetFixedBlockSize();
-    RCP<const StridedMap> stridedDomainMap =
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0] = A->GetFixedBlockSize();
+  RCP<const StridedMap> stridedDomainMap =
       StridedMapFactory::Build(prolongatorGraph->getDomainMap(), strideInfo);
 
-    *out << "Call prolongator constructor" << std::endl;
-    using helpers=Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    if(helpers::isTpetraBlockCrs(A)) {
-      LO NSDim = A->GetStorageBlockSize();
-
-      // Build the exploded Map
-      // FIXME: Should look at doing this on device
-      RCP<const Map> BlockMap = prolongatorGraph->getDomainMap();
-      Teuchos::ArrayView<const GO> block_dofs = BlockMap->getLocalElementList();
-      Teuchos::Array<GO> point_dofs(block_dofs.size()*NSDim);
-      for(LO i=0, ct=0; i<block_dofs.size(); i++) {
-        for(LO j=0; j<NSDim; j++) {
-          point_dofs[ct] = block_dofs[i]*NSDim + j;
-          ct++;
-        }
-      }
-      
-      RCP<const Map> PointMap = MapFactory::Build(BlockMap->lib(),
-                                                  BlockMap->getGlobalNumElements() *NSDim,
-                                                  point_dofs(),
-                                                  BlockMap->getIndexBase(),
-                                                  BlockMap->getComm());
-      strideInfo[0]    = A->GetFixedBlockSize();
-      RCP<const StridedMap> stridedPointMap =  StridedMapFactory::Build(PointMap, strideInfo);
-
-     RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > P_xpetra = Xpetra::CrsMatrixFactory<SC,LO,GO,NO>::BuildBlock(prolongatorGraph, PointMap, A->getRangeMap(),NSDim);
-      RCP<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> >(P_xpetra);
-      if(P_tpetra.is_null()) throw std::runtime_error("BuildConstantP: Matrix factory did not return a Tpetra::BlockCrsMatrix");
-      RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
-      
-      const LO stride = strideInfo[0]*strideInfo[0];
-      const LO in_stride = strideInfo[0];
-      typename CrsMatrix::local_graph_type localGraph = prolongatorGraph->getLocalGraphDevice();
-      auto rowptr  = localGraph.row_map;
-      auto indices = localGraph.entries;
-      auto values = P_tpetra->getTpetra_BlockCrsMatrix()->getValuesDeviceNonConst();
-
-      using ISC = typename Tpetra::BlockCrsMatrix<SC,LO,GO,NO>::impl_scalar_type;
-      ISC one = Teuchos::ScalarTraits<ISC>::one();
-
-      const Kokkos::TeamPolicy<execution_space> policy(prolongatorGraph->getLocalNumRows(), 1);
-      
-      Kokkos::parallel_for("MueLu:GeoInterpFact::BuildConstantP::fill", policy,
-                           KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type &thread) {
-                           auto row = thread.league_rank();
-                           for(LO j = (LO)rowptr[row]; j<(LO) rowptr[row+1]; j++) {
-                             LO block_offset = j*stride;
-                             for(LO k=0; k<in_stride; k++)
-                               values[block_offset + k*(in_stride+1) ] = one;
-                           }
-                         });
-
-      P = P_wrap;
-      if (A->IsView("stridedMaps") == true) {
-        P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedPointMap);
-      }
-      else {
-        P->CreateView("stridedMaps", P->getRangeMap(),   PointMap);
+  *out << "Call prolongator constructor" << std::endl;
+  using helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  if (helpers::isTpetraBlockCrs(A)) {
+    LO NSDim = A->GetStorageBlockSize();
+
+    // Build the exploded Map
+    // FIXME: Should look at doing this on device
+    RCP<const Map> BlockMap                 = prolongatorGraph->getDomainMap();
+    Teuchos::ArrayView<const GO> block_dofs = BlockMap->getLocalElementList();
+    Teuchos::Array<GO> point_dofs(block_dofs.size() * NSDim);
+    for (LO i = 0, ct = 0; i < block_dofs.size(); i++) {
+      for (LO j = 0; j < NSDim; j++) {
+        point_dofs[ct] = block_dofs[i] * NSDim + j;
+        ct++;
       }
     }
-    else {
-      // Create the prolongator matrix and its associated objects
-      RCP<ParameterList> dummyList = rcp(new ParameterList());
-      P = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
-      RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-      PCrs->setAllToScalar(1.0);
-      PCrs->fillComplete();
-      
-      // set StridingInformation of P
-      if (A->IsView("stridedMaps") == true) {
-        P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
-      } else {
-        P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
-      }
+
+    RCP<const Map> PointMap               = MapFactory::Build(BlockMap->lib(),
+                                                              BlockMap->getGlobalNumElements() * NSDim,
+                                                              point_dofs(),
+                                                              BlockMap->getIndexBase(),
+                                                              BlockMap->getComm());
+    strideInfo[0]                         = A->GetFixedBlockSize();
+    RCP<const StridedMap> stridedPointMap = StridedMapFactory::Build(PointMap, strideInfo);
+
+    RCP<Xpetra::CrsMatrix<SC, LO, GO, NO> > P_xpetra            = Xpetra::CrsMatrixFactory<SC, LO, GO, NO>::BuildBlock(prolongatorGraph, PointMap, A->getRangeMap(), NSDim);
+    RCP<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> >(P_xpetra);
+    if (P_tpetra.is_null()) throw std::runtime_error("BuildConstantP: Matrix factory did not return a Tpetra::BlockCrsMatrix");
+    RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
+
+    const LO stride                                 = strideInfo[0] * strideInfo[0];
+    const LO in_stride                              = strideInfo[0];
+    typename CrsMatrix::local_graph_type localGraph = prolongatorGraph->getLocalGraphDevice();
+    auto rowptr                                     = localGraph.row_map;
+    auto indices                                    = localGraph.entries;
+    auto values                                     = P_tpetra->getTpetra_BlockCrsMatrix()->getValuesDeviceNonConst();
+
+    using ISC = typename Tpetra::BlockCrsMatrix<SC, LO, GO, NO>::impl_scalar_type;
+    ISC one   = Teuchos::ScalarTraits<ISC>::one();
+
+    const Kokkos::TeamPolicy<execution_space> policy(prolongatorGraph->getLocalNumRows(), 1);
+
+    Kokkos::parallel_for(
+        "MueLu:GeoInterpFact::BuildConstantP::fill", policy,
+        KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread) {
+          auto row = thread.league_rank();
+          for (LO j = (LO)rowptr[row]; j < (LO)rowptr[row + 1]; j++) {
+            LO block_offset = j * stride;
+            for (LO k = 0; k < in_stride; k++)
+              values[block_offset + k * (in_stride + 1)] = one;
+          }
+        });
+
+    P = P_wrap;
+    if (A->IsView("stridedMaps") == true) {
+      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedPointMap);
+    } else {
+      P->CreateView("stridedMaps", P->getRangeMap(), PointMap);
     }
-      
-  } // BuildConstantP
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildLinearP(RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
-               RCP<realvaluedmultivector_type>& fineCoordinates,
-               RCP<realvaluedmultivector_type>& ghostCoordinates,
-               const int numDimensions, RCP<Matrix>& P) const {
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    // Create the prolongator matrix and its associated objects
+    RCP<ParameterList> dummyList = rcp(new ParameterList());
+    P                            = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
+    RCP<CrsMatrix> PCrs          = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+    PCrs->setAllToScalar(1.0);
+    PCrs->fillComplete();
+
+    // set StridingInformation of P
+    if (A->IsView("stridedMaps") == true) {
+      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
     } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
     }
+  }
 
-    *out << "Entering BuildLinearP" << std::endl;
-
-    // Extract coordinates for interpolation stencil calculations
-    const LO numFineNodes  = fineCoordinates->getLocalLength();
-    const LO numGhostNodes = ghostCoordinates->getLocalLength();
-    Array<ArrayRCP<const real_type> > fineCoords(3);
-    Array<ArrayRCP<const real_type> > ghostCoords(3);
-    const real_type realZero = Teuchos::as<real_type>(0.0);
-    ArrayRCP<real_type> fineZero(numFineNodes, realZero);
-    ArrayRCP<real_type> ghostZero(numGhostNodes, realZero);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        fineCoords[dim]  = fineCoordinates->getData(dim);
-        ghostCoords[dim] = ghostCoordinates->getData(dim);
-      } else {
-        fineCoords[dim]  = fineZero;
-        ghostCoords[dim] = ghostZero;
-      }
+}  // BuildConstantP
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildLinearP(RCP<Matrix>& A, RCP<const CrsGraph>& prolongatorGraph,
+                 RCP<realvaluedmultivector_type>& fineCoordinates,
+                 RCP<realvaluedmultivector_type>& ghostCoordinates,
+                 const int numDimensions, RCP<Matrix>& P) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_GEOMETRICINTERPOLATIONPFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  *out << "Entering BuildLinearP" << std::endl;
+
+  // Extract coordinates for interpolation stencil calculations
+  const LO numFineNodes  = fineCoordinates->getLocalLength();
+  const LO numGhostNodes = ghostCoordinates->getLocalLength();
+  Array<ArrayRCP<const real_type> > fineCoords(3);
+  Array<ArrayRCP<const real_type> > ghostCoords(3);
+  const real_type realZero = Teuchos::as<real_type>(0.0);
+  ArrayRCP<real_type> fineZero(numFineNodes, realZero);
+  ArrayRCP<real_type> ghostZero(numGhostNodes, realZero);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      fineCoords[dim]  = fineCoordinates->getData(dim);
+      ghostCoords[dim] = ghostCoordinates->getData(dim);
+    } else {
+      fineCoords[dim]  = fineZero;
+      ghostCoords[dim] = ghostZero;
     }
+  }
 
-    *out << "Coordinates extracted from the multivectors!" << std::endl;
+  *out << "Coordinates extracted from the multivectors!" << std::endl;
 
-    // Compute 2^numDimensions using bit logic to avoid round-off errors
-    const int numInterpolationPoints = 1 << numDimensions;
-    const int dofsPerNode = A->GetFixedBlockSize();
+  // Compute 2^numDimensions using bit logic to avoid round-off errors
+  const int numInterpolationPoints = 1 << numDimensions;
+  const int dofsPerNode            = A->GetFixedBlockSize();
 
-    std::vector<size_t> strideInfo(1);
-    strideInfo[0] = dofsPerNode;
-    RCP<const StridedMap> stridedDomainMap =
+  std::vector<size_t> strideInfo(1);
+  strideInfo[0] = dofsPerNode;
+  RCP<const StridedMap> stridedDomainMap =
       StridedMapFactory::Build(prolongatorGraph->getDomainMap(), strideInfo);
 
-    *out << "The maps of P have been computed" << std::endl;
-
-    RCP<ParameterList> dummyList = rcp(new ParameterList());
-    P = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-    PCrs->resumeFill(); // The Epetra matrix is considered filled at this point.
-
-    LO interpolationNodeIdx = 0, rowIdx = 0;
-    ArrayView<const LO> colIndices;
-    Array<SC> values;
-    Array<Array<real_type> > coords(numInterpolationPoints + 1);
-    Array<real_type> stencil(numInterpolationPoints);
-    for(LO nodeIdx = 0; nodeIdx < numFineNodes; ++nodeIdx) {
-      if(PCrs->getNumEntriesInLocalRow(nodeIdx*dofsPerNode) == 1) {
-        values.resize(1);
-        values[0] = 1.0;
-        for(LO dof = 0; dof < dofsPerNode; ++dof) {
-          rowIdx = nodeIdx*dofsPerNode + dof;
-          prolongatorGraph->getLocalRowView(rowIdx, colIndices);
-          PCrs->replaceLocalValues(rowIdx, colIndices, values());
-        }
-      } else {
-        // Extract the coordinates associated with the current node
-        // and the neighboring coarse nodes
-        coords[0].resize(3);
-        for(int dim = 0; dim < 3; ++dim) {
-          coords[0][dim] = fineCoords[dim][nodeIdx];
-        }
-        prolongatorGraph->getLocalRowView(nodeIdx*dofsPerNode, colIndices);
-        for(int interpolationIdx=0; interpolationIdx < numInterpolationPoints; ++interpolationIdx) {
-          coords[interpolationIdx + 1].resize(3);
-          interpolationNodeIdx = colIndices[interpolationIdx] / dofsPerNode;
-          for(int dim = 0; dim < 3; ++dim) {
-            coords[interpolationIdx + 1][dim] = ghostCoords[dim][interpolationNodeIdx];
-          }
-        }
-        ComputeLinearInterpolationStencil(numDimensions, numInterpolationPoints, coords, stencil);
-        values.resize(numInterpolationPoints);
-        for(LO valueIdx = 0; valueIdx < numInterpolationPoints; ++valueIdx) {
-          values[valueIdx] = Teuchos::as<SC>(stencil[valueIdx]);
+  *out << "The maps of P have been computed" << std::endl;
+
+  RCP<ParameterList> dummyList = rcp(new ParameterList());
+  P                            = rcp(new CrsMatrixWrap(prolongatorGraph, dummyList));
+  RCP<CrsMatrix> PCrs          = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  PCrs->resumeFill();  // The Epetra matrix is considered filled at this point.
+
+  LO interpolationNodeIdx = 0, rowIdx = 0;
+  ArrayView<const LO> colIndices;
+  Array<SC> values;
+  Array<Array<real_type> > coords(numInterpolationPoints + 1);
+  Array<real_type> stencil(numInterpolationPoints);
+  for (LO nodeIdx = 0; nodeIdx < numFineNodes; ++nodeIdx) {
+    if (PCrs->getNumEntriesInLocalRow(nodeIdx * dofsPerNode) == 1) {
+      values.resize(1);
+      values[0] = 1.0;
+      for (LO dof = 0; dof < dofsPerNode; ++dof) {
+        rowIdx = nodeIdx * dofsPerNode + dof;
+        prolongatorGraph->getLocalRowView(rowIdx, colIndices);
+        PCrs->replaceLocalValues(rowIdx, colIndices, values());
+      }
+    } else {
+      // Extract the coordinates associated with the current node
+      // and the neighboring coarse nodes
+      coords[0].resize(3);
+      for (int dim = 0; dim < 3; ++dim) {
+        coords[0][dim] = fineCoords[dim][nodeIdx];
+      }
+      prolongatorGraph->getLocalRowView(nodeIdx * dofsPerNode, colIndices);
+      for (int interpolationIdx = 0; interpolationIdx < numInterpolationPoints; ++interpolationIdx) {
+        coords[interpolationIdx + 1].resize(3);
+        interpolationNodeIdx = colIndices[interpolationIdx] / dofsPerNode;
+        for (int dim = 0; dim < 3; ++dim) {
+          coords[interpolationIdx + 1][dim] = ghostCoords[dim][interpolationNodeIdx];
         }
+      }
+      ComputeLinearInterpolationStencil(numDimensions, numInterpolationPoints, coords, stencil);
+      values.resize(numInterpolationPoints);
+      for (LO valueIdx = 0; valueIdx < numInterpolationPoints; ++valueIdx) {
+        values[valueIdx] = Teuchos::as<SC>(stencil[valueIdx]);
+      }
 
-        // Set values in all the rows corresponding to nodeIdx
-        for(LO dof = 0; dof < dofsPerNode; ++dof) {
-          rowIdx = nodeIdx*dofsPerNode + dof;
-          prolongatorGraph->getLocalRowView(rowIdx, colIndices);
-          PCrs->replaceLocalValues(rowIdx, colIndices, values());
-        }
+      // Set values in all the rows corresponding to nodeIdx
+      for (LO dof = 0; dof < dofsPerNode; ++dof) {
+        rowIdx = nodeIdx * dofsPerNode + dof;
+        prolongatorGraph->getLocalRowView(rowIdx, colIndices);
+        PCrs->replaceLocalValues(rowIdx, colIndices, values());
       }
     }
+  }
 
-    *out << "The calculation of the interpolation stencils has completed." << std::endl;
-
-    PCrs->fillComplete();
+  *out << "The calculation of the interpolation stencils has completed." << std::endl;
 
-    *out << "All values in P have been set and expertStaticFillComplete has been performed." << std::endl;
+  PCrs->fillComplete();
 
-    // set StridingInformation of P
-    if (A->IsView("stridedMaps") == true) {
-      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
-    } else {
-      P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
-    }
+  *out << "All values in P have been set and expertStaticFillComplete has been performed." << std::endl;
 
-  } // BuildLinearP
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
-                                    const Array<Array<real_type> > coord,
-                                    Array<real_type>& stencil) const {
-
-    //                7         8                Find xi, eta and zeta such that
-    //                x---------x
-    //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
-    //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
-    //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
-    //             |  | *P   |  |
-    //             |  x------|--x          We can do this with a Newton solver:
-    //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
-    //             |/        |/            We compute the Jacobian and iterate until convergence...
-    //  z  y       x---------x
-    //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
-    //  |/                                 give us the weights for the interpolation stencil!
-    //  o---x
-    //
-
-    Teuchos::SerialDenseMatrix<LO,real_type> Jacobian(numDimensions, numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> residual(numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> solutionDirection(numDimensions);
-    Teuchos::SerialDenseVector<LO,real_type> paramCoords(numDimensions);
-    Teuchos::SerialDenseSolver<LO,real_type> problem;
-    int iter = 0, max_iter = 5;
-    real_type functions[4][8], norm_ref = 1.0, norm2 = 1.0, tol = 1.0e-5;
-    paramCoords.size(numDimensions);
-
-    while( (iter < max_iter) && (norm2 > tol*norm_ref) ) {
-      ++iter;
-      norm2 = 0.0;
-      solutionDirection.size(numDimensions);
-      residual.size(numDimensions);
-      Jacobian = 0.0;
-
-      // Compute Jacobian and Residual
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        residual(i) = coord[0][i];                 // Add coordinates from point of interest
-        for(LO k = 0; k < numInterpolationPoints; ++k) {
-          residual(i) -= functions[0][k]*coord[k+1][i]; //Remove contribution from all coarse points
-        }
-        if(iter == 1) {
-          norm_ref += residual(i)*residual(i);
-          if(i == numDimensions - 1) {
-            norm_ref = std::sqrt(norm_ref);
-          }
-        }
+  // set StridingInformation of P
+  if (A->IsView("stridedMaps") == true) {
+    P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), stridedDomainMap);
+  } else {
+    P->CreateView("stridedMaps", P->getRangeMap(), stridedDomainMap);
+  }
 
-        for(LO j = 0; j < numDimensions; ++j) {
-          for(LO k = 0; k < numInterpolationPoints; ++k) {
-            Jacobian(i,j) += functions[j+1][k]*coord[k+1][i];
-          }
-        }
+}  // BuildLinearP
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ComputeLinearInterpolationStencil(const int numDimensions, const int numInterpolationPoints,
+                                      const Array<Array<real_type> > coord,
+                                      Array<real_type>& stencil) const {
+  //                7         8                Find xi, eta and zeta such that
+  //                x---------x
+  //               /|        /|          Rx = x_p - sum N_i(xi,eta,zeta)x_i = 0
+  //             5/ |      6/ |          Ry = y_p - sum N_i(xi,eta,zeta)y_i = 0
+  //             x---------x  |          Rz = z_p - sum N_i(xi,eta,zeta)z_i = 0
+  //             |  | *P   |  |
+  //             |  x------|--x          We can do this with a Newton solver:
+  //             | /3      | /4          We will start with initial guess (xi,eta,zeta) = (0,0,0)
+  //             |/        |/            We compute the Jacobian and iterate until convergence...
+  //  z  y       x---------x
+  //  | /        1         2             Once we have (xi,eta,zeta), we can evaluate all N_i which
+  //  |/                                 give us the weights for the interpolation stencil!
+  //  o---x
+  //
+
+  Teuchos::SerialDenseMatrix<LO, real_type> Jacobian(numDimensions, numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> residual(numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> solutionDirection(numDimensions);
+  Teuchos::SerialDenseVector<LO, real_type> paramCoords(numDimensions);
+  Teuchos::SerialDenseSolver<LO, real_type> problem;
+  int iter = 0, max_iter              = 5;
+  real_type functions[4][8], norm_ref = 1.0, norm2 = 1.0, tol = 1.0e-5;
+  paramCoords.size(numDimensions);
+
+  while ((iter < max_iter) && (norm2 > tol * norm_ref)) {
+    ++iter;
+    norm2 = 0.0;
+    solutionDirection.size(numDimensions);
+    residual.size(numDimensions);
+    Jacobian = 0.0;
+
+    // Compute Jacobian and Residual
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      residual(i) = coord[0][i];  // Add coordinates from point of interest
+      for (LO k = 0; k < numInterpolationPoints; ++k) {
+        residual(i) -= functions[0][k] * coord[k + 1][i];  // Remove contribution from all coarse points
       }
-
-      // Set Jacobian, Vectors and solve problem
-      problem.setMatrix(Teuchos::rcp(&Jacobian, false));
-      problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
-      if(problem.shouldEquilibrate()) {problem.factorWithEquilibration(true);}
-      problem.solve();
-
-      for(LO i = 0; i < numDimensions; ++i) {
-        paramCoords(i) = paramCoords(i) + solutionDirection(i);
+      if (iter == 1) {
+        norm_ref += residual(i) * residual(i);
+        if (i == numDimensions - 1) {
+          norm_ref = std::sqrt(norm_ref);
+        }
       }
 
-      // Recompute Residual norm
-      GetInterpolationFunctions(numDimensions, paramCoords, functions);
-      for(LO i = 0; i < numDimensions; ++i) {
-        real_type tmp = coord[0][i];
-        for(LO k = 0; k < numInterpolationPoints; ++k) {
-          tmp -= functions[0][k]*coord[k+1][i];
+      for (LO j = 0; j < numDimensions; ++j) {
+        for (LO k = 0; k < numInterpolationPoints; ++k) {
+          Jacobian(i, j) += functions[j + 1][k] * coord[k + 1][i];
         }
-        norm2 += tmp*tmp;
-        tmp = 0.0;
       }
-      norm2 = std::sqrt(norm2);
     }
 
-    // Load the interpolation values onto the stencil.
-    for(LO i = 0; i < numInterpolationPoints; ++i) {
-      stencil[i] = functions[0][i];
+    // Set Jacobian, Vectors and solve problem
+    problem.setMatrix(Teuchos::rcp(&Jacobian, false));
+    problem.setVectors(Teuchos::rcp(&solutionDirection, false), Teuchos::rcp(&residual, false));
+    if (problem.shouldEquilibrate()) {
+      problem.factorWithEquilibration(true);
     }
+    problem.solve();
 
-  } // End ComputeLinearInterpolationStencil
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetInterpolationFunctions(const LO numDimensions,
-                            const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
-                            real_type functions[4][8]) const {
-    real_type xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
-    if(numDimensions == 1) {
-      xi = parametricCoordinates[0];
-      denominator = 2.0;
-    } else if(numDimensions == 2) {
-      xi  = parametricCoordinates[0];
-      eta = parametricCoordinates[1];
-      denominator = 4.0;
-    } else if(numDimensions == 3) {
-      xi   = parametricCoordinates[0];
-      eta  = parametricCoordinates[1];
-      zeta = parametricCoordinates[2];
-      denominator = 8.0;
+    for (LO i = 0; i < numDimensions; ++i) {
+      paramCoords(i) = paramCoords(i) + solutionDirection(i);
     }
 
-    functions[0][0] = (1.0 - xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][1] = (1.0 + xi)*(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[0][2] = (1.0 - xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][3] = (1.0 + xi)*(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[0][4] = (1.0 - xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][5] = (1.0 + xi)*(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[0][6] = (1.0 - xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[0][7] = (1.0 + xi)*(1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[1][0] = -(1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][1] =  (1.0 - eta)*(1.0 - zeta) / denominator;
-    functions[1][2] = -(1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][3] =  (1.0 + eta)*(1.0 - zeta) / denominator;
-    functions[1][4] = -(1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][5] =  (1.0 - eta)*(1.0 + zeta) / denominator;
-    functions[1][6] = -(1.0 + eta)*(1.0 + zeta) / denominator;
-    functions[1][7] =  (1.0 + eta)*(1.0 + zeta) / denominator;
-
-    functions[2][0] = -(1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][1] = -(1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][2] =  (1.0 - xi)*(1.0 - zeta) / denominator;
-    functions[2][3] =  (1.0 + xi)*(1.0 - zeta) / denominator;
-    functions[2][4] = -(1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][5] = -(1.0 + xi)*(1.0 + zeta) / denominator;
-    functions[2][6] =  (1.0 - xi)*(1.0 + zeta) / denominator;
-    functions[2][7] =  (1.0 + xi)*(1.0 + zeta) / denominator;
-
-    functions[3][0] = -(1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][1] = -(1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][2] = -(1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][3] = -(1.0 + xi)*(1.0 + eta) / denominator;
-    functions[3][4] =  (1.0 - xi)*(1.0 - eta) / denominator;
-    functions[3][5] =  (1.0 + xi)*(1.0 - eta) / denominator;
-    functions[3][6] =  (1.0 - xi)*(1.0 + eta) / denominator;
-    functions[3][7] =  (1.0 + xi)*(1.0 + eta) / denominator;
-
-  } // End GetInterpolationFunctions
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  coarseCoordinatesBuilderFunctor::coarseCoordinatesBuilderFunctor(RCP<IndexManager_kokkos> geoData,
-                                                                   coord_view_type fineCoordView,
-                                                                   coord_view_type coarseCoordView)
-    : geoData_(*geoData), fineCoordView_(fineCoordView), coarseCoordView_(coarseCoordView) {
-
-  } // coarseCoordinatesBuilderFunctor()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  KOKKOS_INLINE_FUNCTION
-  void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  coarseCoordinatesBuilderFunctor::operator() (const LO coarseNodeIdx) const {
-
-    LO fineNodeIdx;
-    LO nodeCoarseTuple[3] = {0, 0, 0};
-    LO nodeFineTuple[3]   = {0, 0, 0};
-    auto coarseningRate    = geoData_.getCoarseningRates();
-    auto fineNodesPerDir   = geoData_.getLocalFineNodesPerDir();
-    auto coarseNodesPerDir = geoData_.getCoarseNodesPerDir();
-    geoData_.getCoarseLID2CoarseTuple(coarseNodeIdx, nodeCoarseTuple);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(nodeCoarseTuple[dim] == coarseNodesPerDir(dim) - 1) {
-        nodeFineTuple[dim] = fineNodesPerDir(dim) - 1;
-      } else {
-        nodeFineTuple[dim] = nodeCoarseTuple[dim]*coarseningRate(dim);
+    // Recompute Residual norm
+    GetInterpolationFunctions(numDimensions, paramCoords, functions);
+    for (LO i = 0; i < numDimensions; ++i) {
+      real_type tmp = coord[0][i];
+      for (LO k = 0; k < numInterpolationPoints; ++k) {
+        tmp -= functions[0][k] * coord[k + 1][i];
       }
+      norm2 += tmp * tmp;
+      tmp = 0.0;
     }
+    norm2 = std::sqrt(norm2);
+  }
+
+  // Load the interpolation values onto the stencil.
+  for (LO i = 0; i < numInterpolationPoints; ++i) {
+    stencil[i] = functions[0][i];
+  }
 
-    fineNodeIdx = nodeFineTuple[2]*fineNodesPerDir(1)*fineNodesPerDir(0)
-      + nodeFineTuple[1]*fineNodesPerDir(0) + nodeFineTuple[0];
+}  // End ComputeLinearInterpolationStencil
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetInterpolationFunctions(const LO numDimensions,
+                              const Teuchos::SerialDenseVector<LO, real_type> parametricCoordinates,
+                              real_type functions[4][8]) const {
+  real_type xi = 0.0, eta = 0.0, zeta = 0.0, denominator = 0.0;
+  if (numDimensions == 1) {
+    xi          = parametricCoordinates[0];
+    denominator = 2.0;
+  } else if (numDimensions == 2) {
+    xi          = parametricCoordinates[0];
+    eta         = parametricCoordinates[1];
+    denominator = 4.0;
+  } else if (numDimensions == 3) {
+    xi          = parametricCoordinates[0];
+    eta         = parametricCoordinates[1];
+    zeta        = parametricCoordinates[2];
+    denominator = 8.0;
+  }
 
-    for(int dim = 0; dim < fineCoordView_.extent_int(1); ++dim) {
-      coarseCoordView_(coarseNodeIdx, dim) = fineCoordView_(fineNodeIdx, dim);
+  functions[0][0] = (1.0 - xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][1] = (1.0 + xi) * (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[0][2] = (1.0 - xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][3] = (1.0 + xi) * (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[0][4] = (1.0 - xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][5] = (1.0 + xi) * (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[0][6] = (1.0 - xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[0][7] = (1.0 + xi) * (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[1][0] = -(1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][1] = (1.0 - eta) * (1.0 - zeta) / denominator;
+  functions[1][2] = -(1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][3] = (1.0 + eta) * (1.0 - zeta) / denominator;
+  functions[1][4] = -(1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][5] = (1.0 - eta) * (1.0 + zeta) / denominator;
+  functions[1][6] = -(1.0 + eta) * (1.0 + zeta) / denominator;
+  functions[1][7] = (1.0 + eta) * (1.0 + zeta) / denominator;
+
+  functions[2][0] = -(1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][1] = -(1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][2] = (1.0 - xi) * (1.0 - zeta) / denominator;
+  functions[2][3] = (1.0 + xi) * (1.0 - zeta) / denominator;
+  functions[2][4] = -(1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][5] = -(1.0 + xi) * (1.0 + zeta) / denominator;
+  functions[2][6] = (1.0 - xi) * (1.0 + zeta) / denominator;
+  functions[2][7] = (1.0 + xi) * (1.0 + zeta) / denominator;
+
+  functions[3][0] = -(1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][1] = -(1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][2] = -(1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][3] = -(1.0 + xi) * (1.0 + eta) / denominator;
+  functions[3][4] = (1.0 - xi) * (1.0 - eta) / denominator;
+  functions[3][5] = (1.0 + xi) * (1.0 - eta) / denominator;
+  functions[3][6] = (1.0 - xi) * (1.0 + eta) / denominator;
+  functions[3][7] = (1.0 + xi) * (1.0 + eta) / denominator;
+
+}  // End GetInterpolationFunctions
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    coarseCoordinatesBuilderFunctor::coarseCoordinatesBuilderFunctor(RCP<IndexManager_kokkos> geoData,
+                                                                     coord_view_type fineCoordView,
+                                                                     coord_view_type coarseCoordView)
+  : geoData_(*geoData)
+  , fineCoordView_(fineCoordView)
+  , coarseCoordView_(coarseCoordView) {
+}  // coarseCoordinatesBuilderFunctor()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+KOKKOS_INLINE_FUNCTION void GeometricInterpolationPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    coarseCoordinatesBuilderFunctor::operator()(const LO coarseNodeIdx) const {
+  LO fineNodeIdx;
+  LO nodeCoarseTuple[3]  = {0, 0, 0};
+  LO nodeFineTuple[3]    = {0, 0, 0};
+  auto coarseningRate    = geoData_.getCoarseningRates();
+  auto fineNodesPerDir   = geoData_.getLocalFineNodesPerDir();
+  auto coarseNodesPerDir = geoData_.getCoarseNodesPerDir();
+  geoData_.getCoarseLID2CoarseTuple(coarseNodeIdx, nodeCoarseTuple);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (nodeCoarseTuple[dim] == coarseNodesPerDir(dim) - 1) {
+      nodeFineTuple[dim] = fineNodesPerDir(dim) - 1;
+    } else {
+      nodeFineTuple[dim] = nodeCoarseTuple[dim] * coarseningRate(dim);
     }
   }
 
-} // namespace MueLu
+  fineNodeIdx = nodeFineTuple[2] * fineNodesPerDir(1) * fineNodesPerDir(0) + nodeFineTuple[1] * fineNodesPerDir(0) + nodeFineTuple[0];
+
+  for (int dim = 0; dim < fineCoordView_.extent_int(1); ++dim) {
+    coarseCoordView_(coarseNodeIdx, dim) = fineCoordView_(fineNodeIdx, dim);
+  }
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_DEF_HPP
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_decl.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_decl.hpp
index e811884d4b3c..a6f3be5ec5ca 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_decl.hpp
@@ -54,59 +54,59 @@
 
 namespace MueLu {
 
-  /*!
-    @class RegionRFactory class
-    @brief Factory that builds a restriction operator for region multigrid
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RegionRFactory : public TwoLevelFactoryBase {
+/*!
+  @class RegionRFactory class
+  @brief Factory that builds a restriction operator for region multigrid
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RegionRFactory : public TwoLevelFactoryBase {
 #undef MUELU_REGIONRFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using realvaluedmultivector_type = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+ public:
+  using real_type                  = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using realvaluedmultivector_type = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Default Constructor
-    RegionRFactory() = default;
+  //! Default Constructor
+  RegionRFactory() = default;
 
-    //!Destructor
-    virtual ~RegionRFactory() = default;
-    //@}
+  //! Destructor
+  virtual ~RegionRFactory() = default;
+  //@}
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build3D(const int numDimensions,
-                 Array<LO>& lFineNodesPerDim,
-                 const RCP<Matrix>& A,
-                 const RCP<realvaluedmultivector_type>& fineCoordinates,
-                 RCP<Matrix>& R,
-                 RCP<realvaluedmultivector_type>& coarseCoordinates,
-                 Array<LO>& lCoarseNodesPerDim) const;
+  void Build3D(const int numDimensions,
+               Array<LO>& lFineNodesPerDim,
+               const RCP<Matrix>& A,
+               const RCP<realvaluedmultivector_type>& fineCoordinates,
+               RCP<Matrix>& R,
+               RCP<realvaluedmultivector_type>& coarseCoordinates,
+               Array<LO>& lCoarseNodesPerDim) const;
 
-    //@}
+  //@}
 
-  }; // class RegionRFactory
+};  // class RegionRFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REGIONRFACTORY_SHORT
-#endif // MUELU_REGIONRFACTORY_DECL_HPP
+#endif  // MUELU_REGIONRFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_def.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_def.hpp
index 16318f14673a..ee892f421897 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_def.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_def.hpp
@@ -56,1090 +56,999 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase> >("A",                            Teuchos::null,
-                                                 "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",                Teuchos::null,
-                                                 "Number of spatial dimensions in the problem.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",                 Teuchos::null,
-                                                 "Number of local nodes per spatial dimension on the fine grid.");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",                    Teuchos::null,
-                                                 "Fine level nullspace used to construct the coarse level nullspace.");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null,
-                                                 "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
-    validParamList->set<bool>                   ("keep coarse coords",           false, "Flag to keep coordinates for special coarse grid solve");
-
-    return validParamList;
-  } // GetValidParameterList()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "numDimensions");
-    Input(fineLevel, "lNodesPerDim");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "Coordinates");
-
-  } // DeclareInput()
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level& fineLevel, Level& coarseLevel) const {
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null,
+                                               "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spatial dimensions in the problem.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Number of local nodes per spatial dimension on the fine grid.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null,
+                                               "Fine level nullspace used to construct the coarse level nullspace.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null,
+                                               "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
+  validParamList->set<bool>("keep coarse coords", false, "Flag to keep coordinates for special coarse grid solve");
+
+  return validParamList;
+}  // GetValidParameterList()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "numDimensions");
+  Input(fineLevel, "lNodesPerDim");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "Coordinates");
+
+}  // DeclareInput()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& fineLevel, Level& coarseLevel) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  *out << "Starting RegionRFactory::Build." << std::endl;
+
+  // First get the inputs from the fineLevel
+  const int numDimensions = Get<int>(fineLevel, "numDimensions");
+  Array<LO> lFineNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
+  {
+    Array<LO> lNodesPerDim = Get<Array<LO> >(fineLevel, "lNodesPerDim");
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      lFineNodesPerDim[dim] = lNodesPerDim[dim];
     }
-
-    *out << "Starting RegionRFactory::Build." << std::endl;
-
-    // First get the inputs from the fineLevel
-    const int numDimensions = Get<int>(fineLevel, "numDimensions");
-    Array<LO> lFineNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
-    {
-      Array<LO> lNodesPerDim = Get<Array<LO> >(fineLevel, "lNodesPerDim");
-      for(int dim = 0; dim < numDimensions; ++dim) {
-        lFineNodesPerDim[dim] = lNodesPerDim[dim];
+  }
+  *out << "numDimensions " << numDimensions << " and lFineNodesPerDim: " << lFineNodesPerDim
+       << std::endl;
+
+  // Let us check that the inputs verify our assumptions
+  if (numDimensions < 1 || numDimensions > 3) {
+    throw std::runtime_error("numDimensions must be 1, 2 or 3!");
+  }
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    if (lFineNodesPerDim[dim] % 3 != 1) {
+      throw std::runtime_error("The number of fine node in each direction need to be 3n+1");
+    }
+  }
+  Array<LO> lCoarseNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
+
+  const RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
+
+  RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
+  fineCoordinates = Get<RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
+  if (static_cast<int>(fineCoordinates->getNumVectors()) != numDimensions) {
+    throw std::runtime_error("The number of vectors in the coordinates is not equal to numDimensions!");
+  }
+
+  // Let us create R and pass it down to the
+  // appropriate specialization and see what we
+  // get back!
+  RCP<Matrix> R;
+
+  if (numDimensions == 1) {
+    throw std::runtime_error("RegionRFactory no implemented for 1D case yet.");
+  } else if (numDimensions == 2) {
+    throw std::runtime_error("RegionRFactory no implemented for 2D case yet.");
+  } else if (numDimensions == 3) {
+    Build3D(numDimensions, lFineNodesPerDim, A, fineCoordinates,
+            R, coarseCoordinates, lCoarseNodesPerDim);
+  }
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  // Reuse pattern if available (multiple solve)
+  RCP<ParameterList> Tparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    Tparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    Tparams = rcp(new ParameterList);
+
+  // R->describe(*out, Teuchos::VERB_EXTREME);
+  *out << "Compute P=R^t" << std::endl;
+  // By default, we don't need global constants for transpose
+  Tparams->set("compute global constants: temporaries", Tparams->get("compute global constants: temporaries", false));
+  Tparams->set("compute global constants", Tparams->get("compute global constants", false));
+  std::string label = "MueLu::RegionR-transR" + Teuchos::toString(coarseLevel.GetLevelID());
+  RCP<Matrix> P     = Utilities::Transpose(*R, true, label, Tparams);
+
+  *out << "Compute coarse nullspace" << std::endl;
+  RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(R->getRowMap(),
+                                                               fineNullspace->getNumVectors());
+  R->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+           Teuchos::ScalarTraits<SC>::zero());
+
+  *out << "Set data on coarse level" << std::endl;
+  Set(coarseLevel, "numDimensions", numDimensions);
+  Set(coarseLevel, "lNodesPerDim", lCoarseNodesPerDim);
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "Coordinates", coarseCoordinates);
+  if (pL.get<bool>("keep coarse coords")) {
+    coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
+  }
+
+  R->SetFixedBlockSize(A->GetFixedBlockSize());
+  P->SetFixedBlockSize(A->GetFixedBlockSize());
+
+  Set(coarseLevel, "R", R);
+  Set(coarseLevel, "P", P);
+
+}  // Build()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build3D(const int numDimensions,
+            Teuchos::Array<LocalOrdinal>& lFineNodesPerDim,
+            const RCP<Matrix>& A,
+            const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& fineCoordinates,
+            RCP<Matrix>& R,
+            RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coarseCoordinates,
+            Teuchos::Array<LocalOrdinal>& lCoarseNodesPerDim) const {
+  using local_matrix_type = typename CrsMatrix::local_matrix_type;
+  using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
+  using row_map_type      = typename local_matrix_type::row_map_type::non_const_type;
+  using entries_type      = typename local_matrix_type::index_type::non_const_type;
+  using values_type       = typename local_matrix_type::values_type::non_const_type;
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  // Now compute number of coarse grid points
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    lCoarseNodesPerDim[dim] = lFineNodesPerDim[dim] / 3 + 1;
+  }
+  *out << "lCoarseNodesPerDim " << lCoarseNodesPerDim << std::endl;
+
+  // Grab the block size here and multiply all existing offsets by it
+  const LO blkSize = A->GetFixedBlockSize();
+  *out << "blkSize " << blkSize << std::endl;
+
+  // Based on lCoarseNodesPerDim and lFineNodesPerDim
+  // we can compute numRows, numCols and NNZ for R
+  const LO numRows = blkSize * lCoarseNodesPerDim[0] * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[2];
+  const LO numCols = blkSize * lFineNodesPerDim[0] * lFineNodesPerDim[1] * lFineNodesPerDim[2];
+
+  // Create the coarse coordinates multivector
+  // so we can fill it on the fly while computing
+  // the restriction operator
+  RCP<Map> rowMap = MapFactory::Build(A->getRowMap()->lib(),
+                                      Teuchos::OrdinalTraits<GO>::invalid(),
+                                      numRows,
+                                      A->getRowMap()->getIndexBase(),
+                                      A->getRowMap()->getComm());
+
+  RCP<Map> coordRowMap = MapFactory::Build(A->getRowMap()->lib(),
+                                           Teuchos::OrdinalTraits<GO>::invalid(),
+                                           lCoarseNodesPerDim[0] * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[2],
+                                           A->getRowMap()->getIndexBase(),
+                                           A->getRowMap()->getComm());
+
+  coarseCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(coordRowMap,
+                                                                               numDimensions);
+  Array<ArrayRCP<const real_type> > fineCoordData(numDimensions);
+  Array<ArrayRCP<real_type> > coarseCoordData(numDimensions);
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    fineCoordData[dim]   = fineCoordinates->getData(dim);
+    coarseCoordData[dim] = coarseCoordinates->getDataNonConst(dim);
+  }
+
+  // Let us set some parameter that will be useful
+  // while constructing R
+
+  // Length of interpolation stencils based on geometry
+  const LO cornerStencilLength   = 27;
+  const LO edgeStencilLength     = 45;
+  const LO faceStencilLength     = 75;
+  const LO interiorStencilLength = 125;
+
+  // Number of corner, edge, face and interior nodes
+  const LO numCorners   = 8;
+  const LO numEdges     = 4 * (lCoarseNodesPerDim[0] - 2) + 4 * (lCoarseNodesPerDim[1] - 2) + 4 * (lCoarseNodesPerDim[2] - 2);
+  const LO numFaces     = 2 * (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[1] - 2) + 2 * (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[2] - 2) + 2 * (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[2] - 2);
+  const LO numInteriors = (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[2] - 2);
+
+  const LO nnz = (numCorners * cornerStencilLength + numEdges * edgeStencilLength + numFaces * faceStencilLength + numInteriors * interiorStencilLength) * blkSize;
+
+  // Having the number of rows and columns we can genrate
+  // the appropriate maps for our operator.
+
+  *out << "R statistics:" << std::endl
+       << "  -numRows= " << numRows << std::endl
+       << "  -numCols= " << numCols << std::endl
+       << "  -nnz=     " << nnz << std::endl;
+
+  row_map_type row_map(Kokkos::ViewAllocateWithoutInitializing("row_map"), numRows + 1);
+  typename row_map_type::HostMirror row_map_h = Kokkos::create_mirror_view(row_map);
+
+  entries_type entries(Kokkos::ViewAllocateWithoutInitializing("entries"), nnz);
+  typename entries_type::HostMirror entries_h = Kokkos::create_mirror_view(entries);
+
+  values_type values(Kokkos::ViewAllocateWithoutInitializing("values"), nnz);
+  typename values_type::HostMirror values_h = Kokkos::create_mirror_view(values);
+
+  // Compute the basic interpolation
+  // coefficients for 1D rate of 3
+  // coarsening.
+  Array<SC> coeffs({1.0 / 3.0, 2.0 / 3.0, 1.0, 2.0 / 3.0, 1.0 / 3.0});
+  row_map_h(0) = 0;
+
+  // Define some offsets that
+  // will be needed often later on
+  const LO edgeLineOffset     = 2 * cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength;
+  const LO faceLineOffset     = 2 * edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength;
+  const LO interiorLineOffset = 2 * faceStencilLength + (lCoarseNodesPerDim[0] - 2) * interiorStencilLength;
+
+  const LO facePlaneOffset     = 2 * edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset;
+  const LO interiorPlaneOffset = 2 * faceLineOffset + (lCoarseNodesPerDim[1] - 2) * interiorLineOffset;
+
+  // Let us take care of the corners
+  // first since we always have
+  // corners to deal with!
+  {
+    // Corner 1
+    LO coordRowIdx = 0, rowIdx = 0, coordColumnOffset = 0, columnOffset = 0, entryOffset = 0;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i + 2];
+          }
+        }
       }
     }
-    *out << "numDimensions " << numDimensions << " and lFineNodesPerDim: " << lFineNodesPerDim
-         << std::endl;
-
-    // Let us check that the inputs verify our assumptions
-    if(numDimensions < 1 || numDimensions > 3) {
-      throw std::runtime_error("numDimensions must be 1, 2 or 3!");
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      if(lFineNodesPerDim[dim] % 3 != 1) {
-        throw std::runtime_error("The number of fine node in each direction need to be 3n+1");
-      }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
-    Array<LO> lCoarseNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
-
-    const RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
 
-    RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
-    fineCoordinates = Get< RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
-    if(static_cast<int>(fineCoordinates->getNumVectors()) != numDimensions) {
-      throw std::runtime_error("The number of vectors in the coordinates is not equal to numDimensions!");
+    // Corner 5
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i + 2];
+          }
+        }
+      }
     }
-
-    // Let us create R and pass it down to the
-    // appropriate specialization and see what we
-    // get back!
-    RCP<Matrix> R;
-
-    if(numDimensions == 1) {
-      throw std::runtime_error("RegionRFactory no implemented for 1D case yet.");
-    } else if(numDimensions == 2) {
-      throw std::runtime_error("RegionRFactory no implemented for 2D case yet.");
-    } else if(numDimensions == 3) {
-      Build3D(numDimensions, lFineNodesPerDim, A, fineCoordinates,
-              R, coarseCoordinates, lCoarseNodesPerDim);
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-
-    // Reuse pattern if available (multiple solve)
-    RCP<ParameterList> Tparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      Tparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-    else
-      Tparams= rcp(new ParameterList);
-
-    // R->describe(*out, Teuchos::VERB_EXTREME);
-    *out << "Compute P=R^t" << std::endl;
-    // By default, we don't need global constants for transpose
-    Tparams->set("compute global constants: temporaries",Tparams->get("compute global constants: temporaries", false));
-    Tparams->set("compute global constants", Tparams->get("compute global constants",false));
-    std::string label = "MueLu::RegionR-transR" + Teuchos::toString(coarseLevel.GetLevelID());
-    RCP<Matrix> P = Utilities::Transpose(*R, true, label, Tparams);
-
-    *out << "Compute coarse nullspace" << std::endl;
-    RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(R->getRowMap(),
-                                                                 fineNullspace->getNumVectors());
-    R->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-             Teuchos::ScalarTraits<SC>::zero());
-
-    *out << "Set data on coarse level" << std::endl;
-    Set(coarseLevel, "numDimensions", numDimensions);
-    Set(coarseLevel, "lNodesPerDim",  lCoarseNodesPerDim);
-    Set(coarseLevel, "Nullspace",     coarseNullspace);
-    Set(coarseLevel, "Coordinates",   coarseCoordinates);
-    if(pL.get<bool>("keep coarse coords")) {
-      coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
 
-    R->SetFixedBlockSize(A->GetFixedBlockSize());
-    P->SetFixedBlockSize(A->GetFixedBlockSize());
-
-    Set(coarseLevel, "R", R);
-    Set(coarseLevel, "P", P);
-
-  } // Build()
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build3D(const int numDimensions,
-          Teuchos::Array<LocalOrdinal>& lFineNodesPerDim,
-          const RCP<Matrix>& A,
-          const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& fineCoordinates,
-          RCP<Matrix>& R,
-          RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coarseCoordinates,
-          Teuchos::Array<LocalOrdinal>& lCoarseNodesPerDim) const {
-    using local_matrix_type = typename CrsMatrix::local_matrix_type;
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using row_map_type      = typename local_matrix_type::row_map_type::non_const_type;
-    using entries_type      = typename local_matrix_type::index_type::non_const_type;
-    using values_type       = typename local_matrix_type::values_type::non_const_type;
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+    // Corner 2
+    coordRowIdx       = (lCoarseNodesPerDim[0] - 1);
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[0] - 1);
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i];
+          }
+        }
+      }
     }
-
-
-    // Now compute number of coarse grid points
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      lCoarseNodesPerDim[dim] = lFineNodesPerDim[dim] / 3 + 1;
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-    *out << "lCoarseNodesPerDim " << lCoarseNodesPerDim << std::endl;
-
-    // Grab the block size here and multiply all existing offsets by it
-    const LO blkSize = A->GetFixedBlockSize();
-    *out << "blkSize " << blkSize << std::endl;
-
-    // Based on lCoarseNodesPerDim and lFineNodesPerDim
-    // we can compute numRows, numCols and NNZ for R
-    const LO numRows = blkSize*lCoarseNodesPerDim[0]*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[2];
-    const LO numCols = blkSize*lFineNodesPerDim[0]*lFineNodesPerDim[1]*lFineNodesPerDim[2];
-
-    // Create the coarse coordinates multivector
-    // so we can fill it on the fly while computing
-    // the restriction operator
-    RCP<Map> rowMap = MapFactory::Build(A->getRowMap()->lib(),
-                                        Teuchos::OrdinalTraits<GO>::invalid(),
-                                        numRows,
-                                        A->getRowMap()->getIndexBase(),
-                                        A->getRowMap()->getComm());
-
-    RCP<Map> coordRowMap = MapFactory::Build(A->getRowMap()->lib(),
-                                        Teuchos::OrdinalTraits<GO>::invalid(),
-                                        lCoarseNodesPerDim[0]*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[2],
-                                        A->getRowMap()->getIndexBase(),
-                                        A->getRowMap()->getComm());
-
-    coarseCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(coordRowMap,
-                                                                                 numDimensions);
-    Array<ArrayRCP<const real_type> > fineCoordData(numDimensions);
-    Array<ArrayRCP<real_type> > coarseCoordData(numDimensions);
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      fineCoordData[dim] = fineCoordinates->getData(dim);
-      coarseCoordData[dim] = coarseCoordinates->getDataNonConst(dim);
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
 
-    // Let us set some parameter that will be useful
-    // while constructing R
-
-    // Length of interpolation stencils based on geometry
-    const LO cornerStencilLength   = 27;
-    const LO edgeStencilLength     = 45;
-    const LO faceStencilLength     = 75;
-    const LO interiorStencilLength = 125;
-
-    // Number of corner, edge, face and interior nodes
-    const LO numCorners   = 8;
-    const LO numEdges     = 4*(lCoarseNodesPerDim[0] - 2)
-      + 4*(lCoarseNodesPerDim[1] - 2)
-      + 4*(lCoarseNodesPerDim[2] - 2);
-    const LO numFaces     = 2*(lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[1] - 2)
-      + 2*(lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[2] - 2)
-      + 2*(lCoarseNodesPerDim[1] - 2)*(lCoarseNodesPerDim[2] - 2);
-    const LO numInteriors = (lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[1] - 2)
-      *(lCoarseNodesPerDim[2] - 2);
-
-    const LO nnz = (numCorners*cornerStencilLength + numEdges*edgeStencilLength
-		    + numFaces*faceStencilLength + numInteriors*interiorStencilLength)*blkSize;
-
-    // Having the number of rows and columns we can genrate
-    // the appropriate maps for our operator.
-
-    *out << "R statistics:" << std::endl
-         << "  -numRows= " << numRows << std::endl
-         << "  -numCols= " << numCols << std::endl
-         << "  -nnz=     " << nnz << std::endl;
-
-    row_map_type row_map(Kokkos::ViewAllocateWithoutInitializing("row_map"), numRows + 1);
-    typename row_map_type::HostMirror row_map_h = Kokkos::create_mirror_view(row_map);
-
-    entries_type entries(Kokkos::ViewAllocateWithoutInitializing("entries"), nnz);
-    typename entries_type::HostMirror entries_h = Kokkos::create_mirror_view(entries);
-
-    values_type values(Kokkos::ViewAllocateWithoutInitializing("values"), nnz);
-    typename values_type::HostMirror values_h = Kokkos::create_mirror_view(values);
-
-    // Compute the basic interpolation
-    // coefficients for 1D rate of 3
-    // coarsening.
-    Array<SC> coeffs({1.0/3.0, 2.0/3.0, 1.0, 2.0/3.0, 1.0/3.0});
-    row_map_h(0) = 0;
-
-    // Define some offsets that
-    // will be needed often later on
-    const LO edgeLineOffset = 2*cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength;
-    const LO faceLineOffset = 2*edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength;
-    const LO interiorLineOffset = 2*faceStencilLength
-      + (lCoarseNodesPerDim[0] - 2)*interiorStencilLength;
-
-    const LO facePlaneOffset = 2*edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset;
-    const LO interiorPlaneOffset = 2*faceLineOffset + (lCoarseNodesPerDim[1] - 2)*interiorLineOffset;
-
-    // Let us take care of the corners
-    // first since we always have
-    // corners to deal with!
-    {
-      // Corner 1
-      LO coordRowIdx = 0, rowIdx = 0, coordColumnOffset = 0, columnOffset = 0, entryOffset = 0;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i + 2];
-	    }
-	  }
+    // Corner 6
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
+          }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
-      }
-      for(int dim = 0; dim <numDimensions; ++dim) {
-        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-      }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
 
-      // Corner 5
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i + 2];
-	    }
+    // Corner 3
+    coordRowIdx       = (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0];
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0];
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i + 2];
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+
+    // Corner 7
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
+          }
+        }
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
-        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+
+    // Corner 4
+    coordRowIdx       = (lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] - 1);
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[1] * lFineNodesPerDim[0] - 1);
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset +
+                   cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength) *
+                  blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
+          }
+        }
       }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
 
-      // Corner 2
-      coordRowIdx = (lCoarseNodesPerDim[0] - 1);
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[0] - 1);
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		   + j*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i];
+    // Corner 8
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
+          }
+        }
+      }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+  }  // Corners are done!
+
+  // Edges along 0 direction
+  if (lCoarseNodesPerDim[0] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[0] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3;
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 6
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
+      // Edge 1
+      coordRowIdx       = ((lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 3
-      coordRowIdx = (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		   + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i + 2];
+      // Edge 2
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 7
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
+      // Edge 3
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-	row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
-
-      // Corner 4
-      coordRowIdx = (lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0] - 1);
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[1]*lFineNodesPerDim[0] - 1);
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset +
-		     cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		   + (j - 2)*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
+    }
+  }
+
+  // Edges along 1 direction
+  if (lCoarseNodesPerDim[1] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[1] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1) * lCoarseNodesPerDim[0];
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3 * lFineNodesPerDim[0];
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + edgeIdx * faceLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-	row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 8
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
+      // Edge 1
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + edgeIdx * faceLineOffset + edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
-    } // Corners are done!
-
-    // Edges along 0 direction
-    if(lCoarseNodesPerDim[0] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[0] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3;
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i];
-              }
-            }
-          }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 1
-        coordRowIdx = ((lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0] + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0] + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset
-			+ cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
 
-        // Edge 2
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-			+ cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
-              }
+      // Edge 2
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (edgeIdx + 1) * lCoarseNodesPerDim[0]);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3 * lFineNodesPerDim[0]);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + edgeIdx * faceLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
             }
           }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
         }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 3
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0] + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0] + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  =  (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-			 + edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset
-			 + cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
+      // Edge 3
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (edgeIdx + 1) * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3 * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + edgeIdx * faceLineOffset + edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
         }
       }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
     }
-
-    // Edges along 1 direction
-    if(lCoarseNodesPerDim[1] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[1] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1)*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + edgeIdx*faceLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i + 2];
-              }
+  }
+
+  // Edges along 2 direction
+  if (lCoarseNodesPerDim[2] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[2] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + edgeIdx * interiorPlaneOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i + 2];
             }
           }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
         }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 1
-	coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + edgeIdx*faceLineOffset
-			+ edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
-              }
+      // Edge 1
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + faceLineOffset - edgeStencilLength + edgeIdx * interiorPlaneOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
             }
           }
-	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
         }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 2
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + (edgeIdx + 1)*lCoarseNodesPerDim[0]);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (edgeIdx + 1)*3*lFineNodesPerDim[0]);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-			+ edgeLineOffset + edgeIdx*faceLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
-              }
+      // Edge 2
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0]);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0]);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + edgeIdx * interiorPlaneOffset + faceLineOffset + (lCoarseNodesPerDim[1] - 2) * interiorLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 3
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + (edgeIdx + 1)*lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (edgeIdx + 1)*3*lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-			+ edgeLineOffset + edgeIdx*faceLineOffset
-			+ edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
+      // Edge 3
+      coordRowIdx       = ((edgeIdx + 2) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + lFineNodesPerDim[1] * lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (edgeIdx + 1) * interiorPlaneOffset - edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
     }
-
-    // Edges along 2 direction
-    if(lCoarseNodesPerDim[2] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[2] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + edgeIdx*interiorPlaneOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-		  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i + 2];
-              }
+  }
+
+  // Faces in 0-1 plane
+  if ((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[1] - 2 > 0)) {
+    Array<LO> gridIdx(3);
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO faceIdx = 0; faceIdx < (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[0] - 2); ++faceIdx) {
+      // Face 0
+      coordRowIdx       = ((gridIdx[1] + 1) * lCoarseNodesPerDim[0] + gridIdx[0] + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = 3 * ((gridIdx[1] + 1) * lFineNodesPerDim[0] + gridIdx[0] + 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + edgeStencilLength + gridIdx[1] * faceLineOffset + gridIdx[0] * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 1
-        coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + faceLineOffset - edgeStencilLength
-                        + edgeIdx*interiorPlaneOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 2
-        coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0]);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                             + (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0]);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + edgeIdx*interiorPlaneOffset + faceLineOffset
-			+ (lCoarseNodesPerDim[1] - 2)*interiorLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
-              }
+      // Face 1
+      coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+      rowIdx = coordRowIdx * blkSize;
+      coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+      columnOffset = coordColumnOffset * blkSize;
+      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 3
-        coordRowIdx = ((edgeIdx + 2)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + lFineNodesPerDim[1]*lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (edgeIdx + 1)*interiorPlaneOffset - edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      // Last step in the loop
+      // update the grid indices
+      // for next grid point
+      ++gridIdx[0];
+      if (gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
+        gridIdx[0] = 0;
+        ++gridIdx[1];
       }
     }
-
-    // Faces in 0-1 plane
-    if((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[1] - 2 > 0)) {
-
-      Array<LO> gridIdx(3);
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO faceIdx=0; faceIdx < (lCoarseNodesPerDim[1]-2)*(lCoarseNodesPerDim[0]-2); ++faceIdx) {
-
-        // Face 0
-        coordRowIdx = ((gridIdx[1] + 1)*lCoarseNodesPerDim[0] + gridIdx[0] + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = 3*((gridIdx[1] + 1)*lFineNodesPerDim[0] + gridIdx[0] + 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + edgeStencilLength
-			+ gridIdx[1]*faceLineOffset + gridIdx[0]*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*25 + j*5 + i + faceStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*25 + j*5 + i + faceStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
-              }
+  }
+
+  // Faces in 0-2 plane
+  if ((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
+    Array<LO> gridIdx(3);
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO faceIdx = 0; faceIdx < (lCoarseNodesPerDim[2] - 2) * (lCoarseNodesPerDim[0] - 2); ++faceIdx) {
+      // Face 0
+      coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (gridIdx[0] + 1));
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((gridIdx[2] + 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + gridIdx[0] + 1) * 3;
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + gridIdx[2] * interiorPlaneOffset + edgeStencilLength + gridIdx[0] * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Face 1
-        coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*25 + j*5 + i + faceStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*25 + j*5 + i + faceStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
+      // Face 1
+      coordRowIdx += (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0];
+      rowIdx = coordRowIdx * blkSize;
+      coordColumnOffset += (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0];
+      columnOffset = coordColumnOffset * blkSize;
+      entryOffset += (faceLineOffset + (lCoarseNodesPerDim[1] - 2) * interiorLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Last step in the loop
-        // update the grid indices
-        // for next grid point
-        ++gridIdx[0];
-        if(gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
-          gridIdx[0] = 0;
-          ++gridIdx[1];
-        }
+      // Last step in the loop
+      // update the grid indices
+      // for next grid point
+      ++gridIdx[0];
+      if (gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
+        gridIdx[0] = 0;
+        ++gridIdx[2];
       }
     }
-
-    // Faces in 0-2 plane
-    if((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
-
-      Array<LO> gridIdx(3);
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO faceIdx=0; faceIdx < (lCoarseNodesPerDim[2]-2)*(lCoarseNodesPerDim[0]-2); ++faceIdx) {
-
-        // Face 0
-        coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0] + (gridIdx[0] + 1));
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((gridIdx[2] + 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                             + gridIdx[0] + 1)*3;
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + gridIdx[2]*interiorPlaneOffset + edgeStencilLength
-			+ gridIdx[0]*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + faceStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + faceStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
-              }
+  }
+
+  // Faces in 1-2 plane
+  if ((lCoarseNodesPerDim[1] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
+    Array<LO> gridIdx(3);
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO faceIdx = 0; faceIdx < (lCoarseNodesPerDim[2] - 2) * (lCoarseNodesPerDim[1] - 2); ++faceIdx) {
+      // Face 0
+      coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (gridIdx[1] + 1) * lCoarseNodesPerDim[0]);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((gridIdx[2] + 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (gridIdx[1] + 1) * lFineNodesPerDim[0]) * 3;
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + gridIdx[2] * interiorPlaneOffset + faceLineOffset + gridIdx[1] * interiorLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Face 1
-        coordRowIdx += (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset += (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  += (faceLineOffset + (lCoarseNodesPerDim[1] - 2)*interiorLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + faceStencilLength*l) = columnOffset
-		  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + faceStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
+      // Face 1
+      coordRowIdx += (lCoarseNodesPerDim[0] - 1);
+      rowIdx = coordRowIdx * blkSize;
+      coordColumnOffset += (lFineNodesPerDim[0] - 1);
+      columnOffset = coordColumnOffset * blkSize;
+      entryOffset += (faceStencilLength + (lCoarseNodesPerDim[0] - 2) * interiorStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-      	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Last step in the loop
-        // update the grid indices
-        // for next grid point
-        ++gridIdx[0];
-        if(gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
-          gridIdx[0] = 0;
-          ++gridIdx[2];
-        }
+      // Last step in the loop
+      // update the grid indices
+      // for next grid point
+      ++gridIdx[1];
+      if (gridIdx[1] == lCoarseNodesPerDim[1] - 2) {
+        gridIdx[1] = 0;
+        ++gridIdx[2];
       }
     }
-
-    // Faces in 1-2 plane
-    if((lCoarseNodesPerDim[1] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
-
-      Array<LO> gridIdx(3);
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO faceIdx=0; faceIdx < (lCoarseNodesPerDim[2]-2)*(lCoarseNodesPerDim[1]-2); ++faceIdx) {
-
-        // Face 0
-        coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-		       + (gridIdx[1] + 1)*lCoarseNodesPerDim[0]);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((gridIdx[2] + 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (gridIdx[1] + 1)*lFineNodesPerDim[0])*3;
-      	columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + gridIdx[2]*interiorPlaneOffset + faceLineOffset
-			+ gridIdx[1]*interiorLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + faceStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + faceStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Face 1
-        coordRowIdx += (lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset += (lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  += (faceStencilLength + (lCoarseNodesPerDim[0] - 2)*interiorStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + faceStencilLength*l) = columnOffset
-		  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-		     + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + faceStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-      	}
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-      	}
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Last step in the loop
-        // update the grid indices
-        // for next grid point
-        ++gridIdx[1];
-        if(gridIdx[1] == lCoarseNodesPerDim[1] - 2) {
-          gridIdx[1] = 0;
-          ++gridIdx[2];
+  }
+
+  if (numInteriors > 0) {
+    // Allocate and compute arrays
+    // containing column offsets
+    // and values associated with
+    // interior points
+    LO countRowEntries = 0;
+    Array<LO> coordColumnOffsets(125);
+    for (LO k = -2; k < 3; ++k) {
+      for (LO j = -2; j < 3; ++j) {
+        for (LO i = -2; i < 3; ++i) {
+          coordColumnOffsets[countRowEntries] = k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i;
+          ++countRowEntries;
         }
       }
     }
 
-    if(numInteriors > 0) {
-      // Allocate and compute arrays
-      // containing column offsets
-      // and values associated with
-      // interior points
-      LO countRowEntries = 0;
-      Array<LO> coordColumnOffsets(125);
-      for(LO k = -2; k < 3; ++k) {
-        for(LO j = -2; j < 3; ++j) {
-          for(LO i = -2; i < 3; ++i) {
-            coordColumnOffsets[countRowEntries] = k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-	      + j*lFineNodesPerDim[0] + i;
-            ++countRowEntries;
-          }
+    LO countValues = 0;
+    Array<SC> interiorValues(125);
+    for (LO k = 0; k < 5; ++k) {
+      for (LO j = 0; j < 5; ++j) {
+        for (LO i = 0; i < 5; ++i) {
+          interiorValues[countValues] = coeffs[k] * coeffs[j] * coeffs[i];
+          ++countValues;
         }
       }
+    }
 
-      LO countValues = 0;
-      Array<SC> interiorValues(125);
-      for(LO k = 0; k < 5; ++k) {
-        for(LO j = 0; j < 5; ++j) {
-          for(LO i = 0; i < 5; ++i) {
-            interiorValues[countValues] = coeffs[k]*coeffs[j]*coeffs[i];
-            ++countValues;
-	  }
-        }
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    Array<LO> gridIdx(3);
+    for (LO interiorIdx = 0; interiorIdx < numInteriors; ++interiorIdx) {
+      coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim[0] * lCoarseNodesPerDim[1] + (gridIdx[1] + 1) * lCoarseNodesPerDim[0] + gridIdx[0] + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((gridIdx[2] + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (gridIdx[1] + 1) * 3 * lFineNodesPerDim[0] + (gridIdx[0] + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + faceLineOffset + faceStencilLength + gridIdx[2] * interiorPlaneOffset + gridIdx[1] * interiorLineOffset + gridIdx[0] * interiorStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + interiorStencilLength * (l + 1);
       }
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      Array<LO> gridIdx(3);
-      for(LO interiorIdx = 0; interiorIdx < numInteriors; ++interiorIdx) {
-        coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim[0]*lCoarseNodesPerDim[1]
-		       + (gridIdx[1] + 1)*lCoarseNodesPerDim[0]
-		       + gridIdx[0] + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((gridIdx[2] + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-			     + (gridIdx[1] + 1)*3*lFineNodesPerDim[0] + (gridIdx[0] + 1)*3);
-      	columnOffset = coordColumnOffset*blkSize;
-        entryOffset = (facePlaneOffset + faceLineOffset + faceStencilLength
-                       + gridIdx[2]*interiorPlaneOffset + gridIdx[1]*interiorLineOffset
-                       + gridIdx[0]*interiorStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + interiorStencilLength*(l+1);
-      	}
-        // Fill the column indices
-        // and values in the approproate
-        // views.
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO entryIdx = 0; entryIdx < interiorStencilLength; ++entryIdx) {
-            entries_h(entryOffset + entryIdx + interiorStencilLength*l) = columnOffset + coordColumnOffsets[entryIdx]*blkSize + l;
-            values_h(entryOffset + entryIdx + interiorStencilLength*l) = interiorValues[entryIdx];
-	  }
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      // Fill the column indices
+      // and values in the approproate
+      // views.
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO entryIdx = 0; entryIdx < interiorStencilLength; ++entryIdx) {
+          entries_h(entryOffset + entryIdx + interiorStencilLength * l) = columnOffset + coordColumnOffsets[entryIdx] * blkSize + l;
+          values_h(entryOffset + entryIdx + interiorStencilLength * l)  = interiorValues[entryIdx];
         }
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Last step in the loop
-        // update the grid indices
-        // for next grid point
-        ++gridIdx[0];
-        if(gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
-          gridIdx[0] = 0;
-          ++gridIdx[1];
-          if(gridIdx[1] == lCoarseNodesPerDim[1] - 2) {
-            gridIdx[1] = 0;
-            ++gridIdx[2];
-          }
+      // Last step in the loop
+      // update the grid indices
+      // for next grid point
+      ++gridIdx[0];
+      if (gridIdx[0] == lCoarseNodesPerDim[0] - 2) {
+        gridIdx[0] = 0;
+        ++gridIdx[1];
+        if (gridIdx[1] == lCoarseNodesPerDim[1] - 2) {
+          gridIdx[1] = 0;
+          ++gridIdx[2];
         }
       }
     }
+  }
 
-    Kokkos::deep_copy(row_map, row_map_h);
-    Kokkos::deep_copy(entries, entries_h);
-    Kokkos::deep_copy(values,  values_h);
+  Kokkos::deep_copy(row_map, row_map_h);
+  Kokkos::deep_copy(entries, entries_h);
+  Kokkos::deep_copy(values, values_h);
 
-    local_graph_type localGraph(entries, row_map);
-    local_matrix_type localR("R", numCols, values, localGraph);
+  local_graph_type localGraph(entries, row_map);
+  local_matrix_type localR("R", numCols, values, localGraph);
 
-    R = MatrixFactory::Build(localR,            // the local data
-                             rowMap,            // rowMap
-                             A->getRowMap(),    // colMap
-                             A->getRowMap(),    // domainMap == colMap
-                             rowMap,            // rangeMap  == rowMap
-                             Teuchos::null);    // params for optimized construction
+  R = MatrixFactory::Build(localR,          // the local data
+                           rowMap,          // rowMap
+                           A->getRowMap(),  // colMap
+                           A->getRowMap(),  // domainMap == colMap
+                           rowMap,          // rangeMap  == rowMap
+                           Teuchos::null);  // params for optimized construction
 
-  } // Build3D()
+}  // Build3D()
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REGIONRFACTORY_SHORT
-#endif // MUELU_REGIONRFACTORY_DEF_HPP
+#endif  // MUELU_REGIONRFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_decl.hpp
index 79f6d8d88dd6..beac8939539b 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_decl.hpp
@@ -58,64 +58,64 @@
 
 namespace MueLu {
 
-  /*!
-    @class RegionRFactory_kokkos class
-    @brief Factory that builds a restriction operator for region multigrid
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RegionRFactory_kokkos : public TwoLevelFactoryBase {
+/*!
+  @class RegionRFactory_kokkos class
+  @brief Factory that builds a restriction operator for region multigrid
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RegionRFactory_kokkos : public TwoLevelFactoryBase {
 #undef MUELU_REGIONRFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using realvaluedmultivector_type = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
-    using execution_space = typename Node::execution_space;
-    using memory_space    = typename Node::memory_space;
-    using device_type     = Kokkos::Device<execution_space, memory_space>;
-    using intTupleView    = typename Kokkos::View<int[3], device_type>;
-    using LOTupleView     = typename Kokkos::View<LO[3],  device_type>;
+ public:
+  using real_type                  = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using realvaluedmultivector_type = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using execution_space            = typename Node::execution_space;
+  using memory_space               = typename Node::memory_space;
+  using device_type                = Kokkos::Device<execution_space, memory_space>;
+  using intTupleView               = typename Kokkos::View<int[3], device_type>;
+  using LOTupleView                = typename Kokkos::View<LO[3], device_type>;
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Default Constructor
-    RegionRFactory_kokkos() = default;
+  //! Default Constructor
+  RegionRFactory_kokkos() = default;
 
-    //!Destructor
-    virtual ~RegionRFactory_kokkos() = default;
-    //@}
+  //! Destructor
+  virtual ~RegionRFactory_kokkos() = default;
+  //@}
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build3D(const int numDimensions,
-                 Array<LO>& lFineNodesPerDim,
-                 const RCP<Matrix>& A,
-                 const RCP<realvaluedmultivector_type>& fineCoordinates,
-                 RCP<Matrix>& R,
-                 RCP<realvaluedmultivector_type>& coarseCoordinates,
-                 Array<LO>& lCoarseNodesPerDim) const;
+  void Build3D(const int numDimensions,
+               Array<LO>& lFineNodesPerDim,
+               const RCP<Matrix>& A,
+               const RCP<realvaluedmultivector_type>& fineCoordinates,
+               RCP<Matrix>& R,
+               RCP<realvaluedmultivector_type>& coarseCoordinates,
+               Array<LO>& lCoarseNodesPerDim) const;
 
-    //@}
+  //@}
 
-  }; // class RegionRFactory_kokkos
+};  // class RegionRFactory_kokkos
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REGIONRFACTORY_KOKKOS_SHORT
-#endif // MUELU_REGIONRFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_REGIONRFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_def.hpp
index 83244b8fa662..9068e20667b7 100644
--- a/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/GeneralGeometric/MueLu_RegionRFactory_kokkos_def.hpp
@@ -55,1144 +55,1054 @@
 
 #include "MueLu_Types.hpp"
 
-
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set<RCP<const FactoryBase> >("A",                            Teuchos::null,
-                                                 "Generating factory of the matrix A");
-    validParamList->set<RCP<const FactoryBase> >("numDimensions",                Teuchos::null,
-                                                 "Number of spatial dimensions in the problem.");
-    validParamList->set<RCP<const FactoryBase> >("lNodesPerDim",                 Teuchos::null,
-                                                 "Number of local nodes per spatial dimension on the fine grid.");
-    validParamList->set<RCP<const FactoryBase> >("Nullspace",                    Teuchos::null,
-                                                 "Fine level nullspace used to construct the coarse level nullspace.");
-    validParamList->set<RCP<const FactoryBase> >("Coordinates",                  Teuchos::null,
-                                                 "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
-    validParamList->set<bool>                   ("keep coarse coords",           false, "Flag to keep coordinates for special coarse grid solve");
-
-    return validParamList;
-  } // GetValidParameterList()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "numDimensions");
-    Input(fineLevel, "lNodesPerDim");
-    Input(fineLevel, "Nullspace");
-    Input(fineLevel, "Coordinates");
-
-  } // DeclareInput()
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build(Level& fineLevel, Level& coarseLevel) const {
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null,
+                                               "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("numDimensions", Teuchos::null,
+                                               "Number of spatial dimensions in the problem.");
+  validParamList->set<RCP<const FactoryBase> >("lNodesPerDim", Teuchos::null,
+                                               "Number of local nodes per spatial dimension on the fine grid.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null,
+                                               "Fine level nullspace used to construct the coarse level nullspace.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null,
+                                               "Fine level coordinates used to construct piece-wise linear prolongator and coarse level coordinates.");
+  validParamList->set<bool>("keep coarse coords", false, "Flag to keep coordinates for special coarse grid solve");
+
+  return validParamList;
+}  // GetValidParameterList()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "numDimensions");
+  Input(fineLevel, "lNodesPerDim");
+  Input(fineLevel, "Nullspace");
+  Input(fineLevel, "Coordinates");
+
+}  // DeclareInput()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build(Level& fineLevel, Level& coarseLevel) const {
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  *out << "Starting RegionRFactory_kokkos::Build." << std::endl;
+
+  // First get the inputs from the fineLevel
+  const int numDimensions = Get<int>(fineLevel, "numDimensions");
+  Array<LO> lFineNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
+  {
+    Array<LO> lNodesPerDim = Get<Array<LO> >(fineLevel, "lNodesPerDim");
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      lFineNodesPerDim[dim] = lNodesPerDim[dim];
     }
-
-    *out << "Starting RegionRFactory_kokkos::Build." << std::endl;
-
-    // First get the inputs from the fineLevel
-    const int numDimensions = Get<int>(fineLevel, "numDimensions");
-    Array<LO> lFineNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
-    {
-      Array<LO> lNodesPerDim = Get<Array<LO> >(fineLevel, "lNodesPerDim");
-      for(int dim = 0; dim < numDimensions; ++dim) {
-        lFineNodesPerDim[dim] = lNodesPerDim[dim];
+  }
+  *out << "numDimensions " << numDimensions << " and lFineNodesPerDim: " << lFineNodesPerDim
+       << std::endl;
+
+  // Let us check that the inputs verify our assumptions
+  if (numDimensions < 1 || numDimensions > 3) {
+    throw std::runtime_error("numDimensions must be 1, 2 or 3!");
+  }
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    if (lFineNodesPerDim[dim] % 3 != 1) {
+      throw std::runtime_error("The number of fine node in each direction need to be 3n+1");
+    }
+  }
+  Array<LO> lCoarseNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
+
+  const RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
+
+  RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
+  fineCoordinates = Get<RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
+  if (static_cast<int>(fineCoordinates->getNumVectors()) != numDimensions) {
+    throw std::runtime_error("The number of vectors in the coordinates is not equal to numDimensions!");
+  }
+
+  // Let us create R and pass it down to the
+  // appropriate specialization and see what we
+  // get back!
+  RCP<Matrix> R;
+
+  if (numDimensions == 1) {
+    throw std::runtime_error("RegionRFactory_kokkos no implemented for 1D case yet.");
+  } else if (numDimensions == 2) {
+    throw std::runtime_error("RegionRFactory_kokkos no implemented for 2D case yet.");
+  } else if (numDimensions == 3) {
+    Build3D(numDimensions, lFineNodesPerDim, A, fineCoordinates,
+            R, coarseCoordinates, lCoarseNodesPerDim);
+  }
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  // Reuse pattern if available (multiple solve)
+  RCP<ParameterList> Tparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    Tparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    Tparams = rcp(new ParameterList);
+
+  // R->describe(*out, Teuchos::VERB_EXTREME);
+  *out << "Compute P=R^t" << std::endl;
+  // By default, we don't need global constants for transpose
+  Tparams->set("compute global constants: temporaries", Tparams->get("compute global constants: temporaries", false));
+  Tparams->set("compute global constants", Tparams->get("compute global constants", false));
+  std::string label = "MueLu::RegionR-transR" + Teuchos::toString(coarseLevel.GetLevelID());
+  RCP<Matrix> P     = Utilities::Transpose(*R, true, label, Tparams);
+
+  *out << "Compute coarse nullspace" << std::endl;
+  RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(R->getRowMap(),
+                                                               fineNullspace->getNumVectors());
+  R->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+           Teuchos::ScalarTraits<SC>::zero());
+
+  *out << "Set data on coarse level" << std::endl;
+  Set(coarseLevel, "numDimensions", numDimensions);
+  Set(coarseLevel, "lNodesPerDim", lCoarseNodesPerDim);
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "Coordinates", coarseCoordinates);
+  if (pL.get<bool>("keep coarse coords")) {
+    coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
+  }
+
+  R->SetFixedBlockSize(A->GetFixedBlockSize());
+  P->SetFixedBlockSize(A->GetFixedBlockSize());
+
+  Set(coarseLevel, "R", R);
+  Set(coarseLevel, "P", P);
+
+}  // Build()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Build3D(const int numDimensions,
+            Teuchos::Array<LocalOrdinal>& lFineNodesPerDim,
+            const RCP<Matrix>& A,
+            const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& fineCoordinates,
+            RCP<Matrix>& R,
+            RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coarseCoordinates,
+            Teuchos::Array<LocalOrdinal>& lCoarseNodesPerDim) const {
+  using local_matrix_type = typename CrsMatrix::local_matrix_type;
+  using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
+  using row_map_type      = typename local_matrix_type::row_map_type::non_const_type;
+  using entries_type      = typename local_matrix_type::index_type::non_const_type;
+  using values_type       = typename local_matrix_type::values_type::non_const_type;
+  using impl_scalar_type  = typename Kokkos::ArithTraits<Scalar>::val_type;
+
+  // Set debug outputs based on environment variable
+  RCP<Teuchos::FancyOStream> out;
+  if (const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
+    out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    out->setShowAllFrontMatter(false).setShowProcRank(true);
+  } else {
+    out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  }
+
+  // Now compute number of coarse grid points
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    lCoarseNodesPerDim[dim] = lFineNodesPerDim[dim] / 3 + 1;
+  }
+  *out << "lCoarseNodesPerDim " << lCoarseNodesPerDim << std::endl;
+
+  // Grab the block size here and multiply all existing offsets by it
+  const LO blkSize = A->GetFixedBlockSize();
+  *out << "blkSize " << blkSize << std::endl;
+
+  // Based on lCoarseNodesPerDim and lFineNodesPerDim
+  // we can compute numRows, numCols and NNZ for R
+  const LO numRows = blkSize * lCoarseNodesPerDim[0] * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[2];
+  const LO numCols = blkSize * lFineNodesPerDim[0] * lFineNodesPerDim[1] * lFineNodesPerDim[2];
+
+  // Create the coarse coordinates multivector
+  // so we can fill it on the fly while computing
+  // the restriction operator
+  RCP<Map> rowMap = MapFactory::Build(A->getRowMap()->lib(),
+                                      Teuchos::OrdinalTraits<GO>::invalid(),
+                                      numRows,
+                                      A->getRowMap()->getIndexBase(),
+                                      A->getRowMap()->getComm());
+
+  RCP<Map> coordRowMap = MapFactory::Build(A->getRowMap()->lib(),
+                                           Teuchos::OrdinalTraits<GO>::invalid(),
+                                           lCoarseNodesPerDim[0] * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[2],
+                                           A->getRowMap()->getIndexBase(),
+                                           A->getRowMap()->getComm());
+
+  coarseCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(coordRowMap,
+                                                                               numDimensions);
+
+  // Get device views of coordinates
+  auto fineCoordsView   = fineCoordinates->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto coarseCoordsView = coarseCoordinates->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+
+  Array<ArrayRCP<const real_type> > fineCoordData(numDimensions);
+  Array<ArrayRCP<real_type> > coarseCoordData(numDimensions);
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    fineCoordData[dim]   = fineCoordinates->getData(dim);
+    coarseCoordData[dim] = coarseCoordinates->getDataNonConst(dim);
+  }
+
+  // Let us set some parameter that will be useful
+  // while constructing R
+
+  // Length of interpolation stencils based on geometry
+  const LO cornerStencilLength   = 27;
+  const LO edgeStencilLength     = 45;
+  const LO faceStencilLength     = 75;
+  const LO interiorStencilLength = 125;
+
+  // Number of corner, edge, face and interior nodes
+  const LO numCorners   = 8;
+  const LO numEdges     = 4 * (lCoarseNodesPerDim[0] - 2) + 4 * (lCoarseNodesPerDim[1] - 2) + 4 * (lCoarseNodesPerDim[2] - 2);
+  const LO numFaces     = 2 * (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[1] - 2) + 2 * (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[2] - 2) + 2 * (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[2] - 2);
+  const LO numInteriors = (lCoarseNodesPerDim[0] - 2) * (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[2] - 2);
+
+  const LO nnz = (numCorners * cornerStencilLength + numEdges * edgeStencilLength + numFaces * faceStencilLength + numInteriors * interiorStencilLength) * blkSize;
+
+  // Having the number of rows and columns we can genrate
+  // the appropriate maps for our operator.
+
+  *out << "R statistics:" << std::endl
+       << "  -numRows= " << numRows << std::endl
+       << "  -numCols= " << numCols << std::endl
+       << "  -nnz=     " << nnz << std::endl;
+
+  row_map_type row_map(Kokkos::ViewAllocateWithoutInitializing("row_map"), numRows + 1);
+  typename row_map_type::HostMirror row_map_h = Kokkos::create_mirror_view(row_map);
+
+  entries_type entries(Kokkos::ViewAllocateWithoutInitializing("entries"), nnz);
+  typename entries_type::HostMirror entries_h = Kokkos::create_mirror_view(entries);
+
+  values_type values(Kokkos::ViewAllocateWithoutInitializing("values"), nnz);
+  typename values_type::HostMirror values_h = Kokkos::create_mirror_view(values);
+
+  // Compute the basic interpolation
+  // coefficients for 1D rate of 3
+  // coarsening.
+  Array<SC> coeffs({1.0 / 3.0, 2.0 / 3.0, 1.0, 2.0 / 3.0, 1.0 / 3.0});
+  row_map_h(0) = 0;
+
+  // Define some offsets that
+  // will be needed often later on
+  const LO edgeLineOffset     = 2 * cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength;
+  const LO faceLineOffset     = 2 * edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength;
+  const LO interiorLineOffset = 2 * faceStencilLength + (lCoarseNodesPerDim[0] - 2) * interiorStencilLength;
+
+  const LO facePlaneOffset     = 2 * edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset;
+  const LO interiorPlaneOffset = 2 * faceLineOffset + (lCoarseNodesPerDim[1] - 2) * interiorLineOffset;
+
+  // Let us take care of the corners
+  // first since we always have
+  // corners to deal with!
+  {
+    // Corner 1
+    LO coordRowIdx = 0, rowIdx = 0, coordColumnOffset = 0, columnOffset = 0, entryOffset = 0;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i + 2];
+          }
+        }
       }
     }
-    *out << "numDimensions " << numDimensions << " and lFineNodesPerDim: " << lFineNodesPerDim
-         << std::endl;
-
-    // Let us check that the inputs verify our assumptions
-    if(numDimensions < 1 || numDimensions > 3) {
-      throw std::runtime_error("numDimensions must be 1, 2 or 3!");
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      if(lFineNodesPerDim[dim] % 3 != 1) {
-        throw std::runtime_error("The number of fine node in each direction need to be 3n+1");
-      }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
-    Array<LO> lCoarseNodesPerDim(3, Teuchos::OrdinalTraits<LO>::one());
-
-    const RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
 
-    RCP<realvaluedmultivector_type> fineCoordinates, coarseCoordinates;
-    fineCoordinates = Get< RCP<realvaluedmultivector_type> >(fineLevel, "Coordinates");
-    if(static_cast<int>(fineCoordinates->getNumVectors()) != numDimensions) {
-      throw std::runtime_error("The number of vectors in the coordinates is not equal to numDimensions!");
+    // Corner 5
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i + 2];
+          }
+        }
+      }
     }
-
-    // Let us create R and pass it down to the
-    // appropriate specialization and see what we
-    // get back!
-    RCP<Matrix> R;
-
-    if(numDimensions == 1) {
-      throw std::runtime_error("RegionRFactory_kokkos no implemented for 1D case yet.");
-    } else if(numDimensions == 2) {
-      throw std::runtime_error("RegionRFactory_kokkos no implemented for 2D case yet.");
-    } else if(numDimensions == 3) {
-      Build3D(numDimensions, lFineNodesPerDim, A, fineCoordinates,
-              R, coarseCoordinates, lCoarseNodesPerDim);
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-
-    // Reuse pattern if available (multiple solve)
-    RCP<ParameterList> Tparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      Tparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-    else
-      Tparams= rcp(new ParameterList);
-
-    // R->describe(*out, Teuchos::VERB_EXTREME);
-    *out << "Compute P=R^t" << std::endl;
-    // By default, we don't need global constants for transpose
-    Tparams->set("compute global constants: temporaries",Tparams->get("compute global constants: temporaries", false));
-    Tparams->set("compute global constants", Tparams->get("compute global constants",false));
-    std::string label = "MueLu::RegionR-transR" + Teuchos::toString(coarseLevel.GetLevelID());
-    RCP<Matrix> P = Utilities::Transpose(*R, true, label, Tparams);
-
-    *out << "Compute coarse nullspace" << std::endl;
-    RCP<MultiVector> fineNullspace   = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(R->getRowMap(),
-                                                                 fineNullspace->getNumVectors());
-    R->apply(*fineNullspace, *coarseNullspace, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-             Teuchos::ScalarTraits<SC>::zero());
-
-    *out << "Set data on coarse level" << std::endl;
-    Set(coarseLevel, "numDimensions", numDimensions);
-    Set(coarseLevel, "lNodesPerDim",  lCoarseNodesPerDim);
-    Set(coarseLevel, "Nullspace",     coarseNullspace);
-    Set(coarseLevel, "Coordinates",   coarseCoordinates);
-    if(pL.get<bool>("keep coarse coords")) {
-      coarseLevel.Set<RCP<realvaluedmultivector_type> >("Coordinates2", coarseCoordinates, NoFactory::get());
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
 
-    R->SetFixedBlockSize(A->GetFixedBlockSize());
-    P->SetFixedBlockSize(A->GetFixedBlockSize());
-
-    Set(coarseLevel, "R", R);
-    Set(coarseLevel, "P", P);
-
-  } // Build()
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RegionRFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Build3D(const int numDimensions,
-          Teuchos::Array<LocalOrdinal>& lFineNodesPerDim,
-          const RCP<Matrix>& A,
-          const RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& fineCoordinates,
-          RCP<Matrix>& R,
-          RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coarseCoordinates,
-          Teuchos::Array<LocalOrdinal>& lCoarseNodesPerDim) const {
-    using local_matrix_type = typename CrsMatrix::local_matrix_type;
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using row_map_type      = typename local_matrix_type::row_map_type::non_const_type;
-    using entries_type      = typename local_matrix_type::index_type::non_const_type;
-    using values_type       = typename local_matrix_type::values_type::non_const_type;
-    using impl_scalar_type = typename Kokkos::ArithTraits<Scalar>::val_type;
-
-    // Set debug outputs based on environment variable
-    RCP<Teuchos::FancyOStream> out;
-    if(const char* dbg = std::getenv("MUELU_REGIONRFACTORY_DEBUG")) {
-      out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      out->setShowAllFrontMatter(false).setShowProcRank(true);
-    } else {
-      out = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+    // Corner 2
+    coordRowIdx       = (lCoarseNodesPerDim[0] - 1);
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[0] - 1);
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i];
+          }
+        }
+      }
     }
-
-    // Now compute number of coarse grid points
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      lCoarseNodesPerDim[dim] = lFineNodesPerDim[dim] / 3 + 1;
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
     }
-    *out << "lCoarseNodesPerDim " << lCoarseNodesPerDim << std::endl;
-
-    // Grab the block size here and multiply all existing offsets by it
-    const LO blkSize = A->GetFixedBlockSize();
-    *out << "blkSize " << blkSize << std::endl;
-
-    // Based on lCoarseNodesPerDim and lFineNodesPerDim
-    // we can compute numRows, numCols and NNZ for R
-    const LO numRows = blkSize*lCoarseNodesPerDim[0]*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[2];
-    const LO numCols = blkSize*lFineNodesPerDim[0]*lFineNodesPerDim[1]*lFineNodesPerDim[2];
-
-    // Create the coarse coordinates multivector
-    // so we can fill it on the fly while computing
-    // the restriction operator
-    RCP<Map> rowMap = MapFactory::Build(A->getRowMap()->lib(),
-                                        Teuchos::OrdinalTraits<GO>::invalid(),
-                                        numRows,
-                                        A->getRowMap()->getIndexBase(),
-                                        A->getRowMap()->getComm());
-
-    RCP<Map> coordRowMap = MapFactory::Build(A->getRowMap()->lib(),
-                                        Teuchos::OrdinalTraits<GO>::invalid(),
-                                        lCoarseNodesPerDim[0]*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[2],
-                                        A->getRowMap()->getIndexBase(),
-                                        A->getRowMap()->getComm());
-
-    coarseCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(coordRowMap,
-                                                                                 numDimensions);
-
-    // Get device views of coordinates
-    auto fineCoordsView   = fineCoordinates  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto coarseCoordsView = coarseCoordinates->getDeviceLocalView(Xpetra::Access::OverwriteAll);
-
-
-    Array<ArrayRCP<const real_type> > fineCoordData(numDimensions);
-    Array<ArrayRCP<real_type> > coarseCoordData(numDimensions);
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      fineCoordData[dim] = fineCoordinates->getData(dim);
-      coarseCoordData[dim] = coarseCoordinates->getDataNonConst(dim);
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
     }
 
-    // Let us set some parameter that will be useful
-    // while constructing R
-
-    // Length of interpolation stencils based on geometry
-    const LO cornerStencilLength   = 27;
-    const LO edgeStencilLength     = 45;
-    const LO faceStencilLength     = 75;
-    const LO interiorStencilLength = 125;
-
-    // Number of corner, edge, face and interior nodes
-    const LO numCorners   = 8;
-    const LO numEdges     = 4*(lCoarseNodesPerDim[0] - 2)
-      + 4*(lCoarseNodesPerDim[1] - 2)
-      + 4*(lCoarseNodesPerDim[2] - 2);
-    const LO numFaces     = 2*(lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[1] - 2)
-      + 2*(lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[2] - 2)
-      + 2*(lCoarseNodesPerDim[1] - 2)*(lCoarseNodesPerDim[2] - 2);
-    const LO numInteriors = (lCoarseNodesPerDim[0] - 2)*(lCoarseNodesPerDim[1] - 2)
-      *(lCoarseNodesPerDim[2] - 2);
-
-    const LO nnz = (numCorners*cornerStencilLength + numEdges*edgeStencilLength
-            + numFaces*faceStencilLength + numInteriors*interiorStencilLength)*blkSize;
-
-    // Having the number of rows and columns we can genrate
-    // the appropriate maps for our operator.
-
-    *out << "R statistics:" << std::endl
-         << "  -numRows= " << numRows << std::endl
-         << "  -numCols= " << numCols << std::endl
-         << "  -nnz=     " << nnz << std::endl;
-
-    row_map_type row_map(Kokkos::ViewAllocateWithoutInitializing("row_map"), numRows + 1);
-    typename row_map_type::HostMirror row_map_h = Kokkos::create_mirror_view(row_map);
-
-    entries_type entries(Kokkos::ViewAllocateWithoutInitializing("entries"), nnz);
-    typename entries_type::HostMirror entries_h = Kokkos::create_mirror_view(entries);
-
-    values_type values(Kokkos::ViewAllocateWithoutInitializing("values"), nnz);
-    typename values_type::HostMirror values_h = Kokkos::create_mirror_view(values);
-
-    // Compute the basic interpolation
-    // coefficients for 1D rate of 3
-    // coarsening.
-    Array<SC> coeffs({1.0/3.0, 2.0/3.0, 1.0, 2.0/3.0, 1.0/3.0});
-    row_map_h(0) = 0;
-
-    // Define some offsets that
-    // will be needed often later on
-    const LO edgeLineOffset = 2*cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength;
-    const LO faceLineOffset = 2*edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength;
-    const LO interiorLineOffset = 2*faceStencilLength
-      + (lCoarseNodesPerDim[0] - 2)*interiorStencilLength;
-
-    const LO facePlaneOffset = 2*edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset;
-    const LO interiorPlaneOffset = 2*faceLineOffset + (lCoarseNodesPerDim[1] - 2)*interiorLineOffset;
-
-    // Let us take care of the corners
-    // first since we always have
-    // corners to deal with!
-    {
-      // Corner 1
-      LO coordRowIdx = 0, rowIdx = 0, coordColumnOffset = 0, columnOffset = 0, entryOffset = 0;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i + 2];
+    // Corner 6
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
+          }
         }
       }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+
+    // Corner 3
+    coordRowIdx       = (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0];
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0];
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i + 2];
+          }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+
+    // Corner 7
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
+          }
+        }
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
-        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+
+    // Corner 4
+    coordRowIdx       = (lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] - 1);
+    rowIdx            = coordRowIdx * blkSize;
+    coordColumnOffset = (lFineNodesPerDim[1] * lFineNodesPerDim[0] - 1);
+    columnOffset      = coordColumnOffset * blkSize;
+    entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset +
+                   cornerStencilLength + (lCoarseNodesPerDim[0] - 2) * edgeStencilLength) *
+                  blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
+          }
+        }
       }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
 
-      // Corner 5
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i + 2];
+    // Corner 8
+    coordRowIdx += (lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+    rowIdx = coordRowIdx * blkSize;
+    coordColumnOffset += (lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+    columnOffset = coordColumnOffset * blkSize;
+    entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset) * blkSize;
+    for (LO l = 0; l < blkSize; ++l) {
+      for (LO k = 0; k < 3; ++k) {
+        for (LO j = 0; j < 3; ++j) {
+          for (LO i = 0; i < 3; ++i) {
+            entries_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + (i - 2)) * blkSize + l;
+            values_h(entryOffset + k * 9 + j * 3 + i + cornerStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
+          }
         }
+      }
+    }
+    for (LO l = 0; l < blkSize; ++l) {
+      row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength * (l + 1);
+    }
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+    }
+  }  // Corners are done!
+
+  // Edges along 0 direction
+  if (lCoarseNodesPerDim[0] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[0] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3;
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j + 2] * coeffs[i];
+            }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 2
-      coordRowIdx = (lCoarseNodesPerDim[0] - 1);
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[0] - 1);
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-           + j*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i];
+      // Edge 1
+      coordRowIdx       = ((lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 6
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
+      // Edge 2
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 3
-      coordRowIdx = (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-           + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i + 2];
+      // Edge 3
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0] + edgeIdx + 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0] + (edgeIdx + 1) * 3);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + (lCoarseNodesPerDim[1] - 2) * faceLineOffset + cornerStencilLength + edgeIdx * edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 5; ++i) {
+              entries_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 5 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
-
-      // Corner 7
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
+    }
+  }
+
+  // Edges along 1 direction
+  if (lCoarseNodesPerDim[1] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[1] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1) * lCoarseNodesPerDim[0];
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3 * lFineNodesPerDim[0];
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + edgeIdx * faceLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 4
-      coordRowIdx = (lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0] - 1);
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset = (lFineNodesPerDim[1]*lFineNodesPerDim[0] - 1);
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset +
-             cornerStencilLength + (lCoarseNodesPerDim[0] - 2)*edgeStencilLength)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-           + (j - 2)*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
+      // Edge 1
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (edgeLineOffset + edgeIdx * faceLineOffset + edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + (k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k + 2] * coeffs[j] * coeffs[i];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
 
-      // Corner 8
-      coordRowIdx += (lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-      rowIdx = coordRowIdx*blkSize;
-      coordColumnOffset += (lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-      columnOffset = coordColumnOffset*blkSize;
-      entryOffset += (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset)*blkSize;
-      for(LO l = 0; l < blkSize; ++l) {
-        for(LO k = 0; k < 3; ++k) {
-          for(LO j = 0; j < 3; ++j) {
-            for(LO i = 0; i < 3; ++i) {
-              entries_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l) = columnOffset
-                + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + (i - 2))*blkSize + l;
-              values_h(entryOffset + k*9 + j*3 + i + cornerStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
+      // Edge 2
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (edgeIdx + 1) * lCoarseNodesPerDim[0]);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3 * lFineNodesPerDim[0]);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + edgeIdx * faceLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
       }
-      for(LO l = 0; l < blkSize; ++l) {
-        row_map_h(rowIdx + 1 + l) = entryOffset + cornerStencilLength*(l+1);
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
       }
-      for(int dim = 0; dim <numDimensions; ++dim) {
+      for (int dim = 0; dim < numDimensions; ++dim) {
         coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
-    } // Corners are done!
-
-    // Edges along 0 direction
-    if(lCoarseNodesPerDim[0] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[0] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3;
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j + 2]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
 
-        // Edge 1
-        coordRowIdx = ((lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0] + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0] + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset
-            + cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
-              }
+      // Edge 3
+      coordRowIdx       = ((lCoarseNodesPerDim[2] - 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (edgeIdx + 1) * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((lFineNodesPerDim[2] - 1) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (edgeIdx + 1) * 3 * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2) * interiorPlaneOffset + edgeLineOffset + edgeIdx * faceLineOffset + edgeStencilLength + (lCoarseNodesPerDim[0] - 2) * faceStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 3; ++k) {
+          for (LO j = 0; j < 5; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 15 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 2
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-            + cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 3
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0] + edgeIdx + 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0] + (edgeIdx + 1)*3);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  =  (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-             + edgeLineOffset + (lCoarseNodesPerDim[1] - 2)*faceLineOffset
-             + cornerStencilLength + edgeIdx*edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 5; ++i) {
-                entries_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-             + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*5 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
       }
-    }
-
-    // Edges along 1 direction
-    if(lCoarseNodesPerDim[1] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[1] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1)*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + edgeIdx*faceLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i + 2];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 1
-        coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (edgeLineOffset + edgeIdx*faceLineOffset
-            + edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + (k*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k + 2]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 2
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + (edgeIdx + 1)*lCoarseNodesPerDim[0]);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + (edgeIdx + 1)*3*lFineNodesPerDim[0]);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-            + edgeLineOffset + edgeIdx*faceLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
-
-        // Edge 3
-        coordRowIdx = ((lCoarseNodesPerDim[2] - 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + (edgeIdx + 1)*lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((lFineNodesPerDim[2] - 1)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + (edgeIdx + 1)*3*lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (lCoarseNodesPerDim[2] - 2)*interiorPlaneOffset
-            + edgeLineOffset + edgeIdx*faceLineOffset
-            + edgeStencilLength + (lCoarseNodesPerDim[0] - 2)*faceStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 3; ++k) {
-            for(LO j = 0; j < 5; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-             + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*15 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
-            }
-          }
-        }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
     }
-
-    // Edges along 2 direction
-    if(lCoarseNodesPerDim[2] - 2 > 0) {
-
-      LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-      for(LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[2] - 2; ++edgeIdx) {
-
-        // Edge 0
-        coordRowIdx = (edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0];
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = (edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0];
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + edgeIdx*interiorPlaneOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-          + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i + 2];
-              }
+  }
+
+  // Edges along 2 direction
+  if (lCoarseNodesPerDim[2] - 2 > 0) {
+    LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+    for (LO edgeIdx = 0; edgeIdx < lCoarseNodesPerDim[2] - 2; ++edgeIdx) {
+      // Edge 0
+      coordRowIdx       = (edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0];
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = (edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0];
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + edgeIdx * interiorPlaneOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i + 2];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 1
-        coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + faceLineOffset - edgeStencilLength
-                        + edgeIdx*interiorPlaneOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + j*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j + 2]*coeffs[i];
-              }
+      // Edge 1
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + faceLineOffset - edgeStencilLength + edgeIdx * interiorPlaneOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j + 2] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 2
-        coordRowIdx = ((edgeIdx + 1)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0]
-               + (lCoarseNodesPerDim[1] - 1)*lCoarseNodesPerDim[0]);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                             + (lFineNodesPerDim[1] - 1)*lFineNodesPerDim[0]);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + edgeIdx*interiorPlaneOffset + faceLineOffset
-            + (lCoarseNodesPerDim[1] - 2)*interiorLineOffset)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0] + (j - 2)*lFineNodesPerDim[0] + i)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i + 2];
-              }
+      // Edge 2
+      coordRowIdx       = ((edgeIdx + 1) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] + (lCoarseNodesPerDim[1] - 1) * lCoarseNodesPerDim[0]);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (lFineNodesPerDim[1] - 1) * lFineNodesPerDim[0]);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + edgeIdx * interiorPlaneOffset + faceLineOffset + (lCoarseNodesPerDim[1] - 2) * interiorLineOffset) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i + 2];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
+      }
 
-        // Edge 3
-        coordRowIdx = ((edgeIdx + 2)*lCoarseNodesPerDim[1]*lCoarseNodesPerDim[0] - 1);
-        rowIdx = coordRowIdx*blkSize;
-        coordColumnOffset = ((edgeIdx + 1)*3*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-                 + lFineNodesPerDim[1]*lFineNodesPerDim[0] - 1);
-        columnOffset = coordColumnOffset*blkSize;
-        entryOffset  = (facePlaneOffset + (edgeIdx + 1)*interiorPlaneOffset - edgeStencilLength)*blkSize;
-        for(LO l = 0; l < blkSize; ++l) {
-          for(LO k = 0; k < 5; ++k) {
-            for(LO j = 0; j < 3; ++j) {
-              for(LO i = 0; i < 3; ++i) {
-                entries_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l) = columnOffset
-                  + ((k - 2)*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-             + (j - 2)*lFineNodesPerDim[0] + i - 2)*blkSize + l;
-                values_h(entryOffset + k*9 + j*3 + i + edgeStencilLength*l)  = coeffs[k]*coeffs[j]*coeffs[i];
-              }
+      // Edge 3
+      coordRowIdx       = ((edgeIdx + 2) * lCoarseNodesPerDim[1] * lCoarseNodesPerDim[0] - 1);
+      rowIdx            = coordRowIdx * blkSize;
+      coordColumnOffset = ((edgeIdx + 1) * 3 * lFineNodesPerDim[1] * lFineNodesPerDim[0] + lFineNodesPerDim[1] * lFineNodesPerDim[0] - 1);
+      columnOffset      = coordColumnOffset * blkSize;
+      entryOffset       = (facePlaneOffset + (edgeIdx + 1) * interiorPlaneOffset - edgeStencilLength) * blkSize;
+      for (LO l = 0; l < blkSize; ++l) {
+        for (LO k = 0; k < 5; ++k) {
+          for (LO j = 0; j < 3; ++j) {
+            for (LO i = 0; i < 3; ++i) {
+              entries_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim[1] * lFineNodesPerDim[0] + (j - 2) * lFineNodesPerDim[0] + i - 2) * blkSize + l;
+              values_h(entryOffset + k * 9 + j * 3 + i + edgeStencilLength * l)  = coeffs[k] * coeffs[j] * coeffs[i];
             }
           }
         }
-        for(LO l = 0; l < blkSize; ++l) {
-          row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength*(l+1);
-        }
-        for(int dim = 0; dim <numDimensions; ++dim) {
-          coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
-        }
+      }
+      for (LO l = 0; l < blkSize; ++l) {
+        row_map_h(rowIdx + 1 + l) = entryOffset + edgeStencilLength * (l + 1);
+      }
+      for (int dim = 0; dim < numDimensions; ++dim) {
+        coarseCoordData[dim][coordRowIdx] = fineCoordData[dim][coordColumnOffset];
       }
     }
+  }
 
-    //TODO: KOKKOS parallel_for used from here. Not sure if it should be used for edges.
-    Kokkos::deep_copy(row_map, row_map_h);
-    Kokkos::deep_copy(entries, entries_h);
-    Kokkos::deep_copy(values,  values_h);
-
-    // Create views on device for nodes per dim
-    LOTupleView lFineNodesPerDim_d("lFineNodesPerDim");
-    LOTupleView lCoarseNodesPerDim_d("lCoarseNodesPerDim");
+  // TODO: KOKKOS parallel_for used from here. Not sure if it should be used for edges.
+  Kokkos::deep_copy(row_map, row_map_h);
+  Kokkos::deep_copy(entries, entries_h);
+  Kokkos::deep_copy(values, values_h);
 
-    typename Kokkos::View<LO[3], device_type>::HostMirror lCoarseNodesPerDim_h = Kokkos::create_mirror_view( lCoarseNodesPerDim_d );
-    typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDim_h   = Kokkos::create_mirror_view( lFineNodesPerDim_d );
+  // Create views on device for nodes per dim
+  LOTupleView lFineNodesPerDim_d("lFineNodesPerDim");
+  LOTupleView lCoarseNodesPerDim_d("lCoarseNodesPerDim");
 
-    for(int dim = 0; dim < numDimensions; ++dim) {
-      lCoarseNodesPerDim_h(dim) = lCoarseNodesPerDim[dim];
-      lFineNodesPerDim_h(dim) = lFineNodesPerDim[dim];
-    }
+  typename Kokkos::View<LO[3], device_type>::HostMirror lCoarseNodesPerDim_h = Kokkos::create_mirror_view(lCoarseNodesPerDim_d);
+  typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDim_h   = Kokkos::create_mirror_view(lFineNodesPerDim_d);
 
-    Kokkos::deep_copy(lCoarseNodesPerDim_d, lCoarseNodesPerDim_h);
-    Kokkos::deep_copy(lFineNodesPerDim_d, lFineNodesPerDim_h);
+  for (int dim = 0; dim < numDimensions; ++dim) {
+    lCoarseNodesPerDim_h(dim) = lCoarseNodesPerDim[dim];
+    lFineNodesPerDim_h(dim)   = lFineNodesPerDim[dim];
+  }
 
+  Kokkos::deep_copy(lCoarseNodesPerDim_d, lCoarseNodesPerDim_h);
+  Kokkos::deep_copy(lFineNodesPerDim_d, lFineNodesPerDim_h);
 
-    // Faces in 0-1 plane
-    if((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[1] - 2 > 0)) {
-
-      Kokkos::parallel_for("Faces in 0-1 plane region R",
-        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[1]-2)*(lCoarseNodesPerDim[0]-2) ),
+  // Faces in 0-1 plane
+  if ((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[1] - 2 > 0)) {
+    Kokkos::parallel_for(
+        "Faces in 0-1 plane region R",
+        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[1] - 2) * (lCoarseNodesPerDim[0] - 2)),
         KOKKOS_LAMBDA(const LO faceIdx) {
-            LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-            LO gridIdx[3] = {0,0,0};
-            impl_scalar_type coeffs_d[5] = {1.0/3.0, 2.0/3.0, 1.0, 2.0/3.0, 1.0/3.0};
-            // Last step in the loop
-            // update the grid indices
-            // for next grid point
-            for(LO i = 0; i < faceIdx; i++){
-              ++gridIdx[0];
-              if(gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
-                gridIdx[0] = 0;
-                ++gridIdx[1];
-              }
+          LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+          LO gridIdx[3]                = {0, 0, 0};
+          impl_scalar_type coeffs_d[5] = {1.0 / 3.0, 2.0 / 3.0, 1.0, 2.0 / 3.0, 1.0 / 3.0};
+          // Last step in the loop
+          // update the grid indices
+          // for next grid point
+          for (LO i = 0; i < faceIdx; i++) {
+            ++gridIdx[0];
+            if (gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
+              gridIdx[0] = 0;
+              ++gridIdx[1];
             }
+          }
 
-            // Face 0
-            coordRowIdx = ((gridIdx[1] + 1)*lCoarseNodesPerDim_d(0) + gridIdx[0] + 1);
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset = 3*((gridIdx[1] + 1)*lFineNodesPerDim_d(0) + gridIdx[0] + 1);
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  = (edgeLineOffset + edgeStencilLength
-                + gridIdx[1]*faceLineOffset + gridIdx[0]*faceStencilLength)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 3; ++k) {
-                for(LO j = 0; j < 5; ++j) {
-                  for(LO i = 0; i < 5; ++i) {
-                    entries(entryOffset + k*25 + j*5 + i + faceStencilLength*l) = columnOffset
-                      + (k*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0) + (j - 2)*lFineNodesPerDim_d(0) + i - 2)*blkSize + l;
-                    values(entryOffset + k*25 + j*5 + i + faceStencilLength*l)  = coeffs_d[k + 2]*coeffs_d[j]*coeffs_d[i];
-                  }
+          // Face 0
+          coordRowIdx       = ((gridIdx[1] + 1) * lCoarseNodesPerDim_d(0) + gridIdx[0] + 1);
+          rowIdx            = coordRowIdx * blkSize;
+          coordColumnOffset = 3 * ((gridIdx[1] + 1) * lFineNodesPerDim_d(0) + gridIdx[0] + 1);
+          columnOffset      = coordColumnOffset * blkSize;
+          entryOffset       = (edgeLineOffset + edgeStencilLength + gridIdx[1] * faceLineOffset + gridIdx[0] * faceStencilLength) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 3; ++k) {
+              for (LO j = 0; j < 5; ++j) {
+                for (LO i = 0; i < 5; ++i) {
+                  entries(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l) = columnOffset + (k * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (j - 2) * lFineNodesPerDim_d(0) + i - 2) * blkSize + l;
+                  values(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l)  = coeffs_d[k + 2] * coeffs_d[j] * coeffs_d[i];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
+          }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
 
-            // Face 1
-            coordRowIdx += (lCoarseNodesPerDim_d(2) - 1)*lCoarseNodesPerDim_d(1)*lCoarseNodesPerDim_d(0);
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset += (lFineNodesPerDim_d(2) - 1)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0);
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  += (facePlaneOffset + (lCoarseNodesPerDim_d(2) - 2)*interiorPlaneOffset)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 3; ++k) {
-                for(LO j = 0; j < 5; ++j) {
-                  for(LO i = 0; i < 5; ++i) {
-                    entries(entryOffset + k*25 + j*5 + i + faceStencilLength*l) = columnOffset
-                      + ((k - 2)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                 + (j - 2)*lFineNodesPerDim_d(0) + i - 2)*blkSize + l;
-                    values(entryOffset + k*25 + j*5 + i + faceStencilLength*l)  = coeffs_d[k]*coeffs_d[j]*coeffs_d[i];
-                  }
+          // Face 1
+          coordRowIdx += (lCoarseNodesPerDim_d(2) - 1) * lCoarseNodesPerDim_d(1) * lCoarseNodesPerDim_d(0);
+          rowIdx = coordRowIdx * blkSize;
+          coordColumnOffset += (lFineNodesPerDim_d(2) - 1) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0);
+          columnOffset = coordColumnOffset * blkSize;
+          entryOffset += (facePlaneOffset + (lCoarseNodesPerDim_d(2) - 2) * interiorPlaneOffset) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 3; ++k) {
+              for (LO j = 0; j < 5; ++j) {
+                for (LO i = 0; i < 5; ++i) {
+                  entries(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (j - 2) * lFineNodesPerDim_d(0) + i - 2) * blkSize + l;
+                  values(entryOffset + k * 25 + j * 5 + i + faceStencilLength * l)  = coeffs_d[k] * coeffs_d[j] * coeffs_d[i];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
-
-      });// parallel_for faces in 0-1 plane
-    }
-
-    // Faces in 0-2 plane
-    if((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
-
-      Kokkos::parallel_for("Faces in 0-2 plane region R",
-        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[2]-2)*(lCoarseNodesPerDim[0]-2) ),
+          }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
+        });  // parallel_for faces in 0-1 plane
+  }
+
+  // Faces in 0-2 plane
+  if ((lCoarseNodesPerDim[0] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
+    Kokkos::parallel_for(
+        "Faces in 0-2 plane region R",
+        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[2] - 2) * (lCoarseNodesPerDim[0] - 2)),
         KOKKOS_LAMBDA(const LO faceIdx) {
-            LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-            LO gridIdx[3] = {0,0,0};
-            impl_scalar_type coeffs_d[5] = {1.0/3.0, 2.0/3.0, 1.0, 2.0/3.0, 1.0/3.0};
-            // Last step in the loop
-            // update the grid indices
-            // for next grid point
-            for(LO i = 0; i < faceIdx; i++){
-              ++gridIdx[0];
-              if(gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
-                gridIdx[0] = 0;
-                ++gridIdx[2];
-              }
+          LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+          LO gridIdx[3]                = {0, 0, 0};
+          impl_scalar_type coeffs_d[5] = {1.0 / 3.0, 2.0 / 3.0, 1.0, 2.0 / 3.0, 1.0 / 3.0};
+          // Last step in the loop
+          // update the grid indices
+          // for next grid point
+          for (LO i = 0; i < faceIdx; i++) {
+            ++gridIdx[0];
+            if (gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
+              gridIdx[0] = 0;
+              ++gridIdx[2];
             }
+          }
 
-            // Face 0
-            coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim_d(1)*lCoarseNodesPerDim_d(0) + (gridIdx[0] + 1));
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset = ((gridIdx[2] + 1)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                                 + gridIdx[0] + 1)*3;
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  = (facePlaneOffset + gridIdx[2]*interiorPlaneOffset + edgeStencilLength
-                + gridIdx[0]*faceStencilLength)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 5; ++k) {
-                for(LO j = 0; j < 3; ++j) {
-                  for(LO i = 0; i < 5; ++i) {
-                    entries(entryOffset + k*15 + j*5 + i + faceStencilLength*l) = columnOffset
-                      + ((k - 2)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0) + j*lFineNodesPerDim_d(0) + i - 2)*blkSize + l;
-                    values(entryOffset + k*15 + j*5 + i + faceStencilLength*l)  = coeffs_d[k]*coeffs_d[j + 2]*coeffs_d[i];
-                  }
+          // Face 0
+          coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim_d(1) * lCoarseNodesPerDim_d(0) + (gridIdx[0] + 1));
+          rowIdx            = coordRowIdx * blkSize;
+          coordColumnOffset = ((gridIdx[2] + 1) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + gridIdx[0] + 1) * 3;
+          columnOffset      = coordColumnOffset * blkSize;
+          entryOffset       = (facePlaneOffset + gridIdx[2] * interiorPlaneOffset + edgeStencilLength + gridIdx[0] * faceStencilLength) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 5; ++k) {
+              for (LO j = 0; j < 3; ++j) {
+                for (LO i = 0; i < 5; ++i) {
+                  entries(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + j * lFineNodesPerDim_d(0) + i - 2) * blkSize + l;
+                  values(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l)  = coeffs_d[k] * coeffs_d[j + 2] * coeffs_d[i];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
+          }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
 
-            // Face 1
-            coordRowIdx += (lCoarseNodesPerDim_d(1) - 1)*lCoarseNodesPerDim_d(0);
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset += (lFineNodesPerDim_d(1) - 1)*lFineNodesPerDim_d(0);
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  += (faceLineOffset + (lCoarseNodesPerDim_d(1) - 2)*interiorLineOffset)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 5; ++k) {
-                for(LO j = 0; j < 3; ++j) {
-                  for(LO i = 0; i < 5; ++i) {
-                    entries(entryOffset + k*15 + j*5 + i + faceStencilLength*l) = columnOffset
-              + ((k - 2)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                 + (j - 2)*lFineNodesPerDim_d(0) + i - 2)*blkSize + l;
-                    values(entryOffset + k*15 + j*5 + i + faceStencilLength*l)  = coeffs_d[k]*coeffs_d[j]*coeffs_d[i];
-                  }
+          // Face 1
+          coordRowIdx += (lCoarseNodesPerDim_d(1) - 1) * lCoarseNodesPerDim_d(0);
+          rowIdx = coordRowIdx * blkSize;
+          coordColumnOffset += (lFineNodesPerDim_d(1) - 1) * lFineNodesPerDim_d(0);
+          columnOffset = coordColumnOffset * blkSize;
+          entryOffset += (faceLineOffset + (lCoarseNodesPerDim_d(1) - 2) * interiorLineOffset) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 5; ++k) {
+              for (LO j = 0; j < 3; ++j) {
+                for (LO i = 0; i < 5; ++i) {
+                  entries(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (j - 2) * lFineNodesPerDim_d(0) + i - 2) * blkSize + l;
+                  values(entryOffset + k * 15 + j * 5 + i + faceStencilLength * l)  = coeffs_d[k] * coeffs_d[j] * coeffs_d[i];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
-
-      });// parallel_for faces in 0-2 plane
-    }
-
-    // Faces in 1-2 plane
-    if((lCoarseNodesPerDim[1] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
-
-      Kokkos::parallel_for("Faces in 1-2 plane region R",
-        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[2]-2)*(lCoarseNodesPerDim[1]-2) ),
+          }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
+        });  // parallel_for faces in 0-2 plane
+  }
+
+  // Faces in 1-2 plane
+  if ((lCoarseNodesPerDim[1] - 2 > 0) && (lCoarseNodesPerDim[2] - 2 > 0)) {
+    Kokkos::parallel_for(
+        "Faces in 1-2 plane region R",
+        Kokkos::RangePolicy<typename device_type::execution_space>(0, (lCoarseNodesPerDim[2] - 2) * (lCoarseNodesPerDim[1] - 2)),
         KOKKOS_LAMBDA(const LO faceIdx) {
-            LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-            LO gridIdx[3] = {0,0,0};
-            impl_scalar_type coeffs_d[5] = {1.0/3.0, 2.0/3.0, 1.0, 2.0/3.0, 1.0/3.0};
-            // Last step in the loop
-            // update the grid indices
-            // for next grid point
-            for(LO i = 0; i < faceIdx; i++){
-              ++gridIdx[1];
-              if(gridIdx[1] == lCoarseNodesPerDim_d(1) - 2) {
-                gridIdx[1] = 0;
-                ++gridIdx[2];
-              }
+          LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+          LO gridIdx[3]                = {0, 0, 0};
+          impl_scalar_type coeffs_d[5] = {1.0 / 3.0, 2.0 / 3.0, 1.0, 2.0 / 3.0, 1.0 / 3.0};
+          // Last step in the loop
+          // update the grid indices
+          // for next grid point
+          for (LO i = 0; i < faceIdx; i++) {
+            ++gridIdx[1];
+            if (gridIdx[1] == lCoarseNodesPerDim_d(1) - 2) {
+              gridIdx[1] = 0;
+              ++gridIdx[2];
             }
+          }
 
-            // Face 0
-            coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim_d(1)*lCoarseNodesPerDim_d(0)
-                   + (gridIdx[1] + 1)*lCoarseNodesPerDim_d(0));
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset = ((gridIdx[2] + 1)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                     + (gridIdx[1] + 1)*lFineNodesPerDim_d(0))*3;
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  = (facePlaneOffset + gridIdx[2]*interiorPlaneOffset + faceLineOffset
-                + gridIdx[1]*interiorLineOffset)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 5; ++k) {
-                for(LO j = 0; j < 5; ++j) {
-                  for(LO i = 0; i < 3; ++i) {
-                    entries(entryOffset + k*15 + j*3 + i + faceStencilLength*l) = columnOffset
-                      + ((k - 2)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0) + (j - 2)*lFineNodesPerDim_d(0) + i)*blkSize + l;
-                    values(entryOffset + k*15 + j*3 + i + faceStencilLength*l)  = coeffs_d[k]*coeffs_d[j]*coeffs_d[i + 2];
-                  }
+          // Face 0
+          coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim_d(1) * lCoarseNodesPerDim_d(0) + (gridIdx[1] + 1) * lCoarseNodesPerDim_d(0));
+          rowIdx            = coordRowIdx * blkSize;
+          coordColumnOffset = ((gridIdx[2] + 1) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (gridIdx[1] + 1) * lFineNodesPerDim_d(0)) * 3;
+          columnOffset      = coordColumnOffset * blkSize;
+          entryOffset       = (facePlaneOffset + gridIdx[2] * interiorPlaneOffset + faceLineOffset + gridIdx[1] * interiorLineOffset) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 5; ++k) {
+              for (LO j = 0; j < 5; ++j) {
+                for (LO i = 0; i < 3; ++i) {
+                  entries(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (j - 2) * lFineNodesPerDim_d(0) + i) * blkSize + l;
+                  values(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l)  = coeffs_d[k] * coeffs_d[j] * coeffs_d[i + 2];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
+          }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
 
-            // Face 1
-            coordRowIdx += (lCoarseNodesPerDim_d(0) - 1);
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset += (lFineNodesPerDim_d(0) - 1);
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset  += (faceStencilLength + (lCoarseNodesPerDim_d(0) - 2)*interiorStencilLength)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO k = 0; k < 5; ++k) {
-                for(LO j = 0; j < 5; ++j) {
-                  for(LO i = 0; i < 3; ++i) {
-                    entries(entryOffset + k*15 + j*3 + i + faceStencilLength*l) = columnOffset
-                    + ((k - 2)*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                    + (j - 2)*lFineNodesPerDim_d(0) + i - 2)*blkSize + l;
-                    values(entryOffset + k*15 + j*3 + i + faceStencilLength*l)  = coeffs_d[k]*coeffs_d[j]*coeffs_d[i];
-                  }
+          // Face 1
+          coordRowIdx += (lCoarseNodesPerDim_d(0) - 1);
+          rowIdx = coordRowIdx * blkSize;
+          coordColumnOffset += (lFineNodesPerDim_d(0) - 1);
+          columnOffset = coordColumnOffset * blkSize;
+          entryOffset += (faceStencilLength + (lCoarseNodesPerDim_d(0) - 2) * interiorStencilLength) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO k = 0; k < 5; ++k) {
+              for (LO j = 0; j < 5; ++j) {
+                for (LO i = 0; i < 3; ++i) {
+                  entries(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l) = columnOffset + ((k - 2) * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (j - 2) * lFineNodesPerDim_d(0) + i - 2) * blkSize + l;
+                  values(entryOffset + k * 15 + j * 3 + i + faceStencilLength * l)  = coeffs_d[k] * coeffs_d[j] * coeffs_d[i];
                 }
               }
             }
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength*(l+1);
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
-            }
-
-      });// parallel_for faces in 1-2 plane
-    }
-
-    if(numInteriors > 0) {
-
-      // Allocate and compute arrays
-      // containing column offsets
-      // and values associated with
-      // interior points
-      LO countRowEntries = 0;
-      Kokkos::View<LO[125]> coordColumnOffsets_d("coordColOffset");
-      auto coordColumnOffsets_h = Kokkos::create_mirror_view( coordColumnOffsets_d );
-
-      for(LO k = -2; k < 3; ++k) {
-        for(LO j = -2; j < 3; ++j) {
-          for(LO i = -2; i < 3; ++i) {
-            coordColumnOffsets_h(countRowEntries) = k*lFineNodesPerDim[1]*lFineNodesPerDim[0]
-          + j*lFineNodesPerDim[0] + i;
-            ++countRowEntries;
           }
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + faceStencilLength * (l + 1);
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
+        });  // parallel_for faces in 1-2 plane
+  }
+
+  if (numInteriors > 0) {
+    // Allocate and compute arrays
+    // containing column offsets
+    // and values associated with
+    // interior points
+    LO countRowEntries = 0;
+    Kokkos::View<LO[125]> coordColumnOffsets_d("coordColOffset");
+    auto coordColumnOffsets_h = Kokkos::create_mirror_view(coordColumnOffsets_d);
+
+    for (LO k = -2; k < 3; ++k) {
+      for (LO j = -2; j < 3; ++j) {
+        for (LO i = -2; i < 3; ++i) {
+          coordColumnOffsets_h(countRowEntries) = k * lFineNodesPerDim[1] * lFineNodesPerDim[0] + j * lFineNodesPerDim[0] + i;
+          ++countRowEntries;
         }
       }
-      Kokkos::deep_copy(coordColumnOffsets_d, coordColumnOffsets_h);
-
-      LO countValues = 0;
-      Kokkos::View<impl_scalar_type*> interiorValues_d("interiorValues",125);
-      auto interiorValues_h = Kokkos::create_mirror_view( interiorValues_d );
-      for(LO k = 0; k < 5; ++k) {
-        for(LO j = 0; j < 5; ++j) {
-          for(LO i = 0; i < 5; ++i) {
-            interiorValues_h(countValues) = coeffs[k]*coeffs[j]*coeffs[i];
-            ++countValues;
-          }
+    }
+    Kokkos::deep_copy(coordColumnOffsets_d, coordColumnOffsets_h);
+
+    LO countValues = 0;
+    Kokkos::View<impl_scalar_type*> interiorValues_d("interiorValues", 125);
+    auto interiorValues_h = Kokkos::create_mirror_view(interiorValues_d);
+    for (LO k = 0; k < 5; ++k) {
+      for (LO j = 0; j < 5; ++j) {
+        for (LO i = 0; i < 5; ++i) {
+          interiorValues_h(countValues) = coeffs[k] * coeffs[j] * coeffs[i];
+          ++countValues;
         }
       }
-      Kokkos::deep_copy(interiorValues_d, interiorValues_h);
+    }
+    Kokkos::deep_copy(interiorValues_d, interiorValues_h);
 
-      Kokkos::parallel_for("interior idx region R",Kokkos::RangePolicy<typename device_type::execution_space>(0, numInteriors),
+    Kokkos::parallel_for(
+        "interior idx region R", Kokkos::RangePolicy<typename device_type::execution_space>(0, numInteriors),
         KOKKOS_LAMBDA(const LO interiorIdx) {
-            LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
-            LO gridIdx[3];
-            gridIdx[0] = 0; gridIdx[1] = 0; gridIdx[2] = 0;
-            // First step in the loop
-            // update the grid indices
-            // for the grid point
-            for(LO i=0; i<interiorIdx; i++){
-              ++gridIdx[0];
-              if(gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
-                gridIdx[0] = 0;
-                ++gridIdx[1];
-                if(gridIdx[1] == lCoarseNodesPerDim_d(1) - 2) {
-                  gridIdx[1] = 0;
-                  ++gridIdx[2];
-                }
+          LO coordRowIdx, rowIdx, coordColumnOffset, columnOffset, entryOffset;
+          LO gridIdx[3];
+          gridIdx[0] = 0;
+          gridIdx[1] = 0;
+          gridIdx[2] = 0;
+          // First step in the loop
+          // update the grid indices
+          // for the grid point
+          for (LO i = 0; i < interiorIdx; i++) {
+            ++gridIdx[0];
+            if (gridIdx[0] == lCoarseNodesPerDim_d(0) - 2) {
+              gridIdx[0] = 0;
+              ++gridIdx[1];
+              if (gridIdx[1] == lCoarseNodesPerDim_d(1) - 2) {
+                gridIdx[1] = 0;
+                ++gridIdx[2];
               }
             }
+          }
 
-            coordRowIdx = ((gridIdx[2] + 1)*lCoarseNodesPerDim_d(0)*lCoarseNodesPerDim_d(1)
-                   + (gridIdx[1] + 1)*lCoarseNodesPerDim_d(0)
-                   + gridIdx[0] + 1);
-            rowIdx = coordRowIdx*blkSize;
-            coordColumnOffset = ((gridIdx[2] + 1)*3*lFineNodesPerDim_d(1)*lFineNodesPerDim_d(0)
-                     + (gridIdx[1] + 1)*3*lFineNodesPerDim_d(0) + (gridIdx[0] + 1)*3);
-            columnOffset = coordColumnOffset*blkSize;
-            entryOffset = (facePlaneOffset + faceLineOffset + faceStencilLength
-                           + gridIdx[2]*interiorPlaneOffset + gridIdx[1]*interiorLineOffset
-                           + gridIdx[0]*interiorStencilLength)*blkSize;
-            for(LO l = 0; l < blkSize; ++l) {
-              row_map(rowIdx + 1 + l) = entryOffset + interiorStencilLength*(l+1);
-            }
-            // Fill the column indices
-            // and values in the approproate
-            // views.
-            for(LO l = 0; l < blkSize; ++l) {
-              for(LO entryIdx = 0; entryIdx < interiorStencilLength; ++entryIdx) {
-                entries(entryOffset + entryIdx + interiorStencilLength*l) = columnOffset + coordColumnOffsets_d(entryIdx)*blkSize + l;
-                values(entryOffset + entryIdx + interiorStencilLength*l) = interiorValues_d(entryIdx);
-              }
-            }
-            for(int dim = 0; dim <numDimensions; ++dim) {
-              coarseCoordsView(coordRowIdx,dim) = fineCoordsView(coordColumnOffset, dim);
+          coordRowIdx       = ((gridIdx[2] + 1) * lCoarseNodesPerDim_d(0) * lCoarseNodesPerDim_d(1) + (gridIdx[1] + 1) * lCoarseNodesPerDim_d(0) + gridIdx[0] + 1);
+          rowIdx            = coordRowIdx * blkSize;
+          coordColumnOffset = ((gridIdx[2] + 1) * 3 * lFineNodesPerDim_d(1) * lFineNodesPerDim_d(0) + (gridIdx[1] + 1) * 3 * lFineNodesPerDim_d(0) + (gridIdx[0] + 1) * 3);
+          columnOffset      = coordColumnOffset * blkSize;
+          entryOffset       = (facePlaneOffset + faceLineOffset + faceStencilLength + gridIdx[2] * interiorPlaneOffset + gridIdx[1] * interiorLineOffset + gridIdx[0] * interiorStencilLength) * blkSize;
+          for (LO l = 0; l < blkSize; ++l) {
+            row_map(rowIdx + 1 + l) = entryOffset + interiorStencilLength * (l + 1);
+          }
+          // Fill the column indices
+          // and values in the approproate
+          // views.
+          for (LO l = 0; l < blkSize; ++l) {
+            for (LO entryIdx = 0; entryIdx < interiorStencilLength; ++entryIdx) {
+              entries(entryOffset + entryIdx + interiorStencilLength * l) = columnOffset + coordColumnOffsets_d(entryIdx) * blkSize + l;
+              values(entryOffset + entryIdx + interiorStencilLength * l)  = interiorValues_d(entryIdx);
             }
+          }
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            coarseCoordsView(coordRowIdx, dim) = fineCoordsView(coordColumnOffset, dim);
+          }
+        });  // Kokkos::parallel_for interior idx
+    //
+  }
 
-      });// Kokkos::parallel_for interior idx
-      //
-    }
-
-    local_graph_type localGraph(entries, row_map);
-    local_matrix_type localR("R", numCols, values, localGraph);
+  local_graph_type localGraph(entries, row_map);
+  local_matrix_type localR("R", numCols, values, localGraph);
 
-    R = MatrixFactory::Build(localR,            // the local data
-                             rowMap,            // rowMap
-                             A->getRowMap(),    // colMap
-                             A->getRowMap(),    // domainMap == colMap
-                             rowMap,            // rangeMap  == rowMap
-                             Teuchos::null);    // params for optimized construction
+  R = MatrixFactory::Build(localR,          // the local data
+                           rowMap,          // rowMap
+                           A->getRowMap(),  // colMap
+                           A->getRowMap(),  // domainMap == colMap
+                           rowMap,          // rangeMap  == rowMap
+                           Teuchos::null);  // params for optimized construction
 
-  } // Build3D()
+}  // Build3D()
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REGIONRFACTORY_KOKKOS_SHORT
-#endif // MUELU_REGIONRFACTORY_DEF_HPP
+#endif  // MUELU_REGIONRFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_decl.hpp b/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_decl.hpp
index ccc002bffdb4..173637cb81da 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_decl.hpp
@@ -58,23 +58,23 @@
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLuTests {
-  // Forward declaration of friend tester class used to UnitTest CombinePFactory
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CombinePFactoryTester;
-}
+// Forward declaration of friend tester class used to UnitTest CombinePFactory
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CombinePFactoryTester;
+}  // namespace MueLuTests
 
 namespace MueLu {
 
 /*!
   @class CombinePFactory
   @ingroup MueLuTransferClasses
-  @brief Prolongator factory that replicates 'Psubblock' matrix to create new prolongator suitable for PDE systems 
+  @brief Prolongator factory that replicates 'Psubblock' matrix to create new prolongator suitable for PDE systems
 
 
   Takes a set of previously generated prolongators each for a subsystem multiphysics PDE and effectively makes a block diagonal prolongator
   by combining the "combo: npdes" times so that it can be used with a PDE system. A normal use case
-  would be to run an existing MueLu multigrid algorithm on a set of scalar PDEs to generate a hierarchy.  Then use 
-  something like 
+  would be to run an existing MueLu multigrid algorithm on a set of scalar PDEs to generate a hierarchy.  Then use
+  something like
 
   MueLu::HierarchyUtils<SC,LO,GO,NO>::CopyBetweenHierarchies(*scalarHierarchy0,*systemHierarchy, "P",  "Psubblock0", "RCP<Matrix>");
                                                                              .                                   .
@@ -86,7 +86,7 @@ namespace MueLu {
   in its parameter list to then invoke ComboPFactory. Currently, this is used in src/Operators/MueLu_Multiphysics_def.hpp with an example
   in test/multiphysics.
 
-  The end result is a block diagonal matrix corresponding to 
+  The end result is a block diagonal matrix corresponding to
                    [Psubblock0                                                                  ]
                    [           Psubblock1                                                       ]
                    [                      Psubblock2                                            ]
@@ -123,52 +123,50 @@ namespace MueLu {
   | P                 | CombinePFactory          | Prolongator                                                                                                      |
 
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class CombinePFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CombinePFactory : public PFactory {
 #undef MUELU_COMBINEPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  friend class MueLuTests::CombinePFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    friend class MueLuTests::CombinePFactoryTester<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  CombinePFactory() {}
 
-    //! Constructor
-    CombinePFactory() { }
+  //! Destructor.
+  virtual ~CombinePFactory() {}
+  //@}
 
-    //! Destructor.
-    virtual ~CombinePFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+ private:
+  int numPDEs_;
 
-  private:
+};  // class CombinePFactory
 
-    int numPDEs_;
-
-  }; //class CombinePFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COMBINEPFACTORY_SHORT
-#endif // MUELU_COMBINEPFACTORY_DECL_HPP
+#endif  // MUELU_COMBINEPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_def.hpp b/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_def.hpp
index e5c07f18713d..fa8e2f949af3 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_CombinePFactory_def.hpp
@@ -49,7 +49,6 @@
 #include <stdlib.h>
 #include <iomanip>
 
-
 // #include <Teuchos_LAPACK.hpp>
 #include <Teuchos_SerialDenseMatrix.hpp>
 #include <Teuchos_SerialDenseVector.hpp>
@@ -72,200 +71,199 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->setEntry("combine: numBlks",ParameterEntry(1));
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-//    Input(fineLevel, "subblock");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
-                                Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
-                                Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    const ParameterList& pL = GetParameterList();
-    const LO nBlks = as<LO>(pL.get<int> ("combine: numBlks"));
-
-    RCP<Matrix>                A             = Get< RCP<Matrix> >               (fineLevel, "A");
-
-    // Record all matrices that each define a block in block diagonal comboP
-    // matrix used for PDE/multiblock interpolation.  Additionally, count 
-    // total number of local rows, nonzeros, coarseDofs, and colDofs.
-
-    Teuchos::ArrayRCP<RCP<Matrix>>       arrayOfMatrices(nBlks);
-    size_t nComboRowMap = 0, nnzCombo = 0,  nComboColMap = 0, nComboDomMap = 0;
-
-    LO nTotalNumberLocalColMapEntries = 0;
-    Teuchos::ArrayRCP<size_t> DomMapSizePerBlk(nBlks);
-    Teuchos::ArrayRCP<size_t> ColMapSizePerBlk(nBlks);
-    Teuchos::ArrayRCP<size_t> ColMapLocalSizePerBlk(nBlks);
-    Teuchos::ArrayRCP<size_t> ColMapRemoteSizePerBlk(nBlks);
-    Teuchos::ArrayRCP<size_t> ColMapLocalCumulativePerBlk(nBlks+1);   // hardwire 0th entry so that it has the value of 0
-    Teuchos::ArrayRCP<size_t> ColMapRemoteCumulativePerBlk(nBlks+1);  // hardwire 0th entry so that it has the value of 0
-    for(int j = 0; j < nBlks; j++) {
-      std::string blockName = "Psubblock" + Teuchos::toString(j);
-      if (coarseLevel.IsAvailable(blockName, NoFactory::get())) {
-        arrayOfMatrices[j] = coarseLevel.Get< RCP<Matrix> >(blockName, NoFactory::get());
-        nComboRowMap       += Teuchos::as<size_t>((arrayOfMatrices[j])->getRowMap()->getLocalNumElements());
-        DomMapSizePerBlk[j] = Teuchos::as<size_t>((arrayOfMatrices[j])->getDomainMap()->getLocalNumElements());
-        ColMapSizePerBlk[j] = Teuchos::as<size_t>((arrayOfMatrices[j])->getColMap()->getLocalNumElements());
-        nComboDomMap       += DomMapSizePerBlk[j];
-        nComboColMap       += ColMapSizePerBlk[j];
-        nnzCombo         += Teuchos::as<size_t>((arrayOfMatrices[j])->getLocalNumEntries());
-        TEUCHOS_TEST_FOR_EXCEPTION((arrayOfMatrices[j])->getDomainMap()->getIndexBase() != 0, Exceptions::RuntimeError, "interpolation subblocks must use 0 indexbase");
-
-
-        // figure out how many empty entries in each column map
-        int tempii = 0;
-        for (int i = 0; i < (int) DomMapSizePerBlk[j]; i++ ) {
-//          if ( (arrayOfMatrices[j])->getDomainMap()->getGlobalElement(i) == (arrayOfMatrices[j])->getColMap()->getGlobalElement(tempii) )  nTotalNumberLocalColMapEntries++;
-          if ( (arrayOfMatrices[j])->getDomainMap()->getGlobalElement(i) == (arrayOfMatrices[j])->getColMap()->getGlobalElement(tempii) )  tempii++;
-        }
-        nTotalNumberLocalColMapEntries += tempii;
-        ColMapLocalSizePerBlk[j] = tempii;
-        ColMapRemoteSizePerBlk[j] = ColMapSizePerBlk[j] - ColMapLocalSizePerBlk[j];
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->setEntry("combine: numBlks", ParameterEntry(1));
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  //    Input(fineLevel, "subblock");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
+                                                                       Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CombinePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
+                                                                        Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  const ParameterList& pL = GetParameterList();
+  const LO nBlks          = as<LO>(pL.get<int>("combine: numBlks"));
+
+  RCP<Matrix> A = Get<RCP<Matrix>>(fineLevel, "A");
+
+  // Record all matrices that each define a block in block diagonal comboP
+  // matrix used for PDE/multiblock interpolation.  Additionally, count
+  // total number of local rows, nonzeros, coarseDofs, and colDofs.
+
+  Teuchos::ArrayRCP<RCP<Matrix>> arrayOfMatrices(nBlks);
+  size_t nComboRowMap = 0, nnzCombo = 0, nComboColMap = 0, nComboDomMap = 0;
+
+  LO nTotalNumberLocalColMapEntries = 0;
+  Teuchos::ArrayRCP<size_t> DomMapSizePerBlk(nBlks);
+  Teuchos::ArrayRCP<size_t> ColMapSizePerBlk(nBlks);
+  Teuchos::ArrayRCP<size_t> ColMapLocalSizePerBlk(nBlks);
+  Teuchos::ArrayRCP<size_t> ColMapRemoteSizePerBlk(nBlks);
+  Teuchos::ArrayRCP<size_t> ColMapLocalCumulativePerBlk(nBlks + 1);   // hardwire 0th entry so that it has the value of 0
+  Teuchos::ArrayRCP<size_t> ColMapRemoteCumulativePerBlk(nBlks + 1);  // hardwire 0th entry so that it has the value of 0
+  for (int j = 0; j < nBlks; j++) {
+    std::string blockName = "Psubblock" + Teuchos::toString(j);
+    if (coarseLevel.IsAvailable(blockName, NoFactory::get())) {
+      arrayOfMatrices[j] = coarseLevel.Get<RCP<Matrix>>(blockName, NoFactory::get());
+      nComboRowMap += Teuchos::as<size_t>((arrayOfMatrices[j])->getRowMap()->getLocalNumElements());
+      DomMapSizePerBlk[j] = Teuchos::as<size_t>((arrayOfMatrices[j])->getDomainMap()->getLocalNumElements());
+      ColMapSizePerBlk[j] = Teuchos::as<size_t>((arrayOfMatrices[j])->getColMap()->getLocalNumElements());
+      nComboDomMap += DomMapSizePerBlk[j];
+      nComboColMap += ColMapSizePerBlk[j];
+      nnzCombo += Teuchos::as<size_t>((arrayOfMatrices[j])->getLocalNumEntries());
+      TEUCHOS_TEST_FOR_EXCEPTION((arrayOfMatrices[j])->getDomainMap()->getIndexBase() != 0, Exceptions::RuntimeError, "interpolation subblocks must use 0 indexbase");
+
+      // figure out how many empty entries in each column map
+      int tempii = 0;
+      for (int i = 0; i < (int)DomMapSizePerBlk[j]; i++) {
+        //          if ( (arrayOfMatrices[j])->getDomainMap()->getGlobalElement(i) == (arrayOfMatrices[j])->getColMap()->getGlobalElement(tempii) )  nTotalNumberLocalColMapEntries++;
+        if ((arrayOfMatrices[j])->getDomainMap()->getGlobalElement(i) == (arrayOfMatrices[j])->getColMap()->getGlobalElement(tempii)) tempii++;
       }
-      else { 
-        arrayOfMatrices[j] = Teuchos::null;
-        ColMapLocalSizePerBlk[j] = 0;
-        ColMapRemoteSizePerBlk[j] = 0;
-      }
-      ColMapLocalCumulativePerBlk[j+1] = ColMapLocalSizePerBlk[j] + ColMapLocalCumulativePerBlk[j];
-      ColMapRemoteCumulativePerBlk[j+1]= ColMapRemoteSizePerBlk[j] + ColMapRemoteCumulativePerBlk[j];
+      nTotalNumberLocalColMapEntries += tempii;
+      ColMapLocalSizePerBlk[j]  = tempii;
+      ColMapRemoteSizePerBlk[j] = ColMapSizePerBlk[j] - ColMapLocalSizePerBlk[j];
+    } else {
+      arrayOfMatrices[j]        = Teuchos::null;
+      ColMapLocalSizePerBlk[j]  = 0;
+      ColMapRemoteSizePerBlk[j] = 0;
     }
-    TEUCHOS_TEST_FOR_EXCEPTION(nComboRowMap != A->getRowMap()->getLocalNumElements(), Exceptions::RuntimeError, "sum of subblock rows != #row's Afine");
-
-    // build up csr arrays for combo block diagonal P
-    Teuchos::ArrayRCP<size_t> comboPRowPtr(nComboRowMap+1);
-    Teuchos::ArrayRCP<LocalOrdinal> comboPCols(nnzCombo);
-    Teuchos::ArrayRCP<Scalar>       comboPVals(nnzCombo);
-
-    size_t nnzCnt = 0, nrowCntFromPrevBlks = 0,ncolCntFromPrevBlks = 0;
-    LO maxNzPerRow = 0;
-    for(int j = 0; j < nBlks; j++) {
-      // grab csr pointers for individual blocks of P
-      if  (arrayOfMatrices[j] != Teuchos::null) {
-        Teuchos::ArrayRCP<const size_t>     subblockRowPtr((arrayOfMatrices[j])->getLocalNumRows());
-        Teuchos::ArrayRCP<const LocalOrdinal> subblockCols((arrayOfMatrices[j])->getLocalNumEntries());
-        Teuchos::ArrayRCP<const Scalar>         subblockVals((arrayOfMatrices[j])->getLocalNumEntries());
-        if ( (int) (arrayOfMatrices[j])->getLocalMaxNumRowEntries() > maxNzPerRow) maxNzPerRow = (int) (arrayOfMatrices[j])->getLocalMaxNumRowEntries();
-        Teuchos::RCP<CrsMatrixWrap> subblockwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>((arrayOfMatrices[j]));
-        Teuchos::RCP<CrsMatrix>     subblockcrs  = subblockwrap->getCrsMatrix();
-        subblockcrs->getAllValues(subblockRowPtr, subblockCols, subblockVals);
-
-        // copy jth block into csr arrays of comboP 
-
-        for (decltype(subblockRowPtr.size()) i = 0; i < subblockRowPtr.size() - 1; i++) {
-          size_t rowLength = subblockRowPtr[i+1] - subblockRowPtr[i];
-          comboPRowPtr[ nrowCntFromPrevBlks+i ] = nnzCnt;
-          for (size_t k = 0; k < rowLength; k++) {
-            if  ( (int) subblockCols[k+subblockRowPtr[i]] < (int) ColMapLocalSizePerBlk[j]) {
-              comboPCols[nnzCnt  ] = subblockCols[k+subblockRowPtr[i]] + ColMapLocalCumulativePerBlk[j];
-              if ( (int) comboPCols[nnzCnt] >= (int)  ColMapLocalCumulativePerBlk[nBlks]) { printf("ERROR1\n"); exit(1); }
+    ColMapLocalCumulativePerBlk[j + 1]  = ColMapLocalSizePerBlk[j] + ColMapLocalCumulativePerBlk[j];
+    ColMapRemoteCumulativePerBlk[j + 1] = ColMapRemoteSizePerBlk[j] + ColMapRemoteCumulativePerBlk[j];
+  }
+  TEUCHOS_TEST_FOR_EXCEPTION(nComboRowMap != A->getRowMap()->getLocalNumElements(), Exceptions::RuntimeError, "sum of subblock rows != #row's Afine");
+
+  // build up csr arrays for combo block diagonal P
+  Teuchos::ArrayRCP<size_t> comboPRowPtr(nComboRowMap + 1);
+  Teuchos::ArrayRCP<LocalOrdinal> comboPCols(nnzCombo);
+  Teuchos::ArrayRCP<Scalar> comboPVals(nnzCombo);
+
+  size_t nnzCnt = 0, nrowCntFromPrevBlks = 0, ncolCntFromPrevBlks = 0;
+  LO maxNzPerRow = 0;
+  for (int j = 0; j < nBlks; j++) {
+    // grab csr pointers for individual blocks of P
+    if (arrayOfMatrices[j] != Teuchos::null) {
+      Teuchos::ArrayRCP<const size_t> subblockRowPtr((arrayOfMatrices[j])->getLocalNumRows());
+      Teuchos::ArrayRCP<const LocalOrdinal> subblockCols((arrayOfMatrices[j])->getLocalNumEntries());
+      Teuchos::ArrayRCP<const Scalar> subblockVals((arrayOfMatrices[j])->getLocalNumEntries());
+      if ((int)(arrayOfMatrices[j])->getLocalMaxNumRowEntries() > maxNzPerRow) maxNzPerRow = (int)(arrayOfMatrices[j])->getLocalMaxNumRowEntries();
+      Teuchos::RCP<CrsMatrixWrap> subblockwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>((arrayOfMatrices[j]));
+      Teuchos::RCP<CrsMatrix> subblockcrs      = subblockwrap->getCrsMatrix();
+      subblockcrs->getAllValues(subblockRowPtr, subblockCols, subblockVals);
+
+      // copy jth block into csr arrays of comboP
+
+      for (decltype(subblockRowPtr.size()) i = 0; i < subblockRowPtr.size() - 1; i++) {
+        size_t rowLength                      = subblockRowPtr[i + 1] - subblockRowPtr[i];
+        comboPRowPtr[nrowCntFromPrevBlks + i] = nnzCnt;
+        for (size_t k = 0; k < rowLength; k++) {
+          if ((int)subblockCols[k + subblockRowPtr[i]] < (int)ColMapLocalSizePerBlk[j]) {
+            comboPCols[nnzCnt] = subblockCols[k + subblockRowPtr[i]] + ColMapLocalCumulativePerBlk[j];
+            if ((int)comboPCols[nnzCnt] >= (int)ColMapLocalCumulativePerBlk[nBlks]) {
+              printf("ERROR1\n");
+              exit(1);
             }
-            else {
-                   // Here we subtract off the number of local colmap guys ... so this tell us where we are among ghost unknowns. We then want to stick this ghost after
-                   // all the Local guys ... so we add ColMapLocalCumulativePerBlk[nBlks] .... finally we need to account for the fact that other ghost blocks may have already
-                   // been handled ... so we then add  + ColMapRemoteCumulativePerBlk[j];
-                   comboPCols[nnzCnt  ] = subblockCols[k+subblockRowPtr[i]] -  ColMapLocalSizePerBlk[j] +  ColMapLocalCumulativePerBlk[nBlks] + ColMapRemoteCumulativePerBlk[j];
-                  if ( (int) comboPCols[nnzCnt] >= (int)  (ColMapLocalCumulativePerBlk[nBlks]+ ColMapRemoteCumulativePerBlk[nBlks])) { printf("ERROR2\n"); exit(1); }
+          } else {
+            // Here we subtract off the number of local colmap guys ... so this tell us where we are among ghost unknowns. We then want to stick this ghost after
+            // all the Local guys ... so we add ColMapLocalCumulativePerBlk[nBlks] .... finally we need to account for the fact that other ghost blocks may have already
+            // been handled ... so we then add  + ColMapRemoteCumulativePerBlk[j];
+            comboPCols[nnzCnt] = subblockCols[k + subblockRowPtr[i]] - ColMapLocalSizePerBlk[j] + ColMapLocalCumulativePerBlk[nBlks] + ColMapRemoteCumulativePerBlk[j];
+            if ((int)comboPCols[nnzCnt] >= (int)(ColMapLocalCumulativePerBlk[nBlks] + ColMapRemoteCumulativePerBlk[nBlks])) {
+              printf("ERROR2\n");
+              exit(1);
             }
-            comboPVals[nnzCnt++] = subblockVals[k+subblockRowPtr[i]];
           }
+          comboPVals[nnzCnt++] = subblockVals[k + subblockRowPtr[i]];
         }
-
-        nrowCntFromPrevBlks += Teuchos::as<size_t>((arrayOfMatrices[j])->getRowMap()->getLocalNumElements());
-        ncolCntFromPrevBlks += DomMapSizePerBlk[j]; // rst: check this
       }
+
+      nrowCntFromPrevBlks += Teuchos::as<size_t>((arrayOfMatrices[j])->getRowMap()->getLocalNumElements());
+      ncolCntFromPrevBlks += DomMapSizePerBlk[j];  // rst: check this
     }
-    comboPRowPtr[nComboRowMap] = nnzCnt; 
-
-
-    // Come up with global IDs for the coarse grid maps. We assume that each xxx
-    // block has a minimum GID of 0.  Since MueLu is generally creating these 
-    // GIDS, this assumption is probably correct, but we'll check it.
-
-
-    Teuchos::Array<GlobalOrdinal> comboDomainMapGIDs(nComboDomMap);
-    Teuchos::Array<GlobalOrdinal> comboColMapGIDs(nComboColMap);
-
-    LO nTotalNumberRemoteColMapEntries = 0;
-    GlobalOrdinal offset = 0; 
-    size_t        domainMapIndex = 0;
-        int nComboColIndex =  0;
-    for(int j = 0; j < nBlks; j++) {
-      int nThisBlkColIndex =  0;
-      
-      GlobalOrdinal tempMax = 0, maxGID = 0;
-      if  (arrayOfMatrices[j] != Teuchos::null) tempMax = (arrayOfMatrices[j])->getDomainMap()->getMaxGlobalIndex();
-      Teuchos::reduceAll( *(A->getDomainMap()->getComm()), Teuchos::REDUCE_MAX, tempMax, Teuchos::ptr(&maxGID));
-
-      if  (arrayOfMatrices[j] != Teuchos::null) {
-        TEUCHOS_TEST_FOR_EXCEPTION(arrayOfMatrices[j]->getDomainMap()->getMinAllGlobalIndex() < 0,Exceptions::RuntimeError, "Global ID assumption for domainMap not met within subblock");
-
-        GO  priorDomGID = 0;
-        for(size_t c = 0; c < DomMapSizePerBlk[j]; ++c) { // check this
-           //  The global ids of jth block are assumed to go between 0 and maxGID_j.  So the 1th blocks DomainGIDs should start at maxGID_0+1. The 2nd 
-           //  block DomainDIGS starts at maxGID_0+1 + maxGID_1 + 1. We use offset to keep track of these starting points.
-          priorDomGID =  (arrayOfMatrices[j])->getDomainMap()->getGlobalElement(c);
-          comboDomainMapGIDs[domainMapIndex++] =  offset + priorDomGID;
-          if ( priorDomGID == (arrayOfMatrices[j])->getColMap()->getGlobalElement(nThisBlkColIndex) )  {
-            comboColMapGIDs[   nComboColIndex++ ] =  offset + priorDomGID;
-            nThisBlkColIndex++;
-          }
-        }
-        
-        for(size_t cc = nThisBlkColIndex; cc <  ColMapSizePerBlk[j]; ++cc) {
-          comboColMapGIDs[nTotalNumberLocalColMapEntries + nTotalNumberRemoteColMapEntries++ ] =  offset + (arrayOfMatrices[j])->getColMap()->getGlobalElement(cc);
+  }
+  comboPRowPtr[nComboRowMap] = nnzCnt;
+
+  // Come up with global IDs for the coarse grid maps. We assume that each xxx
+  // block has a minimum GID of 0.  Since MueLu is generally creating these
+  // GIDS, this assumption is probably correct, but we'll check it.
+
+  Teuchos::Array<GlobalOrdinal> comboDomainMapGIDs(nComboDomMap);
+  Teuchos::Array<GlobalOrdinal> comboColMapGIDs(nComboColMap);
+
+  LO nTotalNumberRemoteColMapEntries = 0;
+  GlobalOrdinal offset               = 0;
+  size_t domainMapIndex              = 0;
+  int nComboColIndex                 = 0;
+  for (int j = 0; j < nBlks; j++) {
+    int nThisBlkColIndex = 0;
+
+    GlobalOrdinal tempMax = 0, maxGID = 0;
+    if (arrayOfMatrices[j] != Teuchos::null) tempMax = (arrayOfMatrices[j])->getDomainMap()->getMaxGlobalIndex();
+    Teuchos::reduceAll(*(A->getDomainMap()->getComm()), Teuchos::REDUCE_MAX, tempMax, Teuchos::ptr(&maxGID));
+
+    if (arrayOfMatrices[j] != Teuchos::null) {
+      TEUCHOS_TEST_FOR_EXCEPTION(arrayOfMatrices[j]->getDomainMap()->getMinAllGlobalIndex() < 0, Exceptions::RuntimeError, "Global ID assumption for domainMap not met within subblock");
+
+      GO priorDomGID = 0;
+      for (size_t c = 0; c < DomMapSizePerBlk[j]; ++c) {  // check this
+                                                          //  The global ids of jth block are assumed to go between 0 and maxGID_j.  So the 1th blocks DomainGIDs should start at maxGID_0+1. The 2nd
+                                                          //  block DomainDIGS starts at maxGID_0+1 + maxGID_1 + 1. We use offset to keep track of these starting points.
+        priorDomGID                          = (arrayOfMatrices[j])->getDomainMap()->getGlobalElement(c);
+        comboDomainMapGIDs[domainMapIndex++] = offset + priorDomGID;
+        if (priorDomGID == (arrayOfMatrices[j])->getColMap()->getGlobalElement(nThisBlkColIndex)) {
+          comboColMapGIDs[nComboColIndex++] = offset + priorDomGID;
+          nThisBlkColIndex++;
         }
       }
-      offset += maxGID+1;
+
+      for (size_t cc = nThisBlkColIndex; cc < ColMapSizePerBlk[j]; ++cc) {
+        comboColMapGIDs[nTotalNumberLocalColMapEntries + nTotalNumberRemoteColMapEntries++] = offset + (arrayOfMatrices[j])->getColMap()->getGlobalElement(cc);
+      }
     }
+    offset += maxGID + 1;
+  }
 #ifdef out
-    RCP<const Map> coarseDomainMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getDomainMap()->lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),comboDomainMapGIDs, 0, A->getDomainMap()->getComm());
-    RCP<const Map> coarseColMap    = Xpetra::MapFactory<LO,GO,NO>::Build(A->getDomainMap()->lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),comboColMapGIDs,    0, A->getDomainMap()->getComm());
+  RCP<const Map> coarseDomainMap = Xpetra::MapFactory<LO, GO, NO>::Build(A->getDomainMap()->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), comboDomainMapGIDs, 0, A->getDomainMap()->getComm());
+  RCP<const Map> coarseColMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getDomainMap()->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), comboColMapGIDs, 0, A->getDomainMap()->getComm());
 #endif
 
-    RCP<const Map> coarseDomainMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getDomainMap()->lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),comboDomainMapGIDs, 0, A->getRowMap()->getComm());
-    RCP<const Map> coarseColMap    = Xpetra::MapFactory<LO,GO,NO>::Build(A->getDomainMap()->lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),comboColMapGIDs,    0, A->getRowMap()->getComm());
-
-
-    Teuchos::RCP<CrsMatrix> comboPCrs = CrsMatrixFactory::Build(A->getRowMap(), coarseColMap,maxNzPerRow);
-//comboPCrs->getCrsGraph(); //.getRowInfo(6122);
-//comboPCrs->getRowInfo(6122);
+  RCP<const Map> coarseDomainMap = Xpetra::MapFactory<LO, GO, NO>::Build(A->getDomainMap()->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), comboDomainMapGIDs, 0, A->getRowMap()->getComm());
+  RCP<const Map> coarseColMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getDomainMap()->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), comboColMapGIDs, 0, A->getRowMap()->getComm());
 
-//    Teuchos::RCP<CrsMatrix> comboPCrs = CrsMatrixFactory::Build(A->getRowMap(), coarseColMap,nnzCombo+1000);
+  Teuchos::RCP<CrsMatrix> comboPCrs = CrsMatrixFactory::Build(A->getRowMap(), coarseColMap, maxNzPerRow);
+  // comboPCrs->getCrsGraph(); //.getRowInfo(6122);
+  // comboPCrs->getRowInfo(6122);
 
-//    for (size_t i = 0; i < nComboRowMap; i++) {
-//printf("FIXME\n"); if (nComboRowMap > 6142)  nComboRowMap = 6142;
-    for (size_t i = 0; i < nComboRowMap; i++) {
-      comboPCrs->insertLocalValues(i, comboPCols.view(comboPRowPtr[i], comboPRowPtr[i+1] - comboPRowPtr[i]),
-                                      comboPVals.view(comboPRowPtr[i], comboPRowPtr[i+1] - comboPRowPtr[i]));
-    }
-    comboPCrs->fillComplete(coarseDomainMap, A->getRowMap());
-
-    Teuchos::RCP<Matrix> comboP = Teuchos::rcp(new CrsMatrixWrap(comboPCrs));
+  //    Teuchos::RCP<CrsMatrix> comboPCrs = CrsMatrixFactory::Build(A->getRowMap(), coarseColMap,nnzCombo+1000);
 
-    Set(coarseLevel, "P", comboP);
+  //    for (size_t i = 0; i < nComboRowMap; i++) {
+  // printf("FIXME\n"); if (nComboRowMap > 6142)  nComboRowMap = 6142;
+  for (size_t i = 0; i < nComboRowMap; i++) {
+    comboPCrs->insertLocalValues(i, comboPCols.view(comboPRowPtr[i], comboPRowPtr[i + 1] - comboPRowPtr[i]),
+                                 comboPVals.view(comboPRowPtr[i], comboPRowPtr[i + 1] - comboPRowPtr[i]));
   }
+  comboPCrs->fillComplete(coarseDomainMap, A->getRowMap());
+
+  Teuchos::RCP<Matrix> comboP = Teuchos::rcp(new CrsMatrixWrap(comboPCrs));
 
+  Set(coarseLevel, "P", comboP);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COMBINEPFACTORY_SHORT
-#endif // MUELU_COMBINEPFACTORY_DEF_HPP
+#endif  // MUELU_COMBINEPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_decl.hpp b/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_decl.hpp
index eca87aadaba8..193c40caf8e9 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_decl.hpp
@@ -62,48 +62,48 @@
 
 namespace MueLu {
 
-  /*!
-    @class GenericRFactory class.
-    @brief Factory for building restriction operators using a prolongator factory
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class GenericRFactory : public TwoLevelFactoryBase {
+/*!
+  @class GenericRFactory class.
+  @brief Factory for building restriction operators using a prolongator factory
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GenericRFactory : public TwoLevelFactoryBase {
 #undef MUELU_GENERICRFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    GenericRFactory() { }
+  //! Constructor.
+  GenericRFactory() {}
 
-    //! Destructor.
-    virtual ~GenericRFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~GenericRFactory() {}
+  //@}
 
-    //! Input
-    //@{
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build(Level & fineLevel, Level & coarseLevel) const;
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; //class GenericRFactory
+};  // class GenericRFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GENERICRFACTORY_SHORT
-#endif // MUELU_GENERICRFACTORY_DECL_HPP
+#endif  // MUELU_GENERICRFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_def.hpp b/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_def.hpp
index ac47b57f8f34..b32acb7db9ac 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_GenericRFactory_def.hpp
@@ -58,59 +58,65 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix P");
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix P");
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level & /* fineLevel */, Level &coarseLevel) const {
+  RCP<const FactoryBase> PFact1 = GetFactory("P");
+  if (PFact1 == Teuchos::null) {
+    PFact1 = coarseLevel.GetFactoryManager()->GetFactory("P");
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &/* fineLevel */, Level &coarseLevel) const {
-    RCP<const FactoryBase> PFact1 = GetFactory("P");
-    if (PFact1 == Teuchos::null) { PFact1 = coarseLevel.GetFactoryManager()->GetFactory("P"); }
-    RCP<PFactory> PFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(PFact1));;
-
-    bool rmode = PFact->isRestrictionModeSet();
-    PFact->setRestrictionMode(true);             // set restriction mode
-
-    // Force request call for PFact
-    // In general, Request is only called once for each factory, since we
-    // can reuse data generated by the factory. However, here we have to
-    // run the code in PFact.Build again, so we have to request the
-    // dependencies of PFact first!  Thedependencies are (automatically)
-    // cleaned up after the second run of PFact.Build in
-    // coarseLevel.Get<RCP<Matrix> >("R",PFact.get())!
-    coarseLevel.DeclareDependencies(PFact.get());
-
-    coarseLevel.DeclareInput("R", PFact.get(), this);  // we expect the prolongation operator factory to produce "R" as output
-    // call declareInput is called within DeclareInput call
-    PFact->setRestrictionMode(rmode);            // reset restriciton mode flag
+  RCP<PFactory> PFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(PFact1));
+  ;
+
+  bool rmode = PFact->isRestrictionModeSet();
+  PFact->setRestrictionMode(true);  // set restriction mode
+
+  // Force request call for PFact
+  // In general, Request is only called once for each factory, since we
+  // can reuse data generated by the factory. However, here we have to
+  // run the code in PFact.Build again, so we have to request the
+  // dependencies of PFact first!  Thedependencies are (automatically)
+  // cleaned up after the second run of PFact.Build in
+  // coarseLevel.Get<RCP<Matrix> >("R",PFact.get())!
+  coarseLevel.DeclareDependencies(PFact.get());
+
+  coarseLevel.DeclareInput("R", PFact.get(), this);  // we expect the prolongation operator factory to produce "R" as output
+  // call declareInput is called within DeclareInput call
+  PFact->setRestrictionMode(rmode);  // reset restriciton mode flag
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Call prolongator factory for calculating restrictor", coarseLevel);
+
+  RCP<const FactoryBase> PFact1 = GetFactory("P");
+  if (PFact1 == Teuchos::null) {
+    PFact1 = coarseLevel.GetFactoryManager()->GetFactory("P");
   }
+  RCP<PFactory> PFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(PFact1));
+  ;
+  MueLu::DisableMultipleCallCheck check(PFact);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GenericRFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level & coarseLevel) const {
-    FactoryMonitor m(*this, "Call prolongator factory for calculating restrictor", coarseLevel);
-
-    RCP<const FactoryBase> PFact1 = GetFactory("P");
-    if (PFact1 == Teuchos::null) { PFact1 = coarseLevel.GetFactoryManager()->GetFactory("P"); }
-    RCP<PFactory> PFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(PFact1));;
-    MueLu::DisableMultipleCallCheck check(PFact);
-
-    // BuildR
-    bool rmode = PFact->isRestrictionModeSet();
-    PFact->setRestrictionMode(true);     // switch prolongator factory to restriction mode
+  // BuildR
+  bool rmode = PFact->isRestrictionModeSet();
+  PFact->setRestrictionMode(true);  // switch prolongator factory to restriction mode
 
-    //PFact->Build(fineLevel, coarseLevel);  // call PFactory::Build explicitely
-    RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R",PFact.get());
+  // PFact->Build(fineLevel, coarseLevel);  // call PFactory::Build explicitely
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", PFact.get());
 
-    PFact->setRestrictionMode(rmode);    // reset restriction mode flag
+  PFact->setRestrictionMode(rmode);  // reset restriction mode flag
 
-    Set(coarseLevel, "R", R);
+  Set(coarseLevel, "R", R);
 
-  } //BuildR
+}  // BuildR
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GENERICRFACTORY_SHORT
-#endif // MUELU_GENERICRFACTORY_DEF_HPP
+#endif  // MUELU_GENERICRFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_decl.hpp b/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_decl.hpp
index 1d13c2abe4b8..229fa352c784 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_decl.hpp
@@ -57,29 +57,29 @@
 #include "MueLu_Level_fwd.hpp"
 
 namespace MueLuTests {
-  // Forward declaration of friend tester class used to UnitTest ReplicatePFactory
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ReplicatePFactoryTester;
-}
+// Forward declaration of friend tester class used to UnitTest ReplicatePFactory
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ReplicatePFactoryTester;
+}  // namespace MueLuTests
 
 namespace MueLu {
 
 /*!
   @class ReplicatePFactory
   @ingroup MueLuTransferClasses
-  @brief Prolongator factory that replicates 'Psubblock' matrix to create new prolongator suitable for PDE systems 
+  @brief Prolongator factory that replicates 'Psubblock' matrix to create new prolongator suitable for PDE systems
 
 
   Takes a previously generated prolongator for a scalar PDE and effectively makes a block diagonal prolongator
   by replicating it "replicate: npdes" times so that it can be used with a PDE system. A normal use case
-  would be to run an existing MueLu multigrid algorithm on a scalar PDE to generate a hierarchy.  Then use 
-  something like 
+  would be to run an existing MueLu multigrid algorithm on a scalar PDE to generate a hierarchy.  Then use
+  something like
 
   MueLu::HierarchyUtils<SC,LO,GO,NO>::CopyBetweenHierarchies(*scalarHierarchy,*systemHierarchy, "P",  "Psubblock", "RCP<Matrix>");
 
   to copy it to the new hierarchy. This new hierarchy would then have <Parameter name="multigrid algorithm" type="string"  value="replicate"/>
   in its parameter list to then invoke RepliatePFactory. Currently, this is used in src/Operators/MueLu_Maxwell1_def.hpp with an example
-  in test/maxwell. 
+  in test/maxwell.
 
   ## Input/output of ReplicatePFactory ##
 
@@ -104,52 +104,50 @@ namespace MueLu {
   | P                 | ReplicatePFactory | Prolongator                                                                                                      |
 
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class ReplicatePFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class ReplicatePFactory : public PFactory {
 #undef MUELU_REPLICATEPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  friend class MueLuTests::ReplicatePFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    friend class MueLuTests::ReplicatePFactoryTester<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  ReplicatePFactory() {}
 
-    //! Constructor
-    ReplicatePFactory() { }
+  //! Destructor.
+  virtual ~ReplicatePFactory() {}
+  //@}
 
-    //! Destructor.
-    virtual ~ReplicatePFactory() { }
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-    //@}
+ private:
+  int numPDEs_;
 
-  private:
+};  // class ReplicatePFactory
 
-    int numPDEs_;
-
-  }; //class ReplicatePFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPLICATEPFACTORY_SHORT
-#endif // MUELU_REPLICATEPFACTORY_DECL_HPP
+#endif  // MUELU_REPLICATEPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_def.hpp b/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_def.hpp
index 09d515c8e1a6..3e757bd27ad0 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_ReplicatePFactory_def.hpp
@@ -49,7 +49,6 @@
 #include <stdlib.h>
 #include <iomanip>
 
-
 // #include <Teuchos_LAPACK.hpp>
 #include <Teuchos_SerialDenseMatrix.hpp>
 #include <Teuchos_SerialDenseVector.hpp>
@@ -70,117 +69,116 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->setEntry("replicate: npdes",ParameterEntry(1));
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-//    Input(fineLevel, "Psubblock");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
-                                Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
-                                Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    RCP<Matrix>      Psubblock             = coarseLevel.Get< RCP<Matrix> >("Psubblock", NoFactory::get());
-    const ParameterList& pL = GetParameterList();
-    const LO dofPerNode = as<LO>(pL.get<int>   ("replicate: npdes"));
-
-    Teuchos::ArrayRCP<const size_t> amalgRowPtr(Psubblock->getLocalNumRows());
-    Teuchos::ArrayRCP<const LocalOrdinal> amalgCols(Psubblock->getLocalNumEntries());
-    Teuchos::ArrayRCP<const Scalar> amalgVals(Psubblock->getLocalNumEntries());
-    Teuchos::RCP<CrsMatrixWrap> Psubblockwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(Psubblock);
-    Teuchos::RCP<CrsMatrix> Psubblockcrs = Psubblockwrap->getCrsMatrix();
-    Psubblockcrs->getAllValues(amalgRowPtr, amalgCols, amalgVals);
-
-    size_t paddedNrows = Psubblock->getRowMap()->getLocalNumElements() * Teuchos::as<size_t>(dofPerNode);
-    Teuchos::ArrayRCP<size_t> newPRowPtr(paddedNrows+1);
-    Teuchos::ArrayRCP<LocalOrdinal> newPCols(Psubblock->getLocalNumEntries() * dofPerNode);
-    Teuchos::ArrayRCP<Scalar> newPVals(Psubblock->getLocalNumEntries() * dofPerNode);
-    size_t cnt = 0; // local id counter
-    for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
-      size_t rowLength = amalgRowPtr[i+1] - amalgRowPtr[i];
-      for(int j = 0; j < dofPerNode; j++) {
-        newPRowPtr[i*dofPerNode+j] = cnt;
-        for (size_t k = 0; k < rowLength; k++) {
-          newPCols[cnt  ] = amalgCols[k+amalgRowPtr[i]] * dofPerNode + j;
-          newPVals[cnt++] = amalgVals[k+amalgRowPtr[i]];
-        }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->setEntry("replicate: npdes", ParameterEntry(1));
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  //    Input(fineLevel, "Psubblock");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel,
+                                                                         Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReplicatePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel,
+                                                                          Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  RCP<Matrix> Psubblock   = coarseLevel.Get<RCP<Matrix> >("Psubblock", NoFactory::get());
+  const ParameterList& pL = GetParameterList();
+  const LO dofPerNode     = as<LO>(pL.get<int>("replicate: npdes"));
+
+  Teuchos::ArrayRCP<const size_t> amalgRowPtr(Psubblock->getLocalNumRows());
+  Teuchos::ArrayRCP<const LocalOrdinal> amalgCols(Psubblock->getLocalNumEntries());
+  Teuchos::ArrayRCP<const Scalar> amalgVals(Psubblock->getLocalNumEntries());
+  Teuchos::RCP<CrsMatrixWrap> Psubblockwrap = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(Psubblock);
+  Teuchos::RCP<CrsMatrix> Psubblockcrs      = Psubblockwrap->getCrsMatrix();
+  Psubblockcrs->getAllValues(amalgRowPtr, amalgCols, amalgVals);
+
+  size_t paddedNrows = Psubblock->getRowMap()->getLocalNumElements() * Teuchos::as<size_t>(dofPerNode);
+  Teuchos::ArrayRCP<size_t> newPRowPtr(paddedNrows + 1);
+  Teuchos::ArrayRCP<LocalOrdinal> newPCols(Psubblock->getLocalNumEntries() * dofPerNode);
+  Teuchos::ArrayRCP<Scalar> newPVals(Psubblock->getLocalNumEntries() * dofPerNode);
+  size_t cnt = 0;  // local id counter
+  for (decltype(amalgRowPtr.size()) i = 0; i < amalgRowPtr.size() - 1; i++) {
+    size_t rowLength = amalgRowPtr[i + 1] - amalgRowPtr[i];
+    for (int j = 0; j < dofPerNode; j++) {
+      newPRowPtr[i * dofPerNode + j] = cnt;
+      for (size_t k = 0; k < rowLength; k++) {
+        newPCols[cnt]   = amalgCols[k + amalgRowPtr[i]] * dofPerNode + j;
+        newPVals[cnt++] = amalgVals[k + amalgRowPtr[i]];
       }
     }
+  }
 
-    newPRowPtr[paddedNrows] = cnt; // close row CSR array
-    std::vector<size_t> stridingInfo(1, dofPerNode);
-
-    GlobalOrdinal nCoarseDofs = Psubblock->getDomainMap()->getLocalNumElements() * dofPerNode;
-    GlobalOrdinal indexBase   = Psubblock->getDomainMap()->getIndexBase();
-    RCP<const Map> coarseDomainMap = StridedMapFactory::Build(Psubblock->getDomainMap()->lib(),
-        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-        nCoarseDofs,
-        indexBase,
-        stridingInfo,
-        Psubblock->getDomainMap()->getComm(),
-        -1 /* stridedBlockId */,
-        0  /*domainGidOffset */);
-
-    size_t nColCoarseDofs = Teuchos::as<size_t>(Psubblock->getColMap()->getLocalNumElements() * dofPerNode);
-    Teuchos::Array<GlobalOrdinal> unsmooshColMapGIDs(nColCoarseDofs);
-    for(size_t c = 0; c < Psubblock->getColMap()->getLocalNumElements(); ++c) {
-      GlobalOrdinal gid = (Psubblock->getColMap()->getGlobalElement(c)-indexBase) * dofPerNode + indexBase;
-
-      for(int i = 0; i < dofPerNode; ++i) {
-        unsmooshColMapGIDs[c * dofPerNode + i] = gid + i;
-      }
-    }
-    Teuchos::RCP<Map> coarseColMap = MapFactory::Build(Psubblock->getDomainMap()->lib(),
-                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                       unsmooshColMapGIDs(), //View,
-                                                       indexBase,
-                                                       Psubblock->getDomainMap()->getComm());
-
-    Teuchos::Array<GlobalOrdinal> unsmooshRowMapGIDs(paddedNrows);
-    for(size_t c = 0; c < Psubblock->getRowMap()->getLocalNumElements(); ++c) {
-      GlobalOrdinal gid = (Psubblock->getRowMap()->getGlobalElement(c)-indexBase) * dofPerNode + indexBase;
-
-      for(int i = 0; i < dofPerNode; ++i) {
-        unsmooshRowMapGIDs[c * dofPerNode + i] = gid + i;
-      }
+  newPRowPtr[paddedNrows] = cnt;  // close row CSR array
+  std::vector<size_t> stridingInfo(1, dofPerNode);
+
+  GlobalOrdinal nCoarseDofs      = Psubblock->getDomainMap()->getLocalNumElements() * dofPerNode;
+  GlobalOrdinal indexBase        = Psubblock->getDomainMap()->getIndexBase();
+  RCP<const Map> coarseDomainMap = StridedMapFactory::Build(Psubblock->getDomainMap()->lib(),
+                                                            Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                            nCoarseDofs,
+                                                            indexBase,
+                                                            stridingInfo,
+                                                            Psubblock->getDomainMap()->getComm(),
+                                                            -1 /* stridedBlockId */,
+                                                            0 /*domainGidOffset */);
+
+  size_t nColCoarseDofs = Teuchos::as<size_t>(Psubblock->getColMap()->getLocalNumElements() * dofPerNode);
+  Teuchos::Array<GlobalOrdinal> unsmooshColMapGIDs(nColCoarseDofs);
+  for (size_t c = 0; c < Psubblock->getColMap()->getLocalNumElements(); ++c) {
+    GlobalOrdinal gid = (Psubblock->getColMap()->getGlobalElement(c) - indexBase) * dofPerNode + indexBase;
+
+    for (int i = 0; i < dofPerNode; ++i) {
+      unsmooshColMapGIDs[c * dofPerNode + i] = gid + i;
     }
-    Teuchos::RCP<Map> fineRowMap = MapFactory::Build(Psubblock->getDomainMap()->lib(),
+  }
+  Teuchos::RCP<Map> coarseColMap = MapFactory::Build(Psubblock->getDomainMap()->lib(),
                                                      Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                     unsmooshRowMapGIDs(), //View,
+                                                     unsmooshColMapGIDs(),  // View,
                                                      indexBase,
                                                      Psubblock->getDomainMap()->getComm());
 
-    Teuchos::RCP<CrsMatrix> bigPCrs = CrsMatrixFactory::Build(fineRowMap, coarseColMap,
-                                                                  dofPerNode*Psubblock->getLocalMaxNumRowEntries());
-    for (size_t i = 0; i < paddedNrows; i++) {
-      bigPCrs->insertLocalValues(i,
-                                     newPCols.view(newPRowPtr[i], newPRowPtr[i+1] - newPRowPtr[i]),
-                                     newPVals.view(newPRowPtr[i], newPRowPtr[i+1] - newPRowPtr[i]));
-    }
-    bigPCrs->fillComplete(coarseDomainMap, fineRowMap);
+  Teuchos::Array<GlobalOrdinal> unsmooshRowMapGIDs(paddedNrows);
+  for (size_t c = 0; c < Psubblock->getRowMap()->getLocalNumElements(); ++c) {
+    GlobalOrdinal gid = (Psubblock->getRowMap()->getGlobalElement(c) - indexBase) * dofPerNode + indexBase;
 
-    Teuchos::RCP<Matrix> bigP = Teuchos::rcp(new CrsMatrixWrap(bigPCrs));
-
-    Set(coarseLevel, "P", bigP);
+    for (int i = 0; i < dofPerNode; ++i) {
+      unsmooshRowMapGIDs[c * dofPerNode + i] = gid + i;
+    }
   }
+  Teuchos::RCP<Map> fineRowMap = MapFactory::Build(Psubblock->getDomainMap()->lib(),
+                                                   Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                   unsmooshRowMapGIDs(),  // View,
+                                                   indexBase,
+                                                   Psubblock->getDomainMap()->getComm());
+
+  Teuchos::RCP<CrsMatrix> bigPCrs = CrsMatrixFactory::Build(fineRowMap, coarseColMap,
+                                                            dofPerNode * Psubblock->getLocalMaxNumRowEntries());
+  for (size_t i = 0; i < paddedNrows; i++) {
+    bigPCrs->insertLocalValues(i,
+                               newPCols.view(newPRowPtr[i], newPRowPtr[i + 1] - newPRowPtr[i]),
+                               newPVals.view(newPRowPtr[i], newPRowPtr[i + 1] - newPRowPtr[i]));
+  }
+  bigPCrs->fillComplete(coarseDomainMap, fineRowMap);
+
+  Teuchos::RCP<Matrix> bigP = Teuchos::rcp(new CrsMatrixWrap(bigPCrs));
 
+  Set(coarseLevel, "P", bigP);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_REPLICATEPFACTORY_SHORT
-#endif // MUELU_REPLICATEPFACTORY_DEF_HPP
+#endif  // MUELU_REPLICATEPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_decl.hpp b/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_decl.hpp
index b6cadd60749a..c1e563668fb0 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_decl.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_decl.hpp
@@ -58,53 +58,53 @@
 
 namespace MueLu {
 
-  /*!
-    @class RfromP_Or_TransP class.
-    @brief Factory for building restriction operators.
-
-    This factory currently depends on an underlying matrix-matrix multiply with the identity
-    matrix to do the transpose.  This should probably be fixed at some point.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RfromP_Or_TransP : public TwoLevelFactoryBase {
+/*!
+  @class RfromP_Or_TransP class.
+  @brief Factory for building restriction operators.
+
+  This factory currently depends on an underlying matrix-matrix multiply with the identity
+  matrix to do the transpose.  This should probably be fixed at some point.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RfromP_Or_TransP : public TwoLevelFactoryBase {
 #undef MUELU_RFROMP_OR_TRANSP_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    RfromP_Or_TransP() { }
+  //! Constructor.
+  RfromP_Or_TransP() {}
 
-    //! Destructor.
-    virtual ~RfromP_Or_TransP() { }
+  //! Destructor.
+  virtual ~RfromP_Or_TransP() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build(Level & fineLevel, Level & coarseLevel) const;
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; //class RfromP_Or_TransP
+};  // class RfromP_Or_TransP
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RFROMP_OR_TRANSP_SHORT
-#endif // MUELU_RFROMP_OR_TRANSP_DECL_HPP
+#endif  // MUELU_RFROMP_OR_TRANSP_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_def.hpp b/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_def.hpp
index 75b8d8ca0db5..d7e75f240c6a 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_def.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_RfromP_Or_TransP_def.hpp
@@ -65,125 +65,128 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix P");
-    validParamList->set< RCP<const FactoryBase> >("RfromPfactory", Teuchos::null, "Generating factory of the matrix R");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase>>("P", Teuchos::null, "Generating factory of the matrix P");
+  validParamList->set<RCP<const FactoryBase>>("RfromPfactory", Teuchos::null, "Generating factory of the matrix R");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  Input(coarseLevel, "RfromPfactory");
+
+  // Using a PgPFactory in conjunction with a TogglePFactory is a bit problematic. Normally, PgPFactory is supposed to be
+  // invoked twice (once in standard mode and a second time in RestrictionMode). Unfortunately, TogglePFactory
+  // is not designed to produce R. A second issue is that TogglePFactory stores prolongators in an array,
+  // and there are some challenges in determining which array entry (i.e., factory) is needed when producing R.
+  // The way this is addressed is a bit clumsy. RfromP_Or_TransP invokes the prolongator factory to produce R.
+  // To do this, it must first check that this is needed (as opposed to just transposing P or using an already computed
+  // R that might be produced by SemiCoarsenPFactory). This check is needed in both DeclareInput() and in Build().
+  // The DeclareInput() check verifies that TogglePFactory was requested and that one of the prolongator factories
+  // within TogglePFactory is a PgPFactory.  RfromP_Or_TransP's DeclareInput then invokes DeclareDependencies, and
+  // DeclareInput for the PgPFactory. The check within Build(), looks at "RfromPFactory" to see if it is an integer. This
+  // integer is used to find the prolongator factory that is invoked in RestrictionMode to produce R. Otherwise,
+  // "RfromPFactory" is used to get the pre-computed restriction matrix. If "RfromPFactory" is not present, then RfromP_Or_TransP
+  // just transposes P to get R.
+
+  RCP<const FactoryBase> PFact = coarseLevel.GetFactoryManager()->GetFactory("P");
+  if (PFact == Teuchos::null) {
+    PFact = GetFactory("P");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
-    Input(coarseLevel, "RfromPfactory");
-
-    // Using a PgPFactory in conjunction with a TogglePFactory is a bit problematic. Normally, PgPFactory is supposed to be
-    // invoked twice (once in standard mode and a second time in RestrictionMode). Unfortunately, TogglePFactory
-    // is not designed to produce R. A second issue is that TogglePFactory stores prolongators in an array,
-    // and there are some challenges in determining which array entry (i.e., factory) is needed when producing R.
-    // The way this is addressed is a bit clumsy. RfromP_Or_TransP invokes the prolongator factory to produce R.
-    // To do this, it must first check that this is needed (as opposed to just transposing P or using an already computed
-    // R that might be produced by SemiCoarsenPFactory). This check is needed in both DeclareInput() and in Build().
-    // The DeclareInput() check verifies that TogglePFactory was requested and that one of the prolongator factories
-    // within TogglePFactory is a PgPFactory.  RfromP_Or_TransP's DeclareInput then invokes DeclareDependencies, and
-    // DeclareInput for the PgPFactory. The check within Build(), looks at "RfromPFactory" to see if it is an integer. This
-    // integer is used to find the prolongator factory that is invoked in RestrictionMode to produce R. Otherwise,
-    // "RfromPFactory" is used to get the pre-computed restriction matrix. If "RfromPFactory" is not present, then RfromP_Or_TransP
-    // just transposes P to get R.
-
-    RCP<const FactoryBase> PFact = coarseLevel.GetFactoryManager()->GetFactory("P");
-    if (PFact == Teuchos::null) { PFact = GetFactory("P"); }
-    coarseLevel.DeclareInput("P", PFact.get(), this);
-    RCP<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> > myToggleFact = Teuchos::rcp_const_cast<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(rcp_dynamic_cast<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(PFact));
-    if (myToggleFact != Teuchos::null) {
-      for (size_t ii = 0; ii < myToggleFact->NumProlongatorFactories();  ii++) {
-        RCP<const MueLu::PgPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> > actualPFact = Teuchos::rcp_const_cast<const MueLu::PgPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>>(rcp_dynamic_cast<const MueLu::PgPFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(myToggleFact->getProlongatorFactory(ii)));
-        if (actualPFact != Teuchos::null) {
-          RCP<PFactory> subFactory = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(myToggleFact->getProlongatorFactory(ii)));;
-          bool rmode = subFactory->isRestrictionModeSet();
-          subFactory->setRestrictionMode(true);
-          // Force request call for actualPFact
-          coarseLevel.DeclareDependencies(actualPFact.get());
-          coarseLevel.DeclareInput("R", actualPFact.get(), this);
-          subFactory->setRestrictionMode(rmode);
-	}
+  coarseLevel.DeclareInput("P", PFact.get(), this);
+  RCP<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>> myToggleFact = Teuchos::rcp_const_cast<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(rcp_dynamic_cast<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(PFact));
+  if (myToggleFact != Teuchos::null) {
+    for (size_t ii = 0; ii < myToggleFact->NumProlongatorFactories(); ii++) {
+      RCP<const MueLu::PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>> actualPFact = Teuchos::rcp_const_cast<const MueLu::PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(rcp_dynamic_cast<const MueLu::PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(myToggleFact->getProlongatorFactory(ii)));
+      if (actualPFact != Teuchos::null) {
+        RCP<PFactory> subFactory = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(myToggleFact->getProlongatorFactory(ii)));
+        ;
+        bool rmode = subFactory->isRestrictionModeSet();
+        subFactory->setRestrictionMode(true);
+        // Force request call for actualPFact
+        coarseLevel.DeclareDependencies(actualPFact.get());
+        coarseLevel.DeclareInput("R", actualPFact.get(), this);
+        subFactory->setRestrictionMode(rmode);
       }
     }
   }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Transpose P", coarseLevel);
+  std::string label = "MueLu::TransP-" + Teuchos::toString(coarseLevel.GetLevelID());
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  // Reuse pattern if available (multiple solve)
+  RCP<ParameterList> Tparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    Tparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    Tparams = rcp(new ParameterList);
+
+  // By default, we don't need global constants for transpose
+  Tparams->set("compute global constants: temporaries", Tparams->get("compute global constants: temporaries", false));
+  Tparams->set("compute global constants", Tparams->get("compute global constants", false));
+
+  RCP<Matrix> R;
+  RCP<const FactoryBase> PFact = coarseLevel.GetFactoryManager()->GetFactory("P");
+  if (PFact == Teuchos::null) {
+    PFact = GetFactory("P");
+  }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RfromP_Or_TransP<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Transpose P", coarseLevel);
-    std::string label = "MueLu::TransP-" + Teuchos::toString(coarseLevel.GetLevelID());
-
-    const Teuchos::ParameterList& pL = GetParameterList();
-
-    // Reuse pattern if available (multiple solve)
-    RCP<ParameterList> Tparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-        Tparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-      else
-        Tparams= rcp(new ParameterList);
-
-    // By default, we don't need global constants for transpose
-    Tparams->set("compute global constants: temporaries",Tparams->get("compute global constants: temporaries",false));
-    Tparams->set("compute global constants",Tparams->get("compute global constants",false));
-
-    RCP<Matrix> R;
-    RCP<const FactoryBase> PFact = coarseLevel.GetFactoryManager()->GetFactory("P");
-    if (PFact == Teuchos::null) { PFact = GetFactory("P"); }
-
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P",             PFact.get());
-
-    if (coarseLevel.IsAvailable("RfromPfactory", PFact.get()))
-    {
-      std::string strType = coarseLevel.GetTypeName("RfromPfactory",  PFact.get());
-      // Address case where a toggle factory is used in conjunction with a PgP factory. Here,
-      // we need to invoke the PgP factory a 2nd time to produce restriction. In this
-      // situation the PgP factory puts an int RfromPfactory on the level.
-      //
-      // See comments aboove in DeclareInput() method  for more detailsd
-      if (strType == "int") {
-        RCP<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> > myToggleFact = Teuchos::rcp_const_cast<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(rcp_dynamic_cast<const MueLu::TogglePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(PFact));;
-        if (myToggleFact != Teuchos::null) {
-          MueLu::DisableMultipleCallCheck check(myToggleFact);
-          RCP<PFactory> actualPFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(myToggleFact->getProlongatorFactory((size_t) coarseLevel.Get<int>("RfromPfactory",  PFact.get()))));
-	                                                                                    // toggle factory sets RfromPfactory to correct index into prolongatorFactory array
-          MueLu::DisableMultipleCallCheck check2(actualPFact);
-          bool rmode = actualPFact->isRestrictionModeSet();
-          actualPFact->setRestrictionMode(true);
-          R = coarseLevel.Get<RCP<Matrix> >("R",actualPFact.get());
-          actualPFact->setRestrictionMode(rmode);
-        }
-        else  R = Utilities::Transpose(*P, true,label,Tparams);
-      }
-      else R  = coarseLevel.Get< RCP<Matrix> >("RfromPfactory", PFact.get());
-    }
-    else
-      R = Utilities::Transpose(*P, true,label,Tparams);
-
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
-    }
-
-    Set(coarseLevel, "R", R);
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix>>("P", PFact.get());
+
+  if (coarseLevel.IsAvailable("RfromPfactory", PFact.get())) {
+    std::string strType = coarseLevel.GetTypeName("RfromPfactory", PFact.get());
+    // Address case where a toggle factory is used in conjunction with a PgP factory. Here,
+    // we need to invoke the PgP factory a 2nd time to produce restriction. In this
+    // situation the PgP factory puts an int RfromPfactory on the level.
+    //
+    // See comments aboove in DeclareInput() method  for more detailsd
+    if (strType == "int") {
+      RCP<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>> myToggleFact = Teuchos::rcp_const_cast<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(rcp_dynamic_cast<const MueLu::TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(PFact));
+      ;
+      if (myToggleFact != Teuchos::null) {
+        MueLu::DisableMultipleCallCheck check(myToggleFact);
+        RCP<PFactory> actualPFact = Teuchos::rcp_const_cast<PFactory>(rcp_dynamic_cast<const PFactory>(myToggleFact->getProlongatorFactory((size_t)coarseLevel.Get<int>("RfromPfactory", PFact.get()))));
+        // toggle factory sets RfromPfactory to correct index into prolongatorFactory array
+        MueLu::DisableMultipleCallCheck check2(actualPFact);
+        bool rmode = actualPFact->isRestrictionModeSet();
+        actualPFact->setRestrictionMode(true);
+        R = coarseLevel.Get<RCP<Matrix>>("R", actualPFact.get());
+        actualPFact->setRestrictionMode(rmode);
+      } else
+        R = Utilities::Transpose(*P, true, label, Tparams);
+    } else
+      R = coarseLevel.Get<RCP<Matrix>>("RfromPfactory", PFact.get());
+  } else
+    R = Utilities::Transpose(*P, true, label, Tparams);
+
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
+  }
 
-    ///////////////////////// EXPERIMENTAL
-    if (P->IsView("stridedMaps"))
-      R->CreateView("stridedMaps", P, true);
-    ///////////////////////// EXPERIMENTAL
+  Set(coarseLevel, "R", R);
 
-  }
+  ///////////////////////// EXPERIMENTAL
+  if (P->IsView("stridedMaps"))
+    R->CreateView("stridedMaps", P, true);
+  ///////////////////////// EXPERIMENTAL
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_RFROMP_OR_TRANSP_DEF_HPP
+#endif  // MUELU_RFROMP_OR_TRANSP_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_decl.hpp b/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_decl.hpp
index d15bf13cf5ad..3fad8f143865 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_decl.hpp
@@ -58,53 +58,53 @@
 
 namespace MueLu {
 
-  /*!
-    @class TransPFactory class.
-    @brief Factory for building restriction operators.
-
-    This factory currently depends on an underlying matrix-matrix multiply with the identity
-    matrix to do the transpose.  This should probably be fixed at some point.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class TransPFactory : public TwoLevelFactoryBase {
+/*!
+  @class TransPFactory class.
+  @brief Factory for building restriction operators.
+
+  This factory currently depends on an underlying matrix-matrix multiply with the identity
+  matrix to do the transpose.  This should probably be fixed at some point.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class TransPFactory : public TwoLevelFactoryBase {
 #undef MUELU_TRANSPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    TransPFactory() { }
+  //! Constructor.
+  TransPFactory() {}
 
-    //! Destructor.
-    virtual ~TransPFactory() { }
+  //! Destructor.
+  virtual ~TransPFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build(Level & fineLevel, Level & coarseLevel) const;
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; //class TransPFactory
+};  // class TransPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TRANSPFACTORY_SHORT
-#endif // MUELU_TRANSPFACTORY_DECL_HPP
+#endif  // MUELU_TRANSPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_def.hpp b/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_def.hpp
index 0ad9ee9f2d23..db32f0e0efc9 100644
--- a/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Generic/MueLu_TransPFactory_def.hpp
@@ -53,72 +53,70 @@
 
 #include "MueLu_TransPFactory_decl.hpp"
 
-
 #include "MueLu_Monitor.hpp"
 #include "MueLu_PerfUtils.hpp"
 #include "MueLu_Utilities.hpp"
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix P");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Generating factory of the matrix P");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  Input(coarseLevel, "P");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* fineLevel */, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Transpose P", coarseLevel);
+  std::string label = "MueLu::TransP-" + Teuchos::toString(coarseLevel.GetLevelID());
+
+  RCP<Matrix> P = Get<RCP<Matrix> >(coarseLevel, "P");
+  // If we failed to create a valid P (e.g., # of global aggregates is zero), then we just bail here
+  //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (P == Teuchos::null) return;
+
+  const Teuchos::ParameterList& pL = GetParameterList();
+
+  // Reuse pattern if available (multiple solve)
+  RCP<ParameterList> Tparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    Tparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    Tparams = rcp(new ParameterList);
+
+  // By default, we don't need global constants for transpose
+  Tparams->set("compute global constants: temporaries", Tparams->get("compute global constants: temporaries", false));
+  Tparams->set("compute global constants", Tparams->get("compute global constants", false));
+
+  RCP<Matrix> R = Utilities::Transpose(*P, true, label, Tparams);
+
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
-    Input(coarseLevel, "P");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TransPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* fineLevel */, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Transpose P", coarseLevel);
-    std::string label = "MueLu::TransP-" + Teuchos::toString(coarseLevel.GetLevelID());
-
-    RCP<Matrix> P = Get< RCP<Matrix> >(coarseLevel, "P");
-    // If we failed to create a valid P (e.g., # of global aggregates is zero), then we just bail here
-    //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-    if (P == Teuchos::null) return;
-
-    const Teuchos::ParameterList& pL = GetParameterList();
+  Set(coarseLevel, "R", R);
 
-    // Reuse pattern if available (multiple solve)
-    RCP<ParameterList> Tparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-        Tparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-      else
-        Tparams= rcp(new ParameterList);
-
-    // By default, we don't need global constants for transpose
-    Tparams->set("compute global constants: temporaries",Tparams->get("compute global constants: temporaries",false));
-    Tparams->set("compute global constants",Tparams->get("compute global constants",false));
-
-    RCP<Matrix> R = Utilities::Transpose(*P, true,label,Tparams);
-
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
-    }
-
-    Set(coarseLevel, "R", R);
-
-    ///////////////////////// EXPERIMENTAL
-    if (P->IsView("stridedMaps"))
-      R->CreateView("stridedMaps", P, true);
-    ///////////////////////// EXPERIMENTAL
-
-  }
+  ///////////////////////// EXPERIMENTAL
+  if (P->IsView("stridedMaps"))
+    R->CreateView("stridedMaps", P, true);
+  ///////////////////////// EXPERIMENTAL
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_TRANSPFACTORY_DEF_HPP
+#endif  // MUELU_TRANSPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_decl.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_decl.hpp
index 2e0b938651e3..a34fb3e02a5d 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_decl.hpp
@@ -80,78 +80,78 @@
 
 namespace MueLu {
 
-  /*!
-    @class GeoInterpFactory class.
-    @brief Factory for GMG Q2-Q1-Q2 interpolation
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class GeoInterpFactory : public PFactory {
+/*!
+  @class GeoInterpFactory class.
+  @brief Factory for GMG Q2-Q1-Q2 interpolation
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class GeoInterpFactory : public PFactory {
 #undef MUELU_GEOINTERPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    GeoInterpFactory();
+  //! Constructor.
+  GeoInterpFactory();
 
-    //! Destructor.
-    virtual ~GeoInterpFactory();
+  //! Destructor.
+  virtual ~GeoInterpFactory();
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the
-        generating factory is not specified in DeclareInput, then this
-        class will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the
+      generating factory is not specified in DeclareInput, then this
+      class will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    void BuildP(Level &fineLevel, Level &coarseLevel) const;
+  void BuildP(Level &fineLevel, Level &coarseLevel) const;
 
-    //void BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const;
+  // void BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const;
 
-    /* For our geometric multigrid needs, we will explicitly build the
-       coarse grid here and store it as level data. There are two
-       things we care about here: coordinates of the dofs and
-       element-dof lists.
+  /* For our geometric multigrid needs, we will explicitly build the
+     coarse grid here and store it as level data. There are two
+     things we care about here: coordinates of the dofs and
+     element-dof lists.
 
-       We will assume only that the elements are numbered
-       lexicographically, left-to-right, bottom-to-top. The order of
-       the degrees of freedom will only be assumed on coarse grids. It
-       can be anything on the finest grid (fineLevel.GetLevelID()==1).
+     We will assume only that the elements are numbered
+     lexicographically, left-to-right, bottom-to-top. The order of
+     the degrees of freedom will only be assumed on coarse grids. It
+     can be anything on the finest grid (fineLevel.GetLevelID()==1).
 
-       We further assume that the x- and y-components of velocity are
-       zippered together: UX1 UY1 UX2 UY2 ETC...
-    */
+     We further assume that the x- and y-components of velocity are
+     zippered together: UX1 UY1 UX2 UY2 ETC...
+  */
 
-    //@}
+  //@}
 
-  private:
-    // No parameters need to be stored... To save "P", use
-    //    coarseLevel.Set("P", finalP, this);
+ private:
+  // No parameters need to be stored... To save "P", use
+  //    coarseLevel.Set("P", finalP, this);
 
-  }; // class GeoInterpFactory
+};  // class GeoInterpFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GEOINTERPFACTORY_SHORT
-#endif // MUELU_GEOINTERPFACTORY_DECL_HPP
+#endif  // MUELU_GEOINTERPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_def.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_def.hpp
index 05b127770de6..c66ab4cfee66 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_GeoInterpFactory_def.hpp
@@ -64,1025 +64,976 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GeoInterpFactory(){
-    GetOStream(Runtime1) << "I constructed a GeoInterpFactory object... Nothing else to do here." << std::endl;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~GeoInterpFactory(){
-    // Should be empty. All destruction should be handled by Level-based get stuff and RCP
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "A00");
-    Input(fineLevel, "A10");
-    Input(fineLevel, "A20");
-
-    Input(fineLevel, "VElementList");
-    Input(fineLevel, "PElementList");
-    Input(fineLevel, "MElementList");
-
-
-    Input(coarseLevel,"VElementList");
-    Input(coarseLevel,"PElementList");
-    Input(coarseLevel,"MElementList");
-
-/*
-    coarseLevel.DeclareInput("VElementList",coarseLevel.GetFactoryManager()->GetFactory("VElementList").get(),this);
-    coarseLevel.DeclareInput("PElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
-    coarseLevel.DeclareInput("MElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
-
-    fineLevel.DeclareInput("VElementList",fineLevel.GetFactoryManager()->GetFactory("VElementList").get(),this);
-    fineLevel.DeclareInput("PElementList",fineLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
-    fineLevel.DeclareInput("MElementList",fineLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
-*/
-
-    //currentLevel.DeclareInput(varName_,factory_,this);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-
-    GetOStream(Runtime1) << "Starting 'build' routine...\n";
-
-    // This will create a list of elements on the coarse grid with a
-    // predictable structure, as well as modify the fine grid list of
-    // elements, if necessary (i.e. if fineLevel.GetLevelID()==0);
-    //BuildCoarseGrid(fineLevel,coarseLevel);
-
-    // This will actually build our prolongator P
-    return BuildP(fineLevel,coarseLevel);
-
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
-
-    typedef Teuchos::SerialDenseMatrix<GO,GO> SerialDenseMatrixType;
-
-    GetOStream(Runtime1) << "Starting 'BuildP' routine...\n";
-
-    //DEBUG
-    //Teuchos::FancyOStream fout(*GetOStream(Runtime1));
-    //fineLevel.print(fout,Teuchos::VERB_HIGH);
-
-    // Get finegrid element lists
-    RCP<SerialDenseMatrixType> fineElementPDOFs = Get< RCP<SerialDenseMatrixType> > (fineLevel, "PElementList");
-    RCP<SerialDenseMatrixType> fineElementVDOFs = Get< RCP<SerialDenseMatrixType> > (fineLevel, "VElementList");
-    RCP<SerialDenseMatrixType> fineElementMDOFs = Get< RCP<SerialDenseMatrixType> > (fineLevel, "MElementList");
-
-    //DEBUG
-    GetOStream(Runtime1) << "done getting fine level elements...\n";
-    GetOStream(Runtime1) << "getting coarse level elements...\n";
-    //coarseLevel.print(fout,Teuchos::VERB_HIGH);
-
-
-    // Get coarse grid element lists
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > coarseElementVDOFs,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GeoInterpFactory() {
+  GetOStream(Runtime1) << "I constructed a GeoInterpFactory object... Nothing else to do here." << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~GeoInterpFactory() {
+  // Should be empty. All destruction should be handled by Level-based get stuff and RCP
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "A00");
+  Input(fineLevel, "A10");
+  Input(fineLevel, "A20");
+
+  Input(fineLevel, "VElementList");
+  Input(fineLevel, "PElementList");
+  Input(fineLevel, "MElementList");
+
+  Input(coarseLevel, "VElementList");
+  Input(coarseLevel, "PElementList");
+  Input(coarseLevel, "MElementList");
+
+  /*
+      coarseLevel.DeclareInput("VElementList",coarseLevel.GetFactoryManager()->GetFactory("VElementList").get(),this);
+      coarseLevel.DeclareInput("PElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
+      coarseLevel.DeclareInput("MElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
+
+      fineLevel.DeclareInput("VElementList",fineLevel.GetFactoryManager()->GetFactory("VElementList").get(),this);
+      fineLevel.DeclareInput("PElementList",fineLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
+      fineLevel.DeclareInput("MElementList",fineLevel.GetFactoryManager()->GetFactory("PElementList").get(),this);
+  */
+
+  // currentLevel.DeclareInput(varName_,factory_,this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  GetOStream(Runtime1) << "Starting 'build' routine...\n";
+
+  // This will create a list of elements on the coarse grid with a
+  // predictable structure, as well as modify the fine grid list of
+  // elements, if necessary (i.e. if fineLevel.GetLevelID()==0);
+  // BuildCoarseGrid(fineLevel,coarseLevel);
+
+  // This will actually build our prolongator P
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GeoInterpFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
+  typedef Teuchos::SerialDenseMatrix<GO, GO> SerialDenseMatrixType;
+
+  GetOStream(Runtime1) << "Starting 'BuildP' routine...\n";
+
+  // DEBUG
+  // Teuchos::FancyOStream fout(*GetOStream(Runtime1));
+  // fineLevel.print(fout,Teuchos::VERB_HIGH);
+
+  // Get finegrid element lists
+  RCP<SerialDenseMatrixType> fineElementPDOFs = Get<RCP<SerialDenseMatrixType> >(fineLevel, "PElementList");
+  RCP<SerialDenseMatrixType> fineElementVDOFs = Get<RCP<SerialDenseMatrixType> >(fineLevel, "VElementList");
+  RCP<SerialDenseMatrixType> fineElementMDOFs = Get<RCP<SerialDenseMatrixType> >(fineLevel, "MElementList");
+
+  // DEBUG
+  GetOStream(Runtime1) << "done getting fine level elements...\n";
+  GetOStream(Runtime1) << "getting coarse level elements...\n";
+  // coarseLevel.print(fout,Teuchos::VERB_HIGH);
+
+  // Get coarse grid element lists
+  RCP<Teuchos::SerialDenseMatrix<GO, GO> > coarseElementVDOFs,
       coarseElementPDOFs,
       coarseElementMDOFs;
 
-    coarseLevel.Get("VElementList",coarseElementVDOFs,coarseLevel.GetFactoryManager()->GetFactory("VElementList").get());
-    coarseLevel.Get("PElementList",coarseElementPDOFs,coarseLevel.GetFactoryManager()->GetFactory("PElementList").get());
-    coarseLevel.Get("MElementList",coarseElementMDOFs,coarseLevel.GetFactoryManager()->GetFactory("MElementList").get());
-
-
-    GetOStream(Runtime1) << "computing various numbers...\n";
-    // Number of elements?
-    GO totalFineElements = fineElementMDOFs->numRows();
-    LO nFineElements = (int) sqrt(totalFineElements);
-    GO totalCoarseElements = coarseElementMDOFs->numRows();
-    LO nCoarseElements = (int) sqrt(totalCoarseElements);
-
-    // Set sizes for *COARSE GRID*
-    GO nM = (2*nCoarseElements+1)*(2*nCoarseElements+1);
-    GO nV = 2*nM;
-    GO nP = (nCoarseElements+1)*(nCoarseElements+1);
-
-    // Get the row maps for the Ps
-    RCP<Matrix> fineA00 = Get<RCP<Matrix> > (fineLevel,"A00");
-    RCP<Matrix> fineA10 = Get<RCP<Matrix> > (fineLevel,"A10");
-    RCP<Matrix> fineA20 = Get<RCP<Matrix> > (fineLevel,"A20");
-
-    GetOStream(Runtime1) << "creating coarse grid maps...\n";
-
-    RCP<const Map> rowMapforPV = fineA00->getRowMap();
-    RCP<const Map> rowMapforPP = fineA10->getRowMap();
-    RCP<const Map> rowMapforPM = fineA20->getRowMap();
-
-    GO fNV = rowMapforPV->getGlobalNumElements();
-    GO fNP = rowMapforPP->getGlobalNumElements();
-    GO fNM = rowMapforPM->getGlobalNumElements();
-
-    // Get the comm for the maps
-    RCP<const Teuchos::Comm<int> > comm = rowMapforPV->getComm();
-
-    // Create rowMap for P
-    RCP<Matrix> FineA = Factory::Get< RCP<Matrix> >(fineLevel, "A");
-    RCP<const Map> rowMapforP  = FineA->getRowMap();
-
-    // Create colMaps for the coarse grid
-    RCP<const Map> colMapforPV = Xpetra::MapFactory<LO,GO>::createUniformContigMap(Xpetra::UseTpetra,nV,comm);
-    RCP<const Map> colMapforPP = Xpetra::MapFactory<LO,GO>::createUniformContigMap(Xpetra::UseTpetra,nP,comm);
-    RCP<const Map> colMapforPM = Xpetra::MapFactory<LO,GO>::createUniformContigMap(Xpetra::UseTpetra,nM,comm);
-
-
-    GetOStream(Runtime1) << "creating coarse grid matrices...\n";
-    //Create our final output Ps for the coarseGrid
-    size_t maxEntriesPerRowV = 9,//No overlap of VX and VY
-      maxEntriesPerRowP = 4,
-      maxEntriesPerRowM = 9;
-
-    RCP<Matrix> P  = rcp(new CrsMatrixWrap(rowMapforP ,maxEntriesPerRowV));
-    RCP<Matrix> PV = rcp(new CrsMatrixWrap(rowMapforPV,maxEntriesPerRowV));
-    RCP<Matrix> PP = rcp(new CrsMatrixWrap(rowMapforPP,maxEntriesPerRowP));
-    RCP<Matrix> PM = rcp(new CrsMatrixWrap(rowMapforPM,maxEntriesPerRowM));
-
-
-    //*****************************************************************/
-    //
-    //All 25 fine grid dofs are completely determined by the coarse
-    //grid element in which they reside! So I should loop over coarse
-    //grid elements and build 25 rows at a time! If that's not
-    //ridiculous... I just have to be careful about duplicates on
-    //future elements! But duplicates are easy - just the bottom and
-    //left edges.
-    //
-    //
-    //Looking at a fine grid patch, define the following Local-Global
-    //relationship (magnetics as an example):
-    //
-    // Bottom-Left Corner:
-    // 0 -> (*fineElementMDOFs)(fineElement[0],0)
-    //
-    // Bottom Edge:
-    // 1 -> (*fineElementMDOFs)(fineElement[0],4)
-    // 2 -> (*fineElementMDOFs)(fineElement[0],2)
-    // 3 -> (*fineElementMDOFs)(fineElement[1],4)
-    // 4 -> (*fineElementMDOFs)(fineElement[1],2)
-    //
-    // Left Edge:
-    // 5 -> (*fineElementMDOFs)(fineElement[0],7)
-    // 6 -> (*fineElementMDOFs)(fineElement[0],3)
-    // 7 -> (*fineElementMDOFs)(fineElement[2],7)
-    // 8 -> (*fineElementMDOFs)(fineElement[2],3)
-    //
-    // All the rest:
-    // 9 -> (*fineElementMDOFs)(fineElement[3],0)
-    // 10 -> (*fineElementMDOFs)(fineElement[3],1)
-    // 11 -> (*fineElementMDOFs)(fineElement[3],2)
-    // 12 -> (*fineElementMDOFs)(fineElement[3],3)
-    // 13 -> (*fineElementMDOFs)(fineElement[0],5)
-    // 14 -> (*fineElementMDOFs)(fineElement[0],6)
-    // 15 -> (*fineElementMDOFs)(fineElement[1],5)
-    // 16 -> (*fineElementMDOFs)(fineElement[1],6)
-    // 17 -> (*fineElementMDOFs)(fineElement[2],5)
-    // 18 -> (*fineElementMDOFs)(fineElement[2],6)
-    // 19 -> (*fineElementMDOFs)(fineElement[3],5)
-    // 20 -> (*fineElementMDOFs)(fineElement[3],6)
-    // 21 -> (*fineElementMDOFs)(fineElement[0],8)
-    // 22 -> (*fineElementMDOFs)(fineElement[1],8)
-    // 23 -> (*fineElementMDOFs)(fineElement[2],8)
-    // 24 -> (*fineElementMDOFs)(fineElement[3],8)
-    //
-    //*****************************************************************/
-
-    size_t nnz = 0; // Just to make my copy-paste life easier...
-    Teuchos::ArrayRCP<GO> colPtrV(maxEntriesPerRowV,0);
-    Teuchos::ArrayRCP<GO> colPtrM(maxEntriesPerRowM,0);
-    Teuchos::ArrayRCP<SC> valPtrM(maxEntriesPerRowM,0.);
-
-    Teuchos::ArrayRCP<GO> colPtrP(maxEntriesPerRowP,0);
-    Teuchos::ArrayRCP<SC> valPtrP(maxEntriesPerRowP,0.);
-
-    //About which fine-grid elements do we care?
-    GO fineElement[4] = {0,1,nFineElements,nFineElements+1};
-
-    GetOStream(Runtime1) << "start building matrices...\n";
-    GetOStream(Runtime1) << "nCoarseElements = " << nCoarseElements << std::endl;
-
-    for ( GO coarseElement=0; coarseElement<totalCoarseElements; coarseElement++)
-      {
-        // We don't really care what is shared with future elements -
-        // we know a priori what needs to be filled for the element
-        // we're dealing with, depending of if it it's on the bottom
-        // edge, the left edge, or in the middle. Thoses should be the
-        // only cases we care about, and they should require a
-        // superset of the work required for an interior node.
-
-        //if (CoarseElement is on bottom edge)
-        if (coarseElement < nCoarseElements)
-          {
-            //fill in the bottom edge of the element patch
-            // FP = 1
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-            valPtrM[0] = 0.375;
-            colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-            valPtrM[1] = -0.125;
-            colPtrM[2] = (*coarseElementMDOFs)(coarseElement,4);
-            valPtrM[2] = 0.75;
-
-            nnz = 3;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],4),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-            colPtrV[1] = 2*colPtrM[1];
-            colPtrV[2] = 2*colPtrM[2];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],8),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-            colPtrV[1] = 2*colPtrM[1]+1;
-            colPtrV[2] = 2*colPtrM[2]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],9),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            // FPr = 1
-            colPtrP[0] = (*coarseElementPDOFs)(coarseElement,0);
-            valPtrP[0] = 0.5;
-            colPtrP[1] = (*coarseElementPDOFs)(coarseElement,1);
-            valPtrP[1] = 0.5;
-
-            nnz = 2;
-            PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0],1),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-            // FP = 2
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,4);
-            valPtrM[0] = 1.0;
-
-            nnz = 1;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],1),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],2),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],3),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            //FPr = 2
-            colPtrP[0] = (*coarseElementPDOFs)(coarseElement,1);
-            valPtrP[0] = 1.0;
-
-            nnz = 1;
-            PP->insertGlobalValues((*fineElementPDOFs)(fineElement[1],1),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-
-            // FP = 3
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-            valPtrM[0] = -0.125;
-            colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-            valPtrM[1] = 0.375;
-            colPtrM[2] = (*coarseElementMDOFs)(coarseElement,4);
-            valPtrM[2] = 0.75;
-
-            nnz = 3;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1],4),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-            colPtrV[1] = 2*colPtrM[1];
-            colPtrV[2] = 2*colPtrM[2];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],8),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-            colPtrV[1] = 2*colPtrM[1]+1;
-            colPtrV[2] = 2*colPtrM[2]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],9),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-
-            // FP = 4
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,1);
-            valPtrM[0] = 1.0;
-
-            nnz = 1;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1],1),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],2),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],3),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            //if (CoarseElement is on the bottom left corner)
-            if (coarseElement == 0)
-              {
+  coarseLevel.Get("VElementList", coarseElementVDOFs, coarseLevel.GetFactoryManager()->GetFactory("VElementList").get());
+  coarseLevel.Get("PElementList", coarseElementPDOFs, coarseLevel.GetFactoryManager()->GetFactory("PElementList").get());
+  coarseLevel.Get("MElementList", coarseElementMDOFs, coarseLevel.GetFactoryManager()->GetFactory("MElementList").get());
+
+  GetOStream(Runtime1) << "computing various numbers...\n";
+  // Number of elements?
+  GO totalFineElements   = fineElementMDOFs->numRows();
+  LO nFineElements       = (int)sqrt(totalFineElements);
+  GO totalCoarseElements = coarseElementMDOFs->numRows();
+  LO nCoarseElements     = (int)sqrt(totalCoarseElements);
+
+  // Set sizes for *COARSE GRID*
+  GO nM = (2 * nCoarseElements + 1) * (2 * nCoarseElements + 1);
+  GO nV = 2 * nM;
+  GO nP = (nCoarseElements + 1) * (nCoarseElements + 1);
+
+  // Get the row maps for the Ps
+  RCP<Matrix> fineA00 = Get<RCP<Matrix> >(fineLevel, "A00");
+  RCP<Matrix> fineA10 = Get<RCP<Matrix> >(fineLevel, "A10");
+  RCP<Matrix> fineA20 = Get<RCP<Matrix> >(fineLevel, "A20");
+
+  GetOStream(Runtime1) << "creating coarse grid maps...\n";
+
+  RCP<const Map> rowMapforPV = fineA00->getRowMap();
+  RCP<const Map> rowMapforPP = fineA10->getRowMap();
+  RCP<const Map> rowMapforPM = fineA20->getRowMap();
+
+  GO fNV = rowMapforPV->getGlobalNumElements();
+  GO fNP = rowMapforPP->getGlobalNumElements();
+  GO fNM = rowMapforPM->getGlobalNumElements();
+
+  // Get the comm for the maps
+  RCP<const Teuchos::Comm<int> > comm = rowMapforPV->getComm();
+
+  // Create rowMap for P
+  RCP<Matrix> FineA         = Factory::Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<const Map> rowMapforP = FineA->getRowMap();
+
+  // Create colMaps for the coarse grid
+  RCP<const Map> colMapforPV = Xpetra::MapFactory<LO, GO>::createUniformContigMap(Xpetra::UseTpetra, nV, comm);
+  RCP<const Map> colMapforPP = Xpetra::MapFactory<LO, GO>::createUniformContigMap(Xpetra::UseTpetra, nP, comm);
+  RCP<const Map> colMapforPM = Xpetra::MapFactory<LO, GO>::createUniformContigMap(Xpetra::UseTpetra, nM, comm);
+
+  GetOStream(Runtime1) << "creating coarse grid matrices...\n";
+  // Create our final output Ps for the coarseGrid
+  size_t maxEntriesPerRowV = 9,  // No overlap of VX and VY
+      maxEntriesPerRowP    = 4,
+         maxEntriesPerRowM = 9;
+
+  RCP<Matrix> P  = rcp(new CrsMatrixWrap(rowMapforP, maxEntriesPerRowV));
+  RCP<Matrix> PV = rcp(new CrsMatrixWrap(rowMapforPV, maxEntriesPerRowV));
+  RCP<Matrix> PP = rcp(new CrsMatrixWrap(rowMapforPP, maxEntriesPerRowP));
+  RCP<Matrix> PM = rcp(new CrsMatrixWrap(rowMapforPM, maxEntriesPerRowM));
+
+  //*****************************************************************/
+  //
+  // All 25 fine grid dofs are completely determined by the coarse
+  // grid element in which they reside! So I should loop over coarse
+  // grid elements and build 25 rows at a time! If that's not
+  // ridiculous... I just have to be careful about duplicates on
+  // future elements! But duplicates are easy - just the bottom and
+  // left edges.
+  //
+  //
+  // Looking at a fine grid patch, define the following Local-Global
+  // relationship (magnetics as an example):
+  //
+  // Bottom-Left Corner:
+  // 0 -> (*fineElementMDOFs)(fineElement[0],0)
+  //
+  // Bottom Edge:
+  // 1 -> (*fineElementMDOFs)(fineElement[0],4)
+  // 2 -> (*fineElementMDOFs)(fineElement[0],2)
+  // 3 -> (*fineElementMDOFs)(fineElement[1],4)
+  // 4 -> (*fineElementMDOFs)(fineElement[1],2)
+  //
+  // Left Edge:
+  // 5 -> (*fineElementMDOFs)(fineElement[0],7)
+  // 6 -> (*fineElementMDOFs)(fineElement[0],3)
+  // 7 -> (*fineElementMDOFs)(fineElement[2],7)
+  // 8 -> (*fineElementMDOFs)(fineElement[2],3)
+  //
+  // All the rest:
+  // 9 -> (*fineElementMDOFs)(fineElement[3],0)
+  // 10 -> (*fineElementMDOFs)(fineElement[3],1)
+  // 11 -> (*fineElementMDOFs)(fineElement[3],2)
+  // 12 -> (*fineElementMDOFs)(fineElement[3],3)
+  // 13 -> (*fineElementMDOFs)(fineElement[0],5)
+  // 14 -> (*fineElementMDOFs)(fineElement[0],6)
+  // 15 -> (*fineElementMDOFs)(fineElement[1],5)
+  // 16 -> (*fineElementMDOFs)(fineElement[1],6)
+  // 17 -> (*fineElementMDOFs)(fineElement[2],5)
+  // 18 -> (*fineElementMDOFs)(fineElement[2],6)
+  // 19 -> (*fineElementMDOFs)(fineElement[3],5)
+  // 20 -> (*fineElementMDOFs)(fineElement[3],6)
+  // 21 -> (*fineElementMDOFs)(fineElement[0],8)
+  // 22 -> (*fineElementMDOFs)(fineElement[1],8)
+  // 23 -> (*fineElementMDOFs)(fineElement[2],8)
+  // 24 -> (*fineElementMDOFs)(fineElement[3],8)
+  //
+  //*****************************************************************/
+
+  size_t nnz = 0;  // Just to make my copy-paste life easier...
+  Teuchos::ArrayRCP<GO> colPtrV(maxEntriesPerRowV, 0);
+  Teuchos::ArrayRCP<GO> colPtrM(maxEntriesPerRowM, 0);
+  Teuchos::ArrayRCP<SC> valPtrM(maxEntriesPerRowM, 0.);
+
+  Teuchos::ArrayRCP<GO> colPtrP(maxEntriesPerRowP, 0);
+  Teuchos::ArrayRCP<SC> valPtrP(maxEntriesPerRowP, 0.);
+
+  // About which fine-grid elements do we care?
+  GO fineElement[4] = {0, 1, nFineElements, nFineElements + 1};
+
+  GetOStream(Runtime1) << "start building matrices...\n";
+  GetOStream(Runtime1) << "nCoarseElements = " << nCoarseElements << std::endl;
+
+  for (GO coarseElement = 0; coarseElement < totalCoarseElements; coarseElement++) {
+    // We don't really care what is shared with future elements -
+    // we know a priori what needs to be filled for the element
+    // we're dealing with, depending of if it it's on the bottom
+    // edge, the left edge, or in the middle. Thoses should be the
+    // only cases we care about, and they should require a
+    // superset of the work required for an interior node.
+
+    // if (CoarseElement is on bottom edge)
+    if (coarseElement < nCoarseElements) {
+      // fill in the bottom edge of the element patch
+      //  FP = 1
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+      valPtrM[0] = 0.375;
+      colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+      valPtrM[1] = -0.125;
+      colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 4);
+      valPtrM[2] = 0.75;
+
+      nnz = 3;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 4), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+      colPtrV[0] = 2 * colPtrM[0];
+      colPtrV[1] = 2 * colPtrM[1];
+      colPtrV[2] = 2 * colPtrM[2];
+
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 8), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+      colPtrV[0] = 2 * colPtrM[0] + 1;
+      colPtrV[1] = 2 * colPtrM[1] + 1;
+      colPtrV[2] = 2 * colPtrM[2] + 1;
+
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 9), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+      // FPr = 1
+      colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 0);
+      valPtrP[0] = 0.5;
+      colPtrP[1] = (*coarseElementPDOFs)(coarseElement, 1);
+      valPtrP[1] = 0.5;
+
+      nnz = 2;
+      PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0], 1), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+      // FP = 2
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 4);
+      valPtrM[0] = 1.0;
+
+      nnz = 1;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 1), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+      colPtrV[0] = 2 * colPtrM[0];
+
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 2), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+      colPtrV[0] = 2 * colPtrM[0] + 1;
+
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 3), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+      // FPr = 2
+      colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 1);
+      valPtrP[0] = 1.0;
+
+      nnz = 1;
+      PP->insertGlobalValues((*fineElementPDOFs)(fineElement[1], 1), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+      // FP = 3
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+      valPtrM[0] = -0.125;
+      colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+      valPtrM[1] = 0.375;
+      colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 4);
+      valPtrM[2] = 0.75;
 
-                //fill in the bottom left corner
-                // FP = 0
-                colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-                valPtrM[0] = 1.0;
+      nnz = 3;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1], 4), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-                nnz = 1;
-                PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],0),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
+      colPtrV[1] = 2 * colPtrM[1];
+      colPtrV[2] = 2 * colPtrM[2];
 
-                colPtrV[0] = 2*colPtrM[0];
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 8), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-                PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],0),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0] + 1;
+      colPtrV[1] = 2 * colPtrM[1] + 1;
+      colPtrV[2] = 2 * colPtrM[2] + 1;
 
-                colPtrV[0] = 2*colPtrM[0]+1;
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 9), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-                PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],1),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      // FP = 4
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 1);
+      valPtrM[0] = 1.0;
 
-                // FPr = 0
-                colPtrP[0] = (*coarseElementPDOFs)(coarseElement,0);
-                valPtrP[0] = 1.0;
+      nnz = 1;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1], 1), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-                nnz = 1;
-                PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0],0),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
 
-              }//if (coarseElement is on the bottom left corner)
-          }//if (coarseElement is on the bottom edge)
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 2), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        //if (CoarseElement is on left edge)
-        if (coarseElement % (nCoarseElements) == 0)
-          {
-            //fill in the left edge of the element patch
-            // FP = 5
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-            valPtrM[0] = 0.375;
-            colPtrM[1] = (*coarseElementMDOFs)(coarseElement,3);
-            valPtrM[1] = -0.125;
-            colPtrM[2] = (*coarseElementMDOFs)(coarseElement,7);
-            valPtrM[2] = 0.75;
+      colPtrV[0] = 2 * colPtrM[0] + 1;
 
-            nnz = 3;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],7),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 3), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-            colPtrV[0] = 2*colPtrM[0];
-            colPtrV[1] = 2*colPtrM[1];
-            colPtrV[2] = 2*colPtrM[2];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],14),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-            colPtrV[1] = 2*colPtrM[1]+1;
-            colPtrV[2] = 2*colPtrM[2]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],15),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            // FP = 6
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,7);
-            valPtrM[0] = 1.0;
-
-            nnz = 1;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],3),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],6),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],7),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            // FP = 7
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-            valPtrM[0] = -0.125;
-            colPtrM[1] = (*coarseElementMDOFs)(coarseElement,3);
-            valPtrM[1] = 0.375;
-            colPtrM[2] = (*coarseElementMDOFs)(coarseElement,7);
-            valPtrM[2] = 0.75;
-
-            nnz = 3;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2],7),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-            colPtrV[1] = 2*colPtrM[1];
-            colPtrV[2] = 2*colPtrM[2];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],14),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-            colPtrV[1] = 2*colPtrM[1]+1;
-            colPtrV[2] = 2*colPtrM[2]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],15),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-
-            // FP = 8
-            colPtrM[0] = (*coarseElementMDOFs)(coarseElement,3);
-            valPtrM[0] = 1.0;
-
-            nnz = 1;
-            PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2],3),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0];
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],6),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-            colPtrV[0] = 2*colPtrM[0]+1;
-
-            PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],7),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-
-
-            // FPr = 3
-            colPtrP[0] = (*coarseElementPDOFs)(coarseElement,0);
-            valPtrP[0] = 0.5;
-            colPtrP[1] = (*coarseElementPDOFs)(coarseElement,3);
-            valPtrP[1] = 0.5;
-
-            nnz = 2;
-            PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0],3),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-            // FPr = 4
-            colPtrP[0] = (*coarseElementPDOFs)(coarseElement,3);
-            valPtrP[0] = 1.0;
-
-            nnz = 1;
-            PP->insertGlobalValues((*fineElementPDOFs)(fineElement[2],3),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-
-          }//endif (coarseElement is on left edge)
-
-        //fill in the rest of the patch
-        // FP = 9
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,8);
+      // if (CoarseElement is on the bottom left corner)
+      if (coarseElement == 0) {
+        // fill in the bottom left corner
+        //  FP = 0
+        colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
         valPtrM[0] = 1.0;
 
         nnz = 1;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],0),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 0), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0];
+        colPtrV[0] = 2 * colPtrM[0];
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],0),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 0), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0]+1;
+        colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],1),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 1), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        // FP = 10
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[0] = 1.0;
+        // FPr = 0
+        colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 0);
+        valPtrP[0] = 1.0;
 
         nnz = 1;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],1),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+        PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0], 0), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0];
+      }  // if (coarseElement is on the bottom left corner)
+    }    // if (coarseElement is on the bottom edge)
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],2),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    // if (CoarseElement is on left edge)
+    if (coarseElement % (nCoarseElements) == 0) {
+      // fill in the left edge of the element patch
+      //  FP = 5
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+      valPtrM[0] = 0.375;
+      colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 3);
+      valPtrM[1] = -0.125;
+      colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 7);
+      valPtrM[2] = 0.75;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
+      nnz = 3;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 7), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],3),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
+      colPtrV[1] = 2 * colPtrM[1];
+      colPtrV[2] = 2 * colPtrM[2];
 
-        // FP = 11
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[0] = 1.0;
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 14), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        nnz = 1;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],2),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0] + 1;
+      colPtrV[1] = 2 * colPtrM[1] + 1;
+      colPtrV[2] = 2 * colPtrM[2] + 1;
 
-        colPtrV[0] = 2*colPtrM[0];
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 15), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],4),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      // FP = 6
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 7);
+      valPtrM[0] = 1.0;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
+      nnz = 1;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 3), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],5),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
 
-        // FP = 12
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[0] = 1.0;
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 6), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        nnz = 1;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],3),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        colPtrV[0] = 2*colPtrM[0];
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 7), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],6),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      // FP = 7
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+      valPtrM[0] = -0.125;
+      colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 3);
+      valPtrM[1] = 0.375;
+      colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 7);
+      valPtrM[2] = 0.75;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
+      nnz = 3;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2], 7), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],7),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
+      colPtrV[1] = 2 * colPtrM[1];
+      colPtrV[2] = 2 * colPtrM[2];
 
-        // FP = 13
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[0] = 0.375;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[1] = -0.125;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[2] = 0.75;
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 14), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],5),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0] + 1;
+      colPtrV[1] = 2 * colPtrM[1] + 1;
+      colPtrV[2] = 2 * colPtrM[2] + 1;
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 15), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],10),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      // FP = 8
+      colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 3);
+      valPtrM[0] = 1.0;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+      nnz = 1;
+      PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2], 3), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],11),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      colPtrV[0] = 2 * colPtrM[0];
 
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 6), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        // FP = 14
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[0] = -0.125;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[1] = 0.375;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[2] = 0.75;
+      colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],6),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+      PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 7), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+      // FPr = 3
+      colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 0);
+      valPtrP[0] = 0.5;
+      colPtrP[1] = (*coarseElementPDOFs)(coarseElement, 3);
+      valPtrP[1] = 0.5;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],12),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      nnz = 2;
+      PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0], 3), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+      // FPr = 4
+      colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 3);
+      valPtrP[0] = 1.0;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],13),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+      nnz = 1;
+      PP->insertGlobalValues((*fineElementPDOFs)(fineElement[2], 3), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
 
+    }  // endif (coarseElement is on left edge)
 
-        // FP = 15
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[0] = 0.375;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[1] = -0.125;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[2] = 0.75;
+    // fill in the rest of the patch
+    //  FP = 9
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[0] = 1.0;
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1],5),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+    nnz = 1;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 0), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+    colPtrV[0] = 2 * colPtrM[0];
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],10),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 0), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+    colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],11),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 1), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
+    // FP = 10
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[0] = 1.0;
 
-        // FP = 16
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[0] = 0.375;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[1] = -0.125;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[2] = 0.75;
+    nnz = 1;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 1), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1],6),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+    colPtrV[0] = 2 * colPtrM[0];
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 2), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],12),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 3), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],13),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    // FP = 11
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[0] = 1.0;
 
+    nnz = 1;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 2), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        // FP = 17
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[0] = -0.125;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[1] = 0.375;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[2] = 0.75;
+    colPtrV[0] = 2 * colPtrM[0];
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2],5),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 4), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+    colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],10),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 5), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+    // FP = 12
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[0] = 1.0;
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],11),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    nnz = 1;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 3), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
+    colPtrV[0] = 2 * colPtrM[0];
 
-        // FP = 18
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[0] = -0.125;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[1] = 0.375;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[2] = 0.75;
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 6), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2],6),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
+    colPtrV[0] = 2 * colPtrM[0] + 1;
 
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 7), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],12),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    // FP = 13
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[0] = 0.375;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[1] = -0.125;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[2] = 0.75;
 
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 5), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],13),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
 
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 10), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        // FP = 19
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[0] = -0.125;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[1] = 0.375;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[2] = 0.75;
-
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],5),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],10),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],11),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        // FP = 20
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[0] = 0.375;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[1] = -0.125;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[2] = 0.75;
-
-        nnz = 3;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],6),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],12),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],13),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-
-        // FP = 21
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-        valPtrM[0] = 0.140625;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[1] = -0.046875;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[2] = 0.015625;
-        colPtrM[3] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[3] = -0.046875;
-        colPtrM[4] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[4] = 0.28125;
-        colPtrM[5] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[5] = -0.09375;
-        colPtrM[6] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[6] = -0.09375;
-        colPtrM[7] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[7] = 0.28125;
-        colPtrM[8] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[8] = 0.5625;
-
-        nnz = 9;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0],8),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-        colPtrV[3] = 2*colPtrM[3];
-        colPtrV[4] = 2*colPtrM[4];
-        colPtrV[5] = 2*colPtrM[5];
-        colPtrV[6] = 2*colPtrM[6];
-        colPtrV[7] = 2*colPtrM[7];
-        colPtrV[8] = 2*colPtrM[8];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],16),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-        colPtrV[3] = 2*colPtrM[3]+1;
-        colPtrV[4] = 2*colPtrM[4]+1;
-        colPtrV[5] = 2*colPtrM[5]+1;
-        colPtrV[6] = 2*colPtrM[6]+1;
-        colPtrV[7] = 2*colPtrM[7]+1;
-        colPtrV[8] = 2*colPtrM[8]+1;
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0],17),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        // FP = 22
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-        valPtrM[0] = -0.046875;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[1] = 0.140625;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[2] = -0.046875;
-        colPtrM[3] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[3] = 0.015625;
-        colPtrM[4] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[4] = 0.28125;
-        colPtrM[5] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[5] = 0.28125;
-        colPtrM[6] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[6] = -0.09375;
-        colPtrM[7] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[7] = -0.09375;
-        colPtrM[8] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[8] = 0.5625;
-
-        nnz = 9;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1],8),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-        colPtrV[3] = 2*colPtrM[3];
-        colPtrV[4] = 2*colPtrM[4];
-        colPtrV[5] = 2*colPtrM[5];
-        colPtrV[6] = 2*colPtrM[6];
-        colPtrV[7] = 2*colPtrM[7];
-        colPtrV[8] = 2*colPtrM[8];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],16),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-        colPtrV[3] = 2*colPtrM[3]+1;
-        colPtrV[4] = 2*colPtrM[4]+1;
-        colPtrV[5] = 2*colPtrM[5]+1;
-        colPtrV[6] = 2*colPtrM[6]+1;
-        colPtrV[7] = 2*colPtrM[7]+1;
-        colPtrV[8] = 2*colPtrM[8]+1;
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1],17),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        // FP = 23
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-        valPtrM[0] = -0.046875;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[1] = 0.015625;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[2] = -0.046875;
-        colPtrM[3] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[3] = 0.140625;
-        colPtrM[4] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[4] = -0.09375;
-        colPtrM[5] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[5] = -0.09375;
-        colPtrM[6] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[6] = 0.28125;
-        colPtrM[7] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[7] = 0.28125;
-        colPtrM[8] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[8] = 0.5625;
-
-        nnz = 9;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2],8),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-        colPtrV[3] = 2*colPtrM[3];
-        colPtrV[4] = 2*colPtrM[4];
-        colPtrV[5] = 2*colPtrM[5];
-        colPtrV[6] = 2*colPtrM[6];
-        colPtrV[7] = 2*colPtrM[7];
-        colPtrV[8] = 2*colPtrM[8];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],16),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-        colPtrV[3] = 2*colPtrM[3]+1;
-        colPtrV[4] = 2*colPtrM[4]+1;
-        colPtrV[5] = 2*colPtrM[5]+1;
-        colPtrV[6] = 2*colPtrM[6]+1;
-        colPtrV[7] = 2*colPtrM[7]+1;
-        colPtrV[8] = 2*colPtrM[8]+1;
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2],17),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        // FP = 24
-        colPtrM[0] = (*coarseElementMDOFs)(coarseElement,0);
-        valPtrM[0] = 0.015625;
-        colPtrM[1] = (*coarseElementMDOFs)(coarseElement,1);
-        valPtrM[1] = -0.046875;
-        colPtrM[2] = (*coarseElementMDOFs)(coarseElement,2);
-        valPtrM[2] = 0.140625;
-        colPtrM[3] = (*coarseElementMDOFs)(coarseElement,3);
-        valPtrM[3] = -0.046875;
-        colPtrM[4] = (*coarseElementMDOFs)(coarseElement,4);
-        valPtrM[4] = -0.09375;
-        colPtrM[5] = (*coarseElementMDOFs)(coarseElement,5);
-        valPtrM[5] = 0.28125;
-        colPtrM[6] = (*coarseElementMDOFs)(coarseElement,6);
-        valPtrM[6] = 0.28125;
-        colPtrM[7] = (*coarseElementMDOFs)(coarseElement,7);
-        valPtrM[7] = -0.09375;
-        colPtrM[8] = (*coarseElementMDOFs)(coarseElement,8);
-        valPtrM[8] = 0.5625;
-
-        nnz = 9;
-        PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3],8),colPtrM.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0];
-        colPtrV[1] = 2*colPtrM[1];
-        colPtrV[2] = 2*colPtrM[2];
-        colPtrV[3] = 2*colPtrM[3];
-        colPtrV[4] = 2*colPtrM[4];
-        colPtrV[5] = 2*colPtrM[5];
-        colPtrV[6] = 2*colPtrM[6];
-        colPtrV[7] = 2*colPtrM[7];
-        colPtrV[8] = 2*colPtrM[8];
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],16),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-        colPtrV[0] = 2*colPtrM[0]+1;
-        colPtrV[1] = 2*colPtrM[1]+1;
-        colPtrV[2] = 2*colPtrM[2]+1;
-        colPtrV[3] = 2*colPtrM[3]+1;
-        colPtrV[4] = 2*colPtrM[4]+1;
-        colPtrV[5] = 2*colPtrM[5]+1;
-        colPtrV[6] = 2*colPtrM[6]+1;
-        colPtrV[7] = 2*colPtrM[7]+1;
-        colPtrV[8] = 2*colPtrM[8]+1;
-
-        PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3],17),colPtrV.view(0,nnz),valPtrM.view(0,nnz));
-
-
-        // FPr = 5
-        colPtrP[0] = (*coarseElementPDOFs)(coarseElement,0);
-        valPtrP[0] = 0.25;
-        colPtrP[1] = (*coarseElementPDOFs)(coarseElement,1);
-        valPtrP[1] = 0.25;
-        colPtrP[2] = (*coarseElementPDOFs)(coarseElement,2);
-        valPtrP[2] = 0.25;
-        colPtrP[3] = (*coarseElementPDOFs)(coarseElement,3);
-        valPtrP[3] = 0.25;
-
-        nnz = 4;
-        PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0],2),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-        // FPr = 6
-        colPtrP[0] = (*coarseElementPDOFs)(coarseElement,1);
-        valPtrP[0] = 0.5;
-        colPtrP[1] = (*coarseElementPDOFs)(coarseElement,2);
-        valPtrP[1] = 0.5;
-
-        nnz = 2;
-        PP->insertGlobalValues((*fineElementPDOFs)(fineElement[1],2),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
-
-        // FPr = 7
-        colPtrP[0] = (*coarseElementPDOFs)(coarseElement,2);
-        valPtrP[0] = 1.0;
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
 
-        nnz = 1;
-        PP->insertGlobalValues((*fineElementPDOFs)(fineElement[3],2),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 11), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        // FPr = 8
-        colPtrP[0] = (*coarseElementPDOFs)(coarseElement,2);
-        valPtrP[0] = 0.5;
-        colPtrP[1] = (*coarseElementPDOFs)(coarseElement,3);
-        valPtrP[1] = 0.5;
+    // FP = 14
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[0] = -0.125;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[1] = 0.375;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[2] = 0.75;
 
-        nnz = 2;
-        PP->insertGlobalValues((*fineElementPDOFs)(fineElement[3],3),colPtrP.view(0,nnz),valPtrP.view(0,nnz));
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 6), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
 
-        // Update counters:
-        if ((coarseElement+1) % (nCoarseElements) == 0)//if the end of a row of c.g. elements
-          {
-            fineElement[0] = fineElement[3]+1;
-            fineElement[1] = fineElement[0]+1;
-            fineElement[2] = fineElement[0]+nFineElements;
-            fineElement[3] = fineElement[2]+1;
-          }
-        else
-          {
-            fineElement[0] = fineElement[1]+1;
-            fineElement[1] = fineElement[0]+1;
-            fineElement[2] = fineElement[3]+1;
-            fineElement[3] = fineElement[2]+1;
-          }
-      }// END OF BUILD LOOP
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 12), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
 
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 13), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-    //Loop over V rows
-    for (GO VRow = 0; VRow < fNV; VRow++)
-      {
-        Teuchos::ArrayView<const LO> colPtr;
-        Teuchos::ArrayView<const SC> valPtr;
+    // FP = 15
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[0] = 0.375;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[1] = -0.125;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[2] = 0.75;
 
-        PV->getGlobalRowView(VRow,colPtr,valPtr);
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1], 5), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-        //Can be directly inserted!
-        P->insertGlobalValues(VRow,colPtr,valPtr);
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
 
-      }
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 10), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-    //Loop over P rows
-    for (GO PRow = 0; PRow < fNP; PRow++)
-      {
-        Teuchos::ArrayView<const LO> colPtr;
-        Teuchos::ArrayView<const SC> valPtr;
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
 
-        //Now do pressure column:
-        PP->getGlobalRowView(PRow,colPtr,valPtr);
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 11), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(),nV);
-        for (LO jj=0; jj<colPtr.size(); jj++)
-          {
-            newColPtr[jj] += colPtr[jj];
-          }
+    // FP = 16
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[0] = 0.375;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[1] = -0.125;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[2] = 0.75;
 
-        //Insert into A
-        P->insertGlobalValues(PRow+fNV,newColPtr.view(0,colPtr.size()),valPtr);
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1], 6), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
 
-      }
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
 
-    //Loop over M rows
-    for (GO MRow = 0; MRow < fNM; MRow++)
-      {
-        Teuchos::ArrayView<const LO> colPtr;
-        Teuchos::ArrayView<const SC> valPtr;
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 12), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        //Now do magnetics column:
-        PM->getGlobalRowView(MRow,colPtr,valPtr);
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
 
-        Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(),nV+nP);
-        for (LO jj=0; jj<colPtr.size(); jj++)
-          {
-            newColPtr[jj] += colPtr[jj];
-          }
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 13), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
 
-        //Insert into A
-        P->insertGlobalValues(MRow+fNV+fNP,newColPtr.view(0,colPtr.size()),valPtr);
+    // FP = 17
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[0] = -0.125;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[1] = 0.375;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[2] = 0.75;
+
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2], 5), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
 
-      }
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 10), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 11), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 18
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[0] = -0.125;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[1] = 0.375;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[2] = 0.75;
+
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2], 6), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 12), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 13), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 19
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[0] = -0.125;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[1] = 0.375;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[2] = 0.75;
+
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 5), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 10), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 11), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 20
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[0] = 0.375;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[1] = -0.125;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[2] = 0.75;
+
+    nnz = 3;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 6), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 12), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 13), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 21
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+    valPtrM[0] = 0.140625;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[1] = -0.046875;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[2] = 0.015625;
+    colPtrM[3] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[3] = -0.046875;
+    colPtrM[4] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[4] = 0.28125;
+    colPtrM[5] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[5] = -0.09375;
+    colPtrM[6] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[6] = -0.09375;
+    colPtrM[7] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[7] = 0.28125;
+    colPtrM[8] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[8] = 0.5625;
+
+    nnz = 9;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[0], 8), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+    colPtrV[3] = 2 * colPtrM[3];
+    colPtrV[4] = 2 * colPtrM[4];
+    colPtrV[5] = 2 * colPtrM[5];
+    colPtrV[6] = 2 * colPtrM[6];
+    colPtrV[7] = 2 * colPtrM[7];
+    colPtrV[8] = 2 * colPtrM[8];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 16), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+    colPtrV[3] = 2 * colPtrM[3] + 1;
+    colPtrV[4] = 2 * colPtrM[4] + 1;
+    colPtrV[5] = 2 * colPtrM[5] + 1;
+    colPtrV[6] = 2 * colPtrM[6] + 1;
+    colPtrV[7] = 2 * colPtrM[7] + 1;
+    colPtrV[8] = 2 * colPtrM[8] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[0], 17), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 22
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+    valPtrM[0] = -0.046875;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[1] = 0.140625;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[2] = -0.046875;
+    colPtrM[3] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[3] = 0.015625;
+    colPtrM[4] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[4] = 0.28125;
+    colPtrM[5] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[5] = 0.28125;
+    colPtrM[6] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[6] = -0.09375;
+    colPtrM[7] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[7] = -0.09375;
+    colPtrM[8] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[8] = 0.5625;
+
+    nnz = 9;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[1], 8), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+    colPtrV[3] = 2 * colPtrM[3];
+    colPtrV[4] = 2 * colPtrM[4];
+    colPtrV[5] = 2 * colPtrM[5];
+    colPtrV[6] = 2 * colPtrM[6];
+    colPtrV[7] = 2 * colPtrM[7];
+    colPtrV[8] = 2 * colPtrM[8];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 16), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+    colPtrV[3] = 2 * colPtrM[3] + 1;
+    colPtrV[4] = 2 * colPtrM[4] + 1;
+    colPtrV[5] = 2 * colPtrM[5] + 1;
+    colPtrV[6] = 2 * colPtrM[6] + 1;
+    colPtrV[7] = 2 * colPtrM[7] + 1;
+    colPtrV[8] = 2 * colPtrM[8] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[1], 17), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 23
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+    valPtrM[0] = -0.046875;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[1] = 0.015625;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[2] = -0.046875;
+    colPtrM[3] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[3] = 0.140625;
+    colPtrM[4] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[4] = -0.09375;
+    colPtrM[5] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[5] = -0.09375;
+    colPtrM[6] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[6] = 0.28125;
+    colPtrM[7] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[7] = 0.28125;
+    colPtrM[8] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[8] = 0.5625;
+
+    nnz = 9;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[2], 8), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+    colPtrV[3] = 2 * colPtrM[3];
+    colPtrV[4] = 2 * colPtrM[4];
+    colPtrV[5] = 2 * colPtrM[5];
+    colPtrV[6] = 2 * colPtrM[6];
+    colPtrV[7] = 2 * colPtrM[7];
+    colPtrV[8] = 2 * colPtrM[8];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 16), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+    colPtrV[3] = 2 * colPtrM[3] + 1;
+    colPtrV[4] = 2 * colPtrM[4] + 1;
+    colPtrV[5] = 2 * colPtrM[5] + 1;
+    colPtrV[6] = 2 * colPtrM[6] + 1;
+    colPtrV[7] = 2 * colPtrM[7] + 1;
+    colPtrV[8] = 2 * colPtrM[8] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[2], 17), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FP = 24
+    colPtrM[0] = (*coarseElementMDOFs)(coarseElement, 0);
+    valPtrM[0] = 0.015625;
+    colPtrM[1] = (*coarseElementMDOFs)(coarseElement, 1);
+    valPtrM[1] = -0.046875;
+    colPtrM[2] = (*coarseElementMDOFs)(coarseElement, 2);
+    valPtrM[2] = 0.140625;
+    colPtrM[3] = (*coarseElementMDOFs)(coarseElement, 3);
+    valPtrM[3] = -0.046875;
+    colPtrM[4] = (*coarseElementMDOFs)(coarseElement, 4);
+    valPtrM[4] = -0.09375;
+    colPtrM[5] = (*coarseElementMDOFs)(coarseElement, 5);
+    valPtrM[5] = 0.28125;
+    colPtrM[6] = (*coarseElementMDOFs)(coarseElement, 6);
+    valPtrM[6] = 0.28125;
+    colPtrM[7] = (*coarseElementMDOFs)(coarseElement, 7);
+    valPtrM[7] = -0.09375;
+    colPtrM[8] = (*coarseElementMDOFs)(coarseElement, 8);
+    valPtrM[8] = 0.5625;
+
+    nnz = 9;
+    PM->insertGlobalValues((*fineElementMDOFs)(fineElement[3], 8), colPtrM.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0];
+    colPtrV[1] = 2 * colPtrM[1];
+    colPtrV[2] = 2 * colPtrM[2];
+    colPtrV[3] = 2 * colPtrM[3];
+    colPtrV[4] = 2 * colPtrM[4];
+    colPtrV[5] = 2 * colPtrM[5];
+    colPtrV[6] = 2 * colPtrM[6];
+    colPtrV[7] = 2 * colPtrM[7];
+    colPtrV[8] = 2 * colPtrM[8];
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 16), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    colPtrV[0] = 2 * colPtrM[0] + 1;
+    colPtrV[1] = 2 * colPtrM[1] + 1;
+    colPtrV[2] = 2 * colPtrM[2] + 1;
+    colPtrV[3] = 2 * colPtrM[3] + 1;
+    colPtrV[4] = 2 * colPtrM[4] + 1;
+    colPtrV[5] = 2 * colPtrM[5] + 1;
+    colPtrV[6] = 2 * colPtrM[6] + 1;
+    colPtrV[7] = 2 * colPtrM[7] + 1;
+    colPtrV[8] = 2 * colPtrM[8] + 1;
+
+    PV->insertGlobalValues((*fineElementVDOFs)(fineElement[3], 17), colPtrV.view(0, nnz), valPtrM.view(0, nnz));
+
+    // FPr = 5
+    colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 0);
+    valPtrP[0] = 0.25;
+    colPtrP[1] = (*coarseElementPDOFs)(coarseElement, 1);
+    valPtrP[1] = 0.25;
+    colPtrP[2] = (*coarseElementPDOFs)(coarseElement, 2);
+    valPtrP[2] = 0.25;
+    colPtrP[3] = (*coarseElementPDOFs)(coarseElement, 3);
+    valPtrP[3] = 0.25;
+
+    nnz = 4;
+    PP->insertGlobalValues((*fineElementPDOFs)(fineElement[0], 2), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+    // FPr = 6
+    colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 1);
+    valPtrP[0] = 0.5;
+    colPtrP[1] = (*coarseElementPDOFs)(coarseElement, 2);
+    valPtrP[1] = 0.5;
+
+    nnz = 2;
+    PP->insertGlobalValues((*fineElementPDOFs)(fineElement[1], 2), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+    // FPr = 7
+    colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 2);
+    valPtrP[0] = 1.0;
+
+    nnz = 1;
+    PP->insertGlobalValues((*fineElementPDOFs)(fineElement[3], 2), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+    // FPr = 8
+    colPtrP[0] = (*coarseElementPDOFs)(coarseElement, 2);
+    valPtrP[0] = 0.5;
+    colPtrP[1] = (*coarseElementPDOFs)(coarseElement, 3);
+    valPtrP[1] = 0.5;
+
+    nnz = 2;
+    PP->insertGlobalValues((*fineElementPDOFs)(fineElement[3], 3), colPtrP.view(0, nnz), valPtrP.view(0, nnz));
+
+    // Update counters:
+    if ((coarseElement + 1) % (nCoarseElements) == 0)  // if the end of a row of c.g. elements
+    {
+      fineElement[0] = fineElement[3] + 1;
+      fineElement[1] = fineElement[0] + 1;
+      fineElement[2] = fineElement[0] + nFineElements;
+      fineElement[3] = fineElement[2] + 1;
+    } else {
+      fineElement[0] = fineElement[1] + 1;
+      fineElement[1] = fineElement[0] + 1;
+      fineElement[2] = fineElement[3] + 1;
+      fineElement[3] = fineElement[2] + 1;
+    }
+  }  // END OF BUILD LOOP
+
+  // Loop over V rows
+  for (GO VRow = 0; VRow < fNV; VRow++) {
+    Teuchos::ArrayView<const LO> colPtr;
+    Teuchos::ArrayView<const SC> valPtr;
+
+    PV->getGlobalRowView(VRow, colPtr, valPtr);
+
+    // Can be directly inserted!
+    P->insertGlobalValues(VRow, colPtr, valPtr);
+  }
 
+  // Loop over P rows
+  for (GO PRow = 0; PRow < fNP; PRow++) {
+    Teuchos::ArrayView<const LO> colPtr;
+    Teuchos::ArrayView<const SC> valPtr;
 
+    // Now do pressure column:
+    PP->getGlobalRowView(PRow, colPtr, valPtr);
 
+    Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(), nV);
+    for (LO jj = 0; jj < colPtr.size(); jj++) {
+      newColPtr[jj] += colPtr[jj];
+    }
 
+    // Insert into A
+    P->insertGlobalValues(PRow + fNV, newColPtr.view(0, colPtr.size()), valPtr);
+  }
 
+  // Loop over M rows
+  for (GO MRow = 0; MRow < fNM; MRow++) {
+    Teuchos::ArrayView<const LO> colPtr;
+    Teuchos::ArrayView<const SC> valPtr;
 
+    // Now do magnetics column:
+    PM->getGlobalRowView(MRow, colPtr, valPtr);
 
-    // Fill-complete all matrices
-    PV->fillComplete(colMapforPV,rowMapforPV);
-    PP->fillComplete(colMapforPP,rowMapforPP);
-    PM->fillComplete(colMapforPM,rowMapforPM);
-    P->fillComplete();
+    Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(), nV + nP);
+    for (LO jj = 0; jj < colPtr.size(); jj++) {
+      newColPtr[jj] += colPtr[jj];
+    }
+
+    // Insert into A
+    P->insertGlobalValues(MRow + fNV + fNP, newColPtr.view(0, colPtr.size()), valPtr);
+  }
 
-    // Set prolongators on the coarse grid
-    Set(coarseLevel,"PV",PV);
-    Set(coarseLevel,"PP",PP);
-    Set(coarseLevel,"PM",PM);
-    Set(coarseLevel,"P" ,P );
+  // Fill-complete all matrices
+  PV->fillComplete(colMapforPV, rowMapforPV);
+  PP->fillComplete(colMapforPP, rowMapforPP);
+  PM->fillComplete(colMapforPM, rowMapforPM);
+  P->fillComplete();
 
-    Set(coarseLevel,"NV",nV);
-    Set(coarseLevel,"NP",nP);
-    Set(coarseLevel,"NM",nM);
+  // Set prolongators on the coarse grid
+  Set(coarseLevel, "PV", PV);
+  Set(coarseLevel, "PP", PP);
+  Set(coarseLevel, "PM", PM);
+  Set(coarseLevel, "P", P);
 
-  }//end buildp
+  Set(coarseLevel, "NV", nV);
+  Set(coarseLevel, "NP", nP);
+  Set(coarseLevel, "NM", nM);
 
+}  // end buildp
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_GEOINTERPFACTORY_SHORT
-#endif // MUELU_GEOINTERPFACTORY_DEF_HPP
+#endif  // MUELU_GEOINTERPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_decl.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_decl.hpp
index 2282d3345398..cf7a87f6ad5e 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_decl.hpp
@@ -58,69 +58,62 @@
 #include "MueLu_FactoryBase_fwd.hpp"
 
 namespace MueLu {
-  /*!
-    @class MHDRAPFactory
-    @brief Factory for building coarse matrices.
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MHDRAPFactory : public TwoLevelFactoryBase {
+/*!
+  @class MHDRAPFactory
+  @brief Factory for building coarse matrices.
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MHDRAPFactory : public TwoLevelFactoryBase {
 #undef MUELU_MHDRAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    MHDRAPFactory();
+  MHDRAPFactory();
 
-    virtual ~MHDRAPFactory() { }
+  virtual ~MHDRAPFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
-    void Build(Level &fineLevel, Level &coarseLevel) const;
-    //@}
+  //! @name Build methods.
+  //@{
+  void Build(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //! @name Handling of user-defined transfer factories
-    //@{
+  //! @name Handling of user-defined transfer factories
+  //@{
 
-    //! Indicate that the restriction operator action should be implicitly defined by the transpose of the prolongator.
-    void SetImplicitTranspose(bool const &implicit) {
-      implicitTranspose_ = implicit;
-    }
+  //! Indicate that the restriction operator action should be implicitly defined by the transpose of the prolongator.
+  void SetImplicitTranspose(bool const &implicit) {
+    implicitTranspose_ = implicit;
+  }
 
+  //@}
 
+  //! @name internal print methods.
+  //    static std::string PerfUtils::PrintMatrixInfo(const Matrix & Ac, const std::string & msgTag);
 
+  static std::string PrintLoadBalancingInfo(const Matrix &Ac, const std::string &msgTag);
 
-    //@}
+ private:
+  //! If true, the action of the restriction operator action is implicitly defined by the transpose of the prolongator.
+  bool implicitTranspose_;
 
-    //! @name internal print methods.
-//    static std::string PerfUtils::PrintMatrixInfo(const Matrix & Ac, const std::string & msgTag);
+};  // class MHDRAPFactory
 
-    static std::string PrintLoadBalancingInfo(const Matrix & Ac, const std::string & msgTag);
-
-  private:
-
-
-    //! If true, the action of the restriction operator action is implicitly defined by the transpose of the prolongator.
-    bool implicitTranspose_;
-
-
-
-  }; //class MHDRAPFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MHDRAPFACTORY_SHORT
-#endif // MUELU_MHDRAPFACTORY_DECL_HPP
+#endif  // MUELU_MHDRAPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_def.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_def.hpp
index 4b28fb2a2ce1..ef62fc85b4e4 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_MHDRAPFactory_def.hpp
@@ -60,167 +60,153 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MHDRAPFactory()
-    : implicitTranspose_(true) { }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-
-    if (implicitTranspose_ == false)
-      {
-        Input(coarseLevel, "R" );
-        Input(coarseLevel, "RV");
-        Input(coarseLevel, "RP");
-        Input(coarseLevel, "RM");
-      }
-
-    Input(fineLevel, "A"  );
-    Input(fineLevel, "A00");
-    Input(fineLevel, "A01");
-    Input(fineLevel, "A02");
-    Input(fineLevel, "A10");
-    Input(fineLevel, "A11");
-    Input(fineLevel, "A12");
-    Input(fineLevel, "A20");
-    Input(fineLevel, "A21");
-    Input(fineLevel, "A22");
-
-    Input(coarseLevel, "P" );
-    Input(coarseLevel, "PV");
-    Input(coarseLevel, "PP");
-    Input(coarseLevel, "PM");
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MHDRAPFactory()
+  : implicitTranspose_(true) {}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  if (implicitTranspose_ == false) {
+    Input(coarseLevel, "R");
+    Input(coarseLevel, "RV");
+    Input(coarseLevel, "RP");
+    Input(coarseLevel, "RM");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const { // FIXME make fineLevel const
-    {
-      FactoryMonitor m(*this, "Computing Ac", coarseLevel);
-
-      //
-      // Inputs: A, P
-      //
-
-      //DEBUG
-      //Teuchos::FancyOStream fout(*GetOStream(Runtime1));
-      //coarseLevel.print(fout,Teuchos::VERB_HIGH);
-
-
-      RCP<Matrix> A   = Get< RCP<Matrix> >(fineLevel, "A"  );
-      RCP<Matrix> A00 = Get< RCP<Matrix> >(fineLevel, "A00");
-      RCP<Matrix> A01 = Get< RCP<Matrix> >(fineLevel, "A01");
-      RCP<Matrix> A02 = Get< RCP<Matrix> >(fineLevel, "A02");
-      RCP<Matrix> A10 = Get< RCP<Matrix> >(fineLevel, "A10");
-      RCP<Matrix> A11 = Get< RCP<Matrix> >(fineLevel, "A11");
-      RCP<Matrix> A12 = Get< RCP<Matrix> >(fineLevel, "A12");
-      RCP<Matrix> A20 = Get< RCP<Matrix> >(fineLevel, "A20");
-      RCP<Matrix> A21 = Get< RCP<Matrix> >(fineLevel, "A21");
-      RCP<Matrix> A22 = Get< RCP<Matrix> >(fineLevel, "A22");
-
-      RCP<Matrix> P  = Get< RCP<Matrix> >(coarseLevel, "P" );
-      RCP<Matrix> PV = Get< RCP<Matrix> >(coarseLevel, "PV");
-      RCP<Matrix> PP = Get< RCP<Matrix> >(coarseLevel, "PP");
-      RCP<Matrix> PM = Get< RCP<Matrix> >(coarseLevel, "PM");
-
-      //
-      // Build Ac = RAP
-      //
-
-      RCP<Matrix> AP;
-      RCP<Matrix> AP00;
-      RCP<Matrix> AP01;
-      RCP<Matrix> AP02;
-      RCP<Matrix> AP10;
-      RCP<Matrix> AP11;
-      RCP<Matrix> AP12;
-      RCP<Matrix> AP20;
-      RCP<Matrix> AP21;
-      RCP<Matrix> AP22;
-
-      {
-        SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
-
-        AP   = Utils::Multiply(*A  , false, *P , false, AP, GetOStream(Statistics2));
-        AP00 = Utils::Multiply(*A00, false, *PV, false, AP00, GetOStream(Statistics2));
-        AP01 = Utils::Multiply(*A01, false, *PP, false, AP01, GetOStream(Statistics2));
-        AP02 = Utils::Multiply(*A02, false, *PM, false, AP02, GetOStream(Statistics2));
-        AP10 = Utils::Multiply(*A10, false, *PV, false, AP10, GetOStream(Statistics2));
-        AP11 = Utils::Multiply(*A11, false, *PP, false, AP11, GetOStream(Statistics2));
-        AP12 = Utils::Multiply(*A12, false, *PM, false, AP12, GetOStream(Statistics2));
-        AP20 = Utils::Multiply(*A20, false, *PV, false, AP20, GetOStream(Statistics2));
-        AP21 = Utils::Multiply(*A21, false, *PP, false, AP21, GetOStream(Statistics2));
-        AP22 = Utils::Multiply(*A22, false, *PM, false, AP22, GetOStream(Statistics2));
-      }
-
-      RCP<Matrix> Ac;
-      RCP<Matrix> Ac00;
-      RCP<Matrix> Ac01;
-      RCP<Matrix> Ac02;
-      RCP<Matrix> Ac10;
-      RCP<Matrix> Ac11;
-      RCP<Matrix> Ac12;
-      RCP<Matrix> Ac20;
-      RCP<Matrix> Ac21;
-      RCP<Matrix> Ac22;
-
-      if (implicitTranspose_)
-        {
-          SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
-
-          Ac   = Utils::Multiply(*P , true, *AP  , false, Ac, GetOStream(Statistics2));
-          Ac00 = Utils::Multiply(*PV, true, *AP00, false, Ac00, GetOStream(Statistics2));
-          Ac01 = Utils::Multiply(*PV, true, *AP01, false, Ac01, GetOStream(Statistics2));
-          Ac02 = Utils::Multiply(*PV, true, *AP02, false, Ac02, GetOStream(Statistics2));
-          Ac10 = Utils::Multiply(*PP, true, *AP10, false, Ac10, GetOStream(Statistics2));
-          Ac11 = Utils::Multiply(*PP, true, *AP11, false, Ac11, GetOStream(Statistics2));
-          Ac12 = Utils::Multiply(*PP, true, *AP12, false, Ac12, GetOStream(Statistics2));
-          Ac20 = Utils::Multiply(*PM, true, *AP20, false, Ac20, GetOStream(Statistics2));
-          Ac21 = Utils::Multiply(*PM, true, *AP21, false, Ac21, GetOStream(Statistics2));
-          Ac22 = Utils::Multiply(*PM, true, *AP22, false, Ac22, GetOStream(Statistics2));
-
-        }
-      else
-        {
-
-          SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
-
-          RCP<Matrix> R  = Get< RCP<Matrix> >(coarseLevel, "R" );
-          RCP<Matrix> RV = Get< RCP<Matrix> >(coarseLevel, "RV");
-          RCP<Matrix> RP = Get< RCP<Matrix> >(coarseLevel, "RP");
-          RCP<Matrix> RM = Get< RCP<Matrix> >(coarseLevel, "RM");
-
-          Ac   = Utils::Multiply(*R , false, *AP  , false, Ac, GetOStream(Statistics2));
-          Ac00 = Utils::Multiply(*RV, false, *AP00, false, Ac00, GetOStream(Statistics2));
-          Ac01 = Utils::Multiply(*RV, false, *AP01, false, Ac01, GetOStream(Statistics2));
-          Ac02 = Utils::Multiply(*RV, false, *AP02, false, Ac02, GetOStream(Statistics2));
-          Ac10 = Utils::Multiply(*RP, false, *AP10, false, Ac10, GetOStream(Statistics2));
-          Ac11 = Utils::Multiply(*RP, false, *AP11, false, Ac11, GetOStream(Statistics2));
-          Ac12 = Utils::Multiply(*RP, false, *AP12, false, Ac12, GetOStream(Statistics2));
-          Ac20 = Utils::Multiply(*RM, false, *AP20, false, Ac20, GetOStream(Statistics2));
-          Ac21 = Utils::Multiply(*RM, false, *AP21, false, Ac21, GetOStream(Statistics2));
-          Ac22 = Utils::Multiply(*RM, false, *AP22, false, Ac22, GetOStream(Statistics2));
-
-        }
-      // FINISHED MAKING COARSE BLOCKS
-
-      Set(coarseLevel, "A"  , Ac  );
-      Set(coarseLevel, "A00", Ac00);
-      Set(coarseLevel, "A01", Ac01);
-      Set(coarseLevel, "A02", Ac02);
-      Set(coarseLevel, "A10", Ac10);
-      Set(coarseLevel, "A11", Ac11);
-      Set(coarseLevel, "A12", Ac12);
-      Set(coarseLevel, "A20", Ac20);
-      Set(coarseLevel, "A21", Ac21);
-      Set(coarseLevel, "A22", Ac22);
-
+  Input(fineLevel, "A");
+  Input(fineLevel, "A00");
+  Input(fineLevel, "A01");
+  Input(fineLevel, "A02");
+  Input(fineLevel, "A10");
+  Input(fineLevel, "A11");
+  Input(fineLevel, "A12");
+  Input(fineLevel, "A20");
+  Input(fineLevel, "A21");
+  Input(fineLevel, "A22");
+
+  Input(coarseLevel, "P");
+  Input(coarseLevel, "PV");
+  Input(coarseLevel, "PP");
+  Input(coarseLevel, "PM");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {  // FIXME make fineLevel const
+  {
+    FactoryMonitor m(*this, "Computing Ac", coarseLevel);
+
+    //
+    // Inputs: A, P
+    //
+
+    // DEBUG
+    // Teuchos::FancyOStream fout(*GetOStream(Runtime1));
+    // coarseLevel.print(fout,Teuchos::VERB_HIGH);
+
+    RCP<Matrix> A   = Get<RCP<Matrix> >(fineLevel, "A");
+    RCP<Matrix> A00 = Get<RCP<Matrix> >(fineLevel, "A00");
+    RCP<Matrix> A01 = Get<RCP<Matrix> >(fineLevel, "A01");
+    RCP<Matrix> A02 = Get<RCP<Matrix> >(fineLevel, "A02");
+    RCP<Matrix> A10 = Get<RCP<Matrix> >(fineLevel, "A10");
+    RCP<Matrix> A11 = Get<RCP<Matrix> >(fineLevel, "A11");
+    RCP<Matrix> A12 = Get<RCP<Matrix> >(fineLevel, "A12");
+    RCP<Matrix> A20 = Get<RCP<Matrix> >(fineLevel, "A20");
+    RCP<Matrix> A21 = Get<RCP<Matrix> >(fineLevel, "A21");
+    RCP<Matrix> A22 = Get<RCP<Matrix> >(fineLevel, "A22");
+
+    RCP<Matrix> P  = Get<RCP<Matrix> >(coarseLevel, "P");
+    RCP<Matrix> PV = Get<RCP<Matrix> >(coarseLevel, "PV");
+    RCP<Matrix> PP = Get<RCP<Matrix> >(coarseLevel, "PP");
+    RCP<Matrix> PM = Get<RCP<Matrix> >(coarseLevel, "PM");
+
+    //
+    // Build Ac = RAP
+    //
+
+    RCP<Matrix> AP;
+    RCP<Matrix> AP00;
+    RCP<Matrix> AP01;
+    RCP<Matrix> AP02;
+    RCP<Matrix> AP10;
+    RCP<Matrix> AP11;
+    RCP<Matrix> AP12;
+    RCP<Matrix> AP20;
+    RCP<Matrix> AP21;
+    RCP<Matrix> AP22;
 
+    {
+      SubFactoryMonitor subM(*this, "MxM: A x P", coarseLevel);
+
+      AP   = Utils::Multiply(*A, false, *P, false, AP, GetOStream(Statistics2));
+      AP00 = Utils::Multiply(*A00, false, *PV, false, AP00, GetOStream(Statistics2));
+      AP01 = Utils::Multiply(*A01, false, *PP, false, AP01, GetOStream(Statistics2));
+      AP02 = Utils::Multiply(*A02, false, *PM, false, AP02, GetOStream(Statistics2));
+      AP10 = Utils::Multiply(*A10, false, *PV, false, AP10, GetOStream(Statistics2));
+      AP11 = Utils::Multiply(*A11, false, *PP, false, AP11, GetOStream(Statistics2));
+      AP12 = Utils::Multiply(*A12, false, *PM, false, AP12, GetOStream(Statistics2));
+      AP20 = Utils::Multiply(*A20, false, *PV, false, AP20, GetOStream(Statistics2));
+      AP21 = Utils::Multiply(*A21, false, *PP, false, AP21, GetOStream(Statistics2));
+      AP22 = Utils::Multiply(*A22, false, *PM, false, AP22, GetOStream(Statistics2));
     }
 
-
+    RCP<Matrix> Ac;
+    RCP<Matrix> Ac00;
+    RCP<Matrix> Ac01;
+    RCP<Matrix> Ac02;
+    RCP<Matrix> Ac10;
+    RCP<Matrix> Ac11;
+    RCP<Matrix> Ac12;
+    RCP<Matrix> Ac20;
+    RCP<Matrix> Ac21;
+    RCP<Matrix> Ac22;
+
+    if (implicitTranspose_) {
+      SubFactoryMonitor m2(*this, "MxM: P' x (AP) (implicit)", coarseLevel);
+
+      Ac   = Utils::Multiply(*P, true, *AP, false, Ac, GetOStream(Statistics2));
+      Ac00 = Utils::Multiply(*PV, true, *AP00, false, Ac00, GetOStream(Statistics2));
+      Ac01 = Utils::Multiply(*PV, true, *AP01, false, Ac01, GetOStream(Statistics2));
+      Ac02 = Utils::Multiply(*PV, true, *AP02, false, Ac02, GetOStream(Statistics2));
+      Ac10 = Utils::Multiply(*PP, true, *AP10, false, Ac10, GetOStream(Statistics2));
+      Ac11 = Utils::Multiply(*PP, true, *AP11, false, Ac11, GetOStream(Statistics2));
+      Ac12 = Utils::Multiply(*PP, true, *AP12, false, Ac12, GetOStream(Statistics2));
+      Ac20 = Utils::Multiply(*PM, true, *AP20, false, Ac20, GetOStream(Statistics2));
+      Ac21 = Utils::Multiply(*PM, true, *AP21, false, Ac21, GetOStream(Statistics2));
+      Ac22 = Utils::Multiply(*PM, true, *AP22, false, Ac22, GetOStream(Statistics2));
+
+    } else {
+      SubFactoryMonitor m2(*this, "MxM: R x (AP) (explicit)", coarseLevel);
+
+      RCP<Matrix> R  = Get<RCP<Matrix> >(coarseLevel, "R");
+      RCP<Matrix> RV = Get<RCP<Matrix> >(coarseLevel, "RV");
+      RCP<Matrix> RP = Get<RCP<Matrix> >(coarseLevel, "RP");
+      RCP<Matrix> RM = Get<RCP<Matrix> >(coarseLevel, "RM");
+
+      Ac   = Utils::Multiply(*R, false, *AP, false, Ac, GetOStream(Statistics2));
+      Ac00 = Utils::Multiply(*RV, false, *AP00, false, Ac00, GetOStream(Statistics2));
+      Ac01 = Utils::Multiply(*RV, false, *AP01, false, Ac01, GetOStream(Statistics2));
+      Ac02 = Utils::Multiply(*RV, false, *AP02, false, Ac02, GetOStream(Statistics2));
+      Ac10 = Utils::Multiply(*RP, false, *AP10, false, Ac10, GetOStream(Statistics2));
+      Ac11 = Utils::Multiply(*RP, false, *AP11, false, Ac11, GetOStream(Statistics2));
+      Ac12 = Utils::Multiply(*RP, false, *AP12, false, Ac12, GetOStream(Statistics2));
+      Ac20 = Utils::Multiply(*RM, false, *AP20, false, Ac20, GetOStream(Statistics2));
+      Ac21 = Utils::Multiply(*RM, false, *AP21, false, Ac21, GetOStream(Statistics2));
+      Ac22 = Utils::Multiply(*RM, false, *AP22, false, Ac22, GetOStream(Statistics2));
+    }
+    // FINISHED MAKING COARSE BLOCKS
+
+    Set(coarseLevel, "A", Ac);
+    Set(coarseLevel, "A00", Ac00);
+    Set(coarseLevel, "A01", Ac01);
+    Set(coarseLevel, "A02", Ac02);
+    Set(coarseLevel, "A10", Ac10);
+    Set(coarseLevel, "A11", Ac11);
+    Set(coarseLevel, "A12", Ac12);
+    Set(coarseLevel, "A20", Ac20);
+    Set(coarseLevel, "A21", Ac21);
+    Set(coarseLevel, "A22", Ac22);
   }
-
+}
 
 /*
   template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
@@ -234,45 +220,43 @@ namespace MueLu {
   }
 */
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintLoadBalancingInfo(const Matrix & Ac, const std::string & msgTag) {
-    std::stringstream ss(std::stringstream::out);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string MHDRAPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintLoadBalancingInfo(const Matrix &Ac, const std::string &msgTag) {
+  std::stringstream ss(std::stringstream::out);
 
-    // TODO: provide a option to skip this (to avoid global communication)
-      // TODO: skip if nproc == 1
-
-    //nonzero imbalance
-    size_t numMyNnz  = Ac.getLocalNumEntries();
-    GO maxNnz, minNnz;
-    RCP<const Teuchos::Comm<int> > comm = Ac.getRowMap()->getComm();
-    MueLu_maxAll(comm,(GO)numMyNnz,maxNnz);
-    //min nnz over all proc (disallow any processors with 0 nnz)
-    MueLu_minAll(comm, (GO)((numMyNnz > 0) ? numMyNnz : maxNnz), minNnz);
-    double imbalance = ((double) maxNnz) / minNnz;
-
-    size_t numMyRows = Ac.getLocalNumRows();
-    //Check whether Ac is spread over more than one process.
-    GO numActiveProcesses=0;
-    MueLu_sumAll(comm, (GO)((numMyRows > 0) ? 1 : 0), numActiveProcesses);
-
-    //min, max, and avg # rows per proc
-    GO minNumRows, maxNumRows;
-    double avgNumRows;
-    MueLu_maxAll(comm, (GO)numMyRows, maxNumRows);
-    MueLu_minAll(comm, (GO)((numMyRows > 0) ? numMyRows : maxNumRows), minNumRows);
-    assert(numActiveProcesses > 0);
-    avgNumRows = Ac.getGlobalNumRows() / numActiveProcesses;
-
-    ss << msgTag << " # processes with rows = " << numActiveProcesses << std::endl;
-    ss << msgTag << " min # rows per proc = " << minNumRows << ", max # rows per proc = " << maxNumRows << ", avg # rows per proc = " << avgNumRows << std::endl;
-    ss << msgTag << " nonzero imbalance = " << imbalance << std::endl;
+  // TODO: provide a option to skip this (to avoid global communication)
+  // TODO: skip if nproc == 1
 
-    return ss.str();
-  }
+  // nonzero imbalance
+  size_t numMyNnz = Ac.getLocalNumEntries();
+  GO maxNnz, minNnz;
+  RCP<const Teuchos::Comm<int> > comm = Ac.getRowMap()->getComm();
+  MueLu_maxAll(comm, (GO)numMyNnz, maxNnz);
+  // min nnz over all proc (disallow any processors with 0 nnz)
+  MueLu_minAll(comm, (GO)((numMyNnz > 0) ? numMyNnz : maxNnz), minNnz);
+  double imbalance = ((double)maxNnz) / minNnz;
 
+  size_t numMyRows = Ac.getLocalNumRows();
+  // Check whether Ac is spread over more than one process.
+  GO numActiveProcesses = 0;
+  MueLu_sumAll(comm, (GO)((numMyRows > 0) ? 1 : 0), numActiveProcesses);
 
-} //namespace MueLu
+  // min, max, and avg # rows per proc
+  GO minNumRows, maxNumRows;
+  double avgNumRows;
+  MueLu_maxAll(comm, (GO)numMyRows, maxNumRows);
+  MueLu_minAll(comm, (GO)((numMyRows > 0) ? numMyRows : maxNumRows), minNumRows);
+  assert(numActiveProcesses > 0);
+  avgNumRows = Ac.getGlobalNumRows() / numActiveProcesses;
 
-#define MUELU_MHDRAPFACTORY_SHORT
-#endif // MUELU_MHDRAPFACTORY_DEF_HPP
+  ss << msgTag << " # processes with rows = " << numActiveProcesses << std::endl;
+  ss << msgTag << " min # rows per proc = " << minNumRows << ", max # rows per proc = " << maxNumRows << ", avg # rows per proc = " << avgNumRows << std::endl;
+  ss << msgTag << " nonzero imbalance = " << imbalance << std::endl;
+
+  return ss.str();
+}
 
+}  // namespace MueLu
+
+#define MUELU_MHDRAPFACTORY_SHORT
+#endif  // MUELU_MHDRAPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_decl.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_decl.hpp
index c515a82b6e57..877dd5c2764b 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_decl.hpp
@@ -58,76 +58,76 @@
 
 namespace MueLu {
 
-  /*!
-    @class Q2Q1Q2CoarseGridFactory class.
-    @brief Factory for GMG Q2-Q1-Q2 interpolation
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Q2Q1Q2CoarseGridFactory : public TwoLevelFactoryBase {
+/*!
+  @class Q2Q1Q2CoarseGridFactory class.
+  @brief Factory for GMG Q2-Q1-Q2 interpolation
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Q2Q1Q2CoarseGridFactory : public TwoLevelFactoryBase {
 #undef MUELU_Q2Q1Q2COARSEGRIDFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    Q2Q1Q2CoarseGridFactory();
+  //! Constructor.
+  Q2Q1Q2CoarseGridFactory();
 
-    //! Destructor.
-    virtual ~Q2Q1Q2CoarseGridFactory();
+  //! Destructor.
+  virtual ~Q2Q1Q2CoarseGridFactory();
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the
-        generating factory is not specified in DeclareInput, then this
-        class will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the
+      generating factory is not specified in DeclareInput, then this
+      class will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    void BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const;
+  void BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const;
 
-    /* For our geometric multigrid needs, we will explicitly build the
-       coarse grid here and store it as level data. There are two
-       things we care about here: coordinates of the dofs and
-       element-dof lists.
+  /* For our geometric multigrid needs, we will explicitly build the
+     coarse grid here and store it as level data. There are two
+     things we care about here: coordinates of the dofs and
+     element-dof lists.
 
-       We will assume only that the elements are numbered
-       lexicographically, left-to-right, bottom-to-top. The order of
-       the degrees of freedom will only be assumed on coarse grids. It
-       can be anything on the finest grid (fineLevel.GetLevelID()==1).
+     We will assume only that the elements are numbered
+     lexicographically, left-to-right, bottom-to-top. The order of
+     the degrees of freedom will only be assumed on coarse grids. It
+     can be anything on the finest grid (fineLevel.GetLevelID()==1).
 
-       We further assume that the x- and y-components of velocity are
-       zippered together: UX1 UY1 UX2 UY2 ETC...
-    */
+     We further assume that the x- and y-components of velocity are
+     zippered together: UX1 UY1 UX2 UY2 ETC...
+  */
 
-    //@}
+  //@}
 
-  private:
-    // No parameters need to be stored... To save "P", use
-    //    coarseLevel.Set("P", finalP, this);
+ private:
+  // No parameters need to be stored... To save "P", use
+  //    coarseLevel.Set("P", finalP, this);
 
-  }; // class Q2Q1Q2CoarseGridFactory
+};  // class Q2Q1Q2CoarseGridFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_Q2Q1Q2COARSEGRIDFACTORY_SHORT
-#endif // MUELU_Q2Q1Q2COARSEGRIDFACTORY_DECL_HPP
+#endif  // MUELU_Q2Q1Q2COARSEGRIDFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_def.hpp b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_def.hpp
index 6ad181ae9cd9..775312db1b93 100644
--- a/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Geometric-Interpolation/MueLu_Q2Q1Q2CoarseGridFactory_def.hpp
@@ -64,178 +64,155 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Q2Q1Q2CoarseGridFactory(){
-    GetOStream(Runtime1) << "I constructed a Q2Q1Q2CoarseGridFactory object... Nothing else to do here." << std::endl;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~Q2Q1Q2CoarseGridFactory(){
-    // Should be empty. All destruction should be handled by Level-based get stuff and RCP
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-
-    Input(fineLevel, "VElementList");
-    Input(fineLevel, "PElementList");
-    Input(fineLevel, "MElementList");
-
-    Input(coarseLevel, "VElementList");
-    Input(coarseLevel, "PElementList");
-    Input(coarseLevel, "MElementList");
-
-
-    //currentLevel.DeclareInput(varName_,factory_,this);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-
-    GetOStream(Runtime1) << "Starting 'build' routine...\n";
-
-    // This will create a list of elements on the coarse grid with a
-    // predictable structure, as well as modify the fine grid list of
-    // elements, if necessary (i.e. if fineLevel.GetLevelID()==0);
-    //BuildCoarseGrid(fineLevel,coarseLevel);
-
-    // This will actually build our prolongator P
-    return BuildCoarseGrid(fineLevel,coarseLevel);
-
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const
-  {
-
-    GetOStream(Runtime1) << "starting 'BuildCoarseGrid' routine...\n";
-
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > fineElementPDOFs = Get< RCP<Teuchos::SerialDenseMatrix<GO,GO> > >(fineLevel,"PElementList");
-
-    GO totalFineElements = fineElementPDOFs->numRows();
-
-    // Compute number of coarse grid elements in total:
-    GO totalCoarseElements = totalFineElements/4;
-    LO nCoarseElements = (int) sqrt(totalCoarseElements);
-
-    // Initialize some counters:
-    size_t EdgeCount = (nCoarseElements + 1) * (nCoarseElements + 1);
-    size_t CenterCount = EdgeCount + 2 * nCoarseElements * (nCoarseElements + 1);
-
-
-    // Initialize arrays of the proper size:
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > coarseElementVDOFs = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalCoarseElements,18));
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > coarseElementPDOFs = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalCoarseElements,4));
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > coarseElementMDOFs = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalCoarseElements,9));
-
-
-    for ( GO coarseElement=0; coarseElement < totalCoarseElements; coarseElement++ )
-      {
-
-        // ***************************************************************
-        // This is less of a pain in the ass for magnetics, so I'm
-        // going to build the magnetics list. The velocity follows
-        // by doubling everything (and adding 1 for the y-components)
-        // and the pressure follows by copying the magnetics nodes.
-        // ***************************************************************
-
-        //if (coarseElement is on the Bottom Edge)
-        if (coarseElement < nCoarseElements)
-          {
-            //Bottom nodes
-            (*coarseElementMDOFs)(coarseElement,0) = coarseElement;
-            (*coarseElementMDOFs)(coarseElement,1) = coarseElement+1;
-
-            //Bottom edge
-            (*coarseElementMDOFs)(coarseElement,4) = EdgeCount++;
-
-          }
-        else
-          {
-            //Bottom Nodes
-            (*coarseElementMDOFs)(coarseElement,0) = (*coarseElementMDOFs)(coarseElement-nCoarseElements,3);
-            (*coarseElementMDOFs)(coarseElement,1) = (*coarseElementMDOFs)(coarseElement-nCoarseElements,2);
-
-            //Bottom Edge
-            (*coarseElementMDOFs)(coarseElement,4) = (*coarseElementMDOFs)(coarseElement-nCoarseElements,6);
-
-          }
-
-
-        //Right and Top Edges -- must be determined before left edge
-        (*coarseElementMDOFs)(coarseElement,5) = EdgeCount++;
-        (*coarseElementMDOFs)(coarseElement,6) = EdgeCount++;
-
-
-        //if (coarseElement is on the Left Edge)
-        if (coarseElement % nCoarseElements == 0)
-          {
-            //Top left node
-            (*coarseElementMDOFs)(coarseElement,3) = (*coarseElementMDOFs)(coarseElement,0)+nCoarseElements+1;
-
-            //Left Edge
-            (*coarseElementMDOFs)(coarseElement,7) = EdgeCount++;
-
-          }
-        else
-          {
-            //Top left node
-            (*coarseElementMDOFs)(coarseElement,3) = (*coarseElementMDOFs)(coarseElement-1,2);
-
-            //Left Edge
-            (*coarseElementMDOFs)(coarseElement,7) = (*coarseElementMDOFs)(coarseElement-1,5);
-          }
-
-        //Top right node -- Must be the last node to be determined!
-        (*coarseElementMDOFs)(coarseElement,2) = (*coarseElementMDOFs)(coarseElement,3)+1;
-
-        //Center Node
-        (*coarseElementMDOFs)(coarseElement,8) = CenterCount++;
-
-
-
-        // With Magnetics built, Pressure and Velocity follow without effort.
-        // First, Velocity:
-        (*coarseElementVDOFs)(coarseElement,0) = 2*(*coarseElementMDOFs)(coarseElement,0);
-        (*coarseElementVDOFs)(coarseElement,1) = 2*(*coarseElementMDOFs)(coarseElement,0)+1;
-        (*coarseElementVDOFs)(coarseElement,2) = 2*(*coarseElementMDOFs)(coarseElement,1);
-        (*coarseElementVDOFs)(coarseElement,3) = 2*(*coarseElementMDOFs)(coarseElement,1)+1;
-        (*coarseElementVDOFs)(coarseElement,4) = 2*(*coarseElementMDOFs)(coarseElement,2);
-        (*coarseElementVDOFs)(coarseElement,5) = 2*(*coarseElementMDOFs)(coarseElement,2)+1;
-        (*coarseElementVDOFs)(coarseElement,6) = 2*(*coarseElementMDOFs)(coarseElement,3);
-        (*coarseElementVDOFs)(coarseElement,7) = 2*(*coarseElementMDOFs)(coarseElement,3)+1;
-        (*coarseElementVDOFs)(coarseElement,8) = 2*(*coarseElementMDOFs)(coarseElement,4);
-        (*coarseElementVDOFs)(coarseElement,9) = 2*(*coarseElementMDOFs)(coarseElement,4)+1;
-        (*coarseElementVDOFs)(coarseElement,10) = 2*(*coarseElementMDOFs)(coarseElement,5);
-        (*coarseElementVDOFs)(coarseElement,11) = 2*(*coarseElementMDOFs)(coarseElement,5)+1;
-        (*coarseElementVDOFs)(coarseElement,12) = 2*(*coarseElementMDOFs)(coarseElement,6);
-        (*coarseElementVDOFs)(coarseElement,13) = 2*(*coarseElementMDOFs)(coarseElement,6)+1;
-        (*coarseElementVDOFs)(coarseElement,14) = 2*(*coarseElementMDOFs)(coarseElement,7);
-        (*coarseElementVDOFs)(coarseElement,15) = 2*(*coarseElementMDOFs)(coarseElement,7)+1;
-        (*coarseElementVDOFs)(coarseElement,16) = 2*(*coarseElementMDOFs)(coarseElement,8);
-        (*coarseElementVDOFs)(coarseElement,17) = 2*(*coarseElementMDOFs)(coarseElement,8)+1;
-
-        // Lastly, Pressure:
-        (*coarseElementPDOFs)(coarseElement,0) = (*coarseElementMDOFs)(coarseElement,0);
-        (*coarseElementPDOFs)(coarseElement,1) = (*coarseElementMDOFs)(coarseElement,1);
-        (*coarseElementPDOFs)(coarseElement,2) = (*coarseElementMDOFs)(coarseElement,2);
-        (*coarseElementPDOFs)(coarseElement,3) = (*coarseElementMDOFs)(coarseElement,3);
-
-      }// Loop over elements
-
-    Set(coarseLevel,"VElementList",coarseElementVDOFs);
-    Set(coarseLevel,"PElementList",coarseElementPDOFs);
-    Set(coarseLevel,"MElementList",coarseElementMDOFs);
-
-    //coarseLevel.Keep("VElementList",coarseLevel.GetFactoryManager()->GetFactory("VElementList").get());
-    //coarseLevel.Keep("PElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get());
-    //coarseLevel.Keep("MElementList",coarseLevel.GetFactoryManager()->GetFactory("MElementList").get());
-
-  }//BuildCoarseGrid
-
-
-} // namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Q2Q1Q2CoarseGridFactory() {
+  GetOStream(Runtime1) << "I constructed a Q2Q1Q2CoarseGridFactory object... Nothing else to do here." << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~Q2Q1Q2CoarseGridFactory() {
+  // Should be empty. All destruction should be handled by Level-based get stuff and RCP
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  Input(fineLevel, "VElementList");
+  Input(fineLevel, "PElementList");
+  Input(fineLevel, "MElementList");
+
+  Input(coarseLevel, "VElementList");
+  Input(coarseLevel, "PElementList");
+  Input(coarseLevel, "MElementList");
+
+  // currentLevel.DeclareInput(varName_,factory_,this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  GetOStream(Runtime1) << "Starting 'build' routine...\n";
+
+  // This will create a list of elements on the coarse grid with a
+  // predictable structure, as well as modify the fine grid list of
+  // elements, if necessary (i.e. if fineLevel.GetLevelID()==0);
+  // BuildCoarseGrid(fineLevel,coarseLevel);
+
+  // This will actually build our prolongator P
+  return BuildCoarseGrid(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Q2Q1Q2CoarseGridFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCoarseGrid(Level &fineLevel, Level &coarseLevel) const {
+  GetOStream(Runtime1) << "starting 'BuildCoarseGrid' routine...\n";
+
+  RCP<Teuchos::SerialDenseMatrix<GO, GO> > fineElementPDOFs = Get<RCP<Teuchos::SerialDenseMatrix<GO, GO> > >(fineLevel, "PElementList");
+
+  GO totalFineElements = fineElementPDOFs->numRows();
+
+  // Compute number of coarse grid elements in total:
+  GO totalCoarseElements = totalFineElements / 4;
+  LO nCoarseElements     = (int)sqrt(totalCoarseElements);
+
+  // Initialize some counters:
+  size_t EdgeCount   = (nCoarseElements + 1) * (nCoarseElements + 1);
+  size_t CenterCount = EdgeCount + 2 * nCoarseElements * (nCoarseElements + 1);
+
+  // Initialize arrays of the proper size:
+  RCP<Teuchos::SerialDenseMatrix<GO, GO> > coarseElementVDOFs = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalCoarseElements, 18));
+  RCP<Teuchos::SerialDenseMatrix<GO, GO> > coarseElementPDOFs = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalCoarseElements, 4));
+  RCP<Teuchos::SerialDenseMatrix<GO, GO> > coarseElementMDOFs = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalCoarseElements, 9));
+
+  for (GO coarseElement = 0; coarseElement < totalCoarseElements; coarseElement++) {
+    // ***************************************************************
+    // This is less of a pain in the ass for magnetics, so I'm
+    // going to build the magnetics list. The velocity follows
+    // by doubling everything (and adding 1 for the y-components)
+    // and the pressure follows by copying the magnetics nodes.
+    // ***************************************************************
+
+    // if (coarseElement is on the Bottom Edge)
+    if (coarseElement < nCoarseElements) {
+      // Bottom nodes
+      (*coarseElementMDOFs)(coarseElement, 0) = coarseElement;
+      (*coarseElementMDOFs)(coarseElement, 1) = coarseElement + 1;
+
+      // Bottom edge
+      (*coarseElementMDOFs)(coarseElement, 4) = EdgeCount++;
+
+    } else {
+      // Bottom Nodes
+      (*coarseElementMDOFs)(coarseElement, 0) = (*coarseElementMDOFs)(coarseElement - nCoarseElements, 3);
+      (*coarseElementMDOFs)(coarseElement, 1) = (*coarseElementMDOFs)(coarseElement - nCoarseElements, 2);
+
+      // Bottom Edge
+      (*coarseElementMDOFs)(coarseElement, 4) = (*coarseElementMDOFs)(coarseElement - nCoarseElements, 6);
+    }
+
+    // Right and Top Edges -- must be determined before left edge
+    (*coarseElementMDOFs)(coarseElement, 5) = EdgeCount++;
+    (*coarseElementMDOFs)(coarseElement, 6) = EdgeCount++;
+
+    // if (coarseElement is on the Left Edge)
+    if (coarseElement % nCoarseElements == 0) {
+      // Top left node
+      (*coarseElementMDOFs)(coarseElement, 3) = (*coarseElementMDOFs)(coarseElement, 0) + nCoarseElements + 1;
+
+      // Left Edge
+      (*coarseElementMDOFs)(coarseElement, 7) = EdgeCount++;
+
+    } else {
+      // Top left node
+      (*coarseElementMDOFs)(coarseElement, 3) = (*coarseElementMDOFs)(coarseElement - 1, 2);
+
+      // Left Edge
+      (*coarseElementMDOFs)(coarseElement, 7) = (*coarseElementMDOFs)(coarseElement - 1, 5);
+    }
+
+    // Top right node -- Must be the last node to be determined!
+    (*coarseElementMDOFs)(coarseElement, 2) = (*coarseElementMDOFs)(coarseElement, 3) + 1;
+
+    // Center Node
+    (*coarseElementMDOFs)(coarseElement, 8) = CenterCount++;
+
+    // With Magnetics built, Pressure and Velocity follow without effort.
+    // First, Velocity:
+    (*coarseElementVDOFs)(coarseElement, 0)  = 2 * (*coarseElementMDOFs)(coarseElement, 0);
+    (*coarseElementVDOFs)(coarseElement, 1)  = 2 * (*coarseElementMDOFs)(coarseElement, 0) + 1;
+    (*coarseElementVDOFs)(coarseElement, 2)  = 2 * (*coarseElementMDOFs)(coarseElement, 1);
+    (*coarseElementVDOFs)(coarseElement, 3)  = 2 * (*coarseElementMDOFs)(coarseElement, 1) + 1;
+    (*coarseElementVDOFs)(coarseElement, 4)  = 2 * (*coarseElementMDOFs)(coarseElement, 2);
+    (*coarseElementVDOFs)(coarseElement, 5)  = 2 * (*coarseElementMDOFs)(coarseElement, 2) + 1;
+    (*coarseElementVDOFs)(coarseElement, 6)  = 2 * (*coarseElementMDOFs)(coarseElement, 3);
+    (*coarseElementVDOFs)(coarseElement, 7)  = 2 * (*coarseElementMDOFs)(coarseElement, 3) + 1;
+    (*coarseElementVDOFs)(coarseElement, 8)  = 2 * (*coarseElementMDOFs)(coarseElement, 4);
+    (*coarseElementVDOFs)(coarseElement, 9)  = 2 * (*coarseElementMDOFs)(coarseElement, 4) + 1;
+    (*coarseElementVDOFs)(coarseElement, 10) = 2 * (*coarseElementMDOFs)(coarseElement, 5);
+    (*coarseElementVDOFs)(coarseElement, 11) = 2 * (*coarseElementMDOFs)(coarseElement, 5) + 1;
+    (*coarseElementVDOFs)(coarseElement, 12) = 2 * (*coarseElementMDOFs)(coarseElement, 6);
+    (*coarseElementVDOFs)(coarseElement, 13) = 2 * (*coarseElementMDOFs)(coarseElement, 6) + 1;
+    (*coarseElementVDOFs)(coarseElement, 14) = 2 * (*coarseElementMDOFs)(coarseElement, 7);
+    (*coarseElementVDOFs)(coarseElement, 15) = 2 * (*coarseElementMDOFs)(coarseElement, 7) + 1;
+    (*coarseElementVDOFs)(coarseElement, 16) = 2 * (*coarseElementMDOFs)(coarseElement, 8);
+    (*coarseElementVDOFs)(coarseElement, 17) = 2 * (*coarseElementMDOFs)(coarseElement, 8) + 1;
+
+    // Lastly, Pressure:
+    (*coarseElementPDOFs)(coarseElement, 0) = (*coarseElementMDOFs)(coarseElement, 0);
+    (*coarseElementPDOFs)(coarseElement, 1) = (*coarseElementMDOFs)(coarseElement, 1);
+    (*coarseElementPDOFs)(coarseElement, 2) = (*coarseElementMDOFs)(coarseElement, 2);
+    (*coarseElementPDOFs)(coarseElement, 3) = (*coarseElementMDOFs)(coarseElement, 3);
+
+  }  // Loop over elements
+
+  Set(coarseLevel, "VElementList", coarseElementVDOFs);
+  Set(coarseLevel, "PElementList", coarseElementPDOFs);
+  Set(coarseLevel, "MElementList", coarseElementMDOFs);
+
+  // coarseLevel.Keep("VElementList",coarseLevel.GetFactoryManager()->GetFactory("VElementList").get());
+  // coarseLevel.Keep("PElementList",coarseLevel.GetFactoryManager()->GetFactory("PElementList").get());
+  // coarseLevel.Keep("MElementList",coarseLevel.GetFactoryManager()->GetFactory("MElementList").get());
+
+}  // BuildCoarseGrid
+
+}  // namespace MueLu
 
 #define MUELU_Q2Q1Q2COARSEGRIDFACTORY_SHORT
-#endif // MUELU_Q2Q1Q2COARSEGRIDFACTORY_DEF_HPP
+#endif  // MUELU_Q2Q1Q2COARSEGRIDFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_decl.hpp b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_decl.hpp
index e88e3b388c3d..657f0bb2b3df 100644
--- a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_decl.hpp
@@ -61,92 +61,92 @@
 
 namespace MueLu {
 
-  /*!
-    @class MatrixFreeTentativePFactory class.
-    @brief Factory for building the matrix-free tentative restrictor.
-
-    Factory for creating the matrix-free tentative restrictor.   Nullspace vectors are split across aggregates so that they
-    have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
-    tentative prolongator, and the R becomes the coarse nullspace.
-
-    Note that the MatrixFreeTentativePFactory also creates the coarse null space vectors, that is, it serves as generating
-    factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
-    level one has to use the NullspaceFactory.
-
-    @ingroup MueLuTransferClasses
-
-    ## Input/output of MatrixFreeTentativePFactory ##
-
-    ### User parameters of MatrixFreeTentativePFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the MatrixFreeTentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the MatrixFreeTentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
-     UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
-     CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see MatrixFreeTentativePFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see MatrixFreeTentativePFactory::DeclareInput).
-
-    ### Variables provided by MatrixFreeTentativePFactory ###
-
-    After MatrixFreeTentativePFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | R       | MatrixFreeTentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
-    | Nullspace | MatrixFreeTentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MatrixFreeTentativePFactory;
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
-  public:
-    typedef LocalOrdinal                                             local_ordinal_type;
-    typedef GlobalOrdinal                                            global_ordinal_type;
-    typedef typename DeviceType::execution_space                     execution_space;
-    typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
-    typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>      node_type;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType   real_type;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class MatrixFreeTentativePFactory class.
+  @brief Factory for building the matrix-free tentative restrictor.
+
+  Factory for creating the matrix-free tentative restrictor.   Nullspace vectors are split across aggregates so that they
+  have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
+  tentative prolongator, and the R becomes the coarse nullspace.
+
+  Note that the MatrixFreeTentativePFactory also creates the coarse null space vectors, that is, it serves as generating
+  factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
+  level one has to use the NullspaceFactory.
+
+  @ingroup MueLuTransferClasses
+
+  ## Input/output of MatrixFreeTentativePFactory ##
+
+  ### User parameters of MatrixFreeTentativePFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the MatrixFreeTentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the MatrixFreeTentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+   UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
+   CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see MatrixFreeTentativePFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see MatrixFreeTentativePFactory::DeclareInput).
+
+  ### Variables provided by MatrixFreeTentativePFactory ###
+
+  After MatrixFreeTentativePFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | R       | MatrixFreeTentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
+  | Nullspace | MatrixFreeTentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MatrixFreeTentativePFactory;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
+ public:
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename DeviceType::execution_space execution_space;
+  typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
+  typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType real_type;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    MatrixFreeTentativePFactory() { }
+  //! Constructor
+  MatrixFreeTentativePFactory() {}
 
-    //! Destructor.
-    virtual ~MatrixFreeTentativePFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~MatrixFreeTentativePFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
-  };
+  //@}
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
-#endif // MUELU_MATRIXFREETENTATIVEPFACTORY_DECL_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_def.hpp b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_def.hpp
index f828bf6363b3..b143f5ae0d8c 100644
--- a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativePFactory_def.hpp
@@ -57,74 +57,73 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  RCP<const ParameterList> MatrixFreeTentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Aggregates",         Teuchos::null, "Generating factory of the aggregates");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",          Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("Scaled Nullspace",   Teuchos::null, "Generating factory of the scaled nullspace");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap",          Teuchos::null, "Generating factory of the coarse map");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory of the coordinates");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void MatrixFreeTentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-
-    const ParameterList& pL = GetParameterList();
-    // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-    Input(fineLevel, "Aggregates");
-    Input(fineLevel, nspName);
-    Input(fineLevel, "UnAmalgamationInfo");
-    Input(fineLevel, "CoarseMap");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void MatrixFreeTentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void MatrixFreeTentativePFactory<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    const ParameterList& pL = GetParameterList();
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-    auto aggregates    = Get< RCP<Aggregates> > (fineLevel, "Aggregates");
-    auto amalgInfo     = Get< RCP<AmalgamationInfo> >  (fineLevel, "UnAmalgamationInfo");
-    auto fineNullspace = Get< RCP<MultiVector> >       (fineLevel, nspName);
-    auto coarseMap     = Get< RCP<const Map> >         (fineLevel, "CoarseMap");
-    Teuchos::RCP<const Map> fineMap = fineNullspace->getMap();
-
-    // Matrix-free should never run with aggregates that cross processors
-    if (aggregates->AggregatesCrossProcessors())
-      TEUCHOS_TEST_FOR_EXCEPTION(true,Exceptions::RuntimeError,"MatrixFreeTentativePFactory does not support aggregates that cross processors!");
-
-    size_t NSDim = fineNullspace->getNumVectors();
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
-
-    Teuchos::RCP<Operator> P = Teuchos::rcp(new MatrixFreeTentativeP<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>(coarseMap, fineMap, aggregates));
-    P->apply(*fineNullspace,*coarseNullspace,Teuchos::TRANS,1.0,0.0); // coarse = alpha*R*fine + beta*coarse
-
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-    Set(coarseLevel, "P",         P);
-  }
-
-} //namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+RCP<const ParameterList> MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase>>("Aggregates", Teuchos::null, "Generating factory of the aggregates");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase>>("Scaled Nullspace", Teuchos::null, "Generating factory of the scaled nullspace");
+  validParamList->set<RCP<const FactoryBase>>("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase>>("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+  validParamList->set<RCP<const FactoryBase>>("Coordinates", Teuchos::null, "Generating factory of the coordinates");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  const ParameterList& pL = GetParameterList();
+  // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
+  std::string nspName = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  Input(fineLevel, "Aggregates");
+  Input(fineLevel, nspName);
+  Input(fineLevel, "UnAmalgamationInfo");
+  Input(fineLevel, "CoarseMap");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void MatrixFreeTentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  const ParameterList& pL = GetParameterList();
+  std::string nspName     = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  auto aggregates                 = Get<RCP<Aggregates>>(fineLevel, "Aggregates");
+  auto amalgInfo                  = Get<RCP<AmalgamationInfo>>(fineLevel, "UnAmalgamationInfo");
+  auto fineNullspace              = Get<RCP<MultiVector>>(fineLevel, nspName);
+  auto coarseMap                  = Get<RCP<const Map>>(fineLevel, "CoarseMap");
+  Teuchos::RCP<const Map> fineMap = fineNullspace->getMap();
+
+  // Matrix-free should never run with aggregates that cross processors
+  if (aggregates->AggregatesCrossProcessors())
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MatrixFreeTentativePFactory does not support aggregates that cross processors!");
+
+  size_t NSDim                     = fineNullspace->getNumVectors();
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
+
+  Teuchos::RCP<Operator> P = Teuchos::rcp(new MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>(coarseMap, fineMap, aggregates));
+  P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, 1.0, 0.0);  // coarse = alpha*R*fine + beta*coarse
+
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "P", P);
+}
+
+}  // namespace MueLu
 
 #define MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
-#endif // MUELU_MATRIXFREETENTATIVEPFACTORY_DEF_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_decl.hpp b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_decl.hpp
index a0183be34cc3..b864ec88d810 100644
--- a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_decl.hpp
+++ b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_decl.hpp
@@ -62,79 +62,77 @@
 
 namespace MueLu {
 
-  /*!
-    @class MatrixFreeTentativeP class.
-    @brief Matrix-free tentative restrictor operator.
-  */
-  // template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  // class MatrixFreeTentativeP;
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> : public Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
-  public:
-    typedef LocalOrdinal                                             local_ordinal_type;
-    typedef GlobalOrdinal                                            global_ordinal_type;
-    typedef typename DeviceType::execution_space                     execution_space;
-    typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
-    typedef Kokkos::MDRangePolicy<local_ordinal_type, execution_space, Kokkos::Rank<2>> md_range_type;
-    typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>      node_type;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType   real_type;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class MatrixFreeTentativeP class.
+  @brief Matrix-free tentative restrictor operator.
+*/
+// template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// class MatrixFreeTentativeP;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> : public Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> {
+ public:
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename DeviceType::execution_space execution_space;
+  typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
+  typedef Kokkos::MDRangePolicy<local_ordinal_type, execution_space, Kokkos::Rank<2>> md_range_type;
+  typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType real_type;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_MATRIXFREETENTATIVEP_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    MatrixFreeTentativeP(Teuchos::RCP<const Map> coarse_map, Teuchos::RCP<const Map> fine_map, Teuchos::RCP<const Aggregates> aggregates)
-    : fine_map_(fine_map),
-      coarse_map_(coarse_map),
-      aggregates_(aggregates)
-    { }
+  //! Constructor
+  MatrixFreeTentativeP(Teuchos::RCP<const Map> coarse_map, Teuchos::RCP<const Map> fine_map, Teuchos::RCP<const Aggregates> aggregates)
+    : fine_map_(fine_map)
+    , coarse_map_(coarse_map)
+    , aggregates_(aggregates) {}
 
-    //! Destructor.
-    ~MatrixFreeTentativeP() = default;
-    //@}
+  //! Destructor.
+  ~MatrixFreeTentativeP() = default;
+  //@}
 
-    // compute the apply operator, Y = alpha*R*X + beta*Y
-    void apply(const MultiVector &X, MultiVector &Y, Teuchos::ETransp mode=Teuchos::NO_TRANS, Scalar alpha=Teuchos::ScalarTraits<Scalar>::one(), Scalar beta=Teuchos::ScalarTraits<Scalar>::zero()) const override;
+  // compute the apply operator, Y = alpha*R*X + beta*Y
+  void apply(const MultiVector &X, MultiVector &Y, Teuchos::ETransp mode = Teuchos::NO_TRANS, Scalar alpha = Teuchos::ScalarTraits<Scalar>::one(), Scalar beta = Teuchos::ScalarTraits<Scalar>::zero()) const override;
 
-    // compute the residual
-    void residual (const MultiVector &X, const MultiVector &B, MultiVector &R) const override;
+  // compute the residual
+  void residual(const MultiVector &X, const MultiVector &B, MultiVector &R) const override;
 
-    // get the range map
-    Teuchos::RCP<const Map> getRangeMap() const override {
-      return fine_map_;
-    }
+  // get the range map
+  Teuchos::RCP<const Map> getRangeMap() const override {
+    return fine_map_;
+  }
 
-    // get the domain map
-    Teuchos::RCP<const Map> getDomainMap() const override {
-      return coarse_map_;
-    }
+  // get the domain map
+  Teuchos::RCP<const Map> getDomainMap() const override {
+    return coarse_map_;
+  }
 
-    // get the aggregates
-    Teuchos::RCP<const Aggregates> getAggregates() const {
-      return aggregates_;
-    }
+  // get the aggregates
+  Teuchos::RCP<const Aggregates> getAggregates() const {
+    return aggregates_;
+  }
 
-  private:
+ private:
+  // the fine map
+  const Teuchos::RCP<const Map> fine_map_;
 
-    // the fine map
-    const Teuchos::RCP<const Map> fine_map_;
+  // the coarse map
+  const Teuchos::RCP<const Map> coarse_map_;
 
-    // the coarse map
-    const Teuchos::RCP<const Map> coarse_map_;
+  // the aggregates required for the grid transfer
+  const Teuchos::RCP<const Aggregates> aggregates_;
+};
 
-    // the aggregates required for the grid transfer
-    const Teuchos::RCP<const Aggregates> aggregates_;
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MATRIXFREETENTATIVEP_SHORT
-#endif // MUELU_MATRIXFREETENTATIVEP_DECL_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEP_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_def.hpp b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_def.hpp
index b5365e80545a..95239e695da3 100644
--- a/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_def.hpp
+++ b/packages/muelu/src/Transfers/Matrix-Free/MueLu_MatrixFreeTentativeP_def.hpp
@@ -61,63 +61,64 @@
 
 namespace MueLu {
 
-  // compute Y = alpha*R*X + beta*Y
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void MatrixFreeTentativeP<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::apply(const MultiVector &X,
-                                          MultiVector &Y,
-                                          Teuchos::ETransp mode,
-                                          Scalar alpha,
-                                          Scalar beta) const {
-
-    using impl_scalar_type = typename Kokkos::ArithTraits<Scalar>::val_type;
-    impl_scalar_type implAlpha = alpha;
-
-    // Step 1: Y*=beta*Y, setup necessary structures
-    Y.scale(beta);
-
-    // TODO: probably smarter to sqrt the whole aggSizes once, but may be slower if it's done in a separate kernel launch?
-    typename Aggregates::aggregates_sizes_type::const_type aggSizes = aggregates_->ComputeAggregateSizes();
-
-    auto kokkos_view_X = X.getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto kokkos_view_Y = Y.getDeviceLocalView(Xpetra::Access::ReadWrite);
-    LO numCols = kokkos_view_X.extent(1);
-
-    if(mode == Teuchos::TRANS) { // if we're in restrictor mode
-      auto vertex2AggId = aggregates_->GetVertex2AggId();
-      auto vertex2AggIdView = vertex2AggId->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      LO numNodes = kokkos_view_X.extent(0);
-
-      // Step 2: Compute Y=Y+alpha*R*X
-      // recall R*X is an average of X over aggregates
-      Kokkos::parallel_for("MueLu:MatrixFreeTentativeR_kokkos:apply", md_range_type({0,0},{numCols,numNodes}),
+// compute Y = alpha*R*X + beta*Y
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::apply(const MultiVector &X,
+                                                                                                                                 MultiVector &Y,
+                                                                                                                                 Teuchos::ETransp mode,
+                                                                                                                                 Scalar alpha,
+                                                                                                                                 Scalar beta) const {
+  using impl_scalar_type     = typename Kokkos::ArithTraits<Scalar>::val_type;
+  impl_scalar_type implAlpha = alpha;
+
+  // Step 1: Y*=beta*Y, setup necessary structures
+  Y.scale(beta);
+
+  // TODO: probably smarter to sqrt the whole aggSizes once, but may be slower if it's done in a separate kernel launch?
+  typename Aggregates::aggregates_sizes_type::const_type aggSizes = aggregates_->ComputeAggregateSizes();
+
+  auto kokkos_view_X = X.getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto kokkos_view_Y = Y.getDeviceLocalView(Xpetra::Access::ReadWrite);
+  LO numCols         = kokkos_view_X.extent(1);
+
+  if (mode == Teuchos::TRANS) {  // if we're in restrictor mode
+    auto vertex2AggId     = aggregates_->GetVertex2AggId();
+    auto vertex2AggIdView = vertex2AggId->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    LO numNodes           = kokkos_view_X.extent(0);
+
+    // Step 2: Compute Y=Y+alpha*R*X
+    // recall R*X is an average of X over aggregates
+    Kokkos::parallel_for(
+        "MueLu:MatrixFreeTentativeR_kokkos:apply", md_range_type({0, 0}, {numCols, numNodes}),
         KOKKOS_LAMBDA(const int colIdx, const int NodeIdx) {
-          LO aggIdx = vertex2AggIdView(NodeIdx,0);
-          if(aggIdx != -1) { // depending on maps, vertex2agg
-            Kokkos::atomic_add(&kokkos_view_Y(aggIdx,colIdx),implAlpha*kokkos_view_X(NodeIdx,colIdx)/Kokkos::sqrt(aggSizes(aggIdx)));
+          LO aggIdx = vertex2AggIdView(NodeIdx, 0);
+          if (aggIdx != -1) {  // depending on maps, vertex2agg
+            Kokkos::atomic_add(&kokkos_view_Y(aggIdx, colIdx), implAlpha * kokkos_view_X(NodeIdx, colIdx) / Kokkos::sqrt(aggSizes(aggIdx)));
           }
         });
-    } else { // if we're in prolongator mode
-      const auto vertex2Agg = aggregates_->GetVertex2AggId();
-      auto vertex2AggView = vertex2Agg->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      LO numNodes = kokkos_view_Y.extent(0);
-
-      // Step 2: Compute Y=Y+alpha*P*X
-      // recall P*X is essentially injection of X, but sum if a node belongs to multiple aggregates
-      Kokkos::parallel_for("MueLu:MatrixFreeTentativeP:apply", md_range_type({0,0},{numCols,numNodes}),
+  } else {  // if we're in prolongator mode
+    const auto vertex2Agg = aggregates_->GetVertex2AggId();
+    auto vertex2AggView   = vertex2Agg->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    LO numNodes           = kokkos_view_Y.extent(0);
+
+    // Step 2: Compute Y=Y+alpha*P*X
+    // recall P*X is essentially injection of X, but sum if a node belongs to multiple aggregates
+    Kokkos::parallel_for(
+        "MueLu:MatrixFreeTentativeP:apply", md_range_type({0, 0}, {numCols, numNodes}),
         KOKKOS_LAMBDA(const int colIdx, const int fineIdx) {
-          LO aggIdx = vertex2AggView(fineIdx,0);
-          kokkos_view_Y(fineIdx,colIdx) += implAlpha*kokkos_view_X(aggIdx,colIdx)/Kokkos::sqrt(aggSizes(aggIdx));
+          LO aggIdx = vertex2AggView(fineIdx, 0);
+          kokkos_view_Y(fineIdx, colIdx) += implAlpha * kokkos_view_X(aggIdx, colIdx) / Kokkos::sqrt(aggSizes(aggIdx));
         });
-    }
   }
+}
 
-  // I don't care
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void MatrixFreeTentativeP<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::residual(const MultiVector &X, const MultiVector &B, MultiVector &R) const {
-    TEUCHOS_TEST_FOR_EXCEPTION(true,Exceptions::RuntimeError,"MatrixFreeTentativeP residual would make no sense as the operator is not square!");
-  }
+// I don't care
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void MatrixFreeTentativeP<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::residual(const MultiVector &X, const MultiVector &B, MultiVector &R) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MatrixFreeTentativeP residual would make no sense as the operator is not square!");
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MATRIXFREETENTATIVEP_SHORT
-#endif // MUELU_MATRIXFREETENTATIVEP_DEF_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEP_DEF_HPP
diff --git a/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_decl.hpp b/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_decl.hpp
index 9b076caa0e07..5a6dd00bd158 100644
--- a/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_decl.hpp
@@ -66,182 +66,178 @@
 
 namespace MueLu {
 
-  /*!
-    @class IntrepidPCoarsenFactory class.
-    @ingroup MueLuTransferClasses
-    @brief Factory for building transfer operators based on coarsening in polynomial degree, following the Intrepid basis functions
+/*!
+  @class IntrepidPCoarsenFactory class.
+  @ingroup MueLuTransferClasses
+  @brief Factory for building transfer operators based on coarsening in polynomial degree, following the Intrepid basis functions
 
-    ## Input/output of IntrepidPCoarsenFactory ##
+  ## Input/output of IntrepidPCoarsenFactory ##
 
-    ### User parameters of IntrepidPCoarsenFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | pcoarsen: schedule                          | string  | ""    | * | * |   | String describing the higher order coarsening scheme to use |
-    | pcoarsen: element                          | string  | ""    | * | * |   | String describing the class of element to use for higher order coarsening |
-    | pcoarsen: hi basis                          | string  | ""    | * | * |   | String describing higher-order basis function used |
-    | pcoarsen: lo basis                          | string  | ""    | * | * |   | String describing lower-order basis function to be used for coarse grid |
-    | pcoarsen: element to node map               | RCP<Intrepid::FieldContainer<LocalOrdinal> > | null | * | * |   | A FieldContainer with the element-to-node map in local ids compatible with the matrix column map |
-    | A                                      | Factory | null      | * | * |   | Generating factory of the matrix A used during the prolongator smoothing process |
+  ### User parameters of IntrepidPCoarsenFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | pcoarsen: schedule                          | string  | ""    | * | * |   | String describing the higher order coarsening scheme to use |
+  | pcoarsen: element                          | string  | ""    | * | * |   | String describing the class of element to use for higher order coarsening |
+  | pcoarsen: hi basis                          | string  | ""    | * | * |   | String describing higher-order basis function used |
+  | pcoarsen: lo basis                          | string  | ""    | * | * |   | String describing lower-order basis function to be used for coarse grid |
+  | pcoarsen: element to node map               | RCP<Intrepid::FieldContainer<LocalOrdinal> > | null | * | * |   | A FieldContainer with the element-to-node map in local ids compatible with the matrix column map |
+  | A                                      | Factory | null      | * | * |   | Generating factory of the matrix A used during the prolongator smoothing process |
 
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see IntrepidPCoarsen::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see IntrepidPCoarsen::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see IntrepidPCoarsen::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see IntrepidPCoarsen::DeclareInput).
 
-    ### Variables provided by IntrepidPCoarsen ###
+  ### Variables provided by IntrepidPCoarsen ###
 
-    After IntrepidPCoarsen::Build the following data is available (if requested)
+  After IntrepidPCoarsen::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | IntrepidPCoarsenFactory   | Smoothed prolongator
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | IntrepidPCoarsenFactory   | Smoothed prolongator
 
 
-    NOTE:  The prolongator generated by this routine strips Dirichlet unknowns out of the coarse grid.  This means that the prolongator cannot
-    transfer *solutions* between grids, only updates (because the update at a Dirichlet node is zero).  This is a subtle distinction, but an
-    important one for testing purposes.
+  NOTE:  The prolongator generated by this routine strips Dirichlet unknowns out of the coarse grid.  This means that the prolongator cannot
+  transfer *solutions* between grids, only updates (because the update at a Dirichlet node is zero).  This is a subtle distinction, but an
+  important one for testing purposes.
 
-    FIXME: There may be implications for the correctness of the nullspace transfer from fine to coarse that I need to understand.
+  FIXME: There may be implications for the correctness of the nullspace transfer from fine to coarse that I need to understand.
 
 
-  */
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class IntrepidPCoarsenFactory : public PFactory {
+          class Node          = DefaultNode>
+class IntrepidPCoarsenFactory : public PFactory {
 #undef MUELU_INTREPIDPCOARSENFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> LOFieldContainer;
-    typedef Kokkos::DynRankView<double,typename Node::device_type> SCFieldContainer;
-    typedef Intrepid2::Basis<typename Node::device_type::execution_space,double,double> Basis; // Hardwired on purpose
+ public:
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> LOFieldContainer;
+  typedef Kokkos::DynRankView<double, typename Node::device_type> SCFieldContainer;
+  typedef Intrepid2::Basis<typename Node::device_type::execution_space, double, double> Basis;  // Hardwired on purpose
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
-      User can supply a factory for generating the tentative prolongator.
-    */
-    IntrepidPCoarsenFactory() { }
+  /*! @brief Constructor.
+    User can supply a factory for generating the tentative prolongator.
+  */
+  IntrepidPCoarsenFactory() {}
+
+  //! Destructor.
+  virtual ~IntrepidPCoarsenFactory() {}
+
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Destructor.
-    virtual ~IntrepidPCoarsenFactory() { }
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Input
+  //@{
 
-    //@}
-
-    //! Input
-    //@{
-
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
-
-    //@}
-
-    //! @name Build methods.
-    //@{
-
-    /*!
-      @brief Build method.
-
-      Builds IPC prolongator and returns it in <tt>coarseLevel</tt>.
-      */
-    void Build(Level& fineLevel, Level &coarseLevel) const;
-
-    void BuildP(Level &fineLevel, Level &coarseLevel) const; //Build()
-
-    //@}
-  private:
-    //! @name Internal Utilities
-    //@{
-    // NOTE: This is hardwired to double on purpose.
-    void GenerateLinearCoarsening_pn_kirby_to_p1(const LOFieldContainer & hi_elemToNode,
-                                                 const std::vector<bool> & hi_nodeIsOwned,
-                                                 const SCFieldContainer & hi_DofCoords,
-                                                 const std::vector<size_t> &lo_node_in_hi,
-                                                 const Basis &lo_Basis,
-                                                 const std::vector<LocalOrdinal> & hi_to_lo_map,
-                                                 const Teuchos::RCP<const Map> & lo_colMap,
-                                                 const Teuchos::RCP<const Map> & lo_domainMap,
-                                                 const Teuchos::RCP<const Map> & hi_map,
-                                                 Teuchos::RCP<Matrix>& P) const;
-    //@}
-
-    // NOTE: This is hardwired to double on purpose.
-    void GenerateLinearCoarsening_pn_kirby_to_pm(const LOFieldContainer & hi_elemToNode,
-                                                 const std::vector<bool> & hi_nodeIsOwned,
-                                                 const SCFieldContainer & hi_DofCoords,
-                                                 const LOFieldContainer & lo_elemToHiRepresentativeNode,
-                                                 const Basis &lo_basis,
-                                                 const std::vector<LocalOrdinal> & hi_to_lo_map,
-                                                 const Teuchos::RCP<const Map> & lo_colMap,
-                                                 const Teuchos::RCP<const Map> & lo_domainMap,
-                                                 const Teuchos::RCP<const Map> & hi_map,
-                                                 Teuchos::RCP<Matrix>& P) const;
-
-
-  }; //class IntrepidPCoarsenFactory
-
-
-  /* Utility functions for use with Intrepid */
-  namespace MueLuIntrepid {
-
-    template<class Scalar,class KokkosExecutionSpace>
-    Teuchos::RCP<Intrepid2::Basis<KokkosExecutionSpace,Scalar,Scalar> >  BasisFactory(const std::string & name, int & degree);
-
-    template<class Scalar,class KokkosDeviceType>
-    void IntrepidGetLoNodeInHi(const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space,Scalar,Scalar> > &hi_basis,
-                               const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space,Scalar,Scalar> > &lo_basis,
-                               std::vector<size_t> & lo_node_in_hi,
-                               Kokkos::DynRankView<Scalar,KokkosDeviceType> & hi_DofCoords);
-
-    template<class LocalOrdinal, class GlobalOrdinal, class Node, class LOFieldContainer>
-    void GenerateLoNodeInHiViaGIDs(const std::vector<std::vector<size_t> > & candidates,const LOFieldContainer & hi_elemToNode,
-                                   RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & hi_columnMap,
-                                   LOFieldContainer & lo_elemToHiRepresentativeNode);
-
-    template <class LocalOrdinal, class LOFieldContainer>
-    void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
-                           const std::vector<bool> & hi_nodeIsOwned,
-                           const std::vector<size_t> & lo_node_in_hi,
-                           const Teuchos::ArrayRCP<const int> & hi_isDirichlet,
-                           LOFieldContainer & lo_elemToNode,
-                           std::vector<bool> & lo_nodeIsOwned,
-                           std::vector<LocalOrdinal> & hi_to_lo_map,
-                           int & lo_numOwnedNodes);
-
-    template <class LocalOrdinal, class LOFieldContainer>
-    void BuildLoElemToNodeViaRepresentatives(const LOFieldContainer & hi_elemToNode,
-                                             const std::vector<bool> & hi_nodeIsOwned,
-                                             const LOFieldContainer & lo_elemToHiRepresentativeNode,
-                                             LOFieldContainer & lo_elemToNode,
-                                             std::vector<bool> & lo_nodeIsOwned,
-                                             std::vector<LocalOrdinal> & hi_to_lo_map,
-                                             int & lo_numOwnedNodes);
-
-
-    template <class LocalOrdinal, class GlobalOrdinal, class Node>
-    void GenerateColMapFromImport(const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> & hi_importer,const std::vector<LocalOrdinal> &hi_to_lo_map,const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> & lo_domainMap, const size_t & lo_columnMapLength, RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & lo_columnMap);
-
-
-    template<class Basis, class SCFieldContainer>
-    void GenerateRepresentativeBasisNodes(const Basis & basis, const SCFieldContainer & ReferenceNodeLocations, const double threshold, std::vector<std::vector<size_t> > & representative_node_candidates);
-
-    // ! Given an element to (global) node map and a basis, determine one global ordinal per geometric entity (vertex, edge, face,
-    // ! interior).  On exit, seeds container is of dimension (spaceDim+1), and contains a sorted vector of local ordinals
-    // ! belonging to entities of that dimension.  Only locally-owned degrees of freedom (as determined by rowMap and columnMap)
-    // ! will be stored in seeds.
-    template<class Basis, class LOFieldContainer, class LocalOrdinal, class GlobalOrdinal, class Node>
-    void FindGeometricSeedOrdinals(Teuchos::RCP<Basis> basis, const LOFieldContainer &elementToNodeMap,
-                                   std::vector<std::vector<LocalOrdinal> > &seeds,
-                                   const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> &rowMap,
-                                   const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> &columnMap);
-
-  }//namespace MueLuIntrepid
-} //namespace MueLu
-
-#define  MUELU_INTREPIDPCOARSENFACTORY_SHORT
-#endif // MUELU_INTREPIDPCOARSENFACTORY_DECL_HPP
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+
+  //@}
+
+  //! @name Build methods.
+  //@{
+
+  /*!
+    @brief Build method.
+
+    Builds IPC prolongator and returns it in <tt>coarseLevel</tt>.
+    */
+  void Build(Level &fineLevel, Level &coarseLevel) const;
+
+  void BuildP(Level &fineLevel, Level &coarseLevel) const;  // Build()
+
+  //@}
+ private:
+  //! @name Internal Utilities
+  //@{
+  // NOTE: This is hardwired to double on purpose.
+  void GenerateLinearCoarsening_pn_kirby_to_p1(const LOFieldContainer &hi_elemToNode,
+                                               const std::vector<bool> &hi_nodeIsOwned,
+                                               const SCFieldContainer &hi_DofCoords,
+                                               const std::vector<size_t> &lo_node_in_hi,
+                                               const Basis &lo_Basis,
+                                               const std::vector<LocalOrdinal> &hi_to_lo_map,
+                                               const Teuchos::RCP<const Map> &lo_colMap,
+                                               const Teuchos::RCP<const Map> &lo_domainMap,
+                                               const Teuchos::RCP<const Map> &hi_map,
+                                               Teuchos::RCP<Matrix> &P) const;
+  //@}
+
+  // NOTE: This is hardwired to double on purpose.
+  void GenerateLinearCoarsening_pn_kirby_to_pm(const LOFieldContainer &hi_elemToNode,
+                                               const std::vector<bool> &hi_nodeIsOwned,
+                                               const SCFieldContainer &hi_DofCoords,
+                                               const LOFieldContainer &lo_elemToHiRepresentativeNode,
+                                               const Basis &lo_basis,
+                                               const std::vector<LocalOrdinal> &hi_to_lo_map,
+                                               const Teuchos::RCP<const Map> &lo_colMap,
+                                               const Teuchos::RCP<const Map> &lo_domainMap,
+                                               const Teuchos::RCP<const Map> &hi_map,
+                                               Teuchos::RCP<Matrix> &P) const;
+
+};  // class IntrepidPCoarsenFactory
+
+/* Utility functions for use with Intrepid */
+namespace MueLuIntrepid {
+
+template <class Scalar, class KokkosExecutionSpace>
+Teuchos::RCP<Intrepid2::Basis<KokkosExecutionSpace, Scalar, Scalar> > BasisFactory(const std::string &name, int &degree);
+
+template <class Scalar, class KokkosDeviceType>
+void IntrepidGetLoNodeInHi(const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space, Scalar, Scalar> > &hi_basis,
+                           const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space, Scalar, Scalar> > &lo_basis,
+                           std::vector<size_t> &lo_node_in_hi,
+                           Kokkos::DynRankView<Scalar, KokkosDeviceType> &hi_DofCoords);
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class LOFieldContainer>
+void GenerateLoNodeInHiViaGIDs(const std::vector<std::vector<size_t> > &candidates, const LOFieldContainer &hi_elemToNode,
+                               RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &hi_columnMap,
+                               LOFieldContainer &lo_elemToHiRepresentativeNode);
+
+template <class LocalOrdinal, class LOFieldContainer>
+void BuildLoElemToNode(const LOFieldContainer &hi_elemToNode,
+                       const std::vector<bool> &hi_nodeIsOwned,
+                       const std::vector<size_t> &lo_node_in_hi,
+                       const Teuchos::ArrayRCP<const int> &hi_isDirichlet,
+                       LOFieldContainer &lo_elemToNode,
+                       std::vector<bool> &lo_nodeIsOwned,
+                       std::vector<LocalOrdinal> &hi_to_lo_map,
+                       int &lo_numOwnedNodes);
+
+template <class LocalOrdinal, class LOFieldContainer>
+void BuildLoElemToNodeViaRepresentatives(const LOFieldContainer &hi_elemToNode,
+                                         const std::vector<bool> &hi_nodeIsOwned,
+                                         const LOFieldContainer &lo_elemToHiRepresentativeNode,
+                                         LOFieldContainer &lo_elemToNode,
+                                         std::vector<bool> &lo_nodeIsOwned,
+                                         std::vector<LocalOrdinal> &hi_to_lo_map,
+                                         int &lo_numOwnedNodes);
+
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GenerateColMapFromImport(const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> &hi_importer, const std::vector<LocalOrdinal> &hi_to_lo_map, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &lo_domainMap, const size_t &lo_columnMapLength, RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &lo_columnMap);
+
+template <class Basis, class SCFieldContainer>
+void GenerateRepresentativeBasisNodes(const Basis &basis, const SCFieldContainer &ReferenceNodeLocations, const double threshold, std::vector<std::vector<size_t> > &representative_node_candidates);
+
+// ! Given an element to (global) node map and a basis, determine one global ordinal per geometric entity (vertex, edge, face,
+// ! interior).  On exit, seeds container is of dimension (spaceDim+1), and contains a sorted vector of local ordinals
+// ! belonging to entities of that dimension.  Only locally-owned degrees of freedom (as determined by rowMap and columnMap)
+// ! will be stored in seeds.
+template <class Basis, class LOFieldContainer, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FindGeometricSeedOrdinals(Teuchos::RCP<Basis> basis, const LOFieldContainer &elementToNodeMap,
+                               std::vector<std::vector<LocalOrdinal> > &seeds,
+                               const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &rowMap,
+                               const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &columnMap);
+
+}  // namespace MueLuIntrepid
+}  // namespace MueLu
+
+#define MUELU_INTREPIDPCOARSENFACTORY_SHORT
+#endif  // MUELU_INTREPIDPCOARSENFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_def.hpp b/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_def.hpp
index 931a668457b2..802b71eb981f 100644
--- a/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_def.hpp
+++ b/packages/muelu/src/Transfers/PCoarsen/MueLu_IntrepidPCoarsenFactory_def.hpp
@@ -63,126 +63,112 @@
 
 // Intrepid Headers
 
-//Intrepid_HGRAD_HEX_C1_FEM.hpp
-//Intrepid_HGRAD_HEX_C2_FEM.hpp
-//Intrepid_HGRAD_HEX_Cn_FEM.hpp
-//Intrepid_HGRAD_HEX_I2_FEM.hpp
+// Intrepid_HGRAD_HEX_C1_FEM.hpp
+// Intrepid_HGRAD_HEX_C2_FEM.hpp
+// Intrepid_HGRAD_HEX_Cn_FEM.hpp
+// Intrepid_HGRAD_HEX_I2_FEM.hpp
 #include "Intrepid2_HGRAD_LINE_C1_FEM.hpp"
 #include "Intrepid2_HGRAD_LINE_Cn_FEM.hpp"
-//Intrepid_HGRAD_LINE_Cn_FEM_JACOBI.hpp
-//Intrepid_HGRAD_POLY_C1_FEM.hpp
-//Intrepid_HGRAD_PYR_C1_FEM.hpp
-//Intrepid_HGRAD_PYR_I2_FEM.hpp
+// Intrepid_HGRAD_LINE_Cn_FEM_JACOBI.hpp
+// Intrepid_HGRAD_POLY_C1_FEM.hpp
+// Intrepid_HGRAD_PYR_C1_FEM.hpp
+// Intrepid_HGRAD_PYR_I2_FEM.hpp
 #include "Intrepid2_HGRAD_QUAD_C1_FEM.hpp"
 //#include Intrepid_HGRAD_QUAD_C2_FEM.hpp
 #include "Intrepid2_HGRAD_QUAD_Cn_FEM.hpp"
-//Intrepid_HGRAD_TET_C1_FEM.hpp
-//Intrepid_HGRAD_TET_C2_FEM.hpp
-//Intrepid_HGRAD_TET_Cn_FEM.hpp
-//Intrepid_HGRAD_TET_Cn_FEM_ORTH.hpp
-//Intrepid_HGRAD_TET_COMP12_FEM.hpp
-//Intrepid_HGRAD_TRI_C1_FEM.hpp
-//Intrepid_HGRAD_TRI_C2_FEM.hpp
-//Intrepid_HGRAD_TRI_Cn_FEM.hpp
-//Intrepid_HGRAD_TRI_Cn_FEM_ORTH.hpp
-//Intrepid_HGRAD_WEDGE_C1_FEM.hpp
-//Intrepid_HGRAD_WEDGE_C2_FEM.hpp
-//Intrepid_HGRAD_WEDGE_I2_FEM.hpp
+// Intrepid_HGRAD_TET_C1_FEM.hpp
+// Intrepid_HGRAD_TET_C2_FEM.hpp
+// Intrepid_HGRAD_TET_Cn_FEM.hpp
+// Intrepid_HGRAD_TET_Cn_FEM_ORTH.hpp
+// Intrepid_HGRAD_TET_COMP12_FEM.hpp
+// Intrepid_HGRAD_TRI_C1_FEM.hpp
+// Intrepid_HGRAD_TRI_C2_FEM.hpp
+// Intrepid_HGRAD_TRI_Cn_FEM.hpp
+// Intrepid_HGRAD_TRI_Cn_FEM_ORTH.hpp
+// Intrepid_HGRAD_WEDGE_C1_FEM.hpp
+// Intrepid_HGRAD_WEDGE_C2_FEM.hpp
+// Intrepid_HGRAD_WEDGE_I2_FEM.hpp
 
 // Helper Macro to avoid "unrequested" warnings
-#define MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(level,ename,entry) \
-  {if (level.IsRequested(ename,this) || level.GetKeepFlag(ename,this) != 0) this->Set(level,ename,entry);}
-
-
+#define MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(level, ename, entry)                                              \
+  {                                                                                                            \
+    if (level.IsRequested(ename, this) || level.GetKeepFlag(ename, this) != 0) this->Set(level, ename, entry); \
+  }
 
 namespace MueLu {
 
-
 /*********************************************************************************************************/
 namespace MueLuIntrepid {
-inline std::string tolower(const std::string & str) {
+inline std::string tolower(const std::string &str) {
   std::string data(str);
   std::transform(data.begin(), data.end(), data.begin(), ::tolower);
   return data;
 }
 
-
 /*********************************************************************************************************/
-  template<class Basis, class LOFieldContainer, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void FindGeometricSeedOrdinals(Teuchos::RCP<Basis> basis, const LOFieldContainer &elementToNodeMap,
-                                 std::vector<std::vector<LocalOrdinal> > &seeds,
-                                 const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> &rowMap,
-                                 const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> &columnMap)
-  {
-
-    // For each subcell represented by the elements in elementToNodeMap, we want to identify a globally
-    // unique degree of freedom.  Because the other "seed" interfaces in MueLu expect a local ordinal, we
-    // store local ordinals in the resulting seeds container.
-
-    // The approach is as follows.  For each element, we iterate through the subcells of the domain topology.
-    // We determine which, if any, of these has the lowest global ID owned that is locally owned.  We then insert
-    // the local ID corresponding to this in a vector<set<int>> container whose outer index is the spatial dimension
-    // of the subcell.  The set lets us conveniently enforce uniqueness of the stored LIDs.
-
-    shards::CellTopology cellTopo = basis->getBaseCellTopology();
-    int spaceDim = cellTopo.getDimension();
-    seeds.clear();
-    seeds.resize(spaceDim + 1);
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-
-    LocalOrdinal  lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
-    GlobalOrdinal go_invalid = Teuchos::OrdinalTraits<GO>::invalid();
-
-
-    std::vector<std::set<LocalOrdinal>> seedSets(spaceDim+1);
+template <class Basis, class LOFieldContainer, class LocalOrdinal, class GlobalOrdinal, class Node>
+void FindGeometricSeedOrdinals(Teuchos::RCP<Basis> basis, const LOFieldContainer &elementToNodeMap,
+                               std::vector<std::vector<LocalOrdinal>> &seeds,
+                               const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &rowMap,
+                               const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &columnMap) {
+  // For each subcell represented by the elements in elementToNodeMap, we want to identify a globally
+  // unique degree of freedom.  Because the other "seed" interfaces in MueLu expect a local ordinal, we
+  // store local ordinals in the resulting seeds container.
+
+  // The approach is as follows.  For each element, we iterate through the subcells of the domain topology.
+  // We determine which, if any, of these has the lowest global ID owned that is locally owned.  We then insert
+  // the local ID corresponding to this in a vector<set<int>> container whose outer index is the spatial dimension
+  // of the subcell.  The set lets us conveniently enforce uniqueness of the stored LIDs.
+
+  shards::CellTopology cellTopo = basis->getBaseCellTopology();
+  int spaceDim                  = cellTopo.getDimension();
+  seeds.clear();
+  seeds.resize(spaceDim + 1);
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
 
-    int numCells = elementToNodeMap.extent(0);
-    auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
-    Kokkos::deep_copy(elementToNodeMap_host,elementToNodeMap);
-    for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-    {
-      for (int d=0; d<=spaceDim; d++)
-      {
-        int subcellCount = cellTopo.getSubcellCount(d);
-        for (int subcord=0; subcord<subcellCount; subcord++)
-        {
-          int dofCount = basis->getDofCount(d,subcord);
-          if (dofCount == 0) continue;
-          // otherwise, we want to insert the LID corresponding to the least globalID that is locally owned
-          GO leastGlobalDofOrdinal = go_invalid;
-          LO LID_leastGlobalDofOrdinal = lo_invalid;
-          for (int basisOrdinalOrdinal=0; basisOrdinalOrdinal<dofCount; basisOrdinalOrdinal++)
-          {
-            int basisOrdinal = basis->getDofOrdinal(d,subcord,basisOrdinalOrdinal);
-            int colLID = elementToNodeMap_host(cellOrdinal,basisOrdinal);
-            if (colLID != Teuchos::OrdinalTraits<LO>::invalid())
-            {
-              GlobalOrdinal colGID = columnMap.getGlobalElement(colLID);
-              LocalOrdinal rowLID = rowMap.getLocalElement(colGID);
-              if (rowLID != lo_invalid)
-              {
-                if ((leastGlobalDofOrdinal == go_invalid) || (colGID < leastGlobalDofOrdinal))
-                {
-                  // replace with rowLID
-                  leastGlobalDofOrdinal = colGID;
-                  LID_leastGlobalDofOrdinal = rowLID;
-                }
+  LocalOrdinal lo_invalid  = Teuchos::OrdinalTraits<LO>::invalid();
+  GlobalOrdinal go_invalid = Teuchos::OrdinalTraits<GO>::invalid();
+
+  std::vector<std::set<LocalOrdinal>> seedSets(spaceDim + 1);
+
+  int numCells               = elementToNodeMap.extent(0);
+  auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
+  Kokkos::deep_copy(elementToNodeMap_host, elementToNodeMap);
+  for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+    for (int d = 0; d <= spaceDim; d++) {
+      int subcellCount = cellTopo.getSubcellCount(d);
+      for (int subcord = 0; subcord < subcellCount; subcord++) {
+        int dofCount = basis->getDofCount(d, subcord);
+        if (dofCount == 0) continue;
+        // otherwise, we want to insert the LID corresponding to the least globalID that is locally owned
+        GO leastGlobalDofOrdinal     = go_invalid;
+        LO LID_leastGlobalDofOrdinal = lo_invalid;
+        for (int basisOrdinalOrdinal = 0; basisOrdinalOrdinal < dofCount; basisOrdinalOrdinal++) {
+          int basisOrdinal = basis->getDofOrdinal(d, subcord, basisOrdinalOrdinal);
+          int colLID       = elementToNodeMap_host(cellOrdinal, basisOrdinal);
+          if (colLID != Teuchos::OrdinalTraits<LO>::invalid()) {
+            GlobalOrdinal colGID = columnMap.getGlobalElement(colLID);
+            LocalOrdinal rowLID  = rowMap.getLocalElement(colGID);
+            if (rowLID != lo_invalid) {
+              if ((leastGlobalDofOrdinal == go_invalid) || (colGID < leastGlobalDofOrdinal)) {
+                // replace with rowLID
+                leastGlobalDofOrdinal     = colGID;
+                LID_leastGlobalDofOrdinal = rowLID;
               }
             }
           }
-          if (leastGlobalDofOrdinal != go_invalid)
-          {
-            seedSets[d].insert(LID_leastGlobalDofOrdinal);
-          }
+        }
+        if (leastGlobalDofOrdinal != go_invalid) {
+          seedSets[d].insert(LID_leastGlobalDofOrdinal);
         }
       }
     }
-    for (int d=0; d<=spaceDim;d++)
-    {
-      seeds[d] = std::vector<LocalOrdinal>(seedSets[d].begin(),seedSets[d].end());
-    }
   }
+  for (int d = 0; d <= spaceDim; d++) {
+    seeds[d] = std::vector<LocalOrdinal>(seedSets[d].begin(), seedSets[d].end());
+  }
+}
 
 /*********************************************************************************************************/
 // Syntax [HGRAD|HCURL|HDIV][_| ][HEX|LINE|POLY|PYR|QUAD|TET|TRI|WEDGE][_| ][C|I][1|2|n]
@@ -191,51 +177,53 @@ inline std::string tolower(const std::string & str) {
 // Outputs:
 //  degree - order of resulting discretization
 //  return value - Intrepid2 basis correspionding to the name
-template<class Scalar,class KokkosExecutionSpace>
-Teuchos::RCP< Intrepid2::Basis<KokkosExecutionSpace,Scalar,Scalar> > BasisFactory(const std::string & name, int & degree)
-{
-    using std::string;
-    using Teuchos::rcp;
-    string myerror("IntrepidBasisFactory: cannot parse string name '"+name+"'");
-
-    // Syntax [HGRAD|HCURL|HDIV][_| ][HEX|LINE|POLY|PYR|QUAD|TET|TRI|WEDGE][_| ][C|I][1|2|n]
-
-    // Get the derivative type
-    size_t pos1 = name.find_first_of(" _");
-    if(pos1==0) throw std::runtime_error(myerror);
-    string deriv = tolower(name.substr(0,pos1));
-    if(deriv!="hgrad" && deriv!="hcurl" && deriv!="hdiv") throw std::runtime_error(myerror);
-
-    // Get the element type
-    pos1++;
-    size_t pos2 = name.find_first_of(" _",pos1);
-    if(pos2==0) throw std::runtime_error(myerror);
-    string el = tolower(name.substr(pos1,pos2-pos1));
-    if(el!="hex" && el!="line" && el!="poly" && el!="pyr" && el!="quad" && el!="tet" && el!="tri" && el!="wedge") throw std::runtime_error(myerror);
-
-    // Get the polynomial type
-    pos2++;
-    string poly = tolower(name.substr(pos2,1));
-    if(poly!="c" && poly!="i") throw std::runtime_error(myerror);
-
-    // Get the degree
-    pos2++;
-    degree=std::stoi(name.substr(pos2,1));
-    if(degree<=0) throw std::runtime_error(myerror);
-
-    // FIXME LATER: Allow for alternative point types for Kirby elements
-    if(deriv=="hgrad" && el=="quad" && poly=="c"){
-      if(degree==1) return rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<KokkosExecutionSpace,Scalar,Scalar>());
-      else          return rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<KokkosExecutionSpace,Scalar,Scalar>(degree,Intrepid2::POINTTYPE_EQUISPACED));
-    }
-    else if(deriv=="hgrad" && el=="line" && poly=="c"){
-      if(degree==1) return rcp(new Intrepid2::Basis_HGRAD_LINE_C1_FEM<KokkosExecutionSpace,Scalar,Scalar>());
-      else          return rcp(new Intrepid2::Basis_HGRAD_LINE_Cn_FEM<KokkosExecutionSpace,Scalar,Scalar>(degree,Intrepid2::POINTTYPE_EQUISPACED));
-    }
+template <class Scalar, class KokkosExecutionSpace>
+Teuchos::RCP<Intrepid2::Basis<KokkosExecutionSpace, Scalar, Scalar>> BasisFactory(const std::string &name, int &degree) {
+  using std::string;
+  using Teuchos::rcp;
+  string myerror("IntrepidBasisFactory: cannot parse string name '" + name + "'");
+
+  // Syntax [HGRAD|HCURL|HDIV][_| ][HEX|LINE|POLY|PYR|QUAD|TET|TRI|WEDGE][_| ][C|I][1|2|n]
+
+  // Get the derivative type
+  size_t pos1 = name.find_first_of(" _");
+  if (pos1 == 0) throw std::runtime_error(myerror);
+  string deriv = tolower(name.substr(0, pos1));
+  if (deriv != "hgrad" && deriv != "hcurl" && deriv != "hdiv") throw std::runtime_error(myerror);
+
+  // Get the element type
+  pos1++;
+  size_t pos2 = name.find_first_of(" _", pos1);
+  if (pos2 == 0) throw std::runtime_error(myerror);
+  string el = tolower(name.substr(pos1, pos2 - pos1));
+  if (el != "hex" && el != "line" && el != "poly" && el != "pyr" && el != "quad" && el != "tet" && el != "tri" && el != "wedge") throw std::runtime_error(myerror);
+
+  // Get the polynomial type
+  pos2++;
+  string poly = tolower(name.substr(pos2, 1));
+  if (poly != "c" && poly != "i") throw std::runtime_error(myerror);
+
+  // Get the degree
+  pos2++;
+  degree = std::stoi(name.substr(pos2, 1));
+  if (degree <= 0) throw std::runtime_error(myerror);
+
+  // FIXME LATER: Allow for alternative point types for Kirby elements
+  if (deriv == "hgrad" && el == "quad" && poly == "c") {
+    if (degree == 1)
+      return rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<KokkosExecutionSpace, Scalar, Scalar>());
+    else
+      return rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<KokkosExecutionSpace, Scalar, Scalar>(degree, Intrepid2::POINTTYPE_EQUISPACED));
+  } else if (deriv == "hgrad" && el == "line" && poly == "c") {
+    if (degree == 1)
+      return rcp(new Intrepid2::Basis_HGRAD_LINE_C1_FEM<KokkosExecutionSpace, Scalar, Scalar>());
+    else
+      return rcp(new Intrepid2::Basis_HGRAD_LINE_Cn_FEM<KokkosExecutionSpace, Scalar, Scalar>(degree, Intrepid2::POINTTYPE_EQUISPACED));
+  }
 
-    // Error out
-    throw std::runtime_error(myerror);
-    TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+  // Error out
+  throw std::runtime_error(myerror);
+  TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
 }
 
 /*********************************************************************************************************/
@@ -245,33 +233,29 @@ Teuchos::RCP< Intrepid2::Basis<KokkosExecutionSpace,Scalar,Scalar> > BasisFactor
 // Outputs:
 //  lo_node_in_hi   - std::vector<size_t> of size lo dofs in the reference element, which describes the coindcident hi dots
 //  hi_DofCoords    - FC<Scalar> of size (#hi dofs, dim) with the coordinate locations of the hi dofs on the reference element
-template<class Scalar,class KokkosDeviceType>
-void IntrepidGetP1NodeInHi(const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space,Scalar,Scalar> >&hi_basis,
-                           std::vector<size_t> & lo_node_in_hi,
-                           Kokkos::DynRankView<Scalar,KokkosDeviceType> & hi_DofCoords) {
-
+template <class Scalar, class KokkosDeviceType>
+void IntrepidGetP1NodeInHi(const Teuchos::RCP<Intrepid2::Basis<typename KokkosDeviceType::execution_space, Scalar, Scalar>> &hi_basis,
+                           std::vector<size_t> &lo_node_in_hi,
+                           Kokkos::DynRankView<Scalar, KokkosDeviceType> &hi_DofCoords) {
   typedef typename KokkosDeviceType::execution_space KokkosExecutionSpace;
   // Figure out which unknowns in hi_basis correspond to nodes on lo_basis. This varies by element type.
-  size_t degree         = hi_basis->getDegree();
+  size_t degree = hi_basis->getDegree();
   lo_node_in_hi.resize(0);
 
-  if(!rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<KokkosExecutionSpace,Scalar,Scalar> >(hi_basis).is_null()) {
+  if (!rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<KokkosExecutionSpace, Scalar, Scalar>>(hi_basis).is_null()) {
     // HGRAD QUAD Cn: Numbering as per the Kirby convention (straight across, bottom to top)
-    lo_node_in_hi.insert(lo_node_in_hi.end(),{0,degree, (degree+1)*(degree+1)-1, degree*(degree+1)});
-  }
-  else if(!rcp_dynamic_cast<Intrepid2::Basis_HGRAD_LINE_Cn_FEM<KokkosExecutionSpace,Scalar,Scalar> >(hi_basis).is_null()) {
+    lo_node_in_hi.insert(lo_node_in_hi.end(), {0, degree, (degree + 1) * (degree + 1) - 1, degree * (degree + 1)});
+  } else if (!rcp_dynamic_cast<Intrepid2::Basis_HGRAD_LINE_Cn_FEM<KokkosExecutionSpace, Scalar, Scalar>>(hi_basis).is_null()) {
     // HGRAD LINE Cn: Numbering as per the Kirby convention (straight across)
-    lo_node_in_hi.insert(lo_node_in_hi.end(),{0,degree});
-  }
-  else
+    lo_node_in_hi.insert(lo_node_in_hi.end(), {0, degree});
+  } else
     throw std::runtime_error("IntrepidPCoarsenFactory: Unknown element type");
 
   // Get coordinates of the hi_basis dof's
-  Kokkos::resize(hi_DofCoords,hi_basis->getCardinality(),hi_basis->getBaseCellTopology().getDimension());
+  Kokkos::resize(hi_DofCoords, hi_basis->getCardinality(), hi_basis->getBaseCellTopology().getDimension());
   hi_basis->getDofCoords(hi_DofCoords);
 }
 
-
 /*********************************************************************************************************/
 // Given a list of candidates picks a definitive list of "representative" higher order nodes for each lo order node via the "smallest GID" rule
 // Input:
@@ -280,40 +264,40 @@ void IntrepidGetP1NodeInHi(const Teuchos::RCP<Intrepid2::Basis<typename KokkosDe
 //  hi_columnMap    - Column map of the higher order matrix
 // Output:
 //  lo_elemToHiRepresentativeNode - FC<LO> of size (# elements, # lo dofs per element) listing the hi unknown chosen as the single representative for each lo unknown for counting purposes
-template<class LocalOrdinal, class GlobalOrdinal, class Node, class LOFieldContainer>
-void GenerateLoNodeInHiViaGIDs(const std::vector<std::vector<size_t> > & candidates,const LOFieldContainer & hi_elemToNode,
-                               RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & hi_columnMap,
-                               LOFieldContainer & lo_elemToHiRepresentativeNode) {
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class LOFieldContainer>
+void GenerateLoNodeInHiViaGIDs(const std::vector<std::vector<size_t>> &candidates, const LOFieldContainer &hi_elemToNode,
+                               RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> &hi_columnMap,
+                               LOFieldContainer &lo_elemToHiRepresentativeNode) {
   typedef GlobalOrdinal GO;
 
   // Given: A set of "candidate" hi-DOFs to serve as the "representative" DOF for each lo-DOF on the reference element.
   // Algorithm:  For each element, we choose the lowest GID of the candidates for each DOF to generate the lo_elemToHiRepresentativeNode map
 
-   size_t numElem     = hi_elemToNode.extent(0);
-   size_t lo_nperel   = candidates.size();
-   Kokkos::resize(lo_elemToHiRepresentativeNode,numElem, lo_nperel);
-
-   auto lo_elemToHiRepresentativeNode_host = Kokkos::create_mirror_view(lo_elemToHiRepresentativeNode);
-   auto hi_elemToNode_host = Kokkos::create_mirror_view(hi_elemToNode);
-   Kokkos::deep_copy(hi_elemToNode_host, hi_elemToNode);
-   for(size_t i=0; i<numElem; i++)
-     for(size_t j=0; j<lo_nperel; j++) {
-       if(candidates[j].size() == 1)
-         lo_elemToHiRepresentativeNode_host(i,j) =  hi_elemToNode_host(i,candidates[j][0]);
-       else {
-         // First we get the GIDs for each candidate
-         std::vector<GO> GID(candidates[j].size());
-         for(size_t k=0; k<(size_t)candidates[j].size(); k++)
-           GID[k] = hi_columnMap->getGlobalElement(hi_elemToNode_host(i,candidates[j][k]));
-
-         // Find the one with smallest GID
-         size_t which = std::distance(GID.begin(),std::min_element(GID.begin(),GID.end()));
-
-         // Record this
-         lo_elemToHiRepresentativeNode_host(i,j) =  hi_elemToNode_host(i,candidates[j][which]);
-       }
-     }
-   Kokkos::deep_copy(lo_elemToHiRepresentativeNode, lo_elemToHiRepresentativeNode_host);
+  size_t numElem   = hi_elemToNode.extent(0);
+  size_t lo_nperel = candidates.size();
+  Kokkos::resize(lo_elemToHiRepresentativeNode, numElem, lo_nperel);
+
+  auto lo_elemToHiRepresentativeNode_host = Kokkos::create_mirror_view(lo_elemToHiRepresentativeNode);
+  auto hi_elemToNode_host                 = Kokkos::create_mirror_view(hi_elemToNode);
+  Kokkos::deep_copy(hi_elemToNode_host, hi_elemToNode);
+  for (size_t i = 0; i < numElem; i++)
+    for (size_t j = 0; j < lo_nperel; j++) {
+      if (candidates[j].size() == 1)
+        lo_elemToHiRepresentativeNode_host(i, j) = hi_elemToNode_host(i, candidates[j][0]);
+      else {
+        // First we get the GIDs for each candidate
+        std::vector<GO> GID(candidates[j].size());
+        for (size_t k = 0; k < (size_t)candidates[j].size(); k++)
+          GID[k] = hi_columnMap->getGlobalElement(hi_elemToNode_host(i, candidates[j][k]));
+
+        // Find the one with smallest GID
+        size_t which = std::distance(GID.begin(), std::min_element(GID.begin(), GID.end()));
+
+        // Record this
+        lo_elemToHiRepresentativeNode_host(i, j) = hi_elemToNode_host(i, candidates[j][which]);
+      }
+    }
+  Kokkos::deep_copy(lo_elemToHiRepresentativeNode, lo_elemToHiRepresentativeNode_host);
 }
 
 /*********************************************************************************************************/
@@ -327,71 +311,68 @@ void GenerateLoNodeInHiViaGIDs(const std::vector<std::vector<size_t> > & candida
 //  hi_to_lo_map    - std::vector<LO> of size equal to hi's column map, which contains the lo id each hi idea maps to (or invalid if it doesn't)
 //  lo_numOwnedNodes- Number of lo owned nodes
 template <class LocalOrdinal, class LOFieldContainer>
-void BuildLoElemToNodeViaRepresentatives(const LOFieldContainer & hi_elemToNode,
-                                         const std::vector<bool> & hi_nodeIsOwned,
-                                         const LOFieldContainer & lo_elemToHiRepresentativeNode,
-                                         LOFieldContainer & lo_elemToNode,
-                                         std::vector<bool> & lo_nodeIsOwned,
-                                         std::vector<LocalOrdinal> & hi_to_lo_map,
-                                         int & lo_numOwnedNodes) {
+void BuildLoElemToNodeViaRepresentatives(const LOFieldContainer &hi_elemToNode,
+                                         const std::vector<bool> &hi_nodeIsOwned,
+                                         const LOFieldContainer &lo_elemToHiRepresentativeNode,
+                                         LOFieldContainer &lo_elemToNode,
+                                         std::vector<bool> &lo_nodeIsOwned,
+                                         std::vector<LocalOrdinal> &hi_to_lo_map,
+                                         int &lo_numOwnedNodes) {
   typedef LocalOrdinal LO;
   using Teuchos::RCP;
   //  printf("CMS:BuildLoElemToNodeViaRepresentatives: hi_elemToNode.rank() = %d hi_elemToNode.size() = %d\n",hi_elemToNode.rank(), hi_elemToNode.size());
   size_t numElem     = hi_elemToNode.extent(0);
   size_t hi_numNodes = hi_nodeIsOwned.size();
-  size_t lo_nperel = lo_elemToHiRepresentativeNode.extent(1);
-  Kokkos::resize(lo_elemToNode,numElem, lo_nperel);
+  size_t lo_nperel   = lo_elemToHiRepresentativeNode.extent(1);
+  Kokkos::resize(lo_elemToNode, numElem, lo_nperel);
 
   // Start by flagginc the representative nodes
   auto lo_elemToHiRepresentativeNode_host = Kokkos::create_mirror_view(lo_elemToHiRepresentativeNode);
   Kokkos::deep_copy(lo_elemToHiRepresentativeNode_host, lo_elemToHiRepresentativeNode);
-  std::vector<bool> is_low_order(hi_numNodes,false);
-  for(size_t i=0; i<numElem; i++)
-    for(size_t j=0; j<lo_nperel; j++) {
-      LO id = lo_elemToHiRepresentativeNode_host(i,j);
-      is_low_order[id] = true; // This can overwrite and that is OK.
+  std::vector<bool> is_low_order(hi_numNodes, false);
+  for (size_t i = 0; i < numElem; i++)
+    for (size_t j = 0; j < lo_nperel; j++) {
+      LO id            = lo_elemToHiRepresentativeNode_host(i, j);
+      is_low_order[id] = true;  // This can overwrite and that is OK.
     }
 
   // Count the number of lo owned nodes, generating a local index for lo nodes
-  lo_numOwnedNodes=0;
-  size_t lo_numNodes=0;
-  hi_to_lo_map.resize(hi_numNodes,Teuchos::OrdinalTraits<LO>::invalid());
+  lo_numOwnedNodes   = 0;
+  size_t lo_numNodes = 0;
+  hi_to_lo_map.resize(hi_numNodes, Teuchos::OrdinalTraits<LO>::invalid());
 
-  for(size_t i=0; i<hi_numNodes; i++)
-    if(is_low_order[i]) {
+  for (size_t i = 0; i < hi_numNodes; i++)
+    if (is_low_order[i]) {
       hi_to_lo_map[i] = lo_numNodes;
       lo_numNodes++;
-      if(hi_nodeIsOwned[i]) lo_numOwnedNodes++;
+      if (hi_nodeIsOwned[i]) lo_numOwnedNodes++;
     }
 
   // Flag the owned lo nodes
-  lo_nodeIsOwned.resize(lo_numNodes,false);
-  for(size_t i=0; i<hi_numNodes; i++) {
-    if(is_low_order[i] && hi_nodeIsOwned[i])
-      lo_nodeIsOwned[hi_to_lo_map[i]]=true;
+  lo_nodeIsOwned.resize(lo_numNodes, false);
+  for (size_t i = 0; i < hi_numNodes; i++) {
+    if (is_low_order[i] && hi_nodeIsOwned[i])
+      lo_nodeIsOwned[hi_to_lo_map[i]] = true;
   }
 
   // Translate lo_elemToNode to a lo local index
   auto lo_elemToNode_host = Kokkos::create_mirror_view(lo_elemToNode);
-  for(size_t i=0; i<numElem; i++)
-    for(size_t j=0; j<lo_nperel; j++)
-      lo_elemToNode_host(i,j) = hi_to_lo_map[lo_elemToHiRepresentativeNode_host(i,j)];
-
+  for (size_t i = 0; i < numElem; i++)
+    for (size_t j = 0; j < lo_nperel; j++)
+      lo_elemToNode_host(i, j) = hi_to_lo_map[lo_elemToHiRepresentativeNode_host(i, j)];
 
   // Check for the [E|T]petra column map ordering property, namely LIDs for owned nodes should all appear first.
   // Since we're injecting from the higher-order mesh, it should be true, but we should add an error check & throw in case.
-  bool map_ordering_test_passed=true;
-  for(size_t i=0; i<lo_numNodes-1; i++)
-    if(!lo_nodeIsOwned[i] && lo_nodeIsOwned[i+1])
-      map_ordering_test_passed=false;
+  bool map_ordering_test_passed = true;
+  for (size_t i = 0; i < lo_numNodes - 1; i++)
+    if (!lo_nodeIsOwned[i] && lo_nodeIsOwned[i + 1])
+      map_ordering_test_passed = false;
 
-  if(!map_ordering_test_passed)
+  if (!map_ordering_test_passed)
     throw std::runtime_error("MueLu::MueLuIntrepid::BuildLoElemToNodeViaRepresentatives failed map ordering test");
   Kokkos::deep_copy(lo_elemToNode, lo_elemToNode_host);
-
 }
 
-
 /*********************************************************************************************************/
 // Inputs:
 //  hi_elemToNode   - FC<LO> containing the high order element-to-node map
@@ -404,14 +385,14 @@ void BuildLoElemToNodeViaRepresentatives(const LOFieldContainer & hi_elemToNode,
 //  hi_to_lo_map    - std::vector<LO> of size equal to hi's column map, which contains the lo id each hi idea maps to (or invalid if it doesn't)
 //  lo_numOwnedNodes- Number of lo owned nodes
 template <class LocalOrdinal, class LOFieldContainer>
-void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
-                       const std::vector<bool> & hi_nodeIsOwned,
-                       const std::vector<size_t> & lo_node_in_hi,
-                       const Teuchos::ArrayRCP<const int> & hi_isDirichlet,
-                       LOFieldContainer & lo_elemToNode,
-                       std::vector<bool> & lo_nodeIsOwned,
-                       std::vector<LocalOrdinal> & hi_to_lo_map,
-                       int & lo_numOwnedNodes) {
+void BuildLoElemToNode(const LOFieldContainer &hi_elemToNode,
+                       const std::vector<bool> &hi_nodeIsOwned,
+                       const std::vector<size_t> &lo_node_in_hi,
+                       const Teuchos::ArrayRCP<const int> &hi_isDirichlet,
+                       LOFieldContainer &lo_elemToNode,
+                       std::vector<bool> &lo_nodeIsOwned,
+                       std::vector<LocalOrdinal> &hi_to_lo_map,
+                       int &lo_numOwnedNodes) {
   typedef LocalOrdinal LO;
   using Teuchos::RCP;
   LocalOrdinal LOINVALID = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
@@ -421,63 +402,62 @@ void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
   size_t hi_numNodes = hi_nodeIsOwned.size();
 
   size_t lo_nperel = lo_node_in_hi.size();
-  Kokkos::resize(lo_elemToNode,numElem, lo_nperel);
+  Kokkos::resize(lo_elemToNode, numElem, lo_nperel);
 
   // Build lo_elemToNode (in the hi local index ordering) and flag owned ones
-  std::vector<bool> is_low_order(hi_numNodes,false);
+  std::vector<bool> is_low_order(hi_numNodes, false);
   auto hi_elemToNode_host = Kokkos::create_mirror_view(hi_elemToNode);
   Kokkos::deep_copy(hi_elemToNode_host, hi_elemToNode);
   auto lo_elemToNode_host = Kokkos::create_mirror_view(lo_elemToNode);
-  for(size_t i=0; i<numElem; i++)
-    for(size_t j=0; j<lo_nperel; j++) {
-      LO lid = hi_elemToNode_host(i,lo_node_in_hi[j]);
+  for (size_t i = 0; i < numElem; i++)
+    for (size_t j = 0; j < lo_nperel; j++) {
+      LO lid = hi_elemToNode_host(i, lo_node_in_hi[j]);
 
       // Remove Dirichlet
-      if(hi_isDirichlet[lid])
-        lo_elemToNode_host(i,j) = LOINVALID;
+      if (hi_isDirichlet[lid])
+        lo_elemToNode_host(i, j) = LOINVALID;
       else {
-        lo_elemToNode_host(i,j)  = lid;
-        is_low_order[hi_elemToNode_host(i,lo_node_in_hi[j])] = true; // This can overwrite and that is OK.
+        lo_elemToNode_host(i, j)                              = lid;
+        is_low_order[hi_elemToNode_host(i, lo_node_in_hi[j])] = true;  // This can overwrite and that is OK.
       }
     }
 
   // Count the number of lo owned nodes, generating a local index for lo nodes
-  lo_numOwnedNodes=0;
-  size_t lo_numNodes=0;
-  hi_to_lo_map.resize(hi_numNodes,Teuchos::OrdinalTraits<LO>::invalid());
+  lo_numOwnedNodes   = 0;
+  size_t lo_numNodes = 0;
+  hi_to_lo_map.resize(hi_numNodes, Teuchos::OrdinalTraits<LO>::invalid());
 
-  for(size_t i=0; i<hi_numNodes; i++)
-    if(is_low_order[i]) {
+  for (size_t i = 0; i < hi_numNodes; i++)
+    if (is_low_order[i]) {
       hi_to_lo_map[i] = lo_numNodes;
       lo_numNodes++;
-      if(hi_nodeIsOwned[i]) lo_numOwnedNodes++;
+      if (hi_nodeIsOwned[i]) lo_numOwnedNodes++;
     }
 
   // Flag the owned lo nodes
-  lo_nodeIsOwned.resize(lo_numNodes,false);
-  for(size_t i=0; i<hi_numNodes; i++) {
-    if(is_low_order[i] && hi_nodeIsOwned[i])
-      lo_nodeIsOwned[hi_to_lo_map[i]]=true;
+  lo_nodeIsOwned.resize(lo_numNodes, false);
+  for (size_t i = 0; i < hi_numNodes; i++) {
+    if (is_low_order[i] && hi_nodeIsOwned[i])
+      lo_nodeIsOwned[hi_to_lo_map[i]] = true;
   }
 
   // Translate lo_elemToNode to a lo local index
-  for(size_t i=0; i<numElem; i++)
-    for(size_t j=0; j<lo_nperel; j++) {
-      if(lo_elemToNode_host(i,j) != LOINVALID)
-        lo_elemToNode_host(i,j) = hi_to_lo_map[lo_elemToNode_host(i,j)];
+  for (size_t i = 0; i < numElem; i++)
+    for (size_t j = 0; j < lo_nperel; j++) {
+      if (lo_elemToNode_host(i, j) != LOINVALID)
+        lo_elemToNode_host(i, j) = hi_to_lo_map[lo_elemToNode_host(i, j)];
     }
   Kokkos::deep_copy(lo_elemToNode, lo_elemToNode_host);
 
   // Check for the [E|T]petra column map ordering property, namely LIDs for owned nodes should all appear first.
   // Since we're injecting from the higher-order mesh, it should be true, but we should add an error check & throw in case.
-  bool map_ordering_test_passed=true;
-  for(size_t i=0; i<lo_numNodes-1; i++)
-    if(!lo_nodeIsOwned[i] && lo_nodeIsOwned[i+1])
-      map_ordering_test_passed=false;
+  bool map_ordering_test_passed = true;
+  for (size_t i = 0; i < lo_numNodes - 1; i++)
+    if (!lo_nodeIsOwned[i] && lo_nodeIsOwned[i + 1])
+      map_ordering_test_passed = false;
 
-  if(!map_ordering_test_passed)
+  if (!map_ordering_test_passed)
     throw std::runtime_error("MueLu::MueLuIntrepid::BuildLoElemToNode failed map ordering test");
-
 }
 
 /*********************************************************************************************************/
@@ -489,13 +469,13 @@ void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
 //  lo_columnMapLength - Number of local columns in the lo column map
 // Output:
 //  lo_columnMap       - Column map of the lower order matrix
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GenerateColMapFromImport(const Xpetra::Import<LocalOrdinal,GlobalOrdinal, Node> & hi_importer,const std::vector<LocalOrdinal> &hi_to_lo_map,const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> & lo_domainMap, const size_t & lo_columnMapLength, RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & lo_columnMap) {
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+void GenerateColMapFromImport(const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> &hi_importer, const std::vector<LocalOrdinal> &hi_to_lo_map, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> &lo_domainMap, const size_t &lo_columnMapLength, RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> &lo_columnMap) {
   typedef LocalOrdinal LO;
   typedef GlobalOrdinal GO;
   typedef Node NO;
-  typedef Xpetra::Map<LO,GO,NO> Map;
-  typedef Xpetra::Vector<GO,LO,GO,NO> GOVector;
+  typedef Xpetra::Map<LO, GO, NO> Map;
+  typedef Xpetra::Vector<GO, LO, GO, NO> GOVector;
 
   GO go_invalid = Teuchos::OrdinalTraits<GO>::invalid();
   LO lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
@@ -510,30 +490,31 @@ void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
   RCP<GOVector> dvec = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(hi_domainMap);
   {
     ArrayRCP<GO> dvec_data = dvec->getDataNonConst(0);
-    for(size_t i=0; i<hi_domainMap->getLocalNumElements(); i++) {
-      if(hi_to_lo_map[i]!=lo_invalid) dvec_data[i] = lo_domainMap.getGlobalElement(hi_to_lo_map[i]);
-      else dvec_data[i] = go_invalid;
+    for (size_t i = 0; i < hi_domainMap->getLocalNumElements(); i++) {
+      if (hi_to_lo_map[i] != lo_invalid)
+        dvec_data[i] = lo_domainMap.getGlobalElement(hi_to_lo_map[i]);
+      else
+        dvec_data[i] = go_invalid;
     }
   }
 
-
-  RCP<GOVector> cvec = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(hi_columnMap,true);
-  cvec->doImport(*dvec,hi_importer,Xpetra::ADD);
+  RCP<GOVector> cvec = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(hi_columnMap, true);
+  cvec->doImport(*dvec, hi_importer, Xpetra::ADD);
 
   // Generate the lo_columnMap
   // HOW: We can use the local hi_to_lo_map from the GID's in cvec to generate the non-contiguous colmap ids.
   Array<GO> lo_col_data(lo_columnMapLength);
   {
     ArrayRCP<GO> cvec_data = cvec->getDataNonConst(0);
-    for(size_t i=0,idx=0; i<hi_columnMap->getLocalNumElements(); i++) {
-      if(hi_to_lo_map[i]!=lo_invalid) {
+    for (size_t i = 0, idx = 0; i < hi_columnMap->getLocalNumElements(); i++) {
+      if (hi_to_lo_map[i] != lo_invalid) {
         lo_col_data[idx] = cvec_data[i];
         idx++;
       }
     }
   }
 
-  lo_columnMap = Xpetra::MapFactory<LO,GO,NO>::Build(lo_domainMap.lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),lo_col_data(),lo_domainMap.getIndexBase(),lo_domainMap.getComm());
+  lo_columnMap = Xpetra::MapFactory<LO, GO, NO>::Build(lo_domainMap.lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), lo_col_data(), lo_domainMap.getIndexBase(), lo_domainMap.getComm());
 }
 
 /*********************************************************************************************************/
@@ -544,21 +525,21 @@ void BuildLoElemToNode(const LOFieldContainer & hi_elemToNode,
 //  threshold              - tolerance for equivalance testing
 // Output:
 //  representative_node_candidates - std::vector<std::vector<size_t> > of lists of "representative candidate" hi dofs for each lo dof
-template<class Basis, class SCFieldContainer>
-void GenerateRepresentativeBasisNodes(const Basis & basis, const SCFieldContainer & ReferenceNodeLocations, const double threshold,std::vector<std::vector<size_t> > & representative_node_candidates) {
- typedef SCFieldContainer FC;
- typedef typename FC::data_type SC;
+template <class Basis, class SCFieldContainer>
+void GenerateRepresentativeBasisNodes(const Basis &basis, const SCFieldContainer &ReferenceNodeLocations, const double threshold, std::vector<std::vector<size_t>> &representative_node_candidates) {
+  typedef SCFieldContainer FC;
+  typedef typename FC::data_type SC;
 
- // Evaluate the linear basis functions at the Pn nodes
- size_t numFieldsHi = ReferenceNodeLocations.extent(0);
- // size_t dim         = ReferenceNodeLocations.extent(1);
- size_t numFieldsLo = basis.getCardinality();
- 
- FC LoValues("LoValues",numFieldsLo,numFieldsHi);
+  // Evaluate the linear basis functions at the Pn nodes
+  size_t numFieldsHi = ReferenceNodeLocations.extent(0);
+  // size_t dim         = ReferenceNodeLocations.extent(1);
+  size_t numFieldsLo = basis.getCardinality();
+
+  FC LoValues("LoValues", numFieldsLo, numFieldsHi);
 
- basis.getValues(LoValues, ReferenceNodeLocations , Intrepid2::OPERATOR_VALUE);
+  basis.getValues(LoValues, ReferenceNodeLocations, Intrepid2::OPERATOR_VALUE);
 
- Kokkos::fence(); // for kernel in getValues
+  Kokkos::fence();  // for kernel in getValues
 
 #if 0
   printf("** LoValues[%d,%d] **\n",(int)numFieldsLo,(int)numFieldsHi);
@@ -573,281 +554,275 @@ void GenerateRepresentativeBasisNodes(const Basis & basis, const SCFieldContaine
   representative_node_candidates.resize(numFieldsLo);
   auto LoValues_host = Kokkos::create_mirror_view(LoValues);
   Kokkos::deep_copy(LoValues_host, LoValues);
-  for(size_t i=0; i<numFieldsLo; i++) {
+  for (size_t i = 0; i < numFieldsLo; i++) {
     // 1st pass: find the max value
     typename Teuchos::ScalarTraits<SC>::magnitudeType vmax = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::zero();
-    for(size_t j=0; j<numFieldsHi; j++)
-      vmax = std::max(vmax,Teuchos::ScalarTraits<SC>::magnitude(LoValues_host(i,j)));
+    for (size_t j = 0; j < numFieldsHi; j++)
+      vmax = std::max(vmax, Teuchos::ScalarTraits<SC>::magnitude(LoValues_host(i, j)));
 
     // 2nd pass: Find all values w/i threshhold of target
-    for(size_t j=0; j<numFieldsHi; j++) {
-      if(Teuchos::ScalarTraits<SC>::magnitude(vmax - LoValues_host(i,j)) < threshold*vmax)
+    for (size_t j = 0; j < numFieldsHi; j++) {
+      if (Teuchos::ScalarTraits<SC>::magnitude(vmax - LoValues_host(i, j)) < threshold * vmax)
         representative_node_candidates[i].push_back(j);
     }
   }
 
   // Sanity check
-  for(size_t i=0; i<numFieldsLo; i++)
-    if(!representative_node_candidates[i].size())
+  for (size_t i = 0; i < numFieldsLo; i++)
+    if (!representative_node_candidates[i].size())
       throw std::runtime_error("ERROR: GenerateRepresentativeBasisNodes: No candidates found!");
-
-
 }
 
-
-
-}//end MueLu::MueLuIntrepid namespace
-
+}  // namespace MueLuIntrepid
 
 /*********************************************************************************************************/
 /*********************************************************************************************************/
 /*********************************************************************************************************/
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GenerateLinearCoarsening_pn_kirby_to_p1(const LOFieldContainer & hi_elemToNode,
-                                                                                                                 const std::vector<bool> & hi_nodeIsOwned,
-                                                                                                                 const SCFieldContainer & hi_DofCoords,
+void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GenerateLinearCoarsening_pn_kirby_to_p1(const LOFieldContainer &hi_elemToNode,
+                                                                                                                 const std::vector<bool> &hi_nodeIsOwned,
+                                                                                                                 const SCFieldContainer &hi_DofCoords,
                                                                                                                  const std::vector<size_t> &lo_node_in_hi,
                                                                                                                  const Basis &lo_basis,
-                                                                                                                 const std::vector<LocalOrdinal> & hi_to_lo_map,
-                                                                                                                 const Teuchos::RCP<const Map> & lo_colMap,
-                                                                                                                 const Teuchos::RCP<const Map> & lo_domainMap,
-                                                                                                                 const Teuchos::RCP<const Map> & hi_map,
-                                                                                                                 Teuchos::RCP<Matrix>& P) const{
+                                                                                                                 const std::vector<LocalOrdinal> &hi_to_lo_map,
+                                                                                                                 const Teuchos::RCP<const Map> &lo_colMap,
+                                                                                                                 const Teuchos::RCP<const Map> &lo_domainMap,
+                                                                                                                 const Teuchos::RCP<const Map> &hi_map,
+                                                                                                                 Teuchos::RCP<Matrix> &P) const {
   typedef SCFieldContainer FC;
   // Evaluate the linear basis functions at the Pn nodes
-  size_t numFieldsHi = hi_elemToNode.extent(1);
-  size_t numFieldsLo = lo_basis.getCardinality();
+  size_t numFieldsHi     = hi_elemToNode.extent(1);
+  size_t numFieldsLo     = lo_basis.getCardinality();
   LocalOrdinal LOINVALID = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
-  FC LoValues_at_HiDofs("LoValues_at_HiDofs",numFieldsLo,numFieldsHi);
+  FC LoValues_at_HiDofs("LoValues_at_HiDofs", numFieldsLo, numFieldsHi);
   lo_basis.getValues(LoValues_at_HiDofs, hi_DofCoords, Intrepid2::OPERATOR_VALUE);
   auto LoValues_at_HiDofs_host = Kokkos::create_mirror_view(LoValues_at_HiDofs);
   Kokkos::deep_copy(LoValues_at_HiDofs_host, LoValues_at_HiDofs);
-  Kokkos::fence(); // for kernel in getValues
+  Kokkos::fence();  // for kernel in getValues
 
   typedef typename Teuchos::ScalarTraits<SC>::halfPrecision SClo;
   typedef typename Teuchos::ScalarTraits<SClo>::magnitudeType MT;
   MT effective_zero = Teuchos::ScalarTraits<MT>::eps();
 
   // Allocate P
-  P = rcp(new CrsMatrixWrap(hi_map,lo_colMap,numFieldsHi)); //FIXLATER: Need faster fill
-  RCP<CrsMatrix> Pcrs   = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  P                   = rcp(new CrsMatrixWrap(hi_map, lo_colMap, numFieldsHi));  // FIXLATER: Need faster fill
+  RCP<CrsMatrix> Pcrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
 
   // Slow-ish fill
-  size_t Nelem=hi_elemToNode.extent(0);
-  std::vector<bool> touched(hi_map->getLocalNumElements(),false);
+  size_t Nelem = hi_elemToNode.extent(0);
+  std::vector<bool> touched(hi_map->getLocalNumElements(), false);
   Teuchos::Array<GO> col_gid(1);
   Teuchos::Array<SC> val(1);
   auto hi_elemToNode_host = Kokkos::create_mirror_view(hi_elemToNode);
   Kokkos::deep_copy(hi_elemToNode_host, hi_elemToNode);
-  for(size_t i=0; i<Nelem; i++) {
-    for(size_t j=0; j<numFieldsHi; j++) {
-      LO row_lid = hi_elemToNode_host(i,j);
+  for (size_t i = 0; i < Nelem; i++) {
+    for (size_t j = 0; j < numFieldsHi; j++) {
+      LO row_lid = hi_elemToNode_host(i, j);
       GO row_gid = hi_map->getGlobalElement(row_lid);
-      if(hi_nodeIsOwned[row_lid] && !touched[row_lid]) {
-        for(size_t k=0; k<numFieldsLo; k++) {
+      if (hi_nodeIsOwned[row_lid] && !touched[row_lid]) {
+        for (size_t k = 0; k < numFieldsLo; k++) {
           // Get the local id in P1's column map
-          LO col_lid = hi_to_lo_map[hi_elemToNode_host(i,lo_node_in_hi[k])];
-          if(col_lid==LOINVALID) continue;
+          LO col_lid = hi_to_lo_map[hi_elemToNode_host(i, lo_node_in_hi[k])];
+          if (col_lid == LOINVALID) continue;
 
           col_gid[0] = {lo_colMap->getGlobalElement(col_lid)};
-          val[0]     = LoValues_at_HiDofs_host(k,j);
+          val[0]     = LoValues_at_HiDofs_host(k, j);
 
           // Skip near-zeros
-          if(Teuchos::ScalarTraits<SC>::magnitude(val[0]) >= effective_zero)
-            P->insertGlobalValues(row_gid,col_gid(),val());
+          if (Teuchos::ScalarTraits<SC>::magnitude(val[0]) >= effective_zero)
+            P->insertGlobalValues(row_gid, col_gid(), val());
         }
-        touched[row_lid]=true;
+        touched[row_lid] = true;
       }
     }
   }
-  P->fillComplete(lo_domainMap,hi_map);
+  P->fillComplete(lo_domainMap, hi_map);
 }
 
 /*********************************************************************************************************/
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GenerateLinearCoarsening_pn_kirby_to_pm(const LOFieldContainer & hi_elemToNode,
-                                                                                                                 const std::vector<bool> & hi_nodeIsOwned,
-                                                                                                                 const SCFieldContainer & hi_DofCoords,
-                                                                                                                 const LOFieldContainer & lo_elemToHiRepresentativeNode,
+void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GenerateLinearCoarsening_pn_kirby_to_pm(const LOFieldContainer &hi_elemToNode,
+                                                                                                                 const std::vector<bool> &hi_nodeIsOwned,
+                                                                                                                 const SCFieldContainer &hi_DofCoords,
+                                                                                                                 const LOFieldContainer &lo_elemToHiRepresentativeNode,
                                                                                                                  const Basis &lo_basis,
-                                                                                                                 const std::vector<LocalOrdinal> & hi_to_lo_map,
-                                                                                                                 const Teuchos::RCP<const Map> & lo_colMap,
-                                                                                                                 const Teuchos::RCP<const Map> & lo_domainMap,
-                                                                                                                 const Teuchos::RCP<const Map> & hi_map,
-                                                                                                                 Teuchos::RCP<Matrix>& P) const{
+                                                                                                                 const std::vector<LocalOrdinal> &hi_to_lo_map,
+                                                                                                                 const Teuchos::RCP<const Map> &lo_colMap,
+                                                                                                                 const Teuchos::RCP<const Map> &lo_domainMap,
+                                                                                                                 const Teuchos::RCP<const Map> &hi_map,
+                                                                                                                 Teuchos::RCP<Matrix> &P) const {
   typedef SCFieldContainer FC;
   // Evaluate the linear basis functions at the Pn nodes
   size_t numFieldsHi = hi_elemToNode.extent(1);
   size_t numFieldsLo = lo_basis.getCardinality();
-  FC LoValues_at_HiDofs("LoValues_at_HiDofs",numFieldsLo,numFieldsHi);
+  FC LoValues_at_HiDofs("LoValues_at_HiDofs", numFieldsLo, numFieldsHi);
   lo_basis.getValues(LoValues_at_HiDofs, hi_DofCoords, Intrepid2::OPERATOR_VALUE);
-  auto LoValues_at_HiDofs_host = Kokkos::create_mirror_view(LoValues_at_HiDofs);
-  auto hi_elemToNode_host = Kokkos::create_mirror_view(hi_elemToNode);
+  auto LoValues_at_HiDofs_host            = Kokkos::create_mirror_view(LoValues_at_HiDofs);
+  auto hi_elemToNode_host                 = Kokkos::create_mirror_view(hi_elemToNode);
   auto lo_elemToHiRepresentativeNode_host = Kokkos::create_mirror_view(lo_elemToHiRepresentativeNode);
   Kokkos::deep_copy(LoValues_at_HiDofs_host, LoValues_at_HiDofs);
   Kokkos::deep_copy(hi_elemToNode_host, hi_elemToNode);
   Kokkos::deep_copy(lo_elemToHiRepresentativeNode_host, lo_elemToHiRepresentativeNode);
-  Kokkos::fence(); // for kernel in getValues
+  Kokkos::fence();  // for kernel in getValues
 
   typedef typename Teuchos::ScalarTraits<SC>::halfPrecision SClo;
   typedef typename Teuchos::ScalarTraits<SClo>::magnitudeType MT;
   MT effective_zero = Teuchos::ScalarTraits<MT>::eps();
 
   // Allocate P
-  P = rcp(new CrsMatrixWrap(hi_map,lo_colMap,numFieldsHi)); //FIXLATER: Need faster fill
-  RCP<CrsMatrix> Pcrs   = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  P                   = rcp(new CrsMatrixWrap(hi_map, lo_colMap, numFieldsHi));  // FIXLATER: Need faster fill
+  RCP<CrsMatrix> Pcrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
 
   // Slow-ish fill
-  size_t Nelem=hi_elemToNode.extent(0);
-  std::vector<bool> touched(hi_map->getLocalNumElements(),false);
+  size_t Nelem = hi_elemToNode.extent(0);
+  std::vector<bool> touched(hi_map->getLocalNumElements(), false);
   Teuchos::Array<GO> col_gid(1);
   Teuchos::Array<SC> val(1);
-  for(size_t i=0; i<Nelem; i++) {
-    for(size_t j=0; j<numFieldsHi; j++) {
-      LO row_lid = hi_elemToNode_host(i,j);
+  for (size_t i = 0; i < Nelem; i++) {
+    for (size_t j = 0; j < numFieldsHi; j++) {
+      LO row_lid = hi_elemToNode_host(i, j);
       GO row_gid = hi_map->getGlobalElement(row_lid);
-      if(hi_nodeIsOwned[row_lid] && !touched[row_lid]) {
-        for(size_t k=0; k<numFieldsLo; k++) {
+      if (hi_nodeIsOwned[row_lid] && !touched[row_lid]) {
+        for (size_t k = 0; k < numFieldsLo; k++) {
           // Get the local id in P1's column map
-          LO col_lid = hi_to_lo_map[lo_elemToHiRepresentativeNode_host(i,k)];
+          LO col_lid = hi_to_lo_map[lo_elemToHiRepresentativeNode_host(i, k)];
           col_gid[0] = {lo_colMap->getGlobalElement(col_lid)};
-          val[0]     = LoValues_at_HiDofs_host(k,j);
+          val[0]     = LoValues_at_HiDofs_host(k, j);
 
           // Skip near-zeros
-          if(Teuchos::ScalarTraits<SC>::magnitude(val[0]) >= effective_zero)
-            P->insertGlobalValues(row_gid,col_gid(),val());
+          if (Teuchos::ScalarTraits<SC>::magnitude(val[0]) >= effective_zero)
+            P->insertGlobalValues(row_gid, col_gid(), val());
         }
-        touched[row_lid]=true;
+        touched[row_lid] = true;
       }
     }
   }
-  P->fillComplete(lo_domainMap,hi_map);
+  P->fillComplete(lo_domainMap, hi_map);
 }
 
 /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("pcoarsen: hi basis");
-    SET_VALID_ENTRY("pcoarsen: lo basis");
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("pcoarsen: hi basis");
+  SET_VALID_ENTRY("pcoarsen: lo basis");
+#undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
 
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",      Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("pcoarsen: element to node map",          Teuchos::null, "Generating factory of the element to node map");
-    return validParamList;
-  }
+  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase>>("pcoarsen: element to node map", Teuchos::null, "Generating factory of the element to node map");
+  return validParamList;
+}
 
 /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &/* coarseLevel */) const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "pcoarsen: element to node map");
-    Input(fineLevel, "Nullspace");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level & /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "pcoarsen: element to node map");
+  Input(fineLevel, "Nullspace");
+}
 
 /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level &coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
 /*********************************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "P Coarsening", coarseLevel);
-    std::string levelIDs = toString(coarseLevel.GetLevelID());
-    const std::string prefix = "MueLu::IntrepidPCoarsenFactory(" + levelIDs + "): ";
-
-    // NOTE: This is hardwired to double on purpose.  See the note below.
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-    typedef Kokkos::DynRankView<double,typename Node::device_type> FC;
-
-    // Level Get
-    RCP<Matrix> A     = Get< RCP<Matrix> >(fineLevel, "A");
-    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> >(fineLevel, "Nullspace");
-    Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Acrs = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*A);
-
-
-    if (restrictionMode_) {
-      SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
-      A = Utilities::Transpose(*A, true); // build transpose of A explicitly
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "P Coarsening", coarseLevel);
+  std::string levelIDs     = toString(coarseLevel.GetLevelID());
+  const std::string prefix = "MueLu::IntrepidPCoarsenFactory(" + levelIDs + "): ";
+
+  // NOTE: This is hardwired to double on purpose.  See the note below.
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+  typedef Kokkos::DynRankView<double, typename Node::device_type> FC;
+
+  // Level Get
+  RCP<Matrix> A                                                          = Get<RCP<Matrix>>(fineLevel, "A");
+  RCP<MultiVector> fineNullspace                                         = Get<RCP<MultiVector>>(fineLevel, "Nullspace");
+  Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> &Acrs = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> &>(*A);
+
+  if (restrictionMode_) {
+    SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
+    A = Utilities::Transpose(*A, true);  // build transpose of A explicitly
+  }
 
-    // Find the Dirichlet rows in A
-    std::vector<LocalOrdinal> A_dirichletRows;
-    Utilities::FindDirichletRows(A,A_dirichletRows);
+  // Find the Dirichlet rows in A
+  std::vector<LocalOrdinal> A_dirichletRows;
+  Utilities::FindDirichletRows(A, A_dirichletRows);
 
-    // Build final prolongator
-    RCP<Matrix> finalP;
+  // Build final prolongator
+  RCP<Matrix> finalP;
 
-    // Reuse pattern if available
-    RCP<ParameterList> APparams = rcp(new ParameterList);
-    if (coarseLevel.IsAvailable("AP reuse data", this)) {
-      GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
+  // Reuse pattern if available
+  RCP<ParameterList> APparams = rcp(new ParameterList);
+  if (coarseLevel.IsAvailable("AP reuse data", this)) {
+    GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
 
-      APparams = coarseLevel.Get< RCP<ParameterList> >("AP reuse data", this);
+    APparams = coarseLevel.Get<RCP<ParameterList>>("AP reuse data", this);
 
-      if (APparams->isParameter("graph"))
-        finalP = APparams->get< RCP<Matrix> >("graph");
-    }
-    const ParameterList& pL = GetParameterList();
-
-    /*******************/
-    // FIXME LATER: Allow these to be manually specified instead of Intrepid
-    // Get the Intrepid bases
-    // NOTE: To make sure Stokhos works we only instantiate these guys with double.  There's a lot
-    // of stuff in the guts of Intrepid2 that doesn't play well with Stokhos as of yet.
-    int lo_degree, hi_degree;
-    RCP<Basis> hi_basis = MueLuIntrepid::BasisFactory<double,typename Node::device_type::execution_space>(pL.get<std::string>("pcoarsen: hi basis"),hi_degree);
-    RCP<Basis> lo_basis = MueLuIntrepid::BasisFactory<double,typename Node::device_type::execution_space>(pL.get<std::string>("pcoarsen: lo basis"),lo_degree);
-
-    // Useful Output
-    GetOStream(Statistics1) << "P-Coarsening from basis "<<pL.get<std::string>("pcoarsen: hi basis")<<" to "<<pL.get<std::string>("pcoarsen: lo basis") <<std::endl;
-
-    /*******************/
-    // Get the higher-order element-to-node map
-    const Teuchos::RCP<FCi> Pn_elemToNode = Get<Teuchos::RCP<FCi> >(fineLevel,"pcoarsen: element to node map");
-
-    // Calculate node ownership (the quick and dirty way)
-    // NOTE: This exploits two things:
-    //  1) domainMap == rowMap
-    //  2) Standard [e|t]petra ordering (namely the local unknowns are always numbered first).
-    // This routine does not work in general.
-    RCP<const Map> rowMap = A->getRowMap();
-    RCP<const Map> colMap = Acrs.getColMap();
-    RCP<const Map> domainMap = A->getDomainMap();
-    int NumProc = rowMap->getComm()->getSize();
-    assert(rowMap->isSameAs(*domainMap));
-    std::vector<bool> Pn_nodeIsOwned(colMap->getLocalNumElements(),false);
-    LO num_owned_rows = 0;
-    for(size_t i=0; i<rowMap->getLocalNumElements(); i++) {
-      if(rowMap->getGlobalElement(i) == colMap->getGlobalElement(i)) {
-        Pn_nodeIsOwned[i] = true;
-        num_owned_rows++;
-      }
+    if (APparams->isParameter("graph"))
+      finalP = APparams->get<RCP<Matrix>>("graph");
+  }
+  const ParameterList &pL = GetParameterList();
+
+  /*******************/
+  // FIXME LATER: Allow these to be manually specified instead of Intrepid
+  // Get the Intrepid bases
+  // NOTE: To make sure Stokhos works we only instantiate these guys with double.  There's a lot
+  // of stuff in the guts of Intrepid2 that doesn't play well with Stokhos as of yet.
+  int lo_degree, hi_degree;
+  RCP<Basis> hi_basis = MueLuIntrepid::BasisFactory<double, typename Node::device_type::execution_space>(pL.get<std::string>("pcoarsen: hi basis"), hi_degree);
+  RCP<Basis> lo_basis = MueLuIntrepid::BasisFactory<double, typename Node::device_type::execution_space>(pL.get<std::string>("pcoarsen: lo basis"), lo_degree);
+
+  // Useful Output
+  GetOStream(Statistics1) << "P-Coarsening from basis " << pL.get<std::string>("pcoarsen: hi basis") << " to " << pL.get<std::string>("pcoarsen: lo basis") << std::endl;
+
+  /*******************/
+  // Get the higher-order element-to-node map
+  const Teuchos::RCP<FCi> Pn_elemToNode = Get<Teuchos::RCP<FCi>>(fineLevel, "pcoarsen: element to node map");
+
+  // Calculate node ownership (the quick and dirty way)
+  // NOTE: This exploits two things:
+  //  1) domainMap == rowMap
+  //  2) Standard [e|t]petra ordering (namely the local unknowns are always numbered first).
+  // This routine does not work in general.
+  RCP<const Map> rowMap    = A->getRowMap();
+  RCP<const Map> colMap    = Acrs.getColMap();
+  RCP<const Map> domainMap = A->getDomainMap();
+  int NumProc              = rowMap->getComm()->getSize();
+  assert(rowMap->isSameAs(*domainMap));
+  std::vector<bool> Pn_nodeIsOwned(colMap->getLocalNumElements(), false);
+  LO num_owned_rows = 0;
+  for (size_t i = 0; i < rowMap->getLocalNumElements(); i++) {
+    if (rowMap->getGlobalElement(i) == colMap->getGlobalElement(i)) {
+      Pn_nodeIsOwned[i] = true;
+      num_owned_rows++;
     }
+  }
 
-    // Used in all cases
-    FC hi_DofCoords;
-    Teuchos::RCP<FCi> P1_elemToNode = rcp(new FCi());
+  // Used in all cases
+  FC hi_DofCoords;
+  Teuchos::RCP<FCi> P1_elemToNode = rcp(new FCi());
 
-    std::vector<bool> P1_nodeIsOwned;
-    int P1_numOwnedNodes;
-    std::vector<LO> hi_to_lo_map;
+  std::vector<bool> P1_nodeIsOwned;
+  int P1_numOwnedNodes;
+  std::vector<LO> hi_to_lo_map;
 
-    // Degree-1 variables
-    std::vector<size_t> lo_node_in_hi;
+  // Degree-1 variables
+  std::vector<size_t> lo_node_in_hi;
 
-    // Degree-n variables
-    FCi lo_elemToHiRepresentativeNode;
+  // Degree-n variables
+  FCi lo_elemToHiRepresentativeNode;
 
-    // Get Dirichlet unknown information
-    RCP<Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> > hi_isDirichletRow, hi_isDirichletCol;
-    Utilities::FindDirichletRowsAndPropagateToCols(A,hi_isDirichletRow, hi_isDirichletCol);
+  // Get Dirichlet unknown information
+  RCP<Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node>> hi_isDirichletRow, hi_isDirichletCol;
+  Utilities::FindDirichletRowsAndPropagateToCols(A, hi_isDirichletRow, hi_isDirichletCol);
 
 #if 0
     printf("[%d] isDirichletRow = ",A->getRowMap()->getComm()->getRank());
@@ -861,93 +836,93 @@ void IntrepidPCoarsenFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Generat
     fflush(stdout);
 #endif
 
-    /*******************/
-    if(lo_degree == 1) {
-      // Get reference coordinates and the lo-to-hi injection list for the reference element
-      MueLuIntrepid::IntrepidGetP1NodeInHi(hi_basis,lo_node_in_hi,hi_DofCoords);
-
-      // Generate lower-order element-to-node map
-      MueLuIntrepid::BuildLoElemToNode(*Pn_elemToNode,Pn_nodeIsOwned,lo_node_in_hi,hi_isDirichletCol->getData(0),*P1_elemToNode,P1_nodeIsOwned,hi_to_lo_map,P1_numOwnedNodes);
-      assert(hi_to_lo_map.size() == colMap->getLocalNumElements());
-   }
-    else {
-        // Get lo-order candidates
-        double threshold = 1e-10;
-        std::vector<std::vector<size_t> > candidates;
-        Kokkos::resize(hi_DofCoords,hi_basis->getCardinality(),hi_basis->getBaseCellTopology().getDimension());
-        hi_basis->getDofCoords(hi_DofCoords);
-
-        MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis,FC>(*lo_basis,hi_DofCoords,threshold,candidates);
-
-        // Generate the representative nodes
-        MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs(candidates,*Pn_elemToNode,colMap,lo_elemToHiRepresentativeNode);
-        MueLu::MueLuIntrepid::BuildLoElemToNodeViaRepresentatives(*Pn_elemToNode,Pn_nodeIsOwned,lo_elemToHiRepresentativeNode,*P1_elemToNode,P1_nodeIsOwned,hi_to_lo_map,P1_numOwnedNodes);
+  /*******************/
+  if (lo_degree == 1) {
+    // Get reference coordinates and the lo-to-hi injection list for the reference element
+    MueLuIntrepid::IntrepidGetP1NodeInHi(hi_basis, lo_node_in_hi, hi_DofCoords);
+
+    // Generate lower-order element-to-node map
+    MueLuIntrepid::BuildLoElemToNode(*Pn_elemToNode, Pn_nodeIsOwned, lo_node_in_hi, hi_isDirichletCol->getData(0), *P1_elemToNode, P1_nodeIsOwned, hi_to_lo_map, P1_numOwnedNodes);
+    assert(hi_to_lo_map.size() == colMap->getLocalNumElements());
+  } else {
+    // Get lo-order candidates
+    double threshold = 1e-10;
+    std::vector<std::vector<size_t>> candidates;
+    Kokkos::resize(hi_DofCoords, hi_basis->getCardinality(), hi_basis->getBaseCellTopology().getDimension());
+    hi_basis->getDofCoords(hi_DofCoords);
+
+    MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis, FC>(*lo_basis, hi_DofCoords, threshold, candidates);
+
+    // Generate the representative nodes
+    MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs(candidates, *Pn_elemToNode, colMap, lo_elemToHiRepresentativeNode);
+    MueLu::MueLuIntrepid::BuildLoElemToNodeViaRepresentatives(*Pn_elemToNode, Pn_nodeIsOwned, lo_elemToHiRepresentativeNode, *P1_elemToNode, P1_nodeIsOwned, hi_to_lo_map, P1_numOwnedNodes);
+  }
+  MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel, "pcoarsen: element to node map", P1_elemToNode);
+
+  /*******************/
+  // Generate the P1_domainMap
+  // HOW: Since we know how many each proc has, we can use the non-uniform contiguous map constructor to do the work for us
+  RCP<const Map> P1_domainMap = MapFactory::Build(rowMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), P1_numOwnedNodes, rowMap->getIndexBase(), rowMap->getComm());
+  MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel, "CoarseMap", P1_domainMap);
+
+  // Generate the P1_columnMap
+  RCP<const Map> P1_colMap;
+  if (NumProc == 1)
+    P1_colMap = P1_domainMap;
+  else
+    MueLuIntrepid::GenerateColMapFromImport<LO, GO, NO>(*Acrs.getCrsGraph()->getImporter(), hi_to_lo_map, *P1_domainMap, P1_nodeIsOwned.size(), P1_colMap);
+
+  /*******************/
+  // Generate the coarsening
+  if (lo_degree == 1)
+    GenerateLinearCoarsening_pn_kirby_to_p1(*Pn_elemToNode, Pn_nodeIsOwned, hi_DofCoords, lo_node_in_hi, *lo_basis, hi_to_lo_map, P1_colMap, P1_domainMap, A->getRowMap(), finalP);
+  else
+    GenerateLinearCoarsening_pn_kirby_to_pm(*Pn_elemToNode, Pn_nodeIsOwned, hi_DofCoords, lo_elemToHiRepresentativeNode, *lo_basis, hi_to_lo_map, P1_colMap, P1_domainMap, A->getRowMap(), finalP);
+
+  /*******************/
+  // Zero out the Dirichlet rows in P
+  Utilities::ZeroDirichletRows(finalP, A_dirichletRows);
+
+  /*******************/
+  // Build the nullspace
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P1_domainMap, fineNullspace->getNumVectors());
+  finalP->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS);
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+
+  // Level Set
+  if (!restrictionMode_) {
+    // The factory is in prolongation mode
+    Set(coarseLevel, "P", finalP);
+
+    APparams->set("graph", finalP);
+    MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel, "AP reuse data", APparams);
+
+    if (IsPrint(Statistics1)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*finalP, "P", params);
+    }
+  } else {
+    // The factory is in restriction mode
+    RCP<Matrix> R;
+    {
+      SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
+      R = Utilities::Transpose(*finalP, true);
     }
-    MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel,"pcoarsen: element to node map",P1_elemToNode);
-
-    /*******************/
-    // Generate the P1_domainMap
-    // HOW: Since we know how many each proc has, we can use the non-uniform contiguous map constructor to do the work for us
-    RCP<const Map> P1_domainMap = MapFactory::Build(rowMap->lib(),Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),P1_numOwnedNodes,rowMap->getIndexBase(),rowMap->getComm());
-    MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel,"CoarseMap",P1_domainMap);
-
-    // Generate the P1_columnMap
-    RCP<const Map> P1_colMap;
-    if(NumProc==1) P1_colMap = P1_domainMap;
-    else MueLuIntrepid::GenerateColMapFromImport<LO,GO,NO>(*Acrs.getCrsGraph()->getImporter(),hi_to_lo_map,*P1_domainMap,P1_nodeIsOwned.size(),P1_colMap);
-
-    /*******************/
-    // Generate the coarsening
-    if(lo_degree == 1)
-      GenerateLinearCoarsening_pn_kirby_to_p1(*Pn_elemToNode,Pn_nodeIsOwned,hi_DofCoords,lo_node_in_hi,*lo_basis,hi_to_lo_map,P1_colMap,P1_domainMap,A->getRowMap(),finalP);
-    else
-      GenerateLinearCoarsening_pn_kirby_to_pm(*Pn_elemToNode,Pn_nodeIsOwned,hi_DofCoords,lo_elemToHiRepresentativeNode,*lo_basis,hi_to_lo_map,P1_colMap,P1_domainMap,A->getRowMap(),finalP);
-
-    /*******************/
-    // Zero out the Dirichlet rows in P
-    Utilities::ZeroDirichletRows(finalP,A_dirichletRows);
-
-    /*******************/
-    // Build the nullspace
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(P1_domainMap, fineNullspace->getNumVectors());
-    finalP->apply(*fineNullspace,*coarseNullspace,Teuchos::TRANS);
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-
-    // Level Set
-    if (!restrictionMode_) {
-      // The factory is in prolongation mode
-      Set(coarseLevel, "P",             finalP);
-
-      APparams->set("graph", finalP);
-      MUELU_LEVEL_SET_IF_REQUESTED_OR_KEPT(coarseLevel,"AP reuse data",APparams);
-
-      if (IsPrint(Statistics1)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*finalP, "P", params);
-      }
-    } else {
-      // The factory is in restriction mode
-      RCP<Matrix> R;
-      {
-        SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
-        R = Utilities::Transpose(*finalP, true);
-      }
 
-      Set(coarseLevel, "R", R);
+    Set(coarseLevel, "R", R);
 
-      if (IsPrint(Statistics2)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
-      }
+    if (IsPrint(Statistics2)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
     }
+  }
 
-  } //Build()
-
-} //namespace MueLu
+}  // Build()
 
-#endif // MUELU_IPCFACTORY_DEF_HPP
+}  // namespace MueLu
 
+#endif  // MUELU_IPCFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_decl.hpp b/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_decl.hpp
index c15b52cbe84a..c7cf8aa3b15b 100644
--- a/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_decl.hpp
@@ -62,91 +62,88 @@
 
 namespace MueLu {
 
-  /* Options defining how to pick-up the next root node in the local aggregation procedure */
-  enum MinimizationNorm {
-    ANORM = 0, /* A norm   */
-    L2NORM = 1, /* L2 norm */
-    DINVANORM  = 2 /* Dinv A norm */
-  };
-
-  /*!
-    @class PgPFactory class.
-    @brief Factory for building Petrov-Galerkin Smoothed Aggregation prolongators.
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+/* Options defining how to pick-up the next root node in the local aggregation procedure */
+enum MinimizationNorm {
+  ANORM     = 0, /* A norm   */
+  L2NORM    = 1, /* L2 norm */
+  DINVANORM = 2  /* Dinv A norm */
+};
+
+/*!
+  @class PgPFactory class.
+  @brief Factory for building Petrov-Galerkin Smoothed Aggregation prolongators.
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class PgPFactory : public PFactory {
+          class Node          = DefaultNode>
+class PgPFactory : public PFactory {
 #undef MUELU_PGPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
-      User can supply a factory for generating the tentative prolongator.
-    */
-    PgPFactory() { }
-
-    //! Destructor.
-    virtual ~PgPFactory() { }
-
-    //@}
+  /*! @brief Constructor.
+    User can supply a factory for generating the tentative prolongator.
+  */
+  PgPFactory() {}
 
-    //! @name Set methods.
-    //@{
+  //! Destructor.
+  virtual ~PgPFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //! Set minimization mode (L2NORM for cheapest, ANORM more expensive, DINVANORM = default)
-    void SetMinimizationMode(MinimizationNorm minnorm);
+  //! @name Set methods.
+  //@{
 
-    //! return minimization mode
-    MinimizationNorm GetMinimizationMode();
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Set minimization mode (L2NORM for cheapest, ANORM more expensive, DINVANORM = default)
+  void SetMinimizationMode(MinimizationNorm minnorm);
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
-    //@}
+  //! return minimization mode
+  MinimizationNorm GetMinimizationMode();
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! Input
+  //@{
 
-    /*!
-      @brief Build method.
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //@}
 
-      Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
-    */
-    void Build(Level& fineLevel, Level& coarseLevel) const;
+  //! @name Build methods.
+  //@{
 
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  /*!
+    @brief Build method.
 
-    //@}
+    Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
+  */
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    void ReUseDampingParameters(bool bReuse);
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-  private:
+  //@}
 
-    void MultiplySelfAll(const RCP<Matrix>& Op, Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& InnerProdVec) const;
+  void ReUseDampingParameters(bool bReuse);
 
-    void MultiplyAll(const RCP<Matrix>& left, const RCP<Matrix>& right, Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& InnerProdVec) const;
+ private:
+  void MultiplySelfAll(const RCP<Matrix>& Op, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& InnerProdVec) const;
 
-    void ComputeRowBasedOmega(Level& fineLevel, Level& coarseLevel, const RCP<Matrix>& A, const RCP<Matrix>& P0, const RCP<Matrix>& DinvAP0, RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & RowBasedOmega) const;
+  void MultiplyAll(const RCP<Matrix>& left, const RCP<Matrix>& right, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& InnerProdVec) const;
 
-  private:
+  void ComputeRowBasedOmega(Level& fineLevel, Level& coarseLevel, const RCP<Matrix>& A, const RCP<Matrix>& P0, const RCP<Matrix>& DinvAP0, RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& RowBasedOmega) const;
 
-    //! Factory parameters
-    std::string diagonalView_; // TODO do we need this?
-  };
+ private:
+  //! Factory parameters
+  std::string diagonalView_;  // TODO do we need this?
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PGPFACTORY_SHORT
 #endif /* MUELU_PGPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_def.hpp b/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_def.hpp
index 27d36fc0472a..f7c714ef29db 100644
--- a/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Petrov-Galerkin-SA/MueLu_PgPFactory_def.hpp
@@ -66,413 +66,408 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Tentative prolongator factory");
-    validParamList->set< MinimizationNorm >      ("Minimization norm", DINVANORM,  "Norm to be minimized");
-    validParamList->set< bool >                  ("ReUseRowBasedOmegas", false,    "Reuse omegas for prolongator for restrictor");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetMinimizationMode(MinimizationNorm minnorm) {
-    SetParameter("Minimization norm", ParameterEntry(minnorm)); // revalidate
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  MueLu::MinimizationNorm PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetMinimizationMode() {
-    const ParameterList& pL = GetParameterList();
-    return pL.get<MueLu::MinimizationNorm>("Minimization norm");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Tentative prolongator factory");
+  validParamList->set<MinimizationNorm>("Minimization norm", DINVANORM, "Norm to be minimized");
+  validParamList->set<bool>("ReUseRowBasedOmegas", false, "Reuse omegas for prolongator for restrictor");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SetMinimizationMode(MinimizationNorm minnorm) {
+  SetParameter("Minimization norm", ParameterEntry(minnorm));  // revalidate
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+MueLu::MinimizationNorm PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetMinimizationMode() {
+  const ParameterList& pL = GetParameterList();
+  return pL.get<MueLu::MinimizationNorm>("Minimization norm");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
+  // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    Input(fineLevel, "A");
-
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
-    // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    coarseLevel.DeclareInput("P", initialPFact.get(), this);
-
-    /* If PgPFactory is reusing the row based damping parameters omega for
-     * restriction, it has to request the data here.
-     * we have the following scenarios:
-     * 1) Reuse omegas:
-     * PgPFactory.DeclareInput for prolongation mode requests A and P0
-     * PgPFactory.DeclareInput for restriction mode requests A, P0 and RowBasedOmega (call triggered by GenericRFactory)
-     * PgPFactory.Build for prolongation mode calculates RowBasedOmega and stores it (as requested)
-     * PgPFactory.Build for restriction mode reuses RowBasedOmega (and Releases the data with the Get call)
-     * 2) do not reuse omegas
-     * PgPFactory.DeclareInput for prolongation mode requests A and P0
-     * PgPFactory.DeclareInput for restriction mode requests A and P0
-     * PgPFactory.Build for prolongation mode calculates RowBasedOmega for prolongation operator
-     * PgPFactory.Build for restriction mode calculates RowBasedOmega for restriction operator
-     */
-    const ParameterList & pL = GetParameterList();
-    bool bReUseRowBasedOmegas = pL.get<bool>("ReUseRowBasedOmegas");
-    if( bReUseRowBasedOmegas == true && restrictionMode_ == true ) {
-      coarseLevel.DeclareInput("RowBasedOmega", this, this); // RowBasedOmega is calculated by this PgPFactory and requested by this PgPFactory
-    }
+  coarseLevel.DeclareInput("P", initialPFact.get(), this);
+
+  /* If PgPFactory is reusing the row based damping parameters omega for
+   * restriction, it has to request the data here.
+   * we have the following scenarios:
+   * 1) Reuse omegas:
+   * PgPFactory.DeclareInput for prolongation mode requests A and P0
+   * PgPFactory.DeclareInput for restriction mode requests A, P0 and RowBasedOmega (call triggered by GenericRFactory)
+   * PgPFactory.Build for prolongation mode calculates RowBasedOmega and stores it (as requested)
+   * PgPFactory.Build for restriction mode reuses RowBasedOmega (and Releases the data with the Get call)
+   * 2) do not reuse omegas
+   * PgPFactory.DeclareInput for prolongation mode requests A and P0
+   * PgPFactory.DeclareInput for restriction mode requests A and P0
+   * PgPFactory.Build for prolongation mode calculates RowBasedOmega for prolongation operator
+   * PgPFactory.Build for restriction mode calculates RowBasedOmega for restriction operator
+   */
+  const ParameterList& pL   = GetParameterList();
+  bool bReUseRowBasedOmegas = pL.get<bool>("ReUseRowBasedOmegas");
+  if (bReUseRowBasedOmegas == true && restrictionMode_ == true) {
+    coarseLevel.DeclareInput("RowBasedOmega", this, this);  // RowBasedOmega is calculated by this PgPFactory and requested by this PgPFactory
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Prolongator smoothing (PG-AMG)", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Prolongator smoothing (PG-AMG)", coarseLevel);
 
-    // Level Get
-    RCP<Matrix> A = Get< RCP<Matrix> >(fineLevel, "A");
+  // Level Get
+  RCP<Matrix> A = Get<RCP<Matrix> >(fineLevel, "A");
 
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
-    // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    RCP<Matrix> Ptent = coarseLevel.Get< RCP<Matrix> >("P", initialPFact.get());
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
+  // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  }
+  RCP<Matrix> Ptent = coarseLevel.Get<RCP<Matrix> >("P", initialPFact.get());
 
-    /////////////////// switch from A to A^T in restriction mode (necessary as long as implicit transpose not working for Epetra)
-    if(restrictionMode_) {
-      SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
-      A = Utilities::Transpose(*A, true); // build transpose of A explicitely
-    }
+  /////////////////// switch from A to A^T in restriction mode (necessary as long as implicit transpose not working for Epetra)
+  if (restrictionMode_) {
+    SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
+    A = Utilities::Transpose(*A, true);  // build transpose of A explicitely
+  }
 
-    /////////////////// calculate D^{-1} A Ptent (needed for smoothing)
-    bool doFillComplete=true;
-    bool optimizeStorage=true;
-    RCP<Matrix> DinvAP0 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *Ptent, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
-
-    doFillComplete=true;
-    optimizeStorage=false;
-    Teuchos::ArrayRCP<Scalar> diag = Utilities::GetMatrixDiagonal_arcp(*A);
-    Utilities::MyOldScaleMatrix(*DinvAP0, diag, true, doFillComplete, optimizeStorage); //scale matrix with reciprocal of diag
-
-    /////////////////// calculate local damping factors omega
-
-    Teuchos::RCP<Vector > RowBasedOmega = Teuchos::null;
-
-    const ParameterList & pL = GetParameterList();
-    bool bReUseRowBasedOmegas = pL.get<bool>("ReUseRowBasedOmegas");
-    if(restrictionMode_ == false || bReUseRowBasedOmegas == false) {
-      // if in prolongation mode: calculate row based omegas
-      // if in restriction mode: calculate omegas only if row based omegas are not used from prolongation mode
-      ComputeRowBasedOmega(fineLevel, coarseLevel, A, Ptent, DinvAP0, RowBasedOmega);
-    } // if(bReUseRowBasedOmegas == false)
-    else  {
-      // reuse row based omegas, calculated by this factory in the run before (with restrictionMode_ == false)
-      RowBasedOmega = coarseLevel.Get<Teuchos::RCP<Vector > >("RowBasedOmega", this);
-
-      // RowBasedOmega is now based on row map of A (not transposed)
-      // for restriction we use A^T instead of A
-      // -> recommunicate row based omega
-
-      // exporter: overlapping row map to nonoverlapping domain map (target map is unique)
-      // since A is already transposed we use the RangeMap of A
-      Teuchos::RCP<const Export> exporter =
+  /////////////////// calculate D^{-1} A Ptent (needed for smoothing)
+  bool doFillComplete  = true;
+  bool optimizeStorage = true;
+  RCP<Matrix> DinvAP0  = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *Ptent, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
+
+  doFillComplete                 = true;
+  optimizeStorage                = false;
+  Teuchos::ArrayRCP<Scalar> diag = Utilities::GetMatrixDiagonal_arcp(*A);
+  Utilities::MyOldScaleMatrix(*DinvAP0, diag, true, doFillComplete, optimizeStorage);  // scale matrix with reciprocal of diag
+
+  /////////////////// calculate local damping factors omega
+
+  Teuchos::RCP<Vector> RowBasedOmega = Teuchos::null;
+
+  const ParameterList& pL   = GetParameterList();
+  bool bReUseRowBasedOmegas = pL.get<bool>("ReUseRowBasedOmegas");
+  if (restrictionMode_ == false || bReUseRowBasedOmegas == false) {
+    // if in prolongation mode: calculate row based omegas
+    // if in restriction mode: calculate omegas only if row based omegas are not used from prolongation mode
+    ComputeRowBasedOmega(fineLevel, coarseLevel, A, Ptent, DinvAP0, RowBasedOmega);
+  }  // if(bReUseRowBasedOmegas == false)
+  else {
+    // reuse row based omegas, calculated by this factory in the run before (with restrictionMode_ == false)
+    RowBasedOmega = coarseLevel.Get<Teuchos::RCP<Vector> >("RowBasedOmega", this);
+
+    // RowBasedOmega is now based on row map of A (not transposed)
+    // for restriction we use A^T instead of A
+    // -> recommunicate row based omega
+
+    // exporter: overlapping row map to nonoverlapping domain map (target map is unique)
+    // since A is already transposed we use the RangeMap of A
+    Teuchos::RCP<const Export> exporter =
         ExportFactory::Build(RowBasedOmega->getMap(), A->getRangeMap());
 
-      Teuchos::RCP<Vector > noRowBasedOmega =
+    Teuchos::RCP<Vector> noRowBasedOmega =
         VectorFactory::Build(A->getRangeMap());
 
-      noRowBasedOmega->doExport(*RowBasedOmega, *exporter, Xpetra::INSERT);
+    noRowBasedOmega->doExport(*RowBasedOmega, *exporter, Xpetra::INSERT);
 
-      // importer: nonoverlapping map to overlapping map
+    // importer: nonoverlapping map to overlapping map
 
-      // importer: source -> target maps
-      Teuchos::RCP<const Import > importer =
+    // importer: source -> target maps
+    Teuchos::RCP<const Import> importer =
         ImportFactory::Build(A->getRangeMap(), A->getRowMap());
 
-      // doImport target->doImport(*source, importer, action)
-      RowBasedOmega->doImport(*noRowBasedOmega, *importer, Xpetra::INSERT);
-    }
+    // doImport target->doImport(*source, importer, action)
+    RowBasedOmega->doImport(*noRowBasedOmega, *importer, Xpetra::INSERT);
+  }
 
-    Teuchos::ArrayRCP< Scalar > RowBasedOmega_local = RowBasedOmega->getDataNonConst(0);
+  Teuchos::ArrayRCP<Scalar> RowBasedOmega_local = RowBasedOmega->getDataNonConst(0);
 
-    /////////////////// prolongator smoothing using local damping parameters omega
-    RCP<Matrix> P_smoothed = Teuchos::null;
-    Utilities::MyOldScaleMatrix(*DinvAP0, RowBasedOmega_local, false, doFillComplete, optimizeStorage); //scale matrix with reciprocal of diag
+  /////////////////// prolongator smoothing using local damping parameters omega
+  RCP<Matrix> P_smoothed = Teuchos::null;
+  Utilities::MyOldScaleMatrix(*DinvAP0, RowBasedOmega_local, false, doFillComplete, optimizeStorage);  // scale matrix with reciprocal of diag
 
-    Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Ptent, false, Teuchos::ScalarTraits<Scalar>::one(),
-                         *DinvAP0, false, -Teuchos::ScalarTraits<Scalar>::one(),
-                         P_smoothed,GetOStream(Statistics2));
-    P_smoothed->fillComplete(Ptent->getDomainMap(), Ptent->getRangeMap());
+  Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*Ptent, false, Teuchos::ScalarTraits<Scalar>::one(),
+                                                                                *DinvAP0, false, -Teuchos::ScalarTraits<Scalar>::one(),
+                                                                                P_smoothed, GetOStream(Statistics2));
+  P_smoothed->fillComplete(Ptent->getDomainMap(), Ptent->getRangeMap());
 
-    //////////////////// store results in Level
+  //////////////////// store results in Level
 
-    RCP<ParameterList> params = rcp(new ParameterList());
-    params->set("printLoadBalancingInfo", true);
+  RCP<ParameterList> params = rcp(new ParameterList());
+  params->set("printLoadBalancingInfo", true);
 
-    // Level Set
-    if (!restrictionMode_) {
-      // prolongation factory is in prolongation mode
-      Set(coarseLevel, "P", P_smoothed);
+  // Level Set
+  if (!restrictionMode_) {
+    // prolongation factory is in prolongation mode
+    Set(coarseLevel, "P", P_smoothed);
 
-      // RfromPFactory used to indicate to TogglePFactory that a factory
-      // capable  or producing R can be invoked later. TogglePFactory
-      // replaces dummy value with an index into it's array of prolongators
-      // pointing to the correct prolongator factory. This is later used by
-      // RfromP_Or_TransP to invoke the prolongatorfactory in RestrictionMode
-      int dummy = 7;
-      Set(coarseLevel,"RfromPfactory", dummy);
+    // RfromPFactory used to indicate to TogglePFactory that a factory
+    // capable  or producing R can be invoked later. TogglePFactory
+    // replaces dummy value with an index into it's array of prolongators
+    // pointing to the correct prolongator factory. This is later used by
+    // RfromP_Or_TransP to invoke the prolongatorfactory in RestrictionMode
+    int dummy = 7;
+    Set(coarseLevel, "RfromPfactory", dummy);
 
-      if (IsPrint(Statistics1))
-        GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P_smoothed, "P", params);
+    if (IsPrint(Statistics1))
+      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P_smoothed, "P", params);
 
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        P_smoothed->CreateView("stridedMaps", Ptent);
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      P_smoothed->CreateView("stridedMaps", Ptent);
 
-    } else {
-      // prolongation factory is in restriction mode
-      RCP<Matrix> R = Utilities::Transpose(*P_smoothed, true);
-      Set(coarseLevel, "R", R);
+  } else {
+    // prolongation factory is in restriction mode
+    RCP<Matrix> R = Utilities::Transpose(*P_smoothed, true);
+    Set(coarseLevel, "R", R);
 
-      if (IsPrint(Statistics1))
-        GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*R, "P", params);
-
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        R->CreateView("stridedMaps", Ptent, true);
-    }
+    if (IsPrint(Statistics1))
+      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*R, "P", params);
 
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      R->CreateView("stridedMaps", Ptent, true);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeRowBasedOmega(Level& /* fineLevel */, Level& coarseLevel, const RCP<Matrix>& A, const RCP<Matrix>& P0, const RCP<Matrix>& DinvAP0, RCP<Vector>& RowBasedOmega) const {
+  FactoryMonitor m(*this, "PgPFactory::ComputeRowBasedOmega", coarseLevel);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  Scalar sZero    = Teuchos::ScalarTraits<Scalar>::zero();
+  Magnitude mZero = Teuchos::ScalarTraits<Scalar>::magnitude(sZero);
+
+  Teuchos::RCP<Vector> Numerator   = Teuchos::null;
+  Teuchos::RCP<Vector> Denominator = Teuchos::null;
+
+  const ParameterList& pL   = GetParameterList();
+  MinimizationNorm min_norm = pL.get<MinimizationNorm>("Minimization norm");
+
+  switch (min_norm) {
+    case ANORM: {
+      // MUEMAT mode (=paper)
+      // Minimize with respect to the (A)' A norm.
+      // Need to be smart here to avoid the construction of A' A
+      //
+      //                   diag( P0' (A' A) D^{-1} A P0)
+      //   omega =   ------------------------------------------
+      //             diag( P0' A' D^{-1}' ( A'  A) D^{-1} A P0)
+      //
+      // expensive, since we have to recalculate AP0 due to the lack of an explicit scaling routine for DinvAP0
+
+      // calculate A * P0
+      bool doFillComplete  = true;
+      bool optimizeStorage = false;
+      RCP<Matrix> AP0      = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
+
+      // compute A * D^{-1} * A * P0
+      RCP<Matrix> ADinvAP0 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *DinvAP0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
+
+      Numerator   = VectorFactory::Build(ADinvAP0->getColMap(), true);
+      Denominator = VectorFactory::Build(ADinvAP0->getColMap(), true);
+      MultiplyAll(AP0, ADinvAP0, Numerator);
+      MultiplySelfAll(ADinvAP0, Denominator);
+    } break;
+    case L2NORM: {
+      // ML mode 1 (cheapest)
+      // Minimize with respect to L2 norm
+      //                  diag( P0' D^{-1} A P0)
+      //   omega =   -----------------------------
+      //             diag( P0' A' D^{-1}' D^{-1} A P0)
+      //
+      Numerator   = VectorFactory::Build(DinvAP0->getColMap(), true);
+      Denominator = VectorFactory::Build(DinvAP0->getColMap(), true);
+      MultiplyAll(P0, DinvAP0, Numerator);
+      MultiplySelfAll(DinvAP0, Denominator);
+    } break;
+    case DINVANORM: {
+      // ML mode 2
+      // Minimize with respect to the (D^{-1} A)' D^{-1} A norm.
+      // Need to be smart here to avoid the construction of A' A
+      //
+      //                   diag( P0' ( A' D^{-1}' D^{-1} A) D^{-1} A P0)
+      //   omega =   --------------------------------------------------------
+      //             diag( P0' A' D^{-1}' ( A' D^{-1}' D^{-1} A) D^{-1} A P0)
+      //
+
+      // compute D^{-1} * A * D^{-1} * A * P0
+      bool doFillComplete             = true;
+      bool optimizeStorage            = true;
+      Teuchos::ArrayRCP<Scalar> diagA = Utilities::GetMatrixDiagonal_arcp(*A);
+      RCP<Matrix> DinvADinvAP0        = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *DinvAP0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
+      Utilities::MyOldScaleMatrix(*DinvADinvAP0, diagA, true, doFillComplete, optimizeStorage);  // scale matrix with reciprocal of diag
+      diagA = Teuchos::ArrayRCP<Scalar>();
+
+      Numerator   = VectorFactory::Build(DinvADinvAP0->getColMap(), true);
+      Denominator = VectorFactory::Build(DinvADinvAP0->getColMap(), true);
+      MultiplyAll(DinvAP0, DinvADinvAP0, Numerator);
+      MultiplySelfAll(DinvADinvAP0, Denominator);
+    } break;
+    default:
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::Build: minimization mode not supported. error");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ComputeRowBasedOmega(Level& /* fineLevel */, Level &coarseLevel, const RCP<Matrix>& A, const RCP<Matrix>& P0, const RCP<Matrix>& DinvAP0, RCP<Vector > & RowBasedOmega) const {
-    FactoryMonitor m(*this, "PgPFactory::ComputeRowBasedOmega", coarseLevel);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
-    Scalar sZero = Teuchos::ScalarTraits<Scalar>::zero();
-    Magnitude mZero = Teuchos::ScalarTraits<Scalar>::magnitude(sZero);
-
-    Teuchos::RCP<Vector > Numerator = Teuchos::null;
-    Teuchos::RCP<Vector > Denominator = Teuchos::null;
-
-    const ParameterList & pL = GetParameterList();
-    MinimizationNorm min_norm = pL.get<MinimizationNorm>("Minimization norm");
-
-    switch (min_norm)
-      {
-      case ANORM: {
-        // MUEMAT mode (=paper)
-        // Minimize with respect to the (A)' A norm.
-        // Need to be smart here to avoid the construction of A' A
-        //
-        //                   diag( P0' (A' A) D^{-1} A P0)
-        //   omega =   ------------------------------------------
-        //             diag( P0' A' D^{-1}' ( A'  A) D^{-1} A P0)
-        //
-        // expensive, since we have to recalculate AP0 due to the lack of an explicit scaling routine for DinvAP0
-
-        // calculate A * P0
-        bool doFillComplete=true;
-        bool optimizeStorage=false;
-        RCP<Matrix> AP0 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
-
-        // compute A * D^{-1} * A * P0
-        RCP<Matrix> ADinvAP0 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *DinvAP0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
-
-        Numerator =   VectorFactory::Build(ADinvAP0->getColMap(), true);
-        Denominator = VectorFactory::Build(ADinvAP0->getColMap(), true);
-        MultiplyAll(AP0, ADinvAP0, Numerator);
-        MultiplySelfAll(ADinvAP0, Denominator);
-      }
-        break;
-      case L2NORM: {
-
-        // ML mode 1 (cheapest)
-        // Minimize with respect to L2 norm
-        //                  diag( P0' D^{-1} A P0)
-        //   omega =   -----------------------------
-        //             diag( P0' A' D^{-1}' D^{-1} A P0)
-        //
-        Numerator =   VectorFactory::Build(DinvAP0->getColMap(), true);
-        Denominator = VectorFactory::Build(DinvAP0->getColMap(), true);
-        MultiplyAll(P0, DinvAP0, Numerator);
-        MultiplySelfAll(DinvAP0, Denominator);
-      }
-        break;
-      case DINVANORM: {
-        // ML mode 2
-        // Minimize with respect to the (D^{-1} A)' D^{-1} A norm.
-        // Need to be smart here to avoid the construction of A' A
-        //
-        //                   diag( P0' ( A' D^{-1}' D^{-1} A) D^{-1} A P0)
-        //   omega =   --------------------------------------------------------
-        //             diag( P0' A' D^{-1}' ( A' D^{-1}' D^{-1} A) D^{-1} A P0)
-        //
-
-        // compute D^{-1} * A * D^{-1} * A * P0
-        bool doFillComplete=true;
-        bool optimizeStorage=true;
-        Teuchos::ArrayRCP<Scalar> diagA = Utilities::GetMatrixDiagonal_arcp(*A);
-        RCP<Matrix> DinvADinvAP0 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *DinvAP0, false, GetOStream(Statistics2), doFillComplete, optimizeStorage);
-        Utilities::MyOldScaleMatrix(*DinvADinvAP0, diagA, true, doFillComplete, optimizeStorage); //scale matrix with reciprocal of diag
-	diagA = Teuchos::ArrayRCP<Scalar>();
-
-        Numerator =   VectorFactory::Build(DinvADinvAP0->getColMap(), true);
-        Denominator = VectorFactory::Build(DinvADinvAP0->getColMap(), true);
-        MultiplyAll(DinvAP0, DinvADinvAP0, Numerator);
-        MultiplySelfAll(DinvADinvAP0, Denominator);
-      }
-        break;
-      default:
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::Build: minimization mode not supported. error");
-      }
-
-
-    //////////// build Column based omegas /////////////
-    Teuchos::RCP<Vector > ColBasedOmega =
-      VectorFactory::Build(Numerator->getMap()/*DinvAP0->getColMap()*/, true);
-
-    ColBasedOmega->putScalar(-666/**Teuchos::ScalarTraits<Scalar>::one()*/);
-
-    Teuchos::ArrayRCP< const Scalar > Numerator_local = Numerator->getData(0);
-    Teuchos::ArrayRCP< const Scalar > Denominator_local = Denominator->getData(0);
-    Teuchos::ArrayRCP< Scalar >       ColBasedOmega_local = ColBasedOmega->getDataNonConst(0);
-    GlobalOrdinal zero_local = 0;  // count negative colbased omegas
-    GlobalOrdinal nan_local = 0;   // count NaNs -> set them to zero
-    Magnitude min_local = 1000000.0; //Teuchos::ScalarTraits<Scalar>::one() * (Scalar) 1000000;
-    Magnitude max_local = 0.0;
-    for(LocalOrdinal i = 0; i < Teuchos::as<LocalOrdinal>(Numerator->getLocalLength()); i++) {
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(Denominator_local[i]) == mZero)
-        {
-          ColBasedOmega_local[i] = 0.0; // fallback: nonsmoothed basis function since denominator == 0.0
-          nan_local++;
-        }
-      else
-        {
-          ColBasedOmega_local[i] = Numerator_local[i] / Denominator_local[i];  // default case
-        }
-
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) < mZero) { // negative omegas are not valid. set them to zero
-        ColBasedOmega_local[i] = Teuchos::ScalarTraits<Scalar>::zero();
-        zero_local++; // count zero omegas
-      }
+  //////////// build Column based omegas /////////////
+  Teuchos::RCP<Vector> ColBasedOmega =
+      VectorFactory::Build(Numerator->getMap() /*DinvAP0->getColMap()*/, true);
+
+  ColBasedOmega->putScalar(-666 /**Teuchos::ScalarTraits<Scalar>::one()*/);
+
+  Teuchos::ArrayRCP<const Scalar> Numerator_local   = Numerator->getData(0);
+  Teuchos::ArrayRCP<const Scalar> Denominator_local = Denominator->getData(0);
+  Teuchos::ArrayRCP<Scalar> ColBasedOmega_local     = ColBasedOmega->getDataNonConst(0);
+  GlobalOrdinal zero_local                          = 0;          // count negative colbased omegas
+  GlobalOrdinal nan_local                           = 0;          // count NaNs -> set them to zero
+  Magnitude min_local                               = 1000000.0;  // Teuchos::ScalarTraits<Scalar>::one() * (Scalar) 1000000;
+  Magnitude max_local                               = 0.0;
+  for (LocalOrdinal i = 0; i < Teuchos::as<LocalOrdinal>(Numerator->getLocalLength()); i++) {
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(Denominator_local[i]) == mZero) {
+      ColBasedOmega_local[i] = 0.0;  // fallback: nonsmoothed basis function since denominator == 0.0
+      nan_local++;
+    } else {
+      ColBasedOmega_local[i] = Numerator_local[i] / Denominator_local[i];  // default case
+    }
 
-      // handle case that Nominator == Denominator -> Dirichlet bcs in A?
-      // fallback if ColBasedOmega == 1 -> very strong smoothing may lead to zero rows in P
-      // TAW: this is somewhat nonstandard and a rough fallback strategy to avoid problems
-      // also avoid "overshooting" with omega > 0.8
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) >= 0.8) {
-        ColBasedOmega_local[i] = 0.0;
-      }
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) < mZero) {  // negative omegas are not valid. set them to zero
+      ColBasedOmega_local[i] = Teuchos::ScalarTraits<Scalar>::zero();
+      zero_local++;  // count zero omegas
+    }
 
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) < min_local)
-      { min_local = Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]); }
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) > max_local)
-      { max_local = Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]); }
+    // handle case that Nominator == Denominator -> Dirichlet bcs in A?
+    // fallback if ColBasedOmega == 1 -> very strong smoothing may lead to zero rows in P
+    // TAW: this is somewhat nonstandard and a rough fallback strategy to avoid problems
+    // also avoid "overshooting" with omega > 0.8
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) >= 0.8) {
+      ColBasedOmega_local[i] = 0.0;
     }
 
-    { // be verbose
-      GlobalOrdinal zero_all;
-      GlobalOrdinal nan_all;
-      Magnitude min_all;
-      Magnitude max_all;
-      MueLu_sumAll(A->getRowMap()->getComm(), zero_local, zero_all);
-      MueLu_sumAll(A->getRowMap()->getComm(), nan_local, nan_all);
-      MueLu_minAll(A->getRowMap()->getComm(), min_local, min_all);
-      MueLu_maxAll(A->getRowMap()->getComm(), max_local, max_all);
-
-      GetOStream(MueLu::Statistics1, 0) << "PgPFactory: smoothed aggregation (scheme: ";
-      switch (min_norm) {
-        case ANORM:     GetOStream(Statistics1) << "Anorm)"     << std::endl; break;
-        case L2NORM:    GetOStream(Statistics1) << "L2norm)"    << std::endl; break;
-        case DINVANORM: GetOStream(Statistics1) << "DinvAnorm)" << std::endl; break;
-        default:        GetOStream(Statistics1) << "unknown)"   << std::endl; break;
-      }
-      GetOStream(Statistics1) << "Damping parameter: min = " << min_all << ", max = " << max_all << std::endl;
-      GetOStream(Statistics)  << "# negative omegas: " << zero_all << " out of " << ColBasedOmega->getGlobalLength() << " column-based omegas" << std::endl;
-      GetOStream(Statistics)  << "# NaNs: " << nan_all << " out of " << ColBasedOmega->getGlobalLength() << " column-based omegas" << std::endl;
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) < min_local) {
+      min_local = Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]);
+    }
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]) > max_local) {
+      max_local = Teuchos::ScalarTraits<Scalar>::magnitude(ColBasedOmega_local[i]);
     }
+  }
 
-    if(coarseLevel.IsRequested("ColBasedOmega", this)) {
-      coarseLevel.Set("ColBasedOmega", ColBasedOmega, this);
+  {  // be verbose
+    GlobalOrdinal zero_all;
+    GlobalOrdinal nan_all;
+    Magnitude min_all;
+    Magnitude max_all;
+    MueLu_sumAll(A->getRowMap()->getComm(), zero_local, zero_all);
+    MueLu_sumAll(A->getRowMap()->getComm(), nan_local, nan_all);
+    MueLu_minAll(A->getRowMap()->getComm(), min_local, min_all);
+    MueLu_maxAll(A->getRowMap()->getComm(), max_local, max_all);
+
+    GetOStream(MueLu::Statistics1, 0) << "PgPFactory: smoothed aggregation (scheme: ";
+    switch (min_norm) {
+      case ANORM: GetOStream(Statistics1) << "Anorm)" << std::endl; break;
+      case L2NORM: GetOStream(Statistics1) << "L2norm)" << std::endl; break;
+      case DINVANORM: GetOStream(Statistics1) << "DinvAnorm)" << std::endl; break;
+      default: GetOStream(Statistics1) << "unknown)" << std::endl; break;
     }
+    GetOStream(Statistics1) << "Damping parameter: min = " << min_all << ", max = " << max_all << std::endl;
+    GetOStream(Statistics) << "# negative omegas: " << zero_all << " out of " << ColBasedOmega->getGlobalLength() << " column-based omegas" << std::endl;
+    GetOStream(Statistics) << "# NaNs: " << nan_all << " out of " << ColBasedOmega->getGlobalLength() << " column-based omegas" << std::endl;
+  }
 
-    //////////// build Row based omegas /////////////
-    // transform column based omegas to row based omegas
-    RowBasedOmega =
+  if (coarseLevel.IsRequested("ColBasedOmega", this)) {
+    coarseLevel.Set("ColBasedOmega", ColBasedOmega, this);
+  }
+
+  //////////// build Row based omegas /////////////
+  // transform column based omegas to row based omegas
+  RowBasedOmega =
       VectorFactory::Build(DinvAP0->getRowMap(), true);
 
-    RowBasedOmega->putScalar(-666); // TODO bad programming style
-
-    bool bAtLeastOneDefined = false;
-    Teuchos::ArrayRCP< Scalar > RowBasedOmega_local = RowBasedOmega->getDataNonConst(0);
-    for(LocalOrdinal row = 0; row<Teuchos::as<LocalOrdinal>(A->getLocalNumRows()); row++) {
-      Teuchos::ArrayView<const LocalOrdinal> lindices;
-      Teuchos::ArrayView<const Scalar> lvals;
-      DinvAP0->getLocalRowView(row, lindices, lvals);
-      bAtLeastOneDefined = false;
-      for(size_t j=0; j<Teuchos::as<size_t>(lindices.size()); j++) {
-        Scalar omega = ColBasedOmega_local[lindices[j]];
-        if (Teuchos::ScalarTraits<Scalar>::magnitude(omega) != -666) { // TODO bad programming style
-          bAtLeastOneDefined = true;
-          if(Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row]) == -666)    RowBasedOmega_local[row] = omega;
-          else if(Teuchos::ScalarTraits<Scalar>::magnitude(omega) < Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row])) RowBasedOmega_local[row] = omega;
-        }
-      }
-      if(bAtLeastOneDefined == true) {
-        if(Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row]) < mZero)
-          RowBasedOmega_local[row] = sZero;
+  RowBasedOmega->putScalar(-666);  // TODO bad programming style
+
+  bool bAtLeastOneDefined                       = false;
+  Teuchos::ArrayRCP<Scalar> RowBasedOmega_local = RowBasedOmega->getDataNonConst(0);
+  for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(A->getLocalNumRows()); row++) {
+    Teuchos::ArrayView<const LocalOrdinal> lindices;
+    Teuchos::ArrayView<const Scalar> lvals;
+    DinvAP0->getLocalRowView(row, lindices, lvals);
+    bAtLeastOneDefined = false;
+    for (size_t j = 0; j < Teuchos::as<size_t>(lindices.size()); j++) {
+      Scalar omega = ColBasedOmega_local[lindices[j]];
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(omega) != -666) {  // TODO bad programming style
+        bAtLeastOneDefined = true;
+        if (Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row]) == -666)
+          RowBasedOmega_local[row] = omega;
+        else if (Teuchos::ScalarTraits<Scalar>::magnitude(omega) < Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row]))
+          RowBasedOmega_local[row] = omega;
       }
     }
-
-    if(coarseLevel.IsRequested("RowBasedOmega", this)) {
-      Set(coarseLevel, "RowBasedOmega", RowBasedOmega);
+    if (bAtLeastOneDefined == true) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(RowBasedOmega_local[row]) < mZero)
+        RowBasedOmega_local[row] = sZero;
     }
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiplySelfAll(const RCP<Matrix>& Op, Teuchos::RCP<Vector >& InnerProdVec) const {
+  if (coarseLevel.IsRequested("RowBasedOmega", this)) {
+    Set(coarseLevel, "RowBasedOmega", RowBasedOmega);
+  }
+}
 
-    // note: InnerProdVec is based on column map of Op
-    TEUCHOS_TEST_FOR_EXCEPTION(!InnerProdVec->getMap()->isSameAs(*Op->getColMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplySelfAll: map of InnerProdVec must be same as column map of operator. error");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiplySelfAll(const RCP<Matrix>& Op, Teuchos::RCP<Vector>& InnerProdVec) const {
+  // note: InnerProdVec is based on column map of Op
+  TEUCHOS_TEST_FOR_EXCEPTION(!InnerProdVec->getMap()->isSameAs(*Op->getColMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplySelfAll: map of InnerProdVec must be same as column map of operator. error");
 
-    Teuchos::ArrayRCP< Scalar > InnerProd_local = InnerProdVec->getDataNonConst(0);
+  Teuchos::ArrayRCP<Scalar> InnerProd_local = InnerProdVec->getDataNonConst(0);
 
-    Teuchos::ArrayView<const LocalOrdinal> lindices;
-    Teuchos::ArrayView<const Scalar> lvals;
+  Teuchos::ArrayView<const LocalOrdinal> lindices;
+  Teuchos::ArrayView<const Scalar> lvals;
 
-    for(size_t n=0; n<Op->getLocalNumRows(); n++) {
-      Op->getLocalRowView(n, lindices, lvals);
-      for(size_t i=0; i<Teuchos::as<size_t>(lindices.size()); i++) {
-        InnerProd_local[lindices[i]] += lvals[i]*lvals[i];
-      }
+  for (size_t n = 0; n < Op->getLocalNumRows(); n++) {
+    Op->getLocalRowView(n, lindices, lvals);
+    for (size_t i = 0; i < Teuchos::as<size_t>(lindices.size()); i++) {
+      InnerProd_local[lindices[i]] += lvals[i] * lvals[i];
     }
-     InnerProd_local = Teuchos::ArrayRCP< Scalar >();
+  }
+  InnerProd_local = Teuchos::ArrayRCP<Scalar>();
 
-    // exporter: overlapping map to nonoverlapping map (target map is unique)
-    Teuchos::RCP<const Export> exporter =
+  // exporter: overlapping map to nonoverlapping map (target map is unique)
+  Teuchos::RCP<const Export> exporter =
       ExportFactory::Build(Op->getColMap(), Op->getDomainMap());
 
-    Teuchos::RCP<Vector > nonoverlap =
+  Teuchos::RCP<Vector> nonoverlap =
       VectorFactory::Build(Op->getDomainMap());
 
-    nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
+  nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
 
-    // importer: nonoverlapping map to overlapping map
+  // importer: nonoverlapping map to overlapping map
 
-    // importer: source -> target maps
-    Teuchos::RCP<const Import > importer =
+  // importer: source -> target maps
+  Teuchos::RCP<const Import> importer =
       ImportFactory::Build(Op->getDomainMap(), Op->getColMap());
 
-    // doImport target->doImport(*source, importer, action)
-    InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
-
-  }
+  // doImport target->doImport(*source, importer, action)
+  InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiplyAll(const RCP<Matrix>& left, const RCP<Matrix>& right, Teuchos::RCP<Vector >& InnerProdVec) const {
-
-    TEUCHOS_TEST_FOR_EXCEPTION(!left->getDomainMap()->isSameAs(*right->getDomainMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: domain maps of left and right do not match. Error.");
-    TEUCHOS_TEST_FOR_EXCEPTION(!left->getRowMap()->isSameAs(*right->getRowMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: row maps of left and right do not match. Error.");
-#if 1 // 1=new "fast code, 0=old "slow", but safe code
-#if 0 // not necessary - remove me
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MultiplyAll(const RCP<Matrix>& left, const RCP<Matrix>& right, Teuchos::RCP<Vector>& InnerProdVec) const {
+  TEUCHOS_TEST_FOR_EXCEPTION(!left->getDomainMap()->isSameAs(*right->getDomainMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: domain maps of left and right do not match. Error.");
+  TEUCHOS_TEST_FOR_EXCEPTION(!left->getRowMap()->isSameAs(*right->getRowMap()), Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: row maps of left and right do not match. Error.");
+#if 1   // 1=new "fast code, 0=old "slow", but safe code
+#if 0   // not necessary - remove me
     if(InnerProdVec->getMap()->isSameAs(*left->getColMap())) {
       // initialize NewRightLocal vector and assign all entries to
       // left->getColMap()->getLocalNumElements() + 1
@@ -529,179 +524,167 @@ namespace MueLu {
 
 
     } else
-#endif // end remove me
-      if(InnerProdVec->getMap()->isSameAs(*right->getColMap())) {
-        size_t szNewLeftLocal = TEUCHOS_MAX(left->getColMap()->getLocalNumElements(), right->getColMap()->getLocalNumElements());
-        Teuchos::RCP<std::vector<LocalOrdinal> > NewLeftLocal = Teuchos::rcp(new std::vector<LocalOrdinal>(szNewLeftLocal, Teuchos::as<LocalOrdinal>(right->getColMap()->getMaxLocalIndex()+1)));
-
-        LocalOrdinal j = 0;
-        for (size_t i=0; i < left->getColMap()->getLocalNumElements(); i++) {
-          while ( (j < Teuchos::as<LocalOrdinal>(right->getColMap()->getLocalNumElements())) &&
-                  (right->getColMap()->getGlobalElement(j) < left->getColMap()->getGlobalElement(i)) ) j++;
-          if (right->getColMap()->getGlobalElement(j) == left->getColMap()->getGlobalElement(i)) {
-            (*NewLeftLocal)[i] = j;
-          }
-        }
-
-        /*for (size_t i=0; i < right->getColMap()->getLocalNumElements(); i++) {
-          std::cout << "left col map: " << (*NewLeftLocal)[i] << " GID: " << left->getColMap()->getGlobalElement((*NewLeftLocal)[i]) << " GID: " << right->getColMap()->getGlobalElement(i) << " right col map: " << i << std::endl;
-          }*/
-
-        Teuchos::ArrayRCP< Scalar > InnerProd_local = InnerProdVec->getDataNonConst(0);
-        Teuchos::RCP<std::vector<Scalar> > temp_array = Teuchos::rcp(new std::vector<Scalar>(right->getColMap()->getMaxLocalIndex()+2, 0.0));
-
-        for(size_t n=0; n<left->getLocalNumRows(); n++) {
-          Teuchos::ArrayView<const LocalOrdinal> lindices_left;
-          Teuchos::ArrayView<const Scalar> lvals_left;
-          Teuchos::ArrayView<const LocalOrdinal> lindices_right;
-          Teuchos::ArrayView<const Scalar> lvals_right;
-
-          left->getLocalRowView (n, lindices_left,  lvals_left);
-          right->getLocalRowView(n, lindices_right, lvals_right);
-
-          for(size_t i=0; i<Teuchos::as<size_t>(lindices_left.size()); i++) {
-            (*temp_array)[(*NewLeftLocal)[lindices_left[i] ] ] = lvals_left[i];
-          }
-          for (size_t i=0; i < Teuchos::as<size_t>(lindices_right.size()); i++) {
-            InnerProd_local[lindices_right[i]] += (*temp_array)[lindices_right[i] ] * lvals_right[i];
-          }
-          for (size_t i=0; i < Teuchos::as<size_t>(lindices_left.size()); i++) {
-            (*temp_array)[(*NewLeftLocal)[lindices_left[i] ] ] = 0.0;
-          }
-        }
-	InnerProd_local = Teuchos::ArrayRCP< Scalar >();
-        // exporter: overlapping map to nonoverlapping map (target map is unique)
-        Teuchos::RCP<const Export> exporter =
-          ExportFactory::Build(right->getColMap(), right->getDomainMap()); // TODO: change left to right?
-
-        Teuchos::RCP<Vector> nonoverlap =
-          VectorFactory::Build(right->getDomainMap()); // TODO: change left to right?
-
-        nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
-
-        // importer: nonoverlapping map to overlapping map
-
-        // importer: source -> target maps
-        Teuchos::RCP<const Import > importer =
-          ImportFactory::Build(right->getDomainMap(), right->getColMap()); // TODO: change left to right?
-        // doImport target->doImport(*source, importer, action)
-        InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: map of InnerProdVec must be same as column map of left operator? Error.");
+#endif  // end remove me
+  if (InnerProdVec->getMap()->isSameAs(*right->getColMap())) {
+    size_t szNewLeftLocal                                 = TEUCHOS_MAX(left->getColMap()->getLocalNumElements(), right->getColMap()->getLocalNumElements());
+    Teuchos::RCP<std::vector<LocalOrdinal> > NewLeftLocal = Teuchos::rcp(new std::vector<LocalOrdinal>(szNewLeftLocal, Teuchos::as<LocalOrdinal>(right->getColMap()->getMaxLocalIndex() + 1)));
+
+    LocalOrdinal j = 0;
+    for (size_t i = 0; i < left->getColMap()->getLocalNumElements(); i++) {
+      while ((j < Teuchos::as<LocalOrdinal>(right->getColMap()->getLocalNumElements())) &&
+             (right->getColMap()->getGlobalElement(j) < left->getColMap()->getGlobalElement(i))) j++;
+      if (right->getColMap()->getGlobalElement(j) == left->getColMap()->getGlobalElement(i)) {
+        (*NewLeftLocal)[i] = j;
       }
+    }
 
-#else // old "safe" code
-    if(InnerProdVec->getMap()->isSameAs(*left->getColMap())) {
+    /*for (size_t i=0; i < right->getColMap()->getLocalNumElements(); i++) {
+      std::cout << "left col map: " << (*NewLeftLocal)[i] << " GID: " << left->getColMap()->getGlobalElement((*NewLeftLocal)[i]) << " GID: " << right->getColMap()->getGlobalElement(i) << " right col map: " << i << std::endl;
+      }*/
 
-      Teuchos::ArrayRCP< Scalar > InnerProd_local = InnerProdVec->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> InnerProd_local     = InnerProdVec->getDataNonConst(0);
+    Teuchos::RCP<std::vector<Scalar> > temp_array = Teuchos::rcp(new std::vector<Scalar>(right->getColMap()->getMaxLocalIndex() + 2, 0.0));
 
-      // declare variables
+    for (size_t n = 0; n < left->getLocalNumRows(); n++) {
       Teuchos::ArrayView<const LocalOrdinal> lindices_left;
       Teuchos::ArrayView<const Scalar> lvals_left;
       Teuchos::ArrayView<const LocalOrdinal> lindices_right;
       Teuchos::ArrayView<const Scalar> lvals_right;
 
-      for(size_t n=0; n<left->getLocalNumRows(); n++)
-        {
-
-          left->getLocalRowView (n, lindices_left,  lvals_left);
-          right->getLocalRowView(n, lindices_right, lvals_right);
-
-          for(size_t i=0; i<Teuchos::as<size_t>(lindices_left.size()); i++)
-            {
-              GlobalOrdinal left_gid = left->getColMap()->getGlobalElement(lindices_left[i]);
-              for(size_t j=0; j<Teuchos::as<size_t>(lindices_right.size()); j++)
-                {
-                  GlobalOrdinal right_gid= right->getColMap()->getGlobalElement(lindices_right[j]);
-                  if(left_gid == right_gid)
-                    {
-                      InnerProd_local[lindices_left[i]] += lvals_left[i]*lvals_right[j];
-                      break; // skip remaining gids of right operator
-                    }
-                }
-            }
-        }
+      left->getLocalRowView(n, lindices_left, lvals_left);
+      right->getLocalRowView(n, lindices_right, lvals_right);
 
-      // exporter: overlapping map to nonoverlapping map (target map is unique)
-      Teuchos::RCP<const Export> exporter =
-        ExportFactory::Build(left->getColMap(), left->getDomainMap()); // TODO: change left to right?
+      for (size_t i = 0; i < Teuchos::as<size_t>(lindices_left.size()); i++) {
+        (*temp_array)[(*NewLeftLocal)[lindices_left[i]]] = lvals_left[i];
+      }
+      for (size_t i = 0; i < Teuchos::as<size_t>(lindices_right.size()); i++) {
+        InnerProd_local[lindices_right[i]] += (*temp_array)[lindices_right[i]] * lvals_right[i];
+      }
+      for (size_t i = 0; i < Teuchos::as<size_t>(lindices_left.size()); i++) {
+        (*temp_array)[(*NewLeftLocal)[lindices_left[i]]] = 0.0;
+      }
+    }
+    InnerProd_local = Teuchos::ArrayRCP<Scalar>();
+    // exporter: overlapping map to nonoverlapping map (target map is unique)
+    Teuchos::RCP<const Export> exporter =
+        ExportFactory::Build(right->getColMap(), right->getDomainMap());  // TODO: change left to right?
 
-      Teuchos::RCP<Vector > nonoverlap =
-        VectorFactory::Build(left->getDomainMap()); // TODO: change left to right?
+    Teuchos::RCP<Vector> nonoverlap =
+        VectorFactory::Build(right->getDomainMap());  // TODO: change left to right?
 
-      nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
+    nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
 
-      // importer: nonoverlapping map to overlapping map
+    // importer: nonoverlapping map to overlapping map
 
-      // importer: source -> target maps
-      Teuchos::RCP<const Import > importer =
-        ImportFactory::Build(left->getDomainMap(), left->getColMap()); // TODO: change left to right?
+    // importer: source -> target maps
+    Teuchos::RCP<const Import> importer =
+        ImportFactory::Build(right->getDomainMap(), right->getColMap());  // TODO: change left to right?
+    // doImport target->doImport(*source, importer, action)
+    InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: map of InnerProdVec must be same as column map of left operator? Error.");
+  }
 
-      // doImport target->doImport(*source, importer, action)
-      InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
+#else  // old "safe" code
+  if (InnerProdVec->getMap()->isSameAs(*left->getColMap())) {
+    Teuchos::ArrayRCP<Scalar> InnerProd_local = InnerProdVec->getDataNonConst(0);
+
+    // declare variables
+    Teuchos::ArrayView<const LocalOrdinal> lindices_left;
+    Teuchos::ArrayView<const Scalar> lvals_left;
+    Teuchos::ArrayView<const LocalOrdinal> lindices_right;
+    Teuchos::ArrayView<const Scalar> lvals_right;
+
+    for (size_t n = 0; n < left->getLocalNumRows(); n++) {
+      left->getLocalRowView(n, lindices_left, lvals_left);
+      right->getLocalRowView(n, lindices_right, lvals_right);
+
+      for (size_t i = 0; i < Teuchos::as<size_t>(lindices_left.size()); i++) {
+        GlobalOrdinal left_gid = left->getColMap()->getGlobalElement(lindices_left[i]);
+        for (size_t j = 0; j < Teuchos::as<size_t>(lindices_right.size()); j++) {
+          GlobalOrdinal right_gid = right->getColMap()->getGlobalElement(lindices_right[j]);
+          if (left_gid == right_gid) {
+            InnerProd_local[lindices_left[i]] += lvals_left[i] * lvals_right[j];
+            break;  // skip remaining gids of right operator
+          }
+        }
+      }
     }
-    else if(InnerProdVec->getMap()->isSameAs(*right->getColMap())) {
-      Teuchos::ArrayRCP< Scalar > InnerProd_local = InnerProdVec->getDataNonConst(0);
 
-      Teuchos::ArrayView<const LocalOrdinal> lindices_left;
-      Teuchos::ArrayView<const Scalar> lvals_left;
-      Teuchos::ArrayView<const LocalOrdinal> lindices_right;
-      Teuchos::ArrayView<const Scalar> lvals_right;
+    // exporter: overlapping map to nonoverlapping map (target map is unique)
+    Teuchos::RCP<const Export> exporter =
+        ExportFactory::Build(left->getColMap(), left->getDomainMap());  // TODO: change left to right?
+
+    Teuchos::RCP<Vector> nonoverlap =
+        VectorFactory::Build(left->getDomainMap());  // TODO: change left to right?
+
+    nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
+
+    // importer: nonoverlapping map to overlapping map
 
-      for(size_t n=0; n<left->getLocalNumRows(); n++)
-        {
-          left->getLocalRowView(n, lindices_left, lvals_left);
-          right->getLocalRowView(n, lindices_right, lvals_right);
-
-          for(size_t i=0; i<Teuchos::as<size_t>(lindices_left.size()); i++)
-            {
-              GlobalOrdinal left_gid = left->getColMap()->getGlobalElement(lindices_left[i]);
-              for(size_t j=0; j<Teuchos::as<size_t>(lindices_right.size()); j++)
-                {
-                  GlobalOrdinal right_gid= right->getColMap()->getGlobalElement(lindices_right[j]);
-                  if(left_gid == right_gid)
-                    {
-                      InnerProd_local[lindices_right[j]] += lvals_left[i]*lvals_right[j];
-                      break; // skip remaining gids of right operator
-                    }
-                }
-            }
+    // importer: source -> target maps
+    Teuchos::RCP<const Import> importer =
+        ImportFactory::Build(left->getDomainMap(), left->getColMap());  // TODO: change left to right?
+
+    // doImport target->doImport(*source, importer, action)
+    InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
+  } else if (InnerProdVec->getMap()->isSameAs(*right->getColMap())) {
+    Teuchos::ArrayRCP<Scalar> InnerProd_local = InnerProdVec->getDataNonConst(0);
+
+    Teuchos::ArrayView<const LocalOrdinal> lindices_left;
+    Teuchos::ArrayView<const Scalar> lvals_left;
+    Teuchos::ArrayView<const LocalOrdinal> lindices_right;
+    Teuchos::ArrayView<const Scalar> lvals_right;
+
+    for (size_t n = 0; n < left->getLocalNumRows(); n++) {
+      left->getLocalRowView(n, lindices_left, lvals_left);
+      right->getLocalRowView(n, lindices_right, lvals_right);
+
+      for (size_t i = 0; i < Teuchos::as<size_t>(lindices_left.size()); i++) {
+        GlobalOrdinal left_gid = left->getColMap()->getGlobalElement(lindices_left[i]);
+        for (size_t j = 0; j < Teuchos::as<size_t>(lindices_right.size()); j++) {
+          GlobalOrdinal right_gid = right->getColMap()->getGlobalElement(lindices_right[j]);
+          if (left_gid == right_gid) {
+            InnerProd_local[lindices_right[j]] += lvals_left[i] * lvals_right[j];
+            break;  // skip remaining gids of right operator
+          }
         }
+      }
+    }
 
-      // exporter: overlapping map to nonoverlapping map (target map is unique)
-      Teuchos::RCP<const Export> exporter =
-        ExportFactory::Build(right->getColMap(), right->getDomainMap()); // TODO: change left to right?
+    // exporter: overlapping map to nonoverlapping map (target map is unique)
+    Teuchos::RCP<const Export> exporter =
+        ExportFactory::Build(right->getColMap(), right->getDomainMap());  // TODO: change left to right?
 
-      Teuchos::RCP<Vector> nonoverlap =
-        VectorFactory::Build(right->getDomainMap()); // TODO: change left to right?
+    Teuchos::RCP<Vector> nonoverlap =
+        VectorFactory::Build(right->getDomainMap());  // TODO: change left to right?
 
-      nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
+    nonoverlap->doExport(*InnerProdVec, *exporter, Xpetra::ADD);
 
-      // importer: nonoverlapping map to overlapping map
+    // importer: nonoverlapping map to overlapping map
 
-      // importer: source -> target maps
-      Teuchos::RCP<const Import > importer =
-        ImportFactory::Build(right->getDomainMap(), right->getColMap()); // TODO: change left to right?
+    // importer: source -> target maps
+    Teuchos::RCP<const Import> importer =
+        ImportFactory::Build(right->getDomainMap(), right->getColMap());  // TODO: change left to right?
 
-      // doImport target->doImport(*source, importer, action)
-      InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
-    }
-    else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: map of InnerProdVec must be same as column map of left or right operator? Error.");
-    }
-#endif
+    // doImport target->doImport(*source, importer, action)
+    InnerProdVec->doImport(*nonoverlap, *importer, Xpetra::INSERT);
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError, "MueLu::PgPFactory::MultiplyAll: map of InnerProdVec must be same as column map of left or right operator? Error.");
   }
+#endif
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &/* fineLevel */, Level &/* coarseLevel */) const {
-    std::cout << "TODO: remove me" << std::endl;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& /* fineLevel */, Level& /* coarseLevel */) const {
+  std::cout << "TODO: remove me" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReUseDampingParameters(bool bReuse) {
-    SetParameter("ReUseRowBasedOmegas", ParameterEntry(bReuse));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PgPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReUseDampingParameters(bool bReuse) {
+  SetParameter("ReUseRowBasedOmegas", ParameterEntry(bReuse));
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_PGPFACTORY_DEF_HPP */
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_decl.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_decl.hpp
index e19830a9fdd0..a1d6a29878d4 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_decl.hpp
@@ -54,15 +54,14 @@
 
 #include "MueLu_Level_fwd.hpp"
 
-#define  VERTICAL       1
-#define  HORIZONTAL     2
-#define  GRID_SUPPLIED -1
-#define  NUM_ZPTS       0
-#define  ORIENTATION    1
+#define VERTICAL 1
+#define HORIZONTAL 2
+#define GRID_SUPPLIED -1
+#define NUM_ZPTS 0
+#define ORIENTATION 1
 
 namespace MueLu {
 
-
 /*!
   @class SemiCoarsenPFactory
   @ingroup MueLuTransferClasses
@@ -104,55 +103,55 @@ namespace MueLu {
   | NumZLayers | NoFactory | Number of z layers after coarsening. Necessary input for LineDetectionFactory. Useful input for TogglePFactory.
 
 */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class SemiCoarsenPFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class SemiCoarsenPFactory : public PFactory {
 #undef MUELU_SEMICOARSENPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor
-    SemiCoarsenPFactory() : bTransferCoordinates_(false) { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~SemiCoarsenPFactory() { }
-    //@}
+  //! Constructor
+  SemiCoarsenPFactory()
+    : bTransferCoordinates_(false) {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~SemiCoarsenPFactory() {}
+  //@}
 
-    //! Input
-    //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-  private:
-    LO FindCpts(LO const PtsPerLine, LO const CoarsenRate, LO const Thin, LO **LayerCpts) const;
-    LO MakeSemiCoarsenP(LO const Ntotal, LO const nz, LO const CoarsenRate, LO const LayerId[],
-                                  LO const VertLineId[], LO const DofsPerNode, RCP<Matrix>& Amat,
-                                  RCP<Matrix>& P, RCP<const Map>& coarseMap,
-                                  const RCP<MultiVector> fineNullspace, RCP<MultiVector>& coarseNullspace, RCP<Matrix>& R, bool buildRestriction, bool doLinear) const;
-    void RevertToPieceWiseConstant( RCP<Matrix>& P, LO BlkSize) const;
+  //@}
 
+ private:
+  LO FindCpts(LO const PtsPerLine, LO const CoarsenRate, LO const Thin, LO** LayerCpts) const;
+  LO MakeSemiCoarsenP(LO const Ntotal, LO const nz, LO const CoarsenRate, LO const LayerId[],
+                      LO const VertLineId[], LO const DofsPerNode, RCP<Matrix>& Amat,
+                      RCP<Matrix>& P, RCP<const Map>& coarseMap,
+                      const RCP<MultiVector> fineNullspace, RCP<MultiVector>& coarseNullspace, RCP<Matrix>& R, bool buildRestriction, bool doLinear) const;
+  void RevertToPieceWiseConstant(RCP<Matrix>& P, LO BlkSize) const;
 
-    mutable bool bTransferCoordinates_; //< boolean which is true if coordinate information is available to be transferred to coarse coordinate information
-  }; //class SemiCoarsenPFactory
+  mutable bool bTransferCoordinates_;  //< boolean which is true if coordinate information is available to be transferred to coarse coordinate information
+};                                     // class SemiCoarsenPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SEMICOARSENPFACTORY_SHORT
-#endif // MUELU_SEMICOARSENPFACTORY_DECL_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_def.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_def.hpp
index 833c1da23e77..37fcc5863f32 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_def.hpp
@@ -64,606 +64,629 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("semicoarsen: piecewise constant");
-    SET_VALID_ENTRY("semicoarsen: piecewise linear");
-    SET_VALID_ENTRY("semicoarsen: coarsen rate");
-    SET_VALID_ENTRY("semicoarsen: calculate nonsym restriction");
-#undef  SET_VALID_ENTRY
-    validParamList->set< RCP<const FactoryBase> >("A",               Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",       Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",     Teuchos::null, "Generating factory for coordinates");
-
-    validParamList->set< RCP<const FactoryBase> >("LineDetection_VertLineIds", Teuchos::null, "Generating factory for LineDetection vertical line ids");
-    validParamList->set< RCP<const FactoryBase> >("LineDetection_Layers",      Teuchos::null, "Generating factory for LineDetection layer ids");
-    validParamList->set< RCP<const FactoryBase> >("CoarseNumZLayers",          Teuchos::null, "Generating factory for number of coarse z-layers");
-
-    return validParamList;
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-
-    Input(fineLevel, "LineDetection_VertLineIds");
-    Input(fineLevel, "LineDetection_Layers");
-    Input(fineLevel, "CoarseNumZLayers");
-
-    // check whether fine level coordinate information is available.
-    // If yes, request the fine level coordinates and generate coarse coordinates
-    // during the Build call
-    if (fineLevel.GetLevelID() == 0) {
-      if (fineLevel.IsAvailable("Coordinates", NoFactory::get())) {
-        fineLevel.DeclareInput("Coordinates", NoFactory::get(), this);
-        bTransferCoordinates_ = true;
-      }
-    } else if (bTransferCoordinates_ == true){
-      // on coarser levels we check the default factory providing "Coordinates"
-      // or the factory declared to provide "Coordinates"
-      // first, check which factory is providing coordinate information
-      RCP<const FactoryBase> myCoordsFact = GetFactory("Coordinates");
-      if (myCoordsFact == Teuchos::null) { myCoordsFact = fineLevel.GetFactoryManager()->GetFactory("Coordinates"); }
-      if (fineLevel.IsAvailable("Coordinates", myCoordsFact.get())) {
-        fineLevel.DeclareInput("Coordinates", myCoordsFact.get(), this);
-      }
+  SET_VALID_ENTRY("semicoarsen: piecewise constant");
+  SET_VALID_ENTRY("semicoarsen: piecewise linear");
+  SET_VALID_ENTRY("semicoarsen: coarsen rate");
+  SET_VALID_ENTRY("semicoarsen: calculate nonsym restriction");
+#undef SET_VALID_ENTRY
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory for coordinates");
+
+  validParamList->set<RCP<const FactoryBase> >("LineDetection_VertLineIds", Teuchos::null, "Generating factory for LineDetection vertical line ids");
+  validParamList->set<RCP<const FactoryBase> >("LineDetection_Layers", Teuchos::null, "Generating factory for LineDetection layer ids");
+  validParamList->set<RCP<const FactoryBase> >("CoarseNumZLayers", Teuchos::null, "Generating factory for number of coarse z-layers");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+
+  Input(fineLevel, "LineDetection_VertLineIds");
+  Input(fineLevel, "LineDetection_Layers");
+  Input(fineLevel, "CoarseNumZLayers");
+
+  // check whether fine level coordinate information is available.
+  // If yes, request the fine level coordinates and generate coarse coordinates
+  // during the Build call
+  if (fineLevel.GetLevelID() == 0) {
+    if (fineLevel.IsAvailable("Coordinates", NoFactory::get())) {
+      fineLevel.DeclareInput("Coordinates", NoFactory::get(), this);
+      bTransferCoordinates_ = true;
+    }
+  } else if (bTransferCoordinates_ == true) {
+    // on coarser levels we check the default factory providing "Coordinates"
+    // or the factory declared to provide "Coordinates"
+    // first, check which factory is providing coordinate information
+    RCP<const FactoryBase> myCoordsFact = GetFactory("Coordinates");
+    if (myCoordsFact == Teuchos::null) {
+      myCoordsFact = fineLevel.GetFactoryManager()->GetFactory("Coordinates");
+    }
+    if (fineLevel.IsAvailable("Coordinates", myCoordsFact.get())) {
+      fineLevel.DeclareInput("Coordinates", myCoordsFact.get(), this);
     }
   }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    // obtain general variables
-    RCP<Matrix>      A             = Get< RCP<Matrix> >      (fineLevel, "A");
-    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
-
-    // get user-provided coarsening rate parameter (constant over all levels)
-    const ParameterList& pL = GetParameterList();
-    LO CoarsenRate = as<LO>(pL.get<int>("semicoarsen: coarsen rate"));
-    bool buildRestriction = pL.get<bool>("semicoarsen: calculate nonsym restriction");
-    bool doLinear         = pL.get<bool>("semicoarsen: piecewise linear");
-
-    // collect general input data
-    LO BlkSize            = A->GetFixedBlockSize();
-    RCP<const Map> rowMap = A->getRowMap();
-    LO Ndofs              = rowMap->getLocalNumElements();
-    LO Nnodes             = Ndofs/BlkSize;
-
-    // collect line detection information generated by the LineDetectionFactory instance
-    LO FineNumZLayers = Get< LO >(fineLevel, "CoarseNumZLayers");
-    Teuchos::ArrayRCP<LO>     TVertLineId = Get< Teuchos::ArrayRCP<LO> > (fineLevel, "LineDetection_VertLineIds");
-    Teuchos::ArrayRCP<LO>     TLayerId    = Get< Teuchos::ArrayRCP<LO> > (fineLevel, "LineDetection_Layers");
-    LO* VertLineId = TVertLineId.getRawPtr();
-    LO* LayerId    = TLayerId.getRawPtr();
-
-    // generate transfer operator with semicoarsening
-    RCP<const Map> theCoarseMap;
-    RCP<Matrix>    P, R;
-    RCP<MultiVector> coarseNullspace;
-    GO Ncoarse = MakeSemiCoarsenP(Nnodes,FineNumZLayers,CoarsenRate,LayerId,VertLineId,
-                               BlkSize, A, P, theCoarseMap, fineNullspace,coarseNullspace,R,buildRestriction,doLinear);
-
-    // set StridingInformation of P
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  // obtain general variables
+  RCP<Matrix> A                  = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<MultiVector> fineNullspace = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
+
+  // get user-provided coarsening rate parameter (constant over all levels)
+  const ParameterList& pL = GetParameterList();
+  LO CoarsenRate          = as<LO>(pL.get<int>("semicoarsen: coarsen rate"));
+  bool buildRestriction   = pL.get<bool>("semicoarsen: calculate nonsym restriction");
+  bool doLinear           = pL.get<bool>("semicoarsen: piecewise linear");
+
+  // collect general input data
+  LO BlkSize            = A->GetFixedBlockSize();
+  RCP<const Map> rowMap = A->getRowMap();
+  LO Ndofs              = rowMap->getLocalNumElements();
+  LO Nnodes             = Ndofs / BlkSize;
+
+  // collect line detection information generated by the LineDetectionFactory instance
+  LO FineNumZLayers                 = Get<LO>(fineLevel, "CoarseNumZLayers");
+  Teuchos::ArrayRCP<LO> TVertLineId = Get<Teuchos::ArrayRCP<LO> >(fineLevel, "LineDetection_VertLineIds");
+  Teuchos::ArrayRCP<LO> TLayerId    = Get<Teuchos::ArrayRCP<LO> >(fineLevel, "LineDetection_Layers");
+  LO* VertLineId                    = TVertLineId.getRawPtr();
+  LO* LayerId                       = TLayerId.getRawPtr();
+
+  // generate transfer operator with semicoarsening
+  RCP<const Map> theCoarseMap;
+  RCP<Matrix> P, R;
+  RCP<MultiVector> coarseNullspace;
+  GO Ncoarse = MakeSemiCoarsenP(Nnodes, FineNumZLayers, CoarsenRate, LayerId, VertLineId,
+                                BlkSize, A, P, theCoarseMap, fineNullspace, coarseNullspace, R, buildRestriction, doLinear);
+
+  // set StridingInformation of P
+  if (A->IsView("stridedMaps") == true)
+    P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), theCoarseMap);
+  else
+    P->CreateView("stridedMaps", P->getRangeMap(), theCoarseMap);
+
+  if (buildRestriction) {
     if (A->IsView("stridedMaps") == true)
-      P->CreateView("stridedMaps", A->getRowMap("stridedMaps"), theCoarseMap);
+      R->CreateView("stridedMaps", theCoarseMap, A->getRowMap("stridedMaps"));
     else
-      P->CreateView("stridedMaps", P->getRangeMap(), theCoarseMap);
-
-    if (buildRestriction) {
-      if (A->IsView("stridedMaps") == true)
-        R->CreateView("stridedMaps", theCoarseMap, A->getRowMap("stridedMaps"));
-      else
-        R->CreateView("stridedMaps", theCoarseMap, R->getDomainMap());
-    }
-    if  (pL.get<bool>("semicoarsen: piecewise constant")) {
-      TEUCHOS_TEST_FOR_EXCEPTION(buildRestriction, Exceptions::RuntimeError, "Cannot use calculate nonsym restriction with piecewise constant.");
-      RevertToPieceWiseConstant(P, BlkSize);
-    }
-    if  (pL.get<bool>("semicoarsen: piecewise linear")) {
-      TEUCHOS_TEST_FOR_EXCEPTION(buildRestriction, Exceptions::RuntimeError, "Cannot use calculate nonsym restriction with piecewise linear.");
-      TEUCHOS_TEST_FOR_EXCEPTION(pL.get<bool>("semicoarsen: piecewise constant"), Exceptions::RuntimeError, "Cannot use piecewise constant with piecewise linear.");
-    }
-
-    // Store number of coarse z-layers on the coarse level container
-    // This information is used by the LineDetectionAlgorithm
-    // TODO get rid of the NoFactory
-    
-    LO CoarseNumZLayers  =  FineNumZLayers*Ncoarse;
-       CoarseNumZLayers  /= Ndofs;
-    coarseLevel.Set("NumZLayers", Teuchos::as<LO>(CoarseNumZLayers), MueLu::NoFactory::get());
-
-    // store semicoarsening transfer on coarse level
-    Set(coarseLevel, "P", P);
-    if (buildRestriction) Set(coarseLevel, "RfromPfactory", R);
-
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-
-    // transfer coordinates
-    if(bTransferCoordinates_) {
-      typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO> xdMV;
-      RCP<xdMV>    fineCoords = Teuchos::null;
-      if (fineLevel.GetLevelID() == 0 &&
-          fineLevel.IsAvailable("Coordinates", NoFactory::get())) {
-        fineCoords = fineLevel.Get< RCP<xdMV> >("Coordinates", NoFactory::get());
-      } else {
-        RCP<const FactoryBase> myCoordsFact = GetFactory("Coordinates");
-        if (myCoordsFact == Teuchos::null) { myCoordsFact = fineLevel.GetFactoryManager()->GetFactory("Coordinates"); }
-        if (fineLevel.IsAvailable("Coordinates", myCoordsFact.get())) {
-          fineCoords = fineLevel.Get< RCP<xdMV> >("Coordinates", myCoordsFact.get());
-        }
-      }
+      R->CreateView("stridedMaps", theCoarseMap, R->getDomainMap());
+  }
+  if (pL.get<bool>("semicoarsen: piecewise constant")) {
+    TEUCHOS_TEST_FOR_EXCEPTION(buildRestriction, Exceptions::RuntimeError, "Cannot use calculate nonsym restriction with piecewise constant.");
+    RevertToPieceWiseConstant(P, BlkSize);
+  }
+  if (pL.get<bool>("semicoarsen: piecewise linear")) {
+    TEUCHOS_TEST_FOR_EXCEPTION(buildRestriction, Exceptions::RuntimeError, "Cannot use calculate nonsym restriction with piecewise linear.");
+    TEUCHOS_TEST_FOR_EXCEPTION(pL.get<bool>("semicoarsen: piecewise constant"), Exceptions::RuntimeError, "Cannot use piecewise constant with piecewise linear.");
+  }
 
-      TEUCHOS_TEST_FOR_EXCEPTION(fineCoords==Teuchos::null, Exceptions::RuntimeError, "No Coordinates found provided by the user.");
+  // Store number of coarse z-layers on the coarse level container
+  // This information is used by the LineDetectionAlgorithm
+  // TODO get rid of the NoFactory
 
-      TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> x = fineCoords->getDataNonConst(0);
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> y = fineCoords->getDataNonConst(1);
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> z = fineCoords->getDataNonConst(2);
+  LO CoarseNumZLayers = FineNumZLayers * Ncoarse;
+  CoarseNumZLayers /= Ndofs;
+  coarseLevel.Set("NumZLayers", Teuchos::as<LO>(CoarseNumZLayers), MueLu::NoFactory::get());
 
-      // determine the maximum and minimum z coordinate value on the current processor.
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType zval_max = -Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType zval_min =  Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
-      for ( auto it = z.begin(); it != z.end(); ++it) {
-        if(*it > zval_max) zval_max = *it;
-        if(*it < zval_min) zval_min = *it;
-      }
+  // store semicoarsening transfer on coarse level
+  Set(coarseLevel, "P", P);
+  if (buildRestriction) Set(coarseLevel, "RfromPfactory", R);
 
-      LO myCoarseZLayers = Teuchos::as<LO>(CoarseNumZLayers);
+  Set(coarseLevel, "Nullspace", coarseNullspace);
 
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> myZLayerCoords = Teuchos::arcp<typename Teuchos::ScalarTraits<Scalar>::coordinateType>(myCoarseZLayers);
-      if(myCoarseZLayers == 1) {
-        myZLayerCoords[0] = zval_min;
-      } else {
-        typename Teuchos::ScalarTraits<Scalar>::coordinateType dz = (zval_max-zval_min)/(myCoarseZLayers-1);
-        for(LO k = 0; k<myCoarseZLayers; ++k) {
-          myZLayerCoords[k] = k*dz;
-        }
+  // transfer coordinates
+  if (bTransferCoordinates_) {
+    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO> xdMV;
+    RCP<xdMV> fineCoords = Teuchos::null;
+    if (fineLevel.GetLevelID() == 0 &&
+        fineLevel.IsAvailable("Coordinates", NoFactory::get())) {
+      fineCoords = fineLevel.Get<RCP<xdMV> >("Coordinates", NoFactory::get());
+    } else {
+      RCP<const FactoryBase> myCoordsFact = GetFactory("Coordinates");
+      if (myCoordsFact == Teuchos::null) {
+        myCoordsFact = fineLevel.GetFactoryManager()->GetFactory("Coordinates");
       }
-
-      // Note, that the coarse level node coordinates have to be in vertical ordering according
-      // to the numbering of the vertical lines
-
-      // number of vertical lines on current node:
-      LO numVertLines = Nnodes / FineNumZLayers;
-      LO numLocalCoarseNodes = numVertLines * myCoarseZLayers;
-
-      RCP<const Map>   coarseCoordMap =
-          MapFactory::Build (fineCoords->getMap()->lib(),
-              Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-              Teuchos::as<size_t>(numLocalCoarseNodes),
-              fineCoords->getMap()->getIndexBase(),
-              fineCoords->getMap()->getComm());
-      RCP<xdMV> coarseCoords   = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LO,GO,NO>::Build(coarseCoordMap, fineCoords->getNumVectors());
-      coarseCoords->putScalar(-1.0);
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cx = coarseCoords->getDataNonConst(0);
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cy = coarseCoords->getDataNonConst(1);
-      ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cz = coarseCoords->getDataNonConst(2);
-
-      // loop over all vert line indices (stop as soon as possible)
-      LO cntCoarseNodes = 0;
-      for( LO vt = 0; vt < TVertLineId.size(); ++vt) {
-        //vertical line id in *vt
-        LO curVertLineId = TVertLineId[vt];
-
-        if(cx[curVertLineId * myCoarseZLayers] == -1.0 &&
-           cy[curVertLineId * myCoarseZLayers] == -1.0) {
-          // loop over all local myCoarseZLayers
-          for (LO n=0; n<myCoarseZLayers; ++n) {
-            cx[curVertLineId * myCoarseZLayers + n] = x[vt];
-            cy[curVertLineId * myCoarseZLayers + n] = y[vt];
-            cz[curVertLineId * myCoarseZLayers + n] = myZLayerCoords[n];
-          }
-          cntCoarseNodes += myCoarseZLayers;
-        }
-
-        TEUCHOS_TEST_FOR_EXCEPTION(cntCoarseNodes > numLocalCoarseNodes, Exceptions::RuntimeError, "number of coarse nodes is inconsistent.");
-        if(cntCoarseNodes == numLocalCoarseNodes) break;
+      if (fineLevel.IsAvailable("Coordinates", myCoordsFact.get())) {
+        fineCoords = fineLevel.Get<RCP<xdMV> >("Coordinates", myCoordsFact.get());
       }
-
-      // set coarse level coordinates
-      Set(coarseLevel, "Coordinates", coarseCoords);
-    } /* end bool bTransferCoordinates */
-
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalOrdinal SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindCpts(LocalOrdinal const PtsPerLine, LocalOrdinal const CoarsenRate, LocalOrdinal const Thin, LocalOrdinal **LayerCpts) const {
-    /*
-     * Given the number of points in the z direction (PtsPerLine) and a
-     * coarsening rate (CoarsenRate), determine which z-points will serve
-     * as Cpts and return the total number of Cpts.
-     *
-     * Input
-     *    PtsPerLine:   Number of fine level points in the z direction
-     *
-     *    CoarsenRate:  Roughly, number of Cpts  = (PtsPerLine+1)/CoarsenRate - 1
-     *
-     *    Thin:         Must be either 0 or 1. Thin decides what to do when
-     *                  (PtsPerLine+1)/CoarsenRate is not an integer.
-     *
-     *                    Thin == 0  ==>   ceil() the above fraction
-     *                    Thin == 1  ==>   floor() the above fraction
-     *
-     * Output
-     *    LayerCpts     Array where LayerCpts[i] indicates that the
-     *                  LayerCpts[i]th fine level layer is a Cpt Layer.
-     *                  Note: fine level layers are assumed to be numbered starting
-     *                        a one.
-     */
-    typename Teuchos::ScalarTraits<Scalar>::coordinateType temp, RestStride, di;
-    LO    NCpts, i;
-    LO    NCLayers = -1;
-    LO    FirstStride;
-
-    temp =  ((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (PtsPerLine+1))/((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (CoarsenRate)) - 1.0;
-    if  (Thin == 1) NCpts = (LO) ceil(temp);
-    else            NCpts = (LO) floor(temp);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(PtsPerLine == 1, Exceptions::RuntimeError, "SemiCoarsenPFactory::FindCpts: cannot coarsen further.");
-
-    if (NCpts < 1) NCpts = 1;
-
-    FirstStride= (LO) ceil( ((typename Teuchos::ScalarTraits<Scalar>::coordinateType) PtsPerLine+1)/( (typename Teuchos::ScalarTraits<Scalar>::coordinateType) (NCpts+1)));
-    RestStride = ((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (PtsPerLine-FirstStride+1))/((typename Teuchos::ScalarTraits<Scalar>::coordinateType) NCpts);
-
-    NCLayers   = (LO) floor((((typename Teuchos::ScalarTraits<Scalar>::coordinateType) (PtsPerLine-FirstStride+1))/RestStride)+.00001);
-
-    TEUCHOS_TEST_FOR_EXCEPTION(NCLayers != NCpts, Exceptions::RuntimeError, "sizes do not match.");
-
-    di  = (typename Teuchos::ScalarTraits<Scalar>::coordinateType) FirstStride;
-    for (i = 1; i <= NCpts; i++) {
-      (*LayerCpts)[i] = (LO) floor(di);
-      di += RestStride;
     }
 
-    return(NCLayers);
-  }
+    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords == Teuchos::null, Exceptions::RuntimeError, "No Coordinates found provided by the user.");
 
-#define MaxHorNeighborNodes 75
+    TEUCHOS_TEST_FOR_EXCEPTION(fineCoords->getNumVectors() != 3, Exceptions::RuntimeError, "Three coordinates arrays must be supplied if line detection orientation not given.");
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> x = fineCoords->getDataNonConst(0);
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> y = fineCoords->getDataNonConst(1);
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> z = fineCoords->getDataNonConst(2);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalOrdinal SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  MakeSemiCoarsenP(LO const Ntotal, LO const nz, LO const CoarsenRate, LO const LayerId[],
-                   LO const VertLineId[], LO const DofsPerNode, RCP<Matrix> & Amat, RCP<Matrix>& P,
-                   RCP<const Map>& coarseMap, const RCP<MultiVector> fineNullspace,
-                   RCP<MultiVector>& coarseNullspace, RCP<Matrix>& R, bool buildRestriction, bool doLinear) const {
-
-    /*
-     * Given a CSR matrix (OrigARowPtr, OrigAcols, OrigAvals), information
-     * describing the z-layer and vertical line (LayerId and VertLineId)
-     * of each matrix block row, a coarsening rate, and dofs/node information,
-     * construct a prolongator that coarsening to semicoarsening in the z-direction
-     * using something like an operator-dependent grid transfer. In particular,
-     * matrix stencils are collapsed to vertical lines. Thus, each vertical line
-     * gives rise to a block tridiagonal matrix. BlkRows corresponding to
-     * Cpts are replaced by identity matrices. This tridiagonal is solved
-     * to determine each interpolation basis functions. Each Blk Rhs corresponds
-     * to all zeros except at the corresponding C-pt which has an identity
-     *
-     * On termination, return the number of local prolongator columns owned by
-     * this processor.
-     *
-     * Note: This code was adapted from a matlab code where offsets/arrays
-     *       start from 1. In most parts of the code, this 1 offset is kept
-     *       (in some cases wasting the first element of the array). The
-     *       input and output matrices of this function has been changed to
-     *       have offsets/rows/columns which start from 0. LayerId[] and
-     *       VertLineId[] currently start from 1.
-     *
-     * Input
-     * =====
-     *    Ntotal       Number of fine level Blk Rows owned by this processor
-     *
-     *    nz           Number of vertical layers. Note: partitioning must be done
-     *                 so that each processor owns an entire vertical line. This
-     *                 means that nz is the global number of layers, which should
-     *                 be equal to the local number of layers.
-     *    CoarsenRate  Rate of z semicoarsening. Smoothed aggregation-like coarsening
-     *                 would correspond to CoarsenRate = 3.
-     *    LayerId      Array from 0 to Ntotal-1 + Ghost. LayerId(BlkRow) gives the
-     *                 layer number associated with the dofs within BlkRow.
-     *    VertLineId   Array from 1 to Ntotal, VertLineId(BlkRow) gives a unique
-     *                 vertical line id (from 0 to Ntotal/nz-1) of BlkRow. All
-     *                 BlkRows associated with nodes along the same vertical line
-     *                 in the mesh should have the same LineId.
-     *    DofsPerNode  Number of degrees-of-freedom per mesh node.
-     *
-     *    OrigARowPtr, CSR arrays corresponding to the fine level matrix.
-     *    OrigAcols,
-     *    OrigAvals
-     *
-     * Output
-     * =====
-     *    ParamPptr,   CSR arrays corresponding to the final prolongation matrix.
-     *    ParamPcols,
-     *    ParamsPvals
-     */
-    int    NLayers, NVertLines, MaxNnz, NCLayers, MyLine, MyLayer;
-    int    *InvLineLayer=NULL, *CptLayers=NULL, StartLayer, NStencilNodes;
-    int    BlkRow=-1, dof_j, node_k, *Sub2FullMap=NULL, RowLeng;
-    int    i, j, iii, col, count, index, loc, PtRow, PtCol;
-    SC     *BandSol=NULL, *BandMat=NULL, TheSum, *RestrictBandMat=NULL, *RestrictBandSol=NULL;
-    int    *IPIV=NULL, KL, KU, KLU, N, NRHS, LDAB,INFO;
-    int    *Pcols;
-    size_t *Pptr;
-    SC     *Pvals;
-    LO     *Rcols=NULL;
-    size_t *Rptr=NULL;
-    SC     *Rvals=NULL;
-    int    MaxStencilSize, MaxNnzPerRow;
-    LO     *LayDiff=NULL;
-    int    CurRow, LastGuy = -1, NewPtr, RLastGuy = -1;
-    int    Ndofs;
-    int    Nghost;
-    LO     *Layerdofs = NULL, *Col2Dof = NULL;
-
-    Teuchos::LAPACK<LO,SC> lapack;
-
-    char notrans[3];
-    notrans[0] = 'N';
-    notrans[1] = 'N';
-    char trans[3];
-    trans[0] = 'T';
-    trans[1] = 'T';
-
-
-    MaxNnzPerRow = MaxHorNeighborNodes*DofsPerNode*3;
-    Teuchos::ArrayRCP<LO> TLayDiff = Teuchos::arcp<LO>(1+MaxNnzPerRow); LayDiff = TLayDiff.getRawPtr();
-
-    Nghost = Amat->getColMap()->getLocalNumElements() - Amat->getDomainMap()->getLocalNumElements();
-    if (Nghost < 0) Nghost = 0;
-    Teuchos::ArrayRCP<LO> TLayerdofs= Teuchos::arcp<LO>(Ntotal*DofsPerNode+Nghost+1); Layerdofs = TLayerdofs.getRawPtr();
-    Teuchos::ArrayRCP<LO> TCol2Dof= Teuchos::arcp<LO>(Ntotal*DofsPerNode+Nghost+1); Col2Dof= TCol2Dof.getRawPtr();
-
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > localdtemp = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(Amat->getDomainMap());
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > dtemp      = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(Amat->getColMap());
-    ArrayRCP<LO> valptr= localdtemp->getDataNonConst(0);
-
-    for (i = 0; i < Ntotal*DofsPerNode; i++)
-      valptr[i]= LayerId[i/DofsPerNode];
-    valptr=ArrayRCP<LO>();
-
-    RCP< const Import> importer;
-    importer = Amat->getCrsGraph()->getImporter();
-    if (importer == Teuchos::null) {
-      importer = ImportFactory::Build(Amat->getDomainMap(), Amat->getColMap());
-    }
-    dtemp->doImport(*localdtemp, *(importer), Xpetra::INSERT);
-
-    valptr= dtemp->getDataNonConst(0);
-    for (i = 0; i < Ntotal*DofsPerNode+Nghost; i++) Layerdofs[i]= valptr[i];
-    valptr= localdtemp->getDataNonConst(0);
-    for (i = 0; i < Ntotal*DofsPerNode;        i++) valptr[i]= i%DofsPerNode;
-    valptr=ArrayRCP<LO>();
-    dtemp->doImport(*localdtemp, *(importer), Xpetra::INSERT);
-
-    valptr= dtemp->getDataNonConst(0);
-    for (i = 0; i < Ntotal*DofsPerNode+Nghost; i++) Col2Dof[i]= valptr[i];
-    valptr=ArrayRCP<LO>();
-
-    if (Ntotal != 0) {
-      NLayers   = LayerId[0];
-      NVertLines= VertLineId[0];
+    // determine the maximum and minimum z coordinate value on the current processor.
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType zval_max = -Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType zval_min = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
+    for (auto it = z.begin(); it != z.end(); ++it) {
+      if (*it > zval_max) zval_max = *it;
+      if (*it < zval_min) zval_min = *it;
     }
-    else { NLayers = -1; NVertLines = -1; }
 
-    for (i = 1; i < Ntotal; i++) {
-      if ( VertLineId[i] > NVertLines ) NVertLines = VertLineId[i];
-      if ( LayerId[i]    >   NLayers  ) NLayers    = LayerId[i];
-    }
-    NLayers++;
-    NVertLines++;
+    LO myCoarseZLayers = Teuchos::as<LO>(CoarseNumZLayers);
 
-    /*
-     * Make an inverse map so that we can quickly find the dof
-     * associated with a particular vertical line and layer.
-     */
-
-    Teuchos::ArrayRCP<LO> TInvLineLayer= Teuchos::arcp<LO>(1+NVertLines*NLayers); InvLineLayer = TInvLineLayer.getRawPtr();
-    for (i=0; i < Ntotal; i++) {
-      InvLineLayer[ VertLineId[i]+1+LayerId[i]*NVertLines ] = i;
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> myZLayerCoords = Teuchos::arcp<typename Teuchos::ScalarTraits<Scalar>::coordinateType>(myCoarseZLayers);
+    if (myCoarseZLayers == 1) {
+      myZLayerCoords[0] = zval_min;
+    } else {
+      typename Teuchos::ScalarTraits<Scalar>::coordinateType dz = (zval_max - zval_min) / (myCoarseZLayers - 1);
+      for (LO k = 0; k < myCoarseZLayers; ++k) {
+        myZLayerCoords[k] = k * dz;
+      }
     }
 
-    /*
-     * Determine coarse layers where injection will be applied.
-     */
-
-    Teuchos::ArrayRCP<LO> TCptLayers = Teuchos::arcp<LO>(nz+1); CptLayers = TCptLayers.getRawPtr();
-    NCLayers = FindCpts(nz,CoarsenRate,0, &CptLayers);
-
-    /*
-     * Compute the largest possible interpolation stencil width based
-     * on the location of the Clayers. This stencil width is actually
-     * nodal (i.e. assuming 1 dof/node). To get the true max stencil width
-     * one needs to multiply this by DofsPerNode.
-     */
-
-    if  (NCLayers < 2) MaxStencilSize = nz;
-    else MaxStencilSize = CptLayers[2];
-
-    for (i = 3; i <= NCLayers; i++) {
-      if (MaxStencilSize < CptLayers[i]- CptLayers[i-2])
-        MaxStencilSize = CptLayers[i]- CptLayers[i-2];
-    }
-    if (NCLayers > 1) {
-      if (MaxStencilSize < nz - CptLayers[NCLayers-1]+1)
-        MaxStencilSize =  nz - CptLayers[NCLayers-1]+1;
-    }
+    // Note, that the coarse level node coordinates have to be in vertical ordering according
+    // to the numbering of the vertical lines
+
+    // number of vertical lines on current node:
+    LO numVertLines        = Nnodes / FineNumZLayers;
+    LO numLocalCoarseNodes = numVertLines * myCoarseZLayers;
+
+    RCP<const Map> coarseCoordMap =
+        MapFactory::Build(fineCoords->getMap()->lib(),
+                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                          Teuchos::as<size_t>(numLocalCoarseNodes),
+                          fineCoords->getMap()->getIndexBase(),
+                          fineCoords->getMap()->getComm());
+    RCP<xdMV> coarseCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LO, GO, NO>::Build(coarseCoordMap, fineCoords->getNumVectors());
+    coarseCoords->putScalar(-1.0);
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cx = coarseCoords->getDataNonConst(0);
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cy = coarseCoords->getDataNonConst(1);
+    ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::coordinateType> cz = coarseCoords->getDataNonConst(2);
+
+    // loop over all vert line indices (stop as soon as possible)
+    LO cntCoarseNodes = 0;
+    for (LO vt = 0; vt < TVertLineId.size(); ++vt) {
+      // vertical line id in *vt
+      LO curVertLineId = TVertLineId[vt];
+
+      if (cx[curVertLineId * myCoarseZLayers] == -1.0 &&
+          cy[curVertLineId * myCoarseZLayers] == -1.0) {
+        // loop over all local myCoarseZLayers
+        for (LO n = 0; n < myCoarseZLayers; ++n) {
+          cx[curVertLineId * myCoarseZLayers + n] = x[vt];
+          cy[curVertLineId * myCoarseZLayers + n] = y[vt];
+          cz[curVertLineId * myCoarseZLayers + n] = myZLayerCoords[n];
+        }
+        cntCoarseNodes += myCoarseZLayers;
+      }
 
-    /*
-     * Allocate storage associated with solving a banded sub-matrix needed to
-     * determine the interpolation stencil. Note: we compute interpolation stencils
-     * for all dofs within a node at the same time, and so the banded solution
-     * must be large enough to hold all DofsPerNode simultaneously.
-     */
-
-    Teuchos::ArrayRCP<LO> TSub2FullMap= Teuchos::arcp<LO>((MaxStencilSize+1)*DofsPerNode); Sub2FullMap= TSub2FullMap.getRawPtr();
-    Teuchos::ArrayRCP<SC> TBandSol= Teuchos::arcp<SC>((MaxStencilSize+1)*DofsPerNode*DofsPerNode); BandSol = TBandSol.getRawPtr();
-    Teuchos::ArrayRCP<SC> TResBandSol;
-    if (buildRestriction) {
-      TResBandSol= Teuchos::arcp<SC>((MaxStencilSize+1)*DofsPerNode*DofsPerNode); RestrictBandSol = TResBandSol.getRawPtr();
-    }
-    /*
-     * Lapack variables. See comments for dgbsv().
-     */
-    KL     = 2*DofsPerNode-1;
-    KU     = 2*DofsPerNode-1;
-    KLU    = KL+KU;
-    LDAB   = 2*KL+KU+1;
-    NRHS   = DofsPerNode;
-    Teuchos::ArrayRCP<SC> TBandMat= Teuchos::arcp<SC>(LDAB*MaxStencilSize*DofsPerNode+1); BandMat = TBandMat.getRawPtr();
-    Teuchos::ArrayRCP<LO> TIPIV= Teuchos::arcp<LO>((MaxStencilSize+1)*DofsPerNode); IPIV = TIPIV.getRawPtr();
-
-    Teuchos::ArrayRCP<SC> TResBandMat;
-    if (buildRestriction) {
-     TResBandMat= Teuchos::arcp<SC>(LDAB*MaxStencilSize*DofsPerNode+1); RestrictBandMat = TResBandMat.getRawPtr();
+      TEUCHOS_TEST_FOR_EXCEPTION(cntCoarseNodes > numLocalCoarseNodes, Exceptions::RuntimeError, "number of coarse nodes is inconsistent.");
+      if (cntCoarseNodes == numLocalCoarseNodes) break;
     }
 
-    /*
-     * Allocate storage for the final interpolation matrix. Note: each prolongator
-     * row might have entries corresponding to at most two nodes.
-     * Note: the total fine level dofs equals DofsPerNode*Ntotal and the max
-     *       nnz per prolongator row is DofsPerNode*2.
-     */
-
-    Ndofs  = DofsPerNode*Ntotal;
-    MaxNnz = 2*DofsPerNode*Ndofs;
-
-    RCP<const Map> rowMap = Amat->getRowMap();
-    Xpetra::global_size_t GNdofs= rowMap->getGlobalNumElements();
-
-    std::vector<size_t> stridingInfo_;
-    stridingInfo_.push_back(DofsPerNode);
-
-   Xpetra::global_size_t     itemp = GNdofs/nz;
-    coarseMap = StridedMapFactory::Build(rowMap->lib(),
-        NCLayers*itemp,
-        NCLayers*NVertLines*DofsPerNode,
-        0, /* index base */
-        stridingInfo_,
-        rowMap->getComm(),
-        -1, /* strided block id */
-        0 /* domain gid offset */);
-
-
-    //coarseMap = MapFactory::createContigMapWithNode(rowMap->lib(),(NCLayers*GNdofs)/nz, NCLayers*NVertLines*DofsPerNode,(rowMap->getComm()), rowMap->getNode());
-
+    // set coarse level coordinates
+    Set(coarseLevel, "Coordinates", coarseCoords);
+  } /* end bool bTransferCoordinates */
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindCpts(LocalOrdinal const PtsPerLine, LocalOrdinal const CoarsenRate, LocalOrdinal const Thin, LocalOrdinal** LayerCpts) const {
+  /*
+   * Given the number of points in the z direction (PtsPerLine) and a
+   * coarsening rate (CoarsenRate), determine which z-points will serve
+   * as Cpts and return the total number of Cpts.
+   *
+   * Input
+   *    PtsPerLine:   Number of fine level points in the z direction
+   *
+   *    CoarsenRate:  Roughly, number of Cpts  = (PtsPerLine+1)/CoarsenRate - 1
+   *
+   *    Thin:         Must be either 0 or 1. Thin decides what to do when
+   *                  (PtsPerLine+1)/CoarsenRate is not an integer.
+   *
+   *                    Thin == 0  ==>   ceil() the above fraction
+   *                    Thin == 1  ==>   floor() the above fraction
+   *
+   * Output
+   *    LayerCpts     Array where LayerCpts[i] indicates that the
+   *                  LayerCpts[i]th fine level layer is a Cpt Layer.
+   *                  Note: fine level layers are assumed to be numbered starting
+   *                        a one.
+   */
+  typename Teuchos::ScalarTraits<Scalar>::coordinateType temp, RestStride, di;
+  LO NCpts, i;
+  LO NCLayers = -1;
+  LO FirstStride;
+
+  temp = ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(PtsPerLine + 1)) / ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(CoarsenRate)) - 1.0;
+  if (Thin == 1)
+    NCpts = (LO)ceil(temp);
+  else
+    NCpts = (LO)floor(temp);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(PtsPerLine == 1, Exceptions::RuntimeError, "SemiCoarsenPFactory::FindCpts: cannot coarsen further.");
+
+  if (NCpts < 1) NCpts = 1;
+
+  FirstStride = (LO)ceil(((typename Teuchos::ScalarTraits<Scalar>::coordinateType)PtsPerLine + 1) / ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(NCpts + 1)));
+  RestStride  = ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(PtsPerLine - FirstStride + 1)) / ((typename Teuchos::ScalarTraits<Scalar>::coordinateType)NCpts);
+
+  NCLayers = (LO)floor((((typename Teuchos::ScalarTraits<Scalar>::coordinateType)(PtsPerLine - FirstStride + 1)) / RestStride) + .00001);
+
+  TEUCHOS_TEST_FOR_EXCEPTION(NCLayers != NCpts, Exceptions::RuntimeError, "sizes do not match.");
+
+  di = (typename Teuchos::ScalarTraits<Scalar>::coordinateType)FirstStride;
+  for (i = 1; i <= NCpts; i++) {
+    (*LayerCpts)[i] = (LO)floor(di);
+    di += RestStride;
+  }
 
-    P       = rcp(new CrsMatrixWrap(rowMap, coarseMap , 0));
-    coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
+  return (NCLayers);
+}
 
+#define MaxHorNeighborNodes 75
 
-    Teuchos::ArrayRCP<SC> TPvals= Teuchos::arcp<SC>(1+MaxNnz);               Pvals= TPvals.getRawPtr();
-    Teuchos::ArrayRCP<size_t> TPptr = Teuchos::arcp<size_t>(DofsPerNode*(2+Ntotal)); Pptr = TPptr.getRawPtr();
-    Teuchos::ArrayRCP<LO>     TPcols= Teuchos::arcp<LO>(1+MaxNnz);                   Pcols= TPcols.getRawPtr();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    MakeSemiCoarsenP(LO const Ntotal, LO const nz, LO const CoarsenRate, LO const LayerId[],
+                     LO const VertLineId[], LO const DofsPerNode, RCP<Matrix>& Amat, RCP<Matrix>& P,
+                     RCP<const Map>& coarseMap, const RCP<MultiVector> fineNullspace,
+                     RCP<MultiVector>& coarseNullspace, RCP<Matrix>& R, bool buildRestriction, bool doLinear) const {
+  /*
+   * Given a CSR matrix (OrigARowPtr, OrigAcols, OrigAvals), information
+   * describing the z-layer and vertical line (LayerId and VertLineId)
+   * of each matrix block row, a coarsening rate, and dofs/node information,
+   * construct a prolongator that coarsening to semicoarsening in the z-direction
+   * using something like an operator-dependent grid transfer. In particular,
+   * matrix stencils are collapsed to vertical lines. Thus, each vertical line
+   * gives rise to a block tridiagonal matrix. BlkRows corresponding to
+   * Cpts are replaced by identity matrices. This tridiagonal is solved
+   * to determine each interpolation basis functions. Each Blk Rhs corresponds
+   * to all zeros except at the corresponding C-pt which has an identity
+   *
+   * On termination, return the number of local prolongator columns owned by
+   * this processor.
+   *
+   * Note: This code was adapted from a matlab code where offsets/arrays
+   *       start from 1. In most parts of the code, this 1 offset is kept
+   *       (in some cases wasting the first element of the array). The
+   *       input and output matrices of this function has been changed to
+   *       have offsets/rows/columns which start from 0. LayerId[] and
+   *       VertLineId[] currently start from 1.
+   *
+   * Input
+   * =====
+   *    Ntotal       Number of fine level Blk Rows owned by this processor
+   *
+   *    nz           Number of vertical layers. Note: partitioning must be done
+   *                 so that each processor owns an entire vertical line. This
+   *                 means that nz is the global number of layers, which should
+   *                 be equal to the local number of layers.
+   *    CoarsenRate  Rate of z semicoarsening. Smoothed aggregation-like coarsening
+   *                 would correspond to CoarsenRate = 3.
+   *    LayerId      Array from 0 to Ntotal-1 + Ghost. LayerId(BlkRow) gives the
+   *                 layer number associated with the dofs within BlkRow.
+   *    VertLineId   Array from 1 to Ntotal, VertLineId(BlkRow) gives a unique
+   *                 vertical line id (from 0 to Ntotal/nz-1) of BlkRow. All
+   *                 BlkRows associated with nodes along the same vertical line
+   *                 in the mesh should have the same LineId.
+   *    DofsPerNode  Number of degrees-of-freedom per mesh node.
+   *
+   *    OrigARowPtr, CSR arrays corresponding to the fine level matrix.
+   *    OrigAcols,
+   *    OrigAvals
+   *
+   * Output
+   * =====
+   *    ParamPptr,   CSR arrays corresponding to the final prolongation matrix.
+   *    ParamPcols,
+   *    ParamsPvals
+   */
+  int NLayers, NVertLines, MaxNnz, NCLayers, MyLine, MyLayer;
+  int *InvLineLayer = NULL, *CptLayers = NULL, StartLayer, NStencilNodes;
+  int BlkRow = -1, dof_j, node_k, *Sub2FullMap = NULL, RowLeng;
+  int i, j, iii, col, count, index, loc, PtRow, PtCol;
+  SC *BandSol = NULL, *BandMat = NULL, TheSum, *RestrictBandMat = NULL, *RestrictBandSol = NULL;
+  int *IPIV = NULL, KL, KU, KLU, N, NRHS, LDAB, INFO;
+  int* Pcols;
+  size_t* Pptr;
+  SC* Pvals;
+  LO* Rcols    = NULL;
+  size_t* Rptr = NULL;
+  SC* Rvals    = NULL;
+  int MaxStencilSize, MaxNnzPerRow;
+  LO* LayDiff = NULL;
+  int CurRow, LastGuy = -1, NewPtr, RLastGuy = -1;
+  int Ndofs;
+  int Nghost;
+  LO *Layerdofs = NULL, *Col2Dof = NULL;
+
+  Teuchos::LAPACK<LO, SC> lapack;
+
+  char notrans[3];
+  notrans[0] = 'N';
+  notrans[1] = 'N';
+  char trans[3];
+  trans[0] = 'T';
+  trans[1] = 'T';
+
+  MaxNnzPerRow                   = MaxHorNeighborNodes * DofsPerNode * 3;
+  Teuchos::ArrayRCP<LO> TLayDiff = Teuchos::arcp<LO>(1 + MaxNnzPerRow);
+  LayDiff                        = TLayDiff.getRawPtr();
+
+  Nghost = Amat->getColMap()->getLocalNumElements() - Amat->getDomainMap()->getLocalNumElements();
+  if (Nghost < 0) Nghost = 0;
+  Teuchos::ArrayRCP<LO> TLayerdofs = Teuchos::arcp<LO>(Ntotal * DofsPerNode + Nghost + 1);
+  Layerdofs                        = TLayerdofs.getRawPtr();
+  Teuchos::ArrayRCP<LO> TCol2Dof   = Teuchos::arcp<LO>(Ntotal * DofsPerNode + Nghost + 1);
+  Col2Dof                          = TCol2Dof.getRawPtr();
+
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > localdtemp = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(Amat->getDomainMap());
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > dtemp      = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(Amat->getColMap());
+  ArrayRCP<LO> valptr                             = localdtemp->getDataNonConst(0);
+
+  for (i = 0; i < Ntotal * DofsPerNode; i++)
+    valptr[i] = LayerId[i / DofsPerNode];
+  valptr = ArrayRCP<LO>();
+
+  RCP<const Import> importer;
+  importer = Amat->getCrsGraph()->getImporter();
+  if (importer == Teuchos::null) {
+    importer = ImportFactory::Build(Amat->getDomainMap(), Amat->getColMap());
+  }
+  dtemp->doImport(*localdtemp, *(importer), Xpetra::INSERT);
+
+  valptr = dtemp->getDataNonConst(0);
+  for (i = 0; i < Ntotal * DofsPerNode + Nghost; i++) Layerdofs[i] = valptr[i];
+  valptr = localdtemp->getDataNonConst(0);
+  for (i = 0; i < Ntotal * DofsPerNode; i++) valptr[i] = i % DofsPerNode;
+  valptr = ArrayRCP<LO>();
+  dtemp->doImport(*localdtemp, *(importer), Xpetra::INSERT);
+
+  valptr = dtemp->getDataNonConst(0);
+  for (i = 0; i < Ntotal * DofsPerNode + Nghost; i++) Col2Dof[i] = valptr[i];
+  valptr = ArrayRCP<LO>();
+
+  if (Ntotal != 0) {
+    NLayers    = LayerId[0];
+    NVertLines = VertLineId[0];
+  } else {
+    NLayers    = -1;
+    NVertLines = -1;
+  }
 
-    TEUCHOS_TEST_FOR_EXCEPTION(Pcols == NULL, Exceptions::RuntimeError, "MakeSemiCoarsenP: Not enough space \n");
-    Pptr[0] = 0; Pptr[1] = 0;
+  for (i = 1; i < Ntotal; i++) {
+    if (VertLineId[i] > NVertLines) NVertLines = VertLineId[i];
+    if (LayerId[i] > NLayers) NLayers = LayerId[i];
+  }
+  NLayers++;
+  NVertLines++;
+
+  /*
+   * Make an inverse map so that we can quickly find the dof
+   * associated with a particular vertical line and layer.
+   */
+
+  Teuchos::ArrayRCP<LO> TInvLineLayer = Teuchos::arcp<LO>(1 + NVertLines * NLayers);
+  InvLineLayer                        = TInvLineLayer.getRawPtr();
+  for (i = 0; i < Ntotal; i++) {
+    InvLineLayer[VertLineId[i] + 1 + LayerId[i] * NVertLines] = i;
+  }
 
+  /*
+   * Determine coarse layers where injection will be applied.
+   */
+
+  Teuchos::ArrayRCP<LO> TCptLayers = Teuchos::arcp<LO>(nz + 1);
+  CptLayers                        = TCptLayers.getRawPtr();
+  NCLayers                         = FindCpts(nz, CoarsenRate, 0, &CptLayers);
+
+  /*
+   * Compute the largest possible interpolation stencil width based
+   * on the location of the Clayers. This stencil width is actually
+   * nodal (i.e. assuming 1 dof/node). To get the true max stencil width
+   * one needs to multiply this by DofsPerNode.
+   */
+
+  if (NCLayers < 2)
+    MaxStencilSize = nz;
+  else
+    MaxStencilSize = CptLayers[2];
+
+  for (i = 3; i <= NCLayers; i++) {
+    if (MaxStencilSize < CptLayers[i] - CptLayers[i - 2])
+      MaxStencilSize = CptLayers[i] - CptLayers[i - 2];
+  }
+  if (NCLayers > 1) {
+    if (MaxStencilSize < nz - CptLayers[NCLayers - 1] + 1)
+      MaxStencilSize = nz - CptLayers[NCLayers - 1] + 1;
+  }
 
-    Teuchos::ArrayRCP<SC> TRvals;
-    Teuchos::ArrayRCP<size_t> TRptr;
-    Teuchos::ArrayRCP<LO>     TRcols;
-    LO RmaxNnz=-1, RmaxPerRow=-1;
-    if (buildRestriction) {
-      RmaxPerRow = ((LO) ceil(((double) (MaxNnz+7))/((double) (coarseMap->getLocalNumElements()))));
-      RmaxNnz = RmaxPerRow*coarseMap->getLocalNumElements();
-      TRvals= Teuchos::arcp<SC>(1+RmaxNnz);               Rvals= TRvals.getRawPtr();
-      TRptr = Teuchos::arcp<size_t>((2+coarseMap->getLocalNumElements())); Rptr = TRptr.getRawPtr();
-      TRcols= Teuchos::arcp<LO>(1+RmaxNnz);                   Rcols= TRcols.getRawPtr();
-      TEUCHOS_TEST_FOR_EXCEPTION(Rcols == NULL, Exceptions::RuntimeError, "MakeSemiCoarsenP: Not enough space \n");
-      Rptr[0] = 0; Rptr[1] = 0;
-    }
-    /*
-     * Setup P's rowptr as if each row had its maximum of 2*DofsPerNode nonzeros.
-     * This will be useful while filling up P, and then later we will squeeze out
-     * the unused nonzeros locations.
-     */
+  /*
+   * Allocate storage associated with solving a banded sub-matrix needed to
+   * determine the interpolation stencil. Note: we compute interpolation stencils
+   * for all dofs within a node at the same time, and so the banded solution
+   * must be large enough to hold all DofsPerNode simultaneously.
+   */
+
+  Teuchos::ArrayRCP<LO> TSub2FullMap = Teuchos::arcp<LO>((MaxStencilSize + 1) * DofsPerNode);
+  Sub2FullMap                        = TSub2FullMap.getRawPtr();
+  Teuchos::ArrayRCP<SC> TBandSol     = Teuchos::arcp<SC>((MaxStencilSize + 1) * DofsPerNode * DofsPerNode);
+  BandSol                            = TBandSol.getRawPtr();
+  Teuchos::ArrayRCP<SC> TResBandSol;
+  if (buildRestriction) {
+    TResBandSol     = Teuchos::arcp<SC>((MaxStencilSize + 1) * DofsPerNode * DofsPerNode);
+    RestrictBandSol = TResBandSol.getRawPtr();
+  }
+  /*
+   * Lapack variables. See comments for dgbsv().
+   */
+  KL                             = 2 * DofsPerNode - 1;
+  KU                             = 2 * DofsPerNode - 1;
+  KLU                            = KL + KU;
+  LDAB                           = 2 * KL + KU + 1;
+  NRHS                           = DofsPerNode;
+  Teuchos::ArrayRCP<SC> TBandMat = Teuchos::arcp<SC>(LDAB * MaxStencilSize * DofsPerNode + 1);
+  BandMat                        = TBandMat.getRawPtr();
+  Teuchos::ArrayRCP<LO> TIPIV    = Teuchos::arcp<LO>((MaxStencilSize + 1) * DofsPerNode);
+  IPIV                           = TIPIV.getRawPtr();
+
+  Teuchos::ArrayRCP<SC> TResBandMat;
+  if (buildRestriction) {
+    TResBandMat     = Teuchos::arcp<SC>(LDAB * MaxStencilSize * DofsPerNode + 1);
+    RestrictBandMat = TResBandMat.getRawPtr();
+  }
 
-    for (i = 1; i <= MaxNnz; i++) Pcols[i] = -1;  /* mark all entries as unused */
+  /*
+   * Allocate storage for the final interpolation matrix. Note: each prolongator
+   * row might have entries corresponding to at most two nodes.
+   * Note: the total fine level dofs equals DofsPerNode*Ntotal and the max
+   *       nnz per prolongator row is DofsPerNode*2.
+   */
+
+  Ndofs  = DofsPerNode * Ntotal;
+  MaxNnz = 2 * DofsPerNode * Ndofs;
+
+  RCP<const Map> rowMap        = Amat->getRowMap();
+  Xpetra::global_size_t GNdofs = rowMap->getGlobalNumElements();
+
+  std::vector<size_t> stridingInfo_;
+  stridingInfo_.push_back(DofsPerNode);
+
+  Xpetra::global_size_t itemp = GNdofs / nz;
+  coarseMap                   = StridedMapFactory::Build(rowMap->lib(),
+                                                         NCLayers * itemp,
+                                                         NCLayers * NVertLines * DofsPerNode,
+                                                         0, /* index base */
+                                                         stridingInfo_,
+                                                         rowMap->getComm(),
+                                                         -1, /* strided block id */
+                                                         0 /* domain gid offset */);
+
+  // coarseMap = MapFactory::createContigMapWithNode(rowMap->lib(),(NCLayers*GNdofs)/nz, NCLayers*NVertLines*DofsPerNode,(rowMap->getComm()), rowMap->getNode());
+
+  P               = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
+  coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
+
+  Teuchos::ArrayRCP<SC> TPvals    = Teuchos::arcp<SC>(1 + MaxNnz);
+  Pvals                           = TPvals.getRawPtr();
+  Teuchos::ArrayRCP<size_t> TPptr = Teuchos::arcp<size_t>(DofsPerNode * (2 + Ntotal));
+  Pptr                            = TPptr.getRawPtr();
+  Teuchos::ArrayRCP<LO> TPcols    = Teuchos::arcp<LO>(1 + MaxNnz);
+  Pcols                           = TPcols.getRawPtr();
+
+  TEUCHOS_TEST_FOR_EXCEPTION(Pcols == NULL, Exceptions::RuntimeError, "MakeSemiCoarsenP: Not enough space \n");
+  Pptr[0] = 0;
+  Pptr[1] = 0;
+
+  Teuchos::ArrayRCP<SC> TRvals;
+  Teuchos::ArrayRCP<size_t> TRptr;
+  Teuchos::ArrayRCP<LO> TRcols;
+  LO RmaxNnz = -1, RmaxPerRow = -1;
+  if (buildRestriction) {
+    RmaxPerRow = ((LO)ceil(((double)(MaxNnz + 7)) / ((double)(coarseMap->getLocalNumElements()))));
+    RmaxNnz    = RmaxPerRow * coarseMap->getLocalNumElements();
+    TRvals     = Teuchos::arcp<SC>(1 + RmaxNnz);
+    Rvals      = TRvals.getRawPtr();
+    TRptr      = Teuchos::arcp<size_t>((2 + coarseMap->getLocalNumElements()));
+    Rptr       = TRptr.getRawPtr();
+    TRcols     = Teuchos::arcp<LO>(1 + RmaxNnz);
+    Rcols      = TRcols.getRawPtr();
+    TEUCHOS_TEST_FOR_EXCEPTION(Rcols == NULL, Exceptions::RuntimeError, "MakeSemiCoarsenP: Not enough space \n");
+    Rptr[0] = 0;
+    Rptr[1] = 0;
+  }
+  /*
+   * Setup P's rowptr as if each row had its maximum of 2*DofsPerNode nonzeros.
+   * This will be useful while filling up P, and then later we will squeeze out
+   * the unused nonzeros locations.
+   */
+
+  for (i = 1; i <= MaxNnz; i++) Pcols[i] = -1; /* mark all entries as unused */
+  count = 1;
+  for (i = 1; i <= DofsPerNode * Ntotal + 1; i++) {
+    Pptr[i] = count;
+    count += (2 * DofsPerNode);
+  }
+  if (buildRestriction) {
+    for (i = 1; i <= RmaxNnz; i++) Rcols[i] = -1; /* mark all entries as unused */
     count = 1;
-    for (i = 1; i <= DofsPerNode*Ntotal+1; i++) {
-      Pptr[i]  = count;
-      count   += (2*DofsPerNode);
-    }
-    if (buildRestriction) {
-      for (i = 1; i <= RmaxNnz; i++) Rcols[i] = -1;  /* mark all entries as unused */
-      count = 1;
-      for (i = 1; i <= (int) (coarseMap->getLocalNumElements()+1); i++) {
-        Rptr[i]  = count;
-        count   += RmaxPerRow;
-      }
+    for (i = 1; i <= (int)(coarseMap->getLocalNumElements() + 1); i++) {
+      Rptr[i] = count;
+      count += RmaxPerRow;
     }
+  }
 
-    /*
-     * Build P column by column. The 1st block column corresponds to the 1st coarse
-     * layer and the first line. The 2nd block column corresponds to the 2nd coarse
-     * layer and the first line. The NCLayers+1 block column corresponds to the
-     * 1st coarse layer and the 2nd line, etc.
-     */
-
-
-    col = 0;
-    for (MyLine=1; MyLine <= NVertLines; MyLine += 1) {
-      for (iii=1; iii <= NCLayers;  iii+= 1) {
-        col = col+1;
-        MyLayer = CptLayers[iii];
-
-        /*
-         * StartLayer gives the layer number of the lowest layer that
-         * is nonzero in the interpolation stencil that is currently
-         * being computed. Normally, if we are not near a boundary this
-         * is simply CptsLayers[iii-1]+1.
-         *
-         * NStencilNodes indicates the number of nonzero nodes in the
-         * interpolation stencil that is currently being computed. Normally,
-         * if we are not near a boundary this is CptLayers[iii+1]-StartLayer.
-         */
-
-        if (iii !=    1    ) StartLayer = CptLayers[iii-1]+1;
-        else                 StartLayer = 1;
-
-        if (iii != NCLayers) NStencilNodes= CptLayers[iii+1]-StartLayer;
-        else                 NStencilNodes= NLayers - StartLayer + 1;
+  /*
+   * Build P column by column. The 1st block column corresponds to the 1st coarse
+   * layer and the first line. The 2nd block column corresponds to the 2nd coarse
+   * layer and the first line. The NCLayers+1 block column corresponds to the
+   * 1st coarse layer and the 2nd line, etc.
+   */
+
+  col = 0;
+  for (MyLine = 1; MyLine <= NVertLines; MyLine += 1) {
+    for (iii = 1; iii <= NCLayers; iii += 1) {
+      col     = col + 1;
+      MyLayer = CptLayers[iii];
+
+      /*
+       * StartLayer gives the layer number of the lowest layer that
+       * is nonzero in the interpolation stencil that is currently
+       * being computed. Normally, if we are not near a boundary this
+       * is simply CptsLayers[iii-1]+1.
+       *
+       * NStencilNodes indicates the number of nonzero nodes in the
+       * interpolation stencil that is currently being computed. Normally,
+       * if we are not near a boundary this is CptLayers[iii+1]-StartLayer.
+       */
+
+      if (iii != 1)
+        StartLayer = CptLayers[iii - 1] + 1;
+      else
+        StartLayer = 1;
 
+      if (iii != NCLayers)
+        NStencilNodes = CptLayers[iii + 1] - StartLayer;
+      else
+        NStencilNodes = NLayers - StartLayer + 1;
 
-        N = NStencilNodes*DofsPerNode;
+      N = NStencilNodes * DofsPerNode;
 
-        /*
-         *  dgbsv() does not require that the first KL rows be initialized,
-         *  so we could avoid zeroing out some entries?
-         */
+      /*
+       *  dgbsv() does not require that the first KL rows be initialized,
+       *  so we could avoid zeroing out some entries?
+       */
 
-        for (i = 0; i < NStencilNodes*DofsPerNode*DofsPerNode; i++) BandSol[ i] = 0.0;
-        for (i = 0; i < LDAB*N; i++) BandMat[ i] = 0.0;
-        if (buildRestriction) {
-          for (i = 0; i < NStencilNodes*DofsPerNode*DofsPerNode; i++) RestrictBandSol[ i] = 0.0;
-          for (i = 0; i < LDAB*N; i++) RestrictBandMat[ i] = 0.0;
-        }
-
-        /*
-         *  Fill BandMat and BandSol (which is initially the rhs) for each
-         *  node in the interpolation stencil that is being computed.
-         */
+      for (i = 0; i < NStencilNodes * DofsPerNode * DofsPerNode; i++) BandSol[i] = 0.0;
+      for (i = 0; i < LDAB * N; i++) BandMat[i] = 0.0;
+      if (buildRestriction) {
+        for (i = 0; i < NStencilNodes * DofsPerNode * DofsPerNode; i++) RestrictBandSol[i] = 0.0;
+        for (i = 0; i < LDAB * N; i++) RestrictBandMat[i] = 0.0;
+      }
 
-	if (!doLinear) {
-        for (node_k=1; node_k <= NStencilNodes ; node_k++) {
+      /*
+       *  Fill BandMat and BandSol (which is initially the rhs) for each
+       *  node in the interpolation stencil that is being computed.
+       */
 
+      if (!doLinear) {
+        for (node_k = 1; node_k <= NStencilNodes; node_k++) {
           /*  Map a Line and Layer number to a BlkRow in the fine level  matrix
            *  and record the mapping from the sub-system to the BlkRow of the
            *  fine level matrix.
            */
-          BlkRow  = InvLineLayer[MyLine+(StartLayer+node_k-2)*NVertLines]+1;
+          BlkRow              = InvLineLayer[MyLine + (StartLayer + node_k - 2) * NVertLines] + 1;
           Sub2FullMap[node_k] = BlkRow;
 
           /* Two cases:
@@ -677,328 +700,332 @@ namespace MueLu {
            *    2) the current layer is a Cpoint layer and so we
            *       stick an identity block in BandMat and rhs.
            */
-            for (int dof_i=0; dof_i < DofsPerNode; dof_i++) {
-
-              j = (BlkRow-1)*DofsPerNode+dof_i;
-              ArrayView<const LO> AAcols;
-              ArrayView<const SC> AAvals;
-              Amat->getLocalRowView(j, AAcols, AAvals);
-              const int *Acols    = AAcols.getRawPtr();
-              const SC *Avals = AAvals.getRawPtr();
-              RowLeng = AAvals.size();
-
-              TEUCHOS_TEST_FOR_EXCEPTION(RowLeng >= MaxNnzPerRow, Exceptions::RuntimeError, "MakeSemiCoarsenP: recompile with larger Max(HorNeighborNodes)\n");
-
+          for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+            j = (BlkRow - 1) * DofsPerNode + dof_i;
+            ArrayView<const LO> AAcols;
+            ArrayView<const SC> AAvals;
+            Amat->getLocalRowView(j, AAcols, AAvals);
+            const int* Acols = AAcols.getRawPtr();
+            const SC* Avals  = AAvals.getRawPtr();
+            RowLeng          = AAvals.size();
+
+            TEUCHOS_TEST_FOR_EXCEPTION(RowLeng >= MaxNnzPerRow, Exceptions::RuntimeError, "MakeSemiCoarsenP: recompile with larger Max(HorNeighborNodes)\n");
+
+            for (i = 0; i < RowLeng; i++) {
+              LayDiff[i] = Layerdofs[Acols[i]] - StartLayer - node_k + 2;
+
+              /* This is the main spot where there might be off- */
+              /* processor communication. That is, when we       */
+              /* average the stencil in the horizontal direction,*/
+              /* we need to know the layer id of some            */
+              /* neighbors that might reside off-processor.      */
+            }
+            PtRow = (node_k - 1) * DofsPerNode + dof_i + 1;
+            for (dof_j = 0; dof_j < DofsPerNode; dof_j++) {
+              PtCol = (node_k - 1) * DofsPerNode + dof_j + 1;
+              /* Stick in entry corresponding to Mat(PtRow,PtCol) */
+              /* see dgbsv() comments for matrix format.          */
+              TheSum = 0.0;
               for (i = 0; i < RowLeng; i++) {
-                LayDiff[i]  = Layerdofs[Acols[i]]-StartLayer-node_k+2;
-
-                /* This is the main spot where there might be off- */
-                /* processor communication. That is, when we       */
-                /* average the stencil in the horizontal direction,*/
-                /* we need to know the layer id of some            */
-                /* neighbors that might reside off-processor.      */
+                if ((LayDiff[i] == 0) && (Col2Dof[Acols[i]] == dof_j))
+                  TheSum += Avals[i];
               }
-              PtRow = (node_k-1)*DofsPerNode+dof_i+1;
-              for (dof_j=0; dof_j < DofsPerNode; dof_j++) {
-                PtCol = (node_k-1)*DofsPerNode+dof_j + 1;
-                /* Stick in entry corresponding to Mat(PtRow,PtCol) */
-                /* see dgbsv() comments for matrix format.          */
+              index          = LDAB * (PtCol - 1) + KLU + PtRow - PtCol;
+              BandMat[index] = TheSum;
+              if (buildRestriction) RestrictBandMat[index] = TheSum;
+              if (node_k != NStencilNodes) {
+                /* Stick Mat(PtRow,PtCol+DofsPerNode) entry  */
+                /* see dgbsv() comments for matrix format.  */
                 TheSum = 0.0;
                 for (i = 0; i < RowLeng; i++) {
-                  if ((LayDiff[i] == 0)  && (Col2Dof[Acols[i]] == dof_j))
+                  if ((LayDiff[i] == 1) && (Col2Dof[Acols[i]] == dof_j))
                     TheSum += Avals[i];
                 }
-                index = LDAB*(PtCol-1)+KLU+PtRow-PtCol;
+                j              = PtCol + DofsPerNode;
+                index          = LDAB * (j - 1) + KLU + PtRow - j;
                 BandMat[index] = TheSum;
                 if (buildRestriction) RestrictBandMat[index] = TheSum;
-                if (node_k != NStencilNodes) {
-                  /* Stick Mat(PtRow,PtCol+DofsPerNode) entry  */
-                  /* see dgbsv() comments for matrix format.  */
-                  TheSum = 0.0;
-                  for (i = 0; i < RowLeng; i++) {
-                    if ((LayDiff[i] == 1) &&(Col2Dof[Acols[i]]== dof_j))
-                      TheSum += Avals[i];
-                  }
-                  j = PtCol+DofsPerNode;
-                  index=LDAB*(j-1)+KLU+PtRow-j;
-                  BandMat[index] = TheSum;
-                  if (buildRestriction) RestrictBandMat[index] = TheSum;
-                }
-                if (node_k != 1) {
-                  /* Stick Mat(PtRow,PtCol-DofsPerNode) entry  */
-                  /* see dgbsv() comments for matrix format.  */
-                  TheSum = 0.0;
-                  for (i = 0; i < RowLeng; i++) {
-                    if ((LayDiff[i]== -1) &&(Col2Dof[Acols[i]]== dof_j))
-                      TheSum += Avals[i];
-                  }
-                  j = PtCol-DofsPerNode;
-                  index=LDAB*(j-1)+KLU+PtRow-j;
-                  BandMat[index] = TheSum;
-                  if (buildRestriction) RestrictBandMat[index] = TheSum;
+              }
+              if (node_k != 1) {
+                /* Stick Mat(PtRow,PtCol-DofsPerNode) entry  */
+                /* see dgbsv() comments for matrix format.  */
+                TheSum = 0.0;
+                for (i = 0; i < RowLeng; i++) {
+                  if ((LayDiff[i] == -1) && (Col2Dof[Acols[i]] == dof_j))
+                    TheSum += Avals[i];
                 }
+                j              = PtCol - DofsPerNode;
+                index          = LDAB * (j - 1) + KLU + PtRow - j;
+                BandMat[index] = TheSum;
+                if (buildRestriction) RestrictBandMat[index] = TheSum;
               }
             }
+          }
         }
-        node_k = MyLayer - StartLayer+1;
-            for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
-              /* Stick Mat(PtRow,PtRow) and Rhs(PtRow,dof_i+1) */
-              /* see dgbsv() comments for matrix format.     */
-              PtRow = (node_k-1)*DofsPerNode+dof_i+1;
-              BandSol[(dof_i)*DofsPerNode*NStencilNodes+PtRow-1] = 1.;
-              if (buildRestriction) RestrictBandSol[(dof_i)*DofsPerNode*NStencilNodes+PtRow-1] = 1.;
-
-              for (dof_j = 0; dof_j < DofsPerNode; dof_j++) {
-                PtCol = (node_k-1)*DofsPerNode+dof_j+1;
-                index = LDAB*(PtCol-1)+KLU+PtRow-PtCol;
-                if (PtCol == PtRow)  BandMat[index] = 1.0;
-                else                 BandMat[index] = 0.0;
-                if (buildRestriction) {
-                  index = LDAB*(PtRow-1)+KLU+PtCol-PtRow;
-                  if (PtCol == PtRow)  RestrictBandMat[index] = 1.0;
-                  else                 RestrictBandMat[index] = 0.0;
-                }
-                if (node_k != NStencilNodes) {
-                  PtCol = (node_k  )*DofsPerNode+dof_j+1;
-                  index = LDAB*(PtCol-1)+KLU+PtRow-PtCol;
-                  BandMat[index] = 0.0;
-                  if (buildRestriction) {
-                    index = LDAB*(PtRow-1)+KLU+PtCol-PtRow;
-                    RestrictBandMat[index] = 0.0;
-                  }
-                }
-                if (node_k != 1) {
-                  PtCol = (node_k-2)*DofsPerNode+dof_j+1;
-                  index = LDAB*(PtCol-1)+KLU+PtRow-PtCol;
-                  BandMat[index] = 0.0;
-                  if (buildRestriction) {
-                    index = LDAB*(PtRow-1)+KLU+PtCol-PtRow;
-                    RestrictBandMat[index] = 0.0;
-                  }
-                }
+        node_k = MyLayer - StartLayer + 1;
+        for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+          /* Stick Mat(PtRow,PtRow) and Rhs(PtRow,dof_i+1) */
+          /* see dgbsv() comments for matrix format.     */
+          PtRow                                                    = (node_k - 1) * DofsPerNode + dof_i + 1;
+          BandSol[(dof_i)*DofsPerNode * NStencilNodes + PtRow - 1] = 1.;
+          if (buildRestriction) RestrictBandSol[(dof_i)*DofsPerNode * NStencilNodes + PtRow - 1] = 1.;
+
+          for (dof_j = 0; dof_j < DofsPerNode; dof_j++) {
+            PtCol = (node_k - 1) * DofsPerNode + dof_j + 1;
+            index = LDAB * (PtCol - 1) + KLU + PtRow - PtCol;
+            if (PtCol == PtRow)
+              BandMat[index] = 1.0;
+            else
+              BandMat[index] = 0.0;
+            if (buildRestriction) {
+              index = LDAB * (PtRow - 1) + KLU + PtCol - PtRow;
+              if (PtCol == PtRow)
+                RestrictBandMat[index] = 1.0;
+              else
+                RestrictBandMat[index] = 0.0;
+            }
+            if (node_k != NStencilNodes) {
+              PtCol          = (node_k)*DofsPerNode + dof_j + 1;
+              index          = LDAB * (PtCol - 1) + KLU + PtRow - PtCol;
+              BandMat[index] = 0.0;
+              if (buildRestriction) {
+                index                  = LDAB * (PtRow - 1) + KLU + PtCol - PtRow;
+                RestrictBandMat[index] = 0.0;
               }
             }
+            if (node_k != 1) {
+              PtCol          = (node_k - 2) * DofsPerNode + dof_j + 1;
+              index          = LDAB * (PtCol - 1) + KLU + PtRow - PtCol;
+              BandMat[index] = 0.0;
+              if (buildRestriction) {
+                index                  = LDAB * (PtRow - 1) + KLU + PtCol - PtRow;
+                RestrictBandMat[index] = 0.0;
+              }
+            }
+          }
+        }
 
         /* Solve banded system and then stick result in Pmatrix arrays */
 
-        lapack.GBTRF( N, N, KL, KU, BandMat, LDAB, IPIV, &INFO);
+        lapack.GBTRF(N, N, KL, KU, BandMat, LDAB, IPIV, &INFO);
 
         TEUCHOS_TEST_FOR_EXCEPTION(INFO != 0, Exceptions::RuntimeError, "Lapack band factorization failed");
 
         lapack.GBTRS(notrans[0], N, KL, KU, NRHS, BandMat, LDAB, IPIV,
-                     BandSol, N, &INFO );
+                     BandSol, N, &INFO);
 
         TEUCHOS_TEST_FOR_EXCEPTION(INFO != 0, Exceptions::RuntimeError, "Lapack band solve back substitution failed");
 
-        for (dof_j=0; dof_j < DofsPerNode; dof_j++) {
-          for (int dof_i=0; dof_i < DofsPerNode; dof_i++) {
-            for (i =1; i <= NStencilNodes ; i++) {
-              index = (Sub2FullMap[i]-1)*DofsPerNode+dof_i+1;
-              loc = Pptr[index];
-              Pcols[loc] = (col-1)*DofsPerNode+dof_j+1;
-              Pvals[loc] = BandSol[dof_j*DofsPerNode*NStencilNodes +
-                  (i-1)*DofsPerNode + dof_i];
-              Pptr[index]= Pptr[index] + 1;
+        for (dof_j = 0; dof_j < DofsPerNode; dof_j++) {
+          for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+            for (i = 1; i <= NStencilNodes; i++) {
+              index       = (Sub2FullMap[i] - 1) * DofsPerNode + dof_i + 1;
+              loc         = Pptr[index];
+              Pcols[loc]  = (col - 1) * DofsPerNode + dof_j + 1;
+              Pvals[loc]  = BandSol[dof_j * DofsPerNode * NStencilNodes +
+                                   (i - 1) * DofsPerNode + dof_i];
+              Pptr[index] = Pptr[index] + 1;
             }
           }
         }
         if (buildRestriction) {
-          lapack.GBTRF( N, N, KL, KU, RestrictBandMat, LDAB, IPIV, &INFO);
+          lapack.GBTRF(N, N, KL, KU, RestrictBandMat, LDAB, IPIV, &INFO);
           TEUCHOS_TEST_FOR_EXCEPTION(INFO != 0, Exceptions::RuntimeError, "Lapack band factorization failed");
-          lapack.GBTRS(trans[0], N, KL, KU, NRHS, RestrictBandMat, LDAB, IPIV, RestrictBandSol, N, &INFO );
+          lapack.GBTRS(trans[0], N, KL, KU, NRHS, RestrictBandMat, LDAB, IPIV, RestrictBandSol, N, &INFO);
           TEUCHOS_TEST_FOR_EXCEPTION(INFO != 0, Exceptions::RuntimeError, "Lapack band solve back substitution failed");
-          for (dof_j=0; dof_j < DofsPerNode; dof_j++) {
-            for (int dof_i=0; dof_i < DofsPerNode; dof_i++) {
-              for (i =1; i <= NStencilNodes ; i++) {
-                index = (col-1)*DofsPerNode+dof_j+1;
-                loc = Rptr[index];
-                Rcols[loc] = (Sub2FullMap[i]-1)*DofsPerNode+dof_i+1;
-                Rvals[loc] = RestrictBandSol[dof_j*DofsPerNode*NStencilNodes +
-                    (i-1)*DofsPerNode + dof_i];
-                Rptr[index]= Rptr[index] + 1;
+          for (dof_j = 0; dof_j < DofsPerNode; dof_j++) {
+            for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+              for (i = 1; i <= NStencilNodes; i++) {
+                index       = (col - 1) * DofsPerNode + dof_j + 1;
+                loc         = Rptr[index];
+                Rcols[loc]  = (Sub2FullMap[i] - 1) * DofsPerNode + dof_i + 1;
+                Rvals[loc]  = RestrictBandSol[dof_j * DofsPerNode * NStencilNodes +
+                                             (i - 1) * DofsPerNode + dof_i];
+                Rptr[index] = Rptr[index] + 1;
               }
             }
           }
         }
-	}
-	else {
-	  int denom1 = MyLayer-StartLayer+1;
-	  int denom2 = StartLayer+NStencilNodes-MyLayer;
-          for (int dof_i=0; dof_i < DofsPerNode; dof_i++) {
-            for (i =1; i <= NStencilNodes ; i++) {
-              index = (InvLineLayer[MyLine+(StartLayer+i-2)*NVertLines])*DofsPerNode+dof_i+1;
-              loc = Pptr[index];
-              Pcols[loc] = (col-1)*DofsPerNode+dof_i+1;
-	      if (i > denom1) Pvals[loc] = 1.0 + ((double)( denom1 - i))/((double) denom2);
-	      else  Pvals[loc] = ((double)( i))/((double) denom1);
-              Pptr[index]= Pptr[index] + 1;
-	    }
-	  }
-	}
-             /* inject the null space */
-            // int FineStride  = Ntotal*DofsPerNode;
-            // int CoarseStride= NVertLines*NCLayers*DofsPerNode;
-
-            BlkRow  = InvLineLayer[MyLine+(MyLayer-1)*NVertLines]+1;
-            for (int k = 0; k < static_cast<int>(fineNullspace->getNumVectors()); k++) {
-              Teuchos::ArrayRCP<SC> OneCNull = coarseNullspace->getDataNonConst(k);
-              Teuchos::ArrayRCP<SC> OneFNull = fineNullspace->getDataNonConst(k);
-              for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
-                 OneCNull[(   col-1)*DofsPerNode+dof_i] = OneFNull[ (BlkRow-1)*DofsPerNode+dof_i];
-              }
-            }
-
+      } else {
+        int denom1 = MyLayer - StartLayer + 1;
+        int denom2 = StartLayer + NStencilNodes - MyLayer;
+        for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+          for (i = 1; i <= NStencilNodes; i++) {
+            index      = (InvLineLayer[MyLine + (StartLayer + i - 2) * NVertLines]) * DofsPerNode + dof_i + 1;
+            loc        = Pptr[index];
+            Pcols[loc] = (col - 1) * DofsPerNode + dof_i + 1;
+            if (i > denom1)
+              Pvals[loc] = 1.0 + ((double)(denom1 - i)) / ((double)denom2);
+            else
+              Pvals[loc] = ((double)(i)) / ((double)denom1);
+            Pptr[index] = Pptr[index] + 1;
+          }
+        }
+      }
+      /* inject the null space */
+      // int FineStride  = Ntotal*DofsPerNode;
+      // int CoarseStride= NVertLines*NCLayers*DofsPerNode;
+
+      BlkRow = InvLineLayer[MyLine + (MyLayer - 1) * NVertLines] + 1;
+      for (int k = 0; k < static_cast<int>(fineNullspace->getNumVectors()); k++) {
+        Teuchos::ArrayRCP<SC> OneCNull = coarseNullspace->getDataNonConst(k);
+        Teuchos::ArrayRCP<SC> OneFNull = fineNullspace->getDataNonConst(k);
+        for (int dof_i = 0; dof_i < DofsPerNode; dof_i++) {
+          OneCNull[(col - 1) * DofsPerNode + dof_i] = OneFNull[(BlkRow - 1) * DofsPerNode + dof_i];
+        }
       }
     }
+  }
 
-    /*
-     * Squeeze the -1's out of the columns. At the same time convert Pcols
-     * so that now the first column is numbered '0' as opposed to '1'.
-     * Also, the arrays Pcols and Pvals should now use the zeroth element
-     * as opposed to just starting with the first element. Pptr will be
-     * fixed in the for loop below so that Pptr[0] = 0, etc.
-     */
-    CurRow = 1;
-    NewPtr = 1;
-    for (size_t ii=1; ii <= Pptr[Ntotal*DofsPerNode]-1; ii++) {
-      if (ii == Pptr[CurRow]) {
+  /*
+   * Squeeze the -1's out of the columns. At the same time convert Pcols
+   * so that now the first column is numbered '0' as opposed to '1'.
+   * Also, the arrays Pcols and Pvals should now use the zeroth element
+   * as opposed to just starting with the first element. Pptr will be
+   * fixed in the for loop below so that Pptr[0] = 0, etc.
+   */
+  CurRow = 1;
+  NewPtr = 1;
+  for (size_t ii = 1; ii <= Pptr[Ntotal * DofsPerNode] - 1; ii++) {
+    if (ii == Pptr[CurRow]) {
+      Pptr[CurRow] = LastGuy;
+      CurRow++;
+      while (ii > Pptr[CurRow]) {
         Pptr[CurRow] = LastGuy;
         CurRow++;
-        while (ii > Pptr[CurRow]) {
-          Pptr[CurRow] = LastGuy;
-          CurRow++;
-        }
-      }
-      if (Pcols[ii] != -1) {
-        Pcols[NewPtr-1] = Pcols[ii]-1;   /* these -1's fix the offset and */
-        Pvals[NewPtr-1] = Pvals[ii];     /* start using the zeroth element */
-        LastGuy = NewPtr;
-        NewPtr++;
       }
     }
-    for (i = CurRow; i <= Ntotal*DofsPerNode; i++) Pptr[CurRow] = LastGuy;
-
-    /* Now move the pointers so that they now point to the beginning of each
-     * row as opposed to the end of each row
-     */
-    for (i=-Ntotal*DofsPerNode+1; i>= 2 ; i--) {
-      Pptr[i-1] = Pptr[i-2];  /* extra -1 added to start from 0 */
+    if (Pcols[ii] != -1) {
+      Pcols[NewPtr - 1] = Pcols[ii] - 1; /* these -1's fix the offset and */
+      Pvals[NewPtr - 1] = Pvals[ii];     /* start using the zeroth element */
+      LastGuy           = NewPtr;
+      NewPtr++;
     }
-    Pptr[0] = 0;
-
-    // do the same for R
-    if (buildRestriction) {
-      CurRow = 1;
-      NewPtr = 1;
-      for (size_t ii=1; ii <= Rptr[coarseMap->getLocalNumElements()]-1; ii++) {
-        if (ii == Rptr[CurRow]) {
+  }
+  for (i = CurRow; i <= Ntotal * DofsPerNode; i++) Pptr[CurRow] = LastGuy;
+
+  /* Now move the pointers so that they now point to the beginning of each
+   * row as opposed to the end of each row
+   */
+  for (i = -Ntotal * DofsPerNode + 1; i >= 2; i--) {
+    Pptr[i - 1] = Pptr[i - 2]; /* extra -1 added to start from 0 */
+  }
+  Pptr[0] = 0;
+
+  // do the same for R
+  if (buildRestriction) {
+    CurRow = 1;
+    NewPtr = 1;
+    for (size_t ii = 1; ii <= Rptr[coarseMap->getLocalNumElements()] - 1; ii++) {
+      if (ii == Rptr[CurRow]) {
+        Rptr[CurRow] = RLastGuy;
+        CurRow++;
+        while (ii > Rptr[CurRow]) {
           Rptr[CurRow] = RLastGuy;
           CurRow++;
-          while (ii > Rptr[CurRow]) {
-            Rptr[CurRow] = RLastGuy;
-            CurRow++;
-          }
-        }
-        if (Rcols[ii] != -1) {
-          Rcols[NewPtr-1] = Rcols[ii]-1;   /* these -1's fix the offset and */
-          Rvals[NewPtr-1] = Rvals[ii];     /* start using the zeroth element */
-          RLastGuy = NewPtr;
-          NewPtr++;
         }
       }
-      for (i = CurRow; i <=  ((int) coarseMap->getLocalNumElements()); i++) Rptr[CurRow] = RLastGuy;
-      for (i=-coarseMap->getLocalNumElements()+1; i>= 2 ; i--) {
-        Rptr[i-1] = Rptr[i-2];  /* extra -1 added to start from 0 */
+      if (Rcols[ii] != -1) {
+        Rcols[NewPtr - 1] = Rcols[ii] - 1; /* these -1's fix the offset and */
+        Rvals[NewPtr - 1] = Rvals[ii];     /* start using the zeroth element */
+        RLastGuy          = NewPtr;
+        NewPtr++;
       }
-      Rptr[0] = 0;
     }
-    ArrayRCP<size_t>  rcpRowPtr;
-    ArrayRCP<LO>      rcpColumns;
-    ArrayRCP<SC>      rcpValues;
-
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-    LO nnz =  Pptr[Ndofs];
-    PCrs->allocateAllValues(nnz, rcpRowPtr, rcpColumns, rcpValues);
-
-    ArrayView<size_t> rowPtr  = rcpRowPtr();
-    ArrayView<LO>     columns = rcpColumns();
-    ArrayView<SC>     values  = rcpValues();
+    for (i = CurRow; i <= ((int)coarseMap->getLocalNumElements()); i++) Rptr[CurRow] = RLastGuy;
+    for (i = -coarseMap->getLocalNumElements() + 1; i >= 2; i--) {
+      Rptr[i - 1] = Rptr[i - 2]; /* extra -1 added to start from 0 */
+    }
+    Rptr[0] = 0;
+  }
+  ArrayRCP<size_t> rcpRowPtr;
+  ArrayRCP<LO> rcpColumns;
+  ArrayRCP<SC> rcpValues;
+
+  RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  LO nnz              = Pptr[Ndofs];
+  PCrs->allocateAllValues(nnz, rcpRowPtr, rcpColumns, rcpValues);
+
+  ArrayView<size_t> rowPtr = rcpRowPtr();
+  ArrayView<LO> columns    = rcpColumns();
+  ArrayView<SC> values     = rcpValues();
+
+  // copy data over
+
+  for (LO ii = 0; ii <= Ndofs; ii++) rowPtr[ii] = Pptr[ii];
+  for (LO ii = 0; ii < nnz; ii++) columns[ii] = Pcols[ii];
+  for (LO ii = 0; ii < nnz; ii++) values[ii] = Pvals[ii];
+  PCrs->setAllValues(rcpRowPtr, rcpColumns, rcpValues);
+  PCrs->expertStaticFillComplete(coarseMap, Amat->getDomainMap());
+  ArrayRCP<size_t> RrcpRowPtr;
+  ArrayRCP<LO> RrcpColumns;
+  ArrayRCP<SC> RrcpValues;
+  RCP<CrsMatrix> RCrs;
+  if (buildRestriction) {
+    R    = rcp(new CrsMatrixWrap(coarseMap, rowMap, 0));
+    RCrs = rcp_dynamic_cast<CrsMatrixWrap>(R)->getCrsMatrix();
+    nnz  = Rptr[coarseMap->getLocalNumElements()];
+    RCrs->allocateAllValues(nnz, RrcpRowPtr, RrcpColumns, RrcpValues);
+
+    ArrayView<size_t> RrowPtr = RrcpRowPtr();
+    ArrayView<LO> Rcolumns    = RrcpColumns();
+    ArrayView<SC> Rvalues     = RrcpValues();
 
     // copy data over
 
-    for (LO ii = 0; ii <= Ndofs; ii++) rowPtr[ii] = Pptr[ii];
-    for (LO ii = 0; ii < nnz; ii++)    columns[ii] = Pcols[ii];
-    for (LO ii = 0; ii < nnz; ii++) values[ii]  = Pvals[ii];
-    PCrs->setAllValues(rcpRowPtr, rcpColumns, rcpValues);
-    PCrs->expertStaticFillComplete(coarseMap, Amat->getDomainMap());
-    ArrayRCP<size_t>  RrcpRowPtr;
-    ArrayRCP<LO>      RrcpColumns;
-    ArrayRCP<SC>      RrcpValues;
-    RCP<CrsMatrix> RCrs;
-    if (buildRestriction) {
-      R      = rcp(new CrsMatrixWrap(coarseMap, rowMap, 0));
-      RCrs = rcp_dynamic_cast<CrsMatrixWrap>(R)->getCrsMatrix();
-      nnz =  Rptr[coarseMap->getLocalNumElements()];
-      RCrs->allocateAllValues(nnz, RrcpRowPtr, RrcpColumns, RrcpValues);
-
-      ArrayView<size_t> RrowPtr  = RrcpRowPtr();
-      ArrayView<LO>     Rcolumns = RrcpColumns();
-      ArrayView<SC>     Rvalues  = RrcpValues();
-
-      // copy data over
-
-      for (LO ii = 0; ii <= ((LO)coarseMap->getLocalNumElements()); ii++) RrowPtr[ii]  = Rptr[ii];
-      for (LO ii = 0; ii < nnz; ii++)    Rcolumns[ii] = Rcols[ii];
-      for (LO ii = 0; ii < nnz; ii++)    Rvalues[ii]  = Rvals[ii];
-      RCrs->setAllValues(RrcpRowPtr, RrcpColumns, RrcpValues);
-      RCrs->expertStaticFillComplete(Amat->getRangeMap(), coarseMap);
-    }
-
-    return NCLayers*NVertLines*DofsPerNode;
+    for (LO ii = 0; ii <= ((LO)coarseMap->getLocalNumElements()); ii++) RrowPtr[ii] = Rptr[ii];
+    for (LO ii = 0; ii < nnz; ii++) Rcolumns[ii] = Rcols[ii];
+    for (LO ii = 0; ii < nnz; ii++) Rvalues[ii] = Rvals[ii];
+    RCrs->setAllValues(RrcpRowPtr, RrcpColumns, RrcpValues);
+    RCrs->expertStaticFillComplete(Amat->getRangeMap(), coarseMap);
   }
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RevertToPieceWiseConstant( RCP<Matrix>& P, LO BlkSize) const {
-    // This function is a bit of a hack. We basically compute the semi-coarsening transfers and then throw
-    // away all the interpolation coefficients, instead replacing them by piecewise constants. The reason for this
-    // is that SemiCoarsening has no notion of aggregates so defining piecewise constants in the "usual way" is
-    // not possible. Instead, we look for the largest entry in each row, make it a one, and zero out the other
-    // non-zero entries
-
-    ArrayView<const LocalOrdinal> inds;
-    ArrayView<const Scalar> vals1;
-    ArrayView<      Scalar> vals;
-    Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    Scalar ONE  = Teuchos::ScalarTraits<Scalar>::one();
-
-    for (size_t i = 0; i < P->getRowMap()->getLocalNumElements(); i++) {
-      P->getLocalRowView(i, inds, vals1);
-
-      size_t nnz = inds.size();
-      if (nnz != 0) vals = ArrayView<Scalar>(const_cast<Scalar*>(vals1.getRawPtr()), nnz);
-
-      LO largestIndex = -1;
-      Scalar largestValue = ZERO;
-      /* find largest value in row and change that one to a 1 while the others are set to 0 */
-
-      LO rowDof = i%BlkSize;
-      for (size_t j =0; j < nnz; j++) {
-        if (Teuchos::ScalarTraits<SC>::magnitude(vals[ j ]) >= Teuchos::ScalarTraits<SC>::magnitude(largestValue)) {
-          if ( inds[j]%BlkSize == rowDof ) {
-            largestValue = vals[j];
-            largestIndex = (int) j;
-          }
+
+  return NCLayers * NVertLines * DofsPerNode;
+}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SemiCoarsenPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::RevertToPieceWiseConstant(RCP<Matrix>& P, LO BlkSize) const {
+  // This function is a bit of a hack. We basically compute the semi-coarsening transfers and then throw
+  // away all the interpolation coefficients, instead replacing them by piecewise constants. The reason for this
+  // is that SemiCoarsening has no notion of aggregates so defining piecewise constants in the "usual way" is
+  // not possible. Instead, we look for the largest entry in each row, make it a one, and zero out the other
+  // non-zero entries
+
+  ArrayView<const LocalOrdinal> inds;
+  ArrayView<const Scalar> vals1;
+  ArrayView<Scalar> vals;
+  Scalar ZERO = Teuchos::ScalarTraits<Scalar>::zero();
+  Scalar ONE  = Teuchos::ScalarTraits<Scalar>::one();
+
+  for (size_t i = 0; i < P->getRowMap()->getLocalNumElements(); i++) {
+    P->getLocalRowView(i, inds, vals1);
+
+    size_t nnz = inds.size();
+    if (nnz != 0) vals = ArrayView<Scalar>(const_cast<Scalar*>(vals1.getRawPtr()), nnz);
+
+    LO largestIndex     = -1;
+    Scalar largestValue = ZERO;
+    /* find largest value in row and change that one to a 1 while the others are set to 0 */
+
+    LO rowDof = i % BlkSize;
+    for (size_t j = 0; j < nnz; j++) {
+      if (Teuchos::ScalarTraits<SC>::magnitude(vals[j]) >= Teuchos::ScalarTraits<SC>::magnitude(largestValue)) {
+        if (inds[j] % BlkSize == rowDof) {
+          largestValue = vals[j];
+          largestIndex = (int)j;
         }
-        vals[j] = ZERO;
       }
-      if (largestIndex != -1) vals[largestIndex] = ONE;
-      else
-        TEUCHOS_TEST_FOR_EXCEPTION(nnz > 0, Exceptions::RuntimeError, "no nonzero column associated with a proper dof within node.");
-
-      if (Teuchos::ScalarTraits<SC>::magnitude(largestValue) == Teuchos::ScalarTraits<SC>::magnitude(ZERO)) vals[largestIndex] = ZERO;
+      vals[j] = ZERO;
     }
+    if (largestIndex != -1)
+      vals[largestIndex] = ONE;
+    else
+      TEUCHOS_TEST_FOR_EXCEPTION(nnz > 0, Exceptions::RuntimeError, "no nonzero column associated with a proper dof within node.");
+
+    if (Teuchos::ScalarTraits<SC>::magnitude(largestValue) == Teuchos::ScalarTraits<SC>::magnitude(ZERO)) vals[largestIndex] = ZERO;
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SEMICOARSENPFACTORY_SHORT
-#endif // MUELU_SEMICOARSENPFACTORY_DEF_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_decl.hpp
index b9d572c1b936..8a4e969bc619 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_decl.hpp
@@ -101,25 +101,26 @@ template <class Scalar, class LocalOrdinal, class GlobalOrdinal,
 class SemiCoarsenPFactory_kokkos<
     Scalar, LocalOrdinal, GlobalOrdinal,
     Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>> : public PFactory {
-public:
+ public:
   typedef LocalOrdinal local_ordinal_type;
   typedef GlobalOrdinal global_ordinal_type;
   typedef typename DeviceType::execution_space execution_space;
   typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
 
-private:
+ private:
   // For compatibility
   typedef node_type Node;
 
 #undef MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-public:
+ public:
   //! @name Constructors/Destructors.
   //@{
 
   //! Constructor
-  SemiCoarsenPFactory_kokkos() : bTransferCoordinates_(false) {}
+  SemiCoarsenPFactory_kokkos()
+    : bTransferCoordinates_(false) {}
 
   //! Destructor.
   virtual ~SemiCoarsenPFactory_kokkos() = default;
@@ -152,9 +153,9 @@ class SemiCoarsenPFactory_kokkos<
   // boolean which is true if coordinate information is available to be
   // transferred to coarse coordinate information
   mutable bool bTransferCoordinates_;
-}; // class SemiCoarsenPFactory_kokkos
+};  // class SemiCoarsenPFactory_kokkos
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
-#endif // MUELU_SEMICOARSENPFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_def.hpp
index 7406af3914a2..e4ac4675278b 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_SemiCoarsenPFactory_kokkos_def.hpp
@@ -161,16 +161,16 @@ void SemiCoarsenPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal,
 
   // get user-provided coarsening rate parameter (constant over all levels)
   const ParameterList &pL = GetParameterList();
-  LO CoarsenRate = as<LO>(pL.get<int>("semicoarsen: coarsen rate"));
+  LO CoarsenRate          = as<LO>(pL.get<int>("semicoarsen: coarsen rate"));
   TEUCHOS_TEST_FOR_EXCEPTION(
       CoarsenRate < 2, Exceptions::RuntimeError,
       "semicoarsen: coarsen rate must be greater than 1");
 
   // collect general input data
-  LO BlkSize = A->GetFixedBlockSize();
+  LO BlkSize            = A->GetFixedBlockSize();
   RCP<const Map> rowMap = A->getRowMap();
-  LO Ndofs = rowMap->getLocalNumElements();
-  LO Nnodes = Ndofs / BlkSize;
+  LO Ndofs              = rowMap->getLocalNumElements();
+  LO Nnodes             = Ndofs / BlkSize;
 
   // collect line detection information generated by the LineDetectionFactory
   // instance
@@ -281,7 +281,7 @@ void SemiCoarsenPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal,
     // ordering according to the numbering of the vertical lines
 
     // number of vertical lines on current node:
-    LO numVertLines = Nnodes / FineNumZLayers;
+    LO numVertLines        = Nnodes / FineNumZLayers;
     LO numLocalCoarseNodes = numVertLines * myCoarseZLayers;
 
     RCP<const Map> coarseCoordMap = MapFactory::Build(
@@ -341,7 +341,7 @@ void SemiCoarsenPFactory_kokkos<
                       const RCP<MultiVector> fineNullspace, RCP<Matrix> &P,
                       RCP<MultiVector> &coarseNullspace) const {
   SubFactoryMonitor m2(*this, "BuildSemiCoarsenP", coarseLevel);
-  using impl_SC = typename Kokkos::ArithTraits<SC>::val_type;
+  using impl_SC  = typename Kokkos::ArithTraits<SC>::val_type;
   using impl_ATS = Kokkos::ArithTraits<impl_SC>;
   using LOView1D = Kokkos::View<LO *, DeviceType>;
   using LOView2D = Kokkos::View<LO **, DeviceType>;
@@ -365,7 +365,7 @@ void SemiCoarsenPFactory_kokkos<
     FCol2LayerVector->doImport(*localTemp, *(importer), Xpetra::INSERT);
   }
   const auto FCol2LayerView = FCol2LayerVector->getDeviceLocalView(Xpetra::Access::ReadOnly);
-  const auto FCol2Layer = Kokkos::subview(FCol2LayerView, Kokkos::ALL(), 0);
+  const auto FCol2Layer     = Kokkos::subview(FCol2LayerView, Kokkos::ALL(), 0);
 
   // Construct a map from fine level column to local dof per node id (including
   // ghost nodes) Note: this is needed to sum all couplings within a layer
@@ -381,7 +381,7 @@ void SemiCoarsenPFactory_kokkos<
     FCol2DofVector->doImport(*localTemp, *(importer), Xpetra::INSERT);
   }
   const auto FCol2DofView = FCol2DofVector->getDeviceLocalView(Xpetra::Access::ReadOnly);
-  const auto FCol2Dof = Kokkos::subview(FCol2DofView, Kokkos::ALL(), 0);
+  const auto FCol2Dof     = Kokkos::subview(FCol2DofView, Kokkos::ALL(), 0);
 
   // Compute NVertLines
   // TODO: Read this from line detection factory
@@ -431,14 +431,14 @@ void SemiCoarsenPFactory_kokkos<
   typename LOView1D::HostMirror CLayer2StencilSizeHost =
       Kokkos::create_mirror_view(CLayer2StencilSize);
   for (int clayer = 0; clayer < NCLayers; ++clayer) {
-    const int startLayer = (clayer > 0) ? CLayer2FLayerHost(clayer - 1) + 1 : 0;
+    const int startLayer  = (clayer > 0) ? CLayer2FLayerHost(clayer - 1) + 1 : 0;
     const int stencilSize = (clayer < NCLayers - 1)
                                 ? CLayer2FLayerHost(clayer + 1) - startLayer
                                 : NFLayers - startLayer;
 
     if (MaxStencilSize < stencilSize)
       MaxStencilSize = stencilSize;
-    CLayer2StartLayerHost(clayer) = startLayer;
+    CLayer2StartLayerHost(clayer)  = startLayer;
     CLayer2StencilSizeHost(clayer) = stencilSize;
   }
   Kokkos::deep_copy(CLayer2StartLayer, CLayer2StartLayerHost);
@@ -475,12 +475,12 @@ void SemiCoarsenPFactory_kokkos<
   for (int line = 0; line < NVertLines; ++line) {
     for (int clayer = 0; clayer < NCLayers; ++clayer) {
       const int stencilSize = CLayer2StencilSizeHost(clayer);
-      const int startLayer = CLayer2StartLayerHost(clayer);
+      const int startLayer  = CLayer2StartLayerHost(clayer);
       for (int snode = 0; snode < stencilSize; ++snode) {
         for (int dofi = 0; dofi < DofsPerNode; ++dofi) {
-          const int layer = startLayer + snode;
+          const int layer      = startLayer + snode;
           const int AmatBlkRow = LineLayer2NodeHost(line, layer);
-          const int AmatRow = AmatBlkRow * DofsPerNode + dofi;
+          const int AmatRow    = AmatBlkRow * DofsPerNode + dofi;
           PptrHost(AmatRow + 1) += DofsPerNode;
         }
       }
@@ -504,28 +504,27 @@ void SemiCoarsenPFactory_kokkos<
       Kokkos::create_mirror_view(CLayerSNode2PptrOffset);
   for (int clayer = 0; clayer < NCLayers; ++clayer) {
     const int stencilSize = CLayer2StencilSizeHost(clayer);
-    const int startLayer = CLayer2StartLayerHost(clayer);
+    const int startLayer  = CLayer2StartLayerHost(clayer);
     for (int snode = 0; snode < stencilSize; ++snode) {
-      const int layer = startLayer + snode;
+      const int layer                           = startLayer + snode;
       CLayerSNode2PptrOffsetHost(clayer, snode) = layerBuckets(layer);
       layerBuckets(layer)++;
     }
   }
   Kokkos::deep_copy(CLayerSNode2PptrOffset, CLayerSNode2PptrOffsetHost);
 
-  { // Fill P - fill and solve each block tridiagonal system and fill P views
+  {  // Fill P - fill and solve each block tridiagonal system and fill P views
     SubFactoryMonitor m3(*this, "Fill P", coarseLevel);
 
     const auto localAmat = Amat->getLocalMatrixDevice();
-    const auto zero = impl_ATS::zero();
-    const auto one = impl_ATS::one();
+    const auto zero      = impl_ATS::zero();
+    const auto one       = impl_ATS::one();
 
     using range_policy = Kokkos::RangePolicy<execution_space>;
     Kokkos::parallel_for(
         "MueLu::SemiCoarsenPFactory_kokkos::BuildSemiCoarsenP Fill P",
         range_policy(0, NVertLines), KOKKOS_LAMBDA(const int line) {
           for (int clayer = 0; clayer < NCLayers; ++clayer) {
-
             // Initialize BandSol
             auto bandSol =
                 Kokkos::subview(BandSol, line, Kokkos::ALL(), Kokkos::ALL());
@@ -535,7 +534,7 @@ void SemiCoarsenPFactory_kokkos<
 
             // Initialize BandMat (set unused row diagonal to 1.0)
             const int stencilSize = CLayer2StencilSize(clayer);
-            const int N = stencilSize * DofsPerNode;
+            const int N           = stencilSize * DofsPerNode;
             auto bandMat =
                 Kokkos::subview(BandMat, line, Kokkos::ALL(), Kokkos::ALL());
             for (int row = 0; row < Nmax; ++row)
@@ -544,25 +543,23 @@ void SemiCoarsenPFactory_kokkos<
                     (row == col && row >= N) ? one : zero;
 
             // Loop over layers in stencil and fill banded matrix and rhs
-            const int flayer = CLayer2FLayer(clayer);
+            const int flayer     = CLayer2FLayer(clayer);
             const int startLayer = CLayer2StartLayer(clayer);
             for (int snode = 0; snode < stencilSize; ++snode) {
-
               const int layer = startLayer + snode;
-              if (layer == flayer) { // If layer in stencil is a coarse layer
+              if (layer == flayer) {  // If layer in stencil is a coarse layer
                 for (int dof = 0; dof < DofsPerNode; ++dof) {
-                  const int row = snode * DofsPerNode + dof;
+                  const int row     = snode * DofsPerNode + dof;
                   bandMat(row, row) = one;
                   bandSol(row, dof) = one;
                 }
-              } else { // Not a coarse layer
+              } else {  // Not a coarse layer
                 const int AmatBlkRow = LineLayer2Node(line, layer);
                 for (int dofi = 0; dofi < DofsPerNode; ++dofi) {
-
                   // Get Amat row info
-                  const int AmatRow = AmatBlkRow * DofsPerNode + dofi;
+                  const int AmatRow       = AmatBlkRow * DofsPerNode + dofi;
                   const auto localAmatRow = localAmat.rowConst(AmatRow);
-                  const int AmatRowLeng = localAmatRow.length;
+                  const int AmatRowLeng   = localAmatRow.length;
 
                   const int row = snode * DofsPerNode + dofi;
                   for (int dofj = 0; dofj < DofsPerNode; ++dofj) {
@@ -610,7 +607,7 @@ void SemiCoarsenPFactory_kokkos<
             }
 
             // Batch LU and triangular solves
-            namespace KB = KokkosBatched;
+            namespace KB  = KokkosBatched;
             using lu_type = typename KB::SerialLU<KB::Algo::LU::Unblocked>;
             lu_type::invoke(bandMat);
             using trsv_l_type =
@@ -627,16 +624,16 @@ void SemiCoarsenPFactory_kokkos<
             for (int snode = 0; snode < stencilSize; ++snode) {
               for (int dofj = 0; dofj < DofsPerNode; ++dofj) {
                 for (int dofi = 0; dofi < DofsPerNode; ++dofi) {
-                  const int layer = startLayer + snode;
+                  const int layer      = startLayer + snode;
                   const int AmatBlkRow = LineLayer2Node(line, layer);
-                  const int AmatRow = AmatBlkRow * DofsPerNode + dofi;
+                  const int AmatRow    = AmatBlkRow * DofsPerNode + dofi;
 
                   const int pptrOffset = CLayerSNode2PptrOffset(clayer, snode);
                   const int pptr =
                       Pptr(AmatRow) + pptrOffset * DofsPerNode + dofj;
 
                   const int col =
-                      line * NCLayers + clayer; // coarse node (block row) index
+                      line * NCLayers + clayer;  // coarse node (block row) index
                   Pcols(pptr) = col * DofsPerNode + dofj;
                   Pvals(pptr) = bandSol(snode * DofsPerNode + dofi, dofj);
                 }
@@ -644,17 +641,17 @@ void SemiCoarsenPFactory_kokkos<
             }
           }
         });
-  } // Fill P
+  }  // Fill P
 
   // Build P
-  RCP<const Map> rowMap = Amat->getRowMap();
+  RCP<const Map> rowMap        = Amat->getRowMap();
   Xpetra::global_size_t GNdofs = rowMap->getGlobalNumElements();
-  Xpetra::global_size_t itemp = GNdofs / NFLayers;
+  Xpetra::global_size_t itemp  = GNdofs / NFLayers;
   std::vector<size_t> stridingInfo_{(size_t)DofsPerNode};
   RCP<const Map> coarseMap = StridedMapFactory::Build(
       rowMap->lib(), NCLayers * itemp, NCLayers * NVertLines * DofsPerNode, 0,
       stridingInfo_, rowMap->getComm(), -1, 0);
-  P = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
+  P                   = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
   RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
   PCrs->setAllValues(Pptr, Pcols, Pvals);
   PCrs->expertStaticFillComplete(coarseMap, Amat->getDomainMap());
@@ -668,23 +665,23 @@ void SemiCoarsenPFactory_kokkos<
   // Construct coarse nullspace and inject fine nullspace
   coarseNullspace =
       MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
-  const int numVectors = fineNullspace->getNumVectors();
-  const auto fineNullspaceView = fineNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  const int numVectors           = fineNullspace->getNumVectors();
+  const auto fineNullspaceView   = fineNullspace->getDeviceLocalView(Xpetra::Access::ReadOnly);
   const auto coarseNullspaceView = coarseNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
-  using range_policy = Kokkos::RangePolicy<execution_space>;
+  using range_policy             = Kokkos::RangePolicy<execution_space>;
   Kokkos::parallel_for(
       "MueLu::SemiCoarsenPFactory_kokkos::BuildSemiCoarsenP Inject Nullspace",
       range_policy(0, NVertLines), KOKKOS_LAMBDA(const int line) {
         for (int clayer = 0; clayer < NCLayers; ++clayer) {
           const int layer = CLayer2FLayer(clayer);
           const int AmatBlkRow =
-              LineLayer2Node(line, layer); // fine node (block row) index
+              LineLayer2Node(line, layer);  // fine node (block row) index
           const int col =
-              line * NCLayers + clayer; // coarse node (block row) index
+              line * NCLayers + clayer;  // coarse node (block row) index
           for (int k = 0; k < numVectors; ++k) {
             for (int dofi = 0; dofi < DofsPerNode; ++dofi) {
-              const int fRow = AmatBlkRow * DofsPerNode + dofi;
-              const int cRow = col * DofsPerNode + dofi;
+              const int fRow               = AmatBlkRow * DofsPerNode + dofi;
+              const int cRow               = col * DofsPerNode + dofi;
               coarseNullspaceView(cRow, k) = fineNullspaceView(fRow, k);
             }
           }
@@ -692,7 +689,7 @@ void SemiCoarsenPFactory_kokkos<
       });
 }
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
-#endif // MUELU_SEMICOARSENPFACTORY_KOKKOS_DEF_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_decl.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_decl.hpp
index 3aaf5aa1340f..9f7c5ca7a021 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_decl.hpp
@@ -58,73 +58,74 @@ namespace MueLu {
   @brief Class for transferring coordinates from a finer level to a coarser one
 
 */
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class ToggleCoordinatesTransferFactory : public TwoLevelFactoryBase {
+          class Node          = DefaultNode>
+class ToggleCoordinatesTransferFactory : public TwoLevelFactoryBase {
 #undef MUELU_TOGGLECOORDINATESTRANSFERFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    /*! @brief Constructor.
+  /*! @brief Constructor.
 
-       @param vectorName The name of the quantity to be restricted.
-       @param restrictionName The name of the restriction Matrix.
+     @param vectorName The name of the quantity to be restricted.
+     @param restrictionName The name of the restriction Matrix.
 
-       The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
-       <tt>vectorName</tt>.
-    */
-    ToggleCoordinatesTransferFactory() : hasDeclaredInput_(false) { }
+     The operator associated with <tt>projectionName</tt> will be applied to the MultiVector associated with
+     <tt>vectorName</tt>.
+  */
+  ToggleCoordinatesTransferFactory()
+    : hasDeclaredInput_(false) {}
 
-    //! Destructor.
-    virtual ~ToggleCoordinatesTransferFactory() { }
+  //! Destructor.
+  virtual ~ToggleCoordinatesTransferFactory() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &finelevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &finelevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level & fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Get/Set functions
-    //@{
+  //! @name Get/Set functions
+  //@{
 
-    /*! @brief Add a coordinate transfer factory in the end of list of coordinate transfer factories */
-    void AddCoordTransferFactory(const RCP<const FactoryBase>& factory);
+  /*! @brief Add a coordinate transfer factory in the end of list of coordinate transfer factories */
+  void AddCoordTransferFactory(const RCP<const FactoryBase> &factory);
 
-    //! Returns number of coordinate transfer factories.
-    size_t NumCoordTransferFactories() const { return coordFacts_.size(); }
-    //@}
+  //! Returns number of coordinate transfer factories.
+  size_t NumCoordTransferFactories() const { return coordFacts_.size(); }
+  //@}
 
-  private:
-    //! list of user-defined transfer coordinate factories which provide coordinates on the coarse level!
-    mutable std::vector<RCP<const FactoryBase> > coordFacts_;
+ private:
+  //! list of user-defined transfer coordinate factories which provide coordinates on the coarse level!
+  mutable std::vector<RCP<const FactoryBase> > coordFacts_;
 
-    mutable bool hasDeclaredInput_;
-  }; // class ToggleCoordinatesTransferFactory
+  mutable bool hasDeclaredInput_;
+};  // class ToggleCoordinatesTransferFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TOGGLECOORDINATESTRANSFERFACTORY_SHORT
-#endif // MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DECL_HPP
+#endif  // MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_def.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_def.hpp
index 757414924e6d..5f798cf0571b 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_def.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_ToggleCoordinatesTransferFactory_def.hpp
@@ -46,7 +46,6 @@
 #ifndef MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DEF_HPP
 #define MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DEF_HPP
 
-
 #include "MueLu_ToggleCoordinatesTransferFactory_decl.hpp"
 
 #include "MueLu_Level.hpp"
@@ -54,71 +53,71 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("Chosen P", Teuchos::null, "Name of TogglePFactory this ToggleCoordinatesTransferFactory is connected to. Parameter provides information which execution path (prolongator) has been chosen.");
+  validParamList->set<RCP<const FactoryBase> >("Chosen P", Teuchos::null, "Name of TogglePFactory this ToggleCoordinatesTransferFactory is connected to. Parameter provides information which execution path (prolongator) has been chosen.");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
-    const ParameterList& pL = GetParameterList();
-    TEUCHOS_TEST_FOR_EXCEPTION(!pL.isParameter("Chosen P"), Exceptions::RuntimeError, "MueLu::ToggleCoordinatesTransferFactory::DeclareInput: You have to set the 'Chosen P' parameter to a factory name of type TogglePFactory. The ToggleCoordinatesTransferFactory must be used together with a TogglePFactory!");
-    Input(coarseLevel,"Chosen P");
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = coordFacts_.begin(); it != coordFacts_.end(); ++it) {
-      coarseLevel.DeclareInput("Coordinates", (*it).get(), this);  // request/release coarse coordinates
-      (*it)->CallDeclareInput(coarseLevel); // request dependencies
-    }
-    hasDeclaredInput_ = true;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& /* fineLevel */, Level& coarseLevel) const {
+  const ParameterList& pL = GetParameterList();
+  TEUCHOS_TEST_FOR_EXCEPTION(!pL.isParameter("Chosen P"), Exceptions::RuntimeError, "MueLu::ToggleCoordinatesTransferFactory::DeclareInput: You have to set the 'Chosen P' parameter to a factory name of type TogglePFactory. The ToggleCoordinatesTransferFactory must be used together with a TogglePFactory!");
+  Input(coarseLevel, "Chosen P");
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = coordFacts_.begin(); it != coordFacts_.end(); ++it) {
+    coarseLevel.DeclareInput("Coordinates", (*it).get(), this);  // request/release coarse coordinates
+    (*it)->CallDeclareInput(coarseLevel);                        // request dependencies
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & /* fineLevel */, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Coordinate transfer toggle", coarseLevel);
-
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> xdMV;
-
-    TEUCHOS_TEST_FOR_EXCEPTION(coordFacts_.size() != 2, Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: ToggleCoordinatesTransferFactory needs two different transfer operator strategies for toggling.");
-
-    int chosenP = Get< int >      (coarseLevel, "Chosen P");
-    GetOStream(Runtime1) << "Transfer Coordinates" << chosenP << " to coarse level" << std::endl;
-    RCP<xdMV> coarseCoords = coarseLevel.Get< RCP<xdMV> >("Coordinates",(coordFacts_[chosenP]).get());
-    Set(coarseLevel, "Coordinates", coarseCoords);
-
-    // loop through all coord facts and check whether the coarse coordinates are available.
-    // This is the coarse coordinate transfer factory which belongs to the execution path
-    // chosen by the TogglePFactory
-    /*RCP<xdMV> coarseCoords = Teuchos::null;
-    for(size_t t=0; t<coordFacts_.size(); ++t) {
-      bool bIsAv = coarseLevel.IsAvailable("Coordinates",(coordFacts_[t]).get());
-      std::cout << "Coordinates generated by " << (coordFacts_[t]).get() << " available? " << bIsAv << std::endl;
-      if ( coarseLevel.IsAvailable("Coordinates",(coordFacts_[t]).get()) ) {
-        GetOStream(Runtime1) << "Choose factory " << t << " (" << (coordFacts_[t]).get() << ")" << std::endl;
-        coarseCoords = coarseLevel.Get< RCP<xdMV> >("Coordinates",(coordFacts_[t]).get());
-        Set(coarseLevel, "Coordinates", coarseCoords);
-      }
-    }*/
-
-    // Release dependencies of all coordinate transfer factories
-    for(size_t t=0; t<coordFacts_.size(); ++t) {
-      coarseLevel.Release(*(coordFacts_[t]));
+  hasDeclaredInput_ = true;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& /* fineLevel */, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Coordinate transfer toggle", coarseLevel);
+
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
+
+  TEUCHOS_TEST_FOR_EXCEPTION(coordFacts_.size() != 2, Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: ToggleCoordinatesTransferFactory needs two different transfer operator strategies for toggling.");
+
+  int chosenP = Get<int>(coarseLevel, "Chosen P");
+  GetOStream(Runtime1) << "Transfer Coordinates" << chosenP << " to coarse level" << std::endl;
+  RCP<xdMV> coarseCoords = coarseLevel.Get<RCP<xdMV> >("Coordinates", (coordFacts_[chosenP]).get());
+  Set(coarseLevel, "Coordinates", coarseCoords);
+
+  // loop through all coord facts and check whether the coarse coordinates are available.
+  // This is the coarse coordinate transfer factory which belongs to the execution path
+  // chosen by the TogglePFactory
+  /*RCP<xdMV> coarseCoords = Teuchos::null;
+  for(size_t t=0; t<coordFacts_.size(); ++t) {
+    bool bIsAv = coarseLevel.IsAvailable("Coordinates",(coordFacts_[t]).get());
+    std::cout << "Coordinates generated by " << (coordFacts_[t]).get() << " available? " << bIsAv << std::endl;
+    if ( coarseLevel.IsAvailable("Coordinates",(coordFacts_[t]).get()) ) {
+      GetOStream(Runtime1) << "Choose factory " << t << " (" << (coordFacts_[t]).get() << ")" << std::endl;
+      coarseCoords = coarseLevel.Get< RCP<xdMV> >("Coordinates",(coordFacts_[t]).get());
+      Set(coarseLevel, "Coordinates", coarseCoords);
     }
+  }*/
 
-    //TODO: exception if coarseCoords == Teuchos::null
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddCoordTransferFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::ToggleCoordinatesTransferFactory::AddCoordTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. Make sure you provide the factory which generates the coarse coordinates. Usually this is a coordinate transfer factory."
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::ToggleCoordinatesTransferFactory::AddCoordTransferFactory: Factory is being added after we have already declared input");
-    coordFacts_.push_back(factory);
+  // Release dependencies of all coordinate transfer factories
+  for (size_t t = 0; t < coordFacts_.size(); ++t) {
+    coarseLevel.Release(*(coordFacts_[t]));
   }
-} // namespace MueLu
 
-#endif // MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DEF_HPP
+  // TODO: exception if coarseCoords == Teuchos::null
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ToggleCoordinatesTransferFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddCoordTransferFactory(const RCP<const FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::ToggleCoordinatesTransferFactory::AddCoordTransferFactory: Transfer factory is not derived from TwoLevelFactoryBase. Make sure you provide the factory which generates the coarse coordinates. Usually this is a coordinate transfer factory."
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::ToggleCoordinatesTransferFactory::AddCoordTransferFactory: Factory is being added after we have already declared input");
+  coordFacts_.push_back(factory);
+}
+}  // namespace MueLu
+
+#endif  // MUELU_TOGGLECOORDINATESTRANSFER_FACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_decl.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_decl.hpp
index d7c8080c8966..3b0be5094c52 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_decl.hpp
@@ -55,113 +55,113 @@
 
 namespace MueLu {
 
-  /*!
-    @class TogglePFactory
-    @ingroup MueLuTransferClasses
-    @brief Prolongator factory which allows switching between two different prolongator strategies.
-
-    ## Input/output of TogglePFactory ##
-
-    ### User parameters of TogglePFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | P         | Factory | null  |   | * | * | Generating factories of prolongators. Note that the TogglePFactory needs at least two different prolongator subfactories.
-    | Ptent     | Factory | null  |   | * | * | Generating factories of base prolongators. Note that the TogglePFactory needs at least two different base prolongator subfactories. Usually, the factory for Ptent is the same as for the coarse null space.
-    | Nullspace | Factory | null  |   | * | * | Generating factories for fine level null space information. Note, that you have to provide a null space information source factory for each prolongator factory.
-    | NumZLayers| Factory | NoFactory | | * | * | Factory which provides information about z layers. Usually it is stored with MueLu::NoFactory as generating factory. The data is generated by the SemiCoarsenPFactory.
-    | semicoarsen: number of levels | int | ? | * | * |  | number of levels which are reserved for semi-coarsening. Note that the TogglePFactory stops the semi-coarsening process either when we have more levels than declared in this parameter or as soon as the number of remaining z-levels is 1.
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TogglePFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see TogglePFactory::DeclareInput).
-
-    ### Variables provided by TogglePFactory ###
-
-    After TogglePFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | TogglePFactory | Prolongator
-    | Ptent   | TogglePFactory | Base prolongator (e.g. tentative prolongator)
-    | Nullspace | TogglePFactory | associated coarse null space
-
-  */
-
-template <class Scalar = DefaultScalar,
-        class LocalOrdinal = DefaultLocalOrdinal,
-        class GlobalOrdinal = DefaultGlobalOrdinal,
-        class Node = DefaultNode>
-  class TogglePFactory : public PFactory {
+/*!
+  @class TogglePFactory
+  @ingroup MueLuTransferClasses
+  @brief Prolongator factory which allows switching between two different prolongator strategies.
+
+  ## Input/output of TogglePFactory ##
+
+  ### User parameters of TogglePFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | P         | Factory | null  |   | * | * | Generating factories of prolongators. Note that the TogglePFactory needs at least two different prolongator subfactories.
+  | Ptent     | Factory | null  |   | * | * | Generating factories of base prolongators. Note that the TogglePFactory needs at least two different base prolongator subfactories. Usually, the factory for Ptent is the same as for the coarse null space.
+  | Nullspace | Factory | null  |   | * | * | Generating factories for fine level null space information. Note, that you have to provide a null space information source factory for each prolongator factory.
+  | NumZLayers| Factory | NoFactory | | * | * | Factory which provides information about z layers. Usually it is stored with MueLu::NoFactory as generating factory. The data is generated by the SemiCoarsenPFactory.
+  | semicoarsen: number of levels | int | ? | * | * |  | number of levels which are reserved for semi-coarsening. Note that the TogglePFactory stops the semi-coarsening process either when we have more levels than declared in this parameter or as soon as the number of remaining z-levels is 1.
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TogglePFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see TogglePFactory::DeclareInput).
+
+  ### Variables provided by TogglePFactory ###
+
+  After TogglePFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | TogglePFactory | Prolongator
+  | Ptent   | TogglePFactory | Base prolongator (e.g. tentative prolongator)
+  | Nullspace | TogglePFactory | associated coarse null space
+
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class TogglePFactory : public PFactory {
 #undef MUELU_TOGGLEPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  /*! @brief Constructor. */
+  TogglePFactory()
+    : hasDeclaredInput_(false) {}
 
-    /*! @brief Constructor. */
-    TogglePFactory() : hasDeclaredInput_(false) { }
+  //! Destructor.
+  virtual ~TogglePFactory() {}
 
-    //! Destructor.
-    virtual ~TogglePFactory() { }
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! Input
-    //@{
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //@}
+  //! @name Build methods.
+  //@{
 
-    //! @name Build methods.
-    //@{
+  /*!  @brief Build method.   */
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    /*!  @brief Build method.   */
-    void Build(Level& fineLevel, Level &coarseLevel) const;
+  void BuildP(Level & /* fineLevel */, Level & /* coarseLevel */) const {/* empty */};
+  //@}
 
-    void BuildP(Level &/* fineLevel */, Level &/* coarseLevel */) const { /* empty */ };
-    //@}
+  //@{
 
-    //@{
+  /*! @brief Add a prolongator factory in the end of list of prolongator factories.    */
+  void AddProlongatorFactory(const RCP<const FactoryBase> &factory);
 
-    /*! @brief Add a prolongator factory in the end of list of prolongator factories.    */
-    void AddProlongatorFactory(const RCP<const FactoryBase>& factory);
+  //! Returns number of prolongator factories.
+  size_t NumProlongatorFactories() const { return prolongatorFacts_.size(); }
 
-    //! Returns number of prolongator factories.
-    size_t NumProlongatorFactories() const { return prolongatorFacts_.size(); }
+  /*! @brief Add a tentative prolongator factory in the end of list of prolongator factories.    */
+  void AddPtentFactory(const RCP<const FactoryBase> &factory);
 
-    /*! @brief Add a tentative prolongator factory in the end of list of prolongator factories.    */
-    void AddPtentFactory(const RCP<const FactoryBase>& factory);
+  //! Returns number of tentative prolongator factories.
+  size_t NumPtentFactories() const { return ptentFacts_.size(); }
 
-    //! Returns number of tentative prolongator factories.
-    size_t NumPtentFactories() const { return ptentFacts_.size(); }
+  /*! @brief Add a coarse nullspace factory in the end of list of coarse nullspace factories.    */
+  void AddCoarseNullspaceFactory(const RCP<const FactoryBase> &factory);
 
-    /*! @brief Add a coarse nullspace factory in the end of list of coarse nullspace factories.    */
-    void AddCoarseNullspaceFactory(const RCP<const FactoryBase>& factory);
+  //! Returns number of coarse null space factories.
+  size_t NumCoarseNullspaceFactories() const { return prolongatorFacts_.size(); }
 
-    //! Returns number of coarse null space factories.
-    size_t NumCoarseNullspaceFactories() const { return prolongatorFacts_.size(); }
+  RCP<const FactoryBase> getProlongatorFactory(size_t t) const { return prolongatorFacts_[t]; }
+  //@}
+ private:
+  //! list of user-defined prolongation operator factories
+  mutable std::vector<RCP<const FactoryBase> > prolongatorFacts_;
 
-    RCP<const FactoryBase>  getProlongatorFactory(size_t t) const {return prolongatorFacts_[t]; }
-    //@}
-  private:
-    //! list of user-defined prolongation operator factories
-    mutable std::vector<RCP<const FactoryBase> > prolongatorFacts_;
+  //! list of user-defined tentative prolongation operator factories
+  mutable std::vector<RCP<const FactoryBase> > ptentFacts_;
 
-    //! list of user-defined tentative prolongation operator factories
-    mutable std::vector<RCP<const FactoryBase> > ptentFacts_;
+  //! list of user-defined nullspace factories (i.e. the prolongator factories which also generate the coarse level nullspace)
+  mutable std::vector<RCP<const FactoryBase> > nspFacts_;
 
-    //! list of user-defined nullspace factories (i.e. the prolongator factories which also generate the coarse level nullspace)
-    mutable std::vector<RCP<const FactoryBase> > nspFacts_;
+  mutable bool hasDeclaredInput_;
+};  // class TogglePFactory
 
-    mutable bool hasDeclaredInput_;
-  }; //class TogglePFactory
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TOGGLEPFACTORY_SHORT
-#endif // MUELU_TOGGLEPFACTORY_DECL_HPP
+#endif  // MUELU_TOGGLEPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_def.hpp b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_def.hpp
index 5d2d9c3287cd..9a782f847578 100644
--- a/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_def.hpp
+++ b/packages/muelu/src/Transfers/SemiCoarsen/MueLu_TogglePFactory_def.hpp
@@ -56,155 +56,158 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("toggle: mode");
-    SET_VALID_ENTRY("semicoarsen: number of levels");
-#undef  SET_VALID_ENTRY
-
-    return validParamList;
+  SET_VALID_ENTRY("toggle: mode");
+  SET_VALID_ENTRY("semicoarsen: number of levels");
+#undef SET_VALID_ENTRY
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  // request/release "P" and coarse level "Nullspace"
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = prolongatorFacts_.begin(); it != prolongatorFacts_.end(); ++it) {
+    coarseLevel.DeclareInput("P", (*it).get(), this);  // request/release "P" (dependencies are not affected)
+    coarseLevel.DeclareInput("RfromPfactory", (*it).get(), this);
+    (*it)->CallDeclareInput(coarseLevel);  // request dependencies
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    // request/release "P" and coarse level "Nullspace"
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = prolongatorFacts_.begin(); it != prolongatorFacts_.end(); ++it) {
-      coarseLevel.DeclareInput("P", (*it).get(), this); // request/release "P" (dependencies are not affected)
-      coarseLevel.DeclareInput("RfromPfactory", (*it).get(), this);
-      (*it)->CallDeclareInput(coarseLevel); // request dependencies
-    }
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = ptentFacts_.begin(); it != ptentFacts_.end(); ++it) {
-      coarseLevel.DeclareInput("P", (*it).get(), this); // request/release "Ptent" (dependencies are not affected)
-      (*it)->CallDeclareInput(coarseLevel); // request dependencies
-    }
-    for (std::vector<RCP<const FactoryBase> >::const_iterator it = nspFacts_.begin(); it != nspFacts_.end(); ++it) {
-      coarseLevel.DeclareInput("Nullspace", (*it).get(), this);  // request/release coarse "Nullspace" (dependencies are not affected)
-      (*it)->CallDeclareInput(coarseLevel); // request dependencies
-    }
-
-    // The factory needs the information about the number of z-layers. While this information is
-    // provided by the user for the finest level, the factory itself is responsible to provide the
-    // corresponding information on the coarser levels. Since a factory cannot be dependent on itself
-    // we use the NoFactory class as generator class, but remove the UserData keep flag, such that
-    // "NumZLayers" is part of the request/release mechanism.
-    // Please note, that this prevents us from having several (independent) CoarsePFactory instances!
-    // TODO: allow factory to dependent on self-generated data for TwoLevelFactories -> introduce ExpertRequest/Release in Level
-    fineLevel.DeclareInput("NumZLayers", NoFactory::get(), this);
-    fineLevel.RemoveKeepFlag("NumZLayers", NoFactory::get(), MueLu::UserData);
-
-    hasDeclaredInput_ = true;
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = ptentFacts_.begin(); it != ptentFacts_.end(); ++it) {
+    coarseLevel.DeclareInput("P", (*it).get(), this);  // request/release "Ptent" (dependencies are not affected)
+    (*it)->CallDeclareInput(coarseLevel);              // request dependencies
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Prolongator toggle", coarseLevel);
-    std::ostringstream levelstr;
-    levelstr << coarseLevel.GetLevelID();
-
-    TEUCHOS_TEST_FOR_EXCEPTION(nspFacts_.size() != prolongatorFacts_.size(), Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: The number of provided prolongator factories and coarse nullspace factories must be identical.");
-    TEUCHOS_TEST_FOR_EXCEPTION(nspFacts_.size() != 2, Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: TogglePFactory needs two different transfer operator strategies for toggling."); // TODO adapt this/weaken this as soon as other toggling strategies are introduced.
-
-    // decision routine which prolongator factory to be used
-    int nProlongatorFactory = 0; // default behavior: use first prolongator in list
-
-    // extract user parameters
-    const Teuchos::ParameterList & pL = GetParameterList();
-    std::string mode = Teuchos::as<std::string>(pL.get<std::string>("toggle: mode"));
-    int semicoarsen_levels = Teuchos::as<int>(pL.get<int>("semicoarsen: number of levels"));
-
-    TEUCHOS_TEST_FOR_EXCEPTION(mode!="semicoarsen", Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: The 'toggle: mode' parameter must be set to 'semicoarsen'. No other mode supported, yet.");
-
-    LO NumZDir = -1;
-    if(fineLevel.IsAvailable("NumZLayers", NoFactory::get())) {
-      NumZDir = fineLevel.Get<LO>("NumZLayers", NoFactory::get()); //obtain info
-      GetOStream(Runtime1) << "Number of layers for semicoarsening: " << NumZDir << std::endl;
-    }
-
-    // Make a decision which prolongator to be used.
-    if(fineLevel.GetLevelID() >= semicoarsen_levels || NumZDir == 1) {
-      nProlongatorFactory = 1;
-    } else {
-      nProlongatorFactory = 0;
-    }
-
-    RCP<Matrix> P     = Teuchos::null;
-    RCP<Matrix> Ptent = Teuchos::null;
-    RCP<MultiVector> coarseNullspace = Teuchos::null;
-
-    // call Build for selected transfer operator
-    GetOStream(Runtime0) << "TogglePFactory: call transfer factory: " << (prolongatorFacts_[nProlongatorFactory])->description() << std::endl;
-    prolongatorFacts_[nProlongatorFactory]->CallBuild(coarseLevel);
-    P = coarseLevel.Get< RCP<Matrix> >("P", (prolongatorFacts_[nProlongatorFactory]).get());
-    RCP<Matrix> R     = Teuchos::null;
-    int Rplaceholder = -1;   // Used to indicate that an R matrix has not been produced by a prolongator factory, but
-                             // that it is capable of producing one and should be invoked a 2nd time in restrictor mode
-                             // (e.g. with PgPFactory).  prolongatorFacts_[Rplaceholder] is factory that can produce R
-                             // matrix, which might be later invoked by  MueLu_RfromP_Or_TransP
-    if (coarseLevel.IsAvailable("RfromPfactory",(prolongatorFacts_[nProlongatorFactory]).get())) {
-	 std::string strType = coarseLevel.GetTypeName("RfromPfactory", (prolongatorFacts_[nProlongatorFactory]).get());
-	 if (strType == "int") Rplaceholder = nProlongatorFactory;
-	 else R = coarseLevel.Get< RCP<Matrix> >("RfromPfactory", (prolongatorFacts_[nProlongatorFactory]).get());
-	               // Need to get R (and set it below) so that TogglePFactory is given credit for creating R
-    }
-    // do not call "Build" for "Ptent" factory since it should automatically be called recursively
-    // through the "Build" call for "P"
-    Ptent = coarseLevel.Get< RCP<Matrix> >("P", (ptentFacts_[nProlongatorFactory]).get());
-    coarseNullspace = coarseLevel.Get< RCP<MultiVector> >("Nullspace", (nspFacts_[nProlongatorFactory]).get());
-
-    // Release dependencies of all prolongator and coarse level null spaces
-    for(size_t t=0; t<nspFacts_.size(); ++t) {
-      coarseLevel.Release(*(prolongatorFacts_[t]));
-      coarseLevel.Release(*(ptentFacts_[t]));
-      coarseLevel.Release(*(nspFacts_[t]));
-    }
-
-    // store prolongator with this factory identification.
-    Set(coarseLevel, "P", P);
-    //  Three cases:
-    //   1) R already computed and TogglePFactory takes credit for constructing it
-    //   2) R not computed but prolongatorFacts_[Rplaceholder] can produce it
-    //   3) R not computed  and prolongator can not produce it
-    if (R !=  Teuchos::null) Set(coarseLevel, "RfromPfactory", R);
-    else if (Rplaceholder !=  -1) Set(coarseLevel, "RfromPfactory",  Teuchos::as<int>(Rplaceholder));
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-    Set(coarseLevel, "Ptent", Ptent);
-    Set(coarseLevel, "Chosen P", nProlongatorFactory);
-  } //Build()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddProlongatorFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::TogglePFactory::AddProlongatorFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddProlongatorFactory: Factory is being added after we have already declared input");
-    prolongatorFacts_.push_back(factory);
+  for (std::vector<RCP<const FactoryBase> >::const_iterator it = nspFacts_.begin(); it != nspFacts_.end(); ++it) {
+    coarseLevel.DeclareInput("Nullspace", (*it).get(), this);  // request/release coarse "Nullspace" (dependencies are not affected)
+    (*it)->CallDeclareInput(coarseLevel);                      // request dependencies
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddPtentFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::TogglePFactory::AddPtentFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddPtentFactory: Factory is being added after we have already declared input");
-    ptentFacts_.push_back(factory);
+  // The factory needs the information about the number of z-layers. While this information is
+  // provided by the user for the finest level, the factory itself is responsible to provide the
+  // corresponding information on the coarser levels. Since a factory cannot be dependent on itself
+  // we use the NoFactory class as generator class, but remove the UserData keep flag, such that
+  // "NumZLayers" is part of the request/release mechanism.
+  // Please note, that this prevents us from having several (independent) CoarsePFactory instances!
+  // TODO: allow factory to dependent on self-generated data for TwoLevelFactories -> introduce ExpertRequest/Release in Level
+  fineLevel.DeclareInput("NumZLayers", NoFactory::get(), this);
+  fineLevel.RemoveKeepFlag("NumZLayers", NoFactory::get(), MueLu::UserData);
+
+  hasDeclaredInput_ = true;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Prolongator toggle", coarseLevel);
+  std::ostringstream levelstr;
+  levelstr << coarseLevel.GetLevelID();
+
+  TEUCHOS_TEST_FOR_EXCEPTION(nspFacts_.size() != prolongatorFacts_.size(), Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: The number of provided prolongator factories and coarse nullspace factories must be identical.");
+  TEUCHOS_TEST_FOR_EXCEPTION(nspFacts_.size() != 2, Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: TogglePFactory needs two different transfer operator strategies for toggling.");  // TODO adapt this/weaken this as soon as other toggling strategies are introduced.
+
+  // decision routine which prolongator factory to be used
+  int nProlongatorFactory = 0;  // default behavior: use first prolongator in list
+
+  // extract user parameters
+  const Teuchos::ParameterList& pL = GetParameterList();
+  std::string mode                 = Teuchos::as<std::string>(pL.get<std::string>("toggle: mode"));
+  int semicoarsen_levels           = Teuchos::as<int>(pL.get<int>("semicoarsen: number of levels"));
+
+  TEUCHOS_TEST_FOR_EXCEPTION(mode != "semicoarsen", Exceptions::RuntimeError, "MueLu::TogglePFactory::Build: The 'toggle: mode' parameter must be set to 'semicoarsen'. No other mode supported, yet.");
+
+  LO NumZDir = -1;
+  if (fineLevel.IsAvailable("NumZLayers", NoFactory::get())) {
+    NumZDir = fineLevel.Get<LO>("NumZLayers", NoFactory::get());  // obtain info
+    GetOStream(Runtime1) << "Number of layers for semicoarsening: " << NumZDir << std::endl;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddCoarseNullspaceFactory(const RCP<const FactoryBase>& factory) {
-    // check if it's a TwoLevelFactoryBase based transfer factory
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
-                               "MueLu::TogglePFactory::AddCoarseNullspaceFactory: Transfer factory is not derived from TwoLevelFactoryBase. Make sure you provide the factory which generates the coarse level nullspace information. Usually this is a prolongator factory."
-                               "This is very strange. (Note: you can remove this exception if there's a good reason for)");
-    TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddCoarseNullspaceFactory: Factory is being added after we have already declared input");
-    nspFacts_.push_back(factory);
+  // Make a decision which prolongator to be used.
+  if (fineLevel.GetLevelID() >= semicoarsen_levels || NumZDir == 1) {
+    nProlongatorFactory = 1;
+  } else {
+    nProlongatorFactory = 0;
   }
 
+  RCP<Matrix> P                    = Teuchos::null;
+  RCP<Matrix> Ptent                = Teuchos::null;
+  RCP<MultiVector> coarseNullspace = Teuchos::null;
+
+  // call Build for selected transfer operator
+  GetOStream(Runtime0) << "TogglePFactory: call transfer factory: " << (prolongatorFacts_[nProlongatorFactory])->description() << std::endl;
+  prolongatorFacts_[nProlongatorFactory]->CallBuild(coarseLevel);
+  P                = coarseLevel.Get<RCP<Matrix> >("P", (prolongatorFacts_[nProlongatorFactory]).get());
+  RCP<Matrix> R    = Teuchos::null;
+  int Rplaceholder = -1;  // Used to indicate that an R matrix has not been produced by a prolongator factory, but
+                          // that it is capable of producing one and should be invoked a 2nd time in restrictor mode
+                          // (e.g. with PgPFactory).  prolongatorFacts_[Rplaceholder] is factory that can produce R
+                          // matrix, which might be later invoked by  MueLu_RfromP_Or_TransP
+  if (coarseLevel.IsAvailable("RfromPfactory", (prolongatorFacts_[nProlongatorFactory]).get())) {
+    std::string strType = coarseLevel.GetTypeName("RfromPfactory", (prolongatorFacts_[nProlongatorFactory]).get());
+    if (strType == "int")
+      Rplaceholder = nProlongatorFactory;
+    else
+      R = coarseLevel.Get<RCP<Matrix> >("RfromPfactory", (prolongatorFacts_[nProlongatorFactory]).get());
+    // Need to get R (and set it below) so that TogglePFactory is given credit for creating R
+  }
+  // do not call "Build" for "Ptent" factory since it should automatically be called recursively
+  // through the "Build" call for "P"
+  Ptent           = coarseLevel.Get<RCP<Matrix> >("P", (ptentFacts_[nProlongatorFactory]).get());
+  coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", (nspFacts_[nProlongatorFactory]).get());
+
+  // Release dependencies of all prolongator and coarse level null spaces
+  for (size_t t = 0; t < nspFacts_.size(); ++t) {
+    coarseLevel.Release(*(prolongatorFacts_[t]));
+    coarseLevel.Release(*(ptentFacts_[t]));
+    coarseLevel.Release(*(nspFacts_[t]));
+  }
 
-} //namespace MueLu
-
-#endif // MUELU_TOGGLEPFACTORY_DEF_HPP
+  // store prolongator with this factory identification.
+  Set(coarseLevel, "P", P);
+  //  Three cases:
+  //   1) R already computed and TogglePFactory takes credit for constructing it
+  //   2) R not computed but prolongatorFacts_[Rplaceholder] can produce it
+  //   3) R not computed  and prolongator can not produce it
+  if (R != Teuchos::null)
+    Set(coarseLevel, "RfromPfactory", R);
+  else if (Rplaceholder != -1)
+    Set(coarseLevel, "RfromPfactory", Teuchos::as<int>(Rplaceholder));
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "Ptent", Ptent);
+  Set(coarseLevel, "Chosen P", nProlongatorFactory);
+}  // Build()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddProlongatorFactory(const RCP<const FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::TogglePFactory::AddProlongatorFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddProlongatorFactory: Factory is being added after we have already declared input");
+  prolongatorFacts_.push_back(factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddPtentFactory(const RCP<const FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::TogglePFactory::AddPtentFactory: Transfer factory is not derived from TwoLevelFactoryBase. "
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddPtentFactory: Factory is being added after we have already declared input");
+  ptentFacts_.push_back(factory);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TogglePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::AddCoarseNullspaceFactory(const RCP<const FactoryBase>& factory) {
+  // check if it's a TwoLevelFactoryBase based transfer factory
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const TwoLevelFactoryBase>(factory) == Teuchos::null, Exceptions::BadCast,
+                             "MueLu::TogglePFactory::AddCoarseNullspaceFactory: Transfer factory is not derived from TwoLevelFactoryBase. Make sure you provide the factory which generates the coarse level nullspace information. Usually this is a prolongator factory."
+                             "This is very strange. (Note: you can remove this exception if there's a good reason for)");
+  TEUCHOS_TEST_FOR_EXCEPTION(hasDeclaredInput_, Exceptions::RuntimeError, "MueLu::TogglePFactory::AddCoarseNullspaceFactory: Factory is being added after we have already declared input");
+  nspFacts_.push_back(factory);
+}
+
+}  // namespace MueLu
+
+#endif  // MUELU_TOGGLEPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_decl.hpp
index f9cee70fec85..afbfac2bcee4 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_decl.hpp
@@ -59,156 +59,152 @@
 
 namespace MueLu {
 
-  /*!
-    @class CoarseMapFactory class.
-    @brief Factory for generating coarse level map. Used by TentativePFactory.
+/*!
+  @class CoarseMapFactory class.
+  @brief Factory for generating coarse level map. Used by TentativePFactory.
 
-    Factory is used by TentativePFactory to generate the coarse level map (=domain map) of
-    the prolongation operator. The number of DOFs is calculated by the following formula
-    \f[
-    n_{Aggs}\cdot d_{blocksize}
-    \f]
-    where \f$ n_{Aggs} \f$ denotes the number of aggregates (generated by the AggregationFactory) and
-    \f$ d_{blocksize}\f$ denotes the number of DOFs per node on the coarse level.
-    For determing the blocksize, the @c CoarseMapFactory supports strided map information.
-    Per default @c stridedBlockId is -1 and the number of nullspace vectors (from @c "Nullspace" ) is used.
-    Otherwise, the block size is calculated using the striding information.
+  Factory is used by TentativePFactory to generate the coarse level map (=domain map) of
+  the prolongation operator. The number of DOFs is calculated by the following formula
+  \f[
+  n_{Aggs}\cdot d_{blocksize}
+  \f]
+  where \f$ n_{Aggs} \f$ denotes the number of aggregates (generated by the AggregationFactory) and
+  \f$ d_{blocksize}\f$ denotes the number of DOFs per node on the coarse level.
+  For determing the blocksize, the @c CoarseMapFactory supports strided map information.
+  Per default @c stridedBlockId is -1 and the number of nullspace vectors (from @c "Nullspace" ) is used.
+  Otherwise, the block size is calculated using the striding information.
 
-    By setting the "Domain GID offsets" parameter, we can specify the GID offset manually for each level.
-    If the number of entries in that list is smaller than the number of levels,
-    we use the default offset 0 for the remaining coarse levels.
+  By setting the "Domain GID offsets" parameter, we can specify the GID offset manually for each level.
+  If the number of entries in that list is smaller than the number of levels,
+  we use the default offset 0 for the remaining coarse levels.
 
-    @ingroup MueLuTransferClasses
+  @ingroup MueLuTransferClasses
 
-    ## Input/output ##
+  ## Input/output ##
 
-    ### User parameters of this factory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    Aggregates | Factory | null | | * | * | Generating factory for aggregates. We basically need only the number of aggregates in this factory.
-    Nullspace  | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
-    Striding info | std::string | {} |  | * | | String containing the striding information, e.g. {2,1} for a fixed block size of 3 with sub blocks of size 2 and 1 in each node.
-    Strided block id | LocalOrdinal | -1 |  | * | | Strided block id in strided information vector or -1 (=default) for the full block information.
-    Domain GID offsets | std::string | {0} |  | * | | String describing a std::vector with offsets for GIDs for each level. If no offset GID value is given for the level we use 0 as default.
+  ### User parameters of this factory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  Aggregates | Factory | null | | * | * | Generating factory for aggregates. We basically need only the number of aggregates in this factory.
+  Nullspace  | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+  Striding info | std::string | {} |  | * | | String containing the striding information, e.g. {2,1} for a fixed block size of 3 with sub blocks of size 2 and 1 in each node.
+  Strided block id | LocalOrdinal | -1 |  | * | | Strided block id in strided information vector or -1 (=default) for the full block information.
+  Domain GID offsets | std::string | {0} |  | * | | String describing a std::vector with offsets for GIDs for each level. If no offset GID value is given for the level we use 0 as default.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see @c GetValidParameters() ).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see @c DeclareInput() ).
 
-    \note To auto-compute the Domain GID offset on each level, consider to use the @c BlockedCoarseMapFactory .
+  \note To auto-compute the Domain GID offset on each level, consider to use the @c BlockedCoarseMapFactory .
 
-    ### Variables provided by this factory ###
+  ### Variables provided by this factory ###
 
-    After @c Build() the following data is available (if requested)
+  After @c Build() the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | CoarseMap | CoarseMapFactory | Map containing the coarse map used as domain map in the tentative prolongation operator
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | CoarseMap | CoarseMapFactory | Map containing the coarse map used as domain map in the tentative prolongation operator
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class CoarseMapFactory : public SingleLevelFactoryBase {
+          class Node          = DefaultNode>
+class CoarseMapFactory : public SingleLevelFactoryBase {
 #undef MUELU_COARSEMAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Input
-    //@{
-
-    RCP<const ParameterList> GetValidParameterList() const override;
-
-    /*!
-      @brief Specifies the data that this class needs, and the factories that generate that data.
-
-      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput,
-      then this class will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const override;
+ public:
+  //! @name Input
+  //@{
 
-    //@}
-
-    //! @name Build methods.
-    //@{
-
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const override;
-
-    //@}
-
-
-    /*! @brief Get information on the fixed block size
-     *
-     * Returns the full block size (number of DOFs per node) of the domain DOF map (= coarse map).
-     * This is the sum of all entries in the striding vector.
-     * e.g. for 2 velocity dofs and 1 pressure dof the return value is 3.
-     */
-    virtual size_t getFixedBlockSize() const {
-      // sum up size of all strided blocks (= number of dofs per node)
-      size_t blkSize = 0;
-      std::vector<size_t>::const_iterator it;
-      for(it = stridingInfo_.begin(); it != stridingInfo_.end(); ++it) {
-        blkSize += *it;
-      }
-      return blkSize;
-    }
+  RCP<const ParameterList> GetValidParameterList() const override;
 
-    //! @name Get/Set functions
-
-    /*! @brief getStridingData
-     * returns vector with size of striding blocks in the domain DOF map (= coarse map).
-     * e.g. for 2 velocity dofs and 1 pressure dof the vector is (2,1)
-     */
-    virtual std::vector<size_t> getStridingData() const { return stridingInfo_; }
-
-    /*! @brief setStridingData
-     * set striding vector for the domain DOF map (= coarse map),
-     * e.g. (2,1) for 2 velocity dofs and 1 pressure dof
-     */
-    virtual void setStridingData(std::vector<size_t> stridingInfo);
-
-    /*! @brief getStridedBlockId
-     * returns strided block id for the domain DOF map of Ptent (= coarse map)
-     * or -1 if full strided map is stored in the domain map of Ptent (= coarse map)
-     */
-    virtual LocalOrdinal getStridedBlockId() const {
-      const ParameterList & pL = GetParameterList();
-      return pL.get<LocalOrdinal>("Strided block id");
-    }
+  /*!
+    @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    /*! @brief setStridedBlockId
-     * set strided block id for the domain DOF map of Ptent (= coarse map)
-     * or -1 if full strided map is stored in the domain map of Ptent (= coarse map)
-     */
-    virtual void setStridedBlockId(LocalOrdinal stridedBlockId) {
-      SetParameter("Strided block id", ParameterEntry(stridedBlockId));
+    If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput,
+    then this class will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& currentLevel) const override;
+
+  //@}
+
+  //! @name Build methods.
+  //@{
+
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const override;
+
+  //@}
+
+  /*! @brief Get information on the fixed block size
+   *
+   * Returns the full block size (number of DOFs per node) of the domain DOF map (= coarse map).
+   * This is the sum of all entries in the striding vector.
+   * e.g. for 2 velocity dofs and 1 pressure dof the return value is 3.
+   */
+  virtual size_t getFixedBlockSize() const {
+    // sum up size of all strided blocks (= number of dofs per node)
+    size_t blkSize = 0;
+    std::vector<size_t>::const_iterator it;
+    for (it = stridingInfo_.begin(); it != stridingInfo_.end(); ++it) {
+      blkSize += *it;
     }
-
-    //@}
-
-  protected:
-
-    //! Build the coarse map using the domain GID offset
-    virtual void BuildCoarseMap(Level& currentLevel, const GlobalOrdinal domainGIDOffset) const;
-
-    //! Extract domain GID offset from user data
-    virtual GlobalOrdinal GetDomainGIDOffset(Level& currentLevel) const;
-
-  private:
-
-    virtual void CheckForConsistentStridingInformation(
-        LocalOrdinal stridedBlockId, const size_t nullspaceDimension) const;
-
-    //! Vector with size of strided blocks (dofs)
-    mutable std::vector<size_t> stridingInfo_;
-
-  }; //class CoarseMapFactory
-
-} //namespace MueLu
+    return blkSize;
+  }
+
+  //! @name Get/Set functions
+
+  /*! @brief getStridingData
+   * returns vector with size of striding blocks in the domain DOF map (= coarse map).
+   * e.g. for 2 velocity dofs and 1 pressure dof the vector is (2,1)
+   */
+  virtual std::vector<size_t> getStridingData() const { return stridingInfo_; }
+
+  /*! @brief setStridingData
+   * set striding vector for the domain DOF map (= coarse map),
+   * e.g. (2,1) for 2 velocity dofs and 1 pressure dof
+   */
+  virtual void setStridingData(std::vector<size_t> stridingInfo);
+
+  /*! @brief getStridedBlockId
+   * returns strided block id for the domain DOF map of Ptent (= coarse map)
+   * or -1 if full strided map is stored in the domain map of Ptent (= coarse map)
+   */
+  virtual LocalOrdinal getStridedBlockId() const {
+    const ParameterList& pL = GetParameterList();
+    return pL.get<LocalOrdinal>("Strided block id");
+  }
+
+  /*! @brief setStridedBlockId
+   * set strided block id for the domain DOF map of Ptent (= coarse map)
+   * or -1 if full strided map is stored in the domain map of Ptent (= coarse map)
+   */
+  virtual void setStridedBlockId(LocalOrdinal stridedBlockId) {
+    SetParameter("Strided block id", ParameterEntry(stridedBlockId));
+  }
+
+  //@}
+
+ protected:
+  //! Build the coarse map using the domain GID offset
+  virtual void BuildCoarseMap(Level& currentLevel, const GlobalOrdinal domainGIDOffset) const;
+
+  //! Extract domain GID offset from user data
+  virtual GlobalOrdinal GetDomainGIDOffset(Level& currentLevel) const;
+
+ private:
+  virtual void CheckForConsistentStridingInformation(
+      LocalOrdinal stridedBlockId, const size_t nullspaceDimension) const;
+
+  //! Vector with size of strided blocks (dofs)
+  mutable std::vector<size_t> stridingInfo_;
+
+};  // class CoarseMapFactory
+
+}  // namespace MueLu
 
 #define MUELU_COARSEMAPFACTORY_SHORT
 #endif /* MUELU_COARSEMAPFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_def.hpp
index 4b737762ec0a..f876e97de531 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_CoarseMapFactory_def.hpp
@@ -65,148 +65,141 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const
-  {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Generating factory for aggregates.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",  Teuchos::null, "Generating factory for null space.");
-
-    validParamList->set< std::string  >("Striding info", "{}", "Striding information");
-    validParamList->set< LocalOrdinal >("Strided block id", -1, "Strided block id");
-
-    // Domain GID offsets: list of offset values for domain gids (coarse gids) of tentative prolongator  (default = 0).
-    // For each multigrid level we can define a different offset value, ie for the prolongator between
-    // level 0 and level 1 we use the offset entry domainGidOffsets_[0], for the prolongator between
-    // level 1 and level 2 we use domainGidOffsets_[1]...
-    // If the vector domainGidOffsets_ is too short and does not contain a sufficient number of gid offset
-    // values for all levels, a gid offset of 0 is used for all the remaining levels!
-    // The GIDs for the domain dofs of Ptent start with domainGidOffset, are contiguous and distributed
-    // equally over the procs (unless some reordering is done).
-    validParamList->set< std::string > ("Domain GID offsets", "{0}", "vector with offsets for GIDs for each level. If no offset GID value is given for the level we use 0 as default.");
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const
-  {
-    Input(currentLevel, "Aggregates");
-    Input(currentLevel, "Nullspace");
-  }
-
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setStridingData(std::vector<size_t> stridingInfo)
-  {
-    // store striding map in internal variable
-    stridingInfo_ = stridingInfo;
-
-    // try to remove string "Striding info" from parameter list to make sure,
-    // that stridingInfo_ is not replaced in the Build call.
-    std::string strStridingInfo; strStridingInfo.clear();
-    SetParameter("Striding info", ParameterEntry(strStridingInfo));
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const
-  {
-    FactoryMonitor m(*this, "Build", currentLevel);
-
-    GlobalOrdinal domainGIDOffset = GetDomainGIDOffset(currentLevel);
-    BuildCoarseMap(currentLevel, domainGIDOffset);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCoarseMap(
-    Level& currentLevel, const GlobalOrdinal domainGIDOffset) const
-  {
-
-    RCP<Aggregates> aggregates = Get< RCP<Aggregates> >(currentLevel, "Aggregates");
-    GlobalOrdinal numAggs = aggregates->GetNumAggregates();
-    RCP<const Map> aggMap = aggregates->GetMap();
-
-    RCP<MultiVector> nullspace = Get< RCP<MultiVector> >(currentLevel, "Nullspace");
-
-    const size_t NSDim = nullspace->getNumVectors();
-    RCP<const Teuchos::Comm<int> > comm = aggMap->getComm();
-    const ParameterList & pL = GetParameterList();
-
-    LocalOrdinal stridedBlockId = pL.get<LocalOrdinal>("Strided block id");
-
-    // read in stridingInfo from parameter list and fill the internal member variable
-    // read the data only if the parameter "Striding info" exists and is non-empty
-    if(pL.isParameter("Striding info")) {
-      std::string strStridingInfo = pL.get<std::string>("Striding info");
-      if(strStridingInfo.empty() == false) {
-        Teuchos::Array<size_t> arrayVal = Teuchos::fromStringToArray<size_t>(strStridingInfo);
-        stridingInfo_ = Teuchos::createVector(arrayVal);
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Generating factory for aggregates.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory for null space.");
+
+  validParamList->set<std::string>("Striding info", "{}", "Striding information");
+  validParamList->set<LocalOrdinal>("Strided block id", -1, "Strided block id");
+
+  // Domain GID offsets: list of offset values for domain gids (coarse gids) of tentative prolongator  (default = 0).
+  // For each multigrid level we can define a different offset value, ie for the prolongator between
+  // level 0 and level 1 we use the offset entry domainGidOffsets_[0], for the prolongator between
+  // level 1 and level 2 we use domainGidOffsets_[1]...
+  // If the vector domainGidOffsets_ is too short and does not contain a sufficient number of gid offset
+  // values for all levels, a gid offset of 0 is used for all the remaining levels!
+  // The GIDs for the domain dofs of Ptent start with domainGidOffset, are contiguous and distributed
+  // equally over the procs (unless some reordering is done).
+  validParamList->set<std::string>("Domain GID offsets", "{0}", "vector with offsets for GIDs for each level. If no offset GID value is given for the level we use 0 as default.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  Input(currentLevel, "Aggregates");
+  Input(currentLevel, "Nullspace");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::setStridingData(std::vector<size_t> stridingInfo) {
+  // store striding map in internal variable
+  stridingInfo_ = stridingInfo;
+
+  // try to remove string "Striding info" from parameter list to make sure,
+  // that stridingInfo_ is not replaced in the Build call.
+  std::string strStridingInfo;
+  strStridingInfo.clear();
+  SetParameter("Striding info", ParameterEntry(strStridingInfo));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  GlobalOrdinal domainGIDOffset = GetDomainGIDOffset(currentLevel);
+  BuildCoarseMap(currentLevel, domainGIDOffset);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildCoarseMap(
+    Level& currentLevel, const GlobalOrdinal domainGIDOffset) const {
+  RCP<Aggregates> aggregates = Get<RCP<Aggregates> >(currentLevel, "Aggregates");
+  GlobalOrdinal numAggs      = aggregates->GetNumAggregates();
+  RCP<const Map> aggMap      = aggregates->GetMap();
+
+  RCP<MultiVector> nullspace = Get<RCP<MultiVector> >(currentLevel, "Nullspace");
+
+  const size_t NSDim                  = nullspace->getNumVectors();
+  RCP<const Teuchos::Comm<int> > comm = aggMap->getComm();
+  const ParameterList& pL             = GetParameterList();
+
+  LocalOrdinal stridedBlockId = pL.get<LocalOrdinal>("Strided block id");
+
+  // read in stridingInfo from parameter list and fill the internal member variable
+  // read the data only if the parameter "Striding info" exists and is non-empty
+  if (pL.isParameter("Striding info")) {
+    std::string strStridingInfo = pL.get<std::string>("Striding info");
+    if (strStridingInfo.empty() == false) {
+      Teuchos::Array<size_t> arrayVal = Teuchos::fromStringToArray<size_t>(strStridingInfo);
+      stridingInfo_                   = Teuchos::createVector(arrayVal);
     }
-
-    CheckForConsistentStridingInformation(stridedBlockId, NSDim);
-
-    GetOStream(Statistics2) << "domainGIDOffset: " << domainGIDOffset << " block size: " << getFixedBlockSize() << " stridedBlockId: " << stridedBlockId << std::endl;
-
-    // number of coarse level dofs (fixed by number of aggregates and blocksize data)
-    GlobalOrdinal nCoarseDofs = numAggs * getFixedBlockSize();
-    GlobalOrdinal indexBase = aggMap->getIndexBase();
-
-    RCP<const Map> coarseMap = StridedMapFactory::Build(aggMap->lib(),
-        Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-        nCoarseDofs,
-        indexBase,
-        stridingInfo_,
-        comm,
-        stridedBlockId,
-        domainGIDOffset);
-
-    Set(currentLevel, "CoarseMap", coarseMap);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDomainGIDOffset(
-    Level& currentLevel) const
-  {
-    GlobalOrdinal domainGidOffset = Teuchos::ScalarTraits<GlobalOrdinal>::zero();
-
-    std::vector<GlobalOrdinal> domainGidOffsets;
-    domainGidOffsets.clear();
-    const ParameterList & pL = GetParameterList();
-    if(pL.isParameter("Domain GID offsets")) {
-      std::string strDomainGIDs = pL.get<std::string>("Domain GID offsets");
-      if(strDomainGIDs.empty() == false) {
-        Teuchos::Array<GlobalOrdinal> arrayVal = Teuchos::fromStringToArray<GlobalOrdinal>(strDomainGIDs);
-        domainGidOffsets = Teuchos::createVector(arrayVal);
-        if(currentLevel.GetLevelID() < Teuchos::as<int>(domainGidOffsets.size()) ) {
-          domainGidOffset = domainGidOffsets[currentLevel.GetLevelID()];
-        }
+  CheckForConsistentStridingInformation(stridedBlockId, NSDim);
+
+  GetOStream(Statistics2) << "domainGIDOffset: " << domainGIDOffset << " block size: " << getFixedBlockSize() << " stridedBlockId: " << stridedBlockId << std::endl;
+
+  // number of coarse level dofs (fixed by number of aggregates and blocksize data)
+  GlobalOrdinal nCoarseDofs = numAggs * getFixedBlockSize();
+  GlobalOrdinal indexBase   = aggMap->getIndexBase();
+
+  RCP<const Map> coarseMap = StridedMapFactory::Build(aggMap->lib(),
+                                                      Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                      nCoarseDofs,
+                                                      indexBase,
+                                                      stridingInfo_,
+                                                      comm,
+                                                      stridedBlockId,
+                                                      domainGIDOffset);
+
+  Set(currentLevel, "CoarseMap", coarseMap);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetDomainGIDOffset(
+    Level& currentLevel) const {
+  GlobalOrdinal domainGidOffset = Teuchos::ScalarTraits<GlobalOrdinal>::zero();
+
+  std::vector<GlobalOrdinal> domainGidOffsets;
+  domainGidOffsets.clear();
+  const ParameterList& pL = GetParameterList();
+  if (pL.isParameter("Domain GID offsets")) {
+    std::string strDomainGIDs = pL.get<std::string>("Domain GID offsets");
+    if (strDomainGIDs.empty() == false) {
+      Teuchos::Array<GlobalOrdinal> arrayVal = Teuchos::fromStringToArray<GlobalOrdinal>(strDomainGIDs);
+      domainGidOffsets                       = Teuchos::createVector(arrayVal);
+      if (currentLevel.GetLevelID() < Teuchos::as<int>(domainGidOffsets.size())) {
+        domainGidOffset = domainGidOffsets[currentLevel.GetLevelID()];
       }
     }
-
-    return domainGidOffset;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckForConsistentStridingInformation(
-    const LocalOrdinal stridedBlockId, const size_t nullspaceDimension) const
-  {
-    // check for consistency of striding information with NSDim and nCoarseDofs
-    if (stridedBlockId == -1) {
-      // this means we have no real strided map but only a block map with constant blockSize "nullspaceDimension"
-      TEUCHOS_TEST_FOR_EXCEPTION(stridingInfo_.size() > 1, Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): stridingInfo_.size() but must be one");
-      stridingInfo_.clear();
-      stridingInfo_.push_back(nullspaceDimension);
-      TEUCHOS_TEST_FOR_EXCEPTION(stridingInfo_.size() != 1, Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): stridingInfo_.size() but must be one");
-
-    } else {
-      // stridedBlockId > -1, set by user
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedBlockId > Teuchos::as<LO>(stridingInfo_.size() - 1), Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): it is stridingInfo_.size() <= stridedBlockId_. error.");
-      size_t stridedBlockSize = stridingInfo_[stridedBlockId];
-      TEUCHOS_TEST_FOR_EXCEPTION(stridedBlockSize != nullspaceDimension , Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): dimension of strided block != nullspaceDimension. error.");
-    }
+  return domainGidOffset;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseMapFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckForConsistentStridingInformation(
+    const LocalOrdinal stridedBlockId, const size_t nullspaceDimension) const {
+  // check for consistency of striding information with NSDim and nCoarseDofs
+  if (stridedBlockId == -1) {
+    // this means we have no real strided map but only a block map with constant blockSize "nullspaceDimension"
+    TEUCHOS_TEST_FOR_EXCEPTION(stridingInfo_.size() > 1, Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): stridingInfo_.size() but must be one");
+    stridingInfo_.clear();
+    stridingInfo_.push_back(nullspaceDimension);
+    TEUCHOS_TEST_FOR_EXCEPTION(stridingInfo_.size() != 1, Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): stridingInfo_.size() but must be one");
+
+  } else {
+    // stridedBlockId > -1, set by user
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedBlockId > Teuchos::as<LO>(stridingInfo_.size() - 1), Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): it is stridingInfo_.size() <= stridedBlockId_. error.");
+    size_t stridedBlockSize = stridingInfo_[stridedBlockId];
+    TEUCHOS_TEST_FOR_EXCEPTION(stridedBlockSize != nullspaceDimension, Exceptions::RuntimeError, "MueLu::CoarseMapFactory::Build(): dimension of strided block != nullspaceDimension. error.");
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_COARSEMAPFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_decl.hpp
index 3ca71cd3a534..217ad2785fcf 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_decl.hpp
@@ -58,107 +58,106 @@
 
 namespace MueLu {
 
-  /*!
-     @class NullspaceFactory class.
-     @brief Factory for generating nullspace
+/*!
+   @class NullspaceFactory class.
+   @brief Factory for generating nullspace
 
-      The NullspaceFactory is meant to generate a default approximation for
-      the fine level nullspace (Level 0 only). For all other levels it is used
-      only to act as "generating factory" for the "Nullspace", which is actually
-      handled by the TentativePFactory.
+    The NullspaceFactory is meant to generate a default approximation for
+    the fine level nullspace (Level 0 only). For all other levels it is used
+    only to act as "generating factory" for the "Nullspace", which is actually
+    handled by the TentativePFactory.
 
-      Constructor:
-      \code{.cpp}
-      NullspaceFactory(const std::string & nspName = "Nullspace")
-      \endcode
-      This constructor uses the variable with the variable nspName on the finest level
-      as null space for the finest multigrid level.
+    Constructor:
+    \code{.cpp}
+    NullspaceFactory(const std::string & nspName = "Nullspace")
+    \endcode
+    This constructor uses the variable with the variable nspName on the finest level
+    as null space for the finest multigrid level.
 
-      Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
-      in case of multiphysics problems, "Nullspace1" and "Nullspace2"
+    Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
+    in case of multiphysics problems, "Nullspace1" and "Nullspace2"
 
-     @ingroup MueLuTransferClasses
+   @ingroup MueLuTransferClasses
 
-    ## Input/output of NullspaceFactory ##
+  ## Input/output of NullspaceFactory ##
 
-    ### User parameters of NullspaceFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
-     A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and NullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+  ### User parameters of NullspaceFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
+   A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and NullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NullspaceFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see NullspaceFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NullspaceFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see NullspaceFactory::DeclareInput).
 
-    ### Variables provided by NullspaceFactory ###
+  ### Variables provided by NullspaceFactory ###
 
-    After NullspaceFactory::Build the following data is available (if requested)
+  After NullspaceFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Nullspace | NullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors generated by the TentativePFactory (generating factory of "Nullspace") is provided as variable "Nullspace" in the level class. Note that the output variable is always "Nullspace" even if the parameter "Fine level nullspace" is different than "Nullspace". This way, the user can use any name for the fine level null space vectors, but the factories which need "Nullspace" information are guaranteed to get a null space.
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Nullspace | NullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors generated by the TentativePFactory (generating factory of "Nullspace") is provided as variable "Nullspace" in the level class. Note that the output variable is always "Nullspace" even if the parameter "Fine level nullspace" is different than "Nullspace". This way, the user can use any name for the fine level null space vectors, but the factories which need "Nullspace" information are guaranteed to get a null space.
 
-  */
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class NullspaceFactory : public SingleLevelFactoryBase {
+          class Node          = DefaultNode>
+class NullspaceFactory : public SingleLevelFactoryBase {
 #undef MUELU_NULLSPACEFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    NullspaceFactory(const std::string & nspName = "Nullspace") {
-      SetParameter("Fine level nullspace", ParameterEntry(nspName));
-    }
+  //! Constructor
+  NullspaceFactory(const std::string &nspName = "Nullspace") {
+    SetParameter("Fine level nullspace", ParameterEntry(nspName));
+  }
 
-    //! Destructor
-    virtual ~NullspaceFactory() { }
+  //! Destructor
+  virtual ~NullspaceFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
+  //! @name Input
 
-    //@{
+  //@{
 
-    /*! @brief Define valid parameters for internal factory parameters */
-    RCP<const ParameterList> GetValidParameterList() const;
+  /*! @brief Define valid parameters for internal factory parameters */
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
 
-    void DeclareInput(Level &currentLevel) const;
+  void DeclareInput(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    //! Helper function to recursively fill BlockedMultiVector with default null space vectors
-    void fillNullspaceVector(const RCP<BlockedMultiVector>& nsp, LocalOrdinal numPDEs, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zvals, LocalOrdinal nullspaceDim, Scalar cx, Scalar cy, Scalar cz) const;
+ private:
+  //! Helper function to recursively fill BlockedMultiVector with default null space vectors
+  void fillNullspaceVector(const RCP<BlockedMultiVector> &nsp, LocalOrdinal numPDEs, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zvals, LocalOrdinal nullspaceDim, Scalar cx, Scalar cy, Scalar cz) const;
 
-    mutable bool calculateRotations_ = false;
+  mutable bool calculateRotations_ = false;
 
-  }; //class NullspaceFactory
+};  // class NullspaceFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_NULLSPACEFACTORY_SHORT
-#endif // MUELU_NULLSPACEFACTORY_DECL_HPP
+#endif  // MUELU_NULLSPACEFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_def.hpp
index 0f8ce846f340..2d8684c85335 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_def.hpp
@@ -58,256 +58,255 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("nullspace: calculate rotations");
-#undef  SET_VALID_ENTRY
-    validParamList->set< std::string >("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\"). For block matrices also \"Nullspace1\" to \"Nullspace9\" are accepted to describe the null space vectors for the (i,i) block (i=1..9).");
-
-    validParamList->set< RCP<const FactoryBase> >("A",                          Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",                  Teuchos::null, "Generating factory of the fine level null space");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",                Teuchos::null, "Generating factory of the coordinates");
-
-    // TODO not very elegant.
-    // 1/20/2016: we could add a sublist (e.g. "Nullspaces" which is excluded from parameter validation)
-    validParamList->set< RCP<const FactoryBase> >("Nullspace1", Teuchos::null, "Generating factory of the fine level null space associated with the (1,1) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace2", Teuchos::null, "Generating factory of the fine level null space associated with the (2,2) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace3", Teuchos::null, "Generating factory of the fine level null space associated with the (3,3) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace4", Teuchos::null, "Generating factory of the fine level null space associated with the (4,4) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace5", Teuchos::null, "Generating factory of the fine level null space associated with the (5,5) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace6", Teuchos::null, "Generating factory of the fine level null space associated with the (6,6) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace7", Teuchos::null, "Generating factory of the fine level null space associated with the (7,7) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace8", Teuchos::null, "Generating factory of the fine level null space associated with the (8,8) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace9", Teuchos::null, "Generating factory of the fine level null space associated with the (9,9) block in your n x n block matrix.");
-
-    return validParamList;
+  SET_VALID_ENTRY("nullspace: calculate rotations");
+#undef SET_VALID_ENTRY
+  validParamList->set<std::string>("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\"). For block matrices also \"Nullspace1\" to \"Nullspace9\" are accepted to describe the null space vectors for the (i,i) block (i=1..9).");
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the fine level null space");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory of the coordinates");
+
+  // TODO not very elegant.
+  // 1/20/2016: we could add a sublist (e.g. "Nullspaces" which is excluded from parameter validation)
+  validParamList->set<RCP<const FactoryBase> >("Nullspace1", Teuchos::null, "Generating factory of the fine level null space associated with the (1,1) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace2", Teuchos::null, "Generating factory of the fine level null space associated with the (2,2) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace3", Teuchos::null, "Generating factory of the fine level null space associated with the (3,3) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace4", Teuchos::null, "Generating factory of the fine level null space associated with the (4,4) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace5", Teuchos::null, "Generating factory of the fine level null space associated with the (5,5) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace6", Teuchos::null, "Generating factory of the fine level null space associated with the (6,6) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace7", Teuchos::null, "Generating factory of the fine level null space associated with the (7,7) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace8", Teuchos::null, "Generating factory of the fine level null space associated with the (8,8) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace9", Teuchos::null, "Generating factory of the fine level null space associated with the (9,9) block in your n x n block matrix.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  const ParameterList &pL = GetParameterList();
+  std::string nspName     = pL.get<std::string>("Fine level nullspace");
+
+  // only request "A" in DeclareInput if
+  // 1) there is not nspName (e.g. "Nullspace") is available in Level, AND
+  // 2) it is the finest level (i.e. LevelID == 0)
+  if (currentLevel.IsAvailable(nspName, NoFactory::get()) == false && currentLevel.GetLevelID() == 0)
+    Input(currentLevel, "A");
+
+  if (currentLevel.GetLevelID() == 0 &&
+      currentLevel.IsAvailable("Coordinates", NoFactory::get()) &&  // we have coordinates (provided by user app)
+      pL.get<bool>("nullspace: calculate rotations")) {             // and we want to calculate rotation modes
+    calculateRotations_ = true;
+    Input(currentLevel, "Coordinates");
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-
-    const ParameterList & pL = GetParameterList();
-    std::string nspName = pL.get<std::string>("Fine level nullspace");
-
-    // only request "A" in DeclareInput if
-    // 1) there is not nspName (e.g. "Nullspace") is available in Level, AND
-    // 2) it is the finest level (i.e. LevelID == 0)
-    if (currentLevel.IsAvailable(nspName, NoFactory::get()) == false && currentLevel.GetLevelID() == 0)
-      Input(currentLevel, "A");
-
-    if( currentLevel.GetLevelID() == 0 &&
-        currentLevel.IsAvailable("Coordinates", NoFactory::get()) &&     // we have coordinates (provided by user app)
-        pL.get<bool>("nullspace: calculate rotations") ) {               // and we want to calculate rotation modes
-      calculateRotations_ = true;
-      Input(currentLevel, "Coordinates");
-    }
-
-    if (currentLevel.GetLevelID() != 0) {
-      // validate nullspaceFact_
-      // 1) The factory for "Nullspace" (or nspName) must not be Teuchos::null, since the default factory
-      //    for "Nullspace" is a NullspaceFactory
-      // 2) The factory for "Nullspace" (or nspName) must be a TentativePFactory or any other TwoLevelFactoryBase derived object
-      //    which generates the variable "Nullspace" as output
-      TEUCHOS_TEST_FOR_EXCEPTION(GetFactory(nspName)==Teuchos::null, Exceptions::RuntimeError, "MueLu::NullspaceFactory::DeclareInput(): You must declare an existing factory which produces the variable \"Nullspace\" in the NullspaceFactory (e.g. a TentativePFactory).");
-      currentLevel.DeclareInput("Nullspace", GetFactory(nspName).get(), this); /* ! "Nullspace" and nspName mismatch possible here */
-    }
+  if (currentLevel.GetLevelID() != 0) {
+    // validate nullspaceFact_
+    // 1) The factory for "Nullspace" (or nspName) must not be Teuchos::null, since the default factory
+    //    for "Nullspace" is a NullspaceFactory
+    // 2) The factory for "Nullspace" (or nspName) must be a TentativePFactory or any other TwoLevelFactoryBase derived object
+    //    which generates the variable "Nullspace" as output
+    TEUCHOS_TEST_FOR_EXCEPTION(GetFactory(nspName) == Teuchos::null, Exceptions::RuntimeError, "MueLu::NullspaceFactory::DeclareInput(): You must declare an existing factory which produces the variable \"Nullspace\" in the NullspaceFactory (e.g. a TentativePFactory).");
+    currentLevel.DeclareInput("Nullspace", GetFactory(nspName).get(), this); /* ! "Nullspace" and nspName mismatch possible here */
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Nullspace factory", currentLevel);
+
+  RCP<MultiVector> nullspace;
+
+  // TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): NullspaceFactory can be used for finest level (LevelID == 0) only.");
+  const ParameterList &pL = GetParameterList();
+  std::string nspName     = pL.get<std::string>("Fine level nullspace");
+
+  // get coordinates and compute mean of coordinates. (or centroid).
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
+  typedef Xpetra::MultiVector<coordinate_type, LO, GO, NO> RealValuedMultiVector;
+  RCP<RealValuedMultiVector> Coords;
+  ArrayRCP<const coordinate_type> xvals, yvals, zvals;
+  Scalar cx, cy, cz;
+
+  cx = 0.;
+  cy = 0.;
+  cz = 0.;
+  if (calculateRotations_) {
+    Coords = Get<RCP<RealValuedMultiVector> >(currentLevel, "Coordinates");
+
+    cx = Coords->getVector(0)->meanValue();
+    if (Coords->getNumVectors() > 1)
+      cy = Coords->getVector(1)->meanValue();
+    if (Coords->getNumVectors() > 2)
+      cz = Coords->getVector(2)->meanValue();
+
+    xvals = Coords->getData(0);
+    if (Coords->getNumVectors() > 1)
+      yvals = Coords->getData(1);
+    if (Coords->getNumVectors() > 2)
+      zvals = Coords->getData(2);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Nullspace factory", currentLevel);
-
-    RCP<MultiVector> nullspace;
-
-    //TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): NullspaceFactory can be used for finest level (LevelID == 0) only.");
-    const ParameterList & pL = GetParameterList();
-    std::string nspName = pL.get<std::string>("Fine level nullspace");
-
-    // get coordinates and compute mean of coordinates. (or centroid).
-    
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
-    typedef Xpetra::MultiVector<coordinate_type,LO,GO,NO> RealValuedMultiVector;
-    RCP<RealValuedMultiVector> Coords;
-    ArrayRCP<const coordinate_type> xvals, yvals, zvals;
-    Scalar cx, cy, cz;
-
-    cx = 0.; cy = 0.; cz = 0.;
-    if(calculateRotations_) {
-      Coords = Get< RCP<RealValuedMultiVector> >(currentLevel, "Coordinates");
-
-      cx     = Coords->getVector(0)->meanValue();
-      if (Coords->getNumVectors() > 1)
-        cy    = Coords->getVector(1)->meanValue();
-      if (Coords->getNumVectors() > 2)
-        cz    = Coords->getVector(2)->meanValue();
-
-      xvals  = Coords->getData(0);
-      if (Coords->getNumVectors() > 1)
-        yvals = Coords->getData(1);
-      if (Coords->getNumVectors() > 2)
-        zvals = Coords->getData(2);
-    }
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable(nspName, NoFactory::get())) {
+      // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
+      // Set("Nullspace1", ...), we use it.
+      nullspace = currentLevel.Get<RCP<MultiVector> >(nspName, NoFactory::get());
+      GetOStream(Runtime1) << "Use user-given nullspace " << nspName << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
+    } else {
+      // "Nullspace" (nspName) is not available
+      RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+
+      // determine numPDEs
+      LocalOrdinal numPDEs = 1;
+      if (A->IsView("stridedMaps") == true) {
+        Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+        TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap()) == Teuchos::null, Exceptions::BadCast, "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
+        numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
+        oldView = A->SwitchToView(oldView);
+      }
 
-    if (currentLevel.GetLevelID() == 0) {
+      LO nullspaceDim = numPDEs;
 
-      if (currentLevel.IsAvailable(nspName, NoFactory::get())) {
-        // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
-        // Set("Nullspace1", ...), we use it.
-        nullspace = currentLevel.Get< RCP<MultiVector> >(nspName, NoFactory::get());
-        GetOStream(Runtime1) << "Use user-given nullspace " << nspName << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
-      } else {
-        // "Nullspace" (nspName) is not available
-        RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-
-        // determine numPDEs
-        LocalOrdinal numPDEs = 1;
-        if(A->IsView("stridedMaps")==true) {
-          Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-          TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap()) == Teuchos::null, Exceptions::BadCast, "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
-          numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
-          oldView = A->SwitchToView(oldView);
-        }
+      if (calculateRotations_) {
+        if (Coords->getNumVectors() > 1) nullspaceDim++;
+        if (Coords->getNumVectors() > 2) nullspaceDim += 2;
+        GetOStream(Runtime1) << "Generating nullspace with rotations: dimension = " << nullspaceDim << std::endl;
+      } else
+        GetOStream(Runtime1) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
 
-        LO nullspaceDim = numPDEs;
+      nullspace = MultiVectorFactory::Build(A->getDomainMap(), nullspaceDim);
 
-        if(calculateRotations_) {
-          if (Coords->getNumVectors() > 1)  nullspaceDim++;
-          if (Coords->getNumVectors() > 2)  nullspaceDim += 2;
-          GetOStream(Runtime1) << "Generating nullspace with rotations: dimension = " << nullspaceDim << std::endl;
-        }
-        else GetOStream(Runtime1) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
-
-        nullspace = MultiVectorFactory::Build(A->getDomainMap(), nullspaceDim);
-
-        RCP<BlockedMultiVector> bnsp = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(nullspace);
-        if(bnsp.is_null() == true) {
-          for (int i=0; i<numPDEs; ++i) {
-            ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-            int numBlocks = nsValues.size() / numPDEs;
-            for (int j=0; j< numBlocks; ++j) {
-              nsValues[j*numPDEs + i] = 1.0;
-            }
+      RCP<BlockedMultiVector> bnsp = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(nullspace);
+      if (bnsp.is_null() == true) {
+        for (int i = 0; i < numPDEs; ++i) {
+          ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
+          int numBlocks             = nsValues.size() / numPDEs;
+          for (int j = 0; j < numBlocks; ++j) {
+            nsValues[j * numPDEs + i] = 1.0;
           }
-          if (( (int) nullspaceDim > numPDEs ) && ((int) numPDEs > 1)) {
-            /* xy rotation */
-            ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(numPDEs);
-            int numBlocks = nsValues.size() / numPDEs;
-            TEUCHOS_TEST_FOR_EXCEPTION(numBlocks != (int) xvals.size(), Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): number of coordinates does not match  ndofs/numPDEs.");
-            for (int j=0; j< numBlocks; ++j) {
-              nsValues[j*numPDEs + 0] = -(yvals[j]-cy);
-              nsValues[j*numPDEs + 1] =  (xvals[j]-cx);
-            }
+        }
+        if (((int)nullspaceDim > numPDEs) && ((int)numPDEs > 1)) {
+          /* xy rotation */
+          ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(numPDEs);
+          int numBlocks             = nsValues.size() / numPDEs;
+          TEUCHOS_TEST_FOR_EXCEPTION(numBlocks != (int)xvals.size(), Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): number of coordinates does not match  ndofs/numPDEs.");
+          for (int j = 0; j < numBlocks; ++j) {
+            nsValues[j * numPDEs + 0] = -(yvals[j] - cy);
+            nsValues[j * numPDEs + 1] = (xvals[j] - cx);
           }
-          if (( (int) nullspaceDim == numPDEs+3 )  && ((int) numPDEs > 2)) {
-            /* xz rotation */
-            ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(numPDEs+1);
-            int numBlocks = nsValues.size() / numPDEs;
-            for (int j=0; j< numBlocks; ++j) {
-              nsValues[j*numPDEs + 1] = -(zvals[j]-cz);
-              nsValues[j*numPDEs + 2] =  (yvals[j]-cy);
-            }
-            /* yz rotation */
-            nsValues = nullspace->getDataNonConst(numPDEs+2);
-            numBlocks = nsValues.size() / numPDEs;
-            for (int j=0; j< numBlocks; ++j) {
-              nsValues[j*numPDEs + 0] = -(zvals[j]-cz);
-              nsValues[j*numPDEs + 2] =  (xvals[j]-cx);
-            }
+        }
+        if (((int)nullspaceDim == numPDEs + 3) && ((int)numPDEs > 2)) {
+          /* xz rotation */
+          ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(numPDEs + 1);
+          int numBlocks             = nsValues.size() / numPDEs;
+          for (int j = 0; j < numBlocks; ++j) {
+            nsValues[j * numPDEs + 1] = -(zvals[j] - cz);
+            nsValues[j * numPDEs + 2] = (yvals[j] - cy);
           }
-           /*
-          // Scale columns to match what Galeri does. Not sure how important this is  as QR factorization
-          // when creating ptent should take care of scaling issues ... but leaving it just in case.
-           if ( (int) nullspaceDim > numPDEs ) {
-            Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms2(nullspaceDim);
-            nullspace->norm2(norms2);
-            Teuchos::Array<Scalar> norms2scalar(nullspaceDim);
-            for (int i = 0; i < nullspaceDim; i++)
-              norms2scalar[i] = norms2[0] / norms2[i];
-            nullspace->scale(norms2scalar);
+          /* yz rotation */
+          nsValues  = nullspace->getDataNonConst(numPDEs + 2);
+          numBlocks = nsValues.size() / numPDEs;
+          for (int j = 0; j < numBlocks; ++j) {
+            nsValues[j * numPDEs + 0] = -(zvals[j] - cz);
+            nsValues[j * numPDEs + 2] = (xvals[j] - cx);
           }
-          */
-
-        } else {
-          fillNullspaceVector(bnsp,numPDEs, xvals, yvals, zvals, nullspaceDim, cx, cy, cz);
         }
-      } // end if "Nullspace" not available
-    } else {
-      // on coarser levels always use "Nullspace" as variable name, since it is expected by
-      // tentative P factory to be "Nullspace"
+        /*
+       // Scale columns to match what Galeri does. Not sure how important this is  as QR factorization
+       // when creating ptent should take care of scaling issues ... but leaving it just in case.
+        if ( (int) nullspaceDim > numPDEs ) {
+         Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms2(nullspaceDim);
+         nullspace->norm2(norms2);
+         Teuchos::Array<Scalar> norms2scalar(nullspaceDim);
+         for (int i = 0; i < nullspaceDim; i++)
+           norms2scalar[i] = norms2[0] / norms2[i];
+         nullspace->scale(norms2scalar);
+       }
+       */
 
-      nullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory(nspName).get()); /* ! "Nullspace" and nspName mismatch possible here */
+      } else {
+        fillNullspaceVector(bnsp, numPDEs, xvals, yvals, zvals, nullspaceDim, cx, cy, cz);
+      }
+    }  // end if "Nullspace" not available
+  } else {
+    // on coarser levels always use "Nullspace" as variable name, since it is expected by
+    // tentative P factory to be "Nullspace"
 
-    }
+    nullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", GetFactory(nspName).get()); /* ! "Nullspace" and nspName mismatch possible here */
+  }
 
-    // provide "Nullspace" variable on current level (used by TentativePFactory)
-    Set(currentLevel, "Nullspace", nullspace);
+  // provide "Nullspace" variable on current level (used by TentativePFactory)
+  Set(currentLevel, "Nullspace", nullspace);
 
-  } // Build
+}  // Build
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::fillNullspaceVector(const RCP<BlockedMultiVector>& nsp, LocalOrdinal numPDEs, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zvals, LocalOrdinal nullspaceDim, Scalar cx, Scalar cy, Scalar cz) const {
-    RCP< const BlockedMap> bmap = nsp->getBlockedMap();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void NullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::fillNullspaceVector(const RCP<BlockedMultiVector> &nsp, LocalOrdinal numPDEs, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yvals, ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zvals, LocalOrdinal nullspaceDim, Scalar cx, Scalar cy, Scalar cz) const {
+  RCP<const BlockedMap> bmap = nsp->getBlockedMap();
 
-    for(size_t r = 0; r < bmap->getNumMaps(); r++) {
-      Teuchos::RCP<MultiVector> part = nsp->getMultiVector(r);
-      Teuchos::RCP<BlockedMultiVector> bpart = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(part);
-      if(bpart.is_null() == true) {
-        for (int i=0; i<numPDEs; ++i) {
-          ArrayRCP<Scalar> nsValues = part->getDataNonConst(i);
-          int numBlocks = nsValues.size() / numPDEs;
-          for (int j=0; j< numBlocks; ++j) {
-            nsValues[j*numPDEs + i] = 1.0;
-          }
+  for (size_t r = 0; r < bmap->getNumMaps(); r++) {
+    Teuchos::RCP<MultiVector> part         = nsp->getMultiVector(r);
+    Teuchos::RCP<BlockedMultiVector> bpart = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(part);
+    if (bpart.is_null() == true) {
+      for (int i = 0; i < numPDEs; ++i) {
+        ArrayRCP<Scalar> nsValues = part->getDataNonConst(i);
+        int numBlocks             = nsValues.size() / numPDEs;
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + i] = 1.0;
+        }
+      }
+      if ((int)nullspaceDim > numPDEs) {
+        /* xy rotation */
+        ArrayRCP<Scalar> nsValues = part->getDataNonConst(numPDEs);
+        int numBlocks             = nsValues.size() / numPDEs;
+        TEUCHOS_TEST_FOR_EXCEPTION(numBlocks != (int)xvals.size(), Exceptions::RuntimeError, "MueLu::NullspaceFactory::fillNullspaceVector(): number of coordinates does not match  ndofs/numPDEs.");
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + 0] = -(yvals[j] - cy);
+          nsValues[j * numPDEs + 1] = (xvals[j] - cx);
         }
-        if ( (int) nullspaceDim > numPDEs ) {
-           /* xy rotation */
-           ArrayRCP<Scalar> nsValues = part->getDataNonConst(numPDEs);
-           int numBlocks = nsValues.size() / numPDEs;
-           TEUCHOS_TEST_FOR_EXCEPTION(numBlocks != (int) xvals.size(), Exceptions::RuntimeError, "MueLu::NullspaceFactory::fillNullspaceVector(): number of coordinates does not match  ndofs/numPDEs.");
-           for (int j=0; j< numBlocks; ++j) {
-             nsValues[j*numPDEs + 0] = -(yvals[j]-cy);
-             nsValues[j*numPDEs + 1] =  (xvals[j]-cx);
-           }
-         }
-         if ( (int) nullspaceDim == numPDEs+3 ) {
-           /* xz rotation */
-           ArrayRCP<Scalar> nsValues = part->getDataNonConst(numPDEs+1);
-           int numBlocks = nsValues.size() / numPDEs;
-           for (int j=0; j< numBlocks; ++j) {
-             nsValues[j*numPDEs + 1] = -(zvals[j]-cz);
-             nsValues[j*numPDEs + 2] =  (yvals[j]-cy);
-           }
-           /* yz rotation */
-           nsValues = part->getDataNonConst(numPDEs+2);
-           numBlocks = nsValues.size() / numPDEs;
-           for (int j=0; j< numBlocks; ++j) {
-             nsValues[j*numPDEs + 0] = -(zvals[j]-cz);
-             nsValues[j*numPDEs + 2] =  (xvals[j]-cx);
-           }
-         }
-         /*
-          // Scale columns to match what Galeri does. Not sure that this is necessary as the qr factorizatoin
-          // of the tentative prolongator also takes care of scaling issues. I'm leaving the code here
-          // just in case.
-          if ( (int) nullspaceDim > numPDEs ) {
-            Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms2(nullspaceDim);
-            nullspace->norm2(norms2);
-            Teuchos::Array<Scalar> norms2scalar(nullspaceDim);
-            for (int i = 0; i < nullspaceDim; i++)
-              norms2scalar[i] = norms2[0] / norms2[i];
-            nullspace->scale(norms2scalar);
-          }
-          */
-      } else {
-        // call this routine recursively
-        fillNullspaceVector(bpart,numPDEs,xvals, yvals, zvals, nullspaceDim, cx, cy, cz);
       }
+      if ((int)nullspaceDim == numPDEs + 3) {
+        /* xz rotation */
+        ArrayRCP<Scalar> nsValues = part->getDataNonConst(numPDEs + 1);
+        int numBlocks             = nsValues.size() / numPDEs;
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + 1] = -(zvals[j] - cz);
+          nsValues[j * numPDEs + 2] = (yvals[j] - cy);
+        }
+        /* yz rotation */
+        nsValues  = part->getDataNonConst(numPDEs + 2);
+        numBlocks = nsValues.size() / numPDEs;
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + 0] = -(zvals[j] - cz);
+          nsValues[j * numPDEs + 2] = (xvals[j] - cx);
+        }
+      }
+      /*
+       // Scale columns to match what Galeri does. Not sure that this is necessary as the qr factorizatoin
+       // of the tentative prolongator also takes care of scaling issues. I'm leaving the code here
+       // just in case.
+       if ( (int) nullspaceDim > numPDEs ) {
+         Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms2(nullspaceDim);
+         nullspace->norm2(norms2);
+         Teuchos::Array<Scalar> norms2scalar(nullspaceDim);
+         for (int i = 0; i < nullspaceDim; i++)
+           norms2scalar[i] = norms2[0] / norms2[i];
+         nullspace->scale(norms2scalar);
+       }
+       */
+    } else {
+      // call this routine recursively
+      fillNullspaceVector(bpart, numPDEs, xvals, yvals, zvals, nullspaceDim, cx, cy, cz);
     }
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_NULLSPACEFACTORY_DEF_HPP
+#endif  // MUELU_NULLSPACEFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_decl.hpp
index 990596a44d50..2fb52ff69e24 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_decl.hpp
@@ -55,110 +55,108 @@
 
 namespace MueLu {
 
-  /*!
-     @class NullspaceFactory class.
-     @brief Factory for generating nullspace
-
-      The NullspaceFactory is meant to generate a default approximation for
-      the fine level nullspace (Level 0 only). For all other levels it is used
-      only to act as "generating factory" for the "Nullspace", which is actually
-      handled by the TentativePFactory.
-
-      Constructor:
-      \code{.cpp}
-      NullspaceFactory(const std::string & nspName = "Nullspace")
-      \endcode
-      This constructor uses the variable with the variable nspName on the finest level
-      as null space for the finest multigrid level.
-
-      Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
-      in case of multiphysics problems, "Nullspace1" and "Nullspace2"
-
-     @ingroup MueLuTransferClasses
-
-    ## Input/output of NullspaceFactory ##
-
-    ### User parameters of NullspaceFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
-     A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and NullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NullspaceFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see NullspaceFactory::DeclareInput).
-
-    ### Variables provided by NullspaceFactory ###
-
-    After NullspaceFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Nullspace | NullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors generated by the TentativePFactory (generating factory of "Nullspace") is provided as variable "Nullspace" in the level class. Note that the output variable is always "Nullspace" even if the parameter "Fine level nullspace" is different than "Nullspace". This way, the user can use any name for the fine level null space vectors, but the factories which need "Nullspace" information are guaranteed to get a null space.
-
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NullspaceFactory_kokkos;
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public SingleLevelFactoryBase {
-  public:
-    typedef LocalOrdinal                                             local_ordinal_type;
-    typedef GlobalOrdinal                                            global_ordinal_type;
-    typedef typename DeviceType::execution_space                     execution_space;
-    typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
-    typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>      node_type;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+   @class NullspaceFactory class.
+   @brief Factory for generating nullspace
+
+    The NullspaceFactory is meant to generate a default approximation for
+    the fine level nullspace (Level 0 only). For all other levels it is used
+    only to act as "generating factory" for the "Nullspace", which is actually
+    handled by the TentativePFactory.
+
+    Constructor:
+    \code{.cpp}
+    NullspaceFactory(const std::string & nspName = "Nullspace")
+    \endcode
+    This constructor uses the variable with the variable nspName on the finest level
+    as null space for the finest multigrid level.
+
+    Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
+    in case of multiphysics problems, "Nullspace1" and "Nullspace2"
+
+   @ingroup MueLuTransferClasses
+
+  ## Input/output of NullspaceFactory ##
+
+  ### User parameters of NullspaceFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
+   A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and NullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see NullspaceFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see NullspaceFactory::DeclareInput).
+
+  ### Variables provided by NullspaceFactory ###
+
+  After NullspaceFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Nullspace | NullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors generated by the TentativePFactory (generating factory of "Nullspace") is provided as variable "Nullspace" in the level class. Note that the output variable is always "Nullspace" even if the parameter "Fine level nullspace" is different than "Nullspace". This way, the user can use any name for the fine level null space vectors, but the factories which need "Nullspace" information are guaranteed to get a null space.
+
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NullspaceFactory_kokkos;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public SingleLevelFactoryBase {
+ public:
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename DeviceType::execution_space execution_space;
+  typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
+  typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_NULLSPACEFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! @name Constructors/Destructors.
-    //@{
+  //! Constructor
+  NullspaceFactory_kokkos(const std::string& nspName = "Nullspace") {
+    SetParameter("Fine level nullspace", ParameterEntry(nspName));
+  }
 
-    //! Constructor
-    NullspaceFactory_kokkos(const std::string& nspName = "Nullspace") {
-      SetParameter("Fine level nullspace", ParameterEntry(nspName));
-    }
+  //! Destructor
+  virtual ~NullspaceFactory_kokkos() {}
 
-    //! Destructor
-    virtual ~NullspaceFactory_kokkos() { }
+  //@}
 
-    //@}
+  //! @name Input
 
-    //! @name Input
+  //@{
 
-    //@{
+  /*! @brief Define valid parameters for internal factory parameters */
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    /*! @brief Define valid parameters for internal factory parameters */
-    RCP<const ParameterList> GetValidParameterList() const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
-
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-
-    void DeclareInput(Level& currentLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
 
-    //@}
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level& currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-  };
+  //@}
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_NULLSPACEFACTORY_KOKKOS_SHORT
-#endif // MUELU_NULLSPACEFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_NULLSPACEFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_def.hpp
index 03a3a8de4756..339b25d21bbe 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_NullspaceFactory_kokkos_def.hpp
@@ -56,131 +56,130 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  RCP<const ParameterList> NullspaceFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-
-    validParamList->set< RCP<const FactoryBase> >("A",          Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",  Teuchos::null, "Generating factory of the fine level null space");
-
-    // TODO not very elegant.
-    validParamList->set< std::string >("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\"). For block matrices also \"Nullspace1\" to \"Nullspace9\" are accepted to describe the null space vectors for the (i,i) block (i=1..9).");
-
-    // 1/20/2016: we could add a sublist (e.g. "Nullspaces" which is excluded from parameter validation)
-    validParamList->set< RCP<const FactoryBase> >("Nullspace1", Teuchos::null, "Generating factory of the fine level null space associated with the (1,1) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace2", Teuchos::null, "Generating factory of the fine level null space associated with the (2,2) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace3", Teuchos::null, "Generating factory of the fine level null space associated with the (3,3) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace4", Teuchos::null, "Generating factory of the fine level null space associated with the (4,4) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace5", Teuchos::null, "Generating factory of the fine level null space associated with the (5,5) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace6", Teuchos::null, "Generating factory of the fine level null space associated with the (6,6) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace7", Teuchos::null, "Generating factory of the fine level null space associated with the (7,7) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace8", Teuchos::null, "Generating factory of the fine level null space associated with the (8,8) block in your n x n block matrix.");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace9", Teuchos::null, "Generating factory of the fine level null space associated with the (9,9) block in your n x n block matrix.");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+RCP<const ParameterList> NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Generating factory of the fine level null space");
+
+  // TODO not very elegant.
+  validParamList->set<std::string>("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\"). For block matrices also \"Nullspace1\" to \"Nullspace9\" are accepted to describe the null space vectors for the (i,i) block (i=1..9).");
+
+  // 1/20/2016: we could add a sublist (e.g. "Nullspaces" which is excluded from parameter validation)
+  validParamList->set<RCP<const FactoryBase>>("Nullspace1", Teuchos::null, "Generating factory of the fine level null space associated with the (1,1) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace2", Teuchos::null, "Generating factory of the fine level null space associated with the (2,2) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace3", Teuchos::null, "Generating factory of the fine level null space associated with the (3,3) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace4", Teuchos::null, "Generating factory of the fine level null space associated with the (4,4) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace5", Teuchos::null, "Generating factory of the fine level null space associated with the (5,5) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace6", Teuchos::null, "Generating factory of the fine level null space associated with the (6,6) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace7", Teuchos::null, "Generating factory of the fine level null space associated with the (7,7) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace8", Teuchos::null, "Generating factory of the fine level null space associated with the (8,8) block in your n x n block matrix.");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace9", Teuchos::null, "Generating factory of the fine level null space associated with the (9,9) block in your n x n block matrix.");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& currentLevel) const {
+  const ParameterList& pL = GetParameterList();
+  std::string nspName     = pL.get<std::string>("Fine level nullspace");
+
+  // only request "A" in DeclareInput if
+  // 1) there is not nspName (e.g. "Nullspace") is available in Level, AND
+  // 2) it is the finest level (i.e. LevelID == 0)
+  if (currentLevel.IsAvailable(nspName, NoFactory::get()) == false && currentLevel.GetLevelID() == 0)
+    Input(currentLevel, "A");
+
+  if (currentLevel.GetLevelID() != 0) {
+    // validate nullspaceFact_
+    // 1) The factory for "Nullspace" (or nspName) must not be Teuchos::null, since the default factory
+    //    for "Nullspace" is a NullspaceFactory
+    // 2) The factory for "Nullspace" (or nspName) must be a TentativePFactory or any other TwoLevelFactoryBase derived object
+    //    which generates the variable "Nullspace" as output
+    TEUCHOS_TEST_FOR_EXCEPTION(GetFactory(nspName).is_null(), Exceptions::RuntimeError,
+                               "MueLu::NullspaceFactory::DeclareInput(): You must declare an existing factory which "
+                               "produces the variable \"Nullspace\" in the NullspaceFactory (e.g. a TentativePFactory).");
+    currentLevel.DeclareInput("Nullspace", GetFactory(nspName).get(), this); /* ! "Nullspace" and nspName mismatch possible here */
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void NullspaceFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& currentLevel) const {
-    const ParameterList& pL = GetParameterList();
-    std::string nspName = pL.get<std::string>("Fine level nullspace");
-
-    // only request "A" in DeclareInput if
-    // 1) there is not nspName (e.g. "Nullspace") is available in Level, AND
-    // 2) it is the finest level (i.e. LevelID == 0)
-    if (currentLevel.IsAvailable(nspName, NoFactory::get()) == false && currentLevel.GetLevelID() == 0)
-      Input(currentLevel, "A");
-
-    if (currentLevel.GetLevelID() != 0) {
-      // validate nullspaceFact_
-      // 1) The factory for "Nullspace" (or nspName) must not be Teuchos::null, since the default factory
-      //    for "Nullspace" is a NullspaceFactory
-      // 2) The factory for "Nullspace" (or nspName) must be a TentativePFactory or any other TwoLevelFactoryBase derived object
-      //    which generates the variable "Nullspace" as output
-      TEUCHOS_TEST_FOR_EXCEPTION(GetFactory(nspName).is_null(), Exceptions::RuntimeError,
-        "MueLu::NullspaceFactory::DeclareInput(): You must declare an existing factory which "
-        "produces the variable \"Nullspace\" in the NullspaceFactory (e.g. a TentativePFactory).");
-      currentLevel.DeclareInput("Nullspace", GetFactory(nspName).get(), this); /* ! "Nullspace" and nspName mismatch possible here */
-    }
+}
+
+template <class NullspaceType, class LO>
+class NullspaceFunctor {
+ private:
+  NullspaceType nullspace;
+  LO numPDEs;
+  typedef typename NullspaceType::value_type SC;
+  typedef Kokkos::ArithTraits<SC> ATS;
+
+ public:
+  NullspaceFunctor(NullspaceType nullspace_, LO numPDEs_)
+    : nullspace(nullspace_)
+    , numPDEs(numPDEs_) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO j) const {
+    SC one = ATS::one();
+    for (LO i = 0; i < numPDEs; i++)
+      nullspace(j * numPDEs + i, i) = one;
   }
+};
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void NullspaceFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Nullspace factory", currentLevel);
+
+  RCP<MultiVector> nullspace;
+
+  // TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): NullspaceFactory can be used for finest level (LevelID == 0) only.");
+  const ParameterList& pL = GetParameterList();
+  std::string nspName     = pL.get<std::string>("Fine level nullspace");
+
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable(nspName, NoFactory::get())) {
+      // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
+      // Set("Nullspace1", ...), we use it.
+      nullspace = currentLevel.Get<RCP<MultiVector>>(nspName, NoFactory::get());
+      GetOStream(Runtime1) << "Use user-given nullspace " << nspName << ":"
+                           << " nullspace dimension=" << nullspace->getNumVectors()
+                           << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
 
-  template<class NullspaceType, class LO>
-  class NullspaceFunctor {
-  private:
-    NullspaceType                               nullspace;
-    LO                                          numPDEs;
-    typedef typename NullspaceType::value_type  SC;
-    typedef Kokkos::ArithTraits<SC>             ATS;
-
-  public:
-    NullspaceFunctor(NullspaceType nullspace_, LO numPDEs_) :
-      nullspace(nullspace_),
-      numPDEs(numPDEs_)
-    { }
-
-    KOKKOS_INLINE_FUNCTION
-    void operator()(const LO j) const {
-      SC one = ATS::one();
-      for  (LO i = 0; i < numPDEs; i++)
-        nullspace(j*numPDEs + i, i) = one;
-    }
-  };
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void NullspaceFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& currentLevel) const {
-    FactoryMonitor m(*this, "Nullspace factory", currentLevel);
-
-    RCP<MultiVector> nullspace;
-
-    //TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::NullspaceFactory::Build(): NullspaceFactory can be used for finest level (LevelID == 0) only.");
-    const ParameterList& pL = GetParameterList();
-    std::string nspName = pL.get<std::string>("Fine level nullspace");
-
-    if (currentLevel.GetLevelID() == 0) {
-      if (currentLevel.IsAvailable(nspName, NoFactory::get())) {
-        // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
-        // Set("Nullspace1", ...), we use it.
-        nullspace = currentLevel.Get< RCP<MultiVector> >(nspName, NoFactory::get());
-        GetOStream(Runtime1) << "Use user-given nullspace " << nspName << ":"
-            << " nullspace dimension=" << nullspace->getNumVectors()
-            << " nullspace length="    << nullspace->getGlobalLength() << std::endl;
-
-      } else {
-        // "Nullspace" (nspName) is not available
-        auto A = Get< RCP<Matrix> >(currentLevel, "A");
-
-        // Determine numPDEs
-        LO numPDEs = 1;
-        if (A->IsView("stridedMaps") == true) {
-          Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-          TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<const StridedMap>(A->getRowMap()).is_null(), Exceptions::BadCast,
-                                     "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
-          numPDEs = rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
-          oldView = A->SwitchToView(oldView);
-        }
-
-        GetOStream(Runtime1) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
-
-        nullspace = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
-        auto nullspaceView = nullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
-
-        int numBlocks = nullspace->getLocalLength() / numPDEs;
-
-        NullspaceFunctor<decltype(nullspaceView), LO> nullspaceFunctor(nullspaceView, numPDEs);
-        Kokkos::parallel_for("MueLu:NullspaceF:Build:for", range_type(0,numBlocks), nullspaceFunctor);
-        // TODO extend null space factory for blocked multi vectors
+    } else {
+      // "Nullspace" (nspName) is not available
+      auto A = Get<RCP<Matrix>>(currentLevel, "A");
+
+      // Determine numPDEs
+      LO numPDEs = 1;
+      if (A->IsView("stridedMaps") == true) {
+        Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+        TEUCHOS_TEST_FOR_EXCEPTION(rcp_dynamic_cast<const StridedMap>(A->getRowMap()).is_null(), Exceptions::BadCast,
+                                   "MueLu::CoalesceFactory::Build: cast to strided row map failed.");
+        numPDEs = rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
+        oldView = A->SwitchToView(oldView);
       }
 
-    } else {
-      // On coarser levels always use "Nullspace" as variable name, since it is expected by
-      // tentative P factory to be "Nullspace"
-      nullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory(nspName).get()); // NOTE: "Nullspace" and nspName mismatch possible here
+      GetOStream(Runtime1) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
+
+      nullspace          = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
+      auto nullspaceView = nullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+
+      int numBlocks = nullspace->getLocalLength() / numPDEs;
+
+      NullspaceFunctor<decltype(nullspaceView), LO> nullspaceFunctor(nullspaceView, numPDEs);
+      Kokkos::parallel_for("MueLu:NullspaceF:Build:for", range_type(0, numBlocks), nullspaceFunctor);
+      // TODO extend null space factory for blocked multi vectors
     }
 
-    // provide "Nullspace" variable on current level (used by TentativePFactory)
-    Set(currentLevel, "Nullspace", nullspace);
+  } else {
+    // On coarser levels always use "Nullspace" as variable name, since it is expected by
+    // tentative P factory to be "Nullspace"
+    nullspace = currentLevel.Get<RCP<MultiVector>>("Nullspace", GetFactory(nspName).get());  // NOTE: "Nullspace" and nspName mismatch possible here
+  }
+
+  // provide "Nullspace" variable on current level (used by TentativePFactory)
+  Set(currentLevel, "Nullspace", nullspace);
 
-  } // Build
+}  // Build
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_NULLSPACEFACTORY_KOKKOS_DEF_HPP
+#endif  // MUELU_NULLSPACEFACTORY_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_decl.hpp
index 4b16d0b09c75..60c56ca02800 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_decl.hpp
@@ -66,86 +66,83 @@
 
 namespace MueLu {
 
-  /*!
-    @class ReitzingerPFactory class.
-    @brief Factory for building tentative prolongator.
+/*!
+  @class ReitzingerPFactory class.
+  @brief Factory for building tentative prolongator.
 
-    Factory for creating Reitzinger and Schoebel tentative prolongator for Maxwell's equations.
+  Factory for creating Reitzinger and Schoebel tentative prolongator for Maxwell's equations.
 
-    @ingroup MueLuTransferClasses
+  @ingroup MueLuTransferClasses
 
-    ## Input/output of ReitzingerPFactory ##
+  ## Input/output of ReitzingerPFactory ##
 
-    ### User parameters of ReitzingerPFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A        | Factory | null |   | * | * | Generating factory of the matrix A
-     P        | Factory | null |   | * | * | Generating factory of the nodal Prolongator
-     D0       | Factory | null |   | * | * | Generating factory of the discrete gradient operator
-     NodeMatrix  | Factory | null |   | * | * | Generating factory of the nodal A matrix
+  ### User parameters of ReitzingerPFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A        | Factory | null |   | * | * | Generating factory of the matrix A
+   P        | Factory | null |   | * | * | Generating factory of the nodal Prolongator
+   D0       | Factory | null |   | * | * | Generating factory of the discrete gradient operator
+   NodeMatrix  | Factory | null |   | * | * | Generating factory of the nodal A matrix
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ReitzingerPFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see ReitzingerPFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ReitzingerPFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see ReitzingerPFactory::DeclareInput).
 
-    ### Variables provided by ReitzingerPFactory ###
+  ### Variables provided by ReitzingerPFactory ###
 
-    After ReitzingerPFactory::Build the following data is available (if requested)
+  After ReitzingerPFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | ReitzingerPFactory   | Non-smoothed "tentative" Reitzinger edge prolongation operator
-    | D0      | ReitzingerPFactory   | Discrete gradient operator
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | ReitzingerPFactory   | Non-smoothed "tentative" Reitzinger edge prolongation operator
+  | D0      | ReitzingerPFactory   | Discrete gradient operator
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class ReitzingerPFactory : public PFactory {
+          class Node          = DefaultNode>
+class ReitzingerPFactory : public PFactory {
 #undef MUELU_REITZINGERPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    ReitzingerPFactory() { }
+  //! Constructor
+  ReitzingerPFactory() {}
 
-    //! Destructor.
-    virtual ~ReitzingerPFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~ReitzingerPFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! Utility method
-    //@{
-    void CheckCommutingProperty(const Matrix & Pe, const Matrix & D0_c, const Matrix& D0_f, const Matrix & Pn) const;
-    //@}
-
-  private:
-
-  }; //class ReitzingerPFactory
-
-} //namespace MueLu
+  //! Utility method
+  //@{
+  void CheckCommutingProperty(const Matrix& Pe, const Matrix& D0_c, const Matrix& D0_f, const Matrix& Pn) const;
+  //@}
 
+ private:
+};  // class ReitzingerPFactory
 
+}  // namespace MueLu
 
 #define MUELU_REITZINGERPFACTORY_SHORT
-#endif // MUELU_REITZINGERPFACTORY_DECL_HPP
+#endif  // MUELU_REITZINGERPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_def.hpp
index b36ac9bdebeb..bef431ea6969 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ReitzingerPFactory_def.hpp
@@ -70,289 +70,278 @@
 
 namespace MueLu {
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("repartition: enable");
-    SET_VALID_ENTRY("repartition: use subcommunicators");
-    SET_VALID_ENTRY("tentative: calculate qr"); 
-    SET_VALID_ENTRY("tentative: constant column sums");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("D0",                 Teuchos::null, "Generating factory of the matrix D0");
-    validParamList->set< RCP<const FactoryBase> >("NodeAggMatrix",      Teuchos::null, "Generating factory of the matrix NodeAggMatrix");
-    validParamList->set< RCP<const FactoryBase> >("Pnodal",             Teuchos::null, "Generating factory of the matrix P");
-    validParamList->set< RCP<const FactoryBase> >("NodeImporter",       Teuchos::null, "Generating factory of the matrix NodeImporter");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+  SET_VALID_ENTRY("repartition: enable");
+  SET_VALID_ENTRY("repartition: use subcommunicators");
+  SET_VALID_ENTRY("tentative: calculate qr");
+  SET_VALID_ENTRY("tentative: constant column sums");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("D0", Teuchos::null, "Generating factory of the matrix D0");
+  validParamList->set<RCP<const FactoryBase> >("NodeAggMatrix", Teuchos::null, "Generating factory of the matrix NodeAggMatrix");
+  validParamList->set<RCP<const FactoryBase> >("Pnodal", Teuchos::null, "Generating factory of the matrix P");
+  validParamList->set<RCP<const FactoryBase> >("NodeImporter", Teuchos::null, "Generating factory of the matrix NodeImporter");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "D0");
+  Input(fineLevel, "NodeAggMatrix");
+  Input(coarseLevel, "NodeAggMatrix");
+  Input(coarseLevel, "Pnodal");
+  //    Input(coarseLevel, "NodeImporter");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+  using Teuchos::arcp_const_cast;
+  using MT                    = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using XMM                   = Xpetra::MatrixMatrix<SC, LO, GO, NO>;
+  Teuchos::FancyOStream& out0 = GetBlackHole();
+  const ParameterList& pL     = GetParameterList();
+
+  bool update_communicators = pL.get<bool>("repartition: enable") && pL.get<bool>("repartition: use subcommunicators");
+
+  // If these are set correctly we assume that the nodal P contains only ones
+  bool nodal_p_is_all_ones = !pL.get<bool>("tentative: constant column sums") && !pL.get<bool>("tentative: calculate qr");
+
+  RCP<Matrix> EdgeMatrix = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<Matrix> D0         = Get<RCP<Matrix> >(fineLevel, "D0");
+  RCP<Matrix> NodeMatrix = Get<RCP<Matrix> >(fineLevel, "NodeAggMatrix");
+  RCP<Matrix> Pn         = Get<RCP<Matrix> >(coarseLevel, "Pnodal");
+
+  const GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
+  const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  // This needs to be an Operator because if NodeMatrix gets repartitioned away, we get an Operator on the level
+  RCP<Operator> CoarseNodeMatrix = Get<RCP<Operator> >(coarseLevel, "NodeAggMatrix");
+  int MyPID                      = EdgeMatrix.is_null() ? -1 : EdgeMatrix->getRowMap()->getComm()->getRank();
+
+  // Matrix matrix params
+  RCP<ParameterList> mm_params = rcp(new ParameterList);
+  ;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    mm_params->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
+
+  // Normalize P
+  if (!nodal_p_is_all_ones) {
+    // The parameters told us the nodal P isn't all ones, so we make a copy that is.
+    GetOStream(Runtime0) << "ReitzingerPFactory::BuildP(): Assuming Pn is not normalized" << std::endl;
+    RCP<Matrix> Pn_old = Pn;
+
+    Pn = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(Pn->getCrsGraph());
+    Pn->setAllToScalar(Teuchos::ScalarTraits<SC>::one());
+    Pn->fillComplete(Pn->getDomainMap(), Pn->getRangeMap());
+  } else {
+    // The parameters claim P is all ones.
+    GetOStream(Runtime0) << "ReitzingerPFactory::BuildP(): Assuming Pn is normalized" << std::endl;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(fineLevel,   "A");
-    Input(fineLevel,   "D0");
-    Input(fineLevel,   "NodeAggMatrix");
-    Input(coarseLevel, "NodeAggMatrix");
-    Input(coarseLevel, "Pnodal");
-    //    Input(coarseLevel, "NodeImporter");
+  // TODO: We need to make sure Pn isn't normalized.  Right now this has to be done explicitly by the user
 
-  }
+  // TODO: We need to look through and see which of these really need importers and which ones don't
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  /* Generate the D0 * Pn matrix and its transpose */
+  RCP<Matrix> D0_Pn, PnT_D0T, D0_Pn_nonghosted;
+  Teuchos::Array<int> D0_Pn_col_pids;
+  {
+    RCP<Matrix> dummy;
+    SubFactoryMonitor m2(*this, "Generate D0*Pn", coarseLevel);
+    D0_Pn = XMM::Multiply(*D0, false, *Pn, false, dummy, out0, true, true, "D0*Pn", mm_params);
+
+    // We don't want this guy getting accidently used later
+    if (!mm_params.is_null()) mm_params->remove("importer", false);
+
+    // Save this so we don't need to do the multiplication again later
+    D0_Pn_nonghosted = D0_Pn;
+
+    // Get owning PID information on columns for tie-breaking
+    if (!D0_Pn->getCrsGraph()->getImporter().is_null()) {
+      Xpetra::ImportUtils<LO, GO, NO> utils;
+      utils.getPids(*D0_Pn->getCrsGraph()->getImporter(), D0_Pn_col_pids, false);
+    } else {
+      D0_Pn_col_pids.resize(D0_Pn->getCrsGraph()->getColMap()->getLocalNumElements(), MyPID);
+    }
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-    using Teuchos::arcp_const_cast;
-    using MT  = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-    using XMM = Xpetra::MatrixMatrix<SC,LO,GO,NO>;
-    Teuchos::FancyOStream& out0=GetBlackHole();
-    const ParameterList& pL = GetParameterList();
-
-    bool update_communicators = pL.get<bool>("repartition: enable") && pL.get<bool>("repartition: use subcommunicators");
-
-    // If these are set correctly we assume that the nodal P contains only ones
-    bool nodal_p_is_all_ones = !pL.get<bool>("tentative: constant column sums") && !pL.get<bool>("tentative: calculate qr");
-
-    RCP<Matrix>                EdgeMatrix    = Get< RCP<Matrix> >               (fineLevel, "A");
-    RCP<Matrix>                D0            = Get< RCP<Matrix> >               (fineLevel, "D0");
-    RCP<Matrix>                NodeMatrix    = Get< RCP<Matrix> >               (fineLevel, "NodeAggMatrix");
-    RCP<Matrix>                Pn            = Get< RCP<Matrix> >               (coarseLevel, "Pnodal");
-
-    const GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
-    const LO LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-
-    // This needs to be an Operator because if NodeMatrix gets repartitioned away, we get an Operator on the level
-    RCP<Operator> CoarseNodeMatrix = Get< RCP<Operator> >(coarseLevel, "NodeAggMatrix");
-    int MyPID  = EdgeMatrix.is_null()? -1 : EdgeMatrix->getRowMap()->getComm()->getRank();
-
-    // Matrix matrix params
-    RCP<ParameterList> mm_params = rcp(new ParameterList);;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      mm_params->sublist("matrixmatrix: kernel params") = pL.sublist("matrixmatrix: kernel params");
+  {
+    // Get the transpose
+    SubFactoryMonitor m2(*this, "Transpose D0*Pn", coarseLevel);
+    PnT_D0T = Utilities::Transpose(*D0_Pn, true);
+  }
 
+  // We really need a ghosted version of D0_Pn here.
+  // The reason is that if there's only one fine edge between two coarse nodes, somebody is going
+  // to own the associated coarse edge.  The sum/sign rule doesn't guarantee the fine owner is the coarse owner.
+  // So you can wind up with a situation that only guy who *can* register the coarse edge isn't the sum/sign
+  // owner.  Adding more ghosting fixes that.
+  if (!PnT_D0T->getCrsGraph()->getImporter().is_null()) {
+    RCP<const Import> Importer     = PnT_D0T->getCrsGraph()->getImporter();
+    RCP<const CrsMatrix> D0_Pn_crs = rcp_dynamic_cast<const CrsMatrixWrap>(D0_Pn)->getCrsMatrix();
+    RCP<Matrix> D0_Pn_new          = rcp(new CrsMatrixWrap(CrsMatrixFactory::Build(D0_Pn_crs, *Importer, D0_Pn->getDomainMap(), Importer->getTargetMap())));
+    D0_Pn                          = D0_Pn_new;
+    // Get owning PID information on columns for tie-breaking
+    if (!D0_Pn->getCrsGraph()->getImporter().is_null()) {
+      Xpetra::ImportUtils<LO, GO, NO> utils;
+      utils.getPids(*D0_Pn->getCrsGraph()->getImporter(), D0_Pn_col_pids, false);
+    } else {
+      D0_Pn_col_pids.resize(D0_Pn->getCrsGraph()->getColMap()->getLocalNumElements(), MyPID);
+    }
+  }
 
-    // Normalize P
-    if(!nodal_p_is_all_ones) {
-      // The parameters told us the nodal P isn't all ones, so we make a copy that is.
-      GetOStream(Runtime0) << "ReitzingerPFactory::BuildP(): Assuming Pn is not normalized" << std::endl;
-      RCP<Matrix> Pn_old = Pn;    
+  // FIXME: This is using deprecated interfaces
+  ArrayView<const LO> colind_E, colind_N;
+  ArrayView<const SC> values_E, values_N;
 
-      Pn = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(Pn->getCrsGraph());
-      Pn->setAllToScalar(Teuchos::ScalarTraits<SC>::one());
-      Pn->fillComplete(Pn->getDomainMap(),Pn->getRangeMap());
-    }
-    else {
-      // The parameters claim P is all ones.
-      GetOStream(Runtime0) << "ReitzingerPFactory::BuildP(): Assuming Pn is normalized" << std::endl;
-    }
+  size_t Ne = EdgeMatrix->getLocalNumRows();
+  size_t Nn = NodeMatrix->getLocalNumRows();
 
-    // TODO: We need to make sure Pn isn't normalized.  Right now this has to be done explicitly by the user
+  // Upper bound on local number of coarse edges
+  size_t max_edges = (NodeMatrix->getLocalNumEntries() + Nn + 1) / 2;
+  ArrayRCP<size_t> D0_rowptr(Ne + 1);
+  ArrayRCP<LO> D0_colind(max_edges);
+  ArrayRCP<SC> D0_values(max_edges);
+  D0_rowptr[0] = 0;
 
-    // TODO: We need to look through and see which of these really need importers and which ones don't
+  LO current = 0;
+  LO Nnc     = PnT_D0T->getRowMap()->getLocalNumElements();
 
-    /* Generate the D0 * Pn matrix and its transpose */
-    RCP<Matrix> D0_Pn, PnT_D0T, D0_Pn_nonghosted;
-    Teuchos::Array<int> D0_Pn_col_pids;
-    {
-      RCP<Matrix> dummy;
-      SubFactoryMonitor m2(*this, "Generate D0*Pn", coarseLevel);
-      D0_Pn = XMM::Multiply(*D0,false,*Pn,false,dummy,out0,true,true,"D0*Pn",mm_params);
+  // Get the node maps for D0_coarse
+  RCP<const Map> ownedCoarseNodeMap           = Pn->getDomainMap();
+  RCP<const Map> ownedPlusSharedCoarseNodeMap = D0_Pn->getCrsGraph()->getColMap();
 
-      // We don't want this guy getting accidently used later
-      if(!mm_params.is_null()) mm_params->remove("importer",false);
+  for (LO i = 0; i < (LO)Nnc; i++) {
+    LO local_column_i = ownedPlusSharedCoarseNodeMap->getLocalElement(PnT_D0T->getRowMap()->getGlobalElement(i));
 
-      // Save this so we don't need to do the multiplication again later
-      D0_Pn_nonghosted = D0_Pn;
+    // FIXME: We don't really want an std::map here.  This is just a first cut implementation
+    using value_type = bool;
+    std::map<LO, value_type> ce_map;
 
-      // Get owning PID information on columns for tie-breaking
-      if(!D0_Pn->getCrsGraph()->getImporter().is_null()) {
-        Xpetra::ImportUtils<LO,GO,NO> utils;
-        utils.getPids(*D0_Pn->getCrsGraph()->getImporter(),D0_Pn_col_pids,false);
+    // FIXME: This is using deprecated interfaces
+    PnT_D0T->getLocalRowView(i, colind_E, values_E);
+
+    for (LO j = 0; j < (LO)colind_E.size(); j++) {
+      // NOTE: Edges between procs will be via handled via the a version
+      // of ML's odd/even rule
+      // For this to function correctly, we make two assumptions:
+      //  (a) The processor that owns a fine edge owns at least one of the attached nodes.
+      //  (b) Aggregation is uncoupled.
+
+      // TODO: Add some debug code to check the assumptions
+
+      // Check to see if we own this edge and continue if we don't
+      GO edge_gid = PnT_D0T->getColMap()->getGlobalElement(colind_E[j]);
+      LO j_row    = D0_Pn->getRowMap()->getLocalElement(edge_gid);
+      int pid0, pid1;
+      D0_Pn->getLocalRowView(j_row, colind_N, values_N);
+
+      // Skip incomplete rows
+      if (colind_N.size() != 2) continue;
+
+      pid0 = D0_Pn_col_pids[colind_N[0]];
+      pid1 = D0_Pn_col_pids[colind_N[1]];
+      //        printf("[%d] Row %d considering edge (%d)%d -> (%d)%d\n",MyPID,global_i,colind_N[0],D0_Pn->getColMap()->getGlobalElement(colind_N[0]),colind_N[1],D0_Pn->getColMap()->getGlobalElement(colind_N[1]));
+
+      // Check to see who owns these nodes
+      // If the sum of owning procs is odd, the lower ranked proc gets it
+
+      bool zero_matches     = pid0 == MyPID;
+      bool one_matches      = pid1 == MyPID;
+      bool keep_shared_edge = false, own_both_nodes = false;
+      if (zero_matches && one_matches) {
+        own_both_nodes = true;
+      } else {
+        bool sum_is_even  = (pid0 + pid1) % 2 == 0;
+        bool i_am_smaller = MyPID == std::min(pid0, pid1);
+        if (sum_is_even && i_am_smaller) keep_shared_edge = true;
+        if (!sum_is_even && !i_am_smaller) keep_shared_edge = true;
       }
-      else {
-        D0_Pn_col_pids.resize(D0_Pn->getCrsGraph()->getColMap()->getLocalNumElements(),MyPID);
+      //        printf("[%d] - matches %d/%d keep_shared = %d own_both = %d\n",MyPID,(int)zero_matches,(int)one_matches,(int)keep_shared_edge,(int)own_both_nodes);
+      if (!keep_shared_edge && !own_both_nodes) continue;
+
+      // We're doing this in GID space, but only because it allows us to explain
+      // the edge orientation as "always goes from lower GID to higher GID".  This could
+      // be done entirely in LIDs, but then the ordering is a little more confusing.
+      // This could be done in local indices later if we need the extra performance.
+      for (LO k = 0; k < (LO)colind_N.size(); k++) {
+        LO my_colind = colind_N[k];
+        if (my_colind != LO_INVALID && ((keep_shared_edge && my_colind != local_column_i) || (own_both_nodes && my_colind > local_column_i))) {
+          ce_map.emplace(std::make_pair(my_colind, true));
+        }
+      }  // end for k < colind_N.size()
+    }    // end for j < colind_E.size()
+
+    // std::map is sorted, so we'll just iterate through this
+    for (auto iter = ce_map.begin(); iter != ce_map.end(); iter++) {
+      LO col = iter->first;
+      if (col == local_column_i) {
+        continue;
       }
-    }
-          
 
-    {
-      // Get the transpose
-      SubFactoryMonitor m2(*this, "Transpose D0*Pn", coarseLevel);
-      PnT_D0T = Utilities::Transpose(*D0_Pn, true);
+      // NOTE: "i" here might not be a valid local column id, so we read it from the map
+      D0_colind[current] = local_column_i;
+      D0_values[current] = -1;
+      current++;
+      D0_colind[current] = col;
+      D0_values[current] = 1;
+      current++;
+      D0_rowptr[current / 2] = current;
     }
 
+  }  // end for i < Nn
 
-    // We really need a ghosted version of D0_Pn here.
-    // The reason is that if there's only one fine edge between two coarse nodes, somebody is going
-    // to own the associated coarse edge.  The sum/sign rule doesn't guarantee the fine owner is the coarse owner.
-    // So you can wind up with a situation that only guy who *can* register the coarse edge isn't the sum/sign
-    // owner.  Adding more ghosting fixes that.
-    if(!PnT_D0T->getCrsGraph()->getImporter().is_null()) {
-      RCP<const Import> Importer = PnT_D0T->getCrsGraph()->getImporter();
-      RCP<const CrsMatrix> D0_Pn_crs = rcp_dynamic_cast<const CrsMatrixWrap>(D0_Pn)->getCrsMatrix();
-      RCP<Matrix> D0_Pn_new = rcp(new CrsMatrixWrap(CrsMatrixFactory::Build(D0_Pn_crs,*Importer,D0_Pn->getDomainMap(),Importer->getTargetMap())));
-      D0_Pn = D0_Pn_new;
-      // Get owning PID information on columns for tie-breaking
-      if(!D0_Pn->getCrsGraph()->getImporter().is_null()) {
-        Xpetra::ImportUtils<LO,GO,NO> utils;
-        utils.getPids(*D0_Pn->getCrsGraph()->getImporter(),D0_Pn_col_pids,false);
-      }
-      else {
-        D0_Pn_col_pids.resize(D0_Pn->getCrsGraph()->getColMap()->getLocalNumElements(),MyPID);
-      }
-    }
+  LO num_coarse_edges = current / 2;
+  D0_rowptr.resize(num_coarse_edges + 1);
+  D0_colind.resize(current);
+  D0_values.resize(current);
 
+  // We're assuming that if the coarse NodeMatrix has no nodes on a rank, the coarse edge guy won't either.
+  // We check that here.
+  TEUCHOS_TEST_FOR_EXCEPTION((num_coarse_edges > 0 && CoarseNodeMatrix.is_null()) ||
+                                 (num_coarse_edges == 0 && !CoarseNodeMatrix.is_null()),
+                             Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: Mismatched num_coarse_edges and NodeMatrix repartition.");
 
-    // FIXME: This is using deprecated interfaces
-    ArrayView<const LO>     colind_E, colind_N;
-    ArrayView<const SC>     values_E, values_N;
-
-    size_t Ne=EdgeMatrix->getLocalNumRows();
-    size_t Nn=NodeMatrix->getLocalNumRows();
-
-    // Upper bound on local number of coarse edges
-    size_t max_edges = (NodeMatrix->getLocalNumEntries() + Nn +1) / 2;
-    ArrayRCP<size_t>  D0_rowptr(Ne+1);
-    ArrayRCP<LO>      D0_colind(max_edges);
-    ArrayRCP<SC>      D0_values(max_edges);
-    D0_rowptr[0] = 0;
-
-    LO current = 0;
-    LO Nnc = PnT_D0T->getRowMap()->getLocalNumElements();
-
-    // Get the node maps for D0_coarse
-    RCP<const Map> ownedCoarseNodeMap = Pn->getDomainMap();
-    RCP<const Map> ownedPlusSharedCoarseNodeMap  = D0_Pn->getCrsGraph()->getColMap();    
-
-    
-    for(LO i=0; i<(LO)Nnc; i++) {
-       LO local_column_i = ownedPlusSharedCoarseNodeMap->getLocalElement(PnT_D0T->getRowMap()->getGlobalElement(i));
- 
-      // FIXME: We don't really want an std::map here.  This is just a first cut implementation
-      using value_type = bool;
-      std::map<LO, value_type> ce_map;
-      
-      // FIXME: This is using deprecated interfaces
-      PnT_D0T->getLocalRowView(i,colind_E,values_E);
-      
-      for(LO j=0; j<(LO)colind_E.size(); j++) {
-
-        // NOTE: Edges between procs will be via handled via the a version
-        // of ML's odd/even rule
-        // For this to function correctly, we make two assumptions:
-        //  (a) The processor that owns a fine edge owns at least one of the attached nodes.
-        //  (b) Aggregation is uncoupled.
-
-        // TODO: Add some debug code to check the assumptions
-
-        // Check to see if we own this edge and continue if we don't
-        GO edge_gid = PnT_D0T->getColMap()->getGlobalElement(colind_E[j]);
-        LO j_row    = D0_Pn->getRowMap()->getLocalElement(edge_gid);
-        int pid0, pid1;
-        D0_Pn->getLocalRowView(j_row,colind_N,values_N);
-        
-        // Skip incomplete rows
-        if(colind_N.size() != 2) continue;
-        
-        pid0 = D0_Pn_col_pids[colind_N[0]];
-        pid1 = D0_Pn_col_pids[colind_N[1]];
-        //        printf("[%d] Row %d considering edge (%d)%d -> (%d)%d\n",MyPID,global_i,colind_N[0],D0_Pn->getColMap()->getGlobalElement(colind_N[0]),colind_N[1],D0_Pn->getColMap()->getGlobalElement(colind_N[1]));
-        
-        // Check to see who owns these nodes
-        // If the sum of owning procs is odd, the lower ranked proc gets it
-        
-        bool zero_matches = pid0 == MyPID;
-        bool one_matches  = pid1 == MyPID;
-        bool keep_shared_edge = false, own_both_nodes = false;
-        if(zero_matches && one_matches) {own_both_nodes=true;}
-        else {
-          bool sum_is_even  = (pid0 + pid1) % 2 == 0;
-          bool i_am_smaller = MyPID == std::min(pid0,pid1);
-          if(sum_is_even  && i_am_smaller)  keep_shared_edge=true;
-          if(!sum_is_even && !i_am_smaller) keep_shared_edge=true;
-        }
-        //        printf("[%d] - matches %d/%d keep_shared = %d own_both = %d\n",MyPID,(int)zero_matches,(int)one_matches,(int)keep_shared_edge,(int)own_both_nodes);
-        if(!keep_shared_edge && !own_both_nodes) continue;
-
-
-        // We're doing this in GID space, but only because it allows us to explain
-        // the edge orientation as "always goes from lower GID to higher GID".  This could
-        // be done entirely in LIDs, but then the ordering is a little more confusing.
-        // This could be done in local indices later if we need the extra performance.
-        for(LO k=0; k<(LO)colind_N.size(); k++) {
-          LO my_colind = colind_N[k];
-          if(my_colind!=LO_INVALID && ((keep_shared_edge && my_colind != local_column_i ) || (own_both_nodes && my_colind > local_column_i)) ) {
-            ce_map.emplace(std::make_pair(my_colind,true));
-          }
-        }//end for k < colind_N.size()
-      }// end for j < colind_E.size()
-      
-      
-      // std::map is sorted, so we'll just iterate through this
-      for(auto iter=ce_map.begin(); iter != ce_map.end(); iter++) {
-        LO col = iter->first;
-        if(col == local_column_i) {
-          continue;
-        }
-
-        // NOTE: "i" here might not be a valid local column id, so we read it from the map
-        D0_colind[current]  = local_column_i;
-        D0_values[current] = -1;
-        current++;
-        D0_colind[current]  = col;
-        D0_values[current] = 1;
-        current++;
-        D0_rowptr[current / 2] = current;
-      }
+  // Count the total number of edges
+  // NOTE: Since we solve the ownership issue above, this should do what we want
+  RCP<const Map> ownedCoarseEdgeMap = Xpetra::MapFactory<LO, GO, NO>::Build(EdgeMatrix->getRowMap()->lib(), GO_INVALID, num_coarse_edges, EdgeMatrix->getRowMap()->getIndexBase(), EdgeMatrix->getRowMap()->getComm());
 
-    }// end for i < Nn
-
-    LO num_coarse_edges = current / 2;
-    D0_rowptr.resize(num_coarse_edges+1);
-    D0_colind.resize(current);
-    D0_values.resize(current);
-
-    // We're assuming that if the coarse NodeMatrix has no nodes on a rank, the coarse edge guy won't either.
-    // We check that here.
-    TEUCHOS_TEST_FOR_EXCEPTION( (num_coarse_edges > 0  && CoarseNodeMatrix.is_null()) ||
-                                (num_coarse_edges == 0 && !CoarseNodeMatrix.is_null())
-                                , Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: Mismatched num_coarse_edges and NodeMatrix repartition.");
-    
-
-    // Count the total number of edges
-    // NOTE: Since we solve the ownership issue above, this should do what we want
-    RCP<const Map> ownedCoarseEdgeMap = Xpetra::MapFactory<LO,GO,NO>::Build(EdgeMatrix->getRowMap()->lib(), GO_INVALID, num_coarse_edges,EdgeMatrix->getRowMap()->getIndexBase(),EdgeMatrix->getRowMap()->getComm());
-
-    // Create the coarse D0
-    RCP<CrsMatrix> D0_coarse;
-    {
-      SubFactoryMonitor m2(*this, "Build D0", coarseLevel);
-      // FIXME: We can be smarter with memory here
-      // TODO: Is there a smarter way to get this importer?
-      D0_coarse = CrsMatrixFactory::Build(ownedCoarseEdgeMap,ownedPlusSharedCoarseNodeMap,0);
-      TEUCHOS_TEST_FOR_EXCEPTION(D0_coarse.is_null(), Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: CrsMatrixFatory failed.");
-
-      // FIXME: Deprecated code
-      ArrayRCP<size_t>  ia;
-      ArrayRCP<LO>      ja;
-      ArrayRCP<SC>     val;
-      D0_coarse->allocateAllValues(current, ia, ja, val);
-      std::copy(D0_rowptr.begin(),D0_rowptr.end(),ia.begin());
-      std::copy(D0_colind.begin(),D0_colind.end(),ja.begin());
-      std::copy(D0_values.begin(),D0_values.end(),val.begin());
-      D0_coarse->setAllValues(ia, ja, val);
+  // Create the coarse D0
+  RCP<CrsMatrix> D0_coarse;
+  {
+    SubFactoryMonitor m2(*this, "Build D0", coarseLevel);
+    // FIXME: We can be smarter with memory here
+    // TODO: Is there a smarter way to get this importer?
+    D0_coarse = CrsMatrixFactory::Build(ownedCoarseEdgeMap, ownedPlusSharedCoarseNodeMap, 0);
+    TEUCHOS_TEST_FOR_EXCEPTION(D0_coarse.is_null(), Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: CrsMatrixFatory failed.");
+
+    // FIXME: Deprecated code
+    ArrayRCP<size_t> ia;
+    ArrayRCP<LO> ja;
+    ArrayRCP<SC> val;
+    D0_coarse->allocateAllValues(current, ia, ja, val);
+    std::copy(D0_rowptr.begin(), D0_rowptr.end(), ia.begin());
+    std::copy(D0_colind.begin(), D0_colind.end(), ja.begin());
+    std::copy(D0_values.begin(), D0_values.end(), val.begin());
+    D0_coarse->setAllValues(ia, ja, val);
 
 #if 0
       {
@@ -383,104 +372,102 @@ namespace MueLu {
         
       }
 #endif
-      D0_coarse->expertStaticFillComplete(ownedCoarseNodeMap,ownedCoarseEdgeMap);
-    }
-    RCP<Matrix> D0_coarse_m = rcp(new CrsMatrixWrap(D0_coarse));
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
- 
-
-    // Create the Pe matrix, but with the extra entries.  From ML's notes:
-    /* The general idea is that the matrix                              */
-    /*                        T_h P_n T_H^*                             */
-    /* is almost Pe. If we make sure that P_n contains 1's and -1's, the*/
-    /* matrix triple product will yield a matrix with +/- 1 and +/- 2's.*/
-    /* If we remove all the 1's and divide the 2's by 2. we arrive at Pe*/
-    RCP<Matrix> Pe;
-    {
-      SubFactoryMonitor m2(*this, "Generate Pe (pre-fix)", coarseLevel);
-
-      RCP<Matrix> dummy;
-      RCP<Matrix> Pn_D0cT = XMM::Multiply(*Pn,false,*D0_coarse_m,true,dummy,out0,true,true,"Pn*D0c'",mm_params);
-
-      // We don't want this guy getting accidently used later
-      if(!mm_params.is_null()) mm_params->remove("importer",false);
-
-      Pe = XMM::Multiply(*D0,false,*Pn_D0cT,false,dummy,out0,true,true,"D0*(Pn*D0c')",mm_params);
-
-      // TODO: Something like this *might* work.  But this specifically, doesn't
-      // Pe = XMM::Multiply(*D0_Pn_nonghosted,false,*D0_coarse_m,true,dummy,out0,true,true,"(D0*Pn)*D0c'",mm_params);
-    }
+    D0_coarse->expertStaticFillComplete(ownedCoarseNodeMap, ownedCoarseEdgeMap);
+  }
+  RCP<Matrix> D0_coarse_m         = rcp(new CrsMatrixWrap(D0_coarse));
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+
+  // Create the Pe matrix, but with the extra entries.  From ML's notes:
+  /* The general idea is that the matrix                              */
+  /*                        T_h P_n T_H^*                             */
+  /* is almost Pe. If we make sure that P_n contains 1's and -1's, the*/
+  /* matrix triple product will yield a matrix with +/- 1 and +/- 2's.*/
+  /* If we remove all the 1's and divide the 2's by 2. we arrive at Pe*/
+  RCP<Matrix> Pe;
+  {
+    SubFactoryMonitor m2(*this, "Generate Pe (pre-fix)", coarseLevel);
+
+    RCP<Matrix> dummy;
+    RCP<Matrix> Pn_D0cT = XMM::Multiply(*Pn, false, *D0_coarse_m, true, dummy, out0, true, true, "Pn*D0c'", mm_params);
+
+    // We don't want this guy getting accidently used later
+    if (!mm_params.is_null()) mm_params->remove("importer", false);
+
+    Pe = XMM::Multiply(*D0, false, *Pn_D0cT, false, dummy, out0, true, true, "D0*(Pn*D0c')", mm_params);
 
-    /* Weed out the +/- entries, shrinking the matrix as we go */
-    { 
-      SubFactoryMonitor m2(*this, "Generate Pe (post-fix)", coarseLevel);
-      Pe->resumeFill();
-      SC one = Teuchos::ScalarTraits<SC>::one();
-      MT two = 2*Teuchos::ScalarTraits<MT>::one();
-      SC zero = Teuchos::ScalarTraits<SC>::zero();
-      SC neg_one = - one;
-
-      RCP<const CrsMatrix> Pe_crs = rcp_dynamic_cast<const CrsMatrixWrap>(Pe)->getCrsMatrix();
-      TEUCHOS_TEST_FOR_EXCEPTION(Pe_crs.is_null(), Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: Pe is not a crs matrix.");
-      ArrayRCP<const size_t > rowptr_const;
-      ArrayRCP<const LO> colind_const;
-      ArrayRCP<const SC> values_const;
-      Pe_crs->getAllValues(rowptr_const,colind_const,values_const);
-      ArrayRCP<size_t> rowptr = arcp_const_cast<size_t>(rowptr_const);
-      ArrayRCP<LO> colind = arcp_const_cast<LO>(colind_const);
-      ArrayRCP<SC> values = arcp_const_cast<SC>(values_const);
-      LO ct = 0;
-      LO lower = rowptr[0];
-      for(LO i=0; i<(LO)Ne; i++) {
-        for(size_t j=lower; j<rowptr[i+1]; j++) {
-          if (values[j] == one || values[j] == neg_one || values[j] == zero) {
-            // drop this guy
-          }
-          else {
-            colind[ct] = colind[j];
-            values[ct] = values[j] / two;
-            ct++;
-          }          
+    // TODO: Something like this *might* work.  But this specifically, doesn't
+    // Pe = XMM::Multiply(*D0_Pn_nonghosted,false,*D0_coarse_m,true,dummy,out0,true,true,"(D0*Pn)*D0c'",mm_params);
+  }
+
+  /* Weed out the +/- entries, shrinking the matrix as we go */
+  {
+    SubFactoryMonitor m2(*this, "Generate Pe (post-fix)", coarseLevel);
+    Pe->resumeFill();
+    SC one     = Teuchos::ScalarTraits<SC>::one();
+    MT two     = 2 * Teuchos::ScalarTraits<MT>::one();
+    SC zero    = Teuchos::ScalarTraits<SC>::zero();
+    SC neg_one = -one;
+
+    RCP<const CrsMatrix> Pe_crs = rcp_dynamic_cast<const CrsMatrixWrap>(Pe)->getCrsMatrix();
+    TEUCHOS_TEST_FOR_EXCEPTION(Pe_crs.is_null(), Exceptions::RuntimeError, "MueLu::ReitzingerPFactory: Pe is not a crs matrix.");
+    ArrayRCP<const size_t> rowptr_const;
+    ArrayRCP<const LO> colind_const;
+    ArrayRCP<const SC> values_const;
+    Pe_crs->getAllValues(rowptr_const, colind_const, values_const);
+    ArrayRCP<size_t> rowptr = arcp_const_cast<size_t>(rowptr_const);
+    ArrayRCP<LO> colind     = arcp_const_cast<LO>(colind_const);
+    ArrayRCP<SC> values     = arcp_const_cast<SC>(values_const);
+    LO ct                   = 0;
+    LO lower                = rowptr[0];
+    for (LO i = 0; i < (LO)Ne; i++) {
+      for (size_t j = lower; j < rowptr[i + 1]; j++) {
+        if (values[j] == one || values[j] == neg_one || values[j] == zero) {
+          // drop this guy
+        } else {
+          colind[ct] = colind[j];
+          values[ct] = values[j] / two;
+          ct++;
         }
-        lower = rowptr[i+1];
-        rowptr[i+1] = ct;
       }
-      rowptr[Ne] = ct;
-      colind.resize(ct);
-      values.resize(ct);
-      rcp_const_cast<CrsMatrix>(Pe_crs)->setAllValues(rowptr,colind,values);
-
-      Pe->fillComplete(Pe->getDomainMap(),Pe->getRangeMap());
+      lower         = rowptr[i + 1];
+      rowptr[i + 1] = ct;
     }
-   
-    /* Check commuting property */
-    CheckCommutingProperty(*Pe,*D0_coarse_m,*D0,*Pn);
-
-    /*  If we're repartitioning here, we need to cut down the communicators */
-    // NOTE: We need to do this *after* checking the commuting property, since
-    // that's going to need to fineLevel's communicators, not the repartitioned ones
-    if(update_communicators) {
-      //NOTE: We can only do D0 here.  We have to do Ke_coarse=(Re Ke_fine Pe) in RebalanceAcFactory
-      RCP<const Teuchos::Comm<int> > newComm;
-      if(!CoarseNodeMatrix.is_null()) newComm = CoarseNodeMatrix->getDomainMap()->getComm();
-      RCP<const Map>  newMap  = Xpetra::MapFactory<LO,GO,NO>::copyMapWithNewComm(D0_coarse_m->getRowMap(),newComm);
-      D0_coarse_m->removeEmptyProcessesInPlace(newMap);
-
-      // The "in place" still leaves a dummy matrix here.  That needs to go
-      if(newMap.is_null()) D0_coarse_m = Teuchos::null;
-
-      Set(coarseLevel,"InPlaceMap",newMap);
-    }
-    
-    /* Set output on the level */
-    Set(coarseLevel,"P",Pe);
-    Set(coarseLevel,"Ptent",Pe);
-
-    Set(coarseLevel,"D0",D0_coarse_m);
-    coarseLevel.Set("D0",D0_coarse_m,NoFactory::get());
-    coarseLevel.AddKeepFlag("D0",NoFactory::get(), MueLu::Final);
-    coarseLevel.RemoveKeepFlag("D0",NoFactory::get(), MueLu::UserData);
-   
+    rowptr[Ne] = ct;
+    colind.resize(ct);
+    values.resize(ct);
+    rcp_const_cast<CrsMatrix>(Pe_crs)->setAllValues(rowptr, colind, values);
+
+    Pe->fillComplete(Pe->getDomainMap(), Pe->getRangeMap());
+  }
+
+  /* Check commuting property */
+  CheckCommutingProperty(*Pe, *D0_coarse_m, *D0, *Pn);
+
+  /*  If we're repartitioning here, we need to cut down the communicators */
+  // NOTE: We need to do this *after* checking the commuting property, since
+  // that's going to need to fineLevel's communicators, not the repartitioned ones
+  if (update_communicators) {
+    // NOTE: We can only do D0 here.  We have to do Ke_coarse=(Re Ke_fine Pe) in RebalanceAcFactory
+    RCP<const Teuchos::Comm<int> > newComm;
+    if (!CoarseNodeMatrix.is_null()) newComm = CoarseNodeMatrix->getDomainMap()->getComm();
+    RCP<const Map> newMap = Xpetra::MapFactory<LO, GO, NO>::copyMapWithNewComm(D0_coarse_m->getRowMap(), newComm);
+    D0_coarse_m->removeEmptyProcessesInPlace(newMap);
+
+    // The "in place" still leaves a dummy matrix here.  That needs to go
+    if (newMap.is_null()) D0_coarse_m = Teuchos::null;
+
+    Set(coarseLevel, "InPlaceMap", newMap);
+  }
+
+  /* Set output on the level */
+  Set(coarseLevel, "P", Pe);
+  Set(coarseLevel, "Ptent", Pe);
+
+  Set(coarseLevel, "D0", D0_coarse_m);
+  coarseLevel.Set("D0", D0_coarse_m, NoFactory::get());
+  coarseLevel.AddKeepFlag("D0", NoFactory::get(), MueLu::Final);
+  coarseLevel.RemoveKeepFlag("D0", NoFactory::get(), MueLu::UserData);
+
 #if 0
   {
     int numProcs = Pe->getRowMap()->getComm()->getSize();
@@ -493,40 +480,35 @@ namespace MueLu {
   }
 #endif
 
-  }// end Build
+}  // end Build
 
- template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
- void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
- CheckCommutingProperty(const Matrix & Pe, const Matrix & D0_c, const Matrix& D0_f, const Matrix & Pn) const {
-   if(IsPrint(Statistics0)) {
-     using XMM = Xpetra::MatrixMatrix<SC,LO,GO,NO>;
-     using MT   = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-     SC one = Teuchos::ScalarTraits<SC>::one();
-     SC zero = Teuchos::ScalarTraits<SC>::zero();
-
-     RCP<Matrix> dummy;
-     Teuchos::FancyOStream &out0=GetBlackHole();
-     RCP<Matrix> left  = XMM::Multiply(Pe,false,D0_c,false,dummy,out0);
-     RCP<Matrix> right = XMM::Multiply(D0_f,false,Pn,false,dummy,out0);
-     
-     // We need a non-FC matrix for the add, sadly
-     RCP<CrsMatrix> sum_c = CrsMatrixFactory::Build(left->getRowMap(),left->getLocalMaxNumRowEntries()+right->getLocalMaxNumRowEntries());
-     RCP<Matrix> summation = rcp(new CrsMatrixWrap(sum_c));
-     XMM::TwoMatrixAdd(*left,  false, one, *summation, zero);
-     XMM::TwoMatrixAdd(*right, false, -one, *summation, one);
-     
-     MT norm = summation->getFrobeniusNorm();
-     GetOStream(Statistics0) << "CheckCommutingProperty: ||Pe D0_c - D0_f Pn || = "<<norm<<std::endl;
-    
-   }
-
- }//end CheckCommutingProperty
-
-
-
-} //namespace MueLu
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReitzingerPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CheckCommutingProperty(const Matrix& Pe, const Matrix& D0_c, const Matrix& D0_f, const Matrix& Pn) const {
+  if (IsPrint(Statistics0)) {
+    using XMM = Xpetra::MatrixMatrix<SC, LO, GO, NO>;
+    using MT  = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+    SC one    = Teuchos::ScalarTraits<SC>::one();
+    SC zero   = Teuchos::ScalarTraits<SC>::zero();
+
+    RCP<Matrix> dummy;
+    Teuchos::FancyOStream& out0 = GetBlackHole();
+    RCP<Matrix> left            = XMM::Multiply(Pe, false, D0_c, false, dummy, out0);
+    RCP<Matrix> right           = XMM::Multiply(D0_f, false, Pn, false, dummy, out0);
+
+    // We need a non-FC matrix for the add, sadly
+    RCP<CrsMatrix> sum_c  = CrsMatrixFactory::Build(left->getRowMap(), left->getLocalMaxNumRowEntries() + right->getLocalMaxNumRowEntries());
+    RCP<Matrix> summation = rcp(new CrsMatrixWrap(sum_c));
+    XMM::TwoMatrixAdd(*left, false, one, *summation, zero);
+    XMM::TwoMatrixAdd(*right, false, -one, *summation, one);
+
+    MT norm = summation->getFrobeniusNorm();
+    GetOStream(Statistics0) << "CheckCommutingProperty: ||Pe D0_c - D0_f Pn || = " << norm << std::endl;
+  }
 
+}  // end CheckCommutingProperty
 
+}  // namespace MueLu
 
 #define MUELU_REITZINGERPFACTORY_SHORT
-#endif // MUELU_REITZINGERPFACTORY_DEF_HPP
+#endif  // MUELU_REITZINGERPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_decl.hpp
index 090f6d552768..34cc9fb29798 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_decl.hpp
@@ -60,102 +60,99 @@
 
 namespace MueLu {
 
-  /*!
-    @class SaPFactory class.
-    @ingroup MueLuTransferClasses
-    @brief Factory for building Smoothed Aggregation prolongators.
+/*!
+  @class SaPFactory class.
+  @ingroup MueLuTransferClasses
+  @brief Factory for building Smoothed Aggregation prolongators.
 
-    ## Input/output of SaPFactory ##
+  ## Input/output of SaPFactory ##
 
-    ### User parameters of SaPFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | sa: damping factor                     | double  | 1.33  | * | * |   | Damping factor for smoothed aggregation transfer operators |
-    | sa: calculate eigenvalue estimate      | bool    | false | * | * |   | Force calculation of eigenvalue estimate during prolongator smoothing |
-    | sa: eigenvalue estimate num iterations | int     | 10    | * | * |   | Number of power iterations to estimate max eigenvalue |
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | P                                      | Factory | null  |   | * | * | Tentative prolongator factory (should be a TentativePFactory object). Note that if "P" is not provided the FactoryManager tries the factory which is generating the variable "Ptent" as input for "P". In the standard case using the default settings this is the non-smoothed tentative prolongation operator. |
+  ### User parameters of SaPFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | sa: damping factor                     | double  | 1.33  | * | * |   | Damping factor for smoothed aggregation transfer operators |
+  | sa: calculate eigenvalue estimate      | bool    | false | * | * |   | Force calculation of eigenvalue estimate during prolongator smoothing |
+  | sa: eigenvalue estimate num iterations | int     | 10    | * | * |   | Number of power iterations to estimate max eigenvalue |
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | P                                      | Factory | null  |   | * | * | Tentative prolongator factory (should be a TentativePFactory object). Note that if "P" is not provided the FactoryManager tries the factory which is generating the variable "Ptent" as input for "P". In the standard case using the default settings this is the non-smoothed tentative prolongation operator. |
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SaPFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see SaPFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SaPFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see SaPFactory::DeclareInput).
 
-    ### Variables provided by SaPFactory ###
+  ### Variables provided by SaPFactory ###
 
-    After SaPFactory::Build the following data is available (if requested)
+  After SaPFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | SaPFactory   | Smoothed prolongator
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | SaPFactory   | Smoothed prolongator
 
 
-  */
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class SaPFactory : public PFactory {
+          class Node          = DefaultNode>
+class SaPFactory : public PFactory {
 #undef MUELU_SAPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    /*! @brief Constructor.
-      User can supply a factory for generating the tentative prolongator.
-    */
-    SaPFactory() { }
-
-    //! Destructor.
-    virtual ~SaPFactory() { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  /*! @brief Constructor.
+    User can supply a factory for generating the tentative prolongator.
+  */
+  SaPFactory() {}
 
-    //@}
+  //! Destructor.
+  virtual ~SaPFactory() {}
 
-    //! Input
-    //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  //@}
 
-    //@}
+  //! Input
+  //@{
 
-    //! @name Build methods.
-    //@{
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    /*!
-      @brief Build method.
+  //@}
 
-      Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
-      //FIXME what does the return code mean (unclear in MueMat)?
-      */
-    void Build(Level& fineLevel, Level &coarseLevel) const;
+  //! @name Build methods.
+  //@{
 
-    void BuildP(Level &fineLevel, Level &coarseLevel) const; //Build()
+  /*!
+    @brief Build method.
 
-    /*!
-      @brief Enforce constraints on prolongator
+    Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
+    //FIXME what does the return code mean (unclear in MueMat)?
+    */
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-      Modifies the prolongator so that all entries lie between 0 and 1. It also
-      ensures that the row sum is between 0 and 1. This option only makes
-      sense in the SA constext if the tentative prolgonator does not use
-      the QR factorization.
-      */
-    void SatisfyPConstraints(RCP<Matrix> A, RCP<Matrix>& P) const;
-    void optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix>& P) const;
+  void BuildP(Level &fineLevel, Level &coarseLevel) const;  // Build()
 
-    bool constrainRow(Scalar *orig, LocalOrdinal nEntries, Scalar leftBound, Scalar rghtBound,Scalar rsumTarget, Scalar *fixedUnsorted, Scalar *scalarData) const;
+  /*!
+    @brief Enforce constraints on prolongator
 
+    Modifies the prolongator so that all entries lie between 0 and 1. It also
+    ensures that the row sum is between 0 and 1. This option only makes
+    sense in the SA constext if the tentative prolgonator does not use
+    the QR factorization.
+    */
+  void SatisfyPConstraints(RCP<Matrix> A, RCP<Matrix> &P) const;
+  void optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix> &P) const;
 
+  bool constrainRow(Scalar *orig, LocalOrdinal nEntries, Scalar leftBound, Scalar rghtBound, Scalar rsumTarget, Scalar *fixedUnsorted, Scalar *scalarData) const;
 
-    //@}
+  //@}
 
-  }; //class SaPFactory
+};  // class SaPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SAPFACTORY_SHORT
-#endif // MUELU_SAPFACTORY_DECL_HPP
+#endif  // MUELU_SAPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_def.hpp
index e870648f571f..7a9180c7d5a5 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_def.hpp
@@ -47,7 +47,7 @@
 #define MUELU_SAPFACTORY_DEF_HPP
 
 #include <Xpetra_Matrix.hpp>
-#include <Xpetra_IteratorOps.hpp> // containing routines to generate Jacobi iterator
+#include <Xpetra_IteratorOps.hpp>  // containing routines to generate Jacobi iterator
 #include <Xpetra_IO.hpp>
 #include <sstream>
 
@@ -63,531 +63,527 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("sa: damping factor");
-    SET_VALID_ENTRY("sa: calculate eigenvalue estimate");
-    SET_VALID_ENTRY("sa: eigenvalue estimate num iterations");
-    SET_VALID_ENTRY("sa: use rowsumabs diagonal scaling");
-    SET_VALID_ENTRY("sa: enforce constraints");
-    SET_VALID_ENTRY("tentative: calculate qr");
-    SET_VALID_ENTRY("sa: max eigenvalue");
-    SET_VALID_ENTRY("sa: rowsumabs diagonal replacement tolerance");
-    SET_VALID_ENTRY("sa: rowsumabs diagonal replacement value");
-    SET_VALID_ENTRY("sa: rowsumabs replace single entry row with zero");
-    SET_VALID_ENTRY("sa: rowsumabs use automatic diagonal tolerance");
-#undef  SET_VALID_ENTRY
-
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("P",              Teuchos::null, "Tentative prolongator factory");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-
-    return validParamList;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    Input(fineLevel, "A");
+  SET_VALID_ENTRY("sa: damping factor");
+  SET_VALID_ENTRY("sa: calculate eigenvalue estimate");
+  SET_VALID_ENTRY("sa: eigenvalue estimate num iterations");
+  SET_VALID_ENTRY("sa: use rowsumabs diagonal scaling");
+  SET_VALID_ENTRY("sa: enforce constraints");
+  SET_VALID_ENTRY("tentative: calculate qr");
+  SET_VALID_ENTRY("sa: max eigenvalue");
+  SET_VALID_ENTRY("sa: rowsumabs diagonal replacement tolerance");
+  SET_VALID_ENTRY("sa: rowsumabs diagonal replacement value");
+  SET_VALID_ENTRY("sa: rowsumabs replace single entry row with zero");
+  SET_VALID_ENTRY("sa: rowsumabs use automatic diagonal tolerance");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Tentative prolongator factory");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
 
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    coarseLevel.DeclareInput("P", initialPFact.get(), this); // --
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  Input(fineLevel, "A");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level &coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
   }
+  coarseLevel.DeclareInput("P", initialPFact.get(), this);  // --
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
-    FactoryMonitor m(*this, "Prolongator smoothing", coarseLevel);
-
-    std::string levelIDs = toString(coarseLevel.GetLevelID());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
-    const std::string prefix = "MueLu::SaPFactory(" + levelIDs + "): ";
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level &fineLevel, Level &coarseLevel) const {
+  FactoryMonitor m(*this, "Prolongator smoothing", coarseLevel);
 
-    typedef typename Teuchos::ScalarTraits<SC>::coordinateType Coordinate;
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  std::string levelIDs = toString(coarseLevel.GetLevelID());
 
+  const std::string prefix = "MueLu::SaPFactory(" + levelIDs + "): ";
 
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
-    // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    const ParameterList& pL = GetParameterList();
+  typedef typename Teuchos::ScalarTraits<SC>::coordinateType Coordinate;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
 
-    // Level Get
-    RCP<Matrix> A     = Get< RCP<Matrix> >(fineLevel, "A");
-    RCP<Matrix> Ptent = coarseLevel.Get< RCP<Matrix> >("P", initialPFact.get());
-    RCP<Matrix> finalP;
-    // If Tentative facctory bailed out (e.g., number of global aggregates is 0), then SaPFactory bails
-   //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-    if (Ptent == Teuchos::null) {
-      finalP = Teuchos::null;
-      Set(coarseLevel, "P", finalP);
-      return;
-    }
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory and NullspaceFactory.
+  // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  }
+  const ParameterList &pL = GetParameterList();
+
+  // Level Get
+  RCP<Matrix> A     = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<Matrix> Ptent = coarseLevel.Get<RCP<Matrix> >("P", initialPFact.get());
+  RCP<Matrix> finalP;
+  // If Tentative facctory bailed out (e.g., number of global aggregates is 0), then SaPFactory bails
+  //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (Ptent == Teuchos::null) {
+    finalP = Teuchos::null;
+    Set(coarseLevel, "P", finalP);
+    return;
+  }
 
-    if (restrictionMode_) {
-      SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
-      A = Utilities::Transpose(*A, true); // build transpose of A explicitely
-    }
+  if (restrictionMode_) {
+    SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
+    A = Utilities::Transpose(*A, true);  // build transpose of A explicitely
+  }
 
-    // Build final prolongator
+  // Build final prolongator
 
-    // Reuse pattern if available
-    RCP<ParameterList> APparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      APparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-    else
-      APparams = rcp(new ParameterList);
+  // Reuse pattern if available
+  RCP<ParameterList> APparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    APparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    APparams = rcp(new ParameterList);
 
-    if (coarseLevel.IsAvailable("AP reuse data", this)) {
-      GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
+  if (coarseLevel.IsAvailable("AP reuse data", this)) {
+    GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
 
-      APparams = coarseLevel.Get< RCP<ParameterList> >("AP reuse data", this);
+    APparams = coarseLevel.Get<RCP<ParameterList> >("AP reuse data", this);
 
-      if (APparams->isParameter("graph"))
-        finalP = APparams->get< RCP<Matrix> >("graph");
-    }
-    // By default, we don't need global constants for SaP
-    APparams->set("compute global constants: temporaries",APparams->get("compute global constants: temporaries",false));
-    APparams->set("compute global constants",APparams->get("compute global constants",false));
-
-    const SC dampingFactor      = as<SC>(pL.get<double>("sa: damping factor"));
-    const LO maxEigenIterations = as<LO>(pL.get<int>   ("sa: eigenvalue estimate num iterations"));
-    const bool estimateMaxEigen =        pL.get<bool>  ("sa: calculate eigenvalue estimate");
-    const bool useAbsValueRowSum=        pL.get<bool>  ("sa: use rowsumabs diagonal scaling");
-    const bool doQRStep         =        pL.get<bool>("tentative: calculate qr");
-    const bool enforceConstraints=       pL.get<bool>("sa: enforce constraints");
-    const MT   userDefinedMaxEigen =  as<MT>(pL.get<double>("sa: max eigenvalue"));
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
-    double dTol = pL.get<double>("sa: rowsumabs diagonal replacement tolerance");
-    const Magnitude diagonalReplacementTolerance = (dTol == as<double>(-1) ? Teuchos::ScalarTraits<Scalar>::eps()*100 : as<Magnitude>(pL.get<double>("sa: rowsumabs diagonal replacement tolerance")));
-    const SC diagonalReplacementValue =  as<SC>(pL.get<double>("sa: rowsumabs diagonal replacement value"));
-    const bool replaceSingleEntryRowWithZero = pL.get<bool>("sa: rowsumabs replace single entry row with zero");
-    const bool useAutomaticDiagTol =       pL.get<bool>("sa: rowsumabs use automatic diagonal tolerance");
-
-    // Sanity checking
-    TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && enforceConstraints,Exceptions::RuntimeError,
-                               "MueLu::TentativePFactory::MakeTentative: cannot use 'enforce constraints' and 'calculate qr' at the same time");
-
-    if (dampingFactor != Teuchos::ScalarTraits<SC>::zero()) {
-
-      Scalar lambdaMax;
-      Teuchos::RCP<Vector> invDiag;
-      if (userDefinedMaxEigen == -1.)
-      {
-        SubFactoryMonitor m2(*this, "Eigenvalue estimate", coarseLevel);
-        lambdaMax = A->GetMaxEigenvalueEstimate();
-        if (lambdaMax == -Teuchos::ScalarTraits<SC>::one() || estimateMaxEigen) {
-          GetOStream(Statistics1) << "Calculating max eigenvalue estimate now (max iters = "<< maxEigenIterations <<
-          ( (useAbsValueRowSum) ?  ", use rowSumAbs diagonal)" :  ", use point diagonal)") << std::endl;
-          Coordinate stopTol = 1e-4;
-          if (useAbsValueRowSum) {
-            const bool returnReciprocal=true;
-            invDiag = Utilities::GetLumpedMatrixDiagonal(*A,returnReciprocal,
-                                                        diagonalReplacementTolerance,
-                                                        diagonalReplacementValue,
-                                                        replaceSingleEntryRowWithZero,
-                                                        useAutomaticDiagTol);
-            TEUCHOS_TEST_FOR_EXCEPTION(invDiag.is_null(), Exceptions::RuntimeError,
-                                       "SaPFactory: eigenvalue estimate: diagonal reciprocal is null.");
-            lambdaMax = Utilities::PowerMethod(*A, invDiag, maxEigenIterations, stopTol);
-          } else
-            lambdaMax = Utilities::PowerMethod(*A, true, maxEigenIterations, stopTol);
-          A->SetMaxEigenvalueEstimate(lambdaMax);
-        } else {
-          GetOStream(Statistics1) << "Using cached max eigenvalue estimate" << std::endl;
-        }
+    if (APparams->isParameter("graph"))
+      finalP = APparams->get<RCP<Matrix> >("graph");
+  }
+  // By default, we don't need global constants for SaP
+  APparams->set("compute global constants: temporaries", APparams->get("compute global constants: temporaries", false));
+  APparams->set("compute global constants", APparams->get("compute global constants", false));
+
+  const SC dampingFactor        = as<SC>(pL.get<double>("sa: damping factor"));
+  const LO maxEigenIterations   = as<LO>(pL.get<int>("sa: eigenvalue estimate num iterations"));
+  const bool estimateMaxEigen   = pL.get<bool>("sa: calculate eigenvalue estimate");
+  const bool useAbsValueRowSum  = pL.get<bool>("sa: use rowsumabs diagonal scaling");
+  const bool doQRStep           = pL.get<bool>("tentative: calculate qr");
+  const bool enforceConstraints = pL.get<bool>("sa: enforce constraints");
+  const MT userDefinedMaxEigen  = as<MT>(pL.get<double>("sa: max eigenvalue"));
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+  double dTol                                  = pL.get<double>("sa: rowsumabs diagonal replacement tolerance");
+  const Magnitude diagonalReplacementTolerance = (dTol == as<double>(-1) ? Teuchos::ScalarTraits<Scalar>::eps() * 100 : as<Magnitude>(pL.get<double>("sa: rowsumabs diagonal replacement tolerance")));
+  const SC diagonalReplacementValue            = as<SC>(pL.get<double>("sa: rowsumabs diagonal replacement value"));
+  const bool replaceSingleEntryRowWithZero     = pL.get<bool>("sa: rowsumabs replace single entry row with zero");
+  const bool useAutomaticDiagTol               = pL.get<bool>("sa: rowsumabs use automatic diagonal tolerance");
+
+  // Sanity checking
+  TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && enforceConstraints, Exceptions::RuntimeError,
+                             "MueLu::TentativePFactory::MakeTentative: cannot use 'enforce constraints' and 'calculate qr' at the same time");
+
+  if (dampingFactor != Teuchos::ScalarTraits<SC>::zero()) {
+    Scalar lambdaMax;
+    Teuchos::RCP<Vector> invDiag;
+    if (userDefinedMaxEigen == -1.) {
+      SubFactoryMonitor m2(*this, "Eigenvalue estimate", coarseLevel);
+      lambdaMax = A->GetMaxEigenvalueEstimate();
+      if (lambdaMax == -Teuchos::ScalarTraits<SC>::one() || estimateMaxEigen) {
+        GetOStream(Statistics1) << "Calculating max eigenvalue estimate now (max iters = " << maxEigenIterations << ((useAbsValueRowSum) ? ", use rowSumAbs diagonal)" : ", use point diagonal)") << std::endl;
+        Coordinate stopTol = 1e-4;
+        if (useAbsValueRowSum) {
+          const bool returnReciprocal = true;
+          invDiag                     = Utilities::GetLumpedMatrixDiagonal(*A, returnReciprocal,
+                                                                           diagonalReplacementTolerance,
+                                                                           diagonalReplacementValue,
+                                                                           replaceSingleEntryRowWithZero,
+                                                                           useAutomaticDiagTol);
+          TEUCHOS_TEST_FOR_EXCEPTION(invDiag.is_null(), Exceptions::RuntimeError,
+                                     "SaPFactory: eigenvalue estimate: diagonal reciprocal is null.");
+          lambdaMax = Utilities::PowerMethod(*A, invDiag, maxEigenIterations, stopTol);
+        } else
+          lambdaMax = Utilities::PowerMethod(*A, true, maxEigenIterations, stopTol);
+        A->SetMaxEigenvalueEstimate(lambdaMax);
+      } else {
+        GetOStream(Statistics1) << "Using cached max eigenvalue estimate" << std::endl;
       }
-      else {
-          lambdaMax = userDefinedMaxEigen;
-          A->SetMaxEigenvalueEstimate(lambdaMax);
+    } else {
+      lambdaMax = userDefinedMaxEigen;
+      A->SetMaxEigenvalueEstimate(lambdaMax);
+    }
+    GetOStream(Statistics1) << "Prolongator damping factor = " << dampingFactor / lambdaMax << " (" << dampingFactor << " / " << lambdaMax << ")" << std::endl;
+
+    {
+      SubFactoryMonitor m2(*this, "Fused (I-omega*D^{-1} A)*Ptent", coarseLevel);
+      if (!useAbsValueRowSum)
+        invDiag = Utilities::GetMatrixDiagonalInverse(*A);  // default
+      else if (invDiag == Teuchos::null) {
+        GetOStream(Runtime0) << "Using rowsumabs diagonal" << std::endl;
+        const bool returnReciprocal = true;
+        invDiag                     = Utilities::GetLumpedMatrixDiagonal(*A, returnReciprocal,
+                                                                         diagonalReplacementTolerance,
+                                                                         diagonalReplacementValue,
+                                                                         replaceSingleEntryRowWithZero,
+                                                                         useAutomaticDiagTol);
+        TEUCHOS_TEST_FOR_EXCEPTION(invDiag.is_null(), Exceptions::RuntimeError, "SaPFactory: diagonal reciprocal is null.");
       }
-      GetOStream(Statistics1) << "Prolongator damping factor = " << dampingFactor/lambdaMax << " (" << dampingFactor << " / " << lambdaMax << ")" << std::endl;
-
-      {
-        SubFactoryMonitor m2(*this, "Fused (I-omega*D^{-1} A)*Ptent", coarseLevel);
-        if (!useAbsValueRowSum)
-          invDiag = Utilities::GetMatrixDiagonalInverse(*A); //default
-        else if (invDiag == Teuchos::null) {
-          GetOStream(Runtime0) << "Using rowsumabs diagonal" << std::endl;
-          const bool returnReciprocal=true;
-          invDiag = Utilities::GetLumpedMatrixDiagonal(*A,returnReciprocal,
-                                                        diagonalReplacementTolerance,
-                                                        diagonalReplacementValue,
-                                                        replaceSingleEntryRowWithZero,
-                                                        useAutomaticDiagTol);
-          TEUCHOS_TEST_FOR_EXCEPTION(invDiag.is_null(), Exceptions::RuntimeError, "SaPFactory: diagonal reciprocal is null.");
-        }	
-	
-	      SC omega = dampingFactor / lambdaMax;
-	      TEUCHOS_TEST_FOR_EXCEPTION(!std::isfinite(Teuchos::ScalarTraits<SC>::magnitude(omega)), Exceptions::RuntimeError, "Prolongator damping factor needs to be finite.");
-
-        // finalP = Ptent + (I - \omega D^{-1}A) Ptent
-        finalP = Xpetra::IteratorOps<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Jacobi(omega, *invDiag, *A, *Ptent, finalP,
-                    GetOStream(Statistics2), std::string("MueLu::SaP-")+levelIDs, APparams);
-        if (enforceConstraints) {
-           if (A->GetFixedBlockSize() == 1) optimalSatisfyPConstraintsForScalarPDEs( finalP);
-           else                             SatisfyPConstraints( A, finalP);
-        }
+
+      SC omega = dampingFactor / lambdaMax;
+      TEUCHOS_TEST_FOR_EXCEPTION(!std::isfinite(Teuchos::ScalarTraits<SC>::magnitude(omega)), Exceptions::RuntimeError, "Prolongator damping factor needs to be finite.");
+
+      // finalP = Ptent + (I - \omega D^{-1}A) Ptent
+      finalP = Xpetra::IteratorOps<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Jacobi(omega, *invDiag, *A, *Ptent, finalP,
+                                                                                      GetOStream(Statistics2), std::string("MueLu::SaP-") + levelIDs, APparams);
+      if (enforceConstraints) {
+        if (A->GetFixedBlockSize() == 1)
+          optimalSatisfyPConstraintsForScalarPDEs(finalP);
+        else
+          SatisfyPConstraints(A, finalP);
       }
+    }
 
-    } else {
-      finalP = Ptent;
+  } else {
+    finalP = Ptent;
+  }
+
+  // Level Set
+  if (!restrictionMode_) {
+    // The factory is in prolongation mode
+    if (!finalP.is_null()) {
+      std::ostringstream oss;
+      oss << "P_" << coarseLevel.GetLevelID();
+      finalP->setObjectLabel(oss.str());
     }
+    Set(coarseLevel, "P", finalP);
 
-    // Level Set
-    if (!restrictionMode_) {
-      // The factory is in prolongation mode
-      if(!finalP.is_null()) {std::ostringstream oss; oss << "P_" << coarseLevel.GetLevelID(); finalP->setObjectLabel(oss.str());}
-      Set(coarseLevel, "P",             finalP);
-    
-      APparams->set("graph", finalP);
-      Set(coarseLevel, "AP reuse data", APparams);
-
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        finalP->CreateView("stridedMaps", Ptent);
-
-      if (IsPrint(Statistics2)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*finalP, "P", params);
-      }
+    APparams->set("graph", finalP);
+    Set(coarseLevel, "AP reuse data", APparams);
 
-    } else {
-      // The factory is in restriction mode
-      RCP<Matrix> R;
-      {
-        SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
-        R = Utilities::Transpose(*finalP, true);
-        if(!R.is_null()) {std::ostringstream oss; oss << "R_" << coarseLevel.GetLevelID(); R->setObjectLabel(oss.str());}
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      finalP->CreateView("stridedMaps", Ptent);
+
+    if (IsPrint(Statistics2)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*finalP, "P", params);
+    }
+
+  } else {
+    // The factory is in restriction mode
+    RCP<Matrix> R;
+    {
+      SubFactoryMonitor m2(*this, "Transpose P", coarseLevel);
+      R = Utilities::Transpose(*finalP, true);
+      if (!R.is_null()) {
+        std::ostringstream oss;
+        oss << "R_" << coarseLevel.GetLevelID();
+        R->setObjectLabel(oss.str());
       }
+    }
 
-      Set(coarseLevel, "R", R);
+    Set(coarseLevel, "R", R);
 
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        R->CreateView("stridedMaps", Ptent, true/*transposeA*/);
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      R->CreateView("stridedMaps", Ptent, true /*transposeA*/);
 
-      if (IsPrint(Statistics2)) {
-        RCP<ParameterList> params = rcp(new ParameterList());
-        params->set("printLoadBalancingInfo", true);
-        params->set("printCommInfo",          true);
-        GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
-      }
+    if (IsPrint(Statistics2)) {
+      RCP<ParameterList> params = rcp(new ParameterList());
+      params->set("printLoadBalancingInfo", true);
+      params->set("printCommInfo", true);
+      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*R, "R", params);
     }
+  }
 
-  } //Build()
-
-  // Analyze the grid transfer produced by smoothed aggregation and make
-  // modifications if it does not look right. In particular, if there are
-  // negative entries or entries larger than 1, modify P's rows. 
-  //
-  // Note: this kind of evaluation probably only makes sense if not doing QR
-  // when constructing tentative P.
-  //
-  // For entries that do not satisfy the two constraints (>= 0  or <=1) we set
-  // these entries to the constraint value and modify the rest of the row
-  // so that the row sum remains the same as before by adding an equal
-  // amount to each remaining entry. However, if the original row sum value
-  // violates the constraints, we set the row sum back to 1 (the row sum of 
-  // tentative P). After doing the modification to a row, we need to check 
-  // again the entire row to make sure that the modified row does not violate
-  // the constraints.
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SatisfyPConstraints(const RCP<Matrix> A, RCP<Matrix>& P) const {
-
-    const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
-    LO nPDEs = A->GetFixedBlockSize();
-    Teuchos::ArrayRCP<Scalar> ConstraintViolationSum(nPDEs);
-    Teuchos::ArrayRCP<Scalar> Rsum(nPDEs);
-    Teuchos::ArrayRCP<size_t> nPositive(nPDEs);
-    for (size_t k=0; k < (size_t) nPDEs; k++) ConstraintViolationSum[k] = zero;
-    for (size_t k=0; k < (size_t) nPDEs; k++) Rsum[k] = zero;
-    for (size_t k=0; k < (size_t) nPDEs; k++) nPositive[k] = 0;
-
-
-    for (size_t i = 0; i < as<size_t>(P->getRowMap()->getLocalNumElements()); i++) {
-
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals1;
-      Teuchos::ArrayView<      Scalar> vals;
-      P->getLocalRowView((LO) i, indices, vals1);
-      size_t nnz = indices.size();
-      if (nnz == 0) continue;
-
-      vals = ArrayView<Scalar>(const_cast<SC*>(vals1.getRawPtr()), nnz);
-
-
-      bool checkRow = true;
-
-      while (checkRow) { 
-
-        // check constraints and compute the row sum
-    
-        for (LO j = 0; j < indices.size(); j++)  {
-          Rsum[ j%nPDEs ] += vals[j]; 
-          if (Teuchos::ScalarTraits<SC>::real(vals[j]) < Teuchos::ScalarTraits<SC>::real(zero)) { 
-            ConstraintViolationSum[ j%nPDEs ] += vals[j]; 
-            vals[j] = zero;
-          }
-          else {
-            if (Teuchos::ScalarTraits<SC>::real(vals[j]) != Teuchos::ScalarTraits<SC>::real(zero))
-              (nPositive[ j%nPDEs])++;
-    
-            if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(1.00001  )) { 
-              ConstraintViolationSum[ j%nPDEs ] += (vals[j] - one); 
-              vals[j] = one;
-            }
-          }
-        }
-    
-        checkRow = false;
+}  // Build()
 
-        // take into account any row sum that violates the contraints
-    
-        for (size_t k=0; k < (size_t) nPDEs; k++) {
+// Analyze the grid transfer produced by smoothed aggregation and make
+// modifications if it does not look right. In particular, if there are
+// negative entries or entries larger than 1, modify P's rows.
+//
+// Note: this kind of evaluation probably only makes sense if not doing QR
+// when constructing tentative P.
+//
+// For entries that do not satisfy the two constraints (>= 0  or <=1) we set
+// these entries to the constraint value and modify the rest of the row
+// so that the row sum remains the same as before by adding an equal
+// amount to each remaining entry. However, if the original row sum value
+// violates the constraints, we set the row sum back to 1 (the row sum of
+// tentative P). After doing the modification to a row, we need to check
+// again the entire row to make sure that the modified row does not violate
+// the constraints.
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SatisfyPConstraints(const RCP<Matrix> A, RCP<Matrix> &P) const {
+  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
+  LO nPDEs          = A->GetFixedBlockSize();
+  Teuchos::ArrayRCP<Scalar> ConstraintViolationSum(nPDEs);
+  Teuchos::ArrayRCP<Scalar> Rsum(nPDEs);
+  Teuchos::ArrayRCP<size_t> nPositive(nPDEs);
+  for (size_t k = 0; k < (size_t)nPDEs; k++) ConstraintViolationSum[k] = zero;
+  for (size_t k = 0; k < (size_t)nPDEs; k++) Rsum[k] = zero;
+  for (size_t k = 0; k < (size_t)nPDEs; k++) nPositive[k] = 0;
+
+  for (size_t i = 0; i < as<size_t>(P->getRowMap()->getLocalNumElements()); i++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals1;
+    Teuchos::ArrayView<Scalar> vals;
+    P->getLocalRowView((LO)i, indices, vals1);
+    size_t nnz = indices.size();
+    if (nnz == 0) continue;
+
+    vals = ArrayView<Scalar>(const_cast<SC *>(vals1.getRawPtr()), nnz);
+
+    bool checkRow = true;
+
+    while (checkRow) {
+      // check constraints and compute the row sum
+
+      for (LO j = 0; j < indices.size(); j++) {
+        Rsum[j % nPDEs] += vals[j];
+        if (Teuchos::ScalarTraits<SC>::real(vals[j]) < Teuchos::ScalarTraits<SC>::real(zero)) {
+          ConstraintViolationSum[j % nPDEs] += vals[j];
+          vals[j] = zero;
+        } else {
+          if (Teuchos::ScalarTraits<SC>::real(vals[j]) != Teuchos::ScalarTraits<SC>::real(zero))
+            (nPositive[j % nPDEs])++;
 
-          if (Teuchos::ScalarTraits<SC>::real(Rsum[ k ]) < Teuchos::ScalarTraits<SC>::magnitude(zero)) {
-              ConstraintViolationSum[k] +=  (-Rsum[k]);  // rstumin 
-          }
-          else if (Teuchos::ScalarTraits<SC>::real(Rsum[ k ]) > Teuchos::ScalarTraits<SC>::magnitude(1.00001)) {
-              ConstraintViolationSum[k] += (one - Rsum[k]);  // rstumin
+          if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(1.00001)) {
+            ConstraintViolationSum[j % nPDEs] += (vals[j] - one);
+            vals[j] = one;
           }
         }
+      }
 
-        // check if row need modification 
-        for (size_t k=0; k < (size_t) nPDEs; k++) {
-          if (Teuchos::ScalarTraits<SC>::magnitude(ConstraintViolationSum[ k ]) != Teuchos::ScalarTraits<SC>::magnitude(zero))
-             checkRow = true;
-        }
-        // modify row
-        if (checkRow) {
-           for (LO j = 0; j < indices.size(); j++)  {
-             if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(zero)) { 
-                vals[j] += (ConstraintViolationSum[j%nPDEs]/ (as<Scalar>(nPositive[j%nPDEs])));
-             }
-           }
-           for (size_t k=0; k < (size_t) nPDEs; k++) ConstraintViolationSum[k] = zero; 
-        }
-        for (size_t k=0; k < (size_t) nPDEs; k++) Rsum[k] = zero; 
-        for (size_t k=0; k < (size_t) nPDEs; k++) nPositive[k] = 0;
-      } // while (checkRow) ...
-    } // for (size_t i = 0; i < as<size_t>(P->getRowMap()->getNumNodeElements()); i++) ...
-  } //SatsifyPConstraints()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix>& P) const {
-
-    const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
-
-    LocalOrdinal  maxEntriesPerRow = 100; // increased later if needed 
-    Teuchos::ArrayRCP<Scalar> scalarData(3*maxEntriesPerRow);
-    bool hasFeasible;
-
-    for (size_t i = 0; i < as<size_t>(P->getRowMap()->getLocalNumElements()); i++) {
-
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals1;
-      Teuchos::ArrayView<      Scalar> vals;
-      P->getLocalRowView((LocalOrdinal) i, indices, vals1);
-      size_t nnz = indices.size();
-      if (nnz != 0) {
-
-        vals = ArrayView<Scalar>(const_cast<SC*>(vals1.getRawPtr()), nnz);
-        Scalar rsumTarget = zero;
-        for (size_t j = 0; j < nnz; j++) rsumTarget += vals[j];
-
-        if (nnz > as<size_t>(maxEntriesPerRow)) {
-          maxEntriesPerRow = nnz*3;
-          scalarData.resize(3*maxEntriesPerRow);
+      checkRow = false;
+
+      // take into account any row sum that violates the contraints
+
+      for (size_t k = 0; k < (size_t)nPDEs; k++) {
+        if (Teuchos::ScalarTraits<SC>::real(Rsum[k]) < Teuchos::ScalarTraits<SC>::magnitude(zero)) {
+          ConstraintViolationSum[k] += (-Rsum[k]);  // rstumin
+        } else if (Teuchos::ScalarTraits<SC>::real(Rsum[k]) > Teuchos::ScalarTraits<SC>::magnitude(1.00001)) {
+          ConstraintViolationSum[k] += (one - Rsum[k]);  // rstumin
         }
-        hasFeasible = constrainRow(vals.getRawPtr(), as<LocalOrdinal>(nnz), zero , one, rsumTarget, vals.getRawPtr(), scalarData.getRawPtr());
+      }
 
-        if (!hasFeasible) { // just set all entries to the same value giving a row sum of 1
-          for (size_t j = 0; j < nnz; j++) vals[j] = one/as<Scalar>(nnz);
+      // check if row need modification
+      for (size_t k = 0; k < (size_t)nPDEs; k++) {
+        if (Teuchos::ScalarTraits<SC>::magnitude(ConstraintViolationSum[k]) != Teuchos::ScalarTraits<SC>::magnitude(zero))
+          checkRow = true;
+      }
+      // modify row
+      if (checkRow) {
+        for (LO j = 0; j < indices.size(); j++) {
+          if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(zero)) {
+            vals[j] += (ConstraintViolationSum[j % nPDEs] / (as<Scalar>(nPositive[j % nPDEs])));
+          }
         }
+        for (size_t k = 0; k < (size_t)nPDEs; k++) ConstraintViolationSum[k] = zero;
+      }
+      for (size_t k = 0; k < (size_t)nPDEs; k++) Rsum[k] = zero;
+      for (size_t k = 0; k < (size_t)nPDEs; k++) nPositive[k] = 0;
+    }  // while (checkRow) ...
+  }    // for (size_t i = 0; i < as<size_t>(P->getRowMap()->getNumNodeElements()); i++) ...
+}  // SatsifyPConstraints()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix> &P) const {
+  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
+
+  LocalOrdinal maxEntriesPerRow = 100;  // increased later if needed
+  Teuchos::ArrayRCP<Scalar> scalarData(3 * maxEntriesPerRow);
+  bool hasFeasible;
+
+  for (size_t i = 0; i < as<size_t>(P->getRowMap()->getLocalNumElements()); i++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals1;
+    Teuchos::ArrayView<Scalar> vals;
+    P->getLocalRowView((LocalOrdinal)i, indices, vals1);
+    size_t nnz = indices.size();
+    if (nnz != 0) {
+      vals              = ArrayView<Scalar>(const_cast<SC *>(vals1.getRawPtr()), nnz);
+      Scalar rsumTarget = zero;
+      for (size_t j = 0; j < nnz; j++) rsumTarget += vals[j];
+
+      if (nnz > as<size_t>(maxEntriesPerRow)) {
+        maxEntriesPerRow = nnz * 3;
+        scalarData.resize(3 * maxEntriesPerRow);
       }
+      hasFeasible = constrainRow(vals.getRawPtr(), as<LocalOrdinal>(nnz), zero, one, rsumTarget, vals.getRawPtr(), scalarData.getRawPtr());
 
-    } // for (size_t i = 0; i < as<size_t>(P->getRowMap()->getNumNodeElements()); i++) ...
-  } //SatsifyPConstraints()
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::constrainRow(Scalar *orig, LocalOrdinal nEntries, Scalar leftBound, Scalar rghtBound,Scalar rsumTarget, Scalar *fixedUnsorted, Scalar *scalarData) const {
-/*
-   Input
-     orig          data that should be adjusted to satisfy bound constraints and
-                   row sum constraint. orig is not modified by this function.
-
-     nEntries      length or 'orig'
-
-     leftBound,    define bound constraints for the results.
-     rghtBound
-
-     rsumTarget    defines an equality constraint for the row sum
-
-     fixedUnsorted  on output, if a feasible solutuion exists then
-                    || orig - fixedUnsorted || = min  when also
-                    leftBound <= fixedUnsorted[i] <= rghtBound for all i 
-                    and  sum(fixedUnsorted) = rsumTarget.
-
-                    Note: it is possible to use the same pointer for 
-                    fixedUnsorted and orig. In this case, orig gets
-                    overwritten with the new constraint satisfying values.
- 
-      scalarData    a work array that should be 3x nEntries.
-
-   On return constrain() indicates whether or not a feasible solution exists.
-*/
-
-/*
-  Given a sequence of numbers    o1  ... on, fix these so that they are as 
-  close as possible to the original but satisfy bound constraints and also 
-  have the same row sum as the oi's. If we know who is going to lie on a  
-  bound, then the "best" answer (i.e., || o - f ||_2 = min)  perturbs 
-  each element that doesn't lie on a bound by the same amount. 
-   
-  We can represent the oi's by considering scattered points on a number line
-
-                 |                                                   |
-                 |                                                   |
-   o      o    o |  o        o     o           o       o     o       |o       o 
-                 |                                                   |
-                                                                      
-                 \____/                                              \____/
-                 <----                                               <----
-                 delta                                               delta
-
-  Bounds are shown by vertical lines. The fi's must lie within the bounds, so 
-  the 3 leftmost points must be shifted to the right and the 2 rightmost must
-  be shifted to the left. If these fi points are all shifted to the bounds
-  while the others remain in the same location, the row sum constraint is 
-  likely not satisfied and so more shifting is necessary.  In the figure, the f
-  rowsum is too large and so there must be more shifting to the left. 
-
-  To minimize || o - f ||_2, we basically shift all "interiors" by the same 
-  amount, denoted delta. The only trick is that some points near bounds are
-  still affected by the bounds and so these points might be shifted more or less
-  than delta. In the example,t he 3 rightmost points are shifted in the opposite
-  direction as delta to the bound. The 4th point is shifted by something less 
-  than delta so it does not violate the lower bound. The rightmost point is 
-  shifted to the bound by some amount larger than delta. However, the 2nd point
-  is shifted by delta (i.e., it lies inside the two bounds). 
-
-  If we know delta, we can figure out everything. If we know which points
-  are special (not shifted by delta), we can also figure out everything.
-  The problem is these two things (delta and the special points) are 
-  inter-connected. An algorithm for computing follows.
-
-  1) move exterior points to the bounds and compute how much the row sum is off
-     (rowSumDeviation). We assume now that the new row sum is high, so interior
-     points must be shifted left. 
-
-  2) Mark closest point just left of the leftmost bound, closestToLeftBound,
-     and compute its distance to the leftmost bound.  Mark closest point to the
-     left of the rightmost bound, closestToRghtBound, and compute its distance to
-     right bound. There are two cases to consider. 
-
-  3) Case 1: closestToLeftBound is closer than closestToRghtBound.
-     Assume that shifting by delta does not move closestToLeftBound past the 
-     left bound. This means that it will be shifted by delta. However, 
-     closestToRghtBound will be shifted by more than delta.  So the total
-     number of points shifted by delta (|interiorToBounds|) includes 
-     closestToLeftBound up to and including the point just to the left of 
-     closestToRghtBound. So 
-
-                  delta = rowSumDeviation/ |interiorToBounds| .
-
-     Recall that rowSumDeviation already accounts for the non-delta shift of 
-     of closestToRightBound.  Now check whether our assumption is valid.
-
-     If delta <= closestToLeftBoundDist, assumption is true so delta can be 
-     applied to interiorToBounds ... and we are done.
-     Else assumption is false. Shift closestToLeftBound to the left bound. 
-     Update rowSumDeviation, interiorToBounds, and identify new 
-     closestToLeftBound. Repeat step 3).
-
-     Case 2: closestToRghtBound is closer than closestToLeftBound.
-     Assume that shifting by delta does not move closestToRghtBound past right 
-     bound. This means that it must be shifted by more than delta to right
-     bound.  So the total number of points shifted by delta again includes
-     closestToLeftBound up to and including the point just to the left of 
-     closestToRghtBound. So again compute
-
-                  delta = rowSumDeviation/ |interiorToBounds| .
-
-     If delta <= closestToRghtBoundDist, assumption is true so delta is
-     can be applied to interiorToBounds ... and we are done
-     Else  assumption is false. Put closestToRghtBound in the
-     interiorToBounds set. Remove it's contribution to rowSumDeviation,
-     identify new closestToRghtBound. Repeat step 3)
-
-
-  To implement, sort the oi's so things like closestToLeftBound is just index
-  into sorted array.  Updaing closestToLeftBound or closestToRghtBound requires 
-  increment by 1.  |interiorToBounds|= closestToRghtBound -  closestToLeftBound 
-  To handle the case when the rowsum is low (requiring right interior shifts),
-  just flip the signs on data and use the left-shift code (and then flip back
-  before exiting function.
-*/
-  bool   hasFeasibleSol;
+      if (!hasFeasible) {  // just set all entries to the same value giving a row sum of 1
+        for (size_t j = 0; j < nnz; j++) vals[j] = one / as<Scalar>(nnz);
+      }
+    }
+
+  }  // for (size_t i = 0; i < as<size_t>(P->getRowMap()->getNumNodeElements()); i++) ...
+}  // SatsifyPConstraints()
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool SaPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::constrainRow(Scalar *orig, LocalOrdinal nEntries, Scalar leftBound, Scalar rghtBound, Scalar rsumTarget, Scalar *fixedUnsorted, Scalar *scalarData) const {
+  /*
+     Input
+       orig          data that should be adjusted to satisfy bound constraints and
+                     row sum constraint. orig is not modified by this function.
+
+       nEntries      length or 'orig'
+
+       leftBound,    define bound constraints for the results.
+       rghtBound
+
+       rsumTarget    defines an equality constraint for the row sum
+
+       fixedUnsorted  on output, if a feasible solutuion exists then
+                      || orig - fixedUnsorted || = min  when also
+                      leftBound <= fixedUnsorted[i] <= rghtBound for all i
+                      and  sum(fixedUnsorted) = rsumTarget.
+
+                      Note: it is possible to use the same pointer for
+                      fixedUnsorted and orig. In this case, orig gets
+                      overwritten with the new constraint satisfying values.
+
+        scalarData    a work array that should be 3x nEntries.
+
+     On return constrain() indicates whether or not a feasible solution exists.
+  */
+
+  /*
+    Given a sequence of numbers    o1  ... on, fix these so that they are as
+    close as possible to the original but satisfy bound constraints and also
+    have the same row sum as the oi's. If we know who is going to lie on a
+    bound, then the "best" answer (i.e., || o - f ||_2 = min)  perturbs
+    each element that doesn't lie on a bound by the same amount.
+
+    We can represent the oi's by considering scattered points on a number line
+
+                   |                                                   |
+                   |                                                   |
+     o      o    o |  o        o     o           o       o     o       |o       o
+                   |                                                   |
+
+                   \____/                                              \____/
+                   <----                                               <----
+                   delta                                               delta
+
+    Bounds are shown by vertical lines. The fi's must lie within the bounds, so
+    the 3 leftmost points must be shifted to the right and the 2 rightmost must
+    be shifted to the left. If these fi points are all shifted to the bounds
+    while the others remain in the same location, the row sum constraint is
+    likely not satisfied and so more shifting is necessary.  In the figure, the f
+    rowsum is too large and so there must be more shifting to the left.
+
+    To minimize || o - f ||_2, we basically shift all "interiors" by the same
+    amount, denoted delta. The only trick is that some points near bounds are
+    still affected by the bounds and so these points might be shifted more or less
+    than delta. In the example,t he 3 rightmost points are shifted in the opposite
+    direction as delta to the bound. The 4th point is shifted by something less
+    than delta so it does not violate the lower bound. The rightmost point is
+    shifted to the bound by some amount larger than delta. However, the 2nd point
+    is shifted by delta (i.e., it lies inside the two bounds).
+
+    If we know delta, we can figure out everything. If we know which points
+    are special (not shifted by delta), we can also figure out everything.
+    The problem is these two things (delta and the special points) are
+    inter-connected. An algorithm for computing follows.
+
+    1) move exterior points to the bounds and compute how much the row sum is off
+       (rowSumDeviation). We assume now that the new row sum is high, so interior
+       points must be shifted left.
+
+    2) Mark closest point just left of the leftmost bound, closestToLeftBound,
+       and compute its distance to the leftmost bound.  Mark closest point to the
+       left of the rightmost bound, closestToRghtBound, and compute its distance to
+       right bound. There are two cases to consider.
+
+    3) Case 1: closestToLeftBound is closer than closestToRghtBound.
+       Assume that shifting by delta does not move closestToLeftBound past the
+       left bound. This means that it will be shifted by delta. However,
+       closestToRghtBound will be shifted by more than delta.  So the total
+       number of points shifted by delta (|interiorToBounds|) includes
+       closestToLeftBound up to and including the point just to the left of
+       closestToRghtBound. So
+
+                    delta = rowSumDeviation/ |interiorToBounds| .
+
+       Recall that rowSumDeviation already accounts for the non-delta shift of
+       of closestToRightBound.  Now check whether our assumption is valid.
+
+       If delta <= closestToLeftBoundDist, assumption is true so delta can be
+       applied to interiorToBounds ... and we are done.
+       Else assumption is false. Shift closestToLeftBound to the left bound.
+       Update rowSumDeviation, interiorToBounds, and identify new
+       closestToLeftBound. Repeat step 3).
+
+       Case 2: closestToRghtBound is closer than closestToLeftBound.
+       Assume that shifting by delta does not move closestToRghtBound past right
+       bound. This means that it must be shifted by more than delta to right
+       bound.  So the total number of points shifted by delta again includes
+       closestToLeftBound up to and including the point just to the left of
+       closestToRghtBound. So again compute
+
+                    delta = rowSumDeviation/ |interiorToBounds| .
+
+       If delta <= closestToRghtBoundDist, assumption is true so delta is
+       can be applied to interiorToBounds ... and we are done
+       Else  assumption is false. Put closestToRghtBound in the
+       interiorToBounds set. Remove it's contribution to rowSumDeviation,
+       identify new closestToRghtBound. Repeat step 3)
+
+
+    To implement, sort the oi's so things like closestToLeftBound is just index
+    into sorted array.  Updaing closestToLeftBound or closestToRghtBound requires
+    increment by 1.  |interiorToBounds|= closestToRghtBound -  closestToLeftBound
+    To handle the case when the rowsum is low (requiring right interior shifts),
+    just flip the signs on data and use the left-shift code (and then flip back
+    before exiting function.
+  */
+  bool hasFeasibleSol;
   Scalar notFlippedLeftBound, notFlippedRghtBound, aBigNumber, *origSorted;
   Scalar rowSumDeviation, temp, *fixedSorted, delta;
   Scalar closestToLeftBoundDist, closestToRghtBoundDist;
-  LocalOrdinal    closestToLeftBound, closestToRghtBound;
-  LocalOrdinal    *inds; 
-  bool   flipped;
+  LocalOrdinal closestToLeftBound, closestToRghtBound;
+  LocalOrdinal *inds;
+  bool flipped;
 
-  notFlippedLeftBound = leftBound; 
-  notFlippedRghtBound = rghtBound; 
+  notFlippedLeftBound = leftBound;
+  notFlippedRghtBound = rghtBound;
 
-  if ((Teuchos::ScalarTraits<SC>::real(rsumTarget) >= Teuchos::ScalarTraits<SC>::real(leftBound*as<Scalar>(nEntries))) && 
-      (Teuchos::ScalarTraits<SC>::real(rsumTarget) <= Teuchos::ScalarTraits<SC>::real(rghtBound*as<Scalar>(nEntries))))
-    hasFeasibleSol = true; 
+  if ((Teuchos::ScalarTraits<SC>::real(rsumTarget) >= Teuchos::ScalarTraits<SC>::real(leftBound * as<Scalar>(nEntries))) &&
+      (Teuchos::ScalarTraits<SC>::real(rsumTarget) <= Teuchos::ScalarTraits<SC>::real(rghtBound * as<Scalar>(nEntries))))
+    hasFeasibleSol = true;
   else {
-    hasFeasibleSol=false; 
-    return hasFeasibleSol; 
+    hasFeasibleSol = false;
+    return hasFeasibleSol;
   }
-  flipped    = false;
+  flipped = false;
   // compute aBigNumber to handle some corner cases where we need
   // something large so that an if statement will be false
   aBigNumber = Teuchos::ScalarTraits<SC>::zero();
   for (LocalOrdinal i = 0; i < nEntries; i++) {
-    if ( Teuchos::ScalarTraits<SC>::magnitude(orig[i]) > Teuchos::ScalarTraits<SC>::magnitude(aBigNumber)) 
+    if (Teuchos::ScalarTraits<SC>::magnitude(orig[i]) > Teuchos::ScalarTraits<SC>::magnitude(aBigNumber))
       aBigNumber = Teuchos::ScalarTraits<SC>::magnitude(orig[i]);
   }
-  aBigNumber = aBigNumber+ (Teuchos::ScalarTraits<SC>::magnitude(leftBound) + Teuchos::ScalarTraits<SC>::magnitude(rghtBound))*as<Scalar>(100.0);
+  aBigNumber = aBigNumber + (Teuchos::ScalarTraits<SC>::magnitude(leftBound) + Teuchos::ScalarTraits<SC>::magnitude(rghtBound)) * as<Scalar>(100.0);
 
   origSorted  = &scalarData[0];
   fixedSorted = &(scalarData[nEntries]);
-  inds        = (LocalOrdinal    *) &(scalarData[2*nEntries]);
+  inds        = (LocalOrdinal *)&(scalarData[2 * nEntries]);
 
-  for (LocalOrdinal i = 0; i < nEntries; i++) inds[i] = i; 
-  for (LocalOrdinal i = 0; i < nEntries; i++) origSorted[i] = orig[i];  /* orig no longer used */
+  for (LocalOrdinal i = 0; i < nEntries; i++) inds[i] = i;
+  for (LocalOrdinal i = 0; i < nEntries; i++) origSorted[i] = orig[i]; /* orig no longer used */
 
-  // sort so that  orig[inds] is sorted. 
-  std::sort(inds, inds+nEntries,
-            [origSorted](LocalOrdinal leftIndex, LocalOrdinal rightIndex)
-                 { return Teuchos::ScalarTraits<SC>::real(origSorted[leftIndex]) < Teuchos::ScalarTraits<SC>::real(origSorted[rightIndex]);});
+  // sort so that  orig[inds] is sorted.
+  std::sort(inds, inds + nEntries,
+            [origSorted](LocalOrdinal leftIndex, LocalOrdinal rightIndex) { return Teuchos::ScalarTraits<SC>::real(origSorted[leftIndex]) < Teuchos::ScalarTraits<SC>::real(origSorted[rightIndex]); });
 
   for (LocalOrdinal i = 0; i < nEntries; i++) origSorted[i] = orig[inds[i]];
   // find entry in origSorted just to the right of the leftBound
@@ -598,84 +594,92 @@ namespace MueLu {
   closestToRghtBound = closestToLeftBound;
   while ((closestToRghtBound < nEntries) && (Teuchos::ScalarTraits<SC>::real(origSorted[closestToRghtBound]) <= Teuchos::ScalarTraits<SC>::real(rghtBound))) closestToRghtBound++;
 
-  // compute distance between closestToLeftBound and the left bound and the 
-  // distance between closestToRghtBound and the right bound. 
+  // compute distance between closestToLeftBound and the left bound and the
+  // distance between closestToRghtBound and the right bound.
 
   closestToLeftBoundDist = origSorted[closestToLeftBound] - leftBound;
-  if (closestToRghtBound==nEntries) closestToRghtBoundDist= aBigNumber;
-  else                              closestToRghtBoundDist= origSorted[closestToRghtBound] - rghtBound;
+  if (closestToRghtBound == nEntries)
+    closestToRghtBoundDist = aBigNumber;
+  else
+    closestToRghtBoundDist = origSorted[closestToRghtBound] - rghtBound;
 
   // compute how far the rowSum is off from the target row sum taking into account
   // numbers that have been shifted to satisfy bound constraint
 
-  rowSumDeviation = leftBound*as<Scalar>(closestToLeftBound) + as<Scalar>((nEntries-closestToRghtBound))*rghtBound - rsumTarget; 
-  for (LocalOrdinal i=closestToLeftBound; i < closestToRghtBound; i++) rowSumDeviation += origSorted[i];
+  rowSumDeviation = leftBound * as<Scalar>(closestToLeftBound) + as<Scalar>((nEntries - closestToRghtBound)) * rghtBound - rsumTarget;
+  for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound; i++) rowSumDeviation += origSorted[i];
 
   // the code that follow after this if statement assumes that rowSumDeviation is positive. If this
-  // is not the case, flip the signs of everything so that rowSumDeviation is now positive. 
+  // is not the case, flip the signs of everything so that rowSumDeviation is now positive.
   // Later we will flip the data back to its original form.
   if (Teuchos::ScalarTraits<SC>::real(rowSumDeviation) < Teuchos::ScalarTraits<SC>::real(Teuchos::ScalarTraits<SC>::zero())) {
-    flipped = true;
-    temp = leftBound; leftBound = -rghtBound; rghtBound = temp; 
+    flipped   = true;
+    temp      = leftBound;
+    leftBound = -rghtBound;
+    rghtBound = temp;
 
     /* flip sign of origSorted and reverse ordering so that the negative version is sorted */
 
-    if ((nEntries%2) == 1) origSorted[(nEntries/2)  ] =  -origSorted[(nEntries/2)  ];
-    for (LocalOrdinal i=0; i < nEntries/2; i++) {
-      temp=origSorted[i];
-      origSorted[i] = -origSorted[nEntries-1-i]; 
-      origSorted[nEntries-i-1] = -temp;
+    if ((nEntries % 2) == 1) origSorted[(nEntries / 2)] = -origSorted[(nEntries / 2)];
+    for (LocalOrdinal i = 0; i < nEntries / 2; i++) {
+      temp                         = origSorted[i];
+      origSorted[i]                = -origSorted[nEntries - 1 - i];
+      origSorted[nEntries - i - 1] = -temp;
     }
-  
+
     /* reverse bounds */
 
-    LocalOrdinal itemp = closestToLeftBound; 
-    closestToLeftBound = nEntries-closestToRghtBound;
-    closestToRghtBound = nEntries-itemp; 
+    LocalOrdinal itemp     = closestToLeftBound;
+    closestToLeftBound     = nEntries - closestToRghtBound;
+    closestToRghtBound     = nEntries - itemp;
     closestToLeftBoundDist = origSorted[closestToLeftBound] - leftBound;
-    if (closestToRghtBound==nEntries) closestToRghtBoundDist= aBigNumber;
-    else                              closestToRghtBoundDist= origSorted[closestToRghtBound] - rghtBound;
+    if (closestToRghtBound == nEntries)
+      closestToRghtBoundDist = aBigNumber;
+    else
+      closestToRghtBoundDist = origSorted[closestToRghtBound] - rghtBound;
 
     rowSumDeviation = -rowSumDeviation;
   }
 
   // initial fixedSorted so bounds are satisfied and interiors correspond to origSorted
 
-  for (LocalOrdinal i =                  0; i < closestToLeftBound; i++) fixedSorted[i] = leftBound;
+  for (LocalOrdinal i = 0; i < closestToLeftBound; i++) fixedSorted[i] = leftBound;
   for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted[i] = origSorted[i];
-  for (LocalOrdinal i = closestToRghtBound; i < nEntries;           i++) fixedSorted[i] = rghtBound;
+  for (LocalOrdinal i = closestToRghtBound; i < nEntries; i++) fixedSorted[i] = rghtBound;
 
-  while ((Teuchos::ScalarTraits<SC>::magnitude(rowSumDeviation) > Teuchos::ScalarTraits<SC>::magnitude(as<Scalar>(1.e-10)*rsumTarget))){ // && ( (closestToLeftBound < nEntries ) || (closestToRghtBound < nEntries))) {
-   if (closestToRghtBound !=  closestToLeftBound)
-        delta = rowSumDeviation/ as<Scalar>(closestToRghtBound -  closestToLeftBound);
-   else delta = aBigNumber; 
+  while ((Teuchos::ScalarTraits<SC>::magnitude(rowSumDeviation) > Teuchos::ScalarTraits<SC>::magnitude(as<Scalar>(1.e-10) * rsumTarget))) {  // && ( (closestToLeftBound < nEntries ) || (closestToRghtBound < nEntries))) {
+    if (closestToRghtBound != closestToLeftBound)
+      delta = rowSumDeviation / as<Scalar>(closestToRghtBound - closestToLeftBound);
+    else
+      delta = aBigNumber;
 
-   if (Teuchos::ScalarTraits<SC>::magnitude(closestToLeftBoundDist) <= Teuchos::ScalarTraits<SC>::magnitude(closestToRghtBoundDist)) {
+    if (Teuchos::ScalarTraits<SC>::magnitude(closestToLeftBoundDist) <= Teuchos::ScalarTraits<SC>::magnitude(closestToRghtBoundDist)) {
       if (Teuchos::ScalarTraits<SC>::magnitude(delta) <= Teuchos::ScalarTraits<SC>::magnitude(closestToLeftBoundDist)) {
-        rowSumDeviation = Teuchos::ScalarTraits<SC>::zero(); 
-        for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound ; i++) fixedSorted[i] = origSorted[i] - delta;
-      }
-      else {
-        rowSumDeviation = rowSumDeviation - closestToLeftBoundDist; 
+        rowSumDeviation = Teuchos::ScalarTraits<SC>::zero();
+        for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted[i] = origSorted[i] - delta;
+      } else {
+        rowSumDeviation                 = rowSumDeviation - closestToLeftBoundDist;
         fixedSorted[closestToLeftBound] = leftBound;
         closestToLeftBound++;
-        if (closestToLeftBound < nEntries) closestToLeftBoundDist = origSorted[closestToLeftBound] - leftBound;
-        else                               closestToLeftBoundDist = aBigNumber;
+        if (closestToLeftBound < nEntries)
+          closestToLeftBoundDist = origSorted[closestToLeftBound] - leftBound;
+        else
+          closestToLeftBoundDist = aBigNumber;
       }
-   }
-   else {
+    } else {
       if (Teuchos::ScalarTraits<SC>::magnitude(delta) <= Teuchos::ScalarTraits<SC>::magnitude(closestToRghtBoundDist)) {
-        rowSumDeviation = 0; 
-        for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound ; i++) fixedSorted[i] = origSorted[i] - delta;
-      }
-      else {
+        rowSumDeviation = 0;
+        for (LocalOrdinal i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted[i] = origSorted[i] - delta;
+      } else {
         rowSumDeviation = rowSumDeviation + closestToRghtBoundDist;
-//        if (closestToRghtBound < nEntries) { 
-          fixedSorted[closestToRghtBound] = origSorted[closestToRghtBound];
-          closestToRghtBound++;
- //       }
-        if (closestToRghtBound >= nEntries) closestToRghtBoundDist = aBigNumber;
-        else                                closestToRghtBoundDist = origSorted[closestToRghtBound] - rghtBound;
+        //        if (closestToRghtBound < nEntries) {
+        fixedSorted[closestToRghtBound] = origSorted[closestToRghtBound];
+        closestToRghtBound++;
+        //       }
+        if (closestToRghtBound >= nEntries)
+          closestToRghtBoundDist = aBigNumber;
+        else
+          closestToRghtBoundDist = origSorted[closestToRghtBound] - rghtBound;
       }
     }
   }
@@ -683,40 +687,39 @@ namespace MueLu {
   if (flipped) {
     /* flip sign of fixedSorted and reverse ordering so that the positve version is sorted */
 
-    if ((nEntries%2) == 1) fixedSorted[(nEntries/2)  ] =  -fixedSorted[(nEntries/2)  ];
-    for (LocalOrdinal i=0; i < nEntries/2; i++) {
-      temp=fixedSorted[i];
-      fixedSorted[i] = -fixedSorted[nEntries-1-i]; 
-      fixedSorted[nEntries-i-1] = -temp;
+    if ((nEntries % 2) == 1) fixedSorted[(nEntries / 2)] = -fixedSorted[(nEntries / 2)];
+    for (LocalOrdinal i = 0; i < nEntries / 2; i++) {
+      temp                          = fixedSorted[i];
+      fixedSorted[i]                = -fixedSorted[nEntries - 1 - i];
+      fixedSorted[nEntries - i - 1] = -temp;
     }
   }
-  for (LocalOrdinal i = 0; i < nEntries; i++)  fixedUnsorted[inds[i]] = fixedSorted[i];
+  for (LocalOrdinal i = 0; i < nEntries; i++) fixedUnsorted[inds[i]] = fixedSorted[i];
 
   /* check that no constraints are violated */
 
   bool lowerViolation = false;
   bool upperViolation = false;
-  bool sumViolation = false;
-  using TST = Teuchos::ScalarTraits<SC>;
-  temp = TST::zero();
-  for (LocalOrdinal i = 0; i < nEntries; i++)  { 
+  bool sumViolation   = false;
+  using TST           = Teuchos::ScalarTraits<SC>;
+  temp                = TST::zero();
+  for (LocalOrdinal i = 0; i < nEntries; i++) {
     if (TST::real(fixedUnsorted[i]) < TST::real(notFlippedLeftBound)) lowerViolation = true;
     if (TST::real(fixedUnsorted[i]) > TST::real(notFlippedRghtBound)) upperViolation = true;
     temp += fixedUnsorted[i];
   }
-  SC tol = as<Scalar>(std::max(1.0e-8, as<double>(100*TST::eps())));
-  if (TST::magnitude(temp - rsumTarget) > TST::magnitude(tol*rsumTarget)) sumViolation = true;
+  SC tol = as<Scalar>(std::max(1.0e-8, as<double>(100 * TST::eps())));
+  if (TST::magnitude(temp - rsumTarget) > TST::magnitude(tol * rsumTarget)) sumViolation = true;
 
   TEUCHOS_TEST_FOR_EXCEPTION(lowerViolation, Exceptions::RuntimeError, "MueLu::SaPFactory::constrainRow: feasible solution but computation resulted in a lower bound violation??? ");
   TEUCHOS_TEST_FOR_EXCEPTION(upperViolation, Exceptions::RuntimeError, "MueLu::SaPFactory::constrainRow: feasible solution but computation resulted in an upper bound violation??? ");
-  TEUCHOS_TEST_FOR_EXCEPTION(sumViolation,   Exceptions::RuntimeError, "MueLu::SaPFactory::constrainRow: feasible solution but computation resulted in a row sum violation??? ");
+  TEUCHOS_TEST_FOR_EXCEPTION(sumViolation, Exceptions::RuntimeError, "MueLu::SaPFactory::constrainRow: feasible solution but computation resulted in a row sum violation??? ");
 
   return hasFeasibleSol;
 }
 
+}  // namespace MueLu
 
-} //namespace MueLu
-
-#endif // MUELU_SAPFACTORY_DEF_HPP
+#endif  // MUELU_SAPFACTORY_DEF_HPP
 
-//TODO: restrictionMode_ should use the parameter list.
+// TODO: restrictionMode_ should use the parameter list.
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_decl.hpp
index b5e553abd664..d9f7307a5289 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_decl.hpp
@@ -62,98 +62,96 @@
 
 namespace MueLu {
 
-  /*!
-    @class SaPFactory_kokkos class.
-    @brief Factory for building Smoothed Aggregation prolongators.
-    @ingroup MueLuTransferClasses
-
-    ## Input/output of SaPFactory_kokkos ##
-
-    ### User parameters of SaPFactory_kokkos ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | sa: damping factor                     | double  | 1.33  | * | * |   | Damping factor for smoothed aggregation transfer operators |
-    | sa: calculate eigenvalue estimate      | bool    | false | * | * |   | Force calculation of eigenvalue estimate during prolongator smoothing |
-    | sa: eigenvalue estimate num iterations | int     | 10    | * | * |   | Number of power iterations to estimate max eigenvalue |
-    | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
-    | P                                      | Factory | null  |   | * | * | Tentative prolongator factory (should be a TentativePFactory object). Note that if "P" is not provided the FactoryManager tries the factory which is generating the variable "Ptent" as input for "P". In the standard case using the default settings this is the non-smoothed tentative prolongation operator. |
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SaPFactory_kokkos::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see SaPFactory_kokkos::DeclareInput).
-
-    ### Variables provided by SaPFactory_kokkos ###
-
-    After SaPFactory_kokkos::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | SaPFactory_kokkos   | Smoothed prolongator
-
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SaPFactory_kokkos;
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
-  public:
-    typedef LocalOrdinal                                        local_ordinal_type;
-    typedef GlobalOrdinal                                       global_ordinal_type;
-    typedef typename DeviceType::execution_space                execution_space;
-    typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class SaPFactory_kokkos class.
+  @brief Factory for building Smoothed Aggregation prolongators.
+  @ingroup MueLuTransferClasses
+
+  ## Input/output of SaPFactory_kokkos ##
+
+  ### User parameters of SaPFactory_kokkos ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | sa: damping factor                     | double  | 1.33  | * | * |   | Damping factor for smoothed aggregation transfer operators |
+  | sa: calculate eigenvalue estimate      | bool    | false | * | * |   | Force calculation of eigenvalue estimate during prolongator smoothing |
+  | sa: eigenvalue estimate num iterations | int     | 10    | * | * |   | Number of power iterations to estimate max eigenvalue |
+  | A                                      | Factory | null  |   | * | * | Generating factory of the matrix A used during the prolongator smoothing process |
+  | P                                      | Factory | null  |   | * | * | Tentative prolongator factory (should be a TentativePFactory object). Note that if "P" is not provided the FactoryManager tries the factory which is generating the variable "Ptent" as input for "P". In the standard case using the default settings this is the non-smoothed tentative prolongation operator. |
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see SaPFactory_kokkos::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see SaPFactory_kokkos::DeclareInput).
+
+  ### Variables provided by SaPFactory_kokkos ###
+
+  After SaPFactory_kokkos::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | SaPFactory_kokkos   | Smoothed prolongator
+
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SaPFactory_kokkos;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
+ public:
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename DeviceType::execution_space execution_space;
+  typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_SAPFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
-
-    /*! @brief Constructor.
-      User can supply a factory for generating the tentative prolongator.
-    */
-    SaPFactory_kokkos() { }
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~SaPFactory_kokkos() { }
+  /*! @brief Constructor.
+    User can supply a factory for generating the tentative prolongator.
+  */
+  SaPFactory_kokkos() {}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //! Destructor.
+  virtual ~SaPFactory_kokkos() {}
 
-    //@}
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //@}
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  //! Input
+  //@{
 
-    //@}
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    /*!
-      @brief Build method.
+  //! @name Build methods.
+  //@{
 
-      Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
-      //FIXME what does the return code mean (unclear in MueMat)?
-      */
-    void Build (Level& fineLevel, Level& coarseLevel) const;
+  /*!
+    @brief Build method.
 
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+    Builds smoothed aggregation prolongator and returns it in <tt>coarseLevel</tt>.
+    //FIXME what does the return code mean (unclear in MueMat)?
+    */
+  void Build(Level& fineLevel, Level& coarseLevel) const;
 
-    void SatisfyPConstraints(RCP<Matrix> A, RCP<Matrix>& P) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    void optimalSatisfyPConstraintsForScalarPDEs( RCP<Matrix>& P) const;
+  void SatisfyPConstraints(RCP<Matrix> A, RCP<Matrix>& P) const;
 
-    //@}
+  void optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix>& P) const;
 
-  };
+  //@}
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SAPFACTORY_KOKKOS_SHORT
-#endif // MUELU_SAPFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_SAPFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_def.hpp
index a4bb8689f55c..8d9ba7b24f61 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_SaPFactory_kokkos_def.hpp
@@ -68,532 +68,531 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  RCP<const ParameterList> SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+RCP<const ParameterList> SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("sa: damping factor");
-    SET_VALID_ENTRY("sa: calculate eigenvalue estimate");
-    SET_VALID_ENTRY("sa: eigenvalue estimate num iterations");
-    SET_VALID_ENTRY("sa: use rowsumabs diagonal scaling");
-    SET_VALID_ENTRY("sa: enforce constraints");
-    SET_VALID_ENTRY("tentative: calculate qr");
-    SET_VALID_ENTRY("sa: max eigenvalue");
-#undef  SET_VALID_ENTRY
+  SET_VALID_ENTRY("sa: damping factor");
+  SET_VALID_ENTRY("sa: calculate eigenvalue estimate");
+  SET_VALID_ENTRY("sa: eigenvalue estimate num iterations");
+  SET_VALID_ENTRY("sa: use rowsumabs diagonal scaling");
+  SET_VALID_ENTRY("sa: enforce constraints");
+  SET_VALID_ENTRY("tentative: calculate qr");
+  SET_VALID_ENTRY("sa: max eigenvalue");
+#undef SET_VALID_ENTRY
+
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
+  validParamList->set<RCP<const FactoryBase>>("P", Teuchos::null, "Tentative prolongator factory");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "A");
+
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory_kokkos and NullspaceFactory.
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  }
+  coarseLevel.DeclareInput("P", initialPFact.get(), this);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Prolongator smoothing", coarseLevel);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType Magnitude;
+
+  // Get default tentative prolongator factory
+  // Getting it that way ensure that the same factory instance will be used for both SaPFactory_kokkos and NullspaceFactory.
+  // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
+  RCP<const FactoryBase> initialPFact = GetFactory("P");
+  if (initialPFact == Teuchos::null) {
+    initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent");
+  }
+  const ParameterList& pL = GetParameterList();
+
+  // Level Get
+  RCP<Matrix> A     = Get<RCP<Matrix>>(fineLevel, "A");
+  RCP<Matrix> Ptent = coarseLevel.Get<RCP<Matrix>>("P", initialPFact.get());
+  RCP<Matrix> finalP;
+  // If Tentative facctory bailed out (e.g., number of global aggregates is 0), then SaPFactory bails
+  //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (Ptent == Teuchos::null) {
+    finalP = Teuchos::null;
+    Set(coarseLevel, "P", finalP);
+    return;
+  }
 
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A used during the prolongator smoothing process");
-    validParamList->set< RCP<const FactoryBase> >("P", Teuchos::null, "Tentative prolongator factory");
+  if (restrictionMode_) {
+    SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
 
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+    A = Utilities::Transpose(*A, true);  // build transpose of A explicitly
+  }
 
+  // Build final prolongator
 
-    return validParamList;
-  }
+  // Reuse pattern if available
+  RCP<ParameterList> APparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    APparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    APparams = rcp(new ParameterList);
+  if (coarseLevel.IsAvailable("AP reuse data", this)) {
+    GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
-    Input(fineLevel, "A");
+    APparams = coarseLevel.Get<RCP<ParameterList>>("AP reuse data", this);
 
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory_kokkos and NullspaceFactory.
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    coarseLevel.DeclareInput("P", initialPFact.get(), this);
+    if (APparams->isParameter("graph"))
+      finalP = APparams->get<RCP<Matrix>>("graph");
   }
+  // By default, we don't need global constants for SaP
+  APparams->set("compute global constants: temporaries", APparams->get("compute global constants: temporaries", false));
+  APparams->set("compute global constants", APparams->get("compute global constants", false));
+
+  const SC dampingFactor        = as<SC>(pL.get<double>("sa: damping factor"));
+  const LO maxEigenIterations   = as<LO>(pL.get<int>("sa: eigenvalue estimate num iterations"));
+  const bool estimateMaxEigen   = pL.get<bool>("sa: calculate eigenvalue estimate");
+  const bool useAbsValueRowSum  = pL.get<bool>("sa: use rowsumabs diagonal scaling");
+  const bool doQRStep           = pL.get<bool>("tentative: calculate qr");
+  const bool enforceConstraints = pL.get<bool>("sa: enforce constraints");
+  const SC userDefinedMaxEigen  = as<SC>(pL.get<double>("sa: max eigenvalue"));
+
+  // Sanity checking
+  TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && enforceConstraints, Exceptions::RuntimeError,
+                             "MueLu::TentativePFactory::MakeTentative: cannot use 'enforce constraints' and 'calculate qr' at the same time");
+
+  if (dampingFactor != Teuchos::ScalarTraits<SC>::zero()) {
+    SC lambdaMax;
+    RCP<Vector> invDiag;
+    if (Teuchos::ScalarTraits<SC>::real(userDefinedMaxEigen) == Teuchos::ScalarTraits<SC>::real(-1.0)) {
+      SubFactoryMonitor m2(*this, "Eigenvalue estimate", coarseLevel);
+      lambdaMax = A->GetMaxEigenvalueEstimate();
+      if (lambdaMax == -Teuchos::ScalarTraits<SC>::one() || estimateMaxEigen) {
+        GetOStream(Statistics1) << "Calculating max eigenvalue estimate now (max iters = " << maxEigenIterations << ((useAbsValueRowSum) ? ", use rowSumAbs diagonal)" : ", use point diagonal)") << std::endl;
+        Magnitude stopTol = 1e-4;
+        invDiag           = Utilities::GetMatrixDiagonalInverse(*A, Teuchos::ScalarTraits<SC>::eps() * 100, Teuchos::ScalarTraits<SC>::zero(), useAbsValueRowSum);
+        if (useAbsValueRowSum)
+          lambdaMax = Utilities::PowerMethod(*A, invDiag, maxEigenIterations, stopTol);
+        else
+          lambdaMax = Utilities::PowerMethod(*A, true, maxEigenIterations, stopTol);
+        A->SetMaxEigenvalueEstimate(lambdaMax);
+      } else {
+        GetOStream(Statistics1) << "Using cached max eigenvalue estimate" << std::endl;
+      }
+    } else {
+      lambdaMax = userDefinedMaxEigen;
+      A->SetMaxEigenvalueEstimate(lambdaMax);
+    }
+    GetOStream(Statistics0) << "Prolongator damping factor = " << dampingFactor / lambdaMax << " (" << dampingFactor << " / " << lambdaMax << ")" << std::endl;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
+    {
+      SubFactoryMonitor m2(*this, "Fused (I-omega*D^{-1} A)*Ptent", coarseLevel);
+      {
+        SubFactoryMonitor m3(*this, "Diagonal Extraction", coarseLevel);
+        if (useAbsValueRowSum)
+          GetOStream(Runtime0) << "Using rowSumAbs diagonal" << std::endl;
+        if (invDiag == Teuchos::null)
+          invDiag = Utilities::GetMatrixDiagonalInverse(*A, Teuchos::ScalarTraits<SC>::eps() * 100, Teuchos::ScalarTraits<SC>::zero(), useAbsValueRowSum);
+      }
+      SC omega = dampingFactor / lambdaMax;
+      TEUCHOS_TEST_FOR_EXCEPTION(!std::isfinite(Teuchos::ScalarTraits<SC>::magnitude(omega)), Exceptions::RuntimeError, "Prolongator damping factor needs to be finite.");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Prolongator smoothing", coarseLevel);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType Magnitude;
-
-    // Get default tentative prolongator factory
-    // Getting it that way ensure that the same factory instance will be used for both SaPFactory_kokkos and NullspaceFactory.
-    // -- Warning: Do not use directly initialPFact_. Use initialPFact instead everywhere!
-    RCP<const FactoryBase> initialPFact = GetFactory("P");
-    if (initialPFact == Teuchos::null) { initialPFact = coarseLevel.GetFactoryManager()->GetFactory("Ptent"); }
-    const ParameterList& pL = GetParameterList();
-
-    // Level Get
-    RCP<Matrix> A     = Get< RCP<Matrix> >(fineLevel, "A");
-    RCP<Matrix> Ptent = coarseLevel.Get< RCP<Matrix> >("P", initialPFact.get());
-    RCP<Matrix> finalP;
-    // If Tentative facctory bailed out (e.g., number of global aggregates is 0), then SaPFactory bails
-    //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-    if (Ptent == Teuchos::null) {
-      finalP = Teuchos::null;
-      Set(coarseLevel, "P", finalP);
-      return;
+      {
+        SubFactoryMonitor m3(*this, "Xpetra::IteratorOps::Jacobi", coarseLevel);
+        // finalP = Ptent + (I - \omega D^{-1}A) Ptent
+        finalP = Xpetra::IteratorOps<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Jacobi(omega, *invDiag, *A, *Ptent, finalP, GetOStream(Statistics2), std::string("MueLu::SaP-") + toString(coarseLevel.GetLevelID()), APparams);
+        if (enforceConstraints) SatisfyPConstraints(A, finalP);
+      }
     }
 
-    if(restrictionMode_) {
-      SubFactoryMonitor m2(*this, "Transpose A", coarseLevel);
+  } else {
+    finalP = Ptent;
+  }
 
-      A = Utilities::Transpose(*A, true); // build transpose of A explicitly
+  // Level Set
+  if (!restrictionMode_) {
+    // prolongation factory is in prolongation mode
+    if (!finalP.is_null()) {
+      std::ostringstream oss;
+      oss << "P_" << coarseLevel.GetLevelID();
+      finalP->setObjectLabel(oss.str());
+    }
+    Set(coarseLevel, "P", finalP);
+
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      finalP->CreateView("stridedMaps", Ptent);
+
+  } else {
+    // prolongation factory is in restriction mode
+    RCP<Matrix> R = Utilities::Transpose(*finalP, true);
+    Set(coarseLevel, "R", R);
+    if (!R.is_null()) {
+      std::ostringstream oss;
+      oss << "R_" << coarseLevel.GetLevelID();
+      R->setObjectLabel(oss.str());
     }
 
-    //Build final prolongator
+    // NOTE: EXPERIMENTAL
+    if (Ptent->IsView("stridedMaps"))
+      R->CreateView("stridedMaps", Ptent, true);
+  }
 
-    // Reuse pattern if available
-    RCP<ParameterList> APparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      APparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-    else
-      APparams= rcp(new ParameterList);
-    if (coarseLevel.IsAvailable("AP reuse data", this)) {
-      GetOStream(static_cast<MsgType>(Runtime0 | Test)) << "Reusing previous AP data" << std::endl;
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    params->set("printCommInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*finalP, (!restrictionMode_ ? "P" : "R"), params);
+  }
 
-      APparams = coarseLevel.Get< RCP<ParameterList> >("AP reuse data", this);
+}  // Build()
 
-      if (APparams->isParameter("graph"))
-        finalP = APparams->get< RCP<Matrix> >("graph");
-    }
-    // By default, we don't need global constants for SaP
-    APparams->set("compute global constants: temporaries",APparams->get("compute global constants: temporaries",false));
-    APparams->set("compute global constants",APparams->get("compute global constants",false));
-
-    const SC dampingFactor      = as<SC>(pL.get<double>("sa: damping factor"));
-    const LO maxEigenIterations = as<LO>(pL.get<int>("sa: eigenvalue estimate num iterations"));
-    const bool estimateMaxEigen = pL.get<bool>("sa: calculate eigenvalue estimate");
-    const bool useAbsValueRowSum = pL.get<bool>  ("sa: use rowsumabs diagonal scaling");
-    const bool doQRStep         =        pL.get<bool>("tentative: calculate qr");
-    const bool enforceConstraints=       pL.get<bool>("sa: enforce constraints");
-    const SC   userDefinedMaxEigen =  as<SC>(pL.get<double>("sa: max eigenvalue"));
-
-    // Sanity checking
-    TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && enforceConstraints,Exceptions::RuntimeError,
-                               "MueLu::TentativePFactory::MakeTentative: cannot use 'enforce constraints' and 'calculate qr' at the same time");
-
-    if (dampingFactor != Teuchos::ScalarTraits<SC>::zero()) {
-
-      SC lambdaMax;
-      RCP<Vector> invDiag;
-      if (Teuchos::ScalarTraits<SC>::real(userDefinedMaxEigen) == Teuchos::ScalarTraits<SC>::real(-1.0))
-      {
-        SubFactoryMonitor m2(*this, "Eigenvalue estimate", coarseLevel);
-        lambdaMax = A->GetMaxEigenvalueEstimate();
-        if (lambdaMax == -Teuchos::ScalarTraits<SC>::one() || estimateMaxEigen) {
-          GetOStream(Statistics1) << "Calculating max eigenvalue estimate now (max iters = "<< maxEigenIterations <<
-          ( (useAbsValueRowSum) ?  ", use rowSumAbs diagonal)" :  ", use point diagonal)") << std::endl;
-          Magnitude stopTol = 1e-4;
-          invDiag = Utilities::GetMatrixDiagonalInverse(*A, Teuchos::ScalarTraits<SC>::eps()*100, Teuchos::ScalarTraits<SC>::zero(), useAbsValueRowSum);
-          if (useAbsValueRowSum)
-            lambdaMax = Utilities::PowerMethod(*A, invDiag, maxEigenIterations, stopTol);
-          else
-            lambdaMax = Utilities::PowerMethod(*A, true, maxEigenIterations, stopTol);
-          A->SetMaxEigenvalueEstimate(lambdaMax);
+// Analyze the grid transfer produced by smoothed aggregation and make
+// modifications if it does not look right. In particular, if there are
+// negative entries or entries larger than 1, modify P's rows.
+//
+// Note: this kind of evaluation probably only makes sense if not doing QR
+// when constructing tentative P.
+//
+// For entries that do not satisfy the two constraints (>= 0  or <=1) we set
+// these entries to the constraint value and modify the rest of the row
+// so that the row sum remains the same as before by adding an equal
+// amount to each remaining entry. However, if the original row sum value
+// violates the constraints, we set the row sum back to 1 (the row sum of
+// tentative P). After doing the modification to a row, we need to check
+// again the entire row to make sure that the modified row does not violate
+// the constraints.
+
+template <typename local_matrix_type>
+struct constraintKernel {
+  using Scalar      = typename local_matrix_type::non_const_value_type;
+  using SC          = Scalar;
+  using LO          = typename local_matrix_type::non_const_ordinal_type;
+  using Device      = typename local_matrix_type::device_type;
+  const Scalar zero = Kokkos::ArithTraits<SC>::zero();
+  const Scalar one  = Kokkos::ArithTraits<SC>::one();
+  LO nPDEs;
+  local_matrix_type localP;
+  Kokkos::View<SC**, Device> ConstraintViolationSum;
+  Kokkos::View<SC**, Device> Rsum;
+  Kokkos::View<size_t**, Device> nPositive;
+
+  constraintKernel(LO nPDEs_, local_matrix_type localP_)
+    : nPDEs(nPDEs_)
+    , localP(localP_) {
+    ConstraintViolationSum = Kokkos::View<SC**, Device>("ConstraintViolationSum", localP_.numRows(), nPDEs);
+    Rsum                   = Kokkos::View<SC**, Device>("Rsum", localP_.numRows(), nPDEs);
+    nPositive              = Kokkos::View<size_t**, Device>("nPositive", localP_.numRows(), nPDEs);
+  }
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const size_t rowIdx) const {
+    auto rowPtr = localP.graph.row_map;
+    auto values = localP.values;
+
+    bool checkRow = true;
+
+    if (rowPtr(rowIdx + 1) == rowPtr(rowIdx)) checkRow = false;
+
+    while (checkRow) {
+      // check constraints and compute the row sum
+
+      for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) {
+        Rsum(rowIdx, entryIdx % nPDEs) += values(entryIdx);
+        if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) < Kokkos::ArithTraits<SC>::real(zero)) {
+          ConstraintViolationSum(rowIdx, entryIdx % nPDEs) += values(entryIdx);
+          values(entryIdx) = zero;
         } else {
-          GetOStream(Statistics1) << "Using cached max eigenvalue estimate" << std::endl;
+          if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) != Kokkos::ArithTraits<SC>::real(zero))
+            nPositive(rowIdx, entryIdx % nPDEs) = nPositive(rowIdx, entryIdx % nPDEs) + 1;
+
+          if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) > Kokkos::ArithTraits<SC>::real(1.00001)) {
+            ConstraintViolationSum(rowIdx, entryIdx % nPDEs) += (values(entryIdx) - one);
+            values(entryIdx) = one;
+          }
         }
       }
-      else {
-          lambdaMax = userDefinedMaxEigen;
-          A->SetMaxEigenvalueEstimate(lambdaMax);
-      }
-      GetOStream(Statistics0) << "Prolongator damping factor = " << dampingFactor/lambdaMax << " (" << dampingFactor << " / " << lambdaMax << ")" << std::endl;
 
-      {
-        SubFactoryMonitor m2(*this, "Fused (I-omega*D^{-1} A)*Ptent", coarseLevel);
-        {
-          SubFactoryMonitor m3(*this, "Diagonal Extraction", coarseLevel);
-          if (useAbsValueRowSum)
-            GetOStream(Runtime0) << "Using rowSumAbs diagonal" << std::endl;
-          if (invDiag == Teuchos::null)
-            invDiag = Utilities::GetMatrixDiagonalInverse(*A, Teuchos::ScalarTraits<SC>::eps()*100, Teuchos::ScalarTraits<SC>::zero(), useAbsValueRowSum);
-        }
-        SC omega = dampingFactor / lambdaMax;
-        TEUCHOS_TEST_FOR_EXCEPTION(!std::isfinite(Teuchos::ScalarTraits<SC>::magnitude(omega)), Exceptions::RuntimeError, "Prolongator damping factor needs to be finite.");
-
-        {
-          SubFactoryMonitor m3(*this, "Xpetra::IteratorOps::Jacobi", coarseLevel);
-          // finalP = Ptent + (I - \omega D^{-1}A) Ptent
-          finalP = Xpetra::IteratorOps<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Jacobi(omega, *invDiag, *A, *Ptent, finalP, GetOStream(Statistics2), std::string("MueLu::SaP-") + toString(coarseLevel.GetLevelID()), APparams);
-        if (enforceConstraints) SatisfyPConstraints( A, finalP);
+      checkRow = false;
+
+      // take into account any row sum that violates the contraints
+
+      for (size_t k = 0; k < (size_t)nPDEs; k++) {
+        if (Kokkos::ArithTraits<SC>::real(Rsum(rowIdx, k)) < Kokkos::ArithTraits<SC>::magnitude(zero)) {
+          ConstraintViolationSum(rowIdx, k) = ConstraintViolationSum(rowIdx, k) - Rsum(rowIdx, k);  // rstumin
+        } else if (Kokkos::ArithTraits<SC>::real(Rsum(rowIdx, k)) > Kokkos::ArithTraits<SC>::magnitude(1.00001)) {
+          ConstraintViolationSum(rowIdx, k) = ConstraintViolationSum(rowIdx, k) + (one - Rsum(rowIdx, k));  // rstumin
         }
       }
 
-    } else {
-      finalP = Ptent;
-    }
-
-    // Level Set
-    if (!restrictionMode_) {
-      // prolongation factory is in prolongation mode
-      if(!finalP.is_null()) {std::ostringstream oss; oss << "P_" << coarseLevel.GetLevelID(); finalP->setObjectLabel(oss.str());}
-      Set(coarseLevel, "P", finalP);
+      // check if row need modification
+      for (size_t k = 0; k < (size_t)nPDEs; k++) {
+        if (Kokkos::ArithTraits<SC>::magnitude(ConstraintViolationSum(rowIdx, k)) != Kokkos::ArithTraits<SC>::magnitude(zero))
+          checkRow = true;
+      }
+      // modify row
+      if (checkRow) {
+        for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) {
+          if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) > Kokkos::ArithTraits<SC>::real(zero)) {
+            values(entryIdx) = values(entryIdx) +
+                               (ConstraintViolationSum(rowIdx, entryIdx % nPDEs) / (Scalar(nPositive(rowIdx, entryIdx % nPDEs)) != zero ? Scalar(nPositive(rowIdx, entryIdx % nPDEs)) : one));
+          }
+        }
+        for (size_t k = 0; k < (size_t)nPDEs; k++) ConstraintViolationSum(rowIdx, k) = zero;
+      }
+      for (size_t k = 0; k < (size_t)nPDEs; k++) Rsum(rowIdx, k) = zero;
+      for (size_t k = 0; k < (size_t)nPDEs; k++) nPositive(rowIdx, k) = 0;
+    }  // while (checkRow) ...
+  }
+};
 
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        finalP->CreateView("stridedMaps", Ptent);
+template <typename local_matrix_type>
+struct optimalSatisfyConstraintsForScalarPDEsKernel {
+  using Scalar      = typename local_matrix_type::non_const_value_type;
+  using SC          = Scalar;
+  using LO          = typename local_matrix_type::non_const_ordinal_type;
+  using Device      = typename local_matrix_type::device_type;
+  using KAT         = Kokkos::ArithTraits<SC>;
+  const Scalar zero = Kokkos::ArithTraits<SC>::zero();
+  const Scalar one  = Kokkos::ArithTraits<SC>::one();
+  LO nPDEs;
+  local_matrix_type localP;
+  Kokkos::View<SC**, Device> origSorted;
+  Kokkos::View<SC**, Device> fixedSorted;
+  Kokkos::View<LO**, Device> inds;
+
+  optimalSatisfyConstraintsForScalarPDEsKernel(LO nPDEs_, LO maxRowEntries_, local_matrix_type localP_)
+    : nPDEs(nPDEs_)
+    , localP(localP_) {
+    origSorted  = Kokkos::View<SC**, Device>("origSorted", localP_.numRows(), maxRowEntries_);
+    fixedSorted = Kokkos::View<SC**, Device>("fixedSorted", localP_.numRows(), maxRowEntries_);
+    inds        = Kokkos::View<LO**, Device>("inds", localP_.numRows(), maxRowEntries_);
+  }
 
-    } else {
-      // prolongation factory is in restriction mode
-      RCP<Matrix> R = Utilities::Transpose(*finalP, true);
-      Set(coarseLevel, "R", R);
-      if(!R.is_null()) {std::ostringstream oss; oss << "R_" << coarseLevel.GetLevelID(); R->setObjectLabel(oss.str());}
-
-      // NOTE: EXPERIMENTAL
-      if (Ptent->IsView("stridedMaps"))
-        R->CreateView("stridedMaps", Ptent, true);
-    }
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const size_t rowIdx) const {
+    auto rowPtr = localP.graph.row_map;
+    auto values = localP.values;
 
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      params->set("printCommInfo",          true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*finalP, (!restrictionMode_ ? "P" : "R"), params);
-    }
+    auto nnz = rowPtr(rowIdx + 1) - rowPtr(rowIdx);
 
-  } //Build()
-
-  // Analyze the grid transfer produced by smoothed aggregation and make
-  // modifications if it does not look right. In particular, if there are
-  // negative entries or entries larger than 1, modify P's rows.
-  //
-  // Note: this kind of evaluation probably only makes sense if not doing QR
-  // when constructing tentative P.
-  //
-  // For entries that do not satisfy the two constraints (>= 0  or <=1) we set
-  // these entries to the constraint value and modify the rest of the row
-  // so that the row sum remains the same as before by adding an equal
-  // amount to each remaining entry. However, if the original row sum value
-  // violates the constraints, we set the row sum back to 1 (the row sum of
-  // tentative P). After doing the modification to a row, we need to check
-  // again the entire row to make sure that the modified row does not violate
-  // the constraints.
-
-template<typename local_matrix_type>
-struct constraintKernel {
+    if (nnz != 0) {
+      Scalar rsumTarget = zero;
+      for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) rsumTarget += values(entryIdx);
 
-   using Scalar= typename local_matrix_type::non_const_value_type;
-   using SC=Scalar;
-   using LO=typename local_matrix_type::non_const_ordinal_type;
-   using Device= typename local_matrix_type::device_type;
-   const Scalar zero = Kokkos::ArithTraits<SC>::zero();
-   const Scalar one  = Kokkos::ArithTraits<SC>::one();
-   LO     nPDEs;
-   local_matrix_type localP;
-   Kokkos::View<SC**, Device> ConstraintViolationSum;
-   Kokkos::View<SC**, Device> Rsum;
-   Kokkos::View<size_t**, Device> nPositive;
+      {
+        SC aBigNumber;
+        SC rowSumDeviation, temp, delta;
+        SC closestToLeftBoundDist, closestToRghtBoundDist;
+        LO closestToLeftBound, closestToRghtBound;
+        bool flipped;
+
+        SC leftBound = zero;
+        SC rghtBound = one;
+        if ((KAT::real(rsumTarget) >= KAT::real(leftBound * (static_cast<SC>(nnz)))) &&
+            (KAT::real(rsumTarget) <= KAT::real(rghtBound * (static_cast<SC>(nnz))))) {  // has Feasible solution
+
+          flipped = false;
+          // compute aBigNumber to handle some corner cases where we need
+          // something large so that an if statement will be false
+          aBigNumber = KAT::zero();
+          for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) {
+            if (KAT::magnitude(values(entryIdx)) > KAT::magnitude(aBigNumber))
+              aBigNumber = KAT::magnitude(values(entryIdx));
+          }
+          aBigNumber = aBigNumber + (KAT::magnitude(leftBound) + KAT::magnitude(rghtBound)) * (100 * one);
 
-  constraintKernel(LO nPDEs_,local_matrix_type localP_) : nPDEs(nPDEs_), localP(localP_)
-   {
-     ConstraintViolationSum = Kokkos::View<SC**, Device>("ConstraintViolationSum", localP_.numRows(), nPDEs);
-     Rsum = Kokkos::View<SC**, Device>("Rsum", localP_.numRows(), nPDEs);
-     nPositive = Kokkos::View<size_t**, Device>("nPositive", localP_.numRows(), nPDEs);
-   }
+          LO ind = 0;
+          for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) {
+            origSorted(rowIdx, ind) = values(entryIdx);
+            inds(rowIdx, ind)       = ind;
+            ind++;
+          }
 
-   KOKKOS_INLINE_FUNCTION
-   void operator() (const size_t rowIdx) const {
+          auto sortVals = Kokkos::subview(origSorted, rowIdx, Kokkos::ALL());
+          auto sortInds = Kokkos::subview(inds, rowIdx, Kokkos::make_pair(0, ind));
+          // Need permutation indices to sort row values from smallest to largest,
+          // and unsort once row constraints are applied.
+          // serial insertion sort workaround from https://github.com/kokkos/kokkos/issues/645
+          for (LO i = 1; i < static_cast<LO>(nnz); ++i) {
+            ind  = sortInds(i);
+            LO j = i;
 
-      auto rowPtr = localP.graph.row_map;
-      auto values = localP.values;
+            if (KAT::real(sortVals(sortInds(i))) < KAT::real(sortVals(sortInds(0)))) {
+              for (; j > 0; --j) sortInds(j) = sortInds(j - 1);
 
-      bool checkRow = true;
+              sortInds(0) = ind;
+            } else {
+              for (; KAT::real(sortVals(ind)) < KAT::real(sortVals(sortInds(j - 1))); --j) sortInds(j) = sortInds(j - 1);
 
-      if (rowPtr(rowIdx + 1) == rowPtr(rowIdx)) checkRow = false;
+              sortInds(j) = ind;
+            }
+          }
 
+          for (LO i = 0; i < static_cast<LO>(nnz); i++) origSorted(rowIdx, i) = values(rowPtr(rowIdx) + inds(rowIdx, i));  // values is no longer used
+          // find entry in origSorted just to the right of the leftBound
+          closestToLeftBound = 0;
+          while ((closestToLeftBound < static_cast<LO>(nnz)) && (KAT::real(origSorted(rowIdx, closestToLeftBound)) <= KAT::real(leftBound))) closestToLeftBound++;
 
-      while (checkRow) {
+          // find entry in origSorted just to the right of the rghtBound
+          closestToRghtBound = closestToLeftBound;
+          while ((closestToRghtBound < static_cast<LO>(nnz)) && (KAT::real(origSorted(rowIdx, closestToRghtBound)) <= KAT::real(rghtBound))) closestToRghtBound++;
 
-        // check constraints and compute the row sum
-    
-        for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++)  {
-          Rsum(rowIdx, entryIdx%nPDEs) += values(entryIdx);
-          if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) < Kokkos::ArithTraits<SC>::real(zero)) {
+          // compute distance between closestToLeftBound and the left bound and the
+          // distance between closestToRghtBound and the right bound.
 
-            ConstraintViolationSum(rowIdx, entryIdx%nPDEs) += values(entryIdx);
-            values(entryIdx) = zero;
-          }
-          else {
-            if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) != Kokkos::ArithTraits<SC>::real(zero))
-              nPositive(rowIdx, entryIdx%nPDEs) = nPositive(rowIdx, entryIdx%nPDEs) + 1;
-    
-            if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) > Kokkos::ArithTraits<SC>::real(1.00001  )) {
-              ConstraintViolationSum(rowIdx, entryIdx%nPDEs) += (values(entryIdx) - one);
-              values(entryIdx) =  one;
+          closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
+          if (closestToRghtBound == static_cast<LO>(nnz))
+            closestToRghtBoundDist = aBigNumber;
+          else
+            closestToRghtBoundDist = origSorted(rowIdx, closestToRghtBound) - rghtBound;
+
+          // compute how far the rowSum is off from the target row sum taking into account
+          // numbers that have been shifted to satisfy bound constraint
+
+          rowSumDeviation = leftBound * (static_cast<SC>(closestToLeftBound)) + (static_cast<SC>(nnz - closestToRghtBound)) * rghtBound - rsumTarget;
+          for (LO i = closestToLeftBound; i < closestToRghtBound; i++) rowSumDeviation += origSorted(rowIdx, i);
+
+          // the code that follow after this if statement assumes that rowSumDeviation is positive. If this
+          // is not the case, flip the signs of everything so that rowSumDeviation is now positive.
+          // Later we will flip the data back to its original form.
+          if (KAT::real(rowSumDeviation) < KAT::real(KAT::zero())) {
+            flipped   = true;
+            temp      = leftBound;
+            leftBound = -rghtBound;
+            rghtBound = temp;
+
+            /* flip sign of origSorted and reverse ordering so that the negative version is sorted */
+
+            // TODO: the following is bad for GPU performance. Switch to bit shifting if brave.
+            if ((nnz % 2) == 1) origSorted(rowIdx, (nnz / 2)) = -origSorted(rowIdx, (nnz / 2));
+            for (LO i = 0; i < static_cast<LO>(nnz / 2); i++) {
+              temp                            = origSorted(rowIdx, i);
+              origSorted(rowIdx, i)           = -origSorted(rowIdx, nnz - 1 - i);
+              origSorted(rowIdx, nnz - i - 1) = -temp;
             }
-          }
-        }
-    
-        checkRow = false;
 
-        // take into account any row sum that violates the contraints
-    
-        for (size_t k=0; k < (size_t) nPDEs; k++) {
+            /* reverse bounds */
 
-          if (Kokkos::ArithTraits<SC>::real(Rsum(rowIdx, k)) < Kokkos::ArithTraits<SC>::magnitude(zero)) {
-              ConstraintViolationSum(rowIdx, k) = ConstraintViolationSum(rowIdx, k) - Rsum(rowIdx, k);  // rstumin
-          }
-          else if (Kokkos::ArithTraits<SC>::real(Rsum(rowIdx, k)) > Kokkos::ArithTraits<SC>::magnitude(1.00001)) {
-              ConstraintViolationSum(rowIdx, k) = ConstraintViolationSum(rowIdx, k)+ (one - Rsum(rowIdx, k));  // rstumin
+            LO itemp               = closestToLeftBound;
+            closestToLeftBound     = nnz - closestToRghtBound;
+            closestToRghtBound     = nnz - itemp;
+            closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
+            if (closestToRghtBound == static_cast<LO>(nnz))
+              closestToRghtBoundDist = aBigNumber;
+            else
+              closestToRghtBoundDist = origSorted(rowIdx, closestToRghtBound) - rghtBound;
+
+            rowSumDeviation = -rowSumDeviation;
           }
-        }
 
-        // check if row need modification
-        for (size_t k=0; k < (size_t) nPDEs; k++) {
-          if (Kokkos::ArithTraits<SC>::magnitude(ConstraintViolationSum(rowIdx, k)) != Kokkos::ArithTraits<SC>::magnitude(zero))
-             checkRow = true;
-        }
-        // modify row
-        if (checkRow) {
-	   for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++)  {
-             if (Kokkos::ArithTraits<SC>::real(values(entryIdx)) > Kokkos::ArithTraits<SC>::real(zero)) {
-	       values(entryIdx) = values(entryIdx) +
-		 (ConstraintViolationSum(rowIdx, entryIdx%nPDEs)/ (Scalar (nPositive(rowIdx, entryIdx%nPDEs)) != zero ? Scalar (nPositive(rowIdx, entryIdx%nPDEs)) : one));
-             }
-           }
-           for (size_t k=0; k < (size_t) nPDEs; k++) ConstraintViolationSum(rowIdx, k) = zero;
-        }
-        for (size_t k=0; k < (size_t) nPDEs; k++) Rsum(rowIdx, k) = zero;
-        for (size_t k=0; k < (size_t) nPDEs; k++) nPositive(rowIdx, k) = 0;
-      } // while (checkRow) ...
+          // initial fixedSorted so bounds are satisfied and interiors correspond to origSorted
 
-   }
-};
+          for (LO i = 0; i < closestToLeftBound; i++) fixedSorted(rowIdx, i) = leftBound;
+          for (LO i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i);
+          for (LO i = closestToRghtBound; i < static_cast<LO>(nnz); i++) fixedSorted(rowIdx, i) = rghtBound;
 
-template<typename local_matrix_type>
-struct optimalSatisfyConstraintsForScalarPDEsKernel {
+          while ((KAT::magnitude(rowSumDeviation) > KAT::magnitude((one * 1.e-10) * rsumTarget))) {  // && ( (closestToLeftBound < nEntries ) || (closestToRghtBound < nEntries))) {
+            if (closestToRghtBound != closestToLeftBound)
+              delta = rowSumDeviation / static_cast<SC>(closestToRghtBound - closestToLeftBound);
+            else
+              delta = aBigNumber;
 
-   using Scalar= typename local_matrix_type::non_const_value_type;
-   using SC=Scalar;
-   using LO=typename local_matrix_type::non_const_ordinal_type;
-   using Device= typename local_matrix_type::device_type;
-   using KAT = Kokkos::ArithTraits<SC>;
-   const Scalar zero = Kokkos::ArithTraits<SC>::zero();
-   const Scalar one  = Kokkos::ArithTraits<SC>::one();
-   LO     nPDEs;
-   local_matrix_type localP;
-   Kokkos::View<SC**, Device> origSorted;
-   Kokkos::View<SC**, Device> fixedSorted;
-   Kokkos::View<LO**, Device> inds;
-
-   optimalSatisfyConstraintsForScalarPDEsKernel(LO nPDEs_, LO maxRowEntries_, local_matrix_type localP_) : nPDEs(nPDEs_), localP(localP_)
-   {
-     origSorted  =  Kokkos::View<SC**, Device>("origSorted" , localP_.numRows(), maxRowEntries_);
-     fixedSorted =  Kokkos::View<SC**, Device>("fixedSorted", localP_.numRows(), maxRowEntries_);
-     inds        =  Kokkos::View<LO**, Device>("inds"       , localP_.numRows(), maxRowEntries_);
-   }
-
-   KOKKOS_INLINE_FUNCTION
-   void operator() (const size_t rowIdx) const {
-
-     auto rowPtr = localP.graph.row_map;
-     auto values = localP.values;
-
-     auto nnz = rowPtr(rowIdx+1) - rowPtr(rowIdx);
-
-     if (nnz != 0) {
-
-       Scalar rsumTarget = zero;
-       for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) rsumTarget += values(entryIdx);
-
-       {
-         SC aBigNumber;
-         SC rowSumDeviation, temp, delta;
-         SC closestToLeftBoundDist, closestToRghtBoundDist;
-         LO closestToLeftBound, closestToRghtBound;
-         bool   flipped;
-
-         SC leftBound = zero;
-         SC rghtBound = one;
-         if ((KAT::real(rsumTarget) >= KAT::real(leftBound*(static_cast<SC>(nnz)))) &&
-             (KAT::real(rsumTarget) <= KAT::real(rghtBound*(static_cast<SC>(nnz))))){ // has Feasible solution
-
-           flipped    = false;
-           // compute aBigNumber to handle some corner cases where we need
-           // something large so that an if statement will be false
-           aBigNumber = KAT::zero();
-           for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++){
-             if ( KAT::magnitude( values(entryIdx) ) > KAT::magnitude(aBigNumber))
-               aBigNumber = KAT::magnitude( values(entryIdx) );
-           }
-           aBigNumber = aBigNumber+ (KAT::magnitude(leftBound) + KAT::magnitude(rghtBound))*(100*one);
-  
-           LO ind = 0;
-           for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++){
-               origSorted(rowIdx, ind) = values(entryIdx);
-               inds(rowIdx, ind) = ind;
-               ind++;
-           }
-  
-           auto sortVals = Kokkos::subview( origSorted, rowIdx, Kokkos::ALL() );
-           auto sortInds = Kokkos::subview( inds, rowIdx, Kokkos::make_pair(0,ind));
-           // Need permutation indices to sort row values from smallest to largest,
-           // and unsort once row constraints are applied.
-           // serial insertion sort workaround from https://github.com/kokkos/kokkos/issues/645
-           for (LO i = 1; i < static_cast<LO>(nnz); ++i){
-             ind = sortInds(i);
-             LO j = i;
-             
-             if ( KAT::real(sortVals(sortInds(i))) < KAT::real(sortVals(sortInds(0))) ){
-               for ( ; j > 0; --j) sortInds(j) = sortInds(j - 1);
-  
-               sortInds(0) = ind;
-             } else {
-               for ( ; KAT::real(sortVals(ind)) < KAT::real(sortVals(sortInds(j-1))); --j) sortInds(j) = sortInds(j-1);
-  
-               sortInds(j) = ind;
-             }
-           }
-  
-  
-           for (LO i = 0; i < static_cast<LO>(nnz); i++) origSorted(rowIdx, i) = values(rowPtr(rowIdx) + inds(rowIdx, i)); //values is no longer used
-           // find entry in origSorted just to the right of the leftBound
-           closestToLeftBound = 0;
-           while ((closestToLeftBound < static_cast<LO>(nnz)) && (KAT::real(origSorted(rowIdx, closestToLeftBound)) <= KAT::real(leftBound))) closestToLeftBound++;
-  
-           // find entry in origSorted just to the right of the rghtBound
-           closestToRghtBound = closestToLeftBound;
-           while ((closestToRghtBound < static_cast<LO>(nnz)) && (KAT::real(origSorted(rowIdx, closestToRghtBound)) <= KAT::real(rghtBound))) closestToRghtBound++;
-  
-           // compute distance between closestToLeftBound and the left bound and the
-           // distance between closestToRghtBound and the right bound.
-        
-           closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
-           if (closestToRghtBound == static_cast<LO>(nnz)) closestToRghtBoundDist= aBigNumber;
-           else                                closestToRghtBoundDist= origSorted(rowIdx, closestToRghtBound) - rghtBound;
-  
-           // compute how far the rowSum is off from the target row sum taking into account
-           // numbers that have been shifted to satisfy bound constraint
-  
-           rowSumDeviation = leftBound*(static_cast<SC>(closestToLeftBound)) + (static_cast<SC>(nnz-closestToRghtBound))*rghtBound - rsumTarget;
-           for (LO i=closestToLeftBound; i < closestToRghtBound; i++) rowSumDeviation += origSorted(rowIdx, i);
-  
-           // the code that follow after this if statement assumes that rowSumDeviation is positive. If this
-           // is not the case, flip the signs of everything so that rowSumDeviation is now positive.
-           // Later we will flip the data back to its original form.
-           if (KAT::real(rowSumDeviation) < KAT::real(KAT::zero())) {
-             flipped = true;
-             temp = leftBound; leftBound = -rghtBound; rghtBound = temp;
-         
-             /* flip sign of origSorted and reverse ordering so that the negative version is sorted */
-         
-             //TODO: the following is bad for GPU performance. Switch to bit shifting if brave.
-             if ((nnz%2) == 1) origSorted(rowIdx, (nnz/2)  ) =  -origSorted(rowIdx, (nnz/2)  );
-             for (LO i=0; i < static_cast<LO>(nnz/2); i++) {
-               temp=origSorted(rowIdx, i);
-               origSorted(rowIdx, i) = -origSorted(rowIdx, nnz-1-i);
-               origSorted(rowIdx, nnz-i-1) = -temp;
-             }
-           
-             /* reverse bounds */
-         
-             LO itemp = closestToLeftBound;
-             closestToLeftBound = nnz-closestToRghtBound;
-             closestToRghtBound = nnz-itemp;
-             closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
-             if (closestToRghtBound == static_cast<LO>(nnz)) closestToRghtBoundDist= aBigNumber;
-             else                                closestToRghtBoundDist= origSorted(rowIdx, closestToRghtBound) - rghtBound;
-         
-             rowSumDeviation = -rowSumDeviation;
-           }
-  
-           // initial fixedSorted so bounds are satisfied and interiors correspond to origSorted
-         
-           for (LO i =                  0; i < closestToLeftBound;   i++) fixedSorted(rowIdx, i) = leftBound;
-           for (LO i = closestToLeftBound; i < closestToRghtBound;   i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i);
-           for (LO i = closestToRghtBound; i < static_cast<LO>(nnz); i++) fixedSorted(rowIdx, i) = rghtBound;
-         
-           while ((KAT::magnitude(rowSumDeviation) > KAT::magnitude((one*1.e-10)*rsumTarget))){ // && ( (closestToLeftBound < nEntries ) || (closestToRghtBound < nEntries))) {
-            if (closestToRghtBound !=  closestToLeftBound)
-                 delta = rowSumDeviation/ static_cast<SC>(closestToRghtBound -  closestToLeftBound);
-            else delta = aBigNumber;
-         
             if (KAT::magnitude(closestToLeftBoundDist) <= KAT::magnitude(closestToRghtBoundDist)) {
-               if (KAT::magnitude(delta) <= KAT::magnitude(closestToLeftBoundDist)) {
-                 rowSumDeviation = zero;
-                 for (LO i = closestToLeftBound; i < closestToRghtBound ; i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i) - delta;
-               }
-               else {
-                 rowSumDeviation = rowSumDeviation - closestToLeftBoundDist;
-                 fixedSorted(rowIdx, closestToLeftBound) = leftBound;
-                 closestToLeftBound++;
-                 if (closestToLeftBound < static_cast<LO>(nnz)) closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
-                 else                               closestToLeftBoundDist = aBigNumber;
-               }
+              if (KAT::magnitude(delta) <= KAT::magnitude(closestToLeftBoundDist)) {
+                rowSumDeviation = zero;
+                for (LO i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i) - delta;
+              } else {
+                rowSumDeviation                         = rowSumDeviation - closestToLeftBoundDist;
+                fixedSorted(rowIdx, closestToLeftBound) = leftBound;
+                closestToLeftBound++;
+                if (closestToLeftBound < static_cast<LO>(nnz))
+                  closestToLeftBoundDist = origSorted(rowIdx, closestToLeftBound) - leftBound;
+                else
+                  closestToLeftBoundDist = aBigNumber;
+              }
+            } else {
+              if (KAT::magnitude(delta) <= KAT::magnitude(closestToRghtBoundDist)) {
+                rowSumDeviation = 0;
+                for (LO i = closestToLeftBound; i < closestToRghtBound; i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i) - delta;
+              } else {
+                rowSumDeviation = rowSumDeviation + closestToRghtBoundDist;
+                //        if (closestToRghtBound < nEntries) {
+                fixedSorted(rowIdx, closestToRghtBound) = origSorted(rowIdx, closestToRghtBound);
+                closestToRghtBound++;
+                //       }
+                if (closestToRghtBound >= static_cast<LO>(nnz))
+                  closestToRghtBoundDist = aBigNumber;
+                else
+                  closestToRghtBoundDist = origSorted(rowIdx, closestToRghtBound) - rghtBound;
+              }
             }
-            else {
-               if (KAT::magnitude(delta) <= KAT::magnitude(closestToRghtBoundDist)) {
-                 rowSumDeviation = 0;
-                 for (LO i = closestToLeftBound; i < closestToRghtBound ; i++) fixedSorted(rowIdx, i) = origSorted(rowIdx, i) - delta;
-               }
-               else {
-                 rowSumDeviation = rowSumDeviation + closestToRghtBoundDist;
-         //        if (closestToRghtBound < nEntries) {
-                   fixedSorted(rowIdx, closestToRghtBound) = origSorted(rowIdx, closestToRghtBound);
-                   closestToRghtBound++;
-          //       }
-                 if (closestToRghtBound >= static_cast<LO>(nnz)) closestToRghtBoundDist = aBigNumber;
-                 else                                closestToRghtBoundDist = origSorted(rowIdx, closestToRghtBound) - rghtBound;
-               }
-             }
-           }
-  
-           auto rowStart = rowPtr(rowIdx);
-           if (flipped) {
-             /* flip sign of fixedSorted and reverse ordering so that the positve version is sorted */
-  
-             if ((nnz%2) == 1) fixedSorted(rowIdx, (nnz/2)  ) =  -fixedSorted(rowIdx, (nnz/2)  );
-             for (LO i=0; i < static_cast<LO>(nnz/2); i++) {
-               temp=fixedSorted(rowIdx, i);
-               fixedSorted(rowIdx, i) = -fixedSorted(rowIdx, nnz-1-i);
-               fixedSorted(rowIdx, nnz-i-1) = -temp;
-             }
-           }
-           // unsort and update row values with new values
-           for (LO i = 0; i < static_cast<LO>(nnz); i++)  values(rowStart + inds(rowIdx, i)) = fixedSorted(rowIdx, i);
-
-         } else { // row does not have feasible solution to match constraint
-           // just set all entries to the same value giving a row sum of 1
-           for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) values(entryIdx) = one/( static_cast<SC>(nnz) );
-         }
-       }
-
-     }
-
-   }
-};
-   
+          }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::SatisfyPConstraints(const RCP<Matrix> A, RCP<Matrix>& P) const {
+          auto rowStart = rowPtr(rowIdx);
+          if (flipped) {
+            /* flip sign of fixedSorted and reverse ordering so that the positve version is sorted */
 
-    using Device = typename Matrix::local_matrix_type::device_type;
-    LO nPDEs = A->GetFixedBlockSize();
+            if ((nnz % 2) == 1) fixedSorted(rowIdx, (nnz / 2)) = -fixedSorted(rowIdx, (nnz / 2));
+            for (LO i = 0; i < static_cast<LO>(nnz / 2); i++) {
+              temp                             = fixedSorted(rowIdx, i);
+              fixedSorted(rowIdx, i)           = -fixedSorted(rowIdx, nnz - 1 - i);
+              fixedSorted(rowIdx, nnz - i - 1) = -temp;
+            }
+          }
+          // unsort and update row values with new values
+          for (LO i = 0; i < static_cast<LO>(nnz); i++) values(rowStart + inds(rowIdx, i)) = fixedSorted(rowIdx, i);
+
+        } else {  // row does not have feasible solution to match constraint
+          // just set all entries to the same value giving a row sum of 1
+          for (auto entryIdx = rowPtr(rowIdx); entryIdx < rowPtr(rowIdx + 1); entryIdx++) values(entryIdx) = one / (static_cast<SC>(nnz));
+        }
+      }
+    }
+  }
+};
 
-    using local_mat_type = typename Matrix::local_matrix_type;
-    constraintKernel<local_mat_type> myKernel(nPDEs,P->getLocalMatrixDevice() );
-    Kokkos::parallel_for("enforce constraint",Kokkos::RangePolicy<typename Device::execution_space>(0, P->getRowMap()->getLocalNumElements() ),
-                        myKernel );
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::SatisfyPConstraints(const RCP<Matrix> A, RCP<Matrix>& P) const {
+  using Device = typename Matrix::local_matrix_type::device_type;
+  LO nPDEs     = A->GetFixedBlockSize();
 
-  } //SatsifyPConstraints()
+  using local_mat_type = typename Matrix::local_matrix_type;
+  constraintKernel<local_mat_type> myKernel(nPDEs, P->getLocalMatrixDevice());
+  Kokkos::parallel_for("enforce constraint", Kokkos::RangePolicy<typename Device::execution_space>(0, P->getRowMap()->getLocalNumElements()),
+                       myKernel);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void SaPFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::optimalSatisfyPConstraintsForScalarPDEs( RCP<Matrix>& P) const {
+}  // SatsifyPConstraints()
 
-    using Device = typename Matrix::local_matrix_type::device_type;
-    LO nPDEs = 1;//A->GetFixedBlockSize();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void SaPFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::optimalSatisfyPConstraintsForScalarPDEs(RCP<Matrix>& P) const {
+  using Device = typename Matrix::local_matrix_type::device_type;
+  LO nPDEs     = 1;  // A->GetFixedBlockSize();
 
-    using local_mat_type = typename Matrix::local_matrix_type;
-    optimalSatisfyConstraintsForScalarPDEsKernel<local_mat_type> myKernel(nPDEs,P->getLocalMaxNumRowEntries(),P->getLocalMatrixDevice() );
-    Kokkos::parallel_for("enforce constraint",Kokkos::RangePolicy<typename Device::execution_space>(0, P->getLocalNumRows() ),
-                        myKernel );
+  using local_mat_type = typename Matrix::local_matrix_type;
+  optimalSatisfyConstraintsForScalarPDEsKernel<local_mat_type> myKernel(nPDEs, P->getLocalMaxNumRowEntries(), P->getLocalMatrixDevice());
+  Kokkos::parallel_for("enforce constraint", Kokkos::RangePolicy<typename Device::execution_space>(0, P->getLocalNumRows()),
+                       myKernel);
 
-  } //SatsifyPConstraints()
+}  // SatsifyPConstraints()
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_SAPFACTORY_KOKKOS_DEF_HPP
+#endif  // MUELU_SAPFACTORY_KOKKOS_DEF_HPP
 
-//TODO: restrictionMode_ should use the parameter list.
+// TODO: restrictionMode_ should use the parameter list.
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_decl.hpp
index d70b8ceb0ade..2678c4ebabed 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_decl.hpp
@@ -58,103 +58,101 @@
 
 namespace MueLu {
 
-  /*!
-     @class ScaledNullspaceFactory class.
-     @brief Factory for generating a very special nullspace
+/*!
+   @class ScaledNullspaceFactory class.
+   @brief Factory for generating a very special nullspace
 
-      The ScaledNullspaceFactory is meant to generate a default approximation for
-      the fine level nullspace (Level 0 only). For all other levels it is used
-      to generate a (block-diagonally) scaled version of the Nullspace designed for non-symmetric problems.
-      only to act as "generating factory" for the "Nullspace", which is actually
-      handled by the TentativePFactory.
+    The ScaledNullspaceFactory is meant to generate a default approximation for
+    the fine level nullspace (Level 0 only). For all other levels it is used
+    to generate a (block-diagonally) scaled version of the Nullspace designed for non-symmetric problems.
+    only to act as "generating factory" for the "Nullspace", which is actually
+    handled by the TentativePFactory.
 
-      Constructor:
-      \code{.cpp}
-      ScaledNullspaceFactory(const std::string & nspName = "Nullspace")
-      \endcode
-      This constructor uses the variable with the variable nspName on the finest level
-      as null space for the finest multigrid level.
+    Constructor:
+    \code{.cpp}
+    ScaledNullspaceFactory(const std::string & nspName = "Nullspace")
+    \endcode
+    This constructor uses the variable with the variable nspName on the finest level
+    as null space for the finest multigrid level.
 
-      Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
-      in case of multiphysics problems, "Nullspace1" and "Nullspace2"
+    Currently, the only accepted names for the null space vectors are "Nullspace" (default) or,
+    in case of multiphysics problems, "Nullspace1" and "Nullspace2"
 
-     @ingroup MueLuTransferClasses
+   @ingroup MueLuTransferClasses
 
-    ## Input/output of NullspaceFactory ##
+  ## Input/output of NullspaceFactory ##
 
-    ### User parameters of NullspaceFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
-     A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and ScaledNullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the ScaledNullspaceFactory.
+  ### User parameters of NullspaceFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   Fine level nullspace | string | "Nullspace" | | * | * | Name of the variable containing the MultiVector with the near null space vectors on the finest level (level 0) provided by the user.
+   A                  | Factory | null |   | * | * | Generating factory of the matrix A. Note that "A" is only requested, if the user does not provide a fine level null space information and ScaledNullspaceFactory is meant to provide a default set of fine level null space vectors. Then, the information of "A" is used to create constant null space vectors compatible with the row map of A.
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the ScaledNullspaceFactory.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ScaledNullspaceFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see ScaledNullspaceFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see ScaledNullspaceFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see ScaledNullspaceFactory::DeclareInput).
 
-    ### Variables provided by ScaledNullspaceFactory ###
+  ### Variables provided by ScaledNullspaceFactory ###
 
-    After ScaledNullspaceFactory::Build the following data is available (if requested)
+  After ScaledNullspaceFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | Nullspace | ScaledNullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors will be a scaled version of the nullspace generated by the TentativePFactory.
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | Nullspace | ScaledNullspaceFactory | MultiVector containing the near null space vectors. On the finest level it can be a user-provided set of null space vectors. Otherwise a default set of constant near nullspace vectors is provided. On the coarser levels the null space vectors will be a scaled version of the nullspace generated by the TentativePFactory.
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class ScaledNullspaceFactory : public SingleLevelFactoryBase {
+          class Node          = DefaultNode>
+class ScaledNullspaceFactory : public SingleLevelFactoryBase {
 #undef MUELU_SCALEDNULLSPACEFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    ScaledNullspaceFactory(const std::string & nspName = "Nullspace") {
-      SetParameter("Fine level nullspace", ParameterEntry(nspName));
-    }
+  //! Constructor
+  ScaledNullspaceFactory(const std::string &nspName = "Nullspace") {
+    SetParameter("Fine level nullspace", ParameterEntry(nspName));
+  }
 
-    //! Destructor
-    virtual ~ScaledNullspaceFactory() { }
+  //! Destructor
+  virtual ~ScaledNullspaceFactory() {}
 
-    //@}
+  //@}
 
-    //! @name Input
+  //! @name Input
 
-    //@{
+  //@{
 
-    /*! @brief Define valid parameters for internal factory parameters */
-    RCP<const ParameterList> GetValidParameterList() const;
+  /*! @brief Define valid parameters for internal factory parameters */
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-
-    void DeclareInput(Level &currentLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
 
-    //@}
+  void DeclareInput(Level &currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-  private:
+  //@}
 
-  }; //class NullspaceFactory
+ private:
+};  // class NullspaceFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_SCALEDNULLSPACEFACTORY_SHORT
-#endif // MUELU_SCALEDNULLSPACEFACTORY_DECL_HPP
+#endif  // MUELU_SCALEDNULLSPACEFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_def.hpp
index 4a5cd4a912eb..d699f2e802ce 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_ScaledNullspaceFactory_def.hpp
@@ -58,72 +58,71 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
-    validParamList->set< std::string >("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\").");
-  
-    validParamList->set< RCP<const FactoryBase> >("A",                          Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",                  Teuchos::null, "Generating factory of the fine level null space");
-
-    return validParamList;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
+  validParamList->set<std::string>("Fine level nullspace", "Nullspace", "Variable name which is used to store null space multi vector on the finest level (default=\"Nullspace\").");
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the fine level matrix (only needed if default null space is generated)");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the fine level null space");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  // Scaled Nullspace always needs A & a Nullspace from somewhere
+  Input(currentLevel, "A");
+  Input(currentLevel, "Nullspace");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Build", currentLevel);
+
+  RCP<MultiVector> nullspace, tentativeNullspace;
+
+  // TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::ScaledNullspaceFactory::Build(): ScaledNullspaceFactory can be used for finest level (LevelID == 0) only.");
+
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable("Nullspace", NoFactory::get())) {
+      // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
+      // Set("Nullspace1", ...), we use it.
+      tentativeNullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", NoFactory::get());
+    } else {
+      // A User "Nullspace" (nspName) is not available (use factory)
+      tentativeNullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", GetFactory("Nullspace").get());
+
+    }  // end if "Nullspace" not available
+  } else {
+    tentativeNullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", GetFactory("Nullspace").get());
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    // Scaled Nullspace always needs A & a Nullspace from somewhere
-    Input(currentLevel, "A");
-    Input(currentLevel, "Nullspace");
+  // Scale!
+  RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+
+  // determine numPDEs
+  LocalOrdinal numPDEs = 1;
+  if (A->IsView("stridedMaps") == true) {
+    Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap()) == Teuchos::null, Exceptions::BadCast, "MueLu::ScaledNullspaceFactory::Build: cast to strided row map failed.");
+    numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
+    oldView = A->SwitchToView(oldView);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ScaledNullspaceFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Build", currentLevel);
+  GetOStream(Runtime1) << "ScaledNullspaceFactory: Generating scaled nullspace, blocksize = " << tentativeNullspace->getNumVectors() << std::endl;
+  nullspace                      = MultiVectorFactory::Build(tentativeNullspace->getMap(), tentativeNullspace->getNumVectors());
+  *nullspace                     = *tentativeNullspace;  // Copy the tentative nullspace
+  RCP<MultiVector> blockDiagonal = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
 
-    RCP<MultiVector> nullspace, tentativeNullspace;
+  Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::extractBlockDiagonal(*A, *blockDiagonal);
+  Xpetra::MatrixUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::inverseScaleBlockDiagonal(*blockDiagonal, true, *nullspace);
 
-    //TEUCHOS_TEST_FOR_EXCEPTION(currentLevel.GetLevelID() != 0, Exceptions::RuntimeError, "MueLu::ScaledNullspaceFactory::Build(): ScaledNullspaceFactory can be used for finest level (LevelID == 0) only.");
+  // provide "Scaled Nullspace" variable on current level (used by TentativePFactory)
+  Set(currentLevel, "Scaled Nullspace", nullspace);
 
-    if (currentLevel.GetLevelID() == 0) {
-      if (currentLevel.IsAvailable("Nullspace", NoFactory::get())) {
-        // When a fine nullspace have already been defined by user using Set("Nullspace", ...) or
-        // Set("Nullspace1", ...), we use it.
-        tentativeNullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", NoFactory::get());
-      } else {
-        // A User "Nullspace" (nspName) is not available (use factory)
-        tentativeNullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory("Nullspace").get());
+}  // Build
 
-      } // end if "Nullspace" not available
-    } else {
-      tentativeNullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory("Nullspace").get());
-    }
-
-  
-    // Scale!
-    RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-    
-    // determine numPDEs
-    LocalOrdinal numPDEs = 1;
-    if(A->IsView("stridedMaps")==true) {
-      Xpetra::viewLabel_t oldView = A->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-      TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap()) == Teuchos::null, Exceptions::BadCast, "MueLu::ScaledNullspaceFactory::Build: cast to strided row map failed.");
-      numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap())->getFixedBlockSize();
-      oldView = A->SwitchToView(oldView);
-    }
-   
-    GetOStream(Runtime1) << "ScaledNullspaceFactory: Generating scaled nullspace, blocksize = "<< tentativeNullspace->getNumVectors() <<std::endl;
-    nullspace = MultiVectorFactory::Build(tentativeNullspace->getMap(), tentativeNullspace->getNumVectors());
-    *nullspace = *tentativeNullspace;  // Copy the tentative nullspace
-    RCP<MultiVector> blockDiagonal = MultiVectorFactory::Build(A->getDomainMap(), numPDEs);
-    
-    Xpetra::MatrixUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::extractBlockDiagonal(*A,*blockDiagonal);
-    Xpetra::MatrixUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::inverseScaleBlockDiagonal(*blockDiagonal,true,*nullspace);
-
-    // provide "Scaled Nullspace" variable on current level (used by TentativePFactory)
-    Set(currentLevel, "Scaled Nullspace", nullspace);
-
-  } // Build
-
-} //namespace MueLu
-
-#endif // MUELU_SCALEDNULLSPACEFACTORY_DEF_HPP
+}  // namespace MueLu
+
+#endif  // MUELU_SCALEDNULLSPACEFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_decl.hpp
index a5993dd08d2b..47433bc13781 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_decl.hpp
@@ -75,97 +75,96 @@
 
 namespace MueLu {
 
-  /*!
-    @class TentativePFactory class.
-    @brief Factory for building tentative prolongator.
+/*!
+  @class TentativePFactory class.
+  @brief Factory for building tentative prolongator.
 
-    Factory for creating tentative prolongator.   Nullspace vectors are split across aggregates so that they
-    have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
-    tentative prolongator, and the R becomes the coarse nullspace.
+  Factory for creating tentative prolongator.   Nullspace vectors are split across aggregates so that they
+  have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
+  tentative prolongator, and the R becomes the coarse nullspace.
 
-    Note that the TentativePFactory also creates the coarse null space vectors, that is, it serves as generating
-    factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
-    level one has to use the NullspaceFactory.
+  Note that the TentativePFactory also creates the coarse null space vectors, that is, it serves as generating
+  factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
+  level one has to use the NullspaceFactory.
 
-    @ingroup MueLuTransferClasses
+  @ingroup MueLuTransferClasses
 
-    ## Input/output of TentativePFactory ##
+  ## Input/output of TentativePFactory ##
 
-    ### User parameters of TentativePFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A                  | Factory | null |   | * | * | Generating factory of the matrix A
-     Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
-     UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
-     CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
+  ### User parameters of TentativePFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A                  | Factory | null |   | * | * | Generating factory of the matrix A
+   Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+   UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
+   CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TentativePFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see TentativePFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TentativePFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see TentativePFactory::DeclareInput).
 
-    ### Variables provided by TentativePFactory ###
+  ### Variables provided by TentativePFactory ###
 
-    After TentativePFactory::Build the following data is available (if requested)
+  After TentativePFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | TentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
-    | Nullspace | TentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
-  */
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | TentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
+  | Nullspace | TentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
+*/
 
-template <class Scalar = DefaultScalar,
-          class LocalOrdinal = DefaultLocalOrdinal,
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
           class GlobalOrdinal = DefaultGlobalOrdinal,
-          class Node = DefaultNode>
-  class TentativePFactory : public PFactory {
+          class Node          = DefaultNode>
+class TentativePFactory : public PFactory {
 #undef MUELU_TENTATIVEPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    TentativePFactory() { }
+  //! Constructor
+  TentativePFactory() {}
 
-    //! Destructor.
-    virtual ~TentativePFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~TentativePFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
+ private:
+  void BuildPuncoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                       RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
+  void BuildPcoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                     RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace) const;
+  void BuildPuncoupledBlockCrs(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                               RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
 
-    void BuildPuncoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                         RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
-    void BuildPcoupled  (RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                         RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace) const;
-    void BuildPuncoupledBlockCrs(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                         RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
+  mutable bool bTransferCoordinates_ = false;
 
-    mutable bool bTransferCoordinates_ = false;
+};  // class TentativePFactory
 
-  }; //class TentativePFactory
+}  // namespace MueLu
 
-} //namespace MueLu
-
-//TODO: noQR_
+// TODO: noQR_
 
 #define MUELU_TENTATIVEPFACTORY_SHORT
-#endif // MUELU_TENTATIVEPFACTORY_DECL_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_def.hpp
index 952b90e4da57..e6246158bde0 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_def.hpp
@@ -73,1111 +73,1084 @@
 #include "MueLu_PerfUtils.hpp"
 #include "MueLu_Utilities.hpp"
 
-
-
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
-    SET_VALID_ENTRY("tentative: calculate qr");
-    SET_VALID_ENTRY("tentative: build coarse coordinates");
-    SET_VALID_ENTRY("tentative: constant column sums");
-#undef  SET_VALID_ENTRY
-    validParamList->set< std::string >("Nullspace name", "Nullspace", "Name for the input nullspace");
-
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Aggregates",         Teuchos::null, "Generating factory of the aggregates");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",          Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("Scaled Nullspace",   Teuchos::null, "Generating factory of the scaled nullspace");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap",          Teuchos::null, "Generating factory of the coarse map");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory of the coordinates");
-    validParamList->set< RCP<const FactoryBase> >("Node Comm",          Teuchos::null, "Generating factory of the node level communicator");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+  SET_VALID_ENTRY("tentative: calculate qr");
+  SET_VALID_ENTRY("tentative: build coarse coordinates");
+  SET_VALID_ENTRY("tentative: constant column sums");
+#undef SET_VALID_ENTRY
+  validParamList->set<std::string>("Nullspace name", "Nullspace", "Name for the input nullspace");
+
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Generating factory of the aggregates");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase> >("Scaled Nullspace", Teuchos::null, "Generating factory of the scaled nullspace");
+  validParamList->set<RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase> >("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Generating factory of the coordinates");
+  validParamList->set<RCP<const FactoryBase> >("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  const ParameterList& pL = GetParameterList();
+  // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
+  std::string nspName = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  Input(fineLevel, "A");
+  Input(fineLevel, "Aggregates");
+  Input(fineLevel, nspName);
+  Input(fineLevel, "UnAmalgamationInfo");
+  Input(fineLevel, "CoarseMap");
+  if (fineLevel.GetLevelID() == 0 &&
+      fineLevel.IsAvailable("Coordinates", NoFactory::get()) &&  // we have coordinates (provided by user app)
+      pL.get<bool>("tentative: build coarse coordinates")) {     // and we want coordinates on other levels
+    bTransferCoordinates_ = true;                                // then set the transfer coordinates flag to true
+    Input(fineLevel, "Coordinates");
+  } else if (bTransferCoordinates_) {
+    Input(fineLevel, "Coordinates");
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
+  typedef Xpetra::MultiVector<coordinate_type, LO, GO, NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVectorFactory<coordinate_type, LO, GO, NO> RealValuedMultiVectorFactory;
+
+  const ParameterList& pL = GetParameterList();
+  std::string nspName     = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  RCP<Matrix> Ptentative;
+  RCP<Matrix> A              = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<Aggregates> aggregates = Get<RCP<Aggregates> >(fineLevel, "Aggregates");
+  // No coarse DoFs so we need to bail by setting Ptentattive to null and returning
+  //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (aggregates->GetNumGlobalAggregatesComputeIfNeeded() == 0) {
+    Ptentative = Teuchos::null;
+    Set(coarseLevel, "P", Ptentative);
+    return;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-
-    const ParameterList& pL = GetParameterList();
-    // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "Aggregates");
-    Input(fineLevel, nspName);
-    Input(fineLevel, "UnAmalgamationInfo");
-    Input(fineLevel, "CoarseMap");
-    if( fineLevel.GetLevelID() == 0 &&
-        fineLevel.IsAvailable("Coordinates", NoFactory::get()) &&     // we have coordinates (provided by user app)
-        pL.get<bool>("tentative: build coarse coordinates") ) {       // and we want coordinates on other levels
-      bTransferCoordinates_ = true;                                   // then set the transfer coordinates flag to true
-      Input(fineLevel, "Coordinates");
-    } else if (bTransferCoordinates_) {
-      Input(fineLevel, "Coordinates");
-    }
+  RCP<AmalgamationInfo> amalgInfo = Get<RCP<AmalgamationInfo> >(fineLevel, "UnAmalgamationInfo");
+  RCP<MultiVector> fineNullspace  = Get<RCP<MultiVector> >(fineLevel, nspName);
+  RCP<const Map> coarseMap        = Get<RCP<const Map> >(fineLevel, "CoarseMap");
+  RCP<RealValuedMultiVector> fineCoords;
+  if (bTransferCoordinates_) {
+    fineCoords = Get<RCP<RealValuedMultiVector> >(fineLevel, "Coordinates");
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  // FIXME: We should remove the NodeComm on levels past the threshold
+  if (fineLevel.IsAvailable("Node Comm")) {
+    RCP<const Teuchos::Comm<int> > nodeComm = Get<RCP<const Teuchos::Comm<int> > >(fineLevel, "Node Comm");
+    Set<RCP<const Teuchos::Comm<int> > >(coarseLevel, "Node Comm", nodeComm);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
-    typedef Xpetra::MultiVector<coordinate_type,LO,GO,NO> RealValuedMultiVector;
-    typedef Xpetra::MultiVectorFactory<coordinate_type,LO,GO,NO> RealValuedMultiVectorFactory;
-
-    const ParameterList& pL = GetParameterList();
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-
-    RCP<Matrix>                Ptentative;
-    RCP<Matrix>                A             = Get< RCP<Matrix> >               (fineLevel, "A");
-    RCP<Aggregates>            aggregates    = Get< RCP<Aggregates> >           (fineLevel, "Aggregates");
-    // No coarse DoFs so we need to bail by setting Ptentattive to null and returning
-    //  This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-    if ( aggregates->GetNumGlobalAggregatesComputeIfNeeded() == 0) {
-      Ptentative = Teuchos::null;
-      Set(coarseLevel, "P", Ptentative);
-      return;
-    }
+  // NOTE:  We check DomainMap here rather than RowMap because those are different for BlockCrs matrices
+  TEUCHOS_TEST_FOR_EXCEPTION(A->getDomainMap()->getLocalNumElements() != fineNullspace->getMap()->getLocalNumElements(),
+                             Exceptions::RuntimeError, "MueLu::TentativePFactory::MakeTentative: Size mismatch between A and Nullspace");
 
-    RCP<AmalgamationInfo>      amalgInfo     = Get< RCP<AmalgamationInfo> >     (fineLevel, "UnAmalgamationInfo");
-    RCP<MultiVector>           fineNullspace = Get< RCP<MultiVector> >          (fineLevel, nspName);
-    RCP<const Map>             coarseMap     = Get< RCP<const Map> >            (fineLevel, "CoarseMap");
-    RCP<RealValuedMultiVector> fineCoords;
-    if(bTransferCoordinates_) {
-      fineCoords = Get< RCP<RealValuedMultiVector> >(fineLevel, "Coordinates");
-    }
+  RCP<MultiVector> coarseNullspace;
+  RCP<RealValuedMultiVector> coarseCoords;
 
-    // FIXME: We should remove the NodeComm on levels past the threshold
-    if(fineLevel.IsAvailable("Node Comm")) {
-      RCP<const Teuchos::Comm<int> > nodeComm = Get<RCP<const Teuchos::Comm<int> > >(fineLevel,"Node Comm");
-      Set<RCP<const Teuchos::Comm<int> > >(coarseLevel, "Node Comm", nodeComm);
+  if (bTransferCoordinates_) {
+    //*** Create the coarse coordinates ***
+    // First create the coarse map and coarse multivector
+    ArrayView<const GO> elementAList = coarseMap->getLocalElementList();
+    LO blkSize                       = 1;
+    if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null) {
+      blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
     }
+    GO indexBase      = coarseMap->getIndexBase();
+    LO numCoarseNodes = Teuchos::as<LO>(elementAList.size() / blkSize);
+    Array<GO> nodeList(numCoarseNodes);
+    const int numDimensions = fineCoords->getNumVectors();
 
-    // NOTE:  We check DomainMap here rather than RowMap because those are different for BlockCrs matrices
-    TEUCHOS_TEST_FOR_EXCEPTION( A->getDomainMap()->getLocalNumElements() != fineNullspace->getMap()->getLocalNumElements(),
-			       Exceptions::RuntimeError,"MueLu::TentativePFactory::MakeTentative: Size mismatch between A and Nullspace");
-
-    RCP<MultiVector>           coarseNullspace;
-    RCP<RealValuedMultiVector> coarseCoords;
-
-    if(bTransferCoordinates_) {
-      //*** Create the coarse coordinates ***
-      // First create the coarse map and coarse multivector
-      ArrayView<const GO> elementAList   = coarseMap->getLocalElementList();
-      LO                  blkSize        = 1;
-      if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null) {
-        blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
-      }
-      GO                  indexBase      = coarseMap->getIndexBase();
-      LO                  numCoarseNodes = Teuchos::as<LO>(elementAList.size() / blkSize);
-      Array<GO>           nodeList(numCoarseNodes);
-      const int           numDimensions  = fineCoords->getNumVectors();
-
-      for (LO i = 0; i < numCoarseNodes; i++) {
-        nodeList[i] = (elementAList[i*blkSize]-indexBase)/blkSize + indexBase;
-      }
-      RCP<const Map> coarseCoordsMap = MapFactory::Build(fineCoords->getMap()->lib(),
-                                                         Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                         nodeList,
-                                                         indexBase,
-                                                         fineCoords->getMap()->getComm());
-      coarseCoords = RealValuedMultiVectorFactory::Build(coarseCoordsMap, numDimensions);
-
-      // Create overlapped fine coordinates to reduce global communication
-      RCP<RealValuedMultiVector> ghostedCoords;
-      if (aggregates->AggregatesCrossProcessors()) {
-        RCP<const Map>    aggMap = aggregates->GetMap();
-        RCP<const Import> importer     = ImportFactory::Build(fineCoords->getMap(), aggMap);
-
-        ghostedCoords = RealValuedMultiVectorFactory::Build(aggMap, numDimensions);
-        ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
-      } else {
-        ghostedCoords = fineCoords;
-      }
-
-      // Get some info about aggregates
-      int                         myPID        = coarseCoordsMap->getComm()->getRank();
-      LO                          numAggs      = aggregates->GetNumAggregates();
-      ArrayRCP<LO>                aggSizes     = aggregates->ComputeAggregateSizesArrayRCP();
-      const ArrayRCP<const LO>    vertex2AggID = aggregates->GetVertex2AggId()->getData(0);
-      const ArrayRCP<const LO>    procWinner   = aggregates->GetProcWinner()->getData(0);
-
-      // Fill in coarse coordinates
-      for (int dim = 0; dim < numDimensions; ++dim) {
-        ArrayRCP<const coordinate_type> fineCoordsData = ghostedCoords->getData(dim);
-        ArrayRCP<coordinate_type>     coarseCoordsData = coarseCoords->getDataNonConst(dim);
-
-        for (LO lnode = 0; lnode < Teuchos::as<LO>(vertex2AggID.size()); lnode++) {
-          if (procWinner[lnode] == myPID &&
-              lnode < fineCoordsData.size() &&
-              vertex2AggID[lnode] < coarseCoordsData.size() &&
-              Teuchos::ScalarTraits<coordinate_type>::isnaninf(fineCoordsData[lnode]) == false) {
-            coarseCoordsData[vertex2AggID[lnode]] += fineCoordsData[lnode];
-          }
-        }
-        for (LO agg = 0; agg < numAggs; agg++) {
-          coarseCoordsData[agg] /= aggSizes[agg];
-        }
-      }
+    for (LO i = 0; i < numCoarseNodes; i++) {
+      nodeList[i] = (elementAList[i * blkSize] - indexBase) / blkSize + indexBase;
+    }
+    RCP<const Map> coarseCoordsMap = MapFactory::Build(fineCoords->getMap()->lib(),
+                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                       nodeList,
+                                                       indexBase,
+                                                       fineCoords->getMap()->getComm());
+    coarseCoords                   = RealValuedMultiVectorFactory::Build(coarseCoordsMap, numDimensions);
+
+    // Create overlapped fine coordinates to reduce global communication
+    RCP<RealValuedMultiVector> ghostedCoords;
+    if (aggregates->AggregatesCrossProcessors()) {
+      RCP<const Map> aggMap      = aggregates->GetMap();
+      RCP<const Import> importer = ImportFactory::Build(fineCoords->getMap(), aggMap);
+
+      ghostedCoords = RealValuedMultiVectorFactory::Build(aggMap, numDimensions);
+      ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
+    } else {
+      ghostedCoords = fineCoords;
     }
 
-    if (!aggregates->AggregatesCrossProcessors()) {
-      if(Xpetra::Helpers<SC,LO,GO,NO>::isTpetraBlockCrs(A)) {
-        BuildPuncoupledBlockCrs(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace,coarseLevel.GetLevelID());
+    // Get some info about aggregates
+    int myPID                             = coarseCoordsMap->getComm()->getRank();
+    LO numAggs                            = aggregates->GetNumAggregates();
+    ArrayRCP<LO> aggSizes                 = aggregates->ComputeAggregateSizesArrayRCP();
+    const ArrayRCP<const LO> vertex2AggID = aggregates->GetVertex2AggId()->getData(0);
+    const ArrayRCP<const LO> procWinner   = aggregates->GetProcWinner()->getData(0);
+
+    // Fill in coarse coordinates
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      ArrayRCP<const coordinate_type> fineCoordsData = ghostedCoords->getData(dim);
+      ArrayRCP<coordinate_type> coarseCoordsData     = coarseCoords->getDataNonConst(dim);
+
+      for (LO lnode = 0; lnode < Teuchos::as<LO>(vertex2AggID.size()); lnode++) {
+        if (procWinner[lnode] == myPID &&
+            lnode < fineCoordsData.size() &&
+            vertex2AggID[lnode] < coarseCoordsData.size() &&
+            Teuchos::ScalarTraits<coordinate_type>::isnaninf(fineCoordsData[lnode]) == false) {
+          coarseCoordsData[vertex2AggID[lnode]] += fineCoordsData[lnode];
+        }
       }
-      else {
-        BuildPuncoupled(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace,coarseLevel.GetLevelID());
+      for (LO agg = 0; agg < numAggs; agg++) {
+        coarseCoordsData[agg] /= aggSizes[agg];
       }
     }
-    else
-      BuildPcoupled(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace);
-
-    // If available, use striding information of fine level matrix A for range
-    // map and coarseMap as domain map; otherwise use plain range map of
-    // Ptent = plain range map of A for range map and coarseMap as domain map.
-    // NOTE:
-    // The latter is not really safe, since there is no striding information
-    // for the range map. This is not really a problem, since striding
-    // information is always available on the intermedium levels and the
-    // coarsest levels.
-    if (A->IsView("stridedMaps") == true)
-      Ptentative->CreateView("stridedMaps", A->getRowMap("stridedMaps"), coarseMap);
-    else
-      Ptentative->CreateView("stridedMaps", Ptentative->getRangeMap(),   coarseMap);
-
-    if(bTransferCoordinates_) {
-      Set(coarseLevel, "Coordinates", coarseCoords);
-    }
-    Set(coarseLevel, "Nullspace",   coarseNullspace);
-    Set(coarseLevel, "P",           Ptentative);
+  }
 
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ptentative, "Ptent", params);
+  if (!aggregates->AggregatesCrossProcessors()) {
+    if (Xpetra::Helpers<SC, LO, GO, NO>::isTpetraBlockCrs(A)) {
+      BuildPuncoupledBlockCrs(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace, coarseLevel.GetLevelID());
+    } else {
+      BuildPuncoupled(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace, coarseLevel.GetLevelID());
     }
+  } else
+    BuildPcoupled(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace);
+
+  // If available, use striding information of fine level matrix A for range
+  // map and coarseMap as domain map; otherwise use plain range map of
+  // Ptent = plain range map of A for range map and coarseMap as domain map.
+  // NOTE:
+  // The latter is not really safe, since there is no striding information
+  // for the range map. This is not really a problem, since striding
+  // information is always available on the intermedium levels and the
+  // coarsest levels.
+  if (A->IsView("stridedMaps") == true)
+    Ptentative->CreateView("stridedMaps", A->getRowMap("stridedMaps"), coarseMap);
+  else
+    Ptentative->CreateView("stridedMaps", Ptentative->getRangeMap(), coarseMap);
+
+  if (bTransferCoordinates_) {
+    Set(coarseLevel, "Coordinates", coarseCoords);
   }
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "P", Ptentative);
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildPuncoupledBlockCrs(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                          RCP<const Map> coarsePointMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const {
-    /* This routine generates a BlockCrs P for a BlockCrs A.  There are a few assumptions here, which meet the use cases we care about, but could 
-       be generalized later, if we ever need to do so:
-       1) Null space dimension === block size of matrix:  So no elasticity right now
-       2) QR is not supported:  Under assumption #1, this shouldn't cause problems.
-       3) Maps are "good": Aka the first chunk of the ColMap is the RowMap.
-
-       These assumptions keep our code way simpler and still support the use cases we actually care about.
-     */
-
-    RCP<const Map> rowMap     = A->getRowMap();
-    RCP<const Map> rangeMap   = A->getRangeMap();
-    RCP<const Map> colMap     = A->getColMap();
-    //    const size_t numFinePointRows = rangeMap->getLocalNumElements();
-    const size_t numFineBlockRows = rowMap->getLocalNumElements();
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    // typedef typename STS::magnitudeType Magnitude;
-    const SC     zero      = STS::zero();
-    const SC     one       = STS::one();
-    const LO     INVALID   = Teuchos::OrdinalTraits<LO>::invalid();
-
-    const GO     numAggs   = aggregates->GetNumAggregates();
-    const size_t NSDim     = fineNullspace->getNumVectors();
-    ArrayRCP<LO> aggSizes  = aggregates->ComputeAggregateSizesArrayRCP();
-
-    // Need to generate the coarse block map
-    // NOTE: We assume NSDim == block size here
-    // NOTE: We also assume that coarseMap has contiguous GIDs
-    //const size_t numCoarsePointRows = coarsePointMap->getLocalNumElements();
-    const size_t numCoarseBlockRows = coarsePointMap->getLocalNumElements() / NSDim;
-    RCP<const Map> coarseBlockMap = MapFactory::Build(coarsePointMap->lib(),
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ptentative, "Ptent", params);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildPuncoupledBlockCrs(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                            RCP<const Map> coarsePointMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const {
+  /* This routine generates a BlockCrs P for a BlockCrs A.  There are a few assumptions here, which meet the use cases we care about, but could
+     be generalized later, if we ever need to do so:
+     1) Null space dimension === block size of matrix:  So no elasticity right now
+     2) QR is not supported:  Under assumption #1, this shouldn't cause problems.
+     3) Maps are "good": Aka the first chunk of the ColMap is the RowMap.
+
+     These assumptions keep our code way simpler and still support the use cases we actually care about.
+   */
+
+  RCP<const Map> rowMap   = A->getRowMap();
+  RCP<const Map> rangeMap = A->getRangeMap();
+  RCP<const Map> colMap   = A->getColMap();
+  //    const size_t numFinePointRows = rangeMap->getLocalNumElements();
+  const size_t numFineBlockRows = rowMap->getLocalNumElements();
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  // typedef typename STS::magnitudeType Magnitude;
+  const SC zero    = STS::zero();
+  const SC one     = STS::one();
+  const LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  const GO numAggs      = aggregates->GetNumAggregates();
+  const size_t NSDim    = fineNullspace->getNumVectors();
+  ArrayRCP<LO> aggSizes = aggregates->ComputeAggregateSizesArrayRCP();
+
+  // Need to generate the coarse block map
+  // NOTE: We assume NSDim == block size here
+  // NOTE: We also assume that coarseMap has contiguous GIDs
+  // const size_t numCoarsePointRows = coarsePointMap->getLocalNumElements();
+  const size_t numCoarseBlockRows = coarsePointMap->getLocalNumElements() / NSDim;
+  RCP<const Map> coarseBlockMap   = MapFactory::Build(coarsePointMap->lib(),
                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
                                                       numCoarseBlockRows,
                                                       coarsePointMap->getIndexBase(),
-                                                      coarsePointMap->getComm());    
-    // Sanity checking
-    const ParameterList& pL = GetParameterList();
-    const bool &doQRStep = pL.get<bool>("tentative: calculate qr");
-    const bool &constantColSums = pL.get<bool>("tentative: constant column sums");
-
-    TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && constantColSums,Exceptions::RuntimeError,
-                               "MueLu::TentativePFactory::MakeTentative: cannot use 'constant column sums' and 'calculate qr' at the same time");
-
-    // The aggregates use the amalgamated column map, which in this case is what we want
-
-    // Aggregates map is based on the amalgamated column map
-    // We can skip global-to-local conversion if LIDs in row map are
-    // same as LIDs in column map
-    bool goodMap = MueLu::Utilities<SC,LO,GO,NO>::MapsAreNested(*rowMap, *colMap);
-
-    // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
-    // aggStart is a pointer into aggToRowMapLO
-    // aggStart[i]..aggStart[i+1] are indices into aggToRowMapLO
-    // aggToRowMapLO[aggStart[i]]..aggToRowMapLO[aggStart[i+1]-1] are the DOFs in aggregate i
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<LO> aggToRowMapLO;
-    ArrayRCP<GO> aggToRowMapGO;
-    if (goodMap) {
-      amalgInfo->UnamalgamateAggregatesLO(*aggregates, aggStart, aggToRowMapLO);
-      GetOStream(Runtime1) << "Column map is consistent with the row map, good." << std::endl;
-    } else {
-      throw std::runtime_error("TentativePFactory::PuncoupledBlockCrs: Inconsistent maps not currently supported");
-    }
-  
-    coarseNullspace = MultiVectorFactory::Build(coarsePointMap, NSDim);
-
-    // Pull out the nullspace vectors so that we can have random access.
-    ArrayRCP<ArrayRCP<const SC> > fineNS  (NSDim);
-    ArrayRCP<ArrayRCP<SC> >       coarseNS(NSDim);
-    for (size_t i = 0; i < NSDim; i++) {
-      fineNS[i] = fineNullspace->getData(i);
-      if (coarsePointMap->getLocalNumElements() > 0)
-        coarseNS[i] = coarseNullspace->getDataNonConst(i);
-    }
-
-    // BlockCrs requires that we build the (block) graph first, so let's do that...
-    // NOTE: Because we're assuming that the NSDim == BlockSize, we only have one 
-    // block non-zero per row in the matrix;
-    RCP<CrsGraph> BlockGraph = CrsGraphFactory::Build(rowMap,coarseBlockMap,0);
-    ArrayRCP<size_t>  iaPtent;
-    ArrayRCP<LO>      jaPtent;
-    BlockGraph->allocateAllIndices(numFineBlockRows, iaPtent, jaPtent);
-    ArrayView<size_t> ia  = iaPtent();
-    ArrayView<LO>     ja  = jaPtent();
-
-    for (size_t i = 0; i < numFineBlockRows; i++) {
-      ia[i] = i;
-      ja[i] = INVALID;
-    }
-    ia[numCoarseBlockRows] = numCoarseBlockRows;
-
+                                                      coarsePointMap->getComm());
+  // Sanity checking
+  const ParameterList& pL     = GetParameterList();
+  const bool& doQRStep        = pL.get<bool>("tentative: calculate qr");
+  const bool& constantColSums = pL.get<bool>("tentative: constant column sums");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && constantColSums, Exceptions::RuntimeError,
+                             "MueLu::TentativePFactory::MakeTentative: cannot use 'constant column sums' and 'calculate qr' at the same time");
+
+  // The aggregates use the amalgamated column map, which in this case is what we want
+
+  // Aggregates map is based on the amalgamated column map
+  // We can skip global-to-local conversion if LIDs in row map are
+  // same as LIDs in column map
+  bool goodMap = MueLu::Utilities<SC, LO, GO, NO>::MapsAreNested(*rowMap, *colMap);
+
+  // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
+  // aggStart is a pointer into aggToRowMapLO
+  // aggStart[i]..aggStart[i+1] are indices into aggToRowMapLO
+  // aggToRowMapLO[aggStart[i]]..aggToRowMapLO[aggStart[i+1]-1] are the DOFs in aggregate i
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<LO> aggToRowMapLO;
+  ArrayRCP<GO> aggToRowMapGO;
+  if (goodMap) {
+    amalgInfo->UnamalgamateAggregatesLO(*aggregates, aggStart, aggToRowMapLO);
+    GetOStream(Runtime1) << "Column map is consistent with the row map, good." << std::endl;
+  } else {
+    throw std::runtime_error("TentativePFactory::PuncoupledBlockCrs: Inconsistent maps not currently supported");
+  }
 
-    for (GO agg = 0; agg < numAggs; agg++) {
-      LO aggSize = aggStart[agg+1] - aggStart[agg];
-      Xpetra::global_size_t offset = agg;
+  coarseNullspace = MultiVectorFactory::Build(coarsePointMap, NSDim);
 
-      for (LO j = 0; j < aggSize; j++) {
-        // FIXME: Allow for bad maps
-        const LO localRow = aggToRowMapLO[aggStart[agg]+j];
-        const size_t rowStart = ia[localRow];
-        ja[rowStart] = offset;
-      }      
-    }
+  // Pull out the nullspace vectors so that we can have random access.
+  ArrayRCP<ArrayRCP<const SC> > fineNS(NSDim);
+  ArrayRCP<ArrayRCP<SC> > coarseNS(NSDim);
+  for (size_t i = 0; i < NSDim; i++) {
+    fineNS[i] = fineNullspace->getData(i);
+    if (coarsePointMap->getLocalNumElements() > 0)
+      coarseNS[i] = coarseNullspace->getDataNonConst(i);
+  }
 
-    // Compress storage (remove all INVALID, which happen when we skip zeros)
-    // We do that in-place
-    size_t ia_tmp = 0, nnz = 0;
-    for (size_t i = 0; i < numFineBlockRows; i++) {
-      for (size_t j = ia_tmp; j < ia[i+1]; j++)
-        if (ja[j] != INVALID) {
-          ja [nnz] = ja [j];
-          nnz++;
-        }
-      ia_tmp  = ia[i+1];
-      ia[i+1] = nnz;
-    }
+  // BlockCrs requires that we build the (block) graph first, so let's do that...
+  // NOTE: Because we're assuming that the NSDim == BlockSize, we only have one
+  // block non-zero per row in the matrix;
+  RCP<CrsGraph> BlockGraph = CrsGraphFactory::Build(rowMap, coarseBlockMap, 0);
+  ArrayRCP<size_t> iaPtent;
+  ArrayRCP<LO> jaPtent;
+  BlockGraph->allocateAllIndices(numFineBlockRows, iaPtent, jaPtent);
+  ArrayView<size_t> ia = iaPtent();
+  ArrayView<LO> ja     = jaPtent();
+
+  for (size_t i = 0; i < numFineBlockRows; i++) {
+    ia[i] = i;
+    ja[i] = INVALID;
+  }
+  ia[numCoarseBlockRows] = numCoarseBlockRows;
 
-    if (rowMap->lib() == Xpetra::UseTpetra) {
-      // - Cannot resize for Epetra, as it checks for same pointers
-      // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
-      // NOTE: these invalidate ja and val views
-      jaPtent .resize(nnz);
-    }
+  for (GO agg = 0; agg < numAggs; agg++) {
+    LO aggSize                   = aggStart[agg + 1] - aggStart[agg];
+    Xpetra::global_size_t offset = agg;
 
-    GetOStream(Runtime1) << "TentativePFactory : generating block graph" << std::endl;
-    BlockGraph->setAllIndices(iaPtent, jaPtent);
-
-    // Managing labels & constants for ESFC
-    {
-      RCP<ParameterList> FCparams;
-      if(pL.isSublist("matrixmatrix: kernel params"))
-        FCparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-      else
-        FCparams= rcp(new ParameterList);
-      // By default, we don't need global constants for TentativeP, but we do want it for the graph
-      // if we're printing statistics, so let's leave it on for now.
-      FCparams->set("compute global constants",FCparams->get("compute global constants",true));
-      std::string levelIDs = toString(levelID);
-      FCparams->set("Timer Label",std::string("MueLu::TentativeP-")+levelIDs);
-      RCP<const Export> dummy_e;
-      RCP<const Import> dummy_i;
-      BlockGraph->expertStaticFillComplete(coarseBlockMap,rowMap,dummy_i,dummy_e,FCparams);
+    for (LO j = 0; j < aggSize; j++) {
+      // FIXME: Allow for bad maps
+      const LO localRow     = aggToRowMapLO[aggStart[agg] + j];
+      const size_t rowStart = ia[localRow];
+      ja[rowStart]          = offset;
     }
+  }
 
-    // Now let's make a BlockCrs Matrix
-    // NOTE: Assumes block size== NSDim
-    RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > P_xpetra = Xpetra::CrsMatrixFactory<SC,LO,GO,NO>::BuildBlock(BlockGraph, coarsePointMap, rangeMap,NSDim);
-    RCP<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> >(P_xpetra);
-    if(P_tpetra.is_null()) throw std::runtime_error("BuildPUncoupled: Matrix factory did not return a Tpetra::BlockCrsMatrix");
-    RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
-
-    /////////////////////////////
-    //      "no-QR" option     //
-    /////////////////////////////
-    // Local Q factor is just the fine nullspace support over the current aggregate.
-    // Local R factor is the identity.
-    // NOTE: We're not going to do a QR here as we're assuming that blocksize == NSDim
-    // NOTE: "goodMap" case only
-    Teuchos::Array<Scalar> block(NSDim*NSDim, zero);
-    Teuchos::Array<LO> bcol(1);
-
-    GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
-    for (LO agg = 0; agg < numAggs; agg++) {
-      bcol[0] = agg;
-      const LO aggSize = aggStart[agg+1] - aggStart[agg];
-      Xpetra::global_size_t offset = agg*NSDim;
-
-      // Process each row in the local Q factor
-      // NOTE: Blocks are in row-major order
-      for (LO j = 0; j < aggSize; j++) {
-        const LO localBlockRow = aggToRowMapLO[aggStart[agg]+j];
+  // Compress storage (remove all INVALID, which happen when we skip zeros)
+  // We do that in-place
+  size_t ia_tmp = 0, nnz = 0;
+  for (size_t i = 0; i < numFineBlockRows; i++) {
+    for (size_t j = ia_tmp; j < ia[i + 1]; j++)
+      if (ja[j] != INVALID) {
+        ja[nnz] = ja[j];
+        nnz++;
+      }
+    ia_tmp    = ia[i + 1];
+    ia[i + 1] = nnz;
+  }
 
-        for (size_t r = 0; r < NSDim; r++) {
-          LO localPointRow = localBlockRow*NSDim + r;
-          for (size_t c = 0; c < NSDim; c++) 
-            block[r*NSDim+c] = fineNS[c][localPointRow];
-        }
-        // NOTE: Assumes columns==aggs and are ordered sequentially
-        P_tpetra->replaceLocalValues(localBlockRow,bcol(),block());
+  if (rowMap->lib() == Xpetra::UseTpetra) {
+    // - Cannot resize for Epetra, as it checks for same pointers
+    // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
+    // NOTE: these invalidate ja and val views
+    jaPtent.resize(nnz);
+  }
 
-      }//end aggSize
-        
-      for (size_t j = 0; j < NSDim; j++)
-        coarseNS[j][offset+j] = one;
-      
-    } //for (GO agg = 0; agg < numAggs; agg++)
+  GetOStream(Runtime1) << "TentativePFactory : generating block graph" << std::endl;
+  BlockGraph->setAllIndices(iaPtent, jaPtent);
 
-    Ptentative = P_wrap;
+  // Managing labels & constants for ESFC
+  {
+    RCP<ParameterList> FCparams;
+    if (pL.isSublist("matrixmatrix: kernel params"))
+      FCparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+    else
+      FCparams = rcp(new ParameterList);
+    // By default, we don't need global constants for TentativeP, but we do want it for the graph
+    // if we're printing statistics, so let's leave it on for now.
+    FCparams->set("compute global constants", FCparams->get("compute global constants", true));
+    std::string levelIDs = toString(levelID);
+    FCparams->set("Timer Label", std::string("MueLu::TentativeP-") + levelIDs);
+    RCP<const Export> dummy_e;
+    RCP<const Import> dummy_i;
+    BlockGraph->expertStaticFillComplete(coarseBlockMap, rowMap, dummy_i, dummy_e, FCparams);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildPcoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace) const {
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType Magnitude;
-    const SC     zero      = STS::zero();
-    const SC     one       = STS::one();
-
-    // number of aggregates
-    GO numAggs = aggregates->GetNumAggregates();
-
-    // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
-    // aggStart is a pointer into aggToRowMap
-    // aggStart[i]..aggStart[i+1] are indices into aggToRowMap
-    // aggToRowMap[aggStart[i]]..aggToRowMap[aggStart[i+1]-1] are the DOFs in aggregate i
-    ArrayRCP<LO> aggStart;
-    ArrayRCP< GO > aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-
-    // find size of the largest aggregate
-    LO maxAggSize=0;
-    for (GO i=0; i<numAggs; ++i) {
-      LO sizeOfThisAgg = aggStart[i+1] - aggStart[i];
-      if (sizeOfThisAgg > maxAggSize) maxAggSize = sizeOfThisAgg;
-    }
+  // Now let's make a BlockCrs Matrix
+  // NOTE: Assumes block size== NSDim
+  RCP<Xpetra::CrsMatrix<SC, LO, GO, NO> > P_xpetra            = Xpetra::CrsMatrixFactory<SC, LO, GO, NO>::BuildBlock(BlockGraph, coarsePointMap, rangeMap, NSDim);
+  RCP<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO> >(P_xpetra);
+  if (P_tpetra.is_null()) throw std::runtime_error("BuildPUncoupled: Matrix factory did not return a Tpetra::BlockCrsMatrix");
+  RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
+
+  /////////////////////////////
+  //      "no-QR" option     //
+  /////////////////////////////
+  // Local Q factor is just the fine nullspace support over the current aggregate.
+  // Local R factor is the identity.
+  // NOTE: We're not going to do a QR here as we're assuming that blocksize == NSDim
+  // NOTE: "goodMap" case only
+  Teuchos::Array<Scalar> block(NSDim * NSDim, zero);
+  Teuchos::Array<LO> bcol(1);
+
+  GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
+  for (LO agg = 0; agg < numAggs; agg++) {
+    bcol[0]                      = agg;
+    const LO aggSize             = aggStart[agg + 1] - aggStart[agg];
+    Xpetra::global_size_t offset = agg * NSDim;
+
+    // Process each row in the local Q factor
+    // NOTE: Blocks are in row-major order
+    for (LO j = 0; j < aggSize; j++) {
+      const LO localBlockRow = aggToRowMapLO[aggStart[agg] + j];
+
+      for (size_t r = 0; r < NSDim; r++) {
+        LO localPointRow = localBlockRow * NSDim + r;
+        for (size_t c = 0; c < NSDim; c++)
+          block[r * NSDim + c] = fineNS[c][localPointRow];
+      }
+      // NOTE: Assumes columns==aggs and are ordered sequentially
+      P_tpetra->replaceLocalValues(localBlockRow, bcol(), block());
+
+    }  // end aggSize
+
+    for (size_t j = 0; j < NSDim; j++)
+      coarseNS[j][offset + j] = one;
+
+  }  // for (GO agg = 0; agg < numAggs; agg++)
+
+  Ptentative = P_wrap;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildPcoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                  RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace) const {
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType Magnitude;
+  const SC zero = STS::zero();
+  const SC one  = STS::one();
+
+  // number of aggregates
+  GO numAggs = aggregates->GetNumAggregates();
+
+  // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
+  // aggStart is a pointer into aggToRowMap
+  // aggStart[i]..aggStart[i+1] are indices into aggToRowMap
+  // aggToRowMap[aggStart[i]]..aggToRowMap[aggStart[i+1]-1] are the DOFs in aggregate i
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+
+  // find size of the largest aggregate
+  LO maxAggSize = 0;
+  for (GO i = 0; i < numAggs; ++i) {
+    LO sizeOfThisAgg = aggStart[i + 1] - aggStart[i];
+    if (sizeOfThisAgg > maxAggSize) maxAggSize = sizeOfThisAgg;
+  }
 
-    // dimension of fine level nullspace
-    const size_t NSDim = fineNullspace->getNumVectors();
-
-    // index base for coarse Dof map (usually 0)
-    GO indexBase=A->getRowMap()->getIndexBase();
-
-    const RCP<const Map> nonUniqueMap = amalgInfo->ComputeUnamalgamatedImportDofMap(*aggregates);
-    const RCP<const Map> uniqueMap    = A->getDomainMap();
-    RCP<const Import> importer = ImportFactory::Build(uniqueMap, nonUniqueMap);
-    RCP<MultiVector> fineNullspaceWithOverlap = MultiVectorFactory::Build(nonUniqueMap,NSDim);
-    fineNullspaceWithOverlap->doImport(*fineNullspace,*importer,Xpetra::INSERT);
-
-    // Pull out the nullspace vectors so that we can have random access.
-    ArrayRCP< ArrayRCP<const SC> > fineNS(NSDim);
-    for (size_t i=0; i<NSDim; ++i)
-      fineNS[i] = fineNullspaceWithOverlap->getData(i);
-
-    //Allocate storage for the coarse nullspace.
-    coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
-
-    ArrayRCP< ArrayRCP<SC> > coarseNS(NSDim);
-    for (size_t i=0; i<NSDim; ++i)
-      if (coarseMap->getLocalNumElements() > 0) coarseNS[i] = coarseNullspace->getDataNonConst(i);
-
-    //This makes the rowmap of Ptent the same as that of A->
-    //This requires moving some parts of some local Q's to other processors
-    //because aggregates can span processors.
-    RCP<const Map > rowMapForPtent = A->getRowMap();
-    const Map& rowMapForPtentRef = *rowMapForPtent;
-
-    // Set up storage for the rows of the local Qs that belong to other processors.
-    // FIXME This is inefficient and could be done within the main loop below with std::vector's.
-    RCP<const Map> colMap = A->getColMap();
-
-    RCP<const Map > ghostQMap;
-    RCP<MultiVector> ghostQvalues;
-    Array<RCP<Xpetra::Vector<GO,LO,GO,Node> > > ghostQcolumns;
-    RCP<Xpetra::Vector<GO,LO,GO,Node> > ghostQrowNums;
-    ArrayRCP< ArrayRCP<SC> > ghostQvals;
-    ArrayRCP< ArrayRCP<GO> > ghostQcols;
-    ArrayRCP< GO > ghostQrows;
-
-    Array<GO> ghostGIDs;
-    for (LO j=0; j<numAggs; ++j) {
-      for (LO k=aggStart[j]; k<aggStart[j+1]; ++k) {
-        if (rowMapForPtentRef.isNodeGlobalElement(aggToRowMap[k]) == false) {
-          ghostGIDs.push_back(aggToRowMap[k]);
-        }
+  // dimension of fine level nullspace
+  const size_t NSDim = fineNullspace->getNumVectors();
+
+  // index base for coarse Dof map (usually 0)
+  GO indexBase = A->getRowMap()->getIndexBase();
+
+  const RCP<const Map> nonUniqueMap         = amalgInfo->ComputeUnamalgamatedImportDofMap(*aggregates);
+  const RCP<const Map> uniqueMap            = A->getDomainMap();
+  RCP<const Import> importer                = ImportFactory::Build(uniqueMap, nonUniqueMap);
+  RCP<MultiVector> fineNullspaceWithOverlap = MultiVectorFactory::Build(nonUniqueMap, NSDim);
+  fineNullspaceWithOverlap->doImport(*fineNullspace, *importer, Xpetra::INSERT);
+
+  // Pull out the nullspace vectors so that we can have random access.
+  ArrayRCP<ArrayRCP<const SC> > fineNS(NSDim);
+  for (size_t i = 0; i < NSDim; ++i)
+    fineNS[i] = fineNullspaceWithOverlap->getData(i);
+
+  // Allocate storage for the coarse nullspace.
+  coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
+
+  ArrayRCP<ArrayRCP<SC> > coarseNS(NSDim);
+  for (size_t i = 0; i < NSDim; ++i)
+    if (coarseMap->getLocalNumElements() > 0) coarseNS[i] = coarseNullspace->getDataNonConst(i);
+
+  // This makes the rowmap of Ptent the same as that of A->
+  // This requires moving some parts of some local Q's to other processors
+  // because aggregates can span processors.
+  RCP<const Map> rowMapForPtent = A->getRowMap();
+  const Map& rowMapForPtentRef  = *rowMapForPtent;
+
+  // Set up storage for the rows of the local Qs that belong to other processors.
+  // FIXME This is inefficient and could be done within the main loop below with std::vector's.
+  RCP<const Map> colMap = A->getColMap();
+
+  RCP<const Map> ghostQMap;
+  RCP<MultiVector> ghostQvalues;
+  Array<RCP<Xpetra::Vector<GO, LO, GO, Node> > > ghostQcolumns;
+  RCP<Xpetra::Vector<GO, LO, GO, Node> > ghostQrowNums;
+  ArrayRCP<ArrayRCP<SC> > ghostQvals;
+  ArrayRCP<ArrayRCP<GO> > ghostQcols;
+  ArrayRCP<GO> ghostQrows;
+
+  Array<GO> ghostGIDs;
+  for (LO j = 0; j < numAggs; ++j) {
+    for (LO k = aggStart[j]; k < aggStart[j + 1]; ++k) {
+      if (rowMapForPtentRef.isNodeGlobalElement(aggToRowMap[k]) == false) {
+        ghostGIDs.push_back(aggToRowMap[k]);
       }
     }
-    ghostQMap = MapFactory::Build(A->getRowMap()->lib(),
-                                  Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                  ghostGIDs,
-                                  indexBase, A->getRowMap()->getComm()); //JG:Xpetra::global_size_t>?
-    //Vector to hold bits of Q that go to other processors.
-    ghostQvalues = MultiVectorFactory::Build(ghostQMap,NSDim);
-    //Note that Epetra does not support MultiVectors templated on Scalar != double.
-    //So to work around this, we allocate an array of Vectors.  This shouldn't be too
-    //expensive, as the number of Vectors is NSDim.
-    ghostQcolumns.resize(NSDim);
-    for (size_t i=0; i<NSDim; ++i)
-      ghostQcolumns[i] = Xpetra::VectorFactory<GO,LO,GO,Node>::Build(ghostQMap);
-    ghostQrowNums = Xpetra::VectorFactory<GO,LO,GO,Node>::Build(ghostQMap);
-    if (ghostQvalues->getLocalLength() > 0) {
-      ghostQvals.resize(NSDim);
-      ghostQcols.resize(NSDim);
-      for (size_t i=0; i<NSDim; ++i) {
-        ghostQvals[i] = ghostQvalues->getDataNonConst(i);
-        ghostQcols[i] = ghostQcolumns[i]->getDataNonConst(0);
-      }
-      ghostQrows = ghostQrowNums->getDataNonConst(0);
+  }
+  ghostQMap = MapFactory::Build(A->getRowMap()->lib(),
+                                Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                ghostGIDs,
+                                indexBase, A->getRowMap()->getComm());  // JG:Xpetra::global_size_t>?
+  // Vector to hold bits of Q that go to other processors.
+  ghostQvalues = MultiVectorFactory::Build(ghostQMap, NSDim);
+  // Note that Epetra does not support MultiVectors templated on Scalar != double.
+  // So to work around this, we allocate an array of Vectors.  This shouldn't be too
+  // expensive, as the number of Vectors is NSDim.
+  ghostQcolumns.resize(NSDim);
+  for (size_t i = 0; i < NSDim; ++i)
+    ghostQcolumns[i] = Xpetra::VectorFactory<GO, LO, GO, Node>::Build(ghostQMap);
+  ghostQrowNums = Xpetra::VectorFactory<GO, LO, GO, Node>::Build(ghostQMap);
+  if (ghostQvalues->getLocalLength() > 0) {
+    ghostQvals.resize(NSDim);
+    ghostQcols.resize(NSDim);
+    for (size_t i = 0; i < NSDim; ++i) {
+      ghostQvals[i] = ghostQvalues->getDataNonConst(i);
+      ghostQcols[i] = ghostQcolumns[i]->getDataNonConst(0);
     }
+    ghostQrows = ghostQrowNums->getDataNonConst(0);
+  }
 
-    //importer to handle moving Q
-    importer = ImportFactory::Build(ghostQMap, A->getRowMap());
-
-    // Dense QR solver
-    Teuchos::SerialQRDenseSolver<LO,SC> qrSolver;
-
-    //Allocate temporary storage for the tentative prolongator.
-    Array<GO> globalColPtr(maxAggSize*NSDim,0);
-    Array<LO> localColPtr(maxAggSize*NSDim,0);
-    Array<SC> valPtr(maxAggSize*NSDim,0.);
-
-    //Create column map for Ptent, estimate local #nonzeros in Ptent,  and create Ptent itself.
-    const Map& coarseMapRef = *coarseMap;
-
-    // For the 3-arrays constructor
-    ArrayRCP<size_t>  ptent_rowptr;
-    ArrayRCP<LO>      ptent_colind;
-    ArrayRCP<Scalar>  ptent_values;
-
-    // Because ArrayRCPs are slow...
-    ArrayView<size_t> rowptr_v;
-    ArrayView<LO>     colind_v;
-    ArrayView<Scalar> values_v;
-
-    // For temporary usage
-    Array<size_t>    rowptr_temp;
-    Array<LO>        colind_temp;
-    Array<Scalar>    values_temp;
+  // importer to handle moving Q
+  importer = ImportFactory::Build(ghostQMap, A->getRowMap());
+
+  // Dense QR solver
+  Teuchos::SerialQRDenseSolver<LO, SC> qrSolver;
+
+  // Allocate temporary storage for the tentative prolongator.
+  Array<GO> globalColPtr(maxAggSize * NSDim, 0);
+  Array<LO> localColPtr(maxAggSize * NSDim, 0);
+  Array<SC> valPtr(maxAggSize * NSDim, 0.);
+
+  // Create column map for Ptent, estimate local #nonzeros in Ptent,  and create Ptent itself.
+  const Map& coarseMapRef = *coarseMap;
+
+  // For the 3-arrays constructor
+  ArrayRCP<size_t> ptent_rowptr;
+  ArrayRCP<LO> ptent_colind;
+  ArrayRCP<Scalar> ptent_values;
+
+  // Because ArrayRCPs are slow...
+  ArrayView<size_t> rowptr_v;
+  ArrayView<LO> colind_v;
+  ArrayView<Scalar> values_v;
+
+  // For temporary usage
+  Array<size_t> rowptr_temp;
+  Array<LO> colind_temp;
+  Array<Scalar> values_temp;
+
+  RCP<CrsMatrix> PtentCrs;
+
+  RCP<CrsMatrixWrap> PtentCrsWrap = rcp(new CrsMatrixWrap(rowMapForPtent, NSDim));
+  PtentCrs                        = PtentCrsWrap->getCrsMatrix();
+  Ptentative                      = PtentCrsWrap;
+
+  //*****************************************************************
+  // Loop over all aggregates and calculate local QR decompositions.
+  //*****************************************************************
+  GO qctr                    = 0;  // for indexing into Ptent data vectors
+  const Map& nonUniqueMapRef = *nonUniqueMap;
+
+  size_t total_nnz_count = 0;
+
+  for (GO agg = 0; agg < numAggs; ++agg) {
+    LO myAggSize = aggStart[agg + 1] - aggStart[agg];
+    // For each aggregate, extract the corresponding piece of the nullspace and put it in the flat array,
+    // "localQR" (in column major format) for the QR routine.
+    Teuchos::SerialDenseMatrix<LO, SC> localQR(myAggSize, NSDim);
+    for (size_t j = 0; j < NSDim; ++j) {
+      bool bIsZeroNSColumn = true;
+      for (LO k = 0; k < myAggSize; ++k) {
+        // aggToRowMap[aggPtr[i]+k] is the kth DOF in the ith aggregate
+        // fineNS[j][n] is the nth entry in the jth NS vector
+        try {
+          SC nsVal      = fineNS[j][nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg] + k])];  // extract information from fine level NS
+          localQR(k, j) = nsVal;
+          if (nsVal != zero) bIsZeroNSColumn = false;
+        } catch (...) {
+          GetOStream(Runtime1, -1) << "length of fine level nsp: " << fineNullspace->getGlobalLength() << std::endl;
+          GetOStream(Runtime1, -1) << "length of fine level nsp w overlap: " << fineNullspaceWithOverlap->getGlobalLength() << std::endl;
+          GetOStream(Runtime1, -1) << "(local?) aggId=" << agg << std::endl;
+          GetOStream(Runtime1, -1) << "aggSize=" << myAggSize << std::endl;
+          GetOStream(Runtime1, -1) << "agg DOF=" << k << std::endl;
+          GetOStream(Runtime1, -1) << "NS vector j=" << j << std::endl;
+          GetOStream(Runtime1, -1) << "j*myAggSize + k = " << j * myAggSize + k << std::endl;
+          GetOStream(Runtime1, -1) << "aggToRowMap[" << agg << "][" << k << "] = " << aggToRowMap[aggStart[agg] + k] << std::endl;
+          GetOStream(Runtime1, -1) << "id aggToRowMap[agg][k]=" << aggToRowMap[aggStart[agg] + k] << " is global element in nonUniqueMap = " << nonUniqueMapRef.isNodeGlobalElement(aggToRowMap[aggStart[agg] + k]) << std::endl;
+          GetOStream(Runtime1, -1) << "colMap local id aggToRowMap[agg][k]=" << nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg] + k]) << std::endl;
+          GetOStream(Runtime1, -1) << "fineNS...=" << fineNS[j][nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg] + k])] << std::endl;
+          GetOStream(Errors, -1) << "caught an error!" << std::endl;
+        }
+      }  // for (LO k=0 ...
+      TEUCHOS_TEST_FOR_EXCEPTION(bIsZeroNSColumn == true, Exceptions::RuntimeError, "MueLu::TentativePFactory::MakeTentative: fine level NS part has a zero column. Error.");
+    }  // for (LO j=0 ...
 
-    RCP<CrsMatrix> PtentCrs;
+    Xpetra::global_size_t offset = agg * NSDim;
 
-    RCP<CrsMatrixWrap> PtentCrsWrap = rcp(new CrsMatrixWrap(rowMapForPtent, NSDim));
-    PtentCrs   = PtentCrsWrap->getCrsMatrix();
-    Ptentative = PtentCrsWrap;
+    if (myAggSize >= Teuchos::as<LocalOrdinal>(NSDim)) {
+      // calculate QR decomposition (standard)
+      // R is stored in localQR (size: myAggSize x NSDim)
 
-    //*****************************************************************
-    //Loop over all aggregates and calculate local QR decompositions.
-    //*****************************************************************
-    GO qctr=0; //for indexing into Ptent data vectors
-    const Map& nonUniqueMapRef = *nonUniqueMap;
+      // Householder multiplier
+      SC tau = localQR(0, 0);
 
-    size_t total_nnz_count=0;
+      if (NSDim == 1) {
+        // Only one nullspace vector, so normalize by hand
+        Magnitude dtemp = 0;
+        for (size_t k = 0; k < Teuchos::as<size_t>(myAggSize); ++k) {
+          Magnitude tmag = STS::magnitude(localQR(k, 0));
+          dtemp += tmag * tmag;
+        }
+        dtemp         = Teuchos::ScalarTraits<Magnitude>::squareroot(dtemp);
+        tau           = localQR(0, 0);
+        localQR(0, 0) = dtemp;
+      } else {
+        qrSolver.setMatrix(Teuchos::rcp(&localQR, false));
+        qrSolver.factor();
+      }
 
-    for (GO agg=0; agg<numAggs; ++agg)
-    {
-      LO myAggSize = aggStart[agg+1]-aggStart[agg];
-      // For each aggregate, extract the corresponding piece of the nullspace and put it in the flat array,
-      // "localQR" (in column major format) for the QR routine.
-      Teuchos::SerialDenseMatrix<LO,SC> localQR(myAggSize, NSDim);
-      for (size_t j=0; j<NSDim; ++j) {
-        bool bIsZeroNSColumn = true;
-        for (LO k=0; k<myAggSize; ++k)
-        {
-          // aggToRowMap[aggPtr[i]+k] is the kth DOF in the ith aggregate
-          // fineNS[j][n] is the nth entry in the jth NS vector
-          try{
-            SC nsVal = fineNS[j][ nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg]+k]) ]; // extract information from fine level NS
-            localQR(k,j) = nsVal;
-            if (nsVal != zero) bIsZeroNSColumn = false;
-          }
-          catch(...) {
-            GetOStream(Runtime1,-1) << "length of fine level nsp: " << fineNullspace->getGlobalLength() << std::endl;
-            GetOStream(Runtime1,-1) << "length of fine level nsp w overlap: " << fineNullspaceWithOverlap->getGlobalLength() << std::endl;
-            GetOStream(Runtime1,-1) << "(local?) aggId=" << agg << std::endl;
-            GetOStream(Runtime1,-1) << "aggSize=" << myAggSize << std::endl;
-            GetOStream(Runtime1,-1) << "agg DOF=" << k << std::endl;
-            GetOStream(Runtime1,-1) << "NS vector j=" << j << std::endl;
-            GetOStream(Runtime1,-1) << "j*myAggSize + k = " << j*myAggSize + k << std::endl;
-            GetOStream(Runtime1,-1) << "aggToRowMap["<<agg<<"][" << k << "] = " << aggToRowMap[aggStart[agg]+k] << std::endl;
-            GetOStream(Runtime1,-1) << "id aggToRowMap[agg][k]=" << aggToRowMap[aggStart[agg]+k] << " is global element in nonUniqueMap = " <<
-                nonUniqueMapRef.isNodeGlobalElement(aggToRowMap[aggStart[agg]+k]) << std::endl;
-            GetOStream(Runtime1,-1) << "colMap local id aggToRowMap[agg][k]=" << nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg]+k]) << std::endl;
-            GetOStream(Runtime1,-1) << "fineNS...=" << fineNS[j][ nonUniqueMapRef.getLocalElement(aggToRowMap[aggStart[agg]+k]) ] << std::endl;
-            GetOStream(Errors,-1) << "caught an error!" << std::endl;
+      // Extract R, the coarse nullspace.  This is stored in upper triangular part of localQR.
+      // Note:  coarseNS[i][.] is the ith coarse nullspace vector, which may be counter to your intuition.
+      // This stores the (offset+k)th entry only if it is local according to the coarseMap.
+      for (size_t j = 0; j < NSDim; ++j) {
+        for (size_t k = 0; k <= j; ++k) {
+          try {
+            if (coarseMapRef.isNodeLocalElement(offset + k)) {
+              coarseNS[j][offset + k] = localQR(k, j);  // TODO is offset+k the correct local ID?!
+            }
+          } catch (...) {
+            GetOStream(Errors, -1) << "caught error in coarseNS insert, j=" << j << ", offset+k = " << offset + k << std::endl;
           }
-        } //for (LO k=0 ...
-        TEUCHOS_TEST_FOR_EXCEPTION(bIsZeroNSColumn == true, Exceptions::RuntimeError, "MueLu::TentativePFactory::MakeTentative: fine level NS part has a zero column. Error.");
-      } //for (LO j=0 ...
-
-      Xpetra::global_size_t offset=agg*NSDim;
-
-      if(myAggSize >= Teuchos::as<LocalOrdinal>(NSDim)) {
-        // calculate QR decomposition (standard)
-        // R is stored in localQR (size: myAggSize x NSDim)
+        }
+      }
 
-        // Householder multiplier
-        SC tau = localQR(0,0);
+      // Calculate Q, the tentative prolongator.
+      // The Lapack GEQRF call only works for myAggsize >= NSDim
 
-        if (NSDim == 1) {
-          // Only one nullspace vector, so normalize by hand
-          Magnitude dtemp=0;
-          for (size_t k = 0; k < Teuchos::as<size_t>(myAggSize); ++k) {
-              Magnitude tmag = STS::magnitude(localQR(k,0));
-            dtemp += tmag*tmag;
+      if (NSDim == 1) {
+        // Only one nullspace vector, so calculate Q by hand
+        Magnitude dtemp = Teuchos::ScalarTraits<SC>::magnitude(localQR(0, 0));
+        localQR(0, 0)   = tau;
+        dtemp           = 1 / dtemp;
+        for (LocalOrdinal i = 0; i < myAggSize; ++i) {
+          localQR(i, 0) *= dtemp;
+        }
+      } else {
+        qrSolver.formQ();
+        Teuchos::RCP<Teuchos::SerialDenseMatrix<LO, SC> > qFactor = qrSolver.getQ();
+        for (size_t j = 0; j < NSDim; j++) {
+          for (size_t i = 0; i < Teuchos::as<size_t>(myAggSize); i++) {
+            localQR(i, j) = (*qFactor)(i, j);
           }
-          dtemp = Teuchos::ScalarTraits<Magnitude>::squareroot(dtemp);
-          tau = localQR(0,0);
-          localQR(0,0) = dtemp;
-        } else {
-          qrSolver.setMatrix( Teuchos::rcp(&localQR, false) );
-          qrSolver.factor();
         }
+      }
 
-         // Extract R, the coarse nullspace.  This is stored in upper triangular part of localQR.
-         // Note:  coarseNS[i][.] is the ith coarse nullspace vector, which may be counter to your intuition.
-         // This stores the (offset+k)th entry only if it is local according to the coarseMap.
-         for (size_t j=0; j<NSDim; ++j) {
-           for (size_t k=0; k<=j; ++k) {
-             try {
-               if (coarseMapRef.isNodeLocalElement(offset+k)) {
-                 coarseNS[j][offset+k] = localQR(k, j); //TODO is offset+k the correct local ID?!
-               }
-             }
-             catch(...) {
-               GetOStream(Errors,-1) << "caught error in coarseNS insert, j="<<j<<", offset+k = "<<offset+k<<std::endl;
-             }
-           }
-         }
-
-         // Calculate Q, the tentative prolongator.
-         // The Lapack GEQRF call only works for myAggsize >= NSDim
-
-         if (NSDim == 1) {
-           // Only one nullspace vector, so calculate Q by hand
-           Magnitude dtemp = Teuchos::ScalarTraits<SC>::magnitude(localQR(0,0));
-           localQR(0,0) = tau;
-           dtemp = 1 / dtemp;
-           for (LocalOrdinal i=0; i<myAggSize; ++i) {
-             localQR(i,0) *= dtemp ;
-           }
-         } else {
-           qrSolver.formQ();
-           Teuchos::RCP<Teuchos::SerialDenseMatrix<LO,SC> > qFactor = qrSolver.getQ();
-           for (size_t j=0; j<NSDim; j++) {
-             for (size_t i = 0; i < Teuchos::as<size_t>(myAggSize); i++) {
-               localQR(i,j) = (*qFactor)(i,j);
-             }
-           }
-         }
-
-         // end default case (myAggSize >= NSDim)
-      } else {  // special handling for myAggSize < NSDim (i.e. 1pt nodes)
-        // See comments for the uncoupled case
-
-        // R = extended (by adding identity rows) localQR
-        for (size_t j = 0; j < NSDim; j++)
-          for (size_t k = 0; k < NSDim; k++) {
-            TEUCHOS_TEST_FOR_EXCEPTION(!coarseMapRef.isNodeLocalElement(offset+k), Exceptions::RuntimeError,
-                                       "Caught error in coarseNS insert, j=" << j << ", offset+k = " << offset+k);
+      // end default case (myAggSize >= NSDim)
+    } else {  // special handling for myAggSize < NSDim (i.e. 1pt nodes)
+      // See comments for the uncoupled case
 
-            if (k < as<size_t>(myAggSize))
-              coarseNS[j][offset+k] = localQR(k,j);
-            else
-              coarseNS[j][offset+k] = (k == j ? one : zero);
-          }
+      // R = extended (by adding identity rows) localQR
+      for (size_t j = 0; j < NSDim; j++)
+        for (size_t k = 0; k < NSDim; k++) {
+          TEUCHOS_TEST_FOR_EXCEPTION(!coarseMapRef.isNodeLocalElement(offset + k), Exceptions::RuntimeError,
+                                     "Caught error in coarseNS insert, j=" << j << ", offset+k = " << offset + k);
+
+          if (k < as<size_t>(myAggSize))
+            coarseNS[j][offset + k] = localQR(k, j);
+          else
+            coarseNS[j][offset + k] = (k == j ? one : zero);
+        }
 
-        // Q = I (rectangular)
-        for (size_t i = 0; i < as<size_t>(myAggSize); i++)
-          for (size_t j = 0; j < NSDim; j++)
-            localQR(i,j) = (j == i ? one : zero);
-      } // end else (special handling for 1pt aggregates)
-
-      //Process each row in the local Q factor.  If the row is local to the current processor
-      //according to the rowmap, insert it into Ptentative.  Otherwise, save it in ghostQ
-      //to be communicated later to the owning processor.
-      //FIXME -- what happens if maps are blocked?
-      for (GO j=0; j<myAggSize; ++j) {
-        //This loop checks whether row associated with current DOF is local, according to rowMapForPtent.
-        //If it is, the row is inserted.  If not, the row number, columns, and values are saved in
-        //MultiVectors that will be sent to other processors.
-        GO globalRow = aggToRowMap[aggStart[agg]+j];
-
-        //TODO is the use of Xpetra::global_size_t below correct?
-        if (rowMapForPtentRef.isNodeGlobalElement(globalRow) == false ) {
-          ghostQrows[qctr] = globalRow;
-          for (size_t k=0; k<NSDim; ++k) {
-            ghostQcols[k][qctr] = coarseMapRef.getGlobalElement(agg*NSDim+k);
-            ghostQvals[k][qctr] = localQR(j,k);
-          }
-          ++qctr;
-        } else {
-          size_t nnz=0;
-          for (size_t k=0; k<NSDim; ++k) {
-            try{
-              if (localQR(j,k) != Teuchos::ScalarTraits<SC>::zero()) {
-                localColPtr[nnz]  = agg * NSDim + k;
-                globalColPtr[nnz] = coarseMapRef.getGlobalElement(localColPtr[nnz]);
-                valPtr[nnz] = localQR(j,k);
-                ++total_nnz_count;
-                ++nnz;
-              }
-            }
-            catch(...) {
-              GetOStream(Errors,-1) << "caught error in colPtr/valPtr insert, current index="<<nnz<<std::endl;
+      // Q = I (rectangular)
+      for (size_t i = 0; i < as<size_t>(myAggSize); i++)
+        for (size_t j = 0; j < NSDim; j++)
+          localQR(i, j) = (j == i ? one : zero);
+    }  // end else (special handling for 1pt aggregates)
+
+    // Process each row in the local Q factor.  If the row is local to the current processor
+    // according to the rowmap, insert it into Ptentative.  Otherwise, save it in ghostQ
+    // to be communicated later to the owning processor.
+    // FIXME -- what happens if maps are blocked?
+    for (GO j = 0; j < myAggSize; ++j) {
+      // This loop checks whether row associated with current DOF is local, according to rowMapForPtent.
+      // If it is, the row is inserted.  If not, the row number, columns, and values are saved in
+      // MultiVectors that will be sent to other processors.
+      GO globalRow = aggToRowMap[aggStart[agg] + j];
+
+      // TODO is the use of Xpetra::global_size_t below correct?
+      if (rowMapForPtentRef.isNodeGlobalElement(globalRow) == false) {
+        ghostQrows[qctr] = globalRow;
+        for (size_t k = 0; k < NSDim; ++k) {
+          ghostQcols[k][qctr] = coarseMapRef.getGlobalElement(agg * NSDim + k);
+          ghostQvals[k][qctr] = localQR(j, k);
+        }
+        ++qctr;
+      } else {
+        size_t nnz = 0;
+        for (size_t k = 0; k < NSDim; ++k) {
+          try {
+            if (localQR(j, k) != Teuchos::ScalarTraits<SC>::zero()) {
+              localColPtr[nnz]  = agg * NSDim + k;
+              globalColPtr[nnz] = coarseMapRef.getGlobalElement(localColPtr[nnz]);
+              valPtr[nnz]       = localQR(j, k);
+              ++total_nnz_count;
+              ++nnz;
             }
-          } //for (size_t k=0; k<NSDim; ++k)
-
-          try{
-              Ptentative->insertGlobalValues(globalRow,globalColPtr.view(0,nnz),valPtr.view(0,nnz));
-          }
-          catch(...) {
-            GetOStream(Errors,-1) << "pid " << A->getRowMap()->getComm()->getRank()
-                << "caught error during Ptent row insertion, global row "
-                << globalRow << std::endl;
+          } catch (...) {
+            GetOStream(Errors, -1) << "caught error in colPtr/valPtr insert, current index=" << nnz << std::endl;
           }
+        }  // for (size_t k=0; k<NSDim; ++k)
+
+        try {
+          Ptentative->insertGlobalValues(globalRow, globalColPtr.view(0, nnz), valPtr.view(0, nnz));
+        } catch (...) {
+          GetOStream(Errors, -1) << "pid " << A->getRowMap()->getComm()->getRank()
+                                 << "caught error during Ptent row insertion, global row "
+                                 << globalRow << std::endl;
         }
-      } //for (GO j=0; j<myAggSize; ++j)
-
-    } // for (LO agg=0; agg<numAggs; ++agg)
-
-
-    // ***********************************************************
-    // ************* end of aggregate-wise QR ********************
-    // ***********************************************************
-    GetOStream(Runtime1) << "TentativePFactory : aggregates may cross process boundaries" << std::endl;
-    // Import ghost parts of Q factors and insert into Ptentative.
-    // First import just the global row numbers.
-    RCP<Xpetra::Vector<GO,LO,GO,Node> > targetQrowNums = Xpetra::VectorFactory<GO,LO,GO,Node>::Build(rowMapForPtent);
-    targetQrowNums->putScalar(-1);
-    targetQrowNums->doImport(*ghostQrowNums,*importer,Xpetra::INSERT);
-    ArrayRCP< GO > targetQrows = targetQrowNums->getDataNonConst(0);
-
-    // Now create map based on just the row numbers imported.
-    Array<GO> gidsToImport;
-    gidsToImport.reserve(targetQrows.size());
-    for (typename ArrayRCP<GO>::iterator r=targetQrows.begin(); r!=targetQrows.end(); ++r) {
-      if (*r > -1) {
-        gidsToImport.push_back(*r);
       }
+    }  // for (GO j=0; j<myAggSize; ++j)
+
+  }  // for (LO agg=0; agg<numAggs; ++agg)
+
+  // ***********************************************************
+  // ************* end of aggregate-wise QR ********************
+  // ***********************************************************
+  GetOStream(Runtime1) << "TentativePFactory : aggregates may cross process boundaries" << std::endl;
+  // Import ghost parts of Q factors and insert into Ptentative.
+  // First import just the global row numbers.
+  RCP<Xpetra::Vector<GO, LO, GO, Node> > targetQrowNums = Xpetra::VectorFactory<GO, LO, GO, Node>::Build(rowMapForPtent);
+  targetQrowNums->putScalar(-1);
+  targetQrowNums->doImport(*ghostQrowNums, *importer, Xpetra::INSERT);
+  ArrayRCP<GO> targetQrows = targetQrowNums->getDataNonConst(0);
+
+  // Now create map based on just the row numbers imported.
+  Array<GO> gidsToImport;
+  gidsToImport.reserve(targetQrows.size());
+  for (typename ArrayRCP<GO>::iterator r = targetQrows.begin(); r != targetQrows.end(); ++r) {
+    if (*r > -1) {
+      gidsToImport.push_back(*r);
     }
-    RCP<const Map > reducedMap = MapFactory::Build( A->getRowMap()->lib(),
-                                                    Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                                    gidsToImport, indexBase, A->getRowMap()->getComm()    );
+  }
+  RCP<const Map> reducedMap = MapFactory::Build(A->getRowMap()->lib(),
+                                                Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                gidsToImport, indexBase, A->getRowMap()->getComm());
 
-    // Import using the row numbers that this processor will receive.
-    importer = ImportFactory::Build(ghostQMap, reducedMap);
+  // Import using the row numbers that this processor will receive.
+  importer = ImportFactory::Build(ghostQMap, reducedMap);
 
-    Array<RCP<Xpetra::Vector<GO,LO,GO,Node> > > targetQcolumns(NSDim);
-    for (size_t i=0; i<NSDim; ++i) {
-      targetQcolumns[i] = Xpetra::VectorFactory<GO,LO,GO,Node>::Build(reducedMap);
-      targetQcolumns[i]->doImport(*(ghostQcolumns[i]),*importer,Xpetra::INSERT);
+  Array<RCP<Xpetra::Vector<GO, LO, GO, Node> > > targetQcolumns(NSDim);
+  for (size_t i = 0; i < NSDim; ++i) {
+    targetQcolumns[i] = Xpetra::VectorFactory<GO, LO, GO, Node>::Build(reducedMap);
+    targetQcolumns[i]->doImport(*(ghostQcolumns[i]), *importer, Xpetra::INSERT);
+  }
+  RCP<MultiVector> targetQvalues = MultiVectorFactory::Build(reducedMap, NSDim);
+  targetQvalues->doImport(*ghostQvalues, *importer, Xpetra::INSERT);
+
+  ArrayRCP<ArrayRCP<SC> > targetQvals;
+  ArrayRCP<ArrayRCP<GO> > targetQcols;
+  if (targetQvalues->getLocalLength() > 0) {
+    targetQvals.resize(NSDim);
+    targetQcols.resize(NSDim);
+    for (size_t i = 0; i < NSDim; ++i) {
+      targetQvals[i] = targetQvalues->getDataNonConst(i);
+      targetQcols[i] = targetQcolumns[i]->getDataNonConst(0);
     }
-    RCP<MultiVector> targetQvalues = MultiVectorFactory::Build(reducedMap,NSDim);
-    targetQvalues->doImport(*ghostQvalues,*importer,Xpetra::INSERT);
+  }
 
-    ArrayRCP< ArrayRCP<SC> > targetQvals;
-    ArrayRCP<ArrayRCP<GO> > targetQcols;
+  valPtr       = Array<SC>(NSDim, 0.);
+  globalColPtr = Array<GO>(NSDim, 0);
+  for (typename Array<GO>::iterator r = gidsToImport.begin(); r != gidsToImport.end(); ++r) {
     if (targetQvalues->getLocalLength() > 0) {
-      targetQvals.resize(NSDim);
-      targetQcols.resize(NSDim);
-      for (size_t i=0; i<NSDim; ++i) {
-        targetQvals[i] = targetQvalues->getDataNonConst(i);
-        targetQcols[i] = targetQcolumns[i]->getDataNonConst(0);
+      for (size_t j = 0; j < NSDim; ++j) {
+        valPtr[j]       = targetQvals[j][reducedMap->getLocalElement(*r)];
+        globalColPtr[j] = targetQcols[j][reducedMap->getLocalElement(*r)];
       }
-    }
-
-    valPtr = Array<SC>(NSDim,0.);
-    globalColPtr = Array<GO>(NSDim,0);
-    for (typename Array<GO>::iterator r=gidsToImport.begin(); r!=gidsToImport.end(); ++r) {
-      if (targetQvalues->getLocalLength() > 0) {
-        for (size_t j=0; j<NSDim; ++j) {
-          valPtr[j] = targetQvals[j][reducedMap->getLocalElement(*r)];
-          globalColPtr[j] = targetQcols[j][reducedMap->getLocalElement(*r)];
-        }
-        Ptentative->insertGlobalValues(*r, globalColPtr.view(0,NSDim), valPtr.view(0,NSDim));
-      } //if (targetQvalues->getLocalLength() > 0)
-    }
-
-    Ptentative->fillComplete(coarseMap, A->getDomainMap());
+      Ptentative->insertGlobalValues(*r, globalColPtr.view(0, NSDim), valPtr.view(0, NSDim));
+    }  // if (targetQvalues->getLocalLength() > 0)
   }
 
+  Ptentative->fillComplete(coarseMap, A->getDomainMap());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildPuncoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                    RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const {
+  RCP<const Map> rowMap = A->getRowMap();
+  RCP<const Map> colMap = A->getColMap();
+  const size_t numRows  = rowMap->getLocalNumElements();
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType Magnitude;
+  const SC zero    = STS::zero();
+  const SC one     = STS::one();
+  const LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  const GO numAggs      = aggregates->GetNumAggregates();
+  const size_t NSDim    = fineNullspace->getNumVectors();
+  ArrayRCP<LO> aggSizes = aggregates->ComputeAggregateSizesArrayRCP();
+
+  // Sanity checking
+  const ParameterList& pL     = GetParameterList();
+  const bool& doQRStep        = pL.get<bool>("tentative: calculate qr");
+  const bool& constantColSums = pL.get<bool>("tentative: constant column sums");
+
+  TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && constantColSums, Exceptions::RuntimeError,
+                             "MueLu::TentativePFactory::MakeTentative: cannot use 'constant column sums' and 'calculate qr' at the same time");
+
+  // Aggregates map is based on the amalgamated column map
+  // We can skip global-to-local conversion if LIDs in row map are
+  // same as LIDs in column map
+  bool goodMap = MueLu::Utilities<SC, LO, GO, NO>::MapsAreNested(*rowMap, *colMap);
+
+  // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
+  // aggStart is a pointer into aggToRowMapLO
+  // aggStart[i]..aggStart[i+1] are indices into aggToRowMapLO
+  // aggToRowMapLO[aggStart[i]]..aggToRowMapLO[aggStart[i+1]-1] are the DOFs in aggregate i
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<LO> aggToRowMapLO;
+  ArrayRCP<GO> aggToRowMapGO;
+  if (goodMap) {
+    amalgInfo->UnamalgamateAggregatesLO(*aggregates, aggStart, aggToRowMapLO);
+    GetOStream(Runtime1) << "Column map is consistent with the row map, good." << std::endl;
+
+  } else {
+    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMapGO);
+    GetOStream(Warnings0) << "Column map is not consistent with the row map\n"
+                          << "using GO->LO conversion with performance penalty" << std::endl;
+  }
+  coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
+
+  // Pull out the nullspace vectors so that we can have random access.
+  ArrayRCP<ArrayRCP<const SC> > fineNS(NSDim);
+  ArrayRCP<ArrayRCP<SC> > coarseNS(NSDim);
+  for (size_t i = 0; i < NSDim; i++) {
+    fineNS[i] = fineNullspace->getData(i);
+    if (coarseMap->getLocalNumElements() > 0)
+      coarseNS[i] = coarseNullspace->getDataNonConst(i);
+  }
 
+  size_t nnzEstimate = numRows * NSDim;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void TentativePFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildPuncoupled(RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const {
-    RCP<const Map> rowMap = A->getRowMap();
-    RCP<const Map> colMap = A->getColMap();
-    const size_t numRows   = rowMap->getLocalNumElements();
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType Magnitude;
-    const SC     zero      = STS::zero();
-    const SC     one       = STS::one();
-    const LO     INVALID   = Teuchos::OrdinalTraits<LO>::invalid();
-
-    const GO     numAggs   = aggregates->GetNumAggregates();
-    const size_t NSDim     = fineNullspace->getNumVectors();
-    ArrayRCP<LO> aggSizes  = aggregates->ComputeAggregateSizesArrayRCP();
-
+  // Time to construct the matrix and fill in the values
+  Ptentative              = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
+  RCP<CrsMatrix> PtentCrs = rcp_dynamic_cast<CrsMatrixWrap>(Ptentative)->getCrsMatrix();
 
-    // Sanity checking
-    const ParameterList& pL = GetParameterList();
-    const bool &doQRStep = pL.get<bool>("tentative: calculate qr");
-    const bool &constantColSums = pL.get<bool>("tentative: constant column sums");
+  ArrayRCP<size_t> iaPtent;
+  ArrayRCP<LO> jaPtent;
+  ArrayRCP<SC> valPtent;
 
-    TEUCHOS_TEST_FOR_EXCEPTION(doQRStep && constantColSums,Exceptions::RuntimeError,
-                               "MueLu::TentativePFactory::MakeTentative: cannot use 'constant column sums' and 'calculate qr' at the same time");
+  PtentCrs->allocateAllValues(nnzEstimate, iaPtent, jaPtent, valPtent);
 
-    // Aggregates map is based on the amalgamated column map
-    // We can skip global-to-local conversion if LIDs in row map are
-    // same as LIDs in column map
-    bool goodMap = MueLu::Utilities<SC,LO,GO,NO>::MapsAreNested(*rowMap, *colMap);
+  ArrayView<size_t> ia = iaPtent();
+  ArrayView<LO> ja     = jaPtent();
+  ArrayView<SC> val    = valPtent();
 
-    // Create a lookup table to determine the rows (fine DOFs) that belong to a given aggregate.
-    // aggStart is a pointer into aggToRowMapLO
-    // aggStart[i]..aggStart[i+1] are indices into aggToRowMapLO
-    // aggToRowMapLO[aggStart[i]]..aggToRowMapLO[aggStart[i+1]-1] are the DOFs in aggregate i
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<LO> aggToRowMapLO;
-    ArrayRCP<GO> aggToRowMapGO;
-    if (goodMap) {
-      amalgInfo->UnamalgamateAggregatesLO(*aggregates, aggStart, aggToRowMapLO);
-      GetOStream(Runtime1) << "Column map is consistent with the row map, good." << std::endl;
+  ia[0] = 0;
+  for (size_t i = 1; i <= numRows; i++)
+    ia[i] = ia[i - 1] + NSDim;
 
-    } else {
-      amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMapGO);
-      GetOStream(Warnings0) << "Column map is not consistent with the row map\n"
-                            << "using GO->LO conversion with performance penalty" << std::endl;
-    }
-    coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
-
-    // Pull out the nullspace vectors so that we can have random access.
-    ArrayRCP<ArrayRCP<const SC> > fineNS  (NSDim);
-    ArrayRCP<ArrayRCP<SC> >       coarseNS(NSDim);
-    for (size_t i = 0; i < NSDim; i++) {
-      fineNS[i] = fineNullspace->getData(i);
-      if (coarseMap->getLocalNumElements() > 0)
-        coarseNS[i] = coarseNullspace->getDataNonConst(i);
-    }
+  for (size_t j = 0; j < nnzEstimate; j++) {
+    ja[j]  = INVALID;
+    val[j] = zero;
+  }
 
-    size_t nnzEstimate = numRows * NSDim;
+  if (doQRStep) {
+    ////////////////////////////////
+    // Standard aggregate-wise QR //
+    ////////////////////////////////
+    for (GO agg = 0; agg < numAggs; agg++) {
+      LO aggSize = aggStart[agg + 1] - aggStart[agg];
 
-    // Time to construct the matrix and fill in the values
-    Ptentative = rcp(new CrsMatrixWrap(rowMap, coarseMap, 0));
-    RCP<CrsMatrix> PtentCrs   = rcp_dynamic_cast<CrsMatrixWrap>(Ptentative)->getCrsMatrix();
+      Xpetra::global_size_t offset = agg * NSDim;
 
-    ArrayRCP<size_t>  iaPtent;
-    ArrayRCP<LO>      jaPtent;
-    ArrayRCP<SC>     valPtent;
+      // Extract the piece of the nullspace corresponding to the aggregate, and
+      // put it in the flat array, "localQR" (in column major format) for the
+      // QR routine.
+      Teuchos::SerialDenseMatrix<LO, SC> localQR(aggSize, NSDim);
+      if (goodMap) {
+        for (size_t j = 0; j < NSDim; j++)
+          for (LO k = 0; k < aggSize; k++)
+            localQR(k, j) = fineNS[j][aggToRowMapLO[aggStart[agg] + k]];
+      } else {
+        for (size_t j = 0; j < NSDim; j++)
+          for (LO k = 0; k < aggSize; k++)
+            localQR(k, j) = fineNS[j][rowMap->getLocalElement(aggToRowMapGO[aggStart[agg] + k])];
+      }
 
-    PtentCrs->allocateAllValues(nnzEstimate, iaPtent, jaPtent, valPtent);
+      // Test for zero columns
+      for (size_t j = 0; j < NSDim; j++) {
+        bool bIsZeroNSColumn = true;
 
-    ArrayView<size_t> ia  = iaPtent();
-    ArrayView<LO>     ja  = jaPtent();
-    ArrayView<SC>     val = valPtent();
+        for (LO k = 0; k < aggSize; k++)
+          if (localQR(k, j) != zero)
+            bIsZeroNSColumn = false;
 
-    ia[0] = 0;
-    for (size_t i = 1; i <= numRows; i++)
-      ia[i] = ia[i-1] + NSDim;
+        TEUCHOS_TEST_FOR_EXCEPTION(bIsZeroNSColumn == true, Exceptions::RuntimeError,
+                                   "MueLu::TentativePFactory::MakeTentative: fine level NS part has a zero column in NS column " << j);
+      }
 
-    for (size_t j = 0; j < nnzEstimate; j++) {
-      ja [j] = INVALID;
-      val[j] = zero;
-    }
+      // Calculate QR decomposition (standard)
+      // NOTE: Q is stored in localQR and R is stored in coarseNS
+      if (aggSize >= Teuchos::as<LO>(NSDim)) {
+        if (NSDim == 1) {
+          // Only one nullspace vector, calculate Q and R by hand
+          Magnitude norm = STS::magnitude(zero);
+          for (size_t k = 0; k < Teuchos::as<size_t>(aggSize); k++)
+            norm += STS::magnitude(localQR(k, 0) * localQR(k, 0));
+          norm = Teuchos::ScalarTraits<Magnitude>::squareroot(norm);
 
+          // R = norm
+          coarseNS[0][offset] = norm;
 
-    if (doQRStep) {
-      ////////////////////////////////
-      // Standard aggregate-wise QR //
-      ////////////////////////////////
-      for (GO agg = 0; agg < numAggs; agg++) {
-        LO aggSize = aggStart[agg+1] - aggStart[agg];
+          // Q = localQR(:,0)/norm
+          for (LO i = 0; i < aggSize; i++)
+            localQR(i, 0) /= norm;
 
-        Xpetra::global_size_t offset = agg*NSDim;
+        } else {
+          Teuchos::SerialQRDenseSolver<LO, SC> qrSolver;
+          qrSolver.setMatrix(Teuchos::rcp(&localQR, false));
+          qrSolver.factor();
 
-        // Extract the piece of the nullspace corresponding to the aggregate, and
-        // put it in the flat array, "localQR" (in column major format) for the
-        // QR routine.
-        Teuchos::SerialDenseMatrix<LO,SC> localQR(aggSize, NSDim);
-        if (goodMap) {
+          // R = upper triangular part of localQR
           for (size_t j = 0; j < NSDim; j++)
-            for (LO k = 0; k < aggSize; k++)
-              localQR(k,j) = fineNS[j][aggToRowMapLO[aggStart[agg]+k]];
-        } else {
+            for (size_t k = 0; k <= j; k++)
+              coarseNS[j][offset + k] = localQR(k, j);  // TODO is offset+k the correct local ID?!
+
+          // Calculate Q, the tentative prolongator.
+          // The Lapack GEQRF call only works for myAggsize >= NSDim
+          qrSolver.formQ();
+          Teuchos::RCP<Teuchos::SerialDenseMatrix<LO, SC> > qFactor = qrSolver.getQ();
           for (size_t j = 0; j < NSDim; j++)
-            for (LO k = 0; k < aggSize; k++)
-              localQR(k,j) = fineNS[j][rowMap->getLocalElement(aggToRowMapGO[aggStart[agg]+k])];
+            for (size_t i = 0; i < Teuchos::as<size_t>(aggSize); i++)
+              localQR(i, j) = (*qFactor)(i, j);
         }
 
-        // Test for zero columns
-        for (size_t j = 0; j < NSDim; j++) {
-          bool bIsZeroNSColumn = true;
+      } else {
+        // Special handling for aggSize < NSDim (i.e. single node aggregates in structural mechanics)
+
+        // The local QR decomposition is not possible in the "overconstrained"
+        // case (i.e. number of columns in localQR > number of rows), which
+        // corresponds to #DOFs in Aggregate < NSDim. For usual problems this
+        // is only possible for single node aggregates in structural mechanics.
+        // (Similar problems may arise in discontinuous Galerkin problems...)
+        // We bypass the QR decomposition and use an identity block in the
+        // tentative prolongator for the single node aggregate and transfer the
+        // corresponding fine level null space information 1-to-1 to the coarse
+        // level null space part.
+
+        // NOTE: The resulting tentative prolongation operator has
+        // (aggSize*DofsPerNode-NSDim) zero columns leading to a singular
+        // coarse level operator A.  To deal with that one has the following
+        // options:
+        // - Use the "RepairMainDiagonal" flag in the RAPFactory (default:
+        //   false) to add some identity block to the diagonal of the zero rows
+        //   in the coarse level operator A, such that standard level smoothers
+        //   can be used again.
+        // - Use special (projection-based) level smoothers, which can deal
+        //   with singular matrices (very application specific)
+        // - Adapt the code below to avoid zero columns. However, we do not
+        //   support a variable number of DOFs per node in MueLu/Xpetra which
+        //   makes the implementation really hard.
 
-          for (LO k = 0; k < aggSize; k++)
-            if (localQR(k,j) != zero)
-              bIsZeroNSColumn = false;
+        // R = extended (by adding identity rows) localQR
+        for (size_t j = 0; j < NSDim; j++)
+          for (size_t k = 0; k < NSDim; k++)
+            if (k < as<size_t>(aggSize))
+              coarseNS[j][offset + k] = localQR(k, j);
+            else
+              coarseNS[j][offset + k] = (k == j ? one : zero);
 
-          TEUCHOS_TEST_FOR_EXCEPTION(bIsZeroNSColumn == true, Exceptions::RuntimeError,
-                                     "MueLu::TentativePFactory::MakeTentative: fine level NS part has a zero column in NS column " << j);
-        }
+        // Q = I (rectangular)
+        for (size_t i = 0; i < as<size_t>(aggSize); i++)
+          for (size_t j = 0; j < NSDim; j++)
+            localQR(i, j) = (j == i ? one : zero);
+      }
 
-        // Calculate QR decomposition (standard)
-        // NOTE: Q is stored in localQR and R is stored in coarseNS
-        if (aggSize >= Teuchos::as<LO>(NSDim)) {
-
-          if (NSDim == 1) {
-            // Only one nullspace vector, calculate Q and R by hand
-            Magnitude norm = STS::magnitude(zero);
-            for (size_t k = 0; k < Teuchos::as<size_t>(aggSize); k++)
-              norm += STS::magnitude(localQR(k,0)*localQR(k,0));
-            norm = Teuchos::ScalarTraits<Magnitude>::squareroot(norm);
-
-            // R = norm
-            coarseNS[0][offset] = norm;
-
-            // Q = localQR(:,0)/norm
-            for (LO i = 0; i < aggSize; i++)
-              localQR(i,0) /= norm;
-
-          } else {
-            Teuchos::SerialQRDenseSolver<LO,SC> qrSolver;
-            qrSolver.setMatrix(Teuchos::rcp(&localQR, false));
-            qrSolver.factor();
-
-            // R = upper triangular part of localQR
-            for (size_t j = 0; j < NSDim; j++)
-              for (size_t k = 0; k <= j; k++)
-                coarseNS[j][offset+k] = localQR(k,j); //TODO is offset+k the correct local ID?!
-
-            // Calculate Q, the tentative prolongator.
-            // The Lapack GEQRF call only works for myAggsize >= NSDim
-            qrSolver.formQ();
-            Teuchos::RCP<Teuchos::SerialDenseMatrix<LO,SC> > qFactor = qrSolver.getQ();
-            for (size_t j = 0; j < NSDim; j++)
-              for (size_t i = 0; i < Teuchos::as<size_t>(aggSize); i++)
-                localQR(i,j) = (*qFactor)(i,j);
+      // Process each row in the local Q factor
+      // FIXME: What happens if maps are blocked?
+      for (LO j = 0; j < aggSize; j++) {
+        LO localRow = (goodMap ? aggToRowMapLO[aggStart[agg] + j] : rowMap->getLocalElement(aggToRowMapGO[aggStart[agg] + j]));
+
+        size_t rowStart = ia[localRow];
+        for (size_t k = 0, lnnz = 0; k < NSDim; k++) {
+          // Skip zeros (there may be plenty of them, i.e., NSDim > 1 or boundary conditions)
+          if (localQR(j, k) != zero) {
+            ja[rowStart + lnnz]  = offset + k;
+            val[rowStart + lnnz] = localQR(j, k);
+            lnnz++;
           }
-
-        } else {
-          // Special handling for aggSize < NSDim (i.e. single node aggregates in structural mechanics)
-
-          // The local QR decomposition is not possible in the "overconstrained"
-          // case (i.e. number of columns in localQR > number of rows), which
-          // corresponds to #DOFs in Aggregate < NSDim. For usual problems this
-          // is only possible for single node aggregates in structural mechanics.
-          // (Similar problems may arise in discontinuous Galerkin problems...)
-          // We bypass the QR decomposition and use an identity block in the
-          // tentative prolongator for the single node aggregate and transfer the
-          // corresponding fine level null space information 1-to-1 to the coarse
-          // level null space part.
-
-          // NOTE: The resulting tentative prolongation operator has
-          // (aggSize*DofsPerNode-NSDim) zero columns leading to a singular
-          // coarse level operator A.  To deal with that one has the following
-          // options:
-          // - Use the "RepairMainDiagonal" flag in the RAPFactory (default:
-          //   false) to add some identity block to the diagonal of the zero rows
-          //   in the coarse level operator A, such that standard level smoothers
-          //   can be used again.
-          // - Use special (projection-based) level smoothers, which can deal
-          //   with singular matrices (very application specific)
-          // - Adapt the code below to avoid zero columns. However, we do not
-          //   support a variable number of DOFs per node in MueLu/Xpetra which
-          //   makes the implementation really hard.
-
-          // R = extended (by adding identity rows) localQR
-          for (size_t j = 0; j < NSDim; j++)
-            for (size_t k = 0; k < NSDim; k++)
-              if (k < as<size_t>(aggSize))
-                coarseNS[j][offset+k] = localQR(k,j);
-              else
-                coarseNS[j][offset+k] = (k == j ? one : zero);
-
-          // Q = I (rectangular)
-          for (size_t i = 0; i < as<size_t>(aggSize); i++)
-            for (size_t j = 0; j < NSDim; j++)
-              localQR(i,j) = (j == i ? one : zero);
         }
+      }
+    }
 
+  } else {
+    GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
+    if (NSDim > 1)
+      GetOStream(Warnings0) << "TentativePFactory : for nontrivial nullspace, this may degrade performance" << std::endl;
+    /////////////////////////////
+    //      "no-QR" option     //
+    /////////////////////////////
+    // Local Q factor is just the fine nullspace support over the current aggregate.
+    // Local R factor is the identity.
+    // TODO I have not implemented any special handling for aggregates that are too
+    // TODO small to locally support the nullspace, as is done in the standard QR
+    // TODO case above.
+    if (goodMap) {
+      for (GO agg = 0; agg < numAggs; agg++) {
+        const LO aggSize             = aggStart[agg + 1] - aggStart[agg];
+        Xpetra::global_size_t offset = agg * NSDim;
 
         // Process each row in the local Q factor
         // FIXME: What happens if maps are blocked?
         for (LO j = 0; j < aggSize; j++) {
-          LO localRow = (goodMap ? aggToRowMapLO[aggStart[agg]+j] : rowMap->getLocalElement(aggToRowMapGO[aggStart[agg]+j]));
+          // TODO Here I do not check for a zero nullspace column on the aggregate.
+          //      as is done in the standard QR case.
+
+          const LO localRow = aggToRowMapLO[aggStart[agg] + j];
+
+          const size_t rowStart = ia[localRow];
 
-          size_t rowStart = ia[localRow];
           for (size_t k = 0, lnnz = 0; k < NSDim; k++) {
             // Skip zeros (there may be plenty of them, i.e., NSDim > 1 or boundary conditions)
-            if (localQR(j,k) != zero) {
-              ja [rowStart+lnnz] = offset + k;
-              val[rowStart+lnnz] = localQR(j,k);
+            SC qr_jk = fineNS[k][aggToRowMapLO[aggStart[agg] + j]];
+            if (constantColSums) qr_jk = qr_jk / (Magnitude)aggSizes[agg];
+            if (qr_jk != zero) {
+              ja[rowStart + lnnz]  = offset + k;
+              val[rowStart + lnnz] = qr_jk;
               lnnz++;
             }
           }
         }
-      }
+        for (size_t j = 0; j < NSDim; j++)
+          coarseNS[j][offset + j] = one;
+      }  // for (GO agg = 0; agg < numAggs; agg++)
 
     } else {
-      GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
-      if (NSDim>1)
-        GetOStream(Warnings0) << "TentativePFactory : for nontrivial nullspace, this may degrade performance" << std::endl;
-      /////////////////////////////
-      //      "no-QR" option     //
-      /////////////////////////////
-      // Local Q factor is just the fine nullspace support over the current aggregate.
-      // Local R factor is the identity.
-      // TODO I have not implemented any special handling for aggregates that are too
-      // TODO small to locally support the nullspace, as is done in the standard QR
-      // TODO case above.
-      if (goodMap) {
-        for (GO agg = 0; agg < numAggs; agg++) {
-          const LO aggSize = aggStart[agg+1] - aggStart[agg];
-          Xpetra::global_size_t offset = agg*NSDim;
-
-          // Process each row in the local Q factor
-          // FIXME: What happens if maps are blocked?
-          for (LO j = 0; j < aggSize; j++) {
-
-            //TODO Here I do not check for a zero nullspace column on the aggregate.
-            //     as is done in the standard QR case.
-
-            const LO localRow = aggToRowMapLO[aggStart[agg]+j];
-
-            const size_t rowStart = ia[localRow];
-
-            for (size_t k = 0, lnnz = 0; k < NSDim; k++) {
-              // Skip zeros (there may be plenty of them, i.e., NSDim > 1 or boundary conditions)
-              SC qr_jk = fineNS[k][aggToRowMapLO[aggStart[agg]+j]];
-              if(constantColSums) qr_jk = qr_jk / (Magnitude)aggSizes[agg];
-              if (qr_jk != zero) {
-                ja [rowStart+lnnz] = offset + k;
-                val[rowStart+lnnz] = qr_jk;
-                lnnz++;
-              }
-            }
-          }
-          for (size_t j = 0; j < NSDim; j++)
-            coarseNS[j][offset+j] = one;
-        } //for (GO agg = 0; agg < numAggs; agg++)
+      for (GO agg = 0; agg < numAggs; agg++) {
+        const LO aggSize             = aggStart[agg + 1] - aggStart[agg];
+        Xpetra::global_size_t offset = agg * NSDim;
+        for (LO j = 0; j < aggSize; j++) {
+          const LO localRow = rowMap->getLocalElement(aggToRowMapGO[aggStart[agg] + j]);
 
-      } else {
-        for (GO agg = 0; agg < numAggs; agg++) {
-          const LO aggSize = aggStart[agg+1] - aggStart[agg];
-          Xpetra::global_size_t offset = agg*NSDim;
-          for (LO j = 0; j < aggSize; j++) {
-
-            const LO localRow = rowMap->getLocalElement(aggToRowMapGO[aggStart[agg]+j]);
-
-            const size_t rowStart = ia[localRow];
-
-            for (size_t k = 0, lnnz = 0; k < NSDim; ++k) {
-              // Skip zeros (there may be plenty of them, i.e., NSDim > 1 or boundary conditions)
-              SC qr_jk = fineNS[k][rowMap->getLocalElement(aggToRowMapGO[aggStart[agg]+j])];
-              if(constantColSums) qr_jk = qr_jk / (Magnitude)aggSizes[agg];
-              if (qr_jk != zero) {
-                ja [rowStart+lnnz] = offset + k;
-                val[rowStart+lnnz] = qr_jk;
-                lnnz++;
-              }
+          const size_t rowStart = ia[localRow];
+
+          for (size_t k = 0, lnnz = 0; k < NSDim; ++k) {
+            // Skip zeros (there may be plenty of them, i.e., NSDim > 1 or boundary conditions)
+            SC qr_jk = fineNS[k][rowMap->getLocalElement(aggToRowMapGO[aggStart[agg] + j])];
+            if (constantColSums) qr_jk = qr_jk / (Magnitude)aggSizes[agg];
+            if (qr_jk != zero) {
+              ja[rowStart + lnnz]  = offset + k;
+              val[rowStart + lnnz] = qr_jk;
+              lnnz++;
             }
           }
-          for (size_t j = 0; j < NSDim; j++)
-            coarseNS[j][offset+j] = one;
-        } //for (GO agg = 0; agg < numAggs; agg++)
-
-      } //if (goodmap) else ...
-
-    } //if doQRStep ... else
-
-    // Compress storage (remove all INVALID, which happen when we skip zeros)
-    // We do that in-place
-    size_t ia_tmp = 0, nnz = 0;
-    for (size_t i = 0; i < numRows; i++) {
-      for (size_t j = ia_tmp; j < ia[i+1]; j++)
-        if (ja[j] != INVALID) {
-          ja [nnz] = ja [j];
-          val[nnz] = val[j];
-          nnz++;
         }
-      ia_tmp  = ia[i+1];
-      ia[i+1] = nnz;
-    }
-    if (rowMap->lib() == Xpetra::UseTpetra) {
-      // - Cannot resize for Epetra, as it checks for same pointers
-      // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
-      // NOTE: these invalidate ja and val views
-      jaPtent .resize(nnz);
-      valPtent.resize(nnz);
-    }
-
-    GetOStream(Runtime1) << "TentativePFactory : aggregates do not cross process boundaries" << std::endl;
-
-    PtentCrs->setAllValues(iaPtent, jaPtent, valPtent);
-
+        for (size_t j = 0; j < NSDim; j++)
+          coarseNS[j][offset + j] = one;
+      }  // for (GO agg = 0; agg < numAggs; agg++)
+
+    }  // if (goodmap) else ...
+
+  }  // if doQRStep ... else
+
+  // Compress storage (remove all INVALID, which happen when we skip zeros)
+  // We do that in-place
+  size_t ia_tmp = 0, nnz = 0;
+  for (size_t i = 0; i < numRows; i++) {
+    for (size_t j = ia_tmp; j < ia[i + 1]; j++)
+      if (ja[j] != INVALID) {
+        ja[nnz]  = ja[j];
+        val[nnz] = val[j];
+        nnz++;
+      }
+    ia_tmp    = ia[i + 1];
+    ia[i + 1] = nnz;
+  }
+  if (rowMap->lib() == Xpetra::UseTpetra) {
+    // - Cannot resize for Epetra, as it checks for same pointers
+    // - Need to resize for Tpetra, as it check ().size() == ia[numRows]
+    // NOTE: these invalidate ja and val views
+    jaPtent.resize(nnz);
+    valPtent.resize(nnz);
+  }
 
-    // Managing labels & constants for ESFC
-    RCP<ParameterList> FCparams;
-    if(pL.isSublist("matrixmatrix: kernel params"))
-      FCparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-    else
-      FCparams= rcp(new ParameterList);
-    // By default, we don't need global constants for TentativeP
-    FCparams->set("compute global constants",FCparams->get("compute global constants",false));
-    std::string levelIDs = toString(levelID);
-    FCparams->set("Timer Label",std::string("MueLu::TentativeP-")+levelIDs);
-    RCP<const Export> dummy_e;
-    RCP<const Import> dummy_i;
+  GetOStream(Runtime1) << "TentativePFactory : aggregates do not cross process boundaries" << std::endl;
 
-    PtentCrs->expertStaticFillComplete(coarseMap, A->getDomainMap(),dummy_i,dummy_e,FCparams);
-  }
+  PtentCrs->setAllValues(iaPtent, jaPtent, valPtent);
 
+  // Managing labels & constants for ESFC
+  RCP<ParameterList> FCparams;
+  if (pL.isSublist("matrixmatrix: kernel params"))
+    FCparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+  else
+    FCparams = rcp(new ParameterList);
+  // By default, we don't need global constants for TentativeP
+  FCparams->set("compute global constants", FCparams->get("compute global constants", false));
+  std::string levelIDs = toString(levelID);
+  FCparams->set("Timer Label", std::string("MueLu::TentativeP-") + levelIDs);
+  RCP<const Export> dummy_e;
+  RCP<const Import> dummy_i;
 
+  PtentCrs->expertStaticFillComplete(coarseMap, A->getDomainMap(), dummy_i, dummy_e, FCparams);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 // TODO ReUse: If only P or Nullspace is missing, TentativePFactory can be smart and skip part of the computation.
 
 #define MUELU_TENTATIVEPFACTORY_SHORT
-#endif // MUELU_TENTATIVEPFACTORY_DEF_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_decl.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_decl.hpp
index 3aab375e0ea7..165d260cd9d3 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_decl.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_decl.hpp
@@ -64,118 +64,113 @@
 
 namespace MueLu {
 
-  /*!
-    @class TentativePFactory class.
-    @brief Factory for building tentative prolongator.
-
-    Factory for creating tentative prolongator.   Nullspace vectors are split across aggregates so that they
-    have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
-    tentative prolongator, and the R becomes the coarse nullspace.
-
-    Note that the TentativePFactory also creates the coarse null space vectors, that is, it serves as generating
-    factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
-    level one has to use the NullspaceFactory.
-
-    @ingroup MueLuTransferClasses
-
-    ## Input/output of TentativePFactory ##
-
-    ### User parameters of TentativePFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-     A                  | Factory | null |   | * | * | Generating factory of the matrix A
-     Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
-     Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
-     UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
-     CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TentativePFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see TentativePFactory::DeclareInput).
-
-    ### Variables provided by TentativePFactory ###
-
-    After TentativePFactory::Build the following data is available (if requested)
-
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | P       | TentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
-    | Nullspace | TentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
-  */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TentativePFactory_kokkos;
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  class TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
-  public:
-    typedef LocalOrdinal                                             local_ordinal_type;
-    typedef GlobalOrdinal                                            global_ordinal_type;
-    typedef typename DeviceType::execution_space                     execution_space;
-    typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
-    typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>      node_type;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType   real_type;
-    typedef Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, node_type> RealValuedMultiVector;
-
-  private:
-    // For compatibility
-    typedef node_type                                           Node;
+/*!
+  @class TentativePFactory class.
+  @brief Factory for building tentative prolongator.
+
+  Factory for creating tentative prolongator.   Nullspace vectors are split across aggregates so that they
+  have local support.  The vectors with local support are factored via LAPACK QR.  The Q becomes the
+  tentative prolongator, and the R becomes the coarse nullspace.
+
+  Note that the TentativePFactory also creates the coarse null space vectors, that is, it serves as generating
+  factory for the Nullspace vectors on the coarse levels. To feed in the near null space vectors on the finest
+  level one has to use the NullspaceFactory.
+
+  @ingroup MueLuTransferClasses
+
+  ## Input/output of TentativePFactory ##
+
+  ### User parameters of TentativePFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+   A                  | Factory | null |   | * | * | Generating factory of the matrix A
+   Aggregates         | Factory | null |   | * | * | Generating factory of the aggregates (of type "Aggregates" produced, e.g., by the UncoupledAggregationFactory)
+   Nullspace          | Factory | null |   | * | * | Generating factory of the fine nullspace vectors (of type "MultiVector"). In the default case the same instance of the TentativePFactory is also the generating factory for the null space vectors (on the next coarser levels). Therefore, it is recommended to declare the TentativePFactory to be the generating factory of the "Nullspace" variable globally using the FactoryManager object! For defining the near null space vectors on the finest level one should use the NullspaceFactory.
+   UnAmalgamationInfo | Factory | null |   | * | * | Generating factory of UnAmalgamationInfo. This data (of type "AmalgamationInfo") is produced by the AmalgamationFactory class. The default option should be fine for most standard cases though.
+   CoarseMap          | Factory | null |   | * | * | Generating factory of the coarse map. The map generates the coarse domain map of the prolongation operator. The default choice should be fine as long as you do not wish some offset in the domain map (e.g. for block operators in multiphysics problems). The data is generated by the CoarseMapFactory.
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see TentativePFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see TentativePFactory::DeclareInput).
+
+  ### Variables provided by TentativePFactory ###
+
+  After TentativePFactory::Build the following data is available (if requested)
+
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | P       | TentativePFactory   | Non-smoothed "tentative" prolongation operator (with piece-wise constant transfer operator basis functions)
+  | Nullspace | TentativePFactory | Coarse near null space vectors. Please also check the documentation of the NullspaceFactory for the special dependency tree of the "Nullspace" variable throughout all multigrid levels.
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TentativePFactory_kokkos;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+class TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> > : public PFactory {
+ public:
+  typedef LocalOrdinal local_ordinal_type;
+  typedef GlobalOrdinal global_ordinal_type;
+  typedef typename DeviceType::execution_space execution_space;
+  typedef Kokkos::RangePolicy<local_ordinal_type, execution_space> range_type;
+  typedef Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType> node_type;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType real_type;
+  typedef Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, node_type> RealValuedMultiVector;
+
+ private:
+  // For compatibility
+  typedef node_type Node;
 #undef MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    TentativePFactory_kokkos() { }
+  //! Constructor
+  TentativePFactory_kokkos() {}
 
-    //! Destructor.
-    virtual ~TentativePFactory_kokkos() { }
-    //@}
+  //! Destructor.
+  virtual ~TentativePFactory_kokkos() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
+  // NOTE: All of thess should really be private, but CUDA doesn't like that
 
-    // NOTE: All of thess should really be private, but CUDA doesn't like that
-    
-    void BuildPuncoupledBlockCrs(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo,
-                                 RCP<MultiVector> fineNullspace, RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
+  void BuildPuncoupledBlockCrs(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates, RCP<AmalgamationInfo> amalgInfo,
+                               RCP<MultiVector> fineNullspace, RCP<const Map> coarseMap, RCP<Matrix>& Ptentative, RCP<MultiVector>& coarseNullspace, const int levelID) const;
 
+  bool isGoodMap(const Map& rowMap, const Map& colMap) const;
 
-    bool isGoodMap(const Map& rowMap, const Map& colMap) const;
+  void BuildPcoupled(RCP<Matrix> A, RCP<Aggregates> aggregates,
+                     RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                     RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
+                     RCP<MultiVector>& coarseNullspace) const;
 
+  void BuildPuncoupled(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates,
+                       RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                       RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
+                       RCP<MultiVector>& coarseNullspace, const int levelID) const;
 
+  mutable bool bTransferCoordinates_ = false;
+};
 
-    void BuildPcoupled  (RCP<Matrix> A, RCP<Aggregates> aggregates,
-                         RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                         RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
-                         RCP<MultiVector>& coarseNullspace) const;
-
-    void BuildPuncoupled(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates,
-                         RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                         RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
-                         RCP<MultiVector>& coarseNullspace, const int levelID) const;
-
-    mutable bool bTransferCoordinates_ = false;
-
-  };
-
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
-#endif // MUELU_TENTATIVEPFACTORY_KOKKOS_DECL_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_KOKKOS_DECL_HPP
diff --git a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_def.hpp b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_def.hpp
index 2a4484eac756..7be54dc06867 100644
--- a/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_def.hpp
+++ b/packages/muelu/src/Transfers/Smoothed-Aggregation/MueLu_TentativePFactory_kokkos_def.hpp
@@ -62,103 +62,103 @@
 
 namespace MueLu {
 
-  namespace { // anonymous
-
-    template<class LocalOrdinal, class View>
-    class ReduceMaxFunctor{
-    public:
-      ReduceMaxFunctor(View view) : view_(view) { }
-
-      KOKKOS_INLINE_FUNCTION
-      void operator()(const LocalOrdinal &i, LocalOrdinal& vmax) const {
-        if (vmax < view_(i))
-          vmax = view_(i);
-      }
-
-      KOKKOS_INLINE_FUNCTION
-      void join (LocalOrdinal& dst, const LocalOrdinal& src) const {
-        if (dst < src) {
-          dst = src;
-        }
-      }
-
-      KOKKOS_INLINE_FUNCTION
-      void init (LocalOrdinal& dst) const {
-        dst = 0;
-      }
-    private:
-      View view_;
-    };
-
-    // local QR decomposition
-    template<class LOType, class GOType, class SCType,class DeviceType, class NspType, class aggRowsType, class maxAggDofSizeType, class agg2RowMapLOType, class statusType, class rowsType, class rowsAuxType, class colsAuxType, class valsAuxType>
-    class LocalQRDecompFunctor {
-    private:
-      typedef LOType LO;
-      typedef GOType GO;
-      typedef SCType SC;
-
-      typedef typename DeviceType::execution_space execution_space;
-      typedef typename Kokkos::ArithTraits<SC>::val_type impl_SC;
-      typedef Kokkos::ArithTraits<impl_SC> impl_ATS;
-      typedef typename impl_ATS::magnitudeType Magnitude;
-
-      typedef Kokkos::View<impl_SC**,typename execution_space::scratch_memory_space, Kokkos::MemoryUnmanaged> shared_matrix;
-      typedef Kokkos::View<impl_SC* ,typename execution_space::scratch_memory_space, Kokkos::MemoryUnmanaged> shared_vector;
-
-    private:
-
-      NspType fineNS;
-      NspType coarseNS;
-      aggRowsType aggRows;
-      maxAggDofSizeType maxAggDofSize; //< maximum number of dofs in aggregate (max size of aggregate * numDofsPerNode)
-      agg2RowMapLOType agg2RowMapLO;
-      statusType statusAtomic;
-      rowsType rows;
-      rowsAuxType rowsAux;
-      colsAuxType colsAux;
-      valsAuxType valsAux;
-      bool doQRStep;
-    public:
-      LocalQRDecompFunctor(NspType fineNS_, NspType coarseNS_, aggRowsType aggRows_, maxAggDofSizeType maxAggDofSize_, agg2RowMapLOType agg2RowMapLO_, statusType statusAtomic_, rowsType rows_, rowsAuxType rowsAux_, colsAuxType colsAux_, valsAuxType valsAux_, bool doQRStep_) :
-        fineNS(fineNS_),
-        coarseNS(coarseNS_),
-        aggRows(aggRows_),
-        maxAggDofSize(maxAggDofSize_),
-        agg2RowMapLO(agg2RowMapLO_),
-        statusAtomic(statusAtomic_),
-        rows(rows_),
-        rowsAux(rowsAux_),
-        colsAux(colsAux_),
-        valsAux(valsAux_),
-        doQRStep(doQRStep_)
-        { }
-
-      KOKKOS_INLINE_FUNCTION
-      void operator() ( const typename Kokkos::TeamPolicy<execution_space>::member_type & thread, size_t& nnz) const {
-        auto agg = thread.league_rank();
-
-        // size of aggregate: number of DOFs in aggregate
-        auto aggSize = aggRows(agg+1) - aggRows(agg);
-
-        const impl_SC one     = impl_ATS::one();
-        const impl_SC two     = one + one;
-        const impl_SC zero    = impl_ATS::zero();
-        const auto zeroM = impl_ATS::magnitude(zero);
-
-        int m = aggSize;
-        int n = fineNS.extent(1);
+namespace {  // anonymous
+
+template <class LocalOrdinal, class View>
+class ReduceMaxFunctor {
+ public:
+  ReduceMaxFunctor(View view)
+    : view_(view) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LocalOrdinal& i, LocalOrdinal& vmax) const {
+    if (vmax < view_(i))
+      vmax = view_(i);
+  }
 
-        // calculate row offset for coarse nullspace
-        Xpetra::global_size_t offset = agg * n;
+  KOKKOS_INLINE_FUNCTION
+  void join(LocalOrdinal& dst, const LocalOrdinal& src) const {
+    if (dst < src) {
+      dst = src;
+    }
+  }
 
-        if (doQRStep) {
+  KOKKOS_INLINE_FUNCTION
+  void init(LocalOrdinal& dst) const {
+    dst = 0;
+  }
 
-          // Extract the piece of the nullspace corresponding to the aggregate
-          shared_matrix r(thread.team_shmem(), m, n);     // A (initially), R (at the end)
-          for (int j = 0; j < n; j++)
-            for (int k = 0; k < m; k++)
-              r(k,j) = fineNS(agg2RowMapLO(aggRows(agg)+k),j);
+ private:
+  View view_;
+};
+
+// local QR decomposition
+template <class LOType, class GOType, class SCType, class DeviceType, class NspType, class aggRowsType, class maxAggDofSizeType, class agg2RowMapLOType, class statusType, class rowsType, class rowsAuxType, class colsAuxType, class valsAuxType>
+class LocalQRDecompFunctor {
+ private:
+  typedef LOType LO;
+  typedef GOType GO;
+  typedef SCType SC;
+
+  typedef typename DeviceType::execution_space execution_space;
+  typedef typename Kokkos::ArithTraits<SC>::val_type impl_SC;
+  typedef Kokkos::ArithTraits<impl_SC> impl_ATS;
+  typedef typename impl_ATS::magnitudeType Magnitude;
+
+  typedef Kokkos::View<impl_SC**, typename execution_space::scratch_memory_space, Kokkos::MemoryUnmanaged> shared_matrix;
+  typedef Kokkos::View<impl_SC*, typename execution_space::scratch_memory_space, Kokkos::MemoryUnmanaged> shared_vector;
+
+ private:
+  NspType fineNS;
+  NspType coarseNS;
+  aggRowsType aggRows;
+  maxAggDofSizeType maxAggDofSize;  //< maximum number of dofs in aggregate (max size of aggregate * numDofsPerNode)
+  agg2RowMapLOType agg2RowMapLO;
+  statusType statusAtomic;
+  rowsType rows;
+  rowsAuxType rowsAux;
+  colsAuxType colsAux;
+  valsAuxType valsAux;
+  bool doQRStep;
+
+ public:
+  LocalQRDecompFunctor(NspType fineNS_, NspType coarseNS_, aggRowsType aggRows_, maxAggDofSizeType maxAggDofSize_, agg2RowMapLOType agg2RowMapLO_, statusType statusAtomic_, rowsType rows_, rowsAuxType rowsAux_, colsAuxType colsAux_, valsAuxType valsAux_, bool doQRStep_)
+    : fineNS(fineNS_)
+    , coarseNS(coarseNS_)
+    , aggRows(aggRows_)
+    , maxAggDofSize(maxAggDofSize_)
+    , agg2RowMapLO(agg2RowMapLO_)
+    , statusAtomic(statusAtomic_)
+    , rows(rows_)
+    , rowsAux(rowsAux_)
+    , colsAux(colsAux_)
+    , valsAux(valsAux_)
+    , doQRStep(doQRStep_) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread, size_t& nnz) const {
+    auto agg = thread.league_rank();
+
+    // size of aggregate: number of DOFs in aggregate
+    auto aggSize = aggRows(agg + 1) - aggRows(agg);
+
+    const impl_SC one  = impl_ATS::one();
+    const impl_SC two  = one + one;
+    const impl_SC zero = impl_ATS::zero();
+    const auto zeroM   = impl_ATS::magnitude(zero);
+
+    int m = aggSize;
+    int n = fineNS.extent(1);
+
+    // calculate row offset for coarse nullspace
+    Xpetra::global_size_t offset = agg * n;
+
+    if (doQRStep) {
+      // Extract the piece of the nullspace corresponding to the aggregate
+      shared_matrix r(thread.team_shmem(), m, n);  // A (initially), R (at the end)
+      for (int j = 0; j < n; j++)
+        for (int k = 0; k < m; k++)
+          r(k, j) = fineNS(agg2RowMapLO(aggRows(agg) + k), j);
 #if 0
           printf("A\n");
           for (int i = 0; i < m; i++) {
@@ -168,70 +168,70 @@ namespace MueLu {
           }
 #endif
 
-          // Calculate QR decomposition (standard)
-          shared_matrix q(thread.team_shmem(), m, m);     // Q
-          if (m >= n) {
-            bool isSingular = false;
-
-            // Initialize Q^T
-            auto qt = q;
-            for (int i = 0; i < m; i++) {
-              for (int j = 0; j < m; j++)
-                qt(i,j) = zero;
-              qt(i,i) = one;
-            }
+      // Calculate QR decomposition (standard)
+      shared_matrix q(thread.team_shmem(), m, m);  // Q
+      if (m >= n) {
+        bool isSingular = false;
+
+        // Initialize Q^T
+        auto qt = q;
+        for (int i = 0; i < m; i++) {
+          for (int j = 0; j < m; j++)
+            qt(i, j) = zero;
+          qt(i, i) = one;
+        }
 
-            for (int k = 0; k < n; k++) {  // we ignore "n" instead of "n-1" to normalize
-              // FIXME_KOKKOS: use team
-              Magnitude s = zeroM, norm, norm_x;
-              for (int i = k+1; i < m; i++)
-                s += pow(impl_ATS::magnitude(r(i,k)), 2);
-              norm = sqrt(pow(impl_ATS::magnitude(r(k,k)), 2) + s);
+        for (int k = 0; k < n; k++) {  // we ignore "n" instead of "n-1" to normalize
+          // FIXME_KOKKOS: use team
+          Magnitude s = zeroM, norm, norm_x;
+          for (int i = k + 1; i < m; i++)
+            s += pow(impl_ATS::magnitude(r(i, k)), 2);
+          norm = sqrt(pow(impl_ATS::magnitude(r(k, k)), 2) + s);
 
-              if (norm == zero) {
-                isSingular = true;
-                break;
-              }
+          if (norm == zero) {
+            isSingular = true;
+            break;
+          }
 
-              r(k,k) -= norm*one;
+          r(k, k) -= norm * one;
 
-              norm_x = sqrt(pow(impl_ATS::magnitude(r(k,k)), 2) + s);
-              if (norm_x == zeroM) {
-                // We have a single diagonal element in the column.
-                // No reflections required. Just need to restor r(k,k).
-                r(k,k) = norm*one;
-                continue;
-              }
+          norm_x = sqrt(pow(impl_ATS::magnitude(r(k, k)), 2) + s);
+          if (norm_x == zeroM) {
+            // We have a single diagonal element in the column.
+            // No reflections required. Just need to restor r(k,k).
+            r(k, k) = norm * one;
+            continue;
+          }
 
-              // FIXME_KOKKOS: use team
-              for (int i = k; i < m; i++)
-                r(i,k) /= norm_x;
-
-              // Update R(k:m,k+1:n)
-              for (int j = k+1; j < n; j++) {
-                // FIXME_KOKKOS: use team in the loops
-                impl_SC si = zero;
-                for (int i = k; i < m; i++)
-                  si += r(i,k) * r(i,j);
-                for (int i = k; i < m; i++)
-                  r(i,j) -= two*si * r(i,k);
-              }
+          // FIXME_KOKKOS: use team
+          for (int i = k; i < m; i++)
+            r(i, k) /= norm_x;
+
+          // Update R(k:m,k+1:n)
+          for (int j = k + 1; j < n; j++) {
+            // FIXME_KOKKOS: use team in the loops
+            impl_SC si = zero;
+            for (int i = k; i < m; i++)
+              si += r(i, k) * r(i, j);
+            for (int i = k; i < m; i++)
+              r(i, j) -= two * si * r(i, k);
+          }
 
-              // Update Q^T (k:m,k:m)
-              for (int j = k; j < m; j++) {
-                // FIXME_KOKKOS: use team in the loops
-                impl_SC si = zero;
-                for (int i = k; i < m; i++)
-                  si += r(i,k) * qt(i,j);
-                for (int i = k; i < m; i++)
-                  qt(i,j) -= two*si * r(i,k);
-              }
+          // Update Q^T (k:m,k:m)
+          for (int j = k; j < m; j++) {
+            // FIXME_KOKKOS: use team in the loops
+            impl_SC si = zero;
+            for (int i = k; i < m; i++)
+              si += r(i, k) * qt(i, j);
+            for (int i = k; i < m; i++)
+              qt(i, j) -= two * si * r(i, k);
+          }
 
-              // Fix R(k:m,k)
-              r(k,k) = norm*one;
-              for (int i = k+1; i < m; i++)
-                r(i,k) = zero;
-            }
+          // Fix R(k:m,k)
+          r(k, k) = norm * one;
+          for (int i = k + 1; i < m; i++)
+            r(i, k) = zero;
+        }
 
 #if 0
             // Q = (Q^T)^T
@@ -243,76 +243,76 @@ namespace MueLu {
               }
 #endif
 
-            // Build coarse nullspace using the upper triangular part of R
-            for (int j = 0; j < n; j++)
-              for (int k = 0; k <= j; k++)
-                coarseNS(offset+k,j) = r(k,j);
+        // Build coarse nullspace using the upper triangular part of R
+        for (int j = 0; j < n; j++)
+          for (int k = 0; k <= j; k++)
+            coarseNS(offset + k, j) = r(k, j);
 
-            if (isSingular) {
-              statusAtomic(1) = true;
-              return;
-            }
+        if (isSingular) {
+          statusAtomic(1) = true;
+          return;
+        }
 
-          } else {
-            // Special handling for m < n (i.e. single node aggregates in structural mechanics)
-
-            // The local QR decomposition is not possible in the "overconstrained"
-            // case (i.e. number of columns in qr > number of rowsAux), which
-            // corresponds to #DOFs in Aggregate < n. For usual problems this
-            // is only possible for single node aggregates in structural mechanics.
-            // (Similar problems may arise in discontinuous Galerkin problems...)
-            // We bypass the QR decomposition and use an identity block in the
-            // tentative prolongator for the single node aggregate and transfer the
-            // corresponding fine level null space information 1-to-1 to the coarse
-            // level null space part.
-
-            // NOTE: The resulting tentative prolongation operator has
-            // (m*DofsPerNode-n) zero columns leading to a singular
-            // coarse level operator A.  To deal with that one has the following
-            // options:
-            // - Use the "RepairMainDiagonal" flag in the RAPFactory (default:
-            //   false) to add some identity block to the diagonal of the zero rowsAux
-            //   in the coarse level operator A, such that standard level smoothers
-            //   can be used again.
-            // - Use special (projection-based) level smoothers, which can deal
-            //   with singular matrices (very application specific)
-            // - Adapt the code below to avoid zero columns. However, we do not
-            //   support a variable number of DOFs per node in MueLu/Xpetra which
-            //   makes the implementation really hard.
-            //
-            // FIXME: do we need to check for singularity here somehow? Zero
-            // columns would be easy but linear dependency would require proper QR.
-
-            // R = extended (by adding identity rowsAux) qr
-            for (int j = 0; j < n; j++)
-              for (int k = 0; k < n; k++)
-                if (k < m)
-                  coarseNS(offset+k,j) = r(k,j);
-                else
-                  coarseNS(offset+k,j) = (k == j ? one : zero);
+      } else {
+        // Special handling for m < n (i.e. single node aggregates in structural mechanics)
+
+        // The local QR decomposition is not possible in the "overconstrained"
+        // case (i.e. number of columns in qr > number of rowsAux), which
+        // corresponds to #DOFs in Aggregate < n. For usual problems this
+        // is only possible for single node aggregates in structural mechanics.
+        // (Similar problems may arise in discontinuous Galerkin problems...)
+        // We bypass the QR decomposition and use an identity block in the
+        // tentative prolongator for the single node aggregate and transfer the
+        // corresponding fine level null space information 1-to-1 to the coarse
+        // level null space part.
+
+        // NOTE: The resulting tentative prolongation operator has
+        // (m*DofsPerNode-n) zero columns leading to a singular
+        // coarse level operator A.  To deal with that one has the following
+        // options:
+        // - Use the "RepairMainDiagonal" flag in the RAPFactory (default:
+        //   false) to add some identity block to the diagonal of the zero rowsAux
+        //   in the coarse level operator A, such that standard level smoothers
+        //   can be used again.
+        // - Use special (projection-based) level smoothers, which can deal
+        //   with singular matrices (very application specific)
+        // - Adapt the code below to avoid zero columns. However, we do not
+        //   support a variable number of DOFs per node in MueLu/Xpetra which
+        //   makes the implementation really hard.
+        //
+        // FIXME: do we need to check for singularity here somehow? Zero
+        // columns would be easy but linear dependency would require proper QR.
+
+        // R = extended (by adding identity rowsAux) qr
+        for (int j = 0; j < n; j++)
+          for (int k = 0; k < n; k++)
+            if (k < m)
+              coarseNS(offset + k, j) = r(k, j);
+            else
+              coarseNS(offset + k, j) = (k == j ? one : zero);
+
+        // Q = I (rectangular)
+        for (int i = 0; i < m; i++)
+          for (int j = 0; j < n; j++)
+            q(i, j) = (j == i ? one : zero);
+      }
 
-            // Q = I (rectangular)
-            for (int i = 0; i < m; i++)
-              for (int j = 0; j < n; j++)
-                q(i,j) = (j == i ? one : zero);
-          }
-
-          // Process each row in the local Q factor and fill helper arrays to assemble P
-          for (int j = 0; j < m; j++) {
-            LO localRow = agg2RowMapLO(aggRows(agg)+j);
-            size_t rowStart = rowsAux(localRow);
-            size_t lnnz = 0;
-            for (int k = 0; k < n; k++) {
-              // skip zeros
-              if (q(j,k) != zero) {
-                colsAux(rowStart+lnnz) = offset + k;
-                valsAux(rowStart+lnnz) = q(j,k);
-                lnnz++;
-              }
-            }
-            rows(localRow+1) = lnnz;
-            nnz += lnnz;
+      // Process each row in the local Q factor and fill helper arrays to assemble P
+      for (int j = 0; j < m; j++) {
+        LO localRow     = agg2RowMapLO(aggRows(agg) + j);
+        size_t rowStart = rowsAux(localRow);
+        size_t lnnz     = 0;
+        for (int k = 0; k < n; k++) {
+          // skip zeros
+          if (q(j, k) != zero) {
+            colsAux(rowStart + lnnz) = offset + k;
+            valsAux(rowStart + lnnz) = q(j, k);
+            lnnz++;
           }
+        }
+        rows(localRow + 1) = lnnz;
+        nnz += lnnz;
+      }
 
 #if 0
           printf("R\n");
@@ -329,346 +329,344 @@ namespace MueLu {
             printf("\n");
           }
 #endif
-        } else {
-          /////////////////////////////
-          //      "no-QR" option     //
-          /////////////////////////////
-          // Local Q factor is just the fine nullspace support over the current aggregate.
-          // Local R factor is the identity.
-          // TODO I have not implemented any special handling for aggregates that are too
-          // TODO small to locally support the nullspace, as is done in the standard QR
-          // TODO case above.
-
-          for (int j = 0; j < m; j++) {
-            LO localRow = agg2RowMapLO(aggRows(agg)+j);
-            size_t rowStart = rowsAux(localRow);
-            size_t lnnz = 0;
-            for (int k = 0; k < n; k++) {
-              const impl_SC qr_jk = fineNS(localRow,k);
-              // skip zeros
-              if (qr_jk != zero) {
-                colsAux(rowStart+lnnz) = offset + k;
-                valsAux(rowStart+lnnz) = qr_jk;
-                lnnz++;
-              }
-            }
-            rows(localRow+1) = lnnz;
-            nnz += lnnz;
+    } else {
+      /////////////////////////////
+      //      "no-QR" option     //
+      /////////////////////////////
+      // Local Q factor is just the fine nullspace support over the current aggregate.
+      // Local R factor is the identity.
+      // TODO I have not implemented any special handling for aggregates that are too
+      // TODO small to locally support the nullspace, as is done in the standard QR
+      // TODO case above.
+
+      for (int j = 0; j < m; j++) {
+        LO localRow     = agg2RowMapLO(aggRows(agg) + j);
+        size_t rowStart = rowsAux(localRow);
+        size_t lnnz     = 0;
+        for (int k = 0; k < n; k++) {
+          const impl_SC qr_jk = fineNS(localRow, k);
+          // skip zeros
+          if (qr_jk != zero) {
+            colsAux(rowStart + lnnz) = offset + k;
+            valsAux(rowStart + lnnz) = qr_jk;
+            lnnz++;
           }
-
-          for (int j = 0; j < n; j++)
-            coarseNS(offset+j,j) = one;
-
         }
-
+        rows(localRow + 1) = lnnz;
+        nnz += lnnz;
       }
 
-      // amount of shared memory
-      size_t team_shmem_size( int /* team_size */ ) const {
-        if (doQRStep) {
-          int m = maxAggDofSize;
-          int n = fineNS.extent(1);
-          return shared_matrix::shmem_size(m, n) +    // r
-            shared_matrix::shmem_size(m, m);     // q
-        } else
-          return 0;
-      }
-    };
+      for (int j = 0; j < n; j++)
+        coarseNS(offset + j, j) = one;
+    }
+  }
+
+  // amount of shared memory
+  size_t team_shmem_size(int /* team_size */) const {
+    if (doQRStep) {
+      int m = maxAggDofSize;
+      int n = fineNS.extent(1);
+      return shared_matrix::shmem_size(m, n) +  // r
+             shared_matrix::shmem_size(m, m);   // q
+    } else
+      return 0;
+  }
+};
 
-  } // namespace anonymous
+}  // namespace
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  RCP<const ParameterList> TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+RCP<const ParameterList> TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
 #define SET_VALID_ENTRY(name) validParamList->setEntry(name, MasterList::getEntry(name))
   SET_VALID_ENTRY("tentative: calculate qr");
   SET_VALID_ENTRY("tentative: build coarse coordinates");
 #undef SET_VALID_ENTRY
 
-    validParamList->set< RCP<const FactoryBase> >("A",                  Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Aggregates",         Teuchos::null, "Generating factory of the aggregates");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",          Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< RCP<const FactoryBase> >("Scaled Nullspace",   Teuchos::null, "Generating factory of the scaled nullspace");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
-    validParamList->set< RCP<const FactoryBase> >("CoarseMap",          Teuchos::null, "Generating factory of the coarse map");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates",        Teuchos::null, "Generating factory of the coordinates");
-    validParamList->set< RCP<const FactoryBase> >("Node Comm",          Teuchos::null, "Generating factory of the node level communicator");
-
-    // Make sure we don't recursively validate options for the matrixmatrix kernels
-    ParameterList norecurse;
-    norecurse.disableRecursiveValidation();
-    validParamList->set<ParameterList> ("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
-
-    return validParamList;
+  validParamList->set<RCP<const FactoryBase>>("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase>>("Aggregates", Teuchos::null, "Generating factory of the aggregates");
+  validParamList->set<RCP<const FactoryBase>>("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<RCP<const FactoryBase>>("Scaled Nullspace", Teuchos::null, "Generating factory of the scaled nullspace");
+  validParamList->set<RCP<const FactoryBase>>("UnAmalgamationInfo", Teuchos::null, "Generating factory of UnAmalgamationInfo");
+  validParamList->set<RCP<const FactoryBase>>("CoarseMap", Teuchos::null, "Generating factory of the coarse map");
+  validParamList->set<RCP<const FactoryBase>>("Coordinates", Teuchos::null, "Generating factory of the coordinates");
+  validParamList->set<RCP<const FactoryBase>>("Node Comm", Teuchos::null, "Generating factory of the node level communicator");
+
+  // Make sure we don't recursively validate options for the matrixmatrix kernels
+  ParameterList norecurse;
+  norecurse.disableRecursiveValidation();
+  validParamList->set<ParameterList>("matrixmatrix: kernel params", norecurse, "MatrixMatrix kernel parameters");
+
+  return validParamList;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  const ParameterList& pL = GetParameterList();
+  // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
+  std::string nspName = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  Input(fineLevel, "A");
+  Input(fineLevel, "Aggregates");
+  Input(fineLevel, nspName);
+  Input(fineLevel, "UnAmalgamationInfo");
+  Input(fineLevel, "CoarseMap");
+  if (fineLevel.GetLevelID() == 0 &&
+      fineLevel.IsAvailable("Coordinates", NoFactory::get()) &&  // we have coordinates (provided by user app)
+      pL.get<bool>("tentative: build coarse coordinates")) {     // and we want coordinates on other levels
+    bTransferCoordinates_ = true;                                // then set the transfer coordinates flag to true
+    Input(fineLevel, "Coordinates");
+  } else if (bTransferCoordinates_) {
+    Input(fineLevel, "Coordinates");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-
-    const ParameterList& pL = GetParameterList();
-    // NOTE: This guy can only either be 'Nullspace' or 'Scaled Nullspace' or else the validator above will cause issues
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-    Input(fineLevel, "A");
-    Input(fineLevel, "Aggregates");
-    Input(fineLevel, nspName);
-    Input(fineLevel, "UnAmalgamationInfo");
-    Input(fineLevel, "CoarseMap");
-    if( fineLevel.GetLevelID() == 0 &&
-        fineLevel.IsAvailable("Coordinates", NoFactory::get()) &&     // we have coordinates (provided by user app)
-        pL.get<bool>("tentative: build coarse coordinates") ) {       // and we want coordinates on other levels
-      bTransferCoordinates_ = true;                                   // then set the transfer coordinates flag to true
-      Input(fineLevel, "Coordinates");
-    } else if (bTransferCoordinates_) {
-      Input(fineLevel, "Coordinates");
-    }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
+  typedef Xpetra::MultiVectorFactory<coordinate_type, LO, GO, NO> RealValuedMultiVectorFactory;
+  const ParameterList& pL = GetParameterList();
+  std::string nspName     = "Nullspace";
+  if (pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
+
+  auto A             = Get<RCP<Matrix>>(fineLevel, "A");
+  auto aggregates    = Get<RCP<Aggregates>>(fineLevel, "Aggregates");
+  auto amalgInfo     = Get<RCP<AmalgamationInfo>>(fineLevel, "UnAmalgamationInfo");
+  auto fineNullspace = Get<RCP<MultiVector>>(fineLevel, nspName);
+  auto coarseMap     = Get<RCP<const Map>>(fineLevel, "CoarseMap");
+  RCP<RealValuedMultiVector> fineCoords;
+  if (bTransferCoordinates_) {
+    fineCoords = Get<RCP<RealValuedMultiVector>>(fineLevel, "Coordinates");
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
+  RCP<Matrix> Ptentative;
+  // No coarse DoFs so we need to bail by setting Ptentattive to null and returning
+  // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
+  if (aggregates->GetNumGlobalAggregatesComputeIfNeeded() == 0) {
+    Ptentative = Teuchos::null;
+    Set(coarseLevel, "P", Ptentative);
+    return;
   }
+  RCP<MultiVector> coarseNullspace;
+  RCP<RealValuedMultiVector> coarseCoords;
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
-    typedef Xpetra::MultiVectorFactory<coordinate_type,LO,GO,NO> RealValuedMultiVectorFactory;
-    const ParameterList& pL = GetParameterList();
-    std::string nspName = "Nullspace";
-    if(pL.isParameter("Nullspace name")) nspName = pL.get<std::string>("Nullspace name");
-
-    auto A             = Get< RCP<Matrix> >            (fineLevel, "A");
-    auto aggregates    = Get< RCP<Aggregates> > (fineLevel, "Aggregates");
-    auto amalgInfo     = Get< RCP<AmalgamationInfo> >  (fineLevel, "UnAmalgamationInfo");
-    auto fineNullspace = Get< RCP<MultiVector> >       (fineLevel, nspName);
-    auto coarseMap     = Get< RCP<const Map> >         (fineLevel, "CoarseMap");
-    RCP<RealValuedMultiVector> fineCoords;
-    if(bTransferCoordinates_) {
-      fineCoords = Get< RCP<RealValuedMultiVector> >(fineLevel, "Coordinates");
-    }
+  if (bTransferCoordinates_) {
+    RCP<const Map> coarseCoordMap;
+    using array_type = typename Map::global_indices_array_device_type;
 
-    RCP<Matrix>      Ptentative;
-    // No coarse DoFs so we need to bail by setting Ptentattive to null and returning
-    // This level will ultimately be removed in MueLu_Hierarchy_defs.h via a resize()
-    if ( aggregates->GetNumGlobalAggregatesComputeIfNeeded() == 0) {
-      Ptentative = Teuchos::null;
-      Set(coarseLevel, "P", Ptentative);
-      return;
-    }
-    RCP<MultiVector> coarseNullspace;
-    RCP<RealValuedMultiVector> coarseCoords;
+    LO blkSize = 1;
+    if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null)
+      blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
 
-    if(bTransferCoordinates_) {
-      RCP<const Map> coarseCoordMap;
-      using array_type = typename Map::global_indices_array_device_type;
-
-      LO blkSize = 1;
-      if (rcp_dynamic_cast<const StridedMap>(coarseMap) != Teuchos::null)
-        blkSize = rcp_dynamic_cast<const StridedMap>(coarseMap)->getFixedBlockSize();
-
-      if (blkSize == 1) {
-        // Scalar system
-        // No amalgamation required, we can use the coarseMap
-        coarseCoordMap = coarseMap;
-      } else {
-        // Vector system
-        // NOTE: There could be further optimizations here where we detect contiguous maps and then
-        // create a contiguous amalgamated maps, which bypasses the expense of the getMyGlobalIndicesDevice() 
-        // call (which is free for non-contiguous maps, but costs something if the map is contiguous).
-        using range_policy = Kokkos::RangePolicy<typename Node::execution_space>;         
-        array_type elementAList = coarseMap->getMyGlobalIndicesDevice();
-        GO indexBase = coarseMap->getIndexBase();        
-        auto numElements = elementAList.size() / blkSize;
-        typename array_type::non_const_type elementList_nc("elementList",numElements);
-
-        // Amalgamate the map
-        Kokkos::parallel_for("Amalgamate Element List",range_policy(0,numElements),KOKKOS_LAMBDA(LO i) {
-            elementList_nc[i] = (elementAList[i*blkSize]-indexBase)/blkSize + indexBase;
+    if (blkSize == 1) {
+      // Scalar system
+      // No amalgamation required, we can use the coarseMap
+      coarseCoordMap = coarseMap;
+    } else {
+      // Vector system
+      // NOTE: There could be further optimizations here where we detect contiguous maps and then
+      // create a contiguous amalgamated maps, which bypasses the expense of the getMyGlobalIndicesDevice()
+      // call (which is free for non-contiguous maps, but costs something if the map is contiguous).
+      using range_policy      = Kokkos::RangePolicy<typename Node::execution_space>;
+      array_type elementAList = coarseMap->getMyGlobalIndicesDevice();
+      GO indexBase            = coarseMap->getIndexBase();
+      auto numElements        = elementAList.size() / blkSize;
+      typename array_type::non_const_type elementList_nc("elementList", numElements);
+
+      // Amalgamate the map
+      Kokkos::parallel_for(
+          "Amalgamate Element List", range_policy(0, numElements), KOKKOS_LAMBDA(LO i) {
+            elementList_nc[i] = (elementAList[i * blkSize] - indexBase) / blkSize + indexBase;
           });
-        array_type elementList = elementList_nc;
-        coarseCoordMap = MapFactory::Build(coarseMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                           elementList, indexBase, coarseMap->getComm());
-      }
+      array_type elementList = elementList_nc;
+      coarseCoordMap         = MapFactory::Build(coarseMap->lib(), Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                 elementList, indexBase, coarseMap->getComm());
+    }
 
-      coarseCoords = RealValuedMultiVectorFactory::Build(coarseCoordMap, fineCoords->getNumVectors());
+    coarseCoords = RealValuedMultiVectorFactory::Build(coarseCoordMap, fineCoords->getNumVectors());
 
-      // Create overlapped fine coordinates to reduce global communication
-      auto uniqueMap      = fineCoords->getMap();
-      RCP<RealValuedMultiVector> ghostedCoords = fineCoords;
-      if (aggregates->AggregatesCrossProcessors()) {
-        auto nonUniqueMap = aggregates->GetMap();
-        auto importer     = ImportFactory::Build(uniqueMap, nonUniqueMap);
+    // Create overlapped fine coordinates to reduce global communication
+    auto uniqueMap                           = fineCoords->getMap();
+    RCP<RealValuedMultiVector> ghostedCoords = fineCoords;
+    if (aggregates->AggregatesCrossProcessors()) {
+      auto nonUniqueMap = aggregates->GetMap();
+      auto importer     = ImportFactory::Build(uniqueMap, nonUniqueMap);
 
-        ghostedCoords = RealValuedMultiVectorFactory::Build(nonUniqueMap, fineCoords->getNumVectors());
-        ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
-      }
+      ghostedCoords = RealValuedMultiVectorFactory::Build(nonUniqueMap, fineCoords->getNumVectors());
+      ghostedCoords->doImport(*fineCoords, *importer, Xpetra::INSERT);
+    }
 
-      // The good new is that his graph has already been constructed for the
-      // TentativePFactory and was cached in Aggregates. So this is a no-op.
-      auto aggGraph = aggregates->GetGraph();
-      auto numAggs  = aggGraph.numRows();
+    // The good new is that his graph has already been constructed for the
+    // TentativePFactory and was cached in Aggregates. So this is a no-op.
+    auto aggGraph = aggregates->GetGraph();
+    auto numAggs  = aggGraph.numRows();
 
-      auto fineCoordsView   = fineCoords  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      auto coarseCoordsView = coarseCoords->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+    auto fineCoordsView   = fineCoords->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto coarseCoordsView = coarseCoords->getDeviceLocalView(Xpetra::Access::OverwriteAll);
 
-      // Fill in coarse coordinates
-      {
-        SubFactoryMonitor m2(*this, "AverageCoords", coarseLevel);
+    // Fill in coarse coordinates
+    {
+      SubFactoryMonitor m2(*this, "AverageCoords", coarseLevel);
 
-        const auto dim = fineCoords->getNumVectors();
+      const auto dim = fineCoords->getNumVectors();
 
-        typename AppendTrait<decltype(fineCoordsView), Kokkos::RandomAccess>::type fineCoordsRandomView = fineCoordsView;
-        for (size_t j = 0; j < dim; j++) {
-          Kokkos::parallel_for("MueLu::TentativeP::BuildCoords", Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numAggs),
-                               KOKKOS_LAMBDA(const LO i) {
-                                 // A row in this graph represents all node ids in the aggregate
-                                 // Therefore, averaging is very easy
+      typename AppendTrait<decltype(fineCoordsView), Kokkos::RandomAccess>::type fineCoordsRandomView = fineCoordsView;
+      for (size_t j = 0; j < dim; j++) {
+        Kokkos::parallel_for(
+            "MueLu::TentativeP::BuildCoords", Kokkos::RangePolicy<local_ordinal_type, execution_space>(0, numAggs),
+            KOKKOS_LAMBDA(const LO i) {
+              // A row in this graph represents all node ids in the aggregate
+              // Therefore, averaging is very easy
 
-                                 auto aggregate = aggGraph.rowConst(i);
+              auto aggregate = aggGraph.rowConst(i);
 
-                                 coordinate_type sum = 0.0; // do not use Scalar here (Stokhos)
-                                 for (size_t colID = 0; colID < static_cast<size_t>(aggregate.length); colID++)
-                                   sum += fineCoordsRandomView(aggregate(colID),j);
+              coordinate_type sum = 0.0;  // do not use Scalar here (Stokhos)
+              for (size_t colID = 0; colID < static_cast<size_t>(aggregate.length); colID++)
+                sum += fineCoordsRandomView(aggregate(colID), j);
 
-                                 coarseCoordsView(i,j) = sum / aggregate.length;
-                               });
-        }
+              coarseCoordsView(i, j) = sum / aggregate.length;
+            });
       }
     }
+  }
 
-    if (!aggregates->AggregatesCrossProcessors()) {
-      if(Xpetra::Helpers<SC,LO,GO,NO>::isTpetraBlockCrs(A)) {
-        BuildPuncoupledBlockCrs(coarseLevel,A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace,
-                                coarseLevel.GetLevelID());
-      }
-      else {
-        BuildPuncoupled(coarseLevel, A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace, coarseLevel.GetLevelID());
-      }
-    }
-    else
-      BuildPcoupled  (A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace);
-
-    // If available, use striding information of fine level matrix A for range
-    // map and coarseMap as domain map; otherwise use plain range map of
-    // Ptent = plain range map of A for range map and coarseMap as domain map.
-    // NOTE:
-    // The latter is not really safe, since there is no striding information
-    // for the range map. This is not really a problem, since striding
-    // information is always available on the intermedium levels and the
-    // coarsest levels.
-    if (A->IsView("stridedMaps") == true)
-      Ptentative->CreateView("stridedMaps", A->getRowMap("stridedMaps"), coarseMap);
-    else
-      Ptentative->CreateView("stridedMaps", Ptentative->getRangeMap(),   coarseMap);
-
-    if(bTransferCoordinates_) {
-      Set(coarseLevel, "Coordinates", coarseCoords);
-    }
-
-    // FIXME: We should remove the NodeComm on levels past the threshold
-    if(fineLevel.IsAvailable("Node Comm")) {
-      RCP<const Teuchos::Comm<int> > nodeComm = Get<RCP<const Teuchos::Comm<int> > >(fineLevel,"Node Comm");
-      Set<RCP<const Teuchos::Comm<int> > >(coarseLevel, "Node Comm", nodeComm);
-    }
-
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-    Set(coarseLevel, "P",         Ptentative);
-
-    if (IsPrint(Statistics2)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ptentative, "Ptent", params);
+  if (!aggregates->AggregatesCrossProcessors()) {
+    if (Xpetra::Helpers<SC, LO, GO, NO>::isTpetraBlockCrs(A)) {
+      BuildPuncoupledBlockCrs(coarseLevel, A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace,
+                              coarseLevel.GetLevelID());
+    } else {
+      BuildPuncoupled(coarseLevel, A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace, coarseLevel.GetLevelID());
     }
+  } else
+    BuildPcoupled(A, aggregates, amalgInfo, fineNullspace, coarseMap, Ptentative, coarseNullspace);
+
+  // If available, use striding information of fine level matrix A for range
+  // map and coarseMap as domain map; otherwise use plain range map of
+  // Ptent = plain range map of A for range map and coarseMap as domain map.
+  // NOTE:
+  // The latter is not really safe, since there is no striding information
+  // for the range map. This is not really a problem, since striding
+  // information is always available on the intermedium levels and the
+  // coarsest levels.
+  if (A->IsView("stridedMaps") == true)
+    Ptentative->CreateView("stridedMaps", A->getRowMap("stridedMaps"), coarseMap);
+  else
+    Ptentative->CreateView("stridedMaps", Ptentative->getRangeMap(), coarseMap);
+
+  if (bTransferCoordinates_) {
+    Set(coarseLevel, "Coordinates", coarseCoords);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  BuildPuncoupled(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates,
-                  RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                  RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
-                  RCP<MultiVector>& coarseNullspace, const int levelID) const {
-    auto rowMap = A->getRowMap();
-    auto colMap = A->getColMap();
-
-    const size_t numRows  = rowMap->getLocalNumElements();
-    const size_t NSDim    = fineNullspace->getNumVectors();
-
-    typedef Kokkos::ArithTraits<SC>     ATS;
-    using impl_SC = typename ATS::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_SC>;
-    const impl_SC zero = impl_ATS::zero(), one = impl_ATS::one();
-
-    const LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+  // FIXME: We should remove the NodeComm on levels past the threshold
+  if (fineLevel.IsAvailable("Node Comm")) {
+    RCP<const Teuchos::Comm<int>> nodeComm = Get<RCP<const Teuchos::Comm<int>>>(fineLevel, "Node Comm");
+    Set<RCP<const Teuchos::Comm<int>>>(coarseLevel, "Node Comm", nodeComm);
+  }
 
-    typename Aggregates::local_graph_type aggGraph;
-    {
-      SubFactoryMonitor m2(*this, "Get Aggregates graph", coarseLevel);
-      aggGraph = aggregates->GetGraph();
-    }
-    auto aggRows  = aggGraph.row_map;
-    auto aggCols  = aggGraph.entries;
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+  Set(coarseLevel, "P", Ptentative);
 
-    // Aggregates map is based on the amalgamated column map
-    // We can skip global-to-local conversion if LIDs in row map are
-    // same as LIDs in column map
-    bool goodMap;
-    {
-      SubFactoryMonitor m2(*this, "Check good map", coarseLevel);
-      goodMap = isGoodMap(*rowMap, *colMap);
-    }
-    // FIXME_KOKKOS: need to proofread later code for bad maps
-    TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError,
-        "MueLu: TentativePFactory_kokkos: for now works only with good maps "
-        "(i.e. \"matching\" row and column maps)");
+  if (IsPrint(Statistics2)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    GetOStream(Statistics2) << PerfUtils::PrintMatrixInfo(*Ptentative, "Ptent", params);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    BuildPuncoupled(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates,
+                    RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                    RCP<const Map> coarseMap, RCP<Matrix>& Ptentative,
+                    RCP<MultiVector>& coarseNullspace, const int levelID) const {
+  auto rowMap = A->getRowMap();
+  auto colMap = A->getColMap();
+
+  const size_t numRows = rowMap->getLocalNumElements();
+  const size_t NSDim   = fineNullspace->getNumVectors();
+
+  typedef Kokkos::ArithTraits<SC> ATS;
+  using impl_SC      = typename ATS::val_type;
+  using impl_ATS     = Kokkos::ArithTraits<impl_SC>;
+  const impl_SC zero = impl_ATS::zero(), one = impl_ATS::one();
+
+  const LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  typename Aggregates::local_graph_type aggGraph;
+  {
+    SubFactoryMonitor m2(*this, "Get Aggregates graph", coarseLevel);
+    aggGraph = aggregates->GetGraph();
+  }
+  auto aggRows = aggGraph.row_map;
+  auto aggCols = aggGraph.entries;
+
+  // Aggregates map is based on the amalgamated column map
+  // We can skip global-to-local conversion if LIDs in row map are
+  // same as LIDs in column map
+  bool goodMap;
+  {
+    SubFactoryMonitor m2(*this, "Check good map", coarseLevel);
+    goodMap = isGoodMap(*rowMap, *colMap);
+  }
+  // FIXME_KOKKOS: need to proofread later code for bad maps
+  TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError,
+                             "MueLu: TentativePFactory_kokkos: for now works only with good maps "
+                             "(i.e. \"matching\" row and column maps)");
 
-    // STEP 1: do unamalgamation
-    // The non-kokkos version uses member functions from the AmalgamationInfo
-    // container class to unamalgamate the data. In contrast, the kokkos
-    // version of TentativePFactory does the unamalgamation here and only uses
-    // the data of the AmalgamationInfo container class
+  // STEP 1: do unamalgamation
+  // The non-kokkos version uses member functions from the AmalgamationInfo
+  // container class to unamalgamate the data. In contrast, the kokkos
+  // version of TentativePFactory does the unamalgamation here and only uses
+  // the data of the AmalgamationInfo container class
 
-    // Extract information for unamalgamation
-    LO fullBlockSize, blockID, stridingOffset, stridedBlockSize;
-    GO indexBase;
-    amalgInfo->GetStridingInformation(fullBlockSize, blockID, stridingOffset, stridedBlockSize, indexBase);
-    GO globalOffset = amalgInfo->GlobalOffset();
+  // Extract information for unamalgamation
+  LO fullBlockSize, blockID, stridingOffset, stridedBlockSize;
+  GO indexBase;
+  amalgInfo->GetStridingInformation(fullBlockSize, blockID, stridingOffset, stridedBlockSize, indexBase);
+  GO globalOffset = amalgInfo->GlobalOffset();
 
-    // Extract aggregation info (already in Kokkos host views)
-    auto         procWinner    = aggregates->GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto         vertex2AggId  = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    const size_t numAggregates = aggregates->GetNumAggregates();
+  // Extract aggregation info (already in Kokkos host views)
+  auto procWinner            = aggregates->GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto vertex2AggId          = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  const size_t numAggregates = aggregates->GetNumAggregates();
 
-    int myPID = aggregates->GetMap()->getComm()->getRank();
+  int myPID = aggregates->GetMap()->getComm()->getRank();
 
-    // Create Kokkos::View (on the device) to store the aggreate dof sizes
-    // Later used to get aggregate dof offsets
-    // NOTE: This zeros itself on construction
-    typedef typename Aggregates::aggregates_sizes_type::non_const_type AggSizeType;
-    AggSizeType aggDofSizes;
+  // Create Kokkos::View (on the device) to store the aggreate dof sizes
+  // Later used to get aggregate dof offsets
+  // NOTE: This zeros itself on construction
+  typedef typename Aggregates::aggregates_sizes_type::non_const_type AggSizeType;
+  AggSizeType aggDofSizes;
 
-    if (stridedBlockSize == 1) {
-      SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
+  if (stridedBlockSize == 1) {
+    SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
 
-      // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
-      aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates+1);
+    // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
+    aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates + 1);
 
-      auto sizesConst = aggregates->ComputeAggregateSizes();
-      Kokkos::deep_copy(Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(1), numAggregates+1)), sizesConst);
+    auto sizesConst = aggregates->ComputeAggregateSizes();
+    Kokkos::deep_copy(Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(1), numAggregates + 1)), sizesConst);
 
-    } else {
-      SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
+  } else {
+    SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
 
-      // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
-      aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates + 1);
+    // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
+    aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates + 1);
 
-      auto nodeMap = aggregates->GetMap()->getLocalMap();
-      auto dofMap  = colMap->getLocalMap();
+    auto nodeMap = aggregates->GetMap()->getLocalMap();
+    auto dofMap  = colMap->getLocalMap();
 
-      Kokkos::parallel_for("MueLu:TentativePF:Build:compute_agg_sizes", range_type(0,numAggregates),
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:Build:compute_agg_sizes", range_type(0, numAggregates),
         KOKKOS_LAMBDA(const LO agg) {
           auto aggRowView = aggGraph.rowConst(agg);
 
@@ -683,123 +681,127 @@ namespace MueLu {
                 size++;
             }
           }
-          aggDofSizes(agg+1) = size;
+          aggDofSizes(agg + 1) = size;
         });
-    }
+  }
 
-    // Find maximum dof size for aggregates
-    // Later used to reserve enough scratch space for local QR decompositions
-    LO maxAggSize = 0;
-    ReduceMaxFunctor<LO,decltype(aggDofSizes)> reduceMax(aggDofSizes);
-    Kokkos::parallel_reduce("MueLu:TentativePF:Build:max_agg_size", range_type(0, aggDofSizes.extent(0)), reduceMax, maxAggSize);
+  // Find maximum dof size for aggregates
+  // Later used to reserve enough scratch space for local QR decompositions
+  LO maxAggSize = 0;
+  ReduceMaxFunctor<LO, decltype(aggDofSizes)> reduceMax(aggDofSizes);
+  Kokkos::parallel_reduce("MueLu:TentativePF:Build:max_agg_size", range_type(0, aggDofSizes.extent(0)), reduceMax, maxAggSize);
 
-    // parallel_scan (exclusive)
-    // The aggDofSizes View then contains the aggregate dof offsets
-    Kokkos::parallel_scan("MueLu:TentativePF:Build:aggregate_sizes:stage1_scan", range_type(0,numAggregates+1),
+  // parallel_scan (exclusive)
+  // The aggDofSizes View then contains the aggregate dof offsets
+  Kokkos::parallel_scan(
+      "MueLu:TentativePF:Build:aggregate_sizes:stage1_scan", range_type(0, numAggregates + 1),
       KOKKOS_LAMBDA(const LO i, LO& update, const bool& final_pass) {
         update += aggDofSizes(i);
         if (final_pass)
           aggDofSizes(i) = update;
       });
 
-    // Create Kokkos::View on the device to store mapping
-    // between (local) aggregate id and row map ids (LIDs)
-    Kokkos::View<LO*, DeviceType> agg2RowMapLO(Kokkos::ViewAllocateWithoutInitializing("agg2row_map_LO"), numRows);
-    {
-      SubFactoryMonitor m2(*this, "Create Agg2RowMap", coarseLevel);
+  // Create Kokkos::View on the device to store mapping
+  // between (local) aggregate id and row map ids (LIDs)
+  Kokkos::View<LO*, DeviceType> agg2RowMapLO(Kokkos::ViewAllocateWithoutInitializing("agg2row_map_LO"), numRows);
+  {
+    SubFactoryMonitor m2(*this, "Create Agg2RowMap", coarseLevel);
 
-      AggSizeType aggOffsets(Kokkos::ViewAllocateWithoutInitializing("aggOffsets"), numAggregates);
-      Kokkos::deep_copy(aggOffsets, Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(0), numAggregates)));
+    AggSizeType aggOffsets(Kokkos::ViewAllocateWithoutInitializing("aggOffsets"), numAggregates);
+    Kokkos::deep_copy(aggOffsets, Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(0), numAggregates)));
 
-      Kokkos::parallel_for("MueLu:TentativePF:Build:createAgg2RowMap", range_type(0, vertex2AggId.extent(0)),
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:Build:createAgg2RowMap", range_type(0, vertex2AggId.extent(0)),
         KOKKOS_LAMBDA(const LO lnode) {
           if (procWinner(lnode, 0) == myPID) {
             // No need for atomics, it's one-to-one
-            auto aggID = vertex2AggId(lnode,0);
+            auto aggID = vertex2AggId(lnode, 0);
 
-            auto offset = Kokkos::atomic_fetch_add( &aggOffsets(aggID), stridedBlockSize );
+            auto offset = Kokkos::atomic_fetch_add(&aggOffsets(aggID), stridedBlockSize);
             // FIXME: I think this may be wrong
             // We unconditionally add the whole block here. When we calculated
             // aggDofSizes, we did the isLocalElement check. Something's fishy.
             for (LO k = 0; k < stridedBlockSize; k++)
-              agg2RowMapLO(offset + k) = lnode*stridedBlockSize + k;
+              agg2RowMapLO(offset + k) = lnode * stridedBlockSize + k;
           }
         });
-    }
-
-    // STEP 2: prepare local QR decomposition
-    // Reserve memory for tentative prolongation operator
-    coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
-
-    // Pull out the nullspace vectors so that we can have random access (on the device)
-    auto fineNS   = fineNullspace  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto coarseNS = coarseNullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+  }
 
-    size_t nnz = 0;                       // actual number of nnz
+  // STEP 2: prepare local QR decomposition
+  // Reserve memory for tentative prolongation operator
+  coarseNullspace = MultiVectorFactory::Build(coarseMap, NSDim);
 
-    typedef typename Xpetra::Matrix<SC,LO,GO,NO>::local_matrix_type    local_matrix_type;
-    typedef typename local_matrix_type::row_map_type::non_const_type   rows_type;
-    typedef typename local_matrix_type::index_type::non_const_type     cols_type;
-    typedef typename local_matrix_type::values_type::non_const_type    vals_type;
+  // Pull out the nullspace vectors so that we can have random access (on the device)
+  auto fineNS   = fineNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto coarseNS = coarseNullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
 
+  size_t nnz = 0;  // actual number of nnz
 
-    // Device View for status (error messages...)
-    typedef Kokkos::View<int[10], DeviceType> status_type;
-    status_type status("status");
+  typedef typename Xpetra::Matrix<SC, LO, GO, NO>::local_matrix_type local_matrix_type;
+  typedef typename local_matrix_type::row_map_type::non_const_type rows_type;
+  typedef typename local_matrix_type::index_type::non_const_type cols_type;
+  typedef typename local_matrix_type::values_type::non_const_type vals_type;
 
-    typename AppendTrait<decltype(fineNS), Kokkos::RandomAccess>::type fineNSRandom = fineNS;
-    typename AppendTrait<status_type,      Kokkos::Atomic>      ::type statusAtomic = status;
+  // Device View for status (error messages...)
+  typedef Kokkos::View<int[10], DeviceType> status_type;
+  status_type status("status");
 
-    const ParameterList& pL = GetParameterList();
-    const bool& doQRStep = pL.get<bool>("tentative: calculate qr");
-    if (!doQRStep) {
-      GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
-      if (NSDim>1)
-        GetOStream(Warnings0) << "TentativePFactor : for nontrivial nullspace, this may degrade performance" << std::endl;
-    }
+  typename AppendTrait<decltype(fineNS), Kokkos::RandomAccess>::type fineNSRandom = fineNS;
+  typename AppendTrait<status_type, Kokkos::Atomic>::type statusAtomic            = status;
 
-    size_t nnzEstimate = numRows * NSDim;
-    rows_type rowsAux(Kokkos::ViewAllocateWithoutInitializing("Ptent_aux_rows"), numRows+1);
-    cols_type colsAux(Kokkos::ViewAllocateWithoutInitializing("Ptent_aux_cols"), nnzEstimate);
-    vals_type valsAux("Ptent_aux_vals", nnzEstimate);
-    rows_type rows("Ptent_rows", numRows+1);
-    {
-      // Stage 0: fill in views.
-      SubFactoryMonitor m2(*this, "Stage 0 (InitViews)", coarseLevel);
-
-      // The main thing to notice is initialization of vals with INVALID. These
-      // values will later be used to compress the arrays
-      Kokkos::parallel_for("MueLu:TentativePF:BuildPuncoupled:for1", range_type(0, numRows+1),
-                           KOKKOS_LAMBDA(const LO row) {
-                             rowsAux(row) = row*NSDim;
-                           });
-      Kokkos::parallel_for("MueLu:TentativePF:BuildUncoupled:for2", range_type(0, nnzEstimate),
-                           KOKKOS_LAMBDA(const LO j) {
-                             colsAux(j) = INVALID;
-                           });
-    }
+  const ParameterList& pL = GetParameterList();
+  const bool& doQRStep    = pL.get<bool>("tentative: calculate qr");
+  if (!doQRStep) {
+    GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
+    if (NSDim > 1)
+      GetOStream(Warnings0) << "TentativePFactor : for nontrivial nullspace, this may degrade performance" << std::endl;
+  }
 
-    if (NSDim == 1) {
-      // 1D is special, as it is the easiest. We don't even need to the QR,
-      // just normalize an array. Plus, no worries abot small aggregates.  In
-      // addition, we do not worry about compression. It is unlikely that
-      // nullspace will have zeros. If it does, a prolongator row would be
-      // zero and we'll get singularity anyway.
-      SubFactoryMonitor m2(*this, "Stage 1 (LocalQR)", coarseLevel);
+  size_t nnzEstimate = numRows * NSDim;
+  rows_type rowsAux(Kokkos::ViewAllocateWithoutInitializing("Ptent_aux_rows"), numRows + 1);
+  cols_type colsAux(Kokkos::ViewAllocateWithoutInitializing("Ptent_aux_cols"), nnzEstimate);
+  vals_type valsAux("Ptent_aux_vals", nnzEstimate);
+  rows_type rows("Ptent_rows", numRows + 1);
+  {
+    // Stage 0: fill in views.
+    SubFactoryMonitor m2(*this, "Stage 0 (InitViews)", coarseLevel);
+
+    // The main thing to notice is initialization of vals with INVALID. These
+    // values will later be used to compress the arrays
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:BuildPuncoupled:for1", range_type(0, numRows + 1),
+        KOKKOS_LAMBDA(const LO row) {
+          rowsAux(row) = row * NSDim;
+        });
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:BuildUncoupled:for2", range_type(0, nnzEstimate),
+        KOKKOS_LAMBDA(const LO j) {
+          colsAux(j) = INVALID;
+        });
+  }
 
-      // Set up team policy with numAggregates teams and one thread per team.
-      // Each team handles a slice of the data associated with one aggregate
-      // and performs a local QR decomposition (in this case real QR is
-      // unnecessary).
-      const Kokkos::TeamPolicy<execution_space> policy(numAggregates, 1);
+  if (NSDim == 1) {
+    // 1D is special, as it is the easiest. We don't even need to the QR,
+    // just normalize an array. Plus, no worries abot small aggregates.  In
+    // addition, we do not worry about compression. It is unlikely that
+    // nullspace will have zeros. If it does, a prolongator row would be
+    // zero and we'll get singularity anyway.
+    SubFactoryMonitor m2(*this, "Stage 1 (LocalQR)", coarseLevel);
+
+    // Set up team policy with numAggregates teams and one thread per team.
+    // Each team handles a slice of the data associated with one aggregate
+    // and performs a local QR decomposition (in this case real QR is
+    // unnecessary).
+    const Kokkos::TeamPolicy<execution_space> policy(numAggregates, 1);
 
-      if (doQRStep) {
-        Kokkos::parallel_for("MueLu:TentativePF:BuildUncoupled:main_loop", policy,
-          KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type &thread) {
+    if (doQRStep) {
+      Kokkos::parallel_for(
+          "MueLu:TentativePF:BuildUncoupled:main_loop", policy,
+          KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread) {
             auto agg = thread.league_rank();
 
             // size of the aggregate (number of DOFs in aggregate)
-            LO aggSize = aggRows(agg+1) - aggRows(agg);
+            LO aggSize = aggRows(agg + 1) - aggRows(agg);
 
             // Extract the piece of the nullspace corresponding to the aggregate, and
             // put it in the flat array, "localQR" (in column major format) for the
@@ -810,8 +812,8 @@ namespace MueLu {
             // FIXME: shouldn't there be stridedblock here?
             // FIXME_KOKKOS: shouldn't there be stridedblock here?
             for (decltype(aggSize) k = 0; k < aggSize; k++) {
-              auto dnorm = impl_ATS::magnitude(fineNSRandom(agg2RowMapLO(aggRows(agg)+k),0));
-              norm += dnorm*dnorm;
+              auto dnorm = impl_ATS::magnitude(fineNSRandom(agg2RowMapLO(aggRows(agg) + k), 0));
+              norm += dnorm * dnorm;
             }
             norm = sqrt(norm);
 
@@ -826,176 +828,178 @@ namespace MueLu {
 
             // Q = localQR(:,0)/norm
             for (decltype(aggSize) k = 0; k < aggSize; k++) {
-              LO localRow = agg2RowMapLO(aggRows(agg)+k);
-              impl_SC localVal = fineNSRandom(agg2RowMapLO(aggRows(agg)+k),0) / norm;
-
-              rows(localRow+1) = 1;
-              colsAux(localRow) = agg;
-              valsAux(localRow) = localVal;
+              LO localRow      = agg2RowMapLO(aggRows(agg) + k);
+              impl_SC localVal = fineNSRandom(agg2RowMapLO(aggRows(agg) + k), 0) / norm;
 
+              rows(localRow + 1) = 1;
+              colsAux(localRow)  = agg;
+              valsAux(localRow)  = localVal;
             }
           });
 
-        typename status_type::HostMirror statusHost = Kokkos::create_mirror_view(status);
-        Kokkos::deep_copy(statusHost, status);
-        for (decltype(statusHost.size()) i = 0; i < statusHost.size(); i++)
-          if (statusHost(i)) {
-            std::ostringstream oss;
-            oss << "MueLu::TentativePFactory::MakeTentative: ";
-            switch (i) {
-              case 0: oss << "!goodMap is not implemented";               break;
-              case 1: oss << "fine level NS part has a zero column";      break;
-            }
-            throw Exceptions::RuntimeError(oss.str());
+      typename status_type::HostMirror statusHost = Kokkos::create_mirror_view(status);
+      Kokkos::deep_copy(statusHost, status);
+      for (decltype(statusHost.size()) i = 0; i < statusHost.size(); i++)
+        if (statusHost(i)) {
+          std::ostringstream oss;
+          oss << "MueLu::TentativePFactory::MakeTentative: ";
+          switch (i) {
+            case 0: oss << "!goodMap is not implemented"; break;
+            case 1: oss << "fine level NS part has a zero column"; break;
           }
+          throw Exceptions::RuntimeError(oss.str());
+        }
 
-      } else {
-        Kokkos::parallel_for("MueLu:TentativePF:BuildUncoupled:main_loop_noqr", policy,
-          KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type &thread) {
+    } else {
+      Kokkos::parallel_for(
+          "MueLu:TentativePF:BuildUncoupled:main_loop_noqr", policy,
+          KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread) {
             auto agg = thread.league_rank();
 
             // size of the aggregate (number of DOFs in aggregate)
-            LO aggSize = aggRows(agg+1) - aggRows(agg);
+            LO aggSize = aggRows(agg + 1) - aggRows(agg);
 
             // R = norm
             coarseNS(agg, 0) = one;
 
             // Q = localQR(:,0)/norm
             for (decltype(aggSize) k = 0; k < aggSize; k++) {
-              LO localRow = agg2RowMapLO(aggRows(agg)+k);
-              impl_SC localVal = fineNSRandom(agg2RowMapLO(aggRows(agg)+k),0);
-
-              rows(localRow+1) = 1;
-              colsAux(localRow) = agg;
-              valsAux(localRow) = localVal;
+              LO localRow      = agg2RowMapLO(aggRows(agg) + k);
+              impl_SC localVal = fineNSRandom(agg2RowMapLO(aggRows(agg) + k), 0);
 
+              rows(localRow + 1) = 1;
+              colsAux(localRow)  = agg;
+              valsAux(localRow)  = localVal;
             }
           });
-      }
+    }
 
-      Kokkos::parallel_reduce("MueLu:TentativeP:CountNNZ", range_type(0, numRows+1),
-                              KOKKOS_LAMBDA(const LO i, size_t &nnz_count) {
-                                nnz_count += rows(i);
-                              }, nnz);
-
-    } else { // NSdim > 1
-      // FIXME_KOKKOS: This code branch is completely unoptimized.
-      // Work to do:
-      //   - Optimize QR decomposition
-      //   - Remove INVALID usage similarly to CoalesceDropFactory_kokkos by
-      //     packing new values in the beginning of each row
-      // We do use auxilary view in this case, so keep a second rows view for
-      // counting nonzeros in rows
-
-      {
-        SubFactoryMonitor m2 = SubFactoryMonitor(*this, doQRStep ? "Stage 1 (LocalQR)" : "Stage 1 (Fill coarse nullspace and tentative P)", coarseLevel);
-        // Set up team policy with numAggregates teams and one thread per team.
-        // Each team handles a slice of the data associated with one aggregate
-        // and performs a local QR decomposition
-        const Kokkos::TeamPolicy<execution_space> policy(numAggregates,1); // numAggregates teams a 1 thread
-        LocalQRDecompFunctor<LocalOrdinal, GlobalOrdinal, Scalar, DeviceType, decltype(fineNSRandom),
-            decltype(aggDofSizes /*aggregate sizes in dofs*/), decltype(maxAggSize), decltype(agg2RowMapLO),
-            decltype(statusAtomic), decltype(rows), decltype(rowsAux), decltype(colsAux),
-            decltype(valsAux)>
-                localQRFunctor(fineNSRandom, coarseNS, aggDofSizes, maxAggSize, agg2RowMapLO, statusAtomic,
-                               rows, rowsAux, colsAux, valsAux, doQRStep);
-        Kokkos::parallel_reduce("MueLu:TentativePF:BuildUncoupled:main_qr_loop", policy, localQRFunctor, nnz);
-      }
+    Kokkos::parallel_reduce(
+        "MueLu:TentativeP:CountNNZ", range_type(0, numRows + 1),
+        KOKKOS_LAMBDA(const LO i, size_t& nnz_count) {
+          nnz_count += rows(i);
+        },
+        nnz);
+
+  } else {  // NSdim > 1
+    // FIXME_KOKKOS: This code branch is completely unoptimized.
+    // Work to do:
+    //   - Optimize QR decomposition
+    //   - Remove INVALID usage similarly to CoalesceDropFactory_kokkos by
+    //     packing new values in the beginning of each row
+    // We do use auxilary view in this case, so keep a second rows view for
+    // counting nonzeros in rows
 
-      typename status_type::HostMirror statusHost = Kokkos::create_mirror_view(status);
-      Kokkos::deep_copy(statusHost, status);
-      for (decltype(statusHost.size()) i = 0; i < statusHost.size(); i++)
-        if (statusHost(i)) {
-          std::ostringstream oss;
-          oss << "MueLu::TentativePFactory::MakeTentative: ";
-          switch(i) {
-            case 0: oss << "!goodMap is not implemented";               break;
-            case 1: oss << "fine level NS part has a zero column";      break;
-          }
-          throw Exceptions::RuntimeError(oss.str());
-        }
+    {
+      SubFactoryMonitor m2 = SubFactoryMonitor(*this, doQRStep ? "Stage 1 (LocalQR)" : "Stage 1 (Fill coarse nullspace and tentative P)", coarseLevel);
+      // Set up team policy with numAggregates teams and one thread per team.
+      // Each team handles a slice of the data associated with one aggregate
+      // and performs a local QR decomposition
+      const Kokkos::TeamPolicy<execution_space> policy(numAggregates, 1);  // numAggregates teams a 1 thread
+      LocalQRDecompFunctor<LocalOrdinal, GlobalOrdinal, Scalar, DeviceType, decltype(fineNSRandom),
+                           decltype(aggDofSizes /*aggregate sizes in dofs*/), decltype(maxAggSize), decltype(agg2RowMapLO),
+                           decltype(statusAtomic), decltype(rows), decltype(rowsAux), decltype(colsAux),
+                           decltype(valsAux)>
+          localQRFunctor(fineNSRandom, coarseNS, aggDofSizes, maxAggSize, agg2RowMapLO, statusAtomic,
+                         rows, rowsAux, colsAux, valsAux, doQRStep);
+      Kokkos::parallel_reduce("MueLu:TentativePF:BuildUncoupled:main_qr_loop", policy, localQRFunctor, nnz);
     }
 
-    // Compress the cols and vals by ignoring INVALID column entries that correspond
-    // to 0 in QR.
+    typename status_type::HostMirror statusHost = Kokkos::create_mirror_view(status);
+    Kokkos::deep_copy(statusHost, status);
+    for (decltype(statusHost.size()) i = 0; i < statusHost.size(); i++)
+      if (statusHost(i)) {
+        std::ostringstream oss;
+        oss << "MueLu::TentativePFactory::MakeTentative: ";
+        switch (i) {
+          case 0: oss << "!goodMap is not implemented"; break;
+          case 1: oss << "fine level NS part has a zero column"; break;
+        }
+        throw Exceptions::RuntimeError(oss.str());
+      }
+  }
 
-    // The real cols and vals are constructed using calculated (not estimated) nnz
-    cols_type cols;
-    vals_type vals;
+  // Compress the cols and vals by ignoring INVALID column entries that correspond
+  // to 0 in QR.
 
-    if (nnz != nnzEstimate) {
-      {
-        // Stage 2: compress the arrays
-        SubFactoryMonitor m2(*this, "Stage 2 (CompressRows)", coarseLevel);
+  // The real cols and vals are constructed using calculated (not estimated) nnz
+  cols_type cols;
+  vals_type vals;
 
-        Kokkos::parallel_scan("MueLu:TentativePF:Build:compress_rows", range_type(0,numRows+1),
-                              KOKKOS_LAMBDA(const LO i, LO& upd, const bool& final) {
-                                upd += rows(i);
-                                if (final)
-                                  rows(i) = upd;
-                              });
-      }
+  if (nnz != nnzEstimate) {
+    {
+      // Stage 2: compress the arrays
+      SubFactoryMonitor m2(*this, "Stage 2 (CompressRows)", coarseLevel);
+
+      Kokkos::parallel_scan(
+          "MueLu:TentativePF:Build:compress_rows", range_type(0, numRows + 1),
+          KOKKOS_LAMBDA(const LO i, LO& upd, const bool& final) {
+            upd += rows(i);
+            if (final)
+              rows(i) = upd;
+          });
+    }
 
-      {
-        SubFactoryMonitor m2(*this, "Stage 2 (CompressCols)", coarseLevel);
-
-        cols = cols_type("Ptent_cols", nnz);
-        vals = vals_type("Ptent_vals", nnz);
-
-        // FIXME_KOKKOS: this can be spedup by moving correct cols and vals values
-        // to the beginning of rows. See CoalesceDropFactory_kokkos for
-        // example.
-        Kokkos::parallel_for("MueLu:TentativePF:Build:compress_cols_vals", range_type(0,numRows),
-                             KOKKOS_LAMBDA(const LO i) {
-                               LO rowStart = rows(i);
-
-                               size_t lnnz = 0;
-                               for (auto j = rowsAux(i); j < rowsAux(i+1); j++)
-                                 if (colsAux(j) != INVALID) {
-                                   cols(rowStart+lnnz) = colsAux(j);
-                                   vals(rowStart+lnnz) = valsAux(j);
-                                   lnnz++;
-                                 }
-                             });
-      }
+    {
+      SubFactoryMonitor m2(*this, "Stage 2 (CompressCols)", coarseLevel);
 
-    } else {
-      rows = rowsAux;
-      cols = colsAux;
-      vals = valsAux;
+      cols = cols_type("Ptent_cols", nnz);
+      vals = vals_type("Ptent_vals", nnz);
+
+      // FIXME_KOKKOS: this can be spedup by moving correct cols and vals values
+      // to the beginning of rows. See CoalesceDropFactory_kokkos for
+      // example.
+      Kokkos::parallel_for(
+          "MueLu:TentativePF:Build:compress_cols_vals", range_type(0, numRows),
+          KOKKOS_LAMBDA(const LO i) {
+            LO rowStart = rows(i);
+
+            size_t lnnz = 0;
+            for (auto j = rowsAux(i); j < rowsAux(i + 1); j++)
+              if (colsAux(j) != INVALID) {
+                cols(rowStart + lnnz) = colsAux(j);
+                vals(rowStart + lnnz) = valsAux(j);
+                lnnz++;
+              }
+          });
     }
 
-    GetOStream(Runtime1) << "TentativePFactory : aggregates do not cross process boundaries" << std::endl;
+  } else {
+    rows = rowsAux;
+    cols = colsAux;
+    vals = valsAux;
+  }
 
-    {
-      // Stage 3: construct Xpetra::Matrix
-      SubFactoryMonitor m2(*this, "Stage 3 (LocalMatrix+FillComplete)", coarseLevel);
+  GetOStream(Runtime1) << "TentativePFactory : aggregates do not cross process boundaries" << std::endl;
 
-      local_matrix_type lclMatrix = local_matrix_type("A", numRows, coarseMap->getLocalNumElements(), nnz, vals, rows, cols);
+  {
+    // Stage 3: construct Xpetra::Matrix
+    SubFactoryMonitor m2(*this, "Stage 3 (LocalMatrix+FillComplete)", coarseLevel);
 
-      // Managing labels & constants for ESFC
-      RCP<ParameterList> FCparams;
-      if (pL.isSublist("matrixmatrix: kernel params"))
-        FCparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-      else
-        FCparams = rcp(new ParameterList);
+    local_matrix_type lclMatrix = local_matrix_type("A", numRows, coarseMap->getLocalNumElements(), nnz, vals, rows, cols);
 
-      // By default, we don't need global constants for TentativeP
-      FCparams->set("compute global constants", FCparams->get("compute global constants", false));
-      FCparams->set("Timer Label",              std::string("MueLu::TentativeP-") + toString(levelID));
+    // Managing labels & constants for ESFC
+    RCP<ParameterList> FCparams;
+    if (pL.isSublist("matrixmatrix: kernel params"))
+      FCparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+    else
+      FCparams = rcp(new ParameterList);
 
-      auto PtentCrs = CrsMatrixFactory::Build(lclMatrix, rowMap, coarseMap, coarseMap, A->getDomainMap());
-      Ptentative = rcp(new CrsMatrixWrap(PtentCrs));
-    }
-  }
+    // By default, we don't need global constants for TentativeP
+    FCparams->set("compute global constants", FCparams->get("compute global constants", false));
+    FCparams->set("Timer Label", std::string("MueLu::TentativeP-") + toString(levelID));
 
+    auto PtentCrs = CrsMatrixFactory::Build(lclMatrix, rowMap, coarseMap, coarseMap, A->getDomainMap());
+    Ptentative    = rcp(new CrsMatrixWrap(PtentCrs));
+  }
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
     BuildPuncoupledBlockCrs(Level& coarseLevel, RCP<Matrix> A, RCP<Aggregates> aggregates,
-                  RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
-                  RCP<const Map> coarsePointMap, RCP<Matrix>& Ptentative,
-                  RCP<MultiVector>& coarseNullspace, const int levelID) const {
+                            RCP<AmalgamationInfo> amalgInfo, RCP<MultiVector> fineNullspace,
+                            RCP<const Map> coarsePointMap, RCP<Matrix>& Ptentative,
+                            RCP<MultiVector>& coarseNullspace, const int levelID) const {
   /* This routine generates a BlockCrs P for a BlockCrs A.  There are a few assumptions here, which meet the use cases we care about, but could
        be generalized later, if we ever need to do so:
        1) Null space dimension === block size of matrix:  So no elasticity right now
@@ -1005,323 +1009,326 @@ namespace MueLu {
        These assumptions keep our code way simpler and still support the use cases we actually care about.
      */
 
-    RCP<const Map> rowMap     = A->getRowMap();
-    RCP<const Map> rangeMap   = A->getRangeMap();
-    RCP<const Map> colMap     = A->getColMap();
-    //    const size_t numFinePointRows = rangeMap->getLocalNumElements();
-    const size_t numFineBlockRows = rowMap->getLocalNumElements();
-
-    // typedef Teuchos::ScalarTraits<SC> STS;
-    // typedef typename STS::magnitudeType Magnitude;
-    const LO     INVALID   = Teuchos::OrdinalTraits<LO>::invalid();
-
-    typedef Kokkos::ArithTraits<SC>     ATS;
-    using impl_SC = typename ATS::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_SC>;
-    const impl_SC one = impl_ATS::one();
-
-    //    const GO     numAggs   = aggregates->GetNumAggregates();
-    const size_t NSDim     = fineNullspace->getNumVectors();
-    auto aggSizes          = aggregates->ComputeAggregateSizes();
-
-
-    typename Aggregates::local_graph_type aggGraph;
-    {
-      SubFactoryMonitor m2(*this, "Get Aggregates graph", coarseLevel);
-      aggGraph = aggregates->GetGraph();
-    }
-    auto aggRows  = aggGraph.row_map;
-    auto aggCols  = aggGraph.entries;
-
-
-    // Need to generate the coarse block map
-    // NOTE: We assume NSDim == block size here
-    // NOTE: We also assume that coarseMap has contiguous GIDs
-    //const size_t numCoarsePointRows = coarsePointMap->getLocalNumElements();
-    const size_t numCoarseBlockRows = coarsePointMap->getLocalNumElements() / NSDim;
-    RCP<const Map> coarseBlockMap = MapFactory::Build(coarsePointMap->lib(),
+  RCP<const Map> rowMap   = A->getRowMap();
+  RCP<const Map> rangeMap = A->getRangeMap();
+  RCP<const Map> colMap   = A->getColMap();
+  //    const size_t numFinePointRows = rangeMap->getLocalNumElements();
+  const size_t numFineBlockRows = rowMap->getLocalNumElements();
+
+  // typedef Teuchos::ScalarTraits<SC> STS;
+  // typedef typename STS::magnitudeType Magnitude;
+  const LO INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+
+  typedef Kokkos::ArithTraits<SC> ATS;
+  using impl_SC     = typename ATS::val_type;
+  using impl_ATS    = Kokkos::ArithTraits<impl_SC>;
+  const impl_SC one = impl_ATS::one();
+
+  //    const GO     numAggs   = aggregates->GetNumAggregates();
+  const size_t NSDim = fineNullspace->getNumVectors();
+  auto aggSizes      = aggregates->ComputeAggregateSizes();
+
+  typename Aggregates::local_graph_type aggGraph;
+  {
+    SubFactoryMonitor m2(*this, "Get Aggregates graph", coarseLevel);
+    aggGraph = aggregates->GetGraph();
+  }
+  auto aggRows = aggGraph.row_map;
+  auto aggCols = aggGraph.entries;
+
+  // Need to generate the coarse block map
+  // NOTE: We assume NSDim == block size here
+  // NOTE: We also assume that coarseMap has contiguous GIDs
+  // const size_t numCoarsePointRows = coarsePointMap->getLocalNumElements();
+  const size_t numCoarseBlockRows = coarsePointMap->getLocalNumElements() / NSDim;
+  RCP<const Map> coarseBlockMap   = MapFactory::Build(coarsePointMap->lib(),
                                                       Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
                                                       numCoarseBlockRows,
                                                       coarsePointMap->getIndexBase(),
                                                       coarsePointMap->getComm());
-    // Sanity checking
-    const ParameterList& pL = GetParameterList();
-    //    const bool &doQRStep = pL.get<bool>("tentative: calculate qr");
-
-
-    // The aggregates use the amalgamated column map, which in this case is what we want
-
-    // Aggregates map is based on the amalgamated column map
-    // We can skip global-to-local conversion if LIDs in row map are
-    // same as LIDs in column map
-    bool goodMap = MueLu::Utilities<SC,LO,GO,NO>::MapsAreNested(*rowMap, *colMap);
-    TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError,
-                               "MueLu: TentativePFactory_kokkos: for now works only with good maps "
-                               "(i.e. \"matching\" row and column maps)");
-
-    // STEP 1: do unamalgamation
-    // The non-kokkos version uses member functions from the AmalgamationInfo
-    // container class to unamalgamate the data. In contrast, the kokkos
-    // version of TentativePFactory does the unamalgamation here and only uses
-    // the data of the AmalgamationInfo container class
-
-    // Extract information for unamalgamation
-    LO fullBlockSize, blockID, stridingOffset, stridedBlockSize;
-    GO indexBase;
-    amalgInfo->GetStridingInformation(fullBlockSize, blockID, stridingOffset, stridedBlockSize, indexBase);
-    //GO globalOffset = amalgInfo->GlobalOffset();
-
-    // Extract aggregation info (already in Kokkos host views)
-    auto         procWinner    = aggregates->GetProcWinner()  ->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto         vertex2AggId  = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    const size_t numAggregates = aggregates->GetNumAggregates();
-
-    int myPID = aggregates->GetMap()->getComm()->getRank();
-
-    // Create Kokkos::View (on the device) to store the aggreate dof sizes
-    // Later used to get aggregate dof offsets
-    // NOTE: This zeros itself on construction
-    typedef typename Aggregates::aggregates_sizes_type::non_const_type AggSizeType;
-    AggSizeType aggDofSizes; // This turns into "starts" after the parallel_scan
-
-    {
-      SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
-
-      // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
-      aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates+1);
-
-      Kokkos::deep_copy(Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(1), numAggregates+1)), aggSizes);
-    }
+  // Sanity checking
+  const ParameterList& pL = GetParameterList();
+  //    const bool &doQRStep = pL.get<bool>("tentative: calculate qr");
+
+  // The aggregates use the amalgamated column map, which in this case is what we want
+
+  // Aggregates map is based on the amalgamated column map
+  // We can skip global-to-local conversion if LIDs in row map are
+  // same as LIDs in column map
+  bool goodMap = MueLu::Utilities<SC, LO, GO, NO>::MapsAreNested(*rowMap, *colMap);
+  TEUCHOS_TEST_FOR_EXCEPTION(!goodMap, Exceptions::RuntimeError,
+                             "MueLu: TentativePFactory_kokkos: for now works only with good maps "
+                             "(i.e. \"matching\" row and column maps)");
+
+  // STEP 1: do unamalgamation
+  // The non-kokkos version uses member functions from the AmalgamationInfo
+  // container class to unamalgamate the data. In contrast, the kokkos
+  // version of TentativePFactory does the unamalgamation here and only uses
+  // the data of the AmalgamationInfo container class
+
+  // Extract information for unamalgamation
+  LO fullBlockSize, blockID, stridingOffset, stridedBlockSize;
+  GO indexBase;
+  amalgInfo->GetStridingInformation(fullBlockSize, blockID, stridingOffset, stridedBlockSize, indexBase);
+  // GO globalOffset = amalgInfo->GlobalOffset();
+
+  // Extract aggregation info (already in Kokkos host views)
+  auto procWinner            = aggregates->GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto vertex2AggId          = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  const size_t numAggregates = aggregates->GetNumAggregates();
+
+  int myPID = aggregates->GetMap()->getComm()->getRank();
+
+  // Create Kokkos::View (on the device) to store the aggreate dof sizes
+  // Later used to get aggregate dof offsets
+  // NOTE: This zeros itself on construction
+  typedef typename Aggregates::aggregates_sizes_type::non_const_type AggSizeType;
+  AggSizeType aggDofSizes;  // This turns into "starts" after the parallel_scan
+
+  {
+    SubFactoryMonitor m2(*this, "Calc AggSizes", coarseLevel);
+
+    // FIXME_KOKKOS: use ViewAllocateWithoutInitializing + set a single value
+    aggDofSizes = AggSizeType("agg_dof_sizes", numAggregates + 1);
+
+    Kokkos::deep_copy(Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(1), numAggregates + 1)), aggSizes);
+  }
 
-    // Find maximum dof size for aggregates
-    // Later used to reserve enough scratch space for local QR decompositions
-    LO maxAggSize = 0;
-    ReduceMaxFunctor<LO,decltype(aggDofSizes)> reduceMax(aggDofSizes);
-    Kokkos::parallel_reduce("MueLu:TentativePF:Build:max_agg_size", range_type(0, aggDofSizes.extent(0)), reduceMax, maxAggSize);
+  // Find maximum dof size for aggregates
+  // Later used to reserve enough scratch space for local QR decompositions
+  LO maxAggSize = 0;
+  ReduceMaxFunctor<LO, decltype(aggDofSizes)> reduceMax(aggDofSizes);
+  Kokkos::parallel_reduce("MueLu:TentativePF:Build:max_agg_size", range_type(0, aggDofSizes.extent(0)), reduceMax, maxAggSize);
 
-    // parallel_scan (exclusive)
-    // The aggDofSizes View then contains the aggregate dof offsets
-    Kokkos::parallel_scan("MueLu:TentativePF:Build:aggregate_sizes:stage1_scan", range_type(0,numAggregates+1),
+  // parallel_scan (exclusive)
+  // The aggDofSizes View then contains the aggregate dof offsets
+  Kokkos::parallel_scan(
+      "MueLu:TentativePF:Build:aggregate_sizes:stage1_scan", range_type(0, numAggregates + 1),
       KOKKOS_LAMBDA(const LO i, LO& update, const bool& final_pass) {
         update += aggDofSizes(i);
         if (final_pass)
           aggDofSizes(i) = update;
       });
 
-    // Create Kokkos::View on the device to store mapping
-    // between (local) aggregate id and row map ids (LIDs)
-    Kokkos::View<LO*, DeviceType> aggToRowMapLO(Kokkos::ViewAllocateWithoutInitializing("aggtorow_map_LO"), numFineBlockRows);
-    {
-      SubFactoryMonitor m2(*this, "Create AggToRowMap", coarseLevel);
+  // Create Kokkos::View on the device to store mapping
+  // between (local) aggregate id and row map ids (LIDs)
+  Kokkos::View<LO*, DeviceType> aggToRowMapLO(Kokkos::ViewAllocateWithoutInitializing("aggtorow_map_LO"), numFineBlockRows);
+  {
+    SubFactoryMonitor m2(*this, "Create AggToRowMap", coarseLevel);
 
-      AggSizeType aggOffsets(Kokkos::ViewAllocateWithoutInitializing("aggOffsets"), numAggregates);
-      Kokkos::deep_copy(aggOffsets, Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(0), numAggregates)));
+    AggSizeType aggOffsets(Kokkos::ViewAllocateWithoutInitializing("aggOffsets"), numAggregates);
+    Kokkos::deep_copy(aggOffsets, Kokkos::subview(aggDofSizes, Kokkos::make_pair(static_cast<size_t>(0), numAggregates)));
 
-      Kokkos::parallel_for("MueLu:TentativePF:Build:createAgg2RowMap", range_type(0, vertex2AggId.extent(0)),
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:Build:createAgg2RowMap", range_type(0, vertex2AggId.extent(0)),
         KOKKOS_LAMBDA(const LO lnode) {
           if (procWinner(lnode, 0) == myPID) {
             // No need for atomics, it's one-to-one
-            auto aggID = vertex2AggId(lnode,0);
+            auto aggID = vertex2AggId(lnode, 0);
 
-            auto offset = Kokkos::atomic_fetch_add( &aggOffsets(aggID), stridedBlockSize );
+            auto offset = Kokkos::atomic_fetch_add(&aggOffsets(aggID), stridedBlockSize);
             // FIXME: I think this may be wrong
             // We unconditionally add the whole block here. When we calculated
             // aggDofSizes, we did the isLocalElement check. Something's fishy.
             for (LO k = 0; k < stridedBlockSize; k++)
-              aggToRowMapLO(offset + k) = lnode*stridedBlockSize + k;
+              aggToRowMapLO(offset + k) = lnode * stridedBlockSize + k;
           }
         });
-    }
-
-    // STEP 2: prepare local QR decomposition
-    // Reserve memory for tentative prolongation operator
-    coarseNullspace = MultiVectorFactory::Build(coarsePointMap, NSDim);
+  }
 
-    // Pull out the nullspace vectors so that we can have random access (on the device)
-    auto fineNS   = fineNullspace  ->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto coarseNS = coarseNullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+  // STEP 2: prepare local QR decomposition
+  // Reserve memory for tentative prolongation operator
+  coarseNullspace = MultiVectorFactory::Build(coarsePointMap, NSDim);
 
-    typedef typename Xpetra::Matrix<SC,LO,GO,NO>::local_matrix_type    local_matrix_type;
-    typedef typename local_matrix_type::row_map_type::non_const_type   rows_type;
-    typedef typename local_matrix_type::index_type::non_const_type     cols_type;
-    // typedef typename local_matrix_type::values_type::non_const_type    vals_type;
+  // Pull out the nullspace vectors so that we can have random access (on the device)
+  auto fineNS   = fineNullspace->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto coarseNS = coarseNullspace->getDeviceLocalView(Xpetra::Access::OverwriteAll);
 
+  typedef typename Xpetra::Matrix<SC, LO, GO, NO>::local_matrix_type local_matrix_type;
+  typedef typename local_matrix_type::row_map_type::non_const_type rows_type;
+  typedef typename local_matrix_type::index_type::non_const_type cols_type;
+  // typedef typename local_matrix_type::values_type::non_const_type    vals_type;
 
-    // Device View for status (error messages...)
-    typedef Kokkos::View<int[10], DeviceType> status_type;
-    status_type status("status");
+  // Device View for status (error messages...)
+  typedef Kokkos::View<int[10], DeviceType> status_type;
+  status_type status("status");
 
-    typename AppendTrait<decltype(fineNS), Kokkos::RandomAccess>::type fineNSRandom = fineNS;
-    typename AppendTrait<status_type,      Kokkos::Atomic>      ::type statusAtomic = status;
+  typename AppendTrait<decltype(fineNS), Kokkos::RandomAccess>::type fineNSRandom = fineNS;
+  typename AppendTrait<status_type, Kokkos::Atomic>::type statusAtomic            = status;
 
-    // We're going to bypass QR in the BlockCrs version of the code regardless of what the user asks for
-    GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
+  // We're going to bypass QR in the BlockCrs version of the code regardless of what the user asks for
+  GetOStream(Runtime1) << "TentativePFactory : bypassing local QR phase" << std::endl;
 
-    // BlockCrs requires that we build the (block) graph first, so let's do that...
+  // BlockCrs requires that we build the (block) graph first, so let's do that...
 
-    // NOTE: Because we're assuming that the NSDim == BlockSize, we only have one
-    // block non-zero per row in the matrix;
-    rows_type ia(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_rowptr"), numFineBlockRows+1);
-    cols_type ja(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_colind"), numFineBlockRows);
+  // NOTE: Because we're assuming that the NSDim == BlockSize, we only have one
+  // block non-zero per row in the matrix;
+  rows_type ia(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_rowptr"), numFineBlockRows + 1);
+  cols_type ja(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_colind"), numFineBlockRows);
 
-    Kokkos::parallel_for("MueLu:TentativePF:BlockCrs:graph_init", range_type(0, numFineBlockRows),
-                         KOKKOS_LAMBDA(const LO j) {
-                           ia[j] = j;
-                           ja[j] = INVALID;
+  Kokkos::parallel_for(
+      "MueLu:TentativePF:BlockCrs:graph_init", range_type(0, numFineBlockRows),
+      KOKKOS_LAMBDA(const LO j) {
+        ia[j] = j;
+        ja[j] = INVALID;
 
-                           if(j==(LO)numFineBlockRows-1)
-                             ia[numFineBlockRows] = numFineBlockRows;
-                         });
+        if (j == (LO)numFineBlockRows - 1)
+          ia[numFineBlockRows] = numFineBlockRows;
+      });
 
-    // Fill Graph
-    const Kokkos::TeamPolicy<execution_space> policy(numAggregates, 1);
-    Kokkos::parallel_for("MueLu:TentativePF:BlockCrs:fillGraph", policy,
-                         KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type &thread) {
-                           auto agg = thread.league_rank();
-                           Xpetra::global_size_t offset = agg;
-                           
-                           // size of the aggregate (number of DOFs in aggregate)
-                           LO aggSize = aggRows(agg+1) - aggRows(agg);
-                           
-                           for (LO j = 0; j < aggSize; j++) {
-                             // FIXME: Allow for bad maps
-                             const LO localRow = aggToRowMapLO[aggDofSizes[agg]+j];
-                             const size_t rowStart = ia[localRow];
-                             ja[rowStart] = offset;
-                           }
-                         });
-
-    // Compress storage (remove all INVALID, which happen when we skip zeros)
-    // We do that in-place
-    {
-      // Stage 2: compress the arrays
-      SubFactoryMonitor m2(*this, "Stage 2 (CompressData)", coarseLevel);
-      // Fill i_temp with the correct row starts
-      rows_type i_temp(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_rowptr"), numFineBlockRows+1);
-      LO nnz=0;
-      Kokkos::parallel_scan("MueLu:TentativePF:BlockCrs:compress_rows", range_type(0,numFineBlockRows),
-                            KOKKOS_LAMBDA(const LO i, LO& upd, const bool& final) {
-                              if(final)
-                                i_temp[i] = upd;
-                              for (auto j = ia[i]; j < ia[i+1]; j++)
-                                if (ja[j] != INVALID)
-                                  upd++;
-                              if(final && i == (LO) numFineBlockRows-1)
-                                i_temp[numFineBlockRows] = upd;
-                            },nnz);
-
-      cols_type j_temp(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_colind"), nnz);
-      
-      
-      Kokkos::parallel_for("MueLu:TentativePF:BlockCrs:compress_cols", range_type(0,numFineBlockRows),
-                           KOKKOS_LAMBDA(const LO i) {
-                             size_t rowStart = i_temp[i];
-                             size_t lnnz = 0;
-                             for (auto j = ia[i]; j < ia[i+1]; j++)
-                               if (ja[j] != INVALID) {
-                                 j_temp[rowStart+lnnz] = ja[j];
-                                 lnnz++;
-                               }
-                           });
-      
-      ia = i_temp;
-      ja = j_temp;
-    }
-    
-    RCP<CrsGraph> BlockGraph = CrsGraphFactory::Build(rowMap,coarseBlockMap,ia,ja);
- 
+  // Fill Graph
+  const Kokkos::TeamPolicy<execution_space> policy(numAggregates, 1);
+  Kokkos::parallel_for(
+      "MueLu:TentativePF:BlockCrs:fillGraph", policy,
+      KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread) {
+        auto agg                     = thread.league_rank();
+        Xpetra::global_size_t offset = agg;
+
+        // size of the aggregate (number of DOFs in aggregate)
+        LO aggSize = aggRows(agg + 1) - aggRows(agg);
+
+        for (LO j = 0; j < aggSize; j++) {
+          // FIXME: Allow for bad maps
+          const LO localRow     = aggToRowMapLO[aggDofSizes[agg] + j];
+          const size_t rowStart = ia[localRow];
+          ja[rowStart]          = offset;
+        }
+      });
 
-    // Managing labels & constants for ESFC
-    {
-      RCP<ParameterList> FCparams;
-      if(pL.isSublist("matrixmatrix: kernel params"))
-        FCparams=rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
-      else
-        FCparams= rcp(new ParameterList);
-      // By default, we don't need global constants for TentativeP
-      FCparams->set("compute global constants",FCparams->get("compute global constants",false));
-      std::string levelIDs = toString(levelID);
-      FCparams->set("Timer Label",std::string("MueLu::TentativeP-")+levelIDs);
-      RCP<const Export> dummy_e;
-      RCP<const Import> dummy_i;
-      BlockGraph->expertStaticFillComplete(coarseBlockMap,rowMap,dummy_i,dummy_e,FCparams);
-    }
+  // Compress storage (remove all INVALID, which happen when we skip zeros)
+  // We do that in-place
+  {
+    // Stage 2: compress the arrays
+    SubFactoryMonitor m2(*this, "Stage 2 (CompressData)", coarseLevel);
+    // Fill i_temp with the correct row starts
+    rows_type i_temp(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_rowptr"), numFineBlockRows + 1);
+    LO nnz = 0;
+    Kokkos::parallel_scan(
+        "MueLu:TentativePF:BlockCrs:compress_rows", range_type(0, numFineBlockRows),
+        KOKKOS_LAMBDA(const LO i, LO& upd, const bool& final) {
+          if (final)
+            i_temp[i] = upd;
+          for (auto j = ia[i]; j < ia[i + 1]; j++)
+            if (ja[j] != INVALID)
+              upd++;
+          if (final && i == (LO)numFineBlockRows - 1)
+            i_temp[numFineBlockRows] = upd;
+        },
+        nnz);
+
+    cols_type j_temp(Kokkos::ViewAllocateWithoutInitializing("BlockGraph_colind"), nnz);
+
+    Kokkos::parallel_for(
+        "MueLu:TentativePF:BlockCrs:compress_cols", range_type(0, numFineBlockRows),
+        KOKKOS_LAMBDA(const LO i) {
+          size_t rowStart = i_temp[i];
+          size_t lnnz     = 0;
+          for (auto j = ia[i]; j < ia[i + 1]; j++)
+            if (ja[j] != INVALID) {
+              j_temp[rowStart + lnnz] = ja[j];
+              lnnz++;
+            }
+        });
 
-    // We can't leave the ia/ja pointers floating around, because of host/device view counting, so
-    // we clear them here
-    ia = rows_type();
-    ja = cols_type();
-
-    
-    // Now let's make a BlockCrs Matrix
-    // NOTE: Assumes block size== NSDim
-    RCP<Xpetra::CrsMatrix<SC,LO,GO,NO> > P_xpetra = Xpetra::CrsMatrixFactory<SC,LO,GO,NO>::BuildBlock(BlockGraph, coarsePointMap, rangeMap,NSDim);
-    RCP<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> > P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO> >(P_xpetra);
-    if(P_tpetra.is_null()) throw std::runtime_error("BuildPUncoupled: Matrix factory did not return a Tpetra::BlockCrsMatrix");
-    RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
-
-    auto values = P_tpetra->getTpetra_BlockCrsMatrix()->getValuesDeviceNonConst();
-    const LO stride = NSDim*NSDim;
-
-    Kokkos::parallel_for("MueLu:TentativePF:BlockCrs:main_loop_noqr", policy,
-                         KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type &thread) {
-                           auto agg = thread.league_rank();
-                           
-                           // size of the aggregate (number of DOFs in aggregate)
-                           LO aggSize = aggRows(agg+1) - aggRows(agg);
-                           Xpetra::global_size_t offset = agg*NSDim;
-
-                           // Q = localQR(:,0)/norm
-                           for (LO j = 0; j < aggSize; j++) {
-                             LO localBlockRow = aggToRowMapLO(aggRows(agg)+j);
-                             LO rowStart = localBlockRow * stride;
-                             for (LO r = 0; r < (LO)NSDim; r++) {
-                               LO localPointRow = localBlockRow*NSDim + r;
-                               for (LO c = 0; c < (LO)NSDim; c++) {
-                                 values[rowStart + r*NSDim + c] = fineNSRandom(localPointRow,c);
-                               }
-                             }
-                           }
-
-                           // R = norm
-                           for(LO j=0; j<(LO)NSDim; j++)
-                             coarseNS(offset+j,j) = one;
-                         });
-
-    Ptentative = P_wrap;
+    ia = i_temp;
+    ja = j_temp;
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  void TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  BuildPcoupled(RCP<Matrix> /* A */, RCP<Aggregates> /* aggregates */,
-                RCP<AmalgamationInfo> /* amalgInfo */, RCP<MultiVector> /* fineNullspace */,
-                RCP<const Map> /* coarseMap */, RCP<Matrix>& /* Ptentative */,
-                RCP<MultiVector>& /* coarseNullspace */) const {
-    throw Exceptions::RuntimeError("MueLu: Construction of coupled tentative P is not implemented");
+  RCP<CrsGraph> BlockGraph = CrsGraphFactory::Build(rowMap, coarseBlockMap, ia, ja);
+
+  // Managing labels & constants for ESFC
+  {
+    RCP<ParameterList> FCparams;
+    if (pL.isSublist("matrixmatrix: kernel params"))
+      FCparams = rcp(new ParameterList(pL.sublist("matrixmatrix: kernel params")));
+    else
+      FCparams = rcp(new ParameterList);
+    // By default, we don't need global constants for TentativeP
+    FCparams->set("compute global constants", FCparams->get("compute global constants", false));
+    std::string levelIDs = toString(levelID);
+    FCparams->set("Timer Label", std::string("MueLu::TentativeP-") + levelIDs);
+    RCP<const Export> dummy_e;
+    RCP<const Import> dummy_i;
+    BlockGraph->expertStaticFillComplete(coarseBlockMap, rowMap, dummy_i, dummy_e, FCparams);
   }
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class DeviceType>
-  bool TentativePFactory_kokkos<Scalar,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
-  isGoodMap(const Map& rowMap, const Map& colMap) const {
-    auto rowLocalMap = rowMap.getLocalMap();
-    auto colLocalMap = colMap.getLocalMap();
+  // We can't leave the ia/ja pointers floating around, because of host/device view counting, so
+  // we clear them here
+  ia = rows_type();
+  ja = cols_type();
 
-    const size_t numRows = rowLocalMap.getLocalNumElements();
-    const size_t numCols = colLocalMap.getLocalNumElements();
+  // Now let's make a BlockCrs Matrix
+  // NOTE: Assumes block size== NSDim
+  RCP<Xpetra::CrsMatrix<SC, LO, GO, NO>> P_xpetra            = Xpetra::CrsMatrixFactory<SC, LO, GO, NO>::BuildBlock(BlockGraph, coarsePointMap, rangeMap, NSDim);
+  RCP<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO>> P_tpetra = rcp_dynamic_cast<Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO>>(P_xpetra);
+  if (P_tpetra.is_null()) throw std::runtime_error("BuildPUncoupled: Matrix factory did not return a Tpetra::BlockCrsMatrix");
+  RCP<CrsMatrixWrap> P_wrap = rcp(new CrsMatrixWrap(P_xpetra));
 
-    if (numCols < numRows)
-      return false;
+  auto values     = P_tpetra->getTpetra_BlockCrsMatrix()->getValuesDeviceNonConst();
+  const LO stride = NSDim * NSDim;
 
-    size_t numDiff = 0;
-    Kokkos::parallel_reduce("MueLu:TentativePF:isGoodMap", range_type(0, numRows),
-      KOKKOS_LAMBDA(const LO i, size_t &diff) {
+  Kokkos::parallel_for(
+      "MueLu:TentativePF:BlockCrs:main_loop_noqr", policy,
+      KOKKOS_LAMBDA(const typename Kokkos::TeamPolicy<execution_space>::member_type& thread) {
+        auto agg = thread.league_rank();
+
+        // size of the aggregate (number of DOFs in aggregate)
+        LO aggSize                   = aggRows(agg + 1) - aggRows(agg);
+        Xpetra::global_size_t offset = agg * NSDim;
+
+        // Q = localQR(:,0)/norm
+        for (LO j = 0; j < aggSize; j++) {
+          LO localBlockRow = aggToRowMapLO(aggRows(agg) + j);
+          LO rowStart      = localBlockRow * stride;
+          for (LO r = 0; r < (LO)NSDim; r++) {
+            LO localPointRow = localBlockRow * NSDim + r;
+            for (LO c = 0; c < (LO)NSDim; c++) {
+              values[rowStart + r * NSDim + c] = fineNSRandom(localPointRow, c);
+            }
+          }
+        }
+
+        // R = norm
+        for (LO j = 0; j < (LO)NSDim; j++)
+          coarseNS(offset + j, j) = one;
+      });
+
+  Ptentative = P_wrap;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+void TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    BuildPcoupled(RCP<Matrix> /* A */, RCP<Aggregates> /* aggregates */,
+                  RCP<AmalgamationInfo> /* amalgInfo */, RCP<MultiVector> /* fineNullspace */,
+                  RCP<const Map> /* coarseMap */, RCP<Matrix>& /* Ptentative */,
+                  RCP<MultiVector>& /* coarseNullspace */) const {
+  throw Exceptions::RuntimeError("MueLu: Construction of coupled tentative P is not implemented");
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class DeviceType>
+bool TentativePFactory_kokkos<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosDeviceWrapperNode<DeviceType>>::
+    isGoodMap(const Map& rowMap, const Map& colMap) const {
+  auto rowLocalMap = rowMap.getLocalMap();
+  auto colLocalMap = colMap.getLocalMap();
+
+  const size_t numRows = rowLocalMap.getLocalNumElements();
+  const size_t numCols = colLocalMap.getLocalNumElements();
+
+  if (numCols < numRows)
+    return false;
+
+  size_t numDiff = 0;
+  Kokkos::parallel_reduce(
+      "MueLu:TentativePF:isGoodMap", range_type(0, numRows),
+      KOKKOS_LAMBDA(const LO i, size_t& diff) {
         diff += (rowLocalMap.getGlobalElement(i) != colLocalMap.getGlobalElement(i));
-      }, numDiff);
+      },
+      numDiff);
 
-    return (numDiff == 0);
-  }
+  return (numDiff == 0);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
-#endif // MUELU_TENTATIVEPFACTORY_KOKKOS_DEF_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_KOKKOS_DEF_HPP
diff --git a/packages/muelu/src/Transfers/User/MueLu_UserPFactory_decl.hpp b/packages/muelu/src/Transfers/User/MueLu_UserPFactory_decl.hpp
index f481027f9de2..d85059364717 100644
--- a/packages/muelu/src/Transfers/User/MueLu_UserPFactory_decl.hpp
+++ b/packages/muelu/src/Transfers/User/MueLu_UserPFactory_decl.hpp
@@ -57,45 +57,45 @@
 
 namespace MueLu {
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class UserPFactory : public PFactory {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class UserPFactory : public PFactory {
 #undef MUELU_USERPFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    UserPFactory() { }
+  //! Constructor
+  UserPFactory() {}
 
-    //! Destructor.
-    virtual ~UserPFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~UserPFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    void Build (Level& fineLevel, Level& coarseLevel) const;
-    void BuildP(Level& fineLevel, Level& coarseLevel) const;
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+  void BuildP(Level& fineLevel, Level& coarseLevel) const;
 
-    //@}
+  //@}
 
-  }; //class UserPFactory
+};  // class UserPFactory
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_USERPFACTORY_SHORT
-#endif // MUELU_USERPFACTORY_DECL_HPP
+#endif  // MUELU_USERPFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Transfers/User/MueLu_UserPFactory_def.hpp b/packages/muelu/src/Transfers/User/MueLu_UserPFactory_def.hpp
index cdbd66f4de64..971e9b18273c 100644
--- a/packages/muelu/src/Transfers/User/MueLu_UserPFactory_def.hpp
+++ b/packages/muelu/src/Transfers/User/MueLu_UserPFactory_def.hpp
@@ -56,83 +56,82 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< RCP<const FactoryBase> >("A",              Teuchos::null, "Generating factory of the matrix A");
-    validParamList->set< RCP<const FactoryBase> >("Nullspace",      Teuchos::null, "Generating factory of the nullspace");
-    validParamList->set< std::string >           ("matrixFileName",            "", "File with matrix data");
-    validParamList->set< std::string >           ("mapFileName",               "", "File with coarse map data");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A");
+  validParamList->set<RCP<const FactoryBase> >("Nullspace", Teuchos::null, "Generating factory of the nullspace");
+  validParamList->set<std::string>("matrixFileName", "", "File with matrix data");
+  validParamList->set<std::string>("mapFileName", "", "File with coarse map data");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
-    Input(fineLevel, "A");
-    Input(fineLevel, "Nullspace");
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& /* coarseLevel */) const {
+  Input(fineLevel, "A");
+  Input(fineLevel, "Nullspace");
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
-    return BuildP(fineLevel, coarseLevel);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  return BuildP(fineLevel, coarseLevel);
+}
 
-  template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
-  void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
-    FactoryMonitor m(*this, "Build", coarseLevel);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UserPFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildP(Level& fineLevel, Level& coarseLevel) const {
+  FactoryMonitor m(*this, "Build", coarseLevel);
 
-    RCP<Matrix>      A             = Get< RCP<Matrix> >      (fineLevel, "A");
-    RCP<MultiVector> fineNullspace = Get< RCP<MultiVector> > (fineLevel, "Nullspace");
+  RCP<Matrix> A                  = Get<RCP<Matrix> >(fineLevel, "A");
+  RCP<MultiVector> fineNullspace = Get<RCP<MultiVector> >(fineLevel, "Nullspace");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() != 1, Exceptions::RuntimeError, "Block size > 1 has not been implemented");
+  TEUCHOS_TEST_FOR_EXCEPTION(A->GetFixedBlockSize() != 1, Exceptions::RuntimeError, "Block size > 1 has not been implemented");
 
-    const Teuchos::ParameterList& pL = GetParameterList();
+  const Teuchos::ParameterList& pL = GetParameterList();
 
-    std::string    mapFile   = pL.get<std::string>("mapFileName");
-    RCP<const Map> rowMap    = A->getRowMap();
-    RCP<const Map> coarseMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(mapFile, rowMap->lib(), rowMap->getComm());
-    Set(coarseLevel, "CoarseMap", coarseMap);
+  std::string mapFile      = pL.get<std::string>("mapFileName");
+  RCP<const Map> rowMap    = A->getRowMap();
+  RCP<const Map> coarseMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(mapFile, rowMap->lib(), rowMap->getComm());
+  Set(coarseLevel, "CoarseMap", coarseMap);
 
-    std::string matrixFile = pL.get<std::string>("matrixFileName");
-    RCP<Matrix> P          = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(matrixFile, rowMap, coarseMap, coarseMap, rowMap);
+  std::string matrixFile = pL.get<std::string>("matrixFileName");
+  RCP<Matrix> P          = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(matrixFile, rowMap, coarseMap, coarseMap, rowMap);
 #if 1
-    Set(coarseLevel, "P", P);
+  Set(coarseLevel, "P", P);
 #else
-    // Expand column map by 1
-    RCP<Matrix> P1 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false);
-    P = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(matrixFile, rowMap, P1->getColMap(), coarseMap, rowMap);
-    Set(coarseLevel, "P", P);
+  // Expand column map by 1
+  RCP<Matrix> P1 = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *P, false);
+  P              = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(matrixFile, rowMap, P1->getColMap(), coarseMap, rowMap);
+  Set(coarseLevel, "P", P);
 #endif
 
-    RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
-    P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(), Teuchos::ScalarTraits<SC>::zero());
-    Set(coarseLevel, "Nullspace", coarseNullspace);
-
-    // Coordinates transfer
-    size_t n = Teuchos::as<size_t>(sqrt(coarseMap->getGlobalNumElements()));
-    TEUCHOS_TEST_FOR_EXCEPTION(n*n != coarseMap->getGlobalNumElements(), Exceptions::RuntimeError, "Unfortunately, this is not the case, don't know what to do");
-
-    RCP<MultiVector> coarseCoords = MultiVectorFactory::Build(coarseMap, 2);
-    ArrayRCP<Scalar> x = coarseCoords->getDataNonConst(0), y = coarseCoords->getDataNonConst(1);
-    for (size_t LID = 0; LID < coarseMap->getLocalNumElements(); ++LID) {
-      GlobalOrdinal GID = coarseMap->getGlobalElement(LID) - coarseMap->getIndexBase();
-      GlobalOrdinal i = GID % n, j = GID/n;
-      x[LID] = i;
-      y[LID] = j;
-    }
-    Set(coarseLevel, "Coordinates", coarseCoords);
-
-    if (IsPrint(Statistics1)) {
-      RCP<ParameterList> params = rcp(new ParameterList());
-      params->set("printLoadBalancingInfo", true);
-      GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
-    }
+  RCP<MultiVector> coarseNullspace = MultiVectorFactory::Build(coarseMap, fineNullspace->getNumVectors());
+  P->apply(*fineNullspace, *coarseNullspace, Teuchos::TRANS, Teuchos::ScalarTraits<SC>::one(), Teuchos::ScalarTraits<SC>::zero());
+  Set(coarseLevel, "Nullspace", coarseNullspace);
+
+  // Coordinates transfer
+  size_t n = Teuchos::as<size_t>(sqrt(coarseMap->getGlobalNumElements()));
+  TEUCHOS_TEST_FOR_EXCEPTION(n * n != coarseMap->getGlobalNumElements(), Exceptions::RuntimeError, "Unfortunately, this is not the case, don't know what to do");
+
+  RCP<MultiVector> coarseCoords = MultiVectorFactory::Build(coarseMap, 2);
+  ArrayRCP<Scalar> x = coarseCoords->getDataNonConst(0), y = coarseCoords->getDataNonConst(1);
+  for (size_t LID = 0; LID < coarseMap->getLocalNumElements(); ++LID) {
+    GlobalOrdinal GID = coarseMap->getGlobalElement(LID) - coarseMap->getIndexBase();
+    GlobalOrdinal i = GID % n, j = GID / n;
+    x[LID] = i;
+    y[LID] = j;
   }
+  Set(coarseLevel, "Coordinates", coarseCoords);
 
+  if (IsPrint(Statistics1)) {
+    RCP<ParameterList> params = rcp(new ParameterList());
+    params->set("printLoadBalancingInfo", true);
+    GetOStream(Statistics1) << PerfUtils::PrintMatrixInfo(*P, "P", params);
+  }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_USERPFACTORY_SHORT
-#endif // MUELU_USERPFACTORY_DEF_HPP
+#endif  // MUELU_USERPFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AMGXOperator_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AMGXOperator_fwd.hpp
index a0bebac12f97..48ce2ec7cd48 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AMGXOperator_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AMGXOperator_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_AMGX)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AMGXOperator;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AMGXOperator;
 }
 
 #ifndef MUELU_AMGXOPERATOR_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_AMGXOPERATOR_FWD_HPP
+#endif  // MUELU_AMGXOPERATOR_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AdaptiveSaMLParameterListInterpreter_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AdaptiveSaMLParameterListInterpreter_fwd.hpp
index 100b247c478e..bbc1a3503113 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AdaptiveSaMLParameterListInterpreter_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AdaptiveSaMLParameterListInterpreter_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_FWD_HPP
 #define MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AdaptiveSaMLParameterListInterpreter;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AdaptiveSaMLParameterListInterpreter;
 }
 
 #ifndef MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_SHORT
 #define MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_SHORT
 #endif
 
-
-
-#endif // MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_FWD_HPP
+#endif  // MUELU_ADAPTIVESAMLPARAMETERLISTINTERPRETER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregateQualityEstimateFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregateQualityEstimateFactory_fwd.hpp
index 88e2201367bd..3d4df0e94656 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregateQualityEstimateFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregateQualityEstimateFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATEQUALITYESTIMATEFACTORY_FWD_HPP
 #define MUELU_AGGREGATEQUALITYESTIMATEFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregateQualityEstimateFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregateQualityEstimateFactory;
 }
 
 #ifndef MUELU_AGGREGATEQUALITYESTIMATEFACTORY_SHORT
 #define MUELU_AGGREGATEQUALITYESTIMATEFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATEQUALITYESTIMATEFACTORY_FWD_HPP
+#endif  // MUELU_AGGREGATEQUALITYESTIMATEFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Aggregates_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Aggregates_fwd.hpp
index 136d708eb61b..d089c71455ab 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Aggregates_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Aggregates_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATES_FWD_HPP
 #define MUELU_AGGREGATES_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Aggregates;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class Aggregates;
 }
 
 #ifndef MUELU_AGGREGATES_SHORT
 #define MUELU_AGGREGATES_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATES_FWD_HPP
+#endif  // MUELU_AGGREGATES_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationExportFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationExportFactory_fwd.hpp
index 00290c47dc08..5d137305b324 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationExportFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationExportFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONEXPORTFACTORY_FWD_HPP
 #define MUELU_AGGREGATIONEXPORTFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationExportFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationExportFactory;
 }
 
 #ifndef MUELU_AGGREGATIONEXPORTFACTORY_SHORT
 #define MUELU_AGGREGATIONEXPORTFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONEXPORTFACTORY_FWD_HPP
+#endif  // MUELU_AGGREGATIONEXPORTFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_fwd.hpp
index 62cb23898da3..ed8c4319e562 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE1ALGORITHM_FWD_HPP
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase1Algorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase1Algorithm;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE1ALGORITHM_SHORT
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE1ALGORITHM_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE1ALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_kokkos_fwd.hpp
index 3e9445818740..6a36f01ec7f2 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase1Algorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase1Algorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase1Algorithm_kokkos;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_SHORT
 #define MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE1ALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_fwd.hpp
index b28e74d55064..7c2cb0bac067 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE2AALGORITHM_FWD_HPP
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase2aAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase2aAlgorithm;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE2AALGORITHM_SHORT
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE2AALGORITHM_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE2AALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_kokkos_fwd.hpp
index 9fad5f71fde7..4a32868c666d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2aAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase2aAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase2aAlgorithm_kokkos;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_SHORT
 #define MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE2AALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_fwd.hpp
index 1c20355fdded..804b3b4a422c 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE2BALGORITHM_FWD_HPP
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase2bAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase2bAlgorithm;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE2BALGORITHM_SHORT
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE2BALGORITHM_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE2BALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_kokkos_fwd.hpp
index abc08ae8a252..511e382ec5fe 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase2bAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase2bAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase2bAlgorithm_kokkos;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_SHORT
 #define MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE2BALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_fwd.hpp
index 203fef26fc4c..62ceac249dc9 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE3ALGORITHM_FWD_HPP
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase3Algorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase3Algorithm;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE3ALGORITHM_SHORT
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE3ALGORITHM_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE3ALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_kokkos_fwd.hpp
index 9027db6ff10a..fdec4b4186e0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationPhase3Algorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationPhase3Algorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationPhase3Algorithm_kokkos;
 }
 
 #ifndef MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_SHORT
 #define MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_AGGREGATIONPHASE3ALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_fwd.hpp
index e2f2e722a5a9..2fa01d249610 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_FWD_HPP
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationStructuredAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationStructuredAlgorithm;
 }
 
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_SHORT
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONSTRUCTUREDALGORITHM_FWD_HPP
+#endif  // MUELU_AGGREGATIONSTRUCTUREDALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_kokkos_fwd.hpp
index a480fcf90150..71ed3b797b89 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AggregationStructuredAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AggregationStructuredAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AggregationStructuredAlgorithm_kokkos;
 }
 
 #ifndef MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_SHORT
 #define MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_AGGREGATIONSTRUCTUREDALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AlgebraicPermutationStrategy_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AlgebraicPermutationStrategy_fwd.hpp
index 0df1746f351a..f44b879b097a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AlgebraicPermutationStrategy_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AlgebraicPermutationStrategy_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ALGEBRAICPERMUTATIONSTRATEGY_FWD_HPP
 #define MUELU_ALGEBRAICPERMUTATIONSTRATEGY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AlgebraicPermutationStrategy;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AlgebraicPermutationStrategy;
 }
 
 #ifndef MUELU_ALGEBRAICPERMUTATIONSTRATEGY_SHORT
 #define MUELU_ALGEBRAICPERMUTATIONSTRATEGY_SHORT
 #endif
 
-
-
-#endif // MUELU_ALGEBRAICPERMUTATIONSTRATEGY_FWD_HPP
+#endif  // MUELU_ALGEBRAICPERMUTATIONSTRATEGY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationFactory_fwd.hpp
index 5cbdde9fb4e1..5ee741cdcc07 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AMALGAMATIONFACTORY_FWD_HPP
 #define MUELU_AMALGAMATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AmalgamationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class AmalgamationFactory;
 }
 
 #ifndef MUELU_AMALGAMATIONFACTORY_SHORT
 #define MUELU_AMALGAMATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_AMALGAMATIONFACTORY_FWD_HPP
+#endif  // MUELU_AMALGAMATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationInfo_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationInfo_fwd.hpp
index 159151afeaa8..19edbc4d4020 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationInfo_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmalgamationInfo_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_AMALGAMATIONINFO_FWD_HPP
 #define MUELU_AMALGAMATIONINFO_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class AmalgamationInfo;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class AmalgamationInfo;
 }
 
 #ifndef MUELU_AMALGAMATIONINFO_SHORT
 #define MUELU_AMALGAMATIONINFO_SHORT
 #endif
 
-
-
-#endif // MUELU_AMALGAMATIONINFO_FWD_HPP
+#endif  // MUELU_AMALGAMATIONINFO_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Amesos2Smoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Amesos2Smoother_fwd.hpp
index ada859634212..c7ef5a5cd5bd 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Amesos2Smoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Amesos2Smoother_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_AMESOS2)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Amesos2Smoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Amesos2Smoother;
 }
 
 #ifndef MUELU_AMESOS2SMOOTHER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_AMESOS2SMOOTHER_FWD_HPP
+#endif  // MUELU_AMESOS2SMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmesosSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmesosSmoother_fwd.hpp
index f2c57d7f6b24..540380aa6010 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmesosSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_AmesosSmoother_fwd.hpp
@@ -46,20 +46,13 @@
 #ifndef MUELU_AMESOSSMOOTHER_FWD_HPP
 #define MUELU_AMESOSSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Node>
-  class AmesosSmoother;
+template <class Node>
+class AmesosSmoother;
 }
 
-
-
 #ifndef MUELU_AMESOSSMOOTHER_SHORT
 #define MUELU_AMESOSSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_AMESOSSMOOTHER_FWD_HPP
+#endif  // MUELU_AMESOSSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BelosSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BelosSmoother_fwd.hpp
index fed0f5bf105d..d4a2ed98da71 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BelosSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BelosSmoother_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_BELOS)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BelosSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BelosSmoother;
 }
 
 #ifndef MUELU_BELOSSMOOTHER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_BELOSSMOOTHER_FWD_HPP
+#endif  // MUELU_BELOSSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlackBoxPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlackBoxPFactory_fwd.hpp
index 3c3b53c9973b..746cfef632d2 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlackBoxPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlackBoxPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLACKBOXPFACTORY_FWD_HPP
 #define MUELU_BLACKBOXPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlackBoxPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlackBoxPFactory;
 }
 
 #ifndef MUELU_BLACKBOXPFACTORY_SHORT
 #define MUELU_BLACKBOXPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BLACKBOXPFACTORY_FWD_HPP
+#endif  // MUELU_BLACKBOXPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoarseMapFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoarseMapFactory_fwd.hpp
index 95a555d2d5ef..e9123e52ed9b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoarseMapFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoarseMapFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDCOARSEMAPFACTORY_FWD_HPP
 #define MUELU_BLOCKEDCOARSEMAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedCoarseMapFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedCoarseMapFactory;
 }
 
 #ifndef MUELU_BLOCKEDCOARSEMAPFACTORY_SHORT
 #define MUELU_BLOCKEDCOARSEMAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDCOARSEMAPFACTORY_FWD_HPP
+#endif  // MUELU_BLOCKEDCOARSEMAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoordinatesTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoordinatesTransferFactory_fwd.hpp
index b5efcdbb3868..74101f460b5b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoordinatesTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedCoordinatesTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_FWD_HPP
 #define MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedCoordinatesTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedCoordinatesTransferFactory;
 }
 
 #ifndef MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_SHORT
 #define MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_BLOCKEDCOORDINATESTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedDirectSolver_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedDirectSolver_fwd.hpp
index 705743c954e0..e6829c2b25ce 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedDirectSolver_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedDirectSolver_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDDIRECTSOLVER_FWD_HPP
 #define MUELU_BLOCKEDDIRECTSOLVER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedDirectSolver;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedDirectSolver;
 }
 
 #ifndef MUELU_BLOCKEDDIRECTSOLVER_SHORT
 #define MUELU_BLOCKEDDIRECTSOLVER_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDDIRECTSOLVER_FWD_HPP
+#endif  // MUELU_BLOCKEDDIRECTSOLVER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedGaussSeidelSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedGaussSeidelSmoother_fwd.hpp
index fd897a2003b1..a1c94a983495 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedGaussSeidelSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedGaussSeidelSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_FWD_HPP
 #define MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedGaussSeidelSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedGaussSeidelSmoother;
 }
 
 #ifndef MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_SHORT
 #define MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_FWD_HPP
+#endif  // MUELU_BLOCKEDGAUSSSEIDELSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedJacobiSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedJacobiSmoother_fwd.hpp
index 71425e7c5eb9..5c1be3325e60 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedJacobiSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedJacobiSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDJACOBISMOOTHER_FWD_HPP
 #define MUELU_BLOCKEDJACOBISMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedJacobiSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedJacobiSmoother;
 }
 
 #ifndef MUELU_BLOCKEDJACOBISMOOTHER_SHORT
 #define MUELU_BLOCKEDJACOBISMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDJACOBISMOOTHER_FWD_HPP
+#endif  // MUELU_BLOCKEDJACOBISMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedPFactory_fwd.hpp
index 487243312aed..192f7aab7258 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDPFACTORY_FWD_HPP
 #define MUELU_BLOCKEDPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedPFactory;
 }
 
 #ifndef MUELU_BLOCKEDPFACTORY_SHORT
 #define MUELU_BLOCKEDPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDPFACTORY_FWD_HPP
+#endif  // MUELU_BLOCKEDPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedRAPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedRAPFactory_fwd.hpp
index d41f4e36068c..4ef9583a38ea 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedRAPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BlockedRAPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BLOCKEDRAPFACTORY_FWD_HPP
 #define MUELU_BLOCKEDRAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlockedRAPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlockedRAPFactory;
 }
 
 #ifndef MUELU_BLOCKEDRAPFACTORY_SHORT
 #define MUELU_BLOCKEDRAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BLOCKEDRAPFACTORY_FWD_HPP
+#endif  // MUELU_BLOCKEDRAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BraessSarazinSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BraessSarazinSmoother_fwd.hpp
index 54322d7bc862..0245765c517d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BraessSarazinSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BraessSarazinSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BRAESSSARAZINSMOOTHER_FWD_HPP
 #define MUELU_BRAESSSARAZINSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BraessSarazinSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BraessSarazinSmoother;
 }
 
 #ifndef MUELU_BRAESSSARAZINSMOOTHER_SHORT
 #define MUELU_BRAESSSARAZINSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_BRAESSSARAZINSMOOTHER_FWD_HPP
+#endif  // MUELU_BRAESSSARAZINSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BrickAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BrickAggregationFactory_fwd.hpp
index f844daff7dbe..a756c97eac8d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BrickAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_BrickAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_BRICKAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_BRICKAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BrickAggregationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BrickAggregationFactory;
 }
 
 #ifndef MUELU_BRICKAGGREGATIONFACTORY_SHORT
 #define MUELU_BRICKAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_BRICKAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_BRICKAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CGSolver_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CGSolver_fwd.hpp
index 7e004737187a..964c48d791c6 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CGSolver_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CGSolver_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CGSOLVER_FWD_HPP
 #define MUELU_CGSOLVER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CGSolver;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CGSolver;
 }
 
 #ifndef MUELU_CGSOLVER_SHORT
 #define MUELU_CGSOLVER_SHORT
 #endif
 
-
-
-#endif // MUELU_CGSOLVER_FWD_HPP
+#endif  // MUELU_CGSOLVER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalMapFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalMapFactory_fwd.hpp
index 7e675f6a44bb..cf361dbea053 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalMapFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalMapFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CLASSICALMAPFACTORY_FWD_HPP
 #define MUELU_CLASSICALMAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ClassicalMapFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ClassicalMapFactory;
 }
 
 #ifndef MUELU_CLASSICALMAPFACTORY_SHORT
 #define MUELU_CLASSICALMAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_CLASSICALMAPFACTORY_FWD_HPP
+#endif  // MUELU_CLASSICALMAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalPFactory_fwd.hpp
index 8e3c7f7a5dcd..932f0e5645ea 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ClassicalPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CLASSICALPFACTORY_FWD_HPP
 #define MUELU_CLASSICALPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ClassicalPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ClassicalPFactory;
 }
 
 #ifndef MUELU_CLASSICALPFACTORY_SHORT
 #define MUELU_CLASSICALPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_CLASSICALPFACTORY_FWD_HPP
+#endif  // MUELU_CLASSICALPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CloneRepartitionInterface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CloneRepartitionInterface_fwd.hpp
index 2674975a84d5..d2983fce401b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CloneRepartitionInterface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CloneRepartitionInterface_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CLONEREPARTITIONINTERFACE_FWD_HPP
 #define MUELU_CLONEREPARTITIONINTERFACE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CloneRepartitionInterface;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CloneRepartitionInterface;
 }
 
 #ifndef MUELU_CLONEREPARTITIONINTERFACE_SHORT
 #define MUELU_CLONEREPARTITIONINTERFACE_SHORT
 #endif
 
-
-
-#endif // MUELU_CLONEREPARTITIONINTERFACE_FWD_HPP
+#endif  // MUELU_CLONEREPARTITIONINTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_fwd.hpp
index 983fbfe28a42..cb0643b1d933 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COALESCEDROPFACTORY_FWD_HPP
 #define MUELU_COALESCEDROPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoalesceDropFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoalesceDropFactory;
 }
 
 #ifndef MUELU_COALESCEDROPFACTORY_SHORT
 #define MUELU_COALESCEDROPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COALESCEDROPFACTORY_FWD_HPP
+#endif  // MUELU_COALESCEDROPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_kokkos_fwd.hpp
index 67b85508061d..f4f2a08a1cc8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoalesceDropFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COALESCEDROPFACTORY_KOKKOS_FWD_HPP
 #define MUELU_COALESCEDROPFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoalesceDropFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoalesceDropFactory_kokkos;
 }
 
 #ifndef MUELU_COALESCEDROPFACTORY_KOKKOS_SHORT
 #define MUELU_COALESCEDROPFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_COALESCEDROPFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_COALESCEDROPFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseMapFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseMapFactory_fwd.hpp
index bbe941bc35f0..0051128159be 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseMapFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseMapFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COARSEMAPFACTORY_FWD_HPP
 #define MUELU_COARSEMAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoarseMapFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoarseMapFactory;
 }
 
 #ifndef MUELU_COARSEMAPFACTORY_SHORT
 #define MUELU_COARSEMAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COARSEMAPFACTORY_FWD_HPP
+#endif  // MUELU_COARSEMAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseningVisualizationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseningVisualizationFactory_fwd.hpp
index 6e16b151cb17..1507875c64ff 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseningVisualizationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoarseningVisualizationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COARSENINGVISUALIZATIONFACTORY_FWD_HPP
 #define MUELU_COARSENINGVISUALIZATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoarseningVisualizationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoarseningVisualizationFactory;
 }
 
 #ifndef MUELU_COARSENINGVISUALIZATIONFACTORY_SHORT
 #define MUELU_COARSENINGVISUALIZATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COARSENINGVISUALIZATIONFACTORY_FWD_HPP
+#endif  // MUELU_COARSENINGVISUALIZATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CombinePFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CombinePFactory_fwd.hpp
index 62f88a1ccf7b..72d6b3adc9ed 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CombinePFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CombinePFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COMBINEPFACTORY_FWD_HPP
 #define MUELU_COMBINEPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CombinePFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CombinePFactory;
 }
 
 #ifndef MUELU_COMBINEPFACTORY_SHORT
 #define MUELU_COMBINEPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COMBINEPFACTORY_FWD_HPP
+#endif  // MUELU_COMBINEPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ConstraintFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ConstraintFactory_fwd.hpp
index 28a72f63892c..50e97d853385 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ConstraintFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ConstraintFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CONSTRAINTFACTORY_FWD_HPP
 #define MUELU_CONSTRAINTFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ConstraintFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ConstraintFactory;
 }
 
 #ifndef MUELU_CONSTRAINTFACTORY_SHORT
 #define MUELU_CONSTRAINTFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_CONSTRAINTFACTORY_FWD_HPP
+#endif  // MUELU_CONSTRAINTFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Constraint_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Constraint_fwd.hpp
index 3fb2d5d69f64..6dd24eec4018 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Constraint_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Constraint_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_CONSTRAINT_FWD_HPP
 #define MUELU_CONSTRAINT_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Constraint;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Constraint;
 }
 
 #ifndef MUELU_CONSTRAINT_SHORT
 #define MUELU_CONSTRAINT_SHORT
 #endif
 
-
-
-#endif // MUELU_CONSTRAINT_FWD_HPP
+#endif  // MUELU_CONSTRAINT_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoordinatesTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoordinatesTransferFactory_fwd.hpp
index 66d39f00ad57..b5790d5adf9d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoordinatesTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoordinatesTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COORDINATESTRANSFERFACTORY_FWD_HPP
 #define MUELU_COORDINATESTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoordinatesTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoordinatesTransferFactory;
 }
 
 #ifndef MUELU_COORDINATESTRANSFERFACTORY_SHORT
 #define MUELU_COORDINATESTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COORDINATESTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_COORDINATESTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoupledRBMFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoupledRBMFactory_fwd.hpp
index 0605888e2563..26af7e7228ff 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoupledRBMFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_CoupledRBMFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_COUPLEDRBMFACTORY_FWD_HPP
 #define MUELU_COUPLEDRBMFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CoupledRBMFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CoupledRBMFactory;
 }
 
 #ifndef MUELU_COUPLEDRBMFACTORY_SHORT
 #define MUELU_COUPLEDRBMFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_COUPLEDRBMFACTORY_FWD_HPP
+#endif  // MUELU_COUPLEDRBMFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DemoFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DemoFactory_fwd.hpp
index a1faeba7139e..9e078bbbfc10 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DemoFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DemoFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_DEMOFACTORY_FWD_HPP
 #define MUELU_DEMOFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class DemoFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class DemoFactory;
 }
 
 #ifndef MUELU_DEMOFACTORY_SHORT
 #define MUELU_DEMOFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_DEMOFACTORY_FWD_HPP
+#endif  // MUELU_DEMOFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DirectSolver_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DirectSolver_fwd.hpp
index 19218ebe183e..a27dee6388b4 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DirectSolver_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DirectSolver_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_DIRECTSOLVER_FWD_HPP
 #define MUELU_DIRECTSOLVER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class DirectSolver;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class DirectSolver;
 }
 
 #ifndef MUELU_DIRECTSOLVER_SHORT
 #define MUELU_DIRECTSOLVER_SHORT
 #endif
 
-
-
-#endif // MUELU_DIRECTSOLVER_FWD_HPP
+#endif  // MUELU_DIRECTSOLVER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DropNegativeEntriesFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DropNegativeEntriesFactory_fwd.hpp
index b15258523f86..730ffb05b1e6 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DropNegativeEntriesFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_DropNegativeEntriesFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_DROPNEGATIVEENTRIESFACTORY_FWD_HPP
 #define MUELU_DROPNEGATIVEENTRIESFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class DropNegativeEntriesFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class DropNegativeEntriesFactory;
 }
 
 #ifndef MUELU_DROPNEGATIVEENTRIESFACTORY_SHORT
 #define MUELU_DROPNEGATIVEENTRIESFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_DROPNEGATIVEENTRIESFACTORY_FWD_HPP
+#endif  // MUELU_DROPNEGATIVEENTRIESFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_EminPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_EminPFactory_fwd.hpp
index d3cf70a51538..65e850b4b0d1 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_EminPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_EminPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_EMINPFACTORY_FWD_HPP
 #define MUELU_EMINPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class EminPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class EminPFactory;
 }
 
 #ifndef MUELU_EMINPFACTORY_SHORT
 #define MUELU_EMINPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_EMINPFACTORY_FWD_HPP
+#endif  // MUELU_EMINPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeBGS2x2_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeBGS2x2_fwd.hpp
index 94bf5e8f9c29..65b8b18f6515 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeBGS2x2_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeBGS2x2_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACADEBGS2X2_FWD_HPP
 #define MUELU_FACADEBGS2X2_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FacadeBGS2x2;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FacadeBGS2x2;
 }
 
 #ifndef MUELU_FACADEBGS2X2_SHORT
 #define MUELU_FACADEBGS2X2_SHORT
 #endif
 
-
-
-#endif // MUELU_FACADEBGS2X2_FWD_HPP
+#endif  // MUELU_FACADEBGS2X2_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassBase_fwd.hpp
index cfcc6df1a0ee..4e47e8d28e61 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassBase_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACADECLASSBASE_FWD_HPP
 #define MUELU_FACADECLASSBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FacadeClassBase;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FacadeClassBase;
 }
 
 #ifndef MUELU_FACADECLASSBASE_SHORT
 #define MUELU_FACADECLASSBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_FACADECLASSBASE_FWD_HPP
+#endif  // MUELU_FACADECLASSBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassFactory_fwd.hpp
index 48cf869f3ce2..4cb3a29b38d8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeClassFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACADECLASSFACTORY_FWD_HPP
 #define MUELU_FACADECLASSFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FacadeClassFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FacadeClassFactory;
 }
 
 #ifndef MUELU_FACADECLASSFACTORY_SHORT
 #define MUELU_FACADECLASSFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_FACADECLASSFACTORY_FWD_HPP
+#endif  // MUELU_FACADECLASSFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeSimple_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeSimple_fwd.hpp
index bde83211d74c..5f476be2f987 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeSimple_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FacadeSimple_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACADESIMPLE_FWD_HPP
 #define MUELU_FACADESIMPLE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FacadeSimple;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FacadeSimple;
 }
 
 #ifndef MUELU_FACADESIMPLE_SHORT
 #define MUELU_FACADESIMPLE_SHORT
 #endif
 
-
-
-#endif // MUELU_FACADESIMPLE_FWD_HPP
+#endif  // MUELU_FACADESIMPLE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryBase_fwd.hpp
index 968c9a28e2dc..4bc69ff0e30e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryBase_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_FACTORYBASE_FWD_HPP
 #define MUELU_FACTORYBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class FactoryBase;
+class FactoryBase;
 }
 
 #ifndef MUELU_FACTORYBASE_SHORT
 #define MUELU_FACTORYBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_FACTORYBASE_FWD_HPP
+#endif  // MUELU_FACTORYBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryFactory_fwd.hpp
index 0dc1a8e6cee9..6304538aa490 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACTORYFACTORY_FWD_HPP
 #define MUELU_FACTORYFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FactoryFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FactoryFactory;
 }
 
 #ifndef MUELU_FACTORYFACTORY_SHORT
 #define MUELU_FACTORYFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_FACTORYFACTORY_FWD_HPP
+#endif  // MUELU_FACTORYFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManagerBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManagerBase_fwd.hpp
index d7e1d59bb768..c4889f1adc2a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManagerBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManagerBase_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_FACTORYMANAGERBASE_FWD_HPP
 #define MUELU_FACTORYMANAGERBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class FactoryManagerBase;
+class FactoryManagerBase;
 }
 
 #ifndef MUELU_FACTORYMANAGERBASE_SHORT
 #define MUELU_FACTORYMANAGERBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_FACTORYMANAGERBASE_FWD_HPP
+#endif  // MUELU_FACTORYMANAGERBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManager_fwd.hpp
index d43ac1bd12aa..e8db2e73b285 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FactoryManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FACTORYMANAGER_FWD_HPP
 #define MUELU_FACTORYMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FactoryManager;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FactoryManager;
 }
 
 #ifndef MUELU_FACTORYMANAGER_SHORT
 #define MUELU_FACTORYMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_FACTORYMANAGER_FWD_HPP
+#endif  // MUELU_FACTORYMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Factory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Factory_fwd.hpp
index c448d322e3d0..3951e29d9242 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Factory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Factory_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_FACTORY_FWD_HPP
 #define MUELU_FACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class Factory;
+class Factory;
 }
 
 #ifndef MUELU_FACTORY_SHORT
 #define MUELU_FACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_FACTORY_FWD_HPP
+#endif  // MUELU_FACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FakeSmootherPrototype_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FakeSmootherPrototype_fwd.hpp
index d3ecc179d4bd..3a436af35b3a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FakeSmootherPrototype_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FakeSmootherPrototype_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FAKESMOOTHERPROTOTYPE_FWD_HPP
 #define MUELU_FAKESMOOTHERPROTOTYPE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FakeSmootherPrototype;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FakeSmootherPrototype;
 }
 
 #ifndef MUELU_FAKESMOOTHERPROTOTYPE_SHORT
 #define MUELU_FAKESMOOTHERPROTOTYPE_SHORT
 #endif
 
-
-
-#endif // MUELU_FAKESMOOTHERPROTOTYPE_FWD_HPP
+#endif  // MUELU_FAKESMOOTHERPROTOTYPE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FilteredAFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FilteredAFactory_fwd.hpp
index 096cb1ab3370..4d2b70d6d0e9 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FilteredAFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FilteredAFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FILTEREDAFACTORY_FWD_HPP
 #define MUELU_FILTEREDAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FilteredAFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FilteredAFactory;
 }
 
 #ifndef MUELU_FILTEREDAFACTORY_SHORT
 #define MUELU_FILTEREDAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_FILTEREDAFACTORY_FWD_HPP
+#endif  // MUELU_FILTEREDAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FineLevelInputDataFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FineLevelInputDataFactory_fwd.hpp
index 432037b1d98d..fc735b1f2599 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FineLevelInputDataFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_FineLevelInputDataFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_FINELEVELINPUTDATAFACTORY_FWD_HPP
 #define MUELU_FINELEVELINPUTDATAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FineLevelInputDataFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FineLevelInputDataFactory;
 }
 
 #ifndef MUELU_FINELEVELINPUTDATAFACTORY_SHORT
 #define MUELU_FINELEVELINPUTDATAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_FINELEVELINPUTDATAFACTORY_FWD_HPP
+#endif  // MUELU_FINELEVELINPUTDATAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GMRESSolver_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GMRESSolver_fwd.hpp
index eb56e86ca60e..5ed6ff015c9a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GMRESSolver_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GMRESSolver_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GMRESSOLVER_FWD_HPP
 #define MUELU_GMRESSOLVER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GMRESSolver;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GMRESSolver;
 }
 
 #ifndef MUELU_GMRESSOLVER_SHORT
 #define MUELU_GMRESSOLVER_SHORT
 #endif
 
-
-
-#endif // MUELU_GMRESSOLVER_FWD_HPP
+#endif  // MUELU_GMRESSOLVER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeneralGeometricPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeneralGeometricPFactory_fwd.hpp
index 04e9111c3e17..8ca9dc6e93a8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeneralGeometricPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeneralGeometricPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GENERALGEOMETRICPFACTORY_FWD_HPP
 #define MUELU_GENERALGEOMETRICPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeneralGeometricPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeneralGeometricPFactory;
 }
 
 #ifndef MUELU_GENERALGEOMETRICPFACTORY_SHORT
 #define MUELU_GENERALGEOMETRICPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_GENERALGEOMETRICPFACTORY_FWD_HPP
+#endif  // MUELU_GENERALGEOMETRICPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GenericRFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GenericRFactory_fwd.hpp
index e6145ed0ff3f..b44fb64270dc 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GenericRFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GenericRFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GENERICRFACTORY_FWD_HPP
 #define MUELU_GENERICRFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GenericRFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GenericRFactory;
 }
 
 #ifndef MUELU_GENERICRFACTORY_SHORT
 #define MUELU_GENERICRFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_GENERICRFACTORY_FWD_HPP
+#endif  // MUELU_GENERICRFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_fwd.hpp
index b94867ac798a..718876ef13ea 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GEOMETRICINTERPOLATIONPFACTORY_FWD_HPP
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeometricInterpolationPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeometricInterpolationPFactory;
 }
 
 #ifndef MUELU_GEOMETRICINTERPOLATIONPFACTORY_SHORT
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_FWD_HPP
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_kokkos_fwd.hpp
index 819880fb8786..cd6a78089b61 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GeometricInterpolationPFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_FWD_HPP
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GeometricInterpolationPFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class GeometricInterpolationPFactory_kokkos;
 }
 
 #ifndef MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_SHORT
 #define MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_GEOMETRICINTERPOLATIONPFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GlobalLexicographicIndexManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GlobalLexicographicIndexManager_fwd.hpp
index b3531134142c..4b3b93437169 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GlobalLexicographicIndexManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GlobalLexicographicIndexManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
 #define MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GlobalLexicographicIndexManager;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class GlobalLexicographicIndexManager;
 }
 
 #ifndef MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_SHORT
 #define MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
+#endif  // MUELU_GLOBALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GraphBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GraphBase_fwd.hpp
index e68d2abd81e3..1cdbacf79ae4 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GraphBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_GraphBase_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GRAPHBASE_FWD_HPP
 #define MUELU_GRAPHBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class GraphBase;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class GraphBase;
 }
 
 #ifndef MUELU_GRAPHBASE_SHORT
 #define MUELU_GRAPHBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_GRAPHBASE_FWD_HPP
+#endif  // MUELU_GRAPHBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Graph_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Graph_fwd.hpp
index 78f99cead289..35e06e4b7a1c 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Graph_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Graph_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_GRAPH_FWD_HPP
 #define MUELU_GRAPH_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Graph;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class Graph;
 }
 
 #ifndef MUELU_GRAPH_SHORT
 #define MUELU_GRAPH_SHORT
 #endif
 
-
-
-#endif // MUELU_GRAPH_FWD_HPP
+#endif  // MUELU_GRAPH_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyFactory_fwd.hpp
index 1c4df29311b3..01a73cb0e3d7 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_HIERARCHYFACTORY_FWD_HPP
 #define MUELU_HIERARCHYFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class HierarchyFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class HierarchyFactory;
 }
 
 #ifndef MUELU_HIERARCHYFACTORY_SHORT
 #define MUELU_HIERARCHYFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_HIERARCHYFACTORY_FWD_HPP
+#endif  // MUELU_HIERARCHYFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyHelpers_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyHelpers_fwd.hpp
index 89a922005d05..d1fa73d334d4 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyHelpers_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyHelpers_fwd.hpp
@@ -47,15 +47,15 @@
 #define MUELU_HIERARCHYHELPERS_FWD_HPP
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TopRAPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TopRAPFactory;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TopSmootherFactory;
-}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TopSmootherFactory;
+}  // namespace MueLu
 
 #ifndef MUELU_HIERARCHYHELPERS_SHORT
 #define MUELU_HIERARCHYHELPERS_SHORT
 #endif
 
-#endif // MUELU_HIERARCHYHELPERS_FWD_HPP
+#endif  // MUELU_HIERARCHYHELPERS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyManager_fwd.hpp
index 5ac216709ce2..e2b2da1c9492 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_HIERARCHYMANAGER_FWD_HPP
 #define MUELU_HIERARCHYMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class HierarchyManager;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class HierarchyManager;
 }
 
 #ifndef MUELU_HIERARCHYMANAGER_SHORT
 #define MUELU_HIERARCHYMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_HIERARCHYMANAGER_FWD_HPP
+#endif  // MUELU_HIERARCHYMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyUtils_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyUtils_fwd.hpp
index 181fedb435ab..3ae5d30c8d42 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyUtils_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HierarchyUtils_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_HIERARCHYUTILS_FWD_HPP
 #define MUELU_HIERARCHYUTILS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class HierarchyUtils;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class HierarchyUtils;
 }
 
 #ifndef MUELU_HIERARCHYUTILS_SHORT
 #define MUELU_HIERARCHYUTILS_SHORT
 #endif
 
-
-
-#endif // MUELU_HIERARCHYUTILS_FWD_HPP
+#endif  // MUELU_HIERARCHYUTILS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Hierarchy_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Hierarchy_fwd.hpp
index 9e361dd99f7b..5d6167a3ce14 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Hierarchy_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Hierarchy_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_HIERARCHY_FWD_HPP
 #define MUELU_HIERARCHY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Hierarchy;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Hierarchy;
 }
 
 #ifndef MUELU_HIERARCHY_SHORT
 #define MUELU_HIERARCHY_SHORT
 #endif
 
-
-
-#endif // MUELU_HIERARCHY_FWD_HPP
+#endif  // MUELU_HIERARCHY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HybridAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HybridAggregationFactory_fwd.hpp
index cb6ebca4cfe2..e6180259f346 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HybridAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_HybridAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_HYBRIDAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_HYBRIDAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class HybridAggregationFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class HybridAggregationFactory;
 }
 
 #ifndef MUELU_HYBRIDAGGREGATIONFACTORY_SHORT
 #define MUELU_HYBRIDAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_HYBRIDAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_HYBRIDAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Ifpack2Smoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Ifpack2Smoother_fwd.hpp
index 393e3bd6b9df..38e4eba9b59a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Ifpack2Smoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Ifpack2Smoother_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_IFPACK2)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Ifpack2Smoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Ifpack2Smoother;
 }
 
 #ifndef MUELU_IFPACK2SMOOTHER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_IFPACK2SMOOTHER_FWD_HPP
+#endif  // MUELU_IFPACK2SMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IfpackSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IfpackSmoother_fwd.hpp
index d2a1090535fc..6e4c7a775856 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IfpackSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IfpackSmoother_fwd.hpp
@@ -46,20 +46,17 @@
 #ifndef MUELU_IFPACKSMOOTHER_FWD_HPP
 #define MUELU_IFPACKSMOOTHER_FWD_HPP
 
-
 namespace MueLu {
-  template <class Node>
-  class IfpackSmoother;
+template <class Node>
+class IfpackSmoother;
 }
 
-//namespace MueLu {
-//  class IfpackSmoother;
-//}
+// namespace MueLu {
+//   class IfpackSmoother;
+// }
 
 #ifndef MUELU_IFPACKSMOOTHER_SHORT
 #define MUELU_IFPACKSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_IFPACKSMOOTHER_FWD_HPP
+#endif  // MUELU_IFPACKSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndefBlockedDiagonalSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndefBlockedDiagonalSmoother_fwd.hpp
index 5b838a26d81e..c6e9e2d6f963 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndefBlockedDiagonalSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndefBlockedDiagonalSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_FWD_HPP
 #define MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IndefBlockedDiagonalSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class IndefBlockedDiagonalSmoother;
 }
 
 #ifndef MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_SHORT
 #define MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_FWD_HPP
+#endif  // MUELU_INDEFBLOCKEDDIAGONALSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_fwd.hpp
index 790364d8b84d..a524c8aeacb8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INDEXMANAGER_FWD_HPP
 #define MUELU_INDEXMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IndexManager;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IndexManager;
 }
 
 #ifndef MUELU_INDEXMANAGER_SHORT
 #define MUELU_INDEXMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_INDEXMANAGER_FWD_HPP
+#endif  // MUELU_INDEXMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_kokkos_fwd.hpp
index 825db8a5f8ed..2b8b3b6caa0f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IndexManager_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INDEXMANAGER_KOKKOS_FWD_HPP
 #define MUELU_INDEXMANAGER_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IndexManager_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IndexManager_kokkos;
 }
 
 #ifndef MUELU_INDEXMANAGER_KOKKOS_SHORT
 #define MUELU_INDEXMANAGER_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_INDEXMANAGER_KOKKOS_FWD_HPP
+#endif  // MUELU_INDEXMANAGER_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InitialBlockNumberFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InitialBlockNumberFactory_fwd.hpp
index aecbcd760c63..27899d739542 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InitialBlockNumberFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InitialBlockNumberFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INITIALBLOCKNUMBERFACTORY_FWD_HPP
 #define MUELU_INITIALBLOCKNUMBERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class InitialBlockNumberFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class InitialBlockNumberFactory;
 }
 
 #ifndef MUELU_INITIALBLOCKNUMBERFACTORY_SHORT
 #define MUELU_INITIALBLOCKNUMBERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_INITIALBLOCKNUMBERFACTORY_FWD_HPP
+#endif  // MUELU_INITIALBLOCKNUMBERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationAlgorithm_fwd.hpp
index b3129ddd3392..f4214fe0d8c5 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INTERFACEAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_INTERFACEAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class InterfaceAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class InterfaceAggregationAlgorithm;
 }
 
 #ifndef MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
 #define MUELU_INTERFACEAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_INTERFACEAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_INTERFACEAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationFactory_fwd.hpp
index 2c075fae7be6..2a4b0cab690e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INTERFACEAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_INTERFACEAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class InterfaceAggregationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class InterfaceAggregationFactory;
 }
 
 #ifndef MUELU_INTERFACEAGGREGATIONFACTORY_SHORT
 #define MUELU_INTERFACEAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_INTERFACEAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_INTERFACEAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceMappingTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceMappingTransferFactory_fwd.hpp
index 51600baf8826..c6605ee2f2b9 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceMappingTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InterfaceMappingTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INTERFACEMAPPINGTRANSFERFACTORY_FWD_HPP
 #define MUELU_INTERFACEMAPPINGTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class InterfaceMappingTransferFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class InterfaceMappingTransferFactory;
 }
 
 #ifndef MUELU_INTERFACEMAPPINGTRANSFERFACTORY_SHORT
 #define MUELU_INTERFACEMAPPINGTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_INTERFACEMAPPINGTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_INTERFACEMAPPINGTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IntrepidPCoarsenFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IntrepidPCoarsenFactory_fwd.hpp
index 3852cffe5d14..fc58d9a91ce8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IntrepidPCoarsenFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IntrepidPCoarsenFactory_fwd.hpp
@@ -50,8 +50,8 @@
 #ifdef HAVE_MUELU_INTREPID2
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IntrepidPCoarsenFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class IntrepidPCoarsenFactory;
 }
 
 #ifndef MUELU_INTREPIDPCOARSENFACTORY_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_INTREPIDPCOARSENFACTORY_FWD_HPP
+#endif  // MUELU_INTREPIDPCOARSENFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InverseApproximationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InverseApproximationFactory_fwd.hpp
index 61ef3b26ba4e..95c1cc7ea025 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InverseApproximationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_InverseApproximationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_INVERSEAPPROXIMATIONFACTORY_FWD_HPP
 #define MUELU_INVERSEAPPROXIMATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class InverseApproximationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class InverseApproximationFactory;
 }
 
 #ifndef MUELU_INVERSEAPPROXIMATIONFACTORY_SHORT
 #define MUELU_INVERSEAPPROXIMATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_INVERSEAPPROXIMATIONFACTORY_FWD_HPP
+#endif  // MUELU_INVERSEAPPROXIMATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_fwd.hpp
index a8d7a80a8600..f6844650451f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IsolatedNodeAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IsolatedNodeAggregationAlgorithm;
 }
 
 #ifndef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_SHORT
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_fwd.hpp
index faa159931425..071a27ff07be 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsolatedNodeAggregationAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IsolatedNodeAggregationAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IsolatedNodeAggregationAlgorithm_kokkos;
 }
 
 #ifndef MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_SHORT
 #define MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_ISOLATEDNODEAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsorropiaInterface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsorropiaInterface_fwd.hpp
index 6dfff644a35a..9ed5ab68bbd2 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsorropiaInterface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_IsorropiaInterface_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MPI)
 
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class IsorropiaInterface;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class IsorropiaInterface;
 }
 
 #ifndef MUELU_ISORROPIAINTERFACE_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_ISORROPIAINTERFACE_FWD_HPP
+#endif  // MUELU_ISORROPIAINTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_fwd.hpp
index cadf0b5728fb..4246236aea8d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LWGRAPH_FWD_HPP
 #define MUELU_LWGRAPH_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LWGraph;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LWGraph;
 }
 
 #ifndef MUELU_LWGRAPH_SHORT
 #define MUELU_LWGRAPH_SHORT
 #endif
 
-
-
-#endif // MUELU_LWGRAPH_FWD_HPP
+#endif  // MUELU_LWGRAPH_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_kokkos_fwd.hpp
index 7179cbfd1046..a7a1c99ee0e0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LWGraph_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LWGRAPH_KOKKOS_FWD_HPP
 #define MUELU_LWGRAPH_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LWGraph_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LWGraph_kokkos;
 }
 
 #ifndef MUELU_LWGRAPH_KOKKOS_SHORT
 #define MUELU_LWGRAPH_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_LWGRAPH_KOKKOS_FWD_HPP
+#endif  // MUELU_LWGRAPH_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LeftoverAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LeftoverAggregationAlgorithm_fwd.hpp
index 8b4097b0231e..0b39a0051dd4 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LeftoverAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LeftoverAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LEFTOVERAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_LEFTOVERAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LeftoverAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LeftoverAggregationAlgorithm;
 }
 
 #ifndef MUELU_LEFTOVERAGGREGATIONALGORITHM_SHORT
 #define MUELU_LEFTOVERAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_LEFTOVERAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_LEFTOVERAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Level_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Level_fwd.hpp
index e5cbd7874ad3..114a732d2cee 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Level_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Level_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_LEVEL_FWD_HPP
 #define MUELU_LEVEL_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class Level;
+class Level;
 }
 
 #ifndef MUELU_LEVEL_SHORT
 #define MUELU_LEVEL_SHORT
 #endif
 
-
-
-#endif // MUELU_LEVEL_FWD_HPP
+#endif  // MUELU_LEVEL_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LineDetectionFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LineDetectionFactory_fwd.hpp
index 3580f469586b..53e22d88935e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LineDetectionFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LineDetectionFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LINEDETECTIONFACTORY_FWD_HPP
 #define MUELU_LINEDETECTIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LineDetectionFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class LineDetectionFactory;
 }
 
 #ifndef MUELU_LINEDETECTIONFACTORY_SHORT
 #define MUELU_LINEDETECTIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_LINEDETECTIONFACTORY_FWD_HPP
+#endif  // MUELU_LINEDETECTIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalAggregationAlgorithm_fwd.hpp
index 46a3349901ad..a89505820d42 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOCALAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_LOCALAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalAggregationAlgorithm;
 }
 
 #ifndef MUELU_LOCALAGGREGATIONALGORITHM_SHORT
 #define MUELU_LOCALAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_LOCALAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_LOCALAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLWGraph_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLWGraph_kokkos_fwd.hpp
index aac08d5afe01..aea5dd8242bb 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLWGraph_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLWGraph_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOCALLWGRAPH_KOKKOS_FWD_HPP
 #define MUELU_LOCALLWGRAPH_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalLWGraph_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalLWGraph_kokkos;
 }
 
 #ifndef MUELU_LOCALLWGRAPH_KOKKOS_SHORT
 #define MUELU_LOCALLWGRAPH_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_LOCALLWGRAPH_KOKKOS_FWD_HPP
+#endif  // MUELU_LOCALLWGRAPH_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLexicographicIndexManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLexicographicIndexManager_fwd.hpp
index 4f23a5f014b7..e56a2884e673 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLexicographicIndexManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalLexicographicIndexManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
 #define MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalLexicographicIndexManager;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalLexicographicIndexManager;
 }
 
 #ifndef MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_SHORT
 #define MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
+#endif  // MUELU_LOCALLEXICOGRAPHICINDEXMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalOrdinalTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalOrdinalTransferFactory_fwd.hpp
index 5489953ee93b..b06c8da18b55 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalOrdinalTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalOrdinalTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOCALORDINALTRANSFERFACTORY_FWD_HPP
 #define MUELU_LOCALORDINALTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalOrdinalTransferFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalOrdinalTransferFactory;
 }
 
 #ifndef MUELU_LOCALORDINALTRANSFERFACTORY_SHORT
 #define MUELU_LOCALORDINALTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_LOCALORDINALTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_LOCALORDINALTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalPermutationStrategy_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalPermutationStrategy_fwd.hpp
index 579056d7d40e..f8bf404c5614 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalPermutationStrategy_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LocalPermutationStrategy_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOCALPERMUTATIONSTRATEGY_FWD_HPP
 #define MUELU_LOCALPERMUTATIONSTRATEGY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LocalPermutationStrategy;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class LocalPermutationStrategy;
 }
 
 #ifndef MUELU_LOCALPERMUTATIONSTRATEGY_SHORT
 #define MUELU_LOCALPERMUTATIONSTRATEGY_SHORT
 #endif
 
-
-
-#endif // MUELU_LOCALPERMUTATIONSTRATEGY_FWD_HPP
+#endif  // MUELU_LOCALPERMUTATIONSTRATEGY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LowPrecisionFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LowPrecisionFactory_fwd.hpp
index 8d9feecd64e5..f3bdce0e536b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LowPrecisionFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_LowPrecisionFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_LOWPRECISIONFACTORY_FWD_HPP
 #define MUELU_LOWPRECISIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class LowPrecisionFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class LowPrecisionFactory;
 }
 
 #ifndef MUELU_LOWPRECISIONFACTORY_SHORT
 #define MUELU_LOWPRECISIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_LOWPRECISIONFACTORY_FWD_HPP
+#endif  // MUELU_LOWPRECISIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListIntepreter_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListIntepreter_fwd.hpp
index 4e93333e8fb5..54ef6bcd60b4 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListIntepreter_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListIntepreter_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_DEPRECATED_CODE)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MLParameterListIntepreter;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MLParameterListIntepreter;
 }
 
 #ifndef MUELU_MLPARAMETERLISTINTEPRETER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_MLPARAMETERLISTINTEPRETER_FWD_HPP
+#endif  // MUELU_MLPARAMETERLISTINTEPRETER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListInterpreter_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListInterpreter_fwd.hpp
index b37152707c7e..9c48fe4b1744 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListInterpreter_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MLParameterListInterpreter_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_DEPRECATED_CODE)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MLParameterListInterpreter;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MLParameterListInterpreter;
 }
 
 #ifndef MUELU_MLPARAMETERLISTINTERPRETER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_MLPARAMETERLISTINTERPRETER_FWD_HPP
+#endif  // MUELU_MLPARAMETERLISTINTERPRETER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MapTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MapTransferFactory_fwd.hpp
index 57a1ae81aaef..543f29fb2a22 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MapTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MapTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MAPTRANSFERFACTORY_FWD_HPP
 #define MUELU_MAPTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MapTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MapTransferFactory;
 }
 
 #ifndef MUELU_MAPTRANSFERFACTORY_SHORT
 #define MUELU_MAPTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_MAPTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_MAPTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatlabSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatlabSmoother_fwd.hpp
index 2e72e54e211d..8a406d4db331 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatlabSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatlabSmoother_fwd.hpp
@@ -47,12 +47,12 @@
 #define MUELU_MATLABSMOOTHER_FWD_HPP
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MatlabSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MatlabSmoother;
 }
 
 #ifndef MUELU_MATLABSMOOTHER_SHORT
 #define MUELU_MATLABSMOOTHER_SHORT
 #endif
 
-#endif // MUELU_MATLABSMOOTHER_FWD_HPP
+#endif  // MUELU_MATLABSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixAnalysisFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixAnalysisFactory_fwd.hpp
index 7a2c9be3d51c..b56f9f13089f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixAnalysisFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixAnalysisFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MATRIXANALYSISFACTORY_FWD_HPP
 #define MUELU_MATRIXANALYSISFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MatrixAnalysisFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MatrixAnalysisFactory;
 }
 
 #ifndef MUELU_MATRIXANALYSISFACTORY_SHORT
 #define MUELU_MATRIXANALYSISFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_MATRIXANALYSISFACTORY_FWD_HPP
+#endif  // MUELU_MATRIXANALYSISFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativePFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativePFactory_fwd.hpp
index 6a5b15c5431d..a66a9c6ec95c 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativePFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativePFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MATRIXFREETENTATIVEPFACTORY_FWD_HPP
 #define MUELU_MATRIXFREETENTATIVEPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MatrixFreeTentativePFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MatrixFreeTentativePFactory;
 }
 
 #ifndef MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
 #define MUELU_MATRIXFREETENTATIVEPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_MATRIXFREETENTATIVEPFACTORY_FWD_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativeP_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativeP_fwd.hpp
index 228236c4ac92..33711180e24d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativeP_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MatrixFreeTentativeP_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MATRIXFREETENTATIVEP_FWD_HPP
 #define MUELU_MATRIXFREETENTATIVEP_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MatrixFreeTentativeP;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MatrixFreeTentativeP;
 }
 
 #ifndef MUELU_MATRIXFREETENTATIVEP_SHORT
 #define MUELU_MATRIXFREETENTATIVEP_SHORT
 #endif
 
-
-
-#endif // MUELU_MATRIXFREETENTATIVEP_FWD_HPP
+#endif  // MUELU_MATRIXFREETENTATIVEP_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell1_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell1_fwd.hpp
index c16354ed70e0..4d78097223b9 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell1_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell1_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MAXWELL1_FWD_HPP
 #define MUELU_MAXWELL1_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Maxwell1;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Maxwell1;
 }
 
 #ifndef MUELU_MAXWELL1_SHORT
 #define MUELU_MAXWELL1_SHORT
 #endif
 
-
-
-#endif // MUELU_MAXWELL1_FWD_HPP
+#endif  // MUELU_MAXWELL1_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell_Utils_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell_Utils_fwd.hpp
index 88a394a9b5fc..b0ecbc40fb4f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell_Utils_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Maxwell_Utils_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MAXWELL_UTILS_FWD_HPP
 #define MUELU_MAXWELL_UTILS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Maxwell_Utils;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Maxwell_Utils;
 }
 
 #ifndef MUELU_MAXWELL_UTILS_SHORT
 #define MUELU_MAXWELL_UTILS_SHORT
 #endif
 
-
-
-#endif // MUELU_MAXWELL_UTILS_FWD_HPP
+#endif  // MUELU_MAXWELL_UTILS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedBlockedMatrixFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedBlockedMatrixFactory_fwd.hpp
index 5643f024abfb..6a0ccf943d4e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedBlockedMatrixFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedBlockedMatrixFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MERGEDBLOCKEDMATRIXFACTORY_FWD_HPP
 #define MUELU_MERGEDBLOCKEDMATRIXFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MergedBlockedMatrixFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MergedBlockedMatrixFactory;
 }
 
 #ifndef MUELU_MERGEDBLOCKEDMATRIXFACTORY_SHORT
 #define MUELU_MERGEDBLOCKEDMATRIXFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_MERGEDBLOCKEDMATRIXFACTORY_FWD_HPP
+#endif  // MUELU_MERGEDBLOCKEDMATRIXFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedSmoother_fwd.hpp
index 5b53005a7bfe..5cdabad3a622 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MergedSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MERGEDSMOOTHER_FWD_HPP
 #define MUELU_MERGEDSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MergedSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MergedSmoother;
 }
 
 #ifndef MUELU_MERGEDSMOOTHER_SHORT
 #define MUELU_MERGEDSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_MERGEDSMOOTHER_FWD_HPP
+#endif  // MUELU_MERGEDSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiPhys_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiPhys_fwd.hpp
index 73f88406e0a7..364a160af373 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiPhys_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiPhys_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MULTIPHYS_FWD_HPP
 #define MUELU_MULTIPHYS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MultiPhys;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MultiPhys;
 }
 
 #ifndef MUELU_MULTIPHYS_SHORT
 #define MUELU_MULTIPHYS_SHORT
 #endif
 
-
-
-#endif // MUELU_MULTIPHYS_FWD_HPP
+#endif  // MUELU_MULTIPHYS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiVectorTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiVectorTransferFactory_fwd.hpp
index f35676f6de50..4d2432ad1ebe 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiVectorTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_MultiVectorTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_MULTIVECTORTRANSFERFACTORY_FWD_HPP
 #define MUELU_MULTIVECTORTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class MultiVectorTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class MultiVectorTransferFactory;
 }
 
 #ifndef MUELU_MULTIVECTORTRANSFERFACTORY_SHORT
 #define MUELU_MULTIVECTORTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_MULTIVECTORTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_MULTIVECTORTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NodePartitionInterface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NodePartitionInterface_fwd.hpp
index 81cd2d5e851c..78a997f631fb 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NodePartitionInterface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NodePartitionInterface_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MPI)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NodePartitionInterface;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NodePartitionInterface;
 }
 
 #ifndef MUELU_NODEPARTITIONINTERFACE_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_NODEPARTITIONINTERFACE_FWD_HPP
+#endif  // MUELU_NODEPARTITIONINTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NotayAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NotayAggregationFactory_fwd.hpp
index 08740ee8aaa7..45d0262865df 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NotayAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NotayAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_NOTAYAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_NOTAYAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NotayAggregationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NotayAggregationFactory;
 }
 
 #ifndef MUELU_NOTAYAGGREGATIONFACTORY_SHORT
 #define MUELU_NOTAYAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_NOTAYAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_NOTAYAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_fwd.hpp
index 9bfa12079640..8846ad562b40 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_NULLSPACEFACTORY_FWD_HPP
 #define MUELU_NULLSPACEFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NullspaceFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NullspaceFactory;
 }
 
 #ifndef MUELU_NULLSPACEFACTORY_SHORT
 #define MUELU_NULLSPACEFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_NULLSPACEFACTORY_FWD_HPP
+#endif  // MUELU_NULLSPACEFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_kokkos_fwd.hpp
index 219eb70eece3..d96ce41f4d09 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspaceFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_NULLSPACEFACTORY_KOKKOS_FWD_HPP
 #define MUELU_NULLSPACEFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NullspaceFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NullspaceFactory_kokkos;
 }
 
 #ifndef MUELU_NULLSPACEFACTORY_KOKKOS_SHORT
 #define MUELU_NULLSPACEFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_NULLSPACEFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_NULLSPACEFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspacePresmoothFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspacePresmoothFactory_fwd.hpp
index 08b996adb78b..f4b808ac6630 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspacePresmoothFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_NullspacePresmoothFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_NULLSPACEPRESMOOTHFACTORY_FWD_HPP
 #define MUELU_NULLSPACEPRESMOOTHFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class NullspacePresmoothFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class NullspacePresmoothFactory;
 }
 
 #ifndef MUELU_NULLSPACEPRESMOOTHFACTORY_SHORT
 #define MUELU_NULLSPACEPRESMOOTHFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_NULLSPACEPRESMOOTHFACTORY_FWD_HPP
+#endif  // MUELU_NULLSPACEPRESMOOTHFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_fwd.hpp
index 70a588376ee2..e658dea88b79 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ONEPTAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_ONEPTAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class OnePtAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class OnePtAggregationAlgorithm;
 }
 
 #ifndef MUELU_ONEPTAGGREGATIONALGORITHM_SHORT
 #define MUELU_ONEPTAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_ONEPTAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_ONEPTAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_kokkos_fwd.hpp
index 958caeb2f6bb..805dc5cdc477 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_OnePtAggregationAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class OnePtAggregationAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class OnePtAggregationAlgorithm_kokkos;
 }
 
 #ifndef MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_SHORT
 #define MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_ONEPTAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PFactory_fwd.hpp
index 2c11ad243004..8837bbb7531e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PFactory_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_PFACTORY_FWD_HPP
 #define MUELU_PFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class PFactory;
+class PFactory;
 }
 
 #ifndef MUELU_PFACTORY_SHORT
 #define MUELU_PFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_PFACTORY_FWD_HPP
+#endif  // MUELU_PFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PRFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PRFactory_fwd.hpp
index cfa9f7624a48..6aba34086067 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PRFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PRFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PRFACTORY_FWD_HPP
 #define MUELU_PRFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PRFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class PRFactory;
 }
 
 #ifndef MUELU_PRFACTORY_SHORT
 #define MUELU_PRFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_PRFACTORY_FWD_HPP
+#endif  // MUELU_PRFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListIntepreter_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListIntepreter_fwd.hpp
index 0c9ffe6b412b..68bf0da7e003 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListIntepreter_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListIntepreter_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PARAMETERLISTINTEPRETER_FWD_HPP
 #define MUELU_PARAMETERLISTINTEPRETER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ParameterListIntepreter;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ParameterListIntepreter;
 }
 
 #ifndef MUELU_PARAMETERLISTINTEPRETER_SHORT
 #define MUELU_PARAMETERLISTINTEPRETER_SHORT
 #endif
 
-
-
-#endif // MUELU_PARAMETERLISTINTEPRETER_FWD_HPP
+#endif  // MUELU_PARAMETERLISTINTEPRETER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListInterpreter_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListInterpreter_fwd.hpp
index ea25ada73354..f080c3548454 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListInterpreter_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ParameterListInterpreter_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PARAMETERLISTINTERPRETER_FWD_HPP
 #define MUELU_PARAMETERLISTINTERPRETER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ParameterListInterpreter;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ParameterListInterpreter;
 }
 
 #ifndef MUELU_PARAMETERLISTINTERPRETER_SHORT
 #define MUELU_PARAMETERLISTINTERPRETER_SHORT
 #endif
 
-
-
-#endif // MUELU_PARAMETERLISTINTERPRETER_FWD_HPP
+#endif  // MUELU_PARAMETERLISTINTERPRETER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PatternFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PatternFactory_fwd.hpp
index d452c877d114..3cb3ada3d73f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PatternFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PatternFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PATTERNFACTORY_FWD_HPP
 #define MUELU_PATTERNFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PatternFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PatternFactory;
 }
 
 #ifndef MUELU_PATTERNFACTORY_SHORT
 #define MUELU_PATTERNFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_PATTERNFACTORY_FWD_HPP
+#endif  // MUELU_PATTERNFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfModels_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfModels_fwd.hpp
index 830a9c65b5bc..ee9d29c9a82a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfModels_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfModels_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PERFMODELS_FWD_HPP
 #define MUELU_PERFMODELS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PerfModels;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PerfModels;
 }
 
 #ifndef MUELU_PERFMODELS_SHORT
 #define MUELU_PERFMODELS_SHORT
 #endif
 
-
-
-#endif // MUELU_PERFMODELS_FWD_HPP
+#endif  // MUELU_PERFMODELS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfUtils_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfUtils_fwd.hpp
index 2e890447432e..a1d09fc0dd89 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfUtils_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PerfUtils_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PERFUTILS_FWD_HPP
 #define MUELU_PERFUTILS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PerfUtils;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PerfUtils;
 }
 
 #ifndef MUELU_PERFUTILS_SHORT
 #define MUELU_PERFUTILS_SHORT
 #endif
 
-
-
-#endif // MUELU_PERFUTILS_FWD_HPP
+#endif  // MUELU_PERFUTILS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutationFactory_fwd.hpp
index 3a61968fb870..0e6d991e944f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PERMUTATIONFACTORY_FWD_HPP
 #define MUELU_PERMUTATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PermutationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PermutationFactory;
 }
 
 #ifndef MUELU_PERMUTATIONFACTORY_SHORT
 #define MUELU_PERMUTATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_PERMUTATIONFACTORY_FWD_HPP
+#endif  // MUELU_PERMUTATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutingSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutingSmoother_fwd.hpp
index f1422d19382f..33ec332ec86c 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutingSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PermutingSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PERMUTINGSMOOTHER_FWD_HPP
 #define MUELU_PERMUTINGSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PermutingSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PermutingSmoother;
 }
 
 #ifndef MUELU_PERMUTINGSMOOTHER_SHORT
 #define MUELU_PERMUTINGSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_PERMUTINGSMOOTHER_FWD_HPP
+#endif  // MUELU_PERMUTINGSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PgPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PgPFactory_fwd.hpp
index 7325da3a985f..0b4f70484f7d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PgPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PgPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PGPFACTORY_FWD_HPP
 #define MUELU_PGPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PgPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PgPFactory;
 }
 
 #ifndef MUELU_PGPFACTORY_SHORT
 #define MUELU_PGPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_PGPFACTORY_FWD_HPP
+#endif  // MUELU_PGPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionBaseClass_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionBaseClass_fwd.hpp
index 8b86b48c023a..a30aab61cba3 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionBaseClass_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionBaseClass_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PREDROPFUNCTIONBASECLASS_FWD_HPP
 #define MUELU_PREDROPFUNCTIONBASECLASS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PreDropFunctionBaseClass;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PreDropFunctionBaseClass;
 }
 
 #ifndef MUELU_PREDROPFUNCTIONBASECLASS_SHORT
 #define MUELU_PREDROPFUNCTIONBASECLASS_SHORT
 #endif
 
-
-
-#endif // MUELU_PREDROPFUNCTIONBASECLASS_FWD_HPP
+#endif  // MUELU_PREDROPFUNCTIONBASECLASS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionConstVal_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionConstVal_fwd.hpp
index 47bd19502a31..9ed5760a825e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionConstVal_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreDropFunctionConstVal_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PREDROPFUNCTIONCONSTVAL_FWD_HPP
 #define MUELU_PREDROPFUNCTIONCONSTVAL_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PreDropFunctionConstVal;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class PreDropFunctionConstVal;
 }
 
 #ifndef MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
 #define MUELU_PREDROPFUNCTIONCONSTVAL_SHORT
 #endif
 
-
-
-#endif // MUELU_PREDROPFUNCTIONCONSTVAL_FWD_HPP
+#endif  // MUELU_PREDROPFUNCTIONCONSTVAL_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_fwd.hpp
index 9120e5325101..f923b9f9e674 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_FWD_HPP
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PreserveDirichletAggregationAlgorithm;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class PreserveDirichletAggregationAlgorithm;
 }
 
 #ifndef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_SHORT
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_SHORT
 #endif
 
-
-
-#endif // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_FWD_HPP
+#endif  // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_fwd.hpp
index c10a1a1f763a..fc6780ad441e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_PreserveDirichletAggregationAlgorithm_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class PreserveDirichletAggregationAlgorithm_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class PreserveDirichletAggregationAlgorithm_kokkos;
 }
 
 #ifndef MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_SHORT
 #define MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
+#endif  // MUELU_PRESERVEDIRICHLETAGGREGATIONALGORITHM_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ProjectorSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ProjectorSmoother_fwd.hpp
index acfb8773da5f..63b7b94cd8a0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ProjectorSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ProjectorSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_PROJECTORSMOOTHER_FWD_HPP
 #define MUELU_PROJECTORSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ProjectorSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ProjectorSmoother;
 }
 
 #ifndef MUELU_PROJECTORSMOOTHER_SHORT
 #define MUELU_PROJECTORSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_PROJECTORSMOOTHER_FWD_HPP
+#endif  // MUELU_PROJECTORSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPFactory_fwd.hpp
index 693b067b4e07..b1b656554aa9 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_RAPFACTORY_FWD_HPP
 #define MUELU_RAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RAPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RAPFactory;
 }
 
 #ifndef MUELU_RAPFACTORY_SHORT
 #define MUELU_RAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_RAPFACTORY_FWD_HPP
+#endif  // MUELU_RAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPShiftFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPShiftFactory_fwd.hpp
index 6caab04ceb69..84d53b3487d8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPShiftFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RAPShiftFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_RAPSHIFTFACTORY_FWD_HPP
 #define MUELU_RAPSHIFTFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RAPShiftFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RAPShiftFactory;
 }
 
 #ifndef MUELU_RAPSHIFTFACTORY_SHORT
 #define MUELU_RAPSHIFTFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_RAPSHIFTFACTORY_FWD_HPP
+#endif  // MUELU_RAPSHIFTFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RFactory_fwd.hpp
index bd03d4b567a4..178dc501ebd5 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RFactory_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_RFACTORY_FWD_HPP
 #define MUELU_RFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class RFactory;
+class RFactory;
 }
 
 #ifndef MUELU_RFACTORY_SHORT
 #define MUELU_RFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_RFACTORY_FWD_HPP
+#endif  // MUELU_RFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceAcFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceAcFactory_fwd.hpp
index 1a001e29cd34..0c38a94569d8 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceAcFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceAcFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCEACFACTORY_FWD_HPP
 #define MUELU_REBALANCEACFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceAcFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceAcFactory;
 }
 
 #ifndef MUELU_REBALANCEACFACTORY_SHORT
 #define MUELU_REBALANCEACFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCEACFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCEACFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockAcFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockAcFactory_fwd.hpp
index 3ec97ce4532d..04c6c4c868b3 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockAcFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockAcFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCEBLOCKACFACTORY_FWD_HPP
 #define MUELU_REBALANCEBLOCKACFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceBlockAcFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceBlockAcFactory;
 }
 
 #ifndef MUELU_REBALANCEBLOCKACFACTORY_SHORT
 #define MUELU_REBALANCEBLOCKACFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCEBLOCKACFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCEBLOCKACFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockInterpolationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockInterpolationFactory_fwd.hpp
index 9613b8570a00..c9c2f85ad77f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockInterpolationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockInterpolationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_FWD_HPP
 #define MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceBlockInterpolationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceBlockInterpolationFactory;
 }
 
 #ifndef MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_SHORT
 #define MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCEBLOCKINTERPOLATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockRestrictionFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockRestrictionFactory_fwd.hpp
index da1fe6004a1c..0362d966fe1e 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockRestrictionFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceBlockRestrictionFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_FWD_HPP
 #define MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceBlockRestrictionFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceBlockRestrictionFactory;
 }
 
 #ifndef MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_SHORT
 #define MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCEBLOCKRESTRICTIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceMapFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceMapFactory_fwd.hpp
index 59a0ccf72c18..a9320b1ecf70 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceMapFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceMapFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCEMAPFACTORY_FWD_HPP
 #define MUELU_REBALANCEMAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceMapFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceMapFactory;
 }
 
 #ifndef MUELU_REBALANCEMAPFACTORY_SHORT
 #define MUELU_REBALANCEMAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCEMAPFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCEMAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceTransferFactory_fwd.hpp
index e6fa65c238ec..65ecabcb67c5 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RebalanceTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REBALANCETRANSFERFACTORY_FWD_HPP
 #define MUELU_REBALANCETRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RebalanceTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RebalanceTransferFactory;
 }
 
 #ifndef MUELU_REBALANCETRANSFERFACTORY_SHORT
 #define MUELU_REBALANCETRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REBALANCETRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_REBALANCETRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwellSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwellSmoother_fwd.hpp
index 4ce7e7147bd3..513dd4f98ce0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwellSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwellSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REFMAXWELLSMOOTHER_FWD_HPP
 #define MUELU_REFMAXWELLSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RefMaxwellSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RefMaxwellSmoother;
 }
 
 #ifndef MUELU_REFMAXWELLSMOOTHER_SHORT
 #define MUELU_REFMAXWELLSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_REFMAXWELLSMOOTHER_FWD_HPP
+#endif  // MUELU_REFMAXWELLSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwell_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwell_fwd.hpp
index 50d732df5456..b5d60544ff99 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwell_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RefMaxwell_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REFMAXWELL_FWD_HPP
 #define MUELU_REFMAXWELL_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RefMaxwell;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RefMaxwell;
 }
 
 #ifndef MUELU_REFMAXWELL_SHORT
 #define MUELU_REFMAXWELL_SHORT
 #endif
 
-
-
-#endif // MUELU_REFMAXWELL_FWD_HPP
+#endif  // MUELU_REFMAXWELL_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_fwd.hpp
index a739f50ce373..2fb965372c8d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REGIONRFACTORY_FWD_HPP
 #define MUELU_REGIONRFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RegionRFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RegionRFactory;
 }
 
 #ifndef MUELU_REGIONRFACTORY_SHORT
 #define MUELU_REGIONRFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REGIONRFACTORY_FWD_HPP
+#endif  // MUELU_REGIONRFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_kokkos_fwd.hpp
index 94ef90313f0a..f084b2b4f470 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RegionRFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REGIONRFACTORY_KOKKOS_FWD_HPP
 #define MUELU_REGIONRFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RegionRFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RegionRFactory_kokkos;
 }
 
 #ifndef MUELU_REGIONRFACTORY_KOKKOS_SHORT
 #define MUELU_REGIONRFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_REGIONRFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_REGIONRFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReitzingerPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReitzingerPFactory_fwd.hpp
index 481490d8719b..5d68f3be5f95 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReitzingerPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReitzingerPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REITZINGERPFACTORY_FWD_HPP
 #define MUELU_REITZINGERPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ReitzingerPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ReitzingerPFactory;
 }
 
 #ifndef MUELU_REITZINGERPFACTORY_SHORT
 #define MUELU_REITZINGERPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REITZINGERPFACTORY_FWD_HPP
+#endif  // MUELU_REITZINGERPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReorderBlockAFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReorderBlockAFactory_fwd.hpp
index 3e0c343962b1..1f7f700aa3d0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReorderBlockAFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReorderBlockAFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REORDERBLOCKAFACTORY_FWD_HPP
 #define MUELU_REORDERBLOCKAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ReorderBlockAFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ReorderBlockAFactory;
 }
 
 #ifndef MUELU_REORDERBLOCKAFACTORY_SHORT
 #define MUELU_REORDERBLOCKAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REORDERBLOCKAFACTORY_FWD_HPP
+#endif  // MUELU_REORDERBLOCKAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionBlockDiagonalFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionBlockDiagonalFactory_fwd.hpp
index 3c72b07a8d0a..01aef3b72961 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionBlockDiagonalFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionBlockDiagonalFactory_fwd.hpp
@@ -50,8 +50,8 @@
 #ifdef HAVE_MPI
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RepartitionBlockDiagonalFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RepartitionBlockDiagonalFactory;
 }
 
 #ifndef MUELU_REPARTITIONBLOCKDIAGONALFACTORY_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_REPARTITIONBLOCKDIAGONALFACTORY_FWD_HPP
+#endif  // MUELU_REPARTITIONBLOCKDIAGONALFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionFactory_fwd.hpp
index ddd7cbfe52b5..7fc457588f05 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionFactory_fwd.hpp
@@ -50,8 +50,8 @@
 #ifdef HAVE_MPI
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RepartitionFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RepartitionFactory;
 }
 
 #ifndef MUELU_REPARTITIONFACTORY_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_REPARTITIONFACTORY_FWD_HPP
+#endif  // MUELU_REPARTITIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionHeuristicFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionHeuristicFactory_fwd.hpp
index a8a20730dbcd..155dbff7f897 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionHeuristicFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionHeuristicFactory_fwd.hpp
@@ -50,8 +50,8 @@
 #ifdef HAVE_MPI
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RepartitionHeuristicFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RepartitionHeuristicFactory;
 }
 
 #ifndef MUELU_REPARTITIONHEURISTICFACTORY_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_REPARTITIONHEURISTICFACTORY_FWD_HPP
+#endif  // MUELU_REPARTITIONHEURISTICFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionInterface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionInterface_fwd.hpp
index 9b0741a9fc78..7d9e3b0ea951 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionInterface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RepartitionInterface_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REPARTITIONINTERFACE_FWD_HPP
 #define MUELU_REPARTITIONINTERFACE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RepartitionInterface;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class RepartitionInterface;
 }
 
 #ifndef MUELU_REPARTITIONINTERFACE_SHORT
 #define MUELU_REPARTITIONINTERFACE_SHORT
 #endif
 
-
-
-#endif // MUELU_REPARTITIONINTERFACE_FWD_HPP
+#endif  // MUELU_REPARTITIONINTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReplicatePFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReplicatePFactory_fwd.hpp
index 115ccb7cb69e..c868fe760b2f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReplicatePFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ReplicatePFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_REPLICATEPFACTORY_FWD_HPP
 #define MUELU_REPLICATEPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ReplicatePFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ReplicatePFactory;
 }
 
 #ifndef MUELU_REPLICATEPFACTORY_SHORT
 #define MUELU_REPLICATEPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_REPLICATEPFACTORY_FWD_HPP
+#endif  // MUELU_REPLICATEPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RfromP_Or_TransP_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RfromP_Or_TransP_fwd.hpp
index b67fd2184e83..ae7cc5559f07 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RfromP_Or_TransP_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RfromP_Or_TransP_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_RFROMP_OR_TRANSP_FWD_HPP
 #define MUELU_RFROMP_OR_TRANSP_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RfromP_Or_TransP;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RfromP_Or_TransP;
 }
 
 #ifndef MUELU_RFROMP_OR_TRANSP_SHORT
 #define MUELU_RFROMP_OR_TRANSP_SHORT
 #endif
 
-
-
-#endif // MUELU_RFROMP_OR_TRANSP_FWD_HPP
+#endif  // MUELU_RFROMP_OR_TRANSP_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RigidBodyModeFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RigidBodyModeFactory_fwd.hpp
index 5b418ae45251..17f26b66cb99 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RigidBodyModeFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_RigidBodyModeFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_RIGIDBODYMODEFACTORY_FWD_HPP
 #define MUELU_RIGIDBODYMODEFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class RigidBodyModeFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class RigidBodyModeFactory;
 }
 
 #ifndef MUELU_RIGIDBODYMODEFACTORY_SHORT
 #define MUELU_RIGIDBODYMODEFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_RIGIDBODYMODEFACTORY_FWD_HPP
+#endif  // MUELU_RIGIDBODYMODEFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_fwd.hpp
index 13156483e62d..38a2bffa5fbb 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SAPFACTORY_FWD_HPP
 #define MUELU_SAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SaPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SaPFactory;
 }
 
 #ifndef MUELU_SAPFACTORY_SHORT
 #define MUELU_SAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SAPFACTORY_FWD_HPP
+#endif  // MUELU_SAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_kokkos_fwd.hpp
index 24ddeba11ebd..c93264c32ac3 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SaPFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SAPFACTORY_KOKKOS_FWD_HPP
 #define MUELU_SAPFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SaPFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SaPFactory_kokkos;
 }
 
 #ifndef MUELU_SAPFACTORY_KOKKOS_SHORT
 #define MUELU_SAPFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_SAPFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_SAPFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ScaledNullspaceFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ScaledNullspaceFactory_fwd.hpp
index 79046d331ff5..41236c4494cf 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ScaledNullspaceFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ScaledNullspaceFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SCALEDNULLSPACEFACTORY_FWD_HPP
 #define MUELU_SCALEDNULLSPACEFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ScaledNullspaceFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ScaledNullspaceFactory;
 }
 
 #ifndef MUELU_SCALEDNULLSPACEFACTORY_SHORT
 #define MUELU_SCALEDNULLSPACEFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SCALEDNULLSPACEFACTORY_FWD_HPP
+#endif  // MUELU_SCALEDNULLSPACEFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SchurComplementFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SchurComplementFactory_fwd.hpp
index f3f61174f88a..ca68b9f44e34 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SchurComplementFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SchurComplementFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SCHURCOMPLEMENTFACTORY_FWD_HPP
 #define MUELU_SCHURCOMPLEMENTFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SchurComplementFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SchurComplementFactory;
 }
 
 #ifndef MUELU_SCHURCOMPLEMENTFACTORY_SHORT
 #define MUELU_SCHURCOMPLEMENTFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SCHURCOMPLEMENTFACTORY_FWD_HPP
+#endif  // MUELU_SCHURCOMPLEMENTFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SegregatedAFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SegregatedAFactory_fwd.hpp
index 533653824ec4..e1ad5f9359df 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SegregatedAFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SegregatedAFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SEGREGATEDAFACTORY_FWD_HPP
 #define MUELU_SEGREGATEDAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SegregatedAFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SegregatedAFactory;
 }
 
 #ifndef MUELU_SEGREGATEDAFACTORY_SHORT
 #define MUELU_SEGREGATEDAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SEGREGATEDAFACTORY_FWD_HPP
+#endif  // MUELU_SEGREGATEDAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_fwd.hpp
index 9d82b0c510a4..d40efeb120eb 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SEMICOARSENPFACTORY_FWD_HPP
 #define MUELU_SEMICOARSENPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SemiCoarsenPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SemiCoarsenPFactory;
 }
 
 #ifndef MUELU_SEMICOARSENPFACTORY_SHORT
 #define MUELU_SEMICOARSENPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SEMICOARSENPFACTORY_FWD_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_kokkos_fwd.hpp
index cf498c704b19..4e4ebb8fb501 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SemiCoarsenPFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SEMICOARSENPFACTORY_KOKKOS_FWD_HPP
 #define MUELU_SEMICOARSENPFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SemiCoarsenPFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SemiCoarsenPFactory_kokkos;
 }
 
 #ifndef MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
 #define MUELU_SEMICOARSENPFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_SEMICOARSENPFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_SEMICOARSENPFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacianOperator_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacianOperator_fwd.hpp
index fd2c3d30bb50..3ea8aea07d7d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacianOperator_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacianOperator_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_BELOS) && defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ShiftedLaplacianOperator;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ShiftedLaplacianOperator;
 }
 
 #ifndef MUELU_SHIFTEDLAPLACIANOPERATOR_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_SHIFTEDLAPLACIANOPERATOR_FWD_HPP
+#endif  // MUELU_SHIFTEDLAPLACIANOPERATOR_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacian_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacian_fwd.hpp
index ebc9b3a764f9..2b92bc9332ac 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacian_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ShiftedLaplacian_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_BELOS) && defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ShiftedLaplacian;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ShiftedLaplacian;
 }
 
 #ifndef MUELU_SHIFTEDLAPLACIAN_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_SHIFTEDLAPLACIAN_FWD_HPP
+#endif  // MUELU_SHIFTEDLAPLACIAN_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SimpleSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SimpleSmoother_fwd.hpp
index 794176096cf3..985856dbf4f6 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SimpleSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SimpleSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SIMPLESMOOTHER_FWD_HPP
 #define MUELU_SIMPLESMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SimpleSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SimpleSmoother;
 }
 
 #ifndef MUELU_SIMPLESMOOTHER_SHORT
 #define MUELU_SIMPLESMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_SIMPLESMOOTHER_FWD_HPP
+#endif  // MUELU_SIMPLESMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelFactoryBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelFactoryBase_fwd.hpp
index 6c21521fab5e..9e5c9358b6ae 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelFactoryBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelFactoryBase_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_SINGLELEVELFACTORYBASE_FWD_HPP
 #define MUELU_SINGLELEVELFACTORYBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class SingleLevelFactoryBase;
+class SingleLevelFactoryBase;
 }
 
 #ifndef MUELU_SINGLELEVELFACTORYBASE_SHORT
 #define MUELU_SINGLELEVELFACTORYBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_SINGLELEVELFACTORYBASE_FWD_HPP
+#endif  // MUELU_SINGLELEVELFACTORYBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelMatlabFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelMatlabFactory_fwd.hpp
index ca1469e0996a..35f624050fee 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelMatlabFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SingleLevelMatlabFactory_fwd.hpp
@@ -47,12 +47,12 @@
 #define MUELU_SINGLELEVELMATLABFACTORY_FWD_HPP
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SingleLevelMatlabFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SingleLevelMatlabFactory;
 }
 
 #ifndef MUELU_SINGLELEVELMATLABFACTORY_SHORT
 #define MUELU_SINGLELEVELMATLABFACTORY_SHORT
 #endif
 
-#endif // MUELU_SINGLELEVELMATLABFACTORY_FWD_HPP
+#endif  // MUELU_SINGLELEVELMATLABFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmooVecCoalesceDropFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmooVecCoalesceDropFactory_fwd.hpp
index fad87eadb7d7..555d35d859e5 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmooVecCoalesceDropFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmooVecCoalesceDropFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SMOOVECCOALESCEDROPFACTORY_FWD_HPP
 #define MUELU_SMOOVECCOALESCEDROPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SmooVecCoalesceDropFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SmooVecCoalesceDropFactory;
 }
 
 #ifndef MUELU_SMOOVECCOALESCEDROPFACTORY_SHORT
 #define MUELU_SMOOVECCOALESCEDROPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOVECCOALESCEDROPFACTORY_FWD_HPP
+#endif  // MUELU_SMOOVECCOALESCEDROPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherBase_fwd.hpp
index bd1574cff821..fbdd3c3637a2 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherBase_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SMOOTHERBASE_FWD_HPP
 #define MUELU_SMOOTHERBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SmootherBase;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SmootherBase;
 }
 
 #ifndef MUELU_SMOOTHERBASE_SHORT
 #define MUELU_SMOOTHERBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOTHERBASE_FWD_HPP
+#endif  // MUELU_SMOOTHERBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactoryBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactoryBase_fwd.hpp
index 12ea27adfdaa..04063862101f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactoryBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactoryBase_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_SMOOTHERFACTORYBASE_FWD_HPP
 #define MUELU_SMOOTHERFACTORYBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class SmootherFactoryBase;
+class SmootherFactoryBase;
 }
 
 #ifndef MUELU_SMOOTHERFACTORYBASE_SHORT
 #define MUELU_SMOOTHERFACTORYBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOTHERFACTORYBASE_FWD_HPP
+#endif  // MUELU_SMOOTHERFACTORYBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactory_fwd.hpp
index 8643a8b09f0f..ac0c5653901f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SMOOTHERFACTORY_FWD_HPP
 #define MUELU_SMOOTHERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SmootherFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SmootherFactory;
 }
 
 #ifndef MUELU_SMOOTHERFACTORY_SHORT
 #define MUELU_SMOOTHERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOTHERFACTORY_FWD_HPP
+#endif  // MUELU_SMOOTHERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherPrototype_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherPrototype_fwd.hpp
index 96074e6cdf46..3e9d8b145dcc 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherPrototype_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SmootherPrototype_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SMOOTHERPROTOTYPE_FWD_HPP
 #define MUELU_SMOOTHERPROTOTYPE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SmootherPrototype;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SmootherPrototype;
 }
 
 #ifndef MUELU_SMOOTHERPROTOTYPE_SHORT
 #define MUELU_SMOOTHERPROTOTYPE_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOTHERPROTOTYPE_FWD_HPP
+#endif  // MUELU_SMOOTHERPROTOTYPE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Smoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Smoother_fwd.hpp
index a51573119284..687c66b815b1 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Smoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Smoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SMOOTHER_FWD_HPP
 #define MUELU_SMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Smoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Smoother;
 }
 
 #ifndef MUELU_SMOOTHER_SHORT
 #define MUELU_SMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_SMOOTHER_FWD_HPP
+#endif  // MUELU_SMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SolverBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SolverBase_fwd.hpp
index c70c078af6c8..2f9446f332ee 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SolverBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SolverBase_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SOLVERBASE_FWD_HPP
 #define MUELU_SOLVERBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SolverBase;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SolverBase;
 }
 
 #ifndef MUELU_SOLVERBASE_SHORT
 #define MUELU_SOLVERBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_SOLVERBASE_FWD_HPP
+#endif  // MUELU_SOLVERBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SteepestDescentSolver_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SteepestDescentSolver_fwd.hpp
index 2bf893668225..dfe5e8b5bad0 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SteepestDescentSolver_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SteepestDescentSolver_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_STEEPESTDESCENTSOLVER_FWD_HPP
 #define MUELU_STEEPESTDESCENTSOLVER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SteepestDescentSolver;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SteepestDescentSolver;
 }
 
 #ifndef MUELU_STEEPESTDESCENTSOLVER_SHORT
 #define MUELU_STEEPESTDESCENTSOLVER_SHORT
 #endif
 
-
-
-#endif // MUELU_STEEPESTDESCENTSOLVER_FWD_HPP
+#endif  // MUELU_STEEPESTDESCENTSOLVER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StratimikosSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StratimikosSmoother_fwd.hpp
index 8bf88afda1d8..5df3d58782de 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StratimikosSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StratimikosSmoother_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class StratimikosSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class StratimikosSmoother;
 }
 
 #ifndef MUELU_STRATIMIKOSSMOOTHER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_STRATIMIKOSSMOOTHER_FWD_HPP
+#endif  // MUELU_STRATIMIKOSSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_fwd.hpp
index ee21b4e03b67..89ae9a6e3218 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_STRUCTUREDAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class StructuredAggregationFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class StructuredAggregationFactory;
 }
 
 #ifndef MUELU_STRUCTUREDAGGREGATIONFACTORY_SHORT
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_STRUCTUREDAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_STRUCTUREDAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_kokkos_fwd.hpp
index 6729957f7d54..196b9a294157 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredAggregationFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class StructuredAggregationFactory_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class StructuredAggregationFactory_kokkos;
 }
 
 #ifndef MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_SHORT
 #define MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_STRUCTUREDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredLineDetectionFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredLineDetectionFactory_fwd.hpp
index 3c476a187ede..de02b6ce1d6d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredLineDetectionFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_StructuredLineDetectionFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_STRUCTUREDLINEDETECTIONFACTORY_FWD_HPP
 #define MUELU_STRUCTUREDLINEDETECTIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class StructuredLineDetectionFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class StructuredLineDetectionFactory;
 }
 
 #ifndef MUELU_STRUCTUREDLINEDETECTIONFACTORY_SHORT
 #define MUELU_STRUCTUREDLINEDETECTIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_STRUCTUREDLINEDETECTIONFACTORY_FWD_HPP
+#endif  // MUELU_STRUCTUREDLINEDETECTIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SubBlockAFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SubBlockAFactory_fwd.hpp
index eaab14a73450..ed3a67a3282d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SubBlockAFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_SubBlockAFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_SUBBLOCKAFACTORY_FWD_HPP
 #define MUELU_SUBBLOCKAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class SubBlockAFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SubBlockAFactory;
 }
 
 #ifndef MUELU_SUBBLOCKAFACTORY_SHORT
 #define MUELU_SUBBLOCKAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_SUBBLOCKAFACTORY_FWD_HPP
+#endif  // MUELU_SUBBLOCKAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TekoSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TekoSmoother_fwd.hpp
index c30744392e8b..7e1946d85796 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TekoSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TekoSmoother_fwd.hpp
@@ -50,8 +50,8 @@
 #ifdef HAVE_MUELU_TEKO
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TekoSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TekoSmoother;
 }
 
 #ifndef MUELU_TEKOSMOOTHER_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_TEKOSMOOTHER_FWD_HPP
+#endif  // MUELU_TEKOSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_fwd.hpp
index b52fb0f2957d..c7dcbdc3d96c 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TENTATIVEPFACTORY_FWD_HPP
 #define MUELU_TENTATIVEPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TentativePFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TentativePFactory;
 }
 
 #ifndef MUELU_TENTATIVEPFACTORY_SHORT
 #define MUELU_TENTATIVEPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TENTATIVEPFACTORY_FWD_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_kokkos_fwd.hpp
index 4e799034c21f..446245e1517b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TentativePFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TENTATIVEPFACTORY_KOKKOS_FWD_HPP
 #define MUELU_TENTATIVEPFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TentativePFactory_kokkos;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TentativePFactory_kokkos;
 }
 
 #ifndef MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
 #define MUELU_TENTATIVEPFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_TENTATIVEPFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_TENTATIVEPFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ThresholdAFilterFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ThresholdAFilterFactory_fwd.hpp
index 49f0284020ec..ece878876d73 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ThresholdAFilterFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ThresholdAFilterFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_THRESHOLDAFILTERFACTORY_FWD_HPP
 #define MUELU_THRESHOLDAFILTERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ThresholdAFilterFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ThresholdAFilterFactory;
 }
 
 #ifndef MUELU_THRESHOLDAFILTERFACTORY_SHORT
 #define MUELU_THRESHOLDAFILTERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_THRESHOLDAFILTERFACTORY_FWD_HPP
+#endif  // MUELU_THRESHOLDAFILTERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ToggleCoordinatesTransferFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ToggleCoordinatesTransferFactory_fwd.hpp
index 5f12c299e9a7..f04cfa138785 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ToggleCoordinatesTransferFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ToggleCoordinatesTransferFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TOGGLECOORDINATESTRANSFERFACTORY_FWD_HPP
 #define MUELU_TOGGLECOORDINATESTRANSFERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ToggleCoordinatesTransferFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ToggleCoordinatesTransferFactory;
 }
 
 #ifndef MUELU_TOGGLECOORDINATESTRANSFERFACTORY_SHORT
 #define MUELU_TOGGLECOORDINATESTRANSFERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TOGGLECOORDINATESTRANSFERFACTORY_FWD_HPP
+#endif  // MUELU_TOGGLECOORDINATESTRANSFERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TogglePFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TogglePFactory_fwd.hpp
index 17c2583438e6..da8fe69ec22a 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TogglePFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TogglePFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TOGGLEPFACTORY_FWD_HPP
 #define MUELU_TOGGLEPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TogglePFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TogglePFactory;
 }
 
 #ifndef MUELU_TOGGLEPFACTORY_SHORT
 #define MUELU_TOGGLEPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TOGGLEPFACTORY_FWD_HPP
+#endif  // MUELU_TOGGLEPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopRAPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopRAPFactory_fwd.hpp
index 68170fdf21ff..39783e914d22 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopRAPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopRAPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TOPRAPFACTORY_FWD_HPP
 #define MUELU_TOPRAPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TopRAPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TopRAPFactory;
 }
 
 #ifndef MUELU_TOPRAPFACTORY_SHORT
 #define MUELU_TOPRAPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TOPRAPFACTORY_FWD_HPP
+#endif  // MUELU_TOPRAPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopSmootherFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopSmootherFactory_fwd.hpp
index a245775a57fd..bcabc37b7e2b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopSmootherFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TopSmootherFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TOPSMOOTHERFACTORY_FWD_HPP
 #define MUELU_TOPSMOOTHERFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TopSmootherFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TopSmootherFactory;
 }
 
 #ifndef MUELU_TOPSMOOTHERFACTORY_SHORT
 #define MUELU_TOPSMOOTHERFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TOPSMOOTHERFACTORY_FWD_HPP
+#endif  // MUELU_TOPSMOOTHERFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TpetraOperator_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TpetraOperator_fwd.hpp
index a55ba2c49009..e8590e5c6b28 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TpetraOperator_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TpetraOperator_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TPETRAOPERATOR_FWD_HPP
 #define MUELU_TPETRAOPERATOR_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TpetraOperator;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TpetraOperator;
 }
 
 #ifndef MUELU_TPETRAOPERATOR_SHORT
 #define MUELU_TPETRAOPERATOR_SHORT
 #endif
 
-
-
-#endif // MUELU_TPETRAOPERATOR_FWD_HPP
+#endif  // MUELU_TPETRAOPERATOR_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TransPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TransPFactory_fwd.hpp
index 823f13c78ec8..d9444b57982b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TransPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TransPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TRANSPFACTORY_FWD_HPP
 #define MUELU_TRANSPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TransPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TransPFactory;
 }
 
 #ifndef MUELU_TRANSPFACTORY_SHORT
 #define MUELU_TRANSPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_TRANSPFACTORY_FWD_HPP
+#endif  // MUELU_TRANSPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TrilinosSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TrilinosSmoother_fwd.hpp
index de233531056e..b563d2ed8a2f 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TrilinosSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TrilinosSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_TRILINOSSMOOTHER_FWD_HPP
 #define MUELU_TRILINOSSMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TrilinosSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TrilinosSmoother;
 }
 
 #ifndef MUELU_TRILINOSSMOOTHER_SHORT
 #define MUELU_TRILINOSSMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_TRILINOSSMOOTHER_FWD_HPP
+#endif  // MUELU_TRILINOSSMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelFactoryBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelFactoryBase_fwd.hpp
index 9c22ae3c597e..d90c114dae78 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelFactoryBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelFactoryBase_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_TWOLEVELFACTORYBASE_FWD_HPP
 #define MUELU_TWOLEVELFACTORYBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class TwoLevelFactoryBase;
+class TwoLevelFactoryBase;
 }
 
 #ifndef MUELU_TWOLEVELFACTORYBASE_SHORT
 #define MUELU_TWOLEVELFACTORYBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_TWOLEVELFACTORYBASE_FWD_HPP
+#endif  // MUELU_TWOLEVELFACTORYBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelMatlabFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelMatlabFactory_fwd.hpp
index f17105eaa710..d7bdcb5fb827 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelMatlabFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_TwoLevelMatlabFactory_fwd.hpp
@@ -47,12 +47,12 @@
 #define MUELU_TWOLEVELMATLABFACTORY_FWD_HPP
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class TwoLevelMatlabFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TwoLevelMatlabFactory;
 }
 
 #ifndef MUELU_TWOLEVELMATLABFACTORY_SHORT
 #define MUELU_TWOLEVELMATLABFACTORY_SHORT
 #endif
 
-#endif // MUELU_TWOLEVELMATLABFACTORY_FWD_HPP
+#endif  // MUELU_TWOLEVELMATLABFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_fwd.hpp
index 08b3be47fb90..f14a1373d432 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UNCOUPLEDAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UncoupledAggregationFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class UncoupledAggregationFactory;
 }
 
 #ifndef MUELU_UNCOUPLEDAGGREGATIONFACTORY_SHORT
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_UNCOUPLEDAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_UNCOUPLEDAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_kokkos_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_kokkos_fwd.hpp
index 1ab809335886..f4adc18027d7 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_kokkos_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledAggregationFactory_kokkos_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UncoupledAggregationFactory_kokkos;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class UncoupledAggregationFactory_kokkos;
 }
 
 #ifndef MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_SHORT
 #define MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_SHORT
 #endif
 
-
-
-#endif // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
+#endif  // MUELU_UNCOUPLEDAGGREGATIONFACTORY_KOKKOS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledIndexManager_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledIndexManager_fwd.hpp
index debbaaf5452e..618a367b7e19 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledIndexManager_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UncoupledIndexManager_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UNCOUPLEDINDEXMANAGER_FWD_HPP
 #define MUELU_UNCOUPLEDINDEXMANAGER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UncoupledIndexManager;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class UncoupledIndexManager;
 }
 
 #ifndef MUELU_UNCOUPLEDINDEXMANAGER_SHORT
 #define MUELU_UNCOUPLEDINDEXMANAGER_SHORT
 #endif
 
-
-
-#endif // MUELU_UNCOUPLEDINDEXMANAGER_FWD_HPP
+#endif  // MUELU_UNCOUPLEDINDEXMANAGER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UnsmooshFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UnsmooshFactory_fwd.hpp
index b0d558503b33..9c67ab56c8fd 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UnsmooshFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UnsmooshFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UNSMOOSHFACTORY_FWD_HPP
 #define MUELU_UNSMOOSHFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UnsmooshFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class UnsmooshFactory;
 }
 
 #ifndef MUELU_UNSMOOSHFACTORY_SHORT
 #define MUELU_UNSMOOSHFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_UNSMOOSHFACTORY_FWD_HPP
+#endif  // MUELU_UNSMOOSHFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserAggregationFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserAggregationFactory_fwd.hpp
index b3011f7234d2..e8c14ca30e98 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserAggregationFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserAggregationFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_USERAGGREGATIONFACTORY_FWD_HPP
 #define MUELU_USERAGGREGATIONFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UserAggregationFactory;
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+class UserAggregationFactory;
 }
 
 #ifndef MUELU_USERAGGREGATIONFACTORY_SHORT
 #define MUELU_USERAGGREGATIONFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_USERAGGREGATIONFACTORY_FWD_HPP
+#endif  // MUELU_USERAGGREGATIONFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserPFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserPFactory_fwd.hpp
index 54a58819ee35..e36c98ea2627 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserPFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UserPFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_USERPFACTORY_FWD_HPP
 #define MUELU_USERPFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UserPFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class UserPFactory;
 }
 
 #ifndef MUELU_USERPFACTORY_SHORT
 #define MUELU_USERPFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_USERPFACTORY_FWD_HPP
+#endif  // MUELU_USERPFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UtilitiesBase_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UtilitiesBase_fwd.hpp
index 99b1d5a87ba2..9d3cf8f658ef 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UtilitiesBase_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UtilitiesBase_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UTILITIESBASE_FWD_HPP
 #define MUELU_UTILITIESBASE_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UtilitiesBase;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class UtilitiesBase;
 }
 
 #ifndef MUELU_UTILITIESBASE_SHORT
 #define MUELU_UTILITIESBASE_SHORT
 #endif
 
-
-
-#endif // MUELU_UTILITIESBASE_FWD_HPP
+#endif  // MUELU_UTILITIESBASE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Utilities_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Utilities_fwd.hpp
index 3b0afb9b733b..711c15ec4184 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Utilities_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Utilities_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UTILITIES_FWD_HPP
 #define MUELU_UTILITIES_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Utilities;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Utilities;
 }
 
 #ifndef MUELU_UTILITIES_SHORT
 #define MUELU_UTILITIES_SHORT
 #endif
 
-
-
-#endif // MUELU_UTILITIES_FWD_HPP
+#endif  // MUELU_UTILITIES_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UzawaSmoother_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UzawaSmoother_fwd.hpp
index 2fb335e0d9cd..1690c15a646b 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UzawaSmoother_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_UzawaSmoother_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_UZAWASMOOTHER_FWD_HPP
 #define MUELU_UZAWASMOOTHER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class UzawaSmoother;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class UzawaSmoother;
 }
 
 #ifndef MUELU_UZAWASMOOTHER_SHORT
 #define MUELU_UZAWASMOOTHER_SHORT
 #endif
 
-
-
-#endif // MUELU_UZAWASMOOTHER_FWD_HPP
+#endif  // MUELU_UZAWASMOOTHER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableContainer_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableContainer_fwd.hpp
index 851a8de55a36..ffb21aa47d89 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableContainer_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableContainer_fwd.hpp
@@ -46,17 +46,12 @@
 #ifndef MUELU_VARIABLECONTAINER_FWD_HPP
 #define MUELU_VARIABLECONTAINER_FWD_HPP
 
-
-
-
 namespace MueLu {
-  class VariableContainer;
+class VariableContainer;
 }
 
 #ifndef MUELU_VARIABLECONTAINER_SHORT
 #define MUELU_VARIABLECONTAINER_SHORT
 #endif
 
-
-
-#endif // MUELU_VARIABLECONTAINER_FWD_HPP
+#endif  // MUELU_VARIABLECONTAINER_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableDofLaplacianFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableDofLaplacianFactory_fwd.hpp
index c0c3351542e9..cd7934b5dcd1 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableDofLaplacianFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VariableDofLaplacianFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_VARIABLEDOFLAPLACIANFACTORY_FWD_HPP
 #define MUELU_VARIABLEDOFLAPLACIANFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class VariableDofLaplacianFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class VariableDofLaplacianFactory;
 }
 
 #ifndef MUELU_VARIABLEDOFLAPLACIANFACTORY_SHORT
 #define MUELU_VARIABLEDOFLAPLACIANFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_VARIABLEDOFLAPLACIANFACTORY_FWD_HPP
+#endif  // MUELU_VARIABLEDOFLAPLACIANFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VisualizationHelpers_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VisualizationHelpers_fwd.hpp
index 830c565fbb15..28cd1b1c8f20 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VisualizationHelpers_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_VisualizationHelpers_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_VISUALIZATIONHELPERS_FWD_HPP
 #define MUELU_VISUALIZATIONHELPERS_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class VisualizationHelpers;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class VisualizationHelpers;
 }
 
 #ifndef MUELU_VISUALIZATIONHELPERS_SHORT
 #define MUELU_VISUALIZATIONHELPERS_SHORT
 #endif
 
-
-
-#endif // MUELU_VISUALIZATIONHELPERS_FWD_HPP
+#endif  // MUELU_VISUALIZATIONHELPERS_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_XpetraOperator_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_XpetraOperator_fwd.hpp
index d5fb0c24fc90..bfb652306ba5 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_XpetraOperator_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_XpetraOperator_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_XPETRAOPERATOR_FWD_HPP
 #define MUELU_XPETRAOPERATOR_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class XpetraOperator;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class XpetraOperator;
 }
 
 #ifndef MUELU_XPETRAOPERATOR_SHORT
 #define MUELU_XPETRAOPERATOR_SHORT
 #endif
 
-
-
-#endif // MUELU_XPETRAOPERATOR_FWD_HPP
+#endif  // MUELU_XPETRAOPERATOR_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZeroSubBlockAFactory_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZeroSubBlockAFactory_fwd.hpp
index d2784ce05787..cdd103286337 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZeroSubBlockAFactory_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZeroSubBlockAFactory_fwd.hpp
@@ -46,18 +46,13 @@
 #ifndef MUELU_ZEROSUBBLOCKAFACTORY_FWD_HPP
 #define MUELU_ZEROSUBBLOCKAFACTORY_FWD_HPP
 
-
-
-
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ZeroSubBlockAFactory;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ZeroSubBlockAFactory;
 }
 
 #ifndef MUELU_ZEROSUBBLOCKAFACTORY_SHORT
 #define MUELU_ZEROSUBBLOCKAFACTORY_SHORT
 #endif
 
-
-
-#endif // MUELU_ZEROSUBBLOCKAFACTORY_FWD_HPP
+#endif  // MUELU_ZEROSUBBLOCKAFACTORY_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Zoltan2Interface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Zoltan2Interface_fwd.hpp
index f1608edfc823..2152edd6e343 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Zoltan2Interface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_Zoltan2Interface_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MPI)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class Zoltan2Interface;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Zoltan2Interface;
 }
 
 #ifndef MUELU_ZOLTAN2INTERFACE_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_ZOLTAN2INTERFACE_FWD_HPP
+#endif  // MUELU_ZOLTAN2INTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZoltanInterface_fwd.hpp b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZoltanInterface_fwd.hpp
index 657cbdc5ff64..c3865d568f4d 100644
--- a/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZoltanInterface_fwd.hpp
+++ b/packages/muelu/src/Utils/ForwardDeclaration/MueLu_ZoltanInterface_fwd.hpp
@@ -50,8 +50,8 @@
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
 
 namespace MueLu {
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ZoltanInterface;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ZoltanInterface;
 }
 
 #ifndef MUELU_ZOLTANINTERFACE_SHORT
@@ -60,4 +60,4 @@ namespace MueLu {
 
 #endif
 
-#endif // MUELU_ZOLTANINTERFACE_FWD_HPP
+#endif  // MUELU_ZOLTANINTERFACE_FWD_HPP
diff --git a/packages/muelu/src/Utils/ML_Linker.hpp b/packages/muelu/src/Utils/ML_Linker.hpp
index 36db72d694db..f0fa9a92d0c3 100644
--- a/packages/muelu/src/Utils/ML_Linker.hpp
+++ b/packages/muelu/src/Utils/ML_Linker.hpp
@@ -48,7 +48,7 @@
 #include "ml_aggregate.h"
 #include "ml_epetra_utils.h"
 
-extern MueLu_Graph  *MueLu_BuildGraph(ML_Matrix *Amatrix, char *name);
+extern MueLu_Graph *MueLu_BuildGraph(ML_Matrix *Amatrix, char *name);
 extern int MueLu_DestroyGraph(MueLu_Graph *graph);
 
 /**********************************************************************************/
@@ -57,41 +57,39 @@ extern int MueLu_DestroyGraph(MueLu_Graph *graph);
 /* aggregation inside MueLu.
 /**********************************************************************************/
 int ML_Aggregate_CoarsenUncoupled(ML_Aggregate *mlAggregates,
-           ML_Matrix *Amatrix, ML_Matrix **Pmatrix, ML_Comm *comm)
-{
-   MueLu_Graph     *graph;
-   graph = MueLu_BuildGraph(Amatrix,"ML_Uncoupled");
-
-   if (graph->eGraph->Comm().MyPID() == 0 && mlAggregates->printFlag  < MueLu_PrintLevel())
-       printf("ML_Aggregate_CoarsenUncoupled : \n");
-
-   MueLu_AggOptions aggregateOptions;
+                                  ML_Matrix *Amatrix, ML_Matrix **Pmatrix, ML_Comm *comm) {
+  MueLu_Graph *graph;
+  graph = MueLu_BuildGraph(Amatrix, "ML_Uncoupled");
 
-   aggregateOptions.printFlag               = mlAggregates->print_flag;
-   aggregateOptions.minNodesPerAggregate    = mlAggregates->min_nodes_per_aggregate;
-   aggregateOptions.maxNeighAlreadySelected = mlAggregates->max_neigh_already_selected;
-   aggregateOptions.ordering                = mlAggregates->ordering;
-   aggregateOptions.phase3AggCreation       = mlAggregates->phase3_agg_creation;
+  if (graph->eGraph->Comm().MyPID() == 0 && mlAggregates->printFlag < MueLu_PrintLevel())
+    printf("ML_Aggregate_CoarsenUncoupled : \n");
 
+  MueLu_AggOptions aggregateOptions;
 
-   Aggregates *aggregates = NULL;
+  aggregateOptions.printFlag               = mlAggregates->print_flag;
+  aggregateOptions.minNodesPerAggregate    = mlAggregates->min_nodes_per_aggregate;
+  aggregateOptions.maxNeighAlreadySelected = mlAggregates->max_neigh_already_selected;
+  aggregateOptions.ordering                = mlAggregates->ordering;
+  aggregateOptions.phase3AggCreation       = mlAggregates->phase3_agg_creation;
 
-   aggregates = MueLu_Aggregate_CoarsenUncoupled(&aggregateOptions,graph);
+  Aggregates *aggregates = NULL;
 
+  aggregates = MueLu_Aggregate_CoarsenUncoupled(&aggregateOptions, graph);
 
-   MueLu_AggregateLeftOvers(&aggregateOptions, aggregates, "UC_CleanUp", graph);
+  MueLu_AggregateLeftOvers(&aggregateOptions, aggregates, "UC_CleanUp", graph);
 
-//#ifdef out
-Epetra_IntVector Final( aggregates->vertex2AggId->Map() );
-for (int i = 0; i < aggregates->vertex2AggId->Map().NumMyElements(); i++)
-   Final[i] = (*(aggregates->vertex2AggId))[i] + (*(aggregates->procWinner))[i]*1000;
-printf("finals\n");
-cout << Final << endl; sleep(2);
-//#endif
+  //#ifdef out
+  Epetra_IntVector Final(aggregates->vertex2AggId->Map());
+  for (int i = 0; i < aggregates->vertex2AggId->Map().NumMyElements(); i++)
+    Final[i] = (*(aggregates->vertex2AggId))[i] + (*(aggregates->procWinner))[i] * 1000;
+  printf("finals\n");
+  cout << Final << endl;
+  sleep(2);
+  //#endif
 
-   MueLu_AggregateDestroy(aggregates);
-   MueLu_DestroyGraph(graph);
-   return 0;
+  MueLu_AggregateDestroy(aggregates);
+  MueLu_DestroyGraph(graph);
+  return 0;
 }
 
 /**********************************************************************************/
@@ -99,46 +97,45 @@ cout << Final << endl; sleep(2);
 /* extract out the Epetra_CrsGraph. My guess is that this should be changed soon  */
 /* so that the first argument is some MueLu API Matrix.                        */
 /**********************************************************************************/
-MueLu_Graph *MueLu_BuildGraph(ML_Matrix *Amatrix, char *name)
-{
+MueLu_Graph *MueLu_BuildGraph(ML_Matrix *Amatrix, char *name) {
   MueLu_Graph *graph;
   double *dtmp = NULL;
   Epetra_CrsMatrix *A;
 
-  graph = (MueLu_Graph *) malloc(sizeof(MueLu_Graph));
-  graph->eGraph     = NULL;
-  graph->name = NULL;
-  graph->name       = (char *) malloc(sizeof(char)*80); strcpy(Graph->name,name);
-  graph->nVertices  = Amatrix->invec_leng;
-
-  if ( Amatrix->getrow->nrows == 0) {
-     graph->vertexNeighbors    = NULL;
-     graph->vertexNeighborsPtr = NULL;
-     graph->nEdges             = 0;
+  graph         = (MueLu_Graph *)malloc(sizeof(MueLu_Graph));
+  graph->eGraph = NULL;
+  graph->name   = NULL;
+  graph->name   = (char *)malloc(sizeof(char) * 80);
+  strcpy(Graph->name, name);
+  graph->nVertices = Amatrix->invec_leng;
+
+  if (Amatrix->getrow->nrows == 0) {
+    graph->vertexNeighbors    = NULL;
+    graph->vertexNeighborsPtr = NULL;
+    graph->nEdges             = 0;
+  } else {
+    Epetra_ML_GetCrsDataptrs(Amatrix, &dtmp, &(graph->vertexNeighbors), &(graph->vertexNeighborsPtr));
+    if (graph->vertexNeighborsPtr == NULL) {
+      printf("MueLu_BuildGraph: Only functions for an Epetra_CrsMatrix.\n");
+      exit(1);
+    }
+    graph->nEdges = (graph->vertexNeighborsPtr)[Amatrix->getrow->Nrows];
+    Epetra_ML_GetCrsMatrix(Amatrix, (void **)&A);
+    graph->eGraph = &(A->graph());
   }
+  if (graph->eGraph == NULL)
+    graph->nGhost = 0;
   else {
-     Epetra_ML_GetCrsDataptrs(Amatrix, &dtmp, &(graph->vertexNeighbors),&(graph->vertexNeighborsPtr));
-     if ( graph->vertexNeighborsPtr == NULL) {
-        printf("MueLu_BuildGraph: Only functions for an Epetra_CrsMatrix.\n");
-        exit(1);
-      }
-      graph->nEdges      = (graph->vertexNeighborsPtr)[Amatrix->getrow->Nrows];
-      Epetra_ML_GetCrsMatrix( Amatrix, (void **) &A );
-      graph->eGraph = &(A->graph());
-  }
-  if (graph->eGraph == NULL) graph->nGhost = 0;
-  else {
-     graph->nGhost = A->RowMatrixColMap().NumMyElements() - A->MatrixDomainMap().NumMyElements();
-     if (graph->nGhost < 0) graph->nGhost = 0;
+    graph->nGhost = A->RowMatrixColMap().NumMyElements() - A->MatrixDomainMap().NumMyElements();
+    if (graph->nGhost < 0) graph->nGhost = 0;
   }
   return graph;
 }
 
-int MueLu_DestroyGraph(MueLu_Graph *graph)
-{
-   if ( graph != NULL) {
-      if (graph->name != NULL) free(graph->name);
-      free(graph);
-   }
-   return 0;
+int MueLu_DestroyGraph(MueLu_Graph *graph) {
+  if (graph != NULL) {
+    if (graph->name != NULL) free(graph->name);
+    free(graph);
+  }
+  return 0;
 }
diff --git a/packages/muelu/src/Utils/MueLu_AggregationExportFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_AggregationExportFactory_decl.hpp
index 44083fe65164..2de6446d92e2 100644
--- a/packages/muelu/src/Utils/MueLu_AggregationExportFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_AggregationExportFactory_decl.hpp
@@ -67,120 +67,119 @@
 
 namespace MueLu {
 
-  class Level;
-
-  /*!
-    @class AggregationExportFactory class.
-    @brief Factory to export aggregation info or visualize aggregates using VTK
-
-    Note, that some routines only work for 1 dof per node.
-
-    @ingroup MueLuVisualizationClasses
-
-    ## Input/output of AggregationExportFactory ##
-
-    ### User parameters of AggregationExportFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | aggregation: output filename           | string  |   |  | * |   | filename for VTK-style visualization output |
-    | aggregation: output file: time step    | int     | 0 |  | * |   | time step (overwrites '%TIMESTEP' in output file name) |
-    | aggregation: output file: iter         | int     | 0 |  | * |   | nonlinear iteration (overwrites '%ITER' in output file name) |
-    | aggregation: output file: agg style    | string  | Point Cloud |   | * |  | style of aggregation visualization for VTK output. Can be either "Point Cloud", "Jacks", or "Convex Hulls" |
-    | aggregation: output file: fine graph edges | bool | false  |   | * |  | Draw fine node connections in VTK output (only works for 1 dofs per node!) |
-    | aggregation: output file: build colormap | bool | false  |   | * |  | Output a random color map for paraView in a separate xml file. |
-    | Output filename | string |   |    | * |  | Output file name for aggregation data export (outdated, do not use) |
-    | Output file: time step | int | 0  |   | * |  | time step variable for output filename (outdated, do not use) |
-    | Output file: iter      | int | 0  |   | * |  | nonlinear iteration variable for output filename (outdated, do not use) |
-    | A | Factory | Teuchos::null  |   | * | * | Factory for A |
-    | Coordinates | Factory | Teuchos::null  |   | * | * | Factory for Coordinates (only necessary for vtk output) |
-    | Graph | Factory | Teuchos::null  |   | * | * | Factory for Graph of A (only necessary for vtk output) |
-    | Aggregates | Factory | Teuchos::null  |   | * | * | Factory for Aggregates |
-    | UnAmalgamationInfo | Factory | Teuchos::null  |   | * | * | Factory for UnAmalgamationInfo |
-    | DofsPerNode | Factory | Teuchos::null  |   | * | * | Factory for DofsPerNode |
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see AggregationExportFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see AggregationExportFactory::DeclareInput).
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class AggregationExportFactory : public TwoLevelFactoryBase, public VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+class Level;
+
+/*!
+  @class AggregationExportFactory class.
+  @brief Factory to export aggregation info or visualize aggregates using VTK
+
+  Note, that some routines only work for 1 dof per node.
+
+  @ingroup MueLuVisualizationClasses
+
+  ## Input/output of AggregationExportFactory ##
+
+  ### User parameters of AggregationExportFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | aggregation: output filename           | string  |   |  | * |   | filename for VTK-style visualization output |
+  | aggregation: output file: time step    | int     | 0 |  | * |   | time step (overwrites '%TIMESTEP' in output file name) |
+  | aggregation: output file: iter         | int     | 0 |  | * |   | nonlinear iteration (overwrites '%ITER' in output file name) |
+  | aggregation: output file: agg style    | string  | Point Cloud |   | * |  | style of aggregation visualization for VTK output. Can be either "Point Cloud", "Jacks", or "Convex Hulls" |
+  | aggregation: output file: fine graph edges | bool | false  |   | * |  | Draw fine node connections in VTK output (only works for 1 dofs per node!) |
+  | aggregation: output file: build colormap | bool | false  |   | * |  | Output a random color map for paraView in a separate xml file. |
+  | Output filename | string |   |    | * |  | Output file name for aggregation data export (outdated, do not use) |
+  | Output file: time step | int | 0  |   | * |  | time step variable for output filename (outdated, do not use) |
+  | Output file: iter      | int | 0  |   | * |  | nonlinear iteration variable for output filename (outdated, do not use) |
+  | A | Factory | Teuchos::null  |   | * | * | Factory for A |
+  | Coordinates | Factory | Teuchos::null  |   | * | * | Factory for Coordinates (only necessary for vtk output) |
+  | Graph | Factory | Teuchos::null  |   | * | * | Factory for Graph of A (only necessary for vtk output) |
+  | Aggregates | Factory | Teuchos::null  |   | * | * | Factory for Aggregates |
+  | UnAmalgamationInfo | Factory | Teuchos::null  |   | * | * | Factory for UnAmalgamationInfo |
+  | DofsPerNode | Factory | Teuchos::null  |   | * | * | Factory for DofsPerNode |
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see AggregationExportFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see AggregationExportFactory::DeclareInput).
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AggregationExportFactory : public TwoLevelFactoryBase, public VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_AGGREGATIONEXPORTFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    AggregationExportFactory() :
-      doFineGraphEdges_(false),
-      doCoarseGraphEdges_(false),
-      numNodes_(0),
-      numAggs_(0),
-      dims_(0),
-      myRank_(-1),
-      aggsOffset_(0)
-    { }
-
-    //! Destructor.
-    virtual ~AggregationExportFactory() { }
-    //@}
-
-    RCP<const ParameterList> GetValidParameterList() const;
-
-    //! Input
-    //@{
-
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
-
-    //@}
-
-    //@{
-    //! @name Build methods.
-
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
-
-    using coordinate_type       = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using CoordinateMultiVector = typename Xpetra::MultiVector<coordinate_type, LO, GO, NO>;
-
-    //@}
-
-  private:
-    //Break different viz styles into separate functions for organization:
-    void doJacksPlus_(std::vector<int>& vertices, std::vector<int>& geomSizes) const;
-    void doConvexHulls(std::vector<int>& vertices, std::vector<int>& geomSizes) const;
-    void doGraphEdges_(std::ofstream& fout, Teuchos::RCP<Matrix>& A, Teuchos::RCP<GraphBase>& G, bool fine, int dofs) const; //add geometry to display node connections from a matrix. Connections in graph but not matrix have different color.
-
-    // write VTK data
-    void writeFile_(std::ofstream& fout, std::string styleName, std::vector<int>& vertices, std::vector<int>& geomSizes) const;
-    void buildColormap_() const;
-    void writePVTU_(std::ofstream& pvtu, std::string baseFname, int numProcs) const;
-
-    static const int CONTRAST_1_ = -1;
-    static const int CONTRAST_2_ = -2;
-    static const int CONTRAST_3_ = -3;
-
-    // Data that the different styles need to have available when building geometry
-    mutable Teuchos::RCP<CoordinateMultiVector> coords_; // fine local coordinates
-    mutable Teuchos::RCP<CoordinateMultiVector> coordsCoarse_; // coarse local coordinates
-    mutable Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId_;
-    mutable Teuchos::ArrayRCP<LocalOrdinal> aggSizes_;
-    mutable std::vector<bool> isRoot_;
-    mutable bool doFineGraphEdges_;
-    mutable bool doCoarseGraphEdges_;
-    mutable int numNodes_;
-    mutable int numAggs_;
-    mutable int dims_;
-    mutable int myRank_;
-    mutable Teuchos::RCP<const Map> nodeMap_; //map used in A and Coordinates to map local ordinals to global ordinals. Need the whole map especially if it's not contiguous.
-    mutable Teuchos::RCP<const Map> nodeMapCoarse_; //Map for Ac
-    mutable int aggsOffset_;            //in a global list of aggregates, the offset of local aggregate indices
-  }; // class AggregationExportFactory
-} // namespace MueLu
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor.
+  AggregationExportFactory()
+    : doFineGraphEdges_(false)
+    , doCoarseGraphEdges_(false)
+    , numNodes_(0)
+    , numAggs_(0)
+    , dims_(0)
+    , myRank_(-1)
+    , aggsOffset_(0) {}
+
+  //! Destructor.
+  virtual ~AggregationExportFactory() {}
+  //@}
+
+  RCP<const ParameterList> GetValidParameterList() const;
+
+  //! Input
+  //@{
+
+  void DeclareInput(Level& fineLevel, Level& coarseLevel) const;
+
+  //@}
+
+  //@{
+  //! @name Build methods.
+
+  //! Build an object with this factory.
+  void Build(Level& fineLevel, Level& coarseLevel) const;
+
+  using coordinate_type       = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using CoordinateMultiVector = typename Xpetra::MultiVector<coordinate_type, LO, GO, NO>;
+
+  //@}
+
+ private:
+  // Break different viz styles into separate functions for organization:
+  void doJacksPlus_(std::vector<int>& vertices, std::vector<int>& geomSizes) const;
+  void doConvexHulls(std::vector<int>& vertices, std::vector<int>& geomSizes) const;
+  void doGraphEdges_(std::ofstream& fout, Teuchos::RCP<Matrix>& A, Teuchos::RCP<GraphBase>& G, bool fine, int dofs) const;  // add geometry to display node connections from a matrix. Connections in graph but not matrix have different color.
+
+  // write VTK data
+  void writeFile_(std::ofstream& fout, std::string styleName, std::vector<int>& vertices, std::vector<int>& geomSizes) const;
+  void buildColormap_() const;
+  void writePVTU_(std::ofstream& pvtu, std::string baseFname, int numProcs) const;
+
+  static const int CONTRAST_1_ = -1;
+  static const int CONTRAST_2_ = -2;
+  static const int CONTRAST_3_ = -3;
+
+  // Data that the different styles need to have available when building geometry
+  mutable Teuchos::RCP<CoordinateMultiVector> coords_;        // fine local coordinates
+  mutable Teuchos::RCP<CoordinateMultiVector> coordsCoarse_;  // coarse local coordinates
+  mutable Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId_;
+  mutable Teuchos::ArrayRCP<LocalOrdinal> aggSizes_;
+  mutable std::vector<bool> isRoot_;
+  mutable bool doFineGraphEdges_;
+  mutable bool doCoarseGraphEdges_;
+  mutable int numNodes_;
+  mutable int numAggs_;
+  mutable int dims_;
+  mutable int myRank_;
+  mutable Teuchos::RCP<const Map> nodeMap_;        // map used in A and Coordinates to map local ordinals to global ordinals. Need the whole map especially if it's not contiguous.
+  mutable Teuchos::RCP<const Map> nodeMapCoarse_;  // Map for Ac
+  mutable int aggsOffset_;                         // in a global list of aggregates, the offset of local aggregate indices
+};                                                 // class AggregationExportFactory
+}  // namespace MueLu
 
 #define MUELU_AGGREGATIONEXPORTFACTORY_SHORT
 
diff --git a/packages/muelu/src/Utils/MueLu_AggregationExportFactory_def.hpp b/packages/muelu/src/Utils/MueLu_AggregationExportFactory_def.hpp
index 5f083196d2e6..8fd37580b9b6 100644
--- a/packages/muelu/src/Utils/MueLu_AggregationExportFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_AggregationExportFactory_def.hpp
@@ -74,763 +74,699 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    std::string output_msg = "Output filename template (%TIMESTEP is replaced by \'Output file: time step\' variable,"
-        "%ITER is replaced by \'Output file: iter\' variable, %LEVELID is replaced level id, %PROCID is replaced by processor id)";
-    std::string output_def = "aggs_level%LEVELID_proc%PROCID.out";
+  std::string output_msg =
+      "Output filename template (%TIMESTEP is replaced by \'Output file: time step\' variable,"
+      "%ITER is replaced by \'Output file: iter\' variable, %LEVELID is replaced level id, %PROCID is replaced by processor id)";
+  std::string output_def = "aggs_level%LEVELID_proc%PROCID.out";
 
-    validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Factory for A.");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for Coordinates.");
-    validParamList->set< RCP<const FactoryBase> >("Graph", Teuchos::null, "Factory for Graph.");
-    validParamList->set< RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Factory for Aggregates.");
-    validParamList->set< RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Factory for UnAmalgamationInfo.");
-    validParamList->set< RCP<const FactoryBase> >("DofsPerNode", Teuchos::null, "Factory for DofsPerNode.");
-    // CMS/BMK: Old style factory-only options.  Deprecate me.
-    validParamList->set< std::string >           ("Output filename",           output_def, output_msg);
-    validParamList->set< int >                   ("Output file: time step",             0, "time step variable for output file name");
-    validParamList->set< int >                   ("Output file: iter",                  0, "nonlinear iteration variable for output file name");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Factory for A.");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for Coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("Graph", Teuchos::null, "Factory for Graph.");
+  validParamList->set<RCP<const FactoryBase> >("Aggregates", Teuchos::null, "Factory for Aggregates.");
+  validParamList->set<RCP<const FactoryBase> >("UnAmalgamationInfo", Teuchos::null, "Factory for UnAmalgamationInfo.");
+  validParamList->set<RCP<const FactoryBase> >("DofsPerNode", Teuchos::null, "Factory for DofsPerNode.");
+  // CMS/BMK: Old style factory-only options.  Deprecate me.
+  validParamList->set<std::string>("Output filename", output_def, output_msg);
+  validParamList->set<int>("Output file: time step", 0, "time step variable for output file name");
+  validParamList->set<int>("Output file: iter", 0, "nonlinear iteration variable for output file name");
 
-    // New-style master list options (here are same defaults as in masterList.xml)
-    validParamList->set< std::string >           ("aggregation: output filename",                    "",                    "filename for VTK-style visualization output");
-    validParamList->set< int >                   ("aggregation: output file: time step",             0,                     "time step variable for output file name");// Remove me?
-    validParamList->set< int >                   ("aggregation: output file: iter",                  0,                     "nonlinear iteration variable for output file name");//Remove me?
-    validParamList->set<std::string>             ("aggregation: output file: agg style",             "Point Cloud",         "style of aggregate visualization for VTK output");
-    validParamList->set<bool>                    ("aggregation: output file: fine graph edges",      false,                 "Whether to draw all fine node connections along with the aggregates.");
-    validParamList->set<bool>                    ("aggregation: output file: coarse graph edges",    false,                 "Whether to draw all coarse node connections along with the aggregates.");
-    validParamList->set<bool>                    ("aggregation: output file: build colormap",        false,                 "Whether to output a random colormap for ParaView in a separate XML file.");
-    return validParamList;
-  }
+  // New-style master list options (here are same defaults as in masterList.xml)
+  validParamList->set<std::string>("aggregation: output filename", "", "filename for VTK-style visualization output");
+  validParamList->set<int>("aggregation: output file: time step", 0, "time step variable for output file name");       // Remove me?
+  validParamList->set<int>("aggregation: output file: iter", 0, "nonlinear iteration variable for output file name");  // Remove me?
+  validParamList->set<std::string>("aggregation: output file: agg style", "Point Cloud", "style of aggregate visualization for VTK output");
+  validParamList->set<bool>("aggregation: output file: fine graph edges", false, "Whether to draw all fine node connections along with the aggregates.");
+  validParamList->set<bool>("aggregation: output file: coarse graph edges", false, "Whether to draw all coarse node connections along with the aggregates.");
+  validParamList->set<bool>("aggregation: output file: build colormap", false, "Whether to output a random colormap for ParaView in a separate XML file.");
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    Input(fineLevel, "Aggregates");         //< factory which created aggregates
-    Input(fineLevel, "DofsPerNode");        //< CoalesceAndDropFactory (needed for DofsPerNode variable)
-    Input(fineLevel, "UnAmalgamationInfo"); //< AmalgamationFactory (needed for UnAmalgamationInfo variable)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& fineLevel, Level& coarseLevel) const {
+  Input(fineLevel, "Aggregates");          //< factory which created aggregates
+  Input(fineLevel, "DofsPerNode");         //< CoalesceAndDropFactory (needed for DofsPerNode variable)
+  Input(fineLevel, "UnAmalgamationInfo");  //< AmalgamationFactory (needed for UnAmalgamationInfo variable)
 
-    const ParameterList & pL = GetParameterList();
-    //Only pull in coordinates if the user explicitly requests direct VTK output, so as not to break uses of old code
-    if(pL.isParameter("aggregation: output filename") && pL.get<std::string>("aggregation: output filename").length())
-    {
-      Input(fineLevel, "Coordinates");
-      Input(fineLevel, "A");
-      Input(fineLevel, "Graph");
-      if(pL.get<bool>("aggregation: output file: coarse graph edges"))
-      {
-        Input(coarseLevel, "Coordinates");
-        Input(coarseLevel, "A");
-        Input(coarseLevel, "Graph");
-      }
+  const ParameterList& pL = GetParameterList();
+  // Only pull in coordinates if the user explicitly requests direct VTK output, so as not to break uses of old code
+  if (pL.isParameter("aggregation: output filename") && pL.get<std::string>("aggregation: output filename").length()) {
+    Input(fineLevel, "Coordinates");
+    Input(fineLevel, "A");
+    Input(fineLevel, "Graph");
+    if (pL.get<bool>("aggregation: output file: coarse graph edges")) {
+      Input(coarseLevel, "Coordinates");
+      Input(coarseLevel, "A");
+      Input(coarseLevel, "Graph");
     }
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-    using namespace std;
-    //Decide which build function to follow, based on input params
-    const ParameterList& pL = GetParameterList();
-    FactoryMonitor m(*this, "AggregationExportFactory", coarseLevel);
-    Teuchos::RCP<Aggregates> aggregates      = Get< Teuchos::RCP<Aggregates> >(fineLevel,"Aggregates");
-    Teuchos::RCP<const Teuchos::Comm<int> > comm = aggregates->GetMap()->getComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-    string masterFilename = pL.get<std::string>("aggregation: output filename"); //filename parameter from master list
-    string pvtuFilename = ""; //only root processor will set this
-    string localFilename = pL.get<std::string>("Output filename");
-    string filenameToWrite;
-    bool useVTK = false;
-    doCoarseGraphEdges_ = pL.get<bool>("aggregation: output file: coarse graph edges");
-    doFineGraphEdges_ = pL.get<bool>("aggregation: output file: fine graph edges");
-    if(masterFilename.length())
-    {
-      useVTK = true;
-      filenameToWrite = masterFilename;
-      if(filenameToWrite.rfind(".vtu") == string::npos) //Must have the file extension in the name
-        filenameToWrite.append(".vtu");
-      if(numProcs > 1 && filenameToWrite.rfind("%PROCID") == string::npos) //filename can't be identical between processsors in parallel problem
-        filenameToWrite.insert(filenameToWrite.rfind(".vtu"), "-proc%PROCID");
-    }
-    else
-      filenameToWrite = localFilename;
-    LocalOrdinal DofsPerNode                 = Get< LocalOrdinal >            (fineLevel,"DofsPerNode");
-    Teuchos::RCP<AmalgamationInfo> amalgInfo = Get< RCP<AmalgamationInfo> >   (fineLevel,"UnAmalgamationInfo");
-    Teuchos::RCP<Matrix> Amat = Get<RCP<Matrix> >(fineLevel, "A");
-    Teuchos::RCP<Matrix> Ac;
-    if(doCoarseGraphEdges_)
-      Ac = Get<RCP<Matrix> >(coarseLevel, "A");
-    Teuchos::RCP<CoordinateMultiVector> coords = Teuchos::null;
-    Teuchos::RCP<CoordinateMultiVector> coordsCoarse = Teuchos::null;
-    Teuchos::RCP<GraphBase> fineGraph = Teuchos::null;
-    Teuchos::RCP<GraphBase> coarseGraph = Teuchos::null;
-    if(doFineGraphEdges_)
-      fineGraph = Get<RCP<GraphBase> >(fineLevel, "Graph");
-    if(doCoarseGraphEdges_)
-      coarseGraph = Get<RCP<GraphBase> >(coarseLevel, "Graph");
-    if(useVTK) //otherwise leave null, will not be accessed by non-vtk code
-    {
-      coords = Get<RCP<CoordinateMultiVector> >(fineLevel, "Coordinates");
-      coords_ = coords;
-      if(doCoarseGraphEdges_)
-        coordsCoarse = Get<RCP<CoordinateMultiVector> >(coarseLevel, "Coordinates");
-      dims_ = coords->getNumVectors();  //2D or 3D?
-      if(numProcs > 1)
-      {
-        if (aggregates->AggregatesCrossProcessors())
-        { // Do we want to use the map from aggregates here instead of the map from A? Using the map from A seems to be problematic with multiple dofs per node
-          RCP<Import> coordImporter = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coords->getMap(), Amat->getColMap());
-          RCP<CoordinateMultiVector> ghostedCoords = Xpetra::MultiVectorFactory<coordinate_type, LocalOrdinal, GlobalOrdinal, Node>::Build(Amat->getColMap(), dims_);
-          ghostedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
-          coords = ghostedCoords;
-          coords_ = ghostedCoords;
-        }
-        if(doCoarseGraphEdges_)
-        {
-          RCP<Import> coordImporter = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coordsCoarse->getMap(), Ac->getColMap());
-          RCP<CoordinateMultiVector> ghostedCoords = Xpetra::MultiVectorFactory<coordinate_type, LocalOrdinal, GlobalOrdinal, Node>::Build(Ac->getColMap(), dims_);
-          ghostedCoords->doImport(*coordsCoarse, *coordImporter, Xpetra::INSERT);
-          coordsCoarse = ghostedCoords;
-          coordsCoarse_ = ghostedCoords;
-        }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& fineLevel, Level& coarseLevel) const {
+  using namespace std;
+  // Decide which build function to follow, based on input params
+  const ParameterList& pL = GetParameterList();
+  FactoryMonitor m(*this, "AggregationExportFactory", coarseLevel);
+  Teuchos::RCP<Aggregates> aggregates          = Get<Teuchos::RCP<Aggregates> >(fineLevel, "Aggregates");
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = aggregates->GetMap()->getComm();
+  int numProcs                                 = comm->getSize();
+  int myRank                                   = comm->getRank();
+  string masterFilename                        = pL.get<std::string>("aggregation: output filename");  // filename parameter from master list
+  string pvtuFilename                          = "";                                                   // only root processor will set this
+  string localFilename                         = pL.get<std::string>("Output filename");
+  string filenameToWrite;
+  bool useVTK         = false;
+  doCoarseGraphEdges_ = pL.get<bool>("aggregation: output file: coarse graph edges");
+  doFineGraphEdges_   = pL.get<bool>("aggregation: output file: fine graph edges");
+  if (masterFilename.length()) {
+    useVTK          = true;
+    filenameToWrite = masterFilename;
+    if (filenameToWrite.rfind(".vtu") == string::npos)  // Must have the file extension in the name
+      filenameToWrite.append(".vtu");
+    if (numProcs > 1 && filenameToWrite.rfind("%PROCID") == string::npos)  // filename can't be identical between processsors in parallel problem
+      filenameToWrite.insert(filenameToWrite.rfind(".vtu"), "-proc%PROCID");
+  } else
+    filenameToWrite = localFilename;
+  LocalOrdinal DofsPerNode                 = Get<LocalOrdinal>(fineLevel, "DofsPerNode");
+  Teuchos::RCP<AmalgamationInfo> amalgInfo = Get<RCP<AmalgamationInfo> >(fineLevel, "UnAmalgamationInfo");
+  Teuchos::RCP<Matrix> Amat                = Get<RCP<Matrix> >(fineLevel, "A");
+  Teuchos::RCP<Matrix> Ac;
+  if (doCoarseGraphEdges_)
+    Ac = Get<RCP<Matrix> >(coarseLevel, "A");
+  Teuchos::RCP<CoordinateMultiVector> coords       = Teuchos::null;
+  Teuchos::RCP<CoordinateMultiVector> coordsCoarse = Teuchos::null;
+  Teuchos::RCP<GraphBase> fineGraph                = Teuchos::null;
+  Teuchos::RCP<GraphBase> coarseGraph              = Teuchos::null;
+  if (doFineGraphEdges_)
+    fineGraph = Get<RCP<GraphBase> >(fineLevel, "Graph");
+  if (doCoarseGraphEdges_)
+    coarseGraph = Get<RCP<GraphBase> >(coarseLevel, "Graph");
+  if (useVTK)  // otherwise leave null, will not be accessed by non-vtk code
+  {
+    coords  = Get<RCP<CoordinateMultiVector> >(fineLevel, "Coordinates");
+    coords_ = coords;
+    if (doCoarseGraphEdges_)
+      coordsCoarse = Get<RCP<CoordinateMultiVector> >(coarseLevel, "Coordinates");
+    dims_ = coords->getNumVectors();  // 2D or 3D?
+    if (numProcs > 1) {
+      if (aggregates->AggregatesCrossProcessors()) {  // Do we want to use the map from aggregates here instead of the map from A? Using the map from A seems to be problematic with multiple dofs per node
+        RCP<Import> coordImporter                = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coords->getMap(), Amat->getColMap());
+        RCP<CoordinateMultiVector> ghostedCoords = Xpetra::MultiVectorFactory<coordinate_type, LocalOrdinal, GlobalOrdinal, Node>::Build(Amat->getColMap(), dims_);
+        ghostedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
+        coords  = ghostedCoords;
+        coords_ = ghostedCoords;
+      }
+      if (doCoarseGraphEdges_) {
+        RCP<Import> coordImporter                = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coordsCoarse->getMap(), Ac->getColMap());
+        RCP<CoordinateMultiVector> ghostedCoords = Xpetra::MultiVectorFactory<coordinate_type, LocalOrdinal, GlobalOrdinal, Node>::Build(Ac->getColMap(), dims_);
+        ghostedCoords->doImport(*coordsCoarse, *coordImporter, Xpetra::INSERT);
+        coordsCoarse  = ghostedCoords;
+        coordsCoarse_ = ghostedCoords;
       }
     }
-    GetOStream(Runtime0) << "AggregationExportFactory: DofsPerNode: " << DofsPerNode << std::endl;
-    Teuchos::RCP<LocalOrdinalMultiVector> vertex2AggId_vector = aggregates->GetVertex2AggId();
-    Teuchos::RCP<LocalOrdinalVector>      procWinner_vector   = aggregates->GetProcWinner();
-    Teuchos::ArrayRCP<LocalOrdinal>       vertex2AggId        = aggregates->GetVertex2AggId()->getDataNonConst(0);
-    Teuchos::ArrayRCP<LocalOrdinal>       procWinner          = aggregates->GetProcWinner()->getDataNonConst(0);
+  }
+  GetOStream(Runtime0) << "AggregationExportFactory: DofsPerNode: " << DofsPerNode << std::endl;
+  Teuchos::RCP<LocalOrdinalMultiVector> vertex2AggId_vector = aggregates->GetVertex2AggId();
+  Teuchos::RCP<LocalOrdinalVector> procWinner_vector        = aggregates->GetProcWinner();
+  Teuchos::ArrayRCP<LocalOrdinal> vertex2AggId              = aggregates->GetVertex2AggId()->getDataNonConst(0);
+  Teuchos::ArrayRCP<LocalOrdinal> procWinner                = aggregates->GetProcWinner()->getDataNonConst(0);
 
-    vertex2AggId_ = vertex2AggId;
+  vertex2AggId_ = vertex2AggId;
 
-    // prepare for calculating global aggregate ids
-    std::vector<GlobalOrdinal> numAggsGlobal (numProcs, 0);
-    std::vector<GlobalOrdinal> numAggsLocal  (numProcs, 0);
-    std::vector<GlobalOrdinal> minGlobalAggId(numProcs, 0);
+  // prepare for calculating global aggregate ids
+  std::vector<GlobalOrdinal> numAggsGlobal(numProcs, 0);
+  std::vector<GlobalOrdinal> numAggsLocal(numProcs, 0);
+  std::vector<GlobalOrdinal> minGlobalAggId(numProcs, 0);
 
-    numAggsLocal[myRank] = aggregates->GetNumAggregates();
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, numProcs, &numAggsLocal[0], &numAggsGlobal[0]);
-    for (int i = 1; i < Teuchos::as<int>(numAggsGlobal.size()); ++i)
-    {
-      numAggsGlobal [i] += numAggsGlobal[i-1];
-      minGlobalAggId[i]  = numAggsGlobal[i-1];
-    }
-    if(numProcs == 0)
-      aggsOffset_ = 0;
-    else
-      aggsOffset_ = minGlobalAggId[myRank];
-    ArrayRCP<LO>            aggStart;
-    ArrayRCP<GlobalOrdinal> aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-    int timeStep = pL.get< int > ("Output file: time step");
-    int iter = pL.get< int > ("Output file: iter");
-    filenameToWrite = this->replaceAll(filenameToWrite, "%LEVELID",  toString(fineLevel.GetLevelID()));
-    filenameToWrite = this->replaceAll(filenameToWrite, "%TIMESTEP", toString(timeStep));
-    filenameToWrite = this->replaceAll(filenameToWrite, "%ITER",     toString(iter));
-    //Proc id MUST be included in vtu filenames to distinguish them (if multiple procs)
-    //In all other cases (else), including processor # in filename is optional
-    string masterStem = "";
-    if(useVTK)
-    {
-      masterStem = filenameToWrite.substr(0, filenameToWrite.rfind(".vtu"));
-      masterStem = this->replaceAll(masterStem, "%PROCID", "");
-    }
-    pvtuFilename = masterStem + "-master.pvtu";
-    string baseFname = filenameToWrite;  //get a version of the filename string with the %PROCID token, but without substituting myRank (needed for pvtu output)
-    filenameToWrite = this->replaceAll(filenameToWrite, "%PROCID", toString(myRank));
-    GetOStream(Runtime0) << "AggregationExportFactory: outputfile \"" << filenameToWrite << "\"" << std::endl;
-    ofstream fout(filenameToWrite.c_str());
-    GO numAggs = aggregates->GetNumAggregates();
-    if(!useVTK)
-    {
-      GO indexBase = aggregates->GetMap()->getIndexBase(); // extract indexBase from overlapping map within aggregates structure. The indexBase is constant throughout the whole simulation (either 0 = C++ or 1 = Fortran)
-      GO offset    = amalgInfo->GlobalOffset();            // extract offset for global dof ids
-      vector<GlobalOrdinal> nodeIds;
-      for (int i = 0; i < numAggs; ++i) {
-        fout << "Agg " << minGlobalAggId[myRank] + i << " Proc " << myRank << ":";
+  numAggsLocal[myRank] = aggregates->GetNumAggregates();
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, numProcs, &numAggsLocal[0], &numAggsGlobal[0]);
+  for (int i = 1; i < Teuchos::as<int>(numAggsGlobal.size()); ++i) {
+    numAggsGlobal[i] += numAggsGlobal[i - 1];
+    minGlobalAggId[i] = numAggsGlobal[i - 1];
+  }
+  if (numProcs == 0)
+    aggsOffset_ = 0;
+  else
+    aggsOffset_ = minGlobalAggId[myRank];
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GlobalOrdinal> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+  int timeStep    = pL.get<int>("Output file: time step");
+  int iter        = pL.get<int>("Output file: iter");
+  filenameToWrite = this->replaceAll(filenameToWrite, "%LEVELID", toString(fineLevel.GetLevelID()));
+  filenameToWrite = this->replaceAll(filenameToWrite, "%TIMESTEP", toString(timeStep));
+  filenameToWrite = this->replaceAll(filenameToWrite, "%ITER", toString(iter));
+  // Proc id MUST be included in vtu filenames to distinguish them (if multiple procs)
+  // In all other cases (else), including processor # in filename is optional
+  string masterStem = "";
+  if (useVTK) {
+    masterStem = filenameToWrite.substr(0, filenameToWrite.rfind(".vtu"));
+    masterStem = this->replaceAll(masterStem, "%PROCID", "");
+  }
+  pvtuFilename     = masterStem + "-master.pvtu";
+  string baseFname = filenameToWrite;  // get a version of the filename string with the %PROCID token, but without substituting myRank (needed for pvtu output)
+  filenameToWrite  = this->replaceAll(filenameToWrite, "%PROCID", toString(myRank));
+  GetOStream(Runtime0) << "AggregationExportFactory: outputfile \"" << filenameToWrite << "\"" << std::endl;
+  ofstream fout(filenameToWrite.c_str());
+  GO numAggs = aggregates->GetNumAggregates();
+  if (!useVTK) {
+    GO indexBase = aggregates->GetMap()->getIndexBase();  // extract indexBase from overlapping map within aggregates structure. The indexBase is constant throughout the whole simulation (either 0 = C++ or 1 = Fortran)
+    GO offset    = amalgInfo->GlobalOffset();             // extract offset for global dof ids
+    vector<GlobalOrdinal> nodeIds;
+    for (int i = 0; i < numAggs; ++i) {
+      fout << "Agg " << minGlobalAggId[myRank] + i << " Proc " << myRank << ":";
 
-        // TODO: Use k+=DofsPerNode instead of ++k and get rid of std::unique call afterwards
-        for (int k = aggStart[i]; k < aggStart[i+1]; ++k) {
-          nodeIds.push_back((aggToRowMap[k] - offset - indexBase) / DofsPerNode + indexBase);
-        }
+      // TODO: Use k+=DofsPerNode instead of ++k and get rid of std::unique call afterwards
+      for (int k = aggStart[i]; k < aggStart[i + 1]; ++k) {
+        nodeIds.push_back((aggToRowMap[k] - offset - indexBase) / DofsPerNode + indexBase);
+      }
 
-        // remove duplicate entries from nodeids
-        std::sort(nodeIds.begin(), nodeIds.end());
-        typename std::vector<GlobalOrdinal>::iterator endLocation = std::unique(nodeIds.begin(), nodeIds.end());
-        nodeIds.erase(endLocation, nodeIds.end());
+      // remove duplicate entries from nodeids
+      std::sort(nodeIds.begin(), nodeIds.end());
+      typename std::vector<GlobalOrdinal>::iterator endLocation = std::unique(nodeIds.begin(), nodeIds.end());
+      nodeIds.erase(endLocation, nodeIds.end());
 
-        // print out nodeids
-        for(typename std::vector<GlobalOrdinal>::iterator printIt = nodeIds.begin(); printIt != nodeIds.end(); printIt++)
-          fout << " " << *printIt;
-        nodeIds.clear();
-        fout << std::endl;
-      }
+      // print out nodeids
+      for (typename std::vector<GlobalOrdinal>::iterator printIt = nodeIds.begin(); printIt != nodeIds.end(); printIt++)
+        fout << " " << *printIt;
+      nodeIds.clear();
+      fout << std::endl;
     }
-    else
-    {
-      using namespace std;
-      //Make sure we have coordinates
-      TEUCHOS_TEST_FOR_EXCEPTION(coords.is_null(), Exceptions::RuntimeError,"AggExportFactory could not get coordinates, but they are required for VTK output.");
-      numAggs_ = numAggs;
-      numNodes_ = coords->getLocalLength();
-      //Get the sizes of the aggregates to speed up grabbing node IDs
-      aggSizes_ = aggregates->ComputeAggregateSizesArrayRCP();
-      myRank_ = myRank;
-      string aggStyle = "Point Cloud";
-      try
-      {
-        aggStyle = pL.get<string>("aggregation: output file: agg style"); //Let "Point Cloud" be the default style
-      }
-      catch(std::exception& e) {}
-      vector<int> vertices;
-      vector<int> geomSizes;
-      string indent = "";
-      nodeMap_ = Amat->getMap();
-      for(LocalOrdinal i = 0; i < numNodes_; i++)
-      {
-        isRoot_.push_back(aggregates->IsRoot(i));
-      }
-      //If problem is serial and not outputting fine nor coarse graph edges, don't make pvtu file
-      //Otherwise create it
-      if(myRank == 0 && (numProcs != 1 || doCoarseGraphEdges_ || doFineGraphEdges_))
-      {
-        ofstream pvtu(pvtuFilename.c_str());
-        writePVTU_(pvtu, baseFname, numProcs);
-        pvtu.close();
-      }
-      if(aggStyle == "Point Cloud")
-        this->doPointCloud(vertices, geomSizes, numAggs_, numNodes_);
-      else if(aggStyle == "Jacks")
-        this->doJacks(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_);
-      else if(aggStyle == "Jacks++") //Not actually implemented
-        doJacksPlus_(vertices, geomSizes);
-      else if(aggStyle == "Convex Hulls")
-        doConvexHulls(vertices, geomSizes);
-      else
-      {
-        GetOStream(Warnings0) << "   Unrecognized agg style.\n   Possible values are Point Cloud, Jacks, Jacks++, and Convex Hulls.\n   Defaulting to Point Cloud." << std::endl;
-        aggStyle = "Point Cloud";
-        this->doPointCloud(vertices, geomSizes, numAggs_, numNodes_);
-      }
-      writeFile_(fout, aggStyle, vertices, geomSizes);
-      if(doCoarseGraphEdges_)
-      {
-        string fname = filenameToWrite;
-        string cEdgeFile = fname.insert(fname.rfind(".vtu"), "-coarsegraph");
-        std::ofstream edgeStream(cEdgeFile.c_str());
-        doGraphEdges_(edgeStream, Ac, coarseGraph, false, DofsPerNode);
-      }
-      if(doFineGraphEdges_)
-      {
-        string fname = filenameToWrite;
-        string fEdgeFile = fname.insert(fname.rfind(".vtu"), "-finegraph");
-        std::ofstream edgeStream(fEdgeFile.c_str());
-        doGraphEdges_(edgeStream, Amat, fineGraph, true, DofsPerNode);
-      }
-      if(myRank == 0 && pL.get<bool>("aggregation: output file: build colormap"))
-      {
-        buildColormap_();
-      }
+  } else {
+    using namespace std;
+    // Make sure we have coordinates
+    TEUCHOS_TEST_FOR_EXCEPTION(coords.is_null(), Exceptions::RuntimeError, "AggExportFactory could not get coordinates, but they are required for VTK output.");
+    numAggs_  = numAggs;
+    numNodes_ = coords->getLocalLength();
+    // Get the sizes of the aggregates to speed up grabbing node IDs
+    aggSizes_       = aggregates->ComputeAggregateSizesArrayRCP();
+    myRank_         = myRank;
+    string aggStyle = "Point Cloud";
+    try {
+      aggStyle = pL.get<string>("aggregation: output file: agg style");  // Let "Point Cloud" be the default style
+    } catch (std::exception& e) {
+    }
+    vector<int> vertices;
+    vector<int> geomSizes;
+    string indent = "";
+    nodeMap_      = Amat->getMap();
+    for (LocalOrdinal i = 0; i < numNodes_; i++) {
+      isRoot_.push_back(aggregates->IsRoot(i));
+    }
+    // If problem is serial and not outputting fine nor coarse graph edges, don't make pvtu file
+    // Otherwise create it
+    if (myRank == 0 && (numProcs != 1 || doCoarseGraphEdges_ || doFineGraphEdges_)) {
+      ofstream pvtu(pvtuFilename.c_str());
+      writePVTU_(pvtu, baseFname, numProcs);
+      pvtu.close();
+    }
+    if (aggStyle == "Point Cloud")
+      this->doPointCloud(vertices, geomSizes, numAggs_, numNodes_);
+    else if (aggStyle == "Jacks")
+      this->doJacks(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_);
+    else if (aggStyle == "Jacks++")  // Not actually implemented
+      doJacksPlus_(vertices, geomSizes);
+    else if (aggStyle == "Convex Hulls")
+      doConvexHulls(vertices, geomSizes);
+    else {
+      GetOStream(Warnings0) << "   Unrecognized agg style.\n   Possible values are Point Cloud, Jacks, Jacks++, and Convex Hulls.\n   Defaulting to Point Cloud." << std::endl;
+      aggStyle = "Point Cloud";
+      this->doPointCloud(vertices, geomSizes, numAggs_, numNodes_);
+    }
+    writeFile_(fout, aggStyle, vertices, geomSizes);
+    if (doCoarseGraphEdges_) {
+      string fname     = filenameToWrite;
+      string cEdgeFile = fname.insert(fname.rfind(".vtu"), "-coarsegraph");
+      std::ofstream edgeStream(cEdgeFile.c_str());
+      doGraphEdges_(edgeStream, Ac, coarseGraph, false, DofsPerNode);
+    }
+    if (doFineGraphEdges_) {
+      string fname     = filenameToWrite;
+      string fEdgeFile = fname.insert(fname.rfind(".vtu"), "-finegraph");
+      std::ofstream edgeStream(fEdgeFile.c_str());
+      doGraphEdges_(edgeStream, Amat, fineGraph, true, DofsPerNode);
+    }
+    if (myRank == 0 && pL.get<bool>("aggregation: output file: build colormap")) {
+      buildColormap_();
     }
-    fout.close();
   }
+  fout.close();
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doJacksPlus_(std::vector<int>& /* vertices */, std::vector<int>& /* geomSizes */) const
-  {
-    //TODO
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doJacksPlus_(std::vector<int>& /* vertices */, std::vector<int>& /* geomSizes */) const {
+  // TODO
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls(std::vector<int>& vertices, std::vector<int>& geomSizes) const
-  {
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls(std::vector<int>& vertices, std::vector<int>& geomSizes) const {
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
 
-    if(dims_ == 2) {  
-      this->doConvexHulls2D(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_, xCoords, yCoords);
-    } else {
-      zCoords = coords_->getData(2);
-      this->doConvexHulls3D(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_, xCoords, yCoords, zCoords);
-    }
+  if (dims_ == 2) {
+    this->doConvexHulls2D(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_, xCoords, yCoords);
+  } else {
+    zCoords = coords_->getData(2);
+    this->doConvexHulls3D(vertices, geomSizes, numAggs_, numNodes_, isRoot_, vertex2AggId_, xCoords, yCoords, zCoords);
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doGraphEdges_(std::ofstream& fout, Teuchos::RCP<Matrix>& A, Teuchos::RCP<GraphBase>& G, bool fine, int dofs) const
-  {
-    using namespace std;
-    ArrayView<const Scalar> values;
-    ArrayView<const LocalOrdinal> neighbors;
-    //Allow two different colors of connections (by setting "aggregates" scalar to CONTRAST_1 or CONTRAST_2)
-    vector<pair<int, int> > vert1; //vertices (node indices)
-    vector<pair<int, int> > vert2; //size of every cell is assumed to be 2 vertices, since all edges are drawn as lines
-    
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
-    if(dims_ == 3)
-      zCoords = coords_->getData(2);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doGraphEdges_(std::ofstream& fout, Teuchos::RCP<Matrix>& A, Teuchos::RCP<GraphBase>& G, bool fine, int dofs) const {
+  using namespace std;
+  ArrayView<const Scalar> values;
+  ArrayView<const LocalOrdinal> neighbors;
+  // Allow two different colors of connections (by setting "aggregates" scalar to CONTRAST_1 or CONTRAST_2)
+  vector<pair<int, int> > vert1;  // vertices (node indices)
+  vector<pair<int, int> > vert2;  // size of every cell is assumed to be 2 vertices, since all edges are drawn as lines
 
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cx = coordsCoarse_->getData(0);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cy = coordsCoarse_->getData(1);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cz = Teuchos::null;
-    if(dims_ == 3)
-      cz = coordsCoarse_->getData(2);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
+  if (dims_ == 3)
+    zCoords = coords_->getData(2);
 
-    if(A->isGloballyIndexed())
-    {
-      ArrayView<const GlobalOrdinal> indices;
-      for(GlobalOrdinal globRow = 0; globRow < GlobalOrdinal(A->getGlobalNumRows()); globRow++)
-      {
-        if(dofs == 1)
-          A->getGlobalRowView(globRow, indices, values);
-        neighbors = G->getNeighborVertices((LocalOrdinal) globRow);
-        int gEdge = 0;
-        int aEdge = 0;
-        while(gEdge != int(neighbors.size()))
-        {
-          if(dofs == 1)
-          {
-            if(neighbors[gEdge] == indices[aEdge])
-            {
-              //graph and matrix both have this edge, wasn't filtered, show as color 1
-              vert1.push_back(pair<int, int>(int(globRow), neighbors[gEdge]));
-              gEdge++;
-              aEdge++;
-            }
-            else
-            {
-              //graph contains an edge at gEdge which was filtered from A, show as color 2
-              vert2.push_back(pair<int, int>(int(globRow), neighbors[gEdge]));
-              gEdge++;
-            }
-          }
-          else //for multiple DOF problems, don't try to detect filtered edges and ignore A
-          {
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cx = coordsCoarse_->getData(0);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cy = coordsCoarse_->getData(1);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> cz = Teuchos::null;
+  if (dims_ == 3)
+    cz = coordsCoarse_->getData(2);
+
+  if (A->isGloballyIndexed()) {
+    ArrayView<const GlobalOrdinal> indices;
+    for (GlobalOrdinal globRow = 0; globRow < GlobalOrdinal(A->getGlobalNumRows()); globRow++) {
+      if (dofs == 1)
+        A->getGlobalRowView(globRow, indices, values);
+      neighbors = G->getNeighborVertices((LocalOrdinal)globRow);
+      int gEdge = 0;
+      int aEdge = 0;
+      while (gEdge != int(neighbors.size())) {
+        if (dofs == 1) {
+          if (neighbors[gEdge] == indices[aEdge]) {
+            // graph and matrix both have this edge, wasn't filtered, show as color 1
             vert1.push_back(pair<int, int>(int(globRow), neighbors[gEdge]));
             gEdge++;
+            aEdge++;
+          } else {
+            // graph contains an edge at gEdge which was filtered from A, show as color 2
+            vert2.push_back(pair<int, int>(int(globRow), neighbors[gEdge]));
+            gEdge++;
           }
+        } else  // for multiple DOF problems, don't try to detect filtered edges and ignore A
+        {
+          vert1.push_back(pair<int, int>(int(globRow), neighbors[gEdge]));
+          gEdge++;
         }
       }
     }
-    else
-    {
-      ArrayView<const LocalOrdinal> indices;
-      for(LocalOrdinal locRow = 0; locRow < LocalOrdinal(A->getLocalNumRows()); locRow++)
-      {
-        if(dofs == 1)
-          A->getLocalRowView(locRow, indices, values);
-        neighbors = G->getNeighborVertices(locRow);
-        //Add those local indices (columns) to the list of connections (which are pairs of the form (localM, localN))
-        int gEdge = 0;
-        int aEdge = 0;
-        while(gEdge != int(neighbors.size()))
-        {
-          if(dofs == 1)
-          {
-            if(neighbors[gEdge] == indices[aEdge])
-            {
-              vert1.push_back(pair<int, int>(locRow, neighbors[gEdge]));
-              gEdge++;
-              aEdge++;
-            }
-            else
-            {
-              vert2.push_back(pair<int, int>(locRow, neighbors[gEdge]));
-              gEdge++;
-            }
-          }
-          else
-          {
+  } else {
+    ArrayView<const LocalOrdinal> indices;
+    for (LocalOrdinal locRow = 0; locRow < LocalOrdinal(A->getLocalNumRows()); locRow++) {
+      if (dofs == 1)
+        A->getLocalRowView(locRow, indices, values);
+      neighbors = G->getNeighborVertices(locRow);
+      // Add those local indices (columns) to the list of connections (which are pairs of the form (localM, localN))
+      int gEdge = 0;
+      int aEdge = 0;
+      while (gEdge != int(neighbors.size())) {
+        if (dofs == 1) {
+          if (neighbors[gEdge] == indices[aEdge]) {
             vert1.push_back(pair<int, int>(locRow, neighbors[gEdge]));
             gEdge++;
+            aEdge++;
+          } else {
+            vert2.push_back(pair<int, int>(locRow, neighbors[gEdge]));
+            gEdge++;
           }
+        } else {
+          vert1.push_back(pair<int, int>(locRow, neighbors[gEdge]));
+          gEdge++;
         }
       }
     }
-    for(size_t i = 0; i < vert1.size(); i ++)
-    {
-      if(vert1[i].first > vert1[i].second)
-      {
-        int temp = vert1[i].first;
-        vert1[i].first = vert1[i].second;
-        vert1[i].second = temp;
-      }
-    }
-    for(size_t i = 0; i < vert2.size(); i++)
-    {
-      if(vert2[i].first > vert2[i].second)
-      {
-        int temp = vert2[i].first;
-        vert2[i].first = vert2[i].second;
-        vert2[i].second = temp;
-      }
-    }
-    sort(vert1.begin(), vert1.end());
-    vector<pair<int, int> >::iterator newEnd = unique(vert1.begin(), vert1.end()); //remove duplicate edges
-    vert1.erase(newEnd, vert1.end());
-    sort(vert2.begin(), vert2.end());
-    newEnd = unique(vert2.begin(), vert2.end());
-    vert2.erase(newEnd, vert2.end());
-    vector<int> points1;
-    points1.reserve(2 * vert1.size());
-    for(size_t i = 0; i < vert1.size(); i++)
-    {
-      points1.push_back(vert1[i].first);
-      points1.push_back(vert1[i].second);
-    }
-    vector<int> points2;
-    points2.reserve(2 * vert2.size());
-    for(size_t i = 0; i < vert2.size(); i++)
-    {
-      points2.push_back(vert2[i].first);
-      points2.push_back(vert2[i].second);
-    }
-    vector<int> unique1 = this->makeUnique(points1);
-    vector<int> unique2 = this->makeUnique(points2);
-    fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
-    fout << "  <UnstructuredGrid>" << endl;
-    fout << "    <Piece NumberOfPoints=\"" << unique1.size() + unique2.size() << "\" NumberOfCells=\"" << vert1.size() + vert2.size() << "\">" << endl;
-    fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << endl; //node and aggregate will be set to CONTRAST_1|2, but processor will have its actual value
-    string indent = "          ";
-    fout << indent;
-    for(size_t i = 0; i < unique1.size(); i++)
-    {
-      fout << CONTRAST_1_ << " ";
-      if(i % 25 == 24)
-        fout << endl << indent;
-    }
-    for(size_t i = 0; i < unique2.size(); i++)
-    {
-      fout << CONTRAST_2_ << " ";
-      if((i + 2 * vert1.size()) % 25 == 24)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < unique1.size(); i++)
-    {
-      fout << CONTRAST_1_ << " ";
-      if(i % 25 == 24)
-        fout << endl << indent;
-    }
-    for(size_t i = 0; i < unique2.size(); i++)
-    {
-      fout << CONTRAST_2_ << " ";
-      if((i + 2 * vert2.size()) % 25 == 24)
-        fout << endl << indent;
+  }
+  for (size_t i = 0; i < vert1.size(); i++) {
+    if (vert1[i].first > vert1[i].second) {
+      int temp        = vert1[i].first;
+      vert1[i].first  = vert1[i].second;
+      vert1[i].second = temp;
     }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < unique1.size() + unique2.size(); i++)
-    {
-      fout << myRank_ << " ";
-      if(i % 25 == 24)
-        fout << endl << indent;
+  }
+  for (size_t i = 0; i < vert2.size(); i++) {
+    if (vert2[i].first > vert2[i].second) {
+      int temp        = vert2[i].first;
+      vert2[i].first  = vert2[i].second;
+      vert2[i].second = temp;
     }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "      </PointData>" << endl;
-    fout << "      <Points>" << endl;
-    fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < unique1.size(); i++)
-    {
-      if(fine)
-      {
-        fout << xCoords[unique1[i]] << " " << yCoords[unique1[i]] << " ";
-        if(dims_ == 3)
-          fout << zCoords[unique1[i]] << " ";
-        else
-          fout << "0 ";
-        if(i % 2)
-          fout << endl << indent;
-      }
+  }
+  sort(vert1.begin(), vert1.end());
+  vector<pair<int, int> >::iterator newEnd = unique(vert1.begin(), vert1.end());  // remove duplicate edges
+  vert1.erase(newEnd, vert1.end());
+  sort(vert2.begin(), vert2.end());
+  newEnd = unique(vert2.begin(), vert2.end());
+  vert2.erase(newEnd, vert2.end());
+  vector<int> points1;
+  points1.reserve(2 * vert1.size());
+  for (size_t i = 0; i < vert1.size(); i++) {
+    points1.push_back(vert1[i].first);
+    points1.push_back(vert1[i].second);
+  }
+  vector<int> points2;
+  points2.reserve(2 * vert2.size());
+  for (size_t i = 0; i < vert2.size(); i++) {
+    points2.push_back(vert2[i].first);
+    points2.push_back(vert2[i].second);
+  }
+  vector<int> unique1 = this->makeUnique(points1);
+  vector<int> unique2 = this->makeUnique(points2);
+  fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
+  fout << "  <UnstructuredGrid>" << endl;
+  fout << "    <Piece NumberOfPoints=\"" << unique1.size() + unique2.size() << "\" NumberOfCells=\"" << vert1.size() + vert2.size() << "\">" << endl;
+  fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << endl;  // node and aggregate will be set to CONTRAST_1|2, but processor will have its actual value
+  string indent = "          ";
+  fout << indent;
+  for (size_t i = 0; i < unique1.size(); i++) {
+    fout << CONTRAST_1_ << " ";
+    if (i % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  for (size_t i = 0; i < unique2.size(); i++) {
+    fout << CONTRAST_2_ << " ";
+    if ((i + 2 * vert1.size()) % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < unique1.size(); i++) {
+    fout << CONTRAST_1_ << " ";
+    if (i % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  for (size_t i = 0; i < unique2.size(); i++) {
+    fout << CONTRAST_2_ << " ";
+    if ((i + 2 * vert2.size()) % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < unique1.size() + unique2.size(); i++) {
+    fout << myRank_ << " ";
+    if (i % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </PointData>" << endl;
+  fout << "      <Points>" << endl;
+  fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < unique1.size(); i++) {
+    if (fine) {
+      fout << xCoords[unique1[i]] << " " << yCoords[unique1[i]] << " ";
+      if (dims_ == 3)
+        fout << zCoords[unique1[i]] << " ";
       else
-      {
-        fout << cx[unique1[i]] << " " << cy[unique1[i]] << " ";
-        if(dims_ == 3)
-          fout << cz[unique1[i]] << " ";
-        else
-          fout << "0 ";
-        if(i % 2)
-          fout << endl << indent;
-      }
-    }
-    for(size_t i = 0; i < unique2.size(); i++)
-    {
-      if(fine)
-      {
-        fout << xCoords[unique2[i]] << " " << yCoords[unique2[i]] << " ";
-        if(dims_ == 3)
-          fout << zCoords[unique2[i]] << " ";
-        else
-          fout << "0 ";
-        if(i % 2)
-          fout << endl << indent;
-      }
+        fout << "0 ";
+      if (i % 2)
+        fout << endl
+             << indent;
+    } else {
+      fout << cx[unique1[i]] << " " << cy[unique1[i]] << " ";
+      if (dims_ == 3)
+        fout << cz[unique1[i]] << " ";
       else
-      {
-        fout << cx[unique2[i]] << " " << cy[unique2[i]] << " ";
-        if(dims_ == 3)
-          fout << cz[unique2[i]] << " ";
-        else
-          fout << "0 ";
-        if((i + unique1.size()) % 2)
-          fout << endl << indent;
-      }
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "      </Points>" << endl;
-    fout << "      <Cells>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < points1.size(); i++)
-    {
-      fout << points1[i] << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    for(size_t i = 0; i < points2.size(); i++)
-    {
-      fout << points2[i] + unique1.size() << " ";
-      if((i + points1.size()) % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>"  << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << endl;
-    fout << indent;
-    int offset = 0;
-    for(size_t i = 0; i < vert1.size() + vert2.size(); i++)
-    {
-      offset += 2;
-      fout << offset << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>"  << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < vert1.size() + vert2.size(); i++)
-    {
-      fout << "3 ";
-      if(i % 25 == 24)
-        fout << endl << indent;
+        fout << "0 ";
+      if (i % 2)
+        fout << endl
+             << indent;
     }
-    fout << endl;
-    fout << "        </DataArray>"  << endl;
-    fout << "      </Cells>" << endl;
-    fout << "    </Piece>" << endl;
-    fout << "  </UnstructuredGrid>" << endl;
-    fout << "</VTKFile>" << endl;
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFile_(std::ofstream& fout, std::string styleName, std::vector<int>& vertices, std::vector<int>& geomSizes) const
-  {
-    using namespace std;
-    
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
-    if(dims_ == 3)
-      zCoords = coords_->getData(2);
-    
-    vector<int> uniqueFine = this->makeUnique(vertices);
-    string indent = "      ";
-    fout << "<!--" << styleName << " Aggregates Visualization-->" << endl;
-    fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
-    fout << "  <UnstructuredGrid>" << endl;
-    fout << "    <Piece NumberOfPoints=\"" << uniqueFine.size() << "\" NumberOfCells=\"" << geomSizes.size() << "\">" << endl;
-    fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << endl;
-    indent = "          ";
-    fout << indent;
-    bool localIsGlobal = GlobalOrdinal(nodeMap_->getGlobalNumElements()) == GlobalOrdinal(nodeMap_->getLocalNumElements());
-    for(size_t i = 0; i < uniqueFine.size(); i++)
-    {
-      if(localIsGlobal)
-      {
-        fout << uniqueFine[i] << " "; //if all nodes are on this processor, do not map from local to global
-      }
-      else
-        fout << nodeMap_->getGlobalElement(uniqueFine[i]) << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < uniqueFine.size(); i++)
-    {
-      if(vertex2AggId_[uniqueFine[i]]==-1)
-        fout << vertex2AggId_[uniqueFine[i]] << " ";
+  for (size_t i = 0; i < unique2.size(); i++) {
+    if (fine) {
+      fout << xCoords[unique2[i]] << " " << yCoords[unique2[i]] << " ";
+      if (dims_ == 3)
+        fout << zCoords[unique2[i]] << " ";
       else
-        fout << aggsOffset_ + vertex2AggId_[uniqueFine[i]] << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < uniqueFine.size(); i++)
-    {
-      fout << myRank_ << " ";
-      if(i % 20 == 19)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "      </PointData>" << endl;
-    fout << "      <Points>" << endl;
-    fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < uniqueFine.size(); i++)
-    {
-      fout << xCoords[uniqueFine[i]] << " " << yCoords[uniqueFine[i]] << " ";
-      if(dims_ == 2)
         fout << "0 ";
+      if (i % 2)
+        fout << endl
+             << indent;
+    } else {
+      fout << cx[unique2[i]] << " " << cy[unique2[i]] << " ";
+      if (dims_ == 3)
+        fout << cz[unique2[i]] << " ";
       else
-        fout << zCoords[uniqueFine[i]] << " ";
-      if(i % 3 == 2)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "      </Points>" << endl;
-    fout << "      <Cells>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < vertices.size(); i++)
-    {
-      fout << vertices[i] << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << endl;
-    fout << indent;
-    int accum = 0;
-    for(size_t i = 0; i < geomSizes.size(); i++)
-    {
-      accum += geomSizes[i];
-      fout << accum << " ";
-      if(i % 10 == 9)
-        fout << endl << indent;
-    }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << endl;
-    fout << indent;
-    for(size_t i = 0; i < geomSizes.size(); i++)
-    {
-      switch(geomSizes[i])
-      {
-        case 1:
-          fout << "1 "; //Point
-          break;
-        case 2:
-          fout << "3 "; //Line
-          break;
-        case 3:
-          fout << "5 "; //Triangle
-          break;
-        default:
-          fout << "7 "; //Polygon
-      }
-      if(i % 30 == 29)
-        fout << endl << indent;
+        fout << "0 ";
+      if ((i + unique1.size()) % 2)
+        fout << endl
+             << indent;
     }
-    fout << endl;
-    fout << "        </DataArray>" << endl;
-    fout << "      </Cells>" << endl;
-    fout << "    </Piece>" << endl;
-    fout << "  </UnstructuredGrid>" << endl;
-    fout << "</VTKFile>" << endl;
   }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </Points>" << endl;
+  fout << "      <Cells>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < points1.size(); i++) {
+    fout << points1[i] << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  for (size_t i = 0; i < points2.size(); i++) {
+    fout << points2[i] + unique1.size() << " ";
+    if ((i + points1.size()) % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << endl;
+  fout << indent;
+  int offset = 0;
+  for (size_t i = 0; i < vert1.size() + vert2.size(); i++) {
+    offset += 2;
+    fout << offset << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < vert1.size() + vert2.size(); i++) {
+    fout << "3 ";
+    if (i % 25 == 24)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </Cells>" << endl;
+  fout << "    </Piece>" << endl;
+  fout << "  </UnstructuredGrid>" << endl;
+  fout << "</VTKFile>" << endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildColormap_() const
-  {
-    using namespace std;
-    try
-    {
-      ofstream color("random-colormap.xml");
-      color << "<ColorMap name=\"MueLu-Random\" space=\"RGB\">" << endl;
-      //Give -1, -2, -3 distinctive colors (so that the style functions can have constrasted geometry types)
-      //Do red, orange, yellow to constrast with cool color spectrum for other types
-      color << "  <Point x=\"" << CONTRAST_1_ << "\" o=\"1\" r=\"1\" g=\"0\" b=\"0\"/>" << endl;
-      color << "  <Point x=\"" << CONTRAST_2_ << "\" o=\"1\" r=\"1\" g=\"0.6\" b=\"0\"/>" << endl;
-      color << "  <Point x=\"" << CONTRAST_3_ << "\" o=\"1\" r=\"1\" g=\"1\" b=\"0\"/>" << endl;
-      srand(time(NULL));
-      for(int i = 0; i < 5000; i += 4)
-      {
-        color << "  <Point x=\"" << i << "\" o=\"1\" r=\"" << (rand() % 50) / 256.0 << "\" g=\"" << (rand() % 256) / 256.0 << "\" b=\"" << (rand() % 256) / 256.0 << "\"/>" << endl;
-      }
-      color << "</ColorMap>" << endl;
-      color.close();
-    }
-    catch(std::exception& e)
-    {
-      GetOStream(Warnings0) << "   Error while building colormap file: " << e.what() << endl;
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFile_(std::ofstream& fout, std::string styleName, std::vector<int>& vertices, std::vector<int>& geomSizes) const {
+  using namespace std;
+
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = coords_->getData(0);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = coords_->getData(1);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
+  if (dims_ == 3)
+    zCoords = coords_->getData(2);
+
+  vector<int> uniqueFine = this->makeUnique(vertices);
+  string indent          = "      ";
+  fout << "<!--" << styleName << " Aggregates Visualization-->" << endl;
+  fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
+  fout << "  <UnstructuredGrid>" << endl;
+  fout << "    <Piece NumberOfPoints=\"" << uniqueFine.size() << "\" NumberOfCells=\"" << geomSizes.size() << "\">" << endl;
+  fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << endl;
+  indent = "          ";
+  fout << indent;
+  bool localIsGlobal = GlobalOrdinal(nodeMap_->getGlobalNumElements()) == GlobalOrdinal(nodeMap_->getLocalNumElements());
+  for (size_t i = 0; i < uniqueFine.size(); i++) {
+    if (localIsGlobal) {
+      fout << uniqueFine[i] << " ";  // if all nodes are on this processor, do not map from local to global
+    } else
+      fout << nodeMap_->getGlobalElement(uniqueFine[i]) << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < uniqueFine.size(); i++) {
+    if (vertex2AggId_[uniqueFine[i]] == -1)
+      fout << vertex2AggId_[uniqueFine[i]] << " ";
+    else
+      fout << aggsOffset_ + vertex2AggId_[uniqueFine[i]] << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < uniqueFine.size(); i++) {
+    fout << myRank_ << " ";
+    if (i % 20 == 19)
+      fout << endl
+           << indent;
   }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </PointData>" << endl;
+  fout << "      <Points>" << endl;
+  fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < uniqueFine.size(); i++) {
+    fout << xCoords[uniqueFine[i]] << " " << yCoords[uniqueFine[i]] << " ";
+    if (dims_ == 2)
+      fout << "0 ";
+    else
+      fout << zCoords[uniqueFine[i]] << " ";
+    if (i % 3 == 2)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </Points>" << endl;
+  fout << "      <Cells>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < vertices.size(); i++) {
+    fout << vertices[i] << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << endl;
+  fout << indent;
+  int accum = 0;
+  for (size_t i = 0; i < geomSizes.size(); i++) {
+    accum += geomSizes[i];
+    fout << accum << " ";
+    if (i % 10 == 9)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << endl;
+  fout << indent;
+  for (size_t i = 0; i < geomSizes.size(); i++) {
+    switch (geomSizes[i]) {
+      case 1:
+        fout << "1 ";  // Point
+        break;
+      case 2:
+        fout << "3 ";  // Line
+        break;
+      case 3:
+        fout << "5 ";  // Triangle
+        break;
+      default:
+        fout << "7 ";  // Polygon
+    }
+    if (i % 30 == 29)
+      fout << endl
+           << indent;
+  }
+  fout << endl;
+  fout << "        </DataArray>" << endl;
+  fout << "      </Cells>" << endl;
+  fout << "    </Piece>" << endl;
+  fout << "  </UnstructuredGrid>" << endl;
+  fout << "</VTKFile>" << endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writePVTU_(std::ofstream& pvtu, std::string baseFname, int numProcs) const
-  {
-    using namespace std;
-    //If using vtk, filenameToWrite now contains final, correct ***.vtu filename (for the current proc)
-    //So the root proc will need to use its own filenameToWrite to make a list of the filenames of all other procs to put in
-    //pvtu file.
-    pvtu << "<VTKFile type=\"PUnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
-    pvtu << "  <PUnstructuredGrid GhostLevel=\"0\">" << endl;
-    pvtu << "    <PPointData Scalars=\"Node Aggregate Processor\">" << endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Node\"/>" << endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Aggregate\"/>" << endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Processor\"/>" << endl;
-    pvtu << "    </PPointData>" << endl;
-    pvtu << "    <PPoints>" << endl;
-    pvtu << "      <PDataArray type=\"Float64\" NumberOfComponents=\"3\"/>" << endl;
-    pvtu << "    </PPoints>" << endl;
-    for(int i = 0; i < numProcs; i++)
-    {
-      //specify the piece for each proc (the replaceAll expression matches what the filenames will be on other procs)
-      pvtu << "    <Piece Source=\"" << this->replaceAll(baseFname, "%PROCID", toString(i)) << "\"/>" << endl;
-    }
-    //reference files with graph pieces, if applicable
-    if(doFineGraphEdges_)
-    {
-      for(int i = 0; i < numProcs; i++)
-      {
-        string fn = this->replaceAll(baseFname, "%PROCID", toString(i));
-        pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-finegraph") << "\"/>" << endl;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildColormap_() const {
+  using namespace std;
+  try {
+    ofstream color("random-colormap.xml");
+    color << "<ColorMap name=\"MueLu-Random\" space=\"RGB\">" << endl;
+    // Give -1, -2, -3 distinctive colors (so that the style functions can have constrasted geometry types)
+    // Do red, orange, yellow to constrast with cool color spectrum for other types
+    color << "  <Point x=\"" << CONTRAST_1_ << "\" o=\"1\" r=\"1\" g=\"0\" b=\"0\"/>" << endl;
+    color << "  <Point x=\"" << CONTRAST_2_ << "\" o=\"1\" r=\"1\" g=\"0.6\" b=\"0\"/>" << endl;
+    color << "  <Point x=\"" << CONTRAST_3_ << "\" o=\"1\" r=\"1\" g=\"1\" b=\"0\"/>" << endl;
+    srand(time(NULL));
+    for (int i = 0; i < 5000; i += 4) {
+      color << "  <Point x=\"" << i << "\" o=\"1\" r=\"" << (rand() % 50) / 256.0 << "\" g=\"" << (rand() % 256) / 256.0 << "\" b=\"" << (rand() % 256) / 256.0 << "\"/>" << endl;
+    }
+    color << "</ColorMap>" << endl;
+    color.close();
+  } catch (std::exception& e) {
+    GetOStream(Warnings0) << "   Error while building colormap file: " << e.what() << endl;
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AggregationExportFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writePVTU_(std::ofstream& pvtu, std::string baseFname, int numProcs) const {
+  using namespace std;
+  // If using vtk, filenameToWrite now contains final, correct ***.vtu filename (for the current proc)
+  // So the root proc will need to use its own filenameToWrite to make a list of the filenames of all other procs to put in
+  // pvtu file.
+  pvtu << "<VTKFile type=\"PUnstructuredGrid\" byte_order=\"LittleEndian\">" << endl;
+  pvtu << "  <PUnstructuredGrid GhostLevel=\"0\">" << endl;
+  pvtu << "    <PPointData Scalars=\"Node Aggregate Processor\">" << endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Node\"/>" << endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Aggregate\"/>" << endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Processor\"/>" << endl;
+  pvtu << "    </PPointData>" << endl;
+  pvtu << "    <PPoints>" << endl;
+  pvtu << "      <PDataArray type=\"Float64\" NumberOfComponents=\"3\"/>" << endl;
+  pvtu << "    </PPoints>" << endl;
+  for (int i = 0; i < numProcs; i++) {
+    // specify the piece for each proc (the replaceAll expression matches what the filenames will be on other procs)
+    pvtu << "    <Piece Source=\"" << this->replaceAll(baseFname, "%PROCID", toString(i)) << "\"/>" << endl;
+  }
+  // reference files with graph pieces, if applicable
+  if (doFineGraphEdges_) {
+    for (int i = 0; i < numProcs; i++) {
+      string fn = this->replaceAll(baseFname, "%PROCID", toString(i));
+      pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-finegraph") << "\"/>" << endl;
     }
-    if(doCoarseGraphEdges_)
-    {
-      for(int i = 0; i < numProcs; i++)
-      {
-        string fn = this->replaceAll(baseFname, "%PROCID", toString(i));
-        pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-coarsegraph") << "\"/>" << endl;
-      }
+  }
+  if (doCoarseGraphEdges_) {
+    for (int i = 0; i < numProcs; i++) {
+      string fn = this->replaceAll(baseFname, "%PROCID", toString(i));
+      pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-coarsegraph") << "\"/>" << endl;
     }
-    pvtu << "  </PUnstructuredGrid>" << endl;
-    pvtu << "</VTKFile>" << endl;
-    pvtu.close();
   }
-} // namespace MueLu
+  pvtu << "  </PUnstructuredGrid>" << endl;
+  pvtu << "</VTKFile>" << endl;
+  pvtu.close();
+}
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONEXPORTFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_decl.hpp b/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_decl.hpp
index 0608db028771..e8f58db7c62a 100644
--- a/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_decl.hpp
@@ -23,85 +23,84 @@
 namespace MueLu {
 
 // template struct for comparing pairs
-template<class Scalar = DefaultScalar, class LocalOrdinal = DefaultGlobalOrdinal>
+template <class Scalar = DefaultScalar, class LocalOrdinal = DefaultGlobalOrdinal>
 struct CompPairs {
-  CompPairs(const std::vector<Scalar> & v) : vinternal_(v) {}
+  CompPairs(const std::vector<Scalar>& v)
+    : vinternal_(v) {}
   std::vector<Scalar> vinternal_;
   bool operator()(LocalOrdinal a, LocalOrdinal b) {
-    //return vinternal_[a] < vinternal_[b];
+    // return vinternal_[a] < vinternal_[b];
     return Teuchos::ScalarTraits<Scalar>::magnitude(vinternal_[a]) > Teuchos::ScalarTraits<Scalar>::magnitude(vinternal_[b]);
   }
 };
 
 // template function for comparison
-template<class Scalar, class LocalOrdinal>
-CompPairs<Scalar,LocalOrdinal> CreateCmpPairs(const std::vector<Scalar> & v) {
-  return CompPairs<Scalar,LocalOrdinal>(v);
+template <class Scalar, class LocalOrdinal>
+CompPairs<Scalar, LocalOrdinal> CreateCmpPairs(const std::vector<Scalar>& v) {
+  return CompPairs<Scalar, LocalOrdinal>(v);
 }
 
 // template function for sorting permutations
-template<class Scalar, class LocalOrdinal>
-void sortingPermutation(const std::vector<Scalar> & values, std::vector<LocalOrdinal> & v) {
+template <class Scalar, class LocalOrdinal>
+void sortingPermutation(const std::vector<Scalar>& values, std::vector<LocalOrdinal>& v) {
   size_t size = values.size();
-  v.clear(); v.reserve(size);
-  for(size_t i=0; i<size; ++i)
+  v.clear();
+  v.reserve(size);
+  for (size_t i = 0; i < size; ++i)
     v.push_back(i);
 
-  std::sort(v.begin(),v.end(), MueLu::CreateCmpPairs<Scalar,LocalOrdinal>(values));
+  std::sort(v.begin(), v.end(), MueLu::CreateCmpPairs<Scalar, LocalOrdinal>(values));
 }
 
-  //! @brief Algebraic permutation strategy
-  /*!
-     This class permutes columns of a input matrix A trying to make A
-     a diagonal dominant matrix.
+//! @brief Algebraic permutation strategy
+/*!
+   This class permutes columns of a input matrix A trying to make A
+   a diagonal dominant matrix.
 
-    */
+  */
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class AlgebraicPermutationStrategy : public BaseClass {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class AlgebraicPermutationStrategy : public BaseClass {
 #undef MUELU_ALGEBRAICPERMUTATIONSTRATEGY_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-
+ public:
   /*!
     @class AlgebraicPermutationStrategy class.
     @brief Class which defines local permutations of matrix columns.
     */
 
-    //! @name build permutation methods.
-    //@{
-
-    //! build permutation operators
-    /*!
-     *  The following variables produced
-     *  "A"     :      permuted and scaled A
-     *  "permA" :      permuted A without scaling
-     *  "permP" :      permutation opertor (should be identity)
-     *  "permQT":      transpose permutation operators
-     *  "permScaling": scaling operator
-     *
-     * \param A: input matrix (input)
-     * \param permRowMap: Dof row map permutation shall be restricted on (input)
-     * \param currentLevel: only for output of variables
-     * \param genFactory: const pointer to generating (calling) PermutationFactory
-     // TODO avoid this, not very elegant.
-     // Decide which variables have to be generated, give them back per reference to the PermutationFactory.
-     */
-    void BuildPermutation(const Teuchos::RCP<Matrix> & A, const Teuchos::RCP<const Map>& permRowMap,
-                          Level & currentLevel, const FactoryBase* genFactory) const;
-
-    //@}
+  //! @name build permutation methods.
+  //@{
 
+  //! build permutation operators
+  /*!
+   *  The following variables produced
+   *  "A"     :      permuted and scaled A
+   *  "permA" :      permuted A without scaling
+   *  "permP" :      permutation opertor (should be identity)
+   *  "permQT":      transpose permutation operators
+   *  "permScaling": scaling operator
+   *
+   * \param A: input matrix (input)
+   * \param permRowMap: Dof row map permutation shall be restricted on (input)
+   * \param currentLevel: only for output of variables
+   * \param genFactory: const pointer to generating (calling) PermutationFactory
+   // TODO avoid this, not very elegant.
+   // Decide which variables have to be generated, give them back per reference to the PermutationFactory.
+   */
+  void BuildPermutation(const Teuchos::RCP<Matrix>& A, const Teuchos::RCP<const Map>& permRowMap,
+                        Level& currentLevel, const FactoryBase* genFactory) const;
+
+  //@}
+
+ private:
+};
 
-  private:
-   };
-
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_ALGEBRAICPERMUTATIONSTRATEGY_SHORT
 
-
 #endif /* MUELU_ALGEBRAICPERMUTATIONSTRATEGY_DECL_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_def.hpp b/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_def.hpp
index 716be4f485eb..668ada8ff06f 100644
--- a/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_AlgebraicPermutationStrategy_def.hpp
@@ -30,715 +30,700 @@
 
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void AlgebraicPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  BuildPermutation(const Teuchos::RCP<Matrix> & A, const Teuchos::RCP<const Map>& permRowMap,
-                   Level & currentLevel, const FactoryBase* genFactory) const {
-
-    const Scalar SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
-    const Scalar SC_ONE  = Teuchos::ScalarTraits<Scalar>::one();
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    const MT MT_ZERO = Teuchos::ScalarTraits<MT>::zero();
-    const MT MT_ONE  = Teuchos::ScalarTraits<MT>::one();
-
-    const Teuchos::RCP< const Teuchos::Comm< int > > comm = A->getRowMap()->getComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    size_t nDofsPerNode = 1;
-    if (A->IsView("stridedMaps")) {
-      Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
-      nDofsPerNode = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AlgebraicPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    BuildPermutation(const Teuchos::RCP<Matrix>& A, const Teuchos::RCP<const Map>& permRowMap,
+                     Level& currentLevel, const FactoryBase* genFactory) const {
+  const Scalar SC_ZERO = Teuchos::ScalarTraits<Scalar>::zero();
+  const Scalar SC_ONE  = Teuchos::ScalarTraits<Scalar>::one();
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  const MT MT_ZERO = Teuchos::ScalarTraits<MT>::zero();
+  const MT MT_ONE  = Teuchos::ScalarTraits<MT>::one();
+
+  const Teuchos::RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+  int numProcs                                       = comm->getSize();
+  int myRank                                         = comm->getRank();
+
+  size_t nDofsPerNode = 1;
+  if (A->IsView("stridedMaps")) {
+    Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
+    nDofsPerNode                              = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
+  }
 
-    std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > permutedDiagCandidates;
-    std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > keepDiagonalEntries;
-    std::vector<MT> Weights;
+  std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > permutedDiagCandidates;
+  std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > keepDiagonalEntries;
+  std::vector<MT> Weights;
 
-    // loop over all local rows in matrix A and keep diagonal entries if corresponding
-    // matrix rows are not contained in permRowMap
-    for (size_t row = 0; row < A->getRowMap()->getLocalNumElements(); row++) {
-      GlobalOrdinal grow = A->getRowMap()->getGlobalElement(row);
+  // loop over all local rows in matrix A and keep diagonal entries if corresponding
+  // matrix rows are not contained in permRowMap
+  for (size_t row = 0; row < A->getRowMap()->getLocalNumElements(); row++) {
+    GlobalOrdinal grow = A->getRowMap()->getGlobalElement(row);
 
-      if(permRowMap->isNodeGlobalElement(grow) == true) continue;
+    if (permRowMap->isNodeGlobalElement(grow) == true) continue;
 
-      size_t nnz = A->getNumEntriesInLocalRow(row);
+    size_t nnz = A->getNumEntriesInLocalRow(row);
 
-      // extract local row information from matrix
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals;
-      A->getLocalRowView(row, indices, vals);
+    // extract local row information from matrix
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    A->getLocalRowView(row, indices, vals);
 
-      TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError,
-                                 "MueLu::PermutationFactory::Build: number of nonzeros not equal to number of indices? Error.");
+    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError,
+                               "MueLu::PermutationFactory::Build: number of nonzeros not equal to number of indices? Error.");
 
-      // find column entry with max absolute value
-      GlobalOrdinal gMaxValIdx = 0;
-      MT norm1  = MT_ZERO;
-      MT maxVal = MT_ZERO;
-      for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
-        norm1 += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-        if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
-          maxVal = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-          gMaxValIdx = A->getColMap()->getGlobalElement(indices[j]);
-        }
+    // find column entry with max absolute value
+    GlobalOrdinal gMaxValIdx = 0;
+    MT norm1                 = MT_ZERO;
+    MT maxVal                = MT_ZERO;
+    for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
+      norm1 += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
+        maxVal     = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+        gMaxValIdx = A->getColMap()->getGlobalElement(indices[j]);
       }
-
-      if(grow == gMaxValIdx) // only keep row/col pair if it's diagonal dominant!!!
-        keepDiagonalEntries.push_back(std::make_pair(grow,grow));
     }
 
-    //////////
-    // handle rows that are marked to be relevant for permutations
-    for (size_t row = 0; row < permRowMap->getLocalNumElements(); row++) {
-      GlobalOrdinal grow = permRowMap->getGlobalElement(row);
-      LocalOrdinal lArow = A->getRowMap()->getLocalElement(grow);
-      size_t nnz = A->getNumEntriesInLocalRow(lArow);
-
-      // extract local row information from matrix
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals;
-      A->getLocalRowView(lArow, indices, vals);
-
-      TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError,
-                                 "MueLu::PermutationFactory::Build: number of nonzeros not equal to number of indices? Error.");
-
-      // find column entry with max absolute value
-      GlobalOrdinal gMaxValIdx = 0;
-      MT norm1  = MT_ZERO;
-      MT maxVal = MT_ZERO;
-      for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
-        norm1 += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-        if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
-          maxVal = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-          gMaxValIdx = A->getColMap()->getGlobalElement(indices[j]);
-        }
-      }
+    if (grow == gMaxValIdx)  // only keep row/col pair if it's diagonal dominant!!!
+      keepDiagonalEntries.push_back(std::make_pair(grow, grow));
+  }
 
-      if(maxVal > Teuchos::ScalarTraits<MT>::zero ()) { // keep only max Entries \neq 0.0
-        permutedDiagCandidates.push_back(std::make_pair(grow,gMaxValIdx));
-        Weights.push_back(maxVal/(norm1*Teuchos::as<MT>(nnz)));
-      } else {
-        std::cout << "ATTENTION: row " << grow << " has only zero entries -> singular matrix!" << std::endl;
+  //////////
+  // handle rows that are marked to be relevant for permutations
+  for (size_t row = 0; row < permRowMap->getLocalNumElements(); row++) {
+    GlobalOrdinal grow = permRowMap->getGlobalElement(row);
+    LocalOrdinal lArow = A->getRowMap()->getLocalElement(grow);
+    size_t nnz         = A->getNumEntriesInLocalRow(lArow);
+
+    // extract local row information from matrix
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    A->getLocalRowView(lArow, indices, vals);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError,
+                               "MueLu::PermutationFactory::Build: number of nonzeros not equal to number of indices? Error.");
+
+    // find column entry with max absolute value
+    GlobalOrdinal gMaxValIdx = 0;
+    MT norm1                 = MT_ZERO;
+    MT maxVal                = MT_ZERO;
+    for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
+      norm1 += Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
+        maxVal     = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
+        gMaxValIdx = A->getColMap()->getGlobalElement(indices[j]);
       }
-
     }
 
-    // sort Weights in descending order
-    std::vector<int> permutation;
-    sortingPermutation(Weights,permutation);
-
-    // create new vector with exactly one possible entry for each column
+    if (maxVal > Teuchos::ScalarTraits<MT>::zero()) {  // keep only max Entries \neq 0.0
+      permutedDiagCandidates.push_back(std::make_pair(grow, gMaxValIdx));
+      Weights.push_back(maxVal / (norm1 * Teuchos::as<MT>(nnz)));
+    } else {
+      std::cout << "ATTENTION: row " << grow << " has only zero entries -> singular matrix!" << std::endl;
+    }
+  }
 
-    // each processor which requests the global column id gcid adds 1 to gColVec
-    // gColVec will be summed up over all processors and communicated to gDomVec
-    // which is based on the non-overlapping domain map of A.
+  // sort Weights in descending order
+  std::vector<int> permutation;
+  sortingPermutation(Weights, permutation);
 
-    Teuchos::RCP<Vector> gColVec = VectorFactory::Build(A->getColMap());
-    Teuchos::RCP<Vector> gDomVec = VectorFactory::Build(A->getDomainMap());
-    gColVec->putScalar(SC_ZERO);
-    gDomVec->putScalar(SC_ZERO);
+  // create new vector with exactly one possible entry for each column
 
-    // put in all keep diagonal entries
-    for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = keepDiagonalEntries.begin(); p != keepDiagonalEntries.end(); ++p) {
-      gColVec->sumIntoGlobalValue((*p).second,1.0);
-    }
+  // each processor which requests the global column id gcid adds 1 to gColVec
+  // gColVec will be summed up over all processors and communicated to gDomVec
+  // which is based on the non-overlapping domain map of A.
 
-    Teuchos::RCP<Export> exporter = ExportFactory::Build(gColVec->getMap(), gDomVec->getMap());
-    gDomVec->doExport(*gColVec,*exporter,Xpetra::ADD);  // communicate blocked gcolids to all procs
-    gColVec->doImport(*gDomVec,*exporter,Xpetra::INSERT);
+  Teuchos::RCP<Vector> gColVec = VectorFactory::Build(A->getColMap());
+  Teuchos::RCP<Vector> gDomVec = VectorFactory::Build(A->getDomainMap());
+  gColVec->putScalar(SC_ZERO);
+  gDomVec->putScalar(SC_ZERO);
 
-    std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > permutedDiagCandidatesFiltered; // TODO reserve memory
-    std::map<GlobalOrdinal, MT> gColId2Weight;
+  // put in all keep diagonal entries
+  for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = keepDiagonalEntries.begin(); p != keepDiagonalEntries.end(); ++p) {
+    gColVec->sumIntoGlobalValue((*p).second, 1.0);
+  }
 
-    {
-      Teuchos::ArrayRCP< Scalar > ddata = gColVec->getDataNonConst(0);
-      for(size_t i = 0; i < permutedDiagCandidates.size(); ++i) {
-        // loop over all candidates
-        std::pair<GlobalOrdinal, GlobalOrdinal> pp = permutedDiagCandidates[permutation[i]];
-        GlobalOrdinal grow = pp.first;
-        GlobalOrdinal gcol = pp.second;
-
-        LocalOrdinal lcol = A->getColMap()->getLocalElement(gcol);
-        if(Teuchos::ScalarTraits<Scalar>::real (ddata[lcol]) > MT_ZERO){
-          continue; // skip lcol: column already handled by another row
-        }
+  Teuchos::RCP<Export> exporter = ExportFactory::Build(gColVec->getMap(), gDomVec->getMap());
+  gDomVec->doExport(*gColVec, *exporter, Xpetra::ADD);  // communicate blocked gcolids to all procs
+  gColVec->doImport(*gDomVec, *exporter, Xpetra::INSERT);
 
-        // mark column as already taken
-        ddata[lcol] += SC_ONE;
+  std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > permutedDiagCandidatesFiltered;  // TODO reserve memory
+  std::map<GlobalOrdinal, MT> gColId2Weight;
 
-        permutedDiagCandidatesFiltered.push_back(std::make_pair(grow,gcol));
-        gColId2Weight[gcol] = Weights[permutation[i]];
-      }
-    }
+  {
+    Teuchos::ArrayRCP<Scalar> ddata = gColVec->getDataNonConst(0);
+    for (size_t i = 0; i < permutedDiagCandidates.size(); ++i) {
+      // loop over all candidates
+      std::pair<GlobalOrdinal, GlobalOrdinal> pp = permutedDiagCandidates[permutation[i]];
+      GlobalOrdinal grow                         = pp.first;
+      GlobalOrdinal gcol                         = pp.second;
 
-    // communicate how often each column index is requested by the different procs
-    gDomVec->doExport(*gColVec,*exporter,Xpetra::ADD);
-    gColVec->doImport(*gDomVec,*exporter,Xpetra::INSERT); // probably not needed // TODO check me
-
-    //*****************************************************************************************
-    // first communicate ALL global ids of column indices which are requested by more
-    // than one proc to all other procs
-    // detect which global col indices are requested by more than one proc
-    // and store them in the multipleColRequests vector
-    std::vector<GlobalOrdinal> multipleColRequests; // store all global column indices from current processor that are also
-    // requested by another processor. This is possible, since they are stored
-    // in gDomVec which is based on the nonoverlapping domain map. That is, each
-    // global col id is handled by exactly one proc.
-    std::queue<GlobalOrdinal> unusedColIdx; // unused column indices on current processor
-
-    for(size_t sz = 0; sz < gDomVec->getLocalLength(); ++sz) {
-      Teuchos::ArrayRCP< const Scalar > arrDomVec = gDomVec->getData(0);
-      //
-      // FIXME (mfh 30 Oct 2015) I _think_ it's OK to check just the
-      // real part, because this is a count.  (Shouldn't MueLu use
-      // mag_type instead of Scalar here to save space?)
-      //
-      if(Teuchos::ScalarTraits<Scalar>::real(arrDomVec[sz]) > MT_ONE) {
-        multipleColRequests.push_back(gDomVec->getMap()->getGlobalElement(sz));
-      } else if(Teuchos::ScalarTraits<Scalar>::real (arrDomVec[sz]) == MT_ZERO) {
-        unusedColIdx.push(gDomVec->getMap()->getGlobalElement(sz));
+      LocalOrdinal lcol = A->getColMap()->getLocalElement(gcol);
+      if (Teuchos::ScalarTraits<Scalar>::real(ddata[lcol]) > MT_ZERO) {
+        continue;  // skip lcol: column already handled by another row
       }
-    }
 
-    // communicate the global number of column indices which are requested by more than one proc
-    LocalOrdinal localMultColRequests  = Teuchos::as<LocalOrdinal>(multipleColRequests.size());
-    LocalOrdinal globalMultColRequests = 0;
-
-    // sum up all entries in multipleColRequests over all processors
-      //
-      // FIXME (lbv 11 Feb 2019) why cast localMultColRequests as LocalOrdinal
-      // when it was properly declared as such just a line above?
-      //
-    MueLu_sumAll(gDomVec->getMap()->getComm(), (LocalOrdinal)localMultColRequests, globalMultColRequests);
-
-    if(globalMultColRequests > 0) {
-      // special handling: two processors request the same global column id.
-      // decide which processor gets it
-
-      // distribute number of multipleColRequests to all processors
-      // each processor stores how many column ids for exchange are handled by the cur proc
-      std::vector<GlobalOrdinal> numMyMultColRequests(numProcs, 0);
-      std::vector<GlobalOrdinal> numGlobalMultColRequests(numProcs, 0);
-      numMyMultColRequests[myRank] = localMultColRequests;
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
-                         numMyMultColRequests.data(),
-                         numGlobalMultColRequests.data());
+      // mark column as already taken
+      ddata[lcol] += SC_ONE;
 
-      // communicate multipleColRequests entries to all processors
-      int nMyOffset = 0;
-      for (int i=0; i<myRank-1; i++)
-        nMyOffset += numGlobalMultColRequests[i]; // calculate offset to store the weights on the corresponding place in procOverlappingWeights
+      permutedDiagCandidatesFiltered.push_back(std::make_pair(grow, gcol));
+      gColId2Weight[gcol] = Weights[permutation[i]];
+    }
+  }
 
-      const GlobalOrdinal zero = 0;
-      std::vector<GlobalOrdinal> procMultRequestedColIds(globalMultColRequests, zero);
-      std::vector<GlobalOrdinal> global_procMultRequestedColIds(globalMultColRequests, zero);
+  // communicate how often each column index is requested by the different procs
+  gDomVec->doExport(*gColVec, *exporter, Xpetra::ADD);
+  gColVec->doImport(*gDomVec, *exporter, Xpetra::INSERT);  // probably not needed // TODO check me
+
+  //*****************************************************************************************
+  // first communicate ALL global ids of column indices which are requested by more
+  // than one proc to all other procs
+  // detect which global col indices are requested by more than one proc
+  // and store them in the multipleColRequests vector
+  std::vector<GlobalOrdinal> multipleColRequests;  // store all global column indices from current processor that are also
+  // requested by another processor. This is possible, since they are stored
+  // in gDomVec which is based on the nonoverlapping domain map. That is, each
+  // global col id is handled by exactly one proc.
+  std::queue<GlobalOrdinal> unusedColIdx;  // unused column indices on current processor
+
+  for (size_t sz = 0; sz < gDomVec->getLocalLength(); ++sz) {
+    Teuchos::ArrayRCP<const Scalar> arrDomVec = gDomVec->getData(0);
+    //
+    // FIXME (mfh 30 Oct 2015) I _think_ it's OK to check just the
+    // real part, because this is a count.  (Shouldn't MueLu use
+    // mag_type instead of Scalar here to save space?)
+    //
+    if (Teuchos::ScalarTraits<Scalar>::real(arrDomVec[sz]) > MT_ONE) {
+      multipleColRequests.push_back(gDomVec->getMap()->getGlobalElement(sz));
+    } else if (Teuchos::ScalarTraits<Scalar>::real(arrDomVec[sz]) == MT_ZERO) {
+      unusedColIdx.push(gDomVec->getMap()->getGlobalElement(sz));
+    }
+  }
 
-      // loop over all local column GIDs that are also requested by other procs
-      for(size_t i = 0; i < multipleColRequests.size(); i++) {
-        procMultRequestedColIds[nMyOffset + i] = multipleColRequests[i]; // all weights are > 0 ?
-      }
+  // communicate the global number of column indices which are requested by more than one proc
+  LocalOrdinal localMultColRequests  = Teuchos::as<LocalOrdinal>(multipleColRequests.size());
+  LocalOrdinal globalMultColRequests = 0;
+
+  // sum up all entries in multipleColRequests over all processors
+  //
+  // FIXME (lbv 11 Feb 2019) why cast localMultColRequests as LocalOrdinal
+  // when it was properly declared as such just a line above?
+  //
+  MueLu_sumAll(gDomVec->getMap()->getComm(), (LocalOrdinal)localMultColRequests, globalMultColRequests);
+
+  if (globalMultColRequests > 0) {
+    // special handling: two processors request the same global column id.
+    // decide which processor gets it
+
+    // distribute number of multipleColRequests to all processors
+    // each processor stores how many column ids for exchange are handled by the cur proc
+    std::vector<GlobalOrdinal> numMyMultColRequests(numProcs, 0);
+    std::vector<GlobalOrdinal> numGlobalMultColRequests(numProcs, 0);
+    numMyMultColRequests[myRank] = localMultColRequests;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
+                       numMyMultColRequests.data(),
+                       numGlobalMultColRequests.data());
+
+    // communicate multipleColRequests entries to all processors
+    int nMyOffset = 0;
+    for (int i = 0; i < myRank - 1; i++)
+      nMyOffset += numGlobalMultColRequests[i];  // calculate offset to store the weights on the corresponding place in procOverlappingWeights
+
+    const GlobalOrdinal zero = 0;
+    std::vector<GlobalOrdinal> procMultRequestedColIds(globalMultColRequests, zero);
+    std::vector<GlobalOrdinal> global_procMultRequestedColIds(globalMultColRequests, zero);
+
+    // loop over all local column GIDs that are also requested by other procs
+    for (size_t i = 0; i < multipleColRequests.size(); i++) {
+      procMultRequestedColIds[nMyOffset + i] = multipleColRequests[i];  // all weights are > 0 ?
+    }
 
-      // template ordinal, package (double)
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, Teuchos::as<int>(globalMultColRequests),
-                         procMultRequestedColIds.data(),
-                         global_procMultRequestedColIds.data());
+    // template ordinal, package (double)
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, Teuchos::as<int>(globalMultColRequests),
+                       procMultRequestedColIds.data(),
+                       global_procMultRequestedColIds.data());
 
-      // loop over global_procOverlappingWeights and eliminate wrong entries...
-      for (size_t k = 0; k<global_procMultRequestedColIds.size(); k++) {
-        GlobalOrdinal globColId = global_procMultRequestedColIds[k];
+    // loop over global_procOverlappingWeights and eliminate wrong entries...
+    for (size_t k = 0; k < global_procMultRequestedColIds.size(); k++) {
+      GlobalOrdinal globColId = global_procMultRequestedColIds[k];
 
-        std::vector<MT> MyWeightForColId(numProcs, MT_ZERO);
-        std::vector<MT> GlobalWeightForColId(numProcs, MT_ZERO);
+      std::vector<MT> MyWeightForColId(numProcs, MT_ZERO);
+      std::vector<MT> GlobalWeightForColId(numProcs, MT_ZERO);
 
-        if(gColVec->getMap()->isNodeGlobalElement(globColId)) {
-          MyWeightForColId[myRank] = gColId2Weight[globColId];
-        } else {
-          MyWeightForColId[myRank] = MT_ZERO;
-        }
+      if (gColVec->getMap()->isNodeGlobalElement(globColId)) {
+        MyWeightForColId[myRank] = gColId2Weight[globColId];
+      } else {
+        MyWeightForColId[myRank] = MT_ZERO;
+      }
 
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
-                           MyWeightForColId.data(),
-                           GlobalWeightForColId.data());
-
-        if(gColVec->getMap()->isNodeGlobalElement(globColId)) {
-          // note: 2 procs could have the same weight for a column index.
-          // pick the first one.
-          MT winnerValue = MT_ZERO;
-          int winnerProcRank = 0;
-          for (int proc = 0; proc < numProcs; proc++) {
-            if(GlobalWeightForColId[proc] > winnerValue) {
-              winnerValue = GlobalWeightForColId[proc];
-              winnerProcRank = proc;
-            }
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
+                         MyWeightForColId.data(),
+                         GlobalWeightForColId.data());
+
+      if (gColVec->getMap()->isNodeGlobalElement(globColId)) {
+        // note: 2 procs could have the same weight for a column index.
+        // pick the first one.
+        MT winnerValue     = MT_ZERO;
+        int winnerProcRank = 0;
+        for (int proc = 0; proc < numProcs; proc++) {
+          if (GlobalWeightForColId[proc] > winnerValue) {
+            winnerValue    = GlobalWeightForColId[proc];
+            winnerProcRank = proc;
           }
+        }
 
-          // winnerProcRank is the winner for handling globColId.
-          // winnerProcRank is unique (even if two procs have the same weight for a column index)
-
-          if(myRank != winnerProcRank) {
-            // remove corresponding entry from permutedDiagCandidatesFiltered
-            typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = permutedDiagCandidatesFiltered.begin();
-            while(p != permutedDiagCandidatesFiltered.end() )
-              {
-                if((*p).second == globColId)
-                  p = permutedDiagCandidatesFiltered.erase(p);
-                else
-                  p++;
-              }
+        // winnerProcRank is the winner for handling globColId.
+        // winnerProcRank is unique (even if two procs have the same weight for a column index)
+
+        if (myRank != winnerProcRank) {
+          // remove corresponding entry from permutedDiagCandidatesFiltered
+          typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = permutedDiagCandidatesFiltered.begin();
+          while (p != permutedDiagCandidatesFiltered.end()) {
+            if ((*p).second == globColId)
+              p = permutedDiagCandidatesFiltered.erase(p);
+            else
+              p++;
           }
+        }
 
-        } // end if isNodeGlobalElement
-      } // end loop over global_procOverlappingWeights and eliminate wrong entries...
-    } // end if globalMultColRequests > 0
+      }  // end if isNodeGlobalElement
+    }    // end loop over global_procOverlappingWeights and eliminate wrong entries...
+  }      // end if globalMultColRequests > 0
 
-    // put together all pairs:
-    //size_t sizeRowColPairs = keepDiagonalEntries.size() + permutedDiagCandidatesFiltered.size();
-    std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > RowColPairs;
-    RowColPairs.insert( RowColPairs.end(), keepDiagonalEntries.begin(), keepDiagonalEntries.end());
-    RowColPairs.insert( RowColPairs.end(), permutedDiagCandidatesFiltered.begin(), permutedDiagCandidatesFiltered.end());
+  // put together all pairs:
+  // size_t sizeRowColPairs = keepDiagonalEntries.size() + permutedDiagCandidatesFiltered.size();
+  std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > RowColPairs;
+  RowColPairs.insert(RowColPairs.end(), keepDiagonalEntries.begin(), keepDiagonalEntries.end());
+  RowColPairs.insert(RowColPairs.end(), permutedDiagCandidatesFiltered.begin(), permutedDiagCandidatesFiltered.end());
 
 #ifdef DEBUG_OUTPUT
-    //&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
-    // plausibility check
-    gColVec->putScalar(SC_ZERO);
-    gDomVec->putScalar(SC_ZERO);
-    typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator pl = RowColPairs.begin();
-    while(pl != RowColPairs.end() )
-      {
-        //GlobalOrdinal ik = (*pl).first;
-        GlobalOrdinal jk = (*pl).second;
-
-        gColVec->sumIntoGlobalValue(jk,1.0);
-        pl++;
-      }
-    gDomVec->doExport(*gColVec,*exporter,Xpetra::ADD);
-    for(size_t sz = 0; sz < gDomVec->getLocalLength(); ++sz) {
-      Teuchos::ArrayRCP< const Scalar > arrDomVec = gDomVec->getData(0);
-      if(arrDomVec[sz] > 1.0) {
-        GetOStream(Runtime0) << "RowColPairs has multiple column [" << sz << "]=" << arrDomVec[sz] << std::endl;
-      } else if(arrDomVec[sz] == SC_ZERO) {
-        GetOStream(Runtime0) << "RowColPairs has empty column [" << sz << "]=" << arrDomVec[sz] << std::endl;
-      }
+  //&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
+  // plausibility check
+  gColVec->putScalar(SC_ZERO);
+  gDomVec->putScalar(SC_ZERO);
+  typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator pl = RowColPairs.begin();
+  while (pl != RowColPairs.end()) {
+    // GlobalOrdinal ik = (*pl).first;
+    GlobalOrdinal jk = (*pl).second;
+
+    gColVec->sumIntoGlobalValue(jk, 1.0);
+    pl++;
+  }
+  gDomVec->doExport(*gColVec, *exporter, Xpetra::ADD);
+  for (size_t sz = 0; sz < gDomVec->getLocalLength(); ++sz) {
+    Teuchos::ArrayRCP<const Scalar> arrDomVec = gDomVec->getData(0);
+    if (arrDomVec[sz] > 1.0) {
+      GetOStream(Runtime0) << "RowColPairs has multiple column [" << sz << "]=" << arrDomVec[sz] << std::endl;
+    } else if (arrDomVec[sz] == SC_ZERO) {
+      GetOStream(Runtime0) << "RowColPairs has empty column [" << sz << "]=" << arrDomVec[sz] << std::endl;
     }
-    //&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
+  }
+  //&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
 #endif
 
-    //////////////////////////////////////////////////
-    // assumption: on each processor RowColPairs now contains
-    // a valid set of (row,column) pairs, where the row entries
-    // are a subset of the processor's rows and the column entries
-    // are unique throughout all processors.
-    // Note: the RowColPairs are only defined for a subset of all rows,
-    // so there might be rows without an entry in RowColPairs.
-    // It can be, that some rows seem to be missing in RowColPairs, since
-    // the entry in that row with maximum absolute value has been reserved
-    // by another row already (e.g. as already diagonal dominant row outside
-    // of perRowMap).
-    // In fact, the RowColPairs vector only defines the (row,column) pairs
-    // that will be definitely moved to the diagonal after permutation.
+  //////////////////////////////////////////////////
+  // assumption: on each processor RowColPairs now contains
+  // a valid set of (row,column) pairs, where the row entries
+  // are a subset of the processor's rows and the column entries
+  // are unique throughout all processors.
+  // Note: the RowColPairs are only defined for a subset of all rows,
+  // so there might be rows without an entry in RowColPairs.
+  // It can be, that some rows seem to be missing in RowColPairs, since
+  // the entry in that row with maximum absolute value has been reserved
+  // by another row already (e.g. as already diagonal dominant row outside
+  // of perRowMap).
+  // In fact, the RowColPairs vector only defines the (row,column) pairs
+  // that will be definitely moved to the diagonal after permutation.
 
 #ifdef DEBUG_OUTPUT
-    //  for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = RowColPairs.begin(); p != RowColPairs.end(); ++p) {
-    //    std::cout << "proc: " << myRank << " r/c: " << (*p).first << "/" << (*p).second << std::endl;
-    //  }
-    //    for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = RowColPairs.begin(); p != RowColPairs.end(); ++p)
-    //    {
-    ////      if((*p).first != (*p).second) std::cout << "difference: " << (*p).first << " " << (*p).second << std::endl;
-    //      std::cout << (*p).first +1 << " " << (*p).second+1 << std::endl;
-    //    }
-    //    std::cout << "\n";
+  //  for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = RowColPairs.begin(); p != RowColPairs.end(); ++p) {
+  //    std::cout << "proc: " << myRank << " r/c: " << (*p).first << "/" << (*p).second << std::endl;
+  //  }
+  //    for (typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::const_iterator p = RowColPairs.begin(); p != RowColPairs.end(); ++p)
+  //    {
+  ////      if((*p).first != (*p).second) std::cout << "difference: " << (*p).first << " " << (*p).second << std::endl;
+  //      std::cout << (*p).first +1 << " " << (*p).second+1 << std::endl;
+  //    }
+  //    std::cout << "\n";
 #endif
 
-    // vectors to store permutation information
-    Teuchos::RCP<Vector> Pperm  = VectorFactory::Build(A->getRowMap());
-    Teuchos::RCP<Vector> Qperm  = VectorFactory::Build(A->getDomainMap()); // global variant (based on domain map)
-    Teuchos::RCP<Vector> lQperm = VectorFactory::Build(A->getColMap());  // local variant (based on column map)
-
-    Pperm->putScalar(SC_ZERO);
-    Qperm->putScalar(SC_ZERO);
-    lQperm->putScalar(SC_ZERO);
-
-    // setup exporter for Qperm
-    Teuchos::RCP<Export> QpermExporter = ExportFactory::Build(lQperm->getMap(), Qperm->getMap());
-
-    Teuchos::RCP<Vector> RowIdStatus  = VectorFactory::Build(A->getRowMap());
-    Teuchos::RCP<Vector> ColIdStatus  = VectorFactory::Build(A->getDomainMap()); // global variant (based on domain map)
-    Teuchos::RCP<Vector> lColIdStatus = VectorFactory::Build(A->getColMap()); // local variant (based on column map)
-    Teuchos::RCP<Vector> ColIdUsed   = VectorFactory::Build(A->getDomainMap()); // mark column ids to be already in use
-    RowIdStatus->putScalar(SC_ZERO);
-    ColIdStatus->putScalar(SC_ZERO);
-    lColIdStatus->putScalar(SC_ZERO);
-    ColIdUsed->putScalar(SC_ZERO);   // no column ids are used
-
-
-    // count wide-range permutations
-    // a wide-range permutation is defined as a permutation of rows/columns which do not
-    // belong to the same node
-    LocalOrdinal lWideRangeRowPermutations = 0;
-    GlobalOrdinal gWideRangeRowPermutations = 0;
-    LocalOrdinal lWideRangeColPermutations = 0;
-    GlobalOrdinal gWideRangeColPermutations = 0;
-
-    // run 1: mark all "identity" permutations
-    {
-      Teuchos::ArrayRCP< Scalar > RowIdStatusArray = RowIdStatus->getDataNonConst(0);
-      Teuchos::ArrayRCP< Scalar > lColIdStatusArray = lColIdStatus->getDataNonConst(0);
-
-      typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = RowColPairs.begin();
-      while(p != RowColPairs.end() )
-        {
-          GlobalOrdinal ik = (*p).first;
-          GlobalOrdinal jk = (*p).second;
-
-          LocalOrdinal lik = A->getRowMap()->getLocalElement(ik);
-          LocalOrdinal ljk = A->getColMap()->getLocalElement(jk);
-
-          if(RowIdStatusArray[lik] == SC_ZERO) {
-            RowIdStatusArray[lik] = SC_ONE; // use this row id
-            lColIdStatusArray[ljk] = SC_ONE; // use this column id
-            Pperm->replaceLocalValue(lik, ik);
-            lQperm->replaceLocalValue(ljk, ik); // use column map
-            ColIdUsed->replaceGlobalValue(ik,SC_ONE); // ik is now used
-            p = RowColPairs.erase(p);
-
-            // detect wide range permutations
-            if(floor(ik/nDofsPerNode) != floor(jk/nDofsPerNode)) {
-              lWideRangeColPermutations++;
-            }
-          }
-          else
-            p++;
+  // vectors to store permutation information
+  Teuchos::RCP<Vector> Pperm  = VectorFactory::Build(A->getRowMap());
+  Teuchos::RCP<Vector> Qperm  = VectorFactory::Build(A->getDomainMap());  // global variant (based on domain map)
+  Teuchos::RCP<Vector> lQperm = VectorFactory::Build(A->getColMap());     // local variant (based on column map)
+
+  Pperm->putScalar(SC_ZERO);
+  Qperm->putScalar(SC_ZERO);
+  lQperm->putScalar(SC_ZERO);
+
+  // setup exporter for Qperm
+  Teuchos::RCP<Export> QpermExporter = ExportFactory::Build(lQperm->getMap(), Qperm->getMap());
+
+  Teuchos::RCP<Vector> RowIdStatus  = VectorFactory::Build(A->getRowMap());
+  Teuchos::RCP<Vector> ColIdStatus  = VectorFactory::Build(A->getDomainMap());  // global variant (based on domain map)
+  Teuchos::RCP<Vector> lColIdStatus = VectorFactory::Build(A->getColMap());     // local variant (based on column map)
+  Teuchos::RCP<Vector> ColIdUsed    = VectorFactory::Build(A->getDomainMap());  // mark column ids to be already in use
+  RowIdStatus->putScalar(SC_ZERO);
+  ColIdStatus->putScalar(SC_ZERO);
+  lColIdStatus->putScalar(SC_ZERO);
+  ColIdUsed->putScalar(SC_ZERO);  // no column ids are used
+
+  // count wide-range permutations
+  // a wide-range permutation is defined as a permutation of rows/columns which do not
+  // belong to the same node
+  LocalOrdinal lWideRangeRowPermutations  = 0;
+  GlobalOrdinal gWideRangeRowPermutations = 0;
+  LocalOrdinal lWideRangeColPermutations  = 0;
+  GlobalOrdinal gWideRangeColPermutations = 0;
+
+  // run 1: mark all "identity" permutations
+  {
+    Teuchos::ArrayRCP<Scalar> RowIdStatusArray  = RowIdStatus->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> lColIdStatusArray = lColIdStatus->getDataNonConst(0);
+
+    typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = RowColPairs.begin();
+    while (p != RowColPairs.end()) {
+      GlobalOrdinal ik = (*p).first;
+      GlobalOrdinal jk = (*p).second;
+
+      LocalOrdinal lik = A->getRowMap()->getLocalElement(ik);
+      LocalOrdinal ljk = A->getColMap()->getLocalElement(jk);
+
+      if (RowIdStatusArray[lik] == SC_ZERO) {
+        RowIdStatusArray[lik]  = SC_ONE;  // use this row id
+        lColIdStatusArray[ljk] = SC_ONE;  // use this column id
+        Pperm->replaceLocalValue(lik, ik);
+        lQperm->replaceLocalValue(ljk, ik);         // use column map
+        ColIdUsed->replaceGlobalValue(ik, SC_ONE);  // ik is now used
+        p = RowColPairs.erase(p);
+
+        // detect wide range permutations
+        if (floor(ik / nDofsPerNode) != floor(jk / nDofsPerNode)) {
+          lWideRangeColPermutations++;
         }
+      } else
+        p++;
+    }
 
+    // communicate column map -> domain map
+    Qperm->doExport(*lQperm, *QpermExporter, Xpetra::ABSMAX);
+    ColIdStatus->doExport(*lColIdStatus, *QpermExporter, Xpetra::ABSMAX);
 
-      // communicate column map -> domain map
-      Qperm->doExport(*lQperm, *QpermExporter, Xpetra::ABSMAX);
-      ColIdStatus->doExport(*lColIdStatus, *QpermExporter, Xpetra::ABSMAX);
-
-      // plausibility check
-      if(RowColPairs.size()>0) GetOStream(Warnings0) << "MueLu::PermutationFactory: There are Row/Col pairs left!!!" << std::endl; // TODO fix me
+    // plausibility check
+    if (RowColPairs.size() > 0) GetOStream(Warnings0) << "MueLu::PermutationFactory: There are Row/Col pairs left!!!" << std::endl;  // TODO fix me
 
-      // close Pperm
+    // close Pperm
 
-      // count, how many row permutations are missing on current proc
-      size_t cntFreeRowIdx = 0;
-      std::queue<GlobalOrdinal> qFreeGRowIdx;  // store global row ids of "free" rows
-      for (size_t lik = 0; lik < RowIdStatus->getLocalLength(); ++lik) {
-        if(RowIdStatusArray[lik] == SC_ZERO) {
-          cntFreeRowIdx++;
-          qFreeGRowIdx.push(RowIdStatus->getMap()->getGlobalElement(lik));
-        }
+    // count, how many row permutations are missing on current proc
+    size_t cntFreeRowIdx = 0;
+    std::queue<GlobalOrdinal> qFreeGRowIdx;  // store global row ids of "free" rows
+    for (size_t lik = 0; lik < RowIdStatus->getLocalLength(); ++lik) {
+      if (RowIdStatusArray[lik] == SC_ZERO) {
+        cntFreeRowIdx++;
+        qFreeGRowIdx.push(RowIdStatus->getMap()->getGlobalElement(lik));
       }
+    }
 
-      // fix Pperm
-      for (size_t lik = 0; lik < RowIdStatus->getLocalLength(); ++lik) {
-        if(RowIdStatusArray[lik] == SC_ZERO) {
-          RowIdStatusArray[lik] = SC_ONE; // use this row id
-          Pperm->replaceLocalValue(lik, qFreeGRowIdx.front());
-          // detect wide range permutations
-          if(floor(qFreeGRowIdx.front()/nDofsPerNode) != floor(RowIdStatus->getMap()->getGlobalElement(lik)/nDofsPerNode)) {
-            lWideRangeRowPermutations++;
-          }
-          qFreeGRowIdx.pop();
+    // fix Pperm
+    for (size_t lik = 0; lik < RowIdStatus->getLocalLength(); ++lik) {
+      if (RowIdStatusArray[lik] == SC_ZERO) {
+        RowIdStatusArray[lik] = SC_ONE;  // use this row id
+        Pperm->replaceLocalValue(lik, qFreeGRowIdx.front());
+        // detect wide range permutations
+        if (floor(qFreeGRowIdx.front() / nDofsPerNode) != floor(RowIdStatus->getMap()->getGlobalElement(lik) / nDofsPerNode)) {
+          lWideRangeRowPermutations++;
         }
+        qFreeGRowIdx.pop();
       }
     }
+  }
 
-    size_t cntUnusedColIdx = 0;
-    std::queue<GlobalOrdinal> qUnusedGColIdx;  // store global column ids of "free" available columns
-    size_t cntFreeColIdx = 0;
-    std::queue<GlobalOrdinal> qFreeGColIdx;  // store global column ids of "free" available columns
-    {
-      Teuchos::ArrayRCP< Scalar > ColIdStatusArray = ColIdStatus->getDataNonConst(0);
-      Teuchos::ArrayRCP< Scalar > ColIdUsedArray   = ColIdUsed->getDataNonConst(0); // not sure about this
-
-      // close Qperm (free permutation entries in Qperm)
-      for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
-        if(ColIdStatusArray[ljk] == SC_ZERO) {
-          cntFreeColIdx++;
-          qFreeGColIdx.push(ColIdStatus->getMap()->getGlobalElement(ljk));
-        }
+  size_t cntUnusedColIdx = 0;
+  std::queue<GlobalOrdinal> qUnusedGColIdx;  // store global column ids of "free" available columns
+  size_t cntFreeColIdx = 0;
+  std::queue<GlobalOrdinal> qFreeGColIdx;  // store global column ids of "free" available columns
+  {
+    Teuchos::ArrayRCP<Scalar> ColIdStatusArray = ColIdStatus->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> ColIdUsedArray   = ColIdUsed->getDataNonConst(0);  // not sure about this
+
+    // close Qperm (free permutation entries in Qperm)
+    for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
+      if (ColIdStatusArray[ljk] == SC_ZERO) {
+        cntFreeColIdx++;
+        qFreeGColIdx.push(ColIdStatus->getMap()->getGlobalElement(ljk));
       }
+    }
 
-      for (size_t ljk = 0; ljk < ColIdUsed->getLocalLength(); ++ljk) {
-        if(ColIdUsedArray[ljk] == SC_ZERO) {
-          cntUnusedColIdx++;
-          qUnusedGColIdx.push(ColIdUsed->getMap()->getGlobalElement(ljk));
-        }
+    for (size_t ljk = 0; ljk < ColIdUsed->getLocalLength(); ++ljk) {
+      if (ColIdUsedArray[ljk] == SC_ZERO) {
+        cntUnusedColIdx++;
+        qUnusedGColIdx.push(ColIdUsed->getMap()->getGlobalElement(ljk));
       }
+    }
 
-
-      // fix Qperm with local entries
-      for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
-        // stop if no (local) unused column idx are left
-        if(cntUnusedColIdx == 0) break;
-
-        if(ColIdStatusArray[ljk] == SC_ZERO) {
-          ColIdStatusArray[ljk] = SC_ONE; // use this row id
-          Qperm->replaceLocalValue(ljk, qUnusedGColIdx.front()); // loop over ColIdStatus (lives on domain map)
-          ColIdUsed->replaceGlobalValue(qUnusedGColIdx.front(),SC_ONE); // ljk is now used, too
-          // detect wide range permutations
-          if(floor(qUnusedGColIdx.front()/nDofsPerNode) != floor(ColIdStatus->getMap()->getGlobalElement(ljk)/nDofsPerNode)) {
-            lWideRangeColPermutations++;
-          }
-          qUnusedGColIdx.pop();
-          cntUnusedColIdx--;
-          cntFreeColIdx--;
+    // fix Qperm with local entries
+    for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
+      // stop if no (local) unused column idx are left
+      if (cntUnusedColIdx == 0) break;
+
+      if (ColIdStatusArray[ljk] == SC_ZERO) {
+        ColIdStatusArray[ljk] = SC_ONE;                                 // use this row id
+        Qperm->replaceLocalValue(ljk, qUnusedGColIdx.front());          // loop over ColIdStatus (lives on domain map)
+        ColIdUsed->replaceGlobalValue(qUnusedGColIdx.front(), SC_ONE);  // ljk is now used, too
+        // detect wide range permutations
+        if (floor(qUnusedGColIdx.front() / nDofsPerNode) != floor(ColIdStatus->getMap()->getGlobalElement(ljk) / nDofsPerNode)) {
+          lWideRangeColPermutations++;
         }
+        qUnusedGColIdx.pop();
+        cntUnusedColIdx--;
+        cntFreeColIdx--;
       }
+    }
 
+    // Qperm->doExport(*lQperm,*QpermExporter,Xpetra::ABSMAX); // no export necessary, since changes only locally
+    // ColIdStatus->doExport(*lColIdStatus,*QpermExporter,Xpetra::ABSMAX);
 
-      //Qperm->doExport(*lQperm,*QpermExporter,Xpetra::ABSMAX); // no export necessary, since changes only locally
-      //ColIdStatus->doExport(*lColIdStatus,*QpermExporter,Xpetra::ABSMAX);
-
-      // count, how many unused column idx are needed on current processor
-      // to complete Qperm
-      cntFreeColIdx = 0;
-      for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) { // TODO avoid this loop
-        if(ColIdStatusArray[ljk] == SC_ZERO) {
-          cntFreeColIdx++;
-        }
+    // count, how many unused column idx are needed on current processor
+    // to complete Qperm
+    cntFreeColIdx = 0;
+    for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {  // TODO avoid this loop
+      if (ColIdStatusArray[ljk] == SC_ZERO) {
+        cntFreeColIdx++;
       }
     }
+  }
 
-    GlobalOrdinal global_cntFreeColIdx = 0;
-    LocalOrdinal  local_cntFreeColIdx = cntFreeColIdx;
-    MueLu_sumAll(comm, Teuchos::as<GlobalOrdinal>(local_cntFreeColIdx), global_cntFreeColIdx);
+  GlobalOrdinal global_cntFreeColIdx = 0;
+  LocalOrdinal local_cntFreeColIdx   = cntFreeColIdx;
+  MueLu_sumAll(comm, Teuchos::as<GlobalOrdinal>(local_cntFreeColIdx), global_cntFreeColIdx);
 #ifdef DEBUG_OUTPUT
-    std::cout << "global # of empty column idx entries in Qperm: " << global_cntFreeColIdx << std::endl;
+  std::cout << "global # of empty column idx entries in Qperm: " << global_cntFreeColIdx << std::endl;
 #endif
 
-    // avoid global communication if possible
-    if(global_cntFreeColIdx > 0) {
-
-      // 1) count how many unused column ids are left
-      GlobalOrdinal global_cntUnusedColIdx = 0;
-      LocalOrdinal  local_cntUnusedColIdx = cntUnusedColIdx;
-      MueLu_sumAll(comm, Teuchos::as<GlobalOrdinal>(local_cntUnusedColIdx), global_cntUnusedColIdx);
+  // avoid global communication if possible
+  if (global_cntFreeColIdx > 0) {
+    // 1) count how many unused column ids are left
+    GlobalOrdinal global_cntUnusedColIdx = 0;
+    LocalOrdinal local_cntUnusedColIdx   = cntUnusedColIdx;
+    MueLu_sumAll(comm, Teuchos::as<GlobalOrdinal>(local_cntUnusedColIdx), global_cntUnusedColIdx);
 #ifdef DEBUG_OUTPUT
-      std::cout << "global # of unused column idx: " << global_cntUnusedColIdx << std::endl;
+    std::cout << "global # of unused column idx: " << global_cntUnusedColIdx << std::endl;
 #endif
 
-      // 2) communicate how many unused column ids are available on procs
-      std::vector<LocalOrdinal> local_UnusedColIdxOnProc (numProcs);
-      std::vector<LocalOrdinal> global_UnusedColIdxOnProc(numProcs);
-      local_UnusedColIdxOnProc[myRank] = local_cntUnusedColIdx;
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
-                         local_UnusedColIdxOnProc.data(),
-                         global_UnusedColIdxOnProc.data());
+    // 2) communicate how many unused column ids are available on procs
+    std::vector<LocalOrdinal> local_UnusedColIdxOnProc(numProcs);
+    std::vector<LocalOrdinal> global_UnusedColIdxOnProc(numProcs);
+    local_UnusedColIdxOnProc[myRank] = local_cntUnusedColIdx;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
+                       local_UnusedColIdxOnProc.data(),
+                       global_UnusedColIdxOnProc.data());
 
 #ifdef DEBUG_OUTPUT
-      std::cout << "PROC " << myRank << " global num unused indices per proc: ";
-      for (size_t ljk = 0; ljk < global_UnusedColIdxOnProc.size(); ++ljk) {
-        std::cout << " " << global_UnusedColIdxOnProc[ljk];
-      }
-      std::cout << std::endl;
+    std::cout << "PROC " << myRank << " global num unused indices per proc: ";
+    for (size_t ljk = 0; ljk < global_UnusedColIdxOnProc.size(); ++ljk) {
+      std::cout << " " << global_UnusedColIdxOnProc[ljk];
+    }
+    std::cout << std::endl;
 #endif
 
-      // 3) build array of length global_cntUnusedColIdx to globally replicate unused column idx
-      std::vector<GlobalOrdinal> local_UnusedColIdxVector(Teuchos::as<size_t>(global_cntUnusedColIdx));
-      std::vector<GlobalOrdinal> global_UnusedColIdxVector(Teuchos::as<size_t>(global_cntUnusedColIdx));
-      GlobalOrdinal global_cntUnusedColIdxStartIter = 0;
-      for(int proc=0; proc<myRank; proc++) {
-        global_cntUnusedColIdxStartIter += global_UnusedColIdxOnProc[proc];
-      }
-      for(GlobalOrdinal k = global_cntUnusedColIdxStartIter; k < global_cntUnusedColIdxStartIter+local_cntUnusedColIdx; k++) {
-        local_UnusedColIdxVector[k] = qUnusedGColIdx.front();
-        qUnusedGColIdx.pop();
-      }
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, Teuchos::as<int>(global_cntUnusedColIdx),
-                         local_UnusedColIdxVector.data(),
-                         global_UnusedColIdxVector.data());
+    // 3) build array of length global_cntUnusedColIdx to globally replicate unused column idx
+    std::vector<GlobalOrdinal> local_UnusedColIdxVector(Teuchos::as<size_t>(global_cntUnusedColIdx));
+    std::vector<GlobalOrdinal> global_UnusedColIdxVector(Teuchos::as<size_t>(global_cntUnusedColIdx));
+    GlobalOrdinal global_cntUnusedColIdxStartIter = 0;
+    for (int proc = 0; proc < myRank; proc++) {
+      global_cntUnusedColIdxStartIter += global_UnusedColIdxOnProc[proc];
+    }
+    for (GlobalOrdinal k = global_cntUnusedColIdxStartIter; k < global_cntUnusedColIdxStartIter + local_cntUnusedColIdx; k++) {
+      local_UnusedColIdxVector[k] = qUnusedGColIdx.front();
+      qUnusedGColIdx.pop();
+    }
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, Teuchos::as<int>(global_cntUnusedColIdx),
+                       local_UnusedColIdxVector.data(),
+                       global_UnusedColIdxVector.data());
 #ifdef DEBUG_OUTPUT
-      std::cout << "PROC " << myRank << " global UnusedGColIdx: ";
-      for (size_t ljk = 0; ljk < global_UnusedColIdxVector.size(); ++ljk) {
-        std::cout << " " << global_UnusedColIdxVector[ljk];
-      }
-      std::cout << std::endl;
+    std::cout << "PROC " << myRank << " global UnusedGColIdx: ";
+    for (size_t ljk = 0; ljk < global_UnusedColIdxVector.size(); ++ljk) {
+      std::cout << " " << global_UnusedColIdxVector[ljk];
+    }
+    std::cout << std::endl;
 #endif
 
-
-
-      // 4) communicate, how many column idx are needed on each processor
-      //    to complete Qperm
-      std::vector<LocalOrdinal> local_EmptyColIdxOnProc (numProcs);
-      std::vector<LocalOrdinal> global_EmptyColIdxOnProc(numProcs);
-      local_EmptyColIdxOnProc[myRank] = local_cntFreeColIdx;
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
-                         local_EmptyColIdxOnProc.data(),
-                         global_EmptyColIdxOnProc.data());
+    // 4) communicate, how many column idx are needed on each processor
+    //    to complete Qperm
+    std::vector<LocalOrdinal> local_EmptyColIdxOnProc(numProcs);
+    std::vector<LocalOrdinal> global_EmptyColIdxOnProc(numProcs);
+    local_EmptyColIdxOnProc[myRank] = local_cntFreeColIdx;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, numProcs,
+                       local_EmptyColIdxOnProc.data(),
+                       global_EmptyColIdxOnProc.data());
 
 #ifdef DEBUG_OUTPUT
-      std::cout << "PROC " << myRank << " global num of needed column indices: ";
-      for (size_t ljk = 0; ljk < global_EmptyColIdxOnProc.size(); ++ljk) {
-        std::cout << " " << global_EmptyColIdxOnProc[ljk];
-      }
-      std::cout << std::endl;
+    std::cout << "PROC " << myRank << " global num of needed column indices: ";
+    for (size_t ljk = 0; ljk < global_EmptyColIdxOnProc.size(); ++ljk) {
+      std::cout << " " << global_EmptyColIdxOnProc[ljk];
+    }
+    std::cout << std::endl;
 #endif
 
-      // 5) determine first index in global_UnusedColIdxVector for unused column indices,
-      //    that are marked to be used by this processor
-      GlobalOrdinal global_UnusedColStartIdx = 0;
-      for(int proc=0; proc<myRank; proc++) {
-        global_UnusedColStartIdx += global_EmptyColIdxOnProc[proc];
-      }
+    // 5) determine first index in global_UnusedColIdxVector for unused column indices,
+    //    that are marked to be used by this processor
+    GlobalOrdinal global_UnusedColStartIdx = 0;
+    for (int proc = 0; proc < myRank; proc++) {
+      global_UnusedColStartIdx += global_EmptyColIdxOnProc[proc];
+    }
 
 #ifdef DEBUG_OUTPUT
-      GetOStream(Statistics0) << "PROC " << myRank << " is allowd to use the following column gids: ";
-      for(GlobalOrdinal k = global_UnusedColStartIdx; k < global_UnusedColStartIdx + Teuchos::as<GlobalOrdinal>(cntFreeColIdx); k++) {
-        GetOStream(Statistics0) << global_UnusedColIdxVector[k] << " ";
-      }
-      GetOStream(Statistics0) << std::endl;
+    GetOStream(Statistics0) << "PROC " << myRank << " is allowd to use the following column gids: ";
+    for (GlobalOrdinal k = global_UnusedColStartIdx; k < global_UnusedColStartIdx + Teuchos::as<GlobalOrdinal>(cntFreeColIdx); k++) {
+      GetOStream(Statistics0) << global_UnusedColIdxVector[k] << " ";
+    }
+    GetOStream(Statistics0) << std::endl;
 #endif
 
-      // 6.) fix Qperm with global entries
-      {
-        Teuchos::ArrayRCP< Scalar > ColIdStatusArray = ColIdStatus->getDataNonConst(0);
-        GlobalOrdinal array_iter = 0;
-        for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
-
-          if(ColIdStatusArray[ljk] == SC_ZERO) {
-            ColIdStatusArray[ljk] = SC_ONE; // use this row id
-            Qperm->replaceLocalValue(ljk, global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter]);
-            ColIdUsed->replaceGlobalValue(global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter],SC_ONE);
-            // detect wide range permutations
-            if(floor(global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter]/nDofsPerNode) != floor(ColIdStatus->getMap()->getGlobalElement(ljk)/nDofsPerNode)) {
-              lWideRangeColPermutations++;
-            }
-            array_iter++;
-            //cntUnusedColIdx--; // check me
+    // 6.) fix Qperm with global entries
+    {
+      Teuchos::ArrayRCP<Scalar> ColIdStatusArray = ColIdStatus->getDataNonConst(0);
+      GlobalOrdinal array_iter                   = 0;
+      for (size_t ljk = 0; ljk < ColIdStatus->getLocalLength(); ++ljk) {
+        if (ColIdStatusArray[ljk] == SC_ZERO) {
+          ColIdStatusArray[ljk] = SC_ONE;  // use this row id
+          Qperm->replaceLocalValue(ljk, global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter]);
+          ColIdUsed->replaceGlobalValue(global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter], SC_ONE);
+          // detect wide range permutations
+          if (floor(global_UnusedColIdxVector[global_UnusedColStartIdx + array_iter] / nDofsPerNode) != floor(ColIdStatus->getMap()->getGlobalElement(ljk) / nDofsPerNode)) {
+            lWideRangeColPermutations++;
           }
+          array_iter++;
+          // cntUnusedColIdx--; // check me
         }
       }
-    } // end if global_cntFreeColIdx > 0
-    /////////////////// Qperm should be fine now...
-
-
-    // create new empty Matrix
-    Teuchos::RCP<CrsMatrixWrap> permPTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(),1));
-    Teuchos::RCP<CrsMatrixWrap> permQTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(),1));
-
-    {
-      Teuchos::ArrayRCP< Scalar > PpermData = Pperm->getDataNonConst(0);
-      Teuchos::ArrayRCP< Scalar > QpermData = Qperm->getDataNonConst(0);
-
-      for(size_t row=0; row<A->getLocalNumRows(); row++) {
-        // FIXME (mfh 30 Oct 2015): Teuchos::as doesn't know how to
-        // convert from complex Scalar to GO, so we have to take the real
-        // part first.  I think that's the right thing to do in this case.
-        Teuchos::ArrayRCP<GlobalOrdinal> indoutP(1,Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(PpermData[row]))); // column idx for Perm^T
-        Teuchos::ArrayRCP<GlobalOrdinal> indoutQ(1,Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(QpermData[row]))); // column idx for Qperm
-        Teuchos::ArrayRCP<Scalar> valout(1,SC_ONE);
-        permPTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutP.view(0,indoutP.size()), valout.view(0,valout.size()));
-        permQTmatrix->insertGlobalValues (A->getRowMap()->getGlobalElement(row), indoutQ.view(0,indoutQ.size()), valout.view(0,valout.size()));
-      }
     }
+  }  // end if global_cntFreeColIdx > 0
+  /////////////////// Qperm should be fine now...
+
+  // create new empty Matrix
+  Teuchos::RCP<CrsMatrixWrap> permPTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(), 1));
+  Teuchos::RCP<CrsMatrixWrap> permQTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(), 1));
+
+  {
+    Teuchos::ArrayRCP<Scalar> PpermData = Pperm->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> QpermData = Qperm->getDataNonConst(0);
+
+    for (size_t row = 0; row < A->getLocalNumRows(); row++) {
+      // FIXME (mfh 30 Oct 2015): Teuchos::as doesn't know how to
+      // convert from complex Scalar to GO, so we have to take the real
+      // part first.  I think that's the right thing to do in this case.
+      Teuchos::ArrayRCP<GlobalOrdinal> indoutP(1, Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(PpermData[row])));  // column idx for Perm^T
+      Teuchos::ArrayRCP<GlobalOrdinal> indoutQ(1, Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(QpermData[row])));  // column idx for Qperm
+      Teuchos::ArrayRCP<Scalar> valout(1, SC_ONE);
+      permPTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutP.view(0, indoutP.size()), valout.view(0, valout.size()));
+      permQTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutQ.view(0, indoutQ.size()), valout.view(0, valout.size()));
+    }
+  }
 
-    permPTmatrix->fillComplete();
-    permQTmatrix->fillComplete();
+  permPTmatrix->fillComplete();
+  permQTmatrix->fillComplete();
 
-    Teuchos::RCP<Matrix> permPmatrix = Utilities::Transpose(*permPTmatrix, true);
+  Teuchos::RCP<Matrix> permPmatrix = Utilities::Transpose(*permPTmatrix, true);
 
-    for(size_t row=0; row<permPTmatrix->getLocalNumRows(); row++) {
-      if(permPTmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permPTmatrix is " << permPTmatrix->getNumEntriesInLocalRow(row) << std::endl;
-      if(permPmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permPmatrix is " << permPmatrix->getNumEntriesInLocalRow(row) << std::endl;
-      if(permQTmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permQmatrix is " << permQTmatrix->getNumEntriesInLocalRow(row) << std::endl;
-    }
+  for (size_t row = 0; row < permPTmatrix->getLocalNumRows(); row++) {
+    if (permPTmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) << "#entries in row " << row << " of permPTmatrix is " << permPTmatrix->getNumEntriesInLocalRow(row) << std::endl;
+    if (permPmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) << "#entries in row " << row << " of permPmatrix is " << permPmatrix->getNumEntriesInLocalRow(row) << std::endl;
+    if (permQTmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) << "#entries in row " << row << " of permQmatrix is " << permQTmatrix->getNumEntriesInLocalRow(row) << std::endl;
+  }
 
-    // build permP * A * permQT
-    Teuchos::RCP<Matrix> ApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *permQTmatrix, false, GetOStream(Statistics2));
-    Teuchos::RCP<Matrix> permPApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*permPmatrix, false, *ApermQt, false, GetOStream(Statistics2));
-
-    /*
-      MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mat", *A);
-      MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permP.mat", *permPmatrix);
-      MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permQt.mat", *permQTmatrix);
-      MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permPApermQt.mat", *permPApermQt);
-    */
-    // build scaling matrix
-    Teuchos::RCP<Vector> diagVec = VectorFactory::Build(permPApermQt->getRowMap(),true);
-    Teuchos::RCP<Vector> invDiagVec = VectorFactory::Build(permPApermQt->getRowMap(),true);
-    Teuchos::RCP<CrsMatrixWrap> diagScalingOp = Teuchos::rcp(new CrsMatrixWrap(permPApermQt->getRowMap(),1));
-    {
-      permPApermQt->getLocalDiagCopy(*diagVec);
-      Teuchos::ArrayRCP< const Scalar > diagVecData = diagVec->getData(0);
-      Teuchos::ArrayRCP< Scalar > invDiagVecData = invDiagVec->getDataNonConst(0);
-      for(size_t i = 0; i<diagVec->getMap()->getLocalNumElements(); ++i) {
-        if(diagVecData[i] != SC_ZERO)
-          invDiagVecData[i] = SC_ONE / diagVecData[i];
-        else {
-          invDiagVecData[i] = SC_ONE;
-          GetOStream(Statistics0) << "MueLu::PermutationFactory: found zero on diagonal in row " << i << std::endl;
-        }
+  // build permP * A * permQT
+  Teuchos::RCP<Matrix> ApermQt      = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *permQTmatrix, false, GetOStream(Statistics2));
+  Teuchos::RCP<Matrix> permPApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*permPmatrix, false, *ApermQt, false, GetOStream(Statistics2));
+
+  /*
+    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mat", *A);
+    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permP.mat", *permPmatrix);
+    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permQt.mat", *permQTmatrix);
+    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permPApermQt.mat", *permPApermQt);
+  */
+  // build scaling matrix
+  Teuchos::RCP<Vector> diagVec              = VectorFactory::Build(permPApermQt->getRowMap(), true);
+  Teuchos::RCP<Vector> invDiagVec           = VectorFactory::Build(permPApermQt->getRowMap(), true);
+  Teuchos::RCP<CrsMatrixWrap> diagScalingOp = Teuchos::rcp(new CrsMatrixWrap(permPApermQt->getRowMap(), 1));
+  {
+    permPApermQt->getLocalDiagCopy(*diagVec);
+    Teuchos::ArrayRCP<const Scalar> diagVecData = diagVec->getData(0);
+    Teuchos::ArrayRCP<Scalar> invDiagVecData    = invDiagVec->getDataNonConst(0);
+    for (size_t i = 0; i < diagVec->getMap()->getLocalNumElements(); ++i) {
+      if (diagVecData[i] != SC_ZERO)
+        invDiagVecData[i] = SC_ONE / diagVecData[i];
+      else {
+        invDiagVecData[i] = SC_ONE;
+        GetOStream(Statistics0) << "MueLu::PermutationFactory: found zero on diagonal in row " << i << std::endl;
       }
+    }
 
-      for(size_t row=0; row<A->getLocalNumRows(); row++) {
-        Teuchos::ArrayRCP<GlobalOrdinal> indout(1,permPApermQt->getRowMap()->getGlobalElement(row)); // column idx for Perm^T
-        Teuchos::ArrayRCP<Scalar> valout(1,invDiagVecData[row]);
-        diagScalingOp->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indout.view(0,indout.size()), valout.view(0,valout.size()));
-      }
+    for (size_t row = 0; row < A->getLocalNumRows(); row++) {
+      Teuchos::ArrayRCP<GlobalOrdinal> indout(1, permPApermQt->getRowMap()->getGlobalElement(row));  // column idx for Perm^T
+      Teuchos::ArrayRCP<Scalar> valout(1, invDiagVecData[row]);
+      diagScalingOp->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indout.view(0, indout.size()), valout.view(0, valout.size()));
     }
-    diagScalingOp->fillComplete();
-
-    Teuchos::RCP<Matrix> scaledA = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*diagScalingOp, false, *permPApermQt, false, GetOStream(Statistics2));
-    currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(scaledA), genFactory/*this*/);
-
-    currentLevel.Set("permA", Teuchos::rcp_dynamic_cast<Matrix>(permPApermQt), genFactory/*this*/);  // TODO careful with this!!!
-    currentLevel.Set("permP", Teuchos::rcp_dynamic_cast<Matrix>(permPmatrix), genFactory/*this*/);
-    currentLevel.Set("permQT", Teuchos::rcp_dynamic_cast<Matrix>(permQTmatrix), genFactory/*this*/);
-    currentLevel.Set("permScaling", Teuchos::rcp_dynamic_cast<Matrix>(diagScalingOp), genFactory/*this*/);
-
-    //// count row permutations
-    // count zeros on diagonal in P -> number of row permutations
-    Teuchos::RCP<Vector> diagPVec = VectorFactory::Build(permPmatrix->getRowMap(),true);
-    permPmatrix->getLocalDiagCopy(*diagPVec);
-    LocalOrdinal lNumRowPermutations = 0;
-    GlobalOrdinal gNumRowPermutations = 0;
-    {
-      Teuchos::ArrayRCP< const Scalar > diagPVecData = diagPVec->getData(0);
-      for(size_t i = 0; i<diagPVec->getMap()->getLocalNumElements(); ++i) {
-        if(diagPVecData[i] == SC_ZERO) {
-          lNumRowPermutations++;
-        }
+  }
+  diagScalingOp->fillComplete();
+
+  Teuchos::RCP<Matrix> scaledA = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*diagScalingOp, false, *permPApermQt, false, GetOStream(Statistics2));
+  currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(scaledA), genFactory /*this*/);
+
+  currentLevel.Set("permA", Teuchos::rcp_dynamic_cast<Matrix>(permPApermQt), genFactory /*this*/);  // TODO careful with this!!!
+  currentLevel.Set("permP", Teuchos::rcp_dynamic_cast<Matrix>(permPmatrix), genFactory /*this*/);
+  currentLevel.Set("permQT", Teuchos::rcp_dynamic_cast<Matrix>(permQTmatrix), genFactory /*this*/);
+  currentLevel.Set("permScaling", Teuchos::rcp_dynamic_cast<Matrix>(diagScalingOp), genFactory /*this*/);
+
+  //// count row permutations
+  // count zeros on diagonal in P -> number of row permutations
+  Teuchos::RCP<Vector> diagPVec = VectorFactory::Build(permPmatrix->getRowMap(), true);
+  permPmatrix->getLocalDiagCopy(*diagPVec);
+  LocalOrdinal lNumRowPermutations  = 0;
+  GlobalOrdinal gNumRowPermutations = 0;
+  {
+    Teuchos::ArrayRCP<const Scalar> diagPVecData = diagPVec->getData(0);
+    for (size_t i = 0; i < diagPVec->getMap()->getLocalNumElements(); ++i) {
+      if (diagPVecData[i] == SC_ZERO) {
+        lNumRowPermutations++;
       }
     }
+  }
 
-    // sum up all entries in multipleColRequests over all processors
-    MueLu_sumAll(diagPVec->getMap()->getComm(), Teuchos::as<GlobalOrdinal>(lNumRowPermutations), gNumRowPermutations);
-
-    //// count column permutations
-    // count zeros on diagonal in Q^T -> number of column permutations
-    Teuchos::RCP<Vector> diagQTVec = VectorFactory::Build(permQTmatrix->getRowMap(),true);
-    permQTmatrix->getLocalDiagCopy(*diagQTVec);
-    LocalOrdinal lNumColPermutations = 0;
-    GlobalOrdinal gNumColPermutations = 0;
-    {
-      Teuchos::ArrayRCP< const Scalar > diagQTVecData = diagQTVec->getData(0);
-      for(size_t i = 0; i<diagQTVec->getMap()->getLocalNumElements(); ++i) {
-        if(diagQTVecData[i] == SC_ZERO) {
-          lNumColPermutations++;
-        }
+  // sum up all entries in multipleColRequests over all processors
+  MueLu_sumAll(diagPVec->getMap()->getComm(), Teuchos::as<GlobalOrdinal>(lNumRowPermutations), gNumRowPermutations);
+
+  //// count column permutations
+  // count zeros on diagonal in Q^T -> number of column permutations
+  Teuchos::RCP<Vector> diagQTVec = VectorFactory::Build(permQTmatrix->getRowMap(), true);
+  permQTmatrix->getLocalDiagCopy(*diagQTVec);
+  LocalOrdinal lNumColPermutations  = 0;
+  GlobalOrdinal gNumColPermutations = 0;
+  {
+    Teuchos::ArrayRCP<const Scalar> diagQTVecData = diagQTVec->getData(0);
+    for (size_t i = 0; i < diagQTVec->getMap()->getLocalNumElements(); ++i) {
+      if (diagQTVecData[i] == SC_ZERO) {
+        lNumColPermutations++;
       }
     }
+  }
 
-    // sum up all entries in multipleColRequests over all processors
-    MueLu_sumAll(diagQTVec->getMap()->getComm(), Teuchos::as<GlobalOrdinal>(lNumColPermutations), gNumColPermutations);
+  // sum up all entries in multipleColRequests over all processors
+  MueLu_sumAll(diagQTVec->getMap()->getComm(), Teuchos::as<GlobalOrdinal>(lNumColPermutations), gNumColPermutations);
 
-    currentLevel.Set("#RowPermutations", gNumRowPermutations, genFactory/*this*/);
-    currentLevel.Set("#ColPermutations", gNumColPermutations, genFactory/*this*/);
-    currentLevel.Set("#WideRangeRowPermutations", gWideRangeRowPermutations, genFactory/*this*/);
-    currentLevel.Set("#WideRangeColPermutations", gWideRangeColPermutations, genFactory/*this*/);
+  currentLevel.Set("#RowPermutations", gNumRowPermutations, genFactory /*this*/);
+  currentLevel.Set("#ColPermutations", gNumColPermutations, genFactory /*this*/);
+  currentLevel.Set("#WideRangeRowPermutations", gWideRangeRowPermutations, genFactory /*this*/);
+  currentLevel.Set("#WideRangeColPermutations", gWideRangeColPermutations, genFactory /*this*/);
 
-    GetOStream(Statistics0) << "#Row    permutations/max possible permutations: " << gNumRowPermutations << "/" << diagPVec->getMap()->getGlobalNumElements() << std::endl;
-    GetOStream(Statistics0) << "#Column permutations/max possible permutations: " << gNumColPermutations << "/" << diagQTVec->getMap()->getGlobalNumElements() << std::endl;
-    GetOStream(Runtime1) << "#wide range row permutations: " << gWideRangeRowPermutations << " #wide range column permutations: " << gWideRangeColPermutations << std::endl;
-  }
+  GetOStream(Statistics0) << "#Row    permutations/max possible permutations: " << gNumRowPermutations << "/" << diagPVec->getMap()->getGlobalNumElements() << std::endl;
+  GetOStream(Statistics0) << "#Column permutations/max possible permutations: " << gNumColPermutations << "/" << diagQTVec->getMap()->getGlobalNumElements() << std::endl;
+  GetOStream(Runtime1) << "#wide range row permutations: " << gWideRangeRowPermutations << " #wide range column permutations: " << gWideRangeColPermutations << std::endl;
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_ALGEBRAICPERMUTATIONSTRATEGY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_AvatarInterface.cpp b/packages/muelu/src/Utils/MueLu_AvatarInterface.cpp
index 375a7e4fabb7..bba55e12ee93 100644
--- a/packages/muelu/src/Utils/MueLu_AvatarInterface.cpp
+++ b/packages/muelu/src/Utils/MueLu_AvatarInterface.cpp
@@ -54,7 +54,6 @@
 #include "MueLu_BaseClass.hpp"
 #include "Teuchos_RawParameterListHelpers.hpp"
 
-
 // ***********************************************************************
 /* Notional Parameterlist Structure
    "avatar: filestem"              "{'mystem1','mystem2'}"
@@ -82,7 +81,6 @@
 
  */
 
-
 /*
 TODO List:
 Modify MueLu
@@ -95,11 +93,8 @@ extern "C" {
 #include "avatar_api.h"
 }
 
-
-
 namespace MueLu {
 
-
 // ***********************************************************************
 RCP<const ParameterList> AvatarInterface::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
@@ -109,54 +104,53 @@ RCP<const ParameterList> AvatarInterface::GetValidParameterList() const {
   int int_dummy;
 
   // Files from which to read Avatar trees
-  validParamList->set<Teuchos::Array<std::string> >("avatar: decision tree files",ar_dummy,"Names of Avatar decision tree files");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: decision tree files", ar_dummy, "Names of Avatar decision tree files");
 
   // Strings from which to read Avatar trees
-  validParamList->set<Teuchos::Array<std::string> >("avatar: decision tree strings",ar_dummy,"Avatar decision tree strings");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: decision tree strings", ar_dummy, "Avatar decision tree strings");
 
   // Files from which to read Avatar names
-  validParamList->set<Teuchos::Array<std::string> >("avatar: names files",ar_dummy,"Names of Avatar decision names files");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: names files", ar_dummy, "Names of Avatar decision names files");
 
   // Strings from which to read Avatar names
-  validParamList->set<Teuchos::Array<std::string> >("avatar: names strings",ar_dummy,"Avatar decision names strings");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: names strings", ar_dummy, "Avatar decision names strings");
 
   // Filestem arg for Avatar
-  validParamList->set<Teuchos::Array<std::string> >("avatar: filestem",ar_dummy,"Filestem for the files Avatar requires");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: filestem", ar_dummy, "Filestem for the files Avatar requires");
 
   // This should be a MueLu parameter-to-Avatar parameter mapping (e.g. if Avatar doesn't like spaces)
-  validParamList->set<Teuchos::ParameterList>("avatar: muelu parameter mapping",pl_dummy,"Mapping of MueLu to Avatar Parameters");
+  validParamList->set<Teuchos::ParameterList>("avatar: muelu parameter mapping", pl_dummy, "Mapping of MueLu to Avatar Parameters");
 
   // "Good" Class ID for Avatar
-  validParamList->set<int>("avatar: good class",int_dummy,"Numeric code for class Avatar considers to be good");
+  validParamList->set<int>("avatar: good class", int_dummy, "Numeric code for class Avatar considers to be good");
 
-   // Which drop tol choice heuristic to use
-  validParamList->set<int>("avatar: heuristic",int_dummy,"Numeric code for which heuristic we want to use");
+  // Which drop tol choice heuristic to use
+  validParamList->set<int>("avatar: heuristic", int_dummy, "Numeric code for which heuristic we want to use");
 
   // Bounds file for extrapolation risk
-  validParamList->set<Teuchos::Array<std::string> >("avatar: bounds file",ar_dummy,"Bounds file for Avatar extrapolation risk");
+  validParamList->set<Teuchos::Array<std::string>>("avatar: bounds file", ar_dummy, "Bounds file for Avatar extrapolation risk");
 
   // Add dummy variables at the start
-  validParamList->set<int>("avatar: initial dummy variables",int_dummy,"Number of dummy variables to add at the start");
+  validParamList->set<int>("avatar: initial dummy variables", int_dummy, "Number of dummy variables to add at the start");
 
   // Add dummy variables before the class
-  validParamList->set<int>("avatar: pre-class dummy variables",int_dummy,"Number of dummy variables to add at the before the class");
+  validParamList->set<int>("avatar: pre-class dummy variables", int_dummy, "Number of dummy variables to add at the before the class");
 
   // Value of the dummy variables
-  validParamList->set<int>("avatar: dummy value",int_dummy,"Value of the dummy variables to add at the before/after the class");
+  validParamList->set<int>("avatar: dummy value", int_dummy, "Value of the dummy variables to add at the before/after the class");
 
   return validParamList;
 }
 
-
 // ***********************************************************************
-Teuchos::ArrayRCP<std::string> AvatarInterface::ReadFromFiles(const char * paramName) const {
+Teuchos::ArrayRCP<std::string> AvatarInterface::ReadFromFiles(const char* paramName) const {
   //  const Teuchos::Array<std::string> & tf = params_.get<const Teuchos::Array<std::string> >(paramName);
-  Teuchos::Array<std::string> & tf = params_.get<Teuchos::Array<std::string> >(paramName);
+  Teuchos::Array<std::string>& tf = params_.get<Teuchos::Array<std::string>>(paramName);
   Teuchos::ArrayRCP<std::string> treelist;
   // Only Proc 0 will read the files and print the strings
   if (comm_->getRank() == 0) {
     treelist.resize(tf.size());
-    for(Teuchos_Ordinal i=0; i<tf.size(); i++) {
+    for (Teuchos_Ordinal i = 0; i < tf.size(); i++) {
       std::fstream file;
       std::stringstream ss;
       file.open(tf[i]);
@@ -168,37 +162,33 @@ Teuchos::ArrayRCP<std::string> AvatarInterface::ReadFromFiles(const char * param
   return treelist;
 }
 
-
-
 // ***********************************************************************
 void AvatarInterface::Setup() {
   // Sanity check
-  if(comm_.is_null()) throw std::runtime_error("MueLu::AvatarInterface::Setup(): Communicator cannot be null");
+  if (comm_.is_null()) throw std::runtime_error("MueLu::AvatarInterface::Setup(): Communicator cannot be null");
 
   // Get the avatar strings (NOTE: Only exist on proc 0)
-  if(params_.isParameter("avatar: decision tree strings"))
-    avatarStrings_ = Teuchos::arcpFromArray(params_.get<Teuchos::Array<std::string> >("avatar: decision tree strings"));
+  if (params_.isParameter("avatar: decision tree strings"))
+    avatarStrings_ = Teuchos::arcpFromArray(params_.get<Teuchos::Array<std::string>>("avatar: decision tree strings"));
   else
     avatarStrings_ = ReadFromFiles("avatar: decision tree files");
 
-  if(params_.isParameter("avatar: names strings"))
-    namesStrings_ = Teuchos::arcpFromArray(params_.get<Teuchos::Array<std::string> >("avatar: names strings"));
+  if (params_.isParameter("avatar: names strings"))
+    namesStrings_ = Teuchos::arcpFromArray(params_.get<Teuchos::Array<std::string>>("avatar: names strings"));
   else
     namesStrings_ = ReadFromFiles("avatar: names files");
 
-  if(params_.isParameter("avatar: bounds file"))
+  if (params_.isParameter("avatar: bounds file"))
     boundsString_ = ReadFromFiles("avatar: bounds file");
 
   filestem_ = params_.get<Teuchos::Array<std::string>>("avatar: filestem");
 
-
-  if(comm_->getRank() == 0) {
+  if (comm_->getRank() == 0) {
     // Now actually set up avatar - Avatar's cleanup routine will free the pointer
     //    Avatar_handle* avatar_train(int argc, char **argv, char* names_file, int names_file_is_a_string, char* train_file, int train_file_is_a_string);
     const int namesfile_is_a_string = 1;
     const int treesfile_is_a_string = 1;
-    avatarHandle_ = avatar_load(const_cast<char*>(filestem_[0].c_str()),const_cast<char*>(namesStrings_[0].c_str()),namesfile_is_a_string,const_cast<char*>(avatarStrings_[0].c_str()),treesfile_is_a_string);
-
+    avatarHandle_                   = avatar_load(const_cast<char*>(filestem_[0].c_str()), const_cast<char*>(namesStrings_[0].c_str()), namesfile_is_a_string, const_cast<char*>(avatarStrings_[0].c_str()), treesfile_is_a_string);
   }
 
   // Which class does Avatar consider "good"
@@ -208,46 +198,44 @@ void AvatarInterface::Setup() {
 
   // Unpack the MueLu Mapping into something actionable
   UnpackMueLuMapping();
-
 }
 
 // ***********************************************************************
 void AvatarInterface::Cleanup() {
   avatar_cleanup(avatarHandle_);
-  avatarHandle_=0;
+  avatarHandle_ = 0;
 }
 
-
 // ***********************************************************************
-void AvatarInterface::GenerateFeatureString(const Teuchos::ParameterList & problemFeatures, std::string & featureString) const {
+void AvatarInterface::GenerateFeatureString(const Teuchos::ParameterList& problemFeatures, std::string& featureString) const {
   // NOTE: Assumes that the features are in the same order Avatar wants them.
   std::stringstream ss;
 
   // Initial Dummy Variables
   if (params_.isParameter("avatar: initial dummy variables")) {
-    int num_dummy = params_.get<int>("avatar: initial dummy variables");
-    int dummy_value = params_.get("avatar: dummy value",666);
-  
-    for(int i=0; i<num_dummy; i++)
-      ss<<dummy_value<<",";
+    int num_dummy   = params_.get<int>("avatar: initial dummy variables");
+    int dummy_value = params_.get("avatar: dummy value", 666);
+
+    for (int i = 0; i < num_dummy; i++)
+      ss << dummy_value << ",";
   }
 
-  for(Teuchos::ParameterList::ConstIterator i=problemFeatures.begin(); i != problemFeatures.end(); i++) {
+  for (Teuchos::ParameterList::ConstIterator i = problemFeatures.begin(); i != problemFeatures.end(); i++) {
     //    const std::string& name = problemFeatures.name(i);
     const Teuchos::ParameterEntry& entry = problemFeatures.entry(i);
-    if(i!=problemFeatures.begin()) ss<<",";
-    entry.leftshift(ss,false);  // Because ss<<entry prints out '[unused]' and we don't want that.
+    if (i != problemFeatures.begin()) ss << ",";
+    entry.leftshift(ss, false);  // Because ss<<entry prints out '[unused]' and we don't want that.
   }
   featureString = ss.str();
 }
 
 // ***********************************************************************
 void AvatarInterface::UnpackMueLuMapping() {
-  const Teuchos::ParameterList & mapping = params_.get<Teuchos::ParameterList>("avatar: muelu parameter mapping");
+  const Teuchos::ParameterList& mapping = params_.get<Teuchos::ParameterList>("avatar: muelu parameter mapping");
   // Each MueLu/Avatar parameter pair gets its own sublist.  These must be numerically ordered with no gap
 
-  bool done=false;
-  int idx=0;
+  bool done     = false;
+  int idx       = 0;
   int numParams = mapping.numParams();
 
   mueluParameterName_.resize(numParams);
@@ -255,101 +243,97 @@ void AvatarInterface::UnpackMueLuMapping() {
   mueluParameterValues_.resize(numParams);
   avatarParameterValues_.resize(numParams);
 
-  while(!done) {
+  while (!done) {
     std::stringstream ss;
     ss << "param" << idx;
-    if(mapping.isSublist(ss.str())) {
-      const Teuchos::ParameterList & sublist = mapping.sublist(ss.str());
+    if (mapping.isSublist(ss.str())) {
+      const Teuchos::ParameterList& sublist = mapping.sublist(ss.str());
 
       // Get the names
       mueluParameterName_[idx]  = sublist.get<std::string>("muelu parameter");
       avatarParameterName_[idx] = sublist.get<std::string>("avatar parameter");
 
       // Get the values
-      //FIXME: For now we assume that all of these guys are doubles and their Avatar analogues are doubles
-      mueluParameterValues_[idx]  = sublist.get<Teuchos::Array<double> >("muelu values");
-      avatarParameterValues_[idx] = sublist.get<Teuchos::Array<double> >("avatar values");
+      // FIXME: For now we assume that all of these guys are doubles and their Avatar analogues are doubles
+      mueluParameterValues_[idx]  = sublist.get<Teuchos::Array<double>>("muelu values");
+      avatarParameterValues_[idx] = sublist.get<Teuchos::Array<double>>("avatar values");
 
       idx++;
-    }
-    else {
-      done=true;
+    } else {
+      done = true;
     }
   }
 
-  if(idx!=numParams)
+  if (idx != numParams)
     throw std::runtime_error("MueLu::AvatarInterface::UnpackMueLuMapping(): 'avatar: muelu parameter mapping' has unknown fields");
 }
 // ***********************************************************************
-std::string AvatarInterface::ParamsToString(const std::vector<int> & indices) const {
+std::string AvatarInterface::ParamsToString(const std::vector<int>& indices) const {
   std::stringstream ss;
-  for(Teuchos_Ordinal i=0; i<avatarParameterValues_.size(); i++) {
+  for (Teuchos_Ordinal i = 0; i < avatarParameterValues_.size(); i++) {
     ss << "," << avatarParameterValues_[i][indices[i]];
   }
 
   // Pre-Class dummy variables
   if (params_.isParameter("avatar: pre-class dummy variables")) {
-    int num_dummy = params_.get<int>("avatar: pre-class dummy variables");
-    int dummy_value = params_.get("avatar: dummy value",666);
-    for(int i=0; i<num_dummy; i++)
-      ss<<","<<dummy_value;
+    int num_dummy   = params_.get<int>("avatar: pre-class dummy variables");
+    int dummy_value = params_.get("avatar: dummy value", 666);
+    for (int i = 0; i < num_dummy; i++)
+      ss << "," << dummy_value;
   }
-  
+
   return ss.str();
 }
 
 // ***********************************************************************
-void AvatarInterface::SetIndices(int id,std::vector<int> & indices) const {
+void AvatarInterface::SetIndices(int id, std::vector<int>& indices) const {
   // The combo numbering here starts with the first guy
   int numParams = (int)avatarParameterValues_.size();
-  int curr_id = id;
-  for(int i=0; i<numParams; i++) {
-    int div = avatarParameterValues_[i].size();
-    int mod = curr_id % div;
+  int curr_id   = id;
+  for (int i = 0; i < numParams; i++) {
+    int div    = avatarParameterValues_[i].size();
+    int mod    = curr_id % div;
     indices[i] = mod;
-    curr_id = (curr_id - mod)/div;
+    curr_id    = (curr_id - mod) / div;
   }
 }
 
-
-
 // ***********************************************************************
-void AvatarInterface::GenerateMueLuParametersFromIndex(int id,Teuchos::ParameterList & pl) const {
+void AvatarInterface::GenerateMueLuParametersFromIndex(int id, Teuchos::ParameterList& pl) const {
   // The combo numbering here starts with the first guy
   int numParams = (int)avatarParameterValues_.size();
-  int curr_id = id;
-  for(int i=0; i<numParams; i++) {
+  int curr_id   = id;
+  for (int i = 0; i < numParams; i++) {
     int div = avatarParameterValues_[i].size();
     int mod = curr_id % div;
-    pl.set(mueluParameterName_[i],mueluParameterValues_[i][mod]);
-    curr_id = (curr_id - mod)/div;
+    pl.set(mueluParameterName_[i], mueluParameterValues_[i][mod]);
+    curr_id = (curr_id - mod) / div;
   }
 }
 
-
 // ***********************************************************************
-void AvatarInterface::SetMueLuParameters(const Teuchos::ParameterList & problemFeatures, Teuchos::ParameterList & mueluParams, bool overwrite) const {
+void AvatarInterface::SetMueLuParameters(const Teuchos::ParameterList& problemFeatures, Teuchos::ParameterList& mueluParams, bool overwrite) const {
   Teuchos::ParameterList avatarParams;
   std::string paramString;
 
   if (comm_->getRank() == 0) {
     // Only Rank 0 calls Avatar
-    if(!avatarHandle_) throw std::runtime_error("MueLu::AvatarInterface::SetMueLuParameters(): Setup has not been run");
+    if (!avatarHandle_) throw std::runtime_error("MueLu::AvatarInterface::SetMueLuParameters(): Setup has not been run");
 
     // Turn the problem features into a "trial" string for Avatar
     std::string trialString;
-    GenerateFeatureString(problemFeatures,trialString);
-    
+    GenerateFeatureString(problemFeatures, trialString);
+
     // Compute the number of things we need to test
     int numParams = (int)avatarParameterValues_.size();
     std::vector<int> indices(numParams);
     std::vector<int> sizes(numParams);
     int num_combos = 1;
-    for(int i=0; i<numParams; i++) {
-      sizes[i]    = avatarParameterValues_[i].size();
+    for (int i = 0; i < numParams; i++) {
+      sizes[i] = avatarParameterValues_[i].size();
       num_combos *= avatarParameterValues_[i].size();
     }
-    GetOStream(Runtime0)<< "MueLu::AvatarInterface: Testing "<< num_combos << " option combinations"<<std::endl;
+    GetOStream(Runtime0) << "MueLu::AvatarInterface: Testing " << num_combos << " option combinations" << std::endl;
 
     // For each input parameter to avatar we iterate over its allowable values and then compute the list of options which Avatar
     // views as acceptable
@@ -358,82 +342,79 @@ void AvatarInterface::SetMueLuParameters(const Teuchos::ParameterList & problemF
     std::vector<int> predictions(num_combos, 0);
     std::vector<float> probabilities(num_classes * num_combos, 0);
 
-      std::string testString;
-      for(int i=0; i<num_combos; i++) {
-        SetIndices(i,indices);
-        // Now we add the MueLu parameters into one, enormous Avatar trial string and run avatar once
-        testString += trialString + ParamsToString(indices) + ",0\n";
-      }
+    std::string testString;
+    for (int i = 0; i < num_combos; i++) {
+      SetIndices(i, indices);
+      // Now we add the MueLu parameters into one, enormous Avatar trial string and run avatar once
+      testString += trialString + ParamsToString(indices) + ",0\n";
+    }
 
-      std::cout<<"** Avatar TestString ***\n"<<testString<<std::endl;//DEBUG
+    std::cout << "** Avatar TestString ***\n"
+              << testString << std::endl;  // DEBUG
 
-      int bound_check = true;
-      if(params_.isParameter("avatar: bounds file"))
-         bound_check = checkBounds(testString, boundsString_);
-      
-      // FIXME: Only send in first tree's string
-      //int* avatar_test(Avatar_handle* a, char* test_data_file, int test_data_is_a_string);
-      const int test_data_is_a_string = 1;
-      avatar_test(avatarHandle_,const_cast<char*>(testString.c_str()),test_data_is_a_string,predictions.data(),probabilities.data());
+    int bound_check = true;
+    if (params_.isParameter("avatar: bounds file"))
+      bound_check = checkBounds(testString, boundsString_);
+
+    // FIXME: Only send in first tree's string
+    // int* avatar_test(Avatar_handle* a, char* test_data_file, int test_data_is_a_string);
+    const int test_data_is_a_string = 1;
+    avatar_test(avatarHandle_, const_cast<char*>(testString.c_str()), test_data_is_a_string, predictions.data(), probabilities.data());
 
     // Look at the list of acceptable combinations of options
-    std::vector<int> acceptableCombos; acceptableCombos.reserve(100);
-    for(int i=0; i<num_combos; i++) {
-      if(predictions[i] == avatarGoodClass_) acceptableCombos.push_back(i);
+    std::vector<int> acceptableCombos;
+    acceptableCombos.reserve(100);
+    for (int i = 0; i < num_combos; i++) {
+      if (predictions[i] == avatarGoodClass_) acceptableCombos.push_back(i);
     }
-    GetOStream(Runtime0)<< "MueLu::AvatarInterface: "<< acceptableCombos.size() << " acceptable option combinations found"<<std::endl;
+    GetOStream(Runtime0) << "MueLu::AvatarInterface: " << acceptableCombos.size() << " acceptable option combinations found" << std::endl;
 
     // Did we have any good combos at all?
     int chosen_option_id = 0;
-    if(acceptableCombos.size() == 0) {
-      GetOStream(Runtime0) << "WARNING: MueLu::AvatarInterface: found *no* combinations of options which it believes will perform well on this problem" <<std::endl
-                           << "         An arbitrary set of options will be chosen instead"<<std::endl;
-    }
-    else {
+    if (acceptableCombos.size() == 0) {
+      GetOStream(Runtime0) << "WARNING: MueLu::AvatarInterface: found *no* combinations of options which it believes will perform well on this problem" << std::endl
+                           << "         An arbitrary set of options will be chosen instead" << std::endl;
+    } else {
       // If there is only one acceptable combination, use it;
       // otherwise, find the parameter choice with the highest
       // probability of success
-      if(acceptableCombos.size() == 1){
-	chosen_option_id = acceptableCombos[0];
-      }
-      else {
-	switch (heuristicToUse_){
-	  case 1:
-		chosen_option_id = hybrid(probabilities.data(), acceptableCombos);
-		break;
-	  case 2:
-		chosen_option_id = highProb(probabilities.data(), acceptableCombos);
-		break;
-	  case 3:
-		// Choose the first option in the list of acceptable
-		// combinations; the lowest drop tolerance among the
-		// acceptable combinations
-		chosen_option_id = acceptableCombos[0];
-		break;
-	  case 4:
-		chosen_option_id = lowCrash(probabilities.data(), acceptableCombos);
-		break;
-	  case 5:
-		chosen_option_id = weighted(probabilities.data(), acceptableCombos);
-		break;
+      if (acceptableCombos.size() == 1) {
+        chosen_option_id = acceptableCombos[0];
+      } else {
+        switch (heuristicToUse_) {
+          case 1:
+            chosen_option_id = hybrid(probabilities.data(), acceptableCombos);
+            break;
+          case 2:
+            chosen_option_id = highProb(probabilities.data(), acceptableCombos);
+            break;
+          case 3:
+            // Choose the first option in the list of acceptable
+            // combinations; the lowest drop tolerance among the
+            // acceptable combinations
+            chosen_option_id = acceptableCombos[0];
+            break;
+          case 4:
+            chosen_option_id = lowCrash(probabilities.data(), acceptableCombos);
+            break;
+          case 5:
+            chosen_option_id = weighted(probabilities.data(), acceptableCombos);
+            break;
         }
-
       }
     }
-    
+
     // If mesh parameters are outside bounding box, set drop tolerance
     // to 0, otherwise use avatar recommended drop tolerance
-    if (bound_check == 0){
-      GetOStream(Runtime0) << "WARNING: Extrapolation risk detected, setting drop tolerance to 0" <<std::endl;
-      GenerateMueLuParametersFromIndex(0,avatarParams);
+    if (bound_check == 0) {
+      GetOStream(Runtime0) << "WARNING: Extrapolation risk detected, setting drop tolerance to 0" << std::endl;
+      GenerateMueLuParametersFromIndex(0, avatarParams);
     } else {
-      GenerateMueLuParametersFromIndex(chosen_option_id,avatarParams);
+      GenerateMueLuParametersFromIndex(chosen_option_id, avatarParams);
     }
   }
 
-  Teuchos::updateParametersAndBroadcast(outArg(avatarParams),outArg(mueluParams),*comm_,0,overwrite);
-
-
+  Teuchos::updateParametersAndBroadcast(outArg(avatarParams), outArg(mueluParams), *comm_, 0, overwrite);
 }
 
 int AvatarInterface::checkBounds(std::string trialString, Teuchos::ArrayRCP<std::string> boundsString) const {
@@ -441,11 +422,11 @@ int AvatarInterface::checkBounds(std::string trialString, Teuchos::ArrayRCP<std:
   std::vector<double> vect;
 
   double b;
-  while (ss >> b)  {
+  while (ss >> b) {
     vect.push_back(b);
     if (ss.peek() == ',') ss.ignore();
   }
-  
+
   std::stringstream ssBounds(boundsString[0]);
   std::vector<double> boundsVect;
 
@@ -454,107 +435,105 @@ int AvatarInterface::checkBounds(std::string trialString, Teuchos::ArrayRCP<std:
     if (ssBounds.peek() == ',') ssBounds.ignore();
   }
 
-  int min_idx = (int) std::min(vect.size(),boundsVect.size()/2);
+  int min_idx = (int)std::min(vect.size(), boundsVect.size() / 2);
 
-  bool inbounds=true;
-  for(int i=0; inbounds && i<min_idx; i++)
-    inbounds =  boundsVect[2*i] <= vect[i] && vect[i] <= boundsVect[2*i+1];
+  bool inbounds = true;
+  for (int i = 0; inbounds && i < min_idx; i++)
+    inbounds = boundsVect[2 * i] <= vect[i] && vect[i] <= boundsVect[2 * i + 1];
 
-  return (int) inbounds;
+  return (int)inbounds;
 }
 
-int AvatarInterface::hybrid(float * probabilities, std::vector<int> acceptableCombos) const{
+int AvatarInterface::hybrid(float* probabilities, std::vector<int> acceptableCombos) const {
   float low_crash = probabilities[0];
   float best_prob = probabilities[2];
   float diff;
   int this_combo;
   int chosen_option_id = acceptableCombos[0];
-  for(int x=0; x<(int)acceptableCombos.size(); x++){
+  for (int x = 0; x < (int)acceptableCombos.size(); x++) {
     this_combo = acceptableCombos[x] * 3;
-    diff = probabilities[this_combo] - low_crash;
-     // If this parameter combination has a crash
-     // probability .2 lower than the current "best", we
-     // will use this drop tolerance
-    if(diff < -.2){
-      low_crash =  probabilities[this_combo];
-      best_prob = probabilities[this_combo + 2];
+    diff       = probabilities[this_combo] - low_crash;
+    // If this parameter combination has a crash
+    // probability .2 lower than the current "best", we
+    // will use this drop tolerance
+    if (diff < -.2) {
+      low_crash        = probabilities[this_combo];
+      best_prob        = probabilities[this_combo + 2];
       chosen_option_id = acceptableCombos[x];
     }
     // If this parameter combination has the same
     // or slightly lower crash probability than the
     // current best, we compare their "GOOD" probabilities
-    else if(diff <= 0 && probabilities[this_combo + 2] > best_prob){
-      low_crash =  probabilities[this_combo];
-      best_prob = probabilities[this_combo + 2];
+    else if (diff <= 0 && probabilities[this_combo + 2] > best_prob) {
+      low_crash        = probabilities[this_combo];
+      best_prob        = probabilities[this_combo + 2];
       chosen_option_id = acceptableCombos[x];
     }
   }
   return chosen_option_id;
 }
 
-int AvatarInterface::highProb(float * probabilities, std::vector<int> acceptableCombos) const{
+int AvatarInterface::highProb(float* probabilities, std::vector<int> acceptableCombos) const {
   float high_prob = probabilities[2];
   int this_combo;
   int chosen_option_id = acceptableCombos[0];
-  for(int x=0; x<(int)acceptableCombos.size(); x++){
+  for (int x = 0; x < (int)acceptableCombos.size(); x++) {
     this_combo = acceptableCombos[x] * 3;
     // If this parameter combination has a higher "GOOD"
     // probability, use this combination
-    if(probabilities[this_combo + 2] > high_prob){
-      high_prob = probabilities[this_combo + 2];
+    if (probabilities[this_combo + 2] > high_prob) {
+      high_prob        = probabilities[this_combo + 2];
       chosen_option_id = acceptableCombos[x];
     }
   }
   return chosen_option_id;
 }
 
-int AvatarInterface::lowCrash(float * probabilities, std::vector<int> acceptableCombos) const{
+int AvatarInterface::lowCrash(float* probabilities, std::vector<int> acceptableCombos) const {
   float low_crash = probabilities[0];
   int this_combo;
   int chosen_option_id = acceptableCombos[0];
-  for(int x=0; x<(int)acceptableCombos.size(); x++){
+  for (int x = 0; x < (int)acceptableCombos.size(); x++) {
     this_combo = acceptableCombos[x] * 3;
     // If this parameter combination has a lower "CRASH"
     // probability, use this combination
-    if(probabilities[this_combo] < low_crash){
-      low_crash = probabilities[this_combo];
+    if (probabilities[this_combo] < low_crash) {
+      low_crash        = probabilities[this_combo];
       chosen_option_id = acceptableCombos[x];
     }
   }
   return chosen_option_id;
 }
 
-int AvatarInterface::weighted(float * probabilities, std::vector<int> acceptableCombos) const{
+int AvatarInterface::weighted(float* probabilities, std::vector<int> acceptableCombos) const {
   float low_crash = probabilities[0];
   float best_prob = probabilities[2];
   float diff;
   int this_combo;
   int chosen_option_id = acceptableCombos[0];
-  for(int x=0; x<(int)acceptableCombos.size(); x++){
+  for (int x = 0; x < (int)acceptableCombos.size(); x++) {
     this_combo = acceptableCombos[x] * 3;
-    diff = probabilities[this_combo] - low_crash;
-     // If this parameter combination has a crash
-     // probability .2 lower than the current "best", we
-     // will use this drop tolerance
-    if(diff < -.2){
-      low_crash =  probabilities[this_combo];
-      best_prob = probabilities[this_combo + 2];
+    diff       = probabilities[this_combo] - low_crash;
+    // If this parameter combination has a crash
+    // probability .2 lower than the current "best", we
+    // will use this drop tolerance
+    if (diff < -.2) {
+      low_crash        = probabilities[this_combo];
+      best_prob        = probabilities[this_combo + 2];
       chosen_option_id = acceptableCombos[x];
     }
     // If this parameter combination is within .1
     // or has a slightly lower crash probability than the
     // current best, we compare their "GOOD" probabilities
-    else if(diff <= .1 && probabilities[this_combo + 2] > best_prob){
-      low_crash =  probabilities[this_combo];
-      best_prob = probabilities[this_combo + 2];
+    else if (diff <= .1 && probabilities[this_combo + 2] > best_prob) {
+      low_crash        = probabilities[this_combo];
+      best_prob        = probabilities[this_combo + 2];
       chosen_option_id = acceptableCombos[x];
     }
   }
   return chosen_option_id;
 }
 
+}  // namespace MueLu
 
-}// namespace MueLu
-
-#endif// HAVE_MUELU_AVATAR
-
+#endif  // HAVE_MUELU_AVATAR
diff --git a/packages/muelu/src/Utils/MueLu_AvatarInterface.hpp b/packages/muelu/src/Utils/MueLu_AvatarInterface.hpp
index 7eebedf48247..365b400acae8 100644
--- a/packages/muelu/src/Utils/MueLu_AvatarInterface.hpp
+++ b/packages/muelu/src/Utils/MueLu_AvatarInterface.hpp
@@ -53,91 +53,88 @@
 #include "Teuchos_ParameterList.hpp"
 #include "MueLu_BaseClass.hpp"
 
-
 #ifdef HAVE_MUELU_AVATAR
 
-
 // Forward declaration
 struct Avatar_struct;
 typedef struct Avatar_struct Avatar_handle;
 
 namespace MueLu {
 
-  /*! @class
-    Manages the interface to the Avatar machine learning library.
-
-    Note only proc 0 (as defined by comm) will actually have avatar instantiated.  The options determined
-    by avatar will be broadcast to all processors
-  */
-  class AvatarInterface : public BaseClass {
-
-  public:
-
-    AvatarInterface(Teuchos::RCP<const Teuchos::Comm<int> >& comm ):comm_(comm) {}
+/*! @class
+  Manages the interface to the Avatar machine learning library.
 
-    AvatarInterface(Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList& inParams):comm_(comm),params_(inParams){};
+  Note only proc 0 (as defined by comm) will actually have avatar instantiated.  The options determined
+  by avatar will be broadcast to all processors
+*/
+class AvatarInterface : public BaseClass {
+ public:
+  AvatarInterface(Teuchos::RCP<const Teuchos::Comm<int> >& comm)
+    : comm_(comm) {}
 
-    ~AvatarInterface(){Cleanup();}
+  AvatarInterface(Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList& inParams)
+    : comm_(comm)
+    , params_(inParams){};
 
-    Teuchos::RCP<const Teuchos::ParameterList> GetValidParameterList() const;
+  ~AvatarInterface() { Cleanup(); }
 
-    // Sets the input parameters for the AvatarInterface
-    void SetParameterList(Teuchos::ParameterList& inParams) {params_ = inParams;}
+  Teuchos::RCP<const Teuchos::ParameterList> GetValidParameterList() const;
 
-    // Sets up Avatar
-    void Setup();
+  // Sets the input parameters for the AvatarInterface
+  void SetParameterList(Teuchos::ParameterList& inParams) { params_ = inParams; }
 
-    // Calls Avatar to set MueLu Parameters
-    void SetMueLuParameters(const Teuchos::ParameterList & problemFeatures, Teuchos::ParameterList & mueluParams, bool overwrite=true) const;
+  // Sets up Avatar
+  void Setup();
 
-    // Clean up the handle
-    void Cleanup();
+  // Calls Avatar to set MueLu Parameters
+  void SetMueLuParameters(const Teuchos::ParameterList& problemFeatures, Teuchos::ParameterList& mueluParams, bool overwrite = true) const;
 
-    // Returns 1 if the given parameters are within same
-    // same domain as training data, 0 otherwise
-    int checkBounds(std::string trialString, Teuchos::ArrayRCP<std::string> boundsString) const;
+  // Clean up the handle
+  void Cleanup();
 
-    int hybrid(float * probabilities, std::vector<int> acceptableCombos) const;
+  // Returns 1 if the given parameters are within same
+  // same domain as training data, 0 otherwise
+  int checkBounds(std::string trialString, Teuchos::ArrayRCP<std::string> boundsString) const;
 
-    int highProb(float * probabilities, std::vector<int> acceptableCombos) const;
- 
-    int lowCrash(float * probabilities, std::vector<int> acceptableCombos) const;
+  int hybrid(float* probabilities, std::vector<int> acceptableCombos) const;
 
-    int weighted(float * probabilities, std::vector<int> acceptableCombos) const;
+  int highProb(float* probabilities, std::vector<int> acceptableCombos) const;
 
-  private:
-    // Utility functions
-    Teuchos::ArrayRCP<std::string> ReadFromFiles(const char * param_name) const;
-    void GenerateFeatureString(const Teuchos::ParameterList & problemFeatures, std::string & featureString) const;
-    std::string ParamsToString(const std::vector<int> & indices) const;
-    void SetIndices(int id,std::vector<int> & indices) const;
-    void GenerateMueLuParametersFromIndex(int id,Teuchos::ParameterList & pl) const;
-    void UnpackMueLuMapping();
+  int lowCrash(float* probabilities, std::vector<int> acceptableCombos) const;
 
-    // Cached data
-    Teuchos::RCP<const Teuchos::Comm<int> > comm_;
-    mutable Teuchos::ParameterList params_; // The mutable is a hack to deal with issues in Teuchos
-    Teuchos::ArrayRCP<std::string> avatarStrings_;
-    Teuchos::ArrayRCP<std::string> namesStrings_;
-    Teuchos::Array<std::string> filestem_;
-    Teuchos::ArrayRCP<std::string> boundsString_;
-    int avatarGoodClass_;
-    int heuristicToUse_;
+  int weighted(float* probabilities, std::vector<int> acceptableCombos) const;
 
-    // RCP's don't handle opaque pointers well
-    Avatar_handle * avatarHandle_;
+ private:
+  // Utility functions
+  Teuchos::ArrayRCP<std::string> ReadFromFiles(const char* param_name) const;
+  void GenerateFeatureString(const Teuchos::ParameterList& problemFeatures, std::string& featureString) const;
+  std::string ParamsToString(const std::vector<int>& indices) const;
+  void SetIndices(int id, std::vector<int>& indices) const;
+  void GenerateMueLuParametersFromIndex(int id, Teuchos::ParameterList& pl) const;
+  void UnpackMueLuMapping();
 
-    Teuchos::Array<std::string> mueluParameterName_;
-    Teuchos::Array<std::string> avatarParameterName_;
+  // Cached data
+  Teuchos::RCP<const Teuchos::Comm<int> > comm_;
+  mutable Teuchos::ParameterList params_;  // The mutable is a hack to deal with issues in Teuchos
+  Teuchos::ArrayRCP<std::string> avatarStrings_;
+  Teuchos::ArrayRCP<std::string> namesStrings_;
+  Teuchos::Array<std::string> filestem_;
+  Teuchos::ArrayRCP<std::string> boundsString_;
+  int avatarGoodClass_;
+  int heuristicToUse_;
 
-    Teuchos::ArrayRCP<Teuchos::Array<double> > mueluParameterValues_;
-    Teuchos::ArrayRCP<Teuchos::Array<double> > avatarParameterValues_;
+  // RCP's don't handle opaque pointers well
+  Avatar_handle* avatarHandle_;
 
-  };
+  Teuchos::Array<std::string> mueluParameterName_;
+  Teuchos::Array<std::string> avatarParameterName_;
 
+  Teuchos::ArrayRCP<Teuchos::Array<double> > mueluParameterValues_;
+  Teuchos::ArrayRCP<Teuchos::Array<double> > avatarParameterValues_;
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif// HAVE_MUELU_AVATAR
+#endif  // HAVE_MUELU_AVATAR
 
-#endif // MUELU_AVATARINTERFACE_HPP
+#endif  // MUELU_AVATARINTERFACE_HPP
diff --git a/packages/muelu/src/Utils/MueLu_BoostGraphviz.hpp b/packages/muelu/src/Utils/MueLu_BoostGraphviz.hpp
index 42b8b80ff1ee..c504dcb7676f 100644
--- a/packages/muelu/src/Utils/MueLu_BoostGraphviz.hpp
+++ b/packages/muelu/src/Utils/MueLu_BoostGraphviz.hpp
@@ -13,26 +13,26 @@
 #ifdef __GNUC__
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wshadow"
-#endif // __GNUC__
+#endif  // __GNUC__
 
 #include <boost/graph/graphviz.hpp>
 
 #ifdef __GNUC__
 #pragma GCC diagnostic pop
-#endif // __GNUC__
+#endif  // __GNUC__
 
 // define boost graph types
 typedef boost::adjacency_list<boost::vecS, boost::vecS, boost::directedS,
-        boost::property<boost::vertex_name_t, std::string,
-        boost::property<boost::vertex_color_t, std::string,
-        boost::property<boost::vertex_index_t, std::string> > >,
-        boost::property<boost::edge_name_t, std::string,
-        boost::property<boost::edge_color_t, std::string> > >               BoostGraph;
-typedef boost::dynamic_properties                                           BoostProperties;
-typedef boost::graph_traits<BoostGraph>::vertex_descriptor                  BoostVertex;
-typedef boost::graph_traits<BoostGraph>::edge_descriptor                    BoostEdge;
-
-
-#endif // HAVE_MUELU_BOOST && HAVE_MUELU_BOOST_FOR_REAL
-
-#endif // MUELU_BOOSTGRAPHVIZ_HPP
+                              boost::property<boost::vertex_name_t, std::string,
+                                              boost::property<boost::vertex_color_t, std::string,
+                                                              boost::property<boost::vertex_index_t, std::string> > >,
+                              boost::property<boost::edge_name_t, std::string,
+                                              boost::property<boost::edge_color_t, std::string> > >
+    BoostGraph;
+typedef boost::dynamic_properties BoostProperties;
+typedef boost::graph_traits<BoostGraph>::vertex_descriptor BoostVertex;
+typedef boost::graph_traits<BoostGraph>::edge_descriptor BoostEdge;
+
+#endif  // HAVE_MUELU_BOOST && HAVE_MUELU_BOOST_FOR_REAL
+
+#endif  // MUELU_BOOSTGRAPHVIZ_HPP
diff --git a/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_decl.hpp
index 2603d727df63..40b60724724d 100644
--- a/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_decl.hpp
@@ -54,76 +54,75 @@
 #include "MueLu_VisualizationHelpers.hpp"
 #include "MueLu_CoarseningVisualizationFactory_fwd.hpp"
 
-
 namespace MueLu {
 
-  class Level;
-
-  /*!
-    @class CoarseningVisualizationFactory class.
-    @brief Factory to visualize coarsening information using prolongation operators
-
-    @ingroup MueLuVisualizationClasses
-
-    ## Input/output of CoarseningVisualizationFactory ##
-
-    ### User parameters of CoarseningVisualizationFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | visualization: output filename           | string  | viz%LEVELID  |  | * |   | filename for VTK-style visualization output |
-    | visualization: output file: time step    | int     | 0 |  | * |   | time step (overwrites '%TIMESTEP' in output file name) |
-    | visualization: output file: iter         | int     | 0 |  | * |   | nonlinear iteration (overwrites '%ITER' in output file name) |
-    | visualization: style                     | string  | Point Cloud |   | * |  | style of aggregation visualization for VTK output. Can be either "Point Cloud", "Jacks", or "Convex Hulls" |
-    | visualization: fine graph edges | bool | false  |   | * |  | Draw fine node connections in VTK output (only works for 1 dofs per node!) |
-    | visualization: build colormap | bool | false  |   | * |  | Output a random color map for paraView in a separate xml file. |
-    | P | Factory | Teuchos::null  |   | * | * | Prolongator factory. The user has to declare either P or Ptent but not both at the same time. |
-    | Ptent | Factory | Teuchos::null  |   | * | * | Tentative prolongator factory. The user has to declare either P or Ptent but not both at the same time. |
-    | Coordinates | Factory | Teuchos::null  |   | * | * | Factory for Coordinates on fine level |
-    | Graph | Factory | Teuchos::null  |   | * | * | Factory for Graph of A |
-
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoarseningVisualizationFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see CoarseningVisualizationFactory::DeclareInput).
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class CoarseningVisualizationFactory : public TwoLevelFactoryBase, public VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+class Level;
+
+/*!
+  @class CoarseningVisualizationFactory class.
+  @brief Factory to visualize coarsening information using prolongation operators
+
+  @ingroup MueLuVisualizationClasses
+
+  ## Input/output of CoarseningVisualizationFactory ##
+
+  ### User parameters of CoarseningVisualizationFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | visualization: output filename           | string  | viz%LEVELID  |  | * |   | filename for VTK-style visualization output |
+  | visualization: output file: time step    | int     | 0 |  | * |   | time step (overwrites '%TIMESTEP' in output file name) |
+  | visualization: output file: iter         | int     | 0 |  | * |   | nonlinear iteration (overwrites '%ITER' in output file name) |
+  | visualization: style                     | string  | Point Cloud |   | * |  | style of aggregation visualization for VTK output. Can be either "Point Cloud", "Jacks", or "Convex Hulls" |
+  | visualization: fine graph edges | bool | false  |   | * |  | Draw fine node connections in VTK output (only works for 1 dofs per node!) |
+  | visualization: build colormap | bool | false  |   | * |  | Output a random color map for paraView in a separate xml file. |
+  | P | Factory | Teuchos::null  |   | * | * | Prolongator factory. The user has to declare either P or Ptent but not both at the same time. |
+  | Ptent | Factory | Teuchos::null  |   | * | * | Tentative prolongator factory. The user has to declare either P or Ptent but not both at the same time. |
+  | Coordinates | Factory | Teuchos::null  |   | * | * | Factory for Coordinates on fine level |
+  | Graph | Factory | Teuchos::null  |   | * | * | Factory for Graph of A |
+
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see CoarseningVisualizationFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see CoarseningVisualizationFactory::DeclareInput).
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CoarseningVisualizationFactory : public TwoLevelFactoryBase, public VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #undef MUELU_COARSENINGVISUALIZATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    CoarseningVisualizationFactory() { }
+  //! Constructor.
+  CoarseningVisualizationFactory() {}
 
-    //! Destructor.
-    virtual ~CoarseningVisualizationFactory() { }
-    //@}
+  //! Destructor.
+  virtual ~CoarseningVisualizationFactory() {}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! Input
-    //@{
+  //! Input
+  //@{
 
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //@{
-    //! @name Build methods.
+  //@{
+  //! @name Build methods.
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-   }; // class CoarseningVisualizationFactory
-} // namespace MueLu
+};  // class CoarseningVisualizationFactory
+}  // namespace MueLu
 
 #define MUELU_COARSENINGVISUALIZATIONFACTORY_SHORT
 
diff --git a/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_def.hpp b/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_def.hpp
index f3f4c1a8f38e..2a4a33baf502 100644
--- a/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_CoarseningVisualizationFactory_def.hpp
@@ -53,267 +53,263 @@
 #include "MueLu_CoarseningVisualizationFactory_decl.hpp"
 #include "MueLu_Level.hpp"
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const ParameterList> CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-
-    RCP<ParameterList> validParamList = VisualizationHelpers::GetValidParameterList();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const ParameterList> CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = VisualizationHelpers::GetValidParameterList();
 
-    validParamList->set< int >                   ("visualization: start level",             0,                     "visualize only levels with level ids greater or equal than start level");// Remove me?
+  validParamList->set<int>("visualization: start level", 0, "visualize only levels with level ids greater or equal than start level");  // Remove me?
 
-    validParamList->set< RCP<const FactoryBase> >("P",           Teuchos::null, "Prolongator factory. The user has to declare either P or Ptent but not both at the same time.");
-    validParamList->set< RCP<const FactoryBase> >("Ptent",       Teuchos::null, "Tentative prolongator factory. The user has to declare either P or Ptent as input but not both at the same time");
-    validParamList->set< RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for Coordinates.");
-    validParamList->set< RCP<const FactoryBase> >("Graph",       Teuchos::null, "Factory for Graph.");
+  validParamList->set<RCP<const FactoryBase> >("P", Teuchos::null, "Prolongator factory. The user has to declare either P or Ptent but not both at the same time.");
+  validParamList->set<RCP<const FactoryBase> >("Ptent", Teuchos::null, "Tentative prolongator factory. The user has to declare either P or Ptent as input but not both at the same time");
+  validParamList->set<RCP<const FactoryBase> >("Coordinates", Teuchos::null, "Factory for Coordinates.");
+  validParamList->set<RCP<const FactoryBase> >("Graph", Teuchos::null, "Factory for Graph.");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
-    this->Input(fineLevel, "Coordinates");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &coarseLevel) const {
+  this->Input(fineLevel, "Coordinates");
 
-    const ParameterList & pL = this->GetParameterList();
-    TEUCHOS_TEST_FOR_EXCEPTION(pL.isParameter("P") && GetFactory("P") != Teuchos::null &&
-                               pL.isParameter("Ptent") && GetFactory("Ptent") != Teuchos::null, Exceptions::RuntimeError,
-                "You must not declare both P and Ptent. Use only once for visualization.");
-    TEUCHOS_TEST_FOR_EXCEPTION(GetFactory("P") == Teuchos::null && GetFactory("Ptent") == Teuchos::null, Exceptions::RuntimeError,
-                "You have to either declare P or Ptent for visualization, but not both.");
+  const ParameterList &pL = this->GetParameterList();
+  TEUCHOS_TEST_FOR_EXCEPTION(pL.isParameter("P") && GetFactory("P") != Teuchos::null &&
+                                 pL.isParameter("Ptent") && GetFactory("Ptent") != Teuchos::null,
+                             Exceptions::RuntimeError,
+                             "You must not declare both P and Ptent. Use only once for visualization.");
+  TEUCHOS_TEST_FOR_EXCEPTION(GetFactory("P") == Teuchos::null && GetFactory("Ptent") == Teuchos::null, Exceptions::RuntimeError,
+                             "You have to either declare P or Ptent for visualization, but not both.");
 
-    if (GetFactory("P") != Teuchos::null && GetFactory("Ptent") == Teuchos::null)
-      this->Input(coarseLevel, "P");
-    else if (GetFactory("Ptent") != Teuchos::null && GetFactory("P") == Teuchos::null)
-      this->Input(coarseLevel, "Ptent");
+  if (GetFactory("P") != Teuchos::null && GetFactory("Ptent") == Teuchos::null)
+    this->Input(coarseLevel, "P");
+  else if (GetFactory("Ptent") != Teuchos::null && GetFactory("P") == Teuchos::null)
+    this->Input(coarseLevel, "Ptent");
 
-    if(pL.get<bool>("visualization: fine graph edges"))
-      Input(fineLevel, "Graph");
+  if (pL.get<bool>("visualization: fine graph edges"))
+    Input(fineLevel, "Graph");
 #if 0
     if(pL.get<bool>("visualization: coarse graph edges")) {
       Input(coarseLevel, "Coordinates");
       Input(coarseLevel, "Graph");
     }
 #endif
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
-
-    RCP<GraphBase> fineGraph = Teuchos::null;
-    RCP<Matrix>    P         = Teuchos::null;
-    const ParameterList & pL = this->GetParameterList();
-    if (this->GetFactory("P") != Teuchos::null && this->GetFactory("Ptent") == Teuchos::null)
-      P = Get< RCP<Matrix> >(coarseLevel, "P");
-    if (GetFactory("Ptent") != Teuchos::null && GetFactory("P") == Teuchos::null)
-      P = Get< RCP<Matrix> >(coarseLevel, "Ptent");
-
-    RCP<const Teuchos::Comm<int> > comm = P->getRowMap()->getComm();
-
-    LocalOrdinal dofsPerNode = 1;
-    LocalOrdinal stridedRowOffset = 0;
-    RCP<const StridedMap> strRowMap    = Teuchos::null;
-    if (P->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps")) != Teuchos::null) {
-      strRowMap = Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps"));
-      LocalOrdinal blockid       = strRowMap->getStridedBlockId();
-      if (blockid > -1) {
-        std::vector<size_t> stridingInfo = strRowMap->getStridingData();
-        for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
-          stridedRowOffset += stridingInfo[j];
-        dofsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
-      } else {
-        dofsPerNode = strRowMap->getFixedBlockSize();
-      }
-      GetOStream(Runtime1) << "CoarseningVisualizationFactory::Build():" << " #dofs per node = " << dofsPerNode << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoarseningVisualizationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &coarseLevel) const {
+  RCP<GraphBase> fineGraph = Teuchos::null;
+  RCP<Matrix> P            = Teuchos::null;
+  const ParameterList &pL  = this->GetParameterList();
+  if (this->GetFactory("P") != Teuchos::null && this->GetFactory("Ptent") == Teuchos::null)
+    P = Get<RCP<Matrix> >(coarseLevel, "P");
+  if (GetFactory("Ptent") != Teuchos::null && GetFactory("P") == Teuchos::null)
+    P = Get<RCP<Matrix> >(coarseLevel, "Ptent");
+
+  RCP<const Teuchos::Comm<int> > comm = P->getRowMap()->getComm();
+
+  LocalOrdinal dofsPerNode        = 1;
+  LocalOrdinal stridedRowOffset   = 0;
+  RCP<const StridedMap> strRowMap = Teuchos::null;
+  if (P->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps")) != Teuchos::null) {
+    strRowMap            = Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps"));
+    LocalOrdinal blockid = strRowMap->getStridedBlockId();
+    if (blockid > -1) {
+      std::vector<size_t> stridingInfo = strRowMap->getStridingData();
+      for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
+        stridedRowOffset += stridingInfo[j];
+      dofsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+    } else {
+      dofsPerNode = strRowMap->getFixedBlockSize();
     }
+    GetOStream(Runtime1) << "CoarseningVisualizationFactory::Build():"
+                         << " #dofs per node = " << dofsPerNode << std::endl;
+  }
 
-    LocalOrdinal columnsPerNode = dofsPerNode;
-    LocalOrdinal stridedColumnOffset = 0;
-    RCP<const StridedMap> strDomainMap = Teuchos::null;
-    if (P->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps")) != Teuchos::null) {
-      strDomainMap = Teuchos::rcp_dynamic_cast<const StridedMap>(P->getColMap("stridedMaps"));
-      LocalOrdinal blockid = strDomainMap->getStridedBlockId();
-
-      if (blockid > -1) {
-        std::vector<size_t> stridingInfo = strDomainMap->getStridingData();
-        for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
-          stridedColumnOffset += stridingInfo[j];
-        columnsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
-      } else {
-        columnsPerNode = strDomainMap->getFixedBlockSize();
-      }
-      GetOStream(Runtime1) << "CoarseningVisualizationFactory::Build():" << " #columns per node = " << columnsPerNode << std::endl;
+  LocalOrdinal columnsPerNode        = dofsPerNode;
+  LocalOrdinal stridedColumnOffset   = 0;
+  RCP<const StridedMap> strDomainMap = Teuchos::null;
+  if (P->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(P->getRowMap("stridedMaps")) != Teuchos::null) {
+    strDomainMap         = Teuchos::rcp_dynamic_cast<const StridedMap>(P->getColMap("stridedMaps"));
+    LocalOrdinal blockid = strDomainMap->getStridedBlockId();
+
+    if (blockid > -1) {
+      std::vector<size_t> stridingInfo = strDomainMap->getStridingData();
+      for (size_t j = 0; j < Teuchos::as<size_t>(blockid); j++)
+        stridedColumnOffset += stridingInfo[j];
+      columnsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+    } else {
+      columnsPerNode = strDomainMap->getFixedBlockSize();
     }
+    GetOStream(Runtime1) << "CoarseningVisualizationFactory::Build():"
+                         << " #columns per node = " << columnsPerNode << std::endl;
+  }
 
-    // TODO add support for overlapping aggregates
-    TEUCHOS_TEST_FOR_EXCEPTION(strDomainMap->getLocalNumElements() != P->getColMap()->getLocalNumElements(), Exceptions::RuntimeError,
-                                               "CoarseningVisualization only supports non-overlapping transfers");
+  // TODO add support for overlapping aggregates
+  TEUCHOS_TEST_FOR_EXCEPTION(strDomainMap->getLocalNumElements() != P->getColMap()->getLocalNumElements(), Exceptions::RuntimeError,
+                             "CoarseningVisualization only supports non-overlapping transfers");
 
-    // number of local "aggregates"
-    LocalOrdinal numLocalAggs = strDomainMap->getLocalNumElements() / columnsPerNode;
-    std::vector< std::set<LocalOrdinal> > localAggs(numLocalAggs);
+  // number of local "aggregates"
+  LocalOrdinal numLocalAggs = strDomainMap->getLocalNumElements() / columnsPerNode;
+  std::vector<std::set<LocalOrdinal> > localAggs(numLocalAggs);
 
-    // do loop over all local rows of prolongator and extract column information
-    for (LO row = 0; row < Teuchos::as<LO>(P->getRowMap()->getLocalNumElements()); ++row) {
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      P->getLocalRowView(row, indices, vals);
+  // do loop over all local rows of prolongator and extract column information
+  for (LO row = 0; row < Teuchos::as<LO>(P->getRowMap()->getLocalNumElements()); ++row) {
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    P->getLocalRowView(row, indices, vals);
 
-      for(typename ArrayView<const LO>::iterator c = indices.begin(); c != indices.end(); ++c) {
-        localAggs[(*c)/columnsPerNode].insert(row/dofsPerNode);
-      }
+    for (typename ArrayView<const LO>::iterator c = indices.begin(); c != indices.end(); ++c) {
+      localAggs[(*c) / columnsPerNode].insert(row / dofsPerNode);
     }
+  }
 
-    // determine number of "aggs" per proc and calculate local "agg" offset...
-    std::vector<int> myLocalAggsPerProc(comm->getSize(),0);
-    std::vector<int> numLocalAggsPerProc(comm->getSize(),0);
-    myLocalAggsPerProc[comm->getRank()] = numLocalAggs;
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_MAX,comm->getSize(),&myLocalAggsPerProc[0],&numLocalAggsPerProc[0]);
+  // determine number of "aggs" per proc and calculate local "agg" offset...
+  std::vector<int> myLocalAggsPerProc(comm->getSize(), 0);
+  std::vector<int> numLocalAggsPerProc(comm->getSize(), 0);
+  myLocalAggsPerProc[comm->getRank()] = numLocalAggs;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, comm->getSize(), &myLocalAggsPerProc[0], &numLocalAggsPerProc[0]);
 
-    LocalOrdinal myAggOffset = 0;
-    for(int i = 0; i < comm->getRank(); ++i) {
-      myAggOffset += numLocalAggsPerProc[i];
-    }
+  LocalOrdinal myAggOffset = 0;
+  for (int i = 0; i < comm->getRank(); ++i) {
+    myAggOffset += numLocalAggsPerProc[i];
+  }
 
-    /*for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
+  /*for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
 
-      std::cout << "PROC: " << comm->getRank() << " Local aggregate: " << i + myAggOffset << " with nodes: ";
-      for( typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
-        std::cout << *it << ", ";
-      }
-      std::cout << std::endl;
-    }*/
+    std::cout << "PROC: " << comm->getRank() << " Local aggregate: " << i + myAggOffset << " with nodes: ";
+    for( typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
+      std::cout << *it << ", ";
+    }
+    std::cout << std::endl;
+  }*/
 
-    // get fine level coordinate information
-    Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > coords = Get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >(fineLevel, "Coordinates");
+  // get fine level coordinate information
+  Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > coords = Get<RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > >(fineLevel, "Coordinates");
 
-    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<LO>(P->getRowMap()->getLocalNumElements()) / dofsPerNode != Teuchos::as<LocalOrdinal>(coords->getLocalLength()), Exceptions::RuntimeError,
-                                           "Number of fine level nodes in coordinates is inconsistent with dof based information");
+  TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<LO>(P->getRowMap()->getLocalNumElements()) / dofsPerNode != Teuchos::as<LocalOrdinal>(coords->getLocalLength()), Exceptions::RuntimeError,
+                             "Number of fine level nodes in coordinates is inconsistent with dof based information");
 
-    // communicate fine level coordinates if necessary
-    if (pL.get<bool>("visualization: fine graph edges")) {
-      fineGraph = Get<RCP<GraphBase> >(fineLevel, "Graph");
+  // communicate fine level coordinates if necessary
+  if (pL.get<bool>("visualization: fine graph edges")) {
+    fineGraph = Get<RCP<GraphBase> >(fineLevel, "Graph");
 
-      RCP<Import> coordImporter = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coords->getMap(), fineGraph->GetImportMap());
-      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>::Build(fineGraph->GetImportMap(), coords->getNumVectors());
-      ghostedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
-      coords = ghostedCoords;
-    }
+    RCP<Import> coordImporter                                                                                                          = Xpetra::ImportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(coords->getMap(), fineGraph->GetImportMap());
+    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > ghostedCoords = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>::Build(fineGraph->GetImportMap(), coords->getNumVectors());
+    ghostedCoords->doImport(*coords, *coordImporter, Xpetra::INSERT);
+    coords = ghostedCoords;
+  }
 
-    Teuchos::RCP<const Map> nodeMap = coords->getMap();
+  Teuchos::RCP<const Map> nodeMap = coords->getMap();
 
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(0));
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(1));
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
-    if(coords->getNumVectors() == 3) {
-      zCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(2));
-    }
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> xCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(0));
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> yCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(1));
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> zCoords = Teuchos::null;
+  if (coords->getNumVectors() == 3) {
+    zCoords = Teuchos::arcp_reinterpret_cast<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>(coords->getData(2));
+  }
 
-    // determine number of nodes on fine level
-    LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coords->getLocalLength());
+  // determine number of nodes on fine level
+  LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coords->getLocalLength());
 
-    // create vertex2AggId array
-    ArrayRCP<LocalOrdinal> vertex2AggId(numFineNodes, -1);
-    for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
-      // TODO: check if entry = -1
-      for( typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
-        vertex2AggId[*it] = i;
-      }
+  // create vertex2AggId array
+  ArrayRCP<LocalOrdinal> vertex2AggId(numFineNodes, -1);
+  for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
+    // TODO: check if entry = -1
+    for (typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
+      vertex2AggId[*it] = i;
     }
+  }
 
-    // we have no information which node is root and which is not
-    // we could either look at the entries in P again or build some new root nodes
-    // assuming that root nodes are usually at the center of the aggregate
-    std::vector<bool> isRoot(numFineNodes, false);
-    for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
-
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType xCenter = 0.0;
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType yCenter = 0.0;
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType zCenter = 0.0;
-
-      // loop over all nodes in aggregate i and determine center coordinates of aggregate
-      for( typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
-        xCenter += xCoords[*it];
-        yCenter += yCoords[*it];
-        if(coords->getNumVectors() == 3) zCenter += zCoords[*it];
-      }
-      xCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
-      yCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
-      zCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
-
-      // loop over all nodes in aggregate i and find node which is closest to aggregate center
-      LocalOrdinal rootCandidate = -1;
-      typename Teuchos::ScalarTraits<Scalar>::coordinateType minDistance = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
-      for( typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
-        typename Teuchos::ScalarTraits<Scalar>::coordinateType tempx = xCenter - xCoords[*it];
-        typename Teuchos::ScalarTraits<Scalar>::coordinateType tempy = yCenter - yCoords[*it];
-        typename Teuchos::ScalarTraits<Scalar>::coordinateType tempz = 0.0;
-        if(coords->getNumVectors() == 3) tempz = zCenter - zCoords[*it];
-        typename Teuchos::ScalarTraits<Scalar>::coordinateType mydistance = 0.0;
-        mydistance += tempx*tempx;
-        mydistance += tempy*tempy;
-        mydistance += tempz*tempz;
-        mydistance = sqrt(mydistance);
-        if (mydistance <= minDistance) {
-          minDistance = mydistance;
-          rootCandidate = *it;
-        }
-      }
-
-      isRoot[rootCandidate] = true;
+  // we have no information which node is root and which is not
+  // we could either look at the entries in P again or build some new root nodes
+  // assuming that root nodes are usually at the center of the aggregate
+  std::vector<bool> isRoot(numFineNodes, false);
+  for (LocalOrdinal i = 0; i < numLocalAggs; ++i) {
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType xCenter = 0.0;
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType yCenter = 0.0;
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType zCenter = 0.0;
+
+    // loop over all nodes in aggregate i and determine center coordinates of aggregate
+    for (typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
+      xCenter += xCoords[*it];
+      yCenter += yCoords[*it];
+      if (coords->getNumVectors() == 3) zCenter += zCoords[*it];
     }
-
-    std::vector<LocalOrdinal> vertices;
-    std::vector<LocalOrdinal> geomSize;
-    int vizLevel = pL.get<int>("visualization: start level");
-    if(vizLevel <= fineLevel.GetLevelID()) {
-
-      std::string aggStyle = pL.get<std::string>("visualization: style");
-      if(aggStyle == "Point Cloud")
-        this->doPointCloud(vertices, geomSize, numLocalAggs, numFineNodes);
-      else if(aggStyle == "Jacks")
-        this->doJacks(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId);
-      else if(aggStyle == "Convex Hulls") {
-        // TODO do a smarter distinction and check the number of z levels...
-        // loop over all coordinates and collect coordinate components in sets...
-        if(coords->getNumVectors() == 3)
-          this->doConvexHulls3D(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId, xCoords, yCoords, zCoords);
-        else if(coords->getNumVectors() == 2)
-          this->doConvexHulls2D(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId, xCoords, yCoords);
-      }
-      else
-      {
-        GetOStream(Warnings0) << "   Warning: Unrecognized agg style.\nPossible values are Point Cloud, Jacks, Convex Hulls.\nDefaulting to Point Cloud." << std::endl;
-        aggStyle = "Point Cloud";
-        this->doPointCloud(vertices, geomSize, numLocalAggs, numFineNodes);
+    xCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
+    yCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
+    zCenter /= Teuchos::as<LocalOrdinal>(localAggs[i].size());
+
+    // loop over all nodes in aggregate i and find node which is closest to aggregate center
+    LocalOrdinal rootCandidate                                         = -1;
+    typename Teuchos::ScalarTraits<Scalar>::coordinateType minDistance = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::one() / Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<Scalar>::coordinateType>::sfmin();
+    for (typename std::set<LocalOrdinal>::iterator it = localAggs[i].begin(); it != localAggs[i].end(); ++it) {
+      typename Teuchos::ScalarTraits<Scalar>::coordinateType tempx = xCenter - xCoords[*it];
+      typename Teuchos::ScalarTraits<Scalar>::coordinateType tempy = yCenter - yCoords[*it];
+      typename Teuchos::ScalarTraits<Scalar>::coordinateType tempz = 0.0;
+      if (coords->getNumVectors() == 3) tempz = zCenter - zCoords[*it];
+      typename Teuchos::ScalarTraits<Scalar>::coordinateType mydistance = 0.0;
+      mydistance += tempx * tempx;
+      mydistance += tempy * tempy;
+      mydistance += tempz * tempz;
+      mydistance = sqrt(mydistance);
+      if (mydistance <= minDistance) {
+        minDistance   = mydistance;
+        rootCandidate = *it;
       }
     }
 
-    // write out fine edge information
-    if(pL.get<bool>("visualization: fine graph edges")) {
-      TEUCHOS_TEST_FOR_EXCEPTION(fineGraph == Teuchos::null, Exceptions::RuntimeError,
-         "Could not get information about fine graph.");
-
-      std::vector<LocalOrdinal> fine_edges_vertices;
-      std::vector<LocalOrdinal> fine_edges_geomSize;
-      this->doGraphEdges(fine_edges_vertices, fine_edges_geomSize, fineGraph, xCoords, yCoords, zCoords);
-
-      std::string fEdgeFineFile = this->getFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
-      std::string fEdgeFile = fEdgeFineFile.insert(fEdgeFineFile.rfind(".vtu"), "-finegraph");
-      std::ofstream edgeStream(fEdgeFile.c_str());
-
-      std::vector<int> uniqueFineEdges = this->makeUnique(fine_edges_vertices);
-      this->writeFileVTKOpening(edgeStream, uniqueFineEdges, fine_edges_geomSize);
-      this->writeFileVTKNodes(edgeStream, uniqueFineEdges, nodeMap);
-      this->writeFileVTKData(edgeStream, uniqueFineEdges, myAggOffset, vertex2AggId, comm->getRank());
-      this->writeFileVTKCoordinates(edgeStream, uniqueFineEdges, xCoords, yCoords, zCoords, coords->getNumVectors());
-      this->writeFileVTKCells(edgeStream, uniqueFineEdges, fine_edges_vertices, fine_edges_geomSize);
-      this->writeFileVTKClosing(edgeStream);
-      edgeStream.close();
+    isRoot[rootCandidate] = true;
+  }
+
+  std::vector<LocalOrdinal> vertices;
+  std::vector<LocalOrdinal> geomSize;
+  int vizLevel = pL.get<int>("visualization: start level");
+  if (vizLevel <= fineLevel.GetLevelID()) {
+    std::string aggStyle = pL.get<std::string>("visualization: style");
+    if (aggStyle == "Point Cloud")
+      this->doPointCloud(vertices, geomSize, numLocalAggs, numFineNodes);
+    else if (aggStyle == "Jacks")
+      this->doJacks(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId);
+    else if (aggStyle == "Convex Hulls") {
+      // TODO do a smarter distinction and check the number of z levels...
+      // loop over all coordinates and collect coordinate components in sets...
+      if (coords->getNumVectors() == 3)
+        this->doConvexHulls3D(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId, xCoords, yCoords, zCoords);
+      else if (coords->getNumVectors() == 2)
+        this->doConvexHulls2D(vertices, geomSize, numLocalAggs, numFineNodes, isRoot, vertex2AggId, xCoords, yCoords);
+    } else {
+      GetOStream(Warnings0) << "   Warning: Unrecognized agg style.\nPossible values are Point Cloud, Jacks, Convex Hulls.\nDefaulting to Point Cloud." << std::endl;
+      aggStyle = "Point Cloud";
+      this->doPointCloud(vertices, geomSize, numLocalAggs, numFineNodes);
     }
+  }
+
+  // write out fine edge information
+  if (pL.get<bool>("visualization: fine graph edges")) {
+    TEUCHOS_TEST_FOR_EXCEPTION(fineGraph == Teuchos::null, Exceptions::RuntimeError,
+                               "Could not get information about fine graph.");
+
+    std::vector<LocalOrdinal> fine_edges_vertices;
+    std::vector<LocalOrdinal> fine_edges_geomSize;
+    this->doGraphEdges(fine_edges_vertices, fine_edges_geomSize, fineGraph, xCoords, yCoords, zCoords);
+
+    std::string fEdgeFineFile = this->getFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
+    std::string fEdgeFile     = fEdgeFineFile.insert(fEdgeFineFile.rfind(".vtu"), "-finegraph");
+    std::ofstream edgeStream(fEdgeFile.c_str());
+
+    std::vector<int> uniqueFineEdges = this->makeUnique(fine_edges_vertices);
+    this->writeFileVTKOpening(edgeStream, uniqueFineEdges, fine_edges_geomSize);
+    this->writeFileVTKNodes(edgeStream, uniqueFineEdges, nodeMap);
+    this->writeFileVTKData(edgeStream, uniqueFineEdges, myAggOffset, vertex2AggId, comm->getRank());
+    this->writeFileVTKCoordinates(edgeStream, uniqueFineEdges, xCoords, yCoords, zCoords, coords->getNumVectors());
+    this->writeFileVTKCells(edgeStream, uniqueFineEdges, fine_edges_vertices, fine_edges_geomSize);
+    this->writeFileVTKClosing(edgeStream);
+    edgeStream.close();
+  }
 
-    // communicate fine level coordinates if necessary
-#if 0 // we don't have access to the coarse graph
+  // communicate fine level coordinates if necessary
+#if 0  // we don't have access to the coarse graph
     if (pL.get<bool>("visualization: coarse graph edges")) {
       RCP<GraphBase> coarseGraph = Get<RCP<GraphBase> >(coarseLevel, "Graph");
 
@@ -352,34 +348,34 @@ namespace MueLu {
     }
 #endif
 
-    if(pL.get<int>("visualization: start level") <= fineLevel.GetLevelID()) {
-      // write out coarsening information
-      std::string filenameToWrite = this->getFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
-      std::ofstream fout (filenameToWrite.c_str());
-
-      std::vector<int> uniqueFine = this->makeUnique(vertices);
-      this->writeFileVTKOpening(fout, uniqueFine, geomSize);
-      this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
-      this->writeFileVTKData(fout, uniqueFine, myAggOffset, vertex2AggId, comm->getRank());
-      this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coords->getNumVectors());
-      this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
-      this->writeFileVTKClosing(fout);
-      fout.close();
-
-      // create pvtu file
-      if(comm->getRank() == 0) {
-        std::string pvtuFilename = this->getPVTUFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
-        std::string baseFname = this->getBaseFileName(comm->getSize(), fineLevel.GetLevelID(), pL);
-        std::ofstream pvtu(pvtuFilename.c_str());
-        this->writePVTU(pvtu, baseFname, comm->getSize(), pL.get<bool>("visualization: fine graph edges"));
-        pvtu.close();
-      }
+  if (pL.get<int>("visualization: start level") <= fineLevel.GetLevelID()) {
+    // write out coarsening information
+    std::string filenameToWrite = this->getFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
+    std::ofstream fout(filenameToWrite.c_str());
+
+    std::vector<int> uniqueFine = this->makeUnique(vertices);
+    this->writeFileVTKOpening(fout, uniqueFine, geomSize);
+    this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
+    this->writeFileVTKData(fout, uniqueFine, myAggOffset, vertex2AggId, comm->getRank());
+    this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coords->getNumVectors());
+    this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
+    this->writeFileVTKClosing(fout);
+    fout.close();
+
+    // create pvtu file
+    if (comm->getRank() == 0) {
+      std::string pvtuFilename = this->getPVTUFileName(comm->getSize(), comm->getRank(), fineLevel.GetLevelID(), pL);
+      std::string baseFname    = this->getBaseFileName(comm->getSize(), fineLevel.GetLevelID(), pL);
+      std::ofstream pvtu(pvtuFilename.c_str());
+      this->writePVTU(pvtu, baseFname, comm->getSize(), pL.get<bool>("visualization: fine graph edges"));
+      pvtu.close();
     }
+  }
 
-    if(comm->getRank() == 0 && pL.get<bool>("visualization: build colormap")) {
-      this->buildColormap();
-    }
+  if (comm->getRank() == 0 && pL.get<bool>("visualization: build colormap")) {
+    this->buildColormap();
   }
-} // namespace MueLu
+}
+}  // namespace MueLu
 
 #endif /* MUELU_AGGREGATIONEXPORTFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_decl.hpp
index c63bc81383e2..da48ac05e1f9 100644
--- a/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_decl.hpp
@@ -58,83 +58,80 @@
 
 namespace MueLu {
 
-  /*!
-    @class CoupledRBMFactory
-    @ingroup MueLuTransferClasses
-    @brief Nullspace Factory for coupled acoustic-elastic problems.
-
-     Combines standard nullspace with rigid body modes.
-     Assumes that acoustic pressure DOFs are padded with 2 extra DOFs
-     (so that there are 3 DOFs at each mesh grid point)
-  */
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class CoupledRBMFactory : public SingleLevelFactoryBase {
+/*!
+  @class CoupledRBMFactory
+  @ingroup MueLuTransferClasses
+  @brief Nullspace Factory for coupled acoustic-elastic problems.
+
+   Combines standard nullspace with rigid body modes.
+   Assumes that acoustic pressure DOFs are padded with 2 extra DOFs
+   (so that there are 3 DOFs at each mesh grid point)
+*/
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class CoupledRBMFactory : public SingleLevelFactoryBase {
 #undef MUELU_COUPLEDRBMFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-    //! Constructor
-    CoupledRBMFactory(const int numPDEs)
-      : nspName_("Nullspace"),
-        numPDEs_(numPDEs)
-    { }
-    //! Constructor
-    CoupledRBMFactory(const std::string & nspName = "Nullspace")
-      : nspName_(nspName),
-        numPDEs_(3)
-    { }
-
-    //! Destructor.
-    virtual ~CoupledRBMFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+  //! Constructor
+  CoupledRBMFactory(const int numPDEs)
+    : nspName_("Nullspace")
+    , numPDEs_(numPDEs) {}
+  //! Constructor
+  CoupledRBMFactory(const std::string& nspName = "Nullspace")
+    : nspName_(nspName)
+    , numPDEs_(3) {}
 
-    //@}
+  //! Destructor.
+  virtual ~CoupledRBMFactory();
 
-    //! @name Input
-    //@{
+  //@}
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //! @name Input
+  //@{
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level& currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    void BuildRBM(RCP<Matrix>& A, RCP<MultiVector>& Coords, RCP<MultiVector>& nullspace) const;
+  //! Build an object with this factory.
+  void Build(Level& currentLevel) const;
 
-    //@}
-    void setNumPDEs(int numPDEs) {
-      numPDEs_=numPDEs;
-    }
+  void BuildRBM(RCP<Matrix>& A, RCP<MultiVector>& Coords, RCP<MultiVector>& nullspace) const;
 
-    void setLastAcousticDOF(int lastDOF) {
-      lastAcousticDOF_=lastDOF;
-    }
+  //@}
+  void setNumPDEs(int numPDEs) {
+    numPDEs_ = numPDEs;
+  }
 
-  private:
+  void setLastAcousticDOF(int lastDOF) {
+    lastAcousticDOF_ = lastDOF;
+  }
 
-    //! name of nullspace vector on finest level
-    std::string nspName_;
+ private:
+  //! name of nullspace vector on finest level
+  std::string nspName_;
 
-    int numPDEs_;
+  int numPDEs_;
 
-    int lastAcousticDOF_;
+  int lastAcousticDOF_;
 
-  }; // class CoupledRBMFactory
+};  // class CoupledRBMFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COUPLEDRBMFACTORY_SHORT
-#endif // MUELU_COUPLEDRBMFACTORY_DECL_HPP
+#endif  // MUELU_COUPLEDRBMFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_def.hpp b/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_def.hpp
index 1bf2a71d93e6..110631b3fb4e 100644
--- a/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_CoupledRBMFactory_def.hpp
@@ -56,159 +56,155 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~CoupledRBMFactory() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~CoupledRBMFactory() {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    if (currentLevel.IsAvailable(nspName_, NoFactory::get()) == false && currentLevel.GetLevelID() == 0) {
-      Input(currentLevel, "A");
-      //Input(currentLevel,"Coordinates");
-    }
-    if (currentLevel.GetLevelID() !=0) {
-      currentLevel.DeclareInput("Nullspace", GetFactory(nspName_).get(), this); /* ! "Nullspace" and nspName_ mismatch possible here */
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level& currentLevel) const {
+  if (currentLevel.IsAvailable(nspName_, NoFactory::get()) == false && currentLevel.GetLevelID() == 0) {
+    Input(currentLevel, "A");
+    // Input(currentLevel,"Coordinates");
   }
+  if (currentLevel.GetLevelID() != 0) {
+    currentLevel.DeclareInput("Nullspace", GetFactory(nspName_).get(), this); /* ! "Nullspace" and nspName_ mismatch possible here */
+  }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Structural acoustics nullspace factory", currentLevel);
-    RCP<MultiVector> nullspace;
-    if (currentLevel.GetLevelID() == 0) {
-      if (currentLevel.IsAvailable(nspName_, NoFactory::get())) {
-        nullspace = currentLevel.Get< RCP<MultiVector> >(nspName_, NoFactory::get());
-        GetOStream(Runtime1) << "Use user-given rigid body modes " << nspName_ << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
-      }
-      else {
-        RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-        RCP<MultiVector> Coords = Get< RCP<MultiVector> >(currentLevel,"Coordinates");
-        GetOStream(Runtime1) << "Generating nullspace for structural acoustics: dimension = " << numPDEs_ << std::endl;
-        RCP<const Map> xmap=A->getDomainMap();
-        nullspace = MultiVectorFactory::Build(xmap, 6);
-        Scalar zero = (Scalar) 0.0;
-        nullspace -> putScalar(zero);
-        ArrayRCP<Scalar> xnodes, ynodes, znodes;
-        Scalar cx, cy, cz;
-        ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
-        int nDOFs=xmap->getLocalNumElements();
-        xnodes = Coords->getDataNonConst(0);
-        ynodes = Coords->getDataNonConst(1);
-        znodes = Coords->getDataNonConst(2);
-        cx = Coords->getVector(0)->meanValue();
-        cy = Coords->getVector(1)->meanValue();
-        cz = Coords->getVector(2)->meanValue();
-        nsValues0 = nullspace->getDataNonConst(0);
-        nsValues1 = nullspace->getDataNonConst(1);
-        nsValues2 = nullspace->getDataNonConst(2);
-        nsValues3 = nullspace->getDataNonConst(3);
-        nsValues4 = nullspace->getDataNonConst(4);
-        nsValues5 = nullspace->getDataNonConst(5);
-        for (int j=0; j<nDOFs; j+=numPDEs_) {
-          Scalar one = (Scalar) 1.0;
-          if( xmap->getGlobalElement(j) >= lastAcousticDOF_ ) {
-            Scalar xdiff = xnodes[j]-cx;
-            Scalar ydiff = ynodes[j]-cy;
-            Scalar zdiff = znodes[j]-cz;
-            // translation
-            nsValues0[j+0] = one;
-            nsValues1[j+1] = one;
-            nsValues2[j+2] = one;
-            // rotate around z-axis (x-y plane)
-            nsValues3[j+0] = -ydiff;
-            nsValues3[j+1] =  xdiff;
-            // rotate around x-axis (y-z plane)
-            nsValues4[j+1] = -zdiff;
-            nsValues4[j+2] =  ydiff;
-            // rotate around y-axis (x-z plane)
-            nsValues5[j+0] =  zdiff;
-            nsValues5[j+2] = -xdiff;
-          }
-          else {
-            // translation
-            nsValues0[j+0] = one;
-            // insert random values and keep the top row for this node empty
-            nsValues1[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues1[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues2[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues2[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues3[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues3[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues4[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues4[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues5[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues5[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-          }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level& currentLevel) const {
+  FactoryMonitor m(*this, "Structural acoustics nullspace factory", currentLevel);
+  RCP<MultiVector> nullspace;
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable(nspName_, NoFactory::get())) {
+      nullspace = currentLevel.Get<RCP<MultiVector> >(nspName_, NoFactory::get());
+      GetOStream(Runtime1) << "Use user-given rigid body modes " << nspName_ << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
+    } else {
+      RCP<Matrix> A           = Get<RCP<Matrix> >(currentLevel, "A");
+      RCP<MultiVector> Coords = Get<RCP<MultiVector> >(currentLevel, "Coordinates");
+      GetOStream(Runtime1) << "Generating nullspace for structural acoustics: dimension = " << numPDEs_ << std::endl;
+      RCP<const Map> xmap = A->getDomainMap();
+      nullspace           = MultiVectorFactory::Build(xmap, 6);
+      Scalar zero         = (Scalar)0.0;
+      nullspace->putScalar(zero);
+      ArrayRCP<Scalar> xnodes, ynodes, znodes;
+      Scalar cx, cy, cz;
+      ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
+      int nDOFs = xmap->getLocalNumElements();
+      xnodes    = Coords->getDataNonConst(0);
+      ynodes    = Coords->getDataNonConst(1);
+      znodes    = Coords->getDataNonConst(2);
+      cx        = Coords->getVector(0)->meanValue();
+      cy        = Coords->getVector(1)->meanValue();
+      cz        = Coords->getVector(2)->meanValue();
+      nsValues0 = nullspace->getDataNonConst(0);
+      nsValues1 = nullspace->getDataNonConst(1);
+      nsValues2 = nullspace->getDataNonConst(2);
+      nsValues3 = nullspace->getDataNonConst(3);
+      nsValues4 = nullspace->getDataNonConst(4);
+      nsValues5 = nullspace->getDataNonConst(5);
+      for (int j = 0; j < nDOFs; j += numPDEs_) {
+        Scalar one = (Scalar)1.0;
+        if (xmap->getGlobalElement(j) >= lastAcousticDOF_) {
+          Scalar xdiff = xnodes[j] - cx;
+          Scalar ydiff = ynodes[j] - cy;
+          Scalar zdiff = znodes[j] - cz;
+          // translation
+          nsValues0[j + 0] = one;
+          nsValues1[j + 1] = one;
+          nsValues2[j + 2] = one;
+          // rotate around z-axis (x-y plane)
+          nsValues3[j + 0] = -ydiff;
+          nsValues3[j + 1] = xdiff;
+          // rotate around x-axis (y-z plane)
+          nsValues4[j + 1] = -zdiff;
+          nsValues4[j + 2] = ydiff;
+          // rotate around y-axis (x-z plane)
+          nsValues5[j + 0] = zdiff;
+          nsValues5[j + 2] = -xdiff;
+        } else {
+          // translation
+          nsValues0[j + 0] = one;
+          // insert random values and keep the top row for this node empty
+          nsValues1[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues1[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues2[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues2[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues3[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues3[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues4[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues4[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues5[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+          nsValues5[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
         }
-      } // end if "Nullspace" not available
-    }
-    else {
-      nullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory(nspName_).get());
-    }
-    Set(currentLevel, "Nullspace", nullspace);
+      }
+    }  // end if "Nullspace" not available
+  } else {
+    nullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", GetFactory(nspName_).get());
   }
+  Set(currentLevel, "Nullspace", nullspace);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildRBM(RCP<Matrix>& A, RCP<MultiVector>& Coords, RCP<MultiVector>& nullspace) const {
-        GetOStream(Runtime1) << "Generating nullspace for structural acoustics: dimension = " << numPDEs_ << std::endl;
-        RCP<const Map> xmap=A->getDomainMap();
-        nullspace = MultiVectorFactory::Build(xmap, 6);
-        Scalar zero = (Scalar) 0.0;
-        nullspace -> putScalar(zero);
-        ArrayRCP<Scalar> xnodes, ynodes, znodes;
-        Scalar cx, cy, cz;
-        ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
-        int nDOFs=xmap->getLocalNumElements();
-        xnodes = Coords->getDataNonConst(0);
-        ynodes = Coords->getDataNonConst(1);
-        znodes = Coords->getDataNonConst(2);
-        cx = Coords->getVector(0)->meanValue();
-        cy = Coords->getVector(1)->meanValue();
-        cz = Coords->getVector(2)->meanValue();
-        nsValues0 = nullspace->getDataNonConst(0);
-        nsValues1 = nullspace->getDataNonConst(1);
-        nsValues2 = nullspace->getDataNonConst(2);
-        nsValues3 = nullspace->getDataNonConst(3);
-        nsValues4 = nullspace->getDataNonConst(4);
-        nsValues5 = nullspace->getDataNonConst(5);
-        for (int j=0; j<nDOFs; j+=numPDEs_) {
-          Scalar one = (Scalar) 1.0;
-          if( xmap->getGlobalElement(j) >= lastAcousticDOF_ ) {
-            Scalar xdiff = xnodes[j]-cx;
-            Scalar ydiff = ynodes[j]-cy;
-            Scalar zdiff = znodes[j]-cz;
-            // translation
-            nsValues0[j+0] = one;
-            nsValues1[j+1] = one;
-            nsValues2[j+2] = one;
-            // rotate around z-axis (x-y plane)
-            nsValues3[j+0] = -ydiff;
-            nsValues3[j+1] =  xdiff;
-            // rotate around x-axis (y-z plane)
-            nsValues4[j+1] = -zdiff;
-            nsValues4[j+2] =  ydiff;
-            // rotate around y-axis (x-z plane)
-            nsValues5[j+0] =  zdiff;
-            nsValues5[j+2] = -xdiff;
-          }
-          else {
-            // translation
-            nsValues0[j+0] = one;
-            // insert random values and keep the top row for this node empty
-            nsValues1[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues1[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues2[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues2[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues3[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues3[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues4[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues4[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues5[j+1] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-            nsValues5[j+2] = (Scalar) (((double) rand()) / ((double) RAND_MAX));
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildRBM(RCP<Matrix>& A, RCP<MultiVector>& Coords, RCP<MultiVector>& nullspace) const {
+  GetOStream(Runtime1) << "Generating nullspace for structural acoustics: dimension = " << numPDEs_ << std::endl;
+  RCP<const Map> xmap = A->getDomainMap();
+  nullspace           = MultiVectorFactory::Build(xmap, 6);
+  Scalar zero         = (Scalar)0.0;
+  nullspace->putScalar(zero);
+  ArrayRCP<Scalar> xnodes, ynodes, znodes;
+  Scalar cx, cy, cz;
+  ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
+  int nDOFs = xmap->getLocalNumElements();
+  xnodes    = Coords->getDataNonConst(0);
+  ynodes    = Coords->getDataNonConst(1);
+  znodes    = Coords->getDataNonConst(2);
+  cx        = Coords->getVector(0)->meanValue();
+  cy        = Coords->getVector(1)->meanValue();
+  cz        = Coords->getVector(2)->meanValue();
+  nsValues0 = nullspace->getDataNonConst(0);
+  nsValues1 = nullspace->getDataNonConst(1);
+  nsValues2 = nullspace->getDataNonConst(2);
+  nsValues3 = nullspace->getDataNonConst(3);
+  nsValues4 = nullspace->getDataNonConst(4);
+  nsValues5 = nullspace->getDataNonConst(5);
+  for (int j = 0; j < nDOFs; j += numPDEs_) {
+    Scalar one = (Scalar)1.0;
+    if (xmap->getGlobalElement(j) >= lastAcousticDOF_) {
+      Scalar xdiff = xnodes[j] - cx;
+      Scalar ydiff = ynodes[j] - cy;
+      Scalar zdiff = znodes[j] - cz;
+      // translation
+      nsValues0[j + 0] = one;
+      nsValues1[j + 1] = one;
+      nsValues2[j + 2] = one;
+      // rotate around z-axis (x-y plane)
+      nsValues3[j + 0] = -ydiff;
+      nsValues3[j + 1] = xdiff;
+      // rotate around x-axis (y-z plane)
+      nsValues4[j + 1] = -zdiff;
+      nsValues4[j + 2] = ydiff;
+      // rotate around y-axis (x-z plane)
+      nsValues5[j + 0] = zdiff;
+      nsValues5[j + 2] = -xdiff;
+    } else {
+      // translation
+      nsValues0[j + 0] = one;
+      // insert random values and keep the top row for this node empty
+      nsValues1[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues1[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues2[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues2[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues3[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues3[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues4[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues4[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues5[j + 1] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
+      nsValues5[j + 2] = (Scalar)(((double)rand()) / ((double)RAND_MAX));
     }
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_COUPLEDRBMFACTORY_SHORT
-#endif // MUELU_COUPLEDRBMFACTORY_DEF_HPP
+#endif  // MUELU_COUPLEDRBMFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Utils/MueLu_DisableMultipleCallCheck.hpp b/packages/muelu/src/Utils/MueLu_DisableMultipleCallCheck.hpp
index 158053bb474c..1fc0b2eb837a 100644
--- a/packages/muelu/src/Utils/MueLu_DisableMultipleCallCheck.hpp
+++ b/packages/muelu/src/Utils/MueLu_DisableMultipleCallCheck.hpp
@@ -52,15 +52,17 @@
 
 namespace MueLu {
 
-  //! An exception safe way to call the method TwoLevelFactoryBase::DisableMultipleCallCheck
-  class DisableMultipleCallCheck {
-  public:
-    DisableMultipleCallCheck(const RCP<const TwoLevelFactoryBase> & fact) : fact_(fact) { fact_->DisableMultipleCallCheck(); }
-    ~DisableMultipleCallCheck() { fact_->EnableMultipleCallCheck(); }
-  private:
-    const RCP<const TwoLevelFactoryBase> fact_;
-  };
+//! An exception safe way to call the method TwoLevelFactoryBase::DisableMultipleCallCheck
+class DisableMultipleCallCheck {
+ public:
+  DisableMultipleCallCheck(const RCP<const TwoLevelFactoryBase>& fact)
+    : fact_(fact) { fact_->DisableMultipleCallCheck(); }
+  ~DisableMultipleCallCheck() { fact_->EnableMultipleCallCheck(); }
 
-} //namespace MueLu
+ private:
+  const RCP<const TwoLevelFactoryBase> fact_;
+};
 
-#endif // MUELU_DISABLEMULTIPLECALLCHECK_HPP
+}  // namespace MueLu
+
+#endif  // MUELU_DISABLEMULTIPLECALLCHECK_HPP
diff --git a/packages/muelu/src/Utils/MueLu_ETI_3arg.hpp b/packages/muelu/src/Utils/MueLu_ETI_3arg.hpp
index fecac84e9fae..3fc5b4945c2f 100644
--- a/packages/muelu/src/Utils/MueLu_ETI_3arg.hpp
+++ b/packages/muelu/src/Utils/MueLu_ETI_3arg.hpp
@@ -9,29 +9,29 @@
 #include <TpetraCore_ETIHelperMacros.h>
 TPETRA_ETI_MANGLING_TYPEDEFS()
 #if defined(HAVE_MUELU_EPETRA)
-# include <Epetra_config.h>
+#include <Epetra_config.h>
 #endif
 
-#if   (defined(HAVE_MUELU_EPETRA) &&  defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
+#if (defined(HAVE_MUELU_EPETRA) && defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
+// Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
 #elif (defined(HAVE_MUELU_EPETRA) && !defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
+// Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
 #endif
 
 // Epetra = on, Tpetra = on
 #if defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_LGN(MUELU_ETI_GROUP)
-# if((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+TPETRA_INSTANTIATE_LGN(MUELU_ETI_GROUP)
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
      (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  MUELU_ETI_GROUP(int,int,EpetraNode)
-# endif
+MUELU_ETI_GROUP(int, int, EpetraNode)
+#endif
 #endif
 
 // Epetra = off, Tpetra = on
 #if !defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_LGN(MUELU_ETI_GROUP)
+TPETRA_INSTANTIATE_LGN(MUELU_ETI_GROUP)
 #endif
 
-#endif //ifndef MUELU_ETI_3ARGUMENT_HPP
+#endif  // ifndef MUELU_ETI_3ARGUMENT_HPP
diff --git a/packages/muelu/src/Utils/MueLu_ETI_4arg.hpp b/packages/muelu/src/Utils/MueLu_ETI_4arg.hpp
index 5bc71432f311..19c7ac4badda 100644
--- a/packages/muelu/src/Utils/MueLu_ETI_4arg.hpp
+++ b/packages/muelu/src/Utils/MueLu_ETI_4arg.hpp
@@ -9,29 +9,29 @@
 #include <TpetraCore_ETIHelperMacros.h>
 TPETRA_ETI_MANGLING_TYPEDEFS()
 #if defined(HAVE_MUELU_EPETRA)
-# include <Epetra_config.h>
+#include <Epetra_config.h>
 #endif
 
-#if   (defined(HAVE_MUELU_EPETRA) &&  defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
+#if (defined(HAVE_MUELU_EPETRA) && defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
+// Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
 #elif (defined(HAVE_MUELU_EPETRA) && !defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
+// Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
 #endif
 
 // Epetra = on, Tpetra = on
 #if defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP)
-# if((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP)
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
      (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  MUELU_ETI_GROUP(double,int,int,EpetraNode)
-# endif
+MUELU_ETI_GROUP(double, int, int, EpetraNode)
+#endif
 #endif
 
 // Epetra = off, Tpetra = on
 #if !defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP)
+TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP)
 #endif
 
-#endif //ifndef MUELU_ETI_4ARGUMENT_HPP
+#endif  // ifndef MUELU_ETI_4ARGUMENT_HPP
diff --git a/packages/muelu/src/Utils/MueLu_ETI_4arg_Xpetra.hpp b/packages/muelu/src/Utils/MueLu_ETI_4arg_Xpetra.hpp
index 56ee323744b4..2f5245b84fbf 100644
--- a/packages/muelu/src/Utils/MueLu_ETI_4arg_Xpetra.hpp
+++ b/packages/muelu/src/Utils/MueLu_ETI_4arg_Xpetra.hpp
@@ -9,29 +9,29 @@
 #include <TpetraCore_ETIHelperMacros.h>
 TPETRA_ETI_MANGLING_TYPEDEFS()
 #if defined(HAVE_MUELU_EPETRA)
-# include <Epetra_config.h>
+#include <Epetra_config.h>
 #endif
 
-#if   (defined(HAVE_MUELU_EPETRA) &&  defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
+#if (defined(HAVE_MUELU_EPETRA) && defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT)))
+// Epetra is enabled with OpenMP node, but Tpetra is a) not enabled, or b) is not instantiated on OpenMP, or c) is not instantiated on OpenMP with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode EpetraNode;
 #elif (defined(HAVE_MUELU_EPETRA) && !defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT)))
-  // Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
-  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
+// Epetra is enabled with Serial node, but Tpetra is a) not enabled, or b) is not instantiated on Serial, or c) is not instantiated on Serial with <double,int,int>
+typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode EpetraNode;
 #endif
 
 // Epetra = on, Tpetra = on
 #if defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP_TPETRA)
-# if((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP_TPETRA)
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
      (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  MUELU_ETI_GROUP_EPETRA(double,int,int,EpetraNode)
-# endif
+MUELU_ETI_GROUP_EPETRA(double, int, int, EpetraNode)
+#endif
 #endif
 
 // Epetra = off, Tpetra = on
 #if !defined(HAVE_MUELU_EPETRA)
-  TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP_TPETRA)
+TPETRA_INSTANTIATE_SLGN_NO_ORDINAL_SCALAR(MUELU_ETI_GROUP_TPETRA)
 #endif
 
-#endif //ifndef MUELU_ETI_4ARGUMENT_HPP
+#endif  // ifndef MUELU_ETI_4ARGUMENT_HPP
diff --git a/packages/muelu/src/Utils/MueLu_Exceptions.cpp b/packages/muelu/src/Utils/MueLu_Exceptions.cpp
index 8019ba6d6413..bde1bfb3610d 100644
--- a/packages/muelu/src/Utils/MueLu_Exceptions.cpp
+++ b/packages/muelu/src/Utils/MueLu_Exceptions.cpp
@@ -49,21 +49,28 @@
 #include "MueLu_ConfigDefs.hpp"
 
 namespace MueLu {
-  namespace Exceptions {
+namespace Exceptions {
 
-      BadCast::BadCast(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+BadCast::BadCast(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      NotImplemented::NotImplemented(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+NotImplemented::NotImplemented(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      RuntimeError::RuntimeError(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+RuntimeError::RuntimeError(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      Overflow::Overflow(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+Overflow::Overflow(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      Incompatible::Incompatible(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+Incompatible::Incompatible(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      DependencyError::DependencyError(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) {}
+DependencyError::DependencyError(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-      InvalidArgument::InvalidArgument(const std::string& what_arg) : Teuchos::ExceptionBase(what_arg) { }
+InvalidArgument::InvalidArgument(const std::string& what_arg)
+  : Teuchos::ExceptionBase(what_arg) {}
 
-  }
-}
+}  // namespace Exceptions
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_Exceptions.hpp b/packages/muelu/src/Utils/MueLu_Exceptions.hpp
index 85f0908484c6..9d495166a12b 100644
--- a/packages/muelu/src/Utils/MueLu_Exceptions.hpp
+++ b/packages/muelu/src/Utils/MueLu_Exceptions.hpp
@@ -51,55 +51,54 @@
 #include "MueLu_ConfigDefs.hpp"
 
 namespace MueLu {
-  namespace Exceptions {
+namespace Exceptions {
 
-    //! Exception indicating invalid cast attempted
-    class BadCast : public Teuchos::ExceptionBase {
-    public:
-      BadCast(const std::string& what_arg);
-    };
+//! Exception indicating invalid cast attempted
+class BadCast : public Teuchos::ExceptionBase {
+ public:
+  BadCast(const std::string& what_arg);
+};
 
-    //! Exception throws when you call an unimplemented method of MueLu
-    /** Mainly use for development in progress. **/
-    class NotImplemented : public Teuchos::ExceptionBase {
-    public:
-      NotImplemented(const std::string& what_arg);
-    };
+//! Exception throws when you call an unimplemented method of MueLu
+/** Mainly use for development in progress. **/
+class NotImplemented : public Teuchos::ExceptionBase {
+ public:
+  NotImplemented(const std::string& what_arg);
+};
 
-    //! Exception throws to report errors in the internal logical of the program.
-    class RuntimeError : public Teuchos::ExceptionBase {
-    public:
-      RuntimeError(const std::string& what_arg);
-    };
+//! Exception throws to report errors in the internal logical of the program.
+class RuntimeError : public Teuchos::ExceptionBase {
+ public:
+  RuntimeError(const std::string& what_arg);
+};
 
-    //! Exception throws to report overflows.
-    class Overflow : public Teuchos::ExceptionBase {
-    public:
-      Overflow(const std::string& what_arg);
-    };
+//! Exception throws to report overflows.
+class Overflow : public Teuchos::ExceptionBase {
+ public:
+  Overflow(const std::string& what_arg);
+};
 
-    //! Exception throws to report incompatible objects (like maps).
-    class Incompatible : public Teuchos::ExceptionBase {
-    public:
-      Incompatible(const std::string& what_arg);
-    };
+//! Exception throws to report incompatible objects (like maps).
+class Incompatible : public Teuchos::ExceptionBase {
+ public:
+  Incompatible(const std::string& what_arg);
+};
 
-    //! Exception throws to report data dependency problems between factories.
-    class DependencyError : public Teuchos::ExceptionBase {
-    public:
-      DependencyError(const std::string& what_arg);
-    };
+//! Exception throws to report data dependency problems between factories.
+class DependencyError : public Teuchos::ExceptionBase {
+ public:
+  DependencyError(const std::string& what_arg);
+};
 
-    //! Exception throws to report invalid user entry
-    class InvalidArgument: public Teuchos::ExceptionBase {
-    public:
-      InvalidArgument(const std::string& what_arg);
-    };
+//! Exception throws to report invalid user entry
+class InvalidArgument : public Teuchos::ExceptionBase {
+ public:
+  InvalidArgument(const std::string& what_arg);
+};
 
-  }
-}
+}  // namespace Exceptions
+}  // namespace MueLu
 
-#define MUELU_TPETRA_ETI_EXCEPTION(cl,obj,go) TEUCHOS_TEST_FOR_EXCEPTION(1, ::MueLu::Exceptions::BadCast, "Problem in " #cl "! Cannot create new object " #obj " with GO=" #go ". MueLu has been compiled with Tpetra enabled bug GO!=" #go ". Please add TPETRA_INST_INT_INT to your configuration.");
+#define MUELU_TPETRA_ETI_EXCEPTION(cl, obj, go) TEUCHOS_TEST_FOR_EXCEPTION(1, ::MueLu::Exceptions::BadCast, "Problem in " #cl "! Cannot create new object " #obj " with GO=" #go ". MueLu has been compiled with Tpetra enabled bug GO!=" #go ". Please add TPETRA_INST_INT_INT to your configuration.");
 
-
-#endif //ifndef MUELU_EXCEPTIONS_HPP
+#endif  // ifndef MUELU_EXCEPTIONS_HPP
diff --git a/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_decl.hpp b/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_decl.hpp
index e047972b3f41..1c34ef61c307 100644
--- a/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_decl.hpp
@@ -19,56 +19,52 @@
 
 namespace MueLu {
 
-  //! @brief Local permutation strategy
-  /*!
-     This class permutes columns of a input matrix A. No inter-node permutations are allowed,
-     only permutations of columns that correspond to DOFs of the same node.
-    */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class LocalPermutationStrategy : public BaseClass {
+//! @brief Local permutation strategy
+/*!
+   This class permutes columns of a input matrix A. No inter-node permutations are allowed,
+   only permutations of columns that correspond to DOFs of the same node.
+  */
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class LocalPermutationStrategy : public BaseClass {
 #undef MUELU_LOCALPERMUTATIONSTRATEGY_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-
+ public:
   /*!
     @class LocalPermutationStrategy class.
     @brief Class which defines local permutations of matrix columns which correspond to DOFs of the same node.
     */
 
-    //! build permutation operators
-    /*!
-     *  The following variables produced
-     *  "A"     :      permuted and scaled A
-     *  "permA" :      permuted A without scaling
-     *  "permP" :      permutation opertor (should be identity)
-     *  "permQT":      transpose permutation operators
-     *  "permScaling": scaling operator
-     *
-     * \param A: input matrix (input)
-     * \param permRowMap: Dof row map permutation shall be restricted on (input)
-     * \param currentLevel: only for output of variables
-     * \param genFactory: const pointer to generating (calling) PermutationFactory // TODO avoid this, not very elegant. Decide which variables have to be generated, give them back per reference to the PermutationFactory.
-     */
-    void BuildPermutation(const Teuchos::RCP<Matrix> & A, const Teuchos::RCP<const Map> permRowMap, Level & currentLevel, const FactoryBase* genFactory) const;
-
-
-
-  private:
-
-    void BuildPermutations(size_t nDofsPerNode) const;
-
-    mutable std::vector<std::vector<int> > result_permvecs_;
-    mutable size_t permWidth_;
-
-    GlobalOrdinal getGlobalDofId(const Teuchos::RCP<Matrix> & A, LocalOrdinal localNodeId, LocalOrdinal localDof) const;
-    GlobalOrdinal globalDofId2globalNodeId( const Teuchos::RCP<Matrix> & A, GlobalOrdinal grid ) const;
-   };
-
-} // namespace MueLu
+  //! build permutation operators
+  /*!
+   *  The following variables produced
+   *  "A"     :      permuted and scaled A
+   *  "permA" :      permuted A without scaling
+   *  "permP" :      permutation opertor (should be identity)
+   *  "permQT":      transpose permutation operators
+   *  "permScaling": scaling operator
+   *
+   * \param A: input matrix (input)
+   * \param permRowMap: Dof row map permutation shall be restricted on (input)
+   * \param currentLevel: only for output of variables
+   * \param genFactory: const pointer to generating (calling) PermutationFactory // TODO avoid this, not very elegant. Decide which variables have to be generated, give them back per reference to the PermutationFactory.
+   */
+  void BuildPermutation(const Teuchos::RCP<Matrix>& A, const Teuchos::RCP<const Map> permRowMap, Level& currentLevel, const FactoryBase* genFactory) const;
+
+ private:
+  void BuildPermutations(size_t nDofsPerNode) const;
+
+  mutable std::vector<std::vector<int> > result_permvecs_;
+  mutable size_t permWidth_;
+
+  GlobalOrdinal getGlobalDofId(const Teuchos::RCP<Matrix>& A, LocalOrdinal localNodeId, LocalOrdinal localDof) const;
+  GlobalOrdinal globalDofId2globalNodeId(const Teuchos::RCP<Matrix>& A, GlobalOrdinal grid) const;
+};
+
+}  // namespace MueLu
 
 #define MUELU_LOCALPERMUTATIONSTRATEGY_SHORT
 
diff --git a/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_def.hpp b/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_def.hpp
index 69bfc802bf73..b517191d77a4 100644
--- a/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_LocalPermutationStrategy_def.hpp
@@ -23,236 +23,231 @@
 
 namespace MueLu {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildPermutations(size_t nDofsPerNode) const {
-
-    permWidth_ = nDofsPerNode;
-
-    result_permvecs_.clear();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildPermutations(size_t nDofsPerNode) const {
+  permWidth_ = nDofsPerNode;
+
+  result_permvecs_.clear();
+
+  // build permutation string
+  std::stringstream ss;
+  for (size_t t = 0; t < nDofsPerNode; t++)
+    ss << t;
+  std::string cs = ss.str();
+  // std::vector<std::string> result_perms;
+  do {
+    // result_perms.push_back(cs);
+
+    std::vector<int> newPerm(cs.length(), -1);
+    for (size_t len = 0; len < cs.length(); len++) {
+      newPerm[len] = Teuchos::as<int>(cs[len] - '0');
+    }
+    result_permvecs_.push_back(newPerm);
 
-    // build permutation string
-    std::stringstream ss;
-    for(size_t t = 0; t<nDofsPerNode; t++)
-      ss << t;
-    std::string cs = ss.str();
-    //std::vector<std::string> result_perms;
-    do {
-      //result_perms.push_back(cs);
+  } while (std::next_permutation(cs.begin(), cs.end()));
+}
 
-      std::vector<int> newPerm(cs.length(),-1);
-      for(size_t len=0; len<cs.length(); len++) {
-        newPerm[len] = Teuchos::as<int>(cs[len]-'0');
-      }
-      result_permvecs_.push_back(newPerm);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildPermutation(const Teuchos::RCP<Matrix>& A, const Teuchos::RCP<const Map> /* permRowMap */, Level& currentLevel, const FactoryBase* genFactory) const {
+  SC SC_ZERO = Teuchos::ScalarTraits<SC>::zero();
 
-    } while (std::next_permutation(cs.begin(),cs.end()));
+  size_t nDofsPerNode = 1;
+  if (A->IsView("stridedMaps")) {
+    Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
+    nDofsPerNode                              = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildPermutation(const Teuchos::RCP<Matrix> & A, const Teuchos::RCP<const Map> /* permRowMap */, Level & currentLevel, const FactoryBase* genFactory) const {
+  RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
 
-    SC SC_ZERO = Teuchos::ScalarTraits<SC>::zero();
+  //////////////////
+  std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> > RowColPairs;
 
-    size_t nDofsPerNode = 1;
-    if (A->IsView("stridedMaps")) {
-      Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
-      nDofsPerNode = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
-    }
+  // check whether we have to (re)build the permutation vector
+  if (permWidth_ != nDofsPerNode)
+    BuildPermutations(nDofsPerNode);
 
-    RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-
-    //////////////////
-    std::vector<std::pair<GlobalOrdinal,GlobalOrdinal> > RowColPairs;
-
-    // check whether we have to (re)build the permutation vector
-    if(permWidth_ != nDofsPerNode)
-      BuildPermutations(nDofsPerNode);
-
-    // declare local variables used inside loop
-    LocalOrdinal lonDofsPerNode = Teuchos::as<LocalOrdinal> (nDofsPerNode);
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar> vals;
-    Teuchos::SerialDenseMatrix<LocalOrdinal,Scalar> subBlockMatrix(nDofsPerNode, nDofsPerNode, true);
-    std::vector<GlobalOrdinal> growIds(nDofsPerNode);
-
-    // loop over local nodes
-    // TODO what about nOffset?
-    LocalOrdinal numLocalNodes = A->getRowMap()->getLocalNumElements()/nDofsPerNode;
-    for (LocalOrdinal node = 0; node < numLocalNodes; ++node) {
-
-      // zero out block matrix
-      subBlockMatrix.putScalar();
-
-      // loop over all DOFs in current node
-      // Note: were assuming constant number of Dofs per node here!
-      // TODO This is more complicated for variable dofs per node
-      for (LocalOrdinal lrdof = 0; lrdof < lonDofsPerNode; ++lrdof) {
-        GlobalOrdinal grow = getGlobalDofId(A, node, lrdof);
-        growIds[lrdof] = grow;
-
-        // extract local row information from matrix
-        A->getLocalRowView(A->getRowMap()->getLocalElement(grow), indices, vals);
-
-        // find column entry with max absolute value
-        typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-        MT maxVal = 0.0;
-        for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
-          if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
-            maxVal = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
-          }
-        }
+  // declare local variables used inside loop
+  LocalOrdinal lonDofsPerNode = Teuchos::as<LocalOrdinal>(nDofsPerNode);
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> vals;
+  Teuchos::SerialDenseMatrix<LocalOrdinal, Scalar> subBlockMatrix(nDofsPerNode, nDofsPerNode, true);
+  std::vector<GlobalOrdinal> growIds(nDofsPerNode);
 
-        GlobalOrdinal grnodeid = globalDofId2globalNodeId(A,grow);
-
-        for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
-          GlobalOrdinal gcol = A->getColMap()->getGlobalElement(indices[j]);
-          GlobalOrdinal gcnodeid = globalDofId2globalNodeId(A,gcol); // -> global node id
-          if (grnodeid == gcnodeid) {
-            if(maxVal != Teuchos::ScalarTraits<MT>::zero ()) {
-              subBlockMatrix(lrdof, gcol % nDofsPerNode) = vals[j]/maxVal;
-            } else
-            {
-              subBlockMatrix(lrdof, gcol % nDofsPerNode) = vals[j]; // there is a problem
-              std::cout << "maxVal never should be zero!!!!" << std::endl;
-            }
-          }
-        }
-      }
+  // loop over local nodes
+  // TODO what about nOffset?
+  LocalOrdinal numLocalNodes = A->getRowMap()->getLocalNumElements() / nDofsPerNode;
+  for (LocalOrdinal node = 0; node < numLocalNodes; ++node) {
+    // zero out block matrix
+    subBlockMatrix.putScalar();
 
-      // now we have the sub block matrix
-
-      // build permutation string
-      /*std::stringstream ss;
-      for(size_t t = 0; t<nDofsPerNode; t++)
-        ss << t;
-      std::string cs = ss.str();
-      std::vector<std::string> result_perms;
-      do {
-        result_perms.push_back(cs);
-        //std::cout << result_perms.back() << std::endl;
-      } while (std::next_permutation(cs.begin(),cs.end()));*/
-
-      std::vector<Scalar> performance_vector = std::vector<Scalar>(result_permvecs_.size());
-      for (size_t t = 0; t < result_permvecs_.size(); t++) {
-        std::vector<int> vv = result_permvecs_[t];
-        Scalar value = 1.0;
-        for(size_t j=0; j<vv.size(); j++) {
-          value = value * subBlockMatrix(j,vv[j]);
-        }
-        performance_vector[t] = value;
-      }
-      /*for(size_t t = 0; t < result_perms.size(); t++) {
-        std::string s = result_perms[t];
-        Scalar value = 1.0;
-        for(size_t len=0; len<s.length(); len++) {
-          int col = Teuchos::as<int>(s[len]-'0');
-          value = value * subBlockMatrix(len,col);
-        }
-        performance_vector[t] = value;
-      }*/
+    // loop over all DOFs in current node
+    // Note: were assuming constant number of Dofs per node here!
+    // TODO This is more complicated for variable dofs per node
+    for (LocalOrdinal lrdof = 0; lrdof < lonDofsPerNode; ++lrdof) {
+      GlobalOrdinal grow = getGlobalDofId(A, node, lrdof);
+      growIds[lrdof]     = grow;
 
-      // find permutation with maximum performance value
+      // extract local row information from matrix
+      A->getLocalRowView(A->getRowMap()->getLocalElement(grow), indices, vals);
+
+      // find column entry with max absolute value
       typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-      MT maxVal = Teuchos::ScalarTraits<MT>::zero();
-      size_t maxPerformancePermutationIdx = 0;
-      for (size_t j = 0; j < Teuchos::as<size_t>(performance_vector.size()); j++) {
-        if(Teuchos::ScalarTraits<Scalar>::magnitude(performance_vector[j]) > maxVal) {
-          maxVal = Teuchos::ScalarTraits<Scalar>::magnitude(performance_vector[j]);
-          maxPerformancePermutationIdx = j;
+      MT maxVal = 0.0;
+      for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
+        if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]) > maxVal) {
+          maxVal = Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]);
         }
       }
 
-      // build RowColPairs for best permutation
-      std::vector<int> bestPerformancePermutation = result_permvecs_[maxPerformancePermutationIdx];
-      for(size_t t = 0; t<nDofsPerNode; t++) {
-        RowColPairs.push_back(std::make_pair(growIds[t],growIds[bestPerformancePermutation[t]]));
+      GlobalOrdinal grnodeid = globalDofId2globalNodeId(A, grow);
+
+      for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
+        GlobalOrdinal gcol     = A->getColMap()->getGlobalElement(indices[j]);
+        GlobalOrdinal gcnodeid = globalDofId2globalNodeId(A, gcol);  // -> global node id
+        if (grnodeid == gcnodeid) {
+          if (maxVal != Teuchos::ScalarTraits<MT>::zero()) {
+            subBlockMatrix(lrdof, gcol % nDofsPerNode) = vals[j] / maxVal;
+          } else {
+            subBlockMatrix(lrdof, gcol % nDofsPerNode) = vals[j];  // there is a problem
+            std::cout << "maxVal never should be zero!!!!" << std::endl;
+          }
+        }
       }
+    }
 
-    } // end loop over local nodes
+    // now we have the sub block matrix
 
+    // build permutation string
+    /*std::stringstream ss;
+    for(size_t t = 0; t<nDofsPerNode; t++)
+      ss << t;
+    std::string cs = ss.str();
+    std::vector<std::string> result_perms;
+    do {
+      result_perms.push_back(cs);
+      //std::cout << result_perms.back() << std::endl;
+    } while (std::next_permutation(cs.begin(),cs.end()));*/
+
+    std::vector<Scalar> performance_vector = std::vector<Scalar>(result_permvecs_.size());
+    for (size_t t = 0; t < result_permvecs_.size(); t++) {
+      std::vector<int> vv = result_permvecs_[t];
+      Scalar value        = 1.0;
+      for (size_t j = 0; j < vv.size(); j++) {
+        value = value * subBlockMatrix(j, vv[j]);
+      }
+      performance_vector[t] = value;
+    }
+    /*for(size_t t = 0; t < result_perms.size(); t++) {
+      std::string s = result_perms[t];
+      Scalar value = 1.0;
+      for(size_t len=0; len<s.length(); len++) {
+        int col = Teuchos::as<int>(s[len]-'0');
+        value = value * subBlockMatrix(len,col);
+      }
+      performance_vector[t] = value;
+    }*/
 
-    // build Pperm and Qperm vectors
-    Teuchos::RCP<Vector> Pperm = VectorFactory::Build(A->getRowMap());
-    Teuchos::RCP<Vector> Qperm = VectorFactory::Build(A->getDomainMap());
+    // find permutation with maximum performance value
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+    MT maxVal                           = Teuchos::ScalarTraits<MT>::zero();
+    size_t maxPerformancePermutationIdx = 0;
+    for (size_t j = 0; j < Teuchos::as<size_t>(performance_vector.size()); j++) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(performance_vector[j]) > maxVal) {
+        maxVal                       = Teuchos::ScalarTraits<Scalar>::magnitude(performance_vector[j]);
+        maxPerformancePermutationIdx = j;
+      }
+    }
 
-    Pperm->putScalar(SC_ZERO);
-    Qperm->putScalar(SC_ZERO);
+    // build RowColPairs for best permutation
+    std::vector<int> bestPerformancePermutation = result_permvecs_[maxPerformancePermutationIdx];
+    for (size_t t = 0; t < nDofsPerNode; t++) {
+      RowColPairs.push_back(std::make_pair(growIds[t], growIds[bestPerformancePermutation[t]]));
+    }
 
-    Teuchos::ArrayRCP<Scalar> PpermData = Pperm->getDataNonConst(0);
-    Teuchos::ArrayRCP<Scalar> QpermData = Qperm->getDataNonConst(0);
+  }  // end loop over local nodes
 
-    typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = RowColPairs.begin();
-    while(p != RowColPairs.end() ) {
-      GlobalOrdinal ik = (*p).first;
-      GlobalOrdinal jk = (*p).second;
+  // build Pperm and Qperm vectors
+  Teuchos::RCP<Vector> Pperm = VectorFactory::Build(A->getRowMap());
+  Teuchos::RCP<Vector> Qperm = VectorFactory::Build(A->getDomainMap());
 
-      LocalOrdinal lik = A->getRowMap()->getLocalElement(ik);
-      LocalOrdinal ljk = A->getDomainMap()->getLocalElement(jk);
+  Pperm->putScalar(SC_ZERO);
+  Qperm->putScalar(SC_ZERO);
 
-      Pperm->replaceLocalValue(lik,ik);
-      Qperm->replaceLocalValue(ljk,ik);
+  Teuchos::ArrayRCP<Scalar> PpermData = Pperm->getDataNonConst(0);
+  Teuchos::ArrayRCP<Scalar> QpermData = Qperm->getDataNonConst(0);
 
-      p = RowColPairs.erase(p);
-    }
+  typename std::vector<std::pair<GlobalOrdinal, GlobalOrdinal> >::iterator p = RowColPairs.begin();
+  while (p != RowColPairs.end()) {
+    GlobalOrdinal ik = (*p).first;
+    GlobalOrdinal jk = (*p).second;
 
-    if(RowColPairs.size()>0) GetOStream(Warnings0) << "MueLu::LocalPermutationStrategy: There are Row/col pairs left!" << std::endl;
+    LocalOrdinal lik = A->getRowMap()->getLocalElement(ik);
+    LocalOrdinal ljk = A->getDomainMap()->getLocalElement(jk);
 
-    // Qperm should be fine
-    // build matrices
+    Pperm->replaceLocalValue(lik, ik);
+    Qperm->replaceLocalValue(ljk, ik);
 
-    // create new empty Matrix
-    Teuchos::RCP<CrsMatrixWrap> permPTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(),1));
-    Teuchos::RCP<CrsMatrixWrap> permQTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(),1));
+    p = RowColPairs.erase(p);
+  }
 
-    for(size_t row=0; row<A->getLocalNumRows(); row++) {
-      Teuchos::ArrayRCP<GlobalOrdinal> indoutP(1,Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(PpermData[row]))); // column idx for Perm^T
-      Teuchos::ArrayRCP<GlobalOrdinal> indoutQ(1,Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(QpermData[row]))); // column idx for Qperm
-      Teuchos::ArrayRCP<Scalar> valout(1,Teuchos::ScalarTraits<Scalar>::one());
-      permPTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutP.view(0,indoutP.size()), valout.view(0,valout.size()));
-      permQTmatrix->insertGlobalValues (A->getRowMap()->getGlobalElement(row), indoutQ.view(0,indoutQ.size()), valout.view(0,valout.size()));
-    }
+  if (RowColPairs.size() > 0) GetOStream(Warnings0) << "MueLu::LocalPermutationStrategy: There are Row/col pairs left!" << std::endl;
 
-    permPTmatrix->fillComplete();
-    permQTmatrix->fillComplete();
+  // Qperm should be fine
+  // build matrices
 
-    Teuchos::RCP<Matrix> permPmatrix = Utilities::Transpose(*permPTmatrix,true);
+  // create new empty Matrix
+  Teuchos::RCP<CrsMatrixWrap> permPTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(), 1));
+  Teuchos::RCP<CrsMatrixWrap> permQTmatrix = Teuchos::rcp(new CrsMatrixWrap(A->getRowMap(), 1));
 
-    /*for(size_t row=0; row<permPTmatrix->getLocalNumRows(); row++) {
-      if(permPTmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permPTmatrix is " << permPTmatrix->getNumEntriesInLocalRow(row) << std::endl;
-      if(permPmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permPmatrix is " << permPmatrix->getNumEntriesInLocalRow(row) << std::endl;
-      if(permQTmatrix->getNumEntriesInLocalRow(row) != 1)
-        GetOStream(Warnings0) <<"#entries in row " << row << " of permQmatrix is " << permQTmatrix->getNumEntriesInLocalRow(row) << std::endl;
-    }*/
+  for (size_t row = 0; row < A->getLocalNumRows(); row++) {
+    Teuchos::ArrayRCP<GlobalOrdinal> indoutP(1, Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(PpermData[row])));  // column idx for Perm^T
+    Teuchos::ArrayRCP<GlobalOrdinal> indoutQ(1, Teuchos::as<GO>(Teuchos::ScalarTraits<Scalar>::real(QpermData[row])));  // column idx for Qperm
+    Teuchos::ArrayRCP<Scalar> valout(1, Teuchos::ScalarTraits<Scalar>::one());
+    permPTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutP.view(0, indoutP.size()), valout.view(0, valout.size()));
+    permQTmatrix->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indoutQ.view(0, indoutQ.size()), valout.view(0, valout.size()));
+  }
 
-    // build permP * A * permQT
-    Teuchos::RCP<Matrix> ApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *permQTmatrix, false, GetOStream(Statistics2),true,true);
-    Teuchos::RCP<Matrix> permPApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*permPmatrix, false, *ApermQt, false, GetOStream(Statistics2),true,true);
-
-    /*
-    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mat", *A);
-    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permP.mat", *permPmatrix);
-    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permQt.mat", *permQTmatrix);
-    MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permPApermQt.mat", *permPApermQt);
-    */
-    // build scaling matrix
-    Teuchos::RCP<Vector> diagVec = VectorFactory::Build(permPApermQt->getRowMap(),true);
-    Teuchos::RCP<Vector> invDiagVec = VectorFactory::Build(permPApermQt->getRowMap(),true);
-    Teuchos::ArrayRCP< Scalar > invDiagVecData = invDiagVec->getDataNonConst(0);
-
-    LO lCntZeroDiagonals = 0;
-    permPApermQt->getLocalDiagCopy(*diagVec);
-    Teuchos::ArrayRCP< const Scalar > diagVecData = diagVec->getData(0);
-    for(size_t i = 0; i<diagVec->getMap()->getLocalNumElements(); ++i) {
-      if(diagVecData[i] != SC_ZERO)
-        invDiagVecData[i] = Teuchos::ScalarTraits<Scalar>::one()/diagVecData[i];
-      else {
-        invDiagVecData[i] = Teuchos::ScalarTraits<Scalar>::one();
-        lCntZeroDiagonals++;
-        //GetOStream(Statistics0) << "MueLu::LocalPermutationStrategy: found zero on diagonal in row " << i << std::endl;
-      }
+  permPTmatrix->fillComplete();
+  permQTmatrix->fillComplete();
+
+  Teuchos::RCP<Matrix> permPmatrix = Utilities::Transpose(*permPTmatrix, true);
+
+  /*for(size_t row=0; row<permPTmatrix->getLocalNumRows(); row++) {
+    if(permPTmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) <<"#entries in row " << row << " of permPTmatrix is " << permPTmatrix->getNumEntriesInLocalRow(row) << std::endl;
+    if(permPmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) <<"#entries in row " << row << " of permPmatrix is " << permPmatrix->getNumEntriesInLocalRow(row) << std::endl;
+    if(permQTmatrix->getNumEntriesInLocalRow(row) != 1)
+      GetOStream(Warnings0) <<"#entries in row " << row << " of permQmatrix is " << permQTmatrix->getNumEntriesInLocalRow(row) << std::endl;
+  }*/
+
+  // build permP * A * permQT
+  Teuchos::RCP<Matrix> ApermQt      = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *permQTmatrix, false, GetOStream(Statistics2), true, true);
+  Teuchos::RCP<Matrix> permPApermQt = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*permPmatrix, false, *ApermQt, false, GetOStream(Statistics2), true, true);
+
+  /*
+  MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mat", *A);
+  MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permP.mat", *permPmatrix);
+  MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permQt.mat", *permQTmatrix);
+  MueLu::Utils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("permPApermQt.mat", *permPApermQt);
+  */
+  // build scaling matrix
+  Teuchos::RCP<Vector> diagVec             = VectorFactory::Build(permPApermQt->getRowMap(), true);
+  Teuchos::RCP<Vector> invDiagVec          = VectorFactory::Build(permPApermQt->getRowMap(), true);
+  Teuchos::ArrayRCP<Scalar> invDiagVecData = invDiagVec->getDataNonConst(0);
+
+  LO lCntZeroDiagonals = 0;
+  permPApermQt->getLocalDiagCopy(*diagVec);
+  Teuchos::ArrayRCP<const Scalar> diagVecData = diagVec->getData(0);
+  for (size_t i = 0; i < diagVec->getMap()->getLocalNumElements(); ++i) {
+    if (diagVecData[i] != SC_ZERO)
+      invDiagVecData[i] = Teuchos::ScalarTraits<Scalar>::one() / diagVecData[i];
+    else {
+      invDiagVecData[i] = Teuchos::ScalarTraits<Scalar>::one();
+      lCntZeroDiagonals++;
+      // GetOStream(Statistics0) << "MueLu::LocalPermutationStrategy: found zero on diagonal in row " << i << std::endl;
     }
+  }
 
 #if 0
     GO gCntZeroDiagonals  = 0;
@@ -261,89 +256,88 @@ namespace MueLu {
     GetOStream(Statistics0) << "MueLu::LocalPermutationStrategy: found " << gCntZeroDiagonals << " zeros on diagonal" << std::endl;
 #endif
 
-    Teuchos::RCP<CrsMatrixWrap> diagScalingOp = Teuchos::rcp(new CrsMatrixWrap(permPApermQt->getRowMap(),1));
+  Teuchos::RCP<CrsMatrixWrap> diagScalingOp = Teuchos::rcp(new CrsMatrixWrap(permPApermQt->getRowMap(), 1));
 
-    for(size_t row=0; row<A->getLocalNumRows(); row++) {
-      Teuchos::ArrayRCP<GlobalOrdinal> indout(1,permPApermQt->getRowMap()->getGlobalElement(row)); // column idx for Perm^T
-      Teuchos::ArrayRCP<Scalar> valout(1,invDiagVecData[row]);
-      diagScalingOp->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indout.view(0,indout.size()), valout.view(0,valout.size()));
-    }
-    diagScalingOp->fillComplete();
-
-    Teuchos::RCP<Matrix> scaledA = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*diagScalingOp, false, *permPApermQt, false, GetOStream(Statistics2), true, true);
-    currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(scaledA), genFactory);
-
-    currentLevel.Set("permA", Teuchos::rcp_dynamic_cast<Matrix>(permPApermQt), genFactory);
-    currentLevel.Set("permP", Teuchos::rcp_dynamic_cast<Matrix>(permPmatrix), genFactory);
-    currentLevel.Set("permQT", Teuchos::rcp_dynamic_cast<Matrix>(permQTmatrix), genFactory);
-    currentLevel.Set("permScaling", Teuchos::rcp_dynamic_cast<Matrix>(diagScalingOp), genFactory);
-
-    //// count row permutations
-    // count zeros on diagonal in P -> number of row permutations
-    Teuchos::RCP<Vector> diagPVec = VectorFactory::Build(permPmatrix->getRowMap(),true);
-    permPmatrix->getLocalDiagCopy(*diagPVec);
-    Teuchos::ArrayRCP< const Scalar > diagPVecData = diagPVec->getData(0);
-    GlobalOrdinal lNumRowPermutations = 0;
-    GlobalOrdinal gNumRowPermutations = 0;
-    for(size_t i = 0; i<diagPVec->getMap()->getLocalNumElements(); ++i) {
-      if(diagPVecData[i] == SC_ZERO) {
-        lNumRowPermutations++;
-      }
+  for (size_t row = 0; row < A->getLocalNumRows(); row++) {
+    Teuchos::ArrayRCP<GlobalOrdinal> indout(1, permPApermQt->getRowMap()->getGlobalElement(row));  // column idx for Perm^T
+    Teuchos::ArrayRCP<Scalar> valout(1, invDiagVecData[row]);
+    diagScalingOp->insertGlobalValues(A->getRowMap()->getGlobalElement(row), indout.view(0, indout.size()), valout.view(0, valout.size()));
+  }
+  diagScalingOp->fillComplete();
+
+  Teuchos::RCP<Matrix> scaledA = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*diagScalingOp, false, *permPApermQt, false, GetOStream(Statistics2), true, true);
+  currentLevel.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(scaledA), genFactory);
+
+  currentLevel.Set("permA", Teuchos::rcp_dynamic_cast<Matrix>(permPApermQt), genFactory);
+  currentLevel.Set("permP", Teuchos::rcp_dynamic_cast<Matrix>(permPmatrix), genFactory);
+  currentLevel.Set("permQT", Teuchos::rcp_dynamic_cast<Matrix>(permQTmatrix), genFactory);
+  currentLevel.Set("permScaling", Teuchos::rcp_dynamic_cast<Matrix>(diagScalingOp), genFactory);
+
+  //// count row permutations
+  // count zeros on diagonal in P -> number of row permutations
+  Teuchos::RCP<Vector> diagPVec = VectorFactory::Build(permPmatrix->getRowMap(), true);
+  permPmatrix->getLocalDiagCopy(*diagPVec);
+  Teuchos::ArrayRCP<const Scalar> diagPVecData = diagPVec->getData(0);
+  GlobalOrdinal lNumRowPermutations            = 0;
+  GlobalOrdinal gNumRowPermutations            = 0;
+  for (size_t i = 0; i < diagPVec->getMap()->getLocalNumElements(); ++i) {
+    if (diagPVecData[i] == SC_ZERO) {
+      lNumRowPermutations++;
     }
+  }
 
-    // sum up all entries in multipleColRequests over all processors
-    MueLu_sumAll(diagPVec->getMap()->getComm(), lNumRowPermutations, gNumRowPermutations);
-
-    //// count column permutations
-    // count zeros on diagonal in Q^T -> number of column permutations
-    Teuchos::RCP<Vector> diagQTVec = VectorFactory::Build(permQTmatrix->getRowMap(),true);
-    permQTmatrix->getLocalDiagCopy(*diagQTVec);
-    Teuchos::ArrayRCP< const Scalar > diagQTVecData = diagQTVec->getData(0);
-    GlobalOrdinal lNumColPermutations = 0;
-    GlobalOrdinal gNumColPermutations = 0;
-    for(size_t i = 0; i<diagQTVec->getMap()->getLocalNumElements(); ++i) {
-      if(diagQTVecData[i] == SC_ZERO) {
-        lNumColPermutations++;
-      }
+  // sum up all entries in multipleColRequests over all processors
+  MueLu_sumAll(diagPVec->getMap()->getComm(), lNumRowPermutations, gNumRowPermutations);
+
+  //// count column permutations
+  // count zeros on diagonal in Q^T -> number of column permutations
+  Teuchos::RCP<Vector> diagQTVec = VectorFactory::Build(permQTmatrix->getRowMap(), true);
+  permQTmatrix->getLocalDiagCopy(*diagQTVec);
+  Teuchos::ArrayRCP<const Scalar> diagQTVecData = diagQTVec->getData(0);
+  GlobalOrdinal lNumColPermutations             = 0;
+  GlobalOrdinal gNumColPermutations             = 0;
+  for (size_t i = 0; i < diagQTVec->getMap()->getLocalNumElements(); ++i) {
+    if (diagQTVecData[i] == SC_ZERO) {
+      lNumColPermutations++;
     }
-
-    // sum up all entries in multipleColRequests over all processors
-    MueLu_sumAll(diagQTVec->getMap()->getComm(), lNumColPermutations, gNumColPermutations);
-
-    currentLevel.Set("#RowPermutations", gNumRowPermutations, genFactory/*this*/);
-    currentLevel.Set("#ColPermutations", gNumColPermutations, genFactory/*this*/);
-    currentLevel.Set("#WideRangeRowPermutations", 0, genFactory/*this*/);
-    currentLevel.Set("#WideRangeColPermutations", 0, genFactory/*this*/);
-
-    GetOStream(Statistics0) << "#Row    permutations/max possible permutations: " << gNumRowPermutations << "/" << diagPVec->getMap()->getGlobalNumElements() << std::endl;
-    GetOStream(Statistics0) << "#Column permutations/max possible permutations: " << gNumColPermutations << "/" << diagQTVec->getMap()->getGlobalNumElements() << std::endl;
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getGlobalDofId(const Teuchos::RCP<Matrix> & A, LocalOrdinal localNodeId, LocalOrdinal localDof) const {
-    size_t nDofsPerNode = 1;
-    if (A->IsView("stridedMaps")) {
-      Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
-      nDofsPerNode = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
-    }
+  // sum up all entries in multipleColRequests over all processors
+  MueLu_sumAll(diagQTVec->getMap()->getComm(), lNumColPermutations, gNumColPermutations);
 
-    LocalOrdinal localDofId = localNodeId * nDofsPerNode + localDof;
+  currentLevel.Set("#RowPermutations", gNumRowPermutations, genFactory /*this*/);
+  currentLevel.Set("#ColPermutations", gNumColPermutations, genFactory /*this*/);
+  currentLevel.Set("#WideRangeRowPermutations", 0, genFactory /*this*/);
+  currentLevel.Set("#WideRangeColPermutations", 0, genFactory /*this*/);
 
-    return A->getRowMap()->getGlobalElement(localDofId);
+  GetOStream(Statistics0) << "#Row    permutations/max possible permutations: " << gNumRowPermutations << "/" << diagPVec->getMap()->getGlobalNumElements() << std::endl;
+  GetOStream(Statistics0) << "#Column permutations/max possible permutations: " << gNumColPermutations << "/" << diagQTVec->getMap()->getGlobalNumElements() << std::endl;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getGlobalDofId(const Teuchos::RCP<Matrix>& A, LocalOrdinal localNodeId, LocalOrdinal localDof) const {
+  size_t nDofsPerNode = 1;
+  if (A->IsView("stridedMaps")) {
+    Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
+    nDofsPerNode                              = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  GlobalOrdinal LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::globalDofId2globalNodeId( const Teuchos::RCP<Matrix> & A, GlobalOrdinal grid ) const {
-    size_t nDofsPerNode = 1;
-    if (A->IsView("stridedMaps")) {
-      Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
-      nDofsPerNode = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
-    }
+  LocalOrdinal localDofId = localNodeId * nDofsPerNode + localDof;
+
+  return A->getRowMap()->getGlobalElement(localDofId);
+}
 
-    return (GlobalOrdinal) grid / (GlobalOrdinal)nDofsPerNode; // TODO what about nOffset???
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+GlobalOrdinal LocalPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node>::globalDofId2globalNodeId(const Teuchos::RCP<Matrix>& A, GlobalOrdinal grid) const {
+  size_t nDofsPerNode = 1;
+  if (A->IsView("stridedMaps")) {
+    Teuchos::RCP<const Map> permRowMapStrided = A->getRowMap("stridedMaps");
+    nDofsPerNode                              = Teuchos::rcp_dynamic_cast<const StridedMap>(permRowMapStrided)->getFixedBlockSize();
   }
 
-} // namespace MueLu
+  return (GlobalOrdinal)grid / (GlobalOrdinal)nDofsPerNode;  // TODO what about nOffset???
+}
 
+}  // namespace MueLu
 
 #endif /* MUELU_LOCALPERMUTATIONSTRATEGY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_decl.hpp
index bc7faa837dae..95dad8e9c313 100644
--- a/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_decl.hpp
@@ -59,84 +59,83 @@
 
 namespace MueLu {
 
-  /*!
-    @class MatrixAnalysisFactory
-    @brief This class checks matrix properties of A on current level.
-    This factory can be plugged in everywhere
-    in the factory layout of the MueLu preconditioner. It only needs A as input and pipes it through as output.
+/*!
+  @class MatrixAnalysisFactory
+  @brief This class checks matrix properties of A on current level.
+  This factory can be plugged in everywhere
+  in the factory layout of the MueLu preconditioner. It only needs A as input and pipes it through as output.
 
-    ## Input/output of MatrixAnalysisFactory ##
+  ## Input/output of MatrixAnalysisFactory ##
 
-    ### User parameters of MatrixAnalysisFactory ###
-    Parameter | type | default | master.xml | validated | requested | description
-    ----------|------|---------|:----------:|:---------:|:---------:|------------
-    | A       | Factory | null  |   | * | * | Generating factory of the matrix A|
+  ### User parameters of MatrixAnalysisFactory ###
+  Parameter | type | default | master.xml | validated | requested | description
+  ----------|------|---------|:----------:|:---------:|:---------:|------------
+  | A       | Factory | null  |   | * | * | Generating factory of the matrix A|
 
-    The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
-    The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see MatrixAnalysisFactory::GetValidParameters).<br>
-    The * in the @c requested column states that the data is requested as input with all dependencies (see MatrixAnalysisFactory::DeclareInput).
+  The * in the @c master.xml column denotes that the parameter is defined in the @c master.xml file.<br>
+  The * in the @c validated column means that the parameter is declared in the list of valid input parameters (see MatrixAnalysisFactory::GetValidParameters).<br>
+  The * in the @c requested column states that the data is requested as input with all dependencies (see MatrixAnalysisFactory::DeclareInput).
 
-    ### Variables provided by MatrixAnalysisFactory ###
+  ### Variables provided by MatrixAnalysisFactory ###
 
-    After MatrixAnalysisFactory::Build the following data is available (if requested)
+  After MatrixAnalysisFactory::Build the following data is available (if requested)
 
-    Parameter | generated by | description
-    ----------|--------------|------------
-    | A       | MatrixAnalysisFactory   | Original operator (without changes)
+  Parameter | generated by | description
+  ----------|--------------|------------
+  | A       | MatrixAnalysisFactory   | Original operator (without changes)
 
 
-  */
+*/
 
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class MatrixAnalysisFactory : public TwoLevelFactoryBase {
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class MatrixAnalysisFactory : public TwoLevelFactoryBase {
 #undef MUELU_MATRIXANALYSISFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor.
-    MatrixAnalysisFactory();
+  //! Constructor.
+  MatrixAnalysisFactory();
 
-    //! Destructor.
-    virtual ~MatrixAnalysisFactory();
+  //! Destructor.
+  virtual ~MatrixAnalysisFactory();
 
-    //@}
+  //@}
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-    //! @name Build methods.
-    //@{
+  //! @name Build methods.
+  //@{
 
-    //! Build an object with this factory.
-    void Build(Level &fineLevel, Level &coarseLevel) const;
+  //! Build an object with this factory.
+  void Build(Level &fineLevel, Level &coarseLevel) const;
 
-    //@}
+  //@}
 
-  private:
-    void checkVectorEntries(const Teuchos::RCP<Vector>& vec, std::string info) const;
+ private:
+  void checkVectorEntries(const Teuchos::RCP<Vector> &vec, std::string info) const;
 
-  }; // class MatrixAnalysisFactory
+};  // class MatrixAnalysisFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_MATRIXANALYSISFACTORY_SHORT
 
-
 #endif /* MUELU_MATRIXANALYSISFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_def.hpp b/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_def.hpp
index 213a5a73829a..371cc0ef2e57 100644
--- a/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_MatrixAnalysisFactory_def.hpp
@@ -50,7 +50,7 @@
 #include "MueLu_MatrixAnalysisFactory_decl.hpp"
 
 #include <Xpetra_Map.hpp>
-#include <Xpetra_StridedMap.hpp>    // for nDofsPerNode...
+#include <Xpetra_StridedMap.hpp>  // for nDofsPerNode...
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_VectorFactory.hpp>
 #include <Xpetra_Matrix.hpp>
@@ -61,8 +61,7 @@
 
 namespace MueLu {
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MatrixAnalysisFactory()
-  { }
+MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MatrixAnalysisFactory() {}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~MatrixAnalysisFactory() {}
@@ -71,24 +70,24 @@ template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const ParameterList> MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be permuted.");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be permuted.");
 
   return validParamList;
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level &/* coarseLevel */) const {
+void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &fineLevel, Level & /* coarseLevel */) const {
   Input(fineLevel, "A");
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level &/* coarseLevel */) const {
+void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &fineLevel, Level & /* coarseLevel */) const {
   FactoryMonitor m(*this, "MatrixAnalysis Factory ", fineLevel);
 
-  Teuchos::RCP<Matrix> A = Get< Teuchos::RCP<Matrix> > (fineLevel, "A");
+  Teuchos::RCP<Matrix> A                       = Get<Teuchos::RCP<Matrix> >(fineLevel, "A");
   Teuchos::RCP<const Teuchos::Comm<int> > comm = A->getRangeMap()->getComm();
 
-  //const ParameterList & pL = GetParameterList();
+  // const ParameterList & pL = GetParameterList();
 
   // General information
   {
@@ -97,10 +96,10 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     if (A->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps")) != Teuchos::null) {
       Teuchos::RCP<const StridedMap> strRowMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getRowMap("stridedMaps"));
-      LocalOrdinal blockid       = strRowMap->getStridedBlockId();
+      LocalOrdinal blockid                     = strRowMap->getStridedBlockId();
       if (blockid > -1) {
         std::vector<size_t> stridingInfo = strRowMap->getStridingData();
-        LO dofsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+        LO dofsPerNode                   = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
         GetOStream(Runtime0) << "Strided row: " << dofsPerNode << " dofs per node. Strided block id = " << blockid << std::endl;
       } else {
         GetOStream(Runtime0) << "Row: " << strRowMap->getFixedBlockSize() << " dofs per node" << std::endl;
@@ -109,27 +108,26 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     if (A->IsView("stridedMaps") && Teuchos::rcp_dynamic_cast<const StridedMap>(A->getColMap("stridedMaps")) != Teuchos::null) {
       Teuchos::RCP<const StridedMap> strDomMap = Teuchos::rcp_dynamic_cast<const StridedMap>(A->getColMap("stridedMaps"));
-      LocalOrdinal blockid       = strDomMap->getStridedBlockId();
+      LocalOrdinal blockid                     = strDomMap->getStridedBlockId();
       if (blockid > -1) {
         std::vector<size_t> stridingInfo = strDomMap->getStridingData();
-        LO dofsPerNode = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
+        LO dofsPerNode                   = Teuchos::as<LocalOrdinal>(stridingInfo[blockid]);
         GetOStream(Runtime0) << "Strided column: " << dofsPerNode << " dofs per node. Strided block id = " << blockid << std::endl;
       } else {
         GetOStream(Runtime0) << "Column: " << strDomMap->getFixedBlockSize() << " dofs per node" << std::endl;
       }
     }
 
-
     GetOStream(Runtime0) << "A is distributed over " << comm->getSize() << " processors" << std::endl;
 
     // empty processors
-    std::vector<LO> lelePerProc(comm->getSize(),0);
-    std::vector<LO> gelePerProc(comm->getSize(),0);
-    lelePerProc[comm->getRank()] = A->getLocalNumEntries ();
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_MAX,comm->getSize(),&lelePerProc[0],&gelePerProc[0]);
-    if(comm->getRank() == 0) {
-      for(int i=0; i<comm->getSize(); i++) {
-        if(gelePerProc[i] == 0) {
+    std::vector<LO> lelePerProc(comm->getSize(), 0);
+    std::vector<LO> gelePerProc(comm->getSize(), 0);
+    lelePerProc[comm->getRank()] = A->getLocalNumEntries();
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, comm->getSize(), &lelePerProc[0], &gelePerProc[0]);
+    if (comm->getRank() == 0) {
+      for (int i = 0; i < comm->getSize(); i++) {
+        if (gelePerProc[i] == 0) {
           GetOStream(Runtime0) << "Proc " << i << " is empty." << std::endl;
         }
       }
@@ -139,22 +137,22 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
   // Matrix diagonal
   {
     GetOStream(Runtime0) << "~~~~~~~~ MATRIX DIAGONAL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ (Level " << fineLevel.GetLevelID() << ")" << std::endl;
-    Teuchos::RCP<Vector> diagAVec = VectorFactory::Build(A->getRowMap(),true);
+    Teuchos::RCP<Vector> diagAVec = VectorFactory::Build(A->getRowMap(), true);
     A->getLocalDiagCopy(*diagAVec);
-    Teuchos::ArrayRCP< const Scalar > diagAVecData = diagAVec->getData(0);
-    GlobalOrdinal lzerosOnDiagonal = 0;
-    GlobalOrdinal gzerosOnDiagonal = 0;
-    GlobalOrdinal lnearlyzerosOnDiagonal = 0;
-    GlobalOrdinal gnearlyzerosOnDiagonal = 0;
-    GlobalOrdinal lnansOnDiagonal = 0;
-    GlobalOrdinal gnansOnDiagonal = 0;
-
-    for(size_t i = 0; i<diagAVec->getMap()->getLocalNumElements(); ++i) {
-      if(diagAVecData[i] == Teuchos::ScalarTraits<Scalar>::zero()) {
+    Teuchos::ArrayRCP<const Scalar> diagAVecData = diagAVec->getData(0);
+    GlobalOrdinal lzerosOnDiagonal               = 0;
+    GlobalOrdinal gzerosOnDiagonal               = 0;
+    GlobalOrdinal lnearlyzerosOnDiagonal         = 0;
+    GlobalOrdinal gnearlyzerosOnDiagonal         = 0;
+    GlobalOrdinal lnansOnDiagonal                = 0;
+    GlobalOrdinal gnansOnDiagonal                = 0;
+
+    for (size_t i = 0; i < diagAVec->getMap()->getLocalNumElements(); ++i) {
+      if (diagAVecData[i] == Teuchos::ScalarTraits<Scalar>::zero()) {
         lzerosOnDiagonal++;
-      } else if(Teuchos::ScalarTraits<Scalar>::magnitude(diagAVecData[i]) < 1e-6) {
+      } else if (Teuchos::ScalarTraits<Scalar>::magnitude(diagAVecData[i]) < 1e-6) {
         lnearlyzerosOnDiagonal++;
-      } else if(Teuchos::ScalarTraits<Scalar>::isnaninf(diagAVecData[i])) {
+      } else if (Teuchos::ScalarTraits<Scalar>::isnaninf(diagAVecData[i])) {
         lnansOnDiagonal++;
       }
     }
@@ -164,9 +162,9 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
     MueLu_sumAll(comm, lnearlyzerosOnDiagonal, gnearlyzerosOnDiagonal);
     MueLu_sumAll(comm, lnansOnDiagonal, gnansOnDiagonal);
 
-    if(gzerosOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gzerosOnDiagonal << " zeros on diagonal of A" << std::endl;
-    if(gnearlyzerosOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gnearlyzerosOnDiagonal << " entries with absolute value < 1.0e-6 on diagonal of A" << std::endl;
-    if(gnansOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gnansOnDiagonal << " entries with NAN or INF on diagonal of A" << std::endl;
+    if (gzerosOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gzerosOnDiagonal << " zeros on diagonal of A" << std::endl;
+    if (gnearlyzerosOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gnearlyzerosOnDiagonal << " entries with absolute value < 1.0e-6 on diagonal of A" << std::endl;
+    if (gnansOnDiagonal > 0) GetOStream(Runtime0) << "Found " << gnansOnDiagonal << " entries with NAN or INF on diagonal of A" << std::endl;
   }
 
   // Diagonal dominance?
@@ -174,11 +172,11 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
     GetOStream(Runtime0) << "~~~~~~~~ DIAGONAL DOMINANCE ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ (Level " << fineLevel.GetLevelID() << ")" << std::endl;
     // loop over all local rows in matrix A and keep diagonal entries if corresponding
     // matrix rows are not contained in permRowMap
-    GlobalOrdinal lnumWeakDiagDomRows = 0;
-    GlobalOrdinal gnumWeakDiagDomRows = 0;
+    GlobalOrdinal lnumWeakDiagDomRows   = 0;
+    GlobalOrdinal gnumWeakDiagDomRows   = 0;
     GlobalOrdinal lnumStrictDiagDomRows = 0;
     GlobalOrdinal gnumStrictDiagDomRows = 0;
-    double worstRatio = 99999999.9;
+    double worstRatio                   = 99999999.9;
     for (size_t row = 0; row < A->getRowMap()->getLocalNumElements(); row++) {
       GlobalOrdinal grow = A->getRowMap()->getGlobalElement(row);
 
@@ -192,16 +190,20 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
       TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError, "MueLu::MatrixAnalysisFactory::Build: number of nonzeros not equal to number of indices? Error.");
 
       // find column entry with max absolute value
-      double norm1 = 0.0;
+      double norm1    = 0.0;
       double normdiag = 0.0;
       for (size_t j = 0; j < Teuchos::as<size_t>(indices.size()); j++) {
         GO gcol = A->getColMap()->getGlobalElement(indices[j]);
-        if (gcol==grow) normdiag = Teuchos::as<double>(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]));
-        else norm1 +=  Teuchos::as<double>(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]));
+        if (gcol == grow)
+          normdiag = Teuchos::as<double>(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]));
+        else
+          norm1 += Teuchos::as<double>(Teuchos::ScalarTraits<Scalar>::magnitude(vals[j]));
       }
 
-      if      (normdiag >= norm1) lnumWeakDiagDomRows++;
-      else if (normdiag >  norm1) lnumStrictDiagDomRows++;
+      if (normdiag >= norm1)
+        lnumWeakDiagDomRows++;
+      else if (normdiag > norm1)
+        lnumStrictDiagDomRows++;
 
       if (norm1 != 0.0) {
         if (normdiag / norm1 < worstRatio) worstRatio = normdiag / norm1;
@@ -211,15 +213,14 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
     MueLu_sumAll(comm, lnumWeakDiagDomRows, gnumWeakDiagDomRows);
     MueLu_sumAll(comm, lnumStrictDiagDomRows, gnumStrictDiagDomRows);
 
-    GetOStream(Runtime0) << "A has " << gnumWeakDiagDomRows << "/" << A->getRangeMap()->getGlobalNumElements() << " weakly diagonal dominant rows. (" << Teuchos::as<Scalar>(gnumWeakDiagDomRows/A->getRangeMap()->getGlobalNumElements()*100) << "%)" << std::endl;
-    GetOStream(Runtime0) << "A has " << gnumStrictDiagDomRows << "/" << A->getRangeMap()->getGlobalNumElements() << " strictly diagonal dominant rows. (" << Teuchos::as<Scalar>(gnumStrictDiagDomRows/A->getRangeMap()->getGlobalNumElements()*100) << "%)" << std::endl;
+    GetOStream(Runtime0) << "A has " << gnumWeakDiagDomRows << "/" << A->getRangeMap()->getGlobalNumElements() << " weakly diagonal dominant rows. (" << Teuchos::as<Scalar>(gnumWeakDiagDomRows / A->getRangeMap()->getGlobalNumElements() * 100) << "%)" << std::endl;
+    GetOStream(Runtime0) << "A has " << gnumStrictDiagDomRows << "/" << A->getRangeMap()->getGlobalNumElements() << " strictly diagonal dominant rows. (" << Teuchos::as<Scalar>(gnumStrictDiagDomRows / A->getRangeMap()->getGlobalNumElements() * 100) << "%)" << std::endl;
 
     double gworstRatio;
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_MIN,comm->getSize(),&worstRatio,&gworstRatio);
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MIN, comm->getSize(), &worstRatio, &gworstRatio);
     GetOStream(Runtime0) << "The minimum of the ratio of diagonal element/ sum of off-diagonal elements is " << gworstRatio << ". Values about 1.0 are ok." << std::endl;
   }
 
-
   SC one = Teuchos::ScalarTraits<Scalar>::one(), zero = Teuchos::ScalarTraits<Scalar>::zero();
 
   // multiply with one vector
@@ -230,7 +231,7 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     Teuchos::RCP<Vector> res1 = VectorFactory::Build(A->getRangeMap(), false);
     A->apply(*ones, *res1, Teuchos::NO_TRANS, one, zero);
-    checkVectorEntries(res1,std::string("after applying the one vector to A"));
+    checkVectorEntries(res1, std::string("after applying the one vector to A"));
   }
 
   {
@@ -240,7 +241,7 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     Teuchos::RCP<Vector> resrand = VectorFactory::Build(A->getRangeMap(), false);
     A->apply(*randvec, *resrand, Teuchos::NO_TRANS, one, zero);
-    checkVectorEntries(resrand,std::string("after applying random vector to A"));
+    checkVectorEntries(resrand, std::string("after applying random vector to A"));
   }
 
   // apply Jacobi sweep
@@ -251,17 +252,17 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     Teuchos::RCP<Vector> res1 = VectorFactory::Build(A->getRangeMap(), false);
     A->apply(*ones, *res1, Teuchos::NO_TRANS, one, zero);
-    checkVectorEntries(res1,std::string("after applying one vector to A"));
+    checkVectorEntries(res1, std::string("after applying one vector to A"));
 
     Teuchos::RCP<Vector> invDiag = Utilities::GetMatrixDiagonalInverse(*A);
-    checkVectorEntries(invDiag,std::string("in invDiag"));
+    checkVectorEntries(invDiag, std::string("in invDiag"));
 
     Teuchos::RCP<Vector> res2 = VectorFactory::Build(A->getRangeMap(), false);
-    res2->elementWiseMultiply (0.8, *invDiag, *res1, 0.0);
-    checkVectorEntries(res2,std::string("after scaling Av with invDiag (with v the one vector)"));
+    res2->elementWiseMultiply(0.8, *invDiag, *res1, 0.0);
+    checkVectorEntries(res2, std::string("after scaling Av with invDiag (with v the one vector)"));
     res2->update(1.0, *ones, -1.0);
 
-    checkVectorEntries(res2,std::string("after applying one damped Jacobi sweep (with one vector)"));
+    checkVectorEntries(res2, std::string("after applying one damped Jacobi sweep (with one vector)"));
   }
 
   // apply Jacobi sweep
@@ -272,36 +273,36 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Lev
 
     Teuchos::RCP<Vector> res1 = VectorFactory::Build(A->getRangeMap(), false);
     A->apply(*ones, *res1, Teuchos::NO_TRANS, one, zero);
-    checkVectorEntries(res1,std::string("after applying a random vector to A"));
+    checkVectorEntries(res1, std::string("after applying a random vector to A"));
 
     Teuchos::RCP<Vector> invDiag = Utilities::GetMatrixDiagonalInverse(*A);
-    checkVectorEntries(invDiag,std::string("in invDiag"));
+    checkVectorEntries(invDiag, std::string("in invDiag"));
 
     Teuchos::RCP<Vector> res2 = VectorFactory::Build(A->getRangeMap(), false);
-    res2->elementWiseMultiply (0.8, *invDiag, *res1, 0.0);
-    checkVectorEntries(res2,std::string("after scaling Av with invDiag (with v a random vector)"));
+    res2->elementWiseMultiply(0.8, *invDiag, *res1, 0.0);
+    checkVectorEntries(res2, std::string("after scaling Av with invDiag (with v a random vector)"));
     res2->update(1.0, *ones, -1.0);
 
-    checkVectorEntries(res2,std::string("after applying one damped Jacobi sweep (with v a random vector)"));
+    checkVectorEntries(res2, std::string("after applying one damped Jacobi sweep (with v a random vector)"));
   }
 
   GetOStream(Runtime0) << "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ (Level " << fineLevel.GetLevelID() << ")" << std::endl;
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkVectorEntries(const Teuchos::RCP<Vector>& vec, std::string info) const {
-  SC zero = Teuchos::ScalarTraits<Scalar>::zero();
+void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkVectorEntries(const Teuchos::RCP<Vector> &vec, std::string info) const {
+  SC zero                                      = Teuchos::ScalarTraits<Scalar>::zero();
   Teuchos::RCP<const Teuchos::Comm<int> > comm = vec->getMap()->getComm();
-  Teuchos::ArrayRCP< const Scalar > vecData = vec->getData(0);
-  GO lzeros = 0;
-  GO gzeros = 0;
-  GO lnans = 0;
-  GO gnans = 0;
-
-  for(size_t i = 0; i<vec->getMap()->getLocalNumElements(); ++i) {
-    if(vecData[i] == zero) {
+  Teuchos::ArrayRCP<const Scalar> vecData      = vec->getData(0);
+  GO lzeros                                    = 0;
+  GO gzeros                                    = 0;
+  GO lnans                                     = 0;
+  GO gnans                                     = 0;
+
+  for (size_t i = 0; i < vec->getMap()->getLocalNumElements(); ++i) {
+    if (vecData[i] == zero) {
       lzeros++;
-    } else if(Teuchos::ScalarTraits<Scalar>::isnaninf(vecData[i])) {
+    } else if (Teuchos::ScalarTraits<Scalar>::isnaninf(vecData[i])) {
       lnans++;
     }
   }
@@ -310,11 +311,10 @@ void MatrixAnalysisFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::checkVect
   MueLu_sumAll(comm, lzeros, gzeros);
   MueLu_sumAll(comm, lnans, gnans);
 
-  if(gzeros > 0) GetOStream(Runtime0) << "Found " << gzeros << " zeros " << info << std::endl;
-  if(gnans > 0) GetOStream(Runtime0) << "Found " << gnans << " entries " << info << std::endl;
+  if (gzeros > 0) GetOStream(Runtime0) << "Found " << gzeros << " zeros " << info << std::endl;
+  if (gnans > 0) GetOStream(Runtime0) << "Found " << gnans << " entries " << info << std::endl;
 }
 
-} // namespace MueLu
-
+}  // namespace MueLu
 
 #endif /* MUELU_MATRIXANALYSISFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_Memory.cpp b/packages/muelu/src/Utils/MueLu_Memory.cpp
index 77b7cdd32d6a..52400450d9b6 100644
--- a/packages/muelu/src/Utils/MueLu_Memory.cpp
+++ b/packages/muelu/src/Utils/MueLu_Memory.cpp
@@ -47,7 +47,7 @@
 #include <fstream>
 #include "MueLu_Memory.hpp"
 
-#include <iostream> // TODO: remove
+#include <iostream>  // TODO: remove
 #include <unistd.h>
 #include <time.h>
 #ifdef MUELU_USE_MALLINFO
@@ -58,86 +58,80 @@
 
 namespace MueLu {
 
-  namespace MemUtils {
-
-    std::string PrintMemoryUsage() {
-
+namespace MemUtils {
 
+std::string PrintMemoryUsage() {
 #ifdef MUELU_USE_MALLINFO
-      struct mallinfo mem_stats = mallinfo();
-      double memory = mem_stats.hblkhd + mem_stats.usmblks + mem_stats.uordblks;
+  struct mallinfo mem_stats = mallinfo();
+  double memory             = mem_stats.hblkhd + mem_stats.usmblks + mem_stats.uordblks;
 
-      char memchar[128];
-      sprintf(memchar,"%12.1f MB",memory/1048576.0);
-      std::string mem(memchar);
+  char memchar[128];
+  sprintf(memchar, "%12.1f MB", memory / 1048576.0);
+  std::string mem(memchar);
 
-      return mem;
+  return mem;
 #else
-      std::ostringstream mem;
-      std::ifstream proc("/proc/self/status");
-      std::string s;
-
-      mem << PrintMemoryInfo() << " ";
-      while(getline(proc, s), !proc.fail()) {
-        if(s.substr(0, 6) == "VmSize") {
-          mem << s;
-          return mem.str();
-        }
-      }
+  std::ostringstream mem;
+  std::ifstream proc("/proc/self/status");
+  std::string s;
+
+  mem << PrintMemoryInfo() << " ";
+  while (getline(proc, s), !proc.fail()) {
+    if (s.substr(0, 6) == "VmSize") {
+      mem << s;
       return mem.str();
-#endif
-
     }
+  }
+  return mem.str();
+#endif
+}
 
-    std::string PrintMemoryInfo() {
-
+std::string PrintMemoryInfo() {
 #ifdef MUELU_USE_MALLINFO
-      struct mallinfo mem_stats = mallinfo();
-      double memory = mem_stats.hblkhd + mem_stats.usmblks + mem_stats.uordblks;
+  struct mallinfo mem_stats = mallinfo();
+  double memory             = mem_stats.hblkhd + mem_stats.usmblks + mem_stats.uordblks;
 
-      char memchar[128];
-      sprintf(memchar,"%12.1f MB",memory/1048576.0);
-      std::string mem(memchar);
+  char memchar[128];
+  sprintf(memchar, "%12.1f MB", memory / 1048576.0);
+  std::string mem(memchar);
 
-      return mem;
+  return mem;
 #else
-      std::ostringstream mem;
-      std::ifstream proc("/proc/meminfo");
-      std::string s;
-      while(getline(proc, s), !proc.fail()) {
-        if(s.substr(0, 7) == "MemFree") {
-          mem << s;
-          return mem.str();
-        }
-
-      }
+  std::ostringstream mem;
+  std::ifstream proc("/proc/meminfo");
+  std::string s;
+  while (getline(proc, s), !proc.fail()) {
+    if (s.substr(0, 7) == "MemFree") {
+      mem << s;
       return mem.str();
-#endif
     }
+  }
+  return mem.str();
+#endif
+}
 
-    void ReportTimeAndMemory(Teuchos::Time const &timer, Teuchos::Comm<int> const &Comm)
-    {
-      double maxTime=0,minTime=0,avgTime=0;
-      double localTime = timer.totalElapsedTime();
+void ReportTimeAndMemory(Teuchos::Time const &timer, Teuchos::Comm<int> const &Comm) {
+  double maxTime = 0, minTime = 0, avgTime = 0;
+  double localTime = timer.totalElapsedTime();
 #ifdef HAVE_MPI
-      int ntimers=1, root=0;
-      MPI_Reduce(&localTime,&maxTime,ntimers,MPI_DOUBLE,MPI_MAX,root,MPI_COMM_WORLD);
-      MPI_Reduce(&localTime,&minTime,ntimers,MPI_DOUBLE,MPI_MIN,root,MPI_COMM_WORLD);
-      MPI_Reduce(&localTime,&avgTime,ntimers,MPI_DOUBLE,MPI_SUM,root,MPI_COMM_WORLD);
+  int ntimers = 1, root = 0;
+  MPI_Reduce(&localTime, &maxTime, ntimers, MPI_DOUBLE, MPI_MAX, root, MPI_COMM_WORLD);
+  MPI_Reduce(&localTime, &minTime, ntimers, MPI_DOUBLE, MPI_MIN, root, MPI_COMM_WORLD);
+  MPI_Reduce(&localTime, &avgTime, ntimers, MPI_DOUBLE, MPI_SUM, root, MPI_COMM_WORLD);
 #else
-      maxTime = localTime;
-      minTime = localTime;
-      avgTime = localTime;
+  maxTime = localTime;
+  minTime = localTime;
+  avgTime = localTime;
 #endif
-      avgTime /= Comm.getSize();
-      //std::cout << "(" << Comm.getRank() << ") " << localTime << std::endl;
-      if (Comm.getRank()==0) {
-        std::cout << "&&&" << timer.name()
-                 << " max=" << maxTime << " min=" << minTime << " avg=" << avgTime << std::endl;
-        std::cout << "&&&" << timer.name() << " " << MemUtils::PrintMemoryUsage() << std::endl;
-      }
-    } //ReportTimeAndMemory
-
-  } //namespace MemUtils
-
-} //namespace MueLu
+  avgTime /= Comm.getSize();
+  // std::cout << "(" << Comm.getRank() << ") " << localTime << std::endl;
+  if (Comm.getRank() == 0) {
+    std::cout << "&&&" << timer.name()
+              << " max=" << maxTime << " min=" << minTime << " avg=" << avgTime << std::endl;
+    std::cout << "&&&" << timer.name() << " " << MemUtils::PrintMemoryUsage() << std::endl;
+  }
+}  // ReportTimeAndMemory
+
+}  // namespace MemUtils
+
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_Memory.hpp b/packages/muelu/src/Utils/MueLu_Memory.hpp
index ca50b05dffff..5cb425175b9d 100644
--- a/packages/muelu/src/Utils/MueLu_Memory.hpp
+++ b/packages/muelu/src/Utils/MueLu_Memory.hpp
@@ -50,19 +50,24 @@
 #include "Teuchos_Time.hpp"
 #include "Teuchos_Comm.hpp"
 
-namespace Teuchos { class Time; }
-namespace Teuchos { template <typename Ordinal> class Comm; }
+namespace Teuchos {
+class Time;
+}
+namespace Teuchos {
+template <typename Ordinal>
+class Comm;
+}
 
 namespace MueLu {
 
-  namespace MemUtils {
+namespace MemUtils {
 
-    std::string PrintMemoryUsage();
-    std::string PrintMemoryInfo();
-    void ReportTimeAndMemory(Teuchos::Time const &timer, Teuchos::Comm<int> const &Comm);
+std::string PrintMemoryUsage();
+std::string PrintMemoryInfo();
+void ReportTimeAndMemory(Teuchos::Time const &timer, Teuchos::Comm<int> const &Comm);
 
-  } //namespace MemUtils
+}  // namespace MemUtils
 
-} //namespace MueLu
+}  // namespace MueLu
 
-#endif //ifndef MUELU_MEMORY_HPP
+#endif  // ifndef MUELU_MEMORY_HPP
diff --git a/packages/muelu/src/Utils/MueLu_Monitor.cpp b/packages/muelu/src/Utils/MueLu_Monitor.cpp
index 8fd9a0d4b4ba..5ae7ef062074 100644
--- a/packages/muelu/src/Utils/MueLu_Monitor.cpp
+++ b/packages/muelu/src/Utils/MueLu_Monitor.cpp
@@ -1,82 +1,71 @@
 #include "MueLu_Monitor.hpp"
-int MueLu::FactoryMonitor::timerIdentifier_=0;
+int MueLu::FactoryMonitor::timerIdentifier_ = 0;
 
 namespace MueLu {
-  PrintMonitor::PrintMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel) : object_(object) {
-    tabbed = false;
-    if (object_.IsPrint(msgLevel)) {
-      // Print description and new indent
-      object_.GetOStream(msgLevel, 0) << msg << std::endl;
-      object_.getOStream()->pushTab();
-      tabbed = true;
-    }
+PrintMonitor::PrintMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel)
+  : object_(object) {
+  tabbed = false;
+  if (object_.IsPrint(msgLevel)) {
+    // Print description and new indent
+    object_.GetOStream(msgLevel, 0) << msg << std::endl;
+    object_.getOStream()->pushTab();
+    tabbed = true;
   }
+}
 
-  PrintMonitor::~PrintMonitor() { if (tabbed) object_.getOStream()->popTab(); }
+PrintMonitor::~PrintMonitor() {
+  if (tabbed) object_.getOStream()->popTab();
+}
 
-  Monitor::Monitor(const BaseClass& object, const std::string & msg, MsgType msgLevel, MsgType timerLevel)
-    : printMonitor_(object, msg + " (" + object.description() + ")", msgLevel),
-      timerMonitor_(object, object.ShortClassName() + ": " + msg + " (total)",    timerLevel)
-  { }
+Monitor::Monitor(const BaseClass& object, const std::string& msg, MsgType msgLevel, MsgType timerLevel)
+  : printMonitor_(object, msg + " (" + object.description() + ")", msgLevel)
+  , timerMonitor_(object, object.ShortClassName() + ": " + msg + " (total)", timerLevel) {}
 
-  Monitor::Monitor(const BaseClass& object, const std::string & msg, const std::string & label, MsgType msgLevel, MsgType timerLevel)
-    : printMonitor_(object, label + msg + " (" + object.description() + ")", msgLevel),
-      timerMonitor_(object, label + object.ShortClassName() + ": " + msg + " (total)",    timerLevel)
-  { }
+Monitor::Monitor(const BaseClass& object, const std::string& msg, const std::string& label, MsgType msgLevel, MsgType timerLevel)
+  : printMonitor_(object, label + msg + " (" + object.description() + ")", msgLevel)
+  , timerMonitor_(object, label + object.ShortClassName() + ": " + msg + " (total)", timerLevel) {}
 
-  SubMonitor::SubMonitor(const BaseClass& object, const std::string & msg, MsgType msgLevel, MsgType timerLevel)
-    : printMonitor_(object, msg, msgLevel),
-      timerMonitor_(object, object.ShortClassName() + ": " + msg + " (sub, total)", timerLevel)
-  { }
+SubMonitor::SubMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel, MsgType timerLevel)
+  : printMonitor_(object, msg, msgLevel)
+  , timerMonitor_(object, object.ShortClassName() + ": " + msg + " (sub, total)", timerLevel) {}
 
-  SubMonitor::SubMonitor(const BaseClass& object, const std::string & msg, const std::string & label, MsgType msgLevel, MsgType timerLevel)
-    : printMonitor_(object, label + msg, msgLevel),
-      timerMonitor_(object, label + object.ShortClassName() + ": " + msg + " (sub, total)", timerLevel)
-  { }
+SubMonitor::SubMonitor(const BaseClass& object, const std::string& msg, const std::string& label, MsgType msgLevel, MsgType timerLevel)
+  : printMonitor_(object, label + msg, msgLevel)
+  , timerMonitor_(object, label + object.ShortClassName() + ": " + msg + " (sub, total)", timerLevel) {}
 
-  FactoryMonitor::FactoryMonitor(const BaseClass& object, const std::string & msg, int levelID, MsgType msgLevel, MsgType timerLevel)
-    : Monitor(object, msg, msgLevel, timerLevel),
-      timerMonitorExclusive_(object, object.ShortClassName() + ": " + msg, timerLevel)
-  {
-    if (object.IsPrint(TimingsByLevel)) {
-      if (Teuchos::TimeMonitor::getStackedTimer().is_null())
-        levelTimeMonitor_ = rcp(new TimeMonitor(object, object.ShortClassName() + ": " + msg +
-            " (total, level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
-      levelTimeMonitorExclusive_ = rcp(new MutuallyExclusiveTimeMonitor<Level>(object, object.ShortClassName() +
-            ": " + msg + " (level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
-    }
+FactoryMonitor::FactoryMonitor(const BaseClass& object, const std::string& msg, int levelID, MsgType msgLevel, MsgType timerLevel)
+  : Monitor(object, msg, msgLevel, timerLevel)
+  , timerMonitorExclusive_(object, object.ShortClassName() + ": " + msg, timerLevel) {
+  if (object.IsPrint(TimingsByLevel)) {
+    if (Teuchos::TimeMonitor::getStackedTimer().is_null())
+      levelTimeMonitor_ = rcp(new TimeMonitor(object, object.ShortClassName() + ": " + msg + " (total, level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
+    levelTimeMonitorExclusive_ = rcp(new MutuallyExclusiveTimeMonitor<Level>(object, object.ShortClassName() + ": " + msg + " (level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
   }
+}
 
-  FactoryMonitor::FactoryMonitor(const BaseClass& object, const std::string & msg, const Level & level, MsgType msgLevel, MsgType timerLevel)
-    : Monitor(object, msg, FormattingHelper::getColonLabel(level.getObjectLabel()), msgLevel, timerLevel),
-      timerMonitorExclusive_(object, FormattingHelper::getColonLabel(level.getObjectLabel()) + object.ShortClassName() + ": " + msg, timerLevel)
-  {
-    if (object.IsPrint(TimingsByLevel)) {
-      std::string label = FormattingHelper::getColonLabel(level.getObjectLabel());
-      if (Teuchos::TimeMonitor::getStackedTimer().is_null())
-        levelTimeMonitor_ = rcp(new TimeMonitor(object, label+object.ShortClassName() + ": " +  msg +
-            " (total, level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
-      levelTimeMonitorExclusive_ = rcp(new MutuallyExclusiveTimeMonitor<Level>(object, label+object.ShortClassName() +
-            ": " + msg + " (level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
-    }
+FactoryMonitor::FactoryMonitor(const BaseClass& object, const std::string& msg, const Level& level, MsgType msgLevel, MsgType timerLevel)
+  : Monitor(object, msg, FormattingHelper::getColonLabel(level.getObjectLabel()), msgLevel, timerLevel)
+  , timerMonitorExclusive_(object, FormattingHelper::getColonLabel(level.getObjectLabel()) + object.ShortClassName() + ": " + msg, timerLevel) {
+  if (object.IsPrint(TimingsByLevel)) {
+    std::string label = FormattingHelper::getColonLabel(level.getObjectLabel());
+    if (Teuchos::TimeMonitor::getStackedTimer().is_null())
+      levelTimeMonitor_ = rcp(new TimeMonitor(object, label + object.ShortClassName() + ": " + msg + " (total, level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
+    levelTimeMonitorExclusive_ = rcp(new MutuallyExclusiveTimeMonitor<Level>(object, label + object.ShortClassName() + ": " + msg + " (level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
   }
+}
 
-  SubFactoryMonitor::SubFactoryMonitor(const BaseClass& object, const std::string & msg, int levelID, MsgType msgLevel, MsgType timerLevel)
-    : SubMonitor(object, msg, msgLevel, timerLevel)
-  {
-    if (object.IsPrint(TimingsByLevel) && Teuchos::TimeMonitor::getStackedTimer().is_null())
-      levelTimeMonitor_ = rcp(new TimeMonitor(object, object.ShortClassName() + ": " + msg +
-                                              " (sub, total, level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
-  }
+SubFactoryMonitor::SubFactoryMonitor(const BaseClass& object, const std::string& msg, int levelID, MsgType msgLevel, MsgType timerLevel)
+  : SubMonitor(object, msg, msgLevel, timerLevel) {
+  if (object.IsPrint(TimingsByLevel) && Teuchos::TimeMonitor::getStackedTimer().is_null())
+    levelTimeMonitor_ = rcp(new TimeMonitor(object, object.ShortClassName() + ": " + msg + " (sub, total, level=" + Teuchos::Utils::toString(levelID) + ")", timerLevel));
+}
 
-  SubFactoryMonitor::SubFactoryMonitor(const BaseClass& object, const std::string & msg, const Level & level, MsgType msgLevel, MsgType timerLevel)
-    : SubMonitor(object, msg, FormattingHelper::getColonLabel(level.getObjectLabel()), msgLevel, timerLevel)
-  {
-    if (object.IsPrint(TimingsByLevel) && Teuchos::TimeMonitor::getStackedTimer().is_null()) {
-      std::string label = FormattingHelper::getColonLabel(level.getObjectLabel());
-      levelTimeMonitor_ = rcp(new TimeMonitor(object, label+object.ShortClassName() + ": " +  msg +
-                                              " (sub, total, level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
-    }
+SubFactoryMonitor::SubFactoryMonitor(const BaseClass& object, const std::string& msg, const Level& level, MsgType msgLevel, MsgType timerLevel)
+  : SubMonitor(object, msg, FormattingHelper::getColonLabel(level.getObjectLabel()), msgLevel, timerLevel) {
+  if (object.IsPrint(TimingsByLevel) && Teuchos::TimeMonitor::getStackedTimer().is_null()) {
+    std::string label = FormattingHelper::getColonLabel(level.getObjectLabel());
+    levelTimeMonitor_ = rcp(new TimeMonitor(object, label + object.ShortClassName() + ": " + msg + " (sub, total, level=" + Teuchos::Utils::toString(level.GetLevelID()) + ")", timerLevel));
   }
+}
 
-} //namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_Monitor.hpp b/packages/muelu/src/Utils/MueLu_Monitor.hpp
index 232808d5b96e..9fc981000a96 100644
--- a/packages/muelu/src/Utils/MueLu_Monitor.hpp
+++ b/packages/muelu/src/Utils/MueLu_Monitor.hpp
@@ -47,13 +47,13 @@
 #define MUELU_MONITOR_HPP
 
 #include <string>
-#include <algorithm>                    // for swap
-#include <ostream>                      // for basic_ostream, operator<<, etc
-#include "Teuchos_FancyOStream.hpp"     // for OSTab, FancyOStream
-#include "Teuchos_RCPDecl.hpp"          // for RCP
-#include "Teuchos_RCP.hpp"              // for RCP::RCP<T>, RCP::operator=, etc
-#include "Teuchos_Utils.hpp"            // for Utils
-#include "MueLu_VerbosityLevel.hpp"     // for MsgType, MsgType::Runtime0, etc
+#include <algorithm>                 // for swap
+#include <ostream>                   // for basic_ostream, operator<<, etc
+#include "Teuchos_FancyOStream.hpp"  // for OSTab, FancyOStream
+#include "Teuchos_RCPDecl.hpp"       // for RCP
+#include "Teuchos_RCP.hpp"           // for RCP::RCP<T>, RCP::operator=, etc
+#include "Teuchos_Utils.hpp"         // for Utils
+#include "MueLu_VerbosityLevel.hpp"  // for MsgType, MsgType::Runtime0, etc
 #include "MueLu_BaseClass.hpp"
 #include "MueLu_FactoryBase_fwd.hpp"
 #include "MueLu_Level.hpp"
@@ -61,224 +61,219 @@
 
 namespace MueLu {
 
-  struct FormattingHelper {
-    //! Helper function for object label
-    static std::string getColonLabel(const std::string& label) {
-      if (label != "")
-        return label + ": ";
-      else
-        return "";
-    }
-  };
-
-
-  /*! @class PrintMonitor
-    Manages indentation of output using Teuchos::OSTab and verbosity level
+struct FormattingHelper {
+  //! Helper function for object label
+  static std::string getColonLabel(const std::string& label) {
+    if (label != "")
+      return label + ": ";
+    else
+      return "";
+  }
+};
+
+/*! @class PrintMonitor
+  Manages indentation of output using Teuchos::OSTab and verbosity level
+*/
+class PrintMonitor : public BaseClass {
+ public:
+  //! Constructor
+  PrintMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel = Runtime0);
+  ~PrintMonitor();
+
+ private:
+  PrintMonitor();
+
+  bool tabbed;
+  const BaseClass& object_;
+};
+
+/*! @class Monitor
+   @brief Timer to be used in non-factories.
+
+   A timer is created only if 'timerLevel' (Timings0 by default) is enabled.
+   This class manages verbosity, as well as times this object and all its children over all levels.
+
+   The timer yields output such as
+    \verbatim
+     MueLu: SaPFactory: Prolongator smoothing (total)             6.508 (6)        6.551 (6)        6.557 (6)        1.092 (6)
+    \endverbatim
+   Note that the keyword \a total denotes timing of the object and its children.
+
+   @ingroup MueLuTimerClasses
+*/
+class Monitor : public BaseClass {
+ public:
+  /*! @brief Constructor.
+
+      @param[in] object      Reference to the class instance that is creating this Monitor.
+      @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
   */
-  class PrintMonitor : public BaseClass {
-
-  public:
-
-    //! Constructor
-    PrintMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel = Runtime0);
-    ~PrintMonitor();
-
-  private:
-    PrintMonitor();
-
-    bool tabbed;
-    const BaseClass& object_;
-  };
-
-  /*! @class Monitor
-     @brief Timer to be used in non-factories.
-
-     A timer is created only if 'timerLevel' (Timings0 by default) is enabled.
-     This class manages verbosity, as well as times this object and all its children over all levels.
+  Monitor(const BaseClass& object, const std::string& msg, MsgType msgLevel = Runtime0, MsgType timerLevel = Timings0);
 
-     The timer yields output such as
-      \verbatim
-       MueLu: SaPFactory: Prolongator smoothing (total)             6.508 (6)        6.551 (6)        6.557 (6)        1.092 (6)
-      \endverbatim
-     Note that the keyword \a total denotes timing of the object and its children.
+  /*! @brief Constructor.
 
-     @ingroup MueLuTimerClasses
+      @param[in] object      Reference to the class instance that is creating this Monitor.
+      @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+      @param[in] label       An optional prefix label.
   */
-  class Monitor: public BaseClass {
-  public:
-    /*! @brief Constructor.
-
-        @param[in] object      Reference to the class instance that is creating this Monitor.
-        @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    Monitor(const BaseClass& object, const std::string & msg, MsgType msgLevel = Runtime0, MsgType timerLevel = Timings0);
-
-    /*! @brief Constructor.
-
-        @param[in] object      Reference to the class instance that is creating this Monitor.
-        @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-        @param[in] label       An optional prefix label.
-    */
-    Monitor(const BaseClass& object, const std::string & msg, const std::string & label, MsgType msgLevel = Runtime0, MsgType timerLevel = Timings0);
-
-    virtual ~Monitor() = default;
-    
-  private:
-    //! Manages printing.
-    PrintMonitor printMonitor_;
-    //! Records total time spent  in this object and all its children, over all levels.
-    TimeMonitor timerMonitor_;
-  };
-
-  //---------------------------------------------------------------------------------------------------
-
-  /*! @class SubMonitor
-    @brief Timer to be used in non-factories.  Similar to Monitor, but doesn't print object description.
-
-    Should be used in non-factory setting.
-    Times object and all its children over all levels.
-    A timer is created only if 'timerLevel' (Timings1 by default) is enabled.
-
-    The timer yields output such as
-     \verbatim
-     MueLu: SaPFactory: Eigenvalue estimate (sub, total)        0.149 (6)        0.1615 (6)       0.1643 (6)       0.02692 (6)
-     \endverbatim
-     Note that the keyword \a sub denotes that this is output from a SubFactoryMonitor.
-     Note that the keyword \a total denotes timing of the object and its children.
-
-     @ingroup MueLuTimerClasses
+  Monitor(const BaseClass& object, const std::string& msg, const std::string& label, MsgType msgLevel = Runtime0, MsgType timerLevel = Timings0);
+
+  virtual ~Monitor() = default;
+
+ private:
+  //! Manages printing.
+  PrintMonitor printMonitor_;
+  //! Records total time spent  in this object and all its children, over all levels.
+  TimeMonitor timerMonitor_;
+};
+
+//---------------------------------------------------------------------------------------------------
+
+/*! @class SubMonitor
+  @brief Timer to be used in non-factories.  Similar to Monitor, but doesn't print object description.
+
+  Should be used in non-factory setting.
+  Times object and all its children over all levels.
+  A timer is created only if 'timerLevel' (Timings1 by default) is enabled.
+
+  The timer yields output such as
+   \verbatim
+   MueLu: SaPFactory: Eigenvalue estimate (sub, total)        0.149 (6)        0.1615 (6)       0.1643 (6)       0.02692 (6)
+   \endverbatim
+   Note that the keyword \a sub denotes that this is output from a SubFactoryMonitor.
+   Note that the keyword \a total denotes timing of the object and its children.
+
+   @ingroup MueLuTimerClasses
+*/
+class SubMonitor : public BaseClass {
+ public:
+  /*! @brief Constructor.
+
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
   */
-  class SubMonitor: public BaseClass {
-  public:
-    /*! @brief Constructor.
-
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    SubMonitor(const BaseClass& object, const std::string & msg, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
-
-    /*! @brief Constructor.
-
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
-        @param[in] label       An optional prefix label.
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    SubMonitor(const BaseClass& object, const std::string & msg, const std::string & label, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
-
-  private:
-    PrintMonitor printMonitor_;
-    TimeMonitor timerMonitor_;
-  };
-
-  //---------------------------------------------------------------------------------------------------
-
-  /*! @class FactoryMonitor
-      @brief Timer to be used in factories.  Similar to Monitor but with additional timers.
-
-     This class provides the following three timers for an object:
-     - Timer for this object over all levels, excluding calls to children.
-     - Timer for this object per level, excluding calls to children.
-     - Timer for this object over all levels, including calls to children.
-
-     The three timers above yield output like the following:
-     \verbatim
-       MueLu: SaPFactory: Prolongator smoothing                     2.599 (6)        2.602 (6)        2.607 (6)        0.4336 (6)
-       MueLu: SaPFactory: Prolongator smoothing (level=5)           0.4731 (1)       0.4735 (1)       0.4736 (1)       0.4735 (1)
-       MueLu: SaPFactory: Prolongator smoothing (total)             6.508 (6)        6.551 (6)        6.557 (6)        1.092 (6)
-     \endverbatim
-
-     Note that the keyword \a total denotes timing of the object and its children.
-
-     @ingroup MueLuTimerClasses
-  */
-  class FactoryMonitor: public Monitor {
-  public:
-
-    static int timerIdentifier_;
+  SubMonitor(const BaseClass& object, const std::string& msg, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
 
-    /*! @brief Constructor
+  /*! @brief Constructor.
 
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
-        @param[in] levelID     The MueLu Level number.
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    FactoryMonitor(const BaseClass& object, const std::string & msg, int levelID, MsgType msgLevel = static_cast<MsgType>(Test | Runtime0), MsgType timerLevel = Timings0);
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
+      @param[in] label       An optional prefix label.
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+  */
+  SubMonitor(const BaseClass& object, const std::string& msg, const std::string& label, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
 
-    /*! @brief Constructor
+ private:
+  PrintMonitor printMonitor_;
+  TimeMonitor timerMonitor_;
+};
 
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build".
-        @param[in] level       The MueLu Level object.
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+//---------------------------------------------------------------------------------------------------
 
-      TODO: code factorization
-    */
-    FactoryMonitor(const BaseClass& object, const std::string & msg, const Level & level, MsgType msgLevel = static_cast<MsgType>(Test | Runtime0), MsgType timerLevel = Timings0);
+/*! @class FactoryMonitor
+    @brief Timer to be used in factories.  Similar to Monitor but with additional timers.
 
-  private:
-    //! Total time spent on this level in this object and all its children.
-    RCP<TimeMonitor>                           levelTimeMonitor_;
-    //! Total time spent on all levels in this object only, excluding all children.
-    MutuallyExclusiveTimeMonitor<FactoryBase>  timerMonitorExclusive_;
-    //! Total time spent on this level in this object only, excluding all children.
-    RCP<MutuallyExclusiveTimeMonitor<Level> >  levelTimeMonitorExclusive_;
-  };
+   This class provides the following three timers for an object:
+   - Timer for this object over all levels, excluding calls to children.
+   - Timer for this object per level, excluding calls to children.
+   - Timer for this object over all levels, including calls to children.
 
-  //---------------------------------------------------------------------------------------------------
+   The three timers above yield output like the following:
+   \verbatim
+     MueLu: SaPFactory: Prolongator smoothing                     2.599 (6)        2.602 (6)        2.607 (6)        0.4336 (6)
+     MueLu: SaPFactory: Prolongator smoothing (level=5)           0.4731 (1)       0.4735 (1)       0.4736 (1)       0.4735 (1)
+     MueLu: SaPFactory: Prolongator smoothing (total)             6.508 (6)        6.551 (6)        6.557 (6)        1.092 (6)
+   \endverbatim
 
-  /*! @class SubFactoryMonitor
-    @brief Timer to be used in factories.  Similar to SubMonitor but adds a timer level by level.
+   Note that the keyword \a total denotes timing of the object and its children.
 
-    Times an object and all its children on a level-by-level basis.
+   @ingroup MueLuTimerClasses
+*/
+class FactoryMonitor : public Monitor {
+ public:
+  static int timerIdentifier_;
 
-    This timer is useful for timing just a part of a factory; the keyword \a sub denotes that this is output from a SubFactoryMonitor.
-    The timer yields output such as
-     \verbatim
-       MueLu: SaPFactory: Eigenvalue estimate (sub, total, level=4)    0.01999 (1)      0.02103 (1)      0.02121 (1)      0.02103 (1)
-     \endverbatim
-     Note that the keyword \a total denotes timing of the object and its children.
+  /*! @brief Constructor
 
-     @ingroup MueLuTimerClasses
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
+      @param[in] levelID     The MueLu Level number.
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
   */
-  class SubFactoryMonitor: public SubMonitor {
-  public:
+  FactoryMonitor(const BaseClass& object, const std::string& msg, int levelID, MsgType msgLevel = static_cast<MsgType>(Test | Runtime0), MsgType timerLevel = Timings0);
 
-    /*! @brief Constructor
+  /*! @brief Constructor
 
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
-        @param[in] levelID     The Level number.
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    SubFactoryMonitor(const BaseClass& object, const std::string & msg, int levelID, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build".
+      @param[in] level       The MueLu Level object.
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
 
-    /*! @brief Constructor
+    TODO: code factorization
+  */
+  FactoryMonitor(const BaseClass& object, const std::string& msg, const Level& level, MsgType msgLevel = static_cast<MsgType>(Test | Runtime0), MsgType timerLevel = Timings0);
+
+ private:
+  //! Total time spent on this level in this object and all its children.
+  RCP<TimeMonitor> levelTimeMonitor_;
+  //! Total time spent on all levels in this object only, excluding all children.
+  MutuallyExclusiveTimeMonitor<FactoryBase> timerMonitorExclusive_;
+  //! Total time spent on this level in this object only, excluding all children.
+  RCP<MutuallyExclusiveTimeMonitor<Level> > levelTimeMonitorExclusive_;
+};
+
+//---------------------------------------------------------------------------------------------------
+
+/*! @class SubFactoryMonitor
+  @brief Timer to be used in factories.  Similar to SubMonitor but adds a timer level by level.
+
+  Times an object and all its children on a level-by-level basis.
+
+  This timer is useful for timing just a part of a factory; the keyword \a sub denotes that this is output from a SubFactoryMonitor.
+  The timer yields output such as
+   \verbatim
+     MueLu: SaPFactory: Eigenvalue estimate (sub, total, level=4)    0.01999 (1)      0.02103 (1)      0.02121 (1)      0.02103 (1)
+   \endverbatim
+   Note that the keyword \a total denotes timing of the object and its children.
+
+   @ingroup MueLuTimerClasses
+*/
+class SubFactoryMonitor : public SubMonitor {
+ public:
+  /*! @brief Constructor
+
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
+      @param[in] levelID     The Level number.
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+  */
+  SubFactoryMonitor(const BaseClass& object, const std::string& msg, int levelID, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
+
+  /*! @brief Constructor
 
-        @param[in] object      Reference to the class instance that is creating this SubMonitor.
-        @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
-        @param[in] level       The MueLu Level object.
-        @param[in] msgLevel    Governs whether information should be printed.
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    SubFactoryMonitor(const BaseClass& object, const std::string & msg, const Level & level, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
+      @param[in] object      Reference to the class instance that is creating this SubMonitor.
+      @param[in] msg         String that indicates what the SubMonitor is monitoring, e.g., "Build"
+      @param[in] level       The MueLu Level object.
+      @param[in] msgLevel    Governs whether information should be printed.
+      @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+  */
+  SubFactoryMonitor(const BaseClass& object, const std::string& msg, const Level& level, MsgType msgLevel = Runtime1, MsgType timerLevel = Timings1);
 
-  private:
-    //! Total time spent on this level in this object and all children.
-    RCP<TimeMonitor> levelTimeMonitor_;
-  };
+ private:
+  //! Total time spent on this level in this object and all children.
+  RCP<TimeMonitor> levelTimeMonitor_;
+};
 
-} // namespace MueLu
+}  // namespace MueLu
 
-#endif // MUELU_MONITOR_HPP
+#endif  // MUELU_MONITOR_HPP
diff --git a/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.cpp b/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.cpp
index d8934b1ae99f..3a94573da9cd 100644
--- a/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.cpp
+++ b/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.cpp
@@ -45,168 +45,173 @@
 // @HEADER
 
 #include <map>
-#include <iostream>                     // for basic_ostream, etc
-#include <utility>                      // for pair
-#include "Teuchos_FancyOStream.hpp"     // for basic_FancyOStream, etc
-#include "Teuchos_RCP.hpp"              // for RCP::operator->, etc
+#include <iostream>                      // for basic_ostream, etc
+#include <utility>                       // for pair
+#include "Teuchos_FancyOStream.hpp"      // for basic_FancyOStream, etc
+#include "Teuchos_RCP.hpp"               // for RCP::operator->, etc
 #include "Teuchos_TestForException.hpp"  // for TEUCHOS_TEST_FOR_EXCEPTION
 #include "Teuchos_Time.hpp"
 #include "Teuchos_TimeMonitor.hpp"
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_Exceptions.hpp"
 #include "MueLu_BaseClass.hpp"
-#include "MueLu_VerbosityLevel.hpp"     // for MsgType::Debug, etc
+#include "MueLu_VerbosityLevel.hpp"  // for MsgType::Debug, etc
 #include "MueLu_MutuallyExclusiveTime.hpp"
 #include "MueLu_FactoryBase.hpp"
 #include "MueLu_Level.hpp"
 
 namespace MueLu {
 
-  std::map<std::string,std::string> myParent_;
-
-  template <class TagName>
-  MutuallyExclusiveTime<TagName>::MutuallyExclusiveTime(const std::string &name, bool startFlag)
-    : name_(name),
-      timer_(rcp(new Teuchos::Time(name, false))),  // second argument is false in any case, because if start==true,
-                                                    // timer has to be started by MutuallyExclusiveTime::start() instead of Teuchos::Time::start().
-      isPaused_(false)
-  {
-    if (startFlag == true) timer_->start();
-  }
-
-  template <class TagName>
-  MutuallyExclusiveTime<TagName>::~MutuallyExclusiveTime() {
-    // This timer can only be destroyed if it is not in the stack
-    if (isPaused()) {
-      // error message because cannot throw an exception in destructor
-      GetOStream(Errors) << "MutuallyExclusiveTime::~MutuallyExclusiveTime(): Error: destructor called on a paused timer." << std::endl;
-      //TODO: Even if timing results will be wrong, the timer can be removed from the stack to avoid a segmentation fault.
-    }
-
-    stop(); // if isRunning(), remove from the stack, resume previous timer
+std::map<std::string, std::string> myParent_;
+
+template <class TagName>
+MutuallyExclusiveTime<TagName>::MutuallyExclusiveTime(const std::string& name, bool startFlag)
+  : name_(name)
+  , timer_(rcp(new Teuchos::Time(name, false)))
+  ,  // second argument is false in any case, because if start==true,
+     // timer has to be started by MutuallyExclusiveTime::start() instead of Teuchos::Time::start().
+  isPaused_(false) {
+  if (startFlag == true) timer_->start();
+}
+
+template <class TagName>
+MutuallyExclusiveTime<TagName>::~MutuallyExclusiveTime() {
+  // This timer can only be destroyed if it is not in the stack
+  if (isPaused()) {
+    // error message because cannot throw an exception in destructor
+    GetOStream(Errors) << "MutuallyExclusiveTime::~MutuallyExclusiveTime(): Error: destructor called on a paused timer." << std::endl;
+    // TODO: Even if timing results will be wrong, the timer can be removed from the stack to avoid a segmentation fault.
   }
 
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::start(bool reset) {
-    TEUCHOS_TEST_FOR_EXCEPTION(isPaused(), Exceptions::RuntimeError, "MueLu::MutuallyExclusiveTime::start(): timer is paused. Use resume().");
-
-    if (isRunning()) { return; } // If timer is already running, do not pause/push-in-the-stack/start the timer.
-                                 // Otherwise, something bad will happen when this.stop() will be called
-
-    // pause currently running timer
-    if (!timerStack_.empty()) {
-      GetOStream(Debug) << "pausing parent timer " << timerStack_.top()->name_ << std::endl;
-      timerStack_.top()->pause();
-      GetOStream(Debug) << "starting child timer " << this->name_ << std::endl;
-      myParent_[this->name_] = timerStack_.top()->name_;
-    } else {
-      GetOStream(Debug) << "starting orphan timer " << this->name_ << std::endl;
-      myParent_[this->name_] = "no parent";
-    }
-
-    // start this timer
-    timer_->start(reset);
-    timerStack_.push(this);
+  stop();  // if isRunning(), remove from the stack, resume previous timer
+}
+
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::start(bool reset) {
+  TEUCHOS_TEST_FOR_EXCEPTION(isPaused(), Exceptions::RuntimeError, "MueLu::MutuallyExclusiveTime::start(): timer is paused. Use resume().");
+
+  if (isRunning()) {
+    return;
+  }  // If timer is already running, do not pause/push-in-the-stack/start the timer.
+     // Otherwise, something bad will happen when this.stop() will be called
+
+  // pause currently running timer
+  if (!timerStack_.empty()) {
+    GetOStream(Debug) << "pausing parent timer " << timerStack_.top()->name_ << std::endl;
+    timerStack_.top()->pause();
+    GetOStream(Debug) << "starting child timer " << this->name_ << std::endl;
+    myParent_[this->name_] = timerStack_.top()->name_;
+  } else {
+    GetOStream(Debug) << "starting orphan timer " << this->name_ << std::endl;
+    myParent_[this->name_] = "no parent";
   }
 
-  template <class TagName>
-  double MutuallyExclusiveTime<TagName>::stop() {
-    if(isPaused())
-      GetOStream(Errors) << "MueLu::MutuallyExclusiveTime::stop(): timer is paused. Use resume()" << std::endl;
+  // start this timer
+  timer_->start(reset);
+  timerStack_.push(this);
+}
 
-    if (!isRunning()) { return timer_->stop(); } // stop() can be called on stopped timer
+template <class TagName>
+double MutuallyExclusiveTime<TagName>::stop() {
+  if (isPaused())
+    GetOStream(Errors) << "MueLu::MutuallyExclusiveTime::stop(): timer is paused. Use resume()" << std::endl;
 
-    // Here, timer is running, so it is the head of the stack
-    TopOfTheStack();
+  if (!isRunning()) {
+    return timer_->stop();
+  }  // stop() can be called on stopped timer
 
-    timerStack_.pop();
-    double r = timer_->stop();
+  // Here, timer is running, so it is the head of the stack
+  TopOfTheStack();
 
-    if (!timerStack_.empty()) {
-          GetOStream(Debug) << "resuming timer " << timerStack_.top()->name_ << std::endl;
-          timerStack_.top()->resume();
-    }
+  timerStack_.pop();
+  double r = timer_->stop();
 
-    return r;
+  if (!timerStack_.empty()) {
+    GetOStream(Debug) << "resuming timer " << timerStack_.top()->name_ << std::endl;
+    timerStack_.top()->resume();
   }
 
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::pause() {
-    if (isPaused()) // calling twice pause() is allowed
-      return;
+  return r;
+}
 
-    TopOfTheStack();
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::pause() {
+  if (isPaused())  // calling twice pause() is allowed
+    return;
 
-    timer_->stop();
-    isPaused_ = true;
-  }
+  TopOfTheStack();
 
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::resume() {
-    TopOfTheStack();
+  timer_->stop();
+  isPaused_ = true;
+}
 
-    // no 'shortcut' test necessary:
-    // - if timer is stop, it is in pause (cannot be stop and not in pause because this timer is the head of the stack).
-    // - if timer is running, nothing is changed by this function.
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::resume() {
+  TopOfTheStack();
 
-    timer_->start(false);
-    isPaused_ = false;
-  }
+  // no 'shortcut' test necessary:
+  // - if timer is stop, it is in pause (cannot be stop and not in pause because this timer is the head of the stack).
+  // - if timer is running, nothing is changed by this function.
 
-  template <class TagName>
-  bool MutuallyExclusiveTime<TagName>::isRunning() {
-    if (timer_->isRunning()) {
-             // TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.top() != this, Exceptions::RuntimeError,
-             //            "MueLu::MutuallyExclusiveTime::isRunning(): this timer is active so it is supposed to be the head of the stack");
-    }
-    return timer_->isRunning();
-  }
+  timer_->start(false);
+  isPaused_ = false;
+}
 
-  template <class TagName>
-  bool MutuallyExclusiveTime<TagName>::isPaused() {
-    TEUCHOS_TEST_FOR_EXCEPTION(isPaused_ && timer_->isRunning(), Exceptions::RuntimeError, "");
-    return isPaused_;
+template <class TagName>
+bool MutuallyExclusiveTime<TagName>::isRunning() {
+  if (timer_->isRunning()) {
+    // TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.top() != this, Exceptions::RuntimeError,
+    //            "MueLu::MutuallyExclusiveTime::isRunning(): this timer is active so it is supposed to be the head of the stack");
   }
-
-  template <class TagName>
-  RCP<MutuallyExclusiveTime<TagName> > MutuallyExclusiveTime<TagName>::getNewTimer(const std::string& name) {
-    RCP<MutuallyExclusiveTime<TagName> > timer = rcp(new MutuallyExclusiveTime<TagName>(Teuchos::TimeMonitor::getNewTimer(name)));
-    timer->name_ = name;
-    return timer;
+  return timer_->isRunning();
+}
+
+template <class TagName>
+bool MutuallyExclusiveTime<TagName>::isPaused() {
+  TEUCHOS_TEST_FOR_EXCEPTION(isPaused_ && timer_->isRunning(), Exceptions::RuntimeError, "");
+  return isPaused_;
+}
+
+template <class TagName>
+RCP<MutuallyExclusiveTime<TagName> > MutuallyExclusiveTime<TagName>::getNewTimer(const std::string& name) {
+  RCP<MutuallyExclusiveTime<TagName> > timer = rcp(new MutuallyExclusiveTime<TagName>(Teuchos::TimeMonitor::getNewTimer(name)));
+  timer->name_                               = name;
+  return timer;
+}
+
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::incrementNumCalls() { timer_->incrementNumCalls(); }
+
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::PrintParentChildPairs() {
+  // key is child, value is parent
+  RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fos->setOutputToRootOnly(0);
+  *fos << "Parent Child Map" << std::endl;
+  std::map<std::string, std::string>::const_iterator iter;
+  for (iter = ::MueLu::myParent_.begin(); iter != ::MueLu::myParent_.end(); ++iter) {
+    *fos << "Key: " << iter->first << "  Value: " << iter->second << std::endl;
   }
+}
 
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::incrementNumCalls() { timer_->incrementNumCalls(); }
-
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::PrintParentChildPairs() {
-    //key is child, value is parent
-    RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout)); fos->setOutputToRootOnly(0);
-    *fos << "Parent Child Map" << std::endl;
-    std::map<std::string, std::string >::const_iterator iter;
-    for (iter = ::MueLu::myParent_.begin(); iter != ::MueLu::myParent_.end(); ++iter) {
-      *fos << "Key: " << iter->first << "  Value: " << iter->second << std::endl;
-    }
-  }
+template <class TagName>
+MutuallyExclusiveTime<TagName>::MutuallyExclusiveTime(RCP<Teuchos::Time> timer)
+  : timer_(timer)
+  , isPaused_(false) {}
 
-  template <class TagName>
-  MutuallyExclusiveTime<TagName>::MutuallyExclusiveTime(RCP<Teuchos::Time> timer)
-    : timer_(timer), isPaused_(false)
-  { }
-
-  template <class TagName>
-  void MutuallyExclusiveTime<TagName>::TopOfTheStack() {
-    TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.empty(), Exceptions::RuntimeError, "MueLu::MutuallyExclusiveTime::TopOfTheStack(): timer is not the head of the stack (stack is empty).");
-    // TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.top() != this, Exceptions::RuntimeError,  "MueLu::MutuallyExclusiveTime::TopOfTheStack(): timer is not the head of the stack.");
-    TEUCHOS_TEST_FOR_EXCEPTION(!(isRunning() || isPaused()), Exceptions::RuntimeError,  "MueLu::MutuallyExclusiveTime::TopOfTheStack(): head of the stack timer is neither active nor paused.");
-  }
+template <class TagName>
+void MutuallyExclusiveTime<TagName>::TopOfTheStack() {
+  TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.empty(), Exceptions::RuntimeError, "MueLu::MutuallyExclusiveTime::TopOfTheStack(): timer is not the head of the stack (stack is empty).");
+  // TEUCHOS_TEST_FOR_EXCEPTION(timerStack_.top() != this, Exceptions::RuntimeError,  "MueLu::MutuallyExclusiveTime::TopOfTheStack(): timer is not the head of the stack.");
+  TEUCHOS_TEST_FOR_EXCEPTION(!(isRunning() || isPaused()), Exceptions::RuntimeError, "MueLu::MutuallyExclusiveTime::TopOfTheStack(): head of the stack timer is neither active nor paused.");
+}
 
-  template <class TagName>
-  std::stack<MutuallyExclusiveTime<TagName>*> MutuallyExclusiveTime<TagName>::timerStack_;
+template <class TagName>
+std::stack<MutuallyExclusiveTime<TagName>*> MutuallyExclusiveTime<TagName>::timerStack_;
 
-  //FIXME: move this:
-  template class MutuallyExclusiveTime<FactoryBase>;
-  template class MutuallyExclusiveTime<Level>;
-  template class MutuallyExclusiveTime<BaseClass>;
+// FIXME: move this:
+template class MutuallyExclusiveTime<FactoryBase>;
+template class MutuallyExclusiveTime<Level>;
+template class MutuallyExclusiveTime<BaseClass>;
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.hpp b/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.hpp
index a3ae0dba6f3b..4eb528585693 100644
--- a/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.hpp
+++ b/packages/muelu/src/Utils/MueLu_MutuallyExclusiveTime.hpp
@@ -48,7 +48,7 @@
 
 #include <string>
 #include <stack>
-#include "Teuchos_RCPDecl.hpp"          // for RCP
+#include "Teuchos_RCPDecl.hpp"  // for RCP
 
 // types for explicit instantiation
 #include "MueLu_FactoryBase_fwd.hpp"
@@ -57,120 +57,113 @@
 
 namespace MueLu {
 
-  /*! @class MutuallyExclusiveTime
+/*! @class MutuallyExclusiveTime
 
-      @brief This class wraps a Teuchos::Time and maintains a mutually exclusive property between wrapped timers.
+    @brief This class wraps a Teuchos::Time and maintains a mutually exclusive property between wrapped timers.
 
-      This class is useful to exclude from a timer the execution time of a subroutine;
-      when a MutuallyExclusiveTime is running, other timers are not running.
-      Timers have three states (running, stopped, paused) to enforce the mutually exclusive property.
-      When the running timer is stopped, the last active timer is restarted. A stack of timers is used internally to support this functionality.
-      The template parameter of this class can be used to define several sets of mutually exclusive timers.
+    This class is useful to exclude from a timer the execution time of a subroutine;
+    when a MutuallyExclusiveTime is running, other timers are not running.
+    Timers have three states (running, stopped, paused) to enforce the mutually exclusive property.
+    When the running timer is stopped, the last active timer is restarted. A stack of timers is used internally to support this functionality.
+    The template parameter of this class can be used to define several sets of mutually exclusive timers.
 
-      Note: Only one timer can be active at a time but all timers can be inactive at the same time. Timers cannot be destroyed when they are in 'paused'.
+    Note: Only one timer can be active at a time but all timers can be inactive at the same time. Timers cannot be destroyed when they are in 'paused'.
 
-      @todo inheritence from PerformanceMonitorBase<Time> ?
-  */
-  template<class TagName>
-  class MutuallyExclusiveTime : public BaseClass {
-
-  public:
-
-    //! @name Constructor/Destructor
-    //@{
-    //!Constructor
-    MutuallyExclusiveTime(const std::string &name, bool startFlag=false);
-
-    //!Destructor
-    ~MutuallyExclusiveTime();
-    //@}
-
-    //! @brief Starts the timer. If a MutuallyExclusiveTime timer is running, it will be stopped.
-    //! @pre Timer is not already paused.
-    //! @post Timer is running. Other MutuallyExclusiveTime timers are paused or stopped.
-    void start(bool reset=false);
+    @todo inheritence from PerformanceMonitorBase<Time> ?
+*/
+template <class TagName>
+class MutuallyExclusiveTime : public BaseClass {
+ public:
+  //! @name Constructor/Destructor
+  //@{
+  //! Constructor
+  MutuallyExclusiveTime(const std::string& name, bool startFlag = false);
 
-    //! @name Functions that can only be called on the most recent timer (i.e., the running or last paused timer)
-    //@{
+  //! Destructor
+  ~MutuallyExclusiveTime();
+  //@}
 
-    //! @brief Stops the timer.
-    //! The previous MutuallyExclusiveTime that has been paused when this timer was started will be resumed.
-    //! This method can be called on an already stopped timer or on the currently running timer.
-    double stop();
+  //! @brief Starts the timer. If a MutuallyExclusiveTime timer is running, it will be stopped.
+  //! @pre Timer is not already paused.
+  //! @post Timer is running. Other MutuallyExclusiveTime timers are paused or stopped.
+  void start(bool reset = false);
 
-    //! Pause running timer. Used internally by start().
-    void pause();
+  //! @name Functions that can only be called on the most recent timer (i.e., the running or last paused timer)
+  //@{
 
-    //! @brief Resume paused timer.
-    //! Used internally by stop().  Timer is not reset.
-    //! @pre Timer is at the top of the stack.
+  //! @brief Stops the timer.
+  //! The previous MutuallyExclusiveTime that has been paused when this timer was started will be resumed.
+  //! This method can be called on an already stopped timer or on the currently running timer.
+  double stop();
 
-    void resume();
+  //! Pause running timer. Used internally by start().
+  void pause();
 
-    //@}
+  //! @brief Resume paused timer.
+  //! Used internally by stop().  Timer is not reset.
+  //! @pre Timer is at the top of the stack.
 
+  void resume();
 
-    //! @name Query methods.
-    //@{
+  //@}
 
-    bool isRunning();
+  //! @name Query methods.
+  //@{
 
-    bool isPaused();
+  bool isRunning();
 
-    //@}
+  bool isPaused();
 
-    //! Return a new MutuallyExclusiveTime that is registered with the Teuchos::TimeMonitor (for timer summary).
-    // Note: this function is provided by the timer class, not by a monitor (!= Teuchos)
-    static RCP<MutuallyExclusiveTime<TagName> > getNewTimer(const std::string& name);
+  //@}
 
-    //! Increment the number of times this timer has been called.
-    void incrementNumCalls();
+  //! Return a new MutuallyExclusiveTime that is registered with the Teuchos::TimeMonitor (for timer summary).
+  // Note: this function is provided by the timer class, not by a monitor (!= Teuchos)
+  static RCP<MutuallyExclusiveTime<TagName> > getNewTimer(const std::string& name);
 
-    //! Name of this mutually exclusive timer.
-    std::string name_;
+  //! Increment the number of times this timer has been called.
+  void incrementNumCalls();
 
-    /*! @brief Print std::map of (child,parent) pairs for post-run analysis.
+  //! Name of this mutually exclusive timer.
+  std::string name_;
 
-        This method print a std::map.  Each key is a child, and the value is its parent.
-        The (child,parent0) pairs can be used by the MueLu script ``mueprof.sh" to print a
-        nice hierarchical tree that shows both runtime flow and time for each section.
-    */
-    static void PrintParentChildPairs();
+  /*! @brief Print std::map of (child,parent) pairs for post-run analysis.
 
-  private:
-
-    //! This constructor is not public to prevent users from using Teuchos::Time::start()/stop()
-    //! instead of MutuallyExclusiveTime::start()/stop(), if they have access to the underlying Teuchos::Time object.
-    MutuallyExclusiveTime(RCP<Teuchos::Time> timer);
-
-    // MutuallyExclusiveTime() { }
-
-    //! Using an RCP allows to use Teuchos::TimeMonitor to keep track of the timer.
-    RCP<Teuchos::Time> timer_;
-    bool isPaused_;
+      This method print a std::map.  Each key is a child, and the value is its parent.
+      The (child,parent0) pairs can be used by the MueLu script ``mueprof.sh" to print a
+      nice hierarchical tree that shows both runtime flow and time for each section.
+  */
+  static void PrintParentChildPairs();
 
-    //! Stack of created timers (active or paused timers).
-    // - empty when no active timer
-    // - head is the active timer
-    // - other timers are timers paused to enforce the mutually exclusive property of the timer set.
-    static std::stack<MutuallyExclusiveTime<TagName>*> timerStack_;
-    //static std::map<std::string,std::string> myParent_;
+ private:
+  //! This constructor is not public to prevent users from using Teuchos::Time::start()/stop()
+  //! instead of MutuallyExclusiveTime::start()/stop(), if they have access to the underlying Teuchos::Time object.
+  MutuallyExclusiveTime(RCP<Teuchos::Time> timer);
 
-    //! Check if 'this' is the head of the stack.
-    void TopOfTheStack();
+  // MutuallyExclusiveTime() { }
 
-  //TODO: test integrity of the stack:
-  // Head = running or paused
-  // Other timers of the stack = paused
+  //! Using an RCP allows to use Teuchos::TimeMonitor to keep track of the timer.
+  RCP<Teuchos::Time> timer_;
+  bool isPaused_;
 
-  };
+  //! Stack of created timers (active or paused timers).
+  // - empty when no active timer
+  // - head is the active timer
+  // - other timers are timers paused to enforce the mutually exclusive property of the timer set.
+  static std::stack<MutuallyExclusiveTime<TagName>*> timerStack_;
+  // static std::map<std::string,std::string> myParent_;
 
-  extern template class MutuallyExclusiveTime<FactoryBase>;
-  extern template class MutuallyExclusiveTime<Level>;
-  extern template class MutuallyExclusiveTime<BaseClass>;
+  //! Check if 'this' is the head of the stack.
+  void TopOfTheStack();
 
-} // namespace MueLu
+  // TODO: test integrity of the stack:
+  //  Head = running or paused
+  //  Other timers of the stack = paused
+};
 
-#endif // MUELU_MUTUALLYEXCLUSIVETIME_HPP
+extern template class MutuallyExclusiveTime<FactoryBase>;
+extern template class MutuallyExclusiveTime<Level>;
+extern template class MutuallyExclusiveTime<BaseClass>;
 
+}  // namespace MueLu
 
+#endif  // MUELU_MUTUALLYEXCLUSIVETIME_HPP
diff --git a/packages/muelu/src/Utils/MueLu_PerfModels_decl.hpp b/packages/muelu/src/Utils/MueLu_PerfModels_decl.hpp
index bd5eed1f59c1..005822c4a9ce 100644
--- a/packages/muelu/src/Utils/MueLu_PerfModels_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_PerfModels_decl.hpp
@@ -57,116 +57,112 @@
 
 namespace MueLu {
 
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class PerfModels {
-  public:
-    PerfModels();
-
-    /* Single Node tests based upon the STREAM benchmark for measuring memory
-     * bandwith and computation rate. These processes compute either the addition
-     * of two vectors or the multiplication of dense matrices of any given size.
-     * Many iterations occur which then return a vector containing the individual
-     * lengths of time per iteration.
-     *
-     * See further here:
-     *    - https://www.cs.virginia.edu/stream/ref.html
-     *    - https://github.com/UoB-HPC/BabelStream
-     */
-
-    /* This version is for table interpolation and works on chars, so the LOG_MAX_SIZE is for bytes */
-    void stream_vector_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE=20);
-    bool has_stream_vector_table() const {return stream_sizes_.size() > 0;}
-
-    /* Lookup in the stream_vector table */
-    double stream_vector_copy_lookup(int SIZE_IN_BYTES);
-    double stream_vector_add_lookup(int SIZE_IN_BYTES);
-    double latency_corrected_stream_vector_copy_lookup(int SIZE_IN_BYTES);
-    double latency_corrected_stream_vector_add_lookup(int SIZE_IN_BYTES);
-
-    // Uses the faster of the tables.  The time is then divided by the number of memory transactions
-    // per element in the kernel (e.g. 2 for COPY and 3 for ADD).
-    double stream_vector_lookup(int SIZE_IN_BYTES);   
-    double latency_corrected_stream_vector_lookup(int SIZE_IN_BYTES);   
-
-    /* Print table */
-    void print_stream_vector_table(std::ostream & out, const std::string & prefix="");
-    void print_latency_corrected_stream_vector_table(std::ostream & out, const std::string & prefix="");
-
-    /* A latency test between two processes based upon the MVAPICH OSU Micro-Benchmarks.
-     * The sender process sends a message and then waits for confirmation of reception.
-     * Many iterations occur with various message sizes and the average latency values
-     * are returned within a map. Utilizes blocking send and recieve.
-     *
-     * See further: https://mvapich.cse.ohio-state.edu/benchmarks/
-     */    
-    void pingpong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Teuchos::Comm<int> > &comm);
-    bool has_pingpong_table() const {return pingpong_sizes_.size() > 0;}
-
-    /* Lookup in the pingpong_vector table */
-    double pingpong_host_lookup(int SIZE_IN_BYTES);
-    double pingpong_device_lookup(int SIZE_IN_BYTES);
-
-    /* Print table */
-    void print_pingpong_table(std::ostream & out, const std::string & prefix="");
-
-    /* A halo-exchange based ping-pong, inspired by halo-mode in MPPTEST from ANL.
-     * Here we use exactly the communication pattern specified in the import object
-     * and send messages accordingly.  We vary the size in bytes sent per message,
-     * which should capture max-rate effects to some degree.
-     *
-     * See further: https://www.mcs.anl.gov/research/projects/mpi/mpptest/
-     */
-    void halopong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > & import);
-    bool has_halopong_table() const {return halopong_sizes_.size() > 0;}
-
-    /* Lookup in the halopong_vector table */
-    double halopong_host_lookup(int SIZE_IN_BYTES_PER_MESSAGE);
-    double halopong_device_lookup(int SIZE_IN_BYTES_PER_MESSAGE);
-
-    /* Print table */
-    void print_halopong_table(std::ostream & out, const std::string & prefix="");
-
-
-
-    /* Estimate launch latency based on the cost of submitting an empty Kokkos::parallel_for.
-     * This necessary to correct the memory bandwidth costs for models on high latency platforms, 
-     * e.g., GPUS.
-     */
-    void launch_latency_make_table(int KERNEL_REPEATS);
-    bool has_launch_latency_table() const  {return launch_and_wait_latency_ > 0;}
-
-    /* Lookup launch latency */
-    double launch_latency_lookup();
-       
-    /* Print table */
-    void print_launch_latency_table(std::ostream & out, const std::string & prefix="");
-
-  private:
-    void print_stream_vector_table_impl(std::ostream & out,bool use_latency_correction, const std::string & prefix);    
-
-
-    std::vector<int>    stream_sizes_;
-    std::vector<double> stream_copy_times_;
-    std::vector<double> stream_add_times_;
-    std::vector<double> latency_corrected_stream_copy_times_;
-    std::vector<double> latency_corrected_stream_add_times_;
-
-    std::vector<int>     pingpong_sizes_;
-    std::vector<double>  pingpong_host_times_;
-    std::vector<double>  pingpong_device_times_;
-
-    std::vector<int>     halopong_sizes_;
-    std::vector<double>  halopong_host_times_;
-    std::vector<double>  halopong_device_times_;
-
-    double launch_and_wait_latency_;
-
-
-  }; //class PerfModels
-
-} //namespace MueLu
-
-#endif //ifndef MUELU_PERFMODELS_HPP
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PerfModels {
+ public:
+  PerfModels();
+
+  /* Single Node tests based upon the STREAM benchmark for measuring memory
+   * bandwith and computation rate. These processes compute either the addition
+   * of two vectors or the multiplication of dense matrices of any given size.
+   * Many iterations occur which then return a vector containing the individual
+   * lengths of time per iteration.
+   *
+   * See further here:
+   *    - https://www.cs.virginia.edu/stream/ref.html
+   *    - https://github.com/UoB-HPC/BabelStream
+   */
+
+  /* This version is for table interpolation and works on chars, so the LOG_MAX_SIZE is for bytes */
+  void stream_vector_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE = 20);
+  bool has_stream_vector_table() const { return stream_sizes_.size() > 0; }
+
+  /* Lookup in the stream_vector table */
+  double stream_vector_copy_lookup(int SIZE_IN_BYTES);
+  double stream_vector_add_lookup(int SIZE_IN_BYTES);
+  double latency_corrected_stream_vector_copy_lookup(int SIZE_IN_BYTES);
+  double latency_corrected_stream_vector_add_lookup(int SIZE_IN_BYTES);
+
+  // Uses the faster of the tables.  The time is then divided by the number of memory transactions
+  // per element in the kernel (e.g. 2 for COPY and 3 for ADD).
+  double stream_vector_lookup(int SIZE_IN_BYTES);
+  double latency_corrected_stream_vector_lookup(int SIZE_IN_BYTES);
+
+  /* Print table */
+  void print_stream_vector_table(std::ostream &out, const std::string &prefix = "");
+  void print_latency_corrected_stream_vector_table(std::ostream &out, const std::string &prefix = "");
+
+  /* A latency test between two processes based upon the MVAPICH OSU Micro-Benchmarks.
+   * The sender process sends a message and then waits for confirmation of reception.
+   * Many iterations occur with various message sizes and the average latency values
+   * are returned within a map. Utilizes blocking send and recieve.
+   *
+   * See further: https://mvapich.cse.ohio-state.edu/benchmarks/
+   */
+  void pingpong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Teuchos::Comm<int> > &comm);
+  bool has_pingpong_table() const { return pingpong_sizes_.size() > 0; }
+
+  /* Lookup in the pingpong_vector table */
+  double pingpong_host_lookup(int SIZE_IN_BYTES);
+  double pingpong_device_lookup(int SIZE_IN_BYTES);
+
+  /* Print table */
+  void print_pingpong_table(std::ostream &out, const std::string &prefix = "");
+
+  /* A halo-exchange based ping-pong, inspired by halo-mode in MPPTEST from ANL.
+   * Here we use exactly the communication pattern specified in the import object
+   * and send messages accordingly.  We vary the size in bytes sent per message,
+   * which should capture max-rate effects to some degree.
+   *
+   * See further: https://www.mcs.anl.gov/research/projects/mpi/mpptest/
+   */
+  void halopong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > &import);
+  bool has_halopong_table() const { return halopong_sizes_.size() > 0; }
+
+  /* Lookup in the halopong_vector table */
+  double halopong_host_lookup(int SIZE_IN_BYTES_PER_MESSAGE);
+  double halopong_device_lookup(int SIZE_IN_BYTES_PER_MESSAGE);
+
+  /* Print table */
+  void print_halopong_table(std::ostream &out, const std::string &prefix = "");
+
+  /* Estimate launch latency based on the cost of submitting an empty Kokkos::parallel_for.
+   * This necessary to correct the memory bandwidth costs for models on high latency platforms,
+   * e.g., GPUS.
+   */
+  void launch_latency_make_table(int KERNEL_REPEATS);
+  bool has_launch_latency_table() const { return launch_and_wait_latency_ > 0; }
+
+  /* Lookup launch latency */
+  double launch_latency_lookup();
+
+  /* Print table */
+  void print_launch_latency_table(std::ostream &out, const std::string &prefix = "");
+
+ private:
+  void print_stream_vector_table_impl(std::ostream &out, bool use_latency_correction, const std::string &prefix);
+
+  std::vector<int> stream_sizes_;
+  std::vector<double> stream_copy_times_;
+  std::vector<double> stream_add_times_;
+  std::vector<double> latency_corrected_stream_copy_times_;
+  std::vector<double> latency_corrected_stream_add_times_;
+
+  std::vector<int> pingpong_sizes_;
+  std::vector<double> pingpong_host_times_;
+  std::vector<double> pingpong_device_times_;
+
+  std::vector<int> halopong_sizes_;
+  std::vector<double> halopong_host_times_;
+  std::vector<double> halopong_device_times_;
+
+  double launch_and_wait_latency_;
+
+};  // class PerfModels
+
+}  // namespace MueLu
+
+#endif  // ifndef MUELU_PERFMODELS_HPP
diff --git a/packages/muelu/src/Utils/MueLu_PerfModels_def.hpp b/packages/muelu/src/Utils/MueLu_PerfModels_def.hpp
index 9a58ca3cb9c9..1ad77f817296 100644
--- a/packages/muelu/src/Utils/MueLu_PerfModels_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_PerfModels_def.hpp
@@ -61,596 +61,551 @@
 #include <mpi.h>
 #endif
 
-
-
 #ifdef HAVE_MPI
 #include <mpi.h>
 #endif
 
 namespace MueLu {
 
-  namespace PerfDetails {
-    template<class Scalar,class Node>
-    double stream_vector_add(int KERNEL_REPEATS, int VECTOR_SIZE) {      
-      // PerfDetails' STREAM routines need to be instantiatiated on impl_scalar_type, not Scalar
-      using impl_scalar_type  = typename Kokkos::ArithTraits<Scalar>::val_type;
-      
-      using exec_space   = typename Node::execution_space;
-      using memory_space = typename Node::memory_space;
-      using range_policy = Kokkos::RangePolicy<exec_space>;
-
-      Kokkos::View<impl_scalar_type*,memory_space> a("a", VECTOR_SIZE);
-      Kokkos::View<impl_scalar_type*,memory_space> b("b", VECTOR_SIZE);
-      Kokkos::View<impl_scalar_type*,memory_space> c("c", VECTOR_SIZE);
-      double total_test_time = 0.0;
-
-      impl_scalar_type ONE = Teuchos::ScalarTraits<impl_scalar_type>::one();
-
-      Kokkos::parallel_for("stream/fill",range_policy(0,VECTOR_SIZE), KOKKOS_LAMBDA (const size_t i) {
-          a(i) = ONE * (double)i;
-          b(i) = a(i);
-        });
-      exec_space().fence();
+namespace PerfDetails {
+template <class Scalar, class Node>
+double stream_vector_add(int KERNEL_REPEATS, int VECTOR_SIZE) {
+  // PerfDetails' STREAM routines need to be instantiatiated on impl_scalar_type, not Scalar
+  using impl_scalar_type = typename Kokkos::ArithTraits<Scalar>::val_type;
+
+  using exec_space   = typename Node::execution_space;
+  using memory_space = typename Node::memory_space;
+  using range_policy = Kokkos::RangePolicy<exec_space>;
+
+  Kokkos::View<impl_scalar_type *, memory_space> a("a", VECTOR_SIZE);
+  Kokkos::View<impl_scalar_type *, memory_space> b("b", VECTOR_SIZE);
+  Kokkos::View<impl_scalar_type *, memory_space> c("c", VECTOR_SIZE);
+  double total_test_time = 0.0;
+
+  impl_scalar_type ONE = Teuchos::ScalarTraits<impl_scalar_type>::one();
 
-      using clock = std::chrono::high_resolution_clock;
+  Kokkos::parallel_for(
+      "stream/fill", range_policy(0, VECTOR_SIZE), KOKKOS_LAMBDA(const size_t i) {
+        a(i) = ONE * (double)i;
+        b(i) = a(i);
+      });
+  exec_space().fence();
 
-      clock::time_point start, stop;
+  using clock = std::chrono::high_resolution_clock;
 
-      for(int i = 0; i < KERNEL_REPEATS; i++) {       
-        start = clock::now();
-        Kokkos::parallel_for("stream/add",range_policy(0,VECTOR_SIZE), KOKKOS_LAMBDA (const size_t j) { //Vector Addition
-            c(j) = a(j) + b(j);
+  clock::time_point start, stop;
+
+  for (int i = 0; i < KERNEL_REPEATS; i++) {
+    start = clock::now();
+    Kokkos::parallel_for(
+        "stream/add", range_policy(0, VECTOR_SIZE), KOKKOS_LAMBDA(const size_t j) {  // Vector Addition
+          c(j) = a(j) + b(j);
         });
 
-        exec_space().fence();
-        stop = clock::now();
-        double my_test_time = std::chrono::duration<double>(stop - start).count();
-        total_test_time += my_test_time;
-      }
+    exec_space().fence();
+    stop                = clock::now();
+    double my_test_time = std::chrono::duration<double>(stop - start).count();
+    total_test_time += my_test_time;
+  }
 
-      return total_test_time / KERNEL_REPEATS;
-    }
+  return total_test_time / KERNEL_REPEATS;
+}
 
-    template<class Scalar,class Node>
-    double stream_vector_copy(int KERNEL_REPEATS, int VECTOR_SIZE) {      
-      // PerfDetails' STREAM routines need to be instantiatiated on impl_scalar_type, not Scalar
-      using impl_scalar_type  = typename Kokkos::ArithTraits<Scalar>::val_type;
+template <class Scalar, class Node>
+double stream_vector_copy(int KERNEL_REPEATS, int VECTOR_SIZE) {
+  // PerfDetails' STREAM routines need to be instantiatiated on impl_scalar_type, not Scalar
+  using impl_scalar_type = typename Kokkos::ArithTraits<Scalar>::val_type;
 
-      using exec_space   = typename Node::execution_space;
-      using memory_space = typename Node::memory_space;
-      using range_policy = Kokkos::RangePolicy<exec_space>;
+  using exec_space   = typename Node::execution_space;
+  using memory_space = typename Node::memory_space;
+  using range_policy = Kokkos::RangePolicy<exec_space>;
 
-      Kokkos::View<impl_scalar_type*,memory_space> a("a", VECTOR_SIZE);
-      Kokkos::View<impl_scalar_type*,memory_space> b("b", VECTOR_SIZE);
-      double total_test_time = 0.0;
+  Kokkos::View<impl_scalar_type *, memory_space> a("a", VECTOR_SIZE);
+  Kokkos::View<impl_scalar_type *, memory_space> b("b", VECTOR_SIZE);
+  double total_test_time = 0.0;
 
-      impl_scalar_type ONE = Teuchos::ScalarTraits<impl_scalar_type>::one();
+  impl_scalar_type ONE = Teuchos::ScalarTraits<impl_scalar_type>::one();
 
-      Kokkos::parallel_for("stream/fill",range_policy(0,VECTOR_SIZE), KOKKOS_LAMBDA (const size_t i) {
-          a(i) = ONE;
-        });
-      exec_space().fence();
+  Kokkos::parallel_for(
+      "stream/fill", range_policy(0, VECTOR_SIZE), KOKKOS_LAMBDA(const size_t i) {
+        a(i) = ONE;
+      });
+  exec_space().fence();
 
-      using clock = std::chrono::high_resolution_clock;
-      clock::time_point start, stop;
+  using clock = std::chrono::high_resolution_clock;
+  clock::time_point start, stop;
 
-      for(int i = 0; i < KERNEL_REPEATS; i++) {       
-        start = clock::now();
-        Kokkos::parallel_for("stream/copy",range_policy(0,VECTOR_SIZE), KOKKOS_LAMBDA (const size_t j) { //Vector Addition
-            b(j) = a(j);
+  for (int i = 0; i < KERNEL_REPEATS; i++) {
+    start = clock::now();
+    Kokkos::parallel_for(
+        "stream/copy", range_policy(0, VECTOR_SIZE), KOKKOS_LAMBDA(const size_t j) {  // Vector Addition
+          b(j) = a(j);
         });
 
-        exec_space().fence();
-        stop = clock::now();
-        double my_test_time = std::chrono::duration<double>(stop - start).count();
-        total_test_time += my_test_time;
-      }
+    exec_space().fence();
+    stop                = clock::now();
+    double my_test_time = std::chrono::duration<double>(stop - start).count();
+    total_test_time += my_test_time;
+  }
 
-      return total_test_time / KERNEL_REPEATS;
-    }
+  return total_test_time / KERNEL_REPEATS;
+}
 
+double table_lookup(const std::vector<int> &x, const std::vector<double> &y, int value) {
+  // If there's no table, nan
+  if (x.size() == 0) return Teuchos::ScalarTraits<double>::nan();
 
- 
-    double table_lookup(const std::vector<int> & x, const std::vector<double> & y, int value) {
-      // If there's no table, nan
-      if(x.size() == 0) return Teuchos::ScalarTraits<double>::nan();
-      
-      // NOTE:  This should probably be a binary search, but this isn't performance sensitive, so we'll go simple
-      int N = (int) x.size();
-      int hi = 0;
-      for(  ; hi < N; hi++) {
-        if (x[hi] > value) 
-          break;
-      }
-      
-      if(hi == 0) {
-        // Lower end (return the min time)
-        //printf("Lower end: %d < %d\n",value,x[0]);
-        return y[0];
-      }
-      else if (hi == N) {
-        // Higher end (extrapolate from the last two points)
-        //printf("Upper end: %d > %d\n",value,x[N-1]);
-        hi = N-1;
-        int run     = x[hi] - x[hi-1];
-        double rise = y[hi] - y[hi-1];
-        double slope = rise / run;     
-        int diff = value - x[hi-1];
-        
-        return y[hi-1] + slope * diff;
-      }
-      else {
-        // Interpolate
-        //printf("Middle: %d < %d < %d\n",x[hi-1],value,x[hi]);
-        int run     = x[hi] - x[hi-1];
-        double rise = y[hi] - y[hi-1];
-        double slope = rise / run;     
-        int diff = value - x[hi-1];
-        
-        return y[hi-1] + slope * diff;
-      }
-    }
+  // NOTE:  This should probably be a binary search, but this isn't performance sensitive, so we'll go simple
+  int N  = (int)x.size();
+  int hi = 0;
+  for (; hi < N; hi++) {
+    if (x[hi] > value)
+      break;
+  }
 
-    // Report bandwidth in GB / sec
-    const double GB = 1024.0 * 1024.0 * 1024.0;
-    double convert_time_to_bandwidth_gbs(double time, int num_calls, double memory_per_call_bytes) {
-      double time_per_call = time / num_calls;      
-      return memory_per_call_bytes / GB / time_per_call;
-    }
+  if (hi == 0) {
+    // Lower end (return the min time)
+    // printf("Lower end: %d < %d\n",value,x[0]);
+    return y[0];
+  } else if (hi == N) {
+    // Higher end (extrapolate from the last two points)
+    // printf("Upper end: %d > %d\n",value,x[N-1]);
+    hi           = N - 1;
+    int run      = x[hi] - x[hi - 1];
+    double rise  = y[hi] - y[hi - 1];
+    double slope = rise / run;
+    int diff     = value - x[hi - 1];
+
+    return y[hi - 1] + slope * diff;
+  } else {
+    // Interpolate
+    // printf("Middle: %d < %d < %d\n",x[hi-1],value,x[hi]);
+    int run      = x[hi] - x[hi - 1];
+    double rise  = y[hi] - y[hi - 1];
+    double slope = rise / run;
+    int diff     = value - x[hi - 1];
+
+    return y[hi - 1] + slope * diff;
+  }
+}
 
+// Report bandwidth in GB / sec
+const double GB = 1024.0 * 1024.0 * 1024.0;
+double convert_time_to_bandwidth_gbs(double time, int num_calls, double memory_per_call_bytes) {
+  double time_per_call = time / num_calls;
+  return memory_per_call_bytes / GB / time_per_call;
+}
 
-    template <class exec_space, class memory_space>
-    void pingpong_basic(int KERNEL_REPEATS, int MAX_SIZE,const Teuchos::Comm<int> &comm, std::vector<int> & sizes, std::vector<double> & times) {
+template <class exec_space, class memory_space>
+void pingpong_basic(int KERNEL_REPEATS, int MAX_SIZE, const Teuchos::Comm<int> &comm, std::vector<int> &sizes, std::vector<double> &times) {
 #ifdef HAVE_MPI
-      int rank = comm.getRank();
-      int nproc = comm.getSize();
-
-      if(nproc < 2) return;
-
-      const int buff_size = (int) pow(2,MAX_SIZE);
-      
-      sizes.resize(MAX_SIZE+1);
-      times.resize(MAX_SIZE+1);
-            
-      // Allocate memory for the buffers (and fill send)
-      Kokkos::View<char*,memory_space> r_buf("recv",buff_size), s_buf("send",buff_size);
-      Kokkos::deep_copy(s_buf,1);
-      
-      //Send and recieve.   
-      // NOTE:  Do consectutive pair buddies here for simplicity.  We should be smart later
-      int odd = rank % 2;
-      int buddy = odd ? rank - 1 : rank + 1;
-      
-      for(int i = 0; i < MAX_SIZE + 1 ;i ++) {
-        int msg_size = (int) pow(2,i);
-        comm.barrier();
-
-        double t0 = MPI_Wtime();      
-        for(int j = 0; j < KERNEL_REPEATS; j++) {
-          if (buddy < nproc) {
-            if (odd) {
-              comm.send(msg_size, (char*)s_buf.data(), buddy);
-              comm.receive(buddy, msg_size, (char*)r_buf.data());
-            }
-            else {
-              comm.receive(buddy, msg_size,(char*)r_buf.data());
-              comm.send(msg_size, (char*)s_buf.data(), buddy);
-            }
-          }
+  int rank  = comm.getRank();
+  int nproc = comm.getSize();
+
+  if (nproc < 2) return;
+
+  const int buff_size = (int)pow(2, MAX_SIZE);
+
+  sizes.resize(MAX_SIZE + 1);
+  times.resize(MAX_SIZE + 1);
+
+  // Allocate memory for the buffers (and fill send)
+  Kokkos::View<char *, memory_space> r_buf("recv", buff_size), s_buf("send", buff_size);
+  Kokkos::deep_copy(s_buf, 1);
+
+  // Send and recieve.
+  //  NOTE:  Do consectutive pair buddies here for simplicity.  We should be smart later
+  int odd   = rank % 2;
+  int buddy = odd ? rank - 1 : rank + 1;
+
+  for (int i = 0; i < MAX_SIZE + 1; i++) {
+    int msg_size = (int)pow(2, i);
+    comm.barrier();
+
+    double t0 = MPI_Wtime();
+    for (int j = 0; j < KERNEL_REPEATS; j++) {
+      if (buddy < nproc) {
+        if (odd) {
+          comm.send(msg_size, (char *)s_buf.data(), buddy);
+          comm.receive(buddy, msg_size, (char *)r_buf.data());
+        } else {
+          comm.receive(buddy, msg_size, (char *)r_buf.data());
+          comm.send(msg_size, (char *)s_buf.data(), buddy);
         }
-
-        double time_per_call = (MPI_Wtime() - t0) / (2.0 * KERNEL_REPEATS);
-        sizes[i] = msg_size;
-        times[i] = time_per_call;
       }
-#else
-      return;
-#endif
     }
 
+    double time_per_call = (MPI_Wtime() - t0) / (2.0 * KERNEL_REPEATS);
+    sizes[i]             = msg_size;
+    times[i]             = time_per_call;
+  }
+#else
+  return;
+#endif
+}
 
-    template <class exec_space, class memory_space, class LocalOrdinal, class GlobalOrdinal, class Node>
-    void halopong_basic(int KERNEL_REPEATS, int MAX_SIZE,const RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > & import, std::vector<int> & sizes, std::vector<double> & times) {
-      int nproc = import->getSourceMap()->getComm()->getSize();      
-      if(nproc < 2) return;     
+template <class exec_space, class memory_space, class LocalOrdinal, class GlobalOrdinal, class Node>
+void halopong_basic(int KERNEL_REPEATS, int MAX_SIZE, const RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > &import, std::vector<int> &sizes, std::vector<double> &times) {
+  int nproc = import->getSourceMap()->getComm()->getSize();
+  if (nproc < 2) return;
 #if defined(HAVE_MUELU_TPETRA) && defined(HAVE_MPI)
-      // NOTE: We need to get the distributer here, which means we need Tpetra, since Xpetra does 
-      // not have a distributor interface
-      using x_import_type = Xpetra::TpetraImport<LocalOrdinal,GlobalOrdinal,Node>;
-      RCP<const x_import_type> Ximport = Teuchos::rcp_dynamic_cast<const x_import_type>(import);
-      RCP<const Teuchos::MpiComm<int> > mcomm = Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(import->getSourceMap()->getComm());
-      MPI_Comm communicator = *mcomm->getRawMpiComm();
-
-      if(Ximport.is_null() || mcomm.is_null()) return;
-      auto Timport = Ximport->getTpetra_Import();
-      auto distor = Timport->getDistributor();
-
-      // Distributor innards
-      Teuchos::ArrayView<const int> procsFrom = distor.getProcsFrom();
-      Teuchos::ArrayView<const int> procsTo   = distor.getProcsTo();      
-      int num_recvs = (int)distor.getNumReceives();
-      int num_sends = (int)distor.getNumSends();
-            
-
-      const int buff_size_per_msg = (int) pow(2,MAX_SIZE);      
-      sizes.resize(MAX_SIZE+1);
-      times.resize(MAX_SIZE+1);
-      
-      // Allocate memory for the buffers (and fill send)
-      Kokkos::View<char*,memory_space> f_recv_buf("forward_recv",buff_size_per_msg*num_recvs), f_send_buf("forward_send",buff_size_per_msg*num_sends);
-      Kokkos::View<char*,memory_space> r_recv_buf("reverse_recv",buff_size_per_msg*num_sends), r_send_buf("reverse_send",buff_size_per_msg*num_recvs);
-      Kokkos::deep_copy(f_send_buf,1);
-      Kokkos::deep_copy(r_send_buf,1);
-      
-      std::vector<MPI_Request> requests(num_sends+num_recvs);
-      std::vector<MPI_Status>  status(num_sends+num_recvs);
-
-
-      for(int i = 0; i < MAX_SIZE + 1 ;i ++) {
-        int msg_size = (int) pow(2,i);
-
-        MPI_Barrier(communicator);
-        
-        double t0 = MPI_Wtime();      
-        for(int j = 0; j < KERNEL_REPEATS; j++) {
-          int ct=0;
-          // Recv/Send the forward messsages
-          for(int r=0; r<num_recvs;r++) {
-            const int tag = 1000+j;
-            MPI_Irecv(f_recv_buf.data()+msg_size*r,msg_size,MPI_CHAR,procsFrom[r],tag,communicator,&requests[ct]);
-            ct++;
-          }
-          for(int s=0; s<num_sends;s++) {
-            const int tag = 1000+j;
-            MPI_Isend(f_send_buf.data()+msg_size*s,msg_size,MPI_CHAR,procsTo[s],tag,communicator,&requests[ct]);
-            ct++;
-          }
-          // Wait for the forward messsages
-          MPI_Waitall(ct,requests.data(),status.data());
-
-          ct=0;
-          // Recv/Send the reverse messsages
-          for(int r=0; r<num_sends;r++) {
-            const int tag = 2000+j;
-            MPI_Irecv(r_recv_buf.data()+msg_size*r,msg_size,MPI_CHAR,procsTo[r],tag,communicator,&requests[ct]);
-            ct++;
-          }
-          for(int s=0; s<num_recvs;s++) {
-            const int tag = 2000+j;
-            MPI_Isend(r_send_buf.data()+msg_size*s,msg_size,MPI_CHAR,procsFrom[s],tag,communicator,&requests[ct]);
-            ct++;
-          }
-          // Wait for the reverse messsages
-          MPI_Waitall(ct,requests.data(),status.data());
-        }
-
-        double time_per_call = (MPI_Wtime() - t0) / (2.0 * KERNEL_REPEATS);
-        sizes[i] = msg_size;
-        times[i] = time_per_call;
+  // NOTE: We need to get the distributer here, which means we need Tpetra, since Xpetra does
+  // not have a distributor interface
+  using x_import_type                     = Xpetra::TpetraImport<LocalOrdinal, GlobalOrdinal, Node>;
+  RCP<const x_import_type> Ximport        = Teuchos::rcp_dynamic_cast<const x_import_type>(import);
+  RCP<const Teuchos::MpiComm<int> > mcomm = Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(import->getSourceMap()->getComm());
+  MPI_Comm communicator                   = *mcomm->getRawMpiComm();
+
+  if (Ximport.is_null() || mcomm.is_null()) return;
+  auto Timport = Ximport->getTpetra_Import();
+  auto distor  = Timport->getDistributor();
+
+  // Distributor innards
+  Teuchos::ArrayView<const int> procsFrom = distor.getProcsFrom();
+  Teuchos::ArrayView<const int> procsTo   = distor.getProcsTo();
+  int num_recvs                           = (int)distor.getNumReceives();
+  int num_sends                           = (int)distor.getNumSends();
+
+  const int buff_size_per_msg = (int)pow(2, MAX_SIZE);
+  sizes.resize(MAX_SIZE + 1);
+  times.resize(MAX_SIZE + 1);
+
+  // Allocate memory for the buffers (and fill send)
+  Kokkos::View<char *, memory_space> f_recv_buf("forward_recv", buff_size_per_msg * num_recvs), f_send_buf("forward_send", buff_size_per_msg * num_sends);
+  Kokkos::View<char *, memory_space> r_recv_buf("reverse_recv", buff_size_per_msg * num_sends), r_send_buf("reverse_send", buff_size_per_msg * num_recvs);
+  Kokkos::deep_copy(f_send_buf, 1);
+  Kokkos::deep_copy(r_send_buf, 1);
+
+  std::vector<MPI_Request> requests(num_sends + num_recvs);
+  std::vector<MPI_Status> status(num_sends + num_recvs);
+
+  for (int i = 0; i < MAX_SIZE + 1; i++) {
+    int msg_size = (int)pow(2, i);
+
+    MPI_Barrier(communicator);
+
+    double t0 = MPI_Wtime();
+    for (int j = 0; j < KERNEL_REPEATS; j++) {
+      int ct = 0;
+      // Recv/Send the forward messsages
+      for (int r = 0; r < num_recvs; r++) {
+        const int tag = 1000 + j;
+        MPI_Irecv(f_recv_buf.data() + msg_size * r, msg_size, MPI_CHAR, procsFrom[r], tag, communicator, &requests[ct]);
+        ct++;
       }
-
-  
-
-#endif
-
+      for (int s = 0; s < num_sends; s++) {
+        const int tag = 1000 + j;
+        MPI_Isend(f_send_buf.data() + msg_size * s, msg_size, MPI_CHAR, procsTo[s], tag, communicator, &requests[ct]);
+        ct++;
+      }
+      // Wait for the forward messsages
+      MPI_Waitall(ct, requests.data(), status.data());
+
+      ct = 0;
+      // Recv/Send the reverse messsages
+      for (int r = 0; r < num_sends; r++) {
+        const int tag = 2000 + j;
+        MPI_Irecv(r_recv_buf.data() + msg_size * r, msg_size, MPI_CHAR, procsTo[r], tag, communicator, &requests[ct]);
+        ct++;
+      }
+      for (int s = 0; s < num_recvs; s++) {
+        const int tag = 2000 + j;
+        MPI_Isend(r_send_buf.data() + msg_size * s, msg_size, MPI_CHAR, procsFrom[s], tag, communicator, &requests[ct]);
+        ct++;
+      }
+      // Wait for the reverse messsages
+      MPI_Waitall(ct, requests.data(), status.data());
     }
 
-
-  }// end namespace PerfDetails
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PerfModels():launch_and_wait_latency_(-1.0){}
-
-  /****************************************************************************************/
-  /****************************************************************************************/
-  /****************************************************************************************/
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void 
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_make_table(int KERNEL_REPEATS,  int LOG_MAX_SIZE) {
-
-    // We need launch/waits latency estimates for corrected stream
-    launch_latency_make_table(KERNEL_REPEATS);
-    double latency = launch_latency_lookup();
-    
-    if(LOG_MAX_SIZE < 2)
-      LOG_MAX_SIZE=20;
-
-    stream_sizes_.resize(LOG_MAX_SIZE+1);    
-    stream_copy_times_.resize(LOG_MAX_SIZE+1);    
-    stream_add_times_.resize(LOG_MAX_SIZE+1);    
-    latency_corrected_stream_copy_times_.resize(LOG_MAX_SIZE+1);    
-    latency_corrected_stream_add_times_.resize(LOG_MAX_SIZE+1);    
-        
-    for(int i=0; i<LOG_MAX_SIZE+1; i++) {
-      int size = (int) pow(2,i);
-      double c_time = PerfDetails::stream_vector_copy<Scalar,Node>(KERNEL_REPEATS,size);
-      double a_time = PerfDetails::stream_vector_add<Scalar,Node>(KERNEL_REPEATS,size);
-
-      stream_sizes_[i] = size;
-
-      // Correct for the difference in memory transactions per element
-      stream_copy_times_[i] = c_time / 2.0;
-      stream_add_times_[i] = a_time / 3.0;
-
-      // Correct for launch latency too.  We'll note that sometimes the latency estimate
-      // is higher than the actual copy/add time estimate.  If so, we don't correct
-      latency_corrected_stream_copy_times_[i] = (c_time - latency <= 0.0) ? c_time / 2.0 : ( (c_time-latency)/2.0 );
-      latency_corrected_stream_add_times_[i]  = (a_time - latency <= 0.0) ? a_time / 3.0 : ( (a_time-latency)/3.0 );
-
-
-    }
+    double time_per_call = (MPI_Wtime() - t0) / (2.0 * KERNEL_REPEATS);
+    sizes[i]             = msg_size;
+    times[i]             = time_per_call;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_copy_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(stream_sizes_,stream_copy_times_,SIZE_IN_BYTES/sizeof(Scalar));
-  }
+#endif
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_add_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(stream_sizes_,stream_add_times_,SIZE_IN_BYTES/sizeof(Scalar));
-  }
+}  // end namespace PerfDetails
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_lookup(int SIZE_IN_BYTES) {
-    return std::min(stream_vector_copy_lookup(SIZE_IN_BYTES),stream_vector_add_lookup(SIZE_IN_BYTES));
-  }
-  
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PerfModels()
+  : launch_and_wait_latency_(-1.0) {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_copy_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(stream_sizes_,latency_corrected_stream_copy_times_,SIZE_IN_BYTES/sizeof(Scalar));
-  }
+/****************************************************************************************/
+/****************************************************************************************/
+/****************************************************************************************/
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_add_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(stream_sizes_,latency_corrected_stream_add_times_,SIZE_IN_BYTES/sizeof(Scalar));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE) {
+  // We need launch/waits latency estimates for corrected stream
+  launch_latency_make_table(KERNEL_REPEATS);
+  double latency = launch_latency_lookup();
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_lookup(int SIZE_IN_BYTES) {
-    return std::min(latency_corrected_stream_vector_copy_lookup(SIZE_IN_BYTES),latency_corrected_stream_vector_add_lookup(SIZE_IN_BYTES));
-  }
- 
+  if (LOG_MAX_SIZE < 2)
+    LOG_MAX_SIZE = 20;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_stream_vector_table(std::ostream & out, const std::string & prefix) {
-    print_stream_vector_table_impl(out,false,prefix);
-  }
+  stream_sizes_.resize(LOG_MAX_SIZE + 1);
+  stream_copy_times_.resize(LOG_MAX_SIZE + 1);
+  stream_add_times_.resize(LOG_MAX_SIZE + 1);
+  latency_corrected_stream_copy_times_.resize(LOG_MAX_SIZE + 1);
+  latency_corrected_stream_add_times_.resize(LOG_MAX_SIZE + 1);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_latency_corrected_stream_vector_table(std::ostream & out, const std::string & prefix) {
-    print_stream_vector_table_impl(out,true,prefix);
-  }
+  for (int i = 0; i < LOG_MAX_SIZE + 1; i++) {
+    int size      = (int)pow(2, i);
+    double c_time = PerfDetails::stream_vector_copy<Scalar, Node>(KERNEL_REPEATS, size);
+    double a_time = PerfDetails::stream_vector_add<Scalar, Node>(KERNEL_REPEATS, size);
 
+    stream_sizes_[i] = size;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_stream_vector_table_impl(std::ostream & out,bool use_latency_correction, const std::string & prefix) {
-    using namespace std;
-    std::ios old_format(NULL);
-    old_format.copyfmt(out);
+    // Correct for the difference in memory transactions per element
+    stream_copy_times_[i] = c_time / 2.0;
+    stream_add_times_[i]  = a_time / 3.0;
 
-    out << prefix
-        << setw(20) << "Length in Scalars" << setw(1) << " "
-        << setw(20) << "COPY (us)" << setw(1) << " "
-        << setw(20) << "ADD (us)" << setw(1) << " "
-        << setw(20) << "COPY (GB/s)" << setw(1) << " "
-        << setw(20) << "ADD (GB/s)" << std::endl;
+    // Correct for launch latency too.  We'll note that sometimes the latency estimate
+    // is higher than the actual copy/add time estimate.  If so, we don't correct
+    latency_corrected_stream_copy_times_[i] = (c_time - latency <= 0.0) ? c_time / 2.0 : ((c_time - latency) / 2.0);
+    latency_corrected_stream_add_times_[i]  = (a_time - latency <= 0.0) ? a_time / 3.0 : ((a_time - latency) / 3.0);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_copy_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(stream_sizes_, stream_copy_times_, SIZE_IN_BYTES / sizeof(Scalar));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_add_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(stream_sizes_, stream_add_times_, SIZE_IN_BYTES / sizeof(Scalar));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::stream_vector_lookup(int SIZE_IN_BYTES) {
+  return std::min(stream_vector_copy_lookup(SIZE_IN_BYTES), stream_vector_add_lookup(SIZE_IN_BYTES));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_copy_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(stream_sizes_, latency_corrected_stream_copy_times_, SIZE_IN_BYTES / sizeof(Scalar));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_add_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(stream_sizes_, latency_corrected_stream_add_times_, SIZE_IN_BYTES / sizeof(Scalar));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::latency_corrected_stream_vector_lookup(int SIZE_IN_BYTES) {
+  return std::min(latency_corrected_stream_vector_copy_lookup(SIZE_IN_BYTES), latency_corrected_stream_vector_add_lookup(SIZE_IN_BYTES));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_stream_vector_table(std::ostream &out, const std::string &prefix) {
+  print_stream_vector_table_impl(out, false, prefix);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_latency_corrected_stream_vector_table(std::ostream &out, const std::string &prefix) {
+  print_stream_vector_table_impl(out, true, prefix);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_stream_vector_table_impl(std::ostream &out, bool use_latency_correction, const std::string &prefix) {
+  using namespace std;
+  std::ios old_format(NULL);
+  old_format.copyfmt(out);
+
+  out << prefix
+      << setw(20) << "Length in Scalars" << setw(1) << " "
+      << setw(20) << "COPY (us)" << setw(1) << " "
+      << setw(20) << "ADD (us)" << setw(1) << " "
+      << setw(20) << "COPY (GB/s)" << setw(1) << " "
+      << setw(20) << "ADD (GB/s)" << std::endl;
+
+  out << prefix
+      << setw(20) << "-----------------" << setw(1) << " "
+      << setw(20) << "---------" << setw(1) << " "
+      << setw(20) << "--------" << setw(1) << " "
+      << setw(20) << "-----------" << setw(1) << " "
+      << setw(20) << "----------" << std::endl;
+
+  for (int i = 0; i < (int)stream_sizes_.size(); i++) {
+    int size      = stream_sizes_[i];
+    double c_time = use_latency_correction ? latency_corrected_stream_copy_times_[i] : stream_copy_times_[i];
+    double a_time = use_latency_correction ? latency_corrected_stream_add_times_[i] : stream_add_times_[i];
+    // We've already corrected for the transactions per element difference
+    double c_bw = PerfDetails::convert_time_to_bandwidth_gbs(c_time, 1, size * sizeof(Scalar));
+    double a_bw = PerfDetails::convert_time_to_bandwidth_gbs(a_time, 1, size * sizeof(Scalar));
 
     out << prefix
-        << setw(20) << "-----------------" << setw(1) << " "
-        << setw(20) << "---------" << setw(1) << " "
-        << setw(20) << "--------" << setw(1) << " "
-        << setw(20) << "-----------" << setw(1) << " "
-        << setw(20) << "----------" << std::endl;
-    
-
-    for(int i=0; i<(int)stream_sizes_.size(); i++) {
-      int size = stream_sizes_[i];
-      double c_time = use_latency_correction ? latency_corrected_stream_copy_times_[i] : stream_copy_times_[i];
-      double a_time = use_latency_correction ? latency_corrected_stream_add_times_[i]  : stream_add_times_[i];
-      // We've already corrected for the transactions per element difference
-      double c_bw = PerfDetails::convert_time_to_bandwidth_gbs(c_time,1,size*sizeof(Scalar));
-      double a_bw = PerfDetails::convert_time_to_bandwidth_gbs(a_time,1,size*sizeof(Scalar));
-
-
-      out << prefix
-          << setw(20) << size << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (c_time*1e6) << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (a_time*1e6) << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << c_bw << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << a_bw << std::endl;        
-    }
-
-    out.copyfmt(old_format);
+        << setw(20) << size << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (c_time * 1e6) << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (a_time * 1e6) << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << c_bw << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << a_bw << std::endl;
   }
 
+  out.copyfmt(old_format);
+}
 
-  /****************************************************************************************/
-  /****************************************************************************************/
-  /****************************************************************************************/
+/****************************************************************************************/
+/****************************************************************************************/
+/****************************************************************************************/
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
- PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Teuchos::Comm<int> > &comm) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Teuchos::Comm<int> > &comm) {
+  PerfDetails::pingpong_basic<Kokkos::HostSpace::execution_space, Kokkos::HostSpace::memory_space>(KERNEL_REPEATS, LOG_MAX_SIZE, *comm, pingpong_sizes_, pingpong_host_times_);
 
-    PerfDetails::pingpong_basic<Kokkos::HostSpace::execution_space,Kokkos::HostSpace::memory_space>(KERNEL_REPEATS,LOG_MAX_SIZE,*comm,pingpong_sizes_,pingpong_host_times_);
+  PerfDetails::pingpong_basic<typename Node::execution_space, typename Node::memory_space>(KERNEL_REPEATS, LOG_MAX_SIZE, *comm, pingpong_sizes_, pingpong_device_times_);
+}
 
-    PerfDetails::pingpong_basic<typename Node::execution_space,typename Node::memory_space>(KERNEL_REPEATS,LOG_MAX_SIZE,*comm,pingpong_sizes_,pingpong_device_times_);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_host_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(pingpong_sizes_, pingpong_host_times_, SIZE_IN_BYTES);
+}
 
- }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_device_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(pingpong_sizes_, pingpong_device_times_, SIZE_IN_BYTES);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_host_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(pingpong_sizes_,pingpong_host_times_,SIZE_IN_BYTES);
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pingpong_device_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(pingpong_sizes_,pingpong_device_times_,SIZE_IN_BYTES);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_pingpong_table(std::ostream &out, const std::string &prefix) {
+  if (pingpong_sizes_.size() == 0) return;
 
+  using namespace std;
+  std::ios old_format(NULL);
+  old_format.copyfmt(out);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_pingpong_table(std::ostream & out, const std::string & prefix) {
-    if(pingpong_sizes_.size() == 0) return;
+  out << prefix
+      << setw(20) << "Message Size" << setw(1) << " "
+      << setw(20) << "Host (us)" << setw(1) << " "
+      << setw(20) << "Device (us)" << std::endl;
 
-    using namespace std;
-    std::ios old_format(NULL);
-    old_format.copyfmt(out);
+  out << prefix
+      << setw(20) << "------------" << setw(1) << " "
+      << setw(20) << "---------" << setw(1) << " "
+      << setw(20) << "-----------" << std::endl;
 
-    out << prefix
-        << setw(20) << "Message Size" << setw(1) << " "
-        << setw(20) << "Host (us)" << setw(1) << " "
-        << setw(20) << "Device (us)" << std::endl;
+  for (int i = 0; i < (int)pingpong_sizes_.size(); i++) {
+    int size      = pingpong_sizes_[i];
+    double h_time = pingpong_host_times_[i];
+    double d_time = pingpong_device_times_[i];
 
     out << prefix
-        << setw(20) << "------------" << setw(1) << " "
-        << setw(20) << "---------" << setw(1) << " "
-        << setw(20) << "-----------" << std::endl;
-    
-
-    for(int i=0; i<(int)pingpong_sizes_.size(); i++) {
-      int size = pingpong_sizes_[i];
-      double h_time = pingpong_host_times_[i];
-      double d_time = pingpong_device_times_[i];
-
-
-      out << prefix
-          << setw(20) << size << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (h_time*1e6) << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (d_time*1e6) << setw(1) << std::endl;
-    }
-
-    out.copyfmt(old_format);
-  }
-
-
-
-  /****************************************************************************************/
-  /****************************************************************************************/
-  /****************************************************************************************/
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE,  const RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > & import) {
-
-    PerfDetails::halopong_basic<Kokkos::HostSpace::execution_space,Kokkos::HostSpace::memory_space>(KERNEL_REPEATS,LOG_MAX_SIZE,import,halopong_sizes_,halopong_host_times_);
-
-    PerfDetails::halopong_basic<typename Node::execution_space,typename Node::memory_space>(KERNEL_REPEATS,LOG_MAX_SIZE,import,halopong_sizes_,halopong_device_times_);
-
- }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_host_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(halopong_sizes_,halopong_host_times_,SIZE_IN_BYTES);
+        << setw(20) << size << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (h_time * 1e6) << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (d_time * 1e6) << setw(1) << std::endl;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_device_lookup(int SIZE_IN_BYTES) {
-    return PerfDetails::table_lookup(halopong_sizes_,halopong_device_times_,SIZE_IN_BYTES);
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_halopong_table(std::ostream & out, const std::string & prefix) {
-    if(halopong_sizes_.size() == 0) return;
-
-    using namespace std;
-    std::ios old_format(NULL);
-    old_format.copyfmt(out);
+  out.copyfmt(old_format);
+}
+
+/****************************************************************************************/
+/****************************************************************************************/
+/****************************************************************************************/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > &import) {
+  PerfDetails::halopong_basic<Kokkos::HostSpace::execution_space, Kokkos::HostSpace::memory_space>(KERNEL_REPEATS, LOG_MAX_SIZE, import, halopong_sizes_, halopong_host_times_);
+
+  PerfDetails::halopong_basic<typename Node::execution_space, typename Node::memory_space>(KERNEL_REPEATS, LOG_MAX_SIZE, import, halopong_sizes_, halopong_device_times_);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_host_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(halopong_sizes_, halopong_host_times_, SIZE_IN_BYTES);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::halopong_device_lookup(int SIZE_IN_BYTES) {
+  return PerfDetails::table_lookup(halopong_sizes_, halopong_device_times_, SIZE_IN_BYTES);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_halopong_table(std::ostream &out, const std::string &prefix) {
+  if (halopong_sizes_.size() == 0) return;
+
+  using namespace std;
+  std::ios old_format(NULL);
+  old_format.copyfmt(out);
+
+  out << prefix
+      << setw(20) << "Message Size" << setw(1) << " "
+      << setw(20) << "Host (us)" << setw(1) << " "
+      << setw(20) << "Device (us)" << std::endl;
+
+  out << prefix
+      << setw(20) << "------------" << setw(1) << " "
+      << setw(20) << "---------" << setw(1) << " "
+      << setw(20) << "-----------" << std::endl;
+
+  for (int i = 0; i < (int)halopong_sizes_.size(); i++) {
+    int size      = halopong_sizes_[i];
+    double h_time = halopong_host_times_[i];
+    double d_time = halopong_device_times_[i];
 
     out << prefix
-        << setw(20) << "Message Size" << setw(1) << " "
-        << setw(20) << "Host (us)" << setw(1) << " "
-        << setw(20) << "Device (us)" << std::endl;
-
-    out << prefix 
-        << setw(20) << "------------" << setw(1) << " "
-        << setw(20) << "---------" << setw(1) << " "
-        << setw(20) << "-----------" << std::endl;
-    
-
-    for(int i=0; i<(int)halopong_sizes_.size(); i++) {
-      int size = halopong_sizes_[i];
-      double h_time = halopong_host_times_[i];
-      double d_time = halopong_device_times_[i];
-
-
-      out << prefix
-          << setw(20) << size << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (h_time*1e6) << setw(1) << " "
-          << setw(20) << fixed << setprecision(4) << (d_time*1e6) << setw(1) << std::endl;
-    }
-
-    out.copyfmt(old_format);
+        << setw(20) << size << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (h_time * 1e6) << setw(1) << " "
+        << setw(20) << fixed << setprecision(4) << (d_time * 1e6) << setw(1) << std::endl;
   }
 
-  /****************************************************************************************/
-  /****************************************************************************************/
-  /****************************************************************************************/
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::launch_latency_make_table(int KERNEL_REPEATS) {
-    using exec_space   = typename Node::execution_space;
-    using range_policy = Kokkos::RangePolicy<exec_space>;
-    using clock = std::chrono::high_resolution_clock;
-
-    double total_test_time = 0;
-    clock::time_point start, stop;
-    for(int i = 0; i < KERNEL_REPEATS; i++) {       
-      start = clock::now();
-      Kokkos::parallel_for("empty kernel",range_policy(0,1), KOKKOS_LAMBDA (const size_t j) {
+  out.copyfmt(old_format);
+}
+
+/****************************************************************************************/
+/****************************************************************************************/
+/****************************************************************************************/
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::launch_latency_make_table(int KERNEL_REPEATS) {
+  using exec_space   = typename Node::execution_space;
+  using range_policy = Kokkos::RangePolicy<exec_space>;
+  using clock        = std::chrono::high_resolution_clock;
+
+  double total_test_time = 0;
+  clock::time_point start, stop;
+  for (int i = 0; i < KERNEL_REPEATS; i++) {
+    start = clock::now();
+    Kokkos::parallel_for(
+        "empty kernel", range_policy(0, 1), KOKKOS_LAMBDA(const size_t j) {
           ;
         });
-      exec_space().fence();
-      stop = clock::now();
-      double my_test_time = std::chrono::duration<double>(stop - start).count();
-      total_test_time += my_test_time;
-    }
-    
-    launch_and_wait_latency_ = total_test_time / KERNEL_REPEATS;
+    exec_space().fence();
+    stop                = clock::now();
+    double my_test_time = std::chrono::duration<double>(stop - start).count();
+    total_test_time += my_test_time;
   }
 
+  launch_and_wait_latency_ = total_test_time / KERNEL_REPEATS;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::launch_latency_lookup() {
-    return launch_and_wait_latency_;
-  }
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double
+PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::launch_latency_lookup() {
+  return launch_and_wait_latency_;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_launch_latency_table(std::ostream & out, const std::string & prefix) {
-    using namespace std;
-    std::ios old_format(NULL);
-    old_format.copyfmt(out);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PerfModels<Scalar, LocalOrdinal, GlobalOrdinal, Node>::print_launch_latency_table(std::ostream &out, const std::string &prefix) {
+  using namespace std;
+  std::ios old_format(NULL);
+  old_format.copyfmt(out);
 
-    out << prefix
-        << setw(20) << "Launch+Wait Latency (us)" << setw(1) << " " 
-        << setw(20) << fixed << setprecision(4) << (launch_and_wait_latency_*1e6) << std::endl;
-
-    out.copyfmt(old_format);
-  }
+  out << prefix
+      << setw(20) << "Launch+Wait Latency (us)" << setw(1) << " "
+      << setw(20) << fixed << setprecision(4) << (launch_and_wait_latency_ * 1e6) << std::endl;
 
+  out.copyfmt(old_format);
+}
 
-} //namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_PerfUtils_decl.hpp b/packages/muelu/src/Utils/MueLu_PerfUtils_decl.hpp
index a834eaf5258c..a5b0fa72bb4a 100644
--- a/packages/muelu/src/Utils/MueLu_PerfUtils_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_PerfUtils_decl.hpp
@@ -58,33 +58,33 @@
 
 namespace MueLu {
 // MPI helpers
-#define MueLu_sumAll(rcpComm, in, out)                                        \
-    Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_SUM, in, Teuchos::outArg(out))
-#define MueLu_minAll(rcpComm, in, out)                                        \
-    Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MIN, in, Teuchos::outArg(out))
-#define MueLu_maxAll(rcpComm, in, out)                                        \
-    Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MAX, in, Teuchos::outArg(out))
+#define MueLu_sumAll(rcpComm, in, out) \
+  Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_SUM, in, Teuchos::outArg(out))
+#define MueLu_minAll(rcpComm, in, out) \
+  Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MIN, in, Teuchos::outArg(out))
+#define MueLu_maxAll(rcpComm, in, out) \
+  Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MAX, in, Teuchos::outArg(out))
 
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class PerfUtils {
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PerfUtils {
 #undef MUELU_PERFUTILS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    static std::string PrintMatrixInfo(const Matrix& A, const std::string& msgTag, RCP<const Teuchos::ParameterList> params = Teuchos::null);
+ public:
+  static std::string PrintMatrixInfo(const Matrix& A, const std::string& msgTag, RCP<const Teuchos::ParameterList> params = Teuchos::null);
 
-    static std::string PrintImporterInfo(RCP<const Import> importer, const std::string& msgTag);
+  static std::string PrintImporterInfo(RCP<const Import> importer, const std::string& msgTag);
 
-    static std::string CommPattern(const Matrix& A, const std::string& msgTag, RCP<const Teuchos::ParameterList> params = Teuchos::null);
+  static std::string CommPattern(const Matrix& A, const std::string& msgTag, RCP<const Teuchos::ParameterList> params = Teuchos::null);
 
-  private:
-    static bool CheckMatrix(const Matrix& A);
-  };
+ private:
+  static bool CheckMatrix(const Matrix& A);
+};
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PERFUTILS_SHORT
-#endif // MUELU_PERFUTILS_DECL_HPP
+#endif  // MUELU_PERFUTILS_DECL_HPP
diff --git a/packages/muelu/src/Utils/MueLu_PerfUtils_def.hpp b/packages/muelu/src/Utils/MueLu_PerfUtils_def.hpp
index 3f44a7e52d4a..017517a8593f 100644
--- a/packages/muelu/src/Utils/MueLu_PerfUtils_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_PerfUtils_def.hpp
@@ -64,391 +64,390 @@
 
 namespace MueLu {
 
-  template<class Type>
-  void calculateStats(Type& minVal, Type& maxVal, double& avgVal, double& devVal, int& minProc, int& maxProc, const RCP<const Teuchos::Comm<int> >& comm, int numActiveProcs, const Type& v) {
-
-    Type sumVal, sum2Val, v2 = v*v;
-    using MT = typename Teuchos::ScalarTraits<Type>::magnitudeType;
-    double zero = Teuchos::ScalarTraits<double>::zero();
-    
-    MueLu_sumAll(comm,  v, sumVal);
-    MueLu_sumAll(comm, v2, sum2Val);
-    MueLu_minAll(comm,  v, minVal);
-    MueLu_maxAll(comm,  v, maxVal);
-
-    int w;
-    w = (minVal == v) ? comm->getRank() : -1;
-    MueLu_maxAll(comm,   w, maxProc);
-    w = (maxVal == v) ? comm->getRank() : -1;
-    MueLu_maxAll(comm,   w, minProc);
-
-    avgVal = (numActiveProcs > 0 ? (as<double>(Teuchos::ScalarTraits<Type>::real(sumVal)) / numActiveProcs) : zero);
-    MT avgVal_MT = Teuchos::as<MT>(avgVal);
-    devVal = (numActiveProcs > 1 ? sqrt((as<double>(Teuchos::ScalarTraits<Type>::real(sum2Val - sumVal*avgVal_MT)))/(numActiveProcs-1)) : zero);
+template <class Type>
+void calculateStats(Type& minVal, Type& maxVal, double& avgVal, double& devVal, int& minProc, int& maxProc, const RCP<const Teuchos::Comm<int> >& comm, int numActiveProcs, const Type& v) {
+  Type sumVal, sum2Val, v2 = v * v;
+  using MT    = typename Teuchos::ScalarTraits<Type>::magnitudeType;
+  double zero = Teuchos::ScalarTraits<double>::zero();
+
+  MueLu_sumAll(comm, v, sumVal);
+  MueLu_sumAll(comm, v2, sum2Val);
+  MueLu_minAll(comm, v, minVal);
+  MueLu_maxAll(comm, v, maxVal);
+
+  int w;
+  w = (minVal == v) ? comm->getRank() : -1;
+  MueLu_maxAll(comm, w, maxProc);
+  w = (maxVal == v) ? comm->getRank() : -1;
+  MueLu_maxAll(comm, w, minProc);
+
+  avgVal       = (numActiveProcs > 0 ? (as<double>(Teuchos::ScalarTraits<Type>::real(sumVal)) / numActiveProcs) : zero);
+  MT avgVal_MT = Teuchos::as<MT>(avgVal);
+  devVal       = (numActiveProcs > 1 ? sqrt((as<double>(Teuchos::ScalarTraits<Type>::real(sum2Val - sumVal * avgVal_MT))) / (numActiveProcs - 1)) : zero);
+}
+
+template <class Type>
+std::string stringStats(const RCP<const Teuchos::Comm<int> >& comm, int numActiveProcs, const Type& v, RCP<ParameterList> paramList = Teuchos::null) {
+  Type minVal, maxVal;
+  double avgVal, devVal;
+  int minProc, maxProc;
+  calculateStats<Type>(minVal, maxVal, avgVal, devVal, minProc, maxProc, comm, numActiveProcs, v);
+
+  const double zero = Teuchos::ScalarTraits<double>::zero();
+  const double one  = Teuchos::ScalarTraits<double>::one();
+  std::ostringstream buf;
+  buf << std::fixed;
+  if ((avgVal != zero) && (paramList.is_null() || !paramList->isParameter("print abs") || paramList->get<bool>("print abs") == false)) {
+    double relDev = (devVal / avgVal) * 100;
+    double relMin = (as<double>(Teuchos::ScalarTraits<Type>::real(minVal)) / avgVal - one) * 100;
+    double relMax = (as<double>(Teuchos::ScalarTraits<Type>::real(maxVal)) / avgVal - one) * 100;
+    buf << "avg = " << std::scientific << std::setw(10) << std::setprecision(2) << avgVal << ",  "
+        << "dev = " << std::fixed << std::setw(6) << std::setprecision(1) << relDev << "%,  "
+        << "min = " << std::fixed << std::setw(7) << std::setprecision(1) << std::setw(7) << relMin << "%"
+        << " (" << std::scientific << std::setw(10) << std::setprecision(2) << minVal << " on " << std::fixed << std::setw(4) << minProc << "),  "
+        << "max = " << std::fixed << std::setw(7) << std::setprecision(1) << relMax << "%"
+        << " (" << std::scientific << std::setw(10) << std::setprecision(2) << maxVal << " on " << std::fixed << std::setw(4) << maxProc << ")";
+  } else {
+    double relDev = (avgVal != zero ? (devVal / avgVal) * 100 : zero);
+    buf << "avg = " << std::scientific << std::setw(10) << std::setprecision(2) << avgVal << ",  "
+        << "dev = " << std::fixed << std::setw(6) << std::setprecision(1) << relDev << "%,  "
+        << "min = " << std::scientific << std::setw(10) << std::setprecision(2) << minVal
+        << "  (on " << std::fixed << std::setw(4) << minProc << "),          "
+        << "max = " << std::scientific << std::setw(10) << std::setprecision(2) << maxVal
+        << "  (on " << std::fixed << std::setw(4) << maxProc << ")";
   }
+  return buf.str();
+}
 
-  template<class Type>
-  std::string stringStats(const RCP<const Teuchos::Comm<int> >& comm, int numActiveProcs, const Type& v, RCP<ParameterList> paramList = Teuchos::null) {
-    Type minVal, maxVal;
-    double avgVal, devVal;
-    int minProc, maxProc;
-    calculateStats<Type>(minVal, maxVal, avgVal, devVal, minProc, maxProc, comm, numActiveProcs, v);
-
-    const double zero = Teuchos::ScalarTraits<double>::zero();
-    const double one = Teuchos::ScalarTraits<double>::one();
-    std::ostringstream buf;
-    buf << std::fixed;
-    if ((avgVal != zero) && (paramList.is_null() || !paramList->isParameter("print abs") || paramList->get<bool>("print abs") == false)) {
-      double relDev = (devVal/avgVal)*100;
-      double relMin = (as<double>(Teuchos::ScalarTraits<Type>::real(minVal))/avgVal-one)*100;
-      double relMax = (as<double>(Teuchos::ScalarTraits<Type>::real(maxVal))/avgVal-one)*100;
-      buf << "avg = " << std::scientific << std::setw(10) << std::setprecision(2) << avgVal << ",  "
-          << "dev = " << std::fixed << std::setw(6) << std::setprecision(1) << relDev << "%,  "
-          << "min = " << std::fixed << std::setw(7) << std::setprecision(1) << std::setw(7) << relMin << "%"
-          << " (" << std::scientific << std::setw(10) << std::setprecision(2) << minVal << " on " << std::fixed << std::setw(4) << minProc << "),  "
-          << "max = " << std::fixed << std::setw(7) << std::setprecision(1)  << relMax << "%"
-          << " (" << std::scientific << std::setw(10) << std::setprecision(2) << maxVal << " on " << std::fixed << std::setw(4) << maxProc << ")";
-    } else {
-      double relDev = (avgVal != zero ? (devVal/avgVal)*100 : zero);
-      buf << "avg = " << std::scientific << std::setw(10) << std::setprecision(2) << avgVal << ",  "
-          << "dev = " << std::fixed << std::setw(6) << std::setprecision(1) << relDev << "%,  "
-          << "min = " << std::scientific << std::setw(10) << std::setprecision(2) << minVal
-          << "  (on " << std::fixed << std::setw(4) << minProc << "),          "
-          << "max = " << std::scientific << std::setw(10) << std::setprecision(2) << maxVal
-          << "  (on " << std::fixed << std::setw(4) << maxProc << ")";
-    }
-    return buf.str();
-  }
-
-  template<class Map>
-  bool cmp_less(typename Map::value_type& v1, typename Map::value_type& v2) {
-    return v1.second < v2.second;
-  }
+template <class Map>
+bool cmp_less(typename Map::value_type& v1, typename Map::value_type& v2) {
+  return v1.second < v2.second;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintMatrixInfo(const Matrix& A, const std::string& msgTag, RCP<const ParameterList> params) {
-    if (!CheckMatrix(A))
-      return "";
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintMatrixInfo(const Matrix& A, const std::string& msgTag, RCP<const ParameterList> params) {
+  if (!CheckMatrix(A))
+    return "";
 
-    typedef Xpetra::global_size_t global_size_t;
+  typedef Xpetra::global_size_t global_size_t;
 
-    std::ostringstream ss;
+  std::ostringstream ss;
 
-    ss << msgTag << " size =  " << A.getGlobalNumRows() << " x " << A.getGlobalNumCols();
-    if(A.haveGlobalConstants())
-       ss << ", nnz = " << A.getGlobalNumEntries();
-    ss << std::endl;
+  ss << msgTag << " size =  " << A.getGlobalNumRows() << " x " << A.getGlobalNumCols();
+  if (A.haveGlobalConstants())
+    ss << ", nnz = " << A.getGlobalNumEntries();
+  ss << std::endl;
 
-    if (params.is_null())
-      return ss.str();
+  if (params.is_null())
+    return ss.str();
 
-    bool printLoadBalanceInfo = false, printCommInfo = false, printEntryStats = false;
-    if (params->isParameter("printLoadBalancingInfo") && params->get<bool>("printLoadBalancingInfo"))
-      printLoadBalanceInfo = true;
-    if (params->isParameter("printCommInfo") && params->get<bool>("printCommInfo"))
-      printCommInfo = true;
-    if (params->isParameter("printEntryStats") && params->get<bool>("printEntryStats"))
-      printEntryStats = true;
+  bool printLoadBalanceInfo = false, printCommInfo = false, printEntryStats = false;
+  if (params->isParameter("printLoadBalancingInfo") && params->get<bool>("printLoadBalancingInfo"))
+    printLoadBalanceInfo = true;
+  if (params->isParameter("printCommInfo") && params->get<bool>("printCommInfo"))
+    printCommInfo = true;
+  if (params->isParameter("printEntryStats") && params->get<bool>("printEntryStats"))
+    printEntryStats = true;
 
-    if (!printLoadBalanceInfo && !printCommInfo && !printEntryStats)
-      return ss.str();
+  if (!printLoadBalanceInfo && !printCommInfo && !printEntryStats)
+    return ss.str();
 
-    RCP<const Import> importer = A.getCrsGraph()->getImporter();
-    RCP<const Export> exporter = A.getCrsGraph()->getExporter();
+  RCP<const Import> importer = A.getCrsGraph()->getImporter();
+  RCP<const Export> exporter = A.getCrsGraph()->getExporter();
 
-    size_t numMyNnz = A.getLocalNumEntries(), numMyRows = A.getLocalNumRows();
+  size_t numMyNnz = A.getLocalNumEntries(), numMyRows = A.getLocalNumRows();
 
-    // Create communicator only for active processes
-    RCP<const Teuchos::Comm<int> > origComm = A.getRowMap()->getComm();
-    bool activeProc = true;
-    int numProc = origComm->getSize();
-    int numActiveProcs = 0;
+  // Create communicator only for active processes
+  RCP<const Teuchos::Comm<int> > origComm = A.getRowMap()->getComm();
+  bool activeProc                         = true;
+  int numProc                             = origComm->getSize();
+  int numActiveProcs                      = 0;
 #ifdef HAVE_MPI
-    RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(origComm);
-    MPI_Comm rawComm = (*mpiComm->getRawMpiComm())();
-
-    std::vector<size_t> numRowsPerProc(numProc);
-    Teuchos::gatherAll(*origComm, 1, &numMyRows, numProc, &numRowsPerProc[0]);
-
-    int root = 0;
-    bool rootFlag = true;
-    for (int i = 0; i < numProc; i++) {
-      if (numRowsPerProc[i]) {
-        ++numActiveProcs;
-        if(rootFlag) {
-          root = i;
-          rootFlag = false;
-        }
+  RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(origComm);
+  MPI_Comm rawComm                          = (*mpiComm->getRawMpiComm())();
+
+  std::vector<size_t> numRowsPerProc(numProc);
+  Teuchos::gatherAll(*origComm, 1, &numMyRows, numProc, &numRowsPerProc[0]);
+
+  int root      = 0;
+  bool rootFlag = true;
+  for (int i = 0; i < numProc; i++) {
+    if (numRowsPerProc[i]) {
+      ++numActiveProcs;
+      if (rootFlag) {
+        root     = i;
+        rootFlag = false;
       }
     }
+  }
 
-    if(numMyRows == 0) {activeProc = false; numMyNnz = 0;} // Reset numMyNnz to avoid adding it up in reduceAll
+  if (numMyRows == 0) {
+    activeProc = false;
+    numMyNnz   = 0;
+  }  // Reset numMyNnz to avoid adding it up in reduceAll
 #else
-    if(numMyRows == 0) {
-      //FIXME JJH 10-May-2017 Is there any case in serial where numMyRows would be zero?
-      // Reset numMyNnz to avoid adding it up in reduceAll
-      numActiveProcs = 0;
-      activeProc = false;
-      numMyNnz = 0;
-    } else {
-      numActiveProcs = 1;
-    }
+  if (numMyRows == 0) {
+    // FIXME JJH 10-May-2017 Is there any case in serial where numMyRows would be zero?
+    //  Reset numMyNnz to avoid adding it up in reduceAll
+    numActiveProcs = 0;
+    activeProc     = false;
+    numMyNnz       = 0;
+  } else {
+    numActiveProcs = 1;
+  }
 #endif
 
-    std::string outstr;
-    ParameterList absList;
-    absList.set("print abs", true);
-
-    RCP<const Matrix> rcpA = rcpFromRef(A);
-    RCP<const CrsMatrixWrap> crsWrapA = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(rcpA);
-    RCP<const CrsMatrix> crsA;
-    if (!crsWrapA.is_null())
-      crsA = crsWrapA->getCrsMatrix();
-    if (printEntryStats && !crsA.is_null()) {
-      typedef Teuchos::ScalarTraits<Scalar> STS;
-      typedef typename STS::magnitudeType magnitudeType;
-      typedef Teuchos::ScalarTraits<magnitudeType> MTS;
-      ArrayRCP<const size_t>       rowptr_RCP;
-      ArrayRCP<const LocalOrdinal> colind_RCP;
-      ArrayRCP<const Scalar>       vals_RCP;
-      ArrayRCP<size_t>             offsets_RCP;
-      ArrayView<const size_t>      rowptr;
-      ArrayView<const Scalar>      vals;
-      ArrayView<size_t>            offsets;
-
-      crsA->getAllValues(rowptr_RCP, colind_RCP, vals_RCP);
-      crsA->getLocalDiagOffsets(offsets_RCP);
-      rowptr  = rowptr_RCP();
-      vals    = vals_RCP();
-      offsets = offsets_RCP();
-
-      Scalar val, minVal, maxVal;
-      magnitudeType absVal, minAbsVal, maxAbsVal;
-      {
-        minVal = STS::rmax();
-        maxVal = STS::rmin();
-        minAbsVal = MTS::rmax();
-        maxAbsVal = MTS::zero();
-
-        for (int i = 0; i < offsets.size(); i++) {
-          val = vals[rowptr[i]+offsets[i]];
-          if (STS::real(val) < STS::real(minVal))
-            minVal = val;
-          if (STS::real(val) > STS::real(maxVal))
-            maxVal = val;
-          absVal    = STS::magnitude(val);
-          minAbsVal = std::min(minAbsVal, absVal);
-          maxAbsVal = std::max(maxAbsVal, absVal);
-        }
-
-        ss << msgTag << " diag min       : " << stringStats<Scalar>(origComm, numActiveProcs, minVal) << std::endl;
-        ss << msgTag << " diag max       : " << stringStats<Scalar>(origComm, numActiveProcs, maxVal) << std::endl;
-        ss << msgTag << " abs(diag) min  : " << stringStats<Scalar>(origComm, numActiveProcs, minAbsVal) << std::endl;
-        ss << msgTag << " abs(diag) max  : " << stringStats<Scalar>(origComm, numActiveProcs, maxAbsVal) << std::endl;
-      }
-
-      {
-        minVal = STS::rmax();
-        maxVal = STS::rmin();
-        minAbsVal = MTS::rmax();
-        maxAbsVal = MTS::zero();
-
-        for (int i = 0; i < vals.size(); i++) {
-          val = vals[i];
-          if (STS::real(val) < STS::real(minVal))
-            minVal = val;
-          if (STS::real(val) > STS::real(maxVal))
-            maxVal = val;
-          absVal    = STS::magnitude(val);
-          minAbsVal = std::min(minAbsVal, absVal);
-          maxAbsVal = std::max(maxAbsVal, absVal);
-        }
-
-        ss << msgTag << " entry min      : " << stringStats<Scalar>(origComm, numActiveProcs, minVal) << std::endl;
-        ss << msgTag << " entry max      : " << stringStats<Scalar>(origComm, numActiveProcs, maxVal) << std::endl;
-        ss << msgTag << " abs(entry) min : " << stringStats<Scalar>(origComm, numActiveProcs, minAbsVal) << std::endl;
-        ss << msgTag << " abs(entry) max : " << stringStats<Scalar>(origComm, numActiveProcs, maxAbsVal) << std::endl;
+  std::string outstr;
+  ParameterList absList;
+  absList.set("print abs", true);
+
+  RCP<const Matrix> rcpA            = rcpFromRef(A);
+  RCP<const CrsMatrixWrap> crsWrapA = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(rcpA);
+  RCP<const CrsMatrix> crsA;
+  if (!crsWrapA.is_null())
+    crsA = crsWrapA->getCrsMatrix();
+  if (printEntryStats && !crsA.is_null()) {
+    typedef Teuchos::ScalarTraits<Scalar> STS;
+    typedef typename STS::magnitudeType magnitudeType;
+    typedef Teuchos::ScalarTraits<magnitudeType> MTS;
+    ArrayRCP<const size_t> rowptr_RCP;
+    ArrayRCP<const LocalOrdinal> colind_RCP;
+    ArrayRCP<const Scalar> vals_RCP;
+    ArrayRCP<size_t> offsets_RCP;
+    ArrayView<const size_t> rowptr;
+    ArrayView<const Scalar> vals;
+    ArrayView<size_t> offsets;
+
+    crsA->getAllValues(rowptr_RCP, colind_RCP, vals_RCP);
+    crsA->getLocalDiagOffsets(offsets_RCP);
+    rowptr  = rowptr_RCP();
+    vals    = vals_RCP();
+    offsets = offsets_RCP();
+
+    Scalar val, minVal, maxVal;
+    magnitudeType absVal, minAbsVal, maxAbsVal;
+    {
+      minVal    = STS::rmax();
+      maxVal    = STS::rmin();
+      minAbsVal = MTS::rmax();
+      maxAbsVal = MTS::zero();
+
+      for (int i = 0; i < offsets.size(); i++) {
+        val = vals[rowptr[i] + offsets[i]];
+        if (STS::real(val) < STS::real(minVal))
+          minVal = val;
+        if (STS::real(val) > STS::real(maxVal))
+          maxVal = val;
+        absVal    = STS::magnitude(val);
+        minAbsVal = std::min(minAbsVal, absVal);
+        maxAbsVal = std::max(maxAbsVal, absVal);
       }
-    }
-
 
-    if (printLoadBalanceInfo) {
-      ss << msgTag << " Load balancing info"    << std::endl;
-      ss << msgTag << "   # active processes: " << numActiveProcs << "/" << numProc << std::endl;
-      ss << msgTag << "   # rows per proc   : " << stringStats<global_size_t>(origComm, numActiveProcs, numMyRows) << std::endl;
-      ss << msgTag << "   #  nnz per proc   : " << stringStats<global_size_t>(origComm, numActiveProcs,  numMyNnz) << std::endl;
+      ss << msgTag << " diag min       : " << stringStats<Scalar>(origComm, numActiveProcs, minVal) << std::endl;
+      ss << msgTag << " diag max       : " << stringStats<Scalar>(origComm, numActiveProcs, maxVal) << std::endl;
+      ss << msgTag << " abs(diag) min  : " << stringStats<Scalar>(origComm, numActiveProcs, minAbsVal) << std::endl;
+      ss << msgTag << " abs(diag) max  : " << stringStats<Scalar>(origComm, numActiveProcs, maxAbsVal) << std::endl;
     }
 
-    if (printCommInfo && numActiveProcs != 1) {
-      typedef std::map<int,size_t> map_type;
-      map_type neighMap;
-      if (!importer.is_null()) {
-        ArrayView<const int> exportPIDs = importer->getExportPIDs();
-        if (exportPIDs.size())
-          for (int i = 0; i < exportPIDs.size(); i++)
-            neighMap[exportPIDs[i]]++;
+    {
+      minVal    = STS::rmax();
+      maxVal    = STS::rmin();
+      minAbsVal = MTS::rmax();
+      maxAbsVal = MTS::zero();
+
+      for (int i = 0; i < vals.size(); i++) {
+        val = vals[i];
+        if (STS::real(val) < STS::real(minVal))
+          minVal = val;
+        if (STS::real(val) > STS::real(maxVal))
+          maxVal = val;
+        absVal    = STS::magnitude(val);
+        minAbsVal = std::min(minAbsVal, absVal);
+        maxAbsVal = std::max(maxAbsVal, absVal);
       }
 
-      // Communication volume
-      size_t numExportSend = 0;
-      size_t numImportSend = 0;
-      size_t numMsgs       = 0;
-      size_t minMsg        = 0;
-      size_t maxMsg        = 0;
-
-      if(activeProc) {
-        numExportSend = (!exporter.is_null() ? exporter->getNumExportIDs() : 0);
-        numImportSend = (!importer.is_null() ? importer->getNumExportIDs() : 0);
-        numMsgs       = neighMap.size();
-        map_type::const_iterator it = std::min_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
-        minMsg        = (it != neighMap.end() ? it->second : 0);
-        it = std::max_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
-        maxMsg        = (it != neighMap.end() ? it->second : 0);
-      }
-
-      ss << msgTag << " Communication info"     << std::endl;
-      ss << msgTag << "   # num export send : " << stringStats<global_size_t>(origComm, numActiveProcs, numExportSend)                      << std::endl;
-      ss << msgTag << "   # num import send : " << stringStats<global_size_t>(origComm, numActiveProcs, numImportSend)                      << std::endl;
-      ss << msgTag << "   # num msgs        : " << stringStats<global_size_t>(origComm, numActiveProcs,       numMsgs, rcpFromRef(absList)) << std::endl;
-      ss << msgTag << "   # min msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs,        minMsg)                      << std::endl;
-      ss << msgTag << "   # max msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs,        maxMsg)                      << std::endl;
+      ss << msgTag << " entry min      : " << stringStats<Scalar>(origComm, numActiveProcs, minVal) << std::endl;
+      ss << msgTag << " entry max      : " << stringStats<Scalar>(origComm, numActiveProcs, maxVal) << std::endl;
+      ss << msgTag << " abs(entry) min : " << stringStats<Scalar>(origComm, numActiveProcs, minAbsVal) << std::endl;
+      ss << msgTag << " abs(entry) max : " << stringStats<Scalar>(origComm, numActiveProcs, maxAbsVal) << std::endl;
     }
-
-    outstr = ss.str();
-
-#ifdef HAVE_MPI
-    int strLength = outstr.size();
-    MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
-    if (origComm->getRank() != root)
-      outstr.resize(strLength);
-    MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
-#endif
-
-    return outstr;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintImporterInfo(RCP<const Import> importer, const std::string& msgTag) {
-
-    typedef Xpetra::global_size_t global_size_t;
-
-    std::ostringstream ss;
-
-    // Create communicator only for active processes
-    RCP<const Teuchos::Comm<int> > origComm = importer->getSourceMap()->getComm();
-    bool activeProc = true;
-    int numActiveProcs = origComm->getSize();
-#ifdef HAVE_MPI
-    RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(origComm);
-    MPI_Comm rawComm = (*mpiComm->getRawMpiComm())();
-    int root = 0;
-#endif
-
-    std::string outstr;
-    ParameterList absList;
-    absList.set("print abs", true);
+  if (printLoadBalanceInfo) {
+    ss << msgTag << " Load balancing info" << std::endl;
+    ss << msgTag << "   # active processes: " << numActiveProcs << "/" << numProc << std::endl;
+    ss << msgTag << "   # rows per proc   : " << stringStats<global_size_t>(origComm, numActiveProcs, numMyRows) << std::endl;
+    ss << msgTag << "   #  nnz per proc   : " << stringStats<global_size_t>(origComm, numActiveProcs, numMyNnz) << std::endl;
+  }
 
-    typedef std::map<int,size_t> map_type;
+  if (printCommInfo && numActiveProcs != 1) {
+    typedef std::map<int, size_t> map_type;
     map_type neighMap;
-    ArrayView<const int> exportPIDs = importer->getExportPIDs();
-    if (exportPIDs.size())
-      for (int i = 0; i < exportPIDs.size(); i++)
-        neighMap[exportPIDs[i]]++;
+    if (!importer.is_null()) {
+      ArrayView<const int> exportPIDs = importer->getExportPIDs();
+      if (exportPIDs.size())
+        for (int i = 0; i < exportPIDs.size(); i++)
+          neighMap[exportPIDs[i]]++;
+    }
 
     // Communication volume
+    size_t numExportSend = 0;
     size_t numImportSend = 0;
     size_t numMsgs       = 0;
     size_t minMsg        = 0;
     size_t maxMsg        = 0;
 
-    if(activeProc) {
-      numImportSend = importer->getNumExportIDs();
-      numMsgs       = neighMap.size();
+    if (activeProc) {
+      numExportSend               = (!exporter.is_null() ? exporter->getNumExportIDs() : 0);
+      numImportSend               = (!importer.is_null() ? importer->getNumExportIDs() : 0);
+      numMsgs                     = neighMap.size();
       map_type::const_iterator it = std::min_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
-      minMsg        = (it != neighMap.end() ? it->second : 0);
-      it = std::max_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
-      maxMsg        = (it != neighMap.end() ? it->second : 0);
+      minMsg                      = (it != neighMap.end() ? it->second : 0);
+      it                          = std::max_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
+      maxMsg                      = (it != neighMap.end() ? it->second : 0);
     }
 
-    ss << msgTag << " Communication info"     << std::endl;
-    ss << msgTag << "   # num import send : " << stringStats<global_size_t>(origComm, numActiveProcs, numImportSend)                      << std::endl;
-    ss << msgTag << "   # num msgs        : " << stringStats<global_size_t>(origComm, numActiveProcs,       numMsgs, rcpFromRef(absList)) << std::endl;
-    ss << msgTag << "   # min msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs,        minMsg)                      << std::endl;
-    ss << msgTag << "   # max msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs,        maxMsg)                      << std::endl;
+    ss << msgTag << " Communication info" << std::endl;
+    ss << msgTag << "   # num export send : " << stringStats<global_size_t>(origComm, numActiveProcs, numExportSend) << std::endl;
+    ss << msgTag << "   # num import send : " << stringStats<global_size_t>(origComm, numActiveProcs, numImportSend) << std::endl;
+    ss << msgTag << "   # num msgs        : " << stringStats<global_size_t>(origComm, numActiveProcs, numMsgs, rcpFromRef(absList)) << std::endl;
+    ss << msgTag << "   # min msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs, minMsg) << std::endl;
+    ss << msgTag << "   # max msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs, maxMsg) << std::endl;
+  }
 
+  outstr = ss.str();
 
-    outstr = ss.str();
+#ifdef HAVE_MPI
+  int strLength = outstr.size();
+  MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
+  if (origComm->getRank() != root)
+    outstr.resize(strLength);
+  MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+#endif
+
+  return outstr;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PrintImporterInfo(RCP<const Import> importer, const std::string& msgTag) {
+  typedef Xpetra::global_size_t global_size_t;
 
+  std::ostringstream ss;
+
+  // Create communicator only for active processes
+  RCP<const Teuchos::Comm<int> > origComm = importer->getSourceMap()->getComm();
+  bool activeProc                         = true;
+  int numActiveProcs                      = origComm->getSize();
 #ifdef HAVE_MPI
-    int strLength = outstr.size();
-    MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
-    if (origComm->getRank() != root)
-      outstr.resize(strLength);
-    MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+  RCP<const Teuchos::MpiComm<int> > mpiComm = rcp_dynamic_cast<const Teuchos::MpiComm<int> >(origComm);
+  MPI_Comm rawComm                          = (*mpiComm->getRawMpiComm())();
+  int root                                  = 0;
 #endif
 
-    return outstr;
+  std::string outstr;
+  ParameterList absList;
+  absList.set("print abs", true);
+
+  typedef std::map<int, size_t> map_type;
+  map_type neighMap;
+  ArrayView<const int> exportPIDs = importer->getExportPIDs();
+  if (exportPIDs.size())
+    for (int i = 0; i < exportPIDs.size(); i++)
+      neighMap[exportPIDs[i]]++;
+
+  // Communication volume
+  size_t numImportSend = 0;
+  size_t numMsgs       = 0;
+  size_t minMsg        = 0;
+  size_t maxMsg        = 0;
+
+  if (activeProc) {
+    numImportSend               = importer->getNumExportIDs();
+    numMsgs                     = neighMap.size();
+    map_type::const_iterator it = std::min_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
+    minMsg                      = (it != neighMap.end() ? it->second : 0);
+    it                          = std::max_element(neighMap.begin(), neighMap.end(), cmp_less<map_type>);
+    maxMsg                      = (it != neighMap.end() ? it->second : 0);
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CommPattern(const Matrix& A, const std::string& msgTag, RCP<const ParameterList> /* params */) {
-    if (!CheckMatrix(A))
-      return "";
-
-    std::ostringstream out;
+  ss << msgTag << " Communication info" << std::endl;
+  ss << msgTag << "   # num import send : " << stringStats<global_size_t>(origComm, numActiveProcs, numImportSend) << std::endl;
+  ss << msgTag << "   # num msgs        : " << stringStats<global_size_t>(origComm, numActiveProcs, numMsgs, rcpFromRef(absList)) << std::endl;
+  ss << msgTag << "   # min msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs, minMsg) << std::endl;
+  ss << msgTag << "   # max msg size    : " << stringStats<global_size_t>(origComm, numActiveProcs, maxMsg) << std::endl;
 
-    RCP<const Teuchos::Comm<int> > comm = A.getRowMap()->getComm();
-    int myRank = comm->getRank();
+  outstr = ss.str();
 
-    out << msgTag << " " << myRank << ":";
+#ifdef HAVE_MPI
+  int strLength = outstr.size();
+  MPI_Bcast(&strLength, 1, MPI_INT, root, rawComm);
+  if (origComm->getRank() != root)
+    outstr.resize(strLength);
+  MPI_Bcast(&outstr[0], strLength, MPI_CHAR, root, rawComm);
+#endif
 
-    RCP<const Import> importer = (A.getCrsGraph() != Teuchos::null ? A.getCrsGraph()->getImporter() : Teuchos::null);
-    if (importer.is_null()) {
-      out << std::endl;
-      return out.str();
-    }
+  return outstr;
+}
 
-    ArrayView<const int> exportPIDs = importer->getExportPIDs();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CommPattern(const Matrix& A, const std::string& msgTag, RCP<const ParameterList> /* params */) {
+  if (!CheckMatrix(A))
+    return "";
 
-    if (exportPIDs.size()) {
-      // NodeTE: exportPIDs is sorted but not unique ( 1 1 1 2 2 3 4 4 4 )
-      int neigh  = exportPIDs[0];
-      GO  weight = 1;
-      for (int i = 1; i < exportPIDs.size(); i++) {
-        if (exportPIDs[i] != exportPIDs[i-1]) {
-          out << " " << neigh << "(" << weight << ")";
+  std::ostringstream out;
 
-          neigh  = exportPIDs[i];
-          weight = 1;
+  RCP<const Teuchos::Comm<int> > comm = A.getRowMap()->getComm();
+  int myRank                          = comm->getRank();
 
-        } else {
-          weight += 1;
-        }
-      }
-      out << " " << neigh << "(" << weight << ")" << std::endl;
-    }
+  out << msgTag << " " << myRank << ":";
 
+  RCP<const Import> importer = (A.getCrsGraph() != Teuchos::null ? A.getCrsGraph()->getImporter() : Teuchos::null);
+  if (importer.is_null()) {
+    out << std::endl;
     return out.str();
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckMatrix(const Matrix& A) {
-    // We can only print statistics for matrices that have a crs graph. A
-    // potential issue is regarding Xpetra::TpetraBlockCrsMatrix which has no
-    // CrsGraph.  It is held as a private data member by Xpetra::CrsMatrix,
-    // which itself is an Xpetra::Matrix. So we check directly whether the
-    // request for the graph throws.
-    bool hasCrsGraph = true;
-    try {
-      A.getCrsGraph();
-
-    } catch (...) {
-      hasCrsGraph = false;
+  ArrayView<const int> exportPIDs = importer->getExportPIDs();
+
+  if (exportPIDs.size()) {
+    // NodeTE: exportPIDs is sorted but not unique ( 1 1 1 2 2 3 4 4 4 )
+    int neigh = exportPIDs[0];
+    GO weight = 1;
+    for (int i = 1; i < exportPIDs.size(); i++) {
+      if (exportPIDs[i] != exportPIDs[i - 1]) {
+        out << " " << neigh << "(" << weight << ")";
+
+        neigh  = exportPIDs[i];
+        weight = 1;
+
+      } else {
+        weight += 1;
+      }
     }
+    out << " " << neigh << "(" << weight << ")" << std::endl;
+  }
 
-    return hasCrsGraph;
+  return out.str();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool PerfUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CheckMatrix(const Matrix& A) {
+  // We can only print statistics for matrices that have a crs graph. A
+  // potential issue is regarding Xpetra::TpetraBlockCrsMatrix which has no
+  // CrsGraph.  It is held as a private data member by Xpetra::CrsMatrix,
+  // which itself is an Xpetra::Matrix. So we check directly whether the
+  // request for the graph throws.
+  bool hasCrsGraph = true;
+  try {
+    A.getCrsGraph();
+
+  } catch (...) {
+    hasCrsGraph = false;
   }
 
-} //namespace MueLu
+  return hasCrsGraph;
+}
+
+}  // namespace MueLu
 
-#endif // MUELU_PERFUTILS_DEF_HPP
+#endif  // MUELU_PERFUTILS_DEF_HPP
diff --git a/packages/muelu/src/Utils/MueLu_PermutationFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_PermutationFactory_decl.hpp
index fb2a4c3d730e..2592a5f03aa1 100644
--- a/packages/muelu/src/Utils/MueLu_PermutationFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_PermutationFactory_decl.hpp
@@ -64,62 +64,60 @@
 
 namespace MueLu {
 
-  /*!
-    @class PermutationFactory class.
-    @brief factory generates a row- and column permutation operators P and Q such that
-    P*A*Q^T is a (hopefully) diagonal-dominant matrix.
-    It's meant to be used with PermutingSmoother.
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class PermutationFactory : public SingleLevelFactoryBase {
+/*!
+  @class PermutationFactory class.
+  @brief factory generates a row- and column permutation operators P and Q such that
+  P*A*Q^T is a (hopefully) diagonal-dominant matrix.
+  It's meant to be used with PermutingSmoother.
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class PermutationFactory : public SingleLevelFactoryBase {
 #undef MUELU_PERMUTATIONFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    PermutationFactory();
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Destructor.
-    virtual ~PermutationFactory();
+  //! Constructor.
+  PermutationFactory();
 
-    //@}
+  //! Destructor.
+  virtual ~PermutationFactory();
 
-    RCP<const ParameterList> GetValidParameterList() const;
+  //@}
 
-    //! @name Input
-    //@{
+  RCP<const ParameterList> GetValidParameterList() const;
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  //! @name Input
+  //@{
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-    //@}
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level & currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-  private:
+  //@}
 
-  }; // class PermutationFactory
+ private:
+};  // class PermutationFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_PERMUTATIONFACTORY_SHORT
 
-
 #endif /* MUELU_PERMUTATIONFACTORY_DECL_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_PermutationFactory_def.hpp b/packages/muelu/src/Utils/MueLu_PermutationFactory_def.hpp
index 5bef810bf685..286b0ed23fd0 100644
--- a/packages/muelu/src/Utils/MueLu_PermutationFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_PermutationFactory_def.hpp
@@ -59,7 +59,7 @@
 #include "MueLu_PermutationFactory_decl.hpp"
 
 #include <Xpetra_Map.hpp>
-#include <Xpetra_StridedMap.hpp>    // for nDofsPerNode...
+#include <Xpetra_StridedMap.hpp>  // for nDofsPerNode...
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_VectorFactory.hpp>
 #include <Xpetra_Matrix.hpp>
@@ -79,8 +79,7 @@
 
 namespace MueLu {
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PermutationFactory()
-  { }
+PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::PermutationFactory() {}
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~PermutationFactory() {}
@@ -89,12 +88,12 @@ template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 RCP<const ParameterList> PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
   RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-  validParamList->set< RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be permuted.");
+  validParamList->set<RCP<const FactoryBase> >("A", Teuchos::null, "Generating factory of the matrix A to be permuted.");
 
-  validParamList->set< std::string >           ("PermutationRowMapName", "", "Name of input row map for which rows the permutation shall be done. (default='')");
-  validParamList->set< RCP<const FactoryBase> >("PermutationRowMapFactory", Teuchos::null, "Generating factory of the input row map for the permutation.");
+  validParamList->set<std::string>("PermutationRowMapName", "", "Name of input row map for which rows the permutation shall be done. (default='')");
+  validParamList->set<RCP<const FactoryBase> >("PermutationRowMapFactory", Teuchos::null, "Generating factory of the input row map for the permutation.");
 
-  validParamList->set< std::string >           ("PermutationStrategy", "Algebraic", "Permutation strategy (default = 'Algebraic', 'Local'");
+  validParamList->set<std::string>("PermutationStrategy", "Algebraic", "Permutation strategy (default = 'Algebraic', 'Local'");
 
   return validParamList;
 }
@@ -103,50 +102,48 @@ template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
   Input(currentLevel, "A");
 
-  const ParameterList & pL = GetParameterList();
-  std::string mapName                        = pL.get<std::string> ("PermutationRowMapName");
-  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory          ("PermutationRowMapFactory");
+  const ParameterList &pL                    = GetParameterList();
+  std::string mapName                        = pL.get<std::string>("PermutationRowMapName");
+  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory("PermutationRowMapFactory");
 
-  if(mapName.length() > 0 ) {
-    currentLevel.DeclareInput(mapName,mapFactory.get(),this);
+  if (mapName.length() > 0) {
+    currentLevel.DeclareInput(mapName, mapFactory.get(), this);
   }
 }
 
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level & currentLevel) const {
+void PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
   FactoryMonitor m(*this, "Permutation Factory ", currentLevel);
 
-  Teuchos::RCP<Matrix> A = Get< Teuchos::RCP<Matrix> > (currentLevel, "A");
+  Teuchos::RCP<Matrix> A = Get<Teuchos::RCP<Matrix> >(currentLevel, "A");
 
-  const ParameterList & pL = GetParameterList();
-  std::string mapName                        = pL.get<std::string> ("PermutationRowMapName");
-  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory          ("PermutationRowMapFactory");
+  const ParameterList &pL                    = GetParameterList();
+  std::string mapName                        = pL.get<std::string>("PermutationRowMapName");
+  Teuchos::RCP<const FactoryBase> mapFactory = GetFactory("PermutationRowMapFactory");
 
   Teuchos::RCP<const Map> permRowMap = Teuchos::null;
-  if(mapName.length() > 0 ) {
-    permRowMap = currentLevel.Get<RCP<const Map> >(mapName,mapFactory.get());
+  if (mapName.length() > 0) {
+    permRowMap = currentLevel.Get<RCP<const Map> >(mapName, mapFactory.get());
   } else {
-    permRowMap = A->getRowMap(); // use full row map of A
+    permRowMap = A->getRowMap();  // use full row map of A
   }
 
-  std::string strStrategy = pL.get<std::string> ("PermutationStrategy");
-  if( strStrategy == "Algebraic" ) {
+  std::string strStrategy = pL.get<std::string>("PermutationStrategy");
+  if (strStrategy == "Algebraic") {
     Teuchos::RCP<AlgebraicPermutationStrategy> permStrat = Teuchos::rcp(new AlgebraicPermutationStrategy());
-    permStrat->BuildPermutation(A,permRowMap,currentLevel,this);
-  } else if( strStrategy == "Local" ) {
+    permStrat->BuildPermutation(A, permRowMap, currentLevel, this);
+  } else if (strStrategy == "Local") {
     Teuchos::RCP<LocalPermutationStrategy> permStrat = Teuchos::rcp(new LocalPermutationStrategy());
-    permStrat->BuildPermutation(A,permRowMap,currentLevel,this);
+    permStrat->BuildPermutation(A, permRowMap, currentLevel, this);
   } else
     TEUCHOS_TEST_FOR_EXCEPTION(true,
-                                  std::logic_error,
-                                  "`PermutationStrategy' has incorrect value (" << strStrategy << ") in input to PermutationFactory."
-                                  << "Check the documentation for a list of valid choices");
+                               std::logic_error,
+                               "`PermutationStrategy' has incorrect value (" << strStrategy << ") in input to PermutationFactory."
+                                                                             << "Check the documentation for a list of valid choices");
 
   GetOStream(Runtime0) << "Using " << strStrategy << " permutation strategy." << std::endl;
-
 }
 
-} // namespace MueLu
-
+}  // namespace MueLu
 
 #endif /* MUELU_PERMUTATIONFACTORY_DEF_HPP_ */
diff --git a/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_decl.hpp b/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_decl.hpp
index dc173586a9a4..efa39cfbd7b7 100644
--- a/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_decl.hpp
@@ -58,73 +58,70 @@
 
 namespace MueLu {
 
-  /*!
-    @class RigidBodyModeFactory class.
-    @brief Nullspace Factory for building rigid body modes.
-    @ingroup MueLuTransferClasses
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class RigidBodyModeFactory : public SingleLevelFactoryBase {
+/*!
+  @class RigidBodyModeFactory class.
+  @brief Nullspace Factory for building rigid body modes.
+  @ingroup MueLuTransferClasses
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class RigidBodyModeFactory : public SingleLevelFactoryBase {
 #undef MUELU_RIGIDBODYMODEFACTORY_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
+ public:
+  //! @name Constructors/Destructors.
+  //@{
 
-    //! Constructor
-    RigidBodyModeFactory(const int numPDEs)
-      : nspName_("Nullspace"),
-        numPDEs_(numPDEs)
-    { }
-    //! Constructor
-    RigidBodyModeFactory(const std::string & nspName = "Nullspace")
-      : nspName_(nspName),
-        numPDEs_(3)
-    { }
+  //! Constructor
+  RigidBodyModeFactory(const int numPDEs)
+    : nspName_("Nullspace")
+    , numPDEs_(numPDEs) {}
+  //! Constructor
+  RigidBodyModeFactory(const std::string &nspName = "Nullspace")
+    : nspName_(nspName)
+    , numPDEs_(3) {}
 
-    //! Destructor.
-    virtual ~RigidBodyModeFactory();
+  //! Destructor.
+  virtual ~RigidBodyModeFactory();
 
-    //@}
+  //@}
 
-    //! @name Input
-    //@{
+  //! @name Input
+  //@{
 
-    /*! @brief Specifies the data that this class needs, and the factories that generate that data.
+  /*! @brief Specifies the data that this class needs, and the factories that generate that data.
 
-        If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
-        will fall back to the settings in FactoryManager.
-    */
-    void DeclareInput(Level &currentLevel) const;
-
-    //@}
+      If the Build method of this class requires some data, but the generating factory is not specified in DeclareInput, then this class
+      will fall back to the settings in FactoryManager.
+  */
+  void DeclareInput(Level &currentLevel) const;
 
-    //! @name Build methods.
-    //@{
+  //@}
 
-    //! Build an object with this factory.
-    void Build(Level &currentLevel) const;
+  //! @name Build methods.
+  //@{
 
-    //@}
-    void setNumPDEs(int numPDEs) {
-      numPDEs_=numPDEs;
-    }
+  //! Build an object with this factory.
+  void Build(Level &currentLevel) const;
 
-  private:
+  //@}
+  void setNumPDEs(int numPDEs) {
+    numPDEs_ = numPDEs;
+  }
 
-    //! name of nullspace vector on finest level
-    std::string nspName_;
+ private:
+  //! name of nullspace vector on finest level
+  std::string nspName_;
 
-    int numPDEs_;
+  int numPDEs_;
 
-  }; // class RigidBodyModeFactory
+};  // class RigidBodyModeFactory
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RIGIDBODYMODEFACTORY_SHORT
-#endif // MUELU_RIGIDBODYMODEFACTORY_DECL_HPP
+#endif  // MUELU_RIGIDBODYMODEFACTORY_DECL_HPP
diff --git a/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_def.hpp b/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_def.hpp
index 66f056610223..10bad1ce4b3c 100644
--- a/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_RigidBodyModeFactory_def.hpp
@@ -56,105 +56,105 @@
 
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~RigidBodyModeFactory() {}
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::~RigidBodyModeFactory() {}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
-    if (currentLevel.IsAvailable(nspName_, NoFactory::get()) == false && currentLevel.GetLevelID() == 0) {
-      Input(currentLevel, "A");
-      //Input(currentLevel,"Coordinates");
-    }
-    if (currentLevel.GetLevelID() !=0) {
-      currentLevel.DeclareInput("Nullspace", GetFactory(nspName_).get(), this); /* ! "Nullspace" and nspName_ mismatch possible here */
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::DeclareInput(Level &currentLevel) const {
+  if (currentLevel.IsAvailable(nspName_, NoFactory::get()) == false && currentLevel.GetLevelID() == 0) {
+    Input(currentLevel, "A");
+    // Input(currentLevel,"Coordinates");
   }
+  if (currentLevel.GetLevelID() != 0) {
+    currentLevel.DeclareInput("Nullspace", GetFactory(nspName_).get(), this); /* ! "Nullspace" and nspName_ mismatch possible here */
+  }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
-    FactoryMonitor m(*this, "Rigid body mode factory", currentLevel);
-    RCP<MultiVector> nullspace;
-    if (currentLevel.GetLevelID() == 0) {
-      if (currentLevel.IsAvailable(nspName_, NoFactory::get())) {
-        nullspace = currentLevel.Get< RCP<MultiVector> >(nspName_, NoFactory::get());
-        GetOStream(Runtime1) << "Use user-given rigid body modes " << nspName_ << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(Level &currentLevel) const {
+  FactoryMonitor m(*this, "Rigid body mode factory", currentLevel);
+  RCP<MultiVector> nullspace;
+  if (currentLevel.GetLevelID() == 0) {
+    if (currentLevel.IsAvailable(nspName_, NoFactory::get())) {
+      nullspace = currentLevel.Get<RCP<MultiVector> >(nspName_, NoFactory::get());
+      GetOStream(Runtime1) << "Use user-given rigid body modes " << nspName_ << ": nullspace dimension=" << nullspace->getNumVectors() << " nullspace length=" << nullspace->getGlobalLength() << std::endl;
+    } else {
+      RCP<Matrix> A = Get<RCP<Matrix> >(currentLevel, "A");
+      GetOStream(Runtime1) << "Generating rigid body modes: dimension = " << numPDEs_ << std::endl;
+      RCP<const Map> xmap = A->getDomainMap();
+      if (numPDEs_ == 1) {
+        nullspace = MultiVectorFactory::Build(xmap, 1);
+      } else if (numPDEs_ == 2) {
+        nullspace = MultiVectorFactory::Build(xmap, 3);
+      } else if (numPDEs_ == 3) {
+        nullspace = MultiVectorFactory::Build(xmap, 6);
       }
-      else {
-        RCP<Matrix> A = Get< RCP<Matrix> >(currentLevel, "A");
-        GetOStream(Runtime1) << "Generating rigid body modes: dimension = " << numPDEs_ << std::endl;
-        RCP<const Map> xmap=A->getDomainMap();
-        if(numPDEs_==1)      { nullspace = MultiVectorFactory::Build(xmap, 1); }
-        else if(numPDEs_==2) { nullspace = MultiVectorFactory::Build(xmap, 3); }
-        else if(numPDEs_==3) { nullspace = MultiVectorFactory::Build(xmap, 6); }
-        Scalar zero(0.0);
-        nullspace -> putScalar(zero);
-        RCP<MultiVector> Coords = Get< RCP<MultiVector> >(currentLevel,"Coordinates");
-        ArrayRCP<Scalar> xnodes, ynodes, znodes;
-        Scalar cx, cy, cz;
-        ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
-        int nDOFs=xmap->getLocalNumElements();
-        if(numPDEs_==1) {
-          nsValues0 = nullspace->getDataNonConst(0);
-          for(int j=0; j<nDOFs; j++) {
-            // constant null space for scalar PDE
-            nsValues0[j]=1.0;
-          }
+      Scalar zero(0.0);
+      nullspace->putScalar(zero);
+      RCP<MultiVector> Coords = Get<RCP<MultiVector> >(currentLevel, "Coordinates");
+      ArrayRCP<Scalar> xnodes, ynodes, znodes;
+      Scalar cx, cy, cz;
+      ArrayRCP<Scalar> nsValues0, nsValues1, nsValues2, nsValues3, nsValues4, nsValues5;
+      int nDOFs = xmap->getLocalNumElements();
+      if (numPDEs_ == 1) {
+        nsValues0 = nullspace->getDataNonConst(0);
+        for (int j = 0; j < nDOFs; j++) {
+          // constant null space for scalar PDE
+          nsValues0[j] = 1.0;
         }
-        else if(numPDEs_==2) {
-          xnodes = Coords->getDataNonConst(0);
-          ynodes = Coords->getDataNonConst(1);
-          cx = Coords->getVector(0)->meanValue();
-          cy = Coords->getVector(1)->meanValue();
-          nsValues0 = nullspace->getDataNonConst(0);
-          nsValues1 = nullspace->getDataNonConst(1);
-          nsValues2 = nullspace->getDataNonConst(2);
-          for (int j=0; j<nDOFs; j+=numPDEs_) {
-            // translation
-            nsValues0[j+0] = 1.0;
-            nsValues1[j+1] = 1.0;
-            // rotate around z-axis (x-y plane)
-            nsValues2[j+0] = -(ynodes[j]-cy);
-            nsValues2[j+1] =  (xnodes[j]-cx);
-          }
+      } else if (numPDEs_ == 2) {
+        xnodes    = Coords->getDataNonConst(0);
+        ynodes    = Coords->getDataNonConst(1);
+        cx        = Coords->getVector(0)->meanValue();
+        cy        = Coords->getVector(1)->meanValue();
+        nsValues0 = nullspace->getDataNonConst(0);
+        nsValues1 = nullspace->getDataNonConst(1);
+        nsValues2 = nullspace->getDataNonConst(2);
+        for (int j = 0; j < nDOFs; j += numPDEs_) {
+          // translation
+          nsValues0[j + 0] = 1.0;
+          nsValues1[j + 1] = 1.0;
+          // rotate around z-axis (x-y plane)
+          nsValues2[j + 0] = -(ynodes[j] - cy);
+          nsValues2[j + 1] = (xnodes[j] - cx);
         }
-        else if(numPDEs_==3) {
-          xnodes = Coords->getDataNonConst(0);
-          ynodes = Coords->getDataNonConst(1);
-          znodes = Coords->getDataNonConst(2);
-          cx = Coords->getVector(0)->meanValue();
-          cy = Coords->getVector(1)->meanValue();
-          cz = Coords->getVector(2)->meanValue();
-          nsValues0 = nullspace->getDataNonConst(0);
-          nsValues1 = nullspace->getDataNonConst(1);
-          nsValues2 = nullspace->getDataNonConst(2);
-          nsValues3 = nullspace->getDataNonConst(3);
-          nsValues4 = nullspace->getDataNonConst(4);
-          nsValues5 = nullspace->getDataNonConst(5);
-          for (int j=0; j<nDOFs; j+=numPDEs_) {
-            // translation
-            nsValues0[j+0] = 1.0;
-            nsValues1[j+1] = 1.0;
-            nsValues2[j+2] = 1.0;
-            // rotate around z-axis (x-y plane)
-            nsValues3[j+0] = -(ynodes[j]-cy);
-            nsValues3[j+1] =  (xnodes[j]-cx);
-            // rotate around x-axis (y-z plane)
-            nsValues4[j+1] = -(znodes[j]-cz);
-            nsValues4[j+2] =  (ynodes[j]-cy);
-            // rotate around y-axis (x-z plane)
-            nsValues5[j+0] =  (znodes[j]-cz);
-            nsValues5[j+2] = -(xnodes[j]-cx);
-          }
+      } else if (numPDEs_ == 3) {
+        xnodes    = Coords->getDataNonConst(0);
+        ynodes    = Coords->getDataNonConst(1);
+        znodes    = Coords->getDataNonConst(2);
+        cx        = Coords->getVector(0)->meanValue();
+        cy        = Coords->getVector(1)->meanValue();
+        cz        = Coords->getVector(2)->meanValue();
+        nsValues0 = nullspace->getDataNonConst(0);
+        nsValues1 = nullspace->getDataNonConst(1);
+        nsValues2 = nullspace->getDataNonConst(2);
+        nsValues3 = nullspace->getDataNonConst(3);
+        nsValues4 = nullspace->getDataNonConst(4);
+        nsValues5 = nullspace->getDataNonConst(5);
+        for (int j = 0; j < nDOFs; j += numPDEs_) {
+          // translation
+          nsValues0[j + 0] = 1.0;
+          nsValues1[j + 1] = 1.0;
+          nsValues2[j + 2] = 1.0;
+          // rotate around z-axis (x-y plane)
+          nsValues3[j + 0] = -(ynodes[j] - cy);
+          nsValues3[j + 1] = (xnodes[j] - cx);
+          // rotate around x-axis (y-z plane)
+          nsValues4[j + 1] = -(znodes[j] - cz);
+          nsValues4[j + 2] = (ynodes[j] - cy);
+          // rotate around y-axis (x-z plane)
+          nsValues5[j + 0] = (znodes[j] - cz);
+          nsValues5[j + 2] = -(xnodes[j] - cx);
         }
-      } // end if "Nullspace" not available
-    }
-    else {
-      nullspace = currentLevel.Get< RCP<MultiVector> >("Nullspace", GetFactory(nspName_).get());
-    }
-    Set(currentLevel, "Nullspace", nullspace);
+      }
+    }  // end if "Nullspace" not available
+  } else {
+    nullspace = currentLevel.Get<RCP<MultiVector> >("Nullspace", GetFactory(nspName_).get());
   }
+  Set(currentLevel, "Nullspace", nullspace);
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_RIGIDBODYMODEFACTORY_SHORT
-#endif // MUELU_RIGIDBODYMODEFACTORY_DEF_HPP
+#endif  // MUELU_RIGIDBODYMODEFACTORY_DEF_HPP
diff --git a/packages/muelu/src/Utils/MueLu_TimeMonitor.cpp b/packages/muelu/src/Utils/MueLu_TimeMonitor.cpp
index 554b2b71ad95..d8c65668b53a 100644
--- a/packages/muelu/src/Utils/MueLu_TimeMonitor.cpp
+++ b/packages/muelu/src/Utils/MueLu_TimeMonitor.cpp
@@ -51,83 +51,80 @@
 
 namespace MueLu {
 
-  namespace {
-    //! Certain output formats must be observed
-    std::string cleanupLabel(const std::string& label) {
-      // Our main interest is to remove double quotes from the label
-      // since tracing formats can be written in JSON/YAML
-      auto clean_label = label;
-      std::vector<std::string> bad_values =  {"\""};
-      std::vector<std::string> good_values = {"'"};
-      for(size_t i=0; i<bad_values.size(); ++i){
-        const auto & bad_value = bad_values[i];
-        const auto & good_value = good_values[i];
-        size_t pos = 0;
-        while ( ( pos = clean_label.find( bad_value, pos ) ) != std::string::npos ){
-          clean_label.replace( pos, bad_value.size(), good_value );
-          pos += good_value.size();
-        }
-      }
-      return clean_label;
+namespace {
+//! Certain output formats must be observed
+std::string cleanupLabel(const std::string& label) {
+  // Our main interest is to remove double quotes from the label
+  // since tracing formats can be written in JSON/YAML
+  auto clean_label                     = label;
+  std::vector<std::string> bad_values  = {"\""};
+  std::vector<std::string> good_values = {"'"};
+  for (size_t i = 0; i < bad_values.size(); ++i) {
+    const auto& bad_value  = bad_values[i];
+    const auto& good_value = good_values[i];
+    size_t pos             = 0;
+    while ((pos = clean_label.find(bad_value, pos)) != std::string::npos) {
+      clean_label.replace(pos, bad_value.size(), good_value);
+      pos += good_value.size();
     }
   }
+  return clean_label;
+}
+}  // namespace
 
-  TimeMonitor::TimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel) {
-    // Inherit props from 'object'
-    SetVerbLevel      (object.GetVerbLevel());
-    SetProcRankVerbose(object.GetProcRankVerbose());
+TimeMonitor::TimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel) {
+  // Inherit props from 'object'
+  SetVerbLevel(object.GetVerbLevel());
+  SetProcRankVerbose(object.GetProcRankVerbose());
 
-    if (IsPrint(timerLevel) &&
-        /* disable timer if never printed: */ (IsPrint(RuntimeTimings) || (!IsPrint(NoTimeReport)))) {
-
-      if (!IsPrint(NoTimeReport)) {
-        // TODO: there is no function to register a timer in Teuchos::TimeMonitor after the creation of the timer. But would be useful...
-        timer_ = Teuchos::TimeMonitor::getNewTimer(cleanupLabel("MueLu: " + msg));
-      } else {
-        timer_ = rcp(new Teuchos::Time(cleanupLabel("MueLu: " + msg)));
-      }
+  if (IsPrint(timerLevel) &&
+      /* disable timer if never printed: */ (IsPrint(RuntimeTimings) || (!IsPrint(NoTimeReport)))) {
+    if (!IsPrint(NoTimeReport)) {
+      // TODO: there is no function to register a timer in Teuchos::TimeMonitor after the creation of the timer. But would be useful...
+      timer_ = Teuchos::TimeMonitor::getNewTimer(cleanupLabel("MueLu: " + msg));
+    } else {
+      timer_ = rcp(new Teuchos::Time(cleanupLabel("MueLu: " + msg)));
+    }
 
-      // Start the timer (this is what is done by Teuchos::TimeMonitor)
-      timer_->start();
-      timer_->incrementNumCalls();
+    // Start the timer (this is what is done by Teuchos::TimeMonitor)
+    timer_->start();
+    timer_->incrementNumCalls();
 #ifdef HAVE_TEUCHOS_ADD_TIME_MONITOR_TO_STACKED_TIMER
-      const auto stackedTimer = Teuchos::TimeMonitor::getStackedTimer();
-      if (nonnull(stackedTimer))
-        stackedTimer->start(timer_->name());
+    const auto stackedTimer = Teuchos::TimeMonitor::getStackedTimer();
+    if (nonnull(stackedTimer))
+      stackedTimer->start(timer_->name());
 #endif
-    }
-  } //TimeMonitor::TimeMonitor()
+  }
+}  // TimeMonitor::TimeMonitor()
 
-  TimeMonitor::TimeMonitor() {}
+TimeMonitor::TimeMonitor() {}
 
-  TimeMonitor::~TimeMonitor() {
-    // Stop the timer if present
-    if (timer_ != Teuchos::null) {
-      timer_->stop();
+TimeMonitor::~TimeMonitor() {
+  // Stop the timer if present
+  if (timer_ != Teuchos::null) {
+    timer_->stop();
 #ifdef HAVE_TEUCHOS_ADD_TIME_MONITOR_TO_STACKED_TIMER
-      try {
-        const auto stackedTimer = Teuchos::TimeMonitor::getStackedTimer();
-        if (nonnull(stackedTimer))
-          stackedTimer->stop(timer_->name());
-      }
-      catch (std::runtime_error&) {
-        std::ostringstream warning;
-        warning <<
-          "\n*********************************************************************\n"
-          "WARNING: Overlapping timers detected!\n"
-          "A TimeMonitor timer was stopped before a nested subtimer was\n"
-          "stopped. This is not allowed by the StackedTimer. This corner case\n"
-          "typically occurs if the TimeMonitor is stored in an RCP and the RCP is\n"
-          "assigned to a new timer. To disable this warning, either fix the\n"
-          "ordering of timer creation and destuction or disable the StackedTimer\n";
-        std::cout << warning.str() << std::endl;
-        Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
-      }
-#endif
+    try {
+      const auto stackedTimer = Teuchos::TimeMonitor::getStackedTimer();
+      if (nonnull(stackedTimer))
+        stackedTimer->stop(timer_->name());
+    } catch (std::runtime_error&) {
+      std::ostringstream warning;
+      warning << "\n*********************************************************************\n"
+                 "WARNING: Overlapping timers detected!\n"
+                 "A TimeMonitor timer was stopped before a nested subtimer was\n"
+                 "stopped. This is not allowed by the StackedTimer. This corner case\n"
+                 "typically occurs if the TimeMonitor is stored in an RCP and the RCP is\n"
+                 "assigned to a new timer. To disable this warning, either fix the\n"
+                 "ordering of timer creation and destuction or disable the StackedTimer\n";
+      std::cout << warning.str() << std::endl;
+      Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
     }
-  } //TimeMonitor::~TimeMonitor()
+#endif
+  }
+}  // TimeMonitor::~TimeMonitor()
 
-  template class MutuallyExclusiveTimeMonitor<FactoryBase>;
-  template class MutuallyExclusiveTimeMonitor<Level>;
+template class MutuallyExclusiveTimeMonitor<FactoryBase>;
+template class MutuallyExclusiveTimeMonitor<Level>;
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_TimeMonitor.hpp b/packages/muelu/src/Utils/MueLu_TimeMonitor.hpp
index 9a9208b44b1d..d4ac5e654d92 100644
--- a/packages/muelu/src/Utils/MueLu_TimeMonitor.hpp
+++ b/packages/muelu/src/Utils/MueLu_TimeMonitor.hpp
@@ -61,80 +61,73 @@
 
 namespace MueLu {
 
-  /*! @class TimeMonitor
-
-      @brief Integrates Teuchos::TimeMonitor with MueLu verbosity system.
-  */
-  class TimeMonitor : public BaseClass {
-
-  public:
-
-    TimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel = Timings0);
-
-    ~TimeMonitor();
-
-  protected:
-    TimeMonitor();
-
-  private:
-    RCP<Teuchos::Time> timer_;
-  }; //class TimeMonitor
-
-  //TODO: code duplication MutuallyExclusiveTimeMonitor / TimeMonitor
-
-  /*! @class MutuallyExclusiveTimeMonitor
-
-      @brief Similar to TimeMonitor, but uses MutuallyExclusiveTime objects.
-
-  */
-  template <class TagName>
-  class MutuallyExclusiveTimeMonitor : public BaseClass {
-
-  public:
-
-     /*! @brief Constructor
-
-        @param[in] object      Reference to the class instance that is creating this MutuallyExclusiveTimeMonitor.
-        @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
-        @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
-    */
-    MutuallyExclusiveTimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel = Timings0) {
-      // Inherit props from 'object'
-      SetVerbLevel      (object.GetVerbLevel());
-      SetProcRankVerbose(object.GetProcRankVerbose());
-
-      if (IsPrint(timerLevel) &&
-          /* disable timer if never printed: */ (IsPrint(RuntimeTimings) || (!IsPrint(NoTimeReport)))) {
-
-        if (!IsPrint(NoTimeReport)) {
-          timer_ = MutuallyExclusiveTime<TagName>::getNewTimer("MueLu: " + msg /*+ " (MutuallyExclusive)" */);
-        } else {
-          timer_ = rcp(new MutuallyExclusiveTime<TagName>     ("MueLu: " + msg /*+ " (MutuallyExclusive)" */));
-        }
-
-        timer_->start();
-        timer_->incrementNumCalls();
+/*! @class TimeMonitor
+
+    @brief Integrates Teuchos::TimeMonitor with MueLu verbosity system.
+*/
+class TimeMonitor : public BaseClass {
+ public:
+  TimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel = Timings0);
+
+  ~TimeMonitor();
+
+ protected:
+  TimeMonitor();
+
+ private:
+  RCP<Teuchos::Time> timer_;
+};  // class TimeMonitor
+
+// TODO: code duplication MutuallyExclusiveTimeMonitor / TimeMonitor
+
+/*! @class MutuallyExclusiveTimeMonitor
+
+    @brief Similar to TimeMonitor, but uses MutuallyExclusiveTime objects.
+
+*/
+template <class TagName>
+class MutuallyExclusiveTimeMonitor : public BaseClass {
+ public:
+  /*! @brief Constructor
+
+     @param[in] object      Reference to the class instance that is creating this MutuallyExclusiveTimeMonitor.
+     @param[in] msg         String that indicates what the Monitor is monitoring, e.g., "Build"
+     @param[in] timerLevel  Governs whether timing information should be *gathered*.  Setting this to NoTimeReport prevents the creation of timers.
+ */
+  MutuallyExclusiveTimeMonitor(const BaseClass& object, const std::string& msg, MsgType timerLevel = Timings0) {
+    // Inherit props from 'object'
+    SetVerbLevel(object.GetVerbLevel());
+    SetProcRankVerbose(object.GetProcRankVerbose());
+
+    if (IsPrint(timerLevel) &&
+        /* disable timer if never printed: */ (IsPrint(RuntimeTimings) || (!IsPrint(NoTimeReport)))) {
+      if (!IsPrint(NoTimeReport)) {
+        timer_ = MutuallyExclusiveTime<TagName>::getNewTimer("MueLu: " + msg /*+ " (MutuallyExclusive)" */);
+      } else {
+        timer_ = rcp(new MutuallyExclusiveTime<TagName>("MueLu: " + msg /*+ " (MutuallyExclusive)" */));
       }
-    }
 
-    ~MutuallyExclusiveTimeMonitor() {
-      // Stop the timer if present
-      if (timer_ != Teuchos::null)
-        timer_->stop();
+      timer_->start();
+      timer_->incrementNumCalls();
     }
+  }
 
-  protected:
-    MutuallyExclusiveTimeMonitor() { }
-
-  private:
-    RCP<MutuallyExclusiveTime<TagName> > timer_; // keep a reference on the timer to print stats if RuntimeTimings=ON //TODO:use base class instead
-  };
+  ~MutuallyExclusiveTimeMonitor() {
+    // Stop the timer if present
+    if (timer_ != Teuchos::null)
+      timer_->stop();
+  }
 
-  extern template class MutuallyExclusiveTimeMonitor<FactoryBase>;
-  extern template class MutuallyExclusiveTimeMonitor<Level>;
+ protected:
+  MutuallyExclusiveTimeMonitor() {}
 
-} // namespace MueLu
+ private:
+  RCP<MutuallyExclusiveTime<TagName> > timer_;  // keep a reference on the timer to print stats if RuntimeTimings=ON //TODO:use base class instead
+};
 
-#endif // MUELU_TIMEMONITOR_HPP
+extern template class MutuallyExclusiveTimeMonitor<FactoryBase>;
+extern template class MutuallyExclusiveTimeMonitor<Level>;
 
+}  // namespace MueLu
 
+#endif  // MUELU_TIMEMONITOR_HPP
diff --git a/packages/muelu/src/Utils/MueLu_Utilities.cpp b/packages/muelu/src/Utils/MueLu_Utilities.cpp
index ec078bc096d7..4459c8f8fc5e 100644
--- a/packages/muelu/src/Utils/MueLu_Utilities.cpp
+++ b/packages/muelu/src/Utils/MueLu_Utilities.cpp
@@ -62,270 +62,248 @@
 #endif
 #endif
 
-
-
 namespace MueLu {
 
-  long ExtractNonSerializableData(const Teuchos::ParameterList& inList, Teuchos::ParameterList& serialList, Teuchos::ParameterList& nonSerialList) {
-    using Teuchos::ParameterList;
+long ExtractNonSerializableData(const Teuchos::ParameterList& inList, Teuchos::ParameterList& serialList, Teuchos::ParameterList& nonSerialList) {
+  using Teuchos::ParameterList;
 
-    long maxLevel = 0;
+  long maxLevel = 0;
 
-    for (ParameterList::ConstIterator inListEntry = inList.begin(); inListEntry != inList.end(); inListEntry++) {
-      const std::string& levelName = inListEntry->first;
+  for (ParameterList::ConstIterator inListEntry = inList.begin(); inListEntry != inList.end(); inListEntry++) {
+    const std::string& levelName = inListEntry->first;
 
-      // Check for match of the form "level X" where X is a positive integer
-      if (inList.isSublist(levelName) && ((levelName.find("level ") == 0 && levelName.size() > 6) || levelName.find("user data") == 0)) {
-        int levelID = strtol(levelName.substr(6).c_str(), 0, 0);
-        bool userFlag = true;
-        if(levelName.find("user data") == std::string::npos) { // if "user data" is not found in levelName, switch userFlag and set levelID
-          userFlag = false;
-          levelID = strtol(levelName.substr(6).c_str(), 0, 0);
-          if (maxLevel < levelID)
-            maxLevel = levelID;
-        }
+    // Check for match of the form "level X" where X is a positive integer
+    if (inList.isSublist(levelName) && ((levelName.find("level ") == 0 && levelName.size() > 6) || levelName.find("user data") == 0)) {
+      int levelID   = strtol(levelName.substr(6).c_str(), 0, 0);
+      bool userFlag = true;
+      if (levelName.find("user data") == std::string::npos) {  // if "user data" is not found in levelName, switch userFlag and set levelID
+        userFlag = false;
+        levelID  = strtol(levelName.substr(6).c_str(), 0, 0);
+        if (maxLevel < levelID)
+          maxLevel = levelID;
+      }
 
-        // Split the sublist
-        const ParameterList& levelList = inList.sublist(levelName);
-        for (ParameterList::ConstIterator levelListEntry = levelList.begin(); levelListEntry != levelList.end(); levelListEntry++) {
-          const std::string& name = levelListEntry->first;
-          if (name == "A" || name == "P" || name == "R"  || name== "M" || name == "Mdiag" || name == "K" || name == "Nullspace" || name == "Coordinates"
-              || name == "D0" || name == "Dk_1" ||name == "Dk_2"
-              || name == "Mk_one" || name == "Mk_1_one" || name == "M1_beta" || name == "M1_alpha"
-              || name == "invMk_1_invBeta" || name == "invMk_2_invAlpha"
-              || name == "M1" || name == "Ms" || name == "M0inv"
-              || name == "Pnodal" || name == "NodeMatrix" || name == "NodeAggMatrix"
-              || name == "Node Comm" || name == "DualNodeID2PrimalNodeID"
-#ifdef HAVE_MUELU_INTREPID2 // For the IntrepidPCoarsenFactory
-              || name == "pcoarsen: element to node map"
+      // Split the sublist
+      const ParameterList& levelList = inList.sublist(levelName);
+      for (ParameterList::ConstIterator levelListEntry = levelList.begin(); levelListEntry != levelList.end(); levelListEntry++) {
+        const std::string& name = levelListEntry->first;
+        if (name == "A" || name == "P" || name == "R" || name == "M" || name == "Mdiag" || name == "K" || name == "Nullspace" || name == "Coordinates" || name == "D0" || name == "Dk_1" || name == "Dk_2" || name == "Mk_one" || name == "Mk_1_one" || name == "M1_beta" || name == "M1_alpha" || name == "invMk_1_invBeta" || name == "invMk_2_invAlpha" || name == "M1" || name == "Ms" || name == "M0inv" || name == "Pnodal" || name == "NodeMatrix" || name == "NodeAggMatrix" || name == "Node Comm" || name == "DualNodeID2PrimalNodeID"
+#ifdef HAVE_MUELU_INTREPID2  // For the IntrepidPCoarsenFactory
+            || name == "pcoarsen: element to node map"
 #endif
-              || name == "output stream"
-              )
-          {
-            nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
-          }
+            || name == "output stream") {
+          nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
+        }
 #ifdef HAVE_MUELU_MATLAB
-          else if(!userFlag && IsParamMuemexVariable(name))
-          {
-            nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
-          }
+        else if (!userFlag && IsParamMuemexVariable(name)) {
+          nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
+        }
 #endif
-          else if( userFlag && IsParamValidVariable(name)) {
-            nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
-          } else {
-            serialList.sublist(levelName).setEntry(name, levelListEntry->second);
-          }
+        else if (userFlag && IsParamValidVariable(name)) {
+          nonSerialList.sublist(levelName).setEntry(name, levelListEntry->second);
+        } else {
+          serialList.sublist(levelName).setEntry(name, levelListEntry->second);
         }
-
-      } else {
-        serialList.setEntry(inListEntry->first, inListEntry->second);
       }
-    }
 
-    return maxLevel;
+    } else {
+      serialList.setEntry(inListEntry->first, inListEntry->second);
+    }
   }
 
-  void TokenizeStringAndStripWhiteSpace(const std::string& stream, std::vector<std::string>& tokenList, const char* delimChars)
-  {
-    //note: default delimiter string is ","
-    // Take a comma-separated list and tokenize it, stripping out leading & trailing whitespace.  Then add to tokenList
-    char* buf = (char*) malloc(stream.size() + 1);
-    strcpy(buf, stream.c_str());
-    char* token = strtok(buf, delimChars);
-    if(token == NULL)
-    {
-      free(buf);
-      return;
-    }
-    while(token)
-    {
-      //token points to start of string to add to tokenList
-      //remove front whitespace...
-      char* tokStart = token;
-      char* tokEnd = token + strlen(token) - 1;
-      while(*tokStart == ' ' && tokStart < tokEnd)
-        tokStart++;
-      while(*tokEnd == ' ' && tokStart < tokEnd)
-        tokEnd--;
-      tokEnd++;
-      if(tokStart < tokEnd)
-      {
-        std::string finishedToken(tokStart, tokEnd - tokStart); //use the constructor that takes a certain # of chars
-        tokenList.push_back(finishedToken);
-      }
-      token = strtok(NULL, delimChars);
-    }
+  return maxLevel;
+}
+
+void TokenizeStringAndStripWhiteSpace(const std::string& stream, std::vector<std::string>& tokenList, const char* delimChars) {
+  // note: default delimiter string is ","
+  //  Take a comma-separated list and tokenize it, stripping out leading & trailing whitespace.  Then add to tokenList
+  char* buf = (char*)malloc(stream.size() + 1);
+  strcpy(buf, stream.c_str());
+  char* token = strtok(buf, delimChars);
+  if (token == NULL) {
     free(buf);
+    return;
   }
-
-  bool IsParamMuemexVariable(const std::string& name)
-  {
-    //see if paramName is exactly two "words" - like "OrdinalVector myNullspace" or something
-    char* str = (char*) malloc(name.length() + 1);
-    strcpy(str, name.c_str());
-    //Strip leading and trailing whitespace
-    char* firstWord = strtok(str, " ");
-    if(!firstWord) {
-      free(str);
-      return false;
-    }
-    char* secondWord = strtok(NULL, " ");
-    if(!secondWord) {
-      free(str);
-      return false;
-    }
-    char* thirdWord = strtok(NULL, " ");
-    if(thirdWord) {
-      free(str);
-      return false;
-    }
-    //convert first word to all lowercase for case insensitive compare
-    char* tolowerIt = firstWord;
-    while(*tolowerIt)
-    {
-      *tolowerIt = (char) tolower(*tolowerIt);
-      tolowerIt++;
-    }
-    //See if the first word is one of the custom variable names
-    if(strstr(firstWord, "matrix") ||
-       strstr(firstWord, "multivector") ||
-       strstr(firstWord, "map") ||
-       strstr(firstWord, "ordinalvector") ||
-       strstr(firstWord, "int") ||
-       strstr(firstWord, "scalar") ||
-       strstr(firstWord, "double") ||
-       strstr(firstWord, "complex") ||
-       strstr(firstWord, "string"))
-      //Add name to list of keys to remove
-    {
-      free(str);
-      return true;
-    }
-    else
-    {
-      free(str);
-      return false;
+  while (token) {
+    // token points to start of string to add to tokenList
+    // remove front whitespace...
+    char* tokStart = token;
+    char* tokEnd   = token + strlen(token) - 1;
+    while (*tokStart == ' ' && tokStart < tokEnd)
+      tokStart++;
+    while (*tokEnd == ' ' && tokStart < tokEnd)
+      tokEnd--;
+    tokEnd++;
+    if (tokStart < tokEnd) {
+      std::string finishedToken(tokStart, tokEnd - tokStart);  // use the constructor that takes a certain # of chars
+      tokenList.push_back(finishedToken);
     }
+    token = strtok(NULL, delimChars);
   }
+  free(buf);
+}
 
-bool IsParamValidVariable(const std::string& name)
+bool IsParamMuemexVariable(const std::string& name) {
+  // see if paramName is exactly two "words" - like "OrdinalVector myNullspace" or something
+  char* str = (char*)malloc(name.length() + 1);
+  strcpy(str, name.c_str());
+  // Strip leading and trailing whitespace
+  char* firstWord = strtok(str, " ");
+  if (!firstWord) {
+    free(str);
+    return false;
+  }
+  char* secondWord = strtok(NULL, " ");
+  if (!secondWord) {
+    free(str);
+    return false;
+  }
+  char* thirdWord = strtok(NULL, " ");
+  if (thirdWord) {
+    free(str);
+    return false;
+  }
+  // convert first word to all lowercase for case insensitive compare
+  char* tolowerIt = firstWord;
+  while (*tolowerIt) {
+    *tolowerIt = (char)tolower(*tolowerIt);
+    tolowerIt++;
+  }
+  // See if the first word is one of the custom variable names
+  if (strstr(firstWord, "matrix") ||
+      strstr(firstWord, "multivector") ||
+      strstr(firstWord, "map") ||
+      strstr(firstWord, "ordinalvector") ||
+      strstr(firstWord, "int") ||
+      strstr(firstWord, "scalar") ||
+      strstr(firstWord, "double") ||
+      strstr(firstWord, "complex") ||
+      strstr(firstWord, "string"))
+  // Add name to list of keys to remove
   {
-    //see if paramName is exactly two "words" - like "OrdinalVector myNullspace" or something
-    char* str = (char*) malloc(name.length() + 1);
-    strcpy(str, name.c_str());
-    //Strip leading and trailing whitespace
-    char* firstWord = strtok(str, " ");
-    if(!firstWord) {
-      free(str);
-      return false;
-    }
-    char* secondWord = strtok(NULL, " ");
-    if(!secondWord) {
-      free(str);
-      return false;
-    }
-    char* thirdWord = strtok(NULL, " ");
-    if(thirdWord) {
-      free(str);
-      return false;
-    }
-    //convert first word to all lowercase for case insensitive compare
-    char* tolowerIt = firstWord;
-    while(*tolowerIt)
-    {
-      *tolowerIt = (char) tolower(*tolowerIt);
-      tolowerIt++;
-    }
-    //See if the first word is one of the custom variable names
-    if(strstr(firstWord, "matrix") ||
-       strstr(firstWord, "multivector") ||
-       strstr(firstWord, "map") ||
-       strstr(firstWord, "ordinalvector") ||
-       strstr(firstWord, "int") ||
-       strstr(firstWord, "scalar") ||
-       strstr(firstWord, "double") ||
-       strstr(firstWord, "complex") ||
-       strstr(firstWord, "string") ||
-       strstr(firstWord, "array<go>") ||
-       strstr(firstWord, "array<lo>") ||
-       strstr(firstWord, "arrayrcp<lo>") ||
-       strstr(firstWord, "arrayrcp<go>"))
-      //Add name to list of keys to remove
-    {
-      free(str);
-      return true;
-    }
-    else
-    {
-      free(str);
-      return false;
-    }
+    free(str);
+    return true;
+  } else {
+    free(str);
+    return false;
   }
+}
 
+bool IsParamValidVariable(const std::string& name) {
+  // see if paramName is exactly two "words" - like "OrdinalVector myNullspace" or something
+  char* str = (char*)malloc(name.length() + 1);
+  strcpy(str, name.c_str());
+  // Strip leading and trailing whitespace
+  char* firstWord = strtok(str, " ");
+  if (!firstWord) {
+    free(str);
+    return false;
+  }
+  char* secondWord = strtok(NULL, " ");
+  if (!secondWord) {
+    free(str);
+    return false;
+  }
+  char* thirdWord = strtok(NULL, " ");
+  if (thirdWord) {
+    free(str);
+    return false;
+  }
+  // convert first word to all lowercase for case insensitive compare
+  char* tolowerIt = firstWord;
+  while (*tolowerIt) {
+    *tolowerIt = (char)tolower(*tolowerIt);
+    tolowerIt++;
+  }
+  // See if the first word is one of the custom variable names
+  if (strstr(firstWord, "matrix") ||
+      strstr(firstWord, "multivector") ||
+      strstr(firstWord, "map") ||
+      strstr(firstWord, "ordinalvector") ||
+      strstr(firstWord, "int") ||
+      strstr(firstWord, "scalar") ||
+      strstr(firstWord, "double") ||
+      strstr(firstWord, "complex") ||
+      strstr(firstWord, "string") ||
+      strstr(firstWord, "array<go>") ||
+      strstr(firstWord, "array<lo>") ||
+      strstr(firstWord, "arrayrcp<lo>") ||
+      strstr(firstWord, "arrayrcp<go>"))
+  // Add name to list of keys to remove
+  {
+    free(str);
+    return true;
+  } else {
+    free(str);
+    return false;
+  }
+}
 
-   // Generates a communicator whose only members are other ranks of the baseComm on my node
-  Teuchos::RCP<const Teuchos::Comm<int> > GenerateNodeComm(RCP<const Teuchos::Comm<int> > & baseComm, int &NodeId, const int reductionFactor) {
+// Generates a communicator whose only members are other ranks of the baseComm on my node
+Teuchos::RCP<const Teuchos::Comm<int> > GenerateNodeComm(RCP<const Teuchos::Comm<int> >& baseComm, int& NodeId, const int reductionFactor) {
 #ifdef HAVE_MPI
-       int numRanks = baseComm->getSize();
-       if(numRanks == 1) {NodeId = baseComm->getRank(); return baseComm;}
+  int numRanks = baseComm->getSize();
+  if (numRanks == 1) {
+    NodeId = baseComm->getRank();
+    return baseComm;
+  }
 
-       // Get an integer from the hostname
-       char hostname[MPI_MAX_PROCESSOR_NAME];
-       int len;
-       MPI_Get_processor_name(hostname,&len);
-       struct hostent * host = gethostbyname(hostname);
-       int myaddr = (int) inet_addr(inet_ntoa(*(struct in_addr *)host->h_addr));
+  // Get an integer from the hostname
+  char hostname[MPI_MAX_PROCESSOR_NAME];
+  int len;
+  MPI_Get_processor_name(hostname, &len);
+  struct hostent* host = gethostbyname(hostname);
+  int myaddr           = (int)inet_addr(inet_ntoa(*(struct in_addr*)host->h_addr));
 
-       // All-to-all exchange of address integers
-       std::vector<int> addressList(numRanks);
-       Teuchos::gatherAll(*baseComm,1,&myaddr,numRanks,&addressList[0]);
+  // All-to-all exchange of address integers
+  std::vector<int> addressList(numRanks);
+  Teuchos::gatherAll(*baseComm, 1, &myaddr, numRanks, &addressList[0]);
 
-       // Sort!
-       std::sort(addressList.begin(),addressList.end());
+  // Sort!
+  std::sort(addressList.begin(), addressList.end());
 
-       // Find which node I'm on (and stop when I've done that)
-       int numNodes = 0;
-       for(int i=0, prev=addressList[0]; i<numRanks && prev != myaddr; i++) {
-         if(prev != addressList[i]) {
-           prev = addressList[i];
-           numNodes++;
-         }
-       }
-       NodeId = numNodes;
+  // Find which node I'm on (and stop when I've done that)
+  int numNodes = 0;
+  for (int i = 0, prev = addressList[0]; i < numRanks && prev != myaddr; i++) {
+    if (prev != addressList[i]) {
+      prev = addressList[i];
+      numNodes++;
+    }
+  }
+  NodeId = numNodes;
 
-       // Generate nodal communicator
-       Teuchos::RCP<const Teuchos::Comm<int> > newComm =  baseComm->split(NodeId,baseComm->getRank());
+  // Generate nodal communicator
+  Teuchos::RCP<const Teuchos::Comm<int> > newComm = baseComm->split(NodeId, baseComm->getRank());
 
-       // If we want to divide nodes up (for really beefy nodes), we do so here
-       if(reductionFactor != 1) {
-         // Find # cores per node
-         int lastI = 0;
-         int coresPerNode = 0;
-         for(int i=0, prev=addressList[0]; i<numRanks; i++) {
-           if(prev != addressList[i]) {
-             prev = addressList[i];
-             coresPerNode = std::max(i - lastI, coresPerNode);
-             lastI = i;
-           }
-         }
-         coresPerNode = std::max(numRanks - lastI, coresPerNode);
+  // If we want to divide nodes up (for really beefy nodes), we do so here
+  if (reductionFactor != 1) {
+    // Find # cores per node
+    int lastI        = 0;
+    int coresPerNode = 0;
+    for (int i = 0, prev = addressList[0]; i < numRanks; i++) {
+      if (prev != addressList[i]) {
+        prev         = addressList[i];
+        coresPerNode = std::max(i - lastI, coresPerNode);
+        lastI        = i;
+      }
+    }
+    coresPerNode = std::max(numRanks - lastI, coresPerNode);
 
-         // Can we chop that up?
-         if(coresPerNode % reductionFactor != 0)
-           throw std::runtime_error("Reduction factor does not evently divide # cores per node");
-         int reducedCPN = coresPerNode / reductionFactor;
-         int nodeDivision = newComm->getRank() / reducedCPN;
+    // Can we chop that up?
+    if (coresPerNode % reductionFactor != 0)
+      throw std::runtime_error("Reduction factor does not evently divide # cores per node");
+    int reducedCPN   = coresPerNode / reductionFactor;
+    int nodeDivision = newComm->getRank() / reducedCPN;
 
-         NodeId = numNodes * reductionFactor + nodeDivision;
-         newComm =  baseComm->split(NodeId,baseComm->getRank());
-       }
+    NodeId  = numNodes * reductionFactor + nodeDivision;
+    newComm = baseComm->split(NodeId, baseComm->getRank());
+  }
 
-       return newComm;
+  return newComm;
 #else
-       NodeId = baseComm->getRank();
-       return baseComm;
+  NodeId = baseComm->getRank();
+  return baseComm;
 #endif
-    }
-
+}
 
-} // namespace MueLu
+}  // namespace MueLu
diff --git a/packages/muelu/src/Utils/MueLu_UtilitiesBase_decl.hpp b/packages/muelu/src/Utils/MueLu_UtilitiesBase_decl.hpp
index 89394247fd54..dcd82e79d075 100644
--- a/packages/muelu/src/Utils/MueLu_UtilitiesBase_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_UtilitiesBase_decl.hpp
@@ -74,449 +74,429 @@
 #include <Xpetra_Vector_fwd.hpp>
 #include <Xpetra_VectorFactory_fwd.hpp>
 
-
 namespace MueLu {
 
-  // MPI helpers
-#define MueLu_sumAll(rcpComm, in, out)                                  \
+// MPI helpers
+#define MueLu_sumAll(rcpComm, in, out) \
   Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_SUM, in, Teuchos::outArg(out))
-#define MueLu_minAll(rcpComm, in, out)                                  \
+#define MueLu_minAll(rcpComm, in, out) \
   Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MIN, in, Teuchos::outArg(out))
-#define MueLu_maxAll(rcpComm, in, out)                                  \
+#define MueLu_maxAll(rcpComm, in, out) \
   Teuchos::reduceAll(*rcpComm, Teuchos::REDUCE_MAX, in, Teuchos::outArg(out))
 
-  /*!
-    @class UtilitiesBase
-    @brief MueLu utility class.
-
-    This class provides a number of static helper methods. Some are temporary and will eventually
-    go away, while others should be moved to Xpetra.
-  */
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class UtilitiesBase {
-  public:
+/*!
+  @class UtilitiesBase
+  @brief MueLu utility class.
+
+  This class provides a number of static helper methods. Some are temporary and will eventually
+  go away, while others should be moved to Xpetra.
+*/
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class UtilitiesBase {
+ public:
 #undef MUELU_UTILITIESBASE_SHORT
 #include "MueLu_UseShortNames.hpp"
-  public:
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
-
-
-    static RCP<Matrix>                Crs2Op(RCP<CrsMatrix> Op);
-
-    /*! @brief Threshold a matrix
-
-      Returns matrix filtered with a threshold value.
-
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static RCP<CrsMatrixWrap> GetThresholdedMatrix(const RCP<Matrix>& Ain, const Scalar threshold, const bool keepDiagonal=true, const GlobalOrdinal expectedNNZperRow=-1);
-
-    /*! @brief Threshold a graph
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
 
-      Returns graph filtered with a threshold value.
+  static RCP<Matrix> Crs2Op(RCP<CrsMatrix> Op);
 
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > GetThresholdedGraph(const RCP<Matrix>& A, const Magnitude threshold, const GlobalOrdinal expectedNNZperRow=-1);
+  /*! @brief Threshold a matrix
 
-    /*! @brief Extract Matrix Diagonal
-
-      Returns Matrix diagonal in ArrayRCP.
-
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static Teuchos::ArrayRCP<Scalar> GetMatrixDiagonal_arcp(const Matrix& A);
-
-
-     /*! @brief Extract Matrix Diagonal
-
-    Returns Matrix diagonal in RCP<Vector>.
+    Returns matrix filtered with a threshold value.
 
     NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static RCP<Vector> GetMatrixDiagonal(const Matrix& A);
-
-    /*! @brief Extract Matrix Diagonal
-
-      Returns inverse of the Matrix diagonal in ArrayRCP.
-
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    // static RCP<Vector> GetMatrixDiagonalInverse(const Matrix& A, Magnitude tol = Teuchos::ScalarTraits<Scalar>::eps()*100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero());
-    static RCP<Vector> GetMatrixDiagonalInverse(const Matrix& A, Magnitude tol=Teuchos::ScalarTraits<Scalar>::eps()*100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero(), const bool doLumped=false);
-
-    /*! @brief Extract Matrix Diagonal of lumped matrix
-
-      Returns Matrix diagonal of lumped matrix in RCP<Vector>.
-
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static Teuchos::RCP<Vector> GetLumpedMatrixDiagonal(Matrix const & A, const bool doReciprocal = false,
-                                                        Magnitude tol = Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero()),
-                                                        Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero(),
-                                                        const bool replaceSingleEntryRowWithZero = false,
-                                                        const bool useAverageAbsDiagVal = false);
-
-    /*! @brief Return vector containing: max_{i\not=k}(-a_ik), for each for i in the matrix
-     *
-     * @param[in] A: input matrix
-     * @ret: vector containing max_{i\not=k}(-a_ik)
-     */
-
-    static Teuchos::ArrayRCP<Magnitude> GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A);
-
-    static Teuchos::ArrayRCP<Magnitude> GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> &BlockNumber);
-
-    /*! @brief Return vector containing inverse of input vector
-     *
-     * @param[in] v: input vector
-     * @param[in] tol: tolerance. If entries of input vector are smaller than tolerance they are replaced by valReplacement (see below). The default value for tol is 100*eps (machine precision)
-     * @param[in] valReplacement: Value put in for undefined entries in output vector (default: 0.0)
-     * @ret: vector containing inverse values of input vector v
-     */
-    static Teuchos::RCP<Vector> GetInverse(Teuchos::RCP<const Vector> v, Magnitude tol = Teuchos::ScalarTraits<Scalar>::eps()*100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero());
-
-    /*! @brief Extract Overlapped Matrix Diagonal
-
-      Returns overlapped Matrix diagonal in ArrayRCP.
-
-      The local overlapped diagonal has an entry for each index in A's column map.
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-    static RCP<Vector> GetMatrixOverlappedDiagonal(const Matrix& A);
-
-
-    /*! @brief Extract Overlapped Matrix Deleted Rowsum
-
-      Returns overlapped Matrix deleted Rowsum in ArrayRCP.
-
-      The local overlapped deleted Rowsum has an entry for each index in A's column map.
-      NOTE -- it's assumed that A has been fillComplete'd.
-    */
-
-    static RCP<Vector> GetMatrixOverlappedDeletedRowsum(const Matrix& A);
-
-    static RCP<Xpetra::Vector<Magnitude,LocalOrdinal,GlobalOrdinal,Node> >
-    GetMatrixOverlappedAbsDeletedRowsum(const Matrix& A);
-
-
+  */
+  static RCP<CrsMatrixWrap> GetThresholdedMatrix(const RCP<Matrix>& Ain, const Scalar threshold, const bool keepDiagonal = true, const GlobalOrdinal expectedNNZperRow = -1);
 
-    // TODO: should NOT return an Array. Definition must be changed to:
-    // - ArrayRCP<> ResidualNorm(Matrix const &Op, MultiVector const &X, MultiVector const &RHS)
-    // or
-    // - void ResidualNorm(Matrix const &Op, MultiVector const &X, MultiVector const &RHS, Array &)
-    static Teuchos::Array<Magnitude> ResidualNorm(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS);
+  /*! @brief Threshold a graph
 
-    static Teuchos::Array<Magnitude> ResidualNorm(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector & Resid);
+    Returns graph filtered with a threshold value.
 
-    static RCP<MultiVector> Residual(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS);
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
+  static RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node>> GetThresholdedGraph(const RCP<Matrix>& A, const Magnitude threshold, const GlobalOrdinal expectedNNZperRow = -1);
 
+  /*! @brief Extract Matrix Diagonal
 
-    static void Residual(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector & Resid);
+    Returns Matrix diagonal in ArrayRCP.
 
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
+  static Teuchos::ArrayRCP<Scalar> GetMatrixDiagonal_arcp(const Matrix& A);
 
-    /*! @brief Power method.
+  /*! @brief Extract Matrix Diagonal
 
-      @param A matrix
-      @param scaleByDiag if true, estimate the largest eigenvalue of \f$ D^; A \f$.
-      @param niters maximum number of iterations
-      @param tolerance stopping tolerance
-      @verbose if true, print iteration information
-      @seed  seed for randomizing initial guess
+ Returns Matrix diagonal in RCP<Vector>.
 
-      (Shamelessly grabbed from tpetra/examples.)
-    */
-    static Scalar PowerMethod(const Matrix& A, bool scaleByDiag = true,
-                              LocalOrdinal niters = 10, Magnitude tolerance = 1e-2, bool verbose = false, unsigned int seed = 123);
+ NOTE -- it's assumed that A has been fillComplete'd.
+ */
+  static RCP<Vector> GetMatrixDiagonal(const Matrix& A);
 
-    /*! @brief Power method.
+  /*! @brief Extract Matrix Diagonal
 
-      @param A matrix
-      @param diagInvVec reciprocal of matrix diagonal
-      @param niters maximum number of iterations
-      @param tolerance stopping tolerance
-      @verbose if true, print iteration information
-      @seed  seed for randomizing initial guess
+    Returns inverse of the Matrix diagonal in ArrayRCP.
 
-      (Shamelessly grabbed from tpetra/examples.)
-    */
-    static Scalar PowerMethod(const Matrix& A, const RCP<Vector> &diagInvVec,
-                              LocalOrdinal niters = 10, Magnitude tolerance = 1e-2, bool verbose = false, unsigned int seed = 123);
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
+  // static RCP<Vector> GetMatrixDiagonalInverse(const Matrix& A, Magnitude tol = Teuchos::ScalarTraits<Scalar>::eps()*100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero());
+  static RCP<Vector> GetMatrixDiagonalInverse(const Matrix& A, Magnitude tol = Teuchos::ScalarTraits<Scalar>::eps() * 100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero(), const bool doLumped = false);
 
-    static RCP<Teuchos::FancyOStream> MakeFancy(std::ostream& os);
+  /*! @brief Extract Matrix Diagonal of lumped matrix
 
-    /*! @brief Squared distance between two rows in a multivector
+    Returns Matrix diagonal of lumped matrix in RCP<Vector>.
 
-      Used for coordinate vectors.
-    */
-    static typename Teuchos::ScalarTraits<Scalar>::magnitudeType Distance2(const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>> &v, LocalOrdinal i0, LocalOrdinal i1);
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
+  static Teuchos::RCP<Vector> GetLumpedMatrixDiagonal(Matrix const& A, const bool doReciprocal = false,
+                                                      Magnitude tol                            = Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::zero()),
+                                                      Scalar valReplacement                    = Teuchos::ScalarTraits<Scalar>::zero(),
+                                                      const bool replaceSingleEntryRowWithZero = false,
+                                                      const bool useAverageAbsDiagVal          = false);
 
-    /*! @brief Weighted squared distance between two rows in a multivector
+  /*! @brief Return vector containing: max_{i\not=k}(-a_ik), for each for i in the matrix
+   *
+   * @param[in] A: input matrix
+   * @ret: vector containing max_{i\not=k}(-a_ik)
+   */
 
-      Used for coordinate vectors.
-    */
-    static typename Teuchos::ScalarTraits<Scalar>::magnitudeType Distance2(const Teuchos::ArrayView<double> & weight,const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>> &v, LocalOrdinal i0, LocalOrdinal i1);
+  static Teuchos::ArrayRCP<Magnitude> GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A);
 
+  static Teuchos::ArrayRCP<Magnitude> GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>& BlockNumber);
 
-    /*! @brief Detect Dirichlet rows
+  /*! @brief Return vector containing inverse of input vector
+   *
+   * @param[in] v: input vector
+   * @param[in] tol: tolerance. If entries of input vector are smaller than tolerance they are replaced by valReplacement (see below). The default value for tol is 100*eps (machine precision)
+   * @param[in] valReplacement: Value put in for undefined entries in output vector (default: 0.0)
+   * @ret: vector containing inverse values of input vector v
+   */
+  static Teuchos::RCP<Vector> GetInverse(Teuchos::RCP<const Vector> v, Magnitude tol = Teuchos::ScalarTraits<Scalar>::eps() * 100, Scalar valReplacement = Teuchos::ScalarTraits<Scalar>::zero());
 
-      The routine assumes, that if there is only one nonzero per row, it is on the diagonal and therefore a DBC.
-      This is safe for most of our applications, but one should be aware of that.
+  /*! @brief Extract Overlapped Matrix Diagonal
 
-      There is an alternative routine (see DetectDirichletRowsExt)
+    Returns overlapped Matrix diagonal in ArrayRCP.
 
-      @param[in] A matrix
-      @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
+    The local overlapped diagonal has an entry for each index in A's column map.
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
+  static RCP<Vector> GetMatrixOverlappedDiagonal(const Matrix& A);
 
-      @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
-    */
-    static Teuchos::ArrayRCP<const bool> DetectDirichletRows(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Magnitude& tol = Teuchos::ScalarTraits<Magnitude>::zero(), bool count_twos_as_dirichlet=false);
+  /*! @brief Extract Overlapped Matrix Deleted Rowsum
 
-    /*! @brief Detect Dirichlet rows
+    Returns overlapped Matrix deleted Rowsum in ArrayRCP.
 
-      @param[in] A matrix
-      @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
+    The local overlapped deleted Rowsum has an entry for each index in A's column map.
+    NOTE -- it's assumed that A has been fillComplete'd.
+  */
 
-      @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
-    */
-    static Kokkos::View<bool*, typename NO::device_type> DetectDirichletRows_kokkos(const Matrix& A, const Magnitude& tol = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::zero(), const bool count_twos_as_dirichlet=false);
+  static RCP<Vector> GetMatrixOverlappedDeletedRowsum(const Matrix& A);
 
+  static RCP<Xpetra::Vector<Magnitude, LocalOrdinal, GlobalOrdinal, Node>>
+  GetMatrixOverlappedAbsDeletedRowsum(const Matrix& A);
 
-    /*! @brief Detect Dirichlet rows (extended version)
+  // TODO: should NOT return an Array. Definition must be changed to:
+  // - ArrayRCP<> ResidualNorm(Matrix const &Op, MultiVector const &X, MultiVector const &RHS)
+  // or
+  // - void ResidualNorm(Matrix const &Op, MultiVector const &X, MultiVector const &RHS, Array &)
+  static Teuchos::Array<Magnitude> ResidualNorm(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS);
 
-      Look at each matrix row and mark it as Dirichlet if there is only one
-      "not small" nonzero on the diagonal. In determining whether a nonzero
-      is "not small" use
-      \f abs(A(i,j)) / sqrt(abs(diag[i]*diag[j])) > tol
+  static Teuchos::Array<Magnitude> ResidualNorm(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector& Resid);
 
-      @param[in] A matrix
-      @param[in/out] bHasZeroDiagonal Reference to boolean variable. Returns true if there is a zero on the diagonal in the local part of the Matrix. Otherwise it is false. Different processors might return a different value. There is no global reduction!
-      @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
-      @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
-    */
-    static Teuchos::ArrayRCP<const bool> DetectDirichletRowsExt(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, bool & bHasZeroDiagonal, const Magnitude& tol = Teuchos::ScalarTraits<Scalar>::zero());
+  static RCP<MultiVector> Residual(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS);
 
-    /*! @brief Find non-zero values in an ArrayRCP
-      Compares the value to 2 * machine epsilon
+  static void Residual(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector& Resid);
 
-      @param[in]  vals - ArrayRCP<const Scalar> of values to be tested
-      @param[out] nonzeros - ArrayRCP<bool> of true/false values for whether each entry in vals is nonzero
-    */
-    
-    static void FindNonZeros(const Teuchos::ArrayRCP<const Scalar> vals,
-                             Teuchos::ArrayRCP<bool> nonzeros);
+  /*! @brief Power method.
 
-    /*! @brief Find non-zero values in an ArrayRCP
-      Compares the value to 2 * machine epsilon
+    @param A matrix
+    @param scaleByDiag if true, estimate the largest eigenvalue of \f$ D^; A \f$.
+    @param niters maximum number of iterations
+    @param tolerance stopping tolerance
+    @verbose if true, print iteration information
+    @seed  seed for randomizing initial guess
 
-      @param[in]  vals - ArrayRCP<const Scalar> of values to be tested
-      @param[out] nonzeros - ArrayRCP<bool> of true/false values for whether each entry in vals is nonzero
-    */
-    static void FindNonZeros(const typename Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dual_view_type::t_dev_const_um vals,
-                             Kokkos::View<bool*, typename Node::device_type> nonzeros);
+    (Shamelessly grabbed from tpetra/examples.)
+  */
+  static Scalar PowerMethod(const Matrix& A, bool scaleByDiag = true,
+                            LocalOrdinal niters = 10, Magnitude tolerance = 1e-2, bool verbose = false, unsigned int seed = 123);
 
-    /*! @brief Detects Dirichlet columns & domains from a list of Dirichlet rows
+  /*! @brief Power method.
 
-      @param[in] A - Matrix on which to apply Dirichlet column detection
-      @param[in] dirichletRows - ArrayRCP<bool> of indicators as to which rows are Dirichlet
-      @param[out] dirichletCols - ArrayRCP<bool> of indicators as to which cols are Dirichlet
-      @param[out] dirichletDomain - ArrayRCP<bool> of indicators as to which domains are Dirichlet
-    */
+    @param A matrix
+    @param diagInvVec reciprocal of matrix diagonal
+    @param niters maximum number of iterations
+    @param tolerance stopping tolerance
+    @verbose if true, print iteration information
+    @seed  seed for randomizing initial guess
 
-    static void DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                                              const Teuchos::ArrayRCP<bool>& dirichletRows,
-                                              Teuchos::ArrayRCP<bool> dirichletCols,
-                                              Teuchos::ArrayRCP<bool> dirichletDomain);
+    (Shamelessly grabbed from tpetra/examples.)
+  */
+  static Scalar PowerMethod(const Matrix& A, const RCP<Vector>& diagInvVec,
+                            LocalOrdinal niters = 10, Magnitude tolerance = 1e-2, bool verbose = false, unsigned int seed = 123);
 
-    /*! @brief Detects Dirichlet columns & domains from a list of Dirichlet rows
+  static RCP<Teuchos::FancyOStream> MakeFancy(std::ostream& os);
 
-      @param[in] A - Matrix on which to apply Dirichlet column detection
-      @param[in] dirichletRows - View<bool> of indicators as to which rows are Dirichlet
-      @param[out] dirichletCols - View<bool> of indicators as to which cols are Dirichlet
-      @param[out] dirichletDomain - View<bool> of indicators as to which domains are Dirichlet
-    */
-    static void DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                                              const Kokkos::View<bool*, typename Node::device_type> & dirichletRows,
-                                              Kokkos::View<bool*, typename Node::device_type> dirichletCols,
-                                              Kokkos::View<bool*, typename Node::device_type> dirichletDomain);
+  /*! @brief Squared distance between two rows in a multivector
 
+    Used for coordinate vectors.
+  */
+  static typename Teuchos::ScalarTraits<Scalar>::magnitudeType Distance2(const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>>& v, LocalOrdinal i0, LocalOrdinal i1);
 
-    /*! @brief Apply Rowsum Criterion
+  /*! @brief Weighted squared distance between two rows in a multivector
 
-      Flags a row i as dirichlet if:
-    
-      \sum_{j\not=i} A_ij > A_ii * tol
+    Used for coordinate vectors.
+  */
+  static typename Teuchos::ScalarTraits<Scalar>::magnitudeType Distance2(const Teuchos::ArrayView<double>& weight, const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>>& v, LocalOrdinal i0, LocalOrdinal i1);
 
-      @param[in] A matrix
-      @param[in] rowSumTol See above
-      @param[in/out] dirichletRows boolean array.  The ith entry is true if the above criterion is satisfied (or if it was already set to true)
+  /*! @brief Detect Dirichlet rows
 
-    */
-    static void                                                                  ApplyRowSumCriterion(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Magnitude rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows);
+    The routine assumes, that if there is only one nonzero per row, it is on the diagonal and therefore a DBC.
+    This is safe for most of our applications, but one should be aware of that.
 
-    static void ApplyRowSumCriterion(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> &BlockNumber, const Magnitude rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows);
+    There is an alternative routine (see DetectDirichletRowsExt)
 
-    static void ApplyRowSumCriterion(const Matrix& A,
-                                     const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol,
-                                     Kokkos::View<bool*, typename NO::device_type> & dirichletRows);
+    @param[in] A matrix
+    @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
 
+    @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
+  */
+  static Teuchos::ArrayRCP<const bool> DetectDirichletRows(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Magnitude& tol = Teuchos::ScalarTraits<Magnitude>::zero(), bool count_twos_as_dirichlet = false);
 
-    /*! @brief Detect Dirichlet columns based on Dirichlet rows
+  /*! @brief Detect Dirichlet rows
 
-      The routine finds all column indices that are in Dirichlet rows, where Dirichlet rows are described by dirichletRows,
-      as returned by DetectDirichletRows.
+    @param[in] A matrix
+    @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
 
-      @param[in] A matrix
-      @param[in] dirichletRows array of Dirichlet rows.
+    @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
+  */
+  static Kokkos::View<bool*, typename NO::device_type> DetectDirichletRows_kokkos(const Matrix& A, const Magnitude& tol = Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::zero(), const bool count_twos_as_dirichlet = false);
 
-      @return boolean array.  The ith entry is true iff row i is a Dirichlet column.
-    */
-    static Teuchos::ArrayRCP<const bool> DetectDirichletCols(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                                                             const Teuchos::ArrayRCP<const bool>& dirichletRows);
+  /*! @brief Detect Dirichlet rows (extended version)
 
-    /*! @brief Detect Dirichlet columns based on Dirichlet rows
+    Look at each matrix row and mark it as Dirichlet if there is only one
+    "not small" nonzero on the diagonal. In determining whether a nonzero
+    is "not small" use
+    \f abs(A(i,j)) / sqrt(abs(diag[i]*diag[j])) > tol
 
-      The routine finds all column indices that are in Dirichlet rows, where Dirichlet rows are described by dirichletRows,
-      as returned by DetectDirichletRows.
+    @param[in] A matrix
+    @param[in/out] bHasZeroDiagonal Reference to boolean variable. Returns true if there is a zero on the diagonal in the local part of the Matrix. Otherwise it is false. Different processors might return a different value. There is no global reduction!
+    @param[in] tol If a row entry's magnitude is less than or equal to this tolerance, the entry is treated as zero.
+    @return boolean array.  The ith entry is true iff row i is a Dirichlet row.
+  */
+  static Teuchos::ArrayRCP<const bool> DetectDirichletRowsExt(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, bool& bHasZeroDiagonal, const Magnitude& tol = Teuchos::ScalarTraits<Scalar>::zero());
 
-      @param[in] A matrix
-      @param[in] dirichletRows array of Dirichlet rows.
+  /*! @brief Find non-zero values in an ArrayRCP
+    Compares the value to 2 * machine epsilon
 
-      @return boolean array.  The ith entry is true iff row i is a Dirichlet column.
-    */
-    static Kokkos::View<bool*, typename NO::device_type> DetectDirichletCols(const Matrix& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows);
+    @param[in]  vals - ArrayRCP<const Scalar> of values to be tested
+    @param[out] nonzeros - ArrayRCP<bool> of true/false values for whether each entry in vals is nonzero
+  */
 
+  static void FindNonZeros(const Teuchos::ArrayRCP<const Scalar> vals,
+                           Teuchos::ArrayRCP<bool> nonzeros);
 
-    /*! @brief Frobenius inner product of two matrices
+  /*! @brief Find non-zero values in an ArrayRCP
+    Compares the value to 2 * machine epsilon
 
-      Used in energy minimization algorithms
-    */
-    static Scalar Frobenius(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& B);
+    @param[in]  vals - ArrayRCP<const Scalar> of values to be tested
+    @param[out] nonzeros - ArrayRCP<bool> of true/false values for whether each entry in vals is nonzero
+  */
+  static void FindNonZeros(const typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dual_view_type::t_dev_const_um vals,
+                           Kokkos::View<bool*, typename Node::device_type> nonzeros);
 
-    /*! @brief Set seed for random number generator.
+  /*! @brief Detects Dirichlet columns & domains from a list of Dirichlet rows
 
-      Distribute the seeds evenly in [1,INT_MAX-1].  This guarantees nothing
-      about where random number streams on difference processes will intersect.
-      This does avoid overflow situations in parallel when multiplying by a PID.
-      It also avoids the pathological case of having the *same* random number stream
-      on each process.
-    */
+    @param[in] A - Matrix on which to apply Dirichlet column detection
+    @param[in] dirichletRows - ArrayRCP<bool> of indicators as to which rows are Dirichlet
+    @param[out] dirichletCols - ArrayRCP<bool> of indicators as to which cols are Dirichlet
+    @param[out] dirichletDomain - ArrayRCP<bool> of indicators as to which domains are Dirichlet
+  */
 
-    static void SetRandomSeed(const Teuchos::Comm<int> &comm);
+  static void DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                                            const Teuchos::ArrayRCP<bool>& dirichletRows,
+                                            Teuchos::ArrayRCP<bool> dirichletCols,
+                                            Teuchos::ArrayRCP<bool> dirichletDomain);
 
-    // Finds the OAZ Dirichlet rows for this matrix
-    // so far only used in IntrepidPCoarsenFactory
-    // TODO check whether we can use DetectDirichletRows instead
-    static void FindDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A,
-                                  std::vector<LocalOrdinal>& dirichletRows, bool count_twos_as_dirichlet=false);
+  /*! @brief Detects Dirichlet columns & domains from a list of Dirichlet rows
 
-    // Applies Ones-and-Zeros to matrix rows
-    // Takes a vector of row indices
-    static void ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                     const std::vector<LocalOrdinal>& dirichletRows);
+    @param[in] A - Matrix on which to apply Dirichlet column detection
+    @param[in] dirichletRows - View<bool> of indicators as to which rows are Dirichlet
+    @param[out] dirichletCols - View<bool> of indicators as to which cols are Dirichlet
+    @param[out] dirichletDomain - View<bool> of indicators as to which domains are Dirichlet
+  */
+  static void DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                                            const Kokkos::View<bool*, typename Node::device_type>& dirichletRows,
+                                            Kokkos::View<bool*, typename Node::device_type> dirichletCols,
+                                            Kokkos::View<bool*, typename Node::device_type> dirichletDomain);
 
-    // Applies Ones-and-Zeros to matrix rows
-    // Takes a Boolean array.
-    static void ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                     const Teuchos::ArrayRCP<const bool>& dirichletRows);
+  /*! @brief Apply Rowsum Criterion
 
-    static void ApplyOAZToMatrixRows(RCP<Matrix>& A, const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows);
+    Flags a row i as dirichlet if:
 
-    // Zeros out rows
-    // Takes a vector containg Dirichlet row indices
-    static void ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                  const std::vector<LocalOrdinal>& dirichletRows,
-                                  Scalar replaceWith=Teuchos::ScalarTraits<Scalar>::zero());
+    \sum_{j\not=i} A_ij > A_ii * tol
 
-    // Zeros out rows
-    // Takes a Boolean ArrayRCP
-    static void ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                  const Teuchos::ArrayRCP<const bool>& dirichletRows,
-                                  Scalar replaceWith=Teuchos::ScalarTraits<Scalar>::zero());
+    @param[in] A matrix
+    @param[in] rowSumTol See above
+    @param[in/out] dirichletRows boolean array.  The ith entry is true if the above criterion is satisfied (or if it was already set to true)
 
-    // Zeros out rows
-    // Takes a Boolean ArrayRCP
-    static void ZeroDirichletRows(Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& X,
-                                  const Teuchos::ArrayRCP<const bool>& dirichletRows,
-                                  Scalar replaceWith=Teuchos::ScalarTraits<Scalar>::zero());
+  */
+  static void ApplyRowSumCriterion(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Magnitude rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows);
 
-    static void ZeroDirichletRows(RCP<Matrix>& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows, SC replaceWith=Teuchos::ScalarTraits<SC>::zero());
+  static void ApplyRowSumCriterion(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>& BlockNumber, const Magnitude rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows);
 
-    static void ZeroDirichletRows(RCP<MultiVector>& X, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows, SC replaceWith=Teuchos::ScalarTraits<SC>::zero());
+  static void ApplyRowSumCriterion(const Matrix& A,
+                                   const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol,
+                                   Kokkos::View<bool*, typename NO::device_type>& dirichletRows);
 
-    // Zeros out columns
-    // Takes a Boolean vector
-    static void ZeroDirichletCols(Teuchos::RCP<Matrix>& A,
-                                  const Teuchos::ArrayRCP<const bool>& dirichletCols,
-                                  Scalar replaceWith=Teuchos::ScalarTraits<Scalar>::zero());
+  /*! @brief Detect Dirichlet columns based on Dirichlet rows
 
-    static void ZeroDirichletCols(RCP<Matrix>& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletCols, SC replaceWith=Teuchos::ScalarTraits<SC>::zero());
+    The routine finds all column indices that are in Dirichlet rows, where Dirichlet rows are described by dirichletRows,
+    as returned by DetectDirichletRows.
 
-    // Finds the OAZ Dirichlet rows for this matrix
-    static void FindDirichletRowsAndPropagateToCols(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A,
-                                                    Teuchos::RCP<Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> >& isDirichletRow,
-                                                    Teuchos::RCP<Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> >& isDirichletCol);
+    @param[in] A matrix
+    @param[in] dirichletRows array of Dirichlet rows.
 
+    @return boolean array.  The ith entry is true iff row i is a Dirichlet column.
+  */
+  static Teuchos::ArrayRCP<const bool> DetectDirichletCols(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                                                           const Teuchos::ArrayRCP<const bool>& dirichletRows);
 
+  /*! @brief Detect Dirichlet columns based on Dirichlet rows
 
-    //! Creates a copy of a matrix where the non-zero entries are replaced by ones.  
-    // You can use this to de-normalize a tenative prolongator, for instance
-    static RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > ReplaceNonZerosWithOnes(const RCP<Matrix> & original);
-    
+    The routine finds all column indices that are in Dirichlet rows, where Dirichlet rows are described by dirichletRows,
+    as returned by DetectDirichletRows.
 
+    @param[in] A matrix
+    @param[in] dirichletRows array of Dirichlet rows.
 
-    // This routine takes a BlockedMap and an Importer (assuming that the BlockedMap matches the source of the importer) and generates a BlockedMap corresponding
-    // to the Importer's target map.  We assume that the targetMap is unique (which, is not a strict requirement of an Importer, but is here and no, we don't check)
-    // This is largely intended to be used in repartitioning of blocked matrices
-    static RCP<const Xpetra::BlockedMap<LocalOrdinal,GlobalOrdinal,Node> > GeneratedBlockedTargetMap(const Xpetra::BlockedMap<LocalOrdinal,GlobalOrdinal,Node> & sourceBlockedMap,
-                                                                                                     const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> & Importer);
+    @return boolean array.  The ith entry is true iff row i is a Dirichlet column.
+  */
+  static Kokkos::View<bool*, typename NO::device_type> DetectDirichletCols(const Matrix& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows);
 
-    // Checks to see if the first chunk of the colMap is also the row map.  This simiplifies a bunch of
-    // operation in coarsening
-    static bool MapsAreNested(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& rowMap, const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& colMap);
+  /*! @brief Frobenius inner product of two matrices
 
+    Used in energy minimization algorithms
+  */
+  static Scalar Frobenius(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B);
 
-    /*! Perform a Cuthill-McKee (CM) or Reverse Cuthill-McKee (RCM) ordering of the local component of the matrix.
-      Kokkos-Kernels has an RCM implementation, so we reverse that here if we call CM.
-    */
-    static RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > CuthillMcKee(const Matrix &Op);
+  /*! @brief Set seed for random number generator.
 
-    /*! Perform a Reverse Cuthill-McKee (RCM) ordering of the local component of the matrix.
-     */
-    static RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > ReverseCuthillMcKee(const Matrix &Op);
+    Distribute the seeds evenly in [1,INT_MAX-1].  This guarantees nothing
+    about where random number streams on difference processes will intersect.
+    This does avoid overflow situations in parallel when multiplying by a PID.
+    It also avoids the pathological case of having the *same* random number stream
+    on each process.
+  */
 
-  }; // class UtilitiesBase
+  static void SetRandomSeed(const Teuchos::Comm<int>& comm);
+
+  // Finds the OAZ Dirichlet rows for this matrix
+  // so far only used in IntrepidPCoarsenFactory
+  // TODO check whether we can use DetectDirichletRows instead
+  static void FindDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                std::vector<LocalOrdinal>& dirichletRows, bool count_twos_as_dirichlet = false);
+
+  // Applies Ones-and-Zeros to matrix rows
+  // Takes a vector of row indices
+  static void ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                   const std::vector<LocalOrdinal>& dirichletRows);
+
+  // Applies Ones-and-Zeros to matrix rows
+  // Takes a Boolean array.
+  static void ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                   const Teuchos::ArrayRCP<const bool>& dirichletRows);
+
+  static void ApplyOAZToMatrixRows(RCP<Matrix>& A, const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows);
+
+  // Zeros out rows
+  // Takes a vector containg Dirichlet row indices
+  static void ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                const std::vector<LocalOrdinal>& dirichletRows,
+                                Scalar replaceWith = Teuchos::ScalarTraits<Scalar>::zero());
+
+  // Zeros out rows
+  // Takes a Boolean ArrayRCP
+  static void ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                const Teuchos::ArrayRCP<const bool>& dirichletRows,
+                                Scalar replaceWith = Teuchos::ScalarTraits<Scalar>::zero());
+
+  // Zeros out rows
+  // Takes a Boolean ArrayRCP
+  static void ZeroDirichletRows(Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                                const Teuchos::ArrayRCP<const bool>& dirichletRows,
+                                Scalar replaceWith = Teuchos::ScalarTraits<Scalar>::zero());
+
+  static void ZeroDirichletRows(RCP<Matrix>& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows, SC replaceWith = Teuchos::ScalarTraits<SC>::zero());
+
+  static void ZeroDirichletRows(RCP<MultiVector>& X, const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows, SC replaceWith = Teuchos::ScalarTraits<SC>::zero());
+
+  // Zeros out columns
+  // Takes a Boolean vector
+  static void ZeroDirichletCols(Teuchos::RCP<Matrix>& A,
+                                const Teuchos::ArrayRCP<const bool>& dirichletCols,
+                                Scalar replaceWith = Teuchos::ScalarTraits<Scalar>::zero());
+
+  static void ZeroDirichletCols(RCP<Matrix>& A, const Kokkos::View<const bool*, typename NO::device_type>& dirichletCols, SC replaceWith = Teuchos::ScalarTraits<SC>::zero());
+
+  // Finds the OAZ Dirichlet rows for this matrix
+  static void FindDirichletRowsAndPropagateToCols(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                                  Teuchos::RCP<Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node>>& isDirichletRow,
+                                                  Teuchos::RCP<Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node>>& isDirichletCol);
+
+  //! Creates a copy of a matrix where the non-zero entries are replaced by ones.
+  // You can use this to de-normalize a tenative prolongator, for instance
+  static RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> ReplaceNonZerosWithOnes(const RCP<Matrix>& original);
+
+  // This routine takes a BlockedMap and an Importer (assuming that the BlockedMap matches the source of the importer) and generates a BlockedMap corresponding
+  // to the Importer's target map.  We assume that the targetMap is unique (which, is not a strict requirement of an Importer, but is here and no, we don't check)
+  // This is largely intended to be used in repartitioning of blocked matrices
+  static RCP<const Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>> GeneratedBlockedTargetMap(const Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>& sourceBlockedMap,
+                                                                                                    const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>& Importer);
+
+  // Checks to see if the first chunk of the colMap is also the row map.  This simiplifies a bunch of
+  // operation in coarsening
+  static bool MapsAreNested(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& rowMap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& colMap);
+
+  /*! Perform a Cuthill-McKee (CM) or Reverse Cuthill-McKee (RCM) ordering of the local component of the matrix.
+    Kokkos-Kernels has an RCM implementation, so we reverse that here if we call CM.
+  */
+  static RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> CuthillMcKee(const Matrix& Op);
 
+  /*! Perform a Reverse Cuthill-McKee (RCM) ordering of the local component of the matrix.
+   */
+  static RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> ReverseCuthillMcKee(const Matrix& Op);
 
-  // Useful Kokkos conversions
-  template < class View, unsigned AppendValue >
-  struct AppendTrait {
-    // static_assert(false, "Error: NOT a Kokkos::View");
-  };
+};  // class UtilitiesBase
 
-  template < class MT, unsigned T >
-  struct CombineMemoryTraits {
-    // empty
-  };
+// Useful Kokkos conversions
+template <class View, unsigned AppendValue>
+struct AppendTrait {
+  // static_assert(false, "Error: NOT a Kokkos::View");
+};
 
-  template < unsigned U, unsigned T>
-  struct CombineMemoryTraits<Kokkos::MemoryTraits<U>, T> {
-    typedef Kokkos::MemoryTraits<U|T> type;
-  };
+template <class MT, unsigned T>
+struct CombineMemoryTraits {
+  // empty
+};
 
-  template < class DataType, unsigned T, class... Pack >
-  struct AppendTrait< Kokkos::View< DataType, Pack... >, T> {
-    typedef Kokkos::View< DataType, Pack... > view_type;
-    using type = Kokkos::View< DataType, typename view_type::array_layout, typename view_type::device_type, typename CombineMemoryTraits<typename view_type::memory_traits,T>::type >;
-  };
+template <unsigned U, unsigned T>
+struct CombineMemoryTraits<Kokkos::MemoryTraits<U>, T> {
+  typedef Kokkos::MemoryTraits<U | T> type;
+};
 
+template <class DataType, unsigned T, class... Pack>
+struct AppendTrait<Kokkos::View<DataType, Pack...>, T> {
+  typedef Kokkos::View<DataType, Pack...> view_type;
+  using type = Kokkos::View<DataType, typename view_type::array_layout, typename view_type::device_type, typename CombineMemoryTraits<typename view_type::memory_traits, T>::type>;
+};
 
-  ///////////////////////////////////////////
+///////////////////////////////////////////
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UTILITIESBASE_SHORT
-#endif // MUELU_UTILITIESBASE_DECL_HPP
+#endif  // MUELU_UTILITIESBASE_DECL_HPP
diff --git a/packages/muelu/src/Utils/MueLu_UtilitiesBase_def.hpp b/packages/muelu/src/Utils/MueLu_UtilitiesBase_def.hpp
index 25d44daaa599..9d5c37ab1bde 100644
--- a/packages/muelu/src/Utils/MueLu_UtilitiesBase_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_UtilitiesBase_def.hpp
@@ -72,1710 +72,1647 @@
 #include <KokkosKernels_Handle.hpp>
 #include <KokkosGraph_RCM.hpp>
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Crs2Op(RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op) {
-    if (Op.is_null())
-      return Teuchos::null;
-    return rcp(new CrsMatrixWrap(Op));
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetThresholdedMatrix(const RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& Ain, const Scalar threshold, const bool keepDiagonal, const GlobalOrdinal expectedNNZperRow) {
-
-    RCP<const Map> rowmap = Ain->getRowMap();
-    RCP<const Map> colmap = Ain->getColMap();
-    RCP<CrsMatrixWrap> Aout = rcp(new CrsMatrixWrap(rowmap, expectedNNZperRow <= 0 ? Ain->getGlobalMaxNumRowEntries() : expectedNNZperRow));
-    // loop over local rows
-    for(size_t row=0; row<Ain->getLocalNumRows(); row++)
-      {
-        size_t nnz = Ain->getNumEntriesInLocalRow(row);
-
-        Teuchos::ArrayView<const LocalOrdinal> indices;
-        Teuchos::ArrayView<const Scalar> vals;
-        Ain->getLocalRowView(row, indices, vals);
-
-        TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError, "MueLu::ThresholdAFilterFactory::Build: number of nonzeros not equal to number of indices? Error.");
-
-        Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(),Teuchos::ScalarTraits<GlobalOrdinal>::zero());
-        Teuchos::ArrayRCP<Scalar> valout(indices.size(),Teuchos::ScalarTraits<Scalar>::zero());
-        size_t nNonzeros = 0;
-        if (keepDiagonal) {
-          GlobalOrdinal glbRow = rowmap->getGlobalElement(row);
-          LocalOrdinal lclColIdx = colmap->getLocalElement(glbRow);
-          for(size_t i=0; i<(size_t)indices.size(); i++) {
-            if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold) || indices[i]==lclColIdx) {
-              indout[nNonzeros] = colmap->getGlobalElement(indices[i]); // LID -> GID (column)
-              valout[nNonzeros] = vals[i];
-              nNonzeros++;
-            }
-          }
-        } else
-          for(size_t i=0; i<(size_t)indices.size(); i++) {
-            if(Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold)) {
-              indout[nNonzeros] = colmap->getGlobalElement(indices[i]); // LID -> GID (column)
-              valout[nNonzeros] = vals[i];
-              nNonzeros++;
-            }
-          }
-
-        indout.resize(nNonzeros);
-        valout.resize(nNonzeros);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Crs2Op(RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  if (Op.is_null())
+    return Teuchos::null;
+  return rcp(new CrsMatrixWrap(Op));
+}
 
-        Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0,indout.size()), valout.view(0,valout.size()));
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetThresholdedMatrix(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& Ain, const Scalar threshold, const bool keepDiagonal, const GlobalOrdinal expectedNNZperRow) {
+  RCP<const Map> rowmap   = Ain->getRowMap();
+  RCP<const Map> colmap   = Ain->getColMap();
+  RCP<CrsMatrixWrap> Aout = rcp(new CrsMatrixWrap(rowmap, expectedNNZperRow <= 0 ? Ain->getGlobalMaxNumRowEntries() : expectedNNZperRow));
+  // loop over local rows
+  for (size_t row = 0; row < Ain->getLocalNumRows(); row++) {
+    size_t nnz = Ain->getNumEntriesInLocalRow(row);
+
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    Ain->getLocalRowView(row, indices, vals);
+
+    TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::as<size_t>(indices.size()) != nnz, Exceptions::RuntimeError, "MueLu::ThresholdAFilterFactory::Build: number of nonzeros not equal to number of indices? Error.");
+
+    Teuchos::ArrayRCP<GlobalOrdinal> indout(indices.size(), Teuchos::ScalarTraits<GlobalOrdinal>::zero());
+    Teuchos::ArrayRCP<Scalar> valout(indices.size(), Teuchos::ScalarTraits<Scalar>::zero());
+    size_t nNonzeros = 0;
+    if (keepDiagonal) {
+      GlobalOrdinal glbRow   = rowmap->getGlobalElement(row);
+      LocalOrdinal lclColIdx = colmap->getLocalElement(glbRow);
+      for (size_t i = 0; i < (size_t)indices.size(); i++) {
+        if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold) || indices[i] == lclColIdx) {
+          indout[nNonzeros] = colmap->getGlobalElement(indices[i]);  // LID -> GID (column)
+          valout[nNonzeros] = vals[i];
+          nNonzeros++;
+        }
+      }
+    } else
+      for (size_t i = 0; i < (size_t)indices.size(); i++) {
+        if (Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > Teuchos::ScalarTraits<Scalar>::magnitude(threshold)) {
+          indout[nNonzeros] = colmap->getGlobalElement(indices[i]);  // LID -> GID (column)
+          valout[nNonzeros] = vals[i];
+          nNonzeros++;
+        }
       }
-    Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
-
-    return Aout;
-  }
 
+    indout.resize(nNonzeros);
+    valout.resize(nNonzeros);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetThresholdedGraph(const RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A, const Magnitude threshold, const GlobalOrdinal expectedNNZperRow) {
+    Aout->insertGlobalValues(Ain->getRowMap()->getGlobalElement(row), indout.view(0, indout.size()), valout.view(0, valout.size()));
+  }
+  Aout->fillComplete(Ain->getDomainMap(), Ain->getRangeMap());
 
-    using STS = Teuchos::ScalarTraits<Scalar>;
-    RCP<CrsGraph> sparsityPattern = CrsGraphFactory::Build(A->getRowMap(), expectedNNZperRow <= 0 ? A->getGlobalMaxNumRowEntries() : expectedNNZperRow);
+  return Aout;
+}
 
-    RCP<Vector> diag = GetMatrixOverlappedDiagonal(*A);
-    ArrayRCP<const Scalar> D = diag->getData(0);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetThresholdedGraph(const RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A, const Magnitude threshold, const GlobalOrdinal expectedNNZperRow) {
+  using STS                     = Teuchos::ScalarTraits<Scalar>;
+  RCP<CrsGraph> sparsityPattern = CrsGraphFactory::Build(A->getRowMap(), expectedNNZperRow <= 0 ? A->getGlobalMaxNumRowEntries() : expectedNNZperRow);
 
-    for(size_t row=0; row<A->getLocalNumRows(); row++)
-      {
-        ArrayView<const LocalOrdinal> indices;
-        ArrayView<const Scalar> vals;
-        A->getLocalRowView(row, indices, vals);
+  RCP<Vector> diag         = GetMatrixOverlappedDiagonal(*A);
+  ArrayRCP<const Scalar> D = diag->getData(0);
 
-        GlobalOrdinal globalRow = A->getRowMap()->getGlobalElement(row);
-        LocalOrdinal col = A->getColMap()->getLocalElement(globalRow);
+  for (size_t row = 0; row < A->getLocalNumRows(); row++) {
+    ArrayView<const LocalOrdinal> indices;
+    ArrayView<const Scalar> vals;
+    A->getLocalRowView(row, indices, vals);
 
-        const Scalar Dk = STS::magnitude(D[col]) > 0.0 ? STS::magnitude(D[col]) : 1.0;
-        Array<GlobalOrdinal> indicesNew;
+    GlobalOrdinal globalRow = A->getRowMap()->getGlobalElement(row);
+    LocalOrdinal col        = A->getColMap()->getLocalElement(globalRow);
 
-        for(size_t i=0; i<size_t(indices.size()); i++)
-          // keep diagonal per default
-          if(col == indices[i] || STS::magnitude(STS::squareroot(Dk)*vals[i]*STS::squareroot(Dk)) > STS::magnitude(threshold))
-            indicesNew.append(A->getColMap()->getGlobalElement(indices[i]));
+    const Scalar Dk = STS::magnitude(D[col]) > 0.0 ? STS::magnitude(D[col]) : 1.0;
+    Array<GlobalOrdinal> indicesNew;
 
-        sparsityPattern->insertGlobalIndices(globalRow, ArrayView<const GlobalOrdinal>(indicesNew.data(), indicesNew.length()));
-      }
-    sparsityPattern->fillComplete();
+    for (size_t i = 0; i < size_t(indices.size()); i++)
+      // keep diagonal per default
+      if (col == indices[i] || STS::magnitude(STS::squareroot(Dk) * vals[i] * STS::squareroot(Dk)) > STS::magnitude(threshold))
+        indicesNew.append(A->getColMap()->getGlobalElement(indices[i]));
 
-    return sparsityPattern;
+    sparsityPattern->insertGlobalIndices(globalRow, ArrayView<const GlobalOrdinal>(indicesNew.data(), indicesNew.length()));
   }
+  sparsityPattern->fillComplete();
 
+  return sparsityPattern;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<Scalar>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixDiagonal_arcp(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A) {
-    size_t numRows = A.getRowMap()->getLocalNumElements();
-    Teuchos::ArrayRCP<Scalar> diag(numRows);
-    Teuchos::ArrayView<const LocalOrdinal> cols;
-    Teuchos::ArrayView<const Scalar> vals;
-    for (size_t i = 0; i < numRows; ++i) {
-      A.getLocalRowView(i, cols, vals);
-      LocalOrdinal j = 0;
-      for (; j < cols.size(); ++j) {
-        if (Teuchos::as<size_t>(cols[j]) == i) {
-          diag[i] = vals[j];
-          break;
-        }
-      }
-      if (j == cols.size()) {
-        // Diagonal entry is absent
-        diag[i] = Teuchos::ScalarTraits<Scalar>::zero();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<Scalar>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixDiagonal_arcp(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A) {
+  size_t numRows = A.getRowMap()->getLocalNumElements();
+  Teuchos::ArrayRCP<Scalar> diag(numRows);
+  Teuchos::ArrayView<const LocalOrdinal> cols;
+  Teuchos::ArrayView<const Scalar> vals;
+  for (size_t i = 0; i < numRows; ++i) {
+    A.getLocalRowView(i, cols, vals);
+    LocalOrdinal j = 0;
+    for (; j < cols.size(); ++j) {
+      if (Teuchos::as<size_t>(cols[j]) == i) {
+        diag[i] = vals[j];
+        break;
       }
     }
-    return diag;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixDiagonal(const Matrix& A) {
-    const auto rowMap = A.getRowMap();
-    auto diag = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(rowMap, true);
-
-    A.getLocalDiagCopy(*diag);
-
-    return diag;
+    if (j == cols.size()) {
+      // Diagonal entry is absent
+      diag[i] = Teuchos::ScalarTraits<Scalar>::zero();
+    }
   }
+  return diag;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixDiagonal(const Matrix& A) {
+  const auto rowMap = A.getRowMap();
+  auto diag         = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(rowMap, true);
 
-  // template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  // RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  // UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  // GetMatrixDiagonalInverse(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A, Magnitude tol, Scalar valReplacement) {
-  //   Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase::GetMatrixDiagonalInverse");
-
-  //   RCP<const Map> rowMap = A.getRowMap();
-  //   RCP<Vector> diag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap,true);
+  A.getLocalDiagCopy(*diag);
 
-  //   A.getLocalDiagCopy(*diag);
+  return diag;
+}
 
-  //   RCP<Vector> inv = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetInverse(diag, tol, valReplacement);
+// template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
+// UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+// GetMatrixDiagonalInverse(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A, Magnitude tol, Scalar valReplacement) {
+//   Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase::GetMatrixDiagonalInverse");
 
-  //   return inv;
-  // }
+//   RCP<const Map> rowMap = A.getRowMap();
+//   RCP<Vector> diag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap,true);
 
+//   A.getLocalDiagCopy(*diag);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixDiagonalInverse(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                           typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol,
-                           Scalar valReplacement,
-                           const bool doLumped) {
-    Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("Utilities::GetMatrixDiagonalInverse");
+//   RCP<Vector> inv = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetInverse(diag, tol, valReplacement);
 
-    RCP<const BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(rcpFromRef(A));
-    if (!bA.is_null()) {
-      RCP<const Map> rowMap = A.getRowMap();
-      RCP<Vector> diag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap,true);
-      A.getLocalDiagCopy(*diag);
-      RCP<Vector> inv = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetInverse(diag, tol, valReplacement);
-      return inv;
-    }
+//   return inv;
+// }
 
-    // Some useful type definitions
-    using local_matrix_type = typename Matrix::local_matrix_type;
-    // using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using value_type        = typename local_matrix_type::value_type;
-    using values_type       = typename local_matrix_type::values_type;
-    using scalar_type       = typename values_type::non_const_value_type;
-    using ordinal_type      = typename local_matrix_type::ordinal_type;
-    using execution_space   = typename local_matrix_type::execution_space;
-    // using memory_space      = typename local_matrix_type::memory_space;
-    // Be careful with this one, if using Kokkos::ArithTraits<Scalar>
-    // you are likely to run into errors when handling std::complex<>
-    // a good way to work around that is to use the following:
-    // using KAT = Kokkos::ArithTraits<Kokkos::ArithTraits<Scalar>::val_type> >
-    // here we have: value_type = Kokkos::ArithTraits<Scalar>::val_type
-    using KAT               = Kokkos::ArithTraits<value_type>;
-
-    // Get/Create distributed objects
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixDiagonalInverse(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                             typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol,
+                             Scalar valReplacement,
+                             const bool doLumped) {
+  Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("Utilities::GetMatrixDiagonalInverse");
+
+  RCP<const BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(rcpFromRef(A));
+  if (!bA.is_null()) {
     RCP<const Map> rowMap = A.getRowMap();
-    RCP<Vector> diag      = VectorFactory::Build(rowMap,false);
-
-    // Now generate local objects
-    local_matrix_type localMatrix = A.getLocalMatrixDevice();
-    auto diagVals = diag->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-    ordinal_type numRows = localMatrix.graph.numRows();
-
-    scalar_type valReplacement_dev = valReplacement;
-
-    // Note: 2019-11-21, LBV
-    // This could be implemented with a TeamPolicy over the rows
-    // and a TeamVectorRange over the entries in a row if performance
-    // becomes more important here.
-    if (!doLumped)
-      Kokkos::parallel_for("Utilities::GetMatrixDiagonalInverse",
-                           Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
-                           KOKKOS_LAMBDA(const ordinal_type rowIdx) {
-                             bool foundDiagEntry = false;
-                             auto myRow = localMatrix.rowConst(rowIdx);
-                             for(ordinal_type entryIdx = 0; entryIdx < myRow.length; ++entryIdx) {
-                               if(myRow.colidx(entryIdx) == rowIdx) {
-                                 foundDiagEntry = true;
-                                 if(KAT::magnitude(myRow.value(entryIdx)) > KAT::magnitude(tol)) {
-                                   diagVals(rowIdx, 0) = KAT::one() / myRow.value(entryIdx);
-                                 } else {
-                                   diagVals(rowIdx, 0) = valReplacement_dev;
-                                 }
-                                 break;
-                               }
-                             }
-
-                             if(!foundDiagEntry) {diagVals(rowIdx, 0) = KAT::zero();}
-                           });
-    else
-      Kokkos::parallel_for("Utilities::GetMatrixDiagonalInverse",
-                           Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
-                           KOKKOS_LAMBDA(const ordinal_type rowIdx) {
-                             auto myRow = localMatrix.rowConst(rowIdx);
-                             for(ordinal_type entryIdx = 0; entryIdx < myRow.length; ++entryIdx) {
-                               diagVals(rowIdx, 0) += KAT::magnitude(myRow.value(entryIdx));
-                             }
-                             if(KAT::magnitude(diagVals(rowIdx, 0)) > KAT::magnitude(tol))
-                               diagVals(rowIdx, 0) = KAT::one() / diagVals(rowIdx, 0);
-                             else
-                               diagVals(rowIdx, 0) = valReplacement_dev;
-
-                           });
-
-    return diag;
+    RCP<Vector> diag      = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(rowMap, true);
+    A.getLocalDiagCopy(*diag);
+    RCP<Vector> inv = MueLu::UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetInverse(diag, tol, valReplacement);
+    return inv;
   }
 
+  // Some useful type definitions
+  using local_matrix_type = typename Matrix::local_matrix_type;
+  // using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
+  using value_type      = typename local_matrix_type::value_type;
+  using values_type     = typename local_matrix_type::values_type;
+  using scalar_type     = typename values_type::non_const_value_type;
+  using ordinal_type    = typename local_matrix_type::ordinal_type;
+  using execution_space = typename local_matrix_type::execution_space;
+  // using memory_space      = typename local_matrix_type::memory_space;
+  // Be careful with this one, if using Kokkos::ArithTraits<Scalar>
+  // you are likely to run into errors when handling std::complex<>
+  // a good way to work around that is to use the following:
+  // using KAT = Kokkos::ArithTraits<Kokkos::ArithTraits<Scalar>::val_type> >
+  // here we have: value_type = Kokkos::ArithTraits<Scalar>::val_type
+  using KAT = Kokkos::ArithTraits<value_type>;
+
+  // Get/Create distributed objects
+  RCP<const Map> rowMap = A.getRowMap();
+  RCP<Vector> diag      = VectorFactory::Build(rowMap, false);
+
+  // Now generate local objects
+  local_matrix_type localMatrix = A.getLocalMatrixDevice();
+  auto diagVals                 = diag->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+  ordinal_type numRows = localMatrix.graph.numRows();
+
+  scalar_type valReplacement_dev = valReplacement;
+
+  // Note: 2019-11-21, LBV
+  // This could be implemented with a TeamPolicy over the rows
+  // and a TeamVectorRange over the entries in a row if performance
+  // becomes more important here.
+  if (!doLumped)
+    Kokkos::parallel_for(
+        "Utilities::GetMatrixDiagonalInverse",
+        Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const ordinal_type rowIdx) {
+          bool foundDiagEntry = false;
+          auto myRow          = localMatrix.rowConst(rowIdx);
+          for (ordinal_type entryIdx = 0; entryIdx < myRow.length; ++entryIdx) {
+            if (myRow.colidx(entryIdx) == rowIdx) {
+              foundDiagEntry = true;
+              if (KAT::magnitude(myRow.value(entryIdx)) > KAT::magnitude(tol)) {
+                diagVals(rowIdx, 0) = KAT::one() / myRow.value(entryIdx);
+              } else {
+                diagVals(rowIdx, 0) = valReplacement_dev;
+              }
+              break;
+            }
+          }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetLumpedMatrixDiagonal(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>  const & A, const bool doReciprocal,
-                          Magnitude tol,
-                          Scalar valReplacement,
-                          const bool replaceSingleEntryRowWithZero,
-                          const bool useAverageAbsDiagVal) {
-
-    typedef Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<Vector> diag = Teuchos::null;
-    const Scalar zero = TST::zero();
-    const Scalar one = TST::one();
-    const Scalar two = one + one;
-
-    Teuchos::RCP<const Matrix> rcpA = Teuchos::rcpFromRef(A);
-
-    RCP<const Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bA =
-      Teuchos::rcp_dynamic_cast<const Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(rcpA);
-    if(bA == Teuchos::null) {
-      RCP<const Map> rowMap = rcpA->getRowMap();
-      diag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap,true);
-
-      if(rowMap->lib() == Xpetra::UnderlyingLib::UseTpetra) {
-        Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase::GetLumpedMatrixDiagonal (Kokkos implementation)");
-        // Implement using Kokkos
-        using local_vector_type = typename Vector::dual_view_type::t_dev_um;
-        using local_matrix_type = typename Matrix::local_matrix_type;
-        using execution_space   = typename local_vector_type::execution_space;
-        // using rowmap_type       = typename local_matrix_type::row_map_type;
-        // using entries_type      = typename local_matrix_type::index_type;
-        using values_type       = typename local_matrix_type::values_type;
-        using scalar_type       = typename values_type::non_const_value_type;
-        using mag_type          = typename Kokkos::ArithTraits<scalar_type>::mag_type;
-        using KAT_S             = typename Kokkos::ArithTraits<scalar_type>;
-        using KAT_M             = typename Kokkos::ArithTraits<mag_type>;
-        using size_type         = typename local_matrix_type::non_const_size_type;
-
-        local_vector_type diag_dev = diag->getDeviceLocalView(Xpetra::Access::OverwriteAll);
-        local_matrix_type local_mat_dev = rcpA->getLocalMatrixDevice();
-        Kokkos::RangePolicy<execution_space, int> my_policy(0, static_cast<int>(diag_dev.extent(0)));
-        scalar_type valReplacement_dev = valReplacement;
-
-        if(doReciprocal) {
-          Kokkos::View<int*, execution_space> nnzPerRow("nnz per rows", diag_dev.extent(0));
-          Kokkos::View<scalar_type*, execution_space> regSum("regSum", diag_dev.extent(0));
-          Kokkos::View<mag_type, execution_space> avgAbsDiagVal_dev("avgAbsDiagVal");
-          Kokkos::View<int, execution_space> numDiagsEqualToOne_dev("numDiagsEqualToOne");
-
-          {
-          Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: parallel_for (doReciprocal)");
-          Kokkos::parallel_for("GetLumpedMatrixDiagonal", my_policy,
-                               KOKKOS_LAMBDA(const int rowIdx) {
-                                 diag_dev(rowIdx, 0) = KAT_S::zero();
-                                 for(size_type entryIdx = local_mat_dev.graph.row_map(rowIdx);
-                                     entryIdx < local_mat_dev.graph.row_map(rowIdx + 1);
-                                     ++entryIdx) {
-                                   regSum(rowIdx) += local_mat_dev.values(entryIdx);
-                                   if(KAT_M::zero() < KAT_S::abs(local_mat_dev.values(entryIdx))) {
-                                     ++nnzPerRow(rowIdx);
-                                   }
-                                   diag_dev(rowIdx, 0) += KAT_S::abs(local_mat_dev.values(entryIdx));
-                                   if(rowIdx == local_mat_dev.graph.entries(entryIdx)) {
-                                     Kokkos::atomic_add(&avgAbsDiagVal_dev(), KAT_S::abs(local_mat_dev.values(entryIdx)));
-                                   }
-                                 }
-
-                                 if(nnzPerRow(rowIdx) == 1 && KAT_S::magnitude(diag_dev(rowIdx, 0)) == KAT_M::one()) {
-                                   Kokkos::atomic_add(&numDiagsEqualToOne_dev(), 1);
-                                 }
-                               });
-
+          if (!foundDiagEntry) {
+            diagVals(rowIdx, 0) = KAT::zero();
           }
-          if (useAverageAbsDiagVal) {
-            Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: useAverageAbsDiagVal");
-            typename Kokkos::View<mag_type, execution_space>::HostMirror avgAbsDiagVal = Kokkos::create_mirror_view(avgAbsDiagVal_dev);
-            Kokkos::deep_copy(avgAbsDiagVal, avgAbsDiagVal_dev);
-            int numDiagsEqualToOne;
-            Kokkos::deep_copy(numDiagsEqualToOne, numDiagsEqualToOne_dev);
-
-            tol = TST::magnitude(100 * Teuchos::ScalarTraits<Scalar>::eps()) * (avgAbsDiagVal()-numDiagsEqualToOne) / (rowMap->getLocalNumElements()-numDiagsEqualToOne);
+        });
+  else
+    Kokkos::parallel_for(
+        "Utilities::GetMatrixDiagonalInverse",
+        Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
+        KOKKOS_LAMBDA(const ordinal_type rowIdx) {
+          auto myRow = localMatrix.rowConst(rowIdx);
+          for (ordinal_type entryIdx = 0; entryIdx < myRow.length; ++entryIdx) {
+            diagVals(rowIdx, 0) += KAT::magnitude(myRow.value(entryIdx));
           }
+          if (KAT::magnitude(diagVals(rowIdx, 0)) > KAT::magnitude(tol))
+            diagVals(rowIdx, 0) = KAT::one() / diagVals(rowIdx, 0);
+          else
+            diagVals(rowIdx, 0) = valReplacement_dev;
+        });
 
-          {
-          Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("ComputeLumpedDiagonalInverse: parallel_for (doReciprocal)");
-          Kokkos::parallel_for("ComputeLumpedDiagonalInverse", my_policy,
-                               KOKKOS_LAMBDA(const int rowIdx) {
-                                 if (replaceSingleEntryRowWithZero && nnzPerRow(rowIdx) <= 1) {
-                                   diag_dev(rowIdx, 0) = KAT_S::zero();
-                                 } else if((diag_dev(rowIdx, 0) != KAT_S::zero()) && (KAT_S::magnitude(diag_dev(rowIdx, 0)) < KAT_S::magnitude(2*regSum(rowIdx)))) {
-                                   diag_dev(rowIdx, 0) = KAT_S::one() / KAT_S::magnitude(2*regSum(rowIdx));
-                                 } else {
-                                   if(KAT_S::magnitude(diag_dev(rowIdx, 0)) > tol) {
-                                     diag_dev(rowIdx, 0) = KAT_S::one() / diag_dev(rowIdx, 0);
-                                   } else {
-                                     diag_dev(rowIdx, 0) = valReplacement_dev;
-                                   }
-                                 }
-                               });
-          }
+  return diag;
+}
 
-        } else {
-          Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: parallel_for");
-          Kokkos::parallel_for("GetLumpedMatrixDiagonal", my_policy,
-                               KOKKOS_LAMBDA(const int rowIdx) {
-                                 diag_dev(rowIdx, 0) = KAT_S::zero();
-                                 for(size_type entryIdx = local_mat_dev.graph.row_map(rowIdx);
-                                     entryIdx < local_mat_dev.graph.row_map(rowIdx + 1);
-                                     ++entryIdx) {
-                                   diag_dev(rowIdx, 0) += KAT_S::magnitude(local_mat_dev.values(entryIdx));
-                                 }
-                               });
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetLumpedMatrixDiagonal(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> const& A, const bool doReciprocal,
+                            Magnitude tol,
+                            Scalar valReplacement,
+                            const bool replaceSingleEntryRowWithZero,
+                            const bool useAverageAbsDiagVal) {
+  typedef Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<Vector> diag  = Teuchos::null;
+  const Scalar zero = TST::zero();
+  const Scalar one  = TST::one();
+  const Scalar two  = one + one;
+
+  Teuchos::RCP<const Matrix> rcpA = Teuchos::rcpFromRef(A);
+
+  RCP<const Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> bA =
+      Teuchos::rcp_dynamic_cast<const Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(rcpA);
+  if (bA == Teuchos::null) {
+    RCP<const Map> rowMap = rcpA->getRowMap();
+    diag                  = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(rowMap, true);
+
+    if (rowMap->lib() == Xpetra::UnderlyingLib::UseTpetra) {
+      Teuchos::TimeMonitor MM = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase::GetLumpedMatrixDiagonal (Kokkos implementation)");
+      // Implement using Kokkos
+      using local_vector_type = typename Vector::dual_view_type::t_dev_um;
+      using local_matrix_type = typename Matrix::local_matrix_type;
+      using execution_space   = typename local_vector_type::execution_space;
+      // using rowmap_type       = typename local_matrix_type::row_map_type;
+      // using entries_type      = typename local_matrix_type::index_type;
+      using values_type = typename local_matrix_type::values_type;
+      using scalar_type = typename values_type::non_const_value_type;
+      using mag_type    = typename Kokkos::ArithTraits<scalar_type>::mag_type;
+      using KAT_S       = typename Kokkos::ArithTraits<scalar_type>;
+      using KAT_M       = typename Kokkos::ArithTraits<mag_type>;
+      using size_type   = typename local_matrix_type::non_const_size_type;
+
+      local_vector_type diag_dev      = diag->getDeviceLocalView(Xpetra::Access::OverwriteAll);
+      local_matrix_type local_mat_dev = rcpA->getLocalMatrixDevice();
+      Kokkos::RangePolicy<execution_space, int> my_policy(0, static_cast<int>(diag_dev.extent(0)));
+      scalar_type valReplacement_dev = valReplacement;
+
+      if (doReciprocal) {
+        Kokkos::View<int*, execution_space> nnzPerRow("nnz per rows", diag_dev.extent(0));
+        Kokkos::View<scalar_type*, execution_space> regSum("regSum", diag_dev.extent(0));
+        Kokkos::View<mag_type, execution_space> avgAbsDiagVal_dev("avgAbsDiagVal");
+        Kokkos::View<int, execution_space> numDiagsEqualToOne_dev("numDiagsEqualToOne");
+
+        {
+          Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: parallel_for (doReciprocal)");
+          Kokkos::parallel_for(
+              "GetLumpedMatrixDiagonal", my_policy,
+              KOKKOS_LAMBDA(const int rowIdx) {
+                diag_dev(rowIdx, 0) = KAT_S::zero();
+                for (size_type entryIdx = local_mat_dev.graph.row_map(rowIdx);
+                     entryIdx < local_mat_dev.graph.row_map(rowIdx + 1);
+                     ++entryIdx) {
+                  regSum(rowIdx) += local_mat_dev.values(entryIdx);
+                  if (KAT_M::zero() < KAT_S::abs(local_mat_dev.values(entryIdx))) {
+                    ++nnzPerRow(rowIdx);
+                  }
+                  diag_dev(rowIdx, 0) += KAT_S::abs(local_mat_dev.values(entryIdx));
+                  if (rowIdx == local_mat_dev.graph.entries(entryIdx)) {
+                    Kokkos::atomic_add(&avgAbsDiagVal_dev(), KAT_S::abs(local_mat_dev.values(entryIdx)));
+                  }
+                }
+
+                if (nnzPerRow(rowIdx) == 1 && KAT_S::magnitude(diag_dev(rowIdx, 0)) == KAT_M::one()) {
+                  Kokkos::atomic_add(&numDiagsEqualToOne_dev(), 1);
+                }
+              });
+        }
+        if (useAverageAbsDiagVal) {
+          Teuchos::TimeMonitor MMM                                                   = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: useAverageAbsDiagVal");
+          typename Kokkos::View<mag_type, execution_space>::HostMirror avgAbsDiagVal = Kokkos::create_mirror_view(avgAbsDiagVal_dev);
+          Kokkos::deep_copy(avgAbsDiagVal, avgAbsDiagVal_dev);
+          int numDiagsEqualToOne;
+          Kokkos::deep_copy(numDiagsEqualToOne, numDiagsEqualToOne_dev);
+
+          tol = TST::magnitude(100 * Teuchos::ScalarTraits<Scalar>::eps()) * (avgAbsDiagVal() - numDiagsEqualToOne) / (rowMap->getLocalNumElements() - numDiagsEqualToOne);
         }
-      } else {
-        // Implement using Teuchos
-        Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase: GetLumpedMatrixDiagonal: (Teuchos implementation)");
-        ArrayRCP<Scalar> diagVals = diag->getDataNonConst(0);
-        Teuchos::Array<Scalar> regSum(diag->getLocalLength());
-        Teuchos::ArrayView<const LocalOrdinal> cols;
-        Teuchos::ArrayView<const Scalar> vals;
-
-        std::vector<int> nnzPerRow(rowMap->getLocalNumElements());
-
-        //FIXME 2021-10-22 JHU   If this is called with doReciprocal=false, what should the correct behavior be?  Currently,
-        //FIXME 2021-10-22 JHU   the diagonal entry is set to be the sum of the absolute values of the row entries.
 
-        const Magnitude zeroMagn = TST::magnitude(zero);
-        Magnitude avgAbsDiagVal = TST::magnitude(zero);
-        int numDiagsEqualToOne = 0;
-        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-          nnzPerRow[i] = 0;
-          rcpA->getLocalRowView(i, cols, vals);
-          diagVals[i] = zero;
-          for (LocalOrdinal j = 0; j < cols.size(); ++j) {
-            regSum[i] += vals[j];
-            const Magnitude rowEntryMagn = TST::magnitude(vals[j]);
-            if (rowEntryMagn > zeroMagn)
-              nnzPerRow[i]++;
-            diagVals[i] += rowEntryMagn;
-            if (static_cast<size_t>(cols[j]) == i)
-              avgAbsDiagVal += rowEntryMagn;
-          }
-          if (nnzPerRow[i] == 1 && TST::magnitude(diagVals[i])==1.)
-            numDiagsEqualToOne++;
+        {
+          Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("ComputeLumpedDiagonalInverse: parallel_for (doReciprocal)");
+          Kokkos::parallel_for(
+              "ComputeLumpedDiagonalInverse", my_policy,
+              KOKKOS_LAMBDA(const int rowIdx) {
+                if (replaceSingleEntryRowWithZero && nnzPerRow(rowIdx) <= 1) {
+                  diag_dev(rowIdx, 0) = KAT_S::zero();
+                } else if ((diag_dev(rowIdx, 0) != KAT_S::zero()) && (KAT_S::magnitude(diag_dev(rowIdx, 0)) < KAT_S::magnitude(2 * regSum(rowIdx)))) {
+                  diag_dev(rowIdx, 0) = KAT_S::one() / KAT_S::magnitude(2 * regSum(rowIdx));
+                } else {
+                  if (KAT_S::magnitude(diag_dev(rowIdx, 0)) > tol) {
+                    diag_dev(rowIdx, 0) = KAT_S::one() / diag_dev(rowIdx, 0);
+                  } else {
+                    diag_dev(rowIdx, 0) = valReplacement_dev;
+                  }
+                }
+              });
         }
-        if (useAverageAbsDiagVal)
-          tol = TST::magnitude(100 * Teuchos::ScalarTraits<Scalar>::eps()) * (avgAbsDiagVal-numDiagsEqualToOne) / (rowMap->getLocalNumElements()-numDiagsEqualToOne);
-        if (doReciprocal) {
-          for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-            if (replaceSingleEntryRowWithZero && nnzPerRow[i] <= static_cast<int>(1))
-              diagVals[i] = zero;
-            else if ((diagVals[i] != zero) && (TST::magnitude(diagVals[i]) < TST::magnitude(two*regSum[i])))
-              diagVals[i] = one / TST::magnitude((two*regSum[i]));
-            else {
-              if(TST::magnitude(diagVals[i]) > tol)
-                diagVals[i] = one / diagVals[i];
-              else {
-                diagVals[i] = valReplacement;
+
+      } else {
+        Teuchos::TimeMonitor MMM = *Teuchos::TimeMonitor::getNewTimer("GetLumpedMatrixDiagonal: parallel_for");
+        Kokkos::parallel_for(
+            "GetLumpedMatrixDiagonal", my_policy,
+            KOKKOS_LAMBDA(const int rowIdx) {
+              diag_dev(rowIdx, 0) = KAT_S::zero();
+              for (size_type entryIdx = local_mat_dev.graph.row_map(rowIdx);
+                   entryIdx < local_mat_dev.graph.row_map(rowIdx + 1);
+                   ++entryIdx) {
+                diag_dev(rowIdx, 0) += KAT_S::magnitude(local_mat_dev.values(entryIdx));
               }
-            }
-          }
-        }
+            });
       }
     } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(doReciprocal, Xpetra::Exceptions::RuntimeError,
-                                 "UtilitiesBase::GetLumpedMatrixDiagonal(): extracting reciprocal of diagonal of a blocked matrix is not supported");
-      diag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(bA->getRangeMapExtractor()->getFullMap(),true);
-
-      for (size_t row = 0; row < bA->Rows(); ++row) {
-        for (size_t col = 0; col < bA->Cols(); ++col) {
-          if (!bA->getMatrix(row,col).is_null()) {
-            // if we are in Thyra mode, but the block (row,row) is again a blocked operator, we have to use (pseudo) Xpetra-style GIDs with offset!
-            bool bThyraMode = bA->getRangeMapExtractor()->getThyraMode() && (Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(bA->getMatrix(row,col)) == Teuchos::null);
-            RCP<Vector> ddtemp = bA->getRangeMapExtractor()->ExtractVector(diag,row,bThyraMode);
-            RCP<const Vector> dd = GetLumpedMatrixDiagonal(*(bA->getMatrix(row,col)));
-            ddtemp->update(Teuchos::as<Scalar>(1.0),*dd,Teuchos::as<Scalar>(1.0));
-            bA->getRangeMapExtractor()->InsertVector(ddtemp,row,diag,bThyraMode);
+      // Implement using Teuchos
+      Teuchos::TimeMonitor MMM  = *Teuchos::TimeMonitor::getNewTimer("UtilitiesBase: GetLumpedMatrixDiagonal: (Teuchos implementation)");
+      ArrayRCP<Scalar> diagVals = diag->getDataNonConst(0);
+      Teuchos::Array<Scalar> regSum(diag->getLocalLength());
+      Teuchos::ArrayView<const LocalOrdinal> cols;
+      Teuchos::ArrayView<const Scalar> vals;
+
+      std::vector<int> nnzPerRow(rowMap->getLocalNumElements());
+
+      // FIXME 2021-10-22 JHU   If this is called with doReciprocal=false, what should the correct behavior be?  Currently,
+      // FIXME 2021-10-22 JHU   the diagonal entry is set to be the sum of the absolute values of the row entries.
+
+      const Magnitude zeroMagn = TST::magnitude(zero);
+      Magnitude avgAbsDiagVal  = TST::magnitude(zero);
+      int numDiagsEqualToOne   = 0;
+      for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+        nnzPerRow[i] = 0;
+        rcpA->getLocalRowView(i, cols, vals);
+        diagVals[i] = zero;
+        for (LocalOrdinal j = 0; j < cols.size(); ++j) {
+          regSum[i] += vals[j];
+          const Magnitude rowEntryMagn = TST::magnitude(vals[j]);
+          if (rowEntryMagn > zeroMagn)
+            nnzPerRow[i]++;
+          diagVals[i] += rowEntryMagn;
+          if (static_cast<size_t>(cols[j]) == i)
+            avgAbsDiagVal += rowEntryMagn;
+        }
+        if (nnzPerRow[i] == 1 && TST::magnitude(diagVals[i]) == 1.)
+          numDiagsEqualToOne++;
+      }
+      if (useAverageAbsDiagVal)
+        tol = TST::magnitude(100 * Teuchos::ScalarTraits<Scalar>::eps()) * (avgAbsDiagVal - numDiagsEqualToOne) / (rowMap->getLocalNumElements() - numDiagsEqualToOne);
+      if (doReciprocal) {
+        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+          if (replaceSingleEntryRowWithZero && nnzPerRow[i] <= static_cast<int>(1))
+            diagVals[i] = zero;
+          else if ((diagVals[i] != zero) && (TST::magnitude(diagVals[i]) < TST::magnitude(two * regSum[i])))
+            diagVals[i] = one / TST::magnitude((two * regSum[i]));
+          else {
+            if (TST::magnitude(diagVals[i]) > tol)
+              diagVals[i] = one / diagVals[i];
+            else {
+              diagVals[i] = valReplacement;
+            }
           }
         }
       }
-
     }
-
-    return diag;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A) {
-    size_t numRows = A.getRowMap()->getLocalNumElements();
-    Magnitude ZERO = Teuchos::ScalarTraits<Magnitude>::zero();
-    Teuchos::ArrayRCP<Magnitude> maxvec(numRows);
-    Teuchos::ArrayView<const LocalOrdinal> cols;
-    Teuchos::ArrayView<const Scalar> vals;
-    for (size_t i = 0; i < numRows; ++i) {
-      A.getLocalRowView(i, cols, vals);
-      Magnitude mymax = ZERO;
-      for (LocalOrdinal j=0; j < cols.size(); ++j) {
-        if (Teuchos::as<size_t>(cols[j]) != i) {
-          mymax = std::max(mymax,-Teuchos::ScalarTraits<Scalar>::real(vals[j]));
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(doReciprocal, Xpetra::Exceptions::RuntimeError,
+                               "UtilitiesBase::GetLumpedMatrixDiagonal(): extracting reciprocal of diagonal of a blocked matrix is not supported");
+    diag = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(bA->getRangeMapExtractor()->getFullMap(), true);
+
+    for (size_t row = 0; row < bA->Rows(); ++row) {
+      for (size_t col = 0; col < bA->Cols(); ++col) {
+        if (!bA->getMatrix(row, col).is_null()) {
+          // if we are in Thyra mode, but the block (row,row) is again a blocked operator, we have to use (pseudo) Xpetra-style GIDs with offset!
+          bool bThyraMode      = bA->getRangeMapExtractor()->getThyraMode() && (Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(bA->getMatrix(row, col)) == Teuchos::null);
+          RCP<Vector> ddtemp   = bA->getRangeMapExtractor()->ExtractVector(diag, row, bThyraMode);
+          RCP<const Vector> dd = GetLumpedMatrixDiagonal(*(bA->getMatrix(row, col)));
+          ddtemp->update(Teuchos::as<Scalar>(1.0), *dd, Teuchos::as<Scalar>(1.0));
+          bA->getRangeMapExtractor()->InsertVector(ddtemp, row, diag, bThyraMode);
         }
       }
-      maxvec[i] = mymax;
     }
-    return maxvec;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> &BlockNumber) {
-    TEUCHOS_TEST_FOR_EXCEPTION(!A.getColMap()->isSameAs(*BlockNumber.getMap()),std::runtime_error,"GetMatrixMaxMinusOffDiagonal: BlockNumber must match's A's column map.");
-
-    Teuchos::ArrayRCP<const LocalOrdinal> block_id = BlockNumber.getData(0);
+  return diag;
+}
 
-    size_t numRows = A.getRowMap()->getLocalNumElements();
-    Magnitude ZERO = Teuchos::ScalarTraits<Magnitude>::zero();
-    Teuchos::ArrayRCP<Magnitude> maxvec(numRows);
-    Teuchos::ArrayView<const LocalOrdinal> cols;
-    Teuchos::ArrayView<const Scalar> vals;
-    for (size_t i = 0; i < numRows; ++i) {
-      A.getLocalRowView(i, cols, vals);
-      Magnitude mymax = ZERO;
-      for (LocalOrdinal j=0; j < cols.size(); ++j) {
-        if (Teuchos::as<size_t>(cols[j]) != i && block_id[i] == block_id[cols[j]]) {
-          mymax = std::max(mymax,-Teuchos::ScalarTraits<Scalar>::real(vals[j]));
-        }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A) {
+  size_t numRows = A.getRowMap()->getLocalNumElements();
+  Magnitude ZERO = Teuchos::ScalarTraits<Magnitude>::zero();
+  Teuchos::ArrayRCP<Magnitude> maxvec(numRows);
+  Teuchos::ArrayView<const LocalOrdinal> cols;
+  Teuchos::ArrayView<const Scalar> vals;
+  for (size_t i = 0; i < numRows; ++i) {
+    A.getLocalRowView(i, cols, vals);
+    Magnitude mymax = ZERO;
+    for (LocalOrdinal j = 0; j < cols.size(); ++j) {
+      if (Teuchos::as<size_t>(cols[j]) != i) {
+        mymax = std::max(mymax, -Teuchos::ScalarTraits<Scalar>::real(vals[j]));
       }
-      //        printf("A(%d,:) row_scale(block) = %6.4e\n",(int)i,mymax);
-
-      maxvec[i] = mymax;
     }
-    return maxvec;
+    maxvec[i] = mymax;
   }
+  return maxvec;
+}
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetInverse(Teuchos::RCP<const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > v, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol, Scalar valReplacement) {
-
-    RCP<Vector> ret = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(v->getMap(),true);
-
-    // check whether input vector "v" is a BlockedVector
-    RCP<const BlockedVector> bv = Teuchos::rcp_dynamic_cast<const BlockedVector>(v);
-    if(bv.is_null() == false) {
-      RCP<BlockedVector> bret = Teuchos::rcp_dynamic_cast<BlockedVector>(ret);
-      TEUCHOS_TEST_FOR_EXCEPTION(bret.is_null() == true, MueLu::Exceptions::RuntimeError,"MueLu::UtilitiesBase::GetInverse: return vector should be of type BlockedVector");
-      RCP<const BlockedMap> bmap = bv->getBlockedMap();
-      for(size_t r = 0; r < bmap->getNumMaps(); ++r) {
-        RCP<const MultiVector> submvec = bv->getMultiVector(r,bmap->getThyraMode());
-        RCP<const Vector> subvec = submvec->getVector(0);
-        RCP<Vector> subvecinf = MueLu::UtilitiesBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetInverse(subvec,tol,valReplacement);
-        bret->setMultiVector(r, subvecinf, bmap->getThyraMode());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixMaxMinusOffDiagonal(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>& BlockNumber) {
+  TEUCHOS_TEST_FOR_EXCEPTION(!A.getColMap()->isSameAs(*BlockNumber.getMap()), std::runtime_error, "GetMatrixMaxMinusOffDiagonal: BlockNumber must match's A's column map.");
+
+  Teuchos::ArrayRCP<const LocalOrdinal> block_id = BlockNumber.getData(0);
+
+  size_t numRows = A.getRowMap()->getLocalNumElements();
+  Magnitude ZERO = Teuchos::ScalarTraits<Magnitude>::zero();
+  Teuchos::ArrayRCP<Magnitude> maxvec(numRows);
+  Teuchos::ArrayView<const LocalOrdinal> cols;
+  Teuchos::ArrayView<const Scalar> vals;
+  for (size_t i = 0; i < numRows; ++i) {
+    A.getLocalRowView(i, cols, vals);
+    Magnitude mymax = ZERO;
+    for (LocalOrdinal j = 0; j < cols.size(); ++j) {
+      if (Teuchos::as<size_t>(cols[j]) != i && block_id[i] == block_id[cols[j]]) {
+        mymax = std::max(mymax, -Teuchos::ScalarTraits<Scalar>::real(vals[j]));
       }
-      return ret;
     }
+    //        printf("A(%d,:) row_scale(block) = %6.4e\n",(int)i,mymax);
 
-    // v is an {Epetra,Tpetra}Vector: work with the underlying raw data
-    ArrayRCP<Scalar> retVals = ret->getDataNonConst(0);
-    ArrayRCP<const Scalar> inputVals = v->getData(0);
-    for (size_t i = 0; i < v->getMap()->getLocalNumElements(); ++i) {
-      if(Teuchos::ScalarTraits<Scalar>::magnitude(inputVals[i]) > tol)
-        retVals[i] = Teuchos::ScalarTraits<Scalar>::one() / inputVals[i];
-      else
-        retVals[i] = valReplacement;
-    }
-    return ret;
+    maxvec[i] = mymax;
   }
+  return maxvec;
+}
 
-
-  // template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  // RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  // UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  // GetMatrixOverlappedDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A) {
-  //   RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
-
-  //   // Undo block map (if we have one)
-  //   RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
-  //   if(!browMap.is_null()) rowMap = browMap->getMap();
-
-  //   RCP<Vector> localDiag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap);
-  //   try {
-  //     const CrsMatrixWrap* crsOp = dynamic_cast<const CrsMatrixWrap*>(&A);
-  //     if (crsOp == NULL) {
-  //       throw Exceptions::RuntimeError("cast to CrsMatrixWrap failed");
-  //     }
-  //     Teuchos::ArrayRCP<size_t> offsets;
-  //     crsOp->getLocalDiagOffsets(offsets);
-  //     crsOp->getLocalDiagCopy(*localDiag,offsets());
-  //   }
-  //   catch (...) {
-  //     ArrayRCP<Scalar>   localDiagVals = localDiag->getDataNonConst(0);
-  //     Teuchos::ArrayRCP<Scalar> diagVals = GetMatrixDiagonal(A);
-  //     for (LocalOrdinal i = 0; i < localDiagVals.size(); i++)
-  //       localDiagVals[i] = diagVals[i];
-  //     localDiagVals = diagVals = null;
-  //   }
-
-  //   RCP<Vector> diagonal = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap);
-  //   RCP< const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer;
-  //   importer = A.getCrsGraph()->getImporter();
-  //   if (importer == Teuchos::null) {
-  //     importer = Xpetra::ImportFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap, colMap);
-  //   }
-  //   diagonal->doImport(*localDiag, *(importer), Xpetra::INSERT);
-  //   return diagonal;
-  // }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixOverlappedDiagonal(const Matrix& A) {
-    // FIXME_KOKKOS
-    RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
-    RCP<Vector>    localDiag     = VectorFactory::Build(rowMap);
-
-    const CrsMatrixWrap* crsOp = dynamic_cast<const CrsMatrixWrap*>(&A);
-    if ((crsOp != NULL) && (rowMap->lib() == Xpetra::UseTpetra)) {
-       Teuchos::ArrayRCP<size_t> offsets;
-       crsOp->getLocalDiagOffsets(offsets);
-       crsOp->getLocalDiagCopy(*localDiag,offsets());
-    }
-    else {
-      auto localDiagVals = localDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
-      const auto diagVals = GetMatrixDiagonal(A)->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      Kokkos::deep_copy(localDiagVals, diagVals);
-    }
-
-    RCP<Vector> diagonal = VectorFactory::Build(colMap);
-    RCP< const Import> importer;
-    importer = A.getCrsGraph()->getImporter();
-    if (importer == Teuchos::null) {
-      importer = ImportFactory::Build(rowMap, colMap);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetInverse(Teuchos::RCP<const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> v, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tol, Scalar valReplacement) {
+  RCP<Vector> ret = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(v->getMap(), true);
+
+  // check whether input vector "v" is a BlockedVector
+  RCP<const BlockedVector> bv = Teuchos::rcp_dynamic_cast<const BlockedVector>(v);
+  if (bv.is_null() == false) {
+    RCP<BlockedVector> bret = Teuchos::rcp_dynamic_cast<BlockedVector>(ret);
+    TEUCHOS_TEST_FOR_EXCEPTION(bret.is_null() == true, MueLu::Exceptions::RuntimeError, "MueLu::UtilitiesBase::GetInverse: return vector should be of type BlockedVector");
+    RCP<const BlockedMap> bmap = bv->getBlockedMap();
+    for (size_t r = 0; r < bmap->getNumMaps(); ++r) {
+      RCP<const MultiVector> submvec = bv->getMultiVector(r, bmap->getThyraMode());
+      RCP<const Vector> subvec       = submvec->getVector(0);
+      RCP<Vector> subvecinf          = MueLu::UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetInverse(subvec, tol, valReplacement);
+      bret->setMultiVector(r, subvecinf, bmap->getThyraMode());
     }
-    diagonal->doImport(*localDiag, *(importer), Xpetra::INSERT);
-
-    return diagonal;
+    return ret;
   }
 
+  // v is an {Epetra,Tpetra}Vector: work with the underlying raw data
+  ArrayRCP<Scalar> retVals         = ret->getDataNonConst(0);
+  ArrayRCP<const Scalar> inputVals = v->getData(0);
+  for (size_t i = 0; i < v->getMap()->getLocalNumElements(); ++i) {
+    if (Teuchos::ScalarTraits<Scalar>::magnitude(inputVals[i]) > tol)
+      retVals[i] = Teuchos::ScalarTraits<Scalar>::one() / inputVals[i];
+    else
+      retVals[i] = valReplacement;
+  }
+  return ret;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixOverlappedDeletedRowsum(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A) {
-    using STS = typename Teuchos::ScalarTraits<SC>;
-
-    // Undo block map (if we have one)
-    RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
-    RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
-    if(!browMap.is_null()) rowMap = browMap->getMap();
-
-    RCP<Vector> local   = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(rowMap);
-    RCP<Vector> ghosted = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(colMap,true);
-    ArrayRCP<SC> localVals = local->getDataNonConst(0);
-
-    for (LO row = 0; row < static_cast<LO>(A.getRowMap()->getLocalNumElements()); ++row) {
-      size_t nnz = A.getNumEntriesInLocalRow(row);
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      A.getLocalRowView(row, indices, vals);
-
-      SC si = STS::zero();
-
-      for (LO colID = 0; colID < static_cast<LO>(nnz); colID++) {
-        if(indices[colID] != row) {
-          si += vals[colID];
-        }
-      }
-      localVals[row] = si;
-    }
-
-    RCP< const Xpetra::Import<LO,GO,Node> > importer;
-    importer = A.getCrsGraph()->getImporter();
-    if (importer == Teuchos::null) {
-      importer = Xpetra::ImportFactory<LO,GO,Node>::Build(rowMap, colMap);
-    }
-    ghosted->doImport(*local, *(importer), Xpetra::INSERT);
-    return ghosted;
+// template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+// RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
+// UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+// GetMatrixOverlappedDiagonal(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A) {
+//   RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
+
+//   // Undo block map (if we have one)
+//   RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
+//   if(!browMap.is_null()) rowMap = browMap->getMap();
+
+//   RCP<Vector> localDiag = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap);
+//   try {
+//     const CrsMatrixWrap* crsOp = dynamic_cast<const CrsMatrixWrap*>(&A);
+//     if (crsOp == NULL) {
+//       throw Exceptions::RuntimeError("cast to CrsMatrixWrap failed");
+//     }
+//     Teuchos::ArrayRCP<size_t> offsets;
+//     crsOp->getLocalDiagOffsets(offsets);
+//     crsOp->getLocalDiagCopy(*localDiag,offsets());
+//   }
+//   catch (...) {
+//     ArrayRCP<Scalar>   localDiagVals = localDiag->getDataNonConst(0);
+//     Teuchos::ArrayRCP<Scalar> diagVals = GetMatrixDiagonal(A);
+//     for (LocalOrdinal i = 0; i < localDiagVals.size(); i++)
+//       localDiagVals[i] = diagVals[i];
+//     localDiagVals = diagVals = null;
+//   }
+
+//   RCP<Vector> diagonal = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap);
+//   RCP< const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer;
+//   importer = A.getCrsGraph()->getImporter();
+//   if (importer == Teuchos::null) {
+//     importer = Xpetra::ImportFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(rowMap, colMap);
+//   }
+//   diagonal->doImport(*localDiag, *(importer), Xpetra::INSERT);
+//   return diagonal;
+// }
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixOverlappedDiagonal(const Matrix& A) {
+  // FIXME_KOKKOS
+  RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
+  RCP<Vector> localDiag = VectorFactory::Build(rowMap);
+
+  const CrsMatrixWrap* crsOp = dynamic_cast<const CrsMatrixWrap*>(&A);
+  if ((crsOp != NULL) && (rowMap->lib() == Xpetra::UseTpetra)) {
+    Teuchos::ArrayRCP<size_t> offsets;
+    crsOp->getLocalDiagOffsets(offsets);
+    crsOp->getLocalDiagCopy(*localDiag, offsets());
+  } else {
+    auto localDiagVals  = localDiag->getDeviceLocalView(Xpetra::Access::ReadWrite);
+    const auto diagVals = GetMatrixDiagonal(A)->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    Kokkos::deep_copy(localDiagVals, diagVals);
   }
 
+  RCP<Vector> diagonal = VectorFactory::Build(colMap);
+  RCP<const Import> importer;
+  importer = A.getCrsGraph()->getImporter();
+  if (importer == Teuchos::null) {
+    importer = ImportFactory::Build(rowMap, colMap);
+  }
+  diagonal->doImport(*localDiag, *(importer), Xpetra::INSERT);
 
+  return diagonal;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GetMatrixOverlappedAbsDeletedRowsum(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & A) {
-    RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
-    using STS = typename Teuchos::ScalarTraits<Scalar>;
-    using MTS = typename Teuchos::ScalarTraits<Magnitude>;
-    using MT  = Magnitude;
-    using RealValuedVector = Xpetra::Vector<MT,LO,GO,Node>;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixOverlappedDeletedRowsum(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A) {
+  using STS = typename Teuchos::ScalarTraits<SC>;
 
-    // Undo block map (if we have one)
-    RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
-    if(!browMap.is_null()) rowMap = browMap->getMap();
+  // Undo block map (if we have one)
+  RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
+  RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
+  if (!browMap.is_null()) rowMap = browMap->getMap();
 
-    RCP<RealValuedVector> local     = Xpetra::VectorFactory<MT,LO,GO,Node>::Build(rowMap);
-    RCP<RealValuedVector> ghosted   = Xpetra::VectorFactory<MT,LO,GO,Node>::Build(colMap,true);
-    ArrayRCP<MT>          localVals = local->getDataNonConst(0);
+  RCP<Vector> local      = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(rowMap);
+  RCP<Vector> ghosted    = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(colMap, true);
+  ArrayRCP<SC> localVals = local->getDataNonConst(0);
 
-    for (LO rowIdx = 0; rowIdx < static_cast<LO>(A.getRowMap()->getLocalNumElements()); ++rowIdx) {
-      size_t nnz = A.getNumEntriesInLocalRow(rowIdx);
-      ArrayView<const LO> indices;
-      ArrayView<const SC> vals;
-      A.getLocalRowView(rowIdx, indices, vals);
+  for (LO row = 0; row < static_cast<LO>(A.getRowMap()->getLocalNumElements()); ++row) {
+    size_t nnz = A.getNumEntriesInLocalRow(row);
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    A.getLocalRowView(row, indices, vals);
 
-      MT si = MTS::zero();
+    SC si = STS::zero();
 
-      for (LO colID = 0; colID < static_cast<LO>(nnz); ++colID) {
-        if(indices[colID] != rowIdx) {
-          si += STS::magnitude(vals[colID]);
-        }
+    for (LO colID = 0; colID < static_cast<LO>(nnz); colID++) {
+      if (indices[colID] != row) {
+        si += vals[colID];
       }
-      localVals[rowIdx] = si;
     }
-
-    RCP< const Xpetra::Import<LO,GO,Node> > importer;
-    importer = A.getCrsGraph()->getImporter();
-    if (importer == Teuchos::null) {
-      importer = Xpetra::ImportFactory<LO,GO,Node>::Build(rowMap, colMap);
-    }
-    ghosted->doImport(*local, *(importer), Xpetra::INSERT);
-    return ghosted;
+    localVals[row] = si;
   }
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ResidualNorm(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS) {
-    TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
-      const size_t numVecs = X.getNumVectors();
-    RCP<MultiVector> RES = Residual(Op, X, RHS);
-    Teuchos::Array<Magnitude> norms(numVecs);
-    RES->norm2(norms);
-    return norms;
+  RCP<const Xpetra::Import<LO, GO, Node>> importer;
+  importer = A.getCrsGraph()->getImporter();
+  if (importer == Teuchos::null) {
+    importer = Xpetra::ImportFactory<LO, GO, Node>::Build(rowMap, colMap);
   }
+  ghosted->doImport(*local, *(importer), Xpetra::INSERT);
+  return ghosted;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ResidualNorm(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector & Resid) {
-    TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
-      const size_t numVecs = X.getNumVectors();
-    Residual(Op,X,RHS,Resid);
-    Teuchos::Array<Magnitude> norms(numVecs);
-    Resid.norm2(norms);
-    return norms;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GetMatrixOverlappedAbsDeletedRowsum(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A) {
+  RCP<const Map> rowMap = A.getRowMap(), colMap = A.getColMap();
+  using STS              = typename Teuchos::ScalarTraits<Scalar>;
+  using MTS              = typename Teuchos::ScalarTraits<Magnitude>;
+  using MT               = Magnitude;
+  using RealValuedVector = Xpetra::Vector<MT, LO, GO, Node>;
+
+  // Undo block map (if we have one)
+  RCP<const BlockedMap> browMap = Teuchos::rcp_dynamic_cast<const BlockedMap>(rowMap);
+  if (!browMap.is_null()) rowMap = browMap->getMap();
+
+  RCP<RealValuedVector> local   = Xpetra::VectorFactory<MT, LO, GO, Node>::Build(rowMap);
+  RCP<RealValuedVector> ghosted = Xpetra::VectorFactory<MT, LO, GO, Node>::Build(colMap, true);
+  ArrayRCP<MT> localVals        = local->getDataNonConst(0);
+
+  for (LO rowIdx = 0; rowIdx < static_cast<LO>(A.getRowMap()->getLocalNumElements()); ++rowIdx) {
+    size_t nnz = A.getNumEntriesInLocalRow(rowIdx);
+    ArrayView<const LO> indices;
+    ArrayView<const SC> vals;
+    A.getLocalRowView(rowIdx, indices, vals);
+
+    MT si = MTS::zero();
+
+    for (LO colID = 0; colID < static_cast<LO>(nnz); ++colID) {
+      if (indices[colID] != rowIdx) {
+        si += STS::magnitude(vals[colID]);
+      }
+    }
+    localVals[rowIdx] = si;
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Residual(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS) {
-    TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
-      const size_t numVecs = X.getNumVectors();
-    // TODO Op.getRangeMap should return a BlockedMap if it is a BlockedCrsOperator
-    RCP<MultiVector> RES = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(RHS.getMap(), numVecs, false); // no need to initialize to zero
-    Op.residual(X,RHS,*RES);
-    return RES;
+  RCP<const Xpetra::Import<LO, GO, Node>> importer;
+  importer = A.getCrsGraph()->getImporter();
+  if (importer == Teuchos::null) {
+    importer = Xpetra::ImportFactory<LO, GO, Node>::Build(rowMap, colMap);
   }
+  ghosted->doImport(*local, *(importer), Xpetra::INSERT);
+  return ghosted;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ResidualNorm(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS) {
+  TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
+  const size_t numVecs = X.getNumVectors();
+  RCP<MultiVector> RES = Residual(Op, X, RHS);
+  Teuchos::Array<Magnitude> norms(numVecs);
+  RES->norm2(norms);
+  return norms;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Residual(const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector & Resid) {
-    TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides");
-    TEUCHOS_TEST_FOR_EXCEPTION(Resid.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of residual vectors != number of right-hand sides");
-    Op.residual(X,RHS,Resid);
-  }
-
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ResidualNorm(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector& Resid) {
+  TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
+  const size_t numVecs = X.getNumVectors();
+  Residual(Op, X, RHS, Resid);
+  Teuchos::Array<Magnitude> norms(numVecs);
+  Resid.norm2(norms);
+  return norms;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Scalar
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  PowerMethod(const Matrix& A, bool scaleByDiag,
-              LocalOrdinal niters, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tolerance, bool verbose, unsigned int seed) {
-    TEUCHOS_TEST_FOR_EXCEPTION(!(A.getRangeMap()->isSameAs(*(A.getDomainMap()))), Exceptions::Incompatible,
-                               "Utils::PowerMethod: operator must have domain and range maps that are equivalent.");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Residual(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS) {
+  TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides")
+  const size_t numVecs = X.getNumVectors();
+  // TODO Op.getRangeMap should return a BlockedMap if it is a BlockedCrsOperator
+  RCP<MultiVector> RES = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(RHS.getMap(), numVecs, false);  // no need to initialize to zero
+  Op.residual(X, RHS, *RES);
+  return RES;
+}
 
-    // power iteration
-    RCP<Vector> diagInvVec;
-    if (scaleByDiag) {
-      diagInvVec = GetMatrixDiagonalInverse(A);      
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Residual(const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const MultiVector& X, const MultiVector& RHS, MultiVector& Resid) {
+  TEUCHOS_TEST_FOR_EXCEPTION(X.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of solution vectors != number of right-hand sides");
+  TEUCHOS_TEST_FOR_EXCEPTION(Resid.getNumVectors() != RHS.getNumVectors(), Exceptions::RuntimeError, "Number of residual vectors != number of right-hand sides");
+  Op.residual(X, RHS, Resid);
+}
 
-    Scalar lambda = PowerMethod(A, diagInvVec, niters, tolerance, verbose, seed);
-    return lambda;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Scalar
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    PowerMethod(const Matrix& A, bool scaleByDiag,
+                LocalOrdinal niters, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tolerance, bool verbose, unsigned int seed) {
+  TEUCHOS_TEST_FOR_EXCEPTION(!(A.getRangeMap()->isSameAs(*(A.getDomainMap()))), Exceptions::Incompatible,
+                             "Utils::PowerMethod: operator must have domain and range maps that are equivalent.");
+
+  // power iteration
+  RCP<Vector> diagInvVec;
+  if (scaleByDiag) {
+    diagInvVec = GetMatrixDiagonalInverse(A);
   }
 
+  Scalar lambda = PowerMethod(A, diagInvVec, niters, tolerance, verbose, seed);
+  return lambda;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Scalar
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  PowerMethod(const Matrix& A, const RCP<Vector> &diagInvVec,
-              LocalOrdinal niters, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tolerance, bool verbose, unsigned int seed) {
-    TEUCHOS_TEST_FOR_EXCEPTION(!(A.getRangeMap()->isSameAs(*(A.getDomainMap()))), Exceptions::Incompatible,
-                               "Utils::PowerMethod: operator must have domain and range maps that are equivalent.");
-
-    // Create three vectors, fill z with random numbers
-    RCP<Vector> q = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A.getDomainMap());
-    RCP<Vector> r = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A.getRangeMap());
-    RCP<Vector> z = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A.getRangeMap());
-
-    z->setSeed(seed);  // seed random number generator
-    z->randomize(true);// use Xpetra implementation: -> same results for Epetra and Tpetra
-
-    Teuchos::Array<Magnitude> norms(1);
-
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-
-    const Scalar zero = STS::zero(), one = STS::one();
-
-    Scalar lambda = zero;
-    Magnitude residual = STS::magnitude(zero);
-
-    // power iteration
-    for (int iter = 0; iter < niters; ++iter) {
-      z->norm2(norms);                                  // Compute 2-norm of z
-      q->update(one/norms[0], *z, zero);                // Set q = z / normz
-      A.apply(*q, *z);                                  // Compute z = A*q
-      if (diagInvVec != Teuchos::null)
-        z->elementWiseMultiply(one, *diagInvVec, *z, zero);
-      lambda = q->dot(*z);                              // Approximate maximum eigenvalue: lamba = dot(q,z)
-
-      if (iter % 100 == 0 || iter + 1 == niters) {
-        r->update(1.0, *z, -lambda, *q, zero);          // Compute A*q - lambda*q
-        r->norm2(norms);
-        residual = STS::magnitude(norms[0] / lambda);
-        if (verbose) {
-          std::cout << "Iter = " << iter
-                    << "  Lambda = " << lambda
-                    << "  Residual of A*q - lambda*q = " << residual
-                    << std::endl;
-        }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Scalar
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    PowerMethod(const Matrix& A, const RCP<Vector>& diagInvVec,
+                LocalOrdinal niters, typename Teuchos::ScalarTraits<Scalar>::magnitudeType tolerance, bool verbose, unsigned int seed) {
+  TEUCHOS_TEST_FOR_EXCEPTION(!(A.getRangeMap()->isSameAs(*(A.getDomainMap()))), Exceptions::Incompatible,
+                             "Utils::PowerMethod: operator must have domain and range maps that are equivalent.");
+
+  // Create three vectors, fill z with random numbers
+  RCP<Vector> q = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A.getDomainMap());
+  RCP<Vector> r = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A.getRangeMap());
+  RCP<Vector> z = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A.getRangeMap());
+
+  z->setSeed(seed);    // seed random number generator
+  z->randomize(true);  // use Xpetra implementation: -> same results for Epetra and Tpetra
+
+  Teuchos::Array<Magnitude> norms(1);
+
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+
+  const Scalar zero = STS::zero(), one = STS::one();
+
+  Scalar lambda      = zero;
+  Magnitude residual = STS::magnitude(zero);
+
+  // power iteration
+  for (int iter = 0; iter < niters; ++iter) {
+    z->norm2(norms);                      // Compute 2-norm of z
+    q->update(one / norms[0], *z, zero);  // Set q = z / normz
+    A.apply(*q, *z);                      // Compute z = A*q
+    if (diagInvVec != Teuchos::null)
+      z->elementWiseMultiply(one, *diagInvVec, *z, zero);
+    lambda = q->dot(*z);  // Approximate maximum eigenvalue: lamba = dot(q,z)
+
+    if (iter % 100 == 0 || iter + 1 == niters) {
+      r->update(1.0, *z, -lambda, *q, zero);  // Compute A*q - lambda*q
+      r->norm2(norms);
+      residual = STS::magnitude(norms[0] / lambda);
+      if (verbose) {
+        std::cout << "Iter = " << iter
+                  << "  Lambda = " << lambda
+                  << "  Residual of A*q - lambda*q = " << residual
+                  << std::endl;
       }
-      if (residual < tolerance)
-        break;
     }
-    return lambda;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Teuchos::FancyOStream>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  MakeFancy(std::ostream& os) {
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(os));
-    return fancy;
+    if (residual < tolerance)
+      break;
   }
+  return lambda;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Teuchos::FancyOStream>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    MakeFancy(std::ostream& os) {
+  RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(os));
+  return fancy;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Distance2(const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>> &v, LocalOrdinal i0, LocalOrdinal i1) {
-    const size_t numVectors = v.size();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Distance2(const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>>& v, LocalOrdinal i0, LocalOrdinal i1) {
+  const size_t numVectors = v.size();
 
-    Scalar d = Teuchos::ScalarTraits<Scalar>::zero();
-    for (size_t j = 0; j < numVectors; j++) {
-      d += (v[j][i0] - v[j][i1])*(v[j][i0] - v[j][i1]);
-    }
-    return Teuchos::ScalarTraits<Scalar>::magnitude(d);
+  Scalar d = Teuchos::ScalarTraits<Scalar>::zero();
+  for (size_t j = 0; j < numVectors; j++) {
+    d += (v[j][i0] - v[j][i1]) * (v[j][i0] - v[j][i1]);
   }
+  return Teuchos::ScalarTraits<Scalar>::magnitude(d);
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Distance2(const Teuchos::ArrayView<double>& weight, const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>>& v, LocalOrdinal i0, LocalOrdinal i1) {
+  const size_t numVectors = v.size();
+  using MT                = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Distance2(const Teuchos::ArrayView<double> & weight,const Teuchos::Array<Teuchos::ArrayRCP<const Scalar>> &v, LocalOrdinal i0, LocalOrdinal i1) {
-    const size_t numVectors = v.size();
-    using MT = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-
-    Scalar d = Teuchos::ScalarTraits<Scalar>::zero();
-    for (size_t j = 0; j < numVectors; j++) {
-      d += Teuchos::as<MT>(weight[j])*(v[j][i0] - v[j][i1])*(v[j][i0] - v[j][i1]);
-    }
-    return Teuchos::ScalarTraits<Scalar>::magnitude(d);
+  Scalar d = Teuchos::ScalarTraits<Scalar>::zero();
+  for (size_t j = 0; j < numVectors; j++) {
+    d += Teuchos::as<MT>(weight[j]) * (v[j][i0] - v[j][i1]) * (v[j][i0] - v[j][i1]);
   }
+  return Teuchos::ScalarTraits<Scalar>::magnitude(d);
+}
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<const bool>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DetectDirichletRows(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType& tol, bool count_twos_as_dirichlet) {
-    LocalOrdinal numRows = A.getLocalNumRows();
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    ArrayRCP<bool> boundaryNodes(numRows, true);
-    if (count_twos_as_dirichlet) {
-      for (LocalOrdinal row = 0; row < numRows; row++) {
-        ArrayView<const LocalOrdinal> indices;
-        ArrayView<const Scalar> vals;
-        A.getLocalRowView(row, indices, vals);
-        size_t nnz = A.getNumEntriesInLocalRow(row);
-        if (nnz > 2) {
-          size_t col;
-          for (col = 0; col < nnz; col++)
-            if ( (indices[col] != row) && STS::magnitude(vals[col]) > tol) {
-              if (!boundaryNodes[row])
-                break;
-              boundaryNodes[row] = false;
-            }
-          if (col == nnz)
-            boundaryNodes[row] = true;
-        }
-      }
-    } else {
-      for (LocalOrdinal row = 0; row < numRows; row++) {
-        ArrayView<const LocalOrdinal> indices;
-        ArrayView<const Scalar> vals;
-        A.getLocalRowView(row, indices, vals);
-        size_t nnz = A.getNumEntriesInLocalRow(row);
-        if (nnz > 1)
-          for (size_t col = 0; col < nnz; col++)
-            if ( (indices[col] != row) && STS::magnitude(vals[col]) > tol) {
-              boundaryNodes[row] = false;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<const bool>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DetectDirichletRows(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType& tol, bool count_twos_as_dirichlet) {
+  LocalOrdinal numRows = A.getLocalNumRows();
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  ArrayRCP<bool> boundaryNodes(numRows, true);
+  if (count_twos_as_dirichlet) {
+    for (LocalOrdinal row = 0; row < numRows; row++) {
+      ArrayView<const LocalOrdinal> indices;
+      ArrayView<const Scalar> vals;
+      A.getLocalRowView(row, indices, vals);
+      size_t nnz = A.getNumEntriesInLocalRow(row);
+      if (nnz > 2) {
+        size_t col;
+        for (col = 0; col < nnz; col++)
+          if ((indices[col] != row) && STS::magnitude(vals[col]) > tol) {
+            if (!boundaryNodes[row])
               break;
-            }
+            boundaryNodes[row] = false;
+          }
+        if (col == nnz)
+          boundaryNodes[row] = true;
       }
     }
-    return boundaryNodes;
-  }
-
-  template <class SC, class LO, class GO, class NO>
-  Kokkos::View<bool*, typename NO::device_type>
-  UtilitiesBase<SC, LO, GO, NO>::
-  DetectDirichletRows_kokkos(const Xpetra::Matrix<SC,LO,GO,NO>& A,
-                             const typename Teuchos::ScalarTraits<SC>::magnitudeType& tol,
-                             const bool count_twos_as_dirichlet) {
-    using impl_scalar_type = typename Kokkos::ArithTraits<SC>::val_type;
-    using ATS        = Kokkos::ArithTraits<impl_scalar_type>;
-    using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-    using helpers    = Xpetra::Helpers<SC,LO,GO,NO>;
-
-
-    if(helpers::isTpetraBlockCrs(A)) {
-      const Tpetra::BlockCrsMatrix<SC,LO,GO,NO> & Am = helpers::Op2TpetraBlockCrs(A);
-      auto b_graph      = Am.getCrsGraph().getLocalGraphDevice();
-      auto b_rowptr     = Am.getCrsGraph().getLocalRowPtrsDevice();
-      auto values       = Am.getValuesDevice();
-      LO   numBlockRows = Am.getLocalNumRows();
-      const LO stride   = Am.getBlockSize() * Am.getBlockSize();
-
-      Kokkos::View<bool*, typename NO::device_type> boundaryNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numBlockRows);
-
-      if (count_twos_as_dirichlet)
-        throw Exceptions::RuntimeError("BlockCrs does not support counting twos as Dirichlet");
-
-      Kokkos::parallel_for("MueLu:Utils::DetectDirichletRowsBlockCrs", range_type(0,numBlockRows),
-                           KOKKOS_LAMBDA(const LO row) {
-                             auto rowView = b_graph.rowConst(row);
-                             auto length  = rowView.length;
-                             LO valstart  = b_rowptr[row] * stride;
-
-                             boundaryNodes(row) = true;
-                             decltype(length) colID =0;
-                             for (; colID < length; colID++) {
-                               if (rowView.colidx(colID) != row) {
-                                 LO current = valstart + colID*stride;
-                                 for(LO k=0; k<stride; k++) {
-                                   if (ATS::magnitude(values[current+ k]) > tol) {
-                                     boundaryNodes(row) = false;
-                                     break;
-                                   }
-                                 }
-                               }
-                               if(boundaryNodes(row) == false)
-                                 break;
-                             }
-                           });
-
-      return boundaryNodes;
-    }
-    else {
-      auto localMatrix = A.getLocalMatrixDevice();
-      LO   numRows     = A.getLocalNumRows();
-      Kokkos::View<bool*, typename NO::device_type> boundaryNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numRows);
-
-      if (count_twos_as_dirichlet)
-        Kokkos::parallel_for("MueLu:Utils::DetectDirichletRows_Twos_As_Dirichlet", range_type(0,numRows),
-                             KOKKOS_LAMBDA(const LO row) {
-                               auto rowView = localMatrix.row(row);
-                               auto length  = rowView.length;
-
-                               boundaryNodes(row) = true;
-                               if (length > 2) {
-                                 decltype(length) colID =0;
-                                 for ( ; colID < length; colID++)
-                                   if ((rowView.colidx(colID) != row) &&
-                                       (ATS::magnitude(rowView.value(colID)) > tol)) {
-                                     if (!boundaryNodes(row))
-                                       break;
-                                     boundaryNodes(row) = false;
-                                   }
-                                 if (colID == length)
-                                   boundaryNodes(row) = true;
-                               }
-                             });
-      else
-        Kokkos::parallel_for("MueLu:Utils::DetectDirichletRows", range_type(0,numRows),
-                             KOKKOS_LAMBDA(const LO row) {
-                               auto rowView = localMatrix.row(row);
-                               auto length  = rowView.length;
-
-                               boundaryNodes(row) = true;
-                               for (decltype(length) colID = 0; colID < length; colID++)
-                                 if ((rowView.colidx(colID) != row) &&
-                                     (ATS::magnitude(rowView.value(colID)) > tol)) {
-                                   boundaryNodes(row) = false;
-                                   break;
-                                 }
-                             });
-      return boundaryNodes;
-    }
-
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<const bool>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DetectDirichletRowsExt(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, bool & bHasZeroDiagonal, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType& tol) {
-
-    // assume that there is no zero diagonal in matrix
-    bHasZeroDiagonal = false;
-
-    Teuchos::RCP<Vector> diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(A.getRowMap());
-    A.getLocalDiagCopy(*diagVec);
-    Teuchos::ArrayRCP< const Scalar > diagVecData = diagVec->getData(0);
-
-    LocalOrdinal numRows = A.getLocalNumRows();
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    ArrayRCP<bool> boundaryNodes(numRows, false);
+  } else {
     for (LocalOrdinal row = 0; row < numRows; row++) {
       ArrayView<const LocalOrdinal> indices;
       ArrayView<const Scalar> vals;
       A.getLocalRowView(row, indices, vals);
-      size_t nnz = 0; // collect nonzeros in row (excluding the diagonal)
-      bool bHasDiag = false;
-      for (decltype(indices.size()) col = 0; col < indices.size(); col++) {
-        if ( indices[col] != row) {
-          if (STS::magnitude(vals[col] / STS::magnitude(sqrt(STS::magnitude(diagVecData[row]) * STS::magnitude(diagVecData[col])))   ) > tol) {
-            nnz++;
+      size_t nnz = A.getNumEntriesInLocalRow(row);
+      if (nnz > 1)
+        for (size_t col = 0; col < nnz; col++)
+          if ((indices[col] != row) && STS::magnitude(vals[col]) > tol) {
+            boundaryNodes[row] = false;
+            break;
           }
-        } else bHasDiag = true; // found a diagonal entry
-      }
-      if (bHasDiag == false) bHasZeroDiagonal = true; // we found at least one row without a diagonal
-      else if(nnz == 0) boundaryNodes[row] = true;
     }
-    return boundaryNodes;
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindNonZeros(const Teuchos::ArrayRCP<const Scalar> vals,
-               Teuchos::ArrayRCP<bool> nonzeros) {
-    TEUCHOS_ASSERT(vals.size() == nonzeros.size());
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
-    const magnitudeType eps = 2.0*Teuchos::ScalarTraits<magnitudeType>::eps();
-    for(size_t i=0; i<static_cast<size_t>(vals.size()); i++) {
-      nonzeros[i] = (Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > eps);
-    }
-  }
-
-  // Find Nonzeros in a device view
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindNonZeros(const typename Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>::dual_view_type::t_dev_const_um vals,
-               Kokkos::View<bool*, typename Node::device_type> nonzeros) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_ATS = Kokkos::ArithTraits<typename ATS::val_type>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-    TEUCHOS_ASSERT(vals.extent(0) == nonzeros.extent(0));
-    const typename ATS::magnitudeType eps = 2.0*impl_ATS::eps();
-
-    Kokkos::parallel_for("MueLu:Maxwell1::FindNonZeros", range_type(0,vals.extent(0)),
-                         KOKKOS_LAMBDA (const size_t i) {
-                           nonzeros(i) = (impl_ATS::magnitude(vals(i,0)) > eps);
-                         });
   }
+  return boundaryNodes;
+}
 
+template <class SC, class LO, class GO, class NO>
+Kokkos::View<bool*, typename NO::device_type>
+UtilitiesBase<SC, LO, GO, NO>::
+    DetectDirichletRows_kokkos(const Xpetra::Matrix<SC, LO, GO, NO>& A,
+                               const typename Teuchos::ScalarTraits<SC>::magnitudeType& tol,
+                               const bool count_twos_as_dirichlet) {
+  using impl_scalar_type = typename Kokkos::ArithTraits<SC>::val_type;
+  using ATS              = Kokkos::ArithTraits<impl_scalar_type>;
+  using range_type       = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+  using helpers          = Xpetra::Helpers<SC, LO, GO, NO>;
+
+  if (helpers::isTpetraBlockCrs(A)) {
+    const Tpetra::BlockCrsMatrix<SC, LO, GO, NO>& Am = helpers::Op2TpetraBlockCrs(A);
+    auto b_graph                                     = Am.getCrsGraph().getLocalGraphDevice();
+    auto b_rowptr                                    = Am.getCrsGraph().getLocalRowPtrsDevice();
+    auto values                                      = Am.getValuesDevice();
+    LO numBlockRows                                  = Am.getLocalNumRows();
+    const LO stride                                  = Am.getBlockSize() * Am.getBlockSize();
+
+    Kokkos::View<bool*, typename NO::device_type> boundaryNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numBlockRows);
+
+    if (count_twos_as_dirichlet)
+      throw Exceptions::RuntimeError("BlockCrs does not support counting twos as Dirichlet");
+
+    Kokkos::parallel_for(
+        "MueLu:Utils::DetectDirichletRowsBlockCrs", range_type(0, numBlockRows),
+        KOKKOS_LAMBDA(const LO row) {
+          auto rowView = b_graph.rowConst(row);
+          auto length  = rowView.length;
+          LO valstart  = b_rowptr[row] * stride;
+
+          boundaryNodes(row)     = true;
+          decltype(length) colID = 0;
+          for (; colID < length; colID++) {
+            if (rowView.colidx(colID) != row) {
+              LO current = valstart + colID * stride;
+              for (LO k = 0; k < stride; k++) {
+                if (ATS::magnitude(values[current + k]) > tol) {
+                  boundaryNodes(row) = false;
+                  break;
+                }
+              }
+            }
+            if (boundaryNodes(row) == false)
+              break;
+          }
+        });
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                                const Teuchos::ArrayRCP<bool>& dirichletRows,
-                                Teuchos::ArrayRCP<bool> dirichletCols,
-                                Teuchos::ArrayRCP<bool> dirichletDomain) {
-    const Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domMap = A .getDomainMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rowMap = A.getRowMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > colMap = A.getColMap();
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == rowMap->getLocalNumElements());
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletCols.size()) == colMap->getLocalNumElements());
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletDomain.size()) == domMap->getLocalNumElements());
-    RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > myColsToZero = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap, 1, /*zeroOut=*/true);
-    // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
-    for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-      if (dirichletRows[i]) {
-        ArrayView<const LocalOrdinal> indices;
-        ArrayView<const Scalar> values;
-        A.getLocalRowView(i,indices,values);
-        for(size_t j=0; j<static_cast<size_t>(indices.size()); j++)
-          myColsToZero->replaceLocalValue(indices[j],0,one);
-      }
-    }
-
-    RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > globalColsToZero;
-    RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer = A.getCrsGraph()->getImporter();
-    if (!importer.is_null()) {
-      globalColsToZero = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(domMap, 1, /*zeroOut=*/true);
-      // export to domain map
-      globalColsToZero->doExport(*myColsToZero,*importer,Xpetra::ADD);
-      // import to column map
-      myColsToZero->doImport(*globalColsToZero,*importer,Xpetra::INSERT);
-    }
+    return boundaryNodes;
+  } else {
+    auto localMatrix = A.getLocalMatrixDevice();
+    LO numRows       = A.getLocalNumRows();
+    Kokkos::View<bool*, typename NO::device_type> boundaryNodes(Kokkos::ViewAllocateWithoutInitializing("boundaryNodes"), numRows);
+
+    if (count_twos_as_dirichlet)
+      Kokkos::parallel_for(
+          "MueLu:Utils::DetectDirichletRows_Twos_As_Dirichlet", range_type(0, numRows),
+          KOKKOS_LAMBDA(const LO row) {
+            auto rowView = localMatrix.row(row);
+            auto length  = rowView.length;
+
+            boundaryNodes(row) = true;
+            if (length > 2) {
+              decltype(length) colID = 0;
+              for (; colID < length; colID++)
+                if ((rowView.colidx(colID) != row) &&
+                    (ATS::magnitude(rowView.value(colID)) > tol)) {
+                  if (!boundaryNodes(row))
+                    break;
+                  boundaryNodes(row) = false;
+                }
+              if (colID == length)
+                boundaryNodes(row) = true;
+            }
+          });
     else
-      globalColsToZero = myColsToZero;
-
-    FindNonZeros(globalColsToZero->getData(0),dirichletDomain);
-    FindNonZeros(myColsToZero->getData(0),dirichletCols);
+      Kokkos::parallel_for(
+          "MueLu:Utils::DetectDirichletRows", range_type(0, numRows),
+          KOKKOS_LAMBDA(const LO row) {
+            auto rowView = localMatrix.row(row);
+            auto length  = rowView.length;
+
+            boundaryNodes(row) = true;
+            for (decltype(length) colID = 0; colID < length; colID++)
+              if ((rowView.colidx(colID) != row) &&
+                  (ATS::magnitude(rowView.value(colID)) > tol)) {
+                boundaryNodes(row) = false;
+                break;
+              }
+          });
+    return boundaryNodes;
   }
+}
 
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                                const Kokkos::View<bool*, typename Node::device_type> & dirichletRows,
-                                Kokkos::View<bool*, typename Node::device_type> dirichletCols,
-                                Kokkos::View<bool*, typename Node::device_type> dirichletDomain) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_ATS = Kokkos::ArithTraits<typename ATS::val_type>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domMap = A.getDomainMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rowMap = A.getRowMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > colMap = A.getColMap();
-    TEUCHOS_ASSERT(dirichletRows.extent(0) == rowMap->getLocalNumElements());
-    TEUCHOS_ASSERT(dirichletCols.extent(0) == colMap->getLocalNumElements());
-    TEUCHOS_ASSERT(dirichletDomain.extent(0) == domMap->getLocalNumElements());
-    RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > myColsToZero = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap, /*zeroOut=*/true);
-    // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
-    auto myColsToZeroView = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    auto localMatrix = A.getLocalMatrixDevice();
-    Kokkos::parallel_for("MueLu:Maxwell1::DetectDirichletCols", range_type(0,rowMap->getLocalNumElements()),
-                         KOKKOS_LAMBDA(const LocalOrdinal row) {
-                           if (dirichletRows(row)) {
-                             auto rowView = localMatrix.row(row);
-                             auto length  = rowView.length;
-
-                             for (decltype(length) colID = 0; colID < length; colID++)
-                               myColsToZeroView(rowView.colidx(colID),0) = impl_ATS::one();
-                           }
-                         });
-
-    RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > globalColsToZero;
-    RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer = A.getCrsGraph()->getImporter();
-    if (!importer.is_null()) {
-      globalColsToZero = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(domMap, /*zeroOut=*/true);
-      // export to domain map
-      globalColsToZero->doExport(*myColsToZero,*importer,Xpetra::ADD);
-      // import to column map
-      myColsToZero->doImport(*globalColsToZero,*importer,Xpetra::INSERT);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<const bool>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DetectDirichletRowsExt(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, bool& bHasZeroDiagonal, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType& tol) {
+  // assume that there is no zero diagonal in matrix
+  bHasZeroDiagonal = false;
+
+  Teuchos::RCP<Vector> diagVec = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(A.getRowMap());
+  A.getLocalDiagCopy(*diagVec);
+  Teuchos::ArrayRCP<const Scalar> diagVecData = diagVec->getData(0);
+
+  LocalOrdinal numRows = A.getLocalNumRows();
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  ArrayRCP<bool> boundaryNodes(numRows, false);
+  for (LocalOrdinal row = 0; row < numRows; row++) {
+    ArrayView<const LocalOrdinal> indices;
+    ArrayView<const Scalar> vals;
+    A.getLocalRowView(row, indices, vals);
+    size_t nnz    = 0;  // collect nonzeros in row (excluding the diagonal)
+    bool bHasDiag = false;
+    for (decltype(indices.size()) col = 0; col < indices.size(); col++) {
+      if (indices[col] != row) {
+        if (STS::magnitude(vals[col] / STS::magnitude(sqrt(STS::magnitude(diagVecData[row]) * STS::magnitude(diagVecData[col])))) > tol) {
+          nnz++;
+        }
+      } else
+        bHasDiag = true;  // found a diagonal entry
     }
-    else
-      globalColsToZero = myColsToZero;
-    UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindNonZeros(globalColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly),dirichletDomain);
-    UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindNonZeros(myColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly),dirichletCols);
+    if (bHasDiag == false)
+      bHasZeroDiagonal = true;  // we found at least one row without a diagonal
+    else if (nnz == 0)
+      boundaryNodes[row] = true;
   }
+  return boundaryNodes;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindNonZeros(const Teuchos::ArrayRCP<const Scalar> vals,
+                 Teuchos::ArrayRCP<bool> nonzeros) {
+  TEUCHOS_ASSERT(vals.size() == nonzeros.size());
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitudeType;
+  const magnitudeType eps = 2.0 * Teuchos::ScalarTraits<magnitudeType>::eps();
+  for (size_t i = 0; i < static_cast<size_t>(vals.size()); i++) {
+    nonzeros[i] = (Teuchos::ScalarTraits<Scalar>::magnitude(vals[i]) > eps);
+  }
+}
 
+// Find Nonzeros in a device view
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindNonZeros(const typename Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dual_view_type::t_dev_const_um vals,
+                 Kokkos::View<bool*, typename Node::device_type> nonzeros) {
+  using ATS        = Kokkos::ArithTraits<Scalar>;
+  using impl_ATS   = Kokkos::ArithTraits<typename ATS::val_type>;
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+  TEUCHOS_ASSERT(vals.extent(0) == nonzeros.extent(0));
+  const typename ATS::magnitudeType eps = 2.0 * impl_ATS::eps();
+
+  Kokkos::parallel_for(
+      "MueLu:Maxwell1::FindNonZeros", range_type(0, vals.extent(0)),
+      KOKKOS_LAMBDA(const size_t i) {
+        nonzeros(i) = (impl_ATS::magnitude(vals(i, 0)) > eps);
+      });
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyRowSumCriterion(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows) {
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Teuchos::ScalarTraits<MT> MTS;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>> rowmap = A.getRowMap();
-    for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
-      size_t nnz = A.getNumEntriesInLocalRow(row);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                                  const Teuchos::ArrayRCP<bool>& dirichletRows,
+                                  Teuchos::ArrayRCP<bool> dirichletCols,
+                                  Teuchos::ArrayRCP<bool> dirichletDomain) {
+  const Scalar one                                                 = Teuchos::ScalarTraits<Scalar>::one();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> domMap = A.getDomainMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap = A.getRowMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap = A.getColMap();
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == rowMap->getLocalNumElements());
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletCols.size()) == colMap->getLocalNumElements());
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletDomain.size()) == domMap->getLocalNumElements());
+  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> myColsToZero = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(colMap, 1, /*zeroOut=*/true);
+  // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
+  for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+    if (dirichletRows[i]) {
       ArrayView<const LocalOrdinal> indices;
-      ArrayView<const Scalar> vals;
-      A.getLocalRowView(row, indices, vals);
-
-      Scalar rowsum = STS::zero();
-      Scalar diagval = STS::zero();
-
-      for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
-        LocalOrdinal col = indices[colID];
-        if (row == col)
-          diagval = vals[colID];
-        rowsum += vals[colID];
-      }
-      //        printf("A(%d,:) row_sum(point) = %6.4e\n",row,rowsum);
-      if (rowSumTol < MTS::one() && STS::magnitude(rowsum) > STS::magnitude(diagval) * rowSumTol) {
-        //printf("Row %d triggers rowsum\n",(int)row);
-        dirichletRows[row] = true;
-      }
+      ArrayView<const Scalar> values;
+      A.getLocalRowView(i, indices, values);
+      for (size_t j = 0; j < static_cast<size_t>(indices.size()); j++)
+        myColsToZero->replaceLocalValue(indices[j], 0, one);
     }
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyRowSumCriterion(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> &BlockNumber, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows) {
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Teuchos::ScalarTraits<MT> MTS;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rowmap = A.getRowMap();
+  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> globalColsToZero;
+  RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> importer = A.getCrsGraph()->getImporter();
+  if (!importer.is_null()) {
+    globalColsToZero = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(domMap, 1, /*zeroOut=*/true);
+    // export to domain map
+    globalColsToZero->doExport(*myColsToZero, *importer, Xpetra::ADD);
+    // import to column map
+    myColsToZero->doImport(*globalColsToZero, *importer, Xpetra::INSERT);
+  } else
+    globalColsToZero = myColsToZero;
 
-    TEUCHOS_TEST_FOR_EXCEPTION(!A.getColMap()->isSameAs(*BlockNumber.getMap()),std::runtime_error,"ApplyRowSumCriterion: BlockNumber must match's A's column map.");
+  FindNonZeros(globalColsToZero->getData(0), dirichletDomain);
+  FindNonZeros(myColsToZero->getData(0), dirichletCols);
+}
 
-    Teuchos::ArrayRCP<const LocalOrdinal> block_id = BlockNumber.getData(0);
-    for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
-      size_t nnz = A.getNumEntriesInLocalRow(row);
-      ArrayView<const LocalOrdinal> indices;
-      ArrayView<const Scalar> vals;
-      A.getLocalRowView(row, indices, vals);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DetectDirichletColsAndDomains(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                                  const Kokkos::View<bool*, typename Node::device_type>& dirichletRows,
+                                  Kokkos::View<bool*, typename Node::device_type> dirichletCols,
+                                  Kokkos::View<bool*, typename Node::device_type> dirichletDomain) {
+  using ATS                                                        = Kokkos::ArithTraits<Scalar>;
+  using impl_ATS                                                   = Kokkos::ArithTraits<typename ATS::val_type>;
+  using range_type                                                 = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> domMap = A.getDomainMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap = A.getRowMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap = A.getColMap();
+  TEUCHOS_ASSERT(dirichletRows.extent(0) == rowMap->getLocalNumElements());
+  TEUCHOS_ASSERT(dirichletCols.extent(0) == colMap->getLocalNumElements());
+  TEUCHOS_ASSERT(dirichletDomain.extent(0) == domMap->getLocalNumElements());
+  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> myColsToZero = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(colMap, /*zeroOut=*/true);
+  // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
+  auto myColsToZeroView = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  auto localMatrix      = A.getLocalMatrixDevice();
+  Kokkos::parallel_for(
+      "MueLu:Maxwell1::DetectDirichletCols", range_type(0, rowMap->getLocalNumElements()),
+      KOKKOS_LAMBDA(const LocalOrdinal row) {
+        if (dirichletRows(row)) {
+          auto rowView = localMatrix.row(row);
+          auto length  = rowView.length;
+
+          for (decltype(length) colID = 0; colID < length; colID++)
+            myColsToZeroView(rowView.colidx(colID), 0) = impl_ATS::one();
+        }
+      });
 
-      Scalar rowsum = STS::zero();
-      Scalar diagval = STS::zero();
-      for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
-        LocalOrdinal col = indices[colID];
-        if (row == col)
-          diagval = vals[colID];
-        if(block_id[row] == block_id[col])
-          rowsum += vals[colID];
-      }
+  RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> globalColsToZero;
+  RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> importer = A.getCrsGraph()->getImporter();
+  if (!importer.is_null()) {
+    globalColsToZero = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(domMap, /*zeroOut=*/true);
+    // export to domain map
+    globalColsToZero->doExport(*myColsToZero, *importer, Xpetra::ADD);
+    // import to column map
+    myColsToZero->doImport(*globalColsToZero, *importer, Xpetra::INSERT);
+  } else
+    globalColsToZero = myColsToZero;
+  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindNonZeros(globalColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly), dirichletDomain);
+  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::FindNonZeros(myColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly), dirichletCols);
+}
 
-      //        printf("A(%d,:) row_sum(block) = %6.4e\n",row,rowsum);
-      if (rowSumTol < MTS::one() && STS::magnitude(rowsum) > STS::magnitude(diagval) * rowSumTol) {
-        //printf("Row %d triggers rowsum\n",(int)row);
-        dirichletRows[row] = true;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyRowSumCriterion(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows) {
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Teuchos::ScalarTraits<MT> MTS;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowmap = A.getRowMap();
+  for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
+    size_t nnz = A.getNumEntriesInLocalRow(row);
+    ArrayView<const LocalOrdinal> indices;
+    ArrayView<const Scalar> vals;
+    A.getLocalRowView(row, indices, vals);
+
+    Scalar rowsum  = STS::zero();
+    Scalar diagval = STS::zero();
+
+    for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
+      LocalOrdinal col = indices[colID];
+      if (row == col)
+        diagval = vals[colID];
+      rowsum += vals[colID];
+    }
+    //        printf("A(%d,:) row_sum(point) = %6.4e\n",row,rowsum);
+    if (rowSumTol < MTS::one() && STS::magnitude(rowsum) > STS::magnitude(diagval) * rowSumTol) {
+      // printf("Row %d triggers rowsum\n",(int)row);
+      dirichletRows[row] = true;
     }
   }
+}
 
-
-  // Applies rowsum criterion
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyRowSumCriterion(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                       const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol,
-                       Kokkos::View<bool*, typename Node::device_type> & dirichletRows)
-  {
-    typedef Teuchos::ScalarTraits<Scalar> STS;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>> rowmap = A.getRowMap();
-
-    auto dirichletRowsHost = Kokkos::create_mirror_view(dirichletRows);
-    Kokkos::deep_copy(dirichletRowsHost, dirichletRows);
-
-    for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
-      size_t nnz = A.getNumEntriesInLocalRow(row);
-      ArrayView<const LocalOrdinal> indices;
-      ArrayView<const Scalar> vals;
-      A.getLocalRowView(row, indices, vals);
-
-      Scalar rowsum = STS::zero();
-      Scalar diagval = STS::zero();
-      for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
-        LocalOrdinal col = indices[colID];
-        if (row == col)
-          diagval = vals[colID];
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyRowSumCriterion(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>& BlockNumber, const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol, Teuchos::ArrayRCP<bool>& dirichletRows) {
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Teuchos::ScalarTraits<MT> MTS;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowmap = A.getRowMap();
+
+  TEUCHOS_TEST_FOR_EXCEPTION(!A.getColMap()->isSameAs(*BlockNumber.getMap()), std::runtime_error, "ApplyRowSumCriterion: BlockNumber must match's A's column map.");
+
+  Teuchos::ArrayRCP<const LocalOrdinal> block_id = BlockNumber.getData(0);
+  for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
+    size_t nnz = A.getNumEntriesInLocalRow(row);
+    ArrayView<const LocalOrdinal> indices;
+    ArrayView<const Scalar> vals;
+    A.getLocalRowView(row, indices, vals);
+
+    Scalar rowsum  = STS::zero();
+    Scalar diagval = STS::zero();
+    for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
+      LocalOrdinal col = indices[colID];
+      if (row == col)
+        diagval = vals[colID];
+      if (block_id[row] == block_id[col])
         rowsum += vals[colID];
-      }
-      if (STS::real(rowsum) > STS::magnitude(diagval) * rowSumTol)
-        dirichletRowsHost(row) = true;
     }
 
-    Kokkos::deep_copy(dirichletRows, dirichletRowsHost);
-  }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::ArrayRCP<const bool>
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  DetectDirichletCols(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                      const Teuchos::ArrayRCP<const bool>& dirichletRows) {
-    Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domMap = A.getDomainMap();
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > colMap = A.getColMap();
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > myColsToZero = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(colMap,1);
-    myColsToZero->putScalar(zero);
-    // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
-    for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-      if (dirichletRows[i]) {
-        Teuchos::ArrayView<const LocalOrdinal> indices;
-        Teuchos::ArrayView<const Scalar> values;
-        A.getLocalRowView(i,indices,values);
-        for(size_t j=0; j<static_cast<size_t>(indices.size()); j++)
-          myColsToZero->replaceLocalValue(indices[j],0,one);
-      }
+    //        printf("A(%d,:) row_sum(block) = %6.4e\n",row,rowsum);
+    if (rowSumTol < MTS::one() && STS::magnitude(rowsum) > STS::magnitude(diagval) * rowSumTol) {
+      // printf("Row %d triggers rowsum\n",(int)row);
+      dirichletRows[row] = true;
     }
+  }
+}
 
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > globalColsToZero = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(domMap,1);
-    globalColsToZero->putScalar(zero);
-    Teuchos::RCP<Xpetra::Export<LocalOrdinal,GlobalOrdinal,Node> > exporter = Xpetra::ExportFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(colMap,domMap);
-    // export to domain map
-    globalColsToZero->doExport(*myColsToZero,*exporter,Xpetra::ADD);
-    // import to column map
-    myColsToZero->doImport(*globalColsToZero,*exporter,Xpetra::INSERT);
-    Teuchos::ArrayRCP<const Scalar> myCols = myColsToZero->getData(0);
-    Teuchos::ArrayRCP<bool> dirichletCols(colMap->getLocalNumElements(), true);
-    Magnitude eps = Teuchos::ScalarTraits<Magnitude>::eps();
-    for(size_t i=0; i<colMap->getLocalNumElements(); i++) {
-      dirichletCols[i] = Teuchos::ScalarTraits<Scalar>::magnitude(myCols[i])>2.0*eps;
+// Applies rowsum criterion
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyRowSumCriterion(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                         const typename Teuchos::ScalarTraits<Scalar>::magnitudeType rowSumTol,
+                         Kokkos::View<bool*, typename Node::device_type>& dirichletRows) {
+  typedef Teuchos::ScalarTraits<Scalar> STS;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowmap = A.getRowMap();
+
+  auto dirichletRowsHost = Kokkos::create_mirror_view(dirichletRows);
+  Kokkos::deep_copy(dirichletRowsHost, dirichletRows);
+
+  for (LocalOrdinal row = 0; row < Teuchos::as<LocalOrdinal>(rowmap->getLocalNumElements()); ++row) {
+    size_t nnz = A.getNumEntriesInLocalRow(row);
+    ArrayView<const LocalOrdinal> indices;
+    ArrayView<const Scalar> vals;
+    A.getLocalRowView(row, indices, vals);
+
+    Scalar rowsum  = STS::zero();
+    Scalar diagval = STS::zero();
+    for (LocalOrdinal colID = 0; colID < Teuchos::as<LocalOrdinal>(nnz); colID++) {
+      LocalOrdinal col = indices[colID];
+      if (row == col)
+        diagval = vals[colID];
+      rowsum += vals[colID];
     }
-    return dirichletCols;
+    if (STS::real(rowsum) > STS::magnitude(diagval) * rowSumTol)
+      dirichletRowsHost(row) = true;
   }
 
+  Kokkos::deep_copy(dirichletRows, dirichletRowsHost);
+}
 
-  template <class SC, class LO, class GO, class NO>
-  Kokkos::View<bool*, typename NO::device_type>
-  UtilitiesBase<SC, LO, GO, NO>::
-  DetectDirichletCols(const Xpetra::Matrix<SC,LO,GO,NO>& A,
-                      const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows) {
-    using ATS        = Kokkos::ArithTraits<SC>;
-    using impl_ATS = Kokkos::ArithTraits<typename ATS::val_type>;
-    using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
-
-    SC zero = ATS::zero();
-
-    auto localMatrix = A.getLocalMatrixDevice();
-    LO   numRows     = A.getLocalNumRows();
-
-    Teuchos::RCP<const Xpetra::Map<LO,GO,NO> > domMap = A.getDomainMap();
-    Teuchos::RCP<const Xpetra::Map<LO,GO,NO> > colMap = A.getColMap();
-    Teuchos::RCP<Xpetra::MultiVector<SC,LO,GO,NO> > myColsToZero = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(colMap,1);
-    myColsToZero->putScalar(zero);
-    auto myColsToZeroView = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
-    Kokkos::parallel_for("MueLu:Utils::DetectDirichletCols1", range_type(0,numRows),
-                         KOKKOS_LAMBDA(const LO row) {
-                           if (dirichletRows(row)) {
-                             auto rowView = localMatrix.row(row);
-                             auto length  = rowView.length;
-
-                             for (decltype(length) colID = 0; colID < length; colID++)
-                               myColsToZeroView(rowView.colidx(colID),0) = impl_ATS::one();
-                           }
-                         });
-
-    Teuchos::RCP<Xpetra::MultiVector<SC,LO,GO,NO> > globalColsToZero = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(domMap,1);
-    globalColsToZero->putScalar(zero);
-    Teuchos::RCP<Xpetra::Export<LO,GO,NO> > exporter = Xpetra::ExportFactory<LO,GO,NO>::Build(colMap,domMap);
-    // export to domain map
-    globalColsToZero->doExport(*myColsToZero,*exporter,Xpetra::ADD);
-    // import to column map
-    myColsToZero->doImport(*globalColsToZero,*exporter,Xpetra::INSERT);
-
-    auto myCols = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    size_t numColEntries = colMap->getLocalNumElements();
-    Kokkos::View<bool*, typename NO::device_type> dirichletCols(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), numColEntries);
-    const typename ATS::magnitudeType eps = 2.0*ATS::eps();
-
-    Kokkos::parallel_for("MueLu:Utils::DetectDirichletCols2", range_type(0,numColEntries),
-                         KOKKOS_LAMBDA (const size_t i) {
-                           dirichletCols(i) = impl_ATS::magnitude(myCols(i,0))>eps;
-                         });
-    return dirichletCols;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::ArrayRCP<const bool>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    DetectDirichletCols(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                        const Teuchos::ArrayRCP<const bool>& dirichletRows) {
+  Scalar zero                                                                               = Teuchos::ScalarTraits<Scalar>::zero();
+  Scalar one                                                                                = Teuchos::ScalarTraits<Scalar>::one();
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> domMap                 = A.getDomainMap();
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> colMap                 = A.getColMap();
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> myColsToZero = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(colMap, 1);
+  myColsToZero->putScalar(zero);
+  // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
+  for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+    if (dirichletRows[i]) {
+      Teuchos::ArrayView<const LocalOrdinal> indices;
+      Teuchos::ArrayView<const Scalar> values;
+      A.getLocalRowView(i, indices, values);
+      for (size_t j = 0; j < static_cast<size_t>(indices.size()); j++)
+        myColsToZero->replaceLocalValue(indices[j], 0, one);
+    }
   }
 
+  Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> globalColsToZero = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(domMap, 1);
+  globalColsToZero->putScalar(zero);
+  Teuchos::RCP<Xpetra::Export<LocalOrdinal, GlobalOrdinal, Node>> exporter = Xpetra::ExportFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(colMap, domMap);
+  // export to domain map
+  globalColsToZero->doExport(*myColsToZero, *exporter, Xpetra::ADD);
+  // import to column map
+  myColsToZero->doImport(*globalColsToZero, *exporter, Xpetra::INSERT);
+  Teuchos::ArrayRCP<const Scalar> myCols = myColsToZero->getData(0);
+  Teuchos::ArrayRCP<bool> dirichletCols(colMap->getLocalNumElements(), true);
+  Magnitude eps = Teuchos::ScalarTraits<Magnitude>::eps();
+  for (size_t i = 0; i < colMap->getLocalNumElements(); i++) {
+    dirichletCols[i] = Teuchos::ScalarTraits<Scalar>::magnitude(myCols[i]) > 2.0 * eps;
+  }
+  return dirichletCols;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Scalar
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Frobenius(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A, const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& B) {
-    // We check only row maps. Column may be different. One would hope that they are the same, as we typically
-    // calculate frobenius norm of the specified sparsity pattern with an updated matrix from the previous step,
-    // but matrix addition, even when one is submatrix of the other, changes column map (though change may be as
-    // simple as couple of elements swapped)
-    TEUCHOS_TEST_FOR_EXCEPTION(!A.getRowMap()->isSameAs(*B.getRowMap()),   Exceptions::Incompatible, "MueLu::CGSolver::Frobenius: row maps are incompatible");
-    TEUCHOS_TEST_FOR_EXCEPTION(!A.isFillComplete() || !B.isFillComplete(), Exceptions::RuntimeError, "Matrices must be fill completed");
-
-    const Map& AColMap = *A.getColMap();
-    const Map& BColMap = *B.getColMap();
-
-    Teuchos::ArrayView<const LocalOrdinal> indA, indB;
-    Teuchos::ArrayView<const Scalar>       valA, valB;
-    size_t nnzA = 0, nnzB = 0;
-
-    // We use a simple algorithm
-    // for each row we fill valBAll array with the values in the corresponding row of B
-    // as such, it serves as both sorted array and as storage, so we don't need to do a
-    // tricky problem: "find a value in the row of B corresponding to the specific GID"
-    // Once we do that, we translate LID of entries of row of A to LID of B, and multiply
-    // corresponding entries.
-    // The algorithm should be reasonably cheap, as it does not sort anything, provided
-    // that getLocalElement and getGlobalElement functions are reasonably effective. It
-    // *is* possible that the costs are hidden in those functions, but if maps are close
-    // to linear maps, we should be fine
-    Teuchos::Array<Scalar> valBAll(BColMap.getLocalNumElements());
-
-    LocalOrdinal  invalid = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
-    Scalar        zero    = Teuchos::ScalarTraits<Scalar>       ::zero(),    f = zero, gf;
-    size_t numRows = A.getLocalNumRows();
-    for (size_t i = 0; i < numRows; i++) {
-      A.getLocalRowView(i, indA, valA);
-      B.getLocalRowView(i, indB, valB);
-      nnzA = indA.size();
-      nnzB = indB.size();
-
-      // Set up array values
-      for (size_t j = 0; j < nnzB; j++)
-        valBAll[indB[j]] = valB[j];
-
-      for (size_t j = 0; j < nnzA; j++) {
-        // The cost of the whole Frobenius dot product function depends on the
-        // cost of the getLocalElement and getGlobalElement functions here.
-        LocalOrdinal ind = BColMap.getLocalElement(AColMap.getGlobalElement(indA[j]));
-        if (ind != invalid)
-          f += valBAll[ind] * valA[j];
-      }
+template <class SC, class LO, class GO, class NO>
+Kokkos::View<bool*, typename NO::device_type>
+UtilitiesBase<SC, LO, GO, NO>::
+    DetectDirichletCols(const Xpetra::Matrix<SC, LO, GO, NO>& A,
+                        const Kokkos::View<const bool*, typename NO::device_type>& dirichletRows) {
+  using ATS        = Kokkos::ArithTraits<SC>;
+  using impl_ATS   = Kokkos::ArithTraits<typename ATS::val_type>;
+  using range_type = Kokkos::RangePolicy<LO, typename NO::execution_space>;
+
+  SC zero = ATS::zero();
+
+  auto localMatrix = A.getLocalMatrixDevice();
+  LO numRows       = A.getLocalNumRows();
+
+  Teuchos::RCP<const Xpetra::Map<LO, GO, NO>> domMap             = A.getDomainMap();
+  Teuchos::RCP<const Xpetra::Map<LO, GO, NO>> colMap             = A.getColMap();
+  Teuchos::RCP<Xpetra::MultiVector<SC, LO, GO, NO>> myColsToZero = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(colMap, 1);
+  myColsToZero->putScalar(zero);
+  auto myColsToZeroView = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  // Find all local column indices that are in Dirichlet rows, record in myColsToZero as 1.0
+  Kokkos::parallel_for(
+      "MueLu:Utils::DetectDirichletCols1", range_type(0, numRows),
+      KOKKOS_LAMBDA(const LO row) {
+        if (dirichletRows(row)) {
+          auto rowView = localMatrix.row(row);
+          auto length  = rowView.length;
+
+          for (decltype(length) colID = 0; colID < length; colID++)
+            myColsToZeroView(rowView.colidx(colID), 0) = impl_ATS::one();
+        }
+      });
+
+  Teuchos::RCP<Xpetra::MultiVector<SC, LO, GO, NO>> globalColsToZero = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(domMap, 1);
+  globalColsToZero->putScalar(zero);
+  Teuchos::RCP<Xpetra::Export<LO, GO, NO>> exporter = Xpetra::ExportFactory<LO, GO, NO>::Build(colMap, domMap);
+  // export to domain map
+  globalColsToZero->doExport(*myColsToZero, *exporter, Xpetra::ADD);
+  // import to column map
+  myColsToZero->doImport(*globalColsToZero, *exporter, Xpetra::INSERT);
+
+  auto myCols          = myColsToZero->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  size_t numColEntries = colMap->getLocalNumElements();
+  Kokkos::View<bool*, typename NO::device_type> dirichletCols(Kokkos::ViewAllocateWithoutInitializing("dirichletCols"), numColEntries);
+  const typename ATS::magnitudeType eps = 2.0 * ATS::eps();
+
+  Kokkos::parallel_for(
+      "MueLu:Utils::DetectDirichletCols2", range_type(0, numColEntries),
+      KOKKOS_LAMBDA(const size_t i) {
+        dirichletCols(i) = impl_ATS::magnitude(myCols(i, 0)) > eps;
+      });
+  return dirichletCols;
+}
 
-      // Clean up array values
-      for (size_t j = 0; j < nnzB; j++)
-        valBAll[indB[j]] = zero;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Scalar
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Frobenius(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& B) {
+  // We check only row maps. Column may be different. One would hope that they are the same, as we typically
+  // calculate frobenius norm of the specified sparsity pattern with an updated matrix from the previous step,
+  // but matrix addition, even when one is submatrix of the other, changes column map (though change may be as
+  // simple as couple of elements swapped)
+  TEUCHOS_TEST_FOR_EXCEPTION(!A.getRowMap()->isSameAs(*B.getRowMap()), Exceptions::Incompatible, "MueLu::CGSolver::Frobenius: row maps are incompatible");
+  TEUCHOS_TEST_FOR_EXCEPTION(!A.isFillComplete() || !B.isFillComplete(), Exceptions::RuntimeError, "Matrices must be fill completed");
+
+  const Map& AColMap = *A.getColMap();
+  const Map& BColMap = *B.getColMap();
+
+  Teuchos::ArrayView<const LocalOrdinal> indA, indB;
+  Teuchos::ArrayView<const Scalar> valA, valB;
+  size_t nnzA = 0, nnzB = 0;
+
+  // We use a simple algorithm
+  // for each row we fill valBAll array with the values in the corresponding row of B
+  // as such, it serves as both sorted array and as storage, so we don't need to do a
+  // tricky problem: "find a value in the row of B corresponding to the specific GID"
+  // Once we do that, we translate LID of entries of row of A to LID of B, and multiply
+  // corresponding entries.
+  // The algorithm should be reasonably cheap, as it does not sort anything, provided
+  // that getLocalElement and getGlobalElement functions are reasonably effective. It
+  // *is* possible that the costs are hidden in those functions, but if maps are close
+  // to linear maps, we should be fine
+  Teuchos::Array<Scalar> valBAll(BColMap.getLocalNumElements());
+
+  LocalOrdinal invalid = Teuchos::OrdinalTraits<LocalOrdinal>::invalid();
+  Scalar zero = Teuchos::ScalarTraits<Scalar>::zero(), f = zero, gf;
+  size_t numRows = A.getLocalNumRows();
+  for (size_t i = 0; i < numRows; i++) {
+    A.getLocalRowView(i, indA, valA);
+    B.getLocalRowView(i, indB, valB);
+    nnzA = indA.size();
+    nnzB = indB.size();
+
+    // Set up array values
+    for (size_t j = 0; j < nnzB; j++)
+      valBAll[indB[j]] = valB[j];
+
+    for (size_t j = 0; j < nnzA; j++) {
+      // The cost of the whole Frobenius dot product function depends on the
+      // cost of the getLocalElement and getGlobalElement functions here.
+      LocalOrdinal ind = BColMap.getLocalElement(AColMap.getGlobalElement(indA[j]));
+      if (ind != invalid)
+        f += valBAll[ind] * valA[j];
     }
 
-    MueLu_sumAll(AColMap.getComm(), f, gf);
-
-    return gf;
+    // Clean up array values
+    for (size_t j = 0; j < nnzB; j++)
+      valBAll[indB[j]] = zero;
   }
 
+  MueLu_sumAll(AColMap.getComm(), f, gf);
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  SetRandomSeed(const Teuchos::Comm<int> &comm) {
-    // Distribute the seeds evenly in [1,maxint-1].  This guarantees nothing
-    // about where in random number stream we are, but avoids overflow situations
-    // in parallel when multiplying by a PID.  It would be better to use
-    // a good parallel random number generator.
-    double one = 1.0;
-    int maxint = INT_MAX; //= 2^31-1 = 2147483647 for 32-bit integers
-    int mySeed = Teuchos::as<int>((maxint-1) * (one -(comm.getRank()+1)/(comm.getSize()+one)) );
-    if (mySeed < 1 || mySeed == maxint) {
-      std::ostringstream errStr;
-      errStr << "Error detected with random seed = " << mySeed << ". It should be in the interval [1,2^31-2].";
-      throw Exceptions::RuntimeError(errStr.str());
-    }
-    std::srand(mySeed);
-    // For Tpetra, we could use Kokkos' random number generator here.
-    Teuchos::ScalarTraits<Scalar>::seedrandom(mySeed);
-    // Epetra
-    //   MultiVector::Random() -> Epetra_Util::RandomDouble() -> Epetra_Utils::RandomInt()
-    // Its own random number generator, based on Seed_. Seed_ is initialized in Epetra_Util constructor with std::rand()
-    // So our setting std::srand() affects that too
-  }
-
+  return gf;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    SetRandomSeed(const Teuchos::Comm<int>& comm) {
+  // Distribute the seeds evenly in [1,maxint-1].  This guarantees nothing
+  // about where in random number stream we are, but avoids overflow situations
+  // in parallel when multiplying by a PID.  It would be better to use
+  // a good parallel random number generator.
+  double one = 1.0;
+  int maxint = INT_MAX;  //= 2^31-1 = 2147483647 for 32-bit integers
+  int mySeed = Teuchos::as<int>((maxint - 1) * (one - (comm.getRank() + 1) / (comm.getSize() + one)));
+  if (mySeed < 1 || mySeed == maxint) {
+    std::ostringstream errStr;
+    errStr << "Error detected with random seed = " << mySeed << ". It should be in the interval [1,2^31-2].";
+    throw Exceptions::RuntimeError(errStr.str());
+  }
+  std::srand(mySeed);
+  // For Tpetra, we could use Kokkos' random number generator here.
+  Teuchos::ScalarTraits<Scalar>::seedrandom(mySeed);
+  // Epetra
+  //   MultiVector::Random() -> Epetra_Util::RandomDouble() -> Epetra_Utils::RandomInt()
+  // Its own random number generator, based on Seed_. Seed_ is initialized in Epetra_Util constructor with std::rand()
+  // So our setting std::srand() affects that too
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A,
-                    std::vector<LocalOrdinal>& dirichletRows, bool count_twos_as_dirichlet) {
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    dirichletRows.resize(0);
-    for(size_t i=0; i<A->getLocalNumRows(); i++) {
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> values;
-      A->getLocalRowView(i,indices,values);
-      int nnz=0;
-      for (size_t j=0; j<(size_t)indices.size(); j++) {
-        if (Teuchos::ScalarTraits<Scalar>::magnitude(values[j]) > Teuchos::ScalarTraits<MT>::eps()) {
-          nnz++;
-        }
-      }
-      if (nnz == 1 || (count_twos_as_dirichlet && nnz == 2)) {
-        dirichletRows.push_back(i);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                      std::vector<LocalOrdinal>& dirichletRows, bool count_twos_as_dirichlet) {
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  dirichletRows.resize(0);
+  for (size_t i = 0; i < A->getLocalNumRows(); i++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> values;
+    A->getLocalRowView(i, indices, values);
+    int nnz = 0;
+    for (size_t j = 0; j < (size_t)indices.size(); j++) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(values[j]) > Teuchos::ScalarTraits<MT>::eps()) {
+        nnz++;
       }
     }
+    if (nnz == 1 || (count_twos_as_dirichlet && nnz == 2)) {
+      dirichletRows.push_back(i);
+    }
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                         const std::vector<LocalOrdinal>& dirichletRows) {
+  RCP<const Map> Rmap = A->getRowMap();
+  RCP<const Map> Cmap = A->getColMap();
+  Scalar one          = Teuchos::ScalarTraits<Scalar>::one();
+  Scalar zero         = Teuchos::ScalarTraits<Scalar>::zero();
+
+  for (size_t i = 0; i < dirichletRows.size(); i++) {
+    GlobalOrdinal row_gid = Rmap->getGlobalElement(dirichletRows[i]);
+
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> values;
+    A->getLocalRowView(dirichletRows[i], indices, values);
+    // NOTE: This won't work with fancy node types.
+    Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
+    for (size_t j = 0; j < (size_t)indices.size(); j++) {
+      if (Cmap->getGlobalElement(indices[j]) == row_gid)
+        valuesNC[j] = one;
+      else
+        valuesNC[j] = zero;
+    }
+  }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                       const std::vector<LocalOrdinal>& dirichletRows) {
-    RCP<const Map> Rmap = A->getRowMap();
-    RCP<const Map> Cmap = A->getColMap();
-    Scalar one  =Teuchos::ScalarTraits<Scalar>::one();
-    Scalar zero =Teuchos::ScalarTraits<Scalar>::zero();
-
-    for(size_t i=0; i<dirichletRows.size(); i++) {
-      GlobalOrdinal row_gid = Rmap->getGlobalElement(dirichletRows[i]);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                         const Teuchos::ArrayRCP<const bool>& dirichletRows) {
+  TEUCHOS_ASSERT(A->isFillComplete());
+  RCP<const Map> domMap = A->getDomainMap();
+  RCP<const Map> ranMap = A->getRangeMap();
+  RCP<const Map> Rmap   = A->getRowMap();
+  RCP<const Map> Cmap   = A->getColMap();
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == Rmap->getLocalNumElements());
+  const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
+  A->resumeFill();
+  for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+    if (dirichletRows[i]) {
+      GlobalOrdinal row_gid = Rmap->getGlobalElement(i);
 
       Teuchos::ArrayView<const LocalOrdinal> indices;
       Teuchos::ArrayView<const Scalar> values;
-      A->getLocalRowView(dirichletRows[i],indices,values);
-      // NOTE: This won't work with fancy node types.
-      Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
-      for(size_t j=0; j<(size_t)indices.size(); j++) {
-        if(Cmap->getGlobalElement(indices[j])==row_gid)
-          valuesNC[j]=one;
+      A->getLocalRowView(i, indices, values);
+
+      Teuchos::ArrayRCP<Scalar> valuesNC(values.size());
+      for (size_t j = 0; j < (size_t)indices.size(); j++) {
+        if (Cmap->getGlobalElement(indices[j]) == row_gid)
+          valuesNC[j] = one;
         else
-          valuesNC[j]=zero;
+          valuesNC[j] = zero;
       }
+      A->replaceLocalValues(i, indices, valuesNC());
     }
   }
+  A->fillComplete(domMap, ranMap);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                       const Teuchos::ArrayRCP<const bool>& dirichletRows) {
-    TEUCHOS_ASSERT(A->isFillComplete());
-    RCP<const Map> domMap = A->getDomainMap();
-    RCP<const Map> ranMap = A->getRangeMap();
-    RCP<const Map> Rmap = A->getRowMap();
-    RCP<const Map> Cmap = A->getColMap();
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == Rmap->getLocalNumElements());
-    const Scalar one  = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar zero = Teuchos::ScalarTraits<Scalar>::zero();
-    A->resumeFill();
-    for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-      if (dirichletRows[i]){
-        GlobalOrdinal row_gid = Rmap->getGlobalElement(i);
-
-        Teuchos::ArrayView<const LocalOrdinal> indices;
-        Teuchos::ArrayView<const Scalar> values;
-        A->getLocalRowView(i,indices,values);
-
-        Teuchos::ArrayRCP<Scalar> valuesNC(values.size());
-        for(size_t j=0; j<(size_t)indices.size(); j++) {
-          if(Cmap->getGlobalElement(indices[j])==row_gid)
-            valuesNC[j]=one;
-          else
-            valuesNC[j]=zero;
-        }
-        A->replaceLocalValues(i,indices,valuesNC());
-      }
-    }
-    A->fillComplete(domMap, ranMap);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                         const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows) {
+  TEUCHOS_ASSERT(A->isFillComplete());
+  using ATS        = Kokkos::ArithTraits<Scalar>;
+  using impl_ATS   = Kokkos::ArithTraits<typename ATS::val_type>;
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> domMap = A->getDomainMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> ranMap = A->getRangeMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> Rmap   = A->getRowMap();
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> Cmap   = A->getColMap();
+
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == Rmap->getLocalNumElements());
+
+  auto localMatrix = A->getLocalMatrixDevice();
+  auto localRmap   = Rmap->getLocalMap();
+  auto localCmap   = Cmap->getLocalMap();
+
+  Kokkos::parallel_for(
+      "MueLu::Utils::ApplyOAZ", range_type(0, dirichletRows.extent(0)),
+      KOKKOS_LAMBDA(const LocalOrdinal row) {
+        if (dirichletRows(row)) {
+          auto rowView = localMatrix.row(row);
+          auto length  = rowView.length;
+          auto row_gid = localRmap.getGlobalElement(row);
+          auto row_lid = localCmap.getLocalElement(row_gid);
+
+          for (decltype(length) colID = 0; colID < length; colID++)
+            if (rowView.colidx(colID) == row_lid)
+              rowView.value(colID) = impl_ATS::one();
+            else
+              rowView.value(colID) = impl_ATS::zero();
+        }
+      });
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ApplyOAZToMatrixRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                       const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows) {
-    TEUCHOS_ASSERT(A->isFillComplete());
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using impl_ATS = Kokkos::ArithTraits<typename ATS::val_type>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domMap = A->getDomainMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > ranMap = A->getRangeMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > Rmap = A->getRowMap();
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > Cmap = A->getColMap();
-
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == Rmap->getLocalNumElements());
-
-    auto localMatrix = A->getLocalMatrixDevice();
-    auto localRmap = Rmap->getLocalMap();
-    auto localCmap = Cmap->getLocalMap();
-
-    Kokkos::parallel_for("MueLu::Utils::ApplyOAZ",range_type(0,dirichletRows.extent(0)),
-                         KOKKOS_LAMBDA(const LocalOrdinal row) {
-                           if (dirichletRows(row)){
-                             auto rowView = localMatrix.row(row);
-                             auto length = rowView.length;
-                             auto row_gid = localRmap.getGlobalElement(row);
-                             auto row_lid = localCmap.getLocalElement(row_gid);
-
-                             for (decltype(length) colID = 0; colID < length; colID++)
-                               if (rowView.colidx(colID) == row_lid)
-                                 rowView.value(colID) = impl_ATS::one();
-                               else
-                                 rowView.value(colID) = impl_ATS::zero();
-                           }
-                         });
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                      const std::vector<LocalOrdinal>& dirichletRows,
+                      Scalar replaceWith) {
+  for (size_t i = 0; i < dirichletRows.size(); i++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> values;
+    A->getLocalRowView(dirichletRows[i], indices, values);
+    // NOTE: This won't work with fancy node types.
+    Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
+    for (size_t j = 0; j < (size_t)indices.size(); j++)
+      valuesNC[j] = replaceWith;
   }
+}
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                    const std::vector<LocalOrdinal>& dirichletRows,
-                    Scalar replaceWith) {
-    for(size_t i=0; i<dirichletRows.size(); i++) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                      const Teuchos::ArrayRCP<const bool>& dirichletRows,
+                      Scalar replaceWith) {
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == A->getRowMap()->getLocalNumElements());
+  for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+    if (dirichletRows[i]) {
       Teuchos::ArrayView<const LocalOrdinal> indices;
       Teuchos::ArrayView<const Scalar> values;
-      A->getLocalRowView(dirichletRows[i],indices,values);
+      A->getLocalRowView(i, indices, values);
       // NOTE: This won't work with fancy node types.
       Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
-      for(size_t j=0; j<(size_t)indices.size(); j++)
-        valuesNC[j]=replaceWith;
+      for (size_t j = 0; j < (size_t)indices.size(); j++)
+        valuesNC[j] = replaceWith;
     }
   }
+}
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletRows(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                    const Teuchos::ArrayRCP<const bool>& dirichletRows,
-                    Scalar replaceWith) {
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == A->getRowMap()->getLocalNumElements());
-    for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-      if (dirichletRows[i]) {
-        Teuchos::ArrayView<const LocalOrdinal> indices;
-        Teuchos::ArrayView<const Scalar> values;
-        A->getLocalRowView(i,indices,values);
-        // NOTE: This won't work with fancy node types.
-        Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
-        for(size_t j=0; j<(size_t)indices.size(); j++)
-          valuesNC[j]=replaceWith;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletRows(Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                      const Teuchos::ArrayRCP<const bool>& dirichletRows,
+                      Scalar replaceWith) {
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == X->getMap()->getLocalNumElements());
+  for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+    if (dirichletRows[i]) {
+      for (size_t j = 0; j < X->getNumVectors(); j++)
+        X->replaceLocalValue(i, j, replaceWith);
     }
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletRows(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                      const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows,
+                      Scalar replaceWith) {
+  using ATS        = Kokkos::ArithTraits<Scalar>;
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletRows(Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& X,
-                         const Teuchos::ArrayRCP<const bool>& dirichletRows,
-                         Scalar replaceWith) {
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletRows.size()) == X->getMap()->getLocalNumElements());
-    for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-      if (dirichletRows[i]) {
-        for(size_t j=0; j<X->getNumVectors(); j++)
-          X->replaceLocalValue(i,j,replaceWith);
-      }
-    }
-  }
-
+  typename ATS::val_type impl_replaceWith = replaceWith;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletRows(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-                    const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows,
-                    Scalar replaceWith) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    typename ATS::val_type impl_replaceWith = replaceWith;
-
-    auto localMatrix = A->getLocalMatrixDevice();
-    LocalOrdinal numRows = A->getLocalNumRows();
-
-    Kokkos::parallel_for("MueLu:Utils::ZeroDirichletRows", range_type(0,numRows),
-                         KOKKOS_LAMBDA(const LocalOrdinal row) {
-                           if (dirichletRows(row)) {
-                             auto rowView = localMatrix.row(row);
-                             auto length  = rowView.length;
-                             for (decltype(length) colID = 0; colID < length; colID++)
-                               rowView.value(colID) = impl_replaceWith;
-                           }
-                         });
-  }
+  auto localMatrix     = A->getLocalMatrixDevice();
+  LocalOrdinal numRows = A->getLocalNumRows();
 
+  Kokkos::parallel_for(
+      "MueLu:Utils::ZeroDirichletRows", range_type(0, numRows),
+      KOKKOS_LAMBDA(const LocalOrdinal row) {
+        if (dirichletRows(row)) {
+          auto rowView = localMatrix.row(row);
+          auto length  = rowView.length;
+          for (decltype(length) colID = 0; colID < length; colID++)
+            rowView.value(colID) = impl_replaceWith;
+        }
+      });
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletRows(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X,
-                    const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows,
-                    Scalar replaceWith) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    typename ATS::val_type impl_replaceWith = replaceWith;
-
-    auto myCols = X->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    size_t numVecs = X->getNumVectors();
-    Kokkos::parallel_for("MueLu:Utils::ZeroDirichletRows_MV", range_type(0,dirichletRows.size()),
-                         KOKKOS_LAMBDA(const size_t i) {
-                           if (dirichletRows(i)) {
-                             for(size_t j=0; j<numVecs; j++)
-                               myCols(i,j) = impl_replaceWith;
-                           }
-                         });
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletRows(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                      const Kokkos::View<const bool*, typename Node::device_type>& dirichletRows,
+                      Scalar replaceWith) {
+  using ATS        = Kokkos::ArithTraits<Scalar>;
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
+  typename ATS::val_type impl_replaceWith = replaceWith;
 
+  auto myCols    = X->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  size_t numVecs = X->getNumVectors();
+  Kokkos::parallel_for(
+      "MueLu:Utils::ZeroDirichletRows_MV", range_type(0, dirichletRows.size()),
+      KOKKOS_LAMBDA(const size_t i) {
+        if (dirichletRows(i)) {
+          for (size_t j = 0; j < numVecs; j++)
+            myCols(i, j) = impl_replaceWith;
+        }
+      });
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletCols(Teuchos::RCP<Matrix>& A,
-                    const Teuchos::ArrayRCP<const bool>& dirichletCols,
-                    Scalar replaceWith) {
-    TEUCHOS_ASSERT(static_cast<size_t>(dirichletCols.size()) == A->getColMap()->getLocalNumElements());
-    for(size_t i=0; i<A->getLocalNumRows(); i++) {
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> values;
-      A->getLocalRowView(i,indices,values);
-      // NOTE: This won't work with fancy node types.
-      Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
-      for(size_t j=0; j<static_cast<size_t>(indices.size()); j++)
-        if (dirichletCols[indices[j]])
-          valuesNC[j] = replaceWith;
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletCols(Teuchos::RCP<Matrix>& A,
+                      const Teuchos::ArrayRCP<const bool>& dirichletCols,
+                      Scalar replaceWith) {
+  TEUCHOS_ASSERT(static_cast<size_t>(dirichletCols.size()) == A->getColMap()->getLocalNumElements());
+  for (size_t i = 0; i < A->getLocalNumRows(); i++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> values;
+    A->getLocalRowView(i, indices, values);
+    // NOTE: This won't work with fancy node types.
+    Scalar* valuesNC = const_cast<Scalar*>(values.getRawPtr());
+    for (size_t j = 0; j < static_cast<size_t>(indices.size()); j++)
+      if (dirichletCols[indices[j]])
+        valuesNC[j] = replaceWith;
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ZeroDirichletCols(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                      const Kokkos::View<const bool*, typename Node::device_type>& dirichletCols,
+                      Scalar replaceWith) {
+  using ATS        = Kokkos::ArithTraits<Scalar>;
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ZeroDirichletCols(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                    const Kokkos::View<const bool*, typename Node::device_type>& dirichletCols,
-                    Scalar replaceWith) {
-    using ATS        = Kokkos::ArithTraits<Scalar>;
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    typename ATS::val_type impl_replaceWith = replaceWith;
-
-    auto localMatrix = A->getLocalMatrixDevice();
-    LocalOrdinal numRows = A->getLocalNumRows();
-
-    Kokkos::parallel_for("MueLu:Utils::ZeroDirichletCols", range_type(0,numRows),
-                         KOKKOS_LAMBDA(const LocalOrdinal row) {
-                           auto rowView = localMatrix.row(row);
-                           auto length  = rowView.length;
-                           for (decltype(length) colID = 0; colID < length; colID++)
-                             if (dirichletCols(rowView.colidx(colID))) {
-                               rowView.value(colID) = impl_replaceWith;
-                             }
-                         });
-  }
+  typename ATS::val_type impl_replaceWith = replaceWith;
 
+  auto localMatrix     = A->getLocalMatrixDevice();
+  LocalOrdinal numRows = A->getLocalNumRows();
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  FindDirichletRowsAndPropagateToCols(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A,
-                                      Teuchos::RCP<Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> >& isDirichletRow,
-                                      Teuchos::RCP<Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> >& isDirichletCol) {
+  Kokkos::parallel_for(
+      "MueLu:Utils::ZeroDirichletCols", range_type(0, numRows),
+      KOKKOS_LAMBDA(const LocalOrdinal row) {
+        auto rowView = localMatrix.row(row);
+        auto length  = rowView.length;
+        for (decltype(length) colID = 0; colID < length; colID++)
+          if (dirichletCols(rowView.colidx(colID))) {
+            rowView.value(colID) = impl_replaceWith;
+          }
+      });
+}
 
-    // Make sure A's RowMap == DomainMap
-    if(!A->getRowMap()->isSameAs(*A->getDomainMap())) {
-      throw std::runtime_error("UtilitiesBase::FindDirichletRowsAndPropagateToCols row and domain maps must match.");
-    }
-    RCP<const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer = A->getCrsGraph()->getImporter();
-    bool has_import = !importer.is_null();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    FindDirichletRowsAndPropagateToCols(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                                        Teuchos::RCP<Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node>>& isDirichletRow,
+                                        Teuchos::RCP<Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node>>& isDirichletCol) {
+  // Make sure A's RowMap == DomainMap
+  if (!A->getRowMap()->isSameAs(*A->getDomainMap())) {
+    throw std::runtime_error("UtilitiesBase::FindDirichletRowsAndPropagateToCols row and domain maps must match.");
+  }
+  RCP<const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>> importer = A->getCrsGraph()->getImporter();
+  bool has_import                                                       = !importer.is_null();
 
-    // Find the Dirichlet rows
-    std::vector<LocalOrdinal> dirichletRows;
-    FindDirichletRows(A,dirichletRows);
+  // Find the Dirichlet rows
+  std::vector<LocalOrdinal> dirichletRows;
+  FindDirichletRows(A, dirichletRows);
 
 #if 0
     printf("[%d] DirichletRow Ids = ",A->getRowMap()->getComm()->getRank());
@@ -1784,206 +1721,194 @@ namespace MueLu {
     printf("\n");
     fflush(stdout);
 #endif
-    // Allocate all as non-Dirichlet
-    isDirichletRow = Xpetra::VectorFactory<int,LocalOrdinal,GlobalOrdinal,Node>::Build(A->getRowMap(),true);
-    isDirichletCol = Xpetra::VectorFactory<int,LocalOrdinal,GlobalOrdinal,Node>::Build(A->getColMap(),true);
-
-    {
-      Teuchos::ArrayRCP<int> dr_rcp = isDirichletRow->getDataNonConst(0);
-      Teuchos::ArrayView<int> dr    = dr_rcp();
-      Teuchos::ArrayRCP<int> dc_rcp = isDirichletCol->getDataNonConst(0);
-      Teuchos::ArrayView<int> dc    = dc_rcp();
-      for(size_t i=0; i<(size_t) dirichletRows.size(); i++) {
-        dr[dirichletRows[i]] = 1;
-        if(!has_import) dc[dirichletRows[i]] = 1;
-      }
-    }
-
-    if(has_import)
-      isDirichletCol->doImport(*isDirichletRow,*importer,Xpetra::CombineMode::ADD);
+  // Allocate all as non-Dirichlet
+  isDirichletRow = Xpetra::VectorFactory<int, LocalOrdinal, GlobalOrdinal, Node>::Build(A->getRowMap(), true);
+  isDirichletCol = Xpetra::VectorFactory<int, LocalOrdinal, GlobalOrdinal, Node>::Build(A->getColMap(), true);
 
+  {
+    Teuchos::ArrayRCP<int> dr_rcp = isDirichletRow->getDataNonConst(0);
+    Teuchos::ArrayView<int> dr    = dr_rcp();
+    Teuchos::ArrayRCP<int> dc_rcp = isDirichletCol->getDataNonConst(0);
+    Teuchos::ArrayView<int> dc    = dc_rcp();
+    for (size_t i = 0; i < (size_t)dirichletRows.size(); i++) {
+      dr[dirichletRows[i]] = 1;
+      if (!has_import) dc[dirichletRows[i]] = 1;
+    }
   }
 
+  if (has_import)
+    isDirichletCol->doImport(*isDirichletRow, *importer, Xpetra::CombineMode::ADD);
+}
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
 UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-ReplaceNonZerosWithOnes(const RCP<Matrix> & original) {
-  using ISC = typename Kokkos::ArithTraits<Scalar>::val_type;
-  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+    ReplaceNonZerosWithOnes(const RCP<Matrix>& original) {
+  using ISC               = typename Kokkos::ArithTraits<Scalar>::val_type;
+  using range_type        = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
   using local_matrix_type = typename CrsMatrix::local_matrix_type;
-  using values_type = typename local_matrix_type::values_type;
+  using values_type       = typename local_matrix_type::values_type;
 
   const ISC ONE  = Kokkos::ArithTraits<ISC>::one();
   const ISC ZERO = Kokkos::ArithTraits<ISC>::zero();
 
   // Copy the values array of the old matrix to a new array, replacing all the non-zeros with one
   auto localMatrix = original->getLocalMatrixDevice();
-  TEUCHOS_TEST_FOR_EXCEPTION(!original->hasCrsGraph(),Exceptions::RuntimeError, "ReplaceNonZerosWithOnes: Cannot get CrsGraph");
-  values_type new_values("values",localMatrix.nnz());
- 
-  Kokkos::parallel_for("ReplaceNonZerosWithOnes",range_type(0,localMatrix.nnz()),KOKKOS_LAMBDA(const size_t i) {
-      if (localMatrix.values(i) != ZERO)
-        new_values(i) = ONE;
-      else
-        new_values(i) = ZERO;
-    });
+  TEUCHOS_TEST_FOR_EXCEPTION(!original->hasCrsGraph(), Exceptions::RuntimeError, "ReplaceNonZerosWithOnes: Cannot get CrsGraph");
+  values_type new_values("values", localMatrix.nnz());
+
+  Kokkos::parallel_for(
+      "ReplaceNonZerosWithOnes", range_type(0, localMatrix.nnz()), KOKKOS_LAMBDA(const size_t i) {
+        if (localMatrix.values(i) != ZERO)
+          new_values(i) = ONE;
+        else
+          new_values(i) = ZERO;
+      });
 
   // Build the new matrix
-  RCP<Matrix> NewMatrix =  Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(original->getCrsGraph(),new_values);
-  TEUCHOS_TEST_FOR_EXCEPTION(NewMatrix.is_null(),Exceptions::RuntimeError, "ReplaceNonZerosWithOnes: MatrixFactory::Build() did not return matrix");
-  NewMatrix->fillComplete(original->getDomainMap(),original->getRangeMap());
+  RCP<Matrix> NewMatrix = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(original->getCrsGraph(), new_values);
+  TEUCHOS_TEST_FOR_EXCEPTION(NewMatrix.is_null(), Exceptions::RuntimeError, "ReplaceNonZerosWithOnes: MatrixFactory::Build() did not return matrix");
+  NewMatrix->fillComplete(original->getDomainMap(), original->getRangeMap());
   return NewMatrix;
 }
 
-
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Xpetra::BlockedMap<LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  GeneratedBlockedTargetMap(const Xpetra::BlockedMap<LocalOrdinal,GlobalOrdinal,Node> & sourceBlockedMap,
-                            const Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> & Importer) {
-    typedef Xpetra::Vector<int,LocalOrdinal,GlobalOrdinal,Node> IntVector;
-    Xpetra::UnderlyingLib lib = sourceBlockedMap.lib();
-
-    // De-stride the map if we have to (might regret this later)
-    RCP<const Map> fullMap    = sourceBlockedMap.getMap();
-    RCP<const Map> stridedMap = Teuchos::rcp_dynamic_cast<const Xpetra::StridedMap<LocalOrdinal,GlobalOrdinal,Node> >(fullMap);
-    if(!stridedMap.is_null()) fullMap = stridedMap->getMap();
-
-    // Initial sanity checking for map compatibil
-    const size_t numSubMaps = sourceBlockedMap.getNumMaps();
-    if(!Importer.getSourceMap()->isCompatible(*fullMap))
-      throw std::runtime_error("GenerateBlockedTargetMap(): Map compatibility error");
-
-    // Build an indicator vector
-    RCP<IntVector> block_ids = Xpetra::VectorFactory<int,LocalOrdinal,GlobalOrdinal,Node>::Build(fullMap);
-
-    for(size_t i=0; i<numSubMaps; i++) {
-      RCP<const Map> map = sourceBlockedMap.getMap(i);
-
-      for(size_t j=0; j<map->getLocalNumElements(); j++)  {
-        LocalOrdinal jj = fullMap->getLocalElement(map->getGlobalElement(j));
-        block_ids->replaceLocalValue(jj,(int)i);
-      }
-    }
-
-    // Get the block ids for the new map
-    RCP<const Map> targetMap = Importer.getTargetMap();
-    RCP<IntVector> new_block_ids = Xpetra::VectorFactory<int,LocalOrdinal,GlobalOrdinal,Node>::Build(targetMap);
-    new_block_ids->doImport(*block_ids,Importer,Xpetra::CombineMode::ADD);
-    Teuchos::ArrayRCP<const int> dataRCP = new_block_ids->getData(0);
-    Teuchos::ArrayView<const int> data = dataRCP();
-
-
-    // Get the GIDs for each subblock
-    Teuchos::Array<Teuchos::Array<GlobalOrdinal> > elementsInSubMap(numSubMaps);
-    for(size_t i=0; i<targetMap->getLocalNumElements(); i++) {
-      elementsInSubMap[data[i]].push_back(targetMap->getGlobalElement(i));
-    }
-
-    // Generate the new submaps
-    std::vector<RCP<const Map> > subMaps(numSubMaps);
-    for(size_t i=0; i<numSubMaps; i++) {
-      subMaps[i] = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(lib,Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(),elementsInSubMap[i](),targetMap->getIndexBase(),targetMap->getComm());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    GeneratedBlockedTargetMap(const Xpetra::BlockedMap<LocalOrdinal, GlobalOrdinal, Node>& sourceBlockedMap,
+                              const Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node>& Importer) {
+  typedef Xpetra::Vector<int, LocalOrdinal, GlobalOrdinal, Node> IntVector;
+  Xpetra::UnderlyingLib lib = sourceBlockedMap.lib();
+
+  // De-stride the map if we have to (might regret this later)
+  RCP<const Map> fullMap    = sourceBlockedMap.getMap();
+  RCP<const Map> stridedMap = Teuchos::rcp_dynamic_cast<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node>>(fullMap);
+  if (!stridedMap.is_null()) fullMap = stridedMap->getMap();
+
+  // Initial sanity checking for map compatibil
+  const size_t numSubMaps = sourceBlockedMap.getNumMaps();
+  if (!Importer.getSourceMap()->isCompatible(*fullMap))
+    throw std::runtime_error("GenerateBlockedTargetMap(): Map compatibility error");
+
+  // Build an indicator vector
+  RCP<IntVector> block_ids = Xpetra::VectorFactory<int, LocalOrdinal, GlobalOrdinal, Node>::Build(fullMap);
+
+  for (size_t i = 0; i < numSubMaps; i++) {
+    RCP<const Map> map = sourceBlockedMap.getMap(i);
+
+    for (size_t j = 0; j < map->getLocalNumElements(); j++) {
+      LocalOrdinal jj = fullMap->getLocalElement(map->getGlobalElement(j));
+      block_ids->replaceLocalValue(jj, (int)i);
     }
+  }
 
-    // Build the BlockedMap
-    return rcp(new BlockedMap(targetMap,subMaps));
+  // Get the block ids for the new map
+  RCP<const Map> targetMap     = Importer.getTargetMap();
+  RCP<IntVector> new_block_ids = Xpetra::VectorFactory<int, LocalOrdinal, GlobalOrdinal, Node>::Build(targetMap);
+  new_block_ids->doImport(*block_ids, Importer, Xpetra::CombineMode::ADD);
+  Teuchos::ArrayRCP<const int> dataRCP = new_block_ids->getData(0);
+  Teuchos::ArrayView<const int> data   = dataRCP();
+
+  // Get the GIDs for each subblock
+  Teuchos::Array<Teuchos::Array<GlobalOrdinal>> elementsInSubMap(numSubMaps);
+  for (size_t i = 0; i < targetMap->getLocalNumElements(); i++) {
+    elementsInSubMap[data[i]].push_back(targetMap->getGlobalElement(i));
+  }
 
+  // Generate the new submaps
+  std::vector<RCP<const Map>> subMaps(numSubMaps);
+  for (size_t i = 0; i < numSubMaps; i++) {
+    subMaps[i] = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, Teuchos::OrdinalTraits<GlobalOrdinal>::invalid(), elementsInSubMap[i](), targetMap->getIndexBase(), targetMap->getComm());
   }
 
+  // Build the BlockedMap
+  return rcp(new BlockedMap(targetMap, subMaps));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  MapsAreNested(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& rowMap, const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& colMap) {
-    ArrayView<const GlobalOrdinal> rowElements = rowMap.getLocalElementList();
-    ArrayView<const GlobalOrdinal> colElements = colMap.getLocalElementList();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    MapsAreNested(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& rowMap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& colMap) {
+  ArrayView<const GlobalOrdinal> rowElements = rowMap.getLocalElementList();
+  ArrayView<const GlobalOrdinal> colElements = colMap.getLocalElementList();
 
-    const size_t numElements = rowElements.size();
+  const size_t numElements = rowElements.size();
 
-    if (size_t(colElements.size()) < numElements)
-      return false;
-
-    bool goodMap = true;
-    for (size_t i = 0; i < numElements; i++)
-      if (rowElements[i] != colElements[i]) {
-        goodMap = false;
-        break;
-      }
-
-    return goodMap;
-  }
+  if (size_t(colElements.size()) < numElements)
+    return false;
 
+  bool goodMap = true;
+  for (size_t i = 0; i < numElements; i++)
+    if (rowElements[i] != colElements[i]) {
+      goodMap = false;
+      break;
+    }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ReverseCuthillMcKee(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &Op) {
-    using local_matrix_type = typename Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::local_matrix_type;
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using lno_nnz_view_t    = typename local_graph_type::entries_type::non_const_type;
-    using device            = typename local_graph_type::device_type;
-    using execution_space   = typename local_matrix_type::execution_space;
-    using ordinal_type      = typename local_matrix_type::ordinal_type;
-
-    local_graph_type localGraph = Op.getLocalMatrixDevice().graph;
-
-    lno_nnz_view_t rcmOrder = KokkosGraph::Experimental::graph_rcm
-      <device, typename local_graph_type::row_map_type, typename local_graph_type::entries_type, lno_nnz_view_t>
-      (localGraph.row_map, localGraph.entries);
-
-    RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > retval =
-      Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(Op.getRowMap());
-
-    // Copy out and reorder data
-    auto view1D = Kokkos::subview(retval->getDeviceLocalView(Xpetra::Access::ReadWrite),Kokkos::ALL (), 0);
-    Kokkos::parallel_for("Utilities::ReverseCuthillMcKee",
-                         Kokkos::RangePolicy<ordinal_type, execution_space>(0, localGraph.numRows()),
-                         KOKKOS_LAMBDA(const ordinal_type rowIdx) {
-                           view1D(rcmOrder(rowIdx)) = rowIdx;
-                         });
-    return retval;
-  }
+  return goodMap;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> >
-  UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  CuthillMcKee(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &Op) {
-    using local_matrix_type = typename Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::local_matrix_type;
-    using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
-    using lno_nnz_view_t    = typename local_graph_type::entries_type::non_const_type;
-    using device            = typename local_graph_type::device_type;
-    using execution_space   = typename local_matrix_type::execution_space;
-    using ordinal_type      = typename local_matrix_type::ordinal_type;
-
-    local_graph_type localGraph = Op.getLocalMatrixDevice().graph;
-    LocalOrdinal numRows = localGraph.numRows();
-
-    lno_nnz_view_t rcmOrder = KokkosGraph::Experimental::graph_rcm
-      <device, typename local_graph_type::row_map_type, typename local_graph_type::entries_type, lno_nnz_view_t>
-      (localGraph.row_map, localGraph.entries);
-
-    RCP<Xpetra::Vector<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > retval =
-      Xpetra::VectorFactory<LocalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(Op.getRowMap());
-
-    // Copy out data
-    auto view1D = Kokkos::subview(retval->getDeviceLocalView(Xpetra::Access::ReadWrite),Kokkos::ALL (), 0);
-    // Since KokkosKernels produced RCM, also reverse the order of the view to get CM
-    Kokkos::parallel_for("Utilities::ReverseCuthillMcKee",
-                         Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
-                         KOKKOS_LAMBDA(const ordinal_type rowIdx) {
-                           view1D(rcmOrder(numRows - 1 - rowIdx)) = rowIdx;
-                         });
-    return retval;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ReverseCuthillMcKee(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  using local_matrix_type = typename Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_matrix_type;
+  using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
+  using lno_nnz_view_t    = typename local_graph_type::entries_type::non_const_type;
+  using device            = typename local_graph_type::device_type;
+  using execution_space   = typename local_matrix_type::execution_space;
+  using ordinal_type      = typename local_matrix_type::ordinal_type;
+
+  local_graph_type localGraph = Op.getLocalMatrixDevice().graph;
+
+  lno_nnz_view_t rcmOrder = KokkosGraph::Experimental::graph_rcm<device, typename local_graph_type::row_map_type, typename local_graph_type::entries_type, lno_nnz_view_t>(localGraph.row_map, localGraph.entries);
+
+  RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> retval =
+      Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(Op.getRowMap());
+
+  // Copy out and reorder data
+  auto view1D = Kokkos::subview(retval->getDeviceLocalView(Xpetra::Access::ReadWrite), Kokkos::ALL(), 0);
+  Kokkos::parallel_for(
+      "Utilities::ReverseCuthillMcKee",
+      Kokkos::RangePolicy<ordinal_type, execution_space>(0, localGraph.numRows()),
+      KOKKOS_LAMBDA(const ordinal_type rowIdx) {
+        view1D(rcmOrder(rowIdx)) = rowIdx;
+      });
+  return retval;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>>
+UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    CuthillMcKee(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  using local_matrix_type = typename Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_matrix_type;
+  using local_graph_type  = typename local_matrix_type::staticcrsgraph_type;
+  using lno_nnz_view_t    = typename local_graph_type::entries_type::non_const_type;
+  using device            = typename local_graph_type::device_type;
+  using execution_space   = typename local_matrix_type::execution_space;
+  using ordinal_type      = typename local_matrix_type::ordinal_type;
+
+  local_graph_type localGraph = Op.getLocalMatrixDevice().graph;
+  LocalOrdinal numRows        = localGraph.numRows();
+
+  lno_nnz_view_t rcmOrder = KokkosGraph::Experimental::graph_rcm<device, typename local_graph_type::row_map_type, typename local_graph_type::entries_type, lno_nnz_view_t>(localGraph.row_map, localGraph.entries);
+
+  RCP<Xpetra::Vector<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>> retval =
+      Xpetra::VectorFactory<LocalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(Op.getRowMap());
+
+  // Copy out data
+  auto view1D = Kokkos::subview(retval->getDeviceLocalView(Xpetra::Access::ReadWrite), Kokkos::ALL(), 0);
+  // Since KokkosKernels produced RCM, also reverse the order of the view to get CM
+  Kokkos::parallel_for(
+      "Utilities::ReverseCuthillMcKee",
+      Kokkos::RangePolicy<ordinal_type, execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const ordinal_type rowIdx) {
+        view1D(rcmOrder(numRows - 1 - rowIdx)) = rowIdx;
+      });
+  return retval;
+}
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UTILITIESBASE_SHORT
-#endif // MUELU_UTILITIESBASE_DEF_HPP
+#endif  // MUELU_UTILITIESBASE_DEF_HPP
 
 //  LocalWords:  LocalOrdinal
diff --git a/packages/muelu/src/Utils/MueLu_Utilities_decl.hpp b/packages/muelu/src/Utils/MueLu_Utilities_decl.hpp
index 1baf5aae3564..b175b78fc819 100644
--- a/packages/muelu/src/Utils/MueLu_Utilities_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_Utilities_decl.hpp
@@ -72,7 +72,7 @@
 #include <Xpetra_EpetraCrsMatrix.hpp>
 
 // needed because of inlined function
-//TODO: remove inline function?
+// TODO: remove inline function?
 #include <Xpetra_EpetraCrsMatrix_fwd.hpp>
 #include <Xpetra_CrsMatrixWrap_fwd.hpp>
 
@@ -100,1099 +100,1076 @@ class Epetra_Vector;
 
 #include <MueLu_UtilitiesBase.hpp>
 
-
 namespace MueLu {
 
 #ifdef HAVE_MUELU_EPETRA
-  //defined after Utilities class
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::CrsMatrixWrap<SC,LO,GO,NO> >
-  Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap(RCP<Epetra_CrsMatrix> &epAB);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::Matrix<SC, LO, GO, NO> >
-  EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
-  EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V);
+// defined after Utilities class
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::CrsMatrixWrap<SC, LO, GO, NO> >
+Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap(RCP<Epetra_CrsMatrix>& epAB);
+
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::Matrix<SC, LO, GO, NO> >
+EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A);
+
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
+EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V);
 #endif
 
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::Matrix<SC, LO, GO, NO> >
-  TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& Atpetra);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::Matrix<SC, LO, GO, NO> >
-  TpetraFECrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::FECrsMatrix<SC, LO, GO, NO> >& Atpetra);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
-  TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& Vtpetra);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
-  TpetraFEMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::FEMultiVector<SC, LO, GO, NO> >& Vtpetra);
-
-  template<typename SC,typename LO,typename GO,typename NO>
-  void leftRghtDofScalingWithinNode(const Xpetra::Matrix<SC,LO,GO,NO> & Atpetra, size_t blkSize, size_t nSweeps, Teuchos::ArrayRCP<SC> & rowScaling, Teuchos::ArrayRCP<SC> & colScaling);
-
-  /*!
-    @class Utilities
-    @brief MueLu utility class.
-
-    This class provides a number of static helper methods. Some are temporary and will eventually
-    go away, while others should be moved to Xpetra.
-    */
-  template <class Scalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class Utilities : public UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-#undef MUELU_UTILITIES_SHORT
-    #include "MueLu_UseShortNames.hpp"
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::Matrix<SC, LO, GO, NO> >
+TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> >& Atpetra);
 
-  public:
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::Matrix<SC, LO, GO, NO> >
+TpetraFECrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::FECrsMatrix<SC, LO, GO, NO> >& Atpetra);
 
-#ifdef HAVE_MUELU_EPETRA
-    //! Helper utility to pull out the underlying Epetra objects from an Xpetra object
-    // @{
-    static RCP<const Epetra_MultiVector>                    MV2EpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > const vec);
-    static RCP<      Epetra_MultiVector>                    MV2NonConstEpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > vec);
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
+TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<SC, LO, GO, NO> >& Vtpetra);
 
-    static const Epetra_MultiVector&                        MV2EpetraMV(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec);
-    static       Epetra_MultiVector&                        MV2NonConstEpetraMV(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec);
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::MultiVector<SC, LO, GO, NO> >
+TpetraFEMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::FEMultiVector<SC, LO, GO, NO> >& Vtpetra);
 
-    static RCP<const Epetra_CrsMatrix>                      Op2EpetraCrs(RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
-    static RCP<      Epetra_CrsMatrix>                      Op2NonConstEpetraCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
+template <typename SC, typename LO, typename GO, typename NO>
+void leftRghtDofScalingWithinNode(const Xpetra::Matrix<SC, LO, GO, NO>& Atpetra, size_t blkSize, size_t nSweeps, Teuchos::ArrayRCP<SC>& rowScaling, Teuchos::ArrayRCP<SC>& colScaling);
 
-    static const Epetra_CrsMatrix&                          Op2EpetraCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
-    static       Epetra_CrsMatrix&                          Op2NonConstEpetraCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
+/*!
+  @class Utilities
+  @brief MueLu utility class.
 
-    static const Epetra_Map&                                Map2EpetraMap(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& map);
-    // @}
-#endif
+  This class provides a number of static helper methods. Some are temporary and will eventually
+  go away, while others should be moved to Xpetra.
+  */
+template <class Scalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class Utilities : public UtilitiesBase<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+#undef MUELU_UTILITIES_SHORT
+#include "MueLu_UseShortNames.hpp"
 
-    //! Helper utility to pull out the underlying Tpetra objects from an Xpetra object
-    static RCP<const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2TpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > const vec);
-    static RCP<      Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2NonConstTpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > vec);
-    static RCP<      Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2NonConstTpetraMV2(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec);
+ public:
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
 
-    static const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&      MV2TpetraMV(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec);
-    static       Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&      MV2NonConstTpetraMV(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec);
+#ifdef HAVE_MUELU_EPETRA
+  //! Helper utility to pull out the underlying Epetra objects from an Xpetra object
+  // @{
+  static RCP<const Epetra_MultiVector> MV2EpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > const vec);
+  static RCP<Epetra_MultiVector> MV2NonConstEpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > vec);
 
-    static RCP<const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraCrs(RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
-    static RCP<      Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
+  static const Epetra_MultiVector& MV2EpetraMV(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec);
+  static Epetra_MultiVector& MV2NonConstEpetraMV(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec);
 
-    static const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2TpetraCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
-    static       Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2NonConstTpetraCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
+  static RCP<const Epetra_CrsMatrix> Op2EpetraCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
+  static RCP<Epetra_CrsMatrix> Op2NonConstEpetraCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
 
-    static RCP<const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraBlockCrs(RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
-   static RCP<      Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraBlockCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
+  static const Epetra_CrsMatrix& Op2EpetraCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
+  static Epetra_CrsMatrix& Op2NonConstEpetraCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
 
-   static const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2TpetraBlockCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
-    static       Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2NonConstTpetraBlockCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op);
+  static const Epetra_Map& Map2EpetraMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& map);
+  // @}
+#endif
 
+  //! Helper utility to pull out the underlying Tpetra objects from an Xpetra object
+  static RCP<const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2TpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > const vec);
+  static RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2NonConstTpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > vec);
+  static RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2NonConstTpetraMV2(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec);
 
+  static const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& MV2TpetraMV(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec);
+  static Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& MV2NonConstTpetraMV(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec);
 
-    static RCP<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraRow(RCP<const Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
-    static RCP<      Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraRow(RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op);
+  static RCP<const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
+  static RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
 
+  static const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2TpetraCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
+  static Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2NonConstTpetraCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
 
-    static const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >        Map2TpetraMap(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& map);
+  static RCP<const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraBlockCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
+  static RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraBlockCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
 
-    static void MyOldScaleMatrix(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse = true,
-                                 bool doFillComplete = true, bool doOptimizeStorage = true);
+  static const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2TpetraBlockCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
+  static Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2NonConstTpetraBlockCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op);
 
-    static void MyOldScaleMatrix_Epetra(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
-                                        bool doFillComplete, bool doOptimizeStorage);
-    static void MyOldScaleMatrix_Tpetra(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
-                                        bool doFillComplete, bool doOptimizeStorage);
+  static RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraRow(RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
+  static RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraRow(RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op);
 
-    static RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Transpose(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, bool optimizeTranspose = false,const std::string & label = std::string(),const Teuchos::RCP<Teuchos::ParameterList> &params=Teuchos::null);
+  static const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > Map2TpetraMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& map);
 
-    static RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> > X);
+  static void MyOldScaleMatrix(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse = true,
+                               bool doFillComplete = true, bool doOptimizeStorage = true);
 
-    static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > ExtractCoordinatesFromParameterList(ParameterList& paramList);
+  static void MyOldScaleMatrix_Epetra(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
+                                      bool doFillComplete, bool doOptimizeStorage);
+  static void MyOldScaleMatrix_Tpetra(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
+                                      bool doFillComplete, bool doOptimizeStorage);
 
-  }; // class Utilities
+  static RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Transpose(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, bool optimizeTranspose = false, const std::string& label = std::string(), const Teuchos::RCP<Teuchos::ParameterList>& params = Teuchos::null);
 
-  ///////////////////////////////////////////
+  static RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > X);
 
-#ifdef HAVE_MUELU_EPETRA
-  /*!
-    @class Utilities
-    @brief MueLu utility class (specialization SC=double and LO=GO=int).
+  static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > ExtractCoordinatesFromParameterList(ParameterList& paramList);
 
-    This class provides a number of static helper methods. Some are temporary and will eventually
-    go away, while others should be moved to Xpetra.
+};  // class Utilities
 
-  Note: this is the implementation for Epetra. Tpetra throws if TPETRA_INST_INT_INT is disabled!
-  */
-  template <>
-  class Utilities<double,int,int,Xpetra::EpetraNode> : public UtilitiesBase<double,int,int,Xpetra::EpetraNode> {
-  public:
-    typedef double              Scalar;
-    typedef int                 LocalOrdinal;
-    typedef int                 GlobalOrdinal;
-    typedef Xpetra::EpetraNode  Node;
-    typedef Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
+///////////////////////////////////////////
+
+#ifdef HAVE_MUELU_EPETRA
+/*!
+  @class Utilities
+  @brief MueLu utility class (specialization SC=double and LO=GO=int).
+
+  This class provides a number of static helper methods. Some are temporary and will eventually
+  go away, while others should be moved to Xpetra.
+
+Note: this is the implementation for Epetra. Tpetra throws if TPETRA_INST_INT_INT is disabled!
+*/
+template <>
+class Utilities<double, int, int, Xpetra::EpetraNode> : public UtilitiesBase<double, int, int, Xpetra::EpetraNode> {
+ public:
+  typedef double Scalar;
+  typedef int LocalOrdinal;
+  typedef int GlobalOrdinal;
+  typedef Xpetra::EpetraNode Node;
+  typedef Teuchos::ScalarTraits<Scalar>::magnitudeType Magnitude;
 #undef MUELU_UTILITIES_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  private:
-    using EpetraMap = Xpetra::EpetraMapT<GlobalOrdinal,Node>;
-    using EpetraMultiVector = Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node>;
-    // using EpetraCrsMatrix = Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>;
-  public:
-
-    //! Helper utility to pull out the underlying Epetra objects from an Xpetra object
-    // @{
-    static RCP<const Epetra_MultiVector>                    MV2EpetraMV(RCP<MultiVector> const vec) {
-      RCP<const EpetraMultiVector > tmpVec = rcp_dynamic_cast<EpetraMultiVector>(vec);
-      if (tmpVec == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
-      return tmpVec->getEpetra_MultiVector();
-    }
-    static RCP<      Epetra_MultiVector>                    MV2NonConstEpetraMV(RCP<MultiVector> vec) {
-      RCP<const EpetraMultiVector> tmpVec = rcp_dynamic_cast<EpetraMultiVector>(vec);
-      if (tmpVec == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
-      return tmpVec->getEpetra_MultiVector();
-    }
+ private:
+  using EpetraMap         = Xpetra::EpetraMapT<GlobalOrdinal, Node>;
+  using EpetraMultiVector = Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>;
+  // using EpetraCrsMatrix = Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>;
+ public:
+  //! Helper utility to pull out the underlying Epetra objects from an Xpetra object
+  // @{
+  static RCP<const Epetra_MultiVector> MV2EpetraMV(RCP<MultiVector> const vec) {
+    RCP<const EpetraMultiVector> tmpVec = rcp_dynamic_cast<EpetraMultiVector>(vec);
+    if (tmpVec == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
+    return tmpVec->getEpetra_MultiVector();
+  }
+  static RCP<Epetra_MultiVector> MV2NonConstEpetraMV(RCP<MultiVector> vec) {
+    RCP<const EpetraMultiVector> tmpVec = rcp_dynamic_cast<EpetraMultiVector>(vec);
+    if (tmpVec == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
+    return tmpVec->getEpetra_MultiVector();
+  }
 
-    static const Epetra_MultiVector&                        MV2EpetraMV(const MultiVector& vec) {
-      const EpetraMultiVector& tmpVec = dynamic_cast<const EpetraMultiVector&>(vec);
-      return *(tmpVec.getEpetra_MultiVector());
-    }
-    static       Epetra_MultiVector&                        MV2NonConstEpetraMV(MultiVector& vec) {
-      const EpetraMultiVector& tmpVec = dynamic_cast<const EpetraMultiVector&>(vec);
-      return *(tmpVec.getEpetra_MultiVector());
-    }
+  static const Epetra_MultiVector& MV2EpetraMV(const MultiVector& vec) {
+    const EpetraMultiVector& tmpVec = dynamic_cast<const EpetraMultiVector&>(vec);
+    return *(tmpVec.getEpetra_MultiVector());
+  }
+  static Epetra_MultiVector& MV2NonConstEpetraMV(MultiVector& vec) {
+    const EpetraMultiVector& tmpVec = dynamic_cast<const EpetraMultiVector&>(vec);
+    return *(tmpVec.getEpetra_MultiVector());
+  }
 
-    static RCP<const Epetra_CrsMatrix>                      Op2EpetraCrs(RCP<const Matrix> Op) {
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const EpetraCrsMatrix>& tmp_ECrsMtx = rcp_dynamic_cast<const EpetraCrsMatrix>(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-      return tmp_ECrsMtx->getEpetra_CrsMatrix();
-    }
-    static RCP<      Epetra_CrsMatrix>                      Op2NonConstEpetraCrs(RCP<Matrix> Op) {
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const EpetraCrsMatrix> &tmp_ECrsMtx = rcp_dynamic_cast<const EpetraCrsMatrix>(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-      return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst();
-    }
+  static RCP<const Epetra_CrsMatrix> Op2EpetraCrs(RCP<const Matrix> Op) {
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const EpetraCrsMatrix>& tmp_ECrsMtx = rcp_dynamic_cast<const EpetraCrsMatrix>(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
+    return tmp_ECrsMtx->getEpetra_CrsMatrix();
+  }
+  static RCP<Epetra_CrsMatrix> Op2NonConstEpetraCrs(RCP<Matrix> Op) {
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const EpetraCrsMatrix>& tmp_ECrsMtx = rcp_dynamic_cast<const EpetraCrsMatrix>(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
+    return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst();
+  }
 
-    static const Epetra_CrsMatrix&                          Op2EpetraCrs(const Matrix& Op) {
+  static const Epetra_CrsMatrix& Op2EpetraCrs(const Matrix& Op) {
+    try {
+      const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
       try {
-        const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
-        try {
-          const EpetraCrsMatrix& tmp_ECrsMtx = dynamic_cast<const EpetraCrsMatrix&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getEpetra_CrsMatrix();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-        }
+        const EpetraCrsMatrix& tmp_ECrsMtx = dynamic_cast<const EpetraCrsMatrix&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getEpetra_CrsMatrix();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
       }
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
-    static       Epetra_CrsMatrix&                          Op2NonConstEpetraCrs(Matrix& Op) {
+  }
+  static Epetra_CrsMatrix& Op2NonConstEpetraCrs(Matrix& Op) {
+    try {
+      CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
       try {
-        CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
-        try {
-          EpetraCrsMatrix& tmp_ECrsMtx = dynamic_cast<EpetraCrsMatrix&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getEpetra_CrsMatrixNonConst();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-        }
+        EpetraCrsMatrix& tmp_ECrsMtx = dynamic_cast<EpetraCrsMatrix&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getEpetra_CrsMatrixNonConst();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
       }
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
+  }
 
-    static const Epetra_Map&                                Map2EpetraMap(const Map& map) {
-      RCP<const EpetraMap> xeMap = rcp_dynamic_cast<const EpetraMap>(rcpFromRef(map));
-      if (xeMap == Teuchos::null)
-        throw Exceptions::BadCast("Utilities::Map2EpetraMap : Cast from Xpetra::Map to Xpetra::EpetraMap failed");
-      return xeMap->getEpetra_Map();
-    }
-    // @}
+  static const Epetra_Map& Map2EpetraMap(const Map& map) {
+    RCP<const EpetraMap> xeMap = rcp_dynamic_cast<const EpetraMap>(rcpFromRef(map));
+    if (xeMap == Teuchos::null)
+      throw Exceptions::BadCast("Utilities::Map2EpetraMap : Cast from Xpetra::Map to Xpetra::EpetraMap failed");
+    return xeMap->getEpetra_Map();
+  }
+  // @}
 
-    //! Helper utility to pull out the underlying Tpetra objects from an Xpetra object
-    static RCP<const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2TpetraMV(RCP<MultiVector> const vec)   {
+  //! Helper utility to pull out the underlying Tpetra objects from an Xpetra object
+  static RCP<const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2TpetraMV(RCP<MultiVector> const vec) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("MV2TpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("MV2TpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(vec);
-      if (tmpVec == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
-      return tmpVec->getTpetra_MultiVector();
+    RCP<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(vec);
+    if (tmpVec == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
+    return tmpVec->getTpetra_MultiVector();
 #endif
-    }
-    static RCP<      Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2NonConstTpetraMV(RCP<MultiVector> vec) {
+  }
+  static RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2NonConstTpetraMV(RCP<MultiVector> vec) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("MV2NonConstTpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("MV2NonConstTpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(vec);
-      if (tmpVec == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
-      return tmpVec->getTpetra_MultiVector();
+    RCP<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(vec);
+    if (tmpVec == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
+    return tmpVec->getTpetra_MultiVector();
 #endif
-
-    }
-    static RCP<      Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > MV2NonConstTpetraMV2(MultiVector& vec)    {
+  }
+  static RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > MV2NonConstTpetraMV2(MultiVector& vec) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("MV2NonConstTpetraMV2: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("MV2NonConstTpetraMV2: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-      return tmpVec.getTpetra_MultiVector();
+    const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+    return tmpVec.getTpetra_MultiVector();
 #endif
-    }
+  }
 
-    static const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&      MV2TpetraMV(const MultiVector& vec)   {
+  static const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& MV2TpetraMV(const MultiVector& vec) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("MV2TpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("MV2TpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-      return *(tmpVec.getTpetra_MultiVector());
+    const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+    return *(tmpVec.getTpetra_MultiVector());
 #endif
-    }
-    static       Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&      MV2NonConstTpetraMV(MultiVector& vec) {
+  }
+  static Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& MV2NonConstTpetraMV(MultiVector& vec) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("MV2NonConstTpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("MV2NonConstTpetraMV: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-      return *(tmpVec.getTpetra_MultiVector());
+    const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+    return *(tmpVec.getTpetra_MultiVector());
 #endif
-    }
+  }
 
-    static RCP<const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraCrs(RCP<const Matrix> Op)  {
+  static RCP<const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraCrs(RCP<const Matrix> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2TpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2TpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      // Get the underlying Tpetra Mtx
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-      return tmp_ECrsMtx->getTpetra_CrsMatrix();
+    // Get the underlying Tpetra Mtx
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
+    return tmp_ECrsMtx->getTpetra_CrsMatrix();
 #endif
-    }
-    static RCP<      Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraCrs(RCP<Matrix> Op){
+  }
+  static RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraCrs(RCP<Matrix> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-      return tmp_ECrsMtx->getTpetra_CrsMatrixNonConst();
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
+    return tmp_ECrsMtx->getTpetra_CrsMatrixNonConst();
 #endif
-    };
+  };
 
-    static const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2TpetraCrs(const Matrix& Op)   {
+  static const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2TpetraCrs(const Matrix& Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2TpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2TpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
+    try {
+      const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
       try {
-        const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
-        try {
-          const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getTpetra_CrsMatrix();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-        }
+        const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getTpetra_CrsMatrix();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
       }
-#endif
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
-    static       Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2NonConstTpetraCrs(Matrix& Op) {
+#endif
+  }
+  static Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2NonConstTpetraCrs(Matrix& Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
+    try {
+      CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
       try {
-        CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
-        try {
-          Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getTpetra_CrsMatrixNonConst();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-        }
+        Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getTpetra_CrsMatrixNonConst();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
       }
-#endif
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
+#endif
+  }
 
-
-    static RCP<const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraBlockCrs(RCP<const Matrix> Op)  {
+  static RCP<const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraBlockCrs(RCP<const Matrix> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2TpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2TpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      // Get the underlying Tpetra Mtx
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-      return tmp_ECrsMtx->getTpetra_BlockCrsMatrix();
+    // Get the underlying Tpetra Mtx
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
+    return tmp_ECrsMtx->getTpetra_BlockCrsMatrix();
 #endif
-    }
+  }
 
-    static RCP<      Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraBlockCrs(RCP<Matrix> Op){
+  static RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraBlockCrs(RCP<Matrix> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2NonConstTpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2NonConstTpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-      const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-      if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-      return tmp_ECrsMtx->getTpetra_BlockCrsMatrixNonConst();
+    RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(Op);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+    const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsOp->getCrsMatrix());
+    if (tmp_ECrsMtx == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
+    return tmp_ECrsMtx->getTpetra_BlockCrsMatrixNonConst();
 #endif
-    };
+  };
 
-    static const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2TpetraBlockCrs(const Matrix& Op)   {
+  static const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2TpetraBlockCrs(const Matrix& Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2TpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2TpetraBlockCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
+    try {
+      const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
       try {
-        const CrsMatrixWrap& crsOp = dynamic_cast<const CrsMatrixWrap&>(Op);
-        try {
-          const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getTpetra_BlockCrsMatrix();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-        }
+        const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getTpetra_BlockCrsMatrix();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
       }
-#endif
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
-    static       Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Op2NonConstTpetraBlockCrs(Matrix& Op) {
+#endif
+  }
+  static Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op2NonConstTpetraBlockCrs(Matrix& Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2NonConstTpetraCrs: Tpetra has not been compiled with support for LO=GO=int.");
 #else
+    try {
+      CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
       try {
-        CrsMatrixWrap& crsOp = dynamic_cast<CrsMatrixWrap&>(Op);
-        try {
-          Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-          return *tmp_ECrsMtx.getTpetra_BlockCrsMatrixNonConst();
-        } catch (std::bad_cast&) {
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-        }
+        Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+        return *tmp_ECrsMtx.getTpetra_BlockCrsMatrixNonConst();
       } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
       }
-#endif
+    } catch (std::bad_cast&) {
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
     }
+#endif
+  }
 
-
-    static RCP<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2TpetraRow(RCP<const Operator> Op)   {
+  static RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2TpetraRow(RCP<const Operator> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2TpetraRow: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2TpetraRow: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<const Matrix> mat = rcp_dynamic_cast<const Matrix>(Op);
-      if (!mat.is_null()) {
-        RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(mat);
-        if (crsOp == Teuchos::null)
-          throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-
-        RCP<const CrsMatrix> crsMat = crsOp->getCrsMatrix();
-        const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-        RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_BlockCrs;
-        if(!tmp_Crs.is_null()) {
-          return tmp_Crs->getTpetra_CrsMatrixNonConst();
-        }
-        else {
-          tmp_BlockCrs= rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-          if (tmp_BlockCrs.is_null())
-            throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
-          return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
-        }
+    RCP<const Matrix> mat = rcp_dynamic_cast<const Matrix>(Op);
+    if (!mat.is_null()) {
+      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(mat);
+      if (crsOp == Teuchos::null)
+        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+
+      RCP<const CrsMatrix> crsMat                                                                  = crsOp->getCrsMatrix();
+      const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsMat);
+      RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmp_BlockCrs;
+      if (!tmp_Crs.is_null()) {
+        return tmp_Crs->getTpetra_CrsMatrixNonConst();
       } else {
-        RCP<const TpetraOperator> tpOp = rcp_dynamic_cast<const TpetraOperator>(Op, true);
-        RCP<const Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>> tOp = tpOp->getOperatorConst();
-        RCP<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tRow = rcp_dynamic_cast<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tOp, true);
-        return tRow;
+        tmp_BlockCrs = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsMat);
+        if (tmp_BlockCrs.is_null())
+          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
+        return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
       }
-#endif
+    } else {
+      RCP<const TpetraOperator> tpOp                                                = rcp_dynamic_cast<const TpetraOperator>(Op, true);
+      RCP<const Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tOp   = tpOp->getOperatorConst();
+      RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tRow = rcp_dynamic_cast<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(tOp, true);
+      return tRow;
     }
+#endif
+  }
 
-    static RCP<      Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Op2NonConstTpetraRow(RCP<Operator> Op) {
+  static RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op2NonConstTpetraRow(RCP<Operator> Op) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Op2NonConstTpetraRow: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Op2NonConstTpetraRow: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      RCP<Matrix> mat = rcp_dynamic_cast<Matrix>(Op);
-      if (!mat.is_null()) {
-        RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(mat);
-        if (crsOp == Teuchos::null)
-          throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-
-        RCP<const CrsMatrix> crsMat = crsOp->getCrsMatrix();
-        const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-        RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_BlockCrs;
-        if(!tmp_Crs.is_null()) {
-          return tmp_Crs->getTpetra_CrsMatrixNonConst();
-        }
-        else {
-          tmp_BlockCrs= rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-          if (tmp_BlockCrs.is_null())
-            throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
-          return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
-        }
+    RCP<Matrix> mat = rcp_dynamic_cast<Matrix>(Op);
+    if (!mat.is_null()) {
+      RCP<const CrsMatrixWrap> crsOp = rcp_dynamic_cast<const CrsMatrixWrap>(mat);
+      if (crsOp == Teuchos::null)
+        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+
+      RCP<const CrsMatrix> crsMat                                                                  = crsOp->getCrsMatrix();
+      const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsMat);
+      RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tmp_BlockCrs;
+      if (!tmp_Crs.is_null()) {
+        return tmp_Crs->getTpetra_CrsMatrixNonConst();
       } else {
-        RCP<TpetraOperator> tpOp = rcp_dynamic_cast<TpetraOperator>(Op, true);
-        RCP<Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>> tOp = tpOp->getOperator();
-        RCP<Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tRow = rcp_dynamic_cast<Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tOp, true);
-        return tRow;
+        tmp_BlockCrs = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(crsMat);
+        if (tmp_BlockCrs.is_null())
+          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
+        return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
       }
+    } else {
+      RCP<TpetraOperator> tpOp                                                = rcp_dynamic_cast<TpetraOperator>(Op, true);
+      RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tOp   = tpOp->getOperator();
+      RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tRow = rcp_dynamic_cast<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(tOp, true);
+      return tRow;
+    }
 #endif
-    };
-
+  };
 
-    static const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >          Map2TpetraMap(const Map& map) {
+  static const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > Map2TpetraMap(const Map& map) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Map2TpetraMap: Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Map2TpetraMap: Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      const RCP<const Xpetra::TpetraMap<LocalOrdinal,GlobalOrdinal,Node>>& tmp_TMap = rcp_dynamic_cast<const Xpetra::TpetraMap<LocalOrdinal,GlobalOrdinal,Node> >(rcpFromRef(map));
-      if (tmp_TMap == Teuchos::null)
-        throw Exceptions::BadCast("Utilities::Map2TpetraMap : Cast from Xpetra::Map to Xpetra::TpetraMap failed");
-      return tmp_TMap->getTpetra_Map();
+    const RCP<const Xpetra::TpetraMap<LocalOrdinal, GlobalOrdinal, Node> >& tmp_TMap = rcp_dynamic_cast<const Xpetra::TpetraMap<LocalOrdinal, GlobalOrdinal, Node> >(rcpFromRef(map));
+    if (tmp_TMap == Teuchos::null)
+      throw Exceptions::BadCast("Utilities::Map2TpetraMap : Cast from Xpetra::Map to Xpetra::TpetraMap failed");
+    return tmp_TMap->getTpetra_Map();
 #endif
-    };
-
-
-    static void MyOldScaleMatrix(Matrix& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse = true,
-                                 bool doFillComplete = true, bool doOptimizeStorage = true) {
-      Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-      Teuchos::ArrayRCP<Scalar> sv(scalingVector.size());
-      if (doInverse) {
-        for (int i = 0; i < scalingVector.size(); ++i)
-          sv[i] = one / scalingVector[i];
-      } else {
-        for (int i = 0; i < scalingVector.size(); ++i)
-          sv[i] = scalingVector[i];
-      }
+  };
+
+  static void MyOldScaleMatrix(Matrix& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse = true,
+                               bool doFillComplete = true, bool doOptimizeStorage = true) {
+    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+    Teuchos::ArrayRCP<Scalar> sv(scalingVector.size());
+    if (doInverse) {
+      for (int i = 0; i < scalingVector.size(); ++i)
+        sv[i] = one / scalingVector[i];
+    } else {
+      for (int i = 0; i < scalingVector.size(); ++i)
+        sv[i] = scalingVector[i];
+    }
 
-      switch (Op.getRowMap()->lib()) {
-        case Xpetra::UseTpetra:
-          MyOldScaleMatrix_Tpetra(Op, sv, doFillComplete, doOptimizeStorage);
-          break;
+    switch (Op.getRowMap()->lib()) {
+      case Xpetra::UseTpetra:
+        MyOldScaleMatrix_Tpetra(Op, sv, doFillComplete, doOptimizeStorage);
+        break;
 
-        case Xpetra::UseEpetra:
-          MyOldScaleMatrix_Epetra(Op, sv, doFillComplete, doOptimizeStorage);
-          break;
+      case Xpetra::UseEpetra:
+        MyOldScaleMatrix_Epetra(Op, sv, doFillComplete, doOptimizeStorage);
+        break;
 
-        default:
-          throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be scaled.");
-      }
+      default:
+        throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be scaled.");
     }
+  }
 
-    // TODO This is the <double,int,int> specialization
-    static void MyOldScaleMatrix_Tpetra(Matrix& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
-                                        bool doFillComplete, bool doOptimizeStorage) {
+  // TODO This is the <double,int,int> specialization
+  static void MyOldScaleMatrix_Tpetra(Matrix& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
+                                      bool doFillComplete, bool doOptimizeStorage) {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-      throw Exceptions::RuntimeError("Matrix scaling is not possible because Tpetra has not been compiled with support for LO=GO=int.");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+    throw Exceptions::RuntimeError("Matrix scaling is not possible because Tpetra has not been compiled with support for LO=GO=int.");
 #else
-      try {
-        Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpOp = Op2NonConstTpetraCrs(Op);
-
-        const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rowMap    = tpOp.getRowMap();
-        const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domainMap = tpOp.getDomainMap();
-        const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rangeMap  = tpOp.getRangeMap();
-
-        size_t maxRowSize = tpOp.getLocalMaxNumRowEntries();
-        if (maxRowSize == Teuchos::as<size_t>(-1)) // hasn't been determined yet
-          maxRowSize = 20;
-
-        std::vector<Scalar> scaledVals(maxRowSize);
-        if (tpOp.isFillComplete())
-          tpOp.resumeFill();
-
-        if (Op.isLocallyIndexed() == true) {
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
-          for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-            tpOp.getLocalRowView(i, cols, vals);
-            size_t nnz = tpOp.getNumEntriesInLocalRow(i);
-            if (nnz > maxRowSize) {
-              maxRowSize = nnz;
-              scaledVals.resize(maxRowSize);
-            }
-            for (size_t j = 0; j < nnz; ++j)
-              scaledVals[j] = vals[j]*scalingVector[i];
-
-            if (nnz > 0) {
-	      Teuchos::ArrayView<const LocalOrdinal> cols_view(cols.data(), nnz);
-              Teuchos::ArrayView<const Scalar> valview(&scaledVals[0], nnz);
-              tpOp.replaceLocalValues(i, cols_view, valview);
-            }
-          } //for (size_t i=0; ...
-
-        } else {
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
-
-          for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-            GlobalOrdinal gid = rowMap->getGlobalElement(i);
-            tpOp.getGlobalRowView(gid, cols, vals);
-            size_t nnz = tpOp.getNumEntriesInGlobalRow(gid);
-            if (nnz > maxRowSize) {
-              maxRowSize = nnz;
-              scaledVals.resize(maxRowSize);
-            }
-            // FIXME FIXME FIXME FIXME FIXME FIXME
-            for (size_t j = 0; j < nnz; ++j)
-              scaledVals[j] = vals[j]*scalingVector[i]; //FIXME i or gid?
-
-            if (nnz > 0) {
-	      Teuchos::ArrayView<const LocalOrdinal> cols_view(cols.data(), nnz);
-              Teuchos::ArrayView<const Scalar> valview(&scaledVals[0], nnz);
-              tpOp.replaceGlobalValues(gid, cols_view, valview);
-            }
-          } //for (size_t i=0; ...
-        }
+    try {
+      Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpOp = Op2NonConstTpetraCrs(Op);
+
+      const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rowMap    = tpOp.getRowMap();
+      const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainMap = tpOp.getDomainMap();
+      const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangeMap  = tpOp.getRangeMap();
+
+      size_t maxRowSize = tpOp.getLocalMaxNumRowEntries();
+      if (maxRowSize == Teuchos::as<size_t>(-1))  // hasn't been determined yet
+        maxRowSize = 20;
+
+      std::vector<Scalar> scaledVals(maxRowSize);
+      if (tpOp.isFillComplete())
+        tpOp.resumeFill();
+
+      if (Op.isLocallyIndexed() == true) {
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
+        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+          tpOp.getLocalRowView(i, cols, vals);
+          size_t nnz = tpOp.getNumEntriesInLocalRow(i);
+          if (nnz > maxRowSize) {
+            maxRowSize = nnz;
+            scaledVals.resize(maxRowSize);
+          }
+          for (size_t j = 0; j < nnz; ++j)
+            scaledVals[j] = vals[j] * scalingVector[i];
 
-        if (doFillComplete) {
-          if (domainMap == Teuchos::null || rangeMap == Teuchos::null)
-            throw Exceptions::RuntimeError("In Utilities::Scaling: cannot fillComplete because the domain and/or range map hasn't been defined");
+          if (nnz > 0) {
+            Teuchos::ArrayView<const LocalOrdinal> cols_view(cols.data(), nnz);
+            Teuchos::ArrayView<const Scalar> valview(&scaledVals[0], nnz);
+            tpOp.replaceLocalValues(i, cols_view, valview);
+          }
+        }  // for (size_t i=0; ...
 
-          RCP<Teuchos::ParameterList> params = rcp(new Teuchos::ParameterList());
-          params->set("Optimize Storage",    doOptimizeStorage);
-          params->set("No Nonlocal Changes", true);
-          Op.fillComplete(Op.getDomainMap(), Op.getRangeMap(), params);
-        }
-      } catch(...) {
-        throw Exceptions::RuntimeError("Only Tpetra::CrsMatrix types can be scaled (Err.1)");
+      } else {
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
+
+        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+          GlobalOrdinal gid = rowMap->getGlobalElement(i);
+          tpOp.getGlobalRowView(gid, cols, vals);
+          size_t nnz = tpOp.getNumEntriesInGlobalRow(gid);
+          if (nnz > maxRowSize) {
+            maxRowSize = nnz;
+            scaledVals.resize(maxRowSize);
+          }
+          // FIXME FIXME FIXME FIXME FIXME FIXME
+          for (size_t j = 0; j < nnz; ++j)
+            scaledVals[j] = vals[j] * scalingVector[i];  // FIXME i or gid?
+
+          if (nnz > 0) {
+            Teuchos::ArrayView<const LocalOrdinal> cols_view(cols.data(), nnz);
+            Teuchos::ArrayView<const Scalar> valview(&scaledVals[0], nnz);
+            tpOp.replaceGlobalValues(gid, cols_view, valview);
+          }
+        }  // for (size_t i=0; ...
       }
-#endif
+
+      if (doFillComplete) {
+        if (domainMap == Teuchos::null || rangeMap == Teuchos::null)
+          throw Exceptions::RuntimeError("In Utilities::Scaling: cannot fillComplete because the domain and/or range map hasn't been defined");
+
+        RCP<Teuchos::ParameterList> params = rcp(new Teuchos::ParameterList());
+        params->set("Optimize Storage", doOptimizeStorage);
+        params->set("No Nonlocal Changes", true);
+        Op.fillComplete(Op.getDomainMap(), Op.getRangeMap(), params);
+      }
+    } catch (...) {
+      throw Exceptions::RuntimeError("Only Tpetra::CrsMatrix types can be scaled (Err.1)");
     }
+#endif
+  }
 
-    static void MyOldScaleMatrix_Epetra (Matrix& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector, bool /* doFillComplete */, bool /* doOptimizeStorage */) {
+  static void MyOldScaleMatrix_Epetra(Matrix& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector, bool /* doFillComplete */, bool /* doOptimizeStorage */) {
 #ifdef HAVE_MUELU_EPETRA
-      try {
-        //const Epetra_CrsMatrix& epOp = Utilities<double,int,int>::Op2NonConstEpetraCrs(Op);
-        const Epetra_CrsMatrix& epOp = Op2NonConstEpetraCrs(Op);
-
-        Epetra_Map const &rowMap = epOp.RowMap();
-        int nnz;
-        double *vals;
-        int *cols;
-
-        for (int i = 0; i < rowMap.NumMyElements(); ++i) {
-          epOp.ExtractMyRowView(i, nnz, vals, cols);
-          for (int j = 0; j < nnz; ++j)
-            vals[j] *= scalingVector[i];
-        }
-
-      } catch (...){
-        throw Exceptions::RuntimeError("Only Epetra_CrsMatrix types can be scaled");
+    try {
+      // const Epetra_CrsMatrix& epOp = Utilities<double,int,int>::Op2NonConstEpetraCrs(Op);
+      const Epetra_CrsMatrix& epOp = Op2NonConstEpetraCrs(Op);
+
+      Epetra_Map const& rowMap = epOp.RowMap();
+      int nnz;
+      double* vals;
+      int* cols;
+
+      for (int i = 0; i < rowMap.NumMyElements(); ++i) {
+        epOp.ExtractMyRowView(i, nnz, vals, cols);
+        for (int j = 0; j < nnz; ++j)
+          vals[j] *= scalingVector[i];
       }
-#else
-      throw Exceptions::RuntimeError("Matrix scaling is not possible because Epetra has not been enabled.");
-#endif // HAVE_MUELU_EPETRA
+
+    } catch (...) {
+      throw Exceptions::RuntimeError("Only Epetra_CrsMatrix types can be scaled");
     }
+#else
+    throw Exceptions::RuntimeError("Matrix scaling is not possible because Epetra has not been enabled.");
+#endif  // HAVE_MUELU_EPETRA
+  }
 
-    /*! @brief Transpose a Xpetra::Matrix
+  /*! @brief Transpose a Xpetra::Matrix
 
-        Note: Currently, an error is thrown if the matrix isn't a Tpetra::CrsMatrix or Epetra_CrsMatrix.
-        In principle, however, we could allow any Epetra_RowMatrix because the Epetra transposer does.
-    */
-    static RCP<Matrix> Transpose(Matrix& Op, bool /* optimizeTranspose */ = false,const std::string & label = std::string(),const Teuchos::RCP<Teuchos::ParameterList> &params=Teuchos::null) {
-      switch (Op.getRowMap()->lib()) {
-        case Xpetra::UseTpetra: {
+      Note: Currently, an error is thrown if the matrix isn't a Tpetra::CrsMatrix or Epetra_CrsMatrix.
+      In principle, however, we could allow any Epetra_RowMatrix because the Epetra transposer does.
+  */
+  static RCP<Matrix> Transpose(Matrix& Op, bool /* optimizeTranspose */ = false, const std::string& label = std::string(), const Teuchos::RCP<Teuchos::ParameterList>& params = Teuchos::null) {
+    switch (Op.getRowMap()->lib()) {
+      case Xpetra::UseTpetra: {
 #if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
-    (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-            throw Exceptions::RuntimeError("Utilities::Transpose: Tpetra is not compiled with LO=GO=int. Add TPETRA_INST_INT_INT:BOOL=ON to your configuration!");
+     (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
+        throw Exceptions::RuntimeError("Utilities::Transpose: Tpetra is not compiled with LO=GO=int. Add TPETRA_INST_INT_INT:BOOL=ON to your configuration!");
 #else
-            using Helpers = Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-            /***************************************************************/
-            if(Helpers::isTpetraCrs(Op)) {
-              const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(Op);
-              
-              // Compute the transpose A of the Tpetra matrix tpetraOp.
-              RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
-              Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp),label);
-
-              {
-                using Teuchos::ParameterList;
-                using Teuchos::rcp;
-                RCP<ParameterList> transposeParams = params.is_null () ?
-                  rcp (new ParameterList) :
-                  rcp (new ParameterList (*params));
-                transposeParams->set ("sort", false);
-                A = transposer.createTranspose(transposeParams);
-              }
-
-              RCP<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA   = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A));
-              RCP<CrsMatrix>                                                           AAA  = rcp_implicit_cast<CrsMatrix>(AA);
-              RCP<Matrix>                                                              AAAA = rcp( new CrsMatrixWrap(AAA));
-
-              if (Op.IsView("stridedMaps"))
-                AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true/*doTranspose*/);
-
-              return AAAA;
-            }
-              /***************************************************************/
-            else if(Helpers::isTpetraBlockCrs(Op)) {
-              using BCRS = Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-              // using CRS  = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-              const BCRS & tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(Op);
-              RCP<BCRS> At;
-              {
-                Tpetra::BlockCrsMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp),label);
-                
-                using Teuchos::ParameterList;
-                using Teuchos::rcp;
-                RCP<ParameterList> transposeParams = params.is_null () ?
-                  rcp (new ParameterList) :
-                    rcp (new ParameterList (*params));
-                transposeParams->set ("sort", false);
-                At = transposer.createTranspose(transposeParams);
-              }
-
-              RCP<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA   = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(At));
-              RCP<CrsMatrix>                                                           AAA  = rcp_implicit_cast<CrsMatrix>(AA);
-              RCP<Matrix>                                                              AAAA = rcp( new CrsMatrixWrap(AAA));
-              
-              if (Op.IsView("stridedMaps"))
-                AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true/*doTranspose*/);
-
-              return AAAA;
-
-            }
-            /***************************************************************/
-            else {
-              throw Exceptions::RuntimeError("Utilities::Transpose failed, perhaps because matrix is not a Crs or BlockCrs matrix");
-            }
-#endif
+        using Helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+        /***************************************************************/
+        if (Helpers::isTpetraCrs(Op)) {
+          const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(Op);
+
+          // Compute the transpose A of the Tpetra matrix tpetraOp.
+          RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
+          Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp), label);
+
+          {
+            using Teuchos::ParameterList;
+            using Teuchos::rcp;
+            RCP<ParameterList> transposeParams = params.is_null() ? rcp(new ParameterList) : rcp(new ParameterList(*params));
+            transposeParams->set("sort", false);
+            A = transposer.createTranspose(transposeParams);
           }
-        case Xpetra::UseEpetra:
+
+          RCP<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A));
+          RCP<CrsMatrix> AAA                                                          = rcp_implicit_cast<CrsMatrix>(AA);
+          RCP<Matrix> AAAA                                                            = rcp(new CrsMatrixWrap(AAA));
+
+          if (Op.IsView("stridedMaps"))
+            AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true /*doTranspose*/);
+
+          return AAAA;
+        }
+        /***************************************************************/
+        else if (Helpers::isTpetraBlockCrs(Op)) {
+          using BCRS = Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+          // using CRS  = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+          const BCRS& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(Op);
+          RCP<BCRS> At;
           {
+            Tpetra::BlockCrsMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp), label);
+
+            using Teuchos::ParameterList;
+            using Teuchos::rcp;
+            RCP<ParameterList> transposeParams = params.is_null() ? rcp(new ParameterList) : rcp(new ParameterList(*params));
+            transposeParams->set("sort", false);
+            At = transposer.createTranspose(transposeParams);
+          }
+
+          RCP<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(At));
+          RCP<CrsMatrix> AAA                                                               = rcp_implicit_cast<CrsMatrix>(AA);
+          RCP<Matrix> AAAA                                                                 = rcp(new CrsMatrixWrap(AAA));
+
+          if (Op.IsView("stridedMaps"))
+            AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true /*doTranspose*/);
+
+          return AAAA;
+
+        }
+        /***************************************************************/
+        else {
+          throw Exceptions::RuntimeError("Utilities::Transpose failed, perhaps because matrix is not a Crs or BlockCrs matrix");
+        }
+#endif
+      }
+      case Xpetra::UseEpetra: {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
-            Teuchos::TimeMonitor tm(*Teuchos::TimeMonitor::getNewTimer("ZZ Entire Transpose"));
-            // Epetra case
-            Epetra_CrsMatrix& epetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Op);
-            EpetraExt::RowMatrix_Transpose transposer;
-            Epetra_CrsMatrix * A = dynamic_cast<Epetra_CrsMatrix*>(&transposer(epetraOp));
-            transposer.ReleaseTranspose(); // So we can keep A in Muelu...
-
-            RCP<Epetra_CrsMatrix> rcpA(A);
-            RCP<EpetraCrsMatrix> AA   = rcp(new EpetraCrsMatrix(rcpA));
-            RCP<CrsMatrix>       AAA  = rcp_implicit_cast<CrsMatrix>(AA);
-            RCP<Matrix>          AAAA = rcp( new CrsMatrixWrap(AAA));
-
-            if (Op.IsView("stridedMaps"))
-              AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true/*doTranspose*/);
-
-            return AAAA;
+        Teuchos::TimeMonitor tm(*Teuchos::TimeMonitor::getNewTimer("ZZ Entire Transpose"));
+        // Epetra case
+        Epetra_CrsMatrix& epetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Op);
+        EpetraExt::RowMatrix_Transpose transposer;
+        Epetra_CrsMatrix* A = dynamic_cast<Epetra_CrsMatrix*>(&transposer(epetraOp));
+        transposer.ReleaseTranspose();  // So we can keep A in Muelu...
+
+        RCP<Epetra_CrsMatrix> rcpA(A);
+        RCP<EpetraCrsMatrix> AA = rcp(new EpetraCrsMatrix(rcpA));
+        RCP<CrsMatrix> AAA      = rcp_implicit_cast<CrsMatrix>(AA);
+        RCP<Matrix> AAAA        = rcp(new CrsMatrixWrap(AAA));
+
+        if (Op.IsView("stridedMaps"))
+          AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true /*doTranspose*/);
+
+        return AAAA;
 #else
-            throw Exceptions::RuntimeError("Epetra (Err. 2)");
+        throw Exceptions::RuntimeError("Epetra (Err. 2)");
 #endif
-          }
-        default:
-          throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be transposed.");
       }
-
-      TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+      default:
+        throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be transposed.");
     }
 
-    static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >
-    RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> > X) {
-      RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Xscalar = rcp_dynamic_cast<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(X, true);
-      return Xscalar;
-    }
+    TEUCHOS_UNREACHABLE_RETURN(Teuchos::null);
+  }
+
+  static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >
+  RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> > X) {
+    RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Xscalar = rcp_dynamic_cast<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(X, true);
+    return Xscalar;
+  }
 
-    /*! @brief Extract coordinates from parameter list and return them in a Xpetra::MultiVector
-    */
-    static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > ExtractCoordinatesFromParameterList(ParameterList& paramList) {
-      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > coordinates = Teuchos::null;
+  /*! @brief Extract coordinates from parameter list and return them in a Xpetra::MultiVector
+   */
+  static RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > ExtractCoordinatesFromParameterList(ParameterList& paramList) {
+    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > coordinates = Teuchos::null;
 
-      // check whether coordinates are contained in parameter list
-      if(paramList.isParameter ("Coordinates") == false)
-        return coordinates;
+    // check whether coordinates are contained in parameter list
+    if (paramList.isParameter("Coordinates") == false)
+      return coordinates;
 
-  #if ( defined(EPETRA_HAVE_OMP) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)) || \
-      (!defined(EPETRA_HAVE_OMP) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))
+#if (defined(EPETRA_HAVE_OMP) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)) || \
+    (!defined(EPETRA_HAVE_OMP) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT))
 
       // define Tpetra::MultiVector type with Scalar=float only if
       // * ETI is turned off, since then the compiler will instantiate it automatically OR
       // * Tpetra is instantiated on Scalar=float
-  #if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
-      typedef Tpetra::MultiVector<float, LocalOrdinal, GlobalOrdinal, Node> tfMV;
-      RCP<tfMV> floatCoords = Teuchos::null;
-  #endif
-
-      // define Tpetra::MultiVector type with Scalar=double only if
-      // * ETI is turned off, since then the compiler will instantiate it automatically OR
-      // * Tpetra is instantiated on Scalar=double
-      typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> tdMV;
-      RCP<tdMV> doubleCoords = Teuchos::null;
-      if (paramList.isType<RCP<tdMV> >("Coordinates")) {
-        // Coordinates are stored as a double vector
-        doubleCoords = paramList.get<RCP<tdMV> >("Coordinates");
-        paramList.remove("Coordinates");
-      }
-  #if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
-      else if (paramList.isType<RCP<tfMV> >("Coordinates")) {
-        // check if coordinates are stored as a float vector
-        floatCoords = paramList.get<RCP<tfMV> >("Coordinates");
-        paramList.remove("Coordinates");
-        doubleCoords = rcp(new tdMV(floatCoords->getMap(), floatCoords->getNumVectors()));
-        deep_copy(*doubleCoords, *floatCoords);
-      }
-  #endif
-      // We have the coordinates in a Tpetra double vector
-      if(doubleCoords != Teuchos::null) {
-        coordinates = Teuchos::rcp(new Xpetra::TpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>(doubleCoords));
-        TEUCHOS_TEST_FOR_EXCEPT(doubleCoords->getNumVectors() != coordinates->getNumVectors());
-      }
-  #endif // Tpetra instantiated on GO=int and EpetraNode
-
-  #if defined(HAVE_MUELU_EPETRA)
-      RCP<Epetra_MultiVector> doubleEpCoords;
-      if (paramList.isType<RCP<Epetra_MultiVector> >("Coordinates")) {
-        doubleEpCoords = paramList.get<RCP<Epetra_MultiVector> >("Coordinates");
-        paramList.remove("Coordinates");
-        RCP<Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> > epCoordinates = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node>(doubleEpCoords));
-        coordinates = rcp_dynamic_cast<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >(epCoordinates);
-        TEUCHOS_TEST_FOR_EXCEPT(doubleEpCoords->NumVectors() != Teuchos::as<int>(coordinates->getNumVectors()));
-      }
-  #endif
+#if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
+    typedef Tpetra::MultiVector<float, LocalOrdinal, GlobalOrdinal, Node> tfMV;
+    RCP<tfMV> floatCoords = Teuchos::null;
+#endif
 
-      // check for Xpetra coordinates vector
-      if(paramList.isType<decltype(coordinates)>("Coordinates")) {
-        coordinates = paramList.get<decltype(coordinates)>("Coordinates");
-      }
+    // define Tpetra::MultiVector type with Scalar=double only if
+    // * ETI is turned off, since then the compiler will instantiate it automatically OR
+    // * Tpetra is instantiated on Scalar=double
+    typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> tdMV;
+    RCP<tdMV> doubleCoords = Teuchos::null;
+    if (paramList.isType<RCP<tdMV> >("Coordinates")) {
+      // Coordinates are stored as a double vector
+      doubleCoords = paramList.get<RCP<tdMV> >("Coordinates");
+      paramList.remove("Coordinates");
+    }
+#if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
+    else if (paramList.isType<RCP<tfMV> >("Coordinates")) {
+      // check if coordinates are stored as a float vector
+      floatCoords = paramList.get<RCP<tfMV> >("Coordinates");
+      paramList.remove("Coordinates");
+      doubleCoords = rcp(new tdMV(floatCoords->getMap(), floatCoords->getNumVectors()));
+      deep_copy(*doubleCoords, *floatCoords);
+    }
+#endif
+    // We have the coordinates in a Tpetra double vector
+    if (doubleCoords != Teuchos::null) {
+      coordinates = Teuchos::rcp(new Xpetra::TpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>(doubleCoords));
+      TEUCHOS_TEST_FOR_EXCEPT(doubleCoords->getNumVectors() != coordinates->getNumVectors());
+    }
+#endif  // Tpetra instantiated on GO=int and EpetraNode
+
+#if defined(HAVE_MUELU_EPETRA)
+    RCP<Epetra_MultiVector> doubleEpCoords;
+    if (paramList.isType<RCP<Epetra_MultiVector> >("Coordinates")) {
+      doubleEpCoords = paramList.get<RCP<Epetra_MultiVector> >("Coordinates");
+      paramList.remove("Coordinates");
+      RCP<Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node> > epCoordinates = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>(doubleEpCoords));
+      coordinates                                                         = rcp_dynamic_cast<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >(epCoordinates);
+      TEUCHOS_TEST_FOR_EXCEPT(doubleEpCoords->NumVectors() != Teuchos::as<int>(coordinates->getNumVectors()));
+    }
+#endif
 
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(coordinates));
-      return coordinates;
+    // check for Xpetra coordinates vector
+    if (paramList.isType<decltype(coordinates)>("Coordinates")) {
+      coordinates = paramList.get<decltype(coordinates)>("Coordinates");
     }
 
-  }; // class Utilities (specialization SC=double LO=GO=int)
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(coordinates));
+    return coordinates;
+  }
 
-#endif // HAVE_MUELU_EPETRA
+};  // class Utilities (specialization SC=double LO=GO=int)
 
+#endif  // HAVE_MUELU_EPETRA
 
+/*!
+\brief Extract non-serializable data from level-specific sublists and move it to a separate parameter list
 
-  /*!
-  \brief Extract non-serializable data from level-specific sublists and move it to a separate parameter list
+Look through the level-specific sublists form \c inList, extract non-serializable data and move it to \c nonSerialList.
+Everything else is copied to the \c serialList.
 
-  Look through the level-specific sublists form \c inList, extract non-serializable data and move it to \c nonSerialList.
-  Everything else is copied to the \c serialList.
+\note Data is considered "non-serializable" if it is not the same on every rank/processor.
 
-  \note Data is considered "non-serializable" if it is not the same on every rank/processor.
+Non-serializable data to be moved:
+- Operator "A"
+- Prolongator "P"
+- Restrictor "R"
+- "M"
+- "Mdiag"
+- "K"
+- Nullspace information "Nullspace"
+- Coordinate information "Coordinates"
+- "Node Comm"
+- Primal-to-dual node mapping "DualNodeID2PrimalNodeID"
+- "Primal interface DOF map"
+- "pcoarsen: element to node map
 
-  Non-serializable data to be moved:
-  - Operator "A"
-  - Prolongator "P"
-  - Restrictor "R"
-  - "M"
-  - "Mdiag"
-  - "K"
-  - Nullspace information "Nullspace"
-  - Coordinate information "Coordinates"
-  - "Node Comm"
-  - Primal-to-dual node mapping "DualNodeID2PrimalNodeID"
-  - "Primal interface DOF map"
-  - "pcoarsen: element to node map
+@param[in] inList List with all input parameters/data as provided by the user
+@param[out] serialList All serializable data from the input list
+@param[out] nonSerialList All non-serializable, i.e. rank-specific data from the input list
 
-  @param[in] inList List with all input parameters/data as provided by the user
-  @param[out] serialList All serializable data from the input list
-  @param[out] nonSerialList All non-serializable, i.e. rank-specific data from the input list
+@return This function returns the level number of the highest level for which non-serializable data was provided.
 
-  @return This function returns the level number of the highest level for which non-serializable data was provided.
+*/
+long ExtractNonSerializableData(const Teuchos::ParameterList& inList, Teuchos::ParameterList& serialList, Teuchos::ParameterList& nonSerialList);
 
-  */
-  long ExtractNonSerializableData(const Teuchos::ParameterList& inList, Teuchos::ParameterList& serialList, Teuchos::ParameterList& nonSerialList);
+/*! Tokenizes a (comma)-separated string, removing all leading and trailing whitespace
+WARNING: This routine is not threadsafe on most architectures
+*/
+void TokenizeStringAndStripWhiteSpace(const std::string& stream, std::vector<std::string>& tokenList, const char* token = ",");
 
+/*! Returns true if a parameter name is a valid Muemex custom level variable, e.g. "MultiVector myArray"
+ */
+bool IsParamMuemexVariable(const std::string& name);
 
-  /*! Tokenizes a (comma)-separated string, removing all leading and trailing whitespace
-  WARNING: This routine is not threadsafe on most architectures
-  */
-  void TokenizeStringAndStripWhiteSpace(const std::string & stream, std::vector<std::string> & tokenList, const char* token = ",");
+/*! Returns true if a parameter name is a valid user custom level variable, e.g. "MultiVector myArray"
+ */
+bool IsParamValidVariable(const std::string& name);
 
-  /*! Returns true if a parameter name is a valid Muemex custom level variable, e.g. "MultiVector myArray"
+#ifdef HAVE_MUELU_EPETRA
+/*! \fn EpetraCrs_To_XpetraMatrix
+    @brief Helper function to convert a Epetra::CrsMatrix to an Xpetra::Matrix
+    TODO move this function to an Xpetra utility file
   */
-  bool IsParamMuemexVariable(const std::string& name);
-
-  /*! Returns true if a parameter name is a valid user custom level variable, e.g. "MultiVector myArray"
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A) {
+  typedef Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node> XECrsMatrix;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrixWrap;
+
+  RCP<XCrsMatrix> Atmp = rcp(new XECrsMatrix(A));
+  return rcp(new XCrsMatrixWrap(Atmp));
+}
+
+/*! \fn EpetraMultiVector_To_XpetraMultiVector
+  @brief Helper function to convert a Epetra::MultiVector to an Xpetra::MultiVector
+  TODO move this function to an Xpetra utility file
   */
-  bool IsParamValidVariable(const std::string& name);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V) {
+  return rcp(new Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>(V));
+}
+#endif
 
-#ifdef HAVE_MUELU_EPETRA
-  /*! \fn EpetraCrs_To_XpetraMatrix
-      @brief Helper function to convert a Epetra::CrsMatrix to an Xpetra::Matrix
-      TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A) {
-    typedef Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>                      XECrsMatrix;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>       XCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>   XCrsMatrixWrap;
-
-    RCP<XCrsMatrix> Atmp = rcp(new XECrsMatrix(A));
-    return rcp(new XCrsMatrixWrap(Atmp));
-  }
+/*! \fn TpetraCrs_To_XpetraMatrix
+  @brief Helper function to convert a Tpetra::CrsMatrix to an Xpetra::Matrix
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra) {
+  typedef Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTCrsMatrix;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrixWrap;
 
-  /*! \fn EpetraMultiVector_To_XpetraMultiVector
-    @brief Helper function to convert a Epetra::MultiVector to an Xpetra::MultiVector
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V) {
-    return rcp(new Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>(V));
-  }
-#endif
+  RCP<XCrsMatrix> Atmp = rcp(new XTCrsMatrix(Atpetra));
+  return rcp(new XCrsMatrixWrap(Atmp));
+}
 
-  /*! \fn TpetraCrs_To_XpetraMatrix
-    @brief Helper function to convert a Tpetra::CrsMatrix to an Xpetra::Matrix
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra) {
-    typedef Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTCrsMatrix;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>       XCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>   XCrsMatrixWrap;
-
-    RCP<XCrsMatrix> Atmp = rcp(new XTCrsMatrix(Atpetra));
-    return rcp(new XCrsMatrixWrap(Atmp));
-  }
+/*! \fn leftRghtDofScalingWithinNode
+  @brief Helper function computes 2k left/right matrix scaling coefficients for PDE system with k x k blocks
 
-  /*! \fn leftRghtDofScalingWithinNode
-    @brief Helper function computes 2k left/right matrix scaling coefficients for PDE system with k x k blocks
+  Heuristic algorithm computes rowScaling and colScaling so that one can effectively derive matrices
+  rowScalingMatrix and colScalingMatrix such that the abs(rowsums) and abs(colsums) of
 
-    Heuristic algorithm computes rowScaling and colScaling so that one can effectively derive matrices
-    rowScalingMatrix and colScalingMatrix such that the abs(rowsums) and abs(colsums) of
+            rowScalingMatrix * Amat * colScalingMatrix
 
-              rowScalingMatrix * Amat * colScalingMatrix
+  are roughly constant. If D = diag(rowScalingMatrix), then
 
-    are roughly constant. If D = diag(rowScalingMatrix), then
+     D(i:blkSize:end) = rowScaling(i)   for i=1,..,blkSize .
 
-       D(i:blkSize:end) = rowScaling(i)   for i=1,..,blkSize .
+  diag(colScalingMatrix) is defined analogously. This function only computes rowScaling/colScaling.
+  You will need to copy them into a tpetra vector to use tpetra functions such as leftScale() and rightScale()
+  via some kind of loop such as
 
-    diag(colScalingMatrix) is defined analogously. This function only computes rowScaling/colScaling.
-    You will need to copy them into a tpetra vector to use tpetra functions such as leftScale() and rightScale()
-    via some kind of loop such as
+  rghtScaleVec = Teuchos::rcp(new Tpetra::Vector<SC,LO,GO,NO>(tpetraMat->getColMap()));
+  rghtScaleData  = rghtScaleVec->getDataNonConst(0);
+  size_t itemp = 0;
+  for (size_t i = 0; i < tpetraMat->getColMap()->getLocalNumElements(); i++) {
+    rghtScaleData[i] = rghtDofPerNodeScale[itemp++];
+    if (itemp == blkSize) itemp = 0;
+  }
+  followed by tpetraMat->rightScale(*rghtScaleVec);
 
-    rghtScaleVec = Teuchos::rcp(new Tpetra::Vector<SC,LO,GO,NO>(tpetraMat->getColMap()));
-    rghtScaleData  = rghtScaleVec->getDataNonConst(0);
-    size_t itemp = 0;
-    for (size_t i = 0; i < tpetraMat->getColMap()->getLocalNumElements(); i++) {
-      rghtScaleData[i] = rghtDofPerNodeScale[itemp++];
-      if (itemp == blkSize) itemp = 0;
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void leftRghtDofScalingWithinNode(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Amat, size_t blkSize, size_t nSweeps, Teuchos::ArrayRCP<Scalar>& rowScaling, Teuchos::ArrayRCP<Scalar>& colScaling) {
+  LocalOrdinal nBlks = (Amat.getRowMap()->getLocalNumElements()) / blkSize;
+
+  Teuchos::ArrayRCP<Scalar> rowScaleUpdate(blkSize);
+  Teuchos::ArrayRCP<Scalar> colScaleUpdate(blkSize);
+
+  for (size_t i = 0; i < blkSize; i++) rowScaling[i] = 1.0;
+  for (size_t i = 0; i < blkSize; i++) colScaling[i] = 1.0;
+
+  for (size_t k = 0; k < nSweeps; k++) {
+    LocalOrdinal row = 0;
+    for (size_t i = 0; i < blkSize; i++) rowScaleUpdate[i] = 0.0;
+
+    for (LocalOrdinal i = 0; i < nBlks; i++) {
+      for (size_t j = 0; j < blkSize; j++) {
+        Teuchos::ArrayView<const LocalOrdinal> cols;
+        Teuchos::ArrayView<const Scalar> vals;
+        Amat.getLocalRowView(row, cols, vals);
+
+        for (size_t kk = 0; kk < Teuchos::as<size_t>(vals.size()); kk++) {
+          size_t modGuy = (cols[kk] + 1) % blkSize;
+          if (modGuy == 0) modGuy = blkSize;
+          modGuy--;
+          rowScaleUpdate[j] += rowScaling[j] * (Teuchos::ScalarTraits<Scalar>::magnitude(vals[kk])) * colScaling[modGuy];
+        }
+        row++;
+      }
     }
-    followed by tpetraMat->rightScale(*rghtScaleVec);
+    // combine information across processors
+    Teuchos::ArrayRCP<Scalar> tempUpdate(blkSize);
+    Teuchos::reduceAll(*(Amat.getRowMap()->getComm()), Teuchos::REDUCE_SUM, (LocalOrdinal)blkSize, rowScaleUpdate.getRawPtr(), tempUpdate.getRawPtr());
+    for (size_t i = 0; i < blkSize; i++) rowScaleUpdate[i] = tempUpdate[i];
 
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void leftRghtDofScalingWithinNode(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & Amat, size_t blkSize, size_t nSweeps, Teuchos::ArrayRCP<Scalar> & rowScaling, Teuchos::ArrayRCP<Scalar> & colScaling) {
-
-     LocalOrdinal     nBlks = (Amat.getRowMap()->getLocalNumElements())/blkSize;
-
-     Teuchos::ArrayRCP<Scalar>   rowScaleUpdate(blkSize);
-     Teuchos::ArrayRCP<Scalar>   colScaleUpdate(blkSize);
-
-
-     for (size_t i = 0; i < blkSize; i++) rowScaling[i] = 1.0;
-     for (size_t i = 0; i < blkSize; i++) colScaling[i] = 1.0;
-
-     for (size_t k = 0; k < nSweeps; k++) {
-       LocalOrdinal row = 0;
-       for (size_t i = 0; i < blkSize; i++) rowScaleUpdate[i] = 0.0;
-
-       for (LocalOrdinal i = 0; i < nBlks; i++) {
-         for (size_t j = 0; j < blkSize; j++) {
-           Teuchos::ArrayView<const LocalOrdinal> cols;
-           Teuchos::ArrayView<const Scalar> vals;
-           Amat.getLocalRowView(row, cols, vals);
-
-           for (size_t kk = 0; kk < Teuchos::as<size_t>(vals.size()); kk++) {
-             size_t modGuy = (cols[kk]+1)%blkSize;
-             if (modGuy == 0) modGuy = blkSize;
-             modGuy--;
-             rowScaleUpdate[j] += rowScaling[j]*(Teuchos::ScalarTraits<Scalar>::magnitude(vals[kk]))*colScaling[modGuy];
-           }
-           row++;
-         }
-       }
-       // combine information across processors
-       Teuchos::ArrayRCP<Scalar>   tempUpdate(blkSize);
-       Teuchos::reduceAll(*(Amat.getRowMap()->getComm()), Teuchos::REDUCE_SUM, (LocalOrdinal) blkSize, rowScaleUpdate.getRawPtr(), tempUpdate.getRawPtr());
-       for (size_t i = 0; i < blkSize; i++) rowScaleUpdate[i] = tempUpdate[i];
-
-       /* We want to scale by sqrt(1/rowScaleUpdate), but we'll         */
-       /* normalize things by the minimum rowScaleUpdate. That is, the  */
-       /* largest scaling is always one (as normalization is arbitrary).*/
-
-       Scalar minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude((rowScaleUpdate[0]/rowScaling[0])/rowScaling[0]);
-
-       for (size_t i = 1; i < blkSize; i++) {
-          Scalar  temp = (rowScaleUpdate[i]/rowScaling[i])/rowScaling[i];
-          if ( Teuchos::ScalarTraits<Scalar>::magnitude(temp) < Teuchos::ScalarTraits<Scalar>::magnitude(minUpdate))
-            minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude(temp);
-       }
-       for (size_t i = 0; i < blkSize; i++) rowScaling[i] *= sqrt(minUpdate / rowScaleUpdate[i]);
-
-       row = 0;
-       for (size_t i = 0; i < blkSize; i++) colScaleUpdate[i] = 0.0;
-
-       for (LocalOrdinal i = 0; i < nBlks; i++) {
-         for (size_t j = 0; j < blkSize; j++) {
-           Teuchos::ArrayView<const LocalOrdinal> cols;
-           Teuchos::ArrayView<const Scalar> vals;
-           Amat.getLocalRowView(row, cols, vals);
-           for (size_t kk = 0; kk < Teuchos::as<size_t>(vals.size()); kk++) {
-             size_t modGuy = (cols[kk]+1)%blkSize;
-             if (modGuy == 0) modGuy = blkSize;
-             modGuy--;
-             colScaleUpdate[modGuy] += colScaling[modGuy]* (Teuchos::ScalarTraits<Scalar>::magnitude(vals[kk])) *rowScaling[j];
-           }
-           row++;
-         }
-       }
-       Teuchos::reduceAll(*(Amat.getRowMap()->getComm()), Teuchos::REDUCE_SUM, (LocalOrdinal) blkSize, colScaleUpdate.getRawPtr(), tempUpdate.getRawPtr());
-       for (size_t i = 0; i < blkSize; i++) colScaleUpdate[i] = tempUpdate[i];
-
-       /* We want to scale by sqrt(1/colScaleUpdate), but we'll         */
-       /* normalize things by the minimum colScaleUpdate. That is, the  */
-       /* largest scaling is always one (as normalization is arbitrary).*/
-
-
-       minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude((colScaleUpdate[0]/colScaling[0])/colScaling[0]);
-
-       for (size_t i = 1; i < blkSize; i++) {
-          Scalar  temp = (colScaleUpdate[i]/colScaling[i])/colScaling[i];
-          if ( Teuchos::ScalarTraits<Scalar>::magnitude(temp) < Teuchos::ScalarTraits<Scalar>::magnitude(minUpdate))
-            minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude(temp);
-       }
-       for (size_t i = 0; i < blkSize; i++) colScaling[i] *= sqrt(minUpdate/colScaleUpdate[i]);
-     }
-  }
+    /* We want to scale by sqrt(1/rowScaleUpdate), but we'll         */
+    /* normalize things by the minimum rowScaleUpdate. That is, the  */
+    /* largest scaling is always one (as normalization is arbitrary).*/
 
-  /*! \fn TpetraCrs_To_XpetraMatrix
-    @brief Helper function to convert a Tpetra::FECrsMatrix to an Xpetra::Matrix
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraFECrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::FECrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra) {
-    typedef typename Tpetra::FECrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::crs_matrix_type tpetra_crs_matrix_type;
-    typedef Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTCrsMatrix;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>       XCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>   XCrsMatrixWrap;
-
-    RCP<XCrsMatrix> Atmp = rcp(new XTCrsMatrix(rcp_dynamic_cast<tpetra_crs_matrix_type>(Atpetra)));
-    return rcp(new XCrsMatrixWrap(Atmp));
-  }
+    Scalar minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude((rowScaleUpdate[0] / rowScaling[0]) / rowScaling[0]);
 
-  /*! \fn TpetraMultiVector_To_XpetraMultiVector
-    @brief Helper function to convert a Tpetra::MultiVector to an Xpetra::MultiVector
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra) {
-    return rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vtpetra));
-  }
+    for (size_t i = 1; i < blkSize; i++) {
+      Scalar temp = (rowScaleUpdate[i] / rowScaling[i]) / rowScaling[i];
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(temp) < Teuchos::ScalarTraits<Scalar>::magnitude(minUpdate))
+        minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude(temp);
+    }
+    for (size_t i = 0; i < blkSize; i++) rowScaling[i] *= sqrt(minUpdate / rowScaleUpdate[i]);
+
+    row = 0;
+    for (size_t i = 0; i < blkSize; i++) colScaleUpdate[i] = 0.0;
+
+    for (LocalOrdinal i = 0; i < nBlks; i++) {
+      for (size_t j = 0; j < blkSize; j++) {
+        Teuchos::ArrayView<const LocalOrdinal> cols;
+        Teuchos::ArrayView<const Scalar> vals;
+        Amat.getLocalRowView(row, cols, vals);
+        for (size_t kk = 0; kk < Teuchos::as<size_t>(vals.size()); kk++) {
+          size_t modGuy = (cols[kk] + 1) % blkSize;
+          if (modGuy == 0) modGuy = blkSize;
+          modGuy--;
+          colScaleUpdate[modGuy] += colScaling[modGuy] * (Teuchos::ScalarTraits<Scalar>::magnitude(vals[kk])) * rowScaling[j];
+        }
+        row++;
+      }
+    }
+    Teuchos::reduceAll(*(Amat.getRowMap()->getComm()), Teuchos::REDUCE_SUM, (LocalOrdinal)blkSize, colScaleUpdate.getRawPtr(), tempUpdate.getRawPtr());
+    for (size_t i = 0; i < blkSize; i++) colScaleUpdate[i] = tempUpdate[i];
 
-  /*! \fn TpetraFEMultiVector_To_XpetraMultiVector
-  @brief Helper function to convert a Tpetra::FEMultiVector to an Xpetra::MultiVector
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraFEMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::FEMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra) {
-    typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
-    RCP<const MV> Vmv = Teuchos::rcp_dynamic_cast<const MV>(Vtpetra);
-    return rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vmv));
-  }
+    /* We want to scale by sqrt(1/colScaleUpdate), but we'll         */
+    /* normalize things by the minimum colScaleUpdate. That is, the  */
+    /* largest scaling is always one (as normalization is arbitrary).*/
+
+    minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude((colScaleUpdate[0] / colScaling[0]) / colScaling[0]);
 
-  //! Little helper function to convert non-string types to strings
-  template<class T>
-  std::string toString(const T& what) {
-    std::ostringstream buf;
-    buf << what;
-    return buf.str();
+    for (size_t i = 1; i < blkSize; i++) {
+      Scalar temp = (colScaleUpdate[i] / colScaling[i]) / colScaling[i];
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(temp) < Teuchos::ScalarTraits<Scalar>::magnitude(minUpdate))
+        minUpdate = Teuchos::ScalarTraits<Scalar>::magnitude(temp);
+    }
+    for (size_t i = 0; i < blkSize; i++) colScaling[i] *= sqrt(minUpdate / colScaleUpdate[i]);
   }
+}
+
+/*! \fn TpetraCrs_To_XpetraMatrix
+  @brief Helper function to convert a Tpetra::FECrsMatrix to an Xpetra::Matrix
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraFECrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::FECrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra) {
+  typedef typename Tpetra::FECrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::crs_matrix_type tpetra_crs_matrix_type;
+  typedef Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XTCrsMatrix;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> XCrsMatrixWrap;
+
+  RCP<XCrsMatrix> Atmp = rcp(new XTCrsMatrix(rcp_dynamic_cast<tpetra_crs_matrix_type>(Atpetra)));
+  return rcp(new XCrsMatrixWrap(Atmp));
+}
+
+/*! \fn TpetraMultiVector_To_XpetraMultiVector
+  @brief Helper function to convert a Tpetra::MultiVector to an Xpetra::MultiVector
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra) {
+  return rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vtpetra));
+}
+
+/*! \fn TpetraFEMultiVector_To_XpetraMultiVector
+@brief Helper function to convert a Tpetra::FEMultiVector to an Xpetra::MultiVector
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraFEMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::FEMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra) {
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  RCP<const MV> Vmv = Teuchos::rcp_dynamic_cast<const MV>(Vtpetra);
+  return rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vmv));
+}
+
+//! Little helper function to convert non-string types to strings
+template <class T>
+std::string toString(const T& what) {
+  std::ostringstream buf;
+  buf << what;
+  return buf.str();
+}
 
 #ifdef HAVE_MUELU_EPETRA
-  /*! \fn EpetraCrs_To_XpetraMatrix
-    @brief Helper function to convert a Epetra::CrsMatrix to an Xpetra::Matrix
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A);
+/*! \fn EpetraCrs_To_XpetraMatrix
+  @brief Helper function to convert a Epetra::CrsMatrix to an Xpetra::Matrix
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+EpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Epetra_CrsMatrix>& A);
 
-  /*! \fn EpetraMultiVector_To_XpetraMultiVector
-    @brief Helper function to convert a Epetra::MultiVector to an Xpetra::MultiVector
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V);
+/*! \fn EpetraMultiVector_To_XpetraMultiVector
+  @brief Helper function to convert a Epetra::MultiVector to an Xpetra::MultiVector
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+EpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Epetra_MultiVector>& V);
 #endif
 
-  /*! \fn TpetraCrs_To_XpetraMatrix
-    @brief Helper function to convert a Tpetra::CrsMatrix to an Xpetra::Matrix
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra);
+/*! \fn TpetraCrs_To_XpetraMatrix
+  @brief Helper function to convert a Tpetra::CrsMatrix to an Xpetra::Matrix
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraCrs_To_XpetraMatrix(const Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Atpetra);
 
-  /*! \fn TpetraMultiVector_To_XpetraMultiVector
-    @brief Helper function to convert a Tpetra::MultiVector to an Xpetra::MultiVector
-    TODO move this function to an Xpetra utility file
-    */
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra);
+/*! \fn TpetraMultiVector_To_XpetraMultiVector
+  @brief Helper function to convert a Tpetra::MultiVector to an Xpetra::MultiVector
+  TODO move this function to an Xpetra utility file
+  */
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+TpetraMultiVector_To_XpetraMultiVector(const Teuchos::RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Vtpetra);
 
-  // Generates a communicator whose only members are other ranks of the baseComm on my node
-  Teuchos::RCP<const Teuchos::Comm<int> > GenerateNodeComm(RCP<const Teuchos::Comm<int> > & baseComm, int &NodeId, const int reductionFactor);
+// Generates a communicator whose only members are other ranks of the baseComm on my node
+Teuchos::RCP<const Teuchos::Comm<int> > GenerateNodeComm(RCP<const Teuchos::Comm<int> >& baseComm, int& NodeId, const int reductionFactor);
 
-  // Lower case string
-  std::string lowerCase (const std::string& s);
+// Lower case string
+std::string lowerCase(const std::string& s);
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UTILITIES_SHORT
-#endif // MUELU_UTILITIES_DECL_HPP
+#endif  // MUELU_UTILITIES_DECL_HPP
diff --git a/packages/muelu/src/Utils/MueLu_Utilities_def.hpp b/packages/muelu/src/Utils/MueLu_Utilities_def.hpp
index c271bd4d5c8b..0c72919eebe0 100644
--- a/packages/muelu/src/Utils/MueLu_Utilities_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_Utilities_def.hpp
@@ -52,9 +52,9 @@
 #include "MueLu_ConfigDefs.hpp"
 
 #ifdef HAVE_MUELU_EPETRA
-# ifdef HAVE_MPI
-#  include "Epetra_MpiComm.h"
-# endif
+#ifdef HAVE_MPI
+#include "Epetra_MpiComm.h"
+#endif
 #endif
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
@@ -105,608 +105,593 @@
 namespace MueLu {
 
 #ifdef HAVE_MUELU_EPETRA
-  //using Xpetra::EpetraCrsMatrix;   // TODO: mv in Xpetra_UseShortNamesScalar
-  //using Xpetra::EpetraMultiVector;
+// using Xpetra::EpetraCrsMatrix;   // TODO: mv in Xpetra_UseShortNamesScalar
+// using Xpetra::EpetraMultiVector;
 #endif
 
 #ifdef HAVE_MUELU_EPETRA
-  template<typename SC,typename LO,typename GO,typename NO>
-  RCP<Xpetra::CrsMatrixWrap<SC,LO,GO,NO> > Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap(RCP<Epetra_CrsMatrix> &epAB){
-    return Xpetra::Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap<SC,LO,GO,NO>(epAB);
-  }
+template <typename SC, typename LO, typename GO, typename NO>
+RCP<Xpetra::CrsMatrixWrap<SC, LO, GO, NO>> Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap(RCP<Epetra_CrsMatrix>& epAB) {
+  return Xpetra::Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap<SC, LO, GO, NO>(epAB);
+}
 #endif
 
 #ifdef HAVE_MUELU_EPETRA
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Epetra_MultiVector> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2EpetraMV(const RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > vec) {
-    RCP<const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node>  > tmpVec = rcp_dynamic_cast<Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> >(vec);
-    if (tmpVec == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
-    return tmpVec->getEpetra_MultiVector();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Epetra_MultiVector> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstEpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > vec) {
-    RCP<const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> > tmpVec = rcp_dynamic_cast<Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> >(vec);
-    if (tmpVec == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
-    return tmpVec->getEpetra_MultiVector();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Epetra_MultiVector& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstEpetraMV(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &vec) {
-    const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> & tmpVec = dynamic_cast<const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> &>(vec);
-    return *(tmpVec.getEpetra_MultiVector());
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Epetra_MultiVector& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2EpetraMV(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec) {
-    const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> & tmpVec = dynamic_cast<const Xpetra::EpetraMultiVectorT<GlobalOrdinal,Node> &>(vec);
-    return *(tmpVec.getEpetra_MultiVector());
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Epetra_CrsMatrix> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2EpetraCrs(RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op) {
-    RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-    return tmp_ECrsMtx->getEpetra_CrsMatrix();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Epetra_CrsMatrix> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op) {
-    RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-    return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Epetra_CrsMatrix& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2EpetraCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Epetra_MultiVector> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2EpetraMV(const RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> vec) {
+  RCP<const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>> tmpVec = rcp_dynamic_cast<Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>>(vec);
+  if (tmpVec == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
+  return tmpVec->getEpetra_MultiVector();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Epetra_MultiVector> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstEpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> vec) {
+  RCP<const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>> tmpVec = rcp_dynamic_cast<Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>>(vec);
+  if (tmpVec == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::EpetraMultiVector failed");
+  return tmpVec->getEpetra_MultiVector();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Epetra_MultiVector& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstEpetraMV(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec) {
+  const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>&>(vec);
+  return *(tmpVec.getEpetra_MultiVector());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Epetra_MultiVector& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2EpetraMV(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec) {
+  const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::EpetraMultiVectorT<GlobalOrdinal, Node>&>(vec);
+  return *(tmpVec.getEpetra_MultiVector());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Epetra_CrsMatrix> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2EpetraCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
+  return tmp_ECrsMtx->getEpetra_CrsMatrix();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Epetra_CrsMatrix> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
+  return tmp_ECrsMtx->getEpetra_CrsMatrixNonConst();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Epetra_CrsMatrix& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2EpetraCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>& crsOp = dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(Op);
     try {
-      const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>& crsOp = dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(Op);
-      try {
-        const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getEpetra_CrsMatrix();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-      }
+      const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getEpetra_CrsMatrix();
     } catch (std::bad_cast&) {
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Epetra_CrsMatrix& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Epetra_CrsMatrix& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>& crsOp = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(Op);
     try {
-      Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>& crsOp = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(Op);
-      try {
-        Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getEpetra_CrsMatrixNonConst();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
-      }
+      Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getEpetra_CrsMatrixNonConst();
     } catch (std::bad_cast&) {
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::EpetraCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Epetra_Map& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Map2EpetraMap(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& map) {
-    RCP<const Xpetra::EpetraMapT<GlobalOrdinal,Node> > xeMap = rcp_dynamic_cast<const Xpetra::EpetraMapT<GlobalOrdinal,Node> >(rcpFromRef(map));
-    if (xeMap == Teuchos::null)
-      throw Exceptions::BadCast("Utilities::Map2EpetraMap : Cast from Xpetra::Map to Xpetra::EpetraMap failed");
-    return xeMap->getEpetra_Map();
-  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Epetra_Map& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Map2EpetraMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& map) {
+  RCP<const Xpetra::EpetraMapT<GlobalOrdinal, Node>> xeMap = rcp_dynamic_cast<const Xpetra::EpetraMapT<GlobalOrdinal, Node>>(rcpFromRef(map));
+  if (xeMap == Teuchos::null)
+    throw Exceptions::BadCast("Utilities::Map2EpetraMap : Cast from Xpetra::Map to Xpetra::EpetraMap failed");
+  return xeMap->getEpetra_Map();
+}
 #endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > const vec) {
-    RCP<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(vec);
-    if (tmpVec == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
-    return tmpVec->getTpetra_MultiVector();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > vec) {
-    RCP<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(vec);
-    if (tmpVec == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
-    return tmpVec->getTpetra_MultiVector();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> & Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec) {
-    const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-    return *(tmpVec.getTpetra_MultiVector());
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV2(Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &vec) {
-    const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-    return tmpVec.getTpetra_MultiVector();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&
-  Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& vec) {
-    const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(vec);
-    return *(tmpVec.getTpetra_MultiVector());
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op) {
-    // Get the underlying Tpetra Mtx
-    RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-    return tmp_ECrsMtx->getTpetra_CrsMatrix();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op) {
-    RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-    return tmp_ECrsMtx->getTpetra_CrsMatrixNonConst();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> const vec) {
+  RCP<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(vec);
+  if (tmpVec == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
+  return tmpVec->getTpetra_MultiVector();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> vec) {
+  RCP<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmpVec = rcp_dynamic_cast<Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(vec);
+  if (tmpVec == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::MultiVector to Xpetra::TpetraMultiVector failed");
+  return tmpVec->getTpetra_MultiVector();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec) {
+  const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+  return *(tmpVec.getTpetra_MultiVector());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV2(Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec) {
+  const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+  return tmpVec.getTpetra_MultiVector();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&
+Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& vec) {
+  const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmpVec = dynamic_cast<const Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(vec);
+  return *(tmpVec.getTpetra_MultiVector());
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  // Get the underlying Tpetra Mtx
+  RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
+  return tmp_ECrsMtx->getTpetra_CrsMatrix();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
+  return tmp_ECrsMtx->getTpetra_CrsMatrixNonConst();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>& crsOp = dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(Op);
     try {
-      const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>& crsOp = dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(Op);
-      try {
-        const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getTpetra_CrsMatrix();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-      }
+      const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getTpetra_CrsMatrix();
     } catch (std::bad_cast&) {
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>& crsOp = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(Op);
     try {
-      Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>& crsOp = dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(Op);
-      try {
-        Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getTpetra_CrsMatrixNonConst();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
-      }
+      Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getTpetra_CrsMatrixNonConst();
     } catch (std::bad_cast&) {
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op)  {
-    using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    // Get the underlying Tpetra Mtx
-    RCP<const XCrsMatrixWrap> crsOp = rcp_dynamic_cast<const XCrsMatrixWrap>(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-    return tmp_ECrsMtx->getTpetra_BlockCrsMatrix();
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<      Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >   Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraBlockCrs(RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op){
-    using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-    RCP<const XCrsMatrixWrap> crsOp = rcp_dynamic_cast<const XCrsMatrixWrap>(Op);
-    if (crsOp == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-    const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsOp->getCrsMatrix());
-    if (tmp_ECrsMtx == Teuchos::null)
-      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-    return tmp_ECrsMtx->getTpetra_BlockCrsMatrixNonConst();
-  }
-  
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op)   {
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  // Get the underlying Tpetra Mtx
+  RCP<const XCrsMatrixWrap> crsOp = rcp_dynamic_cast<const XCrsMatrixWrap>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
+  return tmp_ECrsMtx->getTpetra_BlockCrsMatrix();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraBlockCrs(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  using XCrsMatrixWrap            = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  RCP<const XCrsMatrixWrap> crsOp = rcp_dynamic_cast<const XCrsMatrixWrap>(Op);
+  if (crsOp == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+  const RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& tmp_ECrsMtx = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsOp->getCrsMatrix());
+  if (tmp_ECrsMtx == Teuchos::null)
+    throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
+  return tmp_ECrsMtx->getTpetra_BlockCrsMatrixNonConst();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    using XCrsMatrixWrap        = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    const XCrsMatrixWrap& crsOp = dynamic_cast<const XCrsMatrixWrap&>(Op);
     try {
-      using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      const XCrsMatrixWrap& crsOp = dynamic_cast<const XCrsMatrixWrap&>(Op);
-      try {
-        const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getTpetra_BlockCrsMatrix();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-      }
+      const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getTpetra_BlockCrsMatrix();
     } catch (std::bad_cast&) {
-      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Tpetra::BlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&        Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraBlockCrs(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraBlockCrs(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op) {
+  try {
+    using XCrsMatrixWrap  = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    XCrsMatrixWrap& crsOp = dynamic_cast<XCrsMatrixWrap&>(Op);
     try {
-      using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      XCrsMatrixWrap& crsOp = dynamic_cast<XCrsMatrixWrap&>(Op);
-      try {
-        Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&>(*crsOp.getCrsMatrix());
-        return *tmp_ECrsMtx.getTpetra_BlockCrsMatrixNonConst();
-      } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
-      }
+      Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tmp_ECrsMtx = dynamic_cast<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>&>(*crsOp.getCrsMatrix());
+      return *tmp_ECrsMtx.getTpetra_BlockCrsMatrixNonConst();
     } catch (std::bad_cast&) {
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+      throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraBlockCrsMatrix failed");
     }
+  } catch (std::bad_cast&) {
+    throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   }
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraRow(RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> mat = rcp_dynamic_cast<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (!mat.is_null()) {
+    RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(mat);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraRow(RCP<const Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Op) {
-    RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > mat = rcp_dynamic_cast<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (!mat.is_null()) {
-      RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(mat);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-
-      RCP<const Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsMat = crsOp->getCrsMatrix();
-      const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-      RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_BlockCrs;
-      if(!tmp_Crs.is_null()) {
-        return tmp_Crs->getTpetra_CrsMatrixNonConst();
-      }
-      else {
-        tmp_BlockCrs= rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-        if (tmp_BlockCrs.is_null())
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
-        return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
-      }
+    RCP<const Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsMat              = crsOp->getCrsMatrix();
+    const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsMat);
+    RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmp_BlockCrs;
+    if (!tmp_Crs.is_null()) {
+      return tmp_Crs->getTpetra_CrsMatrixNonConst();
     } else {
-      RCP<const Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpOp = rcp_dynamic_cast<const Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op, true);
-      RCP<const Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>> tOp = tpOp->getOperatorConst();
-      RCP<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tRow = rcp_dynamic_cast<const Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tOp, true);
-      return tRow;
+      tmp_BlockCrs = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsMat);
+      if (tmp_BlockCrs.is_null())
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
+      return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
     }
+  } else {
+    RCP<const Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tpOp = rcp_dynamic_cast<const Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op, true);
+    RCP<const Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tOp        = tpOp->getOperatorConst();
+    RCP<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tRow      = rcp_dynamic_cast<const Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(tOp, true);
+    return tRow;
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraRow(RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Op) {
-    RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > mat = rcp_dynamic_cast<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op);
-    if (!mat.is_null()) {
-      RCP<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(mat);
-      if (crsOp == Teuchos::null)
-        throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
-
-      RCP<const Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsMat = crsOp->getCrsMatrix();
-      const RCP<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-      RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tmp_BlockCrs;
-      if(!tmp_Crs.is_null()) {
-        return tmp_Crs->getTpetra_CrsMatrixNonConst();
-      }
-      else {
-        tmp_BlockCrs= rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(crsMat);
-        if (tmp_BlockCrs.is_null())
-          throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
-        return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraRow(RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Op) {
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> mat = rcp_dynamic_cast<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op);
+  if (!mat.is_null()) {
+    RCP<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsOp = rcp_dynamic_cast<const Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(mat);
+    if (crsOp == Teuchos::null)
+      throw Exceptions::BadCast("Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+
+    RCP<const Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> crsMat              = crsOp->getCrsMatrix();
+    const RCP<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmp_Crs = rcp_dynamic_cast<const Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsMat);
+    RCP<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tmp_BlockCrs;
+    if (!tmp_Crs.is_null()) {
+      return tmp_Crs->getTpetra_CrsMatrixNonConst();
     } else {
-      RCP<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(Op, true);
-      RCP<Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node>> tOp = tpOp->getOperator();
-      RCP<Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tRow = rcp_dynamic_cast<Tpetra::RowMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(tOp, true);
-      return tRow;
+      tmp_BlockCrs = rcp_dynamic_cast<const Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(crsMat);
+      if (tmp_BlockCrs.is_null())
+        throw Exceptions::BadCast("Cast from Xpetra::CrsMatrix to Xpetra::TpetraCrsMatrix and Xpetra::TpetraBlockCrsMatrix failed");
+      return tmp_BlockCrs->getTpetra_BlockCrsMatrixNonConst();
     }
+  } else {
+    RCP<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tpOp = rcp_dynamic_cast<Xpetra::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(Op, true);
+    RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tOp        = tpOp->getOperator();
+    RCP<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> tRow      = rcp_dynamic_cast<Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(tOp, true);
+    return tRow;
   }
-
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal,Node> > Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Map2TpetraMap(const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>& map) {
-    const RCP<const Xpetra::TpetraMap<LocalOrdinal,GlobalOrdinal,Node>>& tmp_TMap = rcp_dynamic_cast<const Xpetra::TpetraMap<LocalOrdinal,GlobalOrdinal,Node> >(rcpFromRef(map));
-    if (tmp_TMap == Teuchos::null)
-      throw Exceptions::BadCast("Utilities::Map2TpetraMap : Cast from Xpetra::Map to Xpetra::TpetraMap failed");
-    return tmp_TMap->getTpetra_Map();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Map2TpetraMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& map) {
+  const RCP<const Xpetra::TpetraMap<LocalOrdinal, GlobalOrdinal, Node>>& tmp_TMap = rcp_dynamic_cast<const Xpetra::TpetraMap<LocalOrdinal, GlobalOrdinal, Node>>(rcpFromRef(map));
+  if (tmp_TMap == Teuchos::null)
+    throw Exceptions::BadCast("Utilities::Map2TpetraMap : Cast from Xpetra::Map to Xpetra::TpetraMap failed");
+  return tmp_TMap->getTpetra_Map();
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse,
+                                                                            bool doFillComplete,
+                                                                            bool doOptimizeStorage) {
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+  Teuchos::ArrayRCP<Scalar> sv(scalingVector.size());
+  if (doInverse) {
+    for (int i = 0; i < scalingVector.size(); ++i)
+      sv[i] = one / scalingVector[i];
+  } else {
+    for (int i = 0; i < scalingVector.size(); ++i)
+      sv[i] = scalingVector[i];
   }
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const Teuchos::ArrayRCP<const Scalar>& scalingVector, bool doInverse,
-                               bool doFillComplete,
-                               bool doOptimizeStorage)
-  {
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-    Teuchos::ArrayRCP<Scalar> sv(scalingVector.size());
-    if (doInverse) {
-      for (int i = 0; i < scalingVector.size(); ++i)
-        sv[i] = one / scalingVector[i];
-    } else {
-      for (int i = 0; i < scalingVector.size(); ++i)
-        sv[i] = scalingVector[i];
-    }
+  switch (Op.getRowMap()->lib()) {
+    case Xpetra::UseTpetra:
+      MyOldScaleMatrix_Tpetra(Op, sv, doFillComplete, doOptimizeStorage);
+      break;
 
-    switch (Op.getRowMap()->lib()) {
-      case Xpetra::UseTpetra:
-        MyOldScaleMatrix_Tpetra(Op, sv, doFillComplete, doOptimizeStorage);
-        break;
+    case Xpetra::UseEpetra:
+      MyOldScaleMatrix_Epetra(Op, sv, doFillComplete, doOptimizeStorage);
+      break;
 
-      case Xpetra::UseEpetra:
-        MyOldScaleMatrix_Epetra(Op, sv, doFillComplete, doOptimizeStorage);
-        break;
-
-      default:
-        throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be scaled.");
-    }
+    default:
+      throw Exceptions::RuntimeError("Only Epetra and Tpetra matrices can be scaled.");
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix_Epetra(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& /* Op */, const Teuchos::ArrayRCP<Scalar>& /* scalingVector */, bool /* doFillComplete */, bool /* doOptimizeStorage */) {
-    throw Exceptions::RuntimeError("MyOldScaleMatrix_Epetra: Epetra needs SC=double and LO=GO=int.");
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix_Tpetra(Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
-                               bool doFillComplete,
-                               bool doOptimizeStorage)
-  {
-    try {
-      Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpOp = Op2NonConstTpetraCrs(Op);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix_Epetra(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& /* Op */, const Teuchos::ArrayRCP<Scalar>& /* scalingVector */, bool /* doFillComplete */, bool /* doOptimizeStorage */) {
+  throw Exceptions::RuntimeError("MyOldScaleMatrix_Epetra: Epetra needs SC=double and LO=GO=int.");
+}
 
-      const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rowMap    = tpOp.getRowMap();
-      const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > domainMap = tpOp.getDomainMap();
-      const RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > rangeMap  = tpOp.getRangeMap();
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MyOldScaleMatrix_Tpetra(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, const Teuchos::ArrayRCP<Scalar>& scalingVector,
+                                                                                   bool doFillComplete,
+                                                                                   bool doOptimizeStorage) {
+  try {
+    Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpOp = Op2NonConstTpetraCrs(Op);
 
-      size_t maxRowSize = tpOp.getLocalMaxNumRowEntries();
-      if (maxRowSize == Teuchos::as<size_t>(-1)) // hasn't been determined yet
-        maxRowSize = 20;
+    const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rowMap    = tpOp.getRowMap();
+    const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> domainMap = tpOp.getDomainMap();
+    const RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> rangeMap  = tpOp.getRangeMap();
 
+    size_t maxRowSize = tpOp.getLocalMaxNumRowEntries();
+    if (maxRowSize == Teuchos::as<size_t>(-1))  // hasn't been determined yet
+      maxRowSize = 20;
 
-      if (tpOp.isFillComplete())
-        tpOp.resumeFill();
+    if (tpOp.isFillComplete())
+      tpOp.resumeFill();
 
-      if (Op.isLocallyIndexed() == true) {
-	typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
-	typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
+    if (Op.isLocallyIndexed() == true) {
+      typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type cols;
+      typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
 
-        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-          tpOp.getLocalRowView(i, cols, vals);
+      for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+        tpOp.getLocalRowView(i, cols, vals);
 
-          size_t nnz = tpOp.getNumEntriesInLocalRow(i);
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::nonconst_values_host_view_type scaledVals("ScaledVals",nnz);
+        size_t nnz = tpOp.getNumEntriesInLocalRow(i);
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::nonconst_values_host_view_type scaledVals("ScaledVals", nnz);
 
-          for (size_t j = 0; j < nnz; ++j) {
-	    scaledVals[j] = scalingVector[i]*vals[j];
-	  }
+        for (size_t j = 0; j < nnz; ++j) {
+          scaledVals[j] = scalingVector[i] * vals[j];
+        }
 
-          if (nnz > 0) {
-            tpOp.replaceLocalValues(i, cols, scaledVals);
-          }
-        } //for (size_t i=0; ...
+        if (nnz > 0) {
+          tpOp.replaceLocalValues(i, cols, scaledVals);
+        }
+      }  // for (size_t i=0; ...
 
-      } else {
-	typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::global_inds_host_view_type cols;
-	typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
+    } else {
+      typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::global_inds_host_view_type cols;
+      typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
 
-        for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
-          GlobalOrdinal gid = rowMap->getGlobalElement(i);
-          tpOp.getGlobalRowView(gid, cols, vals);
-          size_t nnz = tpOp.getNumEntriesInGlobalRow(gid);
-	  typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::nonconst_values_host_view_type scaledVals("ScaledVals",nnz);
+      for (size_t i = 0; i < rowMap->getLocalNumElements(); ++i) {
+        GlobalOrdinal gid = rowMap->getGlobalElement(i);
+        tpOp.getGlobalRowView(gid, cols, vals);
+        size_t nnz = tpOp.getNumEntriesInGlobalRow(gid);
+        typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::nonconst_values_host_view_type scaledVals("ScaledVals", nnz);
 
-          // FIXME FIXME FIXME FIXME FIXME FIXME
-          for (size_t j = 0; j < nnz; ++j)
-            scaledVals[j] = scalingVector[i]*vals[j]; //FIXME i or gid?
+        // FIXME FIXME FIXME FIXME FIXME FIXME
+        for (size_t j = 0; j < nnz; ++j)
+          scaledVals[j] = scalingVector[i] * vals[j];  // FIXME i or gid?
 
-          if (nnz > 0) {
-            tpOp.replaceGlobalValues(gid, cols, scaledVals);
-          }
-        } //for (size_t i=0; ...
-      }
+        if (nnz > 0) {
+          tpOp.replaceGlobalValues(gid, cols, scaledVals);
+        }
+      }  // for (size_t i=0; ...
+    }
 
-      if (doFillComplete) {
-        if (domainMap == Teuchos::null || rangeMap == Teuchos::null)
-          throw Exceptions::RuntimeError("In Utilities::Scaling: cannot fillComplete because the domain and/or range map hasn't been defined");
+    if (doFillComplete) {
+      if (domainMap == Teuchos::null || rangeMap == Teuchos::null)
+        throw Exceptions::RuntimeError("In Utilities::Scaling: cannot fillComplete because the domain and/or range map hasn't been defined");
 
-        RCP<Teuchos::ParameterList> params = rcp(new Teuchos::ParameterList());
-        params->set("Optimize Storage",    doOptimizeStorage);
-        params->set("No Nonlocal Changes", true);
-        Op.fillComplete(Op.getDomainMap(), Op.getRangeMap(), params);
-      }
-    } catch(...) {
-      throw Exceptions::RuntimeError("Only Tpetra::CrsMatrix types can be scaled (Err.1)");
+      RCP<Teuchos::ParameterList> params = rcp(new Teuchos::ParameterList());
+      params->set("Optimize Storage", doOptimizeStorage);
+      params->set("No Nonlocal Changes", true);
+      Op.fillComplete(Op.getDomainMap(), Op.getRangeMap(), params);
     }
-  } //MyOldScaleMatrix_Tpetra()
+  } catch (...) {
+    throw Exceptions::RuntimeError("Only Tpetra::CrsMatrix types can be scaled (Err.1)");
+  }
+}  // MyOldScaleMatrix_Tpetra()
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  Transpose (Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& Op, bool /* optimizeTranspose */,const std::string & label,const Teuchos::RCP<Teuchos::ParameterList> &params) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    Transpose(Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Op, bool /* optimizeTranspose */, const std::string& label, const Teuchos::RCP<Teuchos::ParameterList>& params) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
-    std::string TorE = "epetra";
+  std::string TorE = "epetra";
 #else
-    std::string TorE = "tpetra";
+  std::string TorE = "tpetra";
 #endif
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
-    try {
-      const Epetra_CrsMatrix& epetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Op);
-      (void) epetraOp; // silence "unused variable" compiler warning
-    } catch (...) {
-      TorE = "tpetra";
-    }
+  try {
+    const Epetra_CrsMatrix& epetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(Op);
+    (void)epetraOp;  // silence "unused variable" compiler warning
+  } catch (...) {
+    TorE = "tpetra";
+  }
 #endif
 
-    if (TorE == "tpetra") {
-      using Helpers = Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-      /***************************************************************/
-      if(Helpers::isTpetraCrs(Op)) {
-        const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(Op);
-        
-        RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A;
-        Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp),label); //more than meets the eye
-        
-        {
-          using Teuchos::ParameterList;
-          using Teuchos::rcp;
-          RCP<ParameterList> transposeParams = params.is_null () ?
-            rcp (new ParameterList) :
-            rcp (new ParameterList (*params));
-          transposeParams->set ("sort", false);
-          A = transposer.createTranspose (transposeParams);
-        }
-        
-        RCP<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA   = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A) );
-        RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >       AAA  = rcp_implicit_cast<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(AA);
-        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >          AAAA = rcp( new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(AAA) );
-        if (!AAAA->isFillComplete())
-          AAAA->fillComplete(Op.getRangeMap(), Op.getDomainMap());
-        
-        if (Op.IsView("stridedMaps"))
-          AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true/*doTranspose*/);
-        
-        return AAAA;
+  if (TorE == "tpetra") {
+    using Helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+    /***************************************************************/
+    if (Helpers::isTpetraCrs(Op)) {
+      const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraCrs(Op);
+
+      RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> A;
+      Tpetra::RowMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp), label);  // more than meets the eye
+
+      {
+        using Teuchos::ParameterList;
+        using Teuchos::rcp;
+        RCP<ParameterList> transposeParams = params.is_null() ? rcp(new ParameterList) : rcp(new ParameterList(*params));
+        transposeParams->set("sort", false);
+        A = transposer.createTranspose(transposeParams);
       }
-      else if(Helpers::isTpetraBlockCrs(Op)) {
-        using XMatrix = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-        using XCrsMatrix = Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-        using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-        using BCRS = Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-        // using CRS  = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-        const BCRS & tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(Op);
-        
-        RCP<BCRS> At;
-        {
-          Tpetra::BlockCrsMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp),label);
-          
-          using Teuchos::ParameterList;
-          using Teuchos::rcp;
-          RCP<ParameterList> transposeParams = params.is_null () ?
-            rcp (new ParameterList) :
-            rcp (new ParameterList (*params));
-          transposeParams->set ("sort", false);
-          At = transposer.createTranspose(transposeParams);
-        }
-        
-        RCP<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > AA   = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(At));
-        RCP<XCrsMatrix>                                                           AAA  = rcp_implicit_cast<XCrsMatrix>(AA);
-        RCP<XMatrix>                                                              AAAA = rcp( new XCrsMatrixWrap(AAA));
-        
-        if (Op.IsView("stridedMaps"))
-          AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true/*doTranspose*/);
-        
-        return AAAA;
-      } else {
-        throw Exceptions::RuntimeError("Utilities::Transpose failed, perhaps because matrix is not a Crs matrix");
-      }
-    } //if
 
-    // Epetra case
-    std::cout << "Utilities::Transpose() not implemented for Epetra" << std::endl;
-    return Teuchos::null;
-
-  } // Transpose
+      RCP<Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AA = rcp(new Xpetra::TpetraCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A));
+      RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AAA      = rcp_implicit_cast<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(AA);
+      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AAAA        = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(AAA));
+      if (!AAAA->isFillComplete())
+        AAAA->fillComplete(Op.getRangeMap(), Op.getDomainMap());
+
+      if (Op.IsView("stridedMaps"))
+        AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true /*doTranspose*/);
+
+      return AAAA;
+    } else if (Helpers::isTpetraBlockCrs(Op)) {
+      using XMatrix        = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      using XCrsMatrix     = Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      using XCrsMatrixWrap = Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      using BCRS           = Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      // using CRS  = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      const BCRS& tpetraOp = Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2TpetraBlockCrs(Op);
+
+      RCP<BCRS> At;
+      {
+        Tpetra::BlockCrsMatrixTransposer<Scalar, LocalOrdinal, GlobalOrdinal, Node> transposer(rcpFromRef(tpetraOp), label);
+
+        using Teuchos::ParameterList;
+        using Teuchos::rcp;
+        RCP<ParameterList> transposeParams = params.is_null() ? rcp(new ParameterList) : rcp(new ParameterList(*params));
+        transposeParams->set("sort", false);
+        At = transposer.createTranspose(transposeParams);
+      }
 
+      RCP<Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> AA = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(At));
+      RCP<XCrsMatrix> AAA                                                             = rcp_implicit_cast<XCrsMatrix>(AA);
+      RCP<XMatrix> AAAA                                                               = rcp(new XCrsMatrixWrap(AAA));
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >
-  Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> > X) {
-    RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > Xscalar;
-#if defined(HAVE_XPETRA_TPETRA) && (defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) || defined(HAVE_TPETRA_INST_COMPLEX_FLOAT))
-    using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    // Need to cast the real-valued multivector to Scalar=complex
-    if ((typeid(Scalar).name() == typeid(std::complex<double>).name()) ||
-        (typeid(Scalar).name() == typeid(std::complex<float>).name())) {
-      size_t numVecs = X->getNumVectors();
-      Xscalar = Xpetra::MultiVectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(X->getMap(),numVecs);
-      auto XVec = X->getDeviceLocalView(Xpetra::Access::ReadOnly);
-      auto XVecScalar = Xscalar->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-      Kokkos::parallel_for("MueLu:Utils::RealValuedToScalarMultiVector", range_type(0,X->getLocalLength()),
-                           KOKKOS_LAMBDA(const size_t i) {
-                             for (size_t j=0; j<numVecs; j++)
-                               XVecScalar(i,j) = XVec(i,j);
-                           });
-    } else
-#endif
-      Xscalar = rcp_dynamic_cast<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(X);
-    return Xscalar;
-  }
+      if (Op.IsView("stridedMaps"))
+        AAAA->CreateView("stridedMaps", Teuchos::rcpFromRef(Op), true /*doTranspose*/);
 
+      return AAAA;
+    } else {
+      throw Exceptions::RuntimeError("Utilities::Transpose failed, perhaps because matrix is not a Crs matrix");
+    }
+  }  // if
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >
-  Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
-  ExtractCoordinatesFromParameterList (ParameterList& paramList) {
+  // Epetra case
+  std::cout << "Utilities::Transpose() not implemented for Epetra" << std::endl;
+  return Teuchos::null;
 
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > coordinates = Teuchos::null;
+}  // Transpose
 
-    // check whether coordinates are contained in parameter list
-    if(paramList.isParameter ("Coordinates") == false)
-      return coordinates;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>
+Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    RealValuedToScalarMultiVector(RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>> X) {
+  RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Xscalar;
+#if defined(HAVE_XPETRA_TPETRA) && (defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) || defined(HAVE_TPETRA_INST_COMPLEX_FLOAT))
+  using range_type = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+
+  // Need to cast the real-valued multivector to Scalar=complex
+  if ((typeid(Scalar).name() == typeid(std::complex<double>).name()) ||
+      (typeid(Scalar).name() == typeid(std::complex<float>).name())) {
+    size_t numVecs  = X->getNumVectors();
+    Xscalar         = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(X->getMap(), numVecs);
+    auto XVec       = X->getDeviceLocalView(Xpetra::Access::ReadOnly);
+    auto XVecScalar = Xscalar->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+    Kokkos::parallel_for(
+        "MueLu:Utils::RealValuedToScalarMultiVector", range_type(0, X->getLocalLength()),
+        KOKKOS_LAMBDA(const size_t i) {
+          for (size_t j = 0; j < numVecs; j++)
+            XVecScalar(i, j) = XVec(i, j);
+        });
+  } else
+#endif
+    Xscalar = rcp_dynamic_cast<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(X);
+  return Xscalar;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>>
+Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::
+    ExtractCoordinatesFromParameterList(ParameterList& paramList) {
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>> coordinates = Teuchos::null;
+
+  // check whether coordinates are contained in parameter list
+  if (paramList.isParameter("Coordinates") == false)
+    return coordinates;
 
     // define Tpetra::MultiVector type with Scalar=float only if
     // * ETI is turned off, since then the compiler will instantiate it automatically OR
     // * Tpetra is instantiated on Scalar=float
 #if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
-    typedef Tpetra::MultiVector<float, LocalOrdinal, GlobalOrdinal, Node> tfMV;
-    RCP<tfMV> floatCoords = Teuchos::null;
+  typedef Tpetra::MultiVector<float, LocalOrdinal, GlobalOrdinal, Node> tfMV;
+  RCP<tfMV> floatCoords = Teuchos::null;
 #endif
 
-    // define Tpetra::MultiVector type with Scalar=double only if
-    // * ETI is turned off, since then the compiler will instantiate it automatically OR
-    // * Tpetra is instantiated on Scalar=double
+  // define Tpetra::MultiVector type with Scalar=double only if
+  // * ETI is turned off, since then the compiler will instantiate it automatically OR
+  // * Tpetra is instantiated on Scalar=double
 #if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_DOUBLE)
-    typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> tdMV;
-    RCP<tdMV> doubleCoords = Teuchos::null;
-    if (paramList.isType<RCP<tdMV> >("Coordinates")) {
-      // Coordinates are stored as a double vector
-      doubleCoords = paramList.get<RCP<tdMV> >("Coordinates");
-      paramList.remove("Coordinates");
-    }
+  typedef Tpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> tdMV;
+  RCP<tdMV> doubleCoords = Teuchos::null;
+  if (paramList.isType<RCP<tdMV>>("Coordinates")) {
+    // Coordinates are stored as a double vector
+    doubleCoords = paramList.get<RCP<tdMV>>("Coordinates");
+    paramList.remove("Coordinates");
+  }
 #if !defined(HAVE_TPETRA_EXPLICIT_INSTANTIATION) || defined(HAVE_TPETRA_INST_FLOAT)
-    else if (paramList.isType<RCP<tfMV> >("Coordinates")) {
-      // check if coordinates are stored as a float vector
-      floatCoords = paramList.get<RCP<tfMV> >("Coordinates");
-      paramList.remove("Coordinates");
-      doubleCoords = rcp(new tdMV(floatCoords->getMap(), floatCoords->getNumVectors()));
-      deep_copy(*doubleCoords, *floatCoords);
-    }
+  else if (paramList.isType<RCP<tfMV>>("Coordinates")) {
+    // check if coordinates are stored as a float vector
+    floatCoords = paramList.get<RCP<tfMV>>("Coordinates");
+    paramList.remove("Coordinates");
+    doubleCoords = rcp(new tdMV(floatCoords->getMap(), floatCoords->getNumVectors()));
+    deep_copy(*doubleCoords, *floatCoords);
+  }
 #endif
-    // We have the coordinates in a Tpetra double vector
-    if(doubleCoords != Teuchos::null) {
-      //rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vtpetra));
-      coordinates = Teuchos::rcp_dynamic_cast<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >(MueLu::TpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node>(doubleCoords));
-      TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(coordinates));
-      TEUCHOS_TEST_FOR_EXCEPT(doubleCoords->getNumVectors() != coordinates->getNumVectors());
-    }
+  // We have the coordinates in a Tpetra double vector
+  if (doubleCoords != Teuchos::null) {
+    // rcp(new Xpetra::TpetraMultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>(Vtpetra));
+    coordinates = Teuchos::rcp_dynamic_cast<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>>(MueLu::TpetraMultiVector_To_XpetraMultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>(doubleCoords));
+    TEUCHOS_TEST_FOR_EXCEPT(Teuchos::is_null(coordinates));
+    TEUCHOS_TEST_FOR_EXCEPT(doubleCoords->getNumVectors() != coordinates->getNumVectors());
+  }
 #else
-    // coordinates usually are stored as double vector
-    // Tpetra is not instantiated on scalar=double
-    throw Exceptions::RuntimeError("ExtractCoordinatesFromParameterList: The coordinates vector in parameter list is expected to be a Tpetra multivector with SC=double or float.");
+  // coordinates usually are stored as double vector
+  // Tpetra is not instantiated on scalar=double
+  throw Exceptions::RuntimeError("ExtractCoordinatesFromParameterList: The coordinates vector in parameter list is expected to be a Tpetra multivector with SC=double or float.");
 #endif
 
-    // check for Xpetra coordinates vector
-    if(paramList.isType<decltype(coordinates)>("Coordinates")) {
-      coordinates = paramList.get<decltype(coordinates)>("Coordinates");
-    }
+  // check for Xpetra coordinates vector
+  if (paramList.isType<decltype(coordinates)>("Coordinates")) {
+    coordinates = paramList.get<decltype(coordinates)>("Coordinates");
+  }
 
-    return coordinates;
-  } // ExtractCoordinatesFromParameterList
+  return coordinates;
+}  // ExtractCoordinatesFromParameterList
 
-} //namespace MueLu
+}  // namespace MueLu
 
 #define MUELU_UTILITIES_SHORT
-#endif // MUELU_UTILITIES_DEF_HPP
+#endif  // MUELU_UTILITIES_DEF_HPP
 
 //  LocalWords:  LocalOrdinal
diff --git a/packages/muelu/src/Utils/MueLu_VisualizationHelpers_decl.hpp b/packages/muelu/src/Utils/MueLu_VisualizationHelpers_decl.hpp
index 321d838c5925..f6dab2b2df66 100644
--- a/packages/muelu/src/Utils/MueLu_VisualizationHelpers_decl.hpp
+++ b/packages/muelu/src/Utils/MueLu_VisualizationHelpers_decl.hpp
@@ -47,7 +47,6 @@
 #ifndef MUELU_VISUALIZATIONHELPERS_DECL_HPP_
 #define MUELU_VISUALIZATIONHELPERS_DECL_HPP_
 
-
 #include "MueLu_ConfigDefs.hpp"
 #include "MueLu_VisualizationHelpers_fwd.hpp"
 #include "MueLu_GraphBase.hpp"
@@ -56,121 +55,132 @@
 
 namespace MueLu {
 
-  class Level;
-  //Utility classes used in convex hull algorithm
-
-  class myTriangle
-  {
-    public:
-      myTriangle() : v1(0), v2(0), v3(0) {}
-      myTriangle(int v1in, int v2in, int v3in) : v1(v1in), v2(v2in), v3(v3in) {}
-      ~myTriangle() {}
-      bool operator==(const myTriangle& l)
-      {
-        if(l.v1 == v1 && l.v2 == v2 && l.v3 == v3)
-          return true;
-        return false;
-      }
-      int v1;
-      int v2;
-      int v3;
-  };
-
-  class myVec3
-  {
-    public:
-      myVec3() : x(0), y(0), z(0) {}
-      myVec3(double xin, double yin, double zin) : x(xin), y(yin), z(zin) {}
-      ~myVec3() {}
-      double x;
-      double y;
-      double z;
-  };
-
-  class myVec2
-  {
-    public:
-      myVec2() : x(0), y(0) {}
-      myVec2(double xin, double yin) : x(xin), y(yin) {}
-      ~myVec2() {}
-      double x;
-      double y;
-  };
-
-  /*!
-    @class VisualizationHelpers class.
-    @brief Base class providing routines to visualize aggregates and coarsening information.
-
-    This class is the base class for the CoarseningVisualizationFactory as well as the AggregationExporterFactory to
-    visualize aggregates or coarsening information from the transfer operators.
-
-  */
-
-  template <class Scalar = DefaultScalar,
-            class LocalOrdinal = DefaultLocalOrdinal,
-            class GlobalOrdinal = DefaultGlobalOrdinal,
-            class Node = DefaultNode>
-  class VisualizationHelpers {
+class Level;
+// Utility classes used in convex hull algorithm
+
+class myTriangle {
+ public:
+  myTriangle()
+    : v1(0)
+    , v2(0)
+    , v3(0) {}
+  myTriangle(int v1in, int v2in, int v3in)
+    : v1(v1in)
+    , v2(v2in)
+    , v3(v3in) {}
+  ~myTriangle() {}
+  bool operator==(const myTriangle& l) {
+    if (l.v1 == v1 && l.v2 == v2 && l.v3 == v3)
+      return true;
+    return false;
+  }
+  int v1;
+  int v2;
+  int v3;
+};
+
+class myVec3 {
+ public:
+  myVec3()
+    : x(0)
+    , y(0)
+    , z(0) {}
+  myVec3(double xin, double yin, double zin)
+    : x(xin)
+    , y(yin)
+    , z(zin) {}
+  ~myVec3() {}
+  double x;
+  double y;
+  double z;
+};
+
+class myVec2 {
+ public:
+  myVec2()
+    : x(0)
+    , y(0) {}
+  myVec2(double xin, double yin)
+    : x(xin)
+    , y(yin) {}
+  ~myVec2() {}
+  double x;
+  double y;
+};
+
+/*!
+  @class VisualizationHelpers class.
+  @brief Base class providing routines to visualize aggregates and coarsening information.
+
+  This class is the base class for the CoarseningVisualizationFactory as well as the AggregationExporterFactory to
+  visualize aggregates or coarsening information from the transfer operators.
+
+*/
+
+template <class Scalar        = DefaultScalar,
+          class LocalOrdinal  = DefaultLocalOrdinal,
+          class GlobalOrdinal = DefaultGlobalOrdinal,
+          class Node          = DefaultNode>
+class VisualizationHelpers {
 #undef MUELU_VISUALIZATIONHELPERS_SHORT
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor.
-    VisualizationHelpers() { }
-
-    //! Destructor.
-    virtual ~VisualizationHelpers() { }
-    //@}
-
-    RCP<ParameterList> GetValidParameterList() const;
-
-  protected:
-
-    void writeFileVTKOpening(std::ofstream & fout, std::vector<int> & uniqueFine, std::vector<int> & geomSizesFine) const;
-    void writeFileVTKNodes(std::ofstream & fout, std::vector<int> & uniqueFine, Teuchos::RCP<const Map> & nodeMap) const;
-    void writeFileVTKData(std::ofstream & fout, std::vector<int> & uniqueFine, LocalOrdinal myAggOffset, ArrayRCP<LocalOrdinal> & vertex2AggId, int myRank) const;
-    void writeFileVTKCoordinates(std::ofstream & fout, std::vector<int> & uniqueFine, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fz, int dim) const;
-    void writeFileVTKCells(std::ofstream & fout, std::vector<int> & uniqueFine, std::vector<LocalOrdinal> & vertices, std::vector<LocalOrdinal> & geomSize) const;
-    void writeFileVTKClosing(std::ofstream & fout) const;
-    void writePVTU(std::ofstream& pvtu, std::string baseFname, int numProcs, bool bFineEdges = false, bool bCoarseEdges = false) const;
-    void buildColormap() const;
-
-    std::string getFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList& pL) const;
-    std::string getBaseFileName(int numProcs, int level, const Teuchos::ParameterList& pL) const;
-    std::string getPVTUFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList& pL) const;
-
-    // move these routines to a common base class for visualization factories?
-    static void doPointCloud(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes);
-    static void doJacks(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId);
-    static void doConvexHulls2D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords);
-    static void doConvexHulls3D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords);
-
-    static void doGraphEdges(std::vector<int>& vertices, std::vector<int>& geomSizes, Teuchos::RCP<GraphBase>& G, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fz);
-
-    static int ccw(const myVec2& a, const myVec2& b, const myVec2& c);
-    static myVec3 crossProduct(myVec3 v1, myVec3 v2);
-    static double dotProduct(myVec2 v1, myVec2 v2);
-    static double dotProduct(myVec3 v1, myVec3 v2);
-    static bool isInFront(myVec3 point, myVec3 inPlane, myVec3 n);
-    static double mymagnitude(myVec2 vec);
-    static double mymagnitude(myVec3 vec);
-    static double distance(myVec2 p1, myVec2 p2);
-    static double distance(myVec3 p1, myVec3 p2);
-    static myVec2 vecSubtract(myVec2 v1, myVec2 v2);
-    static myVec3 vecSubtract(myVec3 v1, myVec3 v2);
-    static myVec2 getNorm(myVec2 v);
-    static myVec3 getNorm(myVec3 v1, myVec3 v2, myVec3 v3);
-    static double pointDistFromTri(myVec3 point, myVec3 v1, myVec3 v2, myVec3 v3);
-    static std::vector<myTriangle> processTriangle(std::list<myTriangle>& tris, myTriangle tri, std::list<int>& pointsInFront, myVec3& barycenter, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords);
-    static std::vector<int> giftWrap(std::vector<myVec2>& points, std::vector<int>& nodes, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & yCoords);
-
-    std::string replaceAll(std::string result, const std::string& replaceWhat, const std::string& replaceWithWhat) const;
-    std::vector<int> makeUnique(std::vector<int>& vertices) const; //!< replaces node indices in vertices with compressed unique indices, and returns list of unique points
-  }; // class VisualizationHelpers
-} // namespace MueLu
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor.
+  VisualizationHelpers() {}
+
+  //! Destructor.
+  virtual ~VisualizationHelpers() {}
+  //@}
+
+  RCP<ParameterList> GetValidParameterList() const;
+
+ protected:
+  void writeFileVTKOpening(std::ofstream& fout, std::vector<int>& uniqueFine, std::vector<int>& geomSizesFine) const;
+  void writeFileVTKNodes(std::ofstream& fout, std::vector<int>& uniqueFine, Teuchos::RCP<const Map>& nodeMap) const;
+  void writeFileVTKData(std::ofstream& fout, std::vector<int>& uniqueFine, LocalOrdinal myAggOffset, ArrayRCP<LocalOrdinal>& vertex2AggId, int myRank) const;
+  void writeFileVTKCoordinates(std::ofstream& fout, std::vector<int>& uniqueFine, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fz, int dim) const;
+  void writeFileVTKCells(std::ofstream& fout, std::vector<int>& uniqueFine, std::vector<LocalOrdinal>& vertices, std::vector<LocalOrdinal>& geomSize) const;
+  void writeFileVTKClosing(std::ofstream& fout) const;
+  void writePVTU(std::ofstream& pvtu, std::string baseFname, int numProcs, bool bFineEdges = false, bool bCoarseEdges = false) const;
+  void buildColormap() const;
+
+  std::string getFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList& pL) const;
+  std::string getBaseFileName(int numProcs, int level, const Teuchos::ParameterList& pL) const;
+  std::string getPVTUFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList& pL) const;
+
+  // move these routines to a common base class for visualization factories?
+  static void doPointCloud(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes);
+  static void doJacks(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId);
+  static void doConvexHulls2D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords);
+  static void doConvexHulls3D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords);
+
+  static void doGraphEdges(std::vector<int>& vertices, std::vector<int>& geomSizes, Teuchos::RCP<GraphBase>& G, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fz);
+
+  static int ccw(const myVec2& a, const myVec2& b, const myVec2& c);
+  static myVec3 crossProduct(myVec3 v1, myVec3 v2);
+  static double dotProduct(myVec2 v1, myVec2 v2);
+  static double dotProduct(myVec3 v1, myVec3 v2);
+  static bool isInFront(myVec3 point, myVec3 inPlane, myVec3 n);
+  static double mymagnitude(myVec2 vec);
+  static double mymagnitude(myVec3 vec);
+  static double distance(myVec2 p1, myVec2 p2);
+  static double distance(myVec3 p1, myVec3 p2);
+  static myVec2 vecSubtract(myVec2 v1, myVec2 v2);
+  static myVec3 vecSubtract(myVec3 v1, myVec3 v2);
+  static myVec2 getNorm(myVec2 v);
+  static myVec3 getNorm(myVec3 v1, myVec3 v2, myVec3 v3);
+  static double pointDistFromTri(myVec3 point, myVec3 v1, myVec3 v2, myVec3 v3);
+  static std::vector<myTriangle> processTriangle(std::list<myTriangle>& tris, myTriangle tri, std::list<int>& pointsInFront, myVec3& barycenter, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords);
+  static std::vector<int> giftWrap(std::vector<myVec2>& points, std::vector<int>& nodes, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords);
+
+  std::string replaceAll(std::string result, const std::string& replaceWhat, const std::string& replaceWithWhat) const;
+  std::vector<int> makeUnique(std::vector<int>& vertices) const;  //!< replaces node indices in vertices with compressed unique indices, and returns list of unique points
+};                                                                // class VisualizationHelpers
+}  // namespace MueLu
 
 #define MUELU_VISUALIZATIONHELPERS_SHORT
 
diff --git a/packages/muelu/src/Utils/MueLu_VisualizationHelpers_def.hpp b/packages/muelu/src/Utils/MueLu_VisualizationHelpers_def.hpp
index eed4bb7c1fa0..62c0fb8b3f55 100644
--- a/packages/muelu/src/Utils/MueLu_VisualizationHelpers_def.hpp
+++ b/packages/muelu/src/Utils/MueLu_VisualizationHelpers_def.hpp
@@ -59,1193 +59,1135 @@
 #include <algorithm>
 #include <string>
 
-
 namespace MueLu {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<ParameterList> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
-    RCP<ParameterList> validParamList = rcp(new ParameterList());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<ParameterList> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetValidParameterList() const {
+  RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-    validParamList->set< std::string >           ("visualization: output filename",                    "viz%LEVELID",                    "filename for VTK-style visualization output");
-    validParamList->set< int >                   ("visualization: output file: time step",             0,                     "time step variable for output file name");// Remove me?
-    validParamList->set< int >                   ("visualization: output file: iter",                  0,                     "nonlinear iteration variable for output file name");//Remove me?
-    validParamList->set<std::string>             ("visualization: style", "Point Cloud", "style of aggregate visualization for VTK output. Can be 'Point Cloud', 'Jacks', 'Convex Hulls'");
-    validParamList->set<bool>                    ("visualization: build colormap",        false,       "Whether to build a random color map in a separate xml file.");
-    validParamList->set<bool>                    ("visualization: fine graph edges",      false,                 "Whether to draw all fine node connections along with the aggregates.");
+  validParamList->set<std::string>("visualization: output filename", "viz%LEVELID", "filename for VTK-style visualization output");
+  validParamList->set<int>("visualization: output file: time step", 0, "time step variable for output file name");       // Remove me?
+  validParamList->set<int>("visualization: output file: iter", 0, "nonlinear iteration variable for output file name");  // Remove me?
+  validParamList->set<std::string>("visualization: style", "Point Cloud", "style of aggregate visualization for VTK output. Can be 'Point Cloud', 'Jacks', 'Convex Hulls'");
+  validParamList->set<bool>("visualization: build colormap", false, "Whether to build a random color map in a separate xml file.");
+  validParamList->set<bool>("visualization: fine graph edges", false, "Whether to draw all fine node connections along with the aggregates.");
 
-    return validParamList;
-  }
+  return validParamList;
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doPointCloud(std::vector<int>& vertices, std::vector<int>& geomSizes, LO /* numLocalAggs */, LO numFineNodes) {
-    vertices.reserve(numFineNodes);
-    geomSizes.reserve(numFineNodes);
-    for(LocalOrdinal i = 0; i < numFineNodes; i++)
-    {
-      vertices.push_back(i);
-      geomSizes.push_back(1);
-    }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doPointCloud(std::vector<int>& vertices, std::vector<int>& geomSizes, LO /* numLocalAggs */, LO numFineNodes) {
+  vertices.reserve(numFineNodes);
+  geomSizes.reserve(numFineNodes);
+  for (LocalOrdinal i = 0; i < numFineNodes; i++) {
+    vertices.push_back(i);
+    geomSizes.push_back(1);
   }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doJacks(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId) {
-    //For each aggregate, find the root node then connect it with the other nodes in the aggregate
-    //Doesn't matter the order, as long as all edges are found.
-    vertices.reserve(vertices.size() + 3 * (numFineNodes - numLocalAggs));
-    geomSizes.reserve(vertices.size() + 2 * (numFineNodes - numLocalAggs));
-    int root = 0;
-    for(int i = 0; i < numLocalAggs; i++) //TODO: Replace this O(n^2) with a better way
-    {
-      while(!isRoot[root])
-        root++;
-      int numInAggFound = 0;
-      for(int j = 0; j < numFineNodes; j++)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doJacks(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& isRoot, const ArrayRCP<LO>& vertex2AggId) {
+  // For each aggregate, find the root node then connect it with the other nodes in the aggregate
+  // Doesn't matter the order, as long as all edges are found.
+  vertices.reserve(vertices.size() + 3 * (numFineNodes - numLocalAggs));
+  geomSizes.reserve(vertices.size() + 2 * (numFineNodes - numLocalAggs));
+  int root = 0;
+  for (int i = 0; i < numLocalAggs; i++)  // TODO: Replace this O(n^2) with a better way
+  {
+    while (!isRoot[root])
+      root++;
+    int numInAggFound = 0;
+    for (int j = 0; j < numFineNodes; j++) {
+      if (j == root)  // don't make a connection from the root to itself
       {
-        if(j == root) //don't make a connection from the root to itself
-        {
-          numInAggFound++;
-          continue;
-        }
-        if(vertex2AggId[root] == vertex2AggId[j])
-        {
-          vertices.push_back(root);
-          vertices.push_back(j);
-          geomSizes.push_back(2);
-          //Also draw the free endpoint explicitly for the current line
-          vertices.push_back(j);
-          geomSizes.push_back(1);
-          numInAggFound++;
-          //if(numInAggFound == aggSizes_[vertex2AggId_[root]]) //don't spend more time looking if done with that root
-          //  break;
-        }
+        numInAggFound++;
+        continue;
+      }
+      if (vertex2AggId[root] == vertex2AggId[j]) {
+        vertices.push_back(root);
+        vertices.push_back(j);
+        geomSizes.push_back(2);
+        // Also draw the free endpoint explicitly for the current line
+        vertices.push_back(j);
+        geomSizes.push_back(1);
+        numInAggFound++;
+        // if(numInAggFound == aggSizes_[vertex2AggId_[root]]) //don't spend more time looking if done with that root
+        //   break;
       }
-      root++; //get set up to look for the next root
     }
+    root++;  // get set up to look for the next root
   }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls2D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& /* isRoot */, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls2D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& /* isRoot */, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords) {
+  // This algorithm is based on Andrew's Monotone Chain variant of the Graham Scan for Convex Hulls.  It adds
+  // a colinearity check which we'll need for our viz.
+  for (int agg = 0; agg < numLocalAggs; agg++) {
+    std::vector<int> aggNodes;
+    for (int i = 0; i < numFineNodes; i++) {
+      if (vertex2AggId[i] == agg)
+        aggNodes.push_back(i);
+    }
 
-    // This algorithm is based on Andrew's Monotone Chain variant of the Graham Scan for Convex Hulls.  It adds
-    // a colinearity check which we'll need for our viz.
-    for(int agg = 0; agg < numLocalAggs; agg++) {
-      std::vector<int> aggNodes;
-      for(int i = 0; i < numFineNodes; i++) {
-        if(vertex2AggId[i] == agg)
-          aggNodes.push_back(i);
+    // have a list of nodes in the aggregate
+    TEUCHOS_TEST_FOR_EXCEPTION(aggNodes.size() == 0, Exceptions::RuntimeError,
+                               "CoarseningVisualization::doConvexHulls2D: aggregate contains zero nodes!");
+    if (aggNodes.size() == 1) {
+      vertices.push_back(aggNodes.front());
+      geomSizes.push_back(1);
+      continue;
+    }
+    if (aggNodes.size() == 2) {
+      vertices.push_back(aggNodes.front());
+      vertices.push_back(aggNodes.back());
+      geomSizes.push_back(2);
+      continue;
+    } else {
+      int N        = (int)aggNodes.size();
+      using MyPair = std::pair<myVec2, int>;
+      std::vector<MyPair> pointsAndIndex(N);
+      for (int i = 0; i < N; i++) {
+        pointsAndIndex[i] = std::make_pair(myVec2(xCoords[aggNodes[i]], yCoords[aggNodes[i]]), i);
       }
 
-      //have a list of nodes in the aggregate
-      TEUCHOS_TEST_FOR_EXCEPTION(aggNodes.size() == 0, Exceptions::RuntimeError,
-               "CoarseningVisualization::doConvexHulls2D: aggregate contains zero nodes!");
-      if(aggNodes.size() == 1) {
-        vertices.push_back(aggNodes.front());
-        geomSizes.push_back(1);
-        continue;
+      // Sort by x coordinate
+      std::sort(pointsAndIndex.begin(), pointsAndIndex.end(), [](const MyPair& a, const MyPair& b) {
+        return a.first.x < b.first.x || (a.first.x == b.first.x && a.first.y < b.first.y);
+      });
+
+      // Colinearity check
+      bool colinear = true;
+      for (int i = 0; i < N; i++) {
+        if (ccw(pointsAndIndex[i].first, pointsAndIndex[(i + 1) % N].first, pointsAndIndex[(i + 2) % N].first) != 0) {
+          colinear = false;
+          break;
+        }
       }
-      if(aggNodes.size() == 2) {
-        vertices.push_back(aggNodes.front());
-        vertices.push_back(aggNodes.back());
+
+      if (colinear) {
+        vertices.push_back(aggNodes[pointsAndIndex.front().second]);
+        vertices.push_back(aggNodes[pointsAndIndex.back().second]);
         geomSizes.push_back(2);
-        continue;
-      }
-      else {	
-	int N = (int) aggNodes.size();
-	using MyPair = std::pair<myVec2,int>;
-	std::vector<MyPair> pointsAndIndex(N);
-	for(int i=0; i<N; i++) {
-	  pointsAndIndex[i] = std::make_pair(myVec2(xCoords[aggNodes[i]], yCoords[aggNodes[i]]),i);
-	}
+      } else {
+        std::vector<int> hull(2 * N);
+        int count = 0;
 
-	// Sort by x coordinate
-	std::sort(pointsAndIndex.begin(),pointsAndIndex.end(),[](const MyPair &a, const MyPair &b) {
-	    return a.first.x < b.first.x || (a.first.x == b.first.x && a.first.y < b.first.y);
-	  });
+        // Build lower hull
+        for (int i = 0; i < N; i++) {
+          while (count >= 2 && ccw(pointsAndIndex[hull[count - 2]].first, pointsAndIndex[hull[count - 1]].first, pointsAndIndex[i].first) <= 0) {
+            count--;
+          }
+          hull[count++] = i;
+        }
 
-	
-	// Colinearity check
-	bool colinear=true;
-	for(int i=0; i<N; i++) {
-	  if(ccw(pointsAndIndex[i].first,pointsAndIndex[(i+1)%N].first,pointsAndIndex[(i+2)%N].first)!=0) {
-	    colinear=false;
-	    break;
-	  }
-	}
+        // Build the upper hull
+        int t = count + 1;
+        for (int i = N - 1; i > 0; i--) {
+          while (count >= t && ccw(pointsAndIndex[hull[count - 2]].first, pointsAndIndex[hull[count - 1]].first, pointsAndIndex[i - 1].first) <= 0) {
+            count--;
+          }
+          hull[count++] = i - 1;
+        }
 
-	if(colinear) {
-	  vertices.push_back(aggNodes[pointsAndIndex.front().second]);
-	  vertices.push_back(aggNodes[pointsAndIndex.back().second]);
-	  geomSizes.push_back(2);
-	}
-	else {
-	  std::vector<int> hull(2*N);
-	  int count=0;
-	  
-	  // Build lower hull
-	  for(int i=0; i<N; i++) {
-	    while (count >=2 && ccw(pointsAndIndex[hull[count-2]].first,pointsAndIndex[hull[count-1]].first,pointsAndIndex[i].first)<=0) {
-	      count--;
-	    }
-	    hull[count++] = i;
-	  }
-	  
-	  // Build the upper hull
-	  int t=count+1;
-	  for(int i=N-1; i>0; i--) {
-	    while(count >= t && ccw(pointsAndIndex[hull[count-2]].first,pointsAndIndex[hull[count-1]].first,pointsAndIndex[i-1].first)<=0) {
-	      count--;
-	    }
-	    hull[count++]=i-1;
-	  }
-	  
-	  // Remove the duplicated point
-	  hull.resize(count-1);
+        // Remove the duplicated point
+        hull.resize(count - 1);
 
-	  // Verify: Check that hull retains CCW order
-	  for(int i=0; i<(int)hull.size(); i++) {
-            TEUCHOS_TEST_FOR_EXCEPTION(ccw(pointsAndIndex[hull[i]].first,pointsAndIndex[hull[(i+1)%hull.size()]].first,pointsAndIndex[hull[(i+1)%hull.size()]].first) == 1, Exceptions::RuntimeError,"CoarseningVisualization::doConvexHulls2D: counter-clockwise verification fails");
-	  }
+        // Verify: Check that hull retains CCW order
+        for (int i = 0; i < (int)hull.size(); i++) {
+          TEUCHOS_TEST_FOR_EXCEPTION(ccw(pointsAndIndex[hull[i]].first, pointsAndIndex[hull[(i + 1) % hull.size()]].first, pointsAndIndex[hull[(i + 1) % hull.size()]].first) == 1, Exceptions::RuntimeError, "CoarseningVisualization::doConvexHulls2D: counter-clockwise verification fails");
+        }
 
-	  // We now need to undo the indices from the sorting
-	  for(int i=0; i<(int)hull.size(); i++) {
-	    hull[i] = aggNodes[pointsAndIndex[hull[i]].second];
-	  }
+        // We now need to undo the indices from the sorting
+        for (int i = 0; i < (int)hull.size(); i++) {
+          hull[i] = aggNodes[pointsAndIndex[hull[i]].second];
+        }
 
-	  vertices.reserve(vertices.size() + hull.size());
-	  for(size_t i = 0; i < hull.size(); i++) {
-	    vertices.push_back(hull[i]);
-	  }
-	  geomSizes.push_back(hull.size());
-	}//else colinear
-      }//else 3 + nodes
-    }
+        vertices.reserve(vertices.size() + hull.size());
+        for (size_t i = 0; i < hull.size(); i++) {
+          vertices.push_back(hull[i]);
+        }
+        geomSizes.push_back(hull.size());
+      }  // else colinear
+    }    // else 3 + nodes
   }
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls3D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& /* isRoot */, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords) {
-    //Use 3D quickhull algo.
-    //Vector of node indices representing triangle vertices
-    //Note: Calculate the hulls first since will only include point data for points in the hulls
-    //Effectively the size() of vertIndices after each hull is added to it
-    typedef std::list<int>::iterator Iter;
-    for(int agg = 0; agg < numLocalAggs; agg++) {
-      std::list<int> aggNodes; //At first, list of all nodes in the aggregate. As nodes are enclosed or included by/in hull, remove them
-      for(int i = 0; i < numFineNodes; i++) {
-        if(vertex2AggId[i] == agg)
-          aggNodes.push_back(i);
-      }
-      //First, check anomalous cases
-      TEUCHOS_TEST_FOR_EXCEPTION(aggNodes.size() == 0, Exceptions::RuntimeError,
-               "CoarseningVisualization::doConvexHulls3D: aggregate contains zero nodes!");
-      if(aggNodes.size() == 1) {
-        vertices.push_back(aggNodes.front());
-        geomSizes.push_back(1);
-        continue;
-      } else if(aggNodes.size() == 2) {
-        vertices.push_back(aggNodes.front());
-        vertices.push_back(aggNodes.back());
-        geomSizes.push_back(2);
-        continue;
-      }
-      //check for collinearity
-      bool areCollinear = true;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doConvexHulls3D(std::vector<int>& vertices, std::vector<int>& geomSizes, LO numLocalAggs, LO numFineNodes, const std::vector<bool>& /* isRoot */, const ArrayRCP<LO>& vertex2AggId, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords) {
+  // Use 3D quickhull algo.
+  // Vector of node indices representing triangle vertices
+  // Note: Calculate the hulls first since will only include point data for points in the hulls
+  // Effectively the size() of vertIndices after each hull is added to it
+  typedef std::list<int>::iterator Iter;
+  for (int agg = 0; agg < numLocalAggs; agg++) {
+    std::list<int> aggNodes;  // At first, list of all nodes in the aggregate. As nodes are enclosed or included by/in hull, remove them
+    for (int i = 0; i < numFineNodes; i++) {
+      if (vertex2AggId[i] == agg)
+        aggNodes.push_back(i);
+    }
+    // First, check anomalous cases
+    TEUCHOS_TEST_FOR_EXCEPTION(aggNodes.size() == 0, Exceptions::RuntimeError,
+                               "CoarseningVisualization::doConvexHulls3D: aggregate contains zero nodes!");
+    if (aggNodes.size() == 1) {
+      vertices.push_back(aggNodes.front());
+      geomSizes.push_back(1);
+      continue;
+    } else if (aggNodes.size() == 2) {
+      vertices.push_back(aggNodes.front());
+      vertices.push_back(aggNodes.back());
+      geomSizes.push_back(2);
+      continue;
+    }
+    // check for collinearity
+    bool areCollinear = true;
+    {
+      Iter it = aggNodes.begin();
+      myVec3 firstVec(xCoords[*it], yCoords[*it], zCoords[*it]);
+      myVec3 comp;
       {
-        Iter it = aggNodes.begin();
-        myVec3 firstVec(xCoords[*it], yCoords[*it], zCoords[*it]);
-        myVec3 comp;
-        {
-          it++;
-          myVec3 secondVec(xCoords[*it], yCoords[*it], zCoords[*it]); //cross this with other vectors to compare
-          comp = vecSubtract(secondVec, firstVec);
-          it++;
-        }
-        for(; it != aggNodes.end(); it++) {
-          myVec3 thisVec(xCoords[*it], yCoords[*it], zCoords[*it]);
-          myVec3 cross = crossProduct(vecSubtract(thisVec, firstVec), comp);
-          if(mymagnitude(cross) > 1e-10) {
-            areCollinear = false;
-            break;
-          }
+        it++;
+        myVec3 secondVec(xCoords[*it], yCoords[*it], zCoords[*it]);  // cross this with other vectors to compare
+        comp = vecSubtract(secondVec, firstVec);
+        it++;
+      }
+      for (; it != aggNodes.end(); it++) {
+        myVec3 thisVec(xCoords[*it], yCoords[*it], zCoords[*it]);
+        myVec3 cross = crossProduct(vecSubtract(thisVec, firstVec), comp);
+        if (mymagnitude(cross) > 1e-10) {
+          areCollinear = false;
+          break;
         }
       }
-      if(areCollinear)
-      {
-        //find the endpoints of segment describing all the points
-        //compare x, if tie compare y, if tie compare z
-        Iter min = aggNodes.begin();
-        Iter max = aggNodes.begin();
-        Iter it = ++aggNodes.begin();
-        for(; it != aggNodes.end(); it++) {
-          if(xCoords[*it] < xCoords[*min]) min = it;
-          else if(xCoords[*it] == xCoords[*min]) {
-            if(yCoords[*it] < yCoords[*min]) min = it;
-            else if(yCoords[*it] == yCoords[*min]) {
-              if(zCoords[*it] < zCoords[*min]) min = it;
-            }
+    }
+    if (areCollinear) {
+      // find the endpoints of segment describing all the points
+      // compare x, if tie compare y, if tie compare z
+      Iter min = aggNodes.begin();
+      Iter max = aggNodes.begin();
+      Iter it  = ++aggNodes.begin();
+      for (; it != aggNodes.end(); it++) {
+        if (xCoords[*it] < xCoords[*min])
+          min = it;
+        else if (xCoords[*it] == xCoords[*min]) {
+          if (yCoords[*it] < yCoords[*min])
+            min = it;
+          else if (yCoords[*it] == yCoords[*min]) {
+            if (zCoords[*it] < zCoords[*min]) min = it;
           }
-          if(xCoords[*it] > xCoords[*max]) max = it;
-          else if(xCoords[*it] == xCoords[*max]) {
-            if(yCoords[*it] > yCoords[*max]) max = it;
-            else if(yCoords[*it] == yCoords[*max]) {
-              if(zCoords[*it] > zCoords[*max])
-                max = it;
-            }
+        }
+        if (xCoords[*it] > xCoords[*max])
+          max = it;
+        else if (xCoords[*it] == xCoords[*max]) {
+          if (yCoords[*it] > yCoords[*max])
+            max = it;
+          else if (yCoords[*it] == yCoords[*max]) {
+            if (zCoords[*it] > zCoords[*max])
+              max = it;
           }
         }
-        vertices.push_back(*min);
-        vertices.push_back(*max);
-        geomSizes.push_back(2);
-        continue;
       }
-      bool areCoplanar = true;
-      {
-        //number of points is known to be >= 3
-        Iter vert = aggNodes.begin();
-        myVec3 v1(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
-        vert++;
-        myVec3 v2(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
-        vert++;
-        myVec3 v3(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
+      vertices.push_back(*min);
+      vertices.push_back(*max);
+      geomSizes.push_back(2);
+      continue;
+    }
+    bool areCoplanar = true;
+    {
+      // number of points is known to be >= 3
+      Iter vert = aggNodes.begin();
+      myVec3 v1(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
+      vert++;
+      myVec3 v2(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
+      vert++;
+      myVec3 v3(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
+      vert++;
+      // Make sure the first three points aren't also collinear (need a non-degenerate triangle to get a normal)
+      while (mymagnitude(crossProduct(vecSubtract(v1, v2), vecSubtract(v1, v3))) < 1e-10) {
+        // Replace the third point with the next point
+        v3 = myVec3(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
         vert++;
-        //Make sure the first three points aren't also collinear (need a non-degenerate triangle to get a normal)
-        while(mymagnitude(crossProduct(vecSubtract(v1, v2), vecSubtract(v1, v3))) < 1e-10) {
-          //Replace the third point with the next point
-          v3 = myVec3(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
-          vert++;
+      }
+      for (; vert != aggNodes.end(); vert++) {
+        myVec3 pt(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
+        if (fabs(pointDistFromTri(pt, v1, v2, v3)) > 1e-12) {
+          areCoplanar = false;
+          break;
         }
-        for(; vert != aggNodes.end(); vert++) {
-          myVec3 pt(xCoords[*vert], yCoords[*vert], zCoords[*vert]);
-          if(fabs(pointDistFromTri(pt, v1, v2, v3)) > 1e-12) {
-            areCoplanar = false;
-            break;
+      }
+      if (areCoplanar) {
+        // do 2D convex hull
+        myVec3 planeNorm = getNorm(v1, v2, v3);
+        planeNorm.x      = fabs(planeNorm.x);
+        planeNorm.y      = fabs(planeNorm.y);
+        planeNorm.z      = fabs(planeNorm.z);
+        std::vector<myVec2> points;
+        std::vector<int> nodes;
+        if (planeNorm.x >= planeNorm.y && planeNorm.x >= planeNorm.z) {
+          // project points to yz plane to make hull
+          for (Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
+            nodes.push_back(*it);
+            points.push_back(myVec2(yCoords[*it], zCoords[*it]));
           }
         }
-        if(areCoplanar) {
-          //do 2D convex hull
-          myVec3 planeNorm = getNorm(v1, v2, v3);
-          planeNorm.x = fabs(planeNorm.x);
-          planeNorm.y = fabs(planeNorm.y);
-          planeNorm.z = fabs(planeNorm.z);
-          std::vector<myVec2> points;
-          std::vector<int> nodes;
-          if(planeNorm.x >= planeNorm.y && planeNorm.x >= planeNorm.z) {
-            //project points to yz plane to make hull
-            for(Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
-              nodes.push_back(*it);
-              points.push_back(myVec2(yCoords[*it], zCoords[*it]));
-            }
+        if (planeNorm.y >= planeNorm.x && planeNorm.y >= planeNorm.z) {
+          // use xz
+          for (Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
+            nodes.push_back(*it);
+            points.push_back(myVec2(xCoords[*it], zCoords[*it]));
           }
-          if(planeNorm.y >= planeNorm.x && planeNorm.y >= planeNorm.z) {
-            //use xz
-            for(Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
-              nodes.push_back(*it);
-              points.push_back(myVec2(xCoords[*it], zCoords[*it]));
-            }
-          }
-          if(planeNorm.z >= planeNorm.x && planeNorm.z >= planeNorm.y) {
-            for(Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
-              nodes.push_back(*it);
-              points.push_back(myVec2(xCoords[*it], yCoords[*it]));
-            }
+        }
+        if (planeNorm.z >= planeNorm.x && planeNorm.z >= planeNorm.y) {
+          for (Iter it = aggNodes.begin(); it != aggNodes.end(); it++) {
+            nodes.push_back(*it);
+            points.push_back(myVec2(xCoords[*it], yCoords[*it]));
           }
-          std::vector<int> convhull2d = giftWrap(points, nodes, xCoords, yCoords);
-          geomSizes.push_back(convhull2d.size());
-          vertices.reserve(vertices.size() + convhull2d.size());
-          for(size_t i = 0; i < convhull2d.size(); i++)
-            vertices.push_back(convhull2d[i]);
-          continue;
         }
+        std::vector<int> convhull2d = giftWrap(points, nodes, xCoords, yCoords);
+        geomSizes.push_back(convhull2d.size());
+        vertices.reserve(vertices.size() + convhull2d.size());
+        for (size_t i = 0; i < convhull2d.size(); i++)
+          vertices.push_back(convhull2d[i]);
+        continue;
       }
-      Iter exIt = aggNodes.begin(); //iterator to be used for searching for min/max x/y/z
-      int extremeSix[] = {*exIt, *exIt, *exIt, *exIt, *exIt, *exIt}; //nodes with minimumX, maxX, minY ...
-      exIt++;
-      for(; exIt != aggNodes.end(); exIt++) {
-        Iter it = exIt;
-        if(xCoords[*it] < xCoords[extremeSix[0]] ||
+    }
+    Iter exIt        = aggNodes.begin();                            // iterator to be used for searching for min/max x/y/z
+    int extremeSix[] = {*exIt, *exIt, *exIt, *exIt, *exIt, *exIt};  // nodes with minimumX, maxX, minY ...
+    exIt++;
+    for (; exIt != aggNodes.end(); exIt++) {
+      Iter it = exIt;
+      if (xCoords[*it] < xCoords[extremeSix[0]] ||
           (xCoords[*it] == xCoords[extremeSix[0]] && yCoords[*it] < yCoords[extremeSix[0]]) ||
           (xCoords[*it] == xCoords[extremeSix[0]] && yCoords[*it] == yCoords[extremeSix[0]] && zCoords[*it] < zCoords[extremeSix[0]]))
-            extremeSix[0] = *it;
-        if(xCoords[*it] > xCoords[extremeSix[1]] ||
+        extremeSix[0] = *it;
+      if (xCoords[*it] > xCoords[extremeSix[1]] ||
           (xCoords[*it] == xCoords[extremeSix[1]] && yCoords[*it] > yCoords[extremeSix[1]]) ||
           (xCoords[*it] == xCoords[extremeSix[1]] && yCoords[*it] == yCoords[extremeSix[1]] && zCoords[*it] > zCoords[extremeSix[1]]))
-            extremeSix[1] = *it;
-        if(yCoords[*it] < yCoords[extremeSix[2]] ||
+        extremeSix[1] = *it;
+      if (yCoords[*it] < yCoords[extremeSix[2]] ||
           (yCoords[*it] == yCoords[extremeSix[2]] && zCoords[*it] < zCoords[extremeSix[2]]) ||
           (yCoords[*it] == yCoords[extremeSix[2]] && zCoords[*it] == zCoords[extremeSix[2]] && xCoords[*it] < xCoords[extremeSix[2]]))
-            extremeSix[2] = *it;
-        if(yCoords[*it] > yCoords[extremeSix[3]] ||
+        extremeSix[2] = *it;
+      if (yCoords[*it] > yCoords[extremeSix[3]] ||
           (yCoords[*it] == yCoords[extremeSix[3]] && zCoords[*it] > zCoords[extremeSix[3]]) ||
           (yCoords[*it] == yCoords[extremeSix[3]] && zCoords[*it] == zCoords[extremeSix[3]] && xCoords[*it] > xCoords[extremeSix[3]]))
-            extremeSix[3] = *it;
-        if(zCoords[*it] < zCoords[extremeSix[4]] ||
+        extremeSix[3] = *it;
+      if (zCoords[*it] < zCoords[extremeSix[4]] ||
           (zCoords[*it] == zCoords[extremeSix[4]] && xCoords[*it] < xCoords[extremeSix[4]]) ||
           (zCoords[*it] == zCoords[extremeSix[4]] && xCoords[*it] == xCoords[extremeSix[4]] && yCoords[*it] < yCoords[extremeSix[4]]))
-            extremeSix[4] = *it;
-        if(zCoords[*it] > zCoords[extremeSix[5]] ||
+        extremeSix[4] = *it;
+      if (zCoords[*it] > zCoords[extremeSix[5]] ||
           (zCoords[*it] == zCoords[extremeSix[5]] && xCoords[*it] > xCoords[extremeSix[5]]) ||
           (zCoords[*it] == zCoords[extremeSix[5]] && xCoords[*it] == xCoords[extremeSix[5]] && yCoords[*it] > zCoords[extremeSix[5]]))
-            extremeSix[5] = *it;
-      }
-      myVec3 extremeVectors[6];
-      for(int i = 0; i < 6; i++) {
-        myVec3 thisExtremeVec(xCoords[extremeSix[i]], yCoords[extremeSix[i]], zCoords[extremeSix[i]]);
-        extremeVectors[i] = thisExtremeVec;
-      }
-      double maxDist = 0;
-      int base1 = 0; //ints from 0-5: which pair out of the 6 extreme points are the most distant? (indices in extremeSix and extremeVectors)
-      int base2 = 0;
-      for(int i = 0; i < 5; i++) {
-        for(int j = i + 1; j < 6; j++) {
-          double thisDist = distance(extremeVectors[i], extremeVectors[j]);
-          // Want to make sure thisDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
-          if(thisDist > maxDist && thisDist - maxDist > 1e-12) {
-            maxDist = thisDist;
-            base1 = i;
-            base2 = j;
-          }
-        }
-      }
-      std::list<myTriangle> hullBuilding;    //each Triangle is a triplet of nodes (int IDs) that form a triangle
-      //remove base1 and base2 iters from aggNodes, they are known to be in the hull
-      aggNodes.remove(extremeSix[base1]);
-      aggNodes.remove(extremeSix[base2]);
-      //extremeSix[base1] and [base2] still have the myVec3 representation
-      myTriangle tri;
-      tri.v1 = extremeSix[base1];
-      tri.v2 = extremeSix[base2];
-      //Now find the point that is furthest away from the line between base1 and base2
-      maxDist = 0;
-      //need the vectors to do "quadruple product" formula
-      myVec3 b1 = extremeVectors[base1];
-      myVec3 b2 = extremeVectors[base2];
-      Iter thirdNode;
-      for(Iter node = aggNodes.begin(); node != aggNodes.end(); node++) {
-        myVec3 nodePos(xCoords[*node], yCoords[*node], zCoords[*node]);
-        double dist = mymagnitude(crossProduct(vecSubtract(nodePos, b1), vecSubtract(nodePos, b2))) / mymagnitude(vecSubtract(b2, b1));
-        // Want to make sure dist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
-        if(dist > maxDist && dist - maxDist > 1e-12) {
-          maxDist = dist;
-          thirdNode = node;
+        extremeSix[5] = *it;
+    }
+    myVec3 extremeVectors[6];
+    for (int i = 0; i < 6; i++) {
+      myVec3 thisExtremeVec(xCoords[extremeSix[i]], yCoords[extremeSix[i]], zCoords[extremeSix[i]]);
+      extremeVectors[i] = thisExtremeVec;
+    }
+    double maxDist = 0;
+    int base1      = 0;  // ints from 0-5: which pair out of the 6 extreme points are the most distant? (indices in extremeSix and extremeVectors)
+    int base2      = 0;
+    for (int i = 0; i < 5; i++) {
+      for (int j = i + 1; j < 6; j++) {
+        double thisDist = distance(extremeVectors[i], extremeVectors[j]);
+        // Want to make sure thisDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
+        if (thisDist > maxDist && thisDist - maxDist > 1e-12) {
+          maxDist = thisDist;
+          base1   = i;
+          base2   = j;
         }
       }
-      //Now know the last node in the first triangle
-      tri.v3 = *thirdNode;
-      hullBuilding.push_back(tri);
-      myVec3 b3(xCoords[*thirdNode], yCoords[*thirdNode], zCoords[*thirdNode]);
-      aggNodes.erase(thirdNode);
-      //Find the fourth node (most distant from triangle) to form tetrahedron
-      maxDist = 0;
-      int fourthVertex = -1;
-      for(Iter node = aggNodes.begin(); node != aggNodes.end(); node++) {
-        myVec3 thisNode(xCoords[*node], yCoords[*node], zCoords[*node]);
-        double nodeDist = pointDistFromTri(thisNode, b1, b2, b3);
-        // Want to make sure nodeDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
-        if(nodeDist > maxDist && nodeDist - maxDist > 1e-12) {
-          maxDist = nodeDist;
-          fourthVertex = *node;
-        }
+    }
+    std::list<myTriangle> hullBuilding;  // each Triangle is a triplet of nodes (int IDs) that form a triangle
+    // remove base1 and base2 iters from aggNodes, they are known to be in the hull
+    aggNodes.remove(extremeSix[base1]);
+    aggNodes.remove(extremeSix[base2]);
+    // extremeSix[base1] and [base2] still have the myVec3 representation
+    myTriangle tri;
+    tri.v1 = extremeSix[base1];
+    tri.v2 = extremeSix[base2];
+    // Now find the point that is furthest away from the line between base1 and base2
+    maxDist = 0;
+    // need the vectors to do "quadruple product" formula
+    myVec3 b1 = extremeVectors[base1];
+    myVec3 b2 = extremeVectors[base2];
+    Iter thirdNode;
+    for (Iter node = aggNodes.begin(); node != aggNodes.end(); node++) {
+      myVec3 nodePos(xCoords[*node], yCoords[*node], zCoords[*node]);
+      double dist = mymagnitude(crossProduct(vecSubtract(nodePos, b1), vecSubtract(nodePos, b2))) / mymagnitude(vecSubtract(b2, b1));
+      // Want to make sure dist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
+      if (dist > maxDist && dist - maxDist > 1e-12) {
+        maxDist   = dist;
+        thirdNode = node;
       }
-      aggNodes.remove(fourthVertex);
-      myVec3 b4(xCoords[fourthVertex], yCoords[fourthVertex], zCoords[fourthVertex]);
-      //Add three new triangles to hullBuilding to form the first tetrahedron
-      //use tri to hold the triangle info temporarily before being added to list
-      tri = hullBuilding.front();
-      tri.v1 = fourthVertex;
-      hullBuilding.push_back(tri);
-      tri = hullBuilding.front();
-      tri.v2 = fourthVertex;
-      hullBuilding.push_back(tri);
-      tri = hullBuilding.front();
-      tri.v3 = fourthVertex;
-      hullBuilding.push_back(tri);
-      //now orient all four triangles so that the vertices are oriented clockwise (so getNorm_ points outward for each)
-      myVec3 barycenter((b1.x + b2.x + b3.x + b4.x) / 4.0, (b1.y + b2.y + b3.y + b4.y) / 4.0, (b1.z + b2.z + b3.z + b4.z) / 4.0);
-      for(std::list<myTriangle>::iterator tetTri = hullBuilding.begin(); tetTri != hullBuilding.end(); tetTri++) {
-        myVec3 triVert1(xCoords[tetTri->v1], yCoords[tetTri->v1], zCoords[tetTri->v1]);
-        myVec3 triVert2(xCoords[tetTri->v2], yCoords[tetTri->v2], zCoords[tetTri->v2]);
-        myVec3 triVert3(xCoords[tetTri->v3], yCoords[tetTri->v3], zCoords[tetTri->v3]);
-        myVec3 trinorm = getNorm(triVert1, triVert2, triVert3);
-        myVec3 ptInPlane = tetTri == hullBuilding.begin() ? b1 : b4; //first triangle definitely has b1 in it, other three definitely have b4
-        if(isInFront(barycenter, ptInPlane, trinorm)) {
-          //don't want the faces of the tetrahedron to face inwards (towards barycenter) so reverse orientation
-          //by swapping two vertices
-          int temp = tetTri->v1;
-          tetTri->v1 = tetTri->v2;
-          tetTri->v2 = temp;
-        }
+    }
+    // Now know the last node in the first triangle
+    tri.v3 = *thirdNode;
+    hullBuilding.push_back(tri);
+    myVec3 b3(xCoords[*thirdNode], yCoords[*thirdNode], zCoords[*thirdNode]);
+    aggNodes.erase(thirdNode);
+    // Find the fourth node (most distant from triangle) to form tetrahedron
+    maxDist          = 0;
+    int fourthVertex = -1;
+    for (Iter node = aggNodes.begin(); node != aggNodes.end(); node++) {
+      myVec3 thisNode(xCoords[*node], yCoords[*node], zCoords[*node]);
+      double nodeDist = pointDistFromTri(thisNode, b1, b2, b3);
+      // Want to make sure nodeDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
+      if (nodeDist > maxDist && nodeDist - maxDist > 1e-12) {
+        maxDist      = nodeDist;
+        fourthVertex = *node;
       }
-      //now, have starting polyhedron in hullBuilding (all faces are facing outwards according to getNorm_) and remaining nodes to process are in aggNodes
-      //recursively, for each triangle, make a list of the points that are 'in front' of the triangle. Find the point with the maximum distance from the triangle.
-      //Add three new triangles, each sharing one edge with the original triangle but now with the most distant point as a vertex. Remove the most distant point from
-      //the list of all points that need to be processed. Also from that list remove all points that are in front of the original triangle but not in front of all three
-      //new triangles, since they are now enclosed in the hull.
-      //Construct point lists for each face of the tetrahedron individually.
-      myVec3 trinorms[4]; //normals to the four tetrahedron faces, now oriented outwards
-      int index = 0;
-      for(std::list<myTriangle>::iterator tetTri = hullBuilding.begin(); tetTri != hullBuilding.end(); tetTri++) {
-        myVec3 triVert1(xCoords[tetTri->v1], yCoords[tetTri->v1], zCoords[tetTri->v1]);
-        myVec3 triVert2(xCoords[tetTri->v2], yCoords[tetTri->v2], zCoords[tetTri->v2]);
-        myVec3 triVert3(xCoords[tetTri->v3], yCoords[tetTri->v3], zCoords[tetTri->v3]);
-        trinorms[index] = getNorm(triVert1, triVert2, triVert3);
-        index++;
+    }
+    aggNodes.remove(fourthVertex);
+    myVec3 b4(xCoords[fourthVertex], yCoords[fourthVertex], zCoords[fourthVertex]);
+    // Add three new triangles to hullBuilding to form the first tetrahedron
+    // use tri to hold the triangle info temporarily before being added to list
+    tri    = hullBuilding.front();
+    tri.v1 = fourthVertex;
+    hullBuilding.push_back(tri);
+    tri    = hullBuilding.front();
+    tri.v2 = fourthVertex;
+    hullBuilding.push_back(tri);
+    tri    = hullBuilding.front();
+    tri.v3 = fourthVertex;
+    hullBuilding.push_back(tri);
+    // now orient all four triangles so that the vertices are oriented clockwise (so getNorm_ points outward for each)
+    myVec3 barycenter((b1.x + b2.x + b3.x + b4.x) / 4.0, (b1.y + b2.y + b3.y + b4.y) / 4.0, (b1.z + b2.z + b3.z + b4.z) / 4.0);
+    for (std::list<myTriangle>::iterator tetTri = hullBuilding.begin(); tetTri != hullBuilding.end(); tetTri++) {
+      myVec3 triVert1(xCoords[tetTri->v1], yCoords[tetTri->v1], zCoords[tetTri->v1]);
+      myVec3 triVert2(xCoords[tetTri->v2], yCoords[tetTri->v2], zCoords[tetTri->v2]);
+      myVec3 triVert3(xCoords[tetTri->v3], yCoords[tetTri->v3], zCoords[tetTri->v3]);
+      myVec3 trinorm   = getNorm(triVert1, triVert2, triVert3);
+      myVec3 ptInPlane = tetTri == hullBuilding.begin() ? b1 : b4;  // first triangle definitely has b1 in it, other three definitely have b4
+      if (isInFront(barycenter, ptInPlane, trinorm)) {
+        // don't want the faces of the tetrahedron to face inwards (towards barycenter) so reverse orientation
+        // by swapping two vertices
+        int temp   = tetTri->v1;
+        tetTri->v1 = tetTri->v2;
+        tetTri->v2 = temp;
       }
-      std::list<int> startPoints1;
-      std::list<int> startPoints2;
-      std::list<int> startPoints3;
-      std::list<int> startPoints4;
-      //scope this so that 'point' is not in function scope
+    }
+    // now, have starting polyhedron in hullBuilding (all faces are facing outwards according to getNorm_) and remaining nodes to process are in aggNodes
+    // recursively, for each triangle, make a list of the points that are 'in front' of the triangle. Find the point with the maximum distance from the triangle.
+    // Add three new triangles, each sharing one edge with the original triangle but now with the most distant point as a vertex. Remove the most distant point from
+    // the list of all points that need to be processed. Also from that list remove all points that are in front of the original triangle but not in front of all three
+    // new triangles, since they are now enclosed in the hull.
+    // Construct point lists for each face of the tetrahedron individually.
+    myVec3 trinorms[4];  // normals to the four tetrahedron faces, now oriented outwards
+    int index = 0;
+    for (std::list<myTriangle>::iterator tetTri = hullBuilding.begin(); tetTri != hullBuilding.end(); tetTri++) {
+      myVec3 triVert1(xCoords[tetTri->v1], yCoords[tetTri->v1], zCoords[tetTri->v1]);
+      myVec3 triVert2(xCoords[tetTri->v2], yCoords[tetTri->v2], zCoords[tetTri->v2]);
+      myVec3 triVert3(xCoords[tetTri->v3], yCoords[tetTri->v3], zCoords[tetTri->v3]);
+      trinorms[index] = getNorm(triVert1, triVert2, triVert3);
+      index++;
+    }
+    std::list<int> startPoints1;
+    std::list<int> startPoints2;
+    std::list<int> startPoints3;
+    std::list<int> startPoints4;
+    // scope this so that 'point' is not in function scope
+    {
+      Iter point = aggNodes.begin();
+      while (!aggNodes.empty())  // this removes points one at a time as they are put in startPointsN or are already done
       {
-        Iter point = aggNodes.begin();
-        while(!aggNodes.empty())  //this removes points one at a time as they are put in startPointsN or are already done
-        {
-          myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
-          //Note: Because of the way the tetrahedron faces are constructed above,
-          //face 1 definitely contains b1 and faces 2-4 definitely contain b4.
-          if(isInFront(pointVec, b1, trinorms[0])) {
-            startPoints1.push_back(*point);
-            point = aggNodes.erase(point);
-          } else if(isInFront(pointVec, b4, trinorms[1])) {
-            startPoints2.push_back(*point);
-            point = aggNodes.erase(point);
-          } else if(isInFront(pointVec, b4, trinorms[2])) {
-            startPoints3.push_back(*point);
-            point = aggNodes.erase(point);
-          } else if(isInFront(pointVec, b4, trinorms[3])) {
-            startPoints4.push_back(*point);
-            point = aggNodes.erase(point);
-          } else {
-            point = aggNodes.erase(point); //points here are already inside tetrahedron.
-          }
+        myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
+        // Note: Because of the way the tetrahedron faces are constructed above,
+        // face 1 definitely contains b1 and faces 2-4 definitely contain b4.
+        if (isInFront(pointVec, b1, trinorms[0])) {
+          startPoints1.push_back(*point);
+          point = aggNodes.erase(point);
+        } else if (isInFront(pointVec, b4, trinorms[1])) {
+          startPoints2.push_back(*point);
+          point = aggNodes.erase(point);
+        } else if (isInFront(pointVec, b4, trinorms[2])) {
+          startPoints3.push_back(*point);
+          point = aggNodes.erase(point);
+        } else if (isInFront(pointVec, b4, trinorms[3])) {
+          startPoints4.push_back(*point);
+          point = aggNodes.erase(point);
+        } else {
+          point = aggNodes.erase(point);  // points here are already inside tetrahedron.
         }
-        //Call processTriangle for each triangle in the initial tetrahedron, one at a time.
-      }
-      typedef std::list<myTriangle>::iterator TriIter;
-      TriIter firstTri = hullBuilding.begin();
-      myTriangle start1 = *firstTri;
-      firstTri++;
-      myTriangle start2 = *firstTri;
-      firstTri++;
-      myTriangle start3 = *firstTri;
-      firstTri++;
-      myTriangle start4 = *firstTri;
-      //kick off depth-first recursive filling of hullBuilding list with all triangles in the convex hull
-      if(!startPoints1.empty())
-        processTriangle(hullBuilding, start1, startPoints1, barycenter, xCoords, yCoords, zCoords);
-      if(!startPoints2.empty())
-        processTriangle(hullBuilding, start2, startPoints2, barycenter, xCoords, yCoords, zCoords);
-      if(!startPoints3.empty())
-        processTriangle(hullBuilding, start3, startPoints3, barycenter, xCoords, yCoords, zCoords);
-      if(!startPoints4.empty())
-        processTriangle(hullBuilding, start4, startPoints4, barycenter, xCoords, yCoords, zCoords);
-      //hullBuilding now has all triangles that make up this hull.
-      //Dump hullBuilding info into the list of all triangles for the scene.
-      vertices.reserve(vertices.size() + 3 * hullBuilding.size());
-      for(TriIter hullTri = hullBuilding.begin(); hullTri != hullBuilding.end(); hullTri++) {
-        vertices.push_back(hullTri->v1);
-        vertices.push_back(hullTri->v2);
-        vertices.push_back(hullTri->v3);
-        geomSizes.push_back(3);
       }
+      // Call processTriangle for each triangle in the initial tetrahedron, one at a time.
+    }
+    typedef std::list<myTriangle>::iterator TriIter;
+    TriIter firstTri  = hullBuilding.begin();
+    myTriangle start1 = *firstTri;
+    firstTri++;
+    myTriangle start2 = *firstTri;
+    firstTri++;
+    myTriangle start3 = *firstTri;
+    firstTri++;
+    myTriangle start4 = *firstTri;
+    // kick off depth-first recursive filling of hullBuilding list with all triangles in the convex hull
+    if (!startPoints1.empty())
+      processTriangle(hullBuilding, start1, startPoints1, barycenter, xCoords, yCoords, zCoords);
+    if (!startPoints2.empty())
+      processTriangle(hullBuilding, start2, startPoints2, barycenter, xCoords, yCoords, zCoords);
+    if (!startPoints3.empty())
+      processTriangle(hullBuilding, start3, startPoints3, barycenter, xCoords, yCoords, zCoords);
+    if (!startPoints4.empty())
+      processTriangle(hullBuilding, start4, startPoints4, barycenter, xCoords, yCoords, zCoords);
+    // hullBuilding now has all triangles that make up this hull.
+    // Dump hullBuilding info into the list of all triangles for the scene.
+    vertices.reserve(vertices.size() + 3 * hullBuilding.size());
+    for (TriIter hullTri = hullBuilding.begin(); hullTri != hullBuilding.end(); hullTri++) {
+      vertices.push_back(hullTri->v1);
+      vertices.push_back(hullTri->v2);
+      vertices.push_back(hullTri->v3);
+      geomSizes.push_back(3);
     }
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doGraphEdges(std::vector<int>& vertices, std::vector<int>& geomSizes, Teuchos::RCP<GraphBase>& G, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & /* fx */, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & /* fy */, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & /* fz */) {
-    ArrayView<const Scalar> values;
-    ArrayView<const LocalOrdinal> neighbors;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::doGraphEdges(std::vector<int>& vertices, std::vector<int>& geomSizes, Teuchos::RCP<GraphBase>& G, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& /* fx */, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& /* fy */, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& /* fz */) {
+  ArrayView<const Scalar> values;
+  ArrayView<const LocalOrdinal> neighbors;
 
-    std::vector<std::pair<int, int> > vert1; //vertices (node indices)
+  std::vector<std::pair<int, int> > vert1;  // vertices (node indices)
 
-    ArrayView<const LocalOrdinal> indices;
-    for(LocalOrdinal locRow = 0; locRow < LocalOrdinal(G->GetNodeNumVertices()); locRow++) {
-      neighbors = G->getNeighborVertices(locRow);
-      //Add those local indices (columns) to the list of connections (which are pairs of the form (localM, localN))
-      for(int gEdge = 0; gEdge < int(neighbors.size()); ++gEdge) {
-        vert1.push_back(std::pair<int, int>(locRow, neighbors[gEdge]));
-      }
+  ArrayView<const LocalOrdinal> indices;
+  for (LocalOrdinal locRow = 0; locRow < LocalOrdinal(G->GetNodeNumVertices()); locRow++) {
+    neighbors = G->getNeighborVertices(locRow);
+    // Add those local indices (columns) to the list of connections (which are pairs of the form (localM, localN))
+    for (int gEdge = 0; gEdge < int(neighbors.size()); ++gEdge) {
+      vert1.push_back(std::pair<int, int>(locRow, neighbors[gEdge]));
     }
-    for(size_t i = 0; i < vert1.size(); i ++) {
-      if(vert1[i].first > vert1[i].second) {
-        int temp = vert1[i].first;
-        vert1[i].first = vert1[i].second;
-        vert1[i].second = temp;
-      }
-    }
-    std::sort(vert1.begin(), vert1.end());
-    std::vector<std::pair<int, int> >::iterator newEnd = unique(vert1.begin(), vert1.end()); //remove duplicate edges
-    vert1.erase(newEnd, vert1.end());
-    //std::vector<int> points1;
-    vertices.reserve(2 * vert1.size());
-    geomSizes.reserve(vert1.size());
-    for(size_t i = 0; i < vert1.size(); i++) {
-      vertices.push_back(vert1[i].first);
-      vertices.push_back(vert1[i].second);
-      geomSizes.push_back(2);
+  }
+  for (size_t i = 0; i < vert1.size(); i++) {
+    if (vert1[i].first > vert1[i].second) {
+      int temp        = vert1[i].first;
+      vert1[i].first  = vert1[i].second;
+      vert1[i].second = temp;
     }
   }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  int VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ccw(const myVec2& a, const myVec2& b, const myVec2& c)
-  {
-    const double ccw_zero_threshold=1e-8;
-    double val = (b.x - a.x) * (c.y - a.y) - (b.y - a.y) * (c.x - a.x);
-    return (val > ccw_zero_threshold) ? 1 : ( (val < -ccw_zero_threshold) ? -1 : 0);
+  std::sort(vert1.begin(), vert1.end());
+  std::vector<std::pair<int, int> >::iterator newEnd = unique(vert1.begin(), vert1.end());  // remove duplicate edges
+  vert1.erase(newEnd, vert1.end());
+  // std::vector<int> points1;
+  vertices.reserve(2 * vert1.size());
+  geomSizes.reserve(vert1.size());
+  for (size_t i = 0; i < vert1.size(); i++) {
+    vertices.push_back(vert1[i].first);
+    vertices.push_back(vert1[i].second);
+    geomSizes.push_back(2);
   }
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::crossProduct(myVec3 v1, myVec3 v2)
-  {
-    return myVec3(v1.y * v2.z - v1.z * v2.y, v1.z * v2.x - v1.x * v2.z, v1.x * v2.y - v1.y * v2.x);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ccw(const myVec2& a, const myVec2& b, const myVec2& c) {
+  const double ccw_zero_threshold = 1e-8;
+  double val                      = (b.x - a.x) * (c.y - a.y) - (b.y - a.y) * (c.x - a.x);
+  return (val > ccw_zero_threshold) ? 1 : ((val < -ccw_zero_threshold) ? -1 : 0);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dotProduct(myVec2 v1, myVec2 v2)
-  {
-    return v1.x * v2.x + v1.y * v2.y;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::crossProduct(myVec3 v1, myVec3 v2) {
+  return myVec3(v1.y * v2.z - v1.z * v2.y, v1.z * v2.x - v1.x * v2.z, v1.x * v2.y - v1.y * v2.x);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dotProduct(myVec3 v1, myVec3 v2)
-  {
-    return v1.x * v2.x + v1.y * v2.y + v1.z * v2.z;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dotProduct(myVec2 v1, myVec2 v2) {
+  return v1.x * v2.x + v1.y * v2.y;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isInFront(myVec3 point, myVec3 inPlane, myVec3 n)
-  {
-    myVec3 rel(point.x - inPlane.x, point.y - inPlane.y, point.z - inPlane.z); //position of the point relative to the plane
-    return dotProduct(rel, n) > 1e-12 ? true : false;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::dotProduct(myVec3 v1, myVec3 v2) {
+  return v1.x * v2.x + v1.y * v2.y + v1.z * v2.z;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::mymagnitude(myVec2 vec)
-  {
-    return sqrt(dotProduct(vec, vec));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::isInFront(myVec3 point, myVec3 inPlane, myVec3 n) {
+  myVec3 rel(point.x - inPlane.x, point.y - inPlane.y, point.z - inPlane.z);  // position of the point relative to the plane
+  return dotProduct(rel, n) > 1e-12 ? true : false;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::mymagnitude(myVec3 vec)
-  {
-    return sqrt(dotProduct(vec, vec));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::mymagnitude(myVec2 vec) {
+  return sqrt(dotProduct(vec, vec));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::distance(myVec2 p1, myVec2 p2)
-  {
-    return mymagnitude(vecSubtract(p1, p2));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::mymagnitude(myVec3 vec) {
+  return sqrt(dotProduct(vec, vec));
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::distance(myVec2 p1, myVec2 p2) {
+  return mymagnitude(vecSubtract(p1, p2));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::distance(myVec3 p1, myVec3 p2)
-  {
-    return mymagnitude(vecSubtract(p1, p2));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::distance(myVec3 p1, myVec3 p2) {
+  return mymagnitude(vecSubtract(p1, p2));
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  myVec2 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::vecSubtract(myVec2 v1, myVec2 v2)
-  {
-    return myVec2(v1.x - v2.x, v1.y - v2.y);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+myVec2 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::vecSubtract(myVec2 v1, myVec2 v2) {
+  return myVec2(v1.x - v2.x, v1.y - v2.y);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::vecSubtract(myVec3 v1, myVec3 v2)
-  {
-    return myVec3(v1.x - v2.x, v1.y - v2.y, v1.z - v2.z);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::vecSubtract(myVec3 v1, myVec3 v2) {
+  return myVec3(v1.x - v2.x, v1.y - v2.y, v1.z - v2.z);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  myVec2 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNorm(myVec2 v) //"normal" to a 2D vector - just rotate 90 degrees to left
-  {
-    return myVec2(v.y, -v.x);
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+myVec2 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNorm(myVec2 v)  //"normal" to a 2D vector - just rotate 90 degrees to left
+{
+  return myVec2(v.y, -v.x);
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNorm(myVec3 v1, myVec3 v2, myVec3 v3) //normal to face of triangle (will be outward rel. to polyhedron) (v1, v2, v3 are in CCW order when normal is toward viewpoint)
-  {
-    return crossProduct(vecSubtract(v2, v1), vecSubtract(v3, v1));
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+myVec3 VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getNorm(myVec3 v1, myVec3 v2, myVec3 v3)  // normal to face of triangle (will be outward rel. to polyhedron) (v1, v2, v3 are in CCW order when normal is toward viewpoint)
+{
+  return crossProduct(vecSubtract(v2, v1), vecSubtract(v3, v1));
+}
 
-  //get minimum distance from 'point' to plane containing v1, v2, v3 (or the triangle with v1, v2, v3 as vertices)
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pointDistFromTri(myVec3 point, myVec3 v1, myVec3 v2, myVec3 v3)
-  {
-    using namespace std;
-    myVec3 norm = getNorm(v1, v2, v3);
-    //must normalize the normal vector
-    double normScl = mymagnitude(norm);
-    double rv = 0.0;
-    if (normScl > 1e-8) {
-      norm.x /= normScl;
-      norm.y /= normScl;
-      norm.z /= normScl;
-      rv = fabs(dotProduct(norm, vecSubtract(point, v1)));
+// get minimum distance from 'point' to plane containing v1, v2, v3 (or the triangle with v1, v2, v3 as vertices)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+double VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::pointDistFromTri(myVec3 point, myVec3 v1, myVec3 v2, myVec3 v3) {
+  using namespace std;
+  myVec3 norm = getNorm(v1, v2, v3);
+  // must normalize the normal vector
+  double normScl = mymagnitude(norm);
+  double rv      = 0.0;
+  if (normScl > 1e-8) {
+    norm.x /= normScl;
+    norm.y /= normScl;
+    norm.z /= normScl;
+    rv = fabs(dotProduct(norm, vecSubtract(point, v1)));
+  } else {
+    // triangle is degenerated
+    myVec3 test1  = vecSubtract(v3, v1);
+    myVec3 test2  = vecSubtract(v2, v1);
+    bool useTest1 = mymagnitude(test1) > 0.0 ? true : false;
+    bool useTest2 = mymagnitude(test2) > 0.0 ? true : false;
+    if (useTest1 == true) {
+      double normScl1 = mymagnitude(test1);
+      test1.x /= normScl1;
+      test1.y /= normScl1;
+      test1.z /= normScl1;
+      rv = fabs(dotProduct(test1, vecSubtract(point, v1)));
+    } else if (useTest2 == true) {
+      double normScl2 = mymagnitude(test2);
+      test2.x /= normScl2;
+      test2.y /= normScl2;
+      test2.z /= normScl2;
+      rv = fabs(dotProduct(test2, vecSubtract(point, v1)));
     } else {
-      // triangle is degenerated
-      myVec3 test1 = vecSubtract(v3, v1);
-      myVec3 test2 = vecSubtract(v2, v1);
-      bool useTest1 = mymagnitude(test1) > 0.0 ? true : false;
-      bool useTest2 = mymagnitude(test2) > 0.0 ? true : false;
-      if(useTest1 == true) {
-        double normScl1 = mymagnitude(test1);
-        test1.x /= normScl1;
-        test1.y /= normScl1;
-        test1.z /= normScl1;
-        rv = fabs(dotProduct(test1, vecSubtract(point, v1)));
-      } else if (useTest2 == true) {
-        double normScl2 = mymagnitude(test2);
-        test2.x /= normScl2;
-        test2.y /= normScl2;
-        test2.z /= normScl2;
-        rv = fabs(dotProduct(test2, vecSubtract(point, v1)));
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-                 "VisualizationHelpers::pointDistFromTri: Could not determine the distance of a point to a triangle.");
-      }
-
+      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                                 "VisualizationHelpers::pointDistFromTri: Could not determine the distance of a point to a triangle.");
     }
-    return rv;
   }
+  return rv;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<myTriangle> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::processTriangle(std::list<myTriangle>& tris, myTriangle tri, std::list<int>& pointsInFront, myVec3& barycenter, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords) {
-    //*tri is in the tris list, and is the triangle to process here. tris is a complete list of all triangles in the hull so far. pointsInFront is only a list of the nodes in front of tri. Need coords also.
-    //precondition: each triangle is already oriented so that getNorm_(v1, v2, v3) points outward (away from interior of hull)
-    //First find the point furthest from triangle.
-    using namespace std;
-    typedef std::list<int>::iterator Iter;
-    typedef std::list<myTriangle>::iterator TriIter;
-    typedef list<pair<int, int> >::iterator EdgeIter;
-    double maxDist = 0;
-    //Need vector representations of triangle's vertices
-    myVec3 v1(xCoords[tri.v1], yCoords[tri.v1], zCoords[tri.v1]);
-    myVec3 v2(xCoords[tri.v2], yCoords[tri.v2], zCoords[tri.v2]);
-    myVec3 v3(xCoords[tri.v3], yCoords[tri.v3], zCoords[tri.v3]);
-    myVec3 farPointVec; //useful to have both the point's coordinates and it's position in the list
-    Iter farPoint = pointsInFront.begin();
-    for(Iter point = pointsInFront.begin(); point != pointsInFront.end(); point++)
-    {
-      myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
-      double dist = pointDistFromTri(pointVec, v1, v2, v3);
-      // Want to make sure nodeDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
-      if(dist > maxDist && dist - maxDist > 1e-12)
-      {
-        dist = maxDist;
-        farPointVec = pointVec;
-        farPoint = point;
-      }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<myTriangle> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::processTriangle(std::list<myTriangle>& tris, myTriangle tri, std::list<int>& pointsInFront, myVec3& barycenter, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& zCoords) {
+  //*tri is in the tris list, and is the triangle to process here. tris is a complete list of all triangles in the hull so far. pointsInFront is only a list of the nodes in front of tri. Need coords also.
+  // precondition: each triangle is already oriented so that getNorm_(v1, v2, v3) points outward (away from interior of hull)
+  // First find the point furthest from triangle.
+  using namespace std;
+  typedef std::list<int>::iterator Iter;
+  typedef std::list<myTriangle>::iterator TriIter;
+  typedef list<pair<int, int> >::iterator EdgeIter;
+  double maxDist = 0;
+  // Need vector representations of triangle's vertices
+  myVec3 v1(xCoords[tri.v1], yCoords[tri.v1], zCoords[tri.v1]);
+  myVec3 v2(xCoords[tri.v2], yCoords[tri.v2], zCoords[tri.v2]);
+  myVec3 v3(xCoords[tri.v3], yCoords[tri.v3], zCoords[tri.v3]);
+  myVec3 farPointVec;  // useful to have both the point's coordinates and it's position in the list
+  Iter farPoint = pointsInFront.begin();
+  for (Iter point = pointsInFront.begin(); point != pointsInFront.end(); point++) {
+    myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
+    double dist = pointDistFromTri(pointVec, v1, v2, v3);
+    // Want to make sure nodeDist is significantly larger than maxDist to prevent roundoff errors from impacting node choice.
+    if (dist > maxDist && dist - maxDist > 1e-12) {
+      dist        = maxDist;
+      farPointVec = pointVec;
+      farPoint    = point;
     }
-    //Find all the triangles that the point is in front of (can be more than 1)
-    //At the same time, remove them from tris, as every one will be replaced later
-    vector<myTriangle> visible; //use a list of iterators so that the underlying object is still in tris
-    for(TriIter it = tris.begin(); it != tris.end();)
-    {
-      myVec3 vec1(xCoords[it->v1], yCoords[it->v1], zCoords[it->v1]);
-      myVec3 vec2(xCoords[it->v2], yCoords[it->v2], zCoords[it->v2]);
-      myVec3 vec3(xCoords[it->v3], yCoords[it->v3], zCoords[it->v3]);
-      myVec3 norm = getNorm(vec1, vec2, vec3);
-      if(isInFront(farPointVec, vec1, norm))
-      {
-        visible.push_back(*it);
-        it = tris.erase(it);
-      }
-      else
-        it++;
+  }
+  // Find all the triangles that the point is in front of (can be more than 1)
+  // At the same time, remove them from tris, as every one will be replaced later
+  vector<myTriangle> visible;  // use a list of iterators so that the underlying object is still in tris
+  for (TriIter it = tris.begin(); it != tris.end();) {
+    myVec3 vec1(xCoords[it->v1], yCoords[it->v1], zCoords[it->v1]);
+    myVec3 vec2(xCoords[it->v2], yCoords[it->v2], zCoords[it->v2]);
+    myVec3 vec3(xCoords[it->v3], yCoords[it->v3], zCoords[it->v3]);
+    myVec3 norm = getNorm(vec1, vec2, vec3);
+    if (isInFront(farPointVec, vec1, norm)) {
+      visible.push_back(*it);
+      it = tris.erase(it);
+    } else
+      it++;
+  }
+  // Figure out what triangles need to be destroyed/created
+  // First create a list of edges (as std::pair<int, int>, where the two ints are the node endpoints)
+  list<pair<int, int> > horizon;
+  // For each triangle, add edges to the list iff the edge only appears once in the set of all
+  // Have members of horizon have the lower node # first, and the higher one second
+  for (vector<myTriangle>::iterator it = visible.begin(); it != visible.end(); it++) {
+    pair<int, int> e1(it->v1, it->v2);
+    pair<int, int> e2(it->v2, it->v3);
+    pair<int, int> e3(it->v1, it->v3);
+    //"sort" the pair values
+    if (e1.first > e1.second) {
+      int temp  = e1.first;
+      e1.first  = e1.second;
+      e1.second = temp;
     }
-    //Figure out what triangles need to be destroyed/created
-    //First create a list of edges (as std::pair<int, int>, where the two ints are the node endpoints)
-    list<pair<int, int> > horizon;
-    //For each triangle, add edges to the list iff the edge only appears once in the set of all
-    //Have members of horizon have the lower node # first, and the higher one second
-    for(vector<myTriangle>::iterator it = visible.begin(); it != visible.end(); it++)
-    {
-      pair<int, int> e1(it->v1, it->v2);
-      pair<int, int> e2(it->v2, it->v3);
-      pair<int, int> e3(it->v1, it->v3);
-      //"sort" the pair values
-      if(e1.first > e1.second)
-      {
-        int temp = e1.first;
-        e1.first = e1.second;
-        e1.second = temp;
-      }
-      if(e2.first > e2.second)
-      {
-        int temp = e2.first;
-        e2.first = e2.second;
-        e2.second = temp;
-      }
-      if(e3.first > e3.second)
-      {
-        int temp = e3.first;
-        e3.first = e3.second;
-        e3.second = temp;
-      }
-      horizon.push_back(e1);
-      horizon.push_back(e2);
-      horizon.push_back(e3);
+    if (e2.first > e2.second) {
+      int temp  = e2.first;
+      e2.first  = e2.second;
+      e2.second = temp;
     }
-    //sort based on lower node first, then higher node (lexicographical ordering built in to pair)
-    horizon.sort();
-    //Remove all edges from horizon, except those that appear exactly once
-    {
-      EdgeIter it = horizon.begin();
-      while(it != horizon.end())
-      {
-        int occur = count(horizon.begin(), horizon.end(), *it);
-        if(occur > 1)
-        {
-          pair<int, int> removeVal = *it;
-          while(removeVal == *it && !(it == horizon.end()))
-            it = horizon.erase(it);
-        }
-        else
-          it++;
-      }
+    if (e3.first > e3.second) {
+      int temp  = e3.first;
+      e3.first  = e3.second;
+      e3.second = temp;
     }
-    //Now make a list of new triangles being created, each of which take 2 vertices from an edge and one from farPoint
-    list<myTriangle> newTris;
-    for(EdgeIter it = horizon.begin(); it != horizon.end(); it++)
-    {
-      myTriangle t(it->first, it->second, *farPoint);
-      newTris.push_back(t);
+    horizon.push_back(e1);
+    horizon.push_back(e2);
+    horizon.push_back(e3);
+  }
+  // sort based on lower node first, then higher node (lexicographical ordering built in to pair)
+  horizon.sort();
+  // Remove all edges from horizon, except those that appear exactly once
+  {
+    EdgeIter it = horizon.begin();
+    while (it != horizon.end()) {
+      int occur = count(horizon.begin(), horizon.end(), *it);
+      if (occur > 1) {
+        pair<int, int> removeVal = *it;
+        while (removeVal == *it && !(it == horizon.end()))
+          it = horizon.erase(it);
+      } else
+        it++;
     }
-    //Ensure every new triangle is oriented outwards, using the barycenter of the initial tetrahedron
-    vector<myTriangle> trisToProcess;
-    vector<list<int> > newFrontPoints;
-    for(TriIter it = newTris.begin(); it != newTris.end(); it++)
-    {
-      myVec3 t1(xCoords[it->v1], yCoords[it->v1], zCoords[it->v1]);
-      myVec3 t2(xCoords[it->v2], yCoords[it->v2], zCoords[it->v2]);
-      myVec3 t3(xCoords[it->v3], yCoords[it->v3], zCoords[it->v3]);
-      if(isInFront(barycenter, t1, getNorm(t1, t2, t3)))
-      {
-        //need to swap two vertices to flip orientation of triangle
-        int temp = it->v1;
-        myVec3 tempVec = t1;
-        it->v1 = it->v2;
-        t1 = t2;
-        it->v2 = temp;
-        t2 = tempVec;
-      }
-      myVec3 outwardNorm = getNorm(t1, t2, t3); //now definitely points outwards
-      //Add the triangle to tris
-      tris.push_back(*it);
-      trisToProcess.push_back(tris.back());
-      //Make a list of the points that are in front of nextToProcess, to be passed in for processing
-      list<int> newInFront;
-      for(Iter point = pointsInFront.begin(); point != pointsInFront.end();)
-      {
-        myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
-        if(isInFront(pointVec, t1, outwardNorm))
-        {
-          newInFront.push_back(*point);
-          point = pointsInFront.erase(point);
-        }
-        else
-          point++;
-      }
-      newFrontPoints.push_back(newInFront);
+  }
+  // Now make a list of new triangles being created, each of which take 2 vertices from an edge and one from farPoint
+  list<myTriangle> newTris;
+  for (EdgeIter it = horizon.begin(); it != horizon.end(); it++) {
+    myTriangle t(it->first, it->second, *farPoint);
+    newTris.push_back(t);
+  }
+  // Ensure every new triangle is oriented outwards, using the barycenter of the initial tetrahedron
+  vector<myTriangle> trisToProcess;
+  vector<list<int> > newFrontPoints;
+  for (TriIter it = newTris.begin(); it != newTris.end(); it++) {
+    myVec3 t1(xCoords[it->v1], yCoords[it->v1], zCoords[it->v1]);
+    myVec3 t2(xCoords[it->v2], yCoords[it->v2], zCoords[it->v2]);
+    myVec3 t3(xCoords[it->v3], yCoords[it->v3], zCoords[it->v3]);
+    if (isInFront(barycenter, t1, getNorm(t1, t2, t3))) {
+      // need to swap two vertices to flip orientation of triangle
+      int temp       = it->v1;
+      myVec3 tempVec = t1;
+      it->v1         = it->v2;
+      t1             = t2;
+      it->v2         = temp;
+      t2             = tempVec;
     }
-    vector<myTriangle> allRemoved; //list of all invalid iterators that were erased by calls to processmyTriangle below
-    for(int i = 0; i < int(trisToProcess.size()); i++)
-    {
-      if(!newFrontPoints[i].empty())
-      {
-        //Comparing the 'triangle to process' to the one for this call prevents infinite recursion/stack overflow.
-        //TODO: Why was it doing that? Rounding error? Make more robust fix. But this does work for the time being.
-        if(find(allRemoved.begin(), allRemoved.end(), trisToProcess[i]) == allRemoved.end() && !(trisToProcess[i] == tri))
-        {
-          vector<myTriangle> removedList = processTriangle(tris, trisToProcess[i], newFrontPoints[i], barycenter, xCoords, yCoords, zCoords);
-          for(int j = 0; j < int(removedList.size()); j++)
-            allRemoved.push_back(removedList[j]);
-        }
+    myVec3 outwardNorm = getNorm(t1, t2, t3);  // now definitely points outwards
+    // Add the triangle to tris
+    tris.push_back(*it);
+    trisToProcess.push_back(tris.back());
+    // Make a list of the points that are in front of nextToProcess, to be passed in for processing
+    list<int> newInFront;
+    for (Iter point = pointsInFront.begin(); point != pointsInFront.end();) {
+      myVec3 pointVec(xCoords[*point], yCoords[*point], zCoords[*point]);
+      if (isInFront(pointVec, t1, outwardNorm)) {
+        newInFront.push_back(*point);
+        point = pointsInFront.erase(point);
+      } else
+        point++;
+    }
+    newFrontPoints.push_back(newInFront);
+  }
+  vector<myTriangle> allRemoved;  // list of all invalid iterators that were erased by calls to processmyTriangle below
+  for (int i = 0; i < int(trisToProcess.size()); i++) {
+    if (!newFrontPoints[i].empty()) {
+      // Comparing the 'triangle to process' to the one for this call prevents infinite recursion/stack overflow.
+      // TODO: Why was it doing that? Rounding error? Make more robust fix. But this does work for the time being.
+      if (find(allRemoved.begin(), allRemoved.end(), trisToProcess[i]) == allRemoved.end() && !(trisToProcess[i] == tri)) {
+        vector<myTriangle> removedList = processTriangle(tris, trisToProcess[i], newFrontPoints[i], barycenter, xCoords, yCoords, zCoords);
+        for (int j = 0; j < int(removedList.size()); j++)
+          allRemoved.push_back(removedList[j]);
       }
     }
-    return visible;
   }
+  return visible;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<int> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::giftWrap(std::vector<myVec2>& points, std::vector<int>& nodes, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & yCoords) {
-    TEUCHOS_TEST_FOR_EXCEPTION(points.size() < 3, Exceptions::RuntimeError,
-             "CoarseningVisualization::giftWrap: Gift wrap algorithm input has to have at least 3 points!");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<int> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::giftWrap(std::vector<myVec2>& points, std::vector<int>& nodes, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& xCoords, const Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& yCoords) {
+  TEUCHOS_TEST_FOR_EXCEPTION(points.size() < 3, Exceptions::RuntimeError,
+                             "CoarseningVisualization::giftWrap: Gift wrap algorithm input has to have at least 3 points!");
 
-#if 1 // TAW's version to determine "minimal" node
-    // determine minimal x and y coordinates
-    double min_x =points[0].x;
-    double min_y =points[0].y;
-    for(std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++) {
-      int i = it - nodes.begin();
-      if(points[i].x < min_x) min_x = points[i].x;
-      if(points[i].y < min_y) min_y = points[i].y;
-    }
-    // create dummy min coordinates
-    min_x -= 1.0;
-    min_y -= 1.0;
-    myVec2 dummy_min(min_x, min_y);
+#if 1  // TAW's version to determine "minimal" node
+  // determine minimal x and y coordinates
+  double min_x = points[0].x;
+  double min_y = points[0].y;
+  for (std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++) {
+    int i = it - nodes.begin();
+    if (points[i].x < min_x) min_x = points[i].x;
+    if (points[i].y < min_y) min_y = points[i].y;
+  }
+  // create dummy min coordinates
+  min_x -= 1.0;
+  min_y -= 1.0;
+  myVec2 dummy_min(min_x, min_y);
 
-    // loop over all nodes and determine nodes with minimal distance to (min_x, min_y)
-    std::vector<int> hull;
-    myVec2 min = points[0];
-    double mindist = distance(min,dummy_min);
-    std::vector<int>::iterator minNode = nodes.begin();
-    for(std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++) {
-      int i = it - nodes.begin();
-      if(distance(points[i],dummy_min) < mindist) {
-        mindist = distance(points[i],dummy_min);
-        min = points[i];
-        minNode = it;
-      }
+  // loop over all nodes and determine nodes with minimal distance to (min_x, min_y)
+  std::vector<int> hull;
+  myVec2 min                         = points[0];
+  double mindist                     = distance(min, dummy_min);
+  std::vector<int>::iterator minNode = nodes.begin();
+  for (std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++) {
+    int i = it - nodes.begin();
+    if (distance(points[i], dummy_min) < mindist) {
+      mindist = distance(points[i], dummy_min);
+      min     = points[i];
+      minNode = it;
     }
-    hull.push_back(*minNode);
-#else // Brian's code
-    std::vector<int> hull;
-    std::vector<int>::iterator minNode = nodes.begin();
-    myVec2 min = points[0];
-    for(std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++)
-    {
-      int i = it - nodes.begin();
-      if(points[i].x < min.x || (fabs(points[i].x - min.x) < 1e-10 && points[i].y < min.y))
-      {
-        min = points[i];
-        minNode = it;
-      }
+  }
+  hull.push_back(*minNode);
+#else  // Brian's code
+  std::vector<int> hull;
+  std::vector<int>::iterator minNode = nodes.begin();
+  myVec2 min                         = points[0];
+  for (std::vector<int>::iterator it = nodes.begin(); it != nodes.end(); it++) {
+    int i = it - nodes.begin();
+    if (points[i].x < min.x || (fabs(points[i].x - min.x) < 1e-10 && points[i].y < min.y)) {
+      min     = points[i];
+      minNode = it;
     }
-    hull.push_back(*minNode);
+  }
+  hull.push_back(*minNode);
 #endif
 
-    bool includeMin = false;
-    //int debug_it = 0;
-    while(1)
-    {
-      std::vector<int>::iterator leftMost = nodes.begin();
-      if(!includeMin && leftMost == minNode)
-      {
-        leftMost++;
-      }
-      std::vector<int>::iterator it = leftMost;
-      it++;
-      for(; it != nodes.end(); it++)
-      {
-        if(it == minNode && !includeMin) //don't compare to min on very first sweep
-          continue;
-        if(*it == hull.back())
-          continue;
-        //see if it is in front of line containing nodes thisHull.back() and leftMost
-        //first get the left normal of leftMost - thisHull.back() (<dy, -dx>)
-        myVec2 leftMostVec = points[leftMost - nodes.begin()];
-        myVec2 lastVec(xCoords[hull.back()], yCoords[hull.back()]);
-        myVec2 testNorm = getNorm(vecSubtract(leftMostVec, lastVec));
-        //now dot testNorm with *it - leftMost. If dot is positive, leftMost becomes it. If dot is zero, take one further from thisHull.back().
-        myVec2 itVec(xCoords[*it], yCoords[*it]);
-        double dotProd = dotProduct(testNorm, vecSubtract(itVec, lastVec));
-        if(-1e-8 < dotProd && dotProd < 1e-8)
-        {
-          //thisHull.back(), it and leftMost are collinear.
-          //Just sum the differences in x and differences in y for each and compare to get further one, don't need distance formula
-          myVec2 itDist = vecSubtract(itVec, lastVec);
-          myVec2 leftMostDist = vecSubtract(leftMostVec, lastVec);
-          if(fabs(itDist.x) + fabs(itDist.y) > fabs(leftMostDist.x) + fabs(leftMostDist.y)) {
-            leftMost = it;
-          }
-        }
-        else if(dotProd > 0) {
+  bool includeMin = false;
+  // int debug_it = 0;
+  while (1) {
+    std::vector<int>::iterator leftMost = nodes.begin();
+    if (!includeMin && leftMost == minNode) {
+      leftMost++;
+    }
+    std::vector<int>::iterator it = leftMost;
+    it++;
+    for (; it != nodes.end(); it++) {
+      if (it == minNode && !includeMin)  // don't compare to min on very first sweep
+        continue;
+      if (*it == hull.back())
+        continue;
+      // see if it is in front of line containing nodes thisHull.back() and leftMost
+      // first get the left normal of leftMost - thisHull.back() (<dy, -dx>)
+      myVec2 leftMostVec = points[leftMost - nodes.begin()];
+      myVec2 lastVec(xCoords[hull.back()], yCoords[hull.back()]);
+      myVec2 testNorm = getNorm(vecSubtract(leftMostVec, lastVec));
+      // now dot testNorm with *it - leftMost. If dot is positive, leftMost becomes it. If dot is zero, take one further from thisHull.back().
+      myVec2 itVec(xCoords[*it], yCoords[*it]);
+      double dotProd = dotProduct(testNorm, vecSubtract(itVec, lastVec));
+      if (-1e-8 < dotProd && dotProd < 1e-8) {
+        // thisHull.back(), it and leftMost are collinear.
+        // Just sum the differences in x and differences in y for each and compare to get further one, don't need distance formula
+        myVec2 itDist       = vecSubtract(itVec, lastVec);
+        myVec2 leftMostDist = vecSubtract(leftMostVec, lastVec);
+        if (fabs(itDist.x) + fabs(itDist.y) > fabs(leftMostDist.x) + fabs(leftMostDist.y)) {
           leftMost = it;
-
         }
+      } else if (dotProd > 0) {
+        leftMost = it;
       }
-      //if leftMost is min, then the loop is complete.
-      if(*leftMost == *minNode)
-        break;
-      hull.push_back(*leftMost);
-      includeMin = true; //have found second point (the one after min) so now include min in the searches
-      //debug_it ++;
-      //if(debug_it > 100) exit(0); //break;
-    }
-    return hull;
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::vector<int> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::makeUnique(std::vector<int>& vertices) const
-  {
-    using namespace std;
-    vector<int> uniqueNodes = vertices;
-    sort(uniqueNodes.begin(), uniqueNodes.end());
-    vector<int>::iterator newUniqueFineEnd = unique(uniqueNodes.begin(), uniqueNodes.end());
-    uniqueNodes.erase(newUniqueFineEnd, uniqueNodes.end());
-    //uniqueNodes is now a sorted list of the nodes whose info actually goes in file
-    //Now replace values in vertices with locations of the old values in uniqueFine
-    for(int i = 0; i < int(vertices.size()); i++)
-    {
-      int lo = 0;
-      int hi = uniqueNodes.size() - 1;
-      int mid = 0;
-      int search = vertices[i];
-      while(lo <= hi)
-      {
-        mid = lo + (hi - lo) / 2;
-        if(uniqueNodes[mid] == search)
-          break;
-        else if(uniqueNodes[mid] > search)
-          hi = mid - 1;
-        else
-          lo = mid + 1;
-      }
-      if(uniqueNodes[mid] != search)
-        throw runtime_error("Issue in makeUnique_() - a point wasn't found in list.");
-      vertices[i] = mid;
     }
-    return uniqueNodes;
+    // if leftMost is min, then the loop is complete.
+    if (*leftMost == *minNode)
+      break;
+    hull.push_back(*leftMost);
+    includeMin = true;  // have found second point (the one after min) so now include min in the searches
+    // debug_it ++;
+    // if(debug_it > 100) exit(0); //break;
   }
+  return hull;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceAll(std::string result, const std::string& replaceWhat, const std::string& replaceWithWhat) const {
-    while(1) {
-      const int pos = result.find(replaceWhat);
-      if (pos == -1)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::vector<int> VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::makeUnique(std::vector<int>& vertices) const {
+  using namespace std;
+  vector<int> uniqueNodes = vertices;
+  sort(uniqueNodes.begin(), uniqueNodes.end());
+  vector<int>::iterator newUniqueFineEnd = unique(uniqueNodes.begin(), uniqueNodes.end());
+  uniqueNodes.erase(newUniqueFineEnd, uniqueNodes.end());
+  // uniqueNodes is now a sorted list of the nodes whose info actually goes in file
+  // Now replace values in vertices with locations of the old values in uniqueFine
+  for (int i = 0; i < int(vertices.size()); i++) {
+    int lo     = 0;
+    int hi     = uniqueNodes.size() - 1;
+    int mid    = 0;
+    int search = vertices[i];
+    while (lo <= hi) {
+      mid = lo + (hi - lo) / 2;
+      if (uniqueNodes[mid] == search)
         break;
-      result.replace(pos, replaceWhat.size(), replaceWithWhat);
+      else if (uniqueNodes[mid] > search)
+        hi = mid - 1;
+      else
+        lo = mid + 1;
     }
-    return result;
+    if (uniqueNodes[mid] != search)
+      throw runtime_error("Issue in makeUnique_() - a point wasn't found in list.");
+    vertices[i] = mid;
   }
+  return uniqueNodes;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList & pL) const {
-    std::string filenameToWrite = getBaseFileName(numProcs, level, pL);
-    filenameToWrite = this->replaceAll(filenameToWrite, "%PROCID",   toString(myRank));
-    return filenameToWrite;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::replaceAll(std::string result, const std::string& replaceWhat, const std::string& replaceWithWhat) const {
+  while (1) {
+    const int pos = result.find(replaceWhat);
+    if (pos == -1)
+      break;
+    result.replace(pos, replaceWhat.size(), replaceWithWhat);
   }
+  return result;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getBaseFileName(int numProcs, int level, const Teuchos::ParameterList & pL) const {
-    std::string filenameToWrite = pL.get<std::string>("visualization: output filename");
-    int timeStep = pL.get<int>("visualization: output file: time step");
-    int iter     = pL.get<int>("visualization: output file: iter");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getFileName(int numProcs, int myRank, int level, const Teuchos::ParameterList& pL) const {
+  std::string filenameToWrite = getBaseFileName(numProcs, level, pL);
+  filenameToWrite             = this->replaceAll(filenameToWrite, "%PROCID", toString(myRank));
+  return filenameToWrite;
+}
 
-    if(filenameToWrite.rfind(".vtu") == std::string::npos)
-      filenameToWrite.append(".vtu");
-    if(numProcs > 1 && filenameToWrite.rfind("%PROCID") == std::string::npos) //filename can't be identical between processsors in parallel problem
-      filenameToWrite.insert(filenameToWrite.rfind(".vtu"), "-proc%PROCID");
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getBaseFileName(int numProcs, int level, const Teuchos::ParameterList& pL) const {
+  std::string filenameToWrite = pL.get<std::string>("visualization: output filename");
+  int timeStep                = pL.get<int>("visualization: output file: time step");
+  int iter                    = pL.get<int>("visualization: output file: iter");
 
-    filenameToWrite = this->replaceAll(filenameToWrite, "%LEVELID",  toString(level));
-    filenameToWrite = this->replaceAll(filenameToWrite, "%TIMESTEP", toString(timeStep));
-    filenameToWrite = this->replaceAll(filenameToWrite, "%ITER",     toString(iter));
-    return filenameToWrite;
-  }
+  if (filenameToWrite.rfind(".vtu") == std::string::npos)
+    filenameToWrite.append(".vtu");
+  if (numProcs > 1 && filenameToWrite.rfind("%PROCID") == std::string::npos)  // filename can't be identical between processsors in parallel problem
+    filenameToWrite.insert(filenameToWrite.rfind(".vtu"), "-proc%PROCID");
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getPVTUFileName(int numProcs, int /* myRank */, int level, const Teuchos::ParameterList & pL) const {
-    std::string filenameToWrite = getBaseFileName(numProcs, level, pL);
-    std::string masterStem = filenameToWrite.substr(0, filenameToWrite.rfind(".vtu"));
-    masterStem = this->replaceAll(masterStem, "%PROCID", "");
-    std::string pvtuFilename = masterStem + "-master.pvtu";
-    return pvtuFilename;
-  }
+  filenameToWrite = this->replaceAll(filenameToWrite, "%LEVELID", toString(level));
+  filenameToWrite = this->replaceAll(filenameToWrite, "%TIMESTEP", toString(timeStep));
+  filenameToWrite = this->replaceAll(filenameToWrite, "%ITER", toString(iter));
+  return filenameToWrite;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKOpening(std::ofstream & fout, std::vector<int> & uniqueFine, std::vector<int> & geomSizesFine) const {
-    std::string styleName = "PointCloud"; // TODO adapt this
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+std::string VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::getPVTUFileName(int numProcs, int /* myRank */, int level, const Teuchos::ParameterList& pL) const {
+  std::string filenameToWrite = getBaseFileName(numProcs, level, pL);
+  std::string masterStem      = filenameToWrite.substr(0, filenameToWrite.rfind(".vtu"));
+  masterStem                  = this->replaceAll(masterStem, "%PROCID", "");
+  std::string pvtuFilename    = masterStem + "-master.pvtu";
+  return pvtuFilename;
+}
 
-    std::string indent = "      ";
-    fout << "<!--" << styleName << " Aggregates Visualization-->" << std::endl;
-    fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << std::endl;
-    fout << "  <UnstructuredGrid>" << std::endl;
-    fout << "    <Piece NumberOfPoints=\"" << uniqueFine.size() << "\" NumberOfCells=\"" << geomSizesFine.size() << "\">" << std::endl;
-    fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << std::endl;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKOpening(std::ofstream& fout, std::vector<int>& uniqueFine, std::vector<int>& geomSizesFine) const {
+  std::string styleName = "PointCloud";  // TODO adapt this
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKNodes(std::ofstream & fout, std::vector<int> & uniqueFine, Teuchos::RCP<const Map> & nodeMap) const {
-    std::string indent = "      ";
-    fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << std::endl;
-    indent = "          ";
-    bool localIsGlobal = GlobalOrdinal(nodeMap->getGlobalNumElements()) == GlobalOrdinal(nodeMap->getLocalNumElements());
-    for(size_t i = 0; i < uniqueFine.size(); i++)
-    {
-      if(localIsGlobal)
-      {
-        fout << uniqueFine[i] << " "; //if all nodes are on this processor, do not map from local to global
-      }
-      else
-        fout << nodeMap->getGlobalElement(uniqueFine[i]) << " ";
-      if(i % 10 == 9)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-  }
+  std::string indent = "      ";
+  fout << "<!--" << styleName << " Aggregates Visualization-->" << std::endl;
+  fout << "<VTKFile type=\"UnstructuredGrid\" byte_order=\"LittleEndian\">" << std::endl;
+  fout << "  <UnstructuredGrid>" << std::endl;
+  fout << "    <Piece NumberOfPoints=\"" << uniqueFine.size() << "\" NumberOfCells=\"" << geomSizesFine.size() << "\">" << std::endl;
+  fout << "      <PointData Scalars=\"Node Aggregate Processor\">" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKData(std::ofstream & fout, std::vector<int> & uniqueFine, LocalOrdinal myAggOffset, ArrayRCP<LocalOrdinal> & vertex2AggId, int myRank) const {
-    std::string indent = "          ";
-    fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    for(int i = 0; i < int(uniqueFine.size()); i++)
-    {
-      fout << myAggOffset + vertex2AggId[uniqueFine[i]] << " ";
-      if(i % 10 == 9)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    for(int i = 0; i < int(uniqueFine.size()); i++)
-    {
-      fout << myRank << " ";
-      if(i % 20 == 19)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "      </PointData>" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKNodes(std::ofstream& fout, std::vector<int>& uniqueFine, Teuchos::RCP<const Map>& nodeMap) const {
+  std::string indent = "      ";
+  fout << "        <DataArray type=\"Int32\" Name=\"Node\" format=\"ascii\">" << std::endl;
+  indent             = "          ";
+  bool localIsGlobal = GlobalOrdinal(nodeMap->getGlobalNumElements()) == GlobalOrdinal(nodeMap->getLocalNumElements());
+  for (size_t i = 0; i < uniqueFine.size(); i++) {
+    if (localIsGlobal) {
+      fout << uniqueFine[i] << " ";  // if all nodes are on this processor, do not map from local to global
+    } else
+      fout << nodeMap->getGlobalElement(uniqueFine[i]) << " ";
+    if (i % 10 == 9)
+      fout << std::endl
+           << indent;
   }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKCoordinates(std::ofstream & fout, std::vector<int> & uniqueFine, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType> & fz, int dim) const {
-    std::string indent = "      ";
-    fout << "      <Points>" << std::endl;
-    fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    for(int i = 0; i < int(uniqueFine.size()); i++)
-    {
-      fout << fx[uniqueFine[i]] << " " << fy[uniqueFine[i]] << " ";
-      if(dim == 2)
-        fout << "0 ";
-      else
-        fout << fz[uniqueFine[i]] << " ";
-      if(i % 3 == 2)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "      </Points>" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKData(std::ofstream& fout, std::vector<int>& uniqueFine, LocalOrdinal myAggOffset, ArrayRCP<LocalOrdinal>& vertex2AggId, int myRank) const {
+  std::string indent = "          ";
+  fout << "        <DataArray type=\"Int32\" Name=\"Aggregate\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  for (int i = 0; i < int(uniqueFine.size()); i++) {
+    fout << myAggOffset + vertex2AggId[uniqueFine[i]] << " ";
+    if (i % 10 == 9)
+      fout << std::endl
+           << indent;
+  }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  for (int i = 0; i < int(uniqueFine.size()); i++) {
+    fout << myRank << " ";
+    if (i % 20 == 19)
+      fout << std::endl
+           << indent;
   }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "      </PointData>" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKCells(std::ofstream & fout, std::vector<int> & /* uniqueFine */, std::vector<LocalOrdinal> & vertices, std::vector<LocalOrdinal> & geomSize) const {
-    std::string indent = "      ";
-    fout << "      <Cells>" << std::endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    for(int i = 0; i < int(vertices.size()); i++)
-    {
-      fout << vertices[i] << " ";
-      if(i % 10 == 9)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    int accum = 0;
-    for(int i = 0; i < int(geomSize.size()); i++)
-    {
-      accum += geomSize[i];
-      fout << accum << " ";
-      if(i % 10 == 9)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << std::endl;
-    fout << indent;
-    for(int i = 0; i < int(geomSize.size()); i++)
-    {
-      switch(geomSize[i])
-      {
-        case 1:
-          fout << "1 "; //Point
-          break;
-        case 2:
-          fout << "3 "; //Line
-          break;
-        case 3:
-          fout << "5 "; //Triangle
-          break;
-        default:
-          fout << "7 "; //Polygon
-      }
-      if(i % 30 == 29)
-        fout << std::endl << indent;
-    }
-    fout << std::endl;
-    fout << "        </DataArray>" << std::endl;
-    fout << "      </Cells>" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKCoordinates(std::ofstream& fout, std::vector<int>& uniqueFine, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fx, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fy, Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::coordinateType>& fz, int dim) const {
+  std::string indent = "      ";
+  fout << "      <Points>" << std::endl;
+  fout << "        <DataArray type=\"Float64\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  for (int i = 0; i < int(uniqueFine.size()); i++) {
+    fout << fx[uniqueFine[i]] << " " << fy[uniqueFine[i]] << " ";
+    if (dim == 2)
+      fout << "0 ";
+    else
+      fout << fz[uniqueFine[i]] << " ";
+    if (i % 3 == 2)
+      fout << std::endl
+           << indent;
   }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "      </Points>" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKClosing(std::ofstream & fout) const {
-    fout << "    </Piece>" << std::endl;
-    fout << "  </UnstructuredGrid>" << std::endl;
-    fout << "</VTKFile>" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKCells(std::ofstream& fout, std::vector<int>& /* uniqueFine */, std::vector<LocalOrdinal>& vertices, std::vector<LocalOrdinal>& geomSize) const {
+  std::string indent = "      ";
+  fout << "      <Cells>" << std::endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"connectivity\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  for (int i = 0; i < int(vertices.size()); i++) {
+    fout << vertices[i] << " ";
+    if (i % 10 == 9)
+      fout << std::endl
+           << indent;
+  }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"offsets\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  int accum = 0;
+  for (int i = 0; i < int(geomSize.size()); i++) {
+    accum += geomSize[i];
+    fout << accum << " ";
+    if (i % 10 == 9)
+      fout << std::endl
+           << indent;
   }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "        <DataArray type=\"Int32\" Name=\"types\" format=\"ascii\">" << std::endl;
+  fout << indent;
+  for (int i = 0; i < int(geomSize.size()); i++) {
+    switch (geomSize[i]) {
+      case 1:
+        fout << "1 ";  // Point
+        break;
+      case 2:
+        fout << "3 ";  // Line
+        break;
+      case 3:
+        fout << "5 ";  // Triangle
+        break;
+      default:
+        fout << "7 ";  // Polygon
+    }
+    if (i % 30 == 29)
+      fout << std::endl
+           << indent;
+  }
+  fout << std::endl;
+  fout << "        </DataArray>" << std::endl;
+  fout << "      </Cells>" << std::endl;
+}
 
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writeFileVTKClosing(std::ofstream& fout) const {
+  fout << "    </Piece>" << std::endl;
+  fout << "  </UnstructuredGrid>" << std::endl;
+  fout << "</VTKFile>" << std::endl;
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writePVTU(std::ofstream& pvtu, std::string baseFname, int numProcs, bool bFineEdges, bool /* bCoarseEdges */) const {
-    //If using vtk, filenameToWrite now contains final, correct ***.vtu filename (for the current proc)
-    //So the root proc will need to use its own filenameToWrite to make a list of the filenames of all other procs to put in
-    //pvtu file.
-    pvtu << "<VTKFile type=\"PUnstructuredGrid\" byte_order=\"LittleEndian\">" << std::endl;
-    pvtu << "  <PUnstructuredGrid GhostLevel=\"0\">" << std::endl;
-    pvtu << "    <PPointData Scalars=\"Node Aggregate Processor\">" << std::endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Node\"/>" << std::endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Aggregate\"/>" << std::endl;
-    pvtu << "      <PDataArray type=\"Int32\" Name=\"Processor\"/>" << std::endl;
-    pvtu << "    </PPointData>" << std::endl;
-    pvtu << "    <PPoints>" << std::endl;
-    pvtu << "      <PDataArray type=\"Float64\" NumberOfComponents=\"3\"/>" << std::endl;
-    pvtu << "    </PPoints>" << std::endl;
-    for(int i = 0; i < numProcs; i++) {
-      //specify the piece for each proc (the replaceAll expression matches what the filenames will be on other procs)
-      pvtu << "    <Piece Source=\"" << replaceAll(baseFname, "%PROCID", toString(i)) << "\"/>" << std::endl;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::writePVTU(std::ofstream& pvtu, std::string baseFname, int numProcs, bool bFineEdges, bool /* bCoarseEdges */) const {
+  // If using vtk, filenameToWrite now contains final, correct ***.vtu filename (for the current proc)
+  // So the root proc will need to use its own filenameToWrite to make a list of the filenames of all other procs to put in
+  // pvtu file.
+  pvtu << "<VTKFile type=\"PUnstructuredGrid\" byte_order=\"LittleEndian\">" << std::endl;
+  pvtu << "  <PUnstructuredGrid GhostLevel=\"0\">" << std::endl;
+  pvtu << "    <PPointData Scalars=\"Node Aggregate Processor\">" << std::endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Node\"/>" << std::endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Aggregate\"/>" << std::endl;
+  pvtu << "      <PDataArray type=\"Int32\" Name=\"Processor\"/>" << std::endl;
+  pvtu << "    </PPointData>" << std::endl;
+  pvtu << "    <PPoints>" << std::endl;
+  pvtu << "      <PDataArray type=\"Float64\" NumberOfComponents=\"3\"/>" << std::endl;
+  pvtu << "    </PPoints>" << std::endl;
+  for (int i = 0; i < numProcs; i++) {
+    // specify the piece for each proc (the replaceAll expression matches what the filenames will be on other procs)
+    pvtu << "    <Piece Source=\"" << replaceAll(baseFname, "%PROCID", toString(i)) << "\"/>" << std::endl;
+  }
+  // reference files with graph pieces, if applicable
+  if (bFineEdges) {
+    for (int i = 0; i < numProcs; i++) {
+      std::string fn = replaceAll(baseFname, "%PROCID", toString(i));
+      pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-finegraph") << "\"/>" << std::endl;
     }
-    //reference files with graph pieces, if applicable
-    if(bFineEdges)
+  }
+  /*if(doCoarseGraphEdges_)
+  {
+    for(int i = 0; i < numProcs; i++)
     {
-      for(int i = 0; i < numProcs; i++)
-      {
-        std::string fn = replaceAll(baseFname, "%PROCID", toString(i));
-        pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-finegraph") << "\"/>" << std::endl;
-      }
+      std::string fn = replaceAll(baseFname, "%PROCID", toString(i));
+      pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-coarsegraph") << "\"/>" << std::endl;
     }
-    /*if(doCoarseGraphEdges_)
-    {
-      for(int i = 0; i < numProcs; i++)
-      {
-        std::string fn = replaceAll(baseFname, "%PROCID", toString(i));
-        pvtu << "    <Piece Source=\"" << fn.insert(fn.rfind(".vtu"), "-coarsegraph") << "\"/>" << std::endl;
-      }
-    }*/
-    pvtu << "  </PUnstructuredGrid>" << std::endl;
-    pvtu << "</VTKFile>" << std::endl;
-    pvtu.close();
-  }
+  }*/
+  pvtu << "  </PUnstructuredGrid>" << std::endl;
+  pvtu << "</VTKFile>" << std::endl;
+  pvtu.close();
+}
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildColormap() const {
-    try {
-      std::ofstream color("random-colormap.xml");
-      color << "<ColorMap name=\"MueLu-Random\" space=\"RGB\">" << std::endl;
-      //Give -1, -2, -3 distinctive colors (so that the style functions can have constrasted geometry types)
-      //Do red, orange, yellow to constrast with cool color spectrum for other types
-      color << "  <Point x=\"" << -1 << "\" o=\"1\" r=\"1\" g=\"0\" b=\"0\"/>" << std::endl;
-      color << "  <Point x=\"" << -2 << "\" o=\"1\" r=\"1\" g=\"0.6\" b=\"0\"/>" << std::endl;
-      color << "  <Point x=\"" << -3 << "\" o=\"1\" r=\"1\" g=\"1\" b=\"0\"/>" << std::endl;
-      srand(time(NULL));
-      for(int i = 0; i < 5000; i += 4) {
-        color << "  <Point x=\"" << i << "\" o=\"1\" r=\"" << (rand() % 50) / 256.0 << "\" g=\"" << (rand() % 256) / 256.0 << "\" b=\"" << (rand() % 256) / 256.0 << "\"/>" << std::endl;
-      }
-      color << "</ColorMap>" << std::endl;
-      color.close();
-    }
-    catch(std::exception& e) {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
-               "VisualizationHelpers::buildColormap: Error while building colormap file: " << e.what());
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::buildColormap() const {
+  try {
+    std::ofstream color("random-colormap.xml");
+    color << "<ColorMap name=\"MueLu-Random\" space=\"RGB\">" << std::endl;
+    // Give -1, -2, -3 distinctive colors (so that the style functions can have constrasted geometry types)
+    // Do red, orange, yellow to constrast with cool color spectrum for other types
+    color << "  <Point x=\"" << -1 << "\" o=\"1\" r=\"1\" g=\"0\" b=\"0\"/>" << std::endl;
+    color << "  <Point x=\"" << -2 << "\" o=\"1\" r=\"1\" g=\"0.6\" b=\"0\"/>" << std::endl;
+    color << "  <Point x=\"" << -3 << "\" o=\"1\" r=\"1\" g=\"1\" b=\"0\"/>" << std::endl;
+    srand(time(NULL));
+    for (int i = 0; i < 5000; i += 4) {
+      color << "  <Point x=\"" << i << "\" o=\"1\" r=\"" << (rand() % 50) / 256.0 << "\" g=\"" << (rand() % 256) / 256.0 << "\" b=\"" << (rand() % 256) / 256.0 << "\"/>" << std::endl;
     }
+    color << "</ColorMap>" << std::endl;
+    color.close();
+  } catch (std::exception& e) {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, Exceptions::RuntimeError,
+                               "VisualizationHelpers::buildColormap: Error while building colormap file: " << e.what());
   }
+}
 
-} // namespace MueLu
+}  // namespace MueLu
 
 #endif /* MUELU_VISUALIZATIONHELPERS_DEF_HPP_ */
diff --git a/packages/muelu/test/blockedtransfer/BlockedTransfer.cpp b/packages/muelu/test/blockedtransfer/BlockedTransfer.cpp
index 30f99c9c1b3f..05a54ac4e864 100644
--- a/packages/muelu/test/blockedtransfer/BlockedTransfer.cpp
+++ b/packages/muelu/test/blockedtransfer/BlockedTransfer.cpp
@@ -101,8 +101,8 @@
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosXpetraAdapter.hpp" // this header defines Belos::XpetraOp()
-#include "BelosMueLuAdapter.hpp" // this header defines Belos::MueLuOp()
+#include "BelosXpetraAdapter.hpp"  // this header defines Belos::XpetraOp()
+#include "BelosMueLuAdapter.hpp"   // this header defines Belos::MueLuOp()
 #endif
 
 /////////////////////////
@@ -110,83 +110,77 @@
 
 namespace MueLuTests {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map, Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map, Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
 #include "MueLu_UseShortNames.hpp"
-    Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(map, 3);
-
-    LocalOrdinal NumMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
-    GlobalOrdinal NumGlobalElements = map->getGlobalNumElements();
-    GlobalOrdinal nIndexBase = map->getIndexBase();
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-    {
-      if (MyGlobalElements[i] == nIndexBase)
-      {
-        // off-diagonal for first row
-        Indices[0] = nIndexBase;
-        NumEntries = 1;
-        Values[0] = c;
-      }
-      else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1)
-      {
-        // off-diagonal for last row
-        Indices[0] = nIndexBase + NumGlobalElements - 2;
-        NumEntries = 1;
-        Values[0] = b;
-      }
-      else
-      {
-        // off-diagonal for internal row
-        Indices[0] = MyGlobalElements[i] - 1;
-        Values[1] = b;
-        Indices[1] = MyGlobalElements[i] + 1;
-        Values[0] = c;
-        NumEntries = 2;
-      }
-
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(MyGlobalElements[i],
-          Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-          Teuchos::tuple<Scalar>(a) );
-
-    } //for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-
-
-    mtx->fillComplete(map,map);
-
-    return mtx;
-  }
+  Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(map, 3);
+
+  LocalOrdinal NumMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
+  GlobalOrdinal NumGlobalElements                          = map->getGlobalNumElements();
+  GlobalOrdinal nIndexBase                                 = map->getIndexBase();
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  for (LocalOrdinal i = 0; i < NumMyElements; ++i) {
+    if (MyGlobalElements[i] == nIndexBase) {
+      // off-diagonal for first row
+      Indices[0] = nIndexBase;
+      NumEntries = 1;
+      Values[0]  = c;
+    } else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1) {
+      // off-diagonal for last row
+      Indices[0] = nIndexBase + NumGlobalElements - 2;
+      NumEntries = 1;
+      Values[0]  = b;
+    } else {
+      // off-diagonal for internal row
+      Indices[0] = MyGlobalElements[i] - 1;
+      Values[1]  = b;
+      Indices[1] = MyGlobalElements[i] + 1;
+      Values[0]  = c;
+      NumEntries = 2;
+    }
+
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(MyGlobalElements[i],
+                            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                            Teuchos::tuple<Scalar>(a));
 
+  }  // for (LocalOrdinal i = 0; i < NumMyElements; ++i)
+
+  mtx->fillComplete(map, map);
+
+  return mtx;
 }
 
+}  // namespace MueLuTests
+
 /////////////////
 // MAIN
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   bool success = false;
   bool verbose = true;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -197,13 +191,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     /**********************************************************************************/
     /* SET TEST PARAMETERS                                                            */
     /**********************************************************************************/
-    Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of xpetra
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     xpetraParameters.check();
@@ -219,39 +213,40 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<const Map> bigMap;
     RCP<const Map> map1;
     RCP<const Map> map2;
-    GO numElements = 500;
+    GO numElements  = 500;
     GO numElements1 = 400;
     GO numElements2 = 100;
 
     std::vector<size_t> stridingInfo;
     stridingInfo.push_back(1);
-    map1   = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm, -1);
-    map2   = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm, -1);
+    map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm, -1);
+    map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm, -1);
 
-    std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-    Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
+    std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+    Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
     localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-    Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
+    Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
     localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-    bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
+    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+    bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
 
     std::vector<Teuchos::RCP<const Map> > maps;
-    maps.push_back(map1); maps.push_back(map2);
+    maps.push_back(map1);
+    maps.push_back(map2);
 
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(bigMap, maps);
+    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(bigMap, maps);
 
-    RCP<CrsMatrixWrap> Op11 = MueLuTests::GenerateProblemMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map1,2,-1,-1);
-    RCP<CrsMatrixWrap> Op22 = MueLuTests::GenerateProblemMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map2,3,-2,-1);
+    RCP<CrsMatrixWrap> Op11 = MueLuTests::GenerateProblemMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map1, 2, -1, -1);
+    RCP<CrsMatrixWrap> Op22 = MueLuTests::GenerateProblemMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map2, 3, -2, -1);
 
     /*Op11->describe(*out,Teuchos::VERB_EXTREME);
       Op22->describe(*out,Teuchos::VERB_EXTREME);*/
 
     // build blocked operator
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(mapExtractor,mapExtractor,10));
+    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(mapExtractor, mapExtractor, 10));
 
-    bOp->setMatrix(0,0,Op11);
-    bOp->setMatrix(1,1,Op22);
+    bOp->setMatrix(0, 0, Op11);
+    bOp->setMatrix(1, 1, Op22);
 
     bOp->fillComplete();
 
@@ -259,46 +254,46 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     Hierarchy H;
     H.SetMaxCoarseSize(50);
     RCP<Level> levelOne = H.GetLevel();
-    levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
+    levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
 
     RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
     RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+    A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
 
     RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
-    RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
+    RCP<TransPFactory> R11Fact     = rcp(new TransPFactory());
 
     RCP<TentativePFactory> P22TentFact = rcp(new TentativePFactory());
-    RCP<PgPFactory> P22Fact = rcp(new PgPFactory());
-    RCP<GenericRFactory> R22Fact = rcp(new GenericRFactory());
+    RCP<PgPFactory> P22Fact            = rcp(new PgPFactory());
+    RCP<GenericRFactory> R22Fact       = rcp(new GenericRFactory());
 
     std::string ifpackType;
     Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 5);
+    ifpackList.set("relaxation: sweeps", (LO)5);
     ifpackList.set("relaxation: damping factor", Teuchos::ScalarTraits<SC>::one());
     ifpackType = "RELAXATION";
     ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    RCP<SmootherPrototype> smoProto11     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+    RCP<SmootherPrototype> smoProto11 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
     smoProto11->SetFactory("A", A11Fact);
-    RCP<SmootherPrototype> smoProto22     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+    RCP<SmootherPrototype> smoProto22 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
     smoProto22->SetFactory("A", A22Fact);
 
-    //RCP<SmootherPrototype> smoProto11     = rcp( new DirectSolver("", Teuchos::ParameterList(), A11Fact) );
-    //RCP<SmootherPrototype> smoProto22     = rcp( new DirectSolver("", Teuchos::ParameterList(), A22Fact) );
+    // RCP<SmootherPrototype> smoProto11     = rcp( new DirectSolver("", Teuchos::ParameterList(), A11Fact) );
+    // RCP<SmootherPrototype> smoProto22     = rcp( new DirectSolver("", Teuchos::ParameterList(), A22Fact) );
 
-    RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
-    RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
+    RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
+    RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
 
     RCP<FactoryManager> M11 = rcp(new FactoryManager());
     M11->SetKokkosRefactor(false);
     M11->SetFactory("A", A11Fact);
     M11->SetFactory("P", P11Fact);
-    M11->SetFactory("Ptent", P11Fact); //for Nullspace
+    M11->SetFactory("Ptent", P11Fact);  // for Nullspace
     M11->SetFactory("R", R11Fact);
     M11->SetFactory("Smoother", Smoo11Fact);
     M11->SetIgnoreUserData(true);
@@ -308,7 +303,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     M22->SetFactory("A", A22Fact);
     M22->SetFactory("P", P22Fact);
     M22->SetFactory("R", R22Fact);
-    M22->SetFactory("Ptent", P22TentFact); //for both P22 and Nullspace
+    M22->SetFactory("Ptent", P22TentFact);  // for both P22 and Nullspace
     M22->SetFactory("Smoother", Smoo22Fact);
     M22->SetIgnoreUserData(true);
 
@@ -321,25 +316,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<Factory> AcFact = rcp(new BlockedRAPFactory());
 
     // Smoothers
-    RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
     smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(2));
     smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::ScalarTraits<Scalar>::one()));
-    smootherPrototype->AddFactoryManager(M11,0);
-    smootherPrototype->AddFactoryManager(M22,1);
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->AddFactoryManager(M11, 0);
+    smootherPrototype->AddFactoryManager(M22, 1);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
     // Coarse grid correction
-    RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp( new BlockedGaussSeidelSmoother() );
-    coarseSolverPrototype->AddFactoryManager(M11,0);
-    coarseSolverPrototype->AddFactoryManager(M22,1);
-    RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, Teuchos::null) );
+    RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp(new BlockedGaussSeidelSmoother());
+    coarseSolverPrototype->AddFactoryManager(M11, 0);
+    coarseSolverPrototype->AddFactoryManager(M22, 1);
+    RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, Teuchos::null));
 
     // main factory manager
     FactoryManager M;
-    M.SetFactory("A",            AcFact);
-    M.SetFactory("P",            PFact);
-    M.SetFactory("R",            RFact);
-    M.SetFactory("Smoother",     smootherFact); // TODO fix me
+    M.SetFactory("A", AcFact);
+    M.SetFactory("P", PFact);
+    M.SetFactory("R", RFact);
+    M.SetFactory("Smoother", smootherFact);  // TODO fix me
     M.SetFactory("CoarseSolver", coarseSolverFact);
 
     H.SetVerbLevel(MueLu::Test);
@@ -349,19 +344,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     RCP<Level> l1 = H.GetLevel(1);
     RCP<Level> l2 = H.GetLevel(2);
 
-    l0->print(*out,Teuchos::VERB_EXTREME);
-    l1->print(*out,Teuchos::VERB_EXTREME);
-    l2->print(*out,Teuchos::VERB_EXTREME);
+    l0->print(*out, Teuchos::VERB_EXTREME);
+    l1->print(*out, Teuchos::VERB_EXTREME);
+    l2->print(*out, Teuchos::VERB_EXTREME);
 
     // Define B
-    RCP<Vector> X = VectorFactory::Build(bigMap,1);
-    RCP<Vector> B = VectorFactory::Build(bigMap,1);
+    RCP<Vector> X = VectorFactory::Build(bigMap, 1);
+    RCP<Vector> B = VectorFactory::Build(bigMap, 1);
     X->setSeed(846930886);
     X->randomize();
     bOp->apply(*X, *B, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
     // X = 0
-    X->putScalar((SC) 0.0);
+    X->putScalar((SC)0.0);
 
     LO nIts = 9;
     H.Iterate(*B, *X, nIts);
@@ -370,7 +365,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
@@ -378,5 +373,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/convergence/Convergence.cpp b/packages/muelu/test/convergence/Convergence.cpp
index 954035d908a4..3ec23a69e625 100644
--- a/packages/muelu/test/convergence/Convergence.cpp
+++ b/packages/muelu/test/convergence/Convergence.cpp
@@ -79,54 +79,55 @@
 #include <BelosPseudoBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockGmresSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
 #ifdef HAVE_MUELU_BELOS
 namespace Belos {
 
-  template<class SC, class MV, class OP>
-  class MyStatusTest : public StatusTestImpResNorm<SC, MV, OP> {
-  private:
-    typedef StatusTestImpResNorm<SC, MV, OP> Base;
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType MagnitudeType;
-
-  public:
-    MyStatusTest(MagnitudeType tol) : Base(tol),
-                                      prevNorm_(-Teuchos::ScalarTraits<MagnitudeType>::one()),
-                                      curNorm_(-Teuchos::ScalarTraits<MagnitudeType>::one()) { }
-
-    StatusType checkStatus(Iteration<SC,MV,OP>* iSolver) {
-      // Get residual
-      std::vector<MagnitudeType> norms(1);
-      Teuchos::RCP<const MV> res = iSolver->getNativeResiduals(&norms);
-      MagnitudeType resNorm = norms[0];
-
-      if (curNorm_ == -Teuchos::ScalarTraits<MagnitudeType>::one()) {
-        prevNorm_ = curNorm_ = resNorm;
-      } else {
-        prevNorm_ = curNorm_;
-        curNorm_  = resNorm;
-      }
-
-      return Base::checkStatus(iSolver);
+template <class SC, class MV, class OP>
+class MyStatusTest : public StatusTestImpResNorm<SC, MV, OP> {
+ private:
+  typedef StatusTestImpResNorm<SC, MV, OP> Base;
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType MagnitudeType;
+
+ public:
+  MyStatusTest(MagnitudeType tol)
+    : Base(tol)
+    , prevNorm_(-Teuchos::ScalarTraits<MagnitudeType>::one())
+    , curNorm_(-Teuchos::ScalarTraits<MagnitudeType>::one()) {}
+
+  StatusType checkStatus(Iteration<SC, MV, OP>* iSolver) {
+    // Get residual
+    std::vector<MagnitudeType> norms(1);
+    Teuchos::RCP<const MV> res = iSolver->getNativeResiduals(&norms);
+    MagnitudeType resNorm      = norms[0];
+
+    if (curNorm_ == -Teuchos::ScalarTraits<MagnitudeType>::one()) {
+      prevNorm_ = curNorm_ = resNorm;
+    } else {
+      prevNorm_ = curNorm_;
+      curNorm_  = resNorm;
     }
 
-    double rate() const {
-      return Teuchos::as<double>(curNorm_/prevNorm_);
-    }
+    return Base::checkStatus(iSolver);
+  }
 
-  private:
-    MagnitudeType prevNorm_;
-    MagnitudeType curNorm_;
-  };
-}
+  double rate() const {
+    return Teuchos::as<double>(curNorm_ / prevNorm_);
+  }
+
+ private:
+  MagnitudeType prevNorm_;
+  MagnitudeType curNorm_;
+};
+}  // namespace Belos
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -138,17 +139,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   typedef Teuchos::ScalarTraits<SC> STS;
   SC zero = STS::zero(), one = STS::one();
   typedef typename STS::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   bool success = true;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int numProc = comm->getSize();
-    int myRank  = comm->getRank();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int numProc                         = comm->getSize();
+    int myRank                          = comm->getRank();
 
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream& out       = *fancy;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
@@ -156,11 +157,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // =========================================================================
     Xpetra::Parameters xpetraParameters(clp);
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     const int numLists = 1;
@@ -170,7 +171,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     bool failed = false;
     for (int k = 0; k < numLists; k++) {
       const std::string& dirName = dirList[k];
-      std::string problemFile = dirName + "problem.xml";
+      std::string problemFile    = dirName + "problem.xml";
 
       Teuchos::ParameterList galeriParameters;
       Teuchos::updateParametersFromXmlFileAndBroadcast(problemFile, Teuchos::Ptr<Teuchos::ParameterList>(&galeriParameters), *comm);
@@ -180,9 +181,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // =========================================================================
       // Problem construction (copy-paste from Driver.cpp)
       // =========================================================================
-      RCP<Matrix>       A;
-      RCP<Map>          map;
-      RCP<MultiVector>  nullspace;
+      RCP<Matrix> A;
+      RCP<Map> map;
+      RCP<MultiVector> nullspace;
       RCP<RealValuedMultiVector> coordinates;
 
       // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
@@ -200,24 +201,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
       // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
       if (matrixType == "Laplace1D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriParameters);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriParameters);
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriParameters);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriParameters);
 
       } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
                  matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriParameters);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriParameters);
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriParameters);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriParameters);
 
       } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriParameters);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriParameters);
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriParameters);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriParameters);
       }
 
       // Expand map to do multiple DOF per node for block problems
       if (matrixType == "Elasticity2D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
       if (matrixType == "Elasticity3D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
 #if 0
       out << "========================================================\n" << xpetraParameters << galeriParameters;
@@ -227,8 +228,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
           << "========================================================" << std::endl;
 #endif
 
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, galeriParameters);
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, galeriParameters);
       A = Pr->BuildMatrix();
 
       if (matrixType == "Elasticity2D" ||
@@ -237,7 +238,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
 
       } else {
-        nullspace = MultiVectorFactory::Build(map, 1);
+        nullspace                    = MultiVectorFactory::Build(map, 1);
         Teuchos::ArrayRCP<SC> nsData = nullspace->getDataNonConst(0);
         for (int i = 0; i < nsData.size(); i++)
           nsData[i] = one;
@@ -247,7 +248,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       // Run different configurations
       // =========================================================================
       Teuchos::ArrayRCP<std::string> fileList = MueLuTests::TestHelpers::GetFileList(dirList[k],
-            (numProc == 1 ? std::string(".xml") : std::string("_np" + Teuchos::toString(numProc) + ".xml")));
+                                                                                     (numProc == 1 ? std::string(".xml") : std::string("_np" + Teuchos::toString(numProc) + ".xml")));
 
       RCP<MultiVector> X = MultiVectorFactory::Build(map, 1);
       RCP<MultiVector> B = MultiVectorFactory::Build(map, 1);
@@ -273,13 +274,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         // Reset (potentially) cached value of the estimate
         A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
 
-        std::string    solveType = paramList.get<std::string>   ("solver", "standalone");
-        double         goldRate  = paramList.get<double>        ("convergence rate");
-        Teuchos::ParameterList& mueluList = paramList.sublist            ("MueLu");
+        std::string solveType             = paramList.get<std::string>("solver", "standalone");
+        double goldRate                   = paramList.get<double>("convergence rate");
+        Teuchos::ParameterList& mueluList = paramList.sublist("MueLu");
 
         TEUCHOS_TEST_FOR_EXCEPTION(solveType != "standalone" && solveType != "cg" && solveType != "gmres", MueLu::Exceptions::RuntimeError,
                                    "Unknown solver type \"" << solveType << "\"");
-        bool           isPrec    = !(solveType == "standalone");
+        bool isPrec = !(solveType == "standalone");
 
         if (!mueluList.isParameter("verbosity"))
           mueluList.set("verbosity", "none");
@@ -298,21 +299,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
           H = mueluFactory.CreateHierarchy();
 
-          H->GetLevel(0)->Set("A",           A);
-          H->GetLevel(0)->Set("Nullspace",   nullspace);
+          H->GetLevel(0)->Set("A", A);
+          H->GetLevel(0)->Set("Nullspace", nullspace);
           H->GetLevel(0)->Set("Coordinates", coordinates);
 
           mueluFactory.SetupHierarchy(*H);
 
         } catch (Teuchos::ExceptionBase& e) {
           std::string msg = e.what();
-          msg = msg.substr(msg.find_last_of('\n')+1);
+          msg             = msg.substr(msg.find_last_of('\n') + 1);
 
           out << "Caught exception: " << msg << std::endl;
 
           if (msg == "Zoltan interface is not available" ||
               msg == "Zoltan2 interface is not available") {
-
             if (myRank == 0)
               out << xmlFile << ": skipped (missing library)" << std::endl;
 
@@ -331,31 +331,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
           Teuchos::Array<typename STS::magnitudeType> norms(1);
           B->norm2(norms);
-          B->scale(one/norms[0]);
+          B->scale(one / norms[0]);
           X->putScalar(zero);
         }
 
-        const int    maxIts = 1000;
+        const int maxIts = 1000;
         typedef typename STS::magnitudeType MagnitudeType;
-        const MagnitudeType tol    = 1e-12;
+        const MagnitudeType tol = 1e-12;
 
         H->IsPreconditioner(isPrec);
         if (isPrec == false) {
-          MueLu::ConvergenceStatus ret = H->Iterate(*B, *X, std::pair<LO,MagnitudeType>(maxIts, tol));
+          MueLu::ConvergenceStatus ret = H->Iterate(*B, *X, std::pair<LO, MagnitudeType>(maxIts, tol));
 
           double rate = H->GetRate();
 
-          if (std::abs(rate-goldRate) < 0.02) {
-            out << xmlFile << ": passed (" <<
-                (ret == MueLu::ConvergenceStatus::Converged ? "converged, " : "unconverged, ") <<
-                "expected rate = " << goldRate << ", real rate = " << rate <<
-                (ret == MueLu::ConvergenceStatus::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
+          if (std::abs(rate - goldRate) < 0.02) {
+            out << xmlFile << ": passed (" << (ret == MueLu::ConvergenceStatus::Converged ? "converged, " : "unconverged, ") << "expected rate = " << goldRate << ", real rate = " << rate << (ret == MueLu::ConvergenceStatus::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
                 << ")" << std::endl;
           } else {
-            out << xmlFile << ": failed (" <<
-                (ret == MueLu::ConvergenceStatus::Converged ? "converged, " : "unconverged, ") <<
-                "expected rate = " << goldRate << ", real rate = " << rate <<
-                (ret == MueLu::ConvergenceStatus::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
+            out << xmlFile << ": failed (" << (ret == MueLu::ConvergenceStatus::Converged ? "converged, " : "unconverged, ") << "expected rate = " << goldRate << ", real rate = " << rate << (ret == MueLu::ConvergenceStatus::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
                 << ")" << std::endl;
             // ap: we need to understand what's going on with the convergence rate
             // At the moment, disable failure state, so that the test passes as long
@@ -366,15 +360,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         } else {
 #ifdef HAVE_MUELU_BELOS
           // Operator and Multivector type that will be used with Belos
-          typedef MultiVector          MV;
+          typedef MultiVector MV;
           typedef Belos::OperatorT<MV> OP;
 
           // Define Operator and Preconditioner
-          RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-          RCP<OP> belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+          RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+          RCP<OP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
           // Construct a Belos LinearProblem object
-          RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+          RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
           belosProblem->setRightPrec(belosPrec);
 
           bool set = belosProblem->setProblem();
@@ -385,14 +379,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
           // Belos parameter list
           Teuchos::ParameterList belosList;
-          belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-          belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
+          belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+          belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
 #if 1
-          belosList.set("Verbosity",             Belos::Errors + Belos::Warnings);
+          belosList.set("Verbosity", Belos::Errors + Belos::Warnings);
 #else
-          belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-          belosList.set("Output Frequency",      1);
-          belosList.set("Output Style",          Belos::Brief);
+          belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+          belosList.set("Output Frequency", 1);
+          belosList.set("Output Style", Belos::Brief);
 #endif
 
           // Belos custom test to store residuals
@@ -402,7 +396,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
           if (solveType == "cg")
             solver = rcp(new Belos::PseudoBlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
           else if (solveType == "gmres")
-            solver = rcp(new Belos::BlockGmresSolMgr   <SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+            solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
           RCP<Belos::MyStatusTest<SC, MV, OP> > status = rcp(new Belos::MyStatusTest<SC, MV, OP>(tol));
           solver->setDebugStatusTest(status);
 
@@ -413,17 +407,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
             double rate = status->rate();
 
-            if (std::abs(rate-goldRate) < 0.02) {
-              out << xmlFile << ": passed (" <<
-                  (ret == Belos::Converged ? "converged, " : "unconverged, ") <<
-                  "expected rate = " << goldRate << ", real rate = " << rate <<
-                  (ret == Belos::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
+            if (std::abs(rate - goldRate) < 0.02) {
+              out << xmlFile << ": passed (" << (ret == Belos::Converged ? "converged, " : "unconverged, ") << "expected rate = " << goldRate << ", real rate = " << rate << (ret == Belos::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
                   << ")" << std::endl;
             } else {
-              out << xmlFile << ": failed (" <<
-                  (ret == Belos::Converged ? "converged, " : "unconverged, ") <<
-                  "expected rate = " << goldRate << ", real rate = " << rate <<
-                  (ret == Belos::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
+              out << xmlFile << ": failed (" << (ret == Belos::Converged ? "converged, " : "unconverged, ") << "expected rate = " << goldRate << ", real rate = " << rate << (ret == Belos::Converged ? "" : " (after " + Teuchos::toString(maxIts) + " iterations)")
                   << ")" << std::endl;
               // ap: we need to understand what's going on with the convergence rate
               // At the moment, disable failure state, so that the test passes as long
@@ -431,11 +419,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
               // failed = true;
             }
 
-          } catch(...) {
+          } catch (...) {
             out << xmlFile << ": failed (exception)" << std::endl;
             failed = true;
           }
-#endif //ifdef HAVE_MUELU_BELOS
+#endif  // ifdef HAVE_MUELU_BELOS
         }
       }
     }
@@ -445,23 +433,21 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-
+int main(int argc, char* argv[]) {
   Teuchos::RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out                = *fancy;
   out.setOutputToRootOnly(0);
 
-  bool status = Automatic_Test_ETI(argc,argv);
-  out << std::endl << "End Result: TEST " << (status ? "FAILED" : "PASSED") << std::endl;
+  bool status = Automatic_Test_ETI(argc, argv);
+  out << std::endl
+      << "End Result: TEST " << (status ? "FAILED" : "PASSED") << std::endl;
 
   return status;
 }
diff --git a/packages/muelu/test/factories/main.cpp b/packages/muelu/test/factories/main.cpp
index 83cb1ea88090..83dfa411df82 100644
--- a/packages/muelu/test/factories/main.cpp
+++ b/packages/muelu/test/factories/main.cpp
@@ -59,100 +59,107 @@
 #include "MueLu_UseDefaultTypes.hpp"
 #include <unistd.h>
 
-
 /**********************************************************************************/
 
 namespace MueLu {
-  /*!
-    @class MiniFactory classes.
-    @brief Mini factories for demonstration purposes only
-  */
-  class MiniFactory : public SingleLevelFactoryBase {
-  public:
-    MiniFactory(std::string factoryName, std::string dependsOn) { factoryName_ = factoryName; dependsOn_ = dependsOn; }
-    virtual ~MiniFactory() { }
-    RCP<const ParameterList> GetValidParameterList() const {
-      RCP<ParameterList> validParamList = rcp(new ParameterList());
-      validParamList->set< std::string >("internal-data", "", "internal data for factory");
-      validParamList->set< bool >("CallDependencyBuild", true, "If true, the dependencies are built");
-      validParamList->set< RCP<const FactoryBase> >(dependsOn_, Teuchos::null, "Generating factory of data this factory depends on");
-      return validParamList;
-    };
-    void DeclareInput(Level &Level) const {
-    	if (dependsOn_ != "")
-    	  Input(Level, dependsOn_);
-    };
-    void Build(Level& Level) const {
-    	std::cout << "MiniFactory" << factoryName_ <<"->Build" << std::endl;
-    	const ParameterList & pL = GetParameterList();
-        std::string data = pL.get<std::string>("internal-data");
-    	std::cout << "Data of MiniFactory" << factoryName_ << " = " << data << std::endl;
-
-    	// call build for dependencies:
-    	bool bBuild = pL.get<bool>("CallDependencyBuild");
-    	if(bBuild) {
-           std::string dep_data = Get< std::string >(Level, dependsOn_);
-    	   std::cout << "Content of variable " << dependsOn_ << " = " << dep_data << std::endl;
-    	}
-
-    	Set(Level, factoryName_, std::string("Data produced by MiniFactory" + factoryName_));
-    };
-  private:
-    std::string factoryName_;
-    std::string dependsOn_;
-  }; //class MiniFactory
-
-  /*!
-    @class MiniSwitchFactory classes.
-    @brief Mini factories for demonstration purposes only
-  */
-  class MiniSwitchFactory : public SingleLevelFactoryBase {
-  public:
-    MiniSwitchFactory(std::string factoryName, std::string dependsOnA, std::string dependsOnB) { factoryName_ = factoryName; dependsOnA_ = dependsOnA; dependsOnB_ = dependsOnB; }
-    virtual ~MiniSwitchFactory() { }
-    RCP<const ParameterList> GetValidParameterList() const {
-      RCP<ParameterList> validParamList = rcp(new ParameterList());
-      validParamList->set< std::string >("internal-data", "", "internal data for factory");
-      validParamList->set< bool >("UseA", true, "If true, the dependencies for A are built");
-      validParamList->set< bool >("UseB", true, "If true, the dependencies for B are built");
-      validParamList->set< RCP<const FactoryBase> >(dependsOnA_, Teuchos::null, "Generating factory of data this factory depends on");
-      validParamList->set< RCP<const FactoryBase> >(dependsOnB_, Teuchos::null, "Generating factory of data this factory depends on");
-      return validParamList;
-    };
-    void DeclareInput(Level &Level) const {
-    	if (dependsOnA_ != "")
-    	  Input(Level, dependsOnA_);
-    	if (dependsOnB_ != "")
-    	  Input(Level, dependsOnB_);
-    };
-    void Build(Level& Level) const {
-    	std::cout << "MiniFactory" << factoryName_ <<"->Build" << std::endl;
-    	const ParameterList & pL = GetParameterList();
-        std::string data = pL.get<std::string>("internal-data");
-    	std::cout << "Data of MiniFactory" << factoryName_ << " = " << data << std::endl;
-
-    	// call build for dependencies:
-    	bool bBuild = pL.get<bool>("UseA");
-    	if(bBuild) {
-           std::string dep_data = Get< std::string >(Level, dependsOnA_);
-    	   std::cout << "Content of variable " << dependsOnA_ << " = " << dep_data << std::endl;
-    	}
-    	bBuild = pL.get<bool>("UseB");
-    	if(bBuild) {
-           std::string dep_data = Get< std::string >(Level, dependsOnB_);
-    	   std::cout << "Content of variable " << dependsOnB_ << " = " << dep_data << std::endl;
-    	}
-
-    	Set(Level, factoryName_, std::string("Data produced by MiniFactory" + factoryName_));
-    };
-  private:
-    std::string factoryName_;
-    std::string dependsOnA_;
-    std::string dependsOnB_;
-  }; //class MiniSwitchFactory
-
-} //namespace MueLu
+/*!
+  @class MiniFactory classes.
+  @brief Mini factories for demonstration purposes only
+*/
+class MiniFactory : public SingleLevelFactoryBase {
+ public:
+  MiniFactory(std::string factoryName, std::string dependsOn) {
+    factoryName_ = factoryName;
+    dependsOn_   = dependsOn;
+  }
+  virtual ~MiniFactory() {}
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
+    validParamList->set<std::string>("internal-data", "", "internal data for factory");
+    validParamList->set<bool>("CallDependencyBuild", true, "If true, the dependencies are built");
+    validParamList->set<RCP<const FactoryBase> >(dependsOn_, Teuchos::null, "Generating factory of data this factory depends on");
+    return validParamList;
+  };
+  void DeclareInput(Level &Level) const {
+    if (dependsOn_ != "")
+      Input(Level, dependsOn_);
+  };
+  void Build(Level &Level) const {
+    std::cout << "MiniFactory" << factoryName_ << "->Build" << std::endl;
+    const ParameterList &pL = GetParameterList();
+    std::string data        = pL.get<std::string>("internal-data");
+    std::cout << "Data of MiniFactory" << factoryName_ << " = " << data << std::endl;
+
+    // call build for dependencies:
+    bool bBuild = pL.get<bool>("CallDependencyBuild");
+    if (bBuild) {
+      std::string dep_data = Get<std::string>(Level, dependsOn_);
+      std::cout << "Content of variable " << dependsOn_ << " = " << dep_data << std::endl;
+    }
+
+    Set(Level, factoryName_, std::string("Data produced by MiniFactory" + factoryName_));
+  };
+
+ private:
+  std::string factoryName_;
+  std::string dependsOn_;
+};  // class MiniFactory
+
+/*!
+  @class MiniSwitchFactory classes.
+  @brief Mini factories for demonstration purposes only
+*/
+class MiniSwitchFactory : public SingleLevelFactoryBase {
+ public:
+  MiniSwitchFactory(std::string factoryName, std::string dependsOnA, std::string dependsOnB) {
+    factoryName_ = factoryName;
+    dependsOnA_  = dependsOnA;
+    dependsOnB_  = dependsOnB;
+  }
+  virtual ~MiniSwitchFactory() {}
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
+    validParamList->set<std::string>("internal-data", "", "internal data for factory");
+    validParamList->set<bool>("UseA", true, "If true, the dependencies for A are built");
+    validParamList->set<bool>("UseB", true, "If true, the dependencies for B are built");
+    validParamList->set<RCP<const FactoryBase> >(dependsOnA_, Teuchos::null, "Generating factory of data this factory depends on");
+    validParamList->set<RCP<const FactoryBase> >(dependsOnB_, Teuchos::null, "Generating factory of data this factory depends on");
+    return validParamList;
+  };
+  void DeclareInput(Level &Level) const {
+    if (dependsOnA_ != "")
+      Input(Level, dependsOnA_);
+    if (dependsOnB_ != "")
+      Input(Level, dependsOnB_);
+  };
+  void Build(Level &Level) const {
+    std::cout << "MiniFactory" << factoryName_ << "->Build" << std::endl;
+    const ParameterList &pL = GetParameterList();
+    std::string data        = pL.get<std::string>("internal-data");
+    std::cout << "Data of MiniFactory" << factoryName_ << " = " << data << std::endl;
+
+    // call build for dependencies:
+    bool bBuild = pL.get<bool>("UseA");
+    if (bBuild) {
+      std::string dep_data = Get<std::string>(Level, dependsOnA_);
+      std::cout << "Content of variable " << dependsOnA_ << " = " << dep_data << std::endl;
+    }
+    bBuild = pL.get<bool>("UseB");
+    if (bBuild) {
+      std::string dep_data = Get<std::string>(Level, dependsOnB_);
+      std::cout << "Content of variable " << dependsOnB_ << " = " << dep_data << std::endl;
+    }
+
+    Set(Level, factoryName_, std::string("Data produced by MiniFactory" + factoryName_));
+  };
 
+ private:
+  std::string factoryName_;
+  std::string dependsOnA_;
+  std::string dependsOnB_;
+};  // class MiniSwitchFactory
+
+}  // namespace MueLu
 
 /**********************************************************************************/
 
@@ -160,11 +167,11 @@ int main(int argc, char *argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success  = true;
-  bool bSuccess = true; // collect success results
-  bool verbose = true;
+  bool bSuccess = true;  // collect success results
+  bool verbose  = true;
   try {
     Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
@@ -172,326 +179,395 @@ int main(int argc, char *argv[]) {
 
     // B depends on A
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		finest->Request("B",facB.get());
-		facB->Build(*finest);
-		//finest->print(std::cout,MueLu::Debug);
-		finest->Release(*facB);
-		//finest->print(std::cout,MueLu::Debug);
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      finest->Request("B", facB.get());
+      facB->Build(*finest);
+      // finest->print(std::cout,MueLu::Debug);
+      finest->Release(*facB);
+      // finest->print(std::cout,MueLu::Debug);
     }
 
     // B depends on A but does not use A
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-		facB->SetFactory("A", facA);
-
-		finest->Request("B",facB.get());
-		facB->Build(*finest);
-		//finest->print(std::cout,MueLu::Debug);
-		finest->Release(*facB);
-		//finest->print(std::cout,MueLu::Debug);
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+      facB->SetFactory("A", facA);
+
+      finest->Request("B", facB.get());
+      facB->Build(*finest);
+      // finest->print(std::cout,MueLu::Debug);
+      finest->Release(*facB);
+      // finest->print(std::cout,MueLu::Debug);
     }
 
     // C depends on B which depends on A
     //              B uses A
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-		facC->SetFactory("B", facB);
-
-		finest->Request("C",facC.get());
-		facC->Build(*finest);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facC);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+      facC->SetFactory("B", facB);
+
+      finest->Request("C", facC.get());
+      facC->Build(*finest);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facC);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
     }
 
     // C depends on B which depends on A
     // C uses       B which uses A
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		finest->Request("C",facC.get());
-		facC->Build(*finest);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facC);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		//finest->print(std::cout,MueLu::Debug);
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      finest->Request("C", facC.get());
+      facC->Build(*finest);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facC);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      // finest->print(std::cout,MueLu::Debug);
     }
 
     // C depends on B which depends on A
     // C uses       B
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		finest->Request("C",facC.get());
-		facC->Build(*finest);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facC);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      finest->Request("C", facC.get());
+      facC->Build(*finest);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facC);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
     }
 
     // D depends on A and C; C depends on B which depends on A
     {
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D","A","C"));
-		facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
-		facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
-		facD->SetParameter("UseB", Teuchos::ParameterEntry(true)); // B means only second factory
-		facD->SetFactory("A", facA);
-		facD->SetFactory("C", facC);
-
-		finest->Request("D",facD.get());
-		facD->Build(*finest);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facD);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-	}
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D", "A", "C"));
+      facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
+      facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
+      facD->SetParameter("UseB", Teuchos::ParameterEntry(true));  // B means only second factory
+      facD->SetFactory("A", facA);
+      facD->SetFactory("C", facC);
+
+      finest->Request("D", facD.get());
+      facD->Build(*finest);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facD);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+    }
 
     // D depends on B and C; C depends on B which depends on A
-	{
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D","B","C"));
-		facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
-		facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
-		facD->SetParameter("UseB", Teuchos::ParameterEntry(true)); // B means only second factory
-		facD->SetFactory("B", facB);
-		facD->SetFactory("C", facC);
-
-		finest->Request("D",facD.get());
-		facD->Build(*finest);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facD);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-	}
+    {
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D", "B", "C"));
+      facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
+      facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
+      facD->SetParameter("UseB", Teuchos::ParameterEntry(true));  // B means only second factory
+      facD->SetFactory("B", facB);
+      facD->SetFactory("C", facC);
+
+      finest->Request("D", facD.get());
+      facD->Build(*finest);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facD);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+    }
 
     // D depends on B and C; C depends on B which depends on A
-	// D only uses B (not C)
-	{
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D","B","C"));
-		facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
-		facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
-		facD->SetParameter("UseB", Teuchos::ParameterEntry(false)); // B means only second factory
-		facD->SetFactory("B", facB);
-		facD->SetFactory("C", facC);
-
-		finest->Request("D",facD.get());
-		facD->Build(*finest);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facD);
-		//finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-	}
+    // D only uses B (not C)
+    {
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D", "B", "C"));
+      facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
+      facD->SetParameter("UseA", Teuchos::ParameterEntry(true));
+      facD->SetParameter("UseB", Teuchos::ParameterEntry(false));  // B means only second factory
+      facD->SetFactory("B", facB);
+      facD->SetFactory("C", facC);
+
+      finest->Request("D", facD.get());
+      facD->Build(*finest);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facD);
+      // finest->print(std::cout,MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+    }
 
     // D depends on B and C; C depends on B which depends on A
-	// D only uses C (not B)
-	{
-		finest = Teuchos::null;
-		finest = Teuchos::rcp(new MueLu::Level());
-
-		Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A",""));
-		facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
-		facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
-
-		Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B","A"));
-		facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
-		facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facB->SetFactory("A", facA);
-
-		Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C","B"));
-		facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
-		facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
-		facC->SetFactory("B", facB);
-
-		Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D","B","C"));
-		facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
-		facD->SetParameter("UseA", Teuchos::ParameterEntry(false));
-		facD->SetParameter("UseB", Teuchos::ParameterEntry(true)); // B means only second factory
-		facD->SetFactory("B", facB);
-		facD->SetFactory("C", facC);
-
-		finest->Request("D",facD.get());
-		facD->Build(*finest);
-		finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("C",facC.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		finest->Release(*facD);
-		finest->print(std::cout,MueLu::Debug);
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C",facC.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B",facB.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A",facA.get()), false, std::cout, success); if (!success) bSuccess=false;
-		TEUCHOS_TEST_EQUALITY(finest->IsRequested("D",facD.get()), true,  std::cout, success); if (!success) bSuccess=false;
-	}
+    // D only uses C (not B)
+    {
+      finest = Teuchos::null;
+      finest = Teuchos::rcp(new MueLu::Level());
+
+      Teuchos::RCP<MueLu::MiniFactory> facA = Teuchos::rcp(new MueLu::MiniFactory("A", ""));
+      facA->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for A")));
+      facA->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(false));
+
+      Teuchos::RCP<MueLu::MiniFactory> facB = Teuchos::rcp(new MueLu::MiniFactory("B", "A"));
+      facB->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for B")));
+      facB->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facB->SetFactory("A", facA);
+
+      Teuchos::RCP<MueLu::MiniFactory> facC = Teuchos::rcp(new MueLu::MiniFactory("C", "B"));
+      facC->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for C")));
+      facC->SetParameter("CallDependencyBuild", Teuchos::ParameterEntry(true));
+      facC->SetFactory("B", facB);
+
+      Teuchos::RCP<MueLu::MiniSwitchFactory> facD = Teuchos::rcp(new MueLu::MiniSwitchFactory("D", "B", "C"));
+      facD->SetParameter("internal-data", Teuchos::ParameterEntry(std::string("data for D")));
+      facD->SetParameter("UseA", Teuchos::ParameterEntry(false));
+      facD->SetParameter("UseB", Teuchos::ParameterEntry(true));  // B means only second factory
+      facD->SetFactory("B", facB);
+      facD->SetFactory("C", facC);
+
+      finest->Request("D", facD.get());
+      facD->Build(*finest);
+      finest->print(std::cout, MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("C", facC.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      finest->Release(*facD);
+      finest->print(std::cout, MueLu::Debug);
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("C", facC.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("B", facB.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsAvailable("A", facA.get()), false, std::cout, success);
+      if (!success) bSuccess = false;
+      TEUCHOS_TEST_EQUALITY(finest->IsRequested("D", facD.get()), true, std::cout, success);
+      if (!success) bSuccess = false;
+    }
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, bSuccess);
 
-  return ( bSuccess ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (bSuccess ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/test/helmholtz/Helmholtz1D.cpp b/packages/muelu/test/helmholtz/Helmholtz1D.cpp
index 4d0ad4d8edc2..bc957b4414a8 100644
--- a/packages/muelu/test/helmholtz/Helmholtz1D.cpp
+++ b/packages/muelu/test/helmholtz/Helmholtz1D.cpp
@@ -74,22 +74,22 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  typedef Tpetra::Vector<SC,LO,GO,NO>              TVEC;
-  typedef Tpetra::MultiVector<SC,LO,GO,NO>         TMV;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>           TCRS;
-  typedef Xpetra::CrsMatrix<SC,LO,GO,NO>           XCRS;
-  typedef Xpetra::TpetraCrsMatrix<SC,LO,GO,NO>     XTCRS;
-  typedef Xpetra::Matrix<SC,LO,GO,NO>              XMAT;
-  typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO>       XWRAP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TCRS;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> XCRS;
+  typedef Xpetra::TpetraCrsMatrix<SC, LO, GO, NO> XTCRS;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> XMAT;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> XWRAP;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-  typedef Belos::OperatorT<TMV>                    TOP;
-  typedef Belos::OperatorTraits<SC,TMV,TOP>        TOPT;
-  typedef Belos::MultiVecTraits<SC,TMV>            TMVT;
-  typedef Belos::LinearProblem<SC,TMV,TOP>         TProblem;
-  typedef Belos::SolverManager<SC,TMV,TOP>         TBelosSolver;
-  typedef Belos::BlockGmresSolMgr<SC,TMV,TOP>      TBelosGMRES;
+  typedef Belos::OperatorT<TMV> TOP;
+  typedef Belos::OperatorTraits<SC, TMV, TOP> TOPT;
+  typedef Belos::MultiVecTraits<SC, TMV> TMVT;
+  typedef Belos::LinearProblem<SC, TMV, TOP> TProblem;
+  typedef Belos::SolverManager<SC, TMV, TOP> TBelosSolver;
+  typedef Belos::BlockGmresSolMgr<SC, TMV, TOP> TBelosGMRES;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -100,26 +100,36 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     Teuchos::CommandLineProcessor clp(false);
 
-    GO nx,ny,nz;
-    nx=100; ny=100; nz=100;
+    GO nx, ny, nz;
+    nx = 100;
+    ny = 100;
+    nz = 100;
     double stretchx, stretchy, stretchz, h, delta;
-    stretchx=1.0; stretchy=1.0; stretchz=1.0;
-    h=0.01; delta=2.0;
+    stretchx = 1.0;
+    stretchy = 1.0;
+    stretchz = 1.0;
+    h        = 0.01;
+    delta    = 2.0;
     int PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR;
-    PMLXL=10; PMLXR=10; PMLYL=10; PMLYR=10; PMLZL=10; PMLZR=10;
+    PMLXL = 10;
+    PMLXR = 10;
+    PMLYL = 10;
+    PMLYR = 10;
+    PMLZL = 10;
+    PMLZR = 10;
     double omega, shift;
-    omega=20.0*M_PI;
-    shift=0.5;
+    omega = 20.0 * M_PI;
+    shift = 0.5;
 
     Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Helmholtz1D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift);
-    Xpetra::Parameters             xpetraParameters(clp);
+                                                    h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift);
+    Xpetra::Parameters xpetraParameters(clp);
 
-    RCP<TimeMonitor> globalTimeMonitor = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
-    RCP<TimeMonitor> tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
+    RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
+    RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
     Teuchos::ParameterList pl = matrixParameters.GetParameterList();
     RCP<RealValuedMultiVector> coordinates;
@@ -129,7 +139,7 @@ int main(int argc, char *argv[]) {
     galeriList.set("nz", pl.get("nz", nz));
     RCP<const Map> map;
 
-    map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+    map         = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
     coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", map, matrixParameters.GetParameterList());
 
     RCP<const Tpetra::Map<LO, GO, NO> > tmap = Xpetra::toTpetra(map);
@@ -137,54 +147,55 @@ int main(int argc, char *argv[]) {
     Teuchos::ParameterList matrixParams = matrixParameters.GetParameterList();
 
     // Build problem
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParams);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParams);
     RCP<Matrix> A = Pr->BuildMatrix();
 
-    RCP<MultiVector> nullspace = MultiVectorFactory::Build(map,1);
-    nullspace->putScalar( (SC) 1.0);
+    RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
+    nullspace->putScalar((SC)1.0);
 
     comm->barrier();
 
     tm = Teuchos::null;
 
     // Construct a multigrid preconditioner
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
 
     // Multigrid Hierarchy
     RCP<Hierarchy> H = rcp(new Hierarchy(A));
-    H->GetLevel(0)->Set("Nullspace",nullspace);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
     FactoryManager Manager;
     H->Setup(Manager, 0, 5);
-    //H->Write(-1,-1);
+    // H->Write(-1,-1);
 
     tm = Teuchos::null;
 
     // Solve Ax = b
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
-    RCP<TVEC> X = Tpetra::createVector<SC,LO,GO,NO>(tmap);
-    RCP<TVEC> B = Tpetra::createVector<SC,LO,GO,NO>(tmap);
-    X->putScalar((SC) 0.0);
-    B->putScalar((SC) 0.0);
-    if(comm->getRank()==0) {
-      B->replaceGlobalValue(0, (SC) 1.0);
+    tm          = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
+    RCP<TVEC> X = Tpetra::createVector<SC, LO, GO, NO>(tmap);
+    RCP<TVEC> B = Tpetra::createVector<SC, LO, GO, NO>(tmap);
+    X->putScalar((SC)0.0);
+    B->putScalar((SC)0.0);
+    if (comm->getRank() == 0) {
+      B->replaceGlobalValue(0, (SC)1.0);
     }
 
     tm = Teuchos::null;
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
 
     // Define Operator and Preconditioner
-    RCP<TOP> belosOp   = rcp(new Belos::XpetraOp<SC,LO,GO,NO> (A) );    // Turns a Xpetra::Matrix object into a Belos operator
-    RCP<TOP> belosPrec = rcp(new Belos::MueLuOp<SC,LO,GO,NO>  (H) );    // Turns a MueLu::Hierarchy object into a Belos operator
+    RCP<TOP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+    RCP<TOP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
-    RCP<TProblem> belosProblem = rcp(new TProblem(belosOp,X,B));
+    RCP<TProblem> belosProblem = rcp(new TProblem(belosOp, X, B));
     belosProblem->setRightPrec(belosPrec);
     bool set = belosProblem->setProblem();
     if (set == false) {
-      if(comm->getRank()==0)
-        std::cout << std::endl << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
+      if (comm->getRank() == 0)
+        std::cout << std::endl
+                  << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
       return EXIT_FAILURE;
     }
 
@@ -192,31 +203,33 @@ int main(int argc, char *argv[]) {
     int maxIts = 100;
     double tol = 1e-6;
     Teuchos::ParameterList belosList;
-    belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-    belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-    belosList.set("Flexible Gmres", false);         // set flexible GMRES on/off
+    belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+    belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+    belosList.set("Flexible Gmres", false);       // set flexible GMRES on/off
     belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    belosList.set("Output Frequency",1);
-    belosList.set("Output Style",Belos::Brief);
+    belosList.set("Output Frequency", 1);
+    belosList.set("Output Style", Belos::Brief);
 
     // Create solver manager
-    RCP<TBelosSolver> solver = rcp( new TBelosGMRES(belosProblem, rcp(&belosList, false)) );
+    RCP<TBelosSolver> solver = rcp(new TBelosGMRES(belosProblem, rcp(&belosList, false)));
 
     // Perform solve
-    Belos::ReturnType ret=Belos::Unconverged;
-    ret = solver->solve();
+    Belos::ReturnType ret = Belos::Unconverged;
+    ret                   = solver->solve();
     if (comm->getRank() == 0)
       std::cout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
     // Check convergence
     if (ret != Belos::Converged) {
-      if (comm->getRank() == 0) std::cout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "ERROR:  Belos did not converge! " << std::endl;
     } else {
-      if (comm->getRank() == 0) std::cout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "SUCCESS:  Belos converged!" << std::endl;
     }
 
     // Get the number of iterations for this solve.
-    if(comm->getRank()==0)
+    if (comm->getRank() == 0)
       std::cout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
     tm = Teuchos::null;
@@ -229,5 +242,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
diff --git a/packages/muelu/test/helmholtz/Helmholtz2D.cpp b/packages/muelu/test/helmholtz/Helmholtz2D.cpp
index b1c503e10bd1..333139e3e849 100644
--- a/packages/muelu/test/helmholtz/Helmholtz2D.cpp
+++ b/packages/muelu/test/helmholtz/Helmholtz2D.cpp
@@ -63,15 +63,15 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  typedef Tpetra::Vector<SC,LO,GO,NO>              TVEC;
-  typedef Tpetra::MultiVector<SC,LO,GO,NO>         TMV;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>           TCRS;
-  typedef Xpetra::CrsMatrix<SC,LO,GO,NO>           XCRS;
-  typedef Xpetra::TpetraCrsMatrix<SC,LO,GO,NO>     XTCRS;
-  typedef Xpetra::Matrix<SC,LO,GO,NO>              XMAT;
-  typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO>       XWRAP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TCRS;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> XCRS;
+  typedef Xpetra::TpetraCrsMatrix<SC, LO, GO, NO> XTCRS;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> XMAT;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> XWRAP;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -88,7 +88,7 @@ int main(int argc, char *argv[]) {
     //   Galeri Parameters   //
     //***********************//
 
-    GO nx,ny,nz;
+    GO nx, ny, nz;
     GO mx, my, mz;
     double stretchx, stretchy, stretchz, h, delta;
     int PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR;
@@ -97,37 +97,37 @@ int main(int argc, char *argv[]) {
 
     std::ifstream inputfile;
     inputfile.open("helm2D.inp");
-    inputfile >> nx       >> ny       >> nz ;
-    if(comm->getRank()==0)
-      std::cout<<"nx: "<<nx<<"  ny: "<<ny<<"  nz: "<<nz<<std::endl;
-    inputfile >> stretchx >> stretchy >> stretchz ;
-    if(comm->getRank()==0)
-      std::cout<<"stretchx: "<<stretchx<<"  stretchy: "<<stretchy<<"  stretchz: "<<stretchz<<std::endl;
-    inputfile >> h        >> delta ;
-    if(comm->getRank()==0)
-      std::cout<<"h: "<<h<<"  delta: "<<delta<<std::endl;
-    inputfile >> PMLXL    >> PMLXR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLXL: "<<PMLXL<<"  PMLXR: "<<PMLXR<<std::endl;
-    inputfile >> PMLYL    >> PMLYR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLYL: "<<PMLYL<<"  PMLYR: "<<PMLYR<<std::endl;
-    inputfile >> PMLZL    >> PMLZR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLZL: "<<PMLZL<<"  PMLZR: "<<PMLZR<<std::endl;
-    inputfile >> omega    >> shift ;
-    if(comm->getRank()==0)
-      std::cout<<"omega: "<<omega<<"  shift: "<<shift<<std::endl;
-    inputfile >> mx       >> my       >> mz ;
-    if(comm->getRank()==0)
-      std::cout<<"mx: "<<mx<<"  my: "<<my<<"  mz: "<<mz<<std::endl;
-    inputfile >> model ;
-    if(comm->getRank()==0)
-      std::cout<<"velocity model: "<<model<<std::endl;
+    inputfile >> nx >> ny >> nz;
+    if (comm->getRank() == 0)
+      std::cout << "nx: " << nx << "  ny: " << ny << "  nz: " << nz << std::endl;
+    inputfile >> stretchx >> stretchy >> stretchz;
+    if (comm->getRank() == 0)
+      std::cout << "stretchx: " << stretchx << "  stretchy: " << stretchy << "  stretchz: " << stretchz << std::endl;
+    inputfile >> h >> delta;
+    if (comm->getRank() == 0)
+      std::cout << "h: " << h << "  delta: " << delta << std::endl;
+    inputfile >> PMLXL >> PMLXR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLXL: " << PMLXL << "  PMLXR: " << PMLXR << std::endl;
+    inputfile >> PMLYL >> PMLYR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLYL: " << PMLYL << "  PMLYR: " << PMLYR << std::endl;
+    inputfile >> PMLZL >> PMLZR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLZL: " << PMLZL << "  PMLZR: " << PMLZR << std::endl;
+    inputfile >> omega >> shift;
+    if (comm->getRank() == 0)
+      std::cout << "omega: " << omega << "  shift: " << shift << std::endl;
+    inputfile >> mx >> my >> mz;
+    if (comm->getRank() == 0)
+      std::cout << "mx: " << mx << "  my: " << my << "  mz: " << mz << std::endl;
+    inputfile >> model;
+    if (comm->getRank() == 0)
+      std::cout << "velocity model: " << model << std::endl;
 
     Galeri::Xpetra::Parameters<GO> matrixParameters_aux(clp, nx, ny, nz, "Helmholtz2D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
-    Xpetra::Parameters             xpetraParameters(clp);
+                                                        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
+    Xpetra::Parameters xpetraParameters(clp);
 
     //****************************************//
     //   Setup Galeri Problems and Matrices   //
@@ -141,39 +141,38 @@ int main(int argc, char *argv[]) {
     galeriList.set("nz", pl.get("nz", nz));
     RCP<const Map> map;
 
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
     coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("2D", map, matrixParameters_aux.GetParameterList());
 
     RCP<const Tpetra::Map<LO, GO, NO> > tmap = Xpetra::toTpetra(map);
 
-    Teuchos::ParameterList matrixParams_aux       = matrixParameters_aux.GetParameterList();
+    Teuchos::ParameterList matrixParams_aux = matrixParameters_aux.GetParameterList();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_aux =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_aux.GetMatrixType(), map, matrixParams_aux);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_aux =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_aux.GetMatrixType(), map, matrixParams_aux);
     RCP<Matrix> Kmat, Mmat;
-    std::pair< RCP<Matrix>, RCP<Matrix> > system = Pr_aux->BuildMatrices();
-    Kmat = system.first;
-    Mmat = system.second;
+    std::pair<RCP<Matrix>, RCP<Matrix> > system = Pr_aux->BuildMatrices();
+    Kmat                                        = system.first;
+    Mmat                                        = system.second;
 
     //******************************************************************//
     //   Initialize Shifted Laplacian Preconditioner and Belos Solver   //
     //******************************************************************//
 
     // Initialize Shifted Laplacian with the stiffness matrix
-    RCP<ShiftedLaplacian> SLSolver = rcp( new ShiftedLaplacian );
-    SLSolver -> setstiff(Kmat);
-    SLSolver -> initialize();
+    RCP<ShiftedLaplacian> SLSolver = rcp(new ShiftedLaplacian);
+    SLSolver->setstiff(Kmat);
+    SLSolver->initialize();
 
     // vector of frequencies
     std::vector<double> omegas;
     omegas.push_back(omega);
-    omegas.push_back(omega+10);
-    omegas.push_back(omega+20);
+    omegas.push_back(omega + 10);
+    omegas.push_back(omega + 20);
     int total_iters = 0;
 
     // loop over frequencies
-    for(size_t i=0; i<omegas.size(); i++) {
-
+    for (size_t i = 0; i < omegas.size(); i++) {
       Galeri::Xpetra::Parameters<GO> matrixParameters_helmholtz(clp, nx, ny, nz, "Helmholtz2D", 0, stretchx, stretchy, stretchz,
                                                                 h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omegas[i], 0.0, mx, my, mz, model);
       Galeri::Xpetra::Parameters<GO> matrixParameters_shifted(clp, nx, ny, nz, "Helmholtz2D", 0, stretchx, stretchy, stretchz,
@@ -182,60 +181,58 @@ int main(int argc, char *argv[]) {
       Teuchos::ParameterList matrixParams_helmholtz = matrixParameters_helmholtz.GetParameterList();
       Teuchos::ParameterList matrixParams_shifted   = matrixParameters_shifted.GetParameterList();
 
-      RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_helmholtz =
-        Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
+      RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_helmholtz =
+          Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
       RCP<Matrix> Amat = Pr_helmholtz->BuildMatrix();
 
-      RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_shifted =
-        Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
+      RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_shifted =
+          Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
       RCP<Matrix> Pmat = Pr_shifted->BuildMatrix();
 
       // set Helmholtz operator (Amat), Shifted Laplacian operator (Pmat),
       // and mass matrix (Mmat)
-      SLSolver -> setmass(Mmat);
-      SLSolver -> setProblemMatrix(Amat);
-      SLSolver -> setPreconditioningMatrix(Pmat);
+      SLSolver->setmass(Mmat);
+      SLSolver->setProblemMatrix(Amat);
+      SLSolver->setPreconditioningMatrix(Pmat);
       // determine shifts for RAPShiftFactory
       std::vector<SC> shifts;
-      int maxLevels=5;
-      for(int j=0; j<maxLevels; j++) {
-        double alpha=1.0;
-        double beta=shift+((double) j)*0.2;
-        SC curshift(alpha,beta);
+      int maxLevels = 5;
+      for (int j = 0; j < maxLevels; j++) {
+        double alpha = 1.0;
+        double beta  = shift + ((double)j) * 0.2;
+        SC curshift(alpha, beta);
         shifts.push_back(-curshift);
       }
-      SLSolver -> setLevelShifts(shifts);
+      SLSolver->setLevelShifts(shifts);
       // setup hierarchy with variable complex shifts
-      SLSolver -> setupSlowRAP();
+      SLSolver->setupSlowRAP();
 
       // solve for RHS
-      RCP<TMV> X = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-      RCP<TMV> B = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-      X->putScalar((SC) 0.0);
-      B->putScalar((SC) 0.0);
-      int pointsourceid=nx*ny/2+nx/2;
-      if(map->isNodeGlobalElement(pointsourceid)==true) {
-        B->replaceGlobalValue(pointsourceid, 0, (SC) 1.0);
+      RCP<TMV> X = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+      RCP<TMV> B = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+      X->putScalar((SC)0.0);
+      B->putScalar((SC)0.0);
+      int pointsourceid = nx * ny / 2 + nx / 2;
+      if (map->isNodeGlobalElement(pointsourceid) == true) {
+        B->replaceGlobalValue(pointsourceid, 0, (SC)1.0);
       }
-      SLSolver -> solve(B,X);
-      
+      SLSolver->solve(B, X);
+
       // sum up number of iterations
       total_iters += SLSolver->GetIterations();
 
-#     ifdef HAVE_MUELU_DEBUG
+#ifdef HAVE_MUELU_DEBUG
       SLSolver->Manager_->ResetDebugData();
-#     endif
-      
+#endif
     }
 
-    if(total_iters<=110)
+    if (total_iters <= 110)
       success = true;
     else
       success = false;
-
   }
 
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
diff --git a/packages/muelu/test/helmholtz/Helmholtz3D.cpp b/packages/muelu/test/helmholtz/Helmholtz3D.cpp
index 5018b7d83305..72a80db35eb8 100644
--- a/packages/muelu/test/helmholtz/Helmholtz3D.cpp
+++ b/packages/muelu/test/helmholtz/Helmholtz3D.cpp
@@ -61,15 +61,15 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  typedef Tpetra::Vector<SC,LO,GO,NO>              TVEC;
-  typedef Tpetra::MultiVector<SC,LO,GO,NO>         TMV;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>           TCRS;
-  typedef Xpetra::CrsMatrix<SC,LO,GO,NO>           XCRS;
-  typedef Xpetra::TpetraCrsMatrix<SC,LO,GO,NO>     XTCRS;
-  typedef Xpetra::Matrix<SC,LO,GO,NO>              XMAT;
-  typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO>       XWRAP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TCRS;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> XCRS;
+  typedef Xpetra::TpetraCrsMatrix<SC, LO, GO, NO> XTCRS;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> XMAT;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> XWRAP;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -80,14 +80,14 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     Teuchos::CommandLineProcessor clp(false);
 
     //***********************//
     //   Galeri Parameters   //
     //***********************//
 
-    GO nx,ny,nz;
+    GO nx, ny, nz;
     GO mx, my, mz;
     double stretchx, stretchy, stretchz, h, delta;
     int PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR;
@@ -96,47 +96,46 @@ int main(int argc, char *argv[]) {
 
     std::ifstream inputfile;
     inputfile.open("helm3D.inp");
-    inputfile >> nx       >> ny       >> nz ;
-    if(comm->getRank()==0)
-      std::cout<<"nx: "<<nx<<"  ny: "<<ny<<"  nz: "<<nz<<std::endl;
-    inputfile >> stretchx >> stretchy >> stretchz ;
-    if(comm->getRank()==0)
-      std::cout<<"stretchx: "<<stretchx<<"  stretchy: "<<stretchy<<"  stretchz: "<<stretchz<<std::endl;
-    inputfile >> h        >> delta ;
-    if(comm->getRank()==0)
-      std::cout<<"h: "<<h<<"  delta: "<<delta<<std::endl;
-    inputfile >> PMLXL    >> PMLXR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLXL: "<<PMLXL<<"  PMLXR: "<<PMLXR<<std::endl;
-    inputfile >> PMLYL    >> PMLYR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLYL: "<<PMLYL<<"  PMLYR: "<<PMLYR<<std::endl;
-    inputfile >> PMLZL    >> PMLZR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLZL: "<<PMLZL<<"  PMLZR: "<<PMLZR<<std::endl;
-    inputfile >> omega    >> shift ;
-    if(comm->getRank()==0)
-      std::cout<<"omega: "<<omega<<"  shift: "<<shift<<std::endl;
-    inputfile >> mx       >> my       >> mz ;
-    if(comm->getRank()==0)
-      std::cout<<"mx: "<<mx<<"  my: "<<my<<"  mz: "<<mz<<std::endl;
-    inputfile >> model ;
-    if(comm->getRank()==0)
-      std::cout<<"velocity model: "<<model<<std::endl;
+    inputfile >> nx >> ny >> nz;
+    if (comm->getRank() == 0)
+      std::cout << "nx: " << nx << "  ny: " << ny << "  nz: " << nz << std::endl;
+    inputfile >> stretchx >> stretchy >> stretchz;
+    if (comm->getRank() == 0)
+      std::cout << "stretchx: " << stretchx << "  stretchy: " << stretchy << "  stretchz: " << stretchz << std::endl;
+    inputfile >> h >> delta;
+    if (comm->getRank() == 0)
+      std::cout << "h: " << h << "  delta: " << delta << std::endl;
+    inputfile >> PMLXL >> PMLXR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLXL: " << PMLXL << "  PMLXR: " << PMLXR << std::endl;
+    inputfile >> PMLYL >> PMLYR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLYL: " << PMLYL << "  PMLYR: " << PMLYR << std::endl;
+    inputfile >> PMLZL >> PMLZR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLZL: " << PMLZL << "  PMLZR: " << PMLZR << std::endl;
+    inputfile >> omega >> shift;
+    if (comm->getRank() == 0)
+      std::cout << "omega: " << omega << "  shift: " << shift << std::endl;
+    inputfile >> mx >> my >> mz;
+    if (comm->getRank() == 0)
+      std::cout << "mx: " << mx << "  my: " << my << "  mz: " << mz << std::endl;
+    inputfile >> model;
+    if (comm->getRank() == 0)
+      std::cout << "velocity model: " << model << std::endl;
 
     Galeri::Xpetra::Parameters<GO> matrixParameters_helmholtz(clp, nx, ny, nz, "Helmholtz3D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
+                                                              h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
     Galeri::Xpetra::Parameters<GO> matrixParameters_shifted(clp, nx, ny, nz, "Helmholtz3D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
-    Xpetra::Parameters             xpetraParameters(clp);
-
+                                                            h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
+    Xpetra::Parameters xpetraParameters(clp);
 
     //****************************************//
     //   Setup Galeri Problems and Matrices   //
     //****************************************//
 
-    RCP<TimeMonitor> globalTimeMonitor = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
-    RCP<TimeMonitor> tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
+    RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
+    RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
     Teuchos::ParameterList pl = matrixParameters_helmholtz.GetParameterList();
     RCP<RealValuedMultiVector> coordinates;
@@ -146,7 +145,7 @@ int main(int argc, char *argv[]) {
     galeriList.set("nz", pl.get("nz", nz));
     RCP<const Map> map;
 
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
     coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", map, matrixParameters_helmholtz.GetParameterList());
 
     RCP<const Tpetra::Map<LO, GO, NO> > tmap = Xpetra::toTpetra(map);
@@ -154,12 +153,12 @@ int main(int argc, char *argv[]) {
     Teuchos::ParameterList matrixParams_helmholtz = matrixParameters_helmholtz.GetParameterList();
     Teuchos::ParameterList matrixParams_shifted   = matrixParameters_shifted.GetParameterList();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_helmholtz =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_helmholtz =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
     RCP<Matrix> Amat = Pr_helmholtz->BuildMatrix();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_shifted =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_shifted =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
     RCP<Matrix> Pmat = Pr_shifted->BuildMatrix();
 
     comm->barrier();
@@ -170,16 +169,16 @@ int main(int argc, char *argv[]) {
     //   Setup Shifted Laplacian Preconditioner and Belos Solver   //
     //*************************************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
 
     // initialize with the Shifted Laplacian operator (Pmat)
-    RCP<ShiftedLaplacian> SLSolver = rcp( new ShiftedLaplacian );
-    SLSolver -> setPreconditioningMatrix(Pmat);
-    SLSolver -> initialize();
+    RCP<ShiftedLaplacian> SLSolver = rcp(new ShiftedLaplacian);
+    SLSolver->setPreconditioningMatrix(Pmat);
+    SLSolver->initialize();
     // set Helmholtz operator (Amat)
-    SLSolver -> setProblemMatrix(Amat);
+    SLSolver->setProblemMatrix(Amat);
     // setup with SL operator
-    SLSolver -> setupNormalRAP();
+    SLSolver->setupNormalRAP();
 
     tm = Teuchos::null;
 
@@ -187,22 +186,22 @@ int main(int argc, char *argv[]) {
     //   Solve linear system with Belos   //
     //************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
 
-    RCP<TMV> X = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    RCP<TMV> B = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    X->putScalar((SC) 0.0);
-    B->putScalar((SC) 0.0);
-    int pointsourceid=nx*ny*nz/2+nx*ny/2+nx/2;
-    if(map->isNodeGlobalElement(pointsourceid)==true) {
-      B->replaceGlobalValue(pointsourceid, 0, (SC) 1.0);
+    RCP<TMV> X = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    RCP<TMV> B = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    X->putScalar((SC)0.0);
+    B->putScalar((SC)0.0);
+    int pointsourceid = nx * ny * nz / 2 + nx * ny / 2 + nx / 2;
+    if (map->isNodeGlobalElement(pointsourceid) == true) {
+      B->replaceGlobalValue(pointsourceid, 0, (SC)1.0);
     }
 
     tm = Teuchos::null;
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
 
-    SLSolver -> solve(B,X);
+    SLSolver->solve(B, X);
 
     tm = Teuchos::null;
 
@@ -214,5 +213,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
diff --git a/packages/muelu/test/helmholtz/HelmholtzFEM2D.cpp b/packages/muelu/test/helmholtz/HelmholtzFEM2D.cpp
index 4bc38d9bde38..957be7a5b54b 100644
--- a/packages/muelu/test/helmholtz/HelmholtzFEM2D.cpp
+++ b/packages/muelu/test/helmholtz/HelmholtzFEM2D.cpp
@@ -64,15 +64,15 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  typedef Tpetra::Vector<SC,LO,GO,NO>              TVEC;
-  typedef Tpetra::MultiVector<SC,LO,GO,NO>         TMV;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>           TCRS;
-  typedef Xpetra::CrsMatrix<SC,LO,GO,NO>           XCRS;
-  typedef Xpetra::TpetraCrsMatrix<SC,LO,GO,NO>     XTCRS;
-  typedef Xpetra::Matrix<SC,LO,GO,NO>              XMAT;
-  typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO>       XWRAP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TCRS;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> XCRS;
+  typedef Xpetra::TpetraCrsMatrix<SC, LO, GO, NO> XTCRS;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> XMAT;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> XWRAP;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -83,7 +83,7 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     Teuchos::CommandLineProcessor clp(false);
 
     //***********************//
@@ -99,46 +99,46 @@ int main(int argc, char *argv[]) {
 
     std::ifstream inputfile;
     inputfile.open("helm2D.inp");
-    inputfile >> nx       >> ny       >> nz ;
-    if(comm->getRank()==0)
-      std::cout<<"nx: "<<nx<<"  ny: "<<ny<<"  nz: "<<nz<<std::endl;
-    inputfile >> stretchx >> stretchy >> stretchz ;
-    if(comm->getRank()==0)
-      std::cout<<"stretchx: "<<stretchx<<"  stretchy: "<<stretchy<<"  stretchz: "<<stretchz<<std::endl;
-    inputfile >> h        >> delta ;
-    if(comm->getRank()==0)
-      std::cout<<"h: "<<h<<"  delta: "<<delta<<std::endl;
-    inputfile >> PMLXL    >> PMLXR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLXL: "<<PMLXL<<"  PMLXR: "<<PMLXR<<std::endl;
-    inputfile >> PMLYL    >> PMLYR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLYL: "<<PMLYL<<"  PMLYR: "<<PMLYR<<std::endl;
-    inputfile >> PMLZL    >> PMLZR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLZL: "<<PMLZL<<"  PMLZR: "<<PMLZR<<std::endl;
-    inputfile >> omega    >> shift ;
-    if(comm->getRank()==0)
-      std::cout<<"omega: "<<omega<<"  shift: "<<shift<<std::endl;
-    inputfile >> mx       >> my       >> mz ;
-    if(comm->getRank()==0)
-      std::cout<<"mx: "<<mx<<"  my: "<<my<<"  mz: "<<mz<<std::endl;
-    inputfile >> model ;
-    if(comm->getRank()==0)
-      std::cout<<"velocity model: "<<model<<std::endl;
+    inputfile >> nx >> ny >> nz;
+    if (comm->getRank() == 0)
+      std::cout << "nx: " << nx << "  ny: " << ny << "  nz: " << nz << std::endl;
+    inputfile >> stretchx >> stretchy >> stretchz;
+    if (comm->getRank() == 0)
+      std::cout << "stretchx: " << stretchx << "  stretchy: " << stretchy << "  stretchz: " << stretchz << std::endl;
+    inputfile >> h >> delta;
+    if (comm->getRank() == 0)
+      std::cout << "h: " << h << "  delta: " << delta << std::endl;
+    inputfile >> PMLXL >> PMLXR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLXL: " << PMLXL << "  PMLXR: " << PMLXR << std::endl;
+    inputfile >> PMLYL >> PMLYR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLYL: " << PMLYL << "  PMLYR: " << PMLYR << std::endl;
+    inputfile >> PMLZL >> PMLZR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLZL: " << PMLZL << "  PMLZR: " << PMLZR << std::endl;
+    inputfile >> omega >> shift;
+    if (comm->getRank() == 0)
+      std::cout << "omega: " << omega << "  shift: " << shift << std::endl;
+    inputfile >> mx >> my >> mz;
+    if (comm->getRank() == 0)
+      std::cout << "mx: " << mx << "  my: " << my << "  mz: " << mz << std::endl;
+    inputfile >> model;
+    if (comm->getRank() == 0)
+      std::cout << "velocity model: " << model << std::endl;
 
     Galeri::Xpetra::Parameters<GO> matrixParameters_helmholtz(clp, nx, ny, nz, "HelmholtzFEM2D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
+                                                              h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
     Galeri::Xpetra::Parameters<GO> matrixParameters_shifted(clp, nx, ny, nz, "HelmholtzFEM2D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
-    Xpetra::Parameters             xpetraParameters(clp);
+                                                            h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
+    Xpetra::Parameters xpetraParameters(clp);
 
     //****************************************//
     //   Setup Galeri Problems and Matrices   //
     //****************************************//
 
-    RCP<TimeMonitor> globalTimeMonitor = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
-    RCP<TimeMonitor> tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
+    RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
+    RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
     Teuchos::ParameterList pl = matrixParameters_helmholtz.GetParameterList();
     RCP<RealValuedMultiVector> coordinates;
@@ -148,7 +148,7 @@ int main(int argc, char *argv[]) {
     galeriList.set("nz", pl.get("nz", nz));
     RCP<const Map> map;
 
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
     coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("2D", map, matrixParameters_helmholtz.GetParameterList());
 
     RCP<const Tpetra::Map<LO, GO, NO> > tmap = Xpetra::toTpetra(map);
@@ -156,16 +156,16 @@ int main(int argc, char *argv[]) {
     Teuchos::ParameterList matrixParams_helmholtz = matrixParameters_helmholtz.GetParameterList();
     Teuchos::ParameterList matrixParams_shifted   = matrixParameters_shifted.GetParameterList();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_helmholtz =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_helmholtz =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
     RCP<Matrix> Kmat, Mmat;
-    std::pair< RCP<Matrix>, RCP<Matrix> > system = Pr_helmholtz->BuildMatrices();
-    Kmat = system.first;
-    Mmat = system.second;
-    RCP<Matrix> Amat = Pr_helmholtz->BuildMatrix();
+    std::pair<RCP<Matrix>, RCP<Matrix> > system = Pr_helmholtz->BuildMatrices();
+    Kmat                                        = system.first;
+    Mmat                                        = system.second;
+    RCP<Matrix> Amat                            = Pr_helmholtz->BuildMatrix();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_shifted =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_shifted =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
     RCP<Matrix> Pmat = Pr_shifted->BuildMatrix();
 
     comm->barrier();
@@ -176,15 +176,15 @@ int main(int argc, char *argv[]) {
     //   Setup Shifted Laplacian Preconditioner and Belos Solver   //
     //*************************************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
 
-    RCP<ShiftedLaplacian> SLSolver = rcp( new ShiftedLaplacian );
-    SLSolver -> setstiff(Kmat);
-    SLSolver -> setmass(Mmat);
-    SLSolver -> setProblemMatrix(Amat);
-    SLSolver -> setPreconditioningMatrix(Pmat);
-    SLSolver -> initialize();
-    SLSolver -> setupFastRAP();
+    RCP<ShiftedLaplacian> SLSolver = rcp(new ShiftedLaplacian);
+    SLSolver->setstiff(Kmat);
+    SLSolver->setmass(Mmat);
+    SLSolver->setProblemMatrix(Amat);
+    SLSolver->setPreconditioningMatrix(Pmat);
+    SLSolver->initialize();
+    SLSolver->setupFastRAP();
 
     tm = Teuchos::null;
 
@@ -192,22 +192,22 @@ int main(int argc, char *argv[]) {
     //   Solve linear system with Belos   //
     //************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
 
-    RCP<TMV> X = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    RCP<TMV> B = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    X->putScalar((SC) 0.0);
-    B->putScalar((SC) 0.0);
-    int pointsourceid=nx*ny/2+nx/2;
-    if(map->isNodeGlobalElement(pointsourceid)==true) {
-      B->replaceGlobalValue(pointsourceid, 0, (SC) 1.0);
+    RCP<TMV> X = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    RCP<TMV> B = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    X->putScalar((SC)0.0);
+    B->putScalar((SC)0.0);
+    int pointsourceid = nx * ny / 2 + nx / 2;
+    if (map->isNodeGlobalElement(pointsourceid) == true) {
+      B->replaceGlobalValue(pointsourceid, 0, (SC)1.0);
     }
 
     tm = Teuchos::null;
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
 
-    SLSolver -> solve(B,X);
+    SLSolver->solve(B, X);
 
     tm = Teuchos::null;
 
@@ -218,5 +218,5 @@ int main(int argc, char *argv[]) {
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
diff --git a/packages/muelu/test/helmholtz/HelmholtzFEM3D.cpp b/packages/muelu/test/helmholtz/HelmholtzFEM3D.cpp
index d8cd318be298..24a3aa4276bf 100644
--- a/packages/muelu/test/helmholtz/HelmholtzFEM3D.cpp
+++ b/packages/muelu/test/helmholtz/HelmholtzFEM3D.cpp
@@ -64,15 +64,15 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  typedef Tpetra::Vector<SC,LO,GO,NO>              TVEC;
-  typedef Tpetra::MultiVector<SC,LO,GO,NO>         TMV;
-  typedef Tpetra::CrsMatrix<SC,LO,GO,NO>           TCRS;
-  typedef Xpetra::CrsMatrix<SC,LO,GO,NO>           XCRS;
-  typedef Xpetra::TpetraCrsMatrix<SC,LO,GO,NO>     XTCRS;
-  typedef Xpetra::Matrix<SC,LO,GO,NO>              XMAT;
-  typedef Xpetra::CrsMatrixWrap<SC,LO,GO,NO>       XWRAP;
+  typedef Tpetra::Vector<SC, LO, GO, NO> TVEC;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> TMV;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> TCRS;
+  typedef Xpetra::CrsMatrix<SC, LO, GO, NO> XCRS;
+  typedef Xpetra::TpetraCrsMatrix<SC, LO, GO, NO> XTCRS;
+  typedef Xpetra::Matrix<SC, LO, GO, NO> XMAT;
+  typedef Xpetra::CrsMatrixWrap<SC, LO, GO, NO> XWRAP;
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -83,7 +83,7 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     Teuchos::CommandLineProcessor clp(false);
 
     //***********************//
@@ -99,46 +99,46 @@ int main(int argc, char *argv[]) {
 
     std::ifstream inputfile;
     inputfile.open("helm3D.inp");
-    inputfile >> nx       >> ny       >> nz ;
-    if(comm->getRank()==0)
-      std::cout<<"nx: "<<nx<<"  ny: "<<ny<<"  nz: "<<nz<<std::endl;
-    inputfile >> stretchx >> stretchy >> stretchz ;
-    if(comm->getRank()==0)
-      std::cout<<"stretchx: "<<stretchx<<"  stretchy: "<<stretchy<<"  stretchz: "<<stretchz<<std::endl;
-    inputfile >> h        >> delta ;
-    if(comm->getRank()==0)
-      std::cout<<"h: "<<h<<"  delta: "<<delta<<std::endl;
-    inputfile >> PMLXL    >> PMLXR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLXL: "<<PMLXL<<"  PMLXR: "<<PMLXR<<std::endl;
-    inputfile >> PMLYL    >> PMLYR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLYL: "<<PMLYL<<"  PMLYR: "<<PMLYR<<std::endl;
-    inputfile >> PMLZL    >> PMLZR ;
-    if(comm->getRank()==0)
-      std::cout<<"PMLZL: "<<PMLZL<<"  PMLZR: "<<PMLZR<<std::endl;
-    inputfile >> omega    >> shift ;
-    if(comm->getRank()==0)
-      std::cout<<"omega: "<<omega<<"  shift: "<<shift<<std::endl;
-    inputfile >> mx       >> my       >> mz ;
-    if(comm->getRank()==0)
-      std::cout<<"mx: "<<mx<<"  my: "<<my<<"  mz: "<<mz<<std::endl;
-    inputfile >> model ;
-    if(comm->getRank()==0)
-      std::cout<<"velocity model: "<<model<<std::endl;
+    inputfile >> nx >> ny >> nz;
+    if (comm->getRank() == 0)
+      std::cout << "nx: " << nx << "  ny: " << ny << "  nz: " << nz << std::endl;
+    inputfile >> stretchx >> stretchy >> stretchz;
+    if (comm->getRank() == 0)
+      std::cout << "stretchx: " << stretchx << "  stretchy: " << stretchy << "  stretchz: " << stretchz << std::endl;
+    inputfile >> h >> delta;
+    if (comm->getRank() == 0)
+      std::cout << "h: " << h << "  delta: " << delta << std::endl;
+    inputfile >> PMLXL >> PMLXR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLXL: " << PMLXL << "  PMLXR: " << PMLXR << std::endl;
+    inputfile >> PMLYL >> PMLYR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLYL: " << PMLYL << "  PMLYR: " << PMLYR << std::endl;
+    inputfile >> PMLZL >> PMLZR;
+    if (comm->getRank() == 0)
+      std::cout << "PMLZL: " << PMLZL << "  PMLZR: " << PMLZR << std::endl;
+    inputfile >> omega >> shift;
+    if (comm->getRank() == 0)
+      std::cout << "omega: " << omega << "  shift: " << shift << std::endl;
+    inputfile >> mx >> my >> mz;
+    if (comm->getRank() == 0)
+      std::cout << "mx: " << mx << "  my: " << my << "  mz: " << mz << std::endl;
+    inputfile >> model;
+    if (comm->getRank() == 0)
+      std::cout << "velocity model: " << model << std::endl;
 
     Galeri::Xpetra::Parameters<GO> matrixParameters_helmholtz(clp, nx, ny, nz, "HelmholtzFEM3D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
+                                                              h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, 0.0, mx, my, mz, model);
     Galeri::Xpetra::Parameters<GO> matrixParameters_shifted(clp, nx, ny, nz, "HelmholtzFEM3D", 0, stretchx, stretchy, stretchz,
-        h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
-    Xpetra::Parameters             xpetraParameters(clp);
+                                                            h, delta, PMLXL, PMLXR, PMLYL, PMLYR, PMLZL, PMLZR, omega, shift, mx, my, mz, model);
+    Xpetra::Parameters xpetraParameters(clp);
 
     //****************************************//
     //   Setup Galeri Problems and Matrices   //
     //****************************************//
 
-    RCP<TimeMonitor> globalTimeMonitor = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
-    RCP<TimeMonitor> tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
+    RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
+    RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
     Teuchos::ParameterList pl = matrixParameters_helmholtz.GetParameterList();
     RCP<RealValuedMultiVector> coordinates;
@@ -148,7 +148,7 @@ int main(int argc, char *argv[]) {
     galeriList.set("nz", pl.get("nz", nz));
     RCP<const Map> map;
 
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
     coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", map, matrixParameters_helmholtz.GetParameterList());
 
     RCP<const Tpetra::Map<LO, GO, NO> > tmap = Xpetra::toTpetra(map);
@@ -156,12 +156,12 @@ int main(int argc, char *argv[]) {
     Teuchos::ParameterList matrixParams_helmholtz = matrixParameters_helmholtz.GetParameterList();
     Teuchos::ParameterList matrixParams_shifted   = matrixParameters_shifted.GetParameterList();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_helmholtz =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_helmholtz =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_helmholtz.GetMatrixType(), map, matrixParams_helmholtz);
     RCP<Matrix> Amat = Pr_helmholtz->BuildMatrix();
 
-    RCP<Galeri::Xpetra::Problem_Helmholtz<Map,CrsMatrixWrap,MultiVector> > Pr_shifted =
-      Galeri::Xpetra::BuildProblem_Helmholtz<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
+    RCP<Galeri::Xpetra::Problem_Helmholtz<Map, CrsMatrixWrap, MultiVector> > Pr_shifted =
+        Galeri::Xpetra::BuildProblem_Helmholtz<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters_shifted.GetMatrixType(), map, matrixParams_shifted);
     RCP<Matrix> Pmat = Pr_shifted->BuildMatrix();
 
     comm->barrier();
@@ -172,16 +172,16 @@ int main(int argc, char *argv[]) {
     //   Setup Shifted Laplacian Preconditioner and Belos Solver   //
     //*************************************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 2 - MueLu Setup")));
 
     // initialize with the Shifted Laplacian operator (Pmat)
-    RCP<ShiftedLaplacian> SLSolver = rcp( new ShiftedLaplacian );
-    SLSolver -> setPreconditioningMatrix(Pmat);
-    SLSolver -> initialize();
+    RCP<ShiftedLaplacian> SLSolver = rcp(new ShiftedLaplacian);
+    SLSolver->setPreconditioningMatrix(Pmat);
+    SLSolver->initialize();
     // set Helmholtz operator (Amat)
-    SLSolver -> setProblemMatrix(Amat);
+    SLSolver->setProblemMatrix(Amat);
     // setup with SL operator
-    SLSolver -> setupNormalRAP();
+    SLSolver->setupNormalRAP();
 
     tm = Teuchos::null;
 
@@ -189,22 +189,22 @@ int main(int argc, char *argv[]) {
     //   Solve linear system with Belos   //
     //************************************//
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
 
-    RCP<TMV> X = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    RCP<TMV> B = Tpetra::createMultiVector<SC,LO,GO,NO>(tmap,1);
-    X->putScalar((SC) 0.0);
-    B->putScalar((SC) 0.0);
-    int pointsourceid=nx*ny*nz/2+nx*ny/2+nx/2;
-    if(map->isNodeGlobalElement(pointsourceid)==true) {
-      B->replaceGlobalValue(pointsourceid, 0, (SC) 1.0);
+    RCP<TMV> X = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    RCP<TMV> B = Tpetra::createMultiVector<SC, LO, GO, NO>(tmap, 1);
+    X->putScalar((SC)0.0);
+    B->putScalar((SC)0.0);
+    int pointsourceid = nx * ny * nz / 2 + nx * ny / 2 + nx / 2;
+    if (map->isNodeGlobalElement(pointsourceid) == true) {
+      B->replaceGlobalValue(pointsourceid, 0, (SC)1.0);
     }
 
     tm = Teuchos::null;
 
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Belos Solve")));
 
-    SLSolver -> solve(B,X);
+    SLSolver->solve(B, X);
 
     tm = Teuchos::null;
 
@@ -216,5 +216,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
diff --git a/packages/muelu/test/interface/CreateOperator.cpp b/packages/muelu/test/interface/CreateOperator.cpp
index 954cc5607f05..4902b6c793e6 100644
--- a/packages/muelu/test/interface/CreateOperator.cpp
+++ b/packages/muelu/test/interface/CreateOperator.cpp
@@ -80,243 +80,245 @@ const std::string prefSeparator = "=====================================";
 
 namespace MueLuExamples {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void setup_system_list(Xpetra::UnderlyingLib& lib, Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A, Teuchos::ParameterList& mueluList, const std::string& fname) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void setup_system_list(Xpetra::UnderlyingLib& lib, Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A, Teuchos::ParameterList& mueluList, const std::string& fname) {
 #include <MueLu_UseShortNames.hpp>
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    int myRank = A->getRowMap()->getComm()->getRank();
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  int myRank = A->getRowMap()->getComm()->getRank();
 
-    std::filebuf    buffer;
-    std::streambuf* oldbuffer = NULL;
+  std::filebuf buffer;
+  std::streambuf* oldbuffer = NULL;
 
-    typedef Tpetra::Operator<SC,LO,GO,NO> Tpetra_Operator;
-    typedef Tpetra::CrsMatrix<SC,LO,GO,NO> Tpetra_CrsMatrix;
-    typedef Tpetra::Vector<SC,LO,GO,NO> Tpetra_Vector;
-    typedef Tpetra::MultiVector<SC,LO,GO,NO> Tpetra_MultiVector;
-    if (lib == Xpetra::UseTpetra) {
-      if (myRank == 0) {
-        // Redirect output
-        buffer.open((fname + ".out").c_str(), std::ios::out);
-        oldbuffer = std::cout.rdbuf(&buffer);
-      }
+  typedef Tpetra::Operator<SC, LO, GO, NO> Tpetra_Operator;
+  typedef Tpetra::CrsMatrix<SC, LO, GO, NO> Tpetra_CrsMatrix;
+  typedef Tpetra::Vector<SC, LO, GO, NO> Tpetra_Vector;
+  typedef Tpetra::MultiVector<SC, LO, GO, NO> Tpetra_MultiVector;
+  if (lib == Xpetra::UseTpetra) {
+    if (myRank == 0) {
+      // Redirect output
+      buffer.open((fname + ".out").c_str(), std::ios::out);
+      oldbuffer = std::cout.rdbuf(&buffer);
+    }
 
-      RCP<Tpetra_CrsMatrix> At = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(A);
-      RCP<Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > opA(At);
-      RCP<Tpetra_Operator>  Mt = MueLu::CreateTpetraPreconditioner(opA, mueluList);
+    RCP<Tpetra_CrsMatrix> At = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(A);
+    RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > opA(At);
+    RCP<Tpetra_Operator> Mt = MueLu::CreateTpetraPreconditioner(opA, mueluList);
 
-      if (myRank == 0) {
-        // Redirect output back
-        std::cout.rdbuf(oldbuffer);
-        buffer.close();
-      }
+    if (myRank == 0) {
+      // Redirect output back
+      std::cout.rdbuf(oldbuffer);
+      buffer.close();
     }
+  }
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_SERIAL)
-    if (lib == Xpetra::UseEpetra) {
-      if (myRank == 0) {
-        // Redirect output
-        buffer.open((fname + ".out").c_str(), std::ios::out);
-        oldbuffer = std::cout.rdbuf(&buffer);
-      }
+  if (lib == Xpetra::UseEpetra) {
+    if (myRank == 0) {
+      // Redirect output
+      buffer.open((fname + ".out").c_str(), std::ios::out);
+      oldbuffer = std::cout.rdbuf(&buffer);
+    }
 
-      RCP<Epetra_CrsMatrix>   Ae = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(A);
-      RCP<Epetra_Operator>    Me = MueLu::CreateEpetraPreconditioner(Ae, mueluList);
+    RCP<Epetra_CrsMatrix> Ae = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstEpetraCrs(A);
+    RCP<Epetra_Operator> Me  = MueLu::CreateEpetraPreconditioner(Ae, mueluList);
 
-      if (myRank == 0) {
-        // Redirect output back
-        std::cout.rdbuf(oldbuffer);
-        buffer.close();
-      }
+    if (myRank == 0) {
+      // Redirect output back
+      std::cout.rdbuf(oldbuffer);
+      buffer.close();
     }
-#endif
   }
+#endif
+}
 
-  // This routine generate's the user's original A matrix and nullspace
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void generate_user_matrix_and_nullspace(std::string &matrixType,  Xpetra::UnderlyingLib & lib, Teuchos::ParameterList &galeriList,  Teuchos::RCP<const Teuchos::Comm<int> > &comm, Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A, Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & nullspace, Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >& coordinates){
+// This routine generate's the user's original A matrix and nullspace
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void generate_user_matrix_and_nullspace(std::string& matrixType, Xpetra::UnderlyingLib& lib, Teuchos::ParameterList& galeriList, Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& nullspace, Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& coordinates) {
 #include <MueLu_UseShortNames.hpp>
-    using Teuchos::RCP;
+  using Teuchos::RCP;
 
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+  RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  Teuchos::FancyOStream& out       = *fancy;
 
-    RCP<const Map>   map;
-    if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+  RCP<const Map> map;
+  if (matrixType == "Laplace1D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
-    } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+  } else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
-    } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
-    }
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(lib, "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
+  }
 
-    // Expand map to do multiple DOF per node for block problems
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
+  // Expand map to do multiple DOF per node for block problems
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, (matrixType == "Elasticity2D" ? 2 : 3));
 
-    out << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
-        << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
-        << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
-        << "========================================================" << std::endl;
+  out << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
+      << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
+      << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
+      << "========================================================" << std::endl;
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, galeriList);
 
-    A = Pr->BuildMatrix();
+  A = Pr->BuildMatrix();
 
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
-      nullspace = Pr->BuildNullspace();
-      A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
-    }
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
+    nullspace = Pr->BuildNullspace();
+    A->SetFixedBlockSize((matrixType == "Elasticity2D") ? 2 : 3);
   }
+}
 
-  void run_sed(const std::string& pattern, const std::string& baseFile) {
-    // sed behaviour differs between Mac and Linux
-    // You can run "sed -i 's//' " in Linux, but you always have to specify
-    // "sed -i "<smth,could be empty>" 's//'" in Mac. Both, however, take '-i<extension>'
-    std::string sed_pref = "sed -i ";
+void run_sed(const std::string& pattern, const std::string& baseFile) {
+  // sed behaviour differs between Mac and Linux
+  // You can run "sed -i 's//' " in Linux, but you always have to specify
+  // "sed -i "<smth,could be empty>" 's//'" in Mac. Both, however, take '-i<extension>'
+  std::string sed_pref = "sed -i ";
 #ifdef __APPLE__
-    sed_pref = sed_pref +  "\"\" ";
+  sed_pref = sed_pref + "\"\" ";
 #endif
 
-    int ret_val = 0; (void) ret_val; // suppress fscanf return value and unused variable warnings
-    ret_val = system((sed_pref + pattern + " " + baseFile + ".gold_filtered").c_str());
-    ret_val = system((sed_pref + pattern + " " + baseFile + ".out_filtered").c_str());
-  }
-
-  bool compare_to_gold(int myRank, const std::string & baseFile) {
-    bool failed=false;
-    if (myRank == 0) {
+  int ret_val = 0;
+  (void)ret_val;  // suppress fscanf return value and unused variable warnings
+  ret_val = system((sed_pref + pattern + " " + baseFile + ".gold_filtered").c_str());
+  ret_val = system((sed_pref + pattern + " " + baseFile + ".out_filtered").c_str());
+}
 
-      // Create a copy of outputs
-      std::string cmd = "cp -f ";
-      int ret_val = 0; (void) ret_val; // suppress fscanf return value and unused variable warnings
-      ret_val = system((cmd + baseFile + ".gold " + baseFile + ".gold_filtered").c_str());
-      ret_val = system((cmd + baseFile + ".out "  + baseFile + ".out_filtered").c_str());
-
-      // Tpetra produces different eigenvalues in Chebyshev due to using
-      // std::rand() for generating random vectors, which may be initialized
-      // using different seed, and may have different algorithm from one
-      // gcc version to another, or to anogther compiler (like clang)
-      // This leads to us always failing this test.
-      // NOTE1 : Epetra, on the other hand, rolls out its out random number
-      // generator, which always produces same results
-
-      // make sure complex tests pass
-      run_sed("'s/relaxation: damping factor = (1,0)/relaxation: damping factor = 1/'", baseFile);
-      run_sed("'s/damping factor: (1,0)/damping factor: 1/'", baseFile);
-      run_sed("'s/relaxation: min diagonal value = (0,0)/relaxation: min diagonal value = 0/'", baseFile);
-
-      // Ignore the value of "lambdaMax"
-      run_sed("'s/lambdaMax: [0-9]*.[0-9]*/lambdaMax = <ignored>/'", baseFile);
-
-      // Ignore the value of "lambdaMin"
-      run_sed("'s/lambdaMin: [0-9]*.[0-9]*/lambdaMin = <ignored>/'", baseFile);
-
-      // Ignore the value of "chebyshev: max eigenvalue"
-      // NOTE: we skip lines with default value ([default])
-      run_sed("'/[default]/! s/chebyshev: max eigenvalue = [0-9]*.[0-9]*/chebyshev: max eigenvalue = <ignored>/'", baseFile);
-
-      // Ignore the exact type of direct solver (it is selected semi-automatically
-      // depending on how Trilinos was configured
-      run_sed("'s/Amesos\\([2]*\\)Smoother{type = .*}/Amesos\\1Smoother{type = <ignored>}/'", baseFile);
-      run_sed("'s/SuperLU solver interface, direct solve/<Direct> solver interface/'", baseFile);
-      run_sed("'s/KLU2 solver interface/<Direct> solver interface/'", baseFile);
-      run_sed("'s/Basker solver interface/<Direct> solver interface/'", baseFile);
-
-      // The smoother complexity depends on the coarse solver.
-      run_sed("'s/Smoother complexity = [0-9]*.[0-9]*/Smoother complexity = <ignored>/'", baseFile);
-
-      // Nuke all pointers
-      run_sed("'s/0x[0-9a-f]*//g'", baseFile);
-
-      // Strip template args for some classes
-      std::vector<std::string> classes;
-      classes.push_back("Xpetra::Matrix");
-      classes.push_back("MueLu::Constraint");
-      classes.push_back("MueLu::SmootherPrototype");
-      for (size_t q = 0; q < classes.size(); q++)
-        run_sed("'s/" + classes[q] + "<.*>/" + classes[q] + "<ignored> >/'", baseFile);
+bool compare_to_gold(int myRank, const std::string& baseFile) {
+  bool failed = false;
+  if (myRank == 0) {
+    // Create a copy of outputs
+    std::string cmd = "cp -f ";
+    int ret_val     = 0;
+    (void)ret_val;  // suppress fscanf return value and unused variable warnings
+    ret_val = system((cmd + baseFile + ".gold " + baseFile + ".gold_filtered").c_str());
+    ret_val = system((cmd + baseFile + ".out " + baseFile + ".out_filtered").c_str());
+
+    // Tpetra produces different eigenvalues in Chebyshev due to using
+    // std::rand() for generating random vectors, which may be initialized
+    // using different seed, and may have different algorithm from one
+    // gcc version to another, or to anogther compiler (like clang)
+    // This leads to us always failing this test.
+    // NOTE1 : Epetra, on the other hand, rolls out its out random number
+    // generator, which always produces same results
+
+    // make sure complex tests pass
+    run_sed("'s/relaxation: damping factor = (1,0)/relaxation: damping factor = 1/'", baseFile);
+    run_sed("'s/damping factor: (1,0)/damping factor: 1/'", baseFile);
+    run_sed("'s/relaxation: min diagonal value = (0,0)/relaxation: min diagonal value = 0/'", baseFile);
+
+    // Ignore the value of "lambdaMax"
+    run_sed("'s/lambdaMax: [0-9]*.[0-9]*/lambdaMax = <ignored>/'", baseFile);
+
+    // Ignore the value of "lambdaMin"
+    run_sed("'s/lambdaMin: [0-9]*.[0-9]*/lambdaMin = <ignored>/'", baseFile);
+
+    // Ignore the value of "chebyshev: max eigenvalue"
+    // NOTE: we skip lines with default value ([default])
+    run_sed("'/[default]/! s/chebyshev: max eigenvalue = [0-9]*.[0-9]*/chebyshev: max eigenvalue = <ignored>/'", baseFile);
+
+    // Ignore the exact type of direct solver (it is selected semi-automatically
+    // depending on how Trilinos was configured
+    run_sed("'s/Amesos\\([2]*\\)Smoother{type = .*}/Amesos\\1Smoother{type = <ignored>}/'", baseFile);
+    run_sed("'s/SuperLU solver interface, direct solve/<Direct> solver interface/'", baseFile);
+    run_sed("'s/KLU2 solver interface/<Direct> solver interface/'", baseFile);
+    run_sed("'s/Basker solver interface/<Direct> solver interface/'", baseFile);
+
+    // The smoother complexity depends on the coarse solver.
+    run_sed("'s/Smoother complexity = [0-9]*.[0-9]*/Smoother complexity = <ignored>/'", baseFile);
+
+    // Nuke all pointers
+    run_sed("'s/0x[0-9a-f]*//g'", baseFile);
+
+    // Strip template args for some classes
+    std::vector<std::string> classes;
+    classes.push_back("Xpetra::Matrix");
+    classes.push_back("MueLu::Constraint");
+    classes.push_back("MueLu::SmootherPrototype");
+    for (size_t q = 0; q < classes.size(); q++)
+      run_sed("'s/" + classes[q] + "<.*>/" + classes[q] + "<ignored> >/'", baseFile);
 
 #ifdef __APPLE__
-      // Some Macs print outs ptrs as 0x0 instead of 0, fix that
-      run_sed("'/RCP/ s/=0x0/=0/g'", baseFile);
+    // Some Macs print outs ptrs as 0x0 instead of 0, fix that
+    run_sed("'/RCP/ s/=0x0/=0/g'", baseFile);
 #endif
 
-      // Run comparison (ignoring whitespaces)
-      cmd = "diff -u -w -I\"^\\s*$\" " + baseFile + ".gold_filtered " + baseFile + ".out_filtered";
-      int ret = system(cmd.c_str());
-      if (ret)
-        failed = true;
-
-      std::cout << baseFile << ": " << (ret ? "failed" : "passed") << std::endl;
-    }
+    // Run comparison (ignoring whitespaces)
+    cmd     = "diff -u -w -I\"^\\s*$\" " + baseFile + ".gold_filtered " + baseFile + ".out_filtered";
+    int ret = system(cmd.c_str());
+    if (ret)
+      failed = true;
 
-    return !failed;
+    std::cout << baseFile << ": " << (ret ? "failed" : "passed") << std::endl;
   }
 
-}//namespace
+  return !failed;
+}
+
+}  // namespace MueLuExamples
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ParameterList;
   using Teuchos::TimeMonitor;
 
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   bool success = true;
   bool verbose = true;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int numProc = comm->getSize();
-    int myRank  = comm->getRank();
-    RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    int numProc                         = comm->getSize();
+    int myRank                          = comm->getRank();
+    RCP<Teuchos::FancyOStream> fancy    = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    Teuchos::FancyOStream& out          = *fancy;
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
     ::Xpetra::Parameters xpetraParameters(clp);
 
     bool useKokkos = false;
-    if(lib == Xpetra::UseTpetra) {
+    if (lib == Xpetra::UseTpetra) {
       useKokkos = !Node::is_serial;
     }
     clp.setOption("useKokkosRefactor", "noKokkosRefactor", &useKokkos, "use kokkos refactor");
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     ParameterList galeriList = galeriParameters.GetParameterList();
-    out << thickSeparator << std::endl << xpetraParameters << galeriParameters;
+    out << thickSeparator << std::endl
+        << xpetraParameters << galeriParameters;
 
     // =========================================================================
     // Problem construction
     // =========================================================================
-    RCP<const Map>   map;
+    RCP<const Map> map;
     RCP<Matrix> A, P, R, Ac;
     RCP<MultiVector> nullspace;
     RCP<RealValuedMultiVector> coordinates0;
     RCP<RealValuedMultiVector> coordinates1;
     std::string matrixType = galeriParameters.GetMatrixType();
-    MueLuExamples::generate_user_matrix_and_nullspace<Scalar,LocalOrdinal,GlobalOrdinal,Node>(matrixType, lib, galeriList, comm, A, nullspace, coordinates0);
+    MueLuExamples::generate_user_matrix_and_nullspace<Scalar, LocalOrdinal, GlobalOrdinal, Node>(matrixType, lib, galeriList, comm, A, nullspace, coordinates0);
     map = A->getRowMap();
 
     std::string prefix;
@@ -338,12 +340,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // Solve #1 (standard MueLu)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 1: Standard "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 1: Standard " << prefSeparator << std::endl;
     {
-      std::string fname = prefix+"Output/operator_solve_1_np" + Teuchos::toString(numProc);
-      fname = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
+      std::string fname = prefix + "Output/operator_solve_1_np" + Teuchos::toString(numProc);
+      fname             = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
 
-      std::filebuf    buffer;
+      std::filebuf buffer;
       std::streambuf* oldbuffer = NULL;
       if (myRank == 0) {
         // Redirect output
@@ -354,13 +356,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       std::srand(12345);
 
       ParameterList mueluList;
-      mueluList.set("verbosity",                           "interfacetest");
-      mueluList.set("coarse: max size",                    100);
-      mueluList.set("use kokkos refactor",                 useKokkos);
-      mueluList.set("aggregation: deterministic",          useKokkos);
+      mueluList.set("verbosity", "interfacetest");
+      mueluList.set("coarse: max size", 100);
+      mueluList.set("use kokkos refactor", useKokkos);
+      mueluList.set("aggregation: deterministic", useKokkos);
 
       ParameterListInterpreter mueLuFactory(mueluList);
-      RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
+      RCP<Hierarchy> H                              = mueLuFactory.CreateHierarchy();
       Teuchos::RCP<FactoryManagerBase> LevelFactory = mueLuFactory.GetFactoryManager(1);
       H->setlib(lib);
       H->AddNewLevel();
@@ -374,123 +376,124 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       H->GetLevel(1)->Get("R", R);
       H->GetLevel(1)->Get("P", P);
       H->GetLevel(1)->Get("A", Ac);
-      nullspace = H->GetLevel(1)->template Get<RCP<MultiVector> >("Nullspace", LevelFactory->GetFactory("Nullspace").get());
+      nullspace    = H->GetLevel(1)->template Get<RCP<MultiVector> >("Nullspace", LevelFactory->GetFactory("Nullspace").get());
       coordinates1 = H->GetLevel(1)->template Get<RCP<RealValuedMultiVector> >("Coordinates", LevelFactory->GetFactory("Coordinates").get());
 
-      if (myRank==0) {
+      if (myRank == 0) {
         // Redirect output back
         std::cout.rdbuf(oldbuffer);
         buffer.close();
       }
 
       bool passed = MueLuExamples::compare_to_gold(myRank, fname);
-      success = success && passed;
+      success     = success && passed;
     }
 
     // =========================================================================
     // Solve #5 (level wrap, the fast way, P, R + Nullspace)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 5: LevelWrap, Fast Way, P, R "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 5: LevelWrap, Fast Way, P, R " << prefSeparator << std::endl;
     {
-      std::string fname = prefix+"Output/operator_solve_5_np" + Teuchos::toString(numProc);
-      fname = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
+      std::string fname = prefix + "Output/operator_solve_5_np" + Teuchos::toString(numProc);
+      fname             = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
 
       std::srand(12345);
 
       ParameterList mueluList;
-      mueluList.set("verbosity",          "interfacetest");
-      mueluList.set("coarse: max size",   100);
+      mueluList.set("verbosity", "interfacetest");
+      mueluList.set("coarse: max size", 100);
       mueluList.set("use kokkos refactor", useKokkos);
       mueluList.set("aggregation: deterministic", useKokkos);
       ParameterList& level0 = mueluList.sublist("level 0");
       level0.set("Coordinates", coordinates0);
       ParameterList& level1 = mueluList.sublist("level 1");
-      level1.set("R",           R);
-      level1.set("P",           P);
-      level1.set("Nullspace",   nullspace);
+      level1.set("R", R);
+      level1.set("P", P);
+      level1.set("Nullspace", nullspace);
       level1.set("Coordinates", coordinates1);
 
-      MueLuExamples::setup_system_list<Scalar,LocalOrdinal,GlobalOrdinal,Node>(lib, A, mueluList, fname);
+      MueLuExamples::setup_system_list<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib, A, mueluList, fname);
 
       bool passed = MueLuExamples::compare_to_gold(myRank, fname);
-      success = success && passed;
+      success     = success && passed;
     }
 
     // =========================================================================
     // Solve #6 (level wrap, the fast way, P only, explicit transpose)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 6: LevelWrap, Fast Way, P only, explicit transpose "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 6: LevelWrap, Fast Way, P only, explicit transpose " << prefSeparator << std::endl;
     {
-      std::string fname = prefix+"Output/operator_solve_6_np" + Teuchos::toString(numProc);
-      fname = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
+      std::string fname = prefix + "Output/operator_solve_6_np" + Teuchos::toString(numProc);
+      fname             = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
 
       std::srand(12345);
 
       ParameterList mueluList;
-      mueluList.set("verbosity",                  "interfacetest");
-      mueluList.set("transpose: use implicit",    false);
-      mueluList.set("max levels",                 4);
-      mueluList.set("coarse: max size",           100);
+      mueluList.set("verbosity", "interfacetest");
+      mueluList.set("transpose: use implicit", false);
+      mueluList.set("max levels", 4);
+      mueluList.set("coarse: max size", 100);
       mueluList.set("use kokkos refactor", useKokkos);
       mueluList.set("aggregation: deterministic", useKokkos);
       ParameterList& level0 = mueluList.sublist("level 0");
       level0.set("Coordinates", coordinates0);
       ParameterList& level1 = mueluList.sublist("level 1");
-      level1.set("P",         P);
+      level1.set("P", P);
       level1.set("Nullspace", nullspace);
       level1.set("Coordinates", coordinates1);
 
-      MueLuExamples::setup_system_list<Scalar,LocalOrdinal,GlobalOrdinal,Node>(lib, A, mueluList, fname);
+      MueLuExamples::setup_system_list<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib, A, mueluList, fname);
 
       bool passed = MueLuExamples::compare_to_gold(myRank, fname);
-      success = success && passed;
+      success     = success && passed;
     }
 
     // =========================================================================
     // Solve #7 (level wrap, the fast way, P only, implicit transpose)
     // =========================================================================
     out << thickSeparator << std::endl;
-    out << prefSeparator << " Solve 7: LevelWrap, Fast Way, P only, implicit transpose "<< prefSeparator <<std::endl;
+    out << prefSeparator << " Solve 7: LevelWrap, Fast Way, P only, implicit transpose " << prefSeparator << std::endl;
     {
-      std::string fname = prefix+"Output/operator_solve_7_np" + Teuchos::toString(numProc);
-      fname = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
+      std::string fname = prefix + "Output/operator_solve_7_np" + Teuchos::toString(numProc);
+      fname             = fname + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
 
       std::srand(12345);
 
       ParameterList mueluList;
-      mueluList.set("verbosity",                  "interfacetest");
-      mueluList.set("coarse: max size",           100);
-      mueluList.set("transpose: use implicit",    true);
-      mueluList.set("max levels",                 2);
+      mueluList.set("verbosity", "interfacetest");
+      mueluList.set("coarse: max size", 100);
+      mueluList.set("transpose: use implicit", true);
+      mueluList.set("max levels", 2);
       mueluList.set("use kokkos refactor", useKokkos);
       mueluList.set("aggregation: deterministic", useKokkos);
       ParameterList& level0 = mueluList.sublist("level 0");
       level0.set("Coordinates", coordinates0);
       ParameterList& level1 = mueluList.sublist("level 1");
-      level1.set("P",         P);
+      level1.set("P", P);
       level1.set("Nullspace", nullspace);
       level1.set("Coordinates", coordinates1);
 
-      MueLuExamples::setup_system_list<Scalar,LocalOrdinal,GlobalOrdinal,Node>(lib, A, mueluList, fname);
+      MueLuExamples::setup_system_list<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib, A, mueluList, fname);
 
       bool passed = MueLuExamples::compare_to_gold(myRank, fname);
-      success = success && passed;
+      success     = success && passed;
     }
 
     if (myRank == 0)
-      std::cout << std::endl << "End Result: TEST " << (!success ? "FAILED" : "PASSED") << std::endl;
+      std::cout << std::endl
+                << "End Result: TEST " << (!success ? "FAILED" : "PASSED") << std::endl;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/interface/ParameterListInterpreter.cpp b/packages/muelu/test/interface/ParameterListInterpreter.cpp
index 12f39f6367ef..472b97ff8a5a 100644
--- a/packages/muelu/test/interface/ParameterListInterpreter.cpp
+++ b/packages/muelu/test/interface/ParameterListInterpreter.cpp
@@ -56,7 +56,7 @@
 #include <MueLu.hpp>
 
 #if defined(HAVE_MUELU_AMESOS2)
-#include <Amesos2_config.h> // needed for check whether KLU2 is available
+#include <Amesos2_config.h>  // needed for check whether KLU2 is available
 #endif
 
 #include <MueLu_Exceptions.hpp>
@@ -71,56 +71,56 @@
 
 void run_sed(const std::string& pattern, const std::string& baseFile);
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
 
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int numProc = comm->getSize();
-  int myRank  = comm->getRank();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  int numProc                         = comm->getSize();
+  int myRank                          = comm->getRank();
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
 
-  bool runHeavyTests = false;
+  bool runHeavyTests       = false;
   std::string xmlForceFile = "";
-  bool useKokkos = false;
-  if(lib == Xpetra::UseTpetra) {
-      useKokkos = !Node::is_serial;
+  bool useKokkos           = false;
+  if (lib == Xpetra::UseTpetra) {
+    useKokkos = !Node::is_serial;
   }
   bool compareWithGold = !Node::is_gpu;
   clp.setOption("useKokkosRefactor", "noKokkosRefactor", &useKokkos, "use kokkos refactor");
-  clp.setOption("heavytests", "noheavytests",  &runHeavyTests, "whether to exercise tests that take a long time to run");
+  clp.setOption("heavytests", "noheavytests", &runHeavyTests, "whether to exercise tests that take a long time to run");
   clp.setOption("xml", &xmlForceFile, "xml input file (useful for debugging)");
   clp.setOption("compareWithGold", "skipCompareWithGold", &compareWithGold, "compare runs against gold files");
   clp.recogniseAllOptions(true);
-  switch (clp.parse(argc,argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   // =========================================================================
   // Problem construction
   // =========================================================================
   Teuchos::ParameterList matrixParameters;
-  matrixParameters.set("nx",         Teuchos::as<GO>(9999));
+  matrixParameters.set("nx", Teuchos::as<GO>(9999));
   matrixParameters.set("matrixType", "Laplace1D");
-  RCP<Matrix>      A           = MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(matrixParameters.get<GO>("nx"), lib);
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), matrixParameters);
-  RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<Matrix> A                          = MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(matrixParameters.get<GO>("nx"), lib);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), matrixParameters);
+  RCP<MultiVector> nullspace             = MultiVectorFactory::Build(A->getRowMap(), 1);
   nullspace->putScalar(1.0);
 
   std::string prefix;
@@ -129,20 +129,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   } else {
     prefix = "default/";
   }
-  std::string outDir = prefix+"Output/";
+  std::string outDir = prefix + "Output/";
 
   std::vector<std::string> dirList;
   if (runHeavyTests) {
-    dirList.push_back(prefix+"EasyParameterListInterpreter-heavy/");
+    dirList.push_back(prefix + "EasyParameterListInterpreter-heavy/");
     if (!useKokkos) {
       // commented since extended xml interface does not support kokkos factories
-      dirList.push_back(prefix+"FactoryParameterListInterpreter-heavy/");
+      dirList.push_back(prefix + "FactoryParameterListInterpreter-heavy/");
     }
   } else {
-    dirList.push_back(prefix+"EasyParameterListInterpreter/");
+    dirList.push_back(prefix + "EasyParameterListInterpreter/");
     if (!useKokkos) {
       // commented since extended xml interface does not support kokkos factories
-      dirList.push_back(prefix+"FactoryParameterListInterpreter/");
+      dirList.push_back(prefix + "FactoryParameterListInterpreter/");
     }
   }
 #if defined(HAVE_MPI) && defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_AMESOS2_KLU2)
@@ -150,20 +150,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // output would depend on configuration (reguarl interpreter does not have
   // that). Therefore, we need to stabilize the configuration here.
   // In addition, we run ML parameter list tests only if KLU is available
-  dirList.push_back(prefix+"MLParameterListInterpreter/");
-  dirList.push_back(prefix+"MLParameterListInterpreter2/");
+  dirList.push_back(prefix + "MLParameterListInterpreter/");
+  dirList.push_back(prefix + "MLParameterListInterpreter2/");
 #endif
   int numLists = dirList.size();
 
-  bool failed = false;
+  bool failed  = false;
   bool jumpOut = false;
   Teuchos::Time timer("Interpreter timer");
-  //double lastTime = timer.wallTime();
+  // double lastTime = timer.wallTime();
   for (int k = 0; k < numLists; k++) {
     Teuchos::ArrayRCP<std::string> fileList = MueLuTests::TestHelpers::GetFileList(dirList[k],
-      (numProc == 1 ? std::string(".xml") : std::string("_np" + Teuchos::toString(numProc) + ".xml")));
+                                                                                   (numProc == 1 ? std::string(".xml") : std::string("_np" + Teuchos::toString(numProc) + ".xml")));
 
-    std::sort(fileList.begin(),fileList.end());
+    std::sort(fileList.begin(), fileList.end());
 
     for (int i = 0; i < fileList.size(); i++) {
       // Set seed
@@ -176,20 +176,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       std::string outFile;
       std::string baseFile;
       std::size_t found;
-      if (xmlForceFile==""){
+      if (xmlForceFile == "") {
         xmlFile  = dirList[k] + fileList[i];
-        outFile  = outDir     + fileList[i];
+        outFile  = outDir + fileList[i];
         baseFile = outFile.substr(0, outFile.find_last_of('.'));
-        found = baseFile.find("_np");
+        found    = baseFile.find("_np");
       } else {
-        xmlFile = xmlForceFile;
-        std::string dir = prefix + xmlForceFile.substr(0, xmlForceFile.find_last_of('/')+1);
-        dirList[k] = dir;
-        outFile = outDir + xmlForceFile.substr(xmlForceFile.find_last_of('/')+1, xmlForceFile.size());
-        baseFile = outFile.substr(0, outFile.find_last_of('.'));
-        found = baseFile.find("_np");
-        jumpOut = true;
-        xmlFile = prefix + xmlFile;
+        xmlFile         = xmlForceFile;
+        std::string dir = prefix + xmlForceFile.substr(0, xmlForceFile.find_last_of('/') + 1);
+        dirList[k]      = dir;
+        outFile         = outDir + xmlForceFile.substr(xmlForceFile.find_last_of('/') + 1, xmlForceFile.size());
+        baseFile        = outFile.substr(0, outFile.find_last_of('.'));
+        found           = baseFile.find("_np");
+        jumpOut         = true;
+        xmlFile         = prefix + xmlFile;
         std::cout << "Test dir: " << dirList[k] << std::endl;
       }
 
@@ -202,9 +202,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #endif
       }
       if (myRank == 0)
-        std::cout << "Testing: "<< xmlFile << std::endl;
+        std::cout << "Testing: " << xmlFile << std::endl;
 
-      baseFile = baseFile + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
+      baseFile             = baseFile + (lib == Xpetra::UseEpetra ? "_epetra" : "_tpetra");
       std::string goldFile = baseFile + ".gold";
       std::ifstream f(goldFile.c_str());
       if (!f.good()) {
@@ -222,12 +222,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       // first to include "test" verbosity
       Teuchos::ParameterList paramList;
       Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFile, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
-      if      (dirList[k] == prefix+"EasyParameterListInterpreter/" || dirList[k] == prefix+"EasyParameterListInterpreter-heavy/")
+      if (dirList[k] == prefix + "EasyParameterListInterpreter/" || dirList[k] == prefix + "EasyParameterListInterpreter-heavy/")
         paramList.set("verbosity", "interfacetest");
-      else if (dirList[k] == prefix+"FactoryParameterListInterpreter/" || dirList[k] == prefix+"FactoryParameterListInterpreter-heavy/")
+      else if (dirList[k] == prefix + "FactoryParameterListInterpreter/" || dirList[k] == prefix + "FactoryParameterListInterpreter-heavy/")
         paramList.sublist("Hierarchy").set("verbosity", "InterfaceTest");
-      else if (dirList[k] == prefix+"MLParameterListInterpreter/" || dirList[k] == prefix+"MLParameterListInterpreter2/")
-        paramList.set("ML output",     666);
+      else if (dirList[k] == prefix + "MLParameterListInterpreter/" || dirList[k] == prefix + "MLParameterListInterpreter2/")
+        paramList.set("ML output", 666);
 
       try {
         timer.start();
@@ -238,24 +238,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         // and the ML parameter list interpreter. Note that the ML paramter interpreter also
         // works with Tpetra matrices.
         bool coordsSet = false;
-        if (dirList[k] == prefix+"EasyParameterListInterpreter/"         ||
-            dirList[k] == prefix+"EasyParameterListInterpreter-heavy/") {
+        if (dirList[k] == prefix + "EasyParameterListInterpreter/" ||
+            dirList[k] == prefix + "EasyParameterListInterpreter-heavy/") {
           paramList.set("use kokkos refactor", useKokkos);
-          mueluFactory = Teuchos::rcp(new ParameterListInterpreter(paramList,comm));
-        } else if (dirList[k] == prefix+"FactoryParameterListInterpreter/"      ||
-                   dirList[k] == prefix+"FactoryParameterListInterpreter-heavy/") {
+          mueluFactory = Teuchos::rcp(new ParameterListInterpreter(paramList, comm));
+        } else if (dirList[k] == prefix + "FactoryParameterListInterpreter/" ||
+                   dirList[k] == prefix + "FactoryParameterListInterpreter-heavy/") {
           paramList.sublist("Hierarchy").set("use kokkos refactor", useKokkos);
-          mueluFactory = Teuchos::rcp(new ParameterListInterpreter(paramList,comm));
+          mueluFactory = Teuchos::rcp(new ParameterListInterpreter(paramList, comm));
 
-        } else if (dirList[k] == prefix+"MLParameterListInterpreter/") {
-          if(paramList.isParameter("parameter list: syntax"))
+        } else if (dirList[k] == prefix + "MLParameterListInterpreter/") {
+          if (paramList.isParameter("parameter list: syntax"))
             paramList.remove("parameter list: syntax");
-          RCP<Teuchos::ParameterList> mueluParamList = Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(paramList,"SA"));
-          mueluParamList->set("multigrid algorithm","sa");
+          RCP<Teuchos::ParameterList> mueluParamList = Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(paramList, "SA"));
+          mueluParamList->set("multigrid algorithm", "sa");
           mueluParamList->set("use kokkos refactor", useKokkos);
-          
+
           // If we are using Kokkos refactor, we need to strip off the options that Kokkos doesn't support
-          if(useKokkos) {
+          if (useKokkos) {
             if (mueluParamList->isParameter("aggregation: match ML phase1"))
               mueluParamList->remove("aggregation: match ML phase1");
             if (mueluParamList->isParameter("aggregation: match ML phase2b"))
@@ -266,10 +266,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
           mueluFactory = Teuchos::rcp(new ParameterListInterpreter(*mueluParamList));
 
-        } else if (dirList[k] == prefix+"MLParameterListInterpreter2/") {
-          RCP<Teuchos::ParameterList> mueluParamList = Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(paramList,"SA"));
+        } else if (dirList[k] == prefix + "MLParameterListInterpreter2/") {
+          RCP<Teuchos::ParameterList> mueluParamList = Teuchos::getParametersFromXmlString(MueLu::ML2MueLuParameterTranslator::translate(paramList, "SA"));
 
-          mueluParamList->set("multigrid algorithm","sa");
+          mueluParamList->set("multigrid algorithm", "sa");
           mueluParamList->set("use kokkos refactor", useKokkos);
           mueluFactory = Teuchos::rcp(new ParameterListInterpreter(*mueluParamList));
         } else
@@ -295,7 +295,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       } catch (Teuchos::ExceptionBase& e) {
         std::string msg = e.what();
-        msg = msg.substr(msg.find_last_of('\n')+1);
+        msg             = msg.substr(msg.find_last_of('\n') + 1);
 
         if (myRank == 0)
           std::cout << "Caught exception: " << msg << std::endl;
@@ -303,7 +303,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         if (msg == "Zoltan interface is not available" ||
             msg == "Zoltan2 interface is not available" ||
             msg == "MueLu::FactoryFactory:BuildFactory(): Cannot create a Zoltan2Interface object: Zoltan2 is disabled: HAVE_MUELU_ZOLTAN2 && HAVE_MPI == false.") {
-
           if (myRank == 0)
             std::cout << xmlFile << ": skipped (missing library)" << std::endl;
 
@@ -323,7 +322,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       MPI_File_open((*mpiComm)(), (baseFile + ".out").c_str(), MPI_MODE_WRONLY | MPI_MODE_CREATE, MPI_INFO_NULL, &logfile);
       MPI_File_set_atomicity(logfile, true);
       const char* msg = logStr.c_str();
-      int err = MPI_File_write_ordered(logfile, msg, logStr.size(), MPI_CHAR, MPI_STATUS_IGNORE);
+      int err         = MPI_File_write_ordered(logfile, msg, logStr.size(), MPI_CHAR, MPI_STATUS_IGNORE);
       TEUCHOS_ASSERT(err == MPI_SUCCESS);
       MPI_File_close(&logfile);
 #else
@@ -336,12 +335,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       std::string cmd;
       if (myRank == 0) {
         // Create a copy of outputs
-        cmd = "cp -f ";
+        cmd     = "cp -f ";
         int ret = 0;
-        ret = system((cmd + baseFile + ".gold " + baseFile + ".gold_filtered").c_str());
-        TEUCHOS_ASSERT_EQUALITY(ret,0);
+        ret     = system((cmd + baseFile + ".gold " + baseFile + ".gold_filtered").c_str());
+        TEUCHOS_ASSERT_EQUALITY(ret, 0);
         ret = system((cmd + baseFile + ".out " + baseFile + ".out_filtered").c_str());
-        TEUCHOS_ASSERT_EQUALITY(ret,0);
+        TEUCHOS_ASSERT_EQUALITY(ret, 0);
 
         // Tpetra produces different eigenvalues in Chebyshev due to using
         // std::rand() for generating random vectors, which may be initialized
@@ -394,7 +393,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
         // Run comparison (ignoring whitespaces)
         cmd = "diff -u -w -I\"^\\s*$\" " + baseFile + ".gold_filtered " + baseFile + ".out_filtered";
-        ret = 0; // GH: to keep the old behavior the same, zero it out here
+        ret = 0;  // GH: to keep the old behavior the same, zero it out here
         if (compareWithGold)
           ret = system(cmd.c_str());
         else
@@ -403,12 +402,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           failed = true;
         }
 
-        //std::ios_base::fmtflags ff(std::cout.flags());
-        //std::cout.precision(2);
-        //std::cout << xmlFile << " (" << std::setiosflags(std::ios::fixed)
-        //          << timer.wallTime() - lastTime << " sec.) : " << (ret ? "failed" : "passed") << std::endl;
-        //lastTime = timer.wallTime();
-        //std::cout.flags(ff); // reset flags to whatever they were prior to printing time
+        // std::ios_base::fmtflags ff(std::cout.flags());
+        // std::cout.precision(2);
+        // std::cout << xmlFile << " (" << std::setiosflags(std::ios::fixed)
+        //           << timer.wallTime() - lastTime << " sec.) : " << (ret ? "failed" : "passed") << std::endl;
+        // lastTime = timer.wallTime();
+        // std::cout.flags(ff); // reset flags to whatever they were prior to printing time
         std::cout << xmlFile << " : " << (ret ? "failed" : "passed") << std::endl;
       }
       if (jumpOut)
@@ -419,36 +418,35 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   }
 
   if (myRank == 0)
-    std::cout << std::endl << "End Result: TEST " << (failed ? "FAILED" : "PASSED") << std::endl;
+    std::cout << std::endl
+              << "End Result: TEST " << (failed ? "FAILED" : "PASSED") << std::endl;
 
   return (failed ? EXIT_FAILURE : EXIT_SUCCESS);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
   Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
 #ifdef KOKKOS_ENABLE_OPENMP
   omp_set_num_threads(1);
 #endif
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
 
-
 void run_sed(const std::string& pattern, const std::string& baseFile) {
   // sed behaviour differs between Mac and Linux
   // You can run "sed -i 's//' " in Linux, but you always have to specify
   // "sed -i "<smth,could be empty>" 's//'" in Mac. Both, however, take '-i<extension>'
   std::string sed_pref = "sed -i ";
 #ifdef __APPLE__
-  sed_pref = sed_pref +  "\"\" ";
+  sed_pref = sed_pref + "\"\" ";
 #endif
   int ret;
   ret = system((sed_pref + pattern + " " + baseFile + ".gold_filtered").c_str());
-  TEUCHOS_ASSERT_EQUALITY(ret,0);
+  TEUCHOS_ASSERT_EQUALITY(ret, 0);
   ret = system((sed_pref + pattern + " " + baseFile + ".out_filtered").c_str());
-  TEUCHOS_ASSERT_EQUALITY(ret,0);
+  TEUCHOS_ASSERT_EQUALITY(ret, 0);
 }
diff --git a/packages/muelu/test/longlong/TwoBillion.cpp b/packages/muelu/test/longlong/TwoBillion.cpp
index f4003c342513..a4b6ebfadd1a 100644
--- a/packages/muelu/test/longlong/TwoBillion.cpp
+++ b/packages/muelu/test/longlong/TwoBillion.cpp
@@ -81,13 +81,13 @@
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosXpetraAdapter.hpp" // this header defines Belos::XpetraOp()
-#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
+#include "BelosXpetraAdapter.hpp"  // this header defines Belos::XpetraOp()
+#include "BelosMueLuAdapter.hpp"   // this header defines Belos::MueLuOp()
 #endif
 
-typedef double                               Scalar;
-typedef int                                  LocalOrdinal;
-typedef long long                            GlobalOrdinal;
+typedef double Scalar;
+typedef int LocalOrdinal;
+typedef long long GlobalOrdinal;
 typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
 using Xpetra::global_size_t;
@@ -95,7 +95,8 @@ using Xpetra::global_size_t;
 int main(int argc, char *argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   Teuchos::oblackholestream blackhole;
@@ -104,54 +105,53 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
 
     Teuchos::CommandLineProcessor clp(false);
-    std::string xmlFileName = "TwoBillion.xml"; clp.setOption("xml", &xmlFileName,  "read parameters from a file");
-    int num_per_proc = 1000; clp.setOption("dpc", &num_per_proc,  "DOFs per core");
+    std::string xmlFileName = "TwoBillion.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file");
+    int num_per_proc = 1000;
+    clp.setOption("dpc", &num_per_proc, "DOFs per core");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     int NumProcs = comm->getSize();
     int MyPID    = comm->getRank();
 
-    if(!MyPID) printf("TwoBillion: Running Test\n");
-
+    if (!MyPID) printf("TwoBillion: Running Test\n");
 
-    const long long FIRST_GID  = 3000000000L;
+    const long long FIRST_GID = 3000000000L;
     //  const long long FIRST_GID  = 0L;
 
-    const long long IndexBase  = 0L;
+    const long long IndexBase = 0L;
     //  const long long IndexBase  = 3000000000L;
 
-    global_size_t NumGlobalElements = NumProcs*num_per_proc;
+    global_size_t NumGlobalElements = NumProcs * num_per_proc;
 
     // Create Map w/ GIDs starting at > 2 billion
     RCP<const Map> map;
     RCP<CrsMatrix> Acrs;
     Teuchos::Array<GlobalOrdinal> mygids(num_per_proc);
 
-    for(int i=0; i<num_per_proc; i++)
-      mygids[i] = FIRST_GID + MyPID*num_per_proc + i;
+    for (int i = 0; i < num_per_proc; i++)
+      mygids[i] = FIRST_GID + MyPID * num_per_proc + i;
 
-    for (int i=0; i<NumProcs; ++i) {
-      if (i==MyPID)
-        std::cout << "pid " <<  i << " : 1st GID = " << mygids[0] << std::endl;
+    for (int i = 0; i < NumProcs; ++i) {
+      if (i == MyPID)
+        std::cout << "pid " << i << " : 1st GID = " << mygids[0] << std::endl;
     }
 
-    //for(int i=0;i<num_per_proc;i++)
-    //  printf("[%d] mygids[%d] = %lld\n",MyPID,i,mygids[i]);
+    // for(int i=0;i<num_per_proc;i++)
+    //   printf("[%d] mygids[%d] = %lld\n",MyPID,i,mygids[i]);
 
-
-    map = MapFactory::Build(Xpetra::UseTpetra, Teuchos::OrdinalTraits<global_size_t>::invalid(),mygids(),IndexBase,comm);
+    map = MapFactory::Build(Xpetra::UseTpetra, Teuchos::OrdinalTraits<global_size_t>::invalid(), mygids(), IndexBase, comm);
 
     //  RCP<Teuchos::FancyOStream> fox = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     //  fox->setOutputToRootOnly(-1);
@@ -163,32 +163,36 @@ int main(int argc, char *argv[]) {
     Teuchos::ArrayView<Scalar> ValView;
     Teuchos::ArrayView<GlobalOrdinal> ColView;
 
-
-    Acrs = CrsMatrixFactory::Build(map,3);
-    for(int i=0; i<num_per_proc; i++) {
-      if(mygids[i]==FIRST_GID ) {
-        mycols[0] = mygids[i];     myvals[0] = 2;
-        mycols[1] = mygids[i]+1;   myvals[1] = -1;
-        ValView=myvals.view(0,2);
-        ColView=mycols.view(0,2);
+    Acrs = CrsMatrixFactory::Build(map, 3);
+    for (int i = 0; i < num_per_proc; i++) {
+      if (mygids[i] == FIRST_GID) {
+        mycols[0] = mygids[i];
+        myvals[0] = 2;
+        mycols[1] = mygids[i] + 1;
+        myvals[1] = -1;
+        ValView   = myvals.view(0, 2);
+        ColView   = mycols.view(0, 2);
         //      printf("[%d %lld] cols %lld %lld\n",MyPID,mygids[i],mycols[0],mycols[1]);
-      }
-      else if(mygids[i] == FIRST_GID + (long long) NumGlobalElements - 1){
-        mycols[0] = mygids[i]-1;   myvals[0] = -1;
-        mycols[1] = mygids[i];     myvals[1] = 2;
-        ValView=myvals.view(0,2);
-        ColView=mycols.view(0,2);
+      } else if (mygids[i] == FIRST_GID + (long long)NumGlobalElements - 1) {
+        mycols[0] = mygids[i] - 1;
+        myvals[0] = -1;
+        mycols[1] = mygids[i];
+        myvals[1] = 2;
+        ValView   = myvals.view(0, 2);
+        ColView   = mycols.view(0, 2);
         //      printf("[%d %lld] cols %lld %lld\n",MyPID,mygids[i],mycols[0],mycols[1]);
-      }
-      else {
-        mycols[0] = mygids[i]-1;   myvals[0] = -1;
-        mycols[1] = mygids[i];     myvals[1] = -2;
-        mycols[2] = mygids[i]+1;   myvals[1] = -1;
-        ValView=myvals();
-        ColView=mycols();
+      } else {
+        mycols[0] = mygids[i] - 1;
+        myvals[0] = -1;
+        mycols[1] = mygids[i];
+        myvals[1] = -2;
+        mycols[2] = mygids[i] + 1;
+        myvals[1] = -1;
+        ValView   = myvals();
+        ColView   = mycols();
         //      printf("[%d %lld] cols %lld %lld %lld\n",MyPID,mygids[i],mycols[0],mycols[1],mycols[2]);
       }
-      Acrs->insertGlobalValues(mygids[i],ColView,ValView);
+      Acrs->insertGlobalValues(mygids[i], ColView, ValView);
     }
     Acrs->fillComplete();
 
@@ -196,11 +200,11 @@ int main(int argc, char *argv[]) {
 
     RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
     nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
-    RCP<MultiVector> coordinates = MultiVectorFactory::Build(map, 1);
+    RCP<MultiVector> coordinates        = MultiVectorFactory::Build(map, 1);
     Teuchos::ArrayRCP<Scalar> coordVals = coordinates->getDataNonConst(0);
-    double h = 1.0 / NumGlobalElements;
-    for (LocalOrdinal i=0; i<num_per_proc; ++i)
-      coordVals[i] = MyPID*num_per_proc*h + i*h;
+    double h                            = 1.0 / NumGlobalElements;
+    for (LocalOrdinal i = 0; i < num_per_proc; ++i)
+      coordVals[i] = MyPID * num_per_proc * h + i * h;
     coordVals = Teuchos::null;
 
     Teuchos::ParameterList paramList;
@@ -211,8 +215,8 @@ int main(int argc, char *argv[]) {
 
     RCP<Hierarchy> H;
     H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A",           A);
-    H->GetLevel(0)->Set("Nullspace",   nullspace);
+    H->GetLevel(0)->Set("A", A);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
     H->GetLevel(0)->Set("Coordinates", coordinates);
     mueLuFactory->SetupHierarchy(*H);
 
@@ -231,45 +235,46 @@ int main(int argc, char *argv[]) {
     X->randomize();
     A->apply(*X, *B, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
     B->norm2(norms);
-    B->scale(1.0/norms[0]);
+    B->scale(1.0 / norms[0]);
 
     //
     // Use AMG as a preconditioner in Belos
     //
 #ifdef HAVE_MUELU_BELOS
     // Operator and Multivector type that will be used with Belos
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
     H->IsPreconditioner(true);
 
     // Define Operator and Preconditioner
-    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Operator object into a Belos operator
-    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Operator object into a Belos operator
+    Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
-    RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+    RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
     belosProblem->setLeftPrec(belosPrec);
 
     bool set = belosProblem->setProblem();
     if (set == false) {
       if (comm->getRank() == 0)
-        std::cout << std::endl << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
+        std::cout << std::endl
+                  << "ERROR:  Belos::LinearProblem failed to set up correctly!" << std::endl;
       return EXIT_FAILURE;
     }
 
     // Belos parameter list
-    int maxIts = 100;
+    int maxIts    = 100;
     double optTol = 1e-8;
     Teuchos::ParameterList belosList;
-    belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-    belosList.set("Convergence Tolerance", optTol);    // Relative convergence tolerance requested
-    //belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails);
+    belosList.set("Maximum Iterations", maxIts);     // Maximum number of iterations allowed
+    belosList.set("Convergence Tolerance", optTol);  // Relative convergence tolerance requested
+    // belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails);
     belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
     belosList.set("Output Frequency", 1);
     belosList.set("Output Style", Belos::Brief);
 
     // Create an iterative solver manager
-    RCP< Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+    RCP<Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
 
     // Perform solve
     Belos::ReturnType ret = Belos::Unconverged;
@@ -281,47 +286,51 @@ int main(int argc, char *argv[]) {
         std::cout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
       if (solver->getNumIters() > 6) {
-        if (comm->getRank() == 0) std::cout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
-        return(EXIT_FAILURE);
+        if (comm->getRank() == 0) std::cout << std::endl
+                                            << "ERROR:  Belos did not converge! " << std::endl;
+        return (EXIT_FAILURE);
       }
 
       // Compute actual residuals.
       int numrhs = 1;
-      std::vector<double> actual_resids( numrhs ); //TODO: double?
-      std::vector<double> rhs_norm( numrhs );
+      std::vector<double> actual_resids(numrhs);  // TODO: double?
+      std::vector<double> rhs_norm(numrhs);
       RCP<MultiVector> resid = MultiVectorFactory::Build(map, numrhs);
 
-      typedef Belos::OperatorTraits<SC, MV, OP>  OPT;
-      typedef Belos::MultiVecTraits<SC, MV>     MVT;
-
-      OPT::Apply( *belosOp, *X, *resid );
-      MVT::MvAddMv( -1.0, *resid, 1.0, *B, *resid );
-      MVT::MvNorm( *resid, actual_resids );
-      MVT::MvNorm( *B, rhs_norm );
-      *out<< "---------- Actual Residuals (normalized) ----------"<<std::endl<<std::endl;
-      for ( int i = 0; i<numrhs; i++) {
-        double actRes = actual_resids[i]/rhs_norm[i];
-        *out<<"Problem "<<i<<" : \t"<< actRes <<std::endl;
-        //if (actRes > tol) { badRes = true; }
+      typedef Belos::OperatorTraits<SC, MV, OP> OPT;
+      typedef Belos::MultiVecTraits<SC, MV> MVT;
+
+      OPT::Apply(*belosOp, *X, *resid);
+      MVT::MvAddMv(-1.0, *resid, 1.0, *B, *resid);
+      MVT::MvNorm(*resid, actual_resids);
+      MVT::MvNorm(*B, rhs_norm);
+      *out << "---------- Actual Residuals (normalized) ----------" << std::endl
+           << std::endl;
+      for (int i = 0; i < numrhs; i++) {
+        double actRes = actual_resids[i] / rhs_norm[i];
+        *out << "Problem " << i << " : \t" << actRes << std::endl;
+        // if (actRes > tol) { badRes = true; }
       }
 
-    } //try
+    }  // try
 
-    catch(...) {
+    catch (...) {
       if (comm->getRank() == 0)
-        std::cout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
-
+        std::cout << std::endl
+                  << "ERROR:  Belos threw an error! " << std::endl;
     }
 
     success = (ret == Belos::Converged);
     // Check convergence
     if (success) {
-      if (comm->getRank() == 0) std::cout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "SUCCESS:  Belos converged!" << std::endl;
     } else {
-      if (comm->getRank() == 0) std::cout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      if (comm->getRank() == 0) std::cout << std::endl
+                                          << "ERROR:  Belos did not converge! " << std::endl;
     }
-#endif // HAVE_MUELU_BELOS
+#endif  // HAVE_MUELU_BELOS
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/test/maxwell/Maxwell3D.cpp b/packages/muelu/test/maxwell/Maxwell3D.cpp
index 62bb6acce7a8..699d592ed0d6 100644
--- a/packages/muelu/test/maxwell/Maxwell3D.cpp
+++ b/packages/muelu/test/maxwell/Maxwell3D.cpp
@@ -81,7 +81,7 @@ using Teuchos::TimeMonitor;
 #include <BelosSolverFactory.hpp>
 #include <BelosTpetraAdapter.hpp>
 #endif
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
 
 // Stratimikos
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
@@ -103,97 +103,97 @@ using Teuchos::TimeMonitor;
 #endif
 
 // Helper functions for compilation purposes
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-struct EpetraSolvers_Wrapper{
-  static void Generate_ML_MaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& SM,
-                                                Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& D0,
-                                                Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& Kn,
-                                                Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& nullspace,
-                                                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> >& coords,
-                                                Teuchos::ParameterList & mueluList,
-                                                Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& mlopX) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct EpetraSolvers_Wrapper {
+  static void Generate_ML_MaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& SM,
+                                                Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& D0,
+                                                Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Kn,
+                                                Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& nullspace,
+                                                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coords,
+                                                Teuchos::ParameterList& mueluList,
+                                                Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& mlopX) {
     throw std::runtime_error("Template parameter mismatch");
   }
 
-  static void Generate_ML_RefMaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& SM,
-                                                   Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& D0,
-                                                   Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& Ms,
-                                                   Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& M0inv,
-                                                   Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& M1,
-                                                   Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& nullspace,
-                                                   Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& node_material,
-                                                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> >& coords,
-                                                   Teuchos::ParameterList & mueluList,
-                                                   Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & mlopX) {
+  static void Generate_ML_RefMaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& SM,
+                                                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& D0,
+                                                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Ms,
+                                                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& M0inv,
+                                                   Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& M1,
+                                                   Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& nullspace,
+                                                   Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& node_material,
+                                                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node> >& coords,
+                                                   Teuchos::ParameterList& mueluList,
+                                                   Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& mlopX) {
     throw std::runtime_error("Template parameter mismatch");
   }
 };
 
 #if defined(HAVE_MUELU_EPETRA)
-template<class GlobalOrdinal>
-struct EpetraSolvers_Wrapper<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
-  static void Generate_ML_MaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& SM,
-                                                Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& D0,
-                                                Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& Kn,
-                                                Teuchos::RCP<Xpetra::MultiVector<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& nullspace,
-                                                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<double>::coordinateType,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& coords,
-                                                Teuchos::ParameterList & mueluList,
-                                                Teuchos::RCP<Xpetra::Operator<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& mlopX) {
+template <class GlobalOrdinal>
+struct EpetraSolvers_Wrapper<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
+  static void Generate_ML_MaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& SM,
+                                                Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& D0,
+                                                Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& Kn,
+                                                Teuchos::RCP<Xpetra::MultiVector<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& nullspace,
+                                                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<double>::coordinateType, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& coords,
+                                                Teuchos::ParameterList& mueluList,
+                                                Teuchos::RCP<Xpetra::Operator<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& mlopX) {
 #if defined(HAVE_MUELU_ML)
     typedef double SC;
     typedef int LO;
     typedef GlobalOrdinal GO;
     typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode NO;
     typedef typename Teuchos::ScalarTraits<SC>::coordinateType coordinate_type;
-    typedef typename Xpetra::Matrix<SC,LO,GO,NO> Matrix;
+    typedef typename Xpetra::Matrix<SC, LO, GO, NO> Matrix;
 
     RCP<const Epetra_CrsMatrix> epetraSM = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(SM);
     RCP<const Epetra_CrsMatrix> epetraD0 = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(D0);
-    if(!coords.is_null()) {
-      RCP<const Epetra_MultiVector> epetraCoord =  MueLu::Utilities<coordinate_type,LO,GO,NO>::MV2EpetraMV(coords);
-      if(epetraCoord->NumVectors() > 0)  mueluList.set("x-coordinates",(*epetraCoord)[0]);
-      if(epetraCoord->NumVectors() > 1)  mueluList.set("y-coordinates",(*epetraCoord)[1]);
-      if(epetraCoord->NumVectors() > 2)  mueluList.set("z-coordinates",(*epetraCoord)[2]);
+    if (!coords.is_null()) {
+      RCP<const Epetra_MultiVector> epetraCoord = MueLu::Utilities<coordinate_type, LO, GO, NO>::MV2EpetraMV(coords);
+      if (epetraCoord->NumVectors() > 0) mueluList.set("x-coordinates", (*epetraCoord)[0]);
+      if (epetraCoord->NumVectors() > 1) mueluList.set("y-coordinates", (*epetraCoord)[1]);
+      if (epetraCoord->NumVectors() > 2) mueluList.set("z-coordinates", (*epetraCoord)[2]);
     }
-    if(!nullspace.is_null()) {
-      RCP<const Epetra_MultiVector> epetraNullspace =  MueLu::Utilities<SC,LO,GO,NO>::MV2EpetraMV(nullspace);
-      mueluList.set("null space: dimension",epetraNullspace->NumVectors());
-      mueluList.set("null space: vectors",(*epetraNullspace)[0]);
-      mueluList.set("null space: type","pre-computed");
+    if (!nullspace.is_null()) {
+      RCP<const Epetra_MultiVector> epetraNullspace = MueLu::Utilities<SC, LO, GO, NO>::MV2EpetraMV(nullspace);
+      mueluList.set("null space: dimension", epetraNullspace->NumVectors());
+      mueluList.set("null space: vectors", (*epetraNullspace)[0]);
+      mueluList.set("null space: type", "pre-computed");
     }
     RCP<const Epetra_CrsMatrix> epetraKn;
     if (Kn.is_null()) {
-      RCP<Matrix> temp = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(SM->getRangeMap());
-      Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*SM,false,*D0,false,*temp,true,true);
-      RCP<Matrix> Kn2 = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(D0->getDomainMap());
-      Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*D0,true,*temp,false,*Kn2,true,true);
+      RCP<Matrix> temp = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(SM->getRangeMap());
+      Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*SM, false, *D0, false, *temp, true, true);
+      RCP<Matrix> Kn2 = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(D0->getDomainMap());
+      Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*D0, true, *temp, false, *Kn2, true, true);
       epetraKn = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(Kn2);
     } else
       epetraKn = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(Kn);
 
-    RCP<Epetra_Operator> mlop  = rcp<Epetra_Operator>(new ML_Epetra::MultiLevelPreconditioner(*epetraSM,*epetraD0,*epetraKn,mueluList,true));
+    RCP<Epetra_Operator> mlop = rcp<Epetra_Operator>(new ML_Epetra::MultiLevelPreconditioner(*epetraSM, *epetraD0, *epetraKn, mueluList, true));
 #if defined(HAVE_MUELU_BELOS)
     // NOTE: Belos needs the Apply() and AppleInverse() routines of ML swapped.  So...
     mlop = rcp<Belos::EpetraPrecOp>(new Belos::EpetraPrecOp(mlop));
 #endif
 
-    mlopX = rcp(new Xpetra::EpetraOperator<GO,NO>(mlop));
+    mlopX = rcp(new Xpetra::EpetraOperator<GO, NO>(mlop));
 #else
     TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error,
                                "Need ML & Epetra support");
 #endif
   }
 
-  static void Generate_ML_RefMaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& SM,
-                                                   Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& D0,
-                                                   Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& Ms,
-                                                   Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& M0inv,
-                                                   Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& M1,
-                                                   Teuchos::RCP<Xpetra::MultiVector<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& nullspace,
-                                                   Teuchos::RCP<Xpetra::MultiVector<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& node_material,
-                                                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<double>::coordinateType,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& coords,
-                                                   Teuchos::ParameterList & mueluList,
-                                                   Teuchos::RCP<Xpetra::Operator<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& mlopX) {
+  static void Generate_ML_RefMaxwellPreconditioner(Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& SM,
+                                                   Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& D0,
+                                                   Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& Ms,
+                                                   Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& M0inv,
+                                                   Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& M1,
+                                                   Teuchos::RCP<Xpetra::MultiVector<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& nullspace,
+                                                   Teuchos::RCP<Xpetra::MultiVector<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& node_material,
+                                                   Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<double>::coordinateType, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& coords,
+                                                   Teuchos::ParameterList& mueluList,
+                                                   Teuchos::RCP<Xpetra::Operator<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& mlopX) {
 #if defined(HAVE_MUELU_ML)
     typedef double SC;
     typedef int LO;
@@ -205,80 +205,80 @@ struct EpetraSolvers_Wrapper<double,int,GlobalOrdinal,Tpetra::KokkosCompat::Kokk
     RCP<const Epetra_CrsMatrix> epetraD0    = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(D0);
     RCP<const Epetra_CrsMatrix> epetraM0inv = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(M0inv);
     RCP<const Epetra_CrsMatrix> epetraMs;
-    RCP<const Epetra_CrsMatrix> epetraM1    = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(M1);
+    RCP<const Epetra_CrsMatrix> epetraM1 = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(M1);
     if (!Ms.is_null())
-      epetraMs    = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(Ms);
+      epetraMs = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(Ms);
     else
       epetraMs = epetraM1;
-    mueluList.set("D0",epetraD0);
-    mueluList.set("Ms",epetraMs);
-    mueluList.set("M0inv",epetraM0inv);
-    mueluList.set("M1",epetraM1);
-    if(!coords.is_null()) {
-      RCP<const Epetra_MultiVector> epetraCoord =  MueLu::Utilities<coordinate_type,LO,GO,NO>::MV2EpetraMV(coords);
-      if(epetraCoord->NumVectors() > 0)  mueluList.sublist("refmaxwell: 11list").set("x-coordinates",(*epetraCoord)[0]);
-      if(epetraCoord->NumVectors() > 1)  mueluList.sublist("refmaxwell: 11list").set("y-coordinates",(*epetraCoord)[1]);
-      if(epetraCoord->NumVectors() > 2)  mueluList.sublist("refmaxwell: 11list").set("z-coordinates",(*epetraCoord)[2]);
+    mueluList.set("D0", epetraD0);
+    mueluList.set("Ms", epetraMs);
+    mueluList.set("M0inv", epetraM0inv);
+    mueluList.set("M1", epetraM1);
+    if (!coords.is_null()) {
+      RCP<const Epetra_MultiVector> epetraCoord = MueLu::Utilities<coordinate_type, LO, GO, NO>::MV2EpetraMV(coords);
+      if (epetraCoord->NumVectors() > 0) mueluList.sublist("refmaxwell: 11list").set("x-coordinates", (*epetraCoord)[0]);
+      if (epetraCoord->NumVectors() > 1) mueluList.sublist("refmaxwell: 11list").set("y-coordinates", (*epetraCoord)[1]);
+      if (epetraCoord->NumVectors() > 2) mueluList.sublist("refmaxwell: 11list").set("z-coordinates", (*epetraCoord)[2]);
     }
-    if(!node_material.is_null()) {
-      RCP<const Epetra_MultiVector> epetraMaterial =  MueLu::Utilities<coordinate_type,LO,GO,NO>::MV2EpetraMV(node_material);
-      mueluList.sublist("refmaxwell: 11list").set("material coordinates",(*epetraMaterial)[0]);
+    if (!node_material.is_null()) {
+      RCP<const Epetra_MultiVector> epetraMaterial = MueLu::Utilities<coordinate_type, LO, GO, NO>::MV2EpetraMV(node_material);
+      mueluList.sublist("refmaxwell: 11list").set("material coordinates", (*epetraMaterial)[0]);
     }
-    if(!nullspace.is_null()) {
-      RCP<const Epetra_MultiVector> epetraNullspace =  MueLu::Utilities<SC,LO,GO,NO>::MV2EpetraMV(nullspace);
-      mueluList.sublist("refmaxwell: 11list").set("null space: dimension",epetraNullspace->NumVectors());
-      mueluList.sublist("refmaxwell: 11list").set("null space: vectors",(*epetraNullspace)[0]);
-      mueluList.sublist("refmaxwell: 11list").set("null space: type","pre-computed");
+    if (!nullspace.is_null()) {
+      RCP<const Epetra_MultiVector> epetraNullspace = MueLu::Utilities<SC, LO, GO, NO>::MV2EpetraMV(nullspace);
+      mueluList.sublist("refmaxwell: 11list").set("null space: dimension", epetraNullspace->NumVectors());
+      mueluList.sublist("refmaxwell: 11list").set("null space: vectors", (*epetraNullspace)[0]);
+      mueluList.sublist("refmaxwell: 11list").set("null space: type", "pre-computed");
     }
 
-    RCP<Epetra_Operator> mlop  = rcp<Epetra_Operator>(new ML_Epetra::RefMaxwellPreconditioner(*epetraSM,mueluList,true));
+    RCP<Epetra_Operator> mlop = rcp<Epetra_Operator>(new ML_Epetra::RefMaxwellPreconditioner(*epetraSM, mueluList, true));
 #if defined(HAVE_MUELU_BELOS)
     // NOTE: Belos needs the Apply() and AppleInverse() routines of ML swapped.  So...
     mlop = rcp<Belos::EpetraPrecOp>(new Belos::EpetraPrecOp(mlop));
 #endif
 
-    mlopX = rcp(new Xpetra::EpetraOperator<GO,NO>(mlop));
+    mlopX = rcp(new Xpetra::EpetraOperator<GO, NO>(mlop));
 #else
     TEUCHOS_TEST_FOR_EXCEPTION(true, std::runtime_error,
                                "Need ML & Epetra support");
 #endif
   }
 };
-#endif // HAVE_MUELU_EPETRA
+#endif  // HAVE_MUELU_EPETRA
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 bool SetupSolve(std::map<std::string, void*> inputs) {
 #include <MueLu_UseShortNames.hpp>
 
   typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> coordMV;
 
-  RCP<Matrix>            SM_Matrix       = *static_cast<RCP<Matrix>*>(inputs["SM"]);
-  RCP<Matrix>            GmhdA_Matrix     = *static_cast<RCP<Matrix>*>(inputs["GmhdA"]);
-  RCP<Matrix>            D0_Matrix       = *static_cast<RCP<Matrix>*>(inputs["D0"]);
-  RCP<Matrix>            M1_Matrix       = *static_cast<RCP<Matrix>*>(inputs["M1"]);
-  RCP<Matrix>            Ms_Matrix       = *static_cast<RCP<Matrix>*>(inputs["Ms"]);
-  RCP<Matrix>            M0inv_Matrix    = *static_cast<RCP<Matrix>*>(inputs["M0inv"]);
-  RCP<Matrix>            Kn_Matrix       = *static_cast<RCP<Matrix>*>(inputs["Kn"]);
+  RCP<Matrix> SM_Matrix    = *static_cast<RCP<Matrix>*>(inputs["SM"]);
+  RCP<Matrix> GmhdA_Matrix = *static_cast<RCP<Matrix>*>(inputs["GmhdA"]);
+  RCP<Matrix> D0_Matrix    = *static_cast<RCP<Matrix>*>(inputs["D0"]);
+  RCP<Matrix> M1_Matrix    = *static_cast<RCP<Matrix>*>(inputs["M1"]);
+  RCP<Matrix> Ms_Matrix    = *static_cast<RCP<Matrix>*>(inputs["Ms"]);
+  RCP<Matrix> M0inv_Matrix = *static_cast<RCP<Matrix>*>(inputs["M0inv"]);
+  RCP<Matrix> Kn_Matrix    = *static_cast<RCP<Matrix>*>(inputs["Kn"]);
 
-  RCP<coordMV>           coords          = *static_cast<RCP<coordMV>*>(inputs["coordinates"]);
-  RCP<MultiVector>       nullspace       = *static_cast<RCP<MultiVector>*>(inputs["nullspace"]);
-  RCP<MultiVector>       material        = *static_cast<RCP<MultiVector>*>(inputs["material"]);
+  RCP<coordMV> coords        = *static_cast<RCP<coordMV>*>(inputs["coordinates"]);
+  RCP<MultiVector> nullspace = *static_cast<RCP<MultiVector>*>(inputs["nullspace"]);
+  RCP<MultiVector> material  = *static_cast<RCP<MultiVector>*>(inputs["material"]);
 
-  RCP<MultiVector>       B               = *static_cast<RCP<MultiVector>*>(inputs["B"]);
-  RCP<MultiVector>       X               = *static_cast<RCP<MultiVector>*>(inputs["X"]);
-  RCP<MultiVector>       X0              = *static_cast<RCP<MultiVector>*>(inputs["X0"]);
+  RCP<MultiVector> B  = *static_cast<RCP<MultiVector>*>(inputs["B"]);
+  RCP<MultiVector> X  = *static_cast<RCP<MultiVector>*>(inputs["X"]);
+  RCP<MultiVector> X0 = *static_cast<RCP<MultiVector>*>(inputs["X0"]);
 
-  Teuchos::ParameterList params          = *static_cast<Teuchos::ParameterList*>(inputs["params"]);
-  Teuchos::ParameterList belosParams     = *static_cast<Teuchos::ParameterList*>(inputs["belosParams"]);
+  Teuchos::ParameterList params            = *static_cast<Teuchos::ParameterList*>(inputs["params"]);
+  Teuchos::ParameterList belosParams       = *static_cast<Teuchos::ParameterList*>(inputs["belosParams"]);
   Teuchos::ParameterList stratimikosParams = *static_cast<Teuchos::ParameterList*>(inputs["stratimikosParams"]);
-  std::string            solverName      = *static_cast<std::string*>(inputs["solverName"]);
-  std::string            belosSolverType = *static_cast<std::string*>(inputs["belosSolverType"]);
-  std::string            precType        = *static_cast<std::string*>(inputs["precType"]);
-  int                    numResolves     = *static_cast<int*>(inputs["numResolves"]);
-  bool                   reuse           = *static_cast<bool*>(inputs["reuse"]);
+  std::string solverName                   = *static_cast<std::string*>(inputs["solverName"]);
+  std::string belosSolverType              = *static_cast<std::string*>(inputs["belosSolverType"]);
+  std::string precType                     = *static_cast<std::string*>(inputs["precType"]);
+  int numResolves                          = *static_cast<int*>(inputs["numResolves"]);
+  bool reuse                               = *static_cast<bool*>(inputs["reuse"]);
 
-  RCP<const Teuchos::Comm<int> > comm    = *static_cast<RCP<const Teuchos::Comm<int> >*>(inputs["comm"]);
-  RCP<Teuchos::FancyOStream> out         = *static_cast<RCP<Teuchos::FancyOStream>*>(inputs["out"]);
+  RCP<const Teuchos::Comm<int> > comm = *static_cast<RCP<const Teuchos::Comm<int> >*>(inputs["comm"]);
+  RCP<Teuchos::FancyOStream> out      = *static_cast<RCP<Teuchos::FancyOStream>*>(inputs["out"]);
 
   bool success = false;
 
@@ -287,54 +287,53 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
   if (solverName == "Belos") {
     // construct preconditioner
     RCP<Operator> preconditioner;
-    if (precType=="MueLu-RefMaxwell") {
-      preconditioner = rcp( new MueLu::RefMaxwell<SC,LO,GO,NO>(SM_Matrix,D0_Matrix,Ms_Matrix,M0inv_Matrix,
-                                                               M1_Matrix,nullspace,coords,params) );
-    }
-    else if (precType=="MueLu-Maxwell1" || precType=="MueLu-Reitzinger") {
+    if (precType == "MueLu-RefMaxwell") {
+      preconditioner = rcp(new MueLu::RefMaxwell<SC, LO, GO, NO>(SM_Matrix, D0_Matrix, Ms_Matrix, M0inv_Matrix,
+                                                                 M1_Matrix, nullspace, coords, params));
+    } else if (precType == "MueLu-Maxwell1" || precType == "MueLu-Reitzinger") {
       if (GmhdA_Matrix.is_null())  // are we doing MHD as opposed to GMHD?
-        preconditioner = rcp( new MueLu::Maxwell1<SC,LO,GO,NO>(SM_Matrix,D0_Matrix,Kn_Matrix,nullspace,coords,params) );
-      else 
-        preconditioner = rcp( new MueLu::Maxwell1<SC,LO,GO,NO>(SM_Matrix,D0_Matrix,Kn_Matrix,nullspace,coords,params,GmhdA_Matrix) );
+        preconditioner = rcp(new MueLu::Maxwell1<SC, LO, GO, NO>(SM_Matrix, D0_Matrix, Kn_Matrix, nullspace, coords, params));
+      else
+        preconditioner = rcp(new MueLu::Maxwell1<SC, LO, GO, NO>(SM_Matrix, D0_Matrix, Kn_Matrix, nullspace, coords, params, GmhdA_Matrix));
 
     }
-
 #ifdef HAVE_MUELU_EPETRA
-    else if (precType=="ML-RefMaxwell") {
-      Xpetra::UnderlyingLib  lib = *static_cast<Xpetra::UnderlyingLib*>(inputs["lib"]);
-      TEUCHOS_ASSERT(lib==Xpetra::UseEpetra);
-      EpetraSolvers_Wrapper<SC, LO, GO, NO>::Generate_ML_RefMaxwellPreconditioner(SM_Matrix,D0_Matrix,Ms_Matrix,M0inv_Matrix,
-                                                                                  M1_Matrix,nullspace,material,coords,params,preconditioner);
-    } else if (precType=="ML-Maxwell") {
-      Xpetra::UnderlyingLib  lib = *static_cast<Xpetra::UnderlyingLib*>(inputs["lib"]);
-      TEUCHOS_ASSERT(lib==Xpetra::UseEpetra);
-      EpetraSolvers_Wrapper<SC, LO, GO, NO>::Generate_ML_MaxwellPreconditioner(SM_Matrix,D0_Matrix,Kn_Matrix,
-                                                                               nullspace,coords,params,preconditioner);
+    else if (precType == "ML-RefMaxwell") {
+      Xpetra::UnderlyingLib lib = *static_cast<Xpetra::UnderlyingLib*>(inputs["lib"]);
+      TEUCHOS_ASSERT(lib == Xpetra::UseEpetra);
+      EpetraSolvers_Wrapper<SC, LO, GO, NO>::Generate_ML_RefMaxwellPreconditioner(SM_Matrix, D0_Matrix, Ms_Matrix, M0inv_Matrix,
+                                                                                  M1_Matrix, nullspace, material, coords, params, preconditioner);
+    } else if (precType == "ML-Maxwell") {
+      Xpetra::UnderlyingLib lib = *static_cast<Xpetra::UnderlyingLib*>(inputs["lib"]);
+      TEUCHOS_ASSERT(lib == Xpetra::UseEpetra);
+      EpetraSolvers_Wrapper<SC, LO, GO, NO>::Generate_ML_MaxwellPreconditioner(SM_Matrix, D0_Matrix, Kn_Matrix,
+                                                                               nullspace, coords, params, preconditioner);
     }
 #endif
 
     {
       // A test to make sure we can wrap this guy as a MueLu::TpetraOperator
       RCP<Operator> precOp = Teuchos::rcp_dynamic_cast<Operator>(preconditioner);
-      MueLu::TpetraOperator<SC,LO,GO,NO> OpT(precOp);
+      MueLu::TpetraOperator<SC, LO, GO, NO> OpT(precOp);
     }
 
     // Belos linear problem
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
     RCP<OP> belosOp;
-    if (GmhdA_Matrix.is_null())  // are we doing MHD as opposed to GMHD?
-           belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(SM_Matrix)); // Turns a Xpetra::Matrix object into a Belos operator
-    else   belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(GmhdA_Matrix)); // Turns a Xpetra::Matrix object into a Belos operator
+    if (GmhdA_Matrix.is_null())                                                // are we doing MHD as opposed to GMHD?
+      belosOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(SM_Matrix));  // Turns a Xpetra::Matrix object into a Belos operator
+    else
+      belosOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(GmhdA_Matrix));  // Turns a Xpetra::Matrix object into a Belos operator
 
-    RCP<Belos::LinearProblem<SC, MV, OP> > problem = rcp( new Belos::LinearProblem<SC, MV, OP>() );
-    problem -> setOperator( belosOp );
+    RCP<Belos::LinearProblem<SC, MV, OP> > problem = rcp(new Belos::LinearProblem<SC, MV, OP>());
+    problem->setOperator(belosOp);
     RCP<OP> belosPrecOp;
     if (precType != "none") {
-      belosPrecOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(preconditioner)); // Turns a Xpetra::Matrix object into a Belos operator
-      problem -> setRightPrec( belosPrecOp );
+      belosPrecOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(preconditioner));  // Turns a Xpetra::Matrix object into a Belos operator
+      problem->setRightPrec(belosPrecOp);
     }
-    problem -> setProblem( X, B );
+    problem->setProblem(X, B);
 
     bool set = problem->setProblem();
     if (set == false) {
@@ -343,9 +342,9 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
     }
 
     // Belos solver
-    RCP< Belos::SolverManager<SC, MV, OP> > solver;
-    RCP< Belos::SolverFactory<SC, MV,OP> > factory = rcp( new  Belos::SolverFactory<SC,MV,OP>() );
-    solver = factory->create(belosSolverType,Teuchos::rcpFromRef(belosParams.sublist(belosSolverType)));
+    RCP<Belos::SolverManager<SC, MV, OP> > solver;
+    RCP<Belos::SolverFactory<SC, MV, OP> > factory = rcp(new Belos::SolverFactory<SC, MV, OP>());
+    solver                                         = factory->create(belosSolverType, Teuchos::rcpFromRef(belosParams.sublist(belosSolverType)));
 
     comm->barrier();
     tm2 = Teuchos::null;
@@ -353,15 +352,15 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
     auto tm3 = TimeMonitor::getNewTimer("Maxwell: 3 - Solve");
 
     // set problem and solve
-    solver -> setProblem( problem );
-    for(int solveno = 0; solveno<=numResolves; solveno++) {
+    solver->setProblem(problem);
+    for (int solveno = 0; solveno <= numResolves; solveno++) {
       if (X0.is_null())
         X->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
       else
         X = X0;
-      Belos::ReturnType status = solver -> solve();
-      int iters = solver -> getNumIters();
-      success = (iters<50 && status == Belos::Converged);
+      Belos::ReturnType status = solver->solve();
+      int iters                = solver->getNumIters();
+      success                  = (iters < 50 && status == Belos::Converged);
       if (success)
         *out << "SUCCESS! Belos converged in " << iters << " iterations." << std::endl;
       else
@@ -370,20 +369,21 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
 
     if (reuse) {
       TEUCHOS_ASSERT(precType == "MueLu-RefMaxwell");
-      for (int solveno = 0; solveno<2; solveno++) {
+      for (int solveno = 0; solveno < 2; solveno++) {
         // SM_Matrix->resumeFill();
         // SM_Matrix->fillComplete();
         if (X0.is_null())
           X->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
         else
           X = X0;
-        problem -> setProblem( X, B );
+        problem->setProblem(X, B);
         if (GmhdA_Matrix.is_null())  // are we doing MHD as opposed to GMHD?
-          Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(preconditioner)->resetMatrix(SM_Matrix);
-        else Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC,LO,GO,NO> >(preconditioner)->resetMatrix(GmhdA_Matrix);
-        Belos::ReturnType status = solver -> solve();
-        int iters = solver -> getNumIters();
-        success = (iters<50 && status == Belos::Converged);
+          Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO> >(preconditioner)->resetMatrix(SM_Matrix);
+        else
+          Teuchos::rcp_dynamic_cast<MueLu::RefMaxwell<SC, LO, GO, NO> >(preconditioner)->resetMatrix(GmhdA_Matrix);
+        Belos::ReturnType status = solver->solve();
+        int iters                = solver->getNumIters();
+        success                  = (iters < 50 && status == Belos::Converged);
         if (success)
           *out << "SUCCESS! Belos converged in " << iters << " iterations." << std::endl;
         else
@@ -391,10 +391,9 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
       }
     }
   }
-#endif // HAVE_MUELU_BELOS
+#endif  // HAVE_MUELU_BELOS
 #if defined(HAVE_MUELU_STRATIMIKOS) && defined(HAVE_MUELU_THYRA)
   if (solverName == "Stratimikos") {
-
     using Teuchos::ParameterList;
     using Teuchos::rcp_dynamic_cast;
 
@@ -407,8 +406,8 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
       sublists.pop_front();
       std::list<std::string> keys;
       for (Teuchos::ParameterList::ConstIterator it = sublist->begin(); it != sublist->end(); ++it) {
-        const std::string &entryName = sublist->name(it);
-        const Teuchos::ParameterEntry &theEntry = sublist->entry(it);
+        const std::string& entryName            = sublist->name(it);
+        const Teuchos::ParameterEntry& theEntry = sublist->entry(it);
         if (theEntry.isList()) {
           Teuchos::ParameterList& sl = sublist->sublist(entryName);
           sublists.push_back(&sl);
@@ -430,8 +429,8 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
           sublist->set(*key_it, Kn_Matrix);
         else if (value == "Coordinates")
           sublist->set(*key_it, coords);
-        // else if (*key_it == "Nullspace")
-        //     sublist->set(*key_it, Nullspace);
+          // else if (*key_it == "Nullspace")
+          //     sublist->set(*key_it, Nullspace);
 #ifdef HAVE_MUELU_EPETRA
         else if (value == "eD0") {
           auto eD0 = Teuchos::rcp_dynamic_cast<EpetraCrsMatrix>(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(D0_Matrix, true)->getCrsMatrix(), true)->getEpetra_CrsMatrix();
@@ -451,22 +450,23 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
     // Build Thyra linear algebra objects
     RCP<const Thyra::LinearOpBase<Scalar> > thyraA;
     if (GmhdA_Matrix.is_null())  // are we doing MHD as opposed to GMHD?
-         thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(SM_Matrix)->getCrsMatrix());
-    else thyraA = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(GmhdA_Matrix)->getCrsMatrix());
-    RCP<      Thyra::VectorBase<Scalar> >thyraX = Teuchos::rcp_const_cast<Thyra::VectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraVector(X->getVectorNonConst(0)));
+      thyraA = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(SM_Matrix)->getCrsMatrix());
+    else
+      thyraA = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyra(Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(GmhdA_Matrix)->getCrsMatrix());
+    RCP<Thyra::VectorBase<Scalar> > thyraX = Teuchos::rcp_const_cast<Thyra::VectorBase<Scalar> >(Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraVector(X->getVectorNonConst(0)));
     // TODO: Why do we loose a reference when running this with Epetra?
-    RCP<const Thyra::VectorBase<Scalar> >thyraB = Xpetra::ThyraUtils<Scalar,LocalOrdinal,GlobalOrdinal,Node>::toThyraVector(B->getVector(0));
+    RCP<const Thyra::VectorBase<Scalar> > thyraB = Xpetra::ThyraUtils<Scalar, LocalOrdinal, GlobalOrdinal, Node>::toThyraVector(B->getVector(0));
 
     // Build Stratimikos solver
-    Stratimikos::LinearSolverBuilder<Scalar> linearSolverBuilder;  // This is the Stratimikos main class (= factory of solver factory).
-    Stratimikos::enableMueLuRefMaxwell<Scalar,LocalOrdinal,GlobalOrdinal,Node>(linearSolverBuilder);                // Register MueLu as a Stratimikos preconditioner strategy.
-    Stratimikos::enableMueLuMaxwell1<Scalar,LocalOrdinal,GlobalOrdinal,Node>(linearSolverBuilder);
-    linearSolverBuilder.setParameterList(rcp(&stratimikosParams,false));              // Setup solver parameters using a Stratimikos parameter list.
+    Stratimikos::LinearSolverBuilder<Scalar> linearSolverBuilder;                                        // This is the Stratimikos main class (= factory of solver factory).
+    Stratimikos::enableMueLuRefMaxwell<Scalar, LocalOrdinal, GlobalOrdinal, Node>(linearSolverBuilder);  // Register MueLu as a Stratimikos preconditioner strategy.
+    Stratimikos::enableMueLuMaxwell1<Scalar, LocalOrdinal, GlobalOrdinal, Node>(linearSolverBuilder);
+    linearSolverBuilder.setParameterList(rcp(&stratimikosParams, false));  // Setup solver parameters using a Stratimikos parameter list.
 
     // Build a new "solver factory" according to the previously specified parameter list.
     RCP<Thyra::LinearOpWithSolveFactoryBase<Scalar> > solverFactory = Thyra::createLinearSolveStrategy(linearSolverBuilder);
-    auto precFactory = solverFactory->getPreconditionerFactory();
-    auto prec = precFactory->createPrec();
+    auto precFactory                                                = solverFactory->getPreconditionerFactory();
+    auto prec                                                       = precFactory->createPrec();
 
     // Build a Thyra operator corresponding to A^{-1} computed using the Stratimikos solver.
     Thyra::initializePrec<Scalar>(*precFactory, thyraA, prec.ptr());
@@ -488,52 +488,66 @@ bool SetupSolve(std::map<std::string, void*> inputs) {
     success = (status.solveStatus == Thyra::SOLVE_STATUS_CONVERGED);
 
     if (reuse) {
-      for (int solveno = 0; solveno<2; solveno++) {
+      for (int solveno = 0; solveno < 2; solveno++) {
         auto tm6 = TimeMonitor::getNewTimer("Maxwell: 4 - Setup Re");
         Thyra::initializePrec<Scalar>(*precFactory, thyraA, prec.ptr());
         comm->barrier();
         tm6 = Teuchos::null;
 
         auto tm7 = TimeMonitor::getNewTimer("Maxwell: 5 - Solve Re");
-        status = Thyra::solve<Scalar>(*thyraInverseA, Thyra::NOTRANS, *thyraB, thyraX.ptr());
+        status   = Thyra::solve<Scalar>(*thyraInverseA, Thyra::NOTRANS, *thyraB, thyraX.ptr());
         comm->barrier();
         tm7 = Teuchos::null;
       }
     }
 
-  } // HAVE_MUELU_STRATIMIKOS && HAVE_MUELU_THYRA
+  }  // HAVE_MUELU_STRATIMIKOS && HAVE_MUELU_THYRA
 #endif
   comm->barrier();
 
   return success;
-} // SetupSolve
-
+}  // SetupSolve
 
-template<typename Scalar,class LocalOrdinal,class GlobalOrdinal,class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <typename Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
   out->setOutputToRootOnly(0);
 
-  bool        printTimings      = true;               clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-  std::string timingsFormat     = "table-fixed";      clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-  double scaling                = 1.0;                clp.setOption("scaling",               &scaling,           "scale mass term");
-  std::string solverName        = "Belos";            clp.setOption("solverName",            &solverName,        "Name of iterative linear solver "
-                                                                                                                 "to use for solving the linear system. "
-                                                                                                                 "(\"Belos\" or \"Stratimikos\")");
-  std::string belosSolverType   = "Block CG";         clp.setOption("belosSolverType",       &belosSolverType,   "Name of the Belos linear solver");
-  std::string precType          = "MueLu-RefMaxwell"; clp.setOption("precType",              &precType,          "preconditioner to use (MueLu-RefMaxwell|ML-RefMaxwell|none)");
-  std::string xml               = "";                 clp.setOption("xml",                   &xml,               "xml file with solver parameters (default: \"Maxwell.xml\")");
-  std::string belos_xml         = "Belos.xml";        clp.setOption("belos_xml",             &belos_xml,         "xml file with Belos solver parameters (default: \"Belos.xml\")");
-  std::string stratimikos_xml   = "Stratimikos.xml";  clp.setOption("stratimikos_xml",       &stratimikos_xml,   "xml file with Stratimikos solver parameters (default: \"Stratimikos.xml\")");
-  int         numResolves       = 0;                  clp.setOption("resolve",               &numResolves,       "#times to redo solve");
-  bool        reuse             = false;              clp.setOption("reuse", "no-reuse",     &reuse,             "test reuse");
-  double      tol               = 1e-10;              clp.setOption("tol",                   &tol,               "solver convergence tolerance");
-  int         maxIts            = 200;                clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-  bool        use_stacked_timer = false;              clp.setOption("stacked-timer",
-                                                                    "no-stacked-timer",      &use_stacked_timer, "use stacked timer");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  std::string timingsFormat = "table-fixed";
+  clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+  double scaling = 1.0;
+  clp.setOption("scaling", &scaling, "scale mass term");
+  std::string solverName = "Belos";
+  clp.setOption("solverName", &solverName,
+                "Name of iterative linear solver "
+                "to use for solving the linear system. "
+                "(\"Belos\" or \"Stratimikos\")");
+  std::string belosSolverType = "Block CG";
+  clp.setOption("belosSolverType", &belosSolverType, "Name of the Belos linear solver");
+  std::string precType = "MueLu-RefMaxwell";
+  clp.setOption("precType", &precType, "preconditioner to use (MueLu-RefMaxwell|ML-RefMaxwell|none)");
+  std::string xml = "";
+  clp.setOption("xml", &xml, "xml file with solver parameters (default: \"Maxwell.xml\")");
+  std::string belos_xml = "Belos.xml";
+  clp.setOption("belos_xml", &belos_xml, "xml file with Belos solver parameters (default: \"Belos.xml\")");
+  std::string stratimikos_xml = "Stratimikos.xml";
+  clp.setOption("stratimikos_xml", &stratimikos_xml, "xml file with Stratimikos solver parameters (default: \"Stratimikos.xml\")");
+  int numResolves = 0;
+  clp.setOption("resolve", &numResolves, "#times to redo solve");
+  bool reuse = false;
+  clp.setOption("reuse", "no-reuse", &reuse, "test reuse");
+  double tol = 1e-10;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  bool use_stacked_timer = false;
+  clp.setOption("stacked-timer",
+                "no-stacked-timer", &use_stacked_timer, "use stacked timer");
 
   std::string S_file, D0_file, M1_file, M0_file;
   if (!Teuchos::ScalarTraits<Scalar>::isComplex) {
@@ -547,29 +561,39 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     M1_file = "M1_complex.mat";
     M0_file = "M0_complex.mat";
   }
-                                                      clp.setOption("S",                     &S_file);
-  std::string SM_file           = "";                 clp.setOption("SM",                    &SM_file);
-  std::string Kn_file           = "";                 clp.setOption("Kn",                    &Kn_file);
-                                                      clp.setOption("D0",                    &D0_file);
-                                                      clp.setOption("M1",                    &M1_file);
-  std::string Ms_file           = "";                 clp.setOption("Ms",                    &Ms_file);
-                                                      clp.setOption("M0",                    &M0_file);
-  std::string M0inv_file        = "";                 clp.setOption("M0inv",                 &M0inv_file);
-
-  std::string coords_file       = "coords.mat";       clp.setOption("coords",                &coords_file);
-  std::string nullspace_file    = "";                 clp.setOption("nullspace",             &nullspace_file);
-  std::string material_file     = "";                 clp.setOption("material",              &material_file);
-
-  std::string rhs_file          = "";                 clp.setOption("rhs",                   &rhs_file);
-  std::string x0_file           = "";                 clp.setOption("x0",                    &x0_file);
-  std::string GmhdA_file         = "";                 clp.setOption("GmhdA",                  &GmhdA_file);
+  clp.setOption("S", &S_file);
+  std::string SM_file = "";
+  clp.setOption("SM", &SM_file);
+  std::string Kn_file = "";
+  clp.setOption("Kn", &Kn_file);
+  clp.setOption("D0", &D0_file);
+  clp.setOption("M1", &M1_file);
+  std::string Ms_file = "";
+  clp.setOption("Ms", &Ms_file);
+  clp.setOption("M0", &M0_file);
+  std::string M0inv_file = "";
+  clp.setOption("M0inv", &M0inv_file);
+
+  std::string coords_file = "coords.mat";
+  clp.setOption("coords", &coords_file);
+  std::string nullspace_file = "";
+  clp.setOption("nullspace", &nullspace_file);
+  std::string material_file = "";
+  clp.setOption("material", &material_file);
+
+  std::string rhs_file = "";
+  clp.setOption("rhs", &rhs_file);
+  std::string x0_file = "";
+  clp.setOption("x0", &x0_file);
+  std::string GmhdA_file = "";
+  clp.setOption("GmhdA", &GmhdA_file);
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   if (precType == "MueLu-Reitzinger" || precType == "MueLu-Maxwell1" || precType == "ML-Maxwell") {
@@ -578,7 +602,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     M0_file = "";
   }
 
-  if (xml == ""){
+  if (xml == "") {
     if (precType == "MueLu-RefMaxwell")
       xml = "Maxwell.xml";
     else if (precType == "MueLu-Reitzinger" || precType == "MueLu-Maxwell1")
@@ -611,13 +635,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
   // gradient matrix
   try {
-    std::string base = D0_file.substr(0, D0_file.find_last_of('/')+1);
-    std::string D0_filename = D0_file.substr(D0_file.find_last_of('/')+1, std::string::npos);
-    std::string edgeMap_file = base + "rowmap_"    + D0_filename;
+    std::string base         = D0_file.substr(0, D0_file.find_last_of('/') + 1);
+    std::string D0_filename  = D0_file.substr(D0_file.find_last_of('/') + 1, std::string::npos);
+    std::string edgeMap_file = base + "rowmap_" + D0_filename;
     std::string nodeMap_file = base + "domainmap_" + D0_filename;
-    std::string colMap_file  = base + "colmap_"    + D0_filename;
-    node_map = Xpetra::IO<SC, LO, GO, NO>::ReadMap(nodeMap_file, lib, comm);
-    edge_map = Xpetra::IO<SC, LO, GO, NO>::ReadMap(edgeMap_file, lib, comm);
+    std::string colMap_file  = base + "colmap_" + D0_filename;
+    node_map                 = Xpetra::IO<SC, LO, GO, NO>::ReadMap(nodeMap_file, lib, comm);
+    edge_map                 = Xpetra::IO<SC, LO, GO, NO>::ReadMap(edgeMap_file, lib, comm);
     RCP<const Map> colMap;
     if (comm->getSize() > 1)
       colMap = Xpetra::IO<SC, LO, GO, NO>::ReadMap(colMap_file, lib, comm);
@@ -625,16 +649,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   } catch (const std::exception&) {
     // *out << "Skipping D0 maps, because: " << e.what() << std::endl;
     D0_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(D0_file, lib, comm);
-    node_map = D0_Matrix->getDomainMap();
-    edge_map = D0_Matrix->getRangeMap();
+    node_map  = D0_Matrix->getDomainMap();
+    edge_map  = D0_Matrix->getRangeMap();
   }
 
   // build stiffness plus mass matrix (SM_Matrix)
   if (SM_file == "") {
     // edge stiffness matrix
     RCP<Matrix> S_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(S_file, edge_map);
-    M1_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(M1_file, edge_map);
-    Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::TwoMatrixAdd(*S_Matrix,false,(SC)1.0,*M1_Matrix,false,scaling,SM_Matrix,*out);
+    M1_Matrix            = Xpetra::IO<SC, LO, GO, NO>::Read(M1_file, edge_map);
+    Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*S_Matrix, false, (SC)1.0, *M1_Matrix, false, scaling, SM_Matrix, *out);
     SM_Matrix->fillComplete();
   } else {
     SM_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(SM_file, edge_map);
@@ -643,39 +667,41 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   if (GmhdA_file != "") {
     Teuchos::ArrayView<const GO> edgeMapEleList = edge_map->getLocalElementList();
-    Teuchos::Array<GO> elementList(edgeMapEleList.size()*2);
+    Teuchos::Array<GO> elementList(edgeMapEleList.size() * 2);
     for (int i = 0; i < edgeMapEleList.size(); i++) {
-       elementList[2*i  ] = 2*edgeMapEleList[i];
-       elementList[2*i+1] = 2*edgeMapEleList[i] + 1;
+      elementList[2 * i]     = 2 * edgeMapEleList[i];
+      elementList[2 * i + 1] = 2 * edgeMapEleList[i] + 1;
     }
-    edgeOrGmhdMap = MapFactory::Build(edge_map->lib(), edge_map->getGlobalNumElements()*2, elementList, edge_map->getIndexBase(), edge_map->getComm());
-    GmhdA_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(GmhdA_file, edgeOrGmhdMap);
-  }
-  else edgeOrGmhdMap = edge_map;
+    edgeOrGmhdMap = MapFactory::Build(edge_map->lib(), edge_map->getGlobalNumElements() * 2, elementList, edge_map->getIndexBase(), edge_map->getComm());
+    GmhdA_Matrix  = Xpetra::IO<SC, LO, GO, NO>::Read(GmhdA_file, edgeOrGmhdMap);
+  } else
+    edgeOrGmhdMap = edge_map;
 
   if (Ms_file != "")
     Ms_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(Ms_file, edge_map);
   else
     Ms_Matrix = M1_Matrix;
-  if (Kn_file!="")
+  if (Kn_file != "")
     Kn_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(Kn_file, node_map);
 
   if ((M0inv_file == "") && (M0_file != "")) {
     // nodal mass matrix
     RCP<Matrix> M0_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(M0_file, node_map);
     // build lumped mass matrix inverse (M0inv_Matrix)
-    RCP<Vector> diag = Utilities::GetLumpedMatrixDiagonal(*M0_Matrix);
+    RCP<Vector> diag                    = Utilities::GetLumpedMatrixDiagonal(*M0_Matrix);
     RCP<CrsMatrixWrap> M0inv_MatrixWrap = rcp(new CrsMatrixWrap(node_map, node_map, 0));
-    RCP<CrsMatrix> M0inv_CrsMatrix = M0inv_MatrixWrap->getCrsMatrix();
+    RCP<CrsMatrix> M0inv_CrsMatrix      = M0inv_MatrixWrap->getCrsMatrix();
     Teuchos::ArrayRCP<size_t> rowPtr;
     Teuchos::ArrayRCP<LO> colInd;
     Teuchos::ArrayRCP<SC> values;
     Teuchos::ArrayRCP<const SC> diags = diag->getData(0);
-    size_t nodeNumElements = node_map->getLocalNumElements();
+    size_t nodeNumElements            = node_map->getLocalNumElements();
     M0inv_CrsMatrix->allocateAllValues(nodeNumElements, rowPtr, colInd, values);
     SC ONE = Teuchos::ScalarTraits<Scalar>::one();
     for (size_t i = 0; i < nodeNumElements; i++) {
-      rowPtr[i] = i;  colInd[i] = i;  values[i] = ONE / diags[i];
+      rowPtr[i] = i;
+      colInd[i] = i;
+      values[i] = ONE / diags[i];
     }
     rowPtr[nodeNumElements] = nodeNumElements;
     M0inv_CrsMatrix->setAllValues(rowPtr, colInd, values);
@@ -694,31 +720,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
   RCP<MultiVector> material = Teuchos::null;
   if (material_file != "") {
-    material  = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(material_file, node_map);
+    material = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(material_file, node_map);
   }
 
   // set parameters
   Teuchos::ParameterList params, belosParams, stratimikosParams;
-  Teuchos::updateParametersFromXmlFileAndBroadcast(xml,            Teuchos::Ptr<Teuchos::ParameterList>(&params),           *comm);
-  Teuchos::updateParametersFromXmlFileAndBroadcast(belos_xml,      Teuchos::Ptr<Teuchos::ParameterList>(&belosParams),      *comm);
-  Teuchos::updateParametersFromXmlFileAndBroadcast(stratimikos_xml,Teuchos::Ptr<Teuchos::ParameterList>(&stratimikosParams),*comm);
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xml, Teuchos::Ptr<Teuchos::ParameterList>(&params), *comm);
+  Teuchos::updateParametersFromXmlFileAndBroadcast(belos_xml, Teuchos::Ptr<Teuchos::ParameterList>(&belosParams), *comm);
+  Teuchos::updateParametersFromXmlFileAndBroadcast(stratimikos_xml, Teuchos::Ptr<Teuchos::ParameterList>(&stratimikosParams), *comm);
   belosParams.sublist(belosSolverType).set("Maximum Iterations", maxIts);
-  belosParams.sublist(belosSolverType).set("Convergence Tolerance",tol);
+  belosParams.sublist(belosSolverType).set("Convergence Tolerance", tol);
 
   // setup LHS, RHS
   RCP<MultiVector> X, X0, B;
   if (rhs_file == "") {
-    B = MultiVectorFactory::Build(edgeOrGmhdMap,1);
-    RCP<MultiVector> vec = MultiVectorFactory::Build(edgeOrGmhdMap,1);
-    vec -> putScalar(Teuchos::ScalarTraits<Scalar>::one());
+    B                    = MultiVectorFactory::Build(edgeOrGmhdMap, 1);
+    RCP<MultiVector> vec = MultiVectorFactory::Build(edgeOrGmhdMap, 1);
+    vec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
     if (GmhdA_Matrix.is_null())  // are we doing MHD as opposed to GMHD?
-         SM_Matrix->apply(*vec,*B);
-    else GmhdA_Matrix->apply(*vec,*B);
+      SM_Matrix->apply(*vec, *B);
+    else
+      GmhdA_Matrix->apply(*vec, *B);
   } else
     B = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(rhs_file, edgeOrGmhdMap);
 
-  X = MultiVectorFactory::Build(edgeOrGmhdMap,1);
-  X -> putScalar(Teuchos::ScalarTraits<Scalar>::zero());
+  X = MultiVectorFactory::Build(edgeOrGmhdMap, 1);
+  X->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
   if (x0_file != "")
     X0 = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(x0_file, edgeOrGmhdMap);
 
@@ -726,36 +753,36 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   tm = Teuchos::null;
 
   std::map<std::string, void*> inputs;
-  inputs["SM"]              = &SM_Matrix;
-  inputs["GmhdA"]           = &GmhdA_Matrix;
-  inputs["D0"]              = &D0_Matrix;
-  inputs["M1"]              = &M1_Matrix;
-  inputs["Ms"]              = &Ms_Matrix;
-  inputs["M0inv"]           = &M0inv_Matrix;
-  inputs["Kn"]              = &Kn_Matrix;
-
-  inputs["coordinates"]     = &coords;
-  inputs["nullspace"]       = &nullspace;
-  inputs["material"]        = &material;
-
-  inputs["B"]               = &B;
-  inputs["X"]               = &X;
-  inputs["X0"]              = &X0;
-
-  inputs["params"]          = &params;
-  inputs["solverName"]      = &solverName;
-  inputs["belosSolverType"] = &belosSolverType;
-  inputs["precType"]        = &precType;
-  inputs["belosParams"]     = &belosParams;
+  inputs["SM"]    = &SM_Matrix;
+  inputs["GmhdA"] = &GmhdA_Matrix;
+  inputs["D0"]    = &D0_Matrix;
+  inputs["M1"]    = &M1_Matrix;
+  inputs["Ms"]    = &Ms_Matrix;
+  inputs["M0inv"] = &M0inv_Matrix;
+  inputs["Kn"]    = &Kn_Matrix;
+
+  inputs["coordinates"] = &coords;
+  inputs["nullspace"]   = &nullspace;
+  inputs["material"]    = &material;
+
+  inputs["B"]  = &B;
+  inputs["X"]  = &X;
+  inputs["X0"] = &X0;
+
+  inputs["params"]            = &params;
+  inputs["solverName"]        = &solverName;
+  inputs["belosSolverType"]   = &belosSolverType;
+  inputs["precType"]          = &precType;
+  inputs["belosParams"]       = &belosParams;
   inputs["stratimikosParams"] = &stratimikosParams;
-  inputs["numResolves"]     = &numResolves;
-  inputs["reuse"]           = &reuse;
+  inputs["numResolves"]       = &numResolves;
+  inputs["reuse"]             = &reuse;
 
-  inputs["lib"]             = &lib;
-  inputs["comm"]            = &comm;
-  inputs["out"]             = &out;
+  inputs["lib"]  = &lib;
+  inputs["comm"] = &comm;
+  inputs["out"]  = &out;
 
-  bool success = SetupSolve<Scalar,LocalOrdinal,GlobalOrdinal,Node>(inputs);
+  bool success = SetupSolve<Scalar, LocalOrdinal, GlobalOrdinal, Node>(inputs);
 
   globalTimeMonitor = Teuchos::null;
 
@@ -768,20 +795,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     } else {
       RCP<Teuchos::ParameterList> reportParams = rcp(new Teuchos::ParameterList);
       if (timingsFormat == "yaml") {
-        reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-        reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
+        reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+        reportParams->set("YAML style", "compact");  // "spacious" or "compact"
       }
-      reportParams->set("How to merge timer sets",   "Union");
-      reportParams->set("alwaysWriteLocal",          false);
-      reportParams->set("writeGlobalStats",          true);
-      reportParams->set("writeZeroTimers",           false);
+      reportParams->set("How to merge timer sets", "Union");
+      reportParams->set("alwaysWriteLocal", false);
+      reportParams->set("writeGlobalStats", true);
+      reportParams->set("writeZeroTimers", false);
       // FIXME: no "ignoreZeroTimers"
 
       const std::string filter = "";
 
       std::ios_base::fmtflags ff(out->flags());
-      if (timingsFormat == "table-fixed") *out << std::fixed;
-      else * out << std::scientific;
+      if (timingsFormat == "table-fixed")
+        *out << std::fixed;
+      else
+        *out << std::scientific;
       TimeMonitor::report(comm.ptr(), *out, filter, reportParams);
       *out << std::setiosflags(ff);
     }
@@ -789,13 +818,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
   TimeMonitor::clearCounters();
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/maxwell/ReitzingerPFactory.cpp b/packages/muelu/test/maxwell/ReitzingerPFactory.cpp
index 958bcf1103a5..60092f618222 100644
--- a/packages/muelu/test/maxwell/ReitzingerPFactory.cpp
+++ b/packages/muelu/test/maxwell/ReitzingerPFactory.cpp
@@ -81,54 +81,50 @@ using Teuchos::rcp;
 // Copy & paste from MueLu_TestHelpers.cpp
 namespace MueLuTests {
 
-  // static members initialization of the class TestHelpers::Parameters
-  Xpetra::Parameters TestHelpers::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
+// static members initialization of the class TestHelpers::Parameters
+Xpetra::Parameters TestHelpers::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
 
-}
+}  // namespace MueLuTests
 
-template <class Scalar,class LocalOrdinal, class GlobalOrdinal, class Node>
- typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-CheckCommutingProperty(MueLu::Level & nodeLevel_coarse, MueLu::Level & edgeLevel_fine, MueLu::Level & edgeLevel_coarse) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+CheckCommutingProperty(MueLu::Level& nodeLevel_coarse, MueLu::Level& edgeLevel_fine, MueLu::Level& edgeLevel_coarse) {
 #include <MueLu_UseShortNames.hpp>
-  RCP<Matrix> Pn = nodeLevel_coarse.Get<RCP<Matrix> >("P");
-  RCP<Matrix> Pe = edgeLevel_coarse.Get<RCP<Matrix> >("P");
+  RCP<Matrix> Pn   = nodeLevel_coarse.Get<RCP<Matrix> >("P");
+  RCP<Matrix> Pe   = edgeLevel_coarse.Get<RCP<Matrix> >("P");
   RCP<Matrix> D0_f = edgeLevel_fine.Get<RCP<Matrix> >("D0");
   RCP<Matrix> D0_c = edgeLevel_coarse.Get<RCP<Matrix> >("D0");
 
-  using XMM = Xpetra::MatrixMatrix<SC,LO,GO,NO>;
-  using MT   = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-  SC one = Teuchos::ScalarTraits<SC>::one();
-  SC zero = Teuchos::ScalarTraits<SC>::zero();
-  
+  using XMM = Xpetra::MatrixMatrix<SC, LO, GO, NO>;
+  using MT  = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  SC one    = Teuchos::ScalarTraits<SC>::one();
+  SC zero   = Teuchos::ScalarTraits<SC>::zero();
+
   RCP<Matrix> dummy;
-  RCP<Teuchos::FancyOStream> out0= Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+  RCP<Teuchos::FancyOStream> out0 = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+
+  RCP<Matrix> left  = XMM::Multiply(*Pe, false, *D0_c, false, dummy, *out0);
+  RCP<Matrix> right = XMM::Multiply(*D0_f, false, *Pn, false, dummy, *out0);
 
-  RCP<Matrix> left  = XMM::Multiply(*Pe,false,*D0_c,false,dummy,*out0);
-  RCP<Matrix> right = XMM::Multiply(*D0_f,false,*Pn,false,dummy,*out0);
-  
   // We need a non-FC matrix for the add, sadly
-  RCP<CrsMatrix> sum_c = CrsMatrixFactory::Build(left->getRowMap(),left->getLocalMaxNumRowEntries()+right->getLocalMaxNumRowEntries());
+  RCP<CrsMatrix> sum_c  = CrsMatrixFactory::Build(left->getRowMap(), left->getLocalMaxNumRowEntries() + right->getLocalMaxNumRowEntries());
   RCP<Matrix> summation = rcp(new CrsMatrixWrap(sum_c));
-  XMM::TwoMatrixAdd(*left,  false, one, *summation, zero);
+  XMM::TwoMatrixAdd(*left, false, one, *summation, zero);
   XMM::TwoMatrixAdd(*right, false, -one, *summation, one);
-  
+
   MT norm = summation->getFrobeniusNorm();
   return norm;
 }
 
-
-
-
-template<typename Scalar,class LocalOrdinal,class GlobalOrdinal,class Node>
-void read_matrix(Xpetra::UnderlyingLib & lib,RCP<const Teuchos::Comm<int> > & comm,
-                 RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & SM_Matrix,
-                 RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & D0_Matrix,
-                 RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & Kn_Matrix,
-                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> > & coords) {
+template <typename Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void read_matrix(Xpetra::UnderlyingLib& lib, RCP<const Teuchos::Comm<int> >& comm,
+                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& SM_Matrix,
+                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& D0_Matrix,
+                 RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Kn_Matrix,
+                 RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& coords) {
 #include <MueLu_UseShortNames.hpp>
   RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
 
-
   std::string S_file, M1_file, D0_file;
   if (!Teuchos::ScalarTraits<Scalar>::isComplex) {
     S_file  = "S.mat";
@@ -147,13 +143,13 @@ void read_matrix(Xpetra::UnderlyingLib & lib,RCP<const Teuchos::Comm<int> > & co
 
   // gradient matrix
   try {
-    std::string base = D0_file.substr(0, D0_file.find_last_of('/')+1);
-    std::string D0_filename = D0_file.substr(D0_file.find_last_of('/')+1, std::string::npos);
-    std::string edgeMap_file = base + "rowmap_"    + D0_filename;
+    std::string base         = D0_file.substr(0, D0_file.find_last_of('/') + 1);
+    std::string D0_filename  = D0_file.substr(D0_file.find_last_of('/') + 1, std::string::npos);
+    std::string edgeMap_file = base + "rowmap_" + D0_filename;
     std::string nodeMap_file = base + "domainmap_" + D0_filename;
-    std::string colMap_file  = base + "colmap_"    + D0_filename;
-    node_map = Xpetra::IO<SC, LO, GO, NO>::ReadMap(nodeMap_file, lib, comm);
-    edge_map = Xpetra::IO<SC, LO, GO, NO>::ReadMap(edgeMap_file, lib, comm);
+    std::string colMap_file  = base + "colmap_" + D0_filename;
+    node_map                 = Xpetra::IO<SC, LO, GO, NO>::ReadMap(nodeMap_file, lib, comm);
+    edge_map                 = Xpetra::IO<SC, LO, GO, NO>::ReadMap(edgeMap_file, lib, comm);
     RCP<const Map> colMap;
     if (comm->getSize() > 1)
       colMap = Xpetra::IO<SC, LO, GO, NO>::ReadMap(colMap_file, lib, comm);
@@ -161,45 +157,42 @@ void read_matrix(Xpetra::UnderlyingLib & lib,RCP<const Teuchos::Comm<int> > & co
   } catch (const std::exception&) {
     // *out << "Skipping D0 maps, because: " << e.what() << std::endl;
     D0_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(D0_file, lib, comm);
-    node_map = D0_Matrix->getDomainMap();
-    edge_map = D0_Matrix->getRangeMap();
+    node_map  = D0_Matrix->getDomainMap();
+    edge_map  = D0_Matrix->getRangeMap();
   }
 
   // build stiffness plus mass matrix (SM_Matrix)
   // edge stiffness matrix
-  RCP<Matrix> S_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(S_file, edge_map);
+  RCP<Matrix> S_Matrix  = Xpetra::IO<SC, LO, GO, NO>::Read(S_file, edge_map);
   RCP<Matrix> M1_Matrix = Xpetra::IO<SC, LO, GO, NO>::Read(M1_file, edge_map);
-  Scalar one = Teuchos::ScalarTraits<SC>::one();
-  Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::TwoMatrixAdd(*S_Matrix,false,one,*M1_Matrix,false,one,SM_Matrix,*out);
+  Scalar one            = Teuchos::ScalarTraits<SC>::one();
+  Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::TwoMatrixAdd(*S_Matrix, false, one, *M1_Matrix, false, one, SM_Matrix, *out);
   SM_Matrix->fillComplete();
 
-
   // coordinates
   coords = Xpetra::IO<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::ReadMultiVector(coords_file, node_map);
 
   // Generate Kn matrix
   {
-    RCP<Matrix> temp = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(SM_Matrix->getRangeMap());
-    Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*SM_Matrix,false,*D0_Matrix,false,*temp,true,true);
-    Kn_Matrix = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(D0_Matrix->getDomainMap());
-    Xpetra::MatrixMatrix<SC,LO,GO,NO>::Multiply(*D0_Matrix,true,*temp,false,*Kn_Matrix,true,true);
+    RCP<Matrix> temp = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(SM_Matrix->getRangeMap());
+    Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*SM_Matrix, false, *D0_Matrix, false, *temp, true, true);
+    Kn_Matrix = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(D0_Matrix->getDomainMap());
+    Xpetra::MatrixMatrix<SC, LO, GO, NO>::Multiply(*D0_Matrix, true, *temp, false, *Kn_Matrix, true, true);
   }
-
 }
 
 namespace MueLuTests {
 
 TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup2Level_Unsmoothed, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-  
   RCP<Matrix> SM_Matrix, D0_Matrix, Kn_Matrix;
   RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coords;
-  read_matrix<SC,LO,GO,NO>(lib,comm,SM_Matrix,D0_Matrix,Kn_Matrix,coords);
+  read_matrix<SC, LO, GO, NO>(lib, comm, SM_Matrix, D0_Matrix, Kn_Matrix, coords);
 
   int NumLevels = 2;
   // This guy works by generating a nodal hierarchy, copying the relevant data to the edge hierarchy
@@ -211,193 +204,193 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup2Level_Unsmoothed, Sc
   NodeH.SetMaxCoarseSize(10);
 
   // Generate Node Hierarchy
-  out<<"*** Setting Up Node Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Node Hierarchy *** " << std::endl;
   {
     RCP<MueLu::Level> Finest = NodeH.GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
     Finest->Set("A", Kn_Matrix);
 
     // Level 0
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     M1.SetKokkosRefactor(false);
     Teuchos::ParameterList tp_list;
-    tp_list.set("tentative: constant column sums",false);
-    tp_list.set("tentative: calculate qr",false);
-    RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+    tp_list.set("tentative: constant column sums", false);
+    tp_list.set("tentative: calculate qr", false);
+    RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
     PnodalFact->SetParameterList(tp_list);
-    M1.SetFactory("P",PnodalFact);
+    M1.SetFactory("P", PnodalFact);
     M1.SetFactory("CoarseSolver", Teuchos::null);
 
-
     bool r;
-    r = NodeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+    r = NodeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+    TEST_EQUALITY(r, true);
     RCP<Level> l0 = NodeH.GetLevel(0);
     RCP<Level> l1 = NodeH.GetLevel(1);
   }
 
-  out<<"*** Copy Node->Edge Data *** "<<std::endl;
+  out << "*** Copy Node->Edge Data *** " << std::endl;
   // Copy Data to Edge Hierarchy
-  for(int i=0; i<NumLevels; i++) {
+  for (int i = 0; i < NumLevels; i++) {
     EdgeH.AddNewLevel();
     RCP<Level> NodeL = NodeH.GetLevel(i);
     RCP<Level> EdgeL = EdgeH.GetLevel(i);
 
-    EdgeL->Set("NodeAggMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    EdgeL->Set("NodeMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    if(i!=0) {
-      EdgeL->Set("Pnodal",NodeL->Get<RCP<Matrix> >("P"));
+    EdgeL->Set("NodeAggMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    EdgeL->Set("NodeMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    if (i != 0) {
+      EdgeL->Set("Pnodal", NodeL->Get<RCP<Matrix> >("P"));
 
       RCP<const Matrix> P = NodeL->Get<RCP<Matrix> >("P");
       //      Xpetra::IO<SC,LO,GO,NO>::Write("Pn.mat",*P);
     }
 
-    if(i==0) {
+    if (i == 0) {
       EdgeL->Set("A", SM_Matrix);
       EdgeL->Set("D0", D0_Matrix);
     }
   }
 
   // Generate the Edge Hierarchy
-  out<<"*** Setting Up Edge Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Edge Hierarchy *** " << std::endl;
   {
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     M1.SetKokkosRefactor(false);
-    RCP<ReitzingerPFactory> PedgeFact  = rcp(new ReitzingerPFactory());
-    M1.SetFactory("P",PedgeFact);
+    RCP<ReitzingerPFactory> PedgeFact = rcp(new ReitzingerPFactory());
+    M1.SetFactory("P", PedgeFact);
     M1.SetFactory("CoarseSolver", Teuchos::null);
 
     // Do the setup
     bool r;
-    r = EdgeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+    r = EdgeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+    TEST_EQUALITY(r, true);
   }
 
-  
   RCP<Level> l0 = EdgeH.GetLevel(0);
   RCP<Level> l1 = EdgeH.GetLevel(1);
 
   RCP<Level> node_l1 = NodeH.GetLevel(1);
 
-  TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
 
   // Check the commuting property
   using MT = typename Teuchos::ScalarTraits<SC>::magnitudeType;
   Teuchos::Array<MT> norm(1), norm0(1);
   norm0[0] = Teuchos::ScalarTraits<MT>::zero();
 
-  norm[0] = CheckCommutingProperty<SC,LO,GO,NO>(*node_l1,*l0,*l1);
-  TEST_COMPARE_FLOATING_ARRAYS(norm,norm0,Teuchos::ScalarTraits<MT>::eps()*100)
-
+  norm[0] = CheckCommutingProperty<SC, LO, GO, NO>(*node_l1, *l0, *l1);
+  TEST_COMPARE_FLOATING_ARRAYS(norm, norm0, Teuchos::ScalarTraits<MT>::eps() * 100)
 }
 
-
-
 TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup2Level_AlphaSmoothed, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-  
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
   RCP<Matrix> SM_Matrix, D0_Matrix, Kn_Matrix;
   RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coords;
-  read_matrix<SC,LO,GO,NO>(lib,comm,SM_Matrix,D0_Matrix,Kn_Matrix,coords);
+  read_matrix<SC, LO, GO, NO>(lib, comm, SM_Matrix, D0_Matrix, Kn_Matrix, coords);
 
   int NumLevels = 2;
   // This guy works by generating a nodal hierarchy, copying the relevant data to the edge hierarchy
   // and then generating the edge hierarchy
   Hierarchy NodeH, EdgeH;
   NodeH.SetMaxCoarseSize(10);
-  //NodeH.EnableGraphDumping("node_graph_0",0);
-  //EdgeH.EnableGraphDumping("edge_graph_0",0);
+  // NodeH.EnableGraphDumping("node_graph_0",0);
+  // EdgeH.EnableGraphDumping("edge_graph_0",0);
 
   // Generate Node Hierarchy
-  out<<"*** Setting Up Node Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Node Hierarchy *** " << std::endl;
   {
     RCP<MueLu::Level> Finest = NodeH.GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
     Finest->Set("A", Kn_Matrix);
 
     // Level 0
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     M1.SetKokkosRefactor(false);
     Teuchos::ParameterList tp_list;
-    tp_list.set("tentative: constant column sums",false);
-    tp_list.set("tentative: calculate qr",false);
-    RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+    tp_list.set("tentative: constant column sums", false);
+    tp_list.set("tentative: calculate qr", false);
+    RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
     PnodalFact->SetParameterList(tp_list);
-    M1.SetFactory("P",PnodalFact);
+    M1.SetFactory("P", PnodalFact);
     M1.SetFactory("CoarseSolver", Teuchos::null);
 
-
     bool r;
-    r = NodeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+    r = NodeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+    TEST_EQUALITY(r, true);
     RCP<Level> l0 = NodeH.GetLevel(0);
     RCP<Level> l1 = NodeH.GetLevel(1);
   }
 
-  out<<"*** Copy Node->Edge Data *** "<<std::endl;
+  out << "*** Copy Node->Edge Data *** " << std::endl;
   // Copy Data to Edge Hierarchy
-  for(int i=0; i<NumLevels; i++) {
+  for (int i = 0; i < NumLevels; i++) {
     EdgeH.AddNewLevel();
     RCP<Level> NodeL = NodeH.GetLevel(i);
     RCP<Level> EdgeL = EdgeH.GetLevel(i);
 
-    EdgeL->Set("NodeAggMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    EdgeL->Set("NodeMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    if(i!=0) {
-      EdgeL->Set("Pnodal",NodeL->Get<RCP<Matrix> >("P"));
+    EdgeL->Set("NodeAggMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    EdgeL->Set("NodeMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    if (i != 0) {
+      EdgeL->Set("Pnodal", NodeL->Get<RCP<Matrix> >("P"));
 
       RCP<const Matrix> P = NodeL->Get<RCP<Matrix> >("P");
       //      Xpetra::IO<SC,LO,GO,NO>::Write("Pn.mat",*P);
     }
 
-    if(i==0) {
+    if (i == 0) {
       EdgeL->Set("A", SM_Matrix);
       EdgeL->Set("D0", D0_Matrix);
     }
   }
 
   // Generate the Edge Hierarchy
-  out<<"*** Setting Up Edge Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Edge Hierarchy *** " << std::endl;
   {
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Smoothed Reitzinger)
     FactoryManager M1;
     M1.SetKokkosRefactor(false);
-    RCP<ReitzingerPFactory> PedgeFact  = rcp(new ReitzingerPFactory());
-    RCP<SaPFactory> PFact = rcp(new SaPFactory());
-    PFact->SetFactory("P",PedgeFact);
-    M1.SetFactory("Ptent",PedgeFact);
-    M1.SetFactory("P",PFact);
+    RCP<ReitzingerPFactory> PedgeFact = rcp(new ReitzingerPFactory());
+    RCP<SaPFactory> PFact             = rcp(new SaPFactory());
+    PFact->SetFactory("P", PedgeFact);
+    M1.SetFactory("Ptent", PedgeFact);
+    M1.SetFactory("P", PFact);
     M1.SetFactory("CoarseSolver", Teuchos::null);
 
-
     // Do the setup
     bool r;
-    r = EdgeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+    r = EdgeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+    TEST_EQUALITY(r, true);
   }
 
-  
   RCP<Level> l0 = EdgeH.GetLevel(0);
   RCP<Level> l1 = EdgeH.GetLevel(1);
 
@@ -405,23 +398,21 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup2Level_AlphaSmoothed,
     RCP<const Matrix> Pe = l1->Get<RCP<Matrix> >("P");
     //    Xpetra::IO<SC,LO,GO,NO>::Write("Pe.mat",*Pe);
   }
-  TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
 }
 
-
-
 TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_AlphaSmoothed, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-  
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
   RCP<Matrix> SM_Matrix, D0_Matrix, Kn_Matrix;
   RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coords;
-  read_matrix<SC,LO,GO,NO>(lib,comm,SM_Matrix,D0_Matrix,Kn_Matrix,coords);
+  read_matrix<SC, LO, GO, NO>(lib, comm, SM_Matrix, D0_Matrix, Kn_Matrix, coords);
 
   int NumLevels = 3;
   // This guy works by generating a nodal hierarchy, copying the relevant data to the edge hierarchy
@@ -429,11 +420,11 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_AlphaSmoothed,
   Hierarchy NodeH, EdgeH;
   NodeH.SetMaxCoarseSize(1);
   EdgeH.SetMaxCoarseSize(1);
-  //NodeH.EnableGraphDumping("node_graph_0",0);
-  //EdgeH.EnableGraphDumping("edge_graph_0",0);
+  // NodeH.EnableGraphDumping("node_graph_0",0);
+  // EdgeH.EnableGraphDumping("edge_graph_0",0);
 
   // Generate Node Hierarchy
-  out<<"*** Setting Up Node Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Node Hierarchy *** " << std::endl;
   {
     RCP<MueLu::Level> Finest = NodeH.GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
@@ -441,20 +432,20 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_AlphaSmoothed,
     Finest->Set("Coordinates", coords);
 
     // Level 0
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     {
       M1.SetKokkosRefactor(false);
       Teuchos::ParameterList tp_list;
-      tp_list.set("tentative: constant column sums",false);
-      tp_list.set("tentative: calculate qr",false);
-      RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+      tp_list.set("tentative: constant column sums", false);
+      tp_list.set("tentative: calculate qr", false);
+      RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
       PnodalFact->SetParameterList(tp_list);
-      M1.SetFactory("P",PnodalFact);
-      M1.SetFactory("Ptent",PnodalFact);// You actually need this
+      M1.SetFactory("P", PnodalFact);
+      M1.SetFactory("Ptent", PnodalFact);  // You actually need this
     }
 
     // Level 2 (Plain aggregation)
@@ -462,113 +453,114 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_AlphaSmoothed,
     {
       M2.SetKokkosRefactor(false);
       Teuchos::ParameterList tp_list;
-      tp_list.set("tentative: constant column sums",false);
-      tp_list.set("tentative: calculate qr",false);
-      RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+      tp_list.set("tentative: constant column sums", false);
+      tp_list.set("tentative: calculate qr", false);
+      RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
       PnodalFact->SetParameterList(tp_list);
-      M2.SetFactory("P",PnodalFact);
+      M2.SetFactory("P", PnodalFact);
       M2.SetFactory("CoarseSolver", Teuchos::null);
     }
 
     bool r;
-    r = NodeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2)); TEST_EQUALITY(r, false);
-    r = NodeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
+    r = NodeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+    TEST_EQUALITY(r, true);
 
     RCP<Level> l0 = NodeH.GetLevel(0);
     RCP<Level> l1 = NodeH.GetLevel(1);
     RCP<Level> l2 = NodeH.GetLevel(2);
   }
 
-  out<<"*** Copy Node->Edge Data *** "<<std::endl;
+  out << "*** Copy Node->Edge Data *** " << std::endl;
   // Copy Data to Edge Hierarchy
-  for(int i=0; i<NumLevels; i++) {
+  for (int i = 0; i < NumLevels; i++) {
     EdgeH.AddNewLevel();
     RCP<Level> NodeL = NodeH.GetLevel(i);
     RCP<Level> EdgeL = EdgeH.GetLevel(i);
 
-    EdgeL->Set("NodeAggMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    EdgeL->Set("NodeMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    if(i!=0) {
-      EdgeL->Set("Pnodal",NodeL->Get<RCP<Matrix> >("P"));
+    EdgeL->Set("NodeAggMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    EdgeL->Set("NodeMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    if (i != 0) {
+      EdgeL->Set("Pnodal", NodeL->Get<RCP<Matrix> >("P"));
 
       RCP<const Matrix> P = NodeL->Get<RCP<Matrix> >("P");
       //      Xpetra::IO<SC,LO,GO,NO>::Write("Pn.mat",*P);
     }
 
-    if(i==0) {
+    if (i == 0) {
       EdgeL->Set("A", SM_Matrix);
       EdgeL->Set("D0", D0_Matrix);
     }
   }
 
   // Generate the Edge Hierarchy
-  out<<"*** Setting Up Edge Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Edge Hierarchy *** " << std::endl;
   {
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Smoothed Reitzinger)
     FactoryManager M1;
     {
       M1.SetKokkosRefactor(false);
-      RCP<ReitzingerPFactory> PedgeFact1  = rcp(new ReitzingerPFactory());
-      RCP<SaPFactory> PFact = rcp(new SaPFactory());
-      PFact->SetFactory("P",PedgeFact1);
-      M1.SetFactory("Ptent",PedgeFact1);
-      M1.SetFactory("D0",PedgeFact1);
-      M1.SetFactory("P",PFact);
+      RCP<ReitzingerPFactory> PedgeFact1 = rcp(new ReitzingerPFactory());
+      RCP<SaPFactory> PFact              = rcp(new SaPFactory());
+      PFact->SetFactory("P", PedgeFact1);
+      M1.SetFactory("Ptent", PedgeFact1);
+      M1.SetFactory("D0", PedgeFact1);
+      M1.SetFactory("P", PFact);
     }
 
     // Level 2 (just like level 1)
     FactoryManager M2;
     {
       M2.SetKokkosRefactor(false);
-      RCP<ReitzingerPFactory> PedgeFact2  = rcp(new ReitzingerPFactory());
-      RCP<SaPFactory> PFact = rcp(new SaPFactory());
-      PFact->SetFactory("P",PedgeFact2);
-      M2.SetFactory("Ptent",PedgeFact2);
-      M2.SetFactory("D0",PedgeFact2);
-      M2.SetFactory("P",PFact);
+      RCP<ReitzingerPFactory> PedgeFact2 = rcp(new ReitzingerPFactory());
+      RCP<SaPFactory> PFact              = rcp(new SaPFactory());
+      PFact->SetFactory("P", PedgeFact2);
+      M2.SetFactory("Ptent", PedgeFact2);
+      M2.SetFactory("D0", PedgeFact2);
+      M2.SetFactory("P", PFact);
       M2.SetFactory("CoarseSolver", Teuchos::null);
     }
-    
+
     // Do the setup
     bool r;
-    r = EdgeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2)); TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);  TEST_EQUALITY(r, true);
+    r = EdgeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+    TEST_EQUALITY(r, true);
   }
 
   RCP<Level> node_l1 = NodeH.GetLevel(1);
   RCP<Level> node_l2 = NodeH.GetLevel(2);
-  
+
   RCP<Level> l0 = EdgeH.GetLevel(0);
   RCP<Level> l1 = EdgeH.GetLevel(1);
   RCP<Level> l2 = EdgeH.GetLevel(2);
 
-  TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-
- 
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
 }
 
-
-
 TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_Unsmoothed, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-  
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
   RCP<Matrix> SM_Matrix, D0_Matrix, Kn_Matrix;
   RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coords;
-  read_matrix<SC,LO,GO,NO>(lib,comm,SM_Matrix,D0_Matrix,Kn_Matrix,coords);
-
+  read_matrix<SC, LO, GO, NO>(lib, comm, SM_Matrix, D0_Matrix, Kn_Matrix, coords);
 
   int NumLevels = 3;
   // This guy works by generating a nodal hierarchy, copying the relevant data to the edge hierarchy
@@ -576,11 +568,11 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_Unsmoothed, Sc
   Hierarchy NodeH, EdgeH;
   NodeH.SetMaxCoarseSize(1);
   EdgeH.SetMaxCoarseSize(1);
-  //NodeH.EnableGraphDumping("node_graph_0",0);
-  //EdgeH.EnableGraphDumping("edge_graph_0",0);
+  // NodeH.EnableGraphDumping("node_graph_0",0);
+  // EdgeH.EnableGraphDumping("edge_graph_0",0);
 
   // Generate Node Hierarchy
-  out<<"*** Setting Up Node Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Node Hierarchy *** " << std::endl;
   {
     RCP<MueLu::Level> Finest = NodeH.GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
@@ -588,20 +580,20 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_Unsmoothed, Sc
     Finest->Set("Coordinates", coords);
 
     // Level 0
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     {
       M1.SetKokkosRefactor(false);
       Teuchos::ParameterList tp_list;
-      tp_list.set("tentative: constant column sums",false);
-      tp_list.set("tentative: calculate qr",false);
-      RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+      tp_list.set("tentative: constant column sums", false);
+      tp_list.set("tentative: calculate qr", false);
+      RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
       PnodalFact->SetParameterList(tp_list);
-      M1.SetFactory("P",PnodalFact);
-      M1.SetFactory("Ptent",PnodalFact);// You actually need this
+      M1.SetFactory("P", PnodalFact);
+      M1.SetFactory("Ptent", PnodalFact);  // You actually need this
     }
 
     // Level 2 (Plain aggregation)
@@ -609,140 +601,134 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Setup3Level_Unsmoothed, Sc
     {
       M2.SetKokkosRefactor(false);
       Teuchos::ParameterList tp_list;
-      tp_list.set("tentative: constant column sums",false);
-      tp_list.set("tentative: calculate qr",false);
-      RCP<TentativePFactory>        PnodalFact = rcp(new TentativePFactory());
+      tp_list.set("tentative: constant column sums", false);
+      tp_list.set("tentative: calculate qr", false);
+      RCP<TentativePFactory> PnodalFact = rcp(new TentativePFactory());
       PnodalFact->SetParameterList(tp_list);
-      M2.SetFactory("P",PnodalFact);
+      M2.SetFactory("P", PnodalFact);
       M2.SetFactory("CoarseSolver", Teuchos::null);
     }
 
     bool r;
-    r = NodeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2)); TEST_EQUALITY(r, false);
-    r = NodeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
+    r = NodeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    TEST_EQUALITY(r, false);
+    r = NodeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+    TEST_EQUALITY(r, true);
 
     RCP<Level> l0 = NodeH.GetLevel(0);
     RCP<Level> l1 = NodeH.GetLevel(1);
     RCP<Level> l2 = NodeH.GetLevel(2);
   }
 
-  out<<"*** Copy Node->Edge Data *** "<<std::endl;
+  out << "*** Copy Node->Edge Data *** " << std::endl;
   // Copy Data to Edge Hierarchy
-  for(int i=0; i<NumLevels; i++) {
+  for (int i = 0; i < NumLevels; i++) {
     EdgeH.AddNewLevel();
     RCP<Level> NodeL = NodeH.GetLevel(i);
     RCP<Level> EdgeL = EdgeH.GetLevel(i);
 
-    EdgeL->Set("NodeAggMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    EdgeL->Set("NodeMatrix",NodeL->Get<RCP<Matrix> >("A"));
-    if(i!=0) {
-      EdgeL->Set("Pnodal",NodeL->Get<RCP<Matrix> >("P"));
+    EdgeL->Set("NodeAggMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    EdgeL->Set("NodeMatrix", NodeL->Get<RCP<Matrix> >("A"));
+    if (i != 0) {
+      EdgeL->Set("Pnodal", NodeL->Get<RCP<Matrix> >("P"));
 
       RCP<const Matrix> P = NodeL->Get<RCP<Matrix> >("P");
       //      Xpetra::IO<SC,LO,GO,NO>::Write("Pn.mat",*P);
     }
 
-    if(i==0) {
+    if (i == 0) {
       EdgeL->Set("A", SM_Matrix);
       EdgeL->Set("D0", D0_Matrix);
     }
   }
 
   // Generate the Edge Hierarchy
-  out<<"*** Setting Up Edge Hierarchy *** "<<std::endl;
+  out << "*** Setting Up Edge Hierarchy *** " << std::endl;
   {
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    
+
     // Level 1 (Plain aggregation)
     FactoryManager M1;
     {
       M1.SetKokkosRefactor(false);
-      RCP<ReitzingerPFactory> PFact  = rcp(new ReitzingerPFactory());
-      M1.SetFactory("Ptent",PFact);
-      M1.SetFactory("D0",PFact);
-      M1.SetFactory("P",PFact);
+      RCP<ReitzingerPFactory> PFact = rcp(new ReitzingerPFactory());
+      M1.SetFactory("Ptent", PFact);
+      M1.SetFactory("D0", PFact);
+      M1.SetFactory("P", PFact);
     }
 
     // Level 2 (just like level 1)
     FactoryManager M2;
     {
       M2.SetKokkosRefactor(false);
-      RCP<ReitzingerPFactory> PFact  = rcp(new ReitzingerPFactory());
-      M2.SetFactory("Ptent",PFact);
-      M2.SetFactory("D0",PFact);
-      M2.SetFactory("P",PFact);
+      RCP<ReitzingerPFactory> PFact = rcp(new ReitzingerPFactory());
+      M2.SetFactory("Ptent", PFact);
+      M2.SetFactory("D0", PFact);
+      M2.SetFactory("P", PFact);
       M2.SetFactory("CoarseSolver", Teuchos::null);
     }
-    
+
     // Do the setup
     bool r;
-    r = EdgeH.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2)); TEST_EQUALITY(r, false);
-    r = EdgeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);  TEST_EQUALITY(r, true);
+    r = EdgeH.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    TEST_EQUALITY(r, false);
+    r = EdgeH.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+    TEST_EQUALITY(r, true);
   }
 
   RCP<Level> node_l1 = NodeH.GetLevel(1);
   RCP<Level> node_l2 = NodeH.GetLevel(2);
-  
+
   RCP<Level> l0 = EdgeH.GetLevel(0);
   RCP<Level> l1 = EdgeH.GetLevel(1);
   RCP<Level> l2 = EdgeH.GetLevel(2);
 
   //  l0->print(std::cout,MueLu::Debug);
   //  l1->print(std::cout,MueLu::Debug);
-  //l2->print(std::cout,MueLu::Debug);
+  // l2->print(std::cout,MueLu::Debug);
 
-  TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-  TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
 
   // Check the commuting property
   using MT = typename Teuchos::ScalarTraits<SC>::magnitudeType;
   Teuchos::Array<MT> norm(1), norm0(1);
   norm0[0] = Teuchos::ScalarTraits<MT>::zero();
 
-  norm[0] = CheckCommutingProperty<SC,LO,GO,NO>(*node_l1,*l0,*l1);
-  TEST_COMPARE_FLOATING_ARRAYS(norm,norm0,Teuchos::ScalarTraits<MT>::eps()*100);
-  norm[0] = CheckCommutingProperty<SC,LO,GO,NO>(*node_l2,*l1,*l2);
-  TEST_COMPARE_FLOATING_ARRAYS(norm,norm0,Teuchos::ScalarTraits<MT>::eps()*100);
-
-
-
+  norm[0] = CheckCommutingProperty<SC, LO, GO, NO>(*node_l1, *l0, *l1);
+  TEST_COMPARE_FLOATING_ARRAYS(norm, norm0, Teuchos::ScalarTraits<MT>::eps() * 100);
+  norm[0] = CheckCommutingProperty<SC, LO, GO, NO>(*node_l2, *l1, *l2);
+  TEST_COMPARE_FLOATING_ARRAYS(norm, norm0, Teuchos::ScalarTraits<MT>::eps() * 100);
 }
 
-
-
-
-# define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup2Level_Unsmoothed, Scalar, LO, GO, Node) \
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup2Level_Unsmoothed, Scalar, LO, GO, Node)    \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup2Level_AlphaSmoothed, Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup3Level_Unsmoothed, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup3Level_Unsmoothed, Scalar, LO, GO, Node)    \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Setup3Level_AlphaSmoothed, Scalar, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
 
+}  // namespace MueLuTests
 
-# include <MueLu_ETI_4arg.hpp>
-
-}//end namespace
-
-
-
-
-int main(int argc, char* argv[])
-{
+int main(int argc, char* argv[]) {
   Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
 #ifdef HAVE_MUELU_KOKKOS
-  Kokkos::initialize(argc , argv);
+  Kokkos::initialize(argc, argv);
 #endif
 
   bool success = false;
   bool verbose = true;
-  int ierr = -1;
+  int ierr     = -1;
   try {
     // Note: the command line parameter --linAlgebra= is take into account.
     // Xpetra parameters are added to the Teuchos::CommandLineProcessor of Teuchos::UnitTestRepository in MueLu_TestHelpers.cpp
@@ -752,25 +738,26 @@ int main(int argc, char* argv[])
 
     int mypid = comm->getRank();
 
-    if (mypid  == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
-    for (int i = 0; i <comm->getSize(); i++) {
-      if (i == mypid ) {
+    if (mypid == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
+    for (int i = 0; i < comm->getSize(); i++) {
+      if (i == mypid) {
         char buf[80];
         char hostname[80];
         gethostname(hostname, sizeof(hostname));
         int pid = getpid();
         sprintf(buf, "Host: %s\tMPI rank: %d,\tPID: %d\n\tattach %d\n\tcontinue\n",
-            hostname, mypid, pid, pid);
-        printf("%s\n",buf);
+                hostname, mypid, pid, pid);
+        printf("%s\n", buf);
         fflush(stdout);
         sleep(1);
       }
     }
 
     if (mypid == 0) {
-      printf( "** Enter a character to continue > "); fflush(stdout);
+      printf("** Enter a character to continue > ");
+      fflush(stdout);
       char go = ' ';
-      scanf("%c",&go);
+      scanf("%c", &go);
     }
     comm->barrier();
 #endif
diff --git a/packages/muelu/test/meshtying/MeshTyingBlocked_SimpleSmoother.cpp b/packages/muelu/test/meshtying/MeshTyingBlocked_SimpleSmoother.cpp
index b8ad1209b055..5366a0516e56 100644
--- a/packages/muelu/test/meshtying/MeshTyingBlocked_SimpleSmoother.cpp
+++ b/packages/muelu/test/meshtying/MeshTyingBlocked_SimpleSmoother.cpp
@@ -60,27 +60,25 @@
 #include <BelosMueLuAdapter.hpp>
 
 template <typename GlobalOrdinal>
-void read_Lagr2Dof(std::string filemane, std::map<GlobalOrdinal, GlobalOrdinal> &lagr2Dof)
-{
+void read_Lagr2Dof(std::string filemane, std::map<GlobalOrdinal, GlobalOrdinal> &lagr2Dof) {
   std::fstream lagr2DofFile;
   lagr2DofFile.open(filemane);
   TEUCHOS_ASSERT(lagr2DofFile.is_open())
 
   GlobalOrdinal key;
   GlobalOrdinal value;
-  while (lagr2DofFile >> key >> value)
-  {
+  while (lagr2DofFile >> key >> value) {
     lagr2Dof[key] = value;
   }
   lagr2DofFile.close();
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
-  #include <MueLu_UseShortNames.hpp>
-  using SparseMatrixType = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
+#include <MueLu_UseShortNames.hpp>
+  using SparseMatrixType    = Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
   using tpetra_mvector_type = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-  using tpetra_map_type = Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+  using tpetra_map_type     = Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
 
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -91,15 +89,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   oblackholestream blackhole;
 
   RCP<const Comm<int>> comm = DefaultComm<int>::getComm();
-  RCP<FancyOStream> out = fancyOStream(rcpFromRef(std::cout));
+  RCP<FancyOStream> out     = fancyOStream(rcpFromRef(std::cout));
   out->setOutputToRootOnly(0);
 
-  const GO globalPrimalNumDofs = 1599;
-  const GO globalDualNumDofs = 100;
-  const GO globalNumDofs = globalPrimalNumDofs + globalDualNumDofs; // used for the maps
+  const GO globalPrimalNumDofs    = 1599;
+  const GO globalDualNumDofs      = 100;
+  const GO globalNumDofs          = globalPrimalNumDofs + globalDualNumDofs;  // used for the maps
   const size_t nPrimalDofsPerNode = 3;
-  const GO globalPrimalNumNodes = globalPrimalNumDofs / nPrimalDofsPerNode;
-  const size_t nDualDofsPerNode = 1;
+  const GO globalPrimalNumNodes   = globalPrimalNumDofs / nPrimalDofsPerNode;
+  const size_t nDualDofsPerNode   = 1;
 
   std::map<GO, GO> lagr2Dof;
   std::map<LO, LO> myLagr2Dof;
@@ -107,11 +105,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Construct the blocked map in Thyra mode
   RCP<const tpetra_map_type> primalNodeMap = Tpetra::createUniformContigMapWithNode<LocalOrdinal, GlobalOrdinal, Node>(globalPrimalNumNodes, comm);
-  const GO indexBase = primalNodeMap->getIndexBase();
-  ArrayView<const GO> myPrimalNodes = primalNodeMap->getLocalElementList();
+  const GO indexBase                       = primalNodeMap->getIndexBase();
+  ArrayView<const GO> myPrimalNodes        = primalNodeMap->getLocalElementList();
 
   const size_t numMyPrimalNodes = primalNodeMap->getLocalNumElements();
-  const size_t numMyPrimalDofs = numMyPrimalNodes * nPrimalDofsPerNode;
+  const size_t numMyPrimalDofs  = numMyPrimalNodes * nPrimalDofsPerNode;
 
   Array<GO> myPrimalDofs(numMyPrimalDofs);
 
@@ -140,14 +138,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   current_i = 0;
   for (auto i = lagr2Dof.begin(); i != lagr2Dof.end(); ++i)
-    if (primalMap->isNodeGlobalElement(nPrimalDofsPerNode * (i->second)))
-    {
-      for (size_t j = 0; j < nDualDofsPerNode; ++j)
-      {
-        myDualDofs[nDualDofsPerNode * current_i + j] = (i->first) * nDualDofsPerNode + j;
+    if (primalMap->isNodeGlobalElement(nPrimalDofsPerNode * (i->second))) {
+      for (size_t j = 0; j < nDualDofsPerNode; ++j) {
+        myDualDofs[nDualDofsPerNode * current_i + j]               = (i->first) * nDualDofsPerNode + j;
         myDofs[numMyPrimalDofs + nDualDofsPerNode * current_i + j] = globalPrimalNumDofs + (i->first) * nDualDofsPerNode + j;
       }
-      GO primalDof = nPrimalDofsPerNode * (i->second);
+      GO primalDof          = nPrimalDofsPerNode * (i->second);
       myLagr2Dof[current_i] = primalMap->getLocalElement(primalDof) / nPrimalDofsPerNode;
       ++current_i;
     }
@@ -155,25 +151,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   RCP<const tpetra_map_type> dualMap = rcp(new tpetra_map_type(globalDualNumDofs, myDualDofs, indexBase, comm));
   RCP<const tpetra_map_type> fullMap = rcp(new tpetra_map_type(globalNumDofs, myDofs, indexBase, comm));
 
-  RCP<const Map> fullXMap = rcp(new TpetraMap(fullMap));
+  RCP<const Map> fullXMap   = rcp(new TpetraMap(fullMap));
   RCP<const Map> primalXMap = rcp(new TpetraMap(primalMap));
-  RCP<const Map> dualXMap = rcp(new TpetraMap(dualMap));
+  RCP<const Map> dualXMap   = rcp(new TpetraMap(dualMap));
 
   std::vector<RCP<const Map>> xsubmaps = {primalXMap, dualXMap};
-  RCP<BlockedMap> blockedMap = rcp(new BlockedMap(fullXMap, xsubmaps, true));
+  RCP<BlockedMap> blockedMap           = rcp(new BlockedMap(fullXMap, xsubmaps, true));
 
   // Read input matrices
   typedef Tpetra::MatrixMarket::Reader<SparseMatrixType> reader_type;
 
-  RCP<Matrix> xQ = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("Q_mm.txt", primalXMap, null, primalXMap, primalXMap);
-  RCP<Matrix> xG = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("G_mm.txt", primalXMap, null, dualXMap, primalXMap);
+  RCP<Matrix> xQ  = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("Q_mm.txt", primalXMap, null, primalXMap, primalXMap);
+  RCP<Matrix> xG  = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("G_mm.txt", primalXMap, null, dualXMap, primalXMap);
   RCP<Matrix> xGT = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("GT_mm.txt", dualXMap, null, primalXMap, dualXMap);
-  RCP<Matrix> xC = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("C_mm.txt", dualXMap, null, dualXMap, dualXMap);
+  RCP<Matrix> xC  = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("C_mm.txt", dualXMap, null, dualXMap, dualXMap);
 
-  RCP<CrsMatrixWrap> xwQ = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xQ);
-  RCP<CrsMatrixWrap> xwG = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xG);
+  RCP<CrsMatrixWrap> xwQ  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xQ);
+  RCP<CrsMatrixWrap> xwG  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xG);
   RCP<CrsMatrixWrap> xwGT = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xGT);
-  RCP<CrsMatrixWrap> xwC = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xC);
+  RCP<CrsMatrixWrap> xwC  = Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(xC);
 
   // Construct the blocked saddle-point matrix
   RCP<BlockedCrsMatrix> blockedMatrix = rcp(new BlockedCrsMatrix(blockedMap, blockedMap, 8));
@@ -187,8 +183,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Create the preconditioner
   std::string xmlFile = "simple_1dof.xml";
 
-  RCP<ParameterList> params = Teuchos::getParametersFromXmlFile(xmlFile);
-  ParameterList& userDataParams = params->sublist("user data");
+  RCP<ParameterList> params     = Teuchos::getParametersFromXmlFile(xmlFile);
+  ParameterList &userDataParams = params->sublist("user data");
   userDataParams.set<RCP<std::map<LO, LO>>>("DualNodeID2PrimalNodeID", rcpFromRef(myLagr2Dof));
 
   RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner(Teuchos::rcp_dynamic_cast<Matrix>(blockedMatrix, true), *params);
@@ -201,11 +197,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   typedef Belos::StatusTestGenResSubNorm<belos_scalar, MultiVector, OP> blockStatusTestClass;
   typedef Belos::StatusTestCombo<belos_scalar, MultiVector, OP> StatusTestComboClass;
 
-  typename ST::magnitudeType tol = 1e-4;
+  typename ST::magnitudeType tol  = 1e-4;
   typename ST::magnitudeType bTol = 1e-5;
 
   RCP<blockStatusTestClass> primalBlockStatusTest = rcp(new blockStatusTestClass(bTol, 0));
-  RCP<blockStatusTestClass> dualBlockStatusTest = rcp(new blockStatusTestClass(bTol, 1));
+  RCP<blockStatusTestClass> dualBlockStatusTest   = rcp(new blockStatusTestClass(bTol, 1));
 
   RCP<StatusTestComboClass> statusTestCombo = rcp(new StatusTestComboClass(StatusTestComboClass::SEQ));
   statusTestCombo->addStatusTest(primalBlockStatusTest);
@@ -221,13 +217,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   belosParams->set("Output Frequency", 1);
 
   typedef Belos::LinearProblem<belos_scalar, MultiVector, OP> BLinProb;
-  RCP<OP> belosOp = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(blockedMatrix));
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(blockedMatrix));
   RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(H));
 
-  RCP<tpetra_mvector_type> rhsMultiVector = reader_type::readDenseFile("f_mm.txt", comm, fullMap);
+  RCP<tpetra_mvector_type> rhsMultiVector      = reader_type::readDenseFile("f_mm.txt", comm, fullMap);
   RCP<tpetra_mvector_type> solutionMultiVector = rcp(new tpetra_mvector_type(fullMap, 1));
 
-  RCP<MultiVector> rhsXMultiVector = rcp(new TpetraMultiVector(rhsMultiVector));
+  RCP<MultiVector> rhsXMultiVector      = rcp(new TpetraMultiVector(rhsMultiVector));
   RCP<MultiVector> solutionXMultiVector = rcp(new TpetraMultiVector(solutionMultiVector));
 
   RCP<BLinProb> blinproblem = rcp(new BLinProb(belosOp, solutionXMultiVector, rhsXMultiVector));
@@ -252,5 +248,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/meshtying/MortarSurfaceCoupling_DofBased.cpp b/packages/muelu/test/meshtying/MortarSurfaceCoupling_DofBased.cpp
index baaf2b9b7030..5cb1c01ba73f 100644
--- a/packages/muelu/test/meshtying/MortarSurfaceCoupling_DofBased.cpp
+++ b/packages/muelu/test/meshtying/MortarSurfaceCoupling_DofBased.cpp
@@ -66,9 +66,8 @@
 #include <Xpetra_MapUtils.hpp>
 #include <Xpetra_MatrixUtils.hpp>
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[])
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::Array;
   using Teuchos::ArrayRCP;
@@ -76,20 +75,27 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   using Teuchos::RCP;
   using Teuchos::rcp;
 
-  using ST = Teuchos::ScalarTraits<Scalar>;
+  using ST  = Teuchos::ScalarTraits<Scalar>;
   using GST = Teuchos::ScalarTraits<GO>;
 
-  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<Teuchos::FancyOStream> out     = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
   out->setOutputToRootOnly(0);
 
-  GO numGlobalDofsPrimal = -GST::one(); clp.setOption("nPrimalDofs", &numGlobalDofsPrimal, "total number of primal DOFs");
-  GO numGlobalDofsDual = -GST::one(); clp.setOption("nDualDofs", &numGlobalDofsDual, "total number of dual DOFs");
-  int numPrimalDofsPerNode = -1; clp.setOption("numPrimalDofsPerNode", &numPrimalDofsPerNode, "number of primal DOFs per mesh node");
-  int numDualDofsPerNode = -1; clp.setOption("numDualDofsPerNode", &numDualDofsPerNode, "number of dual DOFs per interface node");
-  std::string probName = ""; clp.setOption("probName", &probName, "short name of the problem to be done. Used to read-in the problem from files.");
-  std::string xmlFile = ""; clp.setOption("xml", &xmlFile, "xml-file with MueLu configuration");
-  int expectedNumIterations = -1; clp.setOption("expectedNumIts", &expectedNumIterations, "expected number of iterations for this test");
+  GO numGlobalDofsPrimal = -GST::one();
+  clp.setOption("nPrimalDofs", &numGlobalDofsPrimal, "total number of primal DOFs");
+  GO numGlobalDofsDual = -GST::one();
+  clp.setOption("nDualDofs", &numGlobalDofsDual, "total number of dual DOFs");
+  int numPrimalDofsPerNode = -1;
+  clp.setOption("numPrimalDofsPerNode", &numPrimalDofsPerNode, "number of primal DOFs per mesh node");
+  int numDualDofsPerNode = -1;
+  clp.setOption("numDualDofsPerNode", &numDualDofsPerNode, "number of dual DOFs per interface node");
+  std::string probName = "";
+  clp.setOption("probName", &probName, "short name of the problem to be done. Used to read-in the problem from files.");
+  std::string xmlFile = "";
+  clp.setOption("xml", &xmlFile, "xml-file with MueLu configuration");
+  int expectedNumIterations = -1;
+  clp.setOption("expectedNumIts", &expectedNumIterations, "expected number of iterations for this test");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
@@ -99,22 +105,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
-  TEUCHOS_TEST_FOR_EXCEPTION(numGlobalDofsPrimal==-GST::one(), MueLu::Exceptions::InvalidArgument, "Please specify the global number of primal DOFs on the command line.");
-  TEUCHOS_TEST_FOR_EXCEPTION(numGlobalDofsDual==-GST::one(), MueLu::Exceptions::InvalidArgument, "Please specify the global number of dual DOFs on the command line.");
-  TEUCHOS_TEST_FOR_EXCEPTION(numPrimalDofsPerNode==-1, MueLu::Exceptions::InvalidArgument, "Please specify the number of primal DOFs per mesh node on the command line.");
-  TEUCHOS_TEST_FOR_EXCEPTION(numDualDofsPerNode==-1, MueLu::Exceptions::InvalidArgument, "Please specify the number of dual DOFs per interface node on the command line.");
-  TEUCHOS_TEST_FOR_EXCEPTION(probName=="", MueLu::Exceptions::InvalidArgument, "Please specify a valid problem name.");
-  TEUCHOS_TEST_FOR_EXCEPTION(xmlFile=="", MueLu::Exceptions::InvalidArgument, "Please specify a valid xml-file for the MueLu preconditioner.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numGlobalDofsPrimal == -GST::one(), MueLu::Exceptions::InvalidArgument, "Please specify the global number of primal DOFs on the command line.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numGlobalDofsDual == -GST::one(), MueLu::Exceptions::InvalidArgument, "Please specify the global number of dual DOFs on the command line.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numPrimalDofsPerNode == -1, MueLu::Exceptions::InvalidArgument, "Please specify the number of primal DOFs per mesh node on the command line.");
+  TEUCHOS_TEST_FOR_EXCEPTION(numDualDofsPerNode == -1, MueLu::Exceptions::InvalidArgument, "Please specify the number of dual DOFs per interface node on the command line.");
+  TEUCHOS_TEST_FOR_EXCEPTION(probName == "", MueLu::Exceptions::InvalidArgument, "Please specify a valid problem name.");
+  TEUCHOS_TEST_FOR_EXCEPTION(xmlFile == "", MueLu::Exceptions::InvalidArgument, "Please specify a valid xml-file for the MueLu preconditioner.");
 
-  const std::string matrixFileName = probName + "_matrix.mm";
-  const std::string rhsFileName = probName + "_rhs.mm";
-  const std::string nullspace1FileName = probName + "_nullspace1.mm";
+  const std::string matrixFileName           = probName + "_matrix.mm";
+  const std::string rhsFileName              = probName + "_rhs.mm";
+  const std::string nullspace1FileName       = probName + "_nullspace1.mm";
   const std::string dualInterfaceMapFileName = probName + "_interface_dof_map.mm";
 
   // Create maps for primal DOFs
   std::vector<size_t> stridingInfoPrimal;
   stridingInfoPrimal.push_back(numPrimalDofsPerNode);
-  RCP<const StridedMap> dofRowMapPrimal = StridedMapFactory::Build(lib, numGlobalDofsPrimal, Teuchos::ScalarTraits<GO>::zero(), stridingInfoPrimal,comm,-1);
+  RCP<const StridedMap> dofRowMapPrimal = StridedMapFactory::Build(lib, numGlobalDofsPrimal, Teuchos::ScalarTraits<GO>::zero(), stridingInfoPrimal, comm, -1);
 
   // Create maps for dual DOFs
   std::vector<size_t> stridingInfoDual;
@@ -122,62 +128,58 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   RCP<const StridedMap> dofRowMapDual = StridedMapFactory::Build(lib, numGlobalDofsDual, Teuchos::ScalarTraits<GO>::zero(), stridingInfoDual, comm, -1, numGlobalDofsPrimal);
 
   // Construct the blocked map of the global system
-  std::vector<RCP<const Map> > rowmaps;
+  std::vector<RCP<const Map>> rowmaps;
   rowmaps.push_back(dofRowMapPrimal);
   rowmaps.push_back(dofRowMapDual);
-  RCP<const Map> fullRowMap = MapUtils::concatenateMaps(rowmaps);
+  RCP<const Map> fullRowMap        = MapUtils::concatenateMaps(rowmaps);
   RCP<const BlockedMap> blockedMap = rcp(new BlockedMap(fullRowMap, rowmaps));
 
   // Read the matrix from file and transform it into a block matrix
-  RCP<Matrix> mat = Xpetra::IO<SC,LO,GO,NO>::Read(matrixFileName, fullRowMap);
+  RCP<Matrix> mat = Xpetra::IO<SC, LO, GO, NO>::Read(matrixFileName, fullRowMap);
   RCP<MapExtractor> rangeMapExtractor =
-      Xpetra::MapExtractorFactory<SC,LO,GO,NO>::Build(fullRowMap, rowmaps);
+      Xpetra::MapExtractorFactory<SC, LO, GO, NO>::Build(fullRowMap, rowmaps);
   RCP<BlockedCrsMatrix> blockedMatrix =
-      Xpetra::MatrixUtils<SC,LO,GO,NO>::SplitMatrix(*mat,rangeMapExtractor, rangeMapExtractor);
+      Xpetra::MatrixUtils<SC, LO, GO, NO>::SplitMatrix(*mat, rangeMapExtractor, rangeMapExtractor);
   blockedMatrix->fillComplete();
 
   // Read the right-hand side vector from file and transform it into a block vector
-  RCP<MultiVector> rhs = Xpetra::IO<SC,LO,GO,NO>::ReadMultiVector(rhsFileName, fullRowMap);
+  RCP<MultiVector> rhs               = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(rhsFileName, fullRowMap);
   RCP<BlockedMultiVector> blockedRhs = rcp(new BlockedMultiVector(blockedMap, rhs));
 
   // Read the nullspace vector of the (0,0)-block from file
-  RCP<MultiVector> nullspace1 = Xpetra::IO<SC,LO,GO,NO>::ReadMultiVector(nullspace1FileName, dofRowMapPrimal);
+  RCP<MultiVector> nullspace1 = Xpetra::IO<SC, LO, GO, NO>::ReadMultiVector(nullspace1FileName, dofRowMapPrimal);
 
   // Read the interface slave dof row map from file
   TEUCHOS_ASSERT(!dualInterfaceMapFileName.empty());
-  RCP<const Map> primalInterfaceDofMapDofs = Xpetra::IO<SC,LO,GO,NO>::ReadMap(dualInterfaceMapFileName, lib, comm);
+  RCP<const Map> primalInterfaceDofMapDofs = Xpetra::IO<SC, LO, GO, NO>::ReadMap(dualInterfaceMapFileName, lib, comm);
 
   // Distribute interface slave dof row map across all procs
   RCP<const Map> primalInterfaceDofMap = Teuchos::null;
   {
     std::vector<int> interfaceGlobalDofsOnCurrentProc = std::vector<int>(primalInterfaceDofMapDofs->getGlobalNumElements());
-    std::vector<int> interfaceGlobalDofs = std::vector<int>(primalInterfaceDofMapDofs->getGlobalNumElements());
-    if(comm->getRank() == 0)
-    {
-      for(size_t i = 0; i < primalInterfaceDofMapDofs->getGlobalNumElements(); ++i)
-      {
+    std::vector<int> interfaceGlobalDofs              = std::vector<int>(primalInterfaceDofMapDofs->getGlobalNumElements());
+    if (comm->getRank() == 0) {
+      for (size_t i = 0; i < primalInterfaceDofMapDofs->getGlobalNumElements(); ++i) {
         interfaceGlobalDofsOnCurrentProc[i] = primalInterfaceDofMapDofs->getLocalElementList()[i];
       }
     }
     Teuchos::reduceAll<int>(*comm, Teuchos::REDUCE_MAX, interfaceGlobalDofs.size(),
-    &interfaceGlobalDofsOnCurrentProc[0], &interfaceGlobalDofs[0]);
+                            &interfaceGlobalDofsOnCurrentProc[0], &interfaceGlobalDofs[0]);
 
     Array<GlobalOrdinal> primalInterfaceDofMapDofsOnCurProc;
-    for(size_t i = 0; i < interfaceGlobalDofs.size(); ++i)
-    {
-      if(dofRowMapPrimal->isNodeGlobalElement((GlobalOrdinal) interfaceGlobalDofs[i]))
-        primalInterfaceDofMapDofsOnCurProc.push_back((GlobalOrdinal) interfaceGlobalDofs[i]);
+    for (size_t i = 0; i < interfaceGlobalDofs.size(); ++i) {
+      if (dofRowMapPrimal->isNodeGlobalElement((GlobalOrdinal)interfaceGlobalDofs[i]))
+        primalInterfaceDofMapDofsOnCurProc.push_back((GlobalOrdinal)interfaceGlobalDofs[i]);
     }
     primalInterfaceDofMap = MapFactory::Build(lib, primalInterfaceDofMapDofs->getGlobalNumElements(), primalInterfaceDofMapDofsOnCurProc, Teuchos::ScalarTraits<GO>::zero(), comm);
   }
 
   // Create the default nullspace vector of the (1,1)-block
   RCP<MultiVector> nullspace2 = MultiVectorFactory::Build(dofRowMapDual, 2, true);
-  const int dimNS = 2;
-  for (int dim = 0; dim < dimNS; ++dim)
-  {
+  const int dimNS             = 2;
+  for (int dim = 0; dim < dimNS; ++dim) {
     ArrayRCP<Scalar> nsValues = nullspace2->getDataNonConst(dim);
-    const int numBlocks = nsValues.size() / dimNS;
+    const int numBlocks       = nsValues.size() / dimNS;
     for (int j = 0; j < numBlocks; ++j)
       nsValues[j * dimNS + dim] = Teuchos::ScalarTraits<Scalar>::one();
   }
@@ -200,11 +202,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   using blockStatusTestClass = Belos::StatusTestGenResSubNorm<SC, MultiVector, OP>;
   using StatusTestComboClass = Belos::StatusTestCombo<SC, MultiVector, OP>;
 
-  typename ST::magnitudeType tol = 1e-4;
+  typename ST::magnitudeType tol  = 1e-4;
   typename ST::magnitudeType bTol = 1e-5;
 
   RCP<blockStatusTestClass> primalBlockStatusTest = rcp(new blockStatusTestClass(bTol, 0));
-  RCP<blockStatusTestClass> dualBlockStatusTest = rcp(new blockStatusTestClass(bTol, 1));
+  RCP<blockStatusTestClass> dualBlockStatusTest   = rcp(new blockStatusTestClass(bTol, 1));
 
   RCP<StatusTestComboClass> statusTestCombo = rcp(new StatusTestComboClass(StatusTestComboClass::SEQ));
   statusTestCombo->addStatusTest(primalBlockStatusTest);
@@ -219,8 +221,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   belosParams->set("Output Style", 1);
   belosParams->set("Output Frequency", 1);
 
-  using BLinProb = Belos::LinearProblem<SC, MultiVector, OP>;
-  RCP<OP> belosOp = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(blockedMatrix));
+  using BLinProb    = Belos::LinearProblem<SC, MultiVector, OP>;
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(blockedMatrix));
   RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(hierarchy));
 
   RCP<MultiVector> X = MultiVectorFactory::Build(fullRowMap, 1, true);
@@ -234,7 +236,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   blinsolver->setUserConvStatusTest(statusTestCombo);
 
-  Belos::ReturnType ret = blinsolver->solve();
+  Belos::ReturnType ret   = blinsolver->solve();
   const int numIterations = blinsolver->getNumIters();
 
   if (ret == Belos::Converged && (expectedNumIterations == -1 || numIterations == expectedNumIterations))
@@ -243,11 +245,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     return EXIT_FAILURE;
 }
 
-
 //-----------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/mhdvanka/tomtest.cpp b/packages/muelu/test/mhdvanka/tomtest.cpp
index f0c38f78d404..8e8ff2656724 100644
--- a/packages/muelu/test/mhdvanka/tomtest.cpp
+++ b/packages/muelu/test/mhdvanka/tomtest.cpp
@@ -53,11 +53,11 @@
 int main(int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; //reference count pointers
+  using Teuchos::RCP;  // reference count pointers
 
-  typedef MueLu::GeoInterpFactory<SC,LO,GO,NO>          GeoInterpFactory;
-  typedef MueLu::Q2Q1Q2CoarseGridFactory<SC,LO,GO,NO>   Q2Q1Q2CoarseGridFactory;
-  typedef MueLu::MHDRAPFactory<SC,LO,GO,NO>             MHDRAPFactory;
+  typedef MueLu::GeoInterpFactory<SC, LO, GO, NO> GeoInterpFactory;
+  typedef MueLu::Q2Q1Q2CoarseGridFactory<SC, LO, GO, NO> Q2Q1Q2CoarseGridFactory;
+  typedef MueLu::MHDRAPFactory<SC, LO, GO, NO> MHDRAPFactory;
 
   //
   // MPI initialization using Teuchos
@@ -70,13 +70,12 @@ int main(int argc, char *argv[]) {
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // Initialize a "FancyOStream" to output to standard out (cout)
     RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
 
-
     // First, we start with an Xpetra::Map
     Xpetra::UnderlyingLib lib = Xpetra::UseTpetra;
 
@@ -90,15 +89,15 @@ int main(int argc, char *argv[]) {
     string fileNameA21 = "./Matrices/J_21_3.mm";
     string fileNameA22 = "./Matrices/J_22_3.mm";
 
-    RCP<Matrix> A00  = Utils::Read(fileNameA00,lib,comm);
-    RCP<Matrix> A01  = Utils::Read(fileNameA01,lib,comm);
-    RCP<Matrix> A02  = Utils::Read(fileNameA02,lib,comm);
-    RCP<Matrix> A10  = Utils::Read(fileNameA10,lib,comm);
-    RCP<Matrix> A11  = Utils::Read(fileNameA11,lib,comm);
-    RCP<Matrix> A12  = Utils::Read(fileNameA12,lib,comm);
-    RCP<Matrix> A20  = Utils::Read(fileNameA20,lib,comm);
-    RCP<Matrix> A21  = Utils::Read(fileNameA21,lib,comm);
-    RCP<Matrix> A22  = Utils::Read(fileNameA22,lib,comm);
+    RCP<Matrix> A00 = Utils::Read(fileNameA00, lib, comm);
+    RCP<Matrix> A01 = Utils::Read(fileNameA01, lib, comm);
+    RCP<Matrix> A02 = Utils::Read(fileNameA02, lib, comm);
+    RCP<Matrix> A10 = Utils::Read(fileNameA10, lib, comm);
+    RCP<Matrix> A11 = Utils::Read(fileNameA11, lib, comm);
+    RCP<Matrix> A12 = Utils::Read(fileNameA12, lib, comm);
+    RCP<Matrix> A20 = Utils::Read(fileNameA20, lib, comm);
+    RCP<Matrix> A21 = Utils::Read(fileNameA21, lib, comm);
+    RCP<Matrix> A22 = Utils::Read(fileNameA22, lib, comm);
 
     std::cout << A00->isFillComplete() << std::endl;
 
@@ -111,233 +110,212 @@ int main(int argc, char *argv[]) {
     size_t np = PMap->getGlobalNumElements();
     size_t nm = MMap->getGlobalNumElements();
 
-    RCP<const Map> AMap = Xpetra::MapFactory<LO,GO>::createUniformContigMap(Xpetra::UseTpetra,nv+np+nm,comm);
+    RCP<const Map> AMap = Xpetra::MapFactory<LO, GO>::createUniformContigMap(Xpetra::UseTpetra, nv + np + nm, comm);
 
     size_t maxEntriesPerRow = 84;
-    RCP<Matrix> A = rcp(new CrsMatrixWrap(AMap,maxEntriesPerRow));
+    RCP<Matrix> A           = rcp(new CrsMatrixWrap(AMap, maxEntriesPerRow));
 
     Teuchos::ArrayView<const LO> colPtr;
     Teuchos::ArrayView<const SC> valPtr;
 
-    A00->getLocalRowView(7,colPtr,valPtr);
+    A00->getLocalRowView(7, colPtr, valPtr);
 
     std::cout << "valPtr.size() = " << valPtr.size() << std::endl;
 
-    A01->getLocalRowView(7,colPtr,valPtr);
+    A01->getLocalRowView(7, colPtr, valPtr);
 
     std::cout << "valPtr.size() = " << valPtr.size() << std::endl;
 
-
-    //Loop over V rows
-    for (size_t VRow = 0; VRow < nv; VRow++)
-    {
+    // Loop over V rows
+    for (size_t VRow = 0; VRow < nv; VRow++) {
       Teuchos::ArrayView<const LO> colPtr;
       Teuchos::ArrayView<const SC> valPtr;
 
-      A00->getLocalRowView(VRow,colPtr,valPtr);
+      A00->getLocalRowView(VRow, colPtr, valPtr);
 
-      //Can be directly inserted!
-      A->insertGlobalValues(VRow,colPtr,valPtr);
+      // Can be directly inserted!
+      A->insertGlobalValues(VRow, colPtr, valPtr);
 
-      //Now do pressure column:
-      A01->getLocalRowView(VRow,colPtr,valPtr);
+      // Now do pressure column:
+      A01->getLocalRowView(VRow, colPtr, valPtr);
 
-      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(),nv);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(), nv);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(VRow,newColPtr.view(0,colPtr.size()),valPtr);
+      // Insert into A
+      A->insertGlobalValues(VRow, newColPtr.view(0, colPtr.size()), valPtr);
 
-      //Now do magnetics column:
-      A02->getLocalRowView(VRow,colPtr,valPtr);
+      // Now do magnetics column:
+      A02->getLocalRowView(VRow, colPtr, valPtr);
 
       newColPtr.clear();
-      newColPtr.resize(colPtr.size(),nv+np);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      newColPtr.resize(colPtr.size(), nv + np);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(VRow,newColPtr.view(0,colPtr.size()),valPtr);
-
+      // Insert into A
+      A->insertGlobalValues(VRow, newColPtr.view(0, colPtr.size()), valPtr);
     }
 
-    //Loop over P rows
-    for (size_t PRow = 0; PRow < np; PRow++)
-    {
+    // Loop over P rows
+    for (size_t PRow = 0; PRow < np; PRow++) {
       Teuchos::ArrayView<const LO> colPtr;
       Teuchos::ArrayView<const SC> valPtr;
 
-      A10->getLocalRowView(PRow,colPtr,valPtr);
+      A10->getLocalRowView(PRow, colPtr, valPtr);
 
-      //Can be directly inserted!
-      A->insertGlobalValues(PRow+nv,colPtr,valPtr);
+      // Can be directly inserted!
+      A->insertGlobalValues(PRow + nv, colPtr, valPtr);
 
-      //Now do pressure column:
-      A11->getLocalRowView(PRow,colPtr,valPtr);
+      // Now do pressure column:
+      A11->getLocalRowView(PRow, colPtr, valPtr);
 
-      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(),nv);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(), nv);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(PRow+nv,newColPtr.view(0,colPtr.size()),valPtr);
+      // Insert into A
+      A->insertGlobalValues(PRow + nv, newColPtr.view(0, colPtr.size()), valPtr);
 
-      //Now do magnetics column:
-      A12->getLocalRowView(PRow,colPtr,valPtr);
+      // Now do magnetics column:
+      A12->getLocalRowView(PRow, colPtr, valPtr);
 
       newColPtr.clear();
-      newColPtr.resize(colPtr.size(),nv+np);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      newColPtr.resize(colPtr.size(), nv + np);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(PRow+nv,newColPtr.view(0,colPtr.size()),valPtr);
-
+      // Insert into A
+      A->insertGlobalValues(PRow + nv, newColPtr.view(0, colPtr.size()), valPtr);
     }
 
-    //Loop over M rows
-    for (size_t MRow = 0; MRow < nm; MRow++)
-    {
+    // Loop over M rows
+    for (size_t MRow = 0; MRow < nm; MRow++) {
       Teuchos::ArrayView<const LO> colPtr;
       Teuchos::ArrayView<const SC> valPtr;
 
-      A20->getLocalRowView(MRow,colPtr,valPtr);
+      A20->getLocalRowView(MRow, colPtr, valPtr);
 
-      //Can be directly inserted!
-      A->insertGlobalValues(MRow+nv+np,colPtr,valPtr);
+      // Can be directly inserted!
+      A->insertGlobalValues(MRow + nv + np, colPtr, valPtr);
 
-      //Now do pressure column:
-      A21->getLocalRowView(MRow,colPtr,valPtr);
+      // Now do pressure column:
+      A21->getLocalRowView(MRow, colPtr, valPtr);
 
-      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(),nv);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      Teuchos::ArrayRCP<LO> newColPtr(colPtr.size(), nv);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(MRow+nv+np,newColPtr.view(0,colPtr.size()),valPtr);
+      // Insert into A
+      A->insertGlobalValues(MRow + nv + np, newColPtr.view(0, colPtr.size()), valPtr);
 
-      //Now do magnetics column:
-      A22->getLocalRowView(MRow,colPtr,valPtr);
+      // Now do magnetics column:
+      A22->getLocalRowView(MRow, colPtr, valPtr);
 
       newColPtr.clear();
-      newColPtr.resize(colPtr.size(),nv+np);
-      for (LO jj=0; jj<colPtr.size(); jj++)
-      {
+      newColPtr.resize(colPtr.size(), nv + np);
+      for (LO jj = 0; jj < colPtr.size(); jj++) {
         newColPtr[jj] += colPtr[jj];
       }
 
-      //Insert into A
-      A->insertGlobalValues(MRow+nv+np,newColPtr.view(0,colPtr.size()),valPtr);
-
+      // Insert into A
+      A->insertGlobalValues(MRow + nv + np, newColPtr.view(0, colPtr.size()), valPtr);
     }
 
     A->fillComplete();
 
-
     // Let's read in the element connectivity info:
-    GO totalFineElements = 32*32;
+    GO totalFineElements = 32 * 32;
 
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > fineGridVElements = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalFineElements,18));
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > fineGridPElements = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalFineElements,4));
-    RCP<Teuchos::SerialDenseMatrix<GO,GO> > fineGridMElements = rcp(new Teuchos::SerialDenseMatrix<GO,GO>(totalFineElements,9));
+    RCP<Teuchos::SerialDenseMatrix<GO, GO> > fineGridVElements = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalFineElements, 18));
+    RCP<Teuchos::SerialDenseMatrix<GO, GO> > fineGridPElements = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalFineElements, 4));
+    RCP<Teuchos::SerialDenseMatrix<GO, GO> > fineGridMElements = rcp(new Teuchos::SerialDenseMatrix<GO, GO>(totalFineElements, 9));
 
     std::ifstream VElementFile("./Matrices/elements_0_0");
     std::ifstream PElementFile("./Matrices/elements_1_1");
     std::ifstream MElementFile("./Matrices/elements_2_2");
 
-    for (GO ii=0;ii<totalFineElements;ii++)
-    {
-      for (LO jj=0;jj<9;jj++)
-      {
-        VElementFile >> (*fineGridVElements)(ii,2*jj);
-        VElementFile >> (*fineGridVElements)(ii,2*jj+1);
+    for (GO ii = 0; ii < totalFineElements; ii++) {
+      for (LO jj = 0; jj < 9; jj++) {
+        VElementFile >> (*fineGridVElements)(ii, 2 * jj);
+        VElementFile >> (*fineGridVElements)(ii, 2 * jj + 1);
 
-        MElementFile >> (*fineGridMElements)(ii,jj);
+        MElementFile >> (*fineGridMElements)(ii, jj);
       }
 
-      for (LO kk=0;kk<4;kk++)
-      {
-        PElementFile >> (*fineGridPElements)(ii,kk);
+      for (LO kk = 0; kk < 4; kk++) {
+        PElementFile >> (*fineGridPElements)(ii, kk);
       }
-
     }
     VElementFile.close();
     PElementFile.close();
     MElementFile.close();
 
-
     Hierarchy H;
     H.setDefaultVerbLevel(Teuchos::VERB_NONE);
 
     RCP<Level> finest = H.GetLevel();
     finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
 
-    finest->Set("A"  ,A  );
-    finest->Set("A00",A00);
-    finest->Set("A01",A01);
-    finest->Set("A02",A02);
-    finest->Set("A10",A10);
-    finest->Set("A11",A11);
-    finest->Set("A12",A12);
-    finest->Set("A20",A20);
-    finest->Set("A21",A21);
-    finest->Set("A22",A22);
-
-    //Set finegrid elements
-    finest->Set("VElementList",fineGridVElements);
-    finest->Set("PElementList",fineGridPElements);
-    finest->Set("MElementList",fineGridMElements);
-
-    //Set count stuff for Vanka Smoother
-    finest->Set("NV",A00->getGlobalNumRows());
-    finest->Set("NP",A10->getGlobalNumRows());
-    finest->Set("NM",A20->getGlobalNumRows());
+    finest->Set("A", A);
+    finest->Set("A00", A00);
+    finest->Set("A01", A01);
+    finest->Set("A02", A02);
+    finest->Set("A10", A10);
+    finest->Set("A11", A11);
+    finest->Set("A12", A12);
+    finest->Set("A20", A20);
+    finest->Set("A21", A21);
+    finest->Set("A22", A22);
+
+    // Set finegrid elements
+    finest->Set("VElementList", fineGridVElements);
+    finest->Set("PElementList", fineGridPElements);
+    finest->Set("MElementList", fineGridMElements);
+
+    // Set count stuff for Vanka Smoother
+    finest->Set("NV", A00->getGlobalNumRows());
+    finest->Set("NP", A10->getGlobalNumRows());
+    finest->Set("NM", A20->getGlobalNumRows());
 
     // Create a GeoInterpFactory
-    RCP<GeoInterpFactory> geoInterp = rcp( new GeoInterpFactory() );
-    RCP<Q2Q1Q2CoarseGridFactory> coarseElementFact = rcp( new Q2Q1Q2CoarseGridFactory() );
-    RCP<MueLu::MHDRAPFactory<SC,LO,GO,NO> > rapFact = rcp( new MueLu::MHDRAPFactory<SC,LO,GO,NO>() );
-
-
-    RCP<FactoryManager> M = rcp(new FactoryManager() );
-    M->SetFactory("A",rapFact);
-    M->SetFactory("A00",rapFact);
-    M->SetFactory("A01",rapFact);
-    M->SetFactory("A02",rapFact);
-    M->SetFactory("A10",rapFact);
-    M->SetFactory("A11",rapFact);
-    M->SetFactory("A12",rapFact);
-    M->SetFactory("A20",rapFact);
-    M->SetFactory("A21",rapFact);
-    M->SetFactory("A22",rapFact);
-    M->SetFactory("VElementList",coarseElementFact);
-    M->SetFactory("PElementList",coarseElementFact);
-    M->SetFactory("MElementList",coarseElementFact);
-    M->SetFactory("PV",geoInterp);
-    M->SetFactory("PP",geoInterp);
-    M->SetFactory("PM",geoInterp);
-    M->SetFactory("P",geoInterp);
-
-    H.Setup(*M,0,3);
-
-    //Utils::Write("./output/BigAMat.mm",*A);
+    RCP<GeoInterpFactory> geoInterp                    = rcp(new GeoInterpFactory());
+    RCP<Q2Q1Q2CoarseGridFactory> coarseElementFact     = rcp(new Q2Q1Q2CoarseGridFactory());
+    RCP<MueLu::MHDRAPFactory<SC, LO, GO, NO> > rapFact = rcp(new MueLu::MHDRAPFactory<SC, LO, GO, NO>());
+
+    RCP<FactoryManager> M = rcp(new FactoryManager());
+    M->SetFactory("A", rapFact);
+    M->SetFactory("A00", rapFact);
+    M->SetFactory("A01", rapFact);
+    M->SetFactory("A02", rapFact);
+    M->SetFactory("A10", rapFact);
+    M->SetFactory("A11", rapFact);
+    M->SetFactory("A12", rapFact);
+    M->SetFactory("A20", rapFact);
+    M->SetFactory("A21", rapFact);
+    M->SetFactory("A22", rapFact);
+    M->SetFactory("VElementList", coarseElementFact);
+    M->SetFactory("PElementList", coarseElementFact);
+    M->SetFactory("MElementList", coarseElementFact);
+    M->SetFactory("PV", geoInterp);
+    M->SetFactory("PP", geoInterp);
+    M->SetFactory("PM", geoInterp);
+    M->SetFactory("P", geoInterp);
+
+    H.Setup(*M, 0, 3);
+
+    // Utils::Write("./output/BigAMat.mm",*A);
     std::cout << "Hello world!\n";
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
-
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazin.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazin.cpp
index 1babad1466a5..046af096db62 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazin.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazin.cpp
@@ -124,8 +124,7 @@
  *  using block matrices
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -141,10 +140,10 @@ int main(int argc, char *argv[]) {
   using namespace Teuchos;
 
   oblackholestream blackhole;
-  GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  GlobalMPISession mpiSession(&argc, &argv, &blackhole);
   //
   RCP<const Comm<int> > comm = DefaultComm<int>::getComm();
-  RCP<FancyOStream> out = fancyOStream(rcpFromRef(std::cout));
+  RCP<FancyOStream> out      = fancyOStream(rcpFromRef(std::cout));
   out->setOutputToRootOnly(0);
   *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -155,10 +154,10 @@ int main(int argc, char *argv[]) {
   // custom parameters
   LocalOrdinal maxLevels = 3;
 
-  GlobalOrdinal maxCoarseSize=1; //FIXME clp doesn't like long long int
+  GlobalOrdinal maxCoarseSize = 1;  // FIXME clp doesn't like long long int
 
   int globalNumDofs = 1500;  // used for the maps
-  int nDofsPerNode = 3;      // used for generating the fine level null-space
+  int nDofsPerNode  = 3;     // used for generating the fine level null-space
 
   // build strided maps
   // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -171,34 +170,33 @@ int main(int argc, char *argv[]) {
   // xstridedfullmap: full map (velocity and pressure dof gids), continous
   // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
   // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-  Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-  RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-  RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-  RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+  Xpetra::UnderlyingLib lib       = Xpetra::UseEpetra;
+  RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+  RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+  RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
   /////////////////////////////////////// transform Xpetra::Map objects to Epetra
   // this is needed for AztecOO
   const RCP<const Epetra_Map> fullmap = rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-  RCP<const Epetra_Map>       velmap  = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-  RCP<const Epetra_Map>       premap  = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+  RCP<const Epetra_Map> velmap        = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+  RCP<const Epetra_Map> premap        = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
   /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
   // read in problem
-  Epetra_CrsMatrix * ptrA = 0;
-  Epetra_Vector * ptrf = 0;
+  Epetra_CrsMatrix* ptrA    = 0;
+  Epetra_Vector* ptrf       = 0;
   Epetra_MultiVector* ptrNS = 0;
 
   *out << "Reading matrix market file" << std::endl;
 
-  EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-  EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
+  EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+  EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
 
-  RCP<Epetra_CrsMatrix> epA = rcp(ptrA);
-  RCP<Epetra_Vector> epv = rcp(ptrf);
+  RCP<Epetra_CrsMatrix> epA    = rcp(ptrA);
+  RCP<Epetra_Vector> epv       = rcp(ptrf);
   RCP<Epetra_MultiVector> epNS = rcp(ptrNS);
 
-
   /////////////////////////////////////// split system into 2x2 block system
 
   *out << "Split matrix into 2x2 block matrix" << std::endl;
@@ -209,70 +207,70 @@ int main(int argc, char *argv[]) {
   RCP<Epetra_CrsMatrix> A21;
   RCP<Epetra_CrsMatrix> A22;
 
-  if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-    *out << "Problem with splitting matrix"<< std::endl;
+  if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+    *out << "Problem with splitting matrix" << std::endl;
 
   /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
   // build Xpetra objects from Epetra_CrsMatrix objects
-  RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A11));
-  RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A12));
-  RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A21));
-  RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A22));
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A11));
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A12));
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A21));
+  RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A22));
 
   /////////////////////////////////////// generate MapExtractor object
 
-  std::vector<RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > > xmaps;
+  std::vector<RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > > xmaps;
   xmaps.push_back(xstridedvelmap);
   xmaps.push_back(xstridedpremap);
 
-  RCP<const Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(xstridedfullmap,xmaps);
+  RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(xstridedfullmap, xmaps);
 
   /////////////////////////////////////// build blocked transfer operator
   // using the map extractor
-  RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(map_extractor,map_extractor,10));
-  bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-  bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-  bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-  bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+  RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map_extractor, map_extractor, 10));
+  bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+  bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+  bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+  bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
   bOp->fillComplete();
 
   //////////////////////////////////////////////////// create Hierarchy
-  RCP<Hierarchy> H = rcp ( new Hierarchy() );
+  RCP<Hierarchy> H = rcp(new Hierarchy());
   H->setDefaultVerbLevel(VERB_HIGH);
-  //H->setDefaultVerbLevel(VERB_NONE);
+  // H->setDefaultVerbLevel(VERB_NONE);
   H->SetMaxCoarseSize(maxCoarseSize);
 
   //////////////////////////////////////////////////////// finest Level
   RCP<MueLu::Level> Finest = H->GetLevel();
   Finest->setDefaultVerbLevel(VERB_HIGH);
-  Finest->Set("A",rcp_dynamic_cast<Matrix>(bOp));
+  Finest->Set("A", rcp_dynamic_cast<Matrix>(bOp));
 
   /////////////////////////////////////////////// define subblocks of A
   // make A11 block and A22 block available as variable "A" generated
   // by A11Fact and A22Fact
   RCP<SubBlockAFactory> A11Fact = rcp(new SubBlockAFactory());
-  A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-  A11Fact->SetParameter("block row",ParameterEntry(0));
-  A11Fact->SetParameter("block col",ParameterEntry(0));
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", ParameterEntry(0));
+  A11Fact->SetParameter("block col", ParameterEntry(0));
   RCP<SubBlockAFactory> A22Fact = rcp(new SubBlockAFactory());
-  A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-  A22Fact->SetParameter("block row",ParameterEntry(1));
-  A22Fact->SetParameter("block col",ParameterEntry(1));
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", ParameterEntry(1));
+  A22Fact->SetParameter("block col", ParameterEntry(1));
 
   ////////////////////////////////////////// prepare null space for A11
   RCP<MultiVector> nullspace11 = MultiVectorFactory::Build(xstridedvelmap, 2);  // this is a 2D standard null space
 
-  for (int i=0; i<nDofsPerNode-1; ++i) {
+  for (int i = 0; i < nDofsPerNode - 1; ++i) {
     ArrayRCP<Scalar> nsValues = nullspace11->getDataNonConst(i);
-    int numBlocks = nsValues.size() / (nDofsPerNode - 1);
-    for (int j=0; j< numBlocks; ++j) {
-      nsValues[j*(nDofsPerNode - 1) + i] = 1.0;
+    int numBlocks             = nsValues.size() / (nDofsPerNode - 1);
+    for (int j = 0; j < numBlocks; ++j) {
+      nsValues[j * (nDofsPerNode - 1) + i] = 1.0;
     }
   }
 
-  Finest->Set("Nullspace1",nullspace11);
+  Finest->Set("Nullspace1", nullspace11);
 
   ///////////////////////////////////////// define CoalesceDropFactory and Aggregation for A11
   // set up amalgamation for A11. Note: we're using a default null space factory (null)
@@ -320,7 +318,7 @@ int main(int argc, char *argv[]) {
   RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
 
   RCP<NullspaceFactory> nspFact11 = rcp(new NullspaceFactory("Nullspace1"));
-  nspFact11->SetFactory("Nullspace1",P11Fact);
+  nspFact11->SetFactory("Nullspace1", P11Fact);
 
   RCP<CoarseMapFactory> coarseMapFact11 = rcp(new CoarseMapFactory());
   coarseMapFact11->setStridingData(stridingInfo);
@@ -338,16 +336,16 @@ int main(int argc, char *argv[]) {
   M11->SetFactory("CoarseMap", coarseMapFact11);
   M11->SetFactory("Graph", dropFact11);
 #endif
-  M11->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+  M11->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
   ////////////////////////////////////////// prepare null space for A22
   RCP<MultiVector> nullspace22 = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
-  ArrayRCP<Scalar> nsValues22 = nullspace22->getDataNonConst(0);
-  for (int j=0; j< nsValues22.size(); ++j) {
+  ArrayRCP<Scalar> nsValues22  = nullspace22->getDataNonConst(0);
+  for (int j = 0; j < nsValues22.size(); ++j) {
     nsValues22[j] = 1.0;
   }
 
-  Finest->Set("Nullspace2",nullspace22);
+  Finest->Set("Nullspace2", nullspace22);
 
   ///////////////////////////////////////// define transfer ops for A22
 #if 0
@@ -379,7 +377,7 @@ int main(int argc, char *argv[]) {
 #else
   // use TentativePFactory
   RCP<AmalgamationFactory> amalgFact22 = rcp(new AmalgamationFactory());
-  RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory()); // check me (fed with A22) wrong column GIDS!!!
+  RCP<TentativePFactory> P22Fact       = rcp(new TentativePFactory());  // check me (fed with A22) wrong column GIDS!!!
 
   RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
 
@@ -396,14 +394,14 @@ int main(int argc, char *argv[]) {
   M22->SetFactory("R", R22Fact);
   M22->SetFactory("Aggregates", CoupledAggFact11);
   M22->SetFactory("Nullspace", nspFact22);
-  M22->SetFactory("UnAmalgamationInfo", amalgFact22); // TODO oops what about that? it was M11 before?
+  M22->SetFactory("UnAmalgamationInfo", amalgFact22);  // TODO oops what about that? it was M11 before?
   M22->SetFactory("Ptent", P22Fact);
   M22->SetFactory("CoarseMap", coarseMapFact22);
-  M22->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+  M22->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 #endif
 
   /////////////////////////////////////////// define blocked transfer ops
-  RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory()); // use row map index base from bOp
+  RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());  // use row map index base from bOp
   PFact->AddFactoryManager(M11);
   PFact->AddFactoryManager(M22);
 
@@ -419,56 +417,56 @@ int main(int argc, char *argv[]) {
   //////////////////////////////////////////////////////////////////////
   // Smoothers
 
-  //Another factory manager for braes sarazin smoother
-  //Schur Complement Factory, using the factory to generate AcFact
-  SC omega = 1.3;
+  // Another factory manager for braes sarazin smoother
+  // Schur Complement Factory, using the factory to generate AcFact
+  SC omega                          = 1.3;
   RCP<SchurComplementFactory> SFact = rcp(new SchurComplementFactory());
   SFact->SetParameter("omega", ParameterEntry(omega));
   SFact->SetFactory("A", MueLu::NoFactory::getRCP());
 
-  //Smoother Factory, using SFact as a factory for A
+  // Smoother Factory, using SFact as a factory for A
   std::string ifpackSCType;
   ParameterList ifpackSCList;
-  ifpackSCList.set("relaxation: sweeps", (LocalOrdinal) 3);
-  ifpackSCList.set("relaxation: damping factor", (Scalar) 1.0);
+  ifpackSCList.set("relaxation: sweeps", (LocalOrdinal)3);
+  ifpackSCList.set("relaxation: damping factor", (Scalar)1.0);
   ifpackSCType = "RELAXATION";
   ifpackSCList.set("relaxation: type", "Gauss-Seidel");
-  RCP<SmootherPrototype> smoProtoSC     = rcp( new TrilinosSmoother(ifpackSCType, ifpackSCList, 0) );
+  RCP<SmootherPrototype> smoProtoSC = rcp(new TrilinosSmoother(ifpackSCType, ifpackSCList, 0));
   smoProtoSC->SetFactory("A", SFact);
-  RCP<SmootherFactory> SmooSCFact = rcp( new SmootherFactory(smoProtoSC) );
+  RCP<SmootherFactory> SmooSCFact = rcp(new SmootherFactory(smoProtoSC));
 
-  RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
+  RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
   smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(3));
   smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(omega));
 
-  RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+  RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-  RCP<BraessSarazinSmoother> coarseSolverPrototype = rcp( new BraessSarazinSmoother() );
+  RCP<BraessSarazinSmoother> coarseSolverPrototype = rcp(new BraessSarazinSmoother());
   coarseSolverPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(3));
   coarseSolverPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(omega));
 
-  RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, null) );
+  RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, null));
 
   RCP<FactoryManager> MB = rcp(new FactoryManager());
-  MB->SetFactory("A",     SFact);
-  MB->SetFactory("Smoother",    SmooSCFact);
-  MB->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-  smootherPrototype->AddFactoryManager(MB,0);
-  coarseSolverPrototype->AddFactoryManager(MB,0);
+  MB->SetFactory("A", SFact);
+  MB->SetFactory("Smoother", SmooSCFact);
+  MB->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+  smootherPrototype->AddFactoryManager(MB, 0);
+  coarseSolverPrototype->AddFactoryManager(MB, 0);
 
   // main factory manager
   FactoryManager M;
-  M.SetFactory("A",            AcFact);
-  M.SetFactory("P",            PFact);
-  M.SetFactory("R",            RFact);
-  M.SetFactory("Smoother",     smootherFact); // TODO fix me
-  M.SetFactory("PreSmoother",     smootherFact); // TODO fix me
-  M.SetFactory("PostSmoother",     smootherFact); // TODO fix me
+  M.SetFactory("A", AcFact);
+  M.SetFactory("P", PFact);
+  M.SetFactory("R", RFact);
+  M.SetFactory("Smoother", smootherFact);      // TODO fix me
+  M.SetFactory("PreSmoother", smootherFact);   // TODO fix me
+  M.SetFactory("PostSmoother", smootherFact);  // TODO fix me
   M.SetFactory("CoarseSolver", coarseSolverFact);
 
   //////////////////////////////////// setup multigrid
 
-  H->Setup(M,0,maxLevels);
+  H->Setup(M, 0, maxLevels);
 
   *out << std::endl;
   *out << "print content of multigrid levels:" << std::endl;
@@ -481,7 +479,7 @@ int main(int argc, char *argv[]) {
   RCP<Level> coarseLevel2 = H->GetLevel(2);
   coarseLevel2->print(*out);
 
-  RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap,1);
+  RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap, 1);
 
   // Use AMG directly as an iterative method
 #if 0
@@ -533,5 +531,5 @@ int main(int argc, char *argv[]) {
 #else
   std::cout << "Epetra (and/or EpetraExt) not enabled. Skip test." << std::endl;
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
 }
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazinSmoother.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazinSmoother.cpp
index 541afcdd23e2..ccc451abcb13 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazinSmoother.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_BraessSarazinSmoother.cpp
@@ -119,7 +119,7 @@
  *  using block matrices
  */
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -135,13 +135,13 @@ int main(int argc, char *argv[]) {
   using namespace Teuchos;
 
   oblackholestream blackhole;
-  GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
   try {
     RCP<const Comm<int> > comm = DefaultComm<int>::getComm();
-    RCP<FancyOStream> out = fancyOStream(rcpFromRef(std::cout));
+    RCP<FancyOStream> out      = fancyOStream(rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -152,29 +152,29 @@ int main(int argc, char *argv[]) {
     // read in input parameters
 
     // default parameters
-    LO BS_nSweeps = 100;
-    Scalar BS_omega = 1.7;
-    LO SC_nSweeps = 1;
-    Scalar SC_omega = 1.0;
+    LO BS_nSweeps           = 100;
+    Scalar BS_omega         = 1.7;
+    LO SC_nSweeps           = 1;
+    Scalar SC_omega         = 1.0;
     int SC_bUseDirectSolver = 0;
 
     // Note: use --help to list available options.
     CommandLineProcessor clp(false);
-    clp.setOption("BraessSarazin_sweeps",&BS_nSweeps,"number of sweeps with BraessSarazin smoother");
-    clp.setOption("BraessSarazin_omega", &BS_omega,  "scaling factor for BraessSarazin smoother");
-    clp.setOption("SchurComp_sweeps",    &SC_nSweeps,"number of sweeps for BraessSarazin internal SchurComp solver/smoother (GaussSeidel)");
-    clp.setOption("SchurComp_omega",     &SC_omega,  "damping parameter for BraessSarazin internal SchurComp solver/smoother (GaussSeidel)");
-    clp.setOption("SchurComp_solver",    &SC_bUseDirectSolver,  "if 1: use direct solver for SchurComp equation, otherwise use GaussSeidel smoother (=default)");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    clp.setOption("BraessSarazin_sweeps", &BS_nSweeps, "number of sweeps with BraessSarazin smoother");
+    clp.setOption("BraessSarazin_omega", &BS_omega, "scaling factor for BraessSarazin smoother");
+    clp.setOption("SchurComp_sweeps", &SC_nSweeps, "number of sweeps for BraessSarazin internal SchurComp solver/smoother (GaussSeidel)");
+    clp.setOption("SchurComp_omega", &SC_omega, "damping parameter for BraessSarazin internal SchurComp solver/smoother (GaussSeidel)");
+    clp.setOption("SchurComp_solver", &SC_bUseDirectSolver, "if 1: use direct solver for SchurComp equation, otherwise use GaussSeidel smoother (=default)");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     int globalNumDofs = 1500;  // used for the maps
-    //int nDofsPerNode = 3;      // used for generating the fine level null-space
+    // int nDofsPerNode = 3;      // used for generating the fine level null-space
 
     // build strided maps
     // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -187,34 +187,33 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib       = Xpetra::UseEpetra;
+    RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for our splitting routine
     const RCP<const Epetra_Map> fullmap = rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
 
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
 
-    RCP<Epetra_CrsMatrix> epA = rcp(ptrA);
-    RCP<Epetra_Vector> epv = rcp(ptrf);
+    RCP<Epetra_CrsMatrix> epA    = rcp(ptrA);
+    RCP<Epetra_Vector> epv       = rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = rcp(ptrNS);
 
-
     /////////////////////////////////////// split system into 2x2 block system
 
     *out << "Split matrix into 2x2 block matrix" << std::endl;
@@ -225,39 +224,39 @@ int main(int argc, char *argv[]) {
     RCP<Epetra_CrsMatrix> A21;
     RCP<Epetra_CrsMatrix> A22;
 
-    if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A11));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A12));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A21));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A22));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A11));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A12));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A21));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<RCP<const Xpetra::Map<LO,GO,Node> > > xmaps;
+    std::vector<RCP<const Xpetra::Map<LO, GO, Node> > > xmaps;
 
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    RCP<const Xpetra::MapExtractor<Scalar,LO,GO,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(xstridedfullmap,xmaps);
+    RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(xstridedfullmap, xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
     //////////////////////////////////////////////////////// finest Level
     RCP<MueLu::Level> Finest = rcp(new Level());
     Finest->setDefaultVerbLevel(VERB_NONE);
-    Finest->Set("A",rcp_dynamic_cast<Matrix>(bOp));
+    Finest->Set("A", rcp_dynamic_cast<Matrix>(bOp));
 
     ///////////////////////////////////
     // Test Braess Sarazin Smoother as a solver
@@ -271,11 +270,11 @@ int main(int argc, char *argv[]) {
 
     // define BraessSarazin Smoother with BS_nSweeps and BS_omega as scaling factor
     // AFact_ = null (= default) for the 2x2 blocked operator
-    RCP<BraessSarazinSmoother> BraessSarazinSm = rcp( new BraessSarazinSmoother() );
+    RCP<BraessSarazinSmoother> BraessSarazinSm = rcp(new BraessSarazinSmoother());
     BraessSarazinSm->SetParameter("Sweeps", Teuchos::ParameterEntry(BS_nSweeps));
     BraessSarazinSm->SetParameter("Damping factor", Teuchos::ParameterEntry(BS_omega));
 
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(BraessSarazinSm) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(BraessSarazinSm));
 
     /*note that omega must be the same in the SchurComplementFactory and in the BraessSarazinSmoother*/
     // define SchurComplement Factory
@@ -285,55 +284,55 @@ int main(int argc, char *argv[]) {
     // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
     RCP<SchurComplementFactory> SFact = rcp(new SchurComplementFactory());
     SFact->SetParameter("omega", ParameterEntry(BS_omega));
-    SFact->SetFactory("A",MueLu::NoFactory::getRCP());
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());
 
     // define smoother/solver for BraessSarazin
     RCP<SmootherPrototype> smoProtoSC = null;
-    if(SC_bUseDirectSolver != 1) {
-      //Smoother Factory, using SFact as a factory for A
+    if (SC_bUseDirectSolver != 1) {
+      // Smoother Factory, using SFact as a factory for A
       std::string ifpackSCType;
       ParameterList ifpackSCList;
-      ifpackSCList.set("relaxation: sweeps", SC_nSweeps );
-      ifpackSCList.set("relaxation: damping factor", SC_omega );
+      ifpackSCList.set("relaxation: sweeps", SC_nSweeps);
+      ifpackSCList.set("relaxation: damping factor", SC_omega);
       ifpackSCType = "RELAXATION";
       ifpackSCList.set("relaxation: type", "Gauss-Seidel");
-      smoProtoSC     = rcp( new TrilinosSmoother(ifpackSCType, ifpackSCList, 0) );
+      smoProtoSC = rcp(new TrilinosSmoother(ifpackSCType, ifpackSCList, 0));
       smoProtoSC->SetFactory("A", SFact);
-    }
-    else {
+    } else {
       ParameterList ifpackDSList;
       std::string ifpackDSType;
-      smoProtoSC     = rcp( new DirectSolver(ifpackDSType,ifpackDSList) ); smoProtoSC->SetFactory("A", SFact);
+      smoProtoSC = rcp(new DirectSolver(ifpackDSType, ifpackDSList));
+      smoProtoSC->SetFactory("A", SFact);
     }
 
-    RCP<SmootherFactory> SmooSCFact = rcp( new SmootherFactory(smoProtoSC) );
+    RCP<SmootherFactory> SmooSCFact = rcp(new SmootherFactory(smoProtoSC));
 
     // define temporary FactoryManager that is used as input for BraessSarazin smoother
     RCP<FactoryManager> MB = rcp(new FactoryManager());
-    MB->SetFactory("A",                 SFact);         // SchurComplement operator for correction step (defined as "A")
-    MB->SetFactory("Smoother",          SmooSCFact);    // solver/smoother for correction step
-    MB->SetFactory("PreSmoother",               SmooSCFact);
-    MB->SetFactory("PostSmoother",              SmooSCFact);
-    MB->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-    BraessSarazinSm->AddFactoryManager(MB,0);  // set temporary factory manager in BraessSarazin smoother
+    MB->SetFactory("A", SFact);              // SchurComplement operator for correction step (defined as "A")
+    MB->SetFactory("Smoother", SmooSCFact);  // solver/smoother for correction step
+    MB->SetFactory("PreSmoother", SmooSCFact);
+    MB->SetFactory("PostSmoother", SmooSCFact);
+    MB->SetIgnoreUserData(true);                // always use data from factories defined in factory manager
+    BraessSarazinSm->AddFactoryManager(MB, 0);  // set temporary factory manager in BraessSarazin smoother
 
     // setup main factory manager
     RCP<FactoryManager> M = rcp(new FactoryManager());
-    M->SetFactory("A",               MueLu::NoFactory::getRCP()); // this is the 2x2 blocked operator
-    M->SetFactory("Smoother",        smootherFact);               // BraessSarazin block smoother
-    M->SetFactory("PreSmoother",     smootherFact);
-    M->SetFactory("PostSmoother",    smootherFact);
+    M->SetFactory("A", MueLu::NoFactory::getRCP());  // this is the 2x2 blocked operator
+    M->SetFactory("Smoother", smootherFact);         // BraessSarazin block smoother
+    M->SetFactory("PreSmoother", smootherFact);
+    M->SetFactory("PostSmoother", smootherFact);
 
     MueLu::SetFactoryManager SFMCoarse(Finest, M);
-    Finest->Request(MueLu::TopSmootherFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>(M, "Smoother"));
+    Finest->Request(MueLu::TopSmootherFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>(M, "Smoother"));
 
     // call setup (= extract blocks and extract diagonal of F)
     BraessSarazinSm->Setup(*Finest);
 
-    RCP<MultiVector> xtest = MultiVectorFactory::Build(xstridedfullmap,1);
-    xtest->putScalar( (SC) 0.0);
+    RCP<MultiVector> xtest = MultiVectorFactory::Build(xstridedfullmap, 1);
+    xtest->putScalar((SC)0.0);
 
-    RCP<Vector> xR = rcp(new Xpetra::EpetraVectorT<GO,Node>(epv));
+    RCP<Vector> xR = rcp(new Xpetra::EpetraVectorT<GO, Node>(epv));
     // calculate initial (absolute) residual
     Array<ScalarTraits<SC>::magnitudeType> norms(1);
 
@@ -342,7 +341,7 @@ int main(int argc, char *argv[]) {
     *out << "Test: Applying Braess-Sarazin Smoother" << std::endl;
     *out << "Test: START DATA" << std::endl;
     *out << "iterations\tVelocity_residual\tPressure_residual" << std::endl;
-    BraessSarazinSm->Apply(*xtest,*xR);
+    BraessSarazinSm->Apply(*xtest, *xR);
     xtest->norm2(norms);
     *out << "Test: ||x_1|| = " << norms[0] << std::endl;
 
@@ -353,9 +352,9 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
 }
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_Epetra.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_Epetra.cpp
index cfa911292ac0..d138e3fac228 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_Epetra.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_Epetra.cpp
@@ -120,8 +120,7 @@
  *  using block matrices
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -137,29 +136,29 @@ int main(int argc, char *argv[]) {
   using namespace Teuchos;
 
   oblackholestream blackhole;
-  GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
   try {
     // default parameters
-    LO BGS_nSweeps = 3;
+    LO BGS_nSweeps   = 3;
     Scalar BGS_omega = 1.0;
 
     // Note: use --help to list available options.
     CommandLineProcessor clp(false);
-    clp.setOption("BGS_sweeps",&BGS_nSweeps,"number of sweeps with BGS smoother");
-    clp.setOption("BGS_omega", &BGS_omega,  "scaling factor for BGS smoother");
+    clp.setOption("BGS_sweeps", &BGS_nSweeps, "number of sweeps with BGS smoother");
+    clp.setOption("BGS_omega", &BGS_omega, "scaling factor for BGS smoother");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     RCP<const Comm<int> > comm = DefaultComm<int>::getComm();
-    RCP<FancyOStream> out = fancyOStream(rcpFromRef(std::cout));
+    RCP<FancyOStream> out      = fancyOStream(rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -168,12 +167,12 @@ int main(int argc, char *argv[]) {
     TimeMonitor MM(myTime);
 
     // custom parameters
-    LO maxLevels = 2;   // TODO: singular system if MaxLevels > 2?
+    LO maxLevels = 2;  // TODO: singular system if MaxLevels > 2?
 
-    GO maxCoarseSize=1; //FIXME clp doesn't like long long int
+    GO maxCoarseSize = 1;  // FIXME clp doesn't like long long int
 
     int globalNumDofs = 1500;  // used for the maps
-    int nDofsPerNode = 3;            // used for generating the fine level null-space
+    int nDofsPerNode  = 3;     // used for generating the fine level null-space
 
     // build strided maps
     // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -186,29 +185,29 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib             = Xpetra::UseEpetra;
+    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for AztecOO
     const RCP<const Epetra_Map> fullmap = rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
-    RCP<Epetra_CrsMatrix> epA = rcp(ptrA);
-    RCP<Epetra_Vector> epv = rcp(ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
+    RCP<Epetra_CrsMatrix> epA    = rcp(ptrA);
+    RCP<Epetra_Vector> epv       = rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = rcp(ptrNS);
 
     /////////////////////////////////////// split system into 2x2 block system
@@ -221,81 +220,81 @@ int main(int argc, char *argv[]) {
     RCP<Epetra_CrsMatrix> A21;
     RCP<Epetra_CrsMatrix> A22;
 
-    if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A11));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A12));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A21));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A22));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A11));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A12));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A21));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<RCP<const Xpetra::Map<LO,GO,Node> > > xmaps;
+    std::vector<RCP<const Xpetra::Map<LO, GO, Node> > > xmaps;
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    RCP<const Xpetra::MapExtractor<Scalar,LO,GO,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(xstridedfullmap,xmaps);
+    RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(xstridedfullmap, xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
 
     //////////////////////////////////////////////////// create Hierarchy
-    RCP<Hierarchy> H = rcp ( new Hierarchy() );
+    RCP<Hierarchy> H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(VERB_HIGH);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     //////////////////////////////////////////////////////// finest Level
     RCP<MueLu::Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(VERB_HIGH);
-    Finest->Set("A",rcp_dynamic_cast<Matrix>(bOp));
+    Finest->Set("A", rcp_dynamic_cast<Matrix>(bOp));
 
     /////////////////////////////////////////////// define subblocks of A
     // make A11 block and A22 block available as variable "A" generated
     // by A11Fact and A22Fact
     RCP<SubBlockAFactory> A11Fact = rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",ParameterEntry(0));
-    A11Fact->SetParameter("block col",ParameterEntry(0));
+    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A11Fact->SetParameter("block row", ParameterEntry(0));
+    A11Fact->SetParameter("block col", ParameterEntry(0));
     RCP<SubBlockAFactory> A22Fact = rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",ParameterEntry(1));
-    A22Fact->SetParameter("block col",ParameterEntry(1));
+    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A22Fact->SetParameter("block row", ParameterEntry(1));
+    A22Fact->SetParameter("block col", ParameterEntry(1));
 
     ///////////////////////////////////////////// define smoother for A11
     // define block smoother for the first block matrix row in BlockGaussSeidel Smoother
     std::string ifpack11Type;
     ParameterList ifpack11List;
-    ifpack11List.set("relaxation: sweeps", (LO) 3);
-    ifpack11List.set("relaxation: damping factor", (SC) 0.5);
+    ifpack11List.set("relaxation: sweeps", (LO)3);
+    ifpack11List.set("relaxation: damping factor", (SC)0.5);
     ifpack11Type = "RELAXATION";
     ifpack11List.set("relaxation: type", "Gauss-Seidel");
-    RCP<SmootherPrototype> smoProto11     = rcp( new TrilinosSmoother(ifpack11Type, ifpack11List, 0) );
+    RCP<SmootherPrototype> smoProto11 = rcp(new TrilinosSmoother(ifpack11Type, ifpack11List, 0));
     smoProto11->SetFactory("A", A11Fact);
-    RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
+    RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
 
     ////////////////////////////////////////// prepare null space for A11
     RCP<MultiVector> nullspace11 = MultiVectorFactory::Build(xstridedvelmap, 2);  // this is a 2D standard null space
 
-    for (int i=0; i<nDofsPerNode-1; ++i) {
+    for (int i = 0; i < nDofsPerNode - 1; ++i) {
       ArrayRCP<Scalar> nsValues = nullspace11->getDataNonConst(i);
-      int numBlocks = nsValues.size() / (nDofsPerNode - 1);
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*(nDofsPerNode - 1) + i] = 1.0;
+      int numBlocks             = nsValues.size() / (nDofsPerNode - 1);
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * (nDofsPerNode - 1) + i] = 1.0;
       }
     }
 
-    Finest->Set("Nullspace1",nullspace11);
+    Finest->Set("Nullspace1", nullspace11);
 
     ///////////////////////////////////////// define CoalesceDropFactory and Aggregation for A11
     // set up amalgamation for A11. Note: we're using a default null space factory (null)
@@ -345,7 +344,7 @@ int main(int argc, char *argv[]) {
     RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
 
     RCP<NullspaceFactory> nspFact11 = rcp(new NullspaceFactory("Nullspace1"));
-    nspFact11->SetFactory("Nullspace1",P11Fact);
+    nspFact11->SetFactory("Nullspace1", P11Fact);
 
     RCP<CoarseMapFactory> coarseMapFact11 = rcp(new CoarseMapFactory());
     coarseMapFact11->setStridingData(stridingInfo);
@@ -365,11 +364,11 @@ int main(int argc, char *argv[]) {
     M11->SetFactory("CoarseMap", coarseMapFact11);
 
 #if OLD
-    RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory(CoupledAggFact11,amalgFact11)); // check me
+    RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory(CoupledAggFact11, amalgFact11));  // check me
 
     RCP<TransPFactory> R11Fact = rcp(new TransPFactory(P11Fact));
 
-    RCP<NullspaceFactory> nspFact11 = rcp(new NullspaceFactory("Nullspace1",P11Fact));
+    RCP<NullspaceFactory> nspFact11 = rcp(new NullspaceFactory("Nullspace1", P11Fact));
 
     RCP<CoarseMapFactory> coarseMapFact11 = rcp(new CoarseMapFactory(CoupledAggFact11, nspFact11));
     coarseMapFact11->setStridingData(stridingInfo);
@@ -384,30 +383,30 @@ int main(int argc, char *argv[]) {
     M11->SetFactory("Ptent", P11Fact);
     M11->SetFactory("Smoother", Smoo11Fact);
     M11->SetFactory("CoarseMap", coarseMapFact11);
-#endif // TODO remove this
+#endif  // TODO remove this
 
 #endif
-    M11->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    M11->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
     ///////////////////////////////////////////// define smoother for A22
     std::string ifpack22Type;
     ParameterList ifpack22List;
-    ifpack22List.set("relaxation: sweeps", (LO) 1);
-    ifpack22List.set("relaxation: damping factor", (SC) 0.3);
+    ifpack22List.set("relaxation: sweeps", (LO)1);
+    ifpack22List.set("relaxation: damping factor", (SC)0.3);
     ifpack22Type = "RELAXATION";
     ifpack22List.set("relaxation: type", "Gauss-Seidel");
-    RCP<SmootherPrototype> smoProto22     = rcp( new TrilinosSmoother(ifpack22Type, ifpack22List, 0) );
+    RCP<SmootherPrototype> smoProto22 = rcp(new TrilinosSmoother(ifpack22Type, ifpack22List, 0));
     smoProto22->SetFactory("A", A22Fact);
-    RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
+    RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
 
     ////////////////////////////////////////// prepare null space for A22
     RCP<MultiVector> nullspace22 = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
-    ArrayRCP<Scalar> nsValues22 = nullspace22->getDataNonConst(0);
-    for (int j=0; j< nsValues22.size(); ++j) {
+    ArrayRCP<Scalar> nsValues22  = nullspace22->getDataNonConst(0);
+    for (int j = 0; j < nsValues22.size(); ++j) {
       nsValues22[j] = 1.0;
     }
 
-    Finest->Set("Nullspace2",nullspace22);
+    Finest->Set("Nullspace2", nullspace22);
 
     ///////////////////////////////////////// define transfer ops for A22
 #if 0
@@ -442,12 +441,12 @@ int main(int argc, char *argv[]) {
 #else
     // use TentativePFactory
     RCP<AmalgamationFactory> amalgFact22 = rcp(new AmalgamationFactory());
-    RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory()); // check me (fed with A22) wrong column GIDS!!!
+    RCP<TentativePFactory> P22Fact       = rcp(new TentativePFactory());  // check me (fed with A22) wrong column GIDS!!!
 
     RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
 
     RCP<NullspaceFactory> nspFact22 = rcp(new NullspaceFactory("Nullspace2"));
-    nspFact22->SetFactory("Nullspace2",P22Fact);
+    nspFact22->SetFactory("Nullspace2", P22Fact);
 
     RCP<CoarseMapFactory> coarseMapFact22 = rcp(new CoarseMapFactory());
     coarseMapFact22->setStridingData(stridingInfo);
@@ -464,11 +463,11 @@ int main(int argc, char *argv[]) {
     M22->SetFactory("Ptent", P22Fact);
     M22->SetFactory("Smoother", Smoo22Fact);
     M22->SetFactory("CoarseMap", coarseMapFact22);
-    M22->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    M22->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 #endif
 
     /////////////////////////////////////////// define blocked transfer ops
-    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory()); // use row map index base from bOp
+    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());  // use row map index base from bOp
     PFact->AddFactoryManager(M11);
     PFact->AddFactoryManager(M22);
 
@@ -481,35 +480,34 @@ int main(int argc, char *argv[]) {
 
     //////////////////////////////////////////////////////////////////////
     // Smoothers
-    RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
     smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(BGS_nSweeps));
     smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(BGS_omega));
-    smootherPrototype->AddFactoryManager(M11,0);
-    smootherPrototype->AddFactoryManager(M22,1);
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->AddFactoryManager(M11, 0);
+    smootherPrototype->AddFactoryManager(M22, 1);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
     // Coarse grid correction
-    RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp( new BlockedGaussSeidelSmoother() );
+    RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp(new BlockedGaussSeidelSmoother());
     coarseSolverPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(BGS_nSweeps));
     coarseSolverPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(BGS_omega));
-    coarseSolverPrototype->AddFactoryManager(M11,0);
-    coarseSolverPrototype->AddFactoryManager(M22,1);
-    RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, null) );
+    coarseSolverPrototype->AddFactoryManager(M11, 0);
+    coarseSolverPrototype->AddFactoryManager(M22, 1);
+    RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, null));
 
     // main factory manager
     FactoryManager M;
-    M.SetFactory("A",            AcFact);
-    M.SetFactory("P",            PFact);
-    M.SetFactory("R",            RFact);
-    M.SetFactory("Smoother",     smootherFact); // TODO fix me
-    M.SetFactory("PreSmoother",  smootherFact); // TODO fix me
-    M.SetFactory("PostSmoother", smootherFact); // TODO fix me
+    M.SetFactory("A", AcFact);
+    M.SetFactory("P", PFact);
+    M.SetFactory("R", RFact);
+    M.SetFactory("Smoother", smootherFact);      // TODO fix me
+    M.SetFactory("PreSmoother", smootherFact);   // TODO fix me
+    M.SetFactory("PostSmoother", smootherFact);  // TODO fix me
     M.SetFactory("CoarseSolver", coarseSolverFact);
 
     //////////////////////////////////// setup multigrid
 
-    H->Setup(M,0,maxLevels);
-
+    H->Setup(M, 0, maxLevels);
 
     *out << std::endl;
     *out << "print content of multigrid levels:" << std::endl;
@@ -519,17 +517,17 @@ int main(int argc, char *argv[]) {
     RCP<Level> coarseLevel = H->GetLevel(1);
     coarseLevel->print(*out);
 
-    //RCP<Level> coarseLevel2 = H->GetLevel(2);
-    //coarseLevel2->print(*out);
+    // RCP<Level> coarseLevel2 = H->GetLevel(2);
+    // coarseLevel2->print(*out);
 
-    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap,1);
+    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap, 1);
 
     // Use AMG directly as an iterative method
     {
-      xLsg->putScalar( (SC) 0.0);
+      xLsg->putScalar((SC)0.0);
 
       // Epetra_Vector -> Xpetra::Vector
-      RCP<Vector> xRhs = rcp(new Xpetra::EpetraVectorT<int,Node>(epv));
+      RCP<Vector> xRhs = rcp(new Xpetra::EpetraVectorT<int, Node>(epv));
 
       // calculate initial (absolute) residual
       Array<ScalarTraits<SC>::magnitudeType> norms(1);
@@ -537,15 +535,14 @@ int main(int argc, char *argv[]) {
       *out << "||x_0|| = " << norms[0] << std::endl;
 
       // apply ten multigrid iterations
-      H->Iterate(*xRhs,*xLsg,100);
+      H->Iterate(*xRhs, *xLsg, 100);
 
       // calculate and print residual
-      RCP<MultiVector> xTmp = MultiVectorFactory::Build(xstridedfullmap,1);
-      bOp->apply(*xLsg,*xTmp,NO_TRANS,(SC)1.0,(SC)0.0);
-      xRhs->update((SC)-1.0,*xTmp,(SC)1.0);
+      RCP<MultiVector> xTmp = MultiVectorFactory::Build(xstridedfullmap, 1);
+      bOp->apply(*xLsg, *xTmp, NO_TRANS, (SC)1.0, (SC)0.0);
+      xRhs->update((SC)-1.0, *xTmp, (SC)1.0);
       xRhs->norm2(norms);
       *out << "||r|| = " << norms[0] << std::endl;
-
     }
 
     // TODO: don't forget to add Aztec as prerequisite in CMakeLists.txt!
@@ -573,9 +570,9 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
 }
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_Simple.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_Simple.cpp
index 1875ba067431..1f19ebb0b6cb 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_Simple.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_Simple.cpp
@@ -126,8 +126,7 @@
  *  using block matrices
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -141,14 +140,14 @@ int main(int argc, char *argv[]) {
   using Teuchos::rcp;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
   try {
     //
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -159,10 +158,10 @@ int main(int argc, char *argv[]) {
     // custom parameters
     LocalOrdinal maxLevels = 3;
 
-    GlobalOrdinal maxCoarseSize=1; //FIXME clp doesn't like long long int
+    GlobalOrdinal maxCoarseSize = 1;  // FIXME clp doesn't like long long int
 
     int globalNumDofs = 1500;  // used for the maps
-    int nDofsPerNode = 3;      // used for generating the fine level null-space
+    int nDofsPerNode  = 3;     // used for generating the fine level null-space
 
     // build strided maps
     // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -175,34 +174,33 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib       = Xpetra::UseEpetra;
+    RCP<StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for AztecOO
     const RCP<const Epetra_Map> fullmap = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
 
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
 
-    RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    RCP<Epetra_Vector> epv = Teuchos::rcp(ptrf);
+    RCP<Epetra_CrsMatrix> epA    = Teuchos::rcp(ptrA);
+    RCP<Epetra_Vector> epv       = Teuchos::rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = Teuchos::rcp(ptrNS);
 
-
     /////////////////////////////////////// split system into 2x2 block system
 
     *out << "Split matrix into 2x2 block matrix" << std::endl;
@@ -213,70 +211,70 @@ int main(int argc, char *argv[]) {
     Teuchos::RCP<Epetra_CrsMatrix> A21;
     Teuchos::RCP<Epetra_CrsMatrix> A22;
 
-    if(MueLuTests::SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (MueLuTests::SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A11));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A12));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A21));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A22));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A11));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A12));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A21));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > > xmaps;
+    std::vector<Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > > xmaps;
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(xstridedfullmap,xmaps);
+    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(xstridedfullmap, xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
 
     //////////////////////////////////////////////////// create Hierarchy
-    RCP<Hierarchy> H = rcp ( new Hierarchy() );
+    RCP<Hierarchy> H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    //H->setDefaultVerbLevel(Teuchos::VERB_NONE);
+    // H->setDefaultVerbLevel(Teuchos::VERB_NONE);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     //////////////////////////////////////////////////////// finest Level
     RCP<MueLu::Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Teuchos::rcp_dynamic_cast<Matrix>(bOp));
+    Finest->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));
 
     /////////////////////////////////////////////// define subblocks of A
     // make A11 block and A22 block available as variable "A" generated
     // by A11Fact and A22Fact
     RCP<SubBlockAFactory> A11Fact = rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
     RCP<SubBlockAFactory> A22Fact = rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+    A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
 
     ////////////////////////////////////////// prepare null space for A11
     RCP<MultiVector> nullspace11 = MultiVectorFactory::Build(xstridedvelmap, 2);  // this is a 2D standard null space
 
-    for (int i=0; i<nDofsPerNode-1; ++i) {
+    for (int i = 0; i < nDofsPerNode - 1; ++i) {
       Teuchos::ArrayRCP<Scalar> nsValues = nullspace11->getDataNonConst(i);
-      int numBlocks = nsValues.size() / (nDofsPerNode - 1);
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*(nDofsPerNode - 1) + i] = 1.0;
+      int numBlocks                      = nsValues.size() / (nDofsPerNode - 1);
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * (nDofsPerNode - 1) + i] = 1.0;
       }
     }
 
-    Finest->Set("Nullspace1",nullspace11);
+    Finest->Set("Nullspace1", nullspace11);
 
     ///////////////////////////////////////// define CoalesceDropFactory and Aggregation for A11
     // set up amalgamation for A11. Note: we're using a default null space factory (Teuchos::null)
@@ -324,7 +322,7 @@ int main(int argc, char *argv[]) {
     RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
 
     Teuchos::RCP<NullspaceFactory> nspFact11 = Teuchos::rcp(new NullspaceFactory("Nullspace1"));
-    nspFact11->SetFactory("Nullspace1",P11Fact);
+    nspFact11->SetFactory("Nullspace1", P11Fact);
 
     RCP<CoarseMapFactory> coarseMapFact11 = Teuchos::rcp(new CoarseMapFactory());
     coarseMapFact11->setStridingData(stridingInfo);
@@ -342,16 +340,16 @@ int main(int argc, char *argv[]) {
     M11->SetFactory("CoarseMap", coarseMapFact11);
     M11->SetFactory("Graph", dropFact11);
 #endif
-    M11->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    M11->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
     ////////////////////////////////////////// prepare null space for A22
-    RCP<MultiVector> nullspace22 = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
+    RCP<MultiVector> nullspace22         = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
     Teuchos::ArrayRCP<Scalar> nsValues22 = nullspace22->getDataNonConst(0);
-    for (int j=0; j< nsValues22.size(); ++j) {
+    for (int j = 0; j < nsValues22.size(); ++j) {
       nsValues22[j] = 1.0;
     }
 
-    Finest->Set("Nullspace2",nullspace22);
+    Finest->Set("Nullspace2", nullspace22);
 
     ///////////////////////////////////////// define transfer ops for A22
 #if 0
@@ -384,7 +382,7 @@ int main(int argc, char *argv[]) {
 #else
     // use TentativePFactory
     RCP<AmalgamationFactory> amalgFact22 = rcp(new AmalgamationFactory());
-    RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory()); // check me (fed with A22) wrong column GIDS!!!
+    RCP<TentativePFactory> P22Fact       = rcp(new TentativePFactory());  // check me (fed with A22) wrong column GIDS!!!
 
     RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
 
@@ -404,7 +402,7 @@ int main(int argc, char *argv[]) {
     M22->SetFactory("UnAmalgamationInfo", amalgFact22);
     M22->SetFactory("Ptent", P22Fact);
     M22->SetFactory("CoarseMap", coarseMapFact22);
-    M22->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    M22->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 #endif
 
     /////////////////////////////////////////// define blocked transfer ops
@@ -425,25 +423,25 @@ int main(int argc, char *argv[]) {
     // Smoothers
 
     RCP<SubBlockAFactory> A00Fact = Teuchos::rcp(new SubBlockAFactory());
-    A00Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A00Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A00Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+    A00Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A00Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    A00Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
     std::string ifpackTypePredictSmoother;
     Teuchos::ParameterList ifpackListPredictSmoother;
-    ifpackListPredictSmoother.set("relaxation: sweeps", (LocalOrdinal) 1);
-    ifpackListPredictSmoother.set("relaxation: damping factor", (Scalar) 0.5);
+    ifpackListPredictSmoother.set("relaxation: sweeps", (LocalOrdinal)1);
+    ifpackListPredictSmoother.set("relaxation: damping factor", (Scalar)0.5);
     ifpackTypePredictSmoother = "RELAXATION";
     ifpackListPredictSmoother.set("relaxation: type", "Gauss-Seidel");
-    RCP<SmootherPrototype> smoProtoPredict     = rcp( new TrilinosSmoother(ifpackTypePredictSmoother, ifpackListPredictSmoother, 0) );
+    RCP<SmootherPrototype> smoProtoPredict = rcp(new TrilinosSmoother(ifpackTypePredictSmoother, ifpackListPredictSmoother, 0));
     smoProtoPredict->SetFactory("A", A00Fact);
-    RCP<SmootherFactory> SmooPredictFact = rcp( new SmootherFactory(smoProtoPredict) );
+    RCP<SmootherFactory> SmooPredictFact = rcp(new SmootherFactory(smoProtoPredict));
 
     RCP<FactoryManager> MPredict = rcp(new FactoryManager());
-    MPredict->SetFactory("A",                 A00Fact);         // SchurComplement operator for correction step (defined as "A")
-    MPredict->SetFactory("Smoother",          SmooPredictFact);    // solver/smoother for correction step
-    MPredict->SetFactory("PreSmoother",               SmooPredictFact);
-    MPredict->SetFactory("PostSmoother",              SmooPredictFact);
-    MPredict->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    MPredict->SetFactory("A", A00Fact);                 // SchurComplement operator for correction step (defined as "A")
+    MPredict->SetFactory("Smoother", SmooPredictFact);  // solver/smoother for correction step
+    MPredict->SetFactory("PreSmoother", SmooPredictFact);
+    MPredict->SetFactory("PostSmoother", SmooPredictFact);
+    MPredict->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
     ////////////////////////////////////////////////
     // InverseApproximation
@@ -455,64 +453,63 @@ int main(int argc, char *argv[]) {
     // create SchurComp factory (SchurComplement smoother is provided by local FactoryManager)
     Teuchos::RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
     SFact->SetParameter("omega", Teuchos::ParameterEntry(0.8));
-    SFact->SetFactory("A", MueLu::NoFactory::getRCP()); // 2x2 blocked operator
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());  // 2x2 blocked operator
     SFact->SetFactory("Ainv", AinvFact);
 
     // define SchurComplement solver
     std::string ifpackTypeSchurSmoother;
     ifpackTypeSchurSmoother = "RELAXATION";
     Teuchos::ParameterList ifpackListSchurSmoother;
-    ifpackListSchurSmoother.set("relaxation: sweeps", (LocalOrdinal) 10);
-    ifpackListSchurSmoother.set("relaxation: damping factor", (Scalar) 0.8);
+    ifpackListSchurSmoother.set("relaxation: sweeps", (LocalOrdinal)10);
+    ifpackListSchurSmoother.set("relaxation: damping factor", (Scalar)0.8);
     ifpackListSchurSmoother.set("relaxation: type", "Gauss-Seidel");
-    RCP<SmootherPrototype> smoProtoSC     = rcp( new TrilinosSmoother(ifpackTypeSchurSmoother, ifpackListSchurSmoother, 0) );
-    smoProtoSC->SetFactory("A", SFact); // explicitely use SchurComplement matrix as input for smoother
-    RCP<SmootherFactory> SmooSCFact = rcp( new SmootherFactory(smoProtoSC) );
+    RCP<SmootherPrototype> smoProtoSC = rcp(new TrilinosSmoother(ifpackTypeSchurSmoother, ifpackListSchurSmoother, 0));
+    smoProtoSC->SetFactory("A", SFact);  // explicitely use SchurComplement matrix as input for smoother
+    RCP<SmootherFactory> SmooSCFact = rcp(new SmootherFactory(smoProtoSC));
 
     // setup local factory manager for SchurComplementFactory
     Teuchos::RCP<FactoryManager> MSchur = Teuchos::rcp(new FactoryManager());
-    MSchur->SetFactory("A", SFact);              // SchurCompFactory as generating factory for SchurComp equation
+    MSchur->SetFactory("A", SFact);  // SchurCompFactory as generating factory for SchurComp equation
     MSchur->SetFactory("Smoother", SmooSCFact);
     MSchur->SetIgnoreUserData(true);
 
-
     /////////////////////////////////////////////////////
     // create smoother prototype
 
-    RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
     smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(3));
     smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(0.6));
 
-    smootherPrototype->AddFactoryManager(MPredict,0);    // set temporary factory manager for prediction step
-    smootherPrototype->AddFactoryManager(MSchur,1);      // set temporary factory manager for correction step
+    smootherPrototype->AddFactoryManager(MPredict, 0);  // set temporary factory manager for prediction step
+    smootherPrototype->AddFactoryManager(MSchur, 1);    // set temporary factory manager for correction step
     smootherPrototype->SetFactory("A", MueLu::NoFactory::getRCP());
 
     /////////////////////////////////////////////////////
     // create smoother factories
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );                // pre and postsmoothing with SIMPLE on the finest and intermedium levels
-    RCP<SmootherFactory>   coarseSmootherFact    = rcp( new SmootherFactory(smootherPrototype, Teuchos::null) ); // only presmoothing on finest level (we do not want to run two SIMPLE iterations on the coarsest level)
+    RCP<SmootherFactory> smootherFact       = rcp(new SmootherFactory(smootherPrototype));                 // pre and postsmoothing with SIMPLE on the finest and intermedium levels
+    RCP<SmootherFactory> coarseSmootherFact = rcp(new SmootherFactory(smootherPrototype, Teuchos::null));  // only presmoothing on finest level (we do not want to run two SIMPLE iterations on the coarsest level)
 
     // main factory manager
     FactoryManager M;
-    M.SetFactory("A",            AcFact);
-    M.SetFactory("P",            PFact);
-    M.SetFactory("R",            RFact);
-    M.SetFactory("Smoother",     smootherFact); // TODO fix me
+    M.SetFactory("A", AcFact);
+    M.SetFactory("P", PFact);
+    M.SetFactory("R", RFact);
+    M.SetFactory("Smoother", smootherFact);  // TODO fix me
     M.SetFactory("CoarseSolver", coarseSmootherFact);
 
     //////////////////////////////////// setup multigrid
 
-    H->Setup(M,0,maxLevels);
+    H->Setup(M, 0, maxLevels);
 
     Finest->print(*out);
 
     RCP<Level> coarseLevel = H->GetLevel(1);
     coarseLevel->print(*out);
 
-    //RCP<Level> coarseLevel2 = H->GetLevel(2);
-    //coarseLevel2->print(*out);
+    // RCP<Level> coarseLevel2 = H->GetLevel(2);
+    // coarseLevel2->print(*out);
 
-    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap,1);
+    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap, 1);
 
     // Use AMG directly as an iterative method
 #if 0
@@ -564,9 +561,9 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
 }
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_SimpleSmoother.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_SimpleSmoother.cpp
index 60b5f62237aa..dfe06be4c91c 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_SimpleSmoother.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_SimpleSmoother.cpp
@@ -113,8 +113,7 @@
  *  using block matrices
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -129,13 +128,13 @@ int main(int argc, char *argv[]) {
   using namespace MueLuTests;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -146,32 +145,32 @@ int main(int argc, char *argv[]) {
     // read in input parameters
 
     // default parameters
-    LO SIMPLE_nSweeps = 600;
+    LO SIMPLE_nSweeps   = 600;
     Scalar SIMPLE_omega = 0.5;
-    LO SC_nSweeps = 10;
-    Scalar SC_omega = 1.0;
-    LO PRED_nSweeps = 3;
-    Scalar PRED_omega = 1.0;
-    LO useSIMPLEC = 0;
+    LO SC_nSweeps       = 10;
+    Scalar SC_omega     = 1.0;
+    LO PRED_nSweeps     = 3;
+    Scalar PRED_omega   = 1.0;
+    LO useSIMPLEC       = 0;
 
     int SC_bUseDirectSolver = 0;
 
     // Note: use --help to list available options.
     Teuchos::CommandLineProcessor clp(false);
-    clp.setOption("SIMPLE_sweeps",&SIMPLE_nSweeps,"number of sweeps with SIMPLE smoother");
-    clp.setOption("SIMPLE_omega", &SIMPLE_omega,  "scaling factor for SIMPLE smoother");
-    clp.setOption("Predict_sweeps", &PRED_nSweeps,  "number of sweeps for SIMPLE internal velocity prediction smoother (GaussSeidel)");
-    clp.setOption("Predict_omega", &PRED_omega,  "damping parameter for SIMPLE internal velocity prediction smoother (GaussSeidel)");
-    clp.setOption("SchurComp_sweeps",    &SC_nSweeps,"number of sweeps for SIMPLE internal SchurComp solver/smoother (GaussSeidel)");
-    clp.setOption("SchurComp_omega",     &SC_omega,  "damping parameter for SIMPLE internal SchurComp solver/smoother (GaussSeidel)");
-    clp.setOption("SchurComp_solver",    &SC_bUseDirectSolver,  "if 1: use direct solver for SchurComp equation, otherwise use GaussSeidel smoother");
-    clp.setOption("useSIMPLEC",    &useSIMPLEC,  "if 1: use SIMPLEC instead of SIMPLE (default = 0 (SIMPLE))");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    clp.setOption("SIMPLE_sweeps", &SIMPLE_nSweeps, "number of sweeps with SIMPLE smoother");
+    clp.setOption("SIMPLE_omega", &SIMPLE_omega, "scaling factor for SIMPLE smoother");
+    clp.setOption("Predict_sweeps", &PRED_nSweeps, "number of sweeps for SIMPLE internal velocity prediction smoother (GaussSeidel)");
+    clp.setOption("Predict_omega", &PRED_omega, "damping parameter for SIMPLE internal velocity prediction smoother (GaussSeidel)");
+    clp.setOption("SchurComp_sweeps", &SC_nSweeps, "number of sweeps for SIMPLE internal SchurComp solver/smoother (GaussSeidel)");
+    clp.setOption("SchurComp_omega", &SC_omega, "damping parameter for SIMPLE internal SchurComp solver/smoother (GaussSeidel)");
+    clp.setOption("SchurComp_solver", &SC_bUseDirectSolver, "if 1: use direct solver for SchurComp equation, otherwise use GaussSeidel smoother");
+    clp.setOption("useSIMPLEC", &useSIMPLEC, "if 1: use SIMPLEC instead of SIMPLE (default = 0 (SIMPLE))");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     int globalNumDofs = 1500;  // used for the maps
@@ -187,34 +186,33 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib             = Xpetra::UseEpetra;
+    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for our splitting routine
     const RCP<const Epetra_Map> fullmap = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = Teuchos::rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
 
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
 
-    RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    RCP<Epetra_Vector> epv = Teuchos::rcp(ptrf);
+    RCP<Epetra_CrsMatrix> epA    = Teuchos::rcp(ptrA);
+    RCP<Epetra_Vector> epv       = Teuchos::rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = Teuchos::rcp(ptrNS);
 
-
     /////////////////////////////////////// split system into 2x2 block system
 
     *out << "Split matrix into 2x2 block matrix" << std::endl;
@@ -225,40 +223,39 @@ int main(int argc, char *argv[]) {
     Teuchos::RCP<Epetra_CrsMatrix> A21;
     Teuchos::RCP<Epetra_CrsMatrix> A22;
 
-    if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A11));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A12));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A21));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A22));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A11));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A12));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A21));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<Teuchos::RCP<const Xpetra::Map<LO,GO,Node> > > xmaps;
+    std::vector<Teuchos::RCP<const Xpetra::Map<LO, GO, Node> > > xmaps;
 
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar,LO,GO,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(xstridedfullmap,xmaps);
+    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(xstridedfullmap, xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
     //////////////////////////////////////////////////////// finest Level
     RCP<MueLu::Level> Finest = rcp(new Level());
     Finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
-    Finest->Set("A",Teuchos::rcp_dynamic_cast<Matrix>(bOp));
-
+    Finest->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));
 
     ///////////////////////////////////
     // Test Braess Sarazin Smoother as a solver
@@ -272,38 +269,37 @@ int main(int argc, char *argv[]) {
 
     // define SIMPLE Smoother with SIMPLE_nSweeps and SIMPLE_omega as scaling factor
     // AFact_ = Teuchos::null (= default) for the 2x2 blocked operator
-    RCP<SimpleSmoother> SimpleSm = rcp( new SimpleSmoother() );
+    RCP<SimpleSmoother> SimpleSm = rcp(new SimpleSmoother());
     SimpleSm->SetParameter("Sweeps", Teuchos::ParameterEntry(SIMPLE_nSweeps));
     SimpleSm->SetParameter("Damping factor", Teuchos::ParameterEntry(SIMPLE_omega));
-    if(useSIMPLEC==1) SimpleSm->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(true));
+    if (useSIMPLEC == 1) SimpleSm->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(true));
 
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(SimpleSm) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(SimpleSm));
 
     // define smoother for velocity prediction
-    //RCP<SubBlockAFactory> A00Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 0, 0));
+    // RCP<SubBlockAFactory> A00Fact = Teuchos::rcp(new SubBlockAFactory(MueLu::NoFactory::getRCP(), 0, 0));
     RCP<SubBlockAFactory> A00Fact = rcp(new SubBlockAFactory());
-    A00Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A00Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A00Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+    A00Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A00Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    A00Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
     RCP<SmootherPrototype> smoProtoPredict = Teuchos::null;
     std::string ifpackPredictType;
     Teuchos::ParameterList ifpackPredictList;
-    ifpackPredictList.set("relaxation: sweeps", PRED_nSweeps );
-    ifpackPredictList.set("relaxation: damping factor", PRED_omega );
+    ifpackPredictList.set("relaxation: sweeps", PRED_nSweeps);
+    ifpackPredictList.set("relaxation: damping factor", PRED_omega);
     ifpackPredictType = "RELAXATION";
     ifpackPredictList.set("relaxation: type", "Gauss-Seidel");
-    smoProtoPredict = rcp( new TrilinosSmoother(ifpackPredictType, ifpackPredictList, 0) );
+    smoProtoPredict = rcp(new TrilinosSmoother(ifpackPredictType, ifpackPredictList, 0));
     smoProtoPredict->SetFactory("A", A00Fact);
-    RCP<SmootherFactory> SmooPredictFact = rcp( new SmootherFactory(smoProtoPredict) );
+    RCP<SmootherFactory> SmooPredictFact = rcp(new SmootherFactory(smoProtoPredict));
     // define temporary FactoryManager that is used as input for BraessSarazin smoother
     RCP<FactoryManager> MPredict = rcp(new FactoryManager());
-    MPredict->SetFactory("A",                 A00Fact);
-    MPredict->SetFactory("Smoother",          SmooPredictFact);    // solver/smoother for correction step
-    MPredict->SetFactory("PreSmoother",               SmooPredictFact);
-    MPredict->SetFactory("PostSmoother",              SmooPredictFact);
-    MPredict->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-    SimpleSm->SetVelocityPredictionFactoryManager(MPredict);    // set temporary factory manager in BraessSarazin smoother
-
+    MPredict->SetFactory("A", A00Fact);
+    MPredict->SetFactory("Smoother", SmooPredictFact);  // solver/smoother for correction step
+    MPredict->SetFactory("PreSmoother", SmooPredictFact);
+    MPredict->SetFactory("PostSmoother", SmooPredictFact);
+    MPredict->SetIgnoreUserData(true);                        // always use data from factories defined in factory manager
+    SimpleSm->SetVelocityPredictionFactoryManager(MPredict);  // set temporary factory manager in BraessSarazin smoother
 
     // define SchurComplement Factory
     // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
@@ -312,59 +308,59 @@ int main(int argc, char *argv[]) {
     // InverseApproximation
     Teuchos::RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
     AinvFact->SetFactory("A", A00Fact);
-    if(useSIMPLEC == 1) AinvFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
-    
+    if (useSIMPLEC == 1) AinvFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
+
     RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-    SFact->SetParameter("omega", Teuchos::ParameterEntry(1.0)); // for Simple, omega is always 1.0 in the SchurComplement
-    SFact->SetFactory("A",MueLu::NoFactory::getRCP());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(1.0));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());
 
     // define smoother/solver for BraessSarazin
     RCP<SmootherPrototype> smoProtoSC = Teuchos::null;
-    if(SC_bUseDirectSolver != 1) {
-      //Smoother Factory, using SFact as a factory for A
+    if (SC_bUseDirectSolver != 1) {
+      // Smoother Factory, using SFact as a factory for A
       std::string ifpackSCType;
       Teuchos::ParameterList ifpackSCList;
-      ifpackSCList.set("relaxation: sweeps", SC_nSweeps );
-      ifpackSCList.set("relaxation: damping factor", SC_omega );
+      ifpackSCList.set("relaxation: sweeps", SC_nSweeps);
+      ifpackSCList.set("relaxation: damping factor", SC_omega);
       ifpackSCType = "RELAXATION";
       ifpackSCList.set("relaxation: type", "Gauss-Seidel");
-      smoProtoSC     = rcp( new TrilinosSmoother(ifpackSCType, ifpackSCList, 0) );
-      smoProtoSC->SetFactory("A",SFact);
-    }
-    else {
+      smoProtoSC = rcp(new TrilinosSmoother(ifpackSCType, ifpackSCList, 0));
+      smoProtoSC->SetFactory("A", SFact);
+    } else {
       Teuchos::ParameterList ifpackDSList;
       std::string ifpackDSType;
-      smoProtoSC     = rcp( new DirectSolver(ifpackDSType,ifpackDSList) ); smoProtoSC->SetFactory("A", SFact);
+      smoProtoSC = rcp(new DirectSolver(ifpackDSType, ifpackDSList));
+      smoProtoSC->SetFactory("A", SFact);
     }
 
-    RCP<SmootherFactory> SmooSCFact = rcp( new SmootherFactory(smoProtoSC) );
+    RCP<SmootherFactory> SmooSCFact = rcp(new SmootherFactory(smoProtoSC));
 
     // define temporary FactoryManager that is used as input for BraessSarazin smoother
     RCP<FactoryManager> MB = rcp(new FactoryManager());
-    MB->SetFactory("A",                 SFact);         // SchurComplement operator for correction step (defined as "A")
-    MB->SetFactory("Smoother",          SmooSCFact);    // solver/smoother for correction step
-    MB->SetFactory("PreSmoother",               SmooSCFact);
-    MB->SetFactory("PostSmoother",              SmooSCFact);
+    MB->SetFactory("A", SFact);              // SchurComplement operator for correction step (defined as "A")
+    MB->SetFactory("Smoother", SmooSCFact);  // solver/smoother for correction step
+    MB->SetFactory("PreSmoother", SmooSCFact);
+    MB->SetFactory("PostSmoother", SmooSCFact);
     MB->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-    SimpleSm->SetSchurCompFactoryManager(MB);    // set temporary factory manager in BraessSarazin smoother
+    SimpleSm->SetSchurCompFactoryManager(MB);  // set temporary factory manager in BraessSarazin smoother
 
     // setup main factory manager
     RCP<FactoryManager> M = rcp(new FactoryManager());
-    M->SetFactory("A",               MueLu::NoFactory::getRCP()); // this is the 2x2 blocked operator
-    M->SetFactory("Smoother",        smootherFact);               // BraessSarazin block smoother
-    M->SetFactory("PreSmoother",     smootherFact);
-    M->SetFactory("PostSmoother",    smootherFact);
+    M->SetFactory("A", MueLu::NoFactory::getRCP());  // this is the 2x2 blocked operator
+    M->SetFactory("Smoother", smootherFact);         // BraessSarazin block smoother
+    M->SetFactory("PreSmoother", smootherFact);
+    M->SetFactory("PostSmoother", smootherFact);
 
     MueLu::SetFactoryManager SFMCoarse(Finest, M);
-    Finest->Request(MueLu::TopSmootherFactory<Scalar,LO,GO,Node>(M, "Smoother"));
+    Finest->Request(MueLu::TopSmootherFactory<Scalar, LO, GO, Node>(M, "Smoother"));
 
     // call setup (= extract blocks and extract diagonal of F)
     SimpleSm->Setup(*Finest);
 
-    RCP<MultiVector> xtest = MultiVectorFactory::Build(xstridedfullmap,1);
-    xtest->putScalar( (Scalar) 0.0);
+    RCP<MultiVector> xtest = MultiVectorFactory::Build(xstridedfullmap, 1);
+    xtest->putScalar((Scalar)0.0);
 
-    RCP<Vector> xR = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(epv));
+    RCP<Vector> xR = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(epv));
     // calculate initial (absolute) residual
     Teuchos::Array<Teuchos::ScalarTraits<Scalar>::magnitudeType> norms(1);
 
@@ -373,7 +369,7 @@ int main(int argc, char *argv[]) {
     *out << "Test: Applying Simple Smoother" << std::endl;
     *out << "Test: START DATA" << std::endl;
     *out << "iterations\tVelocity_residual\tPressure_residual" << std::endl;
-    SimpleSm->Apply(*xtest,*xR);
+    SimpleSm->Apply(*xtest, *xR);
     xtest->norm2(norms);
     *out << "Test: ||x_1|| = " << norms[0] << std::endl;
 
@@ -386,12 +382,9 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra needs Serial node. Please recompile MueLu with the Serial node enabled." << std::endl;
   return EXIT_SUCCESS;
-#endif // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
+#endif  // #if defined(HAVE_MUELU_SERIAL) && defined(HAVE_MUELU_EPETRA)
 }
-
-
-
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_test2.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_test2.cpp
index 55ab0056cf20..03959fc2fc7e 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_test2.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_test2.cpp
@@ -134,8 +134,7 @@
  *
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -145,20 +144,19 @@ int main(int argc, char *argv[]) {
   typedef Xpetra::EpetraNode Node;
 #include "MueLu_UseShortNames.hpp"
 
-
   using Teuchos::RCP;
   using Teuchos::rcp;
   using Teuchos::rcpFromRef;
   using namespace MueLuTests;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -169,28 +167,27 @@ int main(int argc, char *argv[]) {
     // read in some command line parameters
     Teuchos::CommandLineProcessor clp(false);
 
-    int rebalanceBlocks = 1;      clp.setOption("rebalanceBlocks",       &rebalanceBlocks,     "rebalance blocks (1=yes, else=no)");
+    int rebalanceBlocks = 1;
+    clp.setOption("rebalanceBlocks", &rebalanceBlocks, "rebalance blocks (1=yes, else=no)");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-
 #if defined(HAVE_MPI) && defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MUELU_ISORROPIA)
 #ifndef HAVE_XPETRA_INT_LONG_LONG
     *out << "Warning: scaling test was not compiled with long long int support" << std::endl;
 
-
     // custom parameters
     LocalOrdinal maxLevels = 3;
 
-    GlobalOrdinal maxCoarseSize=1; //FIXME clp doesn't like long long int
+    GlobalOrdinal maxCoarseSize = 1;  // FIXME clp doesn't like long long int
 
     int globalNumDofs = 1500;  // used for the maps
-    int nDofsPerNode = 3;      // used for generating the fine level null-space
+    int nDofsPerNode  = 3;     // used for generating the fine level null-space
 
     // build strided maps
     // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -203,36 +200,35 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib             = Xpetra::UseEpetra;
+    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for AztecOO
     const RCP<const Epetra_Map> fullmap = rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
 
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
-    //EpetraExt::MatrixMarketFileToCrsMatrix("/home/tobias/promotion/trilinos/fc17-dyn/packages/muelu/test/navierstokes/A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    //EpetraExt::MatrixMarketFileToVector("/home/tobias/promotion/trilinos/fc17-dyn/packages/muelu/test/navierstokes/b_re1000_5932.txt",*fullmap,ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
+    // EpetraExt::MatrixMarketFileToCrsMatrix("/home/tobias/promotion/trilinos/fc17-dyn/packages/muelu/test/navierstokes/A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
+    // EpetraExt::MatrixMarketFileToVector("/home/tobias/promotion/trilinos/fc17-dyn/packages/muelu/test/navierstokes/b_re1000_5932.txt",*fullmap,ptrf);
 
-    RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    RCP<Epetra_Vector> epv = Teuchos::rcp(ptrf);
+    RCP<Epetra_CrsMatrix> epA    = Teuchos::rcp(ptrA);
+    RCP<Epetra_Vector> epv       = Teuchos::rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = Teuchos::rcp(ptrNS);
 
-
     /////////////////////////////////////// split system into 2x2 block system
 
     *out << "Split matrix into 2x2 block matrix" << std::endl;
@@ -243,69 +239,67 @@ int main(int argc, char *argv[]) {
     Teuchos::RCP<Epetra_CrsMatrix> A21;
     Teuchos::RCP<Epetra_CrsMatrix> A22;
 
-    if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A11));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A12));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A21));
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node>(A22));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA11 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A11));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA12 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A12));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA21 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A21));
+    Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > xA22 = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > > xmaps;
+    std::vector<Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > > xmaps;
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(xstridedfullmap,xmaps);
+    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(xstridedfullmap, xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
 
     //////////////////////////////////////////////////// create Hierarchy
-    RCP<Hierarchy> H = rcp ( new Hierarchy() );
+    RCP<Hierarchy> H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    //H->setDefaultVerbLevel(Teuchos::VERB_NONE);
+    // H->setDefaultVerbLevel(Teuchos::VERB_NONE);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     //////////////////////////////////////////////////////// finest Level
     RCP<MueLu::Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Teuchos::rcp_dynamic_cast<Matrix>(bOp));
-
+    Finest->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));
 
     ////////////////////////////////////////// prepare null space for A11
     RCP<MultiVector> nullspace11 = MultiVectorFactory::Build(xstridedvelmap, 2);  // this is a 2D standard null space
 
-    for (int i=0; i<nDofsPerNode-1; ++i) {
+    for (int i = 0; i < nDofsPerNode - 1; ++i) {
       Teuchos::ArrayRCP<Scalar> nsValues = nullspace11->getDataNonConst(i);
-      int numBlocks = nsValues.size() / (nDofsPerNode - 1);
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*(nDofsPerNode - 1) + i] = 1.0;
+      int numBlocks                      = nsValues.size() / (nDofsPerNode - 1);
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * (nDofsPerNode - 1) + i] = 1.0;
       }
     }
 
-    Finest->Set("Nullspace1",nullspace11);
+    Finest->Set("Nullspace1", nullspace11);
 
     ////////////////////////////////////////// prepare null space for A22
-    RCP<MultiVector> nullspace22 = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
+    RCP<MultiVector> nullspace22         = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
     Teuchos::ArrayRCP<Scalar> nsValues22 = nullspace22->getDataNonConst(0);
-    for (int j=0; j< nsValues22.size(); ++j) {
+    for (int j = 0; j < nsValues22.size(); ++j) {
       nsValues22[j] = 1.0;
     }
 
-    Finest->Set("Nullspace2",nullspace22);
-
+    Finest->Set("Nullspace2", nullspace22);
 
     /////////////////////////////////////////// define rebalanced block AC factory
     // This is the main factory for "A" and defines the input for
@@ -314,7 +308,7 @@ int main(int argc, char *argv[]) {
     RCP<RebalanceBlockAcFactory> RebalancedAcFact = rcp(new RebalanceBlockAcFactory());
 
     /////////////////////////////////////////// define non-rebalanced blocked transfer ops
-    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory()); // use row map index base from bOp
+    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());  // use row map index base from bOp
     RCP<GenericRFactory> RFact = rcp(new GenericRFactory());
     RFact->SetFactory("P", PFact);
 
@@ -332,31 +326,30 @@ int main(int argc, char *argv[]) {
       Teuchos::ParameterList paramList;
       paramList.set("repartition: min rows per proc", 200);
       paramList.set("repartition: max imbalance", 1.3);
-      if(rebalanceBlocks == 1)
-        paramList.set("repartition: start level",1);
+      if (rebalanceBlocks == 1)
+        paramList.set("repartition: start level", 1);
       else
-        paramList.set("repartition: start level",10); // supress rebalancing
+        paramList.set("repartition: start level", 10);  // supress rebalancing
       RepartitionHeuristicFact->SetParameterList(paramList);
     }
     RepartitionHeuristicFact->SetFactory("A", AcFact);
 
-
     // define matrix sub-blocks of possibly rebalanced block matrix A
     // These are used as input for
     //   - the sub blocks of the transfer operators
     RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
     /*A11Fact->SetParameter("Range map: Striding info",Teuchos::ParameterEntry(std::string("{ 2 1 }")));
     A11Fact->SetParameter("Range map: Strided block id",Teuchos::ParameterEntry(0));
     A11Fact->SetParameter("Domain map: Striding info",Teuchos::ParameterEntry(std::string("{ 2 1 }")));
     A11Fact->SetParameter("Domain map: Strided block id",Teuchos::ParameterEntry(0));*/
 
     RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+    A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+    A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
     /*A22Fact->SetParameter("Range map: Striding info",Teuchos::ParameterEntry(std::string("{ 2 1 }")));
     A22Fact->SetParameter("Range map: Strided block id",Teuchos::ParameterEntry(1));
     A22Fact->SetParameter("Domain map: Striding info",Teuchos::ParameterEntry(std::string("{ 2 1 }")));
@@ -366,22 +359,22 @@ int main(int argc, char *argv[]) {
     // define sub blocks of the coarse non-rebalanced block matrix Ac
     // input is the block operator generated by AcFact
     RCP<SubBlockAFactory> rebA11Fact = Teuchos::rcp(new SubBlockAFactory());
-    rebA11Fact->SetFactory("A",AcFact);
-    rebA11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    rebA11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
-    rebA11Fact->SetParameter("Range map: Striding info",Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
-    rebA11Fact->SetParameter("Range map: Strided block id",Teuchos::ParameterEntry(0));
-    rebA11Fact->SetParameter("Domain map: Striding info",Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
-    rebA11Fact->SetParameter("Domain map: Strided block id",Teuchos::ParameterEntry(0));
+    rebA11Fact->SetFactory("A", AcFact);
+    rebA11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+    rebA11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+    rebA11Fact->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
+    rebA11Fact->SetParameter("Range map: Strided block id", Teuchos::ParameterEntry(0));
+    rebA11Fact->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
+    rebA11Fact->SetParameter("Domain map: Strided block id", Teuchos::ParameterEntry(0));
 
     RCP<SubBlockAFactory> rebA22Fact = Teuchos::rcp(new SubBlockAFactory());
-    rebA22Fact->SetFactory("A",AcFact);
-    rebA22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    rebA22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
-    rebA22Fact->SetParameter("Range map: Striding info",Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
-    rebA22Fact->SetParameter("Range map: Strided block id",Teuchos::ParameterEntry(1));
-    rebA22Fact->SetParameter("Domain map: Striding info",Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
-    rebA22Fact->SetParameter("Domain map: Strided block id",Teuchos::ParameterEntry(1));
+    rebA22Fact->SetFactory("A", AcFact);
+    rebA22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+    rebA22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+    rebA22Fact->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
+    rebA22Fact->SetParameter("Range map: Strided block id", Teuchos::ParameterEntry(1));
+    rebA22Fact->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(std::string("{ 2, 1 }")));
+    rebA22Fact->SetParameter("Domain map: Strided block id", Teuchos::ParameterEntry(1));
 
     // define rebalancing factory for coarse block matrix A(1,1)
     RCP<AmalgamationFactory> rebAmalgFact11 = rcp(new AmalgamationFactory());
@@ -420,11 +413,10 @@ int main(int argc, char *argv[]) {
     RCP<Factory> RepartitionFact2 = rcp(new RepartitionFactory());
     RepartitionFact2->SetFactory("A", rebA22Fact);
     RepartitionFact2->SetFactory("number of partitions", RepartitionHeuristicFact);
-    RepartitionFact2->SetFactory("Partition", repInterface2); // this is not valid
+    RepartitionFact2->SetFactory("Partition", repInterface2);  // this is not valid
     RepartitionFact2->SetParameter("repartition: print partition distribution", Teuchos::ParameterEntry(true));
     RepartitionFact2->SetParameter("repartition: remap parts", Teuchos::ParameterEntry(false)); /* do not remap! */
 
-
     ////////////////////////////////////////// build non-rebalanced matrix blocks
     // build factories for transfer operator P(1,1) and R(1,1)
     RCP<AmalgamationFactory> amalgFact11 = rcp(new AmalgamationFactory());
@@ -447,10 +439,10 @@ int main(int argc, char *argv[]) {
     coarseMapFact11->setStridedBlockId(0);
 
     RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
-    RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
+    RCP<TransPFactory> R11Fact     = rcp(new TransPFactory());
 
     Teuchos::RCP<NullspaceFactory> nspFact11 = Teuchos::rcp(new NullspaceFactory("Nullspace1"));
-    nspFact11->SetFactory("Nullspace1",P11Fact); // pick "Nullspace1" from Finest level
+    nspFact11->SetFactory("Nullspace1", P11Fact);  // pick "Nullspace1" from Finest level
 
     //////////////////////////////// define factory manager for (1,1) block
     RCP<FactoryManager> M11 = rcp(new FactoryManager());
@@ -461,19 +453,19 @@ int main(int argc, char *argv[]) {
     M11->SetFactory("Graph", dropFact11);
     M11->SetFactory("DofsPerNode", dropFact11);
     M11->SetFactory("UnAmalgamationInfo", amalgFact11);
-    M11->SetFactory("Nullspace", nspFact11); // TODO check me?
+    M11->SetFactory("Nullspace", nspFact11);  // TODO check me?
     M11->SetFactory("CoarseMap", coarseMapFact11);
-    M11->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
+    M11->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
 
     ////////////////////////////////////////// build non-rebalanced matrix blocks
     // build factories for transfer operator P(2,2) and R(2,2)
     RCP<AmalgamationFactory> amalgFact22 = rcp(new AmalgamationFactory());
-    RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory());
-    RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
+    RCP<TentativePFactory> P22Fact       = rcp(new TentativePFactory());
+    RCP<TransPFactory> R22Fact           = rcp(new TransPFactory());
 
     // connect null space and tentative PFactory
     Teuchos::RCP<NullspaceFactory> nspFact22 = Teuchos::rcp(new NullspaceFactory("Nullspace2"));
-    nspFact22->SetFactory("Nullspace2", P22Fact); // define null space generated by P22Fact as null space for coarse level (non-rebalanced)
+    nspFact22->SetFactory("Nullspace2", P22Fact);  // define null space generated by P22Fact as null space for coarse level (non-rebalanced)
 
     RCP<CoarseMapFactory> coarseMapFact22 = Teuchos::rcp(new CoarseMapFactory());
     coarseMapFact22->setStridingData(stridingInfo);
@@ -481,10 +473,10 @@ int main(int argc, char *argv[]) {
 
     //////////////////////////////// define factory manager for (2,2) block
     RCP<FactoryManager> M22 = rcp(new FactoryManager());
-    M22->SetFactory("A", A22Fact); // rebalanced fine-level block operator
-    M22->SetFactory("P", P22Fact); // non-rebalanced transfer operator P(2,2)
-    M22->SetFactory("R", R22Fact); // non-rebalanced transfer operator R(2,2)
-    M22->SetFactory("Aggregates", UncoupledAggFact11); // aggregates from block (1,1)
+    M22->SetFactory("A", A22Fact);                      // rebalanced fine-level block operator
+    M22->SetFactory("P", P22Fact);                      // non-rebalanced transfer operator P(2,2)
+    M22->SetFactory("R", R22Fact);                      // non-rebalanced transfer operator R(2,2)
+    M22->SetFactory("Aggregates", UncoupledAggFact11);  // aggregates from block (1,1)
     M22->SetFactory("Nullspace", nspFact22);
     M22->SetFactory("UnAmalgamationInfo", amalgFact22);
     M22->SetFactory("Ptent", P22Fact);
@@ -494,28 +486,28 @@ int main(int argc, char *argv[]) {
     /////////////////////////////////////////// define rebalanced blocked transfer ops
     //////////////////////////////// define factory manager for (1,1) block
     RCP<FactoryManager> rebM11 = rcp(new FactoryManager());
-    rebM11->SetFactory("A", AcFact ); // important: must be a 2x2 block A Factory
+    rebM11->SetFactory("A", AcFact);  // important: must be a 2x2 block A Factory
     rebM11->SetFactory("Importer", RepartitionFact);
     rebM11->SetFactory("number of partitions", RepartitionHeuristicFact);
     rebM11->SetFactory("Nullspace", nspFact11);
-    //rebM11->SetIgnoreUserData(true);
+    // rebM11->SetIgnoreUserData(true);
 
     RCP<FactoryManager> rebM22 = rcp(new FactoryManager());
-    rebM22->SetFactory("A", AcFact ); // important: must be a 2x2 block A Factory
-    rebM22->SetFactory("Importer", RepartitionFact2); // use dummy repartitioning factory
+    rebM22->SetFactory("A", AcFact);                   // important: must be a 2x2 block A Factory
+    rebM22->SetFactory("Importer", RepartitionFact2);  // use dummy repartitioning factory
     rebM22->SetFactory("number of partitions", RepartitionHeuristicFact);
     rebM22->SetFactory("Nullspace", nspFact22);
 
     // Reordering of the transfer operators
     RCP<RebalanceBlockInterpolationFactory> RebalancedBlockPFact = rcp(new RebalanceBlockInterpolationFactory());
-    RebalancedBlockPFact->SetFactory("P", PFact); // use non-rebalanced block P operator as input
+    RebalancedBlockPFact->SetFactory("P", PFact);  // use non-rebalanced block P operator as input
     RebalancedBlockPFact->AddFactoryManager(rebM11);
     RebalancedBlockPFact->AddFactoryManager(rebM22);
 
     RCP<RebalanceBlockRestrictionFactory> RebalancedBlockRFact = rcp(new RebalanceBlockRestrictionFactory());
-    //RebalancedBlockRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
-    RebalancedBlockRFact->SetFactory("R", RFact); // non-rebalanced block P operator
-    RebalancedBlockRFact->SetParameter("repartition: use subcommunicators",Teuchos::ParameterEntry(true));
+    // RebalancedBlockRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
+    RebalancedBlockRFact->SetFactory("R", RFact);  // non-rebalanced block P operator
+    RebalancedBlockRFact->SetParameter("repartition: use subcommunicators", Teuchos::ParameterEntry(true));
     RebalancedBlockRFact->AddFactoryManager(rebM11);
     RebalancedBlockRFact->AddFactoryManager(rebM22);
 
@@ -526,63 +518,62 @@ int main(int argc, char *argv[]) {
     PFact->AddFactoryManager(M22);  // use non-rebalanced information from sub block factory manager M22
 
     ///////////////////////////////////////// initialize rebalanced coarse block AC factory
-    RebalancedAcFact->SetFactory("A", AcFact);   // use non-rebalanced block operator as input
-    RebalancedAcFact->SetParameter("repartition: use subcommunicators",Teuchos::ParameterEntry(true));
+    RebalancedAcFact->SetFactory("A", AcFact);  // use non-rebalanced block operator as input
+    RebalancedAcFact->SetParameter("repartition: use subcommunicators", Teuchos::ParameterEntry(true));
     RebalancedAcFact->AddFactoryManager(rebM11);
     RebalancedAcFact->AddFactoryManager(rebM22);
 
     //////////////////////////////////////////////////////////////////////
     // Smoothers
 
-    //Another factory manager for braes sarazin smoother
-    //Schur Complement Factory, using the factory to generate AcFact
-    SC omega = 1.3;
+    // Another factory manager for braes sarazin smoother
+    // Schur Complement Factory, using the factory to generate AcFact
+    SC omega                          = 1.3;
     RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
     SFact->SetParameter("omega", Teuchos::ParameterEntry(omega));
-    SFact->SetFactory("A", MueLu::NoFactory::getRCP()); // this finally be the rebalanced block operator!
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());  // this finally be the rebalanced block operator!
 
-    //Smoother Factory, using SFact as a factory for A
+    // Smoother Factory, using SFact as a factory for A
     std::string ifpackSCType;
     Teuchos::ParameterList ifpackSCList;
-    ifpackSCList.set("relaxation: sweeps", (LocalOrdinal) 3);
-    ifpackSCList.set("relaxation: damping factor", (Scalar) 1.0);
+    ifpackSCList.set("relaxation: sweeps", (LocalOrdinal)3);
+    ifpackSCList.set("relaxation: damping factor", (Scalar)1.0);
     ifpackSCType = "RELAXATION";
     ifpackSCList.set("relaxation: type", "Gauss-Seidel");
-    RCP<SmootherPrototype> smoProtoSC     = rcp( new TrilinosSmoother(ifpackSCType, ifpackSCList, 0) );
+    RCP<SmootherPrototype> smoProtoSC = rcp(new TrilinosSmoother(ifpackSCType, ifpackSCList, 0));
     smoProtoSC->SetFactory("A", SFact);
-    RCP<SmootherFactory> SmooSCFact = rcp( new SmootherFactory(smoProtoSC) );
+    RCP<SmootherFactory> SmooSCFact = rcp(new SmootherFactory(smoProtoSC));
 
-    RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
     smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(3));
     smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(omega));
-    smootherPrototype->SetFactory("A",MueLu::NoFactory::getRCP());
-    RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->SetFactory("A", MueLu::NoFactory::getRCP());
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-    RCP<BraessSarazinSmoother> coarseSolverPrototype = rcp( new BraessSarazinSmoother() );
+    RCP<BraessSarazinSmoother> coarseSolverPrototype = rcp(new BraessSarazinSmoother());
     coarseSolverPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(3));
     coarseSolverPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(omega));
-    coarseSolverPrototype->SetFactory("A",MueLu::NoFactory::getRCP());
-    RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, Teuchos::null) );
+    coarseSolverPrototype->SetFactory("A", MueLu::NoFactory::getRCP());
+    RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, Teuchos::null));
 
     RCP<FactoryManager> MB = rcp(new FactoryManager());
-    MB->SetFactory("A",     SFact);
-    MB->SetFactory("Smoother",    SmooSCFact);
-    MB->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-    smootherPrototype->AddFactoryManager(MB,0);
-    coarseSolverPrototype->AddFactoryManager(MB,0);
-
+    MB->SetFactory("A", SFact);
+    MB->SetFactory("Smoother", SmooSCFact);
+    MB->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+    smootherPrototype->AddFactoryManager(MB, 0);
+    coarseSolverPrototype->AddFactoryManager(MB, 0);
 
     ////////////////////////////////////////// define main factory manager
     FactoryManager M;
-    M.SetFactory("A",            RebalancedAcFact);     // rebalance block AC Factory using importer
-    M.SetFactory("P",            RebalancedBlockPFact); // rebalance prolongator using non-balanced Ac
-    M.SetFactory("R",            RebalancedBlockRFact); // rebalance restrictor and null space using non-balanced Ac
-    M.SetFactory("Smoother",     smootherFact);
-    M.SetFactory("PreSmoother",     smootherFact);
-    M.SetFactory("PostSmoother",     smootherFact);
+    M.SetFactory("A", RebalancedAcFact);      // rebalance block AC Factory using importer
+    M.SetFactory("P", RebalancedBlockPFact);  // rebalance prolongator using non-balanced Ac
+    M.SetFactory("R", RebalancedBlockRFact);  // rebalance restrictor and null space using non-balanced Ac
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("PreSmoother", smootherFact);
+    M.SetFactory("PostSmoother", smootherFact);
     M.SetFactory("CoarseSolver", coarseSolverFact);
 
-    H->Setup(M,0,maxLevels);
+    H->Setup(M, 0, maxLevels);
 
     /**out << std::endl;
      *out << "print content of multigrid levels:" << std::endl;
@@ -595,7 +586,7 @@ int main(int argc, char *argv[]) {
      RCP<Level> coarseLevel2 = H->GetLevel(2);
      coarseLevel2->print(*out);*/
 
-    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap,1);
+    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap, 1);
 
     // Use AMG directly as an iterative method
 #if 0
@@ -643,13 +634,13 @@ int main(int argc, char *argv[]) {
       aztecSolver.Iterate(maxIts, tol);
     }
 
-#endif // end ifndef HAVE_LONG_LONG_INT
-#endif // #if defined(HAVE_MPI) && defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MUELU_ISORROPIA)
+#endif  // end ifndef HAVE_LONG_LONG_INT
+#endif  // #if defined(HAVE_MPI) && defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MUELU_ISORROPIA)
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
diff --git a/packages/muelu/test/navierstokes/Navier2DBlocked_xml.cpp b/packages/muelu/test/navierstokes/Navier2DBlocked_xml.cpp
index 8f6aabead1c2..7053e3a10d7f 100644
--- a/packages/muelu/test/navierstokes/Navier2DBlocked_xml.cpp
+++ b/packages/muelu/test/navierstokes/Navier2DBlocked_xml.cpp
@@ -121,8 +121,7 @@
  *  using block matrices
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -138,7 +137,7 @@ int main(int argc, char *argv[]) {
   using namespace Teuchos;
 
   oblackholestream blackhole;
-  GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
@@ -150,15 +149,15 @@ int main(int argc, char *argv[]) {
     CommandLineProcessor clp(false);
     clp.setOption("xml", &xmlFile, "xml file with solver parameters for a 2x2 blocked NS example");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     RCP<const Comm<int> > comm = DefaultComm<int>::getComm();
-    RCP<FancyOStream> out = fancyOStream(rcpFromRef(std::cout));
+    RCP<FancyOStream> out      = fancyOStream(rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -166,10 +165,10 @@ int main(int argc, char *argv[]) {
     Time myTime("global");
     TimeMonitor MM(myTime);
 
-    GO maxCoarseSize=1; //FIXME clp doesn't like long long int
+    GO maxCoarseSize = 1;  // FIXME clp doesn't like long long int
 
     int globalNumDofs = 1500;  // used for the maps
-    int nDofsPerNode = 3;            // used for generating the fine level null-space
+    int nDofsPerNode  = 3;     // used for generating the fine level null-space
 
     // build strided maps
     // striding information: 2 velocity dofs and 1 pressure dof = 3 dofs per node
@@ -182,29 +181,29 @@ int main(int argc, char *argv[]) {
     // xstridedfullmap: full map (velocity and pressure dof gids), continous
     // xstridedvelmap: only velocity dof gid maps (i.e. 0,1,3,4,6,7...)
     // xstridedpremap: only pressure dof gid maps (i.e. 2,5,8,...)
-    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib,globalNumDofs,0,stridingInfo,comm,-1);
-    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap,0);
-    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap,1);
+    Xpetra::UnderlyingLib lib             = Xpetra::UseEpetra;
+    RCP<const StridedMap> xstridedfullmap = StridedMapFactory::Build(lib, globalNumDofs, 0, stridingInfo, comm, -1);
+    RCP<const StridedMap> xstridedvelmap  = StridedMapFactory::Build(xstridedfullmap, 0);
+    RCP<const StridedMap> xstridedpremap  = StridedMapFactory::Build(xstridedfullmap, 1);
 
     /////////////////////////////////////// transform Xpetra::Map objects to Epetra
     // this is needed for AztecOO
     const RCP<const Epetra_Map> fullmap = rcpFromRef(Xpetra::toEpetra(*xstridedfullmap));
-    RCP<const Epetra_Map>       velmap  = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
-    RCP<const Epetra_Map>       premap  = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
+    RCP<const Epetra_Map> velmap        = rcpFromRef(Xpetra::toEpetra(*xstridedvelmap));
+    RCP<const Epetra_Map> premap        = rcpFromRef(Xpetra::toEpetra(*xstridedpremap));
 
     /////////////////////////////////////// import problem matrix and RHS from files (-> Epetra)
 
     // read in problem
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     *out << "Reading matrix market file" << std::endl;
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",*fullmap,*fullmap,*fullmap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",*fullmap,ptrf);
-    RCP<Epetra_CrsMatrix> epA = rcp(ptrA);
-    RCP<Epetra_Vector> epv = rcp(ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", *fullmap, *fullmap, *fullmap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", *fullmap, ptrf);
+    RCP<Epetra_CrsMatrix> epA    = rcp(ptrA);
+    RCP<Epetra_Vector> epv       = rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = rcp(ptrNS);
 
     /////////////////////////////////////// split system into 2x2 block system
@@ -217,69 +216,67 @@ int main(int argc, char *argv[]) {
     RCP<Epetra_CrsMatrix> A21;
     RCP<Epetra_CrsMatrix> A22;
 
-    if(SplitMatrix2x2(epA,*velmap,*premap,A11,A12,A21,A22)==false)
-      *out << "Problem with splitting matrix"<< std::endl;
+    if (SplitMatrix2x2(epA, *velmap, *premap, A11, A12, A21, A22) == false)
+      *out << "Problem with splitting matrix" << std::endl;
 
     /////////////////////////////////////// transform Epetra objects to Xpetra (needed for MueLu)
 
     // build Xpetra objects from Epetra_CrsMatrix objects
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A11));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A12));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A21));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO,Node>(A22));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA11 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A11));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA12 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A12));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA21 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A21));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > xA22 = rcp(new Xpetra::EpetraCrsMatrixT<GO, Node>(A22));
 
     /////////////////////////////////////// generate MapExtractor object
 
-    std::vector<RCP<const Xpetra::Map<LO,GO,Node> > > xmaps;
+    std::vector<RCP<const Xpetra::Map<LO, GO, Node> > > xmaps;
     xmaps.push_back(xstridedvelmap);
     xmaps.push_back(xstridedpremap);
 
-    RCP<const Xpetra::MapExtractor<Scalar,LO,GO,Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(xstridedfullmap->getMap(),xmaps);
+    RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > map_extractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(xstridedfullmap->getMap(), xmaps);
 
     /////////////////////////////////////// build blocked transfer operator
     // using the map extractor
-    RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(map_extractor,map_extractor,10));
-    bOp->setMatrix(0,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA11)));
-    bOp->setMatrix(0,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA12)));
-    bOp->setMatrix(1,0,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA21)));
-    bOp->setMatrix(1,1,Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(xA22)));
+    RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(map_extractor, map_extractor, 10));
+    bOp->setMatrix(0, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA11)));
+    bOp->setMatrix(0, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA12)));
+    bOp->setMatrix(1, 0, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA21)));
+    bOp->setMatrix(1, 1, Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(xA22)));
 
     bOp->fillComplete();
 
     //////////////////////////////////////// prepare setup
     ParameterListInterpreter mueLuFactory(xmlFile, *comm);
 
-
     RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
     H->setDefaultVerbLevel(VERB_HIGH);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     RCP<MueLu::Level> Finest = H->GetLevel(0);
     Finest->setDefaultVerbLevel(VERB_HIGH);
-    Finest->Set("A",           rcp_dynamic_cast<Matrix>(bOp));
-
+    Finest->Set("A", rcp_dynamic_cast<Matrix>(bOp));
 
     ////////////////////////////////////////// prepare null space for A11
     RCP<MultiVector> nullspace11 = MultiVectorFactory::Build(xstridedvelmap, 2);  // this is a 2D standard null space
 
-    for (int i=0; i<nDofsPerNode-1; ++i) {
+    for (int i = 0; i < nDofsPerNode - 1; ++i) {
       ArrayRCP<Scalar> nsValues = nullspace11->getDataNonConst(i);
-      int numBlocks = nsValues.size() / (nDofsPerNode - 1);
-      for (int j=0; j< numBlocks; ++j) {
-        nsValues[j*(nDofsPerNode - 1) + i] = 1.0;
+      int numBlocks             = nsValues.size() / (nDofsPerNode - 1);
+      for (int j = 0; j < numBlocks; ++j) {
+        nsValues[j * (nDofsPerNode - 1) + i] = 1.0;
       }
     }
 
-    Finest->Set("Nullspace1",nullspace11);
+    Finest->Set("Nullspace1", nullspace11);
 
     ////////////////////////////////////////// prepare null space for A22
     RCP<MultiVector> nullspace22 = MultiVectorFactory::Build(xstridedpremap, 1);  // this is a 2D standard null space
-    ArrayRCP<Scalar> nsValues22 = nullspace22->getDataNonConst(0);
-    for (int j=0; j< nsValues22.size(); ++j) {
+    ArrayRCP<Scalar> nsValues22  = nullspace22->getDataNonConst(0);
+    for (int j = 0; j < nsValues22.size(); ++j) {
       nsValues22[j] = 1.0;
     }
 
-    Finest->Set("Nullspace2",nullspace22);
+    Finest->Set("Nullspace2", nullspace22);
 
     /////////////////////////////////// BEGIN setup
 
@@ -289,14 +286,14 @@ int main(int argc, char *argv[]) {
 
     *out << std::endl;
 
-    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap,1);
+    RCP<MultiVector> xLsg = MultiVectorFactory::Build(xstridedfullmap, 1);
 
     // Use AMG directly as an iterative method
     {
-      xLsg->putScalar( (SC) 0.0);
+      xLsg->putScalar((SC)0.0);
 
       // Epetra_Vector -> Xpetra::Vector
-      RCP<Vector> xRhs = rcp(new Xpetra::EpetraVectorT<int,Node>(epv));
+      RCP<Vector> xRhs = rcp(new Xpetra::EpetraVectorT<int, Node>(epv));
 
       // calculate initial (absolute) residual
       Array<ScalarTraits<SC>::magnitudeType> norms(1);
@@ -304,15 +301,14 @@ int main(int argc, char *argv[]) {
       *out << "||x_0|| = " << norms[0] << std::endl;
 
       // apply ten multigrid iterations
-      H->Iterate(*xRhs,*xLsg,100);
+      H->Iterate(*xRhs, *xLsg, 100);
 
       // calculate and print residual
-      RCP<MultiVector> xTmp = MultiVectorFactory::Build(xstridedfullmap,1);
-      bOp->apply(*xLsg,*xTmp,NO_TRANS,(SC)1.0,(SC)0.0);
-      xRhs->update((SC)-1.0,*xTmp,(SC)1.0);
+      RCP<MultiVector> xTmp = MultiVectorFactory::Build(xstridedfullmap, 1);
+      bOp->apply(*xLsg, *xTmp, NO_TRANS, (SC)1.0, (SC)0.0);
+      xRhs->update((SC)-1.0, *xTmp, (SC)1.0);
       xRhs->norm2(norms);
       *out << "||r|| = " << norms[0] << std::endl;
-
     }
 
     // TODO: don't forget to add Aztec as prerequisite in CMakeLists.txt!
@@ -340,7 +336,7 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
diff --git a/packages/muelu/test/navierstokes/Navier2D_Epetra.cpp b/packages/muelu/test/navierstokes/Navier2D_Epetra.cpp
index dc4745f94262..adae34b28411 100644
--- a/packages/muelu/test/navierstokes/Navier2D_Epetra.cpp
+++ b/packages/muelu/test/navierstokes/Navier2D_Epetra.cpp
@@ -114,8 +114,7 @@
  *
  */
 
-
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
   typedef double Scalar;
   typedef int LocalOrdinal;
@@ -129,12 +128,12 @@ int main(int argc, char *argv[]) {
   using Teuchos::rcp;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
     *out << MueLu::MemUtils::PrintMemoryUsage() << std::endl;
 
@@ -149,180 +148,189 @@ int main(int argc, char *argv[]) {
     Teuchos::CommandLineProcessor clp(false);
 
     // - Levels
-    LO  optMaxLevels     = 4;              clp.setOption("maxLevels",      &optMaxLevels,          "maximum number of levels allowed");
-    int optMaxCoarseSize = 1;              clp.setOption("maxCoarseSize",  &optMaxCoarseSize,      "maximum #dofs in coarse operator"); //FIXME clp doesn't like long long int
+    LO optMaxLevels = 4;
+    clp.setOption("maxLevels", &optMaxLevels, "maximum number of levels allowed");
+    int optMaxCoarseSize = 1;
+    clp.setOption("maxCoarseSize", &optMaxCoarseSize, "maximum #dofs in coarse operator");  // FIXME clp doesn't like long long int
 
     // - Repartitioning
 #if defined(HAVE_MPI) && defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MUELU_ISORROPIA)
-    int    optRepartition    = 1;             clp.setOption("repartition",    &optRepartition,        "enable repartitioning");
-    LO     optMinRowsPerProc = 50;            clp.setOption("minRowsPerProc", &optMinRowsPerProc,     "min #rows allowable per proc before repartitioning occurs");
-    double optNnzImbalance   = 1.2;           clp.setOption("nnzImbalance",   &optNnzImbalance,       "max allowable nonzero imbalance before repartitioning occurs");
+    int optRepartition = 1;
+    clp.setOption("repartition", &optRepartition, "enable repartitioning");
+    LO optMinRowsPerProc = 50;
+    clp.setOption("minRowsPerProc", &optMinRowsPerProc, "min #rows allowable per proc before repartitioning occurs");
+    double optNnzImbalance = 1.2;
+    clp.setOption("nnzImbalance", &optNnzImbalance, "max allowable nonzero imbalance before repartitioning occurs");
 #else
     int optRepartition = 0;
 #endif
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-
     /////////////////////////////////////////////
     // custom parameters
     LO maxLevels = optMaxLevels;
 
-    GO maxCoarseSize=optMaxCoarseSize;
-    std::string aggOrdering = "natural";
-    int minPerAgg=3;
-    int maxNbrAlreadySelected=0;
+    GO maxCoarseSize          = optMaxCoarseSize;
+    std::string aggOrdering   = "natural";
+    int minPerAgg             = 3;
+    int maxNbrAlreadySelected = 0;
 
     int globalNumDofs = 1500;
-    int nProcs = comm->getSize();
-    int nDofsPerNode = 3;
+    int nProcs        = comm->getSize();
+    int nDofsPerNode  = 3;
 
-    int nLocalDofs = (int) globalNumDofs / nProcs;
-    nLocalDofs = nLocalDofs - (nLocalDofs % nDofsPerNode);
+    int nLocalDofs     = (int)globalNumDofs / nProcs;
+    nLocalDofs         = nLocalDofs - (nLocalDofs % nDofsPerNode);
     int nCumulatedDofs = 0;
-    MueLu_sumAll(comm,nLocalDofs, nCumulatedDofs);
-    //Teuchos::reduceAll<int,int>(*comm,Teuchos::REDUCE_SUM, 1, nLocalDofs, &nCumulatedDofs );
+    MueLu_sumAll(comm, nLocalDofs, nCumulatedDofs);
+    // Teuchos::reduceAll<int,int>(*comm,Teuchos::REDUCE_SUM, 1, nLocalDofs, &nCumulatedDofs );
 
-    if(comm->getRank() == nProcs-1) {
+    if (comm->getRank() == nProcs - 1) {
       nLocalDofs += globalNumDofs - nCumulatedDofs;
     }
 
     std::cout << "PROC: " << comm->getRank() << " numLocalDofs=" << nLocalDofs << std::endl;
 
     // read in problem
-    Epetra_Map emap (globalNumDofs, nLocalDofs, 0, *Xpetra::toEpetra(comm));
-    Epetra_CrsMatrix * ptrA = 0;
-    Epetra_Vector * ptrf = 0;
+    Epetra_Map emap(globalNumDofs, nLocalDofs, 0, *Xpetra::toEpetra(comm));
+    Epetra_CrsMatrix* ptrA    = 0;
+    Epetra_Vector* ptrf       = 0;
     Epetra_MultiVector* ptrNS = 0;
 
     std::cout << "Reading matrix market file" << std::endl;
-    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt",emap,emap,emap,ptrA);
-    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt",emap,ptrf);
-    RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    RCP<Epetra_Vector> epv = Teuchos::rcp(ptrf);
+    EpetraExt::MatrixMarketFileToCrsMatrix("A_re1000_5932.txt", emap, emap, emap, ptrA);
+    EpetraExt::MatrixMarketFileToVector("b_re1000_5932.txt", emap, ptrf);
+    RCP<Epetra_CrsMatrix> epA    = Teuchos::rcp(ptrA);
+    RCP<Epetra_Vector> epv       = Teuchos::rcp(ptrf);
     RCP<Epetra_MultiVector> epNS = Teuchos::rcp(ptrNS);
 
     // Epetra_CrsMatrix -> Xpetra::Matrix
-    RCP<CrsMatrix> exA = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO,NO>(epA));
+    RCP<CrsMatrix> exA       = Teuchos::rcp(new Xpetra::EpetraCrsMatrixT<GO, NO>(epA));
     RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
-    RCP<Matrix> Op = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
+    RCP<Matrix> Op           = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
 
-    Op->SetFixedBlockSize(nDofsPerNode);   // 2 velocity dofs and 1 pressure dof per node.
+    Op->SetFixedBlockSize(nDofsPerNode);  // 2 velocity dofs and 1 pressure dof per node.
 
     // Epetra_Vector -> Xpetra::Vector
-    RCP<Vector> xRhs = Teuchos::rcp(new Xpetra::EpetraVectorT<int,Node>(epv));
+    RCP<Vector> xRhs = Teuchos::rcp(new Xpetra::EpetraVectorT<int, Node>(epv));
 
-    RCP<MultiVector> xNS = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int,Node>(epNS));
+    RCP<MultiVector> xNS = Teuchos::rcp(new Xpetra::EpetraMultiVectorT<int, Node>(epNS));
 
     // Epetra_Map -> Xpetra::Map
-    const RCP< const Map> map = Xpetra::toXpetra<int,Node>(emap);
+    const RCP<const Map> map = Xpetra::toXpetra<int, Node>(emap);
 
-    RCP<Hierarchy> H = rcp ( new Hierarchy() );
+    RCP<Hierarchy> H = rcp(new Hierarchy());
     H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
     H->SetMaxCoarseSize(maxCoarseSize);
 
     // build finest Level
     RCP<MueLu::Level> Finest = H->GetLevel();
     Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);
-    //Finest->Set("Nullspace",xNS);
+    Finest->Set("A", Op);
+    // Finest->Set("Nullspace",xNS);
 
-    if (optRepartition==1) {
+    if (optRepartition == 1) {
       // create null space
 
       RCP<MultiVector> nullspace;
       // determine numPDEs
       LocalOrdinal numPDEs = 1;
-      if(Op->IsView("stridedMaps")==true) {
-        Xpetra::viewLabel_t oldView = Op->SwitchToView("stridedMaps"); // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
-        //TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(Op->getRowMap()) == Teuchos::null, Exceptions::BadCast, "cast to strided row map failed.");
+      if (Op->IsView("stridedMaps") == true) {
+        Xpetra::viewLabel_t oldView = Op->SwitchToView("stridedMaps");  // note: "stridedMaps are always non-overlapping (correspond to range and domain maps!)
+        // TEUCHOS_TEST_FOR_EXCEPTION(Teuchos::rcp_dynamic_cast<const StridedMap>(Op->getRowMap()) == Teuchos::null, Exceptions::BadCast, "cast to strided row map failed.");
         numPDEs = Teuchos::rcp_dynamic_cast<const StridedMap>(Op->getRowMap())->getFixedBlockSize();
         oldView = Op->SwitchToView(oldView);
       }
 
-      //GetOStream(Runtime1, 0) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
+      // GetOStream(Runtime1, 0) << "Generating canonical nullspace: dimension = " << numPDEs << std::endl;
       nullspace = MultiVectorFactory::Build(Op->getDomainMap(), numPDEs);
 
-      for (int i=0; i<numPDEs; ++i) {
+      for (int i = 0; i < numPDEs; ++i) {
         Teuchos::ArrayRCP<Scalar> nsValues = nullspace->getDataNonConst(i);
-        int numBlocks = nsValues.size() / numPDEs;
-        for (int j=0; j< numBlocks; ++j) {
-          nsValues[j*numPDEs + i] = 1.0;
+        int numBlocks                      = nsValues.size() / numPDEs;
+        for (int j = 0; j < numBlocks; ++j) {
+          nsValues[j * numPDEs + i] = 1.0;
         }
       }
-      Finest->Set("Nullspace",nullspace);
+      Finest->Set("Nullspace", nullspace);
     }
 
     RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("lightweight wrap",Teuchos::ParameterEntry(false));
+    dropFact->SetParameter("lightweight wrap", Teuchos::ParameterEntry(false));
     dropFact->SetVerbLevel(MueLu::Extreme);
 
-    //RCP<PreDropFunctionConstVal> predrop = rcp(new PreDropFunctionConstVal(0.00001));
-    //dropFact->SetPreDropFunction(predrop);
+    // RCP<PreDropFunctionConstVal> predrop = rcp(new PreDropFunctionConstVal(0.00001));
+    // dropFact->SetPreDropFunction(predrop);
     RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
     UncoupledAggFact->SetFactory("Graph", dropFact);
     *out << "========================= Aggregate option summary =========================" << std::endl;
     *out << "min DOFs per aggregate :                " << minPerAgg << std::endl;
     *out << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
-    UncoupledAggFact->SetMinNodesPerAggregate(minPerAgg); //TODO should increase if run anything other than 1D
+    UncoupledAggFact->SetMinNodesPerAggregate(minPerAgg);  // TODO should increase if run anything other than 1D
     UncoupledAggFact->SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
     std::transform(aggOrdering.begin(), aggOrdering.end(), aggOrdering.begin(), ::tolower);
     if (aggOrdering == "natural" || aggOrdering == "random" || aggOrdering == "graph") {
       *out << "aggregate ordering :                    " << aggOrdering << std::endl;
       UncoupledAggFact->SetOrdering(aggOrdering);
     } else {
-      std::string msg = "main: bad aggregation option """ + aggOrdering + """.";
+      std::string msg =
+          "main: bad aggregation option "
+          "" +
+          aggOrdering +
+          ""
+          ".";
       throw(MueLu::Exceptions::RuntimeError(msg));
     }
     *out << "=============================================================================" << std::endl;
 
     // build non-rebalanced transfer operators
-    RCP<PgPFactory> Pfact = rcp( new PgPFactory() );
-    RCP<Factory> Rfact  = rcp( new GenericRFactory());
-    //RCP<SaPFactory> Pfact  = rcp( new SaPFactory() );
-    //RCP<Factory>   Rfact  = rcp( new TransPFactory() );
-    RCP<RAPFactory> Acfact = rcp( new RAPFactory() );
+    RCP<PgPFactory> Pfact = rcp(new PgPFactory());
+    RCP<Factory> Rfact    = rcp(new GenericRFactory());
+    // RCP<SaPFactory> Pfact  = rcp( new SaPFactory() );
+    // RCP<Factory>   Rfact  = rcp( new TransPFactory() );
+    RCP<RAPFactory> Acfact = rcp(new RAPFactory());
     Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
     // build level smoothers
     RCP<SmootherPrototype> smooProto;
     std::string ifpackType;
     Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 3);
-    ifpackList.set("relaxation: damping factor", (SC) 0.6); // 0.7
+    ifpackList.set("relaxation: sweeps", (LO)3);
+    ifpackList.set("relaxation: damping factor", (SC)0.6);  // 0.7
     ifpackType = "RELAXATION";
     ifpackList.set("relaxation: type", "Gauss-Seidel");
 
-    smooProto = Teuchos::rcp( new TrilinosSmoother(ifpackType, ifpackList) );
+    smooProto = Teuchos::rcp(new TrilinosSmoother(ifpackType, ifpackList));
     RCP<SmootherFactory> SmooFact;
     if (maxLevels > 1)
-      SmooFact = rcp( new SmootherFactory(smooProto) );
+      SmooFact = rcp(new SmootherFactory(smooProto));
 
     // create coarsest smoother
     RCP<SmootherPrototype> coarsestSmooProto;
     std::string type = "";
     Teuchos::ParameterList coarsestSmooList;
 #if defined(HAVE_AMESOS_SUPERLU)
-    coarsestSmooProto = Teuchos::rcp( new DirectSolver("Superlu", coarsestSmooList) );
+    coarsestSmooProto = Teuchos::rcp(new DirectSolver("Superlu", coarsestSmooList));
 #else
-    coarsestSmooProto = Teuchos::rcp( new DirectSolver("Klu", coarsestSmooList) );
+    coarsestSmooProto  = Teuchos::rcp(new DirectSolver("Klu", coarsestSmooList));
 #endif
     RCP<SmootherFactory> coarsestSmooFact = rcp(new SmootherFactory(coarsestSmooProto, Teuchos::null));
 
     FactoryManager M;
     M.SetKokkosRefactor(false);
     M.SetFactory("Graph", dropFact);
-    //M.SetFactory("UnAmalgamationInfo", amalgFact);
+    // M.SetFactory("UnAmalgamationInfo", amalgFact);
     M.SetFactory("Aggregates", UncoupledAggFact);
     M.SetFactory("Smoother", SmooFact);
     M.SetFactory("CoarseSolver", coarsestSmooFact);
 
-    if(optRepartition == 0) {
+    if (optRepartition == 0) {
       // no rebalancing
       M.SetFactory("P", Pfact);
       M.SetFactory("R", Rfact);
@@ -377,7 +385,7 @@ int main(int argc, char *argv[]) {
       RCP<Factory> RebalancedRFact = rcp(new RebalanceTransferFactory());
       RebalancedRFact->SetParameter("type", Teuchos::ParameterEntry(std::string("Restriction")));
       RebalancedRFact->SetFactory("R", Rfact);
-      //RebalancedRFact->SetFactory("Coordinates", TransferCoordinatesFact);
+      // RebalancedRFact->SetFactory("Coordinates", TransferCoordinatesFact);
 
       // Compute Ac from rebalanced P and R
       RCP<Factory> RebalancedAFact = rcp(new RebalanceAcFactory());
@@ -387,8 +395,8 @@ int main(int argc, char *argv[]) {
       M.SetFactory("A", RebalancedAFact);
       M.SetFactory("P", RebalancedPFact);
       M.SetFactory("R", RebalancedRFact);
-      M.SetFactory("Nullspace",   RebalancedPFact);
-      M.SetFactory("Importer",    RepartitionFact);
+      M.SetFactory("Nullspace", RebalancedPFact);
+      M.SetFactory("Importer", RepartitionFact);
 #else
       // no re-balancing available
       M.SetFactory("P", Pfact);
@@ -399,21 +407,21 @@ int main(int argc, char *argv[]) {
 
     H->Setup(M, 0, maxLevels);
 
-    { // some debug output
+    {  // some debug output
       // print out content of levels
       std::cout << "FINAL CONTENT of multigrid levels" << std::endl;
-      for(LO l = 0; l < H->GetNumLevels(); l++) {
+      for (LO l = 0; l < H->GetNumLevels(); l++) {
         RCP<Level> coarseLevel = H->GetLevel(l);
         coarseLevel->print(*out);
       }
       std::cout << "END FINAL CONTENT of multigrid levels" << std::endl;
-    } // end debug output
+    }  // end debug output
 
-    RCP<MultiVector> xLsg = MultiVectorFactory::Build(map,1);
+    RCP<MultiVector> xLsg = MultiVectorFactory::Build(map, 1);
 
     // Use AMG directly as an iterative method
     {
-      xLsg->putScalar( (SC) 0.0);
+      xLsg->putScalar((SC)0.0);
 
       // calculate initial (absolute) residual
       Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
@@ -421,15 +429,14 @@ int main(int argc, char *argv[]) {
       *out << "||x_0|| = " << norms[0] << std::endl;
 
       // apply ten multigrid iterations
-      H->Iterate(*xRhs,*xLsg,10);
+      H->Iterate(*xRhs, *xLsg, 10);
 
       // calculate and print residual
-      RCP<MultiVector> xTmp = MultiVectorFactory::Build(map,1);
-      Op->apply(*xLsg,*xTmp,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-      xRhs->update((SC)-1.0,*xTmp,(SC)1.0);
+      RCP<MultiVector> xTmp = MultiVectorFactory::Build(map, 1);
+      Op->apply(*xLsg, *xTmp, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+      xRhs->update((SC)-1.0, *xTmp, (SC)1.0);
       xRhs->norm2(norms);
       *out << "||x|| = " << norms[0] << std::endl;
-
     }
 
     //
@@ -464,7 +471,7 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 #else
   std::cout << "Epetra (and/or EpetraExt) are not available. Skip test." << std::endl;
   return EXIT_SUCCESS;
diff --git a/packages/muelu/test/navierstokes/Navier2D_Helpers.cpp b/packages/muelu/test/navierstokes/Navier2D_Helpers.cpp
index 210ff32a7c6b..a2fa315fd8eb 100644
--- a/packages/muelu/test/navierstokes/Navier2D_Helpers.cpp
+++ b/packages/muelu/test/navierstokes/Navier2D_Helpers.cpp
@@ -46,5 +46,3 @@
 
 // the concrete implementation is in the .h file
 #include "Navier2D_Helpers.h"
-
-
diff --git a/packages/muelu/test/paramlist/paramlist.cpp b/packages/muelu/test/paramlist/paramlist.cpp
index a2786ab85b39..693051446f63 100644
--- a/packages/muelu/test/paramlist/paramlist.cpp
+++ b/packages/muelu/test/paramlist/paramlist.cpp
@@ -21,66 +21,66 @@
 
 namespace MueLu {
 
-  using Teuchos::RCP;
-  using Teuchos::ParameterList;
-
-  class MyFactory : public ParameterListAcceptorImpl {
-
-    public:
-
-      MyFactory() { }
-
-      virtual ~MyFactory() { }
-
-      RCP<const ParameterList> GetValidParameterList() const {
-        RCP<ParameterList> validParamList = rcp(new ParameterList());
+using Teuchos::ParameterList;
+using Teuchos::RCP;
 
-        validParamList->set("ParamA", 0.1, "Documentation string for param A");
-        validParamList->set("ParamB", 0.2, "Documentation string for param B");
-        validParamList->set("ParamC", 0.3, "Documentation string for param C");
-        validParamList->set("ParamD", 0.4, "Documentation string for param D");
+class MyFactory : public ParameterListAcceptorImpl {
+ public:
+  MyFactory() {}
 
-        return validParamList;
-      }
+  virtual ~MyFactory() {}
 
-      // Main algorithm
-      //
-      // - We do not want to set default parameters inside of the algorithm.
-      //  => Otherwise, it's difficult to track the defaults. Cf. ML.
-      //  => use ParameterList::get() without the default value input parameter.
-      void Build() {
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<ParameterList> validParamList = rcp(new ParameterList());
 
-        if (GetParameterList().get<double>("ParamA") == 0.5) { } // change "[used]"/["unused"] flag
-        if (GetParameterList().get<double>("ParamC") == 0.5) { }
+    validParamList->set("ParamA", 0.1, "Documentation string for param A");
+    validParamList->set("ParamB", 0.2, "Documentation string for param B");
+    validParamList->set("ParamC", 0.3, "Documentation string for param C");
+    validParamList->set("ParamD", 0.4, "Documentation string for param D");
 
-        // statsParamList_.set(...);
+    return validParamList;
+  }
 
-      }
+  // Main algorithm
+  //
+  // - We do not want to set default parameters inside of the algorithm.
+  //  => Otherwise, it's difficult to track the defaults. Cf. ML.
+  //  => use ParameterList::get() without the default value input parameter.
+  void Build() {
+    if (GetParameterList().get<double>("ParamA") == 0.5) {
+    }  // change "[used]"/["unused"] flag
+    if (GetParameterList().get<double>("ParamC") == 0.5) {
+    }
 
-      // - Do we want to store output stats on the same parameter list?
-      //   => better to distinguish as stats parameters are not valid input.
-      // - RCP? => a view seems enough.
-      Teuchos::ParameterList statsParamList_;
+    // statsParamList_.set(...);
+  }
 
-  };
+  // - Do we want to store output stats on the same parameter list?
+  //   => better to distinguish as stats parameters are not valid input.
+  // - RCP? => a view seems enough.
+  Teuchos::ParameterList statsParamList_;
+};
 
-}
+}  // namespace MueLu
 
 int main(int argc, char* argv[]) {
-  using Teuchos::ParameterList;
   using MueLu::MyFactory;
+  using Teuchos::ParameterList;
 
   bool success = false;
   try {
     //
     // Documentation
     //
-    std::cout << "\n#\n# Documentation\n#\n" << std::endl;
-    MyFactory dummy; dummy.GetDocumentation(std::cout);
+    std::cout << "\n#\n# Documentation\n#\n"
+              << std::endl;
+    MyFactory dummy;
+    dummy.GetDocumentation(std::cout);
 
     //
 
-    std::cout << "#\n# main()\n#\n" << std::endl;
+    std::cout << "#\n# main()\n#\n"
+              << std::endl;
 
     //
     // User parameter list
@@ -93,7 +93,8 @@ int main(int argc, char* argv[]) {
     paramList.set("ParamB", 0.002);
 
     std::cout << "# Input parameter list:" << std::endl;
-    std::cout << paramList << std::endl << std::endl;
+    std::cout << paramList << std::endl
+              << std::endl;
 
     //
     // Validation of the user parameter list
@@ -102,13 +103,14 @@ int main(int argc, char* argv[]) {
     MyFactory f;
     f.SetParameterList(paramList);
 
-    if (0) {// if users want to keep their list untouched:
+    if (0) {  // if users want to keep their list untouched:
       ParameterList tmp(paramList);
       f.SetParameterList(tmp);
     }
 
     std::cout << "# Parameter list after validation:" << std::endl;
-    std::cout << paramList << std::endl << std::endl;
+    std::cout << paramList << std::endl
+              << std::endl;
 
     //
     // Algorithm
@@ -117,10 +119,12 @@ int main(int argc, char* argv[]) {
     f.Build();
 
     std::cout << "# Parameter list after algorithm (flag used/unused):" << std::endl;
-    if (0) // do not work with my design: flags used/unused are not set for the initial parameter list
-      std::cout << paramList << std::endl << std::endl;
+    if (0)  // do not work with my design: flags used/unused are not set for the initial parameter list
+      std::cout << paramList << std::endl
+                << std::endl;
 
-    std::cout << f.GetParameterList() << std::endl << std::endl;
+    std::cout << f.GetParameterList() << std::endl
+              << std::endl;
 
     success = true;
 
@@ -128,5 +132,5 @@ int main(int argc, char* argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/test/paramlist/paramlistAdv.cpp b/packages/muelu/test/paramlist/paramlistAdv.cpp
index c3d82c4fc78e..c1b06cdc4afe 100644
--- a/packages/muelu/test/paramlist/paramlistAdv.cpp
+++ b/packages/muelu/test/paramlist/paramlistAdv.cpp
@@ -6,130 +6,129 @@
 
 namespace MueLu {
 
-  // Another version of ParameterListAcceptorImpl with conditional parameters: parameters "T" and "K" are available only according to the value of the parameter "Solver".
-  //
-  // Note: There is something in Teuchos/Optika to deal with such cases. I haven't figured out yet how to use it.
-  // See for instance: http://trilinos.sandia.gov/packages/docs/r10.2/packages/optika/browser/doc/html/test_2dependencyandexec_2main_8cpp-source.html
+// Another version of ParameterListAcceptorImpl with conditional parameters: parameters "T" and "K" are available only according to the value of the parameter "Solver".
+//
+// Note: There is something in Teuchos/Optika to deal with such cases. I haven't figured out yet how to use it.
+// See for instance: http://trilinos.sandia.gov/packages/docs/r10.2/packages/optika/browser/doc/html/test_2dependencyandexec_2main_8cpp-source.html
 
-  using Teuchos::RCP;
-  using Teuchos::ParameterList;
-
-  class ParameterListAcceptorAdvImpl: public ParameterListAcceptorImpl {
-
-    public:
-
-      ParameterListAcceptorAdvImpl() { }
-
-      virtual ~ParameterListAcceptorAdvImpl() { }
-
-      // This functions add *all* the extra parameters recursively using GetValidParameterListSimple
-      RCP<const ParameterList> GetValidParameterList() const {
-        RCP<const ParameterList> validParamList = GetValidParameterListSimple();
-
-        int numParams;
-        do {
-          numParams = validParamList->numParams();
-          validParamList = GetValidParameterListSimple();
-        } while (validParamList->numParams() != numParams);
-
-        return validParamList;
-      }
-
-      // GetValidParameterListSimple only add one extra level of default parameters. Ex: if "Solver" is not set in the input list "pL",
-      // extra parameters "T" or "K" are not added to the validParamList.
-      virtual RCP<const ParameterList> GetValidParameterListSimple() const = 0;
-
-      void GetDocumentation(std::ostream &os) const {
-        GetAdvancedDocumentation(os);
-      }
-
-    private:
-
-      virtual void GetAdvancedDocumentation(std::ostream &os) const = 0;
-
-  };
+using Teuchos::ParameterList;
+using Teuchos::RCP;
 
+class ParameterListAcceptorAdvImpl : public ParameterListAcceptorImpl {
+ public:
+  ParameterListAcceptorAdvImpl() {}
 
+  virtual ~ParameterListAcceptorAdvImpl() {}
 
-  class MyFactory : public ParameterListAcceptorAdvImpl {
+  // This functions add *all* the extra parameters recursively using GetValidParameterListSimple
+  RCP<const ParameterList> GetValidParameterList() const {
+    RCP<const ParameterList> validParamList = GetValidParameterListSimple();
 
-    public:
+    int numParams;
+    do {
+      numParams      = validParamList->numParams();
+      validParamList = GetValidParameterListSimple();
+    } while (validParamList->numParams() != numParams);
 
-      MyFactory() { }
-
-      virtual ~MyFactory() { }
-
-      RCP<const ParameterList> GetValidParameterListSimple() const {
-        RCP<ParameterList> validParamList = Teuchos::rcp(new ParameterList()); // output list
-
-        typedef Teuchos::StringToIntegralParameterEntryValidator<int> validator_type;
-        validParamList->set("Solver", "ILUT", "The type of solver to use.", Teuchos::rcp(new validator_type(Teuchos::tuple<std::string>("ILUT", "ILUK"), "Solver")));
+    return validParamList;
+  }
 
-        AddILUTParameters(*validParamList);
-        AddILUKParameters(*validParamList);
+  // GetValidParameterListSimple only add one extra level of default parameters. Ex: if "Solver" is not set in the input list "pL",
+  // extra parameters "T" or "K" are not added to the validParamList.
+  virtual RCP<const ParameterList> GetValidParameterListSimple() const = 0;
 
-        return validParamList;
-      }
+  void GetDocumentation(std::ostream& os) const {
+    GetAdvancedDocumentation(os);
+  }
 
-      // Main algorithm
-      void Build() {
-        const ParameterList & pL = GetParameterList();
-        std::string type = pL.get<std::string>("Solver");
-        if (type == "ILUT") {
-          pL.get<double>("T");
+ private:
+  virtual void GetAdvancedDocumentation(std::ostream& os) const = 0;
+};
 
-        } else if (type == "ILUK") {
-          pL.get<int>("K");
-        }
+class MyFactory : public ParameterListAcceptorAdvImpl {
+ public:
+  MyFactory() {}
 
-      }
+  virtual ~MyFactory() {}
 
-    private:
-      // Separates functions to be used by both GetValidParameterListSimple and GetDocumentation.
-      static void AddILUTParameters(ParameterList& paramList) {
-        paramList.set("T", 0.1, "ILUT threshold");
-      }
+  RCP<const ParameterList> GetValidParameterListSimple() const {
+    RCP<ParameterList> validParamList = Teuchos::rcp(new ParameterList());  // output list
 
-      static void AddILUKParameters(ParameterList& paramList) {
-        paramList.set("K", 1, "ILUK level of fill");
-      }
+    typedef Teuchos::StringToIntegralParameterEntryValidator<int> validator_type;
+    validParamList->set("Solver", "ILUT", "The type of solver to use.", Teuchos::rcp(new validator_type(Teuchos::tuple<std::string>("ILUT", "ILUK"), "Solver")));
 
-      void GetAdvancedDocumentation(std::ostream &os) const {
+    AddILUTParameters(*validParamList);
+    AddILUKParameters(*validParamList);
 
-        os << "## Parameters:" << std::endl;
-        printParameterListOptions(os, *GetValidParameterListSimple());
+    return validParamList;
+  }
 
-        os << "# ILUT specific parameters:" << std::endl;
-        { ParameterList p; AddILUKParameters(p); printParameterListOptions(os, p); }
+  // Main algorithm
+  void Build() {
+    const ParameterList& pL = GetParameterList();
+    std::string type        = pL.get<std::string>("Solver");
+    if (type == "ILUT") {
+      pL.get<double>("T");
 
+    } else if (type == "ILUK") {
+      pL.get<int>("K");
+    }
+  }
 
-        os << "# ILUK specific parameters:" << std::endl;
-        { ParameterList p; AddILUTParameters(p); printParameterListOptions(os, p); }
+ private:
+  // Separates functions to be used by both GetValidParameterListSimple and GetDocumentation.
+  static void AddILUTParameters(ParameterList& paramList) {
+    paramList.set("T", 0.1, "ILUT threshold");
+  }
 
-        os << "## Fully described default method:" << std::endl;
-        GetValidParameterList()->print(os, 2, true, false);
-        os << std::endl;
-      }
+  static void AddILUKParameters(ParameterList& paramList) {
+    paramList.set("K", 1, "ILUK level of fill");
+  }
 
-  };
+  void GetAdvancedDocumentation(std::ostream& os) const {
+    os << "## Parameters:" << std::endl;
+    printParameterListOptions(os, *GetValidParameterListSimple());
+
+    os << "# ILUT specific parameters:" << std::endl;
+    {
+      ParameterList p;
+      AddILUKParameters(p);
+      printParameterListOptions(os, p);
+    }
+
+    os << "# ILUK specific parameters:" << std::endl;
+    {
+      ParameterList p;
+      AddILUTParameters(p);
+      printParameterListOptions(os, p);
+    }
+
+    os << "## Fully described default method:" << std::endl;
+    GetValidParameterList()->print(os, 2, true, false);
+    os << std::endl;
+  }
+};
 
-}
+}  // namespace MueLu
 
 int main(int argc, char* argv[]) {
-  using Teuchos::ParameterList;
   using MueLu::MyFactory;
+  using Teuchos::ParameterList;
 
   bool success = false;
   try {
     //
     // Documentation
     //
-    std::cout << "\n#\n# Documentation\n#\n" << std::endl;
-    MyFactory dummy; dummy.GetDocumentation(std::cout);
+    std::cout << "\n#\n# Documentation\n#\n"
+              << std::endl;
+    MyFactory dummy;
+    dummy.GetDocumentation(std::cout);
 
     //
 
-    std::cout << "#\n# main()\n#\n" << std::endl;
+    std::cout << "#\n# main()\n#\n"
+              << std::endl;
 
     //
     // User parameter list
@@ -139,7 +138,8 @@ int main(int argc, char* argv[]) {
     // paramList.set("Solver", "ILUK");
 
     std::cout << "# Input parameter list:" << std::endl;
-    std::cout << paramList << std::endl << std::endl;
+    std::cout << paramList << std::endl
+              << std::endl;
 
     //
     // Validation of the user parameter list
@@ -149,7 +149,8 @@ int main(int argc, char* argv[]) {
     f.SetParameterList(paramList);
 
     std::cout << "# Parameter list after validation:" << std::endl;
-    std::cout << paramList << std::endl << std::endl;
+    std::cout << paramList << std::endl
+              << std::endl;
 
     //
     // Algorithm
@@ -159,11 +160,12 @@ int main(int argc, char* argv[]) {
 
     std::cout << "# Parameter list after algorithm (flag used/unused):" << std::endl;
 
-    std::cout << f.GetParameterList() << std::endl << std::endl;
+    std::cout << f.GetParameterList() << std::endl
+              << std::endl;
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/test/perf_tests_kokkos/KokkosKernels.cpp b/packages/muelu/test/perf_tests_kokkos/KokkosKernels.cpp
index 7d97a42ab3cb..5445754712b7 100644
--- a/packages/muelu/test/perf_tests_kokkos/KokkosKernels.cpp
+++ b/packages/muelu/test/perf_tests_kokkos/KokkosKernels.cpp
@@ -5,15 +5,16 @@
 #include <Kokkos_StaticCrsGraph.hpp>
 #include <Kokkos_Timer.hpp>
 
-template<class scalar_type, class local_ordinal_type, class device_type>
+template <class scalar_type, class local_ordinal_type, class device_type>
 class LWGraph_kokkos {
-private:
-  typedef Kokkos::StaticCrsGraph<local_ordinal_type, Kokkos::LayoutLeft, device_type>   local_graph_type;
-  typedef Kokkos::View<bool*, device_type>                                              boundary_nodes_type;
+ private:
+  typedef Kokkos::StaticCrsGraph<local_ordinal_type, Kokkos::LayoutLeft, device_type> local_graph_type;
+  typedef Kokkos::View<bool*, device_type> boundary_nodes_type;
 
-public:
-  LWGraph_kokkos(const local_graph_type& graph) : graph_(graph) { }
-  ~LWGraph_kokkos() { }
+ public:
+  LWGraph_kokkos(const local_graph_type& graph)
+    : graph_(graph) {}
+  ~LWGraph_kokkos() {}
 
   //! Set boolean array indicating which rows correspond to Dirichlet boundaries.
   KOKKOS_INLINE_FUNCTION
@@ -21,45 +22,44 @@ class LWGraph_kokkos {
     dirichletBoundaries_ = bndry;
   }
 
-private:
-
+ private:
   //! Underlying graph (with label)
   const local_graph_type graph_;
 
   //! Boolean array marking Dirichlet rows.
-  boundary_nodes_type   dirichletBoundaries_;
+  boundary_nodes_type dirichletBoundaries_;
 };
 
-template<class scalar_type, class local_ordinal_type, class device_type>
+template <class scalar_type, class local_ordinal_type, class device_type>
 KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type>
 kernel_construct(local_ordinal_type numRows) {
-  auto numCols         = numRows;
-  auto nnz             = 10*numRows;
+  auto numCols = numRows;
+  auto nnz     = 10 * numRows;
 
-  auto varianz_nel_row = 0.2*nnz/numRows;
-  auto width_row       = 0.01*numRows;
+  auto varianz_nel_row = 0.2 * nnz / numRows;
+  auto width_row       = 0.01 * numRows;
 
-  auto elements_per_row = nnz/numRows;
+  auto elements_per_row = nnz / numRows;
 
-  auto rowPtr = new local_ordinal_type[numRows+1];
-  rowPtr[0] = 0;
+  auto rowPtr = new local_ordinal_type[numRows + 1];
+  rowPtr[0]   = 0;
   for (int row = 0; row < numRows; row++) {
-    int varianz = (1.0*rand()/INT_MAX-0.5)*varianz_nel_row;
+    int varianz = (1.0 * rand() / INT_MAX - 0.5) * varianz_nel_row;
 
-    rowPtr[row+1] = rowPtr[row] + elements_per_row + varianz;
+    rowPtr[row + 1] = rowPtr[row] + elements_per_row + varianz;
   }
   nnz = rowPtr[numRows];
 
   auto colInd = new local_ordinal_type[nnz];
-  auto values = new scalar_type       [nnz];
+  auto values = new scalar_type[nnz];
   for (int row = 0; row < numRows; row++) {
-    for (int k = rowPtr[row]; k < rowPtr[row+1]; k++) {
-      int pos = row + (1.0*rand()/INT_MAX-0.5)*width_row;
+    for (int k = rowPtr[row]; k < rowPtr[row + 1]; k++) {
+      int pos = row + (1.0 * rand() / INT_MAX - 0.5) * width_row;
 
-      if (pos <  0)       pos += numCols;
+      if (pos < 0) pos += numCols;
       if (pos >= numCols) pos -= numCols;
       colInd[k] = pos;
-      values[k] = 100.0*rand()/INT_MAX - 50.0;
+      values[k] = 100.0 * rand() / INT_MAX - 50.0;
     }
   }
 
@@ -68,12 +68,12 @@ kernel_construct(local_ordinal_type numRows) {
   return local_matrix_type("A", numRows, numCols, nnz, values, rowPtr, colInd);
 }
 
-template<class scalar_type, class local_ordinal_type, class device_type>
+template <class scalar_type, class local_ordinal_type, class device_type>
 void kernel_coalesce_drop_device(KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type> A) {
-  typedef KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type>   local_matrix_type;
-  typedef Kokkos::ArithTraits<scalar_type>                                  ATS;
-  typedef typename ATS::mag_type                                            magnitude_type;
-  typedef Kokkos::View<bool*, device_type>                                  boundary_nodes_type;
+  typedef KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type> local_matrix_type;
+  typedef Kokkos::ArithTraits<scalar_type> ATS;
+  typedef typename ATS::mag_type magnitude_type;
+  typedef Kokkos::View<bool*, device_type> boundary_nodes_type;
 
   auto numRows = A.numRows();
 
@@ -85,18 +85,19 @@ void kernel_coalesce_drop_device(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   boundary_nodes_type boundaryNodes("boundaryNodes", numRows);
 
   double tol = 0.0;
-  Kokkos::parallel_for("MueLu:Utils::DetectDirichletRows", RangePolicy(0, numRows),
-   KOKKOS_LAMBDA(const local_ordinal_type row) {
-    auto rowView = A.row(row);
-    auto length  = rowView.length;
-
-    boundaryNodes(row) = true;
-    for (decltype(length) colID = 0; colID < length; colID++)
-      if ((rowView.colidx(colID) != row) && (ATS::magnitude(rowView.value(colID)) > tol)) {
-        boundaryNodes(row) = false;
-        break;
-      }
-   });
+  Kokkos::parallel_for(
+      "MueLu:Utils::DetectDirichletRows", RangePolicy(0, numRows),
+      KOKKOS_LAMBDA(const local_ordinal_type row) {
+        auto rowView = A.row(row);
+        auto length  = rowView.length;
+
+        boundaryNodes(row) = true;
+        for (decltype(length) colID = 0; colID < length; colID++)
+          if ((rowView.colidx(colID) != row) && (ATS::magnitude(rowView.value(colID)) > tol)) {
+            boundaryNodes(row) = false;
+            break;
+          }
+      });
 
   // Stage 1: calculate the number of remaining entries per row
   typedef Kokkos::StaticCrsGraph<local_ordinal_type, Kokkos::LayoutLeft, device_type> local_graph_type;
@@ -104,74 +105,79 @@ void kernel_coalesce_drop_device(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   typedef typename local_graph_type::entries_type entries_type;
 
   typename entries_type::non_const_type diag("ghosted", numRows);
-  typename row_map_type::non_const_type rows("row_map", numRows+1);       // rows(0) = 0 automatically
+  typename row_map_type::non_const_type rows("row_map", numRows + 1);  // rows(0) = 0 automatically
 
   local_ordinal_type realnnz = 0;
 
-  Kokkos::parallel_reduce("kernel_cd:stage1_reduce", RangePolicy(0, numRows),
-    KOKKOS_LAMBDA(const local_ordinal_type row, local_ordinal_type& nnz) {
-      auto rowView = A.row (row);
-      auto length  = rowView.length;
+  Kokkos::parallel_reduce(
+      "kernel_cd:stage1_reduce", RangePolicy(0, numRows),
+      KOKKOS_LAMBDA(const local_ordinal_type row, local_ordinal_type& nnz) {
+        auto rowView = A.row(row);
+        auto length  = rowView.length;
 
-      local_ordinal_type rownnz = 0;
-      for (decltype(length) colID = 0; colID < length; colID++) {
-        auto col = rowView.colidx(colID);
+        local_ordinal_type rownnz = 0;
+        for (decltype(length) colID = 0; colID < length; colID++) {
+          auto col = rowView.colidx(colID);
 
-        // Avoid square root by using squared values
-        auto aiiajj = eps*eps * ATS::magnitude(diag(row))*ATS::magnitude(diag(col));          // eps^2*|a_ii|*|a_jj|
-        auto aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));  // |a_ij|^2
+          // Avoid square root by using squared values
+          auto aiiajj = eps * eps * ATS::magnitude(diag(row)) * ATS::magnitude(diag(col));            // eps^2*|a_ii|*|a_jj|
+          auto aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));  // |a_ij|^2
 
-        if (aij2 > aiiajj || row == col)
-          rownnz++;
-      }
-      rows(row+1) = rownnz;
-      nnz += rownnz;
-  }, realnnz);
+          if (aij2 > aiiajj || row == col)
+            rownnz++;
+        }
+        rows(row + 1) = rownnz;
+        nnz += rownnz;
+      },
+      realnnz);
 
   // parallel_scan (exclusive)
-  Kokkos::parallel_scan("kernel_cd:stage1_scan", RangePolicy(0, numRows+1),
-    KOKKOS_LAMBDA(const local_ordinal_type i, local_ordinal_type& upd, const bool& final) {
-    upd += rows(i);
-    if (final)
-      rows(i) = upd;
-  });
+  Kokkos::parallel_scan(
+      "kernel_cd:stage1_scan", RangePolicy(0, numRows + 1),
+      KOKKOS_LAMBDA(const local_ordinal_type i, local_ordinal_type& upd, const bool& final) {
+        upd += rows(i);
+        if (final)
+          rows(i) = upd;
+      });
 
   // Stage 2: fill in the column indices
   typename boundary_nodes_type::non_const_type bndNodes("boundaryNodes", numRows);
-  typename entries_type::non_const_type        cols    ("entries",       realnnz);
+  typename entries_type::non_const_type cols("entries", realnnz);
 
   local_ordinal_type numDropped = 0;
 
-  Kokkos::parallel_reduce("kernel_cd:stage2_reduce", RangePolicy(0, numRows),
-    KOKKOS_LAMBDA(const local_ordinal_type row, local_ordinal_type& dropped) {
-    auto rowView = A.row (row);
-    auto length = rowView.length;
-
-    local_ordinal_type rownnz = 0;
-    for (decltype(length) colID = 0; colID < length; colID++) {
-      auto col = rowView.colidx(colID);
-
-      // Avoid square root by using squared values
-      auto aiiajj = eps*eps * ATS::magnitude(diag(row))*ATS::magnitude(diag(col));            // eps^2*|a_ii|*|a_jj|
-      auto aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));    // |a_ij|^2
-
-      if (aij2 > aiiajj || row == col) {
-        cols(rows(row) + rownnz) = col;
-        rownnz++;
-      } else {
-        dropped++;
-      }
-      if (rownnz == 1) {
-        // If the only element remaining after filtering is diagonal, mark node as boundary
-        // FIXME: this should really be replaced by the following
-        //    if (indices.size() == 1 && indices[0] == row)
-        //        boundaryNodes[row] = true;
-        // We do not do it this way now because there is no framework for distinguishing isolated
-        // and boundary nodes in the aggregation algorithms
-        bndNodes(row) = true;
-      }
-    }
-  }, numDropped);
+  Kokkos::parallel_reduce(
+      "kernel_cd:stage2_reduce", RangePolicy(0, numRows),
+      KOKKOS_LAMBDA(const local_ordinal_type row, local_ordinal_type& dropped) {
+        auto rowView = A.row(row);
+        auto length  = rowView.length;
+
+        local_ordinal_type rownnz = 0;
+        for (decltype(length) colID = 0; colID < length; colID++) {
+          auto col = rowView.colidx(colID);
+
+          // Avoid square root by using squared values
+          auto aiiajj = eps * eps * ATS::magnitude(diag(row)) * ATS::magnitude(diag(col));            // eps^2*|a_ii|*|a_jj|
+          auto aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));  // |a_ij|^2
+
+          if (aij2 > aiiajj || row == col) {
+            cols(rows(row) + rownnz) = col;
+            rownnz++;
+          } else {
+            dropped++;
+          }
+          if (rownnz == 1) {
+            // If the only element remaining after filtering is diagonal, mark node as boundary
+            // FIXME: this should really be replaced by the following
+            //    if (indices.size() == 1 && indices[0] == row)
+            //        boundaryNodes[row] = true;
+            // We do not do it this way now because there is no framework for distinguishing isolated
+            // and boundary nodes in the aggregation algorithms
+            bndNodes(row) = true;
+          }
+        }
+      },
+      numDropped);
 
   boundaryNodes = bndNodes;
 
@@ -182,12 +188,12 @@ void kernel_coalesce_drop_device(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   graph.setBoundaryNodeMap(boundaryNodes);
 }
 
-template<class scalar_type, class local_ordinal_type, class device_type>
+template <class scalar_type, class local_ordinal_type, class device_type>
 void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type> A) {
-  typedef KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type>   local_matrix_type;
-  typedef Kokkos::ArithTraits<scalar_type>                                  ATS;
-  typedef typename ATS::mag_type                                            magnitude_type;
-  typedef Kokkos::View<bool*, device_type>                                  boundary_nodes_type;
+  typedef KokkosSparse::CrsMatrix<scalar_type, local_ordinal_type, device_type> local_matrix_type;
+  typedef Kokkos::ArithTraits<scalar_type> ATS;
+  typedef typename ATS::mag_type magnitude_type;
+  typedef Kokkos::View<bool*, device_type> boundary_nodes_type;
 
   auto numRows = A.numRows();
 
@@ -198,7 +204,7 @@ void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   // Stage 0: detect Dirichlet rows
   boundary_nodes_type boundaryNodes("boundaryNodes", numRows);
   for (int row = 0; row < numRows; row++) {
-    auto rowView = A.row (row);
+    auto rowView = A.row(row);
     auto length  = rowView.length;
 
     boundaryNodes(row) = true;
@@ -215,15 +221,15 @@ void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   typedef typename local_graph_type::entries_type entries_type;
 
   typename entries_type::non_const_type diag("ghosted", numRows);
-  typename row_map_type::non_const_type rows("row_map", numRows+1);       // rows(0) = 0 automatically
-  typename entries_type::non_const_type cols("entries", 20*numRows);      // estimate
+  typename row_map_type::non_const_type rows("row_map", numRows + 1);   // rows(0) = 0 automatically
+  typename entries_type::non_const_type cols("entries", 20 * numRows);  // estimate
 
   typename boundary_nodes_type::non_const_type bndNodes("boundaryNodes", numRows);
   local_ordinal_type numDropped = 0;
 
   local_ordinal_type realnnz = 0;
   for (int row = 0; row < numRows; row++) {
-    auto rowView = A.row (row);
+    auto rowView = A.row(row);
     auto length  = rowView.length;
 
     local_ordinal_type rownnz = 0;
@@ -231,8 +237,8 @@ void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordi
       local_ordinal_type col = rowView.colidx(colID);
 
       // Avoid square root by using squared values
-      magnitude_type aiiajj = eps*eps * ATS::magnitude(diag(row))*ATS::magnitude(diag(col));                  // eps^2*|a_ii|*|a_jj|
-      magnitude_type aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));    // |a_ij|^2
+      magnitude_type aiiajj = eps * eps * ATS::magnitude(diag(row)) * ATS::magnitude(diag(col));            // eps^2*|a_ii|*|a_jj|
+      magnitude_type aij2   = ATS::magnitude(rowView.value(colID)) * ATS::magnitude(rowView.value(colID));  // |a_ij|^2
 
       if (aij2 > aiiajj || row == col) {
         cols(rownnz++) = col;
@@ -251,7 +257,7 @@ void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordi
         bndNodes(row) = true;
       }
     }
-    rows(row+1) = rows(row) + rownnz;
+    rows(row + 1) = rows(row) + rownnz;
     realnnz += rownnz;
   }
   boundaryNodes = bndNodes;
@@ -263,16 +269,22 @@ void kernel_coalesce_drop_serial(KokkosSparse::CrsMatrix<scalar_type, local_ordi
   graph.setBoundaryNodeMap(boundaryNodes);
 }
 
-template<class scalar_type, class local_ordinal_type, class device_type>
-int main_(int argc, char **argv) {
+template <class scalar_type, class local_ordinal_type, class device_type>
+int main_(int argc, char** argv) {
   int n    = 100000;
   int loop = 10;
 
   for (int i = 1; i < argc; i++) {
-    if      (strcmp(argv[i], "-n") == 0)                                   { n    = atoi(argv[++i]); continue; }
-    else if (strcmp(argv[i], "-l") == 0 || strcmp(argv[i], "--loop") == 0) { loop = atoi(argv[++i]); continue; }
-    else if (strcmp(argv[i], "--node") == 0)                               { i++; continue; }
-    else if (strcmp(argv[i], "-h") == 0 || strcmp(argv[i], "--help") == 0) {
+    if (strcmp(argv[i], "-n") == 0) {
+      n = atoi(argv[++i]);
+      continue;
+    } else if (strcmp(argv[i], "-l") == 0 || strcmp(argv[i], "--loop") == 0) {
+      loop = atoi(argv[++i]);
+      continue;
+    } else if (strcmp(argv[i], "--node") == 0) {
+      i++;
+      continue;
+    } else if (strcmp(argv[i], "-h") == 0 || strcmp(argv[i], "--help") == 0) {
       std::cout << "./MueLu_KokkosKernels.exe [-n <matrix_size>] [-l <number_of_loops>]" << std::endl;
       return 0;
     } else {
@@ -311,7 +323,6 @@ int main_(int argc, char **argv) {
 }
 
 int main(int argc, char* argv[]) {
-
   Kokkos::initialize(argc, argv);
 
   srand(13721);
@@ -330,40 +341,40 @@ int main(int argc, char* argv[]) {
   std::cout << "node = " << (node == "" ? "default" : node) << std::endl;
 
   if (node == "") {
-    return main_<double,int,Kokkos::DefaultExecutionSpace>(argc, argv);
+    return main_<double, int, Kokkos::DefaultExecutionSpace>(argc, argv);
 
   } else if (node == "serial") {
 #ifdef KOKKOS_ENABLE_SERIAL
-    return main_<double,int,Kokkos::Serial>(argc, argv);
+    return main_<double, int, Kokkos::Serial>(argc, argv);
 #else
     std::cout << "Error: Serial node type is disabled" << std::endl;
 #endif
 
   } else if (node == "openmp") {
 #ifdef KOKKOS_ENABLE_OPENMP
-    return main_<double,int,Kokkos::OpenMP>(argc, argv);
+    return main_<double, int, Kokkos::OpenMP>(argc, argv);
 #else
     std::cout << "Error: OpenMP node type is disabled" << std::endl;
 #endif
 
   } else if (node == "cuda") {
 #ifdef KOKKOS_ENABLE_CUDA
-    return main_<double,int,Kokkos::Cuda>(argc, argv);
+    return main_<double, int, Kokkos::Cuda>(argc, argv);
 #else
     std::cout << "Error: CUDA node type is disabled" << std::endl;
 #endif
   } else if (node == "hip") {
 #ifdef KOKKOS_ENABLE_HIP
-    return main_<double,int,Kokkos::HIP>(argc, argv);
+    return main_<double, int, Kokkos::HIP>(argc, argv);
 #else
     std::cout << "Error: HIP node type is disabled" << std::endl;
 #endif
   } else if (node == "sycl") {
 #ifdef KOKKOS_ENABLE_SYCL
-    return main_<double,int,Kokkos::Experimental::SYCL>(argc, argv);
+    return main_<double, int, Kokkos::Experimental::SYCL>(argc, argv);
 #else
     std::cout << "Error: SYCL node type is disabled" << std::endl;
-#endif    
+#endif
   }
 
   Kokkos::finalize();
diff --git a/packages/muelu/test/perf_tests_kokkos/Redirection.cpp b/packages/muelu/test/perf_tests_kokkos/Redirection.cpp
index 63ad4923b1f8..70d679b3bdde 100644
--- a/packages/muelu/test/perf_tests_kokkos/Redirection.cpp
+++ b/packages/muelu/test/perf_tests_kokkos/Redirection.cpp
@@ -67,14 +67,14 @@
 #include <MueLu_ExplicitInstantiation.hpp>
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayView;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayView;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -89,17 +89,18 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Parameters initialization
   // =========================================================================
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
 
-  int loops = 10; clp.setOption("l", &loops, "Number of loops");
+  int loops = 10;
+  clp.setOption("l", &loops, "Number of loops");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   // Retrieve matrix parameters (they may have been changed on the command line)
@@ -120,8 +121,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D")
     map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, galeriList);
   RCP<Matrix> A = Pr->BuildMatrix();
 
   LO numRows = A->getLocalNumRows();
@@ -138,7 +139,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Loop #1: Xpetra")));
     for (int i = 0; i < loops; i++) {
       for (LocalOrdinal row = 0; row < numRows; row++) {
-
         A->getLocalRowView(row, indices, vals);
 
         validation += indices.size();
@@ -154,16 +154,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     GO validation = 0;
 
     if (lib == Xpetra::UseTpetra) {
-      typedef Tpetra::CrsMatrix<SC,LO,GO,NO> tCrsMatrix;
+      typedef Tpetra::CrsMatrix<SC, LO, GO, NO> tCrsMatrix;
       RCP<const tCrsMatrix> tA = Utilities::Op2TpetraCrs(A);
       TEUCHOS_TEST_FOR_EXCEPTION(tA.is_null(), MueLu::Exceptions::RuntimeError,
                                  "A is not a Tpetra CrsMatrix");
 
       typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::local_inds_host_view_type indices;
       typename Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::values_host_view_type vals;
- 
+
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Loop #2: Tpetra/Epetra")));
-      for (int i = 0; i < loops; i++)  {
+      for (int i = 0; i < loops; i++) {
         for (LocalOrdinal row = 0; row < numRows; row++) {
           tA->getLocalRowView(row, indices, vals);
 
@@ -183,9 +183,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       for (int i = 0; i < loops; i++) {
         for (LocalOrdinal row = 0; row < numRows; row++) {
-          int      numEntries;
-          double * eValues;
-          int    * eIndices;
+          int numEntries;
+          double *eValues;
+          int *eIndices;
 
           eA->ExtractMyRowView(row, numEntries, eValues, eIndices);
 
@@ -201,7 +201,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     std::cout << "validation = " << validation << std::endl;
   }
 
-
   // Loop 3
   {
     auto localMatrix = A->getLocalMatrixHost();
@@ -211,7 +210,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Loop #3: Kokkos-serial")));
     for (int i = 0; i < loops; i++) {
       for (LocalOrdinal row = 0; row < numRows; row++) {
-        auto rowView = localMatrix.row (row);
+        auto rowView = localMatrix.row(row);
         auto length  = rowView.length;
 
         validation += length;
@@ -232,13 +231,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Loop #4: Kokkos-parallel (node)")));
     for (int i = 0; i < loops; i++)
-      Kokkos::parallel_reduce("Utils::DetectDirichletRows", RangePolicy(0, numRows),
-        KOKKOS_LAMBDA(const LO row, GO& r) {
-          auto rowView = localMatrix.row (row);
-          auto length  = rowView.length;
-
-          r += length;
-        }, validation);
+      Kokkos::parallel_reduce(
+          "Utils::DetectDirichletRows", RangePolicy(0, numRows),
+          KOKKOS_LAMBDA(const LO row, GO &r) {
+            auto rowView = localMatrix.row(row);
+            auto length  = rowView.length;
+
+            r += length;
+          },
+          validation);
     tm = Teuchos::null;
 
     std::cout << "validation = " << validation << std::endl;
@@ -262,6 +263,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/permutation/Permutation.cpp b/packages/muelu/test/permutation/Permutation.cpp
index 2ee174cfdc47..a9b3e8482c46 100644
--- a/packages/muelu/test/permutation/Permutation.cpp
+++ b/packages/muelu/test/permutation/Permutation.cpp
@@ -76,221 +76,208 @@
 
 namespace MueLuTests {
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Epetra_CrsMatrix> GetEpetraMatrix(std::string name, const Teuchos::RCP<MueLu::Level> level, const Teuchos::RCP<MueLu::Factory>& fct) {
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > result = level->Get<Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > >(name, fct.get());
-    Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsres = Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(result);
-    Teuchos::RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > crsmat = crsres->getCrsMatrix();
-    Teuchos::RCP<Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node> > epcrsmat = Teuchos::rcp_dynamic_cast<Xpetra::EpetraCrsMatrixT<GlobalOrdinal,Node> >(crsmat);
-    Teuchos::RCP<const Epetra_CrsMatrix> epres = epcrsmat->getEpetra_CrsMatrix();
-    return epres;
-  }
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Epetra_CrsMatrix> GetEpetraMatrix(std::string name, const Teuchos::RCP<MueLu::Level> level, const Teuchos::RCP<MueLu::Factory>& fct) {
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > result        = level->Get<Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > >(name, fct.get());
+  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> > crsres = Teuchos::rcp_dynamic_cast<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(result);
+  Teuchos::RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > crsmat     = crsres->getCrsMatrix();
+  Teuchos::RCP<Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node> > epcrsmat                  = Teuchos::rcp_dynamic_cast<Xpetra::EpetraCrsMatrixT<GlobalOrdinal, Node> >(crsmat);
+  Teuchos::RCP<const Epetra_CrsMatrix> epres                                             = epcrsmat->getEpetra_CrsMatrix();
+  return epres;
+}
 
-  // run tests with "Algebraic" permutation strategy and nDofsPerNode = 1
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool runPermutationTest(const std::string input_filename, const std::string expected_filename, const Teuchos::RCP<const Teuchos::Comm<int> >& comm) {
+// run tests with "Algebraic" permutation strategy and nDofsPerNode = 1
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool runPermutationTest(const std::string input_filename, const std::string expected_filename, const Teuchos::RCP<const Teuchos::Comm<int> >& comm) {
 #include <MueLu_UseShortNames.hpp>
 
 #ifndef HAVE_MUELU_INST_COMPLEX_INT_INT
-    Teuchos::RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    out->setOutputToRootOnly(0);
-
-    Epetra_CrsMatrix * ptrA = NULL;
-    Epetra_CrsMatrix * ptrExpected = NULL;
-    int ret = EpetraExt::MatlabFileToCrsMatrix ( input_filename.c_str(),
-        *Xpetra::toEpetra(comm),
-        ptrA
-        );
+  Teuchos::RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  out->setOutputToRootOnly(0);
 
-    if(ret!=0)
-      std::cout << "failed to read matrix from file" << std::endl;
-
-    if(expected_filename.size() > 0)
-    {
-      int ret2 = EpetraExt::MatlabFileToCrsMatrix (expected_filename.c_str(),
-          *Xpetra::toEpetra(comm),
-          ptrExpected
-          );
+  Epetra_CrsMatrix* ptrA        = NULL;
+  Epetra_CrsMatrix* ptrExpected = NULL;
+  int ret                       = EpetraExt::MatlabFileToCrsMatrix(input_filename.c_str(),
+                                                                   *Xpetra::toEpetra(comm),
+                                                                   ptrA);
 
-      if(ret2!=0)
-        std::cout << "failed to read matrix from file" << std::endl;
+  if (ret != 0)
+    std::cout << "failed to read matrix from file" << std::endl;
 
-    }
-    Teuchos::RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    Teuchos::RCP<Epetra_CrsMatrix> epExpected = Teuchos::rcp(ptrExpected);
-
-    // Epetra_CrsMatrix -> Xpetra::Matrix
-    Teuchos::RCP<CrsMatrix> exA = Teuchos::rcp(new EpetraCrsMatrix(epA));
-    Teuchos::RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
-    Teuchos::RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
-    A->SetFixedBlockSize(1);
-
-    Teuchos::RCP<Level> Finest = Teuchos::rcp(new Level());
-    Finest->SetLevelID(0);  // must be level 0 for NullspaceFactory
-    Finest->Set("A", A);
-
-    // permute full matrix
-    Teuchos::RCP<PermutationFactory> PermFact = Teuchos::rcp(new MueLu::PermutationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-    PermFact->SetParameter("PermutationStrategy",Teuchos::ParameterEntry(std::string("Algebraic")));
-    //PermFact->SetParameter("PermutationStrategy",Teuchos::ParameterEntry(std::string("Local")));
-    PermFact->SetParameter("PermutationRowMapName",Teuchos::ParameterEntry(std::string("")));
-    PermFact->SetFactory("PermutationRowMapFactory", Teuchos::null);
-
-    // setup main factory manager
-    Teuchos::RCP<FactoryManager> M = Teuchos::rcp(new FactoryManager());
-    M->SetFactory("permQT",          PermFact);
-    M->SetFactory("A",               MueLu::NoFactory::getRCP()); // this is the input matrix
-    MueLu::SetFactoryManager SFMFinest(Finest, M);
-
-    // prepare building process for permutation operators
-    Finest->Request("A", PermFact.get());
-    Finest->Request("permA", PermFact.get());
-    Finest->Request("permP", PermFact.get());
-    Finest->Request("permQT", PermFact.get());
-    Finest->Request("permScaling", PermFact.get());
-    Finest->Request("#RowPermutations", PermFact.get());
-    Finest->Request("#ColPermutations", PermFact.get());
-    Finest->Request("#WideRangeRowPermutations", PermFact.get());
-    Finest->Request("#WideRangeColPermutations", PermFact.get());
-
-    // build permutation operators
-    PermFact->Build(*Finest);
-
-    //std::cout << "P" <<  *GetEpetraMatrix("permP", Finest, PermFact) << std::endl;
-    //std::cout << "Q^T" << *GetEpetraMatrix("permQT", Finest, PermFact) << std::endl;
-    //std::cout << "permA" <<  *GetEpetraMatrix("A", Finest, PermFact) << std::endl;
-
-    Teuchos::RCP<const Epetra_CrsMatrix> epResult = GetEpetraMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>("A", Finest, PermFact);
-    //std::cout << *epResult << std::endl;
-
-    if(epExpected != Teuchos::null) {
-      Epetra_CrsMatrix* comparison = NULL;
-      EpetraExt::MatrixMatrix::Add(*epResult, false, -1.0, *epExpected, false, 1.0, comparison);
-      comparison->FillComplete();
-      double norm = comparison->NormInf();
-      delete comparison;
-      comparison = NULL;
-
-      if(norm < 1.0e-14) {
-        *out << "** PASSED **: " << input_filename << std::endl;
-        return true;
-      }
-      else {
-        *out << "-- FAILED --: " << input_filename << std::endl;
-        return false;
-      }
+  if (expected_filename.size() > 0) {
+    int ret2 = EpetraExt::MatlabFileToCrsMatrix(expected_filename.c_str(),
+                                                *Xpetra::toEpetra(comm),
+                                                ptrExpected);
 
+    if (ret2 != 0)
+      std::cout << "failed to read matrix from file" << std::endl;
+  }
+  Teuchos::RCP<Epetra_CrsMatrix> epA        = Teuchos::rcp(ptrA);
+  Teuchos::RCP<Epetra_CrsMatrix> epExpected = Teuchos::rcp(ptrExpected);
+
+  // Epetra_CrsMatrix -> Xpetra::Matrix
+  Teuchos::RCP<CrsMatrix> exA       = Teuchos::rcp(new EpetraCrsMatrix(epA));
+  Teuchos::RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
+  Teuchos::RCP<Matrix> A            = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
+  A->SetFixedBlockSize(1);
+
+  Teuchos::RCP<Level> Finest = Teuchos::rcp(new Level());
+  Finest->SetLevelID(0);  // must be level 0 for NullspaceFactory
+  Finest->Set("A", A);
+
+  // permute full matrix
+  Teuchos::RCP<PermutationFactory> PermFact = Teuchos::rcp(new MueLu::PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+  PermFact->SetParameter("PermutationStrategy", Teuchos::ParameterEntry(std::string("Algebraic")));
+  // PermFact->SetParameter("PermutationStrategy",Teuchos::ParameterEntry(std::string("Local")));
+  PermFact->SetParameter("PermutationRowMapName", Teuchos::ParameterEntry(std::string("")));
+  PermFact->SetFactory("PermutationRowMapFactory", Teuchos::null);
+
+  // setup main factory manager
+  Teuchos::RCP<FactoryManager> M = Teuchos::rcp(new FactoryManager());
+  M->SetFactory("permQT", PermFact);
+  M->SetFactory("A", MueLu::NoFactory::getRCP());  // this is the input matrix
+  MueLu::SetFactoryManager SFMFinest(Finest, M);
+
+  // prepare building process for permutation operators
+  Finest->Request("A", PermFact.get());
+  Finest->Request("permA", PermFact.get());
+  Finest->Request("permP", PermFact.get());
+  Finest->Request("permQT", PermFact.get());
+  Finest->Request("permScaling", PermFact.get());
+  Finest->Request("#RowPermutations", PermFact.get());
+  Finest->Request("#ColPermutations", PermFact.get());
+  Finest->Request("#WideRangeRowPermutations", PermFact.get());
+  Finest->Request("#WideRangeColPermutations", PermFact.get());
+
+  // build permutation operators
+  PermFact->Build(*Finest);
+
+  // std::cout << "P" <<  *GetEpetraMatrix("permP", Finest, PermFact) << std::endl;
+  // std::cout << "Q^T" << *GetEpetraMatrix("permQT", Finest, PermFact) << std::endl;
+  // std::cout << "permA" <<  *GetEpetraMatrix("A", Finest, PermFact) << std::endl;
+
+  Teuchos::RCP<const Epetra_CrsMatrix> epResult = GetEpetraMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>("A", Finest, PermFact);
+  // std::cout << *epResult << std::endl;
+
+  if (epExpected != Teuchos::null) {
+    Epetra_CrsMatrix* comparison = NULL;
+    EpetraExt::MatrixMatrix::Add(*epResult, false, -1.0, *epExpected, false, 1.0, comparison);
+    comparison->FillComplete();
+    double norm = comparison->NormInf();
+    delete comparison;
+    comparison = NULL;
+
+    if (norm < 1.0e-14) {
+      *out << "** PASSED **: " << input_filename << std::endl;
+      return true;
+    } else {
+      *out << "-- FAILED --: " << input_filename << std::endl;
+      return false;
     }
-#endif
-    *out << "-- FAILED --: " << input_filename << " no result file found" << std::endl;
-    return false; // no result for comparison available
   }
+#endif
+  *out << "-- FAILED --: " << input_filename << " no result file found" << std::endl;
+  return false;  // no result for comparison available
+}
 
-  // run tests with "Local" permutation strategy and nDofsPerNode = 3
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  bool runPermutationTest2(const std::string input_filename, const std::string expected_filename, const Teuchos::RCP<const Teuchos::Comm<int> >& comm) {
+// run tests with "Local" permutation strategy and nDofsPerNode = 3
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+bool runPermutationTest2(const std::string input_filename, const std::string expected_filename, const Teuchos::RCP<const Teuchos::Comm<int> >& comm) {
 #include <MueLu_UseShortNames.hpp>
 
 #ifndef HAVE_MUELU_INST_COMPLEX_INT_INT
-    Teuchos::RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    out->setOutputToRootOnly(0);
+  Teuchos::RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  out->setOutputToRootOnly(0);
 
-    Epetra_CrsMatrix * ptrA = NULL;
-    Epetra_CrsMatrix * ptrExpected = NULL;
-    int ret = EpetraExt::MatlabFileToCrsMatrix ( input_filename.c_str(),
-        *Xpetra::toEpetra(comm),
-        ptrA
-        );
+  Epetra_CrsMatrix* ptrA        = NULL;
+  Epetra_CrsMatrix* ptrExpected = NULL;
+  int ret                       = EpetraExt::MatlabFileToCrsMatrix(input_filename.c_str(),
+                                                                   *Xpetra::toEpetra(comm),
+                                                                   ptrA);
 
-    if(ret!=0)
-      std::cout << "failed to read matrix from file" << std::endl;
-
-    if(expected_filename.size() > 0)
-    {
-      int ret2 = EpetraExt::MatlabFileToCrsMatrix (expected_filename.c_str(),
-          *Xpetra::toEpetra(comm),
-          ptrExpected
-          );
-
-      if(ret2!=0)
-        std::cout << "failed to read matrix from file" << std::endl;
+  if (ret != 0)
+    std::cout << "failed to read matrix from file" << std::endl;
 
-    }
-    Teuchos::RCP<Epetra_CrsMatrix> epA = Teuchos::rcp(ptrA);
-    Teuchos::RCP<Epetra_CrsMatrix> epExpected = Teuchos::rcp(ptrExpected);
-
-    // Epetra_CrsMatrix -> Xpetra::Matrix
-    Teuchos::RCP<CrsMatrix> exA = Teuchos::rcp(new EpetraCrsMatrix(epA));
-    Teuchos::RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
-    Teuchos::RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
-    A->SetFixedBlockSize(3);
-
-    Teuchos::RCP<Level> Finest = Teuchos::rcp(new Level());
-    Finest->SetLevelID(0);  // must be level 0 for NullspaceFactory
-    Finest->Set("A", A);
-
-    // permute full matrix
-    Teuchos::RCP<PermutationFactory> PermFact = Teuchos::rcp(new MueLu::PermutationFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>());
-    PermFact->SetParameter("PermutationStrategy",Teuchos::ParameterEntry(std::string("Local")));
-    PermFact->SetParameter("PermutationRowMapName",Teuchos::ParameterEntry(std::string("")));
-    PermFact->SetFactory("PermutationRowMapFactory", Teuchos::null);
-
-    // setup main factory manager
-    Teuchos::RCP<FactoryManager> M = Teuchos::rcp(new FactoryManager());
-    M->SetFactory("permQT",          PermFact);
-    M->SetFactory("A",               MueLu::NoFactory::getRCP()); // this is the input matrix
-    MueLu::SetFactoryManager SFMFinest(Finest, M);
-
-    // prepare building process for permutation operators
-    Finest->Request("A", PermFact.get());
-    Finest->Request("permA", PermFact.get());
-    Finest->Request("permP", PermFact.get());
-    Finest->Request("permQT", PermFact.get());
-    Finest->Request("permScaling", PermFact.get());
-    Finest->Request("#RowPermutations", PermFact.get());
-    Finest->Request("#ColPermutations", PermFact.get());
-    Finest->Request("#WideRangeRowPermutations", PermFact.get());
-    Finest->Request("#WideRangeColPermutations", PermFact.get());
-
-    // build permutation operators
-    PermFact->Build(*Finest);
-
-    //std::cout << "P" <<  *GetEpetraMatrix("permP", Finest, PermFact) << std::endl;
-    //std::cout << "Q^T" << *GetEpetraMatrix("permQT", Finest, PermFact) << std::endl;
-    //std::cout << "permA" <<  *GetEpetraMatrix("A", Finest, PermFact) << std::endl;
-
-    Teuchos::RCP<const Epetra_CrsMatrix> epResult = GetEpetraMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>("A", Finest, PermFact);
-    //std::cout << *epResult << std::endl;
-
-    if(epExpected != Teuchos::null) {
-      Epetra_CrsMatrix* comparison = NULL;
-      EpetraExt::MatrixMatrix::Add(*epResult, false, -1.0, *epExpected, false, 1.0, comparison);
-      comparison->FillComplete();
-      //std::cout << *comparison << std::endl;
-      double norm = comparison->NormInf();
-      delete comparison;
-      comparison = NULL;
-
-      if(norm < 1.0e-14) {
-        *out << "** PASSED **: " << input_filename << std::endl;
-        return true;
-      }
-      else {
-        *out << "-- FAILED --: " << input_filename << std::endl;
-        return false;
-      }
+  if (expected_filename.size() > 0) {
+    int ret2 = EpetraExt::MatlabFileToCrsMatrix(expected_filename.c_str(),
+                                                *Xpetra::toEpetra(comm),
+                                                ptrExpected);
 
+    if (ret2 != 0)
+      std::cout << "failed to read matrix from file" << std::endl;
+  }
+  Teuchos::RCP<Epetra_CrsMatrix> epA        = Teuchos::rcp(ptrA);
+  Teuchos::RCP<Epetra_CrsMatrix> epExpected = Teuchos::rcp(ptrExpected);
+
+  // Epetra_CrsMatrix -> Xpetra::Matrix
+  Teuchos::RCP<CrsMatrix> exA       = Teuchos::rcp(new EpetraCrsMatrix(epA));
+  Teuchos::RCP<CrsMatrixWrap> crsOp = Teuchos::rcp(new CrsMatrixWrap(exA));
+  Teuchos::RCP<Matrix> A            = Teuchos::rcp_dynamic_cast<Matrix>(crsOp);
+  A->SetFixedBlockSize(3);
+
+  Teuchos::RCP<Level> Finest = Teuchos::rcp(new Level());
+  Finest->SetLevelID(0);  // must be level 0 for NullspaceFactory
+  Finest->Set("A", A);
+
+  // permute full matrix
+  Teuchos::RCP<PermutationFactory> PermFact = Teuchos::rcp(new MueLu::PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>());
+  PermFact->SetParameter("PermutationStrategy", Teuchos::ParameterEntry(std::string("Local")));
+  PermFact->SetParameter("PermutationRowMapName", Teuchos::ParameterEntry(std::string("")));
+  PermFact->SetFactory("PermutationRowMapFactory", Teuchos::null);
+
+  // setup main factory manager
+  Teuchos::RCP<FactoryManager> M = Teuchos::rcp(new FactoryManager());
+  M->SetFactory("permQT", PermFact);
+  M->SetFactory("A", MueLu::NoFactory::getRCP());  // this is the input matrix
+  MueLu::SetFactoryManager SFMFinest(Finest, M);
+
+  // prepare building process for permutation operators
+  Finest->Request("A", PermFact.get());
+  Finest->Request("permA", PermFact.get());
+  Finest->Request("permP", PermFact.get());
+  Finest->Request("permQT", PermFact.get());
+  Finest->Request("permScaling", PermFact.get());
+  Finest->Request("#RowPermutations", PermFact.get());
+  Finest->Request("#ColPermutations", PermFact.get());
+  Finest->Request("#WideRangeRowPermutations", PermFact.get());
+  Finest->Request("#WideRangeColPermutations", PermFact.get());
+
+  // build permutation operators
+  PermFact->Build(*Finest);
+
+  // std::cout << "P" <<  *GetEpetraMatrix("permP", Finest, PermFact) << std::endl;
+  // std::cout << "Q^T" << *GetEpetraMatrix("permQT", Finest, PermFact) << std::endl;
+  // std::cout << "permA" <<  *GetEpetraMatrix("A", Finest, PermFact) << std::endl;
+
+  Teuchos::RCP<const Epetra_CrsMatrix> epResult = GetEpetraMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>("A", Finest, PermFact);
+  // std::cout << *epResult << std::endl;
+
+  if (epExpected != Teuchos::null) {
+    Epetra_CrsMatrix* comparison = NULL;
+    EpetraExt::MatrixMatrix::Add(*epResult, false, -1.0, *epExpected, false, 1.0, comparison);
+    comparison->FillComplete();
+    // std::cout << *comparison << std::endl;
+    double norm = comparison->NormInf();
+    delete comparison;
+    comparison = NULL;
+
+    if (norm < 1.0e-14) {
+      *out << "** PASSED **: " << input_filename << std::endl;
+      return true;
+    } else {
+      *out << "-- FAILED --: " << input_filename << std::endl;
+      return false;
     }
-#endif
-    *out << "-- FAILED --: " << input_filename << " no result file found" << std::endl;
-    return false; // no result for comparison available
   }
-
+#endif
+  *out << "-- FAILED --: " << input_filename << " no result file found" << std::endl;
+  return false;  // no result for comparison available
 }
 
+}  // namespace MueLuTests
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
   using Teuchos::RCP;
@@ -299,33 +286,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   Teuchos::oblackholestream blackhole;
 
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  int numProcs = comm->getSize();
+  int numProcs                        = comm->getSize();
 
   bool bSuccess = true;
 #ifndef HAVE_MUELU_INST_COMPLEX_INT_INT
-  //runPermutationTest(MatrixFileName, ExpectedFileName, comm)
-  if(runPermutationTest<Scalar,LocalOrdinal,GlobalOrdinal,Node>("test1.txt", "exp1.txt", comm) == false) bSuccess = false;
-  if(runPermutationTest<Scalar,LocalOrdinal,GlobalOrdinal,Node>("test2.txt", "exp2.txt", comm) == false) bSuccess = false;
-  if(runPermutationTest<Scalar,LocalOrdinal,GlobalOrdinal,Node>("test3.txt", "exp3.txt", comm) == false) bSuccess = false;
+  // runPermutationTest(MatrixFileName, ExpectedFileName, comm)
+  if (runPermutationTest<Scalar, LocalOrdinal, GlobalOrdinal, Node>("test1.txt", "exp1.txt", comm) == false) bSuccess = false;
+  if (runPermutationTest<Scalar, LocalOrdinal, GlobalOrdinal, Node>("test2.txt", "exp2.txt", comm) == false) bSuccess = false;
+  if (runPermutationTest<Scalar, LocalOrdinal, GlobalOrdinal, Node>("test3.txt", "exp3.txt", comm) == false) bSuccess = false;
 
   // the following tests work only on 1 or 2 processors
-  if(numProcs == 1 || numProcs == 2) {
-    if(runPermutationTest2<Scalar,LocalOrdinal,GlobalOrdinal,Node>("test4.txt", "exp4.txt", comm) == false) bSuccess = false;
+  if (numProcs == 1 || numProcs == 2) {
+    if (runPermutationTest2<Scalar, LocalOrdinal, GlobalOrdinal, Node>("test4.txt", "exp4.txt", comm) == false) bSuccess = false;
     // test seems to be ok, but matrix addition is not working
     // has wrong entries on the diagonal on proc1... -> wrong handling of index base?
-    //if(runPermutationTest2("test5.txt", "exp5.txt" /*"exp5.txt"*/, comm) == false) bSuccess = false;
+    // if(runPermutationTest2("test5.txt", "exp5.txt" /*"exp5.txt"*/, comm) == false) bSuccess = false;
   }
 #endif
-  if(bSuccess == false)
+  if (bSuccess == false)
     return EXIT_FAILURE;
   return EXIT_SUCCESS;
 }
 
-
 int main(int argc, char* argv[]) {
   bool success = false;
 
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
   const bool throwExceptions     = false;
   const bool recogniseAllOptions = false;
@@ -333,20 +319,21 @@ int main(int argc, char* argv[]) {
   Teuchos::CommandLineProcessor clp(throwExceptions, recogniseAllOptions);
   Xpetra::Parameters xpetraParameters(clp);
 
-  std::string node = "";  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+  std::string node = "";
+  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
 
   switch (clp.parse(argc, argv, NULL)) {
-    case Teuchos::CommandLineProcessor::PARSE_ERROR:               return EXIT_FAILURE;
+    case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
     case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
 
   if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-    return main_<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
+    return main_<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
 #else
     throw MueLu::Exceptions::RuntimeError("Epetra is not available");
 #endif
@@ -356,6 +343,5 @@ int main(int argc, char* argv[]) {
     std::cout << "Skip permutation tests for Tpetra." << std::endl;
   }
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
-
diff --git a/packages/muelu/test/profiling/Aggregation.cpp b/packages/muelu/test/profiling/Aggregation.cpp
index 8895c9714eed..ca40997dde3b 100644
--- a/packages/muelu/test/profiling/Aggregation.cpp
+++ b/packages/muelu/test/profiling/Aggregation.cpp
@@ -82,10 +82,10 @@
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
 #include "BelosBlockGmresSolMgr.hpp"
-#include "BelosMueLuAdapter.hpp" // this header defines Belos::MueLuOp()
+#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -96,7 +96,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   bool success = false;
   try {
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    RCP<Teuchos::FancyOStream> out = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    RCP<Teuchos::FancyOStream> out      = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
     out->setOutputToRootOnly(0);
 
 #ifndef HAVE_XPETRA_INT_LONG_LONG
@@ -105,24 +105,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Default is Laplace1D with nx = 8748.
     // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-    //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-    Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+    // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
     // custom parameters
-    std::string aggOrdering = "natural";
-    int minPerAgg=2;
-    int maxNbrAlreadySelected=0;
+    std::string aggOrdering   = "natural";
+    int minPerAgg             = 2;
+    int maxNbrAlreadySelected = 0;
 
-    clp.setOption("aggOrdering",&aggOrdering,"aggregation ordering strategy (natural,random,graph)");
-    clp.setOption("minPerAgg",&minPerAgg,"minimum #DOFs per aggregate");
-    clp.setOption("maxNbrSel",&maxNbrAlreadySelected,"maximum # of nbrs allowed to be in other aggregates");
+    clp.setOption("aggOrdering", &aggOrdering, "aggregation ordering strategy (natural,random,graph)");
+    clp.setOption("minPerAgg", &minPerAgg, "minimum #DOFs per aggregate");
+    clp.setOption("maxNbrSel", &maxNbrAlreadySelected, "maximum # of nbrs allowed to be in other aggregates");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     matrixParameters.check();
@@ -138,8 +138,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     /* CREATE INITIAL MATRIX                                                          */
     /**********************************************************************************/
     const RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-    Teuchos::RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+    Teuchos::RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
     RCP<Matrix> A = Pr->BuildMatrix();
 
     //  return EXIT_SUCCESS;
@@ -151,7 +151,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     Finest.SetLevelID(0);  // must be level 0 for NullspaceFactory
     Finest.Set("A", A);
 
-    RCP<FactoryManager> factMngr = rcp( new FactoryManager() );
+    RCP<FactoryManager> factMngr = rcp(new FactoryManager());
     factMngr->SetKokkosRefactor(false);
     Finest.SetFactoryManager(factMngr);
 
@@ -160,14 +160,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     *out << "========================= Aggregate option summary  =========================" << std::endl;
     *out << "min DOFs per aggregate :                " << minPerAgg << std::endl;
     *out << "min # of root nbrs already aggregated : " << maxNbrAlreadySelected << std::endl;
-    UncoupledAggFact.SetMinNodesPerAggregate(minPerAgg);  //TODO should increase if run anything other than 1D
+    UncoupledAggFact.SetMinNodesPerAggregate(minPerAgg);  // TODO should increase if run anything other than 1D
     UncoupledAggFact.SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
     std::transform(aggOrdering.begin(), aggOrdering.end(), aggOrdering.begin(), ::tolower);
     if (aggOrdering == "natural" || aggOrdering == "graph" || aggOrdering == "random") {
-      *out << "aggregate ordering :                    " << aggOrdering<< std::endl;
+      *out << "aggregate ordering :                    " << aggOrdering << std::endl;
       UncoupledAggFact.SetOrdering(aggOrdering);
     } else {
-      std::string msg = "main: bad aggregation option """ + aggOrdering + """.";
+      std::string msg =
+          "main: bad aggregation option "
+          "" +
+          aggOrdering +
+          ""
+          ".";
       throw(MueLu::Exceptions::RuntimeError(msg));
     }
     *out << "=============================================================================" << std::endl;
@@ -178,16 +183,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(true, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/scaling/ComboPDriver.cpp b/packages/muelu/test/scaling/ComboPDriver.cpp
index d96117674f18..1f184e837741 100644
--- a/packages/muelu/test/scaling/ComboPDriver.cpp
+++ b/packages/muelu/test/scaling/ComboPDriver.cpp
@@ -57,7 +57,7 @@
 #include <MueLu_Exceptions.hpp>
 #include <MueLu_ParameterListInterpreter.hpp>
 
-#include "MueLu_SemiCoarsenPFactory.hpp" // for semi-coarsening constants
+#include "MueLu_SemiCoarsenPFactory.hpp"  // for semi-coarsening constants
 #include <MueLu_TestHelpers.hpp>
 #include <MueLu_MultiPhys.hpp>
 
@@ -84,43 +84,43 @@
 #ifdef HAVE_MUELU_TPETRA
 #include <BelosTpetraAdapter.hpp>
 #endif
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
 #endif
 #define FOURBYFOUR
 
 #include <unistd.h>
 /**********************************************************************************/
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
 #ifdef BIGVERSION
-  std::string suffix ("big");
+  std::string suffix("big");
 #else
-  std::string suffix ("");
+  std::string suffix("");
 #endif
   Teuchos::oblackholestream blackhole;
 
   bool success = true;
   bool verbose = true;
   try {
-    Teuchos::RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    Teuchos::RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
 
-    // Four different versions of essentially the same problem can be run. 
-    // depending on whether or not FOURBYFOUR is defined and whether or not 
-    // BIGVERSION is defined  BIGVERSION just uses a finer mesh than the 
+    // Four different versions of essentially the same problem can be run.
+    // depending on whether or not FOURBYFOUR is defined and whether or not
+    // BIGVERSION is defined  BIGVERSION just uses a finer mesh than the
     // standard version. Data for the big and the 2x2 versions is kept
     // in https://gitlab-ex.sandia.gov/muelu/mueludata in multiphys directory
-    // In the description below we use terms like 
-    // filename[big].mat.  When BIGVERSION is defined, this refers to 
+    // In the description below we use terms like
+    // filename[big].mat.  When BIGVERSION is defined, this refers to
     // filenamebig.mat while it refers to filename.mat when BIGVERSION is not
     // defined.  FOURBYFOUR decides whether or not a 4x4 or a 2x2 block system
-    // is solved. 
-    // 
+    // is solved.
+    //
     // The 2 x 2 block system
     //
-    // [ A  B ; C  D] 
+    // [ A  B ; C  D]
     //
     // Presently, the file multiPhys2x2[big].mat is read to define the entries of A, B, C, D.
     // A is supposed to be a 2D elasticity problem while D is a Laplace operator. In particular,
@@ -128,14 +128,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // the B and C associated with multiPhys2x2[big].mat are both zero. The larger version
     // of this problem (when FOURBYFOUR is defined) just replicates this system
     //
-    //             [ A  B 0  0 ; C  D 0 0 ;  0 0 A  B ; 0  0 C  D ] 
+    //             [ A  B 0  0 ; C  D 0 0 ;  0 0 A  B ; 0  0 C  D ]
     //
     // and the file read is multiPhys4x4[big].mat.
     //
-    // AMG is first applied twice for the regular problem. Specifically, the 
-    // file aux1[big].mat defines an auxiliary system that AMG is applied to. 
+    // AMG is first applied twice for the regular problem. Specifically, the
+    // file aux1[big].mat defines an auxiliary system that AMG is applied to.
     // This auxiliary matrix should have the same dimensions as A. The file
-    // aux2[big].mat defines a 2nd auxiliary system that AMG is applied to. 
+    // aux2[big].mat defines a 2nd auxiliary system that AMG is applied to.
     // This auxiliary matrix should have the same dimensions as D. The grid
     // transfer for the 2x2 matrix is a block diagonal matrix
     //
@@ -144,130 +144,130 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // where P1 is pulled out of the hierarchy for aux1[big].mat and P2 is pulled
     // out of the hieararchy for aux2[big].mat (on each level). For the 4x4
     // version, 4 hierarchies are built and a 4x4 block diagonal prolongator
-    // is constructed. 
+    // is constructed.
     //
-    // Additionally, the files coords*[big].mat and null*[big].mat are ready to 
+    // Additionally, the files coords*[big].mat and null*[big].mat are ready to
     // define the auxiliary coordinates and null spaces needed by the AMG
     // algorithm. For our specific problem, the meshes for the elasticity
     // and the Poisson operator partially overlap (as this is similar
     // to what happens on some ARIA ablation problems).
-    
-    int nElas,  nPois;
+
+    int nElas, nPois;
 
     // read in sizes of elasticity and Poisson problem
-    
-    FILE *fp = fopen(std::string("aux1" + suffix + ".mat").c_str(),"r");
+
+    FILE *fp = fopen(std::string("aux1" + suffix + ".mat").c_str(), "r");
     if (fp == NULL) {
-      std::cout << "\nERROR:  File aux1"+suffix+".mat not found"  << std::endl;
+      std::cout << "\nERROR:  File aux1" + suffix + ".mat not found" << std::endl;
       return EXIT_FAILURE;
     }
 
     int ch;
-    int ret_val = 0; (void) ret_val; // suppress fscanf return value and unused variable warnings
-    while (  (ch = getc(fp) ) != '\n')
+    int ret_val = 0;
+    (void)ret_val;  // suppress fscanf return value and unused variable warnings
+    while ((ch = getc(fp)) != '\n')
       ;
-    ret_val = fscanf(fp,"%d",&nElas);
+    ret_val = fscanf(fp, "%d", &nElas);
 
-    fp = fopen(std::string("aux2" + suffix + ".mat").c_str(),"r");
+    fp = fopen(std::string("aux2" + suffix + ".mat").c_str(), "r");
     if (fp == NULL) {
-      std::cout << "\nERROR:  File aux2"+suffix+".mat not found"  << std::endl;
+      std::cout << "\nERROR:  File aux2" + suffix + ".mat not found" << std::endl;
       return EXIT_FAILURE;
     }
-    while (  (ch = getc(fp) ) != '\n')
+    while ((ch = getc(fp)) != '\n')
       ;
-    ret_val = fscanf(fp,"%d",&nPois);
-    fclose(fp); 
+    ret_val = fscanf(fp, "%d", &nPois);
+    fclose(fp);
 
     // check that multiphysics problem and aux matrices are consistent with respect to sizes
 
     int itemp;
 
 #ifdef FOURBYFOUR
-    fp = fopen(std::string("multiPhys4x4" + suffix + ".mat").c_str(),"r");
+    fp = fopen(std::string("multiPhys4x4" + suffix + ".mat").c_str(), "r");
     if (fp == NULL) {
-      std::cout << "\nERROR:  File multiPhys4x4"+suffix+".mat not found"  << std::endl;
+      std::cout << "\nERROR:  File multiPhys4x4" + suffix + ".mat not found" << std::endl;
       return EXIT_FAILURE;
     }
-    while (  (ch = getc(fp) ) != '\n')
+    while ((ch = getc(fp)) != '\n')
       ;
-    ret_val = fscanf(fp,"%d",&itemp);
-    if (itemp != nElas+nPois+nElas+nPois) {
-      std::cout << "\nERROR:  multiPhys4x4"+suffix+".mat dimension is" << itemp << "was expecting it to be " << 2*(nElas+nPois) << std::endl;
+    ret_val = fscanf(fp, "%d", &itemp);
+    if (itemp != nElas + nPois + nElas + nPois) {
+      std::cout << "\nERROR:  multiPhys4x4" + suffix + ".mat dimension is" << itemp << "was expecting it to be " << 2 * (nElas + nPois) << std::endl;
       return EXIT_FAILURE;
     }
 #else
-    fp = fopen(std::string("multiPhys2x2" + suffix + ".mat").c_str(),"r");
+    fp = fopen(std::string("multiPhys2x2" + suffix + ".mat").c_str(), "r");
     if (fp == NULL) {
-      std::cout << "\nERROR:  File multiPhys2x2"+suffix+".mat not found"  << std::endl;
+      std::cout << "\nERROR:  File multiPhys2x2" + suffix + ".mat not found" << std::endl;
       return EXIT_FAILURE;
     }
-    while (  (ch = getc(fp) ) != '\n')
+    while ((ch = getc(fp)) != '\n')
       ;
-    ret_val = fscanf(fp,"%d",&itemp);
-    if (itemp != nElas+nPois) {
-      std::cout << "\nERROR:  multiPhys2x2"+suffix+".mat dimension is" << itemp << "was expecting it to be " << (nElas+nPois) << std::endl;
+    ret_val = fscanf(fp, "%d", &itemp);
+    if (itemp != nElas + nPois) {
+      std::cout << "\nERROR:  multiPhys2x2" + suffix + ".mat dimension is" << itemp << "was expecting it to be " << (nElas + nPois) << std::endl;
       return EXIT_FAILURE;
     }
 #endif
 
-    Teuchos::RCP<const Map>   mapPois = MapFactory::Build(lib, nPois, 0, comm);
-    Teuchos::RCP<const Map>   mapElas1DofPerNode= MapFactory::Build(lib, nElas/2, 0, comm);
-    Teuchos::RCP<const Map>   mapElas2DofPerNode = Xpetra::MapFactory<LO,GO,Node>::Build(mapElas1DofPerNode, 2);
+    Teuchos::RCP<const Map> mapPois            = MapFactory::Build(lib, nPois, 0, comm);
+    Teuchos::RCP<const Map> mapElas1DofPerNode = MapFactory::Build(lib, nElas / 2, 0, comm);
+    Teuchos::RCP<const Map> mapElas2DofPerNode = Xpetra::MapFactory<LO, GO, Node>::Build(mapElas1DofPerNode, 2);
 
-    Teuchos::ArrayView<const GO> GIDsElas2DofPerNode  = mapElas2DofPerNode->getLocalElementList();
-    Teuchos::ArrayView<const GO> GIDsPois  = mapPois->getLocalElementList();
+    Teuchos::ArrayView<const GO> GIDsElas2DofPerNode = mapElas2DofPerNode->getLocalElementList();
+    Teuchos::ArrayView<const GO> GIDsPois            = mapPois->getLocalElementList();
 
 #ifdef FOURBYFOUR
-    Teuchos::Array<GO>  GIDsCombo( 2*(GIDsElas2DofPerNode.size() + GIDsPois.size())); 
+    Teuchos::Array<GO> GIDsCombo(2 * (GIDsElas2DofPerNode.size() + GIDsPois.size()));
 #else
-    Teuchos::Array<GO>  GIDsCombo(GIDsElas2DofPerNode.size() + GIDsPois.size()); 
+    Teuchos::Array<GO> GIDsCombo(GIDsElas2DofPerNode.size() + GIDsPois.size());
 #endif
 
-    for (int ii = 0; ii < GIDsElas2DofPerNode.size(); ii++) GIDsCombo[ii                    ] = GIDsElas2DofPerNode[ii];
-    for (int ii = 0; ii < GIDsPois.size(); ii++) GIDsCombo[ii+ GIDsElas2DofPerNode.size()] = GIDsPois[ii]+nElas;
+    for (int ii = 0; ii < GIDsElas2DofPerNode.size(); ii++) GIDsCombo[ii] = GIDsElas2DofPerNode[ii];
+    for (int ii = 0; ii < GIDsPois.size(); ii++) GIDsCombo[ii + GIDsElas2DofPerNode.size()] = GIDsPois[ii] + nElas;
 #ifdef FOURBYFOUR
-    for (int ii = 0; ii < GIDsElas2DofPerNode.size(); ii++) GIDsCombo[ii+  GIDsElas2DofPerNode.size() + GIDsPois.size()] = GIDsElas2DofPerNode[ii] + nElas + nPois;
-    for (int ii = 0; ii < GIDsPois.size(); ii++)            GIDsCombo[ii+2*GIDsElas2DofPerNode.size() + GIDsPois.size()] = GIDsPois[ii]+2*nElas+nPois;
+    for (int ii = 0; ii < GIDsElas2DofPerNode.size(); ii++) GIDsCombo[ii + GIDsElas2DofPerNode.size() + GIDsPois.size()] = GIDsElas2DofPerNode[ii] + nElas + nPois;
+    for (int ii = 0; ii < GIDsPois.size(); ii++) GIDsCombo[ii + 2 * GIDsElas2DofPerNode.size() + GIDsPois.size()] = GIDsPois[ii] + 2 * nElas + nPois;
 #endif
 
-    Teuchos::RCP<const Map>   mapMultiPhysA = Xpetra::MapFactory<LO,GO,Node>::Build(lib, Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), GIDsCombo, mapElas2DofPerNode->getIndexBase(), comm);
+    Teuchos::RCP<const Map> mapMultiPhysA = Xpetra::MapFactory<LO, GO, Node>::Build(lib, Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(), GIDsCombo, mapElas2DofPerNode->getIndexBase(), comm);
 #ifdef FOURBYFOUR
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >     multiPhysA = Xpetra::IO<SC,LO,GO,Node>::Read("multiPhys4x4" + suffix + ".mat",mapMultiPhysA); 
+    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> multiPhysA = Xpetra::IO<SC, LO, GO, Node>::Read("multiPhys4x4" + suffix + ".mat", mapMultiPhysA);
 #else
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >     multiPhysA = Xpetra::IO<SC,LO,GO,Node>::Read("multiPhys2x2" + suffix + ".mat",mapMultiPhysA); 
+    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> multiPhysA = Xpetra::IO<SC, LO, GO, Node>::Read("multiPhys2x2" + suffix + ".mat", mapMultiPhysA);
 #endif
 
-
-    // To solve the block system a block diagonal auxiliary operator such as 
+    // To solve the block system a block diagonal auxiliary operator such as
     //
-    //     [ aux1    0     0      0 ] 
+    //     [ aux1    0     0      0 ]
     //     [ 0    aux1     0      0 ]
     //     [ 0       0  aux1      0 ]
     //     [ 0       0     0   aux2 ]
     //
     //
     // is used in conjunction with multiple invocations of Hierarchy Setup
-    
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >     aux1 = Xpetra::IO<SC,LO,GO,Node>::Read("aux1" + suffix + ".mat",mapElas2DofPerNode);
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >     aux2 = Xpetra::IO<SC,LO,GO,Node>::Read("aux2" + suffix + ".mat",mapPois);  
+
+    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> aux1 = Xpetra::IO<SC, LO, GO, Node>::Read("aux1" + suffix + ".mat", mapElas2DofPerNode);
+    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> aux2 = Xpetra::IO<SC, LO, GO, Node>::Read("aux2" + suffix + ".mat", mapPois);
 
     typedef Teuchos::ScalarTraits<SC> STS;
     typedef typename STS::magnitudeType real_type;
     typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coordinate_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO>              RealValuedMultiVector;
-    typedef Xpetra::MultiVectorFactory<coordinate_type,LO,GO,NO> RealValuedMultiVectorFactory;
+    typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+    typedef Xpetra::MultiVectorFactory<coordinate_type, LO, GO, NO> RealValuedMultiVectorFactory;
 
-    Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > elasCoords = Xpetra::IO<real_type,LO,GO,Node>::ReadMultiVector("coords1" + suffix + ".mat", mapElas1DofPerNode); 
-    Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > poisCoords = Xpetra::IO<real_type,LO,GO,Node>::ReadMultiVector("coords2" + suffix + ".mat", mapPois); 
+    Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>> elasCoords = Xpetra::IO<real_type, LO, GO, Node>::ReadMultiVector("coords1" + suffix + ".mat", mapElas1DofPerNode);
+    Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node>> poisCoords = Xpetra::IO<real_type, LO, GO, Node>::ReadMultiVector("coords2" + suffix + ".mat", mapPois);
 
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > nullspace1 = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector("null1" + suffix + ".mat", mapElas2DofPerNode);
-    Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > nullspace2 = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector("null2" + suffix + ".mat", mapPois);
+    Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> nullspace1 = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector("null1" + suffix + ".mat", mapElas2DofPerNode);
+    Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>> nullspace2 = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector("null2" + suffix + ".mat", mapPois);
 
-    // Note: Entries of arrayOfNullspaces[] may be set to null. 
+    // Note: Entries of arrayOfNullspaces[] may be set to null.
 #ifdef FOURBYFOUR
     Teuchos::ArrayRCP<Teuchos::RCP<Matrix>> arrayOfAuxMatrices(4);
     Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords(4);
-    Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>           arrayOfNullspaces(4);
+    Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces(4);
 
     arrayOfAuxMatrices[2] = aux1;
     arrayOfAuxMatrices[3] = aux2;
@@ -278,7 +278,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #else
     Teuchos::ArrayRCP<Teuchos::RCP<Matrix>> arrayOfAuxMatrices(2);
     Teuchos::ArrayRCP<Teuchos::RCP<RealValuedMultiVector>> arrayOfCoords(2);
-    Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>>           arrayOfNullspaces(2);
+    Teuchos::ArrayRCP<Teuchos::RCP<MultiVector>> arrayOfNullspaces(2);
 #endif
 
     arrayOfAuxMatrices[0] = aux1;
@@ -288,15 +288,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     arrayOfNullspaces[0]  = nullspace1;
     arrayOfNullspaces[1]  = nullspace2;
 
-    Teuchos::ParameterList  comboList;
+    Teuchos::ParameterList comboList;
     Teuchos::updateParametersFromXmlFileAndBroadcast("comboP.xml", Teuchos::Ptr<Teuchos::ParameterList>(&comboList), *comm);
 #ifdef FOURBYFOUR
-    Teuchos::RCP<Operator> preconditioner = rcp(new MueLu::MultiPhys<SC,LO,GO,NO>(multiPhysA,arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, 4, comboList, true));
+    Teuchos::RCP<Operator> preconditioner = rcp(new MueLu::MultiPhys<SC, LO, GO, NO>(multiPhysA, arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, 4, comboList, true));
 #else
-    Teuchos::RCP<Operator> preconditioner = rcp(new MueLu::MultiPhys<SC,LO,GO,NO>(multiPhysA,arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, 2, comboList, true));
+    Teuchos::RCP<Operator> preconditioner = rcp(new MueLu::MultiPhys<SC, LO, GO, NO>(multiPhysA, arrayOfAuxMatrices, arrayOfNullspaces, arrayOfCoords, 2, comboList, true));
 #endif
 
-    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp (new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
+    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
 
     // =========================================================================
     // System solution (Ax = b)
@@ -315,18 +315,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       Teuchos::Array<typename STS::magnitudeType> norms(1);
       B->norm2(norms);
-      B->scale(one/norms[0]);
+      B->scale(one / norms[0]);
       X->putScalar(zero);
     }
 
     // Belos linear problem
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
 
-
-    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(multiPhysA));
-    Teuchos::RCP<OP> belosPrecOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(preconditioner)); // Turns a Xpetra::Matrix object into a Belos operator
-    Teuchos::RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+    Teuchos::RCP<OP> belosOp                                    = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(multiPhysA));
+    Teuchos::RCP<OP> belosPrecOp                                = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(preconditioner));  // Turns a Xpetra::Matrix object into a Belos operator
+    Teuchos::RCP<Belos::LinearProblem<SC, MV, OP>> belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
     belosProblem->setRightPrec(belosPrecOp);
     bool set = belosProblem->setProblem();
     if (set == false) {
@@ -334,21 +333,21 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       return EXIT_FAILURE;
     }
     Teuchos::RCP<Teuchos::ParameterList> belosList = Teuchos::parameterList();
-    belosList->set("Maximum Iterations",    100); 
+    belosList->set("Maximum Iterations", 100);
     belosList->set("Convergence Tolerance", 1e-6);
-    belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    belosList->set("Output Frequency",      1);
-    belosList->set("Output Style",          Belos::Brief);
+    belosList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    belosList->set("Output Frequency", 1);
+    belosList->set("Output Style", Belos::Brief);
     belosList->set("Implicit Residual Scaling", "None");
 
     Belos::SolverFactory<SC, MV, OP> solverFactory;
-    Teuchos::RCP<Belos::SolverManager<SC, MV, OP> > solver = solverFactory.create("gmres", belosList);
+    Teuchos::RCP<Belos::SolverManager<SC, MV, OP>> solver = solverFactory.create("gmres", belosList);
     solver->setProblem(belosProblem);
     Belos::ReturnType retStatus = Belos::Unconverged;
-    retStatus = solver->solve();
+    retStatus                   = solver->solve();
 
-    int iters = solver -> getNumIters();
-    success = (iters<50 && retStatus == Belos::Converged);
+    int iters = solver->getNumIters();
+    success   = (iters < 50 && retStatus == Belos::Converged);
     if (comm->getRank() == 0) {
       if (success)
         std::cout << "SUCCESS! Belos converged in " << iters << " iterations." << std::endl;
@@ -357,12 +356,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     }
 
     // Timer final summaries
-    globalTimeMonitor = Teuchos::null; // stop this timer before summary
+    globalTimeMonitor = Teuchos::null;  // stop this timer before summary
     Teuchos::TimeMonitor::summarize();
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
@@ -370,5 +369,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/scaling/Driver.cpp b/packages/muelu/test/scaling/Driver.cpp
index 92c80a565f01..46dea9d0d319 100644
--- a/packages/muelu/test/scaling/Driver.cpp
+++ b/packages/muelu/test/scaling/Driver.cpp
@@ -89,15 +89,14 @@
 #include <BelosBlockGmresSolMgr.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
-#include <BelosTpetraAdapter.hpp>    // => This header defines Belos::TpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
+#include <BelosTpetraAdapter.hpp>  // => This header defines Belos::TpetraOp
 #ifdef HAVE_MUELU_EPETRA
-#include <BelosEpetraAdapter.hpp>    // => This header defines Belos::EpetraPrecOp
+#include <BelosEpetraAdapter.hpp>  // => This header defines Belos::EpetraPrecOp
 #endif
 #endif
 
-
 #ifdef HAVE_MUELU_CUDA
 #include "cuda_profiler_api.h"
 #endif
@@ -118,101 +117,93 @@
 #include "Xpetra_EpetraMultiVector.hpp"
 #endif
 
-
 /*********************************************************************/
 
 #include "KokkosBlas1_abs_impl.hpp"
-template<class RV, class XV, class SizeType>
+template <class RV, class XV, class SizeType>
 void Temporary_Replacement_For_Kokkos_abs(const RV& R, const XV& X) {
   typedef typename XV::execution_space execution_space;
   const SizeType numRows = X.extent(0);
-  Kokkos::RangePolicy<execution_space, SizeType> policy (0, numRows);
-  KokkosBlas::Impl::V_Abs_Functor<RV, XV, SizeType> op (R, X);
-  Kokkos::parallel_for (policy, op);
+  Kokkos::RangePolicy<execution_space, SizeType> policy(0, numRows);
+  KokkosBlas::Impl::V_Abs_Functor<RV, XV, SizeType> op(R, X);
+  Kokkos::parallel_for(policy, op);
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void equilibrateMatrix(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &Axpetra, std::string equilibrate) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void equilibrateMatrix(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Axpetra, std::string equilibrate) {
 #include <MueLu_UseShortNames.hpp>
   using Tpetra::computeRowAndColumnOneNorms;
   using Tpetra::leftAndOrRightScaleCrsMatrix;
   bool equilibrate_1norm = (equilibrate == "1-norm");
   bool equilibrate_diag  = (equilibrate == "diag");
   bool equilibrate_no    = (equilibrate == "no");
-  bool assumeSymmetric = false;
-  typedef typename Tpetra::Details::EquilibrationInfo<typename Kokkos::ArithTraits<Scalar>::val_type,typename Node::device_type> equil_type;
-
-  Teuchos::RCP<Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A = Utilities::Op2NonConstTpetraCrs(Axpetra);
-
-  if(Axpetra->getRowMap()->lib() == Xpetra::UseTpetra) {
-     equil_type equibResult_ = computeRowAndColumnOneNorms (*A, assumeSymmetric);
-     if (equilibrate_1norm) {
-        using device_type = typename Node::device_type;
-        using mag_type = typename Kokkos::ArithTraits<Scalar>::mag_type;
-        using mag_view_type = Kokkos::View<mag_type*, device_type>;
-        using scalar_view_type = Kokkos::View<typename equil_type::val_type*, device_type>;
-
-        mag_view_type rowDiagAbsVals ("rowDiagAbsVals",equibResult_.rowDiagonalEntries.extent (0));
-        //        KokkosBlas::abs (rowDiagAbsVals, equibResult_.rowDiagonalEntries);
-        Temporary_Replacement_For_Kokkos_abs<mag_view_type,scalar_view_type,LocalOrdinal>(rowDiagAbsVals, equibResult_.rowDiagonalEntries);
-
-        mag_view_type colDiagAbsVals ("colDiagAbsVals",equibResult_.colDiagonalEntries.extent (0));
-
-        //        KokkosBlas::abs (colDiagAbsVals, equibResult_.colDiagonalEntries);
-        Temporary_Replacement_For_Kokkos_abs<mag_view_type,scalar_view_type,LocalOrdinal>(colDiagAbsVals, equibResult_.colDiagonalEntries);
-
-        leftAndOrRightScaleCrsMatrix (*A, rowDiagAbsVals, colDiagAbsVals,
-                                      true, true, equibResult_.assumeSymmetric,
-                                      Tpetra::SCALING_DIVIDE);
-     }
-     else if (equilibrate_diag) {
-        auto colScalingFactors = equibResult_.assumeSymmetric ?
-          equibResult_.colNorms :
-          equibResult_.rowScaledColNorms;
-        leftAndOrRightScaleCrsMatrix (*A, equibResult_.rowNorms,
-                                      colScalingFactors, true, true,
-                                      equibResult_.assumeSymmetric,
-                                      Tpetra::SCALING_DIVIDE);
-      }
-     else if (equilibrate_no) {
-       // no-op
-     }
-     else
-       throw std::runtime_error("Invalid 'equilibrate' option '"+equilibrate+"'");
+  bool assumeSymmetric   = false;
+  typedef typename Tpetra::Details::EquilibrationInfo<typename Kokkos::ArithTraits<Scalar>::val_type, typename Node::device_type> equil_type;
+
+  Teuchos::RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A = Utilities::Op2NonConstTpetraCrs(Axpetra);
+
+  if (Axpetra->getRowMap()->lib() == Xpetra::UseTpetra) {
+    equil_type equibResult_ = computeRowAndColumnOneNorms(*A, assumeSymmetric);
+    if (equilibrate_1norm) {
+      using device_type      = typename Node::device_type;
+      using mag_type         = typename Kokkos::ArithTraits<Scalar>::mag_type;
+      using mag_view_type    = Kokkos::View<mag_type*, device_type>;
+      using scalar_view_type = Kokkos::View<typename equil_type::val_type*, device_type>;
+
+      mag_view_type rowDiagAbsVals("rowDiagAbsVals", equibResult_.rowDiagonalEntries.extent(0));
+      //        KokkosBlas::abs (rowDiagAbsVals, equibResult_.rowDiagonalEntries);
+      Temporary_Replacement_For_Kokkos_abs<mag_view_type, scalar_view_type, LocalOrdinal>(rowDiagAbsVals, equibResult_.rowDiagonalEntries);
+
+      mag_view_type colDiagAbsVals("colDiagAbsVals", equibResult_.colDiagonalEntries.extent(0));
+
+      //        KokkosBlas::abs (colDiagAbsVals, equibResult_.colDiagonalEntries);
+      Temporary_Replacement_For_Kokkos_abs<mag_view_type, scalar_view_type, LocalOrdinal>(colDiagAbsVals, equibResult_.colDiagonalEntries);
+
+      leftAndOrRightScaleCrsMatrix(*A, rowDiagAbsVals, colDiagAbsVals,
+                                   true, true, equibResult_.assumeSymmetric,
+                                   Tpetra::SCALING_DIVIDE);
+    } else if (equilibrate_diag) {
+      auto colScalingFactors = equibResult_.assumeSymmetric ? equibResult_.colNorms : equibResult_.rowScaledColNorms;
+      leftAndOrRightScaleCrsMatrix(*A, equibResult_.rowNorms,
+                                   colScalingFactors, true, true,
+                                   equibResult_.assumeSymmetric,
+                                   Tpetra::SCALING_DIVIDE);
+    } else if (equilibrate_no) {
+      // no-op
+    } else
+      throw std::runtime_error("Invalid 'equilibrate' option '" + equilibrate + "'");
   }
 }
 
-
 /*********************************************************************/
 // Gets current memory usage in kilobytes
-size_t get_current_memory_usage()
-{
-  size_t memory = 0; 
-  
+size_t get_current_memory_usage() {
+  size_t memory = 0;
+
   // darwin reports rusage.ru_maxrss in bytes
 #if defined(__APPLE__) || defined(__MACH__)
-  const size_t RU_MAXRSS_UNITS=1024;
+  const size_t RU_MAXRSS_UNITS = 1024;
 #else
-  const size_t RU_MAXRSS_UNITS=1;
+  const size_t RU_MAXRSS_UNITS = 1;
 #endif
-  
+
   struct rusage sys_resources;
   getrusage(RUSAGE_SELF, &sys_resources);
   memory = (unsigned long)sys_resources.ru_maxrss / RU_MAXRSS_UNITS;
-  
+
   /* Success */
   return memory;
 }
 /*********************************************************************/
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -224,71 +215,109 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // =========================================================================
   typedef Teuchos::ScalarTraits<SC> STS;
   typedef typename STS::coordinateType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "";                 clp.setOption("xml",                   &xmlFileName,       "read parameters from an xml file");
-  std::string yamlFileName      = "";                 clp.setOption("yaml",                  &yamlFileName,      "read parameters from a yaml file");
-  bool        printTimings      = true;               clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-  std::string timingsFormat     = "table-fixed";      clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-  int         writeMatricesOPT  = -2;                 clp.setOption("write",                 &writeMatricesOPT,  "write matrices to file (-1 means all; i>=0 means level i)");
-  std::string dsolveType        = "belos", solveType; clp.setOption("solver",                &dsolveType,        "solve type: (none | belos | standalone | matvec)");
-  std::string belosType         = "cg";               clp.setOption("belosType",             &belosType,         "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
-  double      dtol              = 1e-12, tol;         clp.setOption("tol",                   &dtol,              "solver convergence tolerance");
-  bool        binaryFormat      = false;              clp.setOption("binary", "ascii",       &binaryFormat,      "print timings to screen");
-
-  std::string rowMapFile;                             clp.setOption("rowmap",                &rowMapFile,        "map data file");
-  std::string colMapFile;                             clp.setOption("colmap",                &colMapFile,        "colmap data file");
-  std::string domainMapFile;                          clp.setOption("domainmap",             &domainMapFile,     "domainmap data file");
-  std::string rangeMapFile;                           clp.setOption("rangemap",              &rangeMapFile,      "rangemap data file");
-  std::string matrixFile;                             clp.setOption("matrix",                &matrixFile,        "matrix data file");
-  std::string rhsFile;                                clp.setOption("rhs",                   &rhsFile,           "rhs data file");
-  std::string solFile;                                clp.setOption("sol",                   &solFile,           "write the solution to this file");
-  std::string coordFile;                              clp.setOption("coords",                &coordFile,         "coordinates data file");
-  std::string coordMapFile;                           clp.setOption("coordsmap",             &coordMapFile,      "coordinates map data file");
-  std::string nullFile;                               clp.setOption("nullspace",             &nullFile,          "nullspace data file");
-  std::string materialFile;                           clp.setOption("material",              &materialFile,      "material data file");
-  bool        setNullSpace      = true;               clp.setOption("driver-nullspace","muelu-computed-nullspace", &setNullSpace, "driver sets nullspace");
-  int         numRebuilds       = 0;                  clp.setOption("rebuild",               &numRebuilds,       "#times to rebuild hierarchy");
-  int         numResolves       = 0;                  clp.setOption("resolve",               &numResolves,       "#times to redo solve");
-  int         maxIts            = 200;                clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-  int         numVectors        = 1;                  clp.setOption("multivector",           &numVectors,        "number of rhs to solve simultaneously");
-  bool        scaleResidualHist = true;               clp.setOption("scale", "noscale",      &scaleResidualHist, "scaled Krylov residual history");
-  bool        solvePreconditioned = true;             clp.setOption("solve-preconditioned","no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
-  bool        useStackedTimer   = false;              clp.setOption("stacked-timer","no-stacked-timer", &useStackedTimer, "use stacked timer");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+  std::string yamlFileName = "";
+  clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  std::string timingsFormat = "table-fixed";
+  clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+  int writeMatricesOPT = -2;
+  clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+  std::string dsolveType = "belos", solveType;
+  clp.setOption("solver", &dsolveType, "solve type: (none | belos | standalone | matvec)");
+  std::string belosType = "cg";
+  clp.setOption("belosType", &belosType, "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
+  double dtol = 1e-12, tol;
+  clp.setOption("tol", &dtol, "solver convergence tolerance");
+  bool binaryFormat = false;
+  clp.setOption("binary", "ascii", &binaryFormat, "print timings to screen");
+
+  std::string rowMapFile;
+  clp.setOption("rowmap", &rowMapFile, "map data file");
+  std::string colMapFile;
+  clp.setOption("colmap", &colMapFile, "colmap data file");
+  std::string domainMapFile;
+  clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+  std::string rangeMapFile;
+  clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+  std::string matrixFile;
+  clp.setOption("matrix", &matrixFile, "matrix data file");
+  std::string rhsFile;
+  clp.setOption("rhs", &rhsFile, "rhs data file");
+  std::string solFile;
+  clp.setOption("sol", &solFile, "write the solution to this file");
+  std::string coordFile;
+  clp.setOption("coords", &coordFile, "coordinates data file");
+  std::string coordMapFile;
+  clp.setOption("coordsmap", &coordMapFile, "coordinates map data file");
+  std::string nullFile;
+  clp.setOption("nullspace", &nullFile, "nullspace data file");
+  std::string materialFile;
+  clp.setOption("material", &materialFile, "material data file");
+  bool setNullSpace = true;
+  clp.setOption("driver-nullspace", "muelu-computed-nullspace", &setNullSpace, "driver sets nullspace");
+  int numRebuilds = 0;
+  clp.setOption("rebuild", &numRebuilds, "#times to rebuild hierarchy");
+  int numResolves = 0;
+  clp.setOption("resolve", &numResolves, "#times to redo solve");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  int numVectors = 1;
+  clp.setOption("multivector", &numVectors, "number of rhs to solve simultaneously");
+  bool scaleResidualHist = true;
+  clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
+  bool solvePreconditioned = true;
+  clp.setOption("solve-preconditioned", "no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
+  bool useStackedTimer = false;
+  clp.setOption("stacked-timer", "no-stacked-timer", &useStackedTimer, "use stacked timer");
   std::string watchrProblemName = std::string("MueLu Setup-Solve ") + std::to_string(comm->getSize()) + " ranks";
-                                                      clp.setOption("watchr-problem-name",   &watchrProblemName,   "Problem name for Watchr plot headers");
+  clp.setOption("watchr-problem-name", &watchrProblemName, "Problem name for Watchr plot headers");
 
-  std::string equilibrate = "no" ;                    clp.setOption("equilibrate",           &equilibrate,       "equilibrate the system (no | diag | 1-norm)");
+  std::string equilibrate = "no";
+  clp.setOption("equilibrate", &equilibrate, "equilibrate the system (no | diag | 1-norm)");
 #ifdef HAVE_MUELU_CUDA
-  bool profileSetup = false;                          clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
-  bool profileSolve = false;                          clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
-#else
   bool profileSetup = false;
+  clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
   bool profileSolve = false;
+  clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
+#else
+  bool profileSetup            = false;
+  bool profileSolve            = false;
 #endif
-  int  cacheSize = 0;                                 clp.setOption("cachesize",               &cacheSize,       "cache size (in KB)");
+  int cacheSize = 0;
+  clp.setOption("cachesize", &cacheSize, "cache size (in KB)");
 #ifdef HAVE_MPI
-  int provideNodeComm = 0;                            clp.setOption("nodecomm",          &provideNodeComm,  "make the nodal communicator available w/ reduction factor X");
+  int provideNodeComm = 0;
+  clp.setOption("nodecomm", &provideNodeComm, "make the nodal communicator available w/ reduction factor X");
 #endif
-  std::string userBlkFileName = "";                   clp.setOption("userBlks",              &userBlkFileName,   "read user smoother blocks from MatrixMarket matrix file. nnz (i,j) ==> jth dof in ith block");
-  int numReruns = 1;                                  clp.setOption("reruns",                &numReruns,  "number of reruns");
-  std::string rerunFilePrefix;                             clp.setOption("fileprefix",              &rerunFilePrefix,      "if doing reruns, optional prefix to prepend to output files");
-  std::string rerunFileSuffix;                             clp.setOption("filesuffix",              &rerunFileSuffix,      "if doing reruns, optional suffix to append to output files");
-  std::string  levelPerformanceModel  = "no";          clp.setOption("performance-model", &levelPerformanceModel,  "runs the level-by-level performance mode options- 'no', 'yes' or 'verbose'");
+  std::string userBlkFileName = "";
+  clp.setOption("userBlks", &userBlkFileName, "read user smoother blocks from MatrixMarket matrix file. nnz (i,j) ==> jth dof in ith block");
+  int numReruns = 1;
+  clp.setOption("reruns", &numReruns, "number of reruns");
+  std::string rerunFilePrefix;
+  clp.setOption("fileprefix", &rerunFilePrefix, "if doing reruns, optional prefix to prepend to output files");
+  std::string rerunFileSuffix;
+  clp.setOption("filesuffix", &rerunFileSuffix, "if doing reruns, optional suffix to append to output files");
+  std::string levelPerformanceModel = "no";
+  clp.setOption("performance-model", &levelPerformanceModel, "runs the level-by-level performance mode options- 'no', 'yes' or 'verbose'");
   clp.recogniseAllOptions(true);
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   TEUCHOS_TEST_FOR_EXCEPTION(xmlFileName != "" && yamlFileName != "", std::runtime_error,
@@ -296,7 +325,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   ParameterList paramList;
@@ -314,13 +343,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   if (inst == Xpetra::COMPLEX_INT_INT && dsolveType == "belos") {
     belosType = "gmres";
-    out << "WARNING: CG will not work with COMPLEX scalars, switching to GMRES"<<std::endl;
+    out << "WARNING: CG will not work with COMPLEX scalars, switching to GMRES" << std::endl;
   }
 
 #ifdef HAVE_MUELU_AMGX
-//Initialize AMGX
-MueLu::MueLu_AMGX_initialize();
-MueLu::MueLu_AMGX_initialize_plugins();
+  // Initialize AMGX
+  MueLu::MueLu_AMGX_initialize();
+  MueLu::MueLu_AMGX_initialize_plugins();
 #endif
 
   bool isDriver = paramList.isSublist("Run1");
@@ -335,7 +364,7 @@ MueLu::MueLu_AMGX_initialize_plugins();
     }
 
     // Galeri updates (only works with Run1)
-    if(paramList.sublist("Run1").isSublist("Galeri")) {
+    if (paramList.sublist("Run1").isSublist("Galeri")) {
       ParameterList& moreParams = paramList.sublist("Run1").sublist("Galeri");
       for (ParameterList::ConstIterator it = moreParams.begin(); it != moreParams.end(); it++) {
         const std::string& name = moreParams.name(it);
@@ -344,13 +373,13 @@ MueLu::MueLu_AMGX_initialize_plugins();
       }
     }
   }
-  
+
 #ifdef HAVE_MPI
   // Generate the node-level communicator, if we want one
   Teuchos::RCP<const Teuchos::Comm<int> > nodeComm;
   int NodeId = comm->getRank();
-  if(provideNodeComm) {
-    nodeComm = MueLu::GenerateNodeComm(comm,NodeId,provideNodeComm);
+  if (provideNodeComm) {
+    nodeComm = MueLu::GenerateNodeComm(comm, NodeId, provideNodeComm);
     //    printf("DEBUG: Base rank %d => New, node %d, rank %d\n",comm->getRank(),NodeId,nodeComm->getRank());
   }
 #endif
@@ -365,40 +394,36 @@ MueLu::MueLu_AMGX_initialize_plugins();
   std::ostringstream galeriStream;
 #ifdef HAVE_MUELU_OPENMP
   std::string node_name = Node::name();
-  if(!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
-    galeriStream<<"OpenMP Max Threads = "<<omp_get_max_threads()<<std::endl;
+  if (!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
+    galeriStream << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
 #endif
 
-
   comm->barrier();
   Teuchos::RCP<Teuchos::StackedTimer> stacked_timer;
-  //Constructing the globalTimeMonitor should only be done if not using StackedTimer.
-  //This is because if a StackedTimer is already active, globalTimer will be become a sub-timer of the root.
+  // Constructing the globalTimeMonitor should only be done if not using StackedTimer.
+  // This is because if a StackedTimer is already active, globalTimer will be become a sub-timer of the root.
   RCP<TimeMonitor> globalTimeMonitor = Teuchos::null;
-  if(useStackedTimer)
-  {
+  if (useStackedTimer) {
     stacked_timer = rcp(new Teuchos::StackedTimer("MueLu_Driver"));
     Teuchos::TimeMonitor::setStackedTimer(stacked_timer);
-  }
-  else
+  } else
     globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-
-  RCP<Matrix>      A;
-  RCP<const Map>   map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<RealValuedMultiVector> coordinates;
-  RCP<Xpetra::MultiVector<SC,LO,GO,NO> > nullspace, material;
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > nullspace, material;
   RCP<MultiVector> X, B;
 
   // Load the matrix off disk (or generate it via Galeri)
-  MatrixLoad<SC,LO,GO,NO>(comm,lib,binaryFormat,matrixFile,rhsFile,rowMapFile,colMapFile,domainMapFile,rangeMapFile,coordFile,coordMapFile,nullFile,materialFile,map,A,coordinates,nullspace,material,X,B,numVectors,galeriParameters,xpetraParameters,galeriStream);
+  MatrixLoad<SC, LO, GO, NO>(comm, lib, binaryFormat, matrixFile, rhsFile, rowMapFile, colMapFile, domainMapFile, rangeMapFile, coordFile, coordMapFile, nullFile, materialFile, map, A, coordinates, nullspace, material, X, B, numVectors, galeriParameters, xpetraParameters, galeriStream);
   comm->barrier();
   tm = Teuchos::null;
 
   // Do equilibration if requested
-  if(lib == Xpetra::UseTpetra) {
-    equilibrateMatrix(A,equilibrate);
+  if (lib == Xpetra::UseTpetra) {
+    equilibrateMatrix(A, equilibrate);
   }
 
   bool resetStackedTimer = false;
@@ -410,11 +435,11 @@ MueLu::MueLu_AMGX_initialize_plugins();
     ParameterList mueluList, runList;
     const bool mustAlreadyExist = true;
     if (isDriver) {
-      runList   = paramList.sublist("Run1",  mustAlreadyExist);
-      mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+      runList   = paramList.sublist("Run1", mustAlreadyExist);
+      mueluList = runList.sublist("MueLu", mustAlreadyExist);
     } else {
       mueluList = paramList;
-      stop = true;
+      stop      = true;
     }
 
     if (nullspace.is_null()) {
@@ -433,13 +458,13 @@ MueLu::MueLu_AMGX_initialize_plugins();
       nullspace = MultiVectorFactory::Build(map, blkSize);
       for (int i = 0; i < blkSize; i++) {
         RCP<const Map> domainMap = A->getDomainMap();
-        GO             indexBase = domainMap->getIndexBase();
+        GO indexBase             = domainMap->getIndexBase();
 
         ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
         for (int j = 0; j < nsData.size(); j++) {
           GO GID = domainMap->getGlobalElement(j) - indexBase;
 
-          if ((GID-i) % blkSize == 0)
+          if ((GID - i) % blkSize == 0)
             nsData[j] = Teuchos::ScalarTraits<SC>::one();
         }
       }
@@ -448,17 +473,16 @@ MueLu::MueLu_AMGX_initialize_plugins();
     // If doing user based block smoothing, read block information from file.
     // Must do this for both smoothers and coarse solvers
 
-    readUserBlks<SC,LO,GO,NO> ( userBlkFileName, "coarse", mueluList, A);
-    readUserBlks<SC,LO,GO,NO> ( userBlkFileName, "smoother", mueluList, A);
+    readUserBlks<SC, LO, GO, NO>(userBlkFileName, "coarse", mueluList, A);
+    readUserBlks<SC, LO, GO, NO>(userBlkFileName, "smoother", mueluList, A);
 
-    int runCount = 1;
-    int   savedOut  = -1;
+    int runCount    = 1;
+    int savedOut    = -1;
     FILE* openedOut = NULL;
     do {
       solveType = dsolveType;
       tol       = dtol;
 
-
       if (isDriver) {
         if (runList.isParameter("filename")) {
           // Redirect all output into a filename We have to redirect all output,
@@ -478,7 +502,7 @@ MueLu::MueLu_AMGX_initialize_plugins();
           dup2(fileno(openedOut), STDOUT_FILENO);
         }
         if (runList.isParameter("solver")) solveType = runList.get<std::string>("solver");
-        if (runList.isParameter("tol"))    tol       = runList.get<double>     ("tol");
+        if (runList.isParameter("tol")) tol = runList.get<double>("tol");
 
         if (resetStackedTimer) {
           stacked_timer = rcp(new Teuchos::StackedTimer("MueLu_Driver"));
@@ -487,7 +511,7 @@ MueLu::MueLu_AMGX_initialize_plugins();
       }
 
       RCP<Teuchos::FancyOStream> fancy2 = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      Teuchos::FancyOStream& out2 = *fancy2;
+      Teuchos::FancyOStream& out2       = *fancy2;
       out2.setOutputToRootOnly(0);
       out2 << galeriStream.str();
 
@@ -497,27 +521,26 @@ MueLu::MueLu_AMGX_initialize_plugins();
       bool useAMGX = mueluList.isParameter("use external multigrid package") && (mueluList.get<std::string>("use external multigrid package") == "amgx");
       bool useML   = mueluList.isParameter("use external multigrid package") && (mueluList.get<std::string>("use external multigrid package") == "ml");
 #ifdef HAVE_MPI
-      if(provideNodeComm && !useAMGX && !useML) {
+      if (provideNodeComm && !useAMGX && !useML) {
         Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
-        userParamList.set("Node Comm",nodeComm);
+        userParamList.set("Node Comm", nodeComm);
       }
 #endif
-      out2<<"*********** MueLu ParameterList ***********"<<std::endl;
-      out2<<mueluList;
-      out2<<"*******************************************"<<std::endl;
+      out2 << "*********** MueLu ParameterList ***********" << std::endl;
+      out2 << mueluList;
+      out2 << "*******************************************" << std::endl;
 
       RCP<Hierarchy> H;
       RCP<Operator> Prec;
       // Build the preconditioner numRebuilds+1 times
-      MUELU_SWITCH_TIME_MONITOR(tm,"Driver: 2 - MueLu Setup");
-      PreconditionerSetup(A,coordinates,nullspace,material,mueluList,profileSetup,useAMGX,useML,setNullSpace,numRebuilds,H,Prec);
+      MUELU_SWITCH_TIME_MONITOR(tm, "Driver: 2 - MueLu Setup");
+      PreconditionerSetup(A, coordinates, nullspace, material, mueluList, profileSetup, useAMGX, useML, setNullSpace, numRebuilds, H, Prec);
 
       comm->barrier();
       tm = Teuchos::null;
 
-
       size_t mem = get_current_memory_usage();
-      out2<<"Memory use after preconditioner setup (GB): " << (mem/1024.0/1024.0)<<std::endl;
+      out2 << "Memory use after preconditioner setup (GB): " << (mem / 1024.0 / 1024.0) << std::endl;
 
       // =========================================================================
       // System solution (Ax = b)
@@ -527,44 +550,39 @@ MueLu::MueLu_AMGX_initialize_plugins();
         if (writeMatricesOPT > -2) {
           tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3.5 - Matrix output")));
           H->Write(writeMatricesOPT, writeMatricesOPT);
-          if(writeMatricesOPT == 0 || writeMatricesOPT == -1) {
-            Xpetra::IO<SC,LO,GO,NO>::Write("b_0.m",*B);
+          if (writeMatricesOPT == 0 || writeMatricesOPT == -1) {
+            Xpetra::IO<SC, LO, GO, NO>::Write("b_0.m", *B);
           }
           tm = Teuchos::null;
         }
 
         // Solve the system numResolves+1 times
-        SystemSolve(A,X,B,H,Prec,out2,solveType,belosType,profileSolve,useAMGX,useML,cacheSize,numResolves,scaleResidualHist,solvePreconditioned,maxIts,tol);
+        SystemSolve(A, X, B, H, Prec, out2, solveType, belosType, profileSolve, useAMGX, useML, cacheSize, numResolves, scaleResidualHist, solvePreconditioned, maxIts, tol);
 
         comm->barrier();
-      }
-      catch(const std::exception& e) {
+      } catch (const std::exception& e) {
         if (isDriver)
-          out2<<"MueLu_Driver: solver crashed w/ message:"<<e.what()<<std::endl;
+          out2 << "MueLu_Driver: solver crashed w/ message:" << e.what() << std::endl;
         else
           throw;
       }
 
-
       tm = Teuchos::null;
 
-
       // If we want Level-specific performance model diagnostics, now is the time!
-      if( (levelPerformanceModel=="yes" || levelPerformanceModel=="verbose")
-          && !H.is_null()) {
-        for(int i=0; i < H->GetNumLevels(); i++) {
+      if ((levelPerformanceModel == "yes" || levelPerformanceModel == "verbose") && !H.is_null()) {
+        for (int i = 0; i < H->GetNumLevels(); i++) {
           RCP<Level> level = H->GetLevel(i);
           try {
-            RCP<Matrix> A_level = level->Get<RCP<Matrix> >("A");            
+            RCP<Matrix> A_level    = level->Get<RCP<Matrix> >("A");
             std::string level_name = std::string("Level-") + std::to_string(i) + std::string(": ");
-            std::vector<const char *> timers;//MueLu: Laplace2D: Hierarchy: Solve (level=0)  
-            MueLu::report_spmv_performance_models<Matrix>(A_level,100,timers,globalTimeMonitor,level_name,levelPerformanceModel=="verbose");
+            std::vector<const char*> timers;  // MueLu: Laplace2D: Hierarchy: Solve (level=0)
+            MueLu::report_spmv_performance_models<Matrix>(A_level, 100, timers, globalTimeMonitor, level_name, levelPerformanceModel == "verbose");
+          } catch (...) {
+            ;
           }
-          catch(...) {;}
-        }        
+        }
       }
-     
-
 
       globalTimeMonitor = Teuchos::null;
       if (useStackedTimer)
@@ -573,13 +591,13 @@ MueLu::MueLu_AMGX_initialize_plugins();
       if (printTimings) {
         RCP<ParameterList> reportParams = rcp(new ParameterList);
         if (timingsFormat == "yaml") {
-          reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-          reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
+          reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+          reportParams->set("YAML style", "compact");  // "spacious" or "compact"
         }
-        reportParams->set("How to merge timer sets",   "Union");
-        reportParams->set("alwaysWriteLocal",          false);
-        reportParams->set("writeGlobalStats",          true);
-        reportParams->set("writeZeroTimers",           false);
+        reportParams->set("How to merge timer sets", "Union");
+        reportParams->set("alwaysWriteLocal", false);
+        reportParams->set("writeGlobalStats", true);
+        reportParams->set("writeZeroTimers", false);
         // FIXME: no "ignoreZeroTimers"
 
         const std::string filter = "";
@@ -590,13 +608,14 @@ MueLu::MueLu_AMGX_initialize_plugins();
           options.output_fraction = options.output_histogram = options.output_minmax = true;
           stacked_timer->report(out2, comm, options);
           auto xmlOut = stacked_timer->reportWatchrXML(watchrProblemName, comm);
-          if(xmlOut.length())
+          if (xmlOut.length())
             std::cout << "\nAlso created Watchr performance report " << xmlOut << '\n';
-        }
-        else {
+        } else {
           std::ios_base::fmtflags ff(out2.flags());
-          if (timingsFormat == "table-fixed") out2 << std::fixed;
-          else                                out2 << std::scientific;
+          if (timingsFormat == "table-fixed")
+            out2 << std::fixed;
+          else
+            out2 << std::scientific;
           TimeMonitor::report(comm.ptr(), out, filter, reportParams);
           out2 << std::setiosflags(ff);
         }
@@ -608,7 +627,7 @@ MueLu::MueLu_AMGX_initialize_plugins();
       if (isDriver) {
         try {
           runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
-          mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+          mueluList = runList.sublist("MueLu", mustAlreadyExist);
         } catch (Teuchos::Exceptions::InvalidParameterName&) {
           stop = true;
         }
@@ -625,17 +644,15 @@ MueLu::MueLu_AMGX_initialize_plugins();
       openedOut = NULL;
     }
 
-
-  }//end reruns
+  }  // end reruns
 
   if (solFile != "")
-    Xpetra::IO<SC,LO,GO,Node>::Write(solFile, *X);
-
+    Xpetra::IO<SC, LO, GO, Node>::Write(solFile, *X);
 
 #ifdef HAVE_MUELU_AMGX
-// Finalize AMGX
-MueLu::MueLu_AMGX_finalize_plugins();
-MueLu::MueLu_AMGX_finalize();
+  // Finalize AMGX
+  MueLu::MueLu_AMGX_finalize_plugins();
+  MueLu::MueLu_AMGX_finalize();
 #endif
 
   return EXIT_SUCCESS;
@@ -645,6 +662,6 @@ MueLu::MueLu_AMGX_finalize();
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/scaling/DriverCore.hpp b/packages/muelu/test/scaling/DriverCore.hpp
index 778d37327e9c..de7c00566156 100644
--- a/packages/muelu/test/scaling/DriverCore.hpp
+++ b/packages/muelu/test/scaling/DriverCore.hpp
@@ -54,7 +54,6 @@
 #include <Xpetra_Vector.hpp>
 #include <Xpetra_Matrix.hpp>
 
-
 // Belos
 #ifdef HAVE_MUELU_BELOS
 #include <BelosConfigDefs.hpp>
@@ -63,27 +62,27 @@
 #include <BelosBlockGmresSolMgr.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 
-#include <BelosTpetraAdapter.hpp>    // => This header defines Belos::TpetraOp
+#include <BelosTpetraAdapter.hpp>  // => This header defines Belos::TpetraOp
 #include <Xpetra_TpetraOperator.hpp>
 #include <MueLu_TpetraOperator.hpp>
 #include <MueLu_CreateTpetraPreconditioner.hpp>
 namespace BelosTpetra {
 namespace Impl {
-  extern void register_Cg (const bool verbose);
-  extern void register_CgPipeline (const bool verbose);
-  extern void register_CgSingleReduce (const bool verbose);
-  extern void register_Gmres (const bool verbose);
-  extern void register_GmresS (const bool verbose);
-  extern void register_GmresSstep (const bool verbose);
-  extern void register_GmresSingleReduce (const bool verbose);
-} // namespace Impl
-} // namespace BelosTpetra
+extern void register_Cg(const bool verbose);
+extern void register_CgPipeline(const bool verbose);
+extern void register_CgSingleReduce(const bool verbose);
+extern void register_Gmres(const bool verbose);
+extern void register_GmresS(const bool verbose);
+extern void register_GmresSstep(const bool verbose);
+extern void register_GmresSingleReduce(const bool verbose);
+}  // namespace Impl
+}  // namespace BelosTpetra
 
 #ifdef HAVE_MUELU_EPETRA
-#include <BelosEpetraAdapter.hpp>    // => This header defines Belos::EpetraPrecOp
+#include <BelosEpetraAdapter.hpp>  // => This header defines Belos::EpetraPrecOp
 #endif
 #endif
 
@@ -97,18 +96,17 @@ namespace Impl {
 #include <MueLu_AMGXOperator.hpp>
 #endif
 
-
 #include <MueLu_CreateXpetraPreconditioner.hpp>
 
-
-
 //*************************************************************************************
 //*************************************************************************************
 //*************************************************************************************
 // A handy macro to switch time monitors in a StackedTimer-compatible way
-#define MUELU_SWITCH_TIME_MONITOR(tm,timername)                                  \
-  {tm=Teuchos::null; tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(timername)));}
-
+#define MUELU_SWITCH_TIME_MONITOR(tm, timername)                     \
+  {                                                                  \
+    tm = Teuchos::null;                                              \
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(timername))); \
+  }
 
 //*************************************************************************************
 //*************************************************************************************
@@ -119,117 +117,113 @@ namespace Impl {
 #include <ml_MultiLevelPreconditioner.h>
 
 // Helper functions for compilation purposes
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-struct ML_Wrapper{
-  static void Generate_ML_MultiLevelPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A, Teuchos::ParameterList & mueluList,
-                                                   Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & mlopX) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct ML_Wrapper {
+  static void Generate_ML_MultiLevelPreconditioner(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A, Teuchos::ParameterList& mueluList,
+                                                   Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& mlopX) {
     throw std::runtime_error("Template parameter mismatch");
   }
 };
 
-
-template<class GlobalOrdinal>
-struct ML_Wrapper<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
-  static void Generate_ML_MultiLevelPreconditioner(Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& A,Teuchos::ParameterList & mueluList,
-                                                   Teuchos::RCP<Xpetra::Operator<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& mlopX) {
+template <class GlobalOrdinal>
+struct ML_Wrapper<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
+  static void Generate_ML_MultiLevelPreconditioner(Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>>& A, Teuchos::ParameterList& mueluList,
+                                                   Teuchos::RCP<Xpetra::Operator<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>>& mlopX) {
     typedef double SC;
     typedef int LO;
     typedef GlobalOrdinal GO;
     typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode NO;
-    Teuchos::RCP<const Epetra_CrsMatrix> Aep   = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(A);
-    Teuchos::RCP<Epetra_Operator> mlop  = Teuchos::rcp<Epetra_Operator>(new ML_Epetra::MultiLevelPreconditioner(*Aep,mueluList));
+    Teuchos::RCP<const Epetra_CrsMatrix> Aep = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(A);
+    Teuchos::RCP<Epetra_Operator> mlop       = Teuchos::rcp<Epetra_Operator>(new ML_Epetra::MultiLevelPreconditioner(*Aep, mueluList));
 #if defined(HAVE_MUELU_BELOS)
     // NOTE: Belos needs the Apply() and AppleInverse() routines of ML swapped.  So...
     mlop = Teuchos::rcp<Belos::EpetraPrecOp>(new Belos::EpetraPrecOp(mlop));
 #endif
 
-    mlopX = Teuchos::rcp(new Xpetra::EpetraOperator<GO,NO>(mlop));
+    mlopX = Teuchos::rcp(new Xpetra::EpetraOperator<GO, NO>(mlop));
   }
 };
 #endif
 
-
 //*************************************************************************************
 //*************************************************************************************
 //*************************************************************************************
 // This is a standard setup routine
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void PreconditionerSetup(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A,
-                         Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType,LocalOrdinal,GlobalOrdinal,Node> > & coordinates,
-                         Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & nullspace,
-                         Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & material,
-                         Teuchos::ParameterList & mueluList,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void PreconditionerSetup(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                         Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::coordinateType, LocalOrdinal, GlobalOrdinal, Node>>& coordinates,
+                         Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& nullspace,
+                         Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& material,
+                         Teuchos::ParameterList& mueluList,
                          bool profileSetup,
                          bool useAMGX,
                          bool useML,
                          bool setNullSpace,
                          int numRebuilds,
-                         Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & H,
-                         Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & Prec) {
+                         Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& H,
+                         Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& Prec) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   Xpetra::UnderlyingLib lib = A->getRowMap()->lib();
   typedef typename Teuchos::ScalarTraits<SC>::coordinateType coordinate_type;
-  typedef Xpetra::MultiVector<coordinate_type,LO,GO,NO> CoordinateMultiVector;
+  typedef Xpetra::MultiVector<coordinate_type, LO, GO, NO> CoordinateMultiVector;
   // =========================================================================
   // Preconditioner construction
   // =========================================================================
 #ifdef HAVE_MUELU_CUDA
-  if(profileSetup) cudaProfilerStart();
+  if (profileSetup) cudaProfilerStart();
 #endif
 
-   if(useML && lib != Xpetra::UseEpetra) throw std::runtime_error("Error: Cannot use ML on non-epetra matrices");
+  if (useML && lib != Xpetra::UseEpetra) throw std::runtime_error("Error: Cannot use ML on non-epetra matrices");
 
-   for (int i = 0; i <= numRebuilds; i++) {
-     A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
-     if(useAMGX) {
+  for (int i = 0; i <= numRebuilds; i++) {
+    A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
+    if (useAMGX) {
 #if defined(HAVE_MUELU_AMGX)
-       RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > Ac      = Utilities::Op2NonConstTpetraCrs(A);
-       RCP<Tpetra::Operator<SC,LO,GO,NO> > At       = Teuchos::rcp_dynamic_cast<Tpetra::Operator<SC,LO,GO,NO> >(Ac);
-       RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > Top = MueLu::CreateTpetraPreconditioner(At, mueluList);
-       Prec = Teuchos::rcp(new Xpetra::TpetraOperator<SC,LO,GO,NO>(Top));
+      RCP<Tpetra::CrsMatrix<SC, LO, GO, NO>> Ac      = Utilities::Op2NonConstTpetraCrs(A);
+      RCP<Tpetra::Operator<SC, LO, GO, NO>> At       = Teuchos::rcp_dynamic_cast<Tpetra::Operator<SC, LO, GO, NO>>(Ac);
+      RCP<MueLu::TpetraOperator<SC, LO, GO, NO>> Top = MueLu::CreateTpetraPreconditioner(At, mueluList);
+      Prec                                           = Teuchos::rcp(new Xpetra::TpetraOperator<SC, LO, GO, NO>(Top));
 #endif
-     } else if(useML) {
+    } else if (useML) {
 #if defined(HAVE_MUELU_ML) and defined(HAVE_MUELU_EPETRA)
-       mueluList.remove("use external multigrid package");
-       if(!coordinates.is_null()) {
-         RCP<const Epetra_MultiVector> epetraCoord =  MueLu::Utilities<coordinate_type,LO,GO,NO>::MV2EpetraMV(coordinates);
-         if(epetraCoord->NumVectors() > 0)  mueluList.set("x-coordinates",(*epetraCoord)[0]);
-         if(epetraCoord->NumVectors() > 1)  mueluList.set("y-coordinates",(*epetraCoord)[1]);
-         if(epetraCoord->NumVectors() > 2)  mueluList.set("z-coordinates",(*epetraCoord)[2]);
-       }
-       if(!material.is_null()) {
-         RCP<const Epetra_MultiVector> epetraMat =  MueLu::Utilities<SC,LO,GO,NO>::MV2EpetraMV(material);
-         mueluList.set("material coordinates",(*epetraMat)[0]);
-       }
-       ML_Wrapper<SC, LO, GO, NO>::Generate_ML_MultiLevelPreconditioner(A,mueluList,Prec);
+      mueluList.remove("use external multigrid package");
+      if (!coordinates.is_null()) {
+        RCP<const Epetra_MultiVector> epetraCoord = MueLu::Utilities<coordinate_type, LO, GO, NO>::MV2EpetraMV(coordinates);
+        if (epetraCoord->NumVectors() > 0) mueluList.set("x-coordinates", (*epetraCoord)[0]);
+        if (epetraCoord->NumVectors() > 1) mueluList.set("y-coordinates", (*epetraCoord)[1]);
+        if (epetraCoord->NumVectors() > 2) mueluList.set("z-coordinates", (*epetraCoord)[2]);
+      }
+      if (!material.is_null()) {
+        RCP<const Epetra_MultiVector> epetraMat = MueLu::Utilities<SC, LO, GO, NO>::MV2EpetraMV(material);
+        mueluList.set("material coordinates", (*epetraMat)[0]);
+      }
+      ML_Wrapper<SC, LO, GO, NO>::Generate_ML_MultiLevelPreconditioner(A, mueluList, Prec);
 #endif
-     }
-     else {
-       Teuchos::Array<LO> lNodesPerDim(3, 10);
-       Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
-       if(!coordinates.is_null())
-         userParamList.set<RCP<CoordinateMultiVector> >("Coordinates", coordinates);
-       if (!nullspace.is_null() && setNullSpace )
-         userParamList.set<RCP<Xpetra::MultiVector<SC,LO,GO,NO>> >("Nullspace", nullspace);
-       userParamList.set<Teuchos::Array<LO> >("Array<LO> lNodesPerDim", lNodesPerDim);
-       H = MueLu::CreateXpetraPreconditioner(A, mueluList);
-     }
-   }
+    } else {
+      Teuchos::Array<LO> lNodesPerDim(3, 10);
+      Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
+      if (!coordinates.is_null())
+        userParamList.set<RCP<CoordinateMultiVector>>("Coordinates", coordinates);
+      if (!nullspace.is_null() && setNullSpace)
+        userParamList.set<RCP<Xpetra::MultiVector<SC, LO, GO, NO>>>("Nullspace", nullspace);
+      userParamList.set<Teuchos::Array<LO>>("Array<LO> lNodesPerDim", lNodesPerDim);
+      H = MueLu::CreateXpetraPreconditioner(A, mueluList);
+    }
+  }
 #ifdef HAVE_MUELU_CUDA
-   if(profileSetup) cudaProfilerStop();
+  if (profileSetup) cudaProfilerStop();
 #endif
 }
 
-
 #if defined(HAVE_MUELU_EPETRA)
 
 // Helper functions for compilation purposes
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-struct Matvec_Wrapper{
-  static void UnwrapEpetra(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                           Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& X,
-                           Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& B,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct Matvec_Wrapper {
+  static void UnwrapEpetra(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                           Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                           Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B,
                            Teuchos::RCP<const Epetra_CrsMatrix>& Aepetra,
                            Teuchos::RCP<Epetra_MultiVector>& Xepetra,
                            Teuchos::RCP<Epetra_MultiVector>& Bepetra) {
@@ -237,12 +231,11 @@ struct Matvec_Wrapper{
   }
 };
 
-
-template<class GlobalOrdinal>
-struct Matvec_Wrapper<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
-  static void UnwrapEpetra(Teuchos::RCP<Xpetra::Matrix<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& A,
-                           Teuchos::RCP<Xpetra::MultiVector<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& X,
-                           Teuchos::RCP<Xpetra::MultiVector<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& B,
+template <class GlobalOrdinal>
+struct Matvec_Wrapper<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode> {
+  static void UnwrapEpetra(Teuchos::RCP<Xpetra::Matrix<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>>& A,
+                           Teuchos::RCP<Xpetra::MultiVector<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>>& X,
+                           Teuchos::RCP<Xpetra::MultiVector<double, int, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode>>& B,
                            Teuchos::RCP<const Epetra_CrsMatrix>& Aepetra,
                            Teuchos::RCP<Epetra_MultiVector>& Xepetra,
                            Teuchos::RCP<Epetra_MultiVector>& Bepetra) {
@@ -251,20 +244,20 @@ struct Matvec_Wrapper<double,int,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSeria
     typedef GlobalOrdinal GO;
     typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode NO;
     Aepetra = Xpetra::Helpers<SC, LO, GO, NO>::Op2EpetraCrs(A);
-    Xepetra = Teuchos::rcp(& Xpetra::toEpetra(*X),false);
-    Bepetra = Teuchos::rcp(& Xpetra::toEpetra(*B),false);
+    Xepetra = Teuchos::rcp(&Xpetra::toEpetra(*X), false);
+    Bepetra = Teuchos::rcp(&Xpetra::toEpetra(*B), false);
   }
 };
 #endif
 
 //*************************************************************************************
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & A,
-                 Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & X,
-                 Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & B,
-                 Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node > > & H,
-                 Teuchos::RCP<Xpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node > > & Prec,
-                 Teuchos::FancyOStream & out,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                 Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& X,
+                 Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& B,
+                 Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& H,
+                 Teuchos::RCP<Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& Prec,
+                 Teuchos::FancyOStream& out,
                  std::string solveType,
                  std::string belosType,
                  bool profileSolve,
@@ -289,29 +282,29 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
   int min = 0, max = 10;
   int numInts = 0;
   if (cacheSize > 0) {
-    cacheSize *= 1024; //convert to bytes
-    numInts = cacheSize/sizeof(int) + 1;
+    cacheSize *= 1024;  // convert to bytes
+    numInts = cacheSize / sizeof(int) + 1;
     tempVector.resize(numInts);
   }
 
   // Get the raw matrices for matvec testing
-  Teuchos::RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > Atpetra;
-  Teuchos::RCP<Tpetra::MultiVector<SC,LO,GO,NO> > Xtpetra,Btpetra;
-  if(lib==Xpetra::UseTpetra) {
+  Teuchos::RCP<Tpetra::CrsMatrix<SC, LO, GO, NO>> Atpetra;
+  Teuchos::RCP<Tpetra::MultiVector<SC, LO, GO, NO>> Xtpetra, Btpetra;
+  if (lib == Xpetra::UseTpetra) {
     Atpetra = Utilities::Op2NonConstTpetraCrs(A);
-    Xtpetra = rcp(& Xpetra::toTpetra(*X),false);
-    Btpetra = rcp(& Xpetra::toTpetra(*B),false);
+    Xtpetra = rcp(&Xpetra::toTpetra(*X), false);
+    Btpetra = rcp(&Xpetra::toTpetra(*B), false);
   }
 
 #if defined(HAVE_MUELU_EPETRA)
   Teuchos::RCP<const Epetra_CrsMatrix> Aepetra;
-  Teuchos::RCP<Epetra_MultiVector> Xepetra,Bepetra;
-  if(lib==Xpetra::UseEpetra) {
-    Matvec_Wrapper<SC,LO,GO,NO>::UnwrapEpetra(A,X,B,Aepetra,Xepetra,Bepetra);
+  Teuchos::RCP<Epetra_MultiVector> Xepetra, Bepetra;
+  if (lib == Xpetra::UseEpetra) {
+    Matvec_Wrapper<SC, LO, GO, NO>::UnwrapEpetra(A, X, B, Aepetra, Xepetra, Bepetra);
   }
 #endif
 
-  for(int solveno = 0; solveno<=numResolves; solveno++) {
+  for (int solveno = 0; solveno <= numResolves; solveno++) {
     RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - LHS and RHS initialization")));
     X->putScalar(zero);
     tm = Teuchos::null;
@@ -320,76 +313,73 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
       // Do not perform a solve
     } else if (solveType == "matvec") {
       // Just do matvecs
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 6 - Matvec")));
-      if(lib==Xpetra::UseTpetra)
-        Atpetra->apply(*Btpetra,*Xtpetra);
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 6 - Matvec")));
+      if (lib == Xpetra::UseTpetra)
+        Atpetra->apply(*Btpetra, *Xtpetra);
 
 #if defined(HAVE_MUELU_EPETRA) && !defined(HAVE_MUELU_INST_COMPLEX_INT_INT) && !defined(HAVE_MUELU_INST_FLOAT_INT_INT)
-      if(lib==Xpetra::UseEpetra) Aepetra->Apply(*Bepetra,*Xepetra);
+      if (lib == Xpetra::UseEpetra) Aepetra->Apply(*Bepetra, *Xepetra);
 #endif
-      //clear the cache (and don't time it)
-      tm = Teuchos::null;
+      // clear the cache (and don't time it)
+      tm      = Teuchos::null;
       int ttt = rand();
-      for (int i=0; i<numInts; ++i)
+      for (int i = 0; i < numInts; ++i)
         tempVector[i] += (min + (ttt % static_cast<int>(max - min + 1)));
     } else if (solveType == "standalone") {
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - Fixed Point Solve")));
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - Fixed Point Solve")));
 #ifdef HAVE_MUELU_CUDA
-      if(profileSolve) cudaProfilerStart();
+      if (profileSolve) cudaProfilerStart();
 #endif
-      if(useAMGX) {
+      if (useAMGX) {
 #if defined(HAVE_MUELU_AMGX)
         // Do a fixed-point iteraiton without convergence checks
         RCP<MultiVector> R = MultiVectorFactory::Build(X->getMap(), X->getNumVectors());
-        for(int i=0; i<maxIts; i++) {
+        for (int i = 0; i < maxIts; i++) {
           Utilities::Residual(*A, *X, *B, *R);
-          Prec->apply(*R,*X);
+          Prec->apply(*R, *X);
         }
 #endif
-      }
-      else {
+      } else {
         H->IsPreconditioner(false);
-        std::pair<LocalOrdinal,typename STS::magnitudeType> maxItsTol(maxIts, tol);
+        std::pair<LocalOrdinal, typename STS::magnitudeType> maxItsTol(maxIts, tol);
         H->Iterate(*B, *X, maxItsTol);
       }
 #ifdef HAVE_MUELU_CUDA
-      if(profileSolve) cudaProfilerStop();
+      if (profileSolve) cudaProfilerStop();
 #endif
     } else if (solveType == "belos") {
 #ifdef HAVE_MUELU_BELOS
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Belos Solve")));
 #ifdef HAVE_MUELU_CUDA
-      if(profileSolve) cudaProfilerStart();
+      if (profileSolve) cudaProfilerStart();
 #endif
       // Operator and Multivector type that will be used with Belos
-      typedef MultiVector          MV;
+      typedef MultiVector MV;
       typedef Belos::OperatorT<MV> OP;
 
       // Define Operator and Preconditioner
-      Teuchos::RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-      Teuchos::RCP<OP> belosPrec; // Turns a MueLu::Hierarchy object into a Belos operator
-      if(useAMGX) {
+      Teuchos::RCP<OP> belosOp = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+      Teuchos::RCP<OP> belosPrec;                                              // Turns a MueLu::Hierarchy object into a Belos operator
+      if (useAMGX) {
 #if defined(HAVE_MUELU_AMGX)
-        belosPrec = rcp(new Belos::XpetraOp <SC, LO, GO, NO>(Prec)); // Turns an Xpetra::Operator object into a Belos operator
+        belosPrec = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(Prec));  // Turns an Xpetra::Operator object into a Belos operator
 #endif
-      }
-      else if(useML) {
+      } else if (useML) {
 #if defined(HAVE_MUELU_ML) and defined(HAVE_MUELU_EPETRA)
-        belosPrec = rcp(new Belos::XpetraOp <SC, LO, GO, NO>(Prec)); // Turns an Xpetra::Operator object into a Belos operator
+        belosPrec = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(Prec));  // Turns an Xpetra::Operator object into a Belos operator
 #endif
-      }
-      else {
+      } else {
         H->IsPreconditioner(true);
-        belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+        belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
       }
 
       // Belos parameter list
       RCP<Teuchos::ParameterList> belosList = Teuchos::parameterList();
-      belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-      belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-      belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-      belosList->set("Output Frequency",      1);
-      belosList->set("Output Style",          Belos::Brief);
+      belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+      belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+      belosList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+      belosList->set("Output Frequency", 1);
+      belosList->set("Output Style", Belos::Brief);
       if (!scaleResidualHist)
         belosList->set("Implicit Residual Scaling", "None");
 
@@ -397,35 +387,33 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
       Belos::ReturnType ret = Belos::Unconverged;
 
       constexpr bool verbose = false;
-      BelosTpetra::Impl::register_Cg (verbose);
-      BelosTpetra::Impl::register_CgPipeline (verbose);
-      BelosTpetra::Impl::register_CgSingleReduce (verbose);
-      BelosTpetra::Impl::register_Gmres (verbose);
-      BelosTpetra::Impl::register_GmresS (verbose);
-      BelosTpetra::Impl::register_GmresSstep (verbose);
-      BelosTpetra::Impl::register_GmresSingleReduce (verbose);
+      BelosTpetra::Impl::register_Cg(verbose);
+      BelosTpetra::Impl::register_CgPipeline(verbose);
+      BelosTpetra::Impl::register_CgSingleReduce(verbose);
+      BelosTpetra::Impl::register_Gmres(verbose);
+      BelosTpetra::Impl::register_GmresS(verbose);
+      BelosTpetra::Impl::register_GmresSstep(verbose);
+      BelosTpetra::Impl::register_GmresSingleReduce(verbose);
 
       try {
-
-        TEUCHOS_TEST_FOR_EXCEPTION(lib!=Xpetra::UseTpetra, std::invalid_argument, "Need to use Tpetra backend in order to call a Belos Tpetra solver.");
+        TEUCHOS_TEST_FOR_EXCEPTION(lib != Xpetra::UseTpetra, std::invalid_argument, "Need to use Tpetra backend in order to call a Belos Tpetra solver.");
 
         using tMV = Tpetra::MultiVector<SC, LO, GO, NO>;
         using tOP = Tpetra::Operator<SC, LO, GO, NO>;
 
         Teuchos::RCP<tOP> belosPrecTpetra;
-        if(useAMGX) {
+        if (useAMGX) {
 #if defined(HAVE_MUELU_AMGX)
-          RCP<Xpetra::TpetraOperator<SC,LO,GO,NO> > xto = Teuchos::rcp_dynamic_cast<Xpetra::TpetraOperator<SC,LO,GO,NO> >(Prec);
-          belosPrecTpetra = xto->getOperator();
+          RCP<Xpetra::TpetraOperator<SC, LO, GO, NO>> xto = Teuchos::rcp_dynamic_cast<Xpetra::TpetraOperator<SC, LO, GO, NO>>(Prec);
+          belosPrecTpetra                                 = xto->getOperator();
 #endif
-        }
-        else {
-          belosPrecTpetra = rcp(new MueLu::TpetraOperator<SC,LO,GO,NO>(H));
+        } else {
+          belosPrecTpetra = rcp(new MueLu::TpetraOperator<SC, LO, GO, NO>(H));
         }
 
         // Construct a Belos LinearProblem object
-        RCP<Belos::LinearProblem<SC, tMV, tOP> > belosProblem = rcp(new Belos::LinearProblem<SC, tMV, tOP>(Atpetra, Xtpetra, Btpetra));
-        if(solvePreconditioned) belosProblem->setRightPrec(belosPrecTpetra);
+        RCP<Belos::LinearProblem<SC, tMV, tOP>> belosProblem = rcp(new Belos::LinearProblem<SC, tMV, tOP>(Atpetra, Xtpetra, Btpetra));
+        if (solvePreconditioned) belosProblem->setRightPrec(belosPrecTpetra);
 
         bool set = belosProblem->setProblem();
         if (set == false) {
@@ -434,19 +422,17 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
 
         // Create an iterative solver manager
         Belos::SolverFactory<SC, tMV, tOP> solverFactory;
-        RCP< Belos::SolverManager<SC, tMV, tOP> > solver = solverFactory.create(belosType, belosList);
+        RCP<Belos::SolverManager<SC, tMV, tOP>> solver = solverFactory.create(belosType, belosList);
         solver->setProblem(belosProblem);
 
         // Perform solve
-        ret = solver->solve();
+        ret    = solver->solve();
         numIts = solver->getNumIters();
 
-      } catch (std::invalid_argument&)
-      {
-
+      } catch (std::invalid_argument&) {
         // Construct a Belos LinearProblem object
-        RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
-        if(solvePreconditioned) belosProblem->setRightPrec(belosPrec);
+        RCP<Belos::LinearProblem<SC, MV, OP>> belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+        if (solvePreconditioned) belosProblem->setRightPrec(belosPrec);
 
         bool set = belosProblem->setProblem();
         if (set == false) {
@@ -455,13 +441,12 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
 
         // Create an iterative solver manager
         Belos::SolverFactory<SC, MV, OP> solverFactory;
-        RCP< Belos::SolverManager<SC, MV, OP> > solver = solverFactory.create(belosType, belosList);
+        RCP<Belos::SolverManager<SC, MV, OP>> solver = solverFactory.create(belosType, belosList);
         solver->setProblem(belosProblem);
 
         // Perform solve
-        ret = solver->solve();
+        ret    = solver->solve();
         numIts = solver->getNumIters();
-
       }
 
       // Get the number of iterations for this solve.
@@ -469,18 +454,19 @@ void SystemSolve(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,N
 
       // Check convergence
       if (ret != Belos::Converged)
-        out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+        out << std::endl
+            << "ERROR:  Belos did not converge! " << std::endl;
       else
-        out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+        out << std::endl
+            << "SUCCESS:  Belos converged!" << std::endl;
 #ifdef HAVE_MUELU_CUDA
-      if(profileSolve) cudaProfilerStop();
+      if (profileSolve) cudaProfilerStop();
 #endif
-#endif //ifdef HAVE_MUELU_BELOS
+#endif  // ifdef HAVE_MUELU_BELOS
     } else {
       throw MueLu::Exceptions::RuntimeError("Unknown solver type: \"" + solveType + "\"");
     }
-  }// end resolves
-
+  }  // end resolves
 }
 
 #endif
diff --git a/packages/muelu/test/scaling/ImportPerformance.cpp b/packages/muelu/test/scaling/ImportPerformance.cpp
index 7babdd99b30a..e97a33b90a67 100644
--- a/packages/muelu/test/scaling/ImportPerformance.cpp
+++ b/packages/muelu/test/scaling/ImportPerformance.cpp
@@ -90,13 +90,13 @@ using Teuchos::TimeMonitor;
 // =========================================================================
 // =========================================================================
 // Provide a "resize" operation for double*'s.
-inline void resize_doubles(int nold,int nnew,double*& d){
-  if(nnew > nold){
-    double *tmp = new double[nnew];
-    for(int i=0; i<nold; i++)
-      tmp[i]=d[i];
-    delete [] d;
-    d=tmp;
+inline void resize_doubles(int nold, int nnew, double*& d) {
+  if (nnew > nold) {
+    double* tmp = new double[nnew];
+    for (int i = 0; i < nold; i++)
+      tmp[i] = d[i];
+    delete[] d;
+    d = tmp;
   }
 }
 
@@ -106,19 +106,17 @@ inline void resize_doubles(int nold,int nnew,double*& d){
 #if defined(HAVE_MUELU_EPETRA)
 
 extern void MakeColMapAndReindexSort(int& NumRemoteColGIDs, int*& RemoteColindices,
-                                std::vector<int>& RemotePermuteIDs, std::vector<int>& RemoteOwningPIDs);
+                                     std::vector<int>& RemotePermuteIDs, std::vector<int>& RemoteOwningPIDs);
 
 extern void MakeColMapAndReindexSort(int& NumRemoteColGIDs, long long*& RemoteColindices,
-                                std::vector<int>& RemotePermuteIDs, std::vector<int>& RemoteOwningPIDs);
+                                     std::vector<int>& RemotePermuteIDs, std::vector<int>& RemoteOwningPIDs);
 
-
-
-void build_remote_pids(int MyPID,const std::vector<int> &ColMapOwningPIDs,std::vector<int>&RemotePIDs) {
+void build_remote_pids(int MyPID, const std::vector<int>& ColMapOwningPIDs, std::vector<int>& RemotePIDs) {
   // Presume the column map has Aztec ordering
-  int N=(int)ColMapOwningPIDs.size();
+  int N = (int)ColMapOwningPIDs.size();
   int first_idx;
-  for(first_idx=0;first_idx<N; first_idx++)
-    if(ColMapOwningPIDs[first_idx]!=MyPID)
+  for (first_idx = 0; first_idx < N; first_idx++)
+    if (ColMapOwningPIDs[first_idx] != MyPID)
       break;
 
   /*   printf("[%d] ColMapOwningPIDs(%d) =",MyPID,(int)ColMapOwningPIDs.size());
@@ -127,94 +125,92 @@ void build_remote_pids(int MyPID,const std::vector<int> &ColMapOwningPIDs,std::v
      printf("\n");*/
 
   // Make sure there are some non-local unknowns
-   if(first_idx == N) {
-     printf("[%d] No remotes\n",MyPID);
-     return;
-   }
+  if (first_idx == N) {
+    printf("[%d] No remotes\n", MyPID);
+    return;
+  }
 
   RemotePIDs.resize(ColMapOwningPIDs.size() - first_idx);
-  for(int i=first_idx; i<N; i++)
-    RemotePIDs[i-first_idx] = ColMapOwningPIDs[i];
+  for (int i = first_idx; i < N; i++)
+    RemotePIDs[i - first_idx] = ColMapOwningPIDs[i];
 
   /*   printf("[%d] RemotePIDs(%d) =",MyPID,(int)RemotePIDs.size());
    for(int i=0;i<(int)RemotePIDs.size(); i++)
      printf("%d ",RemotePIDs[i]);
      printf("\n");*/
-
 }
 
-
-Epetra_Map * convert_lightweightmap_to_map(const EpetraExt::LightweightMap & A, const Epetra_Comm & Comm){
-    Epetra_Map *Aout=0;
-    if(A.GlobalIndicesInt()) {
+Epetra_Map* convert_lightweightmap_to_map(const EpetraExt::LightweightMap& A, const Epetra_Comm& Comm) {
+  Epetra_Map* Aout = 0;
+  if (A.GlobalIndicesInt()) {
 #ifndef EPETRA_NO_32BIT_GLOBAL_INDICES
-      Aout = new Epetra_Map(-1,A.NumMyElements(),A.MyGlobalElements(),0,Comm);
+    Aout = new Epetra_Map(-1, A.NumMyElements(), A.MyGlobalElements(), 0, Comm);
 #endif
-    }
-    else if (A.GlobalIndicesLongLong()) {
+  } else if (A.GlobalIndicesLongLong()) {
 #ifndef EPETRA_NO_64BIT_GLOBAL_INDICES
-      Aout = new Epetra_Map((long long)-1,A.NumMyElements(),A.MyGlobalElements64(),(long long)0,Comm);
+    Aout = new Epetra_Map((long long)-1, A.NumMyElements(), A.MyGlobalElements64(), (long long)0, Comm);
 #endif
-    }
+  }
 
-    return Aout;
+  return Aout;
 }
 
-
-Epetra_CrsMatrix* convert_lightweightcrsmatrix_to_crsmatrix(const EpetraExt::LightweightCrsMatrix & A) {
-  auto  tm = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: MapConstructor");
-  const Epetra_Comm & Comm = A.DomainMap_.Comm();
+Epetra_CrsMatrix* convert_lightweightcrsmatrix_to_crsmatrix(const EpetraExt::LightweightCrsMatrix& A) {
+  auto tm                 = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: MapConstructor");
+  const Epetra_Comm& Comm = A.DomainMap_.Comm();
 
   // Build Maps
   Epetra_Map *RowMap, *ColMap;
-  if(A.use_lw) RowMap = convert_lightweightmap_to_map(*A.RowMapLW_,Comm);
-  else throw std::runtime_error("Only works in LW mode");
-  ColMap = convert_lightweightmap_to_map(A.ColMap_,Comm);
-  Epetra_CrsMatrix *Aout = new Epetra_CrsMatrix(Copy,*RowMap,*ColMap,0);
-  int N=RowMap->NumMyElements();
-  int nnz = A.colind_.size();
+  if (A.use_lw)
+    RowMap = convert_lightweightmap_to_map(*A.RowMapLW_, Comm);
+  else
+    throw std::runtime_error("Only works in LW mode");
+  ColMap                 = convert_lightweightmap_to_map(A.ColMap_, Comm);
+  Epetra_CrsMatrix* Aout = new Epetra_CrsMatrix(Copy, *RowMap, *ColMap, 0);
+  int N                  = RowMap->NumMyElements();
+  int nnz                = A.colind_.size();
 
   tm = Teuchos::null;
   // Copy pointers over
-  auto  tm2= TimeMonitor::getNewTimer("OptimizedTransfer: Convert: Data Copy");
-  Epetra_IntSerialDenseVector & rowptr = Aout->ExpertExtractIndexOffset();
-  Epetra_IntSerialDenseVector & colind = Aout->ExpertExtractIndices();
-  double *& vals = Aout->ExpertExtractValues();
-  rowptr.Resize(N+1);
+  auto tm2                            = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: Data Copy");
+  Epetra_IntSerialDenseVector& rowptr = Aout->ExpertExtractIndexOffset();
+  Epetra_IntSerialDenseVector& colind = Aout->ExpertExtractIndices();
+  double*& vals                       = Aout->ExpertExtractValues();
+  rowptr.Resize(N + 1);
   colind.Resize(nnz);
-  resize_doubles(0,nnz,vals);
+  resize_doubles(0, nnz, vals);
 
-  for(int i=0; i<N+1; i++)
+  for (int i = 0; i < N + 1; i++)
     rowptr[i] = A.rowptr_[i];
 
-  for(int i=0; i<nnz; i++) {
+  for (int i = 0; i < nnz; i++) {
     colind[i] = A.colind_[i];
-    vals[i]  = A.vals_[i];
+    vals[i]   = A.vals_[i];
   }
-  tm2 = Teuchos::null;
-  auto tm3= TimeMonitor::getNewTimer("OptimizedTransfer: Convert: BuildRemote");
+  tm2      = Teuchos::null;
+  auto tm3 = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: BuildRemote");
 
   // Get RemotePIDs
   std::vector<int> RemotePIDs_;
-  build_remote_pids(Comm.MyPID(),A.ColMapOwningPIDs_,RemotePIDs_);
+  build_remote_pids(Comm.MyPID(), A.ColMapOwningPIDs_, RemotePIDs_);
 
-  tm3 = Teuchos::null;
-  auto tm4= TimeMonitor::getNewTimer("OptimizedTransfer: Convert: BuildImport");
+  tm3      = Teuchos::null;
+  auto tm4 = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: BuildImport");
 
   // Importer build
-  const int * ExportLIDs = A.ExportLIDs_.size()?&A.ExportLIDs_[0]:0;
-  const int * ExportPIDs = A.ExportPIDs_.size()?&A.ExportPIDs_[0]:0;
-  const int * RemotePIDs = RemotePIDs_.size()?&RemotePIDs_[0]:0;
-  Epetra_Import * Importer = new Epetra_Import(*ColMap,A.DomainMap_,RemotePIDs_.size(),RemotePIDs,A.ExportLIDs_.size(),ExportLIDs,ExportPIDs);
+  const int* ExportLIDs   = A.ExportLIDs_.size() ? &A.ExportLIDs_[0] : 0;
+  const int* ExportPIDs   = A.ExportPIDs_.size() ? &A.ExportPIDs_[0] : 0;
+  const int* RemotePIDs   = RemotePIDs_.size() ? &RemotePIDs_[0] : 0;
+  Epetra_Import* Importer = new Epetra_Import(*ColMap, A.DomainMap_, RemotePIDs_.size(), RemotePIDs, A.ExportLIDs_.size(), ExportLIDs, ExportPIDs);
 
-  tm4 = Teuchos::null;
+  tm4      = Teuchos::null;
   auto tm5 = TimeMonitor::getNewTimer("OptimizedTransfer: Convert: ESFC");
 
   // ESFC
-  Aout->ExpertStaticFillComplete(A.DomainMap_,*RowMap,Importer,0);
+  Aout->ExpertStaticFillComplete(A.DomainMap_, *RowMap, Importer, 0);
 
   // Cleanup
-  if(A.use_lw) delete RowMap;
+  if (A.use_lw) delete RowMap;
   delete ColMap;
 
   return Aout;
@@ -226,149 +222,174 @@ Epetra_CrsMatrix* convert_lightweightcrsmatrix_to_crsmatrix(const EpetraExt::Lig
 // =========================================================================
 // =========================================================================
 #if defined(HAVE_MUELU_EPETRA)
-bool epetra_check_importer_correctness(const Epetra_Import & A, const Epetra_Import & B) {
-  int MyPID = A.SourceMap().Comm().MyPID();
-  bool is_correct=true;
+bool epetra_check_importer_correctness(const Epetra_Import& A, const Epetra_Import& B) {
+  int MyPID       = A.SourceMap().Comm().MyPID();
+  bool is_correct = true;
 
   // Same
-  if(A.NumSameIDs() != B.NumSameIDs())       {printf("[%d] NumSameIDs %d vs. %d\n",MyPID,A.NumSameIDs(),B.NumSameIDs());is_correct=false;}
+  if (A.NumSameIDs() != B.NumSameIDs()) {
+    printf("[%d] NumSameIDs %d vs. %d\n", MyPID, A.NumSameIDs(), B.NumSameIDs());
+    is_correct = false;
+  }
   // Permutes
-  if(A.NumPermuteIDs() != B.NumPermuteIDs()) {printf("[%d] NumPermuteIDs %d vs. %d\n",MyPID,A.NumPermuteIDs(),B.NumPermuteIDs());is_correct=false;}
-  else {
-    int N= A.NumPermuteIDs();
-    bool error_detected=false;
-    for(int i=0; !error_detected && i<N; i++)
+  if (A.NumPermuteIDs() != B.NumPermuteIDs()) {
+    printf("[%d] NumPermuteIDs %d vs. %d\n", MyPID, A.NumPermuteIDs(), B.NumPermuteIDs());
+    is_correct = false;
+  } else {
+    int N               = A.NumPermuteIDs();
+    bool error_detected = false;
+    for (int i = 0; !error_detected && i < N; i++)
       error_detected = (A.PermuteFromLIDs()[i] != B.PermuteFromLIDs()[i]) || (A.PermuteToLIDs()[i] != B.PermuteToLIDs()[i]);
 
-    if(error_detected) {
-      printf("[%d] A Permutes = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",A.PermuteFromLIDs()[i],A.PermuteToLIDs()[i]);
-      printf("\n[%d] B Permutes = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",B.PermuteFromLIDs()[i],B.PermuteToLIDs()[i]);
+    if (error_detected) {
+      printf("[%d] A Permutes = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", A.PermuteFromLIDs()[i], A.PermuteToLIDs()[i]);
+      printf("\n[%d] B Permutes = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", B.PermuteFromLIDs()[i], B.PermuteToLIDs()[i]);
       printf("\n");
-      is_correct=false;
+      is_correct = false;
     }
   }
 
   // Remotes
-  if(A.NumRemoteIDs() != B.NumRemoteIDs()) {printf("[%d] NumRemoteIDs %d vs. %d\n",MyPID,A.NumRemoteIDs(),B.NumRemoteIDs());is_correct=false;}
-  else {
-    int N= A.NumRemoteIDs();
-    bool error_detected=false;
-    for(int i=0; !error_detected && i<N; i++)
+  if (A.NumRemoteIDs() != B.NumRemoteIDs()) {
+    printf("[%d] NumRemoteIDs %d vs. %d\n", MyPID, A.NumRemoteIDs(), B.NumRemoteIDs());
+    is_correct = false;
+  } else {
+    int N               = A.NumRemoteIDs();
+    bool error_detected = false;
+    for (int i = 0; !error_detected && i < N; i++)
       error_detected = A.RemoteLIDs()[i] != B.RemoteLIDs()[i];
 
-    if(error_detected) {
-      printf("[%d] A RemoteLIDs = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d ",A.RemoteLIDs()[i]);
-      printf("\n[%d] B RemoteLIDs = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d ",B.RemoteLIDs()[i]);
+    if (error_detected) {
+      printf("[%d] A RemoteLIDs = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d ", A.RemoteLIDs()[i]);
+      printf("\n[%d] B RemoteLIDs = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d ", B.RemoteLIDs()[i]);
       printf("\n");
-      is_correct=false;
+      is_correct = false;
     }
   }
 
   // Exports
-  if(A.NumExportIDs() != B.NumExportIDs()) {printf("[%d] NumExportIDs %d vs. %d\n",MyPID,A.NumExportIDs(),B.NumExportIDs());is_correct=false;}
-  else {
-    int N= A.NumExportIDs();
-    bool error_detected=false;
-    for(int i=0; !error_detected && i<N; i++)
+  if (A.NumExportIDs() != B.NumExportIDs()) {
+    printf("[%d] NumExportIDs %d vs. %d\n", MyPID, A.NumExportIDs(), B.NumExportIDs());
+    is_correct = false;
+  } else {
+    int N               = A.NumExportIDs();
+    bool error_detected = false;
+    for (int i = 0; !error_detected && i < N; i++)
       error_detected = (A.ExportLIDs()[i] != B.ExportLIDs()[i]) || (A.ExportPIDs()[i] != B.ExportPIDs()[i]);
 
-    if(error_detected) {
-      printf("[%d] A Exports(%d) = ",MyPID,A.NumExportIDs());
-      for(int i=0; i<N; i++)
-        printf("%d(%d)->%d ",A.ExportLIDs()[i],A.SourceMap().GID(A.ExportLIDs()[i]),A.ExportPIDs()[i]);
-      printf("\n[%d] B Exports(%d) = ",MyPID,B.NumExportIDs());
-      for(int i=0; i<N; i++)
-        printf("%d(%d)->%d ",B.ExportLIDs()[i],B.SourceMap().GID(A.ExportLIDs()[i]),B.ExportPIDs()[i]);
+    if (error_detected) {
+      printf("[%d] A Exports(%d) = ", MyPID, A.NumExportIDs());
+      for (int i = 0; i < N; i++)
+        printf("%d(%d)->%d ", A.ExportLIDs()[i], A.SourceMap().GID(A.ExportLIDs()[i]), A.ExportPIDs()[i]);
+      printf("\n[%d] B Exports(%d) = ", MyPID, B.NumExportIDs());
+      for (int i = 0; i < N; i++)
+        printf("%d(%d)->%d ", B.ExportLIDs()[i], B.SourceMap().GID(A.ExportLIDs()[i]), B.ExportPIDs()[i]);
       printf("\n");
-      is_correct=false;
+      is_correct = false;
     }
   }
 
   // Message Counts
-  if(A.NumSend() != B.NumSend())       {printf("[%d] NumSend %d vs. %d\n",MyPID,A.NumSend(),B.NumSend());is_correct=false;}
-  if(A.NumRecv() != B.NumRecv())       {printf("[%d] NumRecv %d vs. %d\n",MyPID,A.NumRecv(),B.NumRecv());is_correct=false;}
+  if (A.NumSend() != B.NumSend()) {
+    printf("[%d] NumSend %d vs. %d\n", MyPID, A.NumSend(), B.NumSend());
+    is_correct = false;
+  }
+  if (A.NumRecv() != B.NumRecv()) {
+    printf("[%d] NumRecv %d vs. %d\n", MyPID, A.NumRecv(), B.NumRecv());
+    is_correct = false;
+  }
 
 #ifdef HAVE_MPI
   const Epetra_MpiDistributor& Ad = *dynamic_cast<Epetra_MpiDistributor*>(&A.Distributor());
   const Epetra_MpiDistributor& Bd = *dynamic_cast<Epetra_MpiDistributor*>(&B.Distributor());
 
-  if(Ad.MaxSendLength() != Bd.MaxSendLength()) {printf("[%d] Distor.MaxSendLength %d vs. %d\n",MyPID,Ad.MaxSendLength(),Bd.MaxSendLength());is_correct=false;}
-  if(Ad.TotalReceiveLength() != Bd.TotalReceiveLength()) {printf("[%d] Distor.TotalReceiveLength %d vs. %d\n",MyPID,Ad.TotalReceiveLength(),Bd.TotalReceiveLength());is_correct=false;}
+  if (Ad.MaxSendLength() != Bd.MaxSendLength()) {
+    printf("[%d] Distor.MaxSendLength %d vs. %d\n", MyPID, Ad.MaxSendLength(), Bd.MaxSendLength());
+    is_correct = false;
+  }
+  if (Ad.TotalReceiveLength() != Bd.TotalReceiveLength()) {
+    printf("[%d] Distor.TotalReceiveLength %d vs. %d\n", MyPID, Ad.TotalReceiveLength(), Bd.TotalReceiveLength());
+    is_correct = false;
+  }
 
-  if(Ad.NumSends() != Bd.NumSends()) {printf("[%d] Distor.NumSends %d vs. %d\n",MyPID,Ad.NumSends(),Bd.NumSends());is_correct=false;}
-  else {
-    int N=Ad.NumSends();
-    bool error_detected=false;
-    for(int i=0; !error_detected && i<N; i++)
+  if (Ad.NumSends() != Bd.NumSends()) {
+    printf("[%d] Distor.NumSends %d vs. %d\n", MyPID, Ad.NumSends(), Bd.NumSends());
+    is_correct = false;
+  } else {
+    int N               = Ad.NumSends();
+    bool error_detected = false;
+    for (int i = 0; !error_detected && i < N; i++)
       error_detected = (Ad.ProcsTo()[i] != Bd.ProcsTo()[i]) || (Ad.LengthsTo()[i] != Bd.LengthsTo()[i]);
 
-    if(error_detected) {
-      printf("[%d] Ad Sends = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",Ad.LengthsTo()[i],Ad.ProcsTo()[i]);
-      printf("\n[%d] Bd Sends = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",Bd.LengthsTo()[i],Bd.ProcsTo()[i]);
+    if (error_detected) {
+      printf("[%d] Ad Sends = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", Ad.LengthsTo()[i], Ad.ProcsTo()[i]);
+      printf("\n[%d] Bd Sends = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", Bd.LengthsTo()[i], Bd.ProcsTo()[i]);
       printf("\n");
-      is_correct=false;
+      is_correct = false;
     }
   }
 
-  if(Ad.NumReceives() != Bd.NumReceives()) {printf("[%d] Distor.NumReceives %d vs. %d\n",MyPID,Ad.NumReceives(),Bd.NumReceives());is_correct=false;}
-  else {
-    int N=Ad.NumReceives();
-    bool error_detected=false;
-    for(int i=0; !error_detected && i<N; i++)
+  if (Ad.NumReceives() != Bd.NumReceives()) {
+    printf("[%d] Distor.NumReceives %d vs. %d\n", MyPID, Ad.NumReceives(), Bd.NumReceives());
+    is_correct = false;
+  } else {
+    int N               = Ad.NumReceives();
+    bool error_detected = false;
+    for (int i = 0; !error_detected && i < N; i++)
       error_detected = (Ad.ProcsFrom()[i] != Bd.ProcsFrom()[i]) || (Ad.LengthsFrom()[i] != Bd.LengthsFrom()[i]);
 
-    if(error_detected) {
-      printf("[%d] Ad Receives = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",Ad.LengthsFrom()[i],Ad.ProcsFrom()[i]);
-      printf("\n[%d] Bd Receives = ",MyPID);
-      for(int i=0; i<N; i++)
-        printf("%d->%d ",Bd.LengthsFrom()[i],Bd.ProcsFrom()[i]);
+    if (error_detected) {
+      printf("[%d] Ad Receives = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", Ad.LengthsFrom()[i], Ad.ProcsFrom()[i]);
+      printf("\n[%d] Bd Receives = ", MyPID);
+      for (int i = 0; i < N; i++)
+        printf("%d->%d ", Bd.LengthsFrom()[i], Bd.ProcsFrom()[i]);
       printf("\n");
-      is_correct=false;
+      is_correct = false;
     }
   }
 #endif
 
   return is_correct;
 }
-#endif //if defined(HAVE_MUELU_EPETRA)
+#endif  // if defined(HAVE_MUELU_EPETRA)
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void TestTransfer(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > A, Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > P) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TestTransfer(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > A, Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > P) {
 #include <MueLu_UseShortNames.hpp>
   Xpetra::UnderlyingLib lib = A->getRowMap()->lib();
 
   if (lib == Xpetra::UseTpetra) {
-    typedef Tpetra::CrsMatrixStruct<SC,LO,GO,NO>  crs_matrix_struct_type;
-    typedef Tpetra::CrsMatrix<SC,LO,GO,NO>        crs_matrix_type;
-    typedef Tpetra::Import<LO,GO,NO>              import_type;
+    typedef Tpetra::CrsMatrixStruct<SC, LO, GO, NO> crs_matrix_struct_type;
+    typedef Tpetra::CrsMatrix<SC, LO, GO, NO> crs_matrix_type;
+    typedef Tpetra::Import<LO, GO, NO> import_type;
 
     RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(A);
     RCP<const crs_matrix_type> Pu = Utilities::Op2TpetraCrs(P);
-    if(Au->getComm()->getSize()==1) return;
+    if (Au->getComm()->getSize() == 1) return;
 
     // ==================
     // Optimized Transfer
     // ==================
     auto tm = TimeMonitor::getNewTimer("OptimizedTransfer: Import");
     crs_matrix_struct_type Pview;
-    Tpetra::MMdetails::import_and_extract_views(*Pu,Au->getColMap(), Pview, Au->getGraph()->getImporter(), false,"ImportPerf: ");
+    Tpetra::MMdetails::import_and_extract_views(*Pu, Au->getColMap(), Pview, Au->getGraph()->getImporter(), false, "ImportPerf: ");
 
     Au->getComm()->barrier();
 
@@ -377,16 +398,15 @@ void TestTransfer(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,
     // ==================
     // Use the columnmap from Aopt and build an importer ex nihilo
 
-    tm = Teuchos::null;
+    tm       = Teuchos::null;
     auto tm2 = TimeMonitor::getNewTimer("NaiveTransfer: BuildImport");
-    import_type NaiveImport(Pview.importMatrix->getColMap(),Pu->getDomainMap());
+    import_type NaiveImport(Pview.importMatrix->getColMap(), Pu->getDomainMap());
     Au->getComm()->barrier();
-  }
-  else if (lib == Xpetra::UseEpetra) {
+  } else if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA)
     RCP<const Epetra_CrsMatrix> Au = Utilities::Op2EpetraCrs(A);
     RCP<const Epetra_CrsMatrix> Pu = Utilities::Op2EpetraCrs(P);
-    if(Au->Comm().NumProc()==1) return;
+    if (Au->Comm().NumProc() == 1) return;
 
     // ==================
     // Optimized Transfer
@@ -395,136 +415,141 @@ void TestTransfer(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,
 
     auto tm3 = TimeMonitor::getNewTimer("OptimizedTransfer: Import");
     EpetraExt::CrsMatrixStruct Pview;
-    bool SortGhosts=true;
+    bool SortGhosts = true;
 
-    if(Au->RowMap().GlobalIndicesInt()) {
+    if (Au->RowMap().GlobalIndicesInt()) {
 #ifndef EPETRA_NO_32BIT_GLOBAL_INDICES
-      EpetraExt::import_only<int> (*Pu,Au->ColMap(),Pview,Au->Importer(),SortGhosts,"ImportPerf: ");
+      EpetraExt::import_only<int>(*Pu, Au->ColMap(), Pview, Au->Importer(), SortGhosts, "ImportPerf: ");
 #endif
-    }
-    else if(Au->RowMap().GlobalIndicesInt()) {
+    } else if (Au->RowMap().GlobalIndicesInt()) {
 #ifndef EPETRA_NO_64BIT_GLOBAL_INDICES
-      EpetraExt::import_only<long long>(*Pu,Au->ColMap(),Pview,Au->Importer(),SortGhosts,"ImportPerf: ");
+      EpetraExt::import_only<long long>(*Pu, Au->ColMap(), Pview, Au->Importer(), SortGhosts, "ImportPerf: ");
 #endif
     }
-    tm3 = Teuchos::null;
-    auto tm4 = TimeMonitor::getNewTimer("OptimizedTransfer: Convert");
-    Epetra_CrsMatrix *Aopt = convert_lightweightcrsmatrix_to_crsmatrix(*Pview.importMatrix);
+    tm3                    = Teuchos::null;
+    auto tm4               = TimeMonitor::getNewTimer("OptimizedTransfer: Convert");
+    Epetra_CrsMatrix* Aopt = convert_lightweightcrsmatrix_to_crsmatrix(*Pview.importMatrix);
 
     Au->Comm().Barrier();
     // ==================
     // Naive Transfer
     // ==================
     // Use the columnmap from Aopt and build an importer ex nihilo
-    tm4 = Teuchos::null;
-    auto tm5 = TimeMonitor::getNewTimer("NaiveTransfer: BuildImport");
-    const Epetra_Map &NaiveColMap = Aopt->ColMap();
-    Epetra_Import NaiveImport(NaiveColMap,Pu->DomainMap());
+    tm4                           = Teuchos::null;
+    auto tm5                      = TimeMonitor::getNewTimer("NaiveTransfer: BuildImport");
+    const Epetra_Map& NaiveColMap = Aopt->ColMap();
+    Epetra_Import NaiveImport(NaiveColMap, Pu->DomainMap());
 
     Au->Comm().Barrier();
 
     // Check importer for correctness
     fflush(stdout);
-    const Epetra_Import * OptImport = Aopt->Importer();
-    bool is_correct = epetra_check_importer_correctness(NaiveImport,*OptImport);
+    const Epetra_Import* OptImport = Aopt->Importer();
+    bool is_correct                = epetra_check_importer_correctness(NaiveImport, *OptImport);
     fflush(stdout);
     int is_OK_local = is_correct, is_OK_global;
-    Au->Comm().MinAll(&is_OK_local,&is_OK_global,1);
-    if(!is_OK_global) throw std::runtime_error("Importer correctness test failed.");
+    Au->Comm().MinAll(&is_OK_local, &is_OK_global, 1);
+    if (!is_OK_global) throw std::runtime_error("Importer correctness test failed.");
 
     // Cleanup
     delete Aopt;
-#endif // defined(HAVE_MUELU_EPETRA)
+#endif  // defined(HAVE_MUELU_EPETRA)
   }
 }
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    using Teuchos::ArrayRCP;
-    using Teuchos::ParameterList;
-
-    // =========================================================================
-    // MPI initialization using Teuchos
-    // =========================================================================
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // =========================================================================
-    // Convenient definitions
-    // =========================================================================
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC one = STS::one();
-    typedef typename STS::coordinateType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    // =========================================================================
-    // Parameters initialization
-    // =========================================================================
-    GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName       = "import.xml";     clp.setOption("xml",                   &xmlFileName,       "read parameters from a file");
-    bool        printTimings      = true;             clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-    std::string timingsFormat     = "table-fixed";    clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-    int         numImports        = 100;              clp.setOption("numImport",             &numImports,        "#times to test");
-    int         MM_TAFC_OptCoreCnt=3000;              clp.setOption("MM_TAFC_OptimizationCoreCount",       &MM_TAFC_OptCoreCnt, "Num Cores above which Optimized MatrixMatrix transferAndFillComplete is used");
-
-    switch (clp.parse(argc, argv)) {
-        case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
-        case Teuchos::CommandLineProcessor::PARSE_ERROR:
-        case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-        case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
-    }
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+
+  // =========================================================================
+  // MPI initialization using Teuchos
+  // =========================================================================
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // =========================================================================
+  // Convenient definitions
+  // =========================================================================
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC one = STS::one();
+  typedef typename STS::coordinateType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  // =========================================================================
+  // Parameters initialization
+  // =========================================================================
+  GO nx = 100, ny = 100, nz = 100;
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "import.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  std::string timingsFormat = "table-fixed";
+  clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+  int numImports = 100;
+  clp.setOption("numImport", &numImports, "#times to test");
+  int MM_TAFC_OptCoreCnt = 3000;
+  clp.setOption("MM_TAFC_OptimizationCoreCount", &MM_TAFC_OptCoreCnt, "Num Cores above which Optimized MatrixMatrix transferAndFillComplete is used");
+
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_ERROR:
+    case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
+  }
 
-    ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), *comm);
+  ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), *comm);
 
-    ParameterList& mmlist = paramList.sublist("matrixmatrix: kernel params",false);
-    int commandcc = mmlist.get("MM_TAFC_OptimizationCoreCount",MM_TAFC_OptCoreCnt);
-    commandcc = paramList.get("MM_TAFC_OptimizationCoreCount",commandcc);
-    paramList.remove("MM_TAFC_OptimizationCoreCount",false);
-    mmlist.set("MM_TAFC_OptimizationCoreCount",commandcc);
+  ParameterList& mmlist = paramList.sublist("matrixmatrix: kernel params", false);
+  int commandcc         = mmlist.get("MM_TAFC_OptimizationCoreCount", MM_TAFC_OptCoreCnt);
+  commandcc             = paramList.get("MM_TAFC_OptimizationCoreCount", commandcc);
+  paramList.remove("MM_TAFC_OptimizationCoreCount", false);
+  mmlist.set("MM_TAFC_OptimizationCoreCount", commandcc);
 
-    bool isDriver = paramList.isSublist("Run1");
-    if (isDriver) {
-        // update galeriParameters with the values from the XML file
-        ParameterList& realParams = galeriParameters.GetParameterList();
+  bool isDriver = paramList.isSublist("Run1");
+  if (isDriver) {
+    // update galeriParameters with the values from the XML file
+    ParameterList& realParams = galeriParameters.GetParameterList();
 
-        for (ParameterList::ConstIterator it = realParams.begin(); it != realParams.end(); it++) {
-            const std::string& name = realParams.name(it);
-            if (paramList.isParameter(name))
-                realParams.setEntry(name, paramList.getEntry(name));
-        }
+    for (ParameterList::ConstIterator it = realParams.begin(); it != realParams.end(); it++) {
+      const std::string& name = realParams.name(it);
+      if (paramList.isParameter(name))
+        realParams.setEntry(name, paramList.getEntry(name));
     }
+  }
 
-    // Retrieve matrix parameters (they may have been changed on the command line)
-    // [for instance, if we changed matrix type from 2D to 3D we need to update nz]
-    ParameterList galeriList = galeriParameters.GetParameterList();
+  // Retrieve matrix parameters (they may have been changed on the command line)
+  // [for instance, if we changed matrix type from 2D to 3D we need to update nz]
+  ParameterList galeriList = galeriParameters.GetParameterList();
 
-    // =========================================================================
-    // Problem construction
-    // =========================================================================
-    std::ostringstream galeriStream;
-    comm->barrier();
+  // =========================================================================
+  // Problem construction
+  // =========================================================================
+  std::ostringstream galeriStream;
+  comm->barrier();
+
+  RCP<Matrix> A;
+  RCP<const Map> map;
+  RCP<RealValuedMultiVector> coordinates;
+  typedef typename RealValuedMultiVector::scalar_type Real;
+  RCP<MultiVector> nullspace;
+  std::string matrixType = galeriParameters.GetMatrixType();
 
-    RCP<Matrix>      A;
-    RCP<const Map>   map;
-    RCP<RealValuedMultiVector> coordinates;
-    typedef typename RealValuedMultiVector::scalar_type Real;
-    RCP<MultiVector> nullspace;
-    std::string matrixType = galeriParameters.GetMatrixType();
-  
-    RCP<TimeMonitor>  globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
-    {
-        auto tm= rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
+  RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
+  {
+    auto tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-    galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+    galeriStream << "========================================================\n"
+                 << xpetraParameters << galeriParameters;
 
     // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
     //                                 d1  d2  d3
@@ -536,29 +561,28 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib &lib, int a
     // size. For example, np=14 will give a 7-by-2 distribution.
     // If you don't want Galeri to do this, specify mx or my on the galeriList.
 
-
     // Create map and coordinates
     // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
     // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
     if (matrixType == "Laplace1D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
     } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
                matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
     } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
     }
 
     // Expand map to do multiple DOF per node for block problems
     if (matrixType == "Elasticity2D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
     if (matrixType == "Elasticity3D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
     galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                  << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -566,203 +590,202 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib &lib, int a
                  << "========================================================" << std::endl;
 
     if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
-        // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-        galeriList.set("right boundary" , "Neumann");
-        galeriList.set("bottom boundary", "Neumann");
-        galeriList.set("top boundary"   , "Neumann");
-        galeriList.set("front boundary" , "Neumann");
-        galeriList.set("back boundary"  , "Neumann");
+      // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
+      galeriList.set("right boundary", "Neumann");
+      galeriList.set("bottom boundary", "Neumann");
+      galeriList.set("top boundary", "Neumann");
+      galeriList.set("front boundary", "Neumann");
+      galeriList.set("back boundary", "Neumann");
     }
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
     A = Pr->BuildMatrix();
 
     if (matrixType == "Elasticity2D" ||
         matrixType == "Elasticity3D") {
-        nullspace = Pr->BuildNullspace();
-        A->SetFixedBlockSize((galeriParameters.GetMatrixType() == "Elasticity2D") ? 2 : 3);
+      nullspace = Pr->BuildNullspace();
+      A->SetFixedBlockSize((galeriParameters.GetMatrixType() == "Elasticity2D") ? 2 : 3);
     }
 
     comm->barrier();
     tm = Teuchos::null;
-    }
-
+  }
 
-    galeriStream << "Galeri complete.\n========================================================" << std::endl;
+  galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
-    int numReruns = 1;
-    if (paramList.isParameter("number of reruns"))
-        numReruns = paramList.get<int>("number of reruns");
+  int numReruns = 1;
+  if (paramList.isParameter("number of reruns"))
+    numReruns = paramList.get<int>("number of reruns");
 
-    const bool mustAlreadyExist = true;
-    for (int rerunCount = 1; rerunCount <= numReruns; rerunCount++) {
+  const bool mustAlreadyExist = true;
+  for (int rerunCount = 1; rerunCount <= numReruns; rerunCount++) {
+    ParameterList mueluList, runList;
 
-        ParameterList mueluList, runList;
+    bool stop = false;
+    if (isDriver) {
+      runList   = paramList.sublist("Run1", mustAlreadyExist);
+      mueluList = runList.sublist("MueLu", mustAlreadyExist);
+    } else {
+      mueluList = paramList;
+      stop      = true;
+    }
 
-        bool stop = false;
-        if (isDriver) {
-            runList   = paramList.sublist("Run1",  mustAlreadyExist);
-            mueluList = runList  .sublist("MueLu", mustAlreadyExist);
-        } else {
-            mueluList = paramList;
-            stop = true;
+    if (nullspace.is_null()) {
+      int blkSize = 1;
+      if (mueluList.isSublist("Matrix")) {
+        // Factory style parameter list
+        const Teuchos::ParameterList& operatorList = paramList.sublist("Matrix");
+        if (operatorList.isParameter("PDE equations"))
+          blkSize = operatorList.get<int>("PDE equations");
+
+      } else if (paramList.isParameter("number of equations")) {
+        // Easy style parameter list
+        blkSize = paramList.get<int>("number of equations");
+      }
+
+      nullspace = MultiVectorFactory::Build(map, blkSize);
+      for (int i = 0; i < blkSize; i++) {
+        RCP<const Map> domainMap = A->getDomainMap();
+        GO indexBase             = domainMap->getIndexBase();
+
+        ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
+        for (int j = 0; j < nsData.size(); j++) {
+          GO GID = domainMap->getGlobalElement(j) - indexBase;
+
+          if ((GID - i) % blkSize == 0)
+            nsData[j] = one;
         }
+      }
+    }
 
-        if (nullspace.is_null()) {
-            int blkSize = 1;
-            if (mueluList.isSublist("Matrix")) {
-                // Factory style parameter list
-                const Teuchos::ParameterList& operatorList = paramList.sublist("Matrix");
-                if (operatorList.isParameter("PDE equations"))
-                    blkSize = operatorList.get<int>("PDE equations");
-
-            } else if (paramList.isParameter("number of equations")) {
-                // Easy style parameter list
-                blkSize = paramList.get<int>("number of equations");
-            }
-
-            nullspace = MultiVectorFactory::Build(map, blkSize);
-            for (int i = 0; i < blkSize; i++) {
-                RCP<const Map> domainMap = A->getDomainMap();
-                GO             indexBase = domainMap->getIndexBase();
-
-                ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
-                for (int j = 0; j < nsData.size(); j++) {
-                    GO GID = domainMap->getGlobalElement(j) - indexBase;
-
-                    if ((GID-i) % blkSize == 0)
-                        nsData[j] = one;
-                }
-            }
+    int runCount = 1;
+    do {
+      int savedOut    = -1;
+      FILE* openedOut = NULL;
+      if (isDriver) {
+        if (runList.isParameter("filename")) {
+          // Redirect all output into a filename We have to redirect all output,
+          // including printf's, therefore we cannot simply replace C++ cout
+          // buffers, and have to use heavy machinary (dup2)
+          std::string filename = runList.get<std::string>("filename");
+          if (numReruns > 1)
+            filename += "_run" + MueLu::toString(rerunCount);
+          filename += (lib == Xpetra::UseEpetra ? ".epetra" : ".tpetra");
+
+          savedOut  = dup(STDOUT_FILENO);
+          openedOut = fopen(filename.c_str(), "w");
+          dup2(fileno(openedOut), STDOUT_FILENO);
         }
+      }
 
-        int runCount = 1;
-        do {
-            int   savedOut  = -1;
-            FILE* openedOut = NULL;
-            if (isDriver) {
-                if (runList.isParameter("filename")) {
-                    // Redirect all output into a filename We have to redirect all output,
-                    // including printf's, therefore we cannot simply replace C++ cout
-                    // buffers, and have to use heavy machinary (dup2)
-                    std::string filename = runList.get<std::string>("filename");
-                    if (numReruns > 1)
-                        filename += "_run" + MueLu::toString(rerunCount);
-                    filename += (lib == Xpetra::UseEpetra ? ".epetra" : ".tpetra");
-
-                    savedOut  = dup(STDOUT_FILENO);
-                    openedOut = fopen(filename.c_str(), "w");
-                    dup2(fileno(openedOut), STDOUT_FILENO);
-                }
-            }
-
-            // Instead of checking each time for rank, create a rank 0 stream
-            RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-            Teuchos::FancyOStream& out = *fancy;
-            out.setOutputToRootOnly(0);
-
-            out << galeriStream.str();
-
-            // =========================================================================
-            // Preconditioner construction
-            // =========================================================================
-            comm->barrier();
-
-
-            RCP<Hierarchy> H;
-            {
-                auto MueLuSU_D2 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup"));
-
-                A->SetMaxEigenvalueEstimate(-one);
-                if(lib == Xpetra::UseEpetra) {mueluList.set("use kokkos refactor", false);}
-                Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
-                userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-                H = MueLu::CreateXpetraPreconditioner(A, mueluList);
-                comm->barrier();
-            }
-
-            // =========================================================================
-            // Grab useful pieces
-            // =========================================================================
-
-            RCP<Matrix> P;
-            {
-                auto MueLuES_D3 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - Extract Stuff"));
-                H->GetLevel(1)->Get("P",P);
-                comm->barrier();
-            }
-
-            RCP<Teuchos::ParameterList> opt_list = rcp(new Teuchos::ParameterList);
-            opt_list->set("Timer Label","OptTAFC");
-
-            for(int i=0; i<numImports; i++) {
-                // =========================================================================
-                // Optimized transfer & fill complete loop for P_1
-                // =========================================================================
-                auto D4 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - TransferAndFillComplete"));
-                TestTransfer(A,P);
-                comm->barrier();
-            }
-
-            globalTimeMonitor  = Teuchos::null;
-
-            // =========================================================================
-            // Timing stuff
-            // =========================================================================
-            if (printTimings) {
-                RCP<ParameterList> reportParams = rcp(new ParameterList);
-                if (timingsFormat == "yaml") {
-                    reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-                    reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
-                }
-                reportParams->set("How to merge timer sets",   "Union");
-                reportParams->set("alwaysWriteLocal",          false);
-                reportParams->set("writeGlobalStats",          true);
-                reportParams->set("writeZeroTimers",           false);
-                // FIXME: no "ignoreZeroTimers"
-
-                const std::string filter = "";
-
-                std::ios_base::fmtflags ff(out.flags());
-                if (timingsFormat == "table-fixed") out << std::fixed;
-                else                                out << std::scientific;
-                TimeMonitor::report(comm.ptr(), out, filter, reportParams);
-                out << std::setiosflags(ff);
-            }
-
-            TimeMonitor::clearCounters();
-
-            if (isDriver) {
-                if (openedOut != NULL) {
-                    TEUCHOS_ASSERT(savedOut >= 0);
-                    dup2(savedOut, STDOUT_FILENO);
-                    fclose(openedOut);
-                    openedOut = NULL;
-                }
-                try {
-                    runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
-                    mueluList = runList  .sublist("MueLu", mustAlreadyExist);
-                } catch (Teuchos::Exceptions::InvalidParameterName&) {
-                    stop = true;
-                }
-            }
-
-        } while (!stop);
-    }
+      // Instead of checking each time for rank, create a rank 0 stream
+      RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+      Teuchos::FancyOStream& out       = *fancy;
+      out.setOutputToRootOnly(0);
 
-    comm->barrier();
+      out << galeriStream.str();
 
-    return EXIT_SUCCESS;
-}
+      // =========================================================================
+      // Preconditioner construction
+      // =========================================================================
+      comm->barrier();
 
+      RCP<Hierarchy> H;
+      {
+        auto MueLuSU_D2 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup"));
 
+        A->SetMaxEigenvalueEstimate(-one);
+        if (lib == Xpetra::UseEpetra) {
+          mueluList.set("use kokkos refactor", false);
+        }
+        Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
+        userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+        H = MueLu::CreateXpetraPreconditioner(A, mueluList);
+        comm->barrier();
+      }
+
+      // =========================================================================
+      // Grab useful pieces
+      // =========================================================================
+
+      RCP<Matrix> P;
+      {
+        auto MueLuES_D3 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - Extract Stuff"));
+        H->GetLevel(1)->Get("P", P);
+        comm->barrier();
+      }
+
+      RCP<Teuchos::ParameterList> opt_list = rcp(new Teuchos::ParameterList);
+      opt_list->set("Timer Label", "OptTAFC");
+
+      for (int i = 0; i < numImports; i++) {
+        // =========================================================================
+        // Optimized transfer & fill complete loop for P_1
+        // =========================================================================
+        auto D4 = TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - TransferAndFillComplete"));
+        TestTransfer(A, P);
+        comm->barrier();
+      }
+
+      globalTimeMonitor = Teuchos::null;
+
+      // =========================================================================
+      // Timing stuff
+      // =========================================================================
+      if (printTimings) {
+        RCP<ParameterList> reportParams = rcp(new ParameterList);
+        if (timingsFormat == "yaml") {
+          reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+          reportParams->set("YAML style", "compact");  // "spacious" or "compact"
+        }
+        reportParams->set("How to merge timer sets", "Union");
+        reportParams->set("alwaysWriteLocal", false);
+        reportParams->set("writeGlobalStats", true);
+        reportParams->set("writeZeroTimers", false);
+        // FIXME: no "ignoreZeroTimers"
+
+        const std::string filter = "";
+
+        std::ios_base::fmtflags ff(out.flags());
+        if (timingsFormat == "table-fixed")
+          out << std::fixed;
+        else
+          out << std::scientific;
+        TimeMonitor::report(comm.ptr(), out, filter, reportParams);
+        out << std::setiosflags(ff);
+      }
+
+      TimeMonitor::clearCounters();
+
+      if (isDriver) {
+        if (openedOut != NULL) {
+          TEUCHOS_ASSERT(savedOut >= 0);
+          dup2(savedOut, STDOUT_FILENO);
+          fclose(openedOut);
+          openedOut = NULL;
+        }
+        try {
+          runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
+          mueluList = runList.sublist("MueLu", mustAlreadyExist);
+        } catch (Teuchos::Exceptions::InvalidParameterName&) {
+          stop = true;
+        }
+      }
+
+    } while (!stop);
+  }
+
+  comm->barrier();
+
+  return EXIT_SUCCESS;
+}
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-    auto val =  Automatic_Test_ETI(argc,argv);
-    return val;
+int main(int argc, char* argv[]) {
+  auto val = Automatic_Test_ETI(argc, argv);
+  return val;
 }
diff --git a/packages/muelu/test/scaling/JacobiKernelDriver.cpp b/packages/muelu/test/scaling/JacobiKernelDriver.cpp
index 2e6a59558646..0167fe5b2c7f 100644
--- a/packages/muelu/test/scaling/JacobiKernelDriver.cpp
+++ b/packages/muelu/test/scaling/JacobiKernelDriver.cpp
@@ -71,29 +71,28 @@
 // =========================================================================
 // Support Routines
 // =========================================================================
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view_n(const int n, const View1 x1, View2 x2) {
-  Kokkos::parallel_for(n,KOKKOS_LAMBDA(const size_t i) {
-      x2[i] = x1[i];
-    });
+  Kokkos::parallel_for(
+      n, KOKKOS_LAMBDA(const size_t i) {
+        x2[i] = x1[i];
+      });
 }
 
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view(const View1 x1, View2 x2) {
-  copy_view_n(x1.extent(0),x1,x2);
+  copy_view_n(x1.extent(0), x1, x2);
 }
 
-
-
 // =========================================================================
 // MKL Testing
 // =========================================================================
 #ifdef HAVE_MUELU_MKL
 #include "mkl.h"
 
-  // mkl_sparse_spmm
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C, Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &D, Scalar omega) {
+// mkl_sparse_spmm
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C, Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &D, Scalar omega) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -104,32 +103,32 @@ void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node
 
   Xpetra::UnderlyingLib lib = A.getRowMap()->lib();
   RCP<TimeMonitor> tm;
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>    vector_type;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
   typedef typename crs_matrix_type::local_matrix_device_type KCRS;
-  typedef typename KCRS::StaticCrsGraphType                  graph_t;
-  typedef typename graph_t::row_map_type::non_const_type     lno_view_t;
-  typedef typename graph_t::row_map_type::const_type         c_lno_view_t;
-  typedef typename graph_t::entries_type::non_const_type     lno_nnz_view_t;
-  typedef typename graph_t::entries_type::const_type         c_lno_nnz_view_t;
-  typedef typename KCRS::values_type::non_const_type         scalar_view_t;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
+  typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+  typedef typename graph_t::row_map_type::const_type c_lno_view_t;
+  typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+  typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
 
-  typedef typename vector_type::device_type              device_type;
-  typedef typename Kokkos::View<MKL_INT*,typename lno_nnz_view_t::array_layout,typename lno_nnz_view_t::device_type> mkl_int_type;
+  typedef typename vector_type::device_type device_type;
+  typedef typename Kokkos::View<MKL_INT *, typename lno_nnz_view_t::array_layout, typename lno_nnz_view_t::device_type> mkl_int_type;
 
-  RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-  RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-  RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-  RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
-  RCP<const vector_type> Du = Xpetra::toTpetra(D);
+  RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+  RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+  RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+  RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+  RCP<const vector_type> Du     = Xpetra::toTpetra(D);
 
-  const KCRS & Amat = Au->getLocalMatrixDevice();
-  const KCRS & Bmat = Bu->getLocalMatrixDevice();
+  const KCRS &Amat = Au->getLocalMatrixDevice();
+  const KCRS &Bmat = Bu->getLocalMatrixDevice();
 
-  if(A.getLocalNumRows()!=C.getLocalNumRows())  throw std::runtime_error("C is not sized correctly");
+  if (A.getLocalNumRows() != C.getLocalNumRows()) throw std::runtime_error("C is not sized correctly");
 
   c_lno_view_t Arowptr = Amat.graph.row_map, Browptr = Bmat.graph.row_map;
-  lno_view_t Crowptr("Crowptr",C.getLocalNumRows()+1);
+  lno_view_t Crowptr("Crowptr", C.getLocalNumRows() + 1);
   c_lno_nnz_view_t Acolind = Amat.graph.entries, Bcolind = Bmat.graph.entries;
   lno_nnz_view_t Ccolind;
   const scalar_view_t Avals = Amat.values, Bvals = Bmat.values;
@@ -142,89 +141,90 @@ void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node
   sparse_matrix_t CMKL;
   sparse_matrix_t DMKL;
 
-  sparse_matrix_t XTempMKL,YTempMKL;
+  sparse_matrix_t XTempMKL, YTempMKL;
 
   sparse_status_t result;
   // **********************************
   // Copy in the matrix data for MKL
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Matrix CopyIn")));
 
-  mkl_int_type ArowptrMKL("Arowptr",Arowptr.extent(0));
-  mkl_int_type BrowptrMKL("Browptr",Browptr.extent(0));
+  mkl_int_type ArowptrMKL("Arowptr", Arowptr.extent(0));
+  mkl_int_type BrowptrMKL("Browptr", Browptr.extent(0));
 
-  mkl_int_type AcolindMKL("Acolind",Acolind.extent(0));
-  mkl_int_type BcolindMKL("Bcolind",Bcolind.extent(0));
+  mkl_int_type AcolindMKL("Acolind", Acolind.extent(0));
+  mkl_int_type BcolindMKL("Bcolind", Bcolind.extent(0));
 
-  copy_view(Arowptr,ArowptrMKL);
-  copy_view(Browptr,BrowptrMKL);
-  copy_view(Acolind,AcolindMKL);
-  copy_view(Bcolind,BcolindMKL);
+  copy_view(Arowptr, ArowptrMKL);
+  copy_view(Browptr, BrowptrMKL);
+  copy_view(Acolind, AcolindMKL);
+  copy_view(Bcolind, BcolindMKL);
 
-  if(std::is_same<Scalar,double>::value) {
-    mkl_sparse_d_create_csr(&AMKL, SPARSE_INDEX_BASE_ZERO, Au->getLocalNumRows(), Au->getLocalNumCols(), ArowptrMKL.data(),ArowptrMKL.data()+1,AcolindMKL.data(),(double*)Avals.data());
-    mkl_sparse_d_create_csr(&BMKL, SPARSE_INDEX_BASE_ZERO, Bu->getLocalNumRows(), Bu->getLocalNumCols(), BrowptrMKL.data(),BrowptrMKL.data()+1,BcolindMKL.data(),(double*)Bvals.data());
-  }
-  else
+  if (std::is_same<Scalar, double>::value) {
+    mkl_sparse_d_create_csr(&AMKL, SPARSE_INDEX_BASE_ZERO, Au->getLocalNumRows(), Au->getLocalNumCols(), ArowptrMKL.data(), ArowptrMKL.data() + 1, AcolindMKL.data(), (double *)Avals.data());
+    mkl_sparse_d_create_csr(&BMKL, SPARSE_INDEX_BASE_ZERO, Bu->getLocalNumRows(), Bu->getLocalNumCols(), BrowptrMKL.data(), BrowptrMKL.data() + 1, BcolindMKL.data(), (double *)Bvals.data());
+  } else
     throw std::runtime_error("MKL Type Mismatch");
   tm = Teuchos::null;
   Au->getComm()->barrier();
 
   // **********************************
   // Copy in the vector-as-matrix data for MKL
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Scaled D-Vector CopyIn")));
+  tm       = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Scaled D-Vector CopyIn")));
   size_t n = Dvals.extent(0);
-  mkl_int_type DrowptrMKL("Drowptr",n+1);
-  mkl_int_type DcolindMKL("Dcolind",n);
-  scalar_view_t DvalsMKL("Dvals",n);
-  Kokkos::parallel_for(n+1,KOKKOS_LAMBDA(const size_t i) {
-      DrowptrMKL[i] = i;
-      if(i < n)  {
-        DcolindMKL[i] = i;
-        DvalsMKL[i]   = - omega * Dvals(i,0);
-      }
-  });
+  mkl_int_type DrowptrMKL("Drowptr", n + 1);
+  mkl_int_type DcolindMKL("Dcolind", n);
+  scalar_view_t DvalsMKL("Dvals", n);
+  Kokkos::parallel_for(
+      n + 1, KOKKOS_LAMBDA(const size_t i) {
+        DrowptrMKL[i] = i;
+        if (i < n) {
+          DcolindMKL[i] = i;
+          DvalsMKL[i]   = -omega * Dvals(i, 0);
+        }
+      });
 
   // NOTE: No sanity checks here.  We did that above
-  mkl_sparse_d_create_csr(&DMKL, SPARSE_INDEX_BASE_ZERO,Dvals.extent(0),Dvals.extent(0), DrowptrMKL.data(),DrowptrMKL.data()+1,DcolindMKL.data(),DvalsMKL.data());
+  mkl_sparse_d_create_csr(&DMKL, SPARSE_INDEX_BASE_ZERO, Dvals.extent(0), Dvals.extent(0), DrowptrMKL.data(), DrowptrMKL.data() + 1, DcolindMKL.data(), DvalsMKL.data());
 
   tm = Teuchos::null;
   Au->getComm()->barrier();
 
   // **********************************
   // Multiply (A*B)
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Multiply")));
+  tm     = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Multiply")));
   result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, BMKL, &XTempMKL);
   typename KCRS::execution_space().fence();
-  if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply failed");
+  if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply failed");
 
   // **********************************
   // Scale (-omegaD) * AB)
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Scale-Via-Multiply")));
+  tm     = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Scale-Via-Multiply")));
   result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, DMKL, XTempMKL, &YTempMKL);
   typename KCRS::execution_space().fence();
-  if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Scale failed");
+  if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Scale failed");
 
   // **********************************
   // Add B - ((-omegaD) * AB))
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Add")));
-  result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE,BMKL,1.0,YTempMKL,&CMKL);
+  tm     = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Add")));
+  result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE, BMKL, 1.0, YTempMKL, &CMKL);
   typename KCRS::execution_space().fence();
-  if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed");
+  if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed");
 
   // **********************************
   // Copy out the data for MKL
   tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi MKL: Copy Out")));
   sparse_index_base_t c_indexing;
   MKL_INT c_rows, c_cols, *rows_start, *rows_end, *columns;
-  double * values;
-  mkl_sparse_d_export_csr(CMKL,&c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
-  size_t cnnz = rows_end[c_rows-1];
-  Kokkos::resize(Ccolind,cnnz);
-  Kokkos::resize(Cvals,cnnz);
-  if(c_rows != A.getLocalNumRows() || c_rows+1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
-  copy_view_n(c_rows,rows_start,Crowptr); Crowptr(c_rows) = rows_end[c_rows-1];
-  copy_view_n(cnnz,columns,Ccolind);
-  copy_view_n(cnnz,values,Cvals);
+  double *values;
+  mkl_sparse_d_export_csr(CMKL, &c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
+  size_t cnnz = rows_end[c_rows - 1];
+  Kokkos::resize(Ccolind, cnnz);
+  Kokkos::resize(Cvals, cnnz);
+  if (c_rows != A.getLocalNumRows() || c_rows + 1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
+  copy_view_n(c_rows, rows_start, Crowptr);
+  Crowptr(c_rows) = rows_end[c_rows - 1];
+  copy_view_n(cnnz, columns, Ccolind);
+  copy_view_n(cnnz, values, Cvals);
 
   Cnc->replaceColMap(Bu->getColMap());
   Cnc->setAllValues(Crowptr,
@@ -237,16 +237,12 @@ void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node
   mkl_sparse_destroy(XTempMKL);
   mkl_sparse_destroy(YTempMKL);
 
-
-
   tm = Teuchos::null;
   Au->getComm()->barrier();
 }
 
 #endif
 
-
-
 // =========================================================================
 // LTG Testing
 // =========================================================================
@@ -255,8 +251,8 @@ void Jacobi_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node
 #include "TpetraExt_MatrixMatrix_ExtraKernels_decl.hpp"
 #include "TpetraExt_MatrixMatrix_ExtraKernels_def.hpp"
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void Jacobi_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B,  Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C,const Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &D, Scalar omega, std::string jacobi_algorithm_name, std::string spgemm_algorithm_name, int team_work_size) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Jacobi_Wrapper(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C, const Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> &D, Scalar omega, std::string jacobi_algorithm_name, std::string spgemm_algorithm_name, int team_work_size) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -269,32 +265,32 @@ void Jacobi_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>
 
   if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_TPETRA_INST_OPENMP)
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef Tpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node>    vector_type;
-    typedef Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node>           import_type;
-    typedef typename crs_matrix_type::local_matrix_device_type    KCRS;
+    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+    typedef Tpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
+    typedef Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> import_type;
+    typedef typename crs_matrix_type::local_matrix_device_type KCRS;
     typedef typename KCRS::device_type device_t;
     typedef typename KCRS::StaticCrsGraphType graph_t;
     typedef typename graph_t::row_map_type::non_const_type lno_view_t;
     typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
     typedef typename KCRS::values_type::non_const_type scalar_view_t;
-    typedef Kokkos::View<LO*, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
-    typedef Tpetra::Map<LO,GO,NO>                                     map_type;
-    typedef typename map_type::local_map_type                         local_map_type;
+    typedef Kokkos::View<LO *, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
+    typedef Tpetra::Map<LO, GO, NO> map_type;
+    typedef typename map_type::local_map_type local_map_type;
     typedef typename Node::execution_space execution_space;
     typedef Kokkos::RangePolicy<execution_space, size_t> range_type;
     LocalOrdinal LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
     RCP<const import_type> Cimport;
-    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-    RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+    RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
 
     RCP<const vector_type> Du = Xpetra::toTpetra(D);
 
     // **********************************
     // Copy in the data for Jacobi Kernel Wrapper
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi "+name+": CopyIn")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Jacobi " + name + ": CopyIn")));
 
     Tpetra::CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node> Aview, Bview;
     Aview.origMatrix   = Au;
@@ -311,34 +307,35 @@ void Jacobi_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>
     Bview.domainMap    = Bu->getDomainMap();
     Bview.importColMap = Teuchos::null;
 
-
     // Because we're in serial...
     Cnc->replaceColMap(Bu->getColMap());
 
     // Bcol2Ccol is trivial
-    lo_view_t Bcol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Bcol2Ccol"),Bview.colMap->getLocalNumElements()), Icol2Ccol;
+    lo_view_t Bcol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Bcol2Ccol"), Bview.colMap->getLocalNumElements()), Icol2Ccol;
     const LO colMapSize = static_cast<LO>(Bview.colMap->getLocalNumElements());
-    Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::Bcol2Ccol_fill",
-                         Kokkos::RangePolicy<execution_space, LO>(0, colMapSize),
-                         KOKKOS_LAMBDA(const LO i) {
-                           Bcol2Ccol(i) = i;
-                         });
+    Kokkos::parallel_for(
+        "Tpetra::mult_A_B_newmatrix::Bcol2Ccol_fill",
+        Kokkos::RangePolicy<execution_space, LO>(0, colMapSize),
+        KOKKOS_LAMBDA(const LO i) {
+          Bcol2Ccol(i) = i;
+        });
 
     // Acol2Brow
     local_map_type Acolmap_local = Aview.colMap->getLocalMap();
     local_map_type Browmap_local = Bview.origMatrix->getRowMap()->getLocalMap();
-    lo_view_t targetMapToOrigRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToOrigRow"),Aview.colMap->getLocalNumElements());
+    lo_view_t targetMapToOrigRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToOrigRow"), Aview.colMap->getLocalNumElements());
     lo_view_t targetMapToImportRow;
-    Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::construct_tables",range_type(Aview.colMap->getMinLocalIndex(), Aview.colMap->getMaxLocalIndex()+1),KOKKOS_LAMBDA(const LO i) {
-      GO aidx = Acolmap_local.getGlobalElement(i);
-      LO B_LID = Browmap_local.getLocalElement(aidx);
-      if (B_LID != LO_INVALID) {
-        targetMapToOrigRow(i)   = B_LID;
-        //        targetMapToImportRow(i) = LO_INVALID;
-      } else {
-        // This shouldn't happen here
-      }
-    });
+    Kokkos::parallel_for(
+        "Tpetra::mult_A_B_newmatrix::construct_tables", range_type(Aview.colMap->getMinLocalIndex(), Aview.colMap->getMaxLocalIndex() + 1), KOKKOS_LAMBDA(const LO i) {
+          GO aidx  = Acolmap_local.getGlobalElement(i);
+          LO B_LID = Browmap_local.getLocalElement(aidx);
+          if (B_LID != LO_INVALID) {
+            targetMapToOrigRow(i) = B_LID;
+            //        targetMapToImportRow(i) = LO_INVALID;
+          } else {
+            // This shouldn't happen here
+          }
+        });
     tm = Teuchos::null;
     Au->getComm()->barrier();
 
@@ -347,36 +344,34 @@ void Jacobi_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>
     Teuchos::RCP<Teuchos::ParameterList> params = Teuchos::rcp(new Teuchos::ParameterList);
     params->set("openmp: jacobi algorithm", jacobi_algorithm_name);
     params->set("openmp: algorithm", spgemm_algorithm_name);
-    params->set("openmp: team work size",team_work_size);
+    params->set("openmp: team work size", team_work_size);
 
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(std::string("Jacobi ")+name+": Kernel")));
-    Tpetra::MMdetails::KernelWrappers2<SC,LO,GO,Node,lno_nnz_view_t>::jacobi_A_B_newmatrix_kernel_wrapper(omega,*Du,Aview,Bview,targetMapToOrigRow,targetMapToImportRow,Bcol2Ccol,Icol2Ccol,*Cnc,Cimport,name,params);
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(std::string("Jacobi ") + name + ": Kernel")));
+    Tpetra::MMdetails::KernelWrappers2<SC, LO, GO, Node, lno_nnz_view_t>::jacobi_A_B_newmatrix_kernel_wrapper(omega, *Du, Aview, Bview, targetMapToOrigRow, targetMapToImportRow, Bcol2Ccol, Icol2Ccol, *Cnc, Cimport, name, params);
 
     tm = Teuchos::null;
     Au->getComm()->barrier();
 
 #endif
-
   }
 }
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using std::endl;
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  using std::endl;
-
 
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -385,7 +380,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> pOut = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *pOut;
+    Teuchos::FancyOStream &out      = *pOut;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
@@ -393,63 +388,71 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // =========================================================================
     //    GO nx = 50, ny = 50, nz = 50;
     // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace3D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
 
-    bool   printTimings = true;  clp.setOption("timings", "notimings",  &printTimings, "print timings to screen");
-    int    nrepeat      = 100;   clp.setOption("nrepeat",               &nrepeat,      "repeat the experiment N times");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int nrepeat = 100;
+    clp.setOption("nrepeat", &nrepeat, "repeat the experiment N times");
     // the kernels
-    bool do_mkl      = true;
-    bool do_kk_mem   = true;
-    bool do_kk_dense = true;
+    bool do_mkl        = true;
+    bool do_kk_mem     = true;
+    bool do_kk_dense   = true;
     bool do_kk_default = true;
     bool do_ltg_msak   = true;
     bool do_ltg_jac    = true;
 
-    #ifndef HAVE_MUELU_MKL
-      do_mkl = false;
-    #endif
-    clp.setOption("mkl",        "nomkl",        &do_mkl,        "Evaluate MKL SpMM");
-    clp.setOption("kk_mem",     "nokk_mem",     &do_kk_mem,     "Evaluate KK Mem");
-    clp.setOption("kk_dense",   "nokk_dense",   &do_kk_dense,   "Evaluate KK Dense");
+#ifndef HAVE_MUELU_MKL
+    do_mkl = false;
+#endif
+    clp.setOption("mkl", "nomkl", &do_mkl, "Evaluate MKL SpMM");
+    clp.setOption("kk_mem", "nokk_mem", &do_kk_mem, "Evaluate KK Mem");
+    clp.setOption("kk_dense", "nokk_dense", &do_kk_dense, "Evaluate KK Dense");
     clp.setOption("kk_default", "nokk_default", &do_kk_default, "Evaluate KK Default");
-    clp.setOption("ltg_msak",   "noltg_msak",   &do_ltg_msak,   "Evaluate LTG (MSAK)");
-    clp.setOption("ltg_jac",    "noltg_jac",    &do_ltg_jac,    "Evaluate LTG (Jacobi)");
+    clp.setOption("ltg_msak", "noltg_msak", &do_ltg_msak, "Evaluate LTG (MSAK)");
+    clp.setOption("ltg_jac", "noltg_jac", &do_ltg_jac, "Evaluate LTG (Jacobi)");
 
-    int kk_team_work_size=16;
+    int kk_team_work_size = 16;
 
-    std::string matrixFileNameA = "A.mm"; clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
-    std::string matrixFileNameB = "B.mm"; clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
+    std::string matrixFileNameA = "A.mm";
+    clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
+    std::string matrixFileNameB = "B.mm";
+    clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    #ifndef HAVE_MUELU_MKL
-    if (do_mkl == true ) {
+#ifndef HAVE_MUELU_MKL
+    if (do_mkl == true) {
       out << "MKL was requested, but this kernel is not available. Disabling..." << endl;
       do_mkl = false;
     }
-    #endif
+#endif
 
     // simple hack to randomize order of experiments
-    enum class Experiments { MKL_SPMM=0, KK_MEM, KK_DENSE, KK_DEFAULT, LTG_MSAK, LTG_JAC };
+    enum class Experiments { MKL_SPMM = 0,
+                             KK_MEM,
+                             KK_DENSE,
+                             KK_DEFAULT,
+                             LTG_MSAK,
+                             LTG_JAC };
     std::vector<Experiments> my_experiments;
     // add the experiments we will run
-  
-    #ifdef HAVE_MUELU_MKL
-    if (do_mkl) my_experiments.push_back(Experiments::MKL_SPMM);   // MKL SPMM
-    #endif
 
-    // assume these are available
-    if (do_kk_mem)     my_experiments.push_back(Experiments::KK_MEM);     // KK Mem
-    if (do_kk_dense)   my_experiments.push_back(Experiments::KK_DENSE);   // KK Dense
-    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT); // KK Default
-    if (do_ltg_msak)   my_experiments.push_back(Experiments::LTG_MSAK);   // LTG MSAK
-    if (do_ltg_jac)    my_experiments.push_back(Experiments::LTG_JAC);    // LTG Jacobi
+#ifdef HAVE_MUELU_MKL
+    if (do_mkl) my_experiments.push_back(Experiments::MKL_SPMM);  // MKL SPMM
+#endif
 
+    // assume these are available
+    if (do_kk_mem) my_experiments.push_back(Experiments::KK_MEM);          // KK Mem
+    if (do_kk_dense) my_experiments.push_back(Experiments::KK_DENSE);      // KK Dense
+    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT);  // KK Default
+    if (do_ltg_msak) my_experiments.push_back(Experiments::LTG_MSAK);      // LTG MSAK
+    if (do_ltg_jac) my_experiments.push_back(Experiments::LTG_JAC);        // LTG Jacobi
 
     out << "========================================================" << endl
         << xpetraParameters
@@ -471,12 +474,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         << "========================================================" << endl;
 
 #if defined(HAVE_TPETRA_INST_OPENMP)
-    out<< "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = "<<Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency()<<endl
-       << "========================================================" << endl;
+    out << "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = " << Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() << endl
+        << "========================================================" << endl;
 #endif
 
     // At the moment, this test only runs on one MPI rank
-    if(comm->getSize() != 1) exit(1);
+    if (comm->getSize() != 1) exit(1);
 
     // =========================================================================
     // Problem construction
@@ -485,18 +488,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     comm->barrier();
 
-
-    RCP<Matrix> A = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameA), lib, comm);
-    RCP<Matrix> B = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameB), lib, comm);
+    RCP<Matrix> A = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameA), lib, comm);
+    RCP<Matrix> B = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameB), lib, comm);
     RCP<Matrix> C;
 
     // Jacobi-specific
-    RCP<Vector> Diagonal = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(A->getRowMap(),true);
+    RCP<Vector> Diagonal = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(A->getRowMap(), true);
     A->getLocalDiagCopy(*Diagonal);
-    RCP<Vector> D = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(A->getRowMap(),true);
+    RCP<Vector> D = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(A->getRowMap(), true);
     D->reciprocal(*Diagonal);
-    SC omega = (4.0/3.0)*Teuchos::ScalarTraits<Scalar>::one();
-      
+    SC omega = (4.0 / 3.0) * Teuchos::ScalarTraits<Scalar>::one();
 
     globalTimeMonitor = Teuchos::null;
     comm->barrier();
@@ -511,84 +512,81 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // random source
     std::random_device rd;
-    std::mt19937 random_source (rd());
+    std::mt19937 random_source(rd());
 
     // no need for a barrier, because the randomization process uses a collective.
-    if ( !my_experiments.empty() ) {
-
-    for (int i=0; i<nrepeat; i++) {
-      
-      // randomize the experiments
-      if (comm->getRank() == 0 ) {
-        std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
-      }
-      // Broadcast this ordering to the other processes
-      comm->broadcast(0,
-                      static_cast<int>(sizeof(Experiments::LTG_MSAK)*my_experiments.size()),
-                      reinterpret_cast<char *>(my_experiments.data()));
-
-      // loop over the randomized experiments
-      for (const auto& experiment_id : my_experiments) {
-        switch (experiment_id) {
- 
-        // MKL_SPMM
-        case Experiments::MKL_SPMM:
-          #ifdef HAVE_MUELU_MKL
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC MKL: Total"));
-            Jacobi_MKL_SPMM(*A,*B,*C,*D,omega);
-          }
-          #endif
-          break;
-        // KK Algorithms (KK Memory)
-        case Experiments::KK_MEM:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/SPGEMM_KK_MEMORY: Total"));
-            Jacobi_Wrapper(*A,*B,*C,*D,omega,std::string("MSAK"),std::string("SPGEMM_KK_MEMORY"),kk_team_work_size);
-          }
-          break;
-        // KK Algorithms (KK Dense)
-        case Experiments::KK_DENSE:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/SPGEMM_KK_DENSE: Total"));
-            Jacobi_Wrapper(*A,*B,*C,*D,omega,std::string("MSAK"),std::string("SPGEMM_KK_DENSE"),kk_team_work_size);
-          }
-          break;
-        // KK Algorithms (KK Default)
-        case Experiments::KK_DEFAULT:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC SPGEMM_KK: Total"));
-            Jacobi_Wrapper(*A,*B,*C,*D,omega,std::string("MSAK"),std::string("SPGEMM_KK"),kk_team_work_size);
-          }
-          break;
-        // LTG (MSAK)
-        case Experiments::LTG_MSAK:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/LTG: Total"));
-            Jacobi_Wrapper(*A,*B,*C,*D,omega,std::string("MSAK"),std::string("LTG"),0);
-          }
-          break;
-        // LTG (Jacobi)
-        case Experiments::LTG_JAC:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("JAC JLTG: Total"));
-            Jacobi_Wrapper(*A,*B,*C,*D,omega,std::string("LTG"),std::string("--"),0);
-          }
-          break;
-
-        default:
-          std::cerr << "Unknown experiment ID encountered: " << (int) experiment_id << std::endl;
+    if (!my_experiments.empty()) {
+      for (int i = 0; i < nrepeat; i++) {
+        // randomize the experiments
+        if (comm->getRank() == 0) {
+          std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
         }
-        comm->barrier();
-      }// end random exp loop
-    } // end repeat
-    } // end ! my_experiments.empty()
+        // Broadcast this ordering to the other processes
+        comm->broadcast(0,
+                        static_cast<int>(sizeof(Experiments::LTG_MSAK) * my_experiments.size()),
+                        reinterpret_cast<char *>(my_experiments.data()));
+
+        // loop over the randomized experiments
+        for (const auto &experiment_id : my_experiments) {
+          switch (experiment_id) {
+            // MKL_SPMM
+            case Experiments::MKL_SPMM:
+#ifdef HAVE_MUELU_MKL
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC MKL: Total"));
+                Jacobi_MKL_SPMM(*A, *B, *C, *D, omega);
+              }
+#endif
+              break;
+            // KK Algorithms (KK Memory)
+            case Experiments::KK_MEM:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/SPGEMM_KK_MEMORY: Total"));
+                Jacobi_Wrapper(*A, *B, *C, *D, omega, std::string("MSAK"), std::string("SPGEMM_KK_MEMORY"), kk_team_work_size);
+              }
+              break;
+            // KK Algorithms (KK Dense)
+            case Experiments::KK_DENSE:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/SPGEMM_KK_DENSE: Total"));
+                Jacobi_Wrapper(*A, *B, *C, *D, omega, std::string("MSAK"), std::string("SPGEMM_KK_DENSE"), kk_team_work_size);
+              }
+              break;
+            // KK Algorithms (KK Default)
+            case Experiments::KK_DEFAULT:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC SPGEMM_KK: Total"));
+                Jacobi_Wrapper(*A, *B, *C, *D, omega, std::string("MSAK"), std::string("SPGEMM_KK"), kk_team_work_size);
+              }
+              break;
+            // LTG (MSAK)
+            case Experiments::LTG_MSAK:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC MSAK/LTG: Total"));
+                Jacobi_Wrapper(*A, *B, *C, *D, omega, std::string("MSAK"), std::string("LTG"), 0);
+              }
+              break;
+            // LTG (Jacobi)
+            case Experiments::LTG_JAC:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("JAC JLTG: Total"));
+                Jacobi_Wrapper(*A, *B, *C, *D, omega, std::string("LTG"), std::string("--"), 0);
+              }
+              break;
+
+            default:
+              std::cerr << "Unknown experiment ID encountered: " << (int)experiment_id << std::endl;
+          }
+          comm->barrier();
+        }  // end random exp loop
+      }    // end repeat
+    }      // end ! my_experiments.empty()
 
     if (printTimings) {
       TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
@@ -598,17 +596,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
-
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/scaling/MMKernelDriver.cpp b/packages/muelu/test/scaling/MMKernelDriver.cpp
index 49dc90bd5e02..700d5ad316bb 100644
--- a/packages/muelu/test/scaling/MMKernelDriver.cpp
+++ b/packages/muelu/test/scaling/MMKernelDriver.cpp
@@ -44,48 +44,57 @@
 //
 // @HEADER
 
-#include <stdlib.h>                                     // for exit, EXIT_FA...
-#include <Teuchos_StandardCatchMacros.hpp>              // for TEUCHOS_STAND...
-#include <iostream>                                     // for endl, ostring...
-#include <limits>                                       // for numeric_limits
-#include <random>                                       // for mt19937, rand...
-#include <stdexcept>                                    // for runtime_error
-#include <string>                                       // for string
-#include <fstream>                                      // write CSV
+#include <stdlib.h>                         // for exit, EXIT_FA...
+#include <Teuchos_StandardCatchMacros.hpp>  // for TEUCHOS_STAND...
+#include <iostream>                         // for endl, ostring...
+#include <limits>                           // for numeric_limits
+#include <random>                           // for mt19937, rand...
+#include <stdexcept>                        // for runtime_error
+#include <string>                           // for string
+#include <fstream>                          // write CSV
 #include "MueLu_ConfigDefs.hpp"
-#include "KokkosKernels_Handle.hpp"                     // for KokkosKernels...
-#include "KokkosSparse_spgemm_handle.hpp"               // for StringToSPGEM...
-#include "KokkosSparse_spgemm_numeric.hpp"              // for spgemm_numeric
-#include "KokkosSparse_spgemm_symbolic.hpp"             // for spgemm_symbolic
-#include "Teuchos_CommandLineProcessor.hpp"             // for CommandLinePr...
-#include "Teuchos_ENull.hpp"                            // for ENull::null
-#include "Teuchos_FancyOStream.hpp"                     // for FancyOStream
-#include "Teuchos_OrdinalTraits.hpp"                    // for OrdinalTraits
-#include "Teuchos_RCP.hpp"                              // for rcp_const_cast
-#include "Teuchos_RCPDecl.hpp"                          // for RCP, rcp
-#include "Teuchos_ScalarTraitsDecl.hpp"                 // for ScalarTraits
-#include "Tpetra_CrsMatrix_decl.hpp"                    // for CrsMatrix
-#include "Tpetra_Import_decl.hpp"                       // for Import
-#include <Tpetra_Import_Util2.hpp> // Import_Util::sortCrsEntries
-#include "Tpetra_Map_decl.hpp"                          // for Map
+#include "KokkosKernels_Handle.hpp"          // for KokkosKernels...
+#include "KokkosSparse_spgemm_handle.hpp"    // for StringToSPGEM...
+#include "KokkosSparse_spgemm_numeric.hpp"   // for spgemm_numeric
+#include "KokkosSparse_spgemm_symbolic.hpp"  // for spgemm_symbolic
+#include "Teuchos_CommandLineProcessor.hpp"  // for CommandLinePr...
+#include "Teuchos_ENull.hpp"                 // for ENull::null
+#include "Teuchos_FancyOStream.hpp"          // for FancyOStream
+#include "Teuchos_OrdinalTraits.hpp"         // for OrdinalTraits
+#include "Teuchos_RCP.hpp"                   // for rcp_const_cast
+#include "Teuchos_RCPDecl.hpp"               // for RCP, rcp
+#include "Teuchos_ScalarTraitsDecl.hpp"      // for ScalarTraits
+#include "Tpetra_CrsMatrix_decl.hpp"         // for CrsMatrix
+#include "Tpetra_Import_decl.hpp"            // for Import
+#include <Tpetra_Import_Util2.hpp>           // Import_Util::sortCrsEntries
+#include "Tpetra_Map_decl.hpp"               // for Map
 #include <Xpetra_IO.hpp>
-#include "Xpetra_BlockedMap.hpp"                        // for LO, GO, NO
-#include "Xpetra_Map.hpp"                               // for UnderlyingLib
-#include "Xpetra_Parameters.hpp"                        // for Parameters
-namespace Teuchos { class ParameterList; }  // lines 92-92
-namespace Teuchos { class TimeMonitor; }  // lines 93-93
-namespace Teuchos { template <typename Ordinal> class Comm; }  // lines 94-94
-namespace Xpetra { template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node> class Matrix; }  // lines 95-95
+#include "Xpetra_BlockedMap.hpp"  // for LO, GO, NO
+#include "Xpetra_Map.hpp"         // for UnderlyingLib
+#include "Xpetra_Parameters.hpp"  // for Parameters
+namespace Teuchos {
+class ParameterList;
+}  // namespace Teuchos
+namespace Teuchos {
+class TimeMonitor;
+}  // namespace Teuchos
+namespace Teuchos {
+template <typename Ordinal>
+class Comm;
+}  // namespace Teuchos
+namespace Xpetra {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class Matrix;
+}  // namespace Xpetra
 
 #ifdef HAVE_VTUNE
 #include "ittnotify.h"
-__itt_domain *        vtune_domain_modLTG;
-__itt_string_handle * vtune_task_modLTG_setup;
-__itt_string_handle * vtune_task_modLTG_mult;
-__itt_string_handle * vtune_task_modLTG_copy;
+__itt_domain *vtune_domain_modLTG;
+__itt_string_handle *vtune_task_modLTG_setup;
+__itt_string_handle *vtune_task_modLTG_mult;
+__itt_string_handle *vtune_task_modLTG_copy;
 #endif
 
-
 // MueLu
 #include "MueLu.hpp"
 #include "MueLu_TestHelpers.hpp"
@@ -93,15 +102,12 @@ __itt_string_handle * vtune_task_modLTG_copy;
 #include <sys/types.h>
 #include <unistd.h>
 
-
-int getRssHWM ()
-{
+int getRssHWM() {
   static const std::string mem_file = "/proc/self/status";
   std::ifstream infile(mem_file);
 
   std::string line;
-  while (std::getline(infile, line))
-  {
+  while (std::getline(infile, line)) {
     if (line.rfind("VmHWM", 0) == 0) {
       // VmHWM:      1764 kB
       std::stringstream ss(line);
@@ -113,7 +119,7 @@ int getRssHWM ()
       return (value);
     }
   }
-  infile.close ();
+  infile.close();
   return (-1);
 }
 
@@ -122,100 +128,105 @@ int getRssHWM ()
 #include <iomanip>
 
 // this will generate a pedantic warning
-#define DescriptiveTime(x) ({\
-            int rssStart = -1; int rssEnd = -1; \
-            if (trackRssHWM) { \
-              rssStart =  getRssHWM (); \
-            } \
-            auto t0 = Time::now ();\
-            x;\
-            auto t1 = Time::now ();\
-            if (trackRssHWM) { \
-              rssEnd =  getRssHWM (); \
-              mem_vector.push_back(static_cast<double>(rssEnd - rssStart)); \
-            } \
-            double_secs ds = t1 - t0;\
-            if (descStats) \
-              timing_vector.push_back(std::chrono::duration_cast<ns>(ds).count ());\
-            })
-
-
-  std::map<std::string, std::vector<double> > my_experiment_timings;
-
-  std::chrono::steady_clock::time_point ltg_core_start;
-  std::chrono::steady_clock::time_point ltg_sort_start;
-  std::chrono::steady_clock::time_point ltg_copy_start;
-  std::chrono::steady_clock::time_point ltg_mult_start;
-
-  std::chrono::steady_clock::time_point ltg_core_stop;
-  std::chrono::steady_clock::time_point ltg_sort_stop;
-  std::chrono::steady_clock::time_point ltg_copy_stop;
-  std::chrono::steady_clock::time_point ltg_mult_stop;
-
-  // MKL
-  std::chrono::steady_clock::time_point mkl_mult_start;
-  std::chrono::steady_clock::time_point mkl_copyin_start;
-  std::chrono::steady_clock::time_point mkl_copyout_start;
-  std::chrono::steady_clock::time_point mkl_sort_start;
-
-  std::chrono::steady_clock::time_point mkl_mult_stop;
-  std::chrono::steady_clock::time_point mkl_copyin_stop;
-  std::chrono::steady_clock::time_point mkl_copyout_stop;
-  std::chrono::steady_clock::time_point mkl_sort_stop;
-
-  // ViennaCL
-  std::chrono::steady_clock::time_point vcl_mult_start;
-  std::chrono::steady_clock::time_point vcl_mult_stop;
-
-  // KK D
-  std::chrono::steady_clock::time_point kk_dense_mult_start;
-  std::chrono::steady_clock::time_point kk_dense_copyout_start;
-
-  std::chrono::steady_clock::time_point kk_dense_mult_stop;
-  std::chrono::steady_clock::time_point kk_dense_copyout_stop;
-
-  // KK M
-  std::chrono::steady_clock::time_point kk_mem_mult_start;
-  std::chrono::steady_clock::time_point kk_mem_copyout_start;
-
-  std::chrono::steady_clock::time_point kk_mem_mult_stop;
-  std::chrono::steady_clock::time_point kk_mem_copyout_stop;
-
-  // KK D
-  std::chrono::steady_clock::time_point kk_def_mult_start;
-  std::chrono::steady_clock::time_point kk_def_copyout_start;
-
-  std::chrono::steady_clock::time_point kk_def_mult_stop;
-  std::chrono::steady_clock::time_point kk_def_copyout_stop;
-
+#define DescriptiveTime(x) ({                                            \
+  int rssStart = -1;                                                     \
+  int rssEnd   = -1;                                                     \
+  if (trackRssHWM) {                                                     \
+    rssStart = getRssHWM();                                              \
+  }                                                                      \
+  auto t0 = Time::now();                                                 \
+  x;                                                                     \
+  auto t1 = Time::now();                                                 \
+  if (trackRssHWM) {                                                     \
+    rssEnd = getRssHWM();                                                \
+    mem_vector.push_back(static_cast<double>(rssEnd - rssStart));        \
+  }                                                                      \
+  double_secs ds = t1 - t0;                                              \
+  if (descStats)                                                         \
+    timing_vector.push_back(std::chrono::duration_cast<ns>(ds).count()); \
+})
+
+std::map<std::string, std::vector<double> > my_experiment_timings;
+
+std::chrono::steady_clock::time_point ltg_core_start;
+std::chrono::steady_clock::time_point ltg_sort_start;
+std::chrono::steady_clock::time_point ltg_copy_start;
+std::chrono::steady_clock::time_point ltg_mult_start;
+
+std::chrono::steady_clock::time_point ltg_core_stop;
+std::chrono::steady_clock::time_point ltg_sort_stop;
+std::chrono::steady_clock::time_point ltg_copy_stop;
+std::chrono::steady_clock::time_point ltg_mult_stop;
+
+// MKL
+std::chrono::steady_clock::time_point mkl_mult_start;
+std::chrono::steady_clock::time_point mkl_copyin_start;
+std::chrono::steady_clock::time_point mkl_copyout_start;
+std::chrono::steady_clock::time_point mkl_sort_start;
+
+std::chrono::steady_clock::time_point mkl_mult_stop;
+std::chrono::steady_clock::time_point mkl_copyin_stop;
+std::chrono::steady_clock::time_point mkl_copyout_stop;
+std::chrono::steady_clock::time_point mkl_sort_stop;
+
+// ViennaCL
+std::chrono::steady_clock::time_point vcl_mult_start;
+std::chrono::steady_clock::time_point vcl_mult_stop;
+
+// KK D
+std::chrono::steady_clock::time_point kk_dense_mult_start;
+std::chrono::steady_clock::time_point kk_dense_copyout_start;
+
+std::chrono::steady_clock::time_point kk_dense_mult_stop;
+std::chrono::steady_clock::time_point kk_dense_copyout_stop;
+
+// KK M
+std::chrono::steady_clock::time_point kk_mem_mult_start;
+std::chrono::steady_clock::time_point kk_mem_copyout_start;
+
+std::chrono::steady_clock::time_point kk_mem_mult_stop;
+std::chrono::steady_clock::time_point kk_mem_copyout_stop;
+
+// KK D
+std::chrono::steady_clock::time_point kk_def_mult_start;
+std::chrono::steady_clock::time_point kk_def_copyout_start;
+
+std::chrono::steady_clock::time_point kk_def_mult_stop;
+std::chrono::steady_clock::time_point kk_def_copyout_stop;
 
 #undef remove_pedantic_macro_warning_type
 #else
 #define DescriptiveTime(x) x
 #endif
 
-
 // =========================================================================
 // Support Routines
 // =========================================================================
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view_n(const int n, const View1 x1, View2 x2) {
-  Kokkos::parallel_for(n,KOKKOS_LAMBDA(const size_t i) {
-      x2[i] = x1[i];
-    });
+  Kokkos::parallel_for(
+      n, KOKKOS_LAMBDA(const size_t i) {
+        x2[i] = x1[i];
+      });
 }
 
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view(const View1 x1, View2 x2) {
-  copy_view_n(x1.extent(0),x1,x2);
+  copy_view_n(x1.extent(0), x1, x2);
 }
 
-enum class Experiments { ViennaCL=0, MKL_SPMM, KK_MEM, KK_DENSE, KK_DEFAULT, LTG, SERIAL};
+enum class Experiments { ViennaCL = 0,
+                         MKL_SPMM,
+                         KK_MEM,
+                         KK_DENSE,
+                         KK_DEFAULT,
+                         LTG,
+                         SERIAL };
 
-static inline std::string to_string (const Experiments& experiment_id) {
+static inline std::string to_string(const Experiments &experiment_id) {
   switch (experiment_id) {
     case Experiments::ViennaCL:
-      return("ViennaCL");
+      return ("ViennaCL");
     case Experiments::MKL_SPMM:
       return ("MKL");
     case Experiments::KK_MEM:
@@ -233,7 +244,6 @@ static inline std::string to_string (const Experiments& experiment_id) {
   }
 }
 
-
 // =========================================================================
 // ViennaCL Testing
 // =========================================================================
@@ -243,8 +253,8 @@ static inline std::string to_string (const Experiments& experiment_id) {
 #include "viennacl/linalg/prod.hpp"
 #include "viennacl/linalg/host_based/common.hpp"
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void Multiply_ViennaCL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B,  Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Multiply_ViennaCL(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -256,28 +266,28 @@ void Multiply_ViennaCL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
   // NOTE: ViennaCL matrices use "unsigned int" for rowptr and colind and are templated on scalar type (yay); which means (for us) POD only.
 
   if (lib == Xpetra::UseTpetra) {
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef typename crs_matrix_type::local_matrix_type    KCRS;
-    typedef typename KCRS::StaticCrsGraphType              graph_t;
+    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+    typedef typename crs_matrix_type::local_matrix_type KCRS;
+    typedef typename KCRS::StaticCrsGraphType graph_t;
     typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-    typedef typename graph_t::row_map_type::const_type     c_lno_view_t;
+    typedef typename graph_t::row_map_type::const_type c_lno_view_t;
     typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-    typedef typename graph_t::entries_type::const_type     c_lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type     scalar_view_t;
-    typedef typename Kokkos::View<unsigned int*,typename lno_nnz_view_t::array_layout,typename lno_nnz_view_t::device_type> vcl_size_t_type;
+    typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+    typedef typename KCRS::values_type::non_const_type scalar_view_t;
+    typedef typename Kokkos::View<unsigned int *, typename lno_nnz_view_t::array_layout, typename lno_nnz_view_t::device_type> vcl_size_t_type;
 
-    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-    RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+    RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
 
-    const KCRS & Amat = Au->getLocalMatrixDevice();
-    const KCRS & Bmat = Bu->getLocalMatrixDevice();
+    const KCRS &Amat = Au->getLocalMatrixDevice();
+    const KCRS &Bmat = Bu->getLocalMatrixDevice();
 
-    using no_init_view=Kokkos::ViewAllocateWithoutInitializing;
+    using no_init_view = Kokkos::ViewAllocateWithoutInitializing;
 
     c_lno_view_t Arowptr = Amat.graph.row_map, Browptr = Bmat.graph.row_map;
-    lno_view_t Crowptr(no_init_view("Crowptr"),C.getLocalNumRows()+1);
+    lno_view_t Crowptr(no_init_view("Crowptr"), C.getLocalNumRows() + 1);
     c_lno_nnz_view_t Acolind = Amat.graph.entries, Bcolind = Bmat.graph.entries;
     lno_nnz_view_t Ccolind;
     const scalar_view_t Avals = Amat.values, Bvals = Bmat.values;
@@ -286,24 +296,24 @@ void Multiply_ViennaCL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
     // **********************************
     // Copy in the data for ViennaCL
 
-    vcl_size_t_type ArowptrVCL(no_init_view("Arowptr"),Arowptr.extent(0));
-    vcl_size_t_type BrowptrVCL(no_init_view("Browptr"),Browptr.extent(0));
+    vcl_size_t_type ArowptrVCL(no_init_view("Arowptr"), Arowptr.extent(0));
+    vcl_size_t_type BrowptrVCL(no_init_view("Browptr"), Browptr.extent(0));
 
-    vcl_size_t_type AcolindVCL(no_init_view("Acolind"),Acolind.extent(0));
-    vcl_size_t_type BcolindVCL(no_init_view("Bcolind"),Bcolind.extent(0));
+    vcl_size_t_type AcolindVCL(no_init_view("Acolind"), Acolind.extent(0));
+    vcl_size_t_type BcolindVCL(no_init_view("Bcolind"), Bcolind.extent(0));
 
     viennacl::compressed_matrix<Scalar> AVCL, BVCL;
     {
-      Teuchos::TimeMonitor tm_copyin (*TimeMonitor::getNewTimer("MM ViennaCL: CopyIn"));
-      copy_view(Arowptr,ArowptrVCL);
-      copy_view(Browptr,BrowptrVCL);
-      copy_view(Acolind,AcolindVCL);
-      copy_view(Bcolind,BcolindVCL);
-  
+      Teuchos::TimeMonitor tm_copyin(*TimeMonitor::getNewTimer("MM ViennaCL: CopyIn"));
+      copy_view(Arowptr, ArowptrVCL);
+      copy_view(Browptr, BrowptrVCL);
+      copy_view(Acolind, AcolindVCL);
+      copy_view(Bcolind, BcolindVCL);
+
       AVCL.set(ArowptrVCL.data(), AcolindVCL.data(), Avals.data(),
-               Au->getLocalNumRows(),Au->getLocalNumCols(),Au->getLocalNumEntries());
+               Au->getLocalNumRows(), Au->getLocalNumCols(), Au->getLocalNumEntries());
       BVCL.set(BrowptrVCL.data(), BcolindVCL.data(), Bvals.data(),
-               Bu->getLocalNumRows(),Bu->getLocalNumCols(),Bu->getLocalNumEntries());
+               Bu->getLocalNumRows(), Bu->getLocalNumCols(), Bu->getLocalNumEntries());
     }
     Au->getComm()->barrier();
 
@@ -311,59 +321,58 @@ void Multiply_ViennaCL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
     // **********************************
     // Multiply
     {
-      Teuchos::TimeMonitor tm_mult (*TimeMonitor::getNewTimer("MM ViennaCL: Multiply"));
-      #ifdef USE_DESCRIPTIVE_STATS
-        vcl_mult_start = std::chrono::steady_clock::now ();
-      #endif
-  
+      Teuchos::TimeMonitor tm_mult(*TimeMonitor::getNewTimer("MM ViennaCL: Multiply"));
+#ifdef USE_DESCRIPTIVE_STATS
+      vcl_mult_start = std::chrono::steady_clock::now();
+#endif
+
       CVCL = viennacl::linalg::prod(AVCL, BVCL);
-  
-      #ifdef USE_DESCRIPTIVE_STATS
-        vcl_mult_stop = std::chrono::steady_clock::now ();
-      #endif
+
+#ifdef USE_DESCRIPTIVE_STATS
+      vcl_mult_stop = std::chrono::steady_clock::now();
+#endif
     }
     KCRS::execution_space::fence();
 
-
     // **********************************
     // Copy out the data for ViennaCL
     {
-      Teuchos::TimeMonitor tm_copy (*TimeMonitor::getNewTimer("MM ViennaCL: Copy Out"));
+      Teuchos::TimeMonitor tm_copy(*TimeMonitor::getNewTimer("MM ViennaCL: Copy Out"));
       size_t cnnz = (size_t)CVCL.nnz();
-      Ccolind = lno_nnz_view_t(Kokkos::ViewAllocateWithoutInitializing("Ccolind"), cnnz);
-      Cvals = scalar_view_t(Kokkos::ViewAllocateWithoutInitializing("Cvals"), cnnz);
-      #ifdef VIENNACL_WITH_CUDA
-        const unsigned int * CrowptrVCL = viennacl::cuda_arg<unsigned int>(CVCL.handle1());
-        const unsigned int * CcolindVCL = viennacl::cuda_arg<unsigned int>(CVCL.handle2());
-        const Scalar * CvalsVCL       = viennacl::cuda_arg<Scalar>(CVCL.handle());
-      #else
-        const unsigned int * CrowptrVCL = viennacl::linalg::host_based::detail::extract_raw_pointer<unsigned int>(CVCL.handle1());
-        const unsigned int * CcolindVCL = viennacl::linalg::host_based::detail::extract_raw_pointer<unsigned int>(CVCL.handle2());
-        const Scalar * CvalsVCL   = viennacl::linalg::host_based::detail::extract_raw_pointer<Scalar>(CVCL.handle());
-      #endif
-  
-      copy_view_n(Crowptr.extent(0),CrowptrVCL,Crowptr);
-      copy_view_n(cnnz,CcolindVCL,Ccolind);
-      copy_view_n(cnnz,CvalsVCL,Cvals);
-  
+      Ccolind     = lno_nnz_view_t(Kokkos::ViewAllocateWithoutInitializing("Ccolind"), cnnz);
+      Cvals       = scalar_view_t(Kokkos::ViewAllocateWithoutInitializing("Cvals"), cnnz);
+#ifdef VIENNACL_WITH_CUDA
+      const unsigned int *CrowptrVCL = viennacl::cuda_arg<unsigned int>(CVCL.handle1());
+      const unsigned int *CcolindVCL = viennacl::cuda_arg<unsigned int>(CVCL.handle2());
+      const Scalar *CvalsVCL         = viennacl::cuda_arg<Scalar>(CVCL.handle());
+#else
+      const unsigned int *CrowptrVCL = viennacl::linalg::host_based::detail::extract_raw_pointer<unsigned int>(CVCL.handle1());
+      const unsigned int *CcolindVCL = viennacl::linalg::host_based::detail::extract_raw_pointer<unsigned int>(CVCL.handle2());
+      const Scalar *CvalsVCL         = viennacl::linalg::host_based::detail::extract_raw_pointer<Scalar>(CVCL.handle());
+#endif
+
+      copy_view_n(Crowptr.extent(0), CrowptrVCL, Crowptr);
+      copy_view_n(cnnz, CcolindVCL, Ccolind);
+      copy_view_n(cnnz, CvalsVCL, Cvals);
+
       // set values
       {
         // Because we're in serial...
         Cnc->replaceColMap(Bu->getColMap());
-  
+
         Cnc->setAllValues(Crowptr,
                           Ccolind,
                           Cvals);
       }
     }
-    #ifdef USE_DESCRIPTIVE_STATS
-      {
-        std::chrono::duration<double> ds;
+#ifdef USE_DESCRIPTIVE_STATS
+    {
+      std::chrono::duration<double> ds;
 
-        ds = vcl_mult_stop - vcl_mult_start;
-        my_experiment_timings["ViennaCL: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
-      }
-    #endif
+      ds = vcl_mult_stop - vcl_mult_start;
+      my_experiment_timings["ViennaCL: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
+    }
+#endif
 
     Au->getComm()->barrier();
   }
@@ -371,48 +380,48 @@ void Multiply_ViennaCL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
 
 #endif
 
-
 // =========================================================================
 // MKL Testing
 // =========================================================================
 #ifdef HAVE_MUELU_MKL
-#include "mkl_spblas.h"                                 // for mkl_sparse_de...
-#include "mkl_types.h"                                  // for MKL_INT
+#include "mkl_spblas.h"  // for mkl_sparse_de...
+#include "mkl_types.h"   // for MKL_INT
 //#include "mkl.h"
 
-#define MMKD_MKL_ERROR_CHECK(rc) { \
-if (mkl_rc != SPARSE_STATUS_SUCCESS) { \
-  std::stringstream ss;  \
-  switch (mkl_rc) { \
-    case SPARSE_STATUS_NOT_INITIALIZED: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_NOT_INITIALIZED"; \
-      break; \
-    case SPARSE_STATUS_ALLOC_FAILED: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_ALLOC_FAILED"; \
-      break; \
-    case SPARSE_STATUS_INVALID_VALUE: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_INVALID_VALUE"; \
-      break; \
-    case SPARSE_STATUS_EXECUTION_FAILED: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_EXECUTION_FAILED"; \
-      break; \
-    case SPARSE_STATUS_INTERNAL_ERROR: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_INTERNAL_ERROR"; \
-      break; \
-    case SPARSE_STATUS_NOT_SUPPORTED: \
-      ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_NOT_SUPPORTED"; \
-      break; \
-    default: \
-      break; \
-  } \
-  std::cerr << ss.str () << std::endl; \
-  return; \
-} \
-}
+#define MMKD_MKL_ERROR_CHECK(rc)                                                \
+  {                                                                             \
+    if (mkl_rc != SPARSE_STATUS_SUCCESS) {                                      \
+      std::stringstream ss;                                                     \
+      switch (mkl_rc) {                                                         \
+        case SPARSE_STATUS_NOT_INITIALIZED:                                     \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_NOT_INITIALIZED";  \
+          break;                                                                \
+        case SPARSE_STATUS_ALLOC_FAILED:                                        \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_ALLOC_FAILED";     \
+          break;                                                                \
+        case SPARSE_STATUS_INVALID_VALUE:                                       \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_INVALID_VALUE";    \
+          break;                                                                \
+        case SPARSE_STATUS_EXECUTION_FAILED:                                    \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_EXECUTION_FAILED"; \
+          break;                                                                \
+        case SPARSE_STATUS_INTERNAL_ERROR:                                      \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_INTERNAL_ERROR";   \
+          break;                                                                \
+        case SPARSE_STATUS_NOT_SUPPORTED:                                       \
+          ss << "[ERROR] mkl::mkl_sparse_spmm: SPARSE_STATUS_NOT_SUPPORTED";    \
+          break;                                                                \
+        default:                                                                \
+          break;                                                                \
+      }                                                                         \
+      std::cerr << ss.str() << std::endl;                                       \
+      return;                                                                   \
+    }                                                                           \
+  }
 
-  // mkl_sparse_spmm
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C) {
+// mkl_sparse_spmm
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -421,30 +430,30 @@ void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
   static constexpr bool sortCSR = false;
 
   RCP<TimeMonitor> tm;
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
   typedef typename crs_matrix_type::local_matrix_device_type KCRS;
-  typedef typename KCRS::StaticCrsGraphType                  graph_t;
-  typedef typename graph_t::row_map_type::non_const_type     lno_view_t;
-  typedef typename graph_t::row_map_type::const_type         c_lno_view_t;
-  typedef typename graph_t::entries_type::non_const_type     lno_nnz_view_t;
-  typedef typename graph_t::entries_type::const_type         c_lno_nnz_view_t;
-  typedef typename KCRS::values_type::non_const_type         scalar_view_t;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
+  typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+  typedef typename graph_t::row_map_type::const_type c_lno_view_t;
+  typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+  typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
 
-  typedef typename Kokkos::View<MKL_INT*,typename lno_nnz_view_t::array_layout,typename lno_nnz_view_t::device_type> mkl_int_type;
+  typedef typename Kokkos::View<MKL_INT *, typename lno_nnz_view_t::array_layout, typename lno_nnz_view_t::device_type> mkl_int_type;
 
-  using no_init_view=Kokkos::ViewAllocateWithoutInitializing;
+  using no_init_view = Kokkos::ViewAllocateWithoutInitializing;
 
-  RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-  RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-  RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-  RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+  RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+  RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+  RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+  RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
 
-  const KCRS & Amat = Au->getLocalMatrixDevice();
-  const KCRS & Bmat = Bu->getLocalMatrixDevice();
-  if(A.getLocalNumRows()!=C.getLocalNumRows())  throw std::runtime_error("C is not sized correctly");
+  const KCRS &Amat = Au->getLocalMatrixDevice();
+  const KCRS &Bmat = Bu->getLocalMatrixDevice();
+  if (A.getLocalNumRows() != C.getLocalNumRows()) throw std::runtime_error("C is not sized correctly");
 
   c_lno_view_t Arowptr = Amat.graph.row_map, Browptr = Bmat.graph.row_map;
-  lno_view_t Crowptr(no_init_view("Crowptr"),C.getLocalNumRows()+1);
+  lno_view_t Crowptr(no_init_view("Crowptr"), C.getLocalNumRows() + 1);
   c_lno_nnz_view_t Acolind = Amat.graph.entries, Bcolind = Bmat.graph.entries;
   const scalar_view_t Avals = Amat.values, Bvals = Bmat.values;
 
@@ -453,116 +462,114 @@ void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
   sparse_matrix_t CMKL;
   sparse_status_t mkl_rc;
 
-  mkl_int_type ArowptrMKL(no_init_view("Arowptr"),Arowptr.extent(0));
-  mkl_int_type BrowptrMKL(no_init_view("Browptr"),Browptr.extent(0));
+  mkl_int_type ArowptrMKL(no_init_view("Arowptr"), Arowptr.extent(0));
+  mkl_int_type BrowptrMKL(no_init_view("Browptr"), Browptr.extent(0));
 
-  mkl_int_type AcolindMKL(no_init_view("Acolind"),Acolind.extent(0));
-  mkl_int_type BcolindMKL(no_init_view("Bcolind"),Bcolind.extent(0));
+  mkl_int_type AcolindMKL(no_init_view("Acolind"), Acolind.extent(0));
+  mkl_int_type BcolindMKL(no_init_view("Bcolind"), Bcolind.extent(0));
 
   {
     // **********************************
     // Copy in the data for MKL
-    Teuchos::TimeMonitor tm_copyin (*TimeMonitor::getNewTimer("MM MKL: CopyIn"));
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_copyin_start = std::chrono::steady_clock::now ();
-    #endif
-    copy_view(Arowptr,ArowptrMKL);
-    copy_view(Browptr,BrowptrMKL);
-    copy_view(Acolind,AcolindMKL);
-    copy_view(Bcolind,BcolindMKL);
-
-    if(std::is_same<Scalar,double>::value) {
+    Teuchos::TimeMonitor tm_copyin(*TimeMonitor::getNewTimer("MM MKL: CopyIn"));
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_copyin_start = std::chrono::steady_clock::now();
+#endif
+    copy_view(Arowptr, ArowptrMKL);
+    copy_view(Browptr, BrowptrMKL);
+    copy_view(Acolind, AcolindMKL);
+    copy_view(Bcolind, BcolindMKL);
+
+    if (std::is_same<Scalar, double>::value) {
       mkl_rc = mkl_sparse_d_create_csr(&AMKL, SPARSE_INDEX_BASE_ZERO,
-                                        Au->getLocalNumRows(),
-                                        Au->getLocalNumCols(),
-                                        ArowptrMKL.data(),
-                                        ArowptrMKL.data()+1,
-                                        AcolindMKL.data(),
-                                        (double*)Avals.data());
+                                       Au->getLocalNumRows(),
+                                       Au->getLocalNumCols(),
+                                       ArowptrMKL.data(),
+                                       ArowptrMKL.data() + 1,
+                                       AcolindMKL.data(),
+                                       (double *)Avals.data());
       MMKD_MKL_ERROR_CHECK(mkl_rc);
       mkl_rc = mkl_sparse_d_create_csr(&BMKL, SPARSE_INDEX_BASE_ZERO,
-                                        Bu->getLocalNumRows(),
-                                        Bu->getLocalNumCols(),
-                                        BrowptrMKL.data(),
-                                        BrowptrMKL.data()+1,
-                                        BcolindMKL.data(),
-                                        (double*)Bvals.data());
+                                       Bu->getLocalNumRows(),
+                                       Bu->getLocalNumCols(),
+                                       BrowptrMKL.data(),
+                                       BrowptrMKL.data() + 1,
+                                       BcolindMKL.data(),
+                                       (double *)Bvals.data());
       MMKD_MKL_ERROR_CHECK(mkl_rc);
-    }
-    else
+    } else
       throw std::runtime_error("MKL Type Mismatch");
 
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_copyin_stop = std::chrono::steady_clock::now ();
-    #endif
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_copyin_stop = std::chrono::steady_clock::now();
+#endif
   }
   Au->getComm()->barrier();
 
   // **********************************
   // Multiply
   {
-    Teuchos::TimeMonitor tm_mult (*TimeMonitor::getNewTimer("MM MKL: Multiply"));
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_mult_start = std::chrono::steady_clock::now ();
-    #endif
+    Teuchos::TimeMonitor tm_mult(*TimeMonitor::getNewTimer("MM MKL: Multiply"));
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_mult_start = std::chrono::steady_clock::now();
+#endif
     mkl_rc = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, BMKL, &CMKL);
     MMKD_MKL_ERROR_CHECK(mkl_rc);
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_mult_stop = std::chrono::steady_clock::now ();
-    #endif
-    //KCRS::execution_space::fence();
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_mult_stop = std::chrono::steady_clock::now();
+#endif
+    // KCRS::execution_space::fence();
   }
   // **********************************
 
   // Copy out the data for MKL
-  Teuchos::TimeMonitor tm_copy (*TimeMonitor::getNewTimer("MM MKL: Copy Out"));
-  #ifdef USE_DESCRIPTIVE_STATS
-    mkl_copyout_start = std::chrono::steady_clock::now ();
-  #endif
+  Teuchos::TimeMonitor tm_copy(*TimeMonitor::getNewTimer("MM MKL: Copy Out"));
+#ifdef USE_DESCRIPTIVE_STATS
+  mkl_copyout_start = std::chrono::steady_clock::now();
+#endif
 
   sparse_index_base_t c_indexing;
   MKL_INT c_rows, c_cols, *rows_start, *rows_end, *columns;
-  double * values;
-  mkl_rc = mkl_sparse_d_export_csr(CMKL,&c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
+  double *values;
+  mkl_rc = mkl_sparse_d_export_csr(CMKL, &c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
   MMKD_MKL_ERROR_CHECK(mkl_rc);
-  size_t cnnz = rows_end[c_rows-1];
+  size_t cnnz            = rows_end[c_rows - 1];
   lno_nnz_view_t Ccolind = lno_nnz_view_t(Kokkos::ViewAllocateWithoutInitializing("Ccolind"), cnnz);
-  scalar_view_t Cvals = scalar_view_t(Kokkos::ViewAllocateWithoutInitializing("Cvals"), cnnz);
-  Kokkos::resize(Ccolind,cnnz);
-  Kokkos::resize(Cvals,cnnz);
-  if((size_t) c_rows != A.getLocalNumRows() || (size_t) c_rows+1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
-  copy_view_n(c_rows,rows_start,Crowptr); Crowptr(c_rows) = rows_end[c_rows-1];
-  copy_view_n(cnnz,columns,Ccolind);
-  copy_view_n(cnnz,values,Cvals);
-
-  #ifdef USE_DESCRIPTIVE_STATS
-    mkl_copyout_stop = std::chrono::steady_clock::now ();
-  #endif
+  scalar_view_t Cvals    = scalar_view_t(Kokkos::ViewAllocateWithoutInitializing("Cvals"), cnnz);
+  Kokkos::resize(Ccolind, cnnz);
+  Kokkos::resize(Cvals, cnnz);
+  if ((size_t)c_rows != A.getLocalNumRows() || (size_t)c_rows + 1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
+  copy_view_n(c_rows, rows_start, Crowptr);
+  Crowptr(c_rows) = rows_end[c_rows - 1];
+  copy_view_n(cnnz, columns, Ccolind);
+  copy_view_n(cnnz, values, Cvals);
+
+#ifdef USE_DESCRIPTIVE_STATS
+  mkl_copyout_stop = std::chrono::steady_clock::now();
+#endif
 
   if (sortCSR == true) {
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_sort_start = std::chrono::steady_clock::now ();
-    #endif
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_sort_start = std::chrono::steady_clock::now();
+#endif
 
     Tpetra::Import_Util::sortCrsEntries(Crowptr,
-                                Ccolind,
-                                Cvals);
+                                        Ccolind,
+                                        Cvals);
 
-    #ifdef USE_DESCRIPTIVE_STATS
-      mkl_sort_stop = std::chrono::steady_clock::now ();
-    #endif
+#ifdef USE_DESCRIPTIVE_STATS
+    mkl_sort_stop = std::chrono::steady_clock::now();
+#endif
   }
 
   // set values
   {
-
     // Because we're in serial...
     Cnc->replaceColMap(Bu->getColMap());
 
     Cnc->setAllValues(Crowptr,
                       Ccolind,
                       Cvals);
-
   }
 
   mkl_rc = mkl_sparse_destroy(AMKL);
@@ -572,23 +579,23 @@ void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
   mkl_rc = mkl_sparse_destroy(CMKL);
   MMKD_MKL_ERROR_CHECK(mkl_rc);
 
-  #ifdef USE_DESCRIPTIVE_STATS
-    {
-      std::chrono::duration<double> ds;
+#ifdef USE_DESCRIPTIVE_STATS
+  {
+    std::chrono::duration<double> ds;
 
-      ds = mkl_mult_stop - mkl_mult_start;
-      my_experiment_timings["MKL: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+    ds = mkl_mult_stop - mkl_mult_start;
+    my_experiment_timings["MKL: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
 
-      ds = mkl_copyin_stop - mkl_copyin_start;
-      my_experiment_timings["MKL: CopyIn"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+    ds = mkl_copyin_stop - mkl_copyin_start;
+    my_experiment_timings["MKL: CopyIn"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
 
-      ds = mkl_copyout_stop - mkl_copyout_start;
-      my_experiment_timings["MKL: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+    ds = mkl_copyout_stop - mkl_copyout_start;
+    my_experiment_timings["MKL: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
 
-      ds = mkl_sort_stop - mkl_sort_start;
-      my_experiment_timings["MKL: Sort"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
-    }
-  #endif
+    ds = mkl_sort_stop - mkl_sort_start;
+    my_experiment_timings["MKL: Sort"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
+  }
+#endif
 }
 
 #undef MMKD_MKL_ERROR_CHECK
@@ -598,8 +605,8 @@ void Multiply_MKL_SPMM(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,No
 // =========================================================================
 // Kokkos Kernels Testing
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B,  Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C, std::string algorithm_name, int team_work_size) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C, std::string algorithm_name, int team_work_size) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -607,29 +614,29 @@ void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdin
 
   Xpetra::UnderlyingLib lib = A.getRowMap()->lib();
 
-  std::string prefix = std::string("MM KokkosKernels ")+algorithm_name + std::string(": ");
+  std::string prefix = std::string("MM KokkosKernels ") + algorithm_name + std::string(": ");
 
   if (lib == Xpetra::UseTpetra) {
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef typename crs_matrix_type::local_matrix_device_type  KCRS;
-    typedef typename KCRS::StaticCrsGraphType                   graph_t;
-    typedef typename graph_t::row_map_type::non_const_type      lno_view_t;
-    typedef typename graph_t::row_map_type::const_type          c_lno_view_t;
-    typedef typename graph_t::entries_type::non_const_type      lno_nnz_view_t;
-    typedef typename graph_t::entries_type::const_type          c_lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type          scalar_view_t;
-    typedef typename KCRS::device_type                          device_t;
-
-    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-    RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
-
-    const KCRS & Amat = Au->getLocalMatrixDevice();
-    const KCRS & Bmat = Bu->getLocalMatrixDevice();
+    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+    typedef typename crs_matrix_type::local_matrix_device_type KCRS;
+    typedef typename KCRS::StaticCrsGraphType graph_t;
+    typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+    typedef typename graph_t::row_map_type::const_type c_lno_view_t;
+    typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+    typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+    typedef typename KCRS::values_type::non_const_type scalar_view_t;
+    typedef typename KCRS::device_type device_t;
+
+    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+    RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+
+    const KCRS &Amat = Au->getLocalMatrixDevice();
+    const KCRS &Bmat = Bu->getLocalMatrixDevice();
 
     c_lno_view_t Arowptr = Amat.graph.row_map, Browptr = Bmat.graph.row_map;
-    lno_view_t Crowptr("Crowptr",A.getLocalNumRows()+1);
+    lno_view_t Crowptr("Crowptr", A.getLocalNumRows() + 1);
     c_lno_nnz_view_t Acolind = Amat.graph.entries, Bcolind = Bmat.graph.entries;
     const scalar_view_t Avals = Amat.values, Bvals = Bmat.values;
     lno_nnz_view_t Ccolind;
@@ -637,9 +644,10 @@ void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdin
 
     // KokkosKernelsHandle
     typedef KokkosKernels::Experimental::KokkosKernelsHandle<
-    typename lno_view_t::const_value_type,typename lno_nnz_view_t::const_value_type, typename scalar_view_t::const_value_type,
-      typename device_t::execution_space, typename device_t::memory_space,typename device_t::memory_space > KernelHandle;
-    KokkosSparse::SPGEMMAlgorithm alg_enum = KokkosSparse::StringToSPGEMMAlgorithm(algorithm_name);
+        typename lno_view_t::const_value_type, typename lno_nnz_view_t::const_value_type, typename scalar_view_t::const_value_type,
+        typename device_t::execution_space, typename device_t::memory_space, typename device_t::memory_space>
+        KernelHandle;
+    KokkosSparse::SPGEMMAlgorithm alg_enum    = KokkosSparse::StringToSPGEMMAlgorithm(algorithm_name);
     typename KernelHandle::nnz_lno_t AnumRows = Au->getLocalNumRows();
     typename KernelHandle::nnz_lno_t BnumRows = Bu->getLocalNumRows();
     typename KernelHandle::nnz_lno_t BnumCols = Bu->getLocalNumCols();
@@ -647,123 +655,123 @@ void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdin
     // **********************************
     // Multiply
     {
-      Teuchos::TimeMonitor tm_mult (*TimeMonitor::getNewTimer(prefix + std::string("Multiply")));
-      #ifdef USE_DESCRIPTIVE_STATS
+      Teuchos::TimeMonitor tm_mult(*TimeMonitor::getNewTimer(prefix + std::string("Multiply")));
+#ifdef USE_DESCRIPTIVE_STATS
       switch (alg_enum) {
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_DENSE:
-          kk_dense_mult_start = std::chrono::steady_clock::now ();
+          kk_dense_mult_start = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_MEMORY:
-          kk_mem_mult_start = std::chrono::steady_clock::now ();
+          kk_mem_mult_start = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK:
-          kk_def_mult_start = std::chrono::steady_clock::now ();
+          kk_def_mult_start = std::chrono::steady_clock::now();
           break;
         default:
           break;
       }
-      #endif
+#endif
       KernelHandle kh;
       kh.create_spgemm_handle(alg_enum);
       kh.set_team_work_size(team_work_size);
-  
-      KokkosSparse::Experimental::spgemm_symbolic(&kh,AnumRows,BnumRows,BnumCols,
-                                                      Arowptr,Acolind,false,
-                                                      Browptr,Bcolind,false,
-                                                      Crowptr);
-  
+
+      KokkosSparse::Experimental::spgemm_symbolic(&kh, AnumRows, BnumRows, BnumCols,
+                                                  Arowptr, Acolind, false,
+                                                  Browptr, Bcolind, false,
+                                                  Crowptr);
+
       size_t c_nnz_size = kh.get_spgemm_handle()->get_c_nnz();
-      if (c_nnz_size){
-        using no_init_view=Kokkos::ViewAllocateWithoutInitializing;
-        Ccolind = lno_nnz_view_t (no_init_view("entriesC"), c_nnz_size);
-        Cvals = scalar_view_t (no_init_view("valuesC"), c_nnz_size);
+      if (c_nnz_size) {
+        using no_init_view = Kokkos::ViewAllocateWithoutInitializing;
+        Ccolind            = lno_nnz_view_t(no_init_view("entriesC"), c_nnz_size);
+        Cvals              = scalar_view_t(no_init_view("valuesC"), c_nnz_size);
       }
-      KokkosSparse::Experimental::spgemm_numeric(&kh,AnumRows,BnumRows,BnumCols,
-                                                     Arowptr,Acolind,Avals,false,
-                                                     Browptr,Bcolind,Bvals,false,
-                                                     Crowptr,Ccolind,Cvals);
+      KokkosSparse::Experimental::spgemm_numeric(&kh, AnumRows, BnumRows, BnumCols,
+                                                 Arowptr, Acolind, Avals, false,
+                                                 Browptr, Bcolind, Bvals, false,
+                                                 Crowptr, Ccolind, Cvals);
       kh.destroy_spgemm_handle();
       typename KCRS::execution_space().fence();
-  
-      #ifdef USE_DESCRIPTIVE_STATS
+
+#ifdef USE_DESCRIPTIVE_STATS
       switch (alg_enum) {
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_DENSE:
-          kk_dense_mult_stop = std::chrono::steady_clock::now ();
+          kk_dense_mult_stop = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_MEMORY:
-          kk_mem_mult_stop = std::chrono::steady_clock::now ();
+          kk_mem_mult_stop = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK:
-          kk_def_mult_stop = std::chrono::steady_clock::now ();
+          kk_def_mult_stop = std::chrono::steady_clock::now();
           break;
         default:
           break;
       }
-      #endif
-    } // scope time monitor mult
+#endif
+    }  // scope time monitor mult
     Au->getComm()->barrier();
 
     // **********************************
     // Copy out the data for KokkosKernels
     {
-      Teuchos::TimeMonitor tm_copy (*TimeMonitor::getNewTimer(prefix + std::string("Copy Out")));
-      #ifdef USE_DESCRIPTIVE_STATS
+      Teuchos::TimeMonitor tm_copy(*TimeMonitor::getNewTimer(prefix + std::string("Copy Out")));
+#ifdef USE_DESCRIPTIVE_STATS
       switch (alg_enum) {
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_DENSE:
-          kk_dense_copyout_start = std::chrono::steady_clock::now ();
+          kk_dense_copyout_start = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_MEMORY:
-          kk_mem_copyout_start = std::chrono::steady_clock::now ();
+          kk_mem_copyout_start = std::chrono::steady_clock::now();
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK:
-          kk_def_copyout_start = std::chrono::steady_clock::now ();
+          kk_def_copyout_start = std::chrono::steady_clock::now();
           break;
         default:
           break;
       }
-      #endif
+#endif
       // Because we're in serial...
       Cnc->replaceColMap(Bu->getColMap());
       Cnc->setAllValues(Crowptr,
                         Ccolind,
                         Cvals);
-  
-      #ifdef USE_DESCRIPTIVE_STATS
+
+#ifdef USE_DESCRIPTIVE_STATS
       std::chrono::duration<double> ds;
       switch (alg_enum) {
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_DENSE:
-          kk_dense_copyout_stop = std::chrono::steady_clock::now ();
-  
+          kk_dense_copyout_stop = std::chrono::steady_clock::now();
+
           ds = kk_dense_mult_stop - kk_dense_mult_start;
-          my_experiment_timings["KK DENSE: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
-  
+          my_experiment_timings["KK DENSE: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
+
           ds = kk_dense_copyout_stop - kk_dense_copyout_start;
-          my_experiment_timings["KK DENSE: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+          my_experiment_timings["KK DENSE: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK_MEMORY:
-          kk_mem_copyout_stop = std::chrono::steady_clock::now ();
-  
+          kk_mem_copyout_stop = std::chrono::steady_clock::now();
+
           ds = kk_mem_mult_stop - kk_mem_mult_start;
-          my_experiment_timings["KK MEM: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
-  
+          my_experiment_timings["KK MEM: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
+
           ds = kk_mem_copyout_stop - kk_mem_copyout_start;
-          my_experiment_timings["KK MEM: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+          my_experiment_timings["KK MEM: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
           break;
         case KokkosSparse::SPGEMMAlgorithm::SPGEMM_KK:
-          kk_def_copyout_stop = std::chrono::steady_clock::now ();
-  
+          kk_def_copyout_stop = std::chrono::steady_clock::now();
+
           ds = kk_def_mult_stop - kk_def_mult_start;
-          my_experiment_timings["KK DEFAULT: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
-  
+          my_experiment_timings["KK DEFAULT: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
+
           ds = kk_def_copyout_stop - kk_def_copyout_start;
-          my_experiment_timings["KK DEFAULT: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+          my_experiment_timings["KK DEFAULT: CopyOut"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
           break;
         default:
           break;
       }
-      #endif
-    } // scope for Teuchos time monitor
-  } // lib
+#endif
+    }  // scope for Teuchos time monitor
+  }    // lib
 }
 
 // =========================================================================
@@ -774,170 +782,167 @@ void Multiply_KokkosKernels(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdin
 #include "TpetraExt_MatrixMatrix_ExtraKernels_decl.hpp"
 #include "TpetraExt_MatrixMatrix_ExtraKernels_def.hpp"
 
-//The LTG kernel is only defined for the Kokkos OpenMP node, so
-//its test must only be enabled for OpenMP
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-struct LTG_Tests
-{
+// The LTG kernel is only defined for the Kokkos OpenMP node, so
+// its test must only be enabled for OpenMP
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+struct LTG_Tests {
   static void Multiply_LTG(
-      const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&,
-      const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&,
-      Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>&)
-  {}
+      const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &,
+      const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &,
+      Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &) {}
 };
 
 #ifdef HAVE_TPETRA_INST_OPENMP
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal>
-struct LTG_Tests<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosOpenMPWrapperNode>
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal>
+struct LTG_Tests<Scalar, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosOpenMPWrapperNode> {
   typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode Node;
-  static void Multiply_LTG(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,
-                  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B,
-                        Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C) {
+  static void Multiply_LTG(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A,
+                           const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B,
+                           Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C) {
 #include <MueLu_UseShortNames.hpp>
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::TimeMonitor;
-
-  Xpetra::UnderlyingLib lib = A.getRowMap()->lib();
-
-  if (lib == Xpetra::UseTpetra) {
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node>          import_type;
-    typedef typename crs_matrix_type::local_matrix_device_type    KCRS;
-    typedef typename KCRS::device_type device_t;
-    typedef typename KCRS::StaticCrsGraphType graph_t;
-    typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-    typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type scalar_view_t;
-    typedef Kokkos::View<LO*, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
-    typedef Tpetra::Map<LO,GO,NO>                                     map_type;
-    typedef typename map_type::local_map_type                         local_map_type;
-    typedef typename Node::execution_space execution_space;
-    typedef Kokkos::RangePolicy<execution_space, size_t> range_type;
-    LocalOrdinal LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
-    RCP<const import_type> Cimport;
-    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-    RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B,false));
-    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-    RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
-
-    using no_init_view=Kokkos::ViewAllocateWithoutInitializing;
+    using Teuchos::RCP;
+    using Teuchos::rcp;
+    using Teuchos::TimeMonitor;
+
+    Xpetra::UnderlyingLib lib = A.getRowMap()->lib();
+
+    if (lib == Xpetra::UseTpetra) {
+      typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+      typedef Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> import_type;
+      typedef typename crs_matrix_type::local_matrix_device_type KCRS;
+      typedef typename KCRS::device_type device_t;
+      typedef typename KCRS::StaticCrsGraphType graph_t;
+      typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+      typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+      typedef typename KCRS::values_type::non_const_type scalar_view_t;
+      typedef Kokkos::View<LO *, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
+      typedef Tpetra::Map<LO, GO, NO> map_type;
+      typedef typename map_type::local_map_type local_map_type;
+      typedef typename Node::execution_space execution_space;
+      typedef Kokkos::RangePolicy<execution_space, size_t> range_type;
+      LocalOrdinal LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
+      RCP<const import_type> Cimport;
+      RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A, false));
+      RCP<const crs_matrix_type> Bu = Utilities::Op2TpetraCrs(rcp(&B, false));
+      RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C, false));
+      RCP<crs_matrix_type> Cnc      = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+
+      using no_init_view = Kokkos::ViewAllocateWithoutInitializing;
+
+      // **********************************
+      // Copy in the data for LTG
+
+      Tpetra::CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node> Aview, Bview;
+      Aview.origMatrix   = Au;
+      Aview.origRowMap   = Au->getRowMap();
+      Aview.rowMap       = Au->getRowMap();
+      Aview.colMap       = Au->getColMap();
+      Aview.domainMap    = Au->getDomainMap();
+      Aview.importColMap = Teuchos::null;
+
+      Bview.origMatrix   = Bu;
+      Bview.origRowMap   = Bu->getRowMap();
+      Bview.rowMap       = Bu->getRowMap();
+      Bview.colMap       = Bu->getColMap();
+      Bview.domainMap    = Bu->getDomainMap();
+      Bview.importColMap = Teuchos::null;
+
+      lo_view_t Bcol2Ccol(no_init_view("Bcol2Ccol"), Bview.colMap->getLocalNumElements()), Icol2Ccol;
+      const LO colMapSize = static_cast<LO>(Bview.colMap->getLocalNumElements());
+
+      local_map_type Acolmap_local = Aview.colMap->getLocalMap();
+      local_map_type Browmap_local = Bview.origMatrix->getRowMap()->getLocalMap();
+      lo_view_t targetMapToOrigRow(no_init_view("targetMapToOrigRow"), Aview.colMap->getLocalNumElements());
+      lo_view_t targetMapToImportRow;
+
+      // Copy in work
+      {
+        Teuchos::TimeMonitor tm_copyin(*TimeMonitor::getNewTimer("MM LTG: CopyIn"));
 
-    // **********************************
-    // Copy in the data for LTG
-
-    Tpetra::CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node> Aview, Bview;
-    Aview.origMatrix   = Au;
-    Aview.origRowMap   = Au->getRowMap();
-    Aview.rowMap       = Au->getRowMap();
-    Aview.colMap       = Au->getColMap();
-    Aview.domainMap    = Au->getDomainMap();
-    Aview.importColMap = Teuchos::null;
-
-    Bview.origMatrix   = Bu;
-    Bview.origRowMap   = Bu->getRowMap();
-    Bview.rowMap       = Bu->getRowMap();
-    Bview.colMap       = Bu->getColMap();
-    Bview.domainMap    = Bu->getDomainMap();
-    Bview.importColMap = Teuchos::null;
-
-    lo_view_t Bcol2Ccol(no_init_view("Bcol2Ccol"),Bview.colMap->getLocalNumElements()), Icol2Ccol;
-    const LO colMapSize = static_cast<LO>(Bview.colMap->getLocalNumElements());
-
-    local_map_type Acolmap_local = Aview.colMap->getLocalMap();
-    local_map_type Browmap_local = Bview.origMatrix->getRowMap()->getLocalMap();
-    lo_view_t targetMapToOrigRow(no_init_view("targetMapToOrigRow"),Aview.colMap->getLocalNumElements());
-    lo_view_t targetMapToImportRow;
-
-    // Copy in work
-    {
-      Teuchos::TimeMonitor tm_copyin (*TimeMonitor::getNewTimer("MM LTG: CopyIn"));
+        // Because we're in serial...
+        Cnc->replaceColMap(Bu->getColMap());
 
-      // Because we're in serial...
-      Cnc->replaceColMap(Bu->getColMap());
-  
-      // Bcol2Ccol is trivial
-      Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::Bcol2Ccol_fill",
-                           Kokkos::RangePolicy<execution_space, LO>(0, colMapSize),
-                           KOKKOS_LAMBDA(const LO i) {
-                             Bcol2Ccol(i) = i;
-                           });
-  
-      // Acol2Brow
-      Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::construct_tables",
-                           range_type(Aview.colMap->getMinLocalIndex(),
-                                      Aview.colMap->getMaxLocalIndex()+1),
-      KOKKOS_LAMBDA(const LO i)
+        // Bcol2Ccol is trivial
+        Kokkos::parallel_for(
+            "Tpetra::mult_A_B_newmatrix::Bcol2Ccol_fill",
+            Kokkos::RangePolicy<execution_space, LO>(0, colMapSize),
+            KOKKOS_LAMBDA(const LO i) {
+              Bcol2Ccol(i) = i;
+            });
+
+        // Acol2Brow
+        Kokkos::parallel_for(
+            "Tpetra::mult_A_B_newmatrix::construct_tables",
+            range_type(Aview.colMap->getMinLocalIndex(),
+                       Aview.colMap->getMaxLocalIndex() + 1),
+            KOKKOS_LAMBDA(const LO i) {
+              GO aidx  = Acolmap_local.getGlobalElement(i);
+              LO B_LID = Browmap_local.getLocalElement(aidx);
+              if (B_LID != LO_INVALID) {
+                targetMapToOrigRow(i) = B_LID;
+                //        targetMapToImportRow(i) = LO_INVALID;
+              } else {
+                // This shouldn't happen here
+              }
+            });
+
+      }  // copyin
+      Au->getComm()->barrier();
+
+      // **********************************
+      // Multiply
+      Teuchos::RCP<Teuchos::ParameterList> params = Teuchos::parameterList();
+      params->set("sort entries", false);
       {
-        GO aidx = Acolmap_local.getGlobalElement(i);
-        LO B_LID = Browmap_local.getLocalElement(aidx);
-        if (B_LID != LO_INVALID) {
-          targetMapToOrigRow(i)   = B_LID;
-          //        targetMapToImportRow(i) = LO_INVALID;
-        } else {
-          // This shouldn't happen here
+        Teuchos::TimeMonitor tm(*TimeMonitor::getNewTimer("MM LTG: Multiply"));
+#ifdef USE_DESCRIPTIVE_STATS
+        ltg_mult_start = std::chrono::steady_clock::now();
+#endif
+#ifdef HAVE_VTUNE
+        __itt_resume();
+#endif
+        Tpetra::MatrixMatrix::ExtraKernels::mult_A_B_newmatrix_LowThreadGustavsonKernel(
+            Aview, Bview,
+            targetMapToOrigRow, targetMapToImportRow,
+            Bcol2Ccol, Icol2Ccol,
+            *Cnc, Cimport,
+            std::string("LTG"), params);
+#ifdef HAVE_VTUNE
+        __itt_pause();
+#endif
+#ifdef USE_DESCRIPTIVE_STATS
+        ltg_mult_stop = std::chrono::steady_clock::now();
+        {
+          std::chrono::duration<double> ds;
+          ds = ltg_mult_stop - ltg_mult_start;
+          my_experiment_timings["LTG: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count());
         }
-      });
-  
-    } //copyin
-    Au->getComm()->barrier();
-
-    // **********************************
-    // Multiply
-    Teuchos::RCP<Teuchos::ParameterList> params = Teuchos::parameterList();
-    params->set("sort entries", false);
-    {
-      Teuchos::TimeMonitor tm (*TimeMonitor::getNewTimer("MM LTG: Multiply"));
-      #ifdef USE_DESCRIPTIVE_STATS
-      ltg_mult_start = std::chrono::steady_clock::now ();
-      #endif
-      #ifdef HAVE_VTUNE
-      __itt_resume();
-      #endif
-      Tpetra::MatrixMatrix::ExtraKernels::mult_A_B_newmatrix_LowThreadGustavsonKernel(
-              Aview,Bview,
-              targetMapToOrigRow,targetMapToImportRow,
-              Bcol2Ccol,Icol2Ccol,
-              *Cnc,Cimport,
-              std::string("LTG"),params);
-      #ifdef HAVE_VTUNE
-      __itt_pause ();
-      #endif
-      #ifdef USE_DESCRIPTIVE_STATS
-      ltg_mult_stop = std::chrono::steady_clock::now ();
-      {
-        std::chrono::duration<double> ds;
-        ds = ltg_mult_stop - ltg_mult_start;
-        my_experiment_timings["LTG: Call"].push_back(std::chrono::duration_cast<std::chrono::nanoseconds>(ds).count ());
+#endif
       }
-      #endif
+      Au->getComm()->barrier();
     }
-    Au->getComm()->barrier();
   }
-}
 };
-#endif //HAVE OPENMP
+#endif  // HAVE OPENMP
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void dump_matrix (const std::string& prefix,
-                  const std::string& kernel_name,
-                  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C)
-{
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void dump_matrix(const std::string &prefix,
+                 const std::string &kernel_name,
+                 const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C) {
   std::ostringstream ss;
   ss << prefix << "_" << kernel_name << ".mtx";
-  Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Write (ss.str(), C);
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write(ss.str(), C);
 }
 
-inline bool file_exists (const std::string& name) {
-    std::ifstream f(name.c_str());
-    return f.good();
+inline bool file_exists(const std::string &name) {
+  std::ifstream f(name.c_str());
+  return f.good();
 }
 
 static int prior_pos;
-void simple_progress_bar (const int iter, std::ostream& out, const int iter_max) {
-  double progress = (double) iter / (double) iter_max;
+void simple_progress_bar(const int iter, std::ostream &out, const int iter_max) {
+  double progress = (double)iter / (double)iter_max;
 
   int barWidth = 70;
 
@@ -948,9 +953,12 @@ void simple_progress_bar (const int iter, std::ostream& out, const int iter_max)
   if (new_pos > prior_pos) {
     out << "[";
     for (int i = 0; i < barWidth; ++i) {
-        if (i < new_pos) out << "=";
-        else if (i == new_pos) out << ">";
-        else out << " ";
+      if (i < new_pos)
+        out << "=";
+      else if (i == new_pos)
+        out << ">";
+      else
+        out << " ";
     }
     out << "] " << int(progress * 100.0) << " %\r";
     out.flush();
@@ -958,10 +966,8 @@ void simple_progress_bar (const int iter, std::ostream& out, const int iter_max)
   prior_pos = new_pos;
 }
 
-
-void print_desc_stats (const std::string& csvFile, const bool writeCSV, std::ostream& out)
-{
-  #if USE_DESCRIPTIVE_STATS == 1
+void print_desc_stats(const std::string &csvFile, const bool writeCSV, std::ostream &out) {
+#if USE_DESCRIPTIVE_STATS == 1
   typedef DescriptiveStats::ns ns;
 
   std::ostringstream oss;
@@ -972,66 +978,66 @@ void print_desc_stats (const std::string& csvFile, const bool writeCSV, std::ost
   const double timer_ratio = double(ns::period::num) / double(ns::period::den);
 
   DescriptiveStats::print_descriptive_stats(oss_header, stats, timer_ratio, "header", true, true);
-  DescriptiveStats::profile_timer (oss);
+  DescriptiveStats::profile_timer(oss);
 
   // loop through and dump the stats to the stringstream
-  for (auto& dataset : my_experiment_timings ) {
-   auto& measurements = dataset.second;
-   const bool compute_median_CI = true;
-   DescriptiveStats::get_descriptive_stats (measurements, measurements.size(), stats, compute_median_CI);
-   DescriptiveStats::print_descriptive_stats(oss, stats, timer_ratio, dataset.first, true);
+  for (auto &dataset : my_experiment_timings) {
+    auto &measurements           = dataset.second;
+    const bool compute_median_CI = true;
+    DescriptiveStats::get_descriptive_stats(measurements, measurements.size(), stats, compute_median_CI);
+    DescriptiveStats::print_descriptive_stats(oss, stats, timer_ratio, dataset.first, true);
   }
 
   // write to 'out'
-  out << oss_header.str () << oss.str ();
+  out << oss_header.str() << oss.str();
 
   // if rank == 0...
   if (writeCSV) {
     std::ofstream csv_fptr;
-    const bool haveExistingCSV = file_exists (csvFile);
-    csv_fptr.open (csvFile, std::ios_base::app);
+    const bool haveExistingCSV = file_exists(csvFile);
+    csv_fptr.open(csvFile, std::ios_base::app);
 
     out << " Writing CSV [" << csvFile << "]...";
-    if (! haveExistingCSV) {
+    if (!haveExistingCSV) {
       out << "adding header...";
-      csv_fptr << oss_header.str ();
+      csv_fptr << oss_header.str();
     }
-    csv_fptr << oss.str ();
-    csv_fptr.close ();
+    csv_fptr << oss.str();
+    csv_fptr.close();
     out << "Wrote data... done" << std::endl;
   }
-  #endif // if have descriptive stats
+#endif  // if have descriptive stats
 
-} // end print desc stats
+}  // end print desc stats
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using std::endl;
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  using std::endl;
 
-  #if USE_DESCRIPTIVE_STATS == 1
+#if USE_DESCRIPTIVE_STATS == 1
   typedef DescriptiveStats::Time Time;
   typedef DescriptiveStats::ns ns;
   typedef DescriptiveStats::double_secs double_secs;
   typedef DescriptiveStats::descriptive_stat_map_type descriptive_stat_map_type;
-  #endif
+#endif
 
-  #ifdef HAVE_VTUNE
-  #pragma message "Pausing Vtune"
-  __itt_pause ();
-  #endif
+#ifdef HAVE_VTUNE
+#pragma message "Pausing Vtune"
+  __itt_pause();
+#endif
 
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
     // disable this annoyance
     Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
 
@@ -1040,97 +1046,108 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // =========================================================================
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> pOut = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *pOut;
+    Teuchos::FancyOStream &out      = *pOut;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
-    clp.throwExceptions (false);
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    bool printTimings  = true; clp.setOption("timings", "notimings",  &printTimings, "print timings to screen");
-    bool descStats     = true; clp.setOption("descStats", "nodescStats",  &descStats, "track all timings and compute descriptive statistics (uses more memory+time)");
-    int  nrepeat       = 100;  clp.setOption("nrepeat",               &nrepeat,      "repeat the experiment N times");
-    bool printFNorm    = true; clp.setOption("fnorm", "nofnorm", &printFNorm, "Compute the Frobenius norm after each multiplication");
-
-    bool printTypes    = true; clp.setOption("types", "noTypes", &printTypes, "Print the types being used");
-    bool printMats     = true; clp.setOption("printMats", "noPrintMats", &printMats, "Describe matrices");
-    bool progressBar   = true; clp.setOption("progressBar", "noProgressBar", &progressBar, "Show an ascii progress bar while executing.");
-    bool writeCSV      = false; clp.setOption("writeCSV", "noWriteCSV", &writeCSV, "Write a CSV file if descriptive stats is enabled. See --csvFile=NAME");
-    std::string csvFile = "desc_stats.csv"; clp.setOption("csvFile", &csvFile, "CSV file to store desc stats in.");
-
-    bool trackRssHWM = false; clp.setOption("trackRssHWM", "noTrackRssHWM", &trackRssHWM, "Track the RSS high water mark (assumes no HP usage)");
+    clp.throwExceptions(false);
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
+
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    bool descStats = true;
+    clp.setOption("descStats", "nodescStats", &descStats, "track all timings and compute descriptive statistics (uses more memory+time)");
+    int nrepeat = 100;
+    clp.setOption("nrepeat", &nrepeat, "repeat the experiment N times");
+    bool printFNorm = true;
+    clp.setOption("fnorm", "nofnorm", &printFNorm, "Compute the Frobenius norm after each multiplication");
+
+    bool printTypes = true;
+    clp.setOption("types", "noTypes", &printTypes, "Print the types being used");
+    bool printMats = true;
+    clp.setOption("printMats", "noPrintMats", &printMats, "Describe matrices");
+    bool progressBar = true;
+    clp.setOption("progressBar", "noProgressBar", &progressBar, "Show an ascii progress bar while executing.");
+    bool writeCSV = false;
+    clp.setOption("writeCSV", "noWriteCSV", &writeCSV, "Write a CSV file if descriptive stats is enabled. See --csvFile=NAME");
+    std::string csvFile = "desc_stats.csv";
+    clp.setOption("csvFile", &csvFile, "CSV file to store desc stats in.");
+
+    bool trackRssHWM = false;
+    clp.setOption("trackRssHWM", "noTrackRssHWM", &trackRssHWM, "Track the RSS high water mark (assumes no HP usage)");
 
     // the kernels
-    bool do_viennaCL = true;
-    bool do_mkl      = true;
-    bool do_kk_mem   = true;
-    bool do_kk_dense = true;
+    bool do_viennaCL   = true;
+    bool do_mkl        = true;
+    bool do_kk_mem     = true;
+    bool do_kk_dense   = true;
     bool do_kk_default = true;
     bool do_ltg        = true;
 
-    #ifndef HAVE_MUELU_VIENNACL
-      do_viennaCL = false;
-    #endif
+#ifndef HAVE_MUELU_VIENNACL
+    do_viennaCL = false;
+#endif
 
-    #ifndef HAVE_MUELU_MKL
-      do_mkl = false;
-    #endif
-    clp.setOption("viennaCL",   "noviennaCL",   &do_viennaCL,   "Evaluate ViennaCL");
-    clp.setOption("mkl",        "nomkl",        &do_mkl,        "Evaluate MKL SpMM");
-    clp.setOption("kk_mem",     "nokk_mem",     &do_kk_mem,     "Evaluate KK Mem");
-    clp.setOption("kk_dense",   "nokk_dense",   &do_kk_dense,   "Evaluate KK Dense");
+#ifndef HAVE_MUELU_MKL
+    do_mkl = false;
+#endif
+    clp.setOption("viennaCL", "noviennaCL", &do_viennaCL, "Evaluate ViennaCL");
+    clp.setOption("mkl", "nomkl", &do_mkl, "Evaluate MKL SpMM");
+    clp.setOption("kk_mem", "nokk_mem", &do_kk_mem, "Evaluate KK Mem");
+    clp.setOption("kk_dense", "nokk_dense", &do_kk_dense, "Evaluate KK Dense");
     clp.setOption("kk_default", "nokk_default", &do_kk_default, "Evaluate KK Default");
-    clp.setOption("ltg",        "noltg",        &do_ltg,        "Evaluate LTG");
+    clp.setOption("ltg", "noltg", &do_ltg, "Evaluate LTG");
 
     std::string dump_result = "";
     clp.setOption("dump_result", &dump_result, "Dump the resulting matrices (once)");
 
-    int kk_team_work_size=16;
+    int kk_team_work_size = 16;
 
-    std::string matrixFileNameA = "A.mm"; clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
-    std::string matrixFileNameB = "B.mm"; clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
+    std::string matrixFileNameA = "A.mm";
+    clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
+    std::string matrixFileNameB = "B.mm";
+    clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
 
     clp.recogniseAllOptions(false);
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    #ifndef HAVE_MUELU_VIENNACL
-    if (do_viennaCL == true ) {
+#ifndef HAVE_MUELU_VIENNACL
+    if (do_viennaCL == true) {
       out << "ViennaCL was requested, but this kernel is not available. Disabling..." << endl;
       do_viennaCL = false;
     }
-    #endif
+#endif
 
-    #ifndef HAVE_MUELU_MKL
-    if (do_mkl == true ) {
+#ifndef HAVE_MUELU_MKL
+    if (do_mkl == true) {
       out << "MKL was requested, but this kernel is not available. Disabling..." << endl;
       do_mkl = false;
     }
-    #endif
+#endif
 
     // simple hack to randomize order of experiments
     std::vector<Experiments> my_experiments;
-    // add the experiments we will run
-    #ifdef HAVE_MUELU_VIENNACL
-    if (do_viennaCL) my_experiments.push_back(Experiments::ViennaCL);   // ViennaCL
-    #endif
+// add the experiments we will run
+#ifdef HAVE_MUELU_VIENNACL
+    if (do_viennaCL) my_experiments.push_back(Experiments::ViennaCL);  // ViennaCL
+#endif
 
-    #ifdef HAVE_MUELU_MKL
-    if (do_mkl) my_experiments.push_back(Experiments::MKL_SPMM);   // MKL SPMM
-    #endif
+#ifdef HAVE_MUELU_MKL
+    if (do_mkl) my_experiments.push_back(Experiments::MKL_SPMM);  // MKL SPMM
+#endif
 
     // assume these are available
-    if (do_kk_mem)     my_experiments.push_back(Experiments::KK_MEM);     // KK Mem
-    if (do_kk_dense)   my_experiments.push_back(Experiments::KK_DENSE);   // KK Dense
-    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT); // KK Default
-    if (do_ltg)        my_experiments.push_back(Experiments::LTG);        // LTG
-
+    if (do_kk_mem) my_experiments.push_back(Experiments::KK_MEM);          // KK Mem
+    if (do_kk_dense) my_experiments.push_back(Experiments::KK_DENSE);      // KK Dense
+    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT);  // KK Default
+    if (do_ltg) my_experiments.push_back(Experiments::LTG);                // LTG
 
     if (printTypes) {
       out << "========================================================" << endl
@@ -1153,76 +1170,70 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           << "========================================================" << endl;
     }
 
-//    bool dump_matrices = false;
-//    if (dump_result != "") dump_matrices = true;
+    //    bool dump_matrices = false;
+    //    if (dump_result != "") dump_matrices = true;
 
-    //Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
+    // Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
 
     // At the moment, this test only runs on one MPI rank
-    if(comm->getSize() != 1) exit(1);
+    if (comm->getSize() != 1) exit(1);
 
     std::stringstream ss;
 
-    // this stuff was designed to report nested parallelism it isn't super helpful
-    // but at some point I was working with OMP_NESTED=true, which is very different
-    // from Kokkos' team based parallelism with OpenMP
-    #ifdef HAVE_MUELU_OPENMP
-    #pragma omp parallel
+// this stuff was designed to report nested parallelism it isn't super helpful
+// but at some point I was working with OMP_NESTED=true, which is very different
+// from Kokkos' team based parallelism with OpenMP
+#ifdef HAVE_MUELU_OPENMP
+#pragma omp parallel
     {
-      int num_threads = omp_get_num_threads ();
+      int num_threads = omp_get_num_threads();
 
-      #pragma omp for ordered
-      for (int n=0;n<num_threads;n++)
-      {
-        #pragma omp critical
+#pragma omp for ordered
+      for (int n = 0; n < num_threads; n++) {
+#pragma omp critical
         {
-          ss  << "Level: " << omp_get_level ()
-              << ", Active Level: " << omp_get_active_level ()
-              << ", Team #: " << omp_get_team_num () + 1 << " / " << omp_get_num_teams ()
-              << ", Thread: " << omp_get_thread_num () + 1 << " / " << omp_get_num_threads()
-              << ", cpu: " << sched_getcpu()
-              << std::endl;
+          ss << "Level: " << omp_get_level()
+             << ", Active Level: " << omp_get_active_level()
+             << ", Team #: " << omp_get_team_num() + 1 << " / " << omp_get_num_teams()
+             << ", Thread: " << omp_get_thread_num() + 1 << " / " << omp_get_num_threads()
+             << ", cpu: " << sched_getcpu()
+             << std::endl;
         }
 
-
-
-      #pragma omp parallel
-      {
-          if (omp_get_active_level () ==  omp_get_level ()) {
-
-          #pragma omp critical
-          {
-            ss << "Level: " << omp_get_level ()
-               << ", Active Level: " << omp_get_active_level ()
-               << ", Parent TID: " << omp_get_ancestor_thread_num(1) + 1
-               << ", Team #: " << omp_get_team_num () + 1<< " / " << omp_get_num_teams ()
-               << ", Team Size: " << omp_get_team_size(2)
-               << ", Thread: " << omp_get_thread_num () + 1<< " / " << omp_get_num_threads()
-               << ", cpu: " << sched_getcpu()
-               << std::endl;
-          }
+#pragma omp parallel
+        {
+          if (omp_get_active_level() == omp_get_level()) {
+#pragma omp critical
+            {
+              ss << "Level: " << omp_get_level()
+                 << ", Active Level: " << omp_get_active_level()
+                 << ", Parent TID: " << omp_get_ancestor_thread_num(1) + 1
+                 << ", Team #: " << omp_get_team_num() + 1 << " / " << omp_get_num_teams()
+                 << ", Team Size: " << omp_get_team_size(2)
+                 << ", Thread: " << omp_get_thread_num() + 1 << " / " << omp_get_num_threads()
+                 << ", cpu: " << sched_getcpu()
+                 << std::endl;
+            }
           } else {
-
-          #pragma omp critical
-          {
-            ss << "Level: " << omp_get_level ()
-               << ", Active Level: " << omp_get_active_level ()
-               << ", Parent TID: " << omp_get_ancestor_thread_num(1) + 1
-               << ", Team #: " << omp_get_team_num () + 1<< " / " << omp_get_num_teams ()
-               << ", Team Size: " << omp_get_team_size(2)
-               << ", Thread: " << omp_get_thread_num () + 1<< " / " << omp_get_num_threads()
-               << ", cpu: " << sched_getcpu()
-               << std::endl;
-          }
-
+#pragma omp critical
+            {
+              ss << "Level: " << omp_get_level()
+                 << ", Active Level: " << omp_get_active_level()
+                 << ", Parent TID: " << omp_get_ancestor_thread_num(1) + 1
+                 << ", Team #: " << omp_get_team_num() + 1 << " / " << omp_get_num_teams()
+                 << ", Team Size: " << omp_get_team_size(2)
+                 << ", Thread: " << omp_get_thread_num() + 1 << " / " << omp_get_num_threads()
+                 << ", cpu: " << sched_getcpu()
+                 << std::endl;
+            }
           }
+        }
       }
-     }
     }
 
-    out << ss.str ();
+    out << ss.str();
     out.flush();
-    #endif // have openmp
+#endif  // have openmp
 
     // =========================================================================
     // Problem construction
@@ -1231,27 +1242,26 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     comm->barrier();
 
-    RCP<Matrix> A = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameA), lib, comm);
+    RCP<Matrix> A = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameA), lib, comm);
     RCP<Matrix> B;
 
-
-    if ( matrixFileNameB == "dupe" ) {
+    if (matrixFileNameB == "dupe") {
       *B = *A;
 
       if (verbose) out << "duplicating A" << endl;
-    }
-    else if ( matrixFileNameB == "alias" ) {
+    } else if (matrixFileNameB == "alias") {
       B = A;
       if (verbose) out << "Aliasing A" << endl;
-    }
-    else {
-      B = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameB), lib, comm);
+    } else {
+      B = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameB), lib, comm);
     }
 
     globalTimeMonitor = Teuchos::null;
     comm->barrier();
 
-    if (verbose) { out << "Matrix Read complete." << endl; }
+    if (verbose) {
+      out << "Matrix Read complete." << endl;
+    }
 
     if (printMats) {
       out << "Matrix A:" << endl
@@ -1260,152 +1270,149 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           << "Matrix B:" << endl
           << *B
           << "========================================================" << endl;
-//      if (do_serial)
-//      out << "Matrix A:" << endl
-//          << *Aserial
-//          << "========================================================" << endl
-//          << "Matrix B:" << endl
-//          << *Bserial
-//          << "========================================================" << endl;
+      //      if (do_serial)
+      //      out << "Matrix A:" << endl
+      //          << *Aserial
+      //          << "========================================================" << endl
+      //          << "Matrix B:" << endl
+      //          << *Bserial
+      //          << "========================================================" << endl;
     }
 
     // random source
     std::random_device rd;
-    std::mt19937 random_source (rd());
+    std::mt19937 random_source(rd());
 
     // no need for a barrier, because the randomization process uses a collective.
-    if (! my_experiments.empty() ) {
-    std::vector<double> mem_dummy;
-    double prior_fnorm = std::numeric_limits<double>::max();
-    int fnorm_failures = 0;
-
-    globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Kernels Total Time")));
-    for (int i=0; i<nrepeat; i++) {
+    if (!my_experiments.empty()) {
+      std::vector<double> mem_dummy;
+      double prior_fnorm = std::numeric_limits<double>::max();
+      int fnorm_failures = 0;
+
+      globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Kernels Total Time")));
+      for (int i = 0; i < nrepeat; i++) {
+        // randomize the experiments
+        if (comm->getRank() == 0) {
+          std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
+        }
+        // Broadcast this ordering to the other processes
+        comm->broadcast(0,
+                        static_cast<int>(sizeof(Experiments::LTG) * my_experiments.size()),
+                        reinterpret_cast<char *>(my_experiments.data()));
+
+        // loop over the randomized experiments
+        for (const auto &experiment_id : my_experiments) {
+          RCP<Matrix> C;
+          std::string kernel_name = to_string(experiment_id);
+#if USE_DESCRIPTIVE_STATS == 1
+          auto &timing_vector = my_experiment_timings[kernel_name + ": Total"];
+          auto &mem_vector    = (trackRssHWM) ? my_experiment_timings[kernel_name + ": RssHWM"] : mem_dummy;
+#endif
 
-      // randomize the experiments
-      if (comm->getRank() == 0 ) {
-        std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
-      }
-      // Broadcast this ordering to the other processes
-      comm->broadcast(0,
-                      static_cast<int>(sizeof(Experiments::LTG)*my_experiments.size()),
-                      reinterpret_cast<char *>(my_experiments.data()));
-
-      // loop over the randomized experiments
-      for (const auto& experiment_id : my_experiments) {
-        RCP<Matrix> C;
-        std::string kernel_name = to_string(experiment_id);
-        #if USE_DESCRIPTIVE_STATS == 1
-        auto& timing_vector = my_experiment_timings[kernel_name + ": Total" ];
-        auto& mem_vector = (trackRssHWM) ? my_experiment_timings[kernel_name+ ": RssHWM"] : mem_dummy;
-        #endif
-
-        switch (experiment_id) {
-        // ViennaCL
-        case Experiments::ViennaCL:
-          #ifdef HAVE_MUELU_VIENNACL
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM ViennaCL: Total"));
-            DescriptiveTime(Multiply_ViennaCL(*A,*B,*C));
-          }
-          #endif
-          break;
-        // MKL_SPMM
-        case Experiments::MKL_SPMM:
-          #ifdef HAVE_MUELU_MKL
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM MKL: Total"));
-            DescriptiveTime(Multiply_MKL_SPMM(*A,*B,*C));
-          }
-          #endif
-          break;
-        // KK Algorithms (KK Memory)
-        case Experiments::KK_MEM:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK_MEMORY: Total"));
-            DescriptiveTime(Multiply_KokkosKernels(*A,*B,*C,std::string("SPGEMM_KK_MEMORY"),kk_team_work_size));
-          }
-          break;
-        // KK Algorithms (KK Dense)
-        case Experiments::KK_DENSE:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK_DENSE: Total"));
-            DescriptiveTime(Multiply_KokkosKernels(*A,*B,*C,std::string("SPGEMM_KK_DENSE"),kk_team_work_size));
-          }
-          break;
-        // KK Algorithms (KK Default)
-        case Experiments::KK_DEFAULT:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK: Total"));
-            DescriptiveTime(Multiply_KokkosKernels(*A,*B,*C,std::string("SPGEMM_KK"),kk_team_work_size));
-          }
-          break;
-        // LTG
-        case Experiments::LTG:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM LTG: Total"));
-            using ltg_tester=LTG_Tests<SC,LO,GO,Node>;
-            DescriptiveTime(ltg_tester::Multiply_LTG(*A,*B,*C));
+          switch (experiment_id) {
+            // ViennaCL
+            case Experiments::ViennaCL:
+#ifdef HAVE_MUELU_VIENNACL
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM ViennaCL: Total"));
+                DescriptiveTime(Multiply_ViennaCL(*A, *B, *C));
+              }
+#endif
+              break;
+            // MKL_SPMM
+            case Experiments::MKL_SPMM:
+#ifdef HAVE_MUELU_MKL
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM MKL: Total"));
+                DescriptiveTime(Multiply_MKL_SPMM(*A, *B, *C));
+              }
+#endif
+              break;
+            // KK Algorithms (KK Memory)
+            case Experiments::KK_MEM:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK_MEMORY: Total"));
+                DescriptiveTime(Multiply_KokkosKernels(*A, *B, *C, std::string("SPGEMM_KK_MEMORY"), kk_team_work_size));
+              }
+              break;
+            // KK Algorithms (KK Dense)
+            case Experiments::KK_DENSE:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK_DENSE: Total"));
+                DescriptiveTime(Multiply_KokkosKernels(*A, *B, *C, std::string("SPGEMM_KK_DENSE"), kk_team_work_size));
+              }
+              break;
+            // KK Algorithms (KK Default)
+            case Experiments::KK_DEFAULT:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM SPGEMM_KK: Total"));
+                DescriptiveTime(Multiply_KokkosKernels(*A, *B, *C, std::string("SPGEMM_KK"), kk_team_work_size));
+              }
+              break;
+            // LTG
+            case Experiments::LTG:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM LTG: Total"));
+                using ltg_tester = LTG_Tests<SC, LO, GO, Node>;
+                DescriptiveTime(ltg_tester::Multiply_LTG(*A, *B, *C));
+              }
+              break;
+
+            default:
+              std::cerr << "Unknown experiment ID encountered: " << (int)experiment_id << std::endl;
           }
-          break;
-
-        default:
-          std::cerr << "Unknown experiment ID encountered: " << (int) experiment_id << std::endl;
-        }
-        comm->barrier();
-
-        if (printFNorm) {
-          const auto current_fnorm = C->getFrobeniusNorm ();
-          const bool fnorm_error = (std::abs(current_fnorm - prior_fnorm)/prior_fnorm) > 1.e-14;
-          if ( i == 0 ) {
-            out << kernel_name << " "
-                << "Current Fnorm: " << current_fnorm
-                << std::endl;
-          } else if (fnorm_error)  {
-            out << kernel_name << " "
-                << "Current Fnorm: " << current_fnorm << ", Difference from Prior: "
-                << std::abs(prior_fnorm - current_fnorm)
-                << std::endl;
-            fnorm_failures++;
+          comm->barrier();
+
+          if (printFNorm) {
+            const auto current_fnorm = C->getFrobeniusNorm();
+            const bool fnorm_error   = (std::abs(current_fnorm - prior_fnorm) / prior_fnorm) > 1.e-14;
+            if (i == 0) {
+              out << kernel_name << " "
+                  << "Current Fnorm: " << current_fnorm
+                  << std::endl;
+            } else if (fnorm_error) {
+              out << kernel_name << " "
+                  << "Current Fnorm: " << current_fnorm << ", Difference from Prior: "
+                  << std::abs(prior_fnorm - current_fnorm)
+                  << std::endl;
+              fnorm_failures++;
+            }
+            prior_fnorm = current_fnorm;
           }
-          prior_fnorm = current_fnorm;
-        }
-        C = Teuchos::null;
-      }// end random exp loop
+          C = Teuchos::null;
+        }  // end random exp loop
 
-      if (progressBar) simple_progress_bar (i, out, nrepeat);
-    } // end repeat
-    } // end ! my_experiments.empty()
+        if (progressBar) simple_progress_bar(i, out, nrepeat);
+      }  // end repeat
+    }    // end ! my_experiments.empty()
 
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings) {
-
       TimeMonitor::summarize(comm.ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
-      #if USE_DESCRIPTIVE_STATS == 1
+#if USE_DESCRIPTIVE_STATS == 1
 
-      if (descStats) print_desc_stats (csvFile, writeCSV, out);
-      #endif
+      if (descStats) print_desc_stats(csvFile, writeCSV, out);
+#endif
     }
 
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/scaling/MatrixLoad.hpp b/packages/muelu/test/scaling/MatrixLoad.hpp
index 11adc92946dd..cce8daf684ef 100644
--- a/packages/muelu/test/scaling/MatrixLoad.hpp
+++ b/packages/muelu/test/scaling/MatrixLoad.hpp
@@ -66,29 +66,27 @@
 
 #include <MueLu.hpp>
 
-
-
 // This is a standard Galeri-or-MatrixFile loading routine designed to be shared between the various scaling tests
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::UnderlyingLib& lib,
-                bool binaryFormat,const std::string & matrixFile, const std::string & rhsFile,
-                const std::string & rowMapFile,
-                const std::string & colMapFile,
-                const std::string & domainMapFile,
-                const std::string & rangeMapFile,
-                const std::string & coordFile,
-                const std::string & coordMapFile, const std::string &nullFile, const std::string &materialFile,
-                Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >          & map,
-                Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >      & A,
-                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> > & coordinates,
-                Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & nullspace,
-                Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & material,
-                Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & X,
-                Teuchos::RCP<Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > & B,
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> >& comm, Xpetra::UnderlyingLib& lib,
+                bool binaryFormat, const std::string& matrixFile, const std::string& rhsFile,
+                const std::string& rowMapFile,
+                const std::string& colMapFile,
+                const std::string& domainMapFile,
+                const std::string& rangeMapFile,
+                const std::string& coordFile,
+                const std::string& coordMapFile, const std::string& nullFile, const std::string& materialFile,
+                Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
+                Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
+                Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& coordinates,
+                Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& nullspace,
+                Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& material,
+                Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& X,
+                Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B,
                 const int numVectors,
-                Galeri::Xpetra::Parameters<GlobalOrdinal> & galeriParameters,  Xpetra::Parameters & xpetraParameters,
-                std::ostringstream & galeriStream) {
+                Galeri::Xpetra::Parameters<GlobalOrdinal>& galeriParameters, Xpetra::Parameters& xpetraParameters,
+                std::ostringstream& galeriStream) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -96,13 +94,12 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
   typedef Teuchos::ScalarTraits<SC> STS;
   SC zero = STS::zero(), one = STS::one();
   typedef typename STS::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  galeriStream << "========================================================\n" << xpetraParameters;
+  galeriStream << "========================================================\n"
+               << xpetraParameters;
   if (matrixFile.empty()) {
-
     galeriStream << galeriParameters;
 
     // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
@@ -120,24 +117,24 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
     // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
     // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
     if (matrixType == "Laplace1D" || matrixType == "Identity") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
     } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
                matrixType == "BigStar2D" || matrixType == "AnisotropicDiffusion" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
     } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+      map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
     }
 
     // Expand map to do multiple DOF per node for block problems
     if (matrixType == "Elasticity2D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
     if (matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
     galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                  << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -146,35 +143,33 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
 
     if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
       // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-      galeriList.set("right boundary" , "Neumann");
+      galeriList.set("right boundary", "Neumann");
       galeriList.set("bottom boundary", "Neumann");
-      galeriList.set("top boundary"   , "Neumann");
-      galeriList.set("front boundary" , "Neumann");
-      galeriList.set("back boundary"  , "Neumann");
+      galeriList.set("top boundary", "Neumann");
+      galeriList.set("front boundary", "Neumann");
+      galeriList.set("back boundary", "Neumann");
     }
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
-    A = Pr->BuildMatrix();
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+    A         = Pr->BuildMatrix();
     nullspace = Pr->BuildNullspace();
     //  The coordinates used by Galeri here might not match the coordinates that come into this function.
     //  In particular, they might correspond to different stretch factors. To fix this, we overwrite the
     //  coordinate array with those that Galeri now provides.
-    
 
-    if(!coordinates.is_null() && (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") ) {
-      Teuchos::RCP<Xpetra::MultiVector<real_type,LocalOrdinal,GlobalOrdinal,Node> > newcoordinates;
+    if (!coordinates.is_null() && (matrixType == "Elasticity2D" || matrixType == "Elasticity3D")) {
+      Teuchos::RCP<Xpetra::MultiVector<real_type, LocalOrdinal, GlobalOrdinal, Node> > newcoordinates;
       newcoordinates = Pr->BuildCoords();
 
       // Galeri makes multiple copies of coordinates to deal with
       // some issues when Ndofs != Nmeshnodes
-     
 
       for (size_t kkk = 0; kkk < coordinates->getNumVectors(); kkk++) {
-        Teuchos::ArrayRCP<real_type> old = coordinates->getDataNonConst(kkk);
-        Teuchos::ArrayRCP<real_type> newvals  = newcoordinates->getDataNonConst(kkk);
-        int numCopies = newvals.size() / old.size();
-        for (int jj=0; jj < old.size(); jj++) old[jj] = newvals[numCopies*jj];
+        Teuchos::ArrayRCP<real_type> old     = coordinates->getDataNonConst(kkk);
+        Teuchos::ArrayRCP<real_type> newvals = newcoordinates->getDataNonConst(kkk);
+        int numCopies                        = newvals.size() / old.size();
+        for (int jj = 0; jj < old.size(); jj++) old[jj] = newvals[numCopies * jj];
       }
     }
 
@@ -185,30 +180,30 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
 
   } else {
     if (!rowMapFile.empty())
-      map = Xpetra::IO<SC,LO,GO,Node>::ReadMap(rowMapFile, lib, comm);
+      map = Xpetra::IO<SC, LO, GO, Node>::ReadMap(rowMapFile, lib, comm);
     comm->barrier();
 
     {
-    RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1a - Matrix read")));
-    if (!binaryFormat && !map.is_null()) {
-      RCP<const Map> colMap    = (!colMapFile.empty()    ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(colMapFile,    lib, comm) : Teuchos::null);
-      RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
-      RCP<const Map> rangeMap  = (!rangeMapFile.empty()  ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(rangeMapFile,  lib, comm) : Teuchos::null);
-      A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
-
-    } else {
-      A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, lib, comm, binaryFormat);
-    }
-    comm->barrier();
-    } //timer scope
+      RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1a - Matrix read")));
+      if (!binaryFormat && !map.is_null()) {
+        RCP<const Map> colMap    = (!colMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(colMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> rangeMap  = (!rangeMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(rangeMapFile, lib, comm) : Teuchos::null);
+        A                        = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
+
+      } else {
+        A = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, lib, comm, binaryFormat);
+      }
+      comm->barrier();
+    }  // timer scope
 
-    //If no rowmap file has been provided and the driver is being run in parallel,
-    //create a uniformly distributed map and use it as A's row map.
+    // If no rowmap file has been provided and the driver is being run in parallel,
+    // create a uniformly distributed map and use it as A's row map.
     if (map.is_null() && comm->getSize() > 1) {
       RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1b - Matrix redistribute")));
-      if (comm->getRank()==0)
+      if (comm->getRank() == 0)
         std::cout << "No rowmap file specified, redistributing matrix using a uniformly distributed rowmap." << std::endl;
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(lib, A->getRowMap()->getGlobalNumElements(), (int) 0, comm);
+      map = Xpetra::MapFactory<LO, GO, Node>::Build(lib, A->getRowMap()->getGlobalNumElements(), (int)0, comm);
 
       RCP<Matrix> newMatrix = MatrixFactory::Build(map, 1);
       RCP<Import> importer  = ImportFactory::Build(A->getRowMap(), map);
@@ -226,18 +221,18 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
       RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1c - Read coordinates")));
       RCP<const Map> coordMap;
       if (!coordMapFile.empty())
-        coordMap = Xpetra::IO<SC,LO,GO,Node>::ReadMap(coordMapFile, lib, comm);
+        coordMap = Xpetra::IO<SC, LO, GO, Node>::ReadMap(coordMapFile, lib, comm);
       else
         coordMap = map;
-      coordinates = Xpetra::IO<real_type,LO,GO,Node>::ReadMultiVector(coordFile, coordMap);
+      coordinates = Xpetra::IO<real_type, LO, GO, Node>::ReadMultiVector(coordFile, coordMap);
       comm->barrier();
     }
 
     if (!nullFile.empty())
-      nullspace = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(nullFile, map);
+      nullspace = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(nullFile, map);
 
     if (!materialFile.empty())
-      material = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(materialFile, map);
+      material = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(materialFile, map);
   }
 
   X = MultiVectorFactory::Build(map, numVectors);
@@ -251,12 +246,12 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
 
     Teuchos::Array<real_type> norms(numVectors);
     B->norm2(norms);
-    B->scale(one/norms[0]);
+    B->scale(one / norms[0]);
 
   } else {
     // read in B
     RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1d - Read RHS")));
-    B = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(rhsFile, map);
+    B                   = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(rhsFile, map);
     comm->barrier();
   }
   galeriStream << "Galeri complete.\n========================================================" << std::endl;
@@ -271,136 +266,153 @@ void MatrixLoad(Teuchos::RCP<const Teuchos::Comm<int> > &comm,  Xpetra::Underlyi
 //     1) The first line in the file is a comment
 //     2) all file entries defining blk k occur before
 //        file entries defining blk k+1.
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 void readUserBlks(const std::string& userBlkFileName, const std::string& smootherOrCoarse, Teuchos::ParameterList& mueluList,
-                  const Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >      & A) {
-
-//     userBlkFileName          MatrixMarket file defining blocks
-//     smootherOrCoarse         Whether blocks are associated with "smoother: params" or "coarse: params"
-//     mueluList                Parameter list that is modified to reflect block informaiton found in file
-//     A                        Matrix that block smoother will be applied to
+                  const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) {
+  //     userBlkFileName          MatrixMarket file defining blocks
+  //     smootherOrCoarse         Whether blocks are associated with "smoother: params" or "coarse: params"
+  //     mueluList                Parameter list that is modified to reflect block informaiton found in file
+  //     A                        Matrix that block smoother will be applied to
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
-
-    RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-
-    TEUCHOS_TEST_FOR_EXCEPTION( (smootherOrCoarse != "smoother") &&  (smootherOrCoarse != "coarse"),
-                                 std::runtime_error, "2nd argument to readUserBlks must be either \"smoother\" or \"coarse\" and not \"" + smootherOrCoarse + "\"");
-    if ( (!userBlkFileName.empty()) && (mueluList.isSublist(smootherOrCoarse + ": params"))) {
-      bool hasSubdomSolver = mueluList.sublist(smootherOrCoarse + ": params").isSublist("subdomain solver parameters");
-      if (hasSubdomSolver) {
-        if (mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").isParameter("partitioner: type")) {
-          if (mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").get<std::string>("partitioner: type") == "user") {
-
-            FILE   *fp;
-            int    retVal;
-            int    nBlks, nRows, nnzs, ch, row, col;
-            int    procId = comm->getRank();
-            double val;
 
-            /* read block information from file only to count the # of rows */
-            /* we own within each block. We'll reopen the file later to fill*/
-            /* the Teuchos::Array for "partitioner: global ID parts"        */
-
-            fp = fopen( &userBlkFileName[0],"r");
-            TEUCHOS_TEST_FOR_EXCEPTION(fp == NULL, std::runtime_error, userBlkFileName + " file not found");
-
-            while ( (ch= getc(fp) != '\n'))  ;  //read first line
+  RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+
+  TEUCHOS_TEST_FOR_EXCEPTION((smootherOrCoarse != "smoother") && (smootherOrCoarse != "coarse"),
+                             std::runtime_error, "2nd argument to readUserBlks must be either \"smoother\" or \"coarse\" and not \"" + smootherOrCoarse + "\"");
+  if ((!userBlkFileName.empty()) && (mueluList.isSublist(smootherOrCoarse + ": params"))) {
+    bool hasSubdomSolver = mueluList.sublist(smootherOrCoarse + ": params").isSublist("subdomain solver parameters");
+    if (hasSubdomSolver) {
+      if (mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").isParameter("partitioner: type")) {
+        if (mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").get<std::string>("partitioner: type") == "user") {
+          FILE* fp;
+          int retVal;
+          int nBlks, nRows, nnzs, ch, row, col;
+          int procId = comm->getRank();
+          double val;
+
+          /* read block information from file only to count the # of rows */
+          /* we own within each block. We'll reopen the file later to fill*/
+          /* the Teuchos::Array for "partitioner: global ID parts"        */
+
+          fp = fopen(&userBlkFileName[0], "r");
+          TEUCHOS_TEST_FOR_EXCEPTION(fp == NULL, std::runtime_error, userBlkFileName + " file not found");
+
+          while ((ch = getc(fp) != '\n'))
+            ;  // read first line
+
+          retVal = fscanf(fp, "%d %d %d\n", &nBlks, &nRows, &nnzs);
+          TEUCHOS_TEST_FOR_EXCEPTION(retVal != 3,
+                                     std::runtime_error, "unable to parse nBlks, nRows, nnzs in user file " + userBlkFileName);
+
+          TEUCHOS_TEST_FOR_EXCEPTION(nRows != (int)A->getRowMap()->getGlobalNumElements(),
+                                     std::runtime_error, "number of global rows in " + userBlkFileName + " does not match those in A");
+
+          Teuchos::ArrayRCP<int> myOwnedRowsPerBlock(nBlks, 0);
+
+          for (int i = 0; i < nnzs; i++) {
+            retVal = fscanf(fp, "%d %d %lf", &row, &col, &val);
+            row--;
+            col--;
+            if (A->getRowMap()->getLocalElement((GlobalOrdinal)col) != Teuchos::OrdinalTraits<LocalOrdinal>::invalid()) (myOwnedRowsPerBlock[row])++;
+          }
+          fclose(fp);
 
-            retVal = fscanf(fp,"%d %d %d\n",&nBlks, &nRows, &nnzs);
-            TEUCHOS_TEST_FOR_EXCEPTION(retVal != 3,
-                std::runtime_error,"unable to parse nBlks, nRows, nnzs in user file " + userBlkFileName);
+          // Assign ownership of each block to one mpirank corresponding
+          // to which mpirank owns the most rows within a block. When
+          // ties occur, highest procId wins.
 
-            TEUCHOS_TEST_FOR_EXCEPTION(nRows != (int) A->getRowMap()->getGlobalNumElements(),
-                std::runtime_error,"number of global rows in " + userBlkFileName + " does not match those in A");
+          // first, set to procId+1 if we own the largest # of rows
+          // in block.  Otherwise, set to 0.
 
-            Teuchos::ArrayRCP<int> myOwnedRowsPerBlock(nBlks, 0);
+          Teuchos::ArrayRCP<int> maxOwnedRowsPerBlock(nBlks, 0);
+          Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, (LocalOrdinal)nBlks, myOwnedRowsPerBlock.getRawPtr(), maxOwnedRowsPerBlock.getRawPtr());
 
-            for (int i = 0; i < nnzs; i++) {
-              retVal = fscanf(fp,"%d %d %lf", &row, &col, &val); row--; col--;
-              if (A->getRowMap()->getLocalElement( (GlobalOrdinal) col) != Teuchos::OrdinalTraits<LocalOrdinal>::invalid()) (myOwnedRowsPerBlock[row])++;
-            }
-            fclose(fp);
-
-            // Assign ownership of each block to one mpirank corresponding
-            // to which mpirank owns the most rows within a block. When
-            // ties occur, highest procId wins.
-
-            // first, set to procId+1 if we own the largest # of rows
-            // in block.  Otherwise, set to 0.
-
-            Teuchos::ArrayRCP<int> maxOwnedRowsPerBlock(nBlks, 0);
-            Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, (LocalOrdinal) nBlks, myOwnedRowsPerBlock.getRawPtr(),maxOwnedRowsPerBlock.getRawPtr());
-      
-            Teuchos::ArrayRCP<int> haveMaxOwned(nBlks, 0);
-            for (int i = 0; i < nBlks; i++) {
-              haveMaxOwned[i] = 0;
-              if ((myOwnedRowsPerBlock[i] == maxOwnedRowsPerBlock[i]) && (myOwnedRowsPerBlock[i] >0)) haveMaxOwned[i] = procId+1;
-            }
-            Teuchos::ArrayRCP<int> maxProcIdPlusOne(nBlks, 0);
-            Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, (LocalOrdinal) nBlks, haveMaxOwned.getRawPtr(), maxProcIdPlusOne.getRawPtr());
-
-            Teuchos::ArrayRCP<bool> ownBlk(nBlks, false);
-            int nOwned = 0;
-            int maxDofsInAnyBlock = 0;
-            for (int i = 0; i < nBlks; i++) {
-              TEUCHOS_TEST_FOR_EXCEPTION(maxProcIdPlusOne[i] == 0, std::runtime_error,
-                "some blocks are not owned by any processor?");
-              if (maxProcIdPlusOne[i] == procId+1) { ownBlk[i] = true;  nOwned++; maxDofsInAnyBlock += myOwnedRowsPerBlock[i];}
+          Teuchos::ArrayRCP<int> haveMaxOwned(nBlks, 0);
+          for (int i = 0; i < nBlks; i++) {
+            haveMaxOwned[i] = 0;
+            if ((myOwnedRowsPerBlock[i] == maxOwnedRowsPerBlock[i]) && (myOwnedRowsPerBlock[i] > 0)) haveMaxOwned[i] = procId + 1;
+          }
+          Teuchos::ArrayRCP<int> maxProcIdPlusOne(nBlks, 0);
+          Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, (LocalOrdinal)nBlks, haveMaxOwned.getRawPtr(), maxProcIdPlusOne.getRawPtr());
+
+          Teuchos::ArrayRCP<bool> ownBlk(nBlks, false);
+          int nOwned            = 0;
+          int maxDofsInAnyBlock = 0;
+          for (int i = 0; i < nBlks; i++) {
+            TEUCHOS_TEST_FOR_EXCEPTION(maxProcIdPlusOne[i] == 0, std::runtime_error,
+                                       "some blocks are not owned by any processor?");
+            if (maxProcIdPlusOne[i] == procId + 1) {
+              ownBlk[i] = true;
+              nOwned++;
+              maxDofsInAnyBlock += myOwnedRowsPerBlock[i];
             }
-            maxDofsInAnyBlock *= 2; // We only have an estimate as to the largest number of dofs in any block (as some dofs
-            maxDofsInAnyBlock++;    // might reside on other processors. So we multiple our estimate by 2 hoping to be large
-                                    // enough. Later, we check to see if things are not large enough and re-allocate space
-                                    // in this case.
-
-            Teuchos::Array<Teuchos::ArrayRCP<GlobalOrdinal> > blockLists(nOwned,Teuchos::null);
+          }
+          maxDofsInAnyBlock *= 2;  // We only have an estimate as to the largest number of dofs in any block (as some dofs
+          maxDofsInAnyBlock++;     // might reside on other processors. So we multiple our estimate by 2 hoping to be large
+                                   // enough. Later, we check to see if things are not large enough and re-allocate space
+                                   // in this case.
 
-            Teuchos::Array<int> buffer(maxDofsInAnyBlock,0);
+          Teuchos::Array<Teuchos::ArrayRCP<GlobalOrdinal> > blockLists(nOwned, Teuchos::null);
 
-            int curRow, currentOwnedRow = 0;
+          Teuchos::Array<int> buffer(maxDofsInAnyBlock, 0);
 
-            // reopen userBlkFilename and record block information in blockLists
+          int curRow, currentOwnedRow = 0;
 
-            fp = fopen( &userBlkFileName[0],"r");
-            TEUCHOS_TEST_FOR_EXCEPTION(fp == NULL, std::runtime_error, userBlkFileName + " file not found");
+          // reopen userBlkFilename and record block information in blockLists
 
-            while ( (ch= getc(fp) != '\n'))  ;
+          fp = fopen(&userBlkFileName[0], "r");
+          TEUCHOS_TEST_FOR_EXCEPTION(fp == NULL, std::runtime_error, userBlkFileName + " file not found");
 
-            retVal = fscanf(fp,"%d %d %d\n",&nBlks, &nRows, &nnzs);
-            TEUCHOS_TEST_FOR_EXCEPTION(retVal != 3,
-                std::runtime_error,"unable to parse nBlks, nRows, nnzs in user file " + userBlkFileName);
+          while ((ch = getc(fp) != '\n'))
+            ;
 
-            retVal = fscanf(fp,"%d %d %lf", &row, &col, &val); row=row-1; col=col-1;
+          retVal = fscanf(fp, "%d %d %d\n", &nBlks, &nRows, &nnzs);
+          TEUCHOS_TEST_FOR_EXCEPTION(retVal != 3,
+                                     std::runtime_error, "unable to parse nBlks, nRows, nnzs in user file " + userBlkFileName);
 
-            int jj = 1;
-            while (jj <= nnzs ) {
-              if (ownBlk[row] == true) {
+          retVal = fscanf(fp, "%d %d %lf", &row, &col, &val);
+          row    = row - 1;
+          col    = col - 1;
 
-                int i = 0;
-                curRow = row;
-                while ((row == curRow) && (jj <= nnzs)){
-                  if (i == maxDofsInAnyBlock) {
-                    maxDofsInAnyBlock *= 2;
-                    buffer.resize(maxDofsInAnyBlock);
-                  }
-                  buffer[i] = col; i++;
-                  retVal = fscanf(fp,"%d %d %lf", &row, &col, &val); jj++; row=row-1; col=col-1;
+          int jj = 1;
+          while (jj <= nnzs) {
+            if (ownBlk[row] == true) {
+              int i  = 0;
+              curRow = row;
+              while ((row == curRow) && (jj <= nnzs)) {
+                if (i == maxDofsInAnyBlock) {
+                  maxDofsInAnyBlock *= 2;
+                  buffer.resize(maxDofsInAnyBlock);
                 }
-                blockLists[currentOwnedRow] = Teuchos::arcp<GlobalOrdinal>(i);
-                for (int k = 0; k < i; k++) { blockLists[currentOwnedRow][k] = (GlobalOrdinal) buffer[k];   }
-                currentOwnedRow++;
+                buffer[i] = col;
+                i++;
+                retVal = fscanf(fp, "%d %d %lf", &row, &col, &val);
+                jj++;
+                row = row - 1;
+                col = col - 1;
+              }
+              blockLists[currentOwnedRow] = Teuchos::arcp<GlobalOrdinal>(i);
+              for (int k = 0; k < i; k++) {
+                blockLists[currentOwnedRow][k] = (GlobalOrdinal)buffer[k];
               }
-              else { retVal = fscanf(fp,"%d %d %lf", &row, &col, &val); jj++;  row=row-1; col=col-1; }
+              currentOwnedRow++;
+            } else {
+              retVal = fscanf(fp, "%d %d %lf", &row, &col, &val);
+              jj++;
+              row = row - 1;
+              col = col - 1;
             }
-            fclose(fp);
-            mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").set< Teuchos::Array<Teuchos::ArrayRCP<GlobalOrdinal> > >("partitioner: global ID parts", blockLists);
           }
+          fclose(fp);
+          mueluList.sublist(smootherOrCoarse + ": params").sublist("subdomain solver parameters").set<Teuchos::Array<Teuchos::ArrayRCP<GlobalOrdinal> > >("partitioner: global ID parts", blockLists);
         }
       }
     }
+  }
 }
 
 #endif
diff --git a/packages/muelu/test/scaling/MatrixMatrixMultiply.cpp b/packages/muelu/test/scaling/MatrixMatrixMultiply.cpp
index 99d3d67b719c..83d0ed561bad 100644
--- a/packages/muelu/test/scaling/MatrixMatrixMultiply.cpp
+++ b/packages/muelu/test/scaling/MatrixMatrixMultiply.cpp
@@ -67,259 +67,253 @@
 
 namespace MueLuTests {
 
-  // generate random matrix
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > generateRandomMatrix(LocalOrdinal const &minEntriesPerRow, LocalOrdinal const &maxEntriesPerRow,
-											     Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > &rowMap,
-											     Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > &domainMap);
-
-  // generate random map
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > generateRandomContiguousMap(size_t const minRowsPerProc, size_t const maxRowsPerProc,
-      Teuchos::RCP<const Teuchos::Comm<int> > const &comm, Xpetra::UnderlyingLib const lib);
-
-  // generate "random" whole number in interval [a,b]
-  template <class T>
-    size_t generateRandomNumber(T a, T b);
-
-  // generate distinct seeds for each process.  The seeds are different each run, but print to screen in case the run needs reproducibility.
-  unsigned int generateSeed(Teuchos::Comm<int> const &comm, const double initSeed=-1);
-
+// generate random matrix
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > generateRandomMatrix(LocalOrdinal const &minEntriesPerRow, LocalOrdinal const &maxEntriesPerRow,
+                                                                                              Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &rowMap,
+                                                                                              Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &domainMap);
 
+// generate random map
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > generateRandomContiguousMap(size_t const minRowsPerProc, size_t const maxRowsPerProc,
+                                                                                                Teuchos::RCP<const Teuchos::Comm<int> > const &comm, Xpetra::UnderlyingLib const lib);
 
+// generate "random" whole number in interval [a,b]
+template <class T>
+size_t generateRandomNumber(T a, T b);
 
+// generate distinct seeds for each process.  The seeds are different each run, but print to screen in case the run needs reproducibility.
+unsigned int generateSeed(Teuchos::Comm<int> const &comm, const double initSeed = -1);
 
-}
+}  // namespace MueLuTests
 
 //- -- --------------------------------------------------------
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
   using Teuchos::RCP;
   using Teuchos::rcp;
+  using Teuchos::StackedTimer;
   using Teuchos::Time;
   using Teuchos::TimeMonitor;
-  using Teuchos::StackedTimer;
   using namespace MueLuTests;
 
-  RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-  
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
 #include <MueLu_UseShortNames.hpp>
-    Xpetra::Parameters            xpetraParameters(clp);
-    int  optNmults         = 1;     clp.setOption("nmults",               &optNmults,           "number of matrix matrix multiplies to perform");
-
-    GO optMaxRowsPerProc   = 1000;  clp.setOption("maxrows",              &optMaxRowsPerProc,   "maximum number of rows per processor");
-    GO optMinRowsPerProc   = 500;   clp.setOption("minrows",              &optMinRowsPerProc,   "minimum number of rows per processor");
-    LO optMaxEntriesPerRow = 50;    clp.setOption("maxnnz",               &optMaxEntriesPerRow, "maximum number of nonzeros per row");
-    LO optMinEntriesPerRow = 25;    clp.setOption("minnnz",               &optMinEntriesPerRow, "minimum number of nonzeros per row");
-
-    bool optDumpMatrices   = false; clp.setOption("dump", "nodump",       &optDumpMatrices,     "write matrix to file");
-    bool optTimings        = false; clp.setOption("timings", "notimings", &optTimings,          "print timings to screen");
-    double optSeed         = -1;    clp.setOption("seed",                 &optSeed,             "random number seed (cast to unsigned int)");
-
-    std::ostringstream description;
-    description << "This is a parallel-only test of the sparse matrix matrix multiply.  Two\n"
-      << "\"random\" matrices are generated. They are most likely rectangular.  Each\n"
-      << "has random contiguous domain and range maps. Each row has a random number of\n"
-      << "entries with random global column IDs. The matrix values are identically 1,\n"
-      << "and success is declared if the multiply finishes.";
-    clp.setDocString(description.str().c_str());
-
-    switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  Xpetra::Parameters xpetraParameters(clp);
+  int optNmults = 1;
+  clp.setOption("nmults", &optNmults, "number of matrix matrix multiplies to perform");
+
+  GO optMaxRowsPerProc = 1000;
+  clp.setOption("maxrows", &optMaxRowsPerProc, "maximum number of rows per processor");
+  GO optMinRowsPerProc = 500;
+  clp.setOption("minrows", &optMinRowsPerProc, "minimum number of rows per processor");
+  LO optMaxEntriesPerRow = 50;
+  clp.setOption("maxnnz", &optMaxEntriesPerRow, "maximum number of nonzeros per row");
+  LO optMinEntriesPerRow = 25;
+  clp.setOption("minnnz", &optMinEntriesPerRow, "minimum number of nonzeros per row");
+
+  bool optDumpMatrices = false;
+  clp.setOption("dump", "nodump", &optDumpMatrices, "write matrix to file");
+  bool optTimings = false;
+  clp.setOption("timings", "notimings", &optTimings, "print timings to screen");
+  double optSeed = -1;
+  clp.setOption("seed", &optSeed, "random number seed (cast to unsigned int)");
+
+  std::ostringstream description;
+  description << "This is a parallel-only test of the sparse matrix matrix multiply.  Two\n"
+              << "\"random\" matrices are generated. They are most likely rectangular.  Each\n"
+              << "has random contiguous domain and range maps. Each row has a random number of\n"
+              << "entries with random global column IDs. The matrix values are identically 1,\n"
+              << "and success is declared if the multiply finishes.";
+  clp.setDocString(description.str().c_str());
+
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
-    }
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
+  }
 
+  unsigned int seed = generateSeed(*comm, optSeed);
+  Teuchos::ScalarTraits<SC>::seedrandom(seed);
 
-    
-    unsigned int seed = generateSeed(*comm, optSeed);
-    Teuchos::ScalarTraits<SC>::seedrandom(seed);
-    
-    RCP<StackedTimer> timer = rcp(new StackedTimer("MatrixMatrix Multiply: Total"));
-    TimeMonitor::setStackedTimer(timer);
-
-    for (int jj=0; jj<optNmults; ++jj) {
-      
-      RCP<Matrix> A;
-      RCP<Matrix> B;
-      {
-	TimeMonitor tm(*TimeMonitor::getNewTimer("MatrixMatrixMultiplyTest: 1 - Matrix creation"));
-	
-	size_t maxRowsPerProc   = optMaxRowsPerProc;
-	size_t minRowsPerProc   = optMinRowsPerProc;
-	if (minRowsPerProc > maxRowsPerProc) minRowsPerProc=maxRowsPerProc;
-	
-	// Create row map.  This will also be used as the range map.
-	RCP<const Map> rowMapForA = generateRandomContiguousMap<LO,GO,NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
-	// Create domain map for A. This will also be the row map for B.
-	RCP<const Map> domainMapForA = generateRandomContiguousMap<LO,GO,NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
-	// Create domain map for B.
-	RCP<const Map> domainMapForB = generateRandomContiguousMap<LO,GO,NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
-	
-	A = generateRandomMatrix<SC,LO,GO,NO>(optMinEntriesPerRow, optMaxEntriesPerRow, rowMapForA, domainMapForA);
-	B = generateRandomMatrix<SC,LO,GO,NO>(optMinEntriesPerRow, optMaxEntriesPerRow, domainMapForA, domainMapForB);
-	
-	if (comm->getRank() == 0) {
-	  std::cout << "case " << jj << " of " << optNmults-1 << " : "
-		    << "seed = " << seed
-		    << ", minEntriesPerRow = " << optMinEntriesPerRow
-		    << ", maxEntriesPerRow = " << optMaxEntriesPerRow
-		    << std::endl
-		    << "  A stats:"
-		    << " #rows = " << rowMapForA->getGlobalNumElements()
-		    << " #cols = " << domainMapForA->getGlobalNumElements()
-		    << ", nnz = " << A->getGlobalNumEntries()
-		    << std::endl
-		    << "  B stats:"
-		    << " #rows = " << domainMapForA->getGlobalNumElements()
-		    << " #cols = " << domainMapForB->getGlobalNumElements()
-		    << ", nnz = " << B->getGlobalNumEntries()
-		    << std::endl;
-	}
-	
-	if (optDumpMatrices) {
-	  std::string fileName="checkA.mm";
-	  Xpetra::IO<SC,LO,GO,Node>::Write( fileName,*A);
-	  fileName="checkB.mm";
-	  Xpetra::IO<SC,LO,GO,Node>::Write( fileName,*B);
-	}
-	
-      }  //scope for timing matrix creation
-
-      {
-	TimeMonitor tm(*TimeMonitor::getNewTimer("MatrixMatrixMultiplyTest: 2 - Multiply"));
-	
-	RCP<Matrix> AB;
-	RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
-	AB = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *B, false, *fos);
-	//if (optDumpMatrices) {
-	//  std::string fileName="checkAB.mm";
-	//  Utils::Write( fileName,*AB);
-	//}
-	
-      } //scope for multiply
-
-      seed = generateSeed(*comm);
-      
-    } //for (int jj=0; jj<optNmults; ++jj)
-        
-    timer->stopBaseTimer();
-    StackedTimer::OutputOptions options;
-    options.print_warnings = false;
-    timer->report(std::cout, comm, options);
-
-    auto xmlOut = timer->reportWatchrXML(std::string("MueLu MatrixMatrix Multiply ") + std::to_string(comm->getSize()) + " ranks", comm);
-    if(xmlOut.length())
-      std::cout << "\nAlso created Watchr performance report " << xmlOut << '\n';
-
-    if (comm->getRank() == 0)
-      std::cout << "End Result: TEST PASSED";
-
-    return 0;
-}
- //main_
+  RCP<StackedTimer> timer = rcp(new StackedTimer("MatrixMatrix Multiply: Total"));
+  TimeMonitor::setStackedTimer(timer);
 
-namespace MueLuTests {
+  for (int jj = 0; jj < optNmults; ++jj) {
+    RCP<Matrix> A;
+    RCP<Matrix> B;
+    {
+      TimeMonitor tm(*TimeMonitor::getNewTimer("MatrixMatrixMultiplyTest: 1 - Matrix creation"));
+
+      size_t maxRowsPerProc = optMaxRowsPerProc;
+      size_t minRowsPerProc = optMinRowsPerProc;
+      if (minRowsPerProc > maxRowsPerProc) minRowsPerProc = maxRowsPerProc;
+
+      // Create row map.  This will also be used as the range map.
+      RCP<const Map> rowMapForA = generateRandomContiguousMap<LO, GO, NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
+      // Create domain map for A. This will also be the row map for B.
+      RCP<const Map> domainMapForA = generateRandomContiguousMap<LO, GO, NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
+      // Create domain map for B.
+      RCP<const Map> domainMapForB = generateRandomContiguousMap<LO, GO, NO>(minRowsPerProc, maxRowsPerProc, comm, xpetraParameters.GetLib());
+
+      A = generateRandomMatrix<SC, LO, GO, NO>(optMinEntriesPerRow, optMaxEntriesPerRow, rowMapForA, domainMapForA);
+      B = generateRandomMatrix<SC, LO, GO, NO>(optMinEntriesPerRow, optMaxEntriesPerRow, domainMapForA, domainMapForB);
+
+      if (comm->getRank() == 0) {
+        std::cout << "case " << jj << " of " << optNmults - 1 << " : "
+                  << "seed = " << seed
+                  << ", minEntriesPerRow = " << optMinEntriesPerRow
+                  << ", maxEntriesPerRow = " << optMaxEntriesPerRow
+                  << std::endl
+                  << "  A stats:"
+                  << " #rows = " << rowMapForA->getGlobalNumElements()
+                  << " #cols = " << domainMapForA->getGlobalNumElements()
+                  << ", nnz = " << A->getGlobalNumEntries()
+                  << std::endl
+                  << "  B stats:"
+                  << " #rows = " << domainMapForA->getGlobalNumElements()
+                  << " #cols = " << domainMapForB->getGlobalNumElements()
+                  << ", nnz = " << B->getGlobalNumEntries()
+                  << std::endl;
+      }
+
+      if (optDumpMatrices) {
+        std::string fileName = "checkA.mm";
+        Xpetra::IO<SC, LO, GO, Node>::Write(fileName, *A);
+        fileName = "checkB.mm";
+        Xpetra::IO<SC, LO, GO, Node>::Write(fileName, *B);
+      }
 
+    }  // scope for timing matrix creation
 
-  template <class T>
-    size_t generateRandomNumber(T a, T b)
     {
-      double rv = (Teuchos::ScalarTraits<double>::random()+1)*0.5; //shift "random" value from interval [-1,1] to [0,1]
-      size_t numMyElements = Teuchos::as<size_t>(a + rv * (b - a));
-      return numMyElements;
-    }
+      TimeMonitor tm(*TimeMonitor::getNewTimer("MatrixMatrixMultiplyTest: 2 - Multiply"));
 
-  //- -- --------------------------------------------------------
+      RCP<Matrix> AB;
+      RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(rcp(new Teuchos::oblackholestream()));
+      AB                             = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*A, false, *B, false, *fos);
+      // if (optDumpMatrices) {
+      //   std::string fileName="checkAB.mm";
+      //   Utils::Write( fileName,*AB);
+      // }
 
-  unsigned int generateSeed(const Teuchos::Comm<int>& comm, const double initSeed)
-  {
-    timeval t1;
-    gettimeofday(&t1, NULL);
+    }  // scope for multiply
 
-    unsigned int seed;
-    if (initSeed > -1) seed = Teuchos::as<unsigned int>(initSeed);
-    else               seed = t1.tv_usec * t1.tv_sec;
+    seed = generateSeed(*comm);
 
-    const Teuchos::MpiComm<int> * mpiComm = dynamic_cast<const Teuchos::MpiComm<int> *>(&comm);
-    TEUCHOS_TEST_FOR_EXCEPTION(mpiComm == 0, MueLu::Exceptions::RuntimeError, "Cast to MpiComm failed");
+  }  // for (int jj=0; jj<optNmults; ++jj)
 
-    // use variant of proc 0's seed so we can always reproduce the results
-    MPI_Bcast((void*)&seed, 1, MPI_UNSIGNED, 0, *(mpiComm->getRawMpiComm()));
-    seed = seed * (1+comm.getRank());
+  timer->stopBaseTimer();
+  StackedTimer::OutputOptions options;
+  options.print_warnings = false;
+  timer->report(std::cout, comm, options);
 
-    return seed;
-  }
+  auto xmlOut = timer->reportWatchrXML(std::string("MueLu MatrixMatrix Multiply ") + std::to_string(comm->getSize()) + " ranks", comm);
+  if (xmlOut.length())
+    std::cout << "\nAlso created Watchr performance report " << xmlOut << '\n';
 
-  //- -- --------------------------------------------------------
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > generateRandomContiguousMap(size_t const minRowsPerProc, size_t const maxRowsPerProc, Teuchos::RCP<const Teuchos::Comm<int> > const &comm, Xpetra::UnderlyingLib const lib)
-  {
-    size_t numMyElements = generateRandomNumber(minRowsPerProc,maxRowsPerProc);
-    Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::createContigMap(lib,
-        Teuchos::OrdinalTraits<size_t>::invalid(),
-        numMyElements,
-        comm);
-    return map;
-  }
+  if (comm->getRank() == 0)
+    std::cout << "End Result: TEST PASSED";
 
-  //- -- --------------------------------------------------------
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > generateRandomMatrix(LocalOrdinal const &minEntriesPerRow, LocalOrdinal const &maxEntriesPerRow,
-											     Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > &rowMap,
-											     Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > &domainMap)
-  {
-    using Teuchos::Array;
-    using Teuchos::ArrayView;
-    using Teuchos::RCP;
-#include <MueLu_UseShortNames.hpp>
+  return 0;
+}
+// main_
 
-    size_t numLocalRowsInA = rowMap->getLocalNumElements();
-    // Now allocate random number of entries per row for each processor, between minEntriesPerRow and maxEntriesPerRow
-    Teuchos::ArrayRCP<size_t> eprData(numLocalRowsInA);
-    for (Teuchos::ArrayRCP<size_t>::iterator i=eprData.begin(); i!=eprData.end(); ++i) {
-      *i = generateRandomNumber(minEntriesPerRow,maxEntriesPerRow);
-    }
+namespace MueLuTests {
 
-    // Populate CrsMatrix with random nnz per row and with random column locations.
-    ArrayView<const GlobalOrdinal> myGlobalElements = rowMap->getLocalElementList();
-    GO numGlobalRows = rowMap->getGlobalNumElements();
-
-    LO realMaxEntriesPerRow = maxEntriesPerRow;
-    if (maxEntriesPerRow > numGlobalRows)
-      realMaxEntriesPerRow = numGlobalRows;
-    RCP<Matrix> A = Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node>(rowMap, eprData));
-
-    Array<Scalar> vals(realMaxEntriesPerRow);
-    //stick in ones for values
-    for (LO j=0; j< realMaxEntriesPerRow; ++j) vals[j] = Teuchos::ScalarTraits<SC>::one();
-    Array<GO> cols(realMaxEntriesPerRow);
-    for (size_t i = 0; i < numLocalRowsInA; ++i) {
-      ArrayView<SC> av(&vals[0],eprData[i]);
-      ArrayView<GO> iv(&cols[0],eprData[i]);
-      for (size_t j=0; j< eprData[i]; ++j) {
-        // we assume there are no gaps in the indices
-        cols[j] = Teuchos::as<GO>(generateRandomNumber(Teuchos::as<GO>(0),domainMap->getMaxAllGlobalIndex()));
-      }
-      A->insertGlobalValues(myGlobalElements[i], iv, av);
-    }
+template <class T>
+size_t generateRandomNumber(T a, T b) {
+  double rv            = (Teuchos::ScalarTraits<double>::random() + 1) * 0.5;  // shift "random" value from interval [-1,1] to [0,1]
+  size_t numMyElements = Teuchos::as<size_t>(a + rv * (b - a));
+  return numMyElements;
+}
+
+//- -- --------------------------------------------------------
+
+unsigned int generateSeed(const Teuchos::Comm<int> &comm, const double initSeed) {
+  timeval t1;
+  gettimeofday(&t1, NULL);
+
+  unsigned int seed;
+  if (initSeed > -1)
+    seed = Teuchos::as<unsigned int>(initSeed);
+  else
+    seed = t1.tv_usec * t1.tv_sec;
+
+  const Teuchos::MpiComm<int> *mpiComm = dynamic_cast<const Teuchos::MpiComm<int> *>(&comm);
+  TEUCHOS_TEST_FOR_EXCEPTION(mpiComm == 0, MueLu::Exceptions::RuntimeError, "Cast to MpiComm failed");
+
+  // use variant of proc 0's seed so we can always reproduce the results
+  MPI_Bcast((void *)&seed, 1, MPI_UNSIGNED, 0, *(mpiComm->getRawMpiComm()));
+  seed = seed * (1 + comm.getRank());
+
+  return seed;
+}
+
+//- -- --------------------------------------------------------
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > generateRandomContiguousMap(size_t const minRowsPerProc, size_t const maxRowsPerProc, Teuchos::RCP<const Teuchos::Comm<int> > const &comm, Xpetra::UnderlyingLib const lib) {
+  size_t numMyElements                                                    = generateRandomNumber(minRowsPerProc, maxRowsPerProc);
+  Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::createContigMap(lib,
+                                                                                                                                                   Teuchos::OrdinalTraits<size_t>::invalid(),
+                                                                                                                                                   numMyElements,
+                                                                                                                                                   comm);
+  return map;
+}
+
+//- -- --------------------------------------------------------
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > generateRandomMatrix(LocalOrdinal const &minEntriesPerRow, LocalOrdinal const &maxEntriesPerRow,
+                                                                                              Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &rowMap,
+                                                                                              Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &domainMap) {
+  using Teuchos::Array;
+  using Teuchos::ArrayView;
+  using Teuchos::RCP;
+#include <MueLu_UseShortNames.hpp>
 
-    A->fillComplete(domainMap,rowMap);
+  size_t numLocalRowsInA = rowMap->getLocalNumElements();
+  // Now allocate random number of entries per row for each processor, between minEntriesPerRow and maxEntriesPerRow
+  Teuchos::ArrayRCP<size_t> eprData(numLocalRowsInA);
+  for (Teuchos::ArrayRCP<size_t>::iterator i = eprData.begin(); i != eprData.end(); ++i) {
+    *i = generateRandomNumber(minEntriesPerRow, maxEntriesPerRow);
+  }
 
-    return A;
+  // Populate CrsMatrix with random nnz per row and with random column locations.
+  ArrayView<const GlobalOrdinal> myGlobalElements = rowMap->getLocalElementList();
+  GO numGlobalRows                                = rowMap->getGlobalNumElements();
+
+  LO realMaxEntriesPerRow = maxEntriesPerRow;
+  if (maxEntriesPerRow > numGlobalRows)
+    realMaxEntriesPerRow = numGlobalRows;
+  RCP<Matrix> A = Teuchos::rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(rowMap, eprData));
+
+  Array<Scalar> vals(realMaxEntriesPerRow);
+  // stick in ones for values
+  for (LO j = 0; j < realMaxEntriesPerRow; ++j) vals[j] = Teuchos::ScalarTraits<SC>::one();
+  Array<GO> cols(realMaxEntriesPerRow);
+  for (size_t i = 0; i < numLocalRowsInA; ++i) {
+    ArrayView<SC> av(&vals[0], eprData[i]);
+    ArrayView<GO> iv(&cols[0], eprData[i]);
+    for (size_t j = 0; j < eprData[i]; ++j) {
+      // we assume there are no gaps in the indices
+      cols[j] = Teuchos::as<GO>(generateRandomNumber(Teuchos::as<GO>(0), domainMap->getMaxAllGlobalIndex()));
+    }
+    A->insertGlobalValues(myGlobalElements[i], iv, av);
   }
 
+  A->fillComplete(domainMap, rowMap);
+
+  return A;
 }
 
+}  // namespace MueLuTests
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
-
diff --git a/packages/muelu/test/scaling/MatvecKernelDriver.cpp b/packages/muelu/test/scaling/MatvecKernelDriver.cpp
index a2b5c07da4fc..bd43960d116a 100644
--- a/packages/muelu/test/scaling/MatvecKernelDriver.cpp
+++ b/packages/muelu/test/scaling/MatvecKernelDriver.cpp
@@ -67,7 +67,6 @@
 #include "MueLu_PerfModelReporter.hpp"
 #include <MatrixLoad.hpp>
 
-
 #include "Xpetra_TpetraMultiVector.hpp"
 #include "Xpetra_TpetraImport.hpp"
 #include "Tpetra_CrsMatrix.hpp"
@@ -80,18 +79,17 @@
 #include "cusparse.h"
 #endif
 
-#if defined (HAVE_MUELU_HYPRE)
+#if defined(HAVE_MUELU_HYPRE)
 #include "HYPRE_IJ_mv.h"
 #include "HYPRE_parcsr_ls.h"
 #include "HYPRE_parcsr_mv.h"
 #include "HYPRE.h"
 #endif
 
-#if defined (HAVE_MUELU_PETSC)
+#if defined(HAVE_MUELU_PETSC)
 #include "petscksp.h"
 #endif
 
-
 Teuchos::RCP<Teuchos::StackedTimer> stacked_timer;
 Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor;
 
@@ -99,49 +97,44 @@ Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor;
 // Support Routines
 // =========================================================================
 
-
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view_n(const int n, const View1 x1, View2 x2) {
-  Kokkos::parallel_for(n,KOKKOS_LAMBDA(const size_t i) {
-      x2[i] = x1[i];
-    });
+  Kokkos::parallel_for(
+      n, KOKKOS_LAMBDA(const size_t i) {
+        x2[i] = x1[i];
+      });
 }
 
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view(const View1 x1, View2 x2) {
-  copy_view_n(x1.extent(0),x1,x2);
+  copy_view_n(x1.extent(0), x1, x2);
 }
 
-
-
 template <class V1, class V2>
 void print_crs_graph(std::string name, const V1 rowptr, const V2 colind) {
-  printf("%s rowptr[%d] = ",name.c_str(),rowptr.extent(0));
-  for(size_t i=0; i<rowptr.extent(0); i++)
-    printf(" %d",(int)rowptr[i]);
-  printf("\n%s colind[%d] = ",name.c_str(),colind.extent(0));
-  for(size_t i=0; i<colind.extent(0); i++)
-    printf(" %d",(int)colind[i]);
+  printf("%s rowptr[%d] = ", name.c_str(), rowptr.extent(0));
+  for (size_t i = 0; i < rowptr.extent(0); i++)
+    printf(" %d", (int)rowptr[i]);
+  printf("\n%s colind[%d] = ", name.c_str(), colind.extent(0));
+  for (size_t i = 0; i < colind.extent(0); i++)
+    printf(" %d", (int)colind[i]);
   printf("\n");
 }
 
-  
-
 //==============================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MakeContiguousMaps(const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> & Matrix,
-                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & ContiguousRowMap,
-                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & ContiguousColumnMap,
-                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & ContiguousDomainMap) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MakeContiguousMaps(const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Matrix,
+                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& ContiguousRowMap,
+                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& ContiguousColumnMap,
+                        Teuchos::RCP<const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& ContiguousDomainMap) {
   using Teuchos::rcp;
   using Teuchos::RCP;
   using LO             = LocalOrdinal;
   using GO             = GlobalOrdinal;
   using NO             = Node;
-  using map_type       = Tpetra::Map<LO,GO,NO>;
-  using import_type    = Tpetra::Import<LO,GO,NO>;
-  using go_vector_type = Tpetra::Vector<GO,LO,GO,NO>;
-
+  using map_type       = Tpetra::Map<LO, GO, NO>;
+  using import_type    = Tpetra::Import<LO, GO, NO>;
+  using go_vector_type = Tpetra::Vector<GO, LO, GO, NO>;
 
   // Must create GloballyContiguous DomainMap (which is a permutation of Matrix_'s
   // DomainMap) and the corresponding permuted ColumnMap.
@@ -152,45 +145,40 @@ void MakeContiguousMaps(const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdina
   RCP<const map_type> ColumnMap   = Matrix.getColMap();
   RCP<const import_type> importer = Matrix.getGraph()->getImporter();
 
-  if(RowMap->isContiguous() ) {
+  if (RowMap->isContiguous()) {
     // If the row map is linear, then we can just use the row map as is.
     ContiguousRowMap = RowMap;
-  }
-  else {
+  } else {
     // The row map isn't linear, so we need a new row map
     ContiguousRowMap = rcp(new map_type(Matrix.getRowMap()->getGlobalNumElements(),
                                         Matrix.getRowMap()->getLocalNumElements(), 0, Matrix.getRowMap()->getComm()));
   }
 
-
-  if(DomainMap->isContiguous() ) {
+  if (DomainMap->isContiguous()) {
     // If the domain map is linear, then we can just use the column map as is.
     ContiguousDomainMap = DomainMap;
     ContiguousColumnMap = ColumnMap;
-  }
-  else {
+  } else {
     // The domain map isn't linear, so we need a new domain map
     ContiguousDomainMap = rcp(new map_type(DomainMap->getGlobalNumElements(),
                                            DomainMap->getLocalNumElements(), 0, DomainMap->getComm()));
-    if(importer) {
+    if (importer) {
       // If there's an importer then we can use it to get a new column map
-      go_vector_type MyGIDsHYPRE(DomainMap,ContiguousDomainMap->getLocalElementList());
+      go_vector_type MyGIDsHYPRE(DomainMap, ContiguousDomainMap->getLocalElementList());
 
       // import the HYPRE GIDs
       go_vector_type ColGIDsHYPRE(ColumnMap);
       ColGIDsHYPRE.doImport(MyGIDsHYPRE, *importer, Tpetra::INSERT);
-   
+
       // Make a HYPRE numbering-based column map.
-      ContiguousColumnMap = rcp(new map_type(ColumnMap->getGlobalNumElements(),ColGIDsHYPRE.getDataNonConst()(),0, ColumnMap->getComm()));
-    }
-    else {
+      ContiguousColumnMap = rcp(new map_type(ColumnMap->getGlobalNumElements(), ColGIDsHYPRE.getDataNonConst()(), 0, ColumnMap->getComm()));
+    } else {
       // The problem has matching domain/column maps, and somehow the domain map isn't linear, so just use the new domain map
-      ContiguousColumnMap = rcp(new map_type(ColumnMap->getGlobalNumElements(),ContiguousDomainMap->getLocalElementList(), 0, ColumnMap->getComm()));
+      ContiguousColumnMap = rcp(new map_type(ColumnMap->getGlobalNumElements(), ContiguousDomainMap->getLocalElementList(), 0, ColumnMap->getComm()));
     }
   }
 }
 
-
 // =========================================================================
 // PETSC Testing
 // =========================================================================
@@ -200,69 +188,68 @@ void MakeContiguousMaps(const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdina
   PetscCall(x); \
   }*/
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class Petsc_SpmV_Pack {
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
-  typedef Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> map_type;
-public:
-  Petsc_SpmV_Pack (const crs_matrix_type& A,
-                   const vector_type& X,
-                   vector_type& Y) {
-    using LO = LocalOrdinal;
-    const MPI_Comm & comm = *Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(A.getRowMap()->getComm())->getRawMpiComm();
-    LO Nx = (LO) X.getMap()->getLocalNumElements();
-    LO Ny = (LO) Y.getMap()->getLocalNumElements();
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
+  typedef Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+
+ public:
+  Petsc_SpmV_Pack(const crs_matrix_type& A,
+                  const vector_type& X,
+                  vector_type& Y) {
+    using LO             = LocalOrdinal;
+    const MPI_Comm& comm = *Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(A.getRowMap()->getComm())->getRawMpiComm();
+    LO Nx                = (LO)X.getMap()->getLocalNumElements();
+    LO Ny                = (LO)Y.getMap()->getLocalNumElements();
 
     // NOTE:  Petsc requires contiguous GIDs for the row map
     Teuchos::RCP<const map_type> c_row_map, c_col_map, c_domain_map;
-    MakeContiguousMaps(A,c_row_map,c_col_map,c_domain_map);
+    MakeContiguousMaps(A, c_row_map, c_col_map, c_domain_map);
 
     // Note: PETSC appears to favor local indices for vector insertion
-    std::vector<PetscInt> l_indices(std::max(Nx,Ny));
-    for(int i=0; i<std::max(Nx,Ny); i++)
+    std::vector<PetscInt> l_indices(std::max(Nx, Ny));
+    for (int i = 0; i < std::max(Nx, Ny); i++)
       l_indices[i] = i;
 
     // x vector
-    VecCreate(comm,&x_p);
-    VecSetType(x_p,VECMPI);
-    VecSetSizes(x_p,Nx, X.getMap()->getGlobalNumElements());
-    VecSetValues(x_p,Nx,l_indices.data(),X.getData(0).getRawPtr(),INSERT_VALUES);
+    VecCreate(comm, &x_p);
+    VecSetType(x_p, VECMPI);
+    VecSetSizes(x_p, Nx, X.getMap()->getGlobalNumElements());
+    VecSetValues(x_p, Nx, l_indices.data(), X.getData(0).getRawPtr(), INSERT_VALUES);
     VecAssemblyBegin(x_p);
     VecAssemblyEnd(x_p);
 
     // y vector
-    VecCreate(comm,&y_p);
-    VecSetType(y_p,VECMPI);
-    VecSetSizes(y_p,Ny, Y.getMap()->getGlobalNumElements());
-    VecSetValues(y_p,Ny,l_indices.data(),Y.getData(0).getRawPtr(),INSERT_VALUES);
+    VecCreate(comm, &y_p);
+    VecSetType(y_p, VECMPI);
+    VecSetSizes(y_p, Ny, Y.getMap()->getGlobalNumElements());
+    VecSetValues(y_p, Ny, l_indices.data(), Y.getData(0).getRawPtr(), INSERT_VALUES);
     VecAssemblyBegin(y_p);
     VecAssemblyEnd(y_p);
 
     // A matrix
     // NOTE: This is an overallocation, but that's OK
     PetscInt max_nnz = A.getLocalMaxNumRowEntries();
-    MatCreateAIJ(comm,c_row_map->getLocalNumElements(),c_domain_map->getLocalNumElements(),PETSC_DECIDE,PETSC_DECIDE,
-                 max_nnz, NULL, max_nnz,NULL, &A_p);
-
+    MatCreateAIJ(comm, c_row_map->getLocalNumElements(), c_domain_map->getLocalNumElements(), PETSC_DECIDE, PETSC_DECIDE,
+                 max_nnz, NULL, max_nnz, NULL, &A_p);
 
     std::vector<PetscInt> new_indices(max_nnz);
-    for(LO i = 0; i < (LO)A.getLocalNumRows(); i++){
-      typename crs_matrix_type::values_host_view_type     values;
+    for (LO i = 0; i < (LO)A.getLocalNumRows(); i++) {
+      typename crs_matrix_type::values_host_view_type values;
       typename crs_matrix_type::local_inds_host_view_type indices;
       A.getLocalRowView(i, indices, values);
-      for(LO j = 0; j < (LO) indices.extent(0); j++){
+      for (LO j = 0; j < (LO)indices.extent(0); j++) {
         new_indices[j] = c_col_map->getGlobalElement(indices(j));
       }
       PetscInt GlobalRow[1];
-      PetscInt numEntries = (PetscInt) indices.extent(0);
-      GlobalRow[0] = c_row_map->getGlobalElement(i);
-      
-      MatSetValues(A_p,1,GlobalRow, numEntries, new_indices.data(), values.data(),INSERT_VALUES);
-    }
-    MatAssemblyBegin(A_p,MAT_FINAL_ASSEMBLY);
-    MatAssemblyEnd(A_p,MAT_FINAL_ASSEMBLY);
+      PetscInt numEntries = (PetscInt)indices.extent(0);
+      GlobalRow[0]        = c_row_map->getGlobalElement(i);
 
+      MatSetValues(A_p, 1, GlobalRow, numEntries, new_indices.data(), values.data(), INSERT_VALUES);
+    }
+    MatAssemblyBegin(A_p, MAT_FINAL_ASSEMBLY);
+    MatAssemblyEnd(A_p, MAT_FINAL_ASSEMBLY);
   }
 
   ~Petsc_SpmV_Pack() {
@@ -272,96 +259,91 @@ class Petsc_SpmV_Pack {
   }
 
   bool spmv(const Scalar alpha, const Scalar beta) {
-    int rv = MatMult(A_p,x_p,y_p);
+    int rv = MatMult(A_p, x_p, y_p);
     Kokkos::fence();
     return (rv != 0);
   }
 
-private:
+ private:
   Mat A_p;
   Vec x_p, y_p;
-  
 };
 
-
 #endif
 
-
-
 // =========================================================================
 // HYPRE Testing
 // =========================================================================
 #if defined(HAVE_MUELU_HYPRE) && defined(HAVE_MPI)
 
-#define HYPRE_CHK_ERR(x)  { \
-    if(x!=0) throw std::runtime_error("ERROR: HYPRE returned non-zero exit code"); \
+#define HYPRE_CHK_ERR(x)                                                              \
+  {                                                                                   \
+    if (x != 0) throw std::runtime_error("ERROR: HYPRE returned non-zero exit code"); \
   }
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class HYPRE_SpmV_Pack {
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
-  typedef Tpetra::Map<LocalOrdinal,GlobalOrdinal,Node> map_type;
-public:
-  HYPRE_SpmV_Pack (const crs_matrix_type& A,
-                   const vector_type& X,
-                   vector_type& Y) {
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
+  typedef Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node> map_type;
+
+ public:
+  HYPRE_SpmV_Pack(const crs_matrix_type& A,
+                  const vector_type& X,
+                  vector_type& Y) {
     using LO = LocalOrdinal;
     using GO = GlobalOrdinal;
-    
-    // Time to copy the matrix / vector over to HYPRE datastructures
-    const MPI_Comm & comm = *Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(A.getRowMap()->getComm())->getRawMpiComm();
 
+    // Time to copy the matrix / vector over to HYPRE datastructures
+    const MPI_Comm& comm = *Teuchos::rcp_dynamic_cast<const Teuchos::MpiComm<int> >(A.getRowMap()->getComm())->getRawMpiComm();
 
     // NOTE:  Hypre requires contiguous GIDs for the row map
     Teuchos::RCP<const map_type> c_row_map, c_col_map, c_domain_map;
-    MakeContiguousMaps(A,c_row_map,c_col_map,c_domain_map);
-    
+    MakeContiguousMaps(A, c_row_map, c_col_map, c_domain_map);
+
     // Create matrix
     int row_lo = c_row_map->getMinGlobalIndex();
     int row_hi = c_row_map->getMaxGlobalIndex();
     int dom_lo = c_domain_map->getMinGlobalIndex();
     int dom_hi = c_domain_map->getMaxGlobalIndex();
-    HYPRE_CHK_ERR(HYPRE_IJMatrixCreate(comm,row_lo,row_hi,dom_lo,dom_hi, &ij_matrix));
-    HYPRE_CHK_ERR(HYPRE_IJMatrixSetObjectType(ij_matrix,HYPRE_PARCSR));
+    HYPRE_CHK_ERR(HYPRE_IJMatrixCreate(comm, row_lo, row_hi, dom_lo, dom_hi, &ij_matrix));
+    HYPRE_CHK_ERR(HYPRE_IJMatrixSetObjectType(ij_matrix, HYPRE_PARCSR));
     HYPRE_CHK_ERR(HYPRE_IJMatrixInitialize(ij_matrix));
 
-
     // Fill matrix
     std::vector<GO> new_indices(A.getLocalMaxNumRowEntries());
-    for(LO i = 0; i < (LO)A.getLocalNumRows(); i++){
-      typename crs_matrix_type::values_host_view_type     values;
+    for (LO i = 0; i < (LO)A.getLocalNumRows(); i++) {
+      typename crs_matrix_type::values_host_view_type values;
       typename crs_matrix_type::local_inds_host_view_type indices;
       A.getLocalRowView(i, indices, values);
-      for(LO j = 0; j < (LO) indices.extent(0); j++){
+      for (LO j = 0; j < (LO)indices.extent(0); j++) {
         new_indices[j] = c_col_map->getGlobalElement(indices(j));
       }
       GO GlobalRow[1];
-      GO numEntries = (GO) indices.extent(0);
-      GlobalRow[0] = c_row_map->getGlobalElement(i);
+      GO numEntries = (GO)indices.extent(0);
+      GlobalRow[0]  = c_row_map->getGlobalElement(i);
       HYPRE_CHK_ERR(HYPRE_IJMatrixSetValues(ij_matrix, 1, &numEntries, GlobalRow, new_indices.data(), values.data()));
     }
-     HYPRE_CHK_ERR(HYPRE_IJMatrixAssemble(ij_matrix));
-     HYPRE_CHK_ERR(HYPRE_IJMatrixGetObject(ij_matrix, (void**)&parcsr_matrix));
+    HYPRE_CHK_ERR(HYPRE_IJMatrixAssemble(ij_matrix));
+    HYPRE_CHK_ERR(HYPRE_IJMatrixGetObject(ij_matrix, (void**)&parcsr_matrix));
 
     // Now the x vector
-    GO * dom_indices = const_cast<GO*>(c_domain_map->getLocalElementList().getRawPtr());
+    GO* dom_indices = const_cast<GO*>(c_domain_map->getLocalElementList().getRawPtr());
     HYPRE_CHK_ERR(HYPRE_IJVectorCreate(comm, dom_lo, dom_hi, &x_ij));
     HYPRE_CHK_ERR(HYPRE_IJVectorSetObjectType(x_ij, HYPRE_PARCSR));
     HYPRE_CHK_ERR(HYPRE_IJVectorInitialize(x_ij));
-    HYPRE_CHK_ERR(HYPRE_IJVectorSetValues(x_ij,X.getLocalLength(),dom_indices,const_cast<vector_type*>(&X)->getDataNonConst(0).getRawPtr()));
+    HYPRE_CHK_ERR(HYPRE_IJVectorSetValues(x_ij, X.getLocalLength(), dom_indices, const_cast<vector_type*>(&X)->getDataNonConst(0).getRawPtr()));
     HYPRE_CHK_ERR(HYPRE_IJVectorAssemble(x_ij));
-    HYPRE_CHK_ERR(HYPRE_IJVectorGetObject(x_ij, (void**) &x_par));
-    
-    
+    HYPRE_CHK_ERR(HYPRE_IJVectorGetObject(x_ij, (void**)&x_par));
+
     // Now the y vector
-    GO * row_indices = const_cast<GO*>(c_row_map->getLocalElementList().getRawPtr());
-    HYPRE_CHK_ERR(HYPRE_IJVectorCreate(comm, row_lo,row_hi, &y_ij));
+    GO* row_indices = const_cast<GO*>(c_row_map->getLocalElementList().getRawPtr());
+    HYPRE_CHK_ERR(HYPRE_IJVectorCreate(comm, row_lo, row_hi, &y_ij));
     HYPRE_CHK_ERR(HYPRE_IJVectorSetObjectType(y_ij, HYPRE_PARCSR));
     HYPRE_CHK_ERR(HYPRE_IJVectorInitialize(y_ij));
-    HYPRE_CHK_ERR(HYPRE_IJVectorSetValues(y_ij,Y.getLocalLength(),row_indices,Y.getDataNonConst(0).getRawPtr()));
+    HYPRE_CHK_ERR(HYPRE_IJVectorSetValues(y_ij, Y.getLocalLength(), row_indices, Y.getDataNonConst(0).getRawPtr()));
     HYPRE_CHK_ERR(HYPRE_IJVectorAssemble(y_ij));
-    HYPRE_CHK_ERR(HYPRE_IJVectorGetObject(y_ij, (void**) &y_par));
+    HYPRE_CHK_ERR(HYPRE_IJVectorGetObject(y_ij, (void**)&y_par));
   }
 
   ~HYPRE_SpmV_Pack() {
@@ -371,12 +353,12 @@ class HYPRE_SpmV_Pack {
   }
 
   bool spmv(const Scalar alpha, const Scalar beta) {
-    int rv = HYPRE_ParCSRMatrixMatvec(alpha,parcsr_matrix, x_par,beta, y_par);
+    int rv = HYPRE_ParCSRMatrixMatvec(alpha, parcsr_matrix, x_par, beta, y_par);
     Kokkos::fence();
     return (rv != 0);
   }
 
-private:
+ private:
   // NOTE:  The par matrix/vectors are just pointers to the internal data of the ij guys.  They do not need to
   // be deallocated on their own.
   HYPRE_IJMatrix ij_matrix;
@@ -384,88 +366,83 @@ class HYPRE_SpmV_Pack {
 
   HYPRE_IJVector x_ij, y_ij;
   HYPRE_ParVector x_par, y_par;
-
-
 };
 
 #endif
 
-
 // =========================================================================
 // MAGMASparse Testing
 // =========================================================================
 #if defined(HAVE_MUELU_MAGMASPARSE)
 #include <magma_v2.h>
 #include <magmasparse.h>
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class MagmaSparse_SpmV_Pack {
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
 
-public:
-  MagmaSparse_SpmV_Pack (const crs_matrix_type& A,
-                         const vector_type& X,
-                         vector_type& Y)
-  {}
+ public:
+  MagmaSparse_SpmV_Pack(const crs_matrix_type& A,
+                        const vector_type& X,
+                        vector_type& Y) {}
 
- ~MagmaSparse_SpmV_Pack() {};
+  ~MagmaSparse_SpmV_Pack(){};
 
   bool spmv(const Scalar alpha, const Scalar beta) { return (true); }
 };
 
-template<typename LocalOrdinal, typename GlobalOrdinal>
-class MagmaSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosCudaWrapperNode> {
+template <typename LocalOrdinal, typename GlobalOrdinal>
+class MagmaSparse_SpmV_Pack<double, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosCudaWrapperNode> {
   // typedefs shared among other TPLs
   typedef double Scalar;
   typedef typename Tpetra::KokkosCompat::KokkosCudaWrapperNode Node;
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
-  typedef typename crs_matrix_type::local_matrix_host_type    KCRS;
-  typedef typename KCRS::StaticCrsGraphType              graph_t;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
+  typedef typename crs_matrix_type::local_matrix_host_type KCRS;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
   typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-  typedef typename graph_t::row_map_type::const_type     c_lno_view_t;
+  typedef typename graph_t::row_map_type::const_type c_lno_view_t;
   typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-  typedef typename graph_t::entries_type::const_type     c_lno_nnz_view_t;
-  typedef typename KCRS::values_type::non_const_type     scalar_view_t;
+  typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
   typedef typename Node::device_type device_type;
 
   typedef typename Kokkos::View<int*,
                                 typename lno_nnz_view_t::array_layout,
-                                typename lno_nnz_view_t::device_type> int_array_type;
+                                typename lno_nnz_view_t::device_type>
+      int_array_type;
 
-public:
-  MagmaSparse_SpmV_Pack (const crs_matrix_type& A,
-                         const vector_type& X,
-                               vector_type& Y)
-  {
+ public:
+  MagmaSparse_SpmV_Pack(const crs_matrix_type& A,
+                        const vector_type& X,
+                        vector_type& Y) {
     // data access common to other TPLs
-    const KCRS & Amat = A.getLocalMatrixHost();
-    c_lno_view_t Arowptr = Amat.graph.row_map;
-    c_lno_nnz_view_t Acolind = Amat.graph.entries;
+    const KCRS& Amat          = A.getLocalMatrixHost();
+    c_lno_view_t Arowptr      = Amat.graph.row_map;
+    c_lno_nnz_view_t Acolind  = Amat.graph.entries;
     const scalar_view_t Avals = Amat.values;
 
     // type conversion
     Arowptr_int = int_array_type("Arowptr", Arowptr.extent(0));
     // copy the ordinals into the local view (type conversion)
-    copy_view(Arowptr,Arowptr_int);
-
+    copy_view(Arowptr, Arowptr_int);
 
-    m   = static_cast<int>(A.getLocalNumRows());
-    n   = static_cast<int>(A.getLocalNumCols());
-    nnz = static_cast<int>(Acolind.extent(0));
+    m      = static_cast<int>(A.getLocalNumRows());
+    n      = static_cast<int>(A.getLocalNumCols());
+    nnz    = static_cast<int>(Acolind.extent(0));
     vals   = reinterpret_cast<Scalar*>(Avals.data());
     cols   = const_cast<int*>(reinterpret_cast<const int*>(Acolind.data()));
     rowptr = reinterpret_cast<int*>(Arowptr_int.data());
 
-    auto X_lcl = X.template getLocalView<device_type> ();
-    auto Y_lcl = Y.template getLocalView<device_type> ();
-    x = reinterpret_cast<Scalar*>(X_lcl.data());
-    y = reinterpret_cast<Scalar*>(Y_lcl.data());
+    auto X_lcl = X.template getLocalView<device_type>();
+    auto Y_lcl = Y.template getLocalView<device_type>();
+    x          = reinterpret_cast<Scalar*>(X_lcl.data());
+    y          = reinterpret_cast<Scalar*>(Y_lcl.data());
 
-    magma_init ();
+    magma_init();
     int device;
-    magma_getdevice( &device );
-    magma_queue_create( device, &queue );
+    magma_getdevice(&device);
+    magma_queue_create(device, &queue);
 
     // create the magma data container
     magma_dev_Acrs = {Magma_CSR};
@@ -474,147 +451,140 @@ class MagmaSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosComp
     magma_dev_y    = {Magma_DENSE};
 
     //
-    magma_dvset_dev (m, 1,         // assume mx1 size
-                     x,            // ptr to data on the device
-                     &magma_dev_x, // magma vector to populate
-                     queue);
-
-    magma_dvset_dev (m, 1,         // assume mx1 size
-                     y,            // ptr to data on the device
-                     &magma_dev_y, // magma vector to populate
-                     queue);
-    magma_dcsrset( m, n, rowptr, cols, vals, &magma_Acrs, queue );
-    magma_dmtransfer( magma_Acrs, &magma_dev_Acrs, Magma_DEV, Magma_DEV, queue );
-
+    magma_dvset_dev(m, 1,          // assume mx1 size
+                    x,             // ptr to data on the device
+                    &magma_dev_x,  // magma vector to populate
+                    queue);
+
+    magma_dvset_dev(m, 1,          // assume mx1 size
+                    y,             // ptr to data on the device
+                    &magma_dev_y,  // magma vector to populate
+                    queue);
+    magma_dcsrset(m, n, rowptr, cols, vals, &magma_Acrs, queue);
+    magma_dmtransfer(magma_Acrs, &magma_dev_Acrs, Magma_DEV, Magma_DEV, queue);
   }
 
-  ~MagmaSparse_SpmV_Pack()
-  {
+  ~MagmaSparse_SpmV_Pack() {
     // we dont' own the data... not sure what magma does here...
     magma_dmfree(&magma_dev_x, queue);
     magma_dmfree(&magma_dev_y, queue);
     magma_dmfree(&magma_dev_Acrs, queue);
 
-    magma_finalize ();
+    magma_finalize();
   };
 
-  bool spmv(const Scalar alpha, const Scalar beta)
-  {
-    magma_d_spmv( 1.0, magma_dev_Acrs, magma_dev_x, 0.0, magma_dev_y, queue );
+  bool spmv(const Scalar alpha, const Scalar beta) {
+    magma_d_spmv(1.0, magma_dev_Acrs, magma_dev_x, 0.0, magma_dev_y, queue);
     Kokkos::fence();
   }
 
-private:
+ private:
   magma_d_matrix magma_dev_Acrs;
   magma_d_matrix magma_Acrs;
   magma_d_matrix magma_dev_x;
   magma_d_matrix magma_dev_y;
   magma_queue_t queue;
 
-  int m = -1;
-  int n = -1;
-  int nnz = -1;
-  Scalar * vals  = nullptr; // aliased
-  int * cols     = nullptr; // aliased
-  int * rowptr   = nullptr; // copied
-  Scalar * x     = nullptr; // aliased
-  Scalar * y     = nullptr; // aliased
+  int m        = -1;
+  int n        = -1;
+  int nnz      = -1;
+  Scalar* vals = nullptr;  // aliased
+  int* cols    = nullptr;  // aliased
+  int* rowptr  = nullptr;  // copied
+  Scalar* x    = nullptr;  // aliased
+  Scalar* y    = nullptr;  // aliased
 
   // handles to the copied data
   int_array_type Arowptr_int;
 };
 #endif
 
-
 // =========================================================================
 // CuSparse Testing
 // =========================================================================
 #if defined(HAVE_MUELU_CUSPARSE)
 
-#define CHECK_CUDA(func)                                                       \
-{                                                                              \
-    cudaError_t status = (func);                                               \
-    if (status != cudaSuccess) {                                               \
-        printf("CUDA API failed at line %d with error: %s (%d)\n",             \
-               __LINE__, cudaGetErrorString(status), status);                  \
-    }                                                                          \
-}
-
-#define CHECK_CUSPARSE(func)                                                   \
-{                                                                              \
-    cusparseStatus_t status = (func);                                          \
-    if (status != CUSPARSE_STATUS_SUCCESS) {                                   \
-        printf("CUSPARSE API failed at line %d with error: %s (%d)\n",         \
-               __LINE__, cusparseGetErrorString(status), status);              \
-    }                                                                          \
-}
+#define CHECK_CUDA(func)                                         \
+  {                                                              \
+    cudaError_t status = (func);                                 \
+    if (status != cudaSuccess) {                                 \
+      printf("CUDA API failed at line %d with error: %s (%d)\n", \
+             __LINE__, cudaGetErrorString(status), status);      \
+    }                                                            \
+  }
 
+#define CHECK_CUSPARSE(func)                                         \
+  {                                                                  \
+    cusparseStatus_t status = (func);                                \
+    if (status != CUSPARSE_STATUS_SUCCESS) {                         \
+      printf("CUSPARSE API failed at line %d with error: %s (%d)\n", \
+             __LINE__, cusparseGetErrorString(status), status);      \
+    }                                                                \
+  }
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class CuSparse_SpmV_Pack {
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
 
-public:
-  CuSparse_SpmV_Pack (const crs_matrix_type& A,
-                      const vector_type& X,
-                            vector_type& Y)
-  {}
+ public:
+  CuSparse_SpmV_Pack(const crs_matrix_type& A,
+                     const vector_type& X,
+                     vector_type& Y) {}
 
- ~CuSparse_SpmV_Pack() {};
+  ~CuSparse_SpmV_Pack(){};
 
   cusparseStatus_t spmv(const Scalar alpha, const Scalar beta) { return CUSPARSE_STATUS_SUCCESS; }
 };
 
-template<typename LocalOrdinal, typename GlobalOrdinal>
-class CuSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat::KokkosCudaWrapperNode> {
+template <typename LocalOrdinal, typename GlobalOrdinal>
+class CuSparse_SpmV_Pack<double, LocalOrdinal, GlobalOrdinal, Tpetra::KokkosCompat::KokkosCudaWrapperNode> {
   // typedefs shared among other TPLs
   typedef double Scalar;
   typedef typename Tpetra::KokkosCompat::KokkosCudaWrapperNode Node;
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-  typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
-  typedef typename crs_matrix_type::local_matrix_device_type    KCRS;
-  typedef typename KCRS::StaticCrsGraphType              graph_t;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+  typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
+  typedef typename crs_matrix_type::local_matrix_device_type KCRS;
+  typedef typename KCRS::StaticCrsGraphType graph_t;
   typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-  typedef typename graph_t::row_map_type::const_type     c_lno_view_t;
+  typedef typename graph_t::row_map_type::const_type c_lno_view_t;
   typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-  typedef typename graph_t::entries_type::const_type     c_lno_nnz_view_t;
-  typedef typename KCRS::values_type::non_const_type     scalar_view_t;
+  typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
   typedef typename Node::device_type device_type;
 
   typedef typename Kokkos::View<int*,
                                 typename lno_nnz_view_t::array_layout,
-                                typename lno_nnz_view_t::device_type> cusparse_int_type;
-public:
-  CuSparse_SpmV_Pack (const crs_matrix_type& A,
-                      const vector_type& X,
-                            vector_type& Y)
-  {
+                                typename lno_nnz_view_t::device_type>
+      cusparse_int_type;
+
+ public:
+  CuSparse_SpmV_Pack(const crs_matrix_type& A,
+                     const vector_type& X,
+                     vector_type& Y) {
     // data access common to other TPLs
-    const KCRS & Amat = A.getLocalMatrixDevice();
-    c_lno_view_t Arowptr = Amat.graph.row_map;
-    c_lno_nnz_view_t Acolind = Amat.graph.entries;
+    const KCRS& Amat          = A.getLocalMatrixDevice();
+    c_lno_view_t Arowptr      = Amat.graph.row_map;
+    c_lno_nnz_view_t Acolind  = Amat.graph.entries;
     const scalar_view_t Avals = Amat.values;
 
-
     Arowptr_cusparse = cusparse_int_type("Arowptr", Arowptr.extent(0));
     Acolind_cusparse = cusparse_int_type("Acolind", Acolind.extent(0));
     // copy the ordinals into the local view (type conversion)
-    copy_view(Arowptr,Arowptr_cusparse);
-    copy_view(Acolind,Acolind_cusparse);
+    copy_view(Arowptr, Arowptr_cusparse);
+    copy_view(Acolind, Acolind_cusparse);
 
-
-    m   = static_cast<int>(A.getLocalNumRows());
-    n   = static_cast<int>(A.getLocalNumCols());
-    nnz = static_cast<int>(Acolind_cusparse.extent(0));
+    m      = static_cast<int>(A.getLocalNumRows());
+    n      = static_cast<int>(A.getLocalNumCols());
+    nnz    = static_cast<int>(Acolind_cusparse.extent(0));
     vals   = reinterpret_cast<Scalar*>(Avals.data());
     cols   = reinterpret_cast<int*>(Acolind_cusparse.data());
     rowptr = reinterpret_cast<int*>(Arowptr_cusparse.data());
 
     auto X_lcl = X.getLocalViewDevice(Tpetra::Access::ReadOnly);
     auto Y_lcl = Y.getLocalViewDevice(Tpetra::Access::ReadWrite);
-    x = const_cast<Scalar*>(reinterpret_cast<const Scalar*>(X_lcl.data()));
-    y = reinterpret_cast<Scalar*>(Y_lcl.data());
+    x          = const_cast<Scalar*>(reinterpret_cast<const Scalar*>(X_lcl.data()));
+    y          = reinterpret_cast<Scalar*>(Y_lcl.data());
 
     /* Get handle to the CUBLAS context */
     cublasCreate(&cublasHandle);
@@ -622,18 +592,18 @@ class CuSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat:
     cusparseCreate(&cusparseHandle);
 
     CHECK_CUSPARSE(cusparseCreateCsr(&descrA, m, n, nnz,
-                                      rowptr, cols, vals,
-                                      CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
-                                      CUSPARSE_INDEX_BASE_ZERO, CUDA_R_64F));
+                                     rowptr, cols, vals,
+                                     CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
+                                     CUSPARSE_INDEX_BASE_ZERO, CUDA_R_64F));
 
     CHECK_CUSPARSE(cusparseCreateDnVec(&vecY, m, y, CUDA_R_64F));
     CHECK_CUSPARSE(cusparseCreateDnVec(&vecX, n, x, CUDA_R_64F));
   }
 
-  ~CuSparse_SpmV_Pack () {
-    CHECK_CUSPARSE( cusparseDestroySpMat(descrA) );
-    CHECK_CUSPARSE( cusparseDestroyDnVec(vecX) );
-    CHECK_CUSPARSE( cusparseDestroyDnVec(vecY) );
+  ~CuSparse_SpmV_Pack() {
+    CHECK_CUSPARSE(cusparseDestroySpMat(descrA));
+    CHECK_CUSPARSE(cusparseDestroyDnVec(vecX));
+    CHECK_CUSPARSE(cusparseDestroyDnVec(vecY));
     cusparseDestroy(cusparseHandle);
     cublasDestroy(cublasHandle);
   }
@@ -677,24 +647,25 @@ class CuSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat:
     Kokkos::fence();
     return (rc);
   }
-private:
-  cublasHandle_t     cublasHandle   = 0;
-  cusparseHandle_t   cusparseHandle = 0;
-  cusparseSpMatDescr_t descrA         = 0;
+
+ private:
+  cublasHandle_t cublasHandle     = 0;
+  cusparseHandle_t cusparseHandle = 0;
+  cusparseSpMatDescr_t descrA     = 0;
   cusparseDnVecDescr_t vecX = 0, vecY = 0;
 
   // CUSPARSE_OPERATION_NON_TRANSPOSE
   // CUSPARSE_OPERATION_TRANSPOSE
   // CUSPARSE_OPERATION_CONJUGATE_TRANSPOSE
   cusparseOperation_t transA = CUSPARSE_OPERATION_NON_TRANSPOSE;
-  int m = -1;
-  int n = -1;
-  int nnz = -1;
-  Scalar * vals  = nullptr; // aliased
-  int * cols     = nullptr; // copied
-  int * rowptr   = nullptr; // copied
-  Scalar * x     = nullptr; // aliased
-  Scalar * y     = nullptr; // aliased
+  int m                      = -1;
+  int n                      = -1;
+  int nnz                    = -1;
+  Scalar* vals               = nullptr;  // aliased
+  int* cols                  = nullptr;  // copied
+  int* rowptr                = nullptr;  // copied
+  Scalar* x                  = nullptr;  // aliased
+  Scalar* y                  = nullptr;  // aliased
 
   // handles to the copied data
   cusparse_int_type Arowptr_cusparse;
@@ -708,83 +679,76 @@ class CuSparse_SpmV_Pack<double,LocalOrdinal,GlobalOrdinal,Tpetra::KokkosCompat:
 #if defined(HAVE_MUELU_MKL)
 #include "mkl.h"
 
-
 std::string mkl_error(sparse_status_t code) {
-  switch(code) {
-  case SPARSE_STATUS_SUCCESS:
-    return std::string("Success");
-  case SPARSE_STATUS_NOT_INITIALIZED:
-    return std::string("Empty handle or matrix array");
-  case SPARSE_STATUS_ALLOC_FAILED:
-    return std::string("Memory allocation failed");
-  case SPARSE_STATUS_INVALID_VALUE:
-    return std::string("Input contains an invalid value");
-  case SPARSE_STATUS_EXECUTION_FAILED:
-    return std::string("Execution failed");
-  case SPARSE_STATUS_INTERNAL_ERROR:
-    return std::string("Internal error");
-  case SPARSE_STATUS_NOT_SUPPORTED:
-    return std::string("Operation not supported");
+  switch (code) {
+    case SPARSE_STATUS_SUCCESS:
+      return std::string("Success");
+    case SPARSE_STATUS_NOT_INITIALIZED:
+      return std::string("Empty handle or matrix array");
+    case SPARSE_STATUS_ALLOC_FAILED:
+      return std::string("Memory allocation failed");
+    case SPARSE_STATUS_INVALID_VALUE:
+      return std::string("Input contains an invalid value");
+    case SPARSE_STATUS_EXECUTION_FAILED:
+      return std::string("Execution failed");
+    case SPARSE_STATUS_INTERNAL_ERROR:
+      return std::string("Internal error");
+    case SPARSE_STATUS_NOT_SUPPORTED:
+      return std::string("Operation not supported");
   };
-
 }
 
 struct matrix_descr mkl_descr;
 
-void MV_MKL(sparse_matrix_t & AMKL, double * x, double * y) {
-  //sparse_status_t mkl_sparse_d_mv (sparse_operation_t operation, double alpha, const sparse_matrix_t A, struct matrix_descr descr, const double *x, double beta, double *y);
+void MV_MKL(sparse_matrix_t& AMKL, double* x, double* y) {
+  // sparse_status_t mkl_sparse_d_mv (sparse_operation_t operation, double alpha, const sparse_matrix_t A, struct matrix_descr descr, const double *x, double beta, double *y);
 
-  mkl_sparse_d_mv(SPARSE_OPERATION_NON_TRANSPOSE,1.0,AMKL,mkl_descr,x,0.0,y);
+  mkl_sparse_d_mv(SPARSE_OPERATION_NON_TRANSPOSE, 1.0, AMKL, mkl_descr, x, 0.0, y);
   Kokkos::fence();
 }
 
 #endif
 
-
-
 // =========================================================================
 // Tpetra Kernel Testing
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MV_Tpetra(const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &x,   Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &y) {
-  A.apply(x,y);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MV_Tpetra(const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y) {
+  A.apply(x, y);
   Kokkos::fence();
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MV_KK(const Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &x,   Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> &y) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MV_KK(const Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A, const Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& x, Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& y) {
   typedef typename Node::device_type device_type;
   const auto& AK = A.getLocalMatrixDevice();
-  auto X_lcl = x.getLocalViewDevice (Tpetra::Access::ReadOnly);
-  auto Y_lcl = y.getLocalViewDevice (Tpetra::Access::OverwriteAll);
-  KokkosSparse::spmv(KokkosSparse::NoTranspose,Teuchos::ScalarTraits<Scalar>::one(),AK,X_lcl,Teuchos::ScalarTraits<Scalar>::zero(),Y_lcl);
+  auto X_lcl     = x.getLocalViewDevice(Tpetra::Access::ReadOnly);
+  auto Y_lcl     = y.getLocalViewDevice(Tpetra::Access::OverwriteAll);
+  KokkosSparse::spmv(KokkosSparse::NoTranspose, Teuchos::ScalarTraits<Scalar>::one(), AK, X_lcl, Teuchos::ScalarTraits<Scalar>::zero(), Y_lcl);
   Kokkos::fence();
 }
 
-
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using std::endl;
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  using std::endl;
 
 #if defined(HAVE_MUELU_PETSC) && defined(HAVE_MPI)
-  PetscInitialize(0,NULL,NULL,NULL);
+  PetscInitialize(0, NULL, NULL, NULL);
 #endif
 
-
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int numProc = comm->getSize();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int numProc                         = comm->getSize();
 
     // =========================================================================
     // Convenient definitions
@@ -793,133 +757,148 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> pOut = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *pOut;
+    Teuchos::FancyOStream& out      = *pOut;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 50;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    bool binaryFormat = false;  clp.setOption("binary", "ascii",  &binaryFormat,   "print timings to screen");
-    std::string matrixFile;     clp.setOption("matrixfile",       &matrixFile,     "matrix market file containing matrix");
-    std::string rowMapFile;     clp.setOption("rowmap",           &rowMapFile,     "map data file");
-    std::string colMapFile;     clp.setOption("colmap",           &colMapFile,     "colmap data file");
-    std::string domainMapFile;  clp.setOption("domainmap",        &domainMapFile,  "domainmap data file");
-    std::string rangeMapFile;   clp.setOption("rangemap",         &rangeMapFile,   "rangemap data file");
-
-    bool printTimings = true;   clp.setOption("timings", "notimings",  &printTimings, "print timings to screen");
-    int  nrepeat      = 1000;   clp.setOption("nrepeat",               &nrepeat,      "repeat the experiment N times");
-
-    bool describeMatrix = true; clp.setOption("showmatrix", "noshowmatrix",  &describeMatrix, "describe matrix");
-    bool useStackedTimer = false; clp.setOption("stackedtimer", "nostackedtimer",  &useStackedTimer, "use stacked timer");
-    bool verboseModel  = false; clp.setOption("verbosemodel", "noverbosemodel",  &verboseModel, "use stacked verbose performance model");
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+    bool binaryFormat = false;
+    clp.setOption("binary", "ascii", &binaryFormat, "print timings to screen");
+    std::string matrixFile;
+    clp.setOption("matrixfile", &matrixFile, "matrix market file containing matrix");
+    std::string rowMapFile;
+    clp.setOption("rowmap", &rowMapFile, "map data file");
+    std::string colMapFile;
+    clp.setOption("colmap", &colMapFile, "colmap data file");
+    std::string domainMapFile;
+    clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+    std::string rangeMapFile;
+    clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int nrepeat = 1000;
+    clp.setOption("nrepeat", &nrepeat, "repeat the experiment N times");
+
+    bool describeMatrix = true;
+    clp.setOption("showmatrix", "noshowmatrix", &describeMatrix, "describe matrix");
+    bool useStackedTimer = false;
+    clp.setOption("stackedtimer", "nostackedtimer", &useStackedTimer, "use stacked timer");
+    bool verboseModel = false;
+    clp.setOption("verbosemodel", "noverbosemodel", &verboseModel, "use stacked verbose performance model");
 
     // the kernels
-    bool do_mkl      = true;
-    bool do_tpetra   = true;
-    bool do_kk       = true;
-    bool do_cusparse = true;
+    bool do_mkl         = true;
+    bool do_tpetra      = true;
+    bool do_kk          = true;
+    bool do_cusparse    = true;
     bool do_magmasparse = true;
-    bool do_hypre    = true;
-    bool do_petsc    = true;
+    bool do_hypre       = true;
+    bool do_petsc       = true;
 
     bool report_error_norms = false;
-    // handle the TPLs
-    #if ! defined(HAVE_MUELU_MKL)
-      do_mkl = false;
-    #endif
-    #if ! defined(HAVE_MUELU_CUSPARSE)
-      do_cusparse = false;
-    #endif
-    #if ! defined(HAVE_MUELU_MAGMASPARSE)
-      do_magmasparse = false;
-    #endif
-    #if ! defined(HAVE_MUELU_HYPRE) || !defined(HAVE_MPI)
-      do_hypre = false;
-    #endif
-    #if ! defined(HAVE_MUELU_PETSC) || !defined(HAVE_MPI)
-      do_petsc = false;
-    #endif
-
+// handle the TPLs
+#if !defined(HAVE_MUELU_MKL)
+    do_mkl = false;
+#endif
+#if !defined(HAVE_MUELU_CUSPARSE)
+    do_cusparse = false;
+#endif
+#if !defined(HAVE_MUELU_MAGMASPARSE)
+    do_magmasparse = false;
+#endif
+#if !defined(HAVE_MUELU_HYPRE) || !defined(HAVE_MPI)
+    do_hypre = false;
+#endif
+#if !defined(HAVE_MUELU_PETSC) || !defined(HAVE_MPI)
+    do_petsc = false;
+#endif
 
-    clp.setOption("mkl",      "nomkl",      &do_mkl,        "Evaluate MKL");
-    clp.setOption("tpetra",   "notpetra",   &do_tpetra,     "Evaluate Tpetra");
-    clp.setOption("kk",       "nokk",       &do_kk,         "Evaluate KokkosKernels");
-    clp.setOption("cusparse", "nocusparse", &do_cusparse,   "Evaluate CuSparse");
-    clp.setOption("magamasparse", "nomagmasparse", &do_magmasparse,   "Evaluate MagmaSparse");
-    clp.setOption("hypre", "nohypre", &do_hypre,   "Evaluate Hypre");
-    clp.setOption("petsc", "nopetsc", &do_petsc,   "Evaluate Petsc");
+    clp.setOption("mkl", "nomkl", &do_mkl, "Evaluate MKL");
+    clp.setOption("tpetra", "notpetra", &do_tpetra, "Evaluate Tpetra");
+    clp.setOption("kk", "nokk", &do_kk, "Evaluate KokkosKernels");
+    clp.setOption("cusparse", "nocusparse", &do_cusparse, "Evaluate CuSparse");
+    clp.setOption("magamasparse", "nomagmasparse", &do_magmasparse, "Evaluate MagmaSparse");
+    clp.setOption("hypre", "nohypre", &do_hypre, "Evaluate Hypre");
+    clp.setOption("petsc", "nopetsc", &do_petsc, "Evaluate Petsc");
 
-    clp.setOption("report_error_norms", "noreport_error_norms", &report_error_norms,   "Report L2 norms for the solution");
+    clp.setOption("report_error_norms", "noreport_error_norms", &report_error_norms, "Report L2 norms for the solution");
 
     std::ostringstream galeriStream;
-    std::string rhsFile,coordFile,coordMapFile,nullFile, materialFile; //unused
+    std::string rhsFile, coordFile, coordMapFile, nullFile, materialFile;  // unused
     typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+    typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
     RCP<RealValuedMultiVector> coordinates;
     RCP<MultiVector> nullspace, material, x, b;
     RCP<Matrix> A;
     RCP<const Map> map;
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:                               break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // Load the matrix off disk (or generate it via Galeri), assuming only one right hand side is loaded.
-    MatrixLoad<SC,LO,GO,NO>(comm, lib, binaryFormat, matrixFile, rhsFile, rowMapFile, colMapFile, domainMapFile, rangeMapFile, coordFile, coordMapFile, nullFile, materialFile, map, A, coordinates, nullspace, material, x, b, 1, galeriParameters, xpetraParameters, galeriStream);
-
+    MatrixLoad<SC, LO, GO, NO>(comm, lib, binaryFormat, matrixFile, rhsFile, rowMapFile, colMapFile, domainMapFile, rangeMapFile, coordFile, coordMapFile, nullFile, materialFile, map, A, coordinates, nullspace, material, x, b, 1, galeriParameters, xpetraParameters, galeriStream);
 
     if (do_kk && comm->getSize() > 1) {
       out << "KK was requested, but this kernel this cannot be run on more than one rank. Disabling..." << endl;
       do_kk = false;
     }
 
-    #ifndef HAVE_MUELU_MKL
+#ifndef HAVE_MUELU_MKL
     if (do_mkl) {
       out << "MKL was requested, but this kernel is not available. Disabling..." << endl;
       do_mkl = false;
     }
-    #endif
+#endif
 
-    #if ! defined(HAVE_MUELU_CUSPARSE)
+#if !defined(HAVE_MUELU_CUSPARSE)
     if (do_cusparse) {
       out << "CuSparse was requested, but this kernel is not available. Disabling..." << endl;
       do_cusparse = false;
     }
-    #else
-    if(! std::is_same<NO, Tpetra::KokkosCompat::KokkosCudaWrapperNode>::value) do_cusparse = false;
-    #endif
+#else
+    if (!std::is_same<NO, Tpetra::KokkosCompat::KokkosCudaWrapperNode>::value) do_cusparse = false;
+#endif
 
-    #if ! defined(HAVE_MUELU_MAGMASPARSE)
+#if !defined(HAVE_MUELU_MAGMASPARSE)
     if (do_magmasparse) {
       out << "MagmaSparse was requested, but this kernel is not available. Disabling..." << endl;
       do_magmasparse = false;
     }
-    #endif
+#endif
 
-    #if ! defined(HAVE_MUELU_HYPRE)
+#if !defined(HAVE_MUELU_HYPRE)
     if (do_hypre) {
       out << "Hypre was requested, but this kernel is not available. Disabling..." << endl;
       do_hypre = false;
     }
-    #endif
+#endif
 
-    #if ! defined(HAVE_MUELU_PETSC)
+#if !defined(HAVE_MUELU_PETSC)
     if (do_hypre) {
       out << "Petsc was requested, but this kernel is not available. Disabling..." << endl;
       do_petsc = false;
     }
-    #endif
+#endif
 
     // simple hack to randomize order of experiments
-    enum class Experiments { MKL=0, TPETRA, KK, CUSPARSE, MAGMASPARSE, HYPRE, PETSC };
-    const char * const experiment_id_to_string[] = {
+    enum class Experiments { MKL = 0,
+                             TPETRA,
+                             KK,
+                             CUSPARSE,
+                             MAGMASPARSE,
+                             HYPRE,
+                             PETSC };
+    const char* const experiment_id_to_string[] = {
         "MKL        ",
         "Tpetra     ",
         "KK         ",
@@ -927,34 +906,33 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         "MagmaSparse",
         "HYPRE",
         "PETSC"};
-      
+
     std::vector<Experiments> my_experiments;
     // add the experiments we will run
 
-    #ifdef HAVE_MUELU_MKL
-    if (do_mkl) my_experiments.push_back(Experiments::MKL);   // MKL
-    #endif
+#ifdef HAVE_MUELU_MKL
+    if (do_mkl) my_experiments.push_back(Experiments::MKL);  // MKL
+#endif
 
-    #ifdef HAVE_MUELU_CUSPARSE
-    if (do_cusparse) my_experiments.push_back(Experiments::CUSPARSE);   // CuSparse
-    #endif
+#ifdef HAVE_MUELU_CUSPARSE
+    if (do_cusparse) my_experiments.push_back(Experiments::CUSPARSE);  // CuSparse
+#endif
 
-    #ifdef HAVE_MUELU_MAGMASPARSE
-    if (do_magmasparse) my_experiments.push_back(Experiments::MAGMASPARSE);   // MagmaSparse
-    #endif
+#ifdef HAVE_MUELU_MAGMASPARSE
+    if (do_magmasparse) my_experiments.push_back(Experiments::MAGMASPARSE);  // MagmaSparse
+#endif
 
-    #if defined(HAVE_MUELU_HYPRE) && defined(HAVE_MPI)
-    if (do_hypre) my_experiments.push_back(Experiments::HYPRE);   // HYPRE
-    #endif
+#if defined(HAVE_MUELU_HYPRE) && defined(HAVE_MPI)
+    if (do_hypre) my_experiments.push_back(Experiments::HYPRE);  // HYPRE
+#endif
 
-    #if defined(HAVE_MUELU_PETSC) && defined(HAVE_MPI)
-    if (do_petsc) my_experiments.push_back(Experiments::PETSC);   // PETSc
-    #endif
+#if defined(HAVE_MUELU_PETSC) && defined(HAVE_MPI)
+    if (do_petsc) my_experiments.push_back(Experiments::PETSC);  // PETSc
+#endif
 
     // assume these are available
-    if (do_tpetra)  my_experiments.push_back(Experiments::TPETRA);     // Tpetra
-    if (do_kk)     my_experiments.push_back(Experiments::KK);     // KK
-
+    if (do_tpetra) my_experiments.push_back(Experiments::TPETRA);  // Tpetra
+    if (do_kk) my_experiments.push_back(Experiments::KK);          // KK
 
     out << "========================================================" << endl
         << xpetraParameters
@@ -974,12 +952,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         << "Vector:        " << Teuchos::demangleName(typeid(MultiVector).name()) << endl
         << "Hierarchy:     " << Teuchos::demangleName(typeid(Hierarchy).name()) << endl
         << "========================================================" << endl
-        << " MPI Ranks:    " << numProc <<endl
+        << " MPI Ranks:    " << numProc << endl
         << "========================================================" << endl;
 
 #if defined(HAVE_TPETRA_INST_OPENMP)
-    out<< "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = "<<Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency()<<endl
-       << "========================================================" << endl;
+    out << "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = " << Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() << endl
+        << "========================================================" << endl;
 #endif
 
     // =========================================================================
@@ -992,102 +970,98 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     comm->barrier();
 
-
-    RCP<MultiVector> y = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(A->getRowMap());
-    RCP<MultiVector> y_baseline = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(A->getRowMap());
+    RCP<MultiVector> y          = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(A->getRowMap());
+    RCP<MultiVector> y_baseline = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(A->getRowMap());
     x->putScalar(Teuchos::ScalarTraits<Scalar>::one());
     y->putScalar(Teuchos::ScalarTraits<Scalar>::nan());
     y_baseline->putScalar(Teuchos::ScalarTraits<Scalar>::nan());
 
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> vector_type;
+    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+    typedef Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> vector_type;
     RCP<const crs_matrix_type> At;
     vector_type xt;
     vector_type yt;
 
-    At = Utilities::Op2TpetraCrs(A);
-    const crs_matrix_type &Att = *At;
-    xt = Xpetra::toTpetra(*x);
-    yt = Xpetra::toTpetra(*y);
+    At                         = Utilities::Op2TpetraCrs(A);
+    const crs_matrix_type& Att = *At;
+    xt                         = Xpetra::toTpetra(*x);
+    yt                         = Xpetra::toTpetra(*y);
 
     size_t l_permutes = 0, g_permutes = 0;
-    if(!At->getGraph()->getImporter().is_null()) {
+    if (!At->getGraph()->getImporter().is_null()) {
       l_permutes = At->getGraph()->getImporter()->getNumPermuteIDs();
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM,1,&l_permutes,&g_permutes);
+      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &l_permutes, &g_permutes);
     }
 
-  #if defined(HAVE_MUELU_CUSPARSE)
-    typedef CuSparse_SpmV_Pack<SC,LO,GO,Node> CuSparse_thing_t;
+#if defined(HAVE_MUELU_CUSPARSE)
+    typedef CuSparse_SpmV_Pack<SC, LO, GO, Node> CuSparse_thing_t;
     CuSparse_thing_t cusparse_spmv(*At, xt, yt);
-  #endif
+#endif
 
-  #if defined(HAVE_MUELU_MAGMASPARSE)
-    typedef MagmaSparse_SpmV_Pack<SC,LO,GO,Node> MagmaSparse_thing_t;
+#if defined(HAVE_MUELU_MAGMASPARSE)
+    typedef MagmaSparse_SpmV_Pack<SC, LO, GO, Node> MagmaSparse_thing_t;
     MagmaSparse_thing_t magmasparse_spmv(*At, xt, yt);
-  #endif
-  #if defined(HAVE_MUELU_HYPRE) && defined(HAVE_MPI)
-    typedef HYPRE_SpmV_Pack<SC,LO,GO,Node> HYPRE_thing_t;
+#endif
+#if defined(HAVE_MUELU_HYPRE) && defined(HAVE_MPI)
+    typedef HYPRE_SpmV_Pack<SC, LO, GO, Node> HYPRE_thing_t;
     HYPRE_thing_t hypre_spmv(*At, xt, yt);
-  #endif
-  #if defined(HAVE_MUELU_PETSC) && defined(HAVE_MPI)
-    typedef Petsc_SpmV_Pack<SC,LO,GO,Node> Petsc_thing_t;
+#endif
+#if defined(HAVE_MUELU_PETSC) && defined(HAVE_MPI)
+    typedef Petsc_SpmV_Pack<SC, LO, GO, Node> Petsc_thing_t;
     Petsc_thing_t petsc_spmv(*At, xt, yt);
-  #endif
-
+#endif
 
-  #if defined(HAVE_MUELU_MKL)
+#if defined(HAVE_MUELU_MKL)
     // typedefs shared among other TPLs
-    typedef typename crs_matrix_type::local_matrix_host_type    KCRS;
-    typedef typename KCRS::StaticCrsGraphType              graph_t;
+    typedef typename crs_matrix_type::local_matrix_host_type KCRS;
+    typedef typename KCRS::StaticCrsGraphType graph_t;
     typedef typename graph_t::row_map_type::non_const_type lno_view_t;
-    typedef typename graph_t::row_map_type::const_type     c_lno_view_t;
+    typedef typename graph_t::row_map_type::const_type c_lno_view_t;
     typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
-    typedef typename graph_t::entries_type::const_type     c_lno_nnz_view_t;
-    typedef typename KCRS::values_type::non_const_type     scalar_view_t;
+    typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+    typedef typename KCRS::values_type::non_const_type scalar_view_t;
     typedef typename Node::device_type device_type;
 
     // data access common to other TPLs
-    const KCRS & Amat = At->getLocalMatrixHost();
-    c_lno_view_t Arowptr = Amat.graph.row_map;
-    c_lno_nnz_view_t Acolind = Amat.graph.entries;
+    const KCRS& Amat          = At->getLocalMatrixHost();
+    c_lno_view_t Arowptr      = Amat.graph.row_map;
+    c_lno_nnz_view_t Acolind  = Amat.graph.entries;
     const scalar_view_t Avals = Amat.values;
 
     // MKL specific things
     sparse_matrix_t mkl_A;
-    typedef typename Kokkos::View<MKL_INT*,typename lno_nnz_view_t::array_layout,typename lno_nnz_view_t::device_type> mkl_int_type;
+    typedef typename Kokkos::View<MKL_INT*, typename lno_nnz_view_t::array_layout, typename lno_nnz_view_t::device_type> mkl_int_type;
 
     mkl_int_type ArowptrMKL("Arowptr", Arowptr.extent(0));
     mkl_int_type AcolindMKL("Acolind", Acolind.extent(0));
-    copy_view(Arowptr,ArowptrMKL);
-    copy_view(Acolind,AcolindMKL);
-    double * mkl_xdouble = nullptr;
-    double * mkl_ydouble = nullptr;
-    mkl_descr.type = SPARSE_MATRIX_TYPE_GENERAL;
+    copy_view(Arowptr, ArowptrMKL);
+    copy_view(Acolind, AcolindMKL);
+    double* mkl_xdouble = nullptr;
+    double* mkl_ydouble = nullptr;
+    mkl_descr.type      = SPARSE_MATRIX_TYPE_GENERAL;
 
-    if(std::is_same<Scalar,double>::value) {
+    if (std::is_same<Scalar, double>::value) {
       mkl_sparse_d_create_csr(&mkl_A,
                               SPARSE_INDEX_BASE_ZERO,
                               At->getLocalNumRows(),
                               At->getLocalNumCols(),
                               ArowptrMKL.data(),
-                              ArowptrMKL.data()+1,
+                              ArowptrMKL.data() + 1,
                               AcolindMKL.data(),
                               (double*)Avals.data());
-    }
-    else
+    } else
       throw std::runtime_error("MKL Type Mismatch");
 
-  #endif // end MKL
-
+#endif  // end MKL
 
     //    globalTimeMonitor = Teuchos::null;
     comm->barrier();
 
     out << "Matrix Read complete." << endl;
     if (describeMatrix) {
-        out << "Matrix A:" << endl
-            << *A
-            << "========================================================" << endl;
+      out << "Matrix A:" << endl
+          << *A
+          << "========================================================" << endl;
     }
     // save std::cout formating
     std::ios cout_default_fmt_flags(NULL);
@@ -1095,164 +1069,146 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // random source
     std::random_device rd;
-    std::mt19937 random_source (rd());
+    std::mt19937 random_source(rd());
 
     // compute the baseline
     vector_type yt_baseline = Xpetra::toTpetra(*y_baseline);
-    if (report_error_norms) MV_Tpetra(*At,xt,yt_baseline);
+    if (report_error_norms) MV_Tpetra(*At, xt, yt_baseline);
     const bool error_check_y = true;
-    std::vector<typename Teuchos::ScalarTraits< Scalar >::magnitudeType> dummy;
+    std::vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> dummy;
     dummy.resize(1);
-    Teuchos::ArrayView< typename Teuchos::ScalarTraits< Scalar >::magnitudeType > y_norms (dummy.data(), 1);
+    Teuchos::ArrayView<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> y_norms(dummy.data(), 1);
 
     // no need for a barrier, because the randomization process uses a collective.
-    if ( !my_experiments.empty() ) {
+    if (!my_experiments.empty()) {
+      for (int i = 0; i < nrepeat; i++) {
+        // randomize the experiments
+        if (comm->getRank() == 0) {
+          std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
+        }
 
+        // Broadcast this ordering to the other processes
+        comm->broadcast(0,
+                        static_cast<int>(sizeof(Experiments::TPETRA) * my_experiments.size()),
+                        reinterpret_cast<char*>(my_experiments.data()));
+
+        // loop over the randomized experiments
+        for (const auto& experiment_id : my_experiments) {
+          switch (experiment_id) {
+#ifdef HAVE_MUELU_MKL
+            // MKL
+            case Experiments::MKL: {
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV MKL: Total"));
+              auto X_lcl  = xt.getLocalViewDevice(Tpetra::Access::ReadOnly);
+              auto Y_lcl  = yt.getLocalViewDevice(Tpetra::Access::OverwriteAll);
+              mkl_xdouble = (double*)X_lcl.data();
+              mkl_ydouble = (double*)Y_lcl.data();
+              MV_MKL(mkl_A, mkl_xdouble, mkl_ydouble);
+            } break;
+#endif
 
-    for (int i=0; i<nrepeat; i++) {
+            // KK Algorithms
+            case Experiments::KK: {
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV KK: Total"));
+              MV_KK(Att, xt, yt);
+            } break;
+            // Tpetra
+            case Experiments::TPETRA: {
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV Tpetra: Total"));
+              MV_Tpetra(*At, xt, yt);
+            } break;
+
+#ifdef HAVE_MUELU_CUSPARSE
+            // CUSPARSE
+            case Experiments::CUSPARSE: {
+              const Scalar alpha = 1.0;
+              const Scalar beta  = 0.0;
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV CuSparse: Total"));
+              cusparse_spmv.spmv(alpha, beta);
+            } break;
+#endif
 
-      // randomize the experiments
-      if (comm->getRank() == 0 ) {
-        std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
-      }
+#ifdef HAVE_MUELU_MAGMASPARSE
+            // Magma CSR
+            case Experiments::MAGMASPARSE: {
+              const Scalar alpha = 1.0;
+              const Scalar beta  = 0.0;
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV MagmaSparse: Total"));
+              magmasparse_spmv.spmv(alpha, beta);
+            } break;
+#endif
 
-      // Broadcast this ordering to the other processes
-      comm->broadcast(0,
-                      static_cast<int>(sizeof(Experiments::TPETRA)*my_experiments.size()),
-                      reinterpret_cast<char *>(my_experiments.data()));
-
-      // loop over the randomized experiments
-      for (const auto& experiment_id : my_experiments) {
-        switch (experiment_id) {
-
-        #ifdef HAVE_MUELU_MKL
-        // MKL
-        case Experiments::MKL:
-        {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MV MKL: Total"));
-            auto X_lcl = xt.getLocalViewDevice(Tpetra::Access::ReadOnly);
-            auto Y_lcl = yt.getLocalViewDevice(Tpetra::Access::OverwriteAll);
-            mkl_xdouble = (double*)X_lcl.data();
-            mkl_ydouble = (double*)Y_lcl.data();
-            MV_MKL(mkl_A,mkl_xdouble,mkl_ydouble);
-        }
-          break;
-        #endif
-
-        // KK Algorithms
-        case Experiments::KK:
-        {
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV KK: Total"));
-           MV_KK(Att,xt,yt);
-        }
-          break;
-        // Tpetra
-        case Experiments::TPETRA:
-        {
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV Tpetra: Total"));
-           MV_Tpetra(*At,xt,yt);
-        }
-          break;
-
-        #ifdef HAVE_MUELU_CUSPARSE
-        // CUSPARSE
-        case Experiments::CUSPARSE:
-        {
-           const Scalar alpha = 1.0;
-           const Scalar beta = 0.0;
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV CuSparse: Total"));
-           cusparse_spmv.spmv(alpha,beta);
-        }
-          break;
-        #endif
-
-        #ifdef HAVE_MUELU_MAGMASPARSE
-        // Magma CSR
-        case Experiments::MAGMASPARSE:
-        {
-           const Scalar alpha = 1.0;
-           const Scalar beta = 0.0;
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV MagmaSparse: Total"));
-           magmasparse_spmv.spmv(alpha,beta);
-        }
-          break;
-        #endif
-
-        #ifdef HAVE_MUELU_HYPRE
-        // HYPRE
-        case Experiments::HYPRE:
-        {
-           const Scalar alpha = 1.0;
-           const Scalar beta = 0.0;
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV HYPRE: Total"));
-           hypre_spmv.spmv(alpha,beta);
-        }
-          break;
-        #endif
-
-        #ifdef HAVE_MUELU_PETSC
-        // PETSc
-        case Experiments::PETSC:
-        {
-           const Scalar alpha = 1.0;
-           const Scalar beta = 0.0;
-           TimeMonitor t(*TimeMonitor::getNewTimer("MV Petsc: Total"));
-           petsc_spmv.spmv(alpha,beta);
-        }
-          break;
-        #endif
+#ifdef HAVE_MUELU_HYPRE
+            // HYPRE
+            case Experiments::HYPRE: {
+              const Scalar alpha = 1.0;
+              const Scalar beta  = 0.0;
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV HYPRE: Total"));
+              hypre_spmv.spmv(alpha, beta);
+            } break;
+#endif
 
-        default:
-          std::cerr << "Unknown experiment ID encountered: " << (int) experiment_id << std::endl;
-        }
-        //TODO: add a correctness check
-        // For now, all things alias x/y, so we can test yt (flakey and scary, but you only live once)
-        if (error_check_y && report_error_norms) {
-          // compute ||y||_2
-          y_norms[0] = -1;
-          y->norm2(y_norms);
-          const auto y_norm2 = y_norms[0];
-
-          y_norms[0] = -1;
-          yt.norm2(y_norms);
-          const auto y_mv_norm2 = y_norms[0];
-
-          y->update(-Teuchos::ScalarTraits<Scalar>::one(), *y_baseline, Teuchos::ScalarTraits<Scalar>::one());
-
-          y_norms[0] = -1;
-          y->norm2(y_norms);
-          const auto y_err =  y_norms[0];
-
-          y->putScalar(Teuchos::ScalarTraits<Scalar>::nan());;
-
-          y_norms[0] = -1;
-          y_baseline->norm2(y_norms);
-          const auto y_baseline_norm2 = y_norms[0];
-
-          y_norms[0] = -1;
-          yt.norm2(y_norms);
-          const auto y_mv_norm2_next_itr = y_norms[0];
-
-          std::cout << "ExperimentID: " << experiment_id_to_string[(int) experiment_id] << ", ||y-y_hat||_2 = "
-                    << std::setprecision(std::numeric_limits<Scalar>::digits10 + 1)
-                    << std::scientific << y_err
-                    << ", ||y||_2 = " << y_norm2
-                    << ", ||y_baseline||_2 = " << y_baseline_norm2
-                    << ", ||y_ptr|| == ||y_mv||:  " << std::boolalpha << (y_mv_norm2 == y_norm2)
-                    << ", setting y to nan ... ||y||_2 for next iter: " << y_mv_norm2_next_itr
-                    << "\n";
-        }
-        
-        // We need to both fence and barrier to make sure the kernels do not overlap
-        Kokkos::fence();
-        comm->barrier();
-      }// end random exp loop
-    } // end repeat
-    } // end ! my_experiments.empty()
+#ifdef HAVE_MUELU_PETSC
+            // PETSc
+            case Experiments::PETSC: {
+              const Scalar alpha = 1.0;
+              const Scalar beta  = 0.0;
+              TimeMonitor t(*TimeMonitor::getNewTimer("MV Petsc: Total"));
+              petsc_spmv.spmv(alpha, beta);
+            } break;
+#endif
+
+            default:
+              std::cerr << "Unknown experiment ID encountered: " << (int)experiment_id << std::endl;
+          }
+          // TODO: add a correctness check
+          //  For now, all things alias x/y, so we can test yt (flakey and scary, but you only live once)
+          if (error_check_y && report_error_norms) {
+            // compute ||y||_2
+            y_norms[0] = -1;
+            y->norm2(y_norms);
+            const auto y_norm2 = y_norms[0];
+
+            y_norms[0] = -1;
+            yt.norm2(y_norms);
+            const auto y_mv_norm2 = y_norms[0];
+
+            y->update(-Teuchos::ScalarTraits<Scalar>::one(), *y_baseline, Teuchos::ScalarTraits<Scalar>::one());
+
+            y_norms[0] = -1;
+            y->norm2(y_norms);
+            const auto y_err = y_norms[0];
+
+            y->putScalar(Teuchos::ScalarTraits<Scalar>::nan());
+            ;
+
+            y_norms[0] = -1;
+            y_baseline->norm2(y_norms);
+            const auto y_baseline_norm2 = y_norms[0];
+
+            y_norms[0] = -1;
+            yt.norm2(y_norms);
+            const auto y_mv_norm2_next_itr = y_norms[0];
+
+            std::cout << "ExperimentID: " << experiment_id_to_string[(int)experiment_id] << ", ||y-y_hat||_2 = "
+                      << std::setprecision(std::numeric_limits<Scalar>::digits10 + 1)
+                      << std::scientific << y_err
+                      << ", ||y||_2 = " << y_norm2
+                      << ", ||y_baseline||_2 = " << y_baseline_norm2
+                      << ", ||y_ptr|| == ||y_mv||:  " << std::boolalpha << (y_mv_norm2 == y_norm2)
+                      << ", setting y to nan ... ||y||_2 for next iter: " << y_mv_norm2_next_itr
+                      << "\n";
+          }
+
+          // We need to both fence and barrier to make sure the kernels do not overlap
+          Kokkos::fence();
+          comm->barrier();
+        }  // end random exp loop
+      }    // end repeat
+    }      // end ! my_experiments.empty()
     // restore the IO stream
     std::cout.copyfmt(cout_default_fmt_flags);
 
-
     if (useStackedTimer) {
       stacked_timer->stop("MueLu_MatvecKernelDriver");
       Teuchos::StackedTimer::OutputOptions options;
@@ -1269,15 +1225,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // ==========================================
     // Performance Models
     // ==========================================
-    std::vector<const char *> timer_names = {"MV MKL: Total",
-                                             "MV KK: Total",
-                                             "MV Tpetra: Total",
-                                             "MV CuSparse: Total",
-                                             "MV MagmaSparse: Total",
-                                             "MV HYPRE: Total",
-                                             "MV Petsc: Total"};
-    
-    MueLu::report_spmv_performance_models<Matrix>(A,nrepeat,timer_names,globalTimeMonitor,"",verboseModel);
+    std::vector<const char*> timer_names = {"MV MKL: Total",
+                                            "MV KK: Total",
+                                            "MV Tpetra: Total",
+                                            "MV CuSparse: Total",
+                                            "MV MagmaSparse: Total",
+                                            "MV HYPRE: Total",
+                                            "MV Petsc: Total"};
+
+    MueLu::report_spmv_performance_models<Matrix>(A, nrepeat, timer_names, globalTimeMonitor, "", verboseModel);
     globalTimeMonitor = Teuchos::null;
 
     success = true;
@@ -1288,18 +1244,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   PetscFinalize();
 #endif
 
-
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
-
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/scaling/MueLu_PerfModelReporter.hpp b/packages/muelu/test/scaling/MueLu_PerfModelReporter.hpp
index 2fe54531ed10..50f2d28817c1 100644
--- a/packages/muelu/test/scaling/MueLu_PerfModelReporter.hpp
+++ b/packages/muelu/test/scaling/MueLu_PerfModelReporter.hpp
@@ -54,173 +54,171 @@
 
 namespace MueLu {
 
-
-
 // =========================================================================
 // Performance Routines
 // =========================================================================
 // Report bandwidth in GB / sec
 
-
 /* @brief Generate performance model reports for SPMV of a given matrix
    Input:
      A       - Matrix to evaluate
      nrepeat - # times to repeat the tests in model construction
      timer_names - Names of Teuchos timers which are SPMVs to compare agains the model
-     verbose - Enable verbose output  
+     verbose - Enable verbose output
 */
-template<class Matrix>
-void report_spmv_performance_models(const Teuchos::RCP<const Matrix> & A, int nrepeat,const std::vector<const char *> & timer_names , Teuchos::RCP<Teuchos::TimeMonitor> & myTimeMonitor, const std::string prefix = "", bool verbose=false) {  
+template <class Matrix>
+void report_spmv_performance_models(const Teuchos::RCP<const Matrix> &A, int nrepeat, const std::vector<const char *> &timer_names, Teuchos::RCP<Teuchos::TimeMonitor> &myTimeMonitor, const std::string prefix = "", bool verbose = false) {
   using Teuchos::RCP;
   using Teuchos::rcp;
   const RCP<const Teuchos::Comm<int> > comm = A->getMap()->getComm();
-  using SC = typename Matrix::scalar_type;
-  using LO = typename Matrix::local_ordinal_type;
-  using GO = typename Matrix::global_ordinal_type;
-  using NO = typename Matrix::node_type;
+  using SC                                  = typename Matrix::scalar_type;
+  using LO                                  = typename Matrix::local_ordinal_type;
+  using GO                                  = typename Matrix::global_ordinal_type;
+  using NO                                  = typename Matrix::node_type;
 
   const double GB = 1024.0 * 1024.0 * 1024.0;
 
- 
   // Conversion function as a lambda
   auto convert_time_to_bandwidth_gbs = [=](double time, int num_calls, double memory_per_call_bytes) {
     double time_per_call = time / num_calls;
 
     return memory_per_call_bytes / GB / time_per_call;
   };
-  
+
   // NOTE: We've hardwired this to size_t for the rowptr.  This really should really get read out of a typedef,
   // if Tpetra actually had one
   using rowptr_type = size_t;
 
   // Make a new model if we need one
-  MueLu::PerfModels<SC,LO,GO,NO> PM;
+  MueLu::PerfModels<SC, LO, GO, NO> PM;
 
-  int rank = comm->getRank(); int nproc = comm->getSize();
-  int m   = static_cast<int>(A->getLocalNumRows());
-  int n   = static_cast<int>(A->getColMap()->getLocalNumElements());
-  int nnz = static_cast<int>(A->getLocalMatrixHost().graph.entries.extent(0));
-  
-  // Generate Lookup Tables  
-  int v_log_max = ceil(log(nnz) / log(2))+1;
-  PM.stream_vector_make_table(nrepeat,v_log_max);
+  int rank  = comm->getRank();
+  int nproc = comm->getSize();
+  int m     = static_cast<int>(A->getLocalNumRows());
+  int n     = static_cast<int>(A->getColMap()->getLocalNumElements());
+  int nnz   = static_cast<int>(A->getLocalMatrixHost().graph.entries.extent(0));
+
+  // Generate Lookup Tables
+  int v_log_max = ceil(log(nnz) / log(2)) + 1;
+  PM.stream_vector_make_table(nrepeat, v_log_max);
 
   int m_log_max = 15;
-  PM.pingpong_make_table(nrepeat,m_log_max,comm);
+  PM.pingpong_make_table(nrepeat, m_log_max, comm);
 
   std::string rank_prefix = prefix + std::to_string(rank) + std::string(": ");
 
-
-  if(A->hasCrsGraph()) {
+  if (A->hasCrsGraph()) {
     auto importer = A->getCrsGraph()->getImporter();
-    if(!importer.is_null()) {
-      size_t recv_size = importer->getRemoteLIDs().size() * sizeof(SC);
-      size_t send_size = importer->getExportLIDs().size() * sizeof(SC);
-      int local_log_max = ceil(log(std::max(send_size,recv_size)) / log(2))+1;
-      int global_log_max=local_log_max;
-      Teuchos::reduceAll<int>(*comm,Teuchos::REDUCE_MAX,1,&local_log_max,&global_log_max);
-      PM.halopong_make_table(nrepeat,global_log_max, importer);   
+    if (!importer.is_null()) {
+      size_t recv_size   = importer->getRemoteLIDs().size() * sizeof(SC);
+      size_t send_size   = importer->getExportLIDs().size() * sizeof(SC);
+      int local_log_max  = ceil(log(std::max(send_size, recv_size)) / log(2)) + 1;
+      int global_log_max = local_log_max;
+      Teuchos::reduceAll<int>(*comm, Teuchos::REDUCE_MAX, 1, &local_log_max, &global_log_max);
+      PM.halopong_make_table(nrepeat, global_log_max, importer);
     }
   }
 
-  if(verbose) {
-    std::cout<<prefix<<"********************************************************"<<std::endl;
-    std::cout<<prefix<<"Performance model results on "<<nproc<<" ranks"<<std::endl;
-    if(rank==0) {
-      std::cout<<rank_prefix<<" ****** Launch Latency Table ******"<<std::endl;
-      PM.print_launch_latency_table(std::cout,rank_prefix);
-      std::cout<<rank_prefix<<"****** Stream Table ******"<<std::endl;
-      PM.print_stream_vector_table(std::cout,rank_prefix);
-      std::cout<<rank_prefix<<"****** Latency Corrected Stream Table ******"<<std::endl;
-      PM.print_latency_corrected_stream_vector_table(std::cout,rank_prefix);
-      std::cout<<rank_prefix<<"****** Pingpong Table ******"<<std::endl;
-      PM.print_pingpong_table(std::cout,rank_prefix);
-      std::cout<<rank_prefix<<"****** Halopong Table ******"<<std::endl;
-      PM.print_halopong_table(std::cout,rank_prefix);
+  if (verbose) {
+    std::cout << prefix << "********************************************************" << std::endl;
+    std::cout << prefix << "Performance model results on " << nproc << " ranks" << std::endl;
+    if (rank == 0) {
+      std::cout << rank_prefix << " ****** Launch Latency Table ******" << std::endl;
+      PM.print_launch_latency_table(std::cout, rank_prefix);
+      std::cout << rank_prefix << "****** Stream Table ******" << std::endl;
+      PM.print_stream_vector_table(std::cout, rank_prefix);
+      std::cout << rank_prefix << "****** Latency Corrected Stream Table ******" << std::endl;
+      PM.print_latency_corrected_stream_vector_table(std::cout, rank_prefix);
+      std::cout << rank_prefix << "****** Pingpong Table ******" << std::endl;
+      PM.print_pingpong_table(std::cout, rank_prefix);
+      std::cout << rank_prefix << "****** Halopong Table ******" << std::endl;
+      PM.print_halopong_table(std::cout, rank_prefix);
     }
   }
 
   // For convenience
-  const int NUM_TIMERS = 6;
-  std::string SPMV_test_names[NUM_TIMERS] = {"colind","rowptr","vals","x","y","all"};
+  const int NUM_TIMERS                    = 6;
+  std::string SPMV_test_names[NUM_TIMERS] = {"colind", "rowptr", "vals", "x", "y", "all"};
   std::vector<int> SPMV_num_objects(NUM_TIMERS), SPMV_object_size(NUM_TIMERS), SPMV_corrected(NUM_TIMERS);
 
-
   // Composite model: Use latency correction
-  SPMV_num_objects[0] = nnz;     SPMV_object_size[0] = sizeof(LO);           SPMV_corrected[0] = 1;// colind
-  SPMV_num_objects[1] = (m + 1); SPMV_object_size[1] = sizeof(rowptr_type);  SPMV_corrected[1] = 1;// rowptr 
-  SPMV_num_objects[2] = nnz;     SPMV_object_size[2] = sizeof(SC);           SPMV_corrected[2] = 1;// vals
-  SPMV_num_objects[3] = n;       SPMV_object_size[3] = sizeof(SC);           SPMV_corrected[3] = 1; // x
-  SPMV_num_objects[4] = m;       SPMV_object_size[4] = sizeof(SC);           SPMV_corrected[4] = 1;// y
+  SPMV_num_objects[0] = nnz;
+  SPMV_object_size[0] = sizeof(LO);
+  SPMV_corrected[0]   = 1;  // colind
+  SPMV_num_objects[1] = (m + 1);
+  SPMV_object_size[1] = sizeof(rowptr_type);
+  SPMV_corrected[1]   = 1;  // rowptr
+  SPMV_num_objects[2] = nnz;
+  SPMV_object_size[2] = sizeof(SC);
+  SPMV_corrected[2]   = 1;  // vals
+  SPMV_num_objects[3] = n;
+  SPMV_object_size[3] = sizeof(SC);
+  SPMV_corrected[3]   = 1;  // x
+  SPMV_num_objects[4] = m;
+  SPMV_object_size[4] = sizeof(SC);
+  SPMV_corrected[4]   = 1;  // y
 
   // All-Model: Do not use latency correction
   SPMV_object_size[5] = 1;
-  SPMV_num_objects[5]  = (m+1)*sizeof(rowptr_type) + nnz*sizeof(LO) + nnz*sizeof(SC) + 
-    n*sizeof(SC) + m*sizeof(SC);  
+  SPMV_num_objects[5] = (m + 1) * sizeof(rowptr_type) + nnz * sizeof(LO) + nnz * sizeof(SC) +
+                        n * sizeof(SC) + m * sizeof(SC);
   SPMV_corrected[5] = 0;
 
-
   std::vector<double> gb_per_sec(NUM_TIMERS);
-  if(verbose && rank == 0)
-    std::cout<<rank_prefix<<"****** Local Time Model Results ******"<<std::endl;
-  for(int i = 0; i < NUM_TIMERS; i++) {
+  if (verbose && rank == 0)
+    std::cout << rank_prefix << "****** Local Time Model Results ******" << std::endl;
+  for (int i = 0; i < NUM_TIMERS; i++) {
     double avg_time;
 
     // Table interpolation - Take the faster of the two
-    int size_in_bytes = SPMV_object_size[i] *SPMV_num_objects[i];
-    if(SPMV_corrected[i] == 1)
+    int size_in_bytes = SPMV_object_size[i] * SPMV_num_objects[i];
+    if (SPMV_corrected[i] == 1)
       avg_time = PM.latency_corrected_stream_vector_lookup(size_in_bytes);
     else
       avg_time = PM.stream_vector_lookup(size_in_bytes);
 
-    
     // The lookup divides by transactions-per-element already
-    double memory_traffic = (double)SPMV_object_size[i] *(double)SPMV_num_objects[i];
-    gb_per_sec[i] = convert_time_to_bandwidth_gbs(avg_time,1,memory_traffic);
-    
-    if(verbose) {
-      std::cout<<rank_prefix<< "Local: "<<SPMV_test_names[i] << " # Scalars = "<<memory_traffic/sizeof(SC) << " time per call = "<<avg_time*1e6 << " us. GB/sec = "<<gb_per_sec[i]<<std::endl;
+    double memory_traffic = (double)SPMV_object_size[i] * (double)SPMV_num_objects[i];
+    gb_per_sec[i]         = convert_time_to_bandwidth_gbs(avg_time, 1, memory_traffic);
+
+    if (verbose) {
+      std::cout << rank_prefix << "Local: " << SPMV_test_names[i] << " # Scalars = " << memory_traffic / sizeof(SC) << " time per call = " << avg_time * 1e6 << " us. GB/sec = " << gb_per_sec[i] << std::endl;
     }
   }
-  
-  // Get the latency info
-  double avg_latency = PM.launch_latency_lookup();    
-
 
+  // Get the latency info
+  double avg_latency = PM.launch_latency_lookup();
 
   /***************************************************************************/
   // *** Calculate SPMV minimum time (composite) ***
-  // Model: 
+  // Model:
   // rowptr = One read per row
   // colind = One read per entry
   // values = One read per entry
   // x      = One read per entry in values array (but we assume the cache will work its magic here)
   // y      = One write per row
- 
-  long unsigned int spmv_memory_bytes[NUM_TIMERS] = {
-     (m+1) * sizeof(rowptr_type), //rowptr
-     nnz * sizeof(LO), //colind
-     nnz * sizeof(SC), // values
-     n  * sizeof(SC), //x
-     m * sizeof(SC), // y
-     0
-    };
-  for(int i=0; i<NUM_TIMERS-1; i++)
-    spmv_memory_bytes[NUM_TIMERS-1] += spmv_memory_bytes[i];
 
+  long unsigned int spmv_memory_bytes[NUM_TIMERS] = {
+      (m + 1) * sizeof(rowptr_type),  // rowptr
+      nnz * sizeof(LO),               // colind
+      nnz * sizeof(SC),               // values
+      n * sizeof(SC),                 // x
+      m * sizeof(SC),                 // y
+      0};
+  for (int i = 0; i < NUM_TIMERS - 1; i++)
+    spmv_memory_bytes[NUM_TIMERS - 1] += spmv_memory_bytes[i];
 
   double minimum_local_composite_time = avg_latency;
-  for(int i=0; i<NUM_TIMERS-1; i++)
-    minimum_local_composite_time +=  spmv_memory_bytes[i] / (GB * gb_per_sec[i]);
+  for (int i = 0; i < NUM_TIMERS - 1; i++)
+    minimum_local_composite_time += spmv_memory_bytes[i] / (GB * gb_per_sec[i]);
 
-  double minimum_local_all_time = spmv_memory_bytes[NUM_TIMERS-1] / (GB * gb_per_sec[NUM_TIMERS-1]);
+  double minimum_local_all_time = spmv_memory_bytes[NUM_TIMERS - 1] / (GB * gb_per_sec[NUM_TIMERS - 1]);
 
-  if(verbose) {
-    std::cout<<rank_prefix<< "Local: composite      = "<<minimum_local_composite_time*1e6<<" us.\n"
-             <<rank_prefix<< "Local: all            = "<<minimum_local_all_time*1e6<<" us.\n";
+  if (verbose) {
+    std::cout << rank_prefix << "Local: composite      = " << minimum_local_composite_time * 1e6 << " us.\n"
+              << rank_prefix << "Local: all            = " << minimum_local_all_time * 1e6 << " us.\n";
   }
 
-
   /***************************************************************************/
   // *** Calculate Remote part of the SPMV ***
   double time_pack_unpack_outofplace = 0.0;
@@ -229,87 +227,78 @@ void report_spmv_performance_models(const Teuchos::RCP<const Matrix> & A, int nr
   double time_communicate_halo       = 0.0;
   // Note: We'll assume that each of the permutes, remotes and exports is a unified
   // memory transaction, even though that's not strictly speaking correct.
-  if(A->hasCrsGraph()) {
+  if (A->hasCrsGraph()) {
     auto importer = A->getCrsGraph()->getImporter();
-    if(! importer.is_null()) {
-      // Sames [pack] - 1 read SC, 1 write SC 
+    if (!importer.is_null()) {
+      // Sames [pack] - 1 read SC, 1 write SC
       // NOTE: Only if you're out-of-place
       size_t num_sames = importer->getNumSameIDs();
-      double same_time = (num_sames == 0) ? 0.0 : 
-        2.0 * PM.latency_corrected_stream_vector_lookup(num_sames*sizeof(SC)) + avg_latency;
+      double same_time = (num_sames == 0) ? 0.0 : 2.0 * PM.latency_corrected_stream_vector_lookup(num_sames * sizeof(SC)) + avg_latency;
 
       // Permutes [pack] - 2 reads LO [to, from] , 1 read SC [values], 1 write SC [values]
       size_t num_permutes = importer->getNumPermuteIDs();
-      double permute_time = (num_permutes == 0) ? 0.0 : 
-        2.0 * PM.latency_corrected_stream_vector_lookup(num_permutes*sizeof(LO)) + 
-        2.0 * PM.latency_corrected_stream_vector_lookup(num_permutes*sizeof(SC)) + avg_latency;
+      double permute_time = (num_permutes == 0) ? 0.0 : 2.0 * PM.latency_corrected_stream_vector_lookup(num_permutes * sizeof(LO)) + 2.0 * PM.latency_corrected_stream_vector_lookup(num_permutes * sizeof(SC)) + avg_latency;
 
       // Exports [pack] - 1 read LO [exportLIDs], 1 read SC [values], 1 write SC [buffer]
-      // This is what Epetra does at least      
+      // This is what Epetra does at least
       size_t num_exports = importer->getNumExportIDs();
-      double export_time = (num_exports == 0) ? 0.0 :
-        PM.latency_corrected_stream_vector_lookup(num_exports*sizeof(LO)) + 
-        2.0 * PM.latency_corrected_stream_vector_lookup(num_exports*sizeof(SC)) + avg_latency;
+      double export_time = (num_exports == 0) ? 0.0 : PM.latency_corrected_stream_vector_lookup(num_exports * sizeof(LO)) + 2.0 * PM.latency_corrected_stream_vector_lookup(num_exports * sizeof(SC)) + avg_latency;
 
       // Remotes [unpack] - 1 read LO [remoteLIDs],  1 read SC [buffer], 1 write SC [values]
       // NOTE: Only if you're out of place
       size_t num_remotes = importer->getNumRemoteIDs();
-      double remote_time = (num_remotes == 0) ? 0.0 :
-        PM.latency_corrected_stream_vector_lookup(num_remotes*sizeof(LO)) + 
-        2.0 * PM.latency_corrected_stream_vector_lookup(num_remotes*sizeof(SC)) + avg_latency;
-
+      double remote_time = (num_remotes == 0) ? 0.0 : PM.latency_corrected_stream_vector_lookup(num_remotes * sizeof(LO)) + 2.0 * PM.latency_corrected_stream_vector_lookup(num_remotes * sizeof(SC)) + avg_latency;
 
       // Total pack / unpack time
       time_pack_unpack_outofplace = same_time + permute_time + export_time + remote_time;
       time_pack_unpack_inplace    = permute_time + export_time;
 
       // We now need to get the size of each message for the ping-pong costs.
-      double send_time = 0.0;
-      double recv_time = 0.0;
-      double halo_time = 0.0;
-      size_t total_send_length=0, total_recv_length=0;
-      double avg_size_per_msg = 0.0;
-      RCP<const Xpetra::TpetraImport<LO,GO,NO> > t_importer = Teuchos::rcp_dynamic_cast<const Xpetra::TpetraImport<LO,GO,NO> >(importer);
-      if(!t_importer.is_null()) {
-        RCP<const Tpetra::Import<LO,GO,NO> > tt_i  = t_importer->getTpetra_Import();
-        Tpetra::Distributor & distor = tt_i->getDistributor();
+      double send_time         = 0.0;
+      double recv_time         = 0.0;
+      double halo_time         = 0.0;
+      size_t total_send_length = 0, total_recv_length = 0;
+      double avg_size_per_msg                                 = 0.0;
+      RCP<const Xpetra::TpetraImport<LO, GO, NO> > t_importer = Teuchos::rcp_dynamic_cast<const Xpetra::TpetraImport<LO, GO, NO> >(importer);
+      if (!t_importer.is_null()) {
+        RCP<const Tpetra::Import<LO, GO, NO> > tt_i   = t_importer->getTpetra_Import();
+        Tpetra::Distributor &distor                   = tt_i->getDistributor();
         Teuchos::ArrayView<const size_t> recv_lengths = distor.getLengthsFrom();
         Teuchos::ArrayView<const size_t> send_lengths = distor.getLengthsTo();
-        
-        for (int i=0; i<(int) send_lengths.size(); i++) {
+
+        for (int i = 0; i < (int)send_lengths.size(); i++) {
           send_time += PM.pingpong_device_lookup(send_lengths[i] * sizeof(SC));
-          total_send_length = send_lengths[i]*sizeof(SC);
+          total_send_length = send_lengths[i] * sizeof(SC);
         }
-        
-        for (int i=0; i<(int) recv_lengths.size(); i++)  {
-          recv_time += PM.pingpong_device_lookup(recv_lengths[i] * sizeof(SC));                  
-          total_recv_length = recv_lengths[i]*sizeof(SC);
+
+        for (int i = 0; i < (int)recv_lengths.size(); i++) {
+          recv_time += PM.pingpong_device_lookup(recv_lengths[i] * sizeof(SC));
+          total_recv_length = recv_lengths[i] * sizeof(SC);
         }
- 
-        avg_size_per_msg = (double)total_send_length/(2.0*send_lengths.size()) +  (double)total_recv_length/(2.0*recv_lengths.size());
-        halo_time = PM.halopong_device_lookup(avg_size_per_msg);
-      }
 
+        avg_size_per_msg = (double)total_send_length / (2.0 * send_lengths.size()) + (double)total_recv_length / (2.0 * recv_lengths.size());
+        halo_time        = PM.halopong_device_lookup(avg_size_per_msg);
+      }
 
       // NOTE: For now we'll do comm time as the larger of send/recv.  Not sure this is
       // really the optimal thing to do, but we'll start here.
-      time_communicate_ping = std::max(send_time,recv_time);
+      time_communicate_ping = std::max(send_time, recv_time);
       time_communicate_halo = halo_time;
 
-      if(verbose) {
-        std::cout<<rank_prefix<<"****** Remote Time Model Results ******"<<std::endl;
-        std::cout<<rank_prefix<< "Remote: same      = "<<same_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: permutes  = "<<permute_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: exports   = "<<export_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: remotes   = "<<remote_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: sends len = "<<total_send_length<<" time = "<<send_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: recvs len = "<<total_recv_length<<" time = "<<recv_time*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: pack/unpack OOP = "<< time_pack_unpack_outofplace*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: pack/unpack INP = "<< time_pack_unpack_inplace*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: ping avg  = "<<(size_t)avg_size_per_msg<<" time  = "<<time_communicate_ping*1e6<<" us.\n"
-                 <<rank_prefix<< "Remote: halo avg  = "<<(size_t)avg_size_per_msg<<" time  = "<<time_communicate_halo*1e6<<" us.\n"<<std::endl;
+      if (verbose) {
+        std::cout << rank_prefix << "****** Remote Time Model Results ******" << std::endl;
+        std::cout << rank_prefix << "Remote: same      = " << same_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: permutes  = " << permute_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: exports   = " << export_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: remotes   = " << remote_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: sends len = " << total_send_length << " time = " << send_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: recvs len = " << total_recv_length << " time = " << recv_time * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: pack/unpack OOP = " << time_pack_unpack_outofplace * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: pack/unpack INP = " << time_pack_unpack_inplace * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: ping avg  = " << (size_t)avg_size_per_msg << " time  = " << time_communicate_ping * 1e6 << " us.\n"
+                  << rank_prefix << "Remote: halo avg  = " << (size_t)avg_size_per_msg << " time  = " << time_communicate_halo * 1e6 << " us.\n"
+                  << std::endl;
       }
-      
     }
   }
   double minimum_time_in_place_ping     = time_communicate_ping + time_pack_unpack_inplace;
@@ -317,12 +306,10 @@ void report_spmv_performance_models(const Teuchos::RCP<const Matrix> & A, int nr
   double minimum_time_in_place_halo     = time_communicate_halo + time_pack_unpack_inplace;
   double minimum_time_out_of_place_halo = time_communicate_halo + time_pack_unpack_outofplace;
 
-
   /***************************************************************************/
- 
 
   // Get global average/max time sums
-  constexpr int NUM_TIMES=12;
+  constexpr int NUM_TIMES = 12;
   double avg_times[NUM_TIMES];
   double max_times[NUM_TIMES];
   {
@@ -334,63 +321,56 @@ void report_spmv_performance_models(const Teuchos::RCP<const Matrix> & A, int nr
     double h_ooplace = minimum_time_out_of_place_halo;
 
     double all_times_local[NUM_TIMES] = {
-      comp,comp+p_inplace,comp+p_ooplace,comp+h_inplace,comp+h_ooplace, //0-4
-      alls,alls+p_inplace,alls+p_ooplace,alls+h_inplace,alls+h_ooplace,//5-9
-      time_communicate_ping,time_communicate_halo };//10,11
+        comp, comp + p_inplace, comp + p_ooplace, comp + h_inplace, comp + h_ooplace,  // 0-4
+        alls, alls + p_inplace, alls + p_ooplace, alls + h_inplace, alls + h_ooplace,  // 5-9
+        time_communicate_ping, time_communicate_halo};                                 // 10,11
 
-    if(nproc > 1) {
+    if (nproc > 1) {
       Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, NUM_TIMES, &all_times_local[0], &avg_times[0]);
       Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, NUM_TIMES, &all_times_local[0], &max_times[0]);
-      for(int i=0; i<NUM_TIMES; i++)
+      for (int i = 0; i < NUM_TIMES; i++)
         avg_times[i] /= nproc;
-    }
-    else {
-      for(int i=0; i<NUM_TIMES; i++)
+    } else {
+      for (int i = 0; i < NUM_TIMES; i++)
         avg_times[i] = max_times[i] = all_times_local[i];
     }
 
-    if(rank == 0)
+    if (rank == 0)
       std::cout << "\n\n========================================================\n"
-                <<prefix<< "Minimum time model (composite) : " << avg_times[0]*1e6 << " us.\n"
-                <<prefix<< "Minimum time model (all)       : " << avg_times[5]*1e6 << " us.\n"
-                <<prefix<< "Pack/unpack in-place           : " << (avg_times[1] - avg_times[0] - avg_times[10])*1e6<< "us.\n" // comp+ping_pack_inplace - comp - ping
-                <<prefix<< "Pack/unpack out-of-place       : " << (avg_times[2] - avg_times[0] - avg_times[10])*1e6<< "us.\n" // comp+ping_pack_ooplace - comp - ping
-                <<prefix<< "Communication time (ping)      : " << avg_times[10]*1e6 << "us.\n"
-                <<prefix<< "Communication time (halo)      : " << avg_times[11]*1e6 << "us." <<std::endl;
-    
-
+                << prefix << "Minimum time model (composite) : " << avg_times[0] * 1e6 << " us.\n"
+                << prefix << "Minimum time model (all)       : " << avg_times[5] * 1e6 << " us.\n"
+                << prefix << "Pack/unpack in-place           : " << (avg_times[1] - avg_times[0] - avg_times[10]) * 1e6 << "us.\n"  // comp+ping_pack_inplace - comp - ping
+                << prefix << "Pack/unpack out-of-place       : " << (avg_times[2] - avg_times[0] - avg_times[10]) * 1e6 << "us.\n"  // comp+ping_pack_ooplace - comp - ping
+                << prefix << "Communication time (ping)      : " << avg_times[10] * 1e6 << "us.\n"
+                << prefix << "Communication time (halo)      : " << avg_times[11] * 1e6 << "us." << std::endl;
   }
 
   // Iterate through all of the "MV" timers
- 
-  if(rank == 0) {
-    if(!myTimeMonitor.is_null() && timer_names.size() > 0) {
-      const std::string l[NUM_TIMES]={"Comp","Comp+ping+inplace","Comp+ping+ooplace","Comp+halo+inplace","Comp+halo+ooplace",
-                                    "All", "All+ping+inplace", "All+ping+ooplace", "All+halo+inplace", "All+halo+ooplace"};
-      const std::string div={"-------------------"};
-      printf("%-60s %20s %20s %20s %20s %20s %20s %20s %20s %20s %20s\n","Timer",l[0].c_str(),l[1].c_str(),l[2].c_str(),l[3].c_str(),l[4].c_str(),
-             l[5].c_str(),l[6].c_str(),l[7].c_str(),l[8].c_str(),l[9].c_str());
-      printf("%-60s %20s %20s %20s %20s %20s %20s %20s %20s %20s %20s\n","-----",div.c_str(),div.c_str(),div.c_str(),div.c_str(),div.c_str(),
-             div.c_str(),div.c_str(),div.c_str(),div.c_str(),div.c_str());
-      for(int i=0; i<(int)timer_names.size(); i++) {
+
+  if (rank == 0) {
+    if (!myTimeMonitor.is_null() && timer_names.size() > 0) {
+      const std::string l[NUM_TIMES] = {"Comp", "Comp+ping+inplace", "Comp+ping+ooplace", "Comp+halo+inplace", "Comp+halo+ooplace",
+                                        "All", "All+ping+inplace", "All+ping+ooplace", "All+halo+inplace", "All+halo+ooplace"};
+      const std::string div          = {"-------------------"};
+      printf("%-60s %20s %20s %20s %20s %20s %20s %20s %20s %20s %20s\n", "Timer", l[0].c_str(), l[1].c_str(), l[2].c_str(), l[3].c_str(), l[4].c_str(),
+             l[5].c_str(), l[6].c_str(), l[7].c_str(), l[8].c_str(), l[9].c_str());
+      printf("%-60s %20s %20s %20s %20s %20s %20s %20s %20s %20s %20s\n", "-----", div.c_str(), div.c_str(), div.c_str(), div.c_str(), div.c_str(),
+             div.c_str(), div.c_str(), div.c_str(), div.c_str(), div.c_str());
+      for (int i = 0; i < (int)timer_names.size(); i++) {
         Teuchos::RCP<Teuchos::Time> t = myTimeMonitor->lookupCounter(timer_names[i]);
-        if(!t.is_null()) {
+        if (!t.is_null()) {
           double time_per_call = t->totalElapsedTime() / t->numCalls();
-          printf("%-60s %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f\n",timer_names[i],
-                 max_times[0]/time_per_call,max_times[1]/time_per_call,max_times[2]/time_per_call,max_times[3]/time_per_call,max_times[4]/time_per_call,
-                 max_times[5]/time_per_call,max_times[6]/time_per_call,max_times[7]/time_per_call,max_times[8]/time_per_call,max_times[9]/time_per_call);
-
+          printf("%-60s %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f %20.2f\n", timer_names[i],
+                 max_times[0] / time_per_call, max_times[1] / time_per_call, max_times[2] / time_per_call, max_times[3] / time_per_call, max_times[4] / time_per_call,
+                 max_times[5] / time_per_call, max_times[6] / time_per_call, max_times[7] / time_per_call, max_times[8] / time_per_call, max_times[9] / time_per_call);
         }
       }
-    }
-    else {
-      std::cout<<prefix<<"Note: Minimum time model individual timers only work with stacked timers off."<<std::endl;
+    } else {
+      std::cout << prefix << "Note: Minimum time model individual timers only work with stacked timers off." << std::endl;
     }
   }
-
 }
 
-
-}// end MueLu namespace
+}  // namespace MueLu
 
 #endif
diff --git a/packages/muelu/test/scaling/ReadMatrix.cpp b/packages/muelu/test/scaling/ReadMatrix.cpp
index df505df60fb3..be860708af8f 100644
--- a/packages/muelu/test/scaling/ReadMatrix.cpp
+++ b/packages/muelu/test/scaling/ReadMatrix.cpp
@@ -61,7 +61,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
@@ -72,22 +72,22 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -96,39 +96,52 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& fancyout = *fancy;
+    Teuchos::FancyOStream &fancyout  = *fancy;
     fancyout.setOutputToRootOnly(0);
 
-
     // =========================================================================
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName = "scalingTest.xml"; clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
-    int    amgAsPrecond     = 1;                 clp.setOption("precond",               &amgAsPrecond,     "apply multigrid as preconditioner");
-    int    amgAsSolver      = 0;                 clp.setOption("fixPoint",              &amgAsSolver,      "apply multigrid as solver");
-    bool   printTimings     = true;              clp.setOption("timings", "notimings",  &printTimings,     "print timings to screen");
-    int    writeMatricesOPT = -2;                clp.setOption("write",                 &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
-    double tol              = 1e-12;             clp.setOption("tol",                   &tol,              "solver convergence tolerance");
-    std::string krylovMethod = "cg"; clp.setOption("krylov",                   &krylovMethod,     "outer Krylov method");
-    int maxIts = 100; clp.setOption("maxits",           &maxIts,   "maximum number of Krylov iterations");
-    int output = 1; clp.setOption("output",           &output,   "how often to print Krylov residual history");
-    std::string matrixFileName = "A.mm"; clp.setOption("matrixfile",           &matrixFileName,   "matrix market file containing matrix");
-    std::string rhsFileName = ""; clp.setOption("rhsfile",           &rhsFileName,   "matrix market file containing right-hand side");
-    int nPDE = 1; clp.setOption("numpdes",           &nPDE,   "number of PDE equations");
-    std::string convType = "r0"; clp.setOption("convtype",                   &convType,     "convergence type (r0 or none)");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+    std::string xmlFileName = "scalingTest.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+    int amgAsPrecond = 1;
+    clp.setOption("precond", &amgAsPrecond, "apply multigrid as preconditioner");
+    int amgAsSolver = 0;
+    clp.setOption("fixPoint", &amgAsSolver, "apply multigrid as solver");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int writeMatricesOPT = -2;
+    clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+    double tol = 1e-12;
+    clp.setOption("tol", &tol, "solver convergence tolerance");
+    std::string krylovMethod = "cg";
+    clp.setOption("krylov", &krylovMethod, "outer Krylov method");
+    int maxIts = 100;
+    clp.setOption("maxits", &maxIts, "maximum number of Krylov iterations");
+    int output = 1;
+    clp.setOption("output", &output, "how often to print Krylov residual history");
+    std::string matrixFileName = "A.mm";
+    clp.setOption("matrixfile", &matrixFileName, "matrix market file containing matrix");
+    std::string rhsFileName = "";
+    clp.setOption("rhsfile", &rhsFileName, "matrix market file containing right-hand side");
+    int nPDE = 1;
+    clp.setOption("numpdes", &nPDE, "number of PDE equations");
+    std::string convType = "r0";
+    clp.setOption("convtype", &convType, "convergence type (r0 or none)");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    fancyout << "========================================================\n" << xpetraParameters << matrixParameters;
+    fancyout << "========================================================\n"
+             << xpetraParameters << matrixParameters;
 
     // =========================================================================
     // Problem construction
@@ -138,20 +151,20 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     comm->barrier();
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
-    RCP<Matrix> A = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileName), lib, comm);
-    RCP<const Map>   map = A->getRowMap();
-    RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(),nPDE);
-    //RCP<MultiVector> fakeCoordinates = MultiVectorFactory::Build(A->getDomainMap(),1);
+    RCP<Matrix> A              = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileName), lib, comm);
+    RCP<const Map> map         = A->getRowMap();
+    RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getDomainMap(), nPDE);
+    // RCP<MultiVector> fakeCoordinates = MultiVectorFactory::Build(A->getDomainMap(),1);
     A->SetFixedBlockSize(nPDE);
     std::cout << "#pdes = " << A->GetFixedBlockSize() << std::endl;
     if (nPDE == 1)
-      nullspace->putScalar( Teuchos::ScalarTraits<SC>::one() );
+      nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
     else {
-      for (int i=0; i<nPDE; ++i) {
+      for (int i = 0; i < nPDE; ++i) {
         Teuchos::ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
-        for (int j=0; j<nsData.size(); ++j) {
+        for (int j = 0; j < nsData.size(); ++j) {
           GO gel = A->getDomainMap()->getGlobalElement(j) - A->getDomainMap()->getIndexBase();
-          if ((gel-i) % nPDE == 0)
+          if ((gel - i) % nPDE == 0)
             nsData[j] = Teuchos::ScalarTraits<SC>::one();
         }
       }
@@ -179,9 +192,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // which is used instead
     H->SetDefaultVerbLevel(MueLu::Extreme);
 
-    H->GetLevel(0)->Set("A",           A);
-    H->GetLevel(0)->Set("Nullspace",   nullspace);
-    //H->GetLevel(0)->Set("Coordinates", fakeCoordinates);
+    H->GetLevel(0)->Set("A", A);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
+    // H->GetLevel(0)->Set("Coordinates", fakeCoordinates);
 
     mueLuFactory.SetupHierarchy(*H);
 
@@ -197,15 +210,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // =========================================================================
     comm->barrier();
     tm.reset();
-    tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 3 - LHS and RHS initialization")));
 
-    RCP<Vector> X = VectorFactory::Build(map,1);
-    RCP<MultiVector> B = VectorFactory::Build(map,1);
+    RCP<Vector> X      = VectorFactory::Build(map, 1);
+    RCP<MultiVector> B = VectorFactory::Build(map, 1);
 
     if (rhsFileName != "")
-      B = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(std::string(rhsFileName), A->getRowMap());
-    else
-    {
+      B = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(std::string(rhsFileName), A->getRowMap());
+    else {
       // we set seed for reproducibility
       X->setSeed(846930886);
       bool useSameRandomGen = false;
@@ -214,7 +226,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms(1);
       B->norm2(norms);
-      //B->scale(1.0/norms[0]);
+      // B->scale(1.0/norms[0]);
     }
     X->putScalar(zero);
     tm = Teuchos::null;
@@ -225,16 +237,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     comm->barrier();
     if (amgAsSolver) {
       tm.reset();
-      tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
+      tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
 
       H->IsPreconditioner(false);
       Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms(1);
-      norms = Utilities::ResidualNorm(*A,*X,*B);
+      norms = Utilities::ResidualNorm(*A, *X, *B);
       std::cout << "                iter:    0           residual = " << norms[0] << std::endl;
-      for (int i=0; i< maxIts; ++i) {
+      for (int i = 0; i < maxIts; ++i) {
         H->Iterate(*B, *X);
-        norms = Utilities::ResidualNorm(*A,*X,*B);
-        std::cout << "                iter:    " << i+1 << "           residual = " << norms[0] << std::endl;
+        norms = Utilities::ResidualNorm(*A, *X, *B);
+        std::cout << "                iter:    " << i + 1 << "           residual = " << norms[0] << std::endl;
       }
 
     } else if (amgAsPrecond) {
@@ -242,16 +254,16 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       tm.reset();
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Belos Solve")));
       // Operator and Multivector type that will be used with Belos
-      typedef MultiVector          MV;
+      typedef MultiVector MV;
       typedef Belos::OperatorT<MV> OP;
       H->IsPreconditioner(true);
 
       // Define Operator and Preconditioner
-      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
       // Construct a Belos LinearProblem object
-      RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+      RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
       belosProblem->setRightPrec(belosPrec);
 
       bool set = belosProblem->setProblem();
@@ -262,19 +274,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       // Belos parameter list
       Teuchos::ParameterList belosList;
-      belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-      belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-      belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-      belosList.set("Output Frequency",      output);
-      belosList.set("Output Style",          Belos::Brief);
-      //belosList.set("Orthogonalization",     "ICGS");
+      belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+      belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+      belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+      belosList.set("Output Frequency", output);
+      belosList.set("Output Style", Belos::Brief);
+      // belosList.set("Orthogonalization",     "ICGS");
       if (convType == "none") {
-        belosList.set("Explicit Residual Scaling",  "None");
-        belosList.set("Implicit Residual Scaling",  "None");
+        belosList.set("Explicit Residual Scaling", "None");
+        belosList.set("Implicit Residual Scaling", "None");
       }
 
       // Create an iterative solver manager
-      RCP< Belos::SolverManager<SC, MV, OP> > solver;
+      RCP<Belos::SolverManager<SC, MV, OP> > solver;
       if (krylovMethod == "cg") {
         solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
       } else if (krylovMethod == "gmres") {
@@ -291,19 +303,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         // Get the number of iterations for this solve.
         fancyout << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-      } catch(...) {
-        fancyout << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+      } catch (...) {
+        fancyout << std::endl
+                 << "ERROR:  Belos threw an error! " << std::endl;
       }
 
       // Check convergence
       if (ret != Belos::Converged)
-        fancyout << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+        fancyout << std::endl
+                 << "ERROR:  Belos did not converge! " << std::endl;
       else
-        fancyout << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-#endif //ifdef HAVE_MUELU_BELOS
+        fancyout << std::endl
+                 << "SUCCESS:  Belos converged!" << std::endl;
+#endif  // ifdef HAVE_MUELU_BELOS
     }
     comm->barrier();
-    tm = Teuchos::null;
+    tm                = Teuchos::null;
     globalTimeMonitor = Teuchos::null;
 
     if (printTimings) {
@@ -315,17 +330,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
-
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/scaling/Reuse.cpp b/packages/muelu/test/scaling/Reuse.cpp
index f781c9e4fa2b..dda6e15ba518 100644
--- a/packages/muelu/test/scaling/Reuse.cpp
+++ b/packages/muelu/test/scaling/Reuse.cpp
@@ -63,7 +63,7 @@
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
 #include <MueLu_Exceptions.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_MutuallyExclusiveTime.hpp>
 
@@ -71,31 +71,31 @@
 #include <BelosConfigDefs.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 //- -- --------------------------------------------------------
 template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using Teuchos::Array;
+  using Teuchos::ArrayRCP;
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::Time;
   using Teuchos::TimeMonitor;
-  using Teuchos::Array;
-  using Teuchos::ArrayRCP;
 
   typedef Teuchos::ScalarTraits<SC> STS;
 
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    int mypid = comm->getRank();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    int mypid                           = comm->getRank();
 
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream &out       = *fancy;
     out.setOutputToRootOnly(0);
 
     SC zero = STS::zero(), one = STS::one();
@@ -104,45 +104,54 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
     // Parameters
     //
 
-    Xpetra::Parameters            xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName  = "reuse.xml"; clp.setOption("xml",                   &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'reuse.xml'");
-    std::string matrixPrefix = "jac";       clp.setOption("matrix",               &matrixPrefix, "prefix for matrix file names.  Default = 'jac'");
-    bool        binary       = false;       clp.setOption("binary", "nobinary",         &binary, "matrix files are binary. Default = false");
-    std::string rhsPrefix    = "rhs";       clp.setOption("rhs" ,                    &rhsPrefix, "prefix for rhs file names. Default = 'rhs'");
-    bool        printTimings = true;        clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
-    int         first_matrix = 0;           clp.setOption("firstMatrix",          &first_matrix, "first matrix in the sequence to use");
-    int         last_matrix  = 1;           clp.setOption("lastMatrix",            &last_matrix, "last matrix in the sequence to use");
-    bool        inMemory     = false;       clp.setOption("inmemory", "noinmemory",   &inMemory, "load all matrices in memory");
-    bool        doReuse      = true;        clp.setOption("reuse",  "noreuse",         &doReuse, "if you want to try reuse");
-    const int   numPDEs      = 2;
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
+
+    std::string xmlFileName = "reuse.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'reuse.xml'");
+    std::string matrixPrefix = "jac";
+    clp.setOption("matrix", &matrixPrefix, "prefix for matrix file names.  Default = 'jac'");
+    bool binary = false;
+    clp.setOption("binary", "nobinary", &binary, "matrix files are binary. Default = false");
+    std::string rhsPrefix = "rhs";
+    clp.setOption("rhs", &rhsPrefix, "prefix for rhs file names. Default = 'rhs'");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int first_matrix = 0;
+    clp.setOption("firstMatrix", &first_matrix, "first matrix in the sequence to use");
+    int last_matrix = 1;
+    clp.setOption("lastMatrix", &last_matrix, "last matrix in the sequence to use");
+    bool inMemory = false;
+    clp.setOption("inmemory", "noinmemory", &inMemory, "load all matrices in memory");
+    bool doReuse = true;
+    clp.setOption("reuse", "noreuse", &doReuse, "if you want to try reuse");
+    const int numPDEs = 2;
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Global Time"))), tm;
     RCP<Time> timer;
 
     // Operator and Multivector type that will be used with Belos
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
 
     // Stats tracking
     int numSteps = last_matrix - first_matrix + 1;
-    Array<Array<int>    >       iteration_counts(numSteps);
-    Array<Array<double> >       iteration_times (numSteps);
-    Array<Array<double> >       setup_times     (numSteps);
-    Array<RCP<Matrix> >         matrices        (numSteps);
-    Array<RCP<MultiVector> >    rhss            (numSteps);
+    Array<Array<int> > iteration_counts(numSteps);
+    Array<Array<double> > iteration_times(numSteps);
+    Array<Array<double> > setup_times(numSteps);
+    Array<RCP<Matrix> > matrices(numSteps);
+    Array<RCP<MultiVector> > rhss(numSteps);
 
     for (int i = 0; i < numSteps; i++) {
       iteration_counts[i].resize(numSteps);
-      iteration_times[i] .resize(numSteps);
-      setup_times[i]     .resize(numSteps);
+      iteration_times[i].resize(numSteps);
+      setup_times[i].resize(numSteps);
     }
 
     ParameterListInterpreter mueLuFactory(xmlFileName, *comm);
@@ -151,16 +160,16 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
       out << "==================================================================" << std::endl;
 
       char matrixFileName[80];
-      char rhsFileName   [80];
-      char timerName     [80];
+      char rhsFileName[80];
+      char timerName[80];
 
-      sprintf(matrixFileName,"%s%d.mm", matrixPrefix.c_str(), i);
+      sprintf(matrixFileName, "%s%d.mm", matrixPrefix.c_str(), i);
 
       // Load the matrix
       RCP<Matrix> Aprecond = matrices[i];
       if (Aprecond.is_null()) {
         out << "[" << i << "] Loading matrix \"" << matrixFileName << "\"... ";
-        Aprecond = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm, binary);
+        Aprecond = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm, binary);
         out << "done" << std::endl;
 
         Aprecond->SetFixedBlockSize(numPDEs);
@@ -181,7 +190,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
       data0 = nullspace->getDataNonConst(0);
       data1 = nullspace->getDataNonConst(1);
       for (size_t k = 0; k < Aprecond->getRowMap()->getLocalNumElements(); k += 2)
-        data0[k+0] = data1[k+1] = one;
+        data0[k + 0] = data1[k + 1] = one;
 
       // Build the preconditioner
       out << "[" << i << "] Building preconditioner \"" << matrixFileName << "\" ..." << std::endl;
@@ -190,7 +199,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
 
       RCP<Hierarchy> H = mueLuFactory.CreateHierarchy();
       H->SetDefaultVerbLevel(MueLu::Extreme);
-      H->GetLevel(0)->Set("A",         Aprecond);
+      H->GetLevel(0)->Set("A", Aprecond);
       H->GetLevel(0)->Set("Nullspace", nullspace);
       H->IsPreconditioner(true);
 
@@ -198,11 +207,11 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
       timer = TimeMonitor::getNewTimer(timerName);
       timer->start();
       mueLuFactory.SetupHierarchy(*H);
-      setup_times[i-first_matrix][i-first_matrix] = timer->stop();
-      timer = Teuchos::null;
+      setup_times[i - first_matrix][i - first_matrix] = timer->stop();
+      timer                                           = Teuchos::null;
 
       Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
-      tm = Teuchos::null;
+      tm                         = Teuchos::null;
 
       // Loop over all future matrices
       int j = i;
@@ -210,14 +219,14 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
         out << "------------------------------------------------------------------" << std::endl;
 
         sprintf(matrixFileName, "%s%d.mm", matrixPrefix.c_str(), j);
-        sprintf(rhsFileName,    "%s%d.mm", rhsPrefix.c_str(),    j);
+        sprintf(rhsFileName, "%s%d.mm", rhsPrefix.c_str(), j);
 
         RCP<Matrix> Amatvec = matrices[j];
         if (j != i) {
           if (Amatvec.is_null()) {
             // Load the matrix
             out << "[" << j << "]<-[" << i << "] Loading matrix \"" << matrixFileName << "\"... ";
-            Amatvec = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm, binary);
+            Amatvec = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileName), xpetraParameters.GetLib(), comm, binary);
             out << "done" << std::endl;
 
             Amatvec->SetFixedBlockSize(numPDEs);
@@ -243,8 +252,8 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
           H->GetLevel(0)->Set("A", Amatvec);
           mueLuFactory.SetupHierarchy(*H);
 
-          setup_times[i-first_matrix][j-first_matrix] = timer->stop();
-          timer = Teuchos::null;
+          setup_times[i - first_matrix][j - first_matrix] = timer->stop();
+          timer                                           = Teuchos::null;
 
           tm = Teuchos::null;
 
@@ -256,7 +265,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
         RCP<MultiVector> rhs = rhss[j];
         if (rhs.is_null()) {
           out << "[" << j << "] Loading rhs " << rhsFileName << "\"... ";
-          rhs = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(std::string(rhsFileName), Amatvec->getRowMap());
+          rhs = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(std::string(rhsFileName), Amatvec->getRowMap());
           out << "done" << std::endl;
 
           if (inMemory)
@@ -268,36 +277,36 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
         X->putScalar(0.0);
 
         // Define Operator and Preconditioner
-        Teuchos::RCP<OP> belosOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(Amatvec)); // Turns a Xpetra::Matrix object into a Belos operator
+        Teuchos::RCP<OP> belosOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(Amatvec));  // Turns a Xpetra::Matrix object into a Belos operator
 
         // Construct a Belos LinearProblem object
-        RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, rhs));
+        RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, rhs));
         belosProblem->setLeftPrec(belosPrec);
         belosProblem->setProblem();
 
         // Belos parameter list
-        int    maxIts = 100;
-        double tol    = 1e-4;
+        int maxIts = 100;
+        double tol = 1e-4;
         Teuchos::ParameterList belosList;
-        belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-        belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
+        belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+        belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
         // belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::TimingDetails + Belos::StatusTestDetails);
-        belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-        belosList.set("Output Frequency",      1);
-        belosList.set("Output Style",          Belos::Brief);
+        belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+        belosList.set("Output Frequency", 1);
+        belosList.set("Output Style", Belos::Brief);
 
         // Create an iterative solver manager
-        RCP< Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+        RCP<Belos::SolverManager<SC, MV, OP> > solver = rcp(new Belos::BlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
 
         // Perform solve
         sprintf(timerName, "Reuse: Solve i=%d j=%d", i, j);
         timer = TimeMonitor::getNewTimer(timerName);
         timer->start();
         Belos::ReturnType ret = Belos::Unconverged;
-        ret = solver->solve();
+        ret                   = solver->solve();
 
         double my_time = timer->stop();
-        timer = Teuchos::null;
+        timer          = Teuchos::null;
 
         // Get the number of iterations for this solve.
 
@@ -308,18 +317,18 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
         } else {
           out << "SUCCESS:  Belos converged!" << std::endl;
           out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
-          iteration_counts[i-first_matrix][j-first_matrix] = solver->getNumIters();
-          iteration_times [i-first_matrix][j-first_matrix] = my_time;
+          iteration_counts[i - first_matrix][j - first_matrix] = solver->getNumIters();
+          iteration_times[i - first_matrix][j - first_matrix]  = my_time;
         }
 
         if (doReuse == 0) {
           j++;
           break;
         }
-      } //end j
+      }  // end j
       for (; j <= last_matrix; j++)
-        setup_times[i-first_matrix][j-first_matrix] = Teuchos::ScalarTraits<double>::nan();
-    } // end i
+        setup_times[i - first_matrix][j - first_matrix] = Teuchos::ScalarTraits<double>::nan();
+    }  // end i
 
     globalTimeMonitor = Teuchos::null;
 
@@ -330,7 +339,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
     if (!mypid) {
       printf("************************* Iteration Counts ***********************\n");
       for (int i = 0; i < numSteps; i++) {
-        for (int j = 0; j < i;         j++) printf("        ");
+        for (int j = 0; j < i; j++) printf("        ");
         for (int j = i; j < numSteps; j++) {
           if (STS::isnaninf(setup_times[i][j])) {
             if (i == j)
@@ -348,7 +357,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
 
       printf("************************* Iteration Times ***********************\n");
       for (int i = 0; i < numSteps; i++) {
-        for (int j = 0; j < i;         j++) printf("        ");
+        for (int j = 0; j < i; j++) printf("        ");
         for (int j = i; j < numSteps; j++) {
           if (STS::isnaninf(setup_times[i][j]))
             break;
@@ -363,7 +372,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
 
       printf("************************* Setup Times ***********************\n");
       for (int i = 0; i < numSteps; i++) {
-        for (int j = 0; j < i;         j++) printf("        ");
+        for (int j = 0; j < i; j++) printf("        ");
         for (int j = i; j < numSteps; j++) {
           if (STS::isnaninf(setup_times[i][j]))
             break;
@@ -378,7 +387,7 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
 
       printf("************************* Total Times ***********************\n");
       for (int i = 0; i < numSteps; i++) {
-        for (int j = 0; j < i;         j++) printf("        ");
+        for (int j = 0; j < i; j++) printf("        ");
         for (int j = i; j < numSteps; j++) {
           if (STS::isnaninf(setup_times[i][j]))
             break;
@@ -396,18 +405,13 @@ int main_(Teuchos::CommandLineProcessor &clp,  Xpetra::UnderlyingLib& lib, int a
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
-
-
-
diff --git a/packages/muelu/test/scaling/SpMVPerformance.cpp b/packages/muelu/test/scaling/SpMVPerformance.cpp
index 9123191df945..e7559cc1ce58 100644
--- a/packages/muelu/test/scaling/SpMVPerformance.cpp
+++ b/packages/muelu/test/scaling/SpMVPerformance.cpp
@@ -53,20 +53,18 @@
 
 #include "MatrixLoad.hpp"
 
-
 class Measurement {
-public:
+ public:
   virtual void measure() = 0;
 };
 
 class Reporter {
-public:
+ public:
   Reporter(std::string n)
-    : name(n),
-      comm(Teuchos::DefaultComm<int>::getComm()),
-      out(Teuchos::rcpFromRef(std::cout)),
-      timer(Teuchos::rcp(new Teuchos::StackedTimer(name.c_str())))
-  {
+    : name(n)
+    , comm(Teuchos::DefaultComm<int>::getComm())
+    , out(Teuchos::rcpFromRef(std::cout))
+    , timer(Teuchos::rcp(new Teuchos::StackedTimer(name.c_str()))) {
     out.setOutputToRootOnly(0);
     Teuchos::TimeMonitor::setStackedTimer(timer);
   }
@@ -85,25 +83,26 @@ class Reporter {
     out << "Run Complete!" << std::endl;
   }
 
-private:
+ private:
   void reportCommandLine() {
     Teuchos::StackedTimer::OutputOptions options;
-    options.output_fraction = true;
+    options.output_fraction  = true;
     options.output_histogram = true;
-    options.output_minmax = true;
+    options.output_minmax    = true;
 
     timer->report(out, comm, options);
   }
 
   void reportXML() {
     std::string xmlName = name + " " + std::to_string(comm->getSize()) + " ranks";
-    auto xmlOut = timer->reportWatchrXML(xmlName, comm);
-
-    if(xmlOut.length()) {
-      out << std::endl << "Created Watchr performance report " << xmlOut << std::endl;
-    }
-    else {
-      out << std::endl << "Did not create Watchr performance report" << std::endl;
+    auto xmlOut         = timer->reportWatchrXML(xmlName, comm);
+
+    if (xmlOut.length()) {
+      out << std::endl
+          << "Created Watchr performance report " << xmlOut << std::endl;
+    } else {
+      out << std::endl
+          << "Did not create Watchr performance report" << std::endl;
     }
   }
 
@@ -113,51 +112,47 @@ class Reporter {
   Teuchos::RCP<Teuchos::StackedTimer> timer;
 };
 
-
 class System {
-public:
-  virtual void apply() = 0;
+ public:
+  virtual void apply()             = 0;
   virtual std::string name() const = 0;
 };
 
 class SpMVMeasurement : public Measurement {
-public:
+ public:
   SpMVMeasurement(System& sys, int n)
-    : numRuns(n),
-      system(sys),
-      name(sys.name() + " SpMV")
-  { }
+    : numRuns(n)
+    , system(sys)
+    , name(sys.name() + " SpMV") {}
 
   void measure() override {
-    for (int i=0; i<numRuns; i++) {
+    for (int i = 0; i < numRuns; i++) {
       Teuchos::TimeMonitor subTimer(*Teuchos::TimeMonitor::getNewTimer(name));
       system.apply();
     }
   }
 
-private:
+ private:
   const int numRuns;
   System& system;
   const std::string name;
 };
 
-
 std::string systemName(const Xpetra::UnderlyingLib& lib) {
   if (lib == Xpetra::UseEpetra) return "Epetra";
   if (lib == Xpetra::UseTpetra) return "Tpetra";
   return "Unknown";
 }
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class LinearSystem : public System {
   using MultiVector = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-  using Matrix = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Matrix      = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-public:
+ public:
   LinearSystem(Xpetra::UnderlyingLib l, int n)
-    : lib(l),
-      numVectors(n)
-  { }
+    : lib(l)
+    , numVectors(n) {}
 
   void apply() override {
     A->apply(*X, *B);
@@ -175,20 +170,19 @@ class LinearSystem : public System {
   Teuchos::RCP<MultiVector> B;
 };
 
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 class SystemLoader {
-  using Map = Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
-  using MultiVector = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-  using CoordScalar = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
+  using Map                   = Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>;
+  using MultiVector           = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using CoordScalar           = typename Teuchos::ScalarTraits<Scalar>::coordinateType;
   using RealValuedMultiVector = Xpetra::MultiVector<CoordScalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-public:
+ public:
   SystemLoader(Teuchos::CommandLineProcessor& c)
-    : clp(c),
-      comm(Teuchos::DefaultComm<int>::getComm()),
-      galeriParameters(clp, 100, 100, 100, "Laplace2D"),
-      xpetraParameters(clp)
-  { }
+    : clp(c)
+    , comm(Teuchos::DefaultComm<int>::getComm())
+    , galeriParameters(clp, 100, 100, 100, "Laplace2D")
+    , xpetraParameters(clp) {}
 
   void fill(LinearSystem<Scalar, LocalOrdinal, GlobalOrdinal, Node>& system) {
     MatrixLoad<Scalar, LocalOrdinal, GlobalOrdinal, Node>(
@@ -202,23 +196,23 @@ class SystemLoader {
         galeriStream);
   }
 
-private:
+ private:
   Teuchos::CommandLineProcessor& clp;
   Teuchos::RCP<const Teuchos::Comm<int>> comm;
   Galeri::Xpetra::Parameters<GlobalOrdinal> galeriParameters;
   Xpetra::Parameters xpetraParameters;
 
   bool binaryFormat;
-  std::string matrixFile = "";
-  std::string rhsFile = "";
-  std::string rowMapFile = "";
-  std::string colMapFile = "";
+  std::string matrixFile    = "";
+  std::string rhsFile       = "";
+  std::string rowMapFile    = "";
+  std::string colMapFile    = "";
   std::string domainMapFile = "";
-  std::string rangeMapFile = "";
-  std::string coordFile = "";
-  std::string coordMapFile = "";
-  std::string nullFile = "";
-  std::string materialFile = "";
+  std::string rangeMapFile  = "";
+  std::string coordFile     = "";
+  std::string coordMapFile  = "";
+  std::string nullFile      = "";
+  std::string materialFile  = "";
 
   Teuchos::RCP<const Map> map;
   Teuchos::RCP<RealValuedMultiVector> coordinates;
@@ -227,8 +221,7 @@ class SystemLoader {
   std::ostringstream galeriStream;
 };
 
-
-template<typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
+template <typename Scalar, typename LocalOrdinal, typename GlobalOrdinal, typename Node>
 int main_ETI(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
   Reporter reporter("SpMV Performance " + systemName(lib));
   SystemLoader<Scalar, LocalOrdinal, GlobalOrdinal, Node> systemLoader(clp);
@@ -241,10 +234,10 @@ int main_ETI(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   LinearSystem<Scalar, LocalOrdinal, GlobalOrdinal, Node> system(lib, numVectors);
@@ -260,6 +253,6 @@ int main_ETI(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_ETI
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
   return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/scaling/TAFCPerformance.cpp b/packages/muelu/test/scaling/TAFCPerformance.cpp
index 3a0d37cf614b..b7933bf54f98 100644
--- a/packages/muelu/test/scaling/TAFCPerformance.cpp
+++ b/packages/muelu/test/scaling/TAFCPerformance.cpp
@@ -81,7 +81,6 @@
 #include <EpetraExt_RowMatrixOut.h>
 #endif
 
-
 using Teuchos::RCP;
 using Teuchos::rcp;
 using Teuchos::TimeMonitor;
@@ -89,26 +88,26 @@ using Teuchos::TimeMonitor;
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >  build_map_for_transfer(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > oldMap){
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > build_map_for_transfer(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > oldMap) {
 #include <MueLu_UseShortNames.hpp>
   // 10 unknowns per proc except N-1: Assume Ids run from 0 to MaxGlobalIndex
   int Nproc = oldMap->getComm()->getSize();
   int MyPID = oldMap->getComm()->getRank();
-  int N = (int)oldMap->getGlobalNumElements();
+  int N     = (int)oldMap->getGlobalNumElements();
 
   Xpetra::UnderlyingLib lib = oldMap->lib();
 
-  if(Nproc==1 )return oldMap;
+  if (Nproc == 1) return oldMap;
 
-  int start= MyPID==Nproc-1? 0 : std::max(N-(MyPID+1)*10,0);
-  int stop=std::max(N-MyPID*10,0);
+  int start = MyPID == Nproc - 1 ? 0 : std::max(N - (MyPID + 1) * 10, 0);
+  int stop  = std::max(N - MyPID * 10, 0);
 
   Teuchos::Array<GlobalOrdinal> elems;
-  if(stop-start>0) {
-    elems.resize(stop-start);
-    for(int i=0; i<stop-start; i++)
-      elems[i] = start+i;
+  if (stop - start > 0) {
+    elems.resize(stop - start);
+    for (int i = 0; i < stop - start; i++)
+      elems[i] = start + i;
   }
 
 #if 0
@@ -119,33 +118,30 @@ Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >  build_map_for
   fflush(stdout);
 #endif
 
-  return Xpetra::MapFactory<LO,GO,Node>::Build(lib,oldMap->getGlobalNumElements(),elems(),oldMap->getIndexBase(),oldMap->getComm());
-
+  return Xpetra::MapFactory<LO, GO, Node>::Build(lib, oldMap->getGlobalNumElements(), elems(), oldMap->getIndexBase(), oldMap->getComm());
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >  build_map_for_transfer_repartition(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > oldMap){
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > build_map_for_transfer_repartition(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > oldMap) {
 #include <MueLu_UseShortNames.hpp>
   // Turn P procs into ~P/9 procs
   int Nproc = oldMap->getComm()->getSize();
   int MyPID = oldMap->getComm()->getRank();
-  int N = (int)oldMap->getGlobalNumElements();
+  int N     = (int)oldMap->getGlobalNumElements();
 
   Xpetra::UnderlyingLib lib = oldMap->lib();
 
-  if(Nproc==1 ) return oldMap;
+  if (Nproc == 1) return oldMap;
   int ideal_new_unknowns_per_proc = N / (Nproc / 9.0);
 
-
-  int start= ideal_new_unknowns_per_proc*MyPID < N ? ideal_new_unknowns_per_proc*MyPID : 0;
-  int stop = ideal_new_unknowns_per_proc*MyPID < N ? std::min(N,ideal_new_unknowns_per_proc*(MyPID+1)):0;
+  int start = ideal_new_unknowns_per_proc * MyPID < N ? ideal_new_unknowns_per_proc * MyPID : 0;
+  int stop  = ideal_new_unknowns_per_proc * MyPID < N ? std::min(N, ideal_new_unknowns_per_proc * (MyPID + 1)) : 0;
 
   Teuchos::Array<GlobalOrdinal> elems;
-  if(stop-start>0) {
-    elems.resize(stop-start);
-    for(int i=0; i<stop-start; i++)
-      elems[i] = start+i;
+  if (stop - start > 0) {
+    elems.resize(stop - start);
+    for (int i = 0; i < stop - start; i++)
+      elems[i] = start + i;
   }
 
 #if 0
@@ -156,51 +152,47 @@ Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >  build_map_for
   fflush(stdout);
 #endif
 
-  int i_am_active= (elems.size()>0);
-  int num_active=0;
-  Teuchos::reduce(&i_am_active,&num_active,1,Teuchos::REDUCE_SUM,0,*oldMap->getComm());
-  if(MyPID==0) printf("Repartitioning to %d/%d processors\n",num_active,Nproc);
-
-  return Xpetra::MapFactory<LO,GO,Node>::Build(lib,oldMap->getGlobalNumElements(),elems(),oldMap->getIndexBase(),oldMap->getComm());
+  int i_am_active = (elems.size() > 0);
+  int num_active  = 0;
+  Teuchos::reduce(&i_am_active, &num_active, 1, Teuchos::REDUCE_SUM, 0, *oldMap->getComm());
+  if (MyPID == 0) printf("Repartitioning to %d/%d processors\n", num_active, Nproc);
 
+  return Xpetra::MapFactory<LO, GO, Node>::Build(lib, oldMap->getGlobalNumElements(), elems(), oldMap->getIndexBase(), oldMap->getComm());
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<const Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > toXpetraCrs(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<const Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > toXpetraCrs(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &A) {
 #include <MueLu_UseShortNames.hpp>
   RCP<const CrsMatrixWrap> crswrapA = Teuchos::rcp_dynamic_cast<const CrsMatrixWrap>(A);
   TEUCHOS_TEST_FOR_EXCEPTION(crswrapA == Teuchos::null, MueLu::Exceptions::BadCast,
-			     "Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
+                             "Cast from Xpetra::Matrix to Xpetra::CrsMatrixWrap failed");
   RCP<CrsMatrix> crsA = crswrapA->getCrsMatrix();
   TEUCHOS_TEST_FOR_EXCEPTION(crsA == Teuchos::null, std::runtime_error,
-                               "Xpetra::CrsMatrixWrap doesn't have a CrsMatrix");
+                             "Xpetra::CrsMatrixWrap doesn't have a CrsMatrix");
   return crsA;
 }
 
-
-
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void TestTransferAndFillComplete(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > &A, Teuchos::RCP<Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > &importer) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TestTransferAndFillComplete(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &A, Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > &importer) {
 #include <MueLu_UseShortNames.hpp>
   //  Xpetra::UnderlyingLib lib = A->getRowMap()->lib();
   RCP<TimeMonitor> tm;
 
   // It makes me sad to do this
-  Teuchos::RCP<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > constA = Teuchos::rcp<const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(&*A,false);
+  Teuchos::RCP<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > constA = Teuchos::rcp<const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >(&*A, false);
 
   // Only makes sense in parallel
-  if(A->getRowMap()->getComm()->getSize()==1) return;
+  if (A->getRowMap()->getComm()->getSize() == 1) return;
 
   // ==================
   // Optimized Transfer
   // ==================
   {
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("OptimizedTransfer")));
-    RCP<Xpetra::CrsMatrix<Scalar,LO,GO,Node> > B = Xpetra::CrsMatrixFactory<Scalar,LO,GO,Node>::Build(toXpetraCrs(A),*importer);
+    tm                                              = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("OptimizedTransfer")));
+    RCP<Xpetra::CrsMatrix<Scalar, LO, GO, Node> > B = Xpetra::CrsMatrixFactory<Scalar, LO, GO, Node>::Build(toXpetraCrs(A), *importer);
   }
 
   A->getRowMap()->getComm()->barrier();
@@ -209,28 +201,27 @@ void TestTransferAndFillComplete(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal
   // Naive Transfer
   // ==================
   {
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("NaiveTransfer")));
-    RCP<Xpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > B = Xpetra::CrsMatrixFactory<Scalar,LO,GO,Node>::Build(importer->getTargetMap(), 0);
-    B->doImport(*toXpetraCrs(A),*importer,Xpetra::ADD);
+    tm                                                                   = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("NaiveTransfer")));
+    RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > B = Xpetra::CrsMatrixFactory<Scalar, LO, GO, Node>::Build(importer->getTargetMap(), 0);
+    B->doImport(*toXpetraCrs(A), *importer, Xpetra::ADD);
     B->fillComplete();
   }
   A->getRowMap()->getComm()->barrier();
-
 }
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
-  using Teuchos::RCP;
-  using Teuchos::rcp;
   using Teuchos::ArrayRCP;
   using Teuchos::ParameterList;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
 
   typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -241,21 +232,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // Parameters initialization
   // =========================================================================
   GO nx = 200, ny = 200, nz = 10;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "import.xml";     clp.setOption("xml",                   &xmlFileName,       "read parameters from a file");
-  bool        printTimings      = true;              clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-  std::string timingsFormat     = "table-fixed";     clp.setOption("time-format",           &timingsFormat,     "timings format (table-fixed | table-scientific | yaml)");
-  int         numImports        = 100;               clp.setOption("numImport",              &numImports,        "#times to test");
-  std::string mapmode           = "small";           clp.setOption("mapMode", &mapmode, "map to use ('small' or 'repartition')");
-
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "import.xml";
+  clp.setOption("xml", &xmlFileName, "read parameters from a file");
+  bool printTimings = true;
+  clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+  std::string timingsFormat = "table-fixed";
+  clp.setOption("time-format", &timingsFormat, "timings format (table-fixed | table-scientific | yaml)");
+  int numImports = 100;
+  clp.setOption("numImport", &numImports, "#times to test");
+  std::string mapmode = "small";
+  clp.setOption("mapMode", &mapmode, "map to use ('small' or 'repartition')");
 
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   ParameterList paramList;
@@ -263,10 +258,10 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   bool isDriver = paramList.isSublist("Run1");
   if (isDriver) {
     // update galeriParameters with the values from the XML file
-    ParameterList& realParams = galeriParameters.GetParameterList();
+    ParameterList &realParams = galeriParameters.GetParameterList();
 
     for (ParameterList::ConstIterator it = realParams.begin(); it != realParams.end(); it++) {
-      const std::string& name = realParams.name(it);
+      const std::string &name = realParams.name(it);
       if (paramList.isParameter(name))
         realParams.setEntry(name, paramList.getEntry(name));
     }
@@ -284,72 +279,71 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-  RCP<Xpetra::Matrix<Scalar,LO,GO,Node> >      A;
-  RCP<const Map>   map;
+  RCP<Xpetra::Matrix<Scalar, LO, GO, Node> > A;
+  RCP<const Map> map;
   RCP<RealValuedMultiVector> coordinates;
   typedef typename RealValuedMultiVector::scalar_type Real;
   RCP<MultiVector> nullspace;
-  galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
-
-    // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
-    //                                 d1  d2  d3
-    //                                 d4  d5  d6
-    //                                 d7  d8  d9
-    //                                 d10 d11 d12
-    // A perfect distribution is only possible when the #processors is a perfect square.
-    // This *will* result in "strip" distribution if the #processors is a prime number or if the factors are very different in
-    // size. For example, np=14 will give a 7-by-2 distribution.
-    // If you don't want Galeri to do this, specify mx or my on the galeriList.
-    std::string matrixType = galeriParameters.GetMatrixType();
-
-    // Create map and coordinates
-    // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
-    // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
-    if (matrixType == "Laplace1D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
-
-    } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
-               matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
-
-    } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-      map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-      coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
-    }
-
-    // Expand map to do multiple DOF per node for block problems
-    if (matrixType == "Elasticity2D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
-    if (matrixType == "Elasticity3D")
-      map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
-
-    galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
-                 << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
-                 << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
-                 << "========================================================" << std::endl;
-
-    if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
-      // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-      galeriList.set("right boundary" , "Neumann");
-      galeriList.set("bottom boundary", "Neumann");
-      galeriList.set("top boundary"   , "Neumann");
-      galeriList.set("front boundary" , "Neumann");
-      galeriList.set("back boundary"  , "Neumann");
-    }
+  galeriStream << "========================================================\n"
+               << xpetraParameters << galeriParameters;
+
+  // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
+  //                                 d1  d2  d3
+  //                                 d4  d5  d6
+  //                                 d7  d8  d9
+  //                                 d10 d11 d12
+  // A perfect distribution is only possible when the #processors is a perfect square.
+  // This *will* result in "strip" distribution if the #processors is a prime number or if the factors are very different in
+  // size. For example, np=14 will give a 7-by-2 distribution.
+  // If you don't want Galeri to do this, specify mx or my on the galeriList.
+  std::string matrixType = galeriParameters.GetMatrixType();
+
+  // Create map and coordinates
+  // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
+  // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
+  if (matrixType == "Laplace1D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
+
+  } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
+             matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
+
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+    map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Real, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
+  }
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
-    A = Pr->BuildMatrix();
+  // Expand map to do multiple DOF per node for block problems
+  if (matrixType == "Elasticity2D")
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
+  if (matrixType == "Elasticity3D")
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
+
+  galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
+               << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
+               << "Processor subdomains in z direction: " << galeriList.get<GO>("mz") << std::endl
+               << "========================================================" << std::endl;
+
+  if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
+    // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
+    galeriList.set("right boundary", "Neumann");
+    galeriList.set("bottom boundary", "Neumann");
+    galeriList.set("top boundary", "Neumann");
+    galeriList.set("front boundary", "Neumann");
+    galeriList.set("back boundary", "Neumann");
+  }
 
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+  A = Pr->BuildMatrix();
 
   comm->barrier();
   tm = Teuchos::null;
 
   galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
-
   int numReruns = 1;
   if (paramList.isParameter("number of reruns"))
     numReruns = paramList.get<int>("number of reruns");
@@ -359,17 +353,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     ParameterList mueluList, runList;
     bool stop = false;
     if (isDriver) {
-      runList   = paramList.sublist("Run1",  mustAlreadyExist);
-      mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+      runList   = paramList.sublist("Run1", mustAlreadyExist);
+      mueluList = runList.sublist("MueLu", mustAlreadyExist);
     } else {
       mueluList = paramList;
-      stop = true;
+      stop      = true;
     }
 
     int runCount = 1;
     do {
-      int   savedOut  = -1;
-      FILE* openedOut = NULL;
+      int savedOut    = -1;
+      FILE *openedOut = NULL;
       if (isDriver) {
         if (runList.isParameter("filename")) {
           // Redirect all output into a filename We have to redirect all output,
@@ -388,30 +382,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       // Instead of checking each time for rank, create a rank 0 stream
       RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      Teuchos::FancyOStream& out = *fancy;
+      Teuchos::FancyOStream &out       = *fancy;
       out.setOutputToRootOnly(0);
       out << galeriStream.str();
 
-
       // Build the target map for the importing
-      Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >  importMap;
-      if(mapmode=="small") importMap=build_map_for_transfer<Scalar,LO,GO,Node>(A->getRowMap());
-      else if(mapmode=="repartition") importMap=build_map_for_transfer_repartition<Scalar,LO,GO,Node>(A->getRowMap());
-      else throw std::runtime_error("Invalid map mode");
+      Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > importMap;
+      if (mapmode == "small")
+        importMap = build_map_for_transfer<Scalar, LO, GO, Node>(A->getRowMap());
+      else if (mapmode == "repartition")
+        importMap = build_map_for_transfer_repartition<Scalar, LO, GO, Node>(A->getRowMap());
+      else
+        throw std::runtime_error("Invalid map mode");
 
-      Teuchos::RCP<Xpetra::Import<LocalOrdinal,GlobalOrdinal,Node> > importer = Xpetra::ImportFactory<LO,GO,Node>::Build(A->getRowMap(),importMap);
+      Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> > importer = Xpetra::ImportFactory<LO, GO, Node>::Build(A->getRowMap(), importMap);
 
-      for(int i=0; i<numImports; i++) {
+      for (int i = 0; i < numImports; i++) {
         // =========================================================================
         // Optimized transfer & fill complete loop
         // =========================================================================
         tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - TransferAndFillComplete")));
-	TestTransferAndFillComplete<Scalar,LO,GO,Node>(A,importer);
+        TestTransferAndFillComplete<Scalar, LO, GO, Node>(A, importer);
         comm->barrier();
-        tm= Teuchos::null;
+        tm = Teuchos::null;
       }
 
-      //Cleanup
+      // Cleanup
       globalTimeMonitor = Teuchos::null;
 
       // =========================================================================
@@ -420,20 +416,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       if (printTimings) {
         RCP<ParameterList> reportParams = rcp(new ParameterList);
         if (timingsFormat == "yaml") {
-          reportParams->set("Report format",             "YAML");            // "Table" or "YAML"
-          reportParams->set("YAML style",                "compact");         // "spacious" or "compact"
+          reportParams->set("Report format", "YAML");  // "Table" or "YAML"
+          reportParams->set("YAML style", "compact");  // "spacious" or "compact"
         }
-        reportParams->set("How to merge timer sets",   "Union");
-        reportParams->set("alwaysWriteLocal",          false);
-        reportParams->set("writeGlobalStats",          true);
-        reportParams->set("writeZeroTimers",           false);
+        reportParams->set("How to merge timer sets", "Union");
+        reportParams->set("alwaysWriteLocal", false);
+        reportParams->set("writeGlobalStats", true);
+        reportParams->set("writeZeroTimers", false);
         // FIXME: no "ignoreZeroTimers"
 
         const std::string filter = "";
 
         std::ios_base::fmtflags ff(out.flags());
-        if (timingsFormat == "table-fixed") out << std::fixed;
-        else                                out << std::scientific;
+        if (timingsFormat == "table-fixed")
+          out << std::fixed;
+        else
+          out << std::scientific;
         TimeMonitor::report(comm.ptr(), out, filter, reportParams);
         out << std::setiosflags(ff);
       }
@@ -449,24 +447,22 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         }
         try {
           runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
-          mueluList = runList  .sublist("MueLu", mustAlreadyExist);
-        } catch (Teuchos::Exceptions::InvalidParameterName&) {
+          mueluList = runList.sublist("MueLu", mustAlreadyExist);
+        } catch (Teuchos::Exceptions::InvalidParameterName &) {
           stop = true;
         }
       }
 
-
     } while (!stop);
   }
 
   return EXIT_SUCCESS;
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/scaling/TwoMatrixMMKernelDriver.cpp b/packages/muelu/test/scaling/TwoMatrixMMKernelDriver.cpp
index e2b511600cd6..52127eb284bc 100644
--- a/packages/muelu/test/scaling/TwoMatrixMMKernelDriver.cpp
+++ b/packages/muelu/test/scaling/TwoMatrixMMKernelDriver.cpp
@@ -71,64 +71,58 @@
 // =========================================================================
 // Support Routines
 // =========================================================================
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view_n(const int n, const View1 x1, View2 x2) {
-  Kokkos::parallel_for(n,KOKKOS_LAMBDA(const size_t i) {
-      x2[i] = x1[i];
-    });
+  Kokkos::parallel_for(
+      n, KOKKOS_LAMBDA(const size_t i) {
+        x2[i] = x1[i];
+      });
 }
 
-template<class View1, class View2>
+template <class View1, class View2>
 inline void copy_view(const View1 x1, View2 x2) {
-  copy_view_n(x1.extent(0),x1,x2);
+  copy_view_n(x1.extent(0), x1, x2);
 }
 
-
-
 template <class V1, class V2>
 void print_crs_graph(std::string name, const V1 rowptr, const V2 colind) {
-  printf("%s rowptr[%d] = ",name.c_str(),rowptr.extent(0));
-  for(size_t i=0; i<rowptr.extent(0); i++)
-    printf(" %d",(int)rowptr[i]);
-  printf("\n%s colind[%d] = ",name.c_str(),colind.extent(0));
-  for(size_t i=0; i<colind.extent(0); i++)
-    printf(" %d",(int)colind[i]);
+  printf("%s rowptr[%d] = ", name.c_str(), rowptr.extent(0));
+  for (size_t i = 0; i < rowptr.extent(0); i++)
+    printf(" %d", (int)rowptr[i]);
+  printf("\n%s colind[%d] = ", name.c_str(), colind.extent(0));
+  for (size_t i = 0; i < colind.extent(0); i++)
+    printf(" %d", (int)colind[i]);
   printf("\n");
 }
 
-
 // =========================================================================
 // MKL Testing
 // =========================================================================
 #ifdef HAVE_MUELU_MKL
 #include "mkl.h"
 
-
 std::string mkl_error(sparse_status_t code) {
-  switch(code) {
-  case SPARSE_STATUS_SUCCESS:
-    return std::string("Success");
-  case SPARSE_STATUS_NOT_INITIALIZED:
-    return std::string("Empty handle or matrix array");
-  case SPARSE_STATUS_ALLOC_FAILED:
-    return std::string("Memory allocation failed");
-  case SPARSE_STATUS_INVALID_VALUE:
-    return std::string("Input contains an invalid value");
-  case SPARSE_STATUS_EXECUTION_FAILED:
-    return std::string("Execution failed");
-  case SPARSE_STATUS_INTERNAL_ERROR:
-    return std::string("Internal error");
-  case SPARSE_STATUS_NOT_SUPPORTED:
-    return std::string("Operation not supported");
+  switch (code) {
+    case SPARSE_STATUS_SUCCESS:
+      return std::string("Success");
+    case SPARSE_STATUS_NOT_INITIALIZED:
+      return std::string("Empty handle or matrix array");
+    case SPARSE_STATUS_ALLOC_FAILED:
+      return std::string("Memory allocation failed");
+    case SPARSE_STATUS_INVALID_VALUE:
+      return std::string("Input contains an invalid value");
+    case SPARSE_STATUS_EXECUTION_FAILED:
+      return std::string("Execution failed");
+    case SPARSE_STATUS_INTERNAL_ERROR:
+      return std::string("Internal error");
+    case SPARSE_STATUS_NOT_SUPPORTED:
+      return std::string("Operation not supported");
   };
-
 }
 
-   
-
-  // mkl_sparse_spmm
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B1, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B2, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C,Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >&Ccolmap, std::string algorithm_name) {
+// mkl_sparse_spmm
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MM2_MKL(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B1, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B2, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C, Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &Ccolmap, std::string algorithm_name) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -137,92 +131,93 @@ void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, co
 
   Xpetra::UnderlyingLib lib = A.getRowMap()->lib();
   RCP<TimeMonitor> tm;
-  typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
+  typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
   typedef typename crs_matrix_type::local_matrix_device_type KCRS;
-  typedef typename KCRS::StaticCrsGraphType                  graph_t;
-  typedef typename graph_t::row_map_type::non_const_type     lno_view_t;
-  typedef typename graph_t::row_map_type::const_type         c_lno_view_t;
-  typedef typename graph_t::entries_type::non_const_type     lno_nnz_view_t;
-  typedef typename graph_t::entries_type::const_type         c_lno_nnz_view_t;
-  typedef typename KCRS::values_type::non_const_type         scalar_view_t;
-  typedef Tpetra::Map<LO,GO,NO>                                     map_type;
-  typedef typename map_type::local_map_type                         local_map_type;
-
+  typedef typename KCRS::StaticCrsGraphType graph_t;
+  typedef typename graph_t::row_map_type::non_const_type lno_view_t;
+  typedef typename graph_t::row_map_type::const_type c_lno_view_t;
+  typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
+  typedef typename graph_t::entries_type::const_type c_lno_nnz_view_t;
+  typedef typename KCRS::values_type::non_const_type scalar_view_t;
+  typedef Tpetra::Map<LO, GO, NO> map_type;
+  typedef typename map_type::local_map_type local_map_type;
 
-  typedef typename Kokkos::View<MKL_INT*,typename lno_nnz_view_t::array_layout,typename lno_nnz_view_t::device_type> mkl_int_type;
+  typedef typename Kokkos::View<MKL_INT *, typename lno_nnz_view_t::array_layout, typename lno_nnz_view_t::device_type> mkl_int_type;
 
-  RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-  RCP<const crs_matrix_type> B1u = Utilities::Op2TpetraCrs(rcp(&B1,false));
-  RCP<const crs_matrix_type> B2u = Utilities::Op2TpetraCrs(rcp(&B2,false));
-  RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-  RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
+  RCP<const crs_matrix_type> Au  = Utilities::Op2TpetraCrs(rcp(&A, false));
+  RCP<const crs_matrix_type> B1u = Utilities::Op2TpetraCrs(rcp(&B1, false));
+  RCP<const crs_matrix_type> B2u = Utilities::Op2TpetraCrs(rcp(&B2, false));
+  RCP<const crs_matrix_type> Cu  = Utilities::Op2TpetraCrs(rcp(&C, false));
+  RCP<crs_matrix_type> Cnc       = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
 
-  const KCRS & Amat = Au->getLocalMatrixDevice();
-  const KCRS & B1mat = B1u->getLocalMatrixDevice();
-  const KCRS & B2mat = B2u->getLocalMatrixDevice();
+  const KCRS &Amat  = Au->getLocalMatrixDevice();
+  const KCRS &B1mat = B1u->getLocalMatrixDevice();
+  const KCRS &B2mat = B2u->getLocalMatrixDevice();
 
-  if(A.getLocalNumRows()!=C.getLocalNumRows())  throw std::runtime_error("C is not sized correctly");
+  if (A.getLocalNumRows() != C.getLocalNumRows()) throw std::runtime_error("C is not sized correctly");
 
   c_lno_view_t Arowptr = Amat.graph.row_map, B1rowptr = B1mat.graph.row_map, B2rowptr = B2mat.graph.row_map;
-  lno_view_t Crowptr("Crowptr",C.getLocalNumRows()+1);
+  lno_view_t Crowptr("Crowptr", C.getLocalNumRows() + 1);
   c_lno_nnz_view_t Acolind = Amat.graph.entries, B1colind = B1mat.graph.entries, B2colind = B2mat.graph.entries;
   lno_nnz_view_t Ccolind;
   const scalar_view_t Avals = Amat.values, B1vals = B1mat.values, B2vals = B2mat.values;
   scalar_view_t Cvals;
-  RCP<const Tpetra::Map<LO,GO,Node> > Ccolmap_t = Xpetra::toTpetra(Ccolmap);
-  local_map_type Bcolmap_local = B1u->getColMap()->getLocalMap();
-  local_map_type Icolmap_local = B2u->getColMap()->getLocalMap();
-  local_map_type Ccolmap_local = Ccolmap_t->getLocalMap();
+  RCP<const Tpetra::Map<LO, GO, Node> > Ccolmap_t = Xpetra::toTpetra(Ccolmap);
+  local_map_type Bcolmap_local                    = B1u->getColMap()->getLocalMap();
+  local_map_type Icolmap_local                    = B2u->getColMap()->getLocalMap();
+  local_map_type Ccolmap_local                    = Ccolmap_t->getLocalMap();
 
   sparse_matrix_t AMKL;
   sparse_matrix_t B1MKL;
   sparse_matrix_t B2MKL;
   sparse_matrix_t CMKL;
-  sparse_matrix_t Temp1MKL,Temp2MKL;
+  sparse_matrix_t Temp1MKL, Temp2MKL;
 
   sparse_status_t result;
   // **********************************
   // Copy in the matrix data for MKL
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 MKL "+algorithm_name+": Matrix CopyIn")));
+  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 MKL " + algorithm_name + ": Matrix CopyIn")));
 
-  mkl_int_type ArowptrMKL("Arowptr",Arowptr.extent(0));
-  mkl_int_type AcolindMKL("Acolind",Acolind.extent(0));
-  mkl_int_type B1colindMKL("B1colind",B1colind.extent(0));
-  mkl_int_type B2colindMKL("B2colind",B2colind.extent(0));
+  mkl_int_type ArowptrMKL("Arowptr", Arowptr.extent(0));
+  mkl_int_type AcolindMKL("Acolind", Acolind.extent(0));
+  mkl_int_type B1colindMKL("B1colind", B1colind.extent(0));
+  mkl_int_type B2colindMKL("B2colind", B2colind.extent(0));
 
   // Generate new rowptrs that have all the rows for both B and I (as MKL requires)
-  size_t Nb = B1rowptr.extent(0)-1, Ni=B2rowptr.extent(0)-1;
-  mkl_int_type B1rowptrMKL("B1rowptr",Nb+Ni+1);
-  mkl_int_type B2rowptrMKL("B2rowptr",Nb+Ni+1);
-  Kokkos::parallel_for(Nb+1,KOKKOS_LAMBDA(const size_t i) {
-      B1rowptrMKL[i] = B1rowptr[i];
-      if(i!= Nb) B2rowptrMKL[i] = 0;
-    });
-  Kokkos::parallel_for(Ni+1,KOKKOS_LAMBDA(const size_t i) {
-      if(i!=0) B1rowptrMKL[Nb+i] = B1rowptr[Nb-1];
-      B2rowptrMKL[Nb+i] = B2rowptr[i];
-    });
+  size_t Nb = B1rowptr.extent(0) - 1, Ni = B2rowptr.extent(0) - 1;
+  mkl_int_type B1rowptrMKL("B1rowptr", Nb + Ni + 1);
+  mkl_int_type B2rowptrMKL("B2rowptr", Nb + Ni + 1);
+  Kokkos::parallel_for(
+      Nb + 1, KOKKOS_LAMBDA(const size_t i) {
+        B1rowptrMKL[i] = B1rowptr[i];
+        if (i != Nb) B2rowptrMKL[i] = 0;
+      });
+  Kokkos::parallel_for(
+      Ni + 1, KOKKOS_LAMBDA(const size_t i) {
+        if (i != 0) B1rowptrMKL[Nb + i] = B1rowptr[Nb - 1];
+        B2rowptrMKL[Nb + i] = B2rowptr[i];
+      });
 
   // Reindex the column indices into C's column map
-  Kokkos::parallel_for(B1colind.extent(0),KOKKOS_LAMBDA(const size_t i) {
-      B1colindMKL[i] = Ccolmap_local.getLocalElement(Bcolmap_local.getGlobalElement(B1colind(i)));
-  });
-
-  Kokkos::parallel_for(B2colind.extent(0),KOKKOS_LAMBDA(const size_t i) {
-      B2colindMKL[i] = Ccolmap_local.getLocalElement(Icolmap_local.getGlobalElement(B2colind(i)));
-  });
+  Kokkos::parallel_for(
+      B1colind.extent(0), KOKKOS_LAMBDA(const size_t i) {
+        B1colindMKL[i] = Ccolmap_local.getLocalElement(Bcolmap_local.getGlobalElement(B1colind(i)));
+      });
 
+  Kokkos::parallel_for(
+      B2colind.extent(0), KOKKOS_LAMBDA(const size_t i) {
+        B2colindMKL[i] = Ccolmap_local.getLocalElement(Icolmap_local.getGlobalElement(B2colind(i)));
+      });
 
   // Copy in A
-  copy_view(Arowptr,ArowptrMKL);
-  copy_view(Acolind,AcolindMKL);
-
-  if(std::is_same<Scalar,double>::value) {
-    mkl_sparse_d_create_csr(&AMKL, SPARSE_INDEX_BASE_ZERO, Au->getLocalNumRows(), Au->getLocalNumCols(), ArowptrMKL.data(),ArowptrMKL.data()+1,AcolindMKL.data(),(double*)Avals.data());
-    mkl_sparse_d_create_csr(&B1MKL, SPARSE_INDEX_BASE_ZERO, Nb+Ni, Ccolmap->getLocalNumElements(), B1rowptrMKL.data(),B1rowptrMKL.data()+1,B1colindMKL.data(),(double*)B1vals.data());
-    mkl_sparse_d_create_csr(&B2MKL, SPARSE_INDEX_BASE_ZERO, Nb+Ni, Ccolmap->getLocalNumElements(), B2rowptrMKL.data(),B2rowptrMKL.data()+1,B2colindMKL.data(),(double*)B2vals.data());
-  }
-  else
+  copy_view(Arowptr, ArowptrMKL);
+  copy_view(Acolind, AcolindMKL);
+
+  if (std::is_same<Scalar, double>::value) {
+    mkl_sparse_d_create_csr(&AMKL, SPARSE_INDEX_BASE_ZERO, Au->getLocalNumRows(), Au->getLocalNumCols(), ArowptrMKL.data(), ArowptrMKL.data() + 1, AcolindMKL.data(), (double *)Avals.data());
+    mkl_sparse_d_create_csr(&B1MKL, SPARSE_INDEX_BASE_ZERO, Nb + Ni, Ccolmap->getLocalNumElements(), B1rowptrMKL.data(), B1rowptrMKL.data() + 1, B1colindMKL.data(), (double *)B1vals.data());
+    mkl_sparse_d_create_csr(&B2MKL, SPARSE_INDEX_BASE_ZERO, Nb + Ni, Ccolmap->getLocalNumElements(), B2rowptrMKL.data(), B2rowptrMKL.data() + 1, B2colindMKL.data(), (double *)B2vals.data());
+  } else
     throw std::runtime_error("MKL Type Mismatch");
   tm = Teuchos::null;
   Au->getComm()->barrier();
@@ -230,70 +225,69 @@ void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, co
   tm = Teuchos::null;
   Au->getComm()->barrier();
 
-  if(algorithm_name == "MULT_ADD" ) {
+  if (algorithm_name == "MULT_ADD") {
     // **********************************
     // Multiply #1 (A*B1)
     {
-    TimeMonitor tm1(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Multiply 1"));
-    result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, B1MKL, &Temp1MKL);
-    typename KCRS::execution_space().fence();
-    if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply 1 failed: "+mkl_error(result));
+      TimeMonitor tm1(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Multiply 1"));
+      result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, B1MKL, &Temp1MKL);
+      typename KCRS::execution_space().fence();
+      if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply 1 failed: " + mkl_error(result));
     }
 
     // **********************************
     // Multiply #2 (A*B2)
     {
-    TimeMonitor tm2(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Multiply 2"));
-    result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, B1MKL, &Temp2MKL);
-    typename KCRS::execution_space().fence();
-    if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply 2 failed: "+mkl_error(result));
+      TimeMonitor tm2(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Multiply 2"));
+      result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, B1MKL, &Temp2MKL);
+      typename KCRS::execution_space().fence();
+      if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply 2 failed: " + mkl_error(result));
     }
 
     // **********************************
     // Add (A*B1) + (A*B2)
     {
-    TimeMonitor tm3(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Add"));
-    result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE,Temp1MKL,1.0,Temp2MKL,&CMKL);
-    typename KCRS::execution_space().fence();
-    if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed: "+mkl_error(result));
+      TimeMonitor tm3(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Add"));
+      result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE, Temp1MKL, 1.0, Temp2MKL, &CMKL);
+      typename KCRS::execution_space().fence();
+      if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed: " + mkl_error(result));
     }
-  }
-  else if(algorithm_name == "ADD_MULT" ) {
+  } else if (algorithm_name == "ADD_MULT") {
     // **********************************
     // Add B1 + B2
     {
-    TimeMonitor tm4(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Add"));
-    result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE,B1MKL,1.0,B2MKL,&Temp1MKL);
-    typename KCRS::execution_space().fence();
-    if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed: "+mkl_error(result));
+      TimeMonitor tm4(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Add"));
+      result = mkl_sparse_d_add(SPARSE_OPERATION_NON_TRANSPOSE, B1MKL, 1.0, B2MKL, &Temp1MKL);
+      typename KCRS::execution_space().fence();
+      if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Add failed: " + mkl_error(result));
     }
 
     // **********************************
     // Multiply A*(B1+B2)
     {
-    TimeMonitor tm5(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Multiply"));
-    result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, Temp1MKL, &CMKL);
-    typename KCRS::execution_space().fence();
-    if(result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply failed: "+mkl_error(result));
+      TimeMonitor tm5(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Multiply"));
+      result = mkl_sparse_spmm(SPARSE_OPERATION_NON_TRANSPOSE, AMKL, Temp1MKL, &CMKL);
+      typename KCRS::execution_space().fence();
+      if (result != SPARSE_STATUS_SUCCESS) throw std::runtime_error("MKL Multiply failed: " + mkl_error(result));
     }
-  }
-  else
+  } else
     throw std::runtime_error("Invalid MKL algorithm");
 
   // **********************************
   // Copy out the data for MKL
-  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 MKL "+algorithm_name+": Copy Out")));
+  tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 MKL " + algorithm_name + ": Copy Out")));
   sparse_index_base_t c_indexing;
   MKL_INT c_rows, c_cols, *rows_start, *rows_end, *columns;
-  double * values;
-  mkl_sparse_d_export_csr(CMKL,&c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
-  size_t cnnz = rows_end[c_rows-1];
-  Kokkos::resize(Ccolind,cnnz);
-  Kokkos::resize(Cvals,cnnz);
-  if(c_rows != A.getLocalNumRows() || c_rows+1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
-  copy_view_n(c_rows,rows_start,Crowptr); Crowptr(c_rows) = rows_end[c_rows-1];
-  copy_view_n(cnnz,columns,Ccolind);
-  copy_view_n(cnnz,values,Cvals);
+  double *values;
+  mkl_sparse_d_export_csr(CMKL, &c_indexing, &c_rows, &c_cols, &rows_start, &rows_end, &columns, &values);
+  size_t cnnz = rows_end[c_rows - 1];
+  Kokkos::resize(Ccolind, cnnz);
+  Kokkos::resize(Cvals, cnnz);
+  if (c_rows != A.getLocalNumRows() || c_rows + 1 != Crowptr.extent(0)) throw std::runtime_error("C row size mismatch");
+  copy_view_n(c_rows, rows_start, Crowptr);
+  Crowptr(c_rows) = rows_end[c_rows - 1];
+  copy_view_n(cnnz, columns, Ccolind);
+  copy_view_n(cnnz, values, Cvals);
 
   Cnc->replaceColMap(Ccolmap_t);
   Cnc->setAllValues(Crowptr,
@@ -305,7 +299,7 @@ void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, co
   mkl_sparse_destroy(B2MKL);
   mkl_sparse_destroy(CMKL);
   mkl_sparse_destroy(Temp1MKL);
-  if(algorithm_name == "MULT_ADD" ) mkl_sparse_destroy(Temp2MKL);
+  if (algorithm_name == "MULT_ADD") mkl_sparse_destroy(Temp2MKL);
 
   tm = Teuchos::null;
   Au->getComm()->barrier();
@@ -313,8 +307,6 @@ void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, co
 
 #endif
 
-
-
 // =========================================================================
 // Tpetra Kernel Testing
 // =========================================================================
@@ -323,8 +315,8 @@ void MM2_MKL(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A, co
 #include "TpetraExt_MatrixMatrix_ExtraKernels_decl.hpp"
 #include "TpetraExt_MatrixMatrix_ExtraKernels_def.hpp"
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void MM2_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A,  const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B1,   Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &B2, Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &C,Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> >&Ccolmap, std::string algorithm_name, int team_work_size) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void MM2_Wrapper(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &A, const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B1, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &B2, Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> &C, Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > &Ccolmap, std::string algorithm_name, int team_work_size) {
 #include <MueLu_UseShortNames.hpp>
   using Teuchos::RCP;
   using Teuchos::rcp;
@@ -337,31 +329,30 @@ void MM2_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A
 
   if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_TPETRA_INST_OPENMP)
-    typedef Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> crs_matrix_type;
-    typedef Tpetra::Import<LocalOrdinal,GlobalOrdinal,Node>           import_type;
-    typedef typename crs_matrix_type::local_matrix_device_type    KCRS;
+    typedef Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> crs_matrix_type;
+    typedef Tpetra::Import<LocalOrdinal, GlobalOrdinal, Node> import_type;
+    typedef typename crs_matrix_type::local_matrix_device_type KCRS;
     typedef typename KCRS::device_type device_t;
     typedef typename KCRS::StaticCrsGraphType graph_t;
     typedef typename graph_t::row_map_type::non_const_type lno_view_t;
     typedef typename graph_t::entries_type::non_const_type lno_nnz_view_t;
     typedef typename KCRS::values_type::non_const_type scalar_view_t;
-    typedef Kokkos::View<LO*, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
-    typedef Tpetra::Map<LO,GO,NO>                                     map_type;
-    typedef typename map_type::local_map_type                         local_map_type;
+    typedef Kokkos::View<LO *, typename lno_view_t::array_layout, typename lno_view_t::device_type> lo_view_t;
+    typedef Tpetra::Map<LO, GO, NO> map_type;
+    typedef typename map_type::local_map_type local_map_type;
     typedef typename Node::execution_space execution_space;
     typedef Kokkos::RangePolicy<execution_space, size_t> range_type;
     LocalOrdinal LO_INVALID = Teuchos::OrdinalTraits<LO>::invalid();
     RCP<const import_type> Cimport;
-    RCP<const crs_matrix_type> Au = Utilities::Op2TpetraCrs(rcp(&A,false));
-    RCP<const crs_matrix_type> B1u = Utilities::Op2TpetraCrs(rcp(&B1,false));
-    RCP<crs_matrix_type> B2u = Teuchos::rcp_const_cast<crs_matrix_type>(Utilities::Op2TpetraCrs(rcp(&B2,false)));
-    RCP<const crs_matrix_type> Cu = Utilities::Op2TpetraCrs(rcp(&C,false));
-    RCP<crs_matrix_type> Cnc = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
-
+    RCP<const crs_matrix_type> Au  = Utilities::Op2TpetraCrs(rcp(&A, false));
+    RCP<const crs_matrix_type> B1u = Utilities::Op2TpetraCrs(rcp(&B1, false));
+    RCP<crs_matrix_type> B2u       = Teuchos::rcp_const_cast<crs_matrix_type>(Utilities::Op2TpetraCrs(rcp(&B2, false)));
+    RCP<const crs_matrix_type> Cu  = Utilities::Op2TpetraCrs(rcp(&C, false));
+    RCP<crs_matrix_type> Cnc       = Teuchos::rcp_const_cast<crs_matrix_type>(Cu);
 
     // **********************************
     // Copy in the data for Kernel Wrapper
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 "+name+": CopyIn")));
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MM2 " + name + ": CopyIn")));
 
     Tpetra::CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node> Aview, Bview;
     Aview.origMatrix   = Au;
@@ -382,10 +373,9 @@ void MM2_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A
     Bview.importMatrix = B2u;
     Bview.importColMap = B2u->getColMap();
 
-    RCP<const Tpetra::Map<LO,GO,Node> > Ccolmap_t = Xpetra::toTpetra(Ccolmap);
+    RCP<const Tpetra::Map<LO, GO, Node> > Ccolmap_t = Xpetra::toTpetra(Ccolmap);
     Cnc->replaceColMap(Ccolmap_t);
 
-
     local_map_type Browmap_local = Bview.origMatrix->getRowMap()->getLocalMap();
     local_map_type Irowmap_local = Bview.importMatrix->getRowMap()->getLocalMap();
 
@@ -395,30 +385,33 @@ void MM2_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A
     local_map_type Ccolmap_local = Ccolmap_t->getLocalMap();
 
     // Bcol2Ccol
-    lo_view_t Bcol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Bcol2Ccol"),Bview.origMatrix->getColMap()->getLocalNumElements());
-    lo_view_t Icol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Icol2Ccol"),Bview.importMatrix->getColMap()->getLocalNumElements());
-    Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::Bcol2Ccol_getGlobalElement",range_type(0,Bview.origMatrix->getColMap()->getLocalNumElements()),KOKKOS_LAMBDA(const LO i) {
-        Bcol2Ccol(i) = Ccolmap_local.getLocalElement(Bcolmap_local.getGlobalElement(i));
-      });
-    Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::Icol2Ccol_getGlobalElement",range_type(0,Bview.importMatrix->getColMap()->getLocalNumElements()),KOKKOS_LAMBDA(const LO i) {
-        Icol2Ccol(i) = Ccolmap_local.getLocalElement(Icolmap_local.getGlobalElement(i));
-      });
+    lo_view_t Bcol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Bcol2Ccol"), Bview.origMatrix->getColMap()->getLocalNumElements());
+    lo_view_t Icol2Ccol(Kokkos::ViewAllocateWithoutInitializing("Icol2Ccol"), Bview.importMatrix->getColMap()->getLocalNumElements());
+    Kokkos::parallel_for(
+        "Tpetra::mult_A_B_newmatrix::Bcol2Ccol_getGlobalElement", range_type(0, Bview.origMatrix->getColMap()->getLocalNumElements()), KOKKOS_LAMBDA(const LO i) {
+          Bcol2Ccol(i) = Ccolmap_local.getLocalElement(Bcolmap_local.getGlobalElement(i));
+        });
+    Kokkos::parallel_for(
+        "Tpetra::mult_A_B_newmatrix::Icol2Ccol_getGlobalElement", range_type(0, Bview.importMatrix->getColMap()->getLocalNumElements()), KOKKOS_LAMBDA(const LO i) {
+          Icol2Ccol(i) = Ccolmap_local.getLocalElement(Icolmap_local.getGlobalElement(i));
+        });
 
     // Acol2Brow
-    lo_view_t targetMapToOrigRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToOrigRow"),Aview.colMap->getLocalNumElements());
-    lo_view_t targetMapToImportRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToImportRow"),Aview.colMap->getLocalNumElements());
-    Kokkos::parallel_for("Tpetra::mult_A_B_newmatrix::construct_tables",range_type(Aview.colMap->getMinLocalIndex(), Aview.colMap->getMaxLocalIndex()+1),KOKKOS_LAMBDA(const LO i) {
-      GO aidx = Acolmap_local.getGlobalElement(i);
-      LO B_LID = Browmap_local.getLocalElement(aidx);
-     if (B_LID != LO_INVALID) {
-        targetMapToOrigRow(i)   = B_LID;
-        targetMapToImportRow(i) = LO_INVALID;
-      } else {
-        LO I_LID = Irowmap_local.getLocalElement(aidx);
-        targetMapToOrigRow(i)   = LO_INVALID;
-        targetMapToImportRow(i) = I_LID;
-     }
-    });
+    lo_view_t targetMapToOrigRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToOrigRow"), Aview.colMap->getLocalNumElements());
+    lo_view_t targetMapToImportRow(Kokkos::ViewAllocateWithoutInitializing("targetMapToImportRow"), Aview.colMap->getLocalNumElements());
+    Kokkos::parallel_for(
+        "Tpetra::mult_A_B_newmatrix::construct_tables", range_type(Aview.colMap->getMinLocalIndex(), Aview.colMap->getMaxLocalIndex() + 1), KOKKOS_LAMBDA(const LO i) {
+          GO aidx  = Acolmap_local.getGlobalElement(i);
+          LO B_LID = Browmap_local.getLocalElement(aidx);
+          if (B_LID != LO_INVALID) {
+            targetMapToOrigRow(i)   = B_LID;
+            targetMapToImportRow(i) = LO_INVALID;
+          } else {
+            LO I_LID                = Irowmap_local.getLocalElement(aidx);
+            targetMapToOrigRow(i)   = LO_INVALID;
+            targetMapToImportRow(i) = I_LID;
+          }
+        });
     tm = Teuchos::null;
     Au->getComm()->barrier();
 
@@ -426,36 +419,34 @@ void MM2_Wrapper(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> &A
     // Multiply
     Teuchos::RCP<Teuchos::ParameterList> params = Teuchos::rcp(new Teuchos::ParameterList);
     params->set("openmp: algorithm", algorithm_name);
-    params->set("openmp: team work size",team_work_size);
+    params->set("openmp: team work size", team_work_size);
 
-    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(std::string("MM2 ")+name+": Kernel")));
-    Tpetra::MMdetails::KernelWrappers<SC,LO,GO,Node,lno_nnz_view_t>::mult_A_B_newmatrix_kernel_wrapper(Aview,Bview,targetMapToOrigRow,targetMapToImportRow,Bcol2Ccol,Icol2Ccol,*Cnc,Cimport,name,params);
+    tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer(std::string("MM2 ") + name + ": Kernel")));
+    Tpetra::MMdetails::KernelWrappers<SC, LO, GO, Node, lno_nnz_view_t>::mult_A_B_newmatrix_kernel_wrapper(Aview, Bview, targetMapToOrigRow, targetMapToImportRow, Bcol2Ccol, Icol2Ccol, *Cnc, Cimport, name, params);
 
     tm = Teuchos::null;
     Au->getComm()->barrier();
 
 #endif
-
   }
 }
 
 // =========================================================================
 // =========================================================================
 // =========================================================================
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib &lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; // reference count pointers
+  using std::endl;
+  using Teuchos::RCP;  // reference count pointers
   using Teuchos::rcp;
   using Teuchos::TimeMonitor;
-  using std::endl;
-
 
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -464,7 +455,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> pOut = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *pOut;
+    Teuchos::FancyOStream &out      = *pOut;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
@@ -472,69 +463,77 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // =========================================================================
     //    GO nx = 50, ny = 50, nz = 50;
     // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace3D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
 
-    bool   printTimings = true;  clp.setOption("timings", "notimings",  &printTimings, "print timings to screen");
-    int    nrepeat      = 100;   clp.setOption("nrepeat",               &nrepeat,      "repeat the experiment N times");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int nrepeat = 100;
+    clp.setOption("nrepeat", &nrepeat, "repeat the experiment N times");
     // the kernels
-    bool do_mkl_ma      = true;
-    bool do_mkl_am      = true;
-    bool do_kk_mem   = true;
-    bool do_kk_dense = true;
-    bool do_kk_default = true;
-    bool do_ltg        = true;
-    double percent_rows_to_remove = 10.0; clp.setOption("remove-rows",               &percent_rows_to_remove,      "% of rows to remove from B to make B1 & B2");
-
-    #ifndef HAVE_MUELU_MKL
-      do_mkl_ma = false;
-      do_mkl_am = false;
-    #endif
-    clp.setOption("mkl_ma",      "nomkl_ma",    &do_mkl_ma,     "Evaluate MKL Mult-Add");
-    clp.setOption("mkl_am",      "nomkl_am",    &do_mkl_am,     "Evaluate MKL Add-Mult");
-    clp.setOption("kk_mem",     "nokk_mem",     &do_kk_mem,     "Evaluate KK Mem");
-    clp.setOption("kk_dense",   "nokk_dense",   &do_kk_dense,   "Evaluate KK Dense");
+    bool do_mkl_ma                = true;
+    bool do_mkl_am                = true;
+    bool do_kk_mem                = true;
+    bool do_kk_dense              = true;
+    bool do_kk_default            = true;
+    bool do_ltg                   = true;
+    double percent_rows_to_remove = 10.0;
+    clp.setOption("remove-rows", &percent_rows_to_remove, "% of rows to remove from B to make B1 & B2");
+
+#ifndef HAVE_MUELU_MKL
+    do_mkl_ma = false;
+    do_mkl_am = false;
+#endif
+    clp.setOption("mkl_ma", "nomkl_ma", &do_mkl_ma, "Evaluate MKL Mult-Add");
+    clp.setOption("mkl_am", "nomkl_am", &do_mkl_am, "Evaluate MKL Add-Mult");
+    clp.setOption("kk_mem", "nokk_mem", &do_kk_mem, "Evaluate KK Mem");
+    clp.setOption("kk_dense", "nokk_dense", &do_kk_dense, "Evaluate KK Dense");
     clp.setOption("kk_default", "nokk_default", &do_kk_default, "Evaluate KK Default");
-    clp.setOption("ltg",        "noltg",        &do_ltg,        "Evaluate LTG");
-    int kk_team_work_size=16;
+    clp.setOption("ltg", "noltg", &do_ltg, "Evaluate LTG");
+    int kk_team_work_size = 16;
 
-    std::string matrixFileNameA = "A.mm"; clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
-    std::string matrixFileNameB = "B.mm"; clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
+    std::string matrixFileNameA = "A.mm";
+    clp.setOption("matrixfileA", &matrixFileNameA, "matrix market file containing matrix");
+    std::string matrixFileNameB = "B.mm";
+    clp.setOption("matrixfileB", &matrixFileNameB, "matrix market file containing matrix");
 
-    
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    if(percent_rows_to_remove < 0.0) percent_rows_to_remove = 0.0;
-    if(percent_rows_to_remove > 100.0) percent_rows_to_remove = 100.0;
+    if (percent_rows_to_remove < 0.0) percent_rows_to_remove = 0.0;
+    if (percent_rows_to_remove > 100.0) percent_rows_to_remove = 100.0;
 
-    #ifndef HAVE_MUELU_MKL
+#ifndef HAVE_MUELU_MKL
     if (do_mkl_ma || do_mkl_am) {
       out << "MKL was requested, but this kernel is not available. Disabling..." << endl;
       do_mkl_ma = false;
       do_mkl_am = false;
     }
-    #endif
+#endif
 
     // simple hack to randomize order of experiments
-    enum class Experiments { MKL_MULT_ADD=0, MKL_ADD_MULT, KK_MEM, KK_DENSE, KK_DEFAULT, LTG };
+    enum class Experiments { MKL_MULT_ADD = 0,
+                             MKL_ADD_MULT,
+                             KK_MEM,
+                             KK_DENSE,
+                             KK_DEFAULT,
+                             LTG };
     std::vector<Experiments> my_experiments;
     // add the experiments we will run
-  
-    #ifdef HAVE_MUELU_MKL
-    if (do_mkl_ma) my_experiments.push_back(Experiments::MKL_MULT_ADD);   // MKL Mult-Add
-    if (do_mkl_am) my_experiments.push_back(Experiments::MKL_ADD_MULT);   // MKL Add-Mult
-    #endif
+
+#ifdef HAVE_MUELU_MKL
+    if (do_mkl_ma) my_experiments.push_back(Experiments::MKL_MULT_ADD);  // MKL Mult-Add
+    if (do_mkl_am) my_experiments.push_back(Experiments::MKL_ADD_MULT);  // MKL Add-Mult
+#endif
 
     // assume these are available
-    if (do_kk_mem)     my_experiments.push_back(Experiments::KK_MEM);     // KK Mem
-    if (do_kk_dense)   my_experiments.push_back(Experiments::KK_DENSE);   // KK Dense
-    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT); // KK Default
-    if (do_ltg)        my_experiments.push_back(Experiments::LTG);        // LTG
+    if (do_kk_mem) my_experiments.push_back(Experiments::KK_MEM);          // KK Mem
+    if (do_kk_dense) my_experiments.push_back(Experiments::KK_DENSE);      // KK Dense
+    if (do_kk_default) my_experiments.push_back(Experiments::KK_DEFAULT);  // KK Default
+    if (do_ltg) my_experiments.push_back(Experiments::LTG);                // LTG
 
     out << "========================================================" << endl
         << xpetraParameters
@@ -556,12 +555,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         << "========================================================" << endl;
 
 #if defined(HAVE_TPETRA_INST_OPENMP)
-    out<< "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = "<<Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency()<<endl
-       << "========================================================" << endl;
+    out << "Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() = " << Tpetra::KokkosCompat::KokkosOpenMPWrapperNode::execution_space().concurrency() << endl
+        << "========================================================" << endl;
 #endif
 
     // At the moment, this test only runs on one MPI rank
-    if(comm->getSize() != 1) exit(1);
+    if (comm->getSize() != 1) exit(1);
 
     // =========================================================================
     // Problem construction
@@ -570,41 +569,40 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     comm->barrier();
 
+    RCP<Matrix> A = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameA), lib, comm);
+    RCP<Matrix> B = Xpetra::IO<SC, LO, GO, Node>::Read(std::string(matrixFileNameB), lib, comm);
+    RCP<Matrix> B1, B2, C;
 
-    RCP<Matrix> A = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameA), lib, comm);
-    RCP<Matrix> B = Xpetra::IO<SC,LO,GO,Node>::Read(std::string(matrixFileNameB), lib, comm);
-    RCP<Matrix> B1,B2,C;
-
-    
     RCP<const Map> Browmap = B->getRowMap();
     RCP<const Map> Bcolmap = B->getColMap();
-    
+
     // Do some surgery to generate B1 and B2
     const GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
-    size_t brows = Browmap->getLocalNumElements();
-    size_t b2rows = std::max((size_t)0,std::min((size_t)(brows * percent_rows_to_remove/100.0),brows));
-    size_t b1rows = brows - b2rows;
+    size_t brows        = Browmap->getLocalNumElements();
+    size_t b2rows       = std::max((size_t)0, std::min((size_t)(brows * percent_rows_to_remove / 100.0), brows));
+    size_t b1rows       = brows - b2rows;
     {
-
       Teuchos::Array<GO> b1list(b1rows), b2list(b2rows);
-      for(size_t i=0,i1=0, i2=0; i<brows; i++) {
-        if(i<b1rows) {b1list[i1] = Browmap->getGlobalElement(i); i1++;}
-        else {b2list[i2] = Browmap->getGlobalElement(i); i2++;}
+      for (size_t i = 0, i1 = 0, i2 = 0; i < brows; i++) {
+        if (i < b1rows) {
+          b1list[i1] = Browmap->getGlobalElement(i);
+          i1++;
+        } else {
+          b2list[i2] = Browmap->getGlobalElement(i);
+          i2++;
+        }
       }
-      
-      RCP<const Map> B1rowmap = Xpetra::MapFactory<LO,GO,Node>::Build(lib,GO_INVALID,b1list(),0,comm);
-      RCP<const Map> B2rowmap = Xpetra::MapFactory<LO,GO,Node>::Build(lib,GO_INVALID,b2list(),0,comm);
-      
 
-      RCP<const Import> B1import = Xpetra::ImportFactory<LO,GO,Node>::Build(Browmap,B1rowmap);
-      RCP<const Import> B2import = Xpetra::ImportFactory<LO,GO,Node>::Build(Browmap,B2rowmap);
+      RCP<const Map> B1rowmap = Xpetra::MapFactory<LO, GO, Node>::Build(lib, GO_INVALID, b1list(), 0, comm);
+      RCP<const Map> B2rowmap = Xpetra::MapFactory<LO, GO, Node>::Build(lib, GO_INVALID, b2list(), 0, comm);
+
+      RCP<const Import> B1import = Xpetra::ImportFactory<LO, GO, Node>::Build(Browmap, B1rowmap);
+      RCP<const Import> B2import = Xpetra::ImportFactory<LO, GO, Node>::Build(Browmap, B2rowmap);
       RCP<const Map> dummy;
-      B1 = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(B,*B1import,dummy,B1rowmap);
-      B2 = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(B,*B2import,dummy,B2rowmap);
+      B1 = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(B, *B1import, dummy, B1rowmap);
+      B2 = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(B, *B2import, dummy, B2rowmap);
     }
 
-
-
     globalTimeMonitor = Teuchos::null;
     comm->barrier();
 
@@ -621,87 +619,84 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // random source
     std::random_device rd;
-    std::mt19937 random_source (rd());
+    std::mt19937 random_source(rd());
 
     // no need for a barrier, because the randomization process uses a collective.
-    if ( !my_experiments.empty() ) {
+    if (!my_experiments.empty()) {
+      for (int i = 0; i < nrepeat; i++) {
+        // randomize the experiments
+        if (comm->getRank() == 0) {
+          std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
+        }
 
-    for (int i=0; i<nrepeat; i++) {
-      
-      // randomize the experiments
-      if (comm->getRank() == 0 ) {
-        std::shuffle(my_experiments.begin(), my_experiments.end(), random_source);
-      }
+        // Broadcast this ordering to the other processes
+        comm->broadcast(0,
+                        static_cast<int>(sizeof(Experiments::LTG) * my_experiments.size()),
+                        reinterpret_cast<char *>(my_experiments.data()));
 
-      // Broadcast this ordering to the other processes
-      comm->broadcast(0,
-                      static_cast<int>(sizeof(Experiments::LTG)*my_experiments.size()),
-                      reinterpret_cast<char *>(my_experiments.data()));
-
-      // loop over the randomized experiments
-      for (const auto& experiment_id : my_experiments) {
-        switch (experiment_id) {
- 
-        // MKL MULT_ADD
-        case Experiments::MKL_MULT_ADD:
-         #ifdef HAVE_MUELU_MKL
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Total"));
-            MM2_MKL(*A,*B1,*B2,*C,Bcolmap,std::string("MULT_ADD"));
-          }
-          #endif
-          break;
-        // MKL ADD_MULT
-        case Experiments::MKL_ADD_MULT:
-         #ifdef HAVE_MUELU_MKL
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Total"));
-            MM2_MKL(*A,*B1,*B2,*C,Bcolmap,std::string("ADD_MULT"));
-          }
-          #endif
-          break;
-        // KK Algorithms (KK Memory)
-        case Experiments::KK_MEM:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK_MEMORY: Total"));
-            MM2_Wrapper(*A,*B1,*B2,*C,Bcolmap,std::string("SPGEMM_KK_MEMORY"),kk_team_work_size);
-          }
-          break;
-        // KK Algorithms (KK Dense)
-        case Experiments::KK_DENSE:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK_DENSE: Total"));
-            MM2_Wrapper(*A,*B1,*B2,*C,Bcolmap,std::string("SPGEMM_KK_DENSE"),kk_team_work_size);
-          }
-          break;
-        // KK Algorithms (KK Default)
-        case Experiments::KK_DEFAULT:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK: Total"));
-            MM2_Wrapper(*A,*B1,*B2,*C,Bcolmap,std::string("SPGEMM_KK"),kk_team_work_size);
-          }
-          break;
-        // LTG
-        case Experiments::LTG:
-          C = Xpetra::MatrixFactory<SC,LO,GO,Node>::Build(A->getRowMap(),0);
-          {
-            TimeMonitor t(*TimeMonitor::getNewTimer("MM2 LTG: Total"));
-            MM2_Wrapper(*A,*B1,*B2,*C,Bcolmap,std::string("LTG"),kk_team_work_size);
+        // loop over the randomized experiments
+        for (const auto &experiment_id : my_experiments) {
+          switch (experiment_id) {
+            // MKL MULT_ADD
+            case Experiments::MKL_MULT_ADD:
+#ifdef HAVE_MUELU_MKL
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 MKL MULT_ADD: Total"));
+                MM2_MKL(*A, *B1, *B2, *C, Bcolmap, std::string("MULT_ADD"));
+              }
+#endif
+              break;
+            // MKL ADD_MULT
+            case Experiments::MKL_ADD_MULT:
+#ifdef HAVE_MUELU_MKL
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 MKL ADD_MULT: Total"));
+                MM2_MKL(*A, *B1, *B2, *C, Bcolmap, std::string("ADD_MULT"));
+              }
+#endif
+              break;
+            // KK Algorithms (KK Memory)
+            case Experiments::KK_MEM:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK_MEMORY: Total"));
+                MM2_Wrapper(*A, *B1, *B2, *C, Bcolmap, std::string("SPGEMM_KK_MEMORY"), kk_team_work_size);
+              }
+              break;
+            // KK Algorithms (KK Dense)
+            case Experiments::KK_DENSE:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK_DENSE: Total"));
+                MM2_Wrapper(*A, *B1, *B2, *C, Bcolmap, std::string("SPGEMM_KK_DENSE"), kk_team_work_size);
+              }
+              break;
+            // KK Algorithms (KK Default)
+            case Experiments::KK_DEFAULT:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 SPGEMM_KK: Total"));
+                MM2_Wrapper(*A, *B1, *B2, *C, Bcolmap, std::string("SPGEMM_KK"), kk_team_work_size);
+              }
+              break;
+            // LTG
+            case Experiments::LTG:
+              C = Xpetra::MatrixFactory<SC, LO, GO, Node>::Build(A->getRowMap(), 0);
+              {
+                TimeMonitor t(*TimeMonitor::getNewTimer("MM2 LTG: Total"));
+                MM2_Wrapper(*A, *B1, *B2, *C, Bcolmap, std::string("LTG"), kk_team_work_size);
+              }
+              break;
+
+            default:
+              std::cerr << "Unknown experiment ID encountered: " << (int)experiment_id << std::endl;
           }
-          break;
-
-        default:
-          std::cerr << "Unknown experiment ID encountered: " << (int) experiment_id << std::endl;
-        }
-        comm->barrier();
-      }// end random exp loop
-    } // end repeat
-    } // end ! my_experiments.empty()
+          comm->barrier();
+        }  // end random exp loop
+      }    // end repeat
+    }      // end ! my_experiments.empty()
 
     if (printTimings) {
       TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
@@ -711,17 +706,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
-
-
-
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/simple1D/Emin.cpp b/packages/muelu/test/simple1D/Emin.cpp
index 7d911108d36c..988e7fbf0d68 100644
--- a/packages/muelu/test/simple1D/Emin.cpp
+++ b/packages/muelu/test/simple1D/Emin.cpp
@@ -99,118 +99,115 @@
 #include "BelosConfigDefs.hpp"
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
-#include "BelosMueLuAdapter.hpp" // this header defines Belos::MueLuOp()
+#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
 
 #define NEUMANN
 #define EMIN
 
+using Teuchos::ArrayRCP;
 using Teuchos::RCP;
 using Teuchos::rcp;
-using Teuchos::ArrayRCP;
 
 namespace MueLuTests {
 
 #include "MueLu_UseShortNames.hpp"
 
-  RCP<SmootherPrototype> gimmeGaussSeidelProto(Xpetra::UnderlyingLib lib) {
-
-    RCP<SmootherPrototype> smooProto;
-    Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 1);
-    ifpackList.set("relaxation: damping factor", (SC) 1.0);
+RCP<SmootherPrototype> gimmeGaussSeidelProto(Xpetra::UnderlyingLib lib) {
+  RCP<SmootherPrototype> smooProto;
+  Teuchos::ParameterList ifpackList;
+  ifpackList.set("relaxation: sweeps", (LO)1);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
 
-    if (lib == Xpetra::UseEpetra) {
+  if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
 #if defined(HAVE_MUELU_SERIAL)
-      ifpackList.set("relaxation: type", "symmetric Gauss-Seidel");
-      smooProto = MueLu::GetIfpackSmoother<SC,LO,GO,NO>("point relaxation stand-alone", ifpackList);
+    ifpackList.set("relaxation: type", "symmetric Gauss-Seidel");
+    smooProto = MueLu::GetIfpackSmoother<SC, LO, GO, NO>("point relaxation stand-alone", ifpackList);
 #else
-      throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelProto: IfpackSmoother only available with SerialNode."));
+    throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelProto: IfpackSmoother only available with SerialNode."));
 #endif
 #endif
-    } else if (lib == Xpetra::UseTpetra) {
+  } else if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_IFPACK2)
-      ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-      smooProto = rcp( new Ifpack2Smoother("RELAXATION", ifpackList) );
+    ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
+    smooProto = rcp(new Ifpack2Smoother("RELAXATION", ifpackList));
 #endif
-    }
-    if (smooProto == Teuchos::null) {
-      throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelSmoother: smoother error"));
-    }
-
-    return smooProto;
+  }
+  if (smooProto == Teuchos::null) {
+    throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelSmoother: smoother error"));
   }
 
-  RCP<SmootherPrototype> gimmeCoarseProto(Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
-    RCP<SmootherPrototype> coarseProto;
-    if (lib == Xpetra::UseEpetra) {
+  return smooProto;
+}
+
+RCP<SmootherPrototype> gimmeCoarseProto(Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
+  RCP<SmootherPrototype> coarseProto;
+  if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
 #if defined(HAVE_MUELU_SERIAL)
-      if (rank == 0) std::cout << "CoarseGrid: AMESOS" << std::endl;
-      Teuchos::ParameterList amesosList;
-      amesosList.set("PrintTiming",true);
-      coarseProto = MueLu::GetAmesosSmoother<SC,LO,GO,NO>("Amesos_Klu", amesosList);
+    if (rank == 0) std::cout << "CoarseGrid: AMESOS" << std::endl;
+    Teuchos::ParameterList amesosList;
+    amesosList.set("PrintTiming", true);
+    coarseProto = MueLu::GetAmesosSmoother<SC, LO, GO, NO>("Amesos_Klu", amesosList);
 #else
-      throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelProto: AmesosSmoother only available with SerialNode."));
+    throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelProto: AmesosSmoother only available with SerialNode."));
 #endif
 #endif
-    } else if (lib == Xpetra::UseTpetra) {
-      if (coarseSolver=="amesos2") {
+  } else if (lib == Xpetra::UseTpetra) {
+    if (coarseSolver == "amesos2") {
 #if defined(HAVE_MUELU_AMESOS2)
-        if (rank == 0) std::cout << "CoarseGrid: AMESOS2" << std::endl;
-        Teuchos::ParameterList paramList; //unused
-        coarseProto = rcp( new Amesos2Smoother("Superlu", paramList) );
+      if (rank == 0) std::cout << "CoarseGrid: AMESOS2" << std::endl;
+      Teuchos::ParameterList paramList;  // unused
+      coarseProto = rcp(new Amesos2Smoother("Superlu", paramList));
 #else
-        std::cout  << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
-        return Teuchos::null; // TODO test for exception //EXIT_FAILURE;
-#endif // HAVE_MUELU_AMESOS2
-      } else if(coarseSolver=="ifpack2") {
+      std::cout << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
+      return Teuchos::null;  // TODO test for exception //EXIT_FAILURE;
+#endif  // HAVE_MUELU_AMESOS2
+    } else if (coarseSolver == "ifpack2") {
 #if defined(HAVE_MUELU_IFPACK2)
-        if (rank == 0) std::cout << "CoarseGrid: IFPACK2" << std::endl;
-        Teuchos::ParameterList ifpack2List;
-        ifpack2List.set("fact: ilut level-of-fill",(double)99); // TODO ??
-        ifpack2List.set("fact: drop tolerance", 0.0);
-        ifpack2List.set("fact: absolute threshold", 0.0);
-        ifpack2List.set("fact: relative threshold", 1.0);
-        coarseProto = rcp( new Ifpack2Smoother("ILUT",ifpack2List) );
+      if (rank == 0) std::cout << "CoarseGrid: IFPACK2" << std::endl;
+      Teuchos::ParameterList ifpack2List;
+      ifpack2List.set("fact: ilut level-of-fill", (double)99);  // TODO ??
+      ifpack2List.set("fact: drop tolerance", 0.0);
+      ifpack2List.set("fact: absolute threshold", 0.0);
+      ifpack2List.set("fact: relative threshold", 1.0);
+      coarseProto = rcp(new Ifpack2Smoother("ILUT", ifpack2List));
 #else
-        std::cout  << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
-        //TODO        TEUCHOS_TEST_FOR_EXCEPTION
-        return Teuchos::null;
+      std::cout << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
+      // TODO        TEUCHOS_TEST_FOR_EXCEPTION
+      return Teuchos::null;
 #endif
-      } else {
-        std::cout  << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
-        return Teuchos::null;
-      }
-
-    }
-    if (coarseProto == Teuchos::null) {
-      throw(MueLu::Exceptions::RuntimeError("main: coarse smoother error"));
+    } else {
+      std::cout << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
+      return Teuchos::null;
     }
-
-    return coarseProto;
+  }
+  if (coarseProto == Teuchos::null) {
+    throw(MueLu::Exceptions::RuntimeError("main: coarse smoother error"));
   }
 
+  return coarseProto;
+}
 
-  RCP<SmootherPrototype> gimmeMergedSmoother(int nSmoothers, Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
-    ArrayRCP<RCP<SmootherPrototype> > smootherList(nSmoothers);
-
-    for (int i=0; i<nSmoothers; i++)
-      smootherList[i] = gimmeGaussSeidelProto(lib);
+RCP<SmootherPrototype> gimmeMergedSmoother(int nSmoothers, Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
+  ArrayRCP<RCP<SmootherPrototype> > smootherList(nSmoothers);
 
-    return rcp (new MergedSmoother(smootherList));
-    //verbose mode: return rcp (new MergedSmoother(smootherList, true));
-  }
+  for (int i = 0; i < nSmoothers; i++)
+    smootherList[i] = gimmeGaussSeidelProto(lib);
 
+  return rcp(new MergedSmoother(smootherList));
+  // verbose mode: return rcp (new MergedSmoother(smootherList, true));
 }
 
-int main(int argc, char *argv[]) {
+}  // namespace MueLuTests
+
+int main(int argc, char* argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
   using namespace MueLuTests;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
 
   bool success = false;
   bool verbose = true;
@@ -225,26 +222,26 @@ int main(int argc, char *argv[]) {
 
     // Default is Laplace1D with nx = 8748.
     // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-    //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-    Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+    // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
     // custom parameters
-    int nSmoothers=2;
-    LO maxLevels = 3;
-    LO its=10;
-    std::string coarseSolver="ifpack2";
+    int nSmoothers           = 2;
+    LO maxLevels             = 3;
+    LO its                   = 10;
+    std::string coarseSolver = "ifpack2";
     // std::string coarseSolver="amesos2";
-    clp.setOption("nSmoothers",&nSmoothers,"number of Gauss-Seidel smoothers in the MergedSmoothers");
-    clp.setOption("maxLevels",&maxLevels,"maximum number of levels allowed. If 1, then a MergedSmoother is used on the coarse grid");
-    clp.setOption("its",&its,"number of multigrid cycles");
-    clp.setOption("coarseSolver",&coarseSolver,"amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
+    clp.setOption("nSmoothers", &nSmoothers, "number of Gauss-Seidel smoothers in the MergedSmoothers");
+    clp.setOption("maxLevels", &maxLevels, "maximum number of levels allowed. If 1, then a MergedSmoother is used on the coarse grid");
+    clp.setOption("its", &its, "number of multigrid cycles");
+    clp.setOption("coarseSolver", &coarseSolver, "amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     matrixParameters.check();
@@ -261,8 +258,8 @@ int main(int argc, char *argv[]) {
     /* CREATE INITIAL MATRIX                                                          */
     /**********************************************************************************/
     const RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
     RCP<Matrix> Op = Pr->BuildMatrix();
 
 #ifdef NEUMANN
@@ -279,34 +276,36 @@ int main(int argc, char *argv[]) {
 
     size_t myRank = Op->getRowMap()->getComm()->getRank();
     size_t nCpus  = Op->getRowMap()->getComm()->getSize();
-    if (myRank == 0) { // JG TODO: can we use rowMap->isNodeLocalElement(0) instead for more genericity?
-      //LO firstRow = 0;
-      newValues[0] = 1.0; newValues[1] = -1.0;
+    if (myRank == 0) {  // JG TODO: can we use rowMap->isNodeLocalElement(0) instead for more genericity?
+      // LO firstRow = 0;
+      newValues[0] = 1.0;
+      newValues[1] = -1.0;
       Op->getLocalRowView(0, indices, values);
       Op->replaceLocalValues(0, indices, newValues);
     }
-    if (myRank == nCpus-1) { // JG TODO: can we use rowMap->isNodeLocalElement(lastRow) instead for more genericity?
-      LO lastRow = Op->getLocalNumRows()-1;
-      newValues[0] = -1.0; newValues[1] = 1.0;
+    if (myRank == nCpus - 1) {  // JG TODO: can we use rowMap->isNodeLocalElement(lastRow) instead for more genericity?
+      LO lastRow   = Op->getLocalNumRows() - 1;
+      newValues[0] = -1.0;
+      newValues[1] = 1.0;
       Op->getLocalRowView(lastRow, indices, values);
       Op->replaceLocalValues(lastRow, indices, newValues);
     }
 
     Op->fillComplete();
-#endif // NEUMANN
+#endif  // NEUMANN
 
     /**********************************************************************************/
     /*                                                                                */
     /**********************************************************************************/
 
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (SC) 1.0);
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+    nullSpace->putScalar((SC)1.0);
     Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
     nullSpace->norm1(norms);
     if (comm->getRank() == 0)
       std::cout << "||NS|| = " << norms[0] << std::endl;
 
-    RCP<MueLu::Hierarchy<SC,LO,GO,NO> > H = rcp( new Hierarchy() );
+    RCP<MueLu::Hierarchy<SC, LO, GO, NO> > H = rcp(new Hierarchy());
     H->SetDefaultVerbLevel(MueLu::Extreme);
     H->IsPreconditioner(false);
 
@@ -318,8 +317,8 @@ int main(int argc, char *argv[]) {
     FactoryManager M;
 
     M.SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));
-    M.SetFactory("Ptent",      rcp(new TentativePFactory()));
-    M.SetFactory("P",          rcp(new SaPFactory()));
+    M.SetFactory("Ptent", rcp(new TentativePFactory()));
+    M.SetFactory("P", rcp(new SaPFactory()));
 
 #ifdef EMIN
     // Energy-minimization
@@ -329,19 +328,19 @@ int main(int argc, char *argv[]) {
 #else
     PatternFact->SetFactory("P", M.GetFactory("P"));
 #endif
-    M.SetFactory("Ppattern",   PatternFact);
+    M.SetFactory("Ppattern", PatternFact);
 
     RCP<EminPFactory> EminPFact = rcp(new EminPFactory());
     EminPFact->SetFactory("P", M.GetFactory("Ptent"));
-    M.SetFactory("P",          EminPFact);
+    M.SetFactory("P", EminPFact);
 
     RCP<NullspacePresmoothFactory> NullPreFact = rcp(new NullspacePresmoothFactory());
     NullPreFact->SetFactory("Nullspace", M.GetFactory("Nullspace"));
-    M.SetFactory("Nullspace",  NullPreFact);
+    M.SetFactory("Nullspace", NullPreFact);
 #endif
 
     RCP<SmootherPrototype> smooProto = gimmeMergedSmoother(nSmoothers, xpetraParameters.GetLib(), coarseSolver, comm->getRank());
-    M.SetFactory("Smoother",   rcp(new SmootherFactory(smooProto)));
+    M.SetFactory("Smoother", rcp(new SmootherFactory(smooProto)));
 
     Teuchos::ParameterList status;
 
@@ -357,23 +356,23 @@ int main(int argc, char *argv[]) {
 #ifdef NEUMANN
     // Use coarse level projection solver
     RCP<SmootherPrototype> projectedSolver = rcp(new ProjectorSmoother(coarseProto));
-    RCP<SmootherFactory>   coarseSolveFact = rcp(new SmootherFactory(projectedSolver, Teuchos::null));
+    RCP<SmootherFactory> coarseSolveFact   = rcp(new SmootherFactory(projectedSolver, Teuchos::null));
 #else
-    RCP<SmootherFactory> coarseSolveFact   = rcp(new SmootherFactory(coarseProto, Teuchos::null));
+    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
 #endif
     M.SetFactory("CoarseSolver", coarseSolveFact);
 
     H->EnableGraphDumping("graph.dot", 2);
 
     H->Setup(M, 0, maxLevels);
-    //if (comm->getRank() == 0) {
-    //  std::cout  << "======================\n Multigrid statistics \n======================" << std::endl;
-    //  status.print(std::cout,Teuchos::ParameterList::PrintOptions().indent(2));
-    //}
+    // if (comm->getRank() == 0) {
+    //   std::cout  << "======================\n Multigrid statistics \n======================" << std::endl;
+    //   status.print(std::cout,Teuchos::ParameterList::PrintOptions().indent(2));
+    // }
 
     // Define RHS
-    RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
     X->setSeed(846930886);
     X->randomize();
@@ -381,14 +380,13 @@ int main(int argc, char *argv[]) {
     if (comm->getRank() == 0)
       std::cout << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-
-    Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
     // Use AMG directly as an iterative method
     {
-      X->putScalar( (SC) 0.0);
+      X->putScalar((SC)0.0);
 
-      H->Iterate(*RHS,*X,its);
+      H->Iterate(*RHS, *X, its);
 
       X->norm2(norms);
       if (comm->getRank() == 0)
@@ -399,5 +397,5 @@ int main(int argc, char *argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/test/simple1D/aggregation.cpp b/packages/muelu/test/simple1D/aggregation.cpp
index b4f76f3c9abc..6cf43f6dfed6 100644
--- a/packages/muelu/test/simple1D/aggregation.cpp
+++ b/packages/muelu/test/simple1D/aggregation.cpp
@@ -78,7 +78,7 @@ int main(int argc, char *argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   /**********************************************************************************/
@@ -89,21 +89,21 @@ int main(int argc, char *argv[]) {
 
   // Default is Laplace1D with nx = 8748.
   // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-  //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-  Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+  // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
   // custom parameters
   LO maxLevels = 3;
-  LO its=10;
-  clp.setOption("maxLevels",&maxLevels,"maximum number of levels allowed");
-  clp.setOption("its",&its,"number of multigrid cycles");
+  LO its       = 10;
+  clp.setOption("maxLevels", &maxLevels, "maximum number of levels allowed");
+  clp.setOption("its", &its, "number of multigrid cycles");
 
-  switch (clp.parse(argc,argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   matrixParameters.check();
@@ -120,8 +120,8 @@ int main(int argc, char *argv[]) {
   /* CREATE INITIAL MATRIX                                                          */
   /**********************************************************************************/
   const RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
   RCP<Matrix> Op = Pr->BuildMatrix();
   /**********************************************************************************/
   /*                                                                                */
@@ -133,12 +133,11 @@ int main(int argc, char *argv[]) {
   aggOptions.SetMaxNeighAlreadySelected(0);
   aggOptions.SetOrdering("natural");
   UncoupledAggregationFactory aggFact(aggOptions);
-  RCP<Graph> graph = rcp(new Graph(Op->getCrsGraph(), "someGraphLabel"));
-  double t0 = MPI_Wtime();
+  RCP<Graph> graph           = rcp(new Graph(Op->getCrsGraph(), "someGraphLabel"));
+  double t0                  = MPI_Wtime();
   RCP<Aggregates> aggregates = aggFact.Build(*graph);
-  double t1 = MPI_Wtime() - t0;
-  printf("Aggregation time only: %g\n",t1);
+  double t1                  = MPI_Wtime() - t0;
+  printf("Aggregation time only: %g\n", t1);
 
   return EXIT_SUCCESS;
-
 }
diff --git a/packages/muelu/test/simple1D/amesos2.cpp b/packages/muelu/test/simple1D/amesos2.cpp
index 9adba196ea4c..79760c20c7c1 100644
--- a/packages/muelu/test/simple1D/amesos2.cpp
+++ b/packages/muelu/test/simple1D/amesos2.cpp
@@ -82,13 +82,13 @@
 #include "BelosConfigDefs.hpp"
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
-#include "BelosMueLuAdapter.hpp" // this header defines Belos::MueLuOp()
+#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
 
 int main(int argc, char *argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   /**********************************************************************************/
@@ -99,23 +99,23 @@ int main(int argc, char *argv[]) {
 
   // Default is Laplace1D with nx = 8748.
   // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-  //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-  Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+  // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
   // custom parameters
-  LO maxLevels = 1;
-  LO its=2;
-  std::string coarseSolver="amesos2";
-  clp.setOption("maxLevels",&maxLevels,"maximum number of levels allowed");
-  clp.setOption("its",&its,"number of multigrid cycles");
-  clp.setOption("coarseSolver",&coarseSolver,"amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
-
-  switch (clp.parse(argc,argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  LO maxLevels             = 1;
+  LO its                   = 2;
+  std::string coarseSolver = "amesos2";
+  clp.setOption("maxLevels", &maxLevels, "maximum number of levels allowed");
+  clp.setOption("its", &its, "number of multigrid cycles");
+  clp.setOption("coarseSolver", &coarseSolver, "amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
+
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   matrixParameters.check();
@@ -128,31 +128,31 @@ int main(int argc, char *argv[]) {
     // TODO: print custom parameters
   }
 
-
 #ifdef FOR_PARALLEL_DEBUGGING
-  //Utils::BreakForDebugger(*comm);
+  // Utils::BreakForDebugger(*comm);
 
   LO mypid = comm->getRank();
 
-  if (mypid  == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
-  for (LO i = 0; i <comm->getSize(); i++) {
-    if (i == mypid ) {
+  if (mypid == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
+  for (LO i = 0; i < comm->getSize(); i++) {
+    if (i == mypid) {
       char buf[80];
       char hostname[80];
       gethostname(hostname, sizeof(hostname));
       LO pid = getpid();
       sprintf(buf, "Host: %s\tMPI rank: %d,\tPID: %d\n\tattach %d\n\tcontinue\n",
-          hostname, mypid, pid, pid);
-      printf("%s\n",buf);
+              hostname, mypid, pid, pid);
+      printf("%s\n", buf);
       fflush(stdout);
       sleep(1);
     }
   }
 
   if (mypid == 0) {
-    printf( "** Enter a character to continue > "); fflush(stdout);
+    printf("** Enter a character to continue > ");
+    fflush(stdout);
     char go = ' ';
-    scanf("%c",&go);
+    scanf("%c", &go);
   }
   comm->barrier();
 #endif
@@ -161,35 +161,35 @@ int main(int argc, char *argv[]) {
   /* CREATE INITIAL MATRIX                                                          */
   /**********************************************************************************/
   const RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
   RCP<Matrix> Op = Pr->BuildMatrix();
   /**********************************************************************************/
   /*                                                                                */
   /**********************************************************************************/
 
   // dump matrix to file
-  //std::string fileName = "Amat.mm";
-  //Utils::Write(fileName,Op);
+  // std::string fileName = "Amat.mm";
+  // Utils::Write(fileName,Op);
 
-  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-  nullSpace->putScalar( (SC) 1.0);
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
   Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
   nullSpace->norm1(norms);
   if (comm->getRank() == 0)
     std::cout << "||NS|| = " << norms[0] << std::endl;
 
-  RCP<MueLu::Hierarchy<SC,LO,GO,NO> > H = rcp( new Hierarchy() );
+  RCP<MueLu::Hierarchy<SC, LO, GO, NO> > H = rcp(new Hierarchy());
   H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-  RCP<MueLu::Level> Finest = rcp( new MueLu::Level() );
+  RCP<MueLu::Level> Finest = rcp(new MueLu::Level());
   Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
-  Finest->Set("A",Op);
-  Finest->Set("Nullspace",nullSpace);
-  Finest->Request("Nullspace"); //FIXME putting this in to avoid error until Merge needs business
-  //FIXME is implemented
+  Finest->Set("A", Op);
+  Finest->Set("Nullspace", nullSpace);
+  Finest->Request("Nullspace");  // FIXME putting this in to avoid error until Merge needs business
+  // FIXME is implemented
 
-  Finest->Set("NullSpace",nullSpace);
+  Finest->Set("NullSpace", nullSpace);
   H->SetLevel(Finest);
 
   RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
@@ -199,15 +199,15 @@ int main(int argc, char *argv[]) {
 
   RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory(UncoupledAggFact));
 
-  RCP<SaPFactory>       Pfact = rcp( new SaPFactory(TentPFact) );
-  RCP<Factory>         Rfact = rcp( new TransPFactory() );
-  RCP<GenericPRFactory> PRfact = rcp( new GenericPRFactory(Pfact,Rfact));
-  RCP<RAPFactory>       Acfact = rcp( new RAPFactory() );
+  RCP<SaPFactory> Pfact        = rcp(new SaPFactory(TentPFact));
+  RCP<Factory> Rfact           = rcp(new TransPFactory());
+  RCP<GenericPRFactory> PRfact = rcp(new GenericPRFactory(Pfact, Rfact));
+  RCP<RAPFactory> Acfact       = rcp(new RAPFactory());
 
   RCP<SmootherPrototype> smooProto;
   Teuchos::ParameterList ifpackList;
-  ifpackList.set("relaxation: sweeps", (LO) 1);
-  ifpackList.set("relaxation: damping factor", (SC) 1.0);
+  ifpackList.set("relaxation: sweeps", (LO)1);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
   /*
      ifpackList.set("type", "Chebyshev");
      ifpackList.set("chebyshev: degree", (int) 1);
@@ -218,82 +218,81 @@ int main(int argc, char *argv[]) {
   if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
     ifpackList.set("relaxation: type", "symmetric Gauss-Seidel");
-    smooProto = rcp( new IfpackSmoother("point relaxation stand-alone",ifpackList) );
+    smooProto = rcp(new IfpackSmoother("point relaxation stand-alone", ifpackList));
 #endif
   } else if (xpetraParameters.GetLib() == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_IFPACK2)
     ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smooProto = rcp( new Ifpack2Smoother("RELAXATION",ifpackList) );
+    smooProto = rcp(new Ifpack2Smoother("RELAXATION", ifpackList));
 #endif
   }
   if (smooProto == Teuchos::null) {
     throw(MueLu::Exceptions::RuntimeError("main: smoother error"));
   }
 
-  RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
+  RCP<SmootherFactory> SmooFact = rcp(new SmootherFactory(smooProto));
   Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
   Teuchos::ParameterList status;
-  status = H->FullPopulate(PRfact,Acfact,SmooFact,0,maxLevels);
-  //RCP<MueLu::Level> coarseLevel = H.GetLevel(1);
-  //RCP<Matrix> P = coarseLevel->template Get< RCP<Matrix> >("P");
-  //fileName = "Pfinal.mm";
-  //Utils::Write(fileName,P);
+  status = H->FullPopulate(PRfact, Acfact, SmooFact, 0, maxLevels);
+  // RCP<MueLu::Level> coarseLevel = H.GetLevel(1);
+  // RCP<Matrix> P = coarseLevel->template Get< RCP<Matrix> >("P");
+  // fileName = "Pfinal.mm";
+  // Utils::Write(fileName,P);
   if (comm->getRank() == 0) {
-    std::cout  << "======================\n Multigrid statistics \n======================" << std::endl;
-    status.print(std::cout,Teuchos::ParameterList::PrintOptions().indent(2));
+    std::cout << "======================\n Multigrid statistics \n======================" << std::endl;
+    status.print(std::cout, Teuchos::ParameterList::PrintOptions().indent(2));
   }
 
-  //FIXME we should be able to just call smoother->SetNIts(50) ... but right now an exception gets thrown
+  // FIXME we should be able to just call smoother->SetNIts(50) ... but right now an exception gets thrown
 
   RCP<SmootherPrototype> coarseProto;
   if (xpetraParameters.GetLib() == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
     if (comm->getRank() == 0) std::cout << "CoarseGrid: AMESOS" << std::endl;
     Teuchos::ParameterList amesosList;
-    amesosList.set("PrintTiming",true);
-    coarseProto = rcp( new AmesosSmoother("Amesos_Klu",amesosList) );
+    amesosList.set("PrintTiming", true);
+    coarseProto = rcp(new AmesosSmoother("Amesos_Klu", amesosList));
     //#elif
 #endif
   } else if (xpetraParameters.GetLib() == Xpetra::UseTpetra) {
-    if (coarseSolver=="amesos2") {
+    if (coarseSolver == "amesos2") {
 #if defined(HAVE_MUELU_AMESOS2)
       if (comm->getRank() == 0) std::cout << "CoarseGrid: AMESOS2" << std::endl;
-      Teuchos::ParameterList paramList; //unused
-      coarseProto = rcp( new Amesos2Smoother("Superlu", paramList) );
+      Teuchos::ParameterList paramList;  // unused
+      coarseProto = rcp(new Amesos2Smoother("Superlu", paramList));
 #else
-      std::cout  << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
+      std::cout << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
       return EXIT_FAILURE;
-#endif // HAVE_MUELU_AMESOS2
-    } else if(coarseSolver=="ifpack2") {
+#endif  // HAVE_MUELU_AMESOS2
+    } else if (coarseSolver == "ifpack2") {
 #if defined(HAVE_MUELU_IFPACK2)
       if (comm->getRank() == 0) std::cout << "CoarseGrid: IFPACK2" << std::endl;
       Teuchos::ParameterList ifpack2List;
-      ifpack2List.set("fact: ilut level-of-fill",99); // TODO ??
+      ifpack2List.set("fact: ilut level-of-fill", 99);  // TODO ??
       ifpack2List.set("fact: drop tolerance", 0);
       ifpack2List.set("fact: absolute threshold", 0);
       ifpack2List.set("fact: relative threshold", 0);
-      coarseProto = rcp( new Ifpack2Smoother("ILUT",ifpack2List) );
+      coarseProto = rcp(new Ifpack2Smoother("ILUT", ifpack2List));
 #else
-      std::cout  << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
+      std::cout << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
       return EXIT_FAILURE;
 #endif
     } else {
-      std::cout  << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
+      std::cout << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
       return EXIT_FAILURE;
     }
-
   }
   if (coarseProto == Teuchos::null) {
     throw(MueLu::Exceptions::RuntimeError("main: coarse smoother error"));
   }
 
   SmootherFactory coarseSolveFact(coarseProto);
-  H->SetCoarsestSolver(coarseSolveFact,MueLu::PRE);
+  H->SetCoarsestSolver(coarseSolveFact, MueLu::PRE);
 
   // Define RHS
-  RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-  RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
   X->setSeed(846930886);
   X->randomize();
@@ -301,14 +300,14 @@ int main(int argc, char *argv[]) {
   if (comm->getRank() == 0)
     std::cout << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-  Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
   // Use AMG directly as an iterative method
   {
-    X->putScalar( (SC) 0.0);
+    X->putScalar((SC)0.0);
 
     H->PrintResidualHistory(true);
-    H->Iterate(*RHS,*X,its);
+    H->Iterate(*RHS, *X, its);
 
     X->norm2(norms);
     if (comm->getRank() == 0)
@@ -316,5 +315,4 @@ int main(int argc, char *argv[]) {
   }
 
   return EXIT_SUCCESS;
-
 }
diff --git a/packages/muelu/test/simple1D/amesos2StandAlone.cpp b/packages/muelu/test/simple1D/amesos2StandAlone.cpp
index 5f488e88f0a7..711e40751613 100644
--- a/packages/muelu/test/simple1D/amesos2StandAlone.cpp
+++ b/packages/muelu/test/simple1D/amesos2StandAlone.cpp
@@ -112,9 +112,8 @@
 #include "Amesos2.hpp"
 #include "Amesos2_Version.hpp"
 
-
 int main(int argc, char *argv[]) {
-  Tpetra::ScopeGuard mpiSession(&argc,&argv);
+  Tpetra::ScopeGuard mpiSession(&argc, &argv);
 
   typedef Tpetra::MultiVector<> MV;
   typedef MV::scalar_type Scalar;
@@ -122,51 +121,50 @@ int main(int argc, char *argv[]) {
   typedef MV::global_ordinal_type GO;
 
   typedef Teuchos::ScalarTraits<Scalar> ST;
-  typedef Tpetra::CrsMatrix<Scalar,LO,GO> MAT;
+  typedef Tpetra::CrsMatrix<Scalar, LO, GO> MAT;
 
-  using Tpetra::global_size_t;
-  using Teuchos::tuple;
   using std::endl;
+  using Teuchos::tuple;
+  using Tpetra::global_size_t;
 
-  std::ostream &out = std::cout;
+  std::ostream &out              = std::cout;
   RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(rcpFromRef(out));
 
   RCP<const Teuchos::Comm<int> > comm = Tpetra::getDefaultComm();
-  size_t myRank = comm->getRank();
+  size_t myRank                       = comm->getRank();
 
-  out << "Amesos2 stand-alone test" << endl << endl;
+  out << "Amesos2 stand-alone test" << endl
+      << endl;
 
   const size_t numVectors = 1;
 
-  int numGlobalElements = 1000;
-  RCP<const Tpetra::Map<> > map = Tpetra::createUniformContigMap<LO, GO> (numGlobalElements, comm);
-  const size_t numMyElements = map->getLocalNumElements();
+  int numGlobalElements                         = 1000;
+  RCP<const Tpetra::Map<> > map                 = Tpetra::createUniformContigMap<LO, GO>(numGlobalElements, comm);
+  const size_t numMyElements                    = map->getLocalNumElements();
   Teuchos::ArrayView<const GO> myGlobalElements = map->getLocalElementList();
 
-  RCP<MAT> A = Tpetra::createCrsMatrix<Scalar>(map,3);
+  RCP<MAT> A = Tpetra::createCrsMatrix<Scalar>(map, 3);
 
   // 1D Laplace
-  for (size_t i=0; i<numMyElements; i++) {
+  for (size_t i = 0; i < numMyElements; i++) {
     if (myGlobalElements[i] == 0) {
-      A->insertGlobalValues( myGlobalElements[i],
-          tuple<GO>( myGlobalElements[i], myGlobalElements[i]+1 ),
-          tuple<Scalar> ( 2.0, -1.0 ) );
-    }
-    else if (myGlobalElements[i] == numGlobalElements-1) {
-      A->insertGlobalValues( myGlobalElements[i],
-          tuple<GO>( myGlobalElements[i]-1, myGlobalElements[i] ),
-          tuple<Scalar> ( -1.0, 2.0 ) );
-    }
-    else {
-      A->insertGlobalValues( myGlobalElements[i],
-          tuple<GO>( myGlobalElements[i]-1, myGlobalElements[i], myGlobalElements[i]+1 ),
-          tuple<Scalar> ( -1.0, 2.0, -1.0 ) );
+      A->insertGlobalValues(myGlobalElements[i],
+                            tuple<GO>(myGlobalElements[i], myGlobalElements[i] + 1),
+                            tuple<Scalar>(2.0, -1.0));
+    } else if (myGlobalElements[i] == numGlobalElements - 1) {
+      A->insertGlobalValues(myGlobalElements[i],
+                            tuple<GO>(myGlobalElements[i] - 1, myGlobalElements[i]),
+                            tuple<Scalar>(-1.0, 2.0));
+    } else {
+      A->insertGlobalValues(myGlobalElements[i],
+                            tuple<GO>(myGlobalElements[i] - 1, myGlobalElements[i], myGlobalElements[i] + 1),
+                            tuple<Scalar>(-1.0, 2.0, -1.0));
     }
   }
   A->fillComplete();
 
   /* Create X */
-  RCP<MV> X = rcp(new MV(map,numVectors));
+  RCP<MV> X = rcp(new MV(map, numVectors));
   Teuchos::ScalarTraits<Scalar>::seedrandom(846930886);
   X->randomize();
 
@@ -179,14 +177,14 @@ int main(int argc, char *argv[]) {
   }
 
   /* Create B  */
-  RCP<MV> B = rcp(new MV(map,numVectors));
-  A->apply(*X,*B,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
+  RCP<MV> B = rcp(new MV(map, numVectors));
+  A->apply(*X, *B, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
 
   /* Reset X */
-  X->putScalar( (Scalar) 0.0);
+  X->putScalar((Scalar)0.0);
 
   // Create solver interface to Superlu through Amesos::Factory
-  RCP<Amesos2::Solver<MAT,MV> > solver = Amesos2::create<MAT,MV>("Superlu", A, X, B);
+  RCP<Amesos2::Solver<MAT, MV> > solver = Amesos2::create<MAT, MV>("Superlu", A, X, B);
 
   // Solve
   solver->symbolicFactorization().numericFactorization().solve();
diff --git a/packages/muelu/test/simple1D/mainMergedSmoo.cpp b/packages/muelu/test/simple1D/mainMergedSmoo.cpp
index 369b5af8840a..a2870f07f448 100644
--- a/packages/muelu/test/simple1D/mainMergedSmoo.cpp
+++ b/packages/muelu/test/simple1D/mainMergedSmoo.cpp
@@ -83,103 +83,100 @@
 #include "BelosConfigDefs.hpp"
 #include "BelosLinearProblem.hpp"
 #include "BelosBlockCGSolMgr.hpp"
-#include "BelosMueLuAdapter.hpp" // this header defines Belos::MueLuOp()
+#include "BelosMueLuAdapter.hpp"  // this header defines Belos::MueLuOp()
 
 namespace MueLuTests {
 #include "MueLu_UseShortNames.hpp"
 
-  RCP<SmootherPrototype> gimmeGaussSeidelProto(Xpetra::UnderlyingLib lib) {
+RCP<SmootherPrototype> gimmeGaussSeidelProto(Xpetra::UnderlyingLib lib) {
+  RCP<SmootherPrototype> smooProto;
+  Teuchos::ParameterList ifpackList;
+  ifpackList.set("relaxation: sweeps", (LO)1);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
 
-    RCP<SmootherPrototype> smooProto;
-    Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LO) 1);
-    ifpackList.set("relaxation: damping factor", (SC) 1.0);
-
-    if (lib == Xpetra::UseEpetra) {
+  if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
-      ifpackList.set("relaxation: type", "symmetric Gauss-Seidel");
-      smooProto = rcp( new IfpackSmoother("point relaxation stand-alone",ifpackList) );
+    ifpackList.set("relaxation: type", "symmetric Gauss-Seidel");
+    smooProto = rcp(new IfpackSmoother("point relaxation stand-alone", ifpackList));
 #endif
-    } else if (lib == Xpetra::UseTpetra) {
+  } else if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_IFPACK2)
-      ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-      smooProto = rcp( new Ifpack2Smoother("RELAXATION", ifpackList) );
+    ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
+    smooProto = rcp(new Ifpack2Smoother("RELAXATION", ifpackList));
 #endif
-    }
-    if (smooProto == Teuchos::null) {
-      throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelSmoother: smoother error"));
-    }
-
-    return smooProto;
+  }
+  if (smooProto == Teuchos::null) {
+    throw(MueLu::Exceptions::RuntimeError("gimmeGaussSeidelSmoother: smoother error"));
   }
 
-  RCP<SmootherPrototype> gimmeCoarseProto(Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
-    RCP<SmootherPrototype> coarseProto;
-    if (lib == Xpetra::UseEpetra) {
+  return smooProto;
+}
+
+RCP<SmootherPrototype> gimmeCoarseProto(Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
+  RCP<SmootherPrototype> coarseProto;
+  if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
-      if (rank == 0) std::cout << "CoarseGrid: AMESOS" << std::endl;
-      Teuchos::ParameterList amesosList;
-      amesosList.set("PrintTiming",true);
-      coarseProto = rcp( new AmesosSmoother("Amesos_Klu",amesosList) );
-      //#elif...
+    if (rank == 0) std::cout << "CoarseGrid: AMESOS" << std::endl;
+    Teuchos::ParameterList amesosList;
+    amesosList.set("PrintTiming", true);
+    coarseProto = rcp(new AmesosSmoother("Amesos_Klu", amesosList));
+    //#elif...
 #endif
-    } else if (lib == Xpetra::UseTpetra) {
-      if (coarseSolver=="amesos2") {
+  } else if (lib == Xpetra::UseTpetra) {
+    if (coarseSolver == "amesos2") {
 #if defined(HAVE_MUELU_AMESOS2)
-        if (rank == 0) std::cout << "CoarseGrid: AMESOS2" << std::endl;
-        Teuchos::ParameterList paramList; //unused
-        coarseProto = rcp( new Amesos2Smoother("Superlu", paramList) );
+      if (rank == 0) std::cout << "CoarseGrid: AMESOS2" << std::endl;
+      Teuchos::ParameterList paramList;  // unused
+      coarseProto = rcp(new Amesos2Smoother("Superlu", paramList));
 #else
-        std::cout  << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
-        return Teuchos::null; // TODO test for exception //EXIT_FAILURE;
-#endif // HAVE_MUELU_AMESOS2
-      } else if(coarseSolver=="ifpack2") {
+      std::cout << "AMESOS2 not available (try --coarseSolver=ifpack2)" << std::endl;
+      return Teuchos::null;  // TODO test for exception //EXIT_FAILURE;
+#endif  // HAVE_MUELU_AMESOS2
+    } else if (coarseSolver == "ifpack2") {
 #if defined(HAVE_MUELU_IFPACK2)
-        if (rank == 0) std::cout << "CoarseGrid: IFPACK2" << std::endl;
-        Teuchos::ParameterList ifpack2List;
-        ifpack2List.set("fact: ilut level-of-fill",99); // TODO ??
-        ifpack2List.set("fact: drop tolerance", 0);
-        ifpack2List.set("fact: absolute threshold", 0);
-        ifpack2List.set("fact: relative threshold", 0);
-        coarseProto = rcp( new Ifpack2Smoother("ILUT",ifpack2List) );
+      if (rank == 0) std::cout << "CoarseGrid: IFPACK2" << std::endl;
+      Teuchos::ParameterList ifpack2List;
+      ifpack2List.set("fact: ilut level-of-fill", 99);  // TODO ??
+      ifpack2List.set("fact: drop tolerance", 0);
+      ifpack2List.set("fact: absolute threshold", 0);
+      ifpack2List.set("fact: relative threshold", 0);
+      coarseProto = rcp(new Ifpack2Smoother("ILUT", ifpack2List));
 #else
-        std::cout  << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
-        //TODO        TEUCHOS_TEST_FOR_EXCEPTION
-        return Teuchos::null;
+      std::cout << "IFPACK2 not available (try --coarseSolver=amesos2)" << std::endl;
+      // TODO        TEUCHOS_TEST_FOR_EXCEPTION
+      return Teuchos::null;
 #endif
-      } else {
-        std::cout  << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
-        return Teuchos::null;
-      }
-
+    } else {
+      std::cout << "Unknow coarse grid solver (try  --coarseSolver=ifpack2 or --coarseSolver=amesos2)" << std::endl;
+      return Teuchos::null;
     }
-    if (coarseProto == Teuchos::null) {
-      throw(MueLu::Exceptions::RuntimeError("main: coarse smoother error"));
-    }
-
-    return coarseProto;
+  }
+  if (coarseProto == Teuchos::null) {
+    throw(MueLu::Exceptions::RuntimeError("main: coarse smoother error"));
   }
 
+  return coarseProto;
+}
 
-  RCP<SmootherPrototype> gimmeMergedSmoother(int nSmoothers, Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
-    ArrayRCP<RCP<SmootherPrototype> > smootherList(nSmoothers);
-
-    for (int i=0; i<nSmoothers; i++)
-      smootherList[i] = gimmeGaussSeidelProto(lib);
+RCP<SmootherPrototype> gimmeMergedSmoother(int nSmoothers, Xpetra::UnderlyingLib lib, const std::string& coarseSolver, int rank) {
+  ArrayRCP<RCP<SmootherPrototype> > smootherList(nSmoothers);
 
-    return rcp (new MergedSmoother(smootherList));
-    //verbose mode: return rcp (new MergedSmoother(smootherList, true));
-  }
+  for (int i = 0; i < nSmoothers; i++)
+    smootherList[i] = gimmeGaussSeidelProto(lib);
 
+  return rcp(new MergedSmoother(smootherList));
+  // verbose mode: return rcp (new MergedSmoother(smootherList, true));
 }
 
-int main(int argc, char *argv[]) {
+}  // namespace MueLuTests
+
+int main(int argc, char* argv[]) {
 #include "MueLu_UseShortNames.hpp"
 
   using namespace MueLuTests;
 
   Teuchos::oblackholestream blackhole;
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv,&blackhole);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv, &blackhole);
   RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
   /**********************************************************************************/
@@ -190,25 +187,25 @@ int main(int argc, char *argv[]) {
 
   // Default is Laplace1D with nx = 8748.
   // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-  //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-  Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+  // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
   // custom parameters
-  int nSmoothers=2;
-  LO maxLevels = 3;
-  LO its=10;
-  std::string coarseSolver="ifpack2";
-  clp.setOption("nSmoothers",&nSmoothers,"number of Gauss-Seidel smoothers in the MergedSmootehrs");
-  clp.setOption("maxLevels",&maxLevels,"maximum number of levels allowed. If 1, then a MergedSmoother is used on the coarse grid");
-  clp.setOption("its",&its,"number of multigrid cycles");
-  clp.setOption("coarseSolver",&coarseSolver,"amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
-
-  switch (clp.parse(argc,argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+  int nSmoothers           = 2;
+  LO maxLevels             = 3;
+  LO its                   = 10;
+  std::string coarseSolver = "ifpack2";
+  clp.setOption("nSmoothers", &nSmoothers, "number of Gauss-Seidel smoothers in the MergedSmootehrs");
+  clp.setOption("maxLevels", &maxLevels, "maximum number of levels allowed. If 1, then a MergedSmoother is used on the coarse grid");
+  clp.setOption("its", &its, "number of multigrid cycles");
+  clp.setOption("coarseSolver", &coarseSolver, "amesos2 or ifpack2 (Tpetra specific. Ignored for Epetra)");
+
+  switch (clp.parse(argc, argv)) {
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   matrixParameters.check();
@@ -225,50 +222,50 @@ int main(int argc, char *argv[]) {
   /* CREATE INITIAL MATRIX                                                          */
   /**********************************************************************************/
   const RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
   RCP<Matrix> Op = Pr->BuildMatrix();
   /**********************************************************************************/
   /*                                                                                */
   /**********************************************************************************/
 
-  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-  nullSpace->putScalar( (SC) 1.0);
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
   Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
   nullSpace->norm1(norms);
   if (comm->getRank() == 0)
     std::cout << "||NS|| = " << norms[0] << std::endl;
 
-  RCP<MueLu::Hierarchy<SC,LO,GO,NO> > H = rcp( new Hierarchy() );
+  RCP<MueLu::Hierarchy<SC, LO, GO, NO> > H = rcp(new Hierarchy());
   H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-  RCP<MueLu::Level> Finest = rcp( new MueLu::Level() );
+  RCP<MueLu::Level> Finest = rcp(new MueLu::Level());
   Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
-  Finest->Set("A",Op);
-  Finest->Set("Nullspace",nullSpace);
-  Finest->Request("Nullspace"); //FIXME putting this in to avoid error until Merge needs business
-  //FIXME is implemented
+  Finest->Set("A", Op);
+  Finest->Set("Nullspace", nullSpace);
+  Finest->Request("Nullspace");  // FIXME putting this in to avoid error until Merge needs business
+  // FIXME is implemented
 
-  Finest->Set("NullSpace",nullSpace);
+  Finest->Set("NullSpace", nullSpace);
   H->SetLevel(Finest);
 
   RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-  RCP<TentativePFactory>    TentPFact = rcp(new TentativePFactory(UncoupledAggFact));
+  RCP<TentativePFactory> TentPFact                  = rcp(new TentativePFactory(UncoupledAggFact));
 
-  RCP<SaPFactory>       Pfact = rcp( new SaPFactory(TentPFact) );
-  RCP<Factory>         Rfact = rcp( new TransPFactory() );
-  RCP<GenericPRFactory> PRfact = rcp( new GenericPRFactory(Pfact,Rfact));
-  RCP<RAPFactory>       Acfact = rcp( new RAPFactory() );
+  RCP<SaPFactory> Pfact        = rcp(new SaPFactory(TentPFact));
+  RCP<Factory> Rfact           = rcp(new TransPFactory());
+  RCP<GenericPRFactory> PRfact = rcp(new GenericPRFactory(Pfact, Rfact));
+  RCP<RAPFactory> Acfact       = rcp(new RAPFactory());
 
   RCP<SmootherPrototype> smooProto = gimmeMergedSmoother(nSmoothers, xpetraParameters.GetLib(), coarseSolver, comm->getRank());
-  RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
   Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
   Teuchos::ParameterList status;
-  status = H->FullPopulate(PRfact,Acfact,SmooFact,0,maxLevels);
+  status = H->FullPopulate(PRfact, Acfact, SmooFact, 0, maxLevels);
   if (comm->getRank() == 0) {
-    std::cout  << "======================\n Multigrid statistics \n======================" << std::endl;
-    status.print(std::cout,Teuchos::ParameterList::PrintOptions().indent(2));
+    std::cout << "======================\n Multigrid statistics \n======================" << std::endl;
+    status.print(std::cout, Teuchos::ParameterList::PrintOptions().indent(2));
   }
 
   RCP<SmootherPrototype> coarseProto;
@@ -279,11 +276,11 @@ int main(int argc, char *argv[]) {
   }
   if (coarseProto == Teuchos::null) return EXIT_FAILURE;
   SmootherFactory coarseSolveFact(coarseProto);
-  H->SetCoarsestSolver(coarseSolveFact,MueLu::PRE);
+  H->SetCoarsestSolver(coarseSolveFact, MueLu::PRE);
 
   // Define RHS
-  RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-  RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
   X->setSeed(846930886);
   X->randomize();
@@ -291,14 +288,14 @@ int main(int argc, char *argv[]) {
   if (comm->getRank() == 0)
     std::cout << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-  Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
   // Use AMG directly as an iterative method
   {
-    X->putScalar( (SC) 0.0);
+    X->putScalar((SC)0.0);
 
     H->PrintResidualHistory(true);
-    H->Iterate(*RHS,*X,its);
+    H->Iterate(*RHS, *X, its);
 
     X->norm2(norms);
     if (comm->getRank() == 0)
@@ -306,5 +303,4 @@ int main(int argc, char *argv[]) {
   }
 
   return EXIT_SUCCESS;
-
 }
diff --git a/packages/muelu/test/simple1D/simple1D-UncoupledAggregation.cpp b/packages/muelu/test/simple1D/simple1D-UncoupledAggregation.cpp
index 3b1a29e6db25..3ae65aa42e1c 100644
--- a/packages/muelu/test/simple1D/simple1D-UncoupledAggregation.cpp
+++ b/packages/muelu/test/simple1D/simple1D-UncoupledAggregation.cpp
@@ -78,7 +78,7 @@
 #include <unistd.h>
 /**********************************************************************************/
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -95,29 +95,29 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Default is Laplace1D with nx = 8748.
     // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-    //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-    Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+    // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
     // custom parameters
-    //std::string aggOrdering = "natural";
-    int minPerAgg=2; //was 3 in simple
-    int maxNbrAlreadySelected=0;
-    int printTimings=0;
-
-    //clp.setOption("aggOrdering",&aggOrdering,"aggregation ordering strategy (natural,graph)");
-    clp.setOption("maxNbrSel",&maxNbrAlreadySelected,"maximum # of nbrs allowed to be in other aggregates");
-    clp.setOption("minPerAgg",&minPerAgg,"minimum #DOFs per aggregate");
-    clp.setOption("timings",&printTimings,"print timings to screen");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    // std::string aggOrdering = "natural";
+    int minPerAgg             = 2;  // was 3 in simple
+    int maxNbrAlreadySelected = 0;
+    int printTimings          = 0;
+
+    // clp.setOption("aggOrdering",&aggOrdering,"aggregation ordering strategy (natural,graph)");
+    clp.setOption("maxNbrSel", &maxNbrAlreadySelected, "maximum # of nbrs allowed to be in other aggregates");
+    clp.setOption("minPerAgg", &minPerAgg, "minimum #DOFs per aggregate");
+    clp.setOption("timings", &printTimings, "print timings to screen");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp (new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
+    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
 
     matrixParameters.check();
     xpetraParameters.check();
@@ -136,16 +136,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       Teuchos::TimeMonitor tm(*Teuchos::TimeMonitor::getNewTimer("Timings: Matrix Build"));
 
       map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
-      Teuchos::RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+      Teuchos::RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
       A = Pr->BuildMatrix();
-
     }
     /**********************************************************************************/
     /*                                                                                */
     /**********************************************************************************/
 
-
     Teuchos::RCP<Hierarchy> hierarchy = Teuchos::rcp(new Hierarchy(A));
     hierarchy->SetDefaultVerbLevel(MueLu::toMueLuVerbLevel(Teuchos::VERB_EXTREME));
     hierarchy->SetMaxCoarseSize(100);
@@ -154,51 +152,49 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     Teuchos::RCP<MueLu::Level> Finest = hierarchy->GetLevel();  // get finest level
 
-    Finest->Set("A",A);
+    Finest->Set("A", A);
 
     Teuchos::RCP<AmalgamationFactory> amalgFact = Teuchos::rcp(new AmalgamationFactory());
-    Teuchos::RCP<CoalesceDropFactory> dropFact = Teuchos::rcp(new CoalesceDropFactory());
+    Teuchos::RCP<CoalesceDropFactory> dropFact  = Teuchos::rcp(new CoalesceDropFactory());
     dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
     // aggregation factory
     Teuchos::RCP<UncoupledAggregationFactory> UnCoupledAggFact = Teuchos::rcp(new UncoupledAggregationFactory(/*dropFact*/));
     UnCoupledAggFact->SetFactory("Graph", dropFact);
     UnCoupledAggFact->SetFactory("DofsPerNode", dropFact);
-    //UnCoupledAggFact->SetFactory("Graph", dropFact); // UnCoupledAggFact not changed to new factory handling
-    //UnCoupledAggFact->SetMinNodesPerAggregate(minPerAgg);
-    //UnCoupledAggFact->SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
-    //UnCoupledAggFact->SetOrdering(MueLu::AggOptions::GRAPH);
+    // UnCoupledAggFact->SetFactory("Graph", dropFact); // UnCoupledAggFact not changed to new factory handling
+    // UnCoupledAggFact->SetMinNodesPerAggregate(minPerAgg);
+    // UnCoupledAggFact->SetMaxNeighAlreadySelected(maxNbrAlreadySelected);
+    // UnCoupledAggFact->SetOrdering(MueLu::AggOptions::GRAPH);
     UnCoupledAggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(maxNbrAlreadySelected));
-    UnCoupledAggFact->SetParameter("aggregation: min agg size",           Teuchos::ParameterEntry(minPerAgg));
-    UnCoupledAggFact->SetParameter("aggregation: ordering",               Teuchos::ParameterEntry(std::string("graph")));
+    UnCoupledAggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(minPerAgg));
+    UnCoupledAggFact->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("graph")));
 
-    Finest->Request("Graph",dropFact.get());
-    Finest->Request("DofsPerNode",dropFact.get());
-    Finest->Request("UnAmalgamationInfo",amalgFact.get());
+    Finest->Request("Graph", dropFact.get());
+    Finest->Request("DofsPerNode", dropFact.get());
+    Finest->Request("UnAmalgamationInfo", amalgFact.get());
     UnCoupledAggFact->Build(*Finest);
 
     // Timer final summaries
-    globalTimeMonitor = Teuchos::null; // stop this timer before summary
+    globalTimeMonitor = Teuchos::null;  // stop this timer before summary
 
     if (printTimings) {
       Teuchos::TimeMonitor::summarize();
 #ifdef HAVE_TEUCHOS_ADD_TIME_MONITOR_TO_STACKED_TIMER
-      Teuchos::TimeMonitor::getStackedTimer()->report(std::cout,comm);
+      Teuchos::TimeMonitor::getStackedTimer()->report(std::cout, comm);
 #endif
     }
     success = true;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
-
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
-
diff --git a/packages/muelu/test/structured/Driver_Structured.cpp b/packages/muelu/test/structured/Driver_Structured.cpp
index 9899aa671e88..7ae799ef5964 100644
--- a/packages/muelu/test/structured/Driver_Structured.cpp
+++ b/packages/muelu/test/structured/Driver_Structured.cpp
@@ -84,12 +84,11 @@
 #include <BelosBlockGmresSolMgr.hpp>
 #include <BelosLinearProblem.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
-#include <BelosTpetraAdapter.hpp>    // => This header defines Belos::TpetraOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
+#include <BelosTpetraAdapter.hpp>  // => This header defines Belos::TpetraOp
 #endif
 
-
 #ifdef HAVE_MUELU_CUDA
 #include "cuda_profiler_api.h"
 #endif
@@ -111,15 +110,14 @@
 
 #include <MueLu_CreateXpetraPreconditioner.hpp>
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -132,37 +130,50 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   typedef Teuchos::ScalarTraits<SC> STS;
   SC zero = STS::zero(), one = STS::one();
   typedef typename STS::magnitudeType real_type;
-  typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
   // =========================================================================
   // Parameters initialization
   // =========================================================================
   GO nx = 100, ny = 100, nz = 100;
-  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-  Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-  std::string xmlFileName       = "";                clp.setOption("xml",                   &xmlFileName,       "read parameters from an xml file");
-  std::string yamlFileName      = "";                clp.setOption("yaml",                  &yamlFileName,      "read parameters from a yaml file");
-  int         rerun             = 1;                 clp.setOption("rerun",                 &rerun,             "number of reruns required");
-  int         maxIts            = 200;               clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-  double      tol               = 1e-12;             clp.setOption("tol",                   &tol,               "solver convergence tolerance");
-  bool        scaleResidualHist = true;              clp.setOption("scale", "noscale",      &scaleResidualHist, "scaled Krylov residual history");
-  bool        solvePreconditioned = true;            clp.setOption("solve-preconditioned","no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
-  std::string operation         = "solve";           clp.setOption("op",                    &operation,         "operation to perform: (matvec | setup | solver)");
-  std::string belosType         = "cg";              clp.setOption("belosType",             &belosType,         "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
-  std::string equilibrate = "no" ;                   clp.setOption("equilibrate",           &equilibrate,       "equilibrate the system (no | diag | 1-norm)");
+  Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+  std::string xmlFileName = "";
+  clp.setOption("xml", &xmlFileName, "read parameters from an xml file");
+  std::string yamlFileName = "";
+  clp.setOption("yaml", &yamlFileName, "read parameters from a yaml file");
+  int rerun = 1;
+  clp.setOption("rerun", &rerun, "number of reruns required");
+  int maxIts = 200;
+  clp.setOption("its", &maxIts, "maximum number of solver iterations");
+  double tol = 1e-12;
+  clp.setOption("tol", &tol, "solver convergence tolerance");
+  bool scaleResidualHist = true;
+  clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
+  bool solvePreconditioned = true;
+  clp.setOption("solve-preconditioned", "no-solve-preconditioned", &solvePreconditioned, "use MueLu preconditioner in solve");
+  std::string operation = "solve";
+  clp.setOption("op", &operation, "operation to perform: (matvec | setup | solver)");
+  std::string belosType = "cg";
+  clp.setOption("belosType", &belosType, "belos solver type: (Pseudoblock CG | Block CG | Pseudoblock GMRES | Block GMRES | ...) see BelosSolverFactory.hpp for exhaustive list of solvers");
+  std::string equilibrate = "no";
+  clp.setOption("equilibrate", &equilibrate, "equilibrate the system (no | diag | 1-norm)");
 #ifdef HAVE_MUELU_CUDA
-  bool profileSetup = false;                         clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
-  bool profileSolve = false;                         clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
+  bool profileSetup = false;
+  clp.setOption("cuda-profile-setup", "no-cuda-profile-setup", &profileSetup, "enable CUDA profiling for setup");
+  bool profileSolve = false;
+  clp.setOption("cuda-profile-solve", "no-cuda-profile-solve", &profileSolve, "enable CUDA profiling for solve");
 #endif
-  int  cacheSize = 0;                                clp.setOption("cachesize",               &cacheSize,       "cache size (in KB)");
+  int cacheSize = 0;
+  clp.setOption("cachesize", &cacheSize, "cache size (in KB)");
 
   clp.recogniseAllOptions(true);
   switch (clp.parse(argc, argv)) {
-    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
     case Teuchos::CommandLineProcessor::PARSE_ERROR:
     case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+    case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
   }
 
   TEUCHOS_TEST_FOR_EXCEPTION(xmlFileName != "" && yamlFileName != "", std::runtime_error,
@@ -170,7 +181,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Instead of checking each time for rank, create a rank 0 stream
   RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-  Teuchos::FancyOStream& out = *fancy;
+  Teuchos::FancyOStream& out       = *fancy;
   out.setOutputToRootOnly(0);
 
   ParameterList paramList;
@@ -186,7 +197,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<ParameterList>(&paramList), *comm);
   }
 
-
   // Retrieve matrix parameters (they may have been changed on the command line)
   // [for instance, if we changed matrix type from 2D to 3D we need to update nz]
   ParameterList galeriList = galeriParameters.GetParameterList();
@@ -197,25 +207,24 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   std::ostringstream galeriStream;
 #ifdef HAVE_MUELU_OPENMP
   std::string node_name = Node::name();
-  if(!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
-    galeriStream<<"OpenMP Max Threads = "<<omp_get_max_threads()<<std::endl;
+  if (!comm->getRank() && !node_name.compare("OpenMP/Wrapper"))
+    galeriStream << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
 #endif
 
-
   comm->barrier();
   Teuchos::TimeMonitor::setStackedTimer(Teuchos::null);
   RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: S - Global Time")));
   RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 1 - Matrix Build")));
 
-
-  RCP<Matrix>      A;
-  RCP<const Map>   map;
+  RCP<Matrix> A;
+  RCP<const Map> map;
   RCP<RealValuedMultiVector> coordinates;
   typedef typename RealValuedMultiVector::scalar_type Real;
-  RCP<Xpetra::MultiVector<SC,LO,GO,NO> > nullspace;
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > nullspace;
   RCP<MultiVector> X, B;
 
-  galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+  galeriStream << "========================================================\n"
+               << xpetraParameters << galeriParameters;
 
   // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
   //                                 d1  d2  d3
@@ -227,7 +236,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // size. For example, np=14 will give a 7-by-2 distribution.
   // If you don't want Galeri to do this, specify mx or my on the galeriList.
   std::string matrixType = galeriParameters.GetMatrixType();
-  int numDimensions = 0;
+  int numDimensions      = 0;
   Teuchos::Array<LO> lNodesPerDim(3);
 
   // Create map and coordinates
@@ -235,30 +244,30 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
   if (matrixType == "Laplace1D") {
     numDimensions = 1;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
   } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
              matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
     numDimensions = 2;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", map, galeriList);
 
   } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
     numDimensions = 3;
-    map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-    coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("3D", map, galeriList);
+    map           = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+    coordinates   = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("3D", map, galeriList);
   }
 
-  if(numDimensions == 1) {
+  if (numDimensions == 1) {
     lNodesPerDim[0] = galeriList.get<LO>("lnx");
     lNodesPerDim[1] = 1;
     lNodesPerDim[2] = 1;
-  } else if(numDimensions == 2) {
+  } else if (numDimensions == 2) {
     lNodesPerDim[0] = galeriList.get<LO>("lnx");
     lNodesPerDim[1] = galeriList.get<LO>("lny");
     lNodesPerDim[2] = 1;
-  } else if(numDimensions == 3) {
+  } else if (numDimensions == 3) {
     lNodesPerDim[0] = galeriList.get<LO>("lnx");
     lNodesPerDim[1] = galeriList.get<LO>("lny");
     lNodesPerDim[2] = galeriList.get<LO>("lnz");
@@ -266,9 +275,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   // Expand map to do multiple DOF per node for block problems
   if (matrixType == "Elasticity2D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
   if (matrixType == "Elasticity3D")
-    map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+    map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
   galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -277,15 +286,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
     // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-    galeriList.set("right boundary" , "Neumann");
+    galeriList.set("right boundary", "Neumann");
     galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
+    galeriList.set("top boundary", "Neumann");
+    galeriList.set("front boundary", "Neumann");
+    galeriList.set("back boundary", "Neumann");
   }
 
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
   A = Pr->BuildMatrix();
 
   if (matrixType == "Elasticity2D" ||
@@ -304,7 +313,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
   Teuchos::Array<typename STS::magnitudeType> norms(1);
   B->norm2(norms);
-  B->scale(one/norms[0]);
+  B->scale(one / norms[0]);
   galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
   out << galeriStream.str();
@@ -313,41 +322,40 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   tm = Teuchos::null;
 
 #ifdef HAVE_MUELU_CUDA
-  if(profileSetup) cudaProfilerStart();
+  if (profileSetup) cudaProfilerStart();
 #endif
 
   // Extract the diagonal of A (for RAPShiftFactory testing)
-  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC,LO,GO,NO>::Build(A->getRowMap(),false);
+  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC, LO, GO, NO>::Build(A->getRowMap(), false);
   A->getLocalDiagCopy(*Mdiag);
 
-  for(int run = 0; run < rerun; ++run) {
-
+  for (int run = 0; run < rerun; ++run) {
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 2 - MueLu Setup")));
     RCP<Hierarchy> H;
     RCP<Operator> Prec;
     A->SetMaxEigenvalueEstimate(-Teuchos::ScalarTraits<SC>::one());
 
-    const std::string userName = "user data";
+    const std::string userName            = "user data";
     Teuchos::ParameterList& userParamList = paramList.sublist(userName);
     userParamList.set<int>("int numDimensions", numDimensions);
     userParamList.set<Teuchos::Array<LO> >("Array<LO> lNodesPerDim", lNodesPerDim);
     userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    userParamList.set<double>("double cfl",1.0);
-    userParamList.set<double>("double deltaT",1.0);
-    userParamList.set("Mdiag",Mdiag);
-
+    userParamList.set<double>("double cfl", 1.0);
+    userParamList.set<double>("double deltaT", 1.0);
+    userParamList.set("Mdiag", Mdiag);
 
     H = MueLu::CreateXpetraPreconditioner(A, paramList);
 
-
     comm->barrier();
     tm = Teuchos::null;
 
 #ifdef HAVE_MUELU_CUDA
-    if(profileSolve) cudaProfilerStop();
+    if (profileSolve) cudaProfilerStop();
 #endif
 
-    if(operation == "setup") {continue;}
+    if (operation == "setup") {
+      continue;
+    }
 
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 3 - LHS and RHS initialization")));
     X->putScalar(zero);
@@ -356,21 +364,21 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #ifdef HAVE_MUELU_BELOS
     tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 5 - Belos Solve")));
 #ifdef HAVE_MUELU_CUDA
-    if(profileSolve) cudaProfilerStart();
+    if (profileSolve) cudaProfilerStart();
 #endif
     // Operator and Multivector type that will be used with Belos
-    typedef MultiVector          MV;
+    typedef MultiVector MV;
     typedef Belos::OperatorT<MV> OP;
 
     // Define Operator and Preconditioner
-    Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-    Teuchos::RCP<OP> belosPrec; // Turns a MueLu::Hierarchy object into a Belos operator
+    Teuchos::RCP<OP> belosOp = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+    Teuchos::RCP<OP> belosPrec;                                                       // Turns a MueLu::Hierarchy object into a Belos operator
     H->IsPreconditioner(true);
-    belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+    belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
 
     // Construct a Belos LinearProblem object
     RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
-    if(solvePreconditioned) belosProblem->setRightPrec(belosPrec);
+    if (solvePreconditioned) belosProblem->setRightPrec(belosPrec);
 
     bool set = belosProblem->setProblem();
     if (set == false) {
@@ -380,11 +388,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Belos parameter list
     RCP<Teuchos::ParameterList> belosList = Teuchos::parameterList();
-    belosList->set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-    belosList->set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-    belosList->set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-    belosList->set("Output Frequency",      1);
-    belosList->set("Output Style",          Belos::Brief);
+    belosList->set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+    belosList->set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+    belosList->set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+    belosList->set("Output Frequency", 1);
+    belosList->set("Output Style", Belos::Brief);
     if (!scaleResidualHist)
       belosList->set("Implicit Residual Scaling", "None");
 
@@ -395,28 +403,30 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Perform solve
     Belos::ReturnType retStatus = Belos::Unconverged;
-    retStatus = solver->solve();
+    retStatus                   = solver->solve();
 
     // Get the number of iterations for this solve.
     out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
     // Check convergence
     if (retStatus != Belos::Converged)
-      out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+      out << std::endl
+          << "ERROR:  Belos did not converge! " << std::endl;
     else
-      out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      out << std::endl
+          << "SUCCESS:  Belos converged!" << std::endl;
 #ifdef HAVE_MUELU_CUDA
-    if(profileSolve) cudaProfilerStop();
+    if (profileSolve) cudaProfilerStop();
 #endif
-#endif //ifdef HAVE_MUELU_BELOS
+#endif  // ifdef HAVE_MUELU_BELOS
 
     comm->barrier();
     tm = Teuchos::null;
-  } // Rerun loop
+  }  // Rerun loop
 
   globalTimeMonitor = Teuchos::null;
 
   RCP<ParameterList> reportParams = rcp(new ParameterList);
-  const std::string filter = "";
+  const std::string filter        = "";
   std::ios_base::fmtflags ff(out.flags());
   TimeMonitor::report(comm.ptr(), out, filter, reportParams);
   out << std::setiosflags(ff);
@@ -430,6 +440,6 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/toggletransfer/main.cpp b/packages/muelu/test/toggletransfer/main.cpp
index 3c15c428f423..7bec3940afc5 100644
--- a/packages/muelu/test/toggletransfer/main.cpp
+++ b/packages/muelu/test/toggletransfer/main.cpp
@@ -57,7 +57,7 @@
 #include <MueLu_Exceptions.hpp>
 #include <MueLu_ParameterListInterpreter.hpp>
 
-#include "MueLu_SemiCoarsenPFactory.hpp" // for semi-coarsening constants
+#include "MueLu_SemiCoarsenPFactory.hpp"  // for semi-coarsening constants
 #include <MueLu_TestHelpers.hpp>
 
 /**********************************************************************************/
@@ -79,7 +79,7 @@
 #include <unistd.h>
 /**********************************************************************************/
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
@@ -96,31 +96,31 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Default is Laplace1D with nx = 8748.
     // It's a nice size for 1D and perfect aggregation. (6561=3^8)
-    //Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-    Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+    // Nice size for 1D and perfect aggregation on small numbers of processors. (8748=4*3^7)
+    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
     // custom parameters
-    //std::string aggOrdering = "natural";
-    int minPerAgg=2; //was 3 in simple
-    int maxNbrAlreadySelected=0;
-    int printTimings=0;
-    std::string xmlFile="parameters.xml";
-
-    //clp.setOption("aggOrdering",&aggOrdering,"aggregation ordering strategy (natural,graph)");
-    clp.setOption("maxNbrSel",&maxNbrAlreadySelected,"maximum # of nbrs allowed to be in other aggregates");
-    clp.setOption("minPerAgg",&minPerAgg,"minimum #DOFs per aggregate");
-    clp.setOption("timings",&printTimings,"print timings to screen");
-    clp.setOption("xmlFile",&xmlFile,"file name containing MueLu multigrid parameters in XML format");
-
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    // std::string aggOrdering = "natural";
+    int minPerAgg             = 2;  // was 3 in simple
+    int maxNbrAlreadySelected = 0;
+    int printTimings          = 0;
+    std::string xmlFile       = "parameters.xml";
+
+    // clp.setOption("aggOrdering",&aggOrdering,"aggregation ordering strategy (natural,graph)");
+    clp.setOption("maxNbrSel", &maxNbrAlreadySelected, "maximum # of nbrs allowed to be in other aggregates");
+    clp.setOption("minPerAgg", &minPerAgg, "minimum #DOFs per aggregate");
+    clp.setOption("timings", &printTimings, "print timings to screen");
+    clp.setOption("xmlFile", &xmlFile, "file name containing MueLu multigrid parameters in XML format");
+
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
-    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp (new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
+    Teuchos::RCP<Teuchos::TimeMonitor> globalTimeMonitor = Teuchos::rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("Timings: Global Time")));
 
     matrixParameters.check();
     xpetraParameters.check();
@@ -139,10 +139,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
       Teuchos::TimeMonitor tm(*Teuchos::TimeMonitor::getNewTimer("Timings: Matrix Build"));
 
       map = MapFactory::Build(lib, matrixParameters.GetNumGlobalElements(), 0, comm);
-      Teuchos::RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList()); //TODO: Matrix vs. CrsMatrixWrap
+      Teuchos::RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixParameters.GetMatrixType(), map, matrixParameters.GetParameterList());  // TODO: Matrix vs. CrsMatrixWrap
       A = Pr->BuildMatrix();
-
     }
     /**********************************************************************************/
     /*                                                                                */
@@ -152,17 +151,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     if (paramList.isSublist("Factories")) {
       Teuchos::ParameterList smootherParams = paramList.sublist("Factories").sublist("myJacobi").sublist("ParameterList");
-      double damping = smootherParams.get<double>("relaxation: damping factor");
+      double damping                        = smootherParams.get<double>("relaxation: damping factor");
       smootherParams.remove("relaxation: damping factor");
-      smootherParams.set<Scalar>("relaxation: damping factor",damping);
-      paramList.sublist("Factories").sublist("myJacobi").set("ParameterList",smootherParams);
+      smootherParams.set<Scalar>("relaxation: damping factor", damping);
+      paramList.sublist("Factories").sublist("myJacobi").set("ParameterList", smootherParams);
     }
     if (paramList.isSublist("smoother: params")) {
       Teuchos::ParameterList smootherParams = paramList.sublist("smoother: params");
-      double damping = smootherParams.get<double>("relaxation: damping factor");
+      double damping                        = smootherParams.get<double>("relaxation: damping factor");
       smootherParams.remove("relaxation: damping factor");
-      smootherParams.set<Scalar>("relaxation: damping factor",damping);
-      paramList.set("smoother: params",smootherParams);
+      smootherParams.set<Scalar>("relaxation: damping factor", damping);
+      paramList.set("smoother: params", smootherParams);
     }
 
     // create parameter list interpreter
@@ -170,7 +169,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     Teuchos::RCP<Hierarchy> H = mueluFactory->CreateHierarchy();
 
-    H->GetLevel(0)->template Set< Teuchos::RCP<Matrix> >("A", A);
+    H->GetLevel(0)->template Set<Teuchos::RCP<Matrix> >("A", A);
 
     Teuchos::RCP<MultiVector> nullspace = MultiVectorFactory::Build(A->getRowMap(), 1);
     nullspace->putScalar(1.0);
@@ -181,19 +180,36 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // parameter: "semicoarsen: num layers"
     // TAW: 3/16/2016: NumZLayers has to be provided as LO
     GO zLayers = matrixParameters.GetParameterList().template get<GO>("nz");
-    H->GetLevel(0)->Set("NumZLayers",Teuchos::as<LO>(zLayers));
-
+    H->GetLevel(0)->Set("NumZLayers", Teuchos::as<LO>(zLayers));
 
     mueluFactory->SetupHierarchy(*H);
 
-    for (int l=0; l<H->GetNumLevels(); l++) {
+    for (int l = 0; l < H->GetNumLevels(); l++) {
       Teuchos::RCP<MueLu::Level> level = H->GetLevel(l);
-      if(level->IsAvailable("A", MueLu::NoFactory::get()) == false) { success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
-      if(level->IsAvailable("P", MueLu::NoFactory::get()) == false && l>0) { success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
-      if(level->IsAvailable("R", MueLu::NoFactory::get()) == false && l>0) { success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
-      if(level->IsAvailable("PreSmoother",  MueLu::NoFactory::get()) == false) { success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
-      if(level->IsAvailable("PostSmoother", MueLu::NoFactory::get()) == false && l<H->GetNumLevels()-1) { success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
-      if(level->IsAvailable("NumZLayers",   MueLu::NoFactory::get()) == true && l>0) {  success = false; H->GetLevel(l)->print(std::cout, MueLu::Debug);}
+      if (level->IsAvailable("A", MueLu::NoFactory::get()) == false) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
+      if (level->IsAvailable("P", MueLu::NoFactory::get()) == false && l > 0) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
+      if (level->IsAvailable("R", MueLu::NoFactory::get()) == false && l > 0) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
+      if (level->IsAvailable("PreSmoother", MueLu::NoFactory::get()) == false) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
+      if (level->IsAvailable("PostSmoother", MueLu::NoFactory::get()) == false && l < H->GetNumLevels() - 1) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
+      if (level->IsAvailable("NumZLayers", MueLu::NoFactory::get()) == true && l > 0) {
+        success = false;
+        H->GetLevel(l)->print(std::cout, MueLu::Debug);
+      }
       H->GetLevel(l)->print(std::cout, MueLu::Debug);
     }
     ///////////////////////////////////////////////////////////
@@ -216,7 +232,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       Teuchos::Array<typename STS::magnitudeType> norms(1);
       B->norm2(norms);
-      B->scale(one/norms[0]);
+      B->scale(one / norms[0]);
       X->putScalar(zero);
     }
 
@@ -226,14 +242,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     H->Iterate(*B, *X, 20);
 
     // Timer final summaries
-    globalTimeMonitor = Teuchos::null; // stop this timer before summary
+    globalTimeMonitor = Teuchos::null;  // stop this timer before summary
 
     if (printTimings)
       Teuchos::TimeMonitor::summarize();
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
@@ -241,5 +257,5 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 #include "MueLu_Test_ETI.hpp"
 
 int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/unit_tests/Adapters/AmgxOperatorAdapter.cpp b/packages/muelu/test/unit_tests/Adapters/AmgxOperatorAdapter.cpp
index 919ce2bfd808..7688678133c4 100644
--- a/packages/muelu/test/unit_tests/Adapters/AmgxOperatorAdapter.cpp
+++ b/packages/muelu/test/unit_tests/Adapters/AmgxOperatorAdapter.cpp
@@ -70,67 +70,62 @@
 
 namespace MueLuTests {
 
- TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AMGXOperator,Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    // NOTE: This test only works for double/int/int
-    if(!TYPE_EQUAL(double,Scalar) || !TYPE_EQUAL(int,LocalOrdinal) || !TYPE_EQUAL(int,GlobalOrdinal)) {
-      out << "This test is enabled only for double/int/int"<<std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AMGXOperator, Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  // NOTE: This test only works for double/int/int
+  if (!TYPE_EQUAL(double, Scalar) || !TYPE_EQUAL(int, LocalOrdinal) || !TYPE_EQUAL(int, GlobalOrdinal)) {
+    out << "This test is enabled only for double/int/int" << std::endl;
+    return;
+  }
+
+  typedef MueLu::AMGXOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> AMGXOperator;
+  out << "version: " << MueLu::Version() << std::endl;
+
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    int nx;
+    // disable amgx test in parallel
+    if (comm->getSize() > 1)
       return;
-    }
+    else
+      nx = 91;
 
-    typedef MueLu::AMGXOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> AMGXOperator;
-    out << "version: " << MueLu::Version() << std::endl;
+    // matrix
+    RCP<Matrix> Op                                                         = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx, -1, Xpetra::UseTpetra);
+    RCP<Tpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tpA = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(Op);
+    RCP<Tpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tOp  = tpA;
+    Teuchos::ParameterList params, dummyList;
+    params.set("use external multigrid package", "amgx");
+    Teuchos::ParameterList subList = params.sublist("amgx:params", false);
+    params.sublist("amgx:params").set("json file", "test.json");
+    RCP<MueLu::TpetraOperator<Scalar, LocalOrdinal, GlobalOrdinal, Node> > tH = MueLu::CreateTpetraPreconditioner(tOp, params);
 
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra)
-    {
+    RCP<AMGXOperator> aH = Teuchos::rcp_dynamic_cast<AMGXOperator>(tH);
+    TEST_EQUALITY(aH->sizeA() == nx * nx / comm->getSize(), true);
 
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      int nx;
-      //disable amgx test in parallel
-      if(comm->getSize() > 1)
-	return;
-      else
-	nx = 91;
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(Op->getRowMap(), 1);
 
-      //matrix
-      RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx, -1, Xpetra::UseTpetra);
-      RCP<Tpetra::CrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tpA = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Op2NonConstTpetraCrs(Op);
-      RCP<Tpetra::Operator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tOp = tpA;
-      Teuchos::ParameterList params, dummyList;
-      params.set("use external multigrid package", "amgx");
-      Teuchos::ParameterList subList = params.sublist("amgx:params", false);
-      params.sublist("amgx:params").set("json file", "test.json");
-      RCP<MueLu::TpetraOperator<Scalar,LocalOrdinal,GlobalOrdinal,Node> > tH = MueLu::CreateTpetraPreconditioner(tOp, params);
+    // RHS=1, zero initial guess
+    RHS->putScalar((double)1.0);
+    X->putScalar((double)0.0);
 
-      RCP<AMGXOperator> aH = Teuchos::rcp_dynamic_cast<AMGXOperator>(tH);
-      TEST_EQUALITY(aH->sizeA()==nx*nx/comm->getSize(), true);
+    aH->apply(*(MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(RHS)), *(MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(X)));
+    // if(comm->getSize() == 1) TEST_EQUALITY(aH->iters()==16,true);
+    TEST_EQUALITY(aH->getStatus() == 0, true);
 
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      RCP<MultiVector> X   = MultiVectorFactory::Build(Op->getRowMap(), 1);
+  } else {
+    out << "This test is enabled only for linAlgebra=Tpetra." << std::endl;
+  }
 
-      //RHS=1, zero initial guess
-      RHS->putScalar( (double) 1.0);
-      X->putScalar( (double) 0.0);
+}  // Apply
 
-      aH->apply(*(MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2TpetraMV(RHS)),*(MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::MV2NonConstTpetraMV(X)));
-      //if(comm->getSize() == 1) TEST_EQUALITY(aH->iters()==16,true);
-      TEST_EQUALITY(aH->getStatus()==0, true);
-
-    } else {
-
-      out << "This test is enabled only for linAlgebra=Tpetra." << std::endl;
-    }
-
-  } //Apply
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AMGXOperator,Apply,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AMGXOperator, Apply, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Adapters/BelosAdapters.cpp b/packages/muelu/test/unit_tests/Adapters/BelosAdapters.cpp
index 68717ffc3705..817752fe2c60 100644
--- a/packages/muelu/test/unit_tests/Adapters/BelosAdapters.cpp
+++ b/packages/muelu/test/unit_tests/Adapters/BelosAdapters.cpp
@@ -56,276 +56,274 @@
 
 namespace MueLuTests {
 
-  //
-  // Helpers function to build tests
-  //
-
-  // Test Belos adapters for the couple <MV,OP>
-  // TODO: add a bunch of 'const' on prototype
-  template <class Scalar, class MV, class OP>
-    int BelosAdaptersTest(RCP<OP> & belosOp, RCP<OP> & belosPrec, RCP<MV> & X, RCP<MV> & B, Teuchos::FancyOStream & out, bool & success) {
-      RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
-      belosProblem->setLeftPrec(belosPrec);
-
-      bool set = belosProblem->setProblem();
-      TEST_EQUALITY(set, true);
-
-      // Belos parameter list
-      Teuchos::ParameterList belosList;
-      belosList.set("Maximum Iterations",    10);   // Maximum number of iterations allowed
-      belosList.set("Convergence Tolerance", 1e-7); // Relative convergence tolerance requested
-
-      // Create an iterative solver manager. (was double before...)
-      RCP<Belos::SolverManager<Scalar, MV, OP> > belosSolver = rcp(new Belos::BlockCGSolMgr<Scalar,MV,OP>(belosProblem, rcp(&belosList,false)));
-
-      // Perform solve
-      Belos::ReturnType ret = belosSolver->solve();
-      TEST_EQUALITY(ret, Belos::Converged);
-
-      // Return number of iterations
-      return belosSolver->getNumIters();
-    }
-
-  //
-  // Helpers function to verify results
-  //
-
-  // Singleton for norm comparisons across tests
-  template <class Scalar>
-    bool BelosAdaptersTestResultsNorm(typename Teuchos::ScalarTraits<Scalar>::magnitudeType r) {
-      static typename Teuchos::ScalarTraits<Scalar>::magnitudeType ref = -1;
-      if (ref == -1) {
-        //std::cout << "BelosAdaptersTestResults(): Set reference results" << std::endl;
-        ref = r;
-        return true;
-      }
-      //std::cout << "BelosAdaptersTestResults(): Compare" << std::endl;
-
-      if (r != ref)
-        std::cout << "ref  norm = " << ref << std::endl
-          << "curr norm = " << r   << std::endl;
-
-      return (r == ref);
-    }
-
-  // Test results
-  template <class Scalar, class MV>
-    bool BelosAdaptersTestResults(int numIters, RCP<MV> & X, Teuchos::FancyOStream & out, bool & success) {
-
-      // Check numIters
-      switch (TestHelpers::Parameters::getDefaultComm()->getSize()) {
-        case 0: TEST_EQUALITY(numIters, 5); break;
-        case 4:
-                // Epetra TEST_EQUALITY(numIters, 6);
-                // Tpetra TEST_EQUALITY(numIters, 7);
-                break;
-        default:;
-      }
-
-      // Compute norm of X (using MV traits)
-      typedef Belos::MultiVecTraits<Scalar, MV> MVT;
-      std::vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms(1);
-      MVT::MvNorm(*X, norms);
-
-      // Test norm equality across the unit tests
-      //return MueLuTests::BelosAdaptersTestResultsNorm<Scalar>(norms[0]); // not working
-      return true;
-    }
-
-  //
-  // Tests
-  //
+//
+// Helpers function to build tests
+//
+
+// Test Belos adapters for the couple <MV,OP>
+// TODO: add a bunch of 'const' on prototype
+template <class Scalar, class MV, class OP>
+int BelosAdaptersTest(RCP<OP>& belosOp, RCP<OP>& belosPrec, RCP<MV>& X, RCP<MV>& B, Teuchos::FancyOStream& out, bool& success) {
+  RCP<Belos::LinearProblem<Scalar, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<Scalar, MV, OP>(belosOp, X, B));
+  belosProblem->setLeftPrec(belosPrec);
+
+  bool set = belosProblem->setProblem();
+  TEST_EQUALITY(set, true);
+
+  // Belos parameter list
+  Teuchos::ParameterList belosList;
+  belosList.set("Maximum Iterations", 10);       // Maximum number of iterations allowed
+  belosList.set("Convergence Tolerance", 1e-7);  // Relative convergence tolerance requested
+
+  // Create an iterative solver manager. (was double before...)
+  RCP<Belos::SolverManager<Scalar, MV, OP> > belosSolver = rcp(new Belos::BlockCGSolMgr<Scalar, MV, OP>(belosProblem, rcp(&belosList, false)));
+
+  // Perform solve
+  Belos::ReturnType ret = belosSolver->solve();
+  TEST_EQUALITY(ret, Belos::Converged);
+
+  // Return number of iterations
+  return belosSolver->getNumIters();
+}
+
+//
+// Helpers function to verify results
+//
+
+// Singleton for norm comparisons across tests
+template <class Scalar>
+bool BelosAdaptersTestResultsNorm(typename Teuchos::ScalarTraits<Scalar>::magnitudeType r) {
+  static typename Teuchos::ScalarTraits<Scalar>::magnitudeType ref = -1;
+  if (ref == -1) {
+    // std::cout << "BelosAdaptersTestResults(): Set reference results" << std::endl;
+    ref = r;
+    return true;
+  }
+  // std::cout << "BelosAdaptersTestResults(): Compare" << std::endl;
+
+  if (r != ref)
+    std::cout << "ref  norm = " << ref << std::endl
+              << "curr norm = " << r << std::endl;
+
+  return (r == ref);
+}
+
+// Test results
+template <class Scalar, class MV>
+bool BelosAdaptersTestResults(int numIters, RCP<MV>& X, Teuchos::FancyOStream& out, bool& success) {
+  // Check numIters
+  switch (TestHelpers::Parameters::getDefaultComm()->getSize()) {
+    case 0: TEST_EQUALITY(numIters, 5); break;
+    case 4:
+      // Epetra TEST_EQUALITY(numIters, 6);
+      // Tpetra TEST_EQUALITY(numIters, 7);
+      break;
+    default:;
+  }
+
+  // Compute norm of X (using MV traits)
+  typedef Belos::MultiVecTraits<Scalar, MV> MVT;
+  std::vector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norms(1);
+  MVT::MvNorm(*X, norms);
+
+  // Test norm equality across the unit tests
+  // return MueLuTests::BelosAdaptersTestResultsNorm<Scalar>(norms[0]); // not working
+  return true;
+}
 
-  // TEST:
-  // - OP: Xpetra::Matrix
-  // - MV: Xpetra::MultiVector
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_XpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+//
+// Tests
+//
+
+// TEST:
+// - OP: Xpetra::Matrix
+// - MV: Xpetra::MultiVector
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_XpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+  Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
 
 #if !defined(HAVE_MUELU_EPETRA) or !defined(HAVE_MUELU_IFPACK) or !defined(HAVE_MUELU_AMESOS)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
 #endif
 
 #if !defined(HAVE_MUELU_IFPACK2) or !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
 #endif
 
-    RCP<TestProblem<SC,LO,GO,NO> > p = rcp (new TestProblem<SC,LO,GO,NO> (lib));
-
-    typedef MultiVector             MV;
-    typedef Belos::OperatorT<MV>    OP;
+  RCP<TestProblem<SC, LO, GO, NO> > p = rcp(new TestProblem<SC, LO, GO, NO>(lib));
 
-    // Construct a Belos LinearProblem object
-    RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(p->GetA()));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(p->GetH()));
+  typedef MultiVector MV;
+  typedef Belos::OperatorT<MV> OP;
 
-    // Run Belos
-    RCP<MultiVector> X = p->GetNewX0();
-    int numIters = MueLuTests::BelosAdaptersTest<Scalar, MV, OP>(belosOp, belosPrec, X, p->GetRHS(), out, success);
+  // Construct a Belos LinearProblem object
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(p->GetA()));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(p->GetH()));
 
-    // Tests
-    TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
-  }
+  // Run Belos
+  RCP<MultiVector> X = p->GetNewX0();
+  int numIters       = MueLuTests::BelosAdaptersTest<Scalar, MV, OP>(belosOp, belosPrec, X, p->GetRHS(), out, success);
 
-  // TEST:
-  // - OP: Xpetra::Matrix
-  // - MV: Epetra::MultiVector
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_EpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  // Tests
+  TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
+}
+
+// TEST:
+// - OP: Xpetra::Matrix
+// - MV: Epetra::MultiVector
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_EpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_AMESOS)
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-    RCP<TestProblem<SC,LO,GO,NO> > p = rcp (new TestProblem<SC,LO,GO,NO> (lib));
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<TestProblem<SC, LO, GO, NO> > p = rcp(new TestProblem<SC, LO, GO, NO>(lib));
 
-    typedef Epetra_MultiVector   MV;
-    typedef Belos::OperatorT<MV> OP;
+  typedef Epetra_MultiVector MV;
+  typedef Belos::OperatorT<MV> OP;
 
-    // Construct a Belos LinearProblem object
-    RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(p->GetA()));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp <SC, LO, GO, NO>(p->GetH()));
+  // Construct a Belos LinearProblem object
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<SC, LO, GO, NO>(p->GetA()));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(p->GetH()));
 
-    // X, B
-    RCP<MV> X = MueLu::Utilities<SC,LO,GO,NO>::MV2NonConstEpetraMV(p->GetNewX0());
-    RCP<MV> B = MueLu::Utilities<SC,LO,GO,NO>::MV2NonConstEpetraMV(p->GetRHS());
+  // X, B
+  RCP<MV> X = MueLu::Utilities<SC, LO, GO, NO>::MV2NonConstEpetraMV(p->GetNewX0());
+  RCP<MV> B = MueLu::Utilities<SC, LO, GO, NO>::MV2NonConstEpetraMV(p->GetRHS());
 
-    // Run Belos
-    int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
+  // Run Belos
+  int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
 
-    // Tests
-    TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
+  // Tests
+  TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
 #endif
-  }
+}
 
-  // TEST:
-  // - OP: Belos::Operator<double>
-  // - MV: Belos::MultiVec<double>
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, BelosMultiVec_BelosMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+// TEST:
+// - OP: Belos::Operator<double>
+// - MV: Belos::MultiVec<double>
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, BelosMultiVec_BelosMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_AMESOS)
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-    RCP<TestProblem<SC,LO,GO,NO> > p = rcp (new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> (lib));
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<TestProblem<SC, LO, GO, NO> > p = rcp(new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib));
 
-    typedef Belos::MultiVec<SC> MV;
-    typedef Belos::Operator<SC> OP;
+  typedef Belos::MultiVec<SC> MV;
+  typedef Belos::Operator<SC> OP;
 
-    // Construct a Belos LinearProblem object
-    RCP<Epetra_CrsMatrix> A = Utilities::Op2NonConstEpetraCrs(p->GetA());
-    RCP<OP> belosOp   = rcp(new Belos::EpetraOp(A));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(p->GetH()));
+  // Construct a Belos LinearProblem object
+  RCP<Epetra_CrsMatrix> A = Utilities::Op2NonConstEpetraCrs(p->GetA());
+  RCP<OP> belosOp         = rcp(new Belos::EpetraOp(A));
+  RCP<OP> belosPrec       = rcp(new Belos::MueLuOp<SC, LO, GO, NO>(p->GetH()));
 
-    // X, B
-    RCP<Epetra_MultiVector> eX = Utilities::MV2NonConstEpetraMV(p->GetNewX0());
-    RCP<Epetra_MultiVector> eB = Utilities::MV2NonConstEpetraMV(p->GetRHS());
-    RCP<MV> X = rcp(new Belos::EpetraMultiVec(*eX));
-    RCP<MV> B = rcp(new Belos::EpetraMultiVec(*eB));
+  // X, B
+  RCP<Epetra_MultiVector> eX = Utilities::MV2NonConstEpetraMV(p->GetNewX0());
+  RCP<Epetra_MultiVector> eB = Utilities::MV2NonConstEpetraMV(p->GetRHS());
+  RCP<MV> X                  = rcp(new Belos::EpetraMultiVec(*eX));
+  RCP<MV> B                  = rcp(new Belos::EpetraMultiVec(*eB));
 
-    // Run Belos
-    int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
+  // Run Belos
+  int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
 
-    // Tests
-    TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
+  // Tests
+  TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
 
-    // TODO: this do not work. Is it a bug?
-    //  double norm;
-    //  eX->Norm2(&norm);
+  // TODO: this do not work. Is it a bug?
+  //  double norm;
+  //  eX->Norm2(&norm);
 #endif
-  }
+}
 
-  // TEST:
-  // - OP: Xpetra::Matrix
-  // - MV: Tpetra::MultiVector
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_TpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+// TEST:
+// - OP: Xpetra::Matrix
+// - MV: Tpetra::MultiVector
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosAdapters, XpetraOp_TpetraMV, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
 #if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS2)
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-    RCP<TestProblem<Scalar,LocalOrdinal,GlobalOrdinal,Node> > p = rcp (new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> (lib));
+  Xpetra::UnderlyingLib lib                                      = TestHelpers::Parameters::getLib();
+  RCP<TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> > p = rcp(new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib));
 
-    typedef typename Tpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> MV;
-    typedef typename Belos::OperatorT<MV>    OP;
+  typedef typename Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> MV;
+  typedef typename Belos::OperatorT<MV> OP;
 
-    // Construct a Belos LinearProblem object
-    RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(p->GetA()));
-    RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(p->GetH()));
+  // Construct a Belos LinearProblem object
+  RCP<OP> belosOp   = rcp(new Belos::XpetraOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(p->GetA()));
+  RCP<OP> belosPrec = rcp(new Belos::MueLuOp<Scalar, LocalOrdinal, GlobalOrdinal, Node>(p->GetH()));
 
-    //X, B
-    RCP<MV> X = Utilities::MV2NonConstTpetraMV(p->GetNewX0());
-    RCP<MV> B = Utilities::MV2NonConstTpetraMV(p->GetRHS());
+  // X, B
+  RCP<MV> X = Utilities::MV2NonConstTpetraMV(p->GetNewX0());
+  RCP<MV> B = Utilities::MV2NonConstTpetraMV(p->GetRHS());
 
-    // Run Belos
-    int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
+  // Run Belos
+  int numIters = MueLuTests::BelosAdaptersTest<SC, MV, OP>(belosOp, belosPrec, X, B, out, success);
 
-    // Tests
-    TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
+  // Tests
+  TEST_EQUALITY(MueLuTests::BelosAdaptersTestResults<Scalar>(numIters, X, out, success), true);
 #endif
-  }
+}
 
 // Instantiate the Tpetra and Xpetra based tests
-  // run Xpetra based tests
-# define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_XpetraMV, Scalar, LO, GO, Node)
+// run Xpetra based tests
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_XpetraMV, Scalar, LO, GO, Node)
 
-# include <MueLu_ETI_4arg.hpp>
+#include <MueLu_ETI_4arg.hpp>
 
-  // run Tpetra based tests
-  // These tests use the original belos/tpetra MultiVecTraits which are not guarded and have no specializations
-  // for Epetra. Therefore, carefully choose valid Tpetra instantiations.
+// run Tpetra based tests
+// These tests use the original belos/tpetra MultiVecTraits which are not guarded and have no specializations
+// for Epetra. Therefore, carefully choose valid Tpetra instantiations.
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT) && defined(HAVE_TPETRA_INST_SERIAL)
-  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode SerialNode;
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, int, SerialNode)
+typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode SerialNode;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, int, SerialNode)
 #endif
 
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG) && defined(HAVE_TPETRA_INST_SERIAL)
-  typedef long long int LongLong;
-  typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode SerialNode;
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, LongLong, SerialNode)
+typedef long long int LongLong;
+typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode SerialNode;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, LongLong, SerialNode)
 #endif
 
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT) && defined(HAVE_TPETRA_INST_OPENMP)
-  typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode OpenMPNode;
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, int, OpenMPNode)
+typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode OpenMPNode;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, int, OpenMPNode)
 #endif
 
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG) && defined(HAVE_TPETRA_INST_OPENMP)
-  typedef long long int LongLong;
-  typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode OpenMPNode;
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, LongLong, OpenMPNode)
+typedef long long int LongLong;
+typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode OpenMPNode;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_TpetraMV, double, int, LongLong, OpenMPNode)
 #endif
 
-
 #if defined(HAVE_MUELU_EPETRA)
 #include "Epetra_config.h"
-#include "Xpetra_Map.hpp" // defines EpetraNode
+#include "Xpetra_Map.hpp"  // defines EpetraNode
 typedef Xpetra::EpetraNode EpetraNode;
 #ifndef EPETRA_NO_32BIT_GLOBAL_INDICES
 TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_EpetraMV, double, int, int, EpetraNode)
 #endif
 #ifndef EPETRA_NO_64BIT_GLOBAL_INDICES
-//typedef long long int LongLong;
-//TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_EpetraMV, double, int, LongLong, EpetraNode)
+// typedef long long int LongLong;
+// TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosAdapters, XpetraOp_EpetraMV, double, int, LongLong, EpetraNode)
 #endif
 #endif
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
-//TODO: norm test can be factorized, using Belos Adapter Norm function.
+// TODO: norm test can be factorized, using Belos Adapter Norm function.
 
-//TODO: generate an hierarchy that do not need ifpack[1,2]/amesos[1,2] and remove safeguards
+// TODO: generate an hierarchy that do not need ifpack[1,2]/amesos[1,2] and remove safeguards
diff --git a/packages/muelu/test/unit_tests/Adapters/CreatePreconditioner.cpp b/packages/muelu/test/unit_tests/Adapters/CreatePreconditioner.cpp
index bcd82c98bc96..6af84159a3ae 100644
--- a/packages/muelu/test/unit_tests/Adapters/CreatePreconditioner.cpp
+++ b/packages/muelu/test/unit_tests/Adapters/CreatePreconditioner.cpp
@@ -76,447 +76,313 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    using Teuchos::RCP;
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  using Teuchos::RCP;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    std::string xmlFileName = "test.xml";
+  std::string xmlFileName = "test.xml";
 
-    if (lib == Xpetra::UseTpetra) {
+  if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
-      typedef Tpetra::CrsMatrix<SC,LO,GO,NO> tpetra_crsmatrix_type;
-      typedef Tpetra::Operator<SC,LO,GO,NO> tpetra_operator_type;
-      typedef Tpetra::MultiVector<SC,LO,GO,NO> tpetra_multivector_type;
-      typedef Xpetra::MultiVector<real_type,LO,GO,NO> dMultiVector;
-      typedef Tpetra::MultiVector<real_type,LO,GO,NO> dtpetra_multivector_type;
-
-      // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
-
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
-
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    typedef Tpetra::CrsMatrix<SC, LO, GO, NO> tpetra_crsmatrix_type;
+    typedef Tpetra::Operator<SC, LO, GO, NO> tpetra_operator_type;
+    typedef Tpetra::MultiVector<SC, LO, GO, NO> tpetra_multivector_type;
+    typedef Xpetra::MultiVector<real_type, LO, GO, NO> dMultiVector;
+    typedef Tpetra::MultiVector<real_type, LO, GO, NO> dtpetra_multivector_type;
+
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
+
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
+
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-      RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), 1);
-      nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+    RCP<MultiVector> nullspace             = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), 1);
+    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
 
-      RCP<tpetra_crsmatrix_type> tpA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op);
+    RCP<tpetra_crsmatrix_type> tpA = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op);
 
-      out << "========== Create Preconditioner from xmlFile ==========" << std::endl;
-      out << "xmlFileName: " << xmlFileName << std::endl;
-      RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA), xmlFileName);
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    out << "========== Create Preconditioner from xmlFile ==========" << std::endl;
+    out << "xmlFileName: " << xmlFileName << std::endl;
+    RCP<MueLu::TpetraOperator<SC, LO, GO, NO> > tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), xmlFileName);
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
 #endif
 
 #else
-      std::cout << "Skip PetraOperator::CreatePreconditioner: Tpetra is not available (with GO=int enabled)" << std::endl;
-#endif // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
+    std::cout << "Skip PetraOperator::CreatePreconditioner: Tpetra is not available (with GO=int enabled)" << std::endl;
+#endif  // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
-    } else if (lib == Xpetra::UseEpetra) {
+  } else if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
 
-      // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
 
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
 
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-      RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), 1);
-      nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+    RCP<MultiVector> nullspace             = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), 1);
+    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
 
-      RCP<Epetra_CrsMatrix> epA = Utils::Op2NonConstEpetraCrs(Op);
+    RCP<Epetra_CrsMatrix> epA = Utils::Op2NonConstEpetraCrs(Op);
 
-      RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, xmlFileName);
+    RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, xmlFileName);
 
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
-      xmlFileName = "testWithRebalance.xml";
+    xmlFileName = "testWithRebalance.xml";
 
-      RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type,LO,GO,NO>::MV2NonConstEpetraMV(coordinates);
-      RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
+    RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type, LO, GO, NO>::MV2NonConstEpetraMV(coordinates);
+    RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
 
-      Teuchos::ParameterList paramList;
-      Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *map->getComm());
-      Teuchos::ParameterList& userParamList = paramList.sublist("user data");
-      userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
-      userParamList.set<RCP<Epetra_MultiVector> >("Nullspace", epnullspace);
+    Teuchos::ParameterList paramList;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *map->getComm());
+    Teuchos::ParameterList& userParamList = paramList.sublist("user data");
+    userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
+    userParamList.set<RCP<Epetra_MultiVector> >("Nullspace", epnullspace);
 
-      eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
+    eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
 
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
 #endif
 
 #else
-      std::cout << "Skip PetraOperator::CreatePreconditioner: Epetra is not available" << std::endl;
+    std::cout << "Skip PetraOperator::CreatePreconditioner: Epetra is not available" << std::endl;
 #endif
 
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
-    }
-
-  } //CreatePreconditioner
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
+  }
 
+}  // CreatePreconditioner
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner_XMLOnList, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner_XMLOnList, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    using Teuchos::RCP;
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> dMultiVector;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  using Teuchos::RCP;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> dMultiVector;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    Teuchos::ParameterList mylist;
-    mylist.set("xml parameter file","test.xml");
+  Teuchos::ParameterList mylist;
+  mylist.set("xml parameter file", "test.xml");
 
-    if (lib == Xpetra::UseTpetra) {
+  if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
-      typedef Tpetra::Operator<SC,LO,GO,NO> tpetra_operator_type;
+    typedef Tpetra::Operator<SC, LO, GO, NO> tpetra_operator_type;
 
-      // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
 
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
 
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-      RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), 1);
-      nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
-
-      RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > tpA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op);
-
-      RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA),mylist);
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-
-      mylist.set("xml parameter file","testWithRebalance.xml");
-
-      RCP<Tpetra::MultiVector<real_type,LO,GO,NO> > tpcoordinates = MueLu::Utilities<real_type,LO,GO,NO>::MV2NonConstTpetraMV(coordinates);
-      RCP<Tpetra::MultiVector<SC,LO,GO,NO> > tpnullspace   = Utils::MV2NonConstTpetraMV(nullspace);
-
-      std::string mueluXML = mylist.get("xml parameter file", "");
-      Teuchos::ParameterList mueluList;
-      Teuchos::updateParametersFromXmlFileAndBroadcast(mueluXML, Teuchos::Ptr<Teuchos::ParameterList>(&mueluList), *map->getComm());
-      Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
-      userParamList.set<RCP<Tpetra::MultiVector<real_type,LO,GO,NO> > >("Coordinates", tpcoordinates);
-      userParamList.set<RCP<Tpetra::MultiVector<SC,LO,GO,NO> > >("Nullspace", tpnullspace);
-      tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA), mueluList);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+    RCP<MultiVector> nullspace             = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), 1);
+    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+
+    RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > tpA = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op);
+
+    RCP<MueLu::TpetraOperator<SC, LO, GO, NO> > tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), mylist);
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+
+    mylist.set("xml parameter file", "testWithRebalance.xml");
+
+    RCP<Tpetra::MultiVector<real_type, LO, GO, NO> > tpcoordinates = MueLu::Utilities<real_type, LO, GO, NO>::MV2NonConstTpetraMV(coordinates);
+    RCP<Tpetra::MultiVector<SC, LO, GO, NO> > tpnullspace          = Utils::MV2NonConstTpetraMV(nullspace);
+
+    std::string mueluXML = mylist.get("xml parameter file", "");
+    Teuchos::ParameterList mueluList;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(mueluXML, Teuchos::Ptr<Teuchos::ParameterList>(&mueluList), *map->getComm());
+    Teuchos::ParameterList& userParamList = mueluList.sublist("user data");
+    userParamList.set<RCP<Tpetra::MultiVector<real_type, LO, GO, NO> > >("Coordinates", tpcoordinates);
+    userParamList.set<RCP<Tpetra::MultiVector<SC, LO, GO, NO> > >("Nullspace", tpnullspace);
+    tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), mueluList);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
 #endif
 
 #else
-      std::cout << "Skip PetraOperator::CreatePreconditioner_XMLOnList: Tpetra is not available (with GO=int enabled)" << std::endl;
-#endif // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
+    std::cout << "Skip PetraOperator::CreatePreconditioner_XMLOnList: Tpetra is not available (with GO=int enabled)" << std::endl;
+#endif  // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
-    } else if (lib == Xpetra::UseEpetra) {
+  } else if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
-
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
-
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
+
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
+
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,dMultiVector>("1D", Op->getRowMap(), galeriList);
-      RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), 1);
-      nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, dMultiVector>("1D", Op->getRowMap(), galeriList);
+    RCP<MultiVector> nullspace    = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), 1);
+    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
 
-      RCP<Epetra_CrsMatrix> epA = Utils::Op2NonConstEpetraCrs(Op);
+    RCP<Epetra_CrsMatrix> epA = Utils::Op2NonConstEpetraCrs(Op);
 
-      RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, mylist);
+    RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, mylist);
 
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
-      mylist.set("xml parameter file","testWithRebalance.xml");
+    mylist.set("xml parameter file", "testWithRebalance.xml");
 
-      RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type,LO,GO,NO>::MV2NonConstEpetraMV(coordinates);
-      RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
+    RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type, LO, GO, NO>::MV2NonConstEpetraMV(coordinates);
+    RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
 
-      Teuchos::ParameterList paramList = mylist;
-      Teuchos::ParameterList& userParamList = paramList.sublist("user data");
-      userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
-      userParamList.set<RCP<Epetra_MultiVector> >("Nullspace", epnullspace);
-      eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
+    Teuchos::ParameterList paramList      = mylist;
+    Teuchos::ParameterList& userParamList = paramList.sublist("user data");
+    userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
+    userParamList.set<RCP<Epetra_MultiVector> >("Nullspace", epnullspace);
+    eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
 
 #endif
 
 #else
-      std::cout << "Skip PetraOperator::CreatePreconditioner_XMLOnList: Epetra is not available" << std::endl;
+    std::cout << "Skip PetraOperator::CreatePreconditioner_XMLOnList: Epetra is not available" << std::endl;
 #endif
 
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
-    }
-
-  } //CreatePreconditioner_XMLOnList
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
+  }
 
+}  // CreatePreconditioner_XMLOnList
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner_PDESystem, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, CreatePreconditioner_PDESystem, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    using Teuchos::RCP;
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> dMultiVector;
+  using Teuchos::RCP;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> dMultiVector;
 
 #if defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
 
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    for (int k = 0; k < 2; k++) {
-      std::string xmlFileName;
-      if (k == 0) xmlFileName = "testPDE.xml";
-      if (k == 1) xmlFileName = "testPDE1.xml";
+  for (int k = 0; k < 2; k++) {
+    std::string xmlFileName;
+    if (k == 0) xmlFileName = "testPDE.xml";
+    if (k == 1) xmlFileName = "testPDE1.xml";
 
-      if (lib == Xpetra::UseTpetra) {
+    if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
-        typedef Tpetra::Operator<SC,LO,GO,NO> tpetra_operator_type;
-
-        int numPDEs=3;
-
-        // Matrix
-        GO nx = 972;
-        RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-        RCP<const Map > map = Op->getRowMap();
-
-        Teuchos::ParameterList clist;
-        clist.set("nx", (nx * comm->getSize())/numPDEs);
-        RCP<const Map>   cmap        = MapFactory::Build(lib, Teuchos::as<size_t>((nx * comm->getSize())/numPDEs), Teuchos::as<int>(0), comm);
-        RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,dMultiVector>("1D", cmap, clist);
-        RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), numPDEs);
-        if (numPDEs == 1) {
-          nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-        } else {
-          for (int i = 0; i < numPDEs; i++) {
-            Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
-            for (int j = 0; j < nsData.size(); j++) {
-              GlobalOrdinal GID = Op->getDomainMap()->getGlobalElement(j) - Op->getDomainMap()->getIndexBase();
-              if ((GID-i) % numPDEs == 0)
-                nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
-            }
-          }
-        }
+      typedef Tpetra::Operator<SC, LO, GO, NO> tpetra_operator_type;
 
-        // Normalized RHS
-        RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-        RHS1->setSeed(846930886);
-        RHS1->randomize();
-        Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-        RHS1->norm2(norms);
-        RHS1->scale(1/norms[0]);
-
-        // Zero initial guess
-        RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-        X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-        RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> >   tpA           = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op);
-        RCP<Tpetra::MultiVector<real_type,LO,GO,NO> > tpcoordinates = MueLu::Utilities<real_type,LO,GO,NO>::MV2NonConstTpetraMV(coordinates);
-        RCP<Tpetra::MultiVector<SC,LO,GO,NO> > tpnullspace   = Utils::MV2NonConstTpetraMV(nullspace);
-
-        Teuchos::ParameterList paramList;
-        Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *tpA->getDomainMap()->getComm());
-        Teuchos::ParameterList& userParamList = paramList.sublist("user data");
-        userParamList.set<RCP<Tpetra::MultiVector<real_type,LO,GO,NO> > >("Coordinates", tpcoordinates);
-        userParamList.set<RCP<Tpetra::MultiVector<SC,LO,GO,NO> > >("Nullspace", tpnullspace);
-        RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA), paramList);
-        tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-        out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-            Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-#else
-        std::cout << "Skip PetraOperator::CreatePreconditioner_PDESystem: Tpetra is not available (with GO=int enabled)" << std::endl;
-#endif // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
+      int numPDEs = 3;
 
-      } else if (lib == Xpetra::UseEpetra) {
-#ifdef HAVE_MUELU_EPETRA
-        int numPDEs=3;
-
-        // Matrix
-        GO nx = 972;
-        RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-        RCP<const Map > map = Op->getRowMap();
-
-        Teuchos::ParameterList clist;
-        clist.set("nx", (nx * comm->getSize())/numPDEs);
-        RCP<const Map>   cmap        = MapFactory::Build(lib, Teuchos::as<size_t>((nx * comm->getSize())/numPDEs), Teuchos::as<int>(0), comm);
-        RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type,LO,GO,Map,dMultiVector>("1D", cmap, clist);
-        RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), numPDEs);
-        if (numPDEs == 1) {
-          nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-        } else {
-          for (int i = 0; i < numPDEs; i++) {
-            Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
-            for (int j = 0; j < nsData.size(); j++) {
-              GlobalOrdinal GID = Op->getDomainMap()->getGlobalElement(j) - Op->getDomainMap()->getIndexBase();
-              if ((GID-i) % numPDEs == 0)
-                nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
-            }
+      // Matrix
+      GO nx              = 972;
+      RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+      RCP<const Map> map = Op->getRowMap();
+
+      Teuchos::ParameterList clist;
+      clist.set("nx", (nx * comm->getSize()) / numPDEs);
+      RCP<const Map> cmap           = MapFactory::Build(lib, Teuchos::as<size_t>((nx * comm->getSize()) / numPDEs), Teuchos::as<int>(0), comm);
+      RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, dMultiVector>("1D", cmap, clist);
+      RCP<MultiVector> nullspace    = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), numPDEs);
+      if (numPDEs == 1) {
+        nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+      } else {
+        for (int i = 0; i < numPDEs; i++) {
+          Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
+          for (int j = 0; j < nsData.size(); j++) {
+            GlobalOrdinal GID = Op->getDomainMap()->getGlobalElement(j) - Op->getDomainMap()->getIndexBase();
+            if ((GID - i) % numPDEs == 0)
+              nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
           }
         }
-
-        // Normalized RHS
-        RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-        RHS1->setSeed(846930886);
-        RHS1->randomize();
-        Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-        RHS1->norm2(norms);
-        RHS1->scale(1/norms[0]);
-
-        // Zero initial guess
-        RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-        X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-        RCP<Epetra_CrsMatrix>   epA           = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstEpetraCrs(Op);
-        RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type,LO,GO,NO>::MV2NonConstEpetraMV(coordinates);
-        RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
-
-        Teuchos::ParameterList paramList;
-        Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName,
-                                                         Teuchos::Ptr<Teuchos::ParameterList>(&paramList),
-                                                         *map->getComm());
-        paramList.set("use kokkos refactor", false); // Done to avoid having kokkos factories called with Epetra
-        Teuchos::ParameterList& userParamList = paramList.sublist("user data");
-        userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
-        userParamList.set<RCP<Epetra_MultiVector> >("Nullspace",   epnullspace);
-        RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
-
-        eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-        out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-            Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-
-#else
-        std::cout << "Skip PetraOperator::CreatePreconditioner_PDESystem: Epetra is not available" << std::endl;
-#endif
-      } else {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
       }
-    }
-#endif // defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, ReusePreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using Teuchos::RCP;
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> dMultiVector;
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    std::string xmlFileName = "testReuse.xml";
-
-    if (lib == Xpetra::UseTpetra) {
-#if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
-      typedef Tpetra::Operator<SC,LO,GO,NO> tpetra_operator_type;
-
-      // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
 
       // Normalized RHS
       RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
@@ -524,36 +390,54 @@ namespace MueLuTests {
       RHS1->randomize();
       Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
       RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+      RHS1->scale(1 / norms[0]);
 
       // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
+      RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
       X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
-      RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > tpA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op);
-
-      RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA), xmlFileName);
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-
-      // Reuse preconditioner
-      MueLu::ReuseTpetraPreconditioner(tpA, *tH);
+      RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > tpA                    = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op);
+      RCP<Tpetra::MultiVector<real_type, LO, GO, NO> > tpcoordinates = MueLu::Utilities<real_type, LO, GO, NO>::MV2NonConstTpetraMV(coordinates);
+      RCP<Tpetra::MultiVector<SC, LO, GO, NO> > tpnullspace          = Utils::MV2NonConstTpetraMV(nullspace);
 
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+      Teuchos::ParameterList paramList;
+      Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *tpA->getDomainMap()->getComm());
+      Teuchos::ParameterList& userParamList = paramList.sublist("user data");
+      userParamList.set<RCP<Tpetra::MultiVector<real_type, LO, GO, NO> > >("Coordinates", tpcoordinates);
+      userParamList.set<RCP<Tpetra::MultiVector<SC, LO, GO, NO> > >("Nullspace", tpnullspace);
+      RCP<MueLu::TpetraOperator<SC, LO, GO, NO> > tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), paramList);
       tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 #else
-      std::cout << "Skip PetraOperator::ReusePreconditioner: Tpetra is not available (with GO=int enabled)" << std::endl;
-#endif // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
+      std::cout << "Skip PetraOperator::CreatePreconditioner_PDESystem: Tpetra is not available (with GO=int enabled)" << std::endl;
+#endif  // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
     } else if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
+      int numPDEs = 3;
+
       // Matrix
-      GO nx = 1000;
-      RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-      RCP<const Map > map = Op->getRowMap();
+      GO nx              = 972;
+      RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+      RCP<const Map> map = Op->getRowMap();
+
+      Teuchos::ParameterList clist;
+      clist.set("nx", (nx * comm->getSize()) / numPDEs);
+      RCP<const Map> cmap           = MapFactory::Build(lib, Teuchos::as<size_t>((nx * comm->getSize()) / numPDEs), Teuchos::as<int>(0), comm);
+      RCP<dMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real_type, LO, GO, Map, dMultiVector>("1D", cmap, clist);
+      RCP<MultiVector> nullspace    = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), numPDEs);
+      if (numPDEs == 1) {
+        nullspace->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+      } else {
+        for (int i = 0; i < numPDEs; i++) {
+          Teuchos::ArrayRCP<Scalar> nsData = nullspace->getDataNonConst(i);
+          for (int j = 0; j < nsData.size(); j++) {
+            GlobalOrdinal GID = Op->getDomainMap()->getGlobalElement(j) - Op->getDomainMap()->getIndexBase();
+            if ((GID - i) % numPDEs == 0)
+              nsData[j] = Teuchos::ScalarTraits<Scalar>::one();
+          }
+        }
+      }
 
       // Normalized RHS
       RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
@@ -561,162 +445,255 @@ namespace MueLuTests {
       RHS1->randomize();
       Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
       RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+      RHS1->scale(1 / norms[0]);
 
       // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
+      RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
       X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
-      RCP<Epetra_CrsMatrix> epA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstEpetraCrs(Op);
+      RCP<Epetra_CrsMatrix> epA             = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstEpetraCrs(Op);
+      RCP<Epetra_MultiVector> epcoordinates = MueLu::Utilities<real_type, LO, GO, NO>::MV2NonConstEpetraMV(coordinates);
+      RCP<Epetra_MultiVector> epnullspace   = Utils::MV2NonConstEpetraMV(nullspace);
 
       Teuchos::ParameterList paramList;
       Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName,
                                                        Teuchos::Ptr<Teuchos::ParameterList>(&paramList),
                                                        *map->getComm());
-      paramList.set("use kokkos refactor", false);
+      paramList.set("use kokkos refactor", false);  // Done to avoid having kokkos factories called with Epetra
+      Teuchos::ParameterList& userParamList = paramList.sublist("user data");
+      userParamList.set<RCP<Epetra_MultiVector> >("Coordinates", epcoordinates);
+      userParamList.set<RCP<Epetra_MultiVector> >("Nullspace", epnullspace);
       RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
 
       eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
-      // Reuse preconditioner
-      MueLu::ReuseEpetraPreconditioner(epA, *eH);
-
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 #else
-      std::cout << "Skip PetraOperator::ReusePreconditioner: Epetra is not available" << std::endl;
+      std::cout << "Skip PetraOperator::CreatePreconditioner_PDESystem: Epetra is not available" << std::endl;
 #endif
-
     } else {
       TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
     }
   }
+#endif  // defined(HAVE_MUELU_ZOLTAN) && defined(HAVE_MPI)
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, ReusePreconditioner2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, ReusePreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    if (Teuchos::ScalarTraits<SC>::isComplex)
-      return;
+  using Teuchos::RCP;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> dMultiVector;
 
-    using Teuchos::RCP;
-    using Teuchos::null;
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> dMultiVector;
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  std::string xmlFileName = "testReuse.xml";
 
-    Teuchos::ParameterList params;
-    params.set("aggregation: type","uncoupled");
-    params.set("aggregation: drop tol", 0.02);
-    params.set("coarse: max size", Teuchos::as<int>(500));
+  if (lib == Xpetra::UseTpetra) {
+#if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
+    typedef Tpetra::Operator<SC, LO, GO, NO> tpetra_operator_type;
 
-    if (lib == Xpetra::UseTpetra) {
-      typedef Tpetra::Operator<SC,LO,GO,NO> tpetra_operator_type;
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
 
-      // Matrix
-      std::string matrixFile("TestMatrices/fuego0.mm");
-      RCP<const Map> rowmap        = MapFactory::Build(lib, Teuchos::as<size_t>(1500), Teuchos::as<int>(0), comm);
-      RCP<Matrix>     Op  = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, rowmap, null, null, null);
-      RCP<const Map > map = Op->getRowMap();
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
 
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
 
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-      RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > tpA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op);
-      RCP<MueLu::TpetraOperator<SC,LO,GO,NO> > tH = MueLu::CreateTpetraPreconditioner<SC,LO,GO,NO>(RCP<tpetra_operator_type>(tpA), params);
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+    RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > tpA = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op);
 
-      // Reuse preconditioner
+    RCP<MueLu::TpetraOperator<SC, LO, GO, NO> > tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), xmlFileName);
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 
-      matrixFile = "TestMatrices/fuego1.mm";
-      RCP<Matrix>     Op2  = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, rowmap, null, null, null);
-      RCP<Tpetra::CrsMatrix<SC,LO,GO,NO> > tpA2 = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Op2);
+    // Reuse preconditioner
+    MueLu::ReuseTpetraPreconditioner(tpA, *tH);
 
-      MueLu::ReuseTpetraPreconditioner(tpA2, *tH);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+#else
+    std::cout << "Skip PetraOperator::ReusePreconditioner: Tpetra is not available (with GO=int enabled)" << std::endl;
+#endif  // #if defined(HAVE_MUELU_TPETRA_INST_INT_INT)
 
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-    } else if (lib == Xpetra::UseEpetra) {
+  } else if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      // Matrix
-      std::string matrixFile("TestMatrices/fuego0.mm");
-      RCP<const Map> rowmap        = MapFactory::Build(lib, Teuchos::as<size_t>(1500), Teuchos::as<int>(0), comm);
-      RCP<Matrix>     Op  = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, rowmap, null, null, null);
-      RCP<const Map > map = Op->getRowMap();
-
-      // Normalized RHS
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
-
-      // Zero initial guess
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-      params.set("use kokkos refactor", false);
-      RCP<Epetra_CrsMatrix> epA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstEpetraCrs(Op);
-      RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, params);
-
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
-
-      // Reuse preconditioner
-
-      matrixFile = "TestMatrices/fuego1.mm";
-      RCP<Matrix>     Op2  = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, rowmap, null, null, null);
-      epA = MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstEpetraCrs(Op);
-      MueLu::ReuseEpetraPreconditioner(epA, *eH);
+    // Matrix
+    GO nx              = 1000;
+    RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+    RCP<const Map> map = Op->getRowMap();
+
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
+
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+    RCP<Epetra_CrsMatrix> epA = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstEpetraCrs(Op);
+
+    Teuchos::ParameterList paramList;
+    Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName,
+                                                     Teuchos::Ptr<Teuchos::ParameterList>(&paramList),
+                                                     *map->getComm());
+    paramList.set("use kokkos refactor", false);
+    RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, paramList);
+
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+
+    // Reuse preconditioner
+    MueLu::ReuseEpetraPreconditioner(epA, *eH);
+
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+#else
+    std::cout << "Skip PetraOperator::ReusePreconditioner: Epetra is not available" << std::endl;
+#endif
 
-      X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-      eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
-      out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) <<
-          Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
+  }
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PetraOperator, ReusePreconditioner2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  if (Teuchos::ScalarTraits<SC>::isComplex)
+    return;
+
+  using Teuchos::null;
+  using Teuchos::RCP;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> dMultiVector;
+
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  Teuchos::ParameterList params;
+  params.set("aggregation: type", "uncoupled");
+  params.set("aggregation: drop tol", 0.02);
+  params.set("coarse: max size", Teuchos::as<int>(500));
+
+  if (lib == Xpetra::UseTpetra) {
+    typedef Tpetra::Operator<SC, LO, GO, NO> tpetra_operator_type;
+
+    // Matrix
+    std::string matrixFile("TestMatrices/fuego0.mm");
+    RCP<const Map> rowmap = MapFactory::Build(lib, Teuchos::as<size_t>(1500), Teuchos::as<int>(0), comm);
+    RCP<Matrix> Op        = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, rowmap, null, null, null);
+    RCP<const Map> map    = Op->getRowMap();
+
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
+
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+    RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > tpA    = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op);
+    RCP<MueLu::TpetraOperator<SC, LO, GO, NO> > tH = MueLu::CreateTpetraPreconditioner<SC, LO, GO, NO>(RCP<tpetra_operator_type>(tpA), params);
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+
+    // Reuse preconditioner
+
+    matrixFile                                   = "TestMatrices/fuego1.mm";
+    RCP<Matrix> Op2                              = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, rowmap, null, null, null);
+    RCP<Tpetra::CrsMatrix<SC, LO, GO, NO> > tpA2 = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstTpetraCrs(Op2);
+
+    MueLu::ReuseTpetraPreconditioner(tpA2, *tH);
+
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+  } else if (lib == Xpetra::UseEpetra) {
+#ifdef HAVE_MUELU_EPETRA
+    // Matrix
+    std::string matrixFile("TestMatrices/fuego0.mm");
+    RCP<const Map> rowmap = MapFactory::Build(lib, Teuchos::as<size_t>(1500), Teuchos::as<int>(0), comm);
+    RCP<Matrix> Op        = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, rowmap, null, null, null);
+    RCP<const Map> map    = Op->getRowMap();
+
+    // Normalized RHS
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
+
+    // Zero initial guess
+    RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+    params.set("use kokkos refactor", false);
+    RCP<Epetra_CrsMatrix> epA     = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstEpetraCrs(Op);
+    RCP<MueLu::EpetraOperator> eH = MueLu::CreateEpetraPreconditioner(epA, params);
+
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
+
+    // Reuse preconditioner
+
+    matrixFile      = "TestMatrices/fuego1.mm";
+    RCP<Matrix> Op2 = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, rowmap, null, null, null);
+    epA             = MueLu::Utilities<SC, LO, GO, NO>::Op2NonConstEpetraCrs(Op);
+    MueLu::ReuseEpetraPreconditioner(epA, *eH);
+
+    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+    eH->Apply(*(Utils::MV2EpetraMV(RHS1)), *(Utils::MV2NonConstEpetraMV(X1)));
+    out << "after apply, ||b-A*x||_2 = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << Utils::ResidualNorm(*Op, *X1, *RHS1) << std::endl;
 #else
-      std::cout << "Skip PetraOperator::ReusePreconditioner: Epetra is not available" << std::endl;
+    std::cout << "Skip PetraOperator::ReusePreconditioner: Epetra is not available" << std::endl;
 #endif
 
-    } else {
-      TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
-    }
+  } else {
+    TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::InvalidArgument, "Unknown Xpetra lib");
   }
+}
 
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, CreatePreconditioner, Scalar, LO, GO, Node) \
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, CreatePreconditioner, Scalar, LO, GO, Node)           \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, CreatePreconditioner_XMLOnList, Scalar, LO, GO, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, CreatePreconditioner_PDESystem, Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, ReusePreconditioner, Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, ReusePreconditioner2, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, ReusePreconditioner, Scalar, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PetraOperator, ReusePreconditioner2, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
 
 // #endif
diff --git a/packages/muelu/test/unit_tests/Adapters/TpetraOperatorAdapter.cpp b/packages/muelu/test/unit_tests/Adapters/TpetraOperatorAdapter.cpp
index 46347c6f2bc9..9ded59059907 100644
--- a/packages/muelu/test/unit_tests/Adapters/TpetraOperatorAdapter.cpp
+++ b/packages/muelu/test/unit_tests/Adapters/TpetraOperatorAdapter.cpp
@@ -67,160 +67,153 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TpetraOperator, Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TpetraOperator, Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
 #if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS2)
-    typedef MueLu::Utilities<SC,LO,GO,NO> Utils;
-    typedef MueLu::TpetraOperator<SC,LO,GO,NO> muelu_tpetra_operator_type;
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  typedef MueLu::Utilities<SC, LO, GO, NO> Utils;
+  typedef MueLu::TpetraOperator<SC, LO, GO, NO> muelu_tpetra_operator_type;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
 
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra )
-    {
-      //matrix
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      RCP<Matrix> Op = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(6561*comm->getSize());  //=8*3^6
-      RCP<const Map > map = Op->getRowMap();
-
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
-      nullSpace->putScalar( (SC) 1.0);
-      //Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-      Teuchos::Array<magnitude_type> norms(1);
-      nullSpace->norm1(norms);
-
-      RCP<Hierarchy> H = rcp(new Hierarchy());
-      H->setDefaultVerbLevel(Teuchos::VERB_NONE);
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra) {
+    // matrix
+    RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+    RCP<Matrix> Op                     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(6561 * comm->getSize());  //=8*3^6
+    RCP<const Map> map                 = Op->getRowMap();
 
-      RCP<MueLu::Level> Finest = H->GetLevel();
-      Finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
-      Finest->Set("A", Op);
-      H->Setup();
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+    nullSpace->putScalar((SC)1.0);
+    // Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+    Teuchos::Array<magnitude_type> norms(1);
+    nullSpace->norm1(norms);
 
+    RCP<Hierarchy> H = rcp(new Hierarchy());
+    H->setDefaultVerbLevel(Teuchos::VERB_NONE);
 
-      // ------------- test Tpetra Operator wrapping MueLu hierarchy ------------
-      RCP<muelu_tpetra_operator_type> tH = rcp(new muelu_tpetra_operator_type(H));
+    RCP<MueLu::Level> Finest = H->GetLevel();
+    Finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
+    Finest->Set("A", Op);
+    H->Setup();
 
-      RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    // ------------- test Tpetra Operator wrapping MueLu hierarchy ------------
+    RCP<muelu_tpetra_operator_type> tH = rcp(new muelu_tpetra_operator_type(H));
 
-      //normalized RHS, zero initial guess
-      RHS1->setSeed(846930886);
-      RHS1->randomize();
-      RHS1->norm2(norms);
-      RHS1->scale(1/norms[0]);
+    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
 
-      X1->putScalar( (SC) 0.0);
+    // normalized RHS, zero initial guess
+    RHS1->setSeed(846930886);
+    RHS1->randomize();
+    RHS1->norm2(norms);
+    RHS1->scale(1 / norms[0]);
 
-      tH->apply(*(Utils::MV2TpetraMV(RHS1)),*(Utils::MV2NonConstTpetraMV(X1)));
+    X1->putScalar((SC)0.0);
 
-      X1->norm2(norms);
-      out << "after apply, ||X1|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    tH->apply(*(Utils::MV2TpetraMV(RHS1)), *(Utils::MV2NonConstTpetraMV(X1)));
 
-      // -------------- test MueLu Hierarchy directly -----------------------
-      RCP<MultiVector> RHS2 = MultiVectorFactory::Build(Op->getRowMap(), 1);
-      RCP<MultiVector> X2   = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    X1->norm2(norms);
+    out << "after apply, ||X1|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      //normalized RHS, zero initial guess
-      RHS2->setSeed(846930886);
-      RHS2->randomize();
-      RHS2->norm2(norms);
-      RHS2->scale(1/norms[0]);
+    // -------------- test MueLu Hierarchy directly -----------------------
+    RCP<MultiVector> RHS2 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    RCP<MultiVector> X2   = MultiVectorFactory::Build(Op->getRowMap(), 1);
 
-      X2->putScalar( (SC) 0.0);
+    // normalized RHS, zero initial guess
+    RHS2->setSeed(846930886);
+    RHS2->randomize();
+    RHS2->norm2(norms);
+    RHS2->scale(1 / norms[0]);
 
-      int iterations=1;
-      H->Iterate(*RHS2, *X2, iterations);
+    X2->putScalar((SC)0.0);
 
-      X2->norm2(norms);
-      out << "after apply, ||X2|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    int iterations = 1;
+    H->Iterate(*RHS2, *X2, iterations);
 
-      RCP<MultiVector> diff = MultiVectorFactory::Build(Op->getRowMap(),1);
-      diff->putScalar(0.0);
+    X2->norm2(norms);
+    out << "after apply, ||X2|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      diff->update(1.0,*X1,-1.0,*X2,0.0);
-      diff->norm2(norms);
-      TEST_EQUALITY(norms[0]<1e-10, true);
+    RCP<MultiVector> diff = MultiVectorFactory::Build(Op->getRowMap(), 1);
+    diff->putScalar(0.0);
 
-    } else {
+    diff->update(1.0, *X1, -1.0, *X2, 0.0);
+    diff->norm2(norms);
+    TEST_EQUALITY(norms[0] < 1e-10, true);
 
-      out << "This test is enabled only for linAlgebra=Tpetra." << std::endl;
-
-    }
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Ifpack2, Amesos2)." << std::endl;
-#   endif
-  } //Apply
+  } else {
+    out << "This test is enabled only for linAlgebra=Tpetra." << std::endl;
+  }
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, Ifpack2, Amesos2)." << std::endl;
+#endif
+}  // Apply
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TpetraOperator, Getters, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TpetraOperator, Getters, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
 #if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS2)
-    using Utils = MueLu::Utilities<SC, LO, GO, NO>;
-    using TpetraOperatorType =  MueLu::TpetraOperator<SC, LO, GO, NO>;
-    using MagnitudeType = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using Utils              = MueLu::Utilities<SC, LO, GO, NO>;
+  using TpetraOperatorType = MueLu::TpetraOperator<SC, LO, GO, NO>;
+  using MagnitudeType      = typename Teuchos::ScalarTraits<SC>::magnitudeType;
 
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra)
-    {
-      const auto Op = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra) {
+    const auto Op = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
 
-      ////////////////////////////////////////
-      //////////   WITH HIERARCHY   //////////
-      ////////////////////////////////////////
-      {
-        const auto map = Op->getRowMap();
-        auto H = rcp(new Hierarchy());
-        H->setDefaultVerbLevel(Teuchos::VERB_NONE);
+    ////////////////////////////////////////
+    //////////   WITH HIERARCHY   //////////
+    ////////////////////////////////////////
+    {
+      const auto map = Op->getRowMap();
+      auto H         = rcp(new Hierarchy());
+      H->setDefaultVerbLevel(Teuchos::VERB_NONE);
 
-        auto Finest = H->GetLevel();
-        Finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
-        Finest->Set("A", Op);
+      auto Finest = H->GetLevel();
+      Finest->setDefaultVerbLevel(Teuchos::VERB_NONE);
+      Finest->Set("A", Op);
 
-        H->Setup();
-        auto tH = rcp(new TpetraOperatorType(H));
+      H->Setup();
+      auto tH = rcp(new TpetraOperatorType(H));
 
-        TEST_EQUALITY(tH->GetOperator(), Teuchos::null);
-        TEST_INEQUALITY(tH->GetHierarchy(), Teuchos::null);
+      TEST_EQUALITY(tH->GetOperator(), Teuchos::null);
+      TEST_INEQUALITY(tH->GetHierarchy(), Teuchos::null);
 
-        TEST_INEQUALITY(tH->getRangeMap(), Teuchos::null);
-        TEST_INEQUALITY(tH->getDomainMap(), Teuchos::null);
+      TEST_INEQUALITY(tH->getRangeMap(), Teuchos::null);
+      TEST_INEQUALITY(tH->getDomainMap(), Teuchos::null);
 
-        // Hardcoded false
-        TEST_EQUALITY(tH->hasTransposeApply(), false);
-      }
+      // Hardcoded false
+      TEST_EQUALITY(tH->hasTransposeApply(), false);
+    }
 
-      ///////////////////////////////////////
-      //////////   WITH OPERATOR   //////////
-      ///////////////////////////////////////
-      {
-        auto tO = rcp(new TpetraOperatorType((Teuchos::RCP<Xpetra::Operator<SC, LO, GO, NO>>)(Op)));
+    ///////////////////////////////////////
+    //////////   WITH OPERATOR   //////////
+    ///////////////////////////////////////
+    {
+      auto tO = rcp(new TpetraOperatorType((Teuchos::RCP<Xpetra::Operator<SC, LO, GO, NO>>)(Op)));
 
-        TEST_INEQUALITY(tO->GetOperator(), Teuchos::null);
-        TEST_EQUALITY(tO->GetHierarchy(), Teuchos::null);
+      TEST_INEQUALITY(tO->GetOperator(), Teuchos::null);
+      TEST_EQUALITY(tO->GetHierarchy(), Teuchos::null);
 
-        TEST_INEQUALITY(tO->getRangeMap(), Teuchos::null);
-        TEST_INEQUALITY(tO->getDomainMap(), Teuchos::null);
+      TEST_INEQUALITY(tO->getRangeMap(), Teuchos::null);
+      TEST_INEQUALITY(tO->getDomainMap(), Teuchos::null);
 
-        // Hardcoded false
-        TEST_EQUALITY(tO->hasTransposeApply(), false);
-      }
+      // Hardcoded false
+      TEST_EQUALITY(tO->hasTransposeApply(), false);
     }
+  }
 #else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Ifpack2, Amesos2)." << std::endl;
+  out << "Skipping test because some required packages are not enabled (Tpetra, Ifpack2, Amesos2)." << std::endl;
 #endif
-  }
+}
 
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TpetraOperator, Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TpetraOperator, Getters, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TpetraOperator, Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TpetraOperator, Getters, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/AggregateQualityEstimateFactory.cpp b/packages/muelu/test/unit_tests/AggregateQualityEstimateFactory.cpp
index 646a904c4b06..50f1ceaaaa02 100644
--- a/packages/muelu/test/unit_tests/AggregateQualityEstimateFactory.cpp
+++ b/packages/muelu/test/unit_tests/AggregateQualityEstimateFactory.cpp
@@ -66,539 +66,515 @@
 
 namespace MueLuTests {
 
-  template<typename LocalOrdinal>
-  std::vector<LocalOrdinal> flip_agg_horizontal(const std::vector<LocalOrdinal>& indices, LocalOrdinal nx) {
-	
-    std::vector<LocalOrdinal> flipped;
+template <typename LocalOrdinal>
+std::vector<LocalOrdinal> flip_agg_horizontal(const std::vector<LocalOrdinal>& indices, LocalOrdinal nx) {
+  std::vector<LocalOrdinal> flipped;
 
-    for (LocalOrdinal idx : indices) {
-      LocalOrdinal x = idx % nx;
-      LocalOrdinal y = idx / nx;
-
-      flipped.push_back(nx*y - x);
-    }
-
-    std::sort(flipped.begin(), flipped.end());
-    return flipped;
+  for (LocalOrdinal idx : indices) {
+    LocalOrdinal x = idx % nx;
+    LocalOrdinal y = idx / nx;
 
+    flipped.push_back(nx * y - x);
   }
 
-  template<typename LocalOrdinal>
-  std::vector<LocalOrdinal> flip_agg_vertical(const std::vector<LocalOrdinal>& indices, LocalOrdinal nx) {
-	
-    std::vector<LocalOrdinal> flipped;
+  std::sort(flipped.begin(), flipped.end());
+  return flipped;
+}
 
-    for (LocalOrdinal idx : indices) {
-      LocalOrdinal x = idx % nx;
-      LocalOrdinal y = idx / nx;
-
-      flipped.push_back(-nx*y + x);
-    }
+template <typename LocalOrdinal>
+std::vector<LocalOrdinal> flip_agg_vertical(const std::vector<LocalOrdinal>& indices, LocalOrdinal nx) {
+  std::vector<LocalOrdinal> flipped;
 
-    std::sort(flipped.begin(), flipped.end());
-    return flipped;
+  for (LocalOrdinal idx : indices) {
+    LocalOrdinal x = idx % nx;
+    LocalOrdinal y = idx / nx;
 
+    flipped.push_back(-nx * y + x);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Xpetra::MultiVector<MT,LO,GO,Node> MultiVectorDouble;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
-
-    RCP<AggregateQualityEstimateFactory> aggQualityEstimateFactory = rcp(new AggregateQualityEstimateFactory);
-    TEST_EQUALITY(aggQualityEstimateFactory != Teuchos::null, true);
-
-  } // Constructor test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, Poisson2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Xpetra::MultiVector<MT,LO,GO,NO> MultiVectorDouble;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level level;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
-
-    GO nx = 20*comm->getSize();
-    GO ny = nx;
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build2DPoisson(nx, ny);
-    level.Set("A",Op);
+  std::sort(flipped.begin(), flipped.end());
+  return flipped;
+}
 
-    AggregateQualityEstimateFactory aggQualityEstimateFactory;
-    std::cout<< *(aggQualityEstimateFactory.GetValidParameterList())<<std::endl;
-    aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(false));
-    aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
-    aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Xpetra::MultiVector<MT, LO, GO, Node> MultiVectorDouble;
 
-    level.Request("AggregateQualities", &aggQualityEstimateFactory);
-    level.Request(aggQualityEstimateFactory);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    out << "Getting aggregate qualities...\n\n";
+  typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
 
-    RCP<MultiVectorDouble> aggQualities = level.Get< RCP<MultiVectorDouble> >("AggregateQualities", &aggQualityEstimateFactory);
-    
-    out << "Testing aggregate qualities to make sure all aggregates are of good quality...\n\n";
+  RCP<AggregateQualityEstimateFactory> aggQualityEstimateFactory = rcp(new AggregateQualityEstimateFactory);
+  TEST_EQUALITY(aggQualityEstimateFactory != Teuchos::null, true);
 
-    ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
-
-    for (size_t i=0;i<aggQualities->getLocalLength();++i) {
-      out << "Aggregate " << i << ": " << aggQualitiesLocalData[i] << "\n";
-      TEST_COMPARE(aggQualitiesLocalData[i], > , 0.0);
-      TEST_COMPARE(aggQualitiesLocalData[i], <=, 20.0);
-    }
-    
+}  // Constructor test
 
-  } // Poisson 2D test
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, Poisson2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, AnisotropicDiffusion2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Xpetra::MultiVector<MT,LO,GO,Node> MultiVectorDouble;
-    typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    out << "version: " << MueLu::Version() << std::endl;
+  typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Xpetra::MultiVector<MT, LO, GO, NO> MultiVectorDouble;
 
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
 
-    // Don't test for complex or ordinal - matrix reader won't work
-    if (TST::isComplex || TST::isOrdinal) {success=true; return;}
+  Level level;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  GO nx          = 20 * comm->getSize();
+  GO ny          = nx;
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build2DPoisson(nx, ny);
+  level.Set("A", Op);
 
-    LO nx = (LO) 20;
+  AggregateQualityEstimateFactory aggQualityEstimateFactory;
+  std::cout << *(aggQualityEstimateFactory.GetValidParameterList()) << std::endl;
+  aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(false));
+  aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
+  aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
 
-    LO num_nodes = (LO) nx*nx;
-    RCP<Matrix> A;
-    RCP<Map> map_for_read = MapFactory::Build(TestHelpers::Parameters::getLib(),num_nodes,0,comm);
-    try {
-      A =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/aniso2dx.mat", map_for_read);
-    }
-    catch (...) {
-      // Sometimes the matrix reader just fails.
-      return;
-    };
+  level.Request("AggregateQualities", &aggQualityEstimateFactory);
+  level.Request(aggQualityEstimateFactory);
 
-    std::vector<std::string> test_matrices = { "TestMatrices/aniso2dx.mat", "TestMatrices/aniso2dy.mat", "TestMatrices/iso2d.mat" };
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  out << "Getting aggregate qualities...\n\n";
 
-    const std::vector<LO> AGG0_NODES = {0,1};
-    const std::vector<LO> AGG1_NODES = {0,nx};
-    const std::vector<LO> AGG2_NODES = {0,nx,nx+1};
-    const std::vector<LO> AGG3_NODES = {0,1,nx,nx+1};
-    const std::vector<LO> AGG4_NODES = {0,1,2};
-    const std::vector<LO> AGG5_NODES = {0,nx,2*nx};
-    const std::vector<LO> AGG6_NODES = {0,nx,nx+1,2*nx};
-    const std::vector<LO> AGG7_NODES = {1,nx,nx+1,nx+2};
-    const std::vector<LO> AGG8_NODES = {1,nx,nx+1,nx+2,2*nx+1};
-    const std::vector<LO> AGG9_NODES = {0,1,2,nx,nx+1,nx+2};
-    const std::vector<LO> AGG10_NODES = {0,1,nx,nx+1,2*nx,2*nx+1};
-    const std::vector<LO> AGG11_NODES = {0,1,2,nx,nx+1,nx+2,2*nx,2*nx+1,2*nx+2};
-    const std::vector<LO> AGG12_NODES = {0,1,nx,nx+1,nx+2};
-    const std::vector<LO> AGG13_NODES = {0,1,nx,nx+1,2*nx};
-    const std::vector<LO> AGG14_NODES = {0,1,2,nx,nx+1,nx+2,2*nx};
-    const std::vector<LO> AGG15_NODES = {0,1,2,nx,nx+1,nx+2,2*nx+1};
-    const std::vector<LO> AGG16_NODES = {0,1,2,nx,nx+1,2*nx,2*nx+1};
-    const std::vector<LO> AGG17_NODES = {0,1,nx,nx+1,nx+2,2*nx,2*nx+1};
-    const std::vector<LO> AGG18_NODES = {1,nx,nx+1,nx+2,2*nx+1,2*nx+2};
+  RCP<MultiVectorDouble> aggQualities = level.Get<RCP<MultiVectorDouble>>("AggregateQualities", &aggQualityEstimateFactory);
 
-    const std::vector<std::vector<LO>> AGG_TYPES = {AGG0_NODES, AGG1_NODES, AGG2_NODES,
-						    AGG3_NODES, AGG4_NODES, AGG5_NODES,
-						    AGG6_NODES, AGG7_NODES, AGG8_NODES,
-						    AGG9_NODES, AGG10_NODES, AGG11_NODES,
-						    AGG12_NODES, AGG13_NODES, AGG14_NODES,
-						    AGG15_NODES, AGG16_NODES, AGG17_NODES,
-						    AGG18_NODES};
+  out << "Testing aggregate qualities to make sure all aggregates are of good quality...\n\n";
 
-    const std::vector<MT> AGG_QUALITIES = {2002., 2.002, 2670.000811,
-					       2001.9998, 4003.999998, 4.004,
-					       3003.666728, 4003.999198, 4003.998397,
-					       4003.999198, 2001.999599, 4003.999198,
-					       4004.666568, 2403.067729, 4886.458381,
-					       4003.998397, 4005.666904, 4004.665757,
-					       4004.665841};
+  ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
 
-    for (int i=0;i<3;i++) {
+  for (size_t i = 0; i < aggQualities->getLocalLength(); ++i) {
+    out << "Aggregate " << i << ": " << aggQualitiesLocalData[i] << "\n";
+    TEST_COMPARE(aggQualitiesLocalData[i], >, 0.0);
+    TEST_COMPARE(aggQualitiesLocalData[i], <=, 20.0);
+  }
 
-      out << "Aggregating Matrix " << i << "..." << std::endl;
+}  // Poisson 2D test
 
-      Level level;
-      TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, AnisotropicDiffusion2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Xpetra::MultiVector<MT, LO, GO, Node> MultiVectorDouble;
+  typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
 
-      RCP<Matrix> A_agg =  Xpetra::IO<SC, LO, GO, NO>::Read(test_matrices.at(i), map_for_read);
+  out << "version: " << MueLu::Version() << std::endl;
 
-      level.Set("A",A_agg);
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
 
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory);
-      dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.1));
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  // Don't test for complex or ordinal - matrix reader won't work
+  if (TST::isComplex || TST::isOrdinal) {
+    success = true;
+    return;
+  }
 
-      RCP<UncoupledAggregationFactory> aggFactory = rcp(new UncoupledAggregationFactory);
-      aggFactory->SetFactory("Graph", dropFact);
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
+
+  LO nx = (LO)20;
+
+  LO num_nodes = (LO)nx * nx;
+  RCP<Matrix> A;
+  RCP<Map> map_for_read = MapFactory::Build(TestHelpers::Parameters::getLib(), num_nodes, 0, comm);
+  try {
+    A = Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/aniso2dx.mat", map_for_read);
+  } catch (...) {
+    // Sometimes the matrix reader just fails.
+    return;
+  };
+
+  std::vector<std::string> test_matrices = {"TestMatrices/aniso2dx.mat", "TestMatrices/aniso2dy.mat", "TestMatrices/iso2d.mat"};
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  const std::vector<LO> AGG0_NODES  = {0, 1};
+  const std::vector<LO> AGG1_NODES  = {0, nx};
+  const std::vector<LO> AGG2_NODES  = {0, nx, nx + 1};
+  const std::vector<LO> AGG3_NODES  = {0, 1, nx, nx + 1};
+  const std::vector<LO> AGG4_NODES  = {0, 1, 2};
+  const std::vector<LO> AGG5_NODES  = {0, nx, 2 * nx};
+  const std::vector<LO> AGG6_NODES  = {0, nx, nx + 1, 2 * nx};
+  const std::vector<LO> AGG7_NODES  = {1, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG8_NODES  = {1, nx, nx + 1, nx + 2, 2 * nx + 1};
+  const std::vector<LO> AGG9_NODES  = {0, 1, 2, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG10_NODES = {0, 1, nx, nx + 1, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG11_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx, 2 * nx + 1, 2 * nx + 2};
+  const std::vector<LO> AGG12_NODES = {0, 1, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG13_NODES = {0, 1, nx, nx + 1, 2 * nx};
+  const std::vector<LO> AGG14_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx};
+  const std::vector<LO> AGG15_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx + 1};
+  const std::vector<LO> AGG16_NODES = {0, 1, 2, nx, nx + 1, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG17_NODES = {0, 1, nx, nx + 1, nx + 2, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG18_NODES = {1, nx, nx + 1, nx + 2, 2 * nx + 1, 2 * nx + 2};
+
+  const std::vector<std::vector<LO>> AGG_TYPES = {AGG0_NODES, AGG1_NODES, AGG2_NODES,
+                                                  AGG3_NODES, AGG4_NODES, AGG5_NODES,
+                                                  AGG6_NODES, AGG7_NODES, AGG8_NODES,
+                                                  AGG9_NODES, AGG10_NODES, AGG11_NODES,
+                                                  AGG12_NODES, AGG13_NODES, AGG14_NODES,
+                                                  AGG15_NODES, AGG16_NODES, AGG17_NODES,
+                                                  AGG18_NODES};
+
+  const std::vector<MT> AGG_QUALITIES = {2002., 2.002, 2670.000811,
+                                         2001.9998, 4003.999998, 4.004,
+                                         3003.666728, 4003.999198, 4003.998397,
+                                         4003.999198, 2001.999599, 4003.999198,
+                                         4004.666568, 2403.067729, 4886.458381,
+                                         4003.998397, 4005.666904, 4004.665757,
+                                         4004.665841};
+
+  for (int i = 0; i < 3; i++) {
+    out << "Aggregating Matrix " << i << "..." << std::endl;
 
-      out << "Coalesce Drop: " << dropFact.get() << std::endl;
-      out << "Aggregation: " << aggFactory.get() << std::endl;
+    Level level;
+    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-      level.Request("Aggregates", aggFactory.get());
+    RCP<Matrix> A_agg = Xpetra::IO<SC, LO, GO, NO>::Read(test_matrices.at(i), map_for_read);
 
-      out << "Building aggregates..." << std::endl;
-	
-      aggFactory->Build(level);
+    level.Set("A", A_agg);
 
-      out << "Built aggregates! Getting them...\n" << std::endl;
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory);
+    dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.1));
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-      RCP<Aggregates> aggs = level.Get< RCP<Aggregates> >("Aggregates", aggFactory.get());
+    RCP<UncoupledAggregationFactory> aggFactory = rcp(new UncoupledAggregationFactory);
+    aggFactory->SetFactory("Graph", dropFact);
 
-      out << "Got aggregates! Computing quality estimate...\n" << std::endl;
+    out << "Coalesce Drop: " << dropFact.get() << std::endl;
+    out << "Aggregation: " << aggFactory.get() << std::endl;
 
-      level.print(out, Teuchos::VERB_EXTREME);
+    level.Request("Aggregates", aggFactory.get());
 
-      level.Set("A",A);
+    out << "Building aggregates..." << std::endl;
 
-      RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory);
-      coarseMapFact->SetFactory("Aggregates", aggFactory);
+    aggFactory->Build(level);
 
-      AggregateQualityEstimateFactory aggQualityEstimateFactory;
-      aggQualityEstimateFactory.SetFactory("Aggregates", aggFactory);
-      aggQualityEstimateFactory.SetFactory("CoarseMap", coarseMapFact);
+    out << "Built aggregates! Getting them...\n"
+        << std::endl;
 
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(false));
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
+    RCP<Aggregates> aggs = level.Get<RCP<Aggregates>>("Aggregates", aggFactory.get());
 
-      level.Request("AggregateQualities", &aggQualityEstimateFactory);
-      // level.Request(aggQualityEstimateFactory);
+    out << "Got aggregates! Computing quality estimate...\n"
+        << std::endl;
 
-      aggQualityEstimateFactory.Build(level);	
+    level.print(out, Teuchos::VERB_EXTREME);
 
-      RCP<MultiVectorDouble> aggQualities = level.Get< RCP<MultiVectorDouble> >("AggregateQualities", &aggQualityEstimateFactory);
-    
-      ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
+    level.Set("A", A);
 
-      ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
-      AggregateQualityEstimateFactory::ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
+    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory);
+    coarseMapFact->SetFactory("Aggregates", aggFactory);
 
-      for (size_t j=0;j<aggQualities->getLocalLength();++j) {
+    AggregateQualityEstimateFactory aggQualityEstimateFactory;
+    aggQualityEstimateFactory.SetFactory("Aggregates", aggFactory);
+    aggQualityEstimateFactory.SetFactory("CoarseMap", coarseMapFact);
 
-	std::vector<LO> nodes;
-	for (LO k=0;k<aggSizes[j];++k) {
-	  nodes.push_back(aggSortedVertices[aggsToIndices[j]+k]);
-	}
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(false));
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
 
-	std::sort(nodes.begin(), nodes.end());
+    level.Request("AggregateQualities", &aggQualityEstimateFactory);
+    // level.Request(aggQualityEstimateFactory);
 
-	bool onBoundary = false;
+    aggQualityEstimateFactory.Build(level);
 
-	for (size_t k=0;k<nodes.size();++k) {
+    RCP<MultiVectorDouble> aggQualities = level.Get<RCP<MultiVectorDouble>>("AggregateQualities", &aggQualityEstimateFactory);
 
-	  if (nodes[k]%nx == 0 || nodes[k]%nx == nx-1 || nodes[k]/nx == 0 || nodes[k]/nx == nx-1) {
-	    onBoundary = true;
-	    break;
-	  }
-			
-	}
+    ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
 
-	if (onBoundary) continue;
+    ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
+    AggregateQualityEstimateFactory::ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
 
-	bool assert_performed = false;
+    for (size_t j = 0; j < aggQualities->getLocalLength(); ++j) {
+      std::vector<LO> nodes;
+      for (LO k = 0; k < aggSizes[j]; ++k) {
+        nodes.push_back(aggSortedVertices[aggsToIndices[j] + k]);
+      }
 
-	for (size_t agg_id=0;agg_id<AGG_TYPES.size();++agg_id) {
+      std::sort(nodes.begin(), nodes.end());
 
-	  if (AGG_TYPES[agg_id].size() != nodes.size()) continue;
+      bool onBoundary = false;
 
-	  const std::vector<LO>& unflipped_agg = AGG_TYPES[agg_id];
+      for (size_t k = 0; k < nodes.size(); ++k) {
+        if (nodes[k] % nx == 0 || nodes[k] % nx == nx - 1 || nodes[k] / nx == 0 || nodes[k] / nx == nx - 1) {
+          onBoundary = true;
+          break;
+        }
+      }
 
-	  for (int flip_id=0;flip_id<4;++flip_id) {
+      if (onBoundary) continue;
 
-	    std::vector<LO> flipped_agg;
+      bool assert_performed = false;
 
-	    switch(flip_id) {
-	    case 0:
-	      flipped_agg = unflipped_agg;
-	      break;
-	    case 1:
-	      flipped_agg = flip_agg_horizontal(unflipped_agg, nx);
-	      break;
-	    case 2:
-	      flipped_agg = flip_agg_vertical(unflipped_agg, nx);
-	      break;
-	    case 3:
-	      flipped_agg = flip_agg_horizontal(flip_agg_vertical(unflipped_agg,nx),nx);
-	      break;
-	    }
+      for (size_t agg_id = 0; agg_id < AGG_TYPES.size(); ++agg_id) {
+        if (AGG_TYPES[agg_id].size() != nodes.size()) continue;
 
-	    LO difference = nodes[0] - flipped_agg[0];
+        const std::vector<LO>& unflipped_agg = AGG_TYPES[agg_id];
 
-	    bool aggFound = true;
+        for (int flip_id = 0; flip_id < 4; ++flip_id) {
+          std::vector<LO> flipped_agg;
 
-	    for (size_t k=1;k<nodes.size();++k) {
-	      if (difference != nodes[k] - flipped_agg[k]) {
-		aggFound = false;
-		break;
-	      }
+          switch (flip_id) {
+            case 0:
+              flipped_agg = unflipped_agg;
+              break;
+            case 1:
+              flipped_agg = flip_agg_horizontal(unflipped_agg, nx);
+              break;
+            case 2:
+              flipped_agg = flip_agg_vertical(unflipped_agg, nx);
+              break;
+            case 3:
+              flipped_agg = flip_agg_horizontal(flip_agg_vertical(unflipped_agg, nx), nx);
+              break;
+          }
 
-	      if (nodes[k]%nx == 0 || nodes[k]%nx == nx-1 || nodes[k]/nx == 0 || nodes[k]/nx == nx-1) {
-		aggFound = false;
-		break;
-	      }
-			
-	    }
+          LO difference = nodes[0] - flipped_agg[0];
 
-	    if (!aggFound) continue;
+          bool aggFound = true;
 
-	    TEST_FLOATING_EQUALITY(aggQualitiesLocalData[j], AGG_QUALITIES[agg_id], 1e-3);
+          for (size_t k = 1; k < nodes.size(); ++k) {
+            if (difference != nodes[k] - flipped_agg[k]) {
+              aggFound = false;
+              break;
+            }
 
-	    assert_performed = true;
-	    break;
+            if (nodes[k] % nx == 0 || nodes[k] % nx == nx - 1 || nodes[k] / nx == 0 || nodes[k] / nx == nx - 1) {
+              aggFound = false;
+              break;
+            }
+          }
 
-	  }
+          if (!aggFound) continue;
 
-	  if (assert_performed) break;
+          TEST_FLOATING_EQUALITY(aggQualitiesLocalData[j], AGG_QUALITIES[agg_id], 1e-3);
 
-	}
+          assert_performed = true;
+          break;
+        }
 
+        if (assert_performed) break;
       }
-
     }
+  }
 
-  } // Anisotropic Diffusion 2D test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, ConvectionDiffusion2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Xpetra::MultiVector<MT,LO,GO,Node> MultiVectorDouble;
-    typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    // Don't test for complex - matrix reader won't work
-    if (TST::isComplex) {success=true; return;}
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    LO nx = (LO) 20;
-
-    LO num_nodes = (LO) nx*nx;
-    RCP<Map> map_for_read = MapFactory::Build(TestHelpers::Parameters::getLib(),num_nodes,0,comm);
-    RCP<Matrix> A =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/cd2dx.mat", map_for_read);
-
-    std::vector<std::string> test_matrices = { "TestMatrices/cd2dx.mat", "TestMatrices/cd2dy.mat", "TestMatrices/aniso2dx.mat", "TestMatrices/aniso2dy.mat", "TestMatrices/iso2d.mat" };
-
-    const std::vector<LO> AGG0_NODES = {0,1};
-    const std::vector<LO> AGG1_NODES = {0,nx};
-    const std::vector<LO> AGG2_NODES = {0,nx,nx+1};
-    const std::vector<LO> AGG3_NODES = {0,1,nx,nx+1};
-    const std::vector<LO> AGG4_NODES = {0,1,2};
-    const std::vector<LO> AGG5_NODES = {0,nx,2*nx};
-    const std::vector<LO> AGG6_NODES = {0,nx,nx+1,2*nx};
-    const std::vector<LO> AGG7_NODES = {1,nx,nx+1,nx+2};
-    const std::vector<LO> AGG8_NODES = {1,nx,nx+1,nx+2,2*nx+1};
-    const std::vector<LO> AGG9_NODES = {0,1,2,nx,nx+1,nx+2};
-    const std::vector<LO> AGG10_NODES = {0,1,nx,nx+1,2*nx,2*nx+1};
-    const std::vector<LO> AGG11_NODES = {0,1,2,nx,nx+1,nx+2,2*nx,2*nx+1,2*nx+2};
-    const std::vector<LO> AGG12_NODES = {0,1,nx,nx+1,nx+2};
-    const std::vector<LO> AGG13_NODES = {0,1,nx,nx+1,2*nx};
-    const std::vector<LO> AGG14_NODES = {0,1,2,nx,nx+1,nx+2,2*nx};
-    const std::vector<LO> AGG15_NODES = {0,1,2,nx,nx+1,nx+2,2*nx+1};
-    const std::vector<LO> AGG16_NODES = {0,1,2,nx,nx+1,2*nx,2*nx+1};
-    const std::vector<LO> AGG17_NODES = {0,1,nx,nx+1,nx+2,2*nx,2*nx+1};
-    const std::vector<LO> AGG18_NODES = {1,nx,nx+1,nx+2,2*nx+1,2*nx+2};
-
-    const std::vector<std::vector<LO>> AGG_TYPES = {AGG0_NODES, AGG1_NODES, AGG2_NODES,
-						    AGG3_NODES, AGG4_NODES, AGG5_NODES,
-						    AGG6_NODES, AGG7_NODES, AGG8_NODES,
-						    AGG9_NODES, AGG10_NODES, AGG11_NODES,
-						    AGG12_NODES, AGG13_NODES, AGG14_NODES,
-						    AGG15_NODES, AGG16_NODES, AGG17_NODES,
-						    AGG18_NODES};
-
-    const std::vector<MT> AGG_QUALITIES = {2.008365, 480.190476, 640.925507,
-					       480.190465, 4.016730, 960.380952,
-					       960.380906, 720.956363, 960.380860,
-					       480.190453, 960.380906, 960.380860,
-					       576.902089, 961.050518, 962.055240,
-					       961.050432, 1172.649129, 960.380860,
-					       961.050785};
-
-    for (size_t i=0;i<test_matrices.size();i++) {
-
-      out << "Aggregating Matrix " << i << "..." << std::endl;
+}  // Anisotropic Diffusion 2D test
 
-      Level level;
-      TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AggregateQualityEstimateFactory, ConvectionDiffusion2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Xpetra::MultiVector<MT, LO, GO, Node> MultiVectorDouble;
+  typedef MueLu::AggregateQualityEstimateFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> AggregateQualityEstimateFactory;
 
-      RCP<Matrix> A_agg =  Xpetra::IO<SC, LO, GO, NO>::Read(test_matrices.at(i), map_for_read);
+  out << "version: " << MueLu::Version() << std::endl;
 
-      level.Set("A",A_agg);
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
 
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory);
-      dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.1));
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  // Don't test for complex - matrix reader won't work
+  if (TST::isComplex) {
+    success = true;
+    return;
+  }
 
-      RCP<UncoupledAggregationFactory> aggFactory = rcp(new UncoupledAggregationFactory);
-      aggFactory->SetFactory("Graph", dropFact);
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
+
+  LO nx = (LO)20;
+
+  LO num_nodes          = (LO)nx * nx;
+  RCP<Map> map_for_read = MapFactory::Build(TestHelpers::Parameters::getLib(), num_nodes, 0, comm);
+  RCP<Matrix> A         = Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/cd2dx.mat", map_for_read);
+
+  std::vector<std::string> test_matrices = {"TestMatrices/cd2dx.mat", "TestMatrices/cd2dy.mat", "TestMatrices/aniso2dx.mat", "TestMatrices/aniso2dy.mat", "TestMatrices/iso2d.mat"};
+
+  const std::vector<LO> AGG0_NODES  = {0, 1};
+  const std::vector<LO> AGG1_NODES  = {0, nx};
+  const std::vector<LO> AGG2_NODES  = {0, nx, nx + 1};
+  const std::vector<LO> AGG3_NODES  = {0, 1, nx, nx + 1};
+  const std::vector<LO> AGG4_NODES  = {0, 1, 2};
+  const std::vector<LO> AGG5_NODES  = {0, nx, 2 * nx};
+  const std::vector<LO> AGG6_NODES  = {0, nx, nx + 1, 2 * nx};
+  const std::vector<LO> AGG7_NODES  = {1, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG8_NODES  = {1, nx, nx + 1, nx + 2, 2 * nx + 1};
+  const std::vector<LO> AGG9_NODES  = {0, 1, 2, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG10_NODES = {0, 1, nx, nx + 1, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG11_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx, 2 * nx + 1, 2 * nx + 2};
+  const std::vector<LO> AGG12_NODES = {0, 1, nx, nx + 1, nx + 2};
+  const std::vector<LO> AGG13_NODES = {0, 1, nx, nx + 1, 2 * nx};
+  const std::vector<LO> AGG14_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx};
+  const std::vector<LO> AGG15_NODES = {0, 1, 2, nx, nx + 1, nx + 2, 2 * nx + 1};
+  const std::vector<LO> AGG16_NODES = {0, 1, 2, nx, nx + 1, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG17_NODES = {0, 1, nx, nx + 1, nx + 2, 2 * nx, 2 * nx + 1};
+  const std::vector<LO> AGG18_NODES = {1, nx, nx + 1, nx + 2, 2 * nx + 1, 2 * nx + 2};
+
+  const std::vector<std::vector<LO>> AGG_TYPES = {AGG0_NODES, AGG1_NODES, AGG2_NODES,
+                                                  AGG3_NODES, AGG4_NODES, AGG5_NODES,
+                                                  AGG6_NODES, AGG7_NODES, AGG8_NODES,
+                                                  AGG9_NODES, AGG10_NODES, AGG11_NODES,
+                                                  AGG12_NODES, AGG13_NODES, AGG14_NODES,
+                                                  AGG15_NODES, AGG16_NODES, AGG17_NODES,
+                                                  AGG18_NODES};
+
+  const std::vector<MT> AGG_QUALITIES = {2.008365, 480.190476, 640.925507,
+                                         480.190465, 4.016730, 960.380952,
+                                         960.380906, 720.956363, 960.380860,
+                                         480.190453, 960.380906, 960.380860,
+                                         576.902089, 961.050518, 962.055240,
+                                         961.050432, 1172.649129, 960.380860,
+                                         961.050785};
+
+  for (size_t i = 0; i < test_matrices.size(); i++) {
+    out << "Aggregating Matrix " << i << "..." << std::endl;
 
-      out << "Coalesce Drop: " << dropFact.get() << std::endl;
-      out << "Aggregation: " << aggFactory.get() << std::endl;
+    Level level;
+    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-      level.Request("Aggregates", aggFactory.get());
+    RCP<Matrix> A_agg = Xpetra::IO<SC, LO, GO, NO>::Read(test_matrices.at(i), map_for_read);
 
-      out << "Building aggregates..." << std::endl;
-	
-      aggFactory->Build(level);
+    level.Set("A", A_agg);
 
-      out << "Built aggregates! Getting them...\n" << std::endl;
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory);
+    dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.1));
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-      RCP<Aggregates> aggs = level.Get< RCP<Aggregates> >("Aggregates", aggFactory.get());
+    RCP<UncoupledAggregationFactory> aggFactory = rcp(new UncoupledAggregationFactory);
+    aggFactory->SetFactory("Graph", dropFact);
 
-      out << "Got aggregates! Computing quality estimate...\n" << std::endl;
+    out << "Coalesce Drop: " << dropFact.get() << std::endl;
+    out << "Aggregation: " << aggFactory.get() << std::endl;
 
-      level.print(out, Teuchos::VERB_EXTREME);
+    level.Request("Aggregates", aggFactory.get());
 
-      level.Set("A",A);
+    out << "Building aggregates..." << std::endl;
 
-      RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory);
-      coarseMapFact->SetFactory("Aggregates", aggFactory);
+    aggFactory->Build(level);
 
-      AggregateQualityEstimateFactory aggQualityEstimateFactory;
-      aggQualityEstimateFactory.SetFactory("Aggregates", aggFactory);
-      aggQualityEstimateFactory.SetFactory("CoarseMap", coarseMapFact);
+    out << "Built aggregates! Getting them...\n"
+        << std::endl;
 
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(true));
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
-      aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
+    RCP<Aggregates> aggs = level.Get<RCP<Aggregates>>("Aggregates", aggFactory.get());
 
-      level.Request("AggregateQualities", &aggQualityEstimateFactory);
-      // level.Request(aggQualityEstimateFactory);
+    out << "Got aggregates! Computing quality estimate...\n"
+        << std::endl;
 
-      aggQualityEstimateFactory.Build(level);	
+    level.print(out, Teuchos::VERB_EXTREME);
 
-      RCP<MultiVectorDouble> aggQualities = level.Get< RCP<MultiVectorDouble> >("AggregateQualities", &aggQualityEstimateFactory);
-    
-      ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
+    level.Set("A", A);
 
-      ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
-      AggregateQualityEstimateFactory::ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
+    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory);
+    coarseMapFact->SetFactory("Aggregates", aggFactory);
 
-      for (size_t j=0;j<aggQualities->getLocalLength();++j) {
+    AggregateQualityEstimateFactory aggQualityEstimateFactory;
+    aggQualityEstimateFactory.SetFactory("Aggregates", aggFactory);
+    aggQualityEstimateFactory.SetFactory("CoarseMap", coarseMapFact);
 
-	std::vector<LO> nodes;
-	for (LO k=0;k<aggSizes[j];++k) {
-	  nodes.push_back(aggSortedVertices[aggsToIndices[j]+k]);
-	}
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: check symmetry", Teuchos::ParameterEntry(true));
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: good aggregate threshold", Teuchos::ParameterEntry(100.0));
+    aggQualityEstimateFactory.SetParameter("aggregate qualities: file output", Teuchos::ParameterEntry(false));
 
-	std::sort(nodes.begin(), nodes.end());
+    level.Request("AggregateQualities", &aggQualityEstimateFactory);
+    // level.Request(aggQualityEstimateFactory);
 
-	bool onBoundary = false;
+    aggQualityEstimateFactory.Build(level);
 
-	for (size_t k=0;k<nodes.size();++k) {
+    RCP<MultiVectorDouble> aggQualities = level.Get<RCP<MultiVectorDouble>>("AggregateQualities", &aggQualityEstimateFactory);
 
-	  if (nodes[k]%nx == 0 || nodes[k]%nx == nx-1 || nodes[k]/nx == 0 || nodes[k]/nx == nx-1) {
-	    onBoundary = true;
-	    break;
-	  }
-			
-	}
+    ArrayRCP<const MT> aggQualitiesLocalData = aggQualities->getData(0);
 
-	if (onBoundary) continue;
+    ArrayRCP<LO> aggSortedVertices, aggsToIndices, aggSizes;
+    AggregateQualityEstimateFactory::ConvertAggregatesData(aggs, aggSortedVertices, aggsToIndices, aggSizes);
 
-	bool assert_performed = false;
+    for (size_t j = 0; j < aggQualities->getLocalLength(); ++j) {
+      std::vector<LO> nodes;
+      for (LO k = 0; k < aggSizes[j]; ++k) {
+        nodes.push_back(aggSortedVertices[aggsToIndices[j] + k]);
+      }
 
-	for (size_t agg_id=0;agg_id<AGG_TYPES.size();++agg_id) {
+      std::sort(nodes.begin(), nodes.end());
 
-	  if (AGG_TYPES[agg_id].size() != nodes.size()) continue;
+      bool onBoundary = false;
 
-	  const std::vector<LO>& unflipped_agg = AGG_TYPES[agg_id];
+      for (size_t k = 0; k < nodes.size(); ++k) {
+        if (nodes[k] % nx == 0 || nodes[k] % nx == nx - 1 || nodes[k] / nx == 0 || nodes[k] / nx == nx - 1) {
+          onBoundary = true;
+          break;
+        }
+      }
 
-	  for (int flip_id=0;flip_id<4;++flip_id) {
+      if (onBoundary) continue;
 
-	    std::vector<LO> flipped_agg;
+      bool assert_performed = false;
 
-	    switch(flip_id) {
-	    case 0:
-	      flipped_agg = unflipped_agg;
-	      break;
-	    case 1:
-	      flipped_agg = flip_agg_horizontal(unflipped_agg, nx);
-	      break;
-	    case 2:
-	      flipped_agg = flip_agg_vertical(unflipped_agg, nx);
-	      break;
-	    case 3:
-	      flipped_agg = flip_agg_horizontal(flip_agg_vertical(unflipped_agg,nx),nx);
-	      break;
-	    }
+      for (size_t agg_id = 0; agg_id < AGG_TYPES.size(); ++agg_id) {
+        if (AGG_TYPES[agg_id].size() != nodes.size()) continue;
 
-	    LO difference = nodes[0] - flipped_agg[0];
+        const std::vector<LO>& unflipped_agg = AGG_TYPES[agg_id];
 
-	    bool aggFound = true;
+        for (int flip_id = 0; flip_id < 4; ++flip_id) {
+          std::vector<LO> flipped_agg;
 
-	    for (size_t k=1;k<nodes.size();++k) {
-	      if (difference != nodes[k] - flipped_agg[k]) {
-		aggFound = false;
-		break;
-	      }
+          switch (flip_id) {
+            case 0:
+              flipped_agg = unflipped_agg;
+              break;
+            case 1:
+              flipped_agg = flip_agg_horizontal(unflipped_agg, nx);
+              break;
+            case 2:
+              flipped_agg = flip_agg_vertical(unflipped_agg, nx);
+              break;
+            case 3:
+              flipped_agg = flip_agg_horizontal(flip_agg_vertical(unflipped_agg, nx), nx);
+              break;
+          }
 
-	      if (nodes[k]%nx == 0 || nodes[k]%nx == nx-1 || nodes[k]/nx == 0 || nodes[k]/nx == nx-1) {
-		aggFound = false;
-		break;
-	      }
-			
-	    }
+          LO difference = nodes[0] - flipped_agg[0];
 
-	    if (!aggFound) continue;
+          bool aggFound = true;
 
-	    TEST_FLOATING_EQUALITY(aggQualitiesLocalData[j], AGG_QUALITIES[agg_id], 1e-3);
+          for (size_t k = 1; k < nodes.size(); ++k) {
+            if (difference != nodes[k] - flipped_agg[k]) {
+              aggFound = false;
+              break;
+            }
 
-	    assert_performed = true;
-	    break;
+            if (nodes[k] % nx == 0 || nodes[k] % nx == nx - 1 || nodes[k] / nx == 0 || nodes[k] / nx == nx - 1) {
+              aggFound = false;
+              break;
+            }
+          }
 
-	  }
+          if (!aggFound) continue;
 
-	  if (assert_performed) break;
+          TEST_FLOATING_EQUALITY(aggQualitiesLocalData[j], AGG_QUALITIES[agg_id], 1e-3);
 
-	}
+          assert_performed = true;
+          break;
+        }
 
+        if (assert_performed) break;
       }
-
     }
+  }
 
-  } // Convection Diffusion 2D test
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)				\
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,Constructor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,Poisson2D,Scalar,LO,GO,Node)
-  //  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,AnisotropicDiffusion2D,Scalar,LO,GO,Node)
+}  // Convection Diffusion 2D test
 
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory, Constructor, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory, Poisson2D, Scalar, LO, GO, Node)
+//  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,AnisotropicDiffusion2D,Scalar,LO,GO,Node)
 
-  //  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,ConvectionDiffusion2D,Scalar,LO,GO,Node)
+//  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AggregateQualityEstimateFactory,ConvectionDiffusion2D,Scalar,LO,GO,Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Aggregates.cpp b/packages/muelu/test/unit_tests/Aggregates.cpp
index 4c3b11698173..c6fc5091e72c 100644
--- a/packages/muelu/test/unit_tests/Aggregates.cpp
+++ b/packages/muelu/test/unit_tests/Aggregates.cpp
@@ -62,1382 +62,1337 @@
 
 namespace MueLuTests {
 
-  template<class T>
-  void print_matrix(const char * label,const Teuchos::RCP<T> &Mat) {
-    ArrayRCP<const size_t> rowptr;
-    ArrayRCP<const typename T::local_ordinal_type> colind;
-    ArrayRCP<const typename T::scalar_type> vals;
-    Mat->getAllValues(rowptr,colind,vals);
-
-    printf("*** %s ***\n",label);
-    printf("rowptr = ");
-    for(int i=0; i<(int)rowptr.size();i++)
-      printf("%d ",(int)rowptr[i]);
-    printf("\ncolind =");
-    for(int i=0; i<(int)colind.size();i++)
-      printf("%d ",(int)colind[i]);
-    printf("\nvalues =");
-    for(int i=0; i<(int)vals.size();i++)
-      printf("%8.1e ",vals[i]);
-    printf("\n");
-  }
-
+template <class T>
+void print_matrix(const char* label, const Teuchos::RCP<T>& Mat) {
+  ArrayRCP<const size_t> rowptr;
+  ArrayRCP<const typename T::local_ordinal_type> colind;
+  ArrayRCP<const typename T::scalar_type> vals;
+  Mat->getAllValues(rowptr, colind, vals);
+
+  printf("*** %s ***\n", label);
+  printf("rowptr = ");
+  for (int i = 0; i < (int)rowptr.size(); i++)
+    printf("%d ", (int)rowptr[i]);
+  printf("\ncolind =");
+  for (int i = 0; i < (int)colind.size(); i++)
+    printf("%d ", (int)colind[i]);
+  printf("\nvalues =");
+  for (int i = 0; i < (int)vals.size(); i++)
+    printf("%8.1e ", vals[i]);
+  printf("\n");
+}
 
-template
-<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
 class AggregateGenerator {
-  //typedef MueLu::AmalgamationInfo<LocalOrdinal,GlobalOrdinal,Node> amalgamation_info_type;
-  //typedef MueLu::Aggregates<LocalOrdinal,GlobalOrdinal,Node> aggregates_type;
-  //typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> xpetra_matrix_type;
-#   include <MueLu_UseShortNames.hpp>
-
-public:
-
-    // Little utility to generate uncoupled aggregates.
-    static RCP<Aggregates>
-    gimmeUncoupledAggregates(const RCP<Matrix> & A, RCP<AmalgamationInfo> & amalgInfo, bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true)
+  // typedef MueLu::AmalgamationInfo<LocalOrdinal,GlobalOrdinal,Node> amalgamation_info_type;
+  // typedef MueLu::Aggregates<LocalOrdinal,GlobalOrdinal,Node> aggregates_type;
+  // typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> xpetra_matrix_type;
+#include <MueLu_UseShortNames.hpp>
+
+ public:
+  // Little utility to generate uncoupled aggregates.
+  static RCP<Aggregates>
+  gimmeUncoupledAggregates(const RCP<Matrix>& A, RCP<AmalgamationInfo>& amalgInfo, bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true)
   //  RCP<MueLu::Aggregates> gimmeUncoupledAggregates(const RCP<xpetra_matrix_type> & A, RCP<AmalgamationInfo> & amalgInfo, bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true)
-    {
-      Level level;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      // Setup aggregation factory (use default factory for graph)
-      RCP<UncoupledAggregationFactory> aggFact  = rcp(new UncoupledAggregationFactory());
-      aggFact->SetFactory("Graph", dropFact);
-      aggFact->SetParameter("aggregation: max agg size",Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: min agg size",Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: max selected neighbors",Teuchos::ParameterEntry(0));
-      aggFact->SetParameter("aggregation: ordering",Teuchos::ParameterEntry(std::string("natural")));
-      aggFact->SetParameter("aggregation: allow user-specified singletons", Teuchos::ParameterEntry(true));
-
-      aggFact->SetParameter("aggregation: enable phase 1",  Teuchos::ParameterEntry(bPhase1));
-      aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(bPhase2a));
-      aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(bPhase2b));
-      aggFact->SetParameter("aggregation: enable phase 3",  Teuchos::ParameterEntry(bPhase3));
-      aggFact->SetParameter("aggregation: phase3 avoid singletons",  Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: use interface aggregation",Teuchos::ParameterEntry(false));
-      level.Request("Aggregates", aggFact.get());
-      level.Request("UnAmalgamationInfo", amalgFact.get());
-
-      level.Request(*aggFact);
-      aggFact->Build(level);
-      RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-      amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo",amalgFact.get()); // fix me
-      level.Release("UnAmalgamationInfo", amalgFact.get());
-      level.Release("Aggregates", aggFact.get());
-      return aggregates;
-    }  // gimmeUncoupledAggregates
-
-
-    // Little utility to generate uncoupled aggregates.
-    static RCP<Aggregates>
-    gimmeStructuredAggregates(const RCP<Matrix>& A,
-                              Array<GO> gNodesPerDir, Array<LO> lNodesPerDir, const bool coupled,
-                              const LO numDimensions, const std::string meshLayout,
-                              const Array<GO> meshData)
-    {
-      Level level;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-
-      const std::string coupling = (coupled ? "coupled" : "uncoupled");
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      amalgFact->SetDefaultVerbLevel(MueLu::None);
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      level.Set("A", A);
-      level.Set("numDimensions", numDimensions);
-      level.Set("gNodesPerDim",  gNodesPerDir);
-      level.Set("lNodesPerDim",  lNodesPerDir);
-      level.Set("aggregation: mesh data", meshData);
-
-      // Setup aggregation factory (use default factory for graph)
-      RCP<StructuredAggregationFactory> aggFact = rcp(new StructuredAggregationFactory());
-      aggFact->SetFactory("Graph", dropFact);
-      aggFact->SetParameter("aggregation: mode", Teuchos::ParameterEntry(coupling));
-      aggFact->SetParameter("aggregation: mesh layout", Teuchos::ParameterEntry(meshLayout));
-      aggFact->SetParameter("aggregation: coarsening order", Teuchos::ParameterEntry(0));
-      aggFact->SetParameter("aggregation: coarsening rate",
-                            Teuchos::ParameterEntry(std::string("{3}")));
-
-      level.Request("Aggregates", aggFact.get());
-
-      level.Request(*aggFact);
-      aggFact->Build(level);
-
-      RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-      level.Release("Aggregates", aggFact.get());
-
-      return aggregates;
-    }  // gimmeStructuredAggregates
-
-    // Little utility to generate uncoupled aggregates with some specified root nodes on interface.
-    static RCP<Aggregates>
-    gimmeInterfaceAggregates(const RCP<Matrix>&A, RCP<AmalgamationInfo> & amalgInfo, Teuchos::Array<LO>& nodeOnInterface)
-    {
-      Level level;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      // Setup aggregation factory (use default factory for graph)
-      RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-      aggFact->SetFactory("Graph", dropFact);
-      aggFact->SetParameter("aggregation: use interface aggregation",Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: max agg size",Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: min agg size",Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: max selected neighbors",Teuchos::ParameterEntry(0));
-      aggFact->SetParameter("aggregation: ordering",Teuchos::ParameterEntry(std::string("natural")));
-      aggFact->SetParameter("aggregation: enable phase 1",  Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 3",  Teuchos::ParameterEntry(true));
-
-      level.Set("nodeOnInterface",nodeOnInterface);
-
-
-
-      level.Request("Aggregates", aggFact.get());
-      level.Request("UnAmalgamationInfo", amalgFact.get());
-
-      level.Request(*aggFact);
-      aggFact->Build(level);
-
-      RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-      level.Release("Aggregates", aggFact.get());
-      return aggregates;
-    }  // gimmeInterfaceAggregates
-
-    // Little utility to generate hybrid structured and uncoupled aggregates
-    static RCP<Aggregates>
-    gimmeHybridAggregates(const RCP<Matrix>&A, RCP<AmalgamationInfo> & amalgInfo,
-                          const std::string regionType,
-                          Array<GO> gNodesPerDir, Array<LO> lNodesPerDir,
-                          const LO numDimensions)
-    {
-      Level level;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.Set("numDimensions", numDimensions);
-      level.Set("lNodesPerDim", lNodesPerDir);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      amalgFact->SetDefaultVerbLevel(MueLu::None);
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      // Setup aggregation factory (use default factory for graph)
-      RCP<HybridAggregationFactory> aggFact = rcp(new HybridAggregationFactory());
-      aggFact->SetFactory("Graph", dropFact);
-      // Structured
-      aggFact->SetParameter("aggregation: coarsening order",             Teuchos::ParameterEntry(0));
-      aggFact->SetParameter("aggregation: coarsening rate",              Teuchos::ParameterEntry(std::string("{3}")));
-      // Uncoupled
-      aggFact->SetParameter("aggregation: use interface aggregation",    Teuchos::ParameterEntry(false));
-      aggFact->SetParameter("aggregation: max agg size",                 Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: min agg size",                 Teuchos::ParameterEntry(3));
-      aggFact->SetParameter("aggregation: max selected neighbors",       Teuchos::ParameterEntry(0));
-      aggFact->SetParameter("aggregation: ordering",                     Teuchos::ParameterEntry(std::string("natural")));
-      aggFact->SetParameter("aggregation: enable phase 1",               Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 2a",              Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 2b",              Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: enable phase 3",               Teuchos::ParameterEntry(true));
-
-      aggFact->SetParameter("aggregation: match ML phase2a",             Teuchos::ParameterEntry(true));
-
-      // Hybrid
-      level.Set("aggregationRegionType", regionType);
-
-
-      level.Request("Aggregates", aggFact.get());
-      level.Request("UnAmalgamationInfo", amalgFact.get());
-
-      level.Request(*aggFact);
-      aggFact->Build(level);
-
-      RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-      level.Release("Aggregates", aggFact.get());
-      return aggregates;
-    }  // gimmeHybridAggregates
-};
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo);
-    TEST_EQUALITY(aggregates != Teuchos::null, true);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-  }
-
-  ///////////////////////////////////////////////////////////////////////////
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GetNumAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<const Map> rowmap = A->getRowMap();
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo);
-    GO numAggs = aggregates->GetNumAggregates();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<GO> aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-    for (LO i = 0; i < numAggs; ++i)
-      aggSizes[i] = aggStart[i+1] - aggStart[i];
-
-    bool foundAggNotSize3=false;
-    for (int i=0; i<aggSizes.size(); ++i)
-      if (aggSizes[i] != 3) {
-        foundAggNotSize3=true;
-        break;
-      }
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    switch (comm->getSize()) {
-
-      case 1 :
-        TEST_EQUALITY(numAggs, 12);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 2:
-        TEST_EQUALITY(numAggs, 6);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 3:
-        TEST_EQUALITY(numAggs, 4);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 4:
-        TEST_EQUALITY(numAggs, 3);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      default:
-        std::string msg = "Only 1-4 MPI processes are supported.";
-        //throw(MueLu::Exceptions::NotImplemented(msg));
-        out << msg << std::endl;
-        break;
-    }
+    // Setup aggregation factory (use default factory for graph)
+    RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+    aggFact->SetFactory("Graph", dropFact);
+    aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+    aggFact->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("natural")));
+    aggFact->SetParameter("aggregation: allow user-specified singletons", Teuchos::ParameterEntry(true));
+
+    aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(bPhase1));
+    aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(bPhase2a));
+    aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(bPhase2b));
+    aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(bPhase3));
+    aggFact->SetParameter("aggregation: phase3 avoid singletons", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: use interface aggregation", Teuchos::ParameterEntry(false));
+    level.Request("Aggregates", aggFact.get());
+    level.Request("UnAmalgamationInfo", amalgFact.get());
 
-    //ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
-    int root = out.getOutputToRootOnly();
-    out.setOutputToRootOnly(-1);
-    for (int j=0; j<comm->getSize(); ++j) {
-      if (comm->getRank() == j) {
-        out << "++ pid " << j << " ++" << std::endl;
-        out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
-        for (int i=0; i< numAggs; ++i) {
-          out << "   aggregate " << i << ": ";
-          for (int k=aggStart[i]; k< aggStart[i+1]; ++k)
-            out << aggToRowMap[k] << " ";
-          out << std::endl;
-        }
-      }
-      comm->barrier();
-    }
-    out.setOutputToRootOnly(root);
+    level.Request(*aggFact);
+    aggFact->Build(level);
+    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());                  // fix me
+    amalgInfo                  = level.Get<RCP<AmalgamationInfo>>("UnAmalgamationInfo", amalgFact.get());  // fix me
+    level.Release("UnAmalgamationInfo", amalgFact.get());
+    level.Release("Aggregates", aggFact.get());
+    return aggregates;
+  }  // gimmeUncoupledAggregates
+
+  // Little utility to generate uncoupled aggregates.
+  static RCP<Aggregates>
+  gimmeStructuredAggregates(const RCP<Matrix>& A,
+                            Array<GO> gNodesPerDir, Array<LO> lNodesPerDir, const bool coupled,
+                            const LO numDimensions, const std::string meshLayout,
+                            const Array<GO> meshData) {
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-  } //GetNumAggregates
+    const std::string coupling = (coupled ? "coupled" : "uncoupled");
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo);
-    TEST_EQUALITY(aggregates != Teuchos::null, true);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-  }
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    amalgFact->SetDefaultVerbLevel(MueLu::None);
+    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustUncoupledAggregationFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+    level.Set("A", A);
+    level.Set("numDimensions", numDimensions);
+    level.Set("gNodesPerDim", gNodesPerDir);
+    level.Set("lNodesPerDim", lNodesPerDir);
+    level.Set("aggregation: mesh data", meshData);
 
     // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory);
-    aggFact->SetOrdering("graph");
-    TEST_EQUALITY(aggFact->GetOrdering()== "graph",             true);
-    aggFact->SetOrdering("natural");
-    TEST_EQUALITY(aggFact->GetOrdering()== "natural",           true);
-    aggFact->SetOrdering("random");
-    TEST_EQUALITY(aggFact->GetOrdering()== "random",            true);
+    RCP<StructuredAggregationFactory> aggFact = rcp(new StructuredAggregationFactory());
+    aggFact->SetFactory("Graph", dropFact);
+    aggFact->SetParameter("aggregation: mode", Teuchos::ParameterEntry(coupling));
+    aggFact->SetParameter("aggregation: mesh layout", Teuchos::ParameterEntry(meshLayout));
+    aggFact->SetParameter("aggregation: coarsening order", Teuchos::ParameterEntry(0));
+    aggFact->SetParameter("aggregation: coarsening rate",
+                          Teuchos::ParameterEntry(std::string("{3}")));
 
-    aggFact->SetMaxNeighAlreadySelected(12);
-    TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 12,   true);
+    level.Request("Aggregates", aggFact.get());
 
-    aggFact->SetMaxNeighAlreadySelected(0);
-    TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 0,   true);
+    level.Request(*aggFact);
+    aggFact->Build(level);
 
-    aggFact->SetMinNodesPerAggregate(0);
-    TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 0,      true);
+    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());  // fix me
+    level.Release("Aggregates", aggFact.get());
 
-    aggFact->SetMinNodesPerAggregate(3);
-    TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 3,      true);
+    return aggregates;
+  }  // gimmeStructuredAggregates
 
+  // Little utility to generate uncoupled aggregates with some specified root nodes on interface.
+  static RCP<Aggregates>
+  gimmeInterfaceAggregates(const RCP<Matrix>& A, RCP<AmalgamationInfo>& amalgInfo, Teuchos::Array<LO>& nodeOnInterface) {
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-}
-  ///////////////////////////////////////////////////////////////////////////
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GetNumUncoupledAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<const Map> rowmap = A->getRowMap();
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo);
-    GO numAggs = aggregates->GetNumAggregates();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<GO> aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-    for (LO i = 0; i < numAggs; ++i)
-      aggSizes[i] = aggStart[i+1] - aggStart[i];
-
-    bool foundAggNotSize3=false;
-    for (int i=0; i<aggSizes.size(); ++i)
-      if (aggSizes[i] != 3) {
-        foundAggNotSize3=true;
-        break;
-      }
+    // Setup aggregation factory (use default factory for graph)
+    RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+    aggFact->SetFactory("Graph", dropFact);
+    aggFact->SetParameter("aggregation: use interface aggregation", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+    aggFact->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("natural")));
+    aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+    level.Set("nodeOnInterface", nodeOnInterface);
 
-    switch (comm->getSize()) {
-
-      case 1 :
-        TEST_EQUALITY(numAggs, 12);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 2:
-        TEST_EQUALITY(numAggs, 6);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 3:
-        TEST_EQUALITY(numAggs, 4);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 4:
-        TEST_EQUALITY(numAggs, 3);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      default:
-        std::string msg = "Only 1-4 MPI processes are supported.";
-        //throw(MueLu::Exceptions::NotImplemented(msg));
-        out << msg << std::endl;
-        break;
-    }
+    level.Request("Aggregates", aggFact.get());
+    level.Request("UnAmalgamationInfo", amalgFact.get());
 
-    //ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
-    int root = out.getOutputToRootOnly();
-    out.setOutputToRootOnly(-1);
-    for (int j=0; j<comm->getSize(); ++j) {
-      if (comm->getRank() == j) {
-        out << "++ pid " << j << " ++" << std::endl;
-        out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
-        for (int i=0; i< numAggs; ++i) {
-          out << "   aggregate " << i << ": ";
-          for (int k=aggStart[i]; k< aggStart[i+1]; ++k)
-            out << aggToRowMap[k] << " ";
-          out << std::endl;
-        }
-      }
-      comm->barrier();
-    }
-    out.setOutputToRootOnly(root);
+    level.Request(*aggFact);
+    aggFact->Build(level);
 
-  } //GetNumAggregates
+    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+    level.Release("Aggregates", aggFact.get());
+    return aggregates;
+  }  // gimmeInterfaceAggregates
+
+  // Little utility to generate hybrid structured and uncoupled aggregates
+  static RCP<Aggregates>
+  gimmeHybridAggregates(const RCP<Matrix>& A, RCP<AmalgamationInfo>& amalgInfo,
+                        const std::string regionType,
+                        Array<GO> gNodesPerDir, Array<LO> lNodesPerDir,
+                        const LO numDimensions) {
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.Set("numDimensions", numDimensions);
+    level.Set("lNodesPerDim", lNodesPerDir);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<const Map> rowmap = A->getRowMap();
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo,true,false,false,false);
-    GO numAggs = aggregates->GetNumAggregates();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<GO> aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-    for (LO i = 0; i < numAggs; ++i)
-      aggSizes[i] = aggStart[i+1] - aggStart[i];
-
-    bool foundAggNotSize3=false;
-    for (int i=0; i<aggSizes.size(); ++i)
-      if (aggSizes[i] != 3) {
-        foundAggNotSize3=true;
-        break;
-      }
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    amalgFact->SetDefaultVerbLevel(MueLu::None);
+    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    switch (comm->getSize()) {
-
-      case 1 :
-        TEST_EQUALITY(numAggs, 12);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 2:
-        TEST_EQUALITY(numAggs, 6);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 3:
-        TEST_EQUALITY(numAggs, 4);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      case 4:
-        TEST_EQUALITY(numAggs, 3);
-        TEST_EQUALITY(foundAggNotSize3, false);
-        break;
-
-      default:
-        std::string msg = "Only 1-4 MPI processes are supported.";
-        //throw(MueLu::Exceptions::NotImplemented(msg));
-        out << msg << std::endl;
-        break;
-    }
+    // Setup aggregation factory (use default factory for graph)
+    RCP<HybridAggregationFactory> aggFact = rcp(new HybridAggregationFactory());
+    aggFact->SetFactory("Graph", dropFact);
+    // Structured
+    aggFact->SetParameter("aggregation: coarsening order", Teuchos::ParameterEntry(0));
+    aggFact->SetParameter("aggregation: coarsening rate", Teuchos::ParameterEntry(std::string("{3}")));
+    // Uncoupled
+    aggFact->SetParameter("aggregation: use interface aggregation", Teuchos::ParameterEntry(false));
+    aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+    aggFact->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("natural")));
+    aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+    aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+    aggFact->SetParameter("aggregation: match ML phase2a", Teuchos::ParameterEntry(true));
+
+    // Hybrid
+    level.Set("aggregationRegionType", regionType);
 
-    //ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
-    int root = out.getOutputToRootOnly();
-    out.setOutputToRootOnly(-1);
-    for (int j=0; j<comm->getSize(); ++j) {
-      if (comm->getRank() == j) {
-        out << "++ pid " << j << " ++" << std::endl;
-        out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
-        for (int i=0; i< numAggs; ++i) {
-          out << "   aggregate " << i << ": ";
-          for (int k=aggStart[i]; k< aggStart[i+1]; ++k)
-            out << aggToRowMap[k] << " ";
-          out << std::endl;
-        }
-      }
-      comm->barrier();
-    }
-    out.setOutputToRootOnly(root);
+    level.Request("Aggregates", aggFact.get());
+    level.Request("UnAmalgamationInfo", amalgFact.get());
 
-  } //UncoupledPhase1
+    level.Request(*aggFact);
+    aggFact->Build(level);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<const Map> rowmap = A->getRowMap();
-    RCP<AmalgamationInfo> amalgInfo;
-    bool bSuccess = true;
-    TEUCHOS_TEST_THROW((AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo,false,true,true,false)),
-                        MueLu::Exceptions::RuntimeError, out, bSuccess);
-    TEST_EQUALITY(bSuccess, true);
-  } //UncoupledPhase2
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<const Map> rowmap = A->getRowMap();
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeUncoupledAggregates(A, amalgInfo,false,false,false,true); GO numAggs = aggregates->GetNumAggregates();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<GO> aggToRowMap;
-    amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-    for (LO i = 0; i < numAggs; ++i)
-      aggSizes[i] = aggStart[i+1] - aggStart[i];
-
-    // LBV on 09/27/19: the aggregate size of 2 is only
-    // expected because the problem is 1D and the nodes
-    // are treated in a lexicographic order. Using a
-    // different ordering will almost always result in
-    // at least on aggregate of size 3 and one of size 1.
-    bool foundAggNotSize2=false;
-    for (int i=0; i<aggSizes.size(); ++i)
-      if (aggSizes[i] != 2) {
-        foundAggNotSize2=true;
-        break;
-      }
+    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+    level.Release("Aggregates", aggFact.get());
+    return aggregates;
+  }  // gimmeHybridAggregates
+};
 
-    switch (comm->getSize()) {
-
-      case 1 :
-        TEST_EQUALITY(numAggs, 18);
-        TEST_EQUALITY(foundAggNotSize2, false);
-        break;
-
-      case 2:
-        TEST_EQUALITY(numAggs, 9);
-        TEST_EQUALITY(foundAggNotSize2, false);
-        break;
-
-      case 3:
-        TEST_EQUALITY(numAggs, 6);
-        TEST_EQUALITY(foundAggNotSize2, false);
-        break;
-
-      case 4:
-        TEST_EQUALITY(numAggs, 4);
-        TEST_EQUALITY(foundAggNotSize2, true);
-        break;
-
-      default:
-        std::string msg = "Only 1-4 MPI processes are supported.";
-        //throw(MueLu::Exceptions::NotImplemented(msg));
-        out << msg << std::endl;
-        break;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo);
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+}
 
-    //ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
-    int root = out.getOutputToRootOnly();
-    out.setOutputToRootOnly(-1);
-    for (int j=0; j<comm->getSize(); ++j) {
-      if (comm->getRank() == j) {
-        out << "++ pid " << j << " ++" << std::endl;
-        out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
-        for (int i=0; i< numAggs; ++i) {
-          out << "   aggregate " << i << ": ";
-          for (int k=aggStart[i]; k< aggStart[i+1]; ++k)
-            out << aggToRowMap[k] << " ";
-          out << std::endl;
-        }
-      }
-      comm->barrier();
+///////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GetNumAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<const Map> rowmap = A->getRowMap();
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates         = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo);
+  GO numAggs                         = aggregates->GetNumAggregates();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+  for (LO i = 0; i < numAggs; ++i)
+    aggSizes[i] = aggStart[i + 1] - aggStart[i];
+
+  bool foundAggNotSize3 = false;
+  for (int i = 0; i < aggSizes.size(); ++i)
+    if (aggSizes[i] != 3) {
+      foundAggNotSize3 = true;
+      break;
     }
-    out.setOutputToRootOnly(root);
-
-  } //UncoupledPhase3
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledInterface, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
 
-    // Get MPI parameters
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  switch (comm->getSize()) {
+    case 1:
+      TEST_EQUALITY(numAggs, 12);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 2:
+      TEST_EQUALITY(numAggs, 6);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 3:
+      TEST_EQUALITY(numAggs, 4);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 4:
+      TEST_EQUALITY(numAggs, 3);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    default:
+      std::string msg = "Only 1-4 MPI processes are supported.";
+      // throw(MueLu::Exceptions::NotImplemented(msg));
+      out << msg << std::endl;
+      break;
+  }
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  // ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
+  int root = out.getOutputToRootOnly();
+  out.setOutputToRootOnly(-1);
+  for (int j = 0; j < comm->getSize(); ++j) {
+    if (comm->getRank() == j) {
+      out << "++ pid " << j << " ++" << std::endl;
+      out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
+      for (int i = 0; i < numAggs; ++i) {
+        out << "   aggregate " << i << ": ";
+        for (int k = aggStart[i]; k < aggStart[i + 1]; ++k)
+          out << aggToRowMap[k] << " ";
+        out << std::endl;
+      }
+    }
+    comm->barrier();
+  }
+  out.setOutputToRootOnly(root);
+
+}  // GetNumAggregates
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo);
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+}
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustUncoupledAggregationFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<const Map> rowmap = A->getRowMap();
-    rowmap->describe(out,Teuchos::VERB_EXTREME);
-    out<< "getLocalNumElements() gives: " << rowmap->getLocalNumElements() << std::endl;
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory);
+  aggFact->SetOrdering("graph");
+  TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
+  aggFact->SetOrdering("natural");
+  TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
+  aggFact->SetOrdering("random");
+  TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
 
-    // Specify root nodes on interface
-    Teuchos::Array<LO> nodeOnInterface(rowmap->getLocalNumElements(),0);
-    if( rowmap->getMinAllGlobalIndex() == rowmap->getMinGlobalIndex() )
-      nodeOnInterface[0] = 1;
-    if( rowmap->getMaxAllGlobalIndex() == rowmap->getMaxGlobalIndex() )
-      nodeOnInterface[rowmap->getLocalNumElements()-1] = 1;
+  aggFact->SetMaxNeighAlreadySelected(12);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 12, true);
 
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::gimmeInterfaceAggregates(A, amalgInfo,nodeOnInterface);
+  aggFact->SetMaxNeighAlreadySelected(0);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 0, true);
 
+  aggFact->SetMinNodesPerAggregate(0);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 0, true);
 
-    // Check to see if specified nodes are root nodes
-    for(LO i=0; i<nodeOnInterface.size(); i++){
-      if (nodeOnInterface[i])
-        TEST_EQUALITY(aggregates->IsRoot( i ), true);
+  aggFact->SetMinNodesPerAggregate(3);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 3, true);
+}
+///////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GetNumUncoupledAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<const Map> rowmap = A->getRowMap();
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates         = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo);
+  GO numAggs                         = aggregates->GetNumAggregates();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+  for (LO i = 0; i < numAggs; ++i)
+    aggSizes[i] = aggStart[i + 1] - aggStart[i];
+
+  bool foundAggNotSize3 = false;
+  for (int i = 0; i < aggSizes.size(); ++i)
+    if (aggSizes[i] != 3) {
+      foundAggNotSize3 = true;
+      break;
     }
 
-  } //UncoupledInterface
+  switch (comm->getSize()) {
+    case 1:
+      TEST_EQUALITY(numAggs, 12);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 2:
+      TEST_EQUALITY(numAggs, 6);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 3:
+      TEST_EQUALITY(numAggs, 4);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 4:
+      TEST_EQUALITY(numAggs, 3);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    default:
+      std::string msg = "Only 1-4 MPI processes are supported.";
+      // throw(MueLu::Exceptions::NotImplemented(msg));
+      out << msg << std::endl;
+      break;
+  }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustStructuredAggregationGlobal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
-
-    // Get MPI parameters
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    LO myRank   = comm->getRank();
-
-    // Set global geometric data
-    const bool coupled = true;
-    LO numDimensions = 3;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    Array<GO> meshData;
-    const std::string meshLayout = "Global Lexicographic";
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = ( (numDimensions == 1) ? 15 : 6 );
-      } else {
-        gNodesPerDir[dim] = 1;
+  // ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
+  int root = out.getOutputToRootOnly();
+  out.setOutputToRootOnly(-1);
+  for (int j = 0; j < comm->getSize(); ++j) {
+    if (comm->getRank() == j) {
+      out << "++ pid " << j << " ++" << std::endl;
+      out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
+      for (int i = 0; i < numAggs; ++i) {
+        out << "   aggregate " << i << ": ";
+        for (int k = aggStart[i]; k < aggStart[i + 1]; ++k)
+          out << aggToRowMap[k] << " ";
+        out << std::endl;
       }
     }
+    comm->barrier();
+  }
+  out.setOutputToRootOnly(root);
+
+}  // GetNumAggregates
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<const Map> rowmap = A->getRowMap();
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates         = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo, true, false, false, false);
+  GO numAggs                         = aggregates->GetNumAggregates();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+  for (LO i = 0; i < numAggs; ++i)
+    aggSizes[i] = aggStart[i + 1] - aggStart[i];
+
+  bool foundAggNotSize3 = false;
+  for (int i = 0; i < aggSizes.size(); ++i)
+    if (aggSizes[i] != 3) {
+      foundAggNotSize3 = true;
+      break;
+    }
 
-    if(myRank == 0) std::cout << "About to build the coordinates" << std::endl;
-
-    RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    if(myRank == 0) std::cout << "About to build the aggregates" << std::endl;
-
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::
-      gimmeStructuredAggregates(A, gNodesPerDir, lNodesPerDir, coupled, numDimensions, meshLayout,
-                                meshData);
-
-    TEST_EQUALITY(aggregates != Teuchos::null, true);
+  switch (comm->getSize()) {
+    case 1:
+      TEST_EQUALITY(numAggs, 12);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 2:
+      TEST_EQUALITY(numAggs, 6);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 3:
+      TEST_EQUALITY(numAggs, 4);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    case 4:
+      TEST_EQUALITY(numAggs, 3);
+      TEST_EQUALITY(foundAggNotSize3, false);
+      break;
+
+    default:
+      std::string msg = "Only 1-4 MPI processes are supported.";
+      // throw(MueLu::Exceptions::NotImplemented(msg));
+      out << msg << std::endl;
+      break;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustStructuredAggregationLocal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
-
-    // Get MPI parameter
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    LO numRanks = comm->getSize();
-
-    // Set global geometric data
-    const bool coupled = true;
-    const std::string meshLayout = std::string("Local Lexicographic");
-    LO numDimensions = 2;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    Array<GO> meshData(10*numRanks);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = ( (numDimensions == 1) ? 15 : 6 );
-      } else {
-        gNodesPerDir[dim] = 1;
+  // ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
+  int root = out.getOutputToRootOnly();
+  out.setOutputToRootOnly(-1);
+  for (int j = 0; j < comm->getSize(); ++j) {
+    if (comm->getRank() == j) {
+      out << "++ pid " << j << " ++" << std::endl;
+      out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
+      for (int i = 0; i < numAggs; ++i) {
+        out << "   aggregate " << i << ": ";
+        for (int k = aggStart[i]; k < aggStart[i + 1]; ++k)
+          out << aggToRowMap[k] << " ";
+        out << std::endl;
       }
     }
-
-    RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::
-      gimmeStructuredAggregates(A, gNodesPerDir, lNodesPerDir, coupled, numDimensions, meshLayout,
-                                meshData);
-
-    TEST_EQUALITY(aggregates != Teuchos::null, true);
+    comm->barrier();
   }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, HybridAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    // Get MPI parameters
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    LO numRanks = comm->getSize();
-    LO myRank   = comm->getRank();
-
-    // Set hybrid region type
-    std::string regionType;
-    if( myRank % 2 == 0 ){
-      regionType = "structured";
-    } else {
-      regionType = "uncoupled";
+  out.setOutputToRootOnly(root);
+
+}  // UncoupledPhase1
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<const Map> rowmap = A->getRowMap();
+  RCP<AmalgamationInfo> amalgInfo;
+  bool bSuccess = true;
+  TEUCHOS_TEST_THROW((AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo, false, true, true, false)),
+                     MueLu::Exceptions::RuntimeError, out, bSuccess);
+  TEST_EQUALITY(bSuccess, true);
+}  // UncoupledPhase2
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledPhase3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<const Map> rowmap = A->getRowMap();
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates         = AggregateGenerator<SC, LO, GO, NO>::gimmeUncoupledAggregates(A, amalgInfo, false, false, false, true);
+  GO numAggs                         = aggregates->GetNumAggregates();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgInfo->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+  for (LO i = 0; i < numAggs; ++i)
+    aggSizes[i] = aggStart[i + 1] - aggStart[i];
+
+  // LBV on 09/27/19: the aggregate size of 2 is only
+  // expected because the problem is 1D and the nodes
+  // are treated in a lexicographic order. Using a
+  // different ordering will almost always result in
+  // at least on aggregate of size 3 and one of size 1.
+  bool foundAggNotSize2 = false;
+  for (int i = 0; i < aggSizes.size(); ++i)
+    if (aggSizes[i] != 2) {
+      foundAggNotSize2 = true;
+      break;
     }
 
-    // Set global geometric data
-    const std::string meshLayout = std::string("Global Lexicographic");
-    LO numDimensions = 2;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    Array<GO> meshData(10*numRanks);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = ( (numDimensions == 1) ? 15 : 6 );
-      } else {
-        gNodesPerDir[dim] = 1;
+  switch (comm->getSize()) {
+    case 1:
+      TEST_EQUALITY(numAggs, 18);
+      TEST_EQUALITY(foundAggNotSize2, false);
+      break;
+
+    case 2:
+      TEST_EQUALITY(numAggs, 9);
+      TEST_EQUALITY(foundAggNotSize2, false);
+      break;
+
+    case 3:
+      TEST_EQUALITY(numAggs, 6);
+      TEST_EQUALITY(foundAggNotSize2, false);
+      break;
+
+    case 4:
+      TEST_EQUALITY(numAggs, 4);
+      TEST_EQUALITY(foundAggNotSize2, true);
+      break;
+
+    default:
+      std::string msg = "Only 1-4 MPI processes are supported.";
+      // throw(MueLu::Exceptions::NotImplemented(msg));
+      out << msg << std::endl;
+      break;
+  }
+
+  // ArrayRCP< ArrayRCP<GO> > aggToRowMap(numAggs);
+  int root = out.getOutputToRootOnly();
+  out.setOutputToRootOnly(-1);
+  for (int j = 0; j < comm->getSize(); ++j) {
+    if (comm->getRank() == j) {
+      out << "++ pid " << j << " ++" << std::endl;
+      out << "   num local DOFs = " << rowmap->getLocalNumElements() << std::endl;
+      for (int i = 0; i < numAggs; ++i) {
+        out << "   aggregate " << i << ": ";
+        for (int k = aggStart[i]; k < aggStart[i + 1]; ++k)
+          out << aggToRowMap[k] << " ";
+        out << std::endl;
       }
     }
-
-    RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-    // Build Problem
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    //aggregates
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<Aggregates> aggregates = AggregateGenerator<SC,LO,GO,NO>::
-      gimmeHybridAggregates(A, amalgInfo, regionType,
-                            gNodesPerDir, lNodesPerDir,
-                            numDimensions);
-
-
-    TEST_EQUALITY(aggregates != Teuchos::null, true);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
+    comm->barrier();
   }
+  out.setOutputToRootOnly(root);
 
+}  // UncoupledPhase3
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GreedyDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running greedy Dirichlet" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-//    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(30);
-    // Make a Matrix with multiple degrees of freedom per node
-    GlobalOrdinal nx = 20, ny = 20;
-
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-
-    map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
-
-
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar>  values;
-
-    // Create a dirichlet boundary row.
-    LocalOrdinal localRowToZero = 5; // Corresponds to a Dof on local graph node 2
-
-    A->resumeFill();
-    A->getLocalRowView(localRowToZero, indices, values);
-    Array<Scalar> newvalues(values.size(),TST::zero());
-    for (int j = 0; j < indices.size(); j++)
-      //keep diagonal
-      if (indices[j] == localRowToZero) newvalues[j] = values[j];
-    A->replaceLocalValues(localRowToZero,indices,newvalues);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, UncoupledInterface, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    A->fillComplete();
+  // Get MPI parameters
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
 
-    ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
-    TEST_EQUALITY(drows[localRowToZero], true);
-    TEST_EQUALITY(drows[localRowToZero-1], false);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
 
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(false));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+  RCP<const Map> rowmap = A->getRowMap();
+  rowmap->describe(out, Teuchos::VERB_EXTREME);
+  out << "getLocalNumElements() gives: " << rowmap->getLocalNumElements() << std::endl;
 
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    typename Aggregates::LO_view aggPtr;
-    typename Aggregates::LO_view aggNodes;
-    typename Aggregates::LO_view unaggregated;
+  // Specify root nodes on interface
+  Teuchos::Array<LO> nodeOnInterface(rowmap->getLocalNumElements(), 0);
+  if (rowmap->getMinAllGlobalIndex() == rowmap->getMinGlobalIndex())
+    nodeOnInterface[0] = 1;
+  if (rowmap->getMaxAllGlobalIndex() == rowmap->getMaxGlobalIndex())
+    nodeOnInterface[rowmap->getLocalNumElements() - 1] = 1;
 
-    aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::gimmeInterfaceAggregates(A, amalgInfo, nodeOnInterface);
 
-    typename Aggregates::LO_view::HostMirror unaggregated_h = Kokkos::create_mirror_view(unaggregated);
-    Kokkos::deep_copy(unaggregated_h, unaggregated);
+  // Check to see if specified nodes are root nodes
+  for (LO i = 0; i < nodeOnInterface.size(); i++) {
+    if (nodeOnInterface[i])
+      TEST_EQUALITY(aggregates->IsRoot(i), true);
+  }
 
-    //     Test to check that the dirichlet node is aggregated:
-    for( LO i = 0; i < (LO)unaggregated_h.extent(0); i++){
-      TEST_EQUALITY( unaggregated_h(i) == 2, false);
+}  // UncoupledInterface
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustStructuredAggregationGlobal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
+
+  // Get MPI parameters
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  LO myRank                          = comm->getRank();
+
+  // Set global geometric data
+  const bool coupled = true;
+  LO numDimensions   = 3;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  Array<GO> meshData;
+  const std::string meshLayout = "Global Lexicographic";
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = ((numDimensions == 1) ? 15 : 6);
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-    // Repeat with greedy Dirichlet
-    Level levelGreedyAndNoPreserve;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(levelGreedyAndNoPreserve);
-    levelGreedyAndNoPreserve.Set("A", A);
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
-
-    levelGreedyAndNoPreserve.Request("Aggregates", aggFact.get());
-    levelGreedyAndNoPreserve.Request("UnAmalgamationInfo", amalgFact.get());
-
-    levelGreedyAndNoPreserve.Request(*aggFact);
-    aggFact->Build(levelGreedyAndNoPreserve);
-    aggregates = levelGreedyAndNoPreserve.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    aggFact->SetParameter("aggregation: preserve Dirichlet points",Teuchos::ParameterEntry(false));
-
-    aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
-    unaggregated_h = Kokkos::create_mirror_view(unaggregated);
-    Kokkos::deep_copy(unaggregated_h, unaggregated);
-
-    // GH: loop over the unaggregated list and add to the counter each time node 2 appears
-    int unaggregated_count_node2 = 0;
-    for( LO i = 0; i < (LO)unaggregated_h.extent(0); i++)
-      if(unaggregated_h(i) == (LO)2)
-        unaggregated_count_node2++;
-
-    TEST_EQUALITY(unaggregated_count_node2,1);// check that the node with the Dof flagged as dirichlet appears in the list once
+  }
 
-    // GH: print this to out since the test can be sensitive to ordering issues
-    out << "unaggregated_h = [";
-    for( LO i = 0; i < (LO)unaggregated_h.extent(0)-1; i++){
-      out << unaggregated_h(i) << ", ";
-    }
-    out << unaggregated_h(unaggregated_h.extent(0)-1) << "]\n";
+  if (myRank == 0) std::cout << "About to build the coordinates" << std::endl;
 
-    // Repeat with greedy Dirichlet and preserve Dirichlet points
-    Level levelGreedyAndPreserve;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(levelGreedyAndPreserve);
-    levelGreedyAndPreserve.Set("A", A);
+  RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData);
 
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
 
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: preserve Dirichlet points",Teuchos::ParameterEntry(true));
+  if (myRank == 0) std::cout << "About to build the aggregates" << std::endl;
 
-    levelGreedyAndPreserve.Request("Aggregates", aggFact.get());
-    levelGreedyAndPreserve.Request("UnAmalgamationInfo", amalgFact.get());
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::
+      gimmeStructuredAggregates(A, gNodesPerDir, lNodesPerDir, coupled, numDimensions, meshLayout,
+                                meshData);
 
-    levelGreedyAndPreserve.Request(*aggFact);
-    aggFact->Build(levelGreedyAndPreserve);
-    aggregates = levelGreedyAndPreserve.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+}
 
-    aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
-    unaggregated_h = Kokkos::create_mirror_view(unaggregated);
-    Kokkos::deep_copy(unaggregated_h, unaggregated);
-    for( LO i = 0; i < (LO)unaggregated_h.extent(0); i++){
-      TEST_EQUALITY( unaggregated_h(i) == 2, false);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, JustStructuredAggregationLocal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef typename Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> xdMV;
+
+  // Get MPI parameter
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  LO numRanks                        = comm->getSize();
+
+  // Set global geometric data
+  const bool coupled           = true;
+  const std::string meshLayout = std::string("Local Lexicographic");
+  LO numDimensions             = 2;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  Array<GO> meshData(10 * numRanks);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = ((numDimensions == 1) ? 15 : 6);
+    } else {
+      gNodesPerDir[dim] = 1;
     }
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, RootStencil, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    // Make a "hot dog" matrix
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace2D");
-    double factor=10;
-    matrixParams.set("stretchx",1.0);
-    matrixParams.set("stretchy",factor);
-
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,TestHelpers::Parameters::getLib());
-
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-    double dropTol = 0.025;
-
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+  RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
 
-    RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
-    filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    filterFact->SetFactory("Aggregates", aggFact);
-    filterFact->SetFactory("Graph",dropFact);
-    filterFact->SetFactory("Filtering",dropFact);
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
 
-    Teuchos::ParameterList params;
-    params.set("filtered matrix: use lumping",true);
-    params.set("filtered matrix: use root stencil",true);
-    filterFact->SetParameterList(params);
-    level.Request("A",filterFact.get());
-
-    filterFact->Build(level);
-    RCP<Matrix> Afiltered = level.Get<RCP<Matrix> >("A",filterFact.get());
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::
+      gimmeStructuredAggregates(A, gNodesPerDir, lNodesPerDir, coupled, numDimensions, meshLayout,
+                                meshData);
 
-    // Now check stuff
-    //    print_matrix("A",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(A));
-    //    print_matrix("Afiltered",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Afiltered));
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+}
 
-    // We use the full graph for the filtered matrix, so the notional nnz should be the same
-    TEST_EQUALITY(A->getLocalNumRows()==Afiltered->getLocalNumRows(), true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, HybridAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // Get MPI parameters
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  LO numRanks                        = comm->getSize();
+  LO myRank                          = comm->getRank();
+
+  // Set hybrid region type
+  std::string regionType;
+  if (myRank % 2 == 0) {
+    regionType = "structured";
+  } else {
+    regionType = "uncoupled";
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingRootStencil, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running spread lumping" << std::endl;
-
-    // Make a "hot dog" matrix
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace2D");
-    double factor=10;
-    matrixParams.set("stretchx",1.0);
-    matrixParams.set("stretchy",factor);
-
+  // Set global geometric data
+  const std::string meshLayout = std::string("Global Lexicographic");
+  LO numDimensions             = 2;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  Array<GO> meshData(10 * numRanks);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = ((numDimensions == 1) ? 15 : 6);
+    } else {
+      gNodesPerDir[dim] = 1;
+    }
+  }
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,TestHelpers::Parameters::getLib());
+  RCP<const Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+  // Build Problem
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  // aggregates
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<Aggregates> aggregates = AggregateGenerator<SC, LO, GO, NO>::
+      gimmeHybridAggregates(A, amalgInfo, regionType,
+                            gNodesPerDir, lNodesPerDir,
+                            numDimensions);
 
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+}
 
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-    double dropTol = 0.025;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, GreedyDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running greedy Dirichlet" << std::endl;
 
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
 
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+  //    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(30);
+  // Make a Matrix with multiple degrees of freedom per node
+  GlobalOrdinal nx = 20, ny = 20;
 
-    RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
-    filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    filterFact->SetFactory("Aggregates", aggFact);
-    filterFact->SetFactory("Graph",dropFact);
-    filterFact->SetFactory("Filtering",dropFact);
-
-    Teuchos::ParameterList params;
-    params.set("filtered matrix: use lumping",true);
-    params.set("filtered matrix: use root stencil",true);
-    params.set("filtered matrix: use spread lumping",true);
-    params.set("filtered matrix: spread lumping diag dom growth factor",1.1);
-    params.set("filtered matrix: spread lumping diag dom cap",2.0);
-    filterFact->SetParameterList(params);
-    level.Request("A",filterFact.get());
-
-    filterFact->Build(level);
-    RCP<Matrix> Afiltered = level.Get<RCP<Matrix> >("A",filterFact.get());
-
-    // Now check stuff
-    //    print_matrix("A",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(A));
-    //    print_matrix("Afiltered",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Afiltered));
-
-    // We use the full graph for the filtered matrix, so the notional nnz should be the same
-    TEST_EQUALITY(A->getLocalNumRows()==Afiltered->getLocalNumRows(), true);
-  } //SpreadLumping
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running spread lumping" << std::endl;
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Map> map                 = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
 
-    // Make a "hot dog" matrix
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace2D");
-    double factor=10;
-    matrixParams.set("stretchx",1.0);
-    matrixParams.set("stretchy",factor);
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
 
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,TestHelpers::Parameters::getLib());
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> values;
 
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
+  // Create a dirichlet boundary row.
+  LocalOrdinal localRowToZero = 5;  // Corresponds to a Dof on local graph node 2
 
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-    double dropTol = 0.025;
+  A->resumeFill();
+  A->getLocalRowView(localRowToZero, indices, values);
+  Array<Scalar> newvalues(values.size(), TST::zero());
+  for (int j = 0; j < indices.size(); j++)
+    // keep diagonal
+    if (indices[j] == localRowToZero) newvalues[j] = values[j];
+  A->replaceLocalValues(localRowToZero, indices, newvalues);
 
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  A->fillComplete();
 
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+  ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
+  TEST_EQUALITY(drows[localRowToZero], true);
+  TEST_EQUALITY(drows[localRowToZero - 1], false);
 
-    RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
-    filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    filterFact->SetFactory("Aggregates", aggFact);
-    filterFact->SetFactory("Graph",dropFact);
-    filterFact->SetFactory("Filtering",dropFact);
-
-    Teuchos::ParameterList params;
-    params.set("filtered matrix: use lumping",true);
-    params.set("filtered matrix: use spread lumping",false);
-    params.set("filtered matrix: reuse graph",true);
-    params.set("filtered matrix: spread lumping diag dom growth factor",1.1);
-    params.set("filtered matrix: spread lumping diag dom cap",2.0);
-    filterFact->SetParameterList(params);
-    level.Request("A",filterFact.get());
-
-    filterFact->Build(level);
-    RCP<Matrix> Afiltered = level.Get<RCP<Matrix> >("A",filterFact.get());
-
-    // We use the full graph for the filtered matrix, so the notional nnz should be the same
-    TEST_EQUALITY(A->getLocalNumRows()==Afiltered->getLocalNumRows(), true);
-  } // SpreadLumpingReuseGraph
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingNoStencilRootNoReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running spread lumping" << std::endl;
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
 
-    // Make a "hot dog" matrix
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace2D");
-    double factor=10;
-    matrixParams.set("stretchx",1.0);
-    matrixParams.set("stretchy",factor);
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
 
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(false));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,TestHelpers::Parameters::getLib());
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
 
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
 
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-    double dropTol = 0.025;
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  typename Aggregates::LO_view aggPtr;
+  typename Aggregates::LO_view aggNodes;
+  typename Aggregates::LO_view unaggregated;
 
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
 
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+  typename Aggregates::LO_view::HostMirror unaggregated_h = Kokkos::create_mirror_view(unaggregated);
+  Kokkos::deep_copy(unaggregated_h, unaggregated);
 
-    RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
-    filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    filterFact->SetFactory("Aggregates", aggFact);
-    filterFact->SetFactory("Graph",dropFact);
-    filterFact->SetFactory("Filtering",dropFact);
-
-    Teuchos::ParameterList params;
-    params.set("filtered matrix: use lumping",true);
-    params.set("filtered matrix: use spread lumping",false);
-    params.set("filtered matrix: reuse graph",false);
-    params.set("filtered matrix: spread lumping diag dom growth factor",1.1);
-    params.set("filtered matrix: spread lumping diag dom cap",2.0);
-    filterFact->SetParameterList(params);
-    level.Request("A",filterFact.get());
-
-    filterFact->Build(level);
-    RCP<Matrix> Afiltered = level.Get<RCP<Matrix> >("A",filterFact.get());
-
-    // We use the full graph for the filtered matrix, so the notional nnz should be the same
-    TEST_EQUALITY(A->getLocalNumRows()==Afiltered->getLocalNumRows(), true);
+  //     Test to check that the dirichlet node is aggregated:
+  for (LO i = 0; i < (LO)unaggregated_h.extent(0); i++) {
+    TEST_EQUALITY(unaggregated_h(i) == 2, false);
   }
+  // Repeat with greedy Dirichlet
+  Level levelGreedyAndNoPreserve;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(levelGreedyAndNoPreserve);
+  levelGreedyAndNoPreserve.Set("A", A);
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  levelGreedyAndNoPreserve.Request("Aggregates", aggFact.get());
+  levelGreedyAndNoPreserve.Request("UnAmalgamationInfo", amalgFact.get());
+
+  levelGreedyAndNoPreserve.Request(*aggFact);
+  aggFact->Build(levelGreedyAndNoPreserve);
+  aggregates = levelGreedyAndNoPreserve.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  aggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(false));
+
+  aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
+  unaggregated_h = Kokkos::create_mirror_view(unaggregated);
+  Kokkos::deep_copy(unaggregated_h, unaggregated);
+
+  // GH: loop over the unaggregated list and add to the counter each time node 2 appears
+  int unaggregated_count_node2 = 0;
+  for (LO i = 0; i < (LO)unaggregated_h.extent(0); i++)
+    if (unaggregated_h(i) == (LO)2)
+      unaggregated_count_node2++;
+
+  TEST_EQUALITY(unaggregated_count_node2, 1);  // check that the node with the Dof flagged as dirichlet appears in the list once
+
+  // GH: print this to out since the test can be sensitive to ordering issues
+  out << "unaggregated_h = [";
+  for (LO i = 0; i < (LO)unaggregated_h.extent(0) - 1; i++) {
+    out << unaggregated_h(i) << ", ";
+  }
+  out << unaggregated_h(unaggregated_h.extent(0) - 1) << "]\n";
+
+  // Repeat with greedy Dirichlet and preserve Dirichlet points
+  Level levelGreedyAndPreserve;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(levelGreedyAndPreserve);
+  levelGreedyAndPreserve.Set("A", A);
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(true));
+
+  levelGreedyAndPreserve.Request("Aggregates", aggFact.get());
+  levelGreedyAndPreserve.Request("UnAmalgamationInfo", amalgFact.get());
+
+  levelGreedyAndPreserve.Request(*aggFact);
+  aggFact->Build(levelGreedyAndPreserve);
+  aggregates = levelGreedyAndPreserve.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+
+  aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
+  unaggregated_h = Kokkos::create_mirror_view(unaggregated);
+  Kokkos::deep_copy(unaggregated_h, unaggregated);
+  for (LO i = 0; i < (LO)unaggregated_h.extent(0); i++) {
+    TEST_EQUALITY(unaggregated_h(i) == 2, false);
+  }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadNoLumpingNoStencilRootNoReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running spread lumping" << std::endl;
-
-    // Make a "hot dog" matrix
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace2D");
-    double factor=10;
-    matrixParams.set("stretchx",1.0);
-    matrixParams.set("stretchy",factor);
-
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,TestHelpers::Parameters::getLib());
-
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory> aggFact;
-    double dropTol = 0.025;
-
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(dropTol));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
-
-    RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
-    filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    filterFact->SetFactory("Aggregates", aggFact);
-    filterFact->SetFactory("Graph",dropFact);
-    filterFact->SetFactory("Filtering",dropFact);
-
-    Teuchos::ParameterList params;
-    params.set("filtered matrix: use lumping",false);
-    params.set("filtered matrix: use spread lumping",false);
-    params.set("filtered matrix: reuse graph",false);
-    params.set("filtered matrix: spread lumping diag dom growth factor",1.1);
-    params.set("filtered matrix: spread lumping diag dom cap",2.0);
-    filterFact->SetParameterList(params);
-    level.Request("A",filterFact.get());
-
-    filterFact->Build(level);
-    RCP<Matrix> Afiltered = level.Get<RCP<Matrix> >("A",filterFact.get());
-
-    // We use the full graph for the filtered matrix, so the notional nnz should be the same
-    TEST_EQUALITY(A->getLocalNumRows()==Afiltered->getLocalNumRows(), true);
-  } // SpreadNoLumpingNoStencilRootNoReuseGraph
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, AllowDroppingToCreateAdditionalDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test option that allows dropping during aggregation to create new Dirichlet rows" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    A->resumeFill();
-
-    // Create one row on every processor with small off-diagonal entries that will be dropped with
-    // an appropriately chosen threshold.  Avoid domain boundaries.
-    LocalOrdinal localRowToModify = 1;
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar>  values;
-    A->getLocalRowView(localRowToModify, indices, values);
-    Array<Scalar> newvalues(values.size(),TST::zero());
-    for (int j = 0; j < indices.size(); j++) {
-      if (indices[j] == localRowToModify)
-        newvalues[j] = values[j]; //keep diagonal unmodified
-      else
-        newvalues[j] = -TST::eps();
-    }
-    A->replaceLocalValues(localRowToModify,indices,newvalues);
-    A->fillComplete();
-
-    // Dropping connections will not lead to the creation of new Dirichlet rows.
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: dropping may create Dirichlet",Teuchos::ParameterEntry(false));
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(Teuchos::as<double>(-100*TST::eps())));
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    level.Request("Graph",dropFact.get());
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    RCP<GraphBase> graph = level.Get<RCP<GraphBase>>("Graph",dropFact.get());
-    ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
-    int numDirichletRows=0;
-    LO numRows = graph->GetNodeNumVertices();
-    for (LO i = 0; i < numRows; i++)
-      if (dirichletBoundaryMap[i] == true)
-        numDirichletRows++;
-    TEST_EQUALITY(numDirichletRows, 0);
-
-    Array< LO > aggPtr;
-    Array< LO > aggNodes;
-    Array< LO > unaggregated;
-
-    // Repeat with "aggregation: dropping may create Dirichlet" = TRUE, i.e.,
-    // dropping connections may lead to the creation of new Dirichlet rows.
-    // The second row should be flagged as Dirichlet because all off-diagonal entries are dropped.
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetParameter("aggregation: dropping may create Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(Teuchos::as<double>(-100*TST::eps())));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    level.Request("Graph",dropFact.get());
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", dropFact);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, RootStencil, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // Make a "hot dog" matrix
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace2D");
+  double factor = 10;
+  matrixParams.set("stretchx", 1.0);
+  matrixParams.set("stretchy", factor);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, TestHelpers::Parameters::getLib());
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
+  double dropTol = 0.025;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
+  filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  filterFact->SetFactory("Aggregates", aggFact);
+  filterFact->SetFactory("Graph", dropFact);
+  filterFact->SetFactory("Filtering", dropFact);
+
+  Teuchos::ParameterList params;
+  params.set("filtered matrix: use lumping", true);
+  params.set("filtered matrix: use root stencil", true);
+  filterFact->SetParameterList(params);
+  level.Request("A", filterFact.get());
+
+  filterFact->Build(level);
+  RCP<Matrix> Afiltered = level.Get<RCP<Matrix>>("A", filterFact.get());
+
+  // Now check stuff
+  //    print_matrix("A",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(A));
+  //    print_matrix("Afiltered",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Afiltered));
+
+  // We use the full graph for the filtered matrix, so the notional nnz should be the same
+  TEST_EQUALITY(A->getLocalNumRows() == Afiltered->getLocalNumRows(), true);
+}
 
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingRootStencil, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running spread lumping" << std::endl;
+
+  // Make a "hot dog" matrix
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace2D");
+  double factor = 10;
+  matrixParams.set("stretchx", 1.0);
+  matrixParams.set("stretchy", factor);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, TestHelpers::Parameters::getLib());
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
+  double dropTol = 0.025;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
+  filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  filterFact->SetFactory("Aggregates", aggFact);
+  filterFact->SetFactory("Graph", dropFact);
+  filterFact->SetFactory("Filtering", dropFact);
+
+  Teuchos::ParameterList params;
+  params.set("filtered matrix: use lumping", true);
+  params.set("filtered matrix: use root stencil", true);
+  params.set("filtered matrix: use spread lumping", true);
+  params.set("filtered matrix: spread lumping diag dom growth factor", 1.1);
+  params.set("filtered matrix: spread lumping diag dom cap", 2.0);
+  filterFact->SetParameterList(params);
+  level.Request("A", filterFact.get());
+
+  filterFact->Build(level);
+  RCP<Matrix> Afiltered = level.Get<RCP<Matrix>>("A", filterFact.get());
+
+  // Now check stuff
+  //    print_matrix("A",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(A));
+  //    print_matrix("Afiltered",MueLu::Utilities<SC,LO,GO,NO>::Op2NonConstTpetraCrs(Afiltered));
+
+  // We use the full graph for the filtered matrix, so the notional nnz should be the same
+  TEST_EQUALITY(A->getLocalNumRows() == Afiltered->getLocalNumRows(), true);
+}  // SpreadLumping
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running spread lumping" << std::endl;
+
+  // Make a "hot dog" matrix
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace2D");
+  double factor = 10;
+  matrixParams.set("stretchx", 1.0);
+  matrixParams.set("stretchy", factor);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, TestHelpers::Parameters::getLib());
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
+  double dropTol = 0.025;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
+  filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  filterFact->SetFactory("Aggregates", aggFact);
+  filterFact->SetFactory("Graph", dropFact);
+  filterFact->SetFactory("Filtering", dropFact);
+
+  Teuchos::ParameterList params;
+  params.set("filtered matrix: use lumping", true);
+  params.set("filtered matrix: use spread lumping", false);
+  params.set("filtered matrix: reuse graph", true);
+  params.set("filtered matrix: spread lumping diag dom growth factor", 1.1);
+  params.set("filtered matrix: spread lumping diag dom cap", 2.0);
+  filterFact->SetParameterList(params);
+  level.Request("A", filterFact.get());
+
+  filterFact->Build(level);
+  RCP<Matrix> Afiltered = level.Get<RCP<Matrix>>("A", filterFact.get());
+
+  // We use the full graph for the filtered matrix, so the notional nnz should be the same
+  TEST_EQUALITY(A->getLocalNumRows() == Afiltered->getLocalNumRows(), true);
+}  // SpreadLumpingReuseGraph
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadLumpingNoStencilRootNoReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running spread lumping" << std::endl;
+
+  // Make a "hot dog" matrix
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace2D");
+  double factor = 10;
+  matrixParams.set("stretchx", 1.0);
+  matrixParams.set("stretchy", factor);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, TestHelpers::Parameters::getLib());
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
+  double dropTol = 0.025;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
+  filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  filterFact->SetFactory("Aggregates", aggFact);
+  filterFact->SetFactory("Graph", dropFact);
+  filterFact->SetFactory("Filtering", dropFact);
+
+  Teuchos::ParameterList params;
+  params.set("filtered matrix: use lumping", true);
+  params.set("filtered matrix: use spread lumping", false);
+  params.set("filtered matrix: reuse graph", false);
+  params.set("filtered matrix: spread lumping diag dom growth factor", 1.1);
+  params.set("filtered matrix: spread lumping diag dom cap", 2.0);
+  filterFact->SetParameterList(params);
+  level.Request("A", filterFact.get());
+
+  filterFact->Build(level);
+  RCP<Matrix> Afiltered = level.Get<RCP<Matrix>>("A", filterFact.get());
+
+  // We use the full graph for the filtered matrix, so the notional nnz should be the same
+  TEST_EQUALITY(A->getLocalNumRows() == Afiltered->getLocalNumRows(), true);
+}
 
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    graph = level.Get<RCP<GraphBase>>("Graph",dropFact.get());
-    dirichletBoundaryMap = graph->GetBoundaryNodeMap();
-    numDirichletRows=0;
-    for (LO i = 0; i < numRows; i++)
-      if (dirichletBoundaryMap[i] == true)
-        numDirichletRows++;
-    TEST_EQUALITY(numDirichletRows,1);
-
-  } //AllowDroppingToCreateAdditionalDirichletRows
-
-#define MUELU_ETI_GROUP(Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,JustAggregation,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,GetNumAggregates,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,JustUncoupledAggregation,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,JustUncoupledAggregationFactory,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,GetNumUncoupledAggregates,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,UncoupledPhase1,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,UncoupledPhase2,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,UncoupledPhase3,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,UncoupledInterface,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,JustStructuredAggregationGlobal,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,JustStructuredAggregationLocal,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,GreedyDirichlet,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,AllowDroppingToCreateAdditionalDirichletRows,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA ,RootStencil,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA ,SpreadLumpingRootStencil,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA ,SpreadLumpingReuseGraph,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA ,SpreadLumpingNoStencilRootNoReuseGraph,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA ,SpreadNoLumpingNoStencilRootNoReuseGraph,Scalar,LO,GO,Node)
-
-  //  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,HybridAggregation,Scalar,LO,GO,Node)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FilteredA, SpreadNoLumpingNoStencilRootNoReuseGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running spread lumping" << std::endl;
+
+  // Make a "hot dog" matrix
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace2D");
+  double factor = 10;
+  matrixParams.set("stretchx", 1.0);
+  matrixParams.set("stretchy", factor);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, TestHelpers::Parameters::getLib());
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory> aggFact;
+  double dropTol = 0.025;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(dropTol));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  RCP<FilteredAFactory> filterFact = rcp(new FilteredAFactory);
+  filterFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  filterFact->SetFactory("Aggregates", aggFact);
+  filterFact->SetFactory("Graph", dropFact);
+  filterFact->SetFactory("Filtering", dropFact);
+
+  Teuchos::ParameterList params;
+  params.set("filtered matrix: use lumping", false);
+  params.set("filtered matrix: use spread lumping", false);
+  params.set("filtered matrix: reuse graph", false);
+  params.set("filtered matrix: spread lumping diag dom growth factor", 1.1);
+  params.set("filtered matrix: spread lumping diag dom cap", 2.0);
+  filterFact->SetParameterList(params);
+  level.Request("A", filterFact.get());
+
+  filterFact->Build(level);
+  RCP<Matrix> Afiltered = level.Get<RCP<Matrix>>("A", filterFact.get());
+
+  // We use the full graph for the filtered matrix, so the notional nnz should be the same
+  TEST_EQUALITY(A->getLocalNumRows() == Afiltered->getLocalNumRows(), true);
+}  // SpreadNoLumpingNoStencilRootNoReuseGraph
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates, AllowDroppingToCreateAdditionalDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test option that allows dropping during aggregation to create new Dirichlet rows" << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  A->resumeFill();
+
+  // Create one row on every processor with small off-diagonal entries that will be dropped with
+  // an appropriately chosen threshold.  Avoid domain boundaries.
+  LocalOrdinal localRowToModify = 1;
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> values;
+  A->getLocalRowView(localRowToModify, indices, values);
+  Array<Scalar> newvalues(values.size(), TST::zero());
+  for (int j = 0; j < indices.size(); j++) {
+    if (indices[j] == localRowToModify)
+      newvalues[j] = values[j];  // keep diagonal unmodified
+    else
+      newvalues[j] = -TST::eps();
+  }
+  A->replaceLocalValues(localRowToModify, indices, newvalues);
+  A->fillComplete();
+
+  // Dropping connections will not lead to the creation of new Dirichlet rows.
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: dropping may create Dirichlet", Teuchos::ParameterEntry(false));
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(Teuchos::as<double>(-100 * TST::eps())));
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  level.Request("Graph", dropFact.get());
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates                = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  RCP<GraphBase> graph                      = level.Get<RCP<GraphBase>>("Graph", dropFact.get());
+  ArrayRCP<const bool> dirichletBoundaryMap = graph->GetBoundaryNodeMap();
+  int numDirichletRows                      = 0;
+  LO numRows                                = graph->GetNodeNumVertices();
+  for (LO i = 0; i < numRows; i++)
+    if (dirichletBoundaryMap[i] == true)
+      numDirichletRows++;
+  TEST_EQUALITY(numDirichletRows, 0);
+
+  Array<LO> aggPtr;
+  Array<LO> aggNodes;
+  Array<LO> unaggregated;
+
+  // Repeat with "aggregation: dropping may create Dirichlet" = TRUE, i.e.,
+  // dropping connections may lead to the creation of new Dirichlet rows.
+  // The second row should be flagged as Dirichlet because all off-diagonal entries are dropped.
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetParameter("aggregation: dropping may create Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(Teuchos::as<double>(-100 * TST::eps())));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  level.Request("Graph", dropFact.get());
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  aggregates           = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  graph                = level.Get<RCP<GraphBase>>("Graph", dropFact.get());
+  dirichletBoundaryMap = graph->GetBoundaryNodeMap();
+  numDirichletRows     = 0;
+  for (LO i = 0; i < numRows; i++)
+    if (dirichletBoundaryMap[i] == true)
+      numDirichletRows++;
+  TEST_EQUALITY(numDirichletRows, 1);
+
+}  // AllowDroppingToCreateAdditionalDirichletRows
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, JustAggregation, Scalar, LO, GO, Node)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, GetNumAggregates, Scalar, LO, GO, Node)                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, JustUncoupledAggregation, Scalar, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, JustUncoupledAggregationFactory, Scalar, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, GetNumUncoupledAggregates, Scalar, LO, GO, Node)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, UncoupledPhase1, Scalar, LO, GO, Node)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, UncoupledPhase2, Scalar, LO, GO, Node)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, UncoupledPhase3, Scalar, LO, GO, Node)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, UncoupledInterface, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, JustStructuredAggregationGlobal, Scalar, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, JustStructuredAggregationLocal, Scalar, LO, GO, Node)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, GreedyDirichlet, Scalar, LO, GO, Node)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates, AllowDroppingToCreateAdditionalDirichletRows, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA, RootStencil, Scalar, LO, GO, Node)                                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA, SpreadLumpingRootStencil, Scalar, LO, GO, Node)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA, SpreadLumpingReuseGraph, Scalar, LO, GO, Node)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA, SpreadLumpingNoStencilRootNoReuseGraph, Scalar, LO, GO, Node)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FilteredA, SpreadNoLumpingNoStencilRootNoReuseGraph, Scalar, LO, GO, Node)
+
+//  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates,HybridAggregation,Scalar,LO,GO,Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/AlgebraicPermutationStrategy.cpp b/packages/muelu/test/unit_tests/AlgebraicPermutationStrategy.cpp
index 2e0d580bbc73..c892fad1b509 100644
--- a/packages/muelu/test/unit_tests/AlgebraicPermutationStrategy.cpp
+++ b/packages/muelu/test/unit_tests/AlgebraicPermutationStrategy.cpp
@@ -51,59 +51,56 @@
 
 #include <MueLu_AlgebraicPermutationStrategy.hpp>
 
-
 namespace MueLuTests {
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_AlgebraicPermutationStrategy, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_AlgebraicPermutationStrategy, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    using TestFactory = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TestFactory = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = TestFactory::Build1DPoisson(200);
-    A->describe(out, Teuchos::VERB_EXTREME);
-    auto map = A->getRowMap();
+  auto A = TestFactory::Build1DPoisson(200);
+  A->describe(out, Teuchos::VERB_EXTREME);
+  auto map = A->getRowMap();
 
-    Level currentLevel;
-    MueLu::AlgebraicPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node> algebraicPermutationStrategy;
-    algebraicPermutationStrategy.BuildPermutation(A, map, currentLevel, MueLu::NoFactory::get());
+  Level currentLevel;
+  MueLu::AlgebraicPermutationStrategy<Scalar, LocalOrdinal, GlobalOrdinal, Node> algebraicPermutationStrategy;
+  algebraicPermutationStrategy.BuildPermutation(A, map, currentLevel, MueLu::NoFactory::get());
 
-    TEST_EQUALITY(true, currentLevel.IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_EQUALITY(true, currentLevel.IsAvailable("permA", MueLu::NoFactory::get()));
-    TEST_EQUALITY(true, currentLevel.IsAvailable("permP", MueLu::NoFactory::get()));
-    TEST_EQUALITY(true, currentLevel.IsAvailable("permQT", MueLu::NoFactory::get()));
-    TEST_EQUALITY(true, currentLevel.IsAvailable("permScaling", MueLu::NoFactory::get()));
+  TEST_EQUALITY(true, currentLevel.IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_EQUALITY(true, currentLevel.IsAvailable("permA", MueLu::NoFactory::get()));
+  TEST_EQUALITY(true, currentLevel.IsAvailable("permP", MueLu::NoFactory::get()));
+  TEST_EQUALITY(true, currentLevel.IsAvailable("permQT", MueLu::NoFactory::get()));
+  TEST_EQUALITY(true, currentLevel.IsAvailable("permScaling", MueLu::NoFactory::get()));
 }
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_AlgebraicPermutationStrategy, Utils, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_AlgebraicPermutationStrategy, Utils, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    std::vector<LocalOrdinal> vec;
+  std::vector<LocalOrdinal> vec;
 
-    std::vector<Scalar> values {10, 12, 40, 50, 32, 78};
-    MueLu::sortingPermutation(values, vec);
+  std::vector<Scalar> values{10, 12, 40, 50, 32, 78};
+  MueLu::sortingPermutation(values, vec);
 
-    TEST_EQUALITY(values.size(), vec.size());
+  TEST_EQUALITY(values.size(), vec.size());
 
-    TEST_EQUALITY(5, vec[0]); // 78
-    TEST_EQUALITY(3, vec[1]); // 50
-    TEST_EQUALITY(2, vec[2]); // 40
-    TEST_EQUALITY(4, vec[3]); // 32
-    TEST_EQUALITY(1, vec[4]); // 12
-    TEST_EQUALITY(0, vec[5]); // 10
+  TEST_EQUALITY(5, vec[0]);  // 78
+  TEST_EQUALITY(3, vec[1]);  // 50
+  TEST_EQUALITY(2, vec[2]);  // 40
+  TEST_EQUALITY(4, vec[3]);  // 32
+  TEST_EQUALITY(1, vec[4]);  // 12
+  TEST_EQUALITY(0, vec[5]);  // 10
 }
 
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_AlgebraicPermutationStrategy, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_AlgebraicPermutationStrategy, Utils, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_AlgebraicPermutationStrategy, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_AlgebraicPermutationStrategy, Utils, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/AmalgamationFactory.cpp b/packages/muelu/test/unit_tests/AmalgamationFactory.cpp
index b01af40454af..5aad146ab45c 100644
--- a/packages/muelu/test/unit_tests/AmalgamationFactory.cpp
+++ b/packages/muelu/test/unit_tests/AmalgamationFactory.cpp
@@ -54,96 +54,89 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<AmalgamationFactory> amalgamationFactory = rcp(new AmalgamationFactory());
-
-    TEST_INEQUALITY(amalgamationFactory, Teuchos::null);
-  } // Constructor
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, DOFGid2NodeId, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // Test layout: 1D mesh w/ 3 DOFs per node
-    const GlobalOrdinal numGlobalNodes = 12;
-    const LocalOrdinal numDofsPerNode = 3;
-    const GlobalOrdinal indexBase = 3;
-    RCP<const Map> nodeMap = MapFactory::Build(lib, numGlobalNodes, indexBase, comm);
-    RCP<const Map> dofMap = MapFactory::Build(lib, numGlobalNodes*numDofsPerNode, indexBase, comm);
-
-    // Probe all nodes in the mesh
-    const LocalOrdinal numLocalDofs = dofMap->getLocalNumElements();
-    LocalOrdinal localNodeID = Teuchos::ScalarTraits<LocalOrdinal>::zero();
-    for (LocalOrdinal localDofID = 0; localDofID < numLocalDofs; ++localDofID)
-    {
-      // Ask AmalgamationFactory for global node ID of this DOF and check w/ expected result
-      GlobalOrdinal nodeID = AmalgamationFactory::DOFGid2NodeId(dofMap->getGlobalElement(localDofID), numDofsPerNode, 0, indexBase);
-      TEST_EQUALITY(nodeID, nodeMap->getGlobalElement(localNodeID));
-
-      // Increment localNodeId, if this is the first DOF of a node
-      if (localDofID % numDofsPerNode == numDofsPerNode - Teuchos::ScalarTraits<LocalOrdinal>::one())
-        ++localNodeID;
-    }
-  } // DOFGid2NodeId
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, AmalgamateMap, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // Test static method AmalgamationFactory::AmalgamateMap().
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "Test static method AmalgamationFactory::AmalgamateMap()." << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    const GlobalOrdinal nx = 32;
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", nx);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::BuildMatrix(matrixList,TestHelpers::Parameters::getLib());
-    LO blkSize=2;
-    Op->SetFixedBlockSize(blkSize);
-
-    RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
-    RCP<const Map> uniqueMap;
-    AmalgamationFactory::AmalgamateMap(*(Op->getRowMap()), *Op, uniqueMap, *theRowTranslation);
-
-    Teuchos::ArrayView<const GO> localEltList = uniqueMap->getLocalElementList();
-    for (size_t j=0; j<uniqueMap->getLocalNumElements(); j++) {
-      TEST_EQUALITY(uniqueMap->getLocalElement(localEltList[j]),static_cast<LO>(j));
-    }
-
-    RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
-    RCP<const Map> nonUniqueMap;
-    AmalgamationFactory::AmalgamateMap(*(Op->getColMap()), *Op, nonUniqueMap, *theColTranslation);
-
-    localEltList = nonUniqueMap->getLocalElementList();
-    for (size_t j=0; j<nonUniqueMap->getLocalNumElements(); j++) {
-      TEST_EQUALITY(nonUniqueMap->getLocalElement(localEltList[j]),static_cast<LO>(j));
-    }
-
-  } // AmalgamateMap
-
-
-  # define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, DOFGid2NodeId, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, AmalgamateMap, Scalar, LO, GO, Node)
-
-# include <MueLu_ETI_4arg.hpp>
-
-} // namespace MueLuTests
-
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<AmalgamationFactory> amalgamationFactory = rcp(new AmalgamationFactory());
+
+  TEST_INEQUALITY(amalgamationFactory, Teuchos::null);
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, DOFGid2NodeId, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // Test layout: 1D mesh w/ 3 DOFs per node
+  const GlobalOrdinal numGlobalNodes = 12;
+  const LocalOrdinal numDofsPerNode  = 3;
+  const GlobalOrdinal indexBase      = 3;
+  RCP<const Map> nodeMap             = MapFactory::Build(lib, numGlobalNodes, indexBase, comm);
+  RCP<const Map> dofMap              = MapFactory::Build(lib, numGlobalNodes * numDofsPerNode, indexBase, comm);
+
+  // Probe all nodes in the mesh
+  const LocalOrdinal numLocalDofs = dofMap->getLocalNumElements();
+  LocalOrdinal localNodeID        = Teuchos::ScalarTraits<LocalOrdinal>::zero();
+  for (LocalOrdinal localDofID = 0; localDofID < numLocalDofs; ++localDofID) {
+    // Ask AmalgamationFactory for global node ID of this DOF and check w/ expected result
+    GlobalOrdinal nodeID = AmalgamationFactory::DOFGid2NodeId(dofMap->getGlobalElement(localDofID), numDofsPerNode, 0, indexBase);
+    TEST_EQUALITY(nodeID, nodeMap->getGlobalElement(localNodeID));
+
+    // Increment localNodeId, if this is the first DOF of a node
+    if (localDofID % numDofsPerNode == numDofsPerNode - Teuchos::ScalarTraits<LocalOrdinal>::one())
+      ++localNodeID;
+  }
+}  // DOFGid2NodeId
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationFactory, AmalgamateMap, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // Test static method AmalgamationFactory::AmalgamateMap().
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "Test static method AmalgamationFactory::AmalgamateMap()." << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  const GlobalOrdinal nx = 32;
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::BuildMatrix(matrixList, TestHelpers::Parameters::getLib());
+  LO blkSize     = 2;
+  Op->SetFixedBlockSize(blkSize);
+
+  RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
+  RCP<const Map> uniqueMap;
+  AmalgamationFactory::AmalgamateMap(*(Op->getRowMap()), *Op, uniqueMap, *theRowTranslation);
+
+  Teuchos::ArrayView<const GO> localEltList = uniqueMap->getLocalElementList();
+  for (size_t j = 0; j < uniqueMap->getLocalNumElements(); j++) {
+    TEST_EQUALITY(uniqueMap->getLocalElement(localEltList[j]), static_cast<LO>(j));
+  }
+
+  RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
+  RCP<const Map> nonUniqueMap;
+  AmalgamationFactory::AmalgamateMap(*(Op->getColMap()), *Op, nonUniqueMap, *theColTranslation);
+
+  localEltList = nonUniqueMap->getLocalElementList();
+  for (size_t j = 0; j < nonUniqueMap->getLocalNumElements(); j++) {
+    TEST_EQUALITY(nonUniqueMap->getLocalElement(localEltList[j]), static_cast<LO>(j));
+  }
+
+}  // AmalgamateMap
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, Constructor, Scalar, LO, GO, Node)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, DOFGid2NodeId, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationFactory, AmalgamateMap, Scalar, LO, GO, Node)
+
+#include <MueLu_ETI_4arg.hpp>
+
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/AmalgamationInfo.cpp b/packages/muelu/test/unit_tests/AmalgamationInfo.cpp
index ee6cac6cdd21..ef74e877b758 100644
--- a/packages/muelu/test/unit_tests/AmalgamationInfo.cpp
+++ b/packages/muelu/test/unit_tests/AmalgamationInfo.cpp
@@ -56,59 +56,56 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    LO fullblocksize    = 1;   // block dim for fixed size blocks
-    GO offset           = 0;   // global offset of dof gids
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    LO stridedblocksize = fullblocksize; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  LO fullblocksize    = 1;              // block dim for fixed size blocks
+  GO offset           = 0;              // global offset of dof gids
+  LO blockid          = -1;             // block id in strided map
+  LO nStridedOffset   = 0;              // DOF offset for strided block id "blockid" (default = 0)
+  LO stridedblocksize = fullblocksize;  // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
 
-    const GO nx = 199;
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  const GO nx        = 199;
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  RCP<Matrix> A      = test_factory::Build1DPoisson(nx);
 
-    RCP<AmalgamationInfo> amalgamationData;
-    RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
-    RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
+  RCP<AmalgamationInfo> amalgamationData;
+  RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
+  RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
 
-    RCP<AmalgamationInfo> amalgamationInfo = rcp(new AmalgamationInfo(theRowTranslation,
-                                                                      theColTranslation,
-                                                                      A->getRowMap(),
-                                                                      A->getColMap(),
-                                                                      A->getColMap(),
-                                                                      fullblocksize,
-                                                                      offset,
-                                                                      blockid,
-                                                                      nStridedOffset,
-                                                                      stridedblocksize) );
-    TEST_INEQUALITY(amalgamationInfo, Teuchos::null);
+  RCP<AmalgamationInfo> amalgamationInfo = rcp(new AmalgamationInfo(theRowTranslation,
+                                                                    theColTranslation,
+                                                                    A->getRowMap(),
+                                                                    A->getColMap(),
+                                                                    A->getColMap(),
+                                                                    fullblocksize,
+                                                                    offset,
+                                                                    blockid,
+                                                                    nStridedOffset,
+                                                                    stridedblocksize));
+  TEST_INEQUALITY(amalgamationInfo, Teuchos::null);
 
-    TEST_EQUALITY(amalgamationInfo->description(), "AmalgamationInfo");
-    LO fullBlockSizeExpected;
-    LO blockIDExpected;
-    LO stridingOffsetExpected;
-    LO stridedBlockSizeExpected;
-    GO indexBaseExpected;
-    amalgamationInfo->GetStridingInformation(fullBlockSizeExpected, blockIDExpected, stridingOffsetExpected, stridedBlockSizeExpected, indexBaseExpected);
-    TEST_EQUALITY(fullBlockSizeExpected == fullblocksize, true);
-    TEST_EQUALITY(blockIDExpected == blockid, true);
-    TEST_EQUALITY(stridingOffsetExpected == nStridedOffset, true);
-    TEST_EQUALITY(stridedBlockSizeExpected == stridedblocksize, true);
-    TEST_EQUALITY(indexBaseExpected == A->getColMap()->getIndexBase(), true);
-    amalgamationInfo->print(*Teuchos::VerboseObjectBase::getDefaultOStream(),MueLu::Extreme);
+  TEST_EQUALITY(amalgamationInfo->description(), "AmalgamationInfo");
+  LO fullBlockSizeExpected;
+  LO blockIDExpected;
+  LO stridingOffsetExpected;
+  LO stridedBlockSizeExpected;
+  GO indexBaseExpected;
+  amalgamationInfo->GetStridingInformation(fullBlockSizeExpected, blockIDExpected, stridingOffsetExpected, stridedBlockSizeExpected, indexBaseExpected);
+  TEST_EQUALITY(fullBlockSizeExpected == fullblocksize, true);
+  TEST_EQUALITY(blockIDExpected == blockid, true);
+  TEST_EQUALITY(stridingOffsetExpected == nStridedOffset, true);
+  TEST_EQUALITY(stridedBlockSizeExpected == stridedblocksize, true);
+  TEST_EQUALITY(indexBaseExpected == A->getColMap()->getIndexBase(), true);
+  amalgamationInfo->print(*Teuchos::VerboseObjectBase::getDefaultOStream(), MueLu::Extreme);
 
-  } // Constructor
+}  // Constructor
 
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo, Constructor, Scalar, LO, GO, Node)
 
-  # define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo, Constructor, Scalar, LO, GO, Node) \
-
-# include <MueLu_ETI_4arg.hpp>
-
-} // namespace MueLuTests
+#include <MueLu_ETI_4arg.hpp>
 
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/BlackBoxPFactory.cpp b/packages/muelu/test/unit_tests/BlackBoxPFactory.cpp
index 25e5c054e528..f519be340f56 100644
--- a/packages/muelu/test/unit_tests/BlackBoxPFactory.cpp
+++ b/packages/muelu/test/unit_tests/BlackBoxPFactory.cpp
@@ -69,793 +69,787 @@
 
 namespace MueLuTests {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class BlackBoxPFactoryTester {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class BlackBoxPFactoryTester {
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor
-    BlackBoxPFactoryTester() { }
-
-    //! Destructor.
-    virtual ~BlackBoxPFactoryTester() { }
-    //@}
-
-    void TestGetGeometricData(Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coordinates,
-                              Array<LO> coarseRate, Array<GO> gNodesPerDim, Array<LO> lNodesPerDim,
-                              LO BlkSize, Array<GO>& gIndices, Array<GO>& gCoarseNodesPerDir,
-                              Array<GO>& ghostGIDs, Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs,
-                              GO& gNumCoarseNodes, Array<LO>& myOffset,
-                              Array<LO>& lCoarseNodesPerDir, Array<LO>& glCoarseNodesPerDir,
-                              Array<LO>& endRate, LO& lNumCoarseNodes, Array<bool>& ghostInterface,
-                              ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes, Array<int>& boundaryFlags) const
-    {
-      // Call the method to be tested.
-      RCP<typename MueLu::BlackBoxPFactory<SC,LO,GO,Node>::NodesIDs> ghostedCoarseNodes
-        = rcp(new typename MueLu::BlackBoxPFactory<SC,LO,GO,Node>::NodesIDs{});
-      MueLu::BlackBoxPFactory<SC,LO,GO,Node> mybbmgPFactory;
-      mybbmgPFactory.GetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, BlkSize,
-                                      gIndices, myOffset, ghostInterface, endRate,
-                                      gCoarseNodesPerDir, lCoarseNodesPerDir, glCoarseNodesPerDir,
-                                      ghostGIDs, coarseNodesGIDs, colGIDs, gNumCoarseNodes,
-                                      lNumCoarseNodes, coarseNodes, boundaryFlags,
-                                      ghostedCoarseNodes);
-    };
-
-    void TestComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
-                                 const Array<LO> endRate, const LO BlkSize,const Array<LO> elemInds,
-                                 const Array<LO> lCoarseElementsPerDir,
-                                 const LO numDimensions, const Array<LO> lFineNodesPerDir,
-                                 const Array<GO> gFineNodesPerDir, const Array<GO> gIndices,
-                                 const Array<LO> lCoarseNodesPerDir,
-                                 const Array<bool> ghostInterface,
-                                 const Array<int> elementFlags,
-                                 const std::string stencilType, const std::string blockStrategy,
-                                 const Array<LO> elementNodesPerDir, const LO numNodesInElement,
-                                 const Array<GO> colGIDs,
-                                 Teuchos::SerialDenseMatrix<LO,SC>& Pi,
-                                 Teuchos::SerialDenseMatrix<LO,SC>& Pf,
-                                 Teuchos::SerialDenseMatrix<LO,SC>& Pe,
-                                 Array<LO>& dofType, Array<LO>& lDofInd) const {
-      MueLu::BlackBoxPFactory<SC,LO,GO,NO> mybbmgPFactory;
-      mybbmgPFactory.ComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
-                                         lCoarseElementsPerDir, numDimensions,
-                                         lFineNodesPerDir, gFineNodesPerDir, gIndices,
-                                         lCoarseNodesPerDir, ghostInterface, elementFlags,
-                                         stencilType, blockStrategy, elementNodesPerDir,
-                                         numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
-    };
-
-    void TestFormatStencil(const LO BlkSize, const Array<bool> ghostInterface, const LO ie,
-                           const LO je, const LO ke, const ArrayView<const SC> rowValues,
-                           const Array<LO> elementNodesPerDir, const int collapseFlags[3],
-                           const std::string stencilType, Array<SC>& stencil) const {
-      MueLu::BlackBoxPFactory<SC,LO,GO,Node> mybbmgPFactory;
-      mybbmgPFactory.FormatStencil(BlkSize, ghostInterface, ie, je, ke, rowValues,
-                                   elementNodesPerDir, collapseFlags, stencilType, stencil);
-    };
-
-    void TestGetNodeInfo(const LO ie, const LO je, const LO ke, const Array<LO> elementNodesPerDir,
-                         int* nodeType, LO& nodeIndex) const {
-      MueLu::BlackBoxPFactory<SC,LO,GO,Node> mybbmgPFactory;
-      int dummy;
-      mybbmgPFactory.GetNodeInfo(ie, je, ke, elementNodesPerDir, nodeType, nodeIndex, &dummy);
-    };
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor
+  BlackBoxPFactoryTester() {}
+
+  //! Destructor.
+  virtual ~BlackBoxPFactoryTester() {}
+  //@}
+
+  void TestGetGeometricData(Teuchos::RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> >& coordinates,
+                            Array<LO> coarseRate, Array<GO> gNodesPerDim, Array<LO> lNodesPerDim,
+                            LO BlkSize, Array<GO>& gIndices, Array<GO>& gCoarseNodesPerDir,
+                            Array<GO>& ghostGIDs, Array<GO>& coarseNodesGIDs, Array<GO>& colGIDs,
+                            GO& gNumCoarseNodes, Array<LO>& myOffset,
+                            Array<LO>& lCoarseNodesPerDir, Array<LO>& glCoarseNodesPerDir,
+                            Array<LO>& endRate, LO& lNumCoarseNodes, Array<bool>& ghostInterface,
+                            ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes, Array<int>& boundaryFlags) const {
+    // Call the method to be tested.
+    RCP<typename MueLu::BlackBoxPFactory<SC, LO, GO, Node>::NodesIDs> ghostedCoarseNodes = rcp(new typename MueLu::BlackBoxPFactory<SC, LO, GO, Node>::NodesIDs{});
+    MueLu::BlackBoxPFactory<SC, LO, GO, Node> mybbmgPFactory;
+    mybbmgPFactory.GetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, BlkSize,
+                                    gIndices, myOffset, ghostInterface, endRate,
+                                    gCoarseNodesPerDir, lCoarseNodesPerDir, glCoarseNodesPerDir,
+                                    ghostGIDs, coarseNodesGIDs, colGIDs, gNumCoarseNodes,
+                                    lNumCoarseNodes, coarseNodes, boundaryFlags,
+                                    ghostedCoarseNodes);
+  };
+
+  void TestComputeLocalEntries(const RCP<const Matrix>& Aghost, const Array<LO> coarseRate,
+                               const Array<LO> endRate, const LO BlkSize, const Array<LO> elemInds,
+                               const Array<LO> lCoarseElementsPerDir,
+                               const LO numDimensions, const Array<LO> lFineNodesPerDir,
+                               const Array<GO> gFineNodesPerDir, const Array<GO> gIndices,
+                               const Array<LO> lCoarseNodesPerDir,
+                               const Array<bool> ghostInterface,
+                               const Array<int> elementFlags,
+                               const std::string stencilType, const std::string blockStrategy,
+                               const Array<LO> elementNodesPerDir, const LO numNodesInElement,
+                               const Array<GO> colGIDs,
+                               Teuchos::SerialDenseMatrix<LO, SC>& Pi,
+                               Teuchos::SerialDenseMatrix<LO, SC>& Pf,
+                               Teuchos::SerialDenseMatrix<LO, SC>& Pe,
+                               Array<LO>& dofType, Array<LO>& lDofInd) const {
+    MueLu::BlackBoxPFactory<SC, LO, GO, NO> mybbmgPFactory;
+    mybbmgPFactory.ComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
+                                       lCoarseElementsPerDir, numDimensions,
+                                       lFineNodesPerDir, gFineNodesPerDir, gIndices,
+                                       lCoarseNodesPerDir, ghostInterface, elementFlags,
+                                       stencilType, blockStrategy, elementNodesPerDir,
+                                       numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
+  };
 
+  void TestFormatStencil(const LO BlkSize, const Array<bool> ghostInterface, const LO ie,
+                         const LO je, const LO ke, const ArrayView<const SC> rowValues,
+                         const Array<LO> elementNodesPerDir, const int collapseFlags[3],
+                         const std::string stencilType, Array<SC>& stencil) const {
+    MueLu::BlackBoxPFactory<SC, LO, GO, Node> mybbmgPFactory;
+    mybbmgPFactory.FormatStencil(BlkSize, ghostInterface, ie, je, ke, rowValues,
+                                 elementNodesPerDir, collapseFlags, stencilType, stencil);
   };
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib,
-                      const LocalOrdinal numDimensions, const std::string stencilType,
-                      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
-                      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >&Coordinates,
-                      RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
-                      Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
+  void TestGetNodeInfo(const LO ie, const LO je, const LO ke, const Array<LO> elementNodesPerDir,
+                       int* nodeType, LO& nodeIndex) const {
+    MueLu::BlackBoxPFactory<SC, LO, GO, Node> mybbmgPFactory;
+    int dummy;
+    mybbmgPFactory.GetNodeInfo(ie, je, ke, elementNodesPerDir, nodeType, nodeIndex, &dummy);
+  };
+};
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib,
+                    const LocalOrdinal numDimensions, const std::string stencilType,
+                    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
+                    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& Coordinates,
+                    RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
+                    Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
 #include "MueLu_UseShortNames.hpp"
 
-    GO nx = 7;
-    GO ny = 7;
-    GO nz = (numDimensions < 3 ? 1 : 7);
-    GO gNumPoints = nx*ny*nz;
-
-    gNodesPerDim[0] = nx;
-    gNodesPerDim[1] = ny;
-    gNodesPerDim[2] = nz;
-
-    GO myOffset = 0;
-    if(comm->getSize() == 1) {
-      myOffset = 0;
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(numDimensions == 2) {
-          myOffset = 0;
-          lNodesPerDim[0] = 4;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 0;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 4;
-        }
-      } else if(comm->getRank() == 1) {
-        if(numDimensions == 2) {
-          myOffset = 4;
-          lNodesPerDim[0] = 3;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 28;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 4;
-        }
-      } else if(comm->getRank() == 2) {
-        if(numDimensions == 2) {
-          myOffset = 28;
-          lNodesPerDim[0] = 4;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 196;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 3;
-        }
-      } else if(comm->getRank() == 3) {
-        if(numDimensions == 2) {
-          myOffset = 32;
-          lNodesPerDim[0] = 3;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 224;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 3;
-        }
+  GO nx         = 7;
+  GO ny         = 7;
+  GO nz         = (numDimensions < 3 ? 1 : 7);
+  GO gNumPoints = nx * ny * nz;
+
+  gNodesPerDim[0] = nx;
+  gNodesPerDim[1] = ny;
+  gNodesPerDim[2] = nz;
+
+  GO myOffset = 0;
+  if (comm->getSize() == 1) {
+    myOffset        = 0;
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (numDimensions == 2) {
+        myOffset        = 0;
+        lNodesPerDim[0] = 4;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 0;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 4;
+      }
+    } else if (comm->getRank() == 1) {
+      if (numDimensions == 2) {
+        myOffset        = 4;
+        lNodesPerDim[0] = 3;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 28;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 4;
+      }
+    } else if (comm->getRank() == 2) {
+      if (numDimensions == 2) {
+        myOffset        = 28;
+        lNodesPerDim[0] = 4;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 196;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 3;
+      }
+    } else if (comm->getRank() == 3) {
+      if (numDimensions == 2) {
+        myOffset        = 32;
+        lNodesPerDim[0] = 3;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 224;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 3;
       }
     }
+  }
 
-    GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
-    if(numDimensions == 2) {
-      myZoffset = 0;
-      myYoffset = myOffset / gNodesPerDim[0];
-      myXoffset = myOffset % gNodesPerDim[0];
-    } else if(numDimensions == 3) {
-      myZoffset = myOffset / (gNodesPerDim[1]*gNodesPerDim[0]);
-      myYoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) / gNodesPerDim[0];
-      myXoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) % gNodesPerDim[0];
-    }
-    LO lNumPoints = lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2];
-
-    // Construct map and local coordinates
-    Teuchos::Array<GO>     myGIDs(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
-    for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          myGIDs[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = myOffset
-            + k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0] + i;
-          myXCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+  GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
+  if (numDimensions == 2) {
+    myZoffset = 0;
+    myYoffset = myOffset / gNodesPerDim[0];
+    myXoffset = myOffset % gNodesPerDim[0];
+  } else if (numDimensions == 3) {
+    myZoffset = myOffset / (gNodesPerDim[1] * gNodesPerDim[0]);
+    myYoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) / gNodesPerDim[0];
+    myXoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) % gNodesPerDim[0];
+  }
+  LO lNumPoints = lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2];
+
+  // Construct map and local coordinates
+  Teuchos::Array<GO> myGIDs(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
+  for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+    for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        myGIDs[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = myOffset + k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + i;
+        myXCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (i + myXoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
-          myYCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+        myYCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (j + myYoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
-          myZCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+        myZCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (k + myZoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
-        }
       }
     }
+  }
 
-    Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
-    if(numDimensions == 1) {
-      myCoordinates[0] = myXCoords();
-    } else if(numDimensions == 2) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-    } else if(numDimensions == 3) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-      myCoordinates[2] = myZCoords();
-    }
+  Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
+  if (numDimensions == 1) {
+    myCoordinates[0] = myXCoords();
+  } else if (numDimensions == 2) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+  } else if (numDimensions == 3) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+    myCoordinates[2] = myZCoords();
+  }
 
-    // Create the map and store coordinates using the above array views
-    map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
-    Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map, myCoordinates(),
-                                                                     numDimensions);
-
-    // small parameter list for Galeri
-    Teuchos::ParameterList problemList;
-    if(numDimensions == 1) {
-      problemList.set("nx",gNodesPerDim[0]);
-    } else if(numDimensions == 2) {
-      problemList.set("nx",gNodesPerDim[0]);
-      problemList.set("ny",gNodesPerDim[1]);
-    } else if(numDimensions == 3) {
-      problemList.set("nx",gNodesPerDim[0]);
-      problemList.set("ny",gNodesPerDim[1]);
-      problemList.set("nz",gNodesPerDim[2]);
+  // Create the map and store coordinates using the above array views
+  map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
+  Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, myCoordinates(),
+                                                                                                                     numDimensions);
+
+  // small parameter list for Galeri
+  Teuchos::ParameterList problemList;
+  if (numDimensions == 1) {
+    problemList.set("nx", gNodesPerDim[0]);
+  } else if (numDimensions == 2) {
+    problemList.set("nx", gNodesPerDim[0]);
+    problemList.set("ny", gNodesPerDim[1]);
+  } else if (numDimensions == 3) {
+    problemList.set("nx", gNodesPerDim[0]);
+    problemList.set("ny", gNodesPerDim[1]);
+    problemList.set("nz", gNodesPerDim[2]);
+  }
+  problemList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  if (stencilType == "reduced") {
+    if (numDimensions == 1) {
+      Galeri::Xpetra::
+          Laplace1DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
+    } else if (numDimensions == 2) {
+      Galeri::Xpetra::
+          Laplace2DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
+    } else if (numDimensions == 3) {
+      Galeri::Xpetra::
+          Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
     }
-    problemList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    if(stencilType == "reduced") {
-      if(numDimensions == 1) {
-        Galeri::Xpetra::
-          Laplace1DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      } else if(numDimensions == 2) {
-        Galeri::Xpetra::
-          Laplace2DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      } else if(numDimensions == 3) {
-        Galeri::Xpetra::
-          Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      }
-    } else if(stencilType == "full") {
-      if(numDimensions == 1) {
-        Galeri::Xpetra::
-          Laplace1DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      } else if(numDimensions == 2) {
-        Galeri::Xpetra::
-          Star2DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      } else if(numDimensions == 3) {
-        Galeri::Xpetra::
-          Brick3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList, map);
-        Op = Problem.BuildMatrix();
-      }
+  } else if (stencilType == "full") {
+    if (numDimensions == 1) {
+      Galeri::Xpetra::
+          Laplace1DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
+    } else if (numDimensions == 2) {
+      Galeri::Xpetra::
+          Star2DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
+    } else if (numDimensions == 3) {
+      Galeri::Xpetra::
+          Brick3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>
+              Problem(problemList, map);
+      Op = Problem.BuildMatrix();
     }
-
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, Constructor, Scalar, LocalOrdinal,
-                                    GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, Constructor, Scalar, LocalOrdinal,
+                                  GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<BlackBoxPFactory> bbmgPFact = rcp(new BlackBoxPFactory);
-    TEST_EQUALITY(bbmgPFact != Teuchos::null, true);
+  RCP<BlackBoxPFactory> bbmgPFact = rcp(new BlackBoxPFactory);
+  TEST_EQUALITY(bbmgPFact != Teuchos::null, true);
 
-  } //Constructor
+}  // Constructor
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, BlackBoxGhosts, Scalar, LocalOrdinal,
-                                    GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    const LO numDimensions = 2;
-    const std::string stencilType = "reduced";
-
-    RCP<Matrix> Op;
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates;
-    RCP<Map> map;
-    Array<GO> gNodesPerDim(3);
-    Array<LO> lNodesPerDim(3);
-    GetProblemData<SC,LO,GO,NO>(comm, lib, numDimensions, stencilType, Op, coordinates, map,
-                                gNodesPerDim, lNodesPerDim);
-
-    Array<LO> coarseRate(3);
-    coarseRate[0] = 2;
-    coarseRate[1] = 2;
-    if(numDimensions == 2) {
-      coarseRate[2] = 1;
-    } else if(numDimensions == 3) {
-      coarseRate[2] = 2;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, BlackBoxGhosts, Scalar, LocalOrdinal,
+                                  GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  const LO numDimensions        = 2;
+  const std::string stencilType = "reduced";
+
+  RCP<Matrix> Op;
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates;
+  RCP<Map> map;
+  Array<GO> gNodesPerDim(3);
+  Array<LO> lNodesPerDim(3);
+  GetProblemData<SC, LO, GO, NO>(comm, lib, numDimensions, stencilType, Op, coordinates, map,
+                                 gNodesPerDim, lNodesPerDim);
+
+  Array<LO> coarseRate(3);
+  coarseRate[0] = 2;
+  coarseRate[1] = 2;
+  if (numDimensions == 2) {
+    coarseRate[2] = 1;
+  } else if (numDimensions == 3) {
+    coarseRate[2] = 2;
+  }
 
-    BlackBoxPFactoryTester<SC,LO,GO,Node> factTester;
-    Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
-    Array<LO> myOffset(3), lCoarseNodesPerDir(3), glCoarseNodesPerDir(3), endRate(3);
-    Array<bool> ghostInterface(6);
-    Array<int> boundaryFlags(3);
-    Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
-    ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
-    GO gNumCoarseNodes = 0;
-    LO lNumCoarseNodes = 0;
-    for(LO dim = 0; dim < numDimensions; ++dim) {
-      fineNodes[dim] = coordinates->getData(dim)();
+  BlackBoxPFactoryTester<SC, LO, GO, Node> factTester;
+  Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
+  Array<LO> myOffset(3), lCoarseNodesPerDir(3), glCoarseNodesPerDir(3), endRate(3);
+  Array<bool> ghostInterface(6);
+  Array<int> boundaryFlags(3);
+  Array<ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > fineNodes(numDimensions);
+  ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
+  GO gNumCoarseNodes = 0;
+  LO lNumCoarseNodes = 0;
+  for (LO dim = 0; dim < numDimensions; ++dim) {
+    fineNodes[dim] = coordinates->getData(dim)();
+  }
+
+  factTester.TestGetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, 1,
+                                  gIndices, gCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs,
+                                  colGIDs, gNumCoarseNodes, myOffset, lCoarseNodesPerDir,
+                                  glCoarseNodesPerDir, endRate, lNumCoarseNodes, ghostInterface,
+                                  coarseNodes, boundaryFlags);
+
+  Array<GO> ghostGIDs_check, coarseNodesGIDs_check, colGIDs_check;
+  if (map->getComm()->getSize() == 1) {
+    coarseNodesGIDs_check.resize(16);
+    for (LO i = 0; i < 16; ++i) {
+      coarseNodesGIDs_check[i] = i;
     }
+    colGIDs_check = coarseNodesGIDs_check;
+    ghostGIDs_check.resize(0);
+  } else if (map->getComm()->getSize() == 4) {
+    coarseNodesGIDs_check.resize(4);
+    if (map->getComm()->getRank() == 0) {
+      coarseNodesGIDs_check[0] = 0;
+      coarseNodesGIDs_check[1] = 1;
+      coarseNodesGIDs_check[2] = 4;
+      coarseNodesGIDs_check[3] = 5;
+
+      ghostGIDs_check.resize(5);
+      ghostGIDs_check[0] = 4;
+      ghostGIDs_check[1] = 18;
+      ghostGIDs_check[2] = 28;
+      ghostGIDs_check[3] = 30;
+      ghostGIDs_check[4] = 32;
+
+      colGIDs_check.resize(9);
+      colGIDs_check[0] = 0;
+      colGIDs_check[1] = 1;
+      colGIDs_check[2] = 4;
+      colGIDs_check[3] = 5;
+      colGIDs_check[4] = 2;
+      colGIDs_check[5] = 6;
+      colGIDs_check[6] = 8;
+      colGIDs_check[7] = 9;
+      colGIDs_check[8] = 10;
+    } else if (map->getComm()->getRank() == 1) {
+      coarseNodesGIDs_check[0] = 2;
+      coarseNodesGIDs_check[1] = 3;
+      coarseNodesGIDs_check[2] = 6;
+      coarseNodesGIDs_check[3] = 7;
+
+      ghostGIDs_check.resize(2);
+      ghostGIDs_check[0] = 32;
+      ghostGIDs_check[1] = 34;
+
+      colGIDs_check.resize(6);
+      colGIDs_check[0] = 2;
+      colGIDs_check[1] = 3;
+      colGIDs_check[2] = 6;
+      colGIDs_check[3] = 7;
+      colGIDs_check[4] = 10;
+      colGIDs_check[5] = 11;
+    } else if (map->getComm()->getRank() == 2) {
+      coarseNodesGIDs_check[0] = 8;
+      coarseNodesGIDs_check[1] = 9;
+      coarseNodesGIDs_check[2] = 12;
+      coarseNodesGIDs_check[3] = 13;
+
+      ghostGIDs_check.resize(2);
+      ghostGIDs_check[0] = 32;
+      ghostGIDs_check[1] = 46;
+
+      colGIDs_check.resize(6);
+      colGIDs_check[0] = 8;
+      colGIDs_check[1] = 9;
+      colGIDs_check[2] = 12;
+      colGIDs_check[3] = 13;
+      colGIDs_check[4] = 10;
+      colGIDs_check[5] = 14;
+    } else if (map->getComm()->getRank() == 3) {
+      coarseNodesGIDs_check[0] = 10;
+      coarseNodesGIDs_check[1] = 11;
+      coarseNodesGIDs_check[2] = 14;
+      coarseNodesGIDs_check[3] = 15;
 
-    factTester.TestGetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, 1,
-                                    gIndices, gCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs,
-                                    colGIDs, gNumCoarseNodes, myOffset, lCoarseNodesPerDir,
-                                    glCoarseNodesPerDir, endRate, lNumCoarseNodes, ghostInterface,
-                                    coarseNodes, boundaryFlags);
-
-    Array<GO> ghostGIDs_check, coarseNodesGIDs_check, colGIDs_check;
-    if(map->getComm()->getSize() == 1) {
-      coarseNodesGIDs_check.resize(16);
-      for(LO i = 0; i < 16; ++i) {
-        coarseNodesGIDs_check[i] = i;
-      }
-      colGIDs_check = coarseNodesGIDs_check;
       ghostGIDs_check.resize(0);
-    } else if(map->getComm()->getSize() == 4) {
-      coarseNodesGIDs_check.resize(4);
-      if(map->getComm()->getRank() == 0) {
-        coarseNodesGIDs_check[0] = 0;
-        coarseNodesGIDs_check[1] = 1;
-        coarseNodesGIDs_check[2] = 4;
-        coarseNodesGIDs_check[3] = 5;
-
-        ghostGIDs_check.resize(5);
-        ghostGIDs_check[0] = 4;
-        ghostGIDs_check[1] = 18;
-        ghostGIDs_check[2] = 28;
-        ghostGIDs_check[3] = 30;
-        ghostGIDs_check[4] = 32;
-
-        colGIDs_check.resize(9);
-        colGIDs_check[0] = 0;
-        colGIDs_check[1] = 1;
-        colGIDs_check[2] = 4;
-        colGIDs_check[3] = 5;
-        colGIDs_check[4] = 2;
-        colGIDs_check[5] = 6;
-        colGIDs_check[6] = 8;
-        colGIDs_check[7] = 9;
-        colGIDs_check[8] = 10;
-      } else if(map->getComm()->getRank() == 1) {
-        coarseNodesGIDs_check[0] = 2;
-        coarseNodesGIDs_check[1] = 3;
-        coarseNodesGIDs_check[2] = 6;
-        coarseNodesGIDs_check[3] = 7;
-
-        ghostGIDs_check.resize(2);
-        ghostGIDs_check[0] = 32;
-        ghostGIDs_check[1] = 34;
-
-        colGIDs_check.resize(6);
-        colGIDs_check[0] = 2;
-        colGIDs_check[1] = 3;
-        colGIDs_check[2] = 6;
-        colGIDs_check[3] = 7;
-        colGIDs_check[4] = 10;
-        colGIDs_check[5] = 11;
-      } else if(map->getComm()->getRank() == 2) {
-        coarseNodesGIDs_check[0] = 8;
-        coarseNodesGIDs_check[1] = 9;
-        coarseNodesGIDs_check[2] = 12;
-        coarseNodesGIDs_check[3] = 13;
-
-        ghostGIDs_check.resize(2);
-        ghostGIDs_check[0] = 32;
-        ghostGIDs_check[1] = 46;
-
-        colGIDs_check.resize(6);
-        colGIDs_check[0] = 8;
-        colGIDs_check[1] = 9;
-        colGIDs_check[2] = 12;
-        colGIDs_check[3] = 13;
-        colGIDs_check[4] = 10;
-        colGIDs_check[5] = 14;
-      } else if(map->getComm()->getRank() == 3) {
-        coarseNodesGIDs_check[0] = 10;
-        coarseNodesGIDs_check[1] = 11;
-        coarseNodesGIDs_check[2] = 14;
-        coarseNodesGIDs_check[3] = 15;
-
-        ghostGIDs_check.resize(0);
-
-        colGIDs_check.resize(4);
-        colGIDs_check[0] = 10;
-        colGIDs_check[1] = 11;
-        colGIDs_check[2] = 14;
-        colGIDs_check[3] = 15;
-      }
+
+      colGIDs_check.resize(4);
+      colGIDs_check[0] = 10;
+      colGIDs_check[1] = 11;
+      colGIDs_check[2] = 14;
+      colGIDs_check[3] = 15;
     }
-    TEST_EQUALITY(coarseNodesGIDs_check == coarseNodesGIDs, true);
-    TEST_EQUALITY(ghostGIDs_check == ghostGIDs, true);
-    TEST_EQUALITY(colGIDs_check == colGIDs, true);
+  }
+  TEST_EQUALITY(coarseNodesGIDs_check == coarseNodesGIDs, true);
+  TEST_EQUALITY(ghostGIDs_check == ghostGIDs, true);
+  TEST_EQUALITY(colGIDs_check == colGIDs, true);
 
-  } //BlackBoxGhosts
+}  // BlackBoxGhosts
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, GetNodeInfo, Scalar, LocalOrdinal,
-                                    GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    // Creater tester factory
-    BlackBoxPFactoryTester<SC,LO,GO,Node> factTester;
-
-    // Create coarse element data
-    Teuchos::Array<LO> elementNodesPerDir(3);
-    elementNodesPerDir[0] = 4;
-    elementNodesPerDir[1] = 4;
-    elementNodesPerDir[2] = 3;
-
-    // Instantiate outputs
-    int nodeType = 0;
-    LO  nodeIndex = 0;
-
-    // Create check variables
-    bool checkResult = true;
-    std::vector<int> nodeTypes = {0, 1, 1, 0, 1, 2, 2, 1, 1, 2, 2, 1, 0, 1, 1, 0,
-                                  1, 2, 2, 1, 2, 3, 3, 2, 2, 3, 3, 2, 1, 2, 2, 1,
-                                  0, 1, 1, 0, 1, 2, 2, 1, 1, 2, 2, 1, 0, 1, 1, 0};
-    std::vector<LO>  nodeIndices = {0, 0, 1, 1, 2, 0, 1, 3, 4, 2, 3, 5, 2, 6, 7, 3,
-                                    8, 4, 5, 9, 6, 0, 1, 7, 8, 2, 3, 9, 10, 10, 11, 11,
-                                    4, 12, 13, 5, 14, 12, 13, 15, 16, 14, 15, 17, 6, 18, 19, 7};
-
-    LO currentIndex;
-    for(LO k = 0; k < elementNodesPerDir[2]; ++k) {
-      for(LO j = 0; j < elementNodesPerDir[1]; ++j) {
-        for(LO i = 0; i < elementNodesPerDir[0]; ++i) {
-          currentIndex = k*elementNodesPerDir[1]*elementNodesPerDir[0] + j*elementNodesPerDir[0] +i;
-          factTester.TestGetNodeInfo(i, j, k, elementNodesPerDir, &nodeType, nodeIndex);
-          if((nodeTypes[currentIndex] != nodeType) || (nodeIndices[currentIndex] != nodeIndex)) {
-            checkResult = false;
-          }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, GetNodeInfo, Scalar, LocalOrdinal,
+                                  GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  // Creater tester factory
+  BlackBoxPFactoryTester<SC, LO, GO, Node> factTester;
+
+  // Create coarse element data
+  Teuchos::Array<LO> elementNodesPerDir(3);
+  elementNodesPerDir[0] = 4;
+  elementNodesPerDir[1] = 4;
+  elementNodesPerDir[2] = 3;
+
+  // Instantiate outputs
+  int nodeType = 0;
+  LO nodeIndex = 0;
+
+  // Create check variables
+  bool checkResult            = true;
+  std::vector<int> nodeTypes  = {0, 1, 1, 0, 1, 2, 2, 1, 1, 2, 2, 1, 0, 1, 1, 0,
+                                 1, 2, 2, 1, 2, 3, 3, 2, 2, 3, 3, 2, 1, 2, 2, 1,
+                                 0, 1, 1, 0, 1, 2, 2, 1, 1, 2, 2, 1, 0, 1, 1, 0};
+  std::vector<LO> nodeIndices = {0, 0, 1, 1, 2, 0, 1, 3, 4, 2, 3, 5, 2, 6, 7, 3,
+                                 8, 4, 5, 9, 6, 0, 1, 7, 8, 2, 3, 9, 10, 10, 11, 11,
+                                 4, 12, 13, 5, 14, 12, 13, 15, 16, 14, 15, 17, 6, 18, 19, 7};
+
+  LO currentIndex;
+  for (LO k = 0; k < elementNodesPerDir[2]; ++k) {
+    for (LO j = 0; j < elementNodesPerDir[1]; ++j) {
+      for (LO i = 0; i < elementNodesPerDir[0]; ++i) {
+        currentIndex = k * elementNodesPerDir[1] * elementNodesPerDir[0] + j * elementNodesPerDir[0] + i;
+        factTester.TestGetNodeInfo(i, j, k, elementNodesPerDir, &nodeType, nodeIndex);
+        if ((nodeTypes[currentIndex] != nodeType) || (nodeIndices[currentIndex] != nodeIndex)) {
+          checkResult = false;
         }
       }
     }
-    TEST_EQUALITY(checkResult, true);
   }
+  TEST_EQUALITY(checkResult, true);
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, Prolongator, Scalar, LocalOrdinal,
-                                    GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    const LO numDimensions = 3;
-    const std::string stencilType = "reduced";
-
-    RCP<Matrix> A;
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates;
-    RCP<Map> map;
-    Array<GO> gNodesPerDim(3);
-    Array<LO> lNodesPerDim(3);
-    GetProblemData(comm, lib, numDimensions, stencilType, A, coordinates, map, gNodesPerDim,
-                   lNodesPerDim);
-
-    // Creater tester factory
-    BlackBoxPFactoryTester<SC,LO,GO,Node> factTester;
-
-    LO BlkSize = 1;
-    Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
-    Array<LO> myOffset(3), coarseRate(3), endRate(3), lCoarseNodesPerDir(3);
-    Array<LO> glCoarseNodesPerDir(3), lCoarseElementsPerDir(3);
-    Array<bool> ghostInterface(6);
-    Array<int> boundaryFlags(3);
-    ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
-    for(int i = 0; i < numDimensions; ++i) {
-      coarseRate[i] = 2;
-    }
-    GO gNumCoarseNodes = 0;
-    LO lNumCoarseNodes = 0;
-    factTester.TestGetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, 1,
-                                    gIndices, gCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs,
-                                    colGIDs, gNumCoarseNodes, myOffset, lCoarseNodesPerDir,
-                                    glCoarseNodesPerDir, endRate, lNumCoarseNodes, ghostInterface,
-                                    coarseNodes, boundaryFlags);
-
-    // From A on local rank, let us construct Aghost
-    Array<GO> ghostRowGIDs, ghostColGIDs, nodeSteps(3);
-    nodeSteps[0] = 1;
-    nodeSteps[1] = gNodesPerDim[0];
-    nodeSteps[2] = gNodesPerDim[0]*gNodesPerDim[1];
-    Array<LO> range(3);
-    GO startingGID = A->getRowMap()->getMinGlobalIndex();
-    for(LO dim = 0; dim < 3; ++dim) {
-      LO numCoarseNodes = 0;
-      if(dim < numDimensions) {
-        startingGID -= myOffset[dim]*nodeSteps[dim];
-        numCoarseNodes = lCoarseNodesPerDir[dim];
-        if(ghostInterface[2*dim]) {++numCoarseNodes;}
-        if(ghostInterface[2*dim+1]) {++numCoarseNodes;}
-        if(gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim]) {
-          range[dim] = (numCoarseNodes - 2)*coarseRate[dim] + endRate[dim] + 1;
-        } else {
-          range[dim] = (numCoarseNodes - 1)*coarseRate[dim] + 1;
-        }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, Prolongator, Scalar, LocalOrdinal,
+                                  GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  const LO numDimensions        = 3;
+  const std::string stencilType = "reduced";
+
+  RCP<Matrix> A;
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates;
+  RCP<Map> map;
+  Array<GO> gNodesPerDim(3);
+  Array<LO> lNodesPerDim(3);
+  GetProblemData(comm, lib, numDimensions, stencilType, A, coordinates, map, gNodesPerDim,
+                 lNodesPerDim);
+
+  // Creater tester factory
+  BlackBoxPFactoryTester<SC, LO, GO, Node> factTester;
+
+  LO BlkSize = 1;
+  Array<GO> gIndices(3), gCoarseNodesPerDir(3), ghostGIDs, coarseNodesGIDs, colGIDs;
+  Array<LO> myOffset(3), coarseRate(3), endRate(3), lCoarseNodesPerDir(3);
+  Array<LO> glCoarseNodesPerDir(3), lCoarseElementsPerDir(3);
+  Array<bool> ghostInterface(6);
+  Array<int> boundaryFlags(3);
+  ArrayRCP<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coarseNodes(numDimensions);
+  for (int i = 0; i < numDimensions; ++i) {
+    coarseRate[i] = 2;
+  }
+  GO gNumCoarseNodes = 0;
+  LO lNumCoarseNodes = 0;
+  factTester.TestGetGeometricData(coordinates, coarseRate, gNodesPerDim, lNodesPerDim, 1,
+                                  gIndices, gCoarseNodesPerDir, ghostGIDs, coarseNodesGIDs,
+                                  colGIDs, gNumCoarseNodes, myOffset, lCoarseNodesPerDir,
+                                  glCoarseNodesPerDir, endRate, lNumCoarseNodes, ghostInterface,
+                                  coarseNodes, boundaryFlags);
+
+  // From A on local rank, let us construct Aghost
+  Array<GO> ghostRowGIDs, ghostColGIDs, nodeSteps(3);
+  nodeSteps[0] = 1;
+  nodeSteps[1] = gNodesPerDim[0];
+  nodeSteps[2] = gNodesPerDim[0] * gNodesPerDim[1];
+  Array<LO> range(3);
+  GO startingGID = A->getRowMap()->getMinGlobalIndex();
+  for (LO dim = 0; dim < 3; ++dim) {
+    LO numCoarseNodes = 0;
+    if (dim < numDimensions) {
+      startingGID -= myOffset[dim] * nodeSteps[dim];
+      numCoarseNodes = lCoarseNodesPerDir[dim];
+      if (ghostInterface[2 * dim]) {
+        ++numCoarseNodes;
+      }
+      if (ghostInterface[2 * dim + 1]) {
+        ++numCoarseNodes;
+      }
+      if (gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim]) {
+        range[dim] = (numCoarseNodes - 2) * coarseRate[dim] + endRate[dim] + 1;
       } else {
-        range[dim] = 1;
+        range[dim] = (numCoarseNodes - 1) * coarseRate[dim] + 1;
       }
+    } else {
+      range[dim] = 1;
     }
-    ghostRowGIDs.resize(range[0]*range[1]*range[2]*BlkSize);
-    for(LO k = 0; k < range[2]; ++k) {
-      for(LO j = 0; j < range[1]; ++j) {
-        for(LO i = 0; i < range[0]; ++i) {
-          for(LO l = 0; l < BlkSize; ++l) {
-            ghostRowGIDs[(k*range[1]*range[0] + j*range[0] + i)*BlkSize + l] = startingGID
-              + (k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0] + i)*BlkSize + l;
-          }
+  }
+  ghostRowGIDs.resize(range[0] * range[1] * range[2] * BlkSize);
+  for (LO k = 0; k < range[2]; ++k) {
+    for (LO j = 0; j < range[1]; ++j) {
+      for (LO i = 0; i < range[0]; ++i) {
+        for (LO l = 0; l < BlkSize; ++l) {
+          ghostRowGIDs[(k * range[1] * range[0] + j * range[0] + i) * BlkSize + l] = startingGID + (k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + i) * BlkSize + l;
         }
       }
     }
+  }
 
-    // Looking at the above loops it is easy to find startingGID for the ghostColGIDs
-    Array<GO> startingGlobalIndices(numDimensions), dimStride(numDimensions);
-    Array<GO> startingColIndices(numDimensions), finishingColIndices(numDimensions);
-    GO colMinGID = 0;
-    Array<LO> colRange(numDimensions);
-    dimStride[0] = 1;
-    for(int dim = 1; dim < numDimensions; ++dim) {
-      dimStride[dim] = dimStride[dim - 1]*gNodesPerDim[dim - 1];
-    }
-    {
-      GO tmp = startingGID;
-      for(int dim = numDimensions; dim > 0; --dim) {
-        startingGlobalIndices[dim - 1] = tmp / dimStride[dim - 1];
-        tmp = tmp % dimStride[dim - 1];
-
-        if(startingGlobalIndices[dim - 1] > 0) {
-          startingColIndices[dim - 1] = startingGlobalIndices[dim - 1] - 1;
-        }
-        if(startingGlobalIndices[dim - 1] + range[dim - 1] < gNodesPerDim[dim - 1]) {
-          finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + range[dim - 1];
-        } else {
-          finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + range[dim - 1] - 1;
-        }
-        colRange[dim - 1] = finishingColIndices[dim - 1] - startingColIndices[dim - 1] + 1;
-        colMinGID += startingColIndices[dim - 1]*dimStride[dim - 1];
+  // Looking at the above loops it is easy to find startingGID for the ghostColGIDs
+  Array<GO> startingGlobalIndices(numDimensions), dimStride(numDimensions);
+  Array<GO> startingColIndices(numDimensions), finishingColIndices(numDimensions);
+  GO colMinGID = 0;
+  Array<LO> colRange(numDimensions);
+  dimStride[0] = 1;
+  for (int dim = 1; dim < numDimensions; ++dim) {
+    dimStride[dim] = dimStride[dim - 1] * gNodesPerDim[dim - 1];
+  }
+  {
+    GO tmp = startingGID;
+    for (int dim = numDimensions; dim > 0; --dim) {
+      startingGlobalIndices[dim - 1] = tmp / dimStride[dim - 1];
+      tmp                            = tmp % dimStride[dim - 1];
+
+      if (startingGlobalIndices[dim - 1] > 0) {
+        startingColIndices[dim - 1] = startingGlobalIndices[dim - 1] - 1;
       }
+      if (startingGlobalIndices[dim - 1] + range[dim - 1] < gNodesPerDim[dim - 1]) {
+        finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + range[dim - 1];
+      } else {
+        finishingColIndices[dim - 1] = startingGlobalIndices[dim - 1] + range[dim - 1] - 1;
+      }
+      colRange[dim - 1] = finishingColIndices[dim - 1] - startingColIndices[dim - 1] + 1;
+      colMinGID += startingColIndices[dim - 1] * dimStride[dim - 1];
     }
-    ghostColGIDs.resize(colRange[0]*colRange[1]*colRange[2]*BlkSize);
-    for(LO k = 0; k < colRange[2]; ++k) {
-      for(LO j = 0; j < colRange[1]; ++j) {
-        for(LO i = 0; i < colRange[0]; ++i) {
-          for(LO l = 0; l < BlkSize; ++l) {
-            ghostColGIDs[(k*colRange[1]*colRange[0] + j*colRange[0] + i)*BlkSize + l] = colMinGID
-              + (k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0] + i)*BlkSize + l;
-          }
+  }
+  ghostColGIDs.resize(colRange[0] * colRange[1] * colRange[2] * BlkSize);
+  for (LO k = 0; k < colRange[2]; ++k) {
+    for (LO j = 0; j < colRange[1]; ++j) {
+      for (LO i = 0; i < colRange[0]; ++i) {
+        for (LO l = 0; l < BlkSize; ++l) {
+          ghostColGIDs[(k * colRange[1] * colRange[0] + j * colRange[0] + i) * BlkSize + l] = colMinGID + (k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + i) * BlkSize + l;
         }
       }
     }
+  }
 
-    RCP<const Map> ghostedRowMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getRowMap()->lib(),
-                                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                           ghostRowGIDs(),
-                                           A->getRowMap()->getIndexBase(),
-                                           A->getRowMap()->getComm());
-    RCP<const Map> ghostedColMap = Xpetra::MapFactory<LO,GO,NO>::Build(A->getRowMap()->lib(),
-                                           Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
-                                           ghostColGIDs(),
-                                           A->getRowMap()->getIndexBase(),
-                                           A->getRowMap()->getComm());
-    RCP<const Import> ghostImporter = Xpetra::ImportFactory<LO,GO,NO>::Build(A->getRowMap(),
+  RCP<const Map> ghostedRowMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getRowMap()->lib(),
+                                                                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                                          ghostRowGIDs(),
+                                                                          A->getRowMap()->getIndexBase(),
+                                                                          A->getRowMap()->getComm());
+  RCP<const Map> ghostedColMap    = Xpetra::MapFactory<LO, GO, NO>::Build(A->getRowMap()->lib(),
+                                                                          Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid(),
+                                                                          ghostColGIDs(),
+                                                                          A->getRowMap()->getIndexBase(),
+                                                                          A->getRowMap()->getComm());
+  RCP<const Import> ghostImporter = Xpetra::ImportFactory<LO, GO, NO>::Build(A->getRowMap(),
                                                                              ghostedRowMap);
-    RCP<const Matrix> Aghost        = Xpetra::MatrixFactory<SC,LO,GO,NO>::Build(A, *ghostImporter,
-                                                                                ghostedRowMap,
-                                                                                ghostedColMap);
-
-    if(comm->getRank() == 0) {
-      // Pick on which coarse element the algorithm is tested
-      Array<LO> elemInds(3), elementNodesPerDir(3);
-
-      elemInds[0] = 1;
-      elemInds[1] = 1;
-      elemInds[2] = 1;
-
-      Array<int> elementFlags(3);
-      for(int dim = 0; dim < 3; ++dim) {
-        if(elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
-          elementFlags[dim] = boundaryFlags[dim];
-        } else if(elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
-          elementFlags[dim] += 1;
-        } else if((elemInds[dim] == lCoarseElementsPerDir[dim] - 1)
-                  && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
-          elementFlags[dim] += 2;
-        } else {
-          elementFlags[dim] = 0;
-        }
+  RCP<const Matrix> Aghost        = Xpetra::MatrixFactory<SC, LO, GO, NO>::Build(A, *ghostImporter,
+                                                                                 ghostedRowMap,
+                                                                                 ghostedColMap);
+
+  if (comm->getRank() == 0) {
+    // Pick on which coarse element the algorithm is tested
+    Array<LO> elemInds(3), elementNodesPerDir(3);
+
+    elemInds[0] = 1;
+    elemInds[1] = 1;
+    elemInds[2] = 1;
+
+    Array<int> elementFlags(3);
+    for (int dim = 0; dim < 3; ++dim) {
+      if (elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
+        elementFlags[dim] = boundaryFlags[dim];
+      } else if (elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
+        elementFlags[dim] += 1;
+      } else if ((elemInds[dim] == lCoarseElementsPerDir[dim] - 1) && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
+        elementFlags[dim] += 2;
+      } else {
+        elementFlags[dim] = 0;
+      }
 
-        // Compute the number of nodes in the current element.
-        if(dim < numDimensions) {
-          if((elemInds[dim] == lCoarseElementsPerDir[dim])
-             && (gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim])) {
-            elementNodesPerDir[dim] = endRate[dim] + 1;
-          } else {
-            elementNodesPerDir[dim] = coarseRate[dim] + 1;
-          }
+      // Compute the number of nodes in the current element.
+      if (dim < numDimensions) {
+        if ((elemInds[dim] == lCoarseElementsPerDir[dim]) && (gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim])) {
+          elementNodesPerDir[dim] = endRate[dim] + 1;
         } else {
-          elementNodesPerDir[dim] = 1;
+          elementNodesPerDir[dim] = coarseRate[dim] + 1;
         }
+      } else {
+        elementNodesPerDir[dim] = 1;
+      }
+    }
+    LO numNodesInElement = elementNodesPerDir[0] * elementNodesPerDir[1] * elementNodesPerDir[2];
+
+    Teuchos::SerialDenseMatrix<LO, SC> Pi, Pf, Pe;
+    Array<LO> dofType(numNodesInElement * BlkSize), lDofInd(numNodesInElement * BlkSize);
+    factTester.TestComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
+                                       lCoarseElementsPerDir, numDimensions, range, gNodesPerDim,
+                                       gIndices, lCoarseNodesPerDir, ghostInterface,
+                                       elementFlags, "reduced", "coupled", elementNodesPerDir,
+                                       numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
+
+    elemInds[0] = 0;
+    elemInds[1] = 0;
+    elemInds[2] = 0;
+
+    elementFlags[0] = 0;
+    elementFlags[1] = 0;
+    elementFlags[2] = 0;
+    for (int dim = 0; dim < 3; ++dim) {
+      if (elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
+        elementFlags[dim] = boundaryFlags[dim];
+      } else if (elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
+        elementFlags[dim] += 1;
+      } else if ((elemInds[dim] == lCoarseElementsPerDir[dim] - 1) && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
+        elementFlags[dim] += 2;
+      } else {
+        elementFlags[dim] = 0;
       }
-      LO numNodesInElement = elementNodesPerDir[0]*elementNodesPerDir[1]*elementNodesPerDir[2];
-
-      Teuchos::SerialDenseMatrix<LO,SC> Pi, Pf, Pe;
-      Array<LO> dofType(numNodesInElement*BlkSize), lDofInd(numNodesInElement*BlkSize);
-      factTester.TestComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
-                                         lCoarseElementsPerDir, numDimensions, range, gNodesPerDim,
-                                         gIndices, lCoarseNodesPerDir, ghostInterface,
-                                         elementFlags, "reduced", "coupled", elementNodesPerDir,
-                                         numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
-
-      elemInds[0] = 0;
-      elemInds[1] = 0;
-      elemInds[2] = 0;
-
-      elementFlags[0] = 0; elementFlags[1] = 0; elementFlags[2] = 0;
-      for(int dim = 0; dim < 3; ++dim) {
-        if(elemInds[dim] == 0 && elemInds[dim] == lCoarseElementsPerDir[dim] - 1) {
-          elementFlags[dim] = boundaryFlags[dim];
-        } else if(elemInds[dim] == 0 && (boundaryFlags[dim] == 1 || boundaryFlags[dim] == 3)) {
-          elementFlags[dim] += 1;
-        } else if((elemInds[dim] == lCoarseElementsPerDir[dim] - 1)
-                  && (boundaryFlags[dim] == 2 || boundaryFlags[dim] == 3)) {
-          elementFlags[dim] += 2;
-        } else {
-          elementFlags[dim] = 0;
-        }
 
-        // Compute the number of nodes in the current element.
-        if(dim < numDimensions) {
-          if((elemInds[dim] == lCoarseElementsPerDir[dim])
-             && (gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim])) {
-            elementNodesPerDir[dim] = endRate[dim] + 1;
-          } else {
-            elementNodesPerDir[dim] = coarseRate[dim] + 1;
-          }
+      // Compute the number of nodes in the current element.
+      if (dim < numDimensions) {
+        if ((elemInds[dim] == lCoarseElementsPerDir[dim]) && (gIndices[dim] + lNodesPerDim[dim] == gNodesPerDim[dim])) {
+          elementNodesPerDir[dim] = endRate[dim] + 1;
         } else {
-          elementNodesPerDir[dim] = 1;
+          elementNodesPerDir[dim] = coarseRate[dim] + 1;
         }
+      } else {
+        elementNodesPerDir[dim] = 1;
       }
-      numNodesInElement = elementNodesPerDir[0]*elementNodesPerDir[1]*elementNodesPerDir[2];
-
-      dofType.resize(numNodesInElement*BlkSize); lDofInd.resize(numNodesInElement*BlkSize);
-      factTester.TestComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
-                                         lCoarseElementsPerDir, numDimensions, range, gNodesPerDim,
-                                         gIndices, lCoarseNodesPerDir, ghostInterface,
-                                         elementFlags, "reduced", "coupled", elementNodesPerDir,
-                                         numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
     }
+    numNodesInElement = elementNodesPerDir[0] * elementNodesPerDir[1] * elementNodesPerDir[2];
+
+    dofType.resize(numNodesInElement * BlkSize);
+    lDofInd.resize(numNodesInElement * BlkSize);
+    factTester.TestComputeLocalEntries(Aghost, coarseRate, endRate, BlkSize, elemInds,
+                                       lCoarseElementsPerDir, numDimensions, range, gNodesPerDim,
+                                       gIndices, lCoarseNodesPerDir, ghostInterface,
+                                       elementFlags, "reduced", "coupled", elementNodesPerDir,
+                                       numNodesInElement, colGIDs, Pi, Pf, Pe, dofType, lDofInd);
+  }
 
-  } // Prolongator
-
+}  // Prolongator
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, BBPoisson, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlackBoxPFactory, BBPoisson, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
-    const LO numDimensions = 3;
-    const std::string stencilType = "reduced";
+  const LO numDimensions        = 3;
+  const std::string stencilType = "reduced";
 
-    RCP<Matrix> A;
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates;
-    RCP<Map> map;
-    Array<GO> gNodesPerDim(3);
-    Array<LO> lNodesPerDim(3);
-    GetProblemData(comm, lib, numDimensions, stencilType, A, coordinates, map, gNodesPerDim,
-                   lNodesPerDim);
+  RCP<Matrix> A;
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates;
+  RCP<Map> map;
+  Array<GO> gNodesPerDim(3);
+  Array<LO> lNodesPerDim(3);
+  GetProblemData(comm, lib, numDimensions, stencilType, A, coordinates, map, gNodesPerDim,
+                 lNodesPerDim);
 
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
-    }
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
 
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new BlackBoxPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("stencil type", Teuchos::ParameterEntry(std::string("reduced")));
-    Pfact->SetParameter("block strategy", Teuchos::ParameterEntry(std::string("coupled")));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A", A);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);  // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set BlackBoxPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set BlackBoxPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, 2);
-
-    // Extract the prolongator operator
-    RCP<Level> lvl1 = H->GetLevel(1);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-  } // BBPoisson
-
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory,Constructor,       Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory,BlackBoxGhosts,    Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory,GetNodeInfo,       Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory,Prolongator,       Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory,BBPoisson,         Scalar,LO,GO,Node)
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new BlackBoxPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("stencil type", Teuchos::ParameterEntry(std::string("reduced")));
+  Pfact->SetParameter("block strategy", Teuchos::ParameterEntry(std::string("coupled")));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", A);                        // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set BlackBoxPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set BlackBoxPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, 2);
+
+  // Extract the prolongator operator
+  RCP<Level> lvl1                          = H->GetLevel(1);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> PCrs                      = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+}  // BBPoisson
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory, Constructor, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory, BlackBoxGhosts, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory, GetNodeInfo, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory, Prolongator, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlackBoxPFactory, BBPoisson, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/BlockedCoarseMapFactory.cpp b/packages/muelu/test/unit_tests/BlockedCoarseMapFactory.cpp
index 658de4028327..f3ddd5cb7de7 100644
--- a/packages/muelu/test/unit_tests/BlockedCoarseMapFactory.cpp
+++ b/packages/muelu/test/unit_tests/BlockedCoarseMapFactory.cpp
@@ -61,85 +61,83 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedCoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<BlockedCoarseMapFactory> mapFact = rcp(new BlockedCoarseMapFactory());
-    TEST_EQUALITY(mapFact != Teuchos::null, true);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedCoarseMapFactory, BuildBlockedCoarseMapWithGIDOffset, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel;
-    Level coarseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(29);
-    A->SetFixedBlockSize(1);
-    fineLevel.Set("A", A);
-
-    LO NSdim = 2;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetFactory("Graph", dropFact);
-    UncoupledAggFact->SetFactory("DofsPerNode", dropFact);
-
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
-
-    RCP<BlockedCoarseMapFactory> blockedCoarseMapFact = rcp(new BlockedCoarseMapFactory());
-    blockedCoarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
-    blockedCoarseMapFact->SetFactory("CoarseMap", coarseMapFact);
-
-    // request input for BlockedCoarseMapFactory by hand
-    fineLevel.Request("Aggregates", UncoupledAggFact.get());
-    fineLevel.Request("CoarseMap", coarseMapFact.get());
-    fineLevel.Request("CoarseMap", blockedCoarseMapFact.get());
-    blockedCoarseMapFact->Build(fineLevel);
-    RCP<const Map> map1 = fineLevel.Get<RCP<const Map>>("CoarseMap", coarseMapFact.get());
-    RCP<const Map> map2 = fineLevel.Get<RCP<const Map>>("CoarseMap", blockedCoarseMapFact.get());
-
-    // access aggregates
-    RCP<Aggregates> aggregates = fineLevel.Get<RCP<Aggregates>>("Aggregates", UncoupledAggFact.get());
-    GO numAggs = aggregates->GetNumAggregates();
-    GO numGlobalAggs = 0;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    MueLu_sumAll(comm, numAggs, numGlobalAggs);
-
-    TEST_EQUALITY(map1->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(map1->getMaxAllGlobalIndex(), numGlobalAggs * Teuchos::as<GO>(NSdim) - 1);
-    TEST_EQUALITY(map2->getMinAllGlobalIndex(), numGlobalAggs * Teuchos::as<GO>(NSdim));
-    TEST_EQUALITY(map2->getMaxAllGlobalIndex(), 2 * numGlobalAggs * Teuchos::as<GO>(NSdim) - 1);
-    TEST_EQUALITY(Teuchos::as<GO>(map1->getLocalNumElements()), numAggs * Teuchos::as<GO>(NSdim));
-    TEST_EQUALITY(Teuchos::as<GO>(map2->getLocalNumElements()), numAggs * Teuchos::as<GO>(NSdim));
-  }
-
-#define MUELU_ETI_GROUP(Scalar,LocalOrdinal,GlobalOrdinal,Node) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedCoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<BlockedCoarseMapFactory> mapFact = rcp(new BlockedCoarseMapFactory());
+  TEST_EQUALITY(mapFact != Teuchos::null, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedCoarseMapFactory, BuildBlockedCoarseMapWithGIDOffset, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel;
+  Level coarseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(29);
+  A->SetFixedBlockSize(1);
+  fineLevel.Set("A", A);
+
+  LO NSdim                   = 2;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetFactory("Graph", dropFact);
+  UncoupledAggFact->SetFactory("DofsPerNode", dropFact);
+
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
+
+  RCP<BlockedCoarseMapFactory> blockedCoarseMapFact = rcp(new BlockedCoarseMapFactory());
+  blockedCoarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
+  blockedCoarseMapFact->SetFactory("CoarseMap", coarseMapFact);
+
+  // request input for BlockedCoarseMapFactory by hand
+  fineLevel.Request("Aggregates", UncoupledAggFact.get());
+  fineLevel.Request("CoarseMap", coarseMapFact.get());
+  fineLevel.Request("CoarseMap", blockedCoarseMapFact.get());
+  blockedCoarseMapFact->Build(fineLevel);
+  RCP<const Map> map1 = fineLevel.Get<RCP<const Map>>("CoarseMap", coarseMapFact.get());
+  RCP<const Map> map2 = fineLevel.Get<RCP<const Map>>("CoarseMap", blockedCoarseMapFact.get());
+
+  // access aggregates
+  RCP<Aggregates> aggregates         = fineLevel.Get<RCP<Aggregates>>("Aggregates", UncoupledAggFact.get());
+  GO numAggs                         = aggregates->GetNumAggregates();
+  GO numGlobalAggs                   = 0;
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  MueLu_sumAll(comm, numAggs, numGlobalAggs);
+
+  TEST_EQUALITY(map1->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(map1->getMaxAllGlobalIndex(), numGlobalAggs * Teuchos::as<GO>(NSdim) - 1);
+  TEST_EQUALITY(map2->getMinAllGlobalIndex(), numGlobalAggs * Teuchos::as<GO>(NSdim));
+  TEST_EQUALITY(map2->getMaxAllGlobalIndex(), 2 * numGlobalAggs * Teuchos::as<GO>(NSdim) - 1);
+  TEST_EQUALITY(Teuchos::as<GO>(map1->getLocalNumElements()), numAggs * Teuchos::as<GO>(NSdim));
+  TEST_EQUALITY(Teuchos::as<GO>(map2->getLocalNumElements()), numAggs * Teuchos::as<GO>(NSdim));
+}
+
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                                      \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedCoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedCoarseMapFactory, BuildBlockedCoarseMapWithGIDOffset, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/BlockedPFactory.cpp b/packages/muelu/test/unit_tests/BlockedPFactory.cpp
index 62cef69fb3c1..d26ca1334a50 100644
--- a/packages/muelu/test/unit_tests/BlockedPFactory.cpp
+++ b/packages/muelu/test/unit_tests/BlockedPFactory.cpp
@@ -64,228 +64,225 @@
 
 namespace MueLuTests {
 
-  /////////////////////////
-  // helper function
-  // note: we assume "domainmap" to be linear starting with GIDs from domainmap->getMinAllGlobalIndex() to
-  //       domainmap->getMaxAllGlobalIndex() and build a quadratic triangular matrix with the stencil (b,a,c)
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal, GlobalOrdinal, Node> >
-  GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangemap,
-                        const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainmap,
-                        Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
+/////////////////////////
+// helper function
+// note: we assume "domainmap" to be linear starting with GIDs from domainmap->getMinAllGlobalIndex() to
+//       domainmap->getMaxAllGlobalIndex() and build a quadratic triangular matrix with the stencil (b,a,c)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangemap,
+                      const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainmap,
+                      Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
 #include "MueLu_UseShortNames.hpp"
-    Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(rangemap, 3);
-
-    LocalOrdinal NumMyRowElements = rangemap->getLocalNumElements();
-
-    GlobalOrdinal minGColId = domainmap->getMinAllGlobalIndex();  // minimum over all procs
-    GlobalOrdinal maxGColId = domainmap->getMaxAllGlobalIndex();  // maximum over all procs
-    GlobalOrdinal numGColElements = domainmap->getGlobalNumElements();
-    //std::cout << maxGColId << " " << minGColId << " " << numGColElements <<std::endl;
-    TEUCHOS_TEST_FOR_EXCEPTION(maxGColId-minGColId!=numGColElements-1,MueLu::Exceptions::RuntimeError,"GenerateProblemMatrix: incosistent number of map elements.");
-
-    GlobalOrdinal minGRowId = rangemap->getMinAllGlobalIndex(); // minimum over all procs
-    GlobalOrdinal maxGRowId = rangemap->getMaxAllGlobalIndex(); // maximum over all procs
-    TEUCHOS_TEST_FOR_EXCEPTION(maxGRowId-minGRowId!=maxGColId-minGColId,MueLu::Exceptions::RuntimeError,"GenerateProblemMatrix: incosistent number of map elements between range and domain maps.");
-
-    GlobalOrdinal offset = minGColId - minGRowId;
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    // loop over all local rows
-    for (LocalOrdinal i = 0; i < NumMyRowElements; ++i) {
-      GlobalOrdinal grid = rangemap->getGlobalElement(i);
-      if(grid == minGRowId) {
-        NumEntries = 1;
-        Values[0]  = c;
-        Indices[0] = minGColId+1;
-      } else if (grid == maxGRowId) {
-        NumEntries = 1;
-        Values[0]  = b;
-        Indices[0] = maxGColId-1;
-      } else {
-        NumEntries = 2;
-        Indices[0] = offset + rangemap->getMinGlobalIndex() + i - 1;
-        Indices[1] = offset + rangemap->getMinGlobalIndex() + i + 1;
-        Values[0] = b;
-        Values[1] = c;
-      }
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(rangemap->getGlobalElement(i), iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(grid,
-          Teuchos::tuple<GlobalOrdinal>(offset + rangemap->getMinGlobalIndex() + i),
-          Teuchos::tuple<Scalar>(a) );
+  Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(rangemap, 3);
+
+  LocalOrdinal NumMyRowElements = rangemap->getLocalNumElements();
+
+  GlobalOrdinal minGColId       = domainmap->getMinAllGlobalIndex();  // minimum over all procs
+  GlobalOrdinal maxGColId       = domainmap->getMaxAllGlobalIndex();  // maximum over all procs
+  GlobalOrdinal numGColElements = domainmap->getGlobalNumElements();
+  // std::cout << maxGColId << " " << minGColId << " " << numGColElements <<std::endl;
+  TEUCHOS_TEST_FOR_EXCEPTION(maxGColId - minGColId != numGColElements - 1, MueLu::Exceptions::RuntimeError, "GenerateProblemMatrix: incosistent number of map elements.");
+
+  GlobalOrdinal minGRowId = rangemap->getMinAllGlobalIndex();  // minimum over all procs
+  GlobalOrdinal maxGRowId = rangemap->getMaxAllGlobalIndex();  // maximum over all procs
+  TEUCHOS_TEST_FOR_EXCEPTION(maxGRowId - minGRowId != maxGColId - minGColId, MueLu::Exceptions::RuntimeError, "GenerateProblemMatrix: incosistent number of map elements between range and domain maps.");
+
+  GlobalOrdinal offset = minGColId - minGRowId;
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  // loop over all local rows
+  for (LocalOrdinal i = 0; i < NumMyRowElements; ++i) {
+    GlobalOrdinal grid = rangemap->getGlobalElement(i);
+    if (grid == minGRowId) {
+      NumEntries = 1;
+      Values[0]  = c;
+      Indices[0] = minGColId + 1;
+    } else if (grid == maxGRowId) {
+      NumEntries = 1;
+      Values[0]  = b;
+      Indices[0] = maxGColId - 1;
+    } else {
+      NumEntries = 2;
+      Indices[0] = offset + rangemap->getMinGlobalIndex() + i - 1;
+      Indices[1] = offset + rangemap->getMinGlobalIndex() + i + 1;
+      Values[0]  = b;
+      Values[1]  = c;
     }
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(rangemap->getGlobalElement(i), iv, av);
+
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(grid,
+                            Teuchos::tuple<GlobalOrdinal>(offset + rangemap->getMinGlobalIndex() + i),
+                            Teuchos::tuple<Scalar>(a));
+  }
 
-    mtx->fillComplete(domainmap,rangemap);
-    return mtx;
+  mtx->fillComplete(domainmap, rangemap);
+  return mtx;
+}
 
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<BlockedPFactory> blockedPFactory = rcp(new BlockedPFactory());
-    TEST_EQUALITY(blockedPFactory != Teuchos::null, true);
-  } // Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedPFactory, CreateBlockedP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    /**********************************************************************************/
-    /* CREATE INITIAL MATRIX                                                          */
-    /**********************************************************************************/
-    const GO numElements = 400;
-    const GO numElements1 = 200;
-    const GO numElements2 = 200;
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // the test matrix has to be a nxn block matrix with quadratic blocks
-    // where the subblocks use consequent numbering of global DOF ids.
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(1);
-
-    RCP<const Map> map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-    RCP<const Map> map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
-
-    std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-    Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
-    localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-    Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
-    localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-    RCP<const Map> bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
-    std::vector<Teuchos::RCP<const Map> > maps;
-    maps.push_back(map1); maps.push_back(map2);
-
-    Teuchos::RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(bigMap, maps);
-
-    RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map1,2,-1,-1);
-    RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map2,1, 0, 0);
-    RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map1,1, 0, 0);
-    RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map2,3,-2,-1);
-
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bOp = Teuchos::rcp(new BlockedCrsMatrix(mapExtractor,mapExtractor,10));
-    bOp->setMatrix(0,0,Op11);
-    bOp->setMatrix(0,1,Op12);
-    bOp->setMatrix(1,0,Op21);
-    bOp->setMatrix(1,1,Op22);
-    bOp->fillComplete();
-    TEST_EQUALITY(bOp!=Teuchos::null, true);
-
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    RCP<Level> levelTwo = rcp(new Level()); levelTwo->SetPreviousLevel(levelOne);
-    levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-
-    // define sub block factories for blocked operator "A"
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
-    RCP<SubBlockAFactory> A12Fact = Teuchos::rcp(new SubBlockAFactory());
-    A12Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A12Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A12Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
-    RCP<SubBlockAFactory> A21Fact = Teuchos::rcp(new SubBlockAFactory());
-    A21Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A21Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A21Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
-
-    // request subblocks of A
-    levelOne->Request("A", A11Fact.get(), MueLu::NoFactory::get());
-    levelOne->Request("A", A22Fact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelOne->IsRequested("A", A11Fact.get()),true);
-    TEST_EQUALITY(levelOne->IsRequested("A", A22Fact.get()),true);
-
-    RCP<Matrix> A11 = levelOne->Get<RCP<Matrix> >("A",A11Fact.get());
-    RCP<Matrix> A22 = levelOne->Get<RCP<Matrix> >("A",A22Fact.get());
-    TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()),true);
-    TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()),true);
-
-    // store data A11 and A22 as variable "P" on level 2
-
-    levelTwo->Request("P", A11Fact.get(), MueLu::NoFactory::get());
-    levelTwo->Request("P", A22Fact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("P", A11Fact.get()),true);
-    TEST_EQUALITY(levelTwo->IsRequested("P", A22Fact.get()),true);
-
-    levelTwo-> Set("P", A11, A11Fact.get());
-    levelTwo-> Set("P", A22, A22Fact.get());
-    TEST_EQUALITY(levelTwo->IsAvailable("P", A11Fact.get()),true);
-    TEST_EQUALITY(levelTwo->IsAvailable("P", A22Fact.get()),true);
-
-    levelOne->Release("A", A11Fact.get());
-    levelOne->Release("A", A22Fact.get());
-
-    // set up factory manager
-    RCP<FactoryManager> FC1 = rcp(new FactoryManager());
-    FC1->SetFactory("P", A11Fact);  // fool P to be generated by A11Fact
-    FC1->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    RCP<FactoryManager> FC2 = rcp(new FactoryManager());
-    FC2->SetFactory("P", A22Fact);  // fool P to be generated by A11Fact
-    FC2->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    /////////////////////////////////////////// define blocked transfer ops
-    RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory()); // use row map index base from bOp
-    PFact->AddFactoryManager(FC1);
-    PFact->AddFactoryManager(FC2);
-
-    levelTwo->Request("P", PFact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("P", PFact.get()),true);
-
-    RCP<Matrix> P = levelTwo->Get<RCP<Matrix> >("P", PFact.get());
-    TEST_EQUALITY(P!=Teuchos::null,true);
-
-    RCP<BlockedCrsMatrix> bP = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(P);
-    TEST_EQUALITY(bP!=Teuchos::null,true);
-
-    TEST_EQUALITY(bP->Rows(), 2);
-    TEST_EQUALITY(bP->Cols(), 2);
-
-    // create test and rhs vector
-    RCP<const Map> fullMap = mapExtractor->getFullMap();
-    TEST_EQUALITY(fullMap == bigMap, true);
-    RCP<Vector> iones = VectorFactory::Build(fullMap);
-    RCP<Vector> rones = VectorFactory::Build(fullMap);
-    iones->putScalar(1.0);
-    bP->apply(*iones, *rones); // the subblocks are chosen, such that bP*ones = zero (except for the first and row and the middle row)
-    TEST_EQUALITY(rones->norm1(), 5.0);
-    TEST_EQUALITY(rones->normInf(), 2.0);
-  } //Constructor
-
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedPFactory, Constructor, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedPFactory, CreateBlockedP, SC, LO, GO, Node)
+  out << "version: " << MueLu::Version() << std::endl;
 
-#include <MueLu_ETI_4arg.hpp>
+  RCP<BlockedPFactory> blockedPFactory = rcp(new BlockedPFactory());
+  TEST_EQUALITY(blockedPFactory != Teuchos::null, true);
+}  // Constructor
 
-} // namespace MueLuTests
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedPFactory, CreateBlockedP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  /**********************************************************************************/
+  /* CREATE INITIAL MATRIX                                                          */
+  /**********************************************************************************/
+  const GO numElements  = 400;
+  const GO numElements1 = 200;
+  const GO numElements2 = 200;
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // the test matrix has to be a nxn block matrix with quadratic blocks
+  // where the subblocks use consequent numbering of global DOF ids.
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(1);
+
+  RCP<const Map> map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  RCP<const Map> map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
+
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
+  localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
+  localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  RCP<const Map> bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
+  std::vector<Teuchos::RCP<const Map> > maps;
+  maps.push_back(map1);
+  maps.push_back(map2);
+
+  Teuchos::RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(bigMap, maps);
+
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map2, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map1, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map2, 3, -2, -1);
+
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bOp = Teuchos::rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(0, 1, Op12);
+  bOp->setMatrix(1, 0, Op21);
+  bOp->setMatrix(1, 1, Op22);
+  bOp->fillComplete();
+  TEST_EQUALITY(bOp != Teuchos::null, true);
+
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  RCP<Level> levelTwo = rcp(new Level());
+  levelTwo->SetPreviousLevel(levelOne);
+  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+
+  // define sub block factories for blocked operator "A"
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+  RCP<SubBlockAFactory> A12Fact = Teuchos::rcp(new SubBlockAFactory());
+  A12Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A12Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A12Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+  RCP<SubBlockAFactory> A21Fact = Teuchos::rcp(new SubBlockAFactory());
+  A21Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A21Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A21Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+
+  // request subblocks of A
+  levelOne->Request("A", A11Fact.get(), MueLu::NoFactory::get());
+  levelOne->Request("A", A22Fact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelOne->IsRequested("A", A11Fact.get()), true);
+  TEST_EQUALITY(levelOne->IsRequested("A", A22Fact.get()), true);
+
+  RCP<Matrix> A11 = levelOne->Get<RCP<Matrix> >("A", A11Fact.get());
+  RCP<Matrix> A22 = levelOne->Get<RCP<Matrix> >("A", A22Fact.get());
+  TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()), true);
+  TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()), true);
+
+  // store data A11 and A22 as variable "P" on level 2
+
+  levelTwo->Request("P", A11Fact.get(), MueLu::NoFactory::get());
+  levelTwo->Request("P", A22Fact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("P", A11Fact.get()), true);
+  TEST_EQUALITY(levelTwo->IsRequested("P", A22Fact.get()), true);
+
+  levelTwo->Set("P", A11, A11Fact.get());
+  levelTwo->Set("P", A22, A22Fact.get());
+  TEST_EQUALITY(levelTwo->IsAvailable("P", A11Fact.get()), true);
+  TEST_EQUALITY(levelTwo->IsAvailable("P", A22Fact.get()), true);
+
+  levelOne->Release("A", A11Fact.get());
+  levelOne->Release("A", A22Fact.get());
+
+  // set up factory manager
+  RCP<FactoryManager> FC1 = rcp(new FactoryManager());
+  FC1->SetFactory("P", A11Fact);  // fool P to be generated by A11Fact
+  FC1->SetIgnoreUserData(true);   // always use data from factories defined in factory manager
+
+  RCP<FactoryManager> FC2 = rcp(new FactoryManager());
+  FC2->SetFactory("P", A22Fact);  // fool P to be generated by A11Fact
+  FC2->SetIgnoreUserData(true);   // always use data from factories defined in factory manager
+
+  /////////////////////////////////////////// define blocked transfer ops
+  RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());  // use row map index base from bOp
+  PFact->AddFactoryManager(FC1);
+  PFact->AddFactoryManager(FC2);
+
+  levelTwo->Request("P", PFact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("P", PFact.get()), true);
+
+  RCP<Matrix> P = levelTwo->Get<RCP<Matrix> >("P", PFact.get());
+  TEST_EQUALITY(P != Teuchos::null, true);
+
+  RCP<BlockedCrsMatrix> bP = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(P);
+  TEST_EQUALITY(bP != Teuchos::null, true);
+
+  TEST_EQUALITY(bP->Rows(), 2);
+  TEST_EQUALITY(bP->Cols(), 2);
+
+  // create test and rhs vector
+  RCP<const Map> fullMap = mapExtractor->getFullMap();
+  TEST_EQUALITY(fullMap == bigMap, true);
+  RCP<Vector> iones = VectorFactory::Build(fullMap);
+  RCP<Vector> rones = VectorFactory::Build(fullMap);
+  iones->putScalar(1.0);
+  bP->apply(*iones, *rones);  // the subblocks are chosen, such that bP*ones = zero (except for the first and row and the middle row)
+  TEST_EQUALITY(rones->norm1(), 5.0);
+  TEST_EQUALITY(rones->normInf(), 2.0);
+}  // Constructor
+
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedPFactory, Constructor, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedPFactory, CreateBlockedP, SC, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
 
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/BlockedRAPFactory.cpp b/packages/muelu/test/unit_tests/BlockedRAPFactory.cpp
index 34a8aec466e0..4d3c78e8e063 100644
--- a/packages/muelu/test/unit_tests/BlockedRAPFactory.cpp
+++ b/packages/muelu/test/unit_tests/BlockedRAPFactory.cpp
@@ -56,20 +56,19 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRAPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRAPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<BlockedRAPFactory> blockedRAPFactory = rcp(new BlockedRAPFactory());
-    TEST_EQUALITY(blockedRAPFactory != Teuchos::null, true);
-  } // Constructor
+  RCP<BlockedRAPFactory> blockedRAPFactory = rcp(new BlockedRAPFactory());
+  TEST_EQUALITY(blockedRAPFactory != Teuchos::null, true);
+}  // Constructor
 
-# define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRAPFactory, Constructor, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRAPFactory, Constructor, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/BlockedRepartition.cpp b/packages/muelu/test/unit_tests/BlockedRepartition.cpp
index 407f526a042c..d63f6f3b7025 100644
--- a/packages/muelu/test/unit_tests/BlockedRepartition.cpp
+++ b/packages/muelu/test/unit_tests/BlockedRepartition.cpp
@@ -99,565 +99,589 @@
 
 namespace MueLuTests {
 
-  /////////////////////////
-  // helper function
-  // note: we assume "domainmap" to be linear starting with GIDs from domainmap->getMinAllGlobalIndex() to
-  //       domainmap->getMaxAllGlobalIndex() and build a quadratic triangular matrix with the stencil (b,a,c)
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal, GlobalOrdinal, Node> >
-  GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangemap,
-                        const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainmap,
-                        Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
+/////////////////////////
+// helper function
+// note: we assume "domainmap" to be linear starting with GIDs from domainmap->getMinAllGlobalIndex() to
+//       domainmap->getMaxAllGlobalIndex() and build a quadratic triangular matrix with the stencil (b,a,c)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > rangemap,
+                      const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > domainmap,
+                      Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
 #include "MueLu_UseShortNames.hpp"
-    Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(rangemap, 3);
-
-    LocalOrdinal NumMyRowElements = rangemap->getLocalNumElements();
-
-    GlobalOrdinal minGColId = domainmap->getMinAllGlobalIndex();  // minimum over all procs
-    GlobalOrdinal maxGColId = domainmap->getMaxAllGlobalIndex();  // maximum over all procs
-    GlobalOrdinal numGColElements = domainmap->getGlobalNumElements();
-    //    std::cout << maxGColId << " " << minGColId << " " << numGColElements <<std::endl;
-    TEUCHOS_TEST_FOR_EXCEPTION(maxGColId-minGColId!=numGColElements-1,MueLu::Exceptions::RuntimeError,"GenerateProblemMatrix: incosistent number of map elements.");
-
-    GlobalOrdinal minGRowId = rangemap->getMinAllGlobalIndex(); // minimum over all procs
-    GlobalOrdinal maxGRowId = rangemap->getMaxAllGlobalIndex(); // maximum over all procs
-    TEUCHOS_TEST_FOR_EXCEPTION(maxGRowId-minGRowId!=maxGColId-minGColId,MueLu::Exceptions::RuntimeError,"GenerateProblemMatrix: incosistent number of map elements between range and domain maps.");
-
-    GlobalOrdinal offset = minGColId - minGRowId;
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    // loop over all local rows
-    for (LocalOrdinal i = 0; i < NumMyRowElements; ++i) {
-      GlobalOrdinal grid = rangemap->getGlobalElement(i);
-      if(grid == minGRowId) {
-        NumEntries = 1;
-        Values[0]  = c;
-        Indices[0] = minGColId+1;
-      } else if (grid == maxGRowId) {
-        NumEntries = 1;
-        Values[0]  = b;
-        Indices[0] = maxGColId-1;
-      } else {
-        NumEntries = 2;
-        Indices[0] = offset + rangemap->getMinGlobalIndex() + i - 1;
-        Indices[1] = offset + rangemap->getMinGlobalIndex() + i + 1;
-        Values[0] = b;
-        Values[1] = c;
-      }
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(rangemap->getGlobalElement(i), iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(grid,
-          Teuchos::tuple<GlobalOrdinal>(offset + rangemap->getMinGlobalIndex() + i),
-          Teuchos::tuple<Scalar>(a) );
+  Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(rangemap, 3);
+
+  LocalOrdinal NumMyRowElements = rangemap->getLocalNumElements();
+
+  GlobalOrdinal minGColId       = domainmap->getMinAllGlobalIndex();  // minimum over all procs
+  GlobalOrdinal maxGColId       = domainmap->getMaxAllGlobalIndex();  // maximum over all procs
+  GlobalOrdinal numGColElements = domainmap->getGlobalNumElements();
+  //    std::cout << maxGColId << " " << minGColId << " " << numGColElements <<std::endl;
+  TEUCHOS_TEST_FOR_EXCEPTION(maxGColId - minGColId != numGColElements - 1, MueLu::Exceptions::RuntimeError, "GenerateProblemMatrix: incosistent number of map elements.");
+
+  GlobalOrdinal minGRowId = rangemap->getMinAllGlobalIndex();  // minimum over all procs
+  GlobalOrdinal maxGRowId = rangemap->getMaxAllGlobalIndex();  // maximum over all procs
+  TEUCHOS_TEST_FOR_EXCEPTION(maxGRowId - minGRowId != maxGColId - minGColId, MueLu::Exceptions::RuntimeError, "GenerateProblemMatrix: incosistent number of map elements between range and domain maps.");
+
+  GlobalOrdinal offset = minGColId - minGRowId;
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  // loop over all local rows
+  for (LocalOrdinal i = 0; i < NumMyRowElements; ++i) {
+    GlobalOrdinal grid = rangemap->getGlobalElement(i);
+    if (grid == minGRowId) {
+      NumEntries = 1;
+      Values[0]  = c;
+      Indices[0] = minGColId + 1;
+    } else if (grid == maxGRowId) {
+      NumEntries = 1;
+      Values[0]  = b;
+      Indices[0] = maxGColId - 1;
+    } else {
+      NumEntries = 2;
+      Indices[0] = offset + rangemap->getMinGlobalIndex() + i - 1;
+      Indices[1] = offset + rangemap->getMinGlobalIndex() + i + 1;
+      Values[0]  = b;
+      Values[1]  = c;
     }
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(rangemap->getGlobalElement(i), iv, av);
+
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(grid,
+                            Teuchos::tuple<GlobalOrdinal>(offset + rangemap->getMinGlobalIndex() + i),
+                            Teuchos::tuple<Scalar>(a));
+  }
 
-    mtx->fillComplete(domainmap,rangemap);
-    return mtx;
+  mtx->fillComplete(domainmap, rangemap);
+  return mtx;
 
-  } //GenerateProblemMatrix
+}  // GenerateProblemMatrix
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    if(comm->getSize() == 1) {
-      out << "Skip BlockedRepartion test in serial case" << std::endl;
-      return;
-    }
+  if (comm->getSize() == 1) {
+    out << "Skip BlockedRepartion test in serial case" << std::endl;
+    return;
+  }
+
+  int nNumProcs = comm->getSize();  // number of procs used before rebalancing
+
+  /**********************************************************************************/
+  /* CREATE INITIAL MATRIX                                                          */
+  /**********************************************************************************/
+  RCP<const Map> bigMap;
+  RCP<const Map> map1;
+  RCP<const Map> map2;
+  GO numElements  = 400;
+  GO numElements1 = 200;
+  GO numElements2 = 200;
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // the test matrix has to be a nxn block matrix with quadratic blocks
+  // where the subblocks use consequent numbering of global DOF ids.
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(1);
+
+  map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
+
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
+  localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
+  localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
+  std::vector<Teuchos::RCP<const Map> > maps;
+  maps.push_back(map1);
+  maps.push_back(map2);
+
+  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(bigMap, maps);
 
-    int nNumProcs = comm->getSize(); // number of procs used before rebalancing
-
-    /**********************************************************************************/
-    /* CREATE INITIAL MATRIX                                                          */
-    /**********************************************************************************/
-    RCP<const Map> bigMap;
-    RCP<const Map> map1;
-    RCP<const Map> map2;
-    GO numElements = 400;
-    GO numElements1 = 200;
-    GO numElements2 = 200;
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // the test matrix has to be a nxn block matrix with quadratic blocks
-    // where the subblocks use consequent numbering of global DOF ids.
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(1);
-
-    map1   = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-    map2   = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
-
-    std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-    Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
-    localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-    Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
-    localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-    bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
-    std::vector<Teuchos::RCP<const Map> > maps;
-    maps.push_back(map1); maps.push_back(map2);
-
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(bigMap, maps);
-
-    RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map1,2,-1,-1);
-    RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map2,1, 0, 0);
-    RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map1,1, 0, 0);
-    RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map2,3,-2,-1);
-
-    // store output of simple MV products for OpIJ
-    RCP<Vector> test11  = VectorFactory::Build(Op11->getDomainMap()); test11->putScalar(1.0);
-    RCP<Vector> test12  = VectorFactory::Build(Op12->getDomainMap()); test12->putScalar(1.0);
-    RCP<Vector> test21  = VectorFactory::Build(Op21->getDomainMap()); test21->putScalar(1.0);
-    RCP<Vector> test22  = VectorFactory::Build(Op22->getDomainMap()); test22->putScalar(1.0);
-    RCP<Vector> res11   = VectorFactory::Build(Op11->getRangeMap());
-    RCP<Vector> res12   = VectorFactory::Build(Op12->getRangeMap());
-    RCP<Vector> res21   = VectorFactory::Build(Op21->getRangeMap());
-    RCP<Vector> res22   = VectorFactory::Build(Op22->getRangeMap());
-    Op11->apply(*test11,*res11);
-    Op12->apply(*test12,*res12);
-    Op21->apply(*test21,*res21);
-    Op22->apply(*test22,*res22);
-
-    // build blocked operator
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(mapExtractor,mapExtractor,10));
-
-    bOp->setMatrix(0,0,Op11);
-    bOp->setMatrix(0,1,Op12);
-    bOp->setMatrix(1,0,Op21);
-    bOp->setMatrix(1,1,Op22);
-    bOp->fillComplete();
-    TEST_EQUALITY(bOp!=Teuchos::null, true);
-    TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
-
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    levelOne->SetLevelID(-1);
-    RCP<Level> levelTwo = rcp(new Level()); levelTwo->SetPreviousLevel(levelOne);
-    levelTwo->SetLevelID(0);
-    levelTwo->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-
-    // define repartition heuristics
-    RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
-    RepHeuFact->SetFactory("A", MueLu::NoFactory::getRCP()); // 2x2 blocked operator
-    RepHeuFact->SetParameter("repartition: start level",Teuchos::ParameterEntry(0));
-    RepHeuFact->SetParameter("repartition: min rows per proc",Teuchos::ParameterEntry(200));
-
-    // define sub block factories for blocked operator "A"
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
-
-    RCP<AmalgamationFactory> Amalg11Fact = Teuchos::rcp(new AmalgamationFactory());
-    RCP<AmalgamationFactory> Amalg22Fact = Teuchos::rcp(new AmalgamationFactory());
-    Amalg11Fact->SetFactory("A", A11Fact);
-    Amalg22Fact->SetFactory("A", A22Fact);
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map2, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map1, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map2, 3, -2, -1);
+
+  // store output of simple MV products for OpIJ
+  RCP<Vector> test11 = VectorFactory::Build(Op11->getDomainMap());
+  test11->putScalar(1.0);
+  RCP<Vector> test12 = VectorFactory::Build(Op12->getDomainMap());
+  test12->putScalar(1.0);
+  RCP<Vector> test21 = VectorFactory::Build(Op21->getDomainMap());
+  test21->putScalar(1.0);
+  RCP<Vector> test22 = VectorFactory::Build(Op22->getDomainMap());
+  test22->putScalar(1.0);
+  RCP<Vector> res11 = VectorFactory::Build(Op11->getRangeMap());
+  RCP<Vector> res12 = VectorFactory::Build(Op12->getRangeMap());
+  RCP<Vector> res21 = VectorFactory::Build(Op21->getRangeMap());
+  RCP<Vector> res22 = VectorFactory::Build(Op22->getRangeMap());
+  Op11->apply(*test11, *res11);
+  Op12->apply(*test12, *res12);
+  Op21->apply(*test21, *res21);
+  Op22->apply(*test22, *res22);
+
+  // build blocked operator
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(mapExtractor, mapExtractor, 10));
+
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(0, 1, Op12);
+  bOp->setMatrix(1, 0, Op21);
+  bOp->setMatrix(1, 1, Op22);
+  bOp->fillComplete();
+  TEST_EQUALITY(bOp != Teuchos::null, true);
+  TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
+
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  levelOne->SetLevelID(-1);
+  RCP<Level> levelTwo = rcp(new Level());
+  levelTwo->SetPreviousLevel(levelOne);
+  levelTwo->SetLevelID(0);
+  levelTwo->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+
+  // define repartition heuristics
+  RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
+  RepHeuFact->SetFactory("A", MueLu::NoFactory::getRCP());  // 2x2 blocked operator
+  RepHeuFact->SetParameter("repartition: start level", Teuchos::ParameterEntry(0));
+  RepHeuFact->SetParameter("repartition: min rows per proc", Teuchos::ParameterEntry(200));
+
+  // define sub block factories for blocked operator "A"
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+
+  RCP<AmalgamationFactory> Amalg11Fact = Teuchos::rcp(new AmalgamationFactory());
+  RCP<AmalgamationFactory> Amalg22Fact = Teuchos::rcp(new AmalgamationFactory());
+  Amalg11Fact->SetFactory("A", A11Fact);
+  Amalg22Fact->SetFactory("A", A22Fact);
 
 #ifdef HAVE_MUELU_ISORROPIA
-    RCP<Factory> Rep11Interface = Teuchos::null;
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
-      RCP<IsorropiaInterface> Iso11Interface = Teuchos::rcp(new IsorropiaInterface());
-      Iso11Interface->SetFactory("A", A11Fact);
-      Iso11Interface->SetFactory("number of partitions", RepHeuFact);
-      Iso11Interface->SetFactory("UnAmalgamationInfo", Amalg11Fact);
-
-      Rep11Interface = Teuchos::rcp(new RepartitionInterface());
-      Rep11Interface->SetFactory("A", A11Fact);
-      Rep11Interface->SetFactory("number of partitions", RepHeuFact);
-      Rep11Interface->SetFactory("AmalgamatedPartition", Iso11Interface);
-    } else {
-      // we are in Tpetra mode (even though Isorropia would be available)
-      // create dummy "Partition" array
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
-      ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-      for(size_t r=0; r<decomposition->getMap()->getLocalNumElements(); r++) {
-        if(r%2 == 0) decompEntries[r] = 0;
-        else decompEntries[r] = 1;
-      }
-      Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
-      levelTwo->Request("Partition", Rep11Interface.get());
-      levelTwo->Set("Partition", decomposition, Rep11Interface.get());
-    }
-#else
+  RCP<Factory> Rep11Interface = Teuchos::null;
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    RCP<IsorropiaInterface> Iso11Interface = Teuchos::rcp(new IsorropiaInterface());
+    Iso11Interface->SetFactory("A", A11Fact);
+    Iso11Interface->SetFactory("number of partitions", RepHeuFact);
+    Iso11Interface->SetFactory("UnAmalgamationInfo", Amalg11Fact);
+
+    Rep11Interface = Teuchos::rcp(new RepartitionInterface());
+    Rep11Interface->SetFactory("A", A11Fact);
+    Rep11Interface->SetFactory("number of partitions", RepHeuFact);
+    Rep11Interface->SetFactory("AmalgamatedPartition", Iso11Interface);
+  } else {
+    // we are in Tpetra mode (even though Isorropia would be available)
     // create dummy "Partition" array
     RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
-    ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-    for(size_t r=0; r<decomposition->getMap()->getLocalNumElements(); r++) {
-      if(r%2 == 0) decompEntries[r] = 0;
-      else decompEntries[r] = 1;
+    ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
+    for (size_t r = 0; r < decomposition->getMap()->getLocalNumElements(); r++) {
+      if (r % 2 == 0)
+        decompEntries[r] = 0;
+      else
+        decompEntries[r] = 1;
     }
-    RCP<SubBlockAFactory> Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
+    Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
     levelTwo->Request("Partition", Rep11Interface.get());
     levelTwo->Set("Partition", decomposition, Rep11Interface.get());
+  }
+#else
+  // create dummy "Partition" array
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
+  ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
+  for (size_t r = 0; r < decomposition->getMap()->getLocalNumElements(); r++) {
+    if (r % 2 == 0)
+      decompEntries[r] = 0;
+    else
+      decompEntries[r] = 1;
+  }
+  RCP<SubBlockAFactory> Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
+  levelTwo->Request("Partition", Rep11Interface.get());
+  levelTwo->Set("Partition", decomposition, Rep11Interface.get());
 #endif
 
-    RCP<RepartitionFactory> Rep11Factory = Teuchos::rcp(new RepartitionFactory);
-    Rep11Factory->SetFactory("A", A11Fact);
-    Rep11Factory->SetFactory("number of partitions", RepHeuFact);
-    Rep11Factory->SetFactory("Partition",Rep11Interface);
-
-    RCP<CloneRepartitionInterface> Rep22Interface = Teuchos::rcp(new CloneRepartitionInterface());
-    Rep22Interface->SetFactory("A", A22Fact);
-    Rep22Interface->SetFactory("Partition",Rep11Interface);
-
-    RCP<RepartitionFactory> Rep22Factory = Teuchos::rcp(new RepartitionFactory);
-    Rep22Factory->SetFactory("A", A22Fact);
-    Rep22Factory->SetFactory("number of partitions", RepHeuFact);
-    Rep22Factory->SetFactory("Partition",Rep22Interface);
-
-    // set up factory manager
-    RCP<FactoryManager> FC1 = rcp(new FactoryManager());
-    FC1->SetFactory("A", A11Fact);
-    FC1->SetFactory("Importer", Rep11Factory);
-    FC1->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    RCP<FactoryManager> FC2 = rcp(new FactoryManager());
-    FC2->SetFactory("A", A22Fact);
-    FC2->SetFactory("Importer", Rep22Factory);
-    FC2->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    /////////////////////////////////////////// define blocked transfer ops
-    RCP<RebalanceBlockAcFactory> RebAFact = rcp(new RebalanceBlockAcFactory());
-    RebAFact->SetParameter("repartition: use subcommunicators",Teuchos::ParameterEntry(false));
-    RebAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-    RebAFact->AddFactoryManager(FC1);
-    RebAFact->AddFactoryManager(FC2);
-
-    /////////////////////////////////////////// request rebalanced coarse level matrix A
-    levelTwo->Request("A", RebAFact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("A", RebAFact.get()),true);
-
-    // request Partition data
-    levelTwo->Request("Partition", Rep11Interface.get(), MueLu::NoFactory::get());
-    levelTwo->Request("Partition", Rep22Interface.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep11Interface.get()),true);
-    TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep22Interface.get()),true);
-
-    /////////////////////////////////////////// build rebalanced coarse level matrix A
-    RCP<Matrix> rebA = levelTwo->Get<RCP<Matrix> >("A",RebAFact.get());
-    TEST_EQUALITY(rebA!=Teuchos::null,true);
-    // get number of active processes used in rebalanced matrix
-    std::vector<int> amActive  = std::vector<int>(comm->getSize(),0);
-    std::vector<int> areActive = std::vector<int>(comm->getSize(),0);
-    if(rebA->getLocalNumEntries() > 0) amActive[comm->getRank()] = 1;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX,comm->getSize(),&amActive[0],&areActive[0]);
-    int nNumProcsReb = 0;
-    for(int p = 0; p < comm->getSize(); p++)
-      nNumProcsReb += areActive[p];
-    if(nNumProcsReb == nNumProcs) {
-      out << "Skip detailed tests. Matrix was not rebalanced" << std::endl;
-      return;
-    }
-    int expectedPartitions=2;
-    TEST_EQUALITY(nNumProcsReb, expectedPartitions);
+  RCP<RepartitionFactory> Rep11Factory = Teuchos::rcp(new RepartitionFactory);
+  Rep11Factory->SetFactory("A", A11Fact);
+  Rep11Factory->SetFactory("number of partitions", RepHeuFact);
+  Rep11Factory->SetFactory("Partition", Rep11Interface);
+
+  RCP<CloneRepartitionInterface> Rep22Interface = Teuchos::rcp(new CloneRepartitionInterface());
+  Rep22Interface->SetFactory("A", A22Fact);
+  Rep22Interface->SetFactory("Partition", Rep11Interface);
+
+  RCP<RepartitionFactory> Rep22Factory = Teuchos::rcp(new RepartitionFactory);
+  Rep22Factory->SetFactory("A", A22Fact);
+  Rep22Factory->SetFactory("number of partitions", RepHeuFact);
+  Rep22Factory->SetFactory("Partition", Rep22Interface);
+
+  // set up factory manager
+  RCP<FactoryManager> FC1 = rcp(new FactoryManager());
+  FC1->SetFactory("A", A11Fact);
+  FC1->SetFactory("Importer", Rep11Factory);
+  FC1->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+
+  RCP<FactoryManager> FC2 = rcp(new FactoryManager());
+  FC2->SetFactory("A", A22Fact);
+  FC2->SetFactory("Importer", Rep22Factory);
+  FC2->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+
+  /////////////////////////////////////////// define blocked transfer ops
+  RCP<RebalanceBlockAcFactory> RebAFact = rcp(new RebalanceBlockAcFactory());
+  RebAFact->SetParameter("repartition: use subcommunicators", Teuchos::ParameterEntry(false));
+  RebAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  RebAFact->AddFactoryManager(FC1);
+  RebAFact->AddFactoryManager(FC2);
+
+  /////////////////////////////////////////// request rebalanced coarse level matrix A
+  levelTwo->Request("A", RebAFact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("A", RebAFact.get()), true);
+
+  // request Partition data
+  levelTwo->Request("Partition", Rep11Interface.get(), MueLu::NoFactory::get());
+  levelTwo->Request("Partition", Rep22Interface.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep11Interface.get()), true);
+  TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep22Interface.get()), true);
+
+  /////////////////////////////////////////// build rebalanced coarse level matrix A
+  RCP<Matrix> rebA = levelTwo->Get<RCP<Matrix> >("A", RebAFact.get());
+  TEST_EQUALITY(rebA != Teuchos::null, true);
+  // get number of active processes used in rebalanced matrix
+  std::vector<int> amActive  = std::vector<int>(comm->getSize(), 0);
+  std::vector<int> areActive = std::vector<int>(comm->getSize(), 0);
+  if (rebA->getLocalNumEntries() > 0) amActive[comm->getRank()] = 1;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, comm->getSize(), &amActive[0], &areActive[0]);
+  int nNumProcsReb = 0;
+  for (int p = 0; p < comm->getSize(); p++)
+    nNumProcsReb += areActive[p];
+  if (nNumProcsReb == nNumProcs) {
+    out << "Skip detailed tests. Matrix was not rebalanced" << std::endl;
+    return;
+  }
+  int expectedPartitions = 2;
+  TEST_EQUALITY(nNumProcsReb, expectedPartitions);
+
+  //////////////////////////////////////////////////
+  // extract partitions
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > part1 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", Rep11Interface.get());
+  TEST_EQUALITY(part1 != Teuchos::null, true);
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > part2 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", Rep22Interface.get());
+  TEST_EQUALITY(part2 != Teuchos::null, true);
+  TEST_EQUALITY(part1->getGlobalLength(), part2->getGlobalLength());
+  TEST_EQUALITY(part1->getLocalLength(), part2->getLocalLength());
+
+  Teuchos::ArrayRCP<const GO> part1_data = part1->getData(0);
+  Teuchos::ArrayRCP<const GO> part2_data = part2->getData(0);
+  for (size_t i = 0; i < part1->getLocalLength(); i++) {
+    TEST_EQUALITY(part1->getMap()->getGlobalElement(i) + 200, part2->getMap()->getGlobalElement(i));
+    TEST_EQUALITY(part1_data[i], part2_data[i]);
+  }
 
-    //////////////////////////////////////////////////
-    // extract partitions
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > part1 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition",Rep11Interface.get());
-    TEST_EQUALITY(part1!=Teuchos::null,true);
-    RCP<Xpetra::Vector<GO, LO, GO, NO> > part2 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition",Rep22Interface.get());
-    TEST_EQUALITY(part2!=Teuchos::null,true);
-    TEST_EQUALITY(part1->getGlobalLength(),part2->getGlobalLength());
-    TEST_EQUALITY(part1->getLocalLength(),part2->getLocalLength());
-
-    Teuchos::ArrayRCP< const GO > part1_data = part1->getData(0);
-    Teuchos::ArrayRCP< const GO > part2_data = part2->getData(0);
-    for(size_t i = 0; i < part1->getLocalLength(); i++) {
-      TEST_EQUALITY(part1->getMap()->getGlobalElement(i)+200,part2->getMap()->getGlobalElement(i));
-      TEST_EQUALITY(part1_data[i],part2_data[i]);
-    }
+  // check rebalanced operator
+  RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> >(rebA);
+  TEST_EQUALITY(bA != Teuchos::null, true);
+  TEST_EQUALITY(bA->Rows(), 2);
+  TEST_EQUALITY(bA->Cols(), 2);
+
+  RCP<Vector> rtest11 = VectorFactory::Build(bA->getMatrix(0, 0)->getDomainMap());
+  rtest11->putScalar(1.0);
+  RCP<Vector> rtest12 = VectorFactory::Build(bA->getMatrix(0, 1)->getDomainMap());
+  rtest12->putScalar(1.0);
+  RCP<Vector> rtest21 = VectorFactory::Build(bA->getMatrix(1, 0)->getDomainMap());
+  rtest21->putScalar(1.0);
+  RCP<Vector> rtest22 = VectorFactory::Build(bA->getMatrix(1, 1)->getDomainMap());
+  rtest22->putScalar(1.0);
+  RCP<Vector> rres11 = VectorFactory::Build(bA->getMatrix(0, 0)->getRangeMap());
+  RCP<Vector> rres12 = VectorFactory::Build(bA->getMatrix(0, 1)->getRangeMap());
+  RCP<Vector> rres21 = VectorFactory::Build(bA->getMatrix(1, 0)->getRangeMap());
+  RCP<Vector> rres22 = VectorFactory::Build(bA->getMatrix(1, 1)->getRangeMap());
+  bA->getMatrix(0, 0)->apply(*rtest11, *rres11);
+  bA->getMatrix(0, 1)->apply(*rtest12, *rres12);
+  bA->getMatrix(1, 0)->apply(*rtest21, *rres21);
+  bA->getMatrix(1, 1)->apply(*rtest22, *rres22);
+  TEST_EQUALITY(res11->norm1(), rres11->norm1());
+  TEST_EQUALITY(res12->norm1(), rres12->norm1());
+  TEST_EQUALITY(res21->norm1(), rres21->norm1());
+  TEST_EQUALITY(res22->norm1(), rres22->norm1());
+  TEST_EQUALITY(bA->getGlobalNumEntries(), 2392);
+}  // BlockedRAPFactory
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryWithRestriction, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // check rebalanced operator
-    RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> >(rebA);
-    TEST_EQUALITY(bA!=Teuchos::null,true);
-    TEST_EQUALITY(bA->Rows(),2);
-    TEST_EQUALITY(bA->Cols(),2);
-
-    RCP<Vector> rtest11  = VectorFactory::Build(bA->getMatrix(0,0)->getDomainMap()); rtest11->putScalar(1.0);
-    RCP<Vector> rtest12  = VectorFactory::Build(bA->getMatrix(0,1)->getDomainMap()); rtest12->putScalar(1.0);
-    RCP<Vector> rtest21  = VectorFactory::Build(bA->getMatrix(1,0)->getDomainMap()); rtest21->putScalar(1.0);
-    RCP<Vector> rtest22  = VectorFactory::Build(bA->getMatrix(1,1)->getDomainMap()); rtest22->putScalar(1.0);
-    RCP<Vector> rres11   = VectorFactory::Build(bA->getMatrix(0,0)->getRangeMap());
-    RCP<Vector> rres12   = VectorFactory::Build(bA->getMatrix(0,1)->getRangeMap());
-    RCP<Vector> rres21   = VectorFactory::Build(bA->getMatrix(1,0)->getRangeMap());
-    RCP<Vector> rres22   = VectorFactory::Build(bA->getMatrix(1,1)->getRangeMap());
-    bA->getMatrix(0,0)->apply(*rtest11,*rres11);
-    bA->getMatrix(0,1)->apply(*rtest12,*rres12);
-    bA->getMatrix(1,0)->apply(*rtest21,*rres21);
-    bA->getMatrix(1,1)->apply(*rtest22,*rres22);
-    TEST_EQUALITY(res11->norm1(), rres11->norm1());
-    TEST_EQUALITY(res12->norm1(), rres12->norm1());
-    TEST_EQUALITY(res21->norm1(), rres21->norm1());
-    TEST_EQUALITY(res22->norm1(), rres22->norm1());
-    TEST_EQUALITY(bA->getGlobalNumEntries(), 2392);
-  } //BlockedRAPFactory
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryWithRestriction, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  if (comm->getSize() == 1) {
+    out << "Skip BlockedRepartion test in serial case" << std::endl;
+    return;
+  }
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  /**********************************************************************************/
+  /* CREATE INITIAL MATRIX                                                          */
+  /**********************************************************************************/
+  RCP<const Map> bigMap;
+  RCP<const Map> map1;
+  RCP<const Map> map2;
+  GO numElements  = 400;
+  GO numElements1 = 200;
+  GO numElements2 = 200;
 
-    if(comm->getSize() == 1) {
-      out << "Skip BlockedRepartion test in serial case" << std::endl;
-      return;
-    }
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // the test matrix has to be a nxn block matrix with quadratic blocks
+  // where the subblocks use consequent numbering of global DOF ids.
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(1);
+
+  map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
 
-    /**********************************************************************************/
-    /* CREATE INITIAL MATRIX                                                          */
-    /**********************************************************************************/
-    RCP<const Map> bigMap;
-    RCP<const Map> map1;
-    RCP<const Map> map2;
-    GO numElements = 400;
-    GO numElements1 = 200;
-    GO numElements2 = 200;
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // the test matrix has to be a nxn block matrix with quadratic blocks
-    // where the subblocks use consequent numbering of global DOF ids.
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(1);
-
-    map1   = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-    map2   = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
-
-    std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-    Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
-    localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-    Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
-    localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-    bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
-    std::vector<Teuchos::RCP<const Map> > maps;
-    maps.push_back(map1); maps.push_back(map2);
-
-    Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(bigMap, maps);
-
-    RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map1,2,-1,-1);
-    RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map2,1, 0, 0);
-    RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map1,1, 0, 0);
-    RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map2,3,-2,-1);
-
-    // store output of simple MV products for OpIJ
-    RCP<Vector> test11  = VectorFactory::Build(Op11->getDomainMap()); test11->putScalar(1.0);
-    RCP<Vector> test12  = VectorFactory::Build(Op12->getDomainMap()); test12->putScalar(1.0);
-    RCP<Vector> test21  = VectorFactory::Build(Op21->getDomainMap()); test21->putScalar(1.0);
-    RCP<Vector> test22  = VectorFactory::Build(Op22->getDomainMap()); test22->putScalar(1.0);
-    RCP<Vector> res11   = VectorFactory::Build(Op11->getRangeMap());
-    RCP<Vector> res12   = VectorFactory::Build(Op12->getRangeMap());
-    RCP<Vector> res21   = VectorFactory::Build(Op21->getRangeMap());
-    RCP<Vector> res22   = VectorFactory::Build(Op22->getRangeMap());
-    Op11->apply(*test11,*res11);
-    Op12->apply(*test12,*res12);
-    Op21->apply(*test21,*res21);
-    Op22->apply(*test22,*res22);
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType res11norm = res11->norm1();
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType res12norm = res12->norm1();
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType res21norm = res21->norm1();
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType res22norm = res22->norm1();
-
-    // build blocked operator
-    Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(mapExtractor,mapExtractor,10));
-
-    bOp->setMatrix(0,0,Op11);
-    bOp->setMatrix(0,1,Op12);
-    bOp->setMatrix(1,0,Op21);
-    bOp->setMatrix(1,1,Op22);
-    bOp->fillComplete();
-    TEST_EQUALITY(bOp!=Teuchos::null, true);
-    TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
-
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    levelOne->SetLevelID(-1);
-    RCP<Level> levelTwo = rcp(new Level()); levelTwo->SetPreviousLevel(levelOne);
-    levelTwo->SetLevelID(0);
-    levelTwo->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-
-    // define repartition heuristics
-    RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
-    RepHeuFact->SetFactory("A", MueLu::NoFactory::getRCP()); // 2x2 blocked operator
-    RepHeuFact->SetParameter("repartition: start level",Teuchos::ParameterEntry(0));
-    RepHeuFact->SetParameter("repartition: min rows per proc",Teuchos::ParameterEntry(200));
-
-    // define sub block factories for blocked operator "A"
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
-
-    RCP<AmalgamationFactory> Amalg11Fact = Teuchos::rcp(new AmalgamationFactory());
-    RCP<AmalgamationFactory> Amalg22Fact = Teuchos::rcp(new AmalgamationFactory());
-    Amalg11Fact->SetFactory("A", A11Fact);
-    Amalg22Fact->SetFactory("A", A22Fact);
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
+  localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
+  localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
+  std::vector<Teuchos::RCP<const Map> > maps;
+  maps.push_back(map1);
+  maps.push_back(map2);
+
+  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(bigMap, maps);
+
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map2, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map1, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map2, 3, -2, -1);
+
+  // store output of simple MV products for OpIJ
+  RCP<Vector> test11 = VectorFactory::Build(Op11->getDomainMap());
+  test11->putScalar(1.0);
+  RCP<Vector> test12 = VectorFactory::Build(Op12->getDomainMap());
+  test12->putScalar(1.0);
+  RCP<Vector> test21 = VectorFactory::Build(Op21->getDomainMap());
+  test21->putScalar(1.0);
+  RCP<Vector> test22 = VectorFactory::Build(Op22->getDomainMap());
+  test22->putScalar(1.0);
+  RCP<Vector> res11 = VectorFactory::Build(Op11->getRangeMap());
+  RCP<Vector> res12 = VectorFactory::Build(Op12->getRangeMap());
+  RCP<Vector> res21 = VectorFactory::Build(Op21->getRangeMap());
+  RCP<Vector> res22 = VectorFactory::Build(Op22->getRangeMap());
+  Op11->apply(*test11, *res11);
+  Op12->apply(*test12, *res12);
+  Op21->apply(*test21, *res21);
+  Op22->apply(*test22, *res22);
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType res11norm = res11->norm1();
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType res12norm = res12->norm1();
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType res21norm = res21->norm1();
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType res22norm = res22->norm1();
+
+  // build blocked operator
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(mapExtractor, mapExtractor, 10));
+
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(0, 1, Op12);
+  bOp->setMatrix(1, 0, Op21);
+  bOp->setMatrix(1, 1, Op22);
+  bOp->fillComplete();
+  TEST_EQUALITY(bOp != Teuchos::null, true);
+  TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
+
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  levelOne->SetLevelID(-1);
+  RCP<Level> levelTwo = rcp(new Level());
+  levelTwo->SetPreviousLevel(levelOne);
+  levelTwo->SetLevelID(0);
+  levelTwo->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+
+  // define repartition heuristics
+  RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
+  RepHeuFact->SetFactory("A", MueLu::NoFactory::getRCP());  // 2x2 blocked operator
+  RepHeuFact->SetParameter("repartition: start level", Teuchos::ParameterEntry(0));
+  RepHeuFact->SetParameter("repartition: min rows per proc", Teuchos::ParameterEntry(200));
+
+  // define sub block factories for blocked operator "A"
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+
+  RCP<AmalgamationFactory> Amalg11Fact = Teuchos::rcp(new AmalgamationFactory());
+  RCP<AmalgamationFactory> Amalg22Fact = Teuchos::rcp(new AmalgamationFactory());
+  Amalg11Fact->SetFactory("A", A11Fact);
+  Amalg22Fact->SetFactory("A", A22Fact);
 
 #ifdef HAVE_MUELU_ISORROPIA
-    RCP<Factory> Rep11Interface = Teuchos::null;
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
-      RCP<IsorropiaInterface> Iso11Interface = Teuchos::rcp(new IsorropiaInterface());
-      Iso11Interface->SetFactory("A", A11Fact);
-      Iso11Interface->SetFactory("number of partitions", RepHeuFact);
-      Iso11Interface->SetFactory("UnAmalgamationInfo", Amalg11Fact);
-
-      Rep11Interface = Teuchos::rcp(new RepartitionInterface());
-      Rep11Interface->SetFactory("A", A11Fact);
-      Rep11Interface->SetFactory("number of partitions", RepHeuFact);
-      Rep11Interface->SetFactory("AmalgamatedPartition", Iso11Interface);
-    } else {
-      // we are in Tpetra mode (even though Isorropia would be available)
-      // create dummy "Partition" array
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
-      ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-      for(size_t r=0; r<decomposition->getMap()->getLocalNumElements(); r++) {
-        if(r%2 == 0) decompEntries[r] = 0;
-        else decompEntries[r] = 1;
-      }
-      Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
-      levelTwo->Request("Partition", Rep11Interface.get());
-      levelTwo->Set("Partition", decomposition, Rep11Interface.get());
-    }
-#else
+  RCP<Factory> Rep11Interface = Teuchos::null;
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    RCP<IsorropiaInterface> Iso11Interface = Teuchos::rcp(new IsorropiaInterface());
+    Iso11Interface->SetFactory("A", A11Fact);
+    Iso11Interface->SetFactory("number of partitions", RepHeuFact);
+    Iso11Interface->SetFactory("UnAmalgamationInfo", Amalg11Fact);
+
+    Rep11Interface = Teuchos::rcp(new RepartitionInterface());
+    Rep11Interface->SetFactory("A", A11Fact);
+    Rep11Interface->SetFactory("number of partitions", RepHeuFact);
+    Rep11Interface->SetFactory("AmalgamatedPartition", Iso11Interface);
+  } else {
+    // we are in Tpetra mode (even though Isorropia would be available)
     // create dummy "Partition" array
     RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
-    ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-    for(size_t r=0; r<decomposition->getMap()->getLocalNumElements(); r++) {
-      if(r%2 == 0) decompEntries[r] = 0;
-      else decompEntries[r] = 1;
+    ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
+    for (size_t r = 0; r < decomposition->getMap()->getLocalNumElements(); r++) {
+      if (r % 2 == 0)
+        decompEntries[r] = 0;
+      else
+        decompEntries[r] = 1;
     }
-    RCP<SubBlockAFactory> Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
+    Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
     levelTwo->Request("Partition", Rep11Interface.get());
     levelTwo->Set("Partition", decomposition, Rep11Interface.get());
+  }
+#else
+  // create dummy "Partition" array
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = Xpetra::VectorFactory<GO, LO, GO, NO>::Build(Op11->getRowMap(), false);
+  ArrayRCP<GO> decompEntries                         = decomposition->getDataNonConst(0);
+  for (size_t r = 0; r < decomposition->getMap()->getLocalNumElements(); r++) {
+    if (r % 2 == 0)
+      decompEntries[r] = 0;
+    else
+      decompEntries[r] = 1;
+  }
+  RCP<SubBlockAFactory> Rep11Interface = Teuchos::rcp(new SubBlockAFactory());
+  levelTwo->Request("Partition", Rep11Interface.get());
+  levelTwo->Set("Partition", decomposition, Rep11Interface.get());
 #endif
 
-    RCP<RepartitionFactory> Rep11Factory = Teuchos::rcp(new RepartitionFactory);
-    Rep11Factory->SetFactory("A", A11Fact);
-    Rep11Factory->SetFactory("number of partitions", RepHeuFact);
-    Rep11Factory->SetFactory("Partition",Rep11Interface);
-
-    RCP<CloneRepartitionInterface> Rep22Interface = Teuchos::rcp(new CloneRepartitionInterface());
-    Rep22Interface->SetFactory("A", A22Fact);
-    Rep22Interface->SetFactory("Partition",Rep11Interface);
-
-    RCP<RepartitionFactory> Rep22Factory = Teuchos::rcp(new RepartitionFactory);
-    Rep22Factory->SetFactory("A", A22Fact);
-    Rep22Factory->SetFactory("number of partitions", RepHeuFact);
-    Rep22Factory->SetFactory("Partition",Rep22Interface);
-
-    // set up factory manager
-    RCP<FactoryManager> FC1 = rcp(new FactoryManager());
-    FC1->SetFactory("A", A11Fact);
-    FC1->SetFactory("Importer", Rep11Factory);
-    FC1->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    RCP<FactoryManager> FC2 = rcp(new FactoryManager());
-    FC2->SetFactory("A", A22Fact);
-    FC2->SetFactory("Importer", Rep22Factory);
-    FC2->SetIgnoreUserData(true);               // always use data from factories defined in factory manager
-
-    /////////////////////////////////////////// define blocked transfer ops
-    RCP<RebalanceBlockAcFactory> RebAFact = rcp(new RebalanceBlockAcFactory());
-    RebAFact->SetParameter("repartition: use subcommunicators",Teuchos::ParameterEntry(true));
-    RebAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-    RebAFact->AddFactoryManager(FC1);
-    RebAFact->AddFactoryManager(FC2);
-
-    /////////////////////////////////////////// request rebalanced coarse level matrix A
-    levelTwo->Request("A", RebAFact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("A", RebAFact.get()),true);
-
-    // request Partition data
-    levelTwo->Request("Partition", Rep11Interface.get(), MueLu::NoFactory::get());
-    levelTwo->Request("Partition", Rep22Interface.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep11Interface.get()),true);
-    TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep22Interface.get()),true);
-
-    /////////////////////////////////////////// build rebalanced coarse level matrix A
-    RCP<Matrix> rebA = levelTwo->Get<RCP<Matrix> >("A",RebAFact.get());
-
-    if(rebA != Teuchos::null) {
-      TEST_EQUALITY(rebA->getRowMap()->getComm()->getSize(), 2);
-
-      //////////////////////////////////////////////////
-      // extract partitions
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > part1 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition",Rep11Interface.get());
-      TEST_EQUALITY(part1!=Teuchos::null,true);
-      RCP<Xpetra::Vector<GO, LO, GO, NO> > part2 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition",Rep22Interface.get());
-      TEST_EQUALITY(part2!=Teuchos::null,true);
-      TEST_EQUALITY(part1->getGlobalLength(),part2->getGlobalLength());
-      TEST_EQUALITY(part1->getLocalLength(),part2->getLocalLength());
-
-      Teuchos::ArrayRCP< const GO > part1_data = part1->getData(0);
-      Teuchos::ArrayRCP< const GO > part2_data = part2->getData(0);
-      for(size_t i = 0; i < part1->getLocalLength(); i++) {
-        TEST_EQUALITY(part1->getMap()->getGlobalElement(i)+200,part2->getMap()->getGlobalElement(i));
-        TEST_EQUALITY(part1_data[i],part2_data[i]);
-      }
-
-      // check rebalanced operator
-      RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> >(rebA);
-      TEST_EQUALITY(bA!=Teuchos::null,true);
-      TEST_EQUALITY(bA->Rows(),2);
-      TEST_EQUALITY(bA->Cols(),2);
-
-      RCP<Vector> rtest11  = VectorFactory::Build(bA->getMatrix(0,0)->getDomainMap()); rtest11->putScalar(1.0);
-      RCP<Vector> rtest12  = VectorFactory::Build(bA->getMatrix(0,1)->getDomainMap()); rtest12->putScalar(1.0);
-      RCP<Vector> rtest21  = VectorFactory::Build(bA->getMatrix(1,0)->getDomainMap()); rtest21->putScalar(1.0);
-      RCP<Vector> rtest22  = VectorFactory::Build(bA->getMatrix(1,1)->getDomainMap()); rtest22->putScalar(1.0);
-      RCP<Vector> rres11   = VectorFactory::Build(bA->getMatrix(0,0)->getRangeMap());
-      RCP<Vector> rres12   = VectorFactory::Build(bA->getMatrix(0,1)->getRangeMap());
-      RCP<Vector> rres21   = VectorFactory::Build(bA->getMatrix(1,0)->getRangeMap());
-      RCP<Vector> rres22   = VectorFactory::Build(bA->getMatrix(1,1)->getRangeMap());
-      bA->getMatrix(0,0)->apply(*rtest11,*rres11);
-      bA->getMatrix(0,1)->apply(*rtest12,*rres12);
-      bA->getMatrix(1,0)->apply(*rtest21,*rres21);
-      bA->getMatrix(1,1)->apply(*rtest22,*rres22);
-      TEST_EQUALITY(res11norm, rres11->norm1());
-      TEST_EQUALITY(res12norm, rres12->norm1());
-      TEST_EQUALITY(res21norm, rres21->norm1());
-      TEST_EQUALITY(res22norm, rres22->norm1());
-      TEST_EQUALITY(bA->getGlobalNumEntries(), 2392);
-    } // end if rebA != Teuchos::null
-  } //BlockedRAPFactoryWithRestriction
+  RCP<RepartitionFactory> Rep11Factory = Teuchos::rcp(new RepartitionFactory);
+  Rep11Factory->SetFactory("A", A11Fact);
+  Rep11Factory->SetFactory("number of partitions", RepHeuFact);
+  Rep11Factory->SetFactory("Partition", Rep11Interface);
+
+  RCP<CloneRepartitionInterface> Rep22Interface = Teuchos::rcp(new CloneRepartitionInterface());
+  Rep22Interface->SetFactory("A", A22Fact);
+  Rep22Interface->SetFactory("Partition", Rep11Interface);
+
+  RCP<RepartitionFactory> Rep22Factory = Teuchos::rcp(new RepartitionFactory);
+  Rep22Factory->SetFactory("A", A22Fact);
+  Rep22Factory->SetFactory("number of partitions", RepHeuFact);
+  Rep22Factory->SetFactory("Partition", Rep22Interface);
+
+  // set up factory manager
+  RCP<FactoryManager> FC1 = rcp(new FactoryManager());
+  FC1->SetFactory("A", A11Fact);
+  FC1->SetFactory("Importer", Rep11Factory);
+  FC1->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+
+  RCP<FactoryManager> FC2 = rcp(new FactoryManager());
+  FC2->SetFactory("A", A22Fact);
+  FC2->SetFactory("Importer", Rep22Factory);
+  FC2->SetIgnoreUserData(true);  // always use data from factories defined in factory manager
+
+  /////////////////////////////////////////// define blocked transfer ops
+  RCP<RebalanceBlockAcFactory> RebAFact = rcp(new RebalanceBlockAcFactory());
+  RebAFact->SetParameter("repartition: use subcommunicators", Teuchos::ParameterEntry(true));
+  RebAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  RebAFact->AddFactoryManager(FC1);
+  RebAFact->AddFactoryManager(FC2);
+
+  /////////////////////////////////////////// request rebalanced coarse level matrix A
+  levelTwo->Request("A", RebAFact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("A", RebAFact.get()), true);
+
+  // request Partition data
+  levelTwo->Request("Partition", Rep11Interface.get(), MueLu::NoFactory::get());
+  levelTwo->Request("Partition", Rep22Interface.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep11Interface.get()), true);
+  TEST_EQUALITY(levelTwo->IsRequested("Partition", Rep22Interface.get()), true);
+
+  /////////////////////////////////////////// build rebalanced coarse level matrix A
+  RCP<Matrix> rebA = levelTwo->Get<RCP<Matrix> >("A", RebAFact.get());
+
+  if (rebA != Teuchos::null) {
+    TEST_EQUALITY(rebA->getRowMap()->getComm()->getSize(), 2);
+
+    //////////////////////////////////////////////////
+    // extract partitions
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > part1 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", Rep11Interface.get());
+    TEST_EQUALITY(part1 != Teuchos::null, true);
+    RCP<Xpetra::Vector<GO, LO, GO, NO> > part2 = levelTwo->Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", Rep22Interface.get());
+    TEST_EQUALITY(part2 != Teuchos::null, true);
+    TEST_EQUALITY(part1->getGlobalLength(), part2->getGlobalLength());
+    TEST_EQUALITY(part1->getLocalLength(), part2->getLocalLength());
+
+    Teuchos::ArrayRCP<const GO> part1_data = part1->getData(0);
+    Teuchos::ArrayRCP<const GO> part2_data = part2->getData(0);
+    for (size_t i = 0; i < part1->getLocalLength(); i++) {
+      TEST_EQUALITY(part1->getMap()->getGlobalElement(i) + 200, part2->getMap()->getGlobalElement(i));
+      TEST_EQUALITY(part1_data[i], part2_data[i]);
+    }
 
+    // check rebalanced operator
+    RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bA = Teuchos::rcp_dynamic_cast<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> >(rebA);
+    TEST_EQUALITY(bA != Teuchos::null, true);
+    TEST_EQUALITY(bA->Rows(), 2);
+    TEST_EQUALITY(bA->Cols(), 2);
+
+    RCP<Vector> rtest11 = VectorFactory::Build(bA->getMatrix(0, 0)->getDomainMap());
+    rtest11->putScalar(1.0);
+    RCP<Vector> rtest12 = VectorFactory::Build(bA->getMatrix(0, 1)->getDomainMap());
+    rtest12->putScalar(1.0);
+    RCP<Vector> rtest21 = VectorFactory::Build(bA->getMatrix(1, 0)->getDomainMap());
+    rtest21->putScalar(1.0);
+    RCP<Vector> rtest22 = VectorFactory::Build(bA->getMatrix(1, 1)->getDomainMap());
+    rtest22->putScalar(1.0);
+    RCP<Vector> rres11 = VectorFactory::Build(bA->getMatrix(0, 0)->getRangeMap());
+    RCP<Vector> rres12 = VectorFactory::Build(bA->getMatrix(0, 1)->getRangeMap());
+    RCP<Vector> rres21 = VectorFactory::Build(bA->getMatrix(1, 0)->getRangeMap());
+    RCP<Vector> rres22 = VectorFactory::Build(bA->getMatrix(1, 1)->getRangeMap());
+    bA->getMatrix(0, 0)->apply(*rtest11, *rres11);
+    bA->getMatrix(0, 1)->apply(*rtest12, *rres12);
+    bA->getMatrix(1, 0)->apply(*rtest21, *rres21);
+    bA->getMatrix(1, 1)->apply(*rtest22, *rres22);
+    TEST_EQUALITY(res11norm, rres11->norm1());
+    TEST_EQUALITY(res12norm, rres12->norm1());
+    TEST_EQUALITY(res21norm, rres21->norm1());
+    TEST_EQUALITY(res22norm, rres22->norm1());
+    TEST_EQUALITY(bA->getGlobalNumEntries(), 2392);
+  }  // end if rebA != Teuchos::null
+}  // BlockedRAPFactoryWithRestriction
 
 #if defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MUELU_ZOLTAN)
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryWithDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryWithDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
   out << "version: " << MueLu::Version() << std::endl;
 
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-  if(comm->getSize() == 1) {
+  if (comm->getSize() == 1) {
     out << "Skip BlockedRepartion test in serial case" << std::endl;
     return;
   }
@@ -670,7 +694,7 @@ namespace MueLuTests {
   RCP<const Map> bigMap;
   RCP<const Map> map1;
   RCP<const Map> map2;
-  GO numElements = 400;
+  GO numElements  = 400;
   GO numElements1 = 200;
   GO numElements2 = 200;
 
@@ -681,124 +705,128 @@ namespace MueLuTests {
   std::vector<size_t> stridingInfo;
   stridingInfo.push_back(1);
 
-  map1   = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-  map2   = StridedMapFactory::Build(lib, numElements2, 0, stridingInfo, comm, -1, numElements1);
+  map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  map2 = StridedMapFactory::Build(lib, numElements2, 0, stridingInfo, comm, -1, numElements1);
 
-  std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-  Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
   localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-  Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
   localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
   std::vector<Teuchos::RCP<const Map> > maps;
-  maps.push_back(map1); maps.push_back(map2);
+  maps.push_back(map1);
+  maps.push_back(map2);
 
-  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(bigMap, maps);
+  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(bigMap, maps);
 
-  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map1,2,-1,-1);
-  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map2,1, 0, 0);
-  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map1,1, 0, 0);
-  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map2,3,-2,-1);
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map2, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map1, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map2, 3, -2, -1);
 
   // store output of simple MV products for OpIJ
-  RCP<Vector> test11  = VectorFactory::Build(Op11->getDomainMap()); test11->putScalar(1.0);
-  RCP<Vector> test12  = VectorFactory::Build(Op12->getDomainMap()); test12->putScalar(1.0);
-  RCP<Vector> test21  = VectorFactory::Build(Op21->getDomainMap()); test21->putScalar(1.0);
-  RCP<Vector> test22  = VectorFactory::Build(Op22->getDomainMap()); test22->putScalar(1.0);
-  RCP<Vector> res11   = VectorFactory::Build(Op11->getRangeMap());
-  RCP<Vector> res12   = VectorFactory::Build(Op12->getRangeMap());
-  RCP<Vector> res21   = VectorFactory::Build(Op21->getRangeMap());
-  RCP<Vector> res22   = VectorFactory::Build(Op22->getRangeMap());
-  Op11->apply(*test11,*res11);
-  Op12->apply(*test12,*res12);
-  Op21->apply(*test21,*res21);
-  Op22->apply(*test22,*res22);
+  RCP<Vector> test11 = VectorFactory::Build(Op11->getDomainMap());
+  test11->putScalar(1.0);
+  RCP<Vector> test12 = VectorFactory::Build(Op12->getDomainMap());
+  test12->putScalar(1.0);
+  RCP<Vector> test21 = VectorFactory::Build(Op21->getDomainMap());
+  test21->putScalar(1.0);
+  RCP<Vector> test22 = VectorFactory::Build(Op22->getDomainMap());
+  test22->putScalar(1.0);
+  RCP<Vector> res11 = VectorFactory::Build(Op11->getRangeMap());
+  RCP<Vector> res12 = VectorFactory::Build(Op12->getRangeMap());
+  RCP<Vector> res21 = VectorFactory::Build(Op21->getRangeMap());
+  RCP<Vector> res22 = VectorFactory::Build(Op22->getRangeMap());
+  Op11->apply(*test11, *res11);
+  Op12->apply(*test12, *res12);
+  Op21->apply(*test21, *res21);
+  Op22->apply(*test22, *res22);
 
   // build blocked operator
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(mapExtractor,mapExtractor,10));
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(mapExtractor, mapExtractor, 10));
 
-  bOp->setMatrix(0,0,Op11);
-  bOp->setMatrix(0,1,Op12);
-  bOp->setMatrix(1,0,Op21);
-  bOp->setMatrix(1,1,Op22);
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(0, 1, Op12);
+  bOp->setMatrix(1, 0, Op21);
+  bOp->setMatrix(1, 1, Op22);
   bOp->fillComplete();
-  TEST_EQUALITY(bOp!=Teuchos::null, true);
+  TEST_EQUALITY(bOp != Teuchos::null, true);
   TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
 
   // coordinates
-  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coord = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(bOp->getFullRangeMap(),1);
-  int PID = comm->getRank();
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coord = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(bOp->getFullRangeMap(), 1);
+  int PID                                                                                            = comm->getRank();
   // GH: scope data manipulation because we should not let a pointer to this data live once we call MueLu
   {
     Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coordData = coord->getDataNonConst(0);
-    for(size_t i = 0; i < (size_t) coordData.size(); ++i)
+    for (size_t i = 0; i < (size_t)coordData.size(); ++i)
       coordData[i] = PID + (typename Teuchos::ScalarTraits<Scalar>::magnitudeType)i / coordData.size();
   }
   // nullspace
-  RCP<MultiVector> nullspace = MultiVectorFactory::Build(bOp->getFullRangeMap(),1);
+  RCP<MultiVector> nullspace    = MultiVectorFactory::Build(bOp->getFullRangeMap(), 1);
   Teuchos::ArrayRCP<SC> nspData = nullspace->getDataNonConst(0);
-  for(size_t i=0; i<(size_t)nspData.size(); i++)
+  for (size_t i = 0; i < (size_t)nspData.size(); i++)
     nspData[i] = 1.0;
 
   // Grab sub-blocks for the Tobias-style goodies
-  RCP<MultiVector> nullspace1 = mapExtractor->ExtractVector(nullspace,0);
-  RCP<MultiVector> nullspace2 = mapExtractor->ExtractVector(nullspace,1);
-
+  RCP<MultiVector> nullspace1 = mapExtractor->ExtractVector(nullspace, 0);
+  RCP<MultiVector> nullspace2 = mapExtractor->ExtractVector(nullspace, 1);
 
   // build hierarchy
   Hierarchy H;
-  H.SetMaxCoarseSize(10);    // FIXME: At present this does not work past two levels
+  H.SetMaxCoarseSize(10);  // FIXME: At present this does not work past two levels
   RCP<Level> levelOne = H.GetLevel();
-  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-  levelOne->Set("Coordinates",coord);
-  levelOne->Set("Nullspace1",nullspace1);
-  levelOne->Set("Nullspace2",nullspace2);
+  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+  levelOne->Set("Coordinates", coord);
+  levelOne->Set("Nullspace1", nullspace1);
+  levelOne->Set("Nullspace2", nullspace2);
 
   // Define the operator's subblocks
   RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-  A11Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-  A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-  A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
   RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-  A22Fact->SetFactory("A",MueLu::NoFactory::getRCP());
-  A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-  A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
 
   // Create smoothers for each block
   std::string ifpackType;
   Teuchos::ParameterList ifpackList;
-  ifpackList.set("relaxation: sweeps", (LO) 5);
-  ifpackList.set("relaxation: damping factor", (SC) 1.0);
+  ifpackList.set("relaxation: sweeps", (LO)5);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
   ifpackType = "RELAXATION";
   ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-  RCP<SmootherPrototype> smoProto11     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+  RCP<SmootherPrototype> smoProto11 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
   smoProto11->SetFactory("A", A11Fact);
-  RCP<SmootherPrototype> smoProto22     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+  RCP<SmootherPrototype> smoProto22 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
   smoProto22->SetFactory("A", A22Fact);
 
-  RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
-  RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
+  RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
+  RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
 
   // Define how the transfer operators for block (0,0) is computed
   RCP<Factory> Amalg11  = rcp(new AmalgamationFactory());
   RCP<Factory> CDFact11 = rcp(new CoalesceDropFactory());
-  CDFact11->SetFactory("UnAmalgamationInfo",Amalg11);
+  CDFact11->SetFactory("UnAmalgamationInfo", Amalg11);
   RCP<Factory> AggFact11 = rcp(new UncoupledAggregationFactory());
-  AggFact11->SetFactory("Graph",CDFact11);
-  AggFact11->SetFactory("DofsPerNode",CDFact11);
+  AggFact11->SetFactory("Graph", CDFact11);
+  AggFact11->SetFactory("DofsPerNode", CDFact11);
   RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
-  P11Fact->SetFactory("Aggregates",AggFact11);
-  P11Fact->SetFactory("UnAmalgamationInfo",Amalg11);
+  P11Fact->SetFactory("Aggregates", AggFact11);
+  P11Fact->SetFactory("UnAmalgamationInfo", Amalg11);
   P11Fact->SetParameter("tentative: build coarse coordinates", Teuchos::ParameterEntry(false));
   RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
-  RCP<Factory> Nullspace11 = rcp(new NullspaceFactory());
+  RCP<Factory> Nullspace11   = rcp(new NullspaceFactory());
   Nullspace11->SetParameter("Fine level nullspace", Teuchos::ParameterEntry(std::string("Nullspace1")));
-  Nullspace11->SetFactory("Nullspace1",P11Fact);
+  Nullspace11->SetFactory("Nullspace1", P11Fact);
   RCP<Factory> Coord11 = rcp(new CoordinatesTransferFactory());
-  RCP<Factory> Cmap11 = rcp(new CoarseMapFactory());
-  Coord11->SetFactory("Aggregates",AggFact11);
-  Coord11->SetFactory("CoarseMap",Cmap11);
+  RCP<Factory> Cmap11  = rcp(new CoarseMapFactory());
+  Coord11->SetFactory("Aggregates", AggFact11);
+  Coord11->SetFactory("CoarseMap", Cmap11);
 
   // Create a factory manager for block (0,0)
   RCP<FactoryManager> M11 = rcp(new FactoryManager());
@@ -806,37 +834,37 @@ namespace MueLuTests {
   M11->SetFactory("Graph", CDFact11);
   M11->SetFactory("DofsPerNode", CDFact11);
   M11->SetFactory("Aggregates", AggFact11);
-  M11->SetFactory("UnAmalgamationInfo",Amalg11);
+  M11->SetFactory("UnAmalgamationInfo", Amalg11);
   M11->SetFactory("P", P11Fact);
   M11->SetFactory("R", R11Fact);
-  M11->SetFactory("Ptent", P11Fact); //for Nullspace
+  M11->SetFactory("Ptent", P11Fact);  // for Nullspace
   M11->SetFactory("Coordinates", Coord11);
   M11->SetFactory("Smoother", Smoo11Fact);
-  M11->SetFactory("CoarseMap",Cmap11);
-  M11->SetFactory("Nullspace",Nullspace11);
+  M11->SetFactory("CoarseMap", Cmap11);
+  M11->SetFactory("Nullspace", Nullspace11);
   M11->SetIgnoreUserData(true);
 
   // Define how the transfer operators for block (1,1) is computed
   RCP<Factory> Amalg22  = rcp(new AmalgamationFactory());
   RCP<Factory> CDFact22 = rcp(new CoalesceDropFactory());
-  CDFact22->SetFactory("UnAmalgamationInfo",Amalg22);
+  CDFact22->SetFactory("UnAmalgamationInfo", Amalg22);
   RCP<Factory> AggFact22 = rcp(new UncoupledAggregationFactory());
-  AggFact22->SetFactory("Graph",CDFact22);
-  AggFact22->SetFactory("DofsPerNode",CDFact22);
-  RCP<TentativePFactory>P22Fact = rcp(new TentativePFactory());
-  P22Fact->SetFactory("Aggregates",AggFact22);
-  P22Fact->SetFactory("UnAmalgamationInfo",Amalg22);
+  AggFact22->SetFactory("Graph", CDFact22);
+  AggFact22->SetFactory("DofsPerNode", CDFact22);
+  RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory());
+  P22Fact->SetFactory("Aggregates", AggFact22);
+  P22Fact->SetFactory("UnAmalgamationInfo", Amalg22);
   P22Fact->SetParameter("tentative: build coarse coordinates", Teuchos::ParameterEntry(false));
   RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
-  RCP<Factory> Nullspace22 = rcp(new NullspaceFactory());
+  RCP<Factory> Nullspace22   = rcp(new NullspaceFactory());
   Nullspace22->SetParameter("Fine level nullspace", Teuchos::ParameterEntry(std::string("Nullspace2")));
-  Nullspace22->SetFactory("Nullspace2",P22Fact);
+  Nullspace22->SetFactory("Nullspace2", P22Fact);
   RCP<Factory> Cmap22 = rcp(new BlockedCoarseMapFactory());
-  Cmap22->SetFactory("Aggregates",AggFact22);
-  Cmap22->SetFactory("CoarseMap",Cmap11);
+  Cmap22->SetFactory("Aggregates", AggFact22);
+  Cmap22->SetFactory("CoarseMap", Cmap11);
   RCP<Factory> Coord22 = rcp(new CoordinatesTransferFactory());
-  Coord22->SetFactory("Aggregates",AggFact22);
-  Coord22->SetFactory("CoarseMap",Cmap22);
+  Coord22->SetFactory("Aggregates", AggFact22);
+  Coord22->SetFactory("CoarseMap", Cmap22);
 
   // Create a factory manager for block (1,1)
   RCP<FactoryManager> M22 = rcp(new FactoryManager());
@@ -844,14 +872,14 @@ namespace MueLuTests {
   M22->SetFactory("Graph", CDFact22);
   M22->SetFactory("DofsPerNode", CDFact22);
   M22->SetFactory("Aggregates", AggFact22);
-  M22->SetFactory("UnAmalgamationInfo",Amalg22);
+  M22->SetFactory("UnAmalgamationInfo", Amalg22);
   M22->SetFactory("P", P22Fact);
   M22->SetFactory("R", R22Fact);
-  M22->SetFactory("Ptent", P22Fact); //for Nullspace
+  M22->SetFactory("Ptent", P22Fact);  // for Nullspace
   M22->SetFactory("Coordinates", P22Fact);
   M22->SetFactory("Smoother", Smoo22Fact);
-  M22->SetFactory("CoarseMap",Cmap22);
-  M22->SetFactory("Nullspace",Nullspace22);
+  M22->SetFactory("CoarseMap", Cmap22);
+  M22->SetFactory("Nullspace", Nullspace22);
   M22->SetIgnoreUserData(true);
 
   RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
@@ -865,45 +893,44 @@ namespace MueLuTests {
   RCP<BlockedCoordinatesTransferFactory> CoordXfer = rcp(new BlockedCoordinatesTransferFactory());
   CoordXfer->AddFactory(Coord11);
   CoordXfer->AddFactory(Coord22);
-  CoordXfer->SetFactory("CoarseMap",PFact);
+  CoordXfer->SetFactory("CoarseMap", PFact);
 
   RCP<BlockedRAPFactory> AcFact = rcp(new BlockedRAPFactory());
   AcFact->AddTransferFactory(CoordXfer);
-  AcFact->SetFactory("P",PFact);
-  AcFact->SetFactory("R",RFact);
+  AcFact->SetFactory("P", PFact);
+  AcFact->SetFactory("R", RFact);
 
   // Smoothers
-  RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
+  RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
   smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(2));
   smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<SC>(1.0)));
-  smootherPrototype->AddFactoryManager(M11,0);
-  smootherPrototype->AddFactoryManager(M22,1);
-  RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+  smootherPrototype->AddFactoryManager(M11, 0);
+  smootherPrototype->AddFactoryManager(M22, 1);
+  RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
   // Coarse grid correction
-  RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp( new BlockedGaussSeidelSmoother() );
-  coarseSolverPrototype->AddFactoryManager(M11,0);
-  coarseSolverPrototype->AddFactoryManager(M22,1);
-  RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, Teuchos::null) );
+  RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp(new BlockedGaussSeidelSmoother());
+  coarseSolverPrototype->AddFactoryManager(M11, 0);
+  coarseSolverPrototype->AddFactoryManager(M22, 1);
+  RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, Teuchos::null));
 
   // Repartitioning
   RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
-  RepHeuFact->SetFactory("A",AcFact);//MueLu::NoFactory::getRCP());
-  RepHeuFact->SetParameter("repartition: start level",Teuchos::ParameterEntry(0));
-  RepHeuFact->SetParameter("repartition: min rows per proc",Teuchos::ParameterEntry(40));
+  RepHeuFact->SetFactory("A", AcFact);  // MueLu::NoFactory::getRCP());
+  RepHeuFact->SetParameter("repartition: start level", Teuchos::ParameterEntry(0));
+  RepHeuFact->SetParameter("repartition: min rows per proc", Teuchos::ParameterEntry(40));
 
   RCP<Factory> BDFact = rcp(new RepartitionBlockDiagonalFactory());
-  BDFact->SetFactory("A",AcFact);
-
+  BDFact->SetFactory("A", AcFact);
 
   RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-  zoltan->SetFactory("A",BDFact);
-  zoltan->SetFactory("number of partitions",RepHeuFact);
-  zoltan->SetFactory("Coordinates",CoordXfer);
+  zoltan->SetFactory("A", BDFact);
+  zoltan->SetFactory("number of partitions", RepHeuFact);
+  zoltan->SetFactory("Coordinates", CoordXfer);
 
   RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-  repart->SetFactory("A",BDFact);
-  repart->SetFactory("number of partitions",RepHeuFact);
+  repart->SetFactory("A", BDFact);
+  repart->SetFactory("number of partitions", RepHeuFact);
   repart->SetFactory("Partition", zoltan);
 
   RCP<RebalanceBlockAcFactory> RebAcFact = rcp(new RebalanceBlockAcFactory());
@@ -922,36 +949,32 @@ namespace MueLuTests {
 
   // main factory manager
   FactoryManager M;
-  M.SetFactory("A",            RebAcFact);
-  M.SetFactory("P",            PFact);
-  M.SetFactory("R",            RebRFact);
-  M.SetFactory("Coordinates",  CoordXfer);
-  M.SetFactory("Smoother",     smootherFact); // TODO fix me
+  M.SetFactory("A", RebAcFact);
+  M.SetFactory("P", PFact);
+  M.SetFactory("R", RebRFact);
+  M.SetFactory("Coordinates", CoordXfer);
+  M.SetFactory("Smoother", smootherFact);  // TODO fix me
   M.SetFactory("CoarseSolver", coarseSolverFact);
-  M.SetFactory("Nullspace",    RebRFact);
-
+  M.SetFactory("Nullspace", RebRFact);
 
   // setup hierarchy
   H.SetVerbLevel(MueLu::Test);
-  H.EnableGraphDumping("dep_graph",0);
+  H.EnableGraphDumping("dep_graph", 0);
 
   H.Setup(M);
 
-
-
-  } // BlockedRAPFactoryWithDiagonal
+}  // BlockedRAPFactoryWithDiagonal
 #endif
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryCoords, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedRepartition, BlockedRAPFactoryCoords, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
   out << "version: " << MueLu::Version() << std::endl;
 
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-  if(comm->getSize() == 1) {
+  if (comm->getSize() == 1) {
     out << "Skip BlockedRepartion test in serial case" << std::endl;
     return;
   }
@@ -962,7 +985,7 @@ namespace MueLuTests {
   RCP<const Map> bigMap;
   RCP<const Map> map1;
   RCP<const Map> map2;
-  GO numElements = 400;
+  GO numElements  = 400;
   GO numElements1 = 200;
   GO numElements2 = 200;
 
@@ -973,176 +996,180 @@ namespace MueLuTests {
   std::vector<size_t> stridingInfo;
   stridingInfo.push_back(2);
 
-  map1   = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-  map2   = StridedMapFactory::Build(lib, numElements2, 0, stridingInfo, comm, -1, numElements1);
+  map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  map2 = StridedMapFactory::Build(lib, numElements2, 0, stridingInfo, comm, -1, numElements1);
 
-  std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-  Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
   localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-  Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
   localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm); // create full big map (concatenation of map1 and map2)
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  bigMap = StridedMapFactory::Build(lib, numElements, eleList, 0, stridingInfo, comm);  // create full big map (concatenation of map1 and map2)
   std::vector<Teuchos::RCP<const Map> > maps;
-  maps.push_back(map1); maps.push_back(map2);
+  maps.push_back(map1);
+  maps.push_back(map2);
 
-  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar,LO,GO,Node>::Build(bigMap, maps);
+  Teuchos::RCP<const Xpetra::MapExtractor<Scalar, LO, GO, Node> > mapExtractor = Xpetra::MapExtractorFactory<Scalar, LO, GO, Node>::Build(bigMap, maps);
 
-  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map1,2,-1,-1);
-  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map1,map2,1, 0, 0);
-  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map1,1, 0, 0);
-  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar,LO,GO,Node>(map2,map2,3,-2,-1);
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op12 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, map2, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op21 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map1, 1, 0, 0);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, map2, 3, -2, -1);
 
   // store output of simple MV products for OpIJ
-  RCP<Vector> test11  = VectorFactory::Build(Op11->getDomainMap()); test11->putScalar(1.0);
-  RCP<Vector> test12  = VectorFactory::Build(Op12->getDomainMap()); test12->putScalar(1.0);
-  RCP<Vector> test21  = VectorFactory::Build(Op21->getDomainMap()); test21->putScalar(1.0);
-  RCP<Vector> test22  = VectorFactory::Build(Op22->getDomainMap()); test22->putScalar(1.0);
-  RCP<Vector> res11   = VectorFactory::Build(Op11->getRangeMap());
-  RCP<Vector> res12   = VectorFactory::Build(Op12->getRangeMap());
-  RCP<Vector> res21   = VectorFactory::Build(Op21->getRangeMap());
-  RCP<Vector> res22   = VectorFactory::Build(Op22->getRangeMap());
-  Op11->apply(*test11,*res11);
-  Op12->apply(*test12,*res12);
-  Op21->apply(*test21,*res21);
-  Op22->apply(*test22,*res22);
+  RCP<Vector> test11 = VectorFactory::Build(Op11->getDomainMap());
+  test11->putScalar(1.0);
+  RCP<Vector> test12 = VectorFactory::Build(Op12->getDomainMap());
+  test12->putScalar(1.0);
+  RCP<Vector> test21 = VectorFactory::Build(Op21->getDomainMap());
+  test21->putScalar(1.0);
+  RCP<Vector> test22 = VectorFactory::Build(Op22->getDomainMap());
+  test22->putScalar(1.0);
+  RCP<Vector> res11 = VectorFactory::Build(Op11->getRangeMap());
+  RCP<Vector> res12 = VectorFactory::Build(Op12->getRangeMap());
+  RCP<Vector> res21 = VectorFactory::Build(Op21->getRangeMap());
+  RCP<Vector> res22 = VectorFactory::Build(Op22->getRangeMap());
+  Op11->apply(*test11, *res11);
+  Op12->apply(*test12, *res12);
+  Op21->apply(*test21, *res21);
+  Op22->apply(*test22, *res22);
 
   // build blocked operator
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar,LO,GO,Node>(mapExtractor,mapExtractor,10));
+  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node> > bOp = Teuchos::rcp(new Xpetra::BlockedCrsMatrix<Scalar, LO, GO, Node>(mapExtractor, mapExtractor, 10));
 
-  bOp->setMatrix(0,0,Op11);
-  bOp->setMatrix(0,1,Op12);
-  bOp->setMatrix(1,0,Op21);
-  bOp->setMatrix(1,1,Op22);
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(0, 1, Op12);
+  bOp->setMatrix(1, 0, Op21);
+  bOp->setMatrix(1, 1, Op22);
   bOp->fillComplete();
-  TEST_EQUALITY(bOp!=Teuchos::null, true);
+  TEST_EQUALITY(bOp != Teuchos::null, true);
   TEST_EQUALITY(bOp->getGlobalNumEntries(), 2392);
 
   // coordinates
-  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coord = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(bOp->getFullRangeMap(),1);
-  int PID = comm->getRank();
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coord = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(bOp->getFullRangeMap(), 1);
+  int PID                                                                                            = comm->getRank();
   // GH: scope data manipulation because we should not let a pointer to this data live once we call MueLu
   {
     Teuchos::ArrayRCP<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> coordData = coord->getDataNonConst(0);
-    for(size_t i = 0; i < (size_t) coordData.size(); ++i)
+    for (size_t i = 0; i < (size_t)coordData.size(); ++i)
       coordData[i] = PID + (typename Teuchos::ScalarTraits<Scalar>::magnitudeType)i / coordData.size();
   }
 
   // nullspace
-  RCP<MultiVector> nullspace = MultiVectorFactory::Build(bOp->getFullRangeMap(),1);
+  RCP<MultiVector> nullspace    = MultiVectorFactory::Build(bOp->getFullRangeMap(), 1);
   Teuchos::ArrayRCP<SC> nspData = nullspace->getDataNonConst(0);
-  for(size_t i=0; i<(size_t)nspData.size(); i++)
+  for (size_t i = 0; i < (size_t)nspData.size(); i++)
     nspData[i] = 1.0;
 
   // Grab sub-blocks for the Tobias-style goodies
-  RCP<MultiVector> nullspace1 = mapExtractor->ExtractVector(nullspace,0);
-  RCP<MultiVector> nullspace2 = mapExtractor->ExtractVector(nullspace,1);
-
+  RCP<MultiVector> nullspace1 = mapExtractor->ExtractVector(nullspace, 0);
+  RCP<MultiVector> nullspace2 = mapExtractor->ExtractVector(nullspace, 1);
 
   // build hierarchy
   Hierarchy H;
-  H.SetMaxCoarseSize(10);    // FIXME: At present this does not work past two levels
+  H.SetMaxCoarseSize(10);  // FIXME: At present this does not work past two levels
   RCP<Level> levelOne = H.GetLevel();
-  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-  levelOne->Set("Coordinates",coord);
-  levelOne->Set("Nullspace1",nullspace1);
-  levelOne->Set("Nullspace2",nullspace2);
+  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+  levelOne->Set("Coordinates", coord);
+  levelOne->Set("Nullspace1", nullspace1);
+  levelOne->Set("Nullspace2", nullspace2);
 
   // Define the operator's subblocks
   RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-  A11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-  A11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
   RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-  A22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-  A22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
 
   // Create smoothers for each block
   std::string ifpackType;
   Teuchos::ParameterList ifpackList;
-  ifpackList.set("relaxation: sweeps", (LO) 5);
-  ifpackList.set("relaxation: damping factor", (SC) 1.0);
+  ifpackList.set("relaxation: sweeps", (LO)5);
+  ifpackList.set("relaxation: damping factor", (SC)1.0);
   ifpackType = "RELAXATION";
   ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-  RCP<SmootherPrototype> smoProto11     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+  RCP<SmootherPrototype> smoProto11 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
   smoProto11->SetFactory("A", A11Fact);
-  RCP<SmootherPrototype> smoProto22     = rcp( new TrilinosSmoother(ifpackType, ifpackList, 0) );
+  RCP<SmootherPrototype> smoProto22 = rcp(new TrilinosSmoother(ifpackType, ifpackList, 0));
   smoProto22->SetFactory("A", A22Fact);
 
-  RCP<SmootherFactory> Smoo11Fact = rcp( new SmootherFactory(smoProto11) );
-  RCP<SmootherFactory> Smoo22Fact = rcp( new SmootherFactory(smoProto22) );
+  RCP<SmootherFactory> Smoo11Fact = rcp(new SmootherFactory(smoProto11));
+  RCP<SmootherFactory> Smoo22Fact = rcp(new SmootherFactory(smoProto22));
 
   // Define how the transfer operators for block (0,0) is computed
   RCP<Factory> Amalg11  = rcp(new AmalgamationFactory());
   RCP<Factory> CDFact11 = rcp(new CoalesceDropFactory());
-  CDFact11->SetFactory("UnAmalgamationInfo",Amalg11);
+  CDFact11->SetFactory("UnAmalgamationInfo", Amalg11);
   RCP<Factory> AggFact11 = rcp(new UncoupledAggregationFactory());
-  AggFact11->SetFactory("Graph",CDFact11);
-  AggFact11->SetFactory("DofsPerNode",CDFact11);
-  RCP<Factory> Cmap11 = rcp(new CoarseMapFactory());
+  AggFact11->SetFactory("Graph", CDFact11);
+  AggFact11->SetFactory("DofsPerNode", CDFact11);
+  RCP<Factory> Cmap11            = rcp(new CoarseMapFactory());
   RCP<TentativePFactory> P11Fact = rcp(new TentativePFactory());
-  P11Fact->SetFactory("A",A11Fact);
-  P11Fact->SetFactory("UnAmalgamationInfo",Amalg11);
-  P11Fact->SetFactory("CoarseMap",Cmap11);
-  P11Fact->SetFactory("Aggregates",AggFact11);
+  P11Fact->SetFactory("A", A11Fact);
+  P11Fact->SetFactory("UnAmalgamationInfo", Amalg11);
+  P11Fact->SetFactory("CoarseMap", Cmap11);
+  P11Fact->SetFactory("Aggregates", AggFact11);
   P11Fact->SetParameter("tentative: build coarse coordinates", Teuchos::ParameterEntry(false));
   RCP<TransPFactory> R11Fact = rcp(new TransPFactory());
-  R11Fact->SetFactory("P",P11Fact);
+  R11Fact->SetFactory("P", P11Fact);
   RCP<Factory> Nullspace11 = rcp(new NullspaceFactory());
   Nullspace11->SetParameter("Fine level nullspace", Teuchos::ParameterEntry(std::string("Nullspace1")));
-  Nullspace11->SetFactory("Nullspace1",P11Fact);
+  Nullspace11->SetFactory("Nullspace1", P11Fact);
   RCP<Factory> Coord11 = rcp(new CoordinatesTransferFactory());
-  Coord11->SetFactory("Aggregates",AggFact11);
-  Coord11->SetFactory("CoarseMap",Cmap11);
+  Coord11->SetFactory("Aggregates", AggFact11);
+  Coord11->SetFactory("CoarseMap", Cmap11);
 
   // Create a factory manager for block (0,0)
   RCP<FactoryManager> M11 = rcp(new FactoryManager());
   M11->SetFactory("A", A11Fact);
-  M11->SetFactory("UnAmalgamationInfo",Amalg11);
+  M11->SetFactory("UnAmalgamationInfo", Amalg11);
   M11->SetFactory("Aggregates", AggFact11);
   M11->SetFactory("P", P11Fact);
   M11->SetFactory("R", R11Fact);
   M11->SetFactory("Coordinates", Coord11);
   M11->SetFactory("Smoother", Smoo11Fact);
-  M11->SetFactory("CoarseMap",Cmap11);
-  M11->SetFactory("Nullspace",Nullspace11);
+  M11->SetFactory("CoarseMap", Cmap11);
+  M11->SetFactory("Nullspace", Nullspace11);
   M11->SetIgnoreUserData(true);
 
   // Define how the transfer operators for block (1,1) is computed
   RCP<Factory> Amalg22  = rcp(new AmalgamationFactory());
   RCP<Factory> CDFact22 = rcp(new CoalesceDropFactory());
-  CDFact22->SetFactory("UnAmalgamationInfo",Amalg22);
+  CDFact22->SetFactory("UnAmalgamationInfo", Amalg22);
   RCP<Factory> AggFact22 = rcp(new UncoupledAggregationFactory());
-  AggFact22->SetFactory("Graph",CDFact22);
-  AggFact22->SetFactory("DofsPerNode",CDFact22);
-  RCP<Factory> Cmap22 = rcp(new BlockedCoarseMapFactory());
-  RCP<TentativePFactory>P22Fact = rcp(new TentativePFactory());
-  P22Fact->SetFactory("A",A22Fact);
-  P22Fact->SetFactory("UnAmalgamationInfo",Amalg22);
-  P22Fact->SetFactory("CoarseMap",Cmap22);
-  P22Fact->SetFactory("Aggregates",AggFact22);
+  AggFact22->SetFactory("Graph", CDFact22);
+  AggFact22->SetFactory("DofsPerNode", CDFact22);
+  RCP<Factory> Cmap22            = rcp(new BlockedCoarseMapFactory());
+  RCP<TentativePFactory> P22Fact = rcp(new TentativePFactory());
+  P22Fact->SetFactory("A", A22Fact);
+  P22Fact->SetFactory("UnAmalgamationInfo", Amalg22);
+  P22Fact->SetFactory("CoarseMap", Cmap22);
+  P22Fact->SetFactory("Aggregates", AggFact22);
   P22Fact->SetParameter("tentative: build coarse coordinates", Teuchos::ParameterEntry(false));
   RCP<TransPFactory> R22Fact = rcp(new TransPFactory());
-  R22Fact->SetFactory("P",P22Fact);
+  R22Fact->SetFactory("P", P22Fact);
   RCP<Factory> Nullspace22 = rcp(new NullspaceFactory());
   Nullspace22->SetParameter("Fine level nullspace", Teuchos::ParameterEntry(std::string("Nullspace2")));
-  Nullspace22->SetFactory("Nullspace2",P22Fact);
-  Cmap22->SetFactory("Aggregates",AggFact22);
-  Cmap22->SetFactory("CoarseMap",Cmap11);
+  Nullspace22->SetFactory("Nullspace2", P22Fact);
+  Cmap22->SetFactory("Aggregates", AggFact22);
+  Cmap22->SetFactory("CoarseMap", Cmap11);
   RCP<Factory> Coord22 = rcp(new CoordinatesTransferFactory());
-  Coord22->SetFactory("Aggregates",AggFact22);
-  Coord22->SetFactory("CoarseMap",Cmap22);
+  Coord22->SetFactory("Aggregates", AggFact22);
+  Coord22->SetFactory("CoarseMap", Cmap22);
 
   // Create a factory manager for block (1,1)
   RCP<FactoryManager> M22 = rcp(new FactoryManager());
   M22->SetFactory("A", A22Fact);
-  M22->SetFactory("UnAmalgamationInfo",Amalg22);
+  M22->SetFactory("UnAmalgamationInfo", Amalg22);
   M22->SetFactory("Aggregates", AggFact22);
   M22->SetFactory("P", P22Fact);
   M22->SetFactory("R", R22Fact);
   M22->SetFactory("Coordinates", P22Fact);
   M22->SetFactory("Smoother", Smoo22Fact);
-  M22->SetFactory("CoarseMap",Cmap22);
-  M22->SetFactory("Nullspace",Nullspace22);
+  M22->SetFactory("CoarseMap", Cmap22);
+  M22->SetFactory("Nullspace", Nullspace22);
   M22->SetIgnoreUserData(true);
 
   RCP<BlockedPFactory> PFact = rcp(new BlockedPFactory());
@@ -1156,81 +1183,81 @@ namespace MueLuTests {
   RCP<BlockedCoordinatesTransferFactory> CoordXfer = rcp(new BlockedCoordinatesTransferFactory());
   CoordXfer->AddFactory(Coord11);
   CoordXfer->AddFactory(Coord22);
-  CoordXfer->SetFactory("CoarseMap",PFact);
+  CoordXfer->SetFactory("CoarseMap", PFact);
 
   RCP<BlockedRAPFactory> AcFact = rcp(new BlockedRAPFactory());
-  AcFact->SetFactory("P",PFact);
-  AcFact->SetFactory("R",RFact);
+  AcFact->SetFactory("P", PFact);
+  AcFact->SetFactory("R", RFact);
   AcFact->AddTransferFactory(CoordXfer);
 
   // Smoothers
-  RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
+  RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
   smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(2));
   smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<SC>(1.0)));
-  smootherPrototype->AddFactoryManager(M11,0);
-  smootherPrototype->AddFactoryManager(M22,1);
-  RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+  smootherPrototype->AddFactoryManager(M11, 0);
+  smootherPrototype->AddFactoryManager(M22, 1);
+  RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
   // Coarse grid correction
-  RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp( new BlockedGaussSeidelSmoother() );
-  coarseSolverPrototype->AddFactoryManager(M11,0);
-  coarseSolverPrototype->AddFactoryManager(M22,1);
-  RCP<SmootherFactory>   coarseSolverFact      = rcp( new SmootherFactory(coarseSolverPrototype, Teuchos::null) );
+  RCP<BlockedGaussSeidelSmoother> coarseSolverPrototype = rcp(new BlockedGaussSeidelSmoother());
+  coarseSolverPrototype->AddFactoryManager(M11, 0);
+  coarseSolverPrototype->AddFactoryManager(M22, 1);
+  RCP<SmootherFactory> coarseSolverFact = rcp(new SmootherFactory(coarseSolverPrototype, Teuchos::null));
 
   // Repartitioning
   RCP<SubBlockAFactory> AR11Fact = Teuchos::rcp(new SubBlockAFactory());
-  AR11Fact->SetFactory("A",AcFact);
-  AR11Fact->SetParameter("block row",Teuchos::ParameterEntry(0));
-  AR11Fact->SetParameter("block col",Teuchos::ParameterEntry(0));
+  AR11Fact->SetFactory("A", AcFact);
+  AR11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  AR11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
   RCP<SubBlockAFactory> AR22Fact = Teuchos::rcp(new SubBlockAFactory());
-  AR22Fact->SetFactory("A",AcFact);
-  AR22Fact->SetParameter("block row",Teuchos::ParameterEntry(1));
-  AR22Fact->SetParameter("block col",Teuchos::ParameterEntry(1));
+  AR22Fact->SetFactory("A", AcFact);
+  AR22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  AR22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
 
   RCP<RepartitionHeuristicFactory> RepHeuFact = Teuchos::rcp(new RepartitionHeuristicFactory);
-  RepHeuFact->SetFactory("A",AcFact);
-  RepHeuFact->SetParameter("repartition: start level",Teuchos::ParameterEntry(0));
-  RepHeuFact->SetParameter("repartition: min rows per proc",Teuchos::ParameterEntry(40));
+  RepHeuFact->SetFactory("A", AcFact);
+  RepHeuFact->SetParameter("repartition: start level", Teuchos::ParameterEntry(0));
+  RepHeuFact->SetParameter("repartition: min rows per proc", Teuchos::ParameterEntry(40));
 
 #ifdef HAVE_MUELU_ZOLTAN2
   RCP<Zoltan2Interface> zoltan11 = rcp(new Zoltan2Interface());
-  zoltan11->SetFactory("A",AR11Fact);
-  zoltan11->SetFactory("number of partitions",RepHeuFact);
-  zoltan11->SetFactory("Coordinates",Coord11);
+  zoltan11->SetFactory("A", AR11Fact);
+  zoltan11->SetFactory("number of partitions", RepHeuFact);
+  zoltan11->SetFactory("Coordinates", Coord11);
 #endif
 
   RCP<RepartitionFactory> repart11 = rcp(new RepartitionFactory());
-  repart11->SetFactory("A",AR11Fact);
-  repart11->SetFactory("number of partitions",RepHeuFact);
+  repart11->SetFactory("A", AR11Fact);
+  repart11->SetFactory("number of partitions", RepHeuFact);
 #ifdef HAVE_MUELU_ZOLTAN2
   repart11->SetFactory("Partition", zoltan11);
 #endif
 
 #ifdef HAVE_MUELU_ZOLTAN2
   RCP<Zoltan2Interface> zoltan22 = rcp(new Zoltan2Interface());
-  zoltan22->SetFactory("A",AR22Fact);
-  zoltan22->SetFactory("number of partitions",RepHeuFact);
-  zoltan22->SetFactory("Coordinates",Coord22);
+  zoltan22->SetFactory("A", AR22Fact);
+  zoltan22->SetFactory("number of partitions", RepHeuFact);
+  zoltan22->SetFactory("Coordinates", Coord22);
 #endif
 
   RCP<RepartitionFactory> repart22 = rcp(new RepartitionFactory());
-  repart22->SetFactory("A",AR22Fact);
-  repart22->SetFactory("number of partitions",RepHeuFact);
+  repart22->SetFactory("A", AR22Fact);
+  repart22->SetFactory("number of partitions", RepHeuFact);
 #ifdef HAVE_MUELU_ZOLTAN2
   repart22->SetFactory("Partition", zoltan22);
 #endif
 
   RCP<FactoryManager> MReb11 = rcp(new FactoryManager());
   MReb11->SetFactory("A", AR11Fact);
-  MReb11->SetFactory("number of partitions",RepHeuFact);
+  MReb11->SetFactory("number of partitions", RepHeuFact);
   MReb11->SetFactory("Importer", repart11);
-  MReb11->SetFactory("Nullspace",Nullspace11);
+  MReb11->SetFactory("Nullspace", Nullspace11);
 
   RCP<FactoryManager> MReb22 = rcp(new FactoryManager());
   MReb22->SetFactory("A", AR22Fact);
-  MReb22->SetFactory("number of partitions",RepHeuFact);
+  MReb22->SetFactory("number of partitions", RepHeuFact);
   MReb22->SetFactory("Importer", repart22);
-  MReb22->SetFactory("Nullspace",Nullspace22);
+  MReb22->SetFactory("Nullspace", Nullspace22);
 
   RCP<RebalanceBlockInterpolationFactory> RebPFact = rcp(new RebalanceBlockInterpolationFactory());
   RebPFact->SetFactory("A", AcFact);
@@ -1250,43 +1277,39 @@ namespace MueLuTests {
 
   // main factory manager
   FactoryManager M;
-  M.SetFactory("A",            RebAcFact);
-  M.SetFactory("P",            RebPFact);
-  M.SetFactory("R",            RebRFact);
-  M.SetFactory("Coordinates",  CoordXfer);
-  M.SetFactory("Smoother",     smootherFact); // TODO fix me
+  M.SetFactory("A", RebAcFact);
+  M.SetFactory("P", RebPFact);
+  M.SetFactory("R", RebRFact);
+  M.SetFactory("Coordinates", CoordXfer);
+  M.SetFactory("Smoother", smootherFact);  // TODO fix me
   M.SetFactory("CoarseSolver", coarseSolverFact);
 
-
   // setup hierarchy
   H.SetVerbLevel(MueLu::Test);
-  H.EnableGraphDumping("dep_graph",0);
+  H.EnableGraphDumping("dep_graph", 0);
 
   H.Setup(M);
-  } //BlockedRAPFactoryCoords
-
+}  // BlockedRAPFactoryCoords
 
 #if defined(HAVE_MUELU_ZOLTAN2) && defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MUELU_ZOLTAN)
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithDiagonal, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryCoords, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithDiagonal, SC, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryCoords, SC, LO, GO, Node)
 
 #elif defined(HAVE_MUELU_ISORROPIA) && defined(HAVE_MUELU_ZOLTAN)
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithDiagonal, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithDiagonal, SC, LO, GO, Node)
 
 #else
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactory, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedRepartition, BlockedRAPFactoryWithRestriction, SC, LO, GO, Node)
 #endif
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ClassicalPFactory.cpp b/packages/muelu/test/unit_tests/ClassicalPFactory.cpp
index 20483cd14394..3cb20ca06456 100644
--- a/packages/muelu/test/unit_tests/ClassicalPFactory.cpp
+++ b/packages/muelu/test/unit_tests/ClassicalPFactory.cpp
@@ -62,276 +62,261 @@
 #include "MueLu_ClassicalPFactory.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory);
-    TEST_EQUALITY(PFact != Teuchos::null, true);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  } //Constructor
+  RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory);
+  TEST_EQUALITY(PFact != Teuchos::null, true);
 
+}  // Constructor
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_Direct, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_Direct, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC,LO,GO,NO>;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    out << "version: " << MueLu::Version() << std::endl;
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
 
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
+  GO nx         = 29;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  A->SetFixedBlockSize(1);
+  fineLevel.Set("A", A);
 
-    GO nx = 29;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    A->SetFixedBlockSize(1);
-    fineLevel.Set("A", A);
-   
-    // This test only works in parallel if we have Zoltan2 & Tpetra
+  // This test only works in parallel if we have Zoltan2 & Tpetra
 #ifndef HAVE_MUELU_ZOLTAN2
-    if(A->getRowMap()->getComm()->getSize() > 1)
-      return;
+  if (A->getRowMap()->getComm()->getSize() > 1)
+    return;
 #else
-    if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
-        (A->getRowMap()->getComm()->getSize() > 1))
-      return;
+  if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
+      (A->getRowMap()->getComm()->getSize() > 1))
+    return;
 #endif
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates
-      = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    LocalOrdinal NSdim = 2;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
-    cmFact->SetFactory("Graph", dropFact);
-    cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    Teuchos::ParameterList cp_params;
-    cp_params.set("aggregation: classical scheme","direct");
-    RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
-    PFact->SetParameterList(cp_params);
-    //    PFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    PFact->SetFactory("Graph", dropFact);
-    PFact->SetFactory("DofsPerNode", dropFact);
-    PFact->SetFactory("FC Splitting", cmFact);
-    PFact->SetFactory("CoarseMap", cmFact);
-
-    coarseLevel.Request("P",PFact.get());         // request Ptent
-    coarseLevel.Request(*PFact);
-    PFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> P;
-    coarseLevel.Get("P",P,PFact.get());
-
-    // Check that the matrix is not zero.
-    RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(),1);
-    x->putScalar(TST::one());
-    RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(),1);
-    P->apply(*x, *y,Teuchos::NO_TRANS);
-    const magnitude_type expected = TMT::one();
-    TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10*TMT::eps());
-  } //BuildP
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_ClassicalModified, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC,LO,GO,NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    GO nx = 29;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    A->SetFixedBlockSize(1);
-    fineLevel.Set("A", A);
-
-    // This test only works in parallel if we have Zoltan2 & Tpetra
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  LocalOrdinal NSdim         = 2;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
+  cmFact->SetFactory("Graph", dropFact);
+  cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  Teuchos::ParameterList cp_params;
+  cp_params.set("aggregation: classical scheme", "direct");
+  RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
+  PFact->SetParameterList(cp_params);
+  //    PFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  PFact->SetFactory("Graph", dropFact);
+  PFact->SetFactory("DofsPerNode", dropFact);
+  PFact->SetFactory("FC Splitting", cmFact);
+  PFact->SetFactory("CoarseMap", cmFact);
+
+  coarseLevel.Request("P", PFact.get());  // request Ptent
+  coarseLevel.Request(*PFact);
+  PFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, PFact.get());
+
+  // Check that the matrix is not zero.
+  RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  x->putScalar(TST::one());
+  RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  P->apply(*x, *y, Teuchos::NO_TRANS);
+  const magnitude_type expected = TMT::one();
+  TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10 * TMT::eps());
+}  // BuildP
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_ClassicalModified, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  GO nx         = 29;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  A->SetFixedBlockSize(1);
+  fineLevel.Set("A", A);
+
+  // This test only works in parallel if we have Zoltan2 & Tpetra
 #ifndef HAVE_MUELU_ZOLTAN2
-    if(A->getRowMap()->getComm()->getSize() > 1)
-      return;
+  if (A->getRowMap()->getComm()->getSize() > 1)
+    return;
 #else
-    if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
-        (A->getRowMap()->getComm()->getSize() > 1))
-      return;
+  if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
+      (A->getRowMap()->getComm()->getSize() > 1))
+    return;
 #endif
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates
-      = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    LocalOrdinal NSdim = 2;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-
-
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
-    cmFact->SetFactory("Graph", dropFact);
-    cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    Teuchos::ParameterList cp_params;
-    cp_params.set("aggregation: classical scheme","classical modified");
-    RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
-    PFact->SetParameterList(cp_params);
-    //PFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    PFact->SetFactory("Graph", dropFact);
-    PFact->SetFactory("DofsPerNode", dropFact);
-    PFact->SetFactory("FC Splitting", cmFact);
-    PFact->SetFactory("CoarseMap", cmFact);
-
-    coarseLevel.Request("P",PFact.get());         // request Ptent
-    coarseLevel.Request(*PFact);
-    PFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> P;
-    coarseLevel.Get("P",P,PFact.get());
-
-    // Check that the matrix is not zero.
-    RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(),1);
-    x->putScalar(TST::one());
-    RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(),1);
-    P->apply(*x, *y,Teuchos::NO_TRANS);
-    const magnitude_type expected = TMT::one();
-    TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10*TMT::eps());
-
-  } //BuildP
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_Ext, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC,LO,GO,NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    GO nx = 29;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    A->SetFixedBlockSize(1);
-    fineLevel.Set("A", A);
-
-    // This test only works in parallel if we have Zoltan2 & Tpetra
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  LocalOrdinal NSdim         = 2;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
+  cmFact->SetFactory("Graph", dropFact);
+  cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  Teuchos::ParameterList cp_params;
+  cp_params.set("aggregation: classical scheme", "classical modified");
+  RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
+  PFact->SetParameterList(cp_params);
+  // PFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  PFact->SetFactory("Graph", dropFact);
+  PFact->SetFactory("DofsPerNode", dropFact);
+  PFact->SetFactory("FC Splitting", cmFact);
+  PFact->SetFactory("CoarseMap", cmFact);
+
+  coarseLevel.Request("P", PFact.get());  // request Ptent
+  coarseLevel.Request(*PFact);
+  PFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, PFact.get());
+
+  // Check that the matrix is not zero.
+  RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  x->putScalar(TST::one());
+  RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  P->apply(*x, *y, Teuchos::NO_TRANS);
+  const magnitude_type expected = TMT::one();
+  TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10 * TMT::eps());
+
+}  // BuildP
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ClassicalPFactory, BuildP_Ext, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  GO nx         = 29;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  A->SetFixedBlockSize(1);
+  fineLevel.Set("A", A);
+
+  // This test only works in parallel if we have Zoltan2 & Tpetra
 #ifndef HAVE_MUELU_ZOLTAN2
-    if(A->getRowMap()->getComm()->getSize() > 1)
-      return;
+  if (A->getRowMap()->getComm()->getSize() > 1)
+    return;
 #else
-    if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
-        (A->getRowMap()->getComm()->getSize() > 1))
-      return;
+  if ((A->getRowMap()->lib() == Xpetra::UseEpetra) &&
+      (A->getRowMap()->getComm()->getSize() > 1))
+    return;
 #endif
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates
-      = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    LocalOrdinal NSdim = 2;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
-    cmFact->SetFactory("Graph", dropFact);
-    cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-
-    Teuchos::ParameterList cp_params;
-    cp_params.set("aggregation: classical scheme","ext+i");
-    RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
-    PFact->SetParameterList(cp_params);
-    //    PFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    PFact->SetFactory("Graph", dropFact);
-    PFact->SetFactory("DofsPerNode", dropFact);
-    PFact->SetFactory("FC Splitting", cmFact);
-    PFact->SetFactory("CoarseMap", cmFact);
-
-    coarseLevel.Request("P",PFact.get());         // request Ptent
-    coarseLevel.Request(*PFact);
-    PFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> P;
-    coarseLevel.Get("P",P,PFact.get());
-
-    // Check that the matrix is not zero.
-    RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(),1);
-    x->putScalar(TST::one());
-    RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(),1);
-    P->apply(*x, *y,Teuchos::NO_TRANS);
-    const magnitude_type expected = TMT::one();
-    TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10*TMT::eps());
-
-  } //BuildP
-
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory,Constructor,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory,BuildP_Direct,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory,BuildP_ClassicalModified,Scalar,LO,GO,Node)
-  // Disabled until we actually have code to run these
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  LocalOrdinal NSdim         = 2;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<ClassicalMapFactory> cmFact = rcp(new ClassicalMapFactory());
+  cmFact->SetFactory("Graph", dropFact);
+  cmFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  Teuchos::ParameterList cp_params;
+  cp_params.set("aggregation: classical scheme", "ext+i");
+  RCP<ClassicalPFactory> PFact = rcp(new ClassicalPFactory());
+  PFact->SetParameterList(cp_params);
+  //    PFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  PFact->SetFactory("Graph", dropFact);
+  PFact->SetFactory("DofsPerNode", dropFact);
+  PFact->SetFactory("FC Splitting", cmFact);
+  PFact->SetFactory("CoarseMap", cmFact);
+
+  coarseLevel.Request("P", PFact.get());  // request Ptent
+  coarseLevel.Request(*PFact);
+  PFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, PFact.get());
+
+  // Check that the matrix is not zero.
+  RCP<MultiVector> x = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  x->putScalar(TST::one());
+  RCP<MultiVector> y = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  P->apply(*x, *y, Teuchos::NO_TRANS);
+  const magnitude_type expected = TMT::one();
+  TEST_FLOATING_EQUALITY(y->getVector(0)->normInf(), expected, 10 * TMT::eps());
+
+}  // BuildP
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory, Constructor, Scalar, LO, GO, Node)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory, BuildP_Direct, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory, BuildP_ClassicalModified, Scalar, LO, GO, Node)
+// Disabled until we actually have code to run these
 #if 0
 
       TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ClassicalPFactory,BuildP_Ext,Scalar,LO,GO,Node)
@@ -339,5 +324,4 @@ namespace MueLuTests {
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/CoalesceDropFactory.cpp b/packages/muelu/test/unit_tests/CoalesceDropFactory.cpp
index f92f7a78f7c2..b3b5c8476c61 100644
--- a/packages/muelu/test/unit_tests/CoalesceDropFactory.cpp
+++ b/packages/muelu/test/unit_tests/CoalesceDropFactory.cpp
@@ -60,2331 +60,2268 @@
 
 namespace MueLuTests {
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<CoalesceDropFactory> coalesceDropFact = rcp(new CoalesceDropFactory());
-    coalesceDropFact->SetDefaultVerbLevel(MueLu::Extreme);
-    TEST_EQUALITY(coalesceDropFact != Teuchos::null, true);
-
-  } //Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-  } //Build
-
-  // TODO remove this
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, PreDrop, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(3);
-    fineLevel.Set("A", A);
-    A->describe(out,Teuchos::VERB_EXTREME);
-
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetVerbLevel(MueLu::Extreme);
-    dropFact.SetPreDropFunction(rcp(new PreDropFunctionConstVal(0.00001)));
-
-    fineLevel.Request("Graph", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    std::cout << graph->GetDomainMap()->getGlobalNumElements() << std::endl;
-    graph->print(out, MueLu::Debug);
-
-    //    TEST_EQUALITY(1 == 0, true);
-
-  } //PreDrop
+  RCP<CoalesceDropFactory> coalesceDropFact = rcp(new CoalesceDropFactory());
+  coalesceDropFact->SetDefaultVerbLevel(MueLu::Extreme);
+  TEST_EQUALITY(coalesceDropFact != Teuchos::null, true);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationBasic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3.
-    // lightweight wrap = false
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+}  // Constructor
 
-    int blockSize=3;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    int nx = blockSize*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(nx);
-    A->SetFixedBlockSize(blockSize, 0);
-    fineLevel.Set("A", A);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(false));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+}  // Build
+
+// TODO remove this
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, PreDrop, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(3);
+  fineLevel.Set("A", A);
+  A->describe(out, Teuchos::VERB_EXTREME);
+
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetVerbLevel(MueLu::Extreme);
+  dropFact.SetPreDropFunction(rcp(new PreDropFunctionConstVal(0.00001)));
+
+  fineLevel.Request("Graph", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  std::cout << graph->GetDomainMap()->getGlobalNumElements() << std::endl;
+  graph->print(out, MueLu::Debug);
+
+  //    TEST_EQUALITY(1 == 0, true);
+
+}  // PreDrop
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationBasic, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3.
+  // lightweight wrap = false
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  int blockSize = 3;
+
+  int nx        = blockSize * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+  A->SetFixedBlockSize(blockSize, 0);
+  fineLevel.Set("A", A);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(false));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
 
-    dropFact.Build(fineLevel);
+  dropFact.Build(fineLevel);
 
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1))); //u,v and p
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));  // u,v and p
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-1), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationBasic
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3 using a strided map
-    // lightweight wrap = false
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    int blockSize = 3;
-    int nx = blockSize*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(nx);
-
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(blockSize));
-    LocalOrdinal stridedBlockId = -1;
-
-    RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  A->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  0 /*offset*/
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            A->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            0 /*offset*/
-                                            );
-
-    if(A->IsView("stridedMaps") == true) A->RemoveView("stridedMaps");
-    A->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", A);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(false));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationBasic
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3 using a strided map
+  // lightweight wrap = false
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  int blockSize = 3;
+  int nx        = blockSize * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(blockSize));
+  LocalOrdinal stridedBlockId = -1;
+
+  RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      A->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      A->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+
+  if (A->IsView("stridedMaps") == true) A->RemoveView("stridedMaps");
+  A->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", A);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(false));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-1), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStrided
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3 = (2,1). wrap block 0
-    // lightweight wrap = false
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // create strided map information
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(2));
-    stridingInfo.push_back(Teuchos::as<size_t>(1));
-    LocalOrdinal stridedBlockId = 0;
-
-    RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3*comm->getSize(), 0,
-                                  stridingInfo, comm,
-                                  stridedBlockId /*blockId*/, 0 /*offset*/);
-
-    /////////////////////////////////////////////////////
-
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  mtx->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  0 /*offset*/
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            mtx->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            0 /*offset*/
-                                            );
-
-    if(mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
-    mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);     // should have holes in these maps
-
-    fineLevel.Set("A", mtx);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(false));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 3, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStrided
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3 = (2,1). wrap block 0
+  // lightweight wrap = false
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  // create strided map information
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(2));
+  stridingInfo.push_back(Teuchos::as<size_t>(1));
+  LocalOrdinal stridedBlockId = 0;
+
+  RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3 * comm->getSize(), 0,
+                                                                                                     stridingInfo, comm,
+                                                                                                     stridedBlockId /*blockId*/, 0 /*offset*/);
+
+  /////////////////////////////////////////////////////
+
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+
+  if (mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
+  mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);  // should have holes in these maps
+
+  fineLevel.Set("A", mtx);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(false));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 3, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*3-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*3-1), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * 3 - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * 3 - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStrided2
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffset, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 9 = (2,3,4). wrap block 1.
-    // lightweight wrap = false
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // create strided map information
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(2));
-    stridingInfo.push_back(Teuchos::as<size_t>(3));
-    stridingInfo.push_back(Teuchos::as<size_t>(4));
-    LocalOrdinal stridedBlockId = 1;
-    GlobalOrdinal offset = 19;
-
-    RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9*comm->getSize(), 0,
-                                  stridingInfo, comm,
-                                  stridedBlockId, offset);
-
-    /////////////////////////////////////////////////////
-
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -3.0);
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  mtx->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  offset
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            mtx->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            offset
-                                            );
-
-    if(mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
-    mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", mtx);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(false));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 9, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStrided2
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffset, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 9 = (2,3,4). wrap block 1.
+  // lightweight wrap = false
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  // create strided map information
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(2));
+  stridingInfo.push_back(Teuchos::as<size_t>(3));
+  stridingInfo.push_back(Teuchos::as<size_t>(4));
+  LocalOrdinal stridedBlockId = 1;
+  GlobalOrdinal offset        = 19;
+
+  RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9 * comm->getSize(), 0,
+                                                                                                     stridingInfo, comm,
+                                                                                                     stridedBlockId, offset);
+
+  /////////////////////////////////////////////////////
+
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -3.0);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+
+  if (mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
+  mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", mtx);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(false));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 9, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*3-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*3-1), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * 3 - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * 3 - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStridedOffset
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationLightweight, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    int blockSize=3;
-
-    RCP<const Map> dofMap = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize*comm->getSize(), 0, comm);
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-    mtx->SetFixedBlockSize(blockSize, 0);
-    fineLevel.Set("A", mtx);
-
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStridedOffset
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationLightweight, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  int blockSize = 3;
+
+  RCP<const Map> dofMap    = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize * comm->getSize(), 0, comm);
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
+  mtx->SetFixedBlockSize(blockSize, 0);
+  fineLevel.Set("A", mtx);
+
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
-    }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-1), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationLightweight
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationLightweightDrop, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 1
-    // lightweight wrap = true
-    // drop small values
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Map> dofMap = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3*comm->getSize(), 0, comm);
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 1.0, -1.0, -0.0001);
-    mtx->SetFixedBlockSize(1, 0);
-    fineLevel.Set("A", mtx);
-
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.5));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == 3*comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 2), true);
     } else {
-      if (comm->getRank() == 0 ) {
-        if (graph->getNeighborVertices(0).size() == 1) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 1), true);
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 3*comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(3*comm->getSize()+(comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 3*comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(3*comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==3), true);
-  } // AmalgamationLightweightDrop
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedLW, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3 using a strided map
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    int blockSize=3;
-
-    int nx = blockSize*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(nx);
-
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(blockSize));
-    LocalOrdinal stridedBlockId = -1;
-
-    RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  A->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  0 /*offset*/
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            A->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            0 /*offset*/
-                                            );
-
-    if(A->IsView("stridedMaps") == true) A->RemoveView("stridedMaps");
-    A->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", A);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
+  }
 
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationLightweight
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationLightweightDrop, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 1
+  // lightweight wrap = true
+  // drop small values
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Map> dofMap    = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3 * comm->getSize(), 0, comm);
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 1.0, -1.0, -0.0001);
+  mtx->SetFixedBlockSize(1, 0);
+  fineLevel.Set("A", mtx);
+
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.5));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == 3 * comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0) {
+      if (graph->getNeighborVertices(0).size() == 1) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
-    }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-1), true);
-      }
-    }
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStridedLW
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided2LW, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3 = (2,1). wrap block 0
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // create strided map information
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(2));
-    stridingInfo.push_back(Teuchos::as<size_t>(1));
-    LocalOrdinal stridedBlockId = 0;
-
-    int blockSize=3;
-
-    RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize*comm->getSize(), 0,
-                                  stridingInfo, comm,
-                                  stridedBlockId /*blockId*/, 0 /*offset*/);
-
-    /////////////////////////////////////////////////////
-
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  mtx->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  0 /*offset*/
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            mtx->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            0 /*offset*/
-                                            );
-    if(mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
-    mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", mtx);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 3 * comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(3 * comm->getSize() + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 3 * comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(3 * comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 3), true);
+}  // AmalgamationLightweightDrop
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedLW, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3 using a strided map
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  int blockSize = 3;
+
+  int nx        = blockSize * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(blockSize));
+  LocalOrdinal stridedBlockId = -1;
+
+  RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      A->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      A->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+
+  if (A->IsView("stridedMaps") == true) A->RemoveView("stridedMaps");
+  A->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", A);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-2), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex==numLocalRowMapElts*blockSize-1), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStrided2LW
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffsetLW, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 9 = (2,3,4). wrap block 1.
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // create strided map information
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(2));
-    stridingInfo.push_back(Teuchos::as<size_t>(3));
-    stridingInfo.push_back(Teuchos::as<size_t>(4));
-    LocalOrdinal stridedBlockId = 1;
-    GlobalOrdinal offset = 19;
-
-    int blockSize=9;
-
-    RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize*comm->getSize(), 0,
-                                  stridingInfo, comm,
-                                  stridedBlockId, offset);
-
-    /////////////////////////////////////////////////////
-
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -3.0);
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  mtx->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  offset
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            mtx->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            offset
-                                            );
-
-    if(mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
-    mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", mtx);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStridedLW
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStrided2LW, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3 = (2,1). wrap block 0
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  // create strided map information
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(2));
+  stridingInfo.push_back(Teuchos::as<size_t>(1));
+  LocalOrdinal stridedBlockId = 0;
+
+  int blockSize = 3;
+
+  RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize * comm->getSize(), 0,
+                                                                                                     stridingInfo, comm,
+                                                                                                     stridedBlockId /*blockId*/, 0 /*offset*/);
+
+  /////////////////////////////////////////////////////
+
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Xpetra::StridedMap<LocalOrdinal, GlobalOrdinal, Node> > stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      0 /*offset*/
+  );
+  if (mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
+  mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", mtx);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == blockSize) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStridedOffsetLW
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationDroppingLW, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 3 = (3)
-    // drop small entries
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Map> dofMap = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3*comm->getSize(), 0, comm);
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, 0.00001);
-    mtx->SetFixedBlockSize(3, 0);
-    fineLevel.Set("A", mtx);
-
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(1.0));
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 3, true);
-    TEST_EQUALITY(graph->getNeighborVertices(0).size(), 1);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 2), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(maxLocalIndex == numLocalRowMapElts * blockSize - 1), true);
     }
+  }
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-
-  } // AmalgamationDroppingLW
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffsetDropping2LW, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // unit test for block size 9 = (2,3,4). wrap block 1.
-    // drop small entries
-    // lightweight wrap = true
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // create strided map information
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(Teuchos::as<size_t>(2));
-    stridingInfo.push_back(Teuchos::as<size_t>(3));
-    stridingInfo.push_back(Teuchos::as<size_t>(4));
-    LocalOrdinal stridedBlockId = 1;
-    GlobalOrdinal offset = 19;
-
-    RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9*comm->getSize(), 0,
-                                  stridingInfo, comm,
-                                  stridedBlockId, offset);
-
-    /////////////////////////////////////////////////////
-
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, 1.0, 0.0001);
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                                  mtx->getRangeMap(),
-                                                  stridingInfo,
-                                                  stridedBlockId,
-                                                  offset
-                                                  );
-    RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
-                                            mtx->getDomainMap(),
-                                            stridingInfo,
-                                            stridedBlockId,
-                                            offset
-                                            );
-
-    if(mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
-    mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
-
-    fineLevel.Set("A", mtx);
-    CoalesceDropFactory dropFact = CoalesceDropFactory();
-    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-    dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.3));
-
-    fineLevel.Request("Graph", &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    fineLevel.print(out);
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
-    TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 9, true);
-    bool bCorrectGraph = false;
-    if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
-      bCorrectGraph = true;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStrided2LW
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffsetLW, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 9 = (2,3,4). wrap block 1.
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  // create strided map information
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(2));
+  stridingInfo.push_back(Teuchos::as<size_t>(3));
+  stridingInfo.push_back(Teuchos::as<size_t>(4));
+  LocalOrdinal stridedBlockId = 1;
+  GlobalOrdinal offset        = 19;
+
+  int blockSize = 9;
+
+  RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, blockSize * comm->getSize(), 0,
+                                                                                                     stridingInfo, comm,
+                                                                                                     stridedBlockId, offset);
+
+  /////////////////////////////////////////////////////
+
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -3.0);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+
+  if (mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
+  mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", mtx);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == blockSize, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     } else {
-      if (comm->getRank() == 0) {
-        if (graph->getNeighborVertices(0).size() == 1) bCorrectGraph = true;
-      }
-      else {
-        if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
-      }
+      if (graph->getNeighborVertices(0).size() == 3) bCorrectGraph = true;
     }
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize()>1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+1), true);
-      } else {
-        TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts==numLocalRowMapElts+2), true);
-      }
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),Teuchos::as<size_t>(comm->getSize()));
-    TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements()==1), true);
-  } // AmalgamationStridedOffsetDropping2LW
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),40);
-
-  } // DistanceLaplacian
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  }
 
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
-    if(!comm->getRank()) {
-      auto vals = coordinates->getDataNonConst(0);
-      vals[0] = vals[0] - 2000*36;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStridedOffsetLW
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationDroppingLW, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 3 = (3)
+  // drop small entries
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Map> dofMap    = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 3 * comm->getSize(), 0, comm);
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, 0.00001);
+  mtx->SetFixedBlockSize(3, 0);
+  fineLevel.Set("A", mtx);
+
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(1.0));
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 3, true);
+  TEST_EQUALITY(graph->getNeighborVertices(0).size(), 1);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
+  }
 
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(8.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: distance laplacian algo",Teuchos::ParameterEntry(std::string("scaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),105);
-
-  } // DistanceLaplacianCut
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianUnscaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
-    if(!comm->getRank()) {
-      auto vals = coordinates->getDataNonConst(0);
-      vals[0] = vals[0] - 2000*36;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+
+}  // AmalgamationDroppingLW
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AmalgamationStridedOffsetDropping2LW, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // unit test for block size 9 = (2,3,4). wrap block 1.
+  // drop small entries
+  // lightweight wrap = true
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  // create strided map information
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(Teuchos::as<size_t>(2));
+  stridingInfo.push_back(Teuchos::as<size_t>(3));
+  stridingInfo.push_back(Teuchos::as<size_t>(4));
+  LocalOrdinal stridedBlockId = 1;
+  GlobalOrdinal offset        = 19;
+
+  RCP<const StridedMap> dofMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9 * comm->getSize(), 0,
+                                                                                                     stridingInfo, comm,
+                                                                                                     stridedBlockId, offset);
+
+  /////////////////////////////////////////////////////
+
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, 1.0, 0.0001);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<const Map> stridedRangeMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getRangeMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+  RCP<const Map> stridedDomainMap = Xpetra::StridedMapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(
+      mtx->getDomainMap(),
+      stridingInfo,
+      stridedBlockId,
+      offset);
+
+  if (mtx->IsView("stridedMaps") == true) mtx->RemoveView("stridedMaps");
+  mtx->CreateView("stridedMaps", stridedRangeMap, stridedDomainMap);
+
+  fineLevel.Set("A", mtx);
+  CoalesceDropFactory dropFact = CoalesceDropFactory();
+  dropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.3));
+
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  fineLevel.print(out);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
+
+  LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+  TEST_EQUALITY(Teuchos::as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 9, true);
+  bool bCorrectGraph = false;
+  if (comm->getSize() == 1 && graph->getNeighborVertices(0).size() == 1) {
+    bCorrectGraph = true;
+  } else {
+    if (comm->getRank() == 0) {
+      if (graph->getNeighborVertices(0).size() == 1) bCorrectGraph = true;
+    } else {
+      if (graph->getNeighborVertices(0).size() == 2) bCorrectGraph = true;
     }
-
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(8.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: distance laplacian algo",Teuchos::ParameterEntry(std::string("unscaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),105);
-
-  } // DistanceLaplacianUnscaleCut
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianCutSym, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
-    if(!comm->getRank()) {
-      auto vals = coordinates->getDataNonConst(0);
-      vals[0] = vals[0] - 2000*36;
+  }
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 1), true);
+    } else {
+      TEST_EQUALITY(Teuchos::as<bool>(numLocalImportElts == numLocalRowMapElts + 2), true);
     }
+  }
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), Teuchos::as<size_t>(comm->getSize()));
+  TEST_EQUALITY(Teuchos::as<bool>(myDomainMap->getLocalNumElements() == 1), true);
+}  // AmalgamationStridedOffsetDropping2LW
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 40);
+
+}  // DistanceLaplacian
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
+  if (!comm->getRank()) {
+    auto vals = coordinates->getDataNonConst(0);
+    vals[0]   = vals[0] - 2000 * 36;
+  }
 
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(8.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: distance laplacian algo",Teuchos::ParameterEntry(std::string("scaled cut symmetric")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),106);
-
-  } // DistanceLaplacianCutScaled
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, ClassicalScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("classical")));
-    coalesceDropFact.SetParameter("aggregation: classical algo",Teuchos::ParameterEntry(std::string("scaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),72);
-
-  } // SignaledClassical
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, ClassicalUnScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("classical")));
-    coalesceDropFact.SetParameter("aggregation: classical algo",Teuchos::ParameterEntry(std::string("unscaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),72);
-
-  } // SignaledClassical
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("signed classical")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
-
-  } // SignaledClassical
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledScaledCutClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("signed classical")));
-    coalesceDropFact.SetParameter("aggregation: classical algo",Teuchos::ParameterEntry(std::string("scaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    TEST_THROW(coalesceDropFact.Build(fineLevel), MueLu::Exceptions::RuntimeError);
+  fineLevel.Set("Coordinates", coordinates);
 
-//    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-//    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-//    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(8.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: distance laplacian algo", Teuchos::ParameterEntry(std::string("scaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-//    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-//    const RCP<const Map> myDomainMap = graph->GetDomainMap();
+  coalesceDropFact.Build(fineLevel);
 
-//    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-//    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-//    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-//    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
 
-//    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-//    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-//    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-//    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
 
-//    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
 
-  } // SignaledScaledCutClassical
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledUnscaledCutClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 105);
 
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+}  // DistanceLaplacianCut
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianUnscaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("signed classical")));
-    coalesceDropFact.SetParameter("aggregation: classical algo",Teuchos::ParameterEntry(std::string("unscaled cut")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
 
-    TEST_THROW(coalesceDropFact.Build(fineLevel), MueLu::Exceptions::RuntimeError);
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
 
-//    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-//    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-//    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+  // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
+  if (!comm->getRank()) {
+    auto vals = coordinates->getDataNonConst(0);
+    vals[0]   = vals[0] - 2000 * 36;
+  }
 
-//    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-//    const RCP<const Map> myDomainMap = graph->GetDomainMap();
+  fineLevel.Set("Coordinates", coordinates);
 
-//    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-//    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-//    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-//    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(8.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: distance laplacian algo", Teuchos::ParameterEntry(std::string("unscaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-//    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-//    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-//    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-//    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
+  coalesceDropFact.Build(fineLevel);
 
-//    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
 
-  } // SignaledUnScaledCutClassical
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalColoredSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
 
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 105);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
+}  // DistanceLaplacianUnscaleCut
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianCutSym, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal colored signed classical")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    coalesceDropFact.Build(fineLevel);
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
 
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
 
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
+  // Now we doctor the coordinates so that the off-diagonal pair row 0 will want to keep (0,1) and row 1 will want to drop (1,0)
+  if (!comm->getRank()) {
+    auto vals = coordinates->getDataNonConst(0);
+    vals[0]   = vals[0] - 2000 * 36;
+  }
 
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(8.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: distance laplacian algo", Teuchos::ParameterEntry(std::string("scaled cut symmetric")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-  } // BlockDiagonalColoredSignedClassical
+  coalesceDropFact.Build(fineLevel);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalNoColoredSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    // this test is only compatible with rank higher than 1
-    if(comm->getSize() == 1) {
-        return;
-    }
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
 
-    // Default is Laplace1D with nx = 8748.
-    // It's a nice size for 1D and perfect aggregation. (6561 = 3^8)
-    //Nice size for 1D and perfect aggregation on small numbers of processors. (8748 = 4*3^7)
-    Teuchos::CommandLineProcessor clp(false);
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-    Xpetra::Parameters xpetraParameters(clp);
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 106);
+
+}  // DistanceLaplacianCutScaled
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, ClassicalScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("classical")));
+  coalesceDropFact.SetParameter("aggregation: classical algo", Teuchos::ParameterEntry(std::string("scaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-    RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 72);
+
+}  // SignaledClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, ClassicalUnScaledCut, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
 
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-//    getCrsGraph()->getImporter()
-    RCP<const Import> importer = ImportFactory::Build(A->getRowMap(), map);
-    fineLevel.Set("Importer", importer);
-    auto importerTest = A->getCrsGraph()->getImporter(); // NULL
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal colored signed classical")));
-    coalesceDropFact.SetParameter("aggregation: coloring: localize color graph", Teuchos::ParameterEntry(false));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    // Need an importer
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
-
-  } // BlockDiagonalNoColoredSignedClassical
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("classical")));
+  coalesceDropFact.SetParameter("aggregation: classical algo", Teuchos::ParameterEntry(std::string("unscaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 72);
+
+}  // SignaledClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("signed classical")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
 
-    Level fineLevel;
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    // We're dropping all the interior off-diagonal entries.
-    // dx = 1/36
-    // L_ij = -36
-    // L_ii = 72
-    // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.51));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal signed classical")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-
-    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    const RCP<const Map> myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
-
-    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
-
-  } // BlockDiagonalSignedClassical
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList,lib);
-    A->SetFixedBlockSize(1);// So we can block diagonalize
-    Level fineLevel;
-    fineLevel.Set("A", A);
-   
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+  coalesceDropFact.Build(fineLevel);
 
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 36);
+
+}  // SignaledClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledScaledCutClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
 
-    RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
-    Teuchos::ParameterList ibList;
-    ibList.set("aggregation: block diagonal: interleaved blocksize",3);
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetFactory("BlockNumber",ibFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(8.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal")));
-    coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize",Teuchos::ParameterEntry(3));
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("signed classical")));
+  coalesceDropFact.SetParameter("aggregation: classical algo", Teuchos::ParameterEntry(std::string("scaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  TEST_THROW(coalesceDropFact.Build(fineLevel), MueLu::Exceptions::RuntimeError);
+
+  //    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  //    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  //    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  //    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  //    const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  //    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  //    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  //    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
+  //    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
+
+  //    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  //    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  //    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
+  //    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
+
+  //    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
+
+}  // SignaledScaledCutClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignaledUnscaledCutClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("signed classical")));
+  coalesceDropFact.SetParameter("aggregation: classical algo", Teuchos::ParameterEntry(std::string("unscaled cut")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  TEST_THROW(coalesceDropFact.Build(fineLevel), MueLu::Exceptions::RuntimeError);
+
+  //    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  //    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  //    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  //    const RCP<const Map> myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  //    const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  //    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  //    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  //    TEST_EQUALITY(myImportMap->getMinLocalIndex(),0);
+  //    TEST_EQUALITY(myImportMap->getGlobalNumElements(),Teuchos::as<size_t>(36 + (comm->getSize()-1)*2));
+
+  //    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  //    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  //    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),0);
+  //    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),36);
+
+  //    TEST_EQUALITY(graph->GetGlobalNumEdges(),36);
+
+}  // SignaledUnScaledCutClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalColoredSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal colored signed classical")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 36);
+
+}  // BlockDiagonalColoredSignedClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalNoColoredSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  // this test is only compatible with rank higher than 1
+  if (comm->getSize() == 1) {
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  // Default is Laplace1D with nx = 8748.
+  // It's a nice size for 1D and perfect aggregation. (6561 = 3^8)
+  // Nice size for 1D and perfect aggregation on small numbers of processors. (8748 = 4*3^7)
+  Teuchos::CommandLineProcessor clp(false);
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);
+
+  RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  //    getCrsGraph()->getImporter()
+  RCP<const Import> importer = ImportFactory::Build(A->getRowMap(), map);
+  fineLevel.Set("Importer", importer);
+  auto importerTest = A->getCrsGraph()->getImporter();  // NULL
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal colored signed classical")));
+  coalesceDropFact.SetParameter("aggregation: coloring: localize color graph", Teuchos::ParameterEntry(false));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  // Need an importer
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 36);
+
+}  // BlockDiagonalNoColoredSignedClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalSignedClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", Teuchos::as<GlobalOrdinal>(36));
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  // We're dropping all the interior off-diagonal entries.
+  // dx = 1/36
+  // L_ij = -36
+  // L_ii = 72
+  // criterion for dropping is |L_ij|^2 <= tol^2 * |L_ii*L_jj|
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.51));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal signed classical")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+
+  const RCP<const Map> myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  const RCP<const Map> myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), Teuchos::as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+
+  TEST_EQUALITY(graph->GetGlobalNumEdges(), 36);
+
+}  // BlockDiagonalSignedClassical
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList, lib);
+  A->SetFixedBlockSize(1);  // So we can block diagonalize
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
+  Teuchos::ParameterList ibList;
+  ibList.set("aggregation: block diagonal: interleaved blocksize", 3);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetFactory("BlockNumber", ibFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(8.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal")));
+  coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize", Teuchos::ParameterEntry(3));
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalClassical, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList, lib);
+  A->SetFixedBlockSize(1);  // So we can block diagonalize
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
+  Teuchos::ParameterList ibList;
+  ibList.set("aggregation: block diagonal: interleaved blocksize", 3);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetFactory("BlockNumber", ibFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(8.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal classical")));
+  coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize", Teuchos::ParameterEntry(3));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList, lib);
+  A->SetFixedBlockSize(1);  // So we can block diagonalize
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
+  Teuchos::ParameterList ibList;
+  ibList.set("aggregation: block diagonal: interleaved blocksize", 3);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetFactory("BlockNumber", ibFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.025));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize", Teuchos::ParameterEntry(3));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceDifferentCoordinatesLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList, lib);
+  A->SetFixedBlockSize(1);  // So we can block diagonalize
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  GO bnx = 15 * comm->getSize();
+  Teuchos::ParameterList bMatrixList;
+  matrixList.set("bnx", bnx);
+  RCP<Matrix> B = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(bMatrixList, lib);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", B->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
+  Teuchos::ParameterList ibList;
+  ibList.set("aggregation: block diagonal: interleaved blocksize", 3);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetFactory("BlockNumber", ibFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.025));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize", Teuchos::ParameterEntry(3));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceLaplacianWeighted, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", (GO)10);
+  matrixList.set("nz", (GO)10);
+  matrixList.set("matrixType", "Laplace3D");
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList, lib);
+  A->SetFixedBlockSize(1);  // So we can block diagonalize
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", A->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
+  Teuchos::ParameterList ibList;
+  ibList.set("aggregation: block diagonal: interleaved blocksize", 3);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetFactory("BlockNumber", ibFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.025));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
+  coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize", Teuchos::ParameterEntry(3));
+  std::vector<double> weights_v{100.0, 1.0, 1.0, 1.0, 100, 1.0, 1.0, 1.0, 100.0};
+  Teuchos::Array<double> weights(weights_v);
+  coalesceDropFact.SetParameter("aggregation: distance laplacian directional weights", Teuchos::ParameterEntry(weights));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianWeighted, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", (GO)10);
+  matrixList.set("nz", (GO)10);
+  matrixList.set("matrixType", "Laplace3D");
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixList, lib);
+
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", A->getRowMap(), matrixList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.025));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("distance laplacian")));
+  std::vector<double> weights_v{100.0, 1.0, 1.0};
+  Teuchos::Array<double> weights(weights_v);
+  coalesceDropFact.SetParameter("aggregation: distance laplacian directional weights", Teuchos::ParameterEntry(weights));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AggresiveDroppingIsMarkedAsBoundary, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // Test that when everything but the diagonal is dropped, the node is marked as boundary
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  RCP<const Map> dofMap    = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 12 * comm->getSize(), 0, comm);
+  Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList,lib);
-    A->SetFixedBlockSize(1);// So we can block diagonalize
+  {
     Level fineLevel;
-    fineLevel.Set("A", A);
-
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
+    mtx->SetFixedBlockSize(1);
+    fineLevel.Set("A", mtx);
 
-    RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
-    Teuchos::ParameterList ibList;
-    ibList.set("aggregation: block diagonal: interleaved blocksize",3);
+    CoalesceDropFactory dropFact = CoalesceDropFactory();
+    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
     RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetFactory("BlockNumber",ibFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(8.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal classical")));
-    coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize",Teuchos::ParameterEntry(3));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-  }
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+    dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(4.1));
+    fineLevel.Request("Graph", &dropFact);
+    fineLevel.Request("DofsPerNode", &dropFact);
 
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList,lib);
-    A->SetFixedBlockSize(1);// So we can block diagonalize
-    Level fineLevel;
-    fineLevel.Set("A", A);
-   
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+    dropFact.Build(fineLevel);
 
+    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
 
-    RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
-    Teuchos::ParameterList ibList;
-    ibList.set("aggregation: block diagonal: interleaved blocksize",3);
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetFactory("BlockNumber",ibFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.025));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize",Teuchos::ParameterEntry(3));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+    auto boundaryNodes         = graph->GetBoundaryNodeMap();
+    bool allNodesAreOnBoundary = true;
+    for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
+      allNodesAreOnBoundary &= boundaryNodes[i];
+    TEST_EQUALITY(allNodesAreOnBoundary, true);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceDifferentCoordinatesLaplacian, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList,lib);
-    A->SetFixedBlockSize(1);// So we can block diagonalize
+  {
     Level fineLevel;
-    fineLevel.Set("A", A);
-
-    GO bnx = 15*comm->getSize();
-    Teuchos::ParameterList bMatrixList;
-    matrixList.set("bnx",bnx);
-    RCP<Matrix> B =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(bMatrixList,lib);
-
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", B->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
+    mtx->SetFixedBlockSize(2);
+    fineLevel.Set("A", mtx);
 
-    RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
-    Teuchos::ParameterList ibList;
-    ibList.set("aggregation: block diagonal: interleaved blocksize",3);
+    CoalesceDropFactory dropFact = CoalesceDropFactory();
+    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
     RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetFactory("BlockNumber",ibFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.025));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize",Teuchos::ParameterEntry(3));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-  }
+    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+    dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(4.1));
+    fineLevel.Request("Graph", &dropFact);
+    fineLevel.Request("DofsPerNode", &dropFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, BlockDiagonalDistanceLaplacianWeighted, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    matrixList.set("ny",(GO)10);
-    matrixList.set("nz",(GO)10);
-    matrixList.set("matrixType","Laplace3D");
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixList,lib);
-    A->SetFixedBlockSize(1);// So we can block diagonalize
-    Level fineLevel;
-    fineLevel.Set("A", A);
-   
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("3D", A->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+    dropFact.Build(fineLevel);
 
+    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
 
-    RCP<InitialBlockNumberFactory> ibFact = rcp(new InitialBlockNumberFactory());
-    Teuchos::ParameterList ibList;
-    ibList.set("aggregation: block diagonal: interleaved blocksize",3);
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetFactory("BlockNumber",ibFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.025));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("block diagonal distance laplacian")));
-    coalesceDropFact.SetParameter("aggregation: block diagonal: interleaved blocksize",Teuchos::ParameterEntry(3));
-    std::vector<double> weights_v{100.0, 1.0, 1.0,   1.0, 100, 1.0,   1.0, 1.0, 100.0};
-    Teuchos::Array<double> weights(weights_v);
-    coalesceDropFact.SetParameter("aggregation: distance laplacian directional weights",Teuchos::ParameterEntry(weights));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+    auto boundaryNodes         = graph->GetBoundaryNodeMap();
+    bool allNodesAreOnBoundary = true;
+    for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
+      allNodesAreOnBoundary &= boundaryNodes[i];
+    TEST_EQUALITY(allNodesAreOnBoundary, true);
   }
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, DistanceLaplacianWeighted, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    matrixList.set("ny",(GO)10);
-    matrixList.set("nz",(GO)10);
-    matrixList.set("matrixType","Laplace3D");
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixList,lib);
-
+  {
     Level fineLevel;
-    fineLevel.Set("A", A);
-   
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("3D", A->getRowMap(), matrixList);
-    fineLevel.Set("Coordinates", coordinates);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
+    mtx->SetFixedBlockSize(3);
+    fineLevel.Set("A", mtx);
 
+    CoalesceDropFactory dropFact = CoalesceDropFactory();
+    dropFact.SetDefaultVerbLevel(MueLu::Extreme);
     RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetDefaultVerbLevel(MueLu::Extreme);
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.025));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("distance laplacian")));
-    std::vector<double> weights_v{100.0, 1.0, 1.0};
-    Teuchos::Array<double> weights(weights_v);
-    coalesceDropFact.SetParameter("aggregation: distance laplacian directional weights",Teuchos::ParameterEntry(weights));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
-  }
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, AggresiveDroppingIsMarkedAsBoundary, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // Test that when everything but the diagonal is dropped, the node is marked as boundary
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    RCP<const Map> dofMap = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 12*comm->getSize(), 0, comm);
-    Teuchos::RCP<Matrix> mtx = TestHelpers::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-
-    {
-      Level fineLevel;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-      mtx->SetFixedBlockSize(1);
-      fineLevel.Set("A", mtx);
-
-      CoalesceDropFactory dropFact = CoalesceDropFactory();
-      dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-      dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-      dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-      fineLevel.Request("Graph", &dropFact);
-      fineLevel.Request("DofsPerNode", &dropFact);
-
-      dropFact.Build(fineLevel);
-
-      RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-      auto boundaryNodes = graph->GetBoundaryNodeMap();
-      bool allNodesAreOnBoundary = true;
-      for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
-        allNodesAreOnBoundary &= boundaryNodes[i];
-      TEST_EQUALITY(allNodesAreOnBoundary, true);
-    }
-
-    {
-      Level fineLevel;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-      mtx->SetFixedBlockSize(2);
-      fineLevel.Set("A", mtx);
-
-      CoalesceDropFactory dropFact = CoalesceDropFactory();
-      dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-      dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-      dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-      fineLevel.Request("Graph", &dropFact);
-      fineLevel.Request("DofsPerNode", &dropFact);
-
-      dropFact.Build(fineLevel);
-
-      RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-      auto boundaryNodes = graph->GetBoundaryNodeMap();
-      bool allNodesAreOnBoundary = true;
-      for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
-        allNodesAreOnBoundary &= boundaryNodes[i];
-      TEST_EQUALITY(allNodesAreOnBoundary, true);
-    }
-
-    {
-      Level fineLevel;
-      TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-      mtx->SetFixedBlockSize(3);
-      fineLevel.Set("A", mtx);
-
-      CoalesceDropFactory dropFact = CoalesceDropFactory();
-      dropFact.SetDefaultVerbLevel(MueLu::Extreme);
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-      dropFact.SetParameter("lightweight wrap",Teuchos::ParameterEntry(true));
-      dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-      fineLevel.Request("Graph", &dropFact);
-      fineLevel.Request("DofsPerNode", &dropFact);
-
-      dropFact.Build(fineLevel);
-
-      RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
-
-      auto boundaryNodes = graph->GetBoundaryNodeMap();
-      bool allNodesAreOnBoundary = true;
-      for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
-        allNodesAreOnBoundary &= boundaryNodes[i];
-      TEST_EQUALITY(allNodesAreOnBoundary, true);
-    }
-
-  } // AggresiveDroppingIsMarkedAsBoundary
-
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignedClassicalSA, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
-#   include <MueLu_UseShortNames.hpp>
-    typedef Teuchos::ScalarTraits<SC> STS;
-    typedef typename STS::magnitudeType real_type;
-
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+    dropFact.SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(4.1));
+    fineLevel.Request("Graph", &dropFact);
+    fineLevel.Request("DofsPerNode", &dropFact);
 
-    GO nx = 10*comm->getSize();
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",nx);
-    matrixList.set("ny",(GO)10);
-    matrixList.set("nz",(GO)10);
-    matrixList.set("matrixType","Laplace3D");
-    RCP<Matrix> A =TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixList,lib);
+    dropFact.Build(fineLevel);
 
-    Level fineLevel;
-    fineLevel.Set("A", A);
-   
+    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &dropFact);
 
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    CoalesceDropFactory coalesceDropFact;
-    coalesceDropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    coalesceDropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(0.0));
-    coalesceDropFact.SetParameter("aggregation: drop scheme",Teuchos::ParameterEntry(std::string("signed classical sa")));
-    fineLevel.Request("Graph",&coalesceDropFact);
-    fineLevel.Request("DofsPerNode", &coalesceDropFact);
-
-    coalesceDropFact.Build(fineLevel);
-
-    RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
-    LO myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
-    TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+    auto boundaryNodes         = graph->GetBoundaryNodeMap();
+    bool allNodesAreOnBoundary = true;
+    for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
+      allNodesAreOnBoundary &= boundaryNodes[i];
+    TEST_EQUALITY(allNodesAreOnBoundary, true);
   }
 
-
-
-#define MUELU_ETI_GROUP(SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,Constructor,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,Build,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,PreDrop,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationBasic,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStrided,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStrided2,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStridedOffset,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationLightweight,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationLightweightDrop,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStridedLW,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStrided2LW,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStridedOffsetLW,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationDroppingLW,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AmalgamationStridedOffsetDropping2LW,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,DistanceLaplacian,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,DistanceLaplacianScaledCut,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,DistanceLaplacianUnscaledCut,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,DistanceLaplacianCutSym,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,ClassicalScaledCut,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,ClassicalUnScaledCut,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,SignaledClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,SignaledScaledCutClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,SignaledUnscaledCutClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonalColoredSignedClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonalNoColoredSignedClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonalSignedClassical,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonal,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonalDistanceLaplacian,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,BlockDiagonalDistanceLaplacianWeighted,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,DistanceLaplacianWeighted,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,AggresiveDroppingIsMarkedAsBoundary,SC,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory,SignedClassicalSA,SC,LO,GO,Node) \
+}  // AggresiveDroppingIsMarkedAsBoundary
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory, SignedClassicalSA, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  typedef Teuchos::ScalarTraits<SC> STS;
+  typedef typename STS::magnitudeType real_type;
+
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+
+  GO nx = 10 * comm->getSize();
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", (GO)10);
+  matrixList.set("nz", (GO)10);
+  matrixList.set("matrixType", "Laplace3D");
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixList, lib);
+
+  Level fineLevel;
+  fineLevel.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  CoalesceDropFactory coalesceDropFact;
+  coalesceDropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  coalesceDropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.0));
+  coalesceDropFact.SetParameter("aggregation: drop scheme", Teuchos::ParameterEntry(std::string("signed classical sa")));
+  fineLevel.Request("Graph", &coalesceDropFact);
+  fineLevel.Request("DofsPerNode", &coalesceDropFact);
+
+  coalesceDropFact.Build(fineLevel);
+
+  RCP<GraphBase> graph = fineLevel.Get<RCP<GraphBase> >("Graph", &coalesceDropFact);
+  LO myDofsPerNode     = fineLevel.Get<LO>("DofsPerNode", &coalesceDropFact);
+  TEST_EQUALITY(Teuchos::as<int>(myDofsPerNode) == 1, true);
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, Constructor, SC, LO, GO, Node)                            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, Build, SC, LO, GO, Node)                                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, PreDrop, SC, LO, GO, Node)                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationBasic, SC, LO, GO, Node)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStrided, SC, LO, GO, Node)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStrided2, SC, LO, GO, Node)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStridedOffset, SC, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationLightweight, SC, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationLightweightDrop, SC, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStridedLW, SC, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStrided2LW, SC, LO, GO, Node)                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStridedOffsetLW, SC, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationDroppingLW, SC, LO, GO, Node)                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AmalgamationStridedOffsetDropping2LW, SC, LO, GO, Node)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, DistanceLaplacian, SC, LO, GO, Node)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, DistanceLaplacianScaledCut, SC, LO, GO, Node)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, DistanceLaplacianUnscaledCut, SC, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, DistanceLaplacianCutSym, SC, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, ClassicalScaledCut, SC, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, ClassicalUnScaledCut, SC, LO, GO, Node)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, SignaledClassical, SC, LO, GO, Node)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, SignaledScaledCutClassical, SC, LO, GO, Node)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, SignaledUnscaledCutClassical, SC, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonalColoredSignedClassical, SC, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonalNoColoredSignedClassical, SC, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonalSignedClassical, SC, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonal, SC, LO, GO, Node)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonalDistanceLaplacian, SC, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, BlockDiagonalDistanceLaplacianWeighted, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, DistanceLaplacianWeighted, SC, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, AggresiveDroppingIsMarkedAsBoundary, SC, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory, SignedClassicalSA, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/CoarseMapFactory.cpp b/packages/muelu/test/unit_tests/CoarseMapFactory.cpp
index 93533ff6de03..fe17e236957e 100644
--- a/packages/muelu/test/unit_tests/CoarseMapFactory.cpp
+++ b/packages/muelu/test/unit_tests/CoarseMapFactory.cpp
@@ -58,117 +58,111 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<CoarseMapFactory> myCMF = rcp(new CoarseMapFactory());
-    TEST_ASSERT(!myCMF.is_null());
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, StandardCase, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    Level myLevel;
-    myLevel.SetLevelID(0);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    myLevel.Set("A", A);
-
-    // build dummy aggregate structure
-    Teuchos::RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
-    aggs->SetNumAggregates(10); // set (local!) number of aggregates
-    myLevel.Set("Aggregates", aggs);
-
-    // build dummy nullspace vector
-    Teuchos::RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(),1);
-    nsp->putScalar(1.0);
-    myLevel.Set("Nullspace", nsp);
-
-    RCP<CoarseMapFactory> myCMF = Teuchos::rcp(new CoarseMapFactory());
-    myLevel.Request("CoarseMap",myCMF.get());
-    myCMF->SetFactory("Aggregates",MueLu::NoFactory::getRCP());
-    myCMF->SetFactory("Nullspace",MueLu::NoFactory::getRCP());
-    myCMF->Build(myLevel);
-    Teuchos::RCP<const Map> myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap",myCMF.get());
-
-    TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
-    TEST_EQUALITY(myCoarseMap->getMaxLocalIndex()==9,true);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, NonStandardCaseA, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    Level myLevel;
-    myLevel.SetLevelID(0);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    myLevel.Set("A", A);
-
-    // build dummy aggregate structure
-    Teuchos::RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
-    aggs->SetNumAggregates(10); // set (local!) number of aggregates
-    myLevel.Set("Aggregates", aggs);
-
-    // build dummy nullspace vector
-    Teuchos::RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(),1);
-    nsp->putScalar(1.0);
-    myLevel.Set("Nullspace", nsp);
-
-    RCP<CoarseMapFactory> myCMF = Teuchos::rcp(new CoarseMapFactory());
-    myLevel.Request("CoarseMap",myCMF.get());
-    myCMF->SetParameter("Domain GID offsets",Teuchos::ParameterEntry(std::string("{100,50}")));
-    myCMF->SetFactory("Aggregates",MueLu::NoFactory::getRCP());
-    myCMF->SetFactory("Nullspace",MueLu::NoFactory::getRCP());
-    myCMF->Build(myLevel);
-    Teuchos::RCP<const Map> myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap",myCMF.get());
-
-    TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 100, true);
-    TEST_EQUALITY(myCoarseMap->getMaxLocalIndex()==9,true);
-
-    myLevel.Release("CoarseMap",myCMF.get());
-    myLevel.SetLevelID(1);
-    myLevel.Request("CoarseMap",myCMF.get());
-    myCMF->SetParameter("Domain GID offsets",Teuchos::ParameterEntry(std::string("{100,50}")));
-    myCMF->SetFactory("Aggregates",MueLu::NoFactory::getRCP());
-    myCMF->SetFactory("Nullspace",MueLu::NoFactory::getRCP());
-    myCMF->Build(myLevel);
-    myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap",myCMF.get());
-
-    TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 50, true);
-    TEST_EQUALITY(myCoarseMap->getMaxLocalIndex()==9,true);
-
-    myLevel.Release("CoarseMap",myCMF.get());
-    myLevel.SetLevelID(2);
-    myLevel.Request("CoarseMap",myCMF.get());
-    myCMF->SetParameter("Domain GID offsets",Teuchos::ParameterEntry(std::string("{100,50}")));
-    myCMF->SetFactory("Aggregates",MueLu::NoFactory::getRCP());
-    myCMF->SetFactory("Nullspace",MueLu::NoFactory::getRCP());
-    myCMF->Build(myLevel);
-    myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap",myCMF.get());
-
-    TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
-    TEST_EQUALITY(myCoarseMap->getMaxLocalIndex()==9,true);
-  } // NonStandardCaseA
-
-  ///////////////////////////////////////////////////////////////////////////
-
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<CoarseMapFactory> myCMF = rcp(new CoarseMapFactory());
+  TEST_ASSERT(!myCMF.is_null());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, StandardCase, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  Level myLevel;
+  myLevel.SetLevelID(0);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  myLevel.Set("A", A);
+
+  // build dummy aggregate structure
+  Teuchos::RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
+  aggs->SetNumAggregates(10);  // set (local!) number of aggregates
+  myLevel.Set("Aggregates", aggs);
+
+  // build dummy nullspace vector
+  Teuchos::RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nsp->putScalar(1.0);
+  myLevel.Set("Nullspace", nsp);
+
+  RCP<CoarseMapFactory> myCMF = Teuchos::rcp(new CoarseMapFactory());
+  myLevel.Request("CoarseMap", myCMF.get());
+  myCMF->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
+  myCMF->SetFactory("Nullspace", MueLu::NoFactory::getRCP());
+  myCMF->Build(myLevel);
+  Teuchos::RCP<const Map> myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", myCMF.get());
+
+  TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
+  TEST_EQUALITY(myCoarseMap->getMaxLocalIndex() == 9, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMapFactory, NonStandardCaseA, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  Level myLevel;
+  myLevel.SetLevelID(0);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  myLevel.Set("A", A);
+
+  // build dummy aggregate structure
+  Teuchos::RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
+  aggs->SetNumAggregates(10);  // set (local!) number of aggregates
+  myLevel.Set("Aggregates", aggs);
+
+  // build dummy nullspace vector
+  Teuchos::RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nsp->putScalar(1.0);
+  myLevel.Set("Nullspace", nsp);
+
+  RCP<CoarseMapFactory> myCMF = Teuchos::rcp(new CoarseMapFactory());
+  myLevel.Request("CoarseMap", myCMF.get());
+  myCMF->SetParameter("Domain GID offsets", Teuchos::ParameterEntry(std::string("{100,50}")));
+  myCMF->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
+  myCMF->SetFactory("Nullspace", MueLu::NoFactory::getRCP());
+  myCMF->Build(myLevel);
+  Teuchos::RCP<const Map> myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", myCMF.get());
+
+  TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 100, true);
+  TEST_EQUALITY(myCoarseMap->getMaxLocalIndex() == 9, true);
+
+  myLevel.Release("CoarseMap", myCMF.get());
+  myLevel.SetLevelID(1);
+  myLevel.Request("CoarseMap", myCMF.get());
+  myCMF->SetParameter("Domain GID offsets", Teuchos::ParameterEntry(std::string("{100,50}")));
+  myCMF->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
+  myCMF->SetFactory("Nullspace", MueLu::NoFactory::getRCP());
+  myCMF->Build(myLevel);
+  myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", myCMF.get());
+
+  TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 50, true);
+  TEST_EQUALITY(myCoarseMap->getMaxLocalIndex() == 9, true);
+
+  myLevel.Release("CoarseMap", myCMF.get());
+  myLevel.SetLevelID(2);
+  myLevel.Request("CoarseMap", myCMF.get());
+  myCMF->SetParameter("Domain GID offsets", Teuchos::ParameterEntry(std::string("{100,50}")));
+  myCMF->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
+  myCMF->SetFactory("Nullspace", MueLu::NoFactory::getRCP());
+  myCMF->Build(myLevel);
+  myCoarseMap = myLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", myCMF.get());
+
+  TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
+  TEST_EQUALITY(myCoarseMap->getMaxLocalIndex() == 9, true);
+}  // NonStandardCaseA
+
+///////////////////////////////////////////////////////////////////////////
+
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoarseMapFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)  \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoarseMapFactory, StandardCase, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoarseMapFactory, NonStandardCaseA, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/FineLevelInputDataFactory.cpp b/packages/muelu/test/unit_tests/FineLevelInputDataFactory.cpp
index 540373cc5334..3dcc4f440988 100644
--- a/packages/muelu/test/unit_tests/FineLevelInputDataFactory.cpp
+++ b/packages/muelu/test/unit_tests/FineLevelInputDataFactory.cpp
@@ -54,272 +54,258 @@
 
 namespace MueLuTests {
 
-  /////////////////////////
-  // helper function
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal, GlobalOrdinal, Node> >
-  GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map,
-                        Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
-#   include "MueLu_UseShortNames.hpp"
-
-    Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(map, 3);
-
-    LocalOrdinal NumMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
-    GlobalOrdinal NumGlobalElements = map->getGlobalNumElements();
-    GlobalOrdinal nIndexBase = map->getIndexBase();
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-    {
-      if (MyGlobalElements[i] == nIndexBase)
-      {
-        // off-diagonal for first row
-        Indices[0] = nIndexBase;
-        NumEntries = 1;
-        Values[0] = c;
-      }
-      else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1)
-      {
-        // off-diagonal for last row
-        Indices[0] = nIndexBase + NumGlobalElements - 2;
-        NumEntries = 1;
-        Values[0] = b;
-      }
-      else
-      {
-        // off-diagonal for internal row
-        Indices[0] = MyGlobalElements[i] - 1;
-        Values[1] = b;
-        Indices[1] = MyGlobalElements[i] + 1;
-        Values[0] = c;
-        NumEntries = 2;
-      }
-
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(MyGlobalElements[i],
-          Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-          Teuchos::tuple<Scalar>(a) );
-
-    } //for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-
-    mtx->fillComplete(map,map);
-
-    return mtx;
-  }
+/////////////////////////
+// helper function
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map,
+                      Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
+#include "MueLu_UseShortNames.hpp"
+
+  Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(map, 3);
+
+  LocalOrdinal NumMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
+  GlobalOrdinal NumGlobalElements                          = map->getGlobalNumElements();
+  GlobalOrdinal nIndexBase                                 = map->getIndexBase();
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  for (LocalOrdinal i = 0; i < NumMyElements; ++i) {
+    if (MyGlobalElements[i] == nIndexBase) {
+      // off-diagonal for first row
+      Indices[0] = nIndexBase;
+      NumEntries = 1;
+      Values[0]  = c;
+    } else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1) {
+      // off-diagonal for last row
+      Indices[0] = nIndexBase + NumGlobalElements - 2;
+      NumEntries = 1;
+      Values[0]  = b;
+    } else {
+      // off-diagonal for internal row
+      Indices[0] = MyGlobalElements[i] - 1;
+      Values[1]  = b;
+      Indices[1] = MyGlobalElements[i] + 1;
+      Values[0]  = c;
+      NumEntries = 2;
+    }
 
-  // dummy helper class (placeholder)
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class DummyFactory : public MueLu::SingleLevelFactoryBase {
-  public:
-    DummyFactory() { }
-    virtual ~DummyFactory() { }
-    RCP<const Teuchos::ParameterList> GetValidParameterList() const { return Teuchos::null; };
-    void DeclareInput(MueLu::Level& currentLevel) const {};
-    void Build(MueLu::Level& currentLevel) const {};
-  }; //class DummyFactory
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, InputData, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(MyGlobalElements[i],
+                            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                            Teuchos::tuple<Scalar>(a));
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  }  // for (LocalOrdinal i = 0; i < NumMyElements; ++i)
 
-    /**********************************************************************************/
-    /* CREATE INITIAL MATRIX                                                          */
-    /**********************************************************************************/
-    RCP<const Map> map;
-    GO numElements = 100;
+  mtx->fillComplete(map, map);
 
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  return mtx;
+}
 
-    map   = MapFactory::Build(lib, numElements, 0, comm);
+// dummy helper class (placeholder)
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class DummyFactory : public MueLu::SingleLevelFactoryBase {
+ public:
+  DummyFactory() {}
+  virtual ~DummyFactory() {}
+  RCP<const Teuchos::ParameterList> GetValidParameterList() const { return Teuchos::null; };
+  void DeclareInput(MueLu::Level& currentLevel) const {};
+  void Build(MueLu::Level& currentLevel) const {};
+};  // class DummyFactory
 
-    RCP<Matrix> A = GenerateProblemMatrix<Scalar, LO, GO, Node>(map,2,-1,-1);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, InputData, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    levelOne->SetLevelID(0);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    FineLevelInputDataFactory inputData;
+  /**********************************************************************************/
+  /* CREATE INITIAL MATRIX                                                          */
+  /**********************************************************************************/
+  RCP<const Map> map;
+  GO numElements = 100;
 
-#ifdef HAVE_MUELU_DEBUG
-    inputData.DisableMultipleCallCheck();
-#endif
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
+  map = MapFactory::Build(lib, numElements, 0, comm);
 
-    // Test 1: different fine level variable name and coarse level variable name, default fine level factory (=NoFactory)
-    levelOne->Set("A-output", A);
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output")));
+  RCP<Matrix> A = GenerateProblemMatrix<Scalar, LO, GO, Node>(map, 2, -1, -1);
 
-    levelOne->Request("A-output", &inputData);
-    RCP<Matrix> AA = levelOne->Get<RCP<Matrix> >("A-output", &inputData);
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  levelOne->SetLevelID(0);
 
-    TEST_EQUALITY(AA.get(), A.get());
+  FineLevelInputDataFactory inputData;
 
-    levelOne->Release("A-output", &inputData);
+#ifdef HAVE_MUELU_DEBUG
+  inputData.DisableMultipleCallCheck();
+#endif
 
-    // Test 2: set fine level factory
-    MueLuTests::DummyFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> dummyFact;
-    levelOne->Set("A-output2", A);
-    inputData.SetFactory("Fine level factory", MueLu::NoFactory::getRCP());
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output2")));
+  // Test 1: different fine level variable name and coarse level variable name, default fine level factory (=NoFactory)
+  levelOne->Set("A-output", A);
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output")));
 
-    levelOne->Request("A-output2", &inputData);
-    AA = levelOne->Get<RCP<Matrix> >("A-output2", &inputData);
-    TEST_EQUALITY(AA.get(), A.get());
-    levelOne->Release("A-output2", &inputData);
+  levelOne->Request("A-output", &inputData);
+  RCP<Matrix> AA = levelOne->Get<RCP<Matrix> >("A-output", &inputData);
 
-    // Test 3: same as test 2 on coarse level
-    levelOne->SetLevelID(1);
-    levelOne->Request("A-output4", &dummyFact);
-    levelOne->Set("A-output4", A, &dummyFact);
+  TEST_EQUALITY(AA.get(), A.get());
 
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output4")));
+  levelOne->Release("A-output", &inputData);
 
-    levelOne->Request("A-output4", &inputData);
+  // Test 2: set fine level factory
+  MueLuTests::DummyFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> dummyFact;
+  levelOne->Set("A-output2", A);
+  inputData.SetFactory("Fine level factory", MueLu::NoFactory::getRCP());
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output2")));
 
-    AA = levelOne->Get<RCP<Matrix> >("A-output4", &inputData);
+  levelOne->Request("A-output2", &inputData);
+  AA = levelOne->Get<RCP<Matrix> >("A-output2", &inputData);
+  TEST_EQUALITY(AA.get(), A.get());
+  levelOne->Release("A-output2", &inputData);
 
-    TEST_EQUALITY(AA.get(), A.get());
+  // Test 3: same as test 2 on coarse level
+  levelOne->SetLevelID(1);
+  levelOne->Request("A-output4", &dummyFact);
+  levelOne->Set("A-output4", A, &dummyFact);
 
-    // Test 4: same as test 2 on coarse level
-    levelOne->Request("A-output5", &dummyFact);
-    levelOne->Set("A-output5", A, &dummyFact);
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output4")));
 
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output5")));
-    levelOne->Request("A-output5", &inputData);
-    AA = levelOne->Get<RCP<Matrix> >("A-output5", &inputData);
-    TEST_EQUALITY(AA.get(), A.get());
-  } // InputData
+  levelOne->Request("A-output4", &inputData);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, InputDataMap, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  AA = levelOne->Get<RCP<Matrix> >("A-output4", &inputData);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  TEST_EQUALITY(AA.get(), A.get());
 
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  // Test 4: same as test 2 on coarse level
+  levelOne->Request("A-output5", &dummyFact);
+  levelOne->Set("A-output5", A, &dummyFact);
 
-    RCP<Map> map   = MapFactory::Build(lib, 100, 0, comm);
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("A-output5")));
+  levelOne->Request("A-output5", &inputData);
+  AA = levelOne->Get<RCP<Matrix> >("A-output5", &inputData);
+  TEST_EQUALITY(AA.get(), A.get());
+}  // InputData
 
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    levelOne->SetLevelID(0);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, InputDataMap, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    FineLevelInputDataFactory inputData;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-#ifdef HAVE_MUELU_DEBUG
-    inputData.DisableMultipleCallCheck();
-#endif
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
+  RCP<Map> map = MapFactory::Build(lib, 100, 0, comm);
 
-    // Test 1: different fine level variable name and coarse level variable name, default fine level factory (=NoFactory)
-    levelOne->Set("m", map);
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  levelOne->SetLevelID(0);
 
-    levelOne->Request("m", &inputData);
+  FineLevelInputDataFactory inputData;
 
-    TEST_THROW(levelOne->Get<RCP<Map> >("m", &inputData), Teuchos::bad_any_cast);
+#ifdef HAVE_MUELU_DEBUG
+  inputData.DisableMultipleCallCheck();
+#endif
 
-    inputData.SetParameter("Variable type", Teuchos::ParameterEntry(std::string("Map")));
-    RCP<Map> mm = levelOne->Get<RCP<Map> >("m", &inputData);
-    TEST_EQUALITY(mm.get(), map.get());
+  // Test 1: different fine level variable name and coarse level variable name, default fine level factory (=NoFactory)
+  levelOne->Set("m", map);
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
 
-    levelOne->Release("m", &inputData);
+  levelOne->Request("m", &inputData);
 
-    // Test 2: set fine level factory
-    MueLuTests::DummyFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> dummyFact;
-    levelOne->Set("m", map);
-    inputData.SetFactory("Fine level factory", MueLu::NoFactory::getRCP());
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
+  TEST_THROW(levelOne->Get<RCP<Map> >("m", &inputData), Teuchos::bad_any_cast);
 
-    levelOne->Request("m", &inputData);
-    mm = levelOne->Get<RCP<Map> >("m", &inputData);
-    TEST_EQUALITY(mm.get(), map.get());
-    levelOne->Release("m", &inputData);
+  inputData.SetParameter("Variable type", Teuchos::ParameterEntry(std::string("Map")));
+  RCP<Map> mm = levelOne->Get<RCP<Map> >("m", &inputData);
+  TEST_EQUALITY(mm.get(), map.get());
 
-    // Test 3: same as test 2 on coarse level
-    levelOne->SetLevelID(1);
-    levelOne->Request("m", &dummyFact);
-    levelOne->Set("m", map, &dummyFact);
+  levelOne->Release("m", &inputData);
 
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
+  // Test 2: set fine level factory
+  MueLuTests::DummyFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> dummyFact;
+  levelOne->Set("m", map);
+  inputData.SetFactory("Fine level factory", MueLu::NoFactory::getRCP());
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
 
-    levelOne->Request("m", &inputData);
+  levelOne->Request("m", &inputData);
+  mm = levelOne->Get<RCP<Map> >("m", &inputData);
+  TEST_EQUALITY(mm.get(), map.get());
+  levelOne->Release("m", &inputData);
 
-    mm = levelOne->Get<RCP<Map> >("m", &inputData);
+  // Test 3: same as test 2 on coarse level
+  levelOne->SetLevelID(1);
+  levelOne->Request("m", &dummyFact);
+  levelOne->Set("m", map, &dummyFact);
 
-    TEST_EQUALITY(mm.get(), map.get());
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
 
+  levelOne->Request("m", &inputData);
 
-    // Test 4: same as test 2 on coarse level
-    levelOne->Request("m", &dummyFact);
-    levelOne->Set("m", map, &dummyFact);
+  mm = levelOne->Get<RCP<Map> >("m", &inputData);
 
-    inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
-    inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
-    levelOne->Request("m", &inputData);
-    mm = levelOne->Get<RCP<Map> >("m", &inputData);
-    TEST_EQUALITY(mm.get(), map.get());
-  } // InputDataMap
+  TEST_EQUALITY(mm.get(), map.get());
 
-  // helper class for testing functionality of FineLevelInputDataFactory
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class FineLevelInputDataFactoryTester {
-#   include "MueLu_UseShortNames.hpp"
-  public:
-    void test_testfunc(const FineLevelInputDataFactory& fac) {
-      //std::cout << "FineLevelInputDataFactoryTester" << std::endl;
-      fac.test();
-    }
-  };
+  // Test 4: same as test 2 on coarse level
+  levelOne->Request("m", &dummyFact);
+  levelOne->Set("m", map, &dummyFact);
 
-  // unit test just for demonstration purposes
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, TestFunc, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  inputData.SetFactory("Coarse level factory", Teuchos::rcpFromRef(dummyFact));
+  inputData.SetParameter("Variable", Teuchos::ParameterEntry(std::string("m")));
+  levelOne->Request("m", &inputData);
+  mm = levelOne->Get<RCP<Map> >("m", &inputData);
+  TEST_EQUALITY(mm.get(), map.get());
+}  // InputDataMap
 
-    FineLevelInputDataFactory fac;
+// helper class for testing functionality of FineLevelInputDataFactory
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class FineLevelInputDataFactoryTester {
+#include "MueLu_UseShortNames.hpp"
+ public:
+  void test_testfunc(const FineLevelInputDataFactory& fac) {
+    // std::cout << "FineLevelInputDataFactoryTester" << std::endl;
+    fac.test();
+  }
+};
 
-    FineLevelInputDataFactoryTester<Scalar,LocalOrdinal,GlobalOrdinal,Node> tester;
+// unit test just for demonstration purposes
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FineLevelInputDataFactory, TestFunc, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    tester.test_testfunc(fac);
+  FineLevelInputDataFactory fac;
 
-    //TEST_EQUALITY(AA.get(), A.get());
-  } // TestFunc
+  FineLevelInputDataFactoryTester<Scalar, LocalOrdinal, GlobalOrdinal, Node> tester;
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, InputData, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, InputDataMap, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, TestFunc, SC, LO, GO, Node) \
+  tester.test_testfunc(fac);
 
-#include <MueLu_ETI_4arg.hpp>
-}
+  // TEST_EQUALITY(AA.get(), A.get());
+}  // TestFunc
 
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, InputData, SC, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, InputDataMap, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FineLevelInputDataFactory, TestFunc, SC, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Galeri.cpp b/packages/muelu/test/unit_tests/Galeri.cpp
index fb61f799a664..546097ab5de3 100644
--- a/packages/muelu/test/unit_tests/Galeri.cpp
+++ b/packages/muelu/test/unit_tests/Galeri.cpp
@@ -56,16 +56,16 @@
 TPETRA_ETI_MANGLING_TYPEDEFS()
 
 namespace Galeri {
-  namespace Xpetra {
+namespace Xpetra {
 
 // Add other Galeri functions as needed
-#define MUELU_ETI_GROUP(LO,GO,NO) \
-  template Teuchos::RCP<::Xpetra::Map<LO, GO, NO>> CreateMap<LO,GO,NO>(::Xpetra::UnderlyingLib lib, const std::string & mapType, const Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList & list);
+#define MUELU_ETI_GROUP(LO, GO, NO) \
+  template Teuchos::RCP<::Xpetra::Map<LO, GO, NO>> CreateMap<LO, GO, NO>(::Xpetra::UnderlyingLib lib, const std::string& mapType, const Teuchos::RCP<const Teuchos::Comm<int>>& comm, Teuchos::ParameterList& list);
 
 #include <MueLu_ETI_3arg.hpp>
 
-  } //Xpetra namespace
-} //Galeri namespace
+}  // namespace Xpetra
+}  // namespace Galeri
 
-#endif //ifdef HAVE_GALERI_XPETRA
+#endif  // ifdef HAVE_GALERI_XPETRA
 //#endif //ifdef HAVE_MUELU_EXPLICIT_INSTANTIATION
diff --git a/packages/muelu/test/unit_tests/GeneralGeometricPFactory.cpp b/packages/muelu/test/unit_tests/GeneralGeometricPFactory.cpp
index a19e415c6f26..8b824839ccf0 100644
--- a/packages/muelu/test/unit_tests/GeneralGeometricPFactory.cpp
+++ b/packages/muelu/test/unit_tests/GeneralGeometricPFactory.cpp
@@ -72,1286 +72,1308 @@
 
 namespace MueLuTests {
 
-  template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
-  class GeneralGeometricPFactoryTester {
+template <class Scalar = double, class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = Tpetra::KokkosClassic::DefaultNode::DefaultNodeType>
+class GeneralGeometricPFactoryTester {
 #include "MueLu_UseShortNames.hpp"
 
-  public:
-    //! @name Constructors/Destructors.
-    //@{
-
-    //! Constructor
-    GeneralGeometricPFactoryTester() { }
-
-    //! Destructor.
-    virtual ~GeneralGeometricPFactoryTester() { }
-    //@}
-
-    void TestComputeLinearInterpolationStencil(const GeneralGeometricPFactory& fac,
-                                               const LO numDimension,
-                                               const Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord,
-                                               std::vector<double>& stencil)
-      const{
-      // Call the method to be tested.
-      MueLu::GeneralGeometricPFactory<SC,LO,GO,Node> myGGPFactory;
-      myGGPFactory.ComputeLinearInterpolationStencil(numDimension, coord, stencil);
-    };
+ public:
+  //! @name Constructors/Destructors.
+  //@{
+
+  //! Constructor
+  GeneralGeometricPFactoryTester() {}
+
+  //! Destructor.
+  virtual ~GeneralGeometricPFactoryTester() {}
+  //@}
+
+  void TestComputeLinearInterpolationStencil(const GeneralGeometricPFactory& fac,
+                                             const LO numDimension,
+                                             const Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord,
+                                             std::vector<double>& stencil)
+      const {
+    // Call the method to be tested.
+    MueLu::GeneralGeometricPFactory<SC, LO, GO, Node> myGGPFactory;
+    myGGPFactory.ComputeLinearInterpolationStencil(numDimension, coord, stencil);
   };
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GGGetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib,
-                        const LocalOrdinal numDimensions, const std::string mode,
-                        RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
-                        RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> >&Coordinates,
-                        RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
-                        Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
+};
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GGGetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib,
+                      const LocalOrdinal numDimensions, const std::string mode,
+                      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
+                      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& Coordinates,
+                      RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
+                      Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
 #include "MueLu_UseShortNames.hpp"
 
-    GO nx = 9;
-    GO ny = 9;
-    GO nz = (numDimensions < 3 ? 1 : 9);
-    GO gNumPoints = nx*ny*nz;
-
-    gNodesPerDim[0] = nx;
-    gNodesPerDim[1] = ny;
-    gNodesPerDim[2] = nz;
-
-    GO myOffset = 0, myGIDOffset = 0;
-    if(comm->getSize() == 1) {
-      myOffset = 0;
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(numDimensions == 2) {
-          myOffset = 0;
-          myGIDOffset = 0;
-          lNodesPerDim[0] = 5;
-          lNodesPerDim[1] = 5;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 0;
-          myGIDOffset = 0;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 5;
-          lNodesPerDim[2] = 5;
-        }
-      } else if(comm->getRank() == 1) {
-        if(numDimensions == 2) {
-          myOffset = 4;
-          myGIDOffset = 16;
-          lNodesPerDim[0] = 4;
-          lNodesPerDim[1] = 5;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 45;
-          myGIDOffset = 225;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 5;
-        }
-      } else if(comm->getRank() == 2) {
-        if(numDimensions == 2) {
-          myOffset = 21;
-          myGIDOffset = 28;
-          lNodesPerDim[0] = 5;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 405;
-          myGIDOffset = 405;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 5;
-          lNodesPerDim[2] = 4;
-        }
-      } else if(comm->getRank() == 3) {
-        if(numDimensions == 2) {
-          myOffset = 25;
-          myGIDOffset = 44;
-          lNodesPerDim[0] = 4;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 450;
-          myGIDOffset = 585;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 4;
-          lNodesPerDim[2] = 4;
-        }
+  GO nx         = 9;
+  GO ny         = 9;
+  GO nz         = (numDimensions < 3 ? 1 : 9);
+  GO gNumPoints = nx * ny * nz;
+
+  gNodesPerDim[0] = nx;
+  gNodesPerDim[1] = ny;
+  gNodesPerDim[2] = nz;
+
+  GO myOffset = 0, myGIDOffset = 0;
+  if (comm->getSize() == 1) {
+    myOffset        = 0;
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (numDimensions == 2) {
+        myOffset        = 0;
+        myGIDOffset     = 0;
+        lNodesPerDim[0] = 5;
+        lNodesPerDim[1] = 5;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 0;
+        myGIDOffset     = 0;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 5;
+        lNodesPerDim[2] = 5;
+      }
+    } else if (comm->getRank() == 1) {
+      if (numDimensions == 2) {
+        myOffset        = 4;
+        myGIDOffset     = 16;
+        lNodesPerDim[0] = 4;
+        lNodesPerDim[1] = 5;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 45;
+        myGIDOffset     = 225;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 5;
+      }
+    } else if (comm->getRank() == 2) {
+      if (numDimensions == 2) {
+        myOffset        = 21;
+        myGIDOffset     = 28;
+        lNodesPerDim[0] = 5;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 405;
+        myGIDOffset     = 405;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 5;
+        lNodesPerDim[2] = 4;
+      }
+    } else if (comm->getRank() == 3) {
+      if (numDimensions == 2) {
+        myOffset        = 25;
+        myGIDOffset     = 44;
+        lNodesPerDim[0] = 4;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 450;
+        myGIDOffset     = 585;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 4;
+        lNodesPerDim[2] = 4;
       }
     }
+  }
 
-    GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
-    if(numDimensions == 2) {
-      myZoffset = 0;
-      myYoffset = myOffset / gNodesPerDim[0];
-      myXoffset = myOffset % gNodesPerDim[0];
-    } else if(numDimensions == 3) {
-      myZoffset = myOffset / (gNodesPerDim[1]*gNodesPerDim[0]);
-      myYoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) / gNodesPerDim[0];
-      myXoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) % gNodesPerDim[0];
-    }
-    LO lNumPoints = lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2];
-
-    // Construct map and local coordinates
-    Teuchos::Array<GO>     myGIDs(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
-    for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          if(mode == "Global Lexicographic") {
-            myGIDs[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = myOffset
-              + k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0] + i;
-          } else if(mode == "Local Lexicographic") {
-            myGIDs[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = myGIDOffset
-              + k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i;
-          }
-          myXCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+  GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
+  if (numDimensions == 2) {
+    myZoffset = 0;
+    myYoffset = myOffset / gNodesPerDim[0];
+    myXoffset = myOffset % gNodesPerDim[0];
+  } else if (numDimensions == 3) {
+    myZoffset = myOffset / (gNodesPerDim[1] * gNodesPerDim[0]);
+    myYoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) / gNodesPerDim[0];
+    myXoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) % gNodesPerDim[0];
+  }
+  LO lNumPoints = lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2];
+
+  // Construct map and local coordinates
+  Teuchos::Array<GO> myGIDs(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
+  for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+    for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        if (mode == "Global Lexicographic") {
+          myGIDs[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = myOffset + k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + i;
+        } else if (mode == "Local Lexicographic") {
+          myGIDs[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = myGIDOffset + k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i;
+        }
+        myXCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (i + myXoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
-          myYCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+        myYCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (j + myYoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
-          myZCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] =
+        myZCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] =
             (k + myZoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
-        }
       }
     }
+  }
 
-    Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
-    if(numDimensions == 1) {
-      myCoordinates[0] = myXCoords();
-    } else if(numDimensions == 2) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-    } else if(numDimensions == 3) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-      myCoordinates[2] = myZCoords();
-    }
-
-    // Create the map and store coordinates using the above array views
-    map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
-    Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map, myCoordinates(),
-                                                                     numDimensions);
-
-    // small parameter list for Galeri
-    Teuchos::ParameterList problemList;
-    if(numDimensions == 1) {
-      problemList.set("nx",gNodesPerDim[0]);
-    } else if(numDimensions == 2) {
-      problemList.set("nx",gNodesPerDim[0]);
-      problemList.set("ny",gNodesPerDim[1]);
-    } else if(numDimensions == 3) {
-      problemList.set("nx",gNodesPerDim[0]);
-      problemList.set("ny",gNodesPerDim[1]);
-      problemList.set("nz",gNodesPerDim[2]);
-    }
-    // problemList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    if(numDimensions == 1) {
-      Galeri::Xpetra::Laplace1DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList,
-                                                                                       map);
-      Op = Problem.BuildMatrix();
-    } else if(numDimensions == 2) {
-      Galeri::Xpetra::Laplace2DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList,
-                                                                                       map);
-      Op = Problem.BuildMatrix();
-    } else if(numDimensions == 3) {
-      Galeri::Xpetra::Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemList,
-                                                                                       map);
-      Op = Problem.BuildMatrix();
-    }
+  Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
+  if (numDimensions == 1) {
+    myCoordinates[0] = myXCoords();
+  } else if (numDimensions == 2) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+  } else if (numDimensions == 3) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+    myCoordinates[2] = myZCoords();
+  }
 
+  // Create the map and store coordinates using the above array views
+  map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
+  Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, myCoordinates(),
+                                                                                                                     numDimensions);
+
+  // small parameter list for Galeri
+  Teuchos::ParameterList problemList;
+  if (numDimensions == 1) {
+    problemList.set("nx", gNodesPerDim[0]);
+  } else if (numDimensions == 2) {
+    problemList.set("nx", gNodesPerDim[0]);
+    problemList.set("ny", gNodesPerDim[1]);
+  } else if (numDimensions == 3) {
+    problemList.set("nx", gNodesPerDim[0]);
+    problemList.set("ny", gNodesPerDim[1]);
+    problemList.set("nz", gNodesPerDim[2]);
+  }
+  // problemList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  if (numDimensions == 1) {
+    Galeri::Xpetra::Laplace1DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemList,
+                                                                                          map);
+    Op = Problem.BuildMatrix();
+  } else if (numDimensions == 2) {
+    Galeri::Xpetra::Laplace2DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemList,
+                                                                                          map);
+    Op = Problem.BuildMatrix();
+  } else if (numDimensions == 3) {
+    Galeri::Xpetra::Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemList,
+                                                                                          map);
+    Op = Problem.BuildMatrix();
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, Constructor, Scalar, LocalOrdinal,
-                                    GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, Constructor, Scalar, LocalOrdinal,
+                                  GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<GeneralGeometricPFactory> ggPFact = rcp(new GeneralGeometricPFactory);
-    TEST_EQUALITY(ggPFact != Teuchos::null, true);
+  RCP<GeneralGeometricPFactory> ggPFact = rcp(new GeneralGeometricPFactory);
+  TEST_EQUALITY(ggPFact != Teuchos::null, true);
 
-  } //Constructor
+}  // Constructor
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LinearInterpolation, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    MueLu::GeneralGeometricPFactory<SC,LO,GO,Node> ggPFact;
-    GeneralGeometricPFactoryTester<SC,LO,GO,Node> factTester;
-
-    LO numDimension = 3;
-    Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord(9);
-    for(int node = 0; node < 9; ++node) {coord[node].resize(3);}
-    coord[0][0] = 0.3; coord[0][1] = 0.9; coord[0][2] = 0.1;
-    coord[1][0] = 0.0; coord[1][1] = 0.0; coord[1][2] = 0.0;
-    coord[2][0] = 1.0; coord[2][1] = 0.0; coord[2][2] = 0.0;
-    coord[3][0] = 0.0; coord[3][1] = 1.0; coord[3][2] = 0.0;
-    coord[4][0] = 1.0; coord[4][1] = 1.0; coord[4][2] = 0.0;
-    coord[5][0] = 0.0; coord[5][1] = 0.0; coord[5][2] = 1.0;
-    coord[6][0] = 1.0; coord[6][1] = 0.0; coord[6][2] = 1.0;
-    coord[7][0] = 0.0; coord[7][1] = 1.0; coord[7][2] = 1.0;
-    coord[8][0] = 1.0; coord[8][1] = 1.0; coord[8][2] = 1.0;
-    std::vector<double> stencil(8);
-    factTester.TestComputeLinearInterpolationStencil(ggPFact, numDimension, coord, stencil);
-
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType x = 0.0, y = 0.0, z = 0.0;
-    for(LO i = 0; i < 8; ++i) {
-      x += stencil[i]*coord[i+1][0];
-      y += stencil[i]*coord[i+1][1];
-      z += stencil[i]*coord[i+1][2];
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LinearInterpolation, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    TEST_EQUALITY((std::abs(x - coord[0][0]) < 1e-5)
-                  && (std::abs(y - coord[0][1]) < 1e-5)
-                  && (std::abs(z - coord[0][2]) < 1e-5), true);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  } // LinearInterpolation
+  MueLu::GeneralGeometricPFactory<SC, LO, GO, Node> ggPFact;
+  GeneralGeometricPFactoryTester<SC, LO, GO, Node> factTester;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LinearExtrapolation, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    MueLu::GeneralGeometricPFactory<SC,LO,GO,Node> ggPFact;
-    GeneralGeometricPFactoryTester<SC,LO,GO,Node> factTester;
-
-    LO numDimension = 3;
-    Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord(9);
-    for(int node = 0; node < 9; ++node) {coord[node].resize(3);}
-    coord[0][0] = 1.1; coord[0][1] = 0.3; coord[0][2] = 0.8;
-    coord[1][0] = 0.0; coord[1][1] = 0.0; coord[1][2] = 0.0;
-    coord[2][0] = 1.0; coord[2][1] = 0.0; coord[2][2] = 0.0;
-    coord[3][0] = 0.0; coord[3][1] = 1.0; coord[3][2] = 0.0;
-    coord[4][0] = 1.0; coord[4][1] = 1.0; coord[4][2] = 0.0;
-    coord[5][0] = 0.0; coord[5][1] = 0.0; coord[5][2] = 1.0;
-    coord[6][0] = 1.0; coord[6][1] = 0.0; coord[6][2] = 1.0;
-    coord[7][0] = 0.0; coord[7][1] = 1.0; coord[7][2] = 1.0;
-    coord[8][0] = 1.0; coord[8][1] = 1.0; coord[8][2] = 1.0;
-    std::vector<double> stencil(8);
-    factTester.TestComputeLinearInterpolationStencil(ggPFact, numDimension, coord, stencil);
-
-    typename Teuchos::ScalarTraits<Scalar>::magnitudeType x = 0.0, y = 0.0, z = 0.0;
-    for(LO i = 0; i < 8; ++i) {
-      x += stencil[i]*coord[i+1][0];
-      y += stencil[i]*coord[i+1][1];
-      z += stencil[i]*coord[i+1][2];
-    }
+  LO numDimension = 3;
+  Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord(9);
+  for (int node = 0; node < 9; ++node) {
+    coord[node].resize(3);
+  }
+  coord[0][0] = 0.3;
+  coord[0][1] = 0.9;
+  coord[0][2] = 0.1;
+  coord[1][0] = 0.0;
+  coord[1][1] = 0.0;
+  coord[1][2] = 0.0;
+  coord[2][0] = 1.0;
+  coord[2][1] = 0.0;
+  coord[2][2] = 0.0;
+  coord[3][0] = 0.0;
+  coord[3][1] = 1.0;
+  coord[3][2] = 0.0;
+  coord[4][0] = 1.0;
+  coord[4][1] = 1.0;
+  coord[4][2] = 0.0;
+  coord[5][0] = 0.0;
+  coord[5][1] = 0.0;
+  coord[5][2] = 1.0;
+  coord[6][0] = 1.0;
+  coord[6][1] = 0.0;
+  coord[6][2] = 1.0;
+  coord[7][0] = 0.0;
+  coord[7][1] = 1.0;
+  coord[7][2] = 1.0;
+  coord[8][0] = 1.0;
+  coord[8][1] = 1.0;
+  coord[8][2] = 1.0;
+  std::vector<double> stencil(8);
+  factTester.TestComputeLinearInterpolationStencil(ggPFact, numDimension, coord, stencil);
+
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType x = 0.0, y = 0.0, z = 0.0;
+  for (LO i = 0; i < 8; ++i) {
+    x += stencil[i] * coord[i + 1][0];
+    y += stencil[i] * coord[i + 1][1];
+    z += stencil[i] * coord[i + 1][2];
+  }
 
-    TEST_EQUALITY((std::abs(x - coord[0][0]) < 1e-5) && (std::abs(y - coord[0][1]) < 1e-5)
-                  && (std::abs(z - coord[0][2]) < 1e-5), true);
+  TEST_EQUALITY((std::abs(x - coord[0][0]) < 1e-5) && (std::abs(y - coord[0][1]) < 1e-5) && (std::abs(z - coord[0][2]) < 1e-5), true);
 
-  } // LinearExtrapolation
+}  // LinearInterpolation
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, PoissonOnCubeLinear, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 2;
-    // LocalOrdinal its=10;
-    GO nx = 4;
-    GO ny = 4;
-    GO nz = 4;
-    GO numPoints = nx*ny*nz;
-    Array<GO> gNodesPerDim(3);
-    gNodesPerDim[0] = 4;
-    gNodesPerDim[1] = 4;
-    gNodesPerDim[2] = 4;
-    Array<LO> lNodesPerDim(3);
-    if(comm->getSize() == 1) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-    } else if(comm->getSize() == 4) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz/4;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LinearExtrapolation, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    const RCP<const Map> map = MapFactory::Build(lib,
-                                                 gNodesPerDim[0]*gNodesPerDim[1]*gNodesPerDim[2],
-                                                 lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2],
-                                                 0, comm);
-
-    Teuchos::ParameterList problemParamList;
-    problemParamList.set("nx",gNodesPerDim[0]);
-    problemParamList.set("ny",gNodesPerDim[1]);
-    problemParamList.set("nz",gNodesPerDim[2]);
-    // problemParamList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    Galeri::Xpetra::Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> PoissonOnCube(problemParamList,
-                                                                                           map);
-    RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
-
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
-    }
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // build coordinates on rank 0
-    global_size_t numGlobalElements = numPoints;
-    size_t numLocalElements = (comm->getRank() == 0 ? numPoints : 0);
-    RCP<const Map> exportMap = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > source_coordinates
-      = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(exportMap, 3);
-    if (comm->getRank() == 0) {
-      for(LO k = 0; k < gNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < gNodesPerDim[1]; ++j) {
-          for(LO i = 0; i < gNodesPerDim[0]; ++i) {
-            source_coordinates->getDataNonConst(0)[k*ny*nx+j*nx+i]
-              = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0]-1);
-            source_coordinates->getDataNonConst(1)[k*ny*nx+j*nx+i]
-              = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1]-1);
-            source_coordinates->getDataNonConst(2)[k*ny*nx+j*nx+i]
-              = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2]-1);
-          }
+  MueLu::GeneralGeometricPFactory<SC, LO, GO, Node> ggPFact;
+  GeneralGeometricPFactoryTester<SC, LO, GO, Node> factTester;
+
+  LO numDimension = 3;
+  Array<Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coord(9);
+  for (int node = 0; node < 9; ++node) {
+    coord[node].resize(3);
+  }
+  coord[0][0] = 1.1;
+  coord[0][1] = 0.3;
+  coord[0][2] = 0.8;
+  coord[1][0] = 0.0;
+  coord[1][1] = 0.0;
+  coord[1][2] = 0.0;
+  coord[2][0] = 1.0;
+  coord[2][1] = 0.0;
+  coord[2][2] = 0.0;
+  coord[3][0] = 0.0;
+  coord[3][1] = 1.0;
+  coord[3][2] = 0.0;
+  coord[4][0] = 1.0;
+  coord[4][1] = 1.0;
+  coord[4][2] = 0.0;
+  coord[5][0] = 0.0;
+  coord[5][1] = 0.0;
+  coord[5][2] = 1.0;
+  coord[6][0] = 1.0;
+  coord[6][1] = 0.0;
+  coord[6][2] = 1.0;
+  coord[7][0] = 0.0;
+  coord[7][1] = 1.0;
+  coord[7][2] = 1.0;
+  coord[8][0] = 1.0;
+  coord[8][1] = 1.0;
+  coord[8][2] = 1.0;
+  std::vector<double> stencil(8);
+  factTester.TestComputeLinearInterpolationStencil(ggPFact, numDimension, coord, stencil);
+
+  typename Teuchos::ScalarTraits<Scalar>::magnitudeType x = 0.0, y = 0.0, z = 0.0;
+  for (LO i = 0; i < 8; ++i) {
+    x += stencil[i] * coord[i + 1][0];
+    y += stencil[i] * coord[i + 1][1];
+    z += stencil[i] * coord[i + 1][2];
+  }
+
+  TEST_EQUALITY((std::abs(x - coord[0][0]) < 1e-5) && (std::abs(y - coord[0][1]) < 1e-5) && (std::abs(z - coord[0][2]) < 1e-5), true);
+
+}  // LinearExtrapolation
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, PoissonOnCubeLinear, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels = 2;
+  // LocalOrdinal its=10;
+  GO nx        = 4;
+  GO ny        = 4;
+  GO nz        = 4;
+  GO numPoints = nx * ny * nz;
+  Array<GO> gNodesPerDim(3);
+  gNodesPerDim[0] = 4;
+  gNodesPerDim[1] = 4;
+  gNodesPerDim[2] = 4;
+  Array<LO> lNodesPerDim(3);
+  if (comm->getSize() == 1) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+  } else if (comm->getSize() == 4) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz / 4;
+  }
+
+  const RCP<const Map> map = MapFactory::Build(lib,
+                                               gNodesPerDim[0] * gNodesPerDim[1] * gNodesPerDim[2],
+                                               lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2],
+                                               0, comm);
+
+  Teuchos::ParameterList problemParamList;
+  problemParamList.set("nx", gNodesPerDim[0]);
+  problemParamList.set("ny", gNodesPerDim[1]);
+  problemParamList.set("nz", gNodesPerDim[2]);
+  // problemParamList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  Galeri::Xpetra::Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> PoissonOnCube(problemParamList,
+                                                                                              map);
+  RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
+
+  // build coordinates on rank 0
+  global_size_t numGlobalElements                                                                                 = numPoints;
+  size_t numLocalElements                                                                                         = (comm->getRank() == 0 ? numPoints : 0);
+  RCP<const Map> exportMap                                                                                        = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > source_coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(exportMap, 3);
+  if (comm->getRank() == 0) {
+    for (LO k = 0; k < gNodesPerDim[2]; ++k) {
+      for (LO j = 0; j < gNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < gNodesPerDim[0]; ++i) {
+          source_coordinates->getDataNonConst(0)[k * ny * nx + j * nx + i] = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
+          source_coordinates->getDataNonConst(1)[k * ny * nx + j * nx + i] = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
+          source_coordinates->getDataNonConst(2)[k * ny * nx + j * nx + i] = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
         }
       }
     }
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map,3);
-    RCP<Xpetra::Export<LO,GO,Node> > coords_exporter = Xpetra::ExportFactory<LO,GO,Node>::Build(exportMap, map);
-    coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new GeneralGeometricPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);  // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    // Extract the prolongator operator
-    RCP<Level> lvl1 = H->GetLevel(1);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    // Construct vectors to check that a linear vector remains linear after projection
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getRangeMap(),1);
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getDomainMap(),1);
-    {
-      ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-	for(LO i = 0; i < 3; ++i) {
-	  for(LO j = 0; j < 3; ++j) {
-	    coarse_data[3*i + j] = 2.0*i + j;
-	  }
-	}
+  }
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, 3);
+  RCP<Xpetra::Export<LO, GO, Node> > coords_exporter                                                       = Xpetra::ExportFactory<LO, GO, Node>::Build(exportMap, map);
+  coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new GeneralGeometricPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                       // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  // Extract the prolongator operator
+  RCP<Level> lvl1                          = H->GetLevel(1);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> PCrs                      = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  // Construct vectors to check that a linear vector remains linear after projection
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector0 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getRangeMap(), 1);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector1 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getDomainMap(), 1);
+  {
+    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      for (LO i = 0; i < 3; ++i) {
+        for (LO j = 0; j < 3; ++j) {
+          coarse_data[3 * i + j] = 2.0 * i + j;
+        }
       }
     }
-    PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                Teuchos::ScalarTraits<SC>::zero());
-
-    ArrayRCP<const SC> fine_data   = vector0->getData(0);
-    bool is_linear = true, is_injected = true;
-    Array<LO> fine_inds(9);
-    Array<LO> coarse_inds(9);
-    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-    if(comm->getRank() == 0) {
-      fine_inds[0] = 0;
-      fine_inds[1] = 2;
-      fine_inds[2] = 3;
-      fine_inds[3] = 8;
-      fine_inds[4] = 10;
-      fine_inds[5] = 11;
-      fine_inds[6] = 12;
-      fine_inds[7] = 14;
-      fine_inds[8] = 15;
-      for(LO i = 0; i < 9; ++i) {
-        if(std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) { is_injected = false; }
+  }
+  PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+              Teuchos::ScalarTraits<SC>::zero());
+
+  ArrayRCP<const SC> fine_data = vector0->getData(0);
+  bool is_linear = true, is_injected = true;
+  Array<LO> fine_inds(9);
+  Array<LO> coarse_inds(9);
+  ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+  if (comm->getRank() == 0) {
+    fine_inds[0] = 0;
+    fine_inds[1] = 2;
+    fine_inds[2] = 3;
+    fine_inds[3] = 8;
+    fine_inds[4] = 10;
+    fine_inds[5] = 11;
+    fine_inds[6] = 12;
+    fine_inds[7] = 14;
+    fine_inds[8] = 15;
+    for (LO i = 0; i < 9; ++i) {
+      if (std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) {
+        is_injected = false;
       }
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = std::abs(  coarse_data[0]
-                                   + coarse_data[1]
-                                   + coarse_data[3]
-                                   + coarse_data[4]) / 4.0;
-      if(std::abs(fine_data[5] - linear_avg) > 1.0e-10) { is_linear = false; }
     }
+    typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = std::abs(coarse_data[0] + coarse_data[1] + coarse_data[3] + coarse_data[4]) / 4.0;
+    if (std::abs(fine_data[5] - linear_avg) > 1.0e-10) {
+      is_linear = false;
+    }
+  }
 
+  TEST_EQUALITY(Op != Teuchos::null, true);
+  TEST_EQUALITY(H != Teuchos::null, true);
+  TEST_EQUALITY(nullSpace != Teuchos::null, true);
+  TEST_EQUALITY(is_injected, true);
+  TEST_EQUALITY(is_linear, true);
 
-    TEST_EQUALITY(Op != Teuchos::null, true);
-    TEST_EQUALITY(H  != Teuchos::null, true);
-    TEST_EQUALITY(nullSpace != Teuchos::null, true);
-    TEST_EQUALITY(is_injected, true);
-    TEST_EQUALITY(is_linear, true);
+}  // PoissonOnCubeLinear
 
-  } // PoissonOnCubeLinear
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, PoissonOnCubeConstant, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels = 2;
+  // LocalOrdinal its=10;
+  GO nx        = 4;
+  GO ny        = 4;
+  GO nz        = 4;
+  GO numPoints = nx * ny * nz;
+  Array<GO> gNodesPerDim(3);
+  gNodesPerDim[0] = 4;
+  gNodesPerDim[1] = 4;
+  gNodesPerDim[2] = 4;
+  Array<LO> lNodesPerDim(3);
+  if (comm->getSize() == 1) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+  } else if (comm->getSize() == 4) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz / 4;
+  }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, PoissonOnCubeConstant, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 2;
-    // LocalOrdinal its=10;
-    GO nx = 4;
-    GO ny = 4;
-    GO nz = 4;
-    GO numPoints = nx*ny*nz;
-    Array<GO> gNodesPerDim(3);
-    gNodesPerDim[0] = 4;
-    gNodesPerDim[1] = 4;
-    gNodesPerDim[2] = 4;
-    Array<LO> lNodesPerDim(3);
-    if(comm->getSize() == 1) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-    } else if(comm->getSize() == 4) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz/4;
-    }
+  const RCP<const Map> map = MapFactory::Build(lib,
+                                               gNodesPerDim[0] * gNodesPerDim[1] * gNodesPerDim[2],
+                                               lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2],
+                                               0, comm);
+
+  Teuchos::ParameterList problemParamList;
+  problemParamList.set("nx", gNodesPerDim[0]);
+  problemParamList.set("ny", gNodesPerDim[1]);
+  problemParamList.set("nz", gNodesPerDim[2]);
+  // problemParamList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  Galeri::Xpetra::Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> PoissonOnCube(problemParamList,
+                                                                                              map);
+  RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
 
-    const RCP<const Map> map = MapFactory::Build(lib,
-                                                 gNodesPerDim[0]*gNodesPerDim[1]*gNodesPerDim[2],
-                                                 lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2],
-                                                 0, comm);
-
-    Teuchos::ParameterList problemParamList;
-    problemParamList.set("nx",gNodesPerDim[0]);
-    problemParamList.set("ny",gNodesPerDim[1]);
-    problemParamList.set("nz",gNodesPerDim[2]);
-    // problemParamList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    Galeri::Xpetra::Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> PoissonOnCube(problemParamList,
-                                                                                           map);
-    RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
-
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
+  // build coordinates on rank 0
+  global_size_t numGlobalElements                                                                                 = numPoints;
+  size_t numLocalElements                                                                                         = (comm->getRank() == 0 ? numPoints : 0);
+  RCP<const Map> exportMap                                                                                        = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > source_coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(exportMap, 3);
+  if (comm->getRank() == 0) {
+    for (LO k = 0; k < gNodesPerDim[2]; ++k) {
+      for (LO j = 0; j < gNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < gNodesPerDim[0]; ++i) {
+          source_coordinates->getDataNonConst(0)[k * ny * nx + j * nx + i] = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
+          source_coordinates->getDataNonConst(1)[k * ny * nx + j * nx + i] = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
+          source_coordinates->getDataNonConst(2)[k * ny * nx + j * nx + i] = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
+        }
+      }
     }
-
-    // build coordinates on rank 0
-    global_size_t numGlobalElements = numPoints;
-    size_t numLocalElements = (comm->getRank() == 0 ? numPoints : 0);
-    RCP<const Map> exportMap = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > source_coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(exportMap, 3);
+  }
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, 3);
+  RCP<Xpetra::Export<LO, GO, Node> > coords_exporter                                                       = Xpetra::ExportFactory<LO, GO, Node>::Build(exportMap, map);
+  coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new GeneralGeometricPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("order", Teuchos::ParameterEntry(0));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                       // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  // Extract the prolongator operator
+  RCP<Level> lvl1                          = H->GetLevel(1);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> PCrs                      = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  // Construct vectors to check that a linear vector remains linear after projection
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector0 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getRangeMap(), 1);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector1 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getDomainMap(), 1);
+  {
+    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
     if (comm->getRank() == 0) {
-      for(LO k = 0; k < gNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < gNodesPerDim[1]; ++j) {
-          for(LO i = 0; i < gNodesPerDim[0]; ++i) {
-            source_coordinates->getDataNonConst(0)[k*ny*nx+j*nx+i]
-              = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0]-1);
-            source_coordinates->getDataNonConst(1)[k*ny*nx+j*nx+i]
-              = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1]-1);
-            source_coordinates->getDataNonConst(2)[k*ny*nx+j*nx+i]
-              = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2]-1);
-          }
+      for (LO i = 0; i < 3; ++i) {
+        for (LO j = 0; j < 3; ++j) {
+          coarse_data[3 * i + j] = 2.0 * i + j;
         }
       }
     }
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates
-      = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map,3);
-    RCP<Xpetra::Export<LO,GO,Node> > coords_exporter
-      = Xpetra::ExportFactory<LO,GO,Node>::Build(exportMap, map);
-    coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new GeneralGeometricPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("order", Teuchos::ParameterEntry(0));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);  // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    // Extract the prolongator operator
-    RCP<Level> lvl1 = H->GetLevel(1);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    // Construct vectors to check that a linear vector remains linear after projection
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getRangeMap(),1);
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getDomainMap(),1);
-    {
-      ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-	for(LO i = 0; i < 3; ++i) {
-	  for(LO j = 0; j < 3; ++j) {
-	    coarse_data[3*i + j] = 2.0*i + j;
-	  }
-	}
+  }
+
+  PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+              Teuchos::ScalarTraits<SC>::zero());
+
+  ArrayRCP<const SC> fine_data = vector0->getData(0);
+  bool is_constant = true, is_injected = true;
+  Array<LO> fine_inds(9);
+  Array<LO> coarse_inds(9);
+  ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+  if (comm->getRank() == 0) {
+    fine_inds[0] = 0;
+    fine_inds[1] = 2;
+    fine_inds[2] = 3;
+    fine_inds[3] = 8;
+    fine_inds[4] = 10;
+    fine_inds[5] = 11;
+    fine_inds[6] = 12;
+    fine_inds[7] = 14;
+    fine_inds[8] = 15;
+    for (LO i = 0; i < 9; ++i) {
+      if (std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) {
+        is_injected = false;
       }
     }
+    if ((std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
+        (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
+        (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) {
+      is_constant = false;
+    }
+  }
 
-    PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                Teuchos::ScalarTraits<SC>::zero());
+  TEST_EQUALITY(Op != Teuchos::null, true);
+  TEST_EQUALITY(H != Teuchos::null, true);
+  TEST_EQUALITY(nullSpace != Teuchos::null, true);
+  TEST_EQUALITY(is_injected, true);
+  TEST_EQUALITY(is_constant, true);
 
-    ArrayRCP<const SC> fine_data   = vector0->getData(0);
-    bool is_constant = true, is_injected = true;
-    Array<LO> fine_inds(9);
-    Array<LO> coarse_inds(9);
-    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-    if(comm->getRank() == 0) {
-      fine_inds[0] = 0;
-      fine_inds[1] = 2;
-      fine_inds[2] = 3;
-      fine_inds[3] = 8;
-      fine_inds[4] = 10;
-      fine_inds[5] = 11;
-      fine_inds[6] = 12;
-      fine_inds[7] = 14;
-      fine_inds[8] = 15;
-      for(LO i = 0; i < 9; ++i) {
-        if(std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) { is_injected = false; }
-      }
-      if( (std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
-          (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
-          (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) { is_constant = false; }
-    }
+}  // PoissonOnCubeConstant
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LocalLexicographicLinear, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  LO numDimensions = 3;
+
+  RCP<Matrix> Op;
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates;
+  RCP<Map> map;
+  Array<GO> gNodesPerDim(3);
+  Array<LO> lNodesPerDim(3);
+  GGGetProblemData<SC, LO, GO, NO>(comm, lib, numDimensions, "Local Lexicographic", Op,
+                                   coordinates, map, gNodesPerDim, lNodesPerDim);
+
+  TEST_EQUALITY(Op != Teuchos::null, true);
+
+  Array<Array<GO> > indsBounds(4);
+  Array<GO> minGID(4);
+  indsBounds[0].resize(4);
+  if (comm->getSize() == 1) {
+    indsBounds[0][0] = 0;
+    indsBounds[0][1] = 8;
+    indsBounds[0][2] = 0;
+    indsBounds[0][3] = 8;
+  } else {
+    indsBounds[0][0] = 0;
+    indsBounds[0][1] = 4;
+    indsBounds[0][2] = 0;
+    indsBounds[0][3] = 4;
+  }
+  minGID[0] = 0;
+  indsBounds[1].resize(4);
+  indsBounds[1][0] = 5;
+  indsBounds[1][1] = 8;
+  indsBounds[1][2] = 0;
+  indsBounds[1][3] = 4;
+  minGID[1]        = 225;
+  indsBounds[2].resize(4);
+  indsBounds[2][0] = 0;
+  indsBounds[2][1] = 4;
+  indsBounds[2][2] = 5;
+  indsBounds[2][3] = 8;
+  minGID[2]        = 405;
+  indsBounds[3].resize(4);
+  indsBounds[3][0] = 5;
+  indsBounds[3][1] = 8;
+  indsBounds[3][2] = 5;
+  indsBounds[3][3] = 8;
+  minGID[3]        = 585;
+
+  Array<GO> meshData(comm->getSize() * 10);
+  for (int rank = 0; rank < comm->getSize(); ++rank) {
+    meshData[rank * 10 + 0] = rank;
+    meshData[rank * 10 + 1] = rank;
+    meshData[rank * 10 + 2] = 0;
+    meshData[rank * 10 + 3] = 0;
+    meshData[rank * 10 + 4] = gNodesPerDim[0] - 1;
+    meshData[rank * 10 + 5] = indsBounds[rank][0];
+    meshData[rank * 10 + 6] = indsBounds[rank][1];
+    meshData[rank * 10 + 7] = indsBounds[rank][2];
+    meshData[rank * 10 + 8] = indsBounds[rank][3];
+    meshData[rank * 10 + 9] = minGID[rank];
+  }
 
-    TEST_EQUALITY(Op != Teuchos::null, true);
-    TEST_EQUALITY(H  != Teuchos::null, true);
-    TEST_EQUALITY(nullSpace != Teuchos::null, true);
-    TEST_EQUALITY(is_injected, true);
-    TEST_EQUALITY(is_constant, true);
+  // generate problem
+  LocalOrdinal maxLevels = 3;
 
-  } // PoissonOnCubeConstant
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LocalLexicographicLinear, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new GeneralGeometricPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
+  Pfact->SetParameter("meshLayout", Teuchos::ParameterEntry(std::string("Local Lexicographic")));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                       // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("meshData", meshData);          // set GeneralGeometricPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  // Extract the prolongator operator
+  RCP<Level> lvl1                           = H->GetLevel(1);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P1 = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> P1Crs                      = rcp_dynamic_cast<CrsMatrixWrap>(P1)->getCrsMatrix();
+
+  // Construct vectors to check that a linear vector remains linear after projection
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector0 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(P1Crs->getRangeMap(), 1);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector1 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(P1Crs->getDomainMap(), 1);
+  std::vector<LO> coarse_inds(8);
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    LO numDimensions = 3;
-
-    RCP<Matrix> Op;
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates;
-    RCP<Map> map;
-    Array<GO> gNodesPerDim(3);
-    Array<LO> lNodesPerDim(3);
-    GGGetProblemData<SC,LO,GO,NO>(comm, lib, numDimensions, "Local Lexicographic", Op,
-                                  coordinates, map, gNodesPerDim, lNodesPerDim);
-
-    TEST_EQUALITY(Op != Teuchos::null, true);
-
-    Array<Array<GO> > indsBounds(4);
-    Array<GO> minGID(4);
-    indsBounds[0].resize(4);
-    if(comm->getSize() == 1) {
-      indsBounds[0][0] = 0;
-      indsBounds[0][1] = 8;
-      indsBounds[0][2] = 0;
-      indsBounds[0][3] = 8;
-    } else {
-      indsBounds[0][0] = 0;
-      indsBounds[0][1] = 4;
-      indsBounds[0][2] = 0;
-      indsBounds[0][3] = 4;
-    }
-    minGID[0] = 0;
-    indsBounds[1].resize(4);
-    indsBounds[1][0] = 5;
-    indsBounds[1][1] = 8;
-    indsBounds[1][2] = 0;
-    indsBounds[1][3] = 4;
-    minGID[1] = 225;
-    indsBounds[2].resize(4);
-    indsBounds[2][0] = 0;
-    indsBounds[2][1] = 4;
-    indsBounds[2][2] = 5;
-    indsBounds[2][3] = 8;
-    minGID[2] = 405;
-    indsBounds[3].resize(4);
-    indsBounds[3][0] = 5;
-    indsBounds[3][1] = 8;
-    indsBounds[3][2] = 5;
-    indsBounds[3][3] = 8;
-    minGID[3] = 585;
-
-    Array<GO> meshData(comm->getSize()*10);
-    for(int rank = 0 ; rank < comm->getSize(); ++rank) {
-      meshData[rank*10 + 0] = rank;
-      meshData[rank*10 + 1] = rank;
-      meshData[rank*10 + 2] = 0;
-      meshData[rank*10 + 3] = 0;
-      meshData[rank*10 + 4] = gNodesPerDim[0] - 1;
-      meshData[rank*10 + 5] = indsBounds[rank][0];
-      meshData[rank*10 + 6] = indsBounds[rank][1];
-      meshData[rank*10 + 7] = indsBounds[rank][2];
-      meshData[rank*10 + 8] = indsBounds[rank][3];
-      meshData[rank*10 + 9] = minGID[rank];
+    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+    if (comm->getSize() == 1) {
+      coarse_inds[0] = 0;
+      coarse_inds[1] = 1;
+      coarse_inds[2] = 5;
+      coarse_inds[3] = 6;
+      coarse_inds[4] = 25;
+      coarse_inds[5] = 26;
+      coarse_inds[6] = 30;
+      coarse_inds[7] = 31;
+    } else if (comm->getSize() == 4 && comm->getRank() == 0) {
+      coarse_inds[0] = 0;
+      coarse_inds[1] = 1;
+      coarse_inds[2] = 5;
+      coarse_inds[3] = 6;
+      coarse_inds[4] = 15;
+      coarse_inds[5] = 16;
+      coarse_inds[6] = 20;
+      coarse_inds[7] = 21;
     }
+    coarse_data[coarse_inds[0]] = 5;
+    coarse_data[coarse_inds[1]] = 1;
+    coarse_data[coarse_inds[2]] = 7;
+    coarse_data[coarse_inds[3]] = 8;
+    coarse_data[coarse_inds[4]] = 0;
+    coarse_data[coarse_inds[5]] = 4;
+    coarse_data[coarse_inds[6]] = 0;
+    coarse_data[coarse_inds[7]] = 9;
+  }
 
-    // generate problem
-    LocalOrdinal maxLevels = 3;
+  P1Crs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+               Teuchos::ScalarTraits<SC>::zero());
 
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
+  ArrayRCP<const SC> fine_data = vector0->getData(0);
+  Array<LO> fine_inds(8);
+  bool is_linear_lvl1 = true, is_injected_lvl1 = true;
+  if (comm->getRank() == 0) {
+    if (comm->getSize() == 1) {
+      fine_inds[0] = 0;
+      fine_inds[1] = 2;
+      fine_inds[2] = 18;
+      fine_inds[3] = 20;
+      fine_inds[4] = 162;
+      fine_inds[5] = 164;
+      fine_inds[6] = 180;
+      fine_inds[7] = 182;
+    } else if (comm->getSize() == 4) {
+      fine_inds[0] = 0;
+      fine_inds[1] = 2;
+      fine_inds[2] = 18;
+      fine_inds[3] = 20;
+      fine_inds[4] = 90;
+      fine_inds[5] = 92;
+      fine_inds[6] = 108;
+      fine_inds[7] = 110;
     }
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new GeneralGeometricPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
-    Pfact->SetParameter("meshLayout", Teuchos::ParameterEntry(std::string("Local Lexicographic")));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                        // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);         // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);    // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("meshData", meshData);          // set GeneralGeometricPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    // Extract the prolongator operator
-    RCP<Level> lvl1 = H->GetLevel(1);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P1 = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> P1Crs = rcp_dynamic_cast<CrsMatrixWrap>(P1)->getCrsMatrix();
-
-    // Construct vectors to check that a linear vector remains linear after projection
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(P1Crs->getRangeMap(),1);
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(P1Crs->getDomainMap(),1);
-    std::vector<LO> coarse_inds(8);
-    {
-      ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-      if(comm->getSize() == 1) {
-	coarse_inds[0] =  0;
-	coarse_inds[1] =  1;
-	coarse_inds[2] =  5;
-	coarse_inds[3] =  6;
-	coarse_inds[4] = 25;
-	coarse_inds[5] = 26;
-	coarse_inds[6] = 30;
-	coarse_inds[7] = 31;
-      } else if(comm->getSize() == 4 && comm->getRank() == 0) {
-	coarse_inds[0] =  0;
-	coarse_inds[1] =  1;
-	coarse_inds[2] =  5;
-	coarse_inds[3] =  6;
-	coarse_inds[4] = 15;
-	coarse_inds[5] = 16;
-	coarse_inds[6] = 20;
-	coarse_inds[7] = 21;
+    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+    for (LO i = 0; i < 8; ++i) {
+      if (std::abs(fine_data[fine_inds[i]] - coarse_data[coarse_inds[i]]) > 1.0e-10) {
+        is_injected_lvl1 = false;
       }
-      coarse_data[coarse_inds[0]] = 5;
-      coarse_data[coarse_inds[1]] = 1;
-      coarse_data[coarse_inds[2]] = 7;
-      coarse_data[coarse_inds[3]] = 8;
-      coarse_data[coarse_inds[4]] = 0;
-      coarse_data[coarse_inds[5]] = 4;
-      coarse_data[coarse_inds[6]] = 0;
-      coarse_data[coarse_inds[7]] = 9;
     }
-
-    P1Crs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                Teuchos::ScalarTraits<SC>::zero());
-
-    ArrayRCP<const SC> fine_data   = vector0->getData(0);
-    Array<LO> fine_inds(8);
-    bool is_linear_lvl1 = true, is_injected_lvl1 = true;
-    if(comm->getRank() == 0) {
-      if(comm->getSize() == 1) {
-        fine_inds[0] =   0;
-        fine_inds[1] =   2;
-        fine_inds[2] =  18;
-        fine_inds[3] =  20;
-        fine_inds[4] = 162;
-        fine_inds[5] = 164;
-        fine_inds[6] = 180;
-        fine_inds[7] = 182;
-      } else if(comm->getSize() == 4) {
-        fine_inds[0] =   0;
-        fine_inds[1] =   2;
-        fine_inds[2] =  18;
-        fine_inds[3] =  20;
-        fine_inds[4] =  90;
-        fine_inds[5] =  92;
-        fine_inds[6] = 108;
-        fine_inds[7] = 110;
-      }
-      ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-      for(LO i = 0; i < 8; ++i) {
-        if(std::abs(fine_data[fine_inds[i]] - coarse_data[coarse_inds[i]]) > 1.0e-10) {
-          is_injected_lvl1 = false;
-        }
-      }
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = 0.0;
-      for(auto ind : coarse_inds) {linear_avg += std::abs(coarse_data[ind]) / 8.0;}
-      LO fine_iref = 0;
-      if(comm->getSize() == 1) {fine_iref = 91;} else if(comm->getSize() == 4) {fine_iref = 55;}
-      if(std::abs(fine_data[fine_iref] - linear_avg) > 1.0e-10) { is_linear_lvl1 = false; }
+    typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = 0.0;
+    for (auto ind : coarse_inds) {
+      linear_avg += std::abs(coarse_data[ind]) / 8.0;
     }
-
-    // Extract the prolongator operator
-    RCP<Level> lvl2 = H->GetLevel(2);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P2 = lvl2->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> P2Crs = rcp_dynamic_cast<CrsMatrixWrap>(P2)->getCrsMatrix();
-
-    // Construct vectors to check that a linear vector remains linear after projection
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector2
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(P2Crs->getRangeMap(),1);
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector3
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(P2Crs->getDomainMap(),1);
-    if(comm->getSize() == 1) {
-      coarse_inds[0] =  0;
-      coarse_inds[1] =  1;
-      coarse_inds[2] =  3;
-      coarse_inds[3] =  4;
-      coarse_inds[4] =  9;
-      coarse_inds[5] = 10;
-      coarse_inds[6] = 12;
-      coarse_inds[7] = 13;
-    } else if(comm->getSize() == 4 && comm->getRank() == 0) {
-      coarse_inds[0] =  0;
-      coarse_inds[1] =  1;
-      coarse_inds[2] =  3;
-      coarse_inds[3] =  4;
-      coarse_inds[4] =  6;
-      coarse_inds[5] =  7;
-      coarse_inds[6] =  9;
-      coarse_inds[7] = 10;
+    LO fine_iref = 0;
+    if (comm->getSize() == 1) {
+      fine_iref = 91;
+    } else if (comm->getSize() == 4) {
+      fine_iref = 55;
     }
-    {
-      ArrayRCP<SC> coarse_data = vector3->getDataNonConst(0);
-      coarse_data[coarse_inds[0]] = 5;
-      coarse_data[coarse_inds[1]] = 1;
-      coarse_data[coarse_inds[2]] = 7;
-      coarse_data[coarse_inds[3]] = 8;
-      coarse_data[coarse_inds[4]] = 0;
-      coarse_data[coarse_inds[5]] = 4;
-      coarse_data[coarse_inds[6]] = 0;
-      coarse_data[coarse_inds[7]] = 9;
+    if (std::abs(fine_data[fine_iref] - linear_avg) > 1.0e-10) {
+      is_linear_lvl1 = false;
     }
+  }
 
-    P2Crs->apply(*vector3, *vector2, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                Teuchos::ScalarTraits<SC>::zero());
-
-    fine_data = vector2->getData(0);
-    bool is_linear_lvl2 = true, is_injected_lvl2 = true;
-    if(comm->getRank() == 0) {
-      if(comm->getSize() == 1) {
-        fine_inds[0] =  0;
-        fine_inds[1] =  2;
-        fine_inds[2] = 10;
-        fine_inds[3] = 12;
-        fine_inds[4] = 50;
-        fine_inds[5] = 52;
-        fine_inds[6] = 60;
-        fine_inds[7] = 62;
-      } else if(comm->getSize() == 4) {
-        fine_inds[0] =  0;
-        fine_inds[1] =  2;
-        fine_inds[2] = 10;
-        fine_inds[3] = 12;
-        fine_inds[4] = 30;
-        fine_inds[5] = 32;
-        fine_inds[6] = 40;
-        fine_inds[7] = 42;
-      }
-      ArrayRCP<SC> coarse_data = vector3->getDataNonConst(0);
-      for(LO i = 0; i < 8; ++i) {
-        if(std::abs(fine_data[fine_inds[i]] - coarse_data[coarse_inds[i]]) > 1.0e-10) {
-          is_injected_lvl2 = false;
-        }
+  // Extract the prolongator operator
+  RCP<Level> lvl2                           = H->GetLevel(2);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P2 = lvl2->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> P2Crs                      = rcp_dynamic_cast<CrsMatrixWrap>(P2)->getCrsMatrix();
+
+  // Construct vectors to check that a linear vector remains linear after projection
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector2 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(P2Crs->getRangeMap(), 1);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector3 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(P2Crs->getDomainMap(), 1);
+  if (comm->getSize() == 1) {
+    coarse_inds[0] = 0;
+    coarse_inds[1] = 1;
+    coarse_inds[2] = 3;
+    coarse_inds[3] = 4;
+    coarse_inds[4] = 9;
+    coarse_inds[5] = 10;
+    coarse_inds[6] = 12;
+    coarse_inds[7] = 13;
+  } else if (comm->getSize() == 4 && comm->getRank() == 0) {
+    coarse_inds[0] = 0;
+    coarse_inds[1] = 1;
+    coarse_inds[2] = 3;
+    coarse_inds[3] = 4;
+    coarse_inds[4] = 6;
+    coarse_inds[5] = 7;
+    coarse_inds[6] = 9;
+    coarse_inds[7] = 10;
+  }
+  {
+    ArrayRCP<SC> coarse_data    = vector3->getDataNonConst(0);
+    coarse_data[coarse_inds[0]] = 5;
+    coarse_data[coarse_inds[1]] = 1;
+    coarse_data[coarse_inds[2]] = 7;
+    coarse_data[coarse_inds[3]] = 8;
+    coarse_data[coarse_inds[4]] = 0;
+    coarse_data[coarse_inds[5]] = 4;
+    coarse_data[coarse_inds[6]] = 0;
+    coarse_data[coarse_inds[7]] = 9;
+  }
+
+  P2Crs->apply(*vector3, *vector2, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+               Teuchos::ScalarTraits<SC>::zero());
+
+  fine_data           = vector2->getData(0);
+  bool is_linear_lvl2 = true, is_injected_lvl2 = true;
+  if (comm->getRank() == 0) {
+    if (comm->getSize() == 1) {
+      fine_inds[0] = 0;
+      fine_inds[1] = 2;
+      fine_inds[2] = 10;
+      fine_inds[3] = 12;
+      fine_inds[4] = 50;
+      fine_inds[5] = 52;
+      fine_inds[6] = 60;
+      fine_inds[7] = 62;
+    } else if (comm->getSize() == 4) {
+      fine_inds[0] = 0;
+      fine_inds[1] = 2;
+      fine_inds[2] = 10;
+      fine_inds[3] = 12;
+      fine_inds[4] = 30;
+      fine_inds[5] = 32;
+      fine_inds[6] = 40;
+      fine_inds[7] = 42;
+    }
+    ArrayRCP<SC> coarse_data = vector3->getDataNonConst(0);
+    for (LO i = 0; i < 8; ++i) {
+      if (std::abs(fine_data[fine_inds[i]] - coarse_data[coarse_inds[i]]) > 1.0e-10) {
+        is_injected_lvl2 = false;
       }
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = 0.0;
-      for(auto ind : coarse_inds) {linear_avg += std::abs(coarse_data[ind]) / 8.0;}
-      LO fine_iref = 0;
-      if(comm->getSize() == 1) {fine_iref = 31;} else if(comm->getSize() == 4) {fine_iref = 21;}
-      if(std::abs(fine_data[fine_iref] - linear_avg) > 1.0e-10) { is_linear_lvl2 = false; }
     }
+    typename Teuchos::ScalarTraits<Scalar>::magnitudeType linear_avg = 0.0;
+    for (auto ind : coarse_inds) {
+      linear_avg += std::abs(coarse_data[ind]) / 8.0;
+    }
+    LO fine_iref = 0;
+    if (comm->getSize() == 1) {
+      fine_iref = 31;
+    } else if (comm->getSize() == 4) {
+      fine_iref = 21;
+    }
+    if (std::abs(fine_data[fine_iref] - linear_avg) > 1.0e-10) {
+      is_linear_lvl2 = false;
+    }
+  }
 
+  TEST_EQUALITY(Op != Teuchos::null, true);
+  TEST_EQUALITY(H != Teuchos::null, true);
+  TEST_EQUALITY(nullSpace != Teuchos::null, true);
+  TEST_EQUALITY(is_injected_lvl1, true);
+  TEST_EQUALITY(is_linear_lvl1, true);
+  TEST_EQUALITY(is_injected_lvl2, true);
+  TEST_EQUALITY(is_linear_lvl2, true);
 
-    TEST_EQUALITY(Op != Teuchos::null, true);
-    TEST_EQUALITY(H  != Teuchos::null, true);
-    TEST_EQUALITY(nullSpace != Teuchos::null, true);
-    TEST_EQUALITY(is_injected_lvl1, true);
-    TEST_EQUALITY(is_linear_lvl1, true);
-    TEST_EQUALITY(is_injected_lvl2, true);
-    TEST_EQUALITY(is_linear_lvl2, true);
-
-  } // End LocalLexicographicLinear
+}  // End LocalLexicographicLinear
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LocalLexicographicConstant, Scalar,
-      LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 2;
-    // LocalOrdinal its=10;
-    GO nx = 4;
-    GO ny = 4;
-    GO nz = 4;
-    GO numPoints = nx*ny*nz;
-    Array<GO> gNodesPerDim(3);
-    gNodesPerDim[0] = 4;
-    gNodesPerDim[1] = 4;
-    gNodesPerDim[2] = 4;
-    Array<GO> meshData(comm->getSize()*10);
-    Array<LO> lNodesPerDim(3);
-    if(comm->getSize() == 1) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-
-      meshData[0] = comm->getRank(); // my Parent rank
-      meshData[1] = comm->getRank(); // my rank
-      meshData[2] = 0; // block ID
-      meshData[3] = 0;
-      meshData[4] = nx - 1;
-      meshData[5] = 0;
-      meshData[6] = ny - 1;
-      meshData[7] = 0;
-      meshData[8] = nz - 1;
-      meshData[9] = 0;
-    } else if(comm->getSize() == 4) {
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz/4;
-
-      for(int rank = 0; rank < comm->getSize(); ++rank) {
-        meshData[rank*10 + 0] = rank;       // my Parent rank
-        meshData[rank*10 + 1] = rank;       // my rank
-        meshData[rank*10 + 2] = 0;          // block ID
-        meshData[rank*10 + 3] = 0;          // imin
-        meshData[rank*10 + 4] = nx - 1;     // imax
-        meshData[rank*10 + 5] = 0;          // jmin
-        meshData[rank*10 + 6] = ny - 1;     // jmax
-        meshData[rank*10 + 7] = rank;       // kmin
-        meshData[rank*10 + 8] = rank;       // kmax
-        meshData[rank*10 + 9] = rank*ny*nx; // minGID on rank
-      }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GeneralGeometricPFactory, LocalLexicographicConstant, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels = 2;
+  // LocalOrdinal its=10;
+  GO nx        = 4;
+  GO ny        = 4;
+  GO nz        = 4;
+  GO numPoints = nx * ny * nz;
+  Array<GO> gNodesPerDim(3);
+  gNodesPerDim[0] = 4;
+  gNodesPerDim[1] = 4;
+  gNodesPerDim[2] = 4;
+  Array<GO> meshData(comm->getSize() * 10);
+  Array<LO> lNodesPerDim(3);
+  if (comm->getSize() == 1) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+
+    meshData[0] = comm->getRank();  // my Parent rank
+    meshData[1] = comm->getRank();  // my rank
+    meshData[2] = 0;                // block ID
+    meshData[3] = 0;
+    meshData[4] = nx - 1;
+    meshData[5] = 0;
+    meshData[6] = ny - 1;
+    meshData[7] = 0;
+    meshData[8] = nz - 1;
+    meshData[9] = 0;
+  } else if (comm->getSize() == 4) {
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz / 4;
+
+    for (int rank = 0; rank < comm->getSize(); ++rank) {
+      meshData[rank * 10 + 0] = rank;            // my Parent rank
+      meshData[rank * 10 + 1] = rank;            // my rank
+      meshData[rank * 10 + 2] = 0;               // block ID
+      meshData[rank * 10 + 3] = 0;               // imin
+      meshData[rank * 10 + 4] = nx - 1;          // imax
+      meshData[rank * 10 + 5] = 0;               // jmin
+      meshData[rank * 10 + 6] = ny - 1;          // jmax
+      meshData[rank * 10 + 7] = rank;            // kmin
+      meshData[rank * 10 + 8] = rank;            // kmax
+      meshData[rank * 10 + 9] = rank * ny * nx;  // minGID on rank
     }
+  }
 
-    const RCP<const Map> map = MapFactory::Build(lib,
-                                                 gNodesPerDim[0]*gNodesPerDim[1]*gNodesPerDim[2],
-                                                 lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2],
-                                                 0,
-                                                 comm);
-
-    Teuchos::ParameterList problemParamList;
-    problemParamList.set("nx",gNodesPerDim[0]);
-    problemParamList.set("ny",gNodesPerDim[1]);
-    problemParamList.set("nz",gNodesPerDim[2]);
-    // problemParamList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    Galeri::Xpetra::Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> PoissonOnCube(problemParamList,
-                                                                                           map);
-    RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
-
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
-    }
+  const RCP<const Map> map = MapFactory::Build(lib,
+                                               gNodesPerDim[0] * gNodesPerDim[1] * gNodesPerDim[2],
+                                               lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2],
+                                               0,
+                                               comm);
+
+  Teuchos::ParameterList problemParamList;
+  problemParamList.set("nx", gNodesPerDim[0]);
+  problemParamList.set("ny", gNodesPerDim[1]);
+  problemParamList.set("nz", gNodesPerDim[2]);
+  // problemParamList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  Galeri::Xpetra::Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> PoissonOnCube(problemParamList,
+                                                                                              map);
+  RCP<Matrix> Op = PoissonOnCube.BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
 
-    // build coordinates on rank 0
-    global_size_t numGlobalElements = numPoints;
-    size_t numLocalElements = (comm->getRank() == 0 ? numPoints : 0);
-    RCP<const Map> exportMap = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > source_coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(exportMap, 3);
-    if (comm->getRank() == 0) {
-      for(LO k = 0; k < gNodesPerDim[2]; ++k) {
-        for(LO j = 0; j < gNodesPerDim[1]; ++j) {
-          for(LO i = 0; i < gNodesPerDim[0]; ++i) {
-            source_coordinates->getDataNonConst(0)[k*ny*nx+j*nx+i]
-              = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0]-1);
-            source_coordinates->getDataNonConst(1)[k*ny*nx+j*nx+i]
-              = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1]-1);
-            source_coordinates->getDataNonConst(2)[k*ny*nx+j*nx+i]
-              = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2]-1);
-          }
+  // build coordinates on rank 0
+  global_size_t numGlobalElements                                                                                 = numPoints;
+  size_t numLocalElements                                                                                         = (comm->getRank() == 0 ? numPoints : 0);
+  RCP<const Map> exportMap                                                                                        = MapFactory::Build(lib, numGlobalElements, numLocalElements, 0, comm);
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > source_coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(exportMap, 3);
+  if (comm->getRank() == 0) {
+    for (LO k = 0; k < gNodesPerDim[2]; ++k) {
+      for (LO j = 0; j < gNodesPerDim[1]; ++j) {
+        for (LO i = 0; i < gNodesPerDim[0]; ++i) {
+          source_coordinates->getDataNonConst(0)[k * ny * nx + j * nx + i] = i / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
+          source_coordinates->getDataNonConst(1)[k * ny * nx + j * nx + i] = j / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
+          source_coordinates->getDataNonConst(2)[k * ny * nx + j * nx + i] = k / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
         }
       }
     }
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates
-      = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map,3);
-    RCP<Xpetra::Export<LO,GO,Node> > coords_exporter
-      = Xpetra::ExportFactory<LO,GO,Node>::Build(exportMap, map);
-    coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new GeneralGeometricPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("order", Teuchos::ParameterEntry(0));
-    Pfact->SetParameter("meshLayout", Teuchos::ParameterEntry(std::string("Local Lexicographic")));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                        // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);         // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);    // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("meshData", meshData);          // set GeneralGeometricPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    // Extract the prolongator operator
-    RCP<Level> lvl1 = H->GetLevel(1);
-    RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    // Construct vectors to check that a linear vector remains linear after projection
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getRangeMap(),1);
-    RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
-      = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getDomainMap(),1);
-    {
-      ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-	for(LO i = 0; i < 3; ++i) {
-	  for(LO j = 0; j < 3; ++j) {
-	    coarse_data[3*i + j] = 2.0*i + j;
-	  }
-	}
+  }
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, 3);
+  RCP<Xpetra::Export<LO, GO, Node> > coords_exporter                                                       = Xpetra::ExportFactory<LO, GO, Node>::Build(exportMap, map);
+  coordinates->doExport(*source_coordinates, *coords_exporter, Xpetra::INSERT);
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new GeneralGeometricPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("order", Teuchos::ParameterEntry(0));
+  Pfact->SetParameter("meshLayout", Teuchos::ParameterEntry(std::string("Local Lexicographic")));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                       // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("meshData", meshData);          // set GeneralGeometricPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  // Extract the prolongator operator
+  RCP<Level> lvl1                          = H->GetLevel(1);
+  RCP<Xpetra::Matrix<SC, LO, GO, Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  RCP<CrsMatrix> PCrs                      = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  // Construct vectors to check that a linear vector remains linear after projection
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector0 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getRangeMap(), 1);
+  RCP<Xpetra::MultiVector<SC, LO, GO, NO> > vector1 = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(PCrs->getDomainMap(), 1);
+  {
+    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      for (LO i = 0; i < 3; ++i) {
+        for (LO j = 0; j < 3; ++j) {
+          coarse_data[3 * i + j] = 2.0 * i + j;
+        }
       }
     }
+  }
 
-    PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-                Teuchos::ScalarTraits<SC>::zero());
-
-    ArrayRCP<const SC> fine_data   = vector0->getData(0);
-    bool is_constant = true, is_injected = true;
-    Array<LO> fine_inds(9);
-    Array<LO> coarse_inds(9);
-    ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-    if(comm->getRank() == 0) {
-      fine_inds[0] = 0;
-      fine_inds[1] = 2;
-      fine_inds[2] = 3;
-      fine_inds[3] = 8;
-      fine_inds[4] = 10;
-      fine_inds[5] = 11;
-      fine_inds[6] = 12;
-      fine_inds[7] = 14;
-      fine_inds[8] = 15;
-      for(LO i = 0; i < 9; ++i) {
-        if(std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) { is_injected = false; }
+  PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+              Teuchos::ScalarTraits<SC>::zero());
+
+  ArrayRCP<const SC> fine_data = vector0->getData(0);
+  bool is_constant = true, is_injected = true;
+  Array<LO> fine_inds(9);
+  Array<LO> coarse_inds(9);
+  ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+  if (comm->getRank() == 0) {
+    fine_inds[0] = 0;
+    fine_inds[1] = 2;
+    fine_inds[2] = 3;
+    fine_inds[3] = 8;
+    fine_inds[4] = 10;
+    fine_inds[5] = 11;
+    fine_inds[6] = 12;
+    fine_inds[7] = 14;
+    fine_inds[8] = 15;
+    for (LO i = 0; i < 9; ++i) {
+      if (std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) {
+        is_injected = false;
       }
-      if( (std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
-          (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
-          (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) { is_constant = false; }
     }
+    if ((std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
+        (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
+        (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) {
+      is_constant = false;
+    }
+  }
 
+  TEST_EQUALITY(Op != Teuchos::null, true);
+  TEST_EQUALITY(H != Teuchos::null, true);
+  TEST_EQUALITY(nullSpace != Teuchos::null, true);
+  TEST_EQUALITY(is_injected, true);
+  TEST_EQUALITY(is_constant, true);
 
-    TEST_EQUALITY(Op != Teuchos::null, true);
-    TEST_EQUALITY(H  != Teuchos::null, true);
-    TEST_EQUALITY(nullSpace != Teuchos::null, true);
-    TEST_EQUALITY(is_injected, true);
-    TEST_EQUALITY(is_constant, true);
-
-  } // LocalLexicographicConstant
-
+}  // LocalLexicographicConstant
 
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,Constructor,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,LinearInterpolation,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,LinearExtrapolation,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,PoissonOnCubeLinear,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,PoissonOnCubeConstant,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,LocalLexicographicLinear,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory,LocalLexicographicConstant,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, Constructor, Scalar, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, LinearInterpolation, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, LinearExtrapolation, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, PoissonOnCubeLinear, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, PoissonOnCubeConstant, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, LocalLexicographicLinear, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GeneralGeometricPFactory, LocalLexicographicConstant, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/GenericRFactory.cpp b/packages/muelu/test/unit_tests/GenericRFactory.cpp
index a4b9d2e64c6f..261ad46c9093 100644
--- a/packages/muelu/test/unit_tests/GenericRFactory.cpp
+++ b/packages/muelu/test/unit_tests/GenericRFactory.cpp
@@ -73,296 +73,287 @@
 
 namespace MueLuTests {
 
-  //this macro declares the unit-test-class:
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    //TEST_EQUALITY(needs != Teuchos::null, true);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, SymmetricProblem, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-
+// this macro declares the unit-test-class:
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // TEST_EQUALITY(needs != Teuchos::null, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, SymmetricProblem, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
+#endif
+#if !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+
+  // generate problem
+  LO maxLevels             = 3;
+  GO nEle                  = 63;
+  const RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", nEle);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
+  RCP<Matrix> Op = Pr->BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0)
+    out << "||NS|| = " << norms[0] << std::endl;
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                 // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+
+  // define transfer operators
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  Teuchos::ParameterList sapParamList;
+  sapParamList.sublist("matrixmatrix: kernel params").set("compute global constants", true);
+  RCP<SaPFactory> Pfact = rcp(new SaPFactory());
+  Pfact->SetParameterList(sapParamList);
+  RCP<Factory> Rfact = rcp(new GenericRFactory());
+  H->SetMaxCoarseSize(1);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LO)1);
+  smootherParamList.set("relaxation: damping factor", (SC)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  // Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  RCP<Level> coarseLevel  = H->GetLevel(1);
+  RCP<Matrix> P1          = coarseLevel->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R1          = coarseLevel->Get<RCP<Matrix> >("R");
+  RCP<Level> coarseLevel2 = H->GetLevel(2);
+  RCP<Matrix> P2          = coarseLevel2->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R2          = coarseLevel2->Get<RCP<Matrix> >("R");
+
+  TEST_EQUALITY(Finest->IsAvailable("PreSmoother"), true);
+  TEST_EQUALITY(Finest->IsAvailable("PostSmoother"), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother"), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother"), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother"), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother"), false);
+
+  // test some basic multgrid data
+  TEST_EQUALITY(P1->getGlobalNumEntries(), R1->getGlobalNumEntries());
+  TEST_EQUALITY(P1->getGlobalNumRows(), R1->getGlobalNumCols());
+  TEST_EQUALITY(P1->getGlobalNumCols(), R1->getGlobalNumRows());
+  TEST_EQUALITY(P2->getGlobalNumEntries(), R2->getGlobalNumEntries());
+  TEST_EQUALITY(P2->getGlobalNumRows(), R2->getGlobalNumCols());
+  TEST_EQUALITY(P2->getGlobalNumCols(), R2->getGlobalNumRows());
+
+  // RCP<Teuchos::FancyOStream> fos = getFancyOStream(Teuchos::rcpFromRef(cout));
+
+  // since A is chosen symmetric, it is P^T = R
+  // check P^T * P = R * P
+  // note: the Epetra matrix-matrix multiplication using implicit transpose is buggy in parallel case
+  //       (for multiplication of a square matrix with a rectangular matrix)
+  //       however it seems to work for two rectangular matrices
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LO, GO, Node> > RP  = Xpetra::MatrixMatrix<Scalar, LO, GO, Node>::Multiply(*R1, false, *P1, false, out);
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LO, GO, Node> > PtP = Xpetra::MatrixMatrix<Scalar, LO, GO, Node>::Multiply(*P1, true, *P1, false, out);
+
+  RCP<Vector> x    = VectorFactory::Build(RP->getDomainMap());
+  RCP<Vector> bRP  = VectorFactory::Build(RP->getRangeMap());
+  RCP<Vector> bPtP = VectorFactory::Build(PtP->getRangeMap());
+
+  x->randomize();
+  RP->apply(*x, *bRP);
+  PtP->apply(*x, *bPtP);
+
+  TEST_EQUALITY(bRP->norm1() - bPtP->norm1() < 1e-12, true);
+
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LO, GO, Node> > RP2  = Xpetra::MatrixMatrix<Scalar, LO, GO, Node>::Multiply(*R2, false, *P2, false, out);
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LO, GO, Node> > PtP2 = Xpetra::MatrixMatrix<Scalar, LO, GO, Node>::Multiply(*P2, true, *P2, false, out);
+
+  x    = VectorFactory::Build(RP2->getDomainMap());
+  bRP  = VectorFactory::Build(RP2->getRangeMap());
+  bPtP = VectorFactory::Build(PtP2->getRangeMap());
+
+  x->randomize();
+  RP2->apply(*x, *bRP);
+  PtP2->apply(*x, *bPtP);
+
+  TEST_EQUALITY(bRP->norm1() - bPtP->norm1() < 1e-12, true);
+
+  // R1->describe(*fos,Teuchos::VERB_EXTREME);
+}
+
+// check Hierarchy::Setup routine with GenericRFactory as restriction factory
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, GenericRSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
+#endif
+#if !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+
+  for (int i = 1; i < 5; i++) {
     // generate problem
-    LO maxLevels = 3;
-    GO nEle = 63;
-    const RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), nEle, 0, comm);
-    Teuchos::ParameterList matrixParameters;
-    matrixParameters.set("nx",nEle);
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    RCP<Matrix> A                       = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(50 * i * comm->getSize());
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
-    RCP<Matrix> Op = Pr->BuildMatrix();
+    // Multigrid Hierarchy
+    Hierarchy H(A);
+    H.setVerbLevel(Teuchos::VERB_HIGH);
 
     // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (SC) 1.0);
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+    nullSpace->putScalar((SC)1.0);
     Teuchos::Array<magnitude_type> norms(1);
     nullSpace->norm1(norms);
     if (comm->getRank() == 0)
       out << "||NS|| = " << norms[0] << std::endl;
 
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-
-    // define transfer operators
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    Teuchos::ParameterList sapParamList;
-    sapParamList.sublist("matrixmatrix: kernel params").set("compute global constants", true);
-    RCP<SaPFactory>         Pfact = rcp( new SaPFactory());
-    Pfact->SetParameterList(sapParamList);
-    RCP<Factory>           Rfact = rcp( new GenericRFactory() );
-    H->SetMaxCoarseSize(1);
+    RCP<PgPFactory> Pfact  = rcp(new PgPFactory());
+    RCP<Factory> Rfact     = rcp(new GenericRFactory());
+    RCP<RAPFactory> Acfact = rcp(new RAPFactory());
 
     // setup smoothers
     Teuchos::ParameterList smootherParamList;
     smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LO) 1);
-    smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    //Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    H->Setup(M, 0, maxLevels);
-
-    RCP<Level> coarseLevel = H->GetLevel(1);
-    RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-    RCP<Level> coarseLevel2 = H->GetLevel(2);
-    RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
-
-    TEST_EQUALITY(Finest->IsAvailable("PreSmoother"), true);
-    TEST_EQUALITY(Finest->IsAvailable("PostSmoother"), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother"), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother"), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother"), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother"), false);
-
-    // test some basic multgrid data
-    TEST_EQUALITY(P1->getGlobalNumEntries(), R1->getGlobalNumEntries());
-    TEST_EQUALITY(P1->getGlobalNumRows(), R1->getGlobalNumCols());
-    TEST_EQUALITY(P1->getGlobalNumCols(), R1->getGlobalNumRows());
-    TEST_EQUALITY(P2->getGlobalNumEntries(), R2->getGlobalNumEntries());
-    TEST_EQUALITY(P2->getGlobalNumRows(), R2->getGlobalNumCols());
-    TEST_EQUALITY(P2->getGlobalNumCols(), R2->getGlobalNumRows());
-
-
-    //RCP<Teuchos::FancyOStream> fos = getFancyOStream(Teuchos::rcpFromRef(cout));
-
-    // since A is chosen symmetric, it is P^T = R
-    // check P^T * P = R * P
-    // note: the Epetra matrix-matrix multiplication using implicit transpose is buggy in parallel case
-    //       (for multiplication of a square matrix with a rectangular matrix)
-    //       however it seems to work for two rectangular matrices
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LO,GO,Node> > RP = Xpetra::MatrixMatrix<Scalar,LO,GO,Node>::Multiply(*R1,false,*P1,false,out);
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LO,GO,Node> > PtP = Xpetra::MatrixMatrix<Scalar,LO,GO,Node>::Multiply(*P1,true,*P1,false,out);
-
-    RCP<Vector> x = VectorFactory::Build(RP->getDomainMap());
-    RCP<Vector> bRP  = VectorFactory::Build(RP->getRangeMap());
-    RCP<Vector> bPtP = VectorFactory::Build(PtP->getRangeMap());
-
-    x->randomize();
-    RP->apply(*x,*bRP);
-    PtP->apply(*x,*bPtP);
-
-    TEST_EQUALITY(bRP->norm1() - bPtP->norm1() < 1e-12, true);
-
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LO,GO,Node> > RP2 = Xpetra::MatrixMatrix<Scalar,LO,GO,Node>::Multiply(*R2,false,*P2,false,out);
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LO,GO,Node> > PtP2 = Xpetra::MatrixMatrix<Scalar,LO,GO,Node>::Multiply(*P2,true,*P2,false,out);
-
-    x = VectorFactory::Build(RP2->getDomainMap());
-    bRP  = VectorFactory::Build(RP2->getRangeMap());
-    bPtP = VectorFactory::Build(PtP2->getRangeMap());
-
-    x->randomize();
-    RP2->apply(*x,*bRP);
-    PtP2->apply(*x,*bPtP);
-
-    TEST_EQUALITY(bRP->norm1() - bPtP->norm1() < 1e-12, true);
-
-
-    //R1->describe(*fos,Teuchos::VERB_EXTREME);
-  }
-
-  // check Hierarchy::Setup routine with GenericRFactory as restriction factory
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(GenericRFactory, GenericRSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-
-    for (int i=1; i<5; i++) {
-      // generate problem
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(50*i*comm->getSize());
-
-      // Multigrid Hierarchy
-      Hierarchy H(A);
-      H.setVerbLevel(Teuchos::VERB_HIGH);
-
-      // build nullspace
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),1);
-      nullSpace->putScalar( (SC) 1.0);
-      Teuchos::Array<magnitude_type> norms(1);
-      nullSpace->norm1(norms);
-      if (comm->getRank() == 0)
-        out << "||NS|| = " << norms[0] << std::endl;
-
-      RCP<PgPFactory>        Pfact = rcp( new PgPFactory());
-      RCP<Factory>           Rfact = rcp( new GenericRFactory() );
-      RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-
-      // setup smoothers
-      Teuchos::ParameterList smootherParamList;
-      smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-      smootherParamList.set("relaxation: sweeps", (LO) 1);
-      smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-      RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-      RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-      Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-      // Multigrid setup phase (using default parameters)
-      FactoryManager M0; // how to build aggregates and smoother of the first level
-      M0.SetFactory("A",            Acfact);
-      M0.SetFactory("P",            Pfact);
-      M0.SetFactory("R",            Rfact);
-      M0.SetFactory("Smoother",     SmooFact);
-      M0.SetFactory("CoarseSolver", SmooFact);
-
-      FactoryManager M1; // first coarse level (Plain aggregation)
-      M1.SetFactory("A",            Acfact);
-      M1.SetFactory("P",            Pfact);
-      M1.SetFactory("R",            Rfact);
-      M1.SetFactory("Smoother",     SmooFact);
-      M1.SetFactory("CoarseSolver", SmooFact);
-
-      FactoryManager M2; // last level (SA)
-      M2.SetFactory("A",            Acfact);
-      M2.SetFactory("P",            Pfact);
-      M2.SetFactory("R",            Rfact);
-      M2.SetFactory("Smoother",     SmooFact);
-      M2.SetFactory("CoarseSolver", SmooFact);
-
-      bool bIsLastLevel = false;
-      if(!bIsLastLevel) bIsLastLevel = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));
-      if(!bIsLastLevel) bIsLastLevel = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
-      if(!bIsLastLevel) bIsLastLevel = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null );
-
-      RCP<Level> l0 = H.GetLevel(0);
-      RCP<Level> l1;
-      RCP<Level> l2;
-
-      if (H.GetNumLevels() > 1) l1 = H.GetLevel(1);
-      if (H.GetNumLevels() > 2) l2 = H.GetLevel(2);
-
-      /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-        l0->print(*stdout,Teuchos::VERB_EXTREME);
-        if(l1 != Teuchos::null) l1->print(*stdout,Teuchos::VERB_EXTREME);
-        if(l2 != Teuchos::null) l2->print(*stdout,Teuchos::VERB_EXTREME);*/
-
-      TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 1 ? true : false));
-      TEST_EQUALITY(l0->IsAvailable("R",            MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l0->IsAvailable("P",            MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-      if (H.GetNumLevels() > 1)
-        TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l0->IsRequested("R",            MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-      TEST_EQUALITY(l0->IsRequested("P",            MueLu::NoFactory::get()), false);
-
-      if (l1 != Teuchos::null) {
-        TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 2 ? true : false));
-        TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-        if (H.GetNumLevels() > 2)
-          TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l1->IsRequested("Graph",        MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(l1->IsRequested("Aggregates",   MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(l1->IsRequested("Nullspace",    MueLu::NoFactory::get()), false);
-      }
-      if (l2 != Teuchos::null) {
-        TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 3 ? true : false));
-        TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-        if (H.GetNumLevels() > 3)
-          TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(l2->IsRequested("Graph",        MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(l2->IsRequested("Aggregates",   MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(l2->IsRequested("Nullspace",    MueLu::NoFactory::get()), false);
-      }
-    } // end for i=1..5
-
-  }
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GenericRFactory, Constructor, Scalar, LO, GO, Node) \
+    smootherParamList.set("relaxation: sweeps", (LO)1);
+    smootherParamList.set("relaxation: damping factor", (SC)1.0);
+    RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+    RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+    // Multigrid setup phase (using default parameters)
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
+    M0.SetFactory("A", Acfact);
+    M0.SetFactory("P", Pfact);
+    M0.SetFactory("R", Rfact);
+    M0.SetFactory("Smoother", SmooFact);
+    M0.SetFactory("CoarseSolver", SmooFact);
+
+    FactoryManager M1;  // first coarse level (Plain aggregation)
+    M1.SetFactory("A", Acfact);
+    M1.SetFactory("P", Pfact);
+    M1.SetFactory("R", Rfact);
+    M1.SetFactory("Smoother", SmooFact);
+    M1.SetFactory("CoarseSolver", SmooFact);
+
+    FactoryManager M2;  // last level (SA)
+    M2.SetFactory("A", Acfact);
+    M2.SetFactory("P", Pfact);
+    M2.SetFactory("R", Rfact);
+    M2.SetFactory("Smoother", SmooFact);
+    M2.SetFactory("CoarseSolver", SmooFact);
+
+    bool bIsLastLevel = false;
+    if (!bIsLastLevel) bIsLastLevel = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    if (!bIsLastLevel) bIsLastLevel = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    if (!bIsLastLevel) bIsLastLevel = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+
+    RCP<Level> l0 = H.GetLevel(0);
+    RCP<Level> l1;
+    RCP<Level> l2;
+
+    if (H.GetNumLevels() > 1) l1 = H.GetLevel(1);
+    if (H.GetNumLevels() > 2) l2 = H.GetLevel(2);
+
+    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+      l0->print(*stdout,Teuchos::VERB_EXTREME);
+      if(l1 != Teuchos::null) l1->print(*stdout,Teuchos::VERB_EXTREME);
+      if(l2 != Teuchos::null) l2->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+    TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 1 ? true : false));
+    TEST_EQUALITY(l0->IsAvailable("R", MueLu::NoFactory::get()), false);
+    TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l0->IsAvailable("P", MueLu::NoFactory::get()), false);
+    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+    if (H.GetNumLevels() > 1)
+      TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l0->IsRequested("R", MueLu::NoFactory::get()), false);
+    TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+    TEST_EQUALITY(l0->IsRequested("P", MueLu::NoFactory::get()), false);
+
+    if (l1 != Teuchos::null) {
+      TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 2 ? true : false));
+      TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      if (H.GetNumLevels() > 2)
+        TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l1->IsRequested("Graph", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(l1->IsRequested("Aggregates", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(l1->IsRequested("Nullspace", MueLu::NoFactory::get()), false);
+    }
+    if (l2 != Teuchos::null) {
+      TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), (H.GetNumLevels() > 3 ? true : false));
+      TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      if (H.GetNumLevels() > 3)
+        TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(l2->IsRequested("Graph", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(l2->IsRequested("Aggregates", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(l2->IsRequested("Nullspace", MueLu::NoFactory::get()), false);
+    }
+  }  // end for i=1..5
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GenericRFactory, Constructor, Scalar, LO, GO, Node)      \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GenericRFactory, SymmetricProblem, Scalar, LO, GO, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(GenericRFactory, GenericRSetup, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-}//namespace MueLuTests
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Hierarchy.cpp b/packages/muelu/test/unit_tests/Hierarchy.cpp
index f0b74a1b2204..14cb183970c4 100644
--- a/packages/muelu/test/unit_tests/Hierarchy.cpp
+++ b/packages/muelu/test/unit_tests/Hierarchy.cpp
@@ -77,1031 +77,1008 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Hierarchy> H = rcp(new Hierarchy);
-
-    TEST_INEQUALITY(H, Teuchos::null);
-
-  } //Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, DescriptionCaching, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    /*
-     * Test to confirm that Hierarchy::description():
-     * a) gives the same result before and after a call to SetupRe() when the number of levels has not changed
-     * b) gives a different result before and after a call to SetupRe() when the number of levels has changed
-     *
-     * The occasion for this test is the introduction of caching for the result of description(), to avoid redundant
-     * computation.
-     */
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numRows = 399;
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(numRows);
-    GO nx = numRows;
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    Teuchos::ParameterList MueLuList;
-    MueLuList.set("verbosity","none");
-    MueLuList.set("coarse: max size",numRows-1); // make it so we want two levels
-    MueLuList.set("max levels",2);
-
-    if(A->getRowMap()->lib() == Xpetra::UseEpetra) {
-      MueLuList.set("use kokkos refactor", false);
-    }
-
-    Teuchos::ParameterList userParamList = MueLuList.sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    Teuchos::RCP<MueLu::Hierarchy<Scalar,LocalOrdinal,GlobalOrdinal,Node> > H =
-        MueLu::CreateXpetraPreconditioner<Scalar,LocalOrdinal,GlobalOrdinal,Node>(A, MueLuList);
-
-    // confirm that we did get a hierarchy with two levels -- a sanity check for this test
-    TEST_EQUALITY(2, H->GetGlobalNumLevels());
-
-    using namespace std;
-    string descriptionTwoLevel = H->description();
-
-    // SetupRe() will reset description, but since we haven't changed anything, the result from description() should not change
-    H->SetupRe();
-    string descriptionActual = H->description();
-    TEST_EQUALITY(descriptionActual,descriptionTwoLevel);
-
-    // now, we allow a larger coarse size; we should get just one level during SetupRe()
-    H->SetMaxCoarseSize(numRows+1);
-    H->SetupRe();
-    // as a sanity check for the test, make sure that we do have just one level
-    TEST_EQUALITY(1, H->GetGlobalNumLevels());
-    descriptionActual = H->description();
-
-    // since the number of levels has changed, the new description should differ from the old
-    TEST_INEQUALITY(descriptionActual, descriptionTwoLevel);
-  }//DescriptionCaching
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetAndGetLevel, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Hierarchy H;
-    //   RCP<Level> level = rcp(new Level());
-    //   H.SetLevel(level);
-    //   RCP<Level> dupLevel = H.GetLevel(1);
-
-    //  TEST_EQUALITY(level, dupLevel);
-
-  }//SetAndGetLevel
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, GetNumLevels, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Hierarchy H;
-    RCP<Level> levelOne = rcp(new Level() );
-    RCP<Level> levelTwo = rcp(new Level() );
-    RCP<Level> levelThree = rcp(new Level() );
-    //   H.SetLevel(levelOne);
-    //   H.SetLevel(levelTwo);
-    //   H.SetLevel(levelThree);
-    //   TEST_EQUALITY(H.GetNumLevels(), 3);
-
-  }//GetNumLevels
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, KeepAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 399*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    Hierarchy H(A);
-    H.SetMaxCoarseSize(1);
-    H.GetLevel(0)->Set("Coordinates", coordinates);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", Teuchos::null);
-    M.SetFactory("CoarseSolver", Teuchos::null);
-
-    H.GetLevel(0)->Keep("Aggregates", UncoupledAggFact.get());
-    H.Setup(M, 0, 2);
-
-    for (LocalOrdinal l=0; l<H.GetNumLevels()-1;l++) {
-      TEST_EQUALITY(H.GetLevel(l)->IsAvailable("Aggregates", UncoupledAggFact.get()), true);
-    }
-
-  } //FullPopulate_KeepAggregates
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Iterate, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    //matrix
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 6561*comm->getSize();  //=8*3^6
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-    RCP<const Map > map = Op->getRowMap();
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
-
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-
-    MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
-    H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<MueLu::Level> Finest = H.GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    Finest->Set("Coordinates", coordinates);
-    Finest->Set("Nullspace", nullSpace);
-    Finest->Set("A", Op);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<CoalesceDropFactory> cdFact;
-    RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
-
-    RCP<SaPFactory>    Pfact  = rcp( new SaPFactory() );
-    RCP<TransPFactory> Rfact  = rcp( new TransPFactory());
-    RCP<RAPFactory>    Acfact = rcp( new RAPFactory() );
-
-    RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
-    RCP<SmootherFactory>    SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherPrototype> coarseProto = rcp(new DirectSolver("Klu"));
-    RCP<SmootherFactory> coarseSolveFact = rcp( new SmootherFactory(coarseProto, Teuchos::null));
-
-    int maxLevels = 5;
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("A", Acfact);
-    M.SetFactory("Ptent", TentPFact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-    M.SetFactory("Coordinates", TentPFact);
-
-    H.Setup(M, 0, maxLevels);
-
-    RCP<MultiVector> X = MultiVectorFactory::Build(map, 1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
-
-    X->setSeed(846930886);
-    X->randomize();
-    //Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
-
-    X->norm2(norms);
-    X->scale(1/norms[0]);
-    X->norm2(norms);
-    out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-    RHS->putScalar( (Scalar) 0.0);
-
-    int iterations=15;
-    H.Iterate(*RHS, *X, iterations);
-
-    X->norm2(norms);
-    out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) <<
-      std::setprecision(10) << norms[0] << std::endl;
-
-    norms = Utilities::ResidualNorm(*Op, *X, *RHS);
-    out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
-
-    TEST_EQUALITY(norms[0] < 100*TMT::eps(), true);
-
-  } //Iterate
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, IterateWithFusedProlongationUpdate, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    //matrix
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 6561*comm->getSize();  //=8*3^6
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-    RCP<const Map > map = Op->getRowMap();
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
-
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-
-    MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
-    H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<MueLu::Level> Finest = H.GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    Finest->Set("Coordinates", coordinates);
-    Finest->Set("Nullspace", nullSpace);
-    Finest->Set("A", Op);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<CoalesceDropFactory> cdFact;
-    RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
-
-    RCP<SaPFactory>    Pfact  = rcp( new SaPFactory() );
-    RCP<TransPFactory> Rfact  = rcp( new TransPFactory());
-    RCP<RAPFactory>    Acfact = rcp( new RAPFactory() );
-
-    RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
-    RCP<SmootherFactory>    SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherPrototype> coarseProto = rcp(new DirectSolver("Klu"));
-    RCP<SmootherFactory> coarseSolveFact = rcp( new SmootherFactory(coarseProto, Teuchos::null));
-
-    int maxLevels = 5;
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("A", Acfact);
-    M.SetFactory("Ptent", TentPFact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-    M.SetFactory("Coordinates", TentPFact);
-
-    H.Setup(M, 0, maxLevels);
-
-    H.SetFuseProlongationAndUpdate(true);
-
-    RCP<MultiVector> X = MultiVectorFactory::Build(map, 1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
-
-    X->setSeed(846930886);
-    X->randomize();
-    //Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
-
-    X->norm2(norms);
-    X->scale(1/norms[0]);
-    X->norm2(norms);
-    out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-    RHS->putScalar( (Scalar) 0.0);
-
-    int iterations=15;
-    H.Iterate(*RHS, *X, iterations);
-
-    X->norm2(norms);
-    out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) <<
-      std::setprecision(10) << norms[0] << std::endl;
-
-    norms = Utilities::ResidualNorm(*Op, *X, *RHS);
-    out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
-    TEST_EQUALITY(norms[0] < 100*TMT::eps(), true);
-
-  } //IterateWithFusedProlongationUpdate
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, IterateWithImplicitRestriction, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    //matrix
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 6561*comm->getSize();  //=8*3^6
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-    RCP<const Map > map = Op->getRowMap();
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", map, galeriList);
-
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<typename TST::magnitudeType> norms(1);
-    nullSpace->norm1(norms);
-
-    MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
-    H.SetImplicitTranspose(true);
-    H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<MueLu::Level> Finest = H.GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A", Op);
-    Finest->Set("Nullspace", nullSpace);
-    Finest->Set("Coordinates", coordinates);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-    RCP<CoalesceDropFactory> cdFact;
-    RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
-
-    RCP<SaPFactory>         Pfact = rcp( new SaPFactory() );
-    RCP<TransPFactory>      Rfact = rcp( new TransPFactory());
-    RCP<RAPFactory>         Acfact = rcp( new RAPFactory() );
-    Teuchos::ParameterList Aclist = *(Acfact->GetValidParameterList());
-    Aclist.set("transpose: use implicit", true);
-    Acfact->SetParameterList(Aclist);
-
-    RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
-    RCP<SmootherFactory>    SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherPrototype> coarseProto = rcp(new DirectSolver("Klu"));
-    RCP<SmootherFactory> coarseSolveFact = rcp( new SmootherFactory(coarseProto, Teuchos::null));
-
-    int maxLevels = 5;
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("A", Acfact);
-    M.SetFactory("Ptent", TentPFact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-    M.SetFactory("Coordinates", TentPFact);
-
-    H.Setup(M, 0, maxLevels);
-
-    //  Teuchos::ParameterList status;
-    // status.print(out, Teuchos::ParameterList::PrintOptions().indent(2));
-
-    RCP<MultiVector> X = MultiVectorFactory::Build(map, 1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
-
-    X->setSeed(846930886);
-    X->randomize();
-    //Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
-
-    X->norm2(norms);
-    X->scale(1/norms[0]);
-    X->norm2(norms);
-    out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-    RHS->putScalar( (Scalar) 0.0);
-
-    int iterations=15;
-    H.Iterate(*RHS, *X, iterations);
-
-    X->norm2(norms);
-    out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) <<
-      std::setprecision(10) << norms[0] << std::endl;
-
-    norms = Utilities::ResidualNorm(*Op, *X, *RHS);
-    out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
-    TEST_EQUALITY(norms[0] < 100*TMT::eps(), true);
-
-  } //Iterate
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy1level, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299*comm->getSize());
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-
-    bool r = H.Setup(0, Teuchos::null, rcpFromRef(M0), Teuchos::null); TEST_EQUALITY(r, true); // cf. Teuchos Bug 5214
-
-    RCP<Level> l0 = H.GetLevel(0);
-
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    //TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-
-    X->putScalar( (Scalar) 0.0);
-
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy1levelv2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TESTING_SET_OSTREAM;
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299*comm->getSize());
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-
-    H.Setup(M0, 0, 1);
-
-    RCP<Level> l0 = H.GetLevel(0);
-
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    //TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-
-    X->putScalar( (Scalar) 0.0);
-
-    int iterations = 10;
-    H.Iterate(*RHS, *X, iterations);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Hierarchy> H = rcp(new Hierarchy);
+
+  TEST_INEQUALITY(H, Teuchos::null);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, DescriptionCaching, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  /*
+   * Test to confirm that Hierarchy::description():
+   * a) gives the same result before and after a call to SetupRe() when the number of levels has not changed
+   * b) gives a different result before and after a call to SetupRe() when the number of levels has changed
+   *
+   * The occasion for this test is the introduction of caching for the result of description(), to avoid redundant
+   * computation.
+   */
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numRows                         = 399;
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(numRows);
+  GO nx                               = numRows;
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  Teuchos::ParameterList MueLuList;
+  MueLuList.set("verbosity", "none");
+  MueLuList.set("coarse: max size", numRows - 1);  // make it so we want two levels
+  MueLuList.set("max levels", 2);
+
+  if (A->getRowMap()->lib() == Xpetra::UseEpetra) {
+    MueLuList.set("use kokkos refactor", false);
   }
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299*comm->getSize());
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    M1.SetFactory("P", rcp(new TentativePFactory()));
-
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
-
-    bool r; // cf. Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
-
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
-
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);*/
-
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-
-    X->putScalar( (Scalar) 0.0);
-
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
-#  endif
+  Teuchos::ParameterList userParamList = MueLuList.sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  Teuchos::RCP<MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> > H =
+      MueLu::CreateXpetraPreconditioner<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, MueLuList);
+
+  // confirm that we did get a hierarchy with two levels -- a sanity check for this test
+  TEST_EQUALITY(2, H->GetGlobalNumLevels());
+
+  using namespace std;
+  string descriptionTwoLevel = H->description();
+
+  // SetupRe() will reset description, but since we haven't changed anything, the result from description() should not change
+  H->SetupRe();
+  string descriptionActual = H->description();
+  TEST_EQUALITY(descriptionActual, descriptionTwoLevel);
+
+  // now, we allow a larger coarse size; we should get just one level during SetupRe()
+  H->SetMaxCoarseSize(numRows + 1);
+  H->SetupRe();
+  // as a sanity check for the test, make sure that we do have just one level
+  TEST_EQUALITY(1, H->GetGlobalNumLevels());
+  descriptionActual = H->description();
+
+  // since the number of levels has changed, the new description should differ from the old
+  TEST_INEQUALITY(descriptionActual, descriptionTwoLevel);
+}  // DescriptionCaching
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetAndGetLevel, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Hierarchy H;
+  //   RCP<Level> level = rcp(new Level());
+  //   H.SetLevel(level);
+  //   RCP<Level> dupLevel = H.GetLevel(1);
+
+  //  TEST_EQUALITY(level, dupLevel);
+
+}  // SetAndGetLevel
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, GetNumLevels, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Hierarchy H;
+  RCP<Level> levelOne   = rcp(new Level());
+  RCP<Level> levelTwo   = rcp(new Level());
+  RCP<Level> levelThree = rcp(new Level());
+  //   H.SetLevel(levelOne);
+  //   H.SetLevel(levelTwo);
+  //   H.SetLevel(levelThree);
+  //   TEST_EQUALITY(H.GetNumLevels(), 3);
+
+}  // GetNumLevels
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, KeepAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 399 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  Hierarchy H(A);
+  H.SetMaxCoarseSize(1);
+  H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", Teuchos::null);
+  M.SetFactory("CoarseSolver", Teuchos::null);
+
+  H.GetLevel(0)->Keep("Aggregates", UncoupledAggFact.get());
+  H.Setup(M, 0, 2);
+
+  for (LocalOrdinal l = 0; l < H.GetNumLevels() - 1; l++) {
+    TEST_EQUALITY(H.GetLevel(l)->IsAvailable("Aggregates", UncoupledAggFact.get()), true);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level_AggregateSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
+}  // FullPopulate_KeepAggregates
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Iterate, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  using TST                   = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299*comm->getSize());
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-    Teuchos::ParameterList ifpack2Params;
-    M0.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+  // matrix
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 6561 * comm->getSize();  //=8*3^6
+  RCP<Matrix> Op                      = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  RCP<const Map> map                  = Op->getRowMap();
 
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    M1.SetFactory("P", rcp(new TentativePFactory()));
-    M1.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
 
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
+  MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
+  H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
-    bool r; // cf. Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+  RCP<MueLu::Level> Finest = H.GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
+  Finest->Set("Coordinates", coordinates);
+  Finest->Set("Nullspace", nullSpace);
+  Finest->Set("A", Op);
 
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);*/
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
 
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  RCP<CoalesceDropFactory> cdFact;
+  RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
 
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+  RCP<SaPFactory> Pfact    = rcp(new SaPFactory());
+  RCP<TransPFactory> Rfact = rcp(new TransPFactory());
+  RCP<RAPFactory> Acfact   = rcp(new RAPFactory());
 
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
+  RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
-    X->putScalar( (Scalar) 0.0);
+  RCP<SmootherPrototype> coarseProto   = rcp(new DirectSolver("Klu"));
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
 
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
+  int maxLevels = 5;
 
-  }
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", TentPFact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+  M.SetFactory("Coordinates", TentPFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level_AggregateSmooth_3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
+  H.Setup(M, 0, maxLevels);
 
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace1D");
-    matrixParams.set("nx",(GlobalOrdinal)20);// needs to be even
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams,Xpetra::UseTpetra);
+  X->setSeed(846930886);
+  X->randomize();
+  // Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
 
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-    Teuchos::ParameterList ifpack2Params;
-    M0.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+  X->norm2(norms);
+  X->scale(1 / norms[0]);
+  X->norm2(norms);
+  out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    M1.SetFactory("P", rcp(new TentativePFactory()));
-    M1.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+  RHS->putScalar((Scalar)0.0);
 
+  int iterations = 15;
+  H.Iterate(*RHS, *X, iterations);
 
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
+  X->norm2(norms);
+  out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-    bool r; // cf. Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1));  TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null); TEST_EQUALITY(r, true);
+  norms = Utilities::ResidualNorm(*Op, *X, *RHS);
+  out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
 
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
+  TEST_EQUALITY(norms[0] < 100 * TMT::eps(), true);
 
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);*/
+}  // Iterate
 
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, IterateWithFusedProlongationUpdate, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  using TST                   = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
+  out << "version: " << MueLu::Version() << std::endl;
 
-    X->putScalar( (Scalar) 0.0);
+  // matrix
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 6561 * comm->getSize();  //=8*3^6
+  RCP<Matrix> Op                      = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  RCP<const Map> map                  = Op->getRowMap();
 
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
 
-  }
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
 
+  MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
+  H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_BlockTriDi_3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  RCP<MueLu::Level> Finest = H.GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
+  Finest->Set("Coordinates", coordinates);
+  Finest->Set("Nullspace", nullSpace);
+  Finest->Set("A", Op);
 
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    Teuchos::ParameterList matrixParams;
-    matrixParams.set("matrixType","Laplace1D");
-    matrixParams.set("nx",(GlobalOrdinal)100);// needs to be even
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams,Xpetra::UseTpetra);
+  RCP<CoalesceDropFactory> cdFact;
+  RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
 
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(10);
+  RCP<SaPFactory> Pfact    = rcp(new SaPFactory());
+  RCP<TransPFactory> Rfact = rcp(new TransPFactory());
+  RCP<RAPFactory> Acfact   = rcp(new RAPFactory());
 
+  RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
+  RCP<SmootherPrototype> coarseProto   = rcp(new DirectSolver("Klu"));
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
 
+  int maxLevels = 5;
 
-    //  Hierarchy H(A);
-    //    H.setVerbLevel(Teuchos::VERB_HIGH);
-    //    H.SetMaxCoarseSize(50);
-    //    H.GetLevel(0)->Set("Coordinates", coordinates);
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", TentPFact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+  M.SetFactory("Coordinates", TentPFact);
 
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-    Teuchos::ParameterList ifpack2Params;
-    ifpack2Params.set("smoother: use blockcrsmatrix storage",true);
-    M0.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("LINESMOOTHING_TRIDI_RELAXATION", ifpack2Params, 0)))));
+  H.Setup(M, 0, maxLevels);
 
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    RCP<FactoryBase> P = rcp(new TentativePFactory());
-    M1.SetFactory("P", P);
-    M1.SetFactory("Ptent", P); //FIXME: can it be done automatically in FactoryManager?
-    M1.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("LINESMOOTHING_TRIDI_RELAXATION", ifpack2Params, 0)))));
+  H.SetFuseProlongationAndUpdate(true);
 
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
-    bool r; // cf. bug Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));   TEST_EQUALITY(r, false);
-    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
+  X->setSeed(846930886);
+  X->randomize();
+  // Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
 
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
-    RCP<Level> l2 = H.GetLevel(2);
+  X->norm2(norms);
+  X->scale(1 / norms[0]);
+  X->norm2(norms);
+  out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
+  RHS->putScalar((Scalar)0.0);
 
-    /*
-    RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);
-      l2->print(*stdout,Teuchos::VERB_EXTREME);
-    std::cout<<"*************"<<std::endl;
-    */
+  int iterations = 15;
+  H.Iterate(*RHS, *X, iterations);
 
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
+  X->norm2(norms);
+  out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  norms = Utilities::ResidualNorm(*Op, *X, *RHS);
+  out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
+  TEST_EQUALITY(norms[0] < 100 * TMT::eps(), true);
 
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+}  // IterateWithFusedProlongationUpdate
 
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, IterateWithImplicitRestriction, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+  using TST                   = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  // matrix
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 6561 * comm->getSize();  //=8*3^6
+  RCP<Matrix> Op                      = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  RCP<const Map> map                  = Op->getRowMap();
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", map, galeriList);
+
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<typename TST::magnitudeType> norms(1);
+  nullSpace->norm1(norms);
+
+  MueLu::Hierarchy<Scalar, LocalOrdinal, GlobalOrdinal, Node> H;
+  H.SetImplicitTranspose(true);
+  H.setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<MueLu::Level> Finest = H.GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);
+  Finest->Set("Nullspace", nullSpace);
+  Finest->Set("Coordinates", coordinates);
+
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+  RCP<CoalesceDropFactory> cdFact;
+  RCP<TentativePFactory> TentPFact = rcp(new TentativePFactory());
+
+  RCP<SaPFactory> Pfact         = rcp(new SaPFactory());
+  RCP<TransPFactory> Rfact      = rcp(new TransPFactory());
+  RCP<RAPFactory> Acfact        = rcp(new RAPFactory());
+  Teuchos::ParameterList Aclist = *(Acfact->GetValidParameterList());
+  Aclist.set("transpose: use implicit", true);
+  Acfact->SetParameterList(Aclist);
+
+  RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherPrototype> coarseProto   = rcp(new DirectSolver("Klu"));
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
+
+  int maxLevels = 5;
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", TentPFact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+  M.SetFactory("Coordinates", TentPFact);
+
+  H.Setup(M, 0, maxLevels);
+
+  //  Teuchos::ParameterList status;
+  // status.print(out, Teuchos::ParameterList::PrintOptions().indent(2));
+
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+  X->setSeed(846930886);
+  X->randomize();
+  // Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  X->norm2(norms);
+  X->scale(1 / norms[0]);
+  X->norm2(norms);
+  out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+  RHS->putScalar((Scalar)0.0);
+
+  int iterations = 15;
+  H.Iterate(*RHS, *X, iterations);
+
+  X->norm2(norms);
+  out << "||X_" << std::setprecision(2) << iterations << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+  norms = Utilities::ResidualNorm(*Op, *X, *RHS);
+  out << "||res_" << std::setprecision(2) << iterations << "|| = " << std::setprecision(15) << norms[0] << std::endl;
+  TEST_EQUALITY(norms[0] < 100 * TMT::eps(), true);
+
+}  // Iterate
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy1level, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299 * comm->getSize());
 
-    X->putScalar( (Scalar) 0.0);
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
 
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
 
-  }
+  bool r = H.Setup(0, Teuchos::null, rcpFromRef(M0), Teuchos::null);
+  TEST_EQUALITY(r, true);  // cf. Teuchos Bug 5214
 
+  RCP<Level> l0 = H.GetLevel(0);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TESTING_SET_OSTREAM;
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
 
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
 
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 299*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  X->putScalar((Scalar)0.0);
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
 
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
-    H.GetLevel(0)->Set("Coordinates", coordinates);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy1levelv2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299 * comm->getSize());
 
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
 
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    RCP<FactoryBase> P = rcp(new TentativePFactory());
-    M1.SetFactory("P", P);
-    M1.SetFactory("Ptent", P); //FIXME: can it be done automatically in FactoryManager?
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
 
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
+  H.Setup(M0, 0, 1);
 
-    bool r; // cf. bug Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));   TEST_EQUALITY(r, false);
-    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
+  RCP<Level> l0 = H.GetLevel(0);
 
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
-    RCP<Level> l2 = H.GetLevel(2);
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
 
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);
-      l2->print(*stdout,Teuchos::VERB_EXTREME);*/
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
 
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
 
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  X->putScalar((Scalar)0.0);
 
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
 
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra);
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    X->putScalar( (Scalar) 0.0);
+#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_AMESOS)
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299 * comm->getSize());
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  M1.SetFactory("P", rcp(new TentativePFactory()));
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+#endif
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level_AggregateSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
 
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
-  }
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(299 * comm->getSize());
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+  Teuchos::ParameterList ifpack2Params;
+  M0.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  M1.SetFactory("P", rcp(new TentativePFactory()));
+  M1.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy2level_AggregateSmooth_3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
 
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace1D");
+  matrixParams.set("nx", (GlobalOrdinal)20);  // needs to be even
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams, Xpetra::UseTpetra);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+  Teuchos::ParameterList ifpack2Params;
+  M0.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  M1.SetFactory("P", rcp(new TentativePFactory()));
+  M1.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("AGGREGATE", ifpack2Params, 0)))));
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_BlockTriDi_3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_BlockSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Teuchos::ParameterList matrixParams;
+  matrixParams.set("matrixType", "Laplace1D");
+  matrixParams.set("nx", (GlobalOrdinal)100);  // needs to be even
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams, Xpetra::UseTpetra);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(10);
+
+  //  Hierarchy H(A);
+  //    H.setVerbLevel(Teuchos::VERB_HIGH);
+  //    H.SetMaxCoarseSize(50);
+  //    H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+  Teuchos::ParameterList ifpack2Params;
+  ifpack2Params.set("smoother: use blockcrsmatrix storage", true);
+  M0.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("LINESMOOTHING_TRIDI_RELAXATION", ifpack2Params, 0)))));
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  RCP<FactoryBase> P = rcp(new TentativePFactory());
+  M1.SetFactory("P", P);
+  M1.SetFactory("Ptent", P);  // FIXME: can it be done automatically in FactoryManager?
+  M1.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("LINESMOOTHING_TRIDI_RELAXATION", ifpack2Params, 0)))));
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. bug Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+  RCP<Level> l2 = H.GetLevel(2);
+
+  /*
+  RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);
+    l2->print(*stdout,Teuchos::VERB_EXTREME);
+  std::cout<<"*************"<<std::endl;
+  */
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 299 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+  H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  RCP<FactoryBase> P = rcp(new TentativePFactory());
+  M1.SetFactory("P", P);
+  M1.SetFactory("Ptent", P);  // FIXME: can it be done automatically in FactoryManager?
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. bug Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+  RCP<Level> l2 = H.GetLevel(2);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);
+    l2->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_BlockSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
     MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
     MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
+#endif
 
     RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 10*comm->getSize();
+    GO nx                               = 10 * comm->getSize();
     Teuchos::ParameterList galeriList, ifpack2Params;
     galeriList.set("nx", nx);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(galeriList,Xpetra::UseTpetra);
+    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(galeriList, Xpetra::UseTpetra);
 
-    ifpack2Params.set("smoother: use blockcrsmatrix storage",true);
+    ifpack2Params.set("smoother: use blockcrsmatrix storage", true);
 
     // Multigrid Hierarchy
     Hierarchy H(A);
@@ -1109,60 +1086,63 @@ namespace MueLuTests {
     H.SetMaxCoarseSize(2);
 
     // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
+    FactoryManager M0;  // how to build aggregates and smoother of the first level
     M0.SetKokkosRefactor(false);
-    M0.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", ifpack2Params, 0)))));
+    M0.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", ifpack2Params, 0)))));
 
-    FactoryManager M1; // first coarse level (Plain aggregation)
+    FactoryManager M1;  // first coarse level (Plain aggregation)
     M1.SetKokkosRefactor(false);
     M1.SetFactory("A", rcp(new RAPFactory()));
     RCP<FactoryBase> P = rcp(new TentativePFactory());
     M1.SetFactory("P", P);
-    M1.SetFactory("Ptent", P); //FIXME: can it be done automatically in FactoryManager?
-    M1.SetFactory("Smoother",rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", ifpack2Params, 0)))));
+    M1.SetFactory("Ptent", P);  // FIXME: can it be done automatically in FactoryManager?
+    M1.SetFactory("Smoother", rcp(new SmootherFactory(rcp(new TrilinosSmoother("RELAXATION", ifpack2Params, 0)))));
 
-    FactoryManager M2; // last level (SA)
+    FactoryManager M2;  // last level (SA)
     M2.SetKokkosRefactor(false);
     M2.SetFactory("A", rcp(new RAPFactory()));
     M2.SetFactory("P", rcp(new SaPFactory()));
 
-    bool r; // cf. bug Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));   TEST_EQUALITY(r, false);
-    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
+    bool r;  // cf. bug Teuchos Bug 5214
+    r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+    TEST_EQUALITY(r, false);
+    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+    TEST_EQUALITY(r, false);
+    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+    TEST_EQUALITY(r, true);
 
     RCP<Level> l0 = H.GetLevel(0);
     RCP<Level> l1 = H.GetLevel(1);
     RCP<Level> l2 = H.GetLevel(2);
 
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
     TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
     TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+    TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+    TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+    TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
     TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
 
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
     TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
 
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
     // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
 
     RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
@@ -1170,695 +1150,709 @@ namespace MueLuTests {
     RHS->setSeed(846930886);
     RHS->randomize();
 
-    X->putScalar( (Scalar) 0.0);
+    X->putScalar((Scalar)0.0);
 
-    int iterations=10;
+    int iterations = 10;
     H.Iterate(*RHS, *X, iterations);
-    }
   }
+}
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_NoPreSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3level_NoPreSmooth, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TESTING_SET_OSTREAM;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 299*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
-    H.GetLevel(0)->Set("Coordinates", coordinates);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    RCP<FactoryBase> P = rcp(new TentativePFactory());
-    M1.SetFactory("P", P);
-    M1.SetFactory("Ptent", P); //FIXME: can it be done automatically in FactoryManager?
-
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
-
-    // Here we want no pre-smoothing on levels 0 and 1
-    RCP<SmootherPrototype> noSmooProto;
-    RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
-    RCP<SmootherFactory>   SmooFact = rcp( new SmootherFactory(noSmooProto,smooProto) );
-    M0.SetFactory("Smoother", SmooFact);
-    M1.SetFactory("Smoother", SmooFact);
-
-    bool r; // cf. bug Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2)); TEST_EQUALITY(r, false);
-    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null ); TEST_EQUALITY(r, true);
-
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
-    RCP<Level> l2 = H.GetLevel(2);
-
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);
-      l2->print(*stdout,Teuchos::VERB_EXTREME);*/
-
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), false); // no pre-smooth
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), false); // no pre-smooth
-    TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 299 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+  H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  RCP<FactoryBase> P = rcp(new TentativePFactory());
+  M1.SetFactory("P", P);
+  M1.SetFactory("Ptent", P);  // FIXME: can it be done automatically in FactoryManager?
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  // Here we want no pre-smoothing on levels 0 and 1
+  RCP<SmootherPrototype> noSmooProto;
+  RCP<SmootherPrototype> smooProto = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSmootherPrototype("Gauss-Seidel", 2);
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(noSmooProto, smooProto));
+  M0.SetFactory("Smoother", SmooFact);
+  M1.SetFactory("Smoother", SmooFact);
+
+  bool r;  // cf. bug Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+  RCP<Level> l2 = H.GetLevel(2);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);
+    l2->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), false);  // no pre-smooth
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), false);  // no pre-smooth
+  TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3levelFacManagers, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra)
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    X->putScalar( (Scalar) 0.0);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 299 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.SetMaxCoarseSize(50);
+  H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  // setup smoother factory
+  RCP<SmootherPrototype> smooProto;
+  std::string ifpackType;
+  Teuchos::ParameterList ifpackList;
+  ifpackList.set("relaxation: sweeps", (LocalOrdinal)2);
+  ifpackList.set("relaxation: damping factor", (Scalar)0.9);  // 0.7
+  ifpackType = "RELAXATION";
+  ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smooProto = Teuchos::rcp(new TrilinosSmoother(ifpackType, ifpackList));
+  TEUCHOS_TEST_FOR_EXCEPTION(smooProto == Teuchos::null, MueLu::Exceptions::Incompatible, "MueLu: UnitTest::Hierarchy::SetupHierarchy3levelFacManagers: Dynamic cast to IfpackSmoother failed. Epetra needs Scalar=double, LocalOrdinal=GlobalOrdinal=int and Node=Serial.");
+
+  RCP<SmootherFactory> preSmooFact;
+  RCP<SmootherFactory> postSmooFact;
+  preSmooFact  = rcp(new SmootherFactory(smooProto));
+  postSmooFact = rcp(new SmootherFactory(smooProto));
+  preSmooFact->SetSmootherPrototypes(smooProto, Teuchos::null);
+  postSmooFact->SetSmootherPrototypes(Teuchos::null, smooProto);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+  M0.SetFactory("Smoother", preSmooFact);
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  RCP<FactoryBase> PFact = rcp(new TentativePFactory());
+  M1.SetFactory("P", PFact);
+  M1.SetFactory("Ptent", PFact);  // FIXME: can it be done automatically in FactoryManager?
+  M1.SetFactory("Smoother", postSmooFact);
+
+  FactoryManager M2;  // last level (SA)
+  M2.SetKokkosRefactor(false);
+  M2.SetFactory("A", rcp(new RAPFactory()));
+  M2.SetFactory("P", rcp(new SaPFactory()));
+
+  bool r;  // cf. bug Teuchos Bug 5214
+  H.EnableGraphDumping("hierarchy_test_graph", 0);
+
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));
+  TEST_EQUALITY(r, false);
+  r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);
+  TEST_EQUALITY(r, true);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  RCP<Level> l1 = H.GetLevel(1);
+  RCP<Level> l2 = H.GetLevel(2);
+
+  /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+    l0->print(*stdout,Teuchos::VERB_EXTREME);
+    l1->print(*stdout,Teuchos::VERB_EXTREME);
+    l2->print(*stdout,Teuchos::VERB_EXTREME);*/
+
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchyTestBreakCondition, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
-  }
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 299 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Multigrid Hierarchy
+  Hierarchy H(A);
+  H.SetMaxCoarseSize(299 * comm->getSize());  // set max coarse size to fit problem size (-> 1 level method)
+  H.setVerbLevel(Teuchos::VERB_HIGH);
+  H.GetLevel(0)->Set("Coordinates", coordinates);
+
+  // Multigrid setup phase (using default parameters)
+  FactoryManager M0;  // how to build aggregates and smoother of the first level
+  M0.SetKokkosRefactor(false);
+
+  FactoryManager M1;  // first coarse level (Plain aggregation)
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", rcp(new RAPFactory()));
+  M1.SetFactory("P", rcp(new TentativePFactory()));
+
+  bool r;  // cf. bug Teuchos Bug 5214
+  r = H.Setup(0, Teuchos::null, rcpFromRef(M0), rcpFromRef(M1));
+  TEST_EQUALITY(r, true);
+  TEST_EQUALITY(H.GetNumLevels(), 1);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // direct solve
+  TEST_EQUALITY(l0->IsAvailable("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsAvailable("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  TEST_EQUALITY(l0->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(l0->GetKeepFlag("PostSmoother",  MueLu::NoFactory::get()), MueLu::Final); //direct solve
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+
+  X->putScalar((Scalar)0.0);
+
+  int iterations = 10;
+  H.Iterate(*RHS, *X, iterations);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testMatrices(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
+             RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B,
+             Teuchos::FancyOStream& out,
+             std::string labelA = "A",
+             std::string labelB = "B") {
+#include <MueLu_UseShortNames.hpp>
+  using TST             = Teuchos::ScalarTraits<Scalar>;
+  RCP<Vector> randomVec = VectorFactory::Build(A->getDomainMap(), false);
+  randomVec->randomize();
+  out << "randomVec norm: " << randomVec->norm2() << std::endl;
+  RCP<Vector> A_v = VectorFactory::Build(A->getRangeMap(), false);
+  A->apply(*randomVec, *A_v, Teuchos::NO_TRANS, 1, 0);
+  out << labelA << "_v norm: " << A_v->norm2() << std::endl;
+
+  RCP<Vector> B_v = VectorFactory::Build(B->getRangeMap(), false);
+  B->apply(*randomVec, *B_v, Teuchos::NO_TRANS, 1, 0);
+  out << labelB << "_v norm: " << B_v->norm2() << std::endl;
+
+  RCP<MultiVector> diff = VectorFactory::Build(A->getRangeMap());
+  // diff = A_v + (-1.0)*(B_v) + 0*diff
+  diff->update(1.0, *A_v, -1.0, *B_v, 0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> norms(1);
+  diff->norm2(norms);
+  out << "||diff|| = " << norms[0] << std::endl;
+  return norms[0];
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+loadMatrix(const std::string& basename, Xpetra::UnderlyingLib lib, const RCP<const Teuchos::Comm<int> >& comm) {
+#include <MueLu_UseShortNames.hpp>
+
+  RCP<const Map> rowMap, colMap, domMap, ranMap;
+  std::string infile;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchy3levelFacManagers, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-    //MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra)
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 299*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.SetMaxCoarseSize(50);
-    H.GetLevel(0)->Set("Coordinates", coordinates);
-
-    // setup smoother factory
-    RCP<SmootherPrototype> smooProto;
-    std::string ifpackType;
-    Teuchos::ParameterList ifpackList;
-    ifpackList.set("relaxation: sweeps", (LocalOrdinal) 2);
-    ifpackList.set("relaxation: damping factor", (Scalar) 0.9); // 0.7
-    ifpackType = "RELAXATION";
-    ifpackList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smooProto = Teuchos::rcp( new TrilinosSmoother(ifpackType, ifpackList) );
-    TEUCHOS_TEST_FOR_EXCEPTION(smooProto == Teuchos::null, MueLu::Exceptions::Incompatible, "MueLu: UnitTest::Hierarchy::SetupHierarchy3levelFacManagers: Dynamic cast to IfpackSmoother failed. Epetra needs Scalar=double, LocalOrdinal=GlobalOrdinal=int and Node=Serial.");
-
-    RCP<SmootherFactory> preSmooFact;
-    RCP<SmootherFactory> postSmooFact;
-    preSmooFact = rcp( new SmootherFactory(smooProto) );
-    postSmooFact = rcp( new SmootherFactory(smooProto) );
-    preSmooFact->SetSmootherPrototypes(smooProto,Teuchos::null );
-    postSmooFact->SetSmootherPrototypes(Teuchos::null,smooProto);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-    M0.SetFactory("Smoother", preSmooFact);
-
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    RCP<FactoryBase> PFact = rcp(new TentativePFactory());
-    M1.SetFactory("P", PFact);
-    M1.SetFactory("Ptent", PFact); //FIXME: can it be done automatically in FactoryManager?
-    M1.SetFactory("Smoother", postSmooFact);
-
-    FactoryManager M2; // last level (SA)
-    M2.SetKokkosRefactor(false);
-    M2.SetFactory("A", rcp(new RAPFactory()));
-    M2.SetFactory("P", rcp(new SaPFactory()));
-
-    bool r; // cf. bug Teuchos Bug 5214
-    H.EnableGraphDumping("hierarchy_test_graph",0);
-
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, false);
-    r = H.Setup(1, rcpFromRef(M0), rcpFromRef(M1), rcpFromRef(M2));   TEST_EQUALITY(r, false);
-    r = H.Setup(2, rcpFromRef(M1), rcpFromRef(M2), Teuchos::null);  TEST_EQUALITY(r, true);
-
-    RCP<Level> l0 = H.GetLevel(0);
-    RCP<Level> l1 = H.GetLevel(1);
-    RCP<Level> l2 = H.GetLevel(2);
-
-    /*RCP<Teuchos::FancyOStream> stdout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-      l0->print(*stdout,Teuchos::VERB_EXTREME);
-      l1->print(*stdout,Teuchos::VERB_EXTREME);
-      l2->print(*stdout,Teuchos::VERB_EXTREME);*/
-
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final); // direct solve
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-
-    X->putScalar( (Scalar) 0.0);
-
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
+    infile = "rowmap_" + basename + ".m";
+    std::ifstream ifile(infile);
+    if (ifile)
+      rowMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(infile, lib, comm);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, SetupHierarchyTestBreakCondition, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx  = 299*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Multigrid Hierarchy
-    Hierarchy H(A);
-    H.SetMaxCoarseSize(299*comm->getSize()); // set max coarse size to fit problem size (-> 1 level method)
-    H.setVerbLevel(Teuchos::VERB_HIGH);
-    H.GetLevel(0)->Set("Coordinates", coordinates);
-
-    // Multigrid setup phase (using default parameters)
-    FactoryManager M0; // how to build aggregates and smoother of the first level
-    M0.SetKokkosRefactor(false);
-
-    FactoryManager M1; // first coarse level (Plain aggregation)
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A", rcp(new RAPFactory()));
-    M1.SetFactory("P", rcp(new TentativePFactory()));
-
-    bool r; // cf. bug Teuchos Bug 5214
-    r = H.Setup(0, Teuchos::null,  rcpFromRef(M0), rcpFromRef(M1)); TEST_EQUALITY(r, true);
-    TEST_EQUALITY(H.GetNumLevels(),1);
-
-    RCP<Level> l0 = H.GetLevel(0);
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false); // direct solve
-    TEST_EQUALITY(l0->IsAvailable("P",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsAvailable("R",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    TEST_EQUALITY(l0->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(l0->GetKeepFlag("PostSmoother",  MueLu::NoFactory::get()), MueLu::Final); //direct solve
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-
-    X->putScalar( (Scalar) 0.0);
-
-    int iterations=10;
-    H.Iterate(*RHS, *X, iterations);
+    infile = "colmap_" + basename + ".m";
+    std::ifstream ifile(infile);
+    if (ifile)
+      colMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(infile, lib, comm);
   }
 
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-  testMatrices(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
-               RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B,
-               Teuchos::FancyOStream& out,
-               std::string labelA="A",
-               std::string labelB="B") {
-#   include <MueLu_UseShortNames.hpp>
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    RCP<Vector> randomVec = VectorFactory::Build(A->getDomainMap(),false);
-    randomVec->randomize();
-    out << "randomVec norm: " << randomVec->norm2() << std::endl;
-    RCP<Vector> A_v = VectorFactory::Build(A->getRangeMap(),false);
-    A->apply(*randomVec,*A_v,Teuchos::NO_TRANS,1,0);
-    out << labelA << "_v norm: " << A_v->norm2() << std::endl;
-
-    RCP<Vector> B_v = VectorFactory::Build(B->getRangeMap(),false);
-    B->apply(*randomVec,*B_v,Teuchos::NO_TRANS,1,0);
-    out << labelB << "_v norm: " << B_v->norm2() << std::endl;
-
-    RCP<MultiVector> diff = VectorFactory::Build(A->getRangeMap());
-    //diff = A_v + (-1.0)*(B_v) + 0*diff
-    diff->update(1.0,*A_v,-1.0,*B_v,0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> norms(1);
-    diff->norm2(norms);
-    out << "||diff|| = " << norms[0] << std::endl;
-    return norms[0];
-  }
-
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-  loadMatrix(const std::string& basename, Xpetra::UnderlyingLib lib, const RCP<const Teuchos::Comm<int> >& comm)
   {
-#   include <MueLu_UseShortNames.hpp>
-
-    RCP<const Map> rowMap, colMap, domMap, ranMap;
-    std::string infile;
-
-    {
-      infile = "rowmap_" + basename + ".m";
-      std::ifstream ifile(infile);
-      if (ifile)
-        rowMap = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::ReadMap(infile, lib, comm);
-    }
-
-    {
-      infile = "colmap_" + basename + ".m";
-      std::ifstream ifile(infile);
-      if (ifile)
-        colMap = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::ReadMap(infile, lib, comm);
-    }
-
-    {
-      infile = "domainmap_" + basename + ".m";
-      std::ifstream ifile(infile);
-      if (ifile)
-        domMap = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::ReadMap(infile, lib, comm);
-    }
-
-    {
-      infile = "rangemap_" + basename + ".m";
-      std::ifstream ifile(infile);
-      if (ifile)
-        ranMap = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::ReadMap(infile, lib, comm);
-    }
-
-    infile = basename + ".m";
-    RCP<Matrix> A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read(infile, rowMap, colMap, domMap, ranMap);
-
-    return A;
+    infile = "domainmap_" + basename + ".m";
+    std::ifstream ifile(infile);
+    if (ifile)
+      domMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(infile, lib, comm);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Write, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    Teuchos::RCP<Teuchos::FancyOStream> allOut = Teuchos::rcp(new Teuchos::FancyOStream(Teuchos::rcpFromRef(std::cout)));
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO nx = 30;
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    Teuchos::ParameterList paramList;
-    paramList.set("repartition: enable", true);
-    paramList.set("repartition: start level", 1);
-    paramList.set("repartition: min rows per proc", 6);
-    paramList.set("coarse: max size", 29);
-    //    paramList.sublist("user data").set("Node Comm",nodeComm);
-    paramList.set("verbosity", "high");
-    RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter(paramList));
-
-    // Multigrid Hierarchy
-    RCP<Hierarchy> H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A", A);
-    H->GetLevel(0)->Set("Coordinates", coordinates);
-    mueLuFactory->SetupHierarchy(*H);
-
-    TEST_THROW( H->Write(1,0), MueLu::Exceptions::RuntimeError );    //start level is greater than end level
-    TEST_THROW( H->Write(0,1000), MueLu::Exceptions::RuntimeError ); //end level is too big
-
-    // Write matrices out, read fine A back in, and check that the read was ok
-    // by using a matvec with a random vector.
-    char t[] = "XXXXXX";
-    int filedescriptor = 0;
-    if (comm->getRank() == 0)
-      filedescriptor = mkstemp(t); //mkstemp() creates a temporary file. We use the name of that file as
-                       //the suffix for the various data files produced by Hierarchy::Write().
-                       //A better solution would be to write to a file stream, but this would
-                       //involve writing new interfaces to Epetra's file I/O capabilities.
-    TEST_INEQUALITY(filedescriptor, -1); // if the file descriptor is -1, it failed
-    std::string tname(t);
-    Teuchos::broadcast<int, char>(*comm, 0, tname.size(), &tname[0]);
-    LocalOrdinal zero = Teuchos::OrdinalTraits<LocalOrdinal>::zero();
-    LocalOrdinal one = Teuchos::OrdinalTraits<LocalOrdinal>::one();
-    //Only write out the fine level matrix, since that is the only data file we test against.
-    H->Write(zero,one,tname);
-
-    comm->barrier();
-
-    RCP<Level> lvl = H->GetLevel(one);
-
-    std::string infile;
-    magnitude_type diff;
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    {
-      std::string basename = "A_0" + tname;
-      RCP<Matrix> A0in = loadMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(basename, lib, comm);
-
-      infile = "colmap_" + basename + ".m";    remove(infile.c_str());
-      infile = "domainmap_" + basename + ".m"; remove(infile.c_str());
-      infile = "rangemap_" + basename + ".m";  remove(infile.c_str());
-      infile = "rowmap_" + basename + ".m";    remove(infile.c_str());
-      infile = basename + ".m";                remove(infile.c_str());
-
-      diff = testMatrices(A, A0in, out, "A0", "A0in");
-    }
-    TEST_EQUALITY(diff < 100*TMT::eps(), true);
-    comm->barrier();
-
-    RCP<Operator> A1 = lvl->Get< RCP<Operator> >("A");
-    if (!A1.is_null()) {
-      RCP<Matrix> A1m = rcp_dynamic_cast<Matrix>(A1, true);
-
-      std::string basename = "A_1" + tname;
-      RCP<Matrix> A1in = loadMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(basename, lib, A1m->getRangeMap()->getComm());
-
-      infile = "colmap_" + basename + ".m";    remove(infile.c_str());
-      infile = "domainmap_" + basename + ".m"; remove(infile.c_str());
-      infile = "rangemap_" + basename + ".m";  remove(infile.c_str());
-      infile = "rowmap_" + basename + ".m";    remove(infile.c_str());
-      infile = basename + ".m";                remove(infile.c_str());
-
-      diff = testMatrices(A1m, A1in, *allOut, "A1", "A1in");
-    }
-    TEST_EQUALITY(diff < 100*TMT::eps(), true);
-    comm->barrier();
-
-    RCP<Operator> P1 = lvl->Get< RCP<Operator> >("P");
-    if (!P1.is_null()) {
-      RCP<Matrix> P1m = rcp_dynamic_cast<Matrix>(P1, true);
-
-      std::string basename = "P_1" + tname;
-      RCP<Matrix> P1in = loadMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(basename, lib, P1m->getRangeMap()->getComm());
-
-      infile = "colmap_" + basename + ".m";    remove(infile.c_str());
-      infile = "domainmap_" + basename + ".m"; remove(infile.c_str());
-      infile = "rangemap_" + basename + ".m";  remove(infile.c_str());
-      infile = "rowmap_" + basename + ".m";    remove(infile.c_str());
-      infile = basename + ".m";                remove(infile.c_str());
-
-      diff = testMatrices(P1m, P1in, *allOut, "P1", "P1in");
-    }
-    TEST_EQUALITY(diff < 100*TMT::eps(), true);
-    comm->barrier();
-
-    RCP<Operator> R1 = lvl->Get< RCP<Operator> >("R");
-    if (!R1.is_null()) {
-      RCP<Matrix> R1m = rcp_dynamic_cast<Matrix>(R1, true);
-
-      std::string basename = "R_1" + tname;
-      RCP<Matrix> R1in = loadMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>(basename, lib, R1m->getRangeMap()->getComm());
-
-      infile = "colmap_" + basename + ".m";    remove(infile.c_str());
-      infile = "domainmap_" + basename + ".m"; remove(infile.c_str());
-      infile = "rangemap_" + basename + ".m";  remove(infile.c_str());
-      infile = "rowmap_" + basename + ".m";    remove(infile.c_str());
-      infile = basename + ".m";                remove(infile.c_str());
-
-      diff = testMatrices(R1m, R1in, *allOut, "R1", "R1in");
-    }
-    TEST_EQUALITY(diff < 100*TMT::eps(), true);
-
-    remove(tname.c_str()); //remove file created by mkstemp
-
+    infile = "rangemap_" + basename + ".m";
+    std::ifstream ifile(infile);
+    if (ifile)
+      ranMap = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReadMap(infile, lib, comm);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, BlockCrs_Mixed, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-#if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS2)
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
+  infile        = basename + ".m";
+  RCP<Matrix> A = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read(infile, rowMap, colMap, domMap, ranMap);
 
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  return A;
+}
 
-    out << "===== Generating matrices =====" << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    Teuchos::ParameterList matrixList;
-    matrixList.set<GO>("nx",          100);
-    matrixList.set<GO>("ny",          100);
-    matrixList.set("matrixType",  "Laplace2D");
-
-    // Construct block matrix
-    RCP<Matrix> A = TestHelpers::TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildBlockMatrix(matrixList,Xpetra::UseTpetra);
-    if(A==Teuchos::null) { // if A is Teuchos::null, we could not build the matrix as it is not instantiated in Tpetra
-      out << "Skipping test" << std::endl;
-      return;
-    }
-
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("2D", A->getRowMap(), matrixList);
-
-    // extract information
-    RCP<const Map>        rangeMap = A->getRangeMap();
-    Xpetra::UnderlyingLib lib      = rangeMap->lib();
-
-    // Construct our own point operators for level 1
-    matrixList.set("nx", matrixList.get<GO>("nx")*3);
-    matrixList.set("ny", matrixList.get<GO>("ny"));
-    RCP<Matrix> A_point = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList, lib);;
-
-    out << "===== Setting up helper hierarchy =====" << std::endl;
-
-    // Setup helper hierarchy
-    Hierarchy Haux(A_point);
-    Haux.SetMaxCoarseSize(1);
-    Haux.SetDefaultVerbLevel(MueLu::None);
-    Haux.GetLevel(0)->Set("Coordinates", coordinates);
-
-    FactoryManager Maux;
-    Maux.SetKokkosRefactor(false);
-    const FactoryBase* nullFactory  = Maux.GetFactory("Nullspace").get();
-
-    Haux.Keep("Nullspace", nullFactory);
-
-    // Level 1 nullspace is only constructed/requested by level 2 factories. So
-    // we need at least three level hierarchy here, unless we somehow call
-    // NullspaceFactory::Build after hierarchy setup
-    Haux.Setup(Maux, 0, 3);
-    TEST_EQUALITY(Haux.GetNumLevels(), 3);
-
-    // Build block SGS smoother
-    std::string ifpack2Type;
-    Teuchos::ParameterList ifpack2List;
-    ifpack2Type = "RBILUK";
-    out << ifpack2Type << std::endl;
-    RCP<SmootherPrototype> smooProto = Teuchos::rcp( new Ifpack2Smoother(ifpack2Type) );
-    RCP<SmootherFactory>   SmooFact  = rcp( new SmootherFactory(smooProto) );
-
-    // Setup hierarchy managers
-    // We need three managers due to restriction in HierarchyHelpers: they
-    // request R, A, and P for coarse level even if User provides them The quick
-    // fix proposed here is to simply set corresponding fatories to NoFactory, so
-    // they would fetch user data.
-    FactoryManager M0, M1, M2;
-    M0.SetKokkosRefactor(false);
-    M0.SetFactory("Smoother", SmooFact);
-    M1.SetKokkosRefactor(false);
-    M1.SetFactory("A",        MueLu::NoFactory::getRCP());
-    M1.SetFactory("P",        MueLu::NoFactory::getRCP());
-    M1.SetFactory("R",        MueLu::NoFactory::getRCP());
-    M2.SetKokkosRefactor(false);
-
-    out << "===== Setting up mixed hierarchy =====" << std::endl;
-
-    // Setup mixed hierarchy
-    HierarchyManager mueluManager;
-    mueluManager.AddFactoryManager(0, 1, Teuchos::rcpFromRef(M0));
-    mueluManager.AddFactoryManager(1, 1, Teuchos::rcpFromRef(M1));
-    mueluManager.AddFactoryManager(2, 1, Teuchos::rcpFromRef(M2));
-
-    RCP<Hierarchy> Hrcp = mueluManager.CreateHierarchy();
-    Hierarchy&     H    = *Hrcp;
-    H.SetDefaultVerbLevel(MueLu::Low | MueLu::Debug);
-
-    RCP<Level> l0 = H.GetLevel(0);
-    l0->Set("A",           A);
-    H.AddNewLevel();
-    RCP<Level> l1     = H   .GetLevel(1);
-    RCP<Level> l1_aux = Haux.GetLevel(1);
-    l1->Set("P",          l1_aux->Get<RCP<Operator> >   ("P"));
-    l1->Set("R",          l1_aux->Get<RCP<Operator> >   ("R"));
-    l1->Set("A",          l1_aux->Get<RCP<Matrix> >     ("A"));
-    l1->Set("Nullspace",  l1_aux->Get<RCP<MultiVector> >("Nullspace", nullFactory));
-
-    mueluManager.SetupHierarchy(H);
-
-    out << "===== Solving =====" << std::endl;
-
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
-    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
-    RHS->setSeed(846930886);
-    RHS->randomize();
-    X->putScalar( Teuchos::ScalarTraits<Scalar>::zero());
-
-    int iterations = 10;
-    real_type tol = 1.0e-10;
-    H.IsPreconditioner(false);
-    H.Iterate(*RHS, *X, std::pair<LO,real_type>(iterations,tol));
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, Write, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  Teuchos::RCP<Teuchos::FancyOStream> allOut = Teuchos::rcp(new Teuchos::FancyOStream(Teuchos::rcpFromRef(std::cout)));
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
 #endif
-    TEST_EQUALITY(0,0);
-  }
+  using TST                   = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type        = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  GO nx                               = 30;
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  Teuchos::ParameterList paramList;
+  paramList.set("repartition: enable", true);
+  paramList.set("repartition: start level", 1);
+  paramList.set("repartition: min rows per proc", 6);
+  paramList.set("coarse: max size", 29);
+  //    paramList.sublist("user data").set("Node Comm",nodeComm);
+  paramList.set("verbosity", "high");
+  RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter(paramList));
+
+  // Multigrid Hierarchy
+  RCP<Hierarchy> H = mueLuFactory->CreateHierarchy();
+  H->GetLevel(0)->Set("A", A);
+  H->GetLevel(0)->Set("Coordinates", coordinates);
+  mueLuFactory->SetupHierarchy(*H);
+
+  TEST_THROW(H->Write(1, 0), MueLu::Exceptions::RuntimeError);     // start level is greater than end level
+  TEST_THROW(H->Write(0, 1000), MueLu::Exceptions::RuntimeError);  // end level is too big
+
+  // Write matrices out, read fine A back in, and check that the read was ok
+  // by using a matvec with a random vector.
+  char t[]           = "XXXXXX";
+  int filedescriptor = 0;
+  if (comm->getRank() == 0)
+    filedescriptor = mkstemp(t);        // mkstemp() creates a temporary file. We use the name of that file as
+                                        // the suffix for the various data files produced by Hierarchy::Write().
+                                        // A better solution would be to write to a file stream, but this would
+                                        // involve writing new interfaces to Epetra's file I/O capabilities.
+  TEST_INEQUALITY(filedescriptor, -1);  // if the file descriptor is -1, it failed
+  std::string tname(t);
+  Teuchos::broadcast<int, char>(*comm, 0, tname.size(), &tname[0]);
+  LocalOrdinal zero = Teuchos::OrdinalTraits<LocalOrdinal>::zero();
+  LocalOrdinal one  = Teuchos::OrdinalTraits<LocalOrdinal>::one();
+  // Only write out the fine level matrix, since that is the only data file we test against.
+  H->Write(zero, one, tname);
+
+  comm->barrier();
+
+  RCP<Level> lvl = H->GetLevel(one);
+
+  std::string infile;
+  magnitude_type diff;
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, CheckNullspaceDimension, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-    // Test that HierarchyManager throws if user-supplied nullspace has dimension smaller than numPDEs
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+    std::string basename = "A_0" + tname;
+    RCP<Matrix> A0in     = loadMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(basename, lib, comm);
+
+    infile = "colmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "domainmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rangemap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rowmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = basename + ".m";
+    remove(infile.c_str());
 
-    GO nx = 30;
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
-    A->SetFixedBlockSize(2);
+    diff = testMatrices(A, A0in, out, "A0", "A0in");
+  }
+  TEST_EQUALITY(diff < 100 * TMT::eps(), true);
+  comm->barrier();
+
+  RCP<Operator> A1 = lvl->Get<RCP<Operator> >("A");
+  if (!A1.is_null()) {
+    RCP<Matrix> A1m = rcp_dynamic_cast<Matrix>(A1, true);
+
+    std::string basename = "A_1" + tname;
+    RCP<Matrix> A1in     = loadMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(basename, lib, A1m->getRangeMap()->getComm());
+
+    infile = "colmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "domainmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rangemap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rowmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = basename + ".m";
+    remove(infile.c_str());
 
-    HierarchyManager mueluManager;
-    RCP<Hierarchy> Hrcp = mueluManager.CreateHierarchy();
-    Hierarchy&     H    = *Hrcp;
-    H.SetDefaultVerbLevel(MueLu::Low | MueLu::Debug);
+    diff = testMatrices(A1m, A1in, *allOut, "A1", "A1in");
+  }
+  TEST_EQUALITY(diff < 100 * TMT::eps(), true);
+  comm->barrier();
+
+  RCP<Operator> P1 = lvl->Get<RCP<Operator> >("P");
+  if (!P1.is_null()) {
+    RCP<Matrix> P1m = rcp_dynamic_cast<Matrix>(P1, true);
+
+    std::string basename = "P_1" + tname;
+    RCP<Matrix> P1in     = loadMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(basename, lib, P1m->getRangeMap()->getComm());
+
+    infile = "colmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "domainmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rangemap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rowmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = basename + ".m";
+    remove(infile.c_str());
 
-    RCP<Level> l0 = H.GetLevel(0);
-    l0->Set("A", A);
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    l0->Set("Nullspace", nullSpace);
+    diff = testMatrices(P1m, P1in, *allOut, "P1", "P1in");
+  }
+  TEST_EQUALITY(diff < 100 * TMT::eps(), true);
+  comm->barrier();
+
+  RCP<Operator> R1 = lvl->Get<RCP<Operator> >("R");
+  if (!R1.is_null()) {
+    RCP<Matrix> R1m = rcp_dynamic_cast<Matrix>(R1, true);
+
+    std::string basename = "R_1" + tname;
+    RCP<Matrix> R1in     = loadMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(basename, lib, R1m->getRangeMap()->getComm());
+
+    infile = "colmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "domainmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rangemap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = "rowmap_" + basename + ".m";
+    remove(infile.c_str());
+    infile = basename + ".m";
+    remove(infile.c_str());
 
-    TEST_THROW( mueluManager.SetupHierarchy(H), MueLu::Exceptions::RuntimeError );
+    diff = testMatrices(R1m, R1in, *allOut, "R1", "R1in");
   }
+  TEST_EQUALITY(diff < 100 * TMT::eps(), true);
 
+  remove(tname.c_str());  // remove file created by mkstemp
+}
 
-# define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, DescriptionCaching, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetAndGetLevel, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, GetNumLevels, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, KeepAggregates, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Iterate, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, IterateWithImplicitRestriction, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy1level, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy1levelv2, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level_AggregateSmooth, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level_AggregateSmooth_3PDEs, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_BlockTriDi_3PDEs, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_BlockSmooth, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_NoPreSmooth, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3levelFacManagers, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchyTestBreakCondition, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Write, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, BlockCrs_Mixed, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, CheckNullspaceDimension, Scalar, LO, GO, Node) \
-    
-
-# include <MueLu_ETI_4arg.hpp>
-
-}//namespace MueLuTests
-
-//Note from JG:
-// For UnitTest,  TEST_EQUALITY(H.GetLevel(1)->Get< RCP<Matrix> >("PreSmoother")->GetType(), "Ifpack: Gauss-Seidel");
-// should be replaced by
-// TEST_EQUALITY(H.GetLevel(1)->Get< RCP<Matrix> >("PreSmoother"), preSmoother);
-// testing if preSmoother->GetType() == "Ifpack: Gauss-Seidel" should be a unit test of the class IfpackSmoother
-
-
-//TODO unit test:
-// test if Hierarchy::Iterate(X,Y) works when X == Y (ie: do we need to test explicitly if X==Y and make a temporary copy inside of Iterate() ?)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, BlockCrs_Mixed, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+#if defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS2)
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  out << "===== Generating matrices =====" << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  Teuchos::ParameterList matrixList;
+  matrixList.set<GO>("nx", 100);
+  matrixList.set<GO>("ny", 100);
+  matrixList.set("matrixType", "Laplace2D");
+
+  // Construct block matrix
+  RCP<Matrix> A = TestHelpers::TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildBlockMatrix(matrixList, Xpetra::UseTpetra);
+  if (A == Teuchos::null) {  // if A is Teuchos::null, we could not build the matrix as it is not instantiated in Tpetra
+    out << "Skipping test" << std::endl;
+    return;
+  }
+
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("2D", A->getRowMap(), matrixList);
+
+  // extract information
+  RCP<const Map> rangeMap   = A->getRangeMap();
+  Xpetra::UnderlyingLib lib = rangeMap->lib();
+
+  // Construct our own point operators for level 1
+  matrixList.set("nx", matrixList.get<GO>("nx") * 3);
+  matrixList.set("ny", matrixList.get<GO>("ny"));
+  RCP<Matrix> A_point = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList, lib);
+  ;
+
+  out << "===== Setting up helper hierarchy =====" << std::endl;
+
+  // Setup helper hierarchy
+  Hierarchy Haux(A_point);
+  Haux.SetMaxCoarseSize(1);
+  Haux.SetDefaultVerbLevel(MueLu::None);
+  Haux.GetLevel(0)->Set("Coordinates", coordinates);
+
+  FactoryManager Maux;
+  Maux.SetKokkosRefactor(false);
+  const FactoryBase* nullFactory = Maux.GetFactory("Nullspace").get();
+
+  Haux.Keep("Nullspace", nullFactory);
+
+  // Level 1 nullspace is only constructed/requested by level 2 factories. So
+  // we need at least three level hierarchy here, unless we somehow call
+  // NullspaceFactory::Build after hierarchy setup
+  Haux.Setup(Maux, 0, 3);
+  TEST_EQUALITY(Haux.GetNumLevels(), 3);
+
+  // Build block SGS smoother
+  std::string ifpack2Type;
+  Teuchos::ParameterList ifpack2List;
+  ifpack2Type = "RBILUK";
+  out << ifpack2Type << std::endl;
+  RCP<SmootherPrototype> smooProto = Teuchos::rcp(new Ifpack2Smoother(ifpack2Type));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+
+  // Setup hierarchy managers
+  // We need three managers due to restriction in HierarchyHelpers: they
+  // request R, A, and P for coarse level even if User provides them The quick
+  // fix proposed here is to simply set corresponding fatories to NoFactory, so
+  // they would fetch user data.
+  FactoryManager M0, M1, M2;
+  M0.SetKokkosRefactor(false);
+  M0.SetFactory("Smoother", SmooFact);
+  M1.SetKokkosRefactor(false);
+  M1.SetFactory("A", MueLu::NoFactory::getRCP());
+  M1.SetFactory("P", MueLu::NoFactory::getRCP());
+  M1.SetFactory("R", MueLu::NoFactory::getRCP());
+  M2.SetKokkosRefactor(false);
+
+  out << "===== Setting up mixed hierarchy =====" << std::endl;
+
+  // Setup mixed hierarchy
+  HierarchyManager mueluManager;
+  mueluManager.AddFactoryManager(0, 1, Teuchos::rcpFromRef(M0));
+  mueluManager.AddFactoryManager(1, 1, Teuchos::rcpFromRef(M1));
+  mueluManager.AddFactoryManager(2, 1, Teuchos::rcpFromRef(M2));
+
+  RCP<Hierarchy> Hrcp = mueluManager.CreateHierarchy();
+  Hierarchy& H        = *Hrcp;
+  H.SetDefaultVerbLevel(MueLu::Low | MueLu::Debug);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  l0->Set("A", A);
+  H.AddNewLevel();
+  RCP<Level> l1     = H.GetLevel(1);
+  RCP<Level> l1_aux = Haux.GetLevel(1);
+  l1->Set("P", l1_aux->Get<RCP<Operator> >("P"));
+  l1->Set("R", l1_aux->Get<RCP<Operator> >("R"));
+  l1->Set("A", l1_aux->Get<RCP<Matrix> >("A"));
+  l1->Set("Nullspace", l1_aux->Get<RCP<MultiVector> >("Nullspace", nullFactory));
+
+  mueluManager.SetupHierarchy(H);
+
+  out << "===== Solving =====" << std::endl;
+
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+  RHS->setSeed(846930886);
+  RHS->randomize();
+  X->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
+
+  int iterations = 10;
+  real_type tol  = 1.0e-10;
+  H.IsPreconditioner(false);
+  H.Iterate(*RHS, *X, std::pair<LO, real_type>(iterations, tol));
+#endif
+  TEST_EQUALITY(0, 0);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Hierarchy, CheckNullspaceDimension, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // Test that HierarchyManager throws if user-supplied nullspace has dimension smaller than numPDEs
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  GO nx         = 30;
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(nx);
+  A->SetFixedBlockSize(2);
+
+  HierarchyManager mueluManager;
+  RCP<Hierarchy> Hrcp = mueluManager.CreateHierarchy();
+  Hierarchy& H        = *Hrcp;
+  H.SetDefaultVerbLevel(MueLu::Low | MueLu::Debug);
+
+  RCP<Level> l0 = H.GetLevel(0);
+  l0->Set("A", A);
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar((Scalar)1.0);
+  l0->Set("Nullspace", nullSpace);
+
+  TEST_THROW(mueluManager.SetupHierarchy(H), MueLu::Exceptions::RuntimeError);
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Constructor, Scalar, LO, GO, Node)                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, DescriptionCaching, Scalar, LO, GO, Node)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetAndGetLevel, Scalar, LO, GO, Node)                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, GetNumLevels, Scalar, LO, GO, Node)                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, KeepAggregates, Scalar, LO, GO, Node)                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Iterate, Scalar, LO, GO, Node)                                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, IterateWithImplicitRestriction, Scalar, LO, GO, Node)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy1level, Scalar, LO, GO, Node)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy1levelv2, Scalar, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level, Scalar, LO, GO, Node)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level_AggregateSmooth, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy2level_AggregateSmooth_3PDEs, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_BlockTriDi_3PDEs, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level, Scalar, LO, GO, Node)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_BlockSmooth, Scalar, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3level_NoPreSmooth, Scalar, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchy3levelFacManagers, Scalar, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, SetupHierarchyTestBreakCondition, Scalar, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, Write, Scalar, LO, GO, Node)                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, BlockCrs_Mixed, Scalar, LO, GO, Node)                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Hierarchy, CheckNullspaceDimension, Scalar, LO, GO, Node)
+
+#include <MueLu_ETI_4arg.hpp>
+
+}  // namespace MueLuTests
+
+// Note from JG:
+//  For UnitTest,  TEST_EQUALITY(H.GetLevel(1)->Get< RCP<Matrix> >("PreSmoother")->GetType(), "Ifpack: Gauss-Seidel");
+//  should be replaced by
+//  TEST_EQUALITY(H.GetLevel(1)->Get< RCP<Matrix> >("PreSmoother"), preSmoother);
+//  testing if preSmoother->GetType() == "Ifpack: Gauss-Seidel" should be a unit test of the class IfpackSmoother
+
+// TODO unit test:
+//  test if Hierarchy::Iterate(X,Y) works when X == Y (ie: do we need to test explicitly if X==Y and make a temporary copy inside of Iterate() ?)
diff --git a/packages/muelu/test/unit_tests/IndexManager.cpp b/packages/muelu/test/unit_tests/IndexManager.cpp
index c0ffc53cefd2..19f68ecd4cb0 100644
--- a/packages/muelu/test/unit_tests/IndexManager.cpp
+++ b/packages/muelu/test/unit_tests/IndexManager.cpp
@@ -54,856 +54,1506 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateGlobalLexicographicIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fout->setShowAllFrontMatter(false).setShowProcRank(true);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
-
-    // Set global geometric data
-    const bool coupled = true;
-    const int numDimensions = 3;
-    const int interpolationOrder = 0;
-    Array<GO> meshData;
-    Array<GO> gNodesPerDir(3);
-    Array<LO> lNodesPerDir(3);
-    Array<LO> coarseRate(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-        coarseRate[dim]   = 2;
-      } else {
-        gNodesPerDir[dim] = -1;
-        coarseRate[dim]   = 1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> coords =
-      MueLuTests::TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions,
-                                                                             gNodesPerDir,
-                                                                             lNodesPerDir,
-                                                                             meshData,
-                                                                             "Global Lexicographic");
-
-    RCP<MueLu::GlobalLexicographicIndexManager<LO,GO,NO> > myIndexManager =
-      rcp(new MueLu::GlobalLexicographicIndexManager<LO,GO,NO>(comm, coupled, numDimensions,
-                                                               interpolationOrder, gNodesPerDir,
-                                                               lNodesPerDir, coarseRate,
-                                                               coords->getMap()->getMinGlobalIndex()
-                                                               ));
-
-    Array<LO>  ghostedNodeCoarseLIDs;
-    Array<int> ghostedNodeCoarsePIDs;
-    Array<GO>  ghostedNodeCoarseGIDs;
-    myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
-                                        ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
-
-    Array<GO> coarseNodeCoarseGIDs;
-    Array<GO> coarseNodeFineGIDs;
-    myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
-                                       coarseNodeFineGIDs);
-
-    int chk = 0;
-    if(myIndexManager->isAggregationCoupled()       != coupled)                     {chk = -1;}
-    if(myIndexManager->getInterpolationOrder()      != interpolationOrder)          {chk = -1;}
-    if(myIndexManager->getNumDimensions()           != numDimensions)               {chk = -1;}
-    if(myIndexManager->getNumGlobalFineNodes()      != 125)                         {chk = -1;}
-    if(myIndexManager->getNumGlobalCoarseNodes()    != 27)                          {chk = -1;}
-    for(int dim = 0; dim < 3; ++dim) {
-      if(myIndexManager->getCoarseningRate(dim)         != coarseRate[dim])         {chk = -1;}
-      if(myIndexManager->getGlobalFineNodesInDir(dim)   != gNodesPerDir[dim])       {chk = -1;}
-      if(myIndexManager->getGlobalCoarseNodesInDir(dim) != 3)                       {chk = -1;}
-    }
-    if(comm->getSize() == 1) {
-      if(myIndexManager->getNumLocalFineNodes()       != 125)                       {chk = -1;}
-      if(myIndexManager->getNumLocalCoarseNodes()     != 27)                        {chk = -1;}
-      if(myIndexManager->getNumLocalGhostedNodes()    != 27)                        {chk = -1;}
-      for(int dim = 0; dim < 3; ++dim) {
-        if(myIndexManager->getLocalFineNodesInDir(dim)    != lNodesPerDir[dim])     {chk = -1;}
-        if(myIndexManager->getStartIndex(dim)             != 0)                     {chk = -1;}
-        if(myIndexManager->getStartIndex(dim + 3)         != gNodesPerDir[dim] - 1) {chk = -1;}
-        if(myIndexManager->getOffset(dim)                 != 0)                     {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim)         != false)                 {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim + 3)     != false)                 {chk = -1;}
-        if(myIndexManager->getLocalCoarseNodesInDir(dim)  != 3)                     {chk = -1;}
-        if(myIndexManager->getGhostedNodesInDir(dim)      != 3)                     {chk = -1;}
-      }
-      GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
-                        50, 52, 54, 60, 62, 64, 70, 72, 74,
-                        100, 102, 104, 110, 112, 114, 120, 122, 124};
-      for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-        if(ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(ghostedNodeCoarsePIDs[coarseIdx] != 0)                                   {chk = -1;}
-        if(ghostedNodeCoarseGIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(coarseNodeFineGIDs[coarseIdx]    != cnfGIDs[coarseIdx])                  {chk = -1;}
-        if(coarseNodeCoarseGIDs[coarseIdx]  != coarseIdx)                           {chk = -1;}
-      }
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(myIndexManager->getNumLocalFineNodes()       != 45)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,3,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {0,0,0,2,2,4}; GO offsets[3] = {0,0,0};
-        bool ghostInt[6] = {false, false, false, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-        GO  gncGIDs[12] = {0, 1, 3, 4, 9, 10, 12, 13, 18, 19, 21, 22};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {0, 2, 10, 12, 50, 52, 60, 62, 100, 102, 110, 112};
-        GO  cncGIDs[12] = {0, 1, 3, 4, 9, 10, 12, 13, 18, 19, 21, 22};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 1) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  6)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,3,5}; LO lCNPD[3] = {1,2,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {3,0,0,4,2,4}; GO offsets[3] = {1,0,0};
-        bool ghostInt[6] = {true, false, false, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {1, 0, 3, 1, 5, 2, 7, 3, 9, 4, 11, 5};
-        int gncPIDs[12] = {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1};
-        GO  gncGIDs[12] = {1, 2, 4, 5, 10, 11, 13, 14, 19, 20, 22, 23};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[6] = {4, 14, 54, 64, 104, 114};
-        GO  cncGIDs[6] = {2, 5, 11, 14, 20, 23};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 2) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  6)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,2,5}; LO lCNPD[3] = {2,1,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {0,3,0,2,4,4}; GO offsets[3] = {0,1,0};
-        bool ghostInt[6] = {false, false, true, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {2, 3, 0, 1, 6, 7, 2, 3, 10, 11, 4, 5};
-        int gncPIDs[12] = {0, 0, 2, 2, 0, 0, 2, 2, 0, 0, 2, 2};
-        GO  gncGIDs[12] = {3, 4, 6, 7, 12, 13, 15, 16, 21, 22, 24, 25};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[6] = {20, 22, 70, 72, 120, 122};
-        GO  cncGIDs[6] = {6, 7, 15, 16, 24, 25};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 3) {
-        if(myIndexManager->getNumLocalFineNodes()       != 20)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  3)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,2,5}; LO lCNPD[3] = {1,1,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {3,3,0,4,4,4}; GO offsets[3] = {1,1,0};
-        bool ghostInt[6] = {true, false, true, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {3, 1, 1, 0, 7, 3, 3, 1, 11, 5, 5, 2};
-        int gncPIDs[12] = {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3};
-        GO  gncGIDs[12] = {4, 5, 7, 8, 13, 14, 16, 17, 22, 23, 25, 26};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[3] = {24, 74, 124};
-        GO  cncGIDs[3] = {8, 17, 26};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      }
-    }
-
-    int gbl_chk[1] = {0};
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
-    TEST_EQUALITY(gbl_chk[0], 0);
-
-
-  } // CreateGlobalLexicographicIndexManager
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateLocalLexicographicIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fout->setShowAllFrontMatter(false).setShowProcRank(true);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
-
-    // Set global geometric data
-    const bool coupled = true;
-    const int numDimensions = 3;
-    const int interpolationOrder = 0;
-    Array<GO> meshData;
-    Array<GO> gNodesPerDir(3);
-    Array<LO> lNodesPerDir(3);
-    Array<LO> coarseRate(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-        coarseRate[dim]   = 2;
-      } else {
-        gNodesPerDir[dim] = -1;
-        coarseRate[dim]   =  1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> coords =
-      MueLuTests::TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions,
-                                                                             gNodesPerDir,
-                                                                             lNodesPerDir,
-                                                                             meshData,
-                                                                             "Local Lexicographic");
-
-    RCP<MueLu::LocalLexicographicIndexManager<LO,GO,NO> > myIndexManager =
-      rcp(new MueLu::LocalLexicographicIndexManager<LO,GO,NO>(comm, coupled, numDimensions,
-                                                              interpolationOrder, comm->getRank(),
-                                                              comm->getSize(), gNodesPerDir,
-                                                              lNodesPerDir, coarseRate, meshData));
-
-    Array<LO>  ghostedNodeCoarseLIDs;
-    Array<int> ghostedNodeCoarsePIDs;
-    Array<GO>  ghostedNodeCoarseGIDs;
-    myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
-                                        ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
-
-    Array<GO> coarseNodeCoarseGIDs;
-    Array<GO> coarseNodeFineGIDs;
-    myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
-                                       coarseNodeFineGIDs);
-
-    int chk = 0;
-    if(myIndexManager->isAggregationCoupled()       != coupled)                     {chk = -1;}
-    if(myIndexManager->getInterpolationOrder()      != interpolationOrder)          {chk = -1;}
-    if(myIndexManager->getNumDimensions()           != numDimensions)               {chk = -1;}
-    if(myIndexManager->getNumGlobalFineNodes()      != 125)                         {chk = -1;}
-    if(myIndexManager->getNumGlobalCoarseNodes()    != 27)                          {chk = -1;}
-    for(int dim = 0; dim < 3; ++dim) {
-      if(myIndexManager->getCoarseningRate(dim)         != coarseRate[dim])         {chk = -1;}
-      if(myIndexManager->getGlobalFineNodesInDir(dim)   != gNodesPerDir[dim])       {chk = -1;}
-      if(myIndexManager->getGlobalCoarseNodesInDir(dim) != 3)                       {chk = -1;}
-    }
-    if(comm->getSize() == 1) {
-      if(myIndexManager->getNumLocalFineNodes()       != 125)                       {chk = -1;}
-      if(myIndexManager->getNumLocalCoarseNodes()     != 27)                        {chk = -1;}
-      if(myIndexManager->getNumLocalGhostedNodes()    != 27)                        {chk = -1;}
-      for(int dim = 0; dim < 3; ++dim) {
-        if(myIndexManager->getLocalFineNodesInDir(dim)    != lNodesPerDir[dim])     {chk = -1;}
-        if(myIndexManager->getStartIndex(dim)             != 0)                     {chk = -1;}
-        if(myIndexManager->getStartIndex(dim + 3)         != gNodesPerDir[dim] - 1) {chk = -1;}
-        if(myIndexManager->getOffset(dim)                 != 0)                     {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim)         != false)                 {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim + 3)     != false)                 {chk = -1;}
-        if(myIndexManager->getLocalCoarseNodesInDir(dim)  != 3)                     {chk = -1;}
-        if(myIndexManager->getGhostedNodesInDir(dim)      != 3)                     {chk = -1;}
-      }
-      GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
-                        50, 52, 54, 60, 62, 64, 70, 72, 74,
-                        100, 102, 104, 110, 112, 114, 120, 122, 124};
-      for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-        if(ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(ghostedNodeCoarsePIDs[coarseIdx] != 0)                                   {chk = -1;}
-        if(ghostedNodeCoarseGIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(coarseNodeFineGIDs[coarseIdx]    != cnfGIDs[coarseIdx])                  {chk = -1;}
-        if(coarseNodeCoarseGIDs[coarseIdx]  != coarseIdx)                           {chk = -1;}
-      }
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(myIndexManager->getNumLocalFineNodes()       != 45)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,3,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {0,0,0,2,2,4}; GO offsets[3] = {0,0,0};
-        bool ghostInt[6] = {false, false, false, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-        GO  gncGIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {0, 2, 6, 8, 18, 20, 24, 26, 36, 38, 42, 44};
-        GO  cncGIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 1) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  6)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,3,5}; LO lCNPD[3] = {1,2,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {3,0,0,4,2,4}; GO offsets[3] = {1,0,0};
-        bool ghostInt[6] = {true, false, false, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {1, 0, 3, 1, 5, 2, 7, 3, 9, 4, 11, 5};
-        int gncPIDs[12] = {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1};
-        GO  gncGIDs[12] = {1, 12, 3, 13, 5, 14, 7, 15, 9, 16, 11, 17};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[6] = {46, 50, 58, 62, 70, 74};
-        GO  cncGIDs[6] = {12, 13, 14, 15, 16, 17};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 2) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  6)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,2,5}; LO lCNPD[3] = {2,1,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {0,3,0,2,4,4}; GO offsets[3] = {0,1,0};
-        bool ghostInt[6] = {false, false, true, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {2, 3, 0, 1, 6, 7, 2, 3, 10, 11, 4, 5};
-        int gncPIDs[12] = {0, 0, 2, 2, 0, 0, 2, 2,  0,  0, 2, 2};
-        GO  gncGIDs[12] = {2, 3, 18, 19, 6, 7, 20, 21, 10, 11, 22, 23};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[6] = {78, 80, 90, 92, 102, 104};
-        GO  cncGIDs[6] = {18, 19, 20 ,21, 22, 23};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 3) {
-        if(myIndexManager->getNumLocalFineNodes()       != 20)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  3)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,2,5}; LO lCNPD[3] = {1,1,3}; LO GNPD[3] = {2,2,3};
-        GO startIndex[6] = {3,3,0,4,4,4}; GO offsets[3] = {1,1,0};
-        bool ghostInt[6] = {true, false, true, false, false, false};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getStartIndex(dim)             != startIndex[dim])     {chk = -1;}
-          if(myIndexManager->getStartIndex(dim + 3)         != startIndex[dim + 3]) {chk = -1;}
-          if(myIndexManager->getOffset(dim)                 != offsets[dim])        {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != ghostInt[dim])       {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != ghostInt[dim + 3])   {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-        }
-        LO  gncLIDs[12] = {3, 1, 1, 0, 7, 3, 3, 1, 11, 5, 5, 2};
-        int gncPIDs[12] = {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3};
-        GO  gncGIDs[12] = {3, 13, 19, 24, 7, 15, 21, 25, 11, 17, 23, 26};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[3] = {108, 116, 124};
-        GO  cncGIDs[3] = {24, 25, 26};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-          if(coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx])                 {chk = -1;}
-        }
-      }
-    }
-
-    int gbl_chk[1] = {0};
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
-    TEST_EQUALITY(gbl_chk[0], 0);
-
-  } // CreateLocalLexicographicIndexManager
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateUncoupledIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fout->setShowAllFrontMatter(false).setShowProcRank(true);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
-
-    // Set global geometric data
-    const bool coupled = false;
-    const int numDimensions = 3;
-    const int interpolationOrder = 0;
-    Array<GO> meshData;
-    Array<GO> gNodesPerDir(3);
-    Array<LO> lNodesPerDir(3);
-    Array<LO> coarseRate(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-        coarseRate[dim]   = 2;
-      } else {
-        gNodesPerDir[dim] = -1;
-        coarseRate[dim]   =  1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> coords =
-      MueLuTests::TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions,
-                                                                             gNodesPerDir,
-                                                                             lNodesPerDir,
-                                                                             meshData,
-                                                                             "Local Lexicographic");
-
-    RCP<MueLu::UncoupledIndexManager<LO,GO,NO> > myIndexManager =
-      rcp(new MueLu::UncoupledIndexManager<LO,GO,NO>(comm, coupled, numDimensions,
-                                                     interpolationOrder, comm->getRank(),
-                                                     comm->getSize(), gNodesPerDir,
-                                                     lNodesPerDir, coarseRate, false));
-    myIndexManager->computeGlobalCoarseParameters();
-
-    Array<LO>  ghostedNodeCoarseLIDs;
-    Array<int> ghostedNodeCoarsePIDs;
-    Array<GO>  ghostedNodeCoarseGIDs;
-    myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
-                                        ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
-
-    Array<GO> coarseNodeCoarseGIDs;
-    Array<GO> coarseNodeFineGIDs;
-    myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
-                                       coarseNodeFineGIDs);
-
-    int chk = 0;
-    if(myIndexManager->isAggregationCoupled()       != coupled)                     {chk = -1;}
-    if(myIndexManager->getInterpolationOrder()      != interpolationOrder)          {chk = -1;}
-    if(myIndexManager->getNumDimensions()           != numDimensions)               {chk = -1;}
-    if(myIndexManager->getNumGlobalFineNodes()      != -1)                          {chk = -1;}
-    for(int dim = 0; dim < 3; ++dim) {
-      if(myIndexManager->getCoarseningRate(dim)         != coarseRate[dim])         {chk = -1;}
-    }
-    if(comm->getSize() == 1) {
-      if(myIndexManager->getNumLocalFineNodes()         != 125)                     {chk = -1;}
-      if(myIndexManager->getNumLocalCoarseNodes()       != 27)                      {chk = -1;}
-      if(myIndexManager->getNumLocalGhostedNodes()      != 27)                      {chk = -1;}
-      if(myIndexManager->getNumGlobalCoarseNodes()      != 27)                      {chk = -1;}
-      for(int dim = 0; dim < 3; ++dim) {
-        if(myIndexManager->getLocalFineNodesInDir(dim)    != lNodesPerDir[dim])     {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim)         != false)                 {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim + 3)     != false)                 {chk = -1;}
-        if(myIndexManager->getLocalCoarseNodesInDir(dim)  != 3)                     {chk = -1;}
-        if(myIndexManager->getGhostedNodesInDir(dim)      != 3)                     {chk = -1;}
-        if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                    {chk = -1;}
-      }
-      GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
-                        50, 52, 54, 60, 62, 64, 70, 72, 74,
-                        100, 102, 104, 110, 112, 114, 120, 122, 124};
-      for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-        if(ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(ghostedNodeCoarsePIDs[coarseIdx] != 0)                                   {chk = -1;}
-        if(coarseNodeFineGIDs[coarseIdx]    != cnfGIDs[coarseIdx])                  {chk = -1;}
-      }
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(myIndexManager->getNumLocalFineNodes()       != 45)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,3,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {0, 2, 6, 8, 18, 20, 24, 26, 36, 38, 42, 44};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 1) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,3,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  1};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {45, 46, 49, 50, 57, 58, 61, 62, 69, 70, 73, 74};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 2) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,2,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {75, 77, 78, 80, 87, 89, 90, 92, 99, 101, 102, 104};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 3) {
-        if(myIndexManager->getNumLocalFineNodes()       != 20)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,2,5}; LO lCNPD[3] = {2,2,3}; LO GNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
-        int gncPIDs[12] = {3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[12] = {105, 106, 107, 108, 113, 114, 115, 116, 121, 122, 123, 124};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      }
-    }
-
-    int gbl_chk[1] = {0};
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
-    TEST_EQUALITY(gbl_chk[0], 0);
-
-  } // CreateUncoupledIndexManager
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, UncoupledIndexManagerSingleCoarsePoint, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fout->setShowAllFrontMatter(false).setShowProcRank(true);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
-
-    // Set global geometric data
-    const bool coupled = false;
-    const int numDimensions = 3;
-    const int interpolationOrder = 0;
-    Array<GO> meshData;
-    Array<GO> gNodesPerDir(3);
-    Array<LO> lNodesPerDir(3);
-    Array<LO> coarseRate(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-        coarseRate[dim]   = 2;
-      } else {
-        gNodesPerDir[dim] = -1;
-        coarseRate[dim]   =  1;
-      }
-    }
-    // Change the number of nodes in the third
-    // direction to trigger singleCoarsePoint
-    gNodesPerDir[2] = 2;
-
-    RCP<RealValuedMultiVector> coords =
-      MueLuTests::TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions,
-                                                                             gNodesPerDir,
-                                                                             lNodesPerDir,
-                                                                             meshData,
-                                                                             "Local Lexicographic");
-
-    RCP<MueLu::UncoupledIndexManager<LO,GO,NO> > myIndexManager =
-      rcp(new MueLu::UncoupledIndexManager<LO,GO,NO>(comm, coupled, numDimensions,
-                                                     interpolationOrder, comm->getRank(),
-                                                     comm->getSize(), gNodesPerDir,
-                                                     lNodesPerDir, coarseRate, true));
-    myIndexManager->computeGlobalCoarseParameters();
-
-    Array<LO>  ghostedNodeCoarseLIDs;
-    Array<int> ghostedNodeCoarsePIDs;
-    Array<GO>  ghostedNodeCoarseGIDs;
-    myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
-                                        ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
-
-    Array<GO> coarseNodeCoarseGIDs;
-    Array<GO> coarseNodeFineGIDs;
-    myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
-                                       coarseNodeFineGIDs);
-
-    int chk = 0;
-    if(myIndexManager->isAggregationCoupled()       != coupled)                     {chk = -1;}
-    if(myIndexManager->getInterpolationOrder()      != interpolationOrder)          {chk = -1;}
-    if(myIndexManager->getNumDimensions()           != numDimensions)               {chk = -1;}
-    if(myIndexManager->getNumGlobalFineNodes()      != -1)                          {chk = -1;}
-    for(int dim = 0; dim < 3; ++dim) {
-      if(myIndexManager->getCoarseningRate(dim)         != coarseRate[dim])         {chk = -1;}
-    }
-    if(comm->getSize() == 1) {
-      if(myIndexManager->getNumLocalFineNodes()         != 50)                      {chk = -1;}
-      if(myIndexManager->getNumLocalCoarseNodes()       != 9)                       {chk = -1;}
-      if(myIndexManager->getNumLocalGhostedNodes()      != 9)                       {chk = -1;}
-      if(myIndexManager->getNumGlobalCoarseNodes()      != 9)                       {chk = -1;}
-      LO lFNPD[3] = {5, 5, 2}; LO lCNPD[3] = {3, 3, 1}; LO GNPD[3] = {3, 3, 1};
-      for(int dim = 0; dim < 3; ++dim) {
-        if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])            {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim)         != false)                 {chk = -1;}
-        if(myIndexManager->getGhostInterface(dim + 3)     != false)                 {chk = -1;}
-        if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])            {chk = -1;}
-        if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])             {chk = -1;}
-        if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                    {chk = -1;}
-      }
-      GO cnfGIDs[9] = {0, 2, 4, 10, 12, 14, 20, 22, 24};
-      for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-        if(ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx)                           {chk = -1;}
-        if(ghostedNodeCoarsePIDs[coarseIdx] != 0)                                   {chk = -1;}
-        if(coarseNodeFineGIDs[coarseIdx]    != cnfGIDs[coarseIdx])                  {chk = -1;}
-      }
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(myIndexManager->getNumLocalFineNodes()       != 18)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  4)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    !=  4)                      {chk = -1;}
-        LO lFNPD[3] = {3,3,2}; LO lCNPD[3] = {2,2,1}; LO GNPD[3] = {2,2,1};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[4] = {0, 1, 2, 3};
-        int gncPIDs[4] = {0, 0, 0, 0};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[4] = {0, 2, 6, 8};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 1) {
-        if(myIndexManager->getNumLocalFineNodes()       != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  2)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    !=  2)                      {chk = -1;}
-        LO lFNPD[3] = {2,3,2}; LO lCNPD[3] = {1,2,1}; LO GNPD[3] = {1,2,1};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[2] = {0, 1};
-        int gncPIDs[2] = {1, 1};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[2] = {18, 22};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 2) {
-        if(myIndexManager->getNumLocalFineNodes()       != 12)                      {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     !=  2)                      {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    !=  2)                      {chk = -1;}
-        LO lFNPD[3] = {3,2,2}; LO lCNPD[3] = {2,1,1}; LO GNPD[3] = {2,1,1};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[2] = {0, 1};
-        int gncPIDs[2] = {2, 2};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[2] = {30, 32};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      } else if(comm->getRank() == 3) {
-        if(myIndexManager->getNumLocalFineNodes()       != 8)                       {chk = -1;}
-        if(myIndexManager->getNumLocalCoarseNodes()     != 1)                       {chk = -1;}
-        if(myIndexManager->getNumLocalGhostedNodes()    != 1)                       {chk = -1;}
-        LO lFNPD[3] = {2,2,2}; LO lCNPD[3] = {1,1,1}; LO GNPD[3] = {1,1,1};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(myIndexManager->getLocalFineNodesInDir(dim)    != lFNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim)         != false)               {chk = -1;}
-          if(myIndexManager->getGhostInterface(dim + 3)     != false)               {chk = -1;}
-          if(myIndexManager->getLocalCoarseNodesInDir(dim)  != lCNPD[dim])          {chk = -1;}
-          if(myIndexManager->getGhostedNodesInDir(dim)      != GNPD[dim])           {chk = -1;}
-          if(myIndexManager->getGlobalCoarseNodesInDir(dim) != -1)                  {chk = -1;}
-        }
-        LO  gncLIDs[1] = {0};
-        int gncPIDs[1] = {3};
-        for(int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
-          if(ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx])              {chk = -1;}
-          if(ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx])              {chk = -1;}
-        }
-        GO  cnfGIDs[1] = {42};
-        for(int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
-          if(coarseNodeFineGIDs[coarseIdx]   != cnfGIDs[coarseIdx])                 {chk = -1;}
-        }
-      }
-    }
-
-    int gbl_chk[1] = {0};
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
-    TEST_EQUALITY(gbl_chk[0], 0);
-
-  } // UncoupledIndexManagerSingleCoarsePoint
-
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager,CreateGlobalLexicographicIndexManager,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager,CreateLocalLexicographicIndexManager,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager,CreateUncoupledIndexManager,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager,UncoupledIndexManagerSingleCoarsePoint,Scalar,LO,GO,Node) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateGlobalLexicographicIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-#include <MueLu_ETI_4arg.hpp>
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fout->setShowAllFrontMatter(false).setShowProcRank(true);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
+
+  // Set global geometric data
+  const bool coupled           = true;
+  const int numDimensions      = 3;
+  const int interpolationOrder = 0;
+  Array<GO> meshData;
+  Array<GO> gNodesPerDir(3);
+  Array<LO> lNodesPerDir(3);
+  Array<LO> coarseRate(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+      coarseRate[dim]   = 2;
+    } else {
+      gNodesPerDir[dim] = -1;
+      coarseRate[dim]   = 1;
+    }
+  }
+
+  RCP<RealValuedMultiVector> coords =
+      MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions,
+                                                                                gNodesPerDir,
+                                                                                lNodesPerDir,
+                                                                                meshData,
+                                                                                "Global Lexicographic");
+
+  RCP<MueLu::GlobalLexicographicIndexManager<LO, GO, NO> > myIndexManager =
+      rcp(new MueLu::GlobalLexicographicIndexManager<LO, GO, NO>(comm, coupled, numDimensions,
+                                                                 interpolationOrder, gNodesPerDir,
+                                                                 lNodesPerDir, coarseRate,
+                                                                 coords->getMap()->getMinGlobalIndex()));
+
+  Array<LO> ghostedNodeCoarseLIDs;
+  Array<int> ghostedNodeCoarsePIDs;
+  Array<GO> ghostedNodeCoarseGIDs;
+  myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
+                                      ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
+
+  Array<GO> coarseNodeCoarseGIDs;
+  Array<GO> coarseNodeFineGIDs;
+  myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
+                                     coarseNodeFineGIDs);
+
+  int chk = 0;
+  if (myIndexManager->isAggregationCoupled() != coupled) {
+    chk = -1;
+  }
+  if (myIndexManager->getInterpolationOrder() != interpolationOrder) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumDimensions() != numDimensions) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalFineNodes() != 125) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalCoarseNodes() != 27) {
+    chk = -1;
+  }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (myIndexManager->getCoarseningRate(dim) != coarseRate[dim]) {
+      chk = -1;
+    }
+    if (myIndexManager->getGlobalFineNodesInDir(dim) != gNodesPerDir[dim]) {
+      chk = -1;
+    }
+    if (myIndexManager->getGlobalCoarseNodesInDir(dim) != 3) {
+      chk = -1;
+    }
+  }
+  if (comm->getSize() == 1) {
+    if (myIndexManager->getNumLocalFineNodes() != 125) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalCoarseNodes() != 27) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalGhostedNodes() != 27) {
+      chk = -1;
+    }
+    for (int dim = 0; dim < 3; ++dim) {
+      if (myIndexManager->getLocalFineNodesInDir(dim) != lNodesPerDir[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getStartIndex(dim) != 0) {
+        chk = -1;
+      }
+      if (myIndexManager->getStartIndex(dim + 3) != gNodesPerDir[dim] - 1) {
+        chk = -1;
+      }
+      if (myIndexManager->getOffset(dim) != 0) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim + 3) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getLocalCoarseNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostedNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+    }
+    GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
+                      50, 52, 54, 60, 62, 64, 70, 72, 74,
+                      100, 102, 104, 110, 112, 114, 120, 122, 124};
+    for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+      if (ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarsePIDs[coarseIdx] != 0) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarseGIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+        chk = -1;
+      }
+      if (coarseNodeCoarseGIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+    }
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (myIndexManager->getNumLocalFineNodes() != 45) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {3, 3, 5};
+      LO lCNPD[3]      = {2, 2, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {0, 0, 0, 2, 2, 4};
+      GO offsets[3]    = {0, 0, 0};
+      bool ghostInt[6] = {false, false, false, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+      GO gncGIDs[12]  = {0, 1, 3, 4, 9, 10, 12, 13, 18, 19, 21, 22};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {0, 2, 10, 12, 50, 52, 60, 62, 100, 102, 110, 112};
+      GO cncGIDs[12] = {0, 1, 3, 4, 9, 10, 12, 13, 18, 19, 21, 22};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 1) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 6) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {2, 3, 5};
+      LO lCNPD[3]      = {1, 2, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {3, 0, 0, 4, 2, 4};
+      GO offsets[3]    = {1, 0, 0};
+      bool ghostInt[6] = {true, false, false, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {1, 0, 3, 1, 5, 2, 7, 3, 9, 4, 11, 5};
+      int gncPIDs[12] = {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1};
+      GO gncGIDs[12]  = {1, 2, 4, 5, 10, 11, 13, 14, 19, 20, 22, 23};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[6] = {4, 14, 54, 64, 104, 114};
+      GO cncGIDs[6] = {2, 5, 11, 14, 20, 23};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 2) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 6) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {3, 2, 5};
+      LO lCNPD[3]      = {2, 1, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {0, 3, 0, 2, 4, 4};
+      GO offsets[3]    = {0, 1, 0};
+      bool ghostInt[6] = {false, false, true, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {2, 3, 0, 1, 6, 7, 2, 3, 10, 11, 4, 5};
+      int gncPIDs[12] = {0, 0, 2, 2, 0, 0, 2, 2, 0, 0, 2, 2};
+      GO gncGIDs[12]  = {3, 4, 6, 7, 12, 13, 15, 16, 21, 22, 24, 25};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[6] = {20, 22, 70, 72, 120, 122};
+      GO cncGIDs[6] = {6, 7, 15, 16, 24, 25};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 3) {
+      if (myIndexManager->getNumLocalFineNodes() != 20) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {2, 2, 5};
+      LO lCNPD[3]      = {1, 1, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {3, 3, 0, 4, 4, 4};
+      GO offsets[3]    = {1, 1, 0};
+      bool ghostInt[6] = {true, false, true, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {3, 1, 1, 0, 7, 3, 3, 1, 11, 5, 5, 2};
+      int gncPIDs[12] = {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3};
+      GO gncGIDs[12]  = {4, 5, 7, 8, 13, 14, 16, 17, 22, 23, 25, 26};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[3] = {24, 74, 124};
+      GO cncGIDs[3] = {8, 17, 26};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    }
+  }
+
+  int gbl_chk[1] = {0};
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
+  TEST_EQUALITY(gbl_chk[0], 0);
+
+}  // CreateGlobalLexicographicIndexManager
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateLocalLexicographicIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fout->setShowAllFrontMatter(false).setShowProcRank(true);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
+  RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
+
+  // Set global geometric data
+  const bool coupled           = true;
+  const int numDimensions      = 3;
+  const int interpolationOrder = 0;
+  Array<GO> meshData;
+  Array<GO> gNodesPerDir(3);
+  Array<LO> lNodesPerDir(3);
+  Array<LO> coarseRate(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+      coarseRate[dim]   = 2;
+    } else {
+      gNodesPerDir[dim] = -1;
+      coarseRate[dim]   = 1;
+    }
+  }
+
+  RCP<RealValuedMultiVector> coords =
+      MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions,
+                                                                                gNodesPerDir,
+                                                                                lNodesPerDir,
+                                                                                meshData,
+                                                                                "Local Lexicographic");
+
+  RCP<MueLu::LocalLexicographicIndexManager<LO, GO, NO> > myIndexManager =
+      rcp(new MueLu::LocalLexicographicIndexManager<LO, GO, NO>(comm, coupled, numDimensions,
+                                                                interpolationOrder, comm->getRank(),
+                                                                comm->getSize(), gNodesPerDir,
+                                                                lNodesPerDir, coarseRate, meshData));
+
+  Array<LO> ghostedNodeCoarseLIDs;
+  Array<int> ghostedNodeCoarsePIDs;
+  Array<GO> ghostedNodeCoarseGIDs;
+  myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
+                                      ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
+
+  Array<GO> coarseNodeCoarseGIDs;
+  Array<GO> coarseNodeFineGIDs;
+  myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
+                                     coarseNodeFineGIDs);
+
+  int chk = 0;
+  if (myIndexManager->isAggregationCoupled() != coupled) {
+    chk = -1;
+  }
+  if (myIndexManager->getInterpolationOrder() != interpolationOrder) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumDimensions() != numDimensions) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalFineNodes() != 125) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalCoarseNodes() != 27) {
+    chk = -1;
+  }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (myIndexManager->getCoarseningRate(dim) != coarseRate[dim]) {
+      chk = -1;
+    }
+    if (myIndexManager->getGlobalFineNodesInDir(dim) != gNodesPerDir[dim]) {
+      chk = -1;
+    }
+    if (myIndexManager->getGlobalCoarseNodesInDir(dim) != 3) {
+      chk = -1;
+    }
+  }
+  if (comm->getSize() == 1) {
+    if (myIndexManager->getNumLocalFineNodes() != 125) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalCoarseNodes() != 27) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalGhostedNodes() != 27) {
+      chk = -1;
+    }
+    for (int dim = 0; dim < 3; ++dim) {
+      if (myIndexManager->getLocalFineNodesInDir(dim) != lNodesPerDir[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getStartIndex(dim) != 0) {
+        chk = -1;
+      }
+      if (myIndexManager->getStartIndex(dim + 3) != gNodesPerDir[dim] - 1) {
+        chk = -1;
+      }
+      if (myIndexManager->getOffset(dim) != 0) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim + 3) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getLocalCoarseNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostedNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+    }
+    GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
+                      50, 52, 54, 60, 62, 64, 70, 72, 74,
+                      100, 102, 104, 110, 112, 114, 120, 122, 124};
+    for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+      if (ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarsePIDs[coarseIdx] != 0) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarseGIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+        chk = -1;
+      }
+      if (coarseNodeCoarseGIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+    }
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (myIndexManager->getNumLocalFineNodes() != 45) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {3, 3, 5};
+      LO lCNPD[3]      = {2, 2, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {0, 0, 0, 2, 2, 4};
+      GO offsets[3]    = {0, 0, 0};
+      bool ghostInt[6] = {false, false, false, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+      GO gncGIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {0, 2, 6, 8, 18, 20, 24, 26, 36, 38, 42, 44};
+      GO cncGIDs[12] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 1) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 6) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {2, 3, 5};
+      LO lCNPD[3]      = {1, 2, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {3, 0, 0, 4, 2, 4};
+      GO offsets[3]    = {1, 0, 0};
+      bool ghostInt[6] = {true, false, false, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {1, 0, 3, 1, 5, 2, 7, 3, 9, 4, 11, 5};
+      int gncPIDs[12] = {0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1};
+      GO gncGIDs[12]  = {1, 12, 3, 13, 5, 14, 7, 15, 9, 16, 11, 17};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[6] = {46, 50, 58, 62, 70, 74};
+      GO cncGIDs[6] = {12, 13, 14, 15, 16, 17};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 2) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 6) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {3, 2, 5};
+      LO lCNPD[3]      = {2, 1, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {0, 3, 0, 2, 4, 4};
+      GO offsets[3]    = {0, 1, 0};
+      bool ghostInt[6] = {false, false, true, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {2, 3, 0, 1, 6, 7, 2, 3, 10, 11, 4, 5};
+      int gncPIDs[12] = {0, 0, 2, 2, 0, 0, 2, 2, 0, 0, 2, 2};
+      GO gncGIDs[12]  = {2, 3, 18, 19, 6, 7, 20, 21, 10, 11, 22, 23};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[6] = {78, 80, 90, 92, 102, 104};
+      GO cncGIDs[6] = {18, 19, 20, 21, 22, 23};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 3) {
+      if (myIndexManager->getNumLocalFineNodes() != 20) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3]      = {2, 2, 5};
+      LO lCNPD[3]      = {1, 1, 3};
+      LO GNPD[3]       = {2, 2, 3};
+      GO startIndex[6] = {3, 3, 0, 4, 4, 4};
+      GO offsets[3]    = {1, 1, 0};
+      bool ghostInt[6] = {true, false, true, false, false, false};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim) != startIndex[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getStartIndex(dim + 3) != startIndex[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getOffset(dim) != offsets[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != ghostInt[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != ghostInt[dim + 3]) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {3, 1, 1, 0, 7, 3, 3, 1, 11, 5, 5, 2};
+      int gncPIDs[12] = {0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3};
+      GO gncGIDs[12]  = {3, 13, 19, 24, 7, 15, 21, 25, 11, 17, 23, 26};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarseGIDs[ghostedIdx] != gncGIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[3] = {108, 116, 124};
+      GO cncGIDs[3] = {24, 25, 26};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalCoarseNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+        if (coarseNodeCoarseGIDs[coarseIdx] != cncGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    }
+  }
+
+  int gbl_chk[1] = {0};
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
+  TEST_EQUALITY(gbl_chk[0], 0);
+
+}  // CreateLocalLexicographicIndexManager
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, CreateUncoupledIndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fout->setShowAllFrontMatter(false).setShowProcRank(true);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
+
+  // Set global geometric data
+  const bool coupled           = false;
+  const int numDimensions      = 3;
+  const int interpolationOrder = 0;
+  Array<GO> meshData;
+  Array<GO> gNodesPerDir(3);
+  Array<LO> lNodesPerDir(3);
+  Array<LO> coarseRate(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+      coarseRate[dim]   = 2;
+    } else {
+      gNodesPerDir[dim] = -1;
+      coarseRate[dim]   = 1;
+    }
+  }
+
+  RCP<RealValuedMultiVector> coords =
+      MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions,
+                                                                                gNodesPerDir,
+                                                                                lNodesPerDir,
+                                                                                meshData,
+                                                                                "Local Lexicographic");
+
+  RCP<MueLu::UncoupledIndexManager<LO, GO, NO> > myIndexManager =
+      rcp(new MueLu::UncoupledIndexManager<LO, GO, NO>(comm, coupled, numDimensions,
+                                                       interpolationOrder, comm->getRank(),
+                                                       comm->getSize(), gNodesPerDir,
+                                                       lNodesPerDir, coarseRate, false));
+  myIndexManager->computeGlobalCoarseParameters();
+
+  Array<LO> ghostedNodeCoarseLIDs;
+  Array<int> ghostedNodeCoarsePIDs;
+  Array<GO> ghostedNodeCoarseGIDs;
+  myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
+                                      ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
+
+  Array<GO> coarseNodeCoarseGIDs;
+  Array<GO> coarseNodeFineGIDs;
+  myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
+                                     coarseNodeFineGIDs);
+
+  int chk = 0;
+  if (myIndexManager->isAggregationCoupled() != coupled) {
+    chk = -1;
+  }
+  if (myIndexManager->getInterpolationOrder() != interpolationOrder) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumDimensions() != numDimensions) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalFineNodes() != -1) {
+    chk = -1;
+  }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (myIndexManager->getCoarseningRate(dim) != coarseRate[dim]) {
+      chk = -1;
+    }
+  }
+  if (comm->getSize() == 1) {
+    if (myIndexManager->getNumLocalFineNodes() != 125) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalCoarseNodes() != 27) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalGhostedNodes() != 27) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumGlobalCoarseNodes() != 27) {
+      chk = -1;
+    }
+    for (int dim = 0; dim < 3; ++dim) {
+      if (myIndexManager->getLocalFineNodesInDir(dim) != lNodesPerDir[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim + 3) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getLocalCoarseNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostedNodesInDir(dim) != 3) {
+        chk = -1;
+      }
+      if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+        chk = -1;
+      }
+    }
+    GO cnfGIDs[27] = {0, 2, 4, 10, 12, 14, 20, 22, 24,
+                      50, 52, 54, 60, 62, 64, 70, 72, 74,
+                      100, 102, 104, 110, 112, 114, 120, 122, 124};
+    for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+      if (ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarsePIDs[coarseIdx] != 0) {
+        chk = -1;
+      }
+      if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+        chk = -1;
+      }
+    }
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (myIndexManager->getNumLocalFineNodes() != 45) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 3, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      LO GNPD[3]  = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {0, 2, 6, 8, 18, 20, 24, 26, 36, 38, 42, 44};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 1) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 3, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      LO GNPD[3]  = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {45, 46, 49, 50, 57, 58, 61, 62, 69, 70, 73, 74};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 2) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 2, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      LO GNPD[3]  = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {75, 77, 78, 80, 87, 89, 90, 92, 99, 101, 102, 104};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 3) {
+      if (myIndexManager->getNumLocalFineNodes() != 20) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 2, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      LO GNPD[3]  = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[12]  = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+      int gncPIDs[12] = {3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[12] = {105, 106, 107, 108, 113, 114, 115, 116, 121, 122, 123, 124};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    }
+  }
+
+  int gbl_chk[1] = {0};
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
+  TEST_EQUALITY(gbl_chk[0], 0);
+
+}  // CreateUncoupledIndexManager
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager, UncoupledIndexManagerSingleCoarsePoint, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fout->setShowAllFrontMatter(false).setShowProcRank(true);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers::Parameters::getDefaultComm();
+
+  // Set global geometric data
+  const bool coupled           = false;
+  const int numDimensions      = 3;
+  const int interpolationOrder = 0;
+  Array<GO> meshData;
+  Array<GO> gNodesPerDir(3);
+  Array<LO> lNodesPerDir(3);
+  Array<LO> coarseRate(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+      coarseRate[dim]   = 2;
+    } else {
+      gNodesPerDir[dim] = -1;
+      coarseRate[dim]   = 1;
+    }
+  }
+  // Change the number of nodes in the third
+  // direction to trigger singleCoarsePoint
+  gNodesPerDir[2] = 2;
+
+  RCP<RealValuedMultiVector> coords =
+      MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions,
+                                                                                gNodesPerDir,
+                                                                                lNodesPerDir,
+                                                                                meshData,
+                                                                                "Local Lexicographic");
+
+  RCP<MueLu::UncoupledIndexManager<LO, GO, NO> > myIndexManager =
+      rcp(new MueLu::UncoupledIndexManager<LO, GO, NO>(comm, coupled, numDimensions,
+                                                       interpolationOrder, comm->getRank(),
+                                                       comm->getSize(), gNodesPerDir,
+                                                       lNodesPerDir, coarseRate, true));
+  myIndexManager->computeGlobalCoarseParameters();
+
+  Array<LO> ghostedNodeCoarseLIDs;
+  Array<int> ghostedNodeCoarsePIDs;
+  Array<GO> ghostedNodeCoarseGIDs;
+  myIndexManager->getGhostedNodesData(coords->getMap(), ghostedNodeCoarseLIDs,
+                                      ghostedNodeCoarsePIDs, ghostedNodeCoarseGIDs);
+
+  Array<GO> coarseNodeCoarseGIDs;
+  Array<GO> coarseNodeFineGIDs;
+  myIndexManager->getCoarseNodesData(coords->getMap(), coarseNodeCoarseGIDs,
+                                     coarseNodeFineGIDs);
+
+  int chk = 0;
+  if (myIndexManager->isAggregationCoupled() != coupled) {
+    chk = -1;
+  }
+  if (myIndexManager->getInterpolationOrder() != interpolationOrder) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumDimensions() != numDimensions) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumGlobalFineNodes() != -1) {
+    chk = -1;
+  }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (myIndexManager->getCoarseningRate(dim) != coarseRate[dim]) {
+      chk = -1;
+    }
+  }
+  if (comm->getSize() == 1) {
+    if (myIndexManager->getNumLocalFineNodes() != 50) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalCoarseNodes() != 9) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumLocalGhostedNodes() != 9) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumGlobalCoarseNodes() != 9) {
+      chk = -1;
+    }
+    LO lFNPD[3] = {5, 5, 2};
+    LO lCNPD[3] = {3, 3, 1};
+    LO GNPD[3]  = {3, 3, 1};
+    for (int dim = 0; dim < 3; ++dim) {
+      if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostInterface(dim + 3) != false) {
+        chk = -1;
+      }
+      if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+        chk = -1;
+      }
+      if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+        chk = -1;
+      }
+    }
+    GO cnfGIDs[9] = {0, 2, 4, 10, 12, 14, 20, 22, 24};
+    for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+      if (ghostedNodeCoarseLIDs[coarseIdx] != coarseIdx) {
+        chk = -1;
+      }
+      if (ghostedNodeCoarsePIDs[coarseIdx] != 0) {
+        chk = -1;
+      }
+      if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+        chk = -1;
+      }
+    }
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (myIndexManager->getNumLocalFineNodes() != 18) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 4) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 4) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 3, 2};
+      LO lCNPD[3] = {2, 2, 1};
+      LO GNPD[3]  = {2, 2, 1};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[4]  = {0, 1, 2, 3};
+      int gncPIDs[4] = {0, 0, 0, 0};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[4] = {0, 2, 6, 8};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 1) {
+      if (myIndexManager->getNumLocalFineNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 2) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 2) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 3, 2};
+      LO lCNPD[3] = {1, 2, 1};
+      LO GNPD[3]  = {1, 2, 1};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[2]  = {0, 1};
+      int gncPIDs[2] = {1, 1};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[2] = {18, 22};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 2) {
+      if (myIndexManager->getNumLocalFineNodes() != 12) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 2) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 2) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 2, 2};
+      LO lCNPD[3] = {2, 1, 1};
+      LO GNPD[3]  = {2, 1, 1};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[2]  = {0, 1};
+      int gncPIDs[2] = {2, 2};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[2] = {30, 32};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 3) {
+      if (myIndexManager->getNumLocalFineNodes() != 8) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalCoarseNodes() != 1) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumLocalGhostedNodes() != 1) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 2, 2};
+      LO lCNPD[3] = {1, 1, 1};
+      LO GNPD[3]  = {1, 1, 1};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (myIndexManager->getLocalFineNodesInDir(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostInterface(dim + 3) != false) {
+          chk = -1;
+        }
+        if (myIndexManager->getLocalCoarseNodesInDir(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGhostedNodesInDir(dim) != GNPD[dim]) {
+          chk = -1;
+        }
+        if (myIndexManager->getGlobalCoarseNodesInDir(dim) != -1) {
+          chk = -1;
+        }
+      }
+      LO gncLIDs[1]  = {0};
+      int gncPIDs[1] = {3};
+      for (int ghostedIdx = 0; ghostedIdx < myIndexManager->getNumLocalGhostedNodes(); ++ghostedIdx) {
+        if (ghostedNodeCoarseLIDs[ghostedIdx] != gncLIDs[ghostedIdx]) {
+          chk = -1;
+        }
+        if (ghostedNodeCoarsePIDs[ghostedIdx] != gncPIDs[ghostedIdx]) {
+          chk = -1;
+        }
+      }
+      GO cnfGIDs[1] = {42};
+      for (int coarseIdx = 0; coarseIdx < myIndexManager->getNumLocalGhostedNodes(); ++coarseIdx) {
+        if (coarseNodeFineGIDs[coarseIdx] != cnfGIDs[coarseIdx]) {
+          chk = -1;
+        }
+      }
+    }
+  }
+
+  int gbl_chk[1] = {0};
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
+  TEST_EQUALITY(gbl_chk[0], 0);
+
+}  // UncoupledIndexManagerSingleCoarsePoint
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager, CreateGlobalLexicographicIndexManager, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager, CreateLocalLexicographicIndexManager, Scalar, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager, CreateUncoupledIndexManager, Scalar, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager, UncoupledIndexManagerSingleCoarsePoint, Scalar, LO, GO, Node)
+
+#include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/IntrepidPCoarsenFactory.cpp b/packages/muelu/test/unit_tests/IntrepidPCoarsenFactory.cpp
index eab6466081d5..e491f8106431 100644
--- a/packages/muelu/test/unit_tests/IntrepidPCoarsenFactory.cpp
+++ b/packages/muelu/test/unit_tests/IntrepidPCoarsenFactory.cpp
@@ -44,7 +44,7 @@
 //
 // @HEADER
 #include <Teuchos_UnitTestHarness.hpp>
-#include <MueLu_UnitTestHelpers.hpp> // provides TEUCHOS_UNIT_TEST_TEMPLATE_5_INSTANT, etc.
+#include <MueLu_UnitTestHelpers.hpp>  // provides TEUCHOS_UNIT_TEST_TEMPLATE_5_INSTANT, etc.
 #include <Teuchos_DefaultComm.hpp>
 #include <Teuchos_ScalarTraits.hpp>
 
@@ -69,7 +69,7 @@
 
 #ifdef HAVE_MUELU_INTREPID2
 #include "MueLu_IntrepidPCoarsenFactory.hpp"
-#include "MueLu_IntrepidPCoarsenFactory_def.hpp"   // Why does ETI suddenly decide to hate right here?
+#include "MueLu_IntrepidPCoarsenFactory_def.hpp"  // Why does ETI suddenly decide to hate right here?
 #include "Intrepid2_Types.hpp"
 #include "Intrepid2_HGRAD_QUAD_C1_FEM.hpp"
 #include "Intrepid2_HGRAD_QUAD_Cn_FEM.hpp"
@@ -78,2782 +78,2538 @@
 //#include "Intrepid2_HGRAD_TET_Cn_FEM.hpp"
 #include "Kokkos_DynRankView.hpp"
 
-
 namespace MueLuTests {
 
-  /**** some helper methods and classes by Nate ****/
+/**** some helper methods and classes by Nate ****/
 #ifndef TEST_MORE_COMBINATIONS
-  static const int MAX_LINE_DEGREE = Intrepid2::Parameters::MaxOrder;
-  static const int MAX_QUAD_DEGREE = Intrepid2::Parameters::MaxOrder;
-  static const int MAX_HEX_DEGREE = (Intrepid2::Parameters::MaxOrder < 4 ) ? Intrepid2::Parameters::MaxOrder : 4;
-  static const int MAX_RANK_COUNT = 4;
+static const int MAX_LINE_DEGREE = Intrepid2::Parameters::MaxOrder;
+static const int MAX_QUAD_DEGREE = Intrepid2::Parameters::MaxOrder;
+static const int MAX_HEX_DEGREE  = (Intrepid2::Parameters::MaxOrder < 4) ? Intrepid2::Parameters::MaxOrder : 4;
+static const int MAX_RANK_COUNT  = 4;
 #else
-  static const int MAX_LINE_DEGREE = Intrepid2::Parameters::MaxOrder;
-  static const int MAX_QUAD_DEGREE = Intrepid2::Parameters::MaxOrder;
-  static const int MAX_HEX_DEGREE = Intrepid2::Parameters::MaxOrder;
-  static const int MAX_RANK_COUNT = 16;
+static const int MAX_LINE_DEGREE = Intrepid2::Parameters::MaxOrder;
+static const int MAX_QUAD_DEGREE = Intrepid2::Parameters::MaxOrder;
+static const int MAX_HEX_DEGREE  = Intrepid2::Parameters::MaxOrder;
+static const int MAX_RANK_COUNT  = 16;
 #endif
 
+using namespace std;
+// pair is subcell dim, subcell ordinal in cellTopo.  Includes (spaceDim-1, sideOrdinal), where spaceDim is the extent of the cellTopo
+std::vector<std::pair<int, int>> subcellEntitiesForSide(const shards::CellTopology &cellTopo, int sideOrdinal) {
   using namespace std;
-  // pair is subcell dim, subcell ordinal in cellTopo.  Includes (spaceDim-1, sideOrdinal), where spaceDim is the extent of the cellTopo
-  std::vector<std::pair<int,int>> subcellEntitiesForSide(const shards::CellTopology &cellTopo, int sideOrdinal)
-  {
-    using namespace std;
-    vector<pair<int,int>> subcellEntities;
-    set<int> nodesForSide;
-    int spaceDim = (int) cellTopo.getDimension();
-    if (spaceDim == 0) return {{}};
-    int sideDim = spaceDim - 1;
-    int nodeCount = (int) cellTopo.getNodeCount(sideDim, sideOrdinal);
-    // first, collect all the nodes that match the side
-    for (int nodeOrdinal=0; nodeOrdinal < nodeCount; nodeOrdinal++)
-    {
-      int node = (int) cellTopo.getNodeMap(sideDim, sideOrdinal, nodeOrdinal);
-      nodesForSide.insert(node);
-    }
-    // now, iterate over extents.
-    // Any subcells that only have nodes that match nodesForSide should be included.
-    for (int d=0; d<=sideDim; d++)
-    {
-      int subcellCount = cellTopo.getSubcellCount(d);
-      for (int subcord = 0; subcord < subcellCount; subcord++)
-      {
-        bool allNodesMatch = true;
-        if (d == 0)
-        {
-          // subcord is a node; just check whether that node is in nodesForSide
-          allNodesMatch = (nodesForSide.find(subcord) != nodesForSide.end());
-        }
-        else
-        {
-          int scNodeCount = cellTopo.getNodeCount(d, subcord);
-          for (int scNodeOrdinal = 0; scNodeOrdinal < scNodeCount; scNodeOrdinal++)
-          {
-            int scNode = (int) cellTopo.getNodeMap(d, subcord, scNodeOrdinal);
-            if (nodesForSide.find(scNode) == nodesForSide.end())
-            {
-              allNodesMatch = false;
-              break;
-            }
+  vector<pair<int, int>> subcellEntities;
+  set<int> nodesForSide;
+  int spaceDim = (int)cellTopo.getDimension();
+  if (spaceDim == 0) return {{}};
+  int sideDim   = spaceDim - 1;
+  int nodeCount = (int)cellTopo.getNodeCount(sideDim, sideOrdinal);
+  // first, collect all the nodes that match the side
+  for (int nodeOrdinal = 0; nodeOrdinal < nodeCount; nodeOrdinal++) {
+    int node = (int)cellTopo.getNodeMap(sideDim, sideOrdinal, nodeOrdinal);
+    nodesForSide.insert(node);
+  }
+  // now, iterate over extents.
+  // Any subcells that only have nodes that match nodesForSide should be included.
+  for (int d = 0; d <= sideDim; d++) {
+    int subcellCount = cellTopo.getSubcellCount(d);
+    for (int subcord = 0; subcord < subcellCount; subcord++) {
+      bool allNodesMatch = true;
+      if (d == 0) {
+        // subcord is a node; just check whether that node is in nodesForSide
+        allNodesMatch = (nodesForSide.find(subcord) != nodesForSide.end());
+      } else {
+        int scNodeCount = cellTopo.getNodeCount(d, subcord);
+        for (int scNodeOrdinal = 0; scNodeOrdinal < scNodeCount; scNodeOrdinal++) {
+          int scNode = (int)cellTopo.getNodeMap(d, subcord, scNodeOrdinal);
+          if (nodesForSide.find(scNode) == nodesForSide.end()) {
+            allNodesMatch = false;
+            break;
           }
         }
-        if (allNodesMatch)
-        {
-          subcellEntities.push_back({d,subcord});
-        }
+      }
+      if (allNodesMatch) {
+        subcellEntities.push_back({d, subcord});
       }
     }
-    return subcellEntities;
   }
+  return subcellEntities;
+}
 
-  //! Returns ordinals in the basis that have nodes on the specified side, as a sorted vector<int>
-  template<class Basis>
-  std::vector<int> localDofOrdinalsForSide(RCP<Basis> basis, int sideOrdinal)
-  {
-    using namespace std;
-    // to use dof tags for this, we first need to determine the subcells of the domain that
-    // are part of the specified side
-    auto subcellEntities = subcellEntitiesForSide(basis->getBaseCellTopology(), sideOrdinal);
+//! Returns ordinals in the basis that have nodes on the specified side, as a sorted vector<int>
+template <class Basis>
+std::vector<int> localDofOrdinalsForSide(RCP<Basis> basis, int sideOrdinal) {
+  using namespace std;
+  // to use dof tags for this, we first need to determine the subcells of the domain that
+  // are part of the specified side
+  auto subcellEntities = subcellEntitiesForSide(basis->getBaseCellTopology(), sideOrdinal);
 
-    auto dofOrdinalData = basis->getAllDofOrdinal();
+  auto dofOrdinalData = basis->getAllDofOrdinal();
 
-    // determine size of first two parts of dofOrdinalData container
-    // for extents > 0, there may no entries at all (for lower-order bases)
-    int maxDim = dofOrdinalData.extent(0);
-    int maxSubcellOrdinal = dofOrdinalData.extent(1);
+  // determine size of first two parts of dofOrdinalData container
+  // for extents > 0, there may no entries at all (for lower-order bases)
+  int maxDim            = dofOrdinalData.extent(0);
+  int maxSubcellOrdinal = dofOrdinalData.extent(1);
 
-    vector<int> localDofOrdinals;
+  vector<int> localDofOrdinals;
 
-    for (auto subcellEntity : subcellEntities)
-    {
-      int subcellDim = subcellEntity.first;
-      int subcellOrdinal = subcellEntity.second;
+  for (auto subcellEntity : subcellEntities) {
+    int subcellDim     = subcellEntity.first;
+    int subcellOrdinal = subcellEntity.second;
 
-      if (subcellDim >= maxDim) continue; // no entries
+    if (subcellDim >= maxDim) continue;  // no entries
 
-      int dofContainerSize = dofOrdinalData.extent(2); // 3rd extent: max dof count
+    int dofContainerSize = dofOrdinalData.extent(2);  // 3rd extent: max dof count
 
-      if (subcellOrdinal >= maxSubcellOrdinal) continue; // no entries
+    if (subcellOrdinal >= maxSubcellOrdinal) continue;  // no entries
 
-      for (int entryOrdinal = 0; entryOrdinal < dofContainerSize; entryOrdinal++)
-      {
-        int localDofOrdinal = dofOrdinalData(subcellDim,subcellOrdinal,entryOrdinal);
+    for (int entryOrdinal = 0; entryOrdinal < dofContainerSize; entryOrdinal++) {
+      int localDofOrdinal = dofOrdinalData(subcellDim, subcellOrdinal, entryOrdinal);
 
-        if (localDofOrdinal >= 0)
-          localDofOrdinals.push_back(localDofOrdinal);
-        else
-          break;
-      }
+      if (localDofOrdinal >= 0)
+        localDofOrdinals.push_back(localDofOrdinal);
+      else
+        break;
     }
-
-    std::sort(localDofOrdinals.begin(), localDofOrdinals.end());
-
-    return localDofOrdinals;
   }
 
-  class Symmetries
-  {
-    // the symmetries on a given topology (e.g. cube) are a subset of the
-    // permutations of the nodes: namely, the ones for which nodal connectivities
-    // (edges) are preserved.
-    // This class enumerates the symmetries.  The 0 symmetry is the identity permutation.
-    vector<vector<int>> _symmetries;
-    vector<vector<int>> _inverses;
-    int _N;
-  public:
-    Symmetries(const vector<vector<int>> &symmetriesList)
-    {
-      _symmetries = symmetriesList;
-      // sanity checks on the input:
-      // - require that all entries are of equal length N
-      // - require that all entries contain each of 0, ..., N-1
-
-      if (symmetriesList.size() == 0)
-      {
-        _N = 0;
-      }
-      else
-      {
-        _N = int(symmetriesList[0].size());
-        for (auto symmetry : symmetriesList)
-        {
-          TEUCHOS_TEST_FOR_EXCEPTION(int(symmetry.size()) != _N, std::invalid_argument, "Each symmetry must have the same length as all the others.");
-          vector<int> inverse(_N,-1);
-          for (int i=0; i<_N; i++)
-          {
-            int i_mapped = symmetry[i];
-            TEUCHOS_TEST_FOR_EXCEPTION((i_mapped < 0) || (i_mapped > _N), std::invalid_argument,
-                                       "Each symmetry entry must be between 0 and N-1, inclusive.");
-            inverse[i_mapped] = i;
-          }
-          for (int j=0; j<_N; j++)
-          {
-            TEUCHOS_TEST_FOR_EXCEPTION(j == -1, std::invalid_argument, "Each symmetry must include every integer between 0 and N-1, inclusive.");
-          }
-          _inverses.push_back(inverse);
+  std::sort(localDofOrdinals.begin(), localDofOrdinals.end());
+
+  return localDofOrdinals;
+}
+
+class Symmetries {
+  // the symmetries on a given topology (e.g. cube) are a subset of the
+  // permutations of the nodes: namely, the ones for which nodal connectivities
+  // (edges) are preserved.
+  // This class enumerates the symmetries.  The 0 symmetry is the identity permutation.
+  vector<vector<int>> _symmetries;
+  vector<vector<int>> _inverses;
+  int _N;
+
+ public:
+  Symmetries(const vector<vector<int>> &symmetriesList) {
+    _symmetries = symmetriesList;
+    // sanity checks on the input:
+    // - require that all entries are of equal length N
+    // - require that all entries contain each of 0, ..., N-1
+
+    if (symmetriesList.size() == 0) {
+      _N = 0;
+    } else {
+      _N = int(symmetriesList[0].size());
+      for (auto symmetry : symmetriesList) {
+        TEUCHOS_TEST_FOR_EXCEPTION(int(symmetry.size()) != _N, std::invalid_argument, "Each symmetry must have the same length as all the others.");
+        vector<int> inverse(_N, -1);
+        for (int i = 0; i < _N; i++) {
+          int i_mapped = symmetry[i];
+          TEUCHOS_TEST_FOR_EXCEPTION((i_mapped < 0) || (i_mapped > _N), std::invalid_argument,
+                                     "Each symmetry entry must be between 0 and N-1, inclusive.");
+          inverse[i_mapped] = i;
         }
+        for (int j = 0; j < _N; j++) {
+          TEUCHOS_TEST_FOR_EXCEPTION(j == -1, std::invalid_argument, "Each symmetry must include every integer between 0 and N-1, inclusive.");
+        }
+        _inverses.push_back(inverse);
       }
     }
+  }
 
-    const vector<int>& getPermutation(int permutationOrdinal)
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION(permutationOrdinal < 0, std::invalid_argument, "permutationOrdinal must be positive");
-      TEUCHOS_TEST_FOR_EXCEPTION(permutationOrdinal >= int(_symmetries.size()), std::invalid_argument, "permutationOrdinal out of range");
-      return _symmetries[permutationOrdinal];
-    }
+  const vector<int> &getPermutation(int permutationOrdinal) {
+    TEUCHOS_TEST_FOR_EXCEPTION(permutationOrdinal < 0, std::invalid_argument, "permutationOrdinal must be positive");
+    TEUCHOS_TEST_FOR_EXCEPTION(permutationOrdinal >= int(_symmetries.size()), std::invalid_argument, "permutationOrdinal out of range");
+    return _symmetries[permutationOrdinal];
+  }
 
-    int getPermutationCount()
-    {
-      return _symmetries.size();
-    }
+  int getPermutationCount() {
+    return _symmetries.size();
+  }
 
-    void printAll()
-    {
-      for (auto permutation : _symmetries)
-      {
-        cout << permutationString(permutation) << endl;
-      }
+  void printAll() {
+    for (auto permutation : _symmetries) {
+      cout << permutationString(permutation) << endl;
     }
+  }
 
-    static string permutationString(vector<int> &permutation)
-    {
-      ostringstream permString;
-      permString << "{ ";
-      for (int entry : permutation)
-      {
-        permString << entry << " ";
-      }
-      permString << "}";
-      return permString.str();
+  static string permutationString(vector<int> &permutation) {
+    ostringstream permString;
+    permString << "{ ";
+    for (int entry : permutation) {
+      permString << entry << " ";
     }
+    permString << "}";
+    return permString.str();
+  }
 
-    typedef vector<vector<int>> EdgeContainer;
+  typedef vector<vector<int>> EdgeContainer;
 
-    // assumes that the inner vectors are sorted:
-    static bool hasEdge(const EdgeContainer &edges, int node1, int node2)
-    {
-      return std::find(edges[node1].begin(), edges[node1].end(), node2) != edges[node1].end();
-    }
+  // assumes that the inner vectors are sorted:
+  static bool hasEdge(const EdgeContainer &edges, int node1, int node2) {
+    return std::find(edges[node1].begin(), edges[node1].end(), node2) != edges[node1].end();
+  }
 
-    /*
-     For any topology, we can determine the symmetries from
-     the graph of the edges.  This method takes as argument
-     a representation of said graph.  The edges container has
-     size equal to the node count, and the entries for a given
-     node are exactly the nodes that it is connected to by an edge.
-     */
-    static Symmetries symmetries(const EdgeContainer &unsortedEdges)
-    {
-      EdgeContainer edges = unsortedEdges;
-      // now sort
-      int nodeCount = (int) edges.size();
-      for (int node=0; node<nodeCount; node++)
-      {
-        std::sort(edges[node].begin(), edges[node].end());
-      }
+  /*
+   For any topology, we can determine the symmetries from
+   the graph of the edges.  This method takes as argument
+   a representation of said graph.  The edges container has
+   size equal to the node count, and the entries for a given
+   node are exactly the nodes that it is connected to by an edge.
+   */
+  static Symmetries symmetries(const EdgeContainer &unsortedEdges) {
+    EdgeContainer edges = unsortedEdges;
+    // now sort
+    int nodeCount = (int)edges.size();
+    for (int node = 0; node < nodeCount; node++) {
+      std::sort(edges[node].begin(), edges[node].end());
+    }
 
-      vector<vector<int>> permutations;
-
-      // recursive lambda function to determine valid permutations that start with a specified set of choices:
-      function<void(const vector<int> &)> addPermutationsThatMatch;
-      addPermutationsThatMatch = [nodeCount,edges,&addPermutationsThatMatch,&permutations]
-      (const vector<int> &permutationStart) -> void
-      {
-        int nextNodeToMap = int(permutationStart.size());
-        // iterate through the possible choices for nodes to map to:
-        for (int possibleMappedNode=0; possibleMappedNode<nodeCount; possibleMappedNode++)
-        {
-          // is this node already mapped?
-          bool isAlreadyMapped = false;
-          for (int mappedNode : permutationStart)
-          {
-            if (mappedNode == possibleMappedNode) isAlreadyMapped = true;
-          }
-          if (isAlreadyMapped) continue;
-          // not already mapped: let's check whether the edge relationships agree
-          bool mappingAgrees = true;
-          int numNodesMapped = (int) permutationStart.size();
-          for (int node=0; node<numNodesMapped; node++)
-          {
-            int mappedNode = permutationStart[node];
-            bool originalHasEdge = hasEdge(edges, node, nextNodeToMap);
-            bool mappedHasEdge = hasEdge(edges, mappedNode, possibleMappedNode);
-            if (originalHasEdge != mappedHasEdge)
-            {
-              mappingAgrees = false;
-              break;
-            }
+    vector<vector<int>> permutations;
+
+    // recursive lambda function to determine valid permutations that start with a specified set of choices:
+    function<void(const vector<int> &)> addPermutationsThatMatch;
+    addPermutationsThatMatch = [nodeCount, edges, &addPermutationsThatMatch, &permutations](const vector<int> &permutationStart) -> void {
+      int nextNodeToMap = int(permutationStart.size());
+      // iterate through the possible choices for nodes to map to:
+      for (int possibleMappedNode = 0; possibleMappedNode < nodeCount; possibleMappedNode++) {
+        // is this node already mapped?
+        bool isAlreadyMapped = false;
+        for (int mappedNode : permutationStart) {
+          if (mappedNode == possibleMappedNode) isAlreadyMapped = true;
+        }
+        if (isAlreadyMapped) continue;
+        // not already mapped: let's check whether the edge relationships agree
+        bool mappingAgrees = true;
+        int numNodesMapped = (int)permutationStart.size();
+        for (int node = 0; node < numNodesMapped; node++) {
+          int mappedNode       = permutationStart[node];
+          bool originalHasEdge = hasEdge(edges, node, nextNodeToMap);
+          bool mappedHasEdge   = hasEdge(edges, mappedNode, possibleMappedNode);
+          if (originalHasEdge != mappedHasEdge) {
+            mappingAgrees = false;
+            break;
           }
-          if (mappingAgrees)
-          {
-            // no conflict detected: try this mapping choice
-            vector<int> permutationStartCopy = permutationStart;
-            permutationStartCopy.push_back(possibleMappedNode);
-            if (int(permutationStartCopy.size()) == nodeCount)
-            {
-              permutations.push_back(permutationStartCopy);
-            }
-            else
-            {
-              addPermutationsThatMatch(permutationStartCopy);
-            }
+        }
+        if (mappingAgrees) {
+          // no conflict detected: try this mapping choice
+          vector<int> permutationStartCopy = permutationStart;
+          permutationStartCopy.push_back(possibleMappedNode);
+          if (int(permutationStartCopy.size()) == nodeCount) {
+            permutations.push_back(permutationStartCopy);
+          } else {
+            addPermutationsThatMatch(permutationStartCopy);
           }
         }
-      };
+      }
+    };
 
-      vector<int> permutation;
-      addPermutationsThatMatch(permutation);
+    vector<int> permutation;
+    addPermutationsThatMatch(permutation);
 
-      return Symmetries(permutations);
-    }
+    return Symmetries(permutations);
+  }
 
-    static Symmetries shardsSymmetries(shards::CellTopology &cellTopo)
-    {
-      int edgeDim = 1;
-      int nodeCount = cellTopo.getNodeCount();
-      TEUCHOS_TEST_FOR_EXCEPTION(nodeCount != int(cellTopo.getVertexCount()), std::invalid_argument, "Higher-order topologies are not supported");
-      int edgeCount = cellTopo.getSubcellCount(edgeDim);
-      EdgeContainer edges(nodeCount); // set --> sorted, which is handy
-      for (int edgeOrdinal=0; edgeOrdinal<edgeCount; edgeOrdinal++)
-      {
-        int node0 = cellTopo.getNodeMap(edgeDim, edgeOrdinal, 0);
-        int node1 = cellTopo.getNodeMap(edgeDim, edgeOrdinal, 1);
-        edges[node0].push_back(node1);
-        edges[node1].push_back(node0);
-      }
-      return symmetries(edges);
+  static Symmetries shardsSymmetries(shards::CellTopology &cellTopo) {
+    int edgeDim   = 1;
+    int nodeCount = cellTopo.getNodeCount();
+    TEUCHOS_TEST_FOR_EXCEPTION(nodeCount != int(cellTopo.getVertexCount()), std::invalid_argument, "Higher-order topologies are not supported");
+    int edgeCount = cellTopo.getSubcellCount(edgeDim);
+    EdgeContainer edges(nodeCount);  // set --> sorted, which is handy
+    for (int edgeOrdinal = 0; edgeOrdinal < edgeCount; edgeOrdinal++) {
+      int node0 = cellTopo.getNodeMap(edgeDim, edgeOrdinal, 0);
+      int node1 = cellTopo.getNodeMap(edgeDim, edgeOrdinal, 1);
+      edges[node0].push_back(node1);
+      edges[node1].push_back(node0);
     }
-  };
-
-  class UniqueNumbering
-  {
-    vector<vector<double>> _knownCoords; // x,y,z, depending on spatial extent; inner vector is sorted
-    double _tol; // what counts as a match
-    map<vector<double>, int> _numbering; // maps from tuple selected from the sets in _knownCoords to unique identifier for tuple
-    void getSanitizedCoords(const vector<double> &coords, vector<double> &sanitizedCoords)
-    {
-      sanitizedCoords.resize(coords.size());
-      for (int d=0; d<int(coords.size()); d++)
-      {
-        double val = coords[d];
-        // look to lower_bound to see if we already have something within _tol
-        auto lowerBoundIt = lower_bound(_knownCoords[d].begin(), _knownCoords[d].end(), val);
-        // lower_bound returns iterator to first value that does not compare less than val (i.e. lower_bound >= val)
-        if ((lowerBoundIt != _knownCoords[d].end()) && ((*lowerBoundIt - val) < _tol))
-        {
-          sanitizedCoords[d] = *lowerBoundIt;
-        }
-        else
-        {
-          bool shouldInsert = true; // unless we find that the previous entry is a match
-          if (lowerBoundIt != _knownCoords[d].begin())
-          {
-            // then check the prior guy to see if he's within _tol
-            double previousEntry = *(lowerBoundIt-1);
-            if ((val-previousEntry) < _tol)
-            {
-              sanitizedCoords[d] = previousEntry;
-              shouldInsert = false;
-            }
-          }
-          if (shouldInsert)
-          {
-            // neither prior nor following entry is within _tol
-            // add to our list:
-            _knownCoords[d].insert(lowerBoundIt, val);
-            sanitizedCoords[d] = val;
+    return symmetries(edges);
+  }
+};
+
+class UniqueNumbering {
+  vector<vector<double>> _knownCoords;  // x,y,z, depending on spatial extent; inner vector is sorted
+  double _tol;                          // what counts as a match
+  map<vector<double>, int> _numbering;  // maps from tuple selected from the sets in _knownCoords to unique identifier for tuple
+  void getSanitizedCoords(const vector<double> &coords, vector<double> &sanitizedCoords) {
+    sanitizedCoords.resize(coords.size());
+    for (int d = 0; d < int(coords.size()); d++) {
+      double val = coords[d];
+      // look to lower_bound to see if we already have something within _tol
+      auto lowerBoundIt = lower_bound(_knownCoords[d].begin(), _knownCoords[d].end(), val);
+      // lower_bound returns iterator to first value that does not compare less than val (i.e. lower_bound >= val)
+      if ((lowerBoundIt != _knownCoords[d].end()) && ((*lowerBoundIt - val) < _tol)) {
+        sanitizedCoords[d] = *lowerBoundIt;
+      } else {
+        bool shouldInsert = true;  // unless we find that the previous entry is a match
+        if (lowerBoundIt != _knownCoords[d].begin()) {
+          // then check the prior guy to see if he's within _tol
+          double previousEntry = *(lowerBoundIt - 1);
+          if ((val - previousEntry) < _tol) {
+            sanitizedCoords[d] = previousEntry;
+            shouldInsert       = false;
           }
         }
+        if (shouldInsert) {
+          // neither prior nor following entry is within _tol
+          // add to our list:
+          _knownCoords[d].insert(lowerBoundIt, val);
+          sanitizedCoords[d] = val;
+        }
       }
     }
+  }
 
-  public:
-    UniqueNumbering(int spaceDim, double tol)
-    {
-      _tol = tol;
-      _knownCoords = vector<vector<double>>(spaceDim);
-    }
-    template<class ArrayScalar, class ArrayOrdinal>
-    void getIDs(const ArrayScalar &points, ArrayOrdinal &globalIDs)
-    {
-
-      auto points_host = Kokkos::create_mirror_view(points);
-      auto globalIDs_host = Kokkos::create_mirror_view(globalIDs);
-      Kokkos::deep_copy(points_host, points);
-      int spaceDim = _knownCoords.size();
-      TEUCHOS_TEST_FOR_EXCEPTION(spaceDim != (int) points.extent(points.rank()-1), std::invalid_argument, "final extent of points container must equal spaceDim");
-      if (points.rank() == 2)
-      {
-        int numPoints = points.extent(0);
-        for (int pointOrdinal=0; pointOrdinal<numPoints; pointOrdinal++)
-        {
-          vector<double> coords(spaceDim);
-          for (int d=0; d<spaceDim; d++)
-          {
-            coords[d] = points_host(pointOrdinal,d);
-          }
-          globalIDs_host(pointOrdinal) = getGlobalID(coords);
+ public:
+  UniqueNumbering(int spaceDim, double tol) {
+    _tol         = tol;
+    _knownCoords = vector<vector<double>>(spaceDim);
+  }
+  template <class ArrayScalar, class ArrayOrdinal>
+  void getIDs(const ArrayScalar &points, ArrayOrdinal &globalIDs) {
+    auto points_host    = Kokkos::create_mirror_view(points);
+    auto globalIDs_host = Kokkos::create_mirror_view(globalIDs);
+    Kokkos::deep_copy(points_host, points);
+    int spaceDim = _knownCoords.size();
+    TEUCHOS_TEST_FOR_EXCEPTION(spaceDim != (int)points.extent(points.rank() - 1), std::invalid_argument, "final extent of points container must equal spaceDim");
+    if (points.rank() == 2) {
+      int numPoints = points.extent(0);
+      for (int pointOrdinal = 0; pointOrdinal < numPoints; pointOrdinal++) {
+        vector<double> coords(spaceDim);
+        for (int d = 0; d < spaceDim; d++) {
+          coords[d] = points_host(pointOrdinal, d);
         }
+        globalIDs_host(pointOrdinal) = getGlobalID(coords);
       }
-      else if (points.rank() == 3)
-      {
-        int numCells = points.extent(0);
-        int numPoints = points.extent(1);
-        for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-        {
-          for (int pointOrdinal=0; pointOrdinal<numPoints; pointOrdinal++)
-          {
-            vector<double> coords(spaceDim);
-            for (int d=0; d<spaceDim; d++)
-            {
-              coords[d] = points_host(cellOrdinal,pointOrdinal,d);
-            }
-            globalIDs_host(cellOrdinal,pointOrdinal) = getGlobalID(coords);
+    } else if (points.rank() == 3) {
+      int numCells  = points.extent(0);
+      int numPoints = points.extent(1);
+      for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+        for (int pointOrdinal = 0; pointOrdinal < numPoints; pointOrdinal++) {
+          vector<double> coords(spaceDim);
+          for (int d = 0; d < spaceDim; d++) {
+            coords[d] = points_host(cellOrdinal, pointOrdinal, d);
           }
+          globalIDs_host(cellOrdinal, pointOrdinal) = getGlobalID(coords);
         }
       }
-      else
-      {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "points must be a rank 2 or rank 3 container");
-      }
-      Kokkos::deep_copy(globalIDs, globalIDs_host);
-    }
-    int getGlobalID(const vector<double> &coords)
-    {
-      TEUCHOS_TEST_FOR_EXCEPTION(coords.size() != _knownCoords.size(), std::invalid_argument,
-                                 "coords must be the same size as spaceDim!");
-      // we first filter ("sanitize") the coords to match _knownCoords within _tol
-      // (adding entries to _knownCoords as needed)
-      vector<double> coordsKey;
-      getSanitizedCoords(coords, coordsKey);
-
-      // we look up using the sanitized coordsKey
-      if (_numbering.find(coordsKey) == _numbering.end())
-      {
-        int newID = _numbering.size();
-        _numbering[coordsKey] = newID;
-      }
-      return _numbering[coordsKey];
+    } else {
+      TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "points must be a rank 2 or rank 3 container");
     }
-    int totalCount()
-    {
-      return _numbering.size();
-    }
-  };
-
-  using namespace std;
-  using namespace Teuchos;
-
-  template<class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
-  void buildSampleElementToNodeMapThreeElements(RCP<Basis> basis, ArrayOrdinal &elemToNodeMap,
-                                                vector<vector<int>> &ordinalsForSubcellDimension,
-                                                vector<int> &subcellCountForDimension)
-  {
-    using namespace Kokkos::Experimental;
-    typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
-    double pointTol = 1e-10;
-
-    ArrayScalar cellWorkset;
-    shards::CellTopology cellTopo = basis->getBaseCellTopology();
-    int numCells = 3;
-    int vertexCount = (int) cellTopo.getVertexCount();
-    int spaceDim = (int) cellTopo.getDimension();
-
-    ArrayScalar refDofCoords, cellDofCoords;
-    ArrayOrdinal cellDofIDs; // for each cell, stores the global ordinal assigned to local ordinal
-    resize(cellWorkset,numCells,vertexCount,spaceDim);
-    resize(refDofCoords,basis->getCardinality(),spaceDim);
-    resize(cellDofCoords,numCells,basis->getCardinality(),spaceDim);
-    resize(cellDofIDs,numCells,basis->getCardinality());
-    resize(elemToNodeMap,numCells,basis->getCardinality());
-
-    basis->getDofCoords(refDofCoords);
-
-    subcellCountForDimension.resize(spaceDim+1);
-    subcellCountForDimension[spaceDim] = numCells;
-    auto cellWorkset_host = Kokkos::create_mirror_view(cellWorkset);
-    if (spaceDim == 1)
-    {
-      // line
-      cellWorkset_host(0,0,0) = 0.0;
-      cellWorkset_host(0,1,0) = 1.0;
-      cellWorkset_host(1,0,0) = 1.0;
-      cellWorkset_host(1,1,0) = 2.0;
-      cellWorkset_host(2,0,0) = 2.0;
-      cellWorkset_host(2,1,0) = 3.0;
-      subcellCountForDimension[0] = 4; // 3 cells x 2 vertices - (2 interior vertices)
-    }
-    else if ((spaceDim == 2) && (vertexCount == 4))
-    {
-      // quad
-      // first element: LL @ (0,1), UR @ (1,2)
-      cellWorkset_host(0,0,0) = 0.0;
-      cellWorkset_host(0,0,1) = 1.0;
-      cellWorkset_host(0,1,0) = 1.0;
-      cellWorkset_host(0,1,1) = 1.0;
-      cellWorkset_host(0,2,0) = 1.0;
-      cellWorkset_host(0,2,1) = 2.0;
-      cellWorkset_host(0,3,0) = 0.0;
-      cellWorkset_host(0,3,1) = 2.0;
-      // second element: LL @ (0,0), UR @ (1,1)
-      cellWorkset_host(1,0,0) = 0.0;
-      cellWorkset_host(1,0,1) = 0.0;
-      cellWorkset_host(1,1,0) = 1.0;
-      cellWorkset_host(1,1,1) = 0.0;
-      cellWorkset_host(1,2,0) = 1.0;
-      cellWorkset_host(1,2,1) = 1.0;
-      cellWorkset_host(1,3,0) = 0.0;
-      cellWorkset_host(1,3,1) = 1.0;
-      // third element: LL @ (1,0), UR @ (2,1)
-      cellWorkset_host(2,0,0) = 1.0;
-      cellWorkset_host(2,0,1) = 0.0;
-      cellWorkset_host(2,1,0) = 2.0;
-      cellWorkset_host(2,1,1) = 0.0;
-      cellWorkset_host(2,2,0) = 2.0;
-      cellWorkset_host(2,2,1) = 1.0;
-      cellWorkset_host(2,3,0) = 1.0;
-      cellWorkset_host(2,3,1) = 1.0;
-      subcellCountForDimension[0] = 12-2-1*2; // 3 cells x 4 vertices - (2 vertices seen by 2 cells, 1 vertex seen by 3 cells)
-      subcellCountForDimension[1] = 12-2;   // 3 cells x 4 edges - (2 shared edges)
-    }
-    else if ((spaceDim == 3) && (vertexCount == 8))
-    {
-      // hex: same geometry as quad, but extruded in z from 0 to 1
-      // first element: LL @ (0,1), UR @ (1,2)
-      cellWorkset_host(0,0,0) = 0.0;
-      cellWorkset_host(0,0,1) = 1.0;
-      cellWorkset_host(0,0,2) = 0.0;
-      cellWorkset_host(0,1,0) = 1.0;
-      cellWorkset_host(0,1,1) = 1.0;
-      cellWorkset_host(0,1,2) = 0.0;
-      cellWorkset_host(0,2,0) = 1.0;
-      cellWorkset_host(0,2,1) = 2.0;
-      cellWorkset_host(0,2,2) = 0.0;
-      cellWorkset_host(0,3,0) = 0.0;
-      cellWorkset_host(0,3,1) = 2.0;
-      cellWorkset_host(0,3,2) = 0.0;
-      cellWorkset_host(0,4,0) = 0.0;
-      cellWorkset_host(0,4,1) = 1.0;
-      cellWorkset_host(0,4,2) = 1.0;
-      cellWorkset_host(0,5,0) = 1.0;
-      cellWorkset_host(0,5,1) = 1.0;
-      cellWorkset_host(0,5,2) = 1.0;
-      cellWorkset_host(0,6,0) = 1.0;
-      cellWorkset_host(0,6,1) = 2.0;
-      cellWorkset_host(0,6,2) = 1.0;
-      cellWorkset_host(0,7,0) = 0.0;
-      cellWorkset_host(0,7,1) = 2.0;
-      cellWorkset_host(0,7,2) = 1.0;
-      // second element: LL @ (0,0), UR @ (1,1)
-      cellWorkset_host(1,0,0) = 0.0;
-      cellWorkset_host(1,0,1) = 0.0;
-      cellWorkset_host(1,0,2) = 0.0;
-      cellWorkset_host(1,1,0) = 1.0;
-      cellWorkset_host(1,1,1) = 0.0;
-      cellWorkset_host(1,1,2) = 0.0;
-      cellWorkset_host(1,2,0) = 1.0;
-      cellWorkset_host(1,2,1) = 1.0;
-      cellWorkset_host(1,2,2) = 0.0;
-      cellWorkset_host(1,3,0) = 0.0;
-      cellWorkset_host(1,3,1) = 1.0;
-      cellWorkset_host(1,3,2) = 0.0;
-      cellWorkset_host(1,4,0) = 0.0;
-      cellWorkset_host(1,4,1) = 0.0;
-      cellWorkset_host(1,4,2) = 1.0;
-      cellWorkset_host(1,5,0) = 1.0;
-      cellWorkset_host(1,5,1) = 0.0;
-      cellWorkset_host(1,5,2) = 1.0;
-      cellWorkset_host(1,6,0) = 1.0;
-      cellWorkset_host(1,6,1) = 1.0;
-      cellWorkset_host(1,6,2) = 1.0;
-      cellWorkset_host(1,7,0) = 0.0;
-      cellWorkset_host(1,7,1) = 1.0;
-      cellWorkset_host(1,7,2) = 1.0;
-      // third element: LL @ (1,0), UR @ (2,1)
-      cellWorkset_host(2,0,0) = 1.0;
-      cellWorkset_host(2,0,1) = 0.0;
-      cellWorkset_host(2,0,2) = 0.0;
-      cellWorkset_host(2,1,0) = 2.0;
-      cellWorkset_host(2,1,1) = 0.0;
-      cellWorkset_host(2,1,2) = 0.0;
-      cellWorkset_host(2,2,0) = 2.0;
-      cellWorkset_host(2,2,1) = 1.0;
-      cellWorkset_host(2,2,2) = 0.0;
-      cellWorkset_host(2,3,0) = 1.0;
-      cellWorkset_host(2,3,1) = 1.0;
-      cellWorkset_host(2,3,2) = 0.0;
-      cellWorkset_host(2,4,0) = 1.0;
-      cellWorkset_host(2,4,1) = 0.0;
-      cellWorkset_host(2,4,2) = 1.0;
-      cellWorkset_host(2,5,0) = 2.0;
-      cellWorkset_host(2,5,1) = 0.0;
-      cellWorkset_host(2,5,2) = 1.0;
-      cellWorkset_host(2,6,0) = 2.0;
-      cellWorkset_host(2,6,1) = 1.0;
-      cellWorkset_host(2,6,2) = 1.0;
-      cellWorkset_host(2,7,0) = 1.0;
-      cellWorkset_host(2,7,1) = 1.0;
-      cellWorkset_host(2,7,2) = 1.0;
-      subcellCountForDimension[0] = 24-4*1-2*2; // 3 cells x 8 vertices - (4 vertices seen by 2 cells, 2 vertices seen by 3 cells)
-      subcellCountForDimension[1] = 36-6*1-1*2; // 3 cells x 12 edges - (6 edges seen by 2 cells, 1 edge seen by 3 cells)
-      subcellCountForDimension[2] = 18-2*1;     // 3 cells x 6 faces - (2 faces seen by 2 cells)
-    }
-    else
-    {
-      // we haven't yet set up test geometry for this CellTopology
-      TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "Unimplemented test case");
+    Kokkos::deep_copy(globalIDs, globalIDs_host);
+  }
+  int getGlobalID(const vector<double> &coords) {
+    TEUCHOS_TEST_FOR_EXCEPTION(coords.size() != _knownCoords.size(), std::invalid_argument,
+                               "coords must be the same size as spaceDim!");
+    // we first filter ("sanitize") the coords to match _knownCoords within _tol
+    // (adding entries to _knownCoords as needed)
+    vector<double> coordsKey;
+    getSanitizedCoords(coords, coordsKey);
+
+    // we look up using the sanitized coordsKey
+    if (_numbering.find(coordsKey) == _numbering.end()) {
+      int newID             = _numbering.size();
+      _numbering[coordsKey] = newID;
     }
+    return _numbering[coordsKey];
+  }
+  int totalCount() {
+    return _numbering.size();
+  }
+};
+
+using namespace std;
+using namespace Teuchos;
+
+template <class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
+void buildSampleElementToNodeMapThreeElements(RCP<Basis> basis, ArrayOrdinal &elemToNodeMap,
+                                              vector<vector<int>> &ordinalsForSubcellDimension,
+                                              vector<int> &subcellCountForDimension) {
+  using namespace Kokkos::Experimental;
+  typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
+  double pointTol = 1e-10;
+
+  ArrayScalar cellWorkset;
+  shards::CellTopology cellTopo = basis->getBaseCellTopology();
+  int numCells                  = 3;
+  int vertexCount               = (int)cellTopo.getVertexCount();
+  int spaceDim                  = (int)cellTopo.getDimension();
+
+  ArrayScalar refDofCoords, cellDofCoords;
+  ArrayOrdinal cellDofIDs;  // for each cell, stores the global ordinal assigned to local ordinal
+  resize(cellWorkset, numCells, vertexCount, spaceDim);
+  resize(refDofCoords, basis->getCardinality(), spaceDim);
+  resize(cellDofCoords, numCells, basis->getCardinality(), spaceDim);
+  resize(cellDofIDs, numCells, basis->getCardinality());
+  resize(elemToNodeMap, numCells, basis->getCardinality());
+
+  basis->getDofCoords(refDofCoords);
+
+  subcellCountForDimension.resize(spaceDim + 1);
+  subcellCountForDimension[spaceDim] = numCells;
+  auto cellWorkset_host              = Kokkos::create_mirror_view(cellWorkset);
+  if (spaceDim == 1) {
+    // line
+    cellWorkset_host(0, 0, 0)   = 0.0;
+    cellWorkset_host(0, 1, 0)   = 1.0;
+    cellWorkset_host(1, 0, 0)   = 1.0;
+    cellWorkset_host(1, 1, 0)   = 2.0;
+    cellWorkset_host(2, 0, 0)   = 2.0;
+    cellWorkset_host(2, 1, 0)   = 3.0;
+    subcellCountForDimension[0] = 4;  // 3 cells x 2 vertices - (2 interior vertices)
+  } else if ((spaceDim == 2) && (vertexCount == 4)) {
+    // quad
+    // first element: LL @ (0,1), UR @ (1,2)
+    cellWorkset_host(0, 0, 0) = 0.0;
+    cellWorkset_host(0, 0, 1) = 1.0;
+    cellWorkset_host(0, 1, 0) = 1.0;
+    cellWorkset_host(0, 1, 1) = 1.0;
+    cellWorkset_host(0, 2, 0) = 1.0;
+    cellWorkset_host(0, 2, 1) = 2.0;
+    cellWorkset_host(0, 3, 0) = 0.0;
+    cellWorkset_host(0, 3, 1) = 2.0;
+    // second element: LL @ (0,0), UR @ (1,1)
+    cellWorkset_host(1, 0, 0) = 0.0;
+    cellWorkset_host(1, 0, 1) = 0.0;
+    cellWorkset_host(1, 1, 0) = 1.0;
+    cellWorkset_host(1, 1, 1) = 0.0;
+    cellWorkset_host(1, 2, 0) = 1.0;
+    cellWorkset_host(1, 2, 1) = 1.0;
+    cellWorkset_host(1, 3, 0) = 0.0;
+    cellWorkset_host(1, 3, 1) = 1.0;
+    // third element: LL @ (1,0), UR @ (2,1)
+    cellWorkset_host(2, 0, 0)   = 1.0;
+    cellWorkset_host(2, 0, 1)   = 0.0;
+    cellWorkset_host(2, 1, 0)   = 2.0;
+    cellWorkset_host(2, 1, 1)   = 0.0;
+    cellWorkset_host(2, 2, 0)   = 2.0;
+    cellWorkset_host(2, 2, 1)   = 1.0;
+    cellWorkset_host(2, 3, 0)   = 1.0;
+    cellWorkset_host(2, 3, 1)   = 1.0;
+    subcellCountForDimension[0] = 12 - 2 - 1 * 2;  // 3 cells x 4 vertices - (2 vertices seen by 2 cells, 1 vertex seen by 3 cells)
+    subcellCountForDimension[1] = 12 - 2;          // 3 cells x 4 edges - (2 shared edges)
+  } else if ((spaceDim == 3) && (vertexCount == 8)) {
+    // hex: same geometry as quad, but extruded in z from 0 to 1
+    // first element: LL @ (0,1), UR @ (1,2)
+    cellWorkset_host(0, 0, 0) = 0.0;
+    cellWorkset_host(0, 0, 1) = 1.0;
+    cellWorkset_host(0, 0, 2) = 0.0;
+    cellWorkset_host(0, 1, 0) = 1.0;
+    cellWorkset_host(0, 1, 1) = 1.0;
+    cellWorkset_host(0, 1, 2) = 0.0;
+    cellWorkset_host(0, 2, 0) = 1.0;
+    cellWorkset_host(0, 2, 1) = 2.0;
+    cellWorkset_host(0, 2, 2) = 0.0;
+    cellWorkset_host(0, 3, 0) = 0.0;
+    cellWorkset_host(0, 3, 1) = 2.0;
+    cellWorkset_host(0, 3, 2) = 0.0;
+    cellWorkset_host(0, 4, 0) = 0.0;
+    cellWorkset_host(0, 4, 1) = 1.0;
+    cellWorkset_host(0, 4, 2) = 1.0;
+    cellWorkset_host(0, 5, 0) = 1.0;
+    cellWorkset_host(0, 5, 1) = 1.0;
+    cellWorkset_host(0, 5, 2) = 1.0;
+    cellWorkset_host(0, 6, 0) = 1.0;
+    cellWorkset_host(0, 6, 1) = 2.0;
+    cellWorkset_host(0, 6, 2) = 1.0;
+    cellWorkset_host(0, 7, 0) = 0.0;
+    cellWorkset_host(0, 7, 1) = 2.0;
+    cellWorkset_host(0, 7, 2) = 1.0;
+    // second element: LL @ (0,0), UR @ (1,1)
+    cellWorkset_host(1, 0, 0) = 0.0;
+    cellWorkset_host(1, 0, 1) = 0.0;
+    cellWorkset_host(1, 0, 2) = 0.0;
+    cellWorkset_host(1, 1, 0) = 1.0;
+    cellWorkset_host(1, 1, 1) = 0.0;
+    cellWorkset_host(1, 1, 2) = 0.0;
+    cellWorkset_host(1, 2, 0) = 1.0;
+    cellWorkset_host(1, 2, 1) = 1.0;
+    cellWorkset_host(1, 2, 2) = 0.0;
+    cellWorkset_host(1, 3, 0) = 0.0;
+    cellWorkset_host(1, 3, 1) = 1.0;
+    cellWorkset_host(1, 3, 2) = 0.0;
+    cellWorkset_host(1, 4, 0) = 0.0;
+    cellWorkset_host(1, 4, 1) = 0.0;
+    cellWorkset_host(1, 4, 2) = 1.0;
+    cellWorkset_host(1, 5, 0) = 1.0;
+    cellWorkset_host(1, 5, 1) = 0.0;
+    cellWorkset_host(1, 5, 2) = 1.0;
+    cellWorkset_host(1, 6, 0) = 1.0;
+    cellWorkset_host(1, 6, 1) = 1.0;
+    cellWorkset_host(1, 6, 2) = 1.0;
+    cellWorkset_host(1, 7, 0) = 0.0;
+    cellWorkset_host(1, 7, 1) = 1.0;
+    cellWorkset_host(1, 7, 2) = 1.0;
+    // third element: LL @ (1,0), UR @ (2,1)
+    cellWorkset_host(2, 0, 0)   = 1.0;
+    cellWorkset_host(2, 0, 1)   = 0.0;
+    cellWorkset_host(2, 0, 2)   = 0.0;
+    cellWorkset_host(2, 1, 0)   = 2.0;
+    cellWorkset_host(2, 1, 1)   = 0.0;
+    cellWorkset_host(2, 1, 2)   = 0.0;
+    cellWorkset_host(2, 2, 0)   = 2.0;
+    cellWorkset_host(2, 2, 1)   = 1.0;
+    cellWorkset_host(2, 2, 2)   = 0.0;
+    cellWorkset_host(2, 3, 0)   = 1.0;
+    cellWorkset_host(2, 3, 1)   = 1.0;
+    cellWorkset_host(2, 3, 2)   = 0.0;
+    cellWorkset_host(2, 4, 0)   = 1.0;
+    cellWorkset_host(2, 4, 1)   = 0.0;
+    cellWorkset_host(2, 4, 2)   = 1.0;
+    cellWorkset_host(2, 5, 0)   = 2.0;
+    cellWorkset_host(2, 5, 1)   = 0.0;
+    cellWorkset_host(2, 5, 2)   = 1.0;
+    cellWorkset_host(2, 6, 0)   = 2.0;
+    cellWorkset_host(2, 6, 1)   = 1.0;
+    cellWorkset_host(2, 6, 2)   = 1.0;
+    cellWorkset_host(2, 7, 0)   = 1.0;
+    cellWorkset_host(2, 7, 1)   = 1.0;
+    cellWorkset_host(2, 7, 2)   = 1.0;
+    subcellCountForDimension[0] = 24 - 4 * 1 - 2 * 2;  // 3 cells x 8 vertices - (4 vertices seen by 2 cells, 2 vertices seen by 3 cells)
+    subcellCountForDimension[1] = 36 - 6 * 1 - 1 * 2;  // 3 cells x 12 edges - (6 edges seen by 2 cells, 1 edge seen by 3 cells)
+    subcellCountForDimension[2] = 18 - 2 * 1;          // 3 cells x 6 faces - (2 faces seen by 2 cells)
+  } else {
+    // we haven't yet set up test geometry for this CellTopology
+    TEUCHOS_TEST_FOR_EXCEPTION(true, std::invalid_argument, "Unimplemented test case");
+  }
 
-    Kokkos::deep_copy(cellWorkset, cellWorkset_host);
-    CellTools::mapToPhysicalFrame(cellDofCoords, refDofCoords, cellWorkset, cellTopo);
+  Kokkos::deep_copy(cellWorkset, cellWorkset_host);
+  CellTools::mapToPhysicalFrame(cellDofCoords, refDofCoords, cellWorkset, cellTopo);
 
-    Kokkos::fence(); // mapToPhysicalFrame calls getValues which calls kernels, so fence is required before UVM reads below
+  Kokkos::fence();  // mapToPhysicalFrame calls getValues which calls kernels, so fence is required before UVM reads below
 
-    UniqueNumbering globalNumbering(spaceDim, pointTol);
-    globalNumbering.getIDs<ArrayScalar,ArrayOrdinal>(cellDofCoords,cellDofIDs);
+  UniqueNumbering globalNumbering(spaceDim, pointTol);
+  globalNumbering.getIDs<ArrayScalar, ArrayOrdinal>(cellDofCoords, cellDofIDs);
 
-    int tagOrdSubcellDim = 0; // where to find the subcell extent within the tag
-    int numDofsPerCell = basis->getCardinality();
+  int tagOrdSubcellDim = 0;  // where to find the subcell extent within the tag
+  int numDofsPerCell   = basis->getCardinality();
 
-    // store ordinals in a set for easy uniquing
-    vector<set<int>> ordinalsForSubcellDimensionSet(spaceDim+1);
+  // store ordinals in a set for easy uniquing
+  vector<set<int>> ordinalsForSubcellDimensionSet(spaceDim + 1);
 
-    auto cellDofIDs_host = Kokkos::create_mirror_view(cellDofIDs);
-    auto elemToNodeMap_host = Kokkos::create_mirror_view(elemToNodeMap);
-    Kokkos::deep_copy(cellDofIDs_host,cellDofIDs);
-    for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-    {
-      for (int dofOrdinal=0; dofOrdinal<numDofsPerCell; dofOrdinal++)
-      {
-        int subcellDofDim = basis->getDofTag(dofOrdinal)[tagOrdSubcellDim];
-        int globalID = cellDofIDs_host(cellOrdinal,dofOrdinal);
-        ordinalsForSubcellDimensionSet[subcellDofDim].insert(globalID);
-        elemToNodeMap_host(cellOrdinal,dofOrdinal) = globalID;
-      }
+  auto cellDofIDs_host    = Kokkos::create_mirror_view(cellDofIDs);
+  auto elemToNodeMap_host = Kokkos::create_mirror_view(elemToNodeMap);
+  Kokkos::deep_copy(cellDofIDs_host, cellDofIDs);
+  for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+    for (int dofOrdinal = 0; dofOrdinal < numDofsPerCell; dofOrdinal++) {
+      int subcellDofDim = basis->getDofTag(dofOrdinal)[tagOrdSubcellDim];
+      int globalID      = cellDofIDs_host(cellOrdinal, dofOrdinal);
+      ordinalsForSubcellDimensionSet[subcellDofDim].insert(globalID);
+      elemToNodeMap_host(cellOrdinal, dofOrdinal) = globalID;
     }
-    Kokkos::deep_copy(elemToNodeMap, elemToNodeMap_host);
-    ordinalsForSubcellDimension.clear();
-    for (int d=0; d<spaceDim+1; d++)
-    {
-      vector<int> ordinals(ordinalsForSubcellDimensionSet[d].begin(),ordinalsForSubcellDimensionSet[d].end());
-      ordinalsForSubcellDimension.push_back(ordinals);
+  }
+  Kokkos::deep_copy(elemToNodeMap, elemToNodeMap_host);
+  ordinalsForSubcellDimension.clear();
+  for (int d = 0; d < spaceDim + 1; d++) {
+    vector<int> ordinals(ordinalsForSubcellDimensionSet[d].begin(), ordinalsForSubcellDimensionSet[d].end());
+    ordinalsForSubcellDimension.push_back(ordinals);
+  }
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void testUniqueNumbering(Teuchos::FancyOStream &out, bool &success) {
+  // simple test with dof coords corresponding to our sample three-element quad mesh with a quadratic basis
+  typedef typename Node::device_type::execution_space ES;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Teuchos::ScalarTraits<LocalOrdinal>::magnitudeType OT;  // ordinal type
+
+  using namespace Kokkos::Experimental;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<OT, typename Node::device_type> FCO;  // FC of ordinals
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> Basis;
+
+  typename FC::HostMirror physDofCoords;
+  typename FCO::HostMirror cellGIDs;
+  std::vector<std::vector<LocalOrdinal>> expectedGIDs = {{0, 1, 2, 3, 4, 5, 6, 7, 8},
+                                                         {9, 10, 11, 12, 13, 14, 0, 1, 2},
+                                                         {11, 15, 16, 14, 17, 18, 2, 19, 20}};
+  int numCells = 3, numPointsPerCell = 9, spaceDim = 2;
+  resize(physDofCoords, numCells, numPointsPerCell, spaceDim);
+  resize(cellGIDs, numCells, numPointsPerCell);
+
+  // x,y coords of lower-left vertices for each cell:
+  std::vector<double> x0s = {0.0, 0.0, 1.0};
+  std::vector<double> y0s = {1.0, 0.0, 0.0};
+  for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+    double x0 = x0s[cellOrdinal], y0 = y0s[cellOrdinal];
+    for (int pointOrdinal = 0; pointOrdinal < numPointsPerCell; pointOrdinal++) {
+      double xOffset                              = (pointOrdinal % 3) * 0.5;  // 0, 0.5, 1.0
+      double yOffset                              = (pointOrdinal / 3) * 0.5;  // 0, 0.5, 1.0
+      physDofCoords(cellOrdinal, pointOrdinal, 0) = x0 + xOffset;
+      physDofCoords(cellOrdinal, pointOrdinal, 1) = y0 + yOffset;
     }
   }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void testUniqueNumbering(Teuchos::FancyOStream &out, bool &success)
-  {
-    // simple test with dof coords corresponding to our sample three-element quad mesh with a quadratic basis
-    typedef typename Node::device_type::execution_space ES;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Teuchos::ScalarTraits<LocalOrdinal>::magnitudeType OT; // ordinal type
-
-    using namespace Kokkos::Experimental;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<OT,typename Node::device_type> FCO; // FC of ordinals
-    typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> Basis;
-
-    typename FC::HostMirror physDofCoords;
-    typename FCO::HostMirror cellGIDs;
-    std::vector<std::vector<LocalOrdinal> > expectedGIDs = {{ 0, 1, 2, 3, 4, 5, 6, 7, 8},
-      { 9,10,11,12,13,14, 0, 1, 2},
-      {11,15,16,14,17,18, 2,19,20}};
-    int numCells = 3, numPointsPerCell = 9, spaceDim = 2;
-    resize(physDofCoords,numCells,numPointsPerCell,spaceDim);
-    resize(cellGIDs,numCells,numPointsPerCell);
-
-    // x,y coords of lower-left vertices for each cell:
-    std::vector<double> x0s = {0.0,0.0,1.0};
-    std::vector<double> y0s = {1.0,0.0,0.0};
-    for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-    {
-      double x0 = x0s[cellOrdinal], y0 = y0s[cellOrdinal];
-      for (int pointOrdinal=0; pointOrdinal<numPointsPerCell; pointOrdinal++)
-      {
-        double xOffset = (pointOrdinal % 3) * 0.5; // 0, 0.5, 1.0
-        double yOffset = (pointOrdinal / 3) * 0.5; // 0, 0.5, 1.0
-        physDofCoords(cellOrdinal,pointOrdinal,0) = x0 + xOffset;
-        physDofCoords(cellOrdinal,pointOrdinal,1) = y0 + yOffset;
+  double pointTol = 1e-10;
+  UniqueNumbering numbering(spaceDim, pointTol);
+  numbering.getIDs(physDofCoords, cellGIDs);
+
+  for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+    out << "cell " << cellOrdinal << " GIDs: {";
+    for (int pointOrdinal = 0; pointOrdinal < numPointsPerCell; pointOrdinal++) {
+      out << " " << cellGIDs(cellOrdinal, pointOrdinal);
+      if (cellGIDs(cellOrdinal, pointOrdinal) != expectedGIDs[cellOrdinal][pointOrdinal]) {
+        success = false;
       }
     }
-    double pointTol = 1e-10;
-    UniqueNumbering numbering(spaceDim,pointTol);
-    numbering.getIDs(physDofCoords, cellGIDs);
-
-    for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-    {
-      out << "cell " << cellOrdinal << " GIDs: {";
-      for (int pointOrdinal=0; pointOrdinal<numPointsPerCell; pointOrdinal++)
-      {
-        out << " " << cellGIDs(cellOrdinal,pointOrdinal);
-        if (cellGIDs(cellOrdinal,pointOrdinal) != expectedGIDs[cellOrdinal][pointOrdinal])
-        {
-          success = false;
+    out << " }\n";
+  }
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, UniqueNumbering, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  ;
+  testUniqueNumbering<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void testBuildSampleElementToNodeMapThreeElementQuad(Teuchos::FancyOStream &out, bool &success) {
+  // simple test with quadratic basis on quads: check that we have the right numbering (hard-coded)
+  std::vector<int> vertexGIDs = {0, 2, 6, 8, 9, 11, 16, 20};           // should be 8 of these
+  std::vector<int> edgeGIDs   = {1, 3, 5, 7, 10, 12, 14, 15, 18, 19};  // 10 of these
+  std::vector<int> cellGIDs   = {4, 13, 17};
+
+  vector<vector<int>> expectedGIDs = {vertexGIDs, edgeGIDs, cellGIDs};
+
+  typedef typename Node::device_type::execution_space ES;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Teuchos::ScalarTraits<LocalOrdinal>::magnitudeType OT;  // ordinal type
+
+  using namespace Kokkos::Experimental;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<OT, typename Node::device_type> FCO;  // FC of ordinals
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> Basis;
+
+  int polyOrder = 2, spaceDim = 2;
+  RCP<Basis> basis = rcp(new Basis(polyOrder, Intrepid2::EPointType::POINTTYPE_EQUISPACED));
+  std::vector<int> subcellCountForDimension;
+  std::vector<std::vector<int>> ordinalsForSubcellDimension;
+  FCO elementToNodeMap;
+  buildSampleElementToNodeMapThreeElements<Basis, ES, FC, FCO>(basis, elementToNodeMap, ordinalsForSubcellDimension,
+                                                               subcellCountForDimension);
+  // check subcell counts first
+  for (int d = 0; d <= spaceDim; d++) {
+    if (subcellCountForDimension[d] != int(expectedGIDs[d].size())) {
+      success = false;
+      out << "subcellCountForDimension[" << d << "] != " << expectedGIDs[d].size() << endl;
+    } else {
+      int dofCount     = ordinalsForSubcellDimension[d].size();
+      bool listsDiffer = ordinalsForSubcellDimension[d].size() != expectedGIDs[d].size();
+
+      if (!listsDiffer) {
+        for (int dofOrdinal = 0; dofOrdinal < dofCount; dofOrdinal++) {
+          if (ordinalsForSubcellDimension[d][dofOrdinal] != expectedGIDs[d][dofOrdinal]) {
+            listsDiffer = true;
+            break;
+          }
         }
       }
-      out << " }\n";
-    }
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, UniqueNumbering, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    ;
-    testUniqueNumbering<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,success);
-  }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void testBuildSampleElementToNodeMapThreeElementQuad(Teuchos::FancyOStream &out, bool &success)
-  {
-    // simple test with quadratic basis on quads: check that we have the right numbering (hard-coded)
-    std::vector<int> vertexGIDs = {0,2,6,8,9,11,16,20}; // should be 8 of these
-    std::vector<int> edgeGIDs = {1,3,5,7,10,12,14,15,18,19}; // 10 of these
-    std::vector<int> cellGIDs = {4,13,17};
-
-    vector<vector<int>> expectedGIDs = {vertexGIDs,edgeGIDs,cellGIDs};
-
-    typedef typename Node::device_type::execution_space ES;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Teuchos::ScalarTraits<LocalOrdinal>::magnitudeType OT; // ordinal type
-
-    using namespace Kokkos::Experimental;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<OT,typename Node::device_type> FCO; // FC of ordinals
-    typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> Basis;
-
-    int polyOrder = 2, spaceDim = 2;
-    RCP<Basis> basis = rcp( new Basis(polyOrder, Intrepid2::EPointType::POINTTYPE_EQUISPACED) );
-    std::vector<int> subcellCountForDimension;
-    std::vector<std::vector<int>> ordinalsForSubcellDimension;
-    FCO elementToNodeMap;
-    buildSampleElementToNodeMapThreeElements<Basis,ES,FC,FCO>(basis,elementToNodeMap,ordinalsForSubcellDimension,
-                                                              subcellCountForDimension);
-    // check subcell counts first
-    for (int d=0; d<=spaceDim; d++)
-    {
-      if (subcellCountForDimension[d] != int(expectedGIDs[d].size()))
-      {
+      if (listsDiffer) {
         success = false;
-        out << "subcellCountForDimension[" << d << "] != " << expectedGIDs[d].size() << endl;
-      }
-      else
-      {
-        int dofCount = ordinalsForSubcellDimension[d].size();
-        bool listsDiffer = ordinalsForSubcellDimension[d].size() != expectedGIDs[d].size();
-
-        if (!listsDiffer)
-        {
-          for (int dofOrdinal=0; dofOrdinal<dofCount; dofOrdinal++)
-          {
-            if (ordinalsForSubcellDimension[d][dofOrdinal] != expectedGIDs[d][dofOrdinal])
-            {
-              listsDiffer = true;
-              break;
-            }
-          }
+        out << "lists for d=" << d << " differ; expected: {";
+        for (LocalOrdinal LID : expectedGIDs[d]) {
+          out << " " << LID;
         }
-        if (listsDiffer)
-        {
-          success = false;
-          out << "lists for d=" << d << " differ; expected: {";
-          for (LocalOrdinal LID : expectedGIDs[d])
-          {
-            out << " " << LID;
-          }
-          out << " }; actual: {";
-          for (LocalOrdinal LID : ordinalsForSubcellDimension[d])
-          {
-            out << " " << LID;
-          }
-          out << " }\n";
+        out << " }; actual: {";
+        for (LocalOrdinal LID : ordinalsForSubcellDimension[d]) {
+          out << " " << LID;
         }
+        out << " }\n";
       }
     }
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildSampleElementToNodeMapThreeElementQuad, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    ;
-    testBuildSampleElementToNodeMapThreeElementQuad<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,success);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildSampleElementToNodeMapThreeElementQuad, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  ;
+  testBuildSampleElementToNodeMapThreeElementQuad<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success);
+}
 
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
+void testFindSeeds(Xpetra::UnderlyingLib lib, int max_degree, Intrepid2::EPointType ptype, int numRanks, Teuchos::FancyOStream &out, bool &success) {
+  // "numRanks" is for emulated parallel execution: we set up several maps, one corresponding to each emulated MPI rank.
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
-  void testFindSeeds(Xpetra::UnderlyingLib lib,int max_degree, Intrepid2::EPointType ptype, int numRanks, Teuchos::FancyOStream &out, bool &success)
-  {
-    // "numRanks" is for emulated parallel execution: we set up several maps, one corresponding to each emulated MPI rank.
-
-    typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
-    typedef ArrayScalar FC;
-    typedef ArrayOrdinal FCO;
-
-    FCO elementToNodeMap;
-    vector<vector<int>> ordinalsForSubcellDimension;
-    vector<int> subcellCountForDimension;
-
-    using namespace Teuchos;
-    RCP<Comm<int>> serialComm = rcp(new SerialComm<int>());
-    typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node>> serialMapRCP, rowMapRCP, colMapRCP;
-
-    for (int polyOrder=1; polyOrder<max_degree; polyOrder++)
-    {
-      RCP<Basis> basis = rcp( new Basis(polyOrder,ptype) );
-      shards::CellTopology cellTopo = basis->getBaseCellTopology();
-      int spaceDim = cellTopo.getDimension();
-      buildSampleElementToNodeMapThreeElements<Basis,ExecutionSpace,FC,FCO>(basis,elementToNodeMap,ordinalsForSubcellDimension,
-                                                                            subcellCountForDimension);
-      vector<vector<LocalOrdinal>> seeds;
-
-      int numCells = elementToNodeMap.extent(0);
-      int dofsPerCell = elementToNodeMap.extent(1);
-      int maxGID = -1;
-      auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
-      Kokkos::deep_copy(elementToNodeMap_host, elementToNodeMap);
-      for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-      {
-        for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-        {
-          maxGID = max(maxGID,elementToNodeMap_host(cellOrdinal,dofOrdinal));
-        }
+  typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
+  typedef ArrayScalar FC;
+  typedef ArrayOrdinal FCO;
+
+  FCO elementToNodeMap;
+  vector<vector<int>> ordinalsForSubcellDimension;
+  vector<int> subcellCountForDimension;
+
+  using namespace Teuchos;
+  RCP<Comm<int>> serialComm = rcp(new SerialComm<int>());
+  typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> serialMapRCP, rowMapRCP, colMapRCP;
+
+  for (int polyOrder = 1; polyOrder < max_degree; polyOrder++) {
+    RCP<Basis> basis              = rcp(new Basis(polyOrder, ptype));
+    shards::CellTopology cellTopo = basis->getBaseCellTopology();
+    int spaceDim                  = cellTopo.getDimension();
+    buildSampleElementToNodeMapThreeElements<Basis, ExecutionSpace, FC, FCO>(basis, elementToNodeMap, ordinalsForSubcellDimension,
+                                                                             subcellCountForDimension);
+    vector<vector<LocalOrdinal>> seeds;
+
+    int numCells               = elementToNodeMap.extent(0);
+    int dofsPerCell            = elementToNodeMap.extent(1);
+    int maxGID                 = -1;
+    auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
+    Kokkos::deep_copy(elementToNodeMap_host, elementToNodeMap);
+    for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+      for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+        maxGID = max(maxGID, elementToNodeMap_host(cellOrdinal, dofOrdinal));
       }
-      int numElements = maxGID + 1;
-      // construct a serial map (claiming all rows and cells)
-      serialMapRCP = MapFactory::createLocalMapWithNode(lib, numElements, serialComm);
-
-      int startingGID = 0;
-      for (int rank=0; rank<numRanks; rank++)
-      {
-        // simple distribution: take numElements / numRanks on each rank
-        int numRankLocalElements = numElements / numRanks;
-        int remainder = numElements - (numElements / numRanks) * numRanks;
-        if (remainder > rank) numRankLocalElements++;
-        FCO rankLocalElementToNodeMap;
-        resize(rankLocalElementToNodeMap,numCells,basis->getCardinality());
-	auto rankLocalElementToNodeMap_host = Kokkos::create_mirror_view(rankLocalElementToNodeMap);
-        vector<set<LocalOrdinal>> expectedSeedsSets(spaceDim+1);
-
-        auto isRankLocal = [startingGID,numRankLocalElements](GlobalOrdinal GID) -> bool {
-          return (GID >= startingGID) && (GID < startingGID + numRankLocalElements);
-        };
+    }
+    int numElements = maxGID + 1;
+    // construct a serial map (claiming all rows and cells)
+    serialMapRCP = MapFactory::createLocalMapWithNode(lib, numElements, serialComm);
+
+    int startingGID = 0;
+    for (int rank = 0; rank < numRanks; rank++) {
+      // simple distribution: take numElements / numRanks on each rank
+      int numRankLocalElements = numElements / numRanks;
+      int remainder            = numElements - (numElements / numRanks) * numRanks;
+      if (remainder > rank) numRankLocalElements++;
+      FCO rankLocalElementToNodeMap;
+      resize(rankLocalElementToNodeMap, numCells, basis->getCardinality());
+      auto rankLocalElementToNodeMap_host = Kokkos::create_mirror_view(rankLocalElementToNodeMap);
+      vector<set<LocalOrdinal>> expectedSeedsSets(spaceDim + 1);
+
+      auto isRankLocal = [startingGID, numRankLocalElements](GlobalOrdinal GID) -> bool {
+        return (GID >= startingGID) && (GID < startingGID + numRankLocalElements);
+      };
 
-        Teuchos::Array<GlobalOrdinal> myRowGIDs(numRankLocalElements);
-        for (int i=0; i<numRankLocalElements; i++)
-        {
-          myRowGIDs[i] = startingGID + i;
-        }
-        // colMap sees all local rows, plus any non-local IDs they talk to:
-        Teuchos::Array<GlobalOrdinal> myColGIDs = myRowGIDs;
-        set<GlobalOrdinal> offRankGIDs;
-        for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-        {
-          // first pass: does this cell contain any GIDs we own?
-          bool hasOwnedGIDs = false;
-          for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-          {
-            GlobalOrdinal GID = elementToNodeMap_host(cellOrdinal,dofOrdinal);
-            if (isRankLocal(GID))
-            {
-              hasOwnedGIDs = true;
-            }
-          }
-          // second pass: if there are owned GIDs, add any off-rank guys this cell sees
-          if (hasOwnedGIDs)
-          {
-            for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-            {
-              GlobalOrdinal GID = elementToNodeMap_host(cellOrdinal,dofOrdinal);
-              if (! isRankLocal(GID))
-              {
-                if (GID == -1)
-                {
-                  cout << "hmmm....\n";
-                }
-                offRankGIDs.insert(GID);
-              }
-            }
+      Teuchos::Array<GlobalOrdinal> myRowGIDs(numRankLocalElements);
+      for (int i = 0; i < numRankLocalElements; i++) {
+        myRowGIDs[i] = startingGID + i;
+      }
+      // colMap sees all local rows, plus any non-local IDs they talk to:
+      Teuchos::Array<GlobalOrdinal> myColGIDs = myRowGIDs;
+      set<GlobalOrdinal> offRankGIDs;
+      for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+        // first pass: does this cell contain any GIDs we own?
+        bool hasOwnedGIDs = false;
+        for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+          GlobalOrdinal GID = elementToNodeMap_host(cellOrdinal, dofOrdinal);
+          if (isRankLocal(GID)) {
+            hasOwnedGIDs = true;
           }
         }
-        // add off-rank GIDs to the end of the colMapGIDs:
-        myColGIDs.insert(myColGIDs.end(), offRankGIDs.begin(), offRankGIDs.end());
-
-        GlobalOrdinal indexBase = 0;
-        GlobalOrdinal GO_INVALID = Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
-        rowMapRCP = MapFactory::Build(lib,GO_INVALID,myRowGIDs().getConst(), indexBase,serialComm);
-        colMapRCP = MapFactory::Build(lib,GO_INVALID,myColGIDs().getConst(), indexBase,serialComm);
-
-
-        // rewrite elementToNodeMap to contain only LIDs, and determine expected seeds
-        for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-        {
-          vector<vector<LocalOrdinal>> seedForSubcell(spaceDim+1);
-          for (int d=0; d<=spaceDim; d++)
-          {
-            int subcellCount = cellTopo.getSubcellCount(d);
-            seedForSubcell[d] = vector<LocalOrdinal>(subcellCount,LocalOrdinal(-1));
-          }
-          for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-          {
-            GlobalOrdinal GID = elementToNodeMap_host(cellOrdinal,dofOrdinal);
-            LocalOrdinal LID = colMapRCP->getLocalElement(GID);
-            rankLocalElementToNodeMap_host(cellOrdinal,dofOrdinal) = LID; // may be -1
-            if (LID != -1)
-            {
-              int subcdim = basis->getDofTag(dofOrdinal)[0];
-              int subcord = basis->getDofTag(dofOrdinal)[1];
-              LocalOrdinal existingSeed = seedForSubcell[subcdim][subcord];
-              if ((existingSeed == (LocalOrdinal)-1) || (LID < existingSeed))
-              {
-                seedForSubcell[subcdim][subcord] = LID;
-              }
-            }
-          }
-          for (int d=0; d<=spaceDim; d++)
-          {
-            for (LocalOrdinal seed : seedForSubcell[d])
-            {
-              if (seed != -1)
-              {
-                GlobalOrdinal GID = colMapRCP->getGlobalElement(seed);
-                if (rowMapRCP->getLocalElement(GID) == seed)
-                {
-                  expectedSeedsSets[d].insert(seed);
-                }
+        // second pass: if there are owned GIDs, add any off-rank guys this cell sees
+        if (hasOwnedGIDs) {
+          for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+            GlobalOrdinal GID = elementToNodeMap_host(cellOrdinal, dofOrdinal);
+            if (!isRankLocal(GID)) {
+              if (GID == -1) {
+                cout << "hmmm....\n";
               }
+              offRankGIDs.insert(GID);
             }
           }
         }
-
-	Kokkos::deep_copy(rankLocalElementToNodeMap, rankLocalElementToNodeMap_host);
-        MueLu::MueLuIntrepid::FindGeometricSeedOrdinals<Basis,FCO,LocalOrdinal,GlobalOrdinal,Node>(basis, rankLocalElementToNodeMap,
-                                                                                                   seeds, *rowMapRCP, *colMapRCP);
-
-        if (int(seeds.size()) != spaceDim + 1)
-        {
-          success = false;
-          out << "seeds should have extent = spaceDim + 1; ";
-          out << seeds.size() << " != " << spaceDim + 1 << endl;
+      }
+      // add off-rank GIDs to the end of the colMapGIDs:
+      myColGIDs.insert(myColGIDs.end(), offRankGIDs.begin(), offRankGIDs.end());
+
+      GlobalOrdinal indexBase  = 0;
+      GlobalOrdinal GO_INVALID = Teuchos::OrdinalTraits<GlobalOrdinal>::invalid();
+      rowMapRCP                = MapFactory::Build(lib, GO_INVALID, myRowGIDs().getConst(), indexBase, serialComm);
+      colMapRCP                = MapFactory::Build(lib, GO_INVALID, myColGIDs().getConst(), indexBase, serialComm);
+
+      // rewrite elementToNodeMap to contain only LIDs, and determine expected seeds
+      for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+        vector<vector<LocalOrdinal>> seedForSubcell(spaceDim + 1);
+        for (int d = 0; d <= spaceDim; d++) {
+          int subcellCount  = cellTopo.getSubcellCount(d);
+          seedForSubcell[d] = vector<LocalOrdinal>(subcellCount, LocalOrdinal(-1));
         }
-        else
-        {
-          for (int d=0; d<=spaceDim; d++)
-          {
-            // check that we have exactly one entry for each subcell entity
-            int expectedSeedCount = expectedSeedsSets[d].size();
-            int seedCount = seeds[d].size();
-            if (expectedSeedCount != seedCount)
-            {
-              success = false;
-              out << "expected " << expectedSeedCount << " seeds for extent " << d;
-              out << "; had " << seedCount << endl;
+        for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+          GlobalOrdinal GID                                       = elementToNodeMap_host(cellOrdinal, dofOrdinal);
+          LocalOrdinal LID                                        = colMapRCP->getLocalElement(GID);
+          rankLocalElementToNodeMap_host(cellOrdinal, dofOrdinal) = LID;  // may be -1
+          if (LID != -1) {
+            int subcdim               = basis->getDofTag(dofOrdinal)[0];
+            int subcord               = basis->getDofTag(dofOrdinal)[1];
+            LocalOrdinal existingSeed = seedForSubcell[subcdim][subcord];
+            if ((existingSeed == (LocalOrdinal)-1) || (LID < existingSeed)) {
+              seedForSubcell[subcdim][subcord] = LID;
             }
-            // check that each entry belongs to an entity of the correct type
-            for (LocalOrdinal localDofOrdinal : seeds[d])
-            {
-              if (expectedSeedsSets[d].find(localDofOrdinal) == expectedSeedsSets[d].end())
-              {
-                success = false;
-                out << "Found local dof ordinal " << localDofOrdinal << " as a seed for extent ";
-                out << d << ", but did not find it in expectedSeedsSets[" << d << "]\n";
+          }
+        }
+        for (int d = 0; d <= spaceDim; d++) {
+          for (LocalOrdinal seed : seedForSubcell[d]) {
+            if (seed != -1) {
+              GlobalOrdinal GID = colMapRCP->getGlobalElement(seed);
+              if (rowMapRCP->getLocalElement(GID) == seed) {
+                expectedSeedsSets[d].insert(seed);
               }
             }
           }
         }
-        startingGID += numRankLocalElements;
       }
-    }
-  }
 
+      Kokkos::deep_copy(rankLocalElementToNodeMap, rankLocalElementToNodeMap_host);
+      MueLu::MueLuIntrepid::FindGeometricSeedOrdinals<Basis, FCO, LocalOrdinal, GlobalOrdinal, Node>(basis, rankLocalElementToNodeMap,
+                                                                                                     seeds, *rowMapRCP, *colMapRCP);
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
-  void testFindSeedsSerial(Xpetra::UnderlyingLib lib, int max_degree, Intrepid2::EPointType ptype, Teuchos::FancyOStream &out, bool &success)
-  {
-    typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
-    typedef ArrayScalar FC;
-    typedef ArrayOrdinal FCO;
-
-    FCO elementToNodeMap;
-    vector<vector<int>> ordinalsForSubcellDimension;
-    vector<int> subcellCountForDimension;
-
-    using namespace Teuchos;
-    RCP<Comm<int>> serialComm = rcp(new SerialComm<int>());
-    typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
-    RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > serialMapRCP;
-
-    for (int polyOrder=1; polyOrder<max_degree; polyOrder++)
-    {
-      RCP<Basis> basis = rcp( new Basis(polyOrder,ptype) );
-      buildSampleElementToNodeMapThreeElements<Basis,ExecutionSpace,FC,FCO>(basis,elementToNodeMap,ordinalsForSubcellDimension,
-                                                                            subcellCountForDimension);
-      vector<vector<LocalOrdinal>> seeds;
-
-      // construct a serial map (claiming all rows and cells)
-      int numCells = elementToNodeMap.extent(0);
-      int dofsPerCell = elementToNodeMap.extent(1);
-      int maxLID = -1;
-      auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
-      Kokkos::deep_copy(elementToNodeMap_host, elementToNodeMap);
-      for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-      {
-        for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-        {
-          maxLID = max(maxLID,elementToNodeMap_host(cellOrdinal,dofOrdinal));
-        }
-      }
-      int numElements = maxLID + 1;
-      serialMapRCP = MapFactory::createLocalMapWithNode(lib, numElements, serialComm);
-
-      MueLu::MueLuIntrepid::FindGeometricSeedOrdinals<Basis,FCO,LocalOrdinal,GlobalOrdinal,Node>(basis, elementToNodeMap,
-                                                                                                 seeds, *serialMapRCP, *serialMapRCP);
-
-      int spaceDim = basis->getBaseCellTopology().getDimension();
-      if (int(seeds.size()) != spaceDim + 1)
-      {
+      if (int(seeds.size()) != spaceDim + 1) {
         success = false;
         out << "seeds should have extent = spaceDim + 1; ";
         out << seeds.size() << " != " << spaceDim + 1 << endl;
-      }
-      else
-      {
-        for (int d=0; d<=spaceDim; d++)
-        {
+      } else {
+        for (int d = 0; d <= spaceDim; d++) {
           // check that we have exactly one entry for each subcell entity
-          int expectedSeedCount = subcellCountForDimension[d];
-          if (basis->getDofCount(d,0) == 0)
-          {
-            // no dofs for first subcell of extent d; we assume that all other subcells of extent d also
-            // don't have dofs assigned
-            expectedSeedCount = 0;
-          }
-          int seedCount = seeds[d].size();
-          if (expectedSeedCount != seedCount)
-          {
+          int expectedSeedCount = expectedSeedsSets[d].size();
+          int seedCount         = seeds[d].size();
+          if (expectedSeedCount != seedCount) {
             success = false;
             out << "expected " << expectedSeedCount << " seeds for extent " << d;
             out << "; had " << seedCount << endl;
           }
           // check that each entry belongs to an entity of the correct type
-          for (LocalOrdinal localDofOrdinal : seeds[d])
-          {
-            if (std::find(ordinalsForSubcellDimension[d].begin(),ordinalsForSubcellDimension[d].end(),localDofOrdinal)
-                == ordinalsForSubcellDimension[d].end())
-            {
+          for (LocalOrdinal localDofOrdinal : seeds[d]) {
+            if (expectedSeedsSets[d].find(localDofOrdinal) == expectedSeedsSets[d].end()) {
               success = false;
               out << "Found local dof ordinal " << localDofOrdinal << " as a seed for extent ";
-              out << d << ", but did not find it in ordinalsForSubcellDimension[" << d << "]\n";
+              out << d << ", but did not find it in expectedSeedsSets[" << d << "]\n";
             }
           }
         }
       }
+      startingGID += numRankLocalElements;
     }
   }
+}
 
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
+void testFindSeedsSerial(Xpetra::UnderlyingLib lib, int max_degree, Intrepid2::EPointType ptype, Teuchos::FancyOStream &out, bool &success) {
+  typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
+  typedef ArrayScalar FC;
+  typedef ArrayOrdinal FCO;
 
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
-  void testFindSeedsParallel(Xpetra::UnderlyingLib lib, int max_degree, Intrepid2::EPointType ptype, Teuchos::FancyOStream &out, bool &success)
-  {
-    typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
-    typedef ArrayScalar FC;
-    typedef ArrayOrdinal FCO;
-
-    for (int rankCount=1; rankCount<=MAX_RANK_COUNT; rankCount++)
-    {
-      out << "running testFindSeedsParallel on " << rankCount << " emulated MPI ranks\n";
-      testFindSeeds<LocalOrdinal, GlobalOrdinal, Node, Basis, ExecutionSpace, ArrayScalar, ArrayOrdinal>(lib,max_degree, ptype, rankCount, out, success);
+  FCO elementToNodeMap;
+  vector<vector<int>> ordinalsForSubcellDimension;
+  vector<int> subcellCountForDimension;
+
+  using namespace Teuchos;
+  RCP<Comm<int>> serialComm = rcp(new SerialComm<int>());
+  typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
+  RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>> serialMapRCP;
+
+  for (int polyOrder = 1; polyOrder < max_degree; polyOrder++) {
+    RCP<Basis> basis = rcp(new Basis(polyOrder, ptype));
+    buildSampleElementToNodeMapThreeElements<Basis, ExecutionSpace, FC, FCO>(basis, elementToNodeMap, ordinalsForSubcellDimension,
+                                                                             subcellCountForDimension);
+    vector<vector<LocalOrdinal>> seeds;
+
+    // construct a serial map (claiming all rows and cells)
+    int numCells               = elementToNodeMap.extent(0);
+    int dofsPerCell            = elementToNodeMap.extent(1);
+    int maxLID                 = -1;
+    auto elementToNodeMap_host = Kokkos::create_mirror_view(elementToNodeMap);
+    Kokkos::deep_copy(elementToNodeMap_host, elementToNodeMap);
+    for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+      for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+        maxLID = max(maxLID, elementToNodeMap_host(cellOrdinal, dofOrdinal));
+      }
+    }
+    int numElements = maxLID + 1;
+    serialMapRCP    = MapFactory::createLocalMapWithNode(lib, numElements, serialComm);
+
+    MueLu::MueLuIntrepid::FindGeometricSeedOrdinals<Basis, FCO, LocalOrdinal, GlobalOrdinal, Node>(basis, elementToNodeMap,
+                                                                                                   seeds, *serialMapRCP, *serialMapRCP);
+
+    int spaceDim = basis->getBaseCellTopology().getDimension();
+    if (int(seeds.size()) != spaceDim + 1) {
+      success = false;
+      out << "seeds should have extent = spaceDim + 1; ";
+      out << seeds.size() << " != " << spaceDim + 1 << endl;
+    } else {
+      for (int d = 0; d <= spaceDim; d++) {
+        // check that we have exactly one entry for each subcell entity
+        int expectedSeedCount = subcellCountForDimension[d];
+        if (basis->getDofCount(d, 0) == 0) {
+          // no dofs for first subcell of extent d; we assume that all other subcells of extent d also
+          // don't have dofs assigned
+          expectedSeedCount = 0;
+        }
+        int seedCount = seeds[d].size();
+        if (expectedSeedCount != seedCount) {
+          success = false;
+          out << "expected " << expectedSeedCount << " seeds for extent " << d;
+          out << "; had " << seedCount << endl;
+        }
+        // check that each entry belongs to an entity of the correct type
+        for (LocalOrdinal localDofOrdinal : seeds[d]) {
+          if (std::find(ordinalsForSubcellDimension[d].begin(), ordinalsForSubcellDimension[d].end(), localDofOrdinal) == ordinalsForSubcellDimension[d].end()) {
+            success = false;
+            out << "Found local dof ordinal " << localDofOrdinal << " as a seed for extent ";
+            out << d << ", but did not find it in ordinalsForSubcellDimension[" << d << "]\n";
+          }
+        }
+      }
     }
   }
+}
 
-  /******* End helper methods and classes by Nate ********/
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class Basis, class ExecutionSpace, class ArrayScalar, class ArrayOrdinal>
+void testFindSeedsParallel(Xpetra::UnderlyingLib lib, int max_degree, Intrepid2::EPointType ptype, Teuchos::FancyOStream &out, bool &success) {
+  typedef Intrepid2::CellTools<ExecutionSpace> CellTools;
+  typedef ArrayScalar FC;
+  typedef ArrayOrdinal FCO;
 
+  for (int rankCount = 1; rankCount <= MAX_RANK_COUNT; rankCount++) {
+    out << "running testFindSeedsParallel on " << rankCount << " emulated MPI ranks\n";
+    testFindSeeds<LocalOrdinal, GlobalOrdinal, Node, Basis, ExecutionSpace, ArrayScalar, ArrayOrdinal>(lib, max_degree, ptype, rankCount, out, success);
+  }
+}
+
+/******* End helper methods and classes by Nate ********/
+
+/******* Begin typedefs for FindSeeds tests by Nate ********/
+#define FIND_SEEDS_MACROS                                                    \
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;          \
+  typedef typename Node::device_type::execution_space ES;                    \
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;            \
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCO; \
+  typedef Intrepid2::Basis_HGRAD_LINE_Cn_FEM<ES, MT, MT> LineBasis;          \
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> QuadBasis;          \
+  typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES, MT, MT> HexBasis;
+
+/******* End typedefs for FindSeeds tests by Nate ********/
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Line, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, LineBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_LINE_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, LineBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_LINE_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Line, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);  // Not sure why I have to do this...
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, LineBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_LINE_DEGREE, POINTTYPE_SPECTRAL, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, LineBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_LINE_DEGREE, POINTTYPE_SPECTRAL, out, success);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Quad, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, QuadBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, QuadBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Quad, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);  // Not sure why I have to do this...
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, QuadBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, POINTTYPE_SPECTRAL, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, QuadBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, POINTTYPE_SPECTRAL, out, success);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Hex, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, HexBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, HexBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_QUAD_DEGREE, Intrepid2::POINTTYPE_EQUISPACED, out, success);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Hex, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  FIND_SEEDS_MACROS;
+  const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);  // Not sure why I have to do this...
+  testFindSeedsParallel<LocalOrdinal, GlobalOrdinal, Node, HexBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_HEX_DEGREE, POINTTYPE_SPECTRAL, out, success);
+  testFindSeedsSerial<LocalOrdinal, GlobalOrdinal, Node, HexBasis, ES, FC, FCO>(MueLuTests::TestHelpers::Parameters::getLib(), MAX_HEX_DEGREE, POINTTYPE_SPECTRAL, out, success);
+}
 
-  /******* Begin typedefs for FindSeeds tests by Nate ********/
-#define FIND_SEEDS_MACROS \
-  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;         \
-  typedef typename Node::device_type::execution_space ES;                   \
-  typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;            \
-  typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCO; \
-  typedef Intrepid2::Basis_HGRAD_LINE_Cn_FEM<ES,MT,MT> LineBasis;           \
-  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> QuadBasis;           \
-  typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES,MT,MT> HexBasis;
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GetP1NodeInHi, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Intrepid2::Basis<ES, MT, MT> Basis;
 
-  /******* End typedefs for FindSeeds tests by Nate ********/
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Line, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,LineBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_LINE_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,LineBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_LINE_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
-  }
+  int max_degree = (Intrepid2::Parameters::MaxOrder < 5) ? Intrepid2::Parameters::MaxOrder : 5;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Line, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);// Not sure why I have to do this...
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,LineBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_LINE_DEGREE,POINTTYPE_SPECTRAL,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,LineBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_LINE_DEGREE,POINTTYPE_SPECTRAL,out,success);
-  }
+    // QUAD
+    RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT>());
 
+    for (int i = 0; i < max_degree; i++) {
+      RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>(i, Intrepid2::POINTTYPE_EQUISPACED));
+      std::vector<size_t> lo_node_in_hi;
+      FC hi_dofCoords;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Quad, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,QuadBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,QuadBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
+      MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT, typename Node::device_type>(hi, lo_node_in_hi, hi_dofCoords);
+      TEST_EQUALITY((size_t)hi_dofCoords.extent(0), (size_t)hi->getCardinality());
+      TEST_EQUALITY((size_t)hi_dofCoords.extent(1), (size_t)hi->getBaseCellTopology().getDimension());
+      TEST_EQUALITY(lo_node_in_hi.size(), (size_t)lo->getCardinality());
+    }
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Quad, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BasisFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // QUAD
+  int degree;
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);// Not sure why I have to do this...
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,QuadBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,POINTTYPE_SPECTRAL,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,QuadBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,POINTTYPE_SPECTRAL,out,success);
+    bool test = rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT>>(MueLu::MueLuIntrepid::BasisFactory<MT, ES>("hgrad_quad_c1", degree)) != Teuchos::null;
+    TEST_EQUALITY(test, true);
   }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Hex, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,HexBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,HexBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_QUAD_DEGREE,Intrepid2::POINTTYPE_EQUISPACED,out,success);
+    bool test = rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>>(MueLu::MueLuIntrepid::BasisFactory<MT, ES>("hgrad_quad_c2", degree)) != Teuchos::null;
+    TEST_EQUALITY(test, true);
   }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, FindSeeds_Spectral_Hex, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    FIND_SEEDS_MACROS;
-    const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);// Not sure why I have to do this...
-    testFindSeedsParallel<LocalOrdinal,GlobalOrdinal,Node,HexBasis,ES,FC,FCO>(MueLuTests::TestHelpers::Parameters::getLib(),MAX_HEX_DEGREE,POINTTYPE_SPECTRAL,out,success);
-    testFindSeedsSerial<LocalOrdinal,GlobalOrdinal,Node,HexBasis,ES,FC,FCO>  (MueLuTests::TestHelpers::Parameters::getLib(),MAX_HEX_DEGREE,POINTTYPE_SPECTRAL,out,success);
+    bool test = rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>>(MueLu::MueLuIntrepid::BasisFactory<MT, ES>("hgrad_quad_c3", degree)) != Teuchos::null;
+    TEST_EQUALITY(test, true);
   }
+}
 
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GetP1NodeInHi, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Intrepid2::Basis<ES,MT,MT> Basis;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    int max_degree = (Intrepid2::Parameters::MaxOrder < 5 ) ? Intrepid2::Parameters::MaxOrder : 5;
-
-    {
-      // QUAD
-      RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT>());
-
-
-      for(int i=0;i<max_degree; i++) {
-        RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT>(i,Intrepid2::POINTTYPE_EQUISPACED));
-        std::vector<size_t> lo_node_in_hi;
-        FC hi_dofCoords;
-
-        MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT,typename Node::device_type>(hi,lo_node_in_hi,hi_dofCoords);
-        TEST_EQUALITY((size_t)hi_dofCoords.extent(0),(size_t)hi->getCardinality());
-        TEST_EQUALITY((size_t)hi_dofCoords.extent(1),(size_t)hi->getBaseCellTopology().getDimension());
-        TEST_EQUALITY(lo_node_in_hi.size(),(size_t)lo->getCardinality());
-      }
-    }
-  }
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildLoElemToNode, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Intrepid2::Basis<ES, MT, MT> Basis;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  int max_degree = (Intrepid2::Parameters::MaxOrder < 5) ? Intrepid2::Parameters::MaxOrder : 5;
 
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BasisFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
     // QUAD
-    int degree;
-    {bool test= rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT> >(MueLu::MueLuIntrepid::BasisFactory<MT,ES>("hgrad_quad_c1",degree)) !=Teuchos::null;TEST_EQUALITY(test,true);}
-    {bool test= rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> >(MueLu::MueLuIntrepid::BasisFactory<MT,ES>("hgrad_quad_c2",degree)) !=Teuchos::null;TEST_EQUALITY(test,true);}
-    {bool test= rcp_dynamic_cast<Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> >(MueLu::MueLuIntrepid::BasisFactory<MT,ES>("hgrad_quad_c3",degree)) !=Teuchos::null;TEST_EQUALITY(test,true);}
+    //  A one element test with Kirby-numbered nodes where the top edge is not owned
+    RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT>());
 
-  }
+    for (int degree = 2; degree < max_degree; degree++) {
+      int Nn        = (degree + 1) * (degree + 1);
+      RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>(degree, Intrepid2::POINTTYPE_EQUISPACED));
+      FCi hi_e2n("hi_e2n", 1, Nn), lo_e2n;
 
+      std::vector<bool> hi_owned(Nn, false), lo_owned;
+      std::vector<size_t> lo_node_in_hi;
+      std::vector<LO> hi_to_lo_map;
+      int lo_numOwnedNodes = 0;
+      FC hi_dofCoords;
+      MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT, typename Node::device_type>(hi, lo_node_in_hi, hi_dofCoords);
 
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildLoElemToNode, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Intrepid2::Basis<ES,MT,MT> Basis;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    int max_degree = (Intrepid2::Parameters::MaxOrder < 5 ) ? Intrepid2::Parameters::MaxOrder : 5;
-
-
-    {
-      //QUAD
-      // A one element test with Kirby-numbered nodes where the top edge is not owned
-      RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT>());
-
-      for(int degree=2; degree < max_degree; degree++) {
-        int Nn = (degree+1)*(degree+1);
-        RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT>(degree,Intrepid2::POINTTYPE_EQUISPACED));
-        FCi hi_e2n("hi_e2n",1,Nn), lo_e2n;
-
-        std::vector<bool> hi_owned(Nn,false),lo_owned;
-        std::vector<size_t> lo_node_in_hi;
-        std::vector<LO> hi_to_lo_map;
-        int lo_numOwnedNodes=0;
-        FC hi_dofCoords;
-        MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT,typename Node::device_type>(hi,lo_node_in_hi,hi_dofCoords);
-
-        Kokkos::parallel_for("IntrepidPCoarsenFactory,BuildLoElemToNode", Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA (int i) {
-           hi_e2n(0,i)=i;
-        });
-	Kokkos::fence();
-        for(int i=0; i<Nn; i++) {
-          if(i < Nn-(degree+1)) hi_owned[i]=true;
-        }
+      Kokkos::parallel_for(
+          "IntrepidPCoarsenFactory,BuildLoElemToNode", Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA(int i) {
+            hi_e2n(0, i) = i;
+          });
+      Kokkos::fence();
+      for (int i = 0; i < Nn; i++) {
+        if (i < Nn - (degree + 1)) hi_owned[i] = true;
+      }
 
-        Teuchos::ArrayRCP<const int> is_dirichlet(Nn,false);
+      Teuchos::ArrayRCP<const int> is_dirichlet(Nn, false);
 
-        MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n,hi_owned,lo_node_in_hi,is_dirichlet,lo_e2n,lo_owned,hi_to_lo_map,lo_numOwnedNodes);
+      MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n, hi_owned, lo_node_in_hi, is_dirichlet, lo_e2n, lo_owned, hi_to_lo_map, lo_numOwnedNodes);
 
-        // Checks
-        TEST_EQUALITY(lo_numOwnedNodes,2);
+      // Checks
+      TEST_EQUALITY(lo_numOwnedNodes, 2);
 
-        size_t num_lo_nodes_located=0;
-        for(size_t i=0;i<hi_to_lo_map.size(); i++) {
-          if(hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
-            num_lo_nodes_located++;
-        }
-        TEST_EQUALITY(lo_owned.size(),num_lo_nodes_located);
+      size_t num_lo_nodes_located = 0;
+      for (size_t i = 0; i < hi_to_lo_map.size(); i++) {
+        if (hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
+          num_lo_nodes_located++;
       }
-    }//end QUAD
+      TEST_EQUALITY(lo_owned.size(), num_lo_nodes_located);
+    }
+  }  // end QUAD
+}
 
-  }
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildLoElemToNodeWithDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Intrepid2::Basis<ES, MT, MT> Basis;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  int max_degree = (Intrepid2::Parameters::MaxOrder < 5) ? Intrepid2::Parameters::MaxOrder : 5;
 
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildLoElemToNodeWithDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-  #   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Intrepid2::Basis<ES,MT,MT> Basis;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    int max_degree = (Intrepid2::Parameters::MaxOrder < 5 ) ? Intrepid2::Parameters::MaxOrder : 5;
-
-
-    {
-      //QUAD
-      // A one element test with Kirby-numbered nodes where the top edge is not owned
-      RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT>());
-
-      for(int degree=2; degree < max_degree; degree++) {
-        int Nn = (degree+1)*(degree+1);
-        RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT>(degree,Intrepid2::POINTTYPE_EQUISPACED));
-        FCi hi_e2n("hi_e2n",1,Nn), lo_e2n;
-
-        std::vector<bool> hi_owned(Nn,false),lo_owned;
-        std::vector<size_t> lo_node_in_hi;
-        std::vector<LO> hi_to_lo_map;
-        int lo_numOwnedNodes=0;
-        FC hi_dofCoords;
-        MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT,typename Node::device_type>(hi,lo_node_in_hi,hi_dofCoords);
-
-        Kokkos::parallel_for("IntrepidPCoarsenFactory,BuildLoElemToNodeWithDirichlet",
-			     Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA (int i) {
-           hi_e2n(0,i)=i;
-        });
-	Kokkos::fence();
-        for(int i=0; i<Nn; i++) {
-          if(i < Nn-(degree+1)) hi_owned[i]=true;
-        }
-
-
-        // Mark of the very first node as Dirichlet
-        Teuchos::ArrayRCP<int> is_dirichlet(Nn,false);
-        is_dirichlet[0] = 1;
+    // QUAD
+    //  A one element test with Kirby-numbered nodes where the top edge is not owned
+    RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT>());
 
-        MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n,hi_owned,lo_node_in_hi,is_dirichlet,lo_e2n,lo_owned,hi_to_lo_map,lo_numOwnedNodes);
+    for (int degree = 2; degree < max_degree; degree++) {
+      int Nn        = (degree + 1) * (degree + 1);
+      RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>(degree, Intrepid2::POINTTYPE_EQUISPACED));
+      FCi hi_e2n("hi_e2n", 1, Nn), lo_e2n;
 
-        // Checks
-        TEST_EQUALITY(lo_numOwnedNodes,1);
+      std::vector<bool> hi_owned(Nn, false), lo_owned;
+      std::vector<size_t> lo_node_in_hi;
+      std::vector<LO> hi_to_lo_map;
+      int lo_numOwnedNodes = 0;
+      FC hi_dofCoords;
+      MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT, typename Node::device_type>(hi, lo_node_in_hi, hi_dofCoords);
 
-        size_t num_lo_nodes_located=0;
-        for(size_t i=0;i<hi_to_lo_map.size(); i++) {
-          if(hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
-            num_lo_nodes_located++;
-        }
-        TEST_EQUALITY(lo_owned.size(),num_lo_nodes_located);
+      Kokkos::parallel_for(
+          "IntrepidPCoarsenFactory,BuildLoElemToNodeWithDirichlet",
+          Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA(int i) {
+            hi_e2n(0, i) = i;
+          });
+      Kokkos::fence();
+      for (int i = 0; i < Nn; i++) {
+        if (i < Nn - (degree + 1)) hi_owned[i] = true;
       }
-    }//end QUAD
 
-  }
+      // Mark of the very first node as Dirichlet
+      Teuchos::ArrayRCP<int> is_dirichlet(Nn, false);
+      is_dirichlet[0] = 1;
 
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,GenerateColMapFromImport, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-    GO lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
-    int NumProcs = comm->getSize();
-    // This function basically takes an existing domain->column importer (the "hi order" guy) and using the hi_to_lo_map,
-    // generates "lo order" version.  The domain map is already given to us here, so we just need to make tha column one.
-
-    // We'll test this by starting with some linearly distributed map for the "hi" domain map and a duplicated map as the "hi" column map.
-    // We then do a "lo" domain map, which just grabs the first GID on each proc.
-    GO numGlobalElements =100;
-    Teuchos::Array<GO> hi_Cols(numGlobalElements);
-    for(size_t i=0; i<(size_t)numGlobalElements; i++)
-      hi_Cols[i] = i;
-
-    RCP<Map> hi_domainMap   = MapFactory::Build(lib,numGlobalElements,0,comm);
-    RCP<Map> hi_colMap      = MapFactory::Build(lib,gst_invalid,hi_Cols(),0,comm);
-    RCP<Import> hi_importer = ImportFactory::Build(hi_domainMap,hi_colMap);
-
-    Teuchos::Array<GO> lo_Doms(1); lo_Doms[0]=hi_domainMap->getGlobalElement(0);
-    RCP<Map> lo_domainMap= MapFactory::Build(lib,gst_invalid,lo_Doms(),0,comm);
-
-    // Get the list of the first GID from each proc (aka the lo_colMap)
-    std::vector<GO> send_colgids(1);
-    std::vector<GO> recv_colgids(NumProcs); send_colgids[0] = hi_domainMap->getGlobalElement(0);
-    comm->gatherAll(1*sizeof(GO),(char*)send_colgids.data(),NumProcs*sizeof(GO),(char*)recv_colgids.data());
-
-    // Build the h2l map
-    std::vector<LO> hi_to_lo_map(hi_colMap->getLocalNumElements(),lo_invalid);
-    for(size_t i=0; i<(size_t)NumProcs; i++)
-      hi_to_lo_map[recv_colgids[i]]=i;
-
-    // Import
-    size_t lo_columnMapLength = NumProcs; // One col per proc
-    RCP<const Map> lo_colMap;
-    MueLu::MueLuIntrepid::GenerateColMapFromImport(*hi_importer,hi_to_lo_map,*lo_domainMap,lo_columnMapLength,lo_colMap);
-
-    // Final test
-    for(size_t i=0; i<lo_colMap->getLocalNumElements(); i++)
-      TEST_EQUALITY(recv_colgids[i],lo_colMap->getGlobalElement(i));
+      MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n, hi_owned, lo_node_in_hi, is_dirichlet, lo_e2n, lo_owned, hi_to_lo_map, lo_numOwnedNodes);
 
-  }
+      // Checks
+      TEST_EQUALITY(lo_numOwnedNodes, 1);
 
-  /*********************************************************************************************************************/
-  /* How this guy works:
-     num_p1_nodes - number of nodes in the p=1 mesh
-     p1_gold_in   - input vector for the lo_basis
-     p2_gold_in   - output vector of linear interpolation from lo_basis to hi_basis
-
-  */
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-void TestPseudoPoisson(Teuchos::FancyOStream &out, int num_p1_nodes, int degree, std::vector<Scalar> &lo_gold_in, std::vector<Scalar> &hi_gold_out,const std::string & hi_basis, const std::string lo_basis = "hgrad_line_c1")
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-//    GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-//    GO lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
-    int MyPID = comm->getRank();
-
-    // Setup Levels
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_p1_nodes,degree,elem_to_node,lib);
-    fineLevel.Set("A",A);
-    fineLevel.Set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-
-    // only one NS vector
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->setSeed(846930886);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace",nullSpace);
-
-    // ParameterList
-    ParameterList Params;
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis",lo_basis);
-
-    // Build P
-    RCP<MueLu::IntrepidPCoarsenFactory<SC,LO,GO,NO> > IPCFact = rcp(new MueLu::IntrepidPCoarsenFactory<SC,LO,GO,NO>());
-    IPCFact->SetParameterList(Params);
-    coarseLevel.Request("P",IPCFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",IPCFact.get());
-    coarseLevel.Request("CoarseMap",IPCFact.get());
-    coarseLevel.Request(*IPCFact);
-    IPCFact->Build(fineLevel,coarseLevel);
-
-    // Get P
-    RCP<Matrix> P;
-    coarseLevel.Get("P",P,IPCFact.get());
-    RCP<CrsMatrix> Pcrs   = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-    //if(!MyPID) printf("P size = %d x %d\n",(int)P->getRangeMap()->getGlobalNumElements(),(int)P->getDomainMap()->getGlobalNumElements());
-    out << "P size = "
-        << (int)P->getRangeMap()->getGlobalNumElements()
-        << " x "
-        << (int)P->getDomainMap()->getGlobalNumElements()
-        << std::endl;
-
-    // Sanity
-    if((int)P->getRangeMap()->getGlobalNumElements()!=(int)hi_gold_out.size())
-      throw std::runtime_error("P range size does not match hi_gold_out");
-    if((int)P->getDomainMap()->getGlobalNumElements()!=(int)lo_gold_in.size())
-      throw std::runtime_error("P domain size does not match lo_gold_in");
-
-    // Build serial comparison maps
-    GO hi_num_global_dofs = A->getRowMap()->getGlobalNumElements();
-    GO hi_num_serial_elements = !MyPID ? hi_num_global_dofs : 0;
-    RCP<Map> hi_SerialMap = MapFactory::Build(lib,hi_num_global_dofs,hi_num_serial_elements,0,comm);
-
-    GO lo_num_global_dofs = P->getDomainMap()->getGlobalNumElements();
-    GO lo_num_serial_elements = !MyPID ? lo_num_global_dofs : 0;
-    RCP<Map> lo_SerialMap = MapFactory::Build(lib, lo_num_global_dofs,lo_num_serial_elements,0,comm);
-
-    RCP<Export> lo_importer = ExportFactory::Build(lo_SerialMap,P->getDomainMap());
-    RCP<Export> hi_importer = ExportFactory::Build(A->getRowMap(),hi_SerialMap);
-
-    // Allocate some vectors
-    RCP<Vector> s_InVec = VectorFactory::Build(lo_SerialMap);
-    RCP<Vector> p_InVec = VectorFactory::Build(P->getDomainMap());
-    RCP<Vector> s_OutVec = VectorFactory::Build(hi_SerialMap);
-    RCP<Vector> s_codeOutput = VectorFactory::Build(hi_SerialMap);
-    RCP<Vector> p_codeOutput = VectorFactory::Build(A->getRowMap());
-
-
-    // Fill serial GOLD vecs on Proc 0
-    if(!MyPID) {
-      for(size_t i=0; i<(size_t)lo_gold_in.size(); i++)
-        s_InVec->replaceLocalValue(i,lo_gold_in[i]);
-
-      for(size_t i=0; i<(size_t)hi_gold_out.size(); i++)
-        s_OutVec->replaceLocalValue(i,hi_gold_out[i]);
+      size_t num_lo_nodes_located = 0;
+      for (size_t i = 0; i < hi_to_lo_map.size(); i++) {
+        if (hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
+          num_lo_nodes_located++;
+      }
+      TEST_EQUALITY(lo_owned.size(), num_lo_nodes_located);
     }
+  }  // end QUAD
+}
 
-    // Migrate input data
-    p_InVec->doExport(*s_InVec,*lo_importer,Xpetra::ADD);
-
-    // Apply P
-    P->apply(*p_InVec,*p_codeOutput);
-
-    // Migrate Output data
-    s_codeOutput->doExport(*p_codeOutput,*hi_importer,Xpetra::ADD);
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateColMapFromImport, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  GO gst_invalid                     = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+  GO lo_invalid                      = Teuchos::OrdinalTraits<LO>::invalid();
+  int NumProcs                       = comm->getSize();
+  // This function basically takes an existing domain->column importer (the "hi order" guy) and using the hi_to_lo_map,
+  // generates "lo order" version.  The domain map is already given to us here, so we just need to make tha column one.
+
+  // We'll test this by starting with some linearly distributed map for the "hi" domain map and a duplicated map as the "hi" column map.
+  // We then do a "lo" domain map, which just grabs the first GID on each proc.
+  GO numGlobalElements = 100;
+  Teuchos::Array<GO> hi_Cols(numGlobalElements);
+  for (size_t i = 0; i < (size_t)numGlobalElements; i++)
+    hi_Cols[i] = i;
+
+  RCP<Map> hi_domainMap   = MapFactory::Build(lib, numGlobalElements, 0, comm);
+  RCP<Map> hi_colMap      = MapFactory::Build(lib, gst_invalid, hi_Cols(), 0, comm);
+  RCP<Import> hi_importer = ImportFactory::Build(hi_domainMap, hi_colMap);
+
+  Teuchos::Array<GO> lo_Doms(1);
+  lo_Doms[0]            = hi_domainMap->getGlobalElement(0);
+  RCP<Map> lo_domainMap = MapFactory::Build(lib, gst_invalid, lo_Doms(), 0, comm);
+
+  // Get the list of the first GID from each proc (aka the lo_colMap)
+  std::vector<GO> send_colgids(1);
+  std::vector<GO> recv_colgids(NumProcs);
+  send_colgids[0] = hi_domainMap->getGlobalElement(0);
+  comm->gatherAll(1 * sizeof(GO), (char *)send_colgids.data(), NumProcs * sizeof(GO), (char *)recv_colgids.data());
+
+  // Build the h2l map
+  std::vector<LO> hi_to_lo_map(hi_colMap->getLocalNumElements(), lo_invalid);
+  for (size_t i = 0; i < (size_t)NumProcs; i++)
+    hi_to_lo_map[recv_colgids[i]] = i;
+
+  // Import
+  size_t lo_columnMapLength = NumProcs;  // One col per proc
+  RCP<const Map> lo_colMap;
+  MueLu::MueLuIntrepid::GenerateColMapFromImport(*hi_importer, hi_to_lo_map, *lo_domainMap, lo_columnMapLength, lo_colMap);
+
+  // Final test
+  for (size_t i = 0; i < lo_colMap->getLocalNumElements(); i++)
+    TEST_EQUALITY(recv_colgids[i], lo_colMap->getGlobalElement(i));
+}
 
-    // Compare vs. GOLD
-    s_codeOutput->update(-1.0,*s_OutVec,1.0);
-    Teuchos::Array<MT> norm2(1);
-    s_codeOutput->norm2(norm2());
+/*********************************************************************************************************************/
+/* How this guy works:
+   num_p1_nodes - number of nodes in the p=1 mesh
+   p1_gold_in   - input vector for the lo_basis
+   p2_gold_in   - output vector of linear interpolation from lo_basis to hi_basis
+
+*/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void TestPseudoPoisson(Teuchos::FancyOStream &out, int num_p1_nodes, int degree, std::vector<Scalar> &lo_gold_in, std::vector<Scalar> &hi_gold_out, const std::string &hi_basis, const std::string lo_basis = "hgrad_line_c1") {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  //    GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+  //    GO lo_invalid = Teuchos::OrdinalTraits<LO>::invalid();
+  int MyPID = comm->getRank();
+
+  // Setup Levels
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_p1_nodes, degree, elem_to_node, lib);
+  fineLevel.Set("A", A);
+  fineLevel.Set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+
+  // only one NS vector
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->setSeed(846930886);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+
+  // ParameterList
+  ParameterList Params;
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", lo_basis);
+
+  // Build P
+  RCP<MueLu::IntrepidPCoarsenFactory<SC, LO, GO, NO>> IPCFact = rcp(new MueLu::IntrepidPCoarsenFactory<SC, LO, GO, NO>());
+  IPCFact->SetParameterList(Params);
+  coarseLevel.Request("P", IPCFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", IPCFact.get());
+  coarseLevel.Request("CoarseMap", IPCFact.get());
+  coarseLevel.Request(*IPCFact);
+  IPCFact->Build(fineLevel, coarseLevel);
+
+  // Get P
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, IPCFact.get());
+  RCP<CrsMatrix> Pcrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+  // if(!MyPID) printf("P size = %d x %d\n",(int)P->getRangeMap()->getGlobalNumElements(),(int)P->getDomainMap()->getGlobalNumElements());
+  out << "P size = "
+      << (int)P->getRangeMap()->getGlobalNumElements()
+      << " x "
+      << (int)P->getDomainMap()->getGlobalNumElements()
+      << std::endl;
+
+  // Sanity
+  if ((int)P->getRangeMap()->getGlobalNumElements() != (int)hi_gold_out.size())
+    throw std::runtime_error("P range size does not match hi_gold_out");
+  if ((int)P->getDomainMap()->getGlobalNumElements() != (int)lo_gold_in.size())
+    throw std::runtime_error("P domain size does not match lo_gold_in");
+
+  // Build serial comparison maps
+  GO hi_num_global_dofs     = A->getRowMap()->getGlobalNumElements();
+  GO hi_num_serial_elements = !MyPID ? hi_num_global_dofs : 0;
+  RCP<Map> hi_SerialMap     = MapFactory::Build(lib, hi_num_global_dofs, hi_num_serial_elements, 0, comm);
+
+  GO lo_num_global_dofs     = P->getDomainMap()->getGlobalNumElements();
+  GO lo_num_serial_elements = !MyPID ? lo_num_global_dofs : 0;
+  RCP<Map> lo_SerialMap     = MapFactory::Build(lib, lo_num_global_dofs, lo_num_serial_elements, 0, comm);
+
+  RCP<Export> lo_importer = ExportFactory::Build(lo_SerialMap, P->getDomainMap());
+  RCP<Export> hi_importer = ExportFactory::Build(A->getRowMap(), hi_SerialMap);
+
+  // Allocate some vectors
+  RCP<Vector> s_InVec      = VectorFactory::Build(lo_SerialMap);
+  RCP<Vector> p_InVec      = VectorFactory::Build(P->getDomainMap());
+  RCP<Vector> s_OutVec     = VectorFactory::Build(hi_SerialMap);
+  RCP<Vector> s_codeOutput = VectorFactory::Build(hi_SerialMap);
+  RCP<Vector> p_codeOutput = VectorFactory::Build(A->getRowMap());
+
+  // Fill serial GOLD vecs on Proc 0
+  if (!MyPID) {
+    for (size_t i = 0; i < (size_t)lo_gold_in.size(); i++)
+      s_InVec->replaceLocalValue(i, lo_gold_in[i]);
+
+    for (size_t i = 0; i < (size_t)hi_gold_out.size(); i++)
+      s_OutVec->replaceLocalValue(i, hi_gold_out[i]);
+  }
 
+  // Migrate input data
+  p_InVec->doExport(*s_InVec, *lo_importer, Xpetra::ADD);
 
-    //if(!MyPID) printf("Diff norm = %10.4e\n",norm2[0]);
-    out << "Diff norm = " << norm2[0] << std::endl;
+  // Apply P
+  P->apply(*p_InVec, *p_codeOutput);
 
-  }
+  // Migrate Output data
+  s_codeOutput->doExport(*p_codeOutput, *hi_importer, Xpetra::ADD);
 
+  // Compare vs. GOLD
+  s_codeOutput->update(-1.0, *s_OutVec, 1.0);
+  Teuchos::Array<MT> norm2(1);
+  s_codeOutput->norm2(norm2());
 
- /*********************************************************************************************************************/
- TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    // GOLD vector collection
-    std::vector<Scalar> p2_gold_in = {/*0,*/1,2,3,4,5,6,7,8,/*9*/};//Ignore Dirichlet unknowns
-    std::vector<Scalar> p2_gold_out= {0,1,2,3,4,5,6,7,8,9,
-                                  0.5,1.5,2.5,3.5,4.5,5.5,6.5,7.5,8.5};
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,2+p2_gold_in.size(),2,p2_gold_in,p2_gold_out,std::string("hgrad_line_c2"));
-  }
+  // if(!MyPID) printf("Diff norm = %10.4e\n",norm2[0]);
+  out << "Diff norm = " << norm2[0] << std::endl;
+}
 
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  // GOLD vector collection
+  std::vector<Scalar> p2_gold_in  = {/*0,*/ 1, 2, 3, 4, 5, 6, 7, 8, /*9*/};  // Ignore Dirichlet unknowns
+  std::vector<Scalar> p2_gold_out = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
+                                     0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 6.5, 7.5, 8.5};
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, 2 + p2_gold_in.size(), 2, p2_gold_in, p2_gold_out, std::string("hgrad_line_c2"));
+}
 
 /*********************************************************************************************************************/
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    // GOLD vector collection
-    size_t total_num_points=10;
-    int degree=3;
-    std::vector<Scalar> gold_in(total_num_points-2);
-    std::vector<Scalar> gold_out(total_num_points + (total_num_points-1) *(degree-1));
-    for(size_t i=0; i<total_num_points-2; i++) {
-      gold_in[i] = i+1;
-      gold_out[i+1] = i+1;
-    }
-    gold_out[0] = 0.0;
-    gold_out[total_num_points-1] = total_num_points-1;
-
-    size_t idx=total_num_points;
-    for(size_t i=0; i<total_num_points-1; i++) {
-      for(size_t j=0; j<(size_t)degree-1; j++) {
-        gold_out[idx] = i + ((double)j+1)/degree;
-        idx++;
-      }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  // GOLD vector collection
+  size_t total_num_points = 10;
+  int degree              = 3;
+  std::vector<Scalar> gold_in(total_num_points - 2);
+  std::vector<Scalar> gold_out(total_num_points + (total_num_points - 1) * (degree - 1));
+  for (size_t i = 0; i < total_num_points - 2; i++) {
+    gold_in[i]      = i + 1;
+    gold_out[i + 1] = i + 1;
+  }
+  gold_out[0]                    = 0.0;
+  gold_out[total_num_points - 1] = total_num_points - 1;
+
+  size_t idx = total_num_points;
+  for (size_t i = 0; i < total_num_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)degree - 1; j++) {
+      gold_out[idx] = i + ((double)j + 1) / degree;
+      idx++;
     }
-
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,2+gold_in.size(),3,gold_in,gold_out,std::string("hgrad_line_c3"));
   }
 
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, 2 + gold_in.size(), 3, gold_in, gold_out, std::string("hgrad_line_c3"));
+}
+
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p4, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-
-    // GOLD vector collection
-    size_t total_num_points=10;
-    int degree=4;
-
-    std::vector<Scalar> gold_in(total_num_points-2);
-    std::vector<Scalar> gold_out(total_num_points + (total_num_points-1) *(degree-1));
-    for(size_t i=0; i<total_num_points-2; i++) {
-      gold_in[i] = i+1;
-      gold_out[i+1] = i+1;
-    }
-    gold_out[0] = 0.0;
-    gold_out[total_num_points-1] = total_num_points-1;
-
-    size_t idx=total_num_points;
-    for(size_t i=0; i<total_num_points-1; i++) {
-      for(size_t j=0; j<(size_t)degree-1; j++) {
-        gold_out[idx] = i + ((double)j+1)/degree;
-        idx++;
-      }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p4, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+
+  // GOLD vector collection
+  size_t total_num_points = 10;
+  int degree              = 4;
+
+  std::vector<Scalar> gold_in(total_num_points - 2);
+  std::vector<Scalar> gold_out(total_num_points + (total_num_points - 1) * (degree - 1));
+  for (size_t i = 0; i < total_num_points - 2; i++) {
+    gold_in[i]      = i + 1;
+    gold_out[i + 1] = i + 1;
+  }
+  gold_out[0]                    = 0.0;
+  gold_out[total_num_points - 1] = total_num_points - 1;
+
+  size_t idx = total_num_points;
+  for (size_t i = 0; i < total_num_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)degree - 1; j++) {
+      gold_out[idx] = i + ((double)j + 1) / degree;
+      idx++;
     }
-
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,2+gold_in.size(),4,gold_in,gold_out,std::string("hgrad_line_c4"));
   }
 
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, 2 + gold_in.size(), 4, gold_in, gold_out, std::string("hgrad_line_c4"));
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=2;
-    std::string hi_basis("hgrad_line_c2");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c1");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    //     if(lib==Xpetra::UseEpetra) Params.set("coarse: type","RELAXATION");// FIXME remove when we sort out the OAZ issue
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-/*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=3;
-    std::string hi_basis("hgrad_line_c3");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c1");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 2;
+  std::string hi_basis("hgrad_line_c2");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c1");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  //     if(lib==Xpetra::UseEpetra) Params.set("coarse: type","RELAXATION");// FIXME remove when we sort out the OAZ issue
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=4;
-    std::string hi_basis("hgrad_line_c4");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c1");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
-
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 3;
+  std::string hi_basis("hgrad_line_c3");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c1");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class Basis>
-bool test_representative_basis(Teuchos::FancyOStream &out, const std::string & name, Intrepid2::EPointType ptype, int max_degree)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Intrepid2::CellTools<typename Node::device_type::execution_space> CellTools;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    bool success = true;
-    int combinationTestedCount = 0;
-
-    // Contruct a container that has the reference coordinates for the domain cell topology
-    RCP<Basis> linearBasis = rcp( new Basis(1,ptype) );
-    shards::CellTopology cellTopo = linearBasis->getBaseCellTopology();
-    int spaceDim = cellTopo.getDimension();
-    int vertexCount = cellTopo.getVertexCount();
-    FC refCellVertices,refCellVertex,physCellVertices,physCellVerticesPermuted;
-    /*
-     How many physical cells do we need to test adequately?  I believe the answer is just 2, even in 3D;
-     by running through all the symmetries we should have all the edge-to-edge pairings covered.
-     */
-    int numCells = 2;
-    double xTranslationForCell1 = 2.0; // shift to the right by 2
-    double pointTol = 1e-12;
-
-    resize(physCellVertices,numCells,vertexCount,spaceDim);
-    resize(physCellVerticesPermuted,numCells,vertexCount,spaceDim);
-    resize(refCellVertices,vertexCount,spaceDim);
-    resize(refCellVertex,3);
-
-    // regardless of spatial extent, CellTools::getReferenceVertex() populates 3 slots
-    for (int vertexOrdinal=0; vertexOrdinal<vertexCount; vertexOrdinal++)
-    {
-      CellTools::getReferenceVertex(refCellVertex, cellTopo, vertexOrdinal);
-      //UVM used here, accessing vertex coordinates on host that were populated on device.
-
-      Kokkos::parallel_for(Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, spaceDim), KOKKOS_LAMBDA (int d)
-      {
-        refCellVertices(vertexOrdinal,d) = refCellVertex(d);
-        //      cout << "refCellVertices(" << vertexOrdinal << "," << d << ") = " << refCellVertex(d) << endl;
-        // cell 0 is just the reference cell:
-        physCellVertices(0,vertexOrdinal,d) = refCellVertex(d);
-        // cell 1 is the reference cell, except that the x coords get translated
-        // NOTE: this will need to change to support non-hypercube topologies
-        physCellVertices(1,vertexOrdinal,d) = refCellVertex(d) + ((d==0) ? xTranslationForCell1 : 0);
-      });
-      Kokkos::fence();
-    }
-    Symmetries cellSymmetries = Symmetries::shardsSymmetries(cellTopo);
-    int symmetryCount = cellSymmetries.getPermutationCount();
-
-    for(int highPolyDegree=1; highPolyDegree<max_degree; highPolyDegree++) {
-      FC hi_DofCoords, lo_DofCoords;
-      RCP<Basis> hi = rcp(new Basis(highPolyDegree,ptype));
-      Kokkos::resize(hi_DofCoords,hi->getCardinality(),hi->getBaseCellTopology().getDimension());
-      hi->getDofCoords(hi_DofCoords);
-      auto hi_DofCoords_host = Kokkos::create_mirror_view(hi_DofCoords);
-      Kokkos::deep_copy(hi_DofCoords_host, hi_DofCoords);
-
-      // we'll want to create a global numbering for both high and low order bases
-      // --> we make a lambda function that accepts FC with dof coords as argument
-      auto getTwoCellNumbering = [pointTol,numCells,spaceDim,xTranslationForCell1](const typename FC::HostMirror &dofCoords) -> UniqueNumbering
-      {
-        int dofsPerCell = dofCoords.extent(0);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
-        vector<double> coords(spaceDim);
-        UniqueNumbering numbering(spaceDim, pointTol);
-
-        // number the guys on cell 0 first, then the ones on cell 1:
-        for (int cellOrdinal=0; cellOrdinal<numCells; cellOrdinal++)
-        {
-          for (int dofOrdinal=0; dofOrdinal<dofsPerCell; dofOrdinal++)
-          {
-            for (int d=0; d<spaceDim; d++)
-            {
-              if ((d==0) && (cellOrdinal == 1))
-              {
-                coords[d] = dofCoords(dofOrdinal,d) + xTranslationForCell1;
-              }
-              else
-              {
-                coords[d] = dofCoords(dofOrdinal,d);
-              }
-            }
-            numbering.getGlobalID(coords);
-          }
-        }
-        return numbering;
-      };
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 4;
+  std::string hi_basis("hgrad_line_c4");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c1");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
-      UniqueNumbering hiNumbering = getTwoCellNumbering(hi_DofCoords_host);
-      out << "Total dof count two cells of degree " << highPolyDegree << ": ";
-      out << hiNumbering.totalCount() << endl;
-
-      for(int lowPolyDegree=1; lowPolyDegree<highPolyDegree; lowPolyDegree++) {
-        out << "Testing with high order " << highPolyDegree << ", low order " << lowPolyDegree << endl;
-        RCP<Basis> lo = rcp(new Basis(lowPolyDegree,ptype));
-        Kokkos::resize(lo_DofCoords,lo->getCardinality(),lo->getBaseCellTopology().getDimension());
-        lo->getDofCoords(lo_DofCoords);
-	auto lo_DofCoords_host = Kokkos::create_mirror_view(lo_DofCoords);
-	Kokkos::deep_copy(lo_DofCoords_host, lo_DofCoords);
-        UniqueNumbering loNumbering = getTwoCellNumbering(lo_DofCoords_host);
-
-        // print out the high/low global numbering along the x=1 interface:
-        out << "Low-order global IDs along intercell interface:\n";
-        for (int lowOrdinal=0; lowOrdinal<lo->getCardinality(); lowOrdinal++)
-        {
-          vector<double> coords(lo_DofCoords.extent(1));
-          for (int d=0; d<int(lo_DofCoords.extent(1)); d++)
-          {
-            coords[d] = lo_DofCoords_host(lowOrdinal,d);
-          }
-          if (coords[0] == 1.0)
-          {
-            int globalOrdinal = loNumbering.getGlobalID(coords);
-            out << globalOrdinal << ": (";
-            for (int d=0; d<int(coords.size())-1; d++)
-            {
-              out << coords[d] << ",";
+/*********************************************************************************************************************/
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class Basis>
+bool test_representative_basis(Teuchos::FancyOStream &out, const std::string &name, Intrepid2::EPointType ptype, int max_degree) {
+#include "MueLu_UseShortNames.hpp"
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Intrepid2::CellTools<typename Node::device_type::execution_space> CellTools;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  bool success               = true;
+  int combinationTestedCount = 0;
+
+  // Contruct a container that has the reference coordinates for the domain cell topology
+  RCP<Basis> linearBasis        = rcp(new Basis(1, ptype));
+  shards::CellTopology cellTopo = linearBasis->getBaseCellTopology();
+  int spaceDim                  = cellTopo.getDimension();
+  int vertexCount               = cellTopo.getVertexCount();
+  FC refCellVertices, refCellVertex, physCellVertices, physCellVerticesPermuted;
+  /*
+   How many physical cells do we need to test adequately?  I believe the answer is just 2, even in 3D;
+   by running through all the symmetries we should have all the edge-to-edge pairings covered.
+   */
+  int numCells                = 2;
+  double xTranslationForCell1 = 2.0;  // shift to the right by 2
+  double pointTol             = 1e-12;
+
+  resize(physCellVertices, numCells, vertexCount, spaceDim);
+  resize(physCellVerticesPermuted, numCells, vertexCount, spaceDim);
+  resize(refCellVertices, vertexCount, spaceDim);
+  resize(refCellVertex, 3);
+
+  // regardless of spatial extent, CellTools::getReferenceVertex() populates 3 slots
+  for (int vertexOrdinal = 0; vertexOrdinal < vertexCount; vertexOrdinal++) {
+    CellTools::getReferenceVertex(refCellVertex, cellTopo, vertexOrdinal);
+    // UVM used here, accessing vertex coordinates on host that were populated on device.
+
+    Kokkos::parallel_for(
+        Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, spaceDim), KOKKOS_LAMBDA(int d) {
+          refCellVertices(vertexOrdinal, d) = refCellVertex(d);
+          //      cout << "refCellVertices(" << vertexOrdinal << "," << d << ") = " << refCellVertex(d) << endl;
+          // cell 0 is just the reference cell:
+          physCellVertices(0, vertexOrdinal, d) = refCellVertex(d);
+          // cell 1 is the reference cell, except that the x coords get translated
+          // NOTE: this will need to change to support non-hypercube topologies
+          physCellVertices(1, vertexOrdinal, d) = refCellVertex(d) + ((d == 0) ? xTranslationForCell1 : 0);
+        });
+    Kokkos::fence();
+  }
+  Symmetries cellSymmetries = Symmetries::shardsSymmetries(cellTopo);
+  int symmetryCount         = cellSymmetries.getPermutationCount();
+
+  for (int highPolyDegree = 1; highPolyDegree < max_degree; highPolyDegree++) {
+    FC hi_DofCoords, lo_DofCoords;
+    RCP<Basis> hi = rcp(new Basis(highPolyDegree, ptype));
+    Kokkos::resize(hi_DofCoords, hi->getCardinality(), hi->getBaseCellTopology().getDimension());
+    hi->getDofCoords(hi_DofCoords);
+    auto hi_DofCoords_host = Kokkos::create_mirror_view(hi_DofCoords);
+    Kokkos::deep_copy(hi_DofCoords_host, hi_DofCoords);
+
+    // we'll want to create a global numbering for both high and low order bases
+    // --> we make a lambda function that accepts FC with dof coords as argument
+    auto getTwoCellNumbering = [pointTol, numCells, spaceDim, xTranslationForCell1](const typename FC::HostMirror &dofCoords) -> UniqueNumbering {
+      int dofsPerCell = dofCoords.extent(0);
+
+      vector<double> coords(spaceDim);
+      UniqueNumbering numbering(spaceDim, pointTol);
+
+      // number the guys on cell 0 first, then the ones on cell 1:
+      for (int cellOrdinal = 0; cellOrdinal < numCells; cellOrdinal++) {
+        for (int dofOrdinal = 0; dofOrdinal < dofsPerCell; dofOrdinal++) {
+          for (int d = 0; d < spaceDim; d++) {
+            if ((d == 0) && (cellOrdinal == 1)) {
+              coords[d] = dofCoords(dofOrdinal, d) + xTranslationForCell1;
+            } else {
+              coords[d] = dofCoords(dofOrdinal, d);
             }
-            out << coords[coords.size()-1] << ")\n";
           }
+          numbering.getGlobalID(coords);
         }
-        // print out the high/low global numbering along the x=1 interface:
-        out << "High-order global IDs along intercell interface:\n";
-        for (int highOrdinal=0; highOrdinal<hi->getCardinality(); highOrdinal++)
-        {
-          vector<double> coords(hi_DofCoords.extent(1));
-          for (int d=0; d<int(hi_DofCoords.extent(1)); d++)
-          {
-            coords[d] = hi_DofCoords_host(highOrdinal,d);
-          }
-          if (coords[0] == 1.0)
-          {
-            int globalOrdinal = hiNumbering.getGlobalID(coords);
-            out << globalOrdinal << ": (";
-            for (int d=0; d<int(coords.size())-1; d++)
-            {
-              out << coords[d] << ",";
-            }
-            out << coords[coords.size()-1] << ")\n";
+      }
+      return numbering;
+    };
+
+    UniqueNumbering hiNumbering = getTwoCellNumbering(hi_DofCoords_host);
+    out << "Total dof count two cells of degree " << highPolyDegree << ": ";
+    out << hiNumbering.totalCount() << endl;
+
+    for (int lowPolyDegree = 1; lowPolyDegree < highPolyDegree; lowPolyDegree++) {
+      out << "Testing with high order " << highPolyDegree << ", low order " << lowPolyDegree << endl;
+      RCP<Basis> lo = rcp(new Basis(lowPolyDegree, ptype));
+      Kokkos::resize(lo_DofCoords, lo->getCardinality(), lo->getBaseCellTopology().getDimension());
+      lo->getDofCoords(lo_DofCoords);
+      auto lo_DofCoords_host = Kokkos::create_mirror_view(lo_DofCoords);
+      Kokkos::deep_copy(lo_DofCoords_host, lo_DofCoords);
+      UniqueNumbering loNumbering = getTwoCellNumbering(lo_DofCoords_host);
+
+      // print out the high/low global numbering along the x=1 interface:
+      out << "Low-order global IDs along intercell interface:\n";
+      for (int lowOrdinal = 0; lowOrdinal < lo->getCardinality(); lowOrdinal++) {
+        vector<double> coords(lo_DofCoords.extent(1));
+        for (int d = 0; d < int(lo_DofCoords.extent(1)); d++) {
+          coords[d] = lo_DofCoords_host(lowOrdinal, d);
+        }
+        if (coords[0] == 1.0) {
+          int globalOrdinal = loNumbering.getGlobalID(coords);
+          out << globalOrdinal << ": (";
+          for (int d = 0; d < int(coords.size()) - 1; d++) {
+            out << coords[d] << ",";
           }
+          out << coords[coords.size() - 1] << ")\n";
         }
-
-        // Get the candidates
-        double threshold = 1e-10;
-        std::vector<std::vector<size_t> > candidates;
-        MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis,FC>(*lo,hi_DofCoords,threshold,candidates);
-
-        // Correctness Test 1: Make sure that there are no duplicates in the representative lists / no low DOF has no candidates
-        std::vector<bool> is_candidate(hi_DofCoords.extent(0),false);
-        bool no_doubles = true;
-        for(int lowOrderDof=0; no_doubles && lowOrderDof<(int)candidates.size(); lowOrderDof++) {
-          if(candidates[lowOrderDof].size()==0) no_doubles=false; // this low DOF has no candidates!
-          for(int l=0; l<(int)candidates[lowOrderDof].size(); l++)
-            if(is_candidate[candidates[lowOrderDof][l]] == false) is_candidate[candidates[lowOrderDof][l]]=true;
-            else {no_doubles=false;break;} // this high-order dof was already claimed by an earlier low DOF!
+      }
+      // print out the high/low global numbering along the x=1 interface:
+      out << "High-order global IDs along intercell interface:\n";
+      for (int highOrdinal = 0; highOrdinal < hi->getCardinality(); highOrdinal++) {
+        vector<double> coords(hi_DofCoords.extent(1));
+        for (int d = 0; d < int(hi_DofCoords.extent(1)); d++) {
+          coords[d] = hi_DofCoords_host(highOrdinal, d);
         }
-        if(!no_doubles) {
-          out<<"ERROR: "<<name<<" The 'no duplicates' test fails w/ lo/hi = "<< lowPolyDegree <<"/"<< highPolyDegree <<std::endl;
-          return false;
+        if (coords[0] == 1.0) {
+          int globalOrdinal = hiNumbering.getGlobalID(coords);
+          out << globalOrdinal << ": (";
+          for (int d = 0; d < int(coords.size()) - 1; d++) {
+            out << coords[d] << ",";
+          }
+          out << coords[coords.size() - 1] << ")\n";
         }
+      }
 
-        // Correctness Test 2: Try 2 elements, in all possible relative orientations, and confirm that the
-        //                     "lowest global ordinal" tie-breaker always returns the same thing for both neighbors
-	auto physCellVertices_host = Kokkos::create_mirror_view(physCellVertices);
-	auto physCellVerticesPermuted_host = Kokkos::create_mirror_view(physCellVerticesPermuted);
-	Kokkos::deep_copy(physCellVertices_host, physCellVertices);
-        for (int permOrdinal0=0; permOrdinal0<symmetryCount; permOrdinal0++)
-        {
-          vector<int> perm0 = cellSymmetries.getPermutation(permOrdinal0);
-          for (int vertexOrdinal=0; vertexOrdinal<vertexCount; vertexOrdinal++)
-          {
-            int mappedVertexOrdinal = perm0[vertexOrdinal];
-            for (int d=0; d<spaceDim; d++)
-            {
-              physCellVerticesPermuted_host(0,vertexOrdinal,d) = physCellVertices_host(0,mappedVertexOrdinal,d);
-            }
+      // Get the candidates
+      double threshold = 1e-10;
+      std::vector<std::vector<size_t>> candidates;
+      MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis, FC>(*lo, hi_DofCoords, threshold, candidates);
+
+      // Correctness Test 1: Make sure that there are no duplicates in the representative lists / no low DOF has no candidates
+      std::vector<bool> is_candidate(hi_DofCoords.extent(0), false);
+      bool no_doubles = true;
+      for (int lowOrderDof = 0; no_doubles && lowOrderDof < (int)candidates.size(); lowOrderDof++) {
+        if (candidates[lowOrderDof].size() == 0) no_doubles = false;  // this low DOF has no candidates!
+        for (int l = 0; l < (int)candidates[lowOrderDof].size(); l++)
+          if (is_candidate[candidates[lowOrderDof][l]] == false)
+            is_candidate[candidates[lowOrderDof][l]] = true;
+          else {
+            no_doubles = false;
+            break;
+          }  // this high-order dof was already claimed by an earlier low DOF!
+      }
+      if (!no_doubles) {
+        out << "ERROR: " << name << " The 'no duplicates' test fails w/ lo/hi = " << lowPolyDegree << "/" << highPolyDegree << std::endl;
+        return false;
+      }
+
+      // Correctness Test 2: Try 2 elements, in all possible relative orientations, and confirm that the
+      //                     "lowest global ordinal" tie-breaker always returns the same thing for both neighbors
+      auto physCellVertices_host         = Kokkos::create_mirror_view(physCellVertices);
+      auto physCellVerticesPermuted_host = Kokkos::create_mirror_view(physCellVerticesPermuted);
+      Kokkos::deep_copy(physCellVertices_host, physCellVertices);
+      for (int permOrdinal0 = 0; permOrdinal0 < symmetryCount; permOrdinal0++) {
+        vector<int> perm0 = cellSymmetries.getPermutation(permOrdinal0);
+        for (int vertexOrdinal = 0; vertexOrdinal < vertexCount; vertexOrdinal++) {
+          int mappedVertexOrdinal = perm0[vertexOrdinal];
+          for (int d = 0; d < spaceDim; d++) {
+            physCellVerticesPermuted_host(0, vertexOrdinal, d) = physCellVertices_host(0, mappedVertexOrdinal, d);
           }
+        }
 
-          // brute force search for the side of cell shared with neighbor
-          // this is the one that has points with x coordinates equal to 1.0
-          // we'll want to do this once for cell 0, and once for cell 1, so we make it a lambda
-          // (NOTE: this will need to change for non-hypercube topology support)
-          auto searchForX1Side = [cellTopo,physCellVerticesPermuted_host,spaceDim](int cellOrdinal) -> int
-          {
-
-            // Line<2> gives wrong answers for getSideCount() and getNodeCount(), so we handle 1D case separately:
-            if (spaceDim == 1)
-            {
-              int sideCount = 2;
-              for (int sideVertexOrdinal=0; sideVertexOrdinal<sideCount; sideVertexOrdinal++)
-              {
-                int cellVertexOrdinal = sideVertexOrdinal;
-                if (physCellVerticesPermuted_host(cellOrdinal,cellVertexOrdinal,0) == 1.0)
-                {
-                  return sideVertexOrdinal;
-                }
+        // brute force search for the side of cell shared with neighbor
+        // this is the one that has points with x coordinates equal to 1.0
+        // we'll want to do this once for cell 0, and once for cell 1, so we make it a lambda
+        // (NOTE: this will need to change for non-hypercube topology support)
+        auto searchForX1Side = [cellTopo, physCellVerticesPermuted_host, spaceDim](int cellOrdinal) -> int {
+          // Line<2> gives wrong answers for getSideCount() and getNodeCount(), so we handle 1D case separately:
+          if (spaceDim == 1) {
+            int sideCount = 2;
+            for (int sideVertexOrdinal = 0; sideVertexOrdinal < sideCount; sideVertexOrdinal++) {
+              int cellVertexOrdinal = sideVertexOrdinal;
+              if (physCellVerticesPermuted_host(cellOrdinal, cellVertexOrdinal, 0) == 1.0) {
+                return sideVertexOrdinal;
               }
-              return -1;
             }
-            int sideCount = (int) cellTopo.getSideCount();
-            for (int sideOrdinal=0; sideOrdinal<sideCount; sideOrdinal++)
-            {
-              int sideVertexCount = cellTopo.getNodeCount(spaceDim-1, sideOrdinal);
-              bool matchFound = true;
-              for (int sideVertexOrdinal=0; sideVertexOrdinal<sideVertexCount; sideVertexOrdinal++)
-              {
-                int cellVertexOrdinal = cellTopo.getNodeMap(spaceDim-1, sideOrdinal, sideVertexOrdinal);
-                if (physCellVerticesPermuted_host(cellOrdinal,cellVertexOrdinal,0) != 1.0)
-                {
-                  matchFound = false;
-                  break;
-                }
-              }
-              if (matchFound)
-              {
-                return sideOrdinal;
+            return -1;
+          }
+          int sideCount = (int)cellTopo.getSideCount();
+          for (int sideOrdinal = 0; sideOrdinal < sideCount; sideOrdinal++) {
+            int sideVertexCount = cellTopo.getNodeCount(spaceDim - 1, sideOrdinal);
+            bool matchFound     = true;
+            for (int sideVertexOrdinal = 0; sideVertexOrdinal < sideVertexCount; sideVertexOrdinal++) {
+              int cellVertexOrdinal = cellTopo.getNodeMap(spaceDim - 1, sideOrdinal, sideVertexOrdinal);
+              if (physCellVerticesPermuted_host(cellOrdinal, cellVertexOrdinal, 0) != 1.0) {
+                matchFound = false;
+                break;
               }
             }
-            return -1;
-          };
+            if (matchFound) {
+              return sideOrdinal;
+            }
+          }
+          return -1;
+        };
 
-          int cell0Side = searchForX1Side(0);
-          //        out << "cell 0 side is " << cell0Side << endl;
-
-          for (int permOrdinal1=0; permOrdinal1<symmetryCount; permOrdinal1++)
-          {
-            vector<int> perm1 = cellSymmetries.getPermutation(permOrdinal1);
-            for (int vertexOrdinal=0; vertexOrdinal<vertexCount; vertexOrdinal++)
-            {
-              int mappedVertexOrdinal = perm1[vertexOrdinal];
-              for (int d=0; d<spaceDim; d++)
-              {
-                physCellVerticesPermuted_host(1,vertexOrdinal,d) = physCellVertices_host(1,mappedVertexOrdinal,d);
-              }
+        int cell0Side = searchForX1Side(0);
+        //        out << "cell 0 side is " << cell0Side << endl;
+
+        for (int permOrdinal1 = 0; permOrdinal1 < symmetryCount; permOrdinal1++) {
+          vector<int> perm1 = cellSymmetries.getPermutation(permOrdinal1);
+          for (int vertexOrdinal = 0; vertexOrdinal < vertexCount; vertexOrdinal++) {
+            int mappedVertexOrdinal = perm1[vertexOrdinal];
+            for (int d = 0; d < spaceDim; d++) {
+              physCellVerticesPermuted_host(1, vertexOrdinal, d) = physCellVertices_host(1, mappedVertexOrdinal, d);
             }
-            // get the mapped dof coords for lo and high bases:
-            FC lo_physDofCoords, hi_physDofCoords;
-            Kokkos::resize(lo_physDofCoords, numCells, lo->getCardinality(), cellTopo.getDimension());
-            Kokkos::resize(hi_physDofCoords, numCells, hi->getCardinality(), cellTopo.getDimension());
-	    
-	    Kokkos::deep_copy(physCellVerticesPermuted,physCellVerticesPermuted_host);
-            CellTools::mapToPhysicalFrame(lo_physDofCoords, lo_DofCoords, physCellVerticesPermuted, cellTopo);
-            CellTools::mapToPhysicalFrame(hi_physDofCoords, hi_DofCoords, physCellVerticesPermuted, cellTopo);
-	    auto lo_physDofCoords_host = Kokkos::create_mirror_view(lo_physDofCoords);
-	    auto hi_physDofCoords_host = Kokkos::create_mirror_view(hi_physDofCoords);
-	    Kokkos::deep_copy(lo_physDofCoords_host, lo_physDofCoords);
-	    Kokkos::deep_copy(hi_physDofCoords_host, hi_physDofCoords);
-            Kokkos::fence(); // mapToPhysicalFrame calls getValues which calls kernels, so fence is required before UVM reads below
-
-            int cell1Side = searchForX1Side(1);
-
-            /*
-             We want to verify that the neighboring cells (as permuted) agree on their
-             representation of the low-order basis in their selection of high-order basis
-             members.  It suffices to check that the *set* of candidate high-order basis functions
-             is the same; then for any global numbering the "lowest global number" tie-breaker
-             will select the same function.
-
-             The logic is this.  We have:
-             - a global numbering for the physical dof coords of the coarse discretization
-             - a global numbering for the physical dof coords of the fine discretization
-             - a local map from lo to hi on the reference cell
-             - a map from local dofs on each cell to the physical dof coords for the {lo|hi} basis
-
-             On each cell, then, we can map from the local low dof ordinal to the local high dof ordinal.
-             At the same time, we can map the local ordinals for high and low to global ordinals, giving
-             us a mapping from global low ordinals to global high ordinals.  We then can compare the mapping
-             to verify that the two cells agree.
-             */
-
-            auto constructMap = [lo, lo_physDofCoords_host, &loNumbering, candidates, hi_physDofCoords_host, &hiNumbering, spaceDim]
-            (int cellOrdinal, int sideOrdinal) -> map<int, set<int>>
-            {
-              map<int,set<int>> globalLowToHighMap;
-
-              vector<int> cell_loDofOrdinals = localDofOrdinalsForSide(lo, sideOrdinal);
-              for (int lowLocalOrdinal : cell_loDofOrdinals)
-              {
-                vector<double> loCoords(spaceDim);
-                for (int d=0; d<spaceDim; d++)
-                {
-                  loCoords[d] = lo_physDofCoords_host(cellOrdinal,lowLocalOrdinal,d);
-                }
-                int lowGlobalNumber = loNumbering.getGlobalID(loCoords);
-                vector<size_t> highLocalOrdinals = candidates[lowLocalOrdinal];
-                for (int highLocalOrdinal : highLocalOrdinals)
-                {
-                  vector<double> hiCoords(spaceDim);
-                  for (int d=0; d<spaceDim; d++)
-                  {
-                    hiCoords[d] = hi_physDofCoords_host(cellOrdinal,highLocalOrdinal,d);
-                  }
-                  int highGlobalNumber = hiNumbering.getGlobalID(hiCoords);
-                  globalLowToHighMap[lowGlobalNumber].insert(highGlobalNumber);
+          }
+          // get the mapped dof coords for lo and high bases:
+          FC lo_physDofCoords, hi_physDofCoords;
+          Kokkos::resize(lo_physDofCoords, numCells, lo->getCardinality(), cellTopo.getDimension());
+          Kokkos::resize(hi_physDofCoords, numCells, hi->getCardinality(), cellTopo.getDimension());
+
+          Kokkos::deep_copy(physCellVerticesPermuted, physCellVerticesPermuted_host);
+          CellTools::mapToPhysicalFrame(lo_physDofCoords, lo_DofCoords, physCellVerticesPermuted, cellTopo);
+          CellTools::mapToPhysicalFrame(hi_physDofCoords, hi_DofCoords, physCellVerticesPermuted, cellTopo);
+          auto lo_physDofCoords_host = Kokkos::create_mirror_view(lo_physDofCoords);
+          auto hi_physDofCoords_host = Kokkos::create_mirror_view(hi_physDofCoords);
+          Kokkos::deep_copy(lo_physDofCoords_host, lo_physDofCoords);
+          Kokkos::deep_copy(hi_physDofCoords_host, hi_physDofCoords);
+          Kokkos::fence();  // mapToPhysicalFrame calls getValues which calls kernels, so fence is required before UVM reads below
+
+          int cell1Side = searchForX1Side(1);
+
+          /*
+           We want to verify that the neighboring cells (as permuted) agree on their
+           representation of the low-order basis in their selection of high-order basis
+           members.  It suffices to check that the *set* of candidate high-order basis functions
+           is the same; then for any global numbering the "lowest global number" tie-breaker
+           will select the same function.
+
+           The logic is this.  We have:
+           - a global numbering for the physical dof coords of the coarse discretization
+           - a global numbering for the physical dof coords of the fine discretization
+           - a local map from lo to hi on the reference cell
+           - a map from local dofs on each cell to the physical dof coords for the {lo|hi} basis
+
+           On each cell, then, we can map from the local low dof ordinal to the local high dof ordinal.
+           At the same time, we can map the local ordinals for high and low to global ordinals, giving
+           us a mapping from global low ordinals to global high ordinals.  We then can compare the mapping
+           to verify that the two cells agree.
+           */
+
+          auto constructMap = [lo, lo_physDofCoords_host, &loNumbering, candidates, hi_physDofCoords_host, &hiNumbering, spaceDim](int cellOrdinal, int sideOrdinal) -> map<int, set<int>> {
+            map<int, set<int>> globalLowToHighMap;
+
+            vector<int> cell_loDofOrdinals = localDofOrdinalsForSide(lo, sideOrdinal);
+            for (int lowLocalOrdinal : cell_loDofOrdinals) {
+              vector<double> loCoords(spaceDim);
+              for (int d = 0; d < spaceDim; d++) {
+                loCoords[d] = lo_physDofCoords_host(cellOrdinal, lowLocalOrdinal, d);
+              }
+              int lowGlobalNumber              = loNumbering.getGlobalID(loCoords);
+              vector<size_t> highLocalOrdinals = candidates[lowLocalOrdinal];
+              for (int highLocalOrdinal : highLocalOrdinals) {
+                vector<double> hiCoords(spaceDim);
+                for (int d = 0; d < spaceDim; d++) {
+                  hiCoords[d] = hi_physDofCoords_host(cellOrdinal, highLocalOrdinal, d);
                 }
+                int highGlobalNumber = hiNumbering.getGlobalID(hiCoords);
+                globalLowToHighMap[lowGlobalNumber].insert(highGlobalNumber);
               }
-              return globalLowToHighMap;
-            };
+            }
+            return globalLowToHighMap;
+          };
 
-            map<int,set<int>> cell0_mapping = constructMap(0,cell0Side);
-            map<int,set<int>> cell1_mapping = constructMap(1,cell1Side);
+          map<int, set<int>> cell0_mapping = constructMap(0, cell0Side);
+          map<int, set<int>> cell1_mapping = constructMap(1, cell1Side);
 
-            // verify the two maps are identical:
-            if (cell0_mapping.size() != cell1_mapping.size())
-            {
-              success = false;
-              out << "cell 0 and cell 1 mapping do not have the same number of low global ordinal entries!\n";
-            }
-            else
-            {
-              auto cell0_it = cell0_mapping.begin();
-              auto cell1_it = cell1_mapping.begin();
-              while (cell0_it != cell0_mapping.end())
-              {
-                if (cell0_it->first != cell1_it->first)
-                {
-                  success = false;
-                  out << "cell 0 and cell 1 have different low global ordinal entries; ";
-                  out << cell0_it->first << " != " << cell1_it->first << endl;
-                }
-                else if (cell0_it->second.size() != cell1_it->second.size())
-                {
-                  success = false;
-                  out << "cell 0 and cell 1 have a different number of high global ordinal entries for low global ordinal ";
-                  out << cell0_it->first << cell0_it->second.size() << " != " << cell1_it->second.size() << endl;
+          // verify the two maps are identical:
+          if (cell0_mapping.size() != cell1_mapping.size()) {
+            success = false;
+            out << "cell 0 and cell 1 mapping do not have the same number of low global ordinal entries!\n";
+          } else {
+            auto cell0_it = cell0_mapping.begin();
+            auto cell1_it = cell1_mapping.begin();
+            while (cell0_it != cell0_mapping.end()) {
+              if (cell0_it->first != cell1_it->first) {
+                success = false;
+                out << "cell 0 and cell 1 have different low global ordinal entries; ";
+                out << cell0_it->first << " != " << cell1_it->first << endl;
+              } else if (cell0_it->second.size() != cell1_it->second.size()) {
+                success = false;
+                out << "cell 0 and cell 1 have a different number of high global ordinal entries for low global ordinal ";
+                out << cell0_it->first << cell0_it->second.size() << " != " << cell1_it->second.size() << endl;
+              } else {
+                // check that the set contents are identical
+                auto cell0_set_it = cell0_it->second.begin();
+                auto cell1_set_it = cell1_it->second.begin();
+                bool setsMatch    = true;
+                while (cell0_set_it != cell0_it->second.end()) {
+                  if (*cell0_set_it != *cell1_set_it) {
+                    setsMatch = false;
+                    break;
+                  }
+                  cell0_set_it++;
+                  cell1_set_it++;
                 }
-                else
-                {
-                  // check that the set contents are identical
-                  auto cell0_set_it = cell0_it->second.begin();
-                  auto cell1_set_it = cell1_it->second.begin();
-                  bool setsMatch = true;
-                  while (cell0_set_it != cell0_it->second.end())
-                  {
-                    if (*cell0_set_it != *cell1_set_it)
-                    {
-                      setsMatch = false;
-                      break;
-                    }
-                    cell0_set_it++;
-                    cell1_set_it++;
+                if (!setsMatch) {
+                  int lowGlobalOrdinal = cell0_it->first;
+                  success              = false;
+                  out << "with permutation selections " << permOrdinal0 << " and " << permOrdinal1 << ", ";
+                  out << "cell 0 and cell 1 maps differ in the high global entries for low global ordinal ";
+                  out << lowGlobalOrdinal << ": { ";
+                  for (int highGlobalOrdinal : cell0_it->second) {
+                    out << highGlobalOrdinal << " ";
                   }
-                  if (!setsMatch)
-                  {
-                    int lowGlobalOrdinal = cell0_it->first;
-                    success = false;
-                    out << "with permutation selections " << permOrdinal0 << " and " << permOrdinal1 << ", ";
-                    out << "cell 0 and cell 1 maps differ in the high global entries for low global ordinal ";
-                    out << lowGlobalOrdinal << ": { ";
-                    for (int highGlobalOrdinal : cell0_it->second) { out << highGlobalOrdinal << " ";}
-                    out << "} != { ";
-                    for (int highGlobalOrdinal : cell1_it->second) { out << highGlobalOrdinal << " ";}
-                    out << "}\n";
+                  out << "} != { ";
+                  for (int highGlobalOrdinal : cell1_it->second) {
+                    out << highGlobalOrdinal << " ";
                   }
+                  out << "}\n";
                 }
-
-                cell0_it++;
-                cell1_it++;
               }
-            }
 
-            combinationTestedCount++;
+              cell0_it++;
+              cell1_it++;
+            }
           }
+
+          combinationTestedCount++;
         }
       }
     }
-    cout << "Tested " << combinationTestedCount << " lo/hi, two-cell permutation combinations.\n";
-    return success;
   }
+  cout << "Tested " << combinationTestedCount << " lo/hi, two-cell permutation combinations.\n";
+  return success;
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_LINE_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_LINE_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Intrepid2::Basis_HGRAD_LINE_Cn_FEM<ES,MT,MT> Basis;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Intrepid2::Basis_HGRAD_LINE_Cn_FEM<ES, MT, MT> Basis;
 
-
-    bool rv = test_representative_basis<Scalar,LocalOrdinal,GlobalOrdinal,Node,Basis>(out," GenerateRepresentativeBasisNodes_LINE_EQUISPACED",Intrepid2::POINTTYPE_EQUISPACED,MAX_LINE_DEGREE);
-    TEST_EQUALITY(rv,true);
-  }
+  bool rv = test_representative_basis<Scalar, LocalOrdinal, GlobalOrdinal, Node, Basis>(out, " GenerateRepresentativeBasisNodes_LINE_EQUISPACED", Intrepid2::POINTTYPE_EQUISPACED, MAX_LINE_DEGREE);
+  TEST_EQUALITY(rv, true);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> Basis;
-
-    bool rv = test_representative_basis<Scalar,LocalOrdinal,GlobalOrdinal,Node,Basis>(out," GenerateRepresentativeBasisNodes_QUAD_EQUISPACED",Intrepid2::POINTTYPE_EQUISPACED,MAX_QUAD_DEGREE);
-    TEST_EQUALITY(rv,true);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-/*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Spectral, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> Basis;
-
-    const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);// Not sure why I have to do this...
-    bool rv = test_representative_basis<Scalar,LocalOrdinal,GlobalOrdinal,Node,Basis>(out," GenerateRepresentativeBasisNodes_QUAD_SPECTRAL",POINTTYPE_SPECTRAL,MAX_QUAD_DEGREE);
-    TEST_EQUALITY(rv,true);
-  }
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> Basis;
 
+  bool rv = test_representative_basis<Scalar, LocalOrdinal, GlobalOrdinal, Node, Basis>(out, " GenerateRepresentativeBasisNodes_QUAD_EQUISPACED", Intrepid2::POINTTYPE_EQUISPACED, MAX_QUAD_DEGREE);
+  TEST_EQUALITY(rv, true);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES,MT,MT> Basis;
-
-    bool rv = test_representative_basis<Scalar,LocalOrdinal,GlobalOrdinal,Node,Basis>(out," GenerateRepresentativeBasisNodes_HEX_EQUISPACED",Intrepid2::POINTTYPE_EQUISPACED,MAX_HEX_DEGREE);
-    TEST_EQUALITY(rv,true);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Spectral, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> Basis;
+
+  const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);  // Not sure why I have to do this...
+  bool rv                                        = test_representative_basis<Scalar, LocalOrdinal, GlobalOrdinal, Node, Basis>(out, " GenerateRepresentativeBasisNodes_QUAD_SPECTRAL", POINTTYPE_SPECTRAL, MAX_QUAD_DEGREE);
+  TEST_EQUALITY(rv, true);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Spectral, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES,MT,MT> Basis;
-
-    const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);// Not sure why I have to do this...
-    bool rv = test_representative_basis<Scalar,LocalOrdinal,GlobalOrdinal,Node,Basis>(out," GenerateRepresentativeBasisNodes_HEX_SPECTRAL",POINTTYPE_SPECTRAL,MAX_HEX_DEGREE);
-    TEST_EQUALITY(rv,true);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Equispaced, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES, MT, MT> Basis;
+
+  bool rv = test_representative_basis<Scalar, LocalOrdinal, GlobalOrdinal, Node, Basis>(out, " GenerateRepresentativeBasisNodes_HEX_EQUISPACED", Intrepid2::POINTTYPE_EQUISPACED, MAX_HEX_DEGREE);
+  TEST_EQUALITY(rv, true);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Spectral, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Intrepid2::Basis_HGRAD_HEX_Cn_FEM<ES, MT, MT> Basis;
+
+  const Intrepid2::EPointType POINTTYPE_SPECTRAL = static_cast<Intrepid2::EPointType>(1);  // Not sure why I have to do this...
+  bool rv                                        = test_representative_basis<Scalar, LocalOrdinal, GlobalOrdinal, Node, Basis>(out, " GenerateRepresentativeBasisNodes_HEX_SPECTRAL", POINTTYPE_SPECTRAL, MAX_HEX_DEGREE);
+  TEST_EQUALITY(rv, true);
+}
 
-    // NOTE: We need more tests for this that do pn to pm
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Node::device_type::execution_space ES;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<LO,typename Node::device_type> FCi;
-    typedef Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT> LoBasis;
-    typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT> HiBasis;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    int max_degree = MAX_QUAD_DEGREE;
-    double threshold = 1e-10;
-    GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-
-    // Lo
-    RCP<LoBasis> lo = rcp(new LoBasis());
-    size_t numLo = lo->getCardinality();
-
-    for(int i=0;i<max_degree; i++) {
-      RCP<HiBasis> hi = rcp(new HiBasis(i,Intrepid2::POINTTYPE_EQUISPACED));
-      size_t numHi    = hi->getCardinality();
-
-      // The quad-only stuff
-      std::vector<size_t> lo_node_in_hi;
-      FC hi_dofCoords;
-      MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT,typename Node::device_type>(hi,lo_node_in_hi,hi_dofCoords);
-      FCi hi_e2n("hi_e2n",1,numHi);
-      FCi lo_e2n("lo_e2n",1,numLo);
-
-      // Dummy elem2node map
-      Kokkos::parallel_for("IntrepidPCoarsenFactory,GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1",
-			   Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, numHi), KOKKOS_LAMBDA (int j) {
-        hi_e2n(0,j)    = j;
-      });
-      Kokkos::fence();
-      Teuchos::Array<GO> hi_colids(numHi);
-      for(size_t j=0; j<numHi; j++) {
-        hi_colids[j] = j;
-      }
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // NOTE: We need more tests for this that do pn to pm
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Node::device_type::execution_space ES;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<LO, typename Node::device_type> FCi;
+  typedef Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT> LoBasis;
+  typedef Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT> HiBasis;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  int max_degree   = MAX_QUAD_DEGREE;
+  double threshold = 1e-10;
+  GO gst_invalid   = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+
+  // Lo
+  RCP<LoBasis> lo = rcp(new LoBasis());
+  size_t numLo    = lo->getCardinality();
+
+  for (int i = 0; i < max_degree; i++) {
+    RCP<HiBasis> hi = rcp(new HiBasis(i, Intrepid2::POINTTYPE_EQUISPACED));
+    size_t numHi    = hi->getCardinality();
+
+    // The quad-only stuff
+    std::vector<size_t> lo_node_in_hi;
+    FC hi_dofCoords;
+    MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT, typename Node::device_type>(hi, lo_node_in_hi, hi_dofCoords);
+    FCi hi_e2n("hi_e2n", 1, numHi);
+    FCi lo_e2n("lo_e2n", 1, numLo);
+
+    // Dummy elem2node map
+    Kokkos::parallel_for(
+        "IntrepidPCoarsenFactory,GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1",
+        Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, numHi), KOKKOS_LAMBDA(int j) {
+          hi_e2n(0, j) = j;
+        });
+    Kokkos::fence();
+    Teuchos::Array<GO> hi_colids(numHi);
+    for (size_t j = 0; j < numHi; j++) {
+      hi_colids[j] = j;
+    }
 
-      // Dummy column map
-      RCP<const Map> hi_colMap      = MapFactory::Build(lib,gst_invalid,hi_colids(),0,comm);
+    // Dummy column map
+    RCP<const Map> hi_colMap = MapFactory::Build(lib, gst_invalid, hi_colids(), 0, comm);
 
-      // The dynamic stuff
-      std::vector<std::vector<size_t> > candidates;
-      MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<LoBasis,FC>(*lo,hi_dofCoords,threshold,candidates);
-      MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs<LO,GO,Node,FCi>(candidates,hi_e2n,hi_colMap,lo_e2n);
+    // The dynamic stuff
+    std::vector<std::vector<size_t>> candidates;
+    MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<LoBasis, FC>(*lo, hi_dofCoords, threshold, candidates);
+    MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs<LO, GO, Node, FCi>(candidates, hi_e2n, hi_colMap, lo_e2n);
 
-      // Compare and make sure we're cool
-      bool node_diff = false;
-      auto lo_e2n_host = Kokkos::create_mirror_view(lo_e2n);
-      Kokkos::deep_copy(lo_e2n_host, lo_e2n);
-      for(size_t j=0; j<numLo; j++)
-        if(lo_node_in_hi[j]!=(size_t)lo_e2n_host(0,j)) node_diff=true;
+    // Compare and make sure we're cool
+    bool node_diff   = false;
+    auto lo_e2n_host = Kokkos::create_mirror_view(lo_e2n);
+    Kokkos::deep_copy(lo_e2n_host, lo_e2n);
+    for (size_t j = 0; j < numLo; j++)
+      if (lo_node_in_hi[j] != (size_t)lo_e2n_host(0, j)) node_diff = true;
 #if 0
       printf("[%d] Comparison = ",i);
       for(size_t j=0; j<numLo; j++)
         printf("%d|%d ",(int)lo_node_in_hi[j],(int)lo_e2n(0,j));
       printf("\n");
 #endif
-      TEST_EQUALITY(node_diff,false);
-      }
+    TEST_EQUALITY(node_diff, false);
   }
+}
 
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // NOTE: We need more tests for this that do pn to pm
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
+  typedef Kokkos::DynRankView<MT, typename Node::device_type> FC;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+  typedef typename Node::device_type::execution_space ES;
+  typedef Intrepid2::Basis<ES, MT, MT> Basis;
+  GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+
+  out << "version: " << MueLu::Version() << std::endl;
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  int max_degree = Intrepid2::Parameters::MaxOrder;
 
-
-  /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-    // NOTE: We need more tests for this that do pn to pm
-  #   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType MT;
-    typedef Kokkos::DynRankView<MT,typename Node::device_type> FC;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-    typedef typename Node::device_type::execution_space ES;
-    typedef Intrepid2::Basis<ES,MT,MT> Basis;
-    GO gst_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-
-    out << "version: " << MueLu::Version() << std::endl;
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    int max_degree=Intrepid2::Parameters::MaxOrder;
-
-    {
-      //QUAD
-      // A one element test with Kirby-numbered nodes where the top edge is not owned
-      RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES,MT,MT>());
-
-
-      for(int degree=2; degree < max_degree; degree++) {
-        int Nn = (degree+1)*(degree+1);
-        RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES,MT,MT>(degree,Intrepid2::POINTTYPE_EQUISPACED));
-        FCi hi_e2n("hi_e2n",1,Nn), lo_e2n, lo_e2n_mk2;
-
-        std::vector<bool> hi_owned(Nn,false),lo_owned, lo_owned_mk2;
-        std::vector<size_t> lo_node_in_hi;
-        std::vector<LO> hi_to_lo_map,  hi_to_lo_map_mk2;
-        int lo_numOwnedNodes=0, lo_numOwnedNodes_mk2=0;
-        FC hi_dofCoords;
-
-        // El2node / ownership / colmap
-        Kokkos::parallel_for("IntrepidPCoarsenFactory,BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1",
-			     Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA (int i) {
-           hi_e2n(0,i)=i;
-        });
-	Kokkos::fence();
-        Teuchos::Array<GO> hi_colids(Nn);
-        for(int i=0; i<Nn; i++) {
-          hi_colids[i] = i;
-          if(i < Nn-(degree+1)) hi_owned[i]=true;
-        }
-
-        Teuchos::ArrayRCP<const int> is_dirichlet(Nn,0);
-
-        /*** Do stuff the injection way ***/
-        MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT,typename Node::device_type>(hi,lo_node_in_hi,hi_dofCoords);
-        MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n,hi_owned,lo_node_in_hi,is_dirichlet,lo_e2n,lo_owned,hi_to_lo_map,lo_numOwnedNodes);
-
-        /*** Do stuff the representative way ***/
-        RCP<const Map> hi_colMap      = MapFactory::Build(lib,gst_invalid,hi_colids(),0,comm);
-        FCi lo_elemToHiRepresentativeNode;
-        double threshold = 1e-10;
-        std::vector<std::vector<size_t> > candidates;
-        MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis,FC>(*lo,hi_dofCoords,threshold,candidates);
-        MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs(candidates,hi_e2n,hi_colMap,lo_elemToHiRepresentativeNode);
-        MueLu::MueLuIntrepid::BuildLoElemToNodeViaRepresentatives(hi_e2n,hi_owned,lo_elemToHiRepresentativeNode,lo_e2n_mk2,lo_owned_mk2,hi_to_lo_map_mk2,lo_numOwnedNodes_mk2);
-
-        // Compare stuff
-        TEST_EQUALITY(lo_numOwnedNodes,2);
-        TEST_EQUALITY(lo_numOwnedNodes_mk2,2);
-
-        size_t num_lo_nodes_located=0;
-        for(size_t i=0;i<hi_to_lo_map.size(); i++) {
-          if(hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
-            num_lo_nodes_located++;
-        }
-        TEST_EQUALITY(lo_owned.size(),num_lo_nodes_located);
-        TEST_EQUALITY(lo_owned_mk2.size(),num_lo_nodes_located);
+    // QUAD
+    //  A one element test with Kirby-numbered nodes where the top edge is not owned
+    RCP<Basis> lo = rcp(new Intrepid2::Basis_HGRAD_QUAD_C1_FEM<ES, MT, MT>());
 
-	auto lo_e2n_host = Kokkos::create_mirror_view(lo_e2n);
-	auto lo_e2n_mk2_host = Kokkos::create_mirror_view(lo_e2n_mk2);
-	Kokkos::deep_copy(lo_e2n_host, lo_e2n);
-	Kokkos::deep_copy(lo_e2n_mk2_host, lo_e2n_mk2);
-        for(size_t i=0; i<lo_e2n.extent(0); i++)
-          for(size_t j=0; j<lo_e2n.extent(1); j++)
-            TEST_EQUALITY(lo_e2n_host(i,j),lo_e2n_mk2_host(i,j));
+    for (int degree = 2; degree < max_degree; degree++) {
+      int Nn        = (degree + 1) * (degree + 1);
+      RCP<Basis> hi = rcp(new Intrepid2::Basis_HGRAD_QUAD_Cn_FEM<ES, MT, MT>(degree, Intrepid2::POINTTYPE_EQUISPACED));
+      FCi hi_e2n("hi_e2n", 1, Nn), lo_e2n, lo_e2n_mk2;
 
-        for(size_t i=0; i<(size_t) lo_owned.size(); i++)
-          TEST_EQUALITY(lo_owned[i],lo_owned_mk2[i]);
+      std::vector<bool> hi_owned(Nn, false), lo_owned, lo_owned_mk2;
+      std::vector<size_t> lo_node_in_hi;
+      std::vector<LO> hi_to_lo_map, hi_to_lo_map_mk2;
+      int lo_numOwnedNodes = 0, lo_numOwnedNodes_mk2 = 0;
+      FC hi_dofCoords;
 
-        TEST_EQUALITY(hi_to_lo_map.size(),hi_to_lo_map_mk2.size());
-        for(size_t i=0; i<(size_t) hi_to_lo_map.size(); i++)
-          TEST_EQUALITY(hi_to_lo_map[i],hi_to_lo_map_mk2[i]);
+      // El2node / ownership / colmap
+      Kokkos::parallel_for(
+          "IntrepidPCoarsenFactory,BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1",
+          Kokkos::RangePolicy<typename Node::device_type::execution_space>(0, Nn), KOKKOS_LAMBDA(int i) {
+            hi_e2n(0, i) = i;
+          });
+      Kokkos::fence();
+      Teuchos::Array<GO> hi_colids(Nn);
+      for (int i = 0; i < Nn; i++) {
+        hi_colids[i] = i;
+        if (i < Nn - (degree + 1)) hi_owned[i] = true;
+      }
 
+      Teuchos::ArrayRCP<const int> is_dirichlet(Nn, 0);
+
+      /*** Do stuff the injection way ***/
+      MueLu::MueLuIntrepid::IntrepidGetP1NodeInHi<MT, typename Node::device_type>(hi, lo_node_in_hi, hi_dofCoords);
+      MueLu::MueLuIntrepid::BuildLoElemToNode(hi_e2n, hi_owned, lo_node_in_hi, is_dirichlet, lo_e2n, lo_owned, hi_to_lo_map, lo_numOwnedNodes);
+
+      /*** Do stuff the representative way ***/
+      RCP<const Map> hi_colMap = MapFactory::Build(lib, gst_invalid, hi_colids(), 0, comm);
+      FCi lo_elemToHiRepresentativeNode;
+      double threshold = 1e-10;
+      std::vector<std::vector<size_t>> candidates;
+      MueLu::MueLuIntrepid::GenerateRepresentativeBasisNodes<Basis, FC>(*lo, hi_dofCoords, threshold, candidates);
+      MueLu::MueLuIntrepid::GenerateLoNodeInHiViaGIDs(candidates, hi_e2n, hi_colMap, lo_elemToHiRepresentativeNode);
+      MueLu::MueLuIntrepid::BuildLoElemToNodeViaRepresentatives(hi_e2n, hi_owned, lo_elemToHiRepresentativeNode, lo_e2n_mk2, lo_owned_mk2, hi_to_lo_map_mk2, lo_numOwnedNodes_mk2);
+
+      // Compare stuff
+      TEST_EQUALITY(lo_numOwnedNodes, 2);
+      TEST_EQUALITY(lo_numOwnedNodes_mk2, 2);
+
+      size_t num_lo_nodes_located = 0;
+      for (size_t i = 0; i < hi_to_lo_map.size(); i++) {
+        if (hi_to_lo_map[i] != Teuchos::OrdinalTraits<LO>::invalid())
+          num_lo_nodes_located++;
       }
-    }//end QUAD
+      TEST_EQUALITY(lo_owned.size(), num_lo_nodes_located);
+      TEST_EQUALITY(lo_owned_mk2.size(), num_lo_nodes_located);
 
-  }
+      auto lo_e2n_host     = Kokkos::create_mirror_view(lo_e2n);
+      auto lo_e2n_mk2_host = Kokkos::create_mirror_view(lo_e2n_mk2);
+      Kokkos::deep_copy(lo_e2n_host, lo_e2n);
+      Kokkos::deep_copy(lo_e2n_mk2_host, lo_e2n_mk2);
+      for (size_t i = 0; i < lo_e2n.extent(0); i++)
+        for (size_t j = 0; j < lo_e2n.extent(1); j++)
+          TEST_EQUALITY(lo_e2n_host(i, j), lo_e2n_mk2_host(i, j));
 
+      for (size_t i = 0; i < (size_t)lo_owned.size(); i++)
+        TEST_EQUALITY(lo_owned[i], lo_owned_mk2[i]);
 
-/*********************************************************************************************************************/
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_LINE_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    int hi_degree=3;
-    int lo_degree=2;
-    // GOLD vector collection
-    // Note: Vectors are exodus-ordered, not Kirby-ordered
-    size_t num_p1_points = 10;
-    size_t num_hi_points = num_p1_points + (num_p1_points-1) *(hi_degree-1);
-    size_t num_lo_points = num_p1_points + (num_p1_points-1) *(lo_degree-1);
-
-    std::vector<Scalar> lo_gold_in(num_lo_points);
-    std::vector<Scalar> hi_gold_out(num_hi_points);
-    for(size_t i=0; i<num_p1_points; i++) {
-      lo_gold_in[i] = i;
-      hi_gold_out[i] = i;
+      TEST_EQUALITY(hi_to_lo_map.size(), hi_to_lo_map_mk2.size());
+      for (size_t i = 0; i < (size_t)hi_to_lo_map.size(); i++)
+        TEST_EQUALITY(hi_to_lo_map[i], hi_to_lo_map_mk2[i]);
     }
+  }  // end QUAD
+}
 
-    size_t idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)hi_degree-1; j++) {
-        hi_gold_out[idx] = i + ((double)j+1)/hi_degree;
-        idx++;
-      }
-    }
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  int hi_degree = 3;
+  int lo_degree = 2;
+  // GOLD vector collection
+  // Note: Vectors are exodus-ordered, not Kirby-ordered
+  size_t num_p1_points = 10;
+  size_t num_hi_points = num_p1_points + (num_p1_points - 1) * (hi_degree - 1);
+  size_t num_lo_points = num_p1_points + (num_p1_points - 1) * (lo_degree - 1);
+
+  std::vector<Scalar> lo_gold_in(num_lo_points);
+  std::vector<Scalar> hi_gold_out(num_hi_points);
+  for (size_t i = 0; i < num_p1_points; i++) {
+    lo_gold_in[i]  = i;
+    hi_gold_out[i] = i;
+  }
 
-    idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)lo_degree-1; j++) {
-        lo_gold_in[idx] = i + ((double)j+1)/lo_degree;
-        idx++;
-      }
+  size_t idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)hi_degree - 1; j++) {
+      hi_gold_out[idx] = i + ((double)j + 1) / hi_degree;
+      idx++;
     }
+  }
 
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,num_p1_points,hi_degree,lo_gold_in,hi_gold_out,std::string("hgrad_line_c3"),std::string("hgrad_line_c2"));
+  idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)lo_degree - 1; j++) {
+      lo_gold_in[idx] = i + ((double)j + 1) / lo_degree;
+      idx++;
+    }
   }
 
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, num_p1_points, hi_degree, lo_gold_in, hi_gold_out, std::string("hgrad_line_c3"), std::string("hgrad_line_c2"));
+}
+
 /*********************************************************************************************************************/
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_LINE_p4_to_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    int hi_degree=4;
-    int lo_degree=3;
-    // GOLD vector collection
-    // Note: Vectors are exodus-ordered, not Kirby-ordered
-    size_t num_p1_points = 10;
-    size_t num_hi_points = num_p1_points + (num_p1_points-1) *(hi_degree-1);
-    size_t num_lo_points = num_p1_points + (num_p1_points-1) *(lo_degree-1);
-
-    std::vector<Scalar> lo_gold_in(num_lo_points);
-    std::vector<Scalar> hi_gold_out(num_hi_points);
-    for(size_t i=0; i<num_p1_points; i++) {
-      lo_gold_in[i] = i;
-      hi_gold_out[i] = i;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  int hi_degree = 4;
+  int lo_degree = 3;
+  // GOLD vector collection
+  // Note: Vectors are exodus-ordered, not Kirby-ordered
+  size_t num_p1_points = 10;
+  size_t num_hi_points = num_p1_points + (num_p1_points - 1) * (hi_degree - 1);
+  size_t num_lo_points = num_p1_points + (num_p1_points - 1) * (lo_degree - 1);
+
+  std::vector<Scalar> lo_gold_in(num_lo_points);
+  std::vector<Scalar> hi_gold_out(num_hi_points);
+  for (size_t i = 0; i < num_p1_points; i++) {
+    lo_gold_in[i]  = i;
+    hi_gold_out[i] = i;
+  }
 
-    size_t idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)hi_degree-1; j++) {
-        hi_gold_out[idx] = i + ((double)j+1)/hi_degree;
-        idx++;
-      }
+  size_t idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)hi_degree - 1; j++) {
+      hi_gold_out[idx] = i + ((double)j + 1) / hi_degree;
+      idx++;
     }
+  }
 
-    idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)lo_degree-1; j++) {
-        lo_gold_in[idx] = i + ((double)j+1)/lo_degree;
-        idx++;
-      }
+  idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)lo_degree - 1; j++) {
+      lo_gold_in[idx] = i + ((double)j + 1) / lo_degree;
+      idx++;
     }
-
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,num_p1_points,hi_degree,lo_gold_in,hi_gold_out,std::string("hgrad_line_c4"),std::string("hgrad_line_c3"));
   }
 
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, num_p1_points, hi_degree, lo_gold_in, hi_gold_out, std::string("hgrad_line_c4"), std::string("hgrad_line_c3"));
+}
+
 /*********************************************************************************************************************/
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_LINE_p4_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    int hi_degree=4;
-    int lo_degree=2;
-    // GOLD vector collection
-    // Note: Vectors are exodus-ordered, not Kirby-ordered
-    size_t num_p1_points = 10;
-    size_t num_hi_points = num_p1_points + (num_p1_points-1) *(hi_degree-1);
-    size_t num_lo_points = num_p1_points + (num_p1_points-1) *(lo_degree-1);
-
-    std::vector<Scalar> lo_gold_in(num_lo_points);
-    std::vector<Scalar> hi_gold_out(num_hi_points);
-    for(size_t i=0; i<num_p1_points; i++) {
-      lo_gold_in[i] = i;
-      hi_gold_out[i] = i;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  int hi_degree = 4;
+  int lo_degree = 2;
+  // GOLD vector collection
+  // Note: Vectors are exodus-ordered, not Kirby-ordered
+  size_t num_p1_points = 10;
+  size_t num_hi_points = num_p1_points + (num_p1_points - 1) * (hi_degree - 1);
+  size_t num_lo_points = num_p1_points + (num_p1_points - 1) * (lo_degree - 1);
+
+  std::vector<Scalar> lo_gold_in(num_lo_points);
+  std::vector<Scalar> hi_gold_out(num_hi_points);
+  for (size_t i = 0; i < num_p1_points; i++) {
+    lo_gold_in[i]  = i;
+    hi_gold_out[i] = i;
+  }
 
-    size_t idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)hi_degree-1; j++) {
-        hi_gold_out[idx] = i + ((double)j+1)/hi_degree;
-        idx++;
-      }
+  size_t idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)hi_degree - 1; j++) {
+      hi_gold_out[idx] = i + ((double)j + 1) / hi_degree;
+      idx++;
     }
+  }
 
-    idx=num_p1_points;
-    for(size_t i=0; i<num_p1_points-1; i++) {
-      for(size_t j=0; j<(size_t)lo_degree-1; j++) {
-        lo_gold_in[idx] = i + ((double)j+1)/lo_degree;
-        idx++;
-      }
+  idx = num_p1_points;
+  for (size_t i = 0; i < num_p1_points - 1; i++) {
+    for (size_t j = 0; j < (size_t)lo_degree - 1; j++) {
+      lo_gold_in[idx] = i + ((double)j + 1) / lo_degree;
+      idx++;
     }
-
-    TestPseudoPoisson<Scalar,LocalOrdinal,GlobalOrdinal,Node>(out,num_p1_points,hi_degree,lo_gold_in,hi_gold_out,std::string("hgrad_line_c4"),std::string("hgrad_line_c2"));
   }
 
+  TestPseudoPoisson<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, num_p1_points, hi_degree, lo_gold_in, hi_gold_out, std::string("hgrad_line_c4"), std::string("hgrad_line_c2"));
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=3;
-    std::string hi_basis("hgrad_line_c3");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c2");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 3;
+  std::string hi_basis("hgrad_line_c3");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c2");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=4;
-    std::string hi_basis("hgrad_line_c4");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c3");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 4;
+  std::string hi_basis("hgrad_line_c4");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c3");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=4;
-    std::string hi_basis("hgrad_line_c4");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    //    Params.set("rap: fix zero diagonals",true);
-    Params.set("pcoarsen: hi basis",hi_basis);
-    Params.set("pcoarsen: lo basis","hgrad_line_c2");
-    Params.set("verbosity","high");
-    Params.set("max levels",2);
-    Params.set("coarse: max size",100);
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 4;
+  std::string hi_basis("hgrad_line_c4");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  //    Params.set("rap: fix zero diagonals",true);
+  Params.set("pcoarsen: hi basis", hi_basis);
+  Params.set("pcoarsen: lo basis", "hgrad_line_c2");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 2);
+  Params.set("coarse: max size", 100);
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=4;
-    std::string hi_basis("hgrad_line_c4");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0, level1, level2, mmm;
-    Params.set("multigrid algorithm","pcoarsen");
-    Params.set("verbosity","high");
-    Params.set("max levels",3);
-    Params.set("coarse: max size",5);
-
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-
-    level1.set("pcoarsen: hi basis",hi_basis);
-    level1.set("pcoarsen: lo basis","hgrad_line_c3");
-    Params.set("level 1",level1);
-
-    level2.set("pcoarsen: hi basis","hgrad_line_c3");
-    level2.set("pcoarsen: lo basis","hgrad_line_c2");
-    Params.set("level 2",level2);
-    Params.set("verbosity","none");
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3_to_p2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 4;
+  std::string hi_basis("hgrad_line_c4");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0, level1, level2, mmm;
+  Params.set("multigrid algorithm", "pcoarsen");
+  Params.set("verbosity", "high");
+  Params.set("max levels", 3);
+  Params.set("coarse: max size", 5);
+
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+
+  level1.set("pcoarsen: hi basis", hi_basis);
+  level1.set("pcoarsen: lo basis", "hgrad_line_c3");
+  Params.set("level 1", level1);
+
+  level2.set("pcoarsen: hi basis", "hgrad_line_c3");
+  level2.set("pcoarsen: lo basis", "hgrad_line_c2");
+  Params.set("level 2", level2);
+  Params.set("verbosity", "none");
 
 #if 0
     // DEBUG
@@ -2865,270 +2621,261 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_L
     Params.set("export data",dump);
 #endif
 
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p2_to_p1_sa, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 2;
+  std::string hi_basis("hgrad_line_c2");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0, level1, level2;
+  Params.set("multigrid algorithm", "pcoarsen");
+  Params.set("rap: fix zero diagonals", true);
+  Params.set("verbosity", "high");
+  Params.set("max levels", 3);
+  Params.set("coarse: max size", 100);
+
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+
+  level1.set("multigrid algorithm", "pcoarsen");
+  level1.set("pcoarsen: hi basis", hi_basis);
+  level1.set("pcoarsen: lo basis", "hgrad_line_c1");
+  Params.set("level 1", level1);
+
+  level2.set("multigrid algorithm", "sa");
+  Params.set("level 2", level2);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p2_to_p1_sa, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=2;
-    std::string hi_basis("hgrad_line_c2");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0, level1, level2;
-    Params.set("multigrid algorithm","pcoarsen");
-    Params.set("rap: fix zero diagonals",true);
-    Params.set("verbosity","high");
-    Params.set("max levels",3);
-    Params.set("coarse: max size",100);
-
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-
-    level1.set("multigrid algorithm","pcoarsen");
-    level1.set("pcoarsen: hi basis",hi_basis);
-    level1.set("pcoarsen: lo basis","hgrad_line_c1");
-    Params.set("level 1",level1);
-
-    level2.set("multigrid algorithm","sa");
-    Params.set("level 2",level2);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_manual, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  ;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 3;
+  std::string hi_basis("hgrad_line_c3");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0, level1, level2, level3;
+  Params.set("multigrid algorithm", "pcoarsen");
+  Params.set("rap: fix zero diagonals", true);
+  Params.set("verbosity", "high");
+  Params.set("max levels", 4);
+  Params.set("coarse: max size", 100);
+
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+
+  level1.set("pcoarsen: hi basis", "hgrad_line_c3");
+  level1.set("pcoarsen: lo basis", "hgrad_line_c2");
+  Params.set("level 1", level1);
+
+  level2.set("pcoarsen: hi basis", "hgrad_line_c2");
+  level2.set("pcoarsen: lo basis", "hgrad_line_c1");
+  Params.set("level 2", level2);
+
+  level3.set("multigrid algorithm", "sa");
+  Params.set("level 3", level3);
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_manual, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    ;
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=3;
-    std::string hi_basis("hgrad_line_c3");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0, level1, level2, level3;
-    Params.set("multigrid algorithm","pcoarsen");
-    Params.set("rap: fix zero diagonals",true);
-    Params.set("verbosity","high");
-    Params.set("max levels",4);
-    Params.set("coarse: max size",100);
-
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-
-    level1.set("pcoarsen: hi basis","hgrad_line_c3");
-    level1.set("pcoarsen: lo basis","hgrad_line_c2");
-    Params.set("level 1",level1);
-
-    level2.set("pcoarsen: hi basis","hgrad_line_c2");
-    level2.set("pcoarsen: lo basis","hgrad_line_c1");
-    Params.set("level 2",level2);
-
-    level3.set("multigrid algorithm","sa");
-    Params.set("level 3",level3);
-    Params.set("verbosity","none");
-
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_schedule, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Tpetra::KokkosCompat::KokkosSerialWrapperNode);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
 
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+  typedef Kokkos::DynRankView<LocalOrdinal, typename Node::device_type> FCi;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  using Teuchos::RCP;
+  int degree = 3;
+  std::string hi_basis("hgrad_line_c3");
+
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO num_nodes = 972;
+  // Build a pseudo-poisson test matrix
+  FCi elem_to_node;
+  RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC, LO, GO, NO>(num_nodes, degree, elem_to_node, lib);
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<MT> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(A->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  // ParameterList
+  ParameterList Params, level0;
+  Params.set("multigrid algorithm", "pcoarsen");
+  Params.set("rap: fix zero diagonals", true);
+  Params.set("verbosity", "high");
+  Params.set("max levels", 5);
+  Params.set("coarse: max size", 10);
+
+  level0.set("pcoarsen: element to node map", rcp(&elem_to_node, false));
+  Params.set("level 0", level0);
+
+  Params.set("pcoarsen: element", "hgrad_line_c");
+  Params.set("pcoarsen: schedule", "{3,2,1}");
+  Params.set("verbosity", "none");
+
+  // Build hierarchy
+  RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, Params);
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_schedule, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Tpetra::KokkosCompat::KokkosSerialWrapperNode);
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-    typedef Kokkos::DynRankView<LocalOrdinal,typename Node::device_type> FCi;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    using Teuchos::RCP;
-    int degree=3;
-    std::string hi_basis("hgrad_line_c3");
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO num_nodes = 972;
-    // Build a pseudo-poisson test matrix
-    FCi elem_to_node;
-    RCP<Matrix> A = TestHelpers::Build1DPseudoPoissonHigherOrder<SC,LO,GO,NO>(num_nodes,degree,elem_to_node,lib);
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(A->getRowMap(), 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<MT> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(A->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    // ParameterList
-    ParameterList Params, level0;
-    Params.set("multigrid algorithm","pcoarsen");
-    Params.set("rap: fix zero diagonals",true);
-    Params.set("verbosity","high");
-    Params.set("max levels",5);
-    Params.set("coarse: max size",10);
-
-    level0.set("pcoarsen: element to node map",rcp(&elem_to_node,false));
-    Params.set("level 0",level0);
-
-    Params.set("pcoarsen: element","hgrad_line_c");
-    Params.set("pcoarsen: schedule","{3,2,1}");
-    Params.set("verbosity","none");
-
-    // Build hierarchy
-    RCP<Hierarchy> tH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,Params);
-  }
 
-  /*********************************************************************************************************************/
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,GetP1NodeInHi,Scalar,LO,GO,Node)  \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BasisFactory,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BuildLoElemToNode,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BuildLoElemToNodeWithDirichlet,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,GenerateColMapFromImport,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p2,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p3,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory,BuildP_PseudoPoisson_p4,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p2, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_LINE_Equispaced,Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Equispaced,Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Spectral,  Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Equispaced, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Spectral,   Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p3_to_p2,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p3,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p2,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p2, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3_to_p2, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p2_to_p1_sa, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_manual, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_schedule, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildSampleElementToNodeMapThreeElementQuad, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, UniqueNumbering, Scalar, LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Line, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Line, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Quad, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Quad, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Hex, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Hex, Scalar, LO, GO, Node) \
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                                            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GetP1NodeInHi, Scalar, LO, GO, Node)                                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BasisFactory, Scalar, LO, GO, Node)                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildLoElemToNode, Scalar, LO, GO, Node)                                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildLoElemToNodeWithDirichlet, Scalar, LO, GO, Node)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateColMapFromImport, Scalar, LO, GO, Node)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p2, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p3, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_p4, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p2, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_LINE_Equispaced, Scalar, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Equispaced, Scalar, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_QUAD_Spectral, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Equispaced, Scalar, LO, GO, Node)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateRepresentativeBasisNodes_HEX_Spectral, Scalar, LO, GO, Node)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, GenerateLoNodeInHighViaGIDs_QUAD_pn_to_p1, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildLoElemToNodeViaRepresentatives_QUAD_pn_to_p1, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p3_to_p2, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p3, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildP_PseudoPoisson_LINE_p4_to_p2, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2, Scalar, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3, Scalar, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p2, Scalar, LO, GO, Node)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p4_to_p3_to_p2, Scalar, LO, GO, Node)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p2_to_p1_sa, Scalar, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_manual, Scalar, LO, GO, Node)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, CreatePreconditioner_p3_to_p2_to_p1_sa_schedule, Scalar, LO, GO, Node)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, BuildSampleElementToNodeMapThreeElementQuad, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, UniqueNumbering, Scalar, LO, GO, Node)                                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Line, Scalar, LO, GO, Node)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Line, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Quad, Scalar, LO, GO, Node)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Quad, Scalar, LO, GO, Node)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Equispaced_Hex, Scalar, LO, GO, Node)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IntrepidPCoarsenFactory, FindSeeds_Spectral_Hex, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
 #endif
diff --git a/packages/muelu/test/unit_tests/InverseApproximationFactory.cpp b/packages/muelu/test/unit_tests/InverseApproximationFactory.cpp
index 2ce430914eb3..4a37c6ec7f22 100644
--- a/packages/muelu/test/unit_tests/InverseApproximationFactory.cpp
+++ b/packages/muelu/test/unit_tests/InverseApproximationFactory.cpp
@@ -63,309 +63,314 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseDiagonalConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseDiagonalConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    {
-      const int n = 20;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build1DPoisson(n, lib);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("diagonal")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-
-      const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
-      Ainv->getLocalDiagCopy(*AinvDiagonal);
-      Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
-      bool bCheck = true;
-      for(int i=0; i<int(n/comm->getSize()); i++) if(AinvData[i] != Teuchos::as<Scalar>(0.5)) bCheck = false;
-      TEST_EQUALITY(bCheck, true);
-    }
+    const int n            = 20;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(n, lib);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("diagonal")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+
+    const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
+    Ainv->getLocalDiagCopy(*AinvDiagonal);
+    Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
+    bool bCheck                              = true;
+    for (int i = 0; i < int(n / comm->getSize()); i++)
+      if (AinvData[i] != Teuchos::as<Scalar>(0.5)) bCheck = false;
+    TEST_EQUALITY(bCheck, true);
+  }
 
-    {
-      const int n = 42;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build2DPoisson(n, n, lib);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("diagonal")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-
-      const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
-      Ainv->getLocalDiagCopy(*AinvDiagonal);
-      Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
-      bool bCheck = true;
-      for(int i=0; i<int(n*n/comm->getSize()); i++) if(AinvData[i] != Teuchos::as<Scalar>(0.25)) bCheck = false;
-      TEST_EQUALITY(bCheck, true);
-    }
+  {
+    const int n            = 42;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(n, n, lib);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("diagonal")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+
+    const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
+    Ainv->getLocalDiagCopy(*AinvDiagonal);
+    Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
+    bool bCheck                              = true;
+    for (int i = 0; i < int(n * n / comm->getSize()); i++)
+      if (AinvData[i] != Teuchos::as<Scalar>(0.25)) bCheck = false;
+    TEST_EQUALITY(bCheck, true);
+  }
+
+}  // InverseDiagonalConstructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseLumpingConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-  } //InverseDiagonalConstructor
+  {
+    const int n            = 20;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(n, lib);
+    A->scale(0.5);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+
+    const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
+    Ainv->getLocalDiagCopy(*AinvDiagonal);
+    Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
+    bool bCheck                              = false;
+    for (int i = 0; i < int(n / comm->getSize()); i++)
+      if (std::abs(AinvData[i] - Teuchos::as<Scalar>(0.66666666667)) < std::abs(Teuchos::as<Scalar>(1e-8)) ||
+          std::abs(AinvData[i] - Teuchos::as<Scalar>(0.5)) < std::abs(Teuchos::as<Scalar>(1e-8)))
+        bCheck = true;
+    TEST_EQUALITY(bCheck, true);
+  }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseLumpingConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    {
-      const int n = 20;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build1DPoisson(n, lib);
-      A->scale(0.5);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-
-      const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
-      Ainv->getLocalDiagCopy(*AinvDiagonal);
-      Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
-      bool bCheck = false;
-      for(int i=0; i<int(n/comm->getSize()); i++)
-        if(std::abs(AinvData[i] - Teuchos::as<Scalar>(0.66666666667)) < std::abs(Teuchos::as<Scalar>(1e-8)) ||
-           std::abs(AinvData[i] - Teuchos::as<Scalar>(0.5)) < std::abs(Teuchos::as<Scalar>(1e-8)))
-          bCheck = true;
-      TEST_EQUALITY(bCheck, true);
-    }
+    const int n            = 42;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(n, n, lib);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+
+    const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
+    Ainv->getLocalDiagCopy(*AinvDiagonal);
+    Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
+    bool bCheck                              = false;
+    for (int i = 0; i < int(n * n / comm->getSize()); i++)
+      if (std::abs(AinvData[i] - Teuchos::as<Scalar>(0.166667)) < std::abs(Teuchos::as<Scalar>(1e-6)) ||
+          std::abs(AinvData[i] - Teuchos::as<Scalar>(0.142857)) < std::abs(Teuchos::as<Scalar>(1e-6)))
+        bCheck = true;
+    TEST_EQUALITY(bCheck, true);
+  }
+
+}  // InverseLumpingConstructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseSpaiConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST            = Teuchos::ScalarTraits<SC>;
+  using magnitude_type = typename TST::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-    {
-      const int n = 42;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build2DPoisson(n, n, lib);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("lumping")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-
-      const RCP<Vector> AinvDiagonal = VectorFactory::Build(Ainv->getRangeMap(), true);
-      Ainv->getLocalDiagCopy(*AinvDiagonal);
-      Teuchos::ArrayRCP<const Scalar> AinvData = AinvDiagonal->getData(0);
-      bool bCheck = false;
-      for(int i=0; i<int(n*n/comm->getSize()); i++)
-        if(std::abs(AinvData[i] - Teuchos::as<Scalar>(0.166667)) < std::abs(Teuchos::as<Scalar>(1e-6)) ||
-           std::abs(AinvData[i] - Teuchos::as<Scalar>(0.142857)) < std::abs(Teuchos::as<Scalar>(1e-6)))
-          bCheck = true;
-      TEST_EQUALITY(bCheck, true);
-    }
+  {
+    const int n            = 20;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(n, lib);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-  } //InverseLumpingConstructor
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n / comm->getSize()) + int(n / comm->getSize() - 1));
+    TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 3.037251711528645, 1e2 * TMT::eps());
+  }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(InverseApproximationFactory, InverseSpaiConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST = Teuchos::ScalarTraits<SC>;
-    using magnitude_type = typename TST::magnitudeType;
-    using TMT = Teuchos::ScalarTraits<magnitude_type>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    {
-      const int n = 20;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build1DPoisson(n, lib);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n/comm->getSize()) + int(n/comm->getSize()-1));
-      TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 3.037251711528645, 1e2*TMT::eps());
-    }
+    const int n            = 42;
+    Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(n, n, lib);
 
-    {
-      const int n = 42;
-      Teuchos::RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build2DPoisson(n, n, lib);
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
 
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
 
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
+    // generate Schur complement operator
+    invapproxFact->Build(level);
 
-      // generate Schur complement operator
-      invapproxFact->Build(level);
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()));
+    TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n * n / comm->getSize()) + int(n * n / comm->getSize() - 1));
+    TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 12.41994595675205, 1e3 * TMT::eps());
+  }
 
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getRangeMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMinGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()));
-      TEST_EQUALITY(Ainv->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * int(n*n/comm->getSize()) + int(n*n/comm->getSize()-1));
-      TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 12.41994595675205, 1e3*TMT::eps());
-    }
+  // test with a highly nonsymmetric matrix
+  {
+    using STS = Teuchos::ScalarTraits<SC>;
 
-    // test with a highly nonsymmetric matrix
-    {
-      using STS = Teuchos::ScalarTraits<SC>;
-
-      // Don't test for complex - matrix reader won't work
-      if (STS::isComplex) {success=true; return;}
-      RCP<Matrix> A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("TestMatrices/nonsym.mm", lib, comm);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
-
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
-
-      // generate Schur complement operator
-      invapproxFact->Build(level);
-
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 0.1235706050986417, 1e2*TMT::eps());
-      // check values of first row only on root
-      if(comm->getRank() == 0) {
-        ArrayView<const LocalOrdinal> indices;
-        ArrayView<const Scalar> values;
-        Ainv->getLocalRowView(0, indices, values);
-        TEST_FLOATING_EQUALITY(values[0], Teuchos::as<Scalar>(1.0000000000000002e-01), 1e2*TMT::eps());
-        TEST_FLOATING_EQUALITY(values[1], Teuchos::as<Scalar>(-1.6666666666666673e-02), 1e2*TMT::eps());
-        TEST_FLOATING_EQUALITY(values[2], Teuchos::as<Scalar>(4.6666666666666688e-03), 1e2*TMT::eps());
-      }
+    // Don't test for complex - matrix reader won't work
+    if (STS::isComplex) {
+      success = true;
+      return;
+    }
+    RCP<Matrix> A = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("TestMatrices/nonsym.mm", lib, comm);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
+
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
+
+    // generate Schur complement operator
+    invapproxFact->Build(level);
+
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 0.1235706050986417, 1e2 * TMT::eps());
+    // check values of first row only on root
+    if (comm->getRank() == 0) {
+      ArrayView<const LocalOrdinal> indices;
+      ArrayView<const Scalar> values;
+      Ainv->getLocalRowView(0, indices, values);
+      TEST_FLOATING_EQUALITY(values[0], Teuchos::as<Scalar>(1.0000000000000002e-01), 1e2 * TMT::eps());
+      TEST_FLOATING_EQUALITY(values[1], Teuchos::as<Scalar>(-1.6666666666666673e-02), 1e2 * TMT::eps());
+      TEST_FLOATING_EQUALITY(values[2], Teuchos::as<Scalar>(4.6666666666666688e-03), 1e2 * TMT::eps());
     }
+  }
 
-    // Test pre and post filtering of approximate inverse
-    {
-      using STS = Teuchos::ScalarTraits<SC>;
+  // Test pre and post filtering of approximate inverse
+  {
+    using STS = Teuchos::ScalarTraits<SC>;
 
-      // Don't test for complex - matrix reader won't work
-      if (STS::isComplex) {success=true; return;}
-      RCP<Matrix> A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("TestMatrices/beam.mm", lib, comm);
+    // Don't test for complex - matrix reader won't work
+    if (STS::isComplex) {
+      success = true;
+      return;
+    }
+    RCP<Matrix> A = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("TestMatrices/beam.mm", lib, comm);
 
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-      RCP<InverseApproximationFactory> invapproxFact = rcp( new InverseApproximationFactory() );
-      invapproxFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      invapproxFact->SetParameter("inverse: drop tolerance", Teuchos::ParameterEntry(Scalar(1e-8)));
-      invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
+    RCP<InverseApproximationFactory> invapproxFact = rcp(new InverseApproximationFactory());
+    invapproxFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    invapproxFact->SetParameter("inverse: drop tolerance", Teuchos::ParameterEntry(Scalar(1e-8)));
+    invapproxFact->SetParameter("inverse: approximation type", Teuchos::ParameterEntry(std::string("sparseapproxinverse")));
 
-      // request InverseApproximation operator
-      level.Request("Ainv", invapproxFact.get());
+    // request InverseApproximation operator
+    level.Request("Ainv", invapproxFact.get());
 
-      // generate Schur complement operator
-      invapproxFact->Build(level);
+    // generate Schur complement operator
+    invapproxFact->Build(level);
 
-      RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
-      TEST_EQUALITY(Ainv.is_null(), false);
-      TEST_EQUALITY(Ainv->getGlobalNumEntries(), 115760);
-      // 8.31688788510637e+06
-      TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 8.31688788510637e+06, 1e5*TMT::eps());
-    }
+    RCP<Matrix> Ainv = level.Get<RCP<Matrix> >("Ainv", invapproxFact.get());
+    TEST_EQUALITY(Ainv.is_null(), false);
+    TEST_EQUALITY(Ainv->getGlobalNumEntries(), 115760);
+    // 8.31688788510637e+06
+    TEST_FLOATING_EQUALITY(Ainv->getFrobeniusNorm(), 8.31688788510637e+06, 1e5 * TMT::eps());
+  }
 
-  } //InverseSpaiConstructor
+}  // InverseSpaiConstructor
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseDiagonalConstructor, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseLumpingConstructor, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseSpaiConstructor, SC, LO, GO, Node) \
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseDiagonalConstructor, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseLumpingConstructor, SC, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(InverseApproximationFactory, InverseSpaiConstructor, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/LWGraph.cpp b/packages/muelu/test/unit_tests/LWGraph.cpp
index 9148b41097b4..95f3d2a6ce22 100644
--- a/packages/muelu/test/unit_tests/LWGraph.cpp
+++ b/packages/muelu/test/unit_tests/LWGraph.cpp
@@ -58,83 +58,81 @@
 
 namespace MueLuTests {
 
-  // Little utility to generate a LWGraph.
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node> >
-  gimmeLWGraph(const Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A) {
-#   include "MueLu_UseShortNames.hpp"
-
-    Level level;
-
-    TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
-    dropFact->SetVerbLevel(MueLu::Extreme);
-    dropFact->SetPreDropFunction(rcp(new PreDropFunctionConstVal(0.00001)));
-
-    level.Request("Graph",              dropFact.get());
-    dropFact->Build(level);
-
-    auto graph = level.Get<RCP<GraphBase> >("Graph", dropFact.get());
-    level.Release("Graph",              dropFact.get());
-    return graph;
+// Little utility to generate a LWGraph.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::GraphBase<LocalOrdinal, GlobalOrdinal, Node> >
+gimmeLWGraph(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A) {
+#include "MueLu_UseShortNames.hpp"
+
+  Level level;
+
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+  dropFact->SetVerbLevel(MueLu::Extreme);
+  dropFact->SetPreDropFunction(rcp(new PreDropFunctionConstVal(0.00001)));
+
+  level.Request("Graph", dropFact.get());
+  dropFact->Build(level);
+
+  auto graph = level.Get<RCP<GraphBase> >("Graph", dropFact.get());
+  level.Release("Graph", dropFact.get());
+  return graph;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph, CreateLWGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph, CreateLWGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
 
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  RCP<GraphBase> graph = gimmeLWGraph(A);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
+  auto comm          = graph->GetComm();
+  const int numRanks = comm->getSize();
 
-    RCP<GraphBase> graph = gimmeLWGraph(A);
+  graph->print(out, MueLu::MsgType::Extreme);
+  //    TEST_EQUALITY( graph->description() == "LWGraph (graph of A)" ,  true);
+  TEST_EQUALITY(graph->description() == "MueLu.description()", true);  // To fix when LWGRaph description method will be fixed
+  TEST_EQUALITY(graph != Teuchos::null, true);
+  auto graphLWK = dynamic_cast<LWGraph*>(graph.get());
+  auto rows     = graphLWK->getRowPtrs();
 
-    auto comm = graph->GetComm();
-    const int numRanks = comm->getSize();
+  if (numRanks == 1) {
+    TEST_EQUALITY(graph->getLocalMaxNumRowEntries() == 3, true);
+    TEST_EQUALITY(graph->GetNodeNumVertices() == 16, true);
+    TEST_EQUALITY(graph->GetNodeNumEdges() == 46, true);
+    TEST_EQUALITY(rows.size() == 17, true);
 
-    graph->print(out, MueLu::MsgType::Extreme);
-    //    TEST_EQUALITY( graph->description() == "LWGraph (graph of A)" ,  true);
-    TEST_EQUALITY( graph->description() == "MueLu.description()",  true); // To fix when LWGRaph description method will be fixed
-    TEST_EQUALITY(graph != Teuchos::null,                true);
-    auto graphLWK = dynamic_cast<LWGraph*>(graph.get());
-    auto rows = graphLWK->getRowPtrs();
-
-    if(numRanks == 1) {
-      TEST_EQUALITY(graph->getLocalMaxNumRowEntries() == 3,   true);
-      TEST_EQUALITY(graph->GetNodeNumVertices() == 16,   true);
-      TEST_EQUALITY(graph->GetNodeNumEdges() == 46,   true);
-      TEST_EQUALITY(rows.size() == 17,   true);
-
-    } else if(numRanks == 4) {
-      TEST_EQUALITY(graph->GetNodeNumVertices() == 4,    true);
-    }
-
-    auto columns = graphLWK->getEntries();
-    for(size_t i = 0; i < graph->GetNodeNumVertices(); ++i) {
-        TEST_EQUALITY(graph->getNeighborVertices(i).size() == columns.view(rows[i], rows[i+1]-rows[i]).size(),   true);
-    }
+  } else if (numRanks == 4) {
+    TEST_EQUALITY(graph->GetNodeNumVertices() == 4, true);
+  }
 
-    auto crsGraph = graphLWK->GetCrsGraph();
-    TEST_EQUALITY(crsGraph->getLocalNumEntries() == graph->GetNodeNumEdges(),    true);
-    TEST_EQUALITY(crsGraph->getLocalNumRows() == graph->GetNodeNumVertices(),    true);
-    TEST_EQUALITY(crsGraph->getLocalMaxNumRowEntries() == graph->getLocalMaxNumRowEntries(),    true);
+  auto columns = graphLWK->getEntries();
+  for (size_t i = 0; i < graph->GetNodeNumVertices(); ++i) {
+    TEST_EQUALITY(graph->getNeighborVertices(i).size() == columns.view(rows[i], rows[i + 1] - rows[i]).size(), true);
+  }
 
-    TEST_THROW(graphLWK->SetBoundaryNodeMap(graphLWK->GetDomainMap()), MueLu::Exceptions::NotImplemented);
+  auto crsGraph = graphLWK->GetCrsGraph();
+  TEST_EQUALITY(crsGraph->getLocalNumEntries() == graph->GetNodeNumEdges(), true);
+  TEST_EQUALITY(crsGraph->getLocalNumRows() == graph->GetNodeNumVertices(), true);
+  TEST_EQUALITY(crsGraph->getLocalMaxNumRowEntries() == graph->getLocalMaxNumRowEntries(), true);
 
-  } // CreateLWGraph
+  TEST_THROW(graphLWK->SetBoundaryNodeMap(graphLWK->GetDomainMap()), MueLu::Exceptions::NotImplemented);
 
+}  // CreateLWGraph
 
 #define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LWGraph, CreateLWGraph,  SC, LO, GO, NO)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LWGraph, CreateLWGraph, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Level.cpp b/packages/muelu/test/unit_tests/Level.cpp
index 95b55e99fd34..59b8a3631257 100644
--- a/packages/muelu/test/unit_tests/Level.cpp
+++ b/packages/muelu/test/unit_tests/Level.cpp
@@ -65,395 +65,381 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, SetCoreData, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, SetCoreData, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level aLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);  // can be an empty operator
+
+  aLevel.Set("Hitchhiker's Guide", 42);
+  int fff = aLevel.Get<int>("Hitchhiker's Guide");
+  TEST_EQUALITY(fff, 42);
 
-    Level aLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Set("PI", 3.14159265);
+  double ggg = aLevel.Get<double>("PI");
+  TEST_EQUALITY(ggg, 3.14159265);
+  TEST_EQUALITY(aLevel.IsAvailable("PI"), true);
+
+  aLevel.Delete("PI", MueLu::NoFactory::get());
+  TEST_EQUALITY(aLevel.IsAvailable("PI"), false);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2); //can be an empty operator
+  aLevel.Set("Hello MueLu", std::string("Greetings to MueMat"));
+  std::string hhh = aLevel.Get<std::string>("Hello MueLu");
+  TEST_EQUALITY(hhh, "Greetings to MueMat");
+
+  aLevel.Set("A", A);
+  RCP<Matrix> newA = aLevel.Get<RCP<Matrix> >("A");
+  TEST_EQUALITY(newA, A);
+
+  aLevel.Set("R", A);
+  RCP<Matrix> newR = aLevel.Get<RCP<Matrix> >("R");
+  TEST_EQUALITY(newR, A);  // TODO from JG: must be tested using another matrix !
+
+  aLevel.Set("P", A);
+  RCP<Matrix> newP = aLevel.Get<RCP<Matrix> >("P");
+  TEST_EQUALITY(newP, A);
+
+  aLevel.SetLevelID(42);
+  TEST_EQUALITY(aLevel.GetLevelID(), 42);  // TODO: test default value of LevelID
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, NumRequests, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level aLevel;
+  aLevel.SetLevelID(0);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
+  aLevel.Set("A", A);
+
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  aLevel.SetFactoryManager(facManager);
+  RCP<FactoryBase> factory = rcp(new CoalesceDropFactory());
+
+  aLevel.Request("Graph", factory.get());
+  aLevel.Request("Graph", factory.get());
+
+  aLevel.Release("Graph", factory.get());
+  TEST_EQUALITY(aLevel.IsRequested("Graph", factory.get()), true);
+
+  aLevel.Release("Graph", factory.get());
+  TEST_EQUALITY(aLevel.IsRequested("Graph", factory.get()), false);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, RequestRelease, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  Level l;
+  l.SetLevelID(0);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
+  l.Set("A", A);
+
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  l.SetFactoryManager(facManager);
+
+  RCP<FactoryBase> factory = rcp(new CoalesceDropFactory());
+
+  l.Request("Graph", factory.get());
+  TEST_EQUALITY(l.IsRequested("Graph", factory.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Graph", factory.get()), false);
+  l.Release("Graph", factory.get());
+  TEST_EQUALITY(l.IsRequested("Graph", factory.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", factory.get()), false);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, RequestReleaseFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  Level l;
+  l.SetLevelID(0);
+
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  l.SetFactoryManager(facManager);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
+  l.Set("A", A);
+
+  RCP<FactoryBase> graphFact = rcp(new CoalesceDropFactory());
+  RCP<Factory> aggFact       = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", graphFact);
+
+  l.Request("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+
+  l.Release("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  Level l;
+  l.SetLevelID(0);
+
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  l.SetFactoryManager(facManager);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
+  l.Set("A", A);
+
+  RCP<Factory> graphFact = rcp(new CoalesceDropFactory());
+  RCP<Factory> aggFact   = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", graphFact);
+
+  l.Keep("Aggregates", aggFact.get());  // set keep flag
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
+  l.Request("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
+
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
+
+  l.Release("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
 
-    aLevel.Set("Hitchhiker's Guide", 42);
-    int fff = aLevel.Get<int>("Hitchhiker's Guide");
-    TEST_EQUALITY(fff, 42);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepAndBuildFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  Level l;
+  l.SetLevelID(0);  // level 0 necessary because of Nullspace factory
+
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  l.SetFactoryManager(facManager);
 
-    aLevel.Set("PI",3.14159265);
-    double ggg = aLevel.Get<double>("PI");
-    TEST_EQUALITY(ggg, 3.14159265);
-    TEST_EQUALITY(aLevel.IsAvailable("PI"), true);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(144);
+  l.Set("A", A);
 
-    aLevel.Delete("PI", MueLu::NoFactory::get());
-    TEST_EQUALITY(aLevel.IsAvailable("PI"), false);
+  RCP<CoalesceDropFactory> graphFact       = rcp(new CoalesceDropFactory());
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", graphFact);
 
-    aLevel.Set("Hello MueLu", std::string("Greetings to MueMat"));
-    std::string hhh = aLevel.Get<std::string>("Hello MueLu");
-    TEST_EQUALITY(hhh, "Greetings to MueMat");
+  l.Keep("Aggregates", aggFact.get());  // set keep flag
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
+  l.Request("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
 
-    aLevel.Set("A",A);
-    RCP<Matrix> newA = aLevel.Get< RCP<Matrix> >("A");
-    TEST_EQUALITY(newA, A);
+  aggFact->Build(l);
 
-    aLevel.Set("R", A);
-    RCP<Matrix> newR = aLevel.Get< RCP<Matrix> >("R");
-    TEST_EQUALITY(newR, A); //TODO from JG: must be tested using another matrix !
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
 
-    aLevel.Set("P", A);
-    RCP<Matrix> newP = aLevel.Get< RCP<Matrix> >("P");
-    TEST_EQUALITY(newP, A);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    aLevel.SetLevelID(42);
-    TEST_EQUALITY(aLevel.GetLevelID(), 42); //TODO: test default value of LevelID
-  }
+  l.Release(*aggFact);  // release dependencies only
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, NumRequests, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
 
-    Level aLevel;
-    aLevel.SetLevelID(0);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
-    aLevel.Set("A", A);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    aLevel.SetFactoryManager(facManager);
-    RCP<FactoryBase> factory = rcp(new CoalesceDropFactory());
+  l.Release("Aggregates", aggFact.get());
 
-    aLevel.Request("Graph", factory.get());
-    aLevel.Request("Graph", factory.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()), MueLu::Keep);
 
-    aLevel.Release("Graph", factory.get());
-    TEST_EQUALITY(aLevel.IsRequested("Graph", factory.get()), true);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    aLevel.Release("Graph", factory.get());
-    TEST_EQUALITY(aLevel.IsRequested("Graph", factory.get()), false);
-  }
+  l.RemoveKeepFlag("Aggregates", aggFact.get(), MueLu::Keep);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, RequestRelease, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    Level l;
-    l.SetLevelID(0);
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
-    l.Set("A", A);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
+}
 
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    l.SetFactoryManager(facManager);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepAndBuildFactory2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  Level l;
+  l.SetLevelID(0);  // level 0 necessary because of Nullspace factory
 
-    RCP<FactoryBase> factory = rcp(new CoalesceDropFactory());
+  RCP<FactoryManager> facManager = rcp(new FactoryManager());
+  l.SetFactoryManager(facManager);
 
-    l.Request("Graph", factory.get());
-    TEST_EQUALITY(l.IsRequested("Graph", factory.get()), true);
-    TEST_EQUALITY(l.IsAvailable("Graph", factory.get()), false);
-    l.Release("Graph", factory.get());
-    TEST_EQUALITY(l.IsRequested("Graph", factory.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph", factory.get()), false);
-  }
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(144);
+  l.Set("A", A);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, RequestReleaseFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    Level l;
-    l.SetLevelID(0);
+  RCP<CoalesceDropFactory> graphFact       = rcp(new CoalesceDropFactory());
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", graphFact);
 
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    l.SetFactoryManager(facManager);
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
-    l.Set("A", A);
+  l.Request("Aggregates", aggFact.get());
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
 
-    RCP<FactoryBase> graphFact = rcp(new CoalesceDropFactory());
-    RCP<Factory> aggFact  = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", graphFact);
+  aggFact->Build(l);
 
-    l.Request("Aggregates", aggFact.get());
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), true);
 
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), true);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    l.Release("Aggregates", aggFact.get());
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
+  l.Release(*aggFact);
 
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-  }
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), true);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), true);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    Level l;
-    l.SetLevelID(0);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    l.SetFactoryManager(facManager);
+  l.Release("Aggregates", aggFact.get());
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);
-    l.Set("A", A);
+  TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()), false);
 
-    RCP<Factory> graphFact = rcp(new CoalesceDropFactory());
-    RCP<Factory> aggFact   = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", graphFact);
+  TEST_EQUALITY(l.IsRequested("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.IsAvailable("Graph", graphFact.get()), false);
+  TEST_EQUALITY(l.GetKeepFlag("Graph", graphFact.get()), 0);
 
-    l.Keep("Aggregates", aggFact.get());      // set keep flag
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
-    l.Request("Aggregates", aggFact.get());
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
+  /*l.RemoveKeepFlag("Aggregates", aggFact.get(), MueLu::Keep);
 
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.Release("Aggregates", aggFact.get());
     TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
     TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
 
     TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
     TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-  }
+    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);*/
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepAndBuildFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    Level l;
-    l.SetLevelID(0); // level 0 necessary because of Nullspace factory
+// Helper class for unit test 'Level/CircularDependency'
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class CircularFactory : public MueLu::SingleLevelFactoryBase {
+#include <MueLu_UseShortNames.hpp>
 
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    l.SetFactoryManager(facManager);
+ public:
+  CircularFactory(int value)
+    : value_(value) {}
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(144);
-    l.Set("A", A);
-
-    RCP<CoalesceDropFactory>  graphFact = rcp(new CoalesceDropFactory());
-    RCP<UncoupledAggregationFactory> aggFact   = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", graphFact);
-
-    l.Keep("Aggregates", aggFact.get());      // set keep flag
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
-    l.Request("Aggregates", aggFact.get());
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
+  virtual ~CircularFactory() {}
 
-    aggFact->Build(l);
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.Release(*aggFact); // release dependencies only
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.Release("Aggregates", aggFact.get());
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.GetKeepFlag("Aggregates", aggFact.get()),   MueLu::Keep);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.RemoveKeepFlag("Aggregates", aggFact.get(), MueLu::Keep);
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
+  void SetCircularFactory(RCP<FactoryBase> circular) { circular_ = circular; }
 
+  void DeclareInput(Level& level) const {
+    level.DeclareInput("data", circular_.get(), this);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, KeepAndBuildFactory2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    Level l;
-    l.SetLevelID(0); // level 0 necessary because of Nullspace factory
-
-    RCP<FactoryManager> facManager = rcp(new FactoryManager());
-    l.SetFactoryManager(facManager);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(144);
-    l.Set("A", A);
-
-    RCP<CoalesceDropFactory>  graphFact = rcp(new CoalesceDropFactory());
-    RCP<UncoupledAggregationFactory> aggFact   = rcp(new UncoupledAggregationFactory());
-    aggFact->SetFactory("Graph", graphFact);
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-
-    l.Request("Aggregates", aggFact.get());
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-
-    aggFact->Build(l);
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   true);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), true);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.Release(*aggFact);
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   true);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   true);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    l.Release("Aggregates", aggFact.get());
-
-    TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-    TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-
-    TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-    TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);
-
-    /*l.RemoveKeepFlag("Aggregates", aggFact.get(), MueLu::Keep);
-
-      TEST_EQUALITY(l.IsRequested("Aggregates", aggFact.get()),   false);
-      TEST_EQUALITY(l.IsAvailable("Aggregates", aggFact.get()),   false);
-
-      TEST_EQUALITY(l.IsRequested("Graph",      graphFact.get()), false);
-      TEST_EQUALITY(l.IsAvailable("Graph",      graphFact.get()), false);
-      TEST_EQUALITY(l.GetKeepFlag("Graph",      graphFact.get()), 0);*/
-
+  void Build(Level& level) const {
+    level.Set("data", value_, this);
+    int value = level.Get<int>("data", circular_.get());
+    level.Set("data", value + value_, this);
   }
 
+ private:
+  int value_;
+  RCP<FactoryBase> circular_;
 
-  // Helper class for unit test 'Level/CircularDependency'
-  template
-  <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class CircularFactory : public MueLu::SingleLevelFactoryBase {
-
-#   include <MueLu_UseShortNames.hpp>
-
-    public:
-
-      CircularFactory(int value) : value_(value) { }
+};  // class CircularFactory
 
-      virtual ~CircularFactory() { }
+//! Even though it's very special, a factory can generate data, that it requests itself.
+//  Level must avoid self-recursive calls of Request
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, CircularDependencyWith1Factory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef CircularFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> circular_factory_type;
+  circular_factory_type A(2);
 
-      void SetCircularFactory(RCP<FactoryBase> circular) { circular_ = circular; }
+  A.SetCircularFactory(rcpFromRef(A));
 
-      void DeclareInput(Level &level) const {
-        level.DeclareInput("data", circular_.get(), this);
-      }
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-      void Build(Level& level) const {
-        level.Set("data", value_, this);
-        int value = level.Get<int>("data", circular_.get());
-        level.Set("data", value+value_, this);
-      }
+  level.Request("data", &A);
 
-    private:
+  TEST_EQUALITY(level.Get<int>("data", &A), (2 + 2));
 
-      int value_;
-      RCP<FactoryBase> circular_;
+  level.Release("data", &A);
+}
 
-  }; //class CircularFactory
+//! Test if circular dependencies between factories are allowed
+//  This test corresponds to a use-case found developping repartitionning capability
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, CircularDependencyWithTwoFactories, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef CircularFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> circular_factory_type;
+  circular_factory_type A(2);
+  circular_factory_type B(3);
 
-  //! Even though it's very special, a factory can generate data, that it requests itself.
-  //  Level must avoid self-recursive calls of Request
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, CircularDependencyWith1Factory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef CircularFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> circular_factory_type;
-    circular_factory_type A(2);
+  A.SetCircularFactory(rcpFromRef(B));
+  B.SetCircularFactory(rcpFromRef(A));
 
-    A.SetCircularFactory(rcpFromRef(A));
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-    Level level; TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Request("data", &A);
 
-    level.Request("data", &A);
+  A.Build(level);
 
-    TEST_EQUALITY(level.Get<int>("data", &A), (2 + 2));
-
-    level.Release("data", &A);
-  }
-
-  //! Test if circular dependencies between factories are allowed
-  //  This test corresponds to a use-case found developping repartitionning capability
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Level, CircularDependencyWithTwoFactories, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef CircularFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> circular_factory_type;
-    circular_factory_type A(2);
-    circular_factory_type B(3);
-
-    A.SetCircularFactory(rcpFromRef(B));
-    B.SetCircularFactory(rcpFromRef(A));
-
-    Level level; TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
-
-    level.Request("data", &A);
-
-    A.Build(level);
-
-    TEST_EQUALITY(level.Get<int>("data", &A), (2 + 3) + 2);
-    TEST_EQUALITY(level.Get<int>("data", &B), (2 + 3));
-
-    level.Release(A); // needed because A.Build(level) have been called manually
-    level.Release("data", &A);
-  }
+  TEST_EQUALITY(level.Get<int>("data", &A), (2 + 3) + 2);
+  TEST_EQUALITY(level.Get<int>("data", &B), (2 + 3));
 
-   #define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,SetCoreData,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,NumRequests,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,RequestRelease,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,RequestReleaseFactory,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,KeepFactory,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,KeepAndBuildFactory,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,KeepAndBuildFactory2,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,CircularDependencyWith1Factory,SC,LO,GO,NO) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level,CircularDependencyWithTwoFactories,SC,LO,GO,NO)
+  level.Release(A);  // needed because A.Build(level) have been called manually
+  level.Release("data", &A);
+}
 
-} // namespace MueLuTests
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, SetCoreData, SC, LO, GO, NO)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, NumRequests, SC, LO, GO, NO)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, RequestRelease, SC, LO, GO, NO)                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, RequestReleaseFactory, SC, LO, GO, NO)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, KeepFactory, SC, LO, GO, NO)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, KeepAndBuildFactory, SC, LO, GO, NO)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, KeepAndBuildFactory2, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, CircularDependencyWith1Factory, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Level, CircularDependencyWithTwoFactories, SC, LO, GO, NO)
 
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/LowPrecision.cpp b/packages/muelu/test/unit_tests/LowPrecision.cpp
index 1a03d0085a33..3ab91360d047 100644
--- a/packages/muelu/test/unit_tests/LowPrecision.cpp
+++ b/packages/muelu/test/unit_tests/LowPrecision.cpp
@@ -61,116 +61,114 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LowPrecisionFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using Teuchos::TimeMonitor;
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using HalfScalar = typename TST::halfPrecision;
-    using HalfTST = Teuchos::ScalarTraits<HalfScalar>;
-
-    // Modify to read in matrix
-    std::string matrix_file = "";
-
-    // for timing
-    int numMatvecs = 100;
-
-    RCP<Matrix> A;
-    if (matrix_file == "") {
-      int nx = 1000;
-      A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build2DPoisson(nx);
-    } else {
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      A = Xpetra::IO<SC, LO, GO, NO>::Read(matrix_file, Xpetra::UseTpetra, comm);
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LowPrecisionFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using Teuchos::TimeMonitor;
+  using TST        = Teuchos::ScalarTraits<Scalar>;
+  using HalfScalar = typename TST::halfPrecision;
+  using HalfTST    = Teuchos::ScalarTraits<HalfScalar>;
+
+  // Modify to read in matrix
+  std::string matrix_file = "";
+
+  // for timing
+  int numMatvecs = 100;
+
+  RCP<Matrix> A;
+  if (matrix_file == "") {
+    int nx = 1000;
+    A      = TestHelpers::TestFactory<SC, LO, GO, NO>::Build2DPoisson(nx);
+  } else {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    A                                   = Xpetra::IO<SC, LO, GO, NO>::Read(matrix_file, Xpetra::UseTpetra, comm);
+  }
 
-    RCP<Operator> lowA;
-    const bool isTpetra = A->getDomainMap()->lib() == Xpetra::UseTpetra;
-    std::list<std::string> matrixKeys = {"A", "P", "R"};
-    for (auto it = matrixKeys.begin(); it != matrixKeys.end(); ++it) {
-      Level aLevel;
-      TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Set(*it,A);
-
-      // Set up LowPrecisionFactory
-      RCP<LowPrecisionFactory> LP = rcp(new LowPrecisionFactory());
-      Teuchos::ParameterList params;
-      params.set("matrix key", *it);
-      LP->SetParameterList(params);
-
-      // Build
-      aLevel.Request(*it,LP.get());
-      LP->Build(aLevel);
-      TEST_EQUALITY(aLevel.IsAvailable(*it,LP.get()), true);
-      lowA = aLevel.Get< RCP<Operator> >(*it,LP.get());
-      aLevel.Release(*it,LP.get());
-
-      // Check that the resulting Op is as expected.
-      if (isTpetra && (false
+  RCP<Operator> lowA;
+  const bool isTpetra               = A->getDomainMap()->lib() == Xpetra::UseTpetra;
+  std::list<std::string> matrixKeys = {"A", "P", "R"};
+  for (auto it = matrixKeys.begin(); it != matrixKeys.end(); ++it) {
+    Level aLevel;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+    aLevel.Set(*it, A);
+
+    // Set up LowPrecisionFactory
+    RCP<LowPrecisionFactory> LP = rcp(new LowPrecisionFactory());
+    Teuchos::ParameterList params;
+    params.set("matrix key", *it);
+    LP->SetParameterList(params);
+
+    // Build
+    aLevel.Request(*it, LP.get());
+    LP->Build(aLevel);
+    TEST_EQUALITY(aLevel.IsAvailable(*it, LP.get()), true);
+    lowA = aLevel.Get<RCP<Operator> >(*it, LP.get());
+    aLevel.Release(*it, LP.get());
+
+    // Check that the resulting Op is as expected.
+    if (isTpetra && (false
 #if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_FLOAT)
-                       || std::is_same<Scalar, double>::value
+                     || std::is_same<Scalar, double>::value
 #endif
 #if defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_COMPLEX_FLOAT)
-                       || std::is_same<Scalar, std::complex<double> >::value
+                     || std::is_same<Scalar, std::complex<double> >::value
 #endif
-                       )) {
-        auto tpCrsMultOp = rcp_dynamic_cast<Tpetra::CrsMatrixMultiplyOp<Scalar,HalfScalar,LocalOrdinal,GlobalOrdinal,Node> >(rcp_dynamic_cast<TpetraOperator>(lowA)->getOperator());
-        TEST_ASSERT(!tpCrsMultOp.is_null());                     // Actually converted
-      } else {
-        TEST_ASSERT(!rcp_dynamic_cast<Matrix>(lowA).is_null());  // Just a regular old matrix
-        TEST_EQUALITY(A, lowA);
-      }
+                     )) {
+      auto tpCrsMultOp = rcp_dynamic_cast<Tpetra::CrsMatrixMultiplyOp<Scalar, HalfScalar, LocalOrdinal, GlobalOrdinal, Node> >(rcp_dynamic_cast<TpetraOperator>(lowA)->getOperator());
+      TEST_ASSERT(!tpCrsMultOp.is_null());  // Actually converted
+    } else {
+      TEST_ASSERT(!rcp_dynamic_cast<Matrix>(lowA).is_null());  // Just a regular old matrix
+      TEST_EQUALITY(A, lowA);
     }
+  }
 
-    // Check apply
-    if (isTpetra && std::is_same<Scalar, double>::value) {
-      RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(),1);
-      X->randomize();
+  // Check apply
+  if (isTpetra && std::is_same<Scalar, double>::value) {
+    RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    X->randomize();
 
-      RCP<MultiVector> B = MultiVectorFactory::Build(A->getRangeMap(),1);
-      RCP<MultiVector> B0 = MultiVectorFactory::Build(lowA->getRangeMap(),1);
+    RCP<MultiVector> B  = MultiVectorFactory::Build(A->getRangeMap(), 1);
+    RCP<MultiVector> B0 = MultiVectorFactory::Build(lowA->getRangeMap(), 1);
 
-      // warm up
-      for (int i = 0; i < 5; i++) {
+    // warm up
+    for (int i = 0; i < 5; i++) {
+      A->apply(*X, *B);
+    }
+    {
+      RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatVec original")));
+      for (int i = 0; i < numMatvecs; i++) {
         A->apply(*X, *B);
       }
-      {
-        RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatVec original")));
-        for (int i = 0; i < numMatvecs; i++) {
-          A->apply(*X, *B);
-        }
-      }
+    }
 
-      // warm up low precision
-      for (int i = 0; i < 5; i++) {
+    // warm up low precision
+    for (int i = 0; i < 5; i++) {
+      lowA->apply(*X, *B0);
+    }
+    {
+      RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatVec low precision")));
+      for (int i = 0; i < numMatvecs; i++) {
         lowA->apply(*X, *B0);
       }
-      {
-        RCP<TimeMonitor> tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatVec low precision")));
-        for (int i = 0; i < numMatvecs; i++) {
-          lowA->apply(*X, *B0);
-        }
-      }
-
-      TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
-      TimeMonitor::zeroOutTimers();
+    }
 
-      B->update(-1, *B0, 1);
+    TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
+    TimeMonitor::zeroOutTimers();
 
-      Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norm(1);
-      B->norm2(norm);
-      TEST_FLOATING_EQUALITY(norm[0], TST::zero(), HalfTST::eps());
-    }
+    B->update(-1, *B0, 1);
 
+    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> norm(1);
+    B->norm2(norm);
+    TEST_FLOATING_EQUALITY(norm[0], TST::zero(), HalfTST::eps());
   }
+}
 
 #define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LowPrecisionFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // end MueLuTests namespace
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/MapTransferFactory.cpp b/packages/muelu/test/unit_tests/MapTransferFactory.cpp
index d7a5c78cc2f0..a092aef3176a 100644
--- a/packages/muelu/test/unit_tests/MapTransferFactory.cpp
+++ b/packages/muelu/test/unit_tests/MapTransferFactory.cpp
@@ -64,734 +64,726 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  TEST_ASSERT(!mapTransferFactory.is_null());
+}  // Constructor
+
+/* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
+ * needs to match the domain map of the prolongator.
+ *
+ * Assume a 1D Poisson discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
+
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+
+  // Create a dummy matrix needed to build a prolongator
+  const GO nx   = 199;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  fineLevel->Set("A", A);
+
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, A->getRowMap());
+
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    out << "version: " << MueLu::Version() << std::endl;
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
+
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
+
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
+
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
 
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    TEST_ASSERT(!mapTransferFactory.is_null());
-  } // Constructor
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
+
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
+
+  TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
+}
+
+/* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
+ * needs to match the domain map of the prolongator.
+ *
+ * Assume a 2D Poisson discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
+
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
 
-  /* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
-   * needs to match the domain map of the prolongator.
-   *
-   * Assume a 1D Poisson discretization.
-   */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  // Create a dummy matrix needed to build a prolongator
+  const GO nx   = 19;
+  const GO ny   = 17;
+  RCP<Matrix> A = test_factory::Build2DPoisson(nx, ny);
+  fineLevel->Set("A", A);
+
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, A->getRowMap());
+
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
 
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
 
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
 
-    // Create a dummy matrix needed to build a prolongator
-    const GO nx = 199;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    fineLevel->Set("A", A);
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
 
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, A->getRowMap());
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
 
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
 
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
+}
 
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
+/* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
+ * needs to match the domain map of the prolongator.
+ *
+ * Assume a 3D elasticity discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap3DElasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
 
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
 
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
 
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
 
-    TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
+  // Create a 3D elsasticity matrix needed to build a prolongator
+  RCP<Matrix> A = Teuchos::null;
+  {
+    const GO nx                  = 5;
+    const GO ny                  = 5;
+    const GO nz                  = 5;
+    const int numDofsPerNode     = 3;  // 3D elasticity
+    const std::string matrixType = "Elasticity3D";
+
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    galeriList.set("ny", ny);
+    galeriList.set("nz", nz);
+    galeriList.set("matrixType", matrixType);
+    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian3D", TestHelpers::Parameters::getDefaultComm(), galeriList);
+    RCP<const Map> dofMap  = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, dofMap, galeriList);
+    A = Pr->BuildMatrix();
+
+    TEST_ASSERT(!A.is_null());
+
+    fineLevel->Set("A", A);
   }
 
-  /* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
-   * needs to match the domain map of the prolongator.
-   *
-   * Assume a 2D Poisson discretization.
-   */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, A->getRowMap());
 
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
 
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
 
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
 
-    // Create a dummy matrix needed to build a prolongator
-    const GO nx = 19;
-    const GO ny = 17;
-    RCP<Matrix> A = test_factory::Build2DPoisson(nx, ny);
-    fineLevel->Set("A", A);
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
 
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, A->getRowMap());
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
 
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
 
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
 
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
+  TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
+}
 
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
+/* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
+ * needs to match the domain map of the prolongator. This particular test targets the special case of
+ * limiting the number of columns from P to be used for the transfer (e.g. when excluding the rotations
+ * from the nullspace when creating the MapTransfer.)
+ *
+ * Assume a 3D elasticity discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap3DElasticityReducedP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
 
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
 
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
 
-    TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
-  }
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
 
-  /* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
-   * needs to match the domain map of the prolongator.
-   *
-   * Assume a 3D elasticity discretization.
-   */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap3DElasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+  // Create a 3D elasticity matrix and its nullspace needed to build a prolongator
+  RCP<Matrix> A = Teuchos::null;
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
-
-    RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
-
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
-
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
-
-    // Create a 3D elsasticity matrix needed to build a prolongator
-    RCP<Matrix> A = Teuchos::null;
-    {
-      const GO nx = 5;
-      const GO ny = 5;
-      const GO nz = 5;
-      const int numDofsPerNode = 3; // 3D elasticity
-      const std::string matrixType = "Elasticity3D";
-
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-      galeriList.set("nz", nz);
-      galeriList.set("matrixType", matrixType);
-      RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian3D", TestHelpers::Parameters::getDefaultComm(), galeriList);
-      RCP<const Map> dofMap = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, dofMap, galeriList);
-      A = Pr->BuildMatrix();
-
-      TEST_ASSERT(!A.is_null());
-
-      fineLevel->Set("A", A);
-    }
+    const GO nx                  = 10;
+    const GO ny                  = 10;
+    const GO nz                  = 5;
+    const int numDofsPerNode     = 3;  // 3D elasticity
+    const std::string matrixType = "Elasticity3D";
+
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    galeriList.set("ny", ny);
+    galeriList.set("nz", nz);
+    galeriList.set("matrixType", matrixType);
+    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian3D", comm, galeriList);
+    RCP<const Map> dofMap  = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, dofMap, galeriList);
+    A                          = Pr->BuildMatrix();
+    RCP<MultiVector> nullspace = Pr->BuildNullspace();
+
+    TEST_ASSERT(!A.is_null());
+    TEST_ASSERT(nullspace->getNumVectors() == 6);
+
+    fineLevel->Set("A", A);
+    fineLevel->Set("Nullspace", nullspace);
+  }
 
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, A->getRowMap());
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, A->getRowMap());
 
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
 
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  tentativePFact->SetParameter("Nullspace name", Teuchos::ParameterEntry(std::string("Nullspace")));
 
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
+  mapTransferFactory->SetParameter("nullspace vectors: limit to", Teuchos::ParameterEntry(std::string("translations")));
 
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
 
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
 
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
 
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
 
-    TEST_ASSERT(coarsenedMap->isSameAs(*Ptent->getDomainMap()));
-  }
+  const global_size_t gNumTranslationalDOFs = Ptent->getDomainMap()->getGlobalNumElements() / 2;
+  TEST_EQUALITY_CONST(coarsenedMap->getMinGlobalIndex(), Ptent->getDomainMap()->getMinGlobalIndex());
+  TEST_EQUALITY_CONST(coarsenedMap->getGlobalNumElements(), gNumTranslationalDOFs);
 
-  /* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
-   * needs to match the domain map of the prolongator. This particular test targets the special case of
-   * limiting the number of columns from P to be used for the transfer (e.g. when excluding the rotations
-   * from the nullspace when creating the MapTransfer.)
-   *
-   * Assume a 3D elasticity discretization.
+  /* Manually extract translational part from domain map of prolongator and compare with coarsenedMap
+   * Thereby, we can exploit the fact, that each nodes carries 6 unknowns on the coarse level: 3 translations
+   * and 3 rotations. Here, we're only interested in the translations, i.e. the first 3 unknowns per node.
    */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap3DElasticityReducedP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
-
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
-
-    // Create a 3D elasticity matrix and its nullspace needed to build a prolongator
-    RCP<Matrix> A = Teuchos::null;
-    {
-      const GO nx = 10;
-      const GO ny = 10;
-      const GO nz = 5;
-      const int numDofsPerNode = 3; // 3D elasticity
-      const std::string matrixType = "Elasticity3D";
-
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-      galeriList.set("nz", nz);
-      galeriList.set("matrixType", matrixType);
-      RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian3D", comm, galeriList);
-      RCP<const Map> dofMap = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, dofMap, galeriList);
-      A = Pr->BuildMatrix();
-      RCP<MultiVector> nullspace = Pr->BuildNullspace();
-
-      TEST_ASSERT(!A.is_null());
-      TEST_ASSERT(nullspace->getNumVectors()==6);
-
-      fineLevel->Set("A", A);
-      fineLevel->Set("Nullspace", nullspace);
+    ArrayView<const GO> domainMapGIDs = Ptent->getDomainMap()->getLocalElementList();
+    Array<GO> gidsOfTranslations;
+    const GO lNumCoarseNodes             = Teuchos::as<GO>(domainMapGIDs.size() / 6);
+    const GO gNumCoarseTranslationalDOFs = Teuchos::as<GO>(Ptent->getDomainMap()->getGlobalNumElements() / 2);
+
+    for (size_t nodeID = 0; nodeID < Teuchos::as<size_t>(lNumCoarseNodes); ++nodeID) {
+      size_t idOfFirstDofOfThisNode = nodeID * 6;
+      gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode]);
+      gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 1]);
+      gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 2]);
     }
+    RCP<const Map> mapForComparison = MapFactory::Build(coarsenedMap->lib(),
+                                                        gNumCoarseTranslationalDOFs, gidsOfTranslations, Teuchos::OrdinalTraits<GO>::zero(), comm);
 
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, A->getRowMap());
-
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
-
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
-    tentativePFact->SetParameter("Nullspace name", Teuchos::ParameterEntry(std::string("Nullspace")));
-
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
-    mapTransferFactory->SetParameter("nullspace vectors: limit to", Teuchos::ParameterEntry(std::string("translations")));
-
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
-
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
-
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
-
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
-
-    const global_size_t gNumTranslationalDOFs = Ptent->getDomainMap()->getGlobalNumElements() / 2;
-    TEST_EQUALITY_CONST(coarsenedMap->getMinGlobalIndex(), Ptent->getDomainMap()->getMinGlobalIndex());
-    TEST_EQUALITY_CONST(coarsenedMap->getGlobalNumElements(), gNumTranslationalDOFs);
-
-    /* Manually extract translational part from domain map of prolongator and compare with coarsenedMap
-     * Thereby, we can exploit the fact, that each nodes carries 6 unknowns on the coarse level: 3 translations
-     * and 3 rotations. Here, we're only interested in the translations, i.e. the first 3 unknowns per node.
-     */
-    {
-      ArrayView<const GO> domainMapGIDs = Ptent->getDomainMap()->getLocalElementList();
-      Array<GO> gidsOfTranslations;
-      const GO lNumCoarseNodes = Teuchos::as<GO>(domainMapGIDs.size() / 6);
-      const GO gNumCoarseTranslationalDOFs = Teuchos::as<GO>(Ptent->getDomainMap()->getGlobalNumElements() / 2);
-
-      for (size_t nodeID = 0; nodeID < Teuchos::as<size_t>(lNumCoarseNodes); ++nodeID) {
-        size_t idOfFirstDofOfThisNode = nodeID * 6;
-        gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode]);
-        gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 1]);
-        gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 2]);
-      }
-      RCP<const Map> mapForComparison = MapFactory::Build(coarsenedMap->lib(),
-          gNumCoarseTranslationalDOFs, gidsOfTranslations, Teuchos::OrdinalTraits<GO>::zero(), comm);
-
-      TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
-    }
+    TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+  }
+}
+
+/* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
+ * needs to match the domain map of the prolongator. This particular test targets the special case of
+ * limiting the number of columns from P to be used for the transfer (e.g. when excluding the rotations
+ * from the nullspace when creating the MapTransfer.)
+ *
+ * Assume a 2D elasticity discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap2DElasticityReducedP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
+
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+
+  // Create a 3D elasticity matrix and its nullspace needed to build a prolongator
+  RCP<Matrix> A = Teuchos::null;
+  {
+    const GO nx                  = 10;
+    const GO ny                  = 5;
+    const int numDofsPerNode     = 2;  // 2D elasticity
+    const std::string matrixType = "Elasticity2D";
+
+    Teuchos::ParameterList galeriList;
+    galeriList.set("nx", nx);
+    galeriList.set("ny", ny);
+    galeriList.set("matrixType", matrixType);
+    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+    RCP<const Map> dofMap  = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, dofMap, galeriList);
+    A                          = Pr->BuildMatrix();
+    RCP<MultiVector> nullspace = Pr->BuildNullspace();
+
+    TEST_ASSERT(!A.is_null());
+    TEST_ASSERT(nullspace->getNumVectors() == 3);
+
+    fineLevel->Set("A", A);
+    fineLevel->Set("Nullspace", nullspace);
   }
 
-  /* This tests coarsens the row map of the fine level operator, so the result from the MapTransferFactory
-   * needs to match the domain map of the prolongator. This particular test targets the special case of
-   * limiting the number of columns from P to be used for the transfer (e.g. when excluding the rotations
-   * from the nullspace when creating the MapTransfer.)
-   *
-   * Assume a 2D elasticity discretization.
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, A->getRowMap());
+
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
+
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  tentativePFact->SetParameter("Nullspace name", Teuchos::ParameterEntry(std::string("Nullspace")));
+
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
+  mapTransferFactory->SetParameter("nullspace vectors: limit to", Teuchos::ParameterEntry(std::string("translations")));
+
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
+
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
+
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
+
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
+
+  const global_size_t gNumTranslationalDOFs = Ptent->getDomainMap()->getGlobalNumElements() * 2 / 3;
+  TEST_EQUALITY_CONST(coarsenedMap->getMinGlobalIndex(), Ptent->getDomainMap()->getMinGlobalIndex());
+  TEST_EQUALITY_CONST(coarsenedMap->getGlobalNumElements(), gNumTranslationalDOFs);
+
+  /* Manually extract translational part from domain map of prolongator and compare with coarsenedMap
+   * Thereby, we can exploit the fact, that each nodes carries 6 unknowns on the coarse level: 2 translations
+   * and 1 rotation. Here, we're only interested in the translations, i.e. the first 2 unknowns per node.
    */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferFullMap2DElasticityReducedP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
-
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
-
-    // Create a 3D elasticity matrix and its nullspace needed to build a prolongator
-    RCP<Matrix> A = Teuchos::null;
-    {
-      const GO nx = 10;
-      const GO ny = 5;
-      const int numDofsPerNode = 2; // 2D elasticity
-      const std::string matrixType = "Elasticity2D";
-
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-      galeriList.set("matrixType", matrixType);
-      RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-      RCP<const Map> dofMap = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, dofMap, galeriList);
-      A = Pr->BuildMatrix();
-      RCP<MultiVector> nullspace = Pr->BuildNullspace();
-
-      TEST_ASSERT(!A.is_null());
-      TEST_ASSERT(nullspace->getNumVectors()==3);
-
-      fineLevel->Set("A", A);
-      fineLevel->Set("Nullspace", nullspace);
+    ArrayView<const GO> domainMapGIDs = Ptent->getDomainMap()->getLocalElementList();
+    Array<GO> gidsOfTranslations;
+    const GO lNumCoarseNodes             = Teuchos::as<GO>(domainMapGIDs.size() / 3);
+    const GO gNumCoarseTranslationalDOFs = Teuchos::as<GO>(Ptent->getDomainMap()->getGlobalNumElements() * 2 / 3);
+
+    for (size_t nodeID = 0; nodeID < Teuchos::as<size_t>(lNumCoarseNodes); ++nodeID) {
+      size_t idOfFirstDofOfThisNode = nodeID * 3;
+      gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode]);
+      gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 1]);
     }
+    RCP<const Map> mapForComparison = MapFactory::Build(coarsenedMap->lib(),
+                                                        gNumCoarseTranslationalDOFs, gidsOfTranslations, Teuchos::OrdinalTraits<GO>::zero(), comm);
 
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, A->getRowMap());
-
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
-
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
-    tentativePFact->SetParameter("Nullspace name", Teuchos::ParameterEntry(std::string("Nullspace")));
-
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
-    mapTransferFactory->SetParameter("nullspace vectors: limit to", Teuchos::ParameterEntry(std::string("translations")));
-
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
-
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
-
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
-
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
-
-    const global_size_t gNumTranslationalDOFs = Ptent->getDomainMap()->getGlobalNumElements() * 2 / 3;
-    TEST_EQUALITY_CONST(coarsenedMap->getMinGlobalIndex(), Ptent->getDomainMap()->getMinGlobalIndex());
-    TEST_EQUALITY_CONST(coarsenedMap->getGlobalNumElements(), gNumTranslationalDOFs);
-
-    /* Manually extract translational part from domain map of prolongator and compare with coarsenedMap
-     * Thereby, we can exploit the fact, that each nodes carries 6 unknowns on the coarse level: 2 translations
-     * and 1 rotation. Here, we're only interested in the translations, i.e. the first 2 unknowns per node.
-     */
-    {
-      ArrayView<const GO> domainMapGIDs = Ptent->getDomainMap()->getLocalElementList();
-      Array<GO> gidsOfTranslations;
-      const GO lNumCoarseNodes = Teuchos::as<GO>(domainMapGIDs.size() / 3);
-      const GO gNumCoarseTranslationalDOFs = Teuchos::as<GO>(Ptent->getDomainMap()->getGlobalNumElements() * 2 / 3);
-
-      for (size_t nodeID = 0; nodeID < Teuchos::as<size_t>(lNumCoarseNodes); ++nodeID) {
-        size_t idOfFirstDofOfThisNode = nodeID * 3;
-        gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode]);
-        gidsOfTranslations.push_back(domainMapGIDs[idOfFirstDofOfThisNode + 1]);
-      }
-      RCP<const Map> mapForComparison = MapFactory::Build(coarsenedMap->lib(),
-          gNumCoarseTranslationalDOFs, gidsOfTranslations, Teuchos::OrdinalTraits<GO>::zero(), comm);
-
-      TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
-    }
+    TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+  }
+}
+
+/* This tests coarsens a subset of the row map of the fine level operator,
+ * so the result from the MapTransferFactory needs to match a subset of the domain map of the prolongator.
+ *
+ * Assume a 1D Poisson discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferPartialMap1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using test_factory  = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
+
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+
+  // Create a dummy matrix needed to build a prolongator
+  const GO nx   = 49;
+  RCP<Matrix> A = test_factory::Build1DPoisson(4 * nx);
+  fineLevel->Set("A", A);
+
+  // Extract a subset of A->getRowMap()'s GIDs to create the map to be transferred
+  ArrayView<const GO> allRowGIDs = A->getRowMap()->getLocalElementList();
+  Array<GO> myMapGIDs;
+  for (LO lid = 0; lid < Teuchos::as<LO>(nx); ++lid) {
+    if (lid % 3 == 0)
+      myMapGIDs.push_back(allRowGIDs[lid]);
   }
+  GO gNumFineEntries = 0;
+  reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myMapGIDs.size()), Teuchos::outArg(gNumFineEntries));
+  RCP<const Map> mapWithHoles = MapFactory::Build(TestHelpers::Parameters::getLib(), gNumFineEntries, myMapGIDs(), Teuchos::ScalarTraits<GO>::zero(), comm);
 
-  /* This tests coarsens a subset of the row map of the fine level operator,
-   * so the result from the MapTransferFactory needs to match a subset of the domain map of the prolongator.
-   *
-   * Assume a 1D Poisson discretization.
-   */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferPartialMap1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, mapWithHoles);
 
-    using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
 
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
 
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
 
-    // Create a dummy matrix needed to build a prolongator
-    const GO nx = 49;
-    RCP<Matrix> A = test_factory::Build1DPoisson(4*nx);
-    fineLevel->Set("A", A);
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
+
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
 
-    // Extract a subset of A->getRowMap()'s GIDs to create the map to be transferred
-    ArrayView<const GO> allRowGIDs = A->getRowMap()->getLocalElementList();
-    Array<GO> myMapGIDs;
-    for (LO lid = 0; lid < Teuchos::as<LO>(nx); ++lid) {
-      if (lid % 3 == 0)
-        myMapGIDs.push_back(allRowGIDs[lid]);
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
+
+  /* Manually construct a coarse version of the mapWithHoles by
+   * 1. populating a vector based on the mapWithHoles
+   * 2. manually restricting that vector to the coarse level
+   * 3. reconstructing the mapForComparison by only taking into accounts vector entries != 0
+   */
+  {
+    // Populate a fine level vector w/ ones according to fineMap
+    RCP<Vector> fullFineVec    = VectorFactory::Build(Ptent->getRangeMap(), true);
+    RCP<Vector> partialFineVec = VectorFactory::Build(mapWithHoles, true);
+    partialFineVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+    RCP<Import> fineImporter = ImportFactory::Build(mapWithHoles, fullFineVec->getMap());
+    fullFineVec->doImport(*partialFineVec, *fineImporter, Xpetra::INSERT);
+
+    // Restrict to coarse level manually
+    RCP<Vector> fullCoarseVec = VectorFactory::Build(Ptent->getDomainMap(), true);
+    Ptent->apply(*fullFineVec, *fullCoarseVec, Teuchos::TRANS);
+
+    // Reconstruct coarse map for result checking
+    ArrayRCP<const Scalar> coarseVecEntries = fullCoarseVec->getData(0);
+    Array<GO> myCoarseGIDs;
+    for (LO lid = 0; lid < static_cast<LO>(coarseVecEntries.size()); ++lid) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(coarseVecEntries[lid]) > Teuchos::ScalarTraits<MagnitudeType>::zero())
+        myCoarseGIDs.push_back(Ptent->getDomainMap()->getGlobalElement(lid));
     }
-    GO gNumFineEntries = 0;
-    reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myMapGIDs.size()), Teuchos::outArg(gNumFineEntries));
-    RCP<const Map> mapWithHoles = MapFactory::Build(TestHelpers::Parameters::getLib(), gNumFineEntries, myMapGIDs(), Teuchos::ScalarTraits<GO>::zero(), comm);
-
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, mapWithHoles);
-
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
-
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
-
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
-
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
-
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
-
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
-
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
-
-    /* Manually construct a coarse version of the mapWithHoles by
-     * 1. populating a vector based on the mapWithHoles
-     * 2. manually restricting that vector to the coarse level
-     * 3. reconstructing the mapForComparison by only taking into accounts vector entries != 0
-     */
-    {
-      // Populate a fine level vector w/ ones according to fineMap
-      RCP<Vector> fullFineVec = VectorFactory::Build(Ptent->getRangeMap(), true);
-      RCP<Vector> partialFineVec = VectorFactory::Build(mapWithHoles, true);
-      partialFineVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-      RCP<Import> fineImporter = ImportFactory::Build(mapWithHoles, fullFineVec->getMap());
-      fullFineVec->doImport(*partialFineVec, *fineImporter, Xpetra::INSERT);
-
-      // Restrict to coarse level manually
-      RCP<Vector> fullCoarseVec = VectorFactory::Build(Ptent->getDomainMap(), true);
-      Ptent->apply(*fullFineVec, *fullCoarseVec, Teuchos::TRANS);
-
-      // Reconstruct coarse map for result checking
-      ArrayRCP<const Scalar> coarseVecEntries = fullCoarseVec->getData(0);
-      Array<GO> myCoarseGIDs;
-      for (LO lid = 0; lid < static_cast<LO>(coarseVecEntries.size()); ++lid) {
-        if (Teuchos::ScalarTraits<Scalar>::magnitude(coarseVecEntries[lid]) > Teuchos::ScalarTraits<MagnitudeType>::zero())
-          myCoarseGIDs.push_back(Ptent->getDomainMap()->getGlobalElement(lid));
-      }
-      GO gNumCoarseEntries = 0;
-      reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myCoarseGIDs.size()), Teuchos::outArg(gNumCoarseEntries));
-      RCP<const Map> mapForComparison = MapFactory::Build(mapWithHoles->lib(), gNumCoarseEntries, myCoarseGIDs,
-          Teuchos::ScalarTraits<GO>::zero(), comm);
-
-      TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+    GO gNumCoarseEntries = 0;
+    reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myCoarseGIDs.size()), Teuchos::outArg(gNumCoarseEntries));
+    RCP<const Map> mapForComparison = MapFactory::Build(mapWithHoles->lib(), gNumCoarseEntries, myCoarseGIDs,
+                                                        Teuchos::ScalarTraits<GO>::zero(), comm);
+
+    TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+  }
+}
+
+/* This tests coarsens a subset of the row map of the fine level operator,
+ * so the result from the MapTransferFactory needs to match a subset of the domain map of the prolongator.
+ *
+ * Assume a 2D Poisson discretization.
+ */
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferPartialMap2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using test_factory  = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  // Manual setup of a two-level hierarchy
+  RCP<Level> fineLevel   = rcp(new Level());
+  RCP<Level> coarseLevel = rcp(new Level());
+  coarseLevel->SetPreviousLevel(fineLevel);
+  fineLevel->SetLevelID(0);
+  coarseLevel->SetLevelID(1);
+
+  TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
+  TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+
+  // Create a dummy matrix needed to build a prolongator
+  const GO nx   = 49;
+  RCP<Matrix> A = test_factory::Build2DPoisson(4 * nx);
+  fineLevel->Set("A", A);
+
+  // Extract a subset of A->getRowMap()'s GIDs to create the map to be transferred
+  ArrayView<const GO> allRowGIDs = A->getRowMap()->getLocalElementList();
+  Array<GO> myMapGIDs;
+  for (LO lid = 0; lid < Teuchos::as<LO>(nx); ++lid) {
+    if (lid % 3 == 0) {
+      myMapGIDs.push_back(allRowGIDs[lid]);
+      myMapGIDs.push_back(allRowGIDs[lid + 1]);
     }
   }
+  GO gNumFineEntries = 0;
+  reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myMapGIDs.size()), Teuchos::outArg(gNumFineEntries));
+  RCP<const Map> mapWithHoles = MapFactory::Build(TestHelpers::Parameters::getLib(), gNumFineEntries, myMapGIDs(), Teuchos::ScalarTraits<GO>::zero(), comm);
 
-  /* This tests coarsens a subset of the row map of the fine level operator,
-   * so the result from the MapTransferFactory needs to match a subset of the domain map of the prolongator.
-   *
-   * Assume a 2D Poisson discretization.
-   */
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MapTransferFactory, TransferPartialMap2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  const std::string mapName = "Dummy Map";
+  fineLevel->Set(mapName, mapWithHoles);
 
-    using MagnitudeType = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using test_factory = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
+  TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test transfer of a map with the MapTransferFactory" << std::endl;
+  RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
+  factoryManager->SetKokkosRefactor(false);
+  factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
+  fineLevel->SetFactoryManager(factoryManager);
+  coarseLevel->SetFactoryManager(factoryManager);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
 
-    // Manual setup of a two-level hierarchy
-    RCP<Level> fineLevel = rcp(new Level());
-    RCP<Level> coarseLevel = rcp(new Level());
-    coarseLevel->SetPreviousLevel(fineLevel);
-    fineLevel->SetLevelID(0);
-    coarseLevel->SetLevelID(1);
+  RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
+  mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+  mapTransferFactory->SetFactory("P", tentativePFact);
 
-    TEST_EQUALITY_CONST(fineLevel->GetLevelID(), 0);
-    TEST_EQUALITY_CONST(coarseLevel->GetLevelID(), 1);
+  coarseLevel->Request(mapName, MueLu::NoFactory::get());
+  coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
+  coarseLevel->Request(*mapTransferFactory);  // This calls DeclareInput() on mapTransferFactory
 
-    // Create a dummy matrix needed to build a prolongator
-    const GO nx = 49;
-    RCP<Matrix> A = test_factory::Build2DPoisson(4*nx);
-    fineLevel->Set("A", A);
+  TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
+  TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
 
-    // Extract a subset of A->getRowMap()'s GIDs to create the map to be transferred
-    ArrayView<const GO> allRowGIDs = A->getRowMap()->getLocalElementList();
-    Array<GO> myMapGIDs;
-    for (LO lid = 0; lid < Teuchos::as<LO>(nx); ++lid) {
-      if (lid % 3 == 0) {
-        myMapGIDs.push_back(allRowGIDs[lid]);
-        myMapGIDs.push_back(allRowGIDs[lid+1]);
-      }
-    }
-    GO gNumFineEntries = 0;
-    reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myMapGIDs.size()), Teuchos::outArg(gNumFineEntries));
-    RCP<const Map> mapWithHoles = MapFactory::Build(TestHelpers::Parameters::getLib(), gNumFineEntries, myMapGIDs(), Teuchos::ScalarTraits<GO>::zero(), comm);
-
-    const std::string mapName = "Dummy Map";
-    fineLevel->Set(mapName, mapWithHoles);
-
-    TEST_ASSERT(fineLevel->IsAvailable("A", MueLu::NoFactory::get()));
-    TEST_ASSERT(fineLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<FactoryManager> factoryManager = rcp(new FactoryManager());
-    factoryManager->SetKokkosRefactor(false);
-    factoryManager->SetFactory(mapName, MueLu::NoFactory::getRCP());
-    fineLevel->SetFactoryManager(factoryManager);
-    coarseLevel->SetFactoryManager(factoryManager);
-
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
-
-    RCP<MapTransferFactory> mapTransferFactory = rcp(new MapTransferFactory());
-    mapTransferFactory->SetParameter("map: factory", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-    mapTransferFactory->SetFactory("P", tentativePFact);
-
-    coarseLevel->Request(mapName, MueLu::NoFactory::get());
-    coarseLevel->Request("P", tentativePFact.get(), mapTransferFactory.get());
-    coarseLevel->Request(*mapTransferFactory); // This calls DeclareInput() on mapTransferFactory
-
-    TEST_ASSERT(coarseLevel->IsRequested(mapName, MueLu::NoFactory::get()));
-    TEST_ASSERT(coarseLevel->IsRequested("P", tentativePFact.get()));
-
-    RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
-    TEST_ASSERT(!Ptent.is_null());
-
-    TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-    mapTransferFactory->Build(*fineLevel, *coarseLevel);
-    TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
-
-    RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
-    TEST_ASSERT(!coarsenedMap.is_null());
-
-    /* Manually construct a coarse version of the mapWithHoles by
-     * 1. populating a vector based on the mapWithHoles
-     * 2. manually restricting that vector to the coarse level
-     * 3. reconstructing the mapForComparison by only taking into accounts vector entries != 0
-     */
-    {
-      // Populate a fine level vector w/ ones according to fineMap
-      RCP<Vector> fullFineVec = VectorFactory::Build(Ptent->getRangeMap(), true);
-      RCP<Vector> partialFineVec = VectorFactory::Build(mapWithHoles, true);
-      partialFineVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-      RCP<Import> fineImporter = ImportFactory::Build(mapWithHoles, fullFineVec->getMap());
-      fullFineVec->doImport(*partialFineVec, *fineImporter, Xpetra::INSERT);
-
-      // Restrict to coarse level manually
-      RCP<Vector> fullCoarseVec = VectorFactory::Build(Ptent->getDomainMap(), true);
-      Ptent->apply(*fullFineVec, *fullCoarseVec, Teuchos::TRANS);
-
-      // Reconstruct coarse map for result checking
-      ArrayRCP<const Scalar> coarseVecEntries = fullCoarseVec->getData(0);
-      Array<GO> myCoarseGIDs;
-      for (LO lid = 0; lid < static_cast<LO>(coarseVecEntries.size()); ++lid) {
-        if (Teuchos::ScalarTraits<Scalar>::magnitude(coarseVecEntries[lid]) > Teuchos::ScalarTraits<MagnitudeType>::zero())
-          myCoarseGIDs.push_back(Ptent->getDomainMap()->getGlobalElement(lid));
-      }
-      GO gNumCoarseEntries = 0;
-      reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myCoarseGIDs.size()), Teuchos::outArg(gNumCoarseEntries));
-      RCP<const Map> mapForComparison = MapFactory::Build(mapWithHoles->lib(), gNumCoarseEntries, myCoarseGIDs,
-          Teuchos::ScalarTraits<GO>::zero(), comm);
-
-      TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+  RCP<Matrix> Ptent = coarseLevel->Get<RCP<Matrix>>("P", tentativePFact.get());
+  TEST_ASSERT(!Ptent.is_null());
+
+  TEST_ASSERT(!coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+  mapTransferFactory->Build(*fineLevel, *coarseLevel);
+  TEST_ASSERT(coarseLevel->IsAvailable(mapName, MueLu::NoFactory::get()));
+
+  RCP<const Map> coarsenedMap = coarseLevel->Get<RCP<const Map>>(mapName, MueLu::NoFactory::get());
+  TEST_ASSERT(!coarsenedMap.is_null());
+
+  /* Manually construct a coarse version of the mapWithHoles by
+   * 1. populating a vector based on the mapWithHoles
+   * 2. manually restricting that vector to the coarse level
+   * 3. reconstructing the mapForComparison by only taking into accounts vector entries != 0
+   */
+  {
+    // Populate a fine level vector w/ ones according to fineMap
+    RCP<Vector> fullFineVec    = VectorFactory::Build(Ptent->getRangeMap(), true);
+    RCP<Vector> partialFineVec = VectorFactory::Build(mapWithHoles, true);
+    partialFineVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+    RCP<Import> fineImporter = ImportFactory::Build(mapWithHoles, fullFineVec->getMap());
+    fullFineVec->doImport(*partialFineVec, *fineImporter, Xpetra::INSERT);
+
+    // Restrict to coarse level manually
+    RCP<Vector> fullCoarseVec = VectorFactory::Build(Ptent->getDomainMap(), true);
+    Ptent->apply(*fullFineVec, *fullCoarseVec, Teuchos::TRANS);
+
+    // Reconstruct coarse map for result checking
+    ArrayRCP<const Scalar> coarseVecEntries = fullCoarseVec->getData(0);
+    Array<GO> myCoarseGIDs;
+    for (LO lid = 0; lid < static_cast<LO>(coarseVecEntries.size()); ++lid) {
+      if (Teuchos::ScalarTraits<Scalar>::magnitude(coarseVecEntries[lid]) > Teuchos::ScalarTraits<MagnitudeType>::zero())
+        myCoarseGIDs.push_back(Ptent->getDomainMap()->getGlobalElement(lid));
     }
-  }
+    GO gNumCoarseEntries = 0;
+    reduceAll(*comm, Teuchos::REDUCE_SUM, Teuchos::as<GO>(myCoarseGIDs.size()), Teuchos::outArg(gNumCoarseEntries));
+    RCP<const Map> mapForComparison = MapFactory::Build(mapWithHoles->lib(), gNumCoarseEntries, myCoarseGIDs,
+                                                        Teuchos::ScalarTraits<GO>::zero(), comm);
 
-  #  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,Constructor,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferFullMap1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferFullMap2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferFullMap3DElasticity,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferFullMap3DElasticityReducedP,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferFullMap2DElasticityReducedP,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferPartialMap1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory,TransferPartialMap2D,Scalar,LO,GO,Node)
+    TEST_ASSERT(coarsenedMap->isSameAs(*mapForComparison));
+  }
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, Constructor, Scalar, LO, GO, Node)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferFullMap1D, Scalar, LO, GO, Node)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferFullMap2D, Scalar, LO, GO, Node)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferFullMap3DElasticity, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferFullMap3DElasticityReducedP, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferFullMap2DElasticityReducedP, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferPartialMap1D, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MapTransferFactory, TransferPartialMap2D, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
\ No newline at end of file
+}  // namespace MueLuTests
\ No newline at end of file
diff --git a/packages/muelu/test/unit_tests/Memory.cpp b/packages/muelu/test/unit_tests/Memory.cpp
index 74c44b6ca97d..02684af217c9 100644
--- a/packages/muelu/test/unit_tests/Memory.cpp
+++ b/packages/muelu/test/unit_tests/Memory.cpp
@@ -51,41 +51,38 @@
 
 namespace {
 
-  TEUCHOS_UNIT_TEST(Memory, Basic)
-  {
-    using namespace MueLu::MemUtils;
+TEUCHOS_UNIT_TEST(Memory, Basic) {
+  using namespace MueLu::MemUtils;
 
-    //    int n = 100000000;
-    int n = 10000;
+  //    int n = 100000000;
+  int n = 10000;
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    out << "Init:" << std::endl;
+  out << "Init:" << std::endl;
 
-    out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
-    out << " PrintMemoryInfo()"  << PrintMemoryInfo()  << std::endl;
+  out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
+  out << " PrintMemoryInfo()" << PrintMemoryInfo() << std::endl;
 
-    out << "Allocation of " << n << " double (sizeof(double=" << sizeof(double) << "), total=" << sizeof(double)*n/(1024*1024) << " MB)" << std::endl;
+  out << "Allocation of " << n << " double (sizeof(double=" << sizeof(double) << "), total=" << sizeof(double) * n / (1024 * 1024) << " MB)" << std::endl;
 
-    double* array = new double[n];
+  double* array = new double[n];
 
-    // Use array to be sure that it will be allocated;
-    for(int i=0; i < n; i++)
-      array[i] = i;
+  // Use array to be sure that it will be allocated;
+  for (int i = 0; i < n; i++)
+    array[i] = i;
 
-    out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
-    out << " PrintMemoryInfo()"  << PrintMemoryInfo()  << std::endl;
+  out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
+  out << " PrintMemoryInfo()" << PrintMemoryInfo() << std::endl;
 
-    out << "Delete:" << std::endl;
+  out << "Delete:" << std::endl;
 
-    delete[] array;
+  delete[] array;
 
-    out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
-    out << " PrintMemoryInfo()"  << PrintMemoryInfo()  << std::endl;
+  out << " PrintMemoryUsage()" << PrintMemoryUsage() << std::endl;
+  out << " PrintMemoryInfo()" << PrintMemoryInfo() << std::endl;
 
-    success=true;
-
-  }
-
-}//namespace <anonymous>
+  success = true;
+}
 
+}  // namespace
diff --git a/packages/muelu/test/unit_tests/MueLu_CoupledRBMFactory.cpp b/packages/muelu/test/unit_tests/MueLu_CoupledRBMFactory.cpp
index beccd15e8be7..3e4df3aec1c0 100644
--- a/packages/muelu/test/unit_tests/MueLu_CoupledRBMFactory.cpp
+++ b/packages/muelu/test/unit_tests/MueLu_CoupledRBMFactory.cpp
@@ -51,75 +51,72 @@
 
 #include <MueLu_CoupledRBMFactory.hpp>
 
-
 namespace MueLuTests {
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_CoupledRBMFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    auto A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
-    const auto map = A->getRangeMap();
-
-    MueLu::CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> coupledRBMFactory("TestName");
-    coupledRBMFactory.setNumPDEs(20);
-    coupledRBMFactory.setLastAcousticDOF(10);
-
-    {
-        MueLu::Level level;
-        level.SetLevelID(0);
-        auto nullSpace = MultiVectorFactory::Build(map, 100);
-        level.Set("TestName", nullSpace);
-        level.Keep("Nullspace", &coupledRBMFactory);
-        coupledRBMFactory.DeclareInput(level);
-
-        TEST_EQUALITY(false, level.IsAvailable("Nullspace", &coupledRBMFactory));
-        coupledRBMFactory.Build(level);
-        TEST_EQUALITY(true, level.IsAvailable("Nullspace", &coupledRBMFactory));
-    }
-
-    {
-        MueLu::Level level;
-        level.SetLevelID(0);
-        level.Set("A", A);
-        level.Keep("Nullspace", &coupledRBMFactory);
-        auto coords = MultiVectorFactory::Build(map, 100);
-        level.Set("Coordinates", coords);
-
-        TEST_EQUALITY(false, level.IsAvailable("Nullspace", &coupledRBMFactory));
-        coupledRBMFactory.Build(level);
-        TEST_EQUALITY(true, level.IsAvailable("Nullspace", &coupledRBMFactory));
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_CoupledRBMFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  auto A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
+  const auto map = A->getRangeMap();
+
+  MueLu::CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> coupledRBMFactory("TestName");
+  coupledRBMFactory.setNumPDEs(20);
+  coupledRBMFactory.setLastAcousticDOF(10);
+
+  {
+    MueLu::Level level;
+    level.SetLevelID(0);
+    auto nullSpace = MultiVectorFactory::Build(map, 100);
+    level.Set("TestName", nullSpace);
+    level.Keep("Nullspace", &coupledRBMFactory);
+    coupledRBMFactory.DeclareInput(level);
+
+    TEST_EQUALITY(false, level.IsAvailable("Nullspace", &coupledRBMFactory));
+    coupledRBMFactory.Build(level);
+    TEST_EQUALITY(true, level.IsAvailable("Nullspace", &coupledRBMFactory));
+  }
+
+  {
+    MueLu::Level level;
+    level.SetLevelID(0);
+    level.Set("A", A);
+    level.Keep("Nullspace", &coupledRBMFactory);
+    auto coords = MultiVectorFactory::Build(map, 100);
+    level.Set("Coordinates", coords);
+
+    TEST_EQUALITY(false, level.IsAvailable("Nullspace", &coupledRBMFactory));
+    coupledRBMFactory.Build(level);
+    TEST_EQUALITY(true, level.IsAvailable("Nullspace", &coupledRBMFactory));
+  }
 }
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_CoupledRBMFactory, BuildRBM, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MueLu_CoupledRBMFactory, BuildRBM, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    MueLu::CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> coupledRBMFactory(20);
-    coupledRBMFactory.setLastAcousticDOF(10);
+  MueLu::CoupledRBMFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> coupledRBMFactory(20);
+  coupledRBMFactory.setLastAcousticDOF(10);
 
-    auto A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
-    const auto map = A->getRangeMap();
+  auto A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
+  const auto map = A->getRangeMap();
 
-    auto coords = MultiVectorFactory::Build(map, 100);
-    decltype(coords) nullSpace;
+  auto coords = MultiVectorFactory::Build(map, 100);
+  decltype(coords) nullSpace;
 
-    coupledRBMFactory.BuildRBM(A, coords, nullSpace);
+  coupledRBMFactory.BuildRBM(A, coords, nullSpace);
 
-    TEST_INEQUALITY(nullSpace, Teuchos::null);
+  TEST_INEQUALITY(nullSpace, Teuchos::null);
 }
 
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_CoupledRBMFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_CoupledRBMFactory, BuildRBM, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_CoupledRBMFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MueLu_CoupledRBMFactory, BuildRBM, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/MueLu_TestHelpers.cpp b/packages/muelu/test/unit_tests/MueLu_TestHelpers.cpp
index 8e5b165a1d8c..60fffc5f1468 100644
--- a/packages/muelu/test/unit_tests/MueLu_TestHelpers.cpp
+++ b/packages/muelu/test/unit_tests/MueLu_TestHelpers.cpp
@@ -49,33 +49,33 @@
 
 namespace MueLuTests {
 
-  // static members initialization of the class TestHelpers::Parameters
-  Xpetra::Parameters TestHelpers::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
+// static members initialization of the class TestHelpers::Parameters
+Xpetra::Parameters TestHelpers::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
 
-}
+}  // namespace MueLuTests
 
 namespace MueLuTests {
-  namespace TestHelpers {
-
-    ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter) {
-      RCP<std::vector<std::string> > files = rcp(new std::vector<std::string>());
+namespace TestHelpers {
 
-      DIR *dir = opendir(dirPath.c_str());
-      TEUCHOS_TEST_FOR_EXCEPTION(dir == NULL, MueLu::Exceptions::RuntimeError, "GetFileList(\"" + dirPath + "\") : " + strerror(errno));
+ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter) {
+  RCP<std::vector<std::string> > files = rcp(new std::vector<std::string>());
 
-      struct dirent *dirEntry;
-      while ((dirEntry = readdir(dir)) != NULL) {
-        std::string dirEntryS(dirEntry->d_name);
+  DIR* dir = opendir(dirPath.c_str());
+  TEUCHOS_TEST_FOR_EXCEPTION(dir == NULL, MueLu::Exceptions::RuntimeError, "GetFileList(\"" + dirPath + "\") : " + strerror(errno));
 
-        size_t pos = dirEntryS.rfind(filter);
-        if (pos != std::string::npos && pos+filter.size() == dirEntryS.size())
-          files->push_back(std::string(dirEntryS));
-      }
+  struct dirent* dirEntry;
+  while ((dirEntry = readdir(dir)) != NULL) {
+    std::string dirEntryS(dirEntry->d_name);
 
-      closedir(dir);
+    size_t pos = dirEntryS.rfind(filter);
+    if (pos != std::string::npos && pos + filter.size() == dirEntryS.size())
+      files->push_back(std::string(dirEntryS));
+  }
 
-      return arcp(files);
-    }
+  closedir(dir);
 
-  }
+  return arcp(files);
 }
+
+}  // namespace TestHelpers
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/MueLu_TestHelpers.hpp b/packages/muelu/test/unit_tests/MueLu_TestHelpers.hpp
index bc24dcd4783e..8ad82912ebf3 100644
--- a/packages/muelu/test/unit_tests/MueLu_TestHelpers.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_TestHelpers.hpp
@@ -45,7 +45,7 @@
 // @HEADER
 #ifndef MUELU_TEST_HELPERS_H
 #define MUELU_TEST_HELPERS_H
-#include <stdio.h> //DEBUG
+#include <stdio.h>  //DEBUG
 #include <string>
 #include <set>
 #ifndef _MSC_VER
@@ -85,16 +85,16 @@
 #include "Galeri_XpetraMatrixTypes.hpp"
 
 namespace Galeri {
-  namespace Xpetra {
-    template <class LocalOrdinal, class GlobalOrdinal, class Map>
-    RCP<Map> CreateMap(const std::string & mapType, const Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList & list);
+namespace Xpetra {
+template <class LocalOrdinal, class GlobalOrdinal, class Map>
+RCP<Map> CreateMap(const std::string& mapType, const Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList& list);
 
 #ifdef HAVE_GALERI_XPETRA
-    template <class LocalOrdinal, class GlobalOrdinal, class Node>
-    Teuchos::RCP< ::Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(::Xpetra::UnderlyingLib lib, const std::string & mapType, const Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList & list);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP< ::Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(::Xpetra::UnderlyingLib lib, const std::string& mapType, const Teuchos::RCP<const Teuchos::Comm<int> >& comm, Teuchos::ParameterList& list);
 #endif
-  }
-}
+}  // namespace Xpetra
+}  // namespace Galeri
 
 #include "MueLu_NoFactory.hpp"
 
@@ -110,873 +110,846 @@ namespace Galeri {
 #include <MueLu_TestHelpers_Common.hpp>
 
 namespace MueLuTests {
-  using Teuchos::RCP;
-  using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
-  using Teuchos::ArrayView;
-  using Teuchos::arcp;
-  using Teuchos::arcpFromArrayView;
-  using Teuchos::rcpFromRef;
-  using Teuchos::null;
-  using Teuchos::arcp_reinterpret_cast;
-  using Teuchos::Array;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcp_implicit_cast;
-  using Teuchos::rcpFromRef;
-
-  namespace TestHelpers {
-
-    using Xpetra::global_size_t;
-
-    class Parameters {
-
-    private:
-      Parameters() {} // static class
+using Teuchos::arcp;
+using Teuchos::arcp_reinterpret_cast;
+using Teuchos::arcpFromArrayView;
+using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::ArrayView;
+using Teuchos::null;
+using Teuchos::RCP;
+using Teuchos::rcp;
+using Teuchos::rcp_dynamic_cast;
+using Teuchos::rcp_implicit_cast;
+using Teuchos::rcpFromRef;
+
+namespace TestHelpers {
+
+using Xpetra::global_size_t;
+
+class Parameters {
+ private:
+  Parameters() {}  // static class
+
+ public:
+  static Xpetra::Parameters xpetraParameters;
+
+  inline static RCP<const Teuchos::Comm<int> > getDefaultComm() {
+    return Xpetra::DefaultPlatform::getDefaultPlatform().getComm();
+  }
 
-    public:
+  inline static Xpetra::UnderlyingLib getLib() {
+    return TestHelpers::Parameters::xpetraParameters.GetLib();
+  }
+};
 
-      static Xpetra::Parameters xpetraParameters;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TestFactory {
+#include "MueLu_UseShortNames.hpp"
 
-      inline static RCP<const Teuchos::Comm<int> > getDefaultComm() {
-        return Xpetra::DefaultPlatform::getDefaultPlatform().getComm();
+ private:
+  TestFactory() {}  // static class
+
+ public:
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVectorFactory<real_type, LO, GO, NO> RealValuedMultiVectorFactory;
+
+  //
+  // Method that creates a map containing a specified number of local elements per process.
+  //
+  static const RCP<const Map> BuildMap(LO numElementsPerProc) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+    const global_size_t INVALID = Teuchos::OrdinalTraits<global_size_t>::invalid();
+
+    return MapFactory::Build(TestHelpers::Parameters::getLib(), INVALID, numElementsPerProc, 0, comm);
+
+  }  // BuildMap()
+
+  // Create a matrix as specified by parameter list options
+  static RCP<Matrix> BuildMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers::Parameters::getLib();
+
+    GO nx, ny, nz;
+    nx = ny = nz = 5;
+    nx           = matrixList.get("nx", nx);
+    ny           = matrixList.get("ny", ny);
+    nz           = matrixList.get("nz", nz);
+
+    std::string matrixType = matrixList.get("matrixType", "Laplace1D");
+    GO numGlobalElements;  // global_size_t
+    if (matrixType == "Laplace1D")
+      numGlobalElements = nx;
+    else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "Cross2D")
+      numGlobalElements = nx * ny;
+    else if (matrixType == "Elasticity2D")
+      numGlobalElements = 2 * nx * ny;
+    else if (matrixType == "Laplace3D" || matrixType == "Brick3D")
+      numGlobalElements = nx * ny * nz;
+    else if (matrixType == "Elasticity3D")
+      numGlobalElements = 3 * nx * ny * nz;
+    else {
+      std::string msg = matrixType + " is unsupported (in unit testing)";
+      throw(MueLu::Exceptions::RuntimeError(msg));
+    }
+
+    RCP<const Map> map = MapFactory::Build(lib, numGlobalElements, 0, comm);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, matrixList);
+    RCP<Matrix> Op = Pr->BuildMatrix();
+
+    return Op;
+  }  // BuildMatrix()
+
+  // Create a tridiagonal matrix (stencil = [b,a,c]) with the specified number of rows
+  // dofMap: row map of matrix
+  static RCP<Matrix> BuildTridiag(RCP<const Map> dofMap, Scalar a, Scalar b, Scalar c, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {  // global_size_t
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers::Parameters::getLib();
+
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    Teuchos::RCP<Matrix> mtx = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(dofMap, 3);
+
+    LocalOrdinal NumMyElements                               = dofMap->getLocalNumElements();
+    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = dofMap->getLocalElementList();
+    GlobalOrdinal indexBase                                  = dofMap->getIndexBase();
+
+    GlobalOrdinal NumEntries;
+    LocalOrdinal nnz = 3;
+    std::vector<Scalar> Values(nnz);
+    std::vector<GlobalOrdinal> Indices(nnz);
+
+    // access information from strided block map
+    std::vector<size_t> strInfo = std::vector<size_t>();
+    size_t blockSize            = 1;
+    LocalOrdinal blockId        = -1;
+    GlobalOrdinal offset        = 0;
+
+    Teuchos::RCP<const StridedMap> strdofMap = Teuchos::rcp_dynamic_cast<const StridedMap>(dofMap);
+    if (strdofMap != Teuchos::null) {
+      strInfo   = strdofMap->getStridingData();
+      blockSize = strdofMap->getFixedBlockSize();
+      blockId   = strdofMap->getStridedBlockId();
+      offset    = strdofMap->getOffset();
+      TEUCHOS_TEST_FOR_EXCEPTION(blockId > -1 && strInfo[blockId] == 1, MueLu::Exceptions::RuntimeError,
+                                 "MueLu::TestHelpers::BuildTridiag: strInfo block size must be > 1.");
+      // todo: write one more special case for a row being first and last bock row
+    } else {
+      // no striding information. emulate block matrix
+      blockId = 0;
+      strInfo.push_back(blockSize);  // default block size = 1
+    }
+
+    GlobalOrdinal rrmax = (dofMap->getMaxAllGlobalIndex() - offset - indexBase) / blockSize;
+
+    // loop over all rows
+    for (LocalOrdinal i = 0; i < NumMyElements; i++) {
+      GlobalOrdinal rr = (MyGlobalElements[i] - offset - indexBase) / blockSize + indexBase;  // node index
+
+      // distinguish 5 different cases
+
+      GlobalOrdinal blockOffset = 0;
+      for (LocalOrdinal k = 0; k < blockId; k++)
+        blockOffset += Teuchos::as<GlobalOrdinal>(strInfo[k]);
+
+      if (MyGlobalElements[i] == blockOffset + offset + indexBase) {
+        // very first row
+        Indices[0] = MyGlobalElements[i];
+        Values[0]  = a;
+        Indices[1] = MyGlobalElements[i] + 1;
+        Values[1]  = c;
+        NumEntries = 2;
+      } else if (MyGlobalElements[i] == rrmax * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
+        // very last row
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        NumEntries = 2;
+      } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
+        // last row in current node block
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = (rr + 1) * blockSize + blockOffset + offset + indexBase;
+        Values[2]  = c;
+        NumEntries = 3;
+      } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + offset + indexBase) {
+        // first row in current node block
+        Indices[0] = (rr - 1) * blockSize + blockOffset + strInfo[blockId] - 1 + offset + indexBase;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = MyGlobalElements[i] + 1;
+        Values[2]  = c;
+        NumEntries = 3;
+      } else {
+        // usual row entries in block rows
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = MyGlobalElements[i] + 1;
+        Values[2]  = c;
+        NumEntries = 3;
       }
 
-      inline static Xpetra::UnderlyingLib getLib() {
-        return TestHelpers::Parameters::xpetraParameters.GetLib();
+      // debug output
+      /*std::cout << "--------> Proc " << comm->getRank() << " lrow " << i << " grow " << MyGlobalElements[i] << " node " << rr << " values = " ;
+      for (size_t k = 0; k < NumEntries; k++) {
+        std::cout << " " << Indices[k] << "->" << Values[k] << "   ";
       }
-    };
-
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class TestFactory {
-#include "MueLu_UseShortNames.hpp"
-
-    private:
-      TestFactory() {} // static class
-
-    public:
-
-      typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-      typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-      typedef Xpetra::MultiVectorFactory<real_type,LO,GO,NO> RealValuedMultiVectorFactory;
-
-      //
-      // Method that creates a map containing a specified number of local elements per process.
-      //
-      static const RCP<const Map> BuildMap(LO numElementsPerProc) {
-
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-        const global_size_t INVALID = Teuchos::OrdinalTraits<global_size_t>::invalid();
-
-        return MapFactory::Build(TestHelpers::Parameters::getLib(), INVALID, numElementsPerProc, 0, comm);
-
-      } // BuildMap()
-
-      // Create a matrix as specified by parameter list options
-      static RCP<Matrix> BuildMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) {
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-        if (lib == Xpetra::NotSpecified)
-          lib = TestHelpers::Parameters::getLib();
-
-        GO nx,ny,nz;
-        nx = ny = nz = 5;
-        nx = matrixList.get("nx",nx);
-        ny = matrixList.get("ny",ny);
-        nz = matrixList.get("nz",nz);
-
-        std::string matrixType = matrixList.get("matrixType","Laplace1D");
-        GO numGlobalElements; //global_size_t
-        if (matrixType == "Laplace1D")
-          numGlobalElements = nx;
-        else if (matrixType == "Laplace2D" || matrixType == "Star2D" || matrixType == "Cross2D")
-          numGlobalElements = nx*ny;
-        else if(matrixType == "Elasticity2D")
-          numGlobalElements = 2*nx*ny;
-        else if (matrixType == "Laplace3D" || matrixType == "Brick3D")
-          numGlobalElements = nx*ny*nz;
-        else if  (matrixType == "Elasticity3D")
-          numGlobalElements = 3*nx*ny*nz;
-        else {
-          std::string msg = matrixType + " is unsupported (in unit testing)";
-          throw(MueLu::Exceptions::RuntimeError(msg));
+      std::cout << std::endl;*/
+
+      // put the off-diagonal entries
+      // Xpetra wants ArrayViews (sigh)
+      Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+    }
+
+    mtx->fillComplete();
+
+    return mtx;
+  }  // BuildTridiag()
+
+  // Create a 1D Poisson matrix with the specified number of rows
+  // nx: global number of rows
+  static RCP<Matrix> Build1DPoisson(GO nx, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {  // global_size_t
+    Teuchos::ParameterList matrixList;
+    matrixList.set("nx", nx);
+    matrixList.set("matrixType", "Laplace1D");
+    RCP<Matrix> A = BuildMatrix(matrixList, lib);
+    return A;
+  }  // Build1DPoisson()
+
+  // Create a 2D Poisson matrix with the specified number of rows
+  // nx: global number of rows
+  // ny: global number of rows
+  static RCP<Matrix> Build2DPoisson(GO nx, GO ny = -1, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {  // global_size_t
+    Teuchos::ParameterList matrixList;
+    if (ny == -1) ny = nx;
+    matrixList.set("nx", nx);
+    matrixList.set("ny", ny);
+    matrixList.set("matrixType", "Laplace2D");
+    RCP<Matrix> A = BuildMatrix(matrixList, lib);
+    return A;
+  }  // Build2DPoisson()
+
+  static RCP<RealValuedMultiVector>
+  BuildGeoCoordinates(const int numDimensions, const Array<GO> gNodesPerDir,
+                      Array<LO>& lNodesPerDir, Array<GO>& meshData,
+                      const std::string meshLayout = "Global Lexicographic") {
+    // Get MPI infos
+    Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    LO numRanks                         = comm->getSize();
+    LO myRank                           = comm->getRank();
+
+    meshData.resize(10 * numRanks);
+
+    ////////////////////////////////////
+    //                                //
+    //   Step 1: Compute map layout   //
+    //                                //
+    ////////////////////////////////////
+    if (numRanks == 1) {
+      if (meshLayout == "Local Lexicographic") {
+        meshData[0] = 0;  // Local Proc
+        meshData[1] = 0;  // Local Proc?
+        meshData[2] = 0;  // Parent Proc
+      }
+      for (int dim = 0; dim < 3; ++dim) {
+        lNodesPerDir[dim] = gNodesPerDir[dim];
+        if (meshLayout == "Local Lexicographic") {
+          meshData[3 + 2 * dim]     = 0;                      // Lowest  index in direction dim
+          meshData[3 + 2 * dim + 1] = gNodesPerDir[dim] - 1;  // Highest index in direction dim
         }
-
-        RCP<const Map> map = MapFactory::Build(lib, numGlobalElements, 0, comm);
-        RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, matrixList);
-        RCP<Matrix> Op = Pr->BuildMatrix();
-
-        return Op;
-      } // BuildMatrix()
-
-      // Create a tridiagonal matrix (stencil = [b,a,c]) with the specified number of rows
-      // dofMap: row map of matrix
-      static RCP<Matrix> BuildTridiag(RCP<const Map> dofMap, Scalar a, Scalar b, Scalar c, Xpetra::UnderlyingLib lib=Xpetra::NotSpecified) { //global_size_t
-
-        if (lib == Xpetra::NotSpecified)
-          lib = TestHelpers::Parameters::getLib();
-
-        RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-        Teuchos::RCP<Matrix> mtx = Xpetra::MatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(dofMap, 3);
-
-        LocalOrdinal NumMyElements = dofMap->getLocalNumElements();
-        Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = dofMap->getLocalElementList();
-        GlobalOrdinal indexBase = dofMap->getIndexBase();
-
-        GlobalOrdinal NumEntries;
-        LocalOrdinal  nnz = 3;
-        std::vector<Scalar>        Values(nnz);
-        std::vector<GlobalOrdinal> Indices(nnz);
-
-        // access information from strided block map
-        std::vector<size_t> strInfo = std::vector<size_t>();
-        size_t blockSize = 1;
-        LocalOrdinal blockId = -1;
-        GlobalOrdinal offset = 0;
-
-        Teuchos::RCP<const StridedMap> strdofMap = Teuchos::rcp_dynamic_cast<const StridedMap>(dofMap);
-        if(strdofMap != Teuchos::null) {
-          strInfo = strdofMap->getStridingData();
-          blockSize = strdofMap->getFixedBlockSize();
-          blockId = strdofMap->getStridedBlockId();
-          offset = strdofMap->getOffset();
-          TEUCHOS_TEST_FOR_EXCEPTION(blockId > -1 && strInfo[blockId]==1, MueLu::Exceptions::RuntimeError,
-                                             "MueLu::TestHelpers::BuildTridiag: strInfo block size must be > 1.");
-          // todo: write one more special case for a row being first and last bock row
-        } else {
-          // no striding information. emulate block matrix
-          blockId = 0;
-          strInfo.push_back(blockSize); // default block size = 1
+      }
+      if (meshLayout == "Local Lexicographic") {
+        meshData[9] = 0;
+      }
+    } else if (numRanks == 4) {
+      if (meshLayout == "Local Lexicographic") {
+        // First fill the rank related info in meshData
+        for (int rank = 0; rank < numRanks; ++rank) {
+          meshData[10 * rank + 0] = rank;  // Local Proc
+          meshData[10 * rank + 1] = rank;  // Local Proc?
+          meshData[10 * rank + 2] = 0;     // Parent Proc
         }
+      }
 
-        GlobalOrdinal rrmax = (dofMap->getMaxAllGlobalIndex()-offset-indexBase) / blockSize;
-
-        // loop over all rows
-        for (LocalOrdinal i = 0; i < NumMyElements; i++) {
-
-          GlobalOrdinal rr = (MyGlobalElements[i]-offset-indexBase) / blockSize + indexBase;  // node index
-
-          // distinguish 5 different cases
-
-          GlobalOrdinal blockOffset = 0;
-          for (LocalOrdinal k=0; k<blockId; k++)
-            blockOffset += Teuchos::as<GlobalOrdinal>(strInfo[k]);
-
-          if (MyGlobalElements[i] == blockOffset + offset + indexBase) {
-            // very first row
-            Indices[0] = MyGlobalElements[i];
-            Values [0] = a;
-            Indices[1] = MyGlobalElements[i] + 1;
-            Values [1] = c;
-            NumEntries = 2;
-          } else if (MyGlobalElements[i] == rrmax * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
-            // very last row
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            NumEntries = 2;
-          } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
-            // last row in current node block
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = (rr+1)*blockSize + blockOffset + offset + indexBase;
-            Values [2] = c;
-            NumEntries = 3;
-          } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + offset + indexBase) {
-            // first row in current node block
-            Indices[0] = (rr-1)*blockSize + blockOffset + strInfo[blockId] - 1 + offset + indexBase;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = MyGlobalElements[i] + 1;
-            Values [2] = c;
-            NumEntries = 3;
-          } else {
-            // usual row entries in block rows
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = MyGlobalElements[i] + 1;
-            Values [2] = c;
-            NumEntries = 3;
-          }
-
-          // debug output
-          /*std::cout << "--------> Proc " << comm->getRank() << " lrow " << i << " grow " << MyGlobalElements[i] << " node " << rr << " values = " ;
-          for (size_t k = 0; k < NumEntries; k++) {
-            std::cout << " " << Indices[k] << "->" << Values[k] << "   ";
-          }
-          std::cout << std::endl;*/
-
-          // put the off-diagonal entries
-          // Xpetra wants ArrayViews (sigh)
-          Teuchos::ArrayView<Scalar>        av(&Values [0], NumEntries);
-          Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
-          mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+      if (numDimensions == 1) {
+        LO numNodesPerRank = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
+        if (myRank == numRanks - 1) {
+          lNodesPerDir[0] = gNodesPerDir[0] - myRank * numNodesPerRank;
+        } else {
+          lNodesPerDir[0] = numNodesPerRank;
         }
-
-        mtx->fillComplete();
-
-        return mtx;
-      } // BuildTridiag()
-
-      // Create a 1D Poisson matrix with the specified number of rows
-      // nx: global number of rows
-      static RCP<Matrix> Build1DPoisson(GO nx, Xpetra::UnderlyingLib lib=Xpetra::NotSpecified) { //global_size_t
-        Teuchos::ParameterList matrixList;
-        matrixList.set("nx", nx);
-        matrixList.set("matrixType","Laplace1D");
-        RCP<Matrix> A = BuildMatrix(matrixList,lib);
-        return A;
-      } // Build1DPoisson()
-
-      // Create a 2D Poisson matrix with the specified number of rows
-      // nx: global number of rows
-      // ny: global number of rows
-      static RCP<Matrix> Build2DPoisson(GO nx, GO ny=-1, Xpetra::UnderlyingLib lib=Xpetra::NotSpecified) { //global_size_t
-        Teuchos::ParameterList matrixList;
-        if (ny==-1) ny=nx;
-        matrixList.set("nx", nx);
-        matrixList.set("ny", ny);
-        matrixList.set("matrixType","Laplace2D");
-        RCP<Matrix> A = BuildMatrix(matrixList,lib);
-        return A;
-      } // Build2DPoisson()
-
-      static RCP<RealValuedMultiVector>
-      BuildGeoCoordinates(const int numDimensions, const Array<GO> gNodesPerDir,
-                          Array<LO>& lNodesPerDir, Array<GO>& meshData,
-                          const std::string meshLayout = "Global Lexicographic") {
-
-        // Get MPI infos
-        Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-        RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-        LO numRanks = comm->getSize();
-        LO myRank   = comm->getRank();
-
-        meshData.resize(10*numRanks);
-
-        ////////////////////////////////////
-        //                                //
-        //   Step 1: Compute map layout   //
-        //                                //
-        ////////////////////////////////////
-        if(numRanks == 1) {
-          if(meshLayout == "Local Lexicographic") {
-            meshData[0] = 0; // Local Proc
-            meshData[1] = 0; // Local Proc?
-            meshData[2] = 0; // Parent Proc
-          }
-          for(int dim = 0; dim < 3; ++dim) {
-            lNodesPerDir[dim] = gNodesPerDir[dim];
-            if(meshLayout == "Local Lexicographic") {
-              meshData[3 + 2*dim]    = 0;                        // Lowest  index in direction dim
-              meshData[3 + 2*dim +1] = gNodesPerDir[dim] - 1;    // Highest index in direction dim
-            }
-          }
-          if(meshLayout == "Local Lexicographic") {
-            meshData[9] = 0;
-          }
-        } else if(numRanks == 4) {
-          if(meshLayout == "Local Lexicographic") {
-            // First fill the rank related info in meshData
-            for(int rank = 0; rank < numRanks; ++rank) {
-              meshData[10*rank + 0] = rank; // Local Proc
-              meshData[10*rank + 1] = rank; // Local Proc?
-              meshData[10*rank + 2] = 0;        // Parent Proc
-            }
-          }
-
-          if(numDimensions == 1) {
-            LO numNodesPerRank = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
-            if(myRank == numRanks - 1) {
-              lNodesPerDir[0] = gNodesPerDir[0] - myRank*numNodesPerRank;
-            } else {
-              lNodesPerDir[0] = numNodesPerRank;
-            }
-            lNodesPerDir[1] = gNodesPerDir[1];
-
-            if(meshLayout == "Local Lexicographic") {
-              for(int rank = 0; rank < numRanks; ++rank) {
-                meshData[10*rank + 3] = rank*numNodesPerRank;
-                meshData[10*rank + 4] =
-                  (rank == numRanks - 1) ? (gNodesPerDir[0] - 1) : ((rank + 1)*numNodesPerRank - 1);
-                meshData[10*rank + 5] = 0;
-                meshData[10*rank + 6] = 0;
-              }
-            }
-          } else {
-            if(myRank == 0 || myRank == 2) {
-              lNodesPerDir[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            } else {
-              lNodesPerDir[0] = std::floor(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            }
-            if(myRank == 0 || myRank == 1) {
-              lNodesPerDir[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            } else {
-              lNodesPerDir[1] = std::floor(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            }
-            if(meshLayout == "Local Lexicographic") {
-              for(int j = 0; j < 2; ++j) {
-                for(int i = 0; i < 2; ++i) {
-                  int rank = 2*j + i;
-                  if(i == 0) {
-                    meshData[10*rank + 3] = 0;
-                    meshData[10*rank + 4] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2) - 1;
-                  } else if(i == 1) {
-                    meshData[10*rank + 3] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-                    meshData[10*rank + 4] = gNodesPerDir[0] - 1;
-                  }
-                  if(j == 0) {
-                    meshData[10*rank + 5] = 0;
-                    meshData[10*rank + 6] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) - 1;
-                  } else if(j == 1) {
-                    meshData[10*rank + 5] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-                    meshData[10*rank + 6] = gNodesPerDir[1] - 1;
-                  }
-                }
-              }
-            }
-          }
-          lNodesPerDir[2] = gNodesPerDir[2];
-          if(meshLayout == "Local Lexicographic") {
-            for(int rank = 0; rank < numRanks; ++rank) {
-              meshData[10*rank + 7] = 0;
-              meshData[10*rank + 8] = gNodesPerDir[2] - 1;
-            }
-            meshData[9] = 0;
-            for(int rank = 1; rank < numRanks; ++rank) {
-              meshData[10*rank + 9] = meshData[10*(rank - 1) + 9]
-                + (meshData[10*(rank - 1) + 4] - meshData[10*(rank - 1) + 3] + 1)
-                *(meshData[10*(rank - 1) + 6] - meshData[10*(rank - 1) + 5] + 1)
-                *(meshData[10*(rank - 1) + 8] - meshData[10*(rank - 1) + 7] + 1);
-            }
+        lNodesPerDir[1] = gNodesPerDir[1];
+
+        if (meshLayout == "Local Lexicographic") {
+          for (int rank = 0; rank < numRanks; ++rank) {
+            meshData[10 * rank + 3] = rank * numNodesPerRank;
+            meshData[10 * rank + 4] =
+                (rank == numRanks - 1) ? (gNodesPerDir[0] - 1) : ((rank + 1) * numNodesPerRank - 1);
+            meshData[10 * rank + 5] = 0;
+            meshData[10 * rank + 6] = 0;
           }
         }
-
-        GO gNumNodes = 1;
-        LO lNumNodes = 1;
-        for(int dim = 0; dim < 3; ++dim) {
-          gNumNodes = gNumNodes*gNodesPerDir[dim];
-          lNumNodes = lNumNodes*lNodesPerDir[dim];
+      } else {
+        if (myRank == 0 || myRank == 2) {
+          lNodesPerDir[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        } else {
+          lNodesPerDir[0] = std::floor(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
         }
-
-        GO myGIDOffset = 0;
-        Array<GO> myLowestTuple(3);
-        if(numDimensions == 1) {
-          myGIDOffset = myRank*std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
-          myLowestTuple[0] = myGIDOffset;
+        if (myRank == 0 || myRank == 1) {
+          lNodesPerDir[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
         } else {
-          if(myRank == 1) {
-            myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = myLowestTuple[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = myLowestTuple[0]*std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2)
-                *gNodesPerDir[2];
-            }
-          } else if(myRank == 2) {
-            myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2)*gNodesPerDir[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = myLowestTuple[1]*gNodesPerDir[0]*gNodesPerDir[2];
-            }
-          } else if(myRank == 3) {
-            myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = myLowestTuple[1]*gNodesPerDir[0] + myLowestTuple[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = (myLowestTuple[0]*(gNodesPerDir[1] - myLowestTuple[1])
-                             + myLowestTuple[1]*gNodesPerDir[0])*gNodesPerDir[2];
-            }
-          }
+          lNodesPerDir[1] = std::floor(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
         }
-
-        ////////////////////////////////////
-        //                                //
-        //    Step 2: Compute map GIDs    //
-        //                                //
-        ////////////////////////////////////
-        Array<GO> myGIDs(lNumNodes);
-        if(meshLayout == "Global Lexicographic") {
-          for(LO k = 0; k < lNodesPerDir[2]; ++k) {
-            for(LO j = 0; j < lNodesPerDir[1]; ++j) {
-              for(LO i = 0; i < lNodesPerDir[0]; ++i) {
-                myGIDs[k*lNodesPerDir[1]*lNodesPerDir[0] + j*lNodesPerDir[0] + i] =
-                  myGIDOffset + k*gNodesPerDir[1]*gNodesPerDir[0] + j*gNodesPerDir[0] + i;
+        if (meshLayout == "Local Lexicographic") {
+          for (int j = 0; j < 2; ++j) {
+            for (int i = 0; i < 2; ++i) {
+              int rank = 2 * j + i;
+              if (i == 0) {
+                meshData[10 * rank + 3] = 0;
+                meshData[10 * rank + 4] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2) - 1;
+              } else if (i == 1) {
+                meshData[10 * rank + 3] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+                meshData[10 * rank + 4] = gNodesPerDir[0] - 1;
               }
-            }
-          }
-        } else if(meshLayout == "Local Lexicographic") {
-          for(LO nodeIdx = 0; nodeIdx < lNumNodes; ++nodeIdx) {
-            myGIDs[nodeIdx] = myGIDOffset + nodeIdx;
-          }
-        }
-
-        RCP<const Map> coordMap = MapFactory::Build(lib, gNumNodes, myGIDs(), 0, comm);
-
-
-        ///////////////////////////////////////
-        //                                   //
-        //    Step 3: Compute coordinates    //
-        //                                   //
-        ///////////////////////////////////////
-        RCP<RealValuedMultiVector> Coordinates = RealValuedMultiVectorFactory::Build(coordMap,
-                                                                                     numDimensions);
-        Array<ArrayRCP<real_type> > myCoords(numDimensions);
-        for(int dim = 0; dim < numDimensions; ++dim) {
-          myCoords[dim] = Coordinates->getDataNonConst(dim);
-        }
-
-        LO nodeIdx = 0;
-        Array<LO> ijk(3);
-        for(ijk[2] = 0; ijk[2] < lNodesPerDir[2]; ++ijk[2]) {
-          for(ijk[1] = 0; ijk[1] < lNodesPerDir[1]; ++ijk[1]) {
-            for(ijk[0] = 0; ijk[0] < lNodesPerDir[0]; ++ijk[0]) {
-              nodeIdx = ijk[2]*lNodesPerDir[1]*lNodesPerDir[0] + ijk[1]*lNodesPerDir[0] + ijk[0];
-              for(int dim = 0; dim < numDimensions; ++dim) {
-                 if(gNodesPerDir[dim] == 1) {
-                  myCoords[dim][nodeIdx] = 0.0;
-                } else {
-                  myCoords[dim][nodeIdx] =
-                    Teuchos::as<real_type>(myLowestTuple[dim] + ijk[dim]) / (gNodesPerDir[dim] - 1);
-                }
+              if (j == 0) {
+                meshData[10 * rank + 5] = 0;
+                meshData[10 * rank + 6] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) - 1;
+              } else if (j == 1) {
+                meshData[10 * rank + 5] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+                meshData[10 * rank + 6] = gNodesPerDir[1] - 1;
               }
             }
           }
         }
-
-        return Coordinates;
-      } // BuildGeoCoordinates
-
-      // Xpetra version of CreateMap
-      static RCP<Map> BuildMap(Xpetra::UnderlyingLib lib, const std::set<GlobalOrdinal>& gids, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
-        Teuchos::Array<GlobalOrdinal> mapvec;
-        mapvec.reserve(gids.size());
-        mapvec.assign(gids.begin(), gids.end());
-        GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
-        GlobalOrdinal gcount;
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
-
-        Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
-            MapFactory::Build(lib,gcount,mapvec(),0,comm);
-        mapvec.clear();
-        return map;
       }
-
-      // Xpetra version of SplitMap
-      static Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(Xpetra::UnderlyingLib lib, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Agiven) {
-        Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
-
-        GlobalOrdinal count=0;
-        Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
-        for (size_t i=0; i<Amap.getLocalNumElements(); ++i) {
-          const GlobalOrdinal gid = Amap.getGlobalElement(i);
-          if (Agiven.isNodeGlobalElement(gid)) continue;
-          myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
-          ++count;
+      lNodesPerDir[2] = gNodesPerDir[2];
+      if (meshLayout == "Local Lexicographic") {
+        for (int rank = 0; rank < numRanks; ++rank) {
+          meshData[10 * rank + 7] = 0;
+          meshData[10 * rank + 8] = gNodesPerDir[2] - 1;
+        }
+        meshData[9] = 0;
+        for (int rank = 1; rank < numRanks; ++rank) {
+          meshData[10 * rank + 9] = meshData[10 * (rank - 1) + 9] + (meshData[10 * (rank - 1) + 4] - meshData[10 * (rank - 1) + 3] + 1) * (meshData[10 * (rank - 1) + 6] - meshData[10 * (rank - 1) + 5] + 1) * (meshData[10 * (rank - 1) + 8] - meshData[10 * (rank - 1) + 7] + 1);
         }
-        myaugids.resize(count);
-        GlobalOrdinal gcount;
-        Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
-        return MapFactory::Build(lib,gcount,myaugids(),0,comm);
       }
-
-      static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrix(Xpetra::UnderlyingLib lib, int noBlocks, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
-        GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,noBlocks-2)) * 10;
-
-        GlobalOrdinal procOffset = comm->getRank() * nOverallDOFGidsPerProc;
-
-        std::set<GlobalOrdinal> myDOFGids;
-        for(GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
-          myDOFGids.insert(i + procOffset);
-
-        Teuchos::RCP<Map> fullmap = TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildMap(lib, myDOFGids, comm);
-
-        std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-        GlobalOrdinal nPartGIDs = nOverallDOFGidsPerProc;
-        Teuchos::RCP<Map> remainingpartmap = fullmap;
-        for (int it=0; it<noBlocks; it++) {
-          if(it == noBlocks - 1) {
-            maps[0] = remainingpartmap;
-            break;
-          }
-          // collect first half of GIDs
-          nPartGIDs = nPartGIDs / 2;
-          std::set<GlobalOrdinal> myHalfGIDs;
-          for(GlobalOrdinal j = 0; j < nPartGIDs; j++)
-            myHalfGIDs.insert(j + procOffset);
-
-          Teuchos::RCP<Map> halfmap = TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::BuildMap(lib, myHalfGIDs, comm);
-
-          Teuchos::RCP<Map> secondmap = TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::SplitMap(lib,*remainingpartmap, *halfmap);
-          remainingpartmap = halfmap;
-
-          maps[noBlocks - 1 - it]  = secondmap;
+    }
+
+    GO gNumNodes = 1;
+    LO lNumNodes = 1;
+    for (int dim = 0; dim < 3; ++dim) {
+      gNumNodes = gNumNodes * gNodesPerDir[dim];
+      lNumNodes = lNumNodes * lNodesPerDir[dim];
+    }
+
+    GO myGIDOffset = 0;
+    Array<GO> myLowestTuple(3);
+    if (numDimensions == 1) {
+      myGIDOffset      = myRank * std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
+      myLowestTuple[0] = myGIDOffset;
+    } else {
+      if (myRank == 1) {
+        myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = myLowestTuple[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = myLowestTuple[0] * std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) * gNodesPerDir[2];
         }
-
-        // create diagonal blocks
-        std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-        for (int it=0; it<noBlocks; it++) {
-          blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
-
-          LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-          Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
-
-          for (LocalOrdinal i = 0; i < NumMyElements; i++)
-            blocks[it]->insertGlobalValues(MyGlobalElements[i],
-                Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-                Teuchos::tuple<Scalar>(it+1));
-          blocks[it]->fillComplete();
+      } else if (myRank == 2) {
+        myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) * gNodesPerDir[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = myLowestTuple[1] * gNodesPerDir[0] * gNodesPerDir[2];
         }
-
-        // create map extractor
-        Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
-        Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
-
-        // build blocked operator
-        Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
-
-        for (int it=0; it<noBlocks; it++) {
-          Teuchos::RCP<CrsMatrixWrap> csrwrap =
-              Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-          bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
+      } else if (myRank == 3) {
+        myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = myLowestTuple[1] * gNodesPerDir[0] + myLowestTuple[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = (myLowestTuple[0] * (gNodesPerDir[1] - myLowestTuple[1]) + myLowestTuple[1] * gNodesPerDir[0]) * gNodesPerDir[2];
         }
-        bop->fillComplete();
-        return bop;
       }
-
-      static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrixThyra(Xpetra::UnderlyingLib lib, int noBlocks, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
-
-        std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-
-        maps[0] = MapFactory::Build (lib, comm->getSize() * 5, 5, 0, comm);
-        for (int it=1; it<noBlocks; it++) {
-          GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,it-1)*5);
-          maps[it]  = MapFactory::Build (lib, comm->getSize() * localDofs, localDofs, 0, comm);
-        }
-
-        // create diagonal blocks
-        std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-        for (int it=0; it<noBlocks; it++) {
-          blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
-
-          LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-          Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
-
-          for (LocalOrdinal i = 0; i < NumMyElements; i++)
-            blocks[it]->insertGlobalValues(MyGlobalElements[i],
-                Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-                Teuchos::tuple<Scalar>(it+1));
-          blocks[it]->fillComplete();
-        }
-
-        // create map extractor
-        // To generate the Thyra style map extractor we do not need a full map but only the
-        // information about the Map details (i.e. lib and indexBase). We can extract this
-        // information from maps[0]
-        Teuchos::RCP<const MapExtractor > rgMapExtractor =
-            Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-        Teuchos::RCP<const MapExtractor > doMapExtractor =
-            Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-        // build blocked operator
-        Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
-
-        for (int it=0; it<noBlocks; it++) {
-          Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-          bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
+    }
+
+    ////////////////////////////////////
+    //                                //
+    //    Step 2: Compute map GIDs    //
+    //                                //
+    ////////////////////////////////////
+    Array<GO> myGIDs(lNumNodes);
+    if (meshLayout == "Global Lexicographic") {
+      for (LO k = 0; k < lNodesPerDir[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDir[1]; ++j) {
+          for (LO i = 0; i < lNodesPerDir[0]; ++i) {
+            myGIDs[k * lNodesPerDir[1] * lNodesPerDir[0] + j * lNodesPerDir[0] + i] =
+                myGIDOffset + k * gNodesPerDir[1] * gNodesPerDir[0] + j * gNodesPerDir[0] + i;
+          }
         }
-        bop->fillComplete();
-        return bop;
       }
-
-      static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlocked3x3MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
-
-        std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
-        maps[0] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-        maps[1] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-        maps[2] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-        RCP<Matrix> A00 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
-        RCP<Matrix> A01 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
-        RCP<Matrix> A10 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-        RCP<Matrix> A11 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
-        RCP<Matrix> A12 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-        RCP<Matrix> A21 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
-        RCP<Matrix> A22 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
-
-        // create map extractor
-        // To generate the Thyra style map extractor we do not need a full map but only the
-        // information about the Map details (i.e. lib and indexBase). We can extract this
-        // information from maps[0]
-        Teuchos::RCP<const MapExtractor > rgMapExtractor =
-            Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-        Teuchos::RCP<const MapExtractor > doMapExtractor =
-            Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-        // build blocked operator
-        Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,5));
-        bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(0),A00);
-        bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(1),A01);
-        bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(0),A10);
-        bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(1),A11);
-        bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(2),A12);
-        bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(1),A21);
-        bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(2),A22);
-        bop->fillComplete();
-        return bop;
+    } else if (meshLayout == "Local Lexicographic") {
+      for (LO nodeIdx = 0; nodeIdx < lNumNodes; ++nodeIdx) {
+        myGIDs[nodeIdx] = myGIDOffset + nodeIdx;
       }
-
-      // Needed to initialize correctly a level used for testing SingleLevel factory Build() methods.
-      // This method initializes LevelID and linked list of level
-      static void createSingleLevelHierarchy(Level& currentLevel) {
-        RCP<FactoryManager> factoryHandler = rcp(new FactoryManager());
-        factoryHandler->SetKokkosRefactor(false);
-        currentLevel.SetFactoryManager(factoryHandler);
-
-        currentLevel.SetLevelID(0);
+    }
+
+    RCP<const Map> coordMap = MapFactory::Build(lib, gNumNodes, myGIDs(), 0, comm);
+
+    ///////////////////////////////////////
+    //                                   //
+    //    Step 3: Compute coordinates    //
+    //                                   //
+    ///////////////////////////////////////
+    RCP<RealValuedMultiVector> Coordinates = RealValuedMultiVectorFactory::Build(coordMap,
+                                                                                 numDimensions);
+    Array<ArrayRCP<real_type> > myCoords(numDimensions);
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      myCoords[dim] = Coordinates->getDataNonConst(dim);
+    }
+
+    LO nodeIdx = 0;
+    Array<LO> ijk(3);
+    for (ijk[2] = 0; ijk[2] < lNodesPerDir[2]; ++ijk[2]) {
+      for (ijk[1] = 0; ijk[1] < lNodesPerDir[1]; ++ijk[1]) {
+        for (ijk[0] = 0; ijk[0] < lNodesPerDir[0]; ++ijk[0]) {
+          nodeIdx = ijk[2] * lNodesPerDir[1] * lNodesPerDir[0] + ijk[1] * lNodesPerDir[0] + ijk[0];
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            if (gNodesPerDir[dim] == 1) {
+              myCoords[dim][nodeIdx] = 0.0;
+            } else {
+              myCoords[dim][nodeIdx] =
+                  Teuchos::as<real_type>(myLowestTuple[dim] + ijk[dim]) / (gNodesPerDir[dim] - 1);
+            }
+          }
+        }
       }
+    }
+
+    return Coordinates;
+  }  // BuildGeoCoordinates
+
+  // Xpetra version of CreateMap
+  static RCP<Map> BuildMap(Xpetra::UnderlyingLib lib, const std::set<GlobalOrdinal>& gids, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+    Teuchos::Array<GlobalOrdinal> mapvec;
+    mapvec.reserve(gids.size());
+    mapvec.assign(gids.begin(), gids.end());
+    GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
+    GlobalOrdinal gcount;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
+
+    Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
+        MapFactory::Build(lib, gcount, mapvec(), 0, comm);
+    mapvec.clear();
+    return map;
+  }
 
-      // Needed to initialize correctly levels used for testing TwoLevel factory Build() methods.
-      // This method initializes LevelID and linked list of level
-      static void createTwoLevelHierarchy(Level& fineLevel, Level& coarseLevel) {
-        RCP<FactoryManager> factoryHandler = rcp(new FactoryManager());
-        factoryHandler->SetKokkosRefactor(false);
-        fineLevel.SetFactoryManager(factoryHandler);
-        coarseLevel.SetFactoryManager(factoryHandler);
-
-        coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
-
-        fineLevel.SetLevelID(0);
-        coarseLevel.SetLevelID(1);
-      }
+  // Xpetra version of SplitMap
+  static Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(Xpetra::UnderlyingLib lib, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Agiven) {
+    Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
+
+    GlobalOrdinal count = 0;
+    Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
+    for (size_t i = 0; i < Amap.getLocalNumElements(); ++i) {
+      const GlobalOrdinal gid = Amap.getGlobalElement(i);
+      if (Agiven.isNodeGlobalElement(gid)) continue;
+      myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
+      ++count;
+    }
+    myaugids.resize(count);
+    GlobalOrdinal gcount;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
+    return MapFactory::Build(lib, gcount, myaugids(), 0, comm);
+  }
 
-      static RCP<SmootherPrototype> createSmootherPrototype(const std::string& type="Gauss-Seidel", LO sweeps=1) {
-        std::string ifpackType = "RELAXATION";
-        Teuchos::ParameterList  ifpackList;
-        ifpackList.set("relaxation: type", type);
-        ifpackList.set("relaxation: sweeps", (LO) sweeps);
-        ifpackList.set("relaxation: damping factor", (SC) 1.0);
-        return Teuchos::rcp( new TrilinosSmoother(ifpackType, ifpackList) );
-      }
+  static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrix(Xpetra::UnderlyingLib lib, int noBlocks, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+    GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, noBlocks - 2)) * 10;
 
+    GlobalOrdinal procOffset = comm->getRank() * nOverallDOFGidsPerProc;
 
-      // Create a matrix as specified by parameter list options
-      static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) {
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-        GO GO_INVALID = Teuchos::OrdinalTraits<GO>::invalid();
-        RCP<Matrix> Op;
-
-         if (lib == Xpetra::NotSpecified)
-           lib = TestHelpers::Parameters::getLib();
-
-
-         // Make the base graph
-         RCP<Matrix> old_matrix    = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList,lib);
-         RCP<const CrsGraph> old_graph   = old_matrix->getCrsGraph();
-         RCP<const Map> old_rowmap = old_graph->getRowMap();
-         RCP<const Map> old_colmap = old_graph->getColMap();
-         int blocksize = 3;
-
-         // Block Map
-         LO orig_num_rows = (LO) old_graph->getRowMap()->getLocalNumElements();
-         Teuchos::Array<GlobalOrdinal> owned_rows(blocksize*orig_num_rows);
-         for(LO i=0; i<orig_num_rows; i++) {
-           GO old_gid = old_rowmap->getGlobalElement(i);
-           for(int j=0; j<blocksize; j++) {
-             owned_rows[i*blocksize+j] = old_gid*blocksize+j;
-           }
-         }
-         RCP<Map> new_map = Xpetra::MapFactory<LO,GO,NO>::Build(lib,GO_INVALID,owned_rows(),0,comm);
-         if(new_map.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: Map constructor failed");
-
-
-         // Block Graph / Matrix
-         RCP<CrsMatrix> new_matrix = Xpetra::CrsMatrixFactory<SC,LO,GO,NO>::Build(new_map,blocksize*old_graph->getLocalMaxNumRowEntries());
-         if(new_matrix.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: Matrix constructor failed");
-         for(LO i=0; i<orig_num_rows; i++) {
-           Teuchos::ArrayView<const LO> old_indices;
-           Teuchos::ArrayView<const SC> old_values;
-           Teuchos::Array<GO> new_indices(1);
-           Teuchos::Array<SC> new_values(1);
-           old_matrix->getLocalRowView(i,old_indices,old_values);
-           for(int ii=0; ii<blocksize; ii++) {
-             GO GRID = new_map->getGlobalElement(i*blocksize+ii);
-             for(LO j=0; j<(LO)old_indices.size(); j++) {
-               for(int jj=0; jj<blocksize; jj++) {
-                new_indices[0] = old_colmap->getGlobalElement(old_indices[j]) * blocksize + jj;
-                new_values[0]  = old_values[j] * (SC)( (ii == jj && i == old_indices[j] ) ? blocksize*blocksize : 1 );
-                new_matrix->insertGlobalValues(GRID,new_indices(),new_values);
-               }
-             }
-           }
-         }
-         new_matrix->fillComplete();
-         Op = rcp(new CrsMatrixWrap(new_matrix));
-         if(new_map.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: CrsMatrixWrap constructor failed");
-         Op->SetFixedBlockSize(blocksize);
-
-         return Op;
-      } // BuildBlockMatrixAsPoint()
-
-    }; // class TestFactory
-
-
-    // Helper class which has some Tpetra specific code inside
-    // We put this into an extra helper class as we need partial specializations and
-    // do not want to introduce partial specializations for the full TestFactory class
-    //
-    // The BuildBlockMatrix is only available with Tpetra. However, if both Epetra
-    // and Tpetra are enabled it may be that Tpetra is not instantiated on either
-    // GO=int/long long and/or Node=Serial/OpenMP. We need partial specializations
-    // with an empty BuildBlockMatrix routine for all instantiations Teptra is not
-    // enabled for, but are existing in Xpetra due to Epetra enabled.
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class TpetraTestFactory {
-#include "MueLu_UseShortNames.hpp"
-    public:
+    std::set<GlobalOrdinal> myDOFGids;
+    for (GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
+      myDOFGids.insert(i + procOffset);
 
-      // Create a matrix as specified by parameter list options
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) {
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-        RCP<Matrix> Op;
+    Teuchos::RCP<Map> fullmap = TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(lib, myDOFGids, comm);
 
-         if (lib == Xpetra::NotSpecified)
-           lib = TestHelpers::Parameters::getLib();
+    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+    GlobalOrdinal nPartGIDs            = nOverallDOFGidsPerProc;
+    Teuchos::RCP<Map> remainingpartmap = fullmap;
+    for (int it = 0; it < noBlocks; it++) {
+      if (it == noBlocks - 1) {
+        maps[0] = remainingpartmap;
+        break;
+      }
+      // collect first half of GIDs
+      nPartGIDs = nPartGIDs / 2;
+      std::set<GlobalOrdinal> myHalfGIDs;
+      for (GlobalOrdinal j = 0; j < nPartGIDs; j++)
+        myHalfGIDs.insert(j + procOffset);
+
+      Teuchos::RCP<Map> halfmap = TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(lib, myHalfGIDs, comm);
+
+      Teuchos::RCP<Map> secondmap = TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::SplitMap(lib, *remainingpartmap, *halfmap);
+      remainingpartmap            = halfmap;
+
+      maps[noBlocks - 1 - it] = secondmap;
+    }
+
+    // create diagonal blocks
+    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+    for (int it = 0; it < noBlocks; it++) {
+      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
+
+      LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+
+      for (LocalOrdinal i = 0; i < NumMyElements; i++)
+        blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                       Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                       Teuchos::tuple<Scalar>(it + 1));
+      blocks[it]->fillComplete();
+    }
+
+    // create map extractor
+    Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+    Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+
+    // build blocked operator
+    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
+
+    for (int it = 0; it < noBlocks; it++) {
+      Teuchos::RCP<CrsMatrixWrap> csrwrap =
+          Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+      bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
+    }
+    bop->fillComplete();
+    return bop;
+  }
 
-         // This only works for Tpetra
-         if (lib!=Xpetra::UseTpetra) return Op;
+  static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrixThyra(Xpetra::UnderlyingLib lib, int noBlocks, Teuchos::RCP<const Teuchos::Comm<int> > comm) {
+    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+
+    maps[0] = MapFactory::Build(lib, comm->getSize() * 5, 5, 0, comm);
+    for (int it = 1; it < noBlocks; it++) {
+      GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, it - 1) * 5);
+      maps[it]                = MapFactory::Build(lib, comm->getSize() * localDofs, localDofs, 0, comm);
+    }
+
+    // create diagonal blocks
+    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+    for (int it = 0; it < noBlocks; it++) {
+      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
+
+      LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+
+      for (LocalOrdinal i = 0; i < NumMyElements; i++)
+        blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                       Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                       Teuchos::tuple<Scalar>(it + 1));
+      blocks[it]->fillComplete();
+    }
+
+    // create map extractor
+    // To generate the Thyra style map extractor we do not need a full map but only the
+    // information about the Map details (i.e. lib and indexBase). We can extract this
+    // information from maps[0]
+    Teuchos::RCP<const MapExtractor> rgMapExtractor =
+        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+    Teuchos::RCP<const MapExtractor> doMapExtractor =
+        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+    // build blocked operator
+    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
+
+    for (int it = 0; it < noBlocks; it++) {
+      Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+      bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
+    }
+    bop->fillComplete();
+    return bop;
+  }
 
-         // Thanks for the code, Travis!
+  static Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlocked3x3MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
+    std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
+    maps[0]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+    maps[1]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+    maps[2]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+    RCP<Matrix> A00                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
+    RCP<Matrix> A01                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
+    RCP<Matrix> A10                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+    RCP<Matrix> A11                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
+    RCP<Matrix> A12                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+    RCP<Matrix> A21                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
+    RCP<Matrix> A22                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
+
+    // create map extractor
+    // To generate the Thyra style map extractor we do not need a full map but only the
+    // information about the Map details (i.e. lib and indexBase). We can extract this
+    // information from maps[0]
+    Teuchos::RCP<const MapExtractor> rgMapExtractor =
+        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+    Teuchos::RCP<const MapExtractor> doMapExtractor =
+        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+    // build blocked operator
+    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 5));
+    bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(0), A00);
+    bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(1), A01);
+    bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(0), A10);
+    bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(1), A11);
+    bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(2), A12);
+    bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(1), A21);
+    bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(2), A22);
+    bop->fillComplete();
+    return bop;
+  }
 
-         // Make the graph
-         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > FirstMatrix = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList,lib);
-         RCP<const Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > FGraph = FirstMatrix->getCrsGraph();
+  // Needed to initialize correctly a level used for testing SingleLevel factory Build() methods.
+  // This method initializes LevelID and linked list of level
+  static void createSingleLevelHierarchy(Level& currentLevel) {
+    RCP<FactoryManager> factoryHandler = rcp(new FactoryManager());
+    factoryHandler->SetKokkosRefactor(false);
+    currentLevel.SetFactoryManager(factoryHandler);
 
-         int blocksize = 3;
-         RCP<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TGraph = rcp_dynamic_cast<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> >(FGraph);
-         RCP<const Tpetra::CrsGraph<LocalOrdinal,GlobalOrdinal,Node> > TTGraph = TGraph->getTpetra_CrsGraph();
+    currentLevel.SetLevelID(0);
+  }
 
-         RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bcrsmatrix = rcp(new Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> (*TTGraph, blocksize));
+  // Needed to initialize correctly levels used for testing TwoLevel factory Build() methods.
+  // This method initializes LevelID and linked list of level
+  static void createTwoLevelHierarchy(Level& fineLevel, Level& coarseLevel) {
+    RCP<FactoryManager> factoryHandler = rcp(new FactoryManager());
+    factoryHandler->SetKokkosRefactor(false);
+    fineLevel.SetFactoryManager(factoryHandler);
+    coarseLevel.SetFactoryManager(factoryHandler);
 
-         const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& meshRowMap = *bcrsmatrix->getRowMap();
-         const Scalar zero   = Teuchos::ScalarTraits<Scalar>::zero();
-         const Scalar one   = Teuchos::ScalarTraits<Scalar>::one();
-         const Scalar two   = one+one;
-         const Scalar three = two+one;
+    coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
 
-         Teuchos::Array<Scalar> basematrix(blocksize*blocksize, zero);
-         basematrix[0] = two;
-         basematrix[2] = three;
-         basematrix[3] = three;
-         basematrix[4] = two;
-         basematrix[7] = three;
-         basematrix[8] = two;
-         Teuchos::Array<Scalar> offmatrix(blocksize*blocksize, zero);
-         offmatrix[0]=offmatrix[4]=offmatrix[8]=-1;
+    fineLevel.SetLevelID(0);
+    coarseLevel.SetLevelID(1);
+  }
 
-         Teuchos::Array<LocalOrdinal> lclColInds(1);
-         for (LocalOrdinal lclRowInd = meshRowMap.getMinLocalIndex (); lclRowInd <= meshRowMap.getMaxLocalIndex(); ++lclRowInd) {
-           lclColInds[0] = lclRowInd;
-           bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &basematrix[0], 1);
-           
-           // Off diagonals
-           if(lclRowInd > meshRowMap.getMinLocalIndex ()) {
-             lclColInds[0] = lclRowInd - 1;
-             bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &offmatrix[0], 1);
-           }
-           if(lclRowInd < meshRowMap.getMaxLocalIndex ()) {
-             lclColInds[0] = lclRowInd + 1;
-             bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &offmatrix[0], 1);
-           }
+  static RCP<SmootherPrototype> createSmootherPrototype(const std::string& type = "Gauss-Seidel", LO sweeps = 1) {
+    std::string ifpackType = "RELAXATION";
+    Teuchos::ParameterList ifpackList;
+    ifpackList.set("relaxation: type", type);
+    ifpackList.set("relaxation: sweeps", (LO)sweeps);
+    ifpackList.set("relaxation: damping factor", (SC)1.0);
+    return Teuchos::rcp(new TrilinosSmoother(ifpackType, ifpackList));
+  }
 
-         }
+  // Create a matrix as specified by parameter list options
+  static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    GO GO_INVALID                       = Teuchos::OrdinalTraits<GO>::invalid();
+    RCP<Matrix> Op;
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers::Parameters::getLib();
+
+    // Make the base graph
+    RCP<Matrix> old_matrix        = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList, lib);
+    RCP<const CrsGraph> old_graph = old_matrix->getCrsGraph();
+    RCP<const Map> old_rowmap     = old_graph->getRowMap();
+    RCP<const Map> old_colmap     = old_graph->getColMap();
+    int blocksize                 = 3;
+
+    // Block Map
+    LO orig_num_rows = (LO)old_graph->getRowMap()->getLocalNumElements();
+    Teuchos::Array<GlobalOrdinal> owned_rows(blocksize * orig_num_rows);
+    for (LO i = 0; i < orig_num_rows; i++) {
+      GO old_gid = old_rowmap->getGlobalElement(i);
+      for (int j = 0; j < blocksize; j++) {
+        owned_rows[i * blocksize + j] = old_gid * blocksize + j;
+      }
+    }
+    RCP<Map> new_map = Xpetra::MapFactory<LO, GO, NO>::Build(lib, GO_INVALID, owned_rows(), 0, comm);
+    if (new_map.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: Map constructor failed");
+
+    // Block Graph / Matrix
+    RCP<CrsMatrix> new_matrix = Xpetra::CrsMatrixFactory<SC, LO, GO, NO>::Build(new_map, blocksize * old_graph->getLocalMaxNumRowEntries());
+    if (new_matrix.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: Matrix constructor failed");
+    for (LO i = 0; i < orig_num_rows; i++) {
+      Teuchos::ArrayView<const LO> old_indices;
+      Teuchos::ArrayView<const SC> old_values;
+      Teuchos::Array<GO> new_indices(1);
+      Teuchos::Array<SC> new_values(1);
+      old_matrix->getLocalRowView(i, old_indices, old_values);
+      for (int ii = 0; ii < blocksize; ii++) {
+        GO GRID = new_map->getGlobalElement(i * blocksize + ii);
+        for (LO j = 0; j < (LO)old_indices.size(); j++) {
+          for (int jj = 0; jj < blocksize; jj++) {
+            new_indices[0] = old_colmap->getGlobalElement(old_indices[j]) * blocksize + jj;
+            new_values[0]  = old_values[j] * (SC)((ii == jj && i == old_indices[j]) ? blocksize * blocksize : 1);
+            new_matrix->insertGlobalValues(GRID, new_indices(), new_values);
+          }
+        }
+      }
+    }
+    new_matrix->fillComplete();
+    Op = rcp(new CrsMatrixWrap(new_matrix));
+    if (new_map.is_null()) throw std::runtime_error("BuildBlockMatrixAsPoint: CrsMatrixWrap constructor failed");
+    Op->SetFixedBlockSize(blocksize);
 
-         RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(bcrsmatrix));
-         Op = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(temp));
-         return Op;
-      } // BuildBlockMatrix()
+    return Op;
+  }  // BuildBlockMatrixAsPoint()
 
+};  // class TestFactory
 
+// Helper class which has some Tpetra specific code inside
+// We put this into an extra helper class as we need partial specializations and
+// do not want to introduce partial specializations for the full TestFactory class
+//
+// The BuildBlockMatrix is only available with Tpetra. However, if both Epetra
+// and Tpetra are enabled it may be that Tpetra is not instantiated on either
+// GO=int/long long and/or Node=Serial/OpenMP. We need partial specializations
+// with an empty BuildBlockMatrix routine for all instantiations Teptra is not
+// enabled for, but are existing in Xpetra due to Epetra enabled.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TpetraTestFactory {
+#include "MueLu_UseShortNames.hpp"
+ public:
+  // Create a matrix as specified by parameter list options
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    RCP<Matrix> Op;
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers::Parameters::getLib();
+
+    // This only works for Tpetra
+    if (lib != Xpetra::UseTpetra) return Op;
+
+    // Thanks for the code, Travis!
+
+    // Make the graph
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > FirstMatrix = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList, lib);
+    RCP<const Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > FGraph      = FirstMatrix->getCrsGraph();
+
+    int blocksize                                                                = 3;
+    RCP<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TGraph = rcp_dynamic_cast<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> >(FGraph);
+    RCP<const Tpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TTGraph      = TGraph->getTpetra_CrsGraph();
+
+    RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bcrsmatrix = rcp(new Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*TTGraph, blocksize));
+
+    const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& meshRowMap = *bcrsmatrix->getRowMap();
+    const Scalar zero                                                = Teuchos::ScalarTraits<Scalar>::zero();
+    const Scalar one                                                 = Teuchos::ScalarTraits<Scalar>::one();
+    const Scalar two                                                 = one + one;
+    const Scalar three                                               = two + one;
+
+    Teuchos::Array<Scalar> basematrix(blocksize * blocksize, zero);
+    basematrix[0] = two;
+    basematrix[2] = three;
+    basematrix[3] = three;
+    basematrix[4] = two;
+    basematrix[7] = three;
+    basematrix[8] = two;
+    Teuchos::Array<Scalar> offmatrix(blocksize * blocksize, zero);
+    offmatrix[0] = offmatrix[4] = offmatrix[8] = -1;
+
+    Teuchos::Array<LocalOrdinal> lclColInds(1);
+    for (LocalOrdinal lclRowInd = meshRowMap.getMinLocalIndex(); lclRowInd <= meshRowMap.getMaxLocalIndex(); ++lclRowInd) {
+      lclColInds[0] = lclRowInd;
+      bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &basematrix[0], 1);
+
+      // Off diagonals
+      if (lclRowInd > meshRowMap.getMinLocalIndex()) {
+        lclColInds[0] = lclRowInd - 1;
+        bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &offmatrix[0], 1);
+      }
+      if (lclRowInd < meshRowMap.getMaxLocalIndex()) {
+        lclColInds[0] = lclRowInd + 1;
+        bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &offmatrix[0], 1);
+      }
+    }
 
+    RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(bcrsmatrix));
+    Op                                                                      = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(temp));
+    return Op;
+  }  // BuildBlockMatrix()
 
-    private:
-      TpetraTestFactory() {} // static class
+ private:
+  TpetraTestFactory() {}  // static class
 
-    }; // class TpetraTestFactory
+};  // class TpetraTestFactory
 
-    // TAW: 3/14/2016: If both Epetra and Tpetra are enabled we need partial specializations
-    //                 on GO=int/long long as well as NO=EpetraNode to disable BuildBlockMatrix
+// TAW: 3/14/2016: If both Epetra and Tpetra are enabled we need partial specializations
+//                 on GO=int/long long as well as NO=EpetraNode to disable BuildBlockMatrix
 #ifdef HAVE_MUELU_EPETRA
-    // partial specializations (GO=int not enabled with Tpetra)
+// partial specializations (GO=int not enabled with Tpetra)
 #if !defined(HAVE_TPETRA_INST_INT_INT)
-    template <class Scalar, class LocalOrdinal, class Node>
-    class TpetraTestFactory<Scalar, LocalOrdinal, int, Node> {
-      typedef int GlobalOrdinal;
+template <class Scalar, class LocalOrdinal, class Node>
+class TpetraTestFactory<Scalar, LocalOrdinal, int, Node> {
+  typedef int GlobalOrdinal;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-      static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+  static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
 #endif
 
-    // partial specializations (GO=long long not enabled with Tpetra)
+// partial specializations (GO=long long not enabled with Tpetra)
 #if !defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-    template <class Scalar, class LocalOrdinal, class Node>
-    class TpetraTestFactory<Scalar, LocalOrdinal, long long, Node> {
-      typedef long long GlobalOrdinal;
+template <class Scalar, class LocalOrdinal, class Node>
+class TpetraTestFactory<Scalar, LocalOrdinal, long long, Node> {
+  typedef long long GlobalOrdinal;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-      static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+  static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
 #endif
 
-    // partial specializations (NO=EpetraNode not enabled with Tpetra)
+// partial specializations (NO=EpetraNode not enabled with Tpetra)
 #if ((defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_OPENMP))) || \
-    (!defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_SERIAL))))
+     (!defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_SERIAL))))
 
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal>
-    class TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Xpetra::EpetraNode> {
-      typedef Xpetra::EpetraNode Node;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal>
+class TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Xpetra::EpetraNode> {
+  typedef Xpetra::EpetraNode Node;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-      static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
-#endif
-#endif // endif HAVE_MUELU_EPETRA
-
-    //! Return the list of files in the directory. Only files that are matching '*filter*' are returned.
-    ArrayRCP<std::string> GetFileList(const std::string & dirPath, const std::string & filter);
-
-
-
-
-  } // namespace TestHelpers
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+  static RCP<Matrix> BuildBlockMatrixAsPoint(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
 
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
+#endif
+#endif  // endif HAVE_MUELU_EPETRA
 
+//! Return the list of files in the directory. Only files that are matching '*filter*' are returned.
+ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter);
 
-} // namespace MueLuTests
+}  // namespace TestHelpers
 
+}  // namespace MueLuTests
 
 // Macro to skip a test when UnderlyingLib==Epetra or Tpetra
-#define MUELU_TEST_ONLY_FOR(UnderlyingLib) \
-  if (TestHelpers::Parameters::getLib() != UnderlyingLib) { \
-    out << "Skipping test for " << ((TestHelpers::Parameters::getLib()==Xpetra::UseEpetra) ? "Epetra" : "Tpetra") << std::endl; \
-    return; \
+#define MUELU_TEST_ONLY_FOR(UnderlyingLib)                                                                                        \
+  if (TestHelpers::Parameters::getLib() != UnderlyingLib) {                                                                       \
+    out << "Skipping test for " << ((TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) ? "Epetra" : "Tpetra") << std::endl; \
+    return;                                                                                                                       \
   }
 
 // Macro to skip a test when Epetra is used with Ordinal != int
@@ -984,23 +957,21 @@ namespace MueLuTests {
   if (!(TestHelpers::Parameters::getLib() == Xpetra::UseEpetra && (Teuchos::OrdinalTraits<LocalOrdinal>::name() != string("int") || Teuchos::OrdinalTraits<GlobalOrdinal>::name() != string("int"))))
 
 // Macro to skip a test when Epetra is used with Scalar != double or Ordinal != int
-#define MUELU_TEST_EPETRA_ONLY_FOR_DOUBLE_AND_INT(Scalar, LocalOrdinal, GlobalOrdinal) \
+#define MUELU_TEST_EPETRA_ONLY_FOR_DOUBLE_AND_INT(Scalar, LocalOrdinal, GlobalOrdinal)                                        \
   if (!(TestHelpers::Parameters::getLib() == Xpetra::UseEpetra && Teuchos::ScalarTraits<Scalar>::name() != string("double"))) \
-    MUELU_TEST_EPETRA_ONLY_FOR_INT(LocalOrdinal, GlobalOrdinal)
+  MUELU_TEST_EPETRA_ONLY_FOR_INT(LocalOrdinal, GlobalOrdinal)
 
 //
 
-//TODO: add directly to Teuchos ?
+// TODO: add directly to Teuchos ?
 //#include "../xpetra/test/Xpetra_UnitTestHelpers.hpp" // declaration of TEUCHOS_UNIT_TEST_TEMPLATE_5_DECL
 
-
 //
 
-
 //! Namespace for MueLu test classes
 namespace MueLuTests {
 
-  using namespace TestHelpers;
+using namespace TestHelpers;
 }
 
-#endif // ifndef MUELU_TEST_HELPERS_H
+#endif  // ifndef MUELU_TEST_HELPERS_H
diff --git a/packages/muelu/test/unit_tests/MueLu_TestHelpers2.hpp b/packages/muelu/test/unit_tests/MueLu_TestHelpers2.hpp
index d635d86843df..e00c56a47176 100644
--- a/packages/muelu/test/unit_tests/MueLu_TestHelpers2.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_TestHelpers2.hpp
@@ -83,102 +83,110 @@
 
 namespace MueLuTests {
 
-  namespace TestHelpers {
+namespace TestHelpers {
 
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class TestProblem {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TestProblem {
 #include "MueLu_UseShortNames.hpp"
 
-    public:
-      TestProblem(Xpetra::UnderlyingLib lib) : lib_(lib) { }
+ public:
+  TestProblem(Xpetra::UnderlyingLib lib)
+    : lib_(lib) {}
 
-      void Init() {
-        if (A_ == Teuchos::null) {
-
-          // Create a matrix
-          {
-            A_ = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
-          }
-
-          // Create a Hierarchy
-          {
-            LO maxLevels = 3;
-
-            H_ = rcp(new Hierarchy(A_));
-            RCP<Level> Finest = H_->GetLevel();
-
-            RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-            aggFact->SetMinNodesPerAggregate(3);
-            aggFact->SetMaxNeighAlreadySelected(0);
-            aggFact->SetOrdering("natural");
+  void Init() {
+    if (A_ == Teuchos::null) {
+      // Create a matrix
+      {
+        A_ = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
+      }
 
-            Teuchos::ParameterList smootherParamList;
-            smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-            smootherParamList.set("relaxation: sweeps", (LO) 1);
-            smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-            RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
-            RCP<SmootherFactory>   smooFact  = rcp(new SmootherFactory(smooProto));
+      // Create a Hierarchy
+      {
+        LO maxLevels = 3;
 
-            RCP<SmootherPrototype> coarseProto = rcp(new DirectSolver());
-            RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
+        H_                = rcp(new Hierarchy(A_));
+        RCP<Level> Finest = H_->GetLevel();
 
-            FactoryManager M;
-            M.SetFactory("Aggregates", aggFact);
-            M.SetFactory("Smoother", smooFact);
-            M.SetFactory("CoarseSolver", coarseSolveFact);
+        RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+        aggFact->SetMinNodesPerAggregate(3);
+        aggFact->SetMaxNeighAlreadySelected(0);
+        aggFact->SetOrdering("natural");
 
-            H_->Setup(M, 0, maxLevels);
-          }
+        Teuchos::ParameterList smootherParamList;
+        smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+        smootherParamList.set("relaxation: sweeps", (LO)1);
+        smootherParamList.set("relaxation: damping factor", (SC)1.0);
+        RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+        RCP<SmootherFactory> smooFact    = rcp(new SmootherFactory(smooProto));
 
-          // Create RHS
-          {
-            RCP<MultiVector> X = MultiVectorFactory::Build(A_->getRowMap(), 1);
-            B_                 = MultiVectorFactory::Build(A_->getRowMap(), 1);
+        RCP<SmootherPrototype> coarseProto   = rcp(new DirectSolver());
+        RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(coarseProto, Teuchos::null));
 
-            X->setSeed(846930886);
-            X->randomize();
+        FactoryManager M;
+        M.SetFactory("Aggregates", aggFact);
+        M.SetFactory("Smoother", smooFact);
+        M.SetFactory("CoarseSolver", coarseSolveFact);
 
-            A_->apply(*X, *B_, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
-          }
-        }
+        H_->Setup(M, 0, maxLevels);
       }
 
-      virtual ~TestProblem() { }
+      // Create RHS
+      {
+        RCP<MultiVector> X = MultiVectorFactory::Build(A_->getRowMap(), 1);
+        B_                 = MultiVectorFactory::Build(A_->getRowMap(), 1);
 
-      RCP<Matrix>    & GetA()   { Init(); return A_; }
-      RCP<Hierarchy>   & GetH()   { Init(); return H_; }
-      RCP<MultiVector> & GetRHS() { Init(); return B_; }
+        X->setSeed(846930886);
+        X->randomize();
 
-      RCP<MultiVector> GetNewX0 () {
-        Init ();
-        RCP<MultiVector> X = MultiVectorFactory::Build (A_->getRowMap (), 1);
-        X->putScalar (0.0);
-        return X;
+        A_->apply(*X, *B_, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
       }
-
-    private:
-      Xpetra::UnderlyingLib lib_;
-      RCP<Matrix>    A_;
-      RCP<Hierarchy>   H_;
-      RCP<MultiVector> B_;
-    };
-
-    // Singleton for TestProblem
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    RCP<TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> > &
-    getTestProblem (Xpetra::UnderlyingLib lib)
-    {
-      static Array<RCP<TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> > > problem_(2);
-
-      int libNum = (Xpetra::UseEpetra) ? 0 : 1;
-      if (problem_[libNum] == Teuchos::null)
-        problem_[libNum] = rcp (new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> (lib));
-
-      return problem_[libNum];
     }
-
-  } // namespace
-
-} // namespace
-
-#endif // ifndef MUELU_TEST_HELPERS2_H
+  }
+
+  virtual ~TestProblem() {}
+
+  RCP<Matrix>& GetA() {
+    Init();
+    return A_;
+  }
+  RCP<Hierarchy>& GetH() {
+    Init();
+    return H_;
+  }
+  RCP<MultiVector>& GetRHS() {
+    Init();
+    return B_;
+  }
+
+  RCP<MultiVector> GetNewX0() {
+    Init();
+    RCP<MultiVector> X = MultiVectorFactory::Build(A_->getRowMap(), 1);
+    X->putScalar(0.0);
+    return X;
+  }
+
+ private:
+  Xpetra::UnderlyingLib lib_;
+  RCP<Matrix> A_;
+  RCP<Hierarchy> H_;
+  RCP<MultiVector> B_;
+};
+
+// Singleton for TestProblem
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+RCP<TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> >&
+getTestProblem(Xpetra::UnderlyingLib lib) {
+  static Array<RCP<TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node> > > problem_(2);
+
+  int libNum = (Xpetra::UseEpetra) ? 0 : 1;
+  if (problem_[libNum] == Teuchos::null)
+    problem_[libNum] = rcp(new TestProblem<Scalar, LocalOrdinal, GlobalOrdinal, Node>(lib));
+
+  return problem_[libNum];
+}
+
+}  // namespace TestHelpers
+
+}  // namespace MueLuTests
+
+#endif  // ifndef MUELU_TEST_HELPERS2_H
diff --git a/packages/muelu/test/unit_tests/MueLu_TestHelpers_Common.hpp b/packages/muelu/test/unit_tests/MueLu_TestHelpers_Common.hpp
index d1ea968953ec..aa45534824a5 100644
--- a/packages/muelu/test/unit_tests/MueLu_TestHelpers_Common.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_TestHelpers_Common.hpp
@@ -47,11 +47,13 @@
 #define MUELU_TEST_HELPERS_COMMON_HPP
 
 #if 1
-#define MUELU_DEBUGGER_MACRO \
+#define MUELU_DEBUGGER_MACRO                                   \
   out << "debug: scalar = " << typeid(GO).name() << std::endl; \
   out << "debug: node   = " << typeid(NO).name() << std::endl; \
-  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) out << "debug: linAlgebra = Epetra" << std::endl; \
-  else                                                        out << "debug: linAlgebra = Tpetra" << std::endl;
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra)  \
+    out << "debug: linAlgebra = Epetra" << std::endl;          \
+  else                                                         \
+    out << "debug: linAlgebra = Tpetra" << std::endl;
 #else
 #define MUELU_DEBUGGER_MACRO
 #endif
@@ -61,42 +63,72 @@
 
 #ifdef HAVE_MUELU_EPETRA
 
-# if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
      (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  // <double,int,int,EpetraNode>: run Epetra, do not run Tpetra
-#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO) \
-  MUELU_DEBUGGER_MACRO \
-  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) { \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; } \
-  } else if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra) { \
+// <double,int,int,EpetraNode>: run Epetra, do not run Tpetra
+#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO)                                                  \
+  MUELU_DEBUGGER_MACRO                                                                         \
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {                                \
+    if (!TYPE_EQUAL(SC, double)) {                                                             \
+      out << "Skipping Epetra for SC != double" << std::endl;                                  \
+      return;                                                                                  \
+    }                                                                                          \
+    if (!TYPE_EQUAL(GO, int)) {                                                                \
+      out << "Skipping Epetra for GO != int" << std::endl;                                     \
+      return;                                                                                  \
+    }                                                                                          \
+    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                                                 \
+      out << "Skipping Epetra for NO != EpetraNode" << std::endl;                              \
+      return;                                                                                  \
+    }                                                                                          \
+  } else if (TestHelpers::Parameters::getLib() == Xpetra::UseTpetra) {                         \
     if (TYPE_EQUAL(SC, double) && TYPE_EQUAL(GO, int) && TYPE_EQUAL(NO, Xpetra::EpetraNode)) { \
-      out << "Skipping Tpetra for <double, int, EpetraNode>" << std::endl; \
-      return; \
-    } \
+      out << "Skipping Tpetra for <double, int, EpetraNode>" << std::endl;                     \
+      return;                                                                                  \
+    }                                                                                          \
   }
 
-#define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO) \
-  { out << "Skipping as we cannot run Epetra and Tpetra for the same combination of template args" << std::endl; return; }
+#define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO)                                           \
+  {                                                                                                              \
+    out << "Skipping as we cannot run Epetra and Tpetra for the same combination of template args" << std::endl; \
+    return;                                                                                                      \
+  }
 
-# else
+#else
 
-#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO) \
-  MUELU_DEBUGGER_MACRO \
-  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) { \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; } \
+#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO)                     \
+  MUELU_DEBUGGER_MACRO                                            \
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {   \
+    if (!TYPE_EQUAL(SC, double)) {                                \
+      out << "Skipping Epetra for SC != double" << std::endl;     \
+      return;                                                     \
+    }                                                             \
+    if (!TYPE_EQUAL(GO, int)) {                                   \
+      out << "Skipping Epetra for GO != int" << std::endl;        \
+      return;                                                     \
+    }                                                             \
+    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                    \
+      out << "Skipping Epetra for NO != EpetraNode" << std::endl; \
+      return;                                                     \
+    }                                                             \
   }
 
 // If linAlgebra==Tpetra, but the test also requires Epetra, this macro will cause the test
 // to return early if SC!=double, GO!={int}, or NO!=Serial.
 #define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO) \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; }
-# endif
+  if (!TYPE_EQUAL(SC, double)) {                                       \
+    out << "Skipping Epetra for SC != double" << std::endl;            \
+    return;                                                            \
+  }                                                                    \
+  if (!TYPE_EQUAL(GO, int)) {                                          \
+    out << "Skipping Epetra for GO != int" << std::endl;               \
+    return;                                                            \
+  }                                                                    \
+  if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                           \
+    out << "Skipping Epetra for NO != EpetraNode" << std::endl;        \
+    return;                                                            \
+  }
+#endif
 
 #else  // HAVE_MUELU_EPETRA
 
@@ -105,16 +137,15 @@
 
 #endif  // HAVE_MUELU_EPETRA
 
-
 // Macro to set MueLu's internal oh-so FancyOStream to be the same as the one used by Teuchos' unit testing framework.
 // This prevents MueLu's output from intermingling with with the unit test pass/fail summary lines.
 #define MUELU_TESTING_SET_OSTREAM \
-   MueLu::VerboseObject::SetMueLuOStream(Teuchos::fancyOStream(out.getOStream()));
+  MueLu::VerboseObject::SetMueLuOStream(Teuchos::fancyOStream(out.getOStream()));
 
-#define MUELU_TESTING_DO_NOT_TEST(lib,packagesNotEnabled) \
-  if (TestHelpers::Parameters::getLib() == lib) { \
+#define MUELU_TESTING_DO_NOT_TEST(lib, packagesNotEnabled)                                                                                                                                                           \
+  if (TestHelpers::Parameters::getLib() == lib) {                                                                                                                                                                    \
     out << "Skipping test for " << (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra ? "Epetra" : "Tpetra") << " because some required packages are not enabled (" << packagesNotEnabled << ")." << std::endl; \
-    return; \
+    return;                                                                                                                                                                                                          \
   }
 
-#endif // ifndef MUELU_TEST_HELPERS_COMMON_HPP
+#endif  // ifndef MUELU_TEST_HELPERS_COMMON_HPP
diff --git a/packages/muelu/test/unit_tests/MueLu_TestHelpers_HO.hpp b/packages/muelu/test/unit_tests/MueLu_TestHelpers_HO.hpp
index 76ee83487047..82f476ba4db1 100644
--- a/packages/muelu/test/unit_tests/MueLu_TestHelpers_HO.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_TestHelpers_HO.hpp
@@ -47,7 +47,7 @@
 #define MUELU_TEST_HELPERS_HO_H
 #include "MueLu_ConfigDefs.hpp"
 
-//Intrepid
+// Intrepid
 #ifdef HAVE_MUELU_INTREPID2
 #include "Kokkos_DynRankView.hpp"
 
@@ -60,128 +60,120 @@
 #include "MueLu_Utilities_def.hpp"
 
 namespace MueLuTests {
-  namespace TestHelpers {
-
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    void AllocateEpetraFECrsMatrix(RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > & pn_rowmap,  RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > pn_colmap, Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > & B)
-    {
-      throw MueLu::Exceptions::RuntimeError("MueLuTests::TestHelpers::AllocateEpetraFECrsMatrix only works for Tpetra::KokkosCompat::KokkosSerialWrapperNode");
-    }
-
-#if defined(HAVE_MUELU_EPETRA) && \
-           (!defined(HAVE_MUELU_EXPLICIT_INSTANTIATION) || \
-            (defined(HAVE_MUELU_EXPLICIT_INSTANTIATION) && defined(HAVE_TPETRA_INST_SERIAL)) \
-           )
-    template <>
-    void AllocateEpetraFECrsMatrix<double,int,int,Tpetra::KokkosCompat::KokkosSerialWrapperNode>(RCP<const Xpetra::Map<int,int,Tpetra::KokkosCompat::KokkosSerialWrapperNode> > & pn_rowmap,  RCP<const Xpetra::Map<int,int,Tpetra::KokkosCompat::KokkosSerialWrapperNode> > pn_colmap, Teuchos::RCP<Xpetra::Matrix<double,int,int,Tpetra::KokkosCompat::KokkosSerialWrapperNode > > & B)
-
-    {
-      // Epetra is hard
-      const Epetra_Map & pn_rowmap_epetra = Xpetra::toEpetra(*pn_rowmap);
-      const Epetra_Map & pn_colmap_epetra = Xpetra::toEpetra(*pn_colmap);
-      RCP<Epetra_CrsMatrix> B_epetra = rcp(new Epetra_FECrsMatrix(Copy,pn_rowmap_epetra,pn_colmap_epetra,0));
-      B = MueLu::Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap<double,int,int,Tpetra::KokkosCompat::KokkosSerialWrapperNode>(B_epetra);
-    }
+namespace TestHelpers {
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void AllocateEpetraFECrsMatrix(RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& pn_rowmap, RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > pn_colmap, Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& B) {
+  throw MueLu::Exceptions::RuntimeError("MueLuTests::TestHelpers::AllocateEpetraFECrsMatrix only works for Tpetra::KokkosCompat::KokkosSerialWrapperNode");
+}
+
+#if defined(HAVE_MUELU_EPETRA) &&                   \
+    (!defined(HAVE_MUELU_EXPLICIT_INSTANTIATION) || \
+     (defined(HAVE_MUELU_EXPLICIT_INSTANTIATION) && defined(HAVE_TPETRA_INST_SERIAL)))
+template <>
+void AllocateEpetraFECrsMatrix<double, int, int, Tpetra::KokkosCompat::KokkosSerialWrapperNode>(RCP<const Xpetra::Map<int, int, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& pn_rowmap, RCP<const Xpetra::Map<int, int, Tpetra::KokkosCompat::KokkosSerialWrapperNode> > pn_colmap, Teuchos::RCP<Xpetra::Matrix<double, int, int, Tpetra::KokkosCompat::KokkosSerialWrapperNode> >& B)
+
+{
+  // Epetra is hard
+  const Epetra_Map& pn_rowmap_epetra = Xpetra::toEpetra(*pn_rowmap);
+  const Epetra_Map& pn_colmap_epetra = Xpetra::toEpetra(*pn_colmap);
+  RCP<Epetra_CrsMatrix> B_epetra     = rcp(new Epetra_FECrsMatrix(Copy, pn_rowmap_epetra, pn_colmap_epetra, 0));
+  B                                  = MueLu::Convert_Epetra_CrsMatrix_ToXpetra_CrsMatrixWrap<double, int, int, Tpetra::KokkosCompat::KokkosSerialWrapperNode>(B_epetra);
+}
 #endif
 
+// Here nx is the number of nodes on the underlying (p=1) mesh.
+// This mesh is then promoted up to degree
+// Teuchos::RCP<Matrix>
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+Build1DPseudoPoissonHigherOrder(GlobalOrdinal nx, int degree,
+                                Kokkos::DynRankView<LocalOrdinal, typename Node::device_type>& elem_to_node,
+                                Xpetra::UnderlyingLib lib) {
+#include "MueLu_UseShortNames.hpp"
+  using Teuchos::arcp;
+  using Teuchos::arcp_reinterpret_cast;
+  using Teuchos::arcpFromArrayView;
+  using Teuchos::Array;
+  using Teuchos::ArrayRCP;
+  using Teuchos::ArrayView;
+  using Teuchos::null;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  using Teuchos::rcp_dynamic_cast;
+  using Teuchos::rcp_implicit_cast;
+  using Teuchos::rcpFromRef;
+  GO go_invalid = Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("matrixType", "Laplace1D");
+  // Build a lower order matrix
+  RCP<Matrix> A                       = MueLuTests::TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixList, lib);
+  RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
+  int MyPID                           = comm->getRank();
+  int Nproc                           = comm->getSize();
+
+  // Get maps
+  RCP<CrsMatrix> Acrs      = rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix();
+  RCP<const Map> p1_colmap = Acrs->getColMap();
+  RCP<const Map> p1_rowmap = Acrs->getRowMap();
+
+  // Count edges.   For shared edges, lower PID gets the owning nodes
+  GO global_num_nodes       = p1_rowmap->getGlobalNumElements();
+  size_t local_num_nodes    = p1_rowmap->getLocalNumElements();
+  GO global_num_elements    = global_num_nodes - 1;
+  size_t local_num_elements = local_num_nodes;
+  if (p1_rowmap->getGlobalElement(local_num_elements - 1) == global_num_nodes - 1) local_num_elements--;
+
+  printf("[%d] P1 Problem Size: nodes=%d/%d elements=%d/%d\n", MyPID, (int)local_num_nodes, (int)global_num_nodes, (int)local_num_elements, (int)global_num_elements);
+
+  int num_edge_dofs            = (degree - 1) * local_num_elements;
+  size_t p1_num_ghost_col_dofs = p1_colmap->getLocalNumElements() - local_num_nodes;
+
+  // Scansum owned edge counts
+  int edge_start = 0;
+  Teuchos::scan(*comm, Teuchos::REDUCE_SUM, 1, &num_edge_dofs, &edge_start);
+  edge_start -= num_edge_dofs;
+  GO go_edge_start = global_num_nodes + edge_start;
+
+  // Build owned pn map
+  Teuchos::Array<GO> pn_owned_dofs(local_num_nodes + num_edge_dofs);
+  for (size_t i = 0; i < local_num_nodes; i++)
+    pn_owned_dofs[i] = p1_rowmap->getGlobalElement(i);
+  for (size_t i = 0; i < (size_t)num_edge_dofs; i++)
+    pn_owned_dofs[local_num_nodes + i] = go_edge_start + i;
+  RCP<const Map> pn_rowmap = MapFactory::Build(lib, go_invalid, pn_owned_dofs(), p1_rowmap->getIndexBase(), comm);
+
+  // Build owned column map in [E|T]petra ordering - offproc nodes last
+  size_t pn_num_col_dofs = pn_owned_dofs.size() + p1_num_ghost_col_dofs;
+  if (MyPID != 0) pn_num_col_dofs += (degree - 1);  // pn ghosts; left side only
+  Teuchos::Array<GO> pn_col_dofs(pn_num_col_dofs);
+  for (size_t i = 0; i < (size_t)pn_owned_dofs.size(); i++)
+    pn_col_dofs[i] = pn_owned_dofs[i];  // onproc
+
+  //      printf("[%d/%d] DEBUG: degree = %d local_num_elements = %d local_num_nodes = %d num_edge_dofs = %d p1_num_ghost_col_dofs =%d  pn_owned_dofs.size() = %d,pn_num_col_dofs = %d mycount = %d\n",MyPID,Nproc,degree,(int)local_num_elements,(int)local_num_nodes,(int)num_edge_dofs,(int)p1_num_ghost_col_dofs, (int)pn_owned_dofs.size(),(int)pn_num_col_dofs,(int)pn_owned_dofs.size()+(MyPID!=0)*(degree) + (MyPID!=Nproc-1) );
+
+  // We have to copy the ghosts from the p1_rowmap as well as the new edge dofs.
+  // This needs to follow [E|T]petra ordering
+  size_t idx = pn_owned_dofs.size();
+  if (MyPID != 0) {
+    // Left side nodal
+    pn_col_dofs[idx] = p1_colmap->getGlobalElement(p1_rowmap->getLocalNumElements());
+    idx++;
+    // Left side, edge
+    for (size_t i = 0; i < (size_t)(degree - 1); i++) {
+      pn_col_dofs[idx] = go_edge_start - (degree - 1) + i;
+      idx++;
+    }
+  }
+  if (MyPID != Nproc - 1) {
+    // Right side nodal
+    pn_col_dofs[idx] = p1_colmap->getGlobalElement(p1_colmap->getLocalNumElements() - 1);
+    idx++;
+  }
 
-
-
-    // Here nx is the number of nodes on the underlying (p=1) mesh.
-    // This mesh is then promoted up to degree
-    //Teuchos::RCP<Matrix>
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
-    Build1DPseudoPoissonHigherOrder(GlobalOrdinal nx, int degree,
-                                    Kokkos::DynRankView<LocalOrdinal,typename Node::device_type>
-                                    & elem_to_node,
-                                    Xpetra::UnderlyingLib lib)
-    {
-#     include "MueLu_UseShortNames.hpp"
-      using Teuchos::RCP;
-      using Teuchos::rcp;
-      using Teuchos::ArrayRCP;
-      using Teuchos::ArrayView;
-      using Teuchos::arcp;
-      using Teuchos::arcpFromArrayView;
-      using Teuchos::rcpFromRef;
-      using Teuchos::null;
-      using Teuchos::arcp_reinterpret_cast;
-      using Teuchos::Array;
-      using Teuchos::rcp_dynamic_cast;
-      using Teuchos::rcp_implicit_cast;
-      using Teuchos::rcpFromRef;
-      GO go_invalid=Teuchos::OrdinalTraits<Xpetra::global_size_t>::invalid();
-
-      Teuchos::ParameterList matrixList;
-      matrixList.set("nx", nx);
-      matrixList.set("matrixType","Laplace1D");
-      // Build a lower order matrix
-      RCP<Matrix> A = MueLuTests::TestHelpers::TestFactory<SC,LO,GO,NO>::BuildMatrix(matrixList,lib);
-      RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
-      int MyPID = comm->getRank();
-      int Nproc = comm->getSize();
-
-      // Get maps
-      RCP<CrsMatrix> Acrs   = rcp_dynamic_cast<CrsMatrixWrap>(A)->getCrsMatrix();
-      RCP<const Map> p1_colmap = Acrs->getColMap();
-      RCP<const Map> p1_rowmap = Acrs->getRowMap();
-
-      // Count edges.   For shared edges, lower PID gets the owning nodes
-      GO global_num_nodes       = p1_rowmap->getGlobalNumElements();
-      size_t local_num_nodes    = p1_rowmap->getLocalNumElements();
-      GO global_num_elements    = global_num_nodes -1;
-      size_t local_num_elements = local_num_nodes;
-      if(p1_rowmap->getGlobalElement(local_num_elements-1) == global_num_nodes-1) local_num_elements--;
-
-      printf("[%d] P1 Problem Size: nodes=%d/%d elements=%d/%d\n",MyPID,(int)local_num_nodes,(int)global_num_nodes,(int)local_num_elements,(int)global_num_elements);
-
-      int num_edge_dofs   = (degree-1)*local_num_elements;
-      size_t p1_num_ghost_col_dofs = p1_colmap->getLocalNumElements() - local_num_nodes;
-
-      // Scansum owned edge counts
-      int edge_start=0;
-      Teuchos::scan(*comm,Teuchos::REDUCE_SUM,1,&num_edge_dofs,&edge_start);
-      edge_start -=num_edge_dofs;
-      GO go_edge_start = global_num_nodes + edge_start;
-
-      // Build owned pn map
-      Teuchos::Array<GO> pn_owned_dofs(local_num_nodes+num_edge_dofs);
-      for(size_t i=0; i<local_num_nodes; i++)
-        pn_owned_dofs[i] =  p1_rowmap->getGlobalElement(i);
-      for(size_t i=0; i<(size_t)num_edge_dofs; i++)
-        pn_owned_dofs[local_num_nodes+i] = go_edge_start+i;
-      RCP<const Map> pn_rowmap = MapFactory::Build(lib,go_invalid,pn_owned_dofs(),p1_rowmap->getIndexBase(),comm);
-
-      // Build owned column map in [E|T]petra ordering - offproc nodes last
-      size_t pn_num_col_dofs = pn_owned_dofs.size() + p1_num_ghost_col_dofs;
-      if(MyPID != 0) pn_num_col_dofs+=(degree-1); // pn ghosts; left side only
-      Teuchos::Array<GO> pn_col_dofs(pn_num_col_dofs);
-      for(size_t i=0; i<(size_t)pn_owned_dofs.size(); i++)
-        pn_col_dofs[i] = pn_owned_dofs[i]; //onproc
-
-      //      printf("[%d/%d] DEBUG: degree = %d local_num_elements = %d local_num_nodes = %d num_edge_dofs = %d p1_num_ghost_col_dofs =%d  pn_owned_dofs.size() = %d,pn_num_col_dofs = %d mycount = %d\n",MyPID,Nproc,degree,(int)local_num_elements,(int)local_num_nodes,(int)num_edge_dofs,(int)p1_num_ghost_col_dofs, (int)pn_owned_dofs.size(),(int)pn_num_col_dofs,(int)pn_owned_dofs.size()+(MyPID!=0)*(degree) + (MyPID!=Nproc-1) );
-
-      // We have to copy the ghosts from the p1_rowmap as well as the new edge dofs.
-      // This needs to follow [E|T]petra ordering
-      size_t idx=pn_owned_dofs.size();
-      if(MyPID!=0) {
-        // Left side nodal
-        pn_col_dofs[idx]=p1_colmap->getGlobalElement(p1_rowmap->getLocalNumElements());
-        idx++;
-        // Left side, edge
-        for(size_t i=0; i<(size_t)(degree-1); i++) {
-          pn_col_dofs[idx] = go_edge_start-(degree-1)+i;
-          idx++;
-        }
-      }
-      if(MyPID!=Nproc-1) {
-        // Right side nodal
-        pn_col_dofs[idx]=p1_colmap->getGlobalElement(p1_colmap->getLocalNumElements()-1);
-        idx++;
-      }
-
-      RCP<const Map> pn_colmap = MapFactory::Build(lib,go_invalid,pn_col_dofs(),p1_rowmap->getIndexBase(),comm);
+  RCP<const Map> pn_colmap = MapFactory::Build(lib, go_invalid, pn_col_dofs(), p1_rowmap->getIndexBase(), comm);
 
 #if 0
       {
@@ -205,67 +197,74 @@ namespace MueLuTests {
       }
 #endif
 
-      // Fill elem_to_node using Kirby-style ordering
-      // Ownership rule: I own the element if I own the left node in said element
-      Kokkos::resize(elem_to_node,local_num_elements,degree+1);
-      auto elem_to_node_host = Kokkos::create_mirror_view(elem_to_node);
-      for(size_t i=0; i<local_num_elements; i++) {
-        // End Nodes
-        // NTS: This only works for lines
-        GO row_gid = pn_colmap->getGlobalElement(i);
-        GO col_gid = row_gid+1;
-        elem_to_node_host(i,0) = i;
-        elem_to_node_host(i,degree) = pn_colmap->getLocalElement(col_gid);
-
-        // Middle nodes (in local ids)
-        for(size_t j=0; j<(size_t)(degree-1); j++)
-          elem_to_node_host(i,1+j) = pn_colmap->getLocalElement(go_edge_start + i*(degree-1)+j);
+  // Fill elem_to_node using Kirby-style ordering
+  // Ownership rule: I own the element if I own the left node in said element
+  Kokkos::resize(elem_to_node, local_num_elements, degree + 1);
+  auto elem_to_node_host = Kokkos::create_mirror_view(elem_to_node);
+  for (size_t i = 0; i < local_num_elements; i++) {
+    // End Nodes
+    // NTS: This only works for lines
+    GO row_gid                   = pn_colmap->getGlobalElement(i);
+    GO col_gid                   = row_gid + 1;
+    elem_to_node_host(i, 0)      = i;
+    elem_to_node_host(i, degree) = pn_colmap->getLocalElement(col_gid);
+
+    // Middle nodes (in local ids)
+    for (size_t j = 0; j < (size_t)(degree - 1); j++)
+      elem_to_node_host(i, 1 + j) = pn_colmap->getLocalElement(go_edge_start + i * (degree - 1) + j);
+  }
+
+  // Since we're inserting off-proc, we really need to use the Epetra_FECrsMatrix here if we're in Epetra mode
+  RCP<Matrix> B;
+  if (lib == Xpetra::UseEpetra) {
+    AllocateEpetraFECrsMatrix(pn_rowmap, pn_colmap, B);
+  } else {
+    // Tpetra is easy
+    B = rcp(new CrsMatrixWrap(pn_rowmap, pn_colmap, global_num_elements));
+  }
+
+  // Assemble pseudo-poisson matrix
+  for (size_t i = 0; i < local_num_elements; i++) {
+    // Fill in a fake stiffness matrix
+    for (int j = 0; j < degree + 1; j++) {
+      // Dirichlet check
+      if ((j == 0 && pn_colmap->getGlobalElement(elem_to_node_host(i, j)) == 0) ||
+          (j == degree && pn_colmap->getGlobalElement(elem_to_node_host(i, j)) == global_num_nodes - 1)) {
+        // Stick a 1 on the diagonal
+        GO row_gid = pn_colmap->getGlobalElement(elem_to_node_host(i, j));
+        Teuchos::Array<GO> index(1);
+        index[0] = row_gid;
+        Teuchos::Array<SC> value(1);
+        value[0] = 1.0;
+        B->insertGlobalValues(row_gid, index(), value());
+        continue;
       }
 
-      // Since we're inserting off-proc, we really need to use the Epetra_FECrsMatrix here if we're in Epetra mode
-      RCP<Matrix> B;
-      if(lib==Xpetra::UseEpetra) {
-        AllocateEpetraFECrsMatrix(pn_rowmap, pn_colmap,B);
-      }
-      else {
-        // Tpetra is easy
-        B = rcp(new CrsMatrixWrap(pn_rowmap,pn_colmap,global_num_elements));
-      }
-
-
-      // Assemble pseudo-poisson matrix
-      for(size_t i=0; i<local_num_elements; i++) {
-        // Fill in a fake stiffness matrix
-        for(int j=0; j<degree+1; j++) {
-          // Dirichlet check
-          if( (j==0 && pn_colmap->getGlobalElement(elem_to_node_host(i,j)) == 0) ||
-              (j==degree &&  pn_colmap->getGlobalElement(elem_to_node_host(i,j)) ==  global_num_nodes-1))  {
-            // Stick a 1 on the diagonal
-            GO row_gid = pn_colmap->getGlobalElement(elem_to_node_host(i,j));
-            Teuchos::Array<GO> index(1); index[0]=row_gid;
-            Teuchos::Array<SC> value(1); value[0]=1.0;
-            B->insertGlobalValues(row_gid,index(),value());
-            continue;
-          }
-
-          GO rowj =  pn_colmap->getGlobalElement(elem_to_node_host(i,j));
-          for(int k=0; k<degree+1; k++) {
-            GO rowk =  pn_colmap->getGlobalElement(elem_to_node_host(i,k));
-            Teuchos::Array<GO> index(1); index[0] = rowk;
-            Teuchos::Array<SC> value(1);
-            if(j==0 && k==0)                value[0]=1.0;
-            else if(j==0 && k==degree)      value[0]=-1.0;
-            else if(j==degree && k==0)      value[0]=-1.0;
-            else if(j==degree && k==degree) value[0]=1.0;
-            else if(j==k)                   value[0] = (degree+1) / 100.0;
-            else                            value[0] = -1.0/100;
-            B->insertGlobalValues(rowj,index(),value());
-          }
-        }
+      GO rowj = pn_colmap->getGlobalElement(elem_to_node_host(i, j));
+      for (int k = 0; k < degree + 1; k++) {
+        GO rowk = pn_colmap->getGlobalElement(elem_to_node_host(i, k));
+        Teuchos::Array<GO> index(1);
+        index[0] = rowk;
+        Teuchos::Array<SC> value(1);
+        if (j == 0 && k == 0)
+          value[0] = 1.0;
+        else if (j == 0 && k == degree)
+          value[0] = -1.0;
+        else if (j == degree && k == 0)
+          value[0] = -1.0;
+        else if (j == degree && k == degree)
+          value[0] = 1.0;
+        else if (j == k)
+          value[0] = (degree + 1) / 100.0;
+        else
+          value[0] = -1.0 / 100;
+        B->insertGlobalValues(rowj, index(), value());
       }
+    }
+  }
 
-      B->fillComplete(pn_rowmap,pn_rowmap);
-      Kokkos::deep_copy(elem_to_node, elem_to_node_host);
+  B->fillComplete(pn_rowmap, pn_rowmap);
+  Kokkos::deep_copy(elem_to_node, elem_to_node_host);
 
 #if 0
       std::cout<<"*** Pseudo Poisson ***"<<std::endl;
@@ -282,14 +281,13 @@ namespace MueLuTests {
       printf("***\n");
 #endif
 
+  return B;
+}  // Build1DPseudoPoissonHigherOrder()
 
-      return B;
-    } // Build1DPseudoPoissonHigherOrder()
-
-} //namespace TestHelpers
+}  // namespace TestHelpers
 
-} //namespace MueLuTests
+}  // namespace MueLuTests
 
-#endif //ifdef HAVE_MUELU_INTREPID2
+#endif  // ifdef HAVE_MUELU_INTREPID2
 
-#endif // ifndef MUELU_TEST_HELPERS_HO_H
+#endif  // ifndef MUELU_TEST_HELPERS_HO_H
diff --git a/packages/muelu/test/unit_tests/MueLu_Test_ETI.hpp b/packages/muelu/test/unit_tests/MueLu_Test_ETI.hpp
index 7ade298942ba..eb26021e3933 100644
--- a/packages/muelu/test/unit_tests/MueLu_Test_ETI.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_Test_ETI.hpp
@@ -69,7 +69,6 @@
 #error "The macro MUELU_AUTOMATIC_TEST_ETI_NAME was not defined"
 #endif
 
-
 // This function serves to manage the ETI (or not) for tests that are *not* unit tests.  This
 // allows us to isolate this particular bit of functionality, rather than have it cut-and-paste
 // duplicated over half of the testing tree.  See documentation at
@@ -79,9 +78,9 @@
 //              and only one instantiation (depending on the choice of --instantiation).
 //              It does NOT test all enabled configurations as the unit tests would do.
 bool Automatic_Test_ETI(int argc, char *argv[]) {
+  using MueLu::Exceptions::RuntimeError;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using MueLu::Exceptions::RuntimeError;
 
   // MPI initialization using Teuchos
   Teuchos::GlobalMPISession mpiSession(&argc, &argv, NULL);
@@ -107,39 +106,44 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
   // created on the first call to these libraries, and that can mess
   // up timings.
   KokkosKernels::Experimental::Controls controls;
-# ifdef KOKKOSKERNELS_ENABLE_TPL_CUBLAS
+#ifdef KOKKOSKERNELS_ENABLE_TPL_CUBLAS
   controls.getCublasHandle();
-# endif
-# ifdef KOKKOSKERNELS_ENABLE_TPL_CUSPARSE
+#endif
+#ifdef KOKKOSKERNELS_ENABLE_TPL_CUSPARSE
   controls.getCusparseHandle();
-# endif
+#endif
   Kokkos::fence();
   bool success = true;
   bool verbose = true;
   try {
     // Parameters
     Teuchos::CommandLineProcessor clp(false);
-    std::string node   = "";    clp.setOption("node",               &node,   "node type (serial | openmp | cuda | hip)");
-    bool        config = false; clp.setOption("config", "noconfig", &config, "display kokkos configuration");
+    std::string node = "";
+    clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+    bool config = false;
+    clp.setOption("config", "noconfig", &config, "display kokkos configuration");
 #ifdef HAVE_TEUCHOS_STACKTRACE
-    bool stacktrace = true;     clp.setOption("stacktrace", "nostacktrace", &stacktrace, "display stacktrace");
+    bool stacktrace = true;
+    clp.setOption("stacktrace", "nostacktrace", &stacktrace, "display stacktrace");
 #endif
 
-    bool timedeepcopy = false;   clp.setOption("timedeepcopy", "notimedeepcopy", &timedeepcopy, "instrument Kokkos::deep_copy() with Teuchos timers.  This can also be done with by setting the environment variable TPETRA_TIME_KOKKOS_DEEP_COPY=ON");
-    bool timefence = false;   clp.setOption("timefence", "notimefence", &timefence, "instrument Kokkos::fence() with Teuchos timers.  This can also be done with by setting the environment variable TPETRA_TIME_KOKKOS_FENCE=ON");
+    bool timedeepcopy = false;
+    clp.setOption("timedeepcopy", "notimedeepcopy", &timedeepcopy, "instrument Kokkos::deep_copy() with Teuchos timers.  This can also be done with by setting the environment variable TPETRA_TIME_KOKKOS_DEEP_COPY=ON");
+    bool timefence = false;
+    clp.setOption("timefence", "notimefence", &timefence, "instrument Kokkos::fence() with Teuchos timers.  This can also be done with by setting the environment variable TPETRA_TIME_KOKKOS_FENCE=ON");
     Xpetra::Parameters xpetraParameters(clp);
 
     clp.recogniseAllOptions(false);
     switch (clp.parse(argc, argv, NULL)) {
-      case Teuchos::CommandLineProcessor::PARSE_ERROR:                return EXIT_FAILURE;
+      case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
       case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:         break;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: break;
     }
 
-    if(timedeepcopy)
+    if (timedeepcopy)
       Tpetra::Details::AddKokkosDeepCopyToTimeMonitor(true);
-    if(timefence)
+    if (timefence)
       Tpetra::Details::AddKokkosFenceToTimeMonitor(true);
 
 #ifdef HAVE_TEUCHOS_STACKTRACE
@@ -155,52 +159,52 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
       //      We might need a feature that allows to run Epetra/Tpetra only
       //      We still need to make sure that the test compiles (i.e., we
       //      need some preprocessor flags/macros RUN_WITH_EPETRA and RUN_WITH_TPETRA
-#  if defined(HAVE_MUELU_INST_DOUBLE_INT_INT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+#if defined(HAVE_MUELU_INST_DOUBLE_INT_INT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
       // Both Epetra and Tpetra (with double, int, int) enabled
-      return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
-#  else
+      return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
+#else
       *out << "Skip running with Epetra since both Epetra and Tpetra are enabled but Tpetra is not instantiated on double, int, int." << std::endl;
-#  endif // end Tpetra instantiated on double, int, int
+#endif  // end Tpetra instantiated on double, int, int
 #else
       throw RuntimeError("Epetra is not available");
 #endif
     }
 
     if (lib == Xpetra::UseTpetra) {
-# ifdef HAVE_MUELU_EXPLICIT_INSTANTIATION
+#ifdef HAVE_MUELU_EXPLICIT_INSTANTIATION
       auto inst = xpetraParameters.GetInstantiation();
-# endif
+#endif
       if (node == "") {
         typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Node::execution_space().print_configuration(*out, true/*details*/);
+          Node::execution_space().print_configuration(*out, true /*details*/);
         }
 
 #ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
 #else
-#  if defined(HAVE_MUELU_INST_DOUBLE_INT_INT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+#if defined(HAVE_MUELU_INST_DOUBLE_INT_INT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp,  lib, argc, argv);
-#  endif
-#  if defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp,  lib, argc, argv);
-#  endif
-#  if defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT) || defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp,  lib, argc, argv);
-#  endif
-#  if defined(HAVE_MUELU_INST_COMPLEX_INT_INT) || defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_MUELU_INST_COMPLEX_INT_INT) || defined(HAVE_TPETRA_INST_COMPLEX_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#  endif
-#  if defined(HAVE_MUELU_INST_FLOAT_INT_INT) || defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_MUELU_INST_FLOAT_INT_INT) || defined(HAVE_TPETRA_INST_FLOAT) && defined(HAVE_TPETRA_INST_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#  endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable Default instantiation");
 #endif
       } else if (node == "serial") {
@@ -209,34 +213,34 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Kokkos::Serial().print_configuration(*out, true/*details*/);
+          Kokkos::Serial().print_configuration(*out, true /*details*/);
         }
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp,  lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#    endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable Serial instantiation");
-#  endif
+#endif
 #else
         throw RuntimeError("Serial node type is disabled");
 #endif
@@ -246,35 +250,35 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Kokkos::OpenMP().print_configuration(*out, true/*details*/);
+          Kokkos::OpenMP().print_configuration(*out, true /*details*/);
           *out << "OpenMP Max Threads = " << omp_get_max_threads() << std::endl;
         }
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp,  lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp, lib,  argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#    endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable OpenMP instantiation");
-#  endif
+#endif
 #else
         throw RuntimeError("OpenMP node type is disabled");
 #endif
@@ -284,111 +288,111 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Kokkos::Cuda().print_configuration(*out, true/*details*/);
+          Kokkos::Cuda().print_configuration(*out, true /*details*/);
         }
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#    endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable Cuda instantiation");
-#  endif
+#endif
 #else
         throw RuntimeError("CUDA node type is disabled");
 #endif
       } else if (node == "hip") {
 #ifdef KOKKOS_ENABLE_HIP
-	typedef Tpetra::KokkosCompat::KokkosHIPWrapperNode Node;
+        typedef Tpetra::KokkosCompat::KokkosHIPWrapperNode Node;
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Kokkos::HIP().print_configuration(*out, true/*details*/);
+          Kokkos::HIP().print_configuration(*out, true /*details*/);
         }
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#    endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable HIP instantiation");
-#  endif
+#endif
 #else
         throw RuntimeError("HIP node type is disabled");
 #endif
       } else if (node == "sycl") {
 #ifdef KOKKOS_ENABLE_SYCL
-	typedef Tpetra::KokkosCompat::KokkosSYCLWrapperNode Node;
+        typedef Tpetra::KokkosCompat::KokkosSYCLWrapperNode Node;
 
         if (config) {
           *out << "Node type: " << Node::execution_space::name() << std::endl;
-          Kokkos::Experimental::SYCL().print_configuration(*out, true/*details*/);
+          Kokkos::Experimental::SYCL().print_configuration(*out, true /*details*/);
         }
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_INT)
         if (inst == Xpetra::DOUBLE_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,int,Node> (clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_DOUBLE_INT_LONGLONGINT)
         if (inst == Xpetra::DOUBLE_INT_LONGLONGINT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<double,int,long long,Node>(clp, lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<double, int, long long, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_COMPLEX_INT_INT)
         if (inst == Xpetra::COMPLEX_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>,int,int,Node>(clp,  lib, argc, argv);
-#    endif
-#    if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<std::complex<double>, int, int, Node>(clp, lib, argc, argv);
+#endif
+#if defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_MUELU_INST_FLOAT_INT_INT)
         if (inst == Xpetra::FLOAT_INT_INT)
-          return MUELU_AUTOMATIC_TEST_ETI_NAME<float,int,int,Node>(clp,  lib, argc, argv);
-#    endif
+          return MUELU_AUTOMATIC_TEST_ETI_NAME<float, int, int, Node>(clp, lib, argc, argv);
+#endif
         throw RuntimeError("Found no suitable SYCL instantiation");
-#  endif
+#endif
 #else
         throw RuntimeError("SYCL node type is disabled");
-#endif	
+#endif
       } else {
         throw RuntimeError("Unrecognized node type");
       }
@@ -398,7 +402,7 @@ bool Automatic_Test_ETI(int argc, char *argv[]) {
 
   Kokkos::finalize();
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 // Undefine the function macro
diff --git a/packages/muelu/test/unit_tests/MueLu_UnitTestHelpers.hpp b/packages/muelu/test/unit_tests/MueLu_UnitTestHelpers.hpp
index 1804c4bffb74..7865b2ad344a 100644
--- a/packages/muelu/test/unit_tests/MueLu_UnitTestHelpers.hpp
+++ b/packages/muelu/test/unit_tests/MueLu_UnitTestHelpers.hpp
@@ -53,105 +53,95 @@
 #include "Teuchos_UnitTestHelpers.hpp"
 
 /** \brief Basic unit test creation macro for templated code on five template parameters. */
-#define TEUCHOS_UNIT_TEST_TEMPLATE_5_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5) \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5>         \
-  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase \
-        { \
-  public: \
-    TEST_GROUP##_##TEST_NAME##_UnitTest( \
-      const std::string& type1Name, \
-      const std::string& type2Name, \
-      const std::string& type3Name, \
-      const std::string& type4Name, \
-      const std::string& type5Name \
-       ) \
-      :Teuchos::UnitTestBase( \
-         std::string(#TEST_GROUP)+"_"+type1Name+"_"+type2Name+"_"+type3Name+"_"+type4Name+"_"+type5Name, #TEST_NAME ) \
-    {} \
-    void runUnitTestImpl( Teuchos::FancyOStream &out, bool &success ) const; \
-    virtual std::string unitTestFile() const { return __FILE__; } \
-    virtual long int unitTestFileLineNumber() const { return __LINE__; } \
-  }; \
-  \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5>         \
-  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1,TYPE2,TYPE3,TYPE4, TYPE5>::runUnitTestImpl( \
-    Teuchos::FancyOStream &out, bool &success ) const \
+#define TEUCHOS_UNIT_TEST_TEMPLATE_5_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5)                                            \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5>                                                                   \
+  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase {                                                                   \
+   public:                                                                                                                                     \
+    TEST_GROUP##_##TEST_NAME##_UnitTest(                                                                                                       \
+        const std::string& type1Name,                                                                                                          \
+        const std::string& type2Name,                                                                                                          \
+        const std::string& type3Name,                                                                                                          \
+        const std::string& type4Name,                                                                                                          \
+        const std::string& type5Name)                                                                                                          \
+      : Teuchos::UnitTestBase(                                                                                                                 \
+            std::string(#TEST_GROUP) + "_" + type1Name + "_" + type2Name + "_" + type3Name + "_" + type4Name + "_" + type5Name, #TEST_NAME) {} \
+    void runUnitTestImpl(Teuchos::FancyOStream& out, bool& success) const;                                                                     \
+    virtual std::string unitTestFile() const { return __FILE__; }                                                                              \
+    virtual long int unitTestFileLineNumber() const { return __LINE__; }                                                                       \
+  };                                                                                                                                           \
+                                                                                                                                               \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5>                                                                   \
+  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5>::runUnitTestImpl(                                                \
+      Teuchos::FancyOStream& out, bool& success) const
 
 /** \brief Template instantiation for five templated types. */
 #define TEUCHOS_UNIT_TEST_TEMPLATE_5_INSTANT(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5) \
-  \
-  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5 >; \
-  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5 >     \
-  instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TEST_NAME##_UnitTest(#TYPE1,#TYPE2,#TYPE3,#TYPE4,#TYPE5);
+                                                                                                       \
+  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5>;               \
+  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5>                               \
+      instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TEST_NAME##_UnitTest(#TYPE1, #TYPE2, #TYPE3, #TYPE4, #TYPE5);
 
 // new unit test macros with support for 6 template parameters
 
 /** \brief Basic unit test creation macro for templated code on six template parameters. */
-#define TEUCHOS_UNIT_TEST_TEMPLATE_6_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6) \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6>         \
-  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase \
-        { \
-  public: \
-    TEST_GROUP##_##TEST_NAME##_UnitTest( \
-      const std::string& type1Name, \
-      const std::string& type2Name, \
-      const std::string& type3Name, \
-      const std::string& type4Name, \
-      const std::string& type5Name, \
-      const std::string& type6Name  \
-       ) \
-      :Teuchos::UnitTestBase( \
-         std::string(#TEST_GROUP)+"_"+type3Name+"_"+type4Name+"_"+type5Name+"_"+type6Name, #TEST_NAME ) \
-    {} \
-    void runUnitTestImpl( Teuchos::FancyOStream &out, bool &success ) const; \
-    virtual std::string unitTestFile() const { return __FILE__; } \
-    virtual long int unitTestFileLineNumber() const { return __LINE__; } \
-  }; \
-  \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6>         \
-  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1,TYPE2,TYPE3,TYPE4,TYPE5,TYPE6>::runUnitTestImpl( \
-    Teuchos::FancyOStream &out, bool &success ) const \
+#define TEUCHOS_UNIT_TEST_TEMPLATE_6_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6)                   \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6>                                    \
+  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase {                                                 \
+   public:                                                                                                                   \
+    TEST_GROUP##_##TEST_NAME##_UnitTest(                                                                                     \
+        const std::string& type1Name,                                                                                        \
+        const std::string& type2Name,                                                                                        \
+        const std::string& type3Name,                                                                                        \
+        const std::string& type4Name,                                                                                        \
+        const std::string& type5Name,                                                                                        \
+        const std::string& type6Name)                                                                                        \
+      : Teuchos::UnitTestBase(                                                                                               \
+            std::string(#TEST_GROUP) + "_" + type3Name + "_" + type4Name + "_" + type5Name + "_" + type6Name, #TEST_NAME) {} \
+    void runUnitTestImpl(Teuchos::FancyOStream& out, bool& success) const;                                                   \
+    virtual std::string unitTestFile() const { return __FILE__; }                                                            \
+    virtual long int unitTestFileLineNumber() const { return __LINE__; }                                                     \
+  };                                                                                                                         \
+                                                                                                                             \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6>                                    \
+  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6>::runUnitTestImpl(                       \
+      Teuchos::FancyOStream& out, bool& success) const
 
 /** \brief Template instantiation for six templated types. */
 #define TEUCHOS_UNIT_TEST_TEMPLATE_6_INSTANT(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6) \
-  \
-  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6 >; \
-  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6 >     \
-  instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TYPE6##_##TEST_NAME##_UnitTest(#TYPE1,#TYPE2,#TYPE3,#TYPE4,#TYPE5,#TYPE6);
-
+                                                                                                              \
+  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6>;               \
+  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6>                               \
+      instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TYPE6##_##TEST_NAME##_UnitTest(#TYPE1, #TYPE2, #TYPE3, #TYPE4, #TYPE5, #TYPE6);
 
 /** \brief Basic unit test creation macro for templated code on seven template parameters. */
-#define TEUCHOS_UNIT_TEST_TEMPLATE_7_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7) \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6, class TYPE7 >         \
-  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase \
-        { \
-  public: \
-    TEST_GROUP##_##TEST_NAME##_UnitTest( \
-      const std::string& type1Name, \
-      const std::string& type2Name, \
-      const std::string& type3Name, \
-      const std::string& type4Name, \
-      const std::string& type5Name, \
-      const std::string& type6Name, \
-      const std::string& type7Name  \
-       ) \
-      :Teuchos::UnitTestBase( \
-         std::string(#TEST_GROUP)+"_"+type4Name+"_"+type5Name+"_"+type6Name+"_"+type7Name, #TEST_NAME ) \
-    {} \
-    void runUnitTestImpl( Teuchos::FancyOStream &out, bool &success ) const; \
-    virtual std::string unitTestFile() const { return __FILE__; } \
-    virtual long int unitTestFileLineNumber() const { return __LINE__; } \
-  }; \
-  \
-  template<class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6, class TYPE7>         \
-  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1,TYPE2,TYPE3,TYPE4,TYPE5,TYPE6,TYPE7>::runUnitTestImpl( \
-    Teuchos::FancyOStream &out, bool &success ) const \
+#define TEUCHOS_UNIT_TEST_TEMPLATE_7_DECL(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7)            \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6, class TYPE7>                       \
+  class TEST_GROUP##_##TEST_NAME##_UnitTest : public Teuchos::UnitTestBase {                                                 \
+   public:                                                                                                                   \
+    TEST_GROUP##_##TEST_NAME##_UnitTest(                                                                                     \
+        const std::string& type1Name,                                                                                        \
+        const std::string& type2Name,                                                                                        \
+        const std::string& type3Name,                                                                                        \
+        const std::string& type4Name,                                                                                        \
+        const std::string& type5Name,                                                                                        \
+        const std::string& type6Name,                                                                                        \
+        const std::string& type7Name)                                                                                        \
+      : Teuchos::UnitTestBase(                                                                                               \
+            std::string(#TEST_GROUP) + "_" + type4Name + "_" + type5Name + "_" + type6Name + "_" + type7Name, #TEST_NAME) {} \
+    void runUnitTestImpl(Teuchos::FancyOStream& out, bool& success) const;                                                   \
+    virtual std::string unitTestFile() const { return __FILE__; }                                                            \
+    virtual long int unitTestFileLineNumber() const { return __LINE__; }                                                     \
+  };                                                                                                                         \
+                                                                                                                             \
+  template <class TYPE1, class TYPE2, class TYPE3, class TYPE4, class TYPE5, class TYPE6, class TYPE7>                       \
+  void TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7>::runUnitTestImpl(                \
+      Teuchos::FancyOStream& out, bool& success) const
 
 /** \brief Template instantiation for seven templated types. */
 #define TEUCHOS_UNIT_TEST_TEMPLATE_7_INSTANT(TEST_GROUP, TEST_NAME, TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7) \
-  \
-  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7 >; \
-  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7 >     \
-  instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TYPE6##_##TYPE7##_##TEST_NAME##_UnitTest(#TYPE1,#TYPE2,#TYPE3,#TYPE4,#TYPE5,#TYPE6,#TYPE7);
+                                                                                                                     \
+  template class TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7>;               \
+  TEST_GROUP##_##TEST_NAME##_UnitTest<TYPE1, TYPE2, TYPE3, TYPE4, TYPE5, TYPE6, TYPE7>                               \
+      instance_##TEST_GROUP##_##TYPE1##_##TYPE2##_##TYPE3##_##TYPE4##_##TYPE5##_##TYPE6##_##TYPE7##_##TEST_NAME##_UnitTest(#TYPE1, #TYPE2, #TYPE3, #TYPE4, #TYPE5, #TYPE6, #TYPE7);
 
 #endif  // MUELU_UNIT_TEST_HELPERS_HPP
diff --git a/packages/muelu/test/unit_tests/MueLu_UnitTests.cpp b/packages/muelu/test/unit_tests/MueLu_UnitTests.cpp
index 094457b5665c..bd355b9a08fa 100644
--- a/packages/muelu/test/unit_tests/MueLu_UnitTests.cpp
+++ b/packages/muelu/test/unit_tests/MueLu_UnitTests.cpp
@@ -62,15 +62,14 @@ specific unit test suites.
 
 #include "MueLu_TestHelpers.hpp"
 
-int main(int argc, char* argv[])
-{
+int main(int argc, char* argv[]) {
   Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
-  Kokkos::initialize(argc , argv);
+  Kokkos::initialize(argc, argv);
 
   bool success = false;
   bool verbose = true;
-  int ierr = -1;
+  int ierr     = -1;
   try {
     // Note: the command line parameter --linAlgebra= is take into account.
     // Xpetra parameters are added to the Teuchos::CommandLineProcessor of Teuchos::UnitTestRepository in MueLu_TestHelpers.cpp
@@ -80,25 +79,26 @@ int main(int argc, char* argv[])
 
     int mypid = comm->getRank();
 
-    if (mypid  == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
-    for (int i = 0; i <comm->getSize(); i++) {
-      if (i == mypid ) {
+    if (mypid == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
+    for (int i = 0; i < comm->getSize(); i++) {
+      if (i == mypid) {
         char buf[80];
         char hostname[80];
         gethostname(hostname, sizeof(hostname));
         int pid = getpid();
         sprintf(buf, "Host: %s\tMPI rank: %d,\tPID: %d\n\tattach %d\n\tcontinue\n",
-            hostname, mypid, pid, pid);
-        printf("%s\n",buf);
+                hostname, mypid, pid, pid);
+        printf("%s\n", buf);
         fflush(stdout);
         sleep(1);
       }
     }
 
     if (mypid == 0) {
-      printf( "** Enter a character to continue > "); fflush(stdout);
+      printf("** Enter a character to continue > ");
+      fflush(stdout);
       char go = ' ';
-      scanf("%c",&go);
+      scanf("%c", &go);
     }
     comm->barrier();
 #endif
diff --git a/packages/muelu/test/unit_tests/MultiVectorTransferFactory.cpp b/packages/muelu/test/unit_tests/MultiVectorTransferFactory.cpp
index cbc43dbd291f..5d624fbf6db4 100644
--- a/packages/muelu/test/unit_tests/MultiVectorTransferFactory.cpp
+++ b/packages/muelu/test/unit_tests/MultiVectorTransferFactory.cpp
@@ -46,7 +46,6 @@
 #include <Teuchos_UnitTestHarness.hpp>
 #include <Teuchos_ScalarTraits.hpp>
 
-
 #include <MueLu_config.hpp>
 
 #include <MueLu_TestHelpers.hpp>
@@ -66,182 +65,176 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Factory> TentativePFact = rcp(new TentativePFactory());
-    RCP<Factory> TentativeRFact = rcp(new TransPFactory());  // Use Ptent for coordinate projection
-
-    RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("Coordinates"));
-    mvtf->SetFactory("R", TentativeRFact);
-
-    TEST_EQUALITY(mvtf != Teuchos::null, true);
-  } // Constructor test
-
-  //------------------------------------------------------------------------------------------
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    out << "Tests the action of the transfer factory on a vector.  In this test, the transfer is the tentative" << std::endl;
-    out << "prolongator, and the vector is all ones.  So the norm of the resulting coarse grid vector should be" << std::endl;
-    out << "equal to the number of fine degrees of freedom." << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-    typedef typename Xpetra::MultiVector<magnitude_type,LO,GO,NO> RealValuedMultiVector;
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    GO nx = 199;
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    RCP<MultiVector> fineOnes = MultiVectorFactory::Build(A->getRowMap(),1);
-    fineOnes->putScalar(1.0);
-    fineLevel.Set("onesVector",fineOnes);
-    fineLevel.Set("Coordinates", coordinates);
-
-    RCP<TentativePFactory>    TentativePFact = rcp(new TentativePFactory());
-    RCP<TransPFactory>        RFact = rcp(new TransPFactory());
-
-    RCP<FactoryManager> M = rcp(new FactoryManager());
-    M->SetKokkosRefactor(false);
-    M->SetFactory("P", TentativePFact);
-    M->SetFactory("Ptent", TentativePFact);
-    M->SetFactory("R", RFact);
-    //    fineLevel.SetFactoryManager(M);
-    coarseLevel.SetFactoryManager(M);
-
-    RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("onesVector"));
-    mvtf->SetFactory("R",RFact);
-
-    coarseLevel.Request("onesVector",mvtf.get());
-    coarseLevel.Request("R",RFact.get());
-    coarseLevel.Request("P",TentativePFact.get());
-
-    mvtf->Build(fineLevel,coarseLevel);
-
-    RCP<MultiVector> coarseOnes = coarseLevel.Get<RCP<MultiVector> >("onesVector",mvtf.get());
-    Teuchos::Array<magnitude_type> vn(1);
-    coarseOnes->norm2(vn);
-    TEST_FLOATING_EQUALITY(vn[0]*vn[0],(Teuchos::as<magnitude_type>(fineOnes->getGlobalLength())),1e-12);
-  } // Build test
-
-  //------------------------------------------------------------------------------------------
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, ThreeLevels, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra,"Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra,"Ifpack2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-
-    out << "Tests usage on a three-level hierarchy." << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    GO nx = 199;
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-
-    // Set up three level hierarchy.
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<Level> fineLevel = H->GetLevel();
-    fineLevel->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    fineLevel->Set("A",A);                       // set fine level matrix
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),1);
-    nullSpace->putScalar( (SC) 1.0);
-    fineLevel->Set("Nullspace", nullSpace);       // set null space information for finest level
-    fineLevel->Set("Coordinates", coordinates);   // set coordinates on finest level
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<TentativePFactory> PFact  = rcp(new TentativePFactory()); //just using plain aggregation
-    RCP<Factory>           RFact  = rcp(new TransPFactory());
-    RCP<RAPFactory>        AcFact = rcp(new RAPFactory());
-    H->SetMaxCoarseSize(1);
-
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LO) 1);
-    smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = Teuchos::null;
-    AcFact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("P", PFact);
-    M.SetFactory("Ptent", PFact); // for nullspace
-    M.SetFactory("R", RFact);
-    M.SetFactory("A", AcFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", SmooFact); // This line avoid dependency to Amesos/Amesos2 for this test.
-
-    //set up the transfer factory
-    RCP<MultiVector> fineOnes = MultiVectorFactory::Build(A->getRowMap(),1);
-    fineOnes->putScalar(1.0);
-    fineLevel->Set("onesVector",fineOnes);
-    RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("onesVector"));
-    mvtf->SetFactory("R",RFact);
-    M.SetFactory("onesVector",mvtf);
-    AcFact->AddTransferFactory(mvtf);
-
-    int maxLevels = 3;
-    H->Setup(M, 0, maxLevels);
-
-    /*
-    //FIXME we probably need to do some requests....
-    coarseLevel.Request("onesVector",mvtf.get());
-    coarseLevel.Request("R",RFact.get());
-    coarseLevel.Request("P",TentativePFact.get());
-    */
-
-    /*
-       RCP<MultiVector> coarseOnes = coarseLevel.Get<RCP<MultiVector> >("onesVector",mvtf.get());
-       Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> vn(1);
-       coarseOnes->norm2(vn);
-
-       TEST_FLOATING_EQUALITY(vn[0]*vn[0],((SC)fineOnes->getGlobalLength()),1e-12);
-       */
-  } // ThreeLevels
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Factory> TentativePFact = rcp(new TentativePFactory());
+  RCP<Factory> TentativeRFact = rcp(new TransPFactory());  // Use Ptent for coordinate projection
+
+  RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("Coordinates"));
+  mvtf->SetFactory("R", TentativeRFact);
+
+  TEST_EQUALITY(mvtf != Teuchos::null, true);
+}  // Constructor test
+
+//------------------------------------------------------------------------------------------
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  out << "Tests the action of the transfer factory on a vector.  In this test, the transfer is the tentative" << std::endl;
+  out << "prolongator, and the vector is all ones.  So the norm of the resulting coarse grid vector should be" << std::endl;
+  out << "equal to the number of fine degrees of freedom." << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  typedef typename Xpetra::MultiVector<magnitude_type, LO, GO, NO> RealValuedMultiVector;
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  GO nx         = 199;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  RCP<MultiVector> fineOnes = MultiVectorFactory::Build(A->getRowMap(), 1);
+  fineOnes->putScalar(1.0);
+  fineLevel.Set("onesVector", fineOnes);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  RCP<TransPFactory> RFact              = rcp(new TransPFactory());
+
+  RCP<FactoryManager> M = rcp(new FactoryManager());
+  M->SetKokkosRefactor(false);
+  M->SetFactory("P", TentativePFact);
+  M->SetFactory("Ptent", TentativePFact);
+  M->SetFactory("R", RFact);
+  //    fineLevel.SetFactoryManager(M);
+  coarseLevel.SetFactoryManager(M);
+
+  RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("onesVector"));
+  mvtf->SetFactory("R", RFact);
+
+  coarseLevel.Request("onesVector", mvtf.get());
+  coarseLevel.Request("R", RFact.get());
+  coarseLevel.Request("P", TentativePFact.get());
+
+  mvtf->Build(fineLevel, coarseLevel);
+
+  RCP<MultiVector> coarseOnes = coarseLevel.Get<RCP<MultiVector> >("onesVector", mvtf.get());
+  Teuchos::Array<magnitude_type> vn(1);
+  coarseOnes->norm2(vn);
+  TEST_FLOATING_EQUALITY(vn[0] * vn[0], (Teuchos::as<magnitude_type>(fineOnes->getGlobalLength())), 1e-12);
+}  // Build test
+
+//------------------------------------------------------------------------------------------
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MultiVectorTransferFactory, ThreeLevels, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
+#endif
+#if !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+
+  out << "Tests usage on a three-level hierarchy." << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  GO nx         = 199;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Set up three level hierarchy.
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> fineLevel = H->GetLevel();
+  fineLevel->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  fineLevel->Set("A", A);  // set fine level matrix
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar((SC)1.0);
+  fineLevel->Set("Nullspace", nullSpace);      // set null space information for finest level
+  fineLevel->Set("Coordinates", coordinates);  // set coordinates on finest level
+
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<TentativePFactory> PFact = rcp(new TentativePFactory());  // just using plain aggregation
+  RCP<Factory> RFact           = rcp(new TransPFactory());
+  RCP<RAPFactory> AcFact       = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LO)1);
+  smootherParamList.set("relaxation: damping factor", (SC)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = Teuchos::null;
+  AcFact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("P", PFact);
+  M.SetFactory("Ptent", PFact);  // for nullspace
+  M.SetFactory("R", RFact);
+  M.SetFactory("A", AcFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", SmooFact);  // This line avoid dependency to Amesos/Amesos2 for this test.
+
+  // set up the transfer factory
+  RCP<MultiVector> fineOnes = MultiVectorFactory::Build(A->getRowMap(), 1);
+  fineOnes->putScalar(1.0);
+  fineLevel->Set("onesVector", fineOnes);
+  RCP<MueLu::MultiVectorTransferFactory<SC, LO, GO, NO> > mvtf = rcp(new MueLu::MultiVectorTransferFactory<SC, LO, GO, NO>("onesVector"));
+  mvtf->SetFactory("R", RFact);
+  M.SetFactory("onesVector", mvtf);
+  AcFact->AddTransferFactory(mvtf);
+
+  int maxLevels = 3;
+  H->Setup(M, 0, maxLevels);
+
+  /*
+  //FIXME we probably need to do some requests....
+  coarseLevel.Request("onesVector",mvtf.get());
+  coarseLevel.Request("R",RFact.get());
+  coarseLevel.Request("P",TentativePFact.get());
+  */
+
+  /*
+     RCP<MultiVector> coarseOnes = coarseLevel.Get<RCP<MultiVector> >("onesVector",mvtf.get());
+     Teuchos::Array<Teuchos::ScalarTraits<SC>::magnitudeType> vn(1);
+     coarseOnes->norm2(vn);
+
+     TEST_FLOATING_EQUALITY(vn[0]*vn[0],((SC)fineOnes->getGlobalLength()),1e-12);
+     */
+}  // ThreeLevels
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                         \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MultiVectorTransferFactory, Constructor, Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MultiVectorTransferFactory, Build, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MultiVectorTransferFactory, Build, Scalar, LO, GO, Node)       \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MultiVectorTransferFactory, ThreeLevels, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/NullspaceFactory.cpp b/packages/muelu/test/unit_tests/NullspaceFactory.cpp
index 2dbddbbc8ec7..63410115ce92 100644
--- a/packages/muelu/test/unit_tests/NullspaceFactory.cpp
+++ b/packages/muelu/test/unit_tests/NullspaceFactory.cpp
@@ -61,91 +61,96 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NullspaceFactory,3DElasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NullspaceFactory, 3DElasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  GO nx = 2, ny = 3, nz = 4, mx = comm->getSize(), my = 1, mz = 1;
+  ;
+  LO blkSize = 3;
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("nz", nz);
+  matrixList.set("mx", mx);
+  matrixList.set("my", my);
+  matrixList.set("mz", mz);
+  matrixList.set("matrixType", "Elasticity3D");
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::BuildMatrix(matrixList, TestHelpers::Parameters::getLib());
+  Op->SetFixedBlockSize(blkSize);
+
+  fineLevel.Set("A", Op);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  galeriList.set("nz", nz);
+  RCP<Matrix> fake                       = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx * ny * nz);  // need scalar PDE to produce proper coordinate array
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("3D", fake->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  RCP<NullspaceFactory> nspace = rcp(new NullspaceFactory());
+  fineLevel.Request("Nullspace", &(*nspace));
+  Teuchos::ParameterList nspaceList = *(nspace->GetValidParameterList());
+  nspaceList.set("nullspace: calculate rotations", true);
+  nspace->SetParameterList(nspaceList);
+  fineLevel.Request(*nspace);
+  nspace->Build(fineLevel);
+  RCP<MultiVector> nullSpace = fineLevel.Get<RCP<MultiVector> >("Nullspace", nspace.get());
+
+  ArrayRCP<real_type> xvals          = coordinates->getDataNonConst(0);
+  ArrayRCP<real_type> yvals          = coordinates->getDataNonConst(1);
+  ArrayRCP<real_type> zvals          = coordinates->getDataNonConst(2);
+  size_t nullDim                     = 6;
+  RCP<MultiVector> handCompNullSpace = MultiVectorFactory::Build(Op->getRowMap(), nullDim);
+  handCompNullSpace->putScalar(0.0);
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    GO nx = 2, ny = 3, nz = 4, mx = comm->getSize(), my = 1, mz = 1; ;
-    LO blkSize = 3;
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", nx);
-    matrixList.set("ny", ny);
-    matrixList.set("nz", nz);
-    matrixList.set("mx", mx);
-    matrixList.set("my", my);
-    matrixList.set("mz", mz);
-    matrixList.set("matrixType","Elasticity3D");
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::BuildMatrix(matrixList,TestHelpers::Parameters::getLib());
-    Op->SetFixedBlockSize(blkSize);
-
-    fineLevel.Set("A",Op);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    galeriList.set("nz", nz);
-    RCP<Matrix> fake = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx*ny*nz); // need scalar PDE to produce proper coordinate array
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("3D", fake->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    RCP<NullspaceFactory> nspace = rcp(new NullspaceFactory());
-    fineLevel.Request("Nullspace", &(*nspace));
-    Teuchos::ParameterList nspaceList = *(nspace->GetValidParameterList());
-    nspaceList.set("nullspace: calculate rotations", true);
-    nspace->SetParameterList(nspaceList);
-    fineLevel.Request(*nspace);
-    nspace->Build(fineLevel);
-    RCP<MultiVector> nullSpace = fineLevel.Get<RCP<MultiVector> >("Nullspace",nspace.get());
-
-    ArrayRCP<real_type> xvals = coordinates->getDataNonConst(0);
-    ArrayRCP<real_type> yvals = coordinates->getDataNonConst(1);
-    ArrayRCP<real_type> zvals = coordinates->getDataNonConst(2);
-    size_t nullDim = 6;
-    RCP<MultiVector> handCompNullSpace = MultiVectorFactory::Build(Op->getRowMap(), nullDim);
-    handCompNullSpace->putScalar(0.0);
-    {
-      ArrayRCP<Scalar> nsValues;
-      nsValues = handCompNullSpace->getDataNonConst(0); for (int j = 0; j < nsValues.size(); j +=3) nsValues[j] = 1.;
-      nsValues = handCompNullSpace->getDataNonConst(1); for (int j = 1; j < nsValues.size(); j +=3) nsValues[j] = 1.;
-      nsValues = handCompNullSpace->getDataNonConst(2); for (int j = 2; j < nsValues.size(); j +=3) nsValues[j] = 1.;
-      nsValues = handCompNullSpace->getDataNonConst(3); for (int j = 0; j < nsValues.size(); j +=3) nsValues[j] = -(yvals[j/3]-.5);
-                                                        for (int j = 1; j < nsValues.size(); j +=3) nsValues[j] =  (xvals[j/3]-.5);
-      nsValues = handCompNullSpace->getDataNonConst(4); for (int j = 1; j < nsValues.size(); j +=3) nsValues[j] = -(zvals[j/3]-.5);
-                                                        for (int j = 2; j < nsValues.size(); j +=3) nsValues[j] =  (yvals[j/3]-.5);
-      nsValues = handCompNullSpace->getDataNonConst(5); for (int j = 0; j < nsValues.size(); j +=3) nsValues[j] = -(zvals[j/3]-.5);
-                                                        for (int j = 2; j < nsValues.size(); j +=3) nsValues[j] =  (xvals[j/3]-.5);
-    }
-
-    RCP<MultiVector> diff = MultiVectorFactory::Build(Op->getRowMap(),6);
-    diff->putScalar(0.0);
-    //diff = fineNS + (-1.0)*(handCompNullSpace) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*handCompNullSpace,0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> norms(6);
-    diff->norm2(norms);
-    for (LocalOrdinal i=0; i<6; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_EQUALITY(norms[i] < 100*TMT::eps(), true);
-    }
-
-  } // NullSpace test
+    ArrayRCP<Scalar> nsValues;
+    nsValues = handCompNullSpace->getDataNonConst(0);
+    for (int j = 0; j < nsValues.size(); j += 3) nsValues[j] = 1.;
+    nsValues = handCompNullSpace->getDataNonConst(1);
+    for (int j = 1; j < nsValues.size(); j += 3) nsValues[j] = 1.;
+    nsValues = handCompNullSpace->getDataNonConst(2);
+    for (int j = 2; j < nsValues.size(); j += 3) nsValues[j] = 1.;
+    nsValues = handCompNullSpace->getDataNonConst(3);
+    for (int j = 0; j < nsValues.size(); j += 3) nsValues[j] = -(yvals[j / 3] - .5);
+    for (int j = 1; j < nsValues.size(); j += 3) nsValues[j] = (xvals[j / 3] - .5);
+    nsValues = handCompNullSpace->getDataNonConst(4);
+    for (int j = 1; j < nsValues.size(); j += 3) nsValues[j] = -(zvals[j / 3] - .5);
+    for (int j = 2; j < nsValues.size(); j += 3) nsValues[j] = (yvals[j / 3] - .5);
+    nsValues = handCompNullSpace->getDataNonConst(5);
+    for (int j = 0; j < nsValues.size(); j += 3) nsValues[j] = -(zvals[j / 3] - .5);
+    for (int j = 2; j < nsValues.size(); j += 3) nsValues[j] = (xvals[j / 3] - .5);
+  }
+
+  RCP<MultiVector> diff = MultiVectorFactory::Build(Op->getRowMap(), 6);
+  diff->putScalar(0.0);
+  // diff = fineNS + (-1.0)*(handCompNullSpace) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *handCompNullSpace, 0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> norms(6);
+  diff->norm2(norms);
+  for (LocalOrdinal i = 0; i < 6; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_EQUALITY(norms[i] < 100 * TMT::eps(), true);
+  }
+
+}  // NullSpace test
 
 #define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NullspaceFactory,3DElasticity,Scalar,LO,GO,Node)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NullspaceFactory, 3DElasticity, Scalar, LO, GO, Node)
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ParameterList/CreateSublists.cpp b/packages/muelu/test/unit_tests/ParameterList/CreateSublists.cpp
index 69188afb5dbc..b3a521eacd08 100644
--- a/packages/muelu/test/unit_tests/ParameterList/CreateSublists.cpp
+++ b/packages/muelu/test/unit_tests/ParameterList/CreateSublists.cpp
@@ -62,46 +62,44 @@ namespace MueLuTests {
 
 #include "MueLu_UseShortNames.hpp"
 
-  TEUCHOS_UNIT_TEST(MueLu_CreateSublists, SetParameterList)
-  {
-    std::string dir("ParameterList/CreateSublists/");
+TEUCHOS_UNIT_TEST(MueLu_CreateSublists, SetParameterList) {
+  std::string dir("ParameterList/CreateSublists/");
 
-    ArrayRCP<std::string> fileList = TestHelpers::GetFileList(dir, std::string(".xml"));
+  ArrayRCP<std::string> fileList = TestHelpers::GetFileList(dir, std::string(".xml"));
 
-    for(int i=0; i< fileList.size(); i++) {
-      out << "Processing file: " << fileList[i] << std::endl;
+  for (int i = 0; i < fileList.size(); i++) {
+    out << "Processing file: " << fileList[i] << std::endl;
 
-      Teuchos::RCP<const Teuchos::ParameterList> inputList = Teuchos::getParametersFromXmlFile(dir + fileList[i]);
-      Teuchos::ParameterList outputList;
+    Teuchos::RCP<const Teuchos::ParameterList> inputList = Teuchos::getParametersFromXmlFile(dir + fileList[i]);
+    Teuchos::ParameterList outputList;
 
-      MueLu::CreateSublists(*inputList, outputList);
+    MueLu::CreateSublists(*inputList, outputList);
 
-      // Test against reference output (replace '.xml' by '.output' to get the filename)
-      Teuchos::RCP<Teuchos::ParameterList> refOutputList = Teuchos::getParametersFromXmlFile(dir + fileList[i].substr(0, fileList[i].find_last_of(".")) + ".output");
-      TEST_EQUALITY(outputList, *refOutputList);
-    }
+    // Test against reference output (replace '.xml' by '.output' to get the filename)
+    Teuchos::RCP<Teuchos::ParameterList> refOutputList = Teuchos::getParametersFromXmlFile(dir + fileList[i].substr(0, fileList[i].find_last_of(".")) + ".output");
+    TEST_EQUALITY(outputList, *refOutputList);
   }
+}
 
 #ifdef HAVE_MUELU_ML
-  TEUCHOS_UNIT_TEST(ML_CreateSublists, SetParameterList)
-  {
-    std::string dir("ParameterList/CreateSublists/");
+TEUCHOS_UNIT_TEST(ML_CreateSublists, SetParameterList) {
+  std::string dir("ParameterList/CreateSublists/");
 
-    ArrayRCP<std::string> fileList = TestHelpers::GetFileList(dir, std::string(".xml"));
+  ArrayRCP<std::string> fileList = TestHelpers::GetFileList(dir, std::string(".xml"));
 
-    for(int i=0; i< fileList.size(); i++) {
-      out << "Processing file: " << fileList[i] << std::endl;
+  for (int i = 0; i < fileList.size(); i++) {
+    out << "Processing file: " << fileList[i] << std::endl;
 
-      Teuchos::RCP<const Teuchos::ParameterList> inputList = Teuchos::getParametersFromXmlFile(dir + fileList[i]);
-      Teuchos::ParameterList outputList;
+    Teuchos::RCP<const Teuchos::ParameterList> inputList = Teuchos::getParametersFromXmlFile(dir + fileList[i]);
+    Teuchos::ParameterList outputList;
 
-      ML_CreateSublists(*inputList, outputList);
+    ML_CreateSublists(*inputList, outputList);
 
-      // Test against reference output (replace '.xml' by '.output' to get the filename)
-      Teuchos::RCP<Teuchos::ParameterList> refOutputList = Teuchos::getParametersFromXmlFile(dir + fileList[i].substr(0, fileList[i].find_last_of(".")) + ".output");
-      TEST_EQUALITY(outputList, *refOutputList);
-    }
+    // Test against reference output (replace '.xml' by '.output' to get the filename)
+    Teuchos::RCP<Teuchos::ParameterList> refOutputList = Teuchos::getParametersFromXmlFile(dir + fileList[i].substr(0, fileList[i].find_last_of(".")) + ".output");
+    TEST_EQUALITY(outputList, *refOutputList);
   }
+}
 #endif
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ParameterList/FactoryFactory.cpp b/packages/muelu/test/unit_tests/ParameterList/FactoryFactory.cpp
index d10833cf4d33..a1f4990fa21b 100644
--- a/packages/muelu/test/unit_tests/ParameterList/FactoryFactory.cpp
+++ b/packages/muelu/test/unit_tests/ParameterList/FactoryFactory.cpp
@@ -65,134 +65,131 @@
 
 namespace MueLuTests {
 
-#define RUN  FactoryFactory().BuildFactory(paramValue, factoryMapIn, factoryManagersIn)
-
-
-  // This is not a real unit test, because output of BuildFactory is not verified. But anyway, it still useful.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FactoryFactory, BuildFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    // Determine Epetra/Tpetra mode
-    Teuchos::CommandLineProcessor clp(false);
-    Xpetra::Parameters xpetraParameters(clp);
-    Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
-
-    typedef std::map<std::string, RCP<const FactoryBase> > FactoryMap;
-    typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
-
-    ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/FactoryFactory/"), std::string(".xml"));
-
-    for(int i=0; i< fileList.size(); i++) {
-      out << "Processing file: " << fileList[i] << std::endl;
-      Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile("ParameterList/FactoryFactory/" + fileList[i]);
-
-      for (Teuchos::ParameterList::ConstIterator param = paramList->begin(); param != paramList->end(); ++param) {
-        Teuchos::OSTab tab(out);
-        const std::string            & paramName  = paramList->name(param);
-        const Teuchos::ParameterEntry& paramValue = paramList->entry(param);
-
-        const FactoryMap factoryMapIn;
-        FactoryManagerMap factoryManagersIn;
-
-        // Test when it is not a sublist
-        try {
-          Teuchos::getValue<std::string>(paramValue);
-          RUN;
-          continue;
-        } catch (Teuchos::bad_any_cast&) { }
+#define RUN FactoryFactory().BuildFactory(paramValue, factoryMapIn, factoryManagersIn)
+
+// This is not a real unit test, because output of BuildFactory is not verified. But anyway, it still useful.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(FactoryFactory, BuildFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  // Determine Epetra/Tpetra mode
+  Teuchos::CommandLineProcessor clp(false);
+  Xpetra::Parameters xpetraParameters(clp);
+  Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
+
+  typedef std::map<std::string, RCP<const FactoryBase> > FactoryMap;
+  typedef std::map<std::string, RCP<FactoryManagerBase> > FactoryManagerMap;
+
+  ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/FactoryFactory/"), std::string(".xml"));
+
+  for (int i = 0; i < fileList.size(); i++) {
+    out << "Processing file: " << fileList[i] << std::endl;
+    Teuchos::RCP<Teuchos::ParameterList> paramList = Teuchos::getParametersFromXmlFile("ParameterList/FactoryFactory/" + fileList[i]);
+
+    for (Teuchos::ParameterList::ConstIterator param = paramList->begin(); param != paramList->end(); ++param) {
+      Teuchos::OSTab tab(out);
+      const std::string& paramName              = paramList->name(param);
+      const Teuchos::ParameterEntry& paramValue = paramList->entry(param);
+
+      const FactoryMap factoryMapIn;
+      FactoryManagerMap factoryManagersIn;
+
+      // Test when it is not a sublist
+      try {
+        Teuchos::getValue<std::string>(paramValue);
+        RUN;
+        continue;
+      } catch (Teuchos::bad_any_cast&) {
+      }
 
-        const Teuchos::ParameterList& sublist = Teuchos::getValue<Teuchos::ParameterList>(paramValue);
+      const Teuchos::ParameterList& sublist = Teuchos::getValue<Teuchos::ParameterList>(paramValue);
 
-        // Test when sublist does not contain type
-        try {
-          sublist.get<std::string>("type");
-        } catch (Teuchos::Exceptions::InvalidParameterName&) {
-          RUN;
-          continue;
-        }
+      // Test when sublist does not contain type
+      try {
+        sublist.get<std::string>("type");
+      } catch (Teuchos::Exceptions::InvalidParameterName&) {
+        RUN;
+        continue;
+      }
 
-        std::string type = sublist.get<std::string>("type");
-        std::transform(type.begin(), type.end(), type.begin(), ::tolower);
+      std::string type = sublist.get<std::string>("type");
+      std::transform(type.begin(), type.end(), type.begin(), ::tolower);
 
-        out << "Building object '" << paramName << std::endl;
-        out << "type = " << type << std::endl;
+      out << "Building object '" << paramName << std::endl;
+      out << "type = " << type << std::endl;
 
-        if (type == "klu") {
-          if (lib == Xpetra::UseEpetra) {
+      if (type == "klu") {
+        if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_AMESOS) and defined(HAVE_AMESOS_KLU)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          } else if (lib == Xpetra::UseTpetra) {
-            // Klu defaults to SuperLu in Amesos2Smoother
-            // Therefore, we need to check against SuperLU
+        } else if (lib == Xpetra::UseTpetra) {
+          // Klu defaults to SuperLu in Amesos2Smoother
+          // Therefore, we need to check against SuperLU
 #if defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_SUPERLU)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          }
-        } else if (type == "superlu") {
-          if (lib == Xpetra::UseEpetra) {
+        }
+      } else if (type == "superlu") {
+        if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_AMESOS) and defined(HAVE_AMESOS_SUPERLU)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          } else if (lib == Xpetra::UseTpetra) {
+        } else if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_SUPERLU)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          }
-        } else if (type == "superlu_dist" || type == "superludist") {
-          if (lib == Xpetra::UseEpetra) {
-            out << "Epetra" << std::endl;
+        }
+      } else if (type == "superlu_dist" || type == "superludist") {
+        if (lib == Xpetra::UseEpetra) {
+          out << "Epetra" << std::endl;
 #if defined(HAVE_MUELU_AMESOS) and defined(HAVE_AMESOS_SUPERLUDIST)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          } else if (lib == Xpetra::UseTpetra) {
-            out << "Tpetra" << std::endl;
+        } else if (lib == Xpetra::UseTpetra) {
+          out << "Tpetra" << std::endl;
 #if defined(HAVE_MUELU_AMESOS2) and defined(HAVE_AMESOS2_SUPERLUDIST)
-            out << "Can run superlu_dist" << std::endl;
-            RUN;
+          out << "Can run superlu_dist" << std::endl;
+          RUN;
 #else
-            out << "Cannot run superlu_dist" << std::endl;
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          out << "Cannot run superlu_dist" << std::endl;
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          }
-        } else if (type == "relaxation") {
-          if (lib == Xpetra::UseEpetra) {
+        }
+      } else if (type == "relaxation") {
+        if (lib == Xpetra::UseEpetra) {
 #if defined(HAVE_MUELU_IFPACK)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          } else if (lib == Xpetra::UseTpetra) {
+        } else if (lib == Xpetra::UseTpetra) {
 #if defined(HAVE_MUELU_IFPACK2)
-            RUN;
+          RUN;
 #else
-            TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
+          TEST_THROW(RUN, MueLu::Exceptions::RuntimeError);
 #endif
-          }
-        } else {
-          RUN;
         }
+      } else {
+        RUN;
       }
     }
   }
+}
 
 #define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(FactoryFactory, BuildFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ParameterList/MLParameterListInterpreter.cpp b/packages/muelu/test/unit_tests/ParameterList/MLParameterListInterpreter.cpp
index 53165a5fd2d9..968a84b8401b 100644
--- a/packages/muelu/test/unit_tests/ParameterList/MLParameterListInterpreter.cpp
+++ b/packages/muelu/test/unit_tests/ParameterList/MLParameterListInterpreter.cpp
@@ -56,56 +56,57 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MLParameterListInterpreter, SetParameterList, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (!TYPE_EQUAL(SC, double)) { out << "Skipping for SC != double" << std::endl; return; }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    //TODO: this test can be done at compilation time
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MLParameterListInterpreter, SetParameterList, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (!TYPE_EQUAL(SC, double)) {
+    out << "Skipping for SC != double" << std::endl;
+    return;
+  }
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // TODO: this test can be done at compilation time
 #if !defined(HAVE_MUELU_EPETRA) or !defined(HAVE_MUELU_IFPACK) or !defined(HAVE_MUELU_AMESOS)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Epetra, Ifpack, Amesos");
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Epetra, Ifpack, Amesos");
 #endif
 
 #if !defined(HAVE_MUELU_IFPACK2) or !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Tpetra, Ifpack2, Amesos2");
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Tpetra, Ifpack2, Amesos2");
 #endif
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
-    Teuchos::ParameterList galeriParameters;
-    galeriParameters.set<GO>("nx", 99);
-    RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", A->getRowMap(), galeriParameters);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
+  Teuchos::ParameterList galeriParameters;
+  galeriParameters.set<GO>("nx", 99);
+  RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", A->getRowMap(), galeriParameters);
 
-    ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/MLParameterListInterpreter/"), std::string(".xml"));
+  ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/MLParameterListInterpreter/"), std::string(".xml"));
 
-    for(int i=0; i< fileList.size(); i++) {
-      out << "Processing file: " << fileList[i] << std::endl;
-      Teuchos::ParameterList myList;
-      myList.set("xml parameter file","ParameterList/MLParameterListInterpreter/" + fileList[i]);
+  for (int i = 0; i < fileList.size(); i++) {
+    out << "Processing file: " << fileList[i] << std::endl;
+    Teuchos::ParameterList myList;
+    myList.set("xml parameter file", "ParameterList/MLParameterListInterpreter/" + fileList[i]);
 
-      Teuchos::ArrayRCP<typename MultiVector::scalar_type> xcoord = coordinates->getDataNonConst(0);
-      myList.set("x-coordinates", xcoord.get());
+    Teuchos::ArrayRCP<typename MultiVector::scalar_type> xcoord = coordinates->getDataNonConst(0);
+    myList.set("x-coordinates", xcoord.get());
 
-      MLParameterListInterpreter mueluFactory(myList,A->getRowMap()->getComm());
+    MLParameterListInterpreter mueluFactory(myList, A->getRowMap()->getComm());
 
-      RCP<Hierarchy> H = mueluFactory.CreateHierarchy();
-      H->GetLevel(0)->template Set<RCP<Matrix> >("A", A);
+    RCP<Hierarchy> H = mueluFactory.CreateHierarchy();
+    H->GetLevel(0)->template Set<RCP<Matrix> >("A", A);
 
-      mueluFactory.SetupHierarchy(*H);
+    mueluFactory.SetupHierarchy(*H);
 
-      //TODO: check no unused parameters
-      //TODO: check results of Iterate()
-    }
+    // TODO: check no unused parameters
+    // TODO: check results of Iterate()
   }
+}
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MLParameterListInterpreter, SetParameterList, SC, LO, GO, Node) \
+#define MUELU_ETI_GROUP(SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MLParameterListInterpreter, SetParameterList, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
 // TODO: some tests of the parameter list parser can be done without building the Hierarchy.
diff --git a/packages/muelu/test/unit_tests/ParameterList/ParameterListInterpreter.cpp b/packages/muelu/test/unit_tests/ParameterList/ParameterListInterpreter.cpp
index 2b03c0e65cf8..5ddb75ae1650 100644
--- a/packages/muelu/test/unit_tests/ParameterList/ParameterListInterpreter.cpp
+++ b/packages/muelu/test/unit_tests/ParameterList/ParameterListInterpreter.cpp
@@ -58,227 +58,218 @@
 #include "Tpetra_BlockCrsMatrix_Helpers.hpp"
 #include "TpetraExt_MatrixMatrix.hpp"
 
-
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, SetParameterList, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, SetParameterList, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2) && defined(HAVE_MUELU_AMESOS) && defined(HAVE_MUELU_AMESOS2)
 
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
+  RCP<Matrix> A                       = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(99);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/ParameterListInterpreter/"), std::string(".xml"));
+
+  for (int i = 0; i < fileList.size(); i++) {
+    // Ignore files with "BlockCrs" in their name
+    auto found = fileList[i].find("BlockCrs");
+    if (found != std::string::npos) continue;
+
+    // Ignore files with "Comparison" in their name
+    found = fileList[i].find("Comparison");
+    if (found != std::string::npos) continue;
+
+    out << "Processing file: " << fileList[i] << std::endl;
+    ParameterListInterpreter mueluFactory("ParameterList/ParameterListInterpreter/" + fileList[i], *comm);
+
+    RCP<Hierarchy> H = mueluFactory.CreateHierarchy();
+    H->GetLevel(0)->Set("A", A);
+
+    mueluFactory.SetupHierarchy(*H);
+
+    // TODO: check no unused parameters
+    // TODO: check results of Iterate()
+  }
+
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2, Amesos, Amesos2)." << std::endl;
+#endif
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList matrixParams;
+    matrixParams.set("matrixType", "Laplace1D");
+    matrixParams.set("nx", (GlobalOrdinal)300);  // needs to be even
+
+    RCP<Matrix> A = TestHelpers::TpetraTestFactory<SC, LO, GO, NO>::BuildBlockMatrix(matrixParams, Xpetra::UseTpetra);
+    out << "Matrix Size (block) = " << A->getGlobalNumRows() << " (point) " << A->getRangeMap()->getGlobalNumElements() << std::endl;
     RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
     ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/ParameterListInterpreter/"), std::string(".xml"));
 
-    for(int i=0; i< fileList.size(); i++) {
-      // Ignore files with "BlockCrs" in their name
+    for (int i = 0; i < fileList.size(); i++) {
+      // Only run files with "BlockCrs" in their name
       auto found = fileList[i].find("BlockCrs");
-      if(found != std::string::npos) continue;
-
-      // Ignore files with "Comparison" in their name
-      found = fileList[i].find("Comparison");
-      if(found != std::string::npos) continue;
+      if (found == std::string::npos) continue;
 
       out << "Processing file: " << fileList[i] << std::endl;
-      ParameterListInterpreter mueluFactory("ParameterList/ParameterListInterpreter/" + fileList[i],*comm);
+      ParameterListInterpreter mueluFactory("ParameterList/ParameterListInterpreter/" + fileList[i], *comm);
 
       RCP<Hierarchy> H = mueluFactory.CreateHierarchy();
       H->GetLevel(0)->Set("A", A);
 
       mueluFactory.SetupHierarchy(*H);
 
-      //TODO: check no unused parameters
-      //TODO: check results of Iterate()
-    }
-
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2, Amesos, Amesos2)." << std::endl;
-#   endif
-  }
-
-
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      Teuchos::ParameterList matrixParams;
-      matrixParams.set("matrixType","Laplace1D");
-      matrixParams.set("nx",(GlobalOrdinal)300);// needs to be even
-
-      RCP<Matrix> A = TestHelpers::TpetraTestFactory<SC, LO, GO, NO>::BuildBlockMatrix(matrixParams,Xpetra::UseTpetra);
-      out<<"Matrix Size (block) = "<<A->getGlobalNumRows()<<" (point) "<<A->getRangeMap()->getGlobalNumElements()<<std::endl;
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      
-      ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/ParameterListInterpreter/"), std::string(".xml"));
-      
-      for(int i=0; i< fileList.size(); i++) {
-        // Only run files with "BlockCrs" in their name
-        auto found = fileList[i].find("BlockCrs");
-        if(found == std::string::npos) continue;
-
-        out << "Processing file: " << fileList[i] << std::endl;
-        ParameterListInterpreter mueluFactory("ParameterList/ParameterListInterpreter/" + fileList[i],*comm);
-        
-        RCP<Hierarchy> H = mueluFactory.CreateHierarchy();
-        H->GetLevel(0)->Set("A", A);
-        
-        mueluFactory.SetupHierarchy(*H);
-
-        // Test to make sure all of the matrices in the Hierarchy are actually Block Matrices
-        using helpers = Xpetra::Helpers<Scalar,LocalOrdinal,GlobalOrdinal,Node>;
-        for(int j=0; j<H->GetNumLevels(); j++) {
-          RCP<Level> level = H->GetLevel(j);
-
-          RCP<Matrix> Am = level->Get<RCP<Matrix> >("A");
-          TEST_EQUALITY(helpers::isTpetraBlockCrs(Am),true);
-          if(j>0) {
-            RCP<Matrix> P = level->Get<RCP<Matrix> >("P");
-            TEST_EQUALITY(helpers::isTpetraBlockCrs(P),true);
-            RCP<Matrix> R = level->Get<RCP<Matrix> >("R");
-            TEST_EQUALITY(helpers::isTpetraBlockCrs(R),true);
-          }
+      // Test to make sure all of the matrices in the Hierarchy are actually Block Matrices
+      using helpers = Xpetra::Helpers<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+      for (int j = 0; j < H->GetNumLevels(); j++) {
+        RCP<Level> level = H->GetLevel(j);
+
+        RCP<Matrix> Am = level->Get<RCP<Matrix> >("A");
+        TEST_EQUALITY(helpers::isTpetraBlockCrs(Am), true);
+        if (j > 0) {
+          RCP<Matrix> P = level->Get<RCP<Matrix> >("P");
+          TEST_EQUALITY(helpers::isTpetraBlockCrs(P), true);
+          RCP<Matrix> R = level->Get<RCP<Matrix> >("R");
+          TEST_EQUALITY(helpers::isTpetraBlockCrs(R), true);
         }
-        
-        //TODO: check no unused parameters
-        //TODO: check results of Iterate()
       }
+
+      // TODO: check no unused parameters
+      // TODO: check results of Iterate()
     }
-    TEST_EQUALITY(1,1);
   }
+  TEST_EQUALITY(1, 1);
+}
 
+template <class Matrix, class MT>
+MT compare_matrices(RCP<Matrix> &Ap, RCP<Matrix> &Ab) {
+  using SC  = typename Matrix::scalar_type;
+  using LO  = typename Matrix::local_ordinal_type;
+  using GO  = typename Matrix::global_ordinal_type;
+  using NO  = typename Matrix::node_type;
+  using CRS = Tpetra::CrsMatrix<SC, LO, GO, NO>;
+  SC one    = Teuchos::ScalarTraits<SC>::one();
+  SC zero   = Teuchos::ScalarTraits<SC>::zero();
 
-
-template<class Matrix,class MT>
-MT compare_matrices(RCP<Matrix> & Ap, RCP<Matrix> &Ab) {
-  using SC = typename Matrix::scalar_type;
-  using LO = typename Matrix::local_ordinal_type;
-  using GO = typename Matrix::global_ordinal_type;
-  using NO = typename Matrix::node_type;
-  using CRS=Tpetra::CrsMatrix<SC,LO,GO,NO>;
-  SC one   = Teuchos::ScalarTraits<SC>::one();
-  SC zero  = Teuchos::ScalarTraits<SC>::zero();
-
-  RCP<const CRS> Ap_t = MueLu::Utilities<SC,LO,GO,NO>::Op2TpetraCrs(Ap);
-  auto Ab_t = MueLu::Utilities<SC,LO,GO,NO>::Op2TpetraBlockCrs(Ab);
-  RCP<CRS> Ab_as_point = Tpetra::convertToCrsMatrix<SC,LO,GO,NO>(*Ab_t);
+  RCP<const CRS> Ap_t  = MueLu::Utilities<SC, LO, GO, NO>::Op2TpetraCrs(Ap);
+  auto Ab_t            = MueLu::Utilities<SC, LO, GO, NO>::Op2TpetraBlockCrs(Ab);
+  RCP<CRS> Ab_as_point = Tpetra::convertToCrsMatrix<SC, LO, GO, NO>(*Ab_t);
 
   RCP<CRS> diff = rcp(new CRS(Ap_t->getCrsGraph()));
   diff->setAllToScalar(zero);
   diff->fillComplete();
-  Tpetra::MatrixMatrix::Add<SC,LO,GO,NO>(*Ap_t,false,one,*Ab_as_point,false,-one,diff);
+  Tpetra::MatrixMatrix::Add<SC, LO, GO, NO>(*Ap_t, false, one, *Ab_as_point, false, -one, diff);
   return diff->getFrobeniusNorm();
 }
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, PointCrs_vs_BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList matrixParams;
+    matrixParams.set("matrixType", "Laplace1D");
+    matrixParams.set("nx", (GlobalOrdinal)300);  // needs to be even
+
+    RCP<Matrix> PointA = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, Xpetra::UseTpetra);
+    RCP<Matrix> BlockA;
+    {
+      using XCRS = Xpetra::TpetraBlockCrsMatrix<SC, LO, GO, NO>;
+
+      auto tA      = MueLu::Utilities<SC, LO, GO, NO>::Op2TpetraCrs(PointA);
+      auto bA      = Tpetra::convertToBlockCrsMatrix<SC, LO, GO, NO>(*tA, 1);
+      RCP<XCRS> AA = rcp(new XCRS(bA));
+      BlockA       = rcp(new CrsMatrixWrap(rcp_implicit_cast<CrsMatrix>(AA)));
+    }
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ParameterListInterpreter, PointCrs_vs_BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      Teuchos::ParameterList matrixParams;
-      matrixParams.set("matrixType","Laplace1D");
-      matrixParams.set("nx",(GlobalOrdinal)300);// needs to be even
-      
-      RCP<Matrix> PointA = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,Xpetra::UseTpetra);
-      RCP<Matrix> BlockA;
-      {
-        using XCRS = Xpetra::TpetraBlockCrsMatrix<SC,LO,GO,NO>;
-
-        auto tA = MueLu::Utilities<SC,LO,GO,NO>::Op2TpetraCrs(PointA);
-        auto bA = Tpetra::convertToBlockCrsMatrix<SC,LO,GO,NO>(*tA,1);
-        RCP<XCRS> AA   = rcp(new XCRS(bA));
-        BlockA = rcp(new CrsMatrixWrap(rcp_implicit_cast<CrsMatrix>(AA)));
-      }
+    out << "Point: Matrix Size (block) = " << PointA->getGlobalNumRows() << " (point) " << PointA->getRangeMap()->getGlobalNumElements() << std::endl;
+    out << "Block: Matrix Size (block) = " << BlockA->getGlobalNumRows() << " (point) " << BlockA->getRangeMap()->getGlobalNumElements() << std::endl;
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-      out<<"Point: Matrix Size (block) = "<<PointA->getGlobalNumRows()<<" (point) "<<PointA->getRangeMap()->getGlobalNumElements()<<std::endl;
-      out<<"Block: Matrix Size (block) = "<<BlockA->getGlobalNumRows()<<" (point) "<<BlockA->getRangeMap()->getGlobalNumElements()<<std::endl;
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      
-      ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/ParameterListInterpreter/"), std::string(".xml"));
-      
-      for(int i=0; i< fileList.size(); i++) {
-        // Only run files with "Comparison" in their name
-        auto found = fileList[i].find("Comparison");
-        if(found == std::string::npos) continue;
-        
-        out << "Processing file: " << fileList[i] << std::endl;
-        
-        // Point Hierarchy
-        ParameterListInterpreter mueluFactory1("ParameterList/ParameterListInterpreter/" + fileList[i],*comm);
-        RCP<Hierarchy> PointH = mueluFactory1.CreateHierarchy();
-        PointH->GetLevel(0)->Set("A", PointA);
-        mueluFactory1.SetupHierarchy(*PointH);
-
-        // Block Hierachy
-        ParameterListInterpreter mueluFactory2("ParameterList/ParameterListInterpreter/" + fileList[i],*comm);
-        RCP<Hierarchy> BlockH = mueluFactory2.CreateHierarchy();
-        BlockH->GetLevel(0)->Set("A", BlockA);
-        mueluFactory2.SetupHierarchy(*BlockH);
-
-        // Check to see that we get the same matrices in both hierarchies
-        TEST_EQUALITY(PointH->GetNumLevels(),BlockH->GetNumLevels());
-
-// TODO BMK: compare_matrices uses MatrixMatrix::Add. This was broken before
-// (see #12565), producing a matrix of all zeros if C is fill-complete on input, like
-// compare_matrices does in this test.
-//
-// After fixing Tpetra::MatrixMatrix::Add, it shows that these pairs of matrices (Ap and Ab, etc.)
-// are actually different so this test is not passing. When this is fixed, uncomment this block.
-
-/*
-        for(int j=0; j<PointH->GetNumLevels(); j++) {
-          using CRS=Tpetra::CrsMatrix<SC,LO,GO,NO>;
-          using MT  = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-          MT tol = Teuchos::ScalarTraits<MT>::squareroot(Teuchos::ScalarTraits<MT>::eps());
-         
-          RCP<Level> Plevel = PointH->GetLevel(j);
-          RCP<Level> Blevel = BlockH->GetLevel(j);
-
-          // Compare A
-          RCP<Matrix> Ap = Plevel->Get<RCP<Matrix> >("A");
-          RCP<Matrix> Ab = Blevel->Get<RCP<Matrix> >("A");
-          MT norm = compare_matrices<Matrix,MT>(Ap,Ab);
-
-          TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
-
-          // Compare P, R
-          if(j>0) {
-            RCP<Matrix> Pp = Plevel->Get<RCP<Matrix> >("P");
-            RCP<Matrix> Pb = Blevel->Get<RCP<Matrix> >("P");
-            norm = compare_matrices<Matrix,MT>(Pp,Pb);
-            TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
-
-            RCP<Matrix> Rp = Plevel->Get<RCP<Matrix> >("R");
-            RCP<Matrix> Rb = Blevel->Get<RCP<Matrix> >("R");
-            norm = compare_matrices<Matrix,MT>(Rp,Rb);
-            TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
-          }
-        }
-*/
+    ArrayRCP<std::string> fileList = TestHelpers::GetFileList(std::string("ParameterList/ParameterListInterpreter/"), std::string(".xml"));
 
-        //TODO: check no unused parameters
-        //TODO: check results of Iterate()
-      }
+    for (int i = 0; i < fileList.size(); i++) {
+      // Only run files with "Comparison" in their name
+      auto found = fileList[i].find("Comparison");
+      if (found == std::string::npos) continue;
+
+      out << "Processing file: " << fileList[i] << std::endl;
+
+      // Point Hierarchy
+      ParameterListInterpreter mueluFactory1("ParameterList/ParameterListInterpreter/" + fileList[i], *comm);
+      RCP<Hierarchy> PointH = mueluFactory1.CreateHierarchy();
+      PointH->GetLevel(0)->Set("A", PointA);
+      mueluFactory1.SetupHierarchy(*PointH);
+
+      // Block Hierachy
+      ParameterListInterpreter mueluFactory2("ParameterList/ParameterListInterpreter/" + fileList[i], *comm);
+      RCP<Hierarchy> BlockH = mueluFactory2.CreateHierarchy();
+      BlockH->GetLevel(0)->Set("A", BlockA);
+      mueluFactory2.SetupHierarchy(*BlockH);
+
+      // Check to see that we get the same matrices in both hierarchies
+      TEST_EQUALITY(PointH->GetNumLevels(), BlockH->GetNumLevels());
+
+      // TODO BMK: compare_matrices uses MatrixMatrix::Add. This was broken before
+      // (see #12565), producing a matrix of all zeros if C is fill-complete on input, like
+      // compare_matrices does in this test.
+      //
+      // After fixing Tpetra::MatrixMatrix::Add, it shows that these pairs of matrices (Ap and Ab, etc.)
+      // are actually different so this test is not passing. When this is fixed, uncomment this block.
+
+      /*
+              for(int j=0; j<PointH->GetNumLevels(); j++) {
+                using CRS=Tpetra::CrsMatrix<SC,LO,GO,NO>;
+                using MT  = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+                MT tol = Teuchos::ScalarTraits<MT>::squareroot(Teuchos::ScalarTraits<MT>::eps());
+
+                RCP<Level> Plevel = PointH->GetLevel(j);
+                RCP<Level> Blevel = BlockH->GetLevel(j);
+
+                // Compare A
+                RCP<Matrix> Ap = Plevel->Get<RCP<Matrix> >("A");
+                RCP<Matrix> Ab = Blevel->Get<RCP<Matrix> >("A");
+                MT norm = compare_matrices<Matrix,MT>(Ap,Ab);
+
+                TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
+
+                // Compare P, R
+                if(j>0) {
+                  RCP<Matrix> Pp = Plevel->Get<RCP<Matrix> >("P");
+                  RCP<Matrix> Pb = Blevel->Get<RCP<Matrix> >("P");
+                  norm = compare_matrices<Matrix,MT>(Pp,Pb);
+                  TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
+
+                  RCP<Matrix> Rp = Plevel->Get<RCP<Matrix> >("R");
+                  RCP<Matrix> Rb = Blevel->Get<RCP<Matrix> >("R");
+                  norm = compare_matrices<Matrix,MT>(Rp,Rb);
+                  TEUCHOS_TEST_COMPARE(norm,<,tol,out,success);
+                }
+              }
+      */
+
+      // TODO: check no unused parameters
+      // TODO: check results of Iterate()
     }
   }
+}
 
-
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                                            \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ParameterListInterpreter, SetParameterList, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ParameterListInterpreter, BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ParameterListInterpreter, BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node)         \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ParameterListInterpreter, PointCrs_vs_BlockCrs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
 // TODO: some tests of the parameter list parser can be done without building the Hierarchy.
diff --git a/packages/muelu/test/unit_tests/PermutationFactory.cpp b/packages/muelu/test/unit_tests/PermutationFactory.cpp
index 859aee7a79c3..f9c395bcf542 100644
--- a/packages/muelu/test/unit_tests/PermutationFactory.cpp
+++ b/packages/muelu/test/unit_tests/PermutationFactory.cpp
@@ -51,37 +51,35 @@
 
 #include <MueLu_PermutationFactory.hpp>
 
-
 namespace MueLuTests {
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PermutationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PermutationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    auto A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
-    const auto map = A->getRangeMap();
+  auto A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
+  const auto map = A->getRangeMap();
 
-    MueLu::PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> permFact;
-    permFact.SetParameter("PermutationStrategy",Teuchos::ParameterEntry(std::string("Local")));
-    permFact.SetParameter("PermutationRowMapName",Teuchos::ParameterEntry(std::string("")));
-    permFact.SetFactory("PermutationRowMapFactory", Teuchos::null);
+  MueLu::PermutationFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> permFact;
+  permFact.SetParameter("PermutationStrategy", Teuchos::ParameterEntry(std::string("Local")));
+  permFact.SetParameter("PermutationRowMapName", Teuchos::ParameterEntry(std::string("")));
+  permFact.SetFactory("PermutationRowMapFactory", Teuchos::null);
 
-    MueLu::Level level;
-    level.SetLevelID(0);
-    level.Set("A", A);
+  MueLu::Level level;
+  level.SetLevelID(0);
+  level.Set("A", A);
 
-    level.Request("A", &permFact);
-    permFact.Build(level);
+  level.Request("A", &permFact);
+  permFact.Build(level);
 
-    auto coords = MultiVectorFactory::Build(map, 100);
+  auto coords = MultiVectorFactory::Build(map, 100);
 }
 
 #define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PermutationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PermutationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/PermutedTransferFactory.cpp b/packages/muelu/test/unit_tests/PermutedTransferFactory.cpp
index 986922b85521..cbc8f8afe526 100644
--- a/packages/muelu/test/unit_tests/PermutedTransferFactory.cpp
+++ b/packages/muelu/test/unit_tests/PermutedTransferFactory.cpp
@@ -69,77 +69,69 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceTransfer, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceTransfer, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<RebalanceTransferFactory> ptFactory = rcp(new RebalanceTransferFactory());
-    TEST_EQUALITY(ptFactory != Teuchos::null, true);
-  } // Constructor test
+  RCP<RebalanceTransferFactory> ptFactory = rcp(new RebalanceTransferFactory());
+  TEST_EQUALITY(ptFactory != Teuchos::null, true);
+}  // Constructor test
 
 #ifdef NEVER_TESTED_TODO
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceTransfer, Build1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    GO nx = 199;
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",A);
-
-    //build coordinates
-    Teuchos::ParameterList list;
-    list.set("nx",nx);
-    RCP<MultiVector> coordVector = Galeri::Xpetra::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D",A->getRowMap(),list);
-    fineLevel.Set("Coordinates",coordVector);
-
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    RCP<TentativePFactory>    Ptentfact = rcp(new TentativePFactory(UncoupledAggFact));
-    RCP<SaPFactory>           Pfact = rcp( new SaPFactory(Ptentfact));
-    RCP<Factory>             Rfact = rcp( new TransPFactory(Pfact) );
-    RCP<RAPFactory>           Acfact = rcp( new RAPFactory(Pfact,Rfact) );
-    RCP<Factory>             Rtentfact = rcp( new TransPFactory(Ptentfact) );
-
-    RCP<MultiVectorTransferFactory> mvTransFact = rcp(new MultiVectorTransferFactory("Coordinates","R",Rtentfact));
-    Acfact->AddTransferFactory(mvTransFact);
-    RCP<ZoltanInterface>      zoltan = rcp(new ZoltanInterface(Acfact,mvTransFact));
-    RCP<RepartitionFactory> RepartitionFact = rcp(new RepartitionFactory(zoltan,Acfact));
-
-    coarseLevel.Request("A",Acfact.get());  // kick off the DeclareInputs
-    coarseLevel.Request("Permutation",RepartitionFact.get());  // request permutation matrix
-    //coarseLevel.Request("P",Pfact.get());
-    coarseLevel.Request("R",Rtentfact.get());
-    coarseLevel.Request("Coordinates",mvTransFact.get());
-
-    RCP<RebalanceTransferFactory> ptFactory = rcp( new RebalanceTransferFactory(RepartitionFact, Acfact, Pfact, MueLu::INTERPOLATION) );
-    ptFactory->SetParameter("type", ParameterEntry("Interpolation"));
-    coarseLevel.Request("P",ptFactory.get());
-    ptFactory->Build(fineLevel,coarseLevel);
-
-    ptFactory = rcp( new RebalanceTransferFactory(RepartitionFact, Acfact, Rfact, MueLu::RESTRICTION) );
-    coarseLevel.Request("R",ptFactory.get());
-    ptFactory->Build(fineLevel,coarseLevel);
-
-
-
-  } // Constructor test
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceTransfer, Build1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  GO nx         = 199;
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", A);
+
+  // build coordinates
+  Teuchos::ParameterList list;
+  list.set("nx", nx);
+  RCP<MultiVector> coordVector = Galeri::Xpetra::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", A->getRowMap(), list);
+  fineLevel.Set("Coordinates", coordVector);
+
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  RCP<TentativePFactory> Ptentfact                  = rcp(new TentativePFactory(UncoupledAggFact));
+  RCP<SaPFactory> Pfact                             = rcp(new SaPFactory(Ptentfact));
+  RCP<Factory> Rfact                                = rcp(new TransPFactory(Pfact));
+  RCP<RAPFactory> Acfact                            = rcp(new RAPFactory(Pfact, Rfact));
+  RCP<Factory> Rtentfact                            = rcp(new TransPFactory(Ptentfact));
+
+  RCP<MultiVectorTransferFactory> mvTransFact = rcp(new MultiVectorTransferFactory("Coordinates", "R", Rtentfact));
+  Acfact->AddTransferFactory(mvTransFact);
+  RCP<ZoltanInterface> zoltan             = rcp(new ZoltanInterface(Acfact, mvTransFact));
+  RCP<RepartitionFactory> RepartitionFact = rcp(new RepartitionFactory(zoltan, Acfact));
+
+  coarseLevel.Request("A", Acfact.get());                     // kick off the DeclareInputs
+  coarseLevel.Request("Permutation", RepartitionFact.get());  // request permutation matrix
+  // coarseLevel.Request("P",Pfact.get());
+  coarseLevel.Request("R", Rtentfact.get());
+  coarseLevel.Request("Coordinates", mvTransFact.get());
+
+  RCP<RebalanceTransferFactory> ptFactory = rcp(new RebalanceTransferFactory(RepartitionFact, Acfact, Pfact, MueLu::INTERPOLATION));
+  ptFactory->SetParameter("type", ParameterEntry("Interpolation"));
+  coarseLevel.Request("P", ptFactory.get());
+  ptFactory->Build(fineLevel, coarseLevel);
+
+  ptFactory = rcp(new RebalanceTransferFactory(RepartitionFact, Acfact, Rfact, MueLu::RESTRICTION));
+  coarseLevel.Request("R", ptFactory.get());
+  ptFactory->Build(fineLevel, coarseLevel);
+
+}  // Constructor test
 
 #endif
 
 #define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-     TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceTransfer, Constructor, SC, LO, GO, NO)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceTransfer, Constructor, SC, LO, GO, NO)
 #include <MueLu_ETI_4arg.hpp>
 
-
-
-} // namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/PgPFactory.cpp b/packages/muelu/test/unit_tests/PgPFactory.cpp
index fcbdb716dbb0..4c7970ff9251 100644
--- a/packages/muelu/test/unit_tests/PgPFactory.cpp
+++ b/packages/muelu/test/unit_tests/PgPFactory.cpp
@@ -71,221 +71,218 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<PgPFactory> pgpFactory = rcp(new PgPFactory);
-    TEST_EQUALITY(pgpFactory != Teuchos::null, true);
+  RCP<PgPFactory> pgpFactory = rcp(new PgPFactory);
+  TEST_EQUALITY(pgpFactory != Teuchos::null, true);
 
-    out << *pgpFactory << std::endl;
+  out << *pgpFactory << std::endl;
 
-    std::ostringstream oss;
+  std::ostringstream oss;
 
-    Level level;
-    std::streambuf* p_cout_streambuf = std::cout.rdbuf();
-    std::cout.rdbuf(oss.rdbuf());
+  Level level;
+  std::streambuf *p_cout_streambuf = std::cout.rdbuf();
+  std::cout.rdbuf(oss.rdbuf());
 
-    // BuildP is currently a method that do nothing else than printing a "TODO: Remove me" message
-    pgpFactory->BuildP(level, level);
+  // BuildP is currently a method that do nothing else than printing a "TODO: Remove me" message
+  pgpFactory->BuildP(level, level);
 
-    std::cout.rdbuf(p_cout_streambuf); // restore
+  std::cout.rdbuf(p_cout_streambuf);  // restore
 
-    // Let's verify this content
-    TEST_EQUALITY(oss.str() == "TODO: remove me\n", true);
-  }
+  // Let's verify this content
+  TEST_EQUALITY(oss.str() == "TODO: remove me\n", true);
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, nonsymExample, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, nonsymExample, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#if defined(IFPACK) && defined(IFPACK2)
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "Skipping test because not all required packages are enabled (Ifpack, Ifpack2)." << std::endl;
+  return;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test PgPFactory within" << std::endl;
+  out << "level AMG solver using Petrov Galerkin smoothed aggregation with" << std::endl;
+  out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  Teuchos::Array<magnitude_type> results(2);
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels   = 3;
+  LocalOrdinal its         = 10;
+  LocalOrdinal nEle        = 63;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", nEle);
+
+  // create nonsymmetric tridiagonal matrix
+  Scalar epsilon = 1e-3;
+  RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap>(map, nEle, 1.0, 1.0 - epsilon, epsilon);
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0)
+    out << "||NS|| = " << norms[0] << std::endl;
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                 // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+
+  // define transfer operators
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
+  RCP<PgPFactory> Pfact            = rcp(new PgPFactory());
+  RCP<Factory> Rfact               = rcp(new GenericRFactory());
+  RCP<RAPFactory> Acfact           = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (SC)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", Ptentfact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  // test some basic multigrid data
+  RCP<Level> coarseLevel = H->GetLevel(1);
+  coarseLevel->print(out);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", Ptentfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", Ptentfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Pfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Ptentfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", Rfact.get()), 0);
+
+  RCP<Matrix> P1          = coarseLevel->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R1          = coarseLevel->Get<RCP<Matrix> >("R");
+  RCP<Level> coarseLevel2 = H->GetLevel(2);
+  coarseLevel2->print(out);
+  TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel2->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("P", Ptentfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("P", Ptentfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Pfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Ptentfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", Rfact.get()), 0);
+  RCP<Matrix> P2 = coarseLevel2->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R2 = coarseLevel2->Get<RCP<Matrix> >("R");
+
+  // Define RHS
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+  X->putScalar(1.0);
+  X->norm2(norms);
+  if (comm->getRank() == 0)
+    out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+  // Use AMG directly as an iterative method
   {
-#   if defined(IFPACK) && defined(IFPACK2)
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "Skipping test because not all required packages are enabled (Ifpack, Ifpack2)." << std::endl;
-    return;
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test PgPFactory within" << std::endl;
-    out << "level AMG solver using Petrov Galerkin smoothed aggregation with" << std::endl;
-    out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
+    X->putScalar((SC)0.0);
 
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+    H->Iterate(*RHS, *X, its);
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    Teuchos::Array<magnitude_type> results(2);
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 3;
-    LocalOrdinal its=10;
-    LocalOrdinal nEle = 63;
-    const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-    Teuchos::ParameterList matrixParameters;
-    matrixParameters.set("nx",nEle);
-
-    // create nonsymmetric tridiagonal matrix
-    Scalar epsilon = 1e-3;
-    RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap>(map, nEle, 1.0, 1.0-epsilon, epsilon);
-
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (SC) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0)
-      out << "||NS|| = " << norms[0] << std::endl;
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-
-    // define transfer operators
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
-    RCP<PgPFactory>        Pfact = rcp( new PgPFactory());
-    RCP<Factory>      Rfact = rcp( new GenericRFactory() );
-    RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-    H->SetMaxCoarseSize(1);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("A", Acfact);
-    M.SetFactory("Ptent", Ptentfact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    H->Setup(M, 0, maxLevels);
-
-    // test some basic multigrid data
-    RCP<Level> coarseLevel = H->GetLevel(1);
-    coarseLevel->print(out);
-    TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("A",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",Ptentfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",Ptentfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Pfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Ptentfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",Rfact.get()), 0);
-
-    RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-    RCP<Level> coarseLevel2 = H->GetLevel(2);
-    coarseLevel2->print(out);
-    TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("P",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("R",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("A",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("P",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("R",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel2->IsRequested("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("P",Ptentfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("P",Ptentfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Pfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Ptentfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",Rfact.get()), 0);
-    RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
-
-    // Define RHS
-    RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
-
-    X->putScalar(1.0);
     X->norm2(norms);
     if (comm->getRank() == 0)
-      out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-    Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-    // Use AMG directly as an iterative method
-    {
-      X->putScalar( (SC) 0.0);
-
-      H->Iterate(*RHS,*X,its);
-
-      X->norm2(norms);
-      if (comm->getRank() == 0)
-        out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-    }
-#   else
-    out << "Skipping test because some required packages are not enabled (Ifpack, Ifpack2)." << std::endl;
-#   endif
-  } //nonsymExample
-
+      out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+  }
+#else
+  out << "Skipping test because some required packages are not enabled (Ifpack, Ifpack2)." << std::endl;
+#endif
+}  // nonsymExample
 
 #if 0
   TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, NonStandardMaps, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
     RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
@@ -463,29 +460,27 @@ namespace MueLuTests {
   }
 #endif
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, MinimizationModes, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test PgPFactory (minimization modes)" << std::endl;
-
-    RCP<PgPFactory>        Pfact = rcp( new PgPFactory());
-    TEST_EQUALITY(Pfact != Teuchos::null, true);
-    Pfact->SetMinimizationMode(MueLu::ANORM);
-    TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::ANORM);
-    Pfact->SetMinimizationMode(MueLu::L2NORM);
-    TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::L2NORM);
-    Pfact->SetMinimizationMode(MueLu::DINVANORM);
-    TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::DINVANORM);
-
-  }
-
-#if 0 // TODO check me
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, MinimizationModes, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test PgPFactory (minimization modes)" << std::endl;
+
+  RCP<PgPFactory> Pfact = rcp(new PgPFactory());
+  TEST_EQUALITY(Pfact != Teuchos::null, true);
+  Pfact->SetMinimizationMode(MueLu::ANORM);
+  TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::ANORM);
+  Pfact->SetMinimizationMode(MueLu::L2NORM);
+  TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::L2NORM);
+  Pfact->SetMinimizationMode(MueLu::DINVANORM);
+  TEST_EQUALITY(Pfact->GetMinimizationMode(), MueLu::DINVANORM);
+}
+
+#if 0  // TODO check me
   TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, ColumnBasedOmegas, Scalar, LocalOrdinal, GlobalOrdinal, Node)
   {
-#   include <MueLu_UseShortNames.hpp>
+#include <MueLu_UseShortNames.hpp>
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
     out << "version: " << MueLu::Version() << std::endl;
@@ -646,253 +641,447 @@ namespace MueLuTests {
   }
 #endif
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReUseOmegasInternal(Teuchos::FancyOStream &out, bool &success, MueLu::MinimizationNorm norm)
-  {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReUseOmegasInternal(Teuchos::FancyOStream &out, bool &success, MueLu::MinimizationNorm norm) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra,"Ifpack2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test PgPFactory (reuse row based omegas for restriction operator)" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 3;
-    LocalOrdinal its       = 10;
-    LocalOrdinal nEle      = 63;
-    const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-    Teuchos::ParameterList matrixParameters;
-    matrixParameters.set("nx", nEle);
-
-    // create nonsymmetric tridiagonal matrix
-    RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap>(map, nEle, 2.0, -1.0, -1.0);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test PgPFactory (reuse row based omegas for restriction operator)" << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels   = 3;
+  LocalOrdinal its         = 10;
+  LocalOrdinal nEle        = 63;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", nEle);
+
+  // create nonsymmetric tridiagonal matrix
+  RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap>(map, nEle, 2.0, -1.0, -1.0);
+
+  GO nx = nEle;
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar, LocalOrdinal, GlobalOrdinal, Map, RealValuedMultiVector>("1D", map, galeriList);
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0)
+    out << "||NS|| = " << norms[0] << std::endl;
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                     // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);      // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
+
+  // define transfer operators
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
+  RCP<PgPFactory> Pfact            = rcp(new PgPFactory());
+  Pfact->SetMinimizationMode(norm);
+  RCP<Factory> Rfact     = rcp(new GenericRFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  Pfact->ReUseDampingParameters(true);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (SC)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", Ptentfact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  // test some basic multigrid data
+  RCP<Level> coarseLevel = H->GetLevel(1);
+  coarseLevel->print(out);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+
+  RCP<Matrix> P1          = coarseLevel->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R1          = coarseLevel->Get<RCP<Matrix> >("R");
+  RCP<Level> coarseLevel2 = H->GetLevel(2);
+  coarseLevel2->print(out);
+  TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+  RCP<Matrix> P2 = coarseLevel2->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R2 = coarseLevel2->Get<RCP<Matrix> >("R");
+
+  // Define RHS
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+  X->putScalar(1.0);
+  X->norm2(norms);
+  if (comm->getRank() == 0)
+    out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+  // Use AMG directly as an iterative method
+  {
+    X->putScalar((SC)0.0);
 
-    GO nx = nEle;
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,Map,RealValuedMultiVector>("1D", map, galeriList);
+    H->Iterate(*RHS, *X, its);
 
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (SC) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
+    X->norm2(norms);
     if (comm->getRank() == 0)
-      out << "||NS|| = " << norms[0] << std::endl;
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",         Op);              // set fine level matrix
-    Finest->Set("Nullspace", nullSpace);       // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);   // set coordinates for finest level
-
-    // define transfer operators
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
-    RCP<PgPFactory>        Pfact     = rcp(new PgPFactory());
-    Pfact->SetMinimizationMode(norm);
-    RCP<Factory>           Rfact     = rcp(new GenericRFactory());
-    RCP<RAPFactory>        Acfact    = rcp(new RAPFactory());
-    H->SetMaxCoarseSize(1);
-
-    Pfact->ReUseDampingParameters(true);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type",           "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps",         (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
-    RCP<SmootherFactory>   SmooFact  = rcp(new SmootherFactory(smooProto));
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P",             Pfact);
-    M.SetFactory("R",             Rfact);
-    M.SetFactory("A",             Acfact);
-    M.SetFactory("Ptent",         Ptentfact);
-    M.SetFactory("Aggregates",    UncoupledAggFact);
-    M.SetFactory("Smoother",      SmooFact);
-    M.SetFactory("CoarseSolver",  coarseSolveFact);
-
-    H->Setup(M, 0, maxLevels);
-
-    // test some basic multigrid data
-    RCP<Level> coarseLevel = H->GetLevel(1);
-    coarseLevel->print(out);
-    TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+      out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+  }
 
-    RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-    RCP<Level> coarseLevel2 = H->GetLevel(2);
-    coarseLevel2->print(out);
-    TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-    RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
+  // Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
+  RCP<Level> l0 = H->GetLevel(0);
+  TEST_EQUALITY(l0->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l0->IsAvailable("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  RCP<Level> l1 = H->GetLevel(1);
+  TEST_EQUALITY(l1->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l1->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  RCP<Level> l2 = H->GetLevel(2);
+  TEST_EQUALITY(l2->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l2->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, ReUseOmegas, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  ReUseOmegasInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::L2NORM);
+  ReUseOmegasInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::DINVANORM);
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void ReUseOmegasTransPInternal(Teuchos::FancyOStream &out, bool &success, MueLu::MinimizationNorm norm) {
+  // reuse row based omegas in PgPFactory activated but not used, since TransPFactory is set as restriction factory
+  // check if RowBasedOmega is not stored in Level!
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test PgPFactory (reuse row based omegas for restriction operator)" << std::endl;
+
+  using magnitude_type        = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels   = 3;
+  LocalOrdinal its         = 10;
+  LocalOrdinal nEle        = 63;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", nEle);
+
+  // create nonsymmetric tridiagonal matrix
+  RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap>(map, nEle, 2.0, -1.0, -1.0);
+
+  GO nx = nEle;
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar, LocalOrdinal, GlobalOrdinal, Map, RealValuedMultiVector>("1D", map, galeriList);
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((SC)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0)
+    out << "||NS|| = " << norms[0] << std::endl;
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                     // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);      // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
+
+  // define transfer operators
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
+  RCP<PgPFactory> Pfact            = rcp(new PgPFactory());
+  Pfact->SetMinimizationMode(norm);
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  Pfact->ReUseDampingParameters(true);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (SC)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", Ptentfact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  // test some basic multigrid data
+  RCP<Level> coarseLevel = H->GetLevel(1);
+  coarseLevel->print(out);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+
+  RCP<Matrix> P1          = coarseLevel->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R1          = coarseLevel->Get<RCP<Matrix> >("R");
+  RCP<Level> coarseLevel2 = H->GetLevel(2);
+  coarseLevel2->print(out);
+  TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+  RCP<Matrix> P2 = coarseLevel2->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R2 = coarseLevel2->Get<RCP<Matrix> >("R");
+
+  // Define RHS
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+  X->putScalar(1.0);
+  X->norm2(norms);
+  if (comm->getRank() == 0)
+    out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+  // Use AMG directly as an iterative method
+  {
+    X->putScalar((SC)0.0);
 
-    // Define RHS
-    RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+    H->Iterate(*RHS, *X, its);
 
-    X->putScalar(1.0);
     X->norm2(norms);
     if (comm->getRank() == 0)
-      out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+      out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+  }
 
-    Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+  // Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
+  RCP<Level> l0 = H->GetLevel(0);
+  TEST_EQUALITY(l0->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l0->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l0->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l0->IsAvailable("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l0->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l0->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::UserData);
+  RCP<Level> l1 = H->GetLevel(1);
+  TEST_EQUALITY(l1->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l1->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l1->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l1->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l1->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l1->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  RCP<Level> l2 = H->GetLevel(2);
+  TEST_EQUALITY(l2->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsRequested("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(l2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(l2->IsAvailable("RowBasedOmega", Pfact.get()), false);
+  TEST_EQUALITY(l2->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(l2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(l2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, ReUseOmegasTransP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  ReUseOmegasTransPInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::L2NORM);
+  ReUseOmegasTransPInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::DINVANORM);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, EpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Compare results of Epetra and Tpetra" << std::endl;
+  out << "for 3 level AMG solver using Petrov Galerkin smoothed aggregation with" << std::endl;
+  out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
 
-    // Use AMG directly as an iterative method
-    {
-      X->putScalar( (SC) 0.0);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-      H->Iterate(*RHS,*X,its);
+  MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar, GlobalOrdinal, Node);
 
-      X->norm2(norms);
-      if (comm->getRank() == 0)
-        out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-    }
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
 
-    //Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
-    RCP<Level> l0 = H->GetLevel(0);
-    TEST_EQUALITY(l0->IsRequested("A",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("P",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("PreSmoother",  MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("R",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l0->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("P",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l0->IsAvailable("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l0->IsAvailable("R",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l0->IsRequested("P",            Pfact.get()),             false);
-    TEST_EQUALITY(l0->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::UserData);
-    RCP<Level> l1 = H->GetLevel(1);
-    TEST_EQUALITY(l1->IsRequested("A",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsRequested("P",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsRequested("PreSmoother",  MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsRequested("R",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l1->IsRequested("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l1->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l1->IsAvailable("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l1->IsRequested("P",            Pfact.get()),             false);
-    TEST_EQUALITY(l1->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l1->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
-    RCP<Level> l2 = H->GetLevel(2);
-    TEST_EQUALITY(l2->IsRequested("A",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsRequested("P",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsRequested("PreSmoother",  MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsRequested("R",            MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsRequested("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l2->IsAvailable("A",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("P",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PreSmoother",  MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(l2->IsAvailable("R",            MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(l2->IsAvailable("RowBasedOmega",Pfact.get()),             false);
-    TEST_EQUALITY(l2->IsRequested("P",            Pfact.get()),             false);
-    TEST_EQUALITY(l2->GetKeepFlag("A",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("P",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("R",            MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",  MueLu::NoFactory::get()), MueLu::Final);
-  }
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
+  Teuchos::Array<magnitude_type> results(2);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, ReUseOmegas, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-      ReUseOmegasInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::L2NORM);
-      ReUseOmegasInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::DINVANORM);
-  }
+  // run test only on 1 procs
+  // then we can check shape of transfer operators
+  // furthermore slightly different results in parallel for Epetra and Tpetra (due to smoothers?)
+  if (comm->getSize() == 1) {
+    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void ReUseOmegasTransPInternal(Teuchos::FancyOStream &out, bool &success, MueLu::MinimizationNorm norm)
-  {
-      // reuse row based omegas in PgPFactory activated but not used, since TransPFactory is set as restriction factory
-      // check if RowBasedOmega is not stored in Level!
-  #   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-  #   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-      MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra,"Amesos2, Ifpack2");
-  #   endif
-      out << "version: " << MueLu::Version() << std::endl;
-      out << "Test PgPFactory (reuse row based omegas for restriction operator)" << std::endl;
-
-      using magnitude_type        = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-      using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-      using RealValuedMultiVector = typename Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-      RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-      // used Xpetra lib (for maps and smoothers)
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+    // run Epetra and Tpetra test
+    for (int run = 0; run < 2; run++)  // TODO: create a subfunction instead or Tuple of UnderlyingLib
+    {
+      if (run == 0)
+        lib = Xpetra::UseEpetra;
+      else
+        lib = Xpetra::UseTpetra;
 
       // generate problem
-      LocalOrdinal maxLevels = 3;
-      LocalOrdinal its=10;
-      LocalOrdinal nEle = 63;
+      LocalOrdinal maxLevels   = 3;
+      LocalOrdinal its         = 10;
+      GlobalOrdinal nEle       = 63;
       const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
       Teuchos::ParameterList matrixParameters;
-      matrixParameters.set("nx",nEle);
+      matrixParameters.set("nx", nEle);
 
-      // create nonsymmetric tridiagonal matrix
-      RCP<Matrix> Op = Galeri::Xpetra::TriDiag<SC,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap>(map, nEle, 2.0, -1.0, -1.0);
-
-      GO nx = nEle;
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,Map,RealValuedMultiVector>("1D", map, galeriList);
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
+      RCP<Matrix> Op                         = Pr->BuildMatrix();
+      RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
 
       // build nullspace
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-      nullSpace->putScalar( (SC) 1.0);
+      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+      nullSpace->putScalar((SC)1.0);
       Teuchos::Array<magnitude_type> norms(1);
       nullSpace->norm1(norms);
       if (comm->getRank() == 0)
         out << "||NS|| = " << norms[0] << std::endl;
 
       // fill hierarchy
-      RCP<Hierarchy> H = rcp( new Hierarchy() );
+      RCP<Hierarchy> H = rcp(new Hierarchy());
       H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
 
       RCP<Level> Finest = H->GetLevel();
       Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-      Finest->Set("A", Op);                      // set fine level matrix
-      Finest->Set("Nullspace", nullSpace);       // set null space information for finest level
-      Finest->Set("Coordinates", coordinates);   // set coordinates for finest level
+      Finest->Set("A", Op);                 // set fine level matrix
+      Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+      // Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
 
       // define transfer operators
       RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
@@ -901,21 +1090,18 @@ namespace MueLuTests {
       UncoupledAggFact->SetOrdering("natural");
 
       RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
-      RCP<PgPFactory>        Pfact = rcp( new PgPFactory());
-      Pfact->SetMinimizationMode(norm);
-      RCP<Factory>           Rfact = rcp( new TransPFactory() );
-      RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
+      RCP<PgPFactory> Pfact            = rcp(new PgPFactory());
+      RCP<Factory> Rfact               = rcp(new GenericRFactory());
+      RCP<RAPFactory> Acfact           = rcp(new RAPFactory());
       H->SetMaxCoarseSize(1);
 
-      Pfact->ReUseDampingParameters(true);
-
       // setup smoothers
       Teuchos::ParameterList smootherParamList;
       smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-      smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-      smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-      RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-      RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
+      smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+      smootherParamList.set("relaxation: damping factor", (SC)1.0);
+      RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+      RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
       Acfact->setVerbLevel(Teuchos::VERB_HIGH);
 
       RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
@@ -934,321 +1120,119 @@ namespace MueLuTests {
 
       // test some basic multigrid data
       RCP<Level> coarseLevel = H->GetLevel(1);
-      coarseLevel->print(out);
-      TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-
-      RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-      RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
+      TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("R", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("R", Rfact.get()), false);
+      RCP<Matrix> P1 = coarseLevel->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R1 = coarseLevel->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P1->getGlobalNumRows(), 63);
+      TEST_EQUALITY(P1->getGlobalNumCols(), 21);
+      TEST_EQUALITY(R1->getGlobalNumRows(), 21);
+      TEST_EQUALITY(R1->getGlobalNumCols(), 63);
       RCP<Level> coarseLevel2 = H->GetLevel(2);
-      coarseLevel2->print(out);
-      TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-      RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-      RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
+      TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("R", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", Pfact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", Ptentfact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", Rfact.get()), 0);
+      RCP<Matrix> P2 = coarseLevel2->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R2 = coarseLevel2->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P2->getGlobalNumRows(), 21);
+      TEST_EQUALITY(P2->getGlobalNumCols(), 7);
+      TEST_EQUALITY(R2->getGlobalNumRows(), 7);
+      TEST_EQUALITY(R2->getGlobalNumCols(), 21);
+
+      Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P1, true, *P1, false, out);
+      TEST_EQUALITY(PtentTPtent->getGlobalMaxNumRowEntries() - 3 < 1e-12, true);
+      TEST_EQUALITY(P1->getGlobalMaxNumRowEntries() - 2 < 1e-12, true);
+      TEST_EQUALITY(P2->getGlobalMaxNumRowEntries() - 2 < 1e-12, true);
 
       // Define RHS
-      RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+      RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+      RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
       X->putScalar(1.0);
       X->norm2(norms);
       if (comm->getRank() == 0)
         out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+      Op->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
       // Use AMG directly as an iterative method
       {
-        X->putScalar( (SC) 0.0);
+        X->putScalar((SC)0.0);
 
-        H->Iterate(*RHS,*X,its);
+        H->Iterate(*RHS, *X, its);
 
         X->norm2(norms);
         if (comm->getRank() == 0)
           out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+        results[run] = norms[0];
       }
+    }
 
-      //Teuchos::RCP<Teuchos::FancyOStream> fos = Teuchos::getFancyOStream(Teuchos::rcpFromRef(std::cout));
-      RCP<Level> l0 = H->GetLevel(0);
-      TEST_EQUALITY(l0->IsRequested("A",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("P",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("R",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l0->IsAvailable("A",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l0->IsAvailable("P",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l0->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l0->IsAvailable("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l0->IsAvailable("R",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l0->IsRequested("P",Pfact.get()), false);
-      TEST_EQUALITY(l0->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::UserData);
-      RCP<Level> l1 = H->GetLevel(1);
-      TEST_EQUALITY(l1->IsRequested("A",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l1->IsRequested("P",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l1->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l1->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l1->IsRequested("R",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l1->IsRequested("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l1->IsAvailable("A",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l1->IsAvailable("P",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l1->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l1->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l1->IsAvailable("R",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l1->IsAvailable("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l1->IsRequested("P",Pfact.get()), false);
-      TEST_EQUALITY(l1->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l1->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l1->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l1->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l1->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-      RCP<Level> l2 = H->GetLevel(2);
-      TEST_EQUALITY(l2->IsRequested("A",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsRequested("P",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsRequested("R",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsRequested("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l2->IsAvailable("A",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l2->IsAvailable("P",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l2->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l2->IsAvailable("PostSmoother",MueLu::NoFactory::get()), false);
-      TEST_EQUALITY(l2->IsAvailable("R",MueLu::NoFactory::get()), true);
-      TEST_EQUALITY(l2->IsAvailable("RowBasedOmega",Pfact.get()), false);
-      TEST_EQUALITY(l2->IsRequested("P",Pfact.get()), false);
-      TEST_EQUALITY(l2->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l2->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l2->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-      TEST_EQUALITY(l2->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-  }
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, ReUseOmegasTransP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-      ReUseOmegasTransPInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::L2NORM);
-      ReUseOmegasTransPInternal<Scalar, LocalOrdinal, GlobalOrdinal, Node>(out, success, MueLu::DINVANORM);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(PgPFactory, EpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Compare results of Epetra and Tpetra" << std::endl;
-    out << "for 3 level AMG solver using Petrov Galerkin smoothed aggregation with" << std::endl;
-    out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    Teuchos::Array<magnitude_type> results(2);
-
-    // run test only on 1 procs
-    // then we can check shape of transfer operators
-    // furthermore slightly different results in parallel for Epetra and Tpetra (due to smoothers?)
-    if(comm->getSize() == 1)
-    {
-      Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-
-      // run Epetra and Tpetra test
-      for (int run = 0; run < 2; run++) //TODO: create a subfunction instead or Tuple of UnderlyingLib
-      {
-        if (run == 0) lib = Xpetra::UseEpetra;
-        else lib = Xpetra::UseTpetra;
-
-        // generate problem
-        LocalOrdinal maxLevels = 3;
-        LocalOrdinal its=10;
-        GlobalOrdinal nEle = 63;
-        const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-        Teuchos::ParameterList matrixParameters;
-        matrixParameters.set("nx",nEle);
-
-        RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
-        RCP<Matrix> Op = Pr->BuildMatrix();
-        RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
-
-        // build nullspace
-        RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-        nullSpace->putScalar( (SC) 1.0);
-        Teuchos::Array<magnitude_type> norms(1);
-        nullSpace->norm1(norms);
-        if (comm->getRank() == 0)
-          out << "||NS|| = " << norms[0] << std::endl;
-
-        // fill hierarchy
-        RCP<Hierarchy> H = rcp( new Hierarchy() );
-        H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-        RCP<Level> Finest = H->GetLevel();
-        Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-        Finest->Set("A",Op);                      // set fine level matrix
-        Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-        // Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
-
-        // define transfer operators
-        RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-        UncoupledAggFact->SetMinNodesPerAggregate(3);
-        UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-        UncoupledAggFact->SetOrdering("natural");
-
-        RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
-        RCP<PgPFactory>         Pfact = rcp( new PgPFactory());
-        RCP<Factory>           Rfact = rcp( new GenericRFactory() );
-        RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-        H->SetMaxCoarseSize(1);
-
-        // setup smoothers
-        Teuchos::ParameterList smootherParamList;
-        smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-        smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-        smootherParamList.set("relaxation: damping factor", (SC) 1.0);
-        RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-        RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-        Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-        RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-        FactoryManager M;
-        M.SetKokkosRefactor(false);
-        M.SetFactory("P", Pfact);
-        M.SetFactory("R", Rfact);
-        M.SetFactory("A", Acfact);
-        M.SetFactory("Ptent", Ptentfact);
-        M.SetFactory("Aggregates", UncoupledAggFact);
-        M.SetFactory("Smoother", SmooFact);
-        M.SetFactory("CoarseSolver", coarseSolveFact);
-
-        H->Setup(M, 0, maxLevels);
-
-        // test some basic multigrid data
-        RCP<Level> coarseLevel = H->GetLevel(1);
-        TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("R",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("A",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("R",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("R",Rfact.get()), false);
-        RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P1->getGlobalNumRows(), 63);
-        TEST_EQUALITY(P1->getGlobalNumCols(), 21);
-        TEST_EQUALITY(R1->getGlobalNumRows(), 21);
-        TEST_EQUALITY(R1->getGlobalNumCols(), 63);
-        RCP<Level> coarseLevel2 = H->GetLevel(2);
-        TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("R",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("A",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("R",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",Pfact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",Ptentfact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",Rfact.get()), 0);
-        RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P2->getGlobalNumRows(), 21);
-        TEST_EQUALITY(P2->getGlobalNumCols(), 7);
-        TEST_EQUALITY(R2->getGlobalNumRows(), 7);
-        TEST_EQUALITY(R2->getGlobalNumCols(), 21);
-
-        Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*P1,true,*P1,false,out);
-        TEST_EQUALITY(PtentTPtent->getGlobalMaxNumRowEntries()-3<1e-12, true);
-        TEST_EQUALITY(P1->getGlobalMaxNumRowEntries()-2<1e-12, true);
-        TEST_EQUALITY(P2->getGlobalMaxNumRowEntries()-2<1e-12, true);
-
-        // Define RHS
-        RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
-
-        X->putScalar(1.0);
-        X->norm2(norms);
-        if (comm->getRank() == 0)
-          out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-        Op->apply(*X,*RHS,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-        // Use AMG directly as an iterative method
-        {
-          X->putScalar( (SC) 0.0);
-
-          H->Iterate(*RHS,*X,its);
-
-          X->norm2(norms);
-          if (comm->getRank() == 0)
-            out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-          results[run] = norms[0];
-        }
-      }
-
-      std::cout << results[0] << " VS " << results[1] << std::endl;
-      TEST_EQUALITY(results[0] - results[1] < 1e-10, true); // check results of EPETRA vs TPETRA
-    } // comm->getSize == 1
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
-#   endif
-
-  } //EpetraVsTpetra
-
-
-#   define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, Test0, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, nonsymExample, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, MinimizationModes, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, ReUseOmegas, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, ReUseOmegasTransP, Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, EpetraVsTpetra, Scalar, LO, GO, Node)
-
-# include <MueLu_ETI_4arg.hpp>
-
-}//namespace MueLuTests
+    std::cout << results[0] << " VS " << results[1] << std::endl;
+    TEST_EQUALITY(results[0] - results[1] < 1e-10, true);  // check results of EPETRA vs TPETRA
+  }                                                        // comm->getSize == 1
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
+#endif
 
+}  // EpetraVsTpetra
 
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, Test0, Scalar, LO, GO, Node)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, nonsymExample, Scalar, LO, GO, Node)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, MinimizationModes, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, ReUseOmegas, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, ReUseOmegasTransP, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(PgPFactory, EpetraVsTpetra, Scalar, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
 
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RAPFactory.cpp b/packages/muelu/test/unit_tests/RAPFactory.cpp
index 2c17fb405090..7307e71ee4f0 100644
--- a/packages/muelu/test/unit_tests/RAPFactory.cpp
+++ b/packages/muelu/test/unit_tests/RAPFactory.cpp
@@ -66,415 +66,408 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<RAPFactory> rapFactory = rcp(new RAPFactory);
-    TEST_EQUALITY(rapFactory != Teuchos::null, true);
-
-    out << *rapFactory << std::endl;
-  } // Constructor test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel, coarseLevel; TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    GO nx = 27*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    TentativePFactory tentpFactory;
-    SaPFactory sapFactory;
-    sapFactory.SetFactory("P",rcpFromRef(tentpFactory));
-    TransPFactory transPFactory;
-    transPFactory.SetFactory("P", rcpFromRef(sapFactory));  //todo:rcpFromRef
-
-    coarseLevel.Request("P",&sapFactory);
-    coarseLevel.Request("R",&transPFactory);
-
-    coarseLevel.Request(sapFactory);
-    coarseLevel.Request(transPFactory);
-    sapFactory.Build(fineLevel,coarseLevel);
-    transPFactory.Build(fineLevel,coarseLevel);
-
-    RAPFactory rap;
-    rap.SetFactory("P", rcpFromRef(sapFactory));
-    rap.SetFactory("R", rcpFromRef(transPFactory));
-
-    coarseLevel.Request(rap);
-
-    coarseLevel.Request("A",&rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P", &sapFactory);
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R", &transPFactory);
-
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(R->getRangeMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    //Calculate result1 = R*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Op->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    R->apply(*workVec2,*result1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(R->getRangeMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the Galerkin triple "
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<RAPFactory> rapFactory = rcp(new RAPFactory);
+  TEST_EQUALITY(rapFactory != Teuchos::null, true);
+
+  out << *rapFactory << std::endl;
+}  // Constructor test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  GO nx          = 27 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  TentativePFactory tentpFactory;
+  SaPFactory sapFactory;
+  sapFactory.SetFactory("P", rcpFromRef(tentpFactory));
+  TransPFactory transPFactory;
+  transPFactory.SetFactory("P", rcpFromRef(sapFactory));  // todo:rcpFromRef
+
+  coarseLevel.Request("P", &sapFactory);
+  coarseLevel.Request("R", &transPFactory);
+
+  coarseLevel.Request(sapFactory);
+  coarseLevel.Request(transPFactory);
+  sapFactory.Build(fineLevel, coarseLevel);
+  transPFactory.Build(fineLevel, coarseLevel);
+
+  RAPFactory rap;
+  rap.SetFactory("P", rcpFromRef(sapFactory));
+  rap.SetFactory("R", rcpFromRef(transPFactory));
+
+  coarseLevel.Request(rap);
+
+  coarseLevel.Request("A", &rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P", &sapFactory);
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", &transPFactory);
+
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(R->getRangeMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  // Calculate result1 = R*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Op->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  R->apply(*workVec2, *result1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(R->getRangeMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X))." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100*TMT::eps());
-
-  } // Correctness test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, ImplicitTranspose, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    if (comm->getSize() > 1 && TestHelpers::Parameters::getLib() == Xpetra::UseEpetra ) {
-      out << "Skipping ImplicitTranspose test for Epetra and #proc>1" << std::endl;
-      return;
-    }
-
-    // build test-specific default factory manager
-    RCP<FactoryManager> defManager = rcp(new FactoryManager());
-    defManager->SetKokkosRefactor(false);
-    defManager->SetFactory("A", rcp(MueLu::NoFactory::get(),false));         // dummy factory for A
-    defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));        // real null space factory for Ptent
-    defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));         // real graph factory for Ptent
-    defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));   // real aggregation factory for Ptent
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // overwrite default factory manager
-    fineLevel.SetFactoryManager(defManager);
-    coarseLevel.SetFactoryManager(defManager);
-
-    GO nx = 19*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    TentativePFactory tentpFactory;
-    SaPFactory sapFactory;
-    sapFactory.SetFactory("P",rcpFromRef(tentpFactory));
-    TransPFactory transPFactory;
-    transPFactory.SetFactory("P", rcpFromRef(sapFactory));
-    coarseLevel.Request("P", &sapFactory);
-    coarseLevel.Request("R", &transPFactory);
-
-    coarseLevel.Request(sapFactory);
-    coarseLevel.Request(transPFactory);
-    sapFactory.Build(fineLevel, coarseLevel);
-    transPFactory.Build(fineLevel,coarseLevel);
-    RAPFactory rap;
-    Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
-    rapList.set("transpose: use implicit", true);
-    rap.SetParameterList(rapList);
-    rap.SetFactory("P", rcpFromRef(sapFactory));
-    rap.SetFactory("R", rcpFromRef(transPFactory));
-    coarseLevel.Request("A", &rap);
-
-    coarseLevel.Request(rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P", &sapFactory);
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R", &transPFactory);
-
-    //std::string filename = "A.dat";
-    //Utils::Write(filename,Op);
-    //filename = "P.dat";
-    //Utils::Write(filename,P);
-
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-    //out.precision(12);
-    //out.setOutputToRootOnly(-1);
-    //X->describe(out,Teuchos::VERB_EXTREME);
-
-    //Calculate result1 = P^T*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Op->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    P->apply(*workVec2,*result1,Teuchos::TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the Galerkin triple "
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100 * TMT::eps());
+
+}  // Correctness test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, ImplicitTranspose, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  if (comm->getSize() > 1 && TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "Skipping ImplicitTranspose test for Epetra and #proc>1" << std::endl;
+    return;
+  }
+
+  // build test-specific default factory manager
+  RCP<FactoryManager> defManager = rcp(new FactoryManager());
+  defManager->SetKokkosRefactor(false);
+  defManager->SetFactory("A", rcp(MueLu::NoFactory::get(), false));              // dummy factory for A
+  defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));              // real null space factory for Ptent
+  defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));               // real graph factory for Ptent
+  defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));  // real aggregation factory for Ptent
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // overwrite default factory manager
+  fineLevel.SetFactoryManager(defManager);
+  coarseLevel.SetFactoryManager(defManager);
+
+  GO nx          = 19 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  TentativePFactory tentpFactory;
+  SaPFactory sapFactory;
+  sapFactory.SetFactory("P", rcpFromRef(tentpFactory));
+  TransPFactory transPFactory;
+  transPFactory.SetFactory("P", rcpFromRef(sapFactory));
+  coarseLevel.Request("P", &sapFactory);
+  coarseLevel.Request("R", &transPFactory);
+
+  coarseLevel.Request(sapFactory);
+  coarseLevel.Request(transPFactory);
+  sapFactory.Build(fineLevel, coarseLevel);
+  transPFactory.Build(fineLevel, coarseLevel);
+  RAPFactory rap;
+  Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
+  rapList.set("transpose: use implicit", true);
+  rap.SetParameterList(rapList);
+  rap.SetFactory("P", rcpFromRef(sapFactory));
+  rap.SetFactory("R", rcpFromRef(transPFactory));
+  coarseLevel.Request("A", &rap);
+
+  coarseLevel.Request(rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P", &sapFactory);
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", &transPFactory);
+
+  // std::string filename = "A.dat";
+  // Utils::Write(filename,Op);
+  // filename = "P.dat";
+  // Utils::Write(filename,P);
+
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+  // out.precision(12);
+  // out.setOutputToRootOnly(-1);
+  // X->describe(out,Teuchos::VERB_EXTREME);
+
+  // Calculate result1 = P^T*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Op->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  P->apply(*workVec2, *result1, Teuchos::TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100*TMT::eps());
-
-  } // ImplicitTranspose test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, FixZeroDiagonals, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    out << "This test checks that the option \"rap: fix zero diagonals\" works properly." << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    // build test-specific default factory manager
-    RCP<FactoryManager> defManager = rcp(new FactoryManager());
-    defManager->SetKokkosRefactor(false);
-    defManager->SetFactory("A", rcp(MueLu::NoFactory::get(),false));         // dummy factory for A
-    defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));        // real null space factory for Ptent
-    defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));         // real graph factory for Ptent
-    defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));   // real aggregation factory for Ptent
-    defManager->SetFactory("P", rcp(MueLu::NoFactory::get(),false));         // dummy factory for P
-    defManager->SetFactory("R", rcp(MueLu::NoFactory::get(),false));         // dummy factory for R
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // overwrite default factory manager
-    fineLevel.SetFactoryManager(defManager);
-    coarseLevel.SetFactoryManager(defManager);
-
-    GO nx = 19*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-
-    //Create diagonal matrix, same dimensions as A, but with some zeros on the diagonal
-    RCP<const Map> rowMap = Op->getRowMap();
-    Teuchos::Array<GO> indout(1);
-    Teuchos::Array<SC> valout(1);
-    RCP<Matrix> diagMat = MatrixFactory::Build(rowMap, 1);
-    RCP<Matrix> zdiagMat = MatrixFactory::Build(rowMap, 1);
-    for (size_t r = 0; r < rowMap->getLocalNumElements(); r++) {
-      GO grid = rowMap->getGlobalElement(r);
-      indout[0] = grid;
-      (r == 0) ?  valout[0] = TST::zero() : valout[0] = TST::one();
-      zdiagMat->insertGlobalValues(grid, indout(), valout());
-      valout[0] = TST::one();
-      diagMat->insertGlobalValues(grid, indout(), valout());
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100 * TMT::eps());
+
+}  // ImplicitTranspose test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, FixZeroDiagonals, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  out << "This test checks that the option \"rap: fix zero diagonals\" works properly." << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  // build test-specific default factory manager
+  RCP<FactoryManager> defManager = rcp(new FactoryManager());
+  defManager->SetKokkosRefactor(false);
+  defManager->SetFactory("A", rcp(MueLu::NoFactory::get(), false));              // dummy factory for A
+  defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));              // real null space factory for Ptent
+  defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));               // real graph factory for Ptent
+  defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));  // real aggregation factory for Ptent
+  defManager->SetFactory("P", rcp(MueLu::NoFactory::get(), false));              // dummy factory for P
+  defManager->SetFactory("R", rcp(MueLu::NoFactory::get(), false));              // dummy factory for R
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // overwrite default factory manager
+  fineLevel.SetFactoryManager(defManager);
+  coarseLevel.SetFactoryManager(defManager);
+
+  GO nx          = 19 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+
+  // Create diagonal matrix, same dimensions as A, but with some zeros on the diagonal
+  RCP<const Map> rowMap = Op->getRowMap();
+  Teuchos::Array<GO> indout(1);
+  Teuchos::Array<SC> valout(1);
+  RCP<Matrix> diagMat  = MatrixFactory::Build(rowMap, 1);
+  RCP<Matrix> zdiagMat = MatrixFactory::Build(rowMap, 1);
+  for (size_t r = 0; r < rowMap->getLocalNumElements(); r++) {
+    GO grid   = rowMap->getGlobalElement(r);
+    indout[0] = grid;
+    (r == 0) ? valout[0] = TST::zero() : valout[0] = TST::one();
+    zdiagMat->insertGlobalValues(grid, indout(), valout());
+    valout[0] = TST::one();
+    diagMat->insertGlobalValues(grid, indout(), valout());
+  }
+  zdiagMat->fillComplete();
+  diagMat->fillComplete();
+
+  // this effectively zero's just the row of R*A*P
+  coarseLevel.Set("P", diagMat, MueLu::NoFactory::get());
+  coarseLevel.Set("R", zdiagMat, MueLu::NoFactory::get());
+
+  RAPFactory rap;
+  Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
+  rapList.set("rap: fix zero diagonals", true);
+  rap.SetParameterList(rapList);
+  coarseLevel.Request("A", &rap);
+
+  coarseLevel.Request(rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P");
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R");
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+  // filename = "Ac.dat";
+  // Xpetra::IO<SC,LO,GO,NO>::Write(filename,*coarseOp);
+
+  // The diagonal repair should result in a single nonzero entry with scalar value one in local row 0,
+  // There should be no repeated column indices.
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> vals;
+  coarseOp->getLocalRowView(0, indices, vals);
+  for (int j = 0; j < indices.size(); ++j) {
+    for (int i = j + 1; i < indices.size(); ++i) {
+      out << "checking indices[" << j << "] and indices[" << i << "]" << std::endl;
+      TEST_INEQUALITY(indices[i], indices[j]);
     }
-    zdiagMat->fillComplete();
-    diagMat->fillComplete();
-
-    // this effectively zero's just the row of R*A*P
-    coarseLevel.Set("P", diagMat, MueLu::NoFactory::get());
-    coarseLevel.Set("R", zdiagMat, MueLu::NoFactory::get());
-
-    RAPFactory rap;
-    Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
-    rapList.set("rap: fix zero diagonals", true);
-    rap.SetParameterList(rapList);
-    coarseLevel.Request("A", &rap);
-
-    coarseLevel.Request(rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R");
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-    //filename = "Ac.dat";
-    //Xpetra::IO<SC,LO,GO,NO>::Write(filename,*coarseOp);
-
-    //The diagonal repair should result in a single nonzero entry with scalar value one in local row 0,
-    //There should be no repeated column indices.
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar> vals;
-    coarseOp->getLocalRowView(0,indices,vals);
-    for (int j=0; j<indices.size(); ++j) {
-      for (int i=j+1; i<indices.size(); ++i) {
-        out << "checking indices[" << j << "] and indices[" << i << "]" << std::endl;
-        TEST_INEQUALITY(indices[i],indices[j]);
-      }
+  }
+  Scalar sum = Teuchos::ScalarTraits<SC>::zero();
+  for (int j = 0; j < vals.size(); ++j)
+    sum += vals[j];
+  TEST_EQUALITY(sum, Teuchos::ScalarTraits<SC>::one());
+  const RCP<const Map> colmap = coarseOp->getColMap();
+  const RCP<const Map> rowmap = coarseOp->getRowMap();
+  for (int j = 0; j < vals.size(); ++j) {
+    if (vals[j] == Teuchos::ScalarTraits<SC>::one()) {
+      out << "checking that nonzero entry is on diagonal" << std::endl;
+      TEST_EQUALITY(colmap->getGlobalElement(indices[j]), rowmap->getGlobalElement(0));
     }
-    Scalar sum = Teuchos::ScalarTraits<SC>::zero();
-    for (int j=0; j<vals.size(); ++j)
-      sum += vals[j];
-    TEST_EQUALITY(sum, Teuchos::ScalarTraits<SC>::one());
-    const RCP<const Map> colmap = coarseOp->getColMap();
-    const RCP<const Map> rowmap = coarseOp->getRowMap();
-    for (int j=0; j<vals.size(); ++j) {
-      if (vals[j] == Teuchos::ScalarTraits<SC>::one()) {
-        out << "checking that nonzero entry is on diagonal" << std::endl;
-        TEST_EQUALITY(colmap->getGlobalElement(indices[j]),rowmap->getGlobalElement(0));
-      }
-    }
-
-  } // FixZeroDiagonals test
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, RelativeBoost, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    out << "This test checks that the option \"rap: relative diagonal floor\" works properly." << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    // build test-specific default factory manager
-    RCP<FactoryManager> defManager = rcp(new FactoryManager());
-    defManager->SetKokkosRefactor(false);
-    defManager->SetFactory("A", rcp(MueLu::NoFactory::get(),false));         // dummy factory for A
-    defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));        // real null space factory for Ptent
-    defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));         // real graph factory for Ptent
-    defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));   // real aggregation factory for Ptent
-    defManager->SetFactory("P", rcp(MueLu::NoFactory::get(),false));         // dummy factory for P
-    defManager->SetFactory("R", rcp(MueLu::NoFactory::get(),false));         // dummy factory for R
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // overwrite default factory manager
-    fineLevel.SetFactoryManager(defManager);
-    coarseLevel.SetFactoryManager(defManager);
-
-    GO nx = 19*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-
-    //Create diagonal matrix, same dimensions as A, but with some zeros on the diagonal
-    RCP<const Map> rowMap = Op->getRowMap();
-    Teuchos::Array<GO> indout(1);
-    Teuchos::Array<SC> valout(1);
-    RCP<Matrix> diagMat = MatrixFactory::Build(rowMap, 1);
-    RCP<Matrix> zdiagMat = MatrixFactory::Build(rowMap, 1);
-    for (size_t r = 0; r < rowMap->getLocalNumElements(); r++) {
-      GO grid = rowMap->getGlobalElement(r);
-      indout[0] = grid;
-      (r == 0) ?  valout[0] = TST::zero() : valout[0] = TST::one();
-      zdiagMat->insertGlobalValues(grid, indout(), valout());
-      valout[0] = TST::one();
-      diagMat->insertGlobalValues(grid, indout(), valout());
+  }
+
+}  // FixZeroDiagonals test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPFactory, RelativeBoost, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  out << "This test checks that the option \"rap: relative diagonal floor\" works properly." << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  // build test-specific default factory manager
+  RCP<FactoryManager> defManager = rcp(new FactoryManager());
+  defManager->SetKokkosRefactor(false);
+  defManager->SetFactory("A", rcp(MueLu::NoFactory::get(), false));              // dummy factory for A
+  defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));              // real null space factory for Ptent
+  defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));               // real graph factory for Ptent
+  defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));  // real aggregation factory for Ptent
+  defManager->SetFactory("P", rcp(MueLu::NoFactory::get(), false));              // dummy factory for P
+  defManager->SetFactory("R", rcp(MueLu::NoFactory::get(), false));              // dummy factory for R
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // overwrite default factory manager
+  fineLevel.SetFactoryManager(defManager);
+  coarseLevel.SetFactoryManager(defManager);
+
+  GO nx          = 19 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+
+  // Create diagonal matrix, same dimensions as A, but with some zeros on the diagonal
+  RCP<const Map> rowMap = Op->getRowMap();
+  Teuchos::Array<GO> indout(1);
+  Teuchos::Array<SC> valout(1);
+  RCP<Matrix> diagMat  = MatrixFactory::Build(rowMap, 1);
+  RCP<Matrix> zdiagMat = MatrixFactory::Build(rowMap, 1);
+  for (size_t r = 0; r < rowMap->getLocalNumElements(); r++) {
+    GO grid   = rowMap->getGlobalElement(r);
+    indout[0] = grid;
+    (r == 0) ? valout[0] = TST::zero() : valout[0] = TST::one();
+    zdiagMat->insertGlobalValues(grid, indout(), valout());
+    valout[0] = TST::one();
+    diagMat->insertGlobalValues(grid, indout(), valout());
+  }
+  zdiagMat->fillComplete();
+  diagMat->fillComplete();
+
+  // this effectively zero's just the row of R*A*P
+  coarseLevel.Set("P", diagMat, MueLu::NoFactory::get());
+  coarseLevel.Set("R", zdiagMat, MueLu::NoFactory::get());
+
+  RAPFactory rap;
+  Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
+  Teuchos::Array<double> relativeFloor(1);
+  relativeFloor[0] = 0.5;
+  rapList.set("rap: relative diagonal floor", relativeFloor);
+  rap.SetParameterList(rapList);
+  coarseLevel.Request("A", &rap);
+
+  coarseLevel.Request(rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P");
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R");
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+  std::string filename = "Ac.dat";
+  Xpetra::IO<SC, LO, GO, NO>::Write(filename, *coarseOp);
+
+  // The diagonal repair should result in a single nonzero entry with scalar value one in local row 0,
+  // There should be no repeated column indices.
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> vals;
+  coarseOp->getLocalRowView(0, indices, vals);
+  for (int j = 0; j < indices.size(); ++j) {
+    for (int i = j + 1; i < indices.size(); ++i) {
+      out << "checking indices[" << j << "] and indices[" << i << "]" << std::endl;
+      TEST_INEQUALITY(indices[i], indices[j]);
     }
-    zdiagMat->fillComplete();
-    diagMat->fillComplete();
-
-    // this effectively zero's just the row of R*A*P
-    coarseLevel.Set("P", diagMat, MueLu::NoFactory::get());
-    coarseLevel.Set("R", zdiagMat, MueLu::NoFactory::get());
-
-    RAPFactory rap;
-    Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
-    Teuchos::Array<double> relativeFloor(1);  relativeFloor[0] = 0.5;
-    rapList.set("rap: relative diagonal floor", relativeFloor);
-    rap.SetParameterList(rapList);
-    coarseLevel.Request("A", &rap);
-
-    coarseLevel.Request(rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R");
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-    std::string filename = "Ac.dat";
-    Xpetra::IO<SC,LO,GO,NO>::Write(filename,*coarseOp);
-
-    //The diagonal repair should result in a single nonzero entry with scalar value one in local row 0,
-    //There should be no repeated column indices.
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar> vals;
-    coarseOp->getLocalRowView(0,indices,vals);
-    for (int j=0; j<indices.size(); ++j) {
-      for (int i=j+1; i<indices.size(); ++i) {
-        out << "checking indices[" << j << "] and indices[" << i << "]" << std::endl;
-        TEST_INEQUALITY(indices[i],indices[j]);
-      }
+  }
+  Scalar sum = Teuchos::ScalarTraits<SC>::zero();
+  for (int j = 0; j < vals.size(); ++j)
+    sum += vals[j];
+  TEST_EQUALITY(sum, Teuchos::ScalarTraits<SC>::one());
+  const RCP<const Map> colmap = coarseOp->getColMap();
+  const RCP<const Map> rowmap = coarseOp->getRowMap();
+  for (int j = 0; j < vals.size(); ++j) {
+    if (vals[j] == Teuchos::ScalarTraits<SC>::one()) {
+      out << "checking that nonzero entry is on diagonal" << std::endl;
+      TEST_EQUALITY(colmap->getGlobalElement(indices[j]), rowmap->getGlobalElement(0));
     }
-    Scalar sum = Teuchos::ScalarTraits<SC>::zero();
-    for (int j=0; j<vals.size(); ++j)
-      sum += vals[j];
-    TEST_EQUALITY(sum, Teuchos::ScalarTraits<SC>::one());
-    const RCP<const Map> colmap = coarseOp->getColMap();
-    const RCP<const Map> rowmap = coarseOp->getRowMap();
-    for (int j=0; j<vals.size(); ++j) {
-      if (vals[j] == Teuchos::ScalarTraits<SC>::one()) {
-        out << "checking that nonzero entry is on diagonal" << std::endl;
-        TEST_EQUALITY(colmap->getGlobalElement(indices[j]),rowmap->getGlobalElement(0));
-      }
-    }
-
-  } // Relative Boost test
-
+  }
 
+}  // Relative Boost test
 
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory,Constructor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory,Correctness,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory,ImplicitTranspose,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory,FixZeroDiagonals,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory,RelativeBoost,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory, Constructor, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory, Correctness, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory, ImplicitTranspose, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory, FixZeroDiagonals, Scalar, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPFactory, RelativeBoost, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RAPShiftFactory.cpp b/packages/muelu/test/unit_tests/RAPShiftFactory.cpp
index f057ecf7a8f2..ce0f39b5aa05 100644
--- a/packages/muelu/test/unit_tests/RAPShiftFactory.cpp
+++ b/packages/muelu/test/unit_tests/RAPShiftFactory.cpp
@@ -66,792 +66,776 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<RAPShiftFactory> rapFactory = rcp(new RAPShiftFactory);
-    TEST_EQUALITY(rapFactory != Teuchos::null, true);
-
-    out << *rapFactory << std::endl;
-  } // Constructor test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel, coarseLevel; TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    GO nx = 27*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    // Set "K" and "M" to be copies of A
-    fineLevel.Set("K",Op);
-    fineLevel.Set("M",Op);
-
-    TentativePFactory tentpFactory;
-    SaPFactory sapFactory;
-    sapFactory.SetFactory("P",rcpFromRef(tentpFactory));
-    TransPFactory transPFactory;
-    transPFactory.SetFactory("P", rcpFromRef(sapFactory));  //todo:rcpFromRef
-
-    coarseLevel.Request("P",&sapFactory);
-    coarseLevel.Request("R",&transPFactory);
-
-    coarseLevel.Request(sapFactory);
-    coarseLevel.Request(transPFactory);
-    sapFactory.Build(fineLevel,coarseLevel);
-    transPFactory.Build(fineLevel,coarseLevel);
-
-    RAPShiftFactory rap;
-    Teuchos::ParameterList rapList;
-    rapList.set("rap: shift", 1.0);
-    rap.SetParameterList(rapList);
-
-    rap.SetFactory("P", rcpFromRef(sapFactory));
-    rap.SetFactory("R", rcpFromRef(transPFactory));
-
-    coarseLevel.Request(rap);
-
-    coarseLevel.Request("A",&rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P", &sapFactory);
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R", &transPFactory);
-
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(R->getRangeMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    //Calculate result1 = 2*R*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)2.0,(Scalar)0.0);
-    Op->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    R->apply(*workVec2,*result1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(R->getRangeMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the Galerkin triple "
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<RAPShiftFactory> rapFactory = rcp(new RAPShiftFactory);
+  TEST_EQUALITY(rapFactory != Teuchos::null, true);
+
+  out << *rapFactory << std::endl;
+}  // Constructor test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  GO nx          = 27 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  // Set "K" and "M" to be copies of A
+  fineLevel.Set("K", Op);
+  fineLevel.Set("M", Op);
+
+  TentativePFactory tentpFactory;
+  SaPFactory sapFactory;
+  sapFactory.SetFactory("P", rcpFromRef(tentpFactory));
+  TransPFactory transPFactory;
+  transPFactory.SetFactory("P", rcpFromRef(sapFactory));  // todo:rcpFromRef
+
+  coarseLevel.Request("P", &sapFactory);
+  coarseLevel.Request("R", &transPFactory);
+
+  coarseLevel.Request(sapFactory);
+  coarseLevel.Request(transPFactory);
+  sapFactory.Build(fineLevel, coarseLevel);
+  transPFactory.Build(fineLevel, coarseLevel);
+
+  RAPShiftFactory rap;
+  Teuchos::ParameterList rapList;
+  rapList.set("rap: shift", 1.0);
+  rap.SetParameterList(rapList);
+
+  rap.SetFactory("P", rcpFromRef(sapFactory));
+  rap.SetFactory("R", rcpFromRef(transPFactory));
+
+  coarseLevel.Request(rap);
+
+  coarseLevel.Request("A", &rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P", &sapFactory);
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", &transPFactory);
+
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(R->getRangeMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  // Calculate result1 = 2*R*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)2.0, (Scalar)0.0);
+  Op->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  R->apply(*workVec2, *result1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(R->getRangeMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X))." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100*TMT::eps());
-
-  } // Correctness test
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, ImplicitTranspose, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    if (comm->getSize() > 1 && TestHelpers::Parameters::getLib() == Xpetra::UseEpetra ) {
-      out << "Skipping ImplicitTranspose test for Epetra and #proc>1" << std::endl;
-      return;
-    }
-
-    // build test-specific default factory manager
-    RCP<FactoryManager> defManager = rcp(new FactoryManager());
-    defManager->SetKokkosRefactor(false);
-    defManager->SetFactory("A", rcp(MueLu::NoFactory::get(),false));         // dummy factory for A
-    defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));        // real null space factory for Ptent
-    defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));         // real graph factory for Ptent
-    defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));   // real aggregation factory for Ptent
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // overwrite default factory manager
-    fineLevel.SetFactoryManager(defManager);
-    coarseLevel.SetFactoryManager(defManager);
-
-    GO nx = 19*comm->getSize();
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-    fineLevel.Set("A",Op);
-    fineLevel.Set("K",Op);
-    fineLevel.Set("M",Op);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    TentativePFactory tentpFactory;
-    SaPFactory sapFactory;
-    sapFactory.SetFactory("P",rcpFromRef(tentpFactory));
-    TransPFactory transPFactory;
-    transPFactory.SetFactory("P", rcpFromRef(sapFactory));
-    coarseLevel.Request("P", &sapFactory);
-    coarseLevel.Request("R", &transPFactory);
-
-    coarseLevel.Request(sapFactory);
-    coarseLevel.Request(transPFactory);
-    sapFactory.Build(fineLevel, coarseLevel);
-    transPFactory.Build(fineLevel,coarseLevel);
-    RAPShiftFactory rap;
-
-    Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
-    rapList.set("transpose: use implicit", true);
-    rapList.set("rap: shift",1.0);
-    rap.SetParameterList(rapList);
-    rap.SetFactory("P", rcpFromRef(sapFactory));
-    rap.SetFactory("R", rcpFromRef(transPFactory));
-    coarseLevel.Request("A", &rap);
-
-    coarseLevel.Request(rap);
-    rap.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> A = fineLevel.Get< RCP<Matrix> >("A");
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P", &sapFactory);
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R", &transPFactory);
-
-
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    //Calculate result1 = 2*P^T*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)2.0,(Scalar)0.0);
-    Op->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    P->apply(*workVec2,*result1,Teuchos::TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = coarseLevel.Get< RCP<Matrix> >("A", &rap);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
-      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100*TMT::eps());
-
-  } // ImplicitTranspose test
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100 * TMT::eps());
+
+}  // Correctness test
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, ImplicitTranspose, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  if (comm->getSize() > 1 && TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "Skipping ImplicitTranspose test for Epetra and #proc>1" << std::endl;
+    return;
+  }
 
+  // build test-specific default factory manager
+  RCP<FactoryManager> defManager = rcp(new FactoryManager());
+  defManager->SetKokkosRefactor(false);
+  defManager->SetFactory("A", rcp(MueLu::NoFactory::get(), false));              // dummy factory for A
+  defManager->SetFactory("Nullspace", rcp(new NullspaceFactory()));              // real null space factory for Ptent
+  defManager->SetFactory("Graph", rcp(new CoalesceDropFactory()));               // real graph factory for Ptent
+  defManager->SetFactory("Aggregates", rcp(new UncoupledAggregationFactory()));  // real aggregation factory for Ptent
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // overwrite default factory manager
+  fineLevel.SetFactoryManager(defManager);
+  coarseLevel.SetFactoryManager(defManager);
+
+  GO nx          = 19 * comm->getSize();
+  RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+  fineLevel.Set("A", Op);
+  fineLevel.Set("K", Op);
+  fineLevel.Set("M", Op);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", Op->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  TentativePFactory tentpFactory;
+  SaPFactory sapFactory;
+  sapFactory.SetFactory("P", rcpFromRef(tentpFactory));
+  TransPFactory transPFactory;
+  transPFactory.SetFactory("P", rcpFromRef(sapFactory));
+  coarseLevel.Request("P", &sapFactory);
+  coarseLevel.Request("R", &transPFactory);
+
+  coarseLevel.Request(sapFactory);
+  coarseLevel.Request(transPFactory);
+  sapFactory.Build(fineLevel, coarseLevel);
+  transPFactory.Build(fineLevel, coarseLevel);
+  RAPShiftFactory rap;
+
+  Teuchos::ParameterList rapList = *(rap.GetValidParameterList());
+  rapList.set("transpose: use implicit", true);
+  rapList.set("rap: shift", 1.0);
+  rap.SetParameterList(rapList);
+  rap.SetFactory("P", rcpFromRef(sapFactory));
+  rap.SetFactory("R", rcpFromRef(transPFactory));
+  coarseLevel.Request("A", &rap);
+
+  coarseLevel.Request(rap);
+  rap.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> A = fineLevel.Get<RCP<Matrix> >("A");
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P", &sapFactory);
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", &transPFactory);
+
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(Op->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  // Calculate result1 = 2*P^T*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)2.0, (Scalar)0.0);
+  Op->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  P->apply(*workVec2, *result1, Teuchos::TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = coarseLevel.Get<RCP<Matrix> >("A", &rap);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
+      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 100 * TMT::eps());
 
+}  // ImplicitTranspose test
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Factory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::coordinateType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // The ugly way of managing input decks
-    std::string myInputString =
-"<ParameterList name=\"MueLu\">"
-"  <ParameterList name=\"Factories\">"
-"   <ParameterList name=\"myFilteredAFact\">"
-"     <Parameter name=\"factory\"                             type=\"string\" value=\"FilteredAFactory\"/>"
-"   </ParameterList>"
-"   <ParameterList name=\"myProlongatorFact\">"
-"     <Parameter name=\"factory\"                             type=\"string\" value=\"SaPFactory\"/>"
-"     <Parameter name=\"A\"                                   type=\"string\" value=\"myFilteredAFact\"/>"
-"   </ParameterList>"
-"   <ParameterList name=\"myRAPFact\">"
-"     <Parameter name=\"factory\"                             type=\"string\" value=\"RAPShiftFactory\"/>"
-"     <Parameter name=\"P\"                                   type=\"string\" value=\"myProlongatorFact\"/>"
-"     <Parameter name=\"rap: shift\"                          type=\"double\" value=\"1.0\"/>"
-"   </ParameterList>"
-"   <ParameterList name=\"mySmoo\">"
-"     <Parameter name=\"factory\"                             type=\"string\" value=\"TrilinosSmoother\"/>"
-"     <Parameter name=\"type\"                                type=\"string\" value=\"RELAXATION\"/>"
-"   </ParameterList>"
-" </ParameterList>"
-" <ParameterList name=\"Hierarchy\">"
-"   <Parameter name=\"max levels\"                            type=\"int\"      value=\"10\"/>"
-"   <Parameter name=\"coarse: max size\"                      type=\"int\"      value=\"1000\"/>"
-"   <Parameter name=\"verbosity\"                             type=\"string\"   value=\"None\"/>"
-"   <Parameter name=\"use kokkos refactor\"                   type=\"bool\"     value=\"false\"/>"
-"   <ParameterList name=\"All\">"
-"     <Parameter name=\"startLevel\"                          type=\"int\"      value=\"0\"/>"
-"     <Parameter name=\"Smoother\"                            type=\"string\"   value=\"mySmoo\"/>"
-"     <Parameter name=\"A\"                                   type=\"string\"   value=\"myRAPFact\"/>"
-"     <Parameter name=\"P\"                                   type=\"string\"   value=\"myProlongatorFact\"/>"
-"     <Parameter name=\"CoarseSolver\"                        type=\"string\"   value=\"DirectSolver\"/>"
-"   </ParameterList>"
-" </ParameterList>"
-"</ParameterList>"
-     ;
-
-    // Get the actual parameter list
-    RCP<Teuchos::ParameterList> Params = Teuchos::getParametersFromXmlString(myInputString);
-    Params->set("verbosity","none");
-
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("K",A);
-    pLevel0.set("M",A);
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    //Calculate result1 = 2*P^T*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)2.0,(Scalar)0.0);
-    A->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    P->apply(*workVec2,*result1,Teuchos::TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Factory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::coordinateType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  GO nx                               = 1999 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // The ugly way of managing input decks
+  std::string myInputString =
+      "<ParameterList name=\"MueLu\">"
+      "  <ParameterList name=\"Factories\">"
+      "   <ParameterList name=\"myFilteredAFact\">"
+      "     <Parameter name=\"factory\"                             type=\"string\" value=\"FilteredAFactory\"/>"
+      "   </ParameterList>"
+      "   <ParameterList name=\"myProlongatorFact\">"
+      "     <Parameter name=\"factory\"                             type=\"string\" value=\"SaPFactory\"/>"
+      "     <Parameter name=\"A\"                                   type=\"string\" value=\"myFilteredAFact\"/>"
+      "   </ParameterList>"
+      "   <ParameterList name=\"myRAPFact\">"
+      "     <Parameter name=\"factory\"                             type=\"string\" value=\"RAPShiftFactory\"/>"
+      "     <Parameter name=\"P\"                                   type=\"string\" value=\"myProlongatorFact\"/>"
+      "     <Parameter name=\"rap: shift\"                          type=\"double\" value=\"1.0\"/>"
+      "   </ParameterList>"
+      "   <ParameterList name=\"mySmoo\">"
+      "     <Parameter name=\"factory\"                             type=\"string\" value=\"TrilinosSmoother\"/>"
+      "     <Parameter name=\"type\"                                type=\"string\" value=\"RELAXATION\"/>"
+      "   </ParameterList>"
+      " </ParameterList>"
+      " <ParameterList name=\"Hierarchy\">"
+      "   <Parameter name=\"max levels\"                            type=\"int\"      value=\"10\"/>"
+      "   <Parameter name=\"coarse: max size\"                      type=\"int\"      value=\"1000\"/>"
+      "   <Parameter name=\"verbosity\"                             type=\"string\"   value=\"None\"/>"
+      "   <Parameter name=\"use kokkos refactor\"                   type=\"bool\"     value=\"false\"/>"
+      "   <ParameterList name=\"All\">"
+      "     <Parameter name=\"startLevel\"                          type=\"int\"      value=\"0\"/>"
+      "     <Parameter name=\"Smoother\"                            type=\"string\"   value=\"mySmoo\"/>"
+      "     <Parameter name=\"A\"                                   type=\"string\"   value=\"myRAPFact\"/>"
+      "     <Parameter name=\"P\"                                   type=\"string\"   value=\"myProlongatorFact\"/>"
+      "     <Parameter name=\"CoarseSolver\"                        type=\"string\"   value=\"DirectSolver\"/>"
+      "   </ParameterList>"
+      " </ParameterList>"
+      "</ParameterList>";
+
+  // Get the actual parameter list
+  RCP<Teuchos::ParameterList> Params = Teuchos::getParametersFromXmlString(myInputString);
+  Params->set("verbosity", "none");
+
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("K", A);
+  pLevel0.set("M", A);
+
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  // Calculate result1 = 2*P^T*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)2.0, (Scalar)0.0);
+  A->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  P->apply(*workVec2, *result1, Teuchos::TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e4*Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
-
-  }
-
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e4 * Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Easy, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // The pretty way of managing input decks
-    RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
-    Params->set("rap: algorithm","shift");
-    Params->set("rap: shift",1.0);
-    Params->set("coarse: max size",1000);
-    Params->set("verbosity","none");
-    Params->set("use kokkos refactor",false);
-
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("K",A);
-    pLevel0.set("M",A);
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    //Calculate result1 = 2*P^T*(A*(P*X))
-    P->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)2.0,(Scalar)0.0);
-    A->apply(*workVec1,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    P->apply(*workVec2,*result1,Teuchos::TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    coarseOp->apply(*X,*result2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Easy, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  GO nx                               = 1999 * comm->getSize();
+  RCP<Matrix> A                       = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // The pretty way of managing input decks
+  RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
+  Params->set("rap: algorithm", "shift");
+  Params->set("rap: shift", 1.0);
+  Params->set("coarse: max size", 1000);
+  Params->set("verbosity", "none");
+  Params->set("use kokkos refactor", false);
+
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("K", A);
+  pLevel0.set("M", A);
+
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  // Calculate result1 = 2*P^T*(A*(P*X))
+  P->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)2.0, (Scalar)0.0);
+  A->apply(*workVec1, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  P->apply(*workVec2, *result1, Teuchos::TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  coarseOp->apply(*X, *result2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e4*Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
-
-  }
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e4 * Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Easy_Diagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Extract the diagonal of A
-    RCP<Vector> Mdiag = Xpetra::VectorFactory<SC,LO,GO,NO>::Build(A->getRowMap(),false);
-    A->getLocalDiagCopy(*Mdiag);
-
-    // The pretty way of managing input decks
-    double shift = 1.0;
-    RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
-    Params->set("rap: algorithm","shift");
-    Params->set("rap: shift",shift);
-    Params->set("rap: shift diagonal M",true);
-    Params->set("coarse: max size",1000);
-    Params->set("verbosity","none");
-    Params->set("use kokkos refactor",false);
-    Params->set("coarse: max size",10);
-    Params->set("max levels",2);
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("K",A);
-    pLevel0.set("Mdiag",Mdiag);
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    //Calculate result1 = (P^T*(A+shift*M)*P)*X
-    P->apply(*X,*workVec1);
-    A->apply(*workVec1,*workVec2);
-    workVec2->elementWiseMultiply((Scalar)shift,*Mdiag,*workVec1,one);
-    P->apply(*workVec2,*result1,Teuchos::TRANS);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-    coarseOp->apply(*X,*result2);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Easy_Diagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  GO nx         = 1999 * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Extract the diagonal of A
+  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC, LO, GO, NO>::Build(A->getRowMap(), false);
+  A->getLocalDiagCopy(*Mdiag);
+
+  // The pretty way of managing input decks
+  double shift                       = 1.0;
+  RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
+  Params->set("rap: algorithm", "shift");
+  Params->set("rap: shift", shift);
+  Params->set("rap: shift diagonal M", true);
+  Params->set("coarse: max size", 1000);
+  Params->set("verbosity", "none");
+  Params->set("use kokkos refactor", false);
+  Params->set("coarse: max size", 10);
+  Params->set("max levels", 2);
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("K", A);
+  pLevel0.set("Mdiag", Mdiag);
+
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result2  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  // Calculate result1 = (P^T*(A+shift*M)*P)*X
+  P->apply(*X, *workVec1);
+  A->apply(*workVec1, *workVec2);
+  workVec2->elementWiseMultiply((Scalar)shift, *Mdiag, *workVec1, one);
+  P->apply(*workVec2, *result1, Teuchos::TRANS);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+  coarseOp->apply(*X, *result2);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000*Teuchos::ScalarTraits<Scalar>::eps());
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000 * Teuchos::ScalarTraits<Scalar>::eps());
+}
+
+/*********************************************************************************************************************/
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Low_Storage, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  GO nx         = 1999 * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Extract the diagonal of A
+  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC, LO, GO, NO>::Build(A->getRowMap(), false);
+  A->getLocalDiagCopy(*Mdiag);
+
+  // The pretty way of managing input decks
+  double shift                       = 2.0;
+  RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
+  Params->set("rap: algorithm", "shift");
+  Params->set("rap: shift", shift);
+  Params->set("rap: shift diagonal M", true);
+  Params->set("rap: shift low storage", true);
+  Params->set("coarse: max size", 1000);
+  Params->set("verbosity", "none");
+  Params->set("coarse: max size", 10);
+  Params->set("max levels", 2);
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("Mdiag", Mdiag);
+  if (A->getRowMap()->lib() == Xpetra::UseEpetra) {
+    Params->set("use kokkos refactor", false);
   }
 
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result2  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  // Calculate result1 = (P^T*(A+shift*M)*P)*X
+  P->apply(*X, *workVec1);
+  A->apply(*workVec1, *workVec2);
+  workVec2->elementWiseMultiply((Scalar)shift, *Mdiag, *workVec1, one);
+  P->apply(*workVec2, *result1, Teuchos::TRANS);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+  coarseOp->apply(*X, *result2);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
+      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000 * Teuchos::ScalarTraits<Scalar>::eps());
+}
 
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Low_Storage, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Extract the diagonal of A
-    RCP<Vector> Mdiag = Xpetra::VectorFactory<SC,LO,GO,NO>::Build(A->getRowMap(),false);
-    A->getLocalDiagCopy(*Mdiag);
-
-    // The pretty way of managing input decks
-    double shift = 2.0;
-    RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
-    Params->set("rap: algorithm","shift");
-    Params->set("rap: shift",shift);
-    Params->set("rap: shift diagonal M",true);
-    Params->set("rap: shift low storage",true);
-    Params->set("coarse: max size",1000);
-    Params->set("verbosity","none");
-    Params->set("coarse: max size",10);
-    Params->set("max levels",2);
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("Mdiag",Mdiag);
-    if(A->getRowMap()->lib() == Xpetra::UseEpetra) {
-      Params->set("use kokkos refactor", false);
-    }
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    //Calculate result1 = (P^T*(A+shift*M)*P)*X
-    P->apply(*X,*workVec1);
-    A->apply(*workVec1,*workVec2);
-    workVec2->elementWiseMultiply((Scalar)shift,*Mdiag,*workVec1,one);
-    P->apply(*workVec2,*result1,Teuchos::TRANS);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-    coarseOp->apply(*X,*result2);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
-      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000*Teuchos::ScalarTraits<Scalar>::eps());
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Low_Storage_ShiftList, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  GO nx         = 1999 * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Extract the diagonal of A
+  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC, LO, GO, NO>::Build(A->getRowMap(), false);
+  A->getLocalDiagCopy(*Mdiag);
+
+  // The pretty way of managing input decks
+  double shift = 2.0;
+  Teuchos::Array<double> shifts(2);
+  shifts[0]                          = 1.0;
+  shifts[1]                          = shifts[0] + shift;
+  RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
+  Params->set("rap: algorithm", "shift");
+  Params->set("rap: shift array", shifts);
+  Params->set("rap: shift diagonal M", true);
+  Params->set("rap: shift low storage", true);
+  Params->set("coarse: max size", 1000);
+  Params->set("verbosity", "none");
+  Params->set("coarse: max size", 10);
+  Params->set("max levels", 2);
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("Mdiag", Mdiag);
+  if (A->getRowMap()->lib() == Xpetra::UseEpetra) {
+    Params->set("use kokkos refactor", false);
   }
 
-/*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Low_Storage_ShiftList, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Extract the diagonal of A
-    RCP<Vector> Mdiag = Xpetra::VectorFactory<SC,LO,GO,NO>::Build(A->getRowMap(),false);
-    A->getLocalDiagCopy(*Mdiag);
-
-    // The pretty way of managing input decks
-    double shift = 2.0;
-    Teuchos::Array<double> shifts(2); shifts[0]=1.0; shifts[1]=shifts[0]+shift;
-    RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
-    Params->set("rap: algorithm","shift");
-    Params->set("rap: shift array",shifts);
-    Params->set("rap: shift diagonal M",true);
-    Params->set("rap: shift low storage",true);
-    Params->set("coarse: max size",1000);
-    Params->set("verbosity","none");
-    Params->set("coarse: max size",10);
-    Params->set("max levels",2);
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("Mdiag",Mdiag);
-    if(A->getRowMap()->lib() == Xpetra::UseEpetra) {
-      Params->set("use kokkos refactor", false);
-    }
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    //Calculate result1 = (P^T*(A+shift*M)*P)*X
-    P->apply(*X,*workVec1);
-    A->apply(*workVec1,*workVec2);
-    workVec2->elementWiseMultiply((Scalar)shift,*Mdiag,*workVec1,one);
-    P->apply(*workVec2,*result1,Teuchos::TRANS);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-    coarseOp->apply(*X,*result2);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result2  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  // Calculate result1 = (P^T*(A+shift*M)*P)*X
+  P->apply(*X, *workVec1);
+  A->apply(*workVec1, *workVec2);
+  workVec2->elementWiseMultiply((Scalar)shift, *Mdiag, *workVec1, one);
+  P->apply(*workVec2, *result1, Teuchos::TRANS);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+  coarseOp->apply(*X, *result2);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
       << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000*Teuchos::ScalarTraits<Scalar>::eps());
-  }
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000 * Teuchos::ScalarTraits<Scalar>::eps());
+}
 /*********************************************************************************************************************/
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory,CreatePreconditioner_Low_Storage_CFLSchedule, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
-#   endif
-    out << "version: " << MueLu::Version() << std::endl;
-    typedef Scalar SC;
-    typedef GlobalOrdinal GO;
-    typedef LocalOrdinal LO;
-    typedef Node  NO;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    using Teuchos::RCP;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    GO nx = 1999*comm->getSize();
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-
-    // Extract the diagonal of A
-    RCP<Vector> Mdiag = Xpetra::VectorFactory<SC,LO,GO,NO>::Build(A->getRowMap(),false);
-    A->getLocalDiagCopy(*Mdiag);
-
-    // The pretty way of managing input decks
-    double shift = 1.0;
-    Teuchos::Array<double> cfls(2); cfls[0]=1.0; cfls[1]=cfls[0] - 1/(1.0+shift);
-    RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
-    Params->set("rap: algorithm","shift");
-    Params->set("rap: shift diagonal M",true);
-    Params->set("rap: shift low storage",true);
-    Params->set("rap: cfl array",cfls);
-    Params->set("coarse: max size",10);
-    Params->set("verbosity","high");
-    Params->set("max levels",4);
-    Teuchos::ParameterList & pLevel0 = Params->sublist("level 0");
-    pLevel0.set("Mdiag",Mdiag);
-    if(A->getRowMap()->lib() == Xpetra::UseEpetra) {
-      Params->set("use kokkos refactor", false);
-    }
-    Teuchos::ParameterList & user = Params->sublist("user data");
-    user.set("double deltaT",1.0);
-    user.set("double cfl",1.0);
-
-
-    // Build hierarchy
-    Teuchos::ParameterList& userParamList = Params->sublist("user data");
-    userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
-    RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(A,*Params);
-
-    // Ready the test vector
-    RCP<Level> Level1 = H->GetLevel(1);
-    RCP<Matrix> P = Level1->Get< RCP<Matrix> >("P");
-    RCP<Matrix> R = Level1->Get< RCP<Matrix> >("R");
-    RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(),1);
-    RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(),1);
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result2 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(),1);
-    X->randomize();
-
-    Scalar one = Teuchos::ScalarTraits<Scalar>::one();
-
-    //Calculate result1 = (P^T*(A+shift*M)*P)*X
-    P->apply(*X,*workVec1);
-    A->apply(*workVec1,*workVec2);
-    workVec2->elementWiseMultiply((Scalar)shift,*Mdiag,*workVec1,one);
-    P->apply(*workVec2,*result1,Teuchos::TRANS);
-
-    //Calculate result2 = (R*A*P)*X
-    RCP<Matrix> coarseOp = Level1->Get< RCP<Matrix> >("A");
-    coarseOp->apply(*X,*result2);
-
-    Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the (Non-)Galerkin triple "
-      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000*Teuchos::ScalarTraits<Scalar>::eps());
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RAPShiftFactory, CreatePreconditioner_Low_Storage_CFLSchedule, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2");
+#endif
+  out << "version: " << MueLu::Version() << std::endl;
+  typedef Scalar SC;
+  typedef GlobalOrdinal GO;
+  typedef LocalOrdinal LO;
+  typedef Node NO;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType MT;
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  using Teuchos::RCP;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  GO nx         = 1999 * comm->getSize();
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LO, GO, NO>::Build1DPoisson(nx);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+
+  // Extract the diagonal of A
+  RCP<Vector> Mdiag = Xpetra::VectorFactory<SC, LO, GO, NO>::Build(A->getRowMap(), false);
+  A->getLocalDiagCopy(*Mdiag);
+
+  // The pretty way of managing input decks
+  double shift = 1.0;
+  Teuchos::Array<double> cfls(2);
+  cfls[0]                            = 1.0;
+  cfls[1]                            = cfls[0] - 1 / (1.0 + shift);
+  RCP<Teuchos::ParameterList> Params = rcp(new Teuchos::ParameterList);
+  Params->set("rap: algorithm", "shift");
+  Params->set("rap: shift diagonal M", true);
+  Params->set("rap: shift low storage", true);
+  Params->set("rap: cfl array", cfls);
+  Params->set("coarse: max size", 10);
+  Params->set("verbosity", "high");
+  Params->set("max levels", 4);
+  Teuchos::ParameterList& pLevel0 = Params->sublist("level 0");
+  pLevel0.set("Mdiag", Mdiag);
+  if (A->getRowMap()->lib() == Xpetra::UseEpetra) {
+    Params->set("use kokkos refactor", false);
   }
-
-
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,Constructor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,Correctness,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,ImplicitTranspose,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Factory,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Easy,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Easy_Diagonal,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Low_Storage,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Low_Storage_ShiftList,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory,CreatePreconditioner_Low_Storage_CFLSchedule,Scalar,LO,GO,Node)
+  Teuchos::ParameterList& user = Params->sublist("user data");
+  user.set("double deltaT", 1.0);
+  user.set("double cfl", 1.0);
+
+  // Build hierarchy
+  Teuchos::ParameterList& userParamList = Params->sublist("user data");
+  userParamList.set<RCP<RealValuedMultiVector> >("Coordinates", coordinates);
+  RCP<Hierarchy> H = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(A, *Params);
+
+  // Ready the test vector
+  RCP<Level> Level1         = H->GetLevel(1);
+  RCP<Matrix> P             = Level1->Get<RCP<Matrix> >("P");
+  RCP<Matrix> R             = Level1->Get<RCP<Matrix> >("R");
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> result1  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result1a = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result2  = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  X->randomize();
+
+  Scalar one = Teuchos::ScalarTraits<Scalar>::one();
+
+  // Calculate result1 = (P^T*(A+shift*M)*P)*X
+  P->apply(*X, *workVec1);
+  A->apply(*workVec1, *workVec2);
+  workVec2->elementWiseMultiply((Scalar)shift, *Mdiag, *workVec1, one);
+  P->apply(*workVec2, *result1, Teuchos::TRANS);
+
+  // Calculate result2 = (R*A*P)*X
+  RCP<Matrix> coarseOp = Level1->Get<RCP<Matrix> >("A");
+  coarseOp->apply(*X, *result2);
+
+  Teuchos::Array<typename TST::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the (Non-)Galerkin triple "
+      << "matrix product by comparing (RAP)*X to R(A(P*X)), where R is the implicit transpose of P." << std::endl;
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1000 * Teuchos::ScalarTraits<Scalar>::eps());
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, Constructor, Scalar, LO, GO, Node)                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, Correctness, Scalar, LO, GO, Node)                                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, ImplicitTranspose, Scalar, LO, GO, Node)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Factory, Scalar, LO, GO, Node)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Easy, Scalar, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Easy_Diagonal, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Low_Storage, Scalar, LO, GO, Node)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Low_Storage_ShiftList, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RAPShiftFactory, CreatePreconditioner_Low_Storage_CFLSchedule, Scalar, LO, GO, Node)
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RebalanceAcFactory.cpp b/packages/muelu/test/unit_tests/RebalanceAcFactory.cpp
index 6c3d69f8ce46..f94cbb1d71a2 100644
--- a/packages/muelu/test/unit_tests/RebalanceAcFactory.cpp
+++ b/packages/muelu/test/unit_tests/RebalanceAcFactory.cpp
@@ -102,91 +102,82 @@
 
 namespace MueLuTests {
 
-  //this macro declares the unit-test-class:
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    auto rAcFactory = rcp(new RebalanceAcFactory());
-    TEST_ASSERT(!rAcFactory.is_null());
-    TEST_EQUALITY(rAcFactory->NumRebalanceFactories() == 0, true);
-
-    // Add Factory
-    rAcFactory->AddRebalanceFactory(rcp(new RebalanceTransferFactory()));
-    rAcFactory->AddRebalanceFactory(rcp(new RebalanceTransferFactory()));
-
-    TEST_EQUALITY(rAcFactory->NumRebalanceFactories() == 2, true);
-
-    auto paramList = rAcFactory->GetValidParameterList();
-    TEST_ASSERT(!paramList.is_null());
-
+// this macro declares the unit-test-class:
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  auto rAcFactory = rcp(new RebalanceAcFactory());
+  TEST_ASSERT(!rAcFactory.is_null());
+  TEST_EQUALITY(rAcFactory->NumRebalanceFactories() == 0, true);
+
+  // Add Factory
+  rAcFactory->AddRebalanceFactory(rcp(new RebalanceTransferFactory()));
+  rAcFactory->AddRebalanceFactory(rcp(new RebalanceTransferFactory()));
+
+  TEST_EQUALITY(rAcFactory->NumRebalanceFactories() == 2, true);
+
+  auto paramList = rAcFactory->GetValidParameterList();
+  TEST_ASSERT(!paramList.is_null());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, BuildWithImporter, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, BuildWithImporter, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    if (TestHelpers::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    Teuchos::CommandLineProcessor clp(false);
-    Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-
-    Level aLevel;
-    Level corseLevel;
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(aLevel, corseLevel);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2); //can be an empty operator
-    corseLevel.Set("A", A);
-    RCP<const Import> importer = ImportFactory::Build(A->getRowMap(), map);
-    corseLevel.Set("Importer", importer);
-
-    RCP<RebalanceAcFactory> RebalancedAFact = rcp(new RebalanceAcFactory());
-    RebalancedAFact->SetDefaultVerbLevel(MueLu::Extreme);
-    RebalancedAFact->Build(aLevel, corseLevel);
-
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, BuildWithoutImporter, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    Level aLevel;
-    Level corseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(aLevel, corseLevel);
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2); //can be an empty operator
-    corseLevel.Set("A", A);
-    RCP<const Import> importer  = Teuchos::null;
-    corseLevel.Set("Importer", importer);
-
-    RCP<RebalanceAcFactory> RebalancedAFact = rcp(new RebalanceAcFactory());
-    RebalancedAFact->SetDefaultVerbLevel(MueLu::Extreme);
-    RebalancedAFact->Build(aLevel, corseLevel);
-
-  }
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, BuildWithoutImporter, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, BuildWithImporter, Scalar, LO, GO, Node)
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  Teuchos::CommandLineProcessor clp(false);
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+
+  Level aLevel;
+  Level corseLevel;
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(aLevel, corseLevel);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);  // can be an empty operator
+  corseLevel.Set("A", A);
+  RCP<const Import> importer = ImportFactory::Build(A->getRowMap(), map);
+  corseLevel.Set("Importer", importer);
+
+  RCP<RebalanceAcFactory> RebalancedAFact = rcp(new RebalanceAcFactory());
+  RebalancedAFact->SetDefaultVerbLevel(MueLu::Extreme);
+  RebalancedAFact->Build(aLevel, corseLevel);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RebalanceAcFactory, BuildWithoutImporter, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  Level aLevel;
+  Level corseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(aLevel, corseLevel);
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2);  // can be an empty operator
+  corseLevel.Set("A", A);
+  RCP<const Import> importer = Teuchos::null;
+  corseLevel.Set("Importer", importer);
+
+  RCP<RebalanceAcFactory> RebalancedAFact = rcp(new RebalanceAcFactory());
+  RebalancedAFact->SetDefaultVerbLevel(MueLu::Extreme);
+  RebalancedAFact->Build(aLevel, corseLevel);
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, Constructor, Scalar, LO, GO, Node)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, BuildWithoutImporter, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RebalanceAcFactory, BuildWithImporter, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-}//namespace MueLuTests
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RegionMatrix.cpp b/packages/muelu/test/unit_tests/RegionMatrix.cpp
index 366e9815704e..7b69f205a92c 100644
--- a/packages/muelu/test/unit_tests/RegionMatrix.cpp
+++ b/packages/muelu/test/unit_tests/RegionMatrix.cpp
@@ -72,18 +72,18 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
                         RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
                         RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& colImport,
                         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regionMats,
-                        Teuchos::ArrayRCP<LocalOrdinal>&  regionMatVecLIDs,
+                        Teuchos::ArrayRCP<LocalOrdinal>& regionMatVecLIDs,
                         Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& regionInterfaceImporter) {
 #include <MueLu_UseShortNames.hpp>
 
   std::string matrixType = galeriList.get<std::string>("matrixType");
-  int numDimensions = 0;
+  int numDimensions      = 0;
   if (matrixType == "Laplace1D") {
     numDimensions = 1;
-  } else if(matrixType == "Laplace2D" || matrixType == "Elasticity2D" ||
-     matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+  } else if (matrixType == "Laplace2D" || matrixType == "Elasticity2D" ||
+             matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
     numDimensions = 2;
-  } else if(matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
     numDimensions = 3;
   }
 
@@ -97,9 +97,9 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   lNodesPerDir[0] = galeriList.get<LO>("lnx");
   lNodesPerDir[1] = (1 < numDimensions) ? galeriList.get<LO>("lny") : 1;
   lNodesPerDir[2] = (2 < numDimensions) ? galeriList.get<LO>("lnz") : 1;
-  procsPerDim[0] = galeriList.get<GO>("mx");
-  procsPerDim[1] = galeriList.get<GO>("my");
-  procsPerDim[2] = galeriList.get<GO>("mz");
+  procsPerDim[0]  = galeriList.get<GO>("mx");
+  procsPerDim[1]  = galeriList.get<GO>("my");
+  procsPerDim[2]  = galeriList.get<GO>("mz");
 
   // std::cout << "p=" << nodeMap->getComm()->getRank() << " | numDimensions=" << numDimensions
   //           << ", useStructured=" << false << ", numDofsPerNode=" << numDofsPerNode
@@ -107,17 +107,17 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   //           << ", procsPerDim=" << procsPerDim << std::endl;
 
   Array<int> boundaryConditions;
-  int maxRegPerGID = 0;
-  int numInterfaces = 0;
+  int maxRegPerGID       = 0;
+  int numInterfaces      = 0;
   LO numLocalRegionNodes = 0;
-  Array<GO>  sendGIDs;
+  Array<GO> sendGIDs;
   Array<int> sendPIDs;
-  Array<LO>  rNodesPerDim(3);
-  Array<LO>  compositeToRegionLIDs(nodeMap->getLocalNumElements()*numDofsPerNode);
-  Array<GO>  quasiRegionGIDs;
-  Array<GO>  quasiRegionCoordGIDs;
-  Array<GO>  interfaceCompositeGIDs, interfaceRegionGIDs;
-  Array<LO>  interfaceRegionLIDs;
+  Array<LO> rNodesPerDim(3);
+  Array<LO> compositeToRegionLIDs(nodeMap->getLocalNumElements() * numDofsPerNode);
+  Array<GO> quasiRegionGIDs;
+  Array<GO> quasiRegionCoordGIDs;
+  Array<GO> interfaceCompositeGIDs, interfaceRegionGIDs;
+  Array<LO> interfaceRegionLIDs;
   createRegionData(numDimensions, false, numDofsPerNode,
                    gNodesPerDir(), lNodesPerDir(), procsPerDim(),
                    nodeMap, dofMap,
@@ -136,18 +136,18 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   // std::cout << "p=" << myRank << " | interfaceCompositeGIDs" << interfaceCompositeGIDs << std::endl;
   // std::cout << "p=" << myRank << " | interfaceRegionLIDs" << interfaceRegionLIDs() << std::endl;
 
-  rowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs(),
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
+  rowMap = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                   Teuchos::OrdinalTraits<GO>::invalid(),
+                                                   quasiRegionGIDs(),
+                                                   A->getRowMap()->getIndexBase(),
+                                                   A->getRowMap()->getComm());
   colMap = rowMap;
 
-  revisedRowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs.size(),
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
+  revisedRowMap = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                          Teuchos::OrdinalTraits<GO>::invalid(),
+                                                          quasiRegionGIDs.size(),
+                                                          A->getRowMap()->getIndexBase(),
+                                                          A->getRowMap()->getComm());
   revisedColMap = revisedRowMap;
 
   ExtractListOfInterfaceRegionGIDs(revisedRowMap, interfaceRegionLIDs, interfaceRegionGIDs);
@@ -174,7 +174,7 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
                      revisedRowMap, revisedColMap,
                      rowImport, quasiRegionGrpMats, regionMats);
 
-} // createRegionMatrix
+}  // createRegionMatrix
 
 // Helper function that creates almost all the data needed to generate a unit-test
 // numDofsPerNode [in]: number of degrees of freedom per grid point
@@ -197,41 +197,41 @@ void createProblem(const LocalOrdinal numDofsPerNode,
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
   std::string mapType, coordinatesType;
-  if((matrixType == "Laplace2D") || (matrixType == "Elasticity2D")) {
-    mapType = "Cartesian2D";
+  if ((matrixType == "Laplace2D") || (matrixType == "Elasticity2D")) {
+    mapType         = "Cartesian2D";
     coordinatesType = "2D";
-  } else if((matrixType == "Laplace3D") || (matrixType == "Elasticity3D")) {
-    mapType = "Cartesian3D";
+  } else if ((matrixType == "Laplace3D") || (matrixType == "Elasticity3D")) {
+    mapType         = "Cartesian3D";
     coordinatesType = "3D";
   }
 
   // Build maps for the problem
   RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              mapType, comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap  = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   A = Pr->BuildMatrix();
   A->SetFixedBlockSize(numDofsPerNode);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>(coordinatesType, nodeMap, galeriList);
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>(coordinatesType, nodeMap, galeriList);
 
   // create the region maps, importer and operator from composite counter parts
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
                      rowMap, colMap, revisedRowMap, revisedColMap,
                      rowImport, colImport, regionMats,
@@ -244,8 +244,7 @@ void createProblem(const LocalOrdinal numDofsPerNode,
   // std::cout << "p=" << comm->getRank() << " | target map element list: "
   //           << regionInterfaceImporter->getTargetMap()->getLocalElementList() << std::endl;
 
-
-} // createProblem
+}  // createProblem
 
 // test_matrix() is checking that performing a MatVec with composite A and region A
 // yields the same vector. It also verifies that regionalToComposite(regA) returns
@@ -260,9 +259,9 @@ void test_matrix(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
                  Teuchos::FancyOStream& out,
                  bool& success) {
 #include <MueLu_UseShortNames.hpp>
-  using TST            = Teuchos::ScalarTraits<SC>;
-  using magnitude_type = typename TST::magnitudeType;
-  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  using TST                           = Teuchos::ScalarTraits<SC>;
+  using magnitude_type                = typename TST::magnitudeType;
+  using TMT                           = Teuchos::ScalarTraits<magnitude_type>;
   RCP<const Teuchos::Comm<int> > comm = A->getRowMap()->getComm();
 
   /************************************/
@@ -284,8 +283,8 @@ void test_matrix(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   // Now build the region X vector
   RCP<Vector> quasiRegX = Teuchos::null;
   RCP<Vector> quasiRegB = Teuchos::null;
-  RCP<Vector> regX = Teuchos::null;
-  RCP<Vector> regB = Teuchos::null;
+  RCP<Vector> regX      = Teuchos::null;
+  RCP<Vector> regB      = Teuchos::null;
   compositeToRegional(X, quasiRegX, regX,
                       revisedRowMap, rowImport);
   regB = VectorFactory::Build(revisedRowMap, true);
@@ -302,13 +301,12 @@ void test_matrix(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   // Extract the data from B and compB to compare it
   ArrayRCP<const SC> dataB     = B->getData(0);
   ArrayRCP<const SC> dataCompB = compB->getData(0);
-  for(size_t idx = 0; idx < B->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < B->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataB[idx]),
                            TST::magnitude(dataCompB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
-
   /************************************/
   /*                                  */
   /*  Second test: composite matrix   */
@@ -328,18 +326,18 @@ void test_matrix(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   using entries_type      = typename local_graph_type::entries_type;
   using values_type       = typename local_matrix_type::values_type;
 
-  local_matrix_type orignalA   = A->getLocalMatrixDevice(); // Local matrix
-  entries_type      refEntries = orignalA.graph.entries;    // view of local column indices
-  values_type       refValues  = orignalA.values;           // view of local values
+  local_matrix_type orignalA = A->getLocalMatrixDevice();  // Local matrix
+  entries_type refEntries    = orignalA.graph.entries;     // view of local column indices
+  values_type refValues      = orignalA.values;            // view of local values
 
   typename entries_type::HostMirror refEntries_h = Kokkos::create_mirror_view(refEntries);
   Kokkos::deep_copy(refEntries_h, refEntries);
   typename values_type::HostMirror refValues_h = Kokkos::create_mirror_view(refValues);
   Kokkos::deep_copy(refValues_h, refValues);
 
-  local_matrix_type compositeA       = compositeMatrix->getLocalMatrixDevice(); // Local matrix
-  entries_type      compositeEntries = compositeA.graph.entries;                // view of local column indices
-  values_type       compositeValues  = compositeA.values;                       // view of local values
+  local_matrix_type compositeA  = compositeMatrix->getLocalMatrixDevice();  // Local matrix
+  entries_type compositeEntries = compositeA.graph.entries;                 // view of local column indices
+  values_type compositeValues   = compositeA.values;                        // view of local values
 
   typename entries_type::HostMirror compositeEntries_h = Kokkos::create_mirror_view(compositeEntries);
   Kokkos::deep_copy(compositeEntries_h, compositeEntries);
@@ -347,54 +345,53 @@ void test_matrix(RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
   Kokkos::deep_copy(compositeValues_h, compositeValues);
 
   TEST_EQUALITY(compositeEntries_h.extent(0), refEntries_h.extent(0));
-  TEST_EQUALITY(compositeValues_h.extent(0),  refValues_h.extent(0));
-  for(LO idx = 0; idx < compositeEntries_h.extent_int(0); ++idx) {
+  TEST_EQUALITY(compositeValues_h.extent(0), refValues_h.extent(0));
+  for (LO idx = 0; idx < compositeEntries_h.extent_int(0); ++idx) {
     TEST_EQUALITY(compositeEntries_h(idx), refEntries_h(idx));
     TEST_FLOATING_EQUALITY(TST::magnitude(compositeValues_h(idx)),
                            TST::magnitude(refValues_h(idx)),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
-} // test_matrix
+}  // test_matrix
 
 // This test only checks the compositeToRegion() operation
 // for matrices. More specifically we compute a region A
 // based on a composite A and check the values in region A
 // against know correct values.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, CompositeToRegionMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, CompositeToRegionMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  LO numRanks = comm->getSize();
-  LO myRank = comm->getRank();
+  LO numRanks                         = comm->getSize();
+  LO myRank                           = comm->getRank();
 
   GO nx = 5, ny = 5, nz = 1;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian2D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
@@ -405,17 +402,17 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, CompositeToRegionMatrix, Scalar,
   RCP<Vector> B = VectorFactory::Build(dofMap);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
 
   // Create the region version of A called regionMats
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  RCP<Matrix> regionMats = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
   Teuchos::ArrayRCP<LO> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
@@ -429,9 +426,9 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, CompositeToRegionMatrix, Scalar,
   using entries_type      = typename local_graph_type::entries_type;
   using values_type       = typename local_matrix_type::values_type;
 
-  local_matrix_type myLocalA  = regionMats->getLocalMatrixDevice(); // Local matrix
-  entries_type      myEntries = myLocalA.graph.entries;             // view of local column indices
-  values_type       myValues  = myLocalA.values;                    // view of local values
+  local_matrix_type myLocalA = regionMats->getLocalMatrixDevice();  // Local matrix
+  entries_type myEntries     = myLocalA.graph.entries;              // view of local column indices
+  values_type myValues       = myLocalA.values;                     // view of local values
 
   typename entries_type::HostMirror myEntries_h = Kokkos::create_mirror_view(myEntries);
   Kokkos::deep_copy(myEntries_h, myEntries);
@@ -439,118 +436,117 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, CompositeToRegionMatrix, Scalar,
   Kokkos::deep_copy(myValues_h, myValues);
 
   // Now do a bunch of checks regarding the values stored in region A
-  if(numRanks == 1) {
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),     25);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),     25);
-    TEST_EQUALITY(regionMats->getLocalNumRows(),       25);
+  if (numRanks == 1) {
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 25);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 25);
+    TEST_EQUALITY(regionMats->getLocalNumRows(), 25);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 105);
-    TEST_EQUALITY(regionMats->getLocalNumEntries(),   105);
+    TEST_EQUALITY(regionMats->getLocalNumEntries(), 105);
 
     // In the serial case we can just compare to the values in A
-    entries_type refEntries = A->getLocalMatrixDevice().graph.entries;
-    values_type  refValues  = A->getLocalMatrixDevice().values;
+    entries_type refEntries                        = A->getLocalMatrixDevice().graph.entries;
+    values_type refValues                          = A->getLocalMatrixDevice().values;
     typename entries_type::HostMirror refEntries_h = Kokkos::create_mirror_view(refEntries);
     Kokkos::deep_copy(refEntries_h, refEntries);
     typename values_type::HostMirror refValues_h = Kokkos::create_mirror_view(refValues);
     Kokkos::deep_copy(refValues_h, refValues);
 
-    for(int idx = 0; idx < 105; ++idx) {
+    for (int idx = 0; idx < 105; ++idx) {
       TEST_EQUALITY(myEntries_h(idx), refEntries_h(idx));
       TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                              TST::magnitude(refValues_h(idx)),
-                             100*TMT::eps());
+                             100 * TMT::eps());
     }
-  } else if(numRanks == 4) {
+  } else if (numRanks == 4) {
     // All ranks will have the same number of rows/cols/entries
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),     36);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),     36);
-    TEST_EQUALITY(regionMats->getLocalNumRows(),        9);
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 36);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 36);
+    TEST_EQUALITY(regionMats->getLocalNumRows(), 9);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 132);
-    TEST_EQUALITY(regionMats->getLocalNumEntries(),    33);
+    TEST_EQUALITY(regionMats->getLocalNumEntries(), 33);
 
     ArrayRCP<LO> refEntries;
     ArrayRCP<SC> refValues;
     refEntries.deepCopy(ArrayView<const LO>({0, 1, 3,
-            0, 1, 2, 4,
-            1, 2, 5,
-            0, 3, 4, 6,
-            1, 3, 4, 5, 7,
-            2, 4, 5, 8,
-            3, 6, 7,
-            4, 6, 7, 8,
-            5, 7, 8}));
-    if(myRank == 0) {
+                                             0, 1, 2, 4,
+                                             1, 2, 5,
+                                             0, 3, 4, 6,
+                                             1, 3, 4, 5, 7,
+                                             2, 4, 5, 8,
+                                             3, 6, 7,
+                                             4, 6, 7, 8,
+                                             5, 7, 8}));
+    if (myRank == 0) {
       refValues.deepCopy(ArrayView<const SC>({4.0, -1.0, -1.0,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, 2.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -0.5, -1.0, 2.0, -0.5,
-              -1.0, 2.0, -0.5,
-              -1.0, -0.5, 2.0, -0.5,
-              -0.5, -0.5, 1.0}));
-
-    } else if(myRank == 1) {
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, 2.0, -0.5,
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -0.5, -1.0, 2.0, -0.5,
+                                              -1.0, 2.0, -0.5,
+                                              -1.0, -0.5, 2.0, -0.5,
+                                              -0.5, -0.5, 1.0}));
+
+    } else if (myRank == 1) {
       refValues.deepCopy(ArrayView<const SC>({2.0, -1.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, 4.0, -1.0,
-              -0.5, 2.0, -1.0, -0.5,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, 1.0, -0.5,
-              -1.0, -0.5, 2.0, -0.5,
-              -1.0, -0.5, 2.0}));
-
-    } else if(myRank == 2) {
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, 4.0, -1.0,
+                                              -0.5, 2.0, -1.0, -0.5,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, 1.0, -0.5,
+                                              -1.0, -0.5, 2.0, -0.5,
+                                              -1.0, -0.5, 2.0}));
+
+    } else if (myRank == 2) {
       refValues.deepCopy(ArrayView<const SC>({2.0, -0.5, -1.0,
-              -0.5, 2.0, -0.5, -1.0,
-              -0.5, 1.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -0.5, -1.0, 2.0, -0.5,
-              -1.0, 4.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, -1.0, 2.0}));
-
-    } else if(myRank == 3) {
+                                              -0.5, 2.0, -0.5, -1.0,
+                                              -0.5, 1.0, -0.5,
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -0.5, -1.0, 2.0, -0.5,
+                                              -1.0, 4.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, -1.0, 2.0}));
+
+    } else if (myRank == 3) {
       refValues.deepCopy(ArrayView<const SC>({1.0, -0.5, -0.5,
-              -0.5, 2.0, -0.5, -1.0,
-              -0.5, 2.0, -1.0,
-              -0.5, 2.0, -1.0, -0.5,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, 2.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -1.0, -1.0, 4.0}));
+                                              -0.5, 2.0, -0.5, -1.0,
+                                              -0.5, 2.0, -1.0,
+                                              -0.5, 2.0, -1.0, -0.5,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, 2.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -1.0, -1.0, 4.0}));
     }
 
     // Loop over region matrix data and compare it to ref data
-    for(int idx = 0; idx < 33; ++idx) {
+    for (int idx = 0; idx < 33; ++idx) {
       TEST_EQUALITY(myEntries_h(idx), refEntries[idx]);
       TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                              TST::magnitude(refValues[idx]),
-                             100*TMT::eps());
+                             100 * TMT::eps());
     }
   }
 
-} // CompositeToRegionMatrix
+}  // CompositeToRegionMatrix
 
 // Here the reverse operation: regionToComposite is tested.
 // For that check the idea is to build a composite problem
 // Then compute the equivalent region problem and finally
 // use the region problem to go back to composite which should
 // lead to a matrix identical to the original composite matrix
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -559,20 +555,20 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar,
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
   GO nx = 5, ny = 5, nz = 1;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian2D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
@@ -583,18 +579,18 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar,
   RCP<Vector> B = VectorFactory::Build(dofMap);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
   RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::
-    CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
+      CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
 
   // From the original composite matrix A, build the region equivalent: regionMats
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  RCP<Matrix> regionMats = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
   Teuchos::ArrayRCP<LO> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
@@ -619,18 +615,18 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar,
   using entries_type      = typename local_graph_type::entries_type;
   using values_type       = typename local_matrix_type::values_type;
 
-  local_matrix_type orignalA   = A->getLocalMatrixDevice(); // Local matrix
-  entries_type      refEntries = orignalA.graph.entries;    // view of local column indices
-  values_type       refValues  = orignalA.values;           // view of local values
+  local_matrix_type orignalA = A->getLocalMatrixDevice();  // Local matrix
+  entries_type refEntries    = orignalA.graph.entries;     // view of local column indices
+  values_type refValues      = orignalA.values;            // view of local values
 
   typename entries_type::HostMirror refEntries_h = Kokkos::create_mirror_view(refEntries);
   Kokkos::deep_copy(refEntries_h, refEntries);
   typename values_type::HostMirror refValues_h = Kokkos::create_mirror_view(refValues);
   Kokkos::deep_copy(refValues_h, refValues);
 
-  local_matrix_type compositeA       = compositeMatrix->getLocalMatrixDevice(); // Local matrix
-  entries_type      compositeEntries = compositeA.graph.entries;                // view of local column indices
-  values_type       compositeValues  = compositeA.values;                       // view of local values
+  local_matrix_type compositeA  = compositeMatrix->getLocalMatrixDevice();  // Local matrix
+  entries_type compositeEntries = compositeA.graph.entries;                 // view of local column indices
+  values_type compositeValues   = compositeA.values;                        // view of local values
 
   typename entries_type::HostMirror compositeEntries_h = Kokkos::create_mirror_view(compositeEntries);
   Kokkos::deep_copy(compositeEntries_h, compositeEntries);
@@ -638,33 +634,31 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, RegionToCompositeMatrix, Scalar,
   Kokkos::deep_copy(compositeValues_h, compositeValues);
 
   TEST_EQUALITY(compositeEntries_h.extent(0), refEntries_h.extent(0));
-  TEST_EQUALITY(compositeValues_h.extent(0),  refValues_h.extent(0));
-  for(LO idx = 0; idx < compositeEntries_h.extent_int(0); ++idx) {
+  TEST_EQUALITY(compositeValues_h.extent(0), refValues_h.extent(0));
+  for (LO idx = 0; idx < compositeEntries_h.extent_int(0); ++idx) {
     TEST_EQUALITY(compositeEntries_h(idx), refEntries_h(idx));
     TEST_FLOATING_EQUALITY(TST::magnitude(compositeValues_h(idx)),
                            TST::magnitude(refValues_h(idx)),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
-} // RegionToCompositeMatrix
-
+}  // RegionToCompositeMatrix
 
 // This test aims at checking that apply regionA to a region vector has the same effect as
 // applying A to a composite vector. Of course the region vector needs to be brought back
 // to composite formate before verifying the equivalence.
 //
 // Do this for 1 DOF per node
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -673,37 +667,37 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
   GO nx = 7, ny = 7, nz = 1;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian2D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
   A->SetFixedBlockSize(numDofsPerNode);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
 
   // create the region maps, importer and operator from composite counter parts
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  RCP<Matrix> regionMats = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
   Teuchos::ArrayRCP<LO> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
@@ -726,8 +720,8 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal
   // Create the region vectors and apply region A
   RCP<Vector> quasiRegX = Teuchos::null;
   RCP<Vector> quasiRegB = Teuchos::null;
-  RCP<Vector> regX = Teuchos::null;
-  RCP<Vector> regB = Teuchos::null;
+  RCP<Vector> regX      = Teuchos::null;
+  RCP<Vector> regB      = Teuchos::null;
   compositeToRegional(X, quasiRegX, regX, revisedRowMap, rowImport);
   regB = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regB, Teuchos::NO_TRANS, TST::one(), TST::zero());
@@ -741,10 +735,10 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal
   // Extract the data from B and compB to compare it
   ArrayRCP<const SC> dataRegB    = regB->getData(0);
   ArrayRCP<const SC> dataRefRegB = refRegB->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegB[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
   // Finally we perform the "fastMatVec" that does not require
@@ -754,34 +748,33 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec, Scalar, LocalOrdinal
   RCP<const Map> regionMap = revisedRowMap;
 
   RCP<Vector> regC = Teuchos::null;
-  regC = VectorFactory::Build(revisedRowMap, true);
+  regC             = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regC, Teuchos::NO_TRANS, TST::one(), TST::zero(), true, regionInterfaceImporter, regionMatVecLIDs);
 
   ArrayRCP<const SC> dataRegC = regC->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegC[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
-} // FastMatVec
+}  // FastMatVec
 
 // This test aims at checking that apply regionA to a region vector has the same effect as
 // applying A to a composite vector. Of course the region vector needs to be brought back
 // to composite formate before verifying the equivalence.
 //
 // Do this for 1 DOF per node
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -790,17 +783,17 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdin
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
   GO nx = 5, ny = 5, nz = 3;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
-  const LO numDofsPerNode = 1;
+  std::string matrixType            = galeriParameters.GetMatrixType();
+  const LO numDofsPerNode           = 1;
 
   // create the region maps, importer and operator from composite counter parts
   RCP<Matrix> regionMats = Teuchos::null;
   RCP<Matrix> A;
   RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
 
@@ -823,8 +816,8 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdin
   // Create the region vectors and apply region A
   RCP<Vector> quasiRegX = Teuchos::null;
   RCP<Vector> quasiRegB = Teuchos::null;
-  RCP<Vector> regX = Teuchos::null;
-  RCP<Vector> regB = Teuchos::null;
+  RCP<Vector> regX      = Teuchos::null;
+  RCP<Vector> regB      = Teuchos::null;
   compositeToRegional(X, quasiRegX, regX,
                       revisedRowMap, rowImport);
   regB = VectorFactory::Build(revisedRowMap, true);
@@ -839,10 +832,10 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdin
   // Extract the data from B and compB to compare it
   ArrayRCP<const SC> dataRegB    = regB->getData(0);
   ArrayRCP<const SC> dataRefRegB = refRegB->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegB[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
   // Finally we perform the "fastMatVec" that does not require
@@ -852,34 +845,33 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D, Scalar, LocalOrdin
   RCP<const Map> regionMap = revisedRowMap;
 
   RCP<Vector> regC = Teuchos::null;
-  regC = VectorFactory::Build(revisedRowMap, true);
+  regC             = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regC, Teuchos::NO_TRANS, TST::one(), TST::zero(), true, regionInterfaceImporter, regionMatVecLIDs);
 
   ArrayRCP<const SC> dataRegC = regC->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegC[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
-} // FastMatVec3D
+}  // FastMatVec3D
 
 // This test aims at checking that apply regionA to a region vector has the same effect as
 // applying A to a composite vector. Of course the region vector needs to be brought back
 // to composite formate before verifying the equivalence.
 //
 // Do this for 2 DOFs per node (two-dimensional elasticity)
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -889,12 +881,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar,
 
   const LO numDofsPerNode = 2;
   GO nx = 7, ny = 7, nz = 1;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Elasticity2D");
   RCP<Matrix> regionMats = Teuchos::null;
   RCP<Matrix> A;
   RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
 
@@ -917,8 +909,8 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar,
   // Create the region vectors and apply region A
   RCP<Vector> quasiRegX = Teuchos::null;
   RCP<Vector> quasiRegB = Teuchos::null;
-  RCP<Vector> regX = Teuchos::null;
-  RCP<Vector> regB = Teuchos::null;
+  RCP<Vector> regX      = Teuchos::null;
+  RCP<Vector> regB      = Teuchos::null;
   compositeToRegional(X, quasiRegX, regX, revisedRowMap, rowImport);
   regB = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regB, Teuchos::NO_TRANS, TST::one(), TST::zero());
@@ -932,10 +924,10 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar,
   // Extract the data from B and compB to compare it
   ArrayRCP<const SC> dataRegB    = regB->getData(0);
   ArrayRCP<const SC> dataRefRegB = refRegB->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegB[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
   // Finally we perform the "fastMatVec" that does not require
@@ -945,34 +937,33 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec2D_Elasticity, Scalar,
   RCP<const Map> regionMap = revisedRowMap;
 
   RCP<Vector> regC = Teuchos::null;
-  regC = VectorFactory::Build(revisedRowMap, true);
+  regC             = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regC, Teuchos::NO_TRANS, TST::one(), TST::zero(), true, regionInterfaceImporter, regionMatVecLIDs);
 
   ArrayRCP<const SC> dataRegC = regC->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegC[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
-} // FastMatVec2D_Elasticity
+}  // FastMatVec2D_Elasticity
 
 // This test aims at checking that apply regionA to a region vector has the same effect as
 // applying A to a composite vector. Of course the region vector needs to be brought back
 // to composite formate before verifying the equivalence.
 //
 // Do this for 3 DOFs per node (three-dimensional elasticity)
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -982,12 +973,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar,
 
   const LO numDofsPerNode = 3;
   GO nx = 5, ny = 5, nz = 3;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Elasticity3D");
-  RCP<Matrix> regionMats = Teuchos::null;
-  RCP<Matrix> A = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
+  RCP<Matrix> A                = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
 
@@ -1010,8 +1001,8 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar,
   // Create the region vectors and apply region A
   RCP<Vector> quasiRegX = Teuchos::null;
   RCP<Vector> quasiRegB = Teuchos::null;
-  RCP<Vector> regX = Teuchos::null;
-  RCP<Vector> regB = Teuchos::null;
+  RCP<Vector> regX      = Teuchos::null;
+  RCP<Vector> regB      = Teuchos::null;
   compositeToRegional(X, quasiRegX, regX, revisedRowMap, rowImport);
   regB = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regB, Teuchos::NO_TRANS, TST::one(), TST::zero());
@@ -1025,10 +1016,10 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar,
   // Extract the data from B and compB to compare it
   ArrayRCP<const SC> dataRegB    = regB->getData(0);
   ArrayRCP<const SC> dataRefRegB = refRegB->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegB[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           100*TMT::eps());
+                           100 * TMT::eps());
   }
 
   // Finally we perform the "fastMatVec" that does not require
@@ -1038,33 +1029,32 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, FastMatVec3D_Elasticity, Scalar,
   RCP<const Map> regionMap = revisedRowMap;
 
   RCP<Vector> regC = Teuchos::null;
-  regC = VectorFactory::Build(revisedRowMap, true);
+  regC             = VectorFactory::Build(revisedRowMap, true);
   regionMats->apply(*regX, *regC, Teuchos::NO_TRANS, TST::one(), TST::zero(), true, regionInterfaceImporter, regionMatVecLIDs);
 
   ArrayRCP<const SC> dataRegC = regC->getData(0);
-  for(size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
+  for (size_t idx = 0; idx < refRegB->getLocalLength(); ++idx) {
     TEST_FLOATING_EQUALITY(TST::magnitude(dataRegC[idx]),
                            TST::magnitude(dataRefRegB[idx]),
-                           200*TMT::eps());
+                           200 * TMT::eps());
   }
 
-} // FastMatVec3D_Elasticity
+}  // FastMatVec3D_Elasticity
 
 // Here a Laplace 2D problem is tested for all the above checks mentioned:
 //   1) the region operator is compared against know values
 //   2) the action of the region MatVec is compared with the composite MatVec
 //   3) compute the composite operator from the region operator leads to the original matrix
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -1073,36 +1063,36 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace2D, Scalar, LocalOrdinal,
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
   GO nx = 6, ny = 5, nz = 1;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian2D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
   A->SetFixedBlockSize(numDofsPerNode);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
 
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  RCP<Matrix> regionMats = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
   Teuchos::ArrayRCP<LO> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
@@ -1120,9 +1110,9 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace2D, Scalar, LocalOrdinal,
   using entries_type      = typename local_graph_type::entries_type;
   using values_type       = typename local_matrix_type::values_type;
 
-  local_matrix_type myLocalA  = regionMats->getLocalMatrixDevice(); // Local matrix
-  entries_type      myEntries = myLocalA.graph.entries;             // view of local column indices
-  values_type       myValues  = myLocalA.values;                    // view of local values
+  local_matrix_type myLocalA = regionMats->getLocalMatrixDevice();  // Local matrix
+  entries_type myEntries     = myLocalA.graph.entries;              // view of local column indices
+  values_type myValues       = myLocalA.values;                     // view of local values
 
   typename entries_type::HostMirror myEntries_h = Kokkos::create_mirror_view(myEntries);
   Kokkos::deep_copy(myEntries_h, myEntries);
@@ -1131,118 +1121,116 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace2D, Scalar, LocalOrdinal,
 
   const int numRanks = comm->getSize();
   const int myRank   = comm->getRank();
-  if(numRanks == 1) {
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),    30);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),    30);
-    TEST_EQUALITY(regionMats->getLocalNumRows(),      30);
+  if (numRanks == 1) {
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 30);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 30);
+    TEST_EQUALITY(regionMats->getLocalNumRows(), 30);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 128);
-    TEST_EQUALITY(regionMats->getLocalNumEntries(),   128);
+    TEST_EQUALITY(regionMats->getLocalNumEntries(), 128);
 
     // In the serial case we can just compare to the values in A
-    entries_type refEntries = A->getLocalMatrixDevice().graph.entries;
-    values_type  refValues  = A->getLocalMatrixDevice().values;
+    entries_type refEntries                        = A->getLocalMatrixDevice().graph.entries;
+    values_type refValues                          = A->getLocalMatrixDevice().values;
     typename entries_type::HostMirror refEntries_h = Kokkos::create_mirror_view(refEntries);
     Kokkos::deep_copy(refEntries_h, refEntries);
     typename values_type::HostMirror refValues_h = Kokkos::create_mirror_view(refValues);
     Kokkos::deep_copy(refValues_h, refValues);
 
-    for(int idx = 0; idx < 128; ++idx) {
+    for (int idx = 0; idx < 128; ++idx) {
       TEST_EQUALITY(myEntries_h(idx), refEntries_h(idx));
       TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                              TST::magnitude(refValues_h(idx)),
-                             100*TMT::eps());
+                             100 * TMT::eps());
     }
-  } else if(numRanks == 4) {
+  } else if (numRanks == 4) {
     // All ranks will have the same number of rows/cols/entries
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),    42);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),    42);
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 42);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 42);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 158);
 
     ArrayRCP<SC> refValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),      9);
-      TEST_EQUALITY(regionMats->getLocalNumEntries(),   33);
+    if (myRank == 0) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 9);
+      TEST_EQUALITY(regionMats->getLocalNumEntries(), 33);
       refValues.deepCopy(ArrayView<const SC>({4.0, -1.0, -1.0,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, 2.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -0.5, -1.0, 2.0, -0.5,
-              -1.0, 2.0, -0.5,
-              -1.0, -0.5, 2.0, -0.5,
-              -0.5, -0.5, 1.0}));
-
-    } else if(myRank == 1) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),      12);
-      TEST_EQUALITY(regionMats->getLocalNumEntries(),   46);
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, 2.0, -0.5,
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -0.5, -1.0, 2.0, -0.5,
+                                              -1.0, 2.0, -0.5,
+                                              -1.0, -0.5, 2.0, -0.5,
+                                              -0.5, -0.5, 1.0}));
+
+    } else if (myRank == 1) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 12);
+      TEST_EQUALITY(regionMats->getLocalNumEntries(), 46);
       refValues.deepCopy(ArrayView<const SC>({2.0, -1.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, 4.0, -1.0,
-              -0.5, 2.0, -1.0, -0.5,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, 1.0, -0.5,
-              -1.0, -0.5, 2.0, -0.5,
-              -1.0, -0.5, 2.0, -0.5,
-              -1.0, -0.5, 2.0}));
-
-    } else if(myRank == 2) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),      9);
-      TEST_EQUALITY(regionMats->getLocalNumEntries(),   33);
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, 4.0, -1.0,
+                                              -0.5, 2.0, -1.0, -0.5,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, 1.0, -0.5,
+                                              -1.0, -0.5, 2.0, -0.5,
+                                              -1.0, -0.5, 2.0, -0.5,
+                                              -1.0, -0.5, 2.0}));
+
+    } else if (myRank == 2) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 9);
+      TEST_EQUALITY(regionMats->getLocalNumEntries(), 33);
       refValues.deepCopy(ArrayView<const SC>({2.0, -0.5, -1.0,
-              -0.5, 2.0, -0.5, -1.0,
-              -0.5, 1.0, -0.5,
-              -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -0.5, -1.0, 2.0, -0.5,
-              -1.0, 4.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, -1.0, 2.0}));
-
-    } else if(myRank == 3) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),      12);
-      TEST_EQUALITY(regionMats->getLocalNumEntries(),   46);
+                                              -0.5, 2.0, -0.5, -1.0,
+                                              -0.5, 1.0, -0.5,
+                                              -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -0.5, -1.0, 2.0, -0.5,
+                                              -1.0, 4.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, -1.0, 2.0}));
+
+    } else if (myRank == 3) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 12);
+      TEST_EQUALITY(regionMats->getLocalNumEntries(), 46);
       refValues.deepCopy(ArrayView<const SC>({1.0, -0.5, -0.5,
-              -0.5, 2.0, -0.5, -1.0,
-              -0.5, 2.0, -0.5, -1.0,
-              -0.5, 2.0, -1.0,
-              -0.5, 2.0, -1.0, -0.5,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -0.5, 2.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -1.0, -1.0, 4.0, -1.0,
-              -1.0, -1.0, 4.0}));
+                                              -0.5, 2.0, -0.5, -1.0,
+                                              -0.5, 2.0, -0.5, -1.0,
+                                              -0.5, 2.0, -1.0,
+                                              -0.5, 2.0, -1.0, -0.5,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -0.5, 2.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -1.0, -1.0, 4.0, -1.0,
+                                              -1.0, -1.0, 4.0}));
     }
 
     // Loop over region matrix data and compare it to ref data
-    for(int idx = 0; idx < 33; ++idx) {
+    for (int idx = 0; idx < 33; ++idx) {
       TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                              TST::magnitude(refValues[idx]),
-                             100*TMT::eps());
+                             100 * TMT::eps());
     }
   }
-} // Laplace2D
-
+}  // Laplace2D
 
 // Here a Laplace 3D problem is tested for all the above checks mentioned:
 //   1) the region operator is compared against know values
 //   2) the action of the region MatVec is compared with the composite MatVec
 //   3) compute the composite operator from the region operator leads to the original matrix
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
@@ -1251,36 +1239,36 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace3D, Scalar, LocalOrdinal,
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
   GO nx = 6, ny = 5, nz = 4;
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian3D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
   A->SetFixedBlockSize(numDofsPerNode);
 
   // Create auxiliary data for MG
-  RCP<MultiVector> nullspace = Pr->BuildNullspace();
-  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>("3D", nodeMap, galeriList);
+  RCP<MultiVector> nullspace             = Pr->BuildNullspace();
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>("3D", nodeMap, galeriList);
 
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  RCP<Matrix> regionMats = Teuchos::null;
+  RCP<Import> rowImport        = Teuchos::null;
+  RCP<Import> colImport        = Teuchos::null;
+  RCP<Matrix> regionMats       = Teuchos::null;
   Teuchos::ArrayRCP<LO> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
@@ -1298,9 +1286,9 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace3D, Scalar, LocalOrdinal,
   using entries_type      = typename local_graph_type::entries_type;
   using values_type       = typename local_matrix_type::values_type;
 
-  local_matrix_type myLocalA  = regionMats->getLocalMatrixDevice();   // Local matrix
-  entries_type      myEntries = myLocalA.graph.entries;               // view of local column indices
-  values_type       myValues  = myLocalA.values;                      // view of local values
+  local_matrix_type myLocalA = regionMats->getLocalMatrixDevice();  // Local matrix
+  entries_type myEntries     = myLocalA.graph.entries;              // view of local column indices
+  values_type myValues       = myLocalA.values;                     // view of local values
 
   typename entries_type::HostMirror myEntries_h = Kokkos::create_mirror_view(myEntries);
   Kokkos::deep_copy(myEntries_h, myEntries);
@@ -1309,260 +1297,259 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionMatrix, Laplace3D, Scalar, LocalOrdinal,
 
   const int numRanks = comm->getSize();
   const int myRank   = comm->getRank();
-  if((numRanks == 1) && (myRank == 0)) {
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),    120);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),    120);
-    TEST_EQUALITY(regionMats->getLocalNumRows(),      120);
+  if ((numRanks == 1) && (myRank == 0)) {
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 120);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 120);
+    TEST_EQUALITY(regionMats->getLocalNumRows(), 120);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 692);
-    TEST_EQUALITY(regionMats->getLocalNumEntries(),   692);
+    TEST_EQUALITY(regionMats->getLocalNumEntries(), 692);
 
     // In the serial case we can just compare to the values in A
-    entries_type refEntries = A->getLocalMatrixDevice().graph.entries;
-    values_type  refValues  = A->getLocalMatrixDevice().values;
+    entries_type refEntries                        = A->getLocalMatrixDevice().graph.entries;
+    values_type refValues                          = A->getLocalMatrixDevice().values;
     typename entries_type::HostMirror refEntries_h = Kokkos::create_mirror_view(refEntries);
     Kokkos::deep_copy(refEntries_h, refEntries);
     typename values_type::HostMirror refValues_h = Kokkos::create_mirror_view(refValues);
     Kokkos::deep_copy(refValues_h, refValues);
 
-    for(int idx = 0; idx < refEntries_h.extent_int(0); ++idx) {
+    for (int idx = 0; idx < refEntries_h.extent_int(0); ++idx) {
       TEST_EQUALITY(myEntries_h(idx), refEntries_h(idx));
       TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                              TST::magnitude(refValues_h(idx)),
-                             100*TMT::eps());
+                             100 * TMT::eps());
     }
-  } else if(numRanks == 4) {
+  } else if (numRanks == 4) {
     // All ranks will have the same number of rows/cols/entries
-    TEST_EQUALITY(regionMats->getGlobalNumRows(),    168);
-    TEST_EQUALITY(regionMats->getGlobalNumCols(),    168);
+    TEST_EQUALITY(regionMats->getGlobalNumRows(), 168);
+    TEST_EQUALITY(regionMats->getGlobalNumCols(), 168);
     TEST_EQUALITY(regionMats->getGlobalNumEntries(), 884);
 
     ArrayRCP<SC> refValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),     36);
+    if (myRank == 0) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 36);
       TEST_EQUALITY(regionMats->getLocalNumEntries(), 186);
       refValues.deepCopy(ArrayView<const SC>({6.0, -1.0, -1.0, -1.0,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, 3.0, -0.5, -0.5,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, 3.0, -0.5, -0.5,
-              -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.5, -0.5, 1.5, -0.25,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.25, -0.5, -0.5, 1.5, -0.25,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.25, -0.5, -0.5, 1.5, -0.25,
-              -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5,
-              -0.5, -1.0, 3.0, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5,
-              -0.25, -0.5, -0.5, 1.5}));
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.5, -0.5, 1.5, -0.25,
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.25, -0.5, -0.5, 1.5, -0.25,
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.25, -0.5, -0.5, 1.5, -0.25,
+                                              -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5,
+                                              -0.5, -1.0, 3.0, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5,
+                                              -0.25, -0.5, -0.5, 1.5}));
 
       // Loop over region matrix data and compare it to ref data
-      for(int idx = 0; idx < 186; ++idx) {
+      for (int idx = 0; idx < 186; ++idx) {
         TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                                TST::magnitude(refValues[idx]),
-                               100*TMT::eps());
+                               100 * TMT::eps());
       }
 
-    } else if(myRank == 1) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),     48);
+    } else if (myRank == 1) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 48);
       TEST_EQUALITY(regionMats->getLocalNumEntries(), 256);
       refValues.deepCopy(ArrayView<const SC>({3.0, -1.0, -0.5, -0.5,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, 6.0, -1.0, -1.0,
-              -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, 1.5, -0.5, -0.25,
-              -1.0, -0.5, 3.0, -0.5, -0.5,
-              -1.0, -0.5, 3.0, -0.5, -0.5,
-              -1.0, -0.5, 3.0, -0.5,
-              -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.25, -0.5, 1.5, -0.5, -0.25,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5,
-              -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.25, -0.5, 1.5, -0.5, -0.25,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5,
-              -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -0.25, -0.5, 1.5, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5,
-              -0.5, -1.0, -0.5, 3.0, -0.5,
-              -0.5, -1.0, -0.5, 3.0}));
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, 1.5, -0.5, -0.25,
+                                              -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -1.0, -0.5, 3.0, -0.5,
+                                              -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.25, -0.5, 1.5, -0.5, -0.25,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5,
+                                              -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.25, -0.5, 1.5, -0.5, -0.25,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5,
+                                              -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -0.25, -0.5, 1.5, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0, -0.5,
+                                              -0.5, -1.0, -0.5, 3.0}));
 
       // Loop over region matrix data and compare it to ref data
-      for(int idx = 0; idx < 256; ++idx) {
+      for (int idx = 0; idx < 256; ++idx) {
         TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                                TST::magnitude(refValues[idx]),
-                               100*TMT::eps());
+                               100 * TMT::eps());
       }
 
-    } else if(myRank == 2) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),     36);
+    } else if (myRank == 2) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 36);
       TEST_EQUALITY(regionMats->getLocalNumEntries(), 186);
       refValues.deepCopy(ArrayView<const SC>({3.0, -0.5, -1.0, -0.5,
-              -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, 1.5, -0.5, -0.25,
-              -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -1.0, 3.0, -0.5,
-              -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.25, -0.5, 1.5, -0.5, -0.25,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5,
-              -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.25, -0.5, 1.5, -0.5, -0.25,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5,
-              -0.5, 3.0, -0.5, -1.0,
-              -0.5, -0.5, 3.0, -0.5, -1.0,
-              -0.25, -0.5, 1.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0, -0.5,
-              -1.0, -1.0, 6.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -0.5, -0.5, -1.0, 3.0}));
+                                              -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, 1.5, -0.5, -0.25,
+                                              -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -1.0, 3.0, -0.5,
+                                              -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.25, -0.5, 1.5, -0.5, -0.25,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5,
+                                              -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.25, -0.5, 1.5, -0.5, -0.25,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5,
+                                              -0.5, 3.0, -0.5, -1.0,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0,
+                                              -0.25, -0.5, 1.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -0.5, -0.5, -1.0, 3.0}));
 
       // Loop over region matrix data and compare it to ref data
-      for(int idx = 0; idx < 186; ++idx) {
+      for (int idx = 0; idx < 186; ++idx) {
         TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                                TST::magnitude(refValues[idx]),
-                               100*TMT::eps());
+                               100 * TMT::eps());
       }
 
-    } else if(myRank == 3) {
-      TEST_EQUALITY(regionMats->getLocalNumRows(),     48);
+    } else if (myRank == 3) {
+      TEST_EQUALITY(regionMats->getLocalNumRows(), 48);
       TEST_EQUALITY(regionMats->getLocalNumEntries(), 256);
       refValues.deepCopy(ArrayView<const SC>({1.5, -0.5, -0.5, -0.25,
-              -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, 3.0, -1.0, -0.5,
-              -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, 6.0, -1.0,
-              -0.25, 1.5, -0.5, -0.5, -0.25,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -0.25, 1.5, -0.5, -0.5, -0.25,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -0.25, 1.5, -0.5, -0.5,
-              -0.5, -0.5, 3.0, -0.5, -1.0,
-              -0.5, -0.5, 3.0, -0.5, -1.0,
-              -0.5, -0.5, 3.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0, -0.5,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -0.5, -0.5, 3.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0, -1.0,
-              -1.0, -1.0, -1.0, 6.0}));
+                                              -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, 3.0, -1.0, -0.5,
+                                              -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, 6.0, -1.0,
+                                              -0.25, 1.5, -0.5, -0.5, -0.25,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -0.25, 1.5, -0.5, -0.5, -0.25,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -0.25, 1.5, -0.5, -0.5,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0,
+                                              -0.5, -0.5, 3.0, -0.5, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0, -0.5,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -0.5, -0.5, 3.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0, -1.0,
+                                              -1.0, -1.0, -1.0, 6.0}));
 
       // Loop over region matrix data and compare it to ref data
-      for(int idx = 0; idx < 256; ++idx) {
+      for (int idx = 0; idx < 256; ++idx) {
         TEST_FLOATING_EQUALITY(TST::magnitude(myValues_h(idx)),
                                TST::magnitude(refValues[idx]),
-                               100*TMT::eps());
+                               100 * TMT::eps());
       }
     }
   }
-} // Laplace3D
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,CompositeToRegionMatrix,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,RegionToCompositeMatrix,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,FastMatVec,Scalar,LO,GO,Node)              \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,FastMatVec3D,Scalar,LO,GO,Node)            \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,FastMatVec2D_Elasticity,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,FastMatVec3D_Elasticity,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,Laplace2D,Scalar,LO,GO,Node)               \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix,Laplace3D,Scalar,LO,GO,Node)
+}  // Laplace3D
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, CompositeToRegionMatrix, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, RegionToCompositeMatrix, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, FastMatVec, Scalar, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, FastMatVec3D, Scalar, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, FastMatVec2D_Elasticity, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, FastMatVec3D_Elasticity, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, Laplace2D, Scalar, LO, GO, Node)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionMatrix, Laplace3D, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RegionRFactory.cpp b/packages/muelu/test/unit_tests/RegionRFactory.cpp
index 71b76596e7d2..d4a09fdfe009 100644
--- a/packages/muelu/test/unit_tests/RegionRFactory.cpp
+++ b/packages/muelu/test/unit_tests/RegionRFactory.cpp
@@ -75,7 +75,7 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
                         RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& rowImport,
                         RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& colImport,
                         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& regionMats,
-                        Teuchos::ArrayRCP<LocalOrdinal>&  regionMatVecLIDs,
+                        Teuchos::ArrayRCP<LocalOrdinal>& regionMatVecLIDs,
                         Teuchos::Array<GlobalOrdinal>& quasiRegionCoordGIDs,
                         Teuchos::RCP<Xpetra::Import<LocalOrdinal, GlobalOrdinal, Node> >& regionInterfaceImporter,
                         Teuchos::Array<LocalOrdinal>& rNodesPerDim,
@@ -83,13 +83,13 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
 #include <MueLu_UseShortNames.hpp>
 
   std::string matrixType = galeriList.get<std::string>("matrixType");
-  int numDimensions = 0;
+  int numDimensions      = 0;
   if (matrixType == "Laplace1D") {
     numDimensions = 1;
-  } else if(matrixType == "Laplace2D" || matrixType == "Elasticity2D" ||
-     matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+  } else if (matrixType == "Laplace2D" || matrixType == "Elasticity2D" ||
+             matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
     numDimensions = 2;
-  } else if(matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
+  } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
     numDimensions = 3;
   }
 
@@ -103,9 +103,9 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   lNodesPerDir[0] = galeriList.get<LO>("lnx");
   lNodesPerDir[1] = (1 < numDimensions) ? galeriList.get<LO>("lny") : 1;
   lNodesPerDir[2] = (2 < numDimensions) ? galeriList.get<LO>("lnz") : 1;
-  procsPerDim[0] = galeriList.get<GO>("mx");
-  procsPerDim[1] = galeriList.get<GO>("my");
-  procsPerDim[2] = galeriList.get<GO>("mz");
+  procsPerDim[0]  = galeriList.get<GO>("mx");
+  procsPerDim[1]  = galeriList.get<GO>("my");
+  procsPerDim[2]  = galeriList.get<GO>("mz");
 
   // std::cout << "p=" << nodeMap->getComm()->getRank() << " | numDimensions=" << numDimensions
   //           << ", useStructured=" << false << ", numDofsPerNode=" << numDofsPerNode
@@ -113,14 +113,14 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   //           << ", procsPerDim=" << procsPerDim << std::endl;
 
   Array<int> boundaryConditions;
-  int maxRegPerGID = 0;
+  int maxRegPerGID  = 0;
   int numInterfaces = 0;
-  Array<GO>  sendGIDs;
+  Array<GO> sendGIDs;
   Array<int> sendPIDs;
-  Array<LO>  compositeToRegionLIDs(nodeMap->getLocalNumElements()*numDofsPerNode);
-  Array<GO>  quasiRegionGIDs;
-  Array<GO>  interfaceCompositeGIDs, interfaceRegionGIDs;
-  Array<LO>  interfaceRegionLIDs;
+  Array<LO> compositeToRegionLIDs(nodeMap->getLocalNumElements() * numDofsPerNode);
+  Array<GO> quasiRegionGIDs;
+  Array<GO> interfaceCompositeGIDs, interfaceRegionGIDs;
+  Array<LO> interfaceRegionLIDs;
   createRegionData(numDimensions, false, numDofsPerNode,
                    gNodesPerDir(), lNodesPerDir(), procsPerDim(),
                    nodeMap, dofMap,
@@ -139,18 +139,18 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
   // std::cout << "p=" << myRank << " | interfaceCompositeGIDs" << interfaceCompositeGIDs << std::endl;
   // std::cout << "p=" << myRank << " | interfaceRegionLIDs" << interfaceRegionLIDs() << std::endl;
 
-  rowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs(),
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
-  colMap= rowMap;
-
-  revisedRowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs.size(),
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
+  rowMap = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                   Teuchos::OrdinalTraits<GO>::invalid(),
+                                                   quasiRegionGIDs(),
+                                                   A->getRowMap()->getIndexBase(),
+                                                   A->getRowMap()->getComm());
+  colMap = rowMap;
+
+  revisedRowMap = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                          Teuchos::OrdinalTraits<GO>::invalid(),
+                                                          quasiRegionGIDs.size(),
+                                                          A->getRowMap()->getIndexBase(),
+                                                          A->getRowMap()->getComm());
   revisedColMap = revisedRowMap;
 
   ExtractListOfInterfaceRegionGIDs(revisedRowMap, interfaceRegionLIDs, interfaceRegionGIDs);
@@ -177,7 +177,7 @@ void createRegionMatrix(const Teuchos::ParameterList galeriList,
                      revisedRowMap, revisedColMap,
                      rowImport, quasiRegionGrpMats, regionMats);
 
-} // createRegionMatrix
+}  // createRegionMatrix
 
 // Helper function that creates almost all the data needed to generate a unit-test
 // numDofsPerNode [in]: number of degrees of freedom per grid point
@@ -203,27 +203,27 @@ void createProblem(const LocalOrdinal numDofsPerNode,
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
 
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
   std::string mapType, coordinatesType;
-  if((matrixType == "Laplace2D") || (matrixType == "Elasticity2D")) {
-    mapType = "Cartesian2D";
+  if ((matrixType == "Laplace2D") || (matrixType == "Elasticity2D")) {
+    mapType         = "Cartesian2D";
     coordinatesType = "2D";
-  } else if((matrixType == "Laplace3D") || (matrixType == "Elasticity3D")) {
-    mapType = "Cartesian3D";
+  } else if ((matrixType == "Laplace3D") || (matrixType == "Elasticity3D")) {
+    mapType         = "Cartesian3D";
     coordinatesType = "3D";
   }
 
   // Build maps for the problem
   RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              mapType, comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap  = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   A = Pr->BuildMatrix();
@@ -232,14 +232,14 @@ void createProblem(const LocalOrdinal numDofsPerNode,
   // Create auxiliary data for MG
   RCP<MultiVector> nullspace = Pr->BuildNullspace();
   RCP<RealValuedMultiVector> coordinates =
-    Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,RealValuedMultiVector>(coordinatesType, nodeMap, galeriList);
+      Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, RealValuedMultiVector>(coordinatesType, nodeMap, galeriList);
 
   // create the region maps, importer and operator from composite counter parts
-  RCP<const Map> rowMap = Teuchos::null;
-  RCP<const Map> colMap = Teuchos::null;
+  RCP<const Map> rowMap        = Teuchos::null;
+  RCP<const Map> colMap        = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  RCP<Import> colImport = Teuchos::null;
-  Array<GO>  quasiRegionCoordGIDs;
+  RCP<Import> colImport        = Teuchos::null;
+  Array<GO> quasiRegionCoordGIDs;
   LO numLocalRegionNodes = 0;
   createRegionMatrix(galeriList, numDofsPerNode, nodeMap, dofMap, A,
                      rowMap, colMap, revisedRowMap, revisedColMap,
@@ -248,33 +248,33 @@ void createProblem(const LocalOrdinal numDofsPerNode,
                      rNodesPerDim, numLocalRegionNodes);
 
   // Build objects needed to construct the region coordinates
-  RCP<Map> quasiRegCoordMap = Teuchos::null;
+  RCP<Map> quasiRegCoordMap  = Teuchos::null;
   RCP<const Map> regCoordMap = Teuchos::null;
-  RCP<Import> coordImporter = Teuchos::null;
-
-  quasiRegCoordMap = Xpetra::MapFactory<LO,GO,Node>::
-    Build(nodeMap->lib(),
-          Teuchos::OrdinalTraits<GO>::invalid(),
-          quasiRegionCoordGIDs(),
-          nodeMap->getIndexBase(),
-          nodeMap->getComm());
-  regCoordMap = Xpetra::MapFactory<LO,GO,Node>::
-    Build(nodeMap->lib(),
-          Teuchos::OrdinalTraits<GO>::invalid(),
-          numLocalRegionNodes,
-          nodeMap->getIndexBase(),
-          nodeMap->getComm());
+  RCP<Import> coordImporter  = Teuchos::null;
+
+  quasiRegCoordMap = Xpetra::MapFactory<LO, GO, Node>::
+      Build(nodeMap->lib(),
+            Teuchos::OrdinalTraits<GO>::invalid(),
+            quasiRegionCoordGIDs(),
+            nodeMap->getIndexBase(),
+            nodeMap->getComm());
+  regCoordMap = Xpetra::MapFactory<LO, GO, Node>::
+      Build(nodeMap->lib(),
+            Teuchos::OrdinalTraits<GO>::invalid(),
+            numLocalRegionNodes,
+            nodeMap->getIndexBase(),
+            nodeMap->getComm());
 
   coordImporter = ImportFactory::Build(nodeMap, quasiRegCoordMap);
 
   // create region coordinates vector
-  regionCoordinates = Xpetra::MultiVectorFactory<real_type,LO,GO,NO>::Build(quasiRegCoordMap,
-      coordinates->getNumVectors());
+  regionCoordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(quasiRegCoordMap,
+                                                                               coordinates->getNumVectors());
   regionCoordinates->doImport(*coordinates, *coordImporter, Xpetra::INSERT);
   regionCoordinates->replaceMap(regCoordMap);
 
   // Create regional nullspace and coordinates
-  RCP<MultiVector> quasiRegionNullspace = Teuchos::null;
+  RCP<MultiVector> quasiRegionNullspace             = Teuchos::null;
   RCP<RealValuedMultiVector> quasiRegionCoordinates = Teuchos::null;
 
   compositeToRegional(nullspace, quasiRegionNullspace, regionNullspace,
@@ -282,70 +282,67 @@ void createProblem(const LocalOrdinal numDofsPerNode,
   compositeToRegional(coordinates, quasiRegionCoordinates, regionCoordinates,
                       regCoordMap, coordImporter);
 
-} // createProblem
+}  // createProblem
 
 // This test aims at checking that RegionRFactory has a working constructor.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactCtor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactCtor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   out << "version: " << MueLu::Version() << std::endl;
 
   RCP<RegionRFactory> myRFact = rcp(new RegionRFactory);
   TEST_EQUALITY(myRFact != Teuchos::null, true);
 
-} // RegionRFactCtor
+}  // RegionRFactCtor
 
 // This test aims at checking that RegionRFactory has a working constructor.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactCtorKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactCtorKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   out << "version: " << MueLu::Version() << std::endl;
 
   RCP<RegionRFactory_kokkos> myRFact = rcp(new RegionRFactory_kokkos);
   TEST_EQUALITY(myRFact != Teuchos::null, true);
 
-} // RegionRFactCtor
+}  // RegionRFactCtor
 
 // This test aims at checking that RegionRFactory produces a reasonable transfer
 // operator on a simple Laplace 3D problem.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  const int numRanks = comm->getSize();
-  const int myRank   = comm->getRank();
+  const int numRanks                  = comm->getSize();
+  const int myRank                    = comm->getRank();
 
   const int numDimensions = 3;
   const LO numDofsPerNode = 1;
   GO nx = 7, ny = 7, nz = 4;
   Teuchos::Array<LO> lNodesPerDim({static_cast<LO>(nx), static_cast<LO>(ny), static_cast<LO>(nz)});
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");
-  RCP<Matrix> regionMats = Teuchos::null;
-  RCP<MultiVector> regionNullspace = Teuchos::null;
+  RCP<Matrix> regionMats                       = Teuchos::null;
+  RCP<MultiVector> regionNullspace             = Teuchos::null;
   RCP<RealValuedMultiVector> regionCoordinates = Teuchos::null;
-  RCP<Matrix> A = Teuchos::null;
-  RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Matrix> A                                = Teuchos::null;
+  RCP<const Map> revisedRowMap                 = Teuchos::null;
+  RCP<Import> rowImport                        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter = Teuchos::null;
   Teuchos::Array<LO> rNodesPerDim(3);
@@ -365,12 +362,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3D, Scalar,
   fineLevel.Request("A");
   fineLevel.Set("A", regionMats);
   fineLevel.Set("numDimensions", numDimensions);
-  fineLevel.Set("lNodesPerDim",  rNodesPerDim);
+  fineLevel.Set("lNodesPerDim", rNodesPerDim);
   fineLevel.Set("Nullspace", regionNullspace);
   fineLevel.Set("Coordinates", regionCoordinates);
 
   // Construct the region R factory
-  RCP<RegionRFactory> myRFact = rcp(new RegionRFactory);
+  RCP<RegionRFactory> myRFact                = rcp(new RegionRFactory);
   RCP<const Teuchos::ParameterList> myParams = myRFact->GetValidParameterList();
 
   // Set requests on coarse level to access
@@ -409,108 +406,107 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3D, Scalar,
 
   // R->describe(out, Teuchos::VERB_EXTREME);
 
-  if(numRanks == 1) {
-    TEST_EQUALITY(R->getGlobalNumRows(),               18);
-    TEST_EQUALITY(R->getGlobalNumCols(),              196);
-    TEST_EQUALITY(R->getLocalNumRows(),                 18);
+  if (numRanks == 1) {
+    TEST_EQUALITY(R->getGlobalNumRows(), 18);
+    TEST_EQUALITY(R->getGlobalNumCols(), 196);
+    TEST_EQUALITY(R->getLocalNumRows(), 18);
     TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 196);
-    TEST_EQUALITY(R->getLocalNumEntries(),             726);
+    TEST_EQUALITY(R->getLocalNumEntries(), 726);
 
     Array<LO> rowLength = {{27, 45, 27, 45, 75, 45, 27, 45, 27,
                             27, 45, 27, 45, 75, 45, 27, 45, 27}};
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+    for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
       R->getLocalRowView(rowIdx, rowEntries, rowValues);
       TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
     }
 
-  } else { // Running with 4 ranks
-    TEST_EQUALITY(R->getGlobalNumRows(),               32);
-    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-    TEST_EQUALITY(R->getLocalNumEntries(),             216);
+  } else {  // Running with 4 ranks
+    TEST_EQUALITY(R->getGlobalNumRows(), 32);
+    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+    TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    if (myRank == 0) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 1) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 1) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 2) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 2) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 3) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 3) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
     }
   }
 
-} // RegionRFactLaplace3D
+}  // RegionRFactLaplace3D
 
 // This test aims at checking that RegionRFactory produces a reasonable transfer
 // operator on a simple Elasticity 3D problem.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  const int numRanks = comm->getSize();
-  const int myRank   = comm->getRank();
+  const int numRanks                  = comm->getSize();
+  const int myRank                    = comm->getRank();
 
   const int numDimensions = 3;
   const LO numDofsPerNode = 3;
   GO nx = 7, ny = 7, nz = 4;
   Teuchos::Array<LO> lNodesPerDim({static_cast<LO>(nx), static_cast<LO>(ny), static_cast<LO>(nz)});
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Elasticity3D");
-  RCP<Matrix> regionMats = Teuchos::null;
-  RCP<MultiVector> regionNullspace = Teuchos::null;
+  RCP<Matrix> regionMats                       = Teuchos::null;
+  RCP<MultiVector> regionNullspace             = Teuchos::null;
   RCP<RealValuedMultiVector> regionCoordinates = Teuchos::null;
-  RCP<Matrix> A = Teuchos::null;
-  RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Matrix> A                                = Teuchos::null;
+  RCP<const Map> revisedRowMap                 = Teuchos::null;
+  RCP<Import> rowImport                        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   Teuchos::Array<LO> rNodesPerDim(4);
@@ -530,12 +526,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3D, Scala
   fineLevel.Request("A");
   fineLevel.Set("A", regionMats);
   fineLevel.Set("numDimensions", numDimensions);
-  fineLevel.Set("lNodesPerDim",  rNodesPerDim);
+  fineLevel.Set("lNodesPerDim", rNodesPerDim);
   fineLevel.Set("Nullspace", regionNullspace);
   fineLevel.Set("Coordinates", regionCoordinates);
 
   // Construct the region R factory
-  RCP<RegionRFactory> myRFact = rcp(new RegionRFactory);
+  RCP<RegionRFactory> myRFact                = rcp(new RegionRFactory);
   RCP<const Teuchos::ParameterList> myParams = myRFact->GetValidParameterList();
 
   // Set requests on coarse level to access
@@ -574,12 +570,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3D, Scala
 
   // R->describe(out, Teuchos::VERB_EXTREME);
 
-  if(numRanks == 1) {
-    TEST_EQUALITY(R->getGlobalNumRows(),                 54);
-    TEST_EQUALITY(R->getGlobalNumCols(),                588);
-    TEST_EQUALITY(R->getLocalNumRows(),                   54);
-    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),   588);
-    TEST_EQUALITY(R->getLocalNumEntries(),              2178);
+  if (numRanks == 1) {
+    TEST_EQUALITY(R->getGlobalNumRows(), 54);
+    TEST_EQUALITY(R->getGlobalNumCols(), 588);
+    TEST_EQUALITY(R->getLocalNumRows(), 54);
+    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 588);
+    TEST_EQUALITY(R->getLocalNumEntries(), 2178);
 
     Array<LO> rowLength = {{27, 27, 27, 45, 45, 45, 27, 27, 27, 45, 45, 45,
                             75, 75, 75, 45, 45, 45, 27, 27, 27, 45, 45, 45,
@@ -588,105 +584,104 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3D, Scala
                             45, 45, 45, 27, 27, 27}};
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+    for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
       R->getLocalRowView(rowIdx, rowEntries, rowValues);
       TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
     }
 
-  } else { // Running with 4 ranks
-    TEST_EQUALITY(R->getGlobalNumRows(),               96);
+  } else {  // Running with 4 ranks
+    TEST_EQUALITY(R->getGlobalNumRows(), 96);
     TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-    TEST_EQUALITY(R->getLocalNumEntries(),             648);
+    TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    if (myRank == 0) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 1) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 1) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 2) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 2) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 3) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 3) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
     }
   }
 
-} // RegionRFactElasticity3D
+}  // RegionRFactElasticity3D
 
 // This test aims at checking that RegionRFactory produces a reasonable transfer
 // operator on a simple Laplace 3D problem.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3DKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3DKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  const int numRanks = comm->getSize();
-  const int myRank   = comm->getRank();
+  const int numRanks                  = comm->getSize();
+  const int myRank                    = comm->getRank();
 
   const int numDimensions = 3;
   const LO numDofsPerNode = 1;
   GO nx = 7, ny = 7, nz = 4;
   Teuchos::Array<LO> lNodesPerDim({static_cast<LO>(nx), static_cast<LO>(ny), static_cast<LO>(nz)});
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace3D");
-  RCP<Matrix> regionMats = Teuchos::null;
-  RCP<MultiVector> regionNullspace = Teuchos::null;
+  RCP<Matrix> regionMats                       = Teuchos::null;
+  RCP<MultiVector> regionNullspace             = Teuchos::null;
   RCP<RealValuedMultiVector> regionCoordinates = Teuchos::null;
-  RCP<Matrix> A = Teuchos::null;
-  RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Matrix> A                                = Teuchos::null;
+  RCP<const Map> revisedRowMap                 = Teuchos::null;
+  RCP<Import> rowImport                        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter = Teuchos::null;
   Teuchos::Array<LO> rNodesPerDim(3);
@@ -706,12 +701,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3DKokkos, Sc
   fineLevel.Request("A");
   fineLevel.Set("A", regionMats);
   fineLevel.Set("numDimensions", numDimensions);
-  fineLevel.Set("lNodesPerDim",  rNodesPerDim);
+  fineLevel.Set("lNodesPerDim", rNodesPerDim);
   fineLevel.Set("Nullspace", regionNullspace);
   fineLevel.Set("Coordinates", regionCoordinates);
 
   // Construct the region R factory
-  RCP<RegionRFactory_kokkos> myRFact = rcp(new RegionRFactory_kokkos);
+  RCP<RegionRFactory_kokkos> myRFact         = rcp(new RegionRFactory_kokkos);
   RCP<const Teuchos::ParameterList> myParams = myRFact->GetValidParameterList();
 
   // Set requests on coarse level to access
@@ -750,108 +745,107 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactLaplace3DKokkos, Sc
 
   // R->describe(out, Teuchos::VERB_EXTREME);
 
-  if(numRanks == 1) {
-    TEST_EQUALITY(R->getGlobalNumRows(),               18);
-    TEST_EQUALITY(R->getGlobalNumCols(),              196);
-    TEST_EQUALITY(R->getLocalNumRows(),                 18);
+  if (numRanks == 1) {
+    TEST_EQUALITY(R->getGlobalNumRows(), 18);
+    TEST_EQUALITY(R->getGlobalNumCols(), 196);
+    TEST_EQUALITY(R->getLocalNumRows(), 18);
     TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 196);
-    TEST_EQUALITY(R->getLocalNumEntries(),             726);
+    TEST_EQUALITY(R->getLocalNumEntries(), 726);
 
     Array<LO> rowLength = {{27, 45, 27, 45, 75, 45, 27, 45, 27,
                             27, 45, 27, 45, 75, 45, 27, 45, 27}};
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+    for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
       R->getLocalRowView(rowIdx, rowEntries, rowValues);
       TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
     }
 
-  } else { // Running with 4 ranks
-    TEST_EQUALITY(R->getGlobalNumRows(),               32);
-    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-    TEST_EQUALITY(R->getLocalNumEntries(),             216);
+  } else {  // Running with 4 ranks
+    TEST_EQUALITY(R->getGlobalNumRows(), 32);
+    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+    TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    if (myRank == 0) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 1) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 1) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 2) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 2) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 3) {
-      TEST_EQUALITY(R->getLocalNumRows(),                  8);
-      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),  64);
-      TEST_EQUALITY(R->getLocalNumEntries(),             216);
+    } else if (myRank == 3) {
+      TEST_EQUALITY(R->getLocalNumRows(), 8);
+      TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 64);
+      TEST_EQUALITY(R->getLocalNumEntries(), 216);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
     }
   }
 
-} // RegionRFactLaplace3DKokkos
+}  // RegionRFactLaplace3DKokkos
 
 // This test aims at checking that RegionRFactory produces a reasonable transfer
 // operator on a simple Elasticity 3D problem.
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3DKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3DKokkos, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  const int numRanks = comm->getSize();
-  const int myRank   = comm->getRank();
+  const int numRanks                  = comm->getSize();
+  const int myRank                    = comm->getRank();
 
   const int numDimensions = 3;
   const LO numDofsPerNode = 3;
   GO nx = 7, ny = 7, nz = 4;
   Teuchos::Array<LO> lNodesPerDim({static_cast<LO>(nx), static_cast<LO>(ny), static_cast<LO>(nz)});
-  Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
+  Teuchos::CommandLineProcessor& clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Elasticity3D");
-  RCP<Matrix> regionMats = Teuchos::null;
-  RCP<MultiVector> regionNullspace = Teuchos::null;
+  RCP<Matrix> regionMats                       = Teuchos::null;
+  RCP<MultiVector> regionNullspace             = Teuchos::null;
   RCP<RealValuedMultiVector> regionCoordinates = Teuchos::null;
-  RCP<Matrix> A = Teuchos::null;
-  RCP<const Map> revisedRowMap = Teuchos::null;
-  RCP<Import> rowImport = Teuchos::null;
+  RCP<Matrix> A                                = Teuchos::null;
+  RCP<const Map> revisedRowMap                 = Teuchos::null;
+  RCP<Import> rowImport                        = Teuchos::null;
   Teuchos::ArrayRCP<LocalOrdinal> regionMatVecLIDs;
   RCP<Import> regionInterfaceImporter;
   Teuchos::Array<LO> rNodesPerDim(4);
@@ -871,12 +865,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3DKokkos,
   fineLevel.Request("A");
   fineLevel.Set("A", regionMats);
   fineLevel.Set("numDimensions", numDimensions);
-  fineLevel.Set("lNodesPerDim",  rNodesPerDim);
+  fineLevel.Set("lNodesPerDim", rNodesPerDim);
   fineLevel.Set("Nullspace", regionNullspace);
   fineLevel.Set("Coordinates", regionCoordinates);
 
   // Construct the region R factory
-  RCP<RegionRFactory_kokkos> myRFact = rcp(new RegionRFactory_kokkos);
+  RCP<RegionRFactory_kokkos> myRFact         = rcp(new RegionRFactory_kokkos);
   RCP<const Teuchos::ParameterList> myParams = myRFact->GetValidParameterList();
 
   // Set requests on coarse level to access
@@ -915,12 +909,12 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3DKokkos,
 
   // R->describe(out, Teuchos::VERB_EXTREME);
 
-  if(numRanks == 1) {
-    TEST_EQUALITY(R->getGlobalNumRows(),                 54);
-    TEST_EQUALITY(R->getGlobalNumCols(),                588);
-    TEST_EQUALITY(R->getLocalNumRows(),                   54);
-    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(),   588);
-    TEST_EQUALITY(R->getLocalNumEntries(),              2178);
+  if (numRanks == 1) {
+    TEST_EQUALITY(R->getGlobalNumRows(), 54);
+    TEST_EQUALITY(R->getGlobalNumCols(), 588);
+    TEST_EQUALITY(R->getLocalNumRows(), 54);
+    TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 588);
+    TEST_EQUALITY(R->getLocalNumEntries(), 2178);
 
     Array<LO> rowLength = {{27, 27, 27, 45, 45, 45, 27, 27, 27, 45, 45, 45,
                             75, 75, 75, 45, 45, 45, 27, 27, 27, 45, 45, 45,
@@ -929,83 +923,79 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionRFactory, RegionRFactElasticity3DKokkos,
                             45, 45, 45, 27, 27, 27}};
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+    for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
       R->getLocalRowView(rowIdx, rowEntries, rowValues);
       TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
     }
 
-  } else { // Running with 4 ranks
-    TEST_EQUALITY(R->getGlobalNumRows(),               96);
+  } else {  // Running with 4 ranks
+    TEST_EQUALITY(R->getGlobalNumRows(), 96);
     TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-    TEST_EQUALITY(R->getLocalNumEntries(),             648);
+    TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
     ArrayView<const LO> rowEntries;
     ArrayView<const SC> rowValues;
-    if(myRank == 0) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    if (myRank == 0) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 1) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 1) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 2) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 2) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
-    } else if(myRank == 3) {
-      TEST_EQUALITY(R->getLocalNumRows(),                 24);
+    } else if (myRank == 3) {
+      TEST_EQUALITY(R->getLocalNumRows(), 24);
       TEST_EQUALITY(R->getCrsGraph()->getLocalNumCols(), 192);
-      TEST_EQUALITY(R->getLocalNumEntries(),             648);
+      TEST_EQUALITY(R->getLocalNumEntries(), 648);
 
       Array<LO> rowLength = {{27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27,
                               27, 27, 27, 27, 27, 27, 27, 27}};
-      for(int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
+      for (int rowIdx = 0; rowIdx < static_cast<int>(R->getLocalNumRows()); ++rowIdx) {
         R->getLocalRowView(rowIdx, rowEntries, rowValues);
         TEST_EQUALITY(static_cast<LO>(rowEntries.size()), rowLength[rowIdx]);
       }
     }
   }
 
-} // RegionRFactElasticity3DKokkos
+}  // RegionRFactElasticity3DKokkos
 
-
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactCtor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactCtorKokkos,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactLaplace3D,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactLaplace3DKokkos,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactElasticity3D,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory,RegionRFactElasticity3DKokkos,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactCtor, Scalar, LO, GO, Node)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactCtorKokkos, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactLaplace3D, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactLaplace3DKokkos, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactElasticity3D, Scalar, LO, GO, Node)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionRFactory, RegionRFactElasticity3DKokkos, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RegionVector.cpp b/packages/muelu/test/unit_tests/RegionVector.cpp
index f5631c6535cf..df7ec3c79105 100644
--- a/packages/muelu/test/unit_tests/RegionVector.cpp
+++ b/packages/muelu/test/unit_tests/RegionVector.cpp
@@ -56,41 +56,40 @@
 
 namespace MueLuTests {
 
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include "MueLu_UseShortNames.hpp"
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
   MUELU_TESTING_SET_OSTREAM;
-  MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
   using TST                   = Teuchos::ScalarTraits<SC>;
   using magnitude_type        = typename TST::magnitudeType;
   using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
   using real_type             = typename TST::coordinateType;
-  using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
   using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
   out << "version: " << MueLu::Version() << std::endl;
 
   // Get MPI parameter
   RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-  LO numRanks = comm->getSize();
-  LO myRank = comm->getRank();
+  LO numRanks                         = comm->getSize();
+  LO myRank                           = comm->getRank();
 
   GO nx = 5, ny = 5, nz = 1;
   Teuchos::CommandLineProcessor &clp = Teuchos::UnitTestRepository::getCLP();
   Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");
   Teuchos::ParameterList galeriList = galeriParameters.GetParameterList();
-  std::string   matrixType = galeriParameters.GetMatrixType();
+  std::string matrixType            = galeriParameters.GetMatrixType();
 
   // Build maps for the problem
   const LO numDofsPerNode = 1;
-  RCP<Map> nodeMap = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
+  RCP<Map> nodeMap        = Galeri::Xpetra::CreateMap<LO, GO, Node>(TestHelpers::Parameters::getLib(),
                                                              "Cartesian2D", comm, galeriList);
-  RCP<Map> dofMap  = Xpetra::MapFactory<LO,GO,Node>::Build(nodeMap, numDofsPerNode);
+  RCP<Map> dofMap         = Xpetra::MapFactory<LO, GO, Node>::Build(nodeMap, numDofsPerNode);
 
   // Build the Xpetra problem
-  RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-    Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), dofMap, galeriList);
 
   // Generate the operator
   RCP<Matrix> A = Pr->BuildMatrix();
@@ -106,22 +105,22 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, L
   lNodesPerDir[0] = galeriList.get<LO>("lnx");
   lNodesPerDir[1] = galeriList.get<LO>("lny");
   lNodesPerDir[2] = 1;
-  procsPerDim[0] = galeriList.get<GO>("mx");
-  procsPerDim[1] = galeriList.get<GO>("my");
-  procsPerDim[2] = 1;
+  procsPerDim[0]  = galeriList.get<GO>("mx");
+  procsPerDim[1]  = galeriList.get<GO>("my");
+  procsPerDim[2]  = 1;
 
   Array<int> boundaryConditions;
-  int maxRegPerGID = 0;
-  int numInterfaces = 0;
+  int maxRegPerGID       = 0;
+  int numInterfaces      = 0;
   LO numLocalRegionNodes = 0;
-  Array<GO>  sendGIDs;
+  Array<GO> sendGIDs;
   Array<int> sendPIDs;
-  Array<LO>  rNodesPerDim(3);
-  Array<LO>  compositeToRegionLIDs(nodeMap->getLocalNumElements()*numDofsPerNode);
-  Array<GO>  quasiRegionGIDs;
-  Array<GO>  quasiRegionCoordGIDs;
-  Array<GO>  interfaceGIDs;
-  Array<LO>  interfaceLIDsData;
+  Array<LO> rNodesPerDim(3);
+  Array<LO> compositeToRegionLIDs(nodeMap->getLocalNumElements() * numDofsPerNode);
+  Array<GO> quasiRegionGIDs;
+  Array<GO> quasiRegionCoordGIDs;
+  Array<GO> interfaceGIDs;
+  Array<LO> interfaceLIDsData;
   createRegionData(2, false, numDofsPerNode,
                    gNodesPerDir(), lNodesPerDir(), procsPerDim(),
                    nodeMap, dofMap,
@@ -132,115 +131,113 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, L
 
   RCP<const Map> rowMap = Teuchos::null;
   RCP<const Map> colMap = Teuchos::null;
-  rowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs(),
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
-  colMap = rowMap;
+  rowMap                = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                                  Teuchos::OrdinalTraits<GO>::invalid(),
+                                                                  quasiRegionGIDs(),
+                                                                  A->getRowMap()->getIndexBase(),
+                                                                  A->getRowMap()->getComm());
+  colMap                = rowMap;
 
   RCP<const Map> revisedRowMap = Teuchos::null;
   RCP<const Map> revisedColMap = Teuchos::null;
-  revisedRowMap = Xpetra::MapFactory<LO,GO,Node>::Build(A->getRowMap()->lib(),
-      Teuchos::OrdinalTraits<GO>::invalid(),
-      quasiRegionGIDs.size()*numDofsPerNode,
-      A->getRowMap()->getIndexBase(),
-      A->getRowMap()->getComm());
-  revisedColMap = revisedRowMap;
+  revisedRowMap                = Xpetra::MapFactory<LO, GO, Node>::Build(A->getRowMap()->lib(),
+                                                                         Teuchos::OrdinalTraits<GO>::invalid(),
+                                                                         quasiRegionGIDs.size() * numDofsPerNode,
+                                                                         A->getRowMap()->getIndexBase(),
+                                                                         A->getRowMap()->getComm());
+  revisedColMap                = revisedRowMap;
 
   RCP<Import> rowImport = ImportFactory::Build(dofMap, rowMap);
   // RCP<Import> colImport = ImportFactory::Build(dofMap, colMap);
 
-/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-  { // test compositeToRegional
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+  {  // test compositeToRegional
     // Create a composite vector
     RCP<Vector> compVec = VectorFactory::Build(dofMap);
     TEUCHOS_ASSERT(!compVec.is_null());
     compVec->putScalar(0.0);
     const size_t localLength = compVec->getLocalLength();
     for (size_t k = 0; k < localLength; ++k) {
-      compVec->replaceLocalValue(k, static_cast<Scalar>(k) + static_cast<magnitude_type>(myRank)/10);
+      compVec->replaceLocalValue(k, static_cast<Scalar>(k) + static_cast<magnitude_type>(myRank) / 10);
     }
 
     // Create a region vector
-    RCP<Vector> regVec = Teuchos::null;
+    RCP<Vector> regVec      = Teuchos::null;
     RCP<Vector> quasiRegVec = Teuchos::null;
 
     compositeToRegional(compVec, quasiRegVec, regVec, revisedRowMap, rowImport);
 
-    if(numRanks == 1){
-      TEST_EQUALITY(regVec->getLocalLength(),  25);
+    if (numRanks == 1) {
+      TEST_EQUALITY(regVec->getLocalLength(), 25);
       TEST_EQUALITY(regVec->getGlobalLength(), 25);
-      TEST_EQUALITY(quasiRegVec->getLocalLength(),  25);
+      TEST_EQUALITY(quasiRegVec->getLocalLength(), 25);
       TEST_EQUALITY(quasiRegVec->getGlobalLength(), 25);
       ArrayRCP<SC> refValues;
       Teuchos::ArrayRCP<const SC> myValues;
       refValues.deepCopy(ArrayView<const SC>({0.0, 1, 2, 3, 4,
-                                               5, 6, 7, 8, 9,
+                                              5, 6, 7, 8, 9,
                                               10, 11, 12, 13, 14,
                                               15, 16, 17, 18, 19,
                                               20, 21, 22, 23, 24}));
 
       myValues = regVec->getData(0);
-      for(size_t idx = 0; idx < regVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < regVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
       myValues = quasiRegVec->getData(0);
-      for(size_t idx = 0; idx < quasiRegVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < quasiRegVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
 
-    } else if(numRanks == 4) {
+    } else if (numRanks == 4) {
       // All ranks will have the same number of rows/cols/entries
-      TEST_EQUALITY(regVec->getLocalLength(),  9);
+      TEST_EQUALITY(regVec->getLocalLength(), 9);
       TEST_EQUALITY(regVec->getGlobalLength(), 36);
-      TEST_EQUALITY(quasiRegVec->getLocalLength(),  9);
+      TEST_EQUALITY(quasiRegVec->getLocalLength(), 9);
       TEST_EQUALITY(quasiRegVec->getGlobalLength(), 36);
 
       ArrayRCP<SC> refValues;
       Teuchos::ArrayRCP<const SC> myValues;
-      if(myRank == 0) {
+      if (myRank == 0) {
         refValues.deepCopy(ArrayView<const SC>({0.0, 1, 2, 3, 4, 5, 6, 7, 8}));
 
-      } else if(myRank == 1) {
+      } else if (myRank == 1) {
         refValues.deepCopy(ArrayView<const SC>({2.0, 0.1, 1.1, 5.0, 2.1, 3.1, 8.0, 4.1, 5.1}));
 
-      } else if(myRank == 2) {
+      } else if (myRank == 2) {
         refValues.deepCopy(ArrayView<const SC>({6, 7, 8, 0.2, 1.2, 2.2, 3.2, 4.2, 5.2}));
 
-      } else if(myRank == 3) {
+      } else if (myRank == 3) {
         refValues.deepCopy(ArrayView<const SC>({8, 4.1, 5.1, 2.2, 0.3, 1.3, 5.2, 2.3, 3.3}));
       }
       // Loop over region matrix data and compare it to ref data
       myValues = regVec->getData(0);
-      for(size_t idx = 0; idx < regVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < regVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
       myValues = quasiRegVec->getData(0);
-      for(size_t idx = 0; idx < quasiRegVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < quasiRegVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
     }
-
   }
-/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
   {
     // Create a region vector
     RCP<Vector> regVec = Teuchos::null;
-    regVec = VectorFactory::Build(revisedRowMap);
-    regVec->putScalar( myRank / 10.0 );
-    for (size_t k = 0; k < regVec->getLocalLength(); ++k){
+    regVec             = VectorFactory::Build(revisedRowMap);
+    regVec->putScalar(myRank / 10.0);
+    for (size_t k = 0; k < regVec->getLocalLength(); ++k) {
       regVec->sumIntoLocalValue(k, k);
     }
 
-
     // Create a composite vector
     RCP<Vector> compVec = VectorFactory::Build(dofMap, true);
 
     regionalToComposite(regVec, compVec, rowImport);
 
-    if(numRanks == 1) {
-      TEST_EQUALITY(compVec->getLocalLength(),  25);
+    if (numRanks == 1) {
+      TEST_EQUALITY(compVec->getLocalLength(), 25);
       TEST_EQUALITY(compVec->getGlobalLength(), 25);
 
       ArrayRCP<SC> refValues;
@@ -251,57 +248,56 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, L
                                               15.0, 16.0, 17.0, 18.0, 19.0,
                                               20.0, 21.0, 22.0, 23.0, 24.0}));
       myValues = compVec->getData(0);
-      for(size_t idx = 0; idx < compVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < compVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
 
-    } else if(numRanks == 4) {
+    } else if (numRanks == 4) {
       // All ranks will have the same global length of the vector, but local length differs.
       // Due to region numbering and region interface duplications,
       // lower rank IDs have more local entries then higher rank IDs.
-      if( myRank == 0){
-        TEST_EQUALITY(compVec->getLocalLength(),  9);
+      if (myRank == 0) {
+        TEST_EQUALITY(compVec->getLocalLength(), 9);
         TEST_EQUALITY(compVec->getGlobalLength(), 25);
       } else if (myRank == 1) {
-        TEST_EQUALITY(compVec->getLocalLength(),  6);
+        TEST_EQUALITY(compVec->getLocalLength(), 6);
         TEST_EQUALITY(compVec->getGlobalLength(), 25);
       } else if (myRank == 2) {
-        TEST_EQUALITY(compVec->getLocalLength(),  6);
+        TEST_EQUALITY(compVec->getLocalLength(), 6);
         TEST_EQUALITY(compVec->getGlobalLength(), 25);
       } else if (myRank == 3) {
-        TEST_EQUALITY(compVec->getLocalLength(),  4);
+        TEST_EQUALITY(compVec->getLocalLength(), 4);
         TEST_EQUALITY(compVec->getGlobalLength(), 25);
       }
 
       ArrayRCP<SC> refValues;
       Teuchos::ArrayRCP<const SC> myValues;
-      if(myRank == 0) {
+      if (myRank == 0) {
         refValues.deepCopy(ArrayView<const SC>({0.0, 1.0, 2.1, 3.0, 4.0, 8.1, 6.2, 8.2, 16.6}));
 
-      } else if(myRank == 1) {
+      } else if (myRank == 1) {
         refValues.deepCopy(ArrayView<const SC>({1.1, 2.1, 4.1, 5.1, 8.4, 10.4}));
 
-      } else if(myRank == 2) {
+      } else if (myRank == 2) {
         refValues.deepCopy(ArrayView<const SC>({3.2, 4.2, 8.5, 6.2, 7.2, 14.5}));
 
-      } else if(myRank == 3) {
+      } else if (myRank == 3) {
         refValues.deepCopy(ArrayView<const SC>({4.3, 5.3, 7.3, 8.3}));
       }
       // Loop over region vector data and compare it to reference data
       myValues = compVec->getData(0);
-      for(size_t idx = 0; idx < compVec->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < compVec->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myValues[idx], refValues[idx], 100 * TMT::eps());
       }
     }
   }
-/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
   {
     // initialize region vector with all ones.
     RCP<Vector> interfaceScaling = Teuchos::null;
-    interfaceScaling = VectorFactory::Build(revisedRowMap);
+    interfaceScaling             = VectorFactory::Build(revisedRowMap);
     interfaceScaling->putScalar(1.0);
 
-
     // transform to composite layout while adding interface values via the Export() combine mode
     RCP<Vector> compInterfaceScalingSum = VectorFactory::Build(dofMap, true);
     regionalToComposite(interfaceScaling, compInterfaceScalingSum, rowImport);
@@ -309,59 +305,55 @@ TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RegionVector, RegionCompositeVector, Scalar, L
     /* transform composite layout back to regional layout. Now, GIDs associated
      * with region interface should carry a scaling factor (!= 1).
      */
-    RCP<Vector> regVec = Teuchos::null;
+    RCP<Vector> regVec                   = Teuchos::null;
     RCP<Vector> quasiRegInterfaceScaling = Teuchos::null;
     compositeToRegional(compInterfaceScalingSum, quasiRegInterfaceScaling,
                         interfaceScaling, revisedRowMap, rowImport);
 
-    if(numRanks == 1) {
-      TEST_EQUALITY(interfaceScaling->getLocalLength(),    25);
-      TEST_EQUALITY(interfaceScaling->getGlobalLength(),   25);
+    if (numRanks == 1) {
+      TEST_EQUALITY(interfaceScaling->getLocalLength(), 25);
+      TEST_EQUALITY(interfaceScaling->getGlobalLength(), 25);
 
       // No scaling on one rank, so all values are 1.0
       Teuchos::ArrayRCP<const SC> myScaling;
       myScaling = interfaceScaling->getData(0);
-      for(size_t idx = 0; idx < interfaceScaling->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myScaling[idx], TST::one(), 100*TMT::eps());
+      for (size_t idx = 0; idx < interfaceScaling->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myScaling[idx], TST::one(), 100 * TMT::eps());
       }
 
-    } else if(numRanks == 4) {
+    } else if (numRanks == 4) {
       // All ranks will have the same number of rows/cols/entries
-      TEST_EQUALITY(interfaceScaling->getLocalLength(),    9);
-      TEST_EQUALITY(interfaceScaling->getGlobalLength(),    36);
+      TEST_EQUALITY(interfaceScaling->getLocalLength(), 9);
+      TEST_EQUALITY(interfaceScaling->getGlobalLength(), 36);
 
       ArrayRCP<SC> refValues;
       Teuchos::ArrayRCP<const SC> myScaling;
-      if(myRank == 0) {
+      if (myRank == 0) {
         refValues.deepCopy(ArrayView<const SC>({1, 1, 2, 1, 1, 2, 2, 2, 4}));
 
-      } else if(myRank == 1) {
+      } else if (myRank == 1) {
         refValues.deepCopy(ArrayView<const SC>({2, 1, 1, 2, 1, 1, 4, 2, 2}));
 
-      } else if(myRank == 2) {
+      } else if (myRank == 2) {
         refValues.deepCopy(ArrayView<const SC>({2, 2, 4, 1, 1, 2, 1, 1, 2}));
 
-      } else if(myRank == 3) {
+      } else if (myRank == 3) {
         refValues.deepCopy(ArrayView<const SC>({4, 2, 2, 2, 1, 1, 2, 1, 1}));
       }
       // Loop over region vector data and compare it to reference data
       myScaling = interfaceScaling->getData(0);
-      for(size_t idx = 0; idx < interfaceScaling->getLocalLength(); ++idx) {
-        TEST_FLOATING_EQUALITY(myScaling[idx], refValues[idx], 100*TMT::eps());
+      for (size_t idx = 0; idx < interfaceScaling->getLocalLength(); ++idx) {
+        TEST_FLOATING_EQUALITY(myScaling[idx], refValues[idx], 100 * TMT::eps());
       }
-
     }
-
   }
-/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
+}  // RegionCompositeVector
 
-} // RegionCompositeVector
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionVector,RegionCompositeVector,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RegionVector, RegionCompositeVector, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ReitzingerPFactory.cpp b/packages/muelu/test/unit_tests/ReitzingerPFactory.cpp
index 56272d1afca2..a07e7590fb83 100644
--- a/packages/muelu/test/unit_tests/ReitzingerPFactory.cpp
+++ b/packages/muelu/test/unit_tests/ReitzingerPFactory.cpp
@@ -59,33 +59,26 @@
 #include "MueLu_ReitzingerPFactory.hpp"
 #include "MueLu_Utilities.hpp"
 
-
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<ReitzingerPFactory> PFact = rcp(new ReitzingerPFactory);
-    TEST_EQUALITY(PFact != Teuchos::null, true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ReitzingerPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-  } //Constructor
+  out << "version: " << MueLu::Version() << std::endl;
 
+  RCP<ReitzingerPFactory> PFact = rcp(new ReitzingerPFactory);
+  TEST_EQUALITY(PFact != Teuchos::null, true);
 
-  // Note: Additional tests live in muelu/test/maxwell, we need to read matrices off disk
-  // to test their functionality properly
-
- 
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory,Constructor,Scalar,LO,GO,Node)
+}  // Constructor
 
+// Note: Additional tests live in muelu/test/maxwell, we need to read matrices off disk
+// to test their functionality properly
 
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ReitzingerPFactory, Constructor, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Repartition.cpp b/packages/muelu/test/unit_tests/Repartition.cpp
index 25b5d229fd70..5a475788ce61 100644
--- a/packages/muelu/test/unit_tests/Repartition.cpp
+++ b/packages/muelu/test/unit_tests/Repartition.cpp
@@ -67,1046 +67,1032 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-    TEST_EQUALITY(repart != Teuchos::null, true);
-
-  } // Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests build of the permutation matrix for repartitioning." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 5, ny = 3;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const GlobalOrdinal     indexBase     = 0;
-    size_t numMyElements = 0;
-    switch (myRank) {
-      case 0:  numMyElements = 6; break;
-      case 1:  numMyElements = 5; break;
-      case 2:  numMyElements = 0; break;
-      case 3:  numMyElements = 4; break;
-    }
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+  TEST_EQUALITY(repart != Teuchos::null, true);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests build of the permutation matrix for repartitioning." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 5, ny = 3;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const GlobalOrdinal indexBase         = 0;
+  size_t numMyElements                  = 0;
+  switch (myRank) {
+    case 0: numMyElements = 6; break;
+    case 1: numMyElements = 5; break;
+    case 2: numMyElements = 0; break;
+    case 3: numMyElements = 4; break;
+  }
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    RCP<MultiVector>                  coords        = MultiVectorFactory::Build(map, 2);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
-    if (decomposition->getLocalLength())
-      partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 4, part2 has 5
-    const int numPartitions = 3;
-    switch (myRank)  {
-      case 0:
-        partitionThisDofBelongsTo[0] = 0;
-        partitionThisDofBelongsTo[3] = 0;
-        partitionThisDofBelongsTo[1] = 1;
-        partitionThisDofBelongsTo[2] = 2;
-        partitionThisDofBelongsTo[4] = 2;
-        partitionThisDofBelongsTo[5] = 2;
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;
-        partitionThisDofBelongsTo[2] = 0;
-        partitionThisDofBelongsTo[1] = 1;
-        partitionThisDofBelongsTo[4] = 1;
-        partitionThisDofBelongsTo[3] = 2;
-        break;
-      case 2:
-        break;
-      case 3:
-        partitionThisDofBelongsTo[1] = 0;
-        partitionThisDofBelongsTo[2] = 0;
-        partitionThisDofBelongsTo[3] = 1;
-        partitionThisDofBelongsTo[0] = 2;
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    // This test uses a made up partitioning  that is given via Level to Repartition. It must be
-    // associated with an instance of the ZoltanInterface so that it can be found inside Repartition.
-    // Furthermore, that same instance must be supplied to MueLu::Repartition. Coordinates must be
-    // provided, even though they are not used, as the Zoltan interface checks for them.
-    Level level;
-    level.SetLevelID(1);
-
-    RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-    level.SetFactoryManager(factoryHandler);
-
-    RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-
-    level.Set<RCP<Matrix> >     ("A",                    A);
-    level.Set<RCP<MultiVector> >("Coordinates",          coords);
-    level.Set<int>              ("number of partitions", numPartitions);
-
-    level.Request("Partition", zoltan.get());
-    level.Set<RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > >("Partition", decomposition, zoltan.get());
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-    Teuchos::ParameterList paramList;
-    paramList.set("repartition: remap parts",       false);
-    repart->SetParameterList(paramList);
-    repart->SetFactory("Partition", zoltan);
-
-    // Build
-    level.Request("Importer", repart.get());
-    repart->Build(level);
-
-    // The local resulting vector should contain the partition number that this pid owns.
-    // In this case, pid 0 owns partition 0
-    //               pid 1 owns partition 1
-    //               pid 2 owns partition 2
-    //               pid 3 does not own a partition
-    RCP<const Import> importer;
-    level.Get("Importer", importer, repart.get());
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > result = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(importer->getTargetMap(), false);
-    result->doImport(*decomposition, *importer, Xpetra::INSERT);
-    Teuchos::ArrayRCP<GlobalOrdinal> resultData;
-    if (result->getLocalLength() > 0)
-      resultData = result->getDataNonConst(0);
-
-    // Local entries in the resulting vector should be equal to the partition number
-    int pidFailed = -1;
-    for (int i = 0; i < resultData.size(); i++) {
-      if (resultData[i] != myRank) {
-        pidFailed = myRank;
-        break;
-      }
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  RCP<MultiVector> coords                                                              = MultiVectorFactory::Build(map, 2);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
+  if (decomposition->getLocalLength())
+    partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 4, part2 has 5
+  const int numPartitions = 3;
+  switch (myRank) {
+    case 0:
+      partitionThisDofBelongsTo[0] = 0;
+      partitionThisDofBelongsTo[3] = 0;
+      partitionThisDofBelongsTo[1] = 1;
+      partitionThisDofBelongsTo[2] = 2;
+      partitionThisDofBelongsTo[4] = 2;
+      partitionThisDofBelongsTo[5] = 2;
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;
+      partitionThisDofBelongsTo[2] = 0;
+      partitionThisDofBelongsTo[1] = 1;
+      partitionThisDofBelongsTo[4] = 1;
+      partitionThisDofBelongsTo[3] = 2;
+      break;
+    case 2:
+      break;
+    case 3:
+      partitionThisDofBelongsTo[1] = 0;
+      partitionThisDofBelongsTo[2] = 0;
+      partitionThisDofBelongsTo[3] = 1;
+      partitionThisDofBelongsTo[0] = 2;
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  // This test uses a made up partitioning  that is given via Level to Repartition. It must be
+  // associated with an instance of the ZoltanInterface so that it can be found inside Repartition.
+  // Furthermore, that same instance must be supplied to MueLu::Repartition. Coordinates must be
+  // provided, even though they are not used, as the Zoltan interface checks for them.
+  Level level;
+  level.SetLevelID(1);
+
+  RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+  level.SetFactoryManager(factoryHandler);
+
+  RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
+
+  level.Set<RCP<Matrix> >("A", A);
+  level.Set<RCP<MultiVector> >("Coordinates", coords);
+  level.Set<int>("number of partitions", numPartitions);
+
+  level.Request("Partition", zoltan.get());
+  level.Set<RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >("Partition", decomposition, zoltan.get());
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+  Teuchos::ParameterList paramList;
+  paramList.set("repartition: remap parts", false);
+  repart->SetParameterList(paramList);
+  repart->SetFactory("Partition", zoltan);
+
+  // Build
+  level.Request("Importer", repart.get());
+  repart->Build(level);
+
+  // The local resulting vector should contain the partition number that this pid owns.
+  // In this case, pid 0 owns partition 0
+  //               pid 1 owns partition 1
+  //               pid 2 owns partition 2
+  //               pid 3 does not own a partition
+  RCP<const Import> importer;
+  level.Get("Importer", importer, repart.get());
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > result = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(importer->getTargetMap(), false);
+  result->doImport(*decomposition, *importer, Xpetra::INSERT);
+  Teuchos::ArrayRCP<GlobalOrdinal> resultData;
+  if (result->getLocalLength() > 0)
+    resultData = result->getDataNonConst(0);
+
+  // Local entries in the resulting vector should be equal to the partition number
+  int pidFailed = -1;
+  for (int i = 0; i < resultData.size(); i++) {
+    if (resultData[i] != myRank) {
+      pidFailed = myRank;
+      break;
     }
-    resultData = Teuchos::null;
-
-    // Test passes if all pid's return -1. If a pid hits an error, it sets
-    // thisPidFailed equal to comm->getRank().  In this way, you can tell with --details=ALL
-    // which pid failed.
-    int whichPidFailed;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, pidFailed, Teuchos::outArg(whichPidFailed));
-    TEST_EQUALITY(whichPidFailed, -1);
-
-  } // Build
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
-    out << "The matrix is distributed across 4 processors, and there are 4 partitions." << std::endl;
-    out << "Process 0 ends up owning a partition for which it originally has no unknowns." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 4, ny = 6;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", true); // keep Dirichlet rows
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const size_t numMyElements = 6;
-    const GlobalOrdinal     indexBase     = 0;
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  }
+  resultData = Teuchos::null;
+
+  // Test passes if all pid's return -1. If a pid hits an error, it sets
+  // thisPidFailed equal to comm->getRank().  In this way, you can tell with --details=ALL
+  // which pid failed.
+  int whichPidFailed;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_MAX, pidFailed, Teuchos::outArg(whichPidFailed));
+  TEST_EQUALITY(whichPidFailed, -1);
+
+}  // Build
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
+  out << "The matrix is distributed across 4 processors, and there are 4 partitions." << std::endl;
+  out << "Process 0 ends up owning a partition for which it originally has no unknowns." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 4, ny = 6;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", true);  // keep Dirichlet rows
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const size_t numMyElements            = 6;
+  const GlobalOrdinal indexBase         = 0;
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, part2 has 9, part3 has 6
-    const int numPartitions = 4;
-    switch (myRank)  {
-      case 0:                                       //  nnz by row    nnz by partition
-        partitionThisDofBelongsTo[2] = 0;           //      1
-        partitionThisDofBelongsTo[3] = 0;           //      1             3
-        partitionThisDofBelongsTo[4] = 0;           //      1
-        partitionThisDofBelongsTo[5] = 1;           //      5             5
-        partitionThisDofBelongsTo[0] = 2;           //      1
-        partitionThisDofBelongsTo[1] = 2;           //      1             2
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;           //      5
-        partitionThisDofBelongsTo[1] = 0;           //      1             6
-        partitionThisDofBelongsTo[2] = 1;           //      1
-        partitionThisDofBelongsTo[3] = 1;           //      5             6
-        partitionThisDofBelongsTo[4] = 2;           //      5
-        partitionThisDofBelongsTo[5] = 2;           //      1             6
-        break;
-      case 2:
-        partitionThisDofBelongsTo[1] = 0;           //      5             5
-        partitionThisDofBelongsTo[0] = 2;           //      1
-        partitionThisDofBelongsTo[2] = 2;           //      5             8
-        partitionThisDofBelongsTo[3] = 2;           //      1
-        partitionThisDofBelongsTo[4] = 2;           //      1
-        partitionThisDofBelongsTo[5] = 3;           //      5             5
-        break;
-      case 3:
-        partitionThisDofBelongsTo[0] = 1;           //      5
-        partitionThisDofBelongsTo[1] = 1;           //      1             7
-        partitionThisDofBelongsTo[2] = 1;           //      1
-        partitionThisDofBelongsTo[3] = 2;           //      1             1
-        partitionThisDofBelongsTo[4] = 3;           //      1
-        partitionThisDofBelongsTo[5] = 3;           //      1             2
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-
-    repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
-
-    Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
-    switch (myRank) {
-      case 0:
-        TEST_EQUALITY(decompEntries[2], 1);
-        TEST_EQUALITY(decompEntries[3], 1);
-        TEST_EQUALITY(decompEntries[4], 1);
-        TEST_EQUALITY(decompEntries[5], 3);
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        break;
-      case 1:
-        TEST_EQUALITY(decompEntries[0], 1);
-        TEST_EQUALITY(decompEntries[1], 1);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 3);
-        TEST_EQUALITY(decompEntries[4], 2);
-        TEST_EQUALITY(decompEntries[5], 2);
-        break;
-      case 2:
-        TEST_EQUALITY(decompEntries[1], 1);
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[2], 2);
-        TEST_EQUALITY(decompEntries[3], 2);
-        TEST_EQUALITY(decompEntries[4], 2);
-        TEST_EQUALITY(decompEntries[5], 0);
-        break;
-      case 3:
-        TEST_EQUALITY(decompEntries[0], 3);
-        TEST_EQUALITY(decompEntries[1], 3);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 2);
-        TEST_EQUALITY(decompEntries[4], 0);
-        TEST_EQUALITY(decompEntries[5], 0);
-        break;
-    }
-  } // DeterminePartitionPlacement1
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 4, ny = 6;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const size_t numMyElements = 6;
-    const GlobalOrdinal     indexBase     = 0;
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo                           = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, part2 has 9, part3 has 6
+  const int numPartitions = 4;
+  switch (myRank) {
+    case 0:                              //  nnz by row    nnz by partition
+      partitionThisDofBelongsTo[2] = 0;  //      1
+      partitionThisDofBelongsTo[3] = 0;  //      1             3
+      partitionThisDofBelongsTo[4] = 0;  //      1
+      partitionThisDofBelongsTo[5] = 1;  //      5             5
+      partitionThisDofBelongsTo[0] = 2;  //      1
+      partitionThisDofBelongsTo[1] = 2;  //      1             2
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;  //      5
+      partitionThisDofBelongsTo[1] = 0;  //      1             6
+      partitionThisDofBelongsTo[2] = 1;  //      1
+      partitionThisDofBelongsTo[3] = 1;  //      5             6
+      partitionThisDofBelongsTo[4] = 2;  //      5
+      partitionThisDofBelongsTo[5] = 2;  //      1             6
+      break;
+    case 2:
+      partitionThisDofBelongsTo[1] = 0;  //      5             5
+      partitionThisDofBelongsTo[0] = 2;  //      1
+      partitionThisDofBelongsTo[2] = 2;  //      5             8
+      partitionThisDofBelongsTo[3] = 2;  //      1
+      partitionThisDofBelongsTo[4] = 2;  //      1
+      partitionThisDofBelongsTo[5] = 3;  //      5             5
+      break;
+    case 3:
+      partitionThisDofBelongsTo[0] = 1;  //      5
+      partitionThisDofBelongsTo[1] = 1;  //      1             7
+      partitionThisDofBelongsTo[2] = 1;  //      1
+      partitionThisDofBelongsTo[3] = 2;  //      1             1
+      partitionThisDofBelongsTo[4] = 3;  //      1
+      partitionThisDofBelongsTo[5] = 3;  //      1             2
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+
+  repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
+
+  Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
+  switch (myRank) {
+    case 0:
+      TEST_EQUALITY(decompEntries[2], 1);
+      TEST_EQUALITY(decompEntries[3], 1);
+      TEST_EQUALITY(decompEntries[4], 1);
+      TEST_EQUALITY(decompEntries[5], 3);
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      break;
+    case 1:
+      TEST_EQUALITY(decompEntries[0], 1);
+      TEST_EQUALITY(decompEntries[1], 1);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 3);
+      TEST_EQUALITY(decompEntries[4], 2);
+      TEST_EQUALITY(decompEntries[5], 2);
+      break;
+    case 2:
+      TEST_EQUALITY(decompEntries[1], 1);
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[2], 2);
+      TEST_EQUALITY(decompEntries[3], 2);
+      TEST_EQUALITY(decompEntries[4], 2);
+      TEST_EQUALITY(decompEntries[5], 0);
+      break;
+    case 3:
+      TEST_EQUALITY(decompEntries[0], 3);
+      TEST_EQUALITY(decompEntries[1], 3);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 2);
+      TEST_EQUALITY(decompEntries[4], 0);
+      TEST_EQUALITY(decompEntries[5], 0);
+      break;
+  }
+}  // DeterminePartitionPlacement1
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 4, ny = 6;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const size_t numMyElements            = 6;
+  const GlobalOrdinal indexBase         = 0;
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, parth2 has 9, part3 has 6
-    const int numPartitions = 4;
-    switch (myRank)  {
-      case 0:                                       //  nnz by row    nnz by partition
-        partitionThisDofBelongsTo[0] = 0;           //       3
-        partitionThisDofBelongsTo[1] = 0;           //       4            7
-        partitionThisDofBelongsTo[2] = 1;           //       4
-        partitionThisDofBelongsTo[3] = 1;           //       3            7
-        partitionThisDofBelongsTo[4] = 3;           //       4
-        partitionThisDofBelongsTo[5] = 3;           //       5            9
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;           //       5
-        partitionThisDofBelongsTo[1] = 0;           //       4            9
-        partitionThisDofBelongsTo[2] = 1;           //       4
-        partitionThisDofBelongsTo[3] = 1;           //       5            9
-        partitionThisDofBelongsTo[4] = 2;           //       5
-        partitionThisDofBelongsTo[5] = 2;           //       4            9
-        break;
-      case 2:
-        partitionThisDofBelongsTo[0] = 0;           //       4
-        partitionThisDofBelongsTo[1] = 0;           //       5            9
-        partitionThisDofBelongsTo[2] = 2;           //       5
-        partitionThisDofBelongsTo[3] = 2;           //       4            9
-        partitionThisDofBelongsTo[4] = 3;           //       4
-        partitionThisDofBelongsTo[5] = 3;           //       5            9
-        break;
-      case 3:
-        partitionThisDofBelongsTo[0] = 1;           //       5
-        partitionThisDofBelongsTo[1] = 1;           //       4            9
-        partitionThisDofBelongsTo[2] = 2;           //       3
-        partitionThisDofBelongsTo[3] = 2;           //       4            7
-        partitionThisDofBelongsTo[4] = 3;           //       4
-        partitionThisDofBelongsTo[5] = 3;           //       3            7
-        break;
-    } //switch
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-
-    repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
-
-    Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
-    switch (myRank)  {
-      case 0:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 3);
-        TEST_EQUALITY(decompEntries[4], 0);
-        TEST_EQUALITY(decompEntries[5], 0);
-        break;
-      case 1:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 3);
-        TEST_EQUALITY(decompEntries[4], 1);
-        TEST_EQUALITY(decompEntries[5], 1);
-        break;
-      case 2:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 1);
-        TEST_EQUALITY(decompEntries[3], 1);
-        TEST_EQUALITY(decompEntries[4], 0);
-        TEST_EQUALITY(decompEntries[5], 0);
-        break;
-      case 3:
-        TEST_EQUALITY(decompEntries[0], 3);
-        TEST_EQUALITY(decompEntries[1], 3);
-        TEST_EQUALITY(decompEntries[2], 1);
-        TEST_EQUALITY(decompEntries[3], 1);
-        TEST_EQUALITY(decompEntries[4], 0);
-        TEST_EQUALITY(decompEntries[5], 0);
-        break;
-    }
-  } // DeterminePartitionPlacement2
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
-    out << "Matrix is initially placed on pids 0 and 1, but there are 4 partitions." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 5, ny = 3;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const GlobalOrdinal     indexBase     = 0;
-    size_t numMyElements = 0;
-    switch (myRank) {
-      case 0:  numMyElements = 8; break;
-      case 1:  numMyElements = 7; break;
-    }
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo                           = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, parth2 has 9, part3 has 6
+  const int numPartitions = 4;
+  switch (myRank) {
+    case 0:                              //  nnz by row    nnz by partition
+      partitionThisDofBelongsTo[0] = 0;  //       3
+      partitionThisDofBelongsTo[1] = 0;  //       4            7
+      partitionThisDofBelongsTo[2] = 1;  //       4
+      partitionThisDofBelongsTo[3] = 1;  //       3            7
+      partitionThisDofBelongsTo[4] = 3;  //       4
+      partitionThisDofBelongsTo[5] = 3;  //       5            9
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;  //       5
+      partitionThisDofBelongsTo[1] = 0;  //       4            9
+      partitionThisDofBelongsTo[2] = 1;  //       4
+      partitionThisDofBelongsTo[3] = 1;  //       5            9
+      partitionThisDofBelongsTo[4] = 2;  //       5
+      partitionThisDofBelongsTo[5] = 2;  //       4            9
+      break;
+    case 2:
+      partitionThisDofBelongsTo[0] = 0;  //       4
+      partitionThisDofBelongsTo[1] = 0;  //       5            9
+      partitionThisDofBelongsTo[2] = 2;  //       5
+      partitionThisDofBelongsTo[3] = 2;  //       4            9
+      partitionThisDofBelongsTo[4] = 3;  //       4
+      partitionThisDofBelongsTo[5] = 3;  //       5            9
+      break;
+    case 3:
+      partitionThisDofBelongsTo[0] = 1;  //       5
+      partitionThisDofBelongsTo[1] = 1;  //       4            9
+      partitionThisDofBelongsTo[2] = 2;  //       3
+      partitionThisDofBelongsTo[3] = 2;  //       4            7
+      partitionThisDofBelongsTo[4] = 3;  //       4
+      partitionThisDofBelongsTo[5] = 3;  //       3            7
+      break;
+  }  // switch
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+
+  repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
+
+  Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
+  switch (myRank) {
+    case 0:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 3);
+      TEST_EQUALITY(decompEntries[4], 0);
+      TEST_EQUALITY(decompEntries[5], 0);
+      break;
+    case 1:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 3);
+      TEST_EQUALITY(decompEntries[4], 1);
+      TEST_EQUALITY(decompEntries[5], 1);
+      break;
+    case 2:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 1);
+      TEST_EQUALITY(decompEntries[3], 1);
+      TEST_EQUALITY(decompEntries[4], 0);
+      TEST_EQUALITY(decompEntries[5], 0);
+      break;
+    case 3:
+      TEST_EQUALITY(decompEntries[0], 3);
+      TEST_EQUALITY(decompEntries[1], 3);
+      TEST_EQUALITY(decompEntries[2], 1);
+      TEST_EQUALITY(decompEntries[3], 1);
+      TEST_EQUALITY(decompEntries[4], 0);
+      TEST_EQUALITY(decompEntries[5], 0);
+      break;
+  }
+}  // DeterminePartitionPlacement2
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
+  out << "Matrix is initially placed on pids 0 and 1, but there are 4 partitions." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 5, ny = 3;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const GlobalOrdinal indexBase         = 0;
+  size_t numMyElements                  = 0;
+  switch (myRank) {
+    case 0: numMyElements = 8; break;
+    case 1: numMyElements = 7; break;
+  }
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
-    if (decomposition->getLocalLength())
-      partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 3, part1 has 4, parth2 has 3, part3 has 5
-    const int numPartitions = 4;
-    switch (myRank)  {
-      case 0:                                       //  nnz by row    nnz by partition
-        partitionThisDofBelongsTo[0] = 0;           //      3
-        partitionThisDofBelongsTo[1] = 0;           //      4             7
-        partitionThisDofBelongsTo[2] = 1;           //      4
-        partitionThisDofBelongsTo[3] = 1;           //      4            11
-        partitionThisDofBelongsTo[4] = 1;           //      3
-        partitionThisDofBelongsTo[5] = 2;           //      4
-        partitionThisDofBelongsTo[6] = 2;           //      5             9
-        partitionThisDofBelongsTo[7] = 3;           //      5             5
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;           //      5             5
-        partitionThisDofBelongsTo[1] = 1;           //      4             4
-        partitionThisDofBelongsTo[2] = 2;           //      3             3
-        partitionThisDofBelongsTo[3] = 3;           //      4
-        partitionThisDofBelongsTo[4] = 3;           //      4            16
-        partitionThisDofBelongsTo[5] = 3;           //      4
-        partitionThisDofBelongsTo[6] = 3;           //      3
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-
-    repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
-
-    Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
-    switch (myRank)  {
-      case 0:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 0);
-        TEST_EQUALITY(decompEntries[3], 0);
-        TEST_EQUALITY(decompEntries[4], 0);
-        TEST_EQUALITY(decompEntries[5], 3);
-        TEST_EQUALITY(decompEntries[6], 3);
-        TEST_EQUALITY(decompEntries[7], 1);
-        break;
-      case 1:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 0);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 1);
-        TEST_EQUALITY(decompEntries[4], 1);
-        TEST_EQUALITY(decompEntries[5], 1);
-        TEST_EQUALITY(decompEntries[6], 1);
-        break;
-    }
-  } // DeterminePartitionPlacement3
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement4, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
-    out << "Matrix is distributed across all four processors, but there are only 3 partitions." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 5, ny = 3;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const GlobalOrdinal     indexBase     = 0;
-    size_t numMyElements = 0;
-    switch (myRank) {
-      case 0:  numMyElements = 3; break;
-      case 1:  numMyElements = 4; break;
-      case 2:  numMyElements = 3; break;
-      case 3:  numMyElements = 5; break;
-    }
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
+  if (decomposition->getLocalLength())
+    partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 3, part1 has 4, parth2 has 3, part3 has 5
+  const int numPartitions = 4;
+  switch (myRank) {
+    case 0:                              //  nnz by row    nnz by partition
+      partitionThisDofBelongsTo[0] = 0;  //      3
+      partitionThisDofBelongsTo[1] = 0;  //      4             7
+      partitionThisDofBelongsTo[2] = 1;  //      4
+      partitionThisDofBelongsTo[3] = 1;  //      4            11
+      partitionThisDofBelongsTo[4] = 1;  //      3
+      partitionThisDofBelongsTo[5] = 2;  //      4
+      partitionThisDofBelongsTo[6] = 2;  //      5             9
+      partitionThisDofBelongsTo[7] = 3;  //      5             5
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;  //      5             5
+      partitionThisDofBelongsTo[1] = 1;  //      4             4
+      partitionThisDofBelongsTo[2] = 2;  //      3             3
+      partitionThisDofBelongsTo[3] = 3;  //      4
+      partitionThisDofBelongsTo[4] = 3;  //      4            16
+      partitionThisDofBelongsTo[5] = 3;  //      4
+      partitionThisDofBelongsTo[6] = 3;  //      3
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+
+  repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
+
+  Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
+  switch (myRank) {
+    case 0:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 0);
+      TEST_EQUALITY(decompEntries[3], 0);
+      TEST_EQUALITY(decompEntries[4], 0);
+      TEST_EQUALITY(decompEntries[5], 3);
+      TEST_EQUALITY(decompEntries[6], 3);
+      TEST_EQUALITY(decompEntries[7], 1);
+      break;
+    case 1:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 0);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 1);
+      TEST_EQUALITY(decompEntries[4], 1);
+      TEST_EQUALITY(decompEntries[5], 1);
+      TEST_EQUALITY(decompEntries[6], 1);
+      break;
+  }
+}  // DeterminePartitionPlacement3
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement4, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
+  out << "Matrix is distributed across all four processors, but there are only 3 partitions." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 5, ny = 3;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const GlobalOrdinal indexBase         = 0;
+  size_t numMyElements                  = 0;
+  switch (myRank) {
+    case 0: numMyElements = 3; break;
+    case 1: numMyElements = 4; break;
+    case 2: numMyElements = 3; break;
+    case 3: numMyElements = 5; break;
+  }
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
-    if (decomposition->getLocalLength())
-      partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, parth2 has 3
-    const int numPartitions = 3;
-    switch (myRank)  {
-      case 0:                                       //  nnz by row    nnz by partition
-        partitionThisDofBelongsTo[0] = 0;           //      3             3
-        partitionThisDofBelongsTo[1] = 1;           //      4             4
-        partitionThisDofBelongsTo[2] = 2;           //      4             4
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;           //      4             4
-        partitionThisDofBelongsTo[1] = 1;           //      3             3
-        partitionThisDofBelongsTo[2] = 2;           //      4
-        partitionThisDofBelongsTo[3] = 2;           //      5             9
-        break;
-      case 2:
-        partitionThisDofBelongsTo[0] = 0;           //      5
-        partitionThisDofBelongsTo[1] = 0;           //      5            10
-        partitionThisDofBelongsTo[2] = 1;           //      4             4
-        break;
-      case 3:
-        partitionThisDofBelongsTo[0] = 0;           //      3
-        partitionThisDofBelongsTo[1] = 0;           //      4             7
-        partitionThisDofBelongsTo[2] = 1;           //      4
-        partitionThisDofBelongsTo[3] = 1;           //      4
-        partitionThisDofBelongsTo[4] = 1;           //      3            11
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-
-    repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
-
-    Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
-    switch (myRank)  {
-      case 0:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 3);
-        TEST_EQUALITY(decompEntries[2], 1);
-        break;
-      case 1:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 3);
-        TEST_EQUALITY(decompEntries[2], 1);
-        TEST_EQUALITY(decompEntries[3], 1);
-        break;
-      case 2:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 3);
-        break;
-      case 3:
-        TEST_EQUALITY(decompEntries[0], 2);
-        TEST_EQUALITY(decompEntries[1], 2);
-        TEST_EQUALITY(decompEntries[2], 3);
-        TEST_EQUALITY(decompEntries[3], 3);
-        TEST_EQUALITY(decompEntries[4], 3);
-        break;
-    }
-  } // DeterminePartitionPlacement4
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement5, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
-    out << "Matrix is distributed across all four processors, but there are only 3 partitions." << std::endl;
-    out << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 5, ny = 3;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const GlobalOrdinal     indexBase     = 0;
-    size_t numMyElements = 0;
-    switch (myRank) {
-      case 0:  numMyElements = 3; break;
-      case 1:  numMyElements = 4; break;
-      case 2:  numMyElements = 3; break;
-      case 3:  numMyElements = 5; break;
-    }
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
+  if (decomposition->getLocalLength())
+    partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, parth2 has 3
+  const int numPartitions = 3;
+  switch (myRank) {
+    case 0:                              //  nnz by row    nnz by partition
+      partitionThisDofBelongsTo[0] = 0;  //      3             3
+      partitionThisDofBelongsTo[1] = 1;  //      4             4
+      partitionThisDofBelongsTo[2] = 2;  //      4             4
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;  //      4             4
+      partitionThisDofBelongsTo[1] = 1;  //      3             3
+      partitionThisDofBelongsTo[2] = 2;  //      4
+      partitionThisDofBelongsTo[3] = 2;  //      5             9
+      break;
+    case 2:
+      partitionThisDofBelongsTo[0] = 0;  //      5
+      partitionThisDofBelongsTo[1] = 0;  //      5            10
+      partitionThisDofBelongsTo[2] = 1;  //      4             4
+      break;
+    case 3:
+      partitionThisDofBelongsTo[0] = 0;  //      3
+      partitionThisDofBelongsTo[1] = 0;  //      4             7
+      partitionThisDofBelongsTo[2] = 1;  //      4
+      partitionThisDofBelongsTo[3] = 1;  //      4
+      partitionThisDofBelongsTo[4] = 1;  //      3            11
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+
+  repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions);
+
+  Teuchos::ArrayRCP<GlobalOrdinal> decompEntries = decomposition->getDataNonConst(0);
+  switch (myRank) {
+    case 0:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 3);
+      TEST_EQUALITY(decompEntries[2], 1);
+      break;
+    case 1:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 3);
+      TEST_EQUALITY(decompEntries[2], 1);
+      TEST_EQUALITY(decompEntries[3], 1);
+      break;
+    case 2:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 3);
+      break;
+    case 3:
+      TEST_EQUALITY(decompEntries[0], 2);
+      TEST_EQUALITY(decompEntries[1], 2);
+      TEST_EQUALITY(decompEntries[2], 3);
+      TEST_EQUALITY(decompEntries[3], 3);
+      TEST_EQUALITY(decompEntries[4], 3);
+      break;
+  }
+}  // DeterminePartitionPlacement4
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, DeterminePartitionPlacement5, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests the algorithm for assigning partitions to PIDs." << std::endl;
+  out << "Matrix is distributed across all four processors, but there are only 3 partitions." << std::endl;
+  out << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 5, ny = 3;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const GlobalOrdinal indexBase         = 0;
+  size_t numMyElements                  = 0;
+  switch (myRank) {
+    case 0: numMyElements = 3; break;
+    case 1: numMyElements = 4; break;
+    case 2: numMyElements = 3; break;
+    case 3: numMyElements = 5; break;
+  }
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
-    if (decomposition->getLocalLength())
-      partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, part2 has 3
-    const int numPartitions = 2;
-    switch (myRank)  {
-      case 0:                                       //  nnz by row    nnz by partition
-        partitionThisDofBelongsTo[0] = 0;           //      3             3
-        partitionThisDofBelongsTo[1] = 1;           //      4
-        partitionThisDofBelongsTo[2] = 1;           //      4             8
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;           //      4
-        partitionThisDofBelongsTo[1] = 1;           //      3
-        partitionThisDofBelongsTo[2] = 0;           //      4             8
-        partitionThisDofBelongsTo[3] = 1;           //      5             8
-        break;
-      case 2:
-        partitionThisDofBelongsTo[0] = 0;           //      5             5
-        partitionThisDofBelongsTo[1] = 1;           //      5
-        partitionThisDofBelongsTo[2] = 1;           //      4             9
-        break;
-      case 3:
-        partitionThisDofBelongsTo[0] = 0;           //      3
-        partitionThisDofBelongsTo[1] = 0;           //      4
-        partitionThisDofBelongsTo[2] = 0;           //      4
-        partitionThisDofBelongsTo[3] = 0;           //      4            15
-        partitionThisDofBelongsTo[4] = 1;           //      3             3
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-    Teuchos::ParameterList paramList;
-    paramList.set("repartition: remap num values",       1);
-    repart->SetParameterList(paramList);
-
-    bool willAcceptPartition;
-    // force rebalancing onto ranks 2 and 3.
-    if (myRank == 1 || myRank == 2)
-      willAcceptPartition = true;
-    else
-      willAcceptPartition = false;
-    const bool allSubdomainsAcceptPartitions = false;
-
-    repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions, willAcceptPartition, allSubdomainsAcceptPartitions);
-
-  } // DeterminePartitionPlacement5
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests application of the permutation matrix to matrix A." << std::endl;
-    out << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    int numProcs = comm->getSize();
-    int myRank   = comm->getRank();
-
-    if (numProcs != 4) {
-      std::cout << "\nThis test must be run on 4 processors!\n" << std::endl;
-      return;
-    }
-
-    const GlobalOrdinal nx = 3, ny = 5;
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx",      nx);
-    matrixList.set("ny",      ny);
-    matrixList.set("keepBCs", false);
-
-    // Describes the initial layout of matrix rows across processors.
-    const GlobalOrdinal     numGlobalElements = nx*ny; // 24
-    const GlobalOrdinal     indexBase     = 0;
-    size_t numMyElements = 0;
-    switch (myRank) {
-      case 0:  numMyElements = 6; break;
-      case 1:  numMyElements = 5; break;
-      case 2:  numMyElements = 0; break;
-      case 3:  numMyElements = 4; break;
-    }
-
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
+  if (decomposition->getLocalLength())
+    partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 6, part2 has 3
+  const int numPartitions = 2;
+  switch (myRank) {
+    case 0:                              //  nnz by row    nnz by partition
+      partitionThisDofBelongsTo[0] = 0;  //      3             3
+      partitionThisDofBelongsTo[1] = 1;  //      4
+      partitionThisDofBelongsTo[2] = 1;  //      4             8
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;  //      4
+      partitionThisDofBelongsTo[1] = 1;  //      3
+      partitionThisDofBelongsTo[2] = 0;  //      4             8
+      partitionThisDofBelongsTo[3] = 1;  //      5             8
+      break;
+    case 2:
+      partitionThisDofBelongsTo[0] = 0;  //      5             5
+      partitionThisDofBelongsTo[1] = 1;  //      5
+      partitionThisDofBelongsTo[2] = 1;  //      4             9
+      break;
+    case 3:
+      partitionThisDofBelongsTo[0] = 0;  //      3
+      partitionThisDofBelongsTo[1] = 0;  //      4
+      partitionThisDofBelongsTo[2] = 0;  //      4
+      partitionThisDofBelongsTo[3] = 0;  //      4            15
+      partitionThisDofBelongsTo[4] = 1;  //      3             3
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+  Teuchos::ParameterList paramList;
+  paramList.set("repartition: remap num values", 1);
+  repart->SetParameterList(paramList);
+
+  bool willAcceptPartition;
+  // force rebalancing onto ranks 2 and 3.
+  if (myRank == 1 || myRank == 2)
+    willAcceptPartition = true;
+  else
+    willAcceptPartition = false;
+  const bool allSubdomainsAcceptPartitions = false;
+
+  repart->DeterminePartitionPlacement(*A, *decomposition, numPartitions, willAcceptPartition, allSubdomainsAcceptPartitions);
+
+}  // DeterminePartitionPlacement5
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests application of the permutation matrix to matrix A." << std::endl;
+  out << std::endl;
+
+  using TST            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename TST::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  int numProcs                        = comm->getSize();
+  int myRank                          = comm->getRank();
+
+  if (numProcs != 4) {
+    std::cout << "\nThis test must be run on 4 processors!\n"
+              << std::endl;
+    return;
+  }
+
+  const GlobalOrdinal nx = 3, ny = 5;
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", nx);
+  matrixList.set("ny", ny);
+  matrixList.set("keepBCs", false);
+
+  // Describes the initial layout of matrix rows across processors.
+  const GlobalOrdinal numGlobalElements = nx * ny;  // 24
+  const GlobalOrdinal indexBase         = 0;
+  size_t numMyElements                  = 0;
+  switch (myRank) {
+    case 0: numMyElements = 6; break;
+    case 1: numMyElements = 5; break;
+    case 2: numMyElements = 0; break;
+    case 3: numMyElements = 4; break;
+  }
+
+  RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyElements, indexBase, comm);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node>::Build(map, false);
-    RCP<MultiVector>                  coords        = MultiVectorFactory::Build(map, 2);
-    Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
-    if (decomposition->getLocalLength())
-      partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
-
-    // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 4, part2 has 5
-    const int numPartitions = 3;
-    switch (myRank)  {
-      case 0:
-        partitionThisDofBelongsTo[0] = 0;
-        partitionThisDofBelongsTo[1] = 1;
-        partitionThisDofBelongsTo[2] = 2;
-        partitionThisDofBelongsTo[3] = 0;
-        partitionThisDofBelongsTo[4] = 2;
-        partitionThisDofBelongsTo[5] = 2;
-        break;
-      case 1:
-        partitionThisDofBelongsTo[0] = 0;
-        partitionThisDofBelongsTo[1] = 1;
-        partitionThisDofBelongsTo[2] = 0;
-        partitionThisDofBelongsTo[3] = 2;
-        partitionThisDofBelongsTo[4] = 1;
-        break;
-      case 2:
-        break;
-      case 3:
-        partitionThisDofBelongsTo[0] = 2;
-        partitionThisDofBelongsTo[1] = 0;
-        partitionThisDofBelongsTo[2] = 0;
-        partitionThisDofBelongsTo[3] = 1;
-        break;
-    }
-    partitionThisDofBelongsTo = Teuchos::null;
-
-    // This test uses a made up partitioning  that is given via Level to Repartition.  It must be
-    // associated with an instance of the ZoltanInterface so that it can be found inside
-    // Repartition.  Furthermore, that same instance must be supplied to MueLu::Repartition.
-    // Coordinates must be provided, even though they are not used, as the Zoltan interface checks for them.
-    Level level;
-    level.SetLevelID(1);
-
-    RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-    level.SetFactoryManager(factoryHandler);
-
-    RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-
-    level.Set<RCP<Matrix> >     ("A",                    A);
-    level.Set<RCP<MultiVector> >("Coordinates",          coords);
-    level.Set<int>              ("number of partitions", numPartitions);
-
-    level.Request("Partition", zoltan.get());
-    level.Set<RCP<Xpetra::Vector<GlobalOrdinal,LocalOrdinal,GlobalOrdinal,Node> > >("Partition", decomposition, zoltan.get());
-
-    RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
-    Teuchos::ParameterList paramList;
-    paramList.set("repartition: remap parts",       false);
-    repart->SetParameterList(paramList);
-    repart->SetFactory("Partition", zoltan);
-
-    // Build
-    level.Request("Importer", repart.get());
-    repart->Build(level);
-
-    RCP<const Import> importer;
-    level.Get("Importer", importer, repart.get());
-
-    RCP<Matrix>        permutedA = MatrixFactory::Build(importer->getTargetMap(), A->getGlobalMaxNumRowEntries());
-    permutedA->doImport(*A, *importer, Xpetra::INSERT);
-    permutedA->fillComplete(A->getDomainMap(), importer->getTargetMap());
-
-    // Calculate vectors y1 = Perm[A*v] and y2 = Perm[A]*v for random vector v. They should be identical (to machine precision)
-    RCP<Vector> randVec = VectorFactory::Build(A->getDomainMap(), false);
-    RCP<Vector> workVec = VectorFactory::Build(A->getRangeMap(),  false);
-    randVec->randomize();
-
-    RCP<Vector> P_Av = VectorFactory::Build(importer->getTargetMap(), false);
-    RCP<Vector> PA_v = VectorFactory::Build(permutedA->getRangeMap(), false);
-
-    const Scalar zero = TST::zero(), one = TST::one();
-
-    permutedA->apply(*randVec, *PA_v, Teuchos::NO_TRANS, one, zero);
-
-    A->apply(*randVec, *workVec, Teuchos::NO_TRANS, one, zero);
-    P_Av->doImport(*workVec,*importer,Xpetra::INSERT);
-
-    RCP<MultiVector> diff = VectorFactory::Build(permutedA->getRangeMap());
-    diff->update(one, *P_Av, -one, *PA_v, zero);
-
-    Teuchos::Array<magnitude_type> norms(1);
-    diff->norm2(norms);
-    out << "||diff|| = " << norms[0] << std::endl;
-    TEST_EQUALITY(norms[0] < 100*TMT::eps(), true);
-
-  } // Correctness
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, CoordinateMap, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests that repartitioning is invariant to map specified in coordinates." << std::endl;
-    out << std::endl;
-
-    //FIXME JJH this is a hack until I can figure out why this test won't pass for Scalar=complex
-    std::string scalarName = Teuchos::ScalarTraits<Scalar>::name();
-    out << "scalar type = " << scalarName << std::endl;
-    if (scalarName.find("complex") != std::string::npos) {
-      out << "Skipping Test for SC=complex" << std::endl;
-      return;
-    }
-
-    /*
-       This test checks that MueLu successfully ignores the map of the coordinate MultiVector (MV).
-       MueLu treats the coordinate data as if the MV is consistent with the linear system A.
-       */
-
-    //
-/*
-    if Teuchos::ScalarTraits<Scalar>::name() != std::string("double") {
-      std::cout << "Skipping for SC other than \"double\"" << std::endl;
-      return;
-    }
-*/
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > decomposition = Xpetra::VectorFactory<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node>::Build(map, false);
+  RCP<MultiVector> coords                                                              = MultiVectorFactory::Build(map, 2);
+  Teuchos::ArrayRCP<GlobalOrdinal> partitionThisDofBelongsTo;
+  if (decomposition->getLocalLength())
+    partitionThisDofBelongsTo = decomposition->getDataNonConst(0);
+
+  // Assign the partition that each unknown belongs to. In this case: part0 has 6, part1 has 4, part2 has 5
+  const int numPartitions = 3;
+  switch (myRank) {
+    case 0:
+      partitionThisDofBelongsTo[0] = 0;
+      partitionThisDofBelongsTo[1] = 1;
+      partitionThisDofBelongsTo[2] = 2;
+      partitionThisDofBelongsTo[3] = 0;
+      partitionThisDofBelongsTo[4] = 2;
+      partitionThisDofBelongsTo[5] = 2;
+      break;
+    case 1:
+      partitionThisDofBelongsTo[0] = 0;
+      partitionThisDofBelongsTo[1] = 1;
+      partitionThisDofBelongsTo[2] = 0;
+      partitionThisDofBelongsTo[3] = 2;
+      partitionThisDofBelongsTo[4] = 1;
+      break;
+    case 2:
+      break;
+    case 3:
+      partitionThisDofBelongsTo[0] = 2;
+      partitionThisDofBelongsTo[1] = 0;
+      partitionThisDofBelongsTo[2] = 0;
+      partitionThisDofBelongsTo[3] = 1;
+      break;
+  }
+  partitionThisDofBelongsTo = Teuchos::null;
+
+  // This test uses a made up partitioning  that is given via Level to Repartition.  It must be
+  // associated with an instance of the ZoltanInterface so that it can be found inside
+  // Repartition.  Furthermore, that same instance must be supplied to MueLu::Repartition.
+  // Coordinates must be provided, even though they are not used, as the Zoltan interface checks for them.
+  Level level;
+  level.SetLevelID(1);
+
+  RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+  level.SetFactoryManager(factoryHandler);
+
+  RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
+
+  level.Set<RCP<Matrix> >("A", A);
+  level.Set<RCP<MultiVector> >("Coordinates", coords);
+  level.Set<int>("number of partitions", numPartitions);
+
+  level.Request("Partition", zoltan.get());
+  level.Set<RCP<Xpetra::Vector<GlobalOrdinal, LocalOrdinal, GlobalOrdinal, Node> > >("Partition", decomposition, zoltan.get());
+
+  RCP<RepartitionFactory> repart = rcp(new RepartitionFactory());
+  Teuchos::ParameterList paramList;
+  paramList.set("repartition: remap parts", false);
+  repart->SetParameterList(paramList);
+  repart->SetFactory("Partition", zoltan);
+
+  // Build
+  level.Request("Importer", repart.get());
+  repart->Build(level);
+
+  RCP<const Import> importer;
+  level.Get("Importer", importer, repart.get());
+
+  RCP<Matrix> permutedA = MatrixFactory::Build(importer->getTargetMap(), A->getGlobalMaxNumRowEntries());
+  permutedA->doImport(*A, *importer, Xpetra::INSERT);
+  permutedA->fillComplete(A->getDomainMap(), importer->getTargetMap());
+
+  // Calculate vectors y1 = Perm[A*v] and y2 = Perm[A]*v for random vector v. They should be identical (to machine precision)
+  RCP<Vector> randVec = VectorFactory::Build(A->getDomainMap(), false);
+  RCP<Vector> workVec = VectorFactory::Build(A->getRangeMap(), false);
+  randVec->randomize();
+
+  RCP<Vector> P_Av = VectorFactory::Build(importer->getTargetMap(), false);
+  RCP<Vector> PA_v = VectorFactory::Build(permutedA->getRangeMap(), false);
+
+  const Scalar zero = TST::zero(), one = TST::one();
+
+  permutedA->apply(*randVec, *PA_v, Teuchos::NO_TRANS, one, zero);
+
+  A->apply(*randVec, *workVec, Teuchos::NO_TRANS, one, zero);
+  P_Av->doImport(*workVec, *importer, Xpetra::INSERT);
+
+  RCP<MultiVector> diff = VectorFactory::Build(permutedA->getRangeMap());
+  diff->update(one, *P_Av, -one, *PA_v, zero);
+
+  Teuchos::Array<magnitude_type> norms(1);
+  diff->norm2(norms);
+  out << "||diff|| = " << norms[0] << std::endl;
+  TEST_EQUALITY(norms[0] < 100 * TMT::eps(), true);
+
+}  // Correctness
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, CoordinateMap, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests that repartitioning is invariant to map specified in coordinates." << std::endl;
+  out << std::endl;
+
+  // FIXME JJH this is a hack until I can figure out why this test won't pass for Scalar=complex
+  std::string scalarName = Teuchos::ScalarTraits<Scalar>::name();
+  out << "scalar type = " << scalarName << std::endl;
+  if (scalarName.find("complex") != std::string::npos) {
+    out << "Skipping Test for SC=complex" << std::endl;
+    return;
+  }
+
+  /*
+     This test checks that MueLu successfully ignores the map of the coordinate MultiVector (MV).
+     MueLu treats the coordinate data as if the MV is consistent with the linear system A.
+     */
+
+  //
+  /*
+      if Teuchos::ScalarTraits<Scalar>::name() != std::string("double") {
+        std::cout << "Skipping for SC other than \"double\"" << std::endl;
+        return;
+      }
+  */
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> mv_type_double;
-    typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> MVFactory_double;
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> mv_type_double;
+  typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> MVFactory_double;
 
-    // Create a matrix and coordinates.
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  // Create a matrix and coordinates.
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    GlobalOrdinal nx = 20, ny = 20;
+  GlobalOrdinal nx = 20, ny = 20;
 
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
 
-    //build coordinates before expanding map (nodal coordinates, not dof-based)
-    //RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,Map,MultiVector>("2D", map, galeriList);
-    RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LocalOrdinal,GlobalOrdinal,Map,mv_type_double>("2D", map, galeriList);
-    map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
+  // build coordinates before expanding map (nodal coordinates, not dof-based)
+  // RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<Scalar,LocalOrdinal,GlobalOrdinal,Map,MultiVector>("2D", map, galeriList);
+  RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LocalOrdinal, GlobalOrdinal, Map, mv_type_double>("2D", map, galeriList);
+  map                             = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
 
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
-
-    Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mm", *A);
-    comm->barrier();
-
-    RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter("testCoordinates.xml", *comm));
-    RCP<Hierarchy> H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A", A);
-    H->GetLevel(0)->Set("Coordinates", coordinates);
-    mueLuFactory->SetupHierarchy(*H);
-    double cplx1 = H->GetOperatorComplexity();
-
-    //build a map that is a "randomly" permuted version of the correct
-    //coordinate map.  This map will be used to build the "bad" coordinates.
-    RCP<const Map> coordMap = coordinates->getMap();
-    std::srand(Teuchos::as<unsigned int>(comm->getRank()*31415));
-    Teuchos::ArrayView<const GlobalOrdinal> correctLocalElts = coordMap->getLocalElementList();
-    std::vector<GlobalOrdinal> eltsToShuffle;
-    for (size_t i=0; i < Teuchos::as<size_t>(correctLocalElts.size()); ++i)
-      eltsToShuffle.push_back(correctLocalElts[i]);
-    std::random_shuffle(eltsToShuffle.begin(),eltsToShuffle.end(), [](GlobalOrdinal i) { return std::rand()%i;});
-    Teuchos::Array<GlobalOrdinal> eltList(eltsToShuffle);
-    RCP<const Map> badMap = MapFactory::Build(TestHelpers::Parameters::getLib(), coordMap->getGlobalNumElements(), eltList(), coordMap->getIndexBase(), comm);
-
-    Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coordVals;
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> xcoords = coordinates->getData(0);
-    Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> ycoords = coordinates->getData(1);
-    coordVals.push_back(xcoords());
-    coordVals.push_back(ycoords());
-    RCP<mv_type_double> badCoordinates = MVFactory_double::Build(badMap, coordVals(), coordinates->getNumVectors());
-    xcoords = Teuchos::null;
-    ycoords = Teuchos::null;
-
-    mueLuFactory = rcp(new ParameterListInterpreter("testCoordinates.xml", *comm));
-    H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A", A);
-    H->GetLevel(0)->Set("Coordinates", badCoordinates);
-    mueLuFactory->SetupHierarchy(*H);
-    double cplx2 = H->GetOperatorComplexity();
-
-    TEST_EQUALITY(cplx1, cplx2);
-
-  } // CoordinateMap
-
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, NodePartition, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
-#   endif
-   
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Tests that node repartitioning works " << std::endl;
-    out << std::endl;
-
-    //FIXME JJH this is a hack until I can figure out why this test won't pass for Scalar=complex
-    std::string scalarName = Teuchos::ScalarTraits<Scalar>::name();
-    out << "scalar type = " << scalarName << std::endl;
-    if (scalarName.find("complex") != std::string::npos) {
-      out << "Skipping Test for SC=complex" << std::endl;
-      return;
-    }
-
- 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> mv_type_double;
-    typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> MVFactory_double;
-
-    // Create a matrix and coordinates.
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GlobalOrdinal nx = 20, ny = 20;
-
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-
-    // build coordinates
-    RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LocalOrdinal,GlobalOrdinal,Map,mv_type_double>("2D", map, galeriList);
-
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
+
+  Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Write("A.mm", *A);
+  comm->barrier();
+
+  RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter("testCoordinates.xml", *comm));
+  RCP<Hierarchy> H                   = mueLuFactory->CreateHierarchy();
+  H->GetLevel(0)->Set("A", A);
+  H->GetLevel(0)->Set("Coordinates", coordinates);
+  mueLuFactory->SetupHierarchy(*H);
+  double cplx1 = H->GetOperatorComplexity();
+
+  // build a map that is a "randomly" permuted version of the correct
+  // coordinate map.  This map will be used to build the "bad" coordinates.
+  RCP<const Map> coordMap = coordinates->getMap();
+  std::srand(Teuchos::as<unsigned int>(comm->getRank() * 31415));
+  Teuchos::ArrayView<const GlobalOrdinal> correctLocalElts = coordMap->getLocalElementList();
+  std::vector<GlobalOrdinal> eltsToShuffle;
+  for (size_t i = 0; i < Teuchos::as<size_t>(correctLocalElts.size()); ++i)
+    eltsToShuffle.push_back(correctLocalElts[i]);
+  std::random_shuffle(eltsToShuffle.begin(), eltsToShuffle.end(), [](GlobalOrdinal i) { return std::rand() % i; });
+  Teuchos::Array<GlobalOrdinal> eltList(eltsToShuffle);
+  RCP<const Map> badMap = MapFactory::Build(TestHelpers::Parameters::getLib(), coordMap->getGlobalNumElements(), eltList(), coordMap->getIndexBase(), comm);
+
+  Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > coordVals;
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> xcoords = coordinates->getData(0);
+  Teuchos::ArrayRCP<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> ycoords = coordinates->getData(1);
+  coordVals.push_back(xcoords());
+  coordVals.push_back(ycoords());
+  RCP<mv_type_double> badCoordinates = MVFactory_double::Build(badMap, coordVals(), coordinates->getNumVectors());
+  xcoords                            = Teuchos::null;
+  ycoords                            = Teuchos::null;
+
+  mueLuFactory = rcp(new ParameterListInterpreter("testCoordinates.xml", *comm));
+  H            = mueLuFactory->CreateHierarchy();
+  H->GetLevel(0)->Set("A", A);
+  H->GetLevel(0)->Set("Coordinates", badCoordinates);
+  mueLuFactory->SetupHierarchy(*H);
+  double cplx2 = H->GetOperatorComplexity();
+
+  TEST_EQUALITY(cplx1, cplx2);
+
+}  // CoordinateMap
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Repartition, NodePartition, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_AMESOS) || !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Amesos, Ifpack");
+#endif
+#if !defined(HAVE_MUELU_AMESOS2) || !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Amesos2, Ifpack2");
+#endif
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Tests that node repartitioning works " << std::endl;
+  out << std::endl;
+
+  // FIXME JJH this is a hack until I can figure out why this test won't pass for Scalar=complex
+  std::string scalarName = Teuchos::ScalarTraits<Scalar>::name();
+  out << "scalar type = " << scalarName << std::endl;
+  if (scalarName.find("complex") != std::string::npos) {
+    out << "Skipping Test for SC=complex" << std::endl;
+    return;
+  }
+
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> mv_type_double;
+  typedef Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> MVFactory_double;
+
+  // Create a matrix and coordinates.
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  GlobalOrdinal nx = 20, ny = 20;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+
+  // build coordinates
+  RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LocalOrdinal, GlobalOrdinal, Map, mv_type_double>("2D", map, galeriList);
+
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    // Generate the node-level communicator
-    Teuchos::RCP<const Teuchos::Comm<int> > nodeComm;
-    int NodeId = comm->getRank();
-    nodeComm = MueLu::GenerateNodeComm(comm,NodeId,1);
-
-    Teuchos::ParameterList paramList;
-    Teuchos::updateParametersFromXmlFileAndBroadcast("testCoordinates.xml", Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
-    paramList.set("repartition: node repartition level",1);
-    paramList.set("repartition: start level",2);
-    //    paramList.sublist("user data").set("Node Comm",nodeComm);
-    paramList.set("verbosity","high");
-    RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter(paramList));
- 
-    RCP<Hierarchy> H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A", A);
-    H->GetLevel(0)->Set("Coordinates", coordinates);
-    H->GetLevel(0)->Set("Node Comm", nodeComm);
-    mueLuFactory->SetupHierarchy(*H);
-
-    
-
-
-  } // NodePartition
-
-
-
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,Constructor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,Build,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,DeterminePartitionPlacement1,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,DeterminePartitionPlacement2,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,DeterminePartitionPlacement3,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,DeterminePartitionPlacement4,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,DeterminePartitionPlacement5,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,Correctness,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,CoordinateMap,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition,NodePartition,Scalar,LO,GO,Node) \
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  // Generate the node-level communicator
+  Teuchos::RCP<const Teuchos::Comm<int> > nodeComm;
+  int NodeId = comm->getRank();
+  nodeComm   = MueLu::GenerateNodeComm(comm, NodeId, 1);
+
+  Teuchos::ParameterList paramList;
+  Teuchos::updateParametersFromXmlFileAndBroadcast("testCoordinates.xml", Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
+  paramList.set("repartition: node repartition level", 1);
+  paramList.set("repartition: start level", 2);
+  //    paramList.sublist("user data").set("Node Comm",nodeComm);
+  paramList.set("verbosity", "high");
+  RCP<HierarchyManager> mueLuFactory = rcp(new ParameterListInterpreter(paramList));
+
+  RCP<Hierarchy> H = mueLuFactory->CreateHierarchy();
+  H->GetLevel(0)->Set("A", A);
+  H->GetLevel(0)->Set("Coordinates", coordinates);
+  H->GetLevel(0)->Set("Node Comm", nodeComm);
+  mueLuFactory->SetupHierarchy(*H);
+
+}  // NodePartition
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, Constructor, Scalar, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, Build, Scalar, LO, GO, Node)                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, DeterminePartitionPlacement1, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, DeterminePartitionPlacement2, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, DeterminePartitionPlacement3, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, DeterminePartitionPlacement4, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, DeterminePartitionPlacement5, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, Correctness, Scalar, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, CoordinateMap, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Repartition, NodePartition, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/RigidBodyModeFactory.cpp b/packages/muelu/test/unit_tests/RigidBodyModeFactory.cpp
index bedd5dfd24e2..df1c3ebda887 100644
--- a/packages/muelu/test/unit_tests/RigidBodyModeFactory.cpp
+++ b/packages/muelu/test/unit_tests/RigidBodyModeFactory.cpp
@@ -51,57 +51,54 @@
 
 #include <MueLu_RigidBodyModeFactory.hpp>
 
-
 namespace MueLuTests {
 
-
-TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RigidBodyModeFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-{
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    auto A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
-    const auto map = A->getRangeMap();
-
-    using RigidBodyModeFactory = MueLu::RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-
-    {
-        RigidBodyModeFactory rigidBodyModeFactory("TestName");
-        rigidBodyModeFactory.setNumPDEs(2);
-
-        MueLu::Level level;
-        level.SetLevelID(0);
-        auto nullSpace = MultiVectorFactory::Build(map, 100);
-        level.Set("TestName", nullSpace);
-        level.Keep("Nullspace", &rigidBodyModeFactory);
-        rigidBodyModeFactory.DeclareInput(level);
-
-        TEST_EQUALITY(false, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
-        rigidBodyModeFactory.Build(level);
-        TEST_EQUALITY(true, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
-    }
-
-    {
-        RigidBodyModeFactory rigidBodyModeFactory(1);
-
-        MueLu::Level level;
-        level.SetLevelID(0);
-        level.Set("A", A);
-        level.Keep("Nullspace", &rigidBodyModeFactory);
-        auto coords = MultiVectorFactory::Build(map, 100);
-        level.Set("Coordinates", coords);
-
-        TEST_EQUALITY(false, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
-        rigidBodyModeFactory.Build(level);
-        TEST_EQUALITY(true, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(RigidBodyModeFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  auto A         = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(100);
+  const auto map = A->getRangeMap();
+
+  using RigidBodyModeFactory = MueLu::RigidBodyModeFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+
+  {
+    RigidBodyModeFactory rigidBodyModeFactory("TestName");
+    rigidBodyModeFactory.setNumPDEs(2);
+
+    MueLu::Level level;
+    level.SetLevelID(0);
+    auto nullSpace = MultiVectorFactory::Build(map, 100);
+    level.Set("TestName", nullSpace);
+    level.Keep("Nullspace", &rigidBodyModeFactory);
+    rigidBodyModeFactory.DeclareInput(level);
+
+    TEST_EQUALITY(false, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
+    rigidBodyModeFactory.Build(level);
+    TEST_EQUALITY(true, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
+  }
+
+  {
+    RigidBodyModeFactory rigidBodyModeFactory(1);
+
+    MueLu::Level level;
+    level.SetLevelID(0);
+    level.Set("A", A);
+    level.Keep("Nullspace", &rigidBodyModeFactory);
+    auto coords = MultiVectorFactory::Build(map, 100);
+    level.Set("Coordinates", coords);
+
+    TEST_EQUALITY(false, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
+    rigidBodyModeFactory.Build(level);
+    TEST_EQUALITY(true, level.IsAvailable("Nullspace", &rigidBodyModeFactory));
+  }
 }
 
 #define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RigidBodyModeFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(RigidBodyModeFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SaPFactory.cpp b/packages/muelu/test/unit_tests/SaPFactory.cpp
index a59eb3c926b0..b4bed1314a9a 100644
--- a/packages/muelu/test/unit_tests/SaPFactory.cpp
+++ b/packages/muelu/test/unit_tests/SaPFactory.cpp
@@ -64,289 +64,286 @@
 
 namespace MueLuTests {
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  RCP<SaPFactory> sapFactory = rcp(new SaPFactory);
+  TEST_EQUALITY(sapFactory != Teuchos::null, true);
 
-    RCP<SaPFactory> sapFactory = rcp(new SaPFactory);
-    TEST_EQUALITY(sapFactory != Teuchos::null, true);
+  out << *sapFactory << std::endl;
+}
 
-    out << *sapFactory << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, EpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Compare results of Epetra and Tpetra" << std::endl;
+  out << "for 3 level AMG solver using smoothed aggregation with" << std::endl;
+  out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
 
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, EpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Compare results of Epetra and Tpetra" << std::endl;
-    out << "for 3 level AMG solver using smoothed aggregation with" << std::endl;
-    out << "one SGS sweep on each multigrid level as pre- and postsmoother" << std::endl;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+  MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar, GlobalOrdinal, Node);
 
-    MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar,GlobalOrdinal,Node);
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  Teuchos::Array<magnitude_type> results(2);
 
-    Teuchos::Array<magnitude_type> results(2);
+  // run test only on 1 proc
+  if (comm->getSize() == 1) {
+    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
 
-    // run test only on 1 proc
-    if(comm->getSize() == 1)
+    // run Epetra and Tpetra test
+    for (int run = 0; run < 2; run++)  // TODO: create a subfunction instead or Tuple of UnderlyingLib
     {
-      Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
-
-      // run Epetra and Tpetra test
-      for (int run = 0; run < 2; run++) //TODO: create a subfunction instead or Tuple of UnderlyingLib
+      if (run == 0)
+        lib = Xpetra::UseEpetra;
+      else
+        lib = Xpetra::UseTpetra;
+
+      // generate problem
+      LocalOrdinal maxLevels   = 3;
+      LocalOrdinal its         = 10;
+      GlobalOrdinal nEle       = 63;
+      const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+      Teuchos::ParameterList matrixParameters;
+      matrixParameters.set("nx", nEle);
+
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
+      RCP<Matrix> Op = Pr->BuildMatrix();
+
+      // build nullspace
+      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+      nullSpace->putScalar((Scalar)1.0);
+      Teuchos::Array<magnitude_type> norms(1);
+      nullSpace->norm1(norms);
+      if (comm->getRank() == 0)
+        out << "||NS|| = " << norms[0] << std::endl;
+
+      // fill hierarchy
+      RCP<Hierarchy> H = rcp(new Hierarchy());
+      H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+      RCP<Level> Finest = H->GetLevel();
+      Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+      Finest->Set("A", Op);                 // set fine level matrix
+      Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+
+      // define transfer operators
+      RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+      UncoupledAggFact->SetMinNodesPerAggregate(3);
+      UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+      UncoupledAggFact->SetOrdering("natural");
+
+      RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
+      RCP<SaPFactory> Pfact            = rcp(new SaPFactory());
+      RCP<Factory> Rfact               = rcp(new TransPFactory());
+      RCP<RAPFactory> Acfact           = rcp(new RAPFactory());
+      H->SetMaxCoarseSize(1);
+
+      // setup smoothers
+      Teuchos::ParameterList smootherParamList;
+      smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+      smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+      smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+      RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+      RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+      Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+      RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+      FactoryManager M;
+      M.SetKokkosRefactor(false);
+      M.SetFactory("P", Pfact);
+      M.SetFactory("R", Rfact);
+      M.SetFactory("A", Acfact);
+      M.SetFactory("Ptent", Ptentfact);
+      M.SetFactory("Aggregates", UncoupledAggFact);
+      M.SetFactory("Smoother", SmooFact);
+      M.SetFactory("CoarseSolver", coarseSolveFact);
+
+      H->Setup(M, 0, maxLevels);
+
+      // test some basic multigrid data
+      RCP<Level> coarseLevel = H->GetLevel(1);
+      TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("R", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsRequested("A", Acfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel->IsAvailable("A", Acfact.get()), false);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Pfact.get()), 0);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Ptentfact.get()), 0);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("R", Rfact.get()), 0);
+      TEST_EQUALITY(coarseLevel->GetKeepFlag("A", Acfact.get()), 0);
+      RCP<Matrix> P1 = coarseLevel->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R1 = coarseLevel->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P1->getGlobalNumRows(), 63);
+      TEST_EQUALITY(P1->getGlobalNumCols(), 21);
+      TEST_EQUALITY(R1->getGlobalNumRows(), 21);
+      TEST_EQUALITY(R1->getGlobalNumCols(), 63);
+      RCP<Level> coarseLevel2 = H->GetLevel(2);
+      TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("R", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("A", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("R", MueLu::NoFactory::get()), true);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsRequested("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", Pfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("P", Ptentfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", SmooFact.get()), false);
+      TEST_EQUALITY(coarseLevel2->IsAvailable("R", Rfact.get()), false);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", Pfact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", Ptentfact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+      TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", Rfact.get()), 0);
+      RCP<Matrix> P2 = coarseLevel2->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R2 = coarseLevel2->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P2->getGlobalNumRows(), 21);
+      TEST_EQUALITY(P2->getGlobalNumCols(), 7);
+      TEST_EQUALITY(R2->getGlobalNumRows(), 7);
+      TEST_EQUALITY(R2->getGlobalNumCols(), 21);
+
+      Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P1, true, *P1, false, out);
+
+      if (PtentTPtent->haveGlobalConstants()) TEST_EQUALITY(PtentTPtent->getGlobalMaxNumRowEntries() - 3 < 1e-12, true);
+      if (P1->haveGlobalConstants()) TEST_EQUALITY(P1->getGlobalMaxNumRowEntries() - 2 < 1e-12, true);
+      if (P2->haveGlobalConstants()) TEST_EQUALITY(P2->getGlobalMaxNumRowEntries() - 2 < 1e-12, true);
+
+      // Define RHS
+      RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+      RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+      X->putScalar(1.0);
+      X->norm2(norms);
+      if (comm->getRank() == 0)
+        out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+      Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+      // Use AMG directly as an iterative method
       {
-        if (run == 0) lib = Xpetra::UseEpetra;
-        else lib = Xpetra::UseTpetra;
-
-        // generate problem
-        LocalOrdinal maxLevels = 3;
-        LocalOrdinal its=10;
-        GlobalOrdinal nEle = 63;
-        const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-        Teuchos::ParameterList matrixParameters;
-        matrixParameters.set("nx",nEle);
-
-        RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap,MultiVector>("Laplace1D", map, matrixParameters);
-        RCP<Matrix> Op = Pr->BuildMatrix();
-
-        // build nullspace
-        RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-        nullSpace->putScalar( (Scalar) 1.0);
-        Teuchos::Array<magnitude_type> norms(1);
-        nullSpace->norm1(norms);
-        if (comm->getRank() == 0)
-          out << "||NS|| = " << norms[0] << std::endl;
-
-        // fill hierarchy
-        RCP<Hierarchy> H = rcp( new Hierarchy() );
-        H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-        RCP<Level> Finest = H->GetLevel();
-        Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-        Finest->Set("A",Op);                      // set fine level matrix
-        Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-
-        // define transfer operators
-        RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-        UncoupledAggFact->SetMinNodesPerAggregate(3);
-        UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-        UncoupledAggFact->SetOrdering("natural");
-
-        RCP<TentativePFactory> Ptentfact = rcp(new TentativePFactory());
-        RCP<SaPFactory>        Pfact = rcp( new SaPFactory());
-        RCP<Factory>      Rfact = rcp( new TransPFactory() );
-        RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-        H->SetMaxCoarseSize(1);
-
-        // setup smoothers
-        Teuchos::ParameterList smootherParamList;
-        smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-        smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-        smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-        RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-        RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-        Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-        RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-        FactoryManager M;
-        M.SetKokkosRefactor(false);
-        M.SetFactory("P", Pfact);
-        M.SetFactory("R", Rfact);
-        M.SetFactory("A", Acfact);
-        M.SetFactory("Ptent", Ptentfact);
-        M.SetFactory("Aggregates", UncoupledAggFact);
-        M.SetFactory("Smoother", SmooFact);
-        M.SetFactory("CoarseSolver", coarseSolveFact);
-
-        H->Setup(M, 0, maxLevels);
-
-        // test some basic multigrid data
-        RCP<Level> coarseLevel = H->GetLevel(1);
-        TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("R",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("A",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->IsAvailable("R",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsRequested("A",Acfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel->IsAvailable("A",Acfact.get()), false);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Pfact.get()), 0);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Ptentfact.get()), 0);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("R",Rfact.get()), 0);
-        TEST_EQUALITY(coarseLevel->GetKeepFlag("A",Acfact.get()), 0);
-        RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P1->getGlobalNumRows(), 63);
-        TEST_EQUALITY(P1->getGlobalNumCols(), 21);
-        TEST_EQUALITY(R1->getGlobalNumRows(), 21);
-        TEST_EQUALITY(R1->getGlobalNumCols(), 63);
-        RCP<Level> coarseLevel2 = H->GetLevel(2);
-        TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("R",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("A",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",MueLu::NoFactory::get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("R",MueLu::NoFactory::get()), true);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsRequested("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",Pfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("P",Ptentfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",SmooFact.get()), false);
-        TEST_EQUALITY(coarseLevel2->IsAvailable("R",Rfact.get()), false);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",Pfact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",Ptentfact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-        TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",Rfact.get()), 0);
-        RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P2->getGlobalNumRows(), 21);
-        TEST_EQUALITY(P2->getGlobalNumCols(), 7);
-        TEST_EQUALITY(R2->getGlobalNumRows(), 7);
-        TEST_EQUALITY(R2->getGlobalNumCols(), 21);
-
-        Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*P1,true,*P1,false,out);
-
-        if(PtentTPtent->haveGlobalConstants())  TEST_EQUALITY(PtentTPtent->getGlobalMaxNumRowEntries()-3<1e-12, true);
-        if(P1->haveGlobalConstants())           TEST_EQUALITY(P1->getGlobalMaxNumRowEntries()-2<1e-12, true);
-        if(P2->haveGlobalConstants())           TEST_EQUALITY(P2->getGlobalMaxNumRowEntries()-2<1e-12, true);
-
-        // Define RHS
-        RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
-
-        X->putScalar(1.0);
+        X->putScalar((Scalar)0.0);
+
+        H->Iterate(*RHS, *X, its);
+
         X->norm2(norms);
         if (comm->getRank() == 0)
-          out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+          out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+        results[run] = norms[0];
+      }
+    }
 
-        Op->apply(*X,*RHS,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
+    TEST_EQUALITY(results[0] - results[1] < 1e-10, true);  // check results of EPETRA vs TPETRA
+  }                                                        // comm->getSize == 1
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
+#endif
 
-        // Use AMG directly as an iterative method
-        {
-          X->putScalar( (Scalar) 0.0);
+}  // SaPFactory_EpetraVsTpetra
 
-          H->Iterate(*RHS,*X,its);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, ConstrainRowOptimalScalarPDE, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
 
-          X->norm2(norms);
-          if (comm->getRank() == 0)
-            out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-          results[run] = norms[0];
-        }
-      }
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero(), one = STS::one();
 
-      TEST_EQUALITY(results[0] - results[1] < 1e-10, true); // check results of EPETRA vs TPETRA
-    } // comm->getSize == 1
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
-#   endif
-
-  } //SaPFactory_EpetraVsTpetra
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory, ConstrainRowOptimalScalarPDE, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero(), one = STS::one();
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // Don't test for complex - matrix reader won't work
-    if (STS::isComplex) {success=true; return;}
-
-    RCP<Matrix> P;
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-    P =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/SaP_constrainTest_P.mat", lib, comm);
-
-    RCP<SaPFactory> sapFactory = rcp(new SaPFactory);
-    sapFactory->optimalSatisfyPConstraintsForScalarPDEs( P );
-
-    // check that row sums are all one by checking the norm of the vector
-    // note: optimalSatisfyPConstraintsForScalarPDEs preserves row sum of original P (one in this case),
-    //       but SatisfyPConstraints normalizes each row sum to one.
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(), 1);
-    RCP<MultiVector> Bfact = MultiVectorFactory::Build(P->getRangeMap(),  1);
-    X->putScalar(one);
-    P->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
-    Array<typename STS::magnitudeType> norms(1);
-    Bfact->norm2(norms);
-    out << "|| B_factory ones || = " << norms[0] << std::endl;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-    TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(6.0), 100*TMT::eps());
-
-    // check that the min and max of each row are in [0,1]
-    bool lowerViolation = false;
-    bool upperViolation = false;
-    for (size_t i = 0; i < (size_t)(P->getRowMap()->getLocalNumElements()); i++) {
-      Teuchos::ArrayView<const LO> indices;
-      Teuchos::ArrayView<const SC> vals;
-      P->getLocalRowView((LO) i, indices, vals);
-      size_t nnz = indices.size();
-      for (size_t j = 0; j < nnz; j++)  {
-        if (Teuchos::ScalarTraits<SC>::real(vals[j]) < Teuchos::ScalarTraits<SC>::real(zero)) lowerViolation = true;
-        if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(one))  upperViolation = true;
-        //if (STS::magnitude(vals[j]) < STS::magnitude(zero)) lowerViolation = true;
-        //if (STS::magnitude(vals[j]) > STS::magnitude(one))  upperViolation = true;
-      }
-    }
-    TEST_EQUALITY(lowerViolation, false);
-    TEST_EQUALITY(upperViolation, false);
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  // Don't test for complex - matrix reader won't work
+  if (STS::isComplex) {
+    success = true;
+    return;
+  }
 
+  RCP<Matrix> P;
+  Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+  P                         = Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/SaP_constrainTest_P.mat", lib, comm);
+
+  RCP<SaPFactory> sapFactory = rcp(new SaPFactory);
+  sapFactory->optimalSatisfyPConstraintsForScalarPDEs(P);
+
+  // check that row sums are all one by checking the norm of the vector
+  // note: optimalSatisfyPConstraintsForScalarPDEs preserves row sum of original P (one in this case),
+  //       but SatisfyPConstraints normalizes each row sum to one.
+  RCP<MultiVector> X     = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> Bfact = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  X->putScalar(one);
+  P->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
+  Array<typename STS::magnitudeType> norms(1);
+  Bfact->norm2(norms);
+  out << "|| B_factory ones || = " << norms[0] << std::endl;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(6.0), 100 * TMT::eps());
+
+  // check that the min and max of each row are in [0,1]
+  bool lowerViolation = false;
+  bool upperViolation = false;
+  for (size_t i = 0; i < (size_t)(P->getRowMap()->getLocalNumElements()); i++) {
+    Teuchos::ArrayView<const LO> indices;
+    Teuchos::ArrayView<const SC> vals;
+    P->getLocalRowView((LO)i, indices, vals);
+    size_t nnz = indices.size();
+    for (size_t j = 0; j < nnz; j++) {
+      if (Teuchos::ScalarTraits<SC>::real(vals[j]) < Teuchos::ScalarTraits<SC>::real(zero)) lowerViolation = true;
+      if (Teuchos::ScalarTraits<SC>::real(vals[j]) > Teuchos::ScalarTraits<SC>::real(one)) upperViolation = true;
+      // if (STS::magnitude(vals[j]) < STS::magnitude(zero)) lowerViolation = true;
+      // if (STS::magnitude(vals[j]) > STS::magnitude(one))  upperViolation = true;
+    }
+  }
+  TEST_EQUALITY(lowerViolation, false);
+  TEST_EQUALITY(upperViolation, false);
 
-  } //SaPFactory_ConstrainRowOptimalScalarPDE
+}  // SaPFactory_ConstrainRowOptimalScalarPDE
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, Test0, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, EpetraVsTpetra, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, ConstrainRowOptimalScalarPDE, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, Test0, SC, LO, GO, Node)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, EpetraVsTpetra, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory, ConstrainRowOptimalScalarPDE, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ScaledNullspaceFactory.cpp b/packages/muelu/test/unit_tests/ScaledNullspaceFactory.cpp
index afd544b5b87b..4222da189200 100644
--- a/packages/muelu/test/unit_tests/ScaledNullspaceFactory.cpp
+++ b/packages/muelu/test/unit_tests/ScaledNullspaceFactory.cpp
@@ -65,129 +65,126 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ScaledNullspaceFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    Teuchos::Array<magnitude_type> results(2);
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-    // Do not run this test for Epetra
-    if (lib == Xpetra::UseEpetra) {
-      TEST_EQUALITY(1,1);
-      return;
-    }
-
-    // Build a Matrix
-    GlobalOrdinal nEle = 20;
-    const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-    Teuchos::ParameterList matrixParameters;
-    matrixParameters.set("nx",nEle);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap,MultiVector>("Laplace1D", map, matrixParameters);
-    RCP<Matrix> Op = Pr->BuildMatrix();
-    
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0)
-      out << "||NS|| = " << norms[0] << std::endl;
-    int maxLevels = 3;
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-    
-    // define transfer operators
-    RCP<NullspaceFactory> NSFact = rcp(new NullspaceFactory());
-    RCP<ScaledNullspaceFactory> SNSFact = rcp(new ScaledNullspaceFactory());
-
-    RCP<TentativePFactory> Ptent1 = rcp(new TentativePFactory());
-    NSFact->SetFactory("Nullspace",Ptent1);
-    RCP<TentativePFactory> Ptent2 = rcp(new TentativePFactory());
-    Teuchos::ParameterList pt2_list; pt2_list.set("Nullspace name","Scaled Nullspace");
-    Ptent2->SetParameterList(pt2_list);
-    
-    RCP<Factory>      Rfact = rcp( new TransPFactory() );
-    Rfact->SetFactory("P",Ptent2);
-    RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-    H->SetMaxCoarseSize(1);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-    
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-    
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Ptent1);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("Nullspace", NSFact);
-    M.SetFactory("Scaled Nullspace", SNSFact);
-    M.SetFactory("Ptent", Ptent1);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-    
-    H->Setup(M, 0, maxLevels);
-    
-    // test some basic multigrid data
-    RCP<Level> coarseLevel = H->GetLevel(1);
-    TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("A",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-    //    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    //    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",Ptent1.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("A",Acfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",Ptent1.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",SmooFact.get()), false);
-    //    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("A",Acfact.get()), false);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Ptent1.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-    //    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",Rfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("A",Acfact.get()), 0);
-
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ScaledNullspaceFactory, Test0, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  Teuchos::Array<magnitude_type> results(2);
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  // Do not run this test for Epetra
+  if (lib == Xpetra::UseEpetra) {
+    TEST_EQUALITY(1, 1);
+    return;
   }
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ScaledNullspaceFactory, Test0, SC, LO, GO, Node)
+  // Build a Matrix
+  GlobalOrdinal nEle       = 20;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", nEle);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
+  RCP<Matrix> Op = Pr->BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0)
+    out << "||NS|| = " << norms[0] << std::endl;
+  int maxLevels = 3;
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                 // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+
+  // define transfer operators
+  RCP<NullspaceFactory> NSFact        = rcp(new NullspaceFactory());
+  RCP<ScaledNullspaceFactory> SNSFact = rcp(new ScaledNullspaceFactory());
+
+  RCP<TentativePFactory> Ptent1 = rcp(new TentativePFactory());
+  NSFact->SetFactory("Nullspace", Ptent1);
+  RCP<TentativePFactory> Ptent2 = rcp(new TentativePFactory());
+  Teuchos::ParameterList pt2_list;
+  pt2_list.set("Nullspace name", "Scaled Nullspace");
+  Ptent2->SetParameterList(pt2_list);
+
+  RCP<Factory> Rfact = rcp(new TransPFactory());
+  Rfact->SetFactory("P", Ptent2);
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Ptent1);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("Nullspace", NSFact);
+  M.SetFactory("Scaled Nullspace", SNSFact);
+  M.SetFactory("Ptent", Ptent1);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  // test some basic multigrid data
+  RCP<Level> coarseLevel = H->GetLevel(1);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  //    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  //    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", Ptent1.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", Acfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", Ptent1.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", SmooFact.get()), false);
+  //    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("A", Acfact.get()), false);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Ptent1.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+  //    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", Rfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("A", Acfact.get()), 0);
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ScaledNullspaceFactory, Test0, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SchurComplementFactory.cpp b/packages/muelu/test/unit_tests/SchurComplementFactory.cpp
index 76b199cd3572..8c47946c40f9 100644
--- a/packages/muelu/test/unit_tests/SchurComplementFactory.cpp
+++ b/packages/muelu/test/unit_tests/SchurComplementFactory.cpp
@@ -63,359 +63,362 @@
 
 namespace MueLuTests {
 
-  /////////////////////////
-  // helper function
+/////////////////////////
+// helper function
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlocked2x2MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
-  typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlocked2x2MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
-  typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-  typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
 
   std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(2, Teuchos::null);
-  maps[0] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-  maps[1] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-  RCP<Matrix> A00 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
-  RCP<Matrix> A01 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A10 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A11 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
+  maps[0]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[1]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  RCP<Matrix> A00                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A01                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A10                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A11                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
 
   // create map extractor
   // To generate the Thyra style map extractor we do not need a full map but only the
   // information about the Map details (i.e. lib and indexBase). We can extract this
   // information from maps[0]
-  Teuchos::RCP<const MapExtractor > rgMapExtractor =
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
       Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-  Teuchos::RCP<const MapExtractor > doMapExtractor =
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
       Teuchos::rcp(new MapExtractor(maps[0], maps, true));
   // build blocked operator
-  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,5));
-  bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(0),A00);
-  bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(1),A01);
-  bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(0),A10);
-  bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(1),A11);
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 5));
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(0), A00);
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(1), A01);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(0), A10);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(1), A11);
   bop->fillComplete();
   return bop;
 }
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlocked3x3MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
-  typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlocked3x3MatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
   typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
-  typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-  typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
 
   std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
-  maps[0] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-  maps[1] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-  maps[2] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-  RCP<Matrix> A00 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
-  RCP<Matrix> A01 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A10 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A11 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
-  RCP<Matrix> A12 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A21 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
-  RCP<Matrix> A22 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
+  maps[0]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[1]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[2]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  RCP<Matrix> A00                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A01                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A10                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A11                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A12                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A21                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A22                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
 
   // create map extractor
   // To generate the Thyra style map extractor we do not need a full map but only the
   // information about the Map details (i.e. lib and indexBase). We can extract this
   // information from maps[0]
-  Teuchos::RCP<const MapExtractor > rgMapExtractor =
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
       Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-  Teuchos::RCP<const MapExtractor > doMapExtractor =
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
       Teuchos::rcp(new MapExtractor(maps[0], maps, true));
   // build blocked operator
-  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,5));
-  bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(0),A00);
-  bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(1),A01);
-  bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(0),A10);
-  bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(1),A11);
-  bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(2),A12);
-  bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(1),A21);
-  bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(2),A22);
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 5));
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(0), A00);
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(1), A01);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(0), A10);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(1), A11);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(2), A12);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(1), A21);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(2), A22);
   bop->fillComplete();
   return bop;
 }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructor2x2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::CreateBlocked2x2MatrixThyra<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*comm, lib);
-    Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-    Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-    Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<SchurComplementFactory> schurFact = rcp( new SchurComplementFactory() );
-    schurFact->SetFactory("A",MueLu::NoFactory::getRCP());
-    schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-    // request SchurComplement operator
-    level.Request("A", schurFact.get());
-
-    // generate Schur complement operator
-    schurFact->Build(level);
-
-    RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
-    TEST_EQUALITY(sOp.is_null(), false);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-    TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
-
-    /*    size_t localEntries = 0;
-    if(comm->getSize() > 2 && comm->getRank() == 0) localEntries = 299;
-    else if(comm->getSize() > 2 && comm->getRank() == comm->getSize()-1) localEntries = 299;
-    else localEntries = 300;
-    if(comm->getSize() == 2) localEntries = 299;
-    if(comm->getSize() == 1) localEntries = 298;*/
-    //    TEST_EQUALITY(sOp->getLocalNumEntries(), localEntries);
-    //    TEST_EQUALITY(sOp->getGlobalNumEntries(),  Teuchos::as<size_t>(comm->getSize() * 300 - 2));
-
-    RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(),true);
-    sOp->getLocalDiagCopy(*v);
-    Teuchos::ArrayRCP<const Scalar> vdata = v->getData(0);
-    bool bCheck = true;
-    for(int i=0; i<100; i++) if(vdata[i] != Teuchos::as<Scalar>(3.75)) bCheck = false;
-    TEST_EQUALITY(bCheck, true);
-
-    // check effect of scaling
-
-    schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0e12)));
-    schurFact->Build(level);
-
-    sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
-    TEST_EQUALITY(sOp.is_null(), false);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-    TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
-
-    /*localEntries = 0;
-    if(comm->getSize() > 2 && comm->getRank() == 0) localEntries = 299;
-    else if(comm->getSize() > 2 && comm->getRank() == comm->getSize()-1) localEntries = 299;
-    else localEntries = 300;
-    if(comm->getSize() == 2) localEntries = 299;
-    if(comm->getSize() == 1) localEntries = 298;*/
-
-    //    TEST_EQUALITY(sOp->getLocalNumEntries(), localEntries);
-    //    TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 300 - 2));
-
-    /*v = VectorFactory::Build(sOp->getRangeMap(),true);
-    sOp->getLocalDiagCopy(*v);
-    vdata = v->getData(0);
-    bCheck = true;
-    for(int i=0; i<100; i++) if(Teuchos::ScalarTraits<Scalar>::magnitude(vdata[i] - Teuchos::as<Scalar>(4.0)) < 1.0e-8) { bCheck = false; std::cout << vdata[i] << std::endl; }
-    TEST_EQUALITY(bCheck, true);*/
-  } //Constructor2x2
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructorXpetra2x2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    Teuchos::RCP<BlockedCrsMatrix> bop = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib, 2, comm);
-    Teuchos::RCP<Matrix> A = Teuchos::rcp_dynamic_cast<Matrix>(bop);
-
-    Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<SchurComplementFactory> schurFact = rcp( new SchurComplementFactory() );
-    schurFact->SetFactory("A",MueLu::NoFactory::getRCP());
-    schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-    // request SchurComplement operator
-    level.Request("A", schurFact.get());
-
-    // generate Schur complement operator
-    schurFact->Build(level);
-
-    RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
-    TEST_EQUALITY(sOp.is_null(), false);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 10 + 5);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 10 + 5);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
-    TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
-
-    TEST_EQUALITY(sOp->getLocalNumEntries(), 5);
-    TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 5));
-
-    RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(),true);
-    sOp->getLocalDiagCopy(*v);
-    Teuchos::ArrayRCP<const Scalar> vdata = v->getData(0);
-    bool bCheck = true;
-    for(int i=0; i<5; i++) if(vdata[i] != Teuchos::as<Scalar>(2.0)) bCheck = false;
-    TEST_EQUALITY(bCheck, true);
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructor3x3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::CreateBlocked3x3MatrixThyra<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*comm, lib);
-    Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-    Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-    Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<InverseApproximationFactory> approxInvFactory = rcp( new InverseApproximationFactory() );
-    approxInvFactory->SetFactory("A",MueLu::NoFactory::getRCP());
-    approxInvFactory->SetParameter("inverse: fixing", Teuchos::ParameterEntry(true));
-
-    RCP<SchurComplementFactory> schurFact = rcp( new SchurComplementFactory() );
-    schurFact->SetFactory("A",MueLu::NoFactory::getRCP());
-    schurFact->SetFactory("Ainv", approxInvFactory);
-    schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.5)));
-
-    // request SchurComplement operator
-    level.Request("A", schurFact.get());
-
-    // generate Schur complement operator
-    TEST_THROW(schurFact->Build(level),MueLu::Exceptions::RuntimeError);
-
-    // generate reordered blocked operator
-    RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 0 [1 2] ]");
-    RCP<const ReorderedBlockedCrsMatrix> brop =
-        Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(buildReorderedBlockedCrsMatrix(brm, bop));
-    RCP<const Matrix> rA = Teuchos::rcp_dynamic_cast<const Matrix>(brop);
-    TEST_EQUALITY(brop->Rows(),2);
-    TEST_EQUALITY(brop->Cols(),2);
-    level.Set("A", Teuchos::rcp_const_cast<Matrix>(rA));
-
-    // try to build Schur complement operator again
-    schurFact->Build(level);
-
-    // consecutive Thyra GIDs with offset comm->getSize() * 100!
-    RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
-
-    TEST_EQUALITY(sOp.is_null(), false);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinAllGlobalIndex(), comm->getSize() * 100);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinAllGlobalIndex(), comm->getSize() * 100);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
-
-    RCP<BlockedCrsMatrix> bsOp = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp);
-    TEST_EQUALITY(bsOp == Teuchos::null, false);
-    TEST_EQUALITY(bsOp->Rows(), 2);
-    TEST_EQUALITY(bsOp->Cols(), 2);
-    TEST_EQUALITY(bsOp->getRangeMapExtractor()->getThyraMode(), true);
-    TEST_EQUALITY(bsOp->getDomainMapExtractor()->getThyraMode(), true);
-
-    RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(),true);
-    sOp->getLocalDiagCopy(*v);
-    RCP<BlockedVector> bv = Teuchos::rcp_dynamic_cast<BlockedVector>(v);
-    TEST_EQUALITY(bv->getBlockedMap()->getNumMaps(),2);
-
-    RCP<MultiVector> mergedv = bv->Merge();
-    Teuchos::ArrayRCP<const Scalar> vdata = mergedv->getData(0);
-    bool bCheck = true;
-    for(int i=0; i<100; i++) if(vdata[i] != Teuchos::as<Scalar>(3.50)) bCheck = false;
-    for(int i=100; i<200; i++) if(vdata[i] != Teuchos::as<Scalar>(4.00)) bCheck = false;
-    TEST_EQUALITY(bCheck, true);
-  } // SchurConstructor3x3
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructorXpetra4x4, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-  #   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib, 4, comm);
-    Teuchos::RCP<const Matrix> cA = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-    Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(cA);
-
-    Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<SchurComplementFactory> schurFact = rcp( new SchurComplementFactory() );
-    schurFact->SetFactory("A",MueLu::NoFactory::getRCP());
-    schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-    // request SchurComplement operator
-    level.Request("A", schurFact.get());
-
-    // generate Schur complement operator
-    TEST_THROW(schurFact->Build(level),MueLu::Exceptions::RuntimeError);
-
-    // generate reordered blocked operator
-    RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [0 3] [1 2]]");
-    RCP<const ReorderedBlockedCrsMatrix> brop =
-        Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(buildReorderedBlockedCrsMatrix(brm, bop));
-    RCP<const Matrix> rA = Teuchos::rcp_dynamic_cast<const Matrix>(brop);
-    TEST_EQUALITY(brop->Rows(),2);
-    TEST_EQUALITY(brop->Cols(),2);
-    level.Set("A", Teuchos::rcp_const_cast<Matrix>(rA));
-
-    // generate Schur complement operator
-    schurFact->Build(level);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructor2x2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::CreateBlocked2x2MatrixThyra<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*comm, lib);
+  Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+  Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+  Level level;
+  TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
+  schurFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+  // request SchurComplement operator
+  level.Request("A", schurFact.get());
+
+  // generate Schur complement operator
+  schurFact->Build(level);
+
+  RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
+  TEST_EQUALITY(sOp.is_null(), false);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+  TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
+
+  /*    size_t localEntries = 0;
+  if(comm->getSize() > 2 && comm->getRank() == 0) localEntries = 299;
+  else if(comm->getSize() > 2 && comm->getRank() == comm->getSize()-1) localEntries = 299;
+  else localEntries = 300;
+  if(comm->getSize() == 2) localEntries = 299;
+  if(comm->getSize() == 1) localEntries = 298;*/
+  //    TEST_EQUALITY(sOp->getLocalNumEntries(), localEntries);
+  //    TEST_EQUALITY(sOp->getGlobalNumEntries(),  Teuchos::as<size_t>(comm->getSize() * 300 - 2));
+
+  RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(), true);
+  sOp->getLocalDiagCopy(*v);
+  Teuchos::ArrayRCP<const Scalar> vdata = v->getData(0);
+  bool bCheck                           = true;
+  for (int i = 0; i < 100; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(3.75)) bCheck = false;
+  TEST_EQUALITY(bCheck, true);
+
+  // check effect of scaling
+
+  schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0e12)));
+  schurFact->Build(level);
+
+  sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
+  TEST_EQUALITY(sOp.is_null(), false);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+  TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
+
+  /*localEntries = 0;
+  if(comm->getSize() > 2 && comm->getRank() == 0) localEntries = 299;
+  else if(comm->getSize() > 2 && comm->getRank() == comm->getSize()-1) localEntries = 299;
+  else localEntries = 300;
+  if(comm->getSize() == 2) localEntries = 299;
+  if(comm->getSize() == 1) localEntries = 298;*/
+
+  //    TEST_EQUALITY(sOp->getLocalNumEntries(), localEntries);
+  //    TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 300 - 2));
+
+  /*v = VectorFactory::Build(sOp->getRangeMap(),true);
+  sOp->getLocalDiagCopy(*v);
+  vdata = v->getData(0);
+  bCheck = true;
+  for(int i=0; i<100; i++) if(Teuchos::ScalarTraits<Scalar>::magnitude(vdata[i] - Teuchos::as<Scalar>(4.0)) < 1.0e-8) { bCheck = false; std::cout << vdata[i] << std::endl; }
+  TEST_EQUALITY(bCheck, true);*/
+}  // Constructor2x2
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructorXpetra2x2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  Teuchos::RCP<BlockedCrsMatrix> bop = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 2, comm);
+  Teuchos::RCP<Matrix> A             = Teuchos::rcp_dynamic_cast<Matrix>(bop);
+
+  Level level;
+  TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
+  schurFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+  // request SchurComplement operator
+  level.Request("A", schurFact.get());
+
+  // generate Schur complement operator
+  schurFact->Build(level);
+
+  RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
+  TEST_EQUALITY(sOp.is_null(), false);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 10 + 5);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 10 + 5);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
+  TEST_EQUALITY(Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp), Teuchos::null);
+
+  TEST_EQUALITY(sOp->getLocalNumEntries(), 5);
+  TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 5));
+
+  RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(), true);
+  sOp->getLocalDiagCopy(*v);
+  Teuchos::ArrayRCP<const Scalar> vdata = v->getData(0);
+  bool bCheck                           = true;
+  for (int i = 0; i < 5; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(2.0)) bCheck = false;
+  TEST_EQUALITY(bCheck, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructor3x3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::CreateBlocked3x3MatrixThyra<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*comm, lib);
+  Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+  Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+  Level level;
+  TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<InverseApproximationFactory> approxInvFactory = rcp(new InverseApproximationFactory());
+  approxInvFactory->SetFactory("A", MueLu::NoFactory::getRCP());
+  approxInvFactory->SetParameter("inverse: fixing", Teuchos::ParameterEntry(true));
+
+  RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
+  schurFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  schurFact->SetFactory("Ainv", approxInvFactory);
+  schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.5)));
+
+  // request SchurComplement operator
+  level.Request("A", schurFact.get());
+
+  // generate Schur complement operator
+  TEST_THROW(schurFact->Build(level), MueLu::Exceptions::RuntimeError);
+
+  // generate reordered blocked operator
+  RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 0 [1 2] ]");
+  RCP<const ReorderedBlockedCrsMatrix> brop =
+      Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(buildReorderedBlockedCrsMatrix(brm, bop));
+  RCP<const Matrix> rA = Teuchos::rcp_dynamic_cast<const Matrix>(brop);
+  TEST_EQUALITY(brop->Rows(), 2);
+  TEST_EQUALITY(brop->Cols(), 2);
+  level.Set("A", Teuchos::rcp_const_cast<Matrix>(rA));
+
+  // try to build Schur complement operator again
+  schurFact->Build(level);
+
+  // consecutive Thyra GIDs with offset comm->getSize() * 100!
+  RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
+
+  TEST_EQUALITY(sOp.is_null(), false);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinAllGlobalIndex(), comm->getSize() * 100);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinAllGlobalIndex(), comm->getSize() * 100);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
+
+  RCP<BlockedCrsMatrix> bsOp = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(sOp);
+  TEST_EQUALITY(bsOp == Teuchos::null, false);
+  TEST_EQUALITY(bsOp->Rows(), 2);
+  TEST_EQUALITY(bsOp->Cols(), 2);
+  TEST_EQUALITY(bsOp->getRangeMapExtractor()->getThyraMode(), true);
+  TEST_EQUALITY(bsOp->getDomainMapExtractor()->getThyraMode(), true);
+
+  RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(), true);
+  sOp->getLocalDiagCopy(*v);
+  RCP<BlockedVector> bv = Teuchos::rcp_dynamic_cast<BlockedVector>(v);
+  TEST_EQUALITY(bv->getBlockedMap()->getNumMaps(), 2);
+
+  RCP<MultiVector> mergedv              = bv->Merge();
+  Teuchos::ArrayRCP<const Scalar> vdata = mergedv->getData(0);
+  bool bCheck                           = true;
+  for (int i = 0; i < 100; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(3.50)) bCheck = false;
+  for (int i = 100; i < 200; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(4.00)) bCheck = false;
+  TEST_EQUALITY(bCheck, true);
+}  // SchurConstructor3x3
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SchurComplementFactory, SchurConstructorXpetra4x4, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  Teuchos::RCP<const BlockedCrsMatrix> bop = MueLuTests::TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 4, comm);
+  Teuchos::RCP<const Matrix> cA            = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+  Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(cA);
+
+  Level level;
+  TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<SchurComplementFactory> schurFact = rcp(new SchurComplementFactory());
+  schurFact->SetFactory("A", MueLu::NoFactory::getRCP());
+  schurFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+  // request SchurComplement operator
+  level.Request("A", schurFact.get());
+
+  // generate Schur complement operator
+  TEST_THROW(schurFact->Build(level), MueLu::Exceptions::RuntimeError);
+
+  // generate reordered blocked operator
+  RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [0 3] [1 2]]");
+  RCP<const ReorderedBlockedCrsMatrix> brop =
+      Teuchos::rcp_dynamic_cast<const ReorderedBlockedCrsMatrix>(buildReorderedBlockedCrsMatrix(brm, bop));
+  RCP<const Matrix> rA = Teuchos::rcp_dynamic_cast<const Matrix>(brop);
+  TEST_EQUALITY(brop->Rows(), 2);
+  TEST_EQUALITY(brop->Cols(), 2);
+  level.Set("A", Teuchos::rcp_const_cast<Matrix>(rA));
+
+  // generate Schur complement operator
+  schurFact->Build(level);
 #if 1
-    RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
-
-    TEST_EQUALITY(sOp.is_null(), false);
-    TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-    TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-    TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-    TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-
-    TEST_EQUALITY(sOp->getLocalNumEntries(), 15);
-    TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 15));
-
-    RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(),true);
-    sOp->getLocalDiagCopy(*v);
-    RCP<BlockedVector> bv = Teuchos::rcp_dynamic_cast<BlockedVector>(v);
-    TEST_EQUALITY(bv.is_null(),false);
-    TEST_EQUALITY(bv->getBlockedMap()->getNumMaps(),2);
-
-    RCP<MultiVector> mergedv = bv->Merge();
-    Teuchos::ArrayRCP<const Scalar> vdata = mergedv->getData(0);
-    bool bCheck = true;
-    for(int i=0; i<5;  i++) if(vdata[i] != Teuchos::as<Scalar>(2.0)) bCheck = false;
-    for(int i=5; i<15; i++) if(vdata[i] != Teuchos::as<Scalar>(3.0)) bCheck = false;
-    TEST_EQUALITY(bCheck, true);
+  RCP<Matrix> sOp = level.Get<RCP<Matrix> >("A", schurFact.get());
+
+  TEST_EQUALITY(sOp.is_null(), false);
+  TEST_EQUALITY(sOp->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+  TEST_EQUALITY(sOp->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+  TEST_EQUALITY(sOp->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+  TEST_EQUALITY(sOp->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+
+  TEST_EQUALITY(sOp->getLocalNumEntries(), 15);
+  TEST_EQUALITY(sOp->getGlobalNumEntries(), Teuchos::as<size_t>(comm->getSize() * 15));
+
+  RCP<Vector> v = VectorFactory::Build(sOp->getRangeMap(), true);
+  sOp->getLocalDiagCopy(*v);
+  RCP<BlockedVector> bv = Teuchos::rcp_dynamic_cast<BlockedVector>(v);
+  TEST_EQUALITY(bv.is_null(), false);
+  TEST_EQUALITY(bv->getBlockedMap()->getNumMaps(), 2);
+
+  RCP<MultiVector> mergedv              = bv->Merge();
+  Teuchos::ArrayRCP<const Scalar> vdata = mergedv->getData(0);
+  bool bCheck                           = true;
+  for (int i = 0; i < 5; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(2.0)) bCheck = false;
+  for (int i = 5; i < 15; i++)
+    if (vdata[i] != Teuchos::as<Scalar>(3.0)) bCheck = false;
+  TEST_EQUALITY(bCheck, true);
 #endif
-  }
-
-
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructor2x2, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructorXpetra2x2, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructor3x3, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructorXpetra4x4, SC, LO, GO, Node) \
-
-#include <MueLu_ETI_4arg.hpp>
 }
 
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructor2x2, SC, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructorXpetra2x2, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructor3x3, SC, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SchurComplementFactory, SchurConstructorXpetra4x4, SC, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SegregatedAFactory.cpp b/packages/muelu/test/unit_tests/SegregatedAFactory.cpp
index cc15f931bcbc..1539f5f9caac 100644
--- a/packages/muelu/test/unit_tests/SegregatedAFactory.cpp
+++ b/packages/muelu/test/unit_tests/SegregatedAFactory.cpp
@@ -64,89 +64,88 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SegregatedAFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SegregatedAFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST            = Teuchos::ScalarTraits<SC>;
+  using magnitude_type = typename TST::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // test with simple example matrix and segregate first 4 rows from rest
+  //     x x x x                x x x x
+  //     x x x x                x x x x
+  //     x x x x x x            x x x x
+  //     x x x x x x      ==>   x x x x
+  //         x x x x x                 x x x
+  //         x x x x x                 x x x
+  //             x x x                 x x x
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST = Teuchos::ScalarTraits<SC>;
-    using magnitude_type = typename TST::magnitudeType;
-    using TMT = Teuchos::ScalarTraits<magnitude_type>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // test with simple example matrix and segregate first 4 rows from rest
-    //     x x x x                x x x x
-    //     x x x x                x x x x
-    //     x x x x x x            x x x x
-    //     x x x x x x      ==>   x x x x
-    //         x x x x x                 x x x
-    //         x x x x x                 x x x
-    //             x x x                 x x x
-    {
-      RCP<Map> rowMap = MapFactory::Build(lib, 7, 0, comm);
-      RCP<CrsGraph> graph = CrsGraphFactory::Build(rowMap, 6);
-
-      graph->insertGlobalIndices(0, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3));
-      graph->insertGlobalIndices(1, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3));
-      graph->insertGlobalIndices(2, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3, 4, 5));
-      graph->insertGlobalIndices(3, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3, 4, 5));
-      graph->insertGlobalIndices(4, Teuchos::tuple<GlobalOrdinal>(2, 3, 4, 5, 6));
-      graph->insertGlobalIndices(5, Teuchos::tuple<GlobalOrdinal>(2, 3, 4, 5, 6));
-      graph->insertGlobalIndices(6, Teuchos::tuple<GlobalOrdinal>(4, 5, 6));
-      graph->fillComplete();
-
-      RCP<Matrix> A = MatrixFactory::Build(graph.getConst());
-      A->setAllToScalar(TST::one());
-      A->fillComplete();
-
-      TEST_EQUALITY(A.is_null(), false);
-      TEST_EQUALITY(A->getGlobalNumEntries(), 33);
-      TEST_FLOATING_EQUALITY(A->getFrobeniusNorm(), 5.744562646538029, 2e1*TMT::eps());
-
-      RCP<Map> map = MapFactory::Build(lib, 4, 0, comm);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      const std::string mapName = "Segregate Map";
-      level.Set(mapName, map);
-
-      TEST_ASSERT(level.IsAvailable("A", MueLu::NoFactory::get()));
-      TEST_ASSERT(level.IsAvailable(mapName, MueLu::NoFactory::get()));
-
-      RCP<SegregatedAFactory> segregateFact = rcp( new SegregatedAFactory() );
-      segregateFact->SetFactory("A", MueLu::NoFactory::getRCP());
-      segregateFact->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
-
-      // request segregated operator
-      level.Request("A", segregateFact.get());
-      segregateFact->Build(level);
-
-      RCP<Matrix> Aout = level.Get<RCP<Matrix> >("A", segregateFact.get());
-
-      // Output (segregated Operator)
-      // Node ID   Global Row  Num Entries(Index,Value)
-      // 0         0           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
-      // 0         1           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
-      // 0         2           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
-      // 0         3           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
-      // 0         4           3 (4, 1)  (5, 1)  (6, 1)
-      // 0         5           3 (4, 1)  (5, 1)  (6, 1)
-      // 0         6           3 (4, 1)  (5, 1)  (6, 1)
-
-      TEST_EQUALITY(Aout.is_null(), false);
-      TEST_EQUALITY(Aout->getGlobalNumEntries(), 25);
-      TEST_FLOATING_EQUALITY(Aout->getFrobeniusNorm(), 5.0, 2e1*TMT::eps());
-    }
+    RCP<Map> rowMap     = MapFactory::Build(lib, 7, 0, comm);
+    RCP<CrsGraph> graph = CrsGraphFactory::Build(rowMap, 6);
+
+    graph->insertGlobalIndices(0, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3));
+    graph->insertGlobalIndices(1, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3));
+    graph->insertGlobalIndices(2, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3, 4, 5));
+    graph->insertGlobalIndices(3, Teuchos::tuple<GlobalOrdinal>(0, 1, 2, 3, 4, 5));
+    graph->insertGlobalIndices(4, Teuchos::tuple<GlobalOrdinal>(2, 3, 4, 5, 6));
+    graph->insertGlobalIndices(5, Teuchos::tuple<GlobalOrdinal>(2, 3, 4, 5, 6));
+    graph->insertGlobalIndices(6, Teuchos::tuple<GlobalOrdinal>(4, 5, 6));
+    graph->fillComplete();
+
+    RCP<Matrix> A = MatrixFactory::Build(graph.getConst());
+    A->setAllToScalar(TST::one());
+    A->fillComplete();
+
+    TEST_EQUALITY(A.is_null(), false);
+    TEST_EQUALITY(A->getGlobalNumEntries(), 33);
+    TEST_FLOATING_EQUALITY(A->getFrobeniusNorm(), 5.744562646538029, 2e1 * TMT::eps());
+
+    RCP<Map> map = MapFactory::Build(lib, 4, 0, comm);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    const std::string mapName = "Segregate Map";
+    level.Set(mapName, map);
+
+    TEST_ASSERT(level.IsAvailable("A", MueLu::NoFactory::get()));
+    TEST_ASSERT(level.IsAvailable(mapName, MueLu::NoFactory::get()));
+
+    RCP<SegregatedAFactory> segregateFact = rcp(new SegregatedAFactory());
+    segregateFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    segregateFact->SetParameter("map: name", Teuchos::ParameterEntry(mapName));
+
+    // request segregated operator
+    level.Request("A", segregateFact.get());
+    segregateFact->Build(level);
+
+    RCP<Matrix> Aout = level.Get<RCP<Matrix> >("A", segregateFact.get());
+
+    // Output (segregated Operator)
+    // Node ID   Global Row  Num Entries(Index,Value)
+    // 0         0           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
+    // 0         1           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
+    // 0         2           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
+    // 0         3           4 (0, 1)  (1, 1)  (2, 1)  (3, 1)
+    // 0         4           3 (4, 1)  (5, 1)  (6, 1)
+    // 0         5           3 (4, 1)  (5, 1)  (6, 1)
+    // 0         6           3 (4, 1)  (5, 1)  (6, 1)
+
+    TEST_EQUALITY(Aout.is_null(), false);
+    TEST_EQUALITY(Aout->getGlobalNumEntries(), 25);
+    TEST_FLOATING_EQUALITY(Aout->getFrobeniusNorm(), 5.0, 2e1 * TMT::eps());
   }
+}
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SegregatedAFactory, Basic, SC, LO, GO, Node) \
+#define MUELU_ETI_GROUP(SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SegregatedAFactory, Basic, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SemiCoarsenPFactory.cpp b/packages/muelu/test/unit_tests/SemiCoarsenPFactory.cpp
index ccc168d78481..9f02a9006fe9 100644
--- a/packages/muelu/test/unit_tests/SemiCoarsenPFactory.cpp
+++ b/packages/muelu/test/unit_tests/SemiCoarsenPFactory.cpp
@@ -56,118 +56,117 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactory, TestSemiCoarsenP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Global Lexicographic";
-    const std::string coupling = "uncoupled";
-    const LO numDimensions = 3;
-    const LO numSize = 5;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(numDimensions);
-    Array<GO> gNodesPerDir(numDimensions);
-    for(int dim = 0; dim < numDimensions; ++dim)
-      gNodesPerDir[dim] = numSize;
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coord_type;
-    typedef Xpetra::MultiVector<coord_type, LO, GO, NO> CoordMV;
-    RCP<CoordMV> fineCoords =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                        lNodesPerDir, meshData,
-                                                                        meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("ny", gNodesPerDir[1]);
-    matrixList.set("nz", gNodesPerDir[2]);
-    matrixList.set("matrixType","Laplace3D");
-    matrixList.set("left boundary" , "Neumann");
-    matrixList.set("right boundary" , "Neumann");
-    matrixList.set("front boundary" , "Neumann");
-    matrixList.set("back boundary"  , "Neumann");
-    matrixList.set("bottom boundary", "Dirichlet");
-    matrixList.set("top boundary"   , "Dirichlet");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", fineCoords->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    nullSpace->putScalar(1.0);
-
-    fineLevel.Set("A", A);
-    fineLevel.Set("Nullspace", nullSpace);
-    fineLevel.Set("Coordinates", fineCoords);
-    fineLevel.Set("CoarseNumZLayers", numSize);
-
-    RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
-    LineDetectionFact->SetParameter("linedetection: orientation",
-                                    Teuchos::ParameterEntry(std::string("coordinates")));
-    LineDetectionFact->SetParameter("linedetection: num layers",
-                                    Teuchos::ParameterEntry(numSize));
-
-    RCP<SemiCoarsenPFactory> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory());
-    SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
-    SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
-    coarseLevel.Request("P", SemiCoarsenPFact.get());
-    coarseLevel.Request("Coordinates", SemiCoarsenPFact.get());
-    SemiCoarsenPFact->Build(fineLevel, coarseLevel);
-
-    RCP<Matrix> P;
-    coarseLevel.Get("P", P, SemiCoarsenPFact.get());
-    RCP<CoordMV> coarseCoords = coarseLevel.Get<RCP<CoordMV>>("Coordinates", SemiCoarsenPFact.get());
-
-    coarseLevel.Release("P", SemiCoarsenPFact.get());
-    coarseLevel.Release("Coordinates", SemiCoarsenPFact.get());
-
-    // Prolongate coarse coords and compute difference
-    using STS = Teuchos::ScalarTraits<SC>;
-    const auto one = STS::one();
-    RCP<MultiVector> coarseCoordsSC = Utilities::RealValuedToScalarMultiVector(coarseCoords);
-    RCP<MultiVector> fineCoordsDiff = Utilities::RealValuedToScalarMultiVector(fineCoords);
-    P->apply(*coarseCoordsSC, *fineCoordsDiff, Teuchos::NO_TRANS, one, -one);
-
-    // check prolongation of coarse coordinates
-    // in this special case, the third layer will have the correct coordinates
-    const int numNodes = fineCoordsDiff->getLocalLength();
-    const int numVectors = fineCoordsDiff->getNumVectors();
-    const auto fineMap = fineCoordsDiff->getMap();
-    const GO gStartLayer3 = 2 * numSize * numSize;
-    const GO gEndLayer3 = 3 * numSize * numSize;
-    LO numBadCoords = 0;
-    for (LO node = 0; node < numNodes; ++node) {
-      const GO gnode = fineMap->getGlobalElement(node);
-      if (gnode >= gStartLayer3 && gnode < gEndLayer3)
-        for (int k = 0; k < numVectors; ++k) {
-          const auto fineCoordsDiffArray = fineCoordsDiff->getData(k);
-          if (STS::magnitude(fineCoordsDiffArray[node]) > 100*STS::eps())
-            numBadCoords += 1;
-        }
-    }
-    const auto comm = Teuchos::DefaultComm<int>::getComm();
-    LO gNumBadCoords;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &numBadCoords, &gNumBadCoords);
-    TEST_EQUALITY(gNumBadCoords, 0);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactory, TestSemiCoarsenP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Global Lexicographic";
+  const std::string coupling   = "uncoupled";
+  const LO numDimensions       = 3;
+  const LO numSize             = 5;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(numDimensions);
+  Array<GO> gNodesPerDir(numDimensions);
+  for (int dim = 0; dim < numDimensions; ++dim)
+    gNodesPerDir[dim] = numSize;
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coord_type;
+  typedef Xpetra::MultiVector<coord_type, LO, GO, NO> CoordMV;
+  RCP<CoordMV> fineCoords =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("ny", gNodesPerDir[1]);
+  matrixList.set("nz", gNodesPerDir[2]);
+  matrixList.set("matrixType", "Laplace3D");
+  matrixList.set("left boundary", "Neumann");
+  matrixList.set("right boundary", "Neumann");
+  matrixList.set("front boundary", "Neumann");
+  matrixList.set("back boundary", "Neumann");
+  matrixList.set("bottom boundary", "Dirichlet");
+  matrixList.set("top boundary", "Dirichlet");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", fineCoords->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+
+  fineLevel.Set("A", A);
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("Coordinates", fineCoords);
+  fineLevel.Set("CoarseNumZLayers", numSize);
+
+  RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
+  LineDetectionFact->SetParameter("linedetection: orientation",
+                                  Teuchos::ParameterEntry(std::string("coordinates")));
+  LineDetectionFact->SetParameter("linedetection: num layers",
+                                  Teuchos::ParameterEntry(numSize));
+
+  RCP<SemiCoarsenPFactory> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory());
+  SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
+  SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
+  coarseLevel.Request("P", SemiCoarsenPFact.get());
+  coarseLevel.Request("Coordinates", SemiCoarsenPFact.get());
+  SemiCoarsenPFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, SemiCoarsenPFact.get());
+  RCP<CoordMV> coarseCoords = coarseLevel.Get<RCP<CoordMV>>("Coordinates", SemiCoarsenPFact.get());
+
+  coarseLevel.Release("P", SemiCoarsenPFact.get());
+  coarseLevel.Release("Coordinates", SemiCoarsenPFact.get());
+
+  // Prolongate coarse coords and compute difference
+  using STS                       = Teuchos::ScalarTraits<SC>;
+  const auto one                  = STS::one();
+  RCP<MultiVector> coarseCoordsSC = Utilities::RealValuedToScalarMultiVector(coarseCoords);
+  RCP<MultiVector> fineCoordsDiff = Utilities::RealValuedToScalarMultiVector(fineCoords);
+  P->apply(*coarseCoordsSC, *fineCoordsDiff, Teuchos::NO_TRANS, one, -one);
+
+  // check prolongation of coarse coordinates
+  // in this special case, the third layer will have the correct coordinates
+  const int numNodes    = fineCoordsDiff->getLocalLength();
+  const int numVectors  = fineCoordsDiff->getNumVectors();
+  const auto fineMap    = fineCoordsDiff->getMap();
+  const GO gStartLayer3 = 2 * numSize * numSize;
+  const GO gEndLayer3   = 3 * numSize * numSize;
+  LO numBadCoords       = 0;
+  for (LO node = 0; node < numNodes; ++node) {
+    const GO gnode = fineMap->getGlobalElement(node);
+    if (gnode >= gStartLayer3 && gnode < gEndLayer3)
+      for (int k = 0; k < numVectors; ++k) {
+        const auto fineCoordsDiffArray = fineCoordsDiff->getData(k);
+        if (STS::magnitude(fineCoordsDiffArray[node]) > 100 * STS::eps())
+          numBadCoords += 1;
+      }
   }
+  const auto comm = Teuchos::DefaultComm<int>::getComm();
+  LO gNumBadCoords;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &numBadCoords, &gNumBadCoords);
+  TEST_EQUALITY(gNumBadCoords, 0);
+}
 
 #define MUELU_ETI_GROUP(SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SemiCoarsenPFactory, TestSemiCoarsenP, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SemiCoarsenPFactoryWithSemiRestriction.cpp b/packages/muelu/test/unit_tests/SemiCoarsenPFactoryWithSemiRestriction.cpp
index 3d5e79a12a5c..820c0bdefd90 100644
--- a/packages/muelu/test/unit_tests/SemiCoarsenPFactoryWithSemiRestriction.cpp
+++ b/packages/muelu/test/unit_tests/SemiCoarsenPFactoryWithSemiRestriction.cpp
@@ -64,128 +64,127 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactoryWithSemiRestriction, TestSemiRestrict, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    // Don't test for complex - matrix reader won't work
-   if (TST::isComplex) {success=true; return;}
-
-    // MAKE SURE lines are not split across processors !!
-    //    ==> nMatrixRows/(nProcs*lineLength*blkSize) must be an integer
-    //    ==> use vertical orientation
-//    GO lineLength = 7, blkSize = 1, nMatrixRows= 84;
-    int lineLength = 7, blkSize = 2, nMatrixRows= 168;
-
-    // Set up factories for test
-
-    RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
-    LineDetectionFact->SetParameter("linedetection: orientation",
-                                    Teuchos::ParameterEntry(std::string("vertical")));
-    LineDetectionFact->SetParameter("linedetection: num layers",
-                                    Teuchos::ParameterEntry(lineLength));
-    LineDetectionFact->SetVerbLevel(MueLu::Extreme);
-    RCP<SemiCoarsenPFactory> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory());
-    SemiCoarsenPFact->SetVerbLevel(MueLu::Extreme);
-    SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
-    SemiCoarsenPFact->SetParameter("semicoarsen: calculate nonsym restriction", Teuchos::ParameterEntry(true));
-    SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
-    // seem to need two factories to avoid failing multipleCallCheck on some machines?
-    RCP<LineDetectionFactory> LineDetectionFact2 = rcp(new LineDetectionFactory());
-    LineDetectionFact2->SetParameter("linedetection: orientation",
-                                    Teuchos::ParameterEntry(std::string("vertical")));
-    LineDetectionFact2->SetParameter("linedetection: num layers",
-                                    Teuchos::ParameterEntry(lineLength));
-    LineDetectionFact2->SetVerbLevel(MueLu::Extreme);
-    RCP<SemiCoarsenPFactory> SemiCoarsenPFact2 = rcp(new SemiCoarsenPFactory());
-    SemiCoarsenPFact2->SetVerbLevel(MueLu::Extreme);
-    SemiCoarsenPFact2->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
-    SemiCoarsenPFact2->SetParameter("semicoarsen: calculate nonsym restriction", Teuchos::ParameterEntry(true));
-    SemiCoarsenPFact2->SetFactory("LineDetection_VertLineIds", LineDetectionFact2);
-    SemiCoarsenPFact2->SetFactory("LineDetection_Layers", LineDetectionFact2);
-    SemiCoarsenPFact2->SetFactory("CoarseNumZLayers", LineDetectionFact2);
-
-    const RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), nMatrixRows, 0, comm);
-//    RCP<Matrix> Op =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/semiRtest2.mm", map);
-    RCP<Matrix> Op =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/semiRblkTestver.mm", map);
-    Op->SetFixedBlockSize(blkSize);
-    RCP<Matrix> transposedOp= Utilities::Transpose(*Op);
-
-    SC one = Teuchos::ScalarTraits<SC> ::one();
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(Op->getRowMap(), blkSize);
-
-    for (int i = 0; i < blkSize; i++) {
-      RCP<const Map> domainMap = Op->getDomainMap();
-      GO             indexBase = domainMap->getIndexBase();
-
-      ArrayRCP<SC> nsData = nullSpace->getDataNonConst(i);
-      for (int j = 0; j < nsData.size(); j++) {
-        GO GID = domainMap->getGlobalElement(j) - indexBase;
-        if ((GID-i) % blkSize == 0) nsData[j] = one;
-      }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactoryWithSemiRestriction, TestSemiRestrict, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  // Don't test for complex - matrix reader won't work
+  if (TST::isComplex) {
+    success = true;
+    return;
+  }
+
+  // MAKE SURE lines are not split across processors !!
+  //    ==> nMatrixRows/(nProcs*lineLength*blkSize) must be an integer
+  //    ==> use vertical orientation
+  //    GO lineLength = 7, blkSize = 1, nMatrixRows= 84;
+  int lineLength = 7, blkSize = 2, nMatrixRows = 168;
+
+  // Set up factories for test
+
+  RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
+  LineDetectionFact->SetParameter("linedetection: orientation",
+                                  Teuchos::ParameterEntry(std::string("vertical")));
+  LineDetectionFact->SetParameter("linedetection: num layers",
+                                  Teuchos::ParameterEntry(lineLength));
+  LineDetectionFact->SetVerbLevel(MueLu::Extreme);
+  RCP<SemiCoarsenPFactory> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory());
+  SemiCoarsenPFact->SetVerbLevel(MueLu::Extreme);
+  SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
+  SemiCoarsenPFact->SetParameter("semicoarsen: calculate nonsym restriction", Teuchos::ParameterEntry(true));
+  SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
+  // seem to need two factories to avoid failing multipleCallCheck on some machines?
+  RCP<LineDetectionFactory> LineDetectionFact2 = rcp(new LineDetectionFactory());
+  LineDetectionFact2->SetParameter("linedetection: orientation",
+                                   Teuchos::ParameterEntry(std::string("vertical")));
+  LineDetectionFact2->SetParameter("linedetection: num layers",
+                                   Teuchos::ParameterEntry(lineLength));
+  LineDetectionFact2->SetVerbLevel(MueLu::Extreme);
+  RCP<SemiCoarsenPFactory> SemiCoarsenPFact2 = rcp(new SemiCoarsenPFactory());
+  SemiCoarsenPFact2->SetVerbLevel(MueLu::Extreme);
+  SemiCoarsenPFact2->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
+  SemiCoarsenPFact2->SetParameter("semicoarsen: calculate nonsym restriction", Teuchos::ParameterEntry(true));
+  SemiCoarsenPFact2->SetFactory("LineDetection_VertLineIds", LineDetectionFact2);
+  SemiCoarsenPFact2->SetFactory("LineDetection_Layers", LineDetectionFact2);
+  SemiCoarsenPFact2->SetFactory("CoarseNumZLayers", LineDetectionFact2);
+
+  const RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), nMatrixRows, 0, comm);
+  //    RCP<Matrix> Op =  Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/semiRtest2.mm", map);
+  RCP<Matrix> Op = Xpetra::IO<SC, LO, GO, NO>::Read("TestMatrices/semiRblkTestver.mm", map);
+  Op->SetFixedBlockSize(blkSize);
+  RCP<Matrix> transposedOp = Utilities::Transpose(*Op);
+
+  SC one                     = Teuchos::ScalarTraits<SC>::one();
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(Op->getRowMap(), blkSize);
+
+  for (int i = 0; i < blkSize; i++) {
+    RCP<const Map> domainMap = Op->getDomainMap();
+    GO indexBase             = domainMap->getIndexBase();
+
+    ArrayRCP<SC> nsData = nullSpace->getDataNonConst(i);
+    for (int j = 0; j < nsData.size(); j++) {
+      GO GID = domainMap->getGlobalElement(j) - indexBase;
+      if ((GID - i) % blkSize == 0) nsData[j] = one;
     }
-
-
-    Level fineLevel1, coarseLevel1, fineLevel2, coarseLevel2;
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel1, coarseLevel1);
-    TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel2, coarseLevel2);
-    fineLevel1.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel1.SetFactoryManager(Teuchos::null);
-    fineLevel1.Set("A",Op);
-    fineLevel1.Set("Nullspace", nullSpace);
-    coarseLevel1.Request("P", SemiCoarsenPFact.get());
-    SemiCoarsenPFact->Build(fineLevel1, coarseLevel1);
-
-
-
-    fineLevel2.Set("A",transposedOp);
-    fineLevel2.Set("Nullspace", nullSpace);
-    fineLevel2.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel2.SetFactoryManager(Teuchos::null);
-    coarseLevel2.Request("P", SemiCoarsenPFact2.get());
-    coarseLevel2.Request("RfromPfactory", SemiCoarsenPFact2.get());
-    SemiCoarsenPFact2->Build(fineLevel2, coarseLevel2);
-
-    RCP<Matrix> firstP  = coarseLevel1.Get< RCP<Matrix> >("P", &(*SemiCoarsenPFact));
-    RCP<Matrix> secondR = coarseLevel2.Get< RCP<Matrix> >("RfromPfactory", &(*SemiCoarsenPFact2));
-    RCP<Matrix> transposedSecondR = Utilities::Transpose(*secondR);
-// Xpetra::IO<SC,LO,GO,NO>::Write("firstP.mm",*firstP);
-// Xpetra::IO<SC,LO,GO,NO>::Write("secondRtrans.mm",*transposedSecondR);
-
- Utilities::SetRandomSeed(*comm);
- RCP<MultiVector> workVec1 = MultiVectorFactory::Build(firstP->getRangeMap(),1);
- RCP<MultiVector> workVec2 = MultiVectorFactory::Build(transposedSecondR->getRangeMap(),1);
- RCP<MultiVector> X = MultiVectorFactory::Build(firstP->getDomainMap(),1);
- X->randomize();
- firstP->apply(*X,*workVec1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
- transposedSecondR->apply(*X,*workVec2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
- RCP<MultiVector> diff = MultiVectorFactory::Build(firstP->getRowMap(),1);
- diff->putScalar(0.0);
- //diff = workVec1 + (-1.0)*(workVec2) + 0*diff
- diff->update(1.0,*workVec1,-1.0,*workVec2,0.0);
- Teuchos::Array<typename TST::magnitudeType> norms(1);
- diff->norm2(norms);
- out << "||diff" << "||_2 = " << norms[0] << std::endl;
- TEST_EQUALITY(norms[0] < 1000*TMT::eps(), true);
-
-  } // NullSpace test
+  }
+
+  Level fineLevel1, coarseLevel1, fineLevel2, coarseLevel2;
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel1, coarseLevel1);
+  TestHelpers::TestFactory<Scalar, LO, GO, NO>::createTwoLevelHierarchy(fineLevel2, coarseLevel2);
+  fineLevel1.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel1.SetFactoryManager(Teuchos::null);
+  fineLevel1.Set("A", Op);
+  fineLevel1.Set("Nullspace", nullSpace);
+  coarseLevel1.Request("P", SemiCoarsenPFact.get());
+  SemiCoarsenPFact->Build(fineLevel1, coarseLevel1);
+
+  fineLevel2.Set("A", transposedOp);
+  fineLevel2.Set("Nullspace", nullSpace);
+  fineLevel2.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel2.SetFactoryManager(Teuchos::null);
+  coarseLevel2.Request("P", SemiCoarsenPFact2.get());
+  coarseLevel2.Request("RfromPfactory", SemiCoarsenPFact2.get());
+  SemiCoarsenPFact2->Build(fineLevel2, coarseLevel2);
+
+  RCP<Matrix> firstP            = coarseLevel1.Get<RCP<Matrix> >("P", &(*SemiCoarsenPFact));
+  RCP<Matrix> secondR           = coarseLevel2.Get<RCP<Matrix> >("RfromPfactory", &(*SemiCoarsenPFact2));
+  RCP<Matrix> transposedSecondR = Utilities::Transpose(*secondR);
+  // Xpetra::IO<SC,LO,GO,NO>::Write("firstP.mm",*firstP);
+  // Xpetra::IO<SC,LO,GO,NO>::Write("secondRtrans.mm",*transposedSecondR);
+
+  Utilities::SetRandomSeed(*comm);
+  RCP<MultiVector> workVec1 = MultiVectorFactory::Build(firstP->getRangeMap(), 1);
+  RCP<MultiVector> workVec2 = MultiVectorFactory::Build(transposedSecondR->getRangeMap(), 1);
+  RCP<MultiVector> X        = MultiVectorFactory::Build(firstP->getDomainMap(), 1);
+  X->randomize();
+  firstP->apply(*X, *workVec1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  transposedSecondR->apply(*X, *workVec2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  RCP<MultiVector> diff = MultiVectorFactory::Build(firstP->getRowMap(), 1);
+  diff->putScalar(0.0);
+  // diff = workVec1 + (-1.0)*(workVec2) + 0*diff
+  diff->update(1.0, *workVec1, -1.0, *workVec2, 0.0);
+  Teuchos::Array<typename TST::magnitudeType> norms(1);
+  diff->norm2(norms);
+  out << "||diff"
+      << "||_2 = " << norms[0] << std::endl;
+  TEST_EQUALITY(norms[0] < 1000 * TMT::eps(), true);
+
+}  // NullSpace test
 
 #define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SemiCoarsenPFactoryWithSemiRestriction,TestSemiRestrict,Scalar,LO,GO,Node)
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SemiCoarsenPFactoryWithSemiRestriction, TestSemiRestrict, Scalar, LO, GO, Node)
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/Amesos2Smoother.cpp b/packages/muelu/test/unit_tests/Smoothers/Amesos2Smoother.cpp
index cd3d518563d8..3287a468e881 100644
--- a/packages/muelu/test/unit_tests/Smoothers/Amesos2Smoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/Amesos2Smoother.cpp
@@ -54,50 +54,46 @@
 
 namespace MueLuTests {
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Amesos2Smoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Amesos2Smoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra)
-    {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
 #if defined HAVE_AMESOS2_KLU2 || defined HAVE_AMESOS2_SUPERLU
-      testApplyNoSetup(Amesos2Smoother(), out, success);
+    testApplyNoSetup(Amesos2Smoother(), out, success);
 #endif
-    }
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Amesos2Smoother, Apply_Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Amesos2Smoother, Apply_Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra)
-    {
-      Teuchos::RCP<Amesos2Smoother> smoother;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::RCP<Amesos2Smoother> smoother;
 #ifdef HAVE_AMESOS2_KLU2
-      smoother = Teuchos::rcp(new Amesos2Smoother("Klu"));
-      testDirectSolver(*smoother, out, success);
+    smoother = Teuchos::rcp(new Amesos2Smoother("Klu"));
+    testDirectSolver(*smoother, out, success);
 #endif
 
 #ifdef HAVE_AMESOS2_SUPERLU
-      smoother = Teuchos::rcp(new Amesos2Smoother("Superlu"));
-      testDirectSolver(*smoother, out, success);
+    smoother = Teuchos::rcp(new Amesos2Smoother("Superlu"));
+    testDirectSolver(*smoother, out, success);
 #endif
-    }
   }
+}
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Amesos2Smoother, NotSetup, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Amesos2Smoother, Apply_Correctness, SC, LO, GO, Node)
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Amesos2Smoother, NotSetup, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Amesos2Smoother, Apply_Correctness, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/AmesosSmoother.cpp b/packages/muelu/test/unit_tests/Smoothers/AmesosSmoother.cpp
index ac38d0a74600..0776f32cf5c0 100644
--- a/packages/muelu/test/unit_tests/Smoothers/AmesosSmoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/AmesosSmoother.cpp
@@ -52,43 +52,39 @@
 
 namespace MueLuTests {
 
-  typedef double                Scalar;
-  typedef int                   LocalOrdinal;
-  typedef int                   GlobalOrdinal;
-  typedef Xpetra::EpetraNode    Node;
+typedef double Scalar;
+typedef int LocalOrdinal;
+typedef int GlobalOrdinal;
+typedef Xpetra::EpetraNode Node;
 #include "MueLu_UseShortNames.hpp"
 
-  using namespace TestHelpers::Smoothers;
+using namespace TestHelpers::Smoothers;
 
-  TEUCHOS_UNIT_TEST(AmesosSmoother, NotSetup)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra)
-    {
-      testApplyNoSetup(AmesosSmoother(), out, success);
-    }
+TEUCHOS_UNIT_TEST(AmesosSmoother, NotSetup) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    testApplyNoSetup(AmesosSmoother(), out, success);
   }
+}
 
-  TEUCHOS_UNIT_TEST(AmesosSmoother, Apply_Correctness)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra)
-    {
-      Teuchos::RCP<AmesosSmoother > smoother;
+TEUCHOS_UNIT_TEST(AmesosSmoother, Apply_Correctness) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    Teuchos::RCP<AmesosSmoother> smoother;
 #ifdef HAVE_AMESOS_KLU
-      smoother = rcp(new AmesosSmoother("Klu"));
-      testDirectSolver(*smoother, out, success);
+    smoother = rcp(new AmesosSmoother("Klu"));
+    testDirectSolver(*smoother, out, success);
 #endif
 
 #ifdef HAVE_AMESOS_SUPERLU
-      smoother = rcp(new AmesosSmoother("Superlu"));
-      testDirectSolver(*smoother, out, success);
+    smoother = rcp(new AmesosSmoother("Superlu"));
+    testDirectSolver(*smoother, out, success);
 #endif
-    }
   }
+}
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
-//TODO
-// Test with invalid std::string
-// Test with invalid parameterList? (== a characterization test for Amesos)
-// Test if paramList takes into account
-// Check if all the defaults that are used by MueLu are tested
+// TODO
+//  Test with invalid std::string
+//  Test with invalid parameterList? (== a characterization test for Amesos)
+//  Test if paramList takes into account
+//  Check if all the defaults that are used by MueLu are tested
diff --git a/packages/muelu/test/unit_tests/Smoothers/BelosSmoother.cpp b/packages/muelu/test/unit_tests/Smoothers/BelosSmoother.cpp
index 0a1ea42cfeec..462a08bdaa59 100644
--- a/packages/muelu/test/unit_tests/Smoothers/BelosSmoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/BelosSmoother.cpp
@@ -50,53 +50,44 @@
 #include <MueLu_BelosSmoother.hpp>
 #include <MueLu_Utilities.hpp>
 
-
 namespace MueLuTests {
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosSmoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      BelosSmoother smoother("Block CG", Teuchos::ParameterList());
-      testApplyNoSetup(smoother, out, success);
-
-    }
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosSmoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    BelosSmoother smoother("Block CG", Teuchos::ParameterList());
+    testApplyNoSetup(smoother, out, success);
+  }
+}
+
+// Tests interface to Belos's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosSmoother, HardCodedResult_BlockCG, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList paramList;
+    paramList.set("Maximum Iterations", 10);
+    BelosSmoother smoother("Block CG", paramList);
+
+    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    RCP<const Teuchos::Comm<int> > comm                                  = TestHelpers::Parameters::getDefaultComm();
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 1.2856486930664495771e-01;
+    TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 1e4 * Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
   }
 
-  // Tests interface to Belos's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BelosSmoother, HardCodedResult_BlockCG, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      Teuchos::ParameterList paramList;
-      paramList.set("Maximum Iterations", 10);
-      BelosSmoother smoother("Block CG", paramList);
-
-
-      typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 1.2856486930664495771e-01;
-      TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,1e4*Teuchos::ScalarTraits<typename Teuchos::ScalarTraits<SC>::magnitudeType>::eps());
-    }
-
-  } // Block CG
-
+}  // Block CG
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosSmoother,NotSetup,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosSmoother,HardCodedResult_BlockCG,SC,LO,GO,NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosSmoother, NotSetup, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BelosSmoother, HardCodedResult_BlockCG, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/BlockedDirectSolver.cpp b/packages/muelu/test/unit_tests/Smoothers/BlockedDirectSolver.cpp
index b89ac162e882..1846ba5f4e75 100644
--- a/packages/muelu/test/unit_tests/Smoothers/BlockedDirectSolver.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/BlockedDirectSolver.cpp
@@ -59,900 +59,925 @@
 #include <MueLu_BlockedDirectSolver.hpp>
 #include <MueLu_Utilities.hpp>
 
-
 namespace MueLuTests {
 
-  using namespace TestHelpers::Smoothers;
-
-  // Xpetra version of CreateMap
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(const std::set<GlobalOrdinal>& gids, const Teuchos::Comm<int>& comm) {
-    Teuchos::Array<GlobalOrdinal> mapvec;
-    mapvec.reserve(gids.size());
-    mapvec.assign(gids.begin(), gids.end());
-    GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
-    GlobalOrdinal gcount;
-    Teuchos::reduceAll(comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
-
-    Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
-        Teuchos::rcp(new MapType(gcount,
-            mapvec(),
-            0,
-            Teuchos::rcpFromRef(comm)));
-    mapvec.clear();
-    return map;
-  }
+using namespace TestHelpers::Smoothers;
+
+// Xpetra version of CreateMap
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(const std::set<GlobalOrdinal>& gids, const Teuchos::Comm<int>& comm) {
+  Teuchos::Array<GlobalOrdinal> mapvec;
+  mapvec.reserve(gids.size());
+  mapvec.assign(gids.begin(), gids.end());
+  GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
+  GlobalOrdinal gcount;
+  Teuchos::reduceAll(comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
+
+  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
+      Teuchos::rcp(new MapType(gcount,
+                               mapvec(),
+                               0,
+                               Teuchos::rcpFromRef(comm)));
+  mapvec.clear();
+  return map;
+}
 
-  // Xpetra version of SplitMap
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Agiven) {
-    Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
-
-    GlobalOrdinal count=0;
-    Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
-    for (size_t i=0; i<Amap.getLocalNumElements(); ++i) {
-      const GlobalOrdinal gid = Amap.getGlobalElement(i);
-      if (Agiven.isNodeGlobalElement(gid)) continue;
-      myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
-      ++count;
-    }
-    myaugids.resize(count);
-    GlobalOrdinal gcount;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
-    return Teuchos::rcp(new MapType(gcount,myaugids(),0,comm));
+// Xpetra version of SplitMap
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Agiven) {
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
+
+  GlobalOrdinal count = 0;
+  Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
+  for (size_t i = 0; i < Amap.getLocalNumElements(); ++i) {
+    const GlobalOrdinal gid = Amap.getGlobalElement(i);
+    if (Agiven.isNodeGlobalElement(gid)) continue;
+    myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
+    ++count;
   }
+  myaugids.resize(count);
+  GlobalOrdinal gcount;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
+  return Teuchos::rcp(new MapType(gcount, myaugids(), 0, comm));
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrix(int noBlocks, const Teuchos::Comm<int>& comm) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
-    typedef Xpetra::CrsMatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixFactory;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixWrap;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrix(int noBlocks, const Teuchos::Comm<int>& comm) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
+  typedef Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixFactory;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixWrap;
 
-    GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,noBlocks-2)) * 10;
+  GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, noBlocks - 2)) * 10;
 
-    GlobalOrdinal procOffset = comm.getRank() * nOverallDOFGidsPerProc;
+  GlobalOrdinal procOffset = comm.getRank() * nOverallDOFGidsPerProc;
 
-    std::set<GlobalOrdinal> myDOFGids;
-    for(GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
-      myDOFGids.insert(i + procOffset);
+  std::set<GlobalOrdinal> myDOFGids;
+  for (GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
+    myDOFGids.insert(i + procOffset);
 
-    Teuchos::RCP<Map> fullmap = CreateMap<LocalOrdinal,GlobalOrdinal,Node,MapType>(myDOFGids, comm);
+  Teuchos::RCP<Map> fullmap = CreateMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(myDOFGids, comm);
 
-    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-    GlobalOrdinal nPartGIDs = nOverallDOFGidsPerProc;
-    Teuchos::RCP<Map> remainingpartmap = fullmap;
-    for (int it=0; it<noBlocks; it++) {
-      if(it == noBlocks - 1) {
-        maps[0] = remainingpartmap;
-        break;
-      }
-      // collect first half of GIDs
-      nPartGIDs = nPartGIDs / 2;
-      std::set<GlobalOrdinal> myHalfGIDs;
-      for(GlobalOrdinal j = 0; j < nPartGIDs; j++)
-        myHalfGIDs.insert(j + procOffset);
+  std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+  GlobalOrdinal nPartGIDs            = nOverallDOFGidsPerProc;
+  Teuchos::RCP<Map> remainingpartmap = fullmap;
+  for (int it = 0; it < noBlocks; it++) {
+    if (it == noBlocks - 1) {
+      maps[0] = remainingpartmap;
+      break;
+    }
+    // collect first half of GIDs
+    nPartGIDs = nPartGIDs / 2;
+    std::set<GlobalOrdinal> myHalfGIDs;
+    for (GlobalOrdinal j = 0; j < nPartGIDs; j++)
+      myHalfGIDs.insert(j + procOffset);
 
-      Teuchos::RCP<Map> halfmap = CreateMap<LocalOrdinal,GlobalOrdinal,Node,MapType> (myHalfGIDs, comm);
+    Teuchos::RCP<Map> halfmap = CreateMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(myHalfGIDs, comm);
 
-      Teuchos::RCP<Map> secondmap = SplitMap<LocalOrdinal,GlobalOrdinal,Node,MapType>(*remainingpartmap, *halfmap);
-      remainingpartmap = halfmap;
+    Teuchos::RCP<Map> secondmap = SplitMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(*remainingpartmap, *halfmap);
+    remainingpartmap            = halfmap;
 
-      maps[noBlocks - 1 - it]  = secondmap;
-    }
+    maps[noBlocks - 1 - it] = secondmap;
+  }
 
-    // create diagonal blocks
-    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-    for (int it=0; it<noBlocks; it++) {
-      //std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
-      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
+  // create diagonal blocks
+  std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+  for (int it = 0; it < noBlocks; it++) {
+    // std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
+    blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
 
-      LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+    LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
 
-      for (LocalOrdinal i = 0; i < NumMyElements; i++)
-        blocks[it]->insertGlobalValues(MyGlobalElements[i],
-            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-            Teuchos::tuple<Scalar>(it+1));
-      blocks[it]->fillComplete();
-    }
+    for (LocalOrdinal i = 0; i < NumMyElements; i++)
+      blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                     Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                     Teuchos::tuple<Scalar>(it + 1));
+    blocks[it]->fillComplete();
+  }
 
-    // create map extractor
-    Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
-    Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+  // create map extractor
+  Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+  Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
 
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
 
-    for (int it=0; it<noBlocks; it++) {
-      Teuchos::RCP<CrsMatrixWrap> csrwrap =
-          Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-      bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
-    }
-    bop->fillComplete();
-    return bop;
+  for (int it = 0; it < noBlocks; it++) {
+    Teuchos::RCP<CrsMatrixWrap> csrwrap =
+        Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+    bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
   }
+  bop->fillComplete();
+  return bop;
+}
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrixThyra(int noBlocks, const Teuchos::Comm<int>& comm) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node> MapFactory;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
-    typedef Xpetra::CrsMatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixFactory;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixWrap;
-
-    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-
-    MapType tm(1,0,Teuchos::rcpFromRef(comm));
-    Xpetra::UnderlyingLib lib = tm.lib();
-
-    maps[0] = MapFactory::Build (lib, comm.getSize() * 5, 5, 0, Teuchos::rcpFromRef(comm));
-    for (int it=1; it<noBlocks; it++) {
-      GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,it-1)*5);
-      maps[it]  = MapFactory::Build (lib, comm.getSize() * localDofs, localDofs, 0, Teuchos::rcpFromRef(comm));
-    }
-
-    // create diagonal blocks
-    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-    for (int it=0; it<noBlocks; it++) {
-      //std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
-      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrixThyra(int noBlocks, const Teuchos::Comm<int>& comm) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
+  typedef Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixFactory;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixWrap;
+
+  std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+
+  MapType tm(1, 0, Teuchos::rcpFromRef(comm));
+  Xpetra::UnderlyingLib lib = tm.lib();
+
+  maps[0] = MapFactory::Build(lib, comm.getSize() * 5, 5, 0, Teuchos::rcpFromRef(comm));
+  for (int it = 1; it < noBlocks; it++) {
+    GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, it - 1) * 5);
+    maps[it]                = MapFactory::Build(lib, comm.getSize() * localDofs, localDofs, 0, Teuchos::rcpFromRef(comm));
+  }
 
-      LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+  // create diagonal blocks
+  std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+  for (int it = 0; it < noBlocks; it++) {
+    // std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
+    blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
 
-      for (LocalOrdinal i = 0; i < NumMyElements; i++)
-        blocks[it]->insertGlobalValues(MyGlobalElements[i],
-            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-            Teuchos::tuple<Scalar>(it+1));
-      blocks[it]->fillComplete();
-    }
+    LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
 
-    // create map extractor
-    // To generate the Thyra style map extractor we do not need a full map but only the
-    // information about the Map details (i.e. lib and indexBase). We can extract this
-    // information from maps[0]
-    Teuchos::RCP<const MapExtractor > rgMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    Teuchos::RCP<const MapExtractor > doMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
-
-    for (int it=0; it<noBlocks; it++) {
-      Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-      bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
-    }
-    bop->fillComplete();
-    return bop;
+    for (LocalOrdinal i = 0; i < NumMyElements; i++)
+      blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                     Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                     Teuchos::tuple<Scalar>(it + 1));
+    blocks[it]->fillComplete();
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockedMatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-
-    std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
-    maps[0] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    maps[1] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    maps[2] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    RCP<Matrix> A00 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
-    RCP<Matrix> A01 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A10 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A11 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
-    RCP<Matrix> A12 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A21 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A22 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
-
-    // create map extractor
-    // To generate the Thyra style map extractor we do not need a full map but only the
-    // information about the Map details (i.e. lib and indexBase). We can extract this
-    // information from maps[0]
-    Teuchos::RCP<const MapExtractor > rgMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    Teuchos::RCP<const MapExtractor > doMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,5));
-    bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(0),A00);
-    bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(1),A01);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(0),A10);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(1),A11);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(2),A12);
-    bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(1),A21);
-    bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(2),A22);
-    bop->fillComplete();
-    return bop;
+  // create map extractor
+  // To generate the Thyra style map extractor we do not need a full map but only the
+  // information about the Map details (i.e. lib and indexBase). We can extract this
+  // information from maps[0]
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
+
+  for (int it = 0; it < noBlocks; it++) {
+    Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+    bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
   }
+  bop->fillComplete();
+  return bop;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockedMatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+
+  std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
+  maps[0]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[1]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[2]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  RCP<Matrix> A00                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A01                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A10                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A11                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A12                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A21                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A22                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
+
+  // create map extractor
+  // To generate the Thyra style map extractor we do not need a full map but only the
+  // information about the Map details (i.e. lib and indexBase). We can extract this
+  // information from maps[0]
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 5));
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(0), A00);
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(1), A01);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(0), A10);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(1), A11);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(2), A12);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(1), A21);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(2), A22);
+  bop->fillComplete();
+  return bop;
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
+  std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
 
-    for(const auto& solver_type : solver_types)
-    {
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  for (const auto& solver_type : solver_types) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 5, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 5, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      Level level;
-      TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.setlib(lib);
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.setlib(lib);
 
-      RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Normalize X
-      Array<magnitude_type> norms(1);
-      X->norm2(norms);
-      X->scale(1 / norms[0]);
-      
-      // Smoothers
-      RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solver_type));
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother", smootherFact);
-      MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // Smoothers
+    RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solver_type));
+    RCP<SmootherFactory> smootherFact          = rcp(new SmootherFactory(smootherPrototype));
 
-      // request smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      smootherFact->Build(level);
+    // request smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    smootherFact->Build(level);
 
-      // Compute RHS corresponding to X
-      A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC) 1.0, (SC) 0.0);
+    RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1);
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      solver->Apply(*X, *RHS, true);  //zero initial guess
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      Teuchos::Array<magnitude_type> finalNorms(1);
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    solver->Apply(*X, *RHS, true);  // zero initial guess
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
-                5e-15, out, success);
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
+                         5e-15, out, success);
 
-      solver->Apply(*X, *RHS, false); //nonzero initial guess
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    solver->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
-                5e-15, out, success);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
+                         5e-15, out, success);
+
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedI53I42II01II_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
-
-    for(const auto& solver_type : solver_types)
-    {
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 6, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.setlib(lib);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[5 3 [4 2] [0 1]]")));
-
-      // Smoothers
-      RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solver_type));
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother", smootherFact);
-      M.SetFactory("A", rAFact);
-
-      MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request block smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 4);
-      TEST_EQUALITY(reorderedbA->Cols(), 4);
-
-      RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      solver->Apply(*X, *RHS, true);  //zero initial guess
-      Teuchos::RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      Teuchos::RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for (size_t i = 0; i < XX->getLocalLength(); i++) {
-        if (i < 5) { if (xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i >= 5 && i < 10) { if (xdata[i] != (SC) (1.0 / 2.0)) bCheck = false; }
-        if (i >= 10 && i < 20) { if (xdata[i] != (SC) (1.0 / 3.0)) bCheck = false; }
-        if (i >= 20 && i < 40) { if (xdata[i] != (SC) (1.0 / 4.0)) bCheck = false; }
-        if (i >= 40 && i < 80) { if (xdata[i] != (SC) (1.0 / 5.0)) bCheck = false; }
-        if (i >= 80 && i < 160) { if (xdata[i] != (SC) (1.0 / 6.0)) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedI53I42II01II_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+  for (const auto& solver_type : solver_types) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 6, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      // Normalize X
-      Array<magnitude_type> norms(1);
-      X->norm2(norms);
-      X->scale(1 / norms[0]);
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.setlib(lib);
 
-      // Compute RHS corresponding to X
-      A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC) 1.0, (SC) 0.0);
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[5 3 [4 2] [0 1]]")));
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // Smoothers
+    RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solver_type));
+    RCP<SmootherFactory> smootherFact          = rcp(new SmootherFactory(smootherPrototype));
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1);
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      solver->Apply(*X, *RHS, true);  //zero initial guess
+    // request block smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      Teuchos::Array<magnitude_type> finalNorms(1);
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    smootherFact->Build(level);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, 5e-15,
-              out, success);
-    }
-  }
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedI53I42II01II_Thyra_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
-
-    for(const auto& solve_type : solver_types)
-    {
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 6, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.setlib(lib);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[5 3 [4 2] [0 1]]")));
-
-      // Smoothers
-      RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solve_type));
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother", smootherFact);
-      M.SetFactory("A", rAFact);
-
-      MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request block smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 4);
-      TEST_EQUALITY(reorderedbA->Cols(), 4);
-
-      RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      solver->Apply(*X, *RHS, true);  //zero initial guess
-      Teuchos::RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      Teuchos::RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for (size_t i = 0; i < XX->getLocalLength(); i++) {
-        if (i < 5) { if (xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i >= 5 && i < 10) { if (xdata[i] != (SC) (1.0 / 2.0)) bCheck = false; }
-        if (i >= 10 && i < 20) { if (xdata[i] != (SC) (1.0 / 3.0)) bCheck = false; }
-        if (i >= 20 && i < 40) { if (xdata[i] != (SC) (1.0 / 4.0)) bCheck = false; }
-        if (i >= 40 && i < 80) { if (xdata[i] != (SC) (1.0 / 5.0)) bCheck = false; }
-        if (i >= 80 && i < 160) { if (xdata[i] != (SC) (1.0 / 6.0)) bCheck = false; }
+    RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 4);
+    TEST_EQUALITY(reorderedbA->Cols(), 4);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    solver->Apply(*X, *RHS, true);  // zero initial guess
+    Teuchos::RCP<BlockedMultiVector> bX   = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    Teuchos::RCP<MultiVector> XX          = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)(1.0 / 2.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 20 && i < 40) {
+        if (xdata[i] != (SC)(1.0 / 4.0)) bCheck = false;
+      }
+      if (i >= 40 && i < 80) {
+        if (xdata[i] != (SC)(1.0 / 5.0)) bCheck = false;
+      }
+      if (i >= 80 && i < 160) {
+        if (xdata[i] != (SC)(1.0 / 6.0)) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Normalize X
-      Array<magnitude_type> norms(1);
-      X->norm2(norms);
-      X->scale(1 / norms[0]);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // Compute RHS corresponding to X
-      A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC) 1.0, (SC) 0.0);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1);
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      solver->Apply(*X, *RHS, true);  //zero initial guess
+    solver->Apply(*X, *RHS, true);  // zero initial guess
 
-      Teuchos::Array<magnitude_type> finalNorms(1);
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, 5e-15,
-              out, success);
-    }
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, 5e-15,
+                         out, success);
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedII20I1I_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedI53I42II01II_Thyra_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
+  std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
 
+  for (const auto& solve_type : solver_types) {
     RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-    if (comm->getSize() > 1) {
-      out << "Skipping test for " << comm->getSize()
-          << " processors as Amesos2 cannot deal with non-standard maps in the non-serial case." << std::endl;
-      return;
-    }
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 6, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-    for(const auto& solver_type : solver_types)
-    {
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.setlib(lib);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
-
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
-      smootherPrototype->SetFactory("A", rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
-      std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A", rAFact);
-      sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<BlockedDirectSolver> smoProtoPredict = Teuchos::rcp(new BlockedDirectSolver(solver_type));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp(new SmootherFactory(smoProtoPredict));
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A", rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      SFact->SetFactory("A", rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new TrilinosSmoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother", smootherFact);
-      M.SetFactory("A", rAFact);
-
-      MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      Teuchos::RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      Teuchos::RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for (size_t i = 0; i < XX->getLocalLength(); i++) {
-        if (i < 10) { if (xdata[i] != (SC) (1.0 / 3.0)) bCheck = false; }
-        if (i >= 10 && i < 15) { if (xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i >= 15 && i < 20) { if (xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.setlib(lib);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[5 3 [4 2] [0 1]]")));
+
+    // Smoothers
+    RCP<BlockedDirectSolver> smootherPrototype = rcp(new BlockedDirectSolver(solve_type));
+    RCP<SmootherFactory> smootherFact          = rcp(new SmootherFactory(smootherPrototype));
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // Normalize X
-      Array<magnitude_type> norms(1);
-      X->norm2(norms);
-      X->scale(1 / norms[0]);
+    // request block smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC) 1.0, (SC) 0.0);
+    smootherFact->Build(level);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RCP<SmootherBase> solver = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1);
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    TEST_EQUALITY(reorderedbA->Rows(), 4);
+    TEST_EQUALITY(reorderedbA->Cols(), 4);
 
-      Teuchos::Array<magnitude_type> finalNorms(1);
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
-                5e-15, out, success);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    solver->Apply(*X, *RHS, true);  // zero initial guess
+    Teuchos::RCP<BlockedMultiVector> bX   = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    Teuchos::RCP<MultiVector> XX          = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)(1.0 / 2.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 20 && i < 40) {
+        if (xdata[i] != (SC)(1.0 / 4.0)) bCheck = false;
+      }
+      if (i >= 40 && i < 80) {
+        if (xdata[i] != (SC)(1.0 / 5.0)) bCheck = false;
+      }
+      if (i >= 80 && i < 160) {
+        if (xdata[i] != (SC)(1.0 / 6.0)) bCheck = false;
+      }
     }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    solver->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, 5e-15,
+                         out, success);
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedII20I1I_Thyra_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedII20I1I_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
+  std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-    if (comm->getSize() > 1) {
-      out << "Skipping test for " << comm->getSize()
-          << " processors as Amesos2 cannot deal with non-standard maps in the non-serial case." << std::endl;
-      return;
-    }
+  if (comm->getSize() > 1) {
+    out << "Skipping test for " << comm->getSize()
+        << " processors as Amesos2 cannot deal with non-standard maps in the non-serial case." << std::endl;
+    return;
+  }
 
-    for(const auto& solver_type : solver_types)
-    {
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      Level level;
-      TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.setlib(lib);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
-
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
-      smootherPrototype->SetFactory("A", rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
-      std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A", rAFact);
-      sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<BlockedDirectSolver> smoProtoPredict = Teuchos::rcp(new BlockedDirectSolver(solver_type));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp(new SmootherFactory(smoProtoPredict));
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A", rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      SFact->SetFactory("A", rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new TrilinosSmoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
-      RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother", smootherFact);
-      M.SetFactory("A", rAFact);
-
-      MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      Teuchos::RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      Teuchos::RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for (size_t i = 0; i < XX->getLocalLength(); i++) {
-        if (i < 10) { if (xdata[i] != (SC) (1.0 / 3.0)) bCheck = false; }
-        if (i >= 10 && i < 15) { if (xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i >= 15 && i < 20) { if (xdata[i] != (SC) 0.5) bCheck = false; }
+  for (const auto& solver_type : solver_types) {
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.setlib(lib);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
+
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<BlockedDirectSolver> smoProtoPredict = Teuchos::rcp(new BlockedDirectSolver(solver_type));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new TrilinosSmoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    Teuchos::RCP<BlockedMultiVector> bX   = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    Teuchos::RCP<MultiVector> XX          = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Normalize X
-      Array<magnitude_type> norms(1);
-      X->norm2(norms);
-      X->scale(1 / norms[0]);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC) 1.0, (SC) 0.0);
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1);
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      Teuchos::Array<magnitude_type> finalNorms(1);
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
-                5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
+                         5e-15, out, success);
+  }
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedDirectSolver, NestedII20I1I_Thyra_BlockedDirectSolver_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  std::array<std::string, 3> solver_types{"", "Klu", "Superlu"};
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  if (comm->getSize() > 1) {
+    out << "Skipping test for " << comm->getSize()
+        << " processors as Amesos2 cannot deal with non-standard maps in the non-serial case." << std::endl;
+    return;
+  }
+
+  for (const auto& solver_type : solver_types) {
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.setlib(lib);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
+
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<BlockedDirectSolver> smoProtoPredict = Teuchos::rcp(new BlockedDirectSolver(solver_type));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new TrilinosSmoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    Teuchos::RCP<BlockedMultiVector> bX   = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    Teuchos::RCP<MultiVector> XX          = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
     }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-15, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <,
+                         5e-15, out, success);
   }
+}
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver,BlockedDirectSolver_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver,NestedI53I42II01II_BlockedDirectSolver_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver,NestedI53I42II01II_Thyra_BlockedDirectSolver_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver,NestedII20I1I_BlockedDirectSolver_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver,NestedII20I1I_Thyra_BlockedDirectSolver_Setup_Apply,SC,LO,GO,NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver, BlockedDirectSolver_Setup_Apply, SC, LO, GO, NO)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver, NestedI53I42II01II_BlockedDirectSolver_Setup_Apply, SC, LO, GO, NO)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver, NestedI53I42II01II_Thyra_BlockedDirectSolver_Setup_Apply, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver, NestedII20I1I_BlockedDirectSolver_Setup_Apply, SC, LO, GO, NO)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedDirectSolver, NestedII20I1I_Thyra_BlockedDirectSolver_Setup_Apply, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/BlockedSmoother.cpp b/packages/muelu/test/unit_tests/Smoothers/BlockedSmoother.cpp
index f186c5201489..d0945c2ab9df 100644
--- a/packages/muelu/test/unit_tests/Smoothers/BlockedSmoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/BlockedSmoother.cpp
@@ -69,6601 +69,6788 @@
 
 namespace MueLuTests {
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
-
-  // Xpetra version of CreateMap
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(const std::set<GlobalOrdinal>& gids, const Teuchos::Comm<int>& comm) {
-    Teuchos::Array<GlobalOrdinal> mapvec;
-    mapvec.reserve(gids.size());
-    mapvec.assign(gids.begin(), gids.end());
-    GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
-    GlobalOrdinal gcount;
-    Teuchos::reduceAll(comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
-
-    Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
-        Teuchos::rcp(new MapType(gcount,
-            mapvec(),
-            0,
-            Teuchos::rcpFromRef(comm)));
-    mapvec.clear();
-    return map;
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
+
+// Xpetra version of CreateMap
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > CreateMap(const std::set<GlobalOrdinal>& gids, const Teuchos::Comm<int>& comm) {
+  Teuchos::Array<GlobalOrdinal> mapvec;
+  mapvec.reserve(gids.size());
+  mapvec.assign(gids.begin(), gids.end());
+  GlobalOrdinal count = Teuchos::as<GlobalOrdinal>(mapvec.size());
+  GlobalOrdinal gcount;
+  Teuchos::reduceAll(comm, Teuchos::REDUCE_SUM, count, Teuchos::outArg(gcount));
+
+  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map =
+      Teuchos::rcp(new MapType(gcount,
+                               mapvec(),
+                               0,
+                               Teuchos::rcpFromRef(comm)));
+  mapvec.clear();
+  return map;
+}
+
+// Xpetra version of SplitMap
+template <class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& Agiven) {
+  Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
+
+  GlobalOrdinal count = 0;
+  Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
+  for (size_t i = 0; i < Amap.getLocalNumElements(); ++i) {
+    const GlobalOrdinal gid = Amap.getGlobalElement(i);
+    if (Agiven.isNodeGlobalElement(gid)) continue;
+    myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
+    ++count;
   }
-
-  // Xpetra version of SplitMap
-  template<class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > SplitMap(const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Amap, const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> & Agiven) {
-    Teuchos::RCP<const Teuchos::Comm<int> > comm = Amap.getComm();
-
-    GlobalOrdinal count=0;
-    Teuchos::Array<GlobalOrdinal> myaugids(Amap.getLocalNumElements());
-    for (size_t i=0; i<Amap.getLocalNumElements(); ++i) {
-      const GlobalOrdinal gid = Amap.getGlobalElement(i);
-      if (Agiven.isNodeGlobalElement(gid)) continue;
-      myaugids[Teuchos::as<GlobalOrdinal>(count)] = gid;
-      ++count;
+  myaugids.resize(count);
+  GlobalOrdinal gcount;
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
+  return Teuchos::rcp(new MapType(gcount, myaugids(), 0, comm));
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrix(int noBlocks, const Teuchos::Comm<int>& comm) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
+  typedef Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixFactory;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixWrap;
+
+  GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, noBlocks - 2)) * 10;
+
+  GlobalOrdinal procOffset = comm.getRank() * nOverallDOFGidsPerProc;
+
+  std::set<GlobalOrdinal> myDOFGids;
+  for (GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
+    myDOFGids.insert(i + procOffset);
+
+  Teuchos::RCP<Map> fullmap = CreateMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(myDOFGids, comm);
+
+  std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+  GlobalOrdinal nPartGIDs            = nOverallDOFGidsPerProc;
+  Teuchos::RCP<Map> remainingpartmap = fullmap;
+  for (int it = 0; it < noBlocks; it++) {
+    if (it == noBlocks - 1) {
+      maps[0] = remainingpartmap;
+      break;
     }
-    myaugids.resize(count);
-    GlobalOrdinal gcount;
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count, &gcount);
-    return Teuchos::rcp(new MapType(gcount,myaugids(),0,comm));
-  }
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrix(int noBlocks, const Teuchos::Comm<int>& comm) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
-    typedef Xpetra::CrsMatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixFactory;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixWrap;
-
-    GlobalOrdinal nOverallDOFGidsPerProc = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,noBlocks-2)) * 10;
-
-    GlobalOrdinal procOffset = comm.getRank() * nOverallDOFGidsPerProc;
-
-    std::set<GlobalOrdinal> myDOFGids;
-    for(GlobalOrdinal i = 0; i < nOverallDOFGidsPerProc; i++)
-      myDOFGids.insert(i + procOffset);
-
-    Teuchos::RCP<Map> fullmap = CreateMap<LocalOrdinal,GlobalOrdinal,Node,MapType>(myDOFGids, comm);
-
-    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-    GlobalOrdinal nPartGIDs = nOverallDOFGidsPerProc;
-    Teuchos::RCP<Map> remainingpartmap = fullmap;
-    for (int it=0; it<noBlocks; it++) {
-      if(it == noBlocks - 1) {
-        maps[0] = remainingpartmap;
-        break;
-      }
-      // collect first half of GIDs
-      nPartGIDs = nPartGIDs / 2;
-      std::set<GlobalOrdinal> myHalfGIDs;
-      for(GlobalOrdinal j = 0; j < nPartGIDs; j++)
-        myHalfGIDs.insert(j + procOffset);
+    // collect first half of GIDs
+    nPartGIDs = nPartGIDs / 2;
+    std::set<GlobalOrdinal> myHalfGIDs;
+    for (GlobalOrdinal j = 0; j < nPartGIDs; j++)
+      myHalfGIDs.insert(j + procOffset);
 
-      Teuchos::RCP<Map> halfmap = CreateMap<LocalOrdinal,GlobalOrdinal,Node,MapType> (myHalfGIDs, comm);
+    Teuchos::RCP<Map> halfmap = CreateMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(myHalfGIDs, comm);
 
-      Teuchos::RCP<Map> secondmap = SplitMap<LocalOrdinal,GlobalOrdinal,Node,MapType>(*remainingpartmap, *halfmap);
-      remainingpartmap = halfmap;
+    Teuchos::RCP<Map> secondmap = SplitMap<LocalOrdinal, GlobalOrdinal, Node, MapType>(*remainingpartmap, *halfmap);
+    remainingpartmap            = halfmap;
 
-      maps[noBlocks - 1 - it]  = secondmap;
-    }
+    maps[noBlocks - 1 - it] = secondmap;
+  }
 
-    // create diagonal blocks
-    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-    for (int it=0; it<noBlocks; it++) {
-      //std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
-      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
+  // create diagonal blocks
+  std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+  for (int it = 0; it < noBlocks; it++) {
+    // std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
+    blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
 
-      LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+    LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
 
-      for (LocalOrdinal i = 0; i < NumMyElements; i++)
-        blocks[it]->insertGlobalValues(MyGlobalElements[i],
-            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-            Teuchos::tuple<Scalar>(it+1));
-      blocks[it]->fillComplete();
-    }
+    for (LocalOrdinal i = 0; i < NumMyElements; i++)
+      blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                     Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                     Teuchos::tuple<Scalar>(it + 1));
+    blocks[it]->fillComplete();
+  }
 
-    // create map extractor
-    Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
-    Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+  // create map extractor
+  Teuchos::RCP<const MapExtractor> rgMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
+  Teuchos::RCP<const MapExtractor> doMapExtractor = Teuchos::rcp(new MapExtractor(fullmap, maps, false));
 
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
 
-    for (int it=0; it<noBlocks; it++) {
-      Teuchos::RCP<CrsMatrixWrap> csrwrap =
-          Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-      bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
-    }
-    bop->fillComplete();
-    return bop;
+  for (int it = 0; it < noBlocks; it++) {
+    Teuchos::RCP<CrsMatrixWrap> csrwrap =
+        Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+    bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
+  }
+  bop->fillComplete();
+  return bop;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockDiagonalExampleMatrixThyra(int noBlocks, const Teuchos::Comm<int>& comm) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node> MapFactory;
+  typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
+  typedef Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixFactory;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+  typedef Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrixWrap;
+
+  std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
+
+  MapType tm(1, 0, Teuchos::rcpFromRef(comm));
+  Xpetra::UnderlyingLib lib = tm.lib();
+
+  maps[0] = MapFactory::Build(lib, comm.getSize() * 5, 5, 0, Teuchos::rcpFromRef(comm));
+  for (int it = 1; it < noBlocks; it++) {
+    GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2, it - 1) * 5);
+    maps[it]                = MapFactory::Build(lib, comm.getSize() * localDofs, localDofs, 0, Teuchos::rcpFromRef(comm));
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class MapType>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockDiagonalExampleMatrixThyra(int noBlocks, const Teuchos::Comm<int>& comm) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node> MapFactory;
-    typedef Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> CrsMatrix;
-    typedef Xpetra::CrsMatrixFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixFactory;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-    typedef Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal,GlobalOrdinal,Node> CrsMatrixWrap;
-
-    std::vector<Teuchos::RCP<const Map> > maps(noBlocks, Teuchos::null);
-
-    MapType tm(1,0,Teuchos::rcpFromRef(comm));
-    Xpetra::UnderlyingLib lib = tm.lib();
-
-    maps[0] = MapFactory::Build (lib, comm.getSize() * 5, 5, 0, Teuchos::rcpFromRef(comm));
-    for (int it=1; it<noBlocks; it++) {
-      GlobalOrdinal localDofs = Teuchos::as<GlobalOrdinal>(Teuchos::ScalarTraits<GlobalOrdinal>::pow(2,it-1)*5);
-      maps[it]  = MapFactory::Build (lib, comm.getSize() * localDofs, localDofs, 0, Teuchos::rcpFromRef(comm));
-    }
-
-    // create diagonal blocks
-    std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks,Teuchos::null);
-    for (int it=0; it<noBlocks; it++) {
-      //std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
-      blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
-
-      LocalOrdinal NumMyElements = maps[it]->getLocalNumElements();
-      Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
+  // create diagonal blocks
+  std::vector<Teuchos::RCP<CrsMatrix> > blocks(noBlocks, Teuchos::null);
+  for (int it = 0; it < noBlocks; it++) {
+    // std::cout << it << " " << maps[it]->getMinAllGlobalIndex() << " - " << maps[it]->getMaxAllGlobalIndex() << std::endl;
+    blocks[it] = CrsMatrixFactory::Build(maps[it], 1);
 
-      for (LocalOrdinal i = 0; i < NumMyElements; i++)
-        blocks[it]->insertGlobalValues(MyGlobalElements[i],
-            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-            Teuchos::tuple<Scalar>(it+1));
-      blocks[it]->fillComplete();
-    }
+    LocalOrdinal NumMyElements                               = maps[it]->getLocalNumElements();
+    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = maps[it]->getLocalElementList();
 
-    // create map extractor
-    // To generate the Thyra style map extractor we do not need a full map but only the
-    // information about the Map details (i.e. lib and indexBase). We can extract this
-    // information from maps[0]
-    Teuchos::RCP<const MapExtractor > rgMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    Teuchos::RCP<const MapExtractor > doMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,1));
-
-    for (int it=0; it<noBlocks; it++) {
-      Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
-      bop->setMatrix(Teuchos::as<size_t>(it),Teuchos::as<size_t>(it),csrwrap);
-    }
-    bop->fillComplete();
-    return bop;
+    for (LocalOrdinal i = 0; i < NumMyElements; i++)
+      blocks[it]->insertGlobalValues(MyGlobalElements[i],
+                                     Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                                     Teuchos::tuple<Scalar>(it + 1));
+    blocks[it]->fillComplete();
   }
 
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > CreateBlockedMatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
-    typedef Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> Map;
-    typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
-    typedef Xpetra::MapExtractor<Scalar,LocalOrdinal,GlobalOrdinal,Node> MapExtractor;
-    typedef Xpetra::BlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> BlockedCrsMatrix;
-
-    std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
-    maps[0] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    maps[1] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    maps[2] = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
-    RCP<Matrix> A00 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
-    RCP<Matrix> A01 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A10 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A11 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
-    RCP<Matrix> A12 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A21 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
-    RCP<Matrix> A22 = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
-
-    // create map extractor
-    // To generate the Thyra style map extractor we do not need a full map but only the
-    // information about the Map details (i.e. lib and indexBase). We can extract this
-    // information from maps[0]
-    Teuchos::RCP<const MapExtractor > rgMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    Teuchos::RCP<const MapExtractor > doMapExtractor =
-        Teuchos::rcp(new MapExtractor(maps[0], maps, true));
-    // build blocked operator
-    Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor,doMapExtractor,5));
-    bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(0),A00);
-    bop->setMatrix(Teuchos::as<size_t>(0),Teuchos::as<size_t>(1),A01);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(0),A10);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(1),A11);
-    bop->setMatrix(Teuchos::as<size_t>(1),Teuchos::as<size_t>(2),A12);
-    bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(1),A21);
-    bop->setMatrix(Teuchos::as<size_t>(2),Teuchos::as<size_t>(2),A22);
-    bop->fillComplete();
-    return bop;
+  // create map extractor
+  // To generate the Thyra style map extractor we do not need a full map but only the
+  // information about the Map details (i.e. lib and indexBase). We can extract this
+  // information from maps[0]
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 1));
+
+  for (int it = 0; it < noBlocks; it++) {
+    Teuchos::RCP<CrsMatrixWrap> csrwrap = Teuchos::rcp(new CrsMatrixWrap(blocks[it]));
+    bop->setMatrix(Teuchos::as<size_t>(it), Teuchos::as<size_t>(it), csrwrap);
   }
+  bop->fillComplete();
+  return bop;
+}
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > CreateBlockedMatrixThyra(const Teuchos::Comm<int>& comm, Xpetra::UnderlyingLib lib) {
+  typedef Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> Map;
+  typedef Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> Matrix;
+  typedef Xpetra::MapExtractor<Scalar, LocalOrdinal, GlobalOrdinal, Node> MapExtractor;
+  typedef Xpetra::BlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> BlockedCrsMatrix;
+
+  std::vector<RCP<const Map> > maps = std::vector<RCP<const Map> >(3, Teuchos::null);
+  maps[0]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[1]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  maps[2]                           = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMap(100);
+  RCP<Matrix> A00                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A01                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[0], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A10                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A11                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], 4.0, -1.0, -1.0, lib);
+  RCP<Matrix> A12                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[1], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A21                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], -1.0, 0.0, 0.0, lib);
+  RCP<Matrix> A22                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildTridiag(maps[2], 4.0, -1.0, -1.0, lib);
+
+  // create map extractor
+  // To generate the Thyra style map extractor we do not need a full map but only the
+  // information about the Map details (i.e. lib and indexBase). We can extract this
+  // information from maps[0]
+  Teuchos::RCP<const MapExtractor> rgMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  Teuchos::RCP<const MapExtractor> doMapExtractor =
+      Teuchos::rcp(new MapExtractor(maps[0], maps, true));
+  // build blocked operator
+  Teuchos::RCP<BlockedCrsMatrix> bop = Teuchos::rcp(new BlockedCrsMatrix(rgMapExtractor, doMapExtractor, 5));
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(0), A00);
+  bop->setMatrix(Teuchos::as<size_t>(0), Teuchos::as<size_t>(1), A01);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(0), A10);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(1), A11);
+  bop->setMatrix(Teuchos::as<size_t>(1), Teuchos::as<size_t>(2), A12);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(1), A21);
+  bop->setMatrix(Teuchos::as<size_t>(2), Teuchos::as<size_t>(2), A22);
+  bop->fillComplete();
+  return bop;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Jacobi_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BlockedJacobiSmoother> smootherPrototype = rcp(new BlockedJacobiSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks, Teuchos::null);
+    for (int k = 0; k < noBlocks; k++) {
+      std::string strInfo = std::string("{ 1 }");
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", MueLu::NoFactory::getRCP());
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
+
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
+
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Jacobi_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BlockedJacobiSmoother> smootherPrototype     = rcp( new BlockedJacobiSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks, Teuchos::null);
-      for (int k = 0; k < noBlocks; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",MueLu::NoFactory::getRCP());
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
-      }
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // request Jacobi smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    // request Jacobi smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RCP<SmootherBase> jacSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<SmootherBase> jacSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      jacSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    jacSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      out << "solve with zero initial guess, and unreliable nonzeroed vector X" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    out << "solve with zero initial guess, and unreliable nonzeroed vector X" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      jacSmoother->Apply(*X, *RHS, true); //zero initial guess with nonzero X
+    jacSmoother->Apply(*X, *RHS, true);  // zero initial guess with nonzero X
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      jacSmoother->Apply(*X, *RHS, false); //nonzero initial guess
+    jacSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm3 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm3[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm3 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm3[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm3[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm3[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
-        TEST_EQUALITY(residualNorm1[0] != residualNorm3[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
+      TEST_EQUALITY(residualNorm1[0] != residualNorm3[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks, Teuchos::null);
+    for (int k = 0; k < noBlocks; k++) {
+      std::string strInfo = std::string("{ 1 }");
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", MueLu::NoFactory::getRCP());
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
+
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
+
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks, Teuchos::null);
-      for (int k = 0; k < noBlocks; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",MueLu::NoFactory::getRCP());
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
-      }
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    out << "solve with zero initial guess, and unreliable nonzeroed vector X" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      out << "solve with zero initial guess, and unreliable nonzeroed vector X" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess with nonzero X
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess with nonzero X
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    bgsSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      bgsSmoother->Apply(*X, *RHS, false); //nonzero initial guess
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm3 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm3[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm3 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm3[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    TEUCHOS_TEST_COMPARE(residualNorm3[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      TEUCHOS_TEST_COMPARE(residualNorm3[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
+      TEST_EQUALITY(residualNorm1[0] != residualNorm3[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Reordered_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 3 1 2 ]")));
+
+    int noBlocks2 = 3;
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smootherPrototype->SetFactory("A", rAFact);
+
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks2, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks2, Teuchos::null);
+    for (int k = 0; k < noBlocks2; k++) {
+      std::string strInfo = std::string("{ 1 }");
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", rAFact);
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
+
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
+
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] == residualNorm2[0], true);
-        TEST_EQUALITY(residualNorm1[0] != residualNorm3[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Reordered_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 3 1 2 ]")));
-
-      int noBlocks2 = 3;
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smootherPrototype->SetFactory("A", rAFact);
-
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks2, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks2, Teuchos::null);
-      for (int k = 0; k < noBlocks2; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",rAFact);
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("A", rAFact);
+    M.SetFactory("Smoother", smootherFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->putScalar(0.0);
+    RHS->putScalar(1.0);
+
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+
+    Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
+
+    Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X, 0);
+    Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X, 1);
+    Teuchos::RCP<MultiVector> v2 = doMapExtractor->ExtractVector(X, 2);
+
+    TEST_EQUALITY((v0->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((v1->getData(0))[0], Teuchos::as<Scalar>(0.5));
+    TEST_EQUALITY((v2->getData(0))[0], Teuchos::as<Scalar>(1.0 / 3.0));
+
+    Teuchos::Array<magnitude_type> n0(1);
+    v0->norm1(n0);
+    Teuchos::Array<magnitude_type> n1(1);
+    v1->norm1(n1);
+    Teuchos::Array<magnitude_type> n2(1);
+    v2->norm1(n2);
+
+    TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(v0->getGlobalLength() * 0.25));
+    TEST_EQUALITY(n1[0], Teuchos::as<Scalar>(v1->getGlobalLength() * 0.5));
+
+    TEST_EQUALITY(X->getMap()->getLocalElement(v0->getMap()->getMinGlobalIndex()), 0);
+    TEST_EQUALITY(X->getMap()->getLocalElement(v0->getMap()->getMaxGlobalIndex()), 19);
+    TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 19);
+    TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 20);
+    TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
+    TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
+    TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+
+    TEST_EQUALITY(X->getMap()->getLocalElement(v1->getMap()->getMinGlobalIndex()), 20);
+    TEST_EQUALITY(X->getMap()->getLocalElement(v1->getMap()->getMaxGlobalIndex()), 24);
+    TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 5);
+    TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+    TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
+    TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
+
+    TEST_EQUALITY(X->getMap()->getLocalElement(v2->getMap()->getMinGlobalIndex()), 25);
+    TEST_EQUALITY(X->getMap()->getLocalElement(v2->getMap()->getMaxGlobalIndex()), 34);
+    TEST_EQUALITY(v2->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v2->getMap()->getMaxLocalIndex(), 9);
+    TEST_EQUALITY(v2->getMap()->getMinAllGlobalIndex(), 10);
+    TEST_EQUALITY(v2->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
+    TEST_EQUALITY(v2->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+    TEST_EQUALITY(v2->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
+
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // global level
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smootherPrototype->SetFactory("A", rAFact);
+
+    for (int k = 0; k < 2; k++) {
+      Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
+      sA->SetFactory("A", rAFact);
+      sA->SetParameter("block row", Teuchos::ParameterEntry(k));  // local block indices relative to size of blocked operator
+      sA->SetParameter("block col", Teuchos::ParameterEntry(k));
+
+      Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+      sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+      sP->SetFactory("A", sA);
+
+      for (int l = 0; l < 2; l++) {
+        std::string strInfo                = std::string("{ 1 }");
+        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+        ssA->SetFactory("A", sA);
+        ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+        ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+        ssP->SetFactory("A", ssA);
+        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+        Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+        ssM->SetFactory("A", ssA);
+        ssM->SetFactory("Smoother", ssF);
+        ssM->SetIgnoreUserData(true);
+        sP->AddFactoryManager(ssM, l);
       }
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("A", rAFact);
-      M.SetFactory("Smoother",     smootherFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
+      Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
+      Teuchos::RCP<FactoryManager> sM  = Teuchos::rcp(new FactoryManager());
+      sM->SetFactory("A", sA);
+      sM->SetFactory("Smoother", sF);
+      sM->SetIgnoreUserData(true);
+      smootherPrototype->AddFactoryManager(sM, k);
+    }
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    // create master smoother factory
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("A", rAFact);
+    M.SetFactory("Smoother", smootherFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->putScalar(0.0);
+    RHS->putScalar(1.0);
+
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+
+    Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
+
+    Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X, 0);
+    Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X, 1);
+
+    Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
+    Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
+    TEST_EQUALITY(bv0.is_null(), false);
+    TEST_EQUALITY(bv1.is_null(), false);
+
+    Teuchos::RCP<MultiVector> bv00 = bv0->getMultiVector(0, false);
+    Teuchos::RCP<MultiVector> bv01 = bv0->getMultiVector(1, false);
+    Teuchos::RCP<MultiVector> bv10 = bv1->getMultiVector(0, false);
+    Teuchos::RCP<MultiVector> bv11 = bv1->getMultiVector(1, false);
+
+    TEST_EQUALITY((bv00->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((bv01->getData(0))[0], Teuchos::as<Scalar>(1.0));
+    TEST_EQUALITY((bv10->getData(0))[0], Teuchos::as<Scalar>(0.5));
+
+    Teuchos::Array<magnitude_type> n0(1);
+    v0->norm1(n0);
+    Teuchos::Array<magnitude_type> n1(1);
+    v1->norm1(n1);
+
+    TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 10.0));
+    TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 5.84, out, success);
+    TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 5.83, out, success);
+
+    TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 24);
+    TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 0);
+    TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
+    TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+
+    TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 14);
+    TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 5);
+    TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+    TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+    TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
+
+    Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0, 0));
+    TEST_EQUALITY(b00->Rows(), 2);
+    TEST_EQUALITY(b00->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v00         = me00->ExtractVector(v0, 0);
+    Teuchos::RCP<MultiVector> v01         = me00->ExtractVector(v0, 1);
+    TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(1.0));
+    TEST_EQUALITY(v00->getLocalLength(), 20);
+    TEST_EQUALITY(v01->getLocalLength(), 5);
+    TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
+    TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+
+    TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 19);
+    TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(), 20);
+    TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
+    TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
+    TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+
+    TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 0);
+    TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 4);
+    TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 36);
+
+    Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1, 1));
+    TEST_EQUALITY(b11->Rows(), 2);
+    TEST_EQUALITY(b11->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v10         = me11->ExtractVector(v1, 0);
+    Teuchos::RCP<MultiVector> v11         = me11->ExtractVector(v1, 1);
+    TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.5));
+    TEST_EQUALITY(v10->getLocalLength(), 5);
+    TEST_EQUALITY(v11->getLocalLength(), 10);
+    TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+    TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
+
+    TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 5);
+    TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+    TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
+    TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
+
+    TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 9);
+    TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 10);
+    TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
+    TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+    TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
+
+    Teuchos::Array<magnitude_type> n00(1);
+    v00->norm1(n00);
+    Teuchos::Array<magnitude_type> n11(1);
+    v01->norm1(n11);
+    Teuchos::Array<magnitude_type> n22(1);
+    v10->norm1(n22);
+    Teuchos::Array<magnitude_type> n33(1);
+    v11->norm1(n33);
+
+    TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
+    TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 1.0));
+    TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.5));
+    TEST_FLOATING_EQUALITY(n33[0], v11->getGlobalLength() / 3.0, Teuchos::ScalarTraits<magnitude_type>::eps());
+
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // global level
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smootherPrototype->SetFactory("A", rAFact);
+
+    for (int k = 0; k < 2; k++) {
+      Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
+      sA->SetFactory("A", rAFact);
+      sA->SetParameter("block row", Teuchos::ParameterEntry(k));  // local block indices relative to size of blocked operator
+      sA->SetParameter("block col", Teuchos::ParameterEntry(k));
+
+      Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+      sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+      sP->SetFactory("A", sA);
+
+      for (int l = 0; l < 2; l++) {
+        std::string strInfo                = std::string("{ 1 }");
+        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+        ssA->SetFactory("A", sA);
+        ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+        ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+        ssP->SetFactory("A", ssA);
+        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+        Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+        ssM->SetFactory("A", ssA);
+        ssM->SetFactory("Smoother", ssF);
+        ssM->SetIgnoreUserData(true);
+        sP->AddFactoryManager(ssM, l);
+      }
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+      Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
+      Teuchos::RCP<FactoryManager> sM  = Teuchos::rcp(new FactoryManager());
+      sM->SetFactory("A", sA);
+      sM->SetFactory("Smoother", sF);
+      sM->SetIgnoreUserData(true);
+      smootherPrototype->AddFactoryManager(sM, k);
+    }
 
-      // Random X
-      X->putScalar(0.0);
-      RHS->putScalar(1.0);
+    // create master smoother factory
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("A", rAFact);
+    M.SetFactory("Smoother", smootherFact);
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    smootherFact->Build(level);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
+    RCP<Matrix> reorderedA                     = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
 
-      Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X,0);
-      Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X,1);
-      Teuchos::RCP<MultiVector> v2 = doMapExtractor->ExtractVector(X,2);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      TEST_EQUALITY((v0->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((v1->getData(0))[0], Teuchos::as<Scalar>(0.5));
-      TEST_EQUALITY((v2->getData(0))[0], Teuchos::as<Scalar>(1.0/3.0));
+    RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
+    RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
 
-      Teuchos::Array<magnitude_type> n0(1); v0->norm1(n0);
-      Teuchos::Array<magnitude_type> n1(1); v1->norm1(n1);
-      Teuchos::Array<magnitude_type> n2(1); v2->norm1(n2);
+    RCP<MultiVector> X   = bX->Merge();
+    RCP<MultiVector> RHS = bRHS->Merge();
 
-      TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(v0->getGlobalLength() * 0.25));
-      TEST_EQUALITY(n1[0], Teuchos::as<Scalar>(v1->getGlobalLength() * 0.5));
+    // Random X
+    X->putScalar(0.0);
+    RHS->putScalar(1.0);
 
-      TEST_EQUALITY(X->getMap()->getLocalElement(v0->getMap()->getMinGlobalIndex()),0);
-      TEST_EQUALITY(X->getMap()->getLocalElement(v0->getMap()->getMaxGlobalIndex()),19);
-      TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 19);
-      TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 20);
-      TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
-      TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
-      TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      TEST_EQUALITY(X->getMap()->getLocalElement(v1->getMap()->getMinGlobalIndex()),20);
-      TEST_EQUALITY(X->getMap()->getLocalElement(v1->getMap()->getMaxGlobalIndex()),24);
-      TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 5);
-      TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-      TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
-      TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEST_EQUALITY(X->getMap()->getLocalElement(v2->getMap()->getMinGlobalIndex()),25);
-      TEST_EQUALITY(X->getMap()->getLocalElement(v2->getMap()->getMaxGlobalIndex()),34);
-      TEST_EQUALITY(v2->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v2->getMap()->getMaxLocalIndex(), 9);
-      TEST_EQUALITY(v2->getMap()->getMinAllGlobalIndex(), 10);
-      TEST_EQUALITY(v2->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
-      TEST_EQUALITY(v2->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-      TEST_EQUALITY(v2->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-    } // end UseTpetra
-  }
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // global level
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smootherPrototype->SetFactory("A", rAFact);
-
-      for(int k = 0; k < 2; k++ ) {
-        Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
-        sA->SetFactory("A",rAFact);
-        sA->SetParameter("block row",Teuchos::ParameterEntry(k)); // local block indices relative to size of blocked operator
-        sA->SetParameter("block col",Teuchos::ParameterEntry(k));
-
-        Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-        sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-        sP->SetFactory("A", sA);
-
-        for(int l = 0; l < 2; l++) {
-          std::string strInfo = std::string("{ 1 }");
-          Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-          ssA->SetFactory("A",sA);
-          ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-          ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-          ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-          ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-          RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-          ssP->SetFactory("A", ssA);
-          Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-          Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-          ssM->SetFactory("A", ssA);
-          ssM->SetFactory("Smoother", ssF);
-          ssM->SetIgnoreUserData(true);
-          sP->AddFactoryManager(ssM,l);
-        }
-
-        Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
-        Teuchos::RCP<FactoryManager> sM = Teuchos::rcp(new FactoryManager());
-        sM->SetFactory("A", sA);
-        sM->SetFactory("Smoother", sF);
-        sM->SetIgnoreUserData(true);
-        smootherPrototype->AddFactoryManager(sM,k);
-      }
+    Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
 
-      // create master smoother factory
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    {
+      RCP<BlockedMultiVector> test = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), X));
+      bX.swap(test);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("A", rAFact);
-      M.SetFactory("Smoother",     smootherFact);
+      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [0 1] [2 3] ]");
+      test                                                = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
+      bX.swap(test);
+    }
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(bX, 0);
+    Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(bX, 1);
+
+    Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
+    Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
+    TEST_EQUALITY(bv0.is_null(), false);
+    TEST_EQUALITY(bv1.is_null(), false);
+
+    Teuchos::RCP<MultiVector> bv00 = bv0->getMultiVector(0, false);
+    Teuchos::RCP<MultiVector> bv01 = bv0->getMultiVector(1, false);
+    Teuchos::RCP<MultiVector> bv10 = bv1->getMultiVector(0, false);
+    Teuchos::RCP<MultiVector> bv11 = bv1->getMultiVector(1, false);
+
+    TEST_EQUALITY((bv00->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((bv01->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((bv10->getData(0))[0], Teuchos::as<Scalar>(0.25));
+
+    Teuchos::Array<magnitude_type> n0(1);
+    v0->norm1(n0);
+    Teuchos::Array<magnitude_type> n1(1);
+    v1->norm1(n1);
+
+    TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 2.5));
+    TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 13.34, out, success);
+    TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 13.33, out, success);
+
+    TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 9);
+    TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40);
+    TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
+    TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
+
+    TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 29);
+    TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 10);
+    TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
+    TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
+    TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+
+    Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0, 0));
+    TEST_EQUALITY(b00->Rows(), 2);
+    TEST_EQUALITY(b00->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v00         = me00->ExtractVector(v0, 0);
+    Teuchos::RCP<MultiVector> v01         = me00->ExtractVector(v0, 1);
+    TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY(v00->getLocalLength(), 5);
+    TEST_EQUALITY(v01->getLocalLength(), 5);
+    TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+    TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+
+    TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 40);
+    TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 4);
+    TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 36);
+
+    TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 5);
+    TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
+    TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
+    TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
+
+    Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1, 1));
+    TEST_EQUALITY(b11->Rows(), 2);
+    TEST_EQUALITY(b11->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v10         = me11->ExtractVector(v1, 0);
+    Teuchos::RCP<MultiVector> v11         = me11->ExtractVector(v1, 1);
+    TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY(v10->getLocalLength(), 10);
+    TEST_EQUALITY(v11->getLocalLength(), 20);
+    TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
+    TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
+
+    TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 9);
+    TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 10);
+    TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
+    TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
+    TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
+
+    TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 19);
+    TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 20);
+    TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
+    TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
+    TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
+
+    Teuchos::Array<magnitude_type> n00(1);
+    v00->norm1(n00);
+    Teuchos::Array<magnitude_type> n11(1);
+    v01->norm1(n11);
+    Teuchos::Array<magnitude_type> n22(1);
+    v10->norm1(n22);
+    Teuchos::Array<magnitude_type> n33(1);
+    v11->norm1(n33);
+
+    TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
+    TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 0.25));
+    TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.25));
+    TEUCHOS_TEST_COMPARE(n33[0], <, v11->getGlobalLength() * 0.54167, out, success);
+    TEUCHOS_TEST_COMPARE(n33[0], >, v11->getGlobalLength() * 0.54166, out, success);
+
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    RCP<BlockedCrsMatrix> A = CreateBlockedMatrixThyra<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*comm, Xpetra::UseTpetra);
+
+    TEST_EQUALITY(A->Rows(), 3);
+    TEST_EQUALITY(A->Cols(), 3);
+    TEST_EQUALITY(A->getRangeMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
+    TEST_EQUALITY(A->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(0, false)->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getRangeMap(0, false)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getRangeMap(0, false)->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(0, false)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(1, false)->getMinAllGlobalIndex(), comm->getSize() * 100);
+    TEST_EQUALITY(A->getRangeMap(1, false)->getMaxAllGlobalIndex(), comm->getSize() * 100 * 2 - 1);
+    TEST_EQUALITY(A->getRangeMap(1, false)->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(1, false)->getMaxGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(2, false)->getMinAllGlobalIndex(), comm->getSize() * 200);
+    TEST_EQUALITY(A->getRangeMap(2, false)->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
+    TEST_EQUALITY(A->getRangeMap(2, false)->getMinGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(2, false)->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(0, true)->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getRangeMap(0, true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getRangeMap(0, true)->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(0, true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getRangeMap(1, true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getMatrix(0, 0)->getRangeMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getMatrix(0, 0)->getDomainMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getMatrix(0, 0)->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    TEST_EQUALITY(A->getMatrix(2, 1)->getDomainMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(A->getMatrix(2, 1)->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
+    TEST_EQUALITY(A->getMatrix(2, 1)->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
+    TEST_EQUALITY(A->getMatrix(2, 1)->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
+
+    // build gloabl vector with one entries
+    Teuchos::RCP<Vector> ones = VectorFactory::Build(A->getRangeMap(), true);
+    Teuchos::RCP<Vector> res  = VectorFactory::Build(A->getRangeMap(), true);
+    ones->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+
+    A->apply(*ones, *res);
+
+    TEST_EQUALITY(res->norm1(), comm->getSize() * 100 * 2 + 6);
+    TEST_EQUALITY(res->normInf(), 2);
+
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(10));
+
+    int noBlocks = Teuchos::as<int>(A->Rows());
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks, Teuchos::null);
+    for (int k = 0; k < noBlocks; k++) {
+      std::string strInfo = std::string("{ 1 }");
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", MueLu::NoFactory::getRCP());
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+      Teuchos::ParameterList paramList;
+      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(10));
+      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
 
-      smootherFact->Build(level);
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-      // Random X
-      X->putScalar(0.0);
-      RHS->putScalar(1.0);
-
-
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-
-      Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
-
-      Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X,0);
-      Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X,1);
-
-      Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
-      Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
-      TEST_EQUALITY(bv0.is_null(),false);
-      TEST_EQUALITY(bv1.is_null(),false);
-
-      Teuchos::RCP<MultiVector> bv00 = bv0->getMultiVector(0,false);
-      Teuchos::RCP<MultiVector> bv01 = bv0->getMultiVector(1,false);
-      Teuchos::RCP<MultiVector> bv10 = bv1->getMultiVector(0,false);
-      Teuchos::RCP<MultiVector> bv11 = bv1->getMultiVector(1,false);
-
-      TEST_EQUALITY((bv00->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((bv01->getData(0))[0], Teuchos::as<Scalar>(1.0));
-      TEST_EQUALITY((bv10->getData(0))[0], Teuchos::as<Scalar>(0.5));
-
-      Teuchos::Array<magnitude_type> n0(1); v0->norm1(n0);
-      Teuchos::Array<magnitude_type> n1(1); v1->norm1(n1);
-
-      TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 10.0));
-      TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 5.84, out, success);
-      TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 5.83, out, success);
-
-      TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 24);
-      TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 0);
-      TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
-      TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
-
-      TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 14);
-      TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 5);
-      TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-      TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-      TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
-
-      Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0,0));
-      TEST_EQUALITY(b00->Rows(), 2);
-      TEST_EQUALITY(b00->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v00 = me00->ExtractVector(v0,0);
-      Teuchos::RCP<MultiVector> v01 = me00->ExtractVector(v0,1);
-      TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(1.0));
-      TEST_EQUALITY(v00->getLocalLength(), 20);
-      TEST_EQUALITY(v01->getLocalLength(), 5);
-      TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
-      TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-
-      TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 19);
-      TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(), 20);
-      TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
-      TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
-      TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
-
-      TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 0);
-      TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 4);
-      TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 36);
-
-
-      Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1,1));
-      TEST_EQUALITY(b11->Rows(), 2);
-      TEST_EQUALITY(b11->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v10 = me11->ExtractVector(v1,0);
-      Teuchos::RCP<MultiVector> v11 = me11->ExtractVector(v1,1);
-      TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.5));
-      TEST_EQUALITY(v10->getLocalLength(), 5);
-      TEST_EQUALITY(v11->getLocalLength(), 10);
-      TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-      TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
-
-      TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 5);
-      TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-      TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
-      TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
-
-      TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 9);
-      TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 10);
-      TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
-      TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-      TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
-
-      Teuchos::Array<magnitude_type> n00(1); v00->norm1(n00);
-      Teuchos::Array<magnitude_type> n11(1); v01->norm1(n11);
-      Teuchos::Array<magnitude_type> n22(1); v10->norm1(n22);
-      Teuchos::Array<magnitude_type> n33(1); v11->norm1(n33);
-
-      TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
-      TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 1.0));
-      TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.5));
-      TEST_FLOATING_EQUALITY(n33[0], v11->getGlobalLength() / 3.0, Teuchos::ScalarTraits<magnitude_type>::eps());
-
-    } // end UseTpetra
-  }
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // global level
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smootherPrototype->SetFactory("A", rAFact);
-
-      for(int k = 0; k < 2; k++ ) {
-        Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
-        sA->SetFactory("A",rAFact);
-        sA->SetParameter("block row",Teuchos::ParameterEntry(k)); // local block indices relative to size of blocked operator
-        sA->SetParameter("block col",Teuchos::ParameterEntry(k));
-
-        Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-        sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-        sP->SetFactory("A", sA);
-
-        for(int l = 0; l < 2; l++) {
-          std::string strInfo = std::string("{ 1 }");
-          Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-          ssA->SetFactory("A",sA);
-          ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-          ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-          ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-          ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-          RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-          ssP->SetFactory("A", ssA);
-          Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-          Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-          ssM->SetFactory("A", ssA);
-          ssM->SetFactory("Smoother", ssF);
-          ssM->SetIgnoreUserData(true);
-          sP->AddFactoryManager(ssM,l);
-        }
-
-        Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
-        Teuchos::RCP<FactoryManager> sM = Teuchos::rcp(new FactoryManager());
-        sM->SetFactory("A", sA);
-        sM->SetFactory("Smoother", sF);
-        sM->SetIgnoreUserData(true);
-        smootherPrototype->AddFactoryManager(sM,k);
-      }
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // create master smoother factory
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherFact->Build(level);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("A", rAFact);
-      M.SetFactory("Smoother",     smootherFact);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      smootherFact->Build(level);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
-      RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      RCP<MultiVector> X = bX->Merge();
-      RCP<MultiVector> RHS = bRHS->Merge();
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      // Random X
-      X->putScalar(0.0);
-      RHS->putScalar(1.0);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-4, out, success);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
+    bgsSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      {
-        RCP<BlockedMultiVector> test = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), X));
-        bX.swap(test);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-        Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [0 1] [2 3] ]");
-        test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
-        bX.swap(test);
-      }
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, 1e-2, out, success);
 
-      Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(bX,0);
-      Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(bX,1);
-
-      Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
-      Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
-      TEST_EQUALITY(bv0.is_null(),false);
-      TEST_EQUALITY(bv1.is_null(),false);
-
-      Teuchos::RCP<MultiVector> bv00 = bv0->getMultiVector(0,false);
-      Teuchos::RCP<MultiVector> bv01 = bv0->getMultiVector(1,false);
-      Teuchos::RCP<MultiVector> bv10 = bv1->getMultiVector(0,false);
-      Teuchos::RCP<MultiVector> bv11 = bv1->getMultiVector(1,false);
-
-      TEST_EQUALITY((bv00->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((bv01->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((bv10->getData(0))[0], Teuchos::as<Scalar>(0.25));
-
-      Teuchos::Array<magnitude_type> n0(1); v0->norm1(n0);
-      Teuchos::Array<magnitude_type> n1(1); v1->norm1(n1);
-
-      TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 2.5));
-      TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 13.34, out, success);
-      TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 13.33, out, success);
-
-      TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 9);
-      TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 40);
-      TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
-      TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
-
-      TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 29);
-      TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 10);
-      TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
-      TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
-      TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
-
-      Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0,0));
-      TEST_EQUALITY(b00->Rows(), 2);
-      TEST_EQUALITY(b00->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v00 = me00->ExtractVector(v0,0);
-      Teuchos::RCP<MultiVector> v01 = me00->ExtractVector(v0,1);
-      TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY(v00->getLocalLength(), 5);
-      TEST_EQUALITY(v01->getLocalLength(), 5);
-      TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-      TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-
-      TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 40);
-      TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 4);
-      TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 36);
-
-      TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 5);
-      TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 5);
-      TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 9);
-      TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 31);
-
-
-      Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1,1));
-      TEST_EQUALITY(b11->Rows(), 2);
-      TEST_EQUALITY(b11->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v10 = me11->ExtractVector(v1,0);
-      Teuchos::RCP<MultiVector> v11 = me11->ExtractVector(v1,1);
-      TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY(v10->getLocalLength(), 10);
-      TEST_EQUALITY(v11->getLocalLength(), 20);
-      TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
-      TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
-
-      TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 9);
-      TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 10);
-      TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 10);
-      TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 19);
-      TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 21);
-
-      TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 19);
-      TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 20);
-      TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 40 + 20);
-      TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 40 + 39);
-      TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 40 - 1);
-
-      Teuchos::Array<magnitude_type> n00(1); v00->norm1(n00);
-      Teuchos::Array<magnitude_type> n11(1); v01->norm1(n11);
-      Teuchos::Array<magnitude_type> n22(1); v10->norm1(n22);
-      Teuchos::Array<magnitude_type> n33(1); v11->norm1(n33);
-
-      TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
-      TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 0.25));
-      TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.25));
-      TEUCHOS_TEST_COMPARE(n33[0], <, v11->getGlobalLength() * 0.54167, out, success);
-      TEUCHOS_TEST_COMPARE(n33[0], >, v11->getGlobalLength() * 0.54166, out, success);
-
-    } // end UseTpetra
-  }
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end Tpetra only
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    RCP<BlockedCrsMatrix> A = CreateBlockedMatrixThyra<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*comm, Xpetra::UseTpetra);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A));
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [1 2] ]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // global level
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(30));
+    smootherPrototype->SetFactory("A", rAFact);
+
+    std::string strInfo = std::string("{ 1 }");
+
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
+    paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+
+    Teuchos::RCP<SubBlockAFactory> sA00 = Teuchos::rcp(new SubBlockAFactory());
+    sA00->SetFactory("A", rAFact);
+    sA00->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA00->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA00->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA00->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> sP00 = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+    sP00->SetFactory("A", sA00);
+    Teuchos::RCP<SmootherFactory> sF00 = Teuchos::rcp(new SmootherFactory(sP00));
+    Teuchos::RCP<FactoryManager> sM00  = Teuchos::rcp(new FactoryManager());
+    sM00->SetFactory("A", sA00);
+    sM00->SetFactory("Smoother", sF00);
+    sM00->SetIgnoreUserData(true);
+    smootherPrototype->AddFactoryManager(sM00, 0);
+
+    // 2x2 blocked operator
+    Teuchos::RCP<SubBlockAFactory> sA11 = Teuchos::rcp(new SubBlockAFactory());
+    sA11->SetFactory("A", rAFact);
+    sA11->SetParameter("block row", Teuchos::ParameterEntry(1));
+    sA11->SetParameter("block col", Teuchos::ParameterEntry(1));
+
+    Teuchos::RCP<BlockedGaussSeidelSmoother> sP11 = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+    sP11->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    sP11->SetFactory("A", sA11);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA11);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      sP11->AddFactoryManager(ssM, l);
+    }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      RCP<BlockedCrsMatrix> A = CreateBlockedMatrixThyra<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*comm, Xpetra::UseTpetra);
-
-      TEST_EQUALITY(A->Rows(), 3);
-      TEST_EQUALITY(A->Cols(), 3);
-      TEST_EQUALITY(A->getRangeMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
-      TEST_EQUALITY(A->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap()->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(0,false)->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getRangeMap(0,false)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getRangeMap(0,false)->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(0,false)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(1,false)->getMinAllGlobalIndex(), comm->getSize() * 100);
-      TEST_EQUALITY(A->getRangeMap(1,false)->getMaxAllGlobalIndex(), comm->getSize() * 100 * 2 - 1);
-      TEST_EQUALITY(A->getRangeMap(1,false)->getMinGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(1,false)->getMaxGlobalIndex(), comm->getSize() * 100 + comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(2,false)->getMinAllGlobalIndex(), comm->getSize() * 200);
-      TEST_EQUALITY(A->getRangeMap(2,false)->getMaxAllGlobalIndex(), comm->getSize() * 300 - 1);
-      TEST_EQUALITY(A->getRangeMap(2,false)->getMinGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(2,false)->getMaxGlobalIndex(), comm->getSize() * 200 + comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(0,true)->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getRangeMap(0,true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getRangeMap(0,true)->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(0,true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getRangeMap(1,true)->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getMatrix(0,0)->getRangeMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getMatrix(0,0)->getRangeMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getMatrix(0,0)->getRangeMap()->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getMatrix(0,0)->getRangeMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getMatrix(0,0)->getDomainMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getMatrix(0,0)->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getMatrix(0,0)->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getMatrix(0,0)->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      TEST_EQUALITY(A->getMatrix(2,1)->getDomainMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(A->getMatrix(2,1)->getDomainMap()->getMaxAllGlobalIndex(), comm->getSize() * 100 - 1);
-      TEST_EQUALITY(A->getMatrix(2,1)->getDomainMap()->getMinGlobalIndex(), comm->getRank() * 100);
-      TEST_EQUALITY(A->getMatrix(2,1)->getDomainMap()->getMaxGlobalIndex(), comm->getRank() * 100 + 99);
-
-      // build gloabl vector with one entries
-      Teuchos::RCP<Vector> ones = VectorFactory::Build(A->getRangeMap(), true);
-      Teuchos::RCP<Vector> res  = VectorFactory::Build(A->getRangeMap(), true);
-      ones->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-
-      A->apply(*ones, *res);
-
-      TEST_EQUALITY(res->norm1(), comm->getSize() * 100 * 2 + 6);
-      TEST_EQUALITY(res->normInf(), 2);
-
-
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype     = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(10));
-
-      int noBlocks = Teuchos::as<int>(A->Rows());
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks, Teuchos::null);
-      for (int k = 0; k < noBlocks; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",MueLu::NoFactory::getRCP());
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        Teuchos::ParameterList paramList;
-        paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(10));
-        paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
+    Teuchos::RCP<SmootherFactory> sF11 = Teuchos::rcp(new SmootherFactory(sP11));
+    Teuchos::RCP<FactoryManager> sM11  = Teuchos::rcp(new FactoryManager());
+    sM11->SetFactory("A", sA11);
+    sM11->SetFactory("Smoother", sF11);
+    sM11->SetIgnoreUserData(true);
+    smootherPrototype->AddFactoryManager(sM11, 1);
+
+    // create master smoother factory
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("A", rAFact);
+    M.SetFactory("Smoother", smootherFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->putScalar(0.0);
+    RHS->putScalar(1.0);
+
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 10 * std::sqrt(Teuchos::ScalarTraits<Scalar>::eps());
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+  }  // end Tpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    int noBlocks                             = 4;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrixThyra<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // global level
+    RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp(new BlockedGaussSeidelSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smootherPrototype->SetFactory("A", rAFact);
+
+    for (int k = 0; k < 2; k++) {
+      Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
+      sA->SetFactory("A", rAFact);
+      sA->SetParameter("block row", Teuchos::ParameterEntry(k));  // local block indices relative to size of blocked operator
+      sA->SetParameter("block col", Teuchos::ParameterEntry(k));
+
+      Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+      sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+      sP->SetFactory("A", sA);
+
+      for (int l = 0; l < 2; l++) {
+        std::string strInfo                = std::string("{ 1 }");
+        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+        ssA->SetFactory("A", sA);
+        ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+        ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+        ssP->SetFactory("A", ssA);
+        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+        Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+        ssM->SetFactory("A", ssA);
+        ssM->SetFactory("Smoother", ssF);
+        ssM->SetIgnoreUserData(true);
+        sP->AddFactoryManager(ssM, l);
       }
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+      Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
+      Teuchos::RCP<FactoryManager> sM  = Teuchos::rcp(new FactoryManager());
+      sM->SetFactory("A", sA);
+      sM->SetFactory("Smoother", sF);
+      sM->SetIgnoreUserData(true);
+      smootherPrototype->AddFactoryManager(sM, k);
+    }
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
+    // create master smoother factory
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("A", rAFact);
+    M.SetFactory("Smoother", smootherFact);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      smootherFact->Build(level);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    smootherFact->Build(level);
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Random X
+    X->putScalar(0.0);
+    RHS->putScalar(1.0);
 
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    bgsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    // We should remove the boolean from ExtractVector. It is not necessary
+#ifdef HAVE_XPETRA_DEBUG
+    TEST_THROW(doMapExtractor->ExtractVector(X, 0, false), Xpetra::Exceptions::RuntimeError);
+    TEST_THROW(doMapExtractor->ExtractVector(X, 1, false), Xpetra::Exceptions::RuntimeError);
+#endif
+    Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X, 0, true);
+    Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X, 1, true);
+
+    Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
+    Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
+    TEST_EQUALITY((bv0->getMultiVector(0, true)->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((bv0->getMultiVector(0, true)->getData(0))[19], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((bv1->getMultiVector(0, true)->getData(0))[0], Teuchos::as<Scalar>(0.5));
+    TEST_EQUALITY((bv1->getMultiVector(0, true)->getData(0))[4], Teuchos::as<Scalar>(0.5));
+    TEST_EQUALITY((bv0->getMultiVector(1, true)->getData(0))[0], Teuchos::as<Scalar>(1.0));
+    TEST_EQUALITY((bv0->getMultiVector(1, true)->getData(0))[4], Teuchos::as<Scalar>(1.0));
+
+    Teuchos::Array<magnitude_type> n0(1);
+    v0->norm1(n0);
+    Teuchos::Array<magnitude_type> n1(1);
+    v1->norm1(n1);
+
+    TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 10.0));
+    TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 5.84, out, success);
+    TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 5.83, out, success);
+
+    TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 24);
+    TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 20);
+    TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getSize() * 20 + comm->getRank() * 5 + 4);
+    TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 25 - 1);
+
+    TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 14);
+    TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
+    TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getSize() * 5 + comm->getRank() * 10 + 9);
+    TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 15 - 1);
+
+    Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0, 0));
+    TEST_EQUALITY(b00->Rows(), 2);
+    TEST_EQUALITY(b00->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v00         = me00->ExtractVector(v0, 0, true);
+    Teuchos::RCP<MultiVector> v01         = me00->ExtractVector(v0, 1, true);
+    TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
+    TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(1.0));
+    TEST_EQUALITY(v00->getLocalLength(), 20);
+    TEST_EQUALITY(v01->getLocalLength(), 5);
+    TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
+    TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+
+    TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 19);
+    TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 20);
+    TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 20 + 19);
+    TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 20 - 1);
+
+    TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
+    TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 5 + 4);
+    TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 5 - 1);
+
+    Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1, 1));
+    TEST_EQUALITY(b11->Rows(), 2);
+    TEST_EQUALITY(b11->Cols(), 2);
+
+    Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
+    Teuchos::RCP<MultiVector> v10         = me11->ExtractVector(v1, 0, true);
+    Teuchos::RCP<MultiVector> v11         = me11->ExtractVector(v1, 1, true);
+    TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.5));
+    TEST_EQUALITY(v10->getLocalLength(), 5);
+    TEST_EQUALITY(v11->getLocalLength(), 10);
+    TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
+    TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
+
+    TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 4);
+    TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
+    TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 5 + 4);
+    TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 5 - 1);
+
+    TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
+    TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 9);
+    TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 0);
+    TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 10);
+    TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
+    TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 10 - 1);
+
+    Teuchos::Array<magnitude_type> n00(1);
+    v00->norm1(n00);
+    Teuchos::Array<magnitude_type> n11(1);
+    v01->norm1(n11);
+    Teuchos::Array<magnitude_type> n22(1);
+    v10->norm1(n22);
+    Teuchos::Array<magnitude_type> n33(1);
+    v11->norm1(n33);
+
+    TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
+    TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 1.0));
+    TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.5));
+    TEST_FLOATING_EQUALITY(n33[0], v11->getGlobalLength() / 3.0, Teuchos::ScalarTraits<magnitude_type>::eps());
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    int noBlocks                             = 2;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, noBlocks, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", MueLu::NoFactory::getRCP());
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 5e-4, out, success);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      bgsSmoother->Apply(*X, *RHS, false); //nonzero initial guess
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, 1e-2, out, success);
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end Tpetra only
-  }
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      RCP<BlockedCrsMatrix> A = CreateBlockedMatrixThyra<Scalar,LocalOrdinal,GlobalOrdinal,Node>(*comm, Xpetra::UseTpetra);
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", Teuchos::rcp_dynamic_cast<Matrix>(A));
+    simpleSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [1 2] ]")));
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      //////////////////////////////////////////////////////////////////////
-      // global level
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(30));
-      smootherPrototype->SetFactory("A", rAFact);
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      std::string strInfo = std::string("{ 1 }");
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp(new SimpleSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoPredict->SetFactory("A", sA[0]);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA[0]);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    }
 
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
-      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
 
-      Teuchos::RCP<SubBlockAFactory> sA00 = Teuchos::rcp(new SubBlockAFactory());
-      sA00->SetFactory("A",rAFact);
-      sA00->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA00->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA00->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA00->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> sP00 = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-      sP00->SetFactory("A", sA00);
-      Teuchos::RCP<SmootherFactory> sF00 = Teuchos::rcp(new SmootherFactory(sP00));
-      Teuchos::RCP<FactoryManager> sM00 = Teuchos::rcp(new FactoryManager());
-      sM00->SetFactory("A", sA00);
-      sM00->SetFactory("Smoother", sF00);
-      sM00->SetIgnoreUserData(true);
-      smootherPrototype->AddFactoryManager(sM00,0);
-
-      // 2x2 blocked operator
-      Teuchos::RCP<SubBlockAFactory> sA11 = Teuchos::rcp(new SubBlockAFactory());
-      sA11->SetFactory("A",rAFact);
-      sA11->SetParameter("block row",Teuchos::ParameterEntry(1));
-      sA11->SetParameter("block col",Teuchos::ParameterEntry(1));
-
-
-      Teuchos::RCP<BlockedGaussSeidelSmoother> sP11 = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-      sP11->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      sP11->SetFactory("A", sA11);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA11);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        sP11->AddFactoryManager(ssM,l);
-      }
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      Teuchos::RCP<SmootherFactory> sF11 = Teuchos::rcp(new SmootherFactory(sP11));
-      Teuchos::RCP<FactoryManager> sM11 = Teuchos::rcp(new FactoryManager());
-      sM11->SetFactory("A", sA11);
-      sM11->SetFactory("Smoother", sF11);
-      sM11->SetIgnoreUserData(true);
-      smootherPrototype->AddFactoryManager(sM11,1);
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
 
-      // create master smoother factory
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("A", rAFact);
-      M.SetFactory("Smoother",     smootherFact);
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      smootherFact->Build(level);
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // Random X
-      X->putScalar(0.0);
-      RHS->putScalar(1.0);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      magnitude_type tol = 10*std::sqrt(Teuchos::ScalarTraits<Scalar>::eps());
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-    } // end Tpetra
-  }
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 4;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrixThyra<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [ 3 0 ] [1 2] ]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // global level
-      RCP<BlockedGaussSeidelSmoother> smootherPrototype = rcp( new BlockedGaussSeidelSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smootherPrototype->SetFactory("A", rAFact);
-
-      for(int k = 0; k < 2; k++ ) {
-        Teuchos::RCP<SubBlockAFactory> sA = Teuchos::rcp(new SubBlockAFactory());
-        sA->SetFactory("A",rAFact);
-        sA->SetParameter("block row",Teuchos::ParameterEntry(k)); // local block indices relative to size of blocked operator
-        sA->SetParameter("block col",Teuchos::ParameterEntry(k));
-
-        Teuchos::RCP<BlockedGaussSeidelSmoother> sP = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-        sP->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-        sP->SetFactory("A", sA);
-
-        for(int l = 0; l < 2; l++) {
-          std::string strInfo = std::string("{ 1 }");
-          Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-          ssA->SetFactory("A",sA);
-          ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-          ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-          ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-          ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-          RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-          ssP->SetFactory("A", ssA);
-          Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-          Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-          ssM->SetFactory("A", ssA);
-          ssM->SetFactory("Smoother", ssF);
-          ssM->SetIgnoreUserData(true);
-          sP->AddFactoryManager(ssM,l);
-        }
-
-        Teuchos::RCP<SmootherFactory> sF = Teuchos::rcp(new SmootherFactory(sP));
-        Teuchos::RCP<FactoryManager> sM = Teuchos::rcp(new FactoryManager());
-        sM->SetFactory("A", sA);
-        sM->SetFactory("Smoother", sF);
-        sM->SetIgnoreUserData(true);
-        smootherPrototype->AddFactoryManager(sM,k);
-      }
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // create master smoother factory
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("A", rAFact);
-      M.SetFactory("Smoother",     smootherFact);
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp(new SimpleSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    }
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      smootherFact->Build(level);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      RCP<SmootherBase> bgsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // Random X
-      X->putScalar(0.0);
-      RHS->putScalar(1.0);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      bgsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      Teuchos::RCP<const MapExtractor> doMapExtractor = reorderedbA->getDomainMapExtractor();
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // We should remove the boolean from ExtractVector. It is not necessary
-#ifdef HAVE_XPETRA_DEBUG
-      TEST_THROW(doMapExtractor->ExtractVector(X,0,false),Xpetra::Exceptions::RuntimeError);
-      TEST_THROW(doMapExtractor->ExtractVector(X,1,false),Xpetra::Exceptions::RuntimeError);
-#endif
-      Teuchos::RCP<MultiVector> v0 = doMapExtractor->ExtractVector(X,0,true);
-      Teuchos::RCP<MultiVector> v1 = doMapExtractor->ExtractVector(X,1,true);
-
-      Teuchos::RCP<BlockedMultiVector> bv0 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v0);
-      Teuchos::RCP<BlockedMultiVector> bv1 = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(v1);
-      TEST_EQUALITY((bv0->getMultiVector(0,true)->getData(0))[0],  Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((bv0->getMultiVector(0,true)->getData(0))[19], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((bv1->getMultiVector(0,true)->getData(0))[0],  Teuchos::as<Scalar>(0.5));
-      TEST_EQUALITY((bv1->getMultiVector(0,true)->getData(0))[4],  Teuchos::as<Scalar>(0.5));
-      TEST_EQUALITY((bv0->getMultiVector(1,true)->getData(0))[0],  Teuchos::as<Scalar>(1.0));
-      TEST_EQUALITY((bv0->getMultiVector(1,true)->getData(0))[4],  Teuchos::as<Scalar>(1.0));
-
-      Teuchos::Array<magnitude_type> n0(1); v0->norm1(n0);
-      Teuchos::Array<magnitude_type> n1(1); v1->norm1(n1);
-
-      TEST_EQUALITY(n0[0], Teuchos::as<Scalar>(comm->getSize() * 10.0));
-      TEUCHOS_TEST_COMPARE(n1[0], <, comm->getSize() * 5.84, out, success);
-      TEUCHOS_TEST_COMPARE(n1[0], >, comm->getSize() * 5.83, out, success);
-
-      TEST_EQUALITY(v0->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMaxLocalIndex(), 24);
-      TEST_EQUALITY(v0->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v0->getMap()->getMinGlobalIndex(), comm->getRank() * 20);
-      TEST_EQUALITY(v0->getMap()->getMaxGlobalIndex(), comm->getSize() * 20 + comm->getRank() * 5 + 4);
-      TEST_EQUALITY(v0->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 25 - 1);
-
-      TEST_EQUALITY(v1->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v1->getMap()->getMaxLocalIndex(), 14);
-      TEST_EQUALITY(v1->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v1->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
-      TEST_EQUALITY(v1->getMap()->getMaxGlobalIndex(), comm->getSize() * 5 + comm->getRank() * 10 + 9);
-      TEST_EQUALITY(v1->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 15 - 1);
-
-      Teuchos::RCP<BlockedCrsMatrix> b00 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(0,0));
-      TEST_EQUALITY(b00->Rows(), 2);
-      TEST_EQUALITY(b00->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me00 = b00->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v00 = me00->ExtractVector(v0,0,true);
-      Teuchos::RCP<MultiVector> v01 = me00->ExtractVector(v0,1,true);
-      TEST_EQUALITY((v00->getData(0))[0], Teuchos::as<Scalar>(0.25));
-      TEST_EQUALITY((v01->getData(0))[0], Teuchos::as<Scalar>(1.0));
-      TEST_EQUALITY(v00->getLocalLength(), 20);
-      TEST_EQUALITY(v01->getLocalLength(), 5);
-      TEST_EQUALITY(v00->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 20));
-      TEST_EQUALITY(v01->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-
-      TEST_EQUALITY(v00->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v00->getMap()->getMaxLocalIndex(), 19);
-      TEST_EQUALITY(v00->getMap()->getMinAllGlobalIndex(),0);
-      TEST_EQUALITY(v00->getMap()->getMinGlobalIndex(), comm->getRank() * 20);
-      TEST_EQUALITY(v00->getMap()->getMaxGlobalIndex(), comm->getRank() * 20 + 19);
-      TEST_EQUALITY(v00->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 20 - 1);
-
-      TEST_EQUALITY(v01->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v01->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v01->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v01->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
-      TEST_EQUALITY(v01->getMap()->getMaxGlobalIndex(), comm->getRank() * 5 + 4);
-      TEST_EQUALITY(v01->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 5 - 1);
-
-      Teuchos::RCP<BlockedCrsMatrix> b11 = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedbA->getMatrix(1,1));
-      TEST_EQUALITY(b11->Rows(), 2);
-      TEST_EQUALITY(b11->Cols(), 2);
-
-      Teuchos::RCP<const MapExtractor> me11 = b11->getDomainMapExtractor();
-      Teuchos::RCP<MultiVector> v10 = me11->ExtractVector(v1,0,true);
-      Teuchos::RCP<MultiVector> v11 = me11->ExtractVector(v1,1,true);
-      TEST_EQUALITY((v10->getData(0))[0], Teuchos::as<Scalar>(0.5));
-      TEST_EQUALITY(v10->getLocalLength(), 5);
-      TEST_EQUALITY(v11->getLocalLength(), 10);
-      TEST_EQUALITY(v10->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 5));
-      TEST_EQUALITY(v11->getGlobalLength(), Teuchos::as<size_t>(comm->getSize() * 10));
-
-      TEST_EQUALITY(v10->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v10->getMap()->getMaxLocalIndex(), 4);
-      TEST_EQUALITY(v10->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v10->getMap()->getMinGlobalIndex(), comm->getRank() * 5);
-      TEST_EQUALITY(v10->getMap()->getMaxGlobalIndex(), comm->getRank() * 5 + 4);
-      TEST_EQUALITY(v10->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 5 - 1);
-
-      TEST_EQUALITY(v11->getMap()->getMinLocalIndex(), 0);
-      TEST_EQUALITY(v11->getMap()->getMaxLocalIndex(), 9);
-      TEST_EQUALITY(v11->getMap()->getMinAllGlobalIndex(), 0);
-      TEST_EQUALITY(v11->getMap()->getMinGlobalIndex(), comm->getRank() * 10);
-      TEST_EQUALITY(v11->getMap()->getMaxGlobalIndex(), comm->getRank() * 10 + 9);
-      TEST_EQUALITY(v11->getMap()->getMaxAllGlobalIndex(), comm->getSize() * 10 - 1);
-
-      Teuchos::Array<magnitude_type> n00(1); v00->norm1(n00);
-      Teuchos::Array<magnitude_type> n11(1); v01->norm1(n11);
-      Teuchos::Array<magnitude_type> n22(1); v10->norm1(n22);
-      Teuchos::Array<magnitude_type> n33(1); v11->norm1(n33);
-
-      TEST_EQUALITY(n00[0], Teuchos::as<Scalar>(v00->getGlobalLength() * 0.25));
-      TEST_EQUALITY(n11[0], Teuchos::as<Scalar>(v01->getGlobalLength() * 1.0));
-      TEST_EQUALITY(n22[0], Teuchos::as<Scalar>(v10->getGlobalLength() * 0.5));
-      TEST_FLOATING_EQUALITY(n33[0], v11->getGlobalLength() / 3.0, Teuchos::ScalarTraits<magnitude_type>::eps());
-    } // end UseTpetra
-  }
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      int noBlocks = 2;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib, noBlocks, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",MueLu::NoFactory::getRCP());
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX = bX->Merge();
 
-      RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->describe(out, Teuchos::VERB_EXTREME);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp(new SimpleSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    }
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",MueLu::NoFactory::getRCP());
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    RCP<Matrix> reorderedA                     = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
+    RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<MultiVector> X   = bX->Merge();
+    RCP<MultiVector> RHS = bRHS->Merge();
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < X->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Random X
+    {
+      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
+      RCP<BlockedMultiVector> test                        = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
+      bX.swap(test);
+      test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
+      bRHS.swap(test);
+    }
 
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    bRHS->putScalar((SC)1.0);
+    bX->setSeed(846930886);
+    bX->randomize();
+
+    RHS = bRHS->Merge();
+    X   = bX->Merge();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    bop->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_Thyra_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp(new SimpleSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoPredict->SetFactory("A", sA[0]);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA[0]);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    }
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      simpleSmoother->Apply(*X, *RHS, false); //nonzero initial guess
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoPredict->SetFactory("A", sA[0]);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA[0]);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoPredict->SetSchurCompFactoryManager(ssM);
-      }
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp(new SimpleSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    }
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i< 10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_SIMPLE_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 BGS for the prediction eq.
+    RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetFactory("A", sA[0]);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA[0]);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoPredict->SetSchurCompFactoryManager(ssM);
-      }
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
+    paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 4e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 2e-3, out, success);
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_SIMPLE_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<SimpleSmoother> smootherPrototype = rcp(new SimpleSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+    smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 Simple for the prediction eq.
+    RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp(new SimpleSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoCorrect->SetSchurCompFactoryManager(ssM);
+    }
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    smootherPrototype->SetSchurCompFactoryManager(sM[1]);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i< 10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      RHS->describe(out, Teuchos::VERB_EXTREME);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 60e-4, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 25e-4, out, success);
+  }  // end UseTpetra
+}
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    int noBlocks                             = 2;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, noBlocks, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      // TODO test only Tpetra because of Ifpack2 smoother!
-      MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-        RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-        Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-        Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-        Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-        Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-        //I don't use the testApply infrastructure because it has no provision for an initial guess.
-        Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-        level.Set("A", A);
-
-        // Test ReorderBlockAFactory
-        Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-        rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-        rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-        //////////////////////////////////////////////////////////////////////
-        // Smoothers
-        RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-        smootherPrototype->SetFactory("A",rAFact);
-        smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-        smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-        smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-        std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-        std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-        std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-        // prediction
-        std::string strInfo = std::string("{ 1 }");
-        sA[0] = rcp(new SubBlockAFactory());
-        sA[0]->SetFactory("A",rAFact);
-        sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-        sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-        sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProtoCorrect->SetFactory("A", sA[0]);
-        sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-        sM[0] = rcp(new FactoryManager());
-        sM[0]->SetFactory("A", sA[0]);
-        sM[0]->SetFactory("Smoother", sF[0]);
-        sM[0]->SetIgnoreUserData(true);
-
-        smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-        // correction
-        // define SchurComplement Factory
-        // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-        // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-        // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-        RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-        AinvFact->SetFactory("A",rAFact);
-
-        RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-        SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-        SFact->SetFactory("A",rAFact);
-        SFact->SetFactory("Ainv", AinvFact);
-
-        // create a 2x2 SIMPLE for the prediction eq.
-        RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp( new SimpleSmoother() );
-        smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-        smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-        smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-        smoProtoPredict->SetFactory("A", SFact);
-
-        for(int l = 0; l < 2; l++) {
-          Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-          ssA->SetFactory("A",SFact);
-          ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-          ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-          ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-          ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-          RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-          ssP->SetFactory("A", ssA);
-          Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-          Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-          ssM->SetFactory("A", ssA);
-          ssM->SetFactory("Smoother", ssF);
-          ssM->SetIgnoreUserData(true);
-          if(l == 0) smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
-          else smoProtoPredict->SetSchurCompFactoryManager(ssM);
-        }
-
-        sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-        sM[1] = rcp(new FactoryManager());
-        sM[1]->SetFactory("A", SFact);
-        sM[1]->SetFactory("Smoother", sF[1]);
-        sM[1]->SetIgnoreUserData(true);
-
-        smootherPrototype->SetSchurCompFactoryManager(sM[1]);
-
-        RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-        // main factory manager
-        FactoryManager M;
-        M.SetFactory("Smoother",     smootherFact);
-        M.SetFactory("A",            rAFact);
-
-        MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-        // request BGS smoother (and all dependencies) on level
-        level.Request("A", rAFact.get());
-        level.Request("Smoother", smootherFact.get());
-        level.Request("PreSmoother", smootherFact.get());
-        level.Request("PostSmoother", smootherFact.get());
-
-        //smootherFact->DeclareInput(level);
-        smootherFact->Build(level);
-
-        level.print(std::cout, Teuchos::VERB_EXTREME);
-
-        RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-        RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-        RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
-
-        TEST_EQUALITY(reorderedbA->Rows(), 2);
-        TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-        RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
-        RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
-
-        RCP<MultiVector> X = bX->Merge();
-        RCP<MultiVector> RHS = bRHS->Merge();
-
-        // apply simple smoother
-        RHS->putScalar((SC) 1.0);
-        X->putScalar((SC) 0.0);
-
-        // solve system
-        simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-        Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
-        bool bCheck = true;
-        for(size_t i=0; i<X->getLocalLength(); i++) {
-          if (i < 5) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-          if (i>=5 && i< 10) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-          if (i>=10 && i< 20) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        }
-        TEST_EQUALITY(bCheck, true);
-
-        // Random X
-        {
-          Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
-          RCP<BlockedMultiVector> test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
-          bX.swap(test);
-          test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
-          bRHS.swap(test);
-        }
-
-        bRHS->putScalar((SC) 1.0);
-        bX->setSeed(846930886);
-        bX->randomize();
-
-        RHS = bRHS->Merge();
-        X = bX->Merge();
-
-        typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-        // Normalize X
-        Array<magnitude_type> norms(1); X->norm2(norms);
-        X->scale(1/norms[0]);
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
 
-        // Compute RHS corresponding to X
-        bop->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
 
-        // Reset X to 0
-        X->putScalar((SC) 0.0);
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", MueLu::NoFactory::getRCP());
 
-        RHS->norm2(norms);
-        out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
 
-        out << "solve with zero initial guess" << std::endl;
-        Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-        out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-        simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    smootherPrototype->AddFactoryManager(sM[0], 0);
 
-        Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-        Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-        out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-        out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-        magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-        TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-        TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-      }// end useTpetra
-    }
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_Thyra_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoPredict->SetFactory("A", sA[0]);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA[0]);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoPredict->SetSchurCompFactoryManager(ssM);
-      }
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    simpleSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i< 10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp(new BraessSarazinSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    std::string strInfo                = std::string("{ 1 }");
+    Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+    ssA->SetFactory("A", SFact);
+    ssA->SetParameter("block row", Teuchos::ParameterEntry(1));  // local block indices relative to size of blocked operator
+    ssA->SetParameter("block col", Teuchos::ParameterEntry(1));
+    ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+    RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    ssP->SetFactory("A", ssA);
+    Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+    Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+    ssM->SetFactory("A", ssA);
+    ssM->SetFactory("Smoother", ssF);
+    ssM->SetIgnoreUserData(true);
+    smoProtoCorrect->AddFactoryManager(ssM, 0);
+
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_BS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp(new BraessSarazinSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    std::string strInfo                = std::string("{ 1 }");
+    Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+    ssA->SetFactory("A", SFact);
+    ssA->SetParameter("block row", Teuchos::ParameterEntry(1));  // local block indices relative to size of blocked operator
+    ssA->SetParameter("block col", Teuchos::ParameterEntry(1));
+    ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+    RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    ssP->SetFactory("A", ssA);
+    Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+    Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+    ssM->SetFactory("A", ssA);
+    ssM->SetFactory("Smoother", ssF);
+    ssM->SetIgnoreUserData(true);
+    smoProtoCorrect->AddFactoryManager(ssM, 0);
+
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA                     = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
+    RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
+
+    RCP<MultiVector> X   = bX->Merge();
+    RCP<MultiVector> RHS = bRHS->Merge();
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < X->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // Random X
+    {
+      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
+      RCP<BlockedMultiVector> test                        = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
+      bX.swap(test);
+      test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
+      bRHS.swap(test);
+    }
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    bRHS->putScalar((SC)1.0);
+    bX->setSeed(846930886);
+    bX->randomize();
+
+    RHS = bRHS->Merge();
+    X   = bX->Merge();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    bop->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_SIMPLE_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<SimpleSmoother> smoProtoPredict = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoPredict->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoPredict->SetSchurCompFactoryManager(ssM);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII02I1I_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 2] 1]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
       }
+      if (i >= 5 && i < 15) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 0 1 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp(new SimpleSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    std::string strInfo = std::string("{ 1 }");
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoCorrect->SetSchurCompFactoryManager(ssM);
+    }
 
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->AddFactoryManager(sM[0], 0);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    smootherFact->Build(level);
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i< 10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
+    // solve system
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp(new BraessSarazinSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    std::string strInfo                = std::string("{ 1 }");
+    Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+    ssA->SetFactory("A", SFact);
+    ssA->SetParameter("block row", Teuchos::ParameterEntry(1));  // local block indices relative to size of blocked operator
+    ssA->SetParameter("block col", Teuchos::ParameterEntry(1));
+    ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+    RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    ssP->SetFactory("A", ssA);
+    Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+    Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+    ssM->SetFactory("A", ssA);
+    ssM->SetFactory("Smoother", ssF);
+    ssM->SetIgnoreUserData(true);
+    smoProtoCorrect->AddFactoryManager(ssM, 0);
+
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    smootherFact->Build(level);
+
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
+
+    // solve system
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedbA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_BS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(30)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.4)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
+    paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
+
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 3e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 1e-3, out, success);
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_BS_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<BraessSarazinSmoother> smootherPrototype = rcp(new BraessSarazinSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+
+    std::vector<RCP<SmootherFactory> > sF(1, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(1, Teuchos::null);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 Simple for the prediction eq.
+    std::string strInfo                 = std::string("{ 1 }");
+    RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp(new SimpleSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      if (l == 0)
+        smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
+      else
+        smoProtoCorrect->SetSchurCompFactoryManager(ssM);
+    }
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
 
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", SFact);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_SIMPLE_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    smootherPrototype->AddFactoryManager(sM[0], 0);
 
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
+    smootherFact->Build(level);
 
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 BGS for the prediction eq.
-      RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetFactory("A", sA[0]);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA[0]);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
+    RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
-      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    bsSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 44e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 15e-3, out, success);
+  }  // end UseTpetra
+}
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A",rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    int noBlocks                             = 2;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks, Teuchos::null);
+    for (int k = 0; k < noBlocks; k++) {
+      std::string strInfo = std::string("{ 1 }");
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", MueLu::NoFactory::getRCP());
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
+
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
+
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RCP<SmootherBase> uzSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 4e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 2e-3, out, success);
-    } // end UseTpetra
-  }
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_SIMPLE_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<SimpleSmoother> smootherPrototype     = rcp( new SimpleSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-      smootherPrototype->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->SetVelocityPredictionFactoryManager(sM[0]);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 Simple for the prediction eq.
-      RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoCorrect->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoCorrect->SetSchurCompFactoryManager(ssM);
-      }
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      smootherPrototype->SetSchurCompFactoryManager(sM[1]);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
+    uzSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    uzSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp(new UzawaSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 60e-4, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 25e-4, out, success);
-    } // end UseTpetra
-  }
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      int noBlocks = 2;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib, noBlocks, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp(new UzawaSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",MueLu::NoFactory::getRCP());
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      smootherPrototype->AddFactoryManager(sM[0],0);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    RCP<Matrix> reorderedA                     = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
+    RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
+    RCP<MultiVector> X   = bX->Merge();
+    RCP<MultiVector> RHS = bRHS->Merge();
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
+    bool bCheck                           = true;
+    X->describe(out, Teuchos::VERB_EXTREME);
+    for (size_t i = 0; i < X->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // Random X
+    {
+      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
+      RCP<BlockedMultiVector> test                        = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
+      bX.swap(test);
+      test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
+      bRHS.swap(test);
+    }
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    bRHS->putScalar((SC)1.0);
+    bX->setSeed(846930886);
+    bX->randomize();
+
+    RHS = bRHS->Merge();
+    X   = bX->Merge();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    bop->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp(new UzawaSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      simpleSmoother->Apply(*X, *RHS, false); //nonzero initial guess
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII20I1I_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [2 0] 1]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10 ) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10  && i<15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i<20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // solve system
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_Uzawa_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 BGS for the prediction eq.
+    RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetFactory("A", sA[0]);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA[0]);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    smootherPrototype->AddFactoryManager(sM[0], 0);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
+    paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp( new BraessSarazinSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetFactory("A", SFact);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      std::string strInfo = std::string("{ 1 }");
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 8e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 5e-3, out, success);
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_Uzawa_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<UzawaSmoother> smootherPrototype = rcp(new UzawaSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 Simple for the prediction eq.
+    RCP<UzawaSmoother> smoProtoCorrect = Teuchos::rcp(new UzawaSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
       Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-      ssA->SetFactory("A",SFact);
-      ssA->SetParameter("block row",Teuchos::ParameterEntry(1)); // local block indices relative to size of blocked operator
-      ssA->SetParameter("block col",Teuchos::ParameterEntry(1));
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
       ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
       ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
       RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
       ssP->SetFactory("A", ssA);
       Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-      Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
       ssM->SetFactory("A", ssA);
       ssM->SetFactory("Smoother", ssF);
       ssM->SetIgnoreUserData(true);
-      smoProtoCorrect->AddFactoryManager(ssM,0);
-
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
+      smoProtoCorrect->AddFactoryManager(ssM, l);
+    }
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      smootherPrototype->AddFactoryManager(sM[0],0);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      smootherFact->Build(level);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // solve system
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10 ) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10  && i<15) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-        if (i>=15 && i<20) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    simpleSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 11e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 5e-3, out, success);
+  }  // end UseTpetra
+}
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    int noBlocks                             = 2;
+    Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node, TpetraMap>(noBlocks, *comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_BS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp( new BraessSarazinSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetFactory("A", SFact);
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
 
+    std::vector<RCP<SubBlockAFactory> > sA(noBlocks, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(noBlocks, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(noBlocks, Teuchos::null);
+    for (int k = 0; k < noBlocks; k++) {
       std::string strInfo = std::string("{ 1 }");
-      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-      ssA->SetFactory("A",SFact);
-      ssA->SetParameter("block row",Teuchos::ParameterEntry(1)); // local block indices relative to size of blocked operator
-      ssA->SetParameter("block col",Teuchos::ParameterEntry(1));
-      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      ssP->SetFactory("A", ssA);
-      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-      Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-      ssM->SetFactory("A", ssA);
-      ssM->SetFactory("Smoother", ssF);
-      ssM->SetIgnoreUserData(true);
-      smoProtoCorrect->AddFactoryManager(ssM,0);
-
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
+      sA[k]               = rcp(new SubBlockAFactory());
+      sA[k]->SetFactory("A", MueLu::NoFactory::getRCP());
+      sA[k]->SetParameter("block row", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("block col", Teuchos::ParameterEntry(k));
+      sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+      RCP<SmootherPrototype> smoProto = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      smoProto->SetFactory("A", sA[k]);
+      sF[k] = rcp(new SmootherFactory(smoProto));
+
+      sM[k] = rcp(new FactoryManager());
+      sM[k]->SetFactory("A", sA[k]);
+      sM[k]->SetFactory("Smoother", sF[k]);
+      sM[k]->SetIgnoreUserData(true);
+
+      smootherPrototype->AddFactoryManager(sM[k], k);
+    }
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    // request BGS smoother (and all dependencies) on level
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      smootherFact->Build(level);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
 
-      RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
-      RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      RCP<MultiVector> X = bX->Merge();
-      RCP<MultiVector> RHS = bRHS->Merge();
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      // solve system
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<X->getLocalLength(); i++) {
-        if (i < 5) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=5 && i< 10) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-        if (i>=10 && i< 20) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      {
-        Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
-        RCP<BlockedMultiVector> test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
-        bX.swap(test);
-        test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
-        bRHS.swap(test);
-      }
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      bRHS->putScalar((SC) 1.0);
-      bX->setSeed(846930886);
-      bX->randomize();
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      RHS = bRHS->Merge();
-      X = bX->Merge();
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      // Compute RHS corresponding to X
-      bop->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    inSmoother->Apply(*X, *RHS, false);  // nonzero initial guess
 
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
+    }
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 block smoother for the prediction eq.
+    RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp(new IndefBlockedDiagonalSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII02I1I_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 2] 1]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<5) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=5  && i< 15) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    // request block smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 0 1 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoCorrect->SetFactory("A", SFact);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      std::string strInfo = std::string("{ 1 }");
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoCorrect->SetSchurCompFactoryManager(ssM);
+    // solve system
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 block smoother for the prediction eq.
+    RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp(new IndefBlockedDiagonalSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      // solve system
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10  && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // request block smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    RCP<Matrix> reorderedA                     = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
+    RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<MultiVector> X   = bX->Merge();
+    RCP<MultiVector> RHS = bRHS->Merge();
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    // solve system
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
+    Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < X->getLocalLength(); i++) {
+      if (i < 5) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 5 && i < 10) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+      if (i >= 10 && i < 20) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I10II_Thyra_BS_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [ 1 0 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      RCP<BraessSarazinSmoother> smoProtoCorrect = Teuchos::rcp( new BraessSarazinSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetFactory("A", SFact);
+    // Random X
+    {
+      Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
+      RCP<BlockedMultiVector> test                        = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
+      bX.swap(test);
+      test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
+      bRHS.swap(test);
+    }
 
-      std::string strInfo = std::string("{ 1 }");
+    bRHS->putScalar((SC)1.0);
+    bX->setSeed(846930886);
+    bX->randomize();
+
+    RHS = bRHS->Merge();
+    X   = bX->Merge();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    bop->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoCorrect->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoCorrect));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 SIMPLE for the prediction eq.
+    RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp(new IndefBlockedDiagonalSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoPredict->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
       Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-      ssA->SetFactory("A",SFact);
-      ssA->SetParameter("block row",Teuchos::ParameterEntry(1)); // local block indices relative to size of blocked operator
-      ssA->SetParameter("block col",Teuchos::ParameterEntry(1));
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
       ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
       ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
       RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
       ssP->SetFactory("A", ssA);
       Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-      Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
       ssM->SetFactory("A", ssA);
       ssM->SetFactory("Smoother", ssF);
       ssM->SetIgnoreUserData(true);
-      smoProtoCorrect->AddFactoryManager(ssM,0);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
+
+    sF[1] = rcp(new SmootherFactory(smoProtoPredict));
 
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      smootherPrototype->AddFactoryManager(sM[0],0);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      smootherFact->Build(level);
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
+    // apply simple smoother
+    RHS->putScalar((SC)1.0);
+    X->putScalar((SC)0.0);
 
-      // solve system
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10  && i< 15) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 1.0) bCheck = false; }
+    // solve system
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+    RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
+    TEST_EQUALITY(bX.is_null(), false);
+    RCP<MultiVector> XX                   = bX->Merge();
+    Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
+    bool bCheck                           = true;
+    for (size_t i = 0; i < XX->getLocalLength(); i++) {
+      if (i < 10) {
+        if (xdata[i] != (SC)(1.0 / 3.0)) bCheck = false;
       }
-      TEST_EQUALITY(bCheck, true);
+      if (i >= 10 && i < 15) {
+        if (xdata[i] != (SC)1.0) bCheck = false;
+      }
+      if (i >= 15 && i < 20) {
+        if (xdata[i] != (SC)0.5) bCheck = false;
+      }
+    }
+    TEST_EQUALITY(bCheck, true);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    magnitude_type tol = 50. * Teuchos::ScalarTraits<Scalar>::eps();
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
+    TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
+  }  // end useTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_Indef_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    // create a 2x2 BGS for the prediction eq.
+    RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp(new BlockedGaussSeidelSmoother());
+    smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoPredict->SetFactory("A", sA[0]);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", sA[0]);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoPredict->AddFactoryManager(ssM, l);
+    }
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    smootherPrototype->AddFactoryManager(sM[0], 0);
 
-      // Compute RHS corresponding to X
-      reorderedbA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
+    paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
+    RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
+    smoProtoCorrect->SetFactory("A", SFact);
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedbA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
+
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
+
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 11e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 7e-3, out, success);
+  }  // end UseTpetra
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_Indef_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+    Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
+    Teuchos::RCP<const Matrix> Aconst        = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
+    Teuchos::RCP<Matrix> A                   = Teuchos::rcp_const_cast<Matrix>(Aconst);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
+
+    //////////////////////////////////////////////////////////////////////
+    // Smoothers
+    RCP<IndefBlockedDiagonalSmoother> smootherPrototype = rcp(new IndefBlockedDiagonalSmoother());
+    smootherPrototype->SetFactory("A", rAFact);
+    smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
+    smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
+
+    std::vector<RCP<SubBlockAFactory> > sA(1, Teuchos::null);
+    std::vector<RCP<SmootherFactory> > sF(2, Teuchos::null);
+    std::vector<RCP<FactoryManager> > sM(2, Teuchos::null);
+
+    // prediction
+    std::string strInfo = std::string("{ 1 }");
+    sA[0]               = rcp(new SubBlockAFactory());
+    sA[0]->SetFactory("A", rAFact);
+    sA[0]->SetParameter("block row", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("block col", Teuchos::ParameterEntry(0));
+    sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+    sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+
+    RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+    smoProtoPredict->SetFactory("A", sA[0]);
+    sF[0] = rcp(new SmootherFactory(smoProtoPredict));
+
+    sM[0] = rcp(new FactoryManager());
+    sM[0]->SetFactory("A", sA[0]);
+    sM[0]->SetFactory("Smoother", sF[0]);
+    sM[0]->SetIgnoreUserData(true);
+
+    smootherPrototype->AddFactoryManager(sM[0], 0);
+
+    // correction
+    // define SchurComplement Factory
+    // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
+    // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
+    // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
+    RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
+    AinvFact->SetFactory("A", rAFact);
+
+    RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
+    SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));  // for Simple, omega is always 1.0 in the SchurComplement
+    SFact->SetFactory("A", rAFact);
+    SFact->SetFactory("Ainv", AinvFact);
+
+    // create a 2x2 Simple for the prediction eq.
+    RCP<IndefBlockedDiagonalSmoother> smoProtoCorrect = Teuchos::rcp(new IndefBlockedDiagonalSmoother());
+    smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
+    smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
+    smoProtoCorrect->SetFactory("A", SFact);
+
+    for (int l = 0; l < 2; l++) {
+      Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
+      ssA->SetFactory("A", SFact);
+      ssA->SetParameter("block row", Teuchos::ParameterEntry(l));  // local block indices relative to size of blocked operator
+      ssA->SetParameter("block col", Teuchos::ParameterEntry(l));
+      ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
+      ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
+      RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
+      ssP->SetFactory("A", ssA);
+      Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
+      Teuchos::RCP<FactoryManager> ssM  = Teuchos::rcp(new FactoryManager());
+      ssM->SetFactory("A", ssA);
+      ssM->SetFactory("Smoother", ssF);
+      ssM->SetIgnoreUserData(true);
+      smoProtoCorrect->AddFactoryManager(ssM, l);
+    }
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
+    sF[1] = rcp(new SmootherFactory(smoProtoCorrect));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_BS_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(30)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.4)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
+    sM[1] = rcp(new FactoryManager());
+    sM[1]->SetFactory("A", SFact);
+    sM[1]->SetFactory("Smoother", sF[1]);
+    sM[1]->SetIgnoreUserData(true);
 
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
-      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
+    smootherPrototype->AddFactoryManager(sM[1], 1);
 
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
+    RCP<SmootherFactory> smootherFact = rcp(new SmootherFactory(smootherPrototype));
 
-      smootherPrototype->AddFactoryManager(sM[0],0);
+    // main factory manager
+    FactoryManager M;
+    M.SetFactory("Smoother", smootherFact);
+    M.SetFactory("A", rAFact);
 
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
+    MueLu::SetFactoryManager SFM(Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
 
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+    level.Request("Smoother", smootherFact.get());
+    level.Request("PreSmoother", smootherFact.get());
+    level.Request("PostSmoother", smootherFact.get());
 
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
+    // smootherFact->DeclareInput(level);
+    smootherFact->Build(level);
 
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A",rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
+    level.print(std::cout, Teuchos::VERB_EXTREME);
 
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
+    RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
 
-      level.print(std::cout, Teuchos::VERB_EXTREME);
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
 
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
+    RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(), 1);
 
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
 
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
+    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
 
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+    // Compute RHS corresponding to X
+    reorderedA->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
+    // Reset X to 0
+    X->putScalar((SC)0.0);
 
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+    inSmoother->Apply(*X, *RHS, true);  // zero initial guess
 
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
 
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 25e-3, out, success);
+    TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 9e-3, out, success);
+  }  // end UseTpetra
+}
 
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, SplitReorder, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  // TODO test only Tpetra because of Ifpack2 smoother!
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+    Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
 
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 3e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 1e-3, out, success);
-    } // end UseTpetra
-  }
+    Teuchos::RCP<const Matrix> Aconst = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(comm->getSize() * 3, -1, lib);
+    Teuchos::RCP<Matrix> A            = Teuchos::rcp_const_cast<Matrix>(Aconst);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_BS_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<BraessSarazinSmoother> smootherPrototype     = rcp( new BraessSarazinSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-
-      std::vector<RCP<SmootherFactory> >  sF (1, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (1, Teuchos::null);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 Simple for the prediction eq.
-      std::string strInfo = std::string("{ 1 }");
-      RCP<SimpleSmoother> smoProtoCorrect = Teuchos::rcp( new SimpleSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetParameter("UseSIMPLEC", Teuchos::ParameterEntry(false));
-      smoProtoCorrect->SetFactory("A", SFact);
+    // split local parts of row map
+    Teuchos::RCP<const Map> map = A->getRowMap();
 
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        if(l == 0) smoProtoCorrect->SetVelocityPredictionFactoryManager(ssM);
-        else smoProtoCorrect->SetSchurCompFactoryManager(ssM);
-      }
-
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", SFact);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> bsSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      bsSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 44e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 15e-3, out, success);
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 2;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks, Teuchos::null);
-      for (int k = 0; k < noBlocks; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",MueLu::NoFactory::getRCP());
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
-      }
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> uzSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      uzSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      uzSmoother->Apply(*X, *RHS, false); //nonzero initial guess
-
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp( new UzawaSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp( new UzawaSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
-      RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
-
-      RCP<MultiVector> X = bX->Merge();
-      RCP<MultiVector> RHS = bRHS->Merge();
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
-      bool bCheck = true;
-      X->describe(out, Teuchos::VERB_EXTREME);
-      for(size_t i=0; i<X->getLocalLength(); i++) {
-        if (i < 5) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=5 && i< 10) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-        if (i>=10 && i< 20) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      {
-        Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
-        RCP<BlockedMultiVector> test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
-        bX.swap(test);
-        test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
-        bRHS.swap(test);
-      }
-
-      bRHS->putScalar((SC) 1.0);
-      bX->setSeed(846930886);
-      bX->randomize();
-
-      RHS = bRHS->Merge();
-      X = bX->Merge();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      bop->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_Uzawa_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<UzawaSmoother> smoProtoPredict = Teuchos::rcp( new UzawaSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_Uzawa_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 BGS for the prediction eq.
-      RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetFactory("A", sA[0]);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA[0]);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
-      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A",rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 8e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 5e-3, out, success);
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_Uzawa_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<UzawaSmoother> smootherPrototype     = rcp( new UzawaSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 Simple for the prediction eq.
-      RCP<UzawaSmoother> smoProtoCorrect = Teuchos::rcp( new UzawaSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoCorrect->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> simpleSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      simpleSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 11e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 5e-3, out, success);
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-      int noBlocks = 2;
-      Teuchos::RCP<const BlockedCrsMatrix> bop = CreateBlockDiagonalExampleMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node,TpetraMap>(noBlocks, *comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-
-      std::vector<RCP<SubBlockAFactory> > sA (noBlocks, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (noBlocks, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (noBlocks, Teuchos::null);
-      for (int k = 0; k < noBlocks; k++) {
-        std::string strInfo = std::string("{ 1 }");
-        sA[k] = rcp(new SubBlockAFactory());
-        sA[k]->SetFactory("A",MueLu::NoFactory::getRCP());
-        sA[k]->SetParameter("block row",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("block col",Teuchos::ParameterEntry(k));
-        sA[k]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        sA[k]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-        RCP<SmootherPrototype> smoProto     = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        smoProto->SetFactory("A", sA[k]);
-        sF[k] = rcp( new SmootherFactory(smoProto) );
-
-        sM[k] = rcp(new FactoryManager());
-        sM[k]->SetFactory("A", sA[k]);
-        sM[k]->SetFactory("Smoother", sF[k]);
-        sM[k]->SetIgnoreUserData(true);
-
-        smootherPrototype->AddFactoryManager(sM[k],k);
-      }
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, false); //nonzero initial guess
-
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm2[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    Teuchos::Array<GlobalOrdinal> myGids1;
+    Teuchos::Array<GlobalOrdinal> myGids2;
+    GlobalOrdinal count1 = 0;
+    GlobalOrdinal count2 = 0;
+    for (size_t i = 0; i < map->getLocalNumElements(); ++i) {
+      const GlobalOrdinal gid = map->getGlobalElement(i);
+      if (gid % 2 == 0) {
+        myGids1.push_back(gid);
+        count1++;
       } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 block smoother for the prediction eq.
-      RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp( new IndefBlockedDiagonalSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
+        myGids2.push_back(gid);
+        count2++;
       }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request block smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrix(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 block smoother for the prediction eq.
-      RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp( new IndefBlockedDiagonalSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request block smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<ReorderedBlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<ReorderedBlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<BlockedMultiVector> bX   = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedDomainMap(), 1, true));
-      RCP<BlockedMultiVector> bRHS = Teuchos::rcp(new BlockedMultiVector(bop->getBlockedRangeMap(), 1, true));
-
-      RCP<MultiVector> X = bX->Merge();
-      RCP<MultiVector> RHS = bRHS->Merge();
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::ArrayRCP<const Scalar> xdata = X->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<X->getLocalLength(); i++) {
-        if (i < 5) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=5 && i< 10) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-        if (i>=10 && i< 20) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      {
-        Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ 2 [0 1]]");
-        RCP<BlockedMultiVector> test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bX));
-        bX.swap(test);
-        test = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(buildReorderedBlockedMultiVector(brm, bRHS));
-        bRHS.swap(test);
-      }
-
-      bRHS->putScalar((SC) 1.0);
-      bX->setSeed(846930886);
-      bX->randomize();
-
-      RHS = bRHS->Merge();
-      X = bX->Merge();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      bop->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI2I01II_Thyra_Indef_Setup_Apply, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlockDiagonalExampleMatrixThyra(lib,3, comm);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 2 [0 1]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(1)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoCorrect->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 SIMPLE for the prediction eq.
-      RCP<IndefBlockedDiagonalSmoother> smoProtoPredict = Teuchos::rcp( new IndefBlockedDiagonalSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoPredict->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // apply simple smoother
-      RHS->putScalar((SC) 1.0);
-      X->putScalar((SC) 0.0);
-
-      // solve system
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-      RCP<BlockedMultiVector> bX = Teuchos::rcp_dynamic_cast<BlockedMultiVector>(X);
-      TEST_EQUALITY(bX.is_null(), false);
-      RCP<MultiVector> XX = bX->Merge();
-      Teuchos::ArrayRCP<const Scalar> xdata = XX->getData(0);
-      bool bCheck = true;
-      for(size_t i=0; i<XX->getLocalLength(); i++) {
-        if (i<10) { if(xdata[i] != (SC) (1.0/3.0)) bCheck = false; }
-        if (i>=10 && i< 15) { if(xdata[i] != (SC) 1.0) bCheck = false; }
-        if (i>=15 && i< 20) { if(xdata[i] != (SC) 0.5) bCheck = false; }
-      }
-      TEST_EQUALITY(bCheck, true);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      magnitude_type tol = 50.*Teuchos::ScalarTraits<Scalar>::eps();
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, tol, out, success);
-      TEUCHOS_TEST_COMPARE(finalNorms[0] - Teuchos::ScalarTraits<Scalar>::magnitude(Teuchos::ScalarTraits<Scalar>::one()), <, tol, out, success);
-    }// end useTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedII01I2I_Thyra_Indef_Setup_Apply2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ [0 1] 2]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      // create a 2x2 BGS for the prediction eq.
-      RCP<BlockedGaussSeidelSmoother> smoProtoPredict = Teuchos::rcp( new BlockedGaussSeidelSmoother() );
-      smoProtoPredict->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoPredict->SetFactory("A", sA[0]);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",sA[0]);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoPredict->AddFactoryManager(ssM,l);
-      }
-
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: sweeps", Teuchos::as<LocalOrdinal>(30));
-      paramList.set("relaxation: damping factor", Teuchos::as<Scalar>(0.9));
-      RCP<SmootherPrototype> smoProtoCorrect = rcp(new Ifpack2Smoother(std::string("RELAXATION"), paramList, 0));
-      smoProtoCorrect->SetFactory("A", SFact);
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A",rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 11e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 7e-3, out, success);
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, NestedI0I21II_Thyra_Indef_Setup_Apply3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const BlockedCrsMatrix> bop = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::CreateBlocked3x3MatrixThyra(*comm, lib);
-      Teuchos::RCP<const Matrix> Aconst = Teuchos::rcp_dynamic_cast<const Matrix>(bop);
-      Teuchos::RCP<      Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 [ 2 1 ]]")));
-
-      //////////////////////////////////////////////////////////////////////
-      // Smoothers
-      RCP<IndefBlockedDiagonalSmoother> smootherPrototype     = rcp( new IndefBlockedDiagonalSmoother() );
-      smootherPrototype->SetFactory("A",rAFact);
-      smootherPrototype->SetParameter("Sweeps", Teuchos::ParameterEntry(Teuchos::as<LocalOrdinal>(15)));
-      smootherPrototype->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(0.9)));
-
-      std::vector<RCP<SubBlockAFactory> > sA (1, Teuchos::null);
-      std::vector<RCP<SmootherFactory> >  sF (2, Teuchos::null);
-      std::vector<RCP<FactoryManager> >   sM (2, Teuchos::null);
-
-      // prediction
-      std::string strInfo = std::string("{ 1 }");
-      sA[0] = rcp(new SubBlockAFactory());
-      sA[0]->SetFactory("A",rAFact);
-      sA[0]->SetParameter("block row",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("block col",Teuchos::ParameterEntry(0));
-      sA[0]->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-      sA[0]->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-
-      RCP<SmootherPrototype> smoProtoPredict = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-      smoProtoPredict->SetFactory("A", sA[0]);
-      sF[0] = rcp( new SmootherFactory(smoProtoPredict) );
-
-      sM[0] = rcp(new FactoryManager());
-      sM[0]->SetFactory("A", sA[0]);
-      sM[0]->SetFactory("Smoother", sF[0]);
-      sM[0]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[0],0);
-
-      // correction
-      // define SchurComplement Factory
-      // SchurComp gets a RCP to AFact_ which has to be the 2x2 blocked operator
-      // It stores the resulting SchurComplement operator as "A" generated by the SchurComplementFactory
-      // Instead of F^{-1} it uses the approximation \hat{F}^{-1} with \hat{F} = diag(F)
-      RCP<InverseApproximationFactory> AinvFact = Teuchos::rcp(new InverseApproximationFactory());
-      AinvFact->SetFactory("A",rAFact);
-
-      RCP<SchurComplementFactory> SFact = Teuchos::rcp(new SchurComplementFactory());
-      SFact->SetParameter("omega", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0))); // for Simple, omega is always 1.0 in the SchurComplement
-      SFact->SetFactory("A",rAFact);
-      SFact->SetFactory("Ainv", AinvFact);
-
-      // create a 2x2 Simple for the prediction eq.
-      RCP<IndefBlockedDiagonalSmoother> smoProtoCorrect = Teuchos::rcp( new IndefBlockedDiagonalSmoother() );
-      smoProtoCorrect->SetParameter("Sweeps", Teuchos::ParameterEntry(1));
-      smoProtoCorrect->SetParameter("Damping factor", Teuchos::ParameterEntry(Teuchos::as<Scalar>(1.0)));
-      smoProtoCorrect->SetFactory("A", SFact);
-
-      for(int l = 0; l < 2; l++) {
-        Teuchos::RCP<SubBlockAFactory> ssA = rcp(new SubBlockAFactory());
-        ssA->SetFactory("A",SFact);
-        ssA->SetParameter("block row",Teuchos::ParameterEntry(l)); // local block indices relative to size of blocked operator
-        ssA->SetParameter("block col",Teuchos::ParameterEntry(l));
-        ssA->SetParameter("Range map: Striding info", Teuchos::ParameterEntry(strInfo));
-        ssA->SetParameter("Domain map: Striding info", Teuchos::ParameterEntry(strInfo));
-        RCP<SmootherPrototype> ssP = rcp(new Ifpack2Smoother(std::string("RELAXATION"), Teuchos::ParameterList(), 0));
-        ssP->SetFactory("A", ssA);
-        Teuchos::RCP<SmootherFactory> ssF = Teuchos::rcp(new SmootherFactory(ssP));
-        Teuchos::RCP<FactoryManager> ssM = Teuchos::rcp(new FactoryManager());
-        ssM->SetFactory("A", ssA);
-        ssM->SetFactory("Smoother", ssF);
-        ssM->SetIgnoreUserData(true);
-        smoProtoCorrect->AddFactoryManager(ssM,l);
-      }
-
-      sF[1] = rcp( new SmootherFactory(smoProtoCorrect) );
-
-      sM[1] = rcp(new FactoryManager());
-      sM[1]->SetFactory("A", SFact);
-      sM[1]->SetFactory("Smoother", sF[1]);
-      sM[1]->SetIgnoreUserData(true);
-
-      smootherPrototype->AddFactoryManager(sM[1],1);
-
-
-      RCP<SmootherFactory>   smootherFact          = rcp( new SmootherFactory(smootherPrototype) );
-
-      // main factory manager
-      FactoryManager M;
-      M.SetFactory("Smoother",     smootherFact);
-      M.SetFactory("A",            rAFact);
-
-      MueLu::SetFactoryManager SFM (Teuchos::rcpFromRef(level), Teuchos::rcpFromRef(M));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-      level.Request("Smoother", smootherFact.get());
-      level.Request("PreSmoother", smootherFact.get());
-      level.Request("PostSmoother", smootherFact.get());
-
-      //smootherFact->DeclareInput(level);
-      smootherFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<SmootherBase> inSmoother = level.Get<RCP<SmootherBase> >("PreSmoother", smootherFact.get());
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(reorderedA->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(reorderedA->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      reorderedA->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      inSmoother->Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*reorderedA, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], <, 25e-3, out, success);
-      TEUCHOS_TEST_COMPARE(residualNorm1[0], >, 9e-3, out, success);
-    } // end UseTpetra
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(BlockedSmoother, SplitReorder, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    // TODO test only Tpetra because of Ifpack2 smoother!
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-      Teuchos::RCP<const Matrix> Aconst = TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build2DPoisson(comm->getSize() * 3,-1,lib);
-      Teuchos::RCP<Matrix> A = Teuchos::rcp_const_cast<Matrix>(Aconst);
-
-      // split local parts of row map
-      Teuchos::RCP<const Map> map = A->getRowMap();
-
-
-      Teuchos::Array<GlobalOrdinal> myGids1;
-      Teuchos::Array<GlobalOrdinal> myGids2;
-      GlobalOrdinal count1 = 0;
-      GlobalOrdinal count2 = 0;
-      for (size_t i=0; i<map->getLocalNumElements(); ++i) {
-        const GlobalOrdinal gid = map->getGlobalElement(i);
-        if (gid % 2 == 0) { myGids1.push_back(gid); count1++; }
-        else              { myGids2.push_back(gid); count2++; }
-      }
-      GlobalOrdinal gcount1 = 0;
-      GlobalOrdinal gcount2 = 0;
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count1, &gcount1);
-      Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count2, &gcount2);
-      Teuchos::RCP<const Map> map1 = MapFactory::Build(lib,gcount1,myGids1(),0,comm);
-      Teuchos::RCP<const Map> map2 = MapFactory::Build(lib,gcount2,myGids2(),0,comm);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Level level; TestHelpers::TestFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      level.Set("Map1", map1);
-      level.Set("Map2", map2);
-
-      // Test ReorderBlockAFactory
-      Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
-      rAFact->SetFactory("A",MueLu::NoFactory::getRCP());
-      rAFact->SetFactory("Map1",MueLu::NoFactory::getRCP());
-      rAFact->SetFactory("Map2",MueLu::NoFactory::getRCP());
-      rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 1 ]")));
-
-      // request BGS smoother (and all dependencies) on level
-      level.Request("A", rAFact.get());
-
-      //smootherFact->DeclareInput(level);
-      rAFact->Build(level);
-
-      level.print(std::cout, Teuchos::VERB_EXTREME);
-
-      RCP<Matrix> reorderedA = level.Get<RCP<Matrix> >("A", rAFact.get());
-      RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
-
-      TEST_EQUALITY(reorderedbA->Rows(), 2);
-      TEST_EQUALITY(reorderedbA->Cols(), 2);
-
-      TEST_EQUALITY(reorderedbA->getRangeMapExtractor()->getFullMap()->isSameAs(*(A->getRowMap())),false);
-      TEST_EQUALITY(reorderedbA->getDomainMapExtractor()->getFullMap()->isSameAs(*(A->getDomainMap())),false);
-      TEUCHOS_TEST_COMPARE(std::abs(A->getFrobeniusNorm()-reorderedA->getFrobeniusNorm()), <, 1e4 * Teuchos::ScalarTraits<Scalar>::eps(), out, success);
-      TEUCHOS_TEST_COMPARE(std::abs(A->getFrobeniusNorm()-reorderedbA->getFrobeniusNorm()), <, 1e4 * Teuchos::ScalarTraits<Scalar>::eps(), out, success);
-    } // end UseTpetra
-  }
-
-
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Jacobi_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,BGS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Reordered_BGS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII30II12II_BGS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII30II12II_BGS_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Thyra_BGS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Thyra_Nested_BGS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Thyra_Nested_BGS_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,SIMPLE_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII20I1I_SIMPLE_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_SIMPLE_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_SIMPLE_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII20I1I_Thyra_SIMPLE_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Thyra_SIMPLE_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII01I2I_Thyra_SIMPLE_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI0I21II_Thyra_SIMPLE_Setup_Apply3,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII20I1I_BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I10II_BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I10II_BS_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII02I1I_Thyra_BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Thyra_BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I10II_Thyra_BS_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII01I2I_Thyra_BS_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI0I21II_Thyra_BS_Setup_Apply3,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Uzawa_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Uzawa_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Uzawa_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Thyra_Uzawa_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII01I2I_Thyra_Uzawa_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI0I21II_Thyra_Uzawa_Setup_Apply3,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,Indef_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Indef_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Indef_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI2I01II_Thyra_Indef_Setup_Apply,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedII01I2I_Thyra_Indef_Setup_Apply2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,NestedI0I21II_Thyra_Indef_Setup_Apply3,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother,SplitReorder,SC,LO,GO,NO)
+    }
+    GlobalOrdinal gcount1 = 0;
+    GlobalOrdinal gcount2 = 0;
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count1, &gcount1);
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &count2, &gcount2);
+    Teuchos::RCP<const Map> map1 = MapFactory::Build(lib, gcount1, myGids1(), 0, comm);
+    Teuchos::RCP<const Map> map2 = MapFactory::Build(lib, gcount2, myGids2(), 0, comm);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Level level;
+    TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    level.Set("Map1", map1);
+    level.Set("Map2", map2);
+
+    // Test ReorderBlockAFactory
+    Teuchos::RCP<ReorderBlockAFactory> rAFact = Teuchos::rcp(new ReorderBlockAFactory());
+    rAFact->SetFactory("A", MueLu::NoFactory::getRCP());
+    rAFact->SetFactory("Map1", MueLu::NoFactory::getRCP());
+    rAFact->SetFactory("Map2", MueLu::NoFactory::getRCP());
+    rAFact->SetParameter(std::string("Reorder Type"), Teuchos::ParameterEntry(std::string("[ 0 1 ]")));
+
+    // request BGS smoother (and all dependencies) on level
+    level.Request("A", rAFact.get());
+
+    // smootherFact->DeclareInput(level);
+    rAFact->Build(level);
+
+    level.print(std::cout, Teuchos::VERB_EXTREME);
+
+    RCP<Matrix> reorderedA            = level.Get<RCP<Matrix> >("A", rAFact.get());
+    RCP<BlockedCrsMatrix> reorderedbA = Teuchos::rcp_dynamic_cast<BlockedCrsMatrix>(reorderedA);
+
+    TEST_EQUALITY(reorderedbA->Rows(), 2);
+    TEST_EQUALITY(reorderedbA->Cols(), 2);
+
+    TEST_EQUALITY(reorderedbA->getRangeMapExtractor()->getFullMap()->isSameAs(*(A->getRowMap())), false);
+    TEST_EQUALITY(reorderedbA->getDomainMapExtractor()->getFullMap()->isSameAs(*(A->getDomainMap())), false);
+    TEUCHOS_TEST_COMPARE(std::abs(A->getFrobeniusNorm() - reorderedA->getFrobeniusNorm()), <, 1e4 * Teuchos::ScalarTraits<Scalar>::eps(), out, success);
+    TEUCHOS_TEST_COMPARE(std::abs(A->getFrobeniusNorm() - reorderedbA->getFrobeniusNorm()), <, 1e4 * Teuchos::ScalarTraits<Scalar>::eps(), out, success);
+  }  // end UseTpetra
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Jacobi_Setup_Apply, SC, LO, GO, NO)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, BGS_Setup_Apply, SC, LO, GO, NO)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Reordered_BGS_Setup_Apply, SC, LO, GO, NO)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII30II12II_BGS_Setup_Apply2, SC, LO, GO, NO)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Thyra_BGS_Setup_Apply, SC, LO, GO, NO)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply, SC, LO, GO, NO)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Thyra_Nested_BGS_Setup_Apply2, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, SIMPLE_Setup_Apply, SC, LO, GO, NO)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII20I1I_SIMPLE_Setup_Apply, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_SIMPLE_Setup_Apply2, SC, LO, GO, NO)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII20I1I_Thyra_SIMPLE_Setup_Apply, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Thyra_SIMPLE_Setup_Apply, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII01I2I_Thyra_SIMPLE_Setup_Apply2, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI0I21II_Thyra_SIMPLE_Setup_Apply3, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, BS_Setup_Apply, SC, LO, GO, NO)                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII20I1I_BS_Setup_Apply, SC, LO, GO, NO)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I10II_BS_Setup_Apply, SC, LO, GO, NO)            \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I10II_BS_Setup_Apply2, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII02I1I_Thyra_BS_Setup_Apply, SC, LO, GO, NO)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Thyra_BS_Setup_Apply, SC, LO, GO, NO)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I10II_Thyra_BS_Setup_Apply, SC, LO, GO, NO)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII01I2I_Thyra_BS_Setup_Apply2, SC, LO, GO, NO)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI0I21II_Thyra_BS_Setup_Apply3, SC, LO, GO, NO)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Uzawa_Setup_Apply, SC, LO, GO, NO)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply, SC, LO, GO, NO)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Uzawa_Setup_Apply2, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Thyra_Uzawa_Setup_Apply, SC, LO, GO, NO)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII01I2I_Thyra_Uzawa_Setup_Apply2, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI0I21II_Thyra_Uzawa_Setup_Apply3, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, Indef_Setup_Apply, SC, LO, GO, NO)                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply, SC, LO, GO, NO)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Indef_Setup_Apply2, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI2I01II_Thyra_Indef_Setup_Apply, SC, LO, GO, NO)   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedII01I2I_Thyra_Indef_Setup_Apply2, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, NestedI0I21II_Thyra_Indef_Setup_Apply3, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(BlockedSmoother, SplitReorder, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/Ifpack2Smoother.cpp b/packages/muelu/test/unit_tests/Smoothers/Ifpack2Smoother.cpp
index 5ac9cf88a537..3961d2c12860 100644
--- a/packages/muelu/test/unit_tests/Smoothers/Ifpack2Smoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/Ifpack2Smoother.cpp
@@ -60,580 +60,561 @@
 
 namespace MueLuTests {
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      Ifpack2Smoother smoother("RELAXATION", Teuchos::ParameterList());
-      testApplyNoSetup(smoother, out, success);
-
-    }
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Ifpack2Smoother smoother("RELAXATION", Teuchos::ParameterList());
+    testApplyNoSetup(smoother, out, success);
   }
-
-  // Tests interface to Ifpack2's Jacobi preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_Jacobi, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      if (Teuchos::ScalarTraits<Scalar>::isComplex) {
-        out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
-        return;
-      }
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: type", "Jacobi");
-      paramList.set("relaxation: sweeps", (int) 1);
-      paramList.set("relaxation: damping factor", (double) 0.8);
-      paramList.set("relaxation: zero starting solution", false);
-
-      Ifpack2Smoother smoother("RELAXATION", paramList);
-
-      typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 6.3245553203367577133e-01;
-      switch (comm->getSize()) {
-        case 1:
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      auto ifpack2prec = smoother.getPreconditioner();
-
-      // reuse setup & solve
-      residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      switch (comm->getSize()) {
-        case 1:
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      // make sure the Ifpack2 preconditioner did not get replaced
-      TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
-
+}
+
+// Tests interface to Ifpack2's Jacobi preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_Jacobi, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    if (Teuchos::ScalarTraits<Scalar>::isComplex) {
+      out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
+      return;
     }
-  } // Jacobi
-
-  // Tests interface to Ifpack2's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_GaussSeidel, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      if (Teuchos::ScalarTraits<Scalar>::isComplex) {
-        out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
-        return;
-      }
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: type", "Gauss-Seidel");
-      paramList.set("relaxation: sweeps", (int) 1);
-      paramList.set("relaxation: damping factor", (double) 1.0);
-      paramList.set("relaxation: zero starting solution", false);
-
-      Ifpack2Smoother smoother("RELAXATION", paramList);
-
-      typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 5.773502691896257e-01;
-      switch (comm->getSize()) {
-        case 1:
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      auto ifpack2prec = smoother.getPreconditioner();
-
-      // reuse setup & solve
-      residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      switch (comm->getSize()) {
-        case 1:
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      // make sure the Ifpack2 preconditioner did not get replaced
-      TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
-
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: type", "Jacobi");
+    paramList.set("relaxation: sweeps", (int)1);
+    paramList.set("relaxation: damping factor", (double)0.8);
+    paramList.set("relaxation: zero starting solution", false);
+
+    Ifpack2Smoother smoother("RELAXATION", paramList);
+
+    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    RCP<const Teuchos::Comm<int> > comm                                  = TestHelpers::Parameters::getDefaultComm();
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 6.3245553203367577133e-01;
+    switch (comm->getSize()) {
+      case 1:
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    auto ifpack2prec = smoother.getPreconditioner();
+
+    // reuse setup & solve
+    residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    switch (comm->getSize()) {
+      case 1:
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    // make sure the Ifpack2 preconditioner did not get replaced
+    TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+  }
+}  // Jacobi
+
+// Tests interface to Ifpack2's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_GaussSeidel, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    if (Teuchos::ScalarTraits<Scalar>::isComplex) {
+      out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
+      return;
     }
-  } // GaussSeidel
-
-  // Tests interface to Ifpack2's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_GaussSeidel2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      if (Teuchos::ScalarTraits<Scalar>::isComplex) {
-        out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
-        return;
-      }
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: type", "Gauss-Seidel");
-      paramList.set("relaxation: sweeps", (int) 10);
-      paramList.set("relaxation: damping factor", (double) 1.0);
-      paramList.set("relaxation: zero starting solution", false);
-
-      Ifpack2Smoother smoother("RELAXATION",paramList);
-
-      typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm1 = 8.326553652741774e-02;
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm4 = 8.326553653078517e-02;
-      switch (comm->getSize()) {
-        case 1:
-          TEST_FLOATING_EQUALITY(residualNorms, expectedNorm1, 1e-12);
-          break;
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms, expectedNorm4, 1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      auto ifpack2prec = smoother.getPreconditioner();
-
-      // reuse setup & solve
-      residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      switch (comm->getSize()) {
-        case 1:
-          TEST_FLOATING_EQUALITY(residualNorms, expectedNorm1, 1e-12);
-          break;
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms, expectedNorm4, 1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-      // make sure the Ifpack2 preconditioner did not get replaced
-      TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
-
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: type", "Gauss-Seidel");
+    paramList.set("relaxation: sweeps", (int)1);
+    paramList.set("relaxation: damping factor", (double)1.0);
+    paramList.set("relaxation: zero starting solution", false);
+
+    Ifpack2Smoother smoother("RELAXATION", paramList);
+
+    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    RCP<const Teuchos::Comm<int> > comm                                  = TestHelpers::Parameters::getDefaultComm();
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 5.773502691896257e-01;
+    switch (comm->getSize()) {
+      case 1:
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    auto ifpack2prec = smoother.getPreconditioner();
+
+    // reuse setup & solve
+    residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    switch (comm->getSize()) {
+      case 1:
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    // make sure the Ifpack2 preconditioner did not get replaced
+    TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+  }
+}  // GaussSeidel
+
+// Tests interface to Ifpack2's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_GaussSeidel2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    if (Teuchos::ScalarTraits<Scalar>::isComplex) {
+      out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
+      return;
     }
-  } // GaussSeidel
-
-  // Tests interface to Ifpack2's Chebyshev preconditioner
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_Chebyshev, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-      using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-
-      if (Teuchos::ScalarTraits<Scalar>::isComplex) {
-        out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
-        return;
-      }
-
-      Teuchos::ParameterList paramList;
-      paramList.set("chebyshev: degree", (int) 3);
-      paramList.set("chebyshev: max eigenvalue", (double) 1.98476);
-      paramList.set("chebyshev: min eigenvalue", (double) 1.0);
-      paramList.set("chebyshev: ratio eigenvalue", (double) 20);
-      paramList.set("chebyshev: zero starting solution", false);
-      Ifpack2Smoother smoother("CHEBYSHEV",paramList);
-
-      typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 5.269156e-01;
-      TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, (1e-7 < TMT::eps() ? 10*TMT::eps() : 1e-7));  // Compare to residual reported by ML
-
-      auto ifpack2prec = smoother.getPreconditioner();
-
-      // reuse setup & solve
-      residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, (1e-7 < TMT::eps() ? 10*TMT::eps() : 1e-7));  // Compare to residual reported by ML
-
-      // make sure the Ifpack2 preconditioner did not get replaced
-      TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: type", "Gauss-Seidel");
+    paramList.set("relaxation: sweeps", (int)10);
+    paramList.set("relaxation: damping factor", (double)1.0);
+    paramList.set("relaxation: zero starting solution", false);
+
+    Ifpack2Smoother smoother("RELAXATION", paramList);
+
+    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    RCP<const Teuchos::Comm<int> > comm                                   = TestHelpers::Parameters::getDefaultComm();
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm1 = 8.326553652741774e-02;
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm4 = 8.326553653078517e-02;
+    switch (comm->getSize()) {
+      case 1:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm1, 1e-12);
+        break;
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm4, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    auto ifpack2prec = smoother.getPreconditioner();
+
+    // reuse setup & solve
+    residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    switch (comm->getSize()) {
+      case 1:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm1, 1e-12);
+        break;
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, expectedNorm4, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
+
+    // make sure the Ifpack2 preconditioner did not get replaced
+    TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+  }
+}  // GaussSeidel
+
+// Tests interface to Ifpack2's Chebyshev preconditioner
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_Chebyshev, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+    if (Teuchos::ScalarTraits<Scalar>::isComplex) {
+      out << "Skipping Tpetra for SC type \"complex\"" << std::endl;
+      return;
     }
-  } // Chebyshev
-
-  // Tests interface to Ifpack2's ILU(0) preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_ILU, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      //FIXME this will probably fail in parallel b/c it becomes block Jacobi
-      using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
 
-      Teuchos::ParameterList paramList;
-      Ifpack2Smoother smoother("ILUT",paramList);
+    Teuchos::ParameterList paramList;
+    paramList.set("chebyshev: degree", (int)3);
+    paramList.set("chebyshev: max eigenvalue", (double)1.98476);
+    paramList.set("chebyshev: min eigenvalue", (double)1.0);
+    paramList.set("chebyshev: ratio eigenvalue", (double)20);
+    paramList.set("chebyshev: zero starting solution", false);
+    Ifpack2Smoother smoother("CHEBYSHEV", paramList);
 
-      magnitude_type residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
+    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms      = testApply_A125_X1_RHS0(smoother, out, success);
+    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 5.269156e-01;
+    TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, (1e-7 < TMT::eps() ? 10 * TMT::eps() : 1e-7));  // Compare to residual reported by ML
 
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorms < 100*Teuchos::ScalarTraits<magnitude_type>::eps(), true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
+    auto ifpack2prec = smoother.getPreconditioner();
 
-      auto ifpack2prec = smoother.getPreconditioner();
-
-      // reuse setup & solve
-      residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorms < 100*Teuchos::ScalarTraits<magnitude_type>::eps(), true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-
-      // make sure the Ifpack2 preconditioner did not get replaced
-      TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+    // reuse setup & solve
+    residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, (1e-7 < TMT::eps() ? 10 * TMT::eps() : 1e-7));  // Compare to residual reported by ML
 
+    // make sure the Ifpack2 preconditioner did not get replaced
+    TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+  }
+}  // Chebyshev
+
+// Tests interface to Ifpack2's ILU(0) preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, HardCodedResult_ILU, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    // FIXME this will probably fail in parallel b/c it becomes block Jacobi
+    using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+
+    Teuchos::ParameterList paramList;
+    Ifpack2Smoother smoother("ILUT", paramList);
+
+    magnitude_type residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
+
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorms < 100 * Teuchos::ScalarTraits<magnitude_type>::eps(), true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // ILU
-
-  // Tests two sweeps of ILUT in Ifpack2
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, ILU_TwoSweeps, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
 
-      //FIXME this will probably fail in parallel b/c it becomes block Jacobi
-
-      Teuchos::ParameterList paramList;
-      Ifpack2Smoother smoother("ILUT",paramList);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, false); //nonzero initial guess
-
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      if (comm->getSize() == 1) {
-        //TEST_EQUALITY(residualNorms < 1e-10, true);
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
+    auto ifpack2prec = smoother.getPreconditioner();
 
+    // reuse setup & solve
+    residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorms < 100 * Teuchos::ScalarTraits<magnitude_type>::eps(), true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // ILU
-
-  // Make sure Ifpack2's Banded relaxation actually gets called
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BandedRelaxation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      Teuchos::ParameterList paramList;
-      paramList.set("partitioner: PDE equations",5);// Warning: This number has to be compatible with the problem size
-      Ifpack2Smoother smoother("BANDED RELAXATION",paramList);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, false); //nonzero initial guess
-
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      if (comm->getSize() == 1) {
-        //TEST_EQUALITY(residualNorms < 1e-10, true);
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
 
+    // make sure the Ifpack2 preconditioner did not get replaced
+    TEUCHOS_ASSERT_EQUALITY(ifpack2prec.ptr(), smoother.getPreconditioner().ptr());
+  }
+}  // ILU
+
+// Tests two sweeps of ILUT in Ifpack2
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, ILU_TwoSweeps, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    // FIXME this will probably fail in parallel b/c it becomes block Jacobi
+
+    Teuchos::ParameterList paramList;
+    Ifpack2Smoother smoother("ILUT", paramList);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, false);  // nonzero initial guess
+
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    if (comm->getSize() == 1) {
+      // TEST_EQUALITY(residualNorms < 1e-10, true);
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // banded
-
-  // Make sure Ifpack2's TriDi relaxation actually gets called
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, TriDiRelaxation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-
-      Teuchos::ParameterList paramList;
-      Ifpack2Smoother smoother("TRIDI RELAXATION",paramList);
-
-      //I don't use the testApply infrastructure because it has no provision for an initial guess.
-      Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
-
-      RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(),1);
-      RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(),1);
-
-      // Random X
-      X->setSeed(846930886);
-      X->randomize();
-
-      // Normalize X
-      Array<magnitude_type> norms(1); X->norm2(norms);
-      X->scale(1/norms[0]);
-
-      // Compute RHS corresponding to X
-      A->apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-      // Reset X to 0
-      X->putScalar((SC) 0.0);
-
-      RHS->norm2(norms);
-      out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-
-      out << "solve with zero initial guess" << std::endl;
-      Teuchos::Array<magnitude_type> initialNorms(1); X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, true);  //zero initial guess
-
-      Teuchos::Array<magnitude_type> finalNorms(1); X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      out << "solve with random initial guess" << std::endl;
-      X->randomize();
-      X->norm2(initialNorms);
-      out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
-
-      smoother.Apply(*X, *RHS, false); //nonzero initial guess
-
-      X->norm2(finalNorms);
-      Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
-      out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
-      out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      if (comm->getSize() == 1) {
-        //TEST_EQUALITY(residualNorms < 1e-10, true);
-        TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
-
+  }
+}  // ILU
+
+// Make sure Ifpack2's Banded relaxation actually gets called
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BandedRelaxation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList paramList;
+    paramList.set("partitioner: PDE equations", 5);  // Warning: This number has to be compatible with the problem size
+    Ifpack2Smoother smoother("BANDED RELAXATION", paramList);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, false);  // nonzero initial guess
+
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    if (comm->getSize() == 1) {
+      // TEST_EQUALITY(residualNorms < 1e-10, true);
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // tridi
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockRelaxation_Autosize, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      Teuchos::ParameterList matrixParams, ifpack2Params;
-
-      matrixParams.set("matrixType","Laplace1D");
-      matrixParams.set("nx",(GlobalOrdinal)20);// needs to be even
-      Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams,Xpetra::UseTpetra);
-      A->SetFixedBlockSize(2);
-
-      ifpack2Params.set("partitioner: type","linear");
-      Ifpack2Smoother smoother("BLOCK RELAXATION",ifpack2Params);
-      
-
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
-
-
-      TEST_EQUALITY(1,1);
+  }
+}  // banded
+
+// Make sure Ifpack2's TriDi relaxation actually gets called
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, TriDiRelaxation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList paramList;
+    Ifpack2Smoother smoother("TRIDI RELAXATION", paramList);
+
+    // I don't use the testApply infrastructure because it has no provision for an initial guess.
+    Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
+
+    RCP<MultiVector> X   = MultiVectorFactory::Build(A->getDomainMap(), 1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+    // Random X
+    X->setSeed(846930886);
+    X->randomize();
+
+    // Normalize X
+    Array<magnitude_type> norms(1);
+    X->norm2(norms);
+    X->scale(1 / norms[0]);
+
+    // Compute RHS corresponding to X
+    A->apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+    // Reset X to 0
+    X->putScalar((SC)0.0);
+
+    RHS->norm2(norms);
+    out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+    out << "solve with zero initial guess" << std::endl;
+    Teuchos::Array<magnitude_type> initialNorms(1);
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, true);  // zero initial guess
+
+    Teuchos::Array<magnitude_type> finalNorms(1);
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm1 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm1[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    out << "solve with random initial guess" << std::endl;
+    X->randomize();
+    X->norm2(initialNorms);
+    out << "  ||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+
+    smoother.Apply(*X, *RHS, false);  // nonzero initial guess
+
+    X->norm2(finalNorms);
+    Teuchos::Array<magnitude_type> residualNorm2 = Utilities::ResidualNorm(*A, *X, *RHS);
+    out << "  ||Residual_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorm2[0] << std::endl;
+    out << "  ||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << finalNorms[0] << std::endl;
+
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    if (comm->getSize() == 1) {
+      // TEST_EQUALITY(residualNorms < 1e-10, true);
+      TEST_EQUALITY(residualNorm1[0] != residualNorm2[0], true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // banded
-
+  }
+}  // tridi
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockRelaxation_Autosize, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList matrixParams, ifpack2Params;
+
+    matrixParams.set("matrixType", "Laplace1D");
+    matrixParams.set("nx", (GlobalOrdinal)20);  // needs to be even
+    Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildMatrix(matrixParams, Xpetra::UseTpetra);
+    A->SetFixedBlockSize(2);
+
+    ifpack2Params.set("partitioner: type", "linear");
+    Ifpack2Smoother smoother("BLOCK RELAXATION", ifpack2Params);
+
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
+
+    TEST_EQUALITY(1, 1);
+  }
+}  // banded
 
- TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockCrsMatrix_Relaxation_ViaPoint, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockCrsMatrix_Relaxation_ViaPoint, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      Teuchos::ParameterList matrixParams, ifpack2Params;
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList matrixParams, ifpack2Params;
 
-      matrixParams.set("matrixType","Laplace1D");
-      matrixParams.set("nx",(GlobalOrdinal)20);// needs to be even
+    matrixParams.set("matrixType", "Laplace1D");
+    matrixParams.set("nx", (GlobalOrdinal)20);  // needs to be even
 
-      RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams,Xpetra::UseTpetra);
-      ifpack2Params.set("smoother: use blockcrsmatrix storage",true);
-      
-      Ifpack2Smoother smoother("RELAXATION",ifpack2Params);
-      
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
+    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildBlockMatrixAsPoint(matrixParams, Xpetra::UseTpetra);
+    ifpack2Params.set("smoother: use blockcrsmatrix storage", true);
 
-      TEST_EQUALITY(1,1);
-    }
-  }
+    Ifpack2Smoother smoother("RELAXATION", ifpack2Params);
 
- TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockCrsMatrix_Relaxation_AsBlock, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
 
-    MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
-      Teuchos::ParameterList matrixParams, ifpack2Params;
+    TEST_EQUALITY(1, 1);
+  }
+}
 
-      matrixParams.set("matrixType","Laplace1D");
-      matrixParams.set("nx",(GlobalOrdinal)20);// needs to be even
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Ifpack2Smoother, BlockCrsMatrix_Relaxation_AsBlock, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-      RCP<Matrix> A = TestHelpers::TpetraTestFactory<SC, LO, GO, NO>::BuildBlockMatrix(matrixParams,Xpetra::UseTpetra);
-      ifpack2Params.set("smoother: use blockcrsmatrix storage",true);
-      
-      Ifpack2Smoother smoother("RELAXATION",ifpack2Params);
-      
-      Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-      level.Set("A", A);
-      smoother.Setup(level);
+  MUELU_TEST_ONLY_FOR(Xpetra::UseTpetra) {
+    Teuchos::ParameterList matrixParams, ifpack2Params;
 
-      TEST_EQUALITY(1,1);
-    }
-  }
+    matrixParams.set("matrixType", "Laplace1D");
+    matrixParams.set("nx", (GlobalOrdinal)20);  // needs to be even
 
+    RCP<Matrix> A = TestHelpers::TpetraTestFactory<SC, LO, GO, NO>::BuildBlockMatrix(matrixParams, Xpetra::UseTpetra);
+    ifpack2Params.set("smoother: use blockcrsmatrix storage", true);
 
+    Ifpack2Smoother smoother("RELAXATION", ifpack2Params);
 
+    Level level;
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+    level.Set("A", A);
+    smoother.Setup(level);
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,NotSetup,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,HardCodedResult_Jacobi,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,HardCodedResult_GaussSeidel,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,HardCodedResult_GaussSeidel2,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,HardCodedResult_Chebyshev,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,HardCodedResult_ILU,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,ILU_TwoSweeps,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,BandedRelaxation,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,TriDiRelaxation,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,BlockRelaxation_Autosize,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,BlockCrsMatrix_Relaxation_ViaPoint,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother,BlockCrsMatrix_Relaxation_AsBlock,SC,LO,GO,NO)
+    TEST_EQUALITY(1, 1);
+  }
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, NotSetup, SC, LO, GO, NO)                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, HardCodedResult_Jacobi, SC, LO, GO, NO)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, HardCodedResult_GaussSeidel, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, HardCodedResult_GaussSeidel2, SC, LO, GO, NO)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, HardCodedResult_Chebyshev, SC, LO, GO, NO)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, HardCodedResult_ILU, SC, LO, GO, NO)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, ILU_TwoSweeps, SC, LO, GO, NO)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, BandedRelaxation, SC, LO, GO, NO)                   \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, TriDiRelaxation, SC, LO, GO, NO)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, BlockRelaxation_Autosize, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, BlockCrsMatrix_Relaxation_ViaPoint, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Ifpack2Smoother, BlockCrsMatrix_Relaxation_AsBlock, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Smoothers/IfpackSmoother.cpp b/packages/muelu/test/unit_tests/Smoothers/IfpackSmoother.cpp
index ca1c4e82faa9..4258269df789 100644
--- a/packages/muelu/test/unit_tests/Smoothers/IfpackSmoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/IfpackSmoother.cpp
@@ -59,134 +59,119 @@
 
 namespace MueLuTests {
 
-  typedef double                Scalar;
-  typedef int                   LocalOrdinal;
-  typedef int                   GlobalOrdinal;
-  typedef Xpetra::EpetraNode    Node;
+typedef double Scalar;
+typedef int LocalOrdinal;
+typedef int GlobalOrdinal;
+typedef Xpetra::EpetraNode Node;
 #include "MueLu_UseShortNames.hpp"
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
 
-  TEUCHOS_UNIT_TEST(IfpackSmoother, NotSetup)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
-
-      IfpackSmoother smoother("point relaxation stand-alone", Teuchos::ParameterList());
-      testApplyNoSetup(smoother, out, success);
-
-    }
+TEUCHOS_UNIT_TEST(IfpackSmoother, NotSetup) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    IfpackSmoother smoother("point relaxation stand-alone", Teuchos::ParameterList());
+    testApplyNoSetup(smoother, out, success);
   }
-
-  // Tests interface to Ifpack's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_GaussSeidel1)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: type", "Gauss-Seidel");
-      paramList.set("relaxation: sweeps", (int) 1);
-      paramList.set("relaxation: damping factor", (double) 1.0);
-      paramList.set("relaxation: zero starting solution", false);
-
-      IfpackSmoother smoother("point relaxation stand-alone", paramList);
-
-      Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      switch (comm->getSize()) {
-        case 1:
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms, 5.773502691896257e-01,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-      } // switch
-
-    }
+}
+
+// Tests interface to Ifpack's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_GaussSeidel1) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: type", "Gauss-Seidel");
+    paramList.set("relaxation: sweeps", (int)1);
+    paramList.set("relaxation: damping factor", (double)1.0);
+    paramList.set("relaxation: zero starting solution", false);
+
+    IfpackSmoother smoother("point relaxation stand-alone", paramList);
+
+    Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    switch (comm->getSize()) {
+      case 1:
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, 5.773502691896257e-01, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+    }  // switch
   }
+}
+
+// Tests interface to Ifpack's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_GaussSeidel2) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+    Teuchos::ParameterList paramList;
+    paramList.set("relaxation: type", "Gauss-Seidel");
+    paramList.set("relaxation: sweeps", (int)10);
+    paramList.set("relaxation: damping factor", (double)1.0);
+    paramList.set("relaxation: zero starting solution", false);
+
+    IfpackSmoother smoother("point relaxation stand-alone", paramList);
+
+    Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+
+    switch (comm->getSize()) {
+      case 1:
+        TEST_FLOATING_EQUALITY(residualNorms, 8.326553652741774e-02, 1e-12);
+        break;
+      case 4:
+        TEST_FLOATING_EQUALITY(residualNorms, 8.326553653078517e-02, 1e-12);
+        break;
+      default:
+        out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
+        break;
+
+    }  // switch
+  }
+}  // GaussSeidel
 
-  // Tests interface to Ifpack's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_GaussSeidel2)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
-
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-      Teuchos::ParameterList paramList;
-      paramList.set("relaxation: type", "Gauss-Seidel");
-      paramList.set("relaxation: sweeps", (int) 10);
-      paramList.set("relaxation: damping factor", (double) 1.0);
-      paramList.set("relaxation: zero starting solution", false);
-
-      IfpackSmoother smoother("point relaxation stand-alone", paramList);
-
-      Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
-
-      switch (comm->getSize()) {
-        case 1:
-          TEST_FLOATING_EQUALITY(residualNorms,8.326553652741774e-02,1e-12);
-          break;
-        case 4:
-          TEST_FLOATING_EQUALITY(residualNorms,8.326553653078517e-02,1e-12);
-          break;
-        default:
-          out << "Pass/Fail is checked only for 1 and 4 processes." << std::endl;
-          break;
-
-      } // switch
-
-    }
-  } // GaussSeidel
-
-  // Tests interface to Ifpack's Chebyshev preconditioner.
-  TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_Chebyshev)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
-
-      // TODO This test should really calculate the reduction analytically.
-
-      Teuchos::ParameterList paramList;
-      paramList.set("chebyshev: degree", (int) 3);
-      paramList.set("chebyshev: max eigenvalue", (double) 1.98476);
-      paramList.set("chebyshev: min eigenvalue", (double) 1.0);
-      paramList.set("chebyshev: ratio eigenvalue", (double) 20);
-      paramList.set("chebyshev: zero starting solution", false);
-
-      IfpackSmoother smoother("Chebyshev", paramList);
+// Tests interface to Ifpack's Chebyshev preconditioner.
+TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_Chebyshev) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    // TODO This test should really calculate the reduction analytically.
 
-      Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
+    Teuchos::ParameterList paramList;
+    paramList.set("chebyshev: degree", (int)3);
+    paramList.set("chebyshev: max eigenvalue", (double)1.98476);
+    paramList.set("chebyshev: min eigenvalue", (double)1.0);
+    paramList.set("chebyshev: ratio eigenvalue", (double)20);
+    paramList.set("chebyshev: zero starting solution", false);
 
-      TEST_FLOATING_EQUALITY(residualNorms, 5.269156e-01, 1e-7); // Compare to residual reported by ML
+    IfpackSmoother smoother("Chebyshev", paramList);
 
-    }
-  } // Chebyshev
+    Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X1_RHS0(smoother, out, success);
 
-  // Tests interface to Ifpack's ILU(0) preconditioner.
-  TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_ILU)
-  {
-    MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
-
-      // FIXME this will probably fail in parallel b/c it becomes block Jacobi
+    TEST_FLOATING_EQUALITY(residualNorms, 5.269156e-01, 1e-7);  // Compare to residual reported by ML
+  }
+}  // Chebyshev
 
-      IfpackSmoother smoother("ILU", Teuchos::ParameterList());
+// Tests interface to Ifpack's ILU(0) preconditioner.
+TEUCHOS_UNIT_TEST(IfpackSmoother, HardCodedResult_ILU) {
+  MUELU_TEST_ONLY_FOR(Xpetra::UseEpetra) {
+    // FIXME this will probably fail in parallel b/c it becomes block Jacobi
 
-      Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
+    IfpackSmoother smoother("ILU", Teuchos::ParameterList());
 
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      if (comm->getSize() == 1) {
-        TEST_EQUALITY(residualNorms < 1e-10, true);
-      } else {
-        out << "Pass/Fail is only checked in serial." << std::endl;
-      }
+    Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
 
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+    if (comm->getSize() == 1) {
+      TEST_EQUALITY(residualNorms < 1e-10, true);
+    } else {
+      out << "Pass/Fail is only checked in serial." << std::endl;
     }
-  } // ILU
+  }
+}  // ILU
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
-//TODO: test if 10 its of ifpack == 10 apply call
-//TODO:     TEST_EQUALITY(smoother->GetNIts(),10);
-//TODO: check failed if TPETRA
+// TODO: test if 10 its of ifpack == 10 apply call
+// TODO:     TEST_EQUALITY(smoother->GetNIts(),10);
+// TODO: check failed if TPETRA
diff --git a/packages/muelu/test/unit_tests/Smoothers/MueLu_TestHelpersSmoothers.hpp b/packages/muelu/test/unit_tests/Smoothers/MueLu_TestHelpersSmoothers.hpp
index c93a06a18f03..71ab72cc01c3 100644
--- a/packages/muelu/test/unit_tests/Smoothers/MueLu_TestHelpersSmoothers.hpp
+++ b/packages/muelu/test/unit_tests/Smoothers/MueLu_TestHelpersSmoothers.hpp
@@ -56,151 +56,155 @@
 // Helper functions to test if derived classes conforms to the SmootherBase and SmootherPrototype interfaces
 
 namespace MueLuTests {
-  namespace TestHelpers {
-    namespace Smoothers {
+namespace TestHelpers {
+namespace Smoothers {
 
-      // SmootherPrototype test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      void testApplyNoSetup(const MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother, Teuchos::FancyOStream & out, bool & success) {
+// SmootherPrototype test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void testApplyNoSetup(const MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother, Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        GO numGlobalElements = 125;
-        RCP<const Map> map = MapFactory::Build(Parameters::getLib(), numGlobalElements, 0, Parameters::getDefaultComm());
+  GO numGlobalElements = 125;
+  RCP<const Map> map   = MapFactory::Build(Parameters::getLib(), numGlobalElements, 0, Parameters::getDefaultComm());
 
-        RCP<MultiVector> X   = MultiVectorFactory::Build(map,1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
 
-        TEST_THROW(smoother.Apply(*X,*RHS), MueLu::Exceptions::RuntimeError);
-      }
+  TEST_THROW(smoother.Apply(*X, *RHS), MueLu::Exceptions::RuntimeError);
+}
 
-      // SmootherBase test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-      testApply(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                const MueLu::SmootherBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node> & X,
-                const Xpetra::MultiVector<Scalar,LocalOrdinal,GlobalOrdinal,Node>& RHS,
-                Teuchos::FancyOStream & out, bool & success) {
+// SmootherBase test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testApply(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+          const MueLu::SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+          Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& X,
+          const Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>& RHS,
+          Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
-        typedef Teuchos::ScalarTraits<SC> ST;
+  typedef Teuchos::ScalarTraits<SC> ST;
 
-        Array<typename ST::magnitudeType> norms(1);
+  Array<typename ST::magnitudeType> norms(1);
 
-        RHS.norm2(norms);
-        out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+  RHS.norm2(norms);
+  out << "||RHS|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
 
-        Teuchos::Array<typename ST::magnitudeType> initialNorms(1); X.norm2(initialNorms);
-        out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
+  Teuchos::Array<typename ST::magnitudeType> initialNorms(1);
+  X.norm2(initialNorms);
+  out << "||X_initial|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << initialNorms[0] << std::endl;
 
-        smoother.Apply(X, RHS); // TODO: bool const &InitialGuessIsZero=false
+  smoother.Apply(X, RHS);  // TODO: bool const &InitialGuessIsZero=false
 
-        Teuchos::Array<typename ST::magnitudeType> finalNorms(1); X.norm2(finalNorms);
-        out << "||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(25) << norms[0] << std::endl;
+  Teuchos::Array<typename ST::magnitudeType> finalNorms(1);
+  X.norm2(finalNorms);
+  out << "||X_final|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(25) << norms[0] << std::endl;
 
-        Teuchos::Array<typename ST::magnitudeType> residualNorms = Utilities::ResidualNorm(A, X, RHS);
-        out << "||Residual|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorms[0] << std::endl;
+  Teuchos::Array<typename ST::magnitudeType> residualNorms = Utilities::ResidualNorm(A, X, RHS);
+  out << "||Residual|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(20) << residualNorms[0] << std::endl;
 
-        return residualNorms[0];
-      }
+  return residualNorms[0];
+}
 
-      // SmootherBase test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-      testApply_X1_RHS0(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                        const MueLu::SmootherBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                        Teuchos::FancyOStream & out, bool & success) {
+// SmootherBase test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testApply_X1_RHS0(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                  const MueLu::SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                  Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        RCP<MultiVector> X   = MultiVectorFactory::Build(A.getDomainMap(),1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(A.getRangeMap(),1);
-        X->putScalar((SC) 1.0);
-        RHS->putScalar((SC) 0.0);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A.getDomainMap(), 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A.getRangeMap(), 1);
+  X->putScalar((SC)1.0);
+  RHS->putScalar((SC)0.0);
 
-        return testApply(A, smoother, *X, *RHS, out, success);
-      }
+  return testApply(A, smoother, *X, *RHS, out, success);
+}
 
-      // SmootherBase test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-      testApply_X0_RandomRHS(const Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>& A,
-                             const MueLu::SmootherBase<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                             Teuchos::FancyOStream & out, bool & success) {
+// SmootherBase test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testApply_X0_RandomRHS(const Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& A,
+                       const MueLu::SmootherBase<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                       Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        RCP<MultiVector> X   = MultiVectorFactory::Build(A.getDomainMap(),1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(A.getRangeMap(),1);
+  RCP<MultiVector> X   = MultiVectorFactory::Build(A.getDomainMap(), 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(A.getRangeMap(), 1);
 
-        // Random X
-        X->setSeed(846930886);
-        X->randomize();
+  // Random X
+  X->setSeed(846930886);
+  X->randomize();
 
-        // Normalize X
-        typedef Teuchos::ScalarTraits<SC> ST;
-        Array<typename ST::magnitudeType> norms(1); X->norm2(norms);
-        X->scale(1/norms[0]);
+  // Normalize X
+  typedef Teuchos::ScalarTraits<SC> ST;
+  Array<typename ST::magnitudeType> norms(1);
+  X->norm2(norms);
+  X->scale(1 / norms[0]);
 
-        // Compute RHS corresponding to X
-        A.apply(*X,*RHS, Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
+  // Compute RHS corresponding to X
+  A.apply(*X, *RHS, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
 
-        // Reset X to 0
-        X->putScalar((SC) 0.0);
+  // Reset X to 0
+  X->putScalar((SC)0.0);
 
-        return testApply(A, smoother, *X, *RHS, out, success);
-      }
+  return testApply(A, smoother, *X, *RHS, out, success);
+}
 
-      // SmootherPrototype helper function
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      void setupSmoother(Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>>& A,
-                         MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                         Teuchos::FancyOStream & out, bool & success) {
+// SmootherPrototype helper function
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void setupSmoother(Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                   MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                   Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        Level level; TestHelpers::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-        level.Set("A", A);
-        smoother.Setup(level);
-      }
-
-      // SmootherPrototype test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-      testApply_A125_X1_RHS0(MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                             Teuchos::FancyOStream & out, bool & success) {
+  Level level;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+  smoother.Setup(level);
+}
+
+// SmootherPrototype test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testApply_A125_X1_RHS0(MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                       Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
+  Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
 
-        setupSmoother(A, smoother, out, success);
-        return testApply_X1_RHS0(*A, smoother, out, success); // in MueLuTests::SmootherBase
-      }
+  setupSmoother(A, smoother, out, success);
+  return testApply_X1_RHS0(*A, smoother, out, success);  // in MueLuTests::SmootherBase
+}
 
-      // SmootherPrototype test
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      typename Teuchos::ScalarTraits<Scalar>::magnitudeType
-      testApply_A125_X0_RandomRHS(MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                                  Teuchos::FancyOStream & out, bool & success) {
+// SmootherPrototype test
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+typename Teuchos::ScalarTraits<Scalar>::magnitudeType
+testApply_A125_X0_RandomRHS(MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                            Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
+  Teuchos::RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(125);
 
-        setupSmoother(A, smoother, out, success);
-        return testApply_X0_RandomRHS(*A, smoother, out, success); // in MueLuTests::SmootherBase
-      }
+  setupSmoother(A, smoother, out, success);
+  return testApply_X0_RandomRHS(*A, smoother, out, success);  // in MueLuTests::SmootherBase
+}
 
-      template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-      void testDirectSolver(MueLu::SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node>& smoother,
-                            Teuchos::FancyOStream & out, bool & success) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void testDirectSolver(MueLu::SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node>& smoother,
+                      Teuchos::FancyOStream& out, bool& success) {
 #include "MueLu_UseShortNames.hpp"
 
-        using ST = Teuchos::ScalarTraits<SC>;
-        using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-        using MT = Teuchos::ScalarTraits<magnitude_type>;
+  using ST             = Teuchos::ScalarTraits<SC>;
+  using magnitude_type = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using MT             = Teuchos::ScalarTraits<magnitude_type>;
 
-        magnitude_type residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
-        TEST_EQUALITY(residualNorms < 100*MT::eps(), true);
-      }
+  magnitude_type residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
+  TEST_EQUALITY(residualNorms < 100 * MT::eps(), true);
+}
 
-    } // namespace Smoothers
-  } // namespace TestHelpers
-} // MueLuTests
+}  // namespace Smoothers
+}  // namespace TestHelpers
+}  // namespace MueLuTests
 
-#endif // MUELU_TEST_HELPERS_SMOOTHERS_H
+#endif  // MUELU_TEST_HELPERS_SMOOTHERS_H
diff --git a/packages/muelu/test/unit_tests/Smoothers/SmootherFactory.cpp b/packages/muelu/test/unit_tests/Smoothers/SmootherFactory.cpp
index 781043eca47d..e8d6e6c60daa 100644
--- a/packages/muelu/test/unit_tests/Smoothers/SmootherFactory.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/SmootherFactory.cpp
@@ -54,402 +54,400 @@
 
 namespace MueLuTests {
 
-  namespace SmootherFactory {
-
-
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    using namespace MueLu::Exceptions;
-
-    template
-    <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class SmootherFactoryHelpers {
-
-#   include <MueLu_UseShortNames.hpp>
-
-      public:
-
-      ////////////////////////////////////////////////////////////////////////////////
-      // Helper functions: class invariant
-      ////////////////////////////////////////////////////////////////////////////////
-
-      // Check the consistency of the internal state of a SmootherFactory object
-      static void check(RCP<const SmootherFactory> smootherFactory, RCP<const SmootherPrototype> preSmooProto, RCP<const SmootherPrototype> postSmooProto, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> getPreSmooProto, getPostSmooProto;
-        smootherFactory->GetSmootherPrototypes(getPreSmooProto, getPostSmooProto);
-
-        TEST_EQUALITY(preSmooProto,  getPreSmooProto);
-        TEST_EQUALITY(postSmooProto, getPostSmooProto);
-
-        // Check if internal prototypes are still prototypes (IsSetup() == false)
-        if (preSmooProto  != Teuchos::null) TEST_EQUALITY(preSmooProto->IsSetup(),  false);
-        if (postSmooProto != Teuchos::null) TEST_EQUALITY(postSmooProto->IsSetup(), false);
-      }
-
-      // Similar to check(), but with INEQUALITY.
-      // Do not check if internal prototypes verifies IsSetup() == false
-      static void ineqcheck(RCP<const SmootherFactory> smootherFactory, RCP<const SmootherBase> preSmooProto, RCP<const SmootherBase> postSmooProto, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> getPreSmooProto, getPostSmooProto;
-        smootherFactory->GetSmootherPrototypes(getPreSmooProto, getPostSmooProto);
-        TEST_INEQUALITY(preSmooProto,  getPreSmooProto);
-        TEST_INEQUALITY(postSmooProto, getPostSmooProto);
-      }
-
-      ////////////////////////////////////////////////////////////////////////////////
-      // Helper functions: apply a test to a collection of test cases
-      ////////////////////////////////////////////////////////////////////////////////
-
-      // Apply a test to a list of cases (one argument)
-      static void testCollection(void (*func)(RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> smooProto = rcp( new FakeSmootherPrototype() );
-
-        // tests
-        func(smooProto, out, success);
-        func(Teuchos::null, out, success);
-      }
-
-      // Apply a test to a list of invalid cases (one argument)
-      static void testInvalidCollection(void (*func)(RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> invalidSmooProto = rcp( new FakeSmootherPrototype() ); Level level; invalidSmooProto->Setup(level);
-
-        // tests
-        func(invalidSmooProto, out, success);
-      }
-
-      // Apply a test to a list of cases (two argument)
-      static void testCollection(
-          void (*func)(RCP<SmootherPrototype>, RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &),
-          Teuchos::FancyOStream & out,
-          bool & success)
-      {
-        RCP<SmootherPrototype> smooProtoA = rcp( new FakeSmootherPrototype() );
-        RCP<SmootherPrototype> smooProtoB = rcp( new FakeSmootherPrototype() );
-
-        // tests
-        func(Teuchos::null, Teuchos::null, out, success);
-        func(smooProtoA, smooProtoB, out, success);
-        func(smooProtoA, Teuchos::null, out, success);
-        func(Teuchos::null, smooProtoB, out, success);
-        func(smooProtoA, smooProtoA, out, success);
-      }
-
-      // Apply a test to a list of invalid cases (two argument)
-      static void testInvalidCollection(void (*func)(RCP<SmootherPrototype>, RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> smooProtoA = rcp( new FakeSmootherPrototype() );
-        RCP<SmootherPrototype> smooProtoB = rcp( new FakeSmootherPrototype() );
-        RCP<SmootherPrototype> invalidSmooProto = rcp( new FakeSmootherPrototype() ); Level level; invalidSmooProto->Setup(level);
-
-        // tests
-        func(invalidSmooProto, smooProtoB, out, success);
-        func(smooProtoA, invalidSmooProto, out, success);
-        func(invalidSmooProto, Teuchos::null, out, success);
-        func(Teuchos::null, invalidSmooProto, out, success);
-        func(invalidSmooProto, invalidSmooProto, out, success);
-      }
-
-      // Build an object and test internal state (one argument)
-      static void testConstructor1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream & out, bool & success) {
-        check(rcp( new SmootherFactory(smooProto) ), smooProto, smooProto, out, success); // One argument == same pre and post smoother prototype
-      }
-
-      // Test with invalid input argument
-      static void testInvalidConstructor1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream & out, bool & success) {
-        TEST_THROW(rcp( new SmootherFactory(smooProto) ), RuntimeError);
-      }
-
-      // Build an object and test internal state (two arguments)
-      static void testConstructor2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        check(rcp( new SmootherFactory(smooProtoA, smooProtoB) ), smooProtoA, smooProtoB, out, success);
-      }
-
-      // Test with invalid input arguments
-      static void testInvalidConstructor2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        TEST_THROW(rcp( new SmootherFactory(smooProtoA, smooProtoB) ), RuntimeError);
-      }
-
-      static void testSetSmootherPrototypes1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> smooProto0 = rcp( new FakeSmootherPrototype() );
-        RCP<SmootherFactory>   smooFact   = rcp( new SmootherFactory(smooProto0) );
-
-        ineqcheck(smooFact, smooProto, smooProto, out, success);
-        smooFact->SetSmootherPrototypes(smooProto);
-        check(smooFact, smooProto, smooProto, out, success); // One argument == same pre and post smoother prototype
-      }
-
-      static void testInvalidSetSmootherPrototypes1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream & out, bool & success) {
-        SmootherFactory smooFact(Teuchos::null);
-        TEST_THROW(smooFact.SetSmootherPrototypes(smooProto), RuntimeError);
-      }
-
-      static void testSetSmootherPrototypes2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherPrototype> smooProto0 = rcp( new FakeSmootherPrototype() );
-        RCP<SmootherFactory>   smooFact   = rcp( new SmootherFactory(smooProto0) );
-
-        ineqcheck(smooFact, smooProtoA, smooProtoB, out, success);
-        smooFact->SetSmootherPrototypes(smooProtoA, smooProtoB);
-        check(smooFact, smooProtoA, smooProtoB, out, success);
-      }
-
-      static void testInvalidSetSmootherPrototypes2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        SmootherFactory smooFact(Teuchos::null);
-        TEST_THROW(smooFact.SetSmootherPrototypes(smooProtoA, smooProtoB), RuntimeError);
-      }
-
-      // test if a smoother created by Build() is correct (check if it corresponds to the prototype)
-      static void testBuildCheckOutput(RCP<SmootherFactory> smooFact, Level& level, RCP<SmootherPrototype> smooProto, const std::string& tag, Teuchos::FancyOStream & out, bool & success) {
-        if (smooProto == Teuchos::null) {
-          TEST_EQUALITY(level.IsAvailable(tag, smooFact.get()), false);
-        } else {
-          RCP<SmootherBase> smoother;
-          TEST_NOTHROW(smoother = level.Get< RCP<SmootherBase> >(tag, smooFact.get()));
-
-          TEST_INEQUALITY(smoother, Teuchos::null);
-          TEST_INEQUALITY(smoother, smooProto);
-
-          if (smooProto != Teuchos::null) {
-            RCP<FakeSmootherPrototype> smootherF;
-
-            // ouput test: smoothers same derived class as prototypes
-            TEST_NOTHROW(smootherF = rcp_dynamic_cast<FakeSmootherPrototype>(smoother,true));
-
-            if (smootherF != Teuchos::null) {
-              // output test: smoother parameters == prototype parameters
-              RCP<FakeSmootherPrototype> smooProtoF = rcp_dynamic_cast<FakeSmootherPrototype>(smooProto,true);
-              TEST_EQUALITY(smootherF->GetParam(), smooProtoF->GetParam());
-
-              // output test: smoothers are ready to be apply
-              TEST_EQUALITY(smootherF->IsSetup(), true);
-
-              // setup done only once.
-              TEST_EQUALITY(smootherF->GetNumOfSetupCall(), 1);
-              TEST_EQUALITY(smootherF->GetNumOfSetup(), 1);
-            }
-          }
+namespace SmootherFactory {
+
+using Teuchos::RCP;
+using Teuchos::rcp;
+using namespace MueLu::Exceptions;
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class SmootherFactoryHelpers {
+#include <MueLu_UseShortNames.hpp>
+
+ public:
+  ////////////////////////////////////////////////////////////////////////////////
+  // Helper functions: class invariant
+  ////////////////////////////////////////////////////////////////////////////////
+
+  // Check the consistency of the internal state of a SmootherFactory object
+  static void check(RCP<const SmootherFactory> smootherFactory, RCP<const SmootherPrototype> preSmooProto, RCP<const SmootherPrototype> postSmooProto, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> getPreSmooProto, getPostSmooProto;
+    smootherFactory->GetSmootherPrototypes(getPreSmooProto, getPostSmooProto);
+
+    TEST_EQUALITY(preSmooProto, getPreSmooProto);
+    TEST_EQUALITY(postSmooProto, getPostSmooProto);
+
+    // Check if internal prototypes are still prototypes (IsSetup() == false)
+    if (preSmooProto != Teuchos::null) TEST_EQUALITY(preSmooProto->IsSetup(), false);
+    if (postSmooProto != Teuchos::null) TEST_EQUALITY(postSmooProto->IsSetup(), false);
+  }
+
+  // Similar to check(), but with INEQUALITY.
+  // Do not check if internal prototypes verifies IsSetup() == false
+  static void ineqcheck(RCP<const SmootherFactory> smootherFactory, RCP<const SmootherBase> preSmooProto, RCP<const SmootherBase> postSmooProto, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> getPreSmooProto, getPostSmooProto;
+    smootherFactory->GetSmootherPrototypes(getPreSmooProto, getPostSmooProto);
+    TEST_INEQUALITY(preSmooProto, getPreSmooProto);
+    TEST_INEQUALITY(postSmooProto, getPostSmooProto);
+  }
+
+  ////////////////////////////////////////////////////////////////////////////////
+  // Helper functions: apply a test to a collection of test cases
+  ////////////////////////////////////////////////////////////////////////////////
+
+  // Apply a test to a list of cases (one argument)
+  static void testCollection(void (*func)(RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> smooProto = rcp(new FakeSmootherPrototype());
+
+    // tests
+    func(smooProto, out, success);
+    func(Teuchos::null, out, success);
+  }
+
+  // Apply a test to a list of invalid cases (one argument)
+  static void testInvalidCollection(void (*func)(RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> invalidSmooProto = rcp(new FakeSmootherPrototype());
+    Level level;
+    invalidSmooProto->Setup(level);
+
+    // tests
+    func(invalidSmooProto, out, success);
+  }
+
+  // Apply a test to a list of cases (two argument)
+  static void testCollection(
+      void (*func)(RCP<SmootherPrototype>, RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &),
+      Teuchos::FancyOStream &out,
+      bool &success) {
+    RCP<SmootherPrototype> smooProtoA = rcp(new FakeSmootherPrototype());
+    RCP<SmootherPrototype> smooProtoB = rcp(new FakeSmootherPrototype());
+
+    // tests
+    func(Teuchos::null, Teuchos::null, out, success);
+    func(smooProtoA, smooProtoB, out, success);
+    func(smooProtoA, Teuchos::null, out, success);
+    func(Teuchos::null, smooProtoB, out, success);
+    func(smooProtoA, smooProtoA, out, success);
+  }
+
+  // Apply a test to a list of invalid cases (two argument)
+  static void testInvalidCollection(void (*func)(RCP<SmootherPrototype>, RCP<SmootherPrototype>, Teuchos::FancyOStream &, bool &), Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> smooProtoA       = rcp(new FakeSmootherPrototype());
+    RCP<SmootherPrototype> smooProtoB       = rcp(new FakeSmootherPrototype());
+    RCP<SmootherPrototype> invalidSmooProto = rcp(new FakeSmootherPrototype());
+    Level level;
+    invalidSmooProto->Setup(level);
+
+    // tests
+    func(invalidSmooProto, smooProtoB, out, success);
+    func(smooProtoA, invalidSmooProto, out, success);
+    func(invalidSmooProto, Teuchos::null, out, success);
+    func(Teuchos::null, invalidSmooProto, out, success);
+    func(invalidSmooProto, invalidSmooProto, out, success);
+  }
+
+  // Build an object and test internal state (one argument)
+  static void testConstructor1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream &out, bool &success) {
+    check(rcp(new SmootherFactory(smooProto)), smooProto, smooProto, out, success);  // One argument == same pre and post smoother prototype
+  }
+
+  // Test with invalid input argument
+  static void testInvalidConstructor1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream &out, bool &success) {
+    TEST_THROW(rcp(new SmootherFactory(smooProto)), RuntimeError);
+  }
+
+  // Build an object and test internal state (two arguments)
+  static void testConstructor2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    check(rcp(new SmootherFactory(smooProtoA, smooProtoB)), smooProtoA, smooProtoB, out, success);
+  }
+
+  // Test with invalid input arguments
+  static void testInvalidConstructor2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    TEST_THROW(rcp(new SmootherFactory(smooProtoA, smooProtoB)), RuntimeError);
+  }
+
+  static void testSetSmootherPrototypes1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> smooProto0 = rcp(new FakeSmootherPrototype());
+    RCP<SmootherFactory> smooFact     = rcp(new SmootherFactory(smooProto0));
+
+    ineqcheck(smooFact, smooProto, smooProto, out, success);
+    smooFact->SetSmootherPrototypes(smooProto);
+    check(smooFact, smooProto, smooProto, out, success);  // One argument == same pre and post smoother prototype
+  }
+
+  static void testInvalidSetSmootherPrototypes1(RCP<SmootherPrototype> smooProto, Teuchos::FancyOStream &out, bool &success) {
+    SmootherFactory smooFact(Teuchos::null);
+    TEST_THROW(smooFact.SetSmootherPrototypes(smooProto), RuntimeError);
+  }
+
+  static void testSetSmootherPrototypes2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherPrototype> smooProto0 = rcp(new FakeSmootherPrototype());
+    RCP<SmootherFactory> smooFact     = rcp(new SmootherFactory(smooProto0));
+
+    ineqcheck(smooFact, smooProtoA, smooProtoB, out, success);
+    smooFact->SetSmootherPrototypes(smooProtoA, smooProtoB);
+    check(smooFact, smooProtoA, smooProtoB, out, success);
+  }
+
+  static void testInvalidSetSmootherPrototypes2(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    SmootherFactory smooFact(Teuchos::null);
+    TEST_THROW(smooFact.SetSmootherPrototypes(smooProtoA, smooProtoB), RuntimeError);
+  }
+
+  // test if a smoother created by Build() is correct (check if it corresponds to the prototype)
+  static void testBuildCheckOutput(RCP<SmootherFactory> smooFact, Level &level, RCP<SmootherPrototype> smooProto, const std::string &tag, Teuchos::FancyOStream &out, bool &success) {
+    if (smooProto == Teuchos::null) {
+      TEST_EQUALITY(level.IsAvailable(tag, smooFact.get()), false);
+    } else {
+      RCP<SmootherBase> smoother;
+      TEST_NOTHROW(smoother = level.Get<RCP<SmootherBase> >(tag, smooFact.get()));
+
+      TEST_INEQUALITY(smoother, Teuchos::null);
+      TEST_INEQUALITY(smoother, smooProto);
+
+      if (smooProto != Teuchos::null) {
+        RCP<FakeSmootherPrototype> smootherF;
+
+        // ouput test: smoothers same derived class as prototypes
+        TEST_NOTHROW(smootherF = rcp_dynamic_cast<FakeSmootherPrototype>(smoother, true));
+
+        if (smootherF != Teuchos::null) {
+          // output test: smoother parameters == prototype parameters
+          RCP<FakeSmootherPrototype> smooProtoF = rcp_dynamic_cast<FakeSmootherPrototype>(smooProto, true);
+          TEST_EQUALITY(smootherF->GetParam(), smooProtoF->GetParam());
+
+          // output test: smoothers are ready to be apply
+          TEST_EQUALITY(smootherF->IsSetup(), true);
+
+          // setup done only once.
+          TEST_EQUALITY(smootherF->GetNumOfSetupCall(), 1);
+          TEST_EQUALITY(smootherF->GetNumOfSetup(), 1);
         }
       }
-
-      // test outpout of Build() and BuildSmoother()
-      static void testBuildCheck(RCP<SmootherFactory> smooFact, Level& level, RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, MueLu::PreOrPost preOrPost, Teuchos::FancyOStream & out, bool & success) {
-        // invariant: smoother prototypes kept unchanged
-        check(smooFact, smooProtoA, smooProtoB, out, success);
-
-        // output test
-        if (preOrPost == MueLu::BOTH) {
-
-          testBuildCheckOutput(smooFact, level, smooProtoA, "PreSmoother", out, success);
-          testBuildCheckOutput(smooFact, level, smooProtoB, "PostSmoother", out, success);
-
-          // ReUse: if pre and post prototype are the same, then pre smoother == post smoother
-          // otherwise, they are different (have not been tested by previous tests)
-          RCP<SmootherBase> smooA, smooB;
-          if (smooProtoA != Teuchos::null) { smooA = level.Get< RCP<SmootherBase> >("PreSmoother",  smooFact.get()); }
-          if (smooProtoB != Teuchos::null) { smooB = level.Get< RCP<SmootherBase> >("PostSmoother", smooFact.get()); }
-          if (smooProtoA == smooProtoB) { TEST_EQUALITY(smooA, smooB); } else { TEST_INEQUALITY(smooA, smooB); }
-
-        } else if (preOrPost == MueLu::PRE) {
-
-          testBuildCheckOutput(smooFact, level, smooProtoA, "PreSmoother", out, success);
-          TEST_EQUALITY(level.IsAvailable("PostSmoother", smooFact.get()), false);
-
-        } else if (preOrPost == MueLu::POST) {
-
-          TEST_EQUALITY(level.IsAvailable("PreSmoother", smooFact.get()), false);
-          testBuildCheckOutput(smooFact, level, smooProtoB, "PostSmoother", out, success);
-
-        } else { TEST_EQUALITY(true, false); }
-
-      }
-
-      //
-      static void testBuild(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherFactory> smooFact = rcp( new SmootherFactory(smooProtoA, smooProtoB) );
-
-        Level level; //level.SetupPhase(true);
-        TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
-
-        level.Request("PreSmoother",smooFact.get());
-        level.Request("PostSmoother", smooFact.get());
-
-        smooFact->Build(level);
-
-        testBuildCheck(smooFact, level, smooProtoA, smooProtoB, MueLu::BOTH, out, success);
-      }
-
-      static void testBuildSmootherPreOrPost(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, MueLu::PreOrPost preOrPost, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherFactory> smooFact = rcp( new SmootherFactory(smooProtoA, smooProtoB) );
-
-        Level level; //level.SetupPhase(true);
-        TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
-
-        level.Request("PreSmoother",smooFact.get());
-        level.Request("PostSmoother", smooFact.get());
-
-        smooFact->BuildSmoother(level, preOrPost);
-
-        testBuildCheck(smooFact, level, smooProtoA, smooProtoB, preOrPost, out, success);
+    }
+  }
+
+  // test outpout of Build() and BuildSmoother()
+  static void testBuildCheck(RCP<SmootherFactory> smooFact, Level &level, RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, MueLu::PreOrPost preOrPost, Teuchos::FancyOStream &out, bool &success) {
+    // invariant: smoother prototypes kept unchanged
+    check(smooFact, smooProtoA, smooProtoB, out, success);
+
+    // output test
+    if (preOrPost == MueLu::BOTH) {
+      testBuildCheckOutput(smooFact, level, smooProtoA, "PreSmoother", out, success);
+      testBuildCheckOutput(smooFact, level, smooProtoB, "PostSmoother", out, success);
+
+      // ReUse: if pre and post prototype are the same, then pre smoother == post smoother
+      // otherwise, they are different (have not been tested by previous tests)
+      RCP<SmootherBase> smooA, smooB;
+      if (smooProtoA != Teuchos::null) {
+        smooA = level.Get<RCP<SmootherBase> >("PreSmoother", smooFact.get());
       }
-
-      static void testBuildSmoother(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::PRE,  out, success);
-        testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::POST, out, success);
-        testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::BOTH, out, success);
+      if (smooProtoB != Teuchos::null) {
+        smooB = level.Get<RCP<SmootherBase> >("PostSmoother", smooFact.get());
       }
-
-      static void testBuildSmootherDefaultArg(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream & out, bool & success) {
-        RCP<SmootherFactory> smooFact = rcp( new SmootherFactory(smooProtoA, smooProtoB) );
-
-        Level level; //level.SetupPhase(true);
-        TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
-
-        level.Request("PreSmoother",smooFact.get());
-        level.Request("PostSmoother", smooFact.get());
-
-        smooFact->BuildSmoother(level);
-
-        testBuildCheck(smooFact, level, smooProtoA, smooProtoB, MueLu::BOTH, out, success);
+      if (smooProtoA == smooProtoB) {
+        TEST_EQUALITY(smooA, smooB);
+      } else {
+        TEST_INEQUALITY(smooA, smooB);
       }
 
-    }; //class SmootherFactoryHelpers
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: Constructor_OneArg
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Constructor_OneArg, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testConstructor1, out, success);               // TEST: Valid input parameter
-      SFH::testInvalidCollection(&SFH::testInvalidConstructor1, out, success); // TEST: Invalid input parameter
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: Constructor_TwoArg
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Constructor_TwoArg, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testConstructor2, out, success); // TEST: Valid input parameter
-      SFH::testInvalidCollection(&SFH::testInvalidConstructor2, out, success); // TEST: Valid input parameter
-    }
+    } else if (preOrPost == MueLu::PRE) {
+      testBuildCheckOutput(smooFact, level, smooProtoA, "PreSmoother", out, success);
+      TEST_EQUALITY(level.IsAvailable("PostSmoother", smooFact.get()), false);
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: SetSmootherPrototypes_OneArg
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, SetSmootherPrototypes_OneArg, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testSetSmootherPrototypes1, out, success); // TEST: Valid input parameter
-      SFH::testInvalidCollection(&SFH::testInvalidSetSmootherPrototypes1, out, success); // TEST: Invalid input parameter
-    }
+    } else if (preOrPost == MueLu::POST) {
+      TEST_EQUALITY(level.IsAvailable("PreSmoother", smooFact.get()), false);
+      testBuildCheckOutput(smooFact, level, smooProtoB, "PostSmoother", out, success);
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: SetSmootherPrototypes_TwoArg
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, SetSmootherPrototypes_TwoArg, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testSetSmootherPrototypes2, out, success); // TEST: Valid input parameter
-      SFH::testInvalidCollection(&SFH::testInvalidSetSmootherPrototypes2, out, success); // TEST: Invalid input parameter
+    } else {
+      TEST_EQUALITY(true, false);
     }
+  }
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: GetSmootherPrototypes
-    ////////////////////////////////////////////////////////////////////////////////
+  //
+  static void testBuild(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherFactory> smooFact = rcp(new SmootherFactory(smooProtoA, smooProtoB));
 
-    /*
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, GetSmootherPrototypes, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-      // Already heavily tested by previous tests.
-    }
-    */
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: DeclareInput
-    ////////////////////////////////////////////////////////////////////////////////
-
-    /*
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, DeclareInput, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-      // TODO:
-      // Create a level, call DeclareInput() and Build() and then check that every counter == 0 at the end
-      // Same test have to be done for each factory...
-    }
-    */
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: Build
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testBuild, out, success);
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: BuildSmoother
-    ////////////////////////////////////////////////////////////////////////////////
-
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, BuildSmoother, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SFH::testCollection(&SFH::testBuildSmootherDefaultArg, out, success); // TEST: default arg
-      SFH::testCollection(&SFH::testBuildSmoother, out, success);           // TEST: PRE, POST and BOTH
-    }
-
-    ////////////////////////////////////////////////////////////////////////////////
-    // Test: Compilation
-    ////////////////////////////////////////////////////////////////////////////////
-
-    // Make sure that Constructors and SetSmootherPrototypes() are not using references as input arguments
-    // (to allow Teuchos::null as input arguments)
-    //TODO: const ref?
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Compilation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-    {
-#     include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      typedef SmootherFactoryHelpers<SC,LO,GO,NO> SFH;
-      SmootherFactory smooFact1(Teuchos::null);
-      SmootherFactory smooFact2(Teuchos::null, Teuchos::null);
-      smooFact1.SetSmootherPrototypes(Teuchos::null);
-      smooFact1.SetSmootherPrototypes(Teuchos::null, Teuchos::null);
-    }
+    Level level;  // level.SetupPhase(true);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
 
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Constructor_OneArg, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Constructor_TwoArg, SC, LO, GO, NO) \
+    level.Request("PreSmoother", smooFact.get());
+    level.Request("PostSmoother", smooFact.get());
+
+    smooFact->Build(level);
+
+    testBuildCheck(smooFact, level, smooProtoA, smooProtoB, MueLu::BOTH, out, success);
+  }
+
+  static void testBuildSmootherPreOrPost(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, MueLu::PreOrPost preOrPost, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherFactory> smooFact = rcp(new SmootherFactory(smooProtoA, smooProtoB));
+
+    Level level;  // level.SetupPhase(true);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+
+    level.Request("PreSmoother", smooFact.get());
+    level.Request("PostSmoother", smooFact.get());
+
+    smooFact->BuildSmoother(level, preOrPost);
+
+    testBuildCheck(smooFact, level, smooProtoA, smooProtoB, preOrPost, out, success);
+  }
+
+  static void testBuildSmoother(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::PRE, out, success);
+    testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::POST, out, success);
+    testBuildSmootherPreOrPost(smooProtoA, smooProtoB, MueLu::BOTH, out, success);
+  }
+
+  static void testBuildSmootherDefaultArg(RCP<SmootherPrototype> smooProtoA, RCP<SmootherPrototype> smooProtoB, Teuchos::FancyOStream &out, bool &success) {
+    RCP<SmootherFactory> smooFact = rcp(new SmootherFactory(smooProtoA, smooProtoB));
+
+    Level level;  // level.SetupPhase(true);
+    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+
+    level.Request("PreSmoother", smooFact.get());
+    level.Request("PostSmoother", smooFact.get());
+
+    smooFact->BuildSmoother(level);
+
+    testBuildCheck(smooFact, level, smooProtoA, smooProtoB, MueLu::BOTH, out, success);
+  }
+
+};  // class SmootherFactoryHelpers
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: Constructor_OneArg
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Constructor_OneArg, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testConstructor1, out, success);                // TEST: Valid input parameter
+  SFH::testInvalidCollection(&SFH::testInvalidConstructor1, out, success);  // TEST: Invalid input parameter
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: Constructor_TwoArg
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Constructor_TwoArg, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testConstructor2, out, success);                // TEST: Valid input parameter
+  SFH::testInvalidCollection(&SFH::testInvalidConstructor2, out, success);  // TEST: Valid input parameter
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: SetSmootherPrototypes_OneArg
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, SetSmootherPrototypes_OneArg, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testSetSmootherPrototypes1, out, success);                // TEST: Valid input parameter
+  SFH::testInvalidCollection(&SFH::testInvalidSetSmootherPrototypes1, out, success);  // TEST: Invalid input parameter
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: SetSmootherPrototypes_TwoArg
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, SetSmootherPrototypes_TwoArg, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testSetSmootherPrototypes2, out, success);                // TEST: Valid input parameter
+  SFH::testInvalidCollection(&SFH::testInvalidSetSmootherPrototypes2, out, success);  // TEST: Invalid input parameter
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: GetSmootherPrototypes
+////////////////////////////////////////////////////////////////////////////////
+
+/*
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, GetSmootherPrototypes, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+{
+  // Already heavily tested by previous tests.
+}
+*/
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: DeclareInput
+////////////////////////////////////////////////////////////////////////////////
+
+/*
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, DeclareInput, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+{
+  // TODO:
+  // Create a level, call DeclareInput() and Build() and then check that every counter == 0 at the end
+  // Same test have to be done for each factory...
+}
+*/
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: Build
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testBuild, out, success);
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: BuildSmoother
+////////////////////////////////////////////////////////////////////////////////
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, BuildSmoother, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SFH::testCollection(&SFH::testBuildSmootherDefaultArg, out, success);  // TEST: default arg
+  SFH::testCollection(&SFH::testBuildSmoother, out, success);            // TEST: PRE, POST and BOTH
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// Test: Compilation
+////////////////////////////////////////////////////////////////////////////////
+
+// Make sure that Constructors and SetSmootherPrototypes() are not using references as input arguments
+// (to allow Teuchos::null as input arguments)
+// TODO: const ref?
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SmootherFactory, Compilation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  typedef SmootherFactoryHelpers<SC, LO, GO, NO> SFH;
+  SmootherFactory smooFact1(Teuchos::null);
+  SmootherFactory smooFact2(Teuchos::null, Teuchos::null);
+  smooFact1.SetSmootherPrototypes(Teuchos::null);
+  smooFact1.SetSmootherPrototypes(Teuchos::null, Teuchos::null);
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Constructor_OneArg, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Constructor_TwoArg, SC, LO, GO, NO)           \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, SetSmootherPrototypes_OneArg, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, SetSmootherPrototypes_TwoArg, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Build, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, BuildSmoother, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Build, SC, LO, GO, NO)                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, BuildSmoother, SC, LO, GO, NO)                \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SmootherFactory, Compilation, SC, LO, GO, NO)
 #include <MueLu_ETI_4arg.hpp>
 
-  } // namespace SmootherFactory
+}  // namespace SmootherFactory
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
 
-//TODO: advance reuse test
-//TODO: remove some useless RCP in helper function prototypes?
+// TODO: advance reuse test
+// TODO: remove some useless RCP in helper function prototypes?
diff --git a/packages/muelu/test/unit_tests/Smoothers/StratimikosSmoother.cpp b/packages/muelu/test/unit_tests/Smoothers/StratimikosSmoother.cpp
index 9652eecdbe16..ed216059df9f 100644
--- a/packages/muelu/test/unit_tests/Smoothers/StratimikosSmoother.cpp
+++ b/packages/muelu/test/unit_tests/Smoothers/StratimikosSmoother.cpp
@@ -50,54 +50,54 @@
 #include <MueLu_StratimikosSmoother.hpp>
 #include <MueLu_Utilities.hpp>
 
-
 namespace MueLuTests {
 
-  // this namespace already has:  #include "MueLu_UseShortNames.hpp"
-  using namespace TestHelpers::Smoothers;
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StratimikosSmoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    if (!TYPE_EQUAL(Scalar, double)) { out << "Skipping for SC != double" << std::endl; return; }
-
-    StratimikosSmoother smoother("STRATIMIKOS", Teuchos::ParameterList());
-    testApplyNoSetup(smoother, out, success);
+// this namespace already has:  #include "MueLu_UseShortNames.hpp"
+using namespace TestHelpers::Smoothers;
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StratimikosSmoother, NotSetup, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  if (!TYPE_EQUAL(Scalar, double)) {
+    out << "Skipping for SC != double" << std::endl;
+    return;
   }
 
-  // Tests interface to Stratimikos's Gauss-Seidel preconditioner.
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StratimikosSmoother, HardCodedResult_BlockCG, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    if (!TYPE_EQUAL(Scalar, double)) { out << "Skipping for SC != double" << std::endl; return; }
-
-    Teuchos::ParameterList paramList;
-    paramList.set("Linear Solver Type", "Belos");
-    paramList.sublist("Linear Solver Types").sublist("Belos").set("Solver Type", "Block CG");
-    paramList.sublist("Linear Solver Types").sublist("Belos").sublist("Solver Types").sublist("Block CG").set("Convergence Tolerance", 1e-2);
-    paramList.sublist("Linear Solver Types").sublist("Belos").sublist("Solver Types").sublist("Block CG").set("Maximum Iterations", 100);
-    paramList.set("Preconditioner Type", "None");
-    StratimikosSmoother smoother("STRATIMIKOS", paramList);
-
+  StratimikosSmoother smoother("STRATIMIKOS", Teuchos::ParameterList());
+  testApplyNoSetup(smoother, out, success);
+}
+
+// Tests interface to Stratimikos's Gauss-Seidel preconditioner.
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StratimikosSmoother, HardCodedResult_BlockCG, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  if (!TYPE_EQUAL(Scalar, double)) {
+    out << "Skipping for SC != double" << std::endl;
+    return;
+  }
 
-    typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
+  Teuchos::ParameterList paramList;
+  paramList.set("Linear Solver Type", "Belos");
+  paramList.sublist("Linear Solver Types").sublist("Belos").set("Solver Type", "Block CG");
+  paramList.sublist("Linear Solver Types").sublist("Belos").sublist("Solver Types").sublist("Block CG").set("Convergence Tolerance", 1e-2);
+  paramList.sublist("Linear Solver Types").sublist("Belos").sublist("Solver Types").sublist("Block CG").set("Maximum Iterations", 100);
+  paramList.set("Preconditioner Type", "None");
+  StratimikosSmoother smoother("STRATIMIKOS", paramList);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 0.02118661536508828144;
-    TEST_FLOATING_EQUALITY(residualNorms,expectedNorm,0.1);
+  typename Teuchos::ScalarTraits<SC>::magnitudeType residualNorms = testApply_A125_X0_RandomRHS(smoother, out, success);
 
-  } // Block CG
+  RCP<const Teuchos::Comm<int> > comm                                  = TestHelpers::Parameters::getDefaultComm();
+  const typename Teuchos::ScalarTraits<SC>::magnitudeType expectedNorm = 0.02118661536508828144;
+  TEST_FLOATING_EQUALITY(residualNorms, expectedNorm, 0.1);
 
+}  // Block CG
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StratimikosSmoother,NotSetup,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StratimikosSmoother,HardCodedResult_BlockCG,SC,LO,GO,NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StratimikosSmoother, NotSetup, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StratimikosSmoother, HardCodedResult_BlockCG, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/StructuredAggregationFactory.cpp b/packages/muelu/test/unit_tests/StructuredAggregationFactory.cpp
index e1d472124bd2..60396ae560a8 100644
--- a/packages/muelu/test/unit_tests/StructuredAggregationFactory.cpp
+++ b/packages/muelu/test/unit_tests/StructuredAggregationFactory.cpp
@@ -64,928 +64,1201 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Global Lexicographic";
-    LO numDimensions = 1;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 20;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Global Lexicographic";
+  LO numDimensions             = 1;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 20;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<SC,LO,GO,Node> > PtentTPtent = Xpetra::MatrixMatrix<SC,LO,GO,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<SC,LO,GO,Node> > diagVec = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // GlobalLexiTentative1D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Global Lexicographic";
-    LO numDimensions = 2;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        gNodesPerDir[dim] = 12;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, Node> > PtentTPtent = Xpetra::MatrixMatrix<SC, LO, GO, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<SC, LO, GO, Node> > diagVec     = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // GlobalLexiTentative1D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Global Lexicographic";
+  LO numDimensions             = 2;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      gNodesPerDir[dim] = 12;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // GlobalLexiTentative2D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Global Lexicographic";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // GlobalLexiTentative2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, GlobalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Global Lexicographic";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // GlobalLexiTentative3D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    LO numDimensions = 1;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 20;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // GlobalLexiTentative3D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  LO numDimensions             = 1;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 20;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // LocalLexiTentative1D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    LO numDimensions = 2;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 12;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // LocalLexiTentative1D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  LO numDimensions             = 2;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 12;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // LocalLexiTentative2D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // LocalLexiTentative2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, LocalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // LocalLexiTentative3D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling   = "uncoupled";
-    LO numDimensions = 1;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 20;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // LocalLexiTentative3D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 1;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 20;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace1D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: mode",
+                                  Teuchos::ParameterEntry(coupling));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // UncoupledLocalLexiTentative1D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 2;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 12;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace1D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates", Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim", gNodesPerDir);
-    fineLevel.Set("lNodesPerDim", lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: mode",
-                                    Teuchos::ParameterEntry(coupling));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // UncoupledLocalLexiTentative1D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 2;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 12;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: mode",
+                                  Teuchos::ParameterEntry(coupling));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // UncoupledLocalLexiTentative2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: mode",
+                                  Teuchos::ParameterEntry(coupling));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // UncoupledLocalLexiTentative3D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledMultilevelScalar, Scalar,
+                                  LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using magnitude_type        = typename Teuchos::ScalarTraits<SC>::magnitudeType;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: mode",
-                                    Teuchos::ParameterEntry(coupling));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // UncoupledLocalLexiTentative2D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledLocalLexiTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("ny", gNodesPerDir[1]);
+  matrixList.set("nz", gNodesPerDir[2]);
+  matrixList.set("matrixType", "Laplace3D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (Coordinates->getMap()->getComm()->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> AmalgFact     = rcp(new AmalgamationFactory());
+  RCP<Factory> CDropfact     = rcp(new CoalesceDropFactory());
+  RCP<Factory> Aggfact       = rcp(new StructuredAggregationFactory());
+  RCP<Factory> coarseMapFact = rcp(new CoarseMapFactory());
+  RCP<Factory> Pfact         = rcp(new TentativePFactory());
+  RCP<Factory> Rfact         = rcp(new TransPFactory());
+  RCP<Factory> Tfact         = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact     = rcp(new RAPFactory());
+  RCP<Factory> NSfact        = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  CDropfact->SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
+  CDropfact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.0));
+
+  Aggfact->SetParameter("aggregation: mode", Teuchos::ParameterEntry(std::string("uncoupled")));
+
+  Tfact->SetParameter("structured aggregation", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  CDropfact->SetFactory("UnAmalgamationInfo", AmalgFact);
+  Aggfact->SetFactory("Graph", CDropfact);
+  Aggfact->SetFactory("DofsPerNode", CDropfact);
+  coarseMapFact->SetFactory("Aggregates", Aggfact);
+  Pfact->SetFactory("UnAmalgamationInfo", AmalgFact);
+  Pfact->SetFactory("Aggregates", Aggfact);
+  Pfact->SetFactory("CoarseMap", coarseMapFact);
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("Graph", CDropfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("numDimensions", Tfact);
+  M.SetFactory("lCoarseNodesPerDim", Aggfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", A);                  // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+  // Finest->Set("Coordinates", Coordinates);    // set fine level coordinates
+  Finest->Set("numDimensions", numDimensions);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDir);    // set GeneralGeometricPFactory specific info
+
+  // Setup the hierarchy
+  LO maxLevels = 5;
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  // // Extract the prolongator operator
+  // RCP<Level> lvl1 = H->GetLevel(1);
+  // RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
+  // RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
+
+  // // Construct vectors to check that a linear vector remains linear after projection
+  // RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
+  //   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getRangeMap(),1);
+  // RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
+  //   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getDomainMap(),1);
+  // ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
+  // if(comm->getRank() == 0) {
+  //   for(LO i = 0; i < 3; ++i) {
+  //     for(LO j = 0; j < 3; ++j) {
+  //       coarse_data[3*i + j] = 2.0*i + j;
+  //     }
+  //   }
+  // }
+
+  // PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
+  //             Teuchos::ScalarTraits<SC>::zero());
+
+  // ArrayRCP<const SC> fine_data   = vector0->getData(0);
+  // bool is_constant = true, is_injected = true;
+  // Array<LO> fine_inds(9);
+  // Array<LO> coarse_inds(9);
+  // if(comm->getRank() == 0) {
+  //   fine_inds[0] = 0;
+  //   fine_inds[1] = 2;
+  //   fine_inds[2] = 3;
+  //   fine_inds[3] = 8;
+  //   fine_inds[4] = 10;
+  //   fine_inds[5] = 11;
+  //   fine_inds[6] = 12;
+  //   fine_inds[7] = 14;
+  //   fine_inds[8] = 15;
+  //   for(LO i = 0; i < 9; ++i) {
+  //     if(std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) { is_injected = false; }
+  //   }
+  //   if( (std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
+  //       (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
+  //       (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) { is_constant = false; }
+  // }
+
+  TEST_EQUALITY(A != Teuchos::null, true);
+  TEST_EQUALITY(H != Teuchos::null, true);
+  TEST_EQUALITY(nullSpace != Teuchos::null, true);
+  // TEST_EQUALITY(is_injected, true);
+  // TEST_EQUALITY(is_constant, true);
+
+}  // UncoupledMultilevelScalar
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, ProlongatorGraphUncoupled, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, Node>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  for (int interpolationOrder = 0; interpolationOrder < 2; ++interpolationOrder) {
+    out << "Tesing 7x7 grid with piece-wise "
+        << (interpolationOrder == 0 ? "constant" : "linear") << " interpolation" << std::endl;
+    Level fineLevel;
+    test_factory::createSingleLevelHierarchy(fineLevel);
     fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
 
     // Set global geometric data
     const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 3;
+    const std::string coupling   = "uncoupled";
+    LO numDimensions             = 2;
     Array<GO> meshData;
     Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
-    }
+    Array<GO> gNodesPerDir = {{7, 7, 1}};
 
     RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
+        TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                      lNodesPerDir, meshData,
+                                                                      meshLayout);
 
     // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
+    for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
       meshData[idx] = -1;
     }
 
     Teuchos::ParameterList matrixList;
     matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
+    matrixList.set("ny", gNodesPerDir[1]);
+    matrixList.set("matrixType", "Laplace2D");
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+        BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                  matrixList);
     RCP<Matrix> A = Pr->BuildMatrix();
     fineLevel.Request("A");
     fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
+    fineLevel.Set("Coordinates", Coordinates);
     fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
+    fineLevel.Set("gNodesPerDim", gNodesPerDir);
+    fineLevel.Set("lNodesPerDim", lNodesPerDir);
     fineLevel.Set("aggregation: mesh data", meshData);
 
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-
     RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
     dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
     RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
     StructuredAggFact->SetFactory("Graph", dropFact);
@@ -994,697 +1267,129 @@ namespace MueLuTests {
                                     Teuchos::ParameterEntry(meshLayout));
     StructuredAggFact->SetParameter("aggregation: mode",
                                     Teuchos::ParameterEntry(coupling));
+    StructuredAggFact->SetParameter("aggregation: output type",
+                                    Teuchos::ParameterEntry(std::string("Graph")));
     StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
+                                    Teuchos::ParameterEntry(interpolationOrder));
     StructuredAggFact->SetParameter("aggregation: coarsening rate",
                                     Teuchos::ParameterEntry(std::string("{3}")));
 
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // UncoupledLocalLexiTentative3D
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledMultilevelScalar, Scalar,
-                                    LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using magnitude_type        = typename Teuchos::ScalarTraits<SC>::magnitudeType;
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType ;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
+    fineLevel.Request("prolongatorGraph", StructuredAggFact.get());
+    fineLevel.Request(*StructuredAggFact);
+    StructuredAggFact->Build(fineLevel);
 
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
-    }
+    RCP<CrsGraph> prolongatorGraph;
+    fineLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
 
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("ny", gNodesPerDir[1]);
-    matrixList.set("nz", gNodesPerDir[2]);
-    matrixList.set("matrixType","Laplace3D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
+    TEST_EQUALITY(prolongatorGraph != Teuchos::null, true);
 
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (Coordinates->getMap()->getComm()->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
+    int numErrors      = 0;
+    const int numRanks = Coordinates->getMap()->getComm()->getSize();
+    TEST_EQUALITY(prolongatorGraph->getGlobalNumRows() == 49, true);
+    if (interpolationOrder == 0) {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 49, true);
+    } else {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == (numRanks == 1 ? 169 : 148), true);
     }
+    if (numRanks == 1) {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 9, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == 49, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == 9, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == (interpolationOrder == 1 ? 169 : 49), true);
 
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      AmalgFact     = rcp( new AmalgamationFactory() );
-    RCP<Factory>      CDropfact     = rcp( new CoalesceDropFactory() );
-    RCP<Factory>      Aggfact       = rcp( new StructuredAggregationFactory() );
-    RCP<Factory>      coarseMapFact = rcp( new CoarseMapFactory() );
-    RCP<Factory>      Pfact         = rcp( new TentativePFactory() );
-    RCP<Factory>      Rfact         = rcp( new TransPFactory() );
-    RCP<Factory>      Tfact         = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact        = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact        = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    CDropfact->SetParameter("lightweight wrap", Teuchos::ParameterEntry(true));
-    CDropfact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.0));
-
-    Aggfact->SetParameter("aggregation: mode", Teuchos::ParameterEntry(std::string("uncoupled")));
-
-    Tfact->SetParameter("structured aggregation", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    CDropfact->SetFactory("UnAmalgamationInfo", AmalgFact);
-    Aggfact->SetFactory("Graph", CDropfact);
-    Aggfact->SetFactory("DofsPerNode", CDropfact);
-    coarseMapFact->SetFactory("Aggregates", Aggfact);
-    Pfact->SetFactory("UnAmalgamationInfo", AmalgFact);
-    Pfact->SetFactory("Aggregates", Aggfact);
-    Pfact->SetFactory("CoarseMap", coarseMapFact);
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",                  Pfact);
-    M.SetFactory("R",                  Rfact);
-    M.SetFactory("A",                  Acfact);
-    M.SetFactory("Nullspace",          NSfact);
-    M.SetFactory("Graph",              CDropfact);
-    M.SetFactory("lNodesPerDim",       Tfact);
-    M.SetFactory("numDimensions",      Tfact);
-    M.SetFactory("lCoarseNodesPerDim", Aggfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A", A);                        // set fine level matrix
-    Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
-    // Finest->Set("Coordinates", Coordinates);    // set fine level coordinates
-    Finest->Set("numDimensions", numDimensions);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim",  lNodesPerDir);  // set GeneralGeometricPFactory specific info
-
-    // Setup the hierarchy
-    LO maxLevels = 5;
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    // // Extract the prolongator operator
-    // RCP<Level> lvl1 = H->GetLevel(1);
-    // RCP<Xpetra::Matrix<SC,LO,GO,Node> > P = lvl1->Get<RCP<Matrix> >("P", MueLu::NoFactory::get());
-    // RCP<CrsMatrix> PCrs = rcp_dynamic_cast<CrsMatrixWrap>(P)->getCrsMatrix();
-
-    // // Construct vectors to check that a linear vector remains linear after projection
-    // RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector0
-    //   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getRangeMap(),1);
-    // RCP<Xpetra::MultiVector<SC,LO,GO,NO> > vector1
-    //   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(PCrs->getDomainMap(),1);
-    // ArrayRCP<SC> coarse_data = vector1->getDataNonConst(0);
-    // if(comm->getRank() == 0) {
-    //   for(LO i = 0; i < 3; ++i) {
-    //     for(LO j = 0; j < 3; ++j) {
-    //       coarse_data[3*i + j] = 2.0*i + j;
-    //     }
-    //   }
-    // }
-
-    // PCrs->apply(*vector1, *vector0, Teuchos::NO_TRANS, Teuchos::ScalarTraits<SC>::one(),
-    //             Teuchos::ScalarTraits<SC>::zero());
-
-    // ArrayRCP<const SC> fine_data   = vector0->getData(0);
-    // bool is_constant = true, is_injected = true;
-    // Array<LO> fine_inds(9);
-    // Array<LO> coarse_inds(9);
-    // if(comm->getRank() == 0) {
-    //   fine_inds[0] = 0;
-    //   fine_inds[1] = 2;
-    //   fine_inds[2] = 3;
-    //   fine_inds[3] = 8;
-    //   fine_inds[4] = 10;
-    //   fine_inds[5] = 11;
-    //   fine_inds[6] = 12;
-    //   fine_inds[7] = 14;
-    //   fine_inds[8] = 15;
-    //   for(LO i = 0; i < 9; ++i) {
-    //     if(std::abs(fine_data[fine_inds[i]] - coarse_data[i]) > 1.0e-10) { is_injected = false; }
-    //   }
-    //   if( (std::abs(fine_data[1] - coarse_data[0]) > 1.0e-10) ||
-    //       (std::abs(fine_data[4] - coarse_data[0]) > 1.0e-10) ||
-    //       (std::abs(fine_data[5] - coarse_data[0]) > 1.0e-10)) { is_constant = false; }
-    // }
-
-
-    TEST_EQUALITY(A != Teuchos::null, true);
-    TEST_EQUALITY(H != Teuchos::null, true);
-    TEST_EQUALITY(nullSpace != Teuchos::null, true);
-    // TEST_EQUALITY(is_injected, true);
-    // TEST_EQUALITY(is_constant, true);
-
-  } // UncoupledMultilevelScalar
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, ProlongatorGraphUncoupled, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, Node>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    for(int interpolationOrder = 0; interpolationOrder < 2; ++interpolationOrder) {
-      out << "Tesing 7x7 grid with piece-wise "
-          << (interpolationOrder == 0 ? "constant" : "linear") << " interpolation" << std::endl;
-      Level fineLevel;
-      test_factory::createSingleLevelHierarchy(fineLevel);
-      fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-
-      // Set global geometric data
-      const std::string meshLayout = "Local Lexicographic";
-      const std::string coupling = "uncoupled";
-      LO numDimensions = 2;
-      Array<GO> meshData;
-      Array<LO> lNodesPerDir(3);
-      Array<GO> gNodesPerDir = {{7, 7, 1}};
-
-      RCP<RealValuedMultiVector> Coordinates =
-        TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                   lNodesPerDir, meshData,
-                                                                   meshLayout);
-
-      // Since we are doing uncoupled aggregation we fill meshData with -1
-      for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-        meshData[idx] = -1;
-      }
+      ArrayView<const LO> rowIndices;
+      const LO indicesConstant[49] = {0, 0, 1, 1, 1, 2, 2,
+                                      0, 0, 1, 1, 1, 2, 2,
+                                      3, 3, 4, 4, 4, 5, 5,
+                                      3, 3, 4, 4, 4, 5, 5,
+                                      3, 3, 4, 4, 4, 5, 5,
+                                      6, 6, 7, 7, 7, 8, 8,
+                                      6, 6, 7, 7, 7, 8, 8};
+      const LO indicesLinear[169]  = {0, 0, 1, 3, 4, 0, 1, 3, 4, 1, 1, 2, 4, 5, 1, 2, 4, 5, 2,
+                                      0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
+                                      0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
+                                      3, 3, 4, 6, 7, 3, 4, 6, 7, 4, 4, 5, 7, 8, 4, 5, 7, 8, 5,
+                                      3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8,
+                                      3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8,
+                                      6, 3, 4, 6, 7, 3, 4, 6, 7, 7, 4, 5, 7, 8, 4, 5, 7, 8, 8};
+      const LO rowOffset[50]       = {0, 1, 5, 9, 10, 14, 18, 19,
+                                      23, 27, 31, 35, 39, 43, 47,
+                                      51, 55, 59, 63, 67, 71, 75,
+                                      76, 80, 84, 85, 89, 93, 94,
+                                      98, 102, 106, 110, 114, 118, 122,
+                                      126, 130, 134, 138, 142, 146, 150,
+                                      151, 155, 159, 160, 164, 168, 169};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
 
-      Teuchos::ParameterList matrixList;
-      matrixList.set("nx", gNodesPerDir[0]);
-      matrixList.set("ny", gNodesPerDir[1]);
-      matrixList.set("matrixType", "Laplace2D");
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-        BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                             matrixList);
-      RCP<Matrix> A = Pr->BuildMatrix();
-      fineLevel.Request("A");
-      fineLevel.Set("A", A);
-      fineLevel.Set("Coordinates",   Coordinates);
-      fineLevel.Set("numDimensions", numDimensions);
-      fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-      fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-      fineLevel.Set("aggregation: mesh data", meshData);
-
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-      StructuredAggFact->SetFactory("Graph", dropFact);
-      StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-      StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                      Teuchos::ParameterEntry(meshLayout));
-      StructuredAggFact->SetParameter("aggregation: mode",
-                                      Teuchos::ParameterEntry(coupling));
-      StructuredAggFact->SetParameter("aggregation: output type",
-                                      Teuchos::ParameterEntry(std::string("Graph")));
-      StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                      Teuchos::ParameterEntry(interpolationOrder));
-      StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                      Teuchos::ParameterEntry(std::string("{3}")));
-
-      fineLevel.Request("prolongatorGraph", StructuredAggFact.get());
-      fineLevel.Request(*StructuredAggFact);
-      StructuredAggFact->Build(fineLevel);
-
-      RCP<CrsGraph> prolongatorGraph;
-      fineLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
-
-      TEST_EQUALITY(prolongatorGraph != Teuchos::null, true);
-
-      int numErrors = 0;
-      const int numRanks = Coordinates->getMap()->getComm()->getSize();
-      TEST_EQUALITY(prolongatorGraph->getGlobalNumRows()    == 49, true);
-      if(interpolationOrder == 0) {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 49, true);
-      } else {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == (numRanks == 1 ? 169 : 148), true);
-      }
-      if(numRanks == 1) {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  ==  9, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == 49, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    ==  9, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == (interpolationOrder == 1 ? 169 : 49), true);
-
-        ArrayView<const LO> rowIndices;
-        const LO indicesConstant[49] = {0, 0, 1, 1, 1, 2, 2,
-                                        0, 0, 1, 1, 1, 2, 2,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        6, 6, 7, 7, 7, 8, 8,
-                                        6, 6, 7, 7, 7, 8, 8};
-        const LO indicesLinear[169]  = {0, 0, 1, 3, 4, 0, 1, 3, 4, 1, 1, 2, 4, 5, 1, 2, 4, 5, 2,
-                                        0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
-                                        0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
-                                        3, 3, 4, 6, 7, 3, 4, 6, 7, 4, 4, 5, 7, 8, 4, 5, 7, 8, 5,
-                                        3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8,
-                                        3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8,
-                                        6, 3, 4, 6, 7, 3, 4, 6, 7, 7, 4, 5, 7, 8, 4, 5, 7, 8, 8};
-        const LO rowOffset[50]       = {0, 1,  5,   9,  10,  14,  18,  19,
-                                        23,   27,  31,  35,  39,  43,  47,
-                                        51,   55,  59,  63,  67,  71,  75,
-                                        76,   80,  84,  85,  89,  93,  94,
-                                        98,  102, 106, 110, 114, 118, 122,
-                                        126, 130, 134, 138, 142, 146, 150,
-                                        151, 155, 159, 160, 164, 168, 169};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-
-          if(interpolationOrder == 0) {
-            if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
-          } else {
-            for(size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
-              if(rowIndices[entryIdx] != indicesLinear[rowOffset[rowIdx] + entryIdx]) {++numErrors;}
-            }
+        if (interpolationOrder == 0) {
+          if (rowIndices[0] != indicesConstant[rowIdx]) {
+            ++numErrors;
           }
-        }
-      } else if(numRanks == 4) {
-        const int myRank = Coordinates->getMap()->getComm()->getRank();
-        Array<size_t> nodeNumRows    = {{16, 12, 12,  9}};
-        Array<int>    rankOffsets    = {{ 0, 16, 28, 40}};
-        Array<size_t> nodeNumEntries = {{52, 36, 36, 24}};
-
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  == 16, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == nodeNumRows[myRank], true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    == 4, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() ==
-                      ((interpolationOrder == 0) ? nodeNumRows[myRank] : nodeNumEntries[myRank]), true);
-
-        ArrayView<const LO> rowIndices;
-        const LO indicesConstant[49] = {0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3,
-                                        0, 0, 1, 0, 0, 1, 2, 2, 3, 2, 2, 3,
-                                        0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
-                                        0, 0, 1, 0, 0, 1, 2, 2, 3};
-        const LO indicesLinear[148]  = {0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 3};
-        const LO rowOffset[50]       = {0, 1, 5, 9, 10, 14, 18, 22, 26, 30, 34, 38, 42, 43, 47, 51 ,52,
-                                        53, 57, 58, 62, 66, 70, 74, 78, 82, 83, 87, 88,
-                                        89, 93, 97, 98, 102, 106, 110, 114, 115, 119, 123, 124,
-                                        125, 129, 130, 134, 138, 142, 143, 147, 148};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-
-          if(interpolationOrder == 0) {
-            if(rowIndices[0] != indicesConstant[rankOffsets[myRank] + rowIdx]) {++numErrors;}
-          } else {
-            for(size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
-              if(rowIndices[entryIdx] != indicesLinear[rowOffset[rankOffsets[myRank] + rowIdx] + entryIdx]) {++numErrors;}
+        } else {
+          for (size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
+            if (rowIndices[entryIdx] != indicesLinear[rowOffset[rowIdx] + entryIdx]) {
+              ++numErrors;
             }
           }
         }
       }
-      TEST_EQUALITY(numErrors == 0, true);
-    } // Loop over interpolationOrder
-
-    for(int interpolationOrder = 0; interpolationOrder < 2; ++interpolationOrder) {
-      out << "Tesing 7x6 grid with piece-wise "
-          << (interpolationOrder == 0 ? "constant" : "linear") << " interpolation" << std::endl;
-      Level fineLevel;
-      test_factory::createSingleLevelHierarchy(fineLevel);
-      fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-
-      // Set global geometric data
-      const std::string meshLayout = "Local Lexicographic";
-      const std::string coupling = "uncoupled";
-      LO numDimensions = 2;
-      Array<GO> meshData;
-      Array<LO> lNodesPerDir(3);
-      Array<GO> gNodesPerDir = {{7, 6, 1}};
-
-      RCP<RealValuedMultiVector> Coordinates =
-        TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                   lNodesPerDir, meshData,
-                                                                   meshLayout);
-
-      // Since we are doing uncoupled aggregation we fill meshData with -1
-      for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-        meshData[idx] = -1;
-      }
+    } else if (numRanks == 4) {
+      const int myRank             = Coordinates->getMap()->getComm()->getRank();
+      Array<size_t> nodeNumRows    = {{16, 12, 12, 9}};
+      Array<int> rankOffsets       = {{0, 16, 28, 40}};
+      Array<size_t> nodeNumEntries = {{52, 36, 36, 24}};
+
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 16, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == nodeNumRows[myRank], true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == 4, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() ==
+                        ((interpolationOrder == 0) ? nodeNumRows[myRank] : nodeNumEntries[myRank]),
+                    true);
 
-      Teuchos::ParameterList matrixList;
-      matrixList.set("nx", gNodesPerDir[0]);
-      matrixList.set("ny", gNodesPerDir[1]);
-      matrixList.set("matrixType", "Laplace2D");
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-        BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                             matrixList);
-      RCP<Matrix> A = Pr->BuildMatrix();
-      fineLevel.Request("A");
-      fineLevel.Set("A", A);
-      fineLevel.Set("Coordinates",   Coordinates);
-      fineLevel.Set("numDimensions", numDimensions);
-      fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-      fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-      fineLevel.Set("aggregation: mesh data", meshData);
-
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-      StructuredAggFact->SetFactory("Graph", dropFact);
-      StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-      StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                      Teuchos::ParameterEntry(meshLayout));
-      StructuredAggFact->SetParameter("aggregation: mode",
-                                      Teuchos::ParameterEntry(coupling));
-      StructuredAggFact->SetParameter("aggregation: output type",
-                                      Teuchos::ParameterEntry(std::string("Graph")));
-      StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                      Teuchos::ParameterEntry(interpolationOrder));
-      StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                      Teuchos::ParameterEntry(std::string("{3}")));
-
-      fineLevel.Request("prolongatorGraph", StructuredAggFact.get());
-      fineLevel.Request(*StructuredAggFact);
-      StructuredAggFact->Build(fineLevel);
-
-      RCP<CrsGraph> prolongatorGraph;
-      fineLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
-
-      TEST_EQUALITY(prolongatorGraph != Teuchos::null, true);
-
-      int numErrors = 0;
-      const int numRanks = Coordinates->getMap()->getComm()->getSize();
-      TEST_EQUALITY(prolongatorGraph->getGlobalNumRows()    == 42, true);
-      if(interpolationOrder == 0) {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 42, true);
-      } else {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == (numRanks == 1 ? 141 : 120), true);
-      }
-      if(numRanks == 1) {
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  ==  9, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == 42, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    ==  9, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == (interpolationOrder == 1 ? 141 : 42), true);
-
-        ArrayView<const LO> rowIndices;
-        const LO indicesConstant[42] = {0, 0, 1, 1, 1, 2, 2,
-                                        0, 0, 1, 1, 1, 2, 2,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        3, 3, 4, 4, 4, 5, 5,
-                                        6, 6, 7, 7, 7, 8, 8};
-        const LO indicesLinear[141]  = {0, 0, 1, 3, 4, 0, 1, 3, 4, 1, 1, 2, 4, 5, 1, 2, 4, 5, 2,
-                                        0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
-                                        0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
-                                        3, 3, 4, 6, 7, 3, 4, 6, 7, 4, 4, 5, 7, 8, 4, 5, 7, 8, 5,
-                                        3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8, 4, 5 ,7, 8,
-                                        6, 3, 4, 6, 7, 3, 4, 6, 7, 7, 4, 5, 7, 8, 4, 5, 7, 8, 8};
-        const LO rowOffset[43]       = {0, 1,  5,   9,  10,  14,  18,  19,
-                                        23,   27,  31,  35,  39,  43,  47,
-                                        51,   55,  59,  63,  67,  71,  75,
-                                        76,   80,  84,  85,  89,  93,  94,
-                                        98,  102, 106, 110, 114, 118, 122,
-                                        123, 127, 131, 132, 136, 140, 141};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-
-          if(interpolationOrder == 0) {
-            if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
-          } else {
-            for(size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
-              if(rowIndices[entryIdx] != indicesLinear[rowOffset[rowIdx] + entryIdx]) {++numErrors;}
-            }
+      ArrayView<const LO> rowIndices;
+      const LO indicesConstant[49] = {0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3,
+                                      0, 0, 1, 0, 0, 1, 2, 2, 3, 2, 2, 3,
+                                      0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
+                                      0, 0, 1, 0, 0, 1, 2, 2, 3};
+      const LO indicesLinear[148]  = {0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 3};
+      const LO rowOffset[50]       = {0, 1, 5, 9, 10, 14, 18, 22, 26, 30, 34, 38, 42, 43, 47, 51, 52,
+                                      53, 57, 58, 62, 66, 70, 74, 78, 82, 83, 87, 88,
+                                      89, 93, 97, 98, 102, 106, 110, 114, 115, 119, 123, 124,
+                                      125, 129, 130, 134, 138, 142, 143, 147, 148};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+
+        if (interpolationOrder == 0) {
+          if (rowIndices[0] != indicesConstant[rankOffsets[myRank] + rowIdx]) {
+            ++numErrors;
           }
-        }
-      } else if(numRanks == 4) {
-        const int myRank = Coordinates->getMap()->getComm()->getRank();
-        Array<size_t> nodeNumRows    = {{12,  9, 12,  9}};
-        Array<int>    rankOffsets    = {{ 0, 12, 21, 33}};
-        Array<size_t> nodeNumEntries = {{36, 24, 36, 24}};
-
-        TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  == 16, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == nodeNumRows[myRank], true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    == 4, true);
-        TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() ==
-                      ((interpolationOrder == 0) ? nodeNumRows[myRank] : nodeNumEntries[myRank]), true);
-
-        ArrayView<const LO> rowIndices;
-        const LO indicesConstant[42] = {0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
-                                        0, 0, 1, 0, 0, 1, 2, 2, 3,
-                                        0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
-                                        0, 0, 1, 0, 0, 1, 2, 2, 3};
-        const LO indicesLinear[120]  = {0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
-                                        0, 0, 1, 2, 3, 1,
-                                        0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-                                        2, 0, 1, 2, 3, 3};
-        const LO rowOffset[43]       = {0, 1, 5, 9, 10, 14, 18, 22, 26, 27, 31, 35, 36,
-                                        37, 41, 42, 46, 50, 54, 55, 59, 60,
-                                        61, 65, 69, 70, 74, 78, 82, 86, 87, 91, 95, 96,
-                                        97, 101, 102, 106, 110, 114, 115, 119, 120};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-
-          if(interpolationOrder == 0) {
-            if(rowIndices[0] != indicesConstant[rankOffsets[myRank] + rowIdx]) {++numErrors;}
-          } else {
-            for(size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
-              if(rowIndices[entryIdx] != indicesLinear[rowOffset[rankOffsets[myRank] + rowIdx] + entryIdx]) {++numErrors;}
+        } else {
+          for (size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
+            if (rowIndices[entryIdx] != indicesLinear[rowOffset[rankOffsets[myRank] + rowIdx] + entryIdx]) {
+              ++numErrors;
             }
           }
         }
       }
-      TEST_EQUALITY(numErrors == 0, true);
-    } // Loop over interpolationOrder
-
-  } // ProlongatorGraphUncoupled
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledAggSingleCoarseNode, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 2;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
     }
+    TEST_EQUALITY(numErrors == 0, true);
+  }  // Loop over interpolationOrder
 
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType", "Laplace2D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
-    fineLevel.Set("aggregation: mesh data", meshData);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
-    StructuredAggFact->SetFactory("Graph", dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: mesh layout",
-                                    Teuchos::ParameterEntry(meshLayout));
-    StructuredAggFact->SetParameter("aggregation: mode",
-                                    Teuchos::ParameterEntry(coupling));
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{2}")));
-    StructuredAggFact->SetParameter("aggregation: single coarse point",
-                                    Teuchos::ParameterEntry(true));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // UncoupledAggSingleCoarseNode
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledGraphSingleCoarseNode, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, Node>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    int interpolationOrder = 0;
+  for (int interpolationOrder = 0; interpolationOrder < 2; ++interpolationOrder) {
     out << "Tesing 7x6 grid with piece-wise "
         << (interpolationOrder == 0 ? "constant" : "linear") << " interpolation" << std::endl;
     Level fineLevel;
@@ -1693,19 +1398,19 @@ namespace MueLuTests {
 
     // Set global geometric data
     const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 2;
+    const std::string coupling   = "uncoupled";
+    LO numDimensions             = 2;
     Array<GO> meshData;
     Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir = {{7, 7, 1}};
+    Array<GO> gNodesPerDir = {{7, 6, 1}};
 
     RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                 lNodesPerDir, meshData,
-                                                                 meshLayout);
+        TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                      lNodesPerDir, meshData,
+                                                                      meshLayout);
 
     // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
+    for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
       meshData[idx] = -1;
     }
 
@@ -1713,21 +1418,20 @@ namespace MueLuTests {
     matrixList.set("nx", gNodesPerDir[0]);
     matrixList.set("ny", gNodesPerDir[1]);
     matrixList.set("matrixType", "Laplace2D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace2D", Coordinates->getMap(),
-                                                           matrixList);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+        BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                  matrixList);
     RCP<Matrix> A = Pr->BuildMatrix();
     fineLevel.Request("A");
     fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",   Coordinates);
+    fineLevel.Set("Coordinates", Coordinates);
     fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("gNodesPerDim",  gNodesPerDir);
-    fineLevel.Set("lNodesPerDim",  lNodesPerDir);
+    fineLevel.Set("gNodesPerDim", gNodesPerDir);
+    fineLevel.Set("lNodesPerDim", lNodesPerDir);
     fineLevel.Set("aggregation: mesh data", meshData);
 
-
     RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
     dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
     RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
     StructuredAggFact->SetFactory("Graph", dropFact);
@@ -1742,8 +1446,6 @@ namespace MueLuTests {
                                     Teuchos::ParameterEntry(interpolationOrder));
     StructuredAggFact->SetParameter("aggregation: coarsening rate",
                                     Teuchos::ParameterEntry(std::string("{3}")));
-    StructuredAggFact->SetParameter("aggregation: single coarse point",
-                                    Teuchos::ParameterEntry(true));
 
     fineLevel.Request("prolongatorGraph", StructuredAggFact.get());
     fineLevel.Request(*StructuredAggFact);
@@ -1754,95 +1456,401 @@ namespace MueLuTests {
 
     TEST_EQUALITY(prolongatorGraph != Teuchos::null, true);
 
-    int numErrors = 0;
+    int numErrors      = 0;
     const int numRanks = Coordinates->getMap()->getComm()->getSize();
-    TEST_EQUALITY(prolongatorGraph->getGlobalNumRows()    == 49, true);
-    TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 49, true);
-    if(numRanks == 1) {
-      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  ==  9, true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == 49, true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    ==  9, true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == 49, true);
+    TEST_EQUALITY(prolongatorGraph->getGlobalNumRows() == 42, true);
+    if (interpolationOrder == 0) {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 42, true);
+    } else {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == (numRanks == 1 ? 141 : 120), true);
+    }
+    if (numRanks == 1) {
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 9, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == 42, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == 9, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == (interpolationOrder == 1 ? 141 : 42), true);
 
       ArrayView<const LO> rowIndices;
-      const LO indicesConstant[49] = {0, 0, 1, 1, 1, 2, 2,
+      const LO indicesConstant[42] = {0, 0, 1, 1, 1, 2, 2,
                                       0, 0, 1, 1, 1, 2, 2,
                                       3, 3, 4, 4, 4, 5, 5,
                                       3, 3, 4, 4, 4, 5, 5,
                                       3, 3, 4, 4, 4, 5, 5,
-                                      6, 6, 7, 7, 7, 8, 8,
                                       6, 6, 7, 7, 7, 8, 8};
-      for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+      const LO indicesLinear[141]  = {0, 0, 1, 3, 4, 0, 1, 3, 4, 1, 1, 2, 4, 5, 1, 2, 4, 5, 2,
+                                      0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
+                                      0, 1, 3, 4, 0, 1, 3, 4, 0, 1, 3, 4, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5, 1, 2, 4, 5,
+                                      3, 3, 4, 6, 7, 3, 4, 6, 7, 4, 4, 5, 7, 8, 4, 5, 7, 8, 5,
+                                      3, 4, 6, 7, 3, 4, 6, 7, 3, 4, 6, 7, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8, 4, 5, 7, 8,
+                                      6, 3, 4, 6, 7, 3, 4, 6, 7, 7, 4, 5, 7, 8, 4, 5, 7, 8, 8};
+      const LO rowOffset[43]       = {0, 1, 5, 9, 10, 14, 18, 19,
+                                      23, 27, 31, 35, 39, 43, 47,
+                                      51, 55, 59, 63, 67, 71, 75,
+                                      76, 80, 84, 85, 89, 93, 94,
+                                      98, 102, 106, 110, 114, 118, 122,
+                                      123, 127, 131, 132, 136, 140, 141};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
         prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-        if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
-      }
-    } else if(numRanks == 4) {
-      const int myRank = Coordinates->getMap()->getComm()->getRank();
-      const Array<size_t> nodeNumRows    = {{16, 12, 12,  9}};
-      const Array<size_t> nodeNumCols    = {{ 4,  2,  2,  1}};
 
-      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols()  == 9, true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumRows()    == nodeNumRows[myRank], true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumCols()    == nodeNumCols[myRank], true);
-      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == nodeNumRows[myRank], true);
+        if (interpolationOrder == 0) {
+          if (rowIndices[0] != indicesConstant[rowIdx]) {
+            ++numErrors;
+          }
+        } else {
+          for (size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
+            if (rowIndices[entryIdx] != indicesLinear[rowOffset[rowIdx] + entryIdx]) {
+              ++numErrors;
+            }
+          }
+        }
+      }
+    } else if (numRanks == 4) {
+      const int myRank             = Coordinates->getMap()->getComm()->getRank();
+      Array<size_t> nodeNumRows    = {{12, 9, 12, 9}};
+      Array<int> rankOffsets       = {{0, 12, 21, 33}};
+      Array<size_t> nodeNumEntries = {{36, 24, 36, 24}};
+
+      TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 16, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == nodeNumRows[myRank], true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == 4, true);
+      TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() ==
+                        ((interpolationOrder == 0) ? nodeNumRows[myRank] : nodeNumEntries[myRank]),
+                    true);
 
       ArrayView<const LO> rowIndices;
-      if(myRank == 0) {
-        const LO indicesConstant[16] = {0, 0, 1, 1,
-                                        0, 0, 1, 1,
-                                        2, 2, 3, 3,
-                                        2, 2, 3, 3};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-          if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
+      const LO indicesConstant[42] = {0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
+                                      0, 0, 1, 0, 0, 1, 2, 2, 3,
+                                      0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3,
+                                      0, 0, 1, 0, 0, 1, 2, 2, 3};
+      const LO indicesLinear[120]  = {0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 0, 1, 2, 3, 3,
+                                      0, 0, 1, 2, 3, 1,
+                                      0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
+                                      2, 0, 1, 2, 3, 3};
+      const LO rowOffset[43]       = {0, 1, 5, 9, 10, 14, 18, 22, 26, 27, 31, 35, 36,
+                                      37, 41, 42, 46, 50, 54, 55, 59, 60,
+                                      61, 65, 69, 70, 74, 78, 82, 86, 87, 91, 95, 96,
+                                      97, 101, 102, 106, 110, 114, 115, 119, 120};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+
+        if (interpolationOrder == 0) {
+          if (rowIndices[0] != indicesConstant[rankOffsets[myRank] + rowIdx]) {
+            ++numErrors;
+          }
+        } else {
+          for (size_t entryIdx = 0; entryIdx < prolongatorGraph->getNumEntriesInLocalRow(rowIdx); ++entryIdx) {
+            if (rowIndices[entryIdx] != indicesLinear[rowOffset[rankOffsets[myRank] + rowIdx] + entryIdx]) {
+              ++numErrors;
+            }
+          }
+        }
+      }
+    }
+    TEST_EQUALITY(numErrors == 0, true);
+  }  // Loop over interpolationOrder
+
+}  // ProlongatorGraphUncoupled
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledAggSingleCoarseNode, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 2;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+    } else {
+      gNodesPerDir[dim] = 1;
+    }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace2D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: mode",
+                                  Teuchos::ParameterEntry(coupling));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{2}")));
+  StructuredAggFact->SetParameter("aggregation: single coarse point",
+                                  Teuchos::ParameterEntry(true));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // UncoupledAggSingleCoarseNode
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation, UncoupledGraphSingleCoarseNode, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, Node>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  int interpolationOrder = 0;
+  out << "Tesing 7x6 grid with piece-wise "
+      << (interpolationOrder == 0 ? "constant" : "linear") << " interpolation" << std::endl;
+  Level fineLevel;
+  test_factory::createSingleLevelHierarchy(fineLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 2;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir = {{7, 7, 1}};
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                    lNodesPerDir, meshData,
+                                                                    meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("ny", gNodesPerDir[1]);
+  matrixList.set("matrixType", "Laplace2D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace2D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("gNodesPerDim", gNodesPerDir);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+  fineLevel.Set("aggregation: mesh data", meshData);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory> StructuredAggFact = rcp(new StructuredAggregationFactory());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: mesh layout",
+                                  Teuchos::ParameterEntry(meshLayout));
+  StructuredAggFact->SetParameter("aggregation: mode",
+                                  Teuchos::ParameterEntry(coupling));
+  StructuredAggFact->SetParameter("aggregation: output type",
+                                  Teuchos::ParameterEntry(std::string("Graph")));
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(interpolationOrder));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+  StructuredAggFact->SetParameter("aggregation: single coarse point",
+                                  Teuchos::ParameterEntry(true));
+
+  fineLevel.Request("prolongatorGraph", StructuredAggFact.get());
+  fineLevel.Request(*StructuredAggFact);
+  StructuredAggFact->Build(fineLevel);
+
+  RCP<CrsGraph> prolongatorGraph;
+  fineLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
+
+  TEST_EQUALITY(prolongatorGraph != Teuchos::null, true);
+
+  int numErrors      = 0;
+  const int numRanks = Coordinates->getMap()->getComm()->getSize();
+  TEST_EQUALITY(prolongatorGraph->getGlobalNumRows() == 49, true);
+  TEST_EQUALITY(prolongatorGraph->getGlobalNumEntries() == 49, true);
+  if (numRanks == 1) {
+    TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 9, true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == 49, true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == 9, true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == 49, true);
+
+    ArrayView<const LO> rowIndices;
+    const LO indicesConstant[49] = {0, 0, 1, 1, 1, 2, 2,
+                                    0, 0, 1, 1, 1, 2, 2,
+                                    3, 3, 4, 4, 4, 5, 5,
+                                    3, 3, 4, 4, 4, 5, 5,
+                                    3, 3, 4, 4, 4, 5, 5,
+                                    6, 6, 7, 7, 7, 8, 8,
+                                    6, 6, 7, 7, 7, 8, 8};
+    for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+      prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+      if (rowIndices[0] != indicesConstant[rowIdx]) {
+        ++numErrors;
+      }
+    }
+  } else if (numRanks == 4) {
+    const int myRank                = Coordinates->getMap()->getComm()->getRank();
+    const Array<size_t> nodeNumRows = {{16, 12, 12, 9}};
+    const Array<size_t> nodeNumCols = {{4, 2, 2, 1}};
+
+    TEST_EQUALITY(prolongatorGraph->getGlobalNumCols() == 9, true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumRows() == nodeNumRows[myRank], true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumCols() == nodeNumCols[myRank], true);
+    TEST_EQUALITY(prolongatorGraph->getLocalNumEntries() == nodeNumRows[myRank], true);
+
+    ArrayView<const LO> rowIndices;
+    if (myRank == 0) {
+      const LO indicesConstant[16] = {0, 0, 1, 1,
+                                      0, 0, 1, 1,
+                                      2, 2, 3, 3,
+                                      2, 2, 3, 3};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+        if (rowIndices[0] != indicesConstant[rowIdx]) {
+          ++numErrors;
         }
-      } else if(myRank == 1) {
-        const LO indicesConstant[12] = {0, 0, 0,
-                                        0, 0, 0,
-                                        1, 1, 1,
-                                        1, 1, 1};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-          if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
+      }
+    } else if (myRank == 1) {
+      const LO indicesConstant[12] = {0, 0, 0,
+                                      0, 0, 0,
+                                      1, 1, 1,
+                                      1, 1, 1};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+        if (rowIndices[0] != indicesConstant[rowIdx]) {
+          ++numErrors;
         }
-      } else if(myRank == 2) {
-        const LO indicesConstant[12] = {0, 0, 1, 1,
-                                        0, 0, 1, 1,
-                                        0, 0, 1, 1};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-          if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
+      }
+    } else if (myRank == 2) {
+      const LO indicesConstant[12] = {0, 0, 1, 1,
+                                      0, 0, 1, 1,
+                                      0, 0, 1, 1};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+        if (rowIndices[0] != indicesConstant[rowIdx]) {
+          ++numErrors;
         }
-      } else if(myRank == 3) {
-        const LO indicesConstant[ 9] = {0, 0, 0,
-                                        0, 0, 0,
-                                        0, 0, 0};
-        for(size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
-          prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
-          if(rowIndices[0] != indicesConstant[rowIdx]) {++numErrors;}
+      }
+    } else if (myRank == 3) {
+      const LO indicesConstant[9] = {0, 0, 0,
+                                     0, 0, 0,
+                                     0, 0, 0};
+      for (size_t rowIdx = 0; rowIdx < prolongatorGraph->getLocalNumRows(); ++rowIdx) {
+        prolongatorGraph->getLocalRowView(rowIdx, rowIndices);
+        if (rowIndices[0] != indicesConstant[rowIdx]) {
+          ++numErrors;
         }
       }
     }
-    TEST_EQUALITY(numErrors == 0, true);
-
-  } // UncoupledGraphSingleCoarseNode
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,GlobalLexiTentative1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,GlobalLexiTentative2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,GlobalLexiTentative3D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,LocalLexiTentative1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,LocalLexiTentative2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,LocalLexiTentative3D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledLocalLexiTentative1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledLocalLexiTentative2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledLocalLexiTentative3D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledMultilevelScalar,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,ProlongatorGraphUncoupled,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledAggSingleCoarseNode,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation,UncoupledGraphSingleCoarseNode,Scalar,LO,GO,Node)
+  }
+  TEST_EQUALITY(numErrors == 0, true);
+
+}  // UncoupledGraphSingleCoarseNode
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, GlobalLexiTentative1D, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, GlobalLexiTentative2D, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, GlobalLexiTentative3D, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, LocalLexiTentative1D, Scalar, LO, GO, Node)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, LocalLexiTentative2D, Scalar, LO, GO, Node)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, LocalLexiTentative3D, Scalar, LO, GO, Node)          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledLocalLexiTentative1D, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledLocalLexiTentative2D, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledLocalLexiTentative3D, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledMultilevelScalar, Scalar, LO, GO, Node)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, ProlongatorGraphUncoupled, Scalar, LO, GO, Node)     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledAggSingleCoarseNode, Scalar, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation, UncoupledGraphSingleCoarseNode, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/StructuredLineDetection.cpp b/packages/muelu/test/unit_tests/StructuredLineDetection.cpp
index b64c41f91885..c84392ce998d 100644
--- a/packages/muelu/test/unit_tests/StructuredLineDetection.cpp
+++ b/packages/muelu/test/unit_tests/StructuredLineDetection.cpp
@@ -62,295 +62,296 @@
 
 namespace MueLuTests {
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void GetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib, const LocalOrdinal numDimensions,
-                      RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
-                      RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& Coordinates,
-                      RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
-                      Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void GetProblemData(RCP<const Teuchos::Comm<int> >& comm, const Xpetra::UnderlyingLib lib, const LocalOrdinal numDimensions,
+                    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& Op,
+                    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> >& Coordinates,
+                    RCP<Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> >& map,
+                    Array<GlobalOrdinal>& gNodesPerDim, Array<LocalOrdinal>& lNodesPerDim) {
 #include "MueLu_UseShortNames.hpp"
 
-    GO nx = 5;
-    GO ny = 5;
-    GO nz = (numDimensions == 2 ? 1 : 5);
-    GO gNumPoints = nx*ny*nz;
-
-    gNodesPerDim[0] = nx;
-    gNodesPerDim[1] = ny;
-    gNodesPerDim[2] = nz;
-
-    GO myOffset = 0;
-    if(comm->getSize() == 1) {
-      myOffset = 0;
-      lNodesPerDim[0] = nx;
-      lNodesPerDim[1] = ny;
-      lNodesPerDim[2] = nz;
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(numDimensions == 2) {
-          myOffset = 0;
-          lNodesPerDim[0] = 3;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 0;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 3;
-        }
-      } else if(comm->getRank() == 1) {
-        if(numDimensions == 2) {
-          myOffset = 4;
-          lNodesPerDim[0] = 2;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 15;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 2;
-          lNodesPerDim[2] = 3;
-        }
-      } else if(comm->getRank() == 2) {
-        if(numDimensions == 2) {
-          myOffset = 21;
-          lNodesPerDim[0] = 3;
-          lNodesPerDim[1] = 2;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 75;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 3;
-          lNodesPerDim[2] = 2;
-        }
-      } else if(comm->getRank() == 3) {
-        if(numDimensions == 2) {
-          myOffset = 25;
-          lNodesPerDim[0] = 2;
-          lNodesPerDim[1] = 2;
-          lNodesPerDim[2] = 1;
-        } else if(numDimensions == 3) {
-          myOffset = 90;
-          lNodesPerDim[0] = nx;
-          lNodesPerDim[1] = 2;
-          lNodesPerDim[2] = 2;
-        }
+  GO nx         = 5;
+  GO ny         = 5;
+  GO nz         = (numDimensions == 2 ? 1 : 5);
+  GO gNumPoints = nx * ny * nz;
+
+  gNodesPerDim[0] = nx;
+  gNodesPerDim[1] = ny;
+  gNodesPerDim[2] = nz;
+
+  GO myOffset = 0;
+  if (comm->getSize() == 1) {
+    myOffset        = 0;
+    lNodesPerDim[0] = nx;
+    lNodesPerDim[1] = ny;
+    lNodesPerDim[2] = nz;
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (numDimensions == 2) {
+        myOffset        = 0;
+        lNodesPerDim[0] = 3;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 0;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 3;
       }
-    }
-
-    GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
-    if(numDimensions == 2) {
-      myZoffset = 0;
-      myYoffset = myOffset / gNodesPerDim[0];
-      myXoffset = myOffset % gNodesPerDim[0];
-    } else if(numDimensions == 3) {
-      myZoffset = myOffset / (gNodesPerDim[1]*gNodesPerDim[0]);
-      myYoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) / gNodesPerDim[0];
-      myXoffset = (myOffset - myZoffset*gNodesPerDim[1]*gNodesPerDim[0]) % gNodesPerDim[0];
-    }
-    LO lNumPoints = lNodesPerDim[0]*lNodesPerDim[1]*lNodesPerDim[2];
-
-    // Construct map and local coordinates
-    Teuchos::Array<GO>     myGIDs(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
-    Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
-    for(LO k = 0; k < lNodesPerDim[2]; ++k) {
-      for(LO j = 0; j < lNodesPerDim[1]; ++j) {
-        for(LO i = 0; i < lNodesPerDim[0]; ++i) {
-          myGIDs[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = myOffset + k*gNodesPerDim[1]*gNodesPerDim[0] + j*gNodesPerDim[0] + i;
-          myXCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = (i + myXoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
-          myYCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = (j + myYoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
-          myZCoords[k*lNodesPerDim[1]*lNodesPerDim[0] + j*lNodesPerDim[0] + i] = (k + myZoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
-        }
+    } else if (comm->getRank() == 1) {
+      if (numDimensions == 2) {
+        myOffset        = 4;
+        lNodesPerDim[0] = 2;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 15;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 2;
+        lNodesPerDim[2] = 3;
+      }
+    } else if (comm->getRank() == 2) {
+      if (numDimensions == 2) {
+        myOffset        = 21;
+        lNodesPerDim[0] = 3;
+        lNodesPerDim[1] = 2;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 75;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 3;
+        lNodesPerDim[2] = 2;
+      }
+    } else if (comm->getRank() == 3) {
+      if (numDimensions == 2) {
+        myOffset        = 25;
+        lNodesPerDim[0] = 2;
+        lNodesPerDim[1] = 2;
+        lNodesPerDim[2] = 1;
+      } else if (numDimensions == 3) {
+        myOffset        = 90;
+        lNodesPerDim[0] = nx;
+        lNodesPerDim[1] = 2;
+        lNodesPerDim[2] = 2;
       }
     }
+  }
 
-    Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
-    if(numDimensions == 1) {
-      myCoordinates[0] = myXCoords();
-    } else if(numDimensions == 2) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-    } else if(numDimensions == 3) {
-      myCoordinates[0] = myXCoords();
-      myCoordinates[1] = myYCoords();
-      myCoordinates[2] = myZCoords();
+  GO myZoffset = 0, myYoffset = 0, myXoffset = 0;
+  if (numDimensions == 2) {
+    myZoffset = 0;
+    myYoffset = myOffset / gNodesPerDim[0];
+    myXoffset = myOffset % gNodesPerDim[0];
+  } else if (numDimensions == 3) {
+    myZoffset = myOffset / (gNodesPerDim[1] * gNodesPerDim[0]);
+    myYoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) / gNodesPerDim[0];
+    myXoffset = (myOffset - myZoffset * gNodesPerDim[1] * gNodesPerDim[0]) % gNodesPerDim[0];
+  }
+  LO lNumPoints = lNodesPerDim[0] * lNodesPerDim[1] * lNodesPerDim[2];
+
+  // Construct map and local coordinates
+  Teuchos::Array<GO> myGIDs(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myXCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myYCoords(lNumPoints);
+  Teuchos::Array<typename Teuchos::ScalarTraits<Scalar>::magnitudeType> myZCoords(lNumPoints);
+  for (LO k = 0; k < lNodesPerDim[2]; ++k) {
+    for (LO j = 0; j < lNodesPerDim[1]; ++j) {
+      for (LO i = 0; i < lNodesPerDim[0]; ++i) {
+        myGIDs[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i]    = myOffset + k * gNodesPerDim[1] * gNodesPerDim[0] + j * gNodesPerDim[0] + i;
+        myXCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = (i + myXoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[0] - 1);
+        myYCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = (j + myYoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[1] - 1);
+        myZCoords[k * lNodesPerDim[1] * lNodesPerDim[0] + j * lNodesPerDim[0] + i] = (k + myZoffset) / Teuchos::as<typename Teuchos::ScalarTraits<Scalar>::magnitudeType>(gNodesPerDim[2] - 1);
+      }
     }
+  }
 
-    // Create the map and store coordinates using the above array views
-    map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
-    Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO>::Build(map, myCoordinates(), numDimensions);
-
-    // small parameter list for Galeri
-    Teuchos::ParameterList problemParamList;
-    if(numDimensions == 1) {
-      problemParamList.set("nx",gNodesPerDim[0]);
-    } else if(numDimensions == 2) {
-      problemParamList.set("nx",gNodesPerDim[0]);
-      problemParamList.set("ny",gNodesPerDim[1]);
-    } else if(numDimensions == 3) {
-      problemParamList.set("nx",gNodesPerDim[0]);
-      problemParamList.set("ny",gNodesPerDim[1]);
-      problemParamList.set("nz",gNodesPerDim[2]);
-    }
-    // problemParamList.set("keepBCs", true);
-
-    // create Poisson problem and matrix
-    if(numDimensions == 1) {
-      Galeri::Xpetra::Laplace1DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemParamList, map);
-      Op = Problem.BuildMatrix();
-    } else if(numDimensions == 2) {
-      Galeri::Xpetra::Laplace2DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemParamList, map);
-      Op = Problem.BuildMatrix();
-    } else if(numDimensions == 3) {
-      Galeri::Xpetra::Laplace3DProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector> Problem(problemParamList, map);
-      Op = Problem.BuildMatrix();
-    }
+  Teuchos::Array<Teuchos::ArrayView<const typename Teuchos::ScalarTraits<Scalar>::magnitudeType> > myCoordinates(numDimensions);
+  if (numDimensions == 1) {
+    myCoordinates[0] = myXCoords();
+  } else if (numDimensions == 2) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+  } else if (numDimensions == 3) {
+    myCoordinates[0] = myXCoords();
+    myCoordinates[1] = myYCoords();
+    myCoordinates[2] = myZCoords();
+  }
 
+  // Create the map and store coordinates using the above array views
+  map         = MapFactory::Build(lib, gNumPoints, myGIDs(), 0, comm);
+  Coordinates = Xpetra::MultiVectorFactory<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO>::Build(map, myCoordinates(), numDimensions);
+
+  // small parameter list for Galeri
+  Teuchos::ParameterList problemParamList;
+  if (numDimensions == 1) {
+    problemParamList.set("nx", gNodesPerDim[0]);
+  } else if (numDimensions == 2) {
+    problemParamList.set("nx", gNodesPerDim[0]);
+    problemParamList.set("ny", gNodesPerDim[1]);
+  } else if (numDimensions == 3) {
+    problemParamList.set("nx", gNodesPerDim[0]);
+    problemParamList.set("ny", gNodesPerDim[1]);
+    problemParamList.set("nz", gNodesPerDim[2]);
+  }
+  // problemParamList.set("keepBCs", true);
+
+  // create Poisson problem and matrix
+  if (numDimensions == 1) {
+    Galeri::Xpetra::Laplace1DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemParamList, map);
+    Op = Problem.BuildMatrix();
+  } else if (numDimensions == 2) {
+    Galeri::Xpetra::Laplace2DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemParamList, map);
+    Op = Problem.BuildMatrix();
+  } else if (numDimensions == 3) {
+    Galeri::Xpetra::Laplace3DProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector> Problem(problemParamList, map);
+    Op = Problem.BuildMatrix();
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredLineDetectionFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredLineDetectionFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<StructuredLineDetectionFactory> lineDetectionFact = rcp(new StructuredLineDetectionFactory);
-    TEST_EQUALITY(lineDetectionFact != Teuchos::null, true);
+  RCP<StructuredLineDetectionFactory> lineDetectionFact = rcp(new StructuredLineDetectionFactory);
+  TEST_EQUALITY(lineDetectionFact != Teuchos::null, true);
 
-  } // Constructor
+}  // Constructor
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredLineDetectionFactory, LabelLines, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredLineDetectionFactory, LabelLines, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
 
-    out << "version: " << MueLu::Version() << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-    // used Xpetra lib (for maps and smoothers)
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  // used Xpetra lib (for maps and smoothers)
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
-    LO numDimensions = 3;
+  LO numDimensions = 3;
 
-    LO maxLevels = 2, maxIter = 10;
-    RCP<Matrix> Op;
-    RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LO,GO,NO> > coordinates;
-    RCP<Xpetra::Map<LO,GO,NO> > map;
-    Array<GO> gNodesPerDim(3);
-    Array<LO> lNodesPerDim(3);
-    GetProblemData<SC,LO,GO,NO>(comm, lib, numDimensions, Op, coordinates, map, gNodesPerDim, lNodesPerDim);
+  LO maxLevels = 2, maxIter = 10;
+  RCP<Matrix> Op;
+  RCP<Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LO, GO, NO> > coordinates;
+  RCP<Xpetra::Map<LO, GO, NO> > map;
+  Array<GO> gNodesPerDim(3);
+  Array<LO> lNodesPerDim(3);
+  GetProblemData<SC, LO, GO, NO>(comm, lib, numDimensions, Op, coordinates, map, gNodesPerDim, lNodesPerDim);
 
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-    Teuchos::Array<magnitude_type> norms(1);
-    nullSpace->norm1(norms);
-    if (comm->getRank() == 0) {
-      out << "||NS|| = " << norms[0] << std::endl;
-    }
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+  Teuchos::Array<magnitude_type> norms(1);
+  nullSpace->norm1(norms);
+  if (comm->getRank() == 0) {
+    out << "||NS|| = " << norms[0] << std::endl;
+  }
 
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-
-    // create the factory manager and the factories
-    FactoryManager M;
-
-    RCP<Factory>      Pfact  = rcp( new GeneralGeometricPFactory() );
-    RCP<Factory>      Rfact  = rcp( new TransPFactory() );
-    RCP<Factory>      LDfact = rcp( new StructuredLineDetectionFactory() );
-    RCP<Factory>      Tfact  = rcp( new CoordinatesTransferFactory() );
-    RCP<RAPFactory>   Acfact = rcp( new RAPFactory() );
-    RCP<Factory>      NSfact = rcp( new NullspaceFactory() );
-
-    // Set paramters needed by the factories
-    Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
-    Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
-
-    LDfact->SetParameter("orientation", Teuchos::ParameterEntry(std::string("X")));
-
-    Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
-
-    // Set interfactory dependencies
-    NSfact->SetFactory("Nullspace", Pfact);
-
-    Acfact->AddTransferFactory(Tfact);
-
-    // Set default factories in the manager
-    M.SetFactory("P",           Pfact);
-    M.SetFactory("R",           Rfact);
-    M.SetFactory("A",           Acfact);
-    M.SetFactory("Nullspace",   NSfact);
-    M.SetFactory("gNodesPerDim", Tfact);
-    M.SetFactory("lNodesPerDim", Tfact);
-    M.SetFactory("Coordinates",  Tfact);
-    M.SetFactory("coarseCoordinates",  Pfact);
-    M.SetFactory("gCoarseNodesPerDim", Pfact);
-    M.SetFactory("lCoarseNodesPerDim", Pfact);
-    M.SetFactory("CoarseNumZLayers",          LDfact);
-    M.SetFactory("LineDetection_VertLineIds", LDfact);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Jacobi");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("LINESMOOTHING_BANDEDRELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-    // Set smoothers and coarse solver in the manager
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", coarseSolveFact);
-
-    // Populate data on the finest level of the hierarchy
-    RCP<Level> Finest = H->GetLevel();
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-    Finest->Set("Coordinates", coordinates);  // set fine level coordinates
-    Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
-    Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
-
-    // Setup the hierarchy
-    H->SetMaxCoarseSize(10);
-    H->Setup(M, 0, maxLevels);
-
-    out << "MueLu setup done!" << std::endl;
-
-    // Define RHS and solution vector
-    RCP<MultiVector> X   = MultiVectorFactory::Build(map,1);
-    RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
-
-    X->putScalar(1.0);
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+
+  // create the factory manager and the factories
+  FactoryManager M;
+
+  RCP<Factory> Pfact     = rcp(new GeneralGeometricPFactory());
+  RCP<Factory> Rfact     = rcp(new TransPFactory());
+  RCP<Factory> LDfact    = rcp(new StructuredLineDetectionFactory());
+  RCP<Factory> Tfact     = rcp(new CoordinatesTransferFactory());
+  RCP<RAPFactory> Acfact = rcp(new RAPFactory());
+  RCP<Factory> NSfact    = rcp(new NullspaceFactory());
+
+  // Set paramters needed by the factories
+  Pfact->SetParameter("Coarsen", Teuchos::ParameterEntry(std::string("{2,2,2}")));
+  Pfact->SetParameter("order", Teuchos::ParameterEntry(1));
+
+  LDfact->SetParameter("orientation", Teuchos::ParameterEntry(std::string("X")));
+
+  Tfact->SetParameter("Geometric", Teuchos::ParameterEntry(true));
+
+  // Set interfactory dependencies
+  NSfact->SetFactory("Nullspace", Pfact);
+
+  Acfact->AddTransferFactory(Tfact);
+
+  // Set default factories in the manager
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Nullspace", NSfact);
+  M.SetFactory("gNodesPerDim", Tfact);
+  M.SetFactory("lNodesPerDim", Tfact);
+  M.SetFactory("Coordinates", Tfact);
+  M.SetFactory("coarseCoordinates", Pfact);
+  M.SetFactory("gCoarseNodesPerDim", Pfact);
+  M.SetFactory("lCoarseNodesPerDim", Pfact);
+  M.SetFactory("CoarseNumZLayers", LDfact);
+  M.SetFactory("LineDetection_VertLineIds", LDfact);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Jacobi");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", Teuchos::ScalarTraits<Scalar>::one());
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("LINESMOOTHING_BANDEDRELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+  // Set smoothers and coarse solver in the manager
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", coarseSolveFact);
+
+  // Populate data on the finest level of the hierarchy
+  RCP<Level> Finest = H->GetLevel();
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                       // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);        // set null space information for finest level
+  Finest->Set("Coordinates", coordinates);    // set fine level coordinates
+  Finest->Set("gNodesPerDim", gNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->Set("lNodesPerDim", lNodesPerDim);  // set GeneralGeometricPFactory specific info
+  Finest->SetFactoryManager(Teuchos::rcpFromRef(M));
+
+  // Setup the hierarchy
+  H->SetMaxCoarseSize(10);
+  H->Setup(M, 0, maxLevels);
+
+  out << "MueLu setup done!" << std::endl;
+
+  // Define RHS and solution vector
+  RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+  RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+  X->putScalar(1.0);
+  X->norm2(norms);
+  Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  X->putScalar((Scalar)0.0);
+  out << "Starting solver iterations" << std::endl;
+  for (LO i = 0; i < 10; ++i) {
+    H->Iterate(*RHS, *X, maxIter);
     X->norm2(norms);
-    Op->apply(*X,*RHS,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    X->putScalar( (Scalar) 0.0);
-    out << "Starting solver iterations" << std::endl;
-    for(LO i = 0; i < 10; ++i) {
-      H->Iterate(*RHS,*X,maxIter);
-      X->norm2(norms);
-      if (comm->getRank() == 0) {out << "||RHS[" << i << "]|| = " << norms << std::endl;}
-      Op->apply(*X,*RHS,Teuchos::NO_TRANS,(Scalar)-1.0,(Scalar)1.0);
-      if(norms[0] < 1.0e-7) { break; }
+    if (comm->getRank() == 0) {
+      out << "||RHS[" << i << "]|| = " << norms << std::endl;
     }
-    TEST_EQUALITY( (norms[0] < 1.0e-7), true);
-  } // LabelLines
+    Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)-1.0, (Scalar)1.0);
+    if (norms[0] < 1.0e-7) {
+      break;
+    }
+  }
+  TEST_EQUALITY((norms[0] < 1.0e-7), true);
+}  // LabelLines
 
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredLineDetectionFactory,Constructor,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredLineDetectionFactory,LabelLines,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredLineDetectionFactory, Constructor, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredLineDetectionFactory, LabelLines, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/SubBlockAFactory.cpp b/packages/muelu/test/unit_tests/SubBlockAFactory.cpp
index 6fdaed7dd326..8d666d407588 100644
--- a/packages/muelu/test/unit_tests/SubBlockAFactory.cpp
+++ b/packages/muelu/test/unit_tests/SubBlockAFactory.cpp
@@ -60,183 +60,172 @@
 
 namespace MueLuTests {
 
-  /////////////////////////
-  // helper function
-
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar,LocalOrdinal, GlobalOrdinal, Node> >
-  GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map,
-                        Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
-#   include "MueLu_UseShortNames.hpp"
-
-    Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(map, 3);
-
-    LocalOrdinal NumMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
-    GlobalOrdinal NumGlobalElements = map->getGlobalNumElements();
-    GlobalOrdinal nIndexBase = map->getIndexBase();
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-    {
-      if (MyGlobalElements[i] == nIndexBase)
-      {
-        // off-diagonal for first row
-        Indices[0] = nIndexBase;
-        NumEntries = 1;
-        Values[0] = c;
-      }
-      else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1)
-      {
-        // off-diagonal for last row
-        Indices[0] = nIndexBase + NumGlobalElements - 2;
-        NumEntries = 1;
-        Values[0] = b;
-      }
-      else
-      {
-        // off-diagonal for internal row
-        Indices[0] = MyGlobalElements[i] - 1;
-        Values[1] = b;
-        Indices[1] = MyGlobalElements[i] + 1;
-        Values[0] = c;
-        NumEntries = 2;
-      }
-
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(MyGlobalElements[i],
-          Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-          Teuchos::tuple<Scalar>(a) );
-
+/////////////////////////
+// helper function
+
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node> >
+GenerateProblemMatrix(const Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > map,
+                      Scalar a = 2.0, Scalar b = -1.0, Scalar c = -1.0) {
+#include "MueLu_UseShortNames.hpp"
+
+  Teuchos::RCP<CrsMatrixWrap> mtx = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(map, 3);
+
+  LocalOrdinal NumMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
+  GlobalOrdinal NumGlobalElements                          = map->getGlobalNumElements();
+  GlobalOrdinal nIndexBase                                 = map->getIndexBase();
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  for (LocalOrdinal i = 0; i < NumMyElements; ++i) {
+    if (MyGlobalElements[i] == nIndexBase) {
+      // off-diagonal for first row
+      Indices[0] = nIndexBase;
+      NumEntries = 1;
+      Values[0]  = c;
+    } else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1) {
+      // off-diagonal for last row
+      Indices[0] = nIndexBase + NumGlobalElements - 2;
+      NumEntries = 1;
+      Values[0]  = b;
+    } else {
+      // off-diagonal for internal row
+      Indices[0] = MyGlobalElements[i] - 1;
+      Values[1]  = b;
+      Indices[1] = MyGlobalElements[i] + 1;
+      Values[0]  = c;
+      NumEntries = 2;
     }
 
-    mtx->fillComplete(map,map);
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
 
-    return mtx;
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(MyGlobalElements[i],
+                            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                            Teuchos::tuple<Scalar>(a));
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SubBlockAFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<SubBlockAFactory> subBlockAFactory = rcp(new SubBlockAFactory());
-    TEST_EQUALITY(subBlockAFactory != Teuchos::null, true);
-  } // Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SubBlockAFactory, ExtractMainDiagBlocks, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    /**********************************************************************************/
-    /* CREATE INITIAL MATRIX                                                          */
-    /**********************************************************************************/
-    const GO numElements = 500;
-    const GO numElements1 = 400;
-    const GO numElements2 = 100;
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    std::vector<size_t> stridingInfo;
-    stridingInfo.push_back(1);
-
-    RCP<const Map> map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
-    RCP<const Map> map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
-
-    std::vector<GlobalOrdinal> localGids; // vector with all local GIDs on cur proc
-    Teuchos::ArrayView< const GlobalOrdinal > map1eleList = map1->getLocalElementList(); // append all local gids from map1 and map2
-    localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
-    Teuchos::ArrayView< const GlobalOrdinal > map2eleList = map2->getLocalElementList();
-    localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
-    Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0],localGids.size());
-    RCP<const Map> bigMap = MapFactory::Build(lib, numElements, eleList, 0, comm); // create full big map (concatenation of map1 and map2)
-
-    std::vector<Teuchos::RCP<const Map> > maps;
-    maps.push_back(map1);
-    maps.push_back(map2);
-
-    RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(bigMap, maps);
-
-    RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, 2, -1, -1);
-    RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, 3, -2, -1);
-
-    // build blocked operator
-    RCP<BlockedCrsMatrix> bOp = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
-    bOp->setMatrix(0, 0, Op11);
-    bOp->setMatrix(1, 1, Op22);
-    bOp->fillComplete();
-    TEST_EQUALITY(bOp!=Teuchos::null, true);
-
-    // build hierarchy
-    RCP<Level> levelOne = rcp(new Level());
-    levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp)); // set blocked operator
-
-    // define sub block factories for blocked operator "A"
-    RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
-    A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
-    A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
-    A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
-    RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
-    A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
-    A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
-    A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
-
-    // Test the request mechanism
-    levelOne->Request("A", A11Fact.get(), MueLu::NoFactory::get());
-    levelOne->Request("A", A22Fact.get(), MueLu::NoFactory::get());
-    TEST_EQUALITY(levelOne->IsRequested("A", A11Fact.get()),true);
-    TEST_EQUALITY(levelOne->IsRequested("A", A22Fact.get()),true);
-
-    // Test availability of subblocks of A
-    RCP<Matrix> A11 = levelOne->Get<RCP<Matrix> >("A", A11Fact.get());
-    RCP<Matrix> A22 = levelOne->Get<RCP<Matrix> >("A", A22Fact.get());
-    TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()), true);
-    TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()), true);
-
-    // Test the release mechanism
-    levelOne->Release("A", A11Fact.get());
-    levelOne->Release("A", A22Fact.get());
-    TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()), false);
-    TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()), false);
-
-    // A11 is supposed to match Op11
-    TEST_EQUALITY(A11->getRowMap()->isSameAs(*(Op11->getRowMap())), true);
-    TEST_EQUALITY(A11->getColMap()->isSameAs(*(Op11->getColMap())), true);
-    TEST_EQUALITY(A11->getRangeMap()->isSameAs(*(Op11->getRangeMap())), true);
-    TEST_EQUALITY(A11->getDomainMap()->isSameAs(*(Op11->getDomainMap())), true);
-    TEST_EQUALITY(A11->getLocalNumEntries(),Op11->getLocalNumEntries());
-
-    // A22 is supposed to match Op22
-    TEST_EQUALITY(A22->getRowMap()->isSameAs(*(Op22->getRowMap())), true);
-    TEST_EQUALITY(A22->getColMap()->isSameAs(*(Op22->getColMap())), true);
-    TEST_EQUALITY(A22->getRangeMap()->isSameAs(*(Op22->getRangeMap())), true);
-    TEST_EQUALITY(A22->getDomainMap()->isSameAs(*(Op22->getDomainMap())), true);
-    TEST_EQUALITY(A22->getLocalNumEntries(), Op22->getLocalNumEntries());
-  } // ExtractMainDiagBlocks
-
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SubBlockAFactory, Constructor, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SubBlockAFactory, ExtractMainDiagBlocks, SC, LO, GO, Node)
+  mtx->fillComplete(map, map);
 
-#include <MueLu_ETI_4arg.hpp>
+  return mtx;
 }
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SubBlockAFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<SubBlockAFactory> subBlockAFactory = rcp(new SubBlockAFactory());
+  TEST_EQUALITY(subBlockAFactory != Teuchos::null, true);
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SubBlockAFactory, ExtractMainDiagBlocks, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  /**********************************************************************************/
+  /* CREATE INITIAL MATRIX                                                          */
+  /**********************************************************************************/
+  const GO numElements  = 500;
+  const GO numElements1 = 400;
+  const GO numElements2 = 100;
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  std::vector<size_t> stridingInfo;
+  stridingInfo.push_back(1);
+
+  RCP<const Map> map1 = StridedMapFactory::Build(lib, numElements1, 0, stridingInfo, comm);
+  RCP<const Map> map2 = StridedMapFactory::Build(lib, numElements2, numElements1, stridingInfo, comm);
+
+  std::vector<GlobalOrdinal> localGids;                                               // vector with all local GIDs on cur proc
+  Teuchos::ArrayView<const GlobalOrdinal> map1eleList = map1->getLocalElementList();  // append all local gids from map1 and map2
+  localGids.insert(localGids.end(), map1eleList.begin(), map1eleList.end());
+  Teuchos::ArrayView<const GlobalOrdinal> map2eleList = map2->getLocalElementList();
+  localGids.insert(localGids.end(), map2eleList.begin(), map2eleList.end());
+  Teuchos::ArrayView<GlobalOrdinal> eleList(&localGids[0], localGids.size());
+  RCP<const Map> bigMap = MapFactory::Build(lib, numElements, eleList, 0, comm);  // create full big map (concatenation of map1 and map2)
+
+  std::vector<Teuchos::RCP<const Map> > maps;
+  maps.push_back(map1);
+  maps.push_back(map2);
+
+  RCP<const MapExtractor> mapExtractor = MapExtractorFactory::Build(bigMap, maps);
+
+  RCP<CrsMatrixWrap> Op11 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map1, 2, -1, -1);
+  RCP<CrsMatrixWrap> Op22 = GenerateProblemMatrix<Scalar, LO, GO, Node>(map2, 3, -2, -1);
+
+  // build blocked operator
+  RCP<BlockedCrsMatrix> bOp = rcp(new BlockedCrsMatrix(mapExtractor, mapExtractor, 10));
+  bOp->setMatrix(0, 0, Op11);
+  bOp->setMatrix(1, 1, Op22);
+  bOp->fillComplete();
+  TEST_EQUALITY(bOp != Teuchos::null, true);
+
+  // build hierarchy
+  RCP<Level> levelOne = rcp(new Level());
+  levelOne->Set("A", Teuchos::rcp_dynamic_cast<Matrix>(bOp));  // set blocked operator
+
+  // define sub block factories for blocked operator "A"
+  RCP<SubBlockAFactory> A11Fact = Teuchos::rcp(new SubBlockAFactory());
+  A11Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A11Fact->SetParameter("block row", Teuchos::ParameterEntry(0));
+  A11Fact->SetParameter("block col", Teuchos::ParameterEntry(0));
+  RCP<SubBlockAFactory> A22Fact = Teuchos::rcp(new SubBlockAFactory());
+  A22Fact->SetFactory("A", MueLu::NoFactory::getRCP());
+  A22Fact->SetParameter("block row", Teuchos::ParameterEntry(1));
+  A22Fact->SetParameter("block col", Teuchos::ParameterEntry(1));
+
+  // Test the request mechanism
+  levelOne->Request("A", A11Fact.get(), MueLu::NoFactory::get());
+  levelOne->Request("A", A22Fact.get(), MueLu::NoFactory::get());
+  TEST_EQUALITY(levelOne->IsRequested("A", A11Fact.get()), true);
+  TEST_EQUALITY(levelOne->IsRequested("A", A22Fact.get()), true);
+
+  // Test availability of subblocks of A
+  RCP<Matrix> A11 = levelOne->Get<RCP<Matrix> >("A", A11Fact.get());
+  RCP<Matrix> A22 = levelOne->Get<RCP<Matrix> >("A", A22Fact.get());
+  TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()), true);
+  TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()), true);
+
+  // Test the release mechanism
+  levelOne->Release("A", A11Fact.get());
+  levelOne->Release("A", A22Fact.get());
+  TEST_EQUALITY(levelOne->IsAvailable("A", A11Fact.get()), false);
+  TEST_EQUALITY(levelOne->IsAvailable("A", A22Fact.get()), false);
+
+  // A11 is supposed to match Op11
+  TEST_EQUALITY(A11->getRowMap()->isSameAs(*(Op11->getRowMap())), true);
+  TEST_EQUALITY(A11->getColMap()->isSameAs(*(Op11->getColMap())), true);
+  TEST_EQUALITY(A11->getRangeMap()->isSameAs(*(Op11->getRangeMap())), true);
+  TEST_EQUALITY(A11->getDomainMap()->isSameAs(*(Op11->getDomainMap())), true);
+  TEST_EQUALITY(A11->getLocalNumEntries(), Op11->getLocalNumEntries());
+
+  // A22 is supposed to match Op22
+  TEST_EQUALITY(A22->getRowMap()->isSameAs(*(Op22->getRowMap())), true);
+  TEST_EQUALITY(A22->getColMap()->isSameAs(*(Op22->getColMap())), true);
+  TEST_EQUALITY(A22->getRangeMap()->isSameAs(*(Op22->getRangeMap())), true);
+  TEST_EQUALITY(A22->getDomainMap()->isSameAs(*(Op22->getDomainMap())), true);
+  TEST_EQUALITY(A22->getLocalNumEntries(), Op22->getLocalNumEntries());
+}  // ExtractMainDiagBlocks
+
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SubBlockAFactory, Constructor, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SubBlockAFactory, ExtractMainDiagBlocks, SC, LO, GO, Node)
 
+#include <MueLu_ETI_4arg.hpp>
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/TentativePFactory.cpp b/packages/muelu/test/unit_tests/TentativePFactory.cpp
index dec7ed42fc87..e4bcd82e7ae4 100644
--- a/packages/muelu/test/unit_tests/TentativePFactory.cpp
+++ b/packages/muelu/test/unit_tests/TentativePFactory.cpp
@@ -69,820 +69,801 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<TentativePFactory> tentPFact = rcp(new TentativePFactory);
-    TEST_EQUALITY(tentPFact != Teuchos::null, true);
-
-  } //Constructor
-
-  //TODO test BuildP
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentative_LapackQR, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC,LO,GO,NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR with user-supplied nullspace" << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    GO nx = 29;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    A->SetFixedBlockSize(1);
-    fineLevel.Set("A", A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates
-      = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    LocalOrdinal NSdim = 2;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-    fineLevel.Set("DofsPerNode",1);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetFactory("Graph", dropFact);
-
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
-
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());         // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get()); // request coarse nullspace
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
-
-    //check interpolation
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(),NSdim);
-    Ptent->apply(*coarseNullSpace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
-
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());   // release coarse nullspace
-
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LocalOrdinal i=0; i<NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_EQUALITY(norms[i] < 100*TMT::eps(), true);
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    Teuchos::ArrayRCP<const Scalar> col1 = coarseNullSpace->getData(0);
-    Teuchos::ArrayRCP<const Scalar> col2 = coarseNullSpace->getData(1);
-    TEST_EQUALITY(col1.size() == col2.size(), true);
-
-  } //MakeTentative  Lapack QR
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR with user-supplied nullspace" << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    GO nx = 199;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
-    fineLevel.Request("A");
-    fineLevel.Set("A",A);
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
-    fineLevel.Set("Coordinates", coordinates);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace",nullSpace);
-    fineLevel.Set("DofsPerNode",1);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetFactory("Graph", dropFact);
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    //check interpolation
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(),NSdim);
-    Ptent->apply(*coarseNullSpace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
-
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LocalOrdinal i=0; i<NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_EQUALITY(norms[i] < 100*TMT::eps(), true);
-    }
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Ptent,true,*Ptent,false,out);
-    Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_EQUALITY(diagVec->norm1(), diagVec->getGlobalLength());
-    TEST_EQUALITY(diagVec->normInf()-1 < 100*TMT::eps(), true);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_EQUALITY(diagVec->meanValue(), TST::one());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+  RCP<TentativePFactory> tentPFact = rcp(new TentativePFactory);
+  TEST_EQUALITY(tentPFact != Teuchos::null, true);
 
-  } //MakeTentative
+}  // Constructor
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentativeUsingDefaultNullSpace, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+// TODO test BuildP
 
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentative_LapackQR, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR when nullspace isn't supplied by user" << std::endl;
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR with user-supplied nullspace" << std::endl;
 
-    GO nx = 199;
-    RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
 
-    fineLevel.Set("A", A);
+  GO nx         = 29;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  A->SetFixedBlockSize(1);
+  fineLevel.Set("A", A);
 
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<real, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
 
-    fineLevel.Set("Coordinates", coordinates);
+  LocalOrdinal NSdim         = 2;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("DofsPerNode", 1);
 
-    RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    coarseLevel.Request("P",tentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace", tentativePFact.get());  // request coarse nullspace
-    coarseLevel.Request(*tentativePFact);
-    tentativePFact->Build(fineLevel,coarseLevel);
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetFactory("Graph", dropFact);
 
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,tentativePFact.get());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
 
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",tentativePFact.get());
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
 
-    coarseLevel.Release("P",tentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",tentativePFact.get());   // release coarse nullspace
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    //grab default fine level nullspace (vector of all ones)
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
-    nullSpace->putScalar(1.0);
+  coarseLevel.Request("P", TentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
 
-    //check interpolation
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    Ptent->apply(*coarseNullSpace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
 
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
 
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
+  // check interpolation
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullSpace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LocalOrdinal i=0; i<NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_EQUALITY(norms[i] < 100*TMT::eps(), true);
-    }
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-  } //MakeTentativeUsingDefaultNullSpace
+  coarseLevel.Release("P", TentativePFact.get());          // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());  // release coarse nullspace
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, NoQROption, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
 
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LocalOrdinal i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_EQUALITY(norms[i] < 100 * TMT::eps(), true);
+  }
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test option that skips local QR factorizations" << std::endl;
+  Teuchos::ArrayRCP<const Scalar> col1 = coarseNullSpace->getData(0);
+  Teuchos::ArrayRCP<const Scalar> col2 = coarseNullSpace->getData(1);
+  TEST_EQUALITY(col1.size() == col2.size(), true);
+
+}  // MakeTentative  Lapack QR
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR with user-supplied nullspace" << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  GO nx         = 199;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
+  fineLevel.Set("Coordinates", coordinates);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("DofsPerNode", 1);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetFactory("Graph", dropFact);
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", UncoupledAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", UncoupledAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  // check interpolation
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullSpace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LocalOrdinal i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_EQUALITY(norms[i] < 100 * TMT::eps(), true);
+  }
 
-    GlobalOrdinal nx = 20, ny = 20;
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_EQUALITY(diagVec->norm1(), diagVec->getGlobalLength());
+  TEST_EQUALITY(diagVec->normInf() - 1 < 100 * TMT::eps(), true);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_EQUALITY(diagVec->meanValue(), TST::one());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
 
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+}  // MakeTentative
 
-    map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, MakeTentativeUsingDefaultNullSpace, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
-    RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    fineLevel.Request("A");
-    fineLevel.Set("A",A);
-    fineLevel.Set("Coordinates", coordinates);
-
-    LocalOrdinal NSdim = 3;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-    fineLevel.Set("DofsPerNode",2);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<UncoupledAggregationFactory> UnCoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UnCoupledAggFact->SetFactory("Graph", dropFact);
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", UnCoupledAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", UnCoupledAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    Teuchos::ParameterList paramList;
-    paramList.set("tentative: calculate qr",       false);
-    TentativePFact->SetParameterList(paramList);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    //check interpolation
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(),NSdim);
-    Ptent->apply(*coarseNullSpace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
-
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
-
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());
-
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
-
-    Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LocalOrdinal i=0; i<NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_EQUALITY(norms[i] < 100*TMT::eps(), true);
-    }
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR when nullspace isn't supplied by user" << std::endl;
 
-  } //NoQR
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, ConstantColumnSum, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  GO nx         = 199;
+  RCP<Matrix> A = test_factory::Build1DPoisson(nx);
 
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real                  = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real,LO,GO,NO>;
-    using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+  fineLevel.Set("A", A);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test option that skips local QR factorizations & uses constant column sum" << std::endl;
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("1D", A->getRowMap(), galeriList);
 
-    GlobalOrdinal nx = 20, ny = 20;
+  fineLevel.Set("Coordinates", coordinates);
 
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+  RCP<TentativePFactory> tentativePFact = rcp(new TentativePFactory());
 
-    map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 1);
+  coarseLevel.Request("P", tentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", tentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*tentativePFact);
+  tentativePFact->Build(fineLevel, coarseLevel);
 
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, tentativePFact.get());
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    fineLevel.Request("A");
-    fineLevel.Set("A",A);
-
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
-    fineLevel.Set("DofsPerNode",1);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<UncoupledAggregationFactory> UnCoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UnCoupledAggFact->SetFactory("Graph", dropFact);
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", UnCoupledAggFact);
-    RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
-    TentativePFact->SetFactory("Aggregates", UnCoupledAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    Teuchos::ParameterList paramList;
-    paramList.set("tentative: calculate qr",       false);
-    paramList.set("tentative: constant column sums",true );
-    paramList.sublist("matrixmatrix: kernel params").set("compute global constants",true);
-    TentativePFact->SetParameterList(paramList);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
-
-    // This is "did it run test" not a "correctness check"
-
-  } //ConstColSum
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, NonStandardMaps, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //#warning Unit test PgPFactory NonStandardMaps disabled
-    //  return;
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-#   if !defined(HAVE_MUELU_IFPACK)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
-#   endif
-#   if !defined(HAVE_MUELU_IFPACK2)
-    MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
-#   endif
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // generate problem
-    LocalOrdinal maxLevels = 3;
-    //LocalOrdinal its=10;
-    GlobalOrdinal nEle = 63;
-    GlobalOrdinal nIndexBase = 10;
-    const RCP<const Map> map = MapFactory::Build(lib, nEle, nIndexBase, comm);
-    RCP<Matrix> mtx = Galeri::Xpetra::MatrixTraits<Map,CrsMatrixWrap>::Build(map, 3);
-
-    LocalOrdinal NumMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
-    GlobalOrdinal NumGlobalElements = map->getGlobalNumElements();
-    assert(NumGlobalElements == nEle);
-
-    GlobalOrdinal NumEntries;
-    LocalOrdinal nnz=2;
-    std::vector<Scalar> Values(nnz);
-    std::vector<GlobalOrdinal> Indices(nnz);
-
-    Scalar a = 2.0;
-    Scalar b = -1.0;
-    Scalar c = -1.0;
-
-    for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-    {
-      if (MyGlobalElements[i] == nIndexBase)
-      {
-        // off-diagonal for first row
-        Indices[0] = nIndexBase;
-        NumEntries = 1;
-        Values[0] = c;
-      }
-      else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1)
-      {
-        // off-diagonal for last row
-        Indices[0] = nIndexBase + NumGlobalElements - 2;
-        NumEntries = 1;
-        Values[0] = b;
-      }
-      else
-      {
-        // off-diagonal for internal row
-        Indices[0] = MyGlobalElements[i] - 1;
-        Values[1] = b;
-        Indices[1] = MyGlobalElements[i] + 1;
-        Values[0] = c;
-        NumEntries = 2;
-      }
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", tentativePFact.get());
+
+  coarseLevel.Release("P", tentativePFact.get());          // release Ptent
+  coarseLevel.Release("Nullspace", tentativePFact.get());  // release coarse nullspace
+
+  // grab default fine level nullspace (vector of all ones)
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar(1.0);
 
-      // put the off-diagonal entries
-      // Xpetra wants ArrayViews (sigh)
-      Teuchos::ArrayView<Scalar> av(&Values[0],NumEntries);
-      Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0],NumEntries);
-      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
-
-      // Put in the diagonal entry
-      mtx->insertGlobalValues(MyGlobalElements[i],
-          Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
-          Teuchos::tuple<Scalar>(a) );
-
-    } //for (LocalOrdinal i = 0; i < NumMyElements; ++i)
-
-
-    mtx->fillComplete(map,map);
-
-    std::cout << map->getIndexBase() << std::endl;
-
-    RCP<Matrix> Op = Teuchos::rcp_dynamic_cast<Matrix>(mtx);
-
-    // build nullspace
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-    nullSpace->putScalar( (Scalar) 1.0);
-
-    // fill hierarchy
-    RCP<Hierarchy> H = rcp( new Hierarchy() );
-    H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    RCP<Level> Finest = H->GetLevel(); // first associate level with hierarchy (for defaultFactoryHandler!)
-
-    Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-    Finest->Set("A",Op);                      // set fine level matrix
-    Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-
-    // define transfer operators
-    RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-    UncoupledAggFact->SetMinNodesPerAggregate(3);
-    UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-    UncoupledAggFact->SetOrdering("natural");
-
-    RCP<TentativePFactory> Pfact  = rcp( new TentativePFactory() );
-    RCP<Factory>           Rfact  = rcp( new TransPFactory() );
-    RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-    H->SetMaxCoarseSize(1);
-
-    // setup smoothers
-    Teuchos::ParameterList smootherParamList;
-    smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-    smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-    smootherParamList.set("relaxation: damping factor", (Scalar) 1.0);
-    RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-    RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-    Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-    FactoryManager M;
-    M.SetKokkosRefactor(false);
-    M.SetFactory("P", Pfact);
-    M.SetFactory("R", Rfact);
-    M.SetFactory("A", Acfact);
-    M.SetFactory("Ptent", Pfact);
-    M.SetFactory("Aggregates", UncoupledAggFact);
-    M.SetFactory("Smoother", SmooFact);
-    M.SetFactory("CoarseSolver", SmooFact);
-
-    H->Setup(M, 0, maxLevels);
-
-    RCP<Level> coarseLevel = H->GetLevel(1);
-    TEST_EQUALITY(coarseLevel->IsRequested("A",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("A",MueLu::NoFactory::get()), true);
-
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel->IsRequested("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsRequested("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel->IsAvailable("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("P",Pfact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel->GetKeepFlag("R",Rfact.get()), 0);
-    RCP<Level> coarseLevel2 = H->GetLevel(2);
-    TEST_EQUALITY(coarseLevel2->IsRequested("A",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("P",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("R",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother",MueLu::NoFactory::get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("A",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("P",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",MueLu::NoFactory::get()), false); // coarse level only has presmoother = coarse solver
-    TEST_EQUALITY(coarseLevel2->IsAvailable("R",MueLu::NoFactory::get()), true);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("A",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",MueLu::NoFactory::get()), MueLu::Final);
-    // TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final); // coarse level only has presmoother = coarse solver
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",MueLu::NoFactory::get()), MueLu::Final);
-    TEST_EQUALITY(coarseLevel2->IsRequested("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsRequested("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("P",Pfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother",SmooFact.get()), false);
-    TEST_EQUALITY(coarseLevel2->IsAvailable("R",Rfact.get()), false);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("P",Pfact.get()), 0);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",SmooFact.get()), 0);
-    TEST_EQUALITY(coarseLevel2->GetKeepFlag("R",Rfact.get()), 0);
+  // check interpolation
+  LocalOrdinal NSdim   = 1;
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  Ptent->apply(*coarseNullSpace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
+
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LocalOrdinal i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_EQUALITY(norms[i] < 100 * TMT::eps(), true);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, PtentEpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar,GlobalOrdinal,Node);
-#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
+}  // MakeTentativeUsingDefaultNullSpace
 
-    using TST            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-    using real           = typename TST::coordinateType;
-    typedef Xpetra::MultiVector<real,LO,GO,NO> RealValuedMultiVector;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, NoQROption, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR when nullspace isn't supplied by user" << std::endl;
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test option that skips local QR factorizations" << std::endl;
 
-    Teuchos::Array<magnitude_type> results(2);
+  GlobalOrdinal nx = 20, ny = 20;
 
-    // run test only on 1 proc
-    if(comm->getSize() == 1)
-    {
-      Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Map> map                  = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
 
-      // run Epetra and Tpetra test
-      for (int run = 0; run < 2; run++)
-      {
-        if (run == 0) lib = Xpetra::UseEpetra;
-        else lib = Xpetra::UseTpetra;
-
-        // generate problem
-        LocalOrdinal maxLevels = 3;
-        LocalOrdinal its=10;
-        GlobalOrdinal nEle = 63;
-        const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-        Teuchos::ParameterList matrixParameters;
-        matrixParameters.set("nx",nEle);
-        RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap,MultiVector>("Laplace1D", map, matrixParameters);
-        RCP<Matrix> Op = Pr->BuildMatrix();
-        RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
-
-        // build nullspace
-        RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map,1);
-        nullSpace->putScalar(TST::one());
-        Teuchos::Array<magnitude_type> norms(1);
-        nullSpace->norm1(norms);
-        if (comm->getRank() == 0)
-          out << "||NS|| = " << norms[0] << std::endl;
-
-        // fill hierarchy
-        RCP<Hierarchy> H = rcp( new Hierarchy() );
-        H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-        RCP<Level> Finest = H->GetLevel(); // first associate level with hierarchy (for defaultFactoryHandler!)
-
-        Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
-        Finest->Set("A",Op);                      // set fine level matrix
-        Finest->Set("Nullspace",nullSpace);       // set null space information for finest level
-        // Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
-
-        // define transfer operators
-        RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
-        UncoupledAggFact->SetMinNodesPerAggregate(3);
-        UncoupledAggFact->SetMaxNeighAlreadySelected(0);
-        UncoupledAggFact->SetOrdering("natural");
-
-        RCP<TentativePFactory> Pfact  = rcp( new TentativePFactory() );
-        RCP<Factory>           Rfact  = rcp( new TransPFactory() );
-        RCP<RAPFactory>        Acfact = rcp( new RAPFactory() );
-        H->SetMaxCoarseSize(1);
-
-        // setup smoothers
-        Teuchos::ParameterList smootherParamList;
-        smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
-        smootherParamList.set("relaxation: sweeps", (LocalOrdinal) 1);
-        smootherParamList.set("relaxation: damping factor", TST::one());
-        RCP<SmootherPrototype> smooProto = rcp( new TrilinosSmoother("RELAXATION", smootherParamList) );
-        RCP<SmootherFactory> SmooFact = rcp( new SmootherFactory(smooProto) );
-        Acfact->setVerbLevel(Teuchos::VERB_HIGH);
-
-        RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
-
-        FactoryManager M;
-        M.SetKokkosRefactor(false);
-        M.SetFactory("P", Pfact);
-        M.SetFactory("R", Rfact);
-        M.SetFactory("A", Acfact);
-        M.SetFactory("Ptent", Pfact);
-        M.SetFactory("Aggregates", UncoupledAggFact);
-        M.SetFactory("Smoother", SmooFact);
-        M.SetFactory("CoarseSolver", coarseSolveFact);
-        // M.SetFactory("Coordinates", Pfact);
-
-        H->Setup(M, 0, maxLevels);
-
-        // test some basic multgrid data
-        RCP<Level> coarseLevel = H->GetLevel(1);
-        RCP<Matrix> P1 = coarseLevel->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R1 = coarseLevel->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P1->getGlobalNumRows(), 63);
-        TEST_EQUALITY(P1->getGlobalNumCols(), 21);
-        TEST_EQUALITY(R1->getGlobalNumRows(), 21);
-        TEST_EQUALITY(R1->getGlobalNumCols(), 63);
-        RCP<Level> coarseLevel2 = H->GetLevel(2);
-        RCP<Matrix> P2 = coarseLevel2->Get< RCP<Matrix> >("P");
-        RCP<Matrix> R2 = coarseLevel2->Get< RCP<Matrix> >("R");
-        TEST_EQUALITY(P2->getGlobalNumRows(), 21);
-        TEST_EQUALITY(P2->getGlobalNumCols(), 7);
-        TEST_EQUALITY(R2->getGlobalNumRows(), 7);
-        TEST_EQUALITY(R2->getGlobalNumCols(), 21);
-
-        Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*P1,true,*P1,false,out);
-        Teuchos::RCP<Xpetra::Vector<Scalar,LocalOrdinal,GlobalOrdinal,Node> > diagVec = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(PtentTPtent->getRowMap());
-        PtentTPtent->getLocalDiagCopy(*diagVec);
-        TEST_EQUALITY(diagVec->norm1()-diagVec->getGlobalLength() < 100*TMT::eps(), true);
-        TEST_EQUALITY(diagVec->normInf()-TMT::one() < 100*TMT::eps(), true);
-        TEST_EQUALITY(TST::magnitude(diagVec->meanValue())-TMT::one() < 100*TMT::eps(), true);
-        TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-        // Define RHS
-        RCP<MultiVector> X = MultiVectorFactory::Build(map,1);
-        RCP<MultiVector> RHS = MultiVectorFactory::Build(map,1);
-
-        X->putScalar(1.0);
-        X->norm2(norms);
-        if (comm->getRank() == 0)
-          out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
 
-        Op->apply(*X,*RHS,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
 
-        // Use AMG directly as an iterative method
-        {
-          X->putScalar( (Scalar) 0.0);
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
+  RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", coordinates);
+
+  LocalOrdinal NSdim         = 3;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("DofsPerNode", 2);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<UncoupledAggregationFactory> UnCoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UnCoupledAggFact->SetFactory("Graph", dropFact);
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", UnCoupledAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", UnCoupledAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  Teuchos::ParameterList paramList;
+  paramList.set("tentative: calculate qr", false);
+  TentativePFact->SetParameterList(paramList);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  // check interpolation
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullSpace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
+
+  Teuchos::Array<typename TST::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LocalOrdinal i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_EQUALITY(norms[i] < 100 * TMT::eps(), true);
+  }
 
-          H->Iterate(*RHS,*X,its);
+}  // NoQR
 
-          X->norm2(norms);
-          if (comm->getRank() == 0)
-            out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
-          results[run] = norms[0];
-        }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, ConstantColumnSum, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real                  = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real, LO, GO, NO>;
+  using test_factory          = TestHelpers::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test option that skips local QR factorizations & uses constant column sum" << std::endl;
+
+  GlobalOrdinal nx = 20, ny = 20;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Map> map                  = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
+
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, galeriList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("DofsPerNode", 1);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<UncoupledAggregationFactory> UnCoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UnCoupledAggFact->SetFactory("Graph", dropFact);
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", UnCoupledAggFact);
+  RCP<TentativePFactory> TentativePFact = rcp(new TentativePFactory());
+  TentativePFact->SetFactory("Aggregates", UnCoupledAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  Teuchos::ParameterList paramList;
+  paramList.set("tentative: calculate qr", false);
+  paramList.set("tentative: constant column sums", true);
+  paramList.sublist("matrixmatrix: kernel params").set("compute global constants", true);
+  TentativePFact->SetParameterList(paramList);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  // This is "did it run test" not a "correctness check"
+
+}  // ConstColSum
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, NonStandardMaps, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  //#warning Unit test PgPFactory NonStandardMaps disabled
+  //  return;
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+#if !defined(HAVE_MUELU_IFPACK)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseEpetra, "Ifpack");
+#endif
+#if !defined(HAVE_MUELU_IFPACK2)
+  MUELU_TESTING_DO_NOT_TEST(Xpetra::UseTpetra, "Ifpack2");
+#endif
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  Xpetra::UnderlyingLib lib           = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // generate problem
+  LocalOrdinal maxLevels = 3;
+  // LocalOrdinal its=10;
+  GlobalOrdinal nEle       = 63;
+  GlobalOrdinal nIndexBase = 10;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, nIndexBase, comm);
+  RCP<Matrix> mtx          = Galeri::Xpetra::MatrixTraits<Map, CrsMatrixWrap>::Build(map, 3);
+
+  LocalOrdinal NumMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> MyGlobalElements = map->getLocalElementList();
+  GlobalOrdinal NumGlobalElements                          = map->getGlobalNumElements();
+  assert(NumGlobalElements == nEle);
+
+  GlobalOrdinal NumEntries;
+  LocalOrdinal nnz = 2;
+  std::vector<Scalar> Values(nnz);
+  std::vector<GlobalOrdinal> Indices(nnz);
+
+  Scalar a = 2.0;
+  Scalar b = -1.0;
+  Scalar c = -1.0;
+
+  for (LocalOrdinal i = 0; i < NumMyElements; ++i) {
+    if (MyGlobalElements[i] == nIndexBase) {
+      // off-diagonal for first row
+      Indices[0] = nIndexBase;
+      NumEntries = 1;
+      Values[0]  = c;
+    } else if (MyGlobalElements[i] == nIndexBase + NumGlobalElements - 1) {
+      // off-diagonal for last row
+      Indices[0] = nIndexBase + NumGlobalElements - 2;
+      NumEntries = 1;
+      Values[0]  = b;
+    } else {
+      // off-diagonal for internal row
+      Indices[0] = MyGlobalElements[i] - 1;
+      Values[1]  = b;
+      Indices[1] = MyGlobalElements[i] + 1;
+      Values[0]  = c;
+      NumEntries = 2;
+    }
+
+    // put the off-diagonal entries
+    // Xpetra wants ArrayViews (sigh)
+    Teuchos::ArrayView<Scalar> av(&Values[0], NumEntries);
+    Teuchos::ArrayView<GlobalOrdinal> iv(&Indices[0], NumEntries);
+    mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+
+    // Put in the diagonal entry
+    mtx->insertGlobalValues(MyGlobalElements[i],
+                            Teuchos::tuple<GlobalOrdinal>(MyGlobalElements[i]),
+                            Teuchos::tuple<Scalar>(a));
+
+  }  // for (LocalOrdinal i = 0; i < NumMyElements; ++i)
+
+  mtx->fillComplete(map, map);
+
+  std::cout << map->getIndexBase() << std::endl;
+
+  RCP<Matrix> Op = Teuchos::rcp_dynamic_cast<Matrix>(mtx);
+
+  // build nullspace
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+  nullSpace->putScalar((Scalar)1.0);
+
+  // fill hierarchy
+  RCP<Hierarchy> H = rcp(new Hierarchy());
+  H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  RCP<Level> Finest = H->GetLevel();  // first associate level with hierarchy (for defaultFactoryHandler!)
+
+  Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+  Finest->Set("A", Op);                 // set fine level matrix
+  Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+
+  // define transfer operators
+  RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+  UncoupledAggFact->SetMinNodesPerAggregate(3);
+  UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+  UncoupledAggFact->SetOrdering("natural");
+
+  RCP<TentativePFactory> Pfact = rcp(new TentativePFactory());
+  RCP<Factory> Rfact           = rcp(new TransPFactory());
+  RCP<RAPFactory> Acfact       = rcp(new RAPFactory());
+  H->SetMaxCoarseSize(1);
+
+  // setup smoothers
+  Teuchos::ParameterList smootherParamList;
+  smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+  smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+  smootherParamList.set("relaxation: damping factor", (Scalar)1.0);
+  RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+  RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+  Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+  FactoryManager M;
+  M.SetKokkosRefactor(false);
+  M.SetFactory("P", Pfact);
+  M.SetFactory("R", Rfact);
+  M.SetFactory("A", Acfact);
+  M.SetFactory("Ptent", Pfact);
+  M.SetFactory("Aggregates", UncoupledAggFact);
+  M.SetFactory("Smoother", SmooFact);
+  M.SetFactory("CoarseSolver", SmooFact);
+
+  H->Setup(M, 0, maxLevels);
+
+  RCP<Level> coarseLevel = H->GetLevel(1);
+  TEST_EQUALITY(coarseLevel->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("A", MueLu::NoFactory::get()), true);
+
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsRequested("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel->IsAvailable("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("P", Pfact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel->GetKeepFlag("R", Rfact.get()), 0);
+  RCP<Level> coarseLevel2 = H->GetLevel(2);
+  TEST_EQUALITY(coarseLevel2->IsRequested("A", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("P", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("R", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PreSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("PostSmoother", MueLu::NoFactory::get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("A", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("P", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", MueLu::NoFactory::get()), false);  // coarse level only has presmoother = coarse solver
+  TEST_EQUALITY(coarseLevel2->IsAvailable("R", MueLu::NoFactory::get()), true);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("A", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", MueLu::NoFactory::get()), MueLu::Final);
+  // TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother",MueLu::NoFactory::get()), MueLu::Final); // coarse level only has presmoother = coarse solver
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", MueLu::NoFactory::get()), MueLu::Final);
+  TEST_EQUALITY(coarseLevel2->IsRequested("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsRequested("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("P", Pfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PreSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("PostSmoother", SmooFact.get()), false);
+  TEST_EQUALITY(coarseLevel2->IsAvailable("R", Rfact.get()), false);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("P", Pfact.get()), 0);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("PreSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("PostSmoother", SmooFact.get()), 0);
+  TEST_EQUALITY(coarseLevel2->GetKeepFlag("R", Rfact.get()), 0);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory, PtentEpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar, GlobalOrdinal, Node);
+#if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
+
+  using TST            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  using real           = typename TST::coordinateType;
+  typedef Xpetra::MultiVector<real, LO, GO, NO> RealValuedMultiVector;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR when nullspace isn't supplied by user" << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+
+  Teuchos::Array<magnitude_type> results(2);
+
+  // run test only on 1 proc
+  if (comm->getSize() == 1) {
+    Xpetra::UnderlyingLib lib = Xpetra::UseEpetra;
+
+    // run Epetra and Tpetra test
+    for (int run = 0; run < 2; run++) {
+      if (run == 0)
+        lib = Xpetra::UseEpetra;
+      else
+        lib = Xpetra::UseTpetra;
+
+      // generate problem
+      LocalOrdinal maxLevels   = 3;
+      LocalOrdinal its         = 10;
+      GlobalOrdinal nEle       = 63;
+      const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+      Teuchos::ParameterList matrixParameters;
+      matrixParameters.set("nx", nEle);
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace1D", map, matrixParameters);
+      RCP<Matrix> Op                         = Pr->BuildMatrix();
+      RCP<RealValuedMultiVector> coordinates = Pr->BuildCoords();
+
+      // build nullspace
+      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(map, 1);
+      nullSpace->putScalar(TST::one());
+      Teuchos::Array<magnitude_type> norms(1);
+      nullSpace->norm1(norms);
+      if (comm->getRank() == 0)
+        out << "||NS|| = " << norms[0] << std::endl;
+
+      // fill hierarchy
+      RCP<Hierarchy> H = rcp(new Hierarchy());
+      H->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+      RCP<Level> Finest = H->GetLevel();  // first associate level with hierarchy (for defaultFactoryHandler!)
+
+      Finest->setDefaultVerbLevel(Teuchos::VERB_HIGH);
+      Finest->Set("A", Op);                 // set fine level matrix
+      Finest->Set("Nullspace", nullSpace);  // set null space information for finest level
+      // Finest->Set("Coordinates", coordinates);  // set coordinates for finest level
+
+      // define transfer operators
+      RCP<UncoupledAggregationFactory> UncoupledAggFact = rcp(new UncoupledAggregationFactory());
+      UncoupledAggFact->SetMinNodesPerAggregate(3);
+      UncoupledAggFact->SetMaxNeighAlreadySelected(0);
+      UncoupledAggFact->SetOrdering("natural");
+
+      RCP<TentativePFactory> Pfact = rcp(new TentativePFactory());
+      RCP<Factory> Rfact           = rcp(new TransPFactory());
+      RCP<RAPFactory> Acfact       = rcp(new RAPFactory());
+      H->SetMaxCoarseSize(1);
+
+      // setup smoothers
+      Teuchos::ParameterList smootherParamList;
+      smootherParamList.set("relaxation: type", "Symmetric Gauss-Seidel");
+      smootherParamList.set("relaxation: sweeps", (LocalOrdinal)1);
+      smootherParamList.set("relaxation: damping factor", TST::one());
+      RCP<SmootherPrototype> smooProto = rcp(new TrilinosSmoother("RELAXATION", smootherParamList));
+      RCP<SmootherFactory> SmooFact    = rcp(new SmootherFactory(smooProto));
+      Acfact->setVerbLevel(Teuchos::VERB_HIGH);
+
+      RCP<SmootherFactory> coarseSolveFact = rcp(new SmootherFactory(smooProto, Teuchos::null));
+
+      FactoryManager M;
+      M.SetKokkosRefactor(false);
+      M.SetFactory("P", Pfact);
+      M.SetFactory("R", Rfact);
+      M.SetFactory("A", Acfact);
+      M.SetFactory("Ptent", Pfact);
+      M.SetFactory("Aggregates", UncoupledAggFact);
+      M.SetFactory("Smoother", SmooFact);
+      M.SetFactory("CoarseSolver", coarseSolveFact);
+      // M.SetFactory("Coordinates", Pfact);
+
+      H->Setup(M, 0, maxLevels);
+
+      // test some basic multgrid data
+      RCP<Level> coarseLevel = H->GetLevel(1);
+      RCP<Matrix> P1         = coarseLevel->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R1         = coarseLevel->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P1->getGlobalNumRows(), 63);
+      TEST_EQUALITY(P1->getGlobalNumCols(), 21);
+      TEST_EQUALITY(R1->getGlobalNumRows(), 21);
+      TEST_EQUALITY(R1->getGlobalNumCols(), 63);
+      RCP<Level> coarseLevel2 = H->GetLevel(2);
+      RCP<Matrix> P2          = coarseLevel2->Get<RCP<Matrix> >("P");
+      RCP<Matrix> R2          = coarseLevel2->Get<RCP<Matrix> >("R");
+      TEST_EQUALITY(P2->getGlobalNumRows(), 21);
+      TEST_EQUALITY(P2->getGlobalNumCols(), 7);
+      TEST_EQUALITY(R2->getGlobalNumRows(), 7);
+      TEST_EQUALITY(R2->getGlobalNumCols(), 21);
+
+      Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > PtentTPtent = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*P1, true, *P1, false, out);
+      Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> > diagVec     = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(PtentTPtent->getRowMap());
+      PtentTPtent->getLocalDiagCopy(*diagVec);
+      TEST_EQUALITY(diagVec->norm1() - diagVec->getGlobalLength() < 100 * TMT::eps(), true);
+      TEST_EQUALITY(diagVec->normInf() - TMT::one() < 100 * TMT::eps(), true);
+      TEST_EQUALITY(TST::magnitude(diagVec->meanValue()) - TMT::one() < 100 * TMT::eps(), true);
+      TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+      // Define RHS
+      RCP<MultiVector> X   = MultiVectorFactory::Build(map, 1);
+      RCP<MultiVector> RHS = MultiVectorFactory::Build(map, 1);
+
+      X->putScalar(1.0);
+      X->norm2(norms);
+      if (comm->getRank() == 0)
+        out << "||X_true|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+
+      Op->apply(*X, *RHS, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+
+      // Use AMG directly as an iterative method
+      {
+        X->putScalar((Scalar)0.0);
+
+        H->Iterate(*RHS, *X, its);
+
+        X->norm2(norms);
+        if (comm->getRank() == 0)
+          out << "||X_" << std::setprecision(2) << its << "|| = " << std::setiosflags(std::ios::fixed) << std::setprecision(10) << norms[0] << std::endl;
+        results[run] = norms[0];
       }
+    }
 
-      TEST_FLOATING_EQUALITY(results[0], results[1], 100*TMT::eps()); // check results of EPETRA vs TPETRA
-    } // comm->getSize == 1
+    TEST_FLOATING_EQUALITY(results[0], results[1], 100 * TMT::eps());  // check results of EPETRA vs TPETRA
+  }                                                                    // comm->getSize == 1
 
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
-#   endif
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, Epetra, EpetraExt, Ifpack, Ifpack2)." << std::endl;
+#endif
 
-  } // TentativePFactory_EpetraVsTpetra
+}  // TentativePFactory_EpetraVsTpetra
 
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,Constructor,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,MakeTentative_LapackQR,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,MakeTentative,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,MakeTentativeUsingDefaultNullSpace,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,NoQROption,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,ConstantColumnSum,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,NonStandardMaps,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory,PtentEpetraVsTpetra,Scalar,LO,GO,Node)
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, Constructor, Scalar, LO, GO, Node)                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, MakeTentative_LapackQR, Scalar, LO, GO, Node)             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, MakeTentative, Scalar, LO, GO, Node)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, MakeTentativeUsingDefaultNullSpace, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, NoQROption, Scalar, LO, GO, Node)                         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, ConstantColumnSum, Scalar, LO, GO, Node)                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, NonStandardMaps, Scalar, LO, GO, Node)                    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory, PtentEpetraVsTpetra, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/ThresholdAFilterFactory.cpp b/packages/muelu/test/unit_tests/ThresholdAFilterFactory.cpp
index 766b3baccd5c..4e8febd78260 100644
--- a/packages/muelu/test/unit_tests/ThresholdAFilterFactory.cpp
+++ b/packages/muelu/test/unit_tests/ThresholdAFilterFactory.cpp
@@ -50,7 +50,6 @@
  *      Author: tobias
  */
 
-
 #include <Teuchos_UnitTestHarness.hpp>
 #include <MueLu_config.hpp>
 #include <MueLu_ConfigDefs.hpp>
@@ -67,56 +66,52 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ThresholdAFilterFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level aLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-
-    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(20); //can be an empty operator
-
-    RCP<ThresholdAFilterFactory> AfilterFactory0 = rcp(new ThresholdAFilterFactory("A",0.1)); // keep all
-    RCP<ThresholdAFilterFactory> AfilterFactory1 = rcp(new ThresholdAFilterFactory("A",1.1)); // keep only diagonal
-    RCP<ThresholdAFilterFactory> AfilterFactory2 = rcp(new ThresholdAFilterFactory("A",3));   // keep only diagonal
-
-    aLevel.Set("A",A);
-
-    aLevel.Request("A",AfilterFactory0.get());
-    AfilterFactory0->Build(aLevel);
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory0.get()), true);
-    RCP<Matrix> A0 = aLevel.Get< RCP<Matrix> >("A",AfilterFactory0.get());
-    aLevel.Release("A",AfilterFactory0.get());
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory0.get()), false);
-    TEST_EQUALITY(A0->getLocalNumEntries(), A->getLocalNumEntries());
-    TEST_EQUALITY(A0->getGlobalNumEntries(), A->getGlobalNumEntries());
-
-    aLevel.Request("A",AfilterFactory1.get());
-    AfilterFactory1->Build(aLevel);
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory1.get()), true);
-    RCP<Matrix> A1 = aLevel.Get< RCP<Matrix> >("A",AfilterFactory1.get());
-    aLevel.Release("A",AfilterFactory1.get());
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory1.get()), false);
-    TEST_EQUALITY(A1->getGlobalNumEntries(), A1->getGlobalNumRows());
-
-    aLevel.Request("A",AfilterFactory2.get());
-    AfilterFactory2->Build(aLevel);
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory2.get()), true);
-    RCP<Matrix> A2 = aLevel.Get< RCP<Matrix> >("A",AfilterFactory2.get());
-    aLevel.Release("A",AfilterFactory2.get());
-    TEST_EQUALITY(aLevel.IsAvailable("A",AfilterFactory2.get()), false);
-    TEST_EQUALITY(A2->getGlobalNumEntries(), A2->getGlobalNumRows());
-
-
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(ThresholdAFilterFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level aLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+
+  RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(20);  // can be an empty operator
+
+  RCP<ThresholdAFilterFactory> AfilterFactory0 = rcp(new ThresholdAFilterFactory("A", 0.1));  // keep all
+  RCP<ThresholdAFilterFactory> AfilterFactory1 = rcp(new ThresholdAFilterFactory("A", 1.1));  // keep only diagonal
+  RCP<ThresholdAFilterFactory> AfilterFactory2 = rcp(new ThresholdAFilterFactory("A", 3));    // keep only diagonal
+
+  aLevel.Set("A", A);
+
+  aLevel.Request("A", AfilterFactory0.get());
+  AfilterFactory0->Build(aLevel);
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory0.get()), true);
+  RCP<Matrix> A0 = aLevel.Get<RCP<Matrix> >("A", AfilterFactory0.get());
+  aLevel.Release("A", AfilterFactory0.get());
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory0.get()), false);
+  TEST_EQUALITY(A0->getLocalNumEntries(), A->getLocalNumEntries());
+  TEST_EQUALITY(A0->getGlobalNumEntries(), A->getGlobalNumEntries());
+
+  aLevel.Request("A", AfilterFactory1.get());
+  AfilterFactory1->Build(aLevel);
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory1.get()), true);
+  RCP<Matrix> A1 = aLevel.Get<RCP<Matrix> >("A", AfilterFactory1.get());
+  aLevel.Release("A", AfilterFactory1.get());
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory1.get()), false);
+  TEST_EQUALITY(A1->getGlobalNumEntries(), A1->getGlobalNumRows());
+
+  aLevel.Request("A", AfilterFactory2.get());
+  AfilterFactory2->Build(aLevel);
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory2.get()), true);
+  RCP<Matrix> A2 = aLevel.Get<RCP<Matrix> >("A", AfilterFactory2.get());
+  aLevel.Release("A", AfilterFactory2.get());
+  TEST_EQUALITY(aLevel.IsAvailable("A", AfilterFactory2.get()), false);
+  TEST_EQUALITY(A2->getGlobalNumEntries(), A2->getGlobalNumRows());
+}
 
 #define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(ThresholdAFilterFactory, Basic, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // end MueLuTests namespace
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/TransPFactory.cpp b/packages/muelu/test/unit_tests/TransPFactory.cpp
index 8c1282abddd8..abca3a29cc5f 100644
--- a/packages/muelu/test/unit_tests/TransPFactory.cpp
+++ b/packages/muelu/test/unit_tests/TransPFactory.cpp
@@ -57,81 +57,78 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TransPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<TransPFactory> transPFact = rcp(new TransPFactory);
-    TEST_EQUALITY(transPFact != Teuchos::null, true);
-
-    out << *transPFact << std::endl;
-
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TransPFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // Test of createTwoLevelHierarchy: to be moved...
-    TEST_EQUALITY(fineLevel.GetLevelID(), 0);
-    TEST_EQUALITY(coarseLevel.GetLevelID(), 1);
-    //compilation warning TEST_EQUALITY(fineLevel.GetPreviousLevel().get(), NULL);
-    //TEST_EQUALITY(coarseLevel.GetPreviousLevel().get(), &fineLevel);
-    // --
-
-    RCP<Matrix> Op = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(27*comm->getSize());
-    fineLevel.Set("A",Op);
-
-    SaPFactory sapFactory;
-    TransPFactory transPFact;
-    transPFact.SetFactory("P", rcpFromRef(sapFactory));
-    coarseLevel.Request(sapFactory);
-    coarseLevel.Request(transPFact);
-
-    coarseLevel.Request("P", &sapFactory);
-    coarseLevel.Request("R", &transPFact);
-
-    sapFactory.BuildP(fineLevel,coarseLevel);
-    transPFact.Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> P = coarseLevel.Get< RCP<Matrix> >("P", &sapFactory);
-    RCP<Matrix> R = coarseLevel.Get< RCP<Matrix> >("R", &transPFact);
-
-    RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(),1);
-    RCP<MultiVector> result2  = MultiVectorFactory::Build(R->getRangeMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getRangeMap(),1);
-    X->randomize();
-
-    //Calculate P^T * X
-    P->apply(*X,*result1,Teuchos::TRANS,(SC)1.0,(SC)0.0);
-    //Calculate R * X
-    R->apply(*X,*result2,Teuchos::NO_TRANS,(SC)1.0,(SC)0.0);
-
-    Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> normX(1), normResult1(1),normResult2(1);
-    X->norm2(normX);
-    out << "This test checks the correctness of the R created by TransPFactory." << std::endl;
-    out << "||X||_2 = " << normX << std::endl;
-    result1->norm2(normResult1);
-    result2->norm2(normResult2);
-    TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e-12);
-
-  } //Correctness test
-
-#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TransPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<TransPFactory> transPFact = rcp(new TransPFactory);
+  TEST_EQUALITY(transPFact != Teuchos::null, true);
+
+  out << *transPFact << std::endl;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TransPFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // Test of createTwoLevelHierarchy: to be moved...
+  TEST_EQUALITY(fineLevel.GetLevelID(), 0);
+  TEST_EQUALITY(coarseLevel.GetLevelID(), 1);
+  // compilation warning TEST_EQUALITY(fineLevel.GetPreviousLevel().get(), NULL);
+  // TEST_EQUALITY(coarseLevel.GetPreviousLevel().get(), &fineLevel);
+  //  --
+
+  RCP<Matrix> Op = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(27 * comm->getSize());
+  fineLevel.Set("A", Op);
+
+  SaPFactory sapFactory;
+  TransPFactory transPFact;
+  transPFact.SetFactory("P", rcpFromRef(sapFactory));
+  coarseLevel.Request(sapFactory);
+  coarseLevel.Request(transPFact);
+
+  coarseLevel.Request("P", &sapFactory);
+  coarseLevel.Request("R", &transPFact);
+
+  sapFactory.BuildP(fineLevel, coarseLevel);
+  transPFact.Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P = coarseLevel.Get<RCP<Matrix> >("P", &sapFactory);
+  RCP<Matrix> R = coarseLevel.Get<RCP<Matrix> >("R", &transPFact);
+
+  RCP<MultiVector> result1 = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> result2 = MultiVectorFactory::Build(R->getRangeMap(), 1);
+  RCP<MultiVector> X       = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  X->randomize();
+
+  // Calculate P^T * X
+  P->apply(*X, *result1, Teuchos::TRANS, (SC)1.0, (SC)0.0);
+  // Calculate R * X
+  R->apply(*X, *result2, Teuchos::NO_TRANS, (SC)1.0, (SC)0.0);
+
+  Teuchos::Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> normX(1), normResult1(1), normResult2(1);
+  X->norm2(normX);
+  out << "This test checks the correctness of the R created by TransPFactory." << std::endl;
+  out << "||X||_2 = " << normX << std::endl;
+  result1->norm2(normResult1);
+  result2->norm2(normResult2);
+  TEST_FLOATING_EQUALITY(normResult1[0], normResult2[0], 1e-12);
+
+}  // Correctness test
+
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node)                                            \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TransPFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TransPFactory, Correctness, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/UncoupledAggregationFactory.cpp b/packages/muelu/test/unit_tests/UncoupledAggregationFactory.cpp
index 811fd0658e0f..b43f3bdc4bdf 100644
--- a/packages/muelu/test/unit_tests/UncoupledAggregationFactory.cpp
+++ b/packages/muelu/test/unit_tests/UncoupledAggregationFactory.cpp
@@ -59,115 +59,106 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-
-      out << "version: " << MueLu::Version() << std::endl;
-      RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-      TEST_EQUALITY(aggFact != Teuchos::null, true);
-      TEST_THROW(aggFact->SetOrdering("unknown_ordering"), Teuchos::Exceptions::InvalidParameterValue);
-      aggFact->SetOrdering("natural");
-      TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
-      aggFact->SetOrdering("graph");
-      TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
-      aggFact->SetOrdering("random");
-      TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
-
-      aggFact->SetMaxNeighAlreadySelected(100);
-      TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 100, true);
-      aggFact->SetMinNodesPerAggregate(10);
-      TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 10, true);
-
-      auto validParamList = aggFact->GetValidParameterList();
-      TEST_EQUALITY(validParamList != Teuchos::null, true);
-
-
-  } // Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-
-      out << "version: " << MueLu::Version() << std::endl;
-      Teuchos::ParameterList galeriList;
-      GlobalOrdinal nx = 20, ny = 20;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-      RCP<Matrix> A = Pr->BuildMatrix();
-      A->SetFixedBlockSize(2);
-
-      Level aLevel;
-      TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Request("A");
-      aLevel.Set("A",A);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-      aggFact->SetFactory("Graph", dropFact);
-
-      aLevel.Request(*aggFact);
-
-      aggFact->Build(aLevel);
-
-
-  } // Build
- TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Build_ML, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-
-      out << "version: " << MueLu::Version() << std::endl;
-      Teuchos::ParameterList galeriList;
-      GlobalOrdinal nx = 20, ny = 20;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-      RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-      RCP<Matrix> A = Pr->BuildMatrix();
-      A->SetFixedBlockSize(2);
-
-      Level aLevel;
-      TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Request("A");
-      aLevel.Set("A",A);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
-
-      // Test the ML style options
-      aggFact->SetParameter("aggregation: match ML phase2a",Teuchos::ParameterEntry(true));
-      aggFact->SetParameter("aggregation: match ML phase2b",Teuchos::ParameterEntry(true));
-      aggFact->SetFactory("Graph", dropFact);
-
-      aLevel.Request(*aggFact);
-
-      aggFact->Build(aLevel);
-
-
-  } // Build
-
-
-
-  # define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Build, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Build_ML, Scalar, LO, GO, Node) 
-
-# include <MueLu_ETI_4arg.hpp>
-
-} // namespace MueLuTests
-
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+  TEST_EQUALITY(aggFact != Teuchos::null, true);
+  TEST_THROW(aggFact->SetOrdering("unknown_ordering"), Teuchos::Exceptions::InvalidParameterValue);
+  aggFact->SetOrdering("natural");
+  TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
+  aggFact->SetOrdering("graph");
+  TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
+  aggFact->SetOrdering("random");
+  TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
+
+  aggFact->SetMaxNeighAlreadySelected(100);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 100, true);
+  aggFact->SetMinNodesPerAggregate(10);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 10, true);
+
+  auto validParamList = aggFact->GetValidParameterList();
+  TEST_EQUALITY(validParamList != Teuchos::null, true);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+
+  out << "version: " << MueLu::Version() << std::endl;
+  Teuchos::ParameterList galeriList;
+  GlobalOrdinal nx = 20, ny = 20;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Map> map                  = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+  map                                 = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
+
+  Level aLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Request("A");
+  aLevel.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+  aggFact->SetFactory("Graph", dropFact);
+
+  aLevel.Request(*aggFact);
+
+  aggFact->Build(aLevel);
+
+}  // Build
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory, Build_ML, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+
+  out << "version: " << MueLu::Version() << std::endl;
+  Teuchos::ParameterList galeriList;
+  GlobalOrdinal nx = 20, ny = 20;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Map> map                  = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+  map                                 = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
+
+  Level aLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Request("A");
+  aLevel.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<UncoupledAggregationFactory> aggFact = rcp(new UncoupledAggregationFactory());
+
+  // Test the ML style options
+  aggFact->SetParameter("aggregation: match ML phase2a", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: match ML phase2b", Teuchos::ParameterEntry(true));
+  aggFact->SetFactory("Graph", dropFact);
+
+  aLevel.Request(*aggFact);
+
+  aggFact->Build(aLevel);
+
+}  // Build
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                          \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Constructor, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Build, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory, Build_ML, Scalar, LO, GO, Node)
+
+#include <MueLu_ETI_4arg.hpp>
+
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/UnsmooshFactory.cpp b/packages/muelu/test/unit_tests/UnsmooshFactory.cpp
index a364b079ade7..0c1de258123b 100644
--- a/packages/muelu/test/unit_tests/UnsmooshFactory.cpp
+++ b/packages/muelu/test/unit_tests/UnsmooshFactory.cpp
@@ -63,125 +63,125 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UnsmooshFactory, UnsmooshTentativeP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (!TYPE_EQUAL(GO, int)) { out << "Skipping test for GO != int"        << std::endl; return; }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
-    if ( comm->getSize() > 2 && lib == Xpetra::UseEpetra) { out << "Skipping test for more than 2 procs when using Epetra"        << std::endl; return; }
-
-    GlobalOrdinal nx = 6, ny = 6;
-
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> mv_type_double;
-    typedef Xpetra::MultiVectorFactory<double,LocalOrdinal,GlobalOrdinal,Node> MVFactory_double;
-
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
-
-    //build coordinates before expanding map (nodal coordinates, not dof-based)
-    RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LocalOrdinal,GlobalOrdinal,Map,mv_type_double>("2D", nodeMap, galeriList);
-    RCP<const Map> dofMap = MapFactory::Build(nodeMap, 2); //expand map for 2 DOFs per node
-
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
-
-    int maxDofPerNode = 2;
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UnsmooshFactory, UnsmooshTentativeP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (!TYPE_EQUAL(GO, int)) {
+    out << "Skipping test for GO != int" << std::endl;
+    return;
+  }
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
+  if (comm->getSize() > 2 && lib == Xpetra::UseEpetra) {
+    out << "Skipping test for more than 2 procs when using Epetra" << std::endl;
+    return;
+  }
+
+  GlobalOrdinal nx = 6, ny = 6;
+
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> mv_type_double;
+  typedef Xpetra::MultiVectorFactory<double, LocalOrdinal, GlobalOrdinal, Node> MVFactory_double;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
+
+  // build coordinates before expanding map (nodal coordinates, not dof-based)
+  RCP<mv_type_double> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LocalOrdinal, GlobalOrdinal, Map, mv_type_double>("2D", nodeMap, galeriList);
+  RCP<const Map> dofMap           = MapFactory::Build(nodeMap, 2);  // expand map for 2 DOFs per node
+
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
+
+  int maxDofPerNode = 2;
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", dofMap, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(maxDofPerNode);
-
-    //->describe(out, Teuchos::VERB_EXTREME);
-
-    TEST_EQUALITY(dofMap->getLocalNumElements(),2*nodeMap->getLocalNumElements());
-
-    // TODO introduce extra factory manager objects
-
-
-    Level fineLevel, coarseLevel;
-    TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // Test of createTwoLevelHierarchy: to be moved...
-    TEST_EQUALITY(fineLevel.GetLevelID(), 0);
-    TEST_EQUALITY(coarseLevel.GetLevelID(), 1);
-
-
-
-    fineLevel.Set("A", A);
-    fineLevel.Set("Coordinates",coordinates);
-
-    // DofPresent for VariableDofLaplacianFactory
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(),1);
-    fineLevel.Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
-    // DofStatus for UnsmooshFactory
-    Teuchos::Array<char> dofStatus = Teuchos::Array<char>(A->getRangeMap()->getLocalNumElements() * maxDofPerNode,'s');
-    fineLevel.Set("DofStatus", dofStatus);
-
-    VariableDofLaplacianFactory lapFact;
-    lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(maxDofPerNode));
-
-    AmalgamationFactory amalgFact;
-    CoalesceDropFactory coalFact;
-    NullspaceFactory nspFact;
-    TentativePFactory Pfact;
-    UnsmooshFactory unsmooFact;
-    FactoryManager innerFactManager;
-    innerFactManager.SetKokkosRefactor(false);
-    innerFactManager.SetFactory("A", Teuchos::rcpFromRef(lapFact));
-    innerFactManager.SetFactory("UnAmalgamationInfo", Teuchos::rcpFromRef(amalgFact));
-    innerFactManager.SetFactory("Graph", Teuchos::rcpFromRef(coalFact));
-    innerFactManager.SetFactory("DofsPerNode", Teuchos::rcpFromRef(coalFact));
-    innerFactManager.SetFactory("Nullspace", Teuchos::rcpFromRef(nspFact));
-    //innerFactManager.SetIgnoreUserData(true);
-
-    fineLevel.SetFactoryManager(Teuchos::rcpFromRef(innerFactManager));
-    coarseLevel.SetFactoryManager(Teuchos::rcpFromRef(innerFactManager));
-
-    // TODO it would be nice if we could use an innerFactoryManager for building Ptent and use
-    //      it for the outer factory manager.
-    //{
-    //  MueLu::SetFactoryManager fineSFM  (Teuchos::rcpFromRef(fineLevel),   Teuchos::rcpFromRef(innerFactManager));
-    //  MueLu::SetFactoryManager coarseSFM(Teuchos::rcpFromRef(coarseLevel), Teuchos::rcpFromRef(innerFactManager));
-      nspFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));   // we have to provide the correct A (since we cannot set ignoreUserData
-      amalgFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
-      coalFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
-      Pfact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
-    //}
-
-    unsmooFact.SetFactory("P", Teuchos::rcpFromRef(Pfact));
-    unsmooFact.SetFactory("DofStatus", MueLu::NoFactory::getRCP());
-    unsmooFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(maxDofPerNode));
-    unsmooFact.SetParameter("fineIsPadded", Teuchos::ParameterEntry(true));
-    unsmooFact.SetFactory("A", MueLu::NoFactory::getRCP());
-    coarseLevel.Request("P",&unsmooFact);
-    RCP<Matrix> Pfinal = coarseLevel.Get<RCP<Matrix> >("P",&unsmooFact);
-
-    //coarseLevel.print(out,MueLu::Extreme);
-    //Pfinal->describe(out, Teuchos::VERB_EXTREME);
-
-    TEST_EQUALITY(Pfinal->getRowMap()->isSameAs(*A->getRowMap()),true);
-  } // UnsmooshTentativeP
-
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UnsmooshFactory, UnsmooshTentativeP, SC, LO, GO, Node) \
-
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(maxDofPerNode);
+
+  //->describe(out, Teuchos::VERB_EXTREME);
+
+  TEST_EQUALITY(dofMap->getLocalNumElements(), 2 * nodeMap->getLocalNumElements());
+
+  // TODO introduce extra factory manager objects
+
+  Level fineLevel, coarseLevel;
+  TestHelpers::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // Test of createTwoLevelHierarchy: to be moved...
+  TEST_EQUALITY(fineLevel.GetLevelID(), 0);
+  TEST_EQUALITY(coarseLevel.GetLevelID(), 1);
+
+  fineLevel.Set("A", A);
+  fineLevel.Set("Coordinates", coordinates);
+
+  // DofPresent for VariableDofLaplacianFactory
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(), 1);
+  fineLevel.Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
+  // DofStatus for UnsmooshFactory
+  Teuchos::Array<char> dofStatus = Teuchos::Array<char>(A->getRangeMap()->getLocalNumElements() * maxDofPerNode, 's');
+  fineLevel.Set("DofStatus", dofStatus);
+
+  VariableDofLaplacianFactory lapFact;
+  lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(maxDofPerNode));
+
+  AmalgamationFactory amalgFact;
+  CoalesceDropFactory coalFact;
+  NullspaceFactory nspFact;
+  TentativePFactory Pfact;
+  UnsmooshFactory unsmooFact;
+  FactoryManager innerFactManager;
+  innerFactManager.SetKokkosRefactor(false);
+  innerFactManager.SetFactory("A", Teuchos::rcpFromRef(lapFact));
+  innerFactManager.SetFactory("UnAmalgamationInfo", Teuchos::rcpFromRef(amalgFact));
+  innerFactManager.SetFactory("Graph", Teuchos::rcpFromRef(coalFact));
+  innerFactManager.SetFactory("DofsPerNode", Teuchos::rcpFromRef(coalFact));
+  innerFactManager.SetFactory("Nullspace", Teuchos::rcpFromRef(nspFact));
+  // innerFactManager.SetIgnoreUserData(true);
+
+  fineLevel.SetFactoryManager(Teuchos::rcpFromRef(innerFactManager));
+  coarseLevel.SetFactoryManager(Teuchos::rcpFromRef(innerFactManager));
+
+  // TODO it would be nice if we could use an innerFactoryManager for building Ptent and use
+  //      it for the outer factory manager.
+  //{
+  //  MueLu::SetFactoryManager fineSFM  (Teuchos::rcpFromRef(fineLevel),   Teuchos::rcpFromRef(innerFactManager));
+  //  MueLu::SetFactoryManager coarseSFM(Teuchos::rcpFromRef(coarseLevel), Teuchos::rcpFromRef(innerFactManager));
+  nspFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));  // we have to provide the correct A (since we cannot set ignoreUserData
+  amalgFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
+  coalFact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
+  Pfact.SetFactory("A", Teuchos::rcpFromRef(lapFact));
+  //}
+
+  unsmooFact.SetFactory("P", Teuchos::rcpFromRef(Pfact));
+  unsmooFact.SetFactory("DofStatus", MueLu::NoFactory::getRCP());
+  unsmooFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(maxDofPerNode));
+  unsmooFact.SetParameter("fineIsPadded", Teuchos::ParameterEntry(true));
+  unsmooFact.SetFactory("A", MueLu::NoFactory::getRCP());
+  coarseLevel.Request("P", &unsmooFact);
+  RCP<Matrix> Pfinal = coarseLevel.Get<RCP<Matrix> >("P", &unsmooFact);
+
+  // coarseLevel.print(out,MueLu::Extreme);
+  // Pfinal->describe(out, Teuchos::VERB_EXTREME);
+
+  TEST_EQUALITY(Pfinal->getRowMap()->isSameAs(*A->getRowMap()), true);
+}  // UnsmooshTentativeP
+
+#define MUELU_ETI_GROUP(SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UnsmooshFactory, UnsmooshTentativeP, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
-}
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/UserData/CreateXpetraPreconditioner.cpp b/packages/muelu/test/unit_tests/UserData/CreateXpetraPreconditioner.cpp
index d36beb54929d..fa9c9cd42d29 100644
--- a/packages/muelu/test/unit_tests/UserData/CreateXpetraPreconditioner.cpp
+++ b/packages/muelu/test/unit_tests/UserData/CreateXpetraPreconditioner.cpp
@@ -76,122 +76,121 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UserData, CreateXpetraPreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UserData, CreateXpetraPreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using Teuchos::RCP;
-
-    Xpetra::UnderlyingLib          lib  = TestHelpers::Parameters::getLib();
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    GO  nx     = 1000;
-
-    std::string xmlFileName = "UserData/test.xml";
-    Teuchos::RCP<Teuchos::ParameterList> inParamList = Teuchos::getParametersFromXmlFile(xmlFileName);
-    if(lib == Xpetra::UseEpetra) {
-      inParamList->sublist("Hierarchy").set("use kokkos refactor", false);
-    }
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
-
-    // Matrix
-    RCP<Matrix>     Op  = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
-    RCP<const Map > map = Op->getRowMap();
-
-    // Normalized RHS
-    RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
-    RHS1->setSeed(846930886);
-    RHS1->randomize();
-    Teuchos::Array<magnitude_type> norms(1);
-    RHS1->norm2(norms);
-    RHS1->scale(1/norms[0]);
-
-    // Zero initial guess
-    RCP<MultiVector> X1   = MultiVectorFactory::Build(Op->getRowMap(), 1);
-    X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
-
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", Op->getRowMap(), galeriList);
-    RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC,LO,GO,NO>::Build(Op->getDomainMap(), 1);
-    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
-
-    // Add sublist "user data" to MueLu's parameter list
-    const std::string userName = "user data";
-    Teuchos::ParameterList& userParamList = inParamList->sublist(userName);
-
-    // Create test variables to be stored on Level 0 of the Hierarchy
-    SC myScalar = 3.14;
-    userParamList.set<SC>("Scalar myScalar", myScalar);
-    double myDouble = 2.71;
-    userParamList.set<double>("Double myDouble", myDouble);
-    int myInt = 42;
-    userParamList.set<int>("int myInt", myInt);
-    std::string myString = "Test string";
-    userParamList.set<std::string>("String myString", myString);
-    Array<GO> myArrayGO(4);
-    myArrayGO[0] = 1;
-    myArrayGO[1] = 4;
-    myArrayGO[2] = 5;
-    myArrayGO[3] = 0;
-    userParamList.set<Array<GO> >("Array<GO> myArray<GO>", myArrayGO);
-    Array<LO> myArrayLO(5);
-    myArrayLO[0] = 8;
-    myArrayLO[1] = 7;
-    myArrayLO[2] = 1;
-    myArrayLO[3] = 2;
-    myArrayLO[4] = 3;
-    userParamList.set<Array<LO> >("Array<LO> myArray<LO>", myArrayLO);
-
-    RCP<Hierarchy> xH = MueLu::CreateXpetraPreconditioner<SC,LO,GO,NO>(Op, *inParamList);
-
-    // Extract variables on level 0 and check that they are unchanged.
-    RCP<MueLu::Level> level0 = xH->GetLevel();
-    bool result = true;
-    std::string errorMsg = "";
-
-    if(! (level0->Get<SC>("myScalar") == myScalar) ) {
-      errorMsg += "myScalar does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    if( !(level0->Get<double>("myDouble") == myDouble) ) {
-      errorMsg += "myDouble does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    if( !(level0->Get<int>("myInt") == myInt) ) {
-      errorMsg += "myInt does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    if( !(level0->Get<std::string>("myString") == myString) ) {
-      errorMsg += "myString does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    if( !(level0->Get<Array<GO> >("myArray<GO>") == myArrayGO) ) {
-      errorMsg += "myArray<GO> does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    if( !(level0->Get<Array<LO> >("myArray<LO>") == myArrayLO) ) {
-      errorMsg += "myArray<LO> does not have correct value on level 0.\n";
-      result = false;
-    }
-
-    TEST_EQUALITY(result, true);
-  } //CreatePreconditioner
-
-#define MUELU_ETI_GROUP(Scalar,LocalOrdinal,GlobalOrdinal,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UserData,CreateXpetraPreconditioner,Scalar,LocalOrdinal,GlobalOrdinal,Node)
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using Teuchos::RCP;
+
+  Xpetra::UnderlyingLib lib           = TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  GO nx = 1000;
+
+  std::string xmlFileName                          = "UserData/test.xml";
+  Teuchos::RCP<Teuchos::ParameterList> inParamList = Teuchos::getParametersFromXmlFile(xmlFileName);
+  if (lib == Xpetra::UseEpetra) {
+    inParamList->sublist("Hierarchy").set("use kokkos refactor", false);
+  }
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType magnitude_type;
+
+  // Matrix
+  RCP<Matrix> Op     = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx * comm->getSize(), lib);
+  RCP<const Map> map = Op->getRowMap();
+
+  // Normalized RHS
+  RCP<MultiVector> RHS1 = MultiVectorFactory::Build(map, 1);
+  RHS1->setSeed(846930886);
+  RHS1->randomize();
+  Teuchos::Array<magnitude_type> norms(1);
+  RHS1->norm2(norms);
+  RHS1->scale(1 / norms[0]);
+
+  // Zero initial guess
+  RCP<MultiVector> X1 = MultiVectorFactory::Build(Op->getRowMap(), 1);
+  X1->putScalar(Teuchos::ScalarTraits<SC>::zero());
+
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  RCP<MultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", Op->getRowMap(), galeriList);
+  RCP<MultiVector> nullspace   = Xpetra::MultiVectorFactory<SC, LO, GO, NO>::Build(Op->getDomainMap(), 1);
+  nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+
+  // Add sublist "user data" to MueLu's parameter list
+  const std::string userName            = "user data";
+  Teuchos::ParameterList& userParamList = inParamList->sublist(userName);
+
+  // Create test variables to be stored on Level 0 of the Hierarchy
+  SC myScalar = 3.14;
+  userParamList.set<SC>("Scalar myScalar", myScalar);
+  double myDouble = 2.71;
+  userParamList.set<double>("Double myDouble", myDouble);
+  int myInt = 42;
+  userParamList.set<int>("int myInt", myInt);
+  std::string myString = "Test string";
+  userParamList.set<std::string>("String myString", myString);
+  Array<GO> myArrayGO(4);
+  myArrayGO[0] = 1;
+  myArrayGO[1] = 4;
+  myArrayGO[2] = 5;
+  myArrayGO[3] = 0;
+  userParamList.set<Array<GO> >("Array<GO> myArray<GO>", myArrayGO);
+  Array<LO> myArrayLO(5);
+  myArrayLO[0] = 8;
+  myArrayLO[1] = 7;
+  myArrayLO[2] = 1;
+  myArrayLO[3] = 2;
+  myArrayLO[4] = 3;
+  userParamList.set<Array<LO> >("Array<LO> myArray<LO>", myArrayLO);
+
+  RCP<Hierarchy> xH = MueLu::CreateXpetraPreconditioner<SC, LO, GO, NO>(Op, *inParamList);
+
+  // Extract variables on level 0 and check that they are unchanged.
+  RCP<MueLu::Level> level0 = xH->GetLevel();
+  bool result              = true;
+  std::string errorMsg     = "";
+
+  if (!(level0->Get<SC>("myScalar") == myScalar)) {
+    errorMsg += "myScalar does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  if (!(level0->Get<double>("myDouble") == myDouble)) {
+    errorMsg += "myDouble does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  if (!(level0->Get<int>("myInt") == myInt)) {
+    errorMsg += "myInt does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  if (!(level0->Get<std::string>("myString") == myString)) {
+    errorMsg += "myString does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  if (!(level0->Get<Array<GO> >("myArray<GO>") == myArrayGO)) {
+    errorMsg += "myArray<GO> does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  if (!(level0->Get<Array<LO> >("myArray<LO>") == myArrayLO)) {
+    errorMsg += "myArray<LO> does not have correct value on level 0.\n";
+    result = false;
+  }
+
+  TEST_EQUALITY(result, true);
+}  // CreatePreconditioner
+
+#define MUELU_ETI_GROUP(Scalar, LocalOrdinal, GlobalOrdinal, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UserData, CreateXpetraPreconditioner, Scalar, LocalOrdinal, GlobalOrdinal, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
 
-#endif // HAVE_MUELU_EPETRA && HAVE_MUELU_SERIAL
+#endif  // HAVE_MUELU_EPETRA && HAVE_MUELU_SERIAL
diff --git a/packages/muelu/test/unit_tests/Utilities.cpp b/packages/muelu/test/unit_tests/Utilities.cpp
index 0765236f45e1..5b2aa32e874f 100644
--- a/packages/muelu/test/unit_tests/Utilities.cpp
+++ b/packages/muelu/test/unit_tests/Utilities.cpp
@@ -63,241 +63,264 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,MatMatMult_EpetraVsTpetra,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, MatMatMult_EpetraVsTpetra, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT)
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "This test compares the matrix matrix multiply between Tpetra and Epetra" << std::endl;
-
-    MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar,GlobalOrdinal,Node);
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
-
-    //Calculate result = (Op*Op)*X for Epetra
-    GO nx = 37*comm->getSize();
-    GO ny = nx;
-    RCP<Matrix> Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx,ny,Xpetra::UseEpetra);
-    RCP<Matrix> OpOp = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Op,false,*Op,false,out);
-    RCP<MultiVector> result = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    RCP<MultiVector> X = MultiVectorFactory::Build(OpOp->getDomainMap(),1);
-    Teuchos::Array<magnitude_type> xnorm(1);
-    X->setSeed(8675309);
-    X->randomize(true);
-    X->norm2(xnorm);
-    OpOp->apply(*X,*result,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Teuchos::Array<magnitude_type> normEpetra(1);
-    result->norm2(normEpetra);
-
-    // aid debugging by calculating Op*(Op*X)
-    RCP<MultiVector> workVec = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    RCP<MultiVector> check1 = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    Op->apply(*X,*workVec,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Op->apply(*workVec,*check1,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Teuchos::Array<magnitude_type> normCheck1(1);
-    check1->norm2(normCheck1);
-
-    //Calculate result = (Op*Op)*X for Tpetra
-    Op = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx,ny,Xpetra::UseTpetra);
-    OpOp = Xpetra::MatrixMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Multiply(*Op,false,*Op,false,out);
-    result = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    X = MultiVectorFactory::Build(OpOp->getDomainMap(),1);
-    X->setSeed(8675309);
-    X->randomize(true);
-    X->norm2(xnorm);
-    OpOp->apply(*X,*result,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Teuchos::Array<magnitude_type> normTpetra(1);
-    result->norm2(normTpetra);
-
-    // aid debugging by calculating Op*(Op*X)
-    workVec = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    RCP<MultiVector> check2 = MultiVectorFactory::Build(OpOp->getRangeMap(),1);
-    Op->apply(*X,*workVec,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Op->apply(*workVec,*check2,Teuchos::NO_TRANS,(Scalar)1.0,(Scalar)0.0);
-    Teuchos::Array<magnitude_type> normCheck2(1);
-    check2->norm2(normCheck2);
-
-    TEST_FLOATING_EQUALITY(normEpetra[0], normTpetra[0], 1e-12);
-    out << "Epetra ||A*(A*x)|| = " << normCheck1[0] << std::endl;
-    out << "Tpetra ||A*(A*x)|| = " << normCheck2[0] << std::endl;
-#   else
-    out << "Skipping test because some required packages are not enabled (Tpetra, EpetraExt)." << std::endl;
-#   endif
-
-  } //EpetraVersusTpetra
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,DetectDirichletRows,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(100);
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar>  values;
-
-    LocalOrdinal localRowToZero = 5;
-    A->resumeFill();
-    A->getLocalRowView(localRowToZero, indices, values);
-    Array<Scalar> newvalues(values.size(),TST::zero());
-    for (int j = 0; j < indices.size(); j++)
-      //keep diagonal
-      if (indices[j] == localRowToZero) newvalues[j] = values[j];
-    A->replaceLocalValues(localRowToZero,indices,newvalues);
-
-    A->fillComplete();
-
-    ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
-    TEST_EQUALITY(drows[localRowToZero], true);
-    TEST_EQUALITY(drows[localRowToZero-1], false);
-
-    A->resumeFill();
-    A->getLocalRowView(localRowToZero, indices, values);
-    for (int j = 0; j < indices.size(); j++)
-      //keep diagonal
-      if (indices[j] == localRowToZero) newvalues[j] = values[j];
-      else newvalues[j] = Teuchos::as<Scalar>(0.25);
-    A->replaceLocalValues(localRowToZero,indices,newvalues);
-
-    //row 5 should not be Dirichlet
-    drows = Utilities::DetectDirichletRows(*A,TST::magnitude(0.24));
-    TEST_EQUALITY(drows[localRowToZero], false);
-    TEST_EQUALITY(drows[localRowToZero-1], false);
-
-    //row 5 should be Dirichlet
-    drows = Utilities::DetectDirichletRows(*A,TST::magnitude(0.26));
-    TEST_EQUALITY(drows[localRowToZero], true);
-    TEST_EQUALITY(drows[localRowToZero-1], false);
-
-  } //DetectDirichletRows
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,GetDiagonalInverse,Scalar,LocalOrdinal,GlobalOrdinal,Node)
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "This test compares the matrix matrix multiply between Tpetra and Epetra" << std::endl;
+
+  MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(Scalar, GlobalOrdinal, Node);
+
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::magnitudeType magnitude_type;
+
+  // Calculate result = (Op*Op)*X for Epetra
+  GO nx                   = 37 * comm->getSize();
+  GO ny                   = nx;
+  RCP<Matrix> Op          = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx, ny, Xpetra::UseEpetra);
+  RCP<Matrix> OpOp        = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Op, false, *Op, false, out);
+  RCP<MultiVector> result = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  RCP<MultiVector> X      = MultiVectorFactory::Build(OpOp->getDomainMap(), 1);
+  Teuchos::Array<magnitude_type> xnorm(1);
+  X->setSeed(8675309);
+  X->randomize(true);
+  X->norm2(xnorm);
+  OpOp->apply(*X, *result, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Teuchos::Array<magnitude_type> normEpetra(1);
+  result->norm2(normEpetra);
+
+  // aid debugging by calculating Op*(Op*X)
+  RCP<MultiVector> workVec = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  RCP<MultiVector> check1  = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  Op->apply(*X, *workVec, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Op->apply(*workVec, *check1, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Teuchos::Array<magnitude_type> normCheck1(1);
+  check1->norm2(normCheck1);
+
+  // Calculate result = (Op*Op)*X for Tpetra
+  Op     = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build2DPoisson(nx, ny, Xpetra::UseTpetra);
+  OpOp   = Xpetra::MatrixMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Multiply(*Op, false, *Op, false, out);
+  result = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  X      = MultiVectorFactory::Build(OpOp->getDomainMap(), 1);
+  X->setSeed(8675309);
+  X->randomize(true);
+  X->norm2(xnorm);
+  OpOp->apply(*X, *result, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Teuchos::Array<magnitude_type> normTpetra(1);
+  result->norm2(normTpetra);
+
+  // aid debugging by calculating Op*(Op*X)
+  workVec                 = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  RCP<MultiVector> check2 = MultiVectorFactory::Build(OpOp->getRangeMap(), 1);
+  Op->apply(*X, *workVec, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Op->apply(*workVec, *check2, Teuchos::NO_TRANS, (Scalar)1.0, (Scalar)0.0);
+  Teuchos::Array<magnitude_type> normCheck2(1);
+  check2->norm2(normCheck2);
+
+  TEST_FLOATING_EQUALITY(normEpetra[0], normTpetra[0], 1e-12);
+  out << "Epetra ||A*(A*x)|| = " << normCheck1[0] << std::endl;
+  out << "Tpetra ||A*(A*x)|| = " << normCheck2[0] << std::endl;
+#else
+  out << "Skipping test because some required packages are not enabled (Tpetra, EpetraExt)." << std::endl;
+#endif
+
+}  // EpetraVersusTpetra
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, DetectDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(100);
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> values;
+
+  LocalOrdinal localRowToZero = 5;
+  A->resumeFill();
+  A->getLocalRowView(localRowToZero, indices, values);
+  Array<Scalar> newvalues(values.size(), TST::zero());
+  for (int j = 0; j < indices.size(); j++)
+    // keep diagonal
+    if (indices[j] == localRowToZero) newvalues[j] = values[j];
+  A->replaceLocalValues(localRowToZero, indices, newvalues);
+
+  A->fillComplete();
+
+  ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
+  TEST_EQUALITY(drows[localRowToZero], true);
+  TEST_EQUALITY(drows[localRowToZero - 1], false);
+
+  A->resumeFill();
+  A->getLocalRowView(localRowToZero, indices, values);
+  for (int j = 0; j < indices.size(); j++)
+    // keep diagonal
+    if (indices[j] == localRowToZero)
+      newvalues[j] = values[j];
+    else
+      newvalues[j] = Teuchos::as<Scalar>(0.25);
+  A->replaceLocalValues(localRowToZero, indices, newvalues);
+
+  // row 5 should not be Dirichlet
+  drows = Utilities::DetectDirichletRows(*A, TST::magnitude(0.24));
+  TEST_EQUALITY(drows[localRowToZero], false);
+  TEST_EQUALITY(drows[localRowToZero - 1], false);
+
+  // row 5 should be Dirichlet
+  drows = Utilities::DetectDirichletRows(*A, TST::magnitude(0.26));
+  TEST_EQUALITY(drows[localRowToZero], true);
+  TEST_EQUALITY(drows[localRowToZero - 1], false);
+
+}  // DetectDirichletRows
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, GetDiagonalInverse, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+
+  Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+
+  // blocked diagonal operator (Xpetra)
+  RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+
+  RCP<const BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
+  TEST_EQUALITY(bA != Teuchos::null, true);
+
   {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    // blocked diagonal operator (Xpetra)
-    RCP<Matrix> A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
-
-    RCP<const BlockedCrsMatrix> bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
-    TEST_EQUALITY(bA != Teuchos::null, true);
-
-    {
-      RCP<Vector> diagInv = Utilities::GetMatrixDiagonalInverse(*bA);
-      //diagInv->describe(out,Teuchos::VERB_EXTREME);
-      RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
-      TEST_EQUALITY(bDiagInv.is_null(), false);
-      TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())),true);
-      RCP<MultiVector> diagInvMerged = bDiagInv->Merge();
-      Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
-      for(size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
-        if(i < 5 ) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0),true);
-        if(i >= 5  && i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5),true);
-        if(i >= 10 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0/3.0),true);
-      }
+    RCP<Vector> diagInv = Utilities::GetMatrixDiagonalInverse(*bA);
+    // diagInv->describe(out,Teuchos::VERB_EXTREME);
+    RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
+    TEST_EQUALITY(bDiagInv.is_null(), false);
+    TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())), true);
+    RCP<MultiVector> diagInvMerged              = bDiagInv->Merge();
+    Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
+    for (size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
+      if (i < 5) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0), true);
+      if (i >= 5 && i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5), true);
+      if (i >= 10 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0 / 3.0), true);
     }
+  }
 
-    A = Teuchos::null; bA = Teuchos::null;
-
-    // blocked diagonal operator (Thyra)
-    A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
-
-    bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
-    TEST_EQUALITY(bA != Teuchos::null, true);
-    {
-      RCP<Vector> diagInv = Utilities::GetMatrixDiagonalInverse(*bA);
-      RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
-      TEST_EQUALITY(bDiagInv.is_null(), false);
-      TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())),true);
-      RCP<MultiVector> diagInvMerged = bDiagInv->Merge();
-      Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
-      for(size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
-        if(i < 5 ) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0),true);
-        if(i >= 5  && i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5),true);
-        if(i >= 10 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0/3.0),true);
-      }
-    }
-    // reordered blocked diagonal operator (Xpetra)
-    A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
-    Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [ 2 0] 1 ]");
-    RCP<const BlockedCrsMatrix> bAA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
-    bA = Teuchos::rcp_dynamic_cast<const Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(buildReorderedBlockedCrsMatrix(brm, bAA));
-
-    TEST_EQUALITY(bA->Rows(),2);
-    TEST_EQUALITY(bA->Cols(),2);
-
-    TEST_EQUALITY(bA->getRangeMapExtractor()->getThyraMode(),false);
-    TEST_EQUALITY(bA->getDomainMapExtractor()->getThyraMode(),false);
-    {
-      RCP<Vector> diagInv = Utilities::GetMatrixDiagonalInverse(*bA);
-      RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
-      TEST_EQUALITY(bDiagInv.is_null(), false);
-      TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())),true);
-      RCP<MultiVector> diagInvMerged = bDiagInv->Merge();
-      Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
-      for(size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
-        if(i >= 10 && i < 15) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0),true);
-        if(i >= 15 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5),true);
-        if(i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0/3.0),true);
-      }
+  A  = Teuchos::null;
+  bA = Teuchos::null;
+
+  // blocked diagonal operator (Thyra)
+  A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+
+  bA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
+  TEST_EQUALITY(bA != Teuchos::null, true);
+  {
+    RCP<Vector> diagInv         = Utilities::GetMatrixDiagonalInverse(*bA);
+    RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
+    TEST_EQUALITY(bDiagInv.is_null(), false);
+    TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())), true);
+    RCP<MultiVector> diagInvMerged              = bDiagInv->Merge();
+    Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
+    for (size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
+      if (i < 5) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0), true);
+      if (i >= 5 && i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5), true);
+      if (i >= 10 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0 / 3.0), true);
     }
-    // reordered blocked diagonal operator (Thyra)
-    A = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
-    brm = Xpetra::blockedReorderFromString("[ [ 2 0] 1 ]");
-    bAA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
-    bA = Teuchos::rcp_dynamic_cast<const Xpetra::ReorderedBlockedCrsMatrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >(buildReorderedBlockedCrsMatrix(brm, bAA));
-
-    TEST_EQUALITY(bA->Rows(),2);
-    TEST_EQUALITY(bA->Cols(),2);
-
-    TEST_EQUALITY(bA->getRangeMapExtractor()->getThyraMode(),true);
-    TEST_EQUALITY(bA->getDomainMapExtractor()->getThyraMode(),true);
-    {
-      RCP<Vector> diagInv = Utilities::GetMatrixDiagonalInverse(*bA);
-      RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
-      TEST_EQUALITY(bDiagInv.is_null(), false);
-      TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())),true);
-      RCP<MultiVector> diagInvMerged = bDiagInv->Merge();
-      Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
-      for(size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
-        if(i >= 10 && i < 15) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0),true);
-        if(i >= 15 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5),true);
-        if(i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0/3.0),true);
-      }
+  }
+  // reordered blocked diagonal operator (Xpetra)
+  A                                                   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrix(lib, 3, comm);
+  Teuchos::RCP<const Xpetra::BlockReorderManager> brm = Xpetra::blockedReorderFromString("[ [ 2 0] 1 ]");
+  RCP<const BlockedCrsMatrix> bAA                     = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
+  bA                                                  = Teuchos::rcp_dynamic_cast<const Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(buildReorderedBlockedCrsMatrix(brm, bAA));
+
+  TEST_EQUALITY(bA->Rows(), 2);
+  TEST_EQUALITY(bA->Cols(), 2);
+
+  TEST_EQUALITY(bA->getRangeMapExtractor()->getThyraMode(), false);
+  TEST_EQUALITY(bA->getDomainMapExtractor()->getThyraMode(), false);
+  {
+    RCP<Vector> diagInv         = Utilities::GetMatrixDiagonalInverse(*bA);
+    RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
+    TEST_EQUALITY(bDiagInv.is_null(), false);
+    TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())), true);
+    RCP<MultiVector> diagInvMerged              = bDiagInv->Merge();
+    Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
+    for (size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
+      if (i >= 10 && i < 15) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0), true);
+      if (i >= 15 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5), true);
+      if (i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0 / 3.0), true);
     }
-  } // GetDiagonalInverse
+  }
+  // reordered blocked diagonal operator (Thyra)
+  A   = TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CreateBlockDiagonalExampleMatrixThyra(lib, 3, comm);
+  brm = Xpetra::blockedReorderFromString("[ [ 2 0] 1 ]");
+  bAA = Teuchos::rcp_dynamic_cast<const BlockedCrsMatrix>(A);
+  bA  = Teuchos::rcp_dynamic_cast<const Xpetra::ReorderedBlockedCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>(buildReorderedBlockedCrsMatrix(brm, bAA));
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,GetLumpedDiagonal,Scalar,LocalOrdinal,GlobalOrdinal,Node)
+  TEST_EQUALITY(bA->Rows(), 2);
+  TEST_EQUALITY(bA->Cols(), 2);
+
+  TEST_EQUALITY(bA->getRangeMapExtractor()->getThyraMode(), true);
+  TEST_EQUALITY(bA->getDomainMapExtractor()->getThyraMode(), true);
   {
-    // Note: Lumped diagonal does not support blocked operations.
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+    RCP<Vector> diagInv         = Utilities::GetMatrixDiagonalInverse(*bA);
+    RCP<BlockedVector> bDiagInv = Teuchos::rcp_dynamic_cast<BlockedVector>(diagInv);
+    TEST_EQUALITY(bDiagInv.is_null(), false);
+    TEST_EQUALITY(bDiagInv->getBlockedMap()->isSameAs(*(bA->getRangeMap())), true);
+    RCP<MultiVector> diagInvMerged              = bDiagInv->Merge();
+    Teuchos::ArrayRCP<const Scalar> diagInvData = diagInvMerged->getData(0);
+    for (size_t i = 0; i < Teuchos::as<size_t>(diagInvData.size()); ++i) {
+      if (i >= 10 && i < 15) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0), true);
+      if (i >= 15 && i < 20) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(0.5), true);
+      if (i < 10) TEST_EQUALITY(diagInvData[i] == Teuchos::as<Scalar>(1.0 / 3.0), true);
+    }
+  }
+}  // GetDiagonalInverse
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, GetLumpedDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // Note: Lumped diagonal does not support blocked operations.
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
 
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+  Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+
+  RCP<CrsMatrix> crs_mat;
+  RCP<const Map> row_map, col_map;
+
+  {
+    ArrayRCP<size_t> nnzPerRow(9, 0);
+    row_map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9 * comm->getSize(), 0, comm);
+    if (comm->getSize() == 1) {
+      col_map      = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 15, 0, comm);
+      nnzPerRow[0] = 3;
+      nnzPerRow[1] = 4;
+      nnzPerRow[2] = 4;
+      nnzPerRow[3] = 4;
+      nnzPerRow[4] = 5;
+      nnzPerRow[5] = 5;
+      nnzPerRow[6] = 4;
+      nnzPerRow[7] = 5;
+      nnzPerRow[8] = 5;
+    } else if (comm->getSize() == 4) {
+      Teuchos::Array<GlobalOrdinal> global_colids(15);
+      const GlobalOrdinal rank_offset = static_cast<GlobalOrdinal>(comm->getRank() * 9);
+      for (GlobalOrdinal colIdx = 0; colIdx < 9; ++colIdx) {
+        global_colids[colIdx] = colIdx + rank_offset;
+      }
 
-    RCP<CrsMatrix> crs_mat;
-    RCP<const Map> row_map, col_map;
+      if (comm->getRank() == 0) {
+        global_colids[9]  = 9;
+        global_colids[10] = 12;
+        global_colids[11] = 15;
+        global_colids[12] = 18;
+        global_colids[13] = 19;
+        global_colids[14] = 20;
 
-    {
-      ArrayRCP<size_t> nnzPerRow(9, 0);
-      row_map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 9*comm->getSize(), 0, comm);
-      if(comm->getSize() == 1) {
-        col_map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 15, 0, comm);
         nnzPerRow[0] = 3;
         nnzPerRow[1] = 4;
         nnzPerRow[2] = 4;
@@ -307,414 +330,398 @@ namespace MueLuTests {
         nnzPerRow[6] = 4;
         nnzPerRow[7] = 5;
         nnzPerRow[8] = 5;
-      } else if(comm->getSize() == 4) {
-        Teuchos::Array<GlobalOrdinal> global_colids(15);
-        const GlobalOrdinal rank_offset = static_cast<GlobalOrdinal>(comm->getRank() * 9);
-        for(GlobalOrdinal colIdx = 0; colIdx < 9; ++colIdx) {
-          global_colids[colIdx] = colIdx + rank_offset;
-        }
-
-        if(comm->getRank() == 0) {
-          global_colids[9]  = 9;
-          global_colids[10] = 12;
-          global_colids[11] = 15;
-          global_colids[12] = 18;
-          global_colids[13] = 19;
-          global_colids[14] = 20;
-
-          nnzPerRow[0] = 3;
-          nnzPerRow[1] = 4;
-          nnzPerRow[2] = 4;
-          nnzPerRow[3] = 4;
-          nnzPerRow[4] = 5;
-          nnzPerRow[5] = 5;
-          nnzPerRow[6] = 4;
-          nnzPerRow[7] = 5;
-          nnzPerRow[8] = 5;
-        } else if(comm->getRank() == 1) {
-          global_colids[9]  = 2;
-          global_colids[10] = 5;
-          global_colids[11] = 8;
-          global_colids[12] = 27;
-          global_colids[13] = 28;
-          global_colids[14] = 29;
-
-          nnzPerRow[0] = 4;
-          nnzPerRow[1] = 4;
-          nnzPerRow[2] = 3;
-          nnzPerRow[3] = 5;
-          nnzPerRow[4] = 5;
-          nnzPerRow[5] = 4;
-          nnzPerRow[6] = 5;
-          nnzPerRow[7] = 5;
-          nnzPerRow[8] = 4;
-        } else if(comm->getRank() == 2) {
-          global_colids[9]  = 6;
-          global_colids[10] = 7;
-          global_colids[11] = 8;
-          global_colids[12] = 27;
-          global_colids[13] = 30;
-          global_colids[14] = 33;
-
-          nnzPerRow[0] = 4;
-          nnzPerRow[1] = 5;
-          nnzPerRow[2] = 5;
-          nnzPerRow[3] = 4;
-          nnzPerRow[4] = 5;
-          nnzPerRow[5] = 5;
-          nnzPerRow[6] = 3;
-          nnzPerRow[7] = 4;
-          nnzPerRow[8] = 4;
-        } else if(comm->getRank() == 3) {
-          global_colids[9]  = 15;
-          global_colids[10] = 16;
-          global_colids[11] = 17;
-          global_colids[12] = 20;
-          global_colids[13] = 23;
-          global_colids[14] = 26;
-
-          nnzPerRow[0] = 5;
-          nnzPerRow[1] = 5;
-          nnzPerRow[2] = 4;
-          nnzPerRow[3] = 5;
-          nnzPerRow[4] = 5;
-          nnzPerRow[5] = 4;
-          nnzPerRow[6] = 4;
-          nnzPerRow[7] = 4;
-          nnzPerRow[8] = 3;
-        }
-        col_map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 36, global_colids.view(0, 15), 0, comm);
+      } else if (comm->getRank() == 1) {
+        global_colids[9]  = 2;
+        global_colids[10] = 5;
+        global_colids[11] = 8;
+        global_colids[12] = 27;
+        global_colids[13] = 28;
+        global_colids[14] = 29;
+
+        nnzPerRow[0] = 4;
+        nnzPerRow[1] = 4;
+        nnzPerRow[2] = 3;
+        nnzPerRow[3] = 5;
+        nnzPerRow[4] = 5;
+        nnzPerRow[5] = 4;
+        nnzPerRow[6] = 5;
+        nnzPerRow[7] = 5;
+        nnzPerRow[8] = 4;
+      } else if (comm->getRank() == 2) {
+        global_colids[9]  = 6;
+        global_colids[10] = 7;
+        global_colids[11] = 8;
+        global_colids[12] = 27;
+        global_colids[13] = 30;
+        global_colids[14] = 33;
+
+        nnzPerRow[0] = 4;
+        nnzPerRow[1] = 5;
+        nnzPerRow[2] = 5;
+        nnzPerRow[3] = 4;
+        nnzPerRow[4] = 5;
+        nnzPerRow[5] = 5;
+        nnzPerRow[6] = 3;
+        nnzPerRow[7] = 4;
+        nnzPerRow[8] = 4;
+      } else if (comm->getRank() == 3) {
+        global_colids[9]  = 15;
+        global_colids[10] = 16;
+        global_colids[11] = 17;
+        global_colids[12] = 20;
+        global_colids[13] = 23;
+        global_colids[14] = 26;
+
+        nnzPerRow[0] = 5;
+        nnzPerRow[1] = 5;
+        nnzPerRow[2] = 4;
+        nnzPerRow[3] = 5;
+        nnzPerRow[4] = 5;
+        nnzPerRow[5] = 4;
+        nnzPerRow[6] = 4;
+        nnzPerRow[7] = 4;
+        nnzPerRow[8] = 3;
       }
-      crs_mat = Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(row_map, col_map, nnzPerRow);
+      col_map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 36, global_colids.view(0, 15), 0, comm);
     }
+    crs_mat = Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(row_map, col_map, nnzPerRow);
+  }
 
-    // Some numbers we need...
-    const Scalar zero  = Teuchos::ScalarTraits<Scalar>::zero();
-    const Scalar one   = Teuchos::ScalarTraits<Scalar>::one();
-    const Scalar two   = one + one;
-    const Scalar four  = one + one + one + one;
-    const Scalar six   = four + two;
-    const Scalar seven = six + one;
-    const Scalar eight = seven + one;
-    const Scalar ten   = four + four + two;
-
-    {
-      using size_type = typename ArrayRCP<size_t>::size_type;
-      ArrayRCP<size_t> row_ptr;
-      ArrayRCP<LocalOrdinal> colInds;
-      ArrayRCP<Scalar> values;
-
-      if(comm->getRank() == 0) {
-        size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
-        ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        row_ptr.deepCopy(row_ptr_view.getConst());
-
-        LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
-        ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        colInds.deepCopy(colInds_view.getConst());
-
-        Scalar raw_values[] = {four, -one, -one,
-                               -one, four, -one, -one,
-                               -one, four, -one, -one,
-                               -one, four, -one, -one,
-                               -one, -one / four, four, -one / four, -one,
-                               zero, zero, two, zero, zero,
-                               -one / ten, four / ten, -one / ten, -one / ten,
-                               -one, -one, four, -one, -one,
-                               -one, -one, four, -one, -one};
-        ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        values.deepCopy(values_view.getConst());
-      } else if(comm->getRank() == 1) {
-        size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
-        ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        row_ptr.deepCopy(row_ptr_view.getConst());
-
-        LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
-        ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        colInds.deepCopy(colInds_view.getConst());
-
-        Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
-        ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        values.deepCopy(values_view.getConst());
-      } else if(comm->getRank() == 2) {
-        size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
-        ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        row_ptr.deepCopy(row_ptr_view.getConst());
-
-        LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
-        ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        colInds.deepCopy(colInds_view.getConst());
-
-        Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one,
-                               -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one,
-                               -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
-        ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        values.deepCopy(values_view.getConst());
-      } else if(comm->getRank() == 3) {
-        size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
-        ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        row_ptr.deepCopy(row_ptr_view.getConst());
-
-        LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
-        ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        colInds.deepCopy(colInds_view.getConst());
-
-        Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one,
-                               -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one,
-                               -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
-        ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
-        values.deepCopy(values_view.getConst());
-      }
-      crs_mat->setAllValues(row_ptr, colInds, values);
-    }
+  // Some numbers we need...
+  const Scalar zero  = Teuchos::ScalarTraits<Scalar>::zero();
+  const Scalar one   = Teuchos::ScalarTraits<Scalar>::one();
+  const Scalar two   = one + one;
+  const Scalar four  = one + one + one + one;
+  const Scalar six   = four + two;
+  const Scalar seven = six + one;
+  const Scalar eight = seven + one;
+  const Scalar ten   = four + four + two;
 
-    RCP<Matrix> mat = Teuchos::rcp(new CrsMatrixWrap(crs_mat));
-
-    { // Regular lumped diag
-      out << std::endl << "== regular lumped diagonal ==" << std::endl;
-      RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat);
-      ArrayRCP<Scalar> diag = diagLumped->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-        Scalar diag_ref[] = {six, seven, seven, seven, six + one / two, two, seven / ten, eight, eight};
-        for(int idx = 0; idx < 9; ++idx) {
-          out << std::endl << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
-          TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100*Teuchos::ScalarTraits<Scalar>::eps());
-        }
-      }
+  {
+    using size_type = typename ArrayRCP<size_t>::size_type;
+    ArrayRCP<size_t> row_ptr;
+    ArrayRCP<LocalOrdinal> colInds;
+    ArrayRCP<Scalar> values;
+
+    if (comm->getRank() == 0) {
+      size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
+      ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      row_ptr.deepCopy(row_ptr_view.getConst());
+
+      LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
+      ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      colInds.deepCopy(colInds_view.getConst());
+
+      Scalar raw_values[] = {four, -one, -one,
+                             -one, four, -one, -one,
+                             -one, four, -one, -one,
+                             -one, four, -one, -one,
+                             -one, -one / four, four, -one / four, -one,
+                             zero, zero, two, zero, zero,
+                             -one / ten, four / ten, -one / ten, -one / ten,
+                             -one, -one, four, -one, -one,
+                             -one, -one, four, -one, -one};
+      ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      values.deepCopy(values_view.getConst());
+    } else if (comm->getRank() == 1) {
+      size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
+      ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      row_ptr.deepCopy(row_ptr_view.getConst());
+
+      LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
+      ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      colInds.deepCopy(colInds_view.getConst());
+
+      Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
+      ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      values.deepCopy(values_view.getConst());
+    } else if (comm->getRank() == 2) {
+      size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
+      ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      row_ptr.deepCopy(row_ptr_view.getConst());
+
+      LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
+      ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      colInds.deepCopy(colInds_view.getConst());
+
+      Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one,
+                             -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one,
+                             -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
+      ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      values.deepCopy(values_view.getConst());
+    } else if (comm->getRank() == 3) {
+      size_t raw_row_ptr[] = {0, 3, 7, 11, 15, 20, 25, 29, 34, 39};
+      ArrayView<size_t> row_ptr_view(raw_row_ptr, 10, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      row_ptr.deepCopy(row_ptr_view.getConst());
+
+      LocalOrdinal raw_colInds[] = {0, 1, 3, 0, 1, 2, 4, 1, 2, 5, 9, 0, 3, 4, 6, 1, 3, 4, 5, 7, 2, 4, 5, 8, 10, 3, 6, 7, 12, 4, 6, 7, 8, 13, 5, 7, 8, 10, 14};
+      ArrayView<LocalOrdinal> colInds_view(raw_colInds, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      colInds.deepCopy(colInds_view.getConst());
+
+      Scalar raw_values[] = {four, -one, -one, -one, four, -one, -one, -one, four, -one, -one,
+                             -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one,
+                             -one, four, -one, -one, -one, -one, four, -one, -one, -one, -one, four, -one, -one};
+      ArrayView<Scalar> values_view(raw_values, 39, Teuchos::ERCPNodeLookup::RCP_ENABLE_NODE_LOOKUP);
+      values.deepCopy(values_view.getConst());
     }
+    crs_mat->setAllValues(row_ptr, colInds, values);
+  }
 
-    { // doReciprocal
-      out << std::endl << "== reciprocal lumped diagonal ==" << std::endl;
-      RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true);
-      ArrayRCP<Scalar> diag = diagLumped->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-        Scalar diag_ref[] = {one / six, one / seven, one / seven,
-                             one / seven, one / (six + one / two), one / four,
-                             ten / seven, one / eight, one / eight};
-        for(int idx = 0; idx < 9; ++idx) {
-          out << std::endl << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
-          TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100*Teuchos::ScalarTraits<Scalar>::eps());
-        }
+  RCP<Matrix> mat = Teuchos::rcp(new CrsMatrixWrap(crs_mat));
+
+  {  // Regular lumped diag
+    out << std::endl
+        << "== regular lumped diagonal ==" << std::endl;
+    RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat);
+    ArrayRCP<Scalar> diag  = diagLumped->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      Scalar diag_ref[] = {six, seven, seven, seven, six + one / two, two, seven / ten, eight, eight};
+      for (int idx = 0; idx < 9; ++idx) {
+        out << std::endl
+            << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
+        TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100 * Teuchos::ScalarTraits<Scalar>::eps());
       }
     }
+  }
 
-    { // val < 0.9 treated as 0.0
-      out << std::endl << "== reciprocal lumped diagonal with dropping ==" << std::endl;
-      RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0.9);
-      ArrayRCP<Scalar> diag = diagLumped->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-        Scalar diag_ref[] = {one / six, one / seven, one / seven,
-                             one / seven, one / (six + one / two), one / four,
-                             zero, one / eight, one / eight};
-        for(int idx = 0; idx < 9; ++idx) {
-          out << std::endl << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
-          TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100*Teuchos::ScalarTraits<Scalar>::eps());
-        }
+  {  // doReciprocal
+    out << std::endl
+        << "== reciprocal lumped diagonal ==" << std::endl;
+    RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true);
+    ArrayRCP<Scalar> diag  = diagLumped->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      Scalar diag_ref[] = {one / six, one / seven, one / seven,
+                           one / seven, one / (six + one / two), one / four,
+                           ten / seven, one / eight, one / eight};
+      for (int idx = 0; idx < 9; ++idx) {
+        out << std::endl
+            << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
+        TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100 * Teuchos::ScalarTraits<Scalar>::eps());
       }
     }
+  }
 
-    { // nnzPerRow(i) <= 1 --> diag(i) = zero
-      out << std::endl << "== reciprocal lumped diagonal with single-entry row ==" << std::endl;
-      RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0, 42, true);
-      ArrayRCP<Scalar> diag = diagLumped->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-        Scalar diag_ref[] = {one / six, one / seven, one / seven,
-                             one / seven, one / (six + one / two), zero,
-                             ten / seven, one / eight, one / eight};
-        for(int idx = 0; idx < 9; ++idx) {
-          out << std::endl << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
-          TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100*Teuchos::ScalarTraits<Scalar>::eps());
-        }
+  {  // val < 0.9 treated as 0.0
+    out << std::endl
+        << "== reciprocal lumped diagonal with dropping ==" << std::endl;
+    RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0.9);
+    ArrayRCP<Scalar> diag  = diagLumped->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      Scalar diag_ref[] = {one / six, one / seven, one / seven,
+                           one / seven, one / (six + one / two), one / four,
+                           zero, one / eight, one / eight};
+      for (int idx = 0; idx < 9; ++idx) {
+        out << std::endl
+            << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
+        TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100 * Teuchos::ScalarTraits<Scalar>::eps());
       }
     }
+  }
 
-    { // nnzPerRow(i) <= 1 --> diag(i) = zero
-      out << std::endl << "== reciprocal lumped diagonal with zero-entry row ==" << std::endl;
-      RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0.9, 42, true);
-      ArrayRCP<Scalar> diag = diagLumped->getDataNonConst(0);
-      if(comm->getRank() == 0) {
-        Scalar diag_ref[] = {one / six, one / seven, one / seven,
-                             one / seven, one / (six + one / two), zero,
-                             ten+ten+ten+ten+two, one / eight, one / eight};
-        for(int idx = 0; idx < 9; ++idx) {
-          out << std::endl << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
-          TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100*Teuchos::ScalarTraits<Scalar>::eps());
-        }
+  {  // nnzPerRow(i) <= 1 --> diag(i) = zero
+    out << std::endl
+        << "== reciprocal lumped diagonal with single-entry row ==" << std::endl;
+    RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0, 42, true);
+    ArrayRCP<Scalar> diag  = diagLumped->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      Scalar diag_ref[] = {one / six, one / seven, one / seven,
+                           one / seven, one / (six + one / two), zero,
+                           ten / seven, one / eight, one / eight};
+      for (int idx = 0; idx < 9; ++idx) {
+        out << std::endl
+            << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
+        TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100 * Teuchos::ScalarTraits<Scalar>::eps());
       }
     }
-
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,GetInverse,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
-
-    RCP<Map> m  = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(lib, 100, 0, comm);
-
-    RCP<Vector> v  = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(m, true);
-    RCP<Vector> tv = Xpetra::VectorFactory<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Build(m, true);
-    {
-      Teuchos::ArrayRCP<Scalar> vData  = v->getDataNonConst(0);
-      Teuchos::ArrayRCP<Scalar> tvData = tv->getDataNonConst(0);
-      for(LocalOrdinal i = 0; i < Teuchos::as<LocalOrdinal>(v->getLocalLength()); ++i) {
-        vData[i] = Teuchos::as<Scalar>(i+1);
-        tvData[i] = Teuchos::ScalarTraits<Scalar>::one() / Teuchos::as<Scalar>(i+1);
+  {  // nnzPerRow(i) <= 1 --> diag(i) = zero
+    out << std::endl
+        << "== reciprocal lumped diagonal with zero-entry row ==" << std::endl;
+    RCP<Vector> diagLumped = Utilities::GetLumpedMatrixDiagonal(*mat, true, 0.9, 42, true);
+    ArrayRCP<Scalar> diag  = diagLumped->getDataNonConst(0);
+    if (comm->getRank() == 0) {
+      Scalar diag_ref[] = {one / six, one / seven, one / seven,
+                           one / seven, one / (six + one / two), zero,
+                           ten + ten + ten + ten + two, one / eight, one / eight};
+      for (int idx = 0; idx < 9; ++idx) {
+        out << std::endl
+            << "comparing diag[" << idx << "] and diag_ref[" << idx << "]";
+        TEST_FLOATING_EQUALITY(diag[idx], diag_ref[idx], 100 * Teuchos::ScalarTraits<Scalar>::eps());
       }
     }
-    RCP<Vector> inv = Utilities::GetInverse(v);
-
-    tv->update(1.0,*inv,-1.0);
-    TEST_EQUALITY(tv->norm1(),Teuchos::ScalarTraits<Scalar>::zero());
-    TEST_EQUALITY(tv->norm2(),Teuchos::ScalarTraits<Scalar>::zero());
-    TEST_EQUALITY(tv->normInf(),Teuchos::ScalarTraits<Scalar>::zero());
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,GetThresholdedMatrix,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, GetInverse, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-      using TST = Teuchos::ScalarTraits<SC>;
-      using magnitude_type = typename TST::magnitudeType;
-      using TMT = Teuchos::ScalarTraits<magnitude_type>;
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
 
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
 
-      // Don't test for complex - matrix reader won't work
-      if (TST::isComplex) {success=true; return;}
-      RCP<Matrix> Ain = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("TestMatrices/filter.mm", lib, comm);
+  Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
 
-      RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node>> Aout =
-              MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetThresholdedMatrix(Ain, 1e-5, true, -1);
+  RCP<Map> m = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 100, 0, comm);
 
-      TEST_EQUALITY(Aout->getCrsGraph()->getGlobalNumEntries(), Teuchos::as<size_t>(13));
-      TEST_FLOATING_EQUALITY(Aout->getFrobeniusNorm(), Teuchos::as<magnitude_type>(7.549834435270750), 1e2*Teuchos::ScalarTraits<Scalar>::eps());
+  RCP<Vector> v  = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(m, true);
+  RCP<Vector> tv = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(m, true);
+  {
+    Teuchos::ArrayRCP<Scalar> vData  = v->getDataNonConst(0);
+    Teuchos::ArrayRCP<Scalar> tvData = tv->getDataNonConst(0);
+    for (LocalOrdinal i = 0; i < Teuchos::as<LocalOrdinal>(v->getLocalLength()); ++i) {
+      vData[i]  = Teuchos::as<Scalar>(i + 1);
+      tvData[i] = Teuchos::ScalarTraits<Scalar>::one() / Teuchos::as<Scalar>(i + 1);
+    }
   }
+  RCP<Vector> inv = Utilities::GetInverse(v);
+
+  tv->update(1.0, *inv, -1.0);
+  TEST_EQUALITY(tv->norm1(), Teuchos::ScalarTraits<Scalar>::zero());
+  TEST_EQUALITY(tv->norm2(), Teuchos::ScalarTraits<Scalar>::zero());
+  TEST_EQUALITY(tv->normInf(), Teuchos::ScalarTraits<Scalar>::zero());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, GetThresholdedMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST            = Teuchos::ScalarTraits<SC>;
+  using magnitude_type = typename TST::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // Don't test for complex - matrix reader won't work
+  if (TST::isComplex) {
+    success = true;
+    return;
+  }
+  RCP<Matrix> Ain = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("TestMatrices/filter.mm", lib, comm);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,GetThresholdedGraph,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
+  RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>> Aout =
+      MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThresholdedMatrix(Ain, 1e-5, true, -1);
 
-      using TST = Teuchos::ScalarTraits<SC>;
-      using magnitude_type = typename TST::magnitudeType;
-      using TMT = Teuchos::ScalarTraits<magnitude_type>;
+  TEST_EQUALITY(Aout->getCrsGraph()->getGlobalNumEntries(), Teuchos::as<size_t>(13));
+  TEST_FLOATING_EQUALITY(Aout->getFrobeniusNorm(), Teuchos::as<magnitude_type>(7.549834435270750), 1e2 * Teuchos::ScalarTraits<Scalar>::eps());
+}
 
-      RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-      Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, GetThresholdedGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-      // Don't test for complex - matrix reader won't work
-      if (TST::isComplex) {success=true; return;}
-      RCP<Matrix> A = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("TestMatrices/filter.mm", lib, comm);
+  using TST            = Teuchos::ScalarTraits<SC>;
+  using magnitude_type = typename TST::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-      RCP<Xpetra::CrsGraph<LocalOrdinal,GlobalOrdinal,Node>> graph =
-              MueLu::Utilities<Scalar,LocalOrdinal,GlobalOrdinal,Node>::GetThresholdedGraph(A, 1e-5, -1);
+  RCP<const Teuchos::Comm<int>> comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib          = MueLuTests::TestHelpers::Parameters::getLib();
 
-      TEST_EQUALITY(graph->getGlobalNumEntries(), Teuchos::as<size_t>(13));
+  // Don't test for complex - matrix reader won't work
+  if (TST::isComplex) {
+    success = true;
+    return;
   }
+  RCP<Matrix> A = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("TestMatrices/filter.mm", lib, comm);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities,TransposeNonsymmetricConstMatrix,Scalar,LocalOrdinal,GlobalOrdinal,Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers::Parameters::getLib();
+  RCP<Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node>> graph =
+      MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::GetThresholdedGraph(A, 1e-5, -1);
 
-    const GO numGlobalElements = 29;
-    RCP<Map> dofMap  = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(lib, numGlobalElements, 0, comm);
+  TEST_EQUALITY(graph->getGlobalNumEntries(), Teuchos::as<size_t>(13));
+}
 
-    TEST_ASSERT(!dofMap.is_null());
-    TEST_EQUALITY_CONST(dofMap->getGlobalNumElements(), numGlobalElements);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities, TransposeNonsymmetricConstMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    RCP<const Matrix> matrix = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 1.0, 2.0, 3.0, lib);
-    TEST_ASSERT(!matrix.is_null());
-    TEST_EQUALITY_CONST(matrix->getGlobalNumRows(), numGlobalElements);
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib          = TestHelpers::Parameters::getLib();
 
-    RCP<const Matrix> transposedMatrix = Utilities::Transpose(const_cast<Matrix&>(*matrix));
-    TEST_ASSERT(!transposedMatrix.is_null());
+  const GO numGlobalElements = 29;
+  RCP<Map> dofMap            = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, numGlobalElements, 0, comm);
 
-    TEST_ASSERT(transposedMatrix->getRangeMap()->isSameAs(*matrix->getDomainMap()));
-    TEST_ASSERT(transposedMatrix->getDomainMap()->isSameAs(*matrix->getRangeMap()));
+  TEST_ASSERT(!dofMap.is_null());
+  TEST_EQUALITY_CONST(dofMap->getGlobalNumElements(), numGlobalElements);
 
-    // Verify, that A^T actually differs from A
-    {
-      RCP<Matrix> diffMatrix = rcp(new CrsMatrixWrap(matrix->getCrsGraph()));
-      MatrixMatrix::TwoMatrixAdd(*matrix, false, 1.0, *transposedMatrix, false, -1.0, diffMatrix, out);
-      diffMatrix->fillComplete();
-      TEST_ASSERT(!diffMatrix.is_null());
+  RCP<const Matrix> matrix = TestHelpers::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 1.0, 2.0, 3.0, lib);
+  TEST_ASSERT(!matrix.is_null());
+  TEST_EQUALITY_CONST(matrix->getGlobalNumRows(), numGlobalElements);
 
-      bool allEntriesAreZero = true;
-      for (LO lRowId = 0; lRowId < Teuchos::as<LO>(diffMatrix->getLocalNumRows()); ++lRowId)
-      {
-        ArrayView<const LO> cols;
-        ArrayView<const Scalar> vals;
-        diffMatrix->getLocalRowView(lRowId, cols, vals);
+  RCP<const Matrix> transposedMatrix = Utilities::Transpose(const_cast<Matrix&>(*matrix));
+  TEST_ASSERT(!transposedMatrix.is_null());
 
-        TEST_INEQUALITY_CONST(cols.size(), 0);
-        TEST_INEQUALITY_CONST(vals.size(), 0);
+  TEST_ASSERT(transposedMatrix->getRangeMap()->isSameAs(*matrix->getDomainMap()));
+  TEST_ASSERT(transposedMatrix->getDomainMap()->isSameAs(*matrix->getRangeMap()));
 
-        for (const auto& entry : vals) {
-          if (entry != Teuchos::ScalarTraits<Scalar>::zero()) allEntriesAreZero = false;
-        }
+  // Verify, that A^T actually differs from A
+  {
+    RCP<Matrix> diffMatrix = rcp(new CrsMatrixWrap(matrix->getCrsGraph()));
+    MatrixMatrix::TwoMatrixAdd(*matrix, false, 1.0, *transposedMatrix, false, -1.0, diffMatrix, out);
+    diffMatrix->fillComplete();
+    TEST_ASSERT(!diffMatrix.is_null());
+
+    bool allEntriesAreZero = true;
+    for (LO lRowId = 0; lRowId < Teuchos::as<LO>(diffMatrix->getLocalNumRows()); ++lRowId) {
+      ArrayView<const LO> cols;
+      ArrayView<const Scalar> vals;
+      diffMatrix->getLocalRowView(lRowId, cols, vals);
+
+      TEST_INEQUALITY_CONST(cols.size(), 0);
+      TEST_INEQUALITY_CONST(vals.size(), 0);
+
+      for (const auto& entry : vals) {
+        if (entry != Teuchos::ScalarTraits<Scalar>::zero()) allEntriesAreZero = false;
       }
-      TEST_ASSERT(!allEntriesAreZero);
     }
+    TEST_ASSERT(!allEntriesAreZero);
+  }
 
-    RCP<const Matrix> doubleTransposedMatrix = Utilities::Transpose(const_cast<Matrix&>(*transposedMatrix));
-    TEST_ASSERT(!doubleTransposedMatrix.is_null());
-
-    TEST_ASSERT(doubleTransposedMatrix->getRangeMap()->isSameAs(*matrix->getRangeMap()));
-    TEST_ASSERT(doubleTransposedMatrix->getDomainMap()->isSameAs(*matrix->getDomainMap()));
-
-    // Transpose twice: A - (A^T)^T needs to be the zero matrix
-    {
-      RCP<Matrix> diffMatrix = rcp(new CrsMatrixWrap(matrix->getCrsGraph()));
-      MatrixMatrix::TwoMatrixAdd(*matrix, false, 1.0, *doubleTransposedMatrix, false, -1.0, diffMatrix, out);
-      diffMatrix->fillComplete();
-      TEST_ASSERT(!diffMatrix.is_null());
-
-      bool allEntriesAreZero = true;
-      for (LO lRowId = 0; lRowId < Teuchos::as<LO>(diffMatrix->getLocalNumRows()); ++lRowId)
-      {
-        ArrayView<const LO> cols;
-        ArrayView<const Scalar> vals;
-        diffMatrix->getLocalRowView(lRowId, cols, vals);
+  RCP<const Matrix> doubleTransposedMatrix = Utilities::Transpose(const_cast<Matrix&>(*transposedMatrix));
+  TEST_ASSERT(!doubleTransposedMatrix.is_null());
 
-        TEST_INEQUALITY_CONST(cols.size(), 0);
-        TEST_INEQUALITY_CONST(vals.size(), 0);
+  TEST_ASSERT(doubleTransposedMatrix->getRangeMap()->isSameAs(*matrix->getRangeMap()));
+  TEST_ASSERT(doubleTransposedMatrix->getDomainMap()->isSameAs(*matrix->getDomainMap()));
 
-        for (const auto& entry : vals) {
-          if (entry != Teuchos::ScalarTraits<Scalar>::zero()) allEntriesAreZero = false;
-        }
+  // Transpose twice: A - (A^T)^T needs to be the zero matrix
+  {
+    RCP<Matrix> diffMatrix = rcp(new CrsMatrixWrap(matrix->getCrsGraph()));
+    MatrixMatrix::TwoMatrixAdd(*matrix, false, 1.0, *doubleTransposedMatrix, false, -1.0, diffMatrix, out);
+    diffMatrix->fillComplete();
+    TEST_ASSERT(!diffMatrix.is_null());
+
+    bool allEntriesAreZero = true;
+    for (LO lRowId = 0; lRowId < Teuchos::as<LO>(diffMatrix->getLocalNumRows()); ++lRowId) {
+      ArrayView<const LO> cols;
+      ArrayView<const Scalar> vals;
+      diffMatrix->getLocalRowView(lRowId, cols, vals);
+
+      TEST_INEQUALITY_CONST(cols.size(), 0);
+      TEST_INEQUALITY_CONST(vals.size(), 0);
+
+      for (const auto& entry : vals) {
+        if (entry != Teuchos::ScalarTraits<Scalar>::zero()) allEntriesAreZero = false;
       }
-      TEST_ASSERT(allEntriesAreZero);
     }
+    TEST_ASSERT(allEntriesAreZero);
   }
-
-#define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,MatMatMult_EpetraVsTpetra,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,DetectDirichletRows,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,GetDiagonalInverse,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,GetLumpedDiagonal,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,GetInverse,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,GetThresholdedMatrix,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,GetThresholdedGraph,Scalar,LO,GO,Node) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities,TransposeNonsymmetricConstMatrix,Scalar,LO,GO,Node)
+}
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, MatMatMult_EpetraVsTpetra, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, DetectDirichletRows, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, GetDiagonalInverse, Scalar, LO, GO, Node)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, GetLumpedDiagonal, Scalar, LO, GO, Node)         \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, GetInverse, Scalar, LO, GO, Node)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, GetThresholdedMatrix, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, GetThresholdedGraph, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities, TransposeNonsymmetricConstMatrix, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/VariableContainer.cpp b/packages/muelu/test/unit_tests/VariableContainer.cpp
index 37d3fdf0ab3f..f65167de5db4 100644
--- a/packages/muelu/test/unit_tests/VariableContainer.cpp
+++ b/packages/muelu/test/unit_tests/VariableContainer.cpp
@@ -50,81 +50,78 @@
 
 namespace MueLuTests {
 
-  using MueLu::VariableContainer;
-  using MueLu::KeepType;
+using MueLu::KeepType;
+using MueLu::VariableContainer;
 
-  TEUCHOS_UNIT_TEST(VariableContainer, Constructor)
-  {
-    VariableContainer vc;
+TEUCHOS_UNIT_TEST(VariableContainer, Constructor) {
+  VariableContainer vc;
 
-    KeepType keepFlag = vc.GetKeepFlag();
-    TEST_EQUALITY(keepFlag, 0);
-  }
+  KeepType keepFlag = vc.GetKeepFlag();
+  TEST_EQUALITY(keepFlag, 0);
+}
 
-  TEUCHOS_UNIT_TEST(VariableContainer, AddKeepFlag_RemoveKeepFlag)
-  {
-    VariableContainer vc;
+TEUCHOS_UNIT_TEST(VariableContainer, AddKeepFlag_RemoveKeepFlag) {
+  VariableContainer vc;
 
-    // Test: default input parameter is UserData
-    vc.AddKeepFlag();
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
-    vc.RemoveKeepFlag();
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: default input parameter is UserData
+  vc.AddKeepFlag();
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
+  vc.RemoveKeepFlag();
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    // Test: add and remove one flag
-    vc.AddKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep);
-    vc.RemoveKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: add and remove one flag
+  vc.AddKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep);
+  vc.RemoveKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    // Test: add and remove are using bitwise operations (| ^ instead of + -)
-    // Test of add
-    vc.AddKeepFlag(MueLu::Keep);    // 2 add
-    vc.AddKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep); // failed if AddKeepFlag uses + instead of |
-    vc.RemoveKeepFlag(MueLu::Keep); // 1 remove
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
-    // Test of remove (removing something twice or something that do not exist)
-    vc.AddKeepFlag(MueLu::Keep);    // 1 add
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep);
-    vc.RemoveKeepFlag(MueLu::Keep); // 2 remove
-    vc.RemoveKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: add and remove are using bitwise operations (| ^ instead of + -)
+  // Test of add
+  vc.AddKeepFlag(MueLu::Keep);  // 2 add
+  vc.AddKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep);  // failed if AddKeepFlag uses + instead of |
+  vc.RemoveKeepFlag(MueLu::Keep);                // 1 remove
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test of remove (removing something twice or something that do not exist)
+  vc.AddKeepFlag(MueLu::Keep);  // 1 add
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Keep);
+  vc.RemoveKeepFlag(MueLu::Keep);  // 2 remove
+  vc.RemoveKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    // Test: internal flag can be a combinaison of flags.
-    vc.AddKeepFlag(MueLu::UserData);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
-    vc.AddKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep);
-    vc.AddKeepFlag(MueLu::Final);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep | MueLu::Final);
-    vc.RemoveKeepFlag(MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Final);
-    vc.RemoveKeepFlag(MueLu::UserData);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Final);
-    vc.RemoveKeepFlag(MueLu::Final);
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: internal flag can be a combinaison of flags.
+  vc.AddKeepFlag(MueLu::UserData);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
+  vc.AddKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep);
+  vc.AddKeepFlag(MueLu::Final);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep | MueLu::Final);
+  vc.RemoveKeepFlag(MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Final);
+  vc.RemoveKeepFlag(MueLu::UserData);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::Final);
+  vc.RemoveKeepFlag(MueLu::Final);
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    // Test: input of add and remove can be a combinaison of flags
-    vc.AddKeepFlag(MueLu::UserData | MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep);
-    vc.RemoveKeepFlag(MueLu::UserData | MueLu::Keep);
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: input of add and remove can be a combinaison of flags
+  vc.AddKeepFlag(MueLu::UserData | MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData | MueLu::Keep);
+  vc.RemoveKeepFlag(MueLu::UserData | MueLu::Keep);
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    // Test: MueLu::All == MueLu::UserData | MueLu::Keep | MueLu::Final
-    TEST_EQUALITY( MueLu::UserData | MueLu::Keep | MueLu::Final, MueLu::All);
+  // Test: MueLu::All == MueLu::UserData | MueLu::Keep | MueLu::Final
+  TEST_EQUALITY(MueLu::UserData | MueLu::Keep | MueLu::Final, MueLu::All);
 
-    // Test: a very common use case: set one flag, remove all flags
-    // (this is what is actually done by Level::Keep() followed by Level::Delete())
-    vc.AddKeepFlag(MueLu::UserData);
-    TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
-    vc.RemoveKeepFlag(MueLu::All);
-    TEST_EQUALITY(vc.GetKeepFlag(), 0);
+  // Test: a very common use case: set one flag, remove all flags
+  // (this is what is actually done by Level::Keep() followed by Level::Delete())
+  vc.AddKeepFlag(MueLu::UserData);
+  TEST_EQUALITY(vc.GetKeepFlag(), MueLu::UserData);
+  vc.RemoveKeepFlag(MueLu::All);
+  TEST_EQUALITY(vc.GetKeepFlag(), 0);
 
-    //TODO: Test: add remove check if input argument is a valid flag
-    //    TEST_THROW(vc.AddKeepFlag(42))
-    //    TEST_THROW(vc.DeleteKeepFlag(42))
+  // TODO: Test: add remove check if input argument is a valid flag
+  //     TEST_THROW(vc.AddKeepFlag(42))
+  //     TEST_THROW(vc.DeleteKeepFlag(42))
+}
 
-  }
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/VariableDofLaplacianFactory.cpp b/packages/muelu/test/unit_tests/VariableDofLaplacianFactory.cpp
index 19421aee52f8..a734ea7a0480 100644
--- a/packages/muelu/test/unit_tests/VariableDofLaplacianFactory.cpp
+++ b/packages/muelu/test/unit_tests/VariableDofLaplacianFactory.cpp
@@ -55,259 +55,271 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (!TYPE_EQUAL(GO, int)) { out << "Skipping test for GO != int"        << std::endl; return; }
-    if (!TYPE_EQUAL(SC, double)) { out << "Skipping test for SC != double; no support for complex SC in Galeri::Xpetra branch" << std::endl; return; }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    GlobalOrdinal nEle = 64;
-    const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
-    Teuchos::ParameterList matrixParameters;
-    matrixParameters.set("nx",8);
-    matrixParameters.set("ny",8);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-      Galeri::Xpetra::BuildProblem<Scalar,LocalOrdinal,GlobalOrdinal,Map,CrsMatrixWrap,MultiVector>("Laplace2D", map, matrixParameters);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    // TODO coords should use SC = double
-    RCP<RealValuedMultiVector> coords = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,RealValuedMultiVector>("2D", A->getRowMap(), matrixParameters);
-
-    // build hierarchy
-    RCP<Level> l = rcp(new Level());
-    l->SetLevelID(0);
-    l->SetComm(comm);
-    l->Set("A", A);
-    l->Set("Coordinates",coords);
-
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(),1);
-    l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
-
-    VariableDofLaplacianFactory lapFact;
-
-    l->Request("A",&lapFact);
-
-    RCP<Matrix> lapA = l->Get<RCP<Matrix> >("A",&lapFact);
-
-    //lapA->describe(out, Teuchos::VERB_EXTREME);
-
-    TEST_EQUALITY(lapA->getRowMap()->isSameAs(*A->getRowMap()),true);
-
-    Teuchos::RCP<Vector> oneVec = VectorFactory::Build(A->getRowMap());
-    Teuchos::RCP<Vector> res = VectorFactory::Build(A->getRowMap());
-    oneVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    res->putScalar(Teuchos::as<Scalar>(27)*Teuchos::ScalarTraits<Scalar>::one());
-    Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA)->apply(*oneVec,*res);
-    TEST_COMPARE(res->normInf(),<, 100*TMT::eps());
-  } // VarLaplConstructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplConstructor2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (!TYPE_EQUAL(GO, int)) { out << "Skipping test for GO != int"        << std::endl; return; }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using real_type             = typename TST::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
-
-    // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
-    if ( comm->getSize() > 2 && lib == Xpetra::UseEpetra) { out << "Skipping test for more than 2 procs when using Epetra"        << std::endl; return; }
-
-    GlobalOrdinal nx = 6, ny = 6;
-
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
-
-    //build coordinates before expanding map (nodal coordinates, not dof-based)
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LocalOrdinal,GlobalOrdinal,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
-    RCP<const Map> dofMap = MapFactory::Build(nodeMap, 2); //expand map for 2 DOFs per node
-
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplConstructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (!TYPE_EQUAL(GO, int)) {
+    out << "Skipping test for GO != int" << std::endl;
+    return;
+  }
+  if (!TYPE_EQUAL(SC, double)) {
+    out << "Skipping test for SC != double; no support for complex SC in Galeri::Xpetra branch" << std::endl;
+    return;
+  }
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  GlobalOrdinal nEle       = 64;
+  const RCP<const Map> map = MapFactory::Build(lib, nEle, 0, comm);
+  Teuchos::ParameterList matrixParameters;
+  matrixParameters.set("nx", 8);
+  matrixParameters.set("ny", 8);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+      Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Laplace2D", map, matrixParameters);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  // TODO coords should use SC = double
+  RCP<RealValuedMultiVector> coords = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, RealValuedMultiVector>("2D", A->getRowMap(), matrixParameters);
+
+  // build hierarchy
+  RCP<Level> l = rcp(new Level());
+  l->SetLevelID(0);
+  l->SetComm(comm);
+  l->Set("A", A);
+  l->Set("Coordinates", coords);
+
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(), 1);
+  l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
+
+  VariableDofLaplacianFactory lapFact;
+
+  l->Request("A", &lapFact);
+
+  RCP<Matrix> lapA = l->Get<RCP<Matrix> >("A", &lapFact);
+
+  // lapA->describe(out, Teuchos::VERB_EXTREME);
+
+  TEST_EQUALITY(lapA->getRowMap()->isSameAs(*A->getRowMap()), true);
+
+  Teuchos::RCP<Vector> oneVec = VectorFactory::Build(A->getRowMap());
+  Teuchos::RCP<Vector> res    = VectorFactory::Build(A->getRowMap());
+  oneVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+  res->putScalar(Teuchos::as<Scalar>(27) * Teuchos::ScalarTraits<Scalar>::one());
+  Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA)->apply(*oneVec, *res);
+  TEST_COMPARE(res->normInf(), <, 100 * TMT::eps());
+}  // VarLaplConstructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplConstructor2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (!TYPE_EQUAL(GO, int)) {
+    out << "Skipping test for GO != int" << std::endl;
+    return;
+  }
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using real_type             = typename TST::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+
+  // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
+  if (comm->getSize() > 2 && lib == Xpetra::UseEpetra) {
+    out << "Skipping test for more than 2 procs when using Epetra" << std::endl;
+    return;
+  }
+
+  GlobalOrdinal nx = 6, ny = 6;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
+
+  // build coordinates before expanding map (nodal coordinates, not dof-based)
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LocalOrdinal, GlobalOrdinal, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
+  RCP<const Map> dofMap                  = MapFactory::Build(nodeMap, 2);  // expand map for 2 DOFs per node
+
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", dofMap, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
 
-    //->describe(out, Teuchos::VERB_EXTREME);
+  //->describe(out, Teuchos::VERB_EXTREME);
 
-    TEST_EQUALITY(dofMap->getLocalNumElements(),2*nodeMap->getLocalNumElements());
+  TEST_EQUALITY(dofMap->getLocalNumElements(), 2 * nodeMap->getLocalNumElements());
 
-    // build hierarchy
-    RCP<Level> l = rcp(new Level());
-    l->SetLevelID(0);
-    l->SetComm(comm);
-    l->Set("A", A);
-    l->Set("Coordinates",coordinates);
+  // build hierarchy
+  RCP<Level> l = rcp(new Level());
+  l->SetLevelID(0);
+  l->SetComm(comm);
+  l->Set("A", A);
+  l->Set("Coordinates", coordinates);
 
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(),1);
-    l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(), 1);
+  l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
 
-    //A->getColMap()->describe(out,Teuchos::VERB_EXTREME);
+  // A->getColMap()->describe(out,Teuchos::VERB_EXTREME);
 
-    VariableDofLaplacianFactory lapFact;
-    lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(2));
-    l->Request("A",&lapFact);
+  VariableDofLaplacianFactory lapFact;
+  lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(2));
+  l->Request("A", &lapFact);
 
-    RCP<Matrix> lapA = l->Get<RCP<Matrix> >("A",&lapFact);
+  RCP<Matrix> lapA = l->Get<RCP<Matrix> >("A", &lapFact);
 
-    //lapA->describe(out, Teuchos::VERB_EXTREME);
+  // lapA->describe(out, Teuchos::VERB_EXTREME);
 
-    Teuchos::RCP<Vector> oneVec = VectorFactory::Build(lapA->getRowMap());
-    Teuchos::RCP<Vector> res = VectorFactory::Build(lapA->getRowMap());
-    oneVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    res->putScalar(Teuchos::as<Scalar>(27)*Teuchos::ScalarTraits<Scalar>::one());
-    Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA)->apply(*oneVec,*res);
-    TEST_COMPARE(res->normInf(),<, 100*TMT::eps());
+  Teuchos::RCP<Vector> oneVec = VectorFactory::Build(lapA->getRowMap());
+  Teuchos::RCP<Vector> res    = VectorFactory::Build(lapA->getRowMap());
+  oneVec->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+  res->putScalar(Teuchos::as<Scalar>(27) * Teuchos::ScalarTraits<Scalar>::one());
+  Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA)->apply(*oneVec, *res);
+  TEST_COMPARE(res->normInf(), <, 100 * TMT::eps());
 
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent2(3 * lapA->getRowMap()->getLocalNumElements(),1);
-    for(decltype(dofPresent2.size()) i = 2; i < dofPresent2.size(); i = i+3) {
-      dofPresent2[i] = 0;
-    }
-    l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent2);
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent2(3 * lapA->getRowMap()->getLocalNumElements(), 1);
+  for (decltype(dofPresent2.size()) i = 2; i < dofPresent2.size(); i = i + 3) {
+    dofPresent2[i] = 0;
+  }
+  l->Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent2);
 
-    //A->getColMap()->describe(out,Teuchos::VERB_EXTREME);
+  // A->getColMap()->describe(out,Teuchos::VERB_EXTREME);
 
-    VariableDofLaplacianFactory lapFact2;
-    lapFact2.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(3));
-    l->Request("A",&lapFact2);
+  VariableDofLaplacianFactory lapFact2;
+  lapFact2.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(3));
+  l->Request("A", &lapFact2);
 
-    RCP<Matrix> lapA2 = l->Get<RCP<Matrix> >("A",&lapFact2);
+  RCP<Matrix> lapA2 = l->Get<RCP<Matrix> >("A", &lapFact2);
 
-    Teuchos::RCP<Vector> oneVec2 = VectorFactory::Build(lapA2->getRowMap());
-    Teuchos::RCP<Vector> res2 = VectorFactory::Build(lapA2->getRowMap());
-    oneVec2->putScalar(Teuchos::ScalarTraits<Scalar>::one());
-    res2->putScalar(Teuchos::as<Scalar>(27)*Teuchos::ScalarTraits<Scalar>::one());
-    Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA2)->apply(*oneVec2,*res2);
-    TEST_COMPARE(res2->normInf(),<, 100*TMT::eps());
-    TEST_EQUALITY(res->getMap()->isSameAs(*(res2->getMap())),true);
+  Teuchos::RCP<Vector> oneVec2 = VectorFactory::Build(lapA2->getRowMap());
+  Teuchos::RCP<Vector> res2    = VectorFactory::Build(lapA2->getRowMap());
+  oneVec2->putScalar(Teuchos::ScalarTraits<Scalar>::one());
+  res2->putScalar(Teuchos::as<Scalar>(27) * Teuchos::ScalarTraits<Scalar>::one());
+  Teuchos::rcp_dynamic_cast<CrsMatrixWrap>(lapA2)->apply(*oneVec2, *res2);
+  TEST_COMPARE(res2->normInf(), <, 100 * TMT::eps());
+  TEST_EQUALITY(res->getMap()->isSameAs(*(res2->getMap())), true);
 
-    //lapA2->describe(out, Teuchos::VERB_EXTREME);
-  } // VarLaplConstructor2
+  // lapA2->describe(out, Teuchos::VERB_EXTREME);
+}  // VarLaplConstructor2
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplPtent, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (!TYPE_EQUAL(GO, int)) { out << "Skipping test for GO != int"        << std::endl; return; }
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(VariableDofLaplacianFactory, VarLaplPtent, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (!TYPE_EQUAL(GO, int)) {
+    out << "Skipping test for GO != int" << std::endl;
+    return;
+  }
+  out << "version: " << MueLu::Version() << std::endl;
 
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
 
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
+  Xpetra::UnderlyingLib lib = MueLuTests::TestHelpers::Parameters::getLib();
 
-    // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
-    if ( comm->getSize() > 2 && lib == Xpetra::UseEpetra) { out << "Skipping test for more than 2 procs when using Epetra"        << std::endl; return; }
+  // TAW 04/21: test is crashing on 4 procs with Epetra due to an unknown reason in the Epetra_BlockMap constructor (MPI communication)
+  if (comm->getSize() > 2 && lib == Xpetra::UseEpetra) {
+    out << "Skipping test for more than 2 procs when using Epetra" << std::endl;
+    return;
+  }
 
-    GlobalOrdinal nx = 6, ny = 6;
+  GlobalOrdinal nx = 6, ny = 6;
 
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType,LocalOrdinal,GlobalOrdinal,Node> mv_type_double;
-    typedef Xpetra::MultiVectorFactory<double,LocalOrdinal,GlobalOrdinal,Node> MVFactory_double;
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> mv_type_double;
+  typedef Xpetra::MultiVectorFactory<double, LocalOrdinal, GlobalOrdinal, Node> MVFactory_double;
 
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Map> nodeMap = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(lib, "Cartesian2D", comm, galeriList);
 
-    //build coordinates before expanding map (nodal coordinates, not dof-based)
-    RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LocalOrdinal,GlobalOrdinal,Map,RealValuedMultiVector>("2D", nodeMap, galeriList);
-    RCP<const Map> dofMap = MapFactory::Build(nodeMap, 2); //expand map for 2 DOFs per node
+  // build coordinates before expanding map (nodal coordinates, not dof-based)
+  RCP<RealValuedMultiVector> coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LocalOrdinal, GlobalOrdinal, Map, RealValuedMultiVector>("2D", nodeMap, galeriList);
+  RCP<const Map> dofMap                  = MapFactory::Build(nodeMap, 2);  // expand map for 2 DOFs per node
 
-    galeriList.set("right boundary" , "Neumann");
-    galeriList.set("bottom boundary", "Neumann");
-    galeriList.set("top boundary"   , "Neumann");
-    galeriList.set("front boundary" , "Neumann");
-    galeriList.set("back boundary"  , "Neumann");
-    galeriList.set("keepBCs",             false);
+  galeriList.set("right boundary", "Neumann");
+  galeriList.set("bottom boundary", "Neumann");
+  galeriList.set("top boundary", "Neumann");
+  galeriList.set("front boundary", "Neumann");
+  galeriList.set("back boundary", "Neumann");
+  galeriList.set("keepBCs", false);
 
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", dofMap, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
 
-    //->describe(out, Teuchos::VERB_EXTREME);
+  //->describe(out, Teuchos::VERB_EXTREME);
 
-    TEST_EQUALITY(dofMap->getLocalNumElements(),2*nodeMap->getLocalNumElements());
+  TEST_EQUALITY(dofMap->getLocalNumElements(), 2 * nodeMap->getLocalNumElements());
 
-    Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(),1);
+  Teuchos::ArrayRCP<LocalOrdinal> dofPresent(A->getRowMap()->getLocalNumElements(), 1);
 
-    // build hierarchy
-    typedef Teuchos::ScalarTraits<Scalar> TST;
-    typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  // build hierarchy
+  typedef Teuchos::ScalarTraits<Scalar> TST;
+  typedef TestHelpers::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
 
-    // generate laplacian matrix using level l
-    Level l;
-    l.Set("A", A);
-    l.Set("Coordinates",coordinates);
-    l.Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
+  // generate laplacian matrix using level l
+  Level l;
+  l.Set("A", A);
+  l.Set("Coordinates", coordinates);
+  l.Set<Teuchos::ArrayRCP<LocalOrdinal> >("DofPresent", dofPresent);
 
-    VariableDofLaplacianFactory lapFact;
-    lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(2));
-    l.Request("A",&lapFact);
+  VariableDofLaplacianFactory lapFact;
+  lapFact.SetParameter("maxDofPerNode", Teuchos::ParameterEntry(2));
+  l.Request("A", &lapFact);
 
-    RCP<Matrix> lapA = l.Get<RCP<Matrix> >("A",&lapFact);
-    lapA->describe(out, Teuchos::VERB_EXTREME);
+  RCP<Matrix> lapA = l.Get<RCP<Matrix> >("A", &lapFact);
+  lapA->describe(out, Teuchos::VERB_EXTREME);
 
-    // create a new two-level hierarchy and set lapA as input "A"
-    // This way we avoid dealing with factory managers
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  // create a new two-level hierarchy and set lapA as input "A"
+  // This way we avoid dealing with factory managers
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    fineLevel.Set("Coordinates", coordinates);
-    fineLevel.Set("A", lapA);
+  fineLevel.Set("Coordinates", coordinates);
+  fineLevel.Set("A", lapA);
 
-    TentativePFactory PFact;
+  TentativePFactory PFact;
 
-    coarseLevel.Request("P", &PFact);
-    RCP<Matrix> pMat = coarseLevel.Get<RCP<Matrix> >("P", &PFact);
+  coarseLevel.Request("P", &PFact);
+  RCP<Matrix> pMat = coarseLevel.Get<RCP<Matrix> >("P", &PFact);
 
-    TEST_EQUALITY(pMat->getRowMap()->isSameAs(*(lapA->getRowMap())),true);
-    TEST_EQUALITY(pMat->getLocalNumEntries(), pMat->getRowMap()->getLocalNumElements());
-  } // VarLaplPtent
+  TEST_EQUALITY(pMat->getRowMap()->isSameAs(*(lapA->getRowMap())), true);
+  TEST_EQUALITY(pMat->getLocalNumEntries(), pMat->getRowMap()->getLocalNumElements());
+}  // VarLaplPtent
 
-#  define MUELU_ETI_GROUP(SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplConstructor, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplConstructor2, SC, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplPtent, SC, LO, GO, Node) \
+#define MUELU_ETI_GROUP(SC, LO, GO, Node)                                                                  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplConstructor, SC, LO, GO, Node)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplConstructor2, SC, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(VariableDofLaplacianFactory, VarLaplPtent, SC, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
-}
-
-
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/Zoltan.cpp b/packages/muelu/test/unit_tests/Zoltan.cpp
index 26f15b02373f..1f11824aca03 100644
--- a/packages/muelu/test/unit_tests/Zoltan.cpp
+++ b/packages/muelu/test/unit_tests/Zoltan.cpp
@@ -56,550 +56,551 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Zoltan, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-    RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-    TEST_EQUALITY(zoltan != Teuchos::null, true);
-
-  } //Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Zoltan, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Zoltan, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+  RCP<ZoltanInterface> zoltan         = rcp(new ZoltanInterface());
+  TEST_EQUALITY(zoltan != Teuchos::null, true);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Zoltan, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << std::endl;
+  out << "This tests that the partitioning produced by Zoltan is \"reasonable\" for a matrix" << std::endl;
+  out << "that has a random number of nonzeros per row.  Good results have been precomputed" << std::endl;
+  out << "for up to 5 processors.  The results are the number of nonzeros in the local matrix" << std::endl;
+  out << "once the Zoltan repartitioning has been applied." << std::endl;
+  out << "The results can be viewed in Paraview by enabling code guarded by the macro MUELU_VISUALIZE_REPARTITIONING" << std::endl;
+  typedef Teuchos::ScalarTraits<Scalar> ST;
+
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
+
+  if (comm->getSize() > 5) {
     out << std::endl;
-    out << "This tests that the partitioning produced by Zoltan is \"reasonable\" for a matrix" << std::endl;
-    out << "that has a random number of nonzeros per row.  Good results have been precomputed" << std::endl;
-    out << "for up to 5 processors.  The results are the number of nonzeros in the local matrix" << std::endl;
-    out << "once the Zoltan repartitioning has been applied." << std::endl;
-    out << "The results can be viewed in Paraview by enabling code guarded by the macro MUELU_VISUALIZE_REPARTITIONING" << std::endl;
-    typedef Teuchos::ScalarTraits<Scalar> ST;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    if (comm->getSize() > 5) {
-      out << std::endl;
-      out << "This test must be run on 1 to 5 processes." << std::endl;
-      TEST_EQUALITY(true, true);
-      return;
+    out << "This test must be run on 1 to 5 processes." << std::endl;
+    TEST_EQUALITY(true, true);
+    return;
+  }
+
+  Level level;
+  RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+  level.SetFactoryManager(factoryHandler);
+  GO nx                                   = 7;
+  GO ny                                   = nx;
+  Xpetra::global_size_t numGlobalElements = nx * ny;
+  size_t maxEntriesPerRow                 = 30;
+
+  // Populate CrsMatrix with random number of entries (up to maxEntriesPerRow) per row.
+  RCP<const Map> map                                       = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, 0 /*indexBase*/, comm);
+  const size_t numMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = map->getLocalElementList();
+  // memory on the fly.  While not super efficient, this
+  // ensures that no zeros are being stored.  Thus, from
+  // Zoltan's perspective the matrix is imbalanced.
+  // Create a vector with random integer entries in [1,maxEntriesPerRow].
+  ST::seedrandom(666 * comm->getRank());
+  LO maxNnzPerRow                                    = 0;
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > entriesPerRow = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(map, false);
+  Teuchos::ArrayRCP<LO> eprData                      = entriesPerRow->getDataNonConst(0);
+  for (typename Teuchos::ArrayRCP<LO>::iterator i = eprData.begin(); i != eprData.end(); ++i) {
+    *i = static_cast<LO>(std::floor(((Teuchos::ScalarTraits<double>::random() + 1) * 0.5 * maxEntriesPerRow) + 1));
+    if (maxNnzPerRow < *i) {
+      maxNnzPerRow = *i;
     }
-
-    Level level;
-    RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-    level.SetFactoryManager(factoryHandler);
-    GO nx=7;
-    GO ny=nx;
-    Xpetra::global_size_t numGlobalElements = nx*ny;
-    size_t maxEntriesPerRow=30;
-
-    // Populate CrsMatrix with random number of entries (up to maxEntriesPerRow) per row.
-    RCP<const Map> map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, 0/*indexBase*/, comm);
-    const size_t numMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = map->getLocalElementList();
-    // memory on the fly.  While not super efficient, this
-    // ensures that no zeros are being stored.  Thus, from
-    // Zoltan's perspective the matrix is imbalanced.
-    // Create a vector with random integer entries in [1,maxEntriesPerRow].
-    ST::seedrandom(666*comm->getRank());
-    LO maxNnzPerRow = 0;
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > entriesPerRow = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(map,false);
-    Teuchos::ArrayRCP<LO> eprData = entriesPerRow->getDataNonConst(0);
-    for (typename Teuchos::ArrayRCP<LO>::iterator i=eprData.begin(); i!=eprData.end(); ++i) {
-      *i = static_cast<LO>(std::floor(((Teuchos::ScalarTraits<double>::random()+1)*0.5*maxEntriesPerRow)+1));
-      if(maxNnzPerRow < *i) {maxNnzPerRow = *i;}
-    }
-    RCP<Matrix> A = rcp(new CrsMatrixWrap(map, maxNnzPerRow)); // Force underlying linear algebra library to allocate more
-
-    RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fos->setOutputToRootOnly(-1);
-
-    GO zero = Teuchos::OrdinalTraits<GO>::zero();
-    GO one = Teuchos::OrdinalTraits<GO>::one();
-
-    Teuchos::Array<Scalar> vals(maxEntriesPerRow);
-    Teuchos::Array<GO> cols(maxEntriesPerRow);
-    for (size_t i = 0; i < numMyElements; ++i) {
-      Teuchos::ArrayView<SC> av(&vals[0],eprData[i]);
-      Teuchos::ArrayView<GO> iv(&cols[0],eprData[i]);
-      //stick in ones for values
-      for (LO j=0; j< eprData[i]; ++j) vals[j] = ST::one();
-      //figure out valid column indices
-      GO start = std::max(myGlobalElements[i]-eprData[i]+one,zero);
-      for (LO j=0; j< eprData[i]; ++j) cols[j] = start+j;
-      A->insertGlobalValues(myGlobalElements[i], iv, av);
-    }
-
-    A->fillComplete();
-    level.Set("A",A);
-
-    //build coordinates
-    RCP<const Map> rowMap = A->getRowMap();
-    Teuchos::ParameterList list;
-    list.set("nx",nx);
-    list.set("ny",ny);
-    typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
-    RCP<double_multivector_type> XYZ = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double,LO,GO,Map,double_multivector_type>("2D",rowMap,list);
-    level.Set("Coordinates",XYZ);
-
-    int numPartitions = comm->getSize();
-    level.Set("number of partitions",numPartitions);
-    RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-    //zoltan->SetNumberOfPartitions(numPartitions);
-    //zoltan->SetOutputLevel(0); //options are 0=none, 1=summary, 2=every pid prints
-    level.Request("Partition",zoltan.get());
-    zoltan->Build(level);
-
-    RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = level.Get<RCP<Xpetra::Vector<GO,LO,GO,NO> > >("Partition",zoltan.get());
-    /* //TODO temporary code to have the trivial decomposition (no change)
-       ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-       for (ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
-     *i = comm->getRank();
-     decompEntries=Teuchos::null;
-     */ //TODO end of temporary code
-
-    //Create vector whose local length is the global number of partitions.
-    //This vector will record the local number of nonzeros associated with each partition.
-    Teuchos::Array<GO> parts(numPartitions);
-    for (int i=0; i<numPartitions; ++i) parts[i] = i;
-    Teuchos::ArrayView<GO> partsView(&parts[0],numPartitions);
-    RCP<const Map> partitionMap = MapFactory::Build(TestHelpers::Parameters::getLib(),
-        Teuchos::OrdinalTraits<global_size_t>::invalid(), partsView,
-        map->getIndexBase(),comm);
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > localPartsVec = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(partitionMap);
-
-    //For the local rows in each partition, tally up the number of nonzeros.  This is what
-    //Zoltan should be load-balancing.
-    Teuchos::ArrayRCP<LO> lpvData = localPartsVec->getDataNonConst(0);
-    Teuchos::ArrayRCP<const GO> decompData = decomposition->getData(0);
-    for (size_t i=0; i<decomposition->getLocalLength();++i) {
-      Teuchos::ArrayView<const LO> c;
-      Teuchos::ArrayView<const SC> v;
-      A->getLocalRowView(i,c,v);
-      lpvData[decompData[i]] += v.size();
-    }
-
-    lpvData = Teuchos::null;
-    decompData = Teuchos::null;
-
-    //localPartsVec->describe(*fos,Teuchos::VERB_EXTREME);
-
-    //Send the local nnz tallies to pid 0, which can report the global sums.
-    size_t mysize=1;
-    if (comm->getRank() == 0) mysize = numPartitions;
-    RCP<const Map> globalTallyMap = MapFactory::Build(TestHelpers::Parameters::getLib(),
-        Teuchos::OrdinalTraits<global_size_t>::invalid(),
-        mysize,
-        map->getIndexBase(),
-        comm);
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > globalTallyVec = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(globalTallyMap);
-    RCP<const Export> exporter = ExportFactory::Build( partitionMap, globalTallyMap);
-    globalTallyVec->doExport(*localPartsVec,*exporter,Xpetra::ADD);
-
-    ArrayRCP<GO> expectedResults(numPartitions);
-    switch (comm->getSize()) {
-      case 1:
-        expectedResults[0] = 807;
-        break;
-
-      case 2:
-        expectedResults[0] = 364;
-        expectedResults[1] = 363;
-        break;
-
-      case 3:
-        expectedResults[0] = 277;
-        expectedResults[1] = 261;
-        expectedResults[2] = 269;
-        break;
-
-      case 4:
-        expectedResults[0] = 195;
-        expectedResults[1] = 186;
-        expectedResults[2] = 177;
-        expectedResults[3] = 168;
-        break;
-
-      case 5:
-        expectedResults[0] = 161;
-        expectedResults[1] = 145;
-        expectedResults[2] = 148;
-        expectedResults[3] = 159;
-        expectedResults[4] = 157;
-        break;
-
-      default:
-        break;
-    };
-
-    //FIXME cool ... this next line causes a hang if locally the globalyTallyVec has no data.
-    //FIXME I get around this by making mysize (above) 1 instead of 0. Is this a bug or feature
-    //FIXME in getData?
-    ArrayRCP<const LO> gtvData = globalTallyVec->getData(0);
+  }
+  RCP<Matrix> A = rcp(new CrsMatrixWrap(map, maxNnzPerRow));  // Force underlying linear algebra library to allocate more
+
+  RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fos->setOutputToRootOnly(-1);
+
+  GO zero = Teuchos::OrdinalTraits<GO>::zero();
+  GO one  = Teuchos::OrdinalTraits<GO>::one();
+
+  Teuchos::Array<Scalar> vals(maxEntriesPerRow);
+  Teuchos::Array<GO> cols(maxEntriesPerRow);
+  for (size_t i = 0; i < numMyElements; ++i) {
+    Teuchos::ArrayView<SC> av(&vals[0], eprData[i]);
+    Teuchos::ArrayView<GO> iv(&cols[0], eprData[i]);
+    // stick in ones for values
+    for (LO j = 0; j < eprData[i]; ++j) vals[j] = ST::one();
+    // figure out valid column indices
+    GO start = std::max(myGlobalElements[i] - eprData[i] + one, zero);
+    for (LO j = 0; j < eprData[i]; ++j) cols[j] = start + j;
+    A->insertGlobalValues(myGlobalElements[i], iv, av);
+  }
+
+  A->fillComplete();
+  level.Set("A", A);
+
+  // build coordinates
+  RCP<const Map> rowMap = A->getRowMap();
+  Teuchos::ParameterList list;
+  list.set("nx", nx);
+  list.set("ny", ny);
+  typedef Xpetra::MultiVector<typename Teuchos::ScalarTraits<Scalar>::magnitudeType, LocalOrdinal, GlobalOrdinal, Node> double_multivector_type;
+  RCP<double_multivector_type> XYZ = Galeri::Xpetra::Utils::CreateCartesianCoordinates<double, LO, GO, Map, double_multivector_type>("2D", rowMap, list);
+  level.Set("Coordinates", XYZ);
+
+  int numPartitions = comm->getSize();
+  level.Set("number of partitions", numPartitions);
+  RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
+  // zoltan->SetNumberOfPartitions(numPartitions);
+  // zoltan->SetOutputLevel(0); //options are 0=none, 1=summary, 2=every pid prints
+  level.Request("Partition", zoltan.get());
+  zoltan->Build(level);
+
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = level.Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", zoltan.get());
+  /* //TODO temporary code to have the trivial decomposition (no change)
+     ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
+     for (ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
+   *i = comm->getRank();
+   decompEntries=Teuchos::null;
+   */
+  // TODO end of temporary code
+
+  // Create vector whose local length is the global number of partitions.
+  // This vector will record the local number of nonzeros associated with each partition.
+  Teuchos::Array<GO> parts(numPartitions);
+  for (int i = 0; i < numPartitions; ++i) parts[i] = i;
+  Teuchos::ArrayView<GO> partsView(&parts[0], numPartitions);
+  RCP<const Map> partitionMap                        = MapFactory::Build(TestHelpers::Parameters::getLib(),
+                                                                         Teuchos::OrdinalTraits<global_size_t>::invalid(), partsView,
+                                                                         map->getIndexBase(), comm);
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > localPartsVec = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(partitionMap);
+
+  // For the local rows in each partition, tally up the number of nonzeros.  This is what
+  // Zoltan should be load-balancing.
+  Teuchos::ArrayRCP<LO> lpvData          = localPartsVec->getDataNonConst(0);
+  Teuchos::ArrayRCP<const GO> decompData = decomposition->getData(0);
+  for (size_t i = 0; i < decomposition->getLocalLength(); ++i) {
+    Teuchos::ArrayView<const LO> c;
+    Teuchos::ArrayView<const SC> v;
+    A->getLocalRowView(i, c, v);
+    lpvData[decompData[i]] += v.size();
+  }
+
+  lpvData    = Teuchos::null;
+  decompData = Teuchos::null;
+
+  // localPartsVec->describe(*fos,Teuchos::VERB_EXTREME);
+
+  // Send the local nnz tallies to pid 0, which can report the global sums.
+  size_t mysize = 1;
+  if (comm->getRank() == 0) mysize = numPartitions;
+  RCP<const Map> globalTallyMap                       = MapFactory::Build(TestHelpers::Parameters::getLib(),
+                                                                          Teuchos::OrdinalTraits<global_size_t>::invalid(),
+                                                                          mysize,
+                                                                          map->getIndexBase(),
+                                                                          comm);
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > globalTallyVec = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(globalTallyMap);
+  RCP<const Export> exporter                          = ExportFactory::Build(partitionMap, globalTallyMap);
+  globalTallyVec->doExport(*localPartsVec, *exporter, Xpetra::ADD);
+
+  ArrayRCP<GO> expectedResults(numPartitions);
+  switch (comm->getSize()) {
+    case 1:
+      expectedResults[0] = 807;
+      break;
+
+    case 2:
+      expectedResults[0] = 364;
+      expectedResults[1] = 363;
+      break;
+
+    case 3:
+      expectedResults[0] = 277;
+      expectedResults[1] = 261;
+      expectedResults[2] = 269;
+      break;
+
+    case 4:
+      expectedResults[0] = 195;
+      expectedResults[1] = 186;
+      expectedResults[2] = 177;
+      expectedResults[3] = 168;
+      break;
+
+    case 5:
+      expectedResults[0] = 161;
+      expectedResults[1] = 145;
+      expectedResults[2] = 148;
+      expectedResults[3] = 159;
+      expectedResults[4] = 157;
+      break;
+
+    default:
+      break;
+  };
+
+  // FIXME cool ... this next line causes a hang if locally the globalyTallyVec has no data.
+  // FIXME I get around this by making mysize (above) 1 instead of 0. Is this a bug or feature
+  // FIXME in getData?
+  ArrayRCP<const LO> gtvData = globalTallyVec->getData(0);
 
 #ifdef __linux__
-    out << "Checking results..." << std::endl;
-    for (int i=0; i<numPartitions; ++i) {
-      if (comm->getRank() == 0) TEST_EQUALITY( expectedResults[i], gtvData[i]);
-    }
+  out << "Checking results..." << std::endl;
+  for (int i = 0; i < numPartitions; ++i) {
+    if (comm->getRank() == 0) TEST_EQUALITY(expectedResults[i], gtvData[i]);
+  }
 #endif
 
 #ifdef MUELU_VISUALIZE_REPARTITIONING
-    //
-    //Now write everything to a comma-separate list that ParaView can grok
-    //
-    Teuchos::ArrayRCP<const double> X = XYZ->getData(0);
-    Teuchos::ArrayRCP<const double> Y = XYZ->getData(1);
-    Teuchos::ArrayRCP<const GO> D = decomposition->getData(0);
-    RCP<std::ofstream> outFile;
-    std::string fileName = "zoltanResults.csv";
-
-    //write header information
-    if (comm->getRank() == 0) {
-      outFile = rcp(new std::ofstream(fileName.c_str()));
-      *outFile << "x coord, y coord, z coord, scalar" << std::endl;
-    }
-    comm->barrier();
-
-    //append coordinates
-    for (int j=0; j<comm->getSize(); ++j) {
-      int mypid = comm->getRank();
-      if (mypid == j) {
-        outFile = rcp(new std::ofstream(fileName.c_str(),std::ios::app));
-        for (int i=0; i < D.size(); ++i) {
-          *outFile << X[i] << ", " << Y[i] << ", " << ST::zero() << ", " << D[i] << std::endl;
-        }
+  //
+  // Now write everything to a comma-separate list that ParaView can grok
+  //
+  Teuchos::ArrayRCP<const double> X = XYZ->getData(0);
+  Teuchos::ArrayRCP<const double> Y = XYZ->getData(1);
+  Teuchos::ArrayRCP<const GO> D     = decomposition->getData(0);
+  RCP<std::ofstream> outFile;
+  std::string fileName = "zoltanResults.csv";
+
+  // write header information
+  if (comm->getRank() == 0) {
+    outFile = rcp(new std::ofstream(fileName.c_str()));
+    *outFile << "x coord, y coord, z coord, scalar" << std::endl;
+  }
+  comm->barrier();
+
+  // append coordinates
+  for (int j = 0; j < comm->getSize(); ++j) {
+    int mypid = comm->getRank();
+    if (mypid == j) {
+      outFile = rcp(new std::ofstream(fileName.c_str(), std::ios::app));
+      for (int i = 0; i < D.size(); ++i) {
+        *outFile << X[i] << ", " << Y[i] << ", " << ST::zero() << ", " << D[i] << std::endl;
       }
-    } //for (int i=0; i<comm->getSize(); ++i)
-
-    out << std::endl;
-    out << "You can view the Zoltan decomposition in ParaView 3.10.1 or later:" << std::endl;
-    out << "   1) Load the data file " << fileName << "." << std::endl;
-    out << "   2) Run the filter Filters/ Alphabetical/ Table To Points." << std::endl;
-    out << "   3) Tell ParaView what columns are the X, Y and Z coordinates." << std::endl;
-    out << "   4) Split screen horizontally (Icon, top right)." << std::endl;
-    out << "   5) Click on the eyeball in the Pipeline Browser to see the points." << std::endl;
-    out << "   6) Under the Display tab, you can color points by scalar value and resize them." << std::endl;
-    out << std::endl;
-    out << " To display row weights next to each point:" << std::endl;
-    out << "   1) Click the \"Select Points Through\" button (2nd row) and select all points." << std::endl;
-    out << "   2) Under View pull-down menu, choose the \"Selection Inspector\"." << std::endl;
-    out << "   3) Under the Point Label, check the Visible box and set the Label Mode to \"row weight\"." << std::endl;
+    }
+  }  // for (int i=0; i<comm->getSize(); ++i)
+
+  out << std::endl;
+  out << "You can view the Zoltan decomposition in ParaView 3.10.1 or later:" << std::endl;
+  out << "   1) Load the data file " << fileName << "." << std::endl;
+  out << "   2) Run the filter Filters/ Alphabetical/ Table To Points." << std::endl;
+  out << "   3) Tell ParaView what columns are the X, Y and Z coordinates." << std::endl;
+  out << "   4) Split screen horizontally (Icon, top right)." << std::endl;
+  out << "   5) Click on the eyeball in the Pipeline Browser to see the points." << std::endl;
+  out << "   6) Under the Display tab, you can color points by scalar value and resize them." << std::endl;
+  out << std::endl;
+  out << " To display row weights next to each point:" << std::endl;
+  out << "   1) Click the \"Select Points Through\" button (2nd row) and select all points." << std::endl;
+  out << "   2) Under View pull-down menu, choose the \"Selection Inspector\"." << std::endl;
+  out << "   3) Under the Point Label, check the Visible box and set the Label Mode to \"row weight\"." << std::endl;
 #endif
 
-  } //Build
+}  // Build
 
-#ifdef DISABLED // JG: FIXME: coordinates format
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan, Build3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+#ifdef DISABLED  // JG: FIXME: coordinates format
+TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan, Build3PDEs, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  typedef Teuchos::ScalarTraits<Scalar> ST;
 
-    typedef Teuchos::ScalarTraits<Scalar> ST;
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << std::endl;
+  out << "This tests that the partitioning produced by Zoltan is \"reasonable\" for a matrix" << std::endl;
+  out << "that has a random number of nonzeros per row and 3 DOFs per mesh point.  Good results have been precomputed" << std::endl;
+  out << "for up to 5 processors.  The results are the number of nonzeros in the local matrix" << std::endl;
+  out << "once the Zoltan repartitioning has been applied." << std::endl;
+  out << "The results can be viewed in Paraview by enabling code guarded by the macro MUELU_VISUALIZE_REPARTITIONING" << std::endl;
 
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << std::endl;
-    out << "This tests that the partitioning produced by Zoltan is \"reasonable\" for a matrix" << std::endl;
-    out << "that has a random number of nonzeros per row and 3 DOFs per mesh point.  Good results have been precomputed" << std::endl;
-    out << "for up to 5 processors.  The results are the number of nonzeros in the local matrix" << std::endl;
-    out << "once the Zoltan repartitioning has been applied." << std::endl;
-    out << "The results can be viewed in Paraview by enabling code guarded by the macro MUELU_VISUALIZE_REPARTITIONING" << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
-
-    if (comm->getSize() > 5) {
-      out << std::endl;
-      out << "This test must be run on 1 to 5 processes." << std::endl;
-      TEST_EQUALITY(true, true);
-      return;
-    }
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers::Parameters::getDefaultComm();
 
-    Level level;
-    RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-    level.SetFactoryManager(factoryHandler);
-    int nx=9;
-    int ny=nx;
-    int dofsPerNode = 3;
-    GO numGlobalElements = nx*ny*dofsPerNode;
-    size_t maxEntriesPerRow=30;
-
-    RCP<const Map> map;
-    int numMyNodes = numGlobalElements / dofsPerNode;
-    if (comm->getSize() > 1) {
-      // In parallel, make sure that the dof's associated with a node all
-      // reside on the same processor.
-      int numNodes = numGlobalElements / dofsPerNode;
-      TEUCHOS_TEST_FOR_EXCEPTION( (numGlobalElements - numNodes * dofsPerNode) != 0, MueLu::Exceptions::RuntimeError,
-          "Number of matrix rows is not divisible by #dofs" );
-      int nproc = comm->getSize();
-      if (comm->getRank() < nproc-1) numMyNodes = numNodes / nproc;
-      else numMyNodes = numNodes - (numNodes/nproc) * (nproc-1);
-      map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyNodes*dofsPerNode, 0/*indexBase*/, comm);
-    } else {
-      map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, 0/*indexBase*/, comm);
-    }
-
-    const size_t numMyElements = map->getLocalNumElements();
-    Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = map->getLocalElementList();
-    RCP<Matrix> A = rcp(new CrsMatrixWrap(map, 1)); // Force underlying linear algebra library to allocate more
-    // memory on the fly.  While not super efficient, this
-    // ensures that no zeros are being stored.  Thus, from
-    // Zoltan's perspective the matrix is imbalanced.
-    // Populate CrsMatrix with random number of entries (up to maxEntriesPerRow) per row.
-    // Create a vector with random integer entries in [1,maxEntriesPerRow].
-    ST::seedrandom(666*comm->getRank());
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > entriesPerRow = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(map,false);
-    Teuchos::ArrayRCP<LO> eprData = entriesPerRow->getDataNonConst(0);
-    for (Teuchos::ArrayRCP<LO>::iterator i=eprData.begin(); i!=eprData.end(); ++i) {
-      *i = (LO)(std::floor(((ST::random()+1)*0.5*maxEntriesPerRow)+1));
-    }
-
-    RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fos->setOutputToRootOnly(-1);
-
-    Teuchos::Array<Scalar> vals(maxEntriesPerRow);
-    Teuchos::Array<GO> cols(maxEntriesPerRow);
-    for (size_t i = 0; i < numMyElements; ++i) {
-      Teuchos::ArrayView<SC> av(&vals[0],eprData[i]);
-      Teuchos::ArrayView<GO> iv(&cols[0],eprData[i]);
-      //stick in ones for values
-      for (LO j=0; j< eprData[i]; ++j) vals[j] = ST::one();
-      //figure out valid column indices
-      GO start = std::max(myGlobalElements[i]-eprData[i]+1,0);
-      for (LO j=0; j< eprData[i]; ++j) cols[j] = start+j;
-      A->insertGlobalValues(myGlobalElements[i], iv, av);
-    }
-
-    A->fillComplete();
-
-    // Now treat the matrix as if it has 3 DOFs per node.
-    A->SetFixedBlockSize(dofsPerNode);
-    level.Set("A",A);
-
-    //build coordinates
-    Teuchos::ParameterList list;
-    list.set("nx",nx);
-    list.set("ny",ny);
-    RCP<const Map> coalescedMap = MapFactory::Build<NO>(TestHelpers::Parameters::getLib(), numGlobalElements/dofsPerNode, numMyNodes, 0/*indexBase*/, comm);
-    RCP<MultiVector> XYZ = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D",coalescedMap,list);
-
-    // XYZ are the "coalesce" coordinates as it has been generated for 1 DOF/node and we are using them for 3 DOFS/node
-    // level.Set("Coordinates",XYZ); "Coordinates" == uncoalesce. "X,Y,ZCoordinates" == coalesce
-    {
-      RCP<MultiVector> coordinates = XYZ;
-
-      // making a copy because I don't want to keep 'open' the Xpetra_MultiVector
-      if (coordinates->getNumVectors() >= 1) {
-        Teuchos::ArrayRCP<const SC> coord = coordinates->getData(0);
-        Teuchos::ArrayRCP<SC> coordCpy(coord.size());
-        for(int i=0; i<coord.size(); i++) {
-          coordCpy[i] = coord[i];
-        }
-        level.Set("XCoordinates", coordCpy);
-        //std::cout << coordCpy << std::endl;
-      }
+  if (comm->getSize() > 5) {
+    out << std::endl;
+    out << "This test must be run on 1 to 5 processes." << std::endl;
+    TEST_EQUALITY(true, true);
+    return;
+  }
+
+  Level level;
+  RCP<FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+  level.SetFactoryManager(factoryHandler);
+  int nx                  = 9;
+  int ny                  = nx;
+  int dofsPerNode         = 3;
+  GO numGlobalElements    = nx * ny * dofsPerNode;
+  size_t maxEntriesPerRow = 30;
+
+  RCP<const Map> map;
+  int numMyNodes = numGlobalElements / dofsPerNode;
+  if (comm->getSize() > 1) {
+    // In parallel, make sure that the dof's associated with a node all
+    // reside on the same processor.
+    int numNodes = numGlobalElements / dofsPerNode;
+    TEUCHOS_TEST_FOR_EXCEPTION((numGlobalElements - numNodes * dofsPerNode) != 0, MueLu::Exceptions::RuntimeError,
+                               "Number of matrix rows is not divisible by #dofs");
+    int nproc = comm->getSize();
+    if (comm->getRank() < nproc - 1)
+      numMyNodes = numNodes / nproc;
+    else
+      numMyNodes = numNodes - (numNodes / nproc) * (nproc - 1);
+    map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, numMyNodes * dofsPerNode, 0 /*indexBase*/, comm);
+  } else {
+    map = MapFactory::Build(TestHelpers::Parameters::getLib(), numGlobalElements, 0 /*indexBase*/, comm);
+  }
+
+  const size_t numMyElements                               = map->getLocalNumElements();
+  Teuchos::ArrayView<const GlobalOrdinal> myGlobalElements = map->getLocalElementList();
+  RCP<Matrix> A                                            = rcp(new CrsMatrixWrap(map, 1));  // Force underlying linear algebra library to allocate more
+  // memory on the fly.  While not super efficient, this
+  // ensures that no zeros are being stored.  Thus, from
+  // Zoltan's perspective the matrix is imbalanced.
+  // Populate CrsMatrix with random number of entries (up to maxEntriesPerRow) per row.
+  // Create a vector with random integer entries in [1,maxEntriesPerRow].
+  ST::seedrandom(666 * comm->getRank());
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > entriesPerRow = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(map, false);
+  Teuchos::ArrayRCP<LO> eprData                      = entriesPerRow->getDataNonConst(0);
+  for (Teuchos::ArrayRCP<LO>::iterator i = eprData.begin(); i != eprData.end(); ++i) {
+    *i = (LO)(std::floor(((ST::random() + 1) * 0.5 * maxEntriesPerRow) + 1));
+  }
+
+  RCP<Teuchos::FancyOStream> fos = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fos->setOutputToRootOnly(-1);
+
+  Teuchos::Array<Scalar> vals(maxEntriesPerRow);
+  Teuchos::Array<GO> cols(maxEntriesPerRow);
+  for (size_t i = 0; i < numMyElements; ++i) {
+    Teuchos::ArrayView<SC> av(&vals[0], eprData[i]);
+    Teuchos::ArrayView<GO> iv(&cols[0], eprData[i]);
+    // stick in ones for values
+    for (LO j = 0; j < eprData[i]; ++j) vals[j] = ST::one();
+    // figure out valid column indices
+    GO start = std::max(myGlobalElements[i] - eprData[i] + 1, 0);
+    for (LO j = 0; j < eprData[i]; ++j) cols[j] = start + j;
+    A->insertGlobalValues(myGlobalElements[i], iv, av);
+  }
+
+  A->fillComplete();
+
+  // Now treat the matrix as if it has 3 DOFs per node.
+  A->SetFixedBlockSize(dofsPerNode);
+  level.Set("A", A);
+
+  // build coordinates
+  Teuchos::ParameterList list;
+  list.set("nx", nx);
+  list.set("ny", ny);
+  RCP<const Map> coalescedMap = MapFactory::Build<NO>(TestHelpers::Parameters::getLib(), numGlobalElements / dofsPerNode, numMyNodes, 0 /*indexBase*/, comm);
+  RCP<MultiVector> XYZ        = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", coalescedMap, list);
+
+  // XYZ are the "coalesce" coordinates as it has been generated for 1 DOF/node and we are using them for 3 DOFS/node
+  // level.Set("Coordinates",XYZ); "Coordinates" == uncoalesce. "X,Y,ZCoordinates" == coalesce
+  {
+    RCP<MultiVector> coordinates = XYZ;
 
-      if (coordinates->getNumVectors() >= 2) {
-        Teuchos::ArrayRCP<const SC> coord = coordinates->getData(1);
-        Teuchos::ArrayRCP<SC> coordCpy(coord.size());
-        for(int i=0; i<coord.size(); i++) {
-          coordCpy[i] = coord[i];
-        }
-        level.Set("YCoordinates", coordCpy);
+    // making a copy because I don't want to keep 'open' the Xpetra_MultiVector
+    if (coordinates->getNumVectors() >= 1) {
+      Teuchos::ArrayRCP<const SC> coord = coordinates->getData(0);
+      Teuchos::ArrayRCP<SC> coordCpy(coord.size());
+      for (int i = 0; i < coord.size(); i++) {
+        coordCpy[i] = coord[i];
       }
+      level.Set("XCoordinates", coordCpy);
+      // std::cout << coordCpy << std::endl;
+    }
 
-      /*if (coordinates->getNumVectors() >= 3) {
-        Teuchos::ArrayRCP<const SC> coord = coordinates->getData(2);
-        Teuchos::ArrayRCP<SC> coordCpy(coord.size());
-        for(int i=0; i<coord.size(); i++) {
+    if (coordinates->getNumVectors() >= 2) {
+      Teuchos::ArrayRCP<const SC> coord = coordinates->getData(1);
+      Teuchos::ArrayRCP<SC> coordCpy(coord.size());
+      for (int i = 0; i < coord.size(); i++) {
         coordCpy[i] = coord[i];
-        }
-        level.Set("ZCoordinates", coordCpy);
-        }*/
+      }
+      level.Set("YCoordinates", coordCpy);
     }
 
-    //coalescedMap->describe(*fos,Teuchos::VERB_EXTREME);
-    //sleep(1); comm->barrier();
-    //XYZ->describe(*fos,Teuchos::VERB_EXTREME);
-
-    int numPartitions = comm->getSize();
-    level.Set("number of partitions",numPartitions);
-    RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
-    //zoltan->SetOutputLevel(0); //options are 0=none, 1=summary, 2=every pid prints
-    level.Request("Partition",zoltan.get());
-    zoltan->Build(level);
-
-    RCP<Xpetra::Vector<GO,LO,GO,NO> > decomposition = level.Get<RCP<Xpetra::Vector<GO,LO,GO,NO> > >("Partition",zoltan.get());
-    /* //temporary code to have the trivial decomposition (no change)
-       ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
-       for (ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
-     *i = comm->getRank();
-     decompEntries=Teuchos::null;
+    /*if (coordinates->getNumVectors() >= 3) {
+      Teuchos::ArrayRCP<const SC> coord = coordinates->getData(2);
+      Teuchos::ArrayRCP<SC> coordCpy(coord.size());
+      for(int i=0; i<coord.size(); i++) {
+      coordCpy[i] = coord[i];
+      }
+      level.Set("ZCoordinates", coordCpy);
+      }*/
+  }
+
+  // coalescedMap->describe(*fos,Teuchos::VERB_EXTREME);
+  // sleep(1); comm->barrier();
+  // XYZ->describe(*fos,Teuchos::VERB_EXTREME);
+
+  int numPartitions = comm->getSize();
+  level.Set("number of partitions", numPartitions);
+  RCP<ZoltanInterface> zoltan = rcp(new ZoltanInterface());
+  // zoltan->SetOutputLevel(0); //options are 0=none, 1=summary, 2=every pid prints
+  level.Request("Partition", zoltan.get());
+  zoltan->Build(level);
+
+  RCP<Xpetra::Vector<GO, LO, GO, NO> > decomposition = level.Get<RCP<Xpetra::Vector<GO, LO, GO, NO> > >("Partition", zoltan.get());
+  /* //temporary code to have the trivial decomposition (no change)
+     ArrayRCP<GO> decompEntries = decomposition->getDataNonConst(0);
+     for (ArrayRCP<GO>::iterator i = decompEntries.begin(); i != decompEntries.end(); ++i)
+   *i = comm->getRank();
+   decompEntries=Teuchos::null;
+   */
+
+  // Create vector whose local length is the global number of partitions.
+  // This vector will record the local number of nonzeros associated with each partition.
+  Teuchos::Array<GO> parts(numPartitions);
+  for (int i = 0; i < numPartitions; ++i) parts[i] = i;
+  Teuchos::ArrayView<GO> partsView(&parts[0], numPartitions);
+  RCP<const Map> partitionMap                        = MapFactory::Build(TestHelpers::Parameters::getLib(),
+                                                                         Teuchos::OrdinalTraits<global_size_t>::invalid(), partsView,
+                                                                         map->getIndexBase(), comm);
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > localPartsVec = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(partitionMap);
+
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > nnzPerRow = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(A->getRowMap());
+  Teuchos::ArrayRCP<GO> nnzData                  = nnzPerRow->getDataNonConst(0);
+  // For the local rows in each partition, tally up the number of nonzeros.  This is what
+  // Zoltan should be load-balancing.
+  Teuchos::ArrayRCP<GO> lpvData          = localPartsVec->getDataNonConst(0);
+  Teuchos::ArrayRCP<const GO> decompData = decomposition->getData(0);
+  for (size_t i = 0; i < decomposition->getLocalLength(); ++i) {
+    Teuchos::ArrayView<const LO> c;
+    Teuchos::ArrayView<const SC> v;
+    A->getLocalRowView(i, c, v);
+    lpvData[decompData[i]] += v.size();
+    nnzData[i] = v.size();
+  }
+
+  lpvData    = Teuchos::null;
+  decompData = Teuchos::null;
+  nnzData    = Teuchos::null;
+
+  /*
+     if (comm->getRank() == 0)
+     std::cout << "nnz per row" << std::endl;
+     nnzPerRow->describe(*fos,Teuchos::VERB_EXTREME);
+
+     if (comm->getRank() == 0)
+     std::cout << "Row-to-partition assignment (from Zoltan)" << std::endl;
+     decomposition->describe(*fos,Teuchos::VERB_EXTREME);
+
+     if (comm->getRank() == 0)
+     std::cout << "#nonzeros per partition" << std::endl;
+     localPartsVec->describe(*fos,Teuchos::VERB_EXTREME);
      */
 
-    //Create vector whose local length is the global number of partitions.
-    //This vector will record the local number of nonzeros associated with each partition.
-    Teuchos::Array<GO> parts(numPartitions);
-    for (int i=0; i<numPartitions; ++i) parts[i] = i;
-    Teuchos::ArrayView<GO> partsView(&parts[0],numPartitions);
-    RCP<const Map> partitionMap = MapFactory::Build(TestHelpers::Parameters::getLib(),
-        Teuchos::OrdinalTraits<global_size_t>::invalid(), partsView,
-        map->getIndexBase(),comm);
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > localPartsVec = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(partitionMap);
-
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > nnzPerRow = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(A->getRowMap());
-    Teuchos::ArrayRCP<GO> nnzData = nnzPerRow->getDataNonConst(0);
-    //For the local rows in each partition, tally up the number of nonzeros.  This is what
-    //Zoltan should be load-balancing.
-    Teuchos::ArrayRCP<GO> lpvData = localPartsVec->getDataNonConst(0);
-    Teuchos::ArrayRCP<const GO> decompData = decomposition->getData(0);
-    for (size_t i=0; i<decomposition->getLocalLength();++i) {
-      Teuchos::ArrayView<const LO> c;
-      Teuchos::ArrayView<const SC> v;
-      A->getLocalRowView(i,c,v);
-      lpvData[decompData[i]] += v.size();
-      nnzData[i] = v.size();
-    }
-
-    lpvData = Teuchos::null;
-    decompData = Teuchos::null;
-    nnzData = Teuchos::null;
-
-    /*
-       if (comm->getRank() == 0)
-       std::cout << "nnz per row" << std::endl;
-       nnzPerRow->describe(*fos,Teuchos::VERB_EXTREME);
-
-       if (comm->getRank() == 0)
-       std::cout << "Row-to-partition assignment (from Zoltan)" << std::endl;
-       decomposition->describe(*fos,Teuchos::VERB_EXTREME);
-
-       if (comm->getRank() == 0)
-       std::cout << "#nonzeros per partition" << std::endl;
-       localPartsVec->describe(*fos,Teuchos::VERB_EXTREME);
-       */
-
-    //Send the local nnz tallies to pid 0, which can report the global sums.
-    size_t mysize=1;
-    if (comm->getRank() == 0) mysize = numPartitions;
-    RCP<const Map> globalTallyMap = MapFactory::Build(TestHelpers::Parameters::getLib(),
-        Teuchos::OrdinalTraits<global_size_t>::invalid(),
-        mysize,
-        map->getIndexBase(),
-        comm);
-    RCP<Xpetra::Vector<LO,LO,GO,NO> > globalTallyVec = Xpetra::VectorFactory<LO,LO,GO,NO>::Build(globalTallyMap);
-    RCP<const Export> exporter = ExportFactory::Build( partitionMap, globalTallyMap);
-    globalTallyVec->doExport(*localPartsVec,*exporter,Xpetra::ADD);
-
-    ArrayRCP<GO> expectedResults(numPartitions);
-    switch (comm->getSize()) {
-      case 1:
-        expectedResults[0] = 3951;
-        break;
-
-      case 2:
-        expectedResults[0] = 1955;
-        expectedResults[1] = 1910;
-        break;
-
-      case 3:
-        expectedResults[0] = 1326;
-        expectedResults[1] = 1340;
-        expectedResults[2] = 1321;
-        break;
-
-      case 4:
-        expectedResults[0] = 950;
-        expectedResults[1] = 922;
-        expectedResults[2] = 908;
-        expectedResults[3] = 936;
-        break;
-
-      case 5:
-        expectedResults[0] = 774;
-        expectedResults[1] = 735;
-        expectedResults[2] = 726;
-        expectedResults[3] = 771;
-        expectedResults[4] = 759;
-        break;
-
-      default:
-        break;
-    };
-
-    ArrayRCP<const LO> gtvData = globalTallyVec->getData(0);
+  // Send the local nnz tallies to pid 0, which can report the global sums.
+  size_t mysize = 1;
+  if (comm->getRank() == 0) mysize = numPartitions;
+  RCP<const Map> globalTallyMap                       = MapFactory::Build(TestHelpers::Parameters::getLib(),
+                                                                          Teuchos::OrdinalTraits<global_size_t>::invalid(),
+                                                                          mysize,
+                                                                          map->getIndexBase(),
+                                                                          comm);
+  RCP<Xpetra::Vector<LO, LO, GO, NO> > globalTallyVec = Xpetra::VectorFactory<LO, LO, GO, NO>::Build(globalTallyMap);
+  RCP<const Export> exporter                          = ExportFactory::Build(partitionMap, globalTallyMap);
+  globalTallyVec->doExport(*localPartsVec, *exporter, Xpetra::ADD);
+
+  ArrayRCP<GO> expectedResults(numPartitions);
+  switch (comm->getSize()) {
+    case 1:
+      expectedResults[0] = 3951;
+      break;
+
+    case 2:
+      expectedResults[0] = 1955;
+      expectedResults[1] = 1910;
+      break;
+
+    case 3:
+      expectedResults[0] = 1326;
+      expectedResults[1] = 1340;
+      expectedResults[2] = 1321;
+      break;
+
+    case 4:
+      expectedResults[0] = 950;
+      expectedResults[1] = 922;
+      expectedResults[2] = 908;
+      expectedResults[3] = 936;
+      break;
+
+    case 5:
+      expectedResults[0] = 774;
+      expectedResults[1] = 735;
+      expectedResults[2] = 726;
+      expectedResults[3] = 771;
+      expectedResults[4] = 759;
+      break;
+
+    default:
+      break;
+  };
+
+  ArrayRCP<const LO> gtvData = globalTallyVec->getData(0);
 
 #ifdef __linux__
-    out << "Checking results..." << std::endl;
-    for (int i=0; i<numPartitions; ++i) {
-      if (comm->getRank() == 0) TEST_EQUALITY( expectedResults[i], gtvData[i]);
-    }
+  out << "Checking results..." << std::endl;
+  for (int i = 0; i < numPartitions; ++i) {
+    if (comm->getRank() == 0) TEST_EQUALITY(expectedResults[i], gtvData[i]);
+  }
 #endif
 
 #ifdef MUELU_VISUALIZE_REPARTITIONING
-    //
-    //Now write everything to a comma-separate list that ParaView can grok
-    //
-    Teuchos::ArrayRCP<const double> X = XYZ->getData(0);
-    Teuchos::ArrayRCP<const double> Y = XYZ->getData(1);
-    Teuchos::ArrayRCP<const GO> D = decomposition->getData(0);
-    RCP<std::ofstream> outFile;
-    std::string fileName = "zoltanResults.csv";
-
-    //write header information
-    if (comm->getRank() == 0) {
-      outFile = rcp(new std::ofstream(fileName.c_str()));
-      *outFile << "x coord, y coord, z coord, partition, row weight" << std::endl;
-    }
-    comm->barrier();
-
-    //append coordinates
-    nnzData = nnzPerRow->getDataNonConst(0);
-    for (int j=0; j<comm->getSize(); ++j) {
-      int mypid = comm->getRank();
-      if (mypid == j) {
-        outFile = rcp(new std::ofstream(fileName.c_str(),std::ios::app));
-        int blockSize = A->GetFixedBlockSize();
-        //Coordinates are for coalesced system, D is for uncoalesced
-        for (int i=0; i < D.size()/blockSize; ++i) {
-          int nnz=0;
-          for (int k=0; k<blockSize; ++k)  nnz += nnzData[i*blockSize+k];
-          *outFile << X[i] << ", " << Y[i] << ", " << ST::zero() << ", "
-            << D[i*blockSize] << ", " << nnz << std::endl;
-        }
+  //
+  // Now write everything to a comma-separate list that ParaView can grok
+  //
+  Teuchos::ArrayRCP<const double> X = XYZ->getData(0);
+  Teuchos::ArrayRCP<const double> Y = XYZ->getData(1);
+  Teuchos::ArrayRCP<const GO> D     = decomposition->getData(0);
+  RCP<std::ofstream> outFile;
+  std::string fileName = "zoltanResults.csv";
+
+  // write header information
+  if (comm->getRank() == 0) {
+    outFile = rcp(new std::ofstream(fileName.c_str()));
+    *outFile << "x coord, y coord, z coord, partition, row weight" << std::endl;
+  }
+  comm->barrier();
+
+  // append coordinates
+  nnzData = nnzPerRow->getDataNonConst(0);
+  for (int j = 0; j < comm->getSize(); ++j) {
+    int mypid = comm->getRank();
+    if (mypid == j) {
+      outFile       = rcp(new std::ofstream(fileName.c_str(), std::ios::app));
+      int blockSize = A->GetFixedBlockSize();
+      // Coordinates are for coalesced system, D is for uncoalesced
+      for (int i = 0; i < D.size() / blockSize; ++i) {
+        int nnz = 0;
+        for (int k = 0; k < blockSize; ++k) nnz += nnzData[i * blockSize + k];
+        *outFile << X[i] << ", " << Y[i] << ", " << ST::zero() << ", "
+                 << D[i * blockSize] << ", " << nnz << std::endl;
       }
-    } //for (int i=0; i<comm->getSize(); ++i)
-
-    out << std::endl;
-    out << "You can view the Zoltan decomposition in ParaView 3.10.1 or later:" << std::endl;
-    out << "   1) Load the data file " << fileName << "." << std::endl;
-    out << "   2) Run the filter Filters/ Alphabetical/ Table To Points." << std::endl;
-    out << "   3) Tell ParaView what columns are the X, Y and Z coordinates." << std::endl;
-    out << "   4) Split screen horizontally (Icon, top right)." << std::endl;
-    out << "   5) Click on the eyeball in the Pipeline Browser to see the points." << std::endl;
-    out << "   6) Under the Display tab, you can color points by scalar value and resize them." << std::endl;
-    out << std::endl;
-    out << " To display row weights next to each point:" << std::endl;
-    out << "   1) Click the \"Select Points Through\" button (2nd row) and select all points." << std::endl;
-    out << "   2) Under View pull-down menu, choose the \"Selection Inspector\"." << std::endl;
-    out << "   3) Under the Point Label, check the Visible box and set the Label Mode to \"row weight\"." << std::endl;
+    }
+  }  // for (int i=0; i<comm->getSize(); ++i)
+
+  out << std::endl;
+  out << "You can view the Zoltan decomposition in ParaView 3.10.1 or later:" << std::endl;
+  out << "   1) Load the data file " << fileName << "." << std::endl;
+  out << "   2) Run the filter Filters/ Alphabetical/ Table To Points." << std::endl;
+  out << "   3) Tell ParaView what columns are the X, Y and Z coordinates." << std::endl;
+  out << "   4) Split screen horizontally (Icon, top right)." << std::endl;
+  out << "   5) Click on the eyeball in the Pipeline Browser to see the points." << std::endl;
+  out << "   6) Under the Display tab, you can color points by scalar value and resize them." << std::endl;
+  out << std::endl;
+  out << " To display row weights next to each point:" << std::endl;
+  out << "   1) Click the \"Select Points Through\" button (2nd row) and select all points." << std::endl;
+  out << "   2) Under View pull-down menu, choose the \"Selection Inspector\"." << std::endl;
+  out << "   3) Under the Point Label, check the Visible box and set the Label Mode to \"row weight\"." << std::endl;
 #endif
 
-  } //Build3PDEs
-#endif // TMP
+}  // Build3PDEs
+#endif  // TMP
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan,Constructor,SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan,Build,SC,LO,GO,NO)
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan, Constructor, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Zoltan, Build, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests/utils/MueLu_FakeSmootherPrototype.hpp b/packages/muelu/test/unit_tests/utils/MueLu_FakeSmootherPrototype.hpp
index bdf41545a316..e7ee267298e3 100644
--- a/packages/muelu/test/unit_tests/utils/MueLu_FakeSmootherPrototype.hpp
+++ b/packages/muelu/test/unit_tests/utils/MueLu_FakeSmootherPrototype.hpp
@@ -53,62 +53,61 @@
 
 namespace MueLu {
 
-  class Level;
-
-  template <class Scalar = SmootherPrototype<>::scalar_type,
-           class LocalOrdinal = typename SmootherPrototype<Scalar>::local_ordinal_type,
-           class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
-           class Node = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
-             class FakeSmootherPrototype :
-               public SmootherPrototype<Scalar,LocalOrdinal,GlobalOrdinal,Node> {
+class Level;
 
+template <class Scalar        = SmootherPrototype<>::scalar_type,
+          class LocalOrdinal  = typename SmootherPrototype<Scalar>::local_ordinal_type,
+          class GlobalOrdinal = typename SmootherPrototype<Scalar, LocalOrdinal>::global_ordinal_type,
+          class Node          = typename SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal>::node_type>
+class FakeSmootherPrototype : public SmootherPrototype<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
 #include "MueLu_UseShortNames.hpp"
 
-                 public:
-
-                   FakeSmootherPrototype(int param=0) : param_(param), numOfSetup_(0), numOfSetupCall_(0) {}
+ public:
+  FakeSmootherPrototype(int param = 0)
+    : param_(param)
+    , numOfSetup_(0)
+    , numOfSetupCall_(0) {}
 
-                   virtual ~FakeSmootherPrototype() {}
+  virtual ~FakeSmootherPrototype() {}
 
-                   virtual RCP<SmootherPrototype> Copy() const {
-                     TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == true, Exceptions::RuntimeError, "Not a prototype. Do not copy"); // test not mandatory, but it is the only use case that we need.
-                     return rcp(new FakeSmootherPrototype(*this));
-                   }
+  virtual RCP<SmootherPrototype> Copy() const {
+    TEUCHOS_TEST_FOR_EXCEPTION(SmootherPrototype::IsSetup() == true, Exceptions::RuntimeError, "Not a prototype. Do not copy");  // test not mandatory, but it is the only use case that we need.
+    return rcp(new FakeSmootherPrototype(*this));
+  }
 
-                   void DeclareInput(Level &currentLevel) const { }
+  void DeclareInput(Level& currentLevel) const {}
 
-                   void Setup(Level &) {
-                     numOfSetupCall_++;
-                     if (SmootherPrototype::IsSetup()) return;
+  void Setup(Level&) {
+    numOfSetupCall_++;
+    if (SmootherPrototype::IsSetup()) return;
 
-                     numOfSetup_++;
+    numOfSetup_++;
 
-                     SmootherPrototype::IsSetup(true);
-                   }
+    SmootherPrototype::IsSetup(true);
+  }
 
-                   void Apply(MultiVector& x, const MultiVector& rhs, bool InitialGuessIsZero) const {
-                     TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::NotImplemented, "MueLu::FakeSmootherPrototype()::Apply(): this class is for test purpose only.")
-                   }
+  void Apply(MultiVector& x, const MultiVector& rhs, bool InitialGuessIsZero) const {
+    TEUCHOS_TEST_FOR_EXCEPTION(1, Exceptions::NotImplemented, "MueLu::FakeSmootherPrototype()::Apply(): this class is for test purpose only.")
+  }
 
-                   void SetParam(int param) { param_ = param; }
+  void SetParam(int param) { param_ = param; }
 
-                   int GetParam() const { return param_; }
+  int GetParam() const { return param_; }
 
-                   int GetNumOfSetup() const { return numOfSetup_; }
-                   int GetNumOfSetupCall() const { return numOfSetupCall_; }
+  int GetNumOfSetup() const { return numOfSetup_; }
+  int GetNumOfSetupCall() const { return numOfSetupCall_; }
 
-    size_t getNodeSmootherComplexity() const {return Teuchos::OrdinalTraits<size_t>::invalid();}
+  size_t getNodeSmootherComplexity() const { return Teuchos::OrdinalTraits<size_t>::invalid(); }
 
-    
-                 private:
-                   int param_;
+ private:
+  int param_;
 
-                   int numOfSetup_;
-                   int numOfSetupCall_;
-               };
+  int numOfSetup_;
+  int numOfSetupCall_;
+};
 
-}
+}  // namespace MueLu
 
 #define MUELU_FAKESMOOTHERPROTOTYPE_SHORT
 
-#endif //ifndef MUELU_FAKESMOOTHERPROTOTYPE_HPP
+#endif  // ifndef MUELU_FAKESMOOTHERPROTOTYPE_HPP
diff --git a/packages/muelu/test/unit_tests_kokkos/Aggregates_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/Aggregates_kokkos.cpp
index 1b4d40b43b53..88cfbfa11208 100644
--- a/packages/muelu/test/unit_tests_kokkos/Aggregates_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/Aggregates_kokkos.cpp
@@ -77,830 +77,809 @@
 
 namespace MueLuTests {
 
-  // Little utility to generate uncoupled aggregates.
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  void gimmeUncoupledAggregates(const Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                       RCP<MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node> >& graph,
-                                       Teuchos::RCP<MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node> >& aggregates,
-                                       bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true) {
-#   include "MueLu_UseShortNames.hpp"
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    level.Request("Graph", dropFact.get());
-    level.Request(*dropFact);
-    dropFact->Build(level);
-    graph = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
-    const LO numNodes = graph->GetNodeNumVertices();
-    aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("UC");
-
-    using graph_t = typename LWGraph_kokkos::local_graph_type;
-    using device_type = typename graph_t::device_type;
-    using KernelHandle = KokkosKernels::Experimental::
+// Little utility to generate uncoupled aggregates.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+void gimmeUncoupledAggregates(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                              RCP<MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node>>& graph,
+                              Teuchos::RCP<MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node>>& aggregates,
+                              bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true) {
+#include "MueLu_UseShortNames.hpp"
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  level.Request("Graph", dropFact.get());
+  level.Request(*dropFact);
+  dropFact->Build(level);
+  graph             = level.Get<RCP<LWGraph_kokkos>>("Graph", dropFact.get());
+  const LO numNodes = graph->GetNodeNumVertices();
+  aggregates        = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("UC");
+
+  using graph_t      = typename LWGraph_kokkos::local_graph_type;
+  using device_type  = typename graph_t::device_type;
+  using KernelHandle = KokkosKernels::Experimental::
       KokkosKernelsHandle<typename graph_t::row_map_type::value_type,
                           typename graph_t::entries_type::value_type,
                           typename graph_t::entries_type::value_type,
                           typename graph_t::device_type::execution_space,
                           typename graph_t::device_type::memory_space,
                           typename graph_t::device_type::memory_space>;
-    KernelHandle kh;
-    //Leave gc algorithm choice as the default:
-    //COLORING_D2_SERIAL for Serial execspace, and COLORING_D2_NB_BIT otherwise.
-    kh.create_distance2_graph_coloring_handle();
-
-    // get the distance-2 graph coloring handle
-    auto coloringHandle = kh.get_distance2_graph_coloring_handle();
-
-    //Create device views for graph rowptrs/colinds
-    typename graph_t::row_map_type aRowptrs = graph->getLocalLWGraph().getRowPtrs();
-    typename graph_t::entries_type aColinds = graph->getLocalLWGraph().getEntries();
-
-    //run d2 graph coloring
-    //graph is symmetric so row map/entries and col map/entries are the same
-    KokkosGraph::Experimental::graph_color_distance2(&kh, numNodes, aRowptrs, aColinds);
-
-    // extract the colors and store them in the aggregates
-    aggregates->SetGraphColors(coloringHandle->get_vertex_colors());
-    aggregates->SetGraphNumColors(static_cast<LO>(coloringHandle->get_num_colors()));
-
-    LO numNonAggregatedNodes = 0;
-    Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNodes);
-    Kokkos::deep_copy(aggStat, MueLu::READY);
-    Teuchos::ParameterList params;
-    params.set<int> ("aggregation: min agg size", 1);
-    params.set<int> ("aggregation: max agg size", 3);
-    params.set<bool>("aggregation: deterministic", false);
-
-    params.set<bool>("aggregation: match ML phase2a", true);
-    params.set<bool>("aggregation: error on nodes with no on-rank neighbors", false);
-    params.set<bool>("aggregation: phase3 avoid singletons", false);
-
-    if(bPhase1) {
-      RCP<MueLu::AggregationAlgorithmBase_kokkos<LO,GO,NO> > phase1
-        = rcp(new AggregationPhase1Algorithm_kokkos(dropFact));
-      phase1->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-    }
-    if(bPhase2a) {
-      RCP<MueLu::AggregationAlgorithmBase_kokkos<LO,GO,NO> > phase2a
-        = rcp(new AggregationPhase2aAlgorithm_kokkos(dropFact));
-      phase2a->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-    }
-    if(bPhase2b) {
-      RCP<MueLu::AggregationAlgorithmBase_kokkos<LO,GO,NO> > phase2b
-        = rcp(new AggregationPhase2bAlgorithm_kokkos(dropFact));
-      phase2b->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-    }
-    if(bPhase3) {
-      RCP<MueLu::AggregationAlgorithmBase_kokkos<LO,GO,NO> > phase3
-        = rcp(new AggregationPhase3Algorithm_kokkos(dropFact));
-      phase3->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
-    }
-    aggregates->AggregatesCrossProcessors(false);
-    aggregates->ComputeAggregateSizes(true/*forceRecompute*/);
-    level.Release("Graph", dropFact.get());
+  KernelHandle kh;
+  // Leave gc algorithm choice as the default:
+  // COLORING_D2_SERIAL for Serial execspace, and COLORING_D2_NB_BIT otherwise.
+  kh.create_distance2_graph_coloring_handle();
+
+  // get the distance-2 graph coloring handle
+  auto coloringHandle = kh.get_distance2_graph_coloring_handle();
+
+  // Create device views for graph rowptrs/colinds
+  typename graph_t::row_map_type aRowptrs = graph->getLocalLWGraph().getRowPtrs();
+  typename graph_t::entries_type aColinds = graph->getLocalLWGraph().getEntries();
+
+  // run d2 graph coloring
+  // graph is symmetric so row map/entries and col map/entries are the same
+  KokkosGraph::Experimental::graph_color_distance2(&kh, numNodes, aRowptrs, aColinds);
+
+  // extract the colors and store them in the aggregates
+  aggregates->SetGraphColors(coloringHandle->get_vertex_colors());
+  aggregates->SetGraphNumColors(static_cast<LO>(coloringHandle->get_num_colors()));
+
+  LO numNonAggregatedNodes = 0;
+  Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNodes);
+  Kokkos::deep_copy(aggStat, MueLu::READY);
+  Teuchos::ParameterList params;
+  params.set<int>("aggregation: min agg size", 1);
+  params.set<int>("aggregation: max agg size", 3);
+  params.set<bool>("aggregation: deterministic", false);
+
+  params.set<bool>("aggregation: match ML phase2a", true);
+  params.set<bool>("aggregation: error on nodes with no on-rank neighbors", false);
+  params.set<bool>("aggregation: phase3 avoid singletons", false);
+
+  if (bPhase1) {
+    RCP<MueLu::AggregationAlgorithmBase_kokkos<LO, GO, NO>> phase1 = rcp(new AggregationPhase1Algorithm_kokkos(dropFact));
+    phase1->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
   }
+  if (bPhase2a) {
+    RCP<MueLu::AggregationAlgorithmBase_kokkos<LO, GO, NO>> phase2a = rcp(new AggregationPhase2aAlgorithm_kokkos(dropFact));
+    phase2a->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+  }
+  if (bPhase2b) {
+    RCP<MueLu::AggregationAlgorithmBase_kokkos<LO, GO, NO>> phase2b = rcp(new AggregationPhase2bAlgorithm_kokkos(dropFact));
+    phase2b->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+  }
+  if (bPhase3) {
+    RCP<MueLu::AggregationAlgorithmBase_kokkos<LO, GO, NO>> phase3 = rcp(new AggregationPhase3Algorithm_kokkos(dropFact));
+    phase3->BuildAggregates(params, *graph, *aggregates, aggStat, numNonAggregatedNodes);
+  }
+  aggregates->AggregatesCrossProcessors(false);
+  aggregates->ComputeAggregateSizes(true /*forceRecompute*/);
+  level.Release("Graph", dropFact.get());
+}
 
-  // Little utility to generate uncoupled aggregates.
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node> >
-  gimmeUncoupledAggregates(const Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-                                  bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true) {
-#   include "MueLu_UseShortNames.hpp"
-    RCP<LWGraph_kokkos> graph;
-    RCP<Aggregates> aggregates;
-
-    gimmeUncoupledAggregates(A, graph, aggregates, bPhase1, bPhase2a, bPhase2b, bPhase3);
+// Little utility to generate uncoupled aggregates.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node>>
+gimmeUncoupledAggregates(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>>& A,
+                         bool bPhase1 = true, bool bPhase2a = true, bool bPhase2b = true, bool bPhase3 = true) {
+#include "MueLu_UseShortNames.hpp"
+  RCP<LWGraph_kokkos> graph;
+  RCP<Aggregates> aggregates;
 
-    return aggregates;
-  }
+  gimmeUncoupledAggregates(A, graph, aggregates, bPhase1, bPhase2a, bPhase2b, bPhase3);
 
-  template <class LocalOrdinal, class GlobalOrdinal, class Node>
-  LocalOrdinal checkAggregatesContiguous(MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node> graph,
-                                         MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node> aggregates) {
-    using LO = LocalOrdinal;
-    using GO = GlobalOrdinal;
-    using Aggregates = MueLu::Aggregates<LO, GO, Node>;
-    using LWGraph_kokkos    = MueLu::LWGraph_kokkos<LO, GO, Node>;
-    using execution_space   = typename LWGraph_kokkos::execution_space;
-    using memory_space      = typename LWGraph_kokkos::memory_space;
-    using device_type       = typename LWGraph_kokkos::device_type;
-
-    const LO numNodes = graph.GetNodeNumVertices();
-    auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    auto aggSizes     = aggregates.ComputeAggregateSizes(true);
-
-    auto lclLWGraph = graph.getLocalLWGraph();
-
-    Kokkos::View<LO*, device_type> discontiguousAggs("discontiguous aggregates",
-						     aggregates.GetNumAggregates());
-    Kokkos::parallel_for("Mark discontiguous aggregates",
-                         Kokkos::RangePolicy<LO, execution_space>(0, numNodes),
-                         KOKKOS_LAMBDA(const LO nodeIdx) {
-                           const LO myAggId   = vertex2AggId(nodeIdx, 0);
-                           // Check that the node is actually aggregated
-                           if(myAggId == -1) {return;}
-                           const LO myAggSize = aggSizes(myAggId);
-
-                           if(myAggSize == 1) {
-                             // Can't have a discontiguous singleton
-                             return;
-                           } else {
-                             auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
-                             for(LO neigh = 0; neigh < neighbors.length; ++neigh) {
-                               const LO neighIdx   = neighbors(neigh);
-                               const LO neighAggId = vertex2AggId(neighIdx, 0);
-                               if((nodeIdx != neighIdx) && (neighAggId == myAggId)) {
-                                 // This aggregate might be discontiguous
-                                 // but at least not because of this node
-                                 return;
-                               }
-                             }
-                             discontiguousAggs(myAggId) = 1;
-                           }
-                         });
-
-    LO numDiscontiguousAggregates = 0;
-    Kokkos::parallel_reduce("Count discontiguous aggregates",
-                            Kokkos::RangePolicy<LO, execution_space>(0, aggregates.GetNumAggregates()),
-                            KOKKOS_LAMBDA(const LO aggIdx, LO& numDiscontiguous) {
-                              if(discontiguousAggs(aggIdx) == 1) {
-                                ++numDiscontiguous;
-                              }
-                            }, numDiscontiguousAggregates);
-
-    return numDiscontiguousAggregates;
-  }
+  return aggregates;
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustUncoupledAggregationFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetOrdering("graph");
-    TEST_EQUALITY(aggFact->GetOrdering()== "graph",             true);
-    aggFact->SetOrdering("natural");
-    TEST_EQUALITY(aggFact->GetOrdering()== "natural",           true);
-    aggFact->SetOrdering("random");
-    TEST_EQUALITY(aggFact->GetOrdering()== "random",            true);
-
-    aggFact->SetMaxNeighAlreadySelected(12);
-    TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 12,   true);
-
-    aggFact->SetMaxNeighAlreadySelected(0);
-    TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 0,   true);
-
-    aggFact->SetMinNodesPerAggregate(0);
-    TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 0,      true);
-
-    aggFact->SetMinNodesPerAggregate(3);
-    TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 3,      true);
+template <class LocalOrdinal, class GlobalOrdinal, class Node>
+LocalOrdinal checkAggregatesContiguous(MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node> graph,
+                                       MueLu::Aggregates<LocalOrdinal, GlobalOrdinal, Node> aggregates) {
+  using LO              = LocalOrdinal;
+  using GO              = GlobalOrdinal;
+  using Aggregates      = MueLu::Aggregates<LO, GO, Node>;
+  using LWGraph_kokkos  = MueLu::LWGraph_kokkos<LO, GO, Node>;
+  using execution_space = typename LWGraph_kokkos::execution_space;
+  using memory_space    = typename LWGraph_kokkos::memory_space;
+  using device_type     = typename LWGraph_kokkos::device_type;
+
+  const LO numNodes = graph.GetNodeNumVertices();
+  auto vertex2AggId = aggregates.GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto aggSizes     = aggregates.ComputeAggregateSizes(true);
+
+  auto lclLWGraph = graph.getLocalLWGraph();
+
+  Kokkos::View<LO*, device_type> discontiguousAggs("discontiguous aggregates",
+                                                   aggregates.GetNumAggregates());
+  Kokkos::parallel_for(
+      "Mark discontiguous aggregates",
+      Kokkos::RangePolicy<LO, execution_space>(0, numNodes),
+      KOKKOS_LAMBDA(const LO nodeIdx) {
+        const LO myAggId = vertex2AggId(nodeIdx, 0);
+        // Check that the node is actually aggregated
+        if (myAggId == -1) {
+          return;
+        }
+        const LO myAggSize = aggSizes(myAggId);
+
+        if (myAggSize == 1) {
+          // Can't have a discontiguous singleton
+          return;
+        } else {
+          auto neighbors = lclLWGraph.getNeighborVertices(nodeIdx);
+          for (LO neigh = 0; neigh < neighbors.length; ++neigh) {
+            const LO neighIdx   = neighbors(neigh);
+            const LO neighAggId = vertex2AggId(neighIdx, 0);
+            if ((nodeIdx != neighIdx) && (neighAggId == myAggId)) {
+              // This aggregate might be discontiguous
+              // but at least not because of this node
+              return;
+            }
+          }
+          discontiguousAggs(myAggId) = 1;
+        }
+      });
+
+  LO numDiscontiguousAggregates = 0;
+  Kokkos::parallel_reduce(
+      "Count discontiguous aggregates",
+      Kokkos::RangePolicy<LO, execution_space>(0, aggregates.GetNumAggregates()),
+      KOKKOS_LAMBDA(const LO aggIdx, LO& numDiscontiguous) {
+        if (discontiguousAggs(aggIdx) == 1) {
+          ++numDiscontiguous;
+        }
+      },
+      numDiscontiguousAggregates);
+
+  return numDiscontiguousAggregates;
 }
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustUncoupledAggregationFactory, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetOrdering("graph");
+  TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
+  aggFact->SetOrdering("natural");
+  TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
+  aggFact->SetOrdering("random");
+  TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
 
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  aggFact->SetMaxNeighAlreadySelected(12);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 12, true);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  aggFact->SetMaxNeighAlreadySelected(0);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 0, true);
 
-    RCP<Aggregates> aggregates = gimmeUncoupledAggregates(A);
+  aggFact->SetMinNodesPerAggregate(0);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 0, true);
 
-    TEST_EQUALITY(aggregates != Teuchos::null,              true);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),  false);
+  aggFact->SetMinNodesPerAggregate(3);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 3, true);
+}
 
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2UncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
-    //because it can't take a custom parameter list, add that as parameter?
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: max agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: min agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
-    aggFact->SetParameter("aggregation: preserve Dirichlet points",Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 1",         Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2a",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2b",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 3",         Teuchos::ParameterEntry(true));
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-    TEST_INEQUALITY(aggregates, Teuchos::null);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
-    amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo",amalgFact.get()); // fix me
-    level.Release("UnAmalgamationInfo", amalgFact.get());
-    level.Release("Aggregates", aggFact.get());
-  }
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2PreserveUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
-    //because it can't take a custom parameter list, add that as parameter?
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: max agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: min agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
-    aggFact->SetParameter("aggregation: enable phase 1",         Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2a",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2b",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 3",         Teuchos::ParameterEntry(true));
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-    TEST_INEQUALITY(aggregates, Teuchos::null);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
-    amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo",amalgFact.get()); // fix me
-    level.Release("UnAmalgamationInfo", amalgFact.get());
-    level.Release("Aggregates", aggFact.get());
-  }
+  RCP<Aggregates> aggregates = gimmeUncoupledAggregates(A);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustOnePt2UncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
-    //because it can't take a custom parameter list, add that as parameter?
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: max agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: min agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
-    aggFact->SetParameter("aggregation: allow user-specified singletons", Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 1",         Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2a",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2b",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 3",         Teuchos::ParameterEntry(true));
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-    TEST_INEQUALITY(aggregates, Teuchos::null);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
-    amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo",amalgFact.get()); // fix me
-    level.Release("UnAmalgamationInfo", amalgFact.get());
-    level.Release("Aggregates", aggFact.get());
-  }
+  TEST_EQUALITY(aggregates != Teuchos::null, true);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+}
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2UncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
+  // because it can't take a custom parameter list, add that as parameter?
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+  aggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());  // fix me
+  TEST_INEQUALITY(aggregates, Teuchos::null);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  amalgInfo = level.Get<RCP<AmalgamationInfo>>("UnAmalgamationInfo", amalgFact.get());  // fix me
+  level.Release("UnAmalgamationInfo", amalgFact.get());
+  level.Release("Aggregates", aggFact.get());
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2DeterUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    //TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
-    //because it can't take a custom parameter list, add that as parameter?
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: max agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: min agg size",           Teuchos::ParameterEntry(3));
-    aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
-    aggFact->SetParameter("aggregation: deterministic",          Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 1",         Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2a",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 2b",        Teuchos::ParameterEntry(true));
-    aggFact->SetParameter("aggregation: enable phase 3",         Teuchos::ParameterEntry(true));
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-
-    // Set some node to ONEPT state
-
-
-
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get()); // fix me
-    TEST_INEQUALITY(aggregates, Teuchos::null);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
-    amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo",amalgFact.get()); // fix me
-    level.Release("UnAmalgamationInfo", amalgFact.get());
-    level.Release("Aggregates", aggFact.get());
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2PreserveUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
+  // because it can't take a custom parameter list, add that as parameter?
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+  aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());  // fix me
+  TEST_INEQUALITY(aggregates, Teuchos::null);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  amalgInfo = level.Get<RCP<AmalgamationInfo>>("UnAmalgamationInfo", amalgFact.get());  // fix me
+  level.Release("UnAmalgamationInfo", amalgFact.get());
+  level.Release("Aggregates", aggFact.get());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustOnePt2UncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
+  // because it can't take a custom parameter list, add that as parameter?
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+  aggFact->SetParameter("aggregation: allow user-specified singletons", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());  // fix me
+  TEST_INEQUALITY(aggregates, Teuchos::null);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  amalgInfo = level.Get<RCP<AmalgamationInfo>>("UnAmalgamationInfo", amalgFact.get());  // fix me
+  level.Release("UnAmalgamationInfo", amalgFact.get());
+  level.Release("Aggregates", aggFact.get());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, JustDist2DeterUncoupledAggregation, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // TODO bmk: A lot of test code duplicated here from gimmeUncoupledAggregates
+  // because it can't take a custom parameter list, add that as parameter?
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: max agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: min agg size", Teuchos::ParameterEntry(3));
+  aggFact->SetParameter("aggregation: max selected neighbors", Teuchos::ParameterEntry(0));
+  aggFact->SetParameter("aggregation: deterministic", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 1", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2a", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 2b", Teuchos::ParameterEntry(true));
+  aggFact->SetParameter("aggregation: enable phase 3", Teuchos::ParameterEntry(true));
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+
+  // Set some node to ONEPT state
+
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());  // fix me
+  TEST_INEQUALITY(aggregates, Teuchos::null);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  amalgInfo = level.Get<RCP<AmalgamationInfo>>("UnAmalgamationInfo", amalgFact.get());  // fix me
+  level.Release("UnAmalgamationInfo", amalgFact.get());
+  level.Release("Aggregates", aggFact.get());
+}
 
 // A test that creates discontiguous aggregates to make sure the detection algorithm works well
 /// pretty much testing the test...
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, DiscontiguousAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using device_type  = typename Aggregates::device_type;
-    using memory_space = typename device_type::memory_space;
-
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(3*comm->getSize());
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<Aggregates> aggregates;
-    RCP<LWGraph_kokkos> graph;
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    level.Request("Graph", dropFact.get());
-    level.Request(*dropFact);
-    dropFact->Build(level);
-    graph = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
-    RCP<const Map> importMap = graph->GetImportMap();
-    const LO numNodes = graph->GetNodeNumVertices();
-    aggregates = rcp(new Aggregates(*graph));
-    aggregates->setObjectLabel("UC");
-
-    Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNodes);
-    Kokkos::deep_copy(aggStat, MueLu::READY);
-
-    // Performing fake aggregates to generate a discontiguous aggregate
-    Kokkos::View<LO**, Kokkos::LayoutLeft, device_type> vertex2AggId = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-    Kokkos::View<LO**, Kokkos::LayoutLeft, device_type> procWinner   = aggregates->GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
-
-    typename Kokkos::View<LO**, Kokkos::LayoutLeft, device_type>::HostMirror vertex2AggId_h
-      = Kokkos::create_mirror_view(vertex2AggId);
-    Kokkos::deep_copy(vertex2AggId_h, vertex2AggId);
-    vertex2AggId_h(0, 0) = 0;
-    vertex2AggId_h(1, 0) = 1;
-    vertex2AggId_h(2, 0) = 0;
-    Kokkos::deep_copy(vertex2AggId, vertex2AggId_h);
-
-    typename Kokkos::View<LO**, Kokkos::LayoutLeft, device_type>::HostMirror procWinner_h
-      = Kokkos::create_mirror_view(procWinner);
-    Kokkos::deep_copy(procWinner_h, procWinner);
-    procWinner_h(0, 0) = comm->getRank();
-    procWinner_h(1, 0) = comm->getRank();
-    procWinner_h(2, 0) = comm->getRank();
-    Kokkos::deep_copy(procWinner, procWinner_h);
-
-    aggregates->SetNumAggregates(2);
-    aggregates->AggregatesCrossProcessors(false);
-    aggregates->ComputeAggregateSizes(true);
-
-    const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
-    TEST_EQUALITY(numDiscontiguous, 1);
-  } //UncoupledPhase1
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase1, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<AmalgamationInfo> amalgInfo;
-
-    RCP<Aggregates> aggregates;
-    RCP<LWGraph_kokkos> graph;
-    gimmeUncoupledAggregates(A, graph, aggregates, true, false, false, false);
-    const GO numAggs = aggregates->GetNumAggregates();
-
-    TEST_INEQUALITY(numAggs, 0);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
-
-    typename Aggregates::aggregates_sizes_type::const_type aggSizes
-      = aggregates->ComputeAggregateSizes(true);
-
-    LO numBadAggregates = 0;
-    Kokkos::parallel_reduce("Checking aggregates sizes",
-                            Kokkos::RangePolicy<LO, typename Aggregates::execution_space>(0, aggSizes.extent(0)),
-                            KOKKOS_LAMBDA(const LO aggIdx, LO& lNumBadAggregates) {
-                              if ((aggSizes(aggIdx) < 1) || (3 < aggSizes(aggIdx))) {
-                                lNumBadAggregates += 1;
-                              }
-                            }, numBadAggregates);
-    TEST_EQUALITY(numBadAggregates, 0);
-
-    const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
-    TEST_EQUALITY(numDiscontiguous, 0);
-  } //UncoupledPhase1
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase2, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-
-    RCP<LWGraph_kokkos> graph;
-    RCP<Aggregates> aggregates;
-    gimmeUncoupledAggregates(A, graph, aggregates, false, true, true, false);
-    GO numAggs = aggregates->GetNumAggregates();
-
-    TEST_INEQUALITY(numAggs, 0);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
-    TEST_EQUALITY(numDiscontiguous, 0);
-  } //UncoupledPhase2
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase3, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
-    RCP<AmalgamationInfo> amalgInfo;
-
-    RCP<Aggregates> aggregates = gimmeUncoupledAggregates(A, false, false, false, true);
-    GO numAggs = aggregates->GetNumAggregates();
-
-    TEST_INEQUALITY(numAggs, 0);
-    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-
-    typename Aggregates::aggregates_sizes_type::const_type aggSizes
-      = aggregates->ComputeAggregateSizes(true);
-
-    LO numBadAggregates = 0;
-    Kokkos::parallel_reduce("Checking aggregates sizes",
-                            Kokkos::RangePolicy<LO, typename Aggregates::execution_space>(0, aggSizes.extent(0)),
-                            KOKKOS_LAMBDA(const LO aggIdx, LO& lNumBadAggregates) {
-                              if ((aggSizes(aggIdx) < 1) || (5 < aggSizes(aggIdx))) {
-                                lNumBadAggregates += 1;
-                              }
-                            }, numBadAggregates);
-    TEST_EQUALITY(numBadAggregates, 0);
-
-    // Check ComputeNodesInAggregate
-    typename Aggregates::LO_view aggPtr, aggNodes, unaggregated;
-    aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
-    TEST_EQUALITY(aggPtr.extent_int(0), numAggs+1);
-    // TEST_EQUALITY(unaggregated.extent_int(0), 0); // 1 unaggregated node in the MPI_4 case
-    
-    // test aggPtr(i)+aggSizes(i)=aggPtr(i+1)
-    typename Aggregates::LO_view::HostMirror aggPtr_h = Kokkos::create_mirror_view(aggPtr);
-    typename Aggregates::aggregates_sizes_type::HostMirror aggSizes_h = Kokkos::create_mirror_view(aggSizes);
-    Kokkos::deep_copy(aggPtr_h, aggPtr);
-    Kokkos::deep_copy(aggSizes_h, aggSizes);
-    for(LO i=0; i<aggSizes_h.extent_int(0); ++i)
-      TEST_EQUALITY(aggPtr_h(i)+aggSizes_h(i), aggPtr_h(i+1));
-
-  } //UncoupledPhase3
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, AllowDroppingToCreateAdditionalDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test option that allows dropping during aggregation to create new Dirichlet rows" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-    using device_type  = typename Aggregates::device_type;
-
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    A->resumeFill();
-
-    // Create one row on every processor with small off-diagonal entries that will be dropped with
-    // an appropriately chosen threshold.  Avoid domain boundaries.
-    LocalOrdinal localRowToModify = 1;
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar>  values;
-    A->getLocalRowView(localRowToModify, indices, values);
-    Array<Scalar> newvalues(values.size(),TST::zero());
-    for (int j = 0; j < indices.size(); j++) {
-      if (indices[j] == localRowToModify)
-        newvalues[j] = values[j]; //keep diagonal unmodified
-      else
-        newvalues[j] = -TST::eps();
-    }
-    A->replaceLocalValues(localRowToModify,indices,newvalues);
-    A->fillComplete();
-
-    // Dropping connections will not lead to the creation of new Dirichlet rows.
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetParameter("aggregation: dropping may create Dirichlet",Teuchos::ParameterEntry(false));
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(Teuchos::as<double>(-100*TST::eps())));
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    level.Request("Graph",dropFact.get());
-
-    // Setup aggregation factory (use default factory for graph)
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    RCP<LWGraph_kokkos> graph = level.Get<RCP<LWGraph_kokkos>>("Graph",dropFact.get());
-    Kokkos::View<const bool*, device_type> dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
-    int numDirichletRows=0;
-    using execution_space   = typename LWGraph_kokkos::execution_space;
-    LO numRows = graph->GetNodeNumVertices();
-    Kokkos::parallel_reduce("Count Dirichlet rows",
-                            Kokkos::RangePolicy<LO, execution_space>(0, numRows),
-                            KOKKOS_LAMBDA(const LO rowIdx, LO& numDirichlet) {
-                              if(dirichletBoundaryMap[rowIdx]) {
-                                ++numDirichlet;
-                              }
-                            }, numDirichletRows);
-    TEST_EQUALITY(numDirichletRows, 0);
-
-    Array< LO > aggPtr;
-    Array< LO > aggNodes;
-    Array< LO > unaggregated;
-
-    // Repeat with "aggregation: dropping may create Dirichlet" = TRUE, i.e.,
-    // dropping connections may lead to the creation of new Dirichlet rows.
-    // The second row should be flagged as Dirichlet because all off-diagonal entries are dropped.
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetParameter("aggregation: dropping may create Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(Teuchos::as<double>(-100*TST::eps())));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    level.Request("Graph",dropFact.get());
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    graph = level.Get<RCP<LWGraph_kokkos>>("Graph",dropFact.get());
-    dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
-    numDirichletRows=0;
-    Kokkos::parallel_reduce("Count Dirichlet rows",
-                            Kokkos::RangePolicy<LO, execution_space>(0, numRows),
-                            KOKKOS_LAMBDA(const LO rowIdx, LO& numDirichlet) {
-                              if(dirichletBoundaryMap[rowIdx]) {
-                                ++numDirichlet;
-                              }
-                            }, numDirichletRows);
-    TEST_EQUALITY(numDirichletRows,1);
-
-  } //AllowDroppingToCreateAdditionalDirichletRows
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, GreedyDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "running greedy Dirichlet" << std::endl;
-
-    typedef typename Teuchos::ScalarTraits<Scalar> TST;
-
-//    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(30);
-    // Make a Matrix with multiple degrees of freedom per node
-    GlobalOrdinal nx = 8, ny = 8;
-
-    // Describes the initial layout of matrix rows across processors.
-    Teuchos::ParameterList galeriList;
-    galeriList.set("nx", nx);
-    galeriList.set("ny", ny);
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    RCP<const Map> map = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers_kokkos::Parameters::getLib(), "Cartesian2D", comm, galeriList);
-
-    map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, DiscontiguousAggregates, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using device_type  = typename Aggregates::device_type;
+  using memory_space = typename device_type::memory_space;
+
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  RCP<Matrix> A                      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(3 * comm->getSize());
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<Aggregates> aggregates;
+  RCP<LWGraph_kokkos> graph;
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  level.Request("Graph", dropFact.get());
+  level.Request(*dropFact);
+  dropFact->Build(level);
+  graph                    = level.Get<RCP<LWGraph_kokkos>>("Graph", dropFact.get());
+  RCP<const Map> importMap = graph->GetImportMap();
+  const LO numNodes        = graph->GetNodeNumVertices();
+  aggregates               = rcp(new Aggregates(*graph));
+  aggregates->setObjectLabel("UC");
+
+  Kokkos::View<unsigned*, device_type> aggStat("aggStat", numNodes);
+  Kokkos::deep_copy(aggStat, MueLu::READY);
+
+  // Performing fake aggregates to generate a discontiguous aggregate
+  Kokkos::View<LO**, Kokkos::LayoutLeft, device_type> vertex2AggId = aggregates->GetVertex2AggId()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+  Kokkos::View<LO**, Kokkos::LayoutLeft, device_type> procWinner   = aggregates->GetProcWinner()->getDeviceLocalView(Xpetra::Access::ReadWrite);
+
+  typename Kokkos::View<LO**, Kokkos::LayoutLeft, device_type>::HostMirror vertex2AggId_h = Kokkos::create_mirror_view(vertex2AggId);
+  Kokkos::deep_copy(vertex2AggId_h, vertex2AggId);
+  vertex2AggId_h(0, 0) = 0;
+  vertex2AggId_h(1, 0) = 1;
+  vertex2AggId_h(2, 0) = 0;
+  Kokkos::deep_copy(vertex2AggId, vertex2AggId_h);
+
+  typename Kokkos::View<LO**, Kokkos::LayoutLeft, device_type>::HostMirror procWinner_h = Kokkos::create_mirror_view(procWinner);
+  Kokkos::deep_copy(procWinner_h, procWinner);
+  procWinner_h(0, 0) = comm->getRank();
+  procWinner_h(1, 0) = comm->getRank();
+  procWinner_h(2, 0) = comm->getRank();
+  Kokkos::deep_copy(procWinner, procWinner_h);
+
+  aggregates->SetNumAggregates(2);
+  aggregates->AggregatesCrossProcessors(false);
+  aggregates->ComputeAggregateSizes(true);
+
+  const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
+  TEST_EQUALITY(numDiscontiguous, 1);
+}  // UncoupledPhase1
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase1, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<AmalgamationInfo> amalgInfo;
+
+  RCP<Aggregates> aggregates;
+  RCP<LWGraph_kokkos> graph;
+  gimmeUncoupledAggregates(A, graph, aggregates, true, false, false, false);
+  const GO numAggs = aggregates->GetNumAggregates();
+
+  TEST_INEQUALITY(numAggs, 0);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  typename Aggregates::aggregates_sizes_type::const_type aggSizes = aggregates->ComputeAggregateSizes(true);
+
+  LO numBadAggregates = 0;
+  Kokkos::parallel_reduce(
+      "Checking aggregates sizes",
+      Kokkos::RangePolicy<LO, typename Aggregates::execution_space>(0, aggSizes.extent(0)),
+      KOKKOS_LAMBDA(const LO aggIdx, LO& lNumBadAggregates) {
+        if ((aggSizes(aggIdx) < 1) || (3 < aggSizes(aggIdx))) {
+          lNumBadAggregates += 1;
+        }
+      },
+      numBadAggregates);
+  TEST_EQUALITY(numBadAggregates, 0);
+
+  const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
+  TEST_EQUALITY(numDiscontiguous, 0);
+}  // UncoupledPhase1
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase2, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+
+  RCP<LWGraph_kokkos> graph;
+  RCP<Aggregates> aggregates;
+  gimmeUncoupledAggregates(A, graph, aggregates, false, true, true, false);
+  GO numAggs = aggregates->GetNumAggregates();
+
+  TEST_INEQUALITY(numAggs, 0);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  const LO numDiscontiguous = checkAggregatesContiguous(*graph, *aggregates);
+  TEST_EQUALITY(numDiscontiguous, 0);
+}  // UncoupledPhase2
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, UncoupledPhase3, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  RCP<AmalgamationInfo> amalgInfo;
+
+  RCP<Aggregates> aggregates = gimmeUncoupledAggregates(A, false, false, false, true);
+  GO numAggs                 = aggregates->GetNumAggregates();
+
+  TEST_INEQUALITY(numAggs, 0);
+  TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+
+  typename Aggregates::aggregates_sizes_type::const_type aggSizes = aggregates->ComputeAggregateSizes(true);
+
+  LO numBadAggregates = 0;
+  Kokkos::parallel_reduce(
+      "Checking aggregates sizes",
+      Kokkos::RangePolicy<LO, typename Aggregates::execution_space>(0, aggSizes.extent(0)),
+      KOKKOS_LAMBDA(const LO aggIdx, LO& lNumBadAggregates) {
+        if ((aggSizes(aggIdx) < 1) || (5 < aggSizes(aggIdx))) {
+          lNumBadAggregates += 1;
+        }
+      },
+      numBadAggregates);
+  TEST_EQUALITY(numBadAggregates, 0);
+
+  // Check ComputeNodesInAggregate
+  typename Aggregates::LO_view aggPtr, aggNodes, unaggregated;
+  aggregates->ComputeNodesInAggregate(aggPtr, aggNodes, unaggregated);
+  TEST_EQUALITY(aggPtr.extent_int(0), numAggs + 1);
+  // TEST_EQUALITY(unaggregated.extent_int(0), 0); // 1 unaggregated node in the MPI_4 case
+
+  // test aggPtr(i)+aggSizes(i)=aggPtr(i+1)
+  typename Aggregates::LO_view::HostMirror aggPtr_h                 = Kokkos::create_mirror_view(aggPtr);
+  typename Aggregates::aggregates_sizes_type::HostMirror aggSizes_h = Kokkos::create_mirror_view(aggSizes);
+  Kokkos::deep_copy(aggPtr_h, aggPtr);
+  Kokkos::deep_copy(aggSizes_h, aggSizes);
+  for (LO i = 0; i < aggSizes_h.extent_int(0); ++i)
+    TEST_EQUALITY(aggPtr_h(i) + aggSizes_h(i), aggPtr_h(i + 1));
+
+}  // UncoupledPhase3
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, AllowDroppingToCreateAdditionalDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test option that allows dropping during aggregation to create new Dirichlet rows" << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+  using device_type = typename Aggregates::device_type;
+
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  A->resumeFill();
+
+  // Create one row on every processor with small off-diagonal entries that will be dropped with
+  // an appropriately chosen threshold.  Avoid domain boundaries.
+  LocalOrdinal localRowToModify = 1;
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> values;
+  A->getLocalRowView(localRowToModify, indices, values);
+  Array<Scalar> newvalues(values.size(), TST::zero());
+  for (int j = 0; j < indices.size(); j++) {
+    if (indices[j] == localRowToModify)
+      newvalues[j] = values[j];  // keep diagonal unmodified
+    else
+      newvalues[j] = -TST::eps();
+  }
+  A->replaceLocalValues(localRowToModify, indices, newvalues);
+  A->fillComplete();
+
+  // Dropping connections will not lead to the creation of new Dirichlet rows.
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetParameter("aggregation: dropping may create Dirichlet", Teuchos::ParameterEntry(false));
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(Teuchos::as<double>(-100 * TST::eps())));
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  level.Request("Graph", dropFact.get());
+
+  // Setup aggregation factory (use default factory for graph)
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates                                  = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  RCP<LWGraph_kokkos> graph                                   = level.Get<RCP<LWGraph_kokkos>>("Graph", dropFact.get());
+  Kokkos::View<const bool*, device_type> dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
+  int numDirichletRows                                        = 0;
+  using execution_space                                       = typename LWGraph_kokkos::execution_space;
+  LO numRows                                                  = graph->GetNodeNumVertices();
+  Kokkos::parallel_reduce(
+      "Count Dirichlet rows",
+      Kokkos::RangePolicy<LO, execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const LO rowIdx, LO& numDirichlet) {
+        if (dirichletBoundaryMap[rowIdx]) {
+          ++numDirichlet;
+        }
+      },
+      numDirichletRows);
+  TEST_EQUALITY(numDirichletRows, 0);
+
+  Array<LO> aggPtr;
+  Array<LO> aggNodes;
+  Array<LO> unaggregated;
+
+  // Repeat with "aggregation: dropping may create Dirichlet" = TRUE, i.e.,
+  // dropping connections may lead to the creation of new Dirichlet rows.
+  // The second row should be flagged as Dirichlet because all off-diagonal entries are dropped.
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetParameter("aggregation: dropping may create Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(Teuchos::as<double>(-100 * TST::eps())));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  level.Request("Graph", dropFact.get());
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  aggregates           = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  graph                = level.Get<RCP<LWGraph_kokkos>>("Graph", dropFact.get());
+  dirichletBoundaryMap = graph->getLocalLWGraph().GetBoundaryNodeMap();
+  numDirichletRows     = 0;
+  Kokkos::parallel_reduce(
+      "Count Dirichlet rows",
+      Kokkos::RangePolicy<LO, execution_space>(0, numRows),
+      KOKKOS_LAMBDA(const LO rowIdx, LO& numDirichlet) {
+        if (dirichletBoundaryMap[rowIdx]) {
+          ++numDirichlet;
+        }
+      },
+      numDirichletRows);
+  TEST_EQUALITY(numDirichletRows, 1);
+
+}  // AllowDroppingToCreateAdditionalDirichletRows
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Aggregates_kokkos, GreedyDirichlet, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "running greedy Dirichlet" << std::endl;
+
+  typedef typename Teuchos::ScalarTraits<Scalar> TST;
+
+  //    RCP<Matrix> A = TestHelpers::TestFactory<SC, LO, GO, NO>::Build1DPoisson(30);
+  // Make a Matrix with multiple degrees of freedom per node
+  GlobalOrdinal nx = 8, ny = 8;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+  RCP<const Teuchos::Comm<int>> comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  RCP<const Map> map                 = Galeri::Xpetra::CreateMap<LocalOrdinal, GlobalOrdinal, Node>(TestHelpers_kokkos::Parameters::getLib(), "Cartesian2D", comm, galeriList);
+
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr =
       Galeri::Xpetra::BuildProblem<Scalar, LocalOrdinal, GlobalOrdinal, Map, CrsMatrixWrap, MultiVector>("Elasticity2D", map, galeriList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    A->SetFixedBlockSize(2);
-
-
-    Teuchos::ArrayView<const LocalOrdinal> indices;
-    Teuchos::ArrayView<const Scalar>  values;
-
-    // Create a dirichlet boundary row.
-    LocalOrdinal localRowToZero = 5; // Corresponds to a Dof on local graph node 2
-
-    A->resumeFill();
-    A->getLocalRowView(localRowToZero, indices, values);
-    Array<Scalar> newvalues(values.size(),TST::zero());
-    for (int j = 0; j < indices.size(); j++)
-      //keep diagonal
-      if (indices[j] == localRowToZero) newvalues[j] = values[j];
-    A->replaceLocalValues(localRowToZero,indices,newvalues);
-
-    A->fillComplete();
-
-    ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
-    TEST_EQUALITY(drows[localRowToZero], true);
-    TEST_EQUALITY(drows[localRowToZero-1], false);
-
-    RCP<AmalgamationInfo> amalgInfo;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-
-    RCP<CoalesceDropFactory_kokkos> dropFact;
-    RCP<AmalgamationFactory> amalgFact;
-    RCP<UncoupledAggregationFactory_kokkos> aggFact;
-
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(false));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-
-    level.Request("Aggregates", aggFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
-
-    level.Request(*aggFact);
-    aggFact->Build(level);
-    RCP<Aggregates> aggregates = level.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    Array< LO > aggPtr;
-    Array< LO > aggNodes;
-    Array< LO > unaggregated;
-
-    typename Aggregates::aggregates_sizes_type::const_type aggSizes
-      = aggregates->ComputeAggregateSizes(true);
-
-    auto vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
-    for( auto i = 0; i<(nx/2*ny/2); i++){
-        TEST_EQUALITY(vertex2AggId(i,0) != MUELU_UNAGGREGATED, true);// check that all nodes are aggregated
-    }
-
-    // Repeat with greedy Dirichlet
-    Level levelGreedyAndNoPreserve;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(levelGreedyAndNoPreserve);
-    levelGreedyAndNoPreserve.Set("A", A);
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-
-    levelGreedyAndNoPreserve.Request("Aggregates", aggFact.get());
-    levelGreedyAndNoPreserve.Request("UnAmalgamationInfo", amalgFact.get());
-
-    levelGreedyAndNoPreserve.Request(*aggFact);
-    aggFact->Build(levelGreedyAndNoPreserve);
-    aggregates = levelGreedyAndNoPreserve.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-    aggFact->SetParameter("aggregation: preserve Dirichlet points",Teuchos::ParameterEntry(false));
-
-    typename Aggregates::aggregates_sizes_type::const_type aggSizesGreedy
-      = aggregates->ComputeAggregateSizes(true);
-    // There should be the same number of aggregates in the greedy dirichlet case as in the standard case
-    // though note the aggregates will be smaller in the greedy dirirchlet case.
-    // This may not be true for all problem setups, but is true for the setup in this test case.
-    TEST_EQUALITY( aggSizesGreedy.extent(0) == aggSizes.extent(0), true)
-
-    vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
-    TEST_EQUALITY(vertex2AggId(2,0) == MUELU_UNAGGREGATED, true);// check that the node with the Dof flagged as dirichlet is unaggregated
-
-    // Repeat with greedy Dirichlet and preserve Dirichlet points
-    Level levelGreedyAndPreserve;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(levelGreedyAndPreserve);
-    levelGreedyAndPreserve.Set("A", A);
-
-    amalgFact = rcp(new AmalgamationFactory());
-    dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetParameter("aggregation: greedy Dirichlet",Teuchos::ParameterEntry(true));
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    // Setup aggregation factory (use default factory for graph)
-    aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    aggFact->SetFactory("Graph", dropFact);
-    aggFact->SetParameter("aggregation: preserve Dirichlet points",Teuchos::ParameterEntry(true));
-
-    levelGreedyAndPreserve.Request("Aggregates", aggFact.get());
-    levelGreedyAndPreserve.Request("UnAmalgamationInfo", amalgFact.get());
-
-    levelGreedyAndPreserve.Request(*aggFact);
-    aggFact->Build(levelGreedyAndPreserve);
-    aggregates = levelGreedyAndPreserve.Get<RCP<Aggregates> >("Aggregates",aggFact.get());
-
-    typename Aggregates::aggregates_sizes_type::const_type aggSizesGreedyPreserve
-      = aggregates->ComputeAggregateSizes(true);
-    // check that dirichlet aggs are preserved
-    // there should be more aggregates in the dirichlet preserved case than in the standard case.
-    // This will always be true.
-    TEST_EQUALITY( aggSizesGreedyPreserve.extent(0) > aggSizesGreedy.extent(0), true)
-
-
-    vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
-    for( auto i = 0; i<(nx/2*ny/2); i++){
-        TEST_EQUALITY(vertex2AggId(i,0) != MUELU_UNAGGREGATED, true);// check that all nodes are aggregated
-    }
-
+  RCP<Matrix> A = Pr->BuildMatrix();
+  A->SetFixedBlockSize(2);
+
+  Teuchos::ArrayView<const LocalOrdinal> indices;
+  Teuchos::ArrayView<const Scalar> values;
+
+  // Create a dirichlet boundary row.
+  LocalOrdinal localRowToZero = 5;  // Corresponds to a Dof on local graph node 2
+
+  A->resumeFill();
+  A->getLocalRowView(localRowToZero, indices, values);
+  Array<Scalar> newvalues(values.size(), TST::zero());
+  for (int j = 0; j < indices.size(); j++)
+    // keep diagonal
+    if (indices[j] == localRowToZero) newvalues[j] = values[j];
+  A->replaceLocalValues(localRowToZero, indices, newvalues);
+
+  A->fillComplete();
+
+  ArrayRCP<const bool> drows = Utilities::DetectDirichletRows(*A);
+  TEST_EQUALITY(drows[localRowToZero], true);
+  TEST_EQUALITY(drows[localRowToZero - 1], false);
+
+  RCP<AmalgamationInfo> amalgInfo;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<CoalesceDropFactory_kokkos> dropFact;
+  RCP<AmalgamationFactory> amalgFact;
+  RCP<UncoupledAggregationFactory_kokkos> aggFact;
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(false));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+
+  level.Request("Aggregates", aggFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*aggFact);
+  aggFact->Build(level);
+  RCP<Aggregates> aggregates = level.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  Array<LO> aggPtr;
+  Array<LO> aggNodes;
+  Array<LO> unaggregated;
+
+  typename Aggregates::aggregates_sizes_type::const_type aggSizes = aggregates->ComputeAggregateSizes(true);
+
+  auto vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
+  for (auto i = 0; i < (nx / 2 * ny / 2); i++) {
+    TEST_EQUALITY(vertex2AggId(i, 0) != MUELU_UNAGGREGATED, true);  // check that all nodes are aggregated
   }
 
+  // Repeat with greedy Dirichlet
+  Level levelGreedyAndNoPreserve;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(levelGreedyAndNoPreserve);
+  levelGreedyAndNoPreserve.Set("A", A);
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+
+  levelGreedyAndNoPreserve.Request("Aggregates", aggFact.get());
+  levelGreedyAndNoPreserve.Request("UnAmalgamationInfo", amalgFact.get());
+
+  levelGreedyAndNoPreserve.Request(*aggFact);
+  aggFact->Build(levelGreedyAndNoPreserve);
+  aggregates = levelGreedyAndNoPreserve.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+  aggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(false));
+
+  typename Aggregates::aggregates_sizes_type::const_type aggSizesGreedy = aggregates->ComputeAggregateSizes(true);
+  // There should be the same number of aggregates in the greedy dirichlet case as in the standard case
+  // though note the aggregates will be smaller in the greedy dirirchlet case.
+  // This may not be true for all problem setups, but is true for the setup in this test case.
+  TEST_EQUALITY(aggSizesGreedy.extent(0) == aggSizes.extent(0), true)
+
+  vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
+  TEST_EQUALITY(vertex2AggId(2, 0) == MUELU_UNAGGREGATED, true);  // check that the node with the Dof flagged as dirichlet is unaggregated
+
+  // Repeat with greedy Dirichlet and preserve Dirichlet points
+  Level levelGreedyAndPreserve;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(levelGreedyAndPreserve);
+  levelGreedyAndPreserve.Set("A", A);
+
+  amalgFact = rcp(new AmalgamationFactory());
+  dropFact  = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetParameter("aggregation: greedy Dirichlet", Teuchos::ParameterEntry(true));
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  // Setup aggregation factory (use default factory for graph)
+  aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: preserve Dirichlet points", Teuchos::ParameterEntry(true));
+
+  levelGreedyAndPreserve.Request("Aggregates", aggFact.get());
+  levelGreedyAndPreserve.Request("UnAmalgamationInfo", amalgFact.get());
+
+  levelGreedyAndPreserve.Request(*aggFact);
+  aggFact->Build(levelGreedyAndPreserve);
+  aggregates = levelGreedyAndPreserve.Get<RCP<Aggregates>>("Aggregates", aggFact.get());
+
+  typename Aggregates::aggregates_sizes_type::const_type aggSizesGreedyPreserve = aggregates->ComputeAggregateSizes(true);
+  // check that dirichlet aggs are preserved
+  // there should be more aggregates in the dirichlet preserved case than in the standard case.
+  // This will always be true.
+  TEST_EQUALITY(aggSizesGreedyPreserve.extent(0) > aggSizesGreedy.extent(0), true)
+
+  vertex2AggId = aggregates->GetVertex2AggId()->getHostLocalView(Xpetra::Access::ReadOnly);
+  for (auto i = 0; i < (nx / 2 * ny / 2); i++) {
+    TEST_EQUALITY(vertex2AggId(i, 0) != MUELU_UNAGGREGATED, true);  // check that all nodes are aggregated
+  }
+}
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustUncoupledAggregationFactory, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustUncoupledAggregation, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2UncoupledAggregation, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2PreserveUncoupledAggregation, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2DeterUncoupledAggregation, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustOnePt2UncoupledAggregation, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, DiscontiguousAggregates, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase1, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase2, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase3, SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustUncoupledAggregationFactory, SC, LO, GO, NO)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustUncoupledAggregation, SC, LO, GO, NO)                     \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2UncoupledAggregation, SC, LO, GO, NO)                \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2PreserveUncoupledAggregation, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustDist2DeterUncoupledAggregation, SC, LO, GO, NO)           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, JustOnePt2UncoupledAggregation, SC, LO, GO, NO)               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, DiscontiguousAggregates, SC, LO, GO, NO)                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase1, SC, LO, GO, NO)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase2, SC, LO, GO, NO)                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, UncoupledPhase3, SC, LO, GO, NO)                              \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, AllowDroppingToCreateAdditionalDirichletRows, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Aggregates_kokkos, GreedyDirichlet, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/AmalgamationInfo_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/AmalgamationInfo_kokkos.cpp
index 9f9cd5dd0609..988469b366de 100644
--- a/packages/muelu/test/unit_tests_kokkos/AmalgamationInfo_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/AmalgamationInfo_kokkos.cpp
@@ -59,118 +59,113 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    LO fullblocksize    = 1;   // block dim for fixed size blocks
-    GO offset           = 0;   // global offset of dof gids
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    LO stridedblocksize = fullblocksize; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
-
-    const GO nx = 199;
-    using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
-
-    RCP<AmalgamationInfo> amalgamationData;
-    RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
-    RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
-
-    RCP<AmalgamationInfo> amalgamationInfo_kokkos = rcp(new AmalgamationInfo(theRowTranslation,
-                                                                      theColTranslation,
-                                                                      A->getRowMap(),
-                                                                      A->getColMap(),
-                                                                      A->getColMap(),
-                                                                      fullblocksize,
-                                                                      offset,
-                                                                      blockid,
-                                                                      nStridedOffset,
-                                                                      stridedblocksize) );
-    TEST_INEQUALITY(amalgamationInfo_kokkos, Teuchos::null);
-
-    TEST_EQUALITY(amalgamationInfo_kokkos->description(), "AmalgamationInfo");
-    LO fullBlockSizeExpected;
-    LO blockIDExpected;
-    LO stridingOffsetExpected;
-    LO stridedBlockSizeExpected;
-    GO indexBaseExpected;
-    amalgamationInfo_kokkos->GetStridingInformation(fullBlockSizeExpected, blockIDExpected, stridingOffsetExpected, stridedBlockSizeExpected, indexBaseExpected);
-    TEST_EQUALITY(fullBlockSizeExpected == fullblocksize, true);
-    TEST_EQUALITY(blockIDExpected == blockid, true);
-    TEST_EQUALITY(stridingOffsetExpected == nStridedOffset, true);
-    TEST_EQUALITY(stridedBlockSizeExpected == stridedblocksize, true);
-    TEST_EQUALITY(indexBaseExpected == A->getColMap()->getIndexBase(), true);
-
-    TEST_EQUALITY(amalgamationInfo_kokkos->getNodeRowMap() == A->getRowMap(), true);
-    TEST_EQUALITY(amalgamationInfo_kokkos->getNodeColMap() == A->getColMap(), true);
-    TEST_EQUALITY(amalgamationInfo_kokkos->getRowTranslation() == theRowTranslation, true);
-    TEST_EQUALITY(amalgamationInfo_kokkos->getColTranslation() == theColTranslation, true);
-    TEST_EQUALITY(amalgamationInfo_kokkos->GlobalOffset() == offset, true);
-
-  } // Constructor
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo_kokkos, UnamalgateAggregate, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-    LO fullblocksize    = 1;   // block dim for fixed size blocks
-    GO offset           = 0;   // global offset of dof gids
-    LO blockid          = -1;  // block id in strided map
-    LO nStridedOffset   = 0;   // DOF offset for strided block id "blockid" (default = 0)
-    LO stridedblocksize = fullblocksize; // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
-
-    const GO nx = 199;
-    using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
-
-    RCP<AmalgamationInfo> amalgamationData;
-    RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
-    RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
-
-    RCP<AmalgamationInfo> amalgamationInfo_kokkos = rcp(new AmalgamationInfo(theRowTranslation,
-                                                                      theColTranslation,
-                                                                      A->getRowMap(),
-                                                                      A->getColMap(),
-                                                                      A->getColMap(),
-                                                                      fullblocksize,
-                                                                      offset,
-                                                                      blockid,
-                                                                      nStridedOffset,
-                                                                      stridedblocksize) );
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-    RCP<LWGraph_kokkos> graph;
-    Level level;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
-    level.Request("Graph", dropFact.get());
-    level.Request(*dropFact);
-    dropFact->Build(level);
-    graph = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
-    RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
-    GO numAggs = aggregates->GetNumAggregates();
-    ArrayRCP<LO> aggSizes = Teuchos::ArrayRCP<LO>(numAggs);
-    ArrayRCP<LO> aggStart;
-    ArrayRCP<GO> aggToRowMap;
-    amalgamationInfo_kokkos->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
-
-
-  } // UnamalgateAggregate
-
-
-  # define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo_kokkos, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo_kokkos, UnamalgateAggregate, Scalar, LO, GO, Node) \
-
-# include <MueLu_ETI_4arg.hpp>
-
-} // namespace MueLuTests
-
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  LO fullblocksize    = 1;              // block dim for fixed size blocks
+  GO offset           = 0;              // global offset of dof gids
+  LO blockid          = -1;             // block id in strided map
+  LO nStridedOffset   = 0;              // DOF offset for strided block id "blockid" (default = 0)
+  LO stridedblocksize = fullblocksize;  // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+
+  const GO nx        = 199;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+
+  RCP<AmalgamationInfo> amalgamationData;
+  RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
+  RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
+
+  RCP<AmalgamationInfo> amalgamationInfo_kokkos = rcp(new AmalgamationInfo(theRowTranslation,
+                                                                           theColTranslation,
+                                                                           A->getRowMap(),
+                                                                           A->getColMap(),
+                                                                           A->getColMap(),
+                                                                           fullblocksize,
+                                                                           offset,
+                                                                           blockid,
+                                                                           nStridedOffset,
+                                                                           stridedblocksize));
+  TEST_INEQUALITY(amalgamationInfo_kokkos, Teuchos::null);
+
+  TEST_EQUALITY(amalgamationInfo_kokkos->description(), "AmalgamationInfo");
+  LO fullBlockSizeExpected;
+  LO blockIDExpected;
+  LO stridingOffsetExpected;
+  LO stridedBlockSizeExpected;
+  GO indexBaseExpected;
+  amalgamationInfo_kokkos->GetStridingInformation(fullBlockSizeExpected, blockIDExpected, stridingOffsetExpected, stridedBlockSizeExpected, indexBaseExpected);
+  TEST_EQUALITY(fullBlockSizeExpected == fullblocksize, true);
+  TEST_EQUALITY(blockIDExpected == blockid, true);
+  TEST_EQUALITY(stridingOffsetExpected == nStridedOffset, true);
+  TEST_EQUALITY(stridedBlockSizeExpected == stridedblocksize, true);
+  TEST_EQUALITY(indexBaseExpected == A->getColMap()->getIndexBase(), true);
+
+  TEST_EQUALITY(amalgamationInfo_kokkos->getNodeRowMap() == A->getRowMap(), true);
+  TEST_EQUALITY(amalgamationInfo_kokkos->getNodeColMap() == A->getColMap(), true);
+  TEST_EQUALITY(amalgamationInfo_kokkos->getRowTranslation() == theRowTranslation, true);
+  TEST_EQUALITY(amalgamationInfo_kokkos->getColTranslation() == theColTranslation, true);
+  TEST_EQUALITY(amalgamationInfo_kokkos->GlobalOffset() == offset, true);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(AmalgamationInfo_kokkos, UnamalgateAggregate, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  LO fullblocksize    = 1;              // block dim for fixed size blocks
+  GO offset           = 0;              // global offset of dof gids
+  LO blockid          = -1;             // block id in strided map
+  LO nStridedOffset   = 0;              // DOF offset for strided block id "blockid" (default = 0)
+  LO stridedblocksize = fullblocksize;  // size of strided block id "blockid" (default = fullblocksize, only if blockid!=-1 stridedblocksize <= fullblocksize)
+
+  const GO nx        = 199;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nx);
+
+  RCP<AmalgamationInfo> amalgamationData;
+  RCP<Array<LO> > theRowTranslation = rcp(new Array<LO>);
+  RCP<Array<LO> > theColTranslation = rcp(new Array<LO>);
+
+  RCP<AmalgamationInfo> amalgamationInfo_kokkos = rcp(new AmalgamationInfo(theRowTranslation,
+                                                                           theColTranslation,
+                                                                           A->getRowMap(),
+                                                                           A->getColMap(),
+                                                                           A->getColMap(),
+                                                                           fullblocksize,
+                                                                           offset,
+                                                                           blockid,
+                                                                           nStridedOffset,
+                                                                           stridedblocksize));
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<LWGraph_kokkos> graph;
+  Level level;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+  level.Request("Graph", dropFact.get());
+  level.Request(*dropFact);
+  dropFact->Build(level);
+  graph                      = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
+  RCP<Aggregates> aggregates = rcp(new Aggregates(*graph));
+  GO numAggs                 = aggregates->GetNumAggregates();
+  ArrayRCP<LO> aggSizes      = Teuchos::ArrayRCP<LO>(numAggs);
+  ArrayRCP<LO> aggStart;
+  ArrayRCP<GO> aggToRowMap;
+  amalgamationInfo_kokkos->UnamalgamateAggregates(*aggregates, aggStart, aggToRowMap);
+
+}  // UnamalgateAggregate
+
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo_kokkos, Constructor, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(AmalgamationInfo_kokkos, UnamalgateAggregate, Scalar, LO, GO, Node)
+
+#include <MueLu_ETI_4arg.hpp>
+
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/CoalesceDropFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/CoalesceDropFactory_kokkos.cpp
index 81cf8a5ccdc3..768d2638f9cd 100644
--- a/packages/muelu/test/unit_tests_kokkos/CoalesceDropFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/CoalesceDropFactory_kokkos.cpp
@@ -54,325 +54,326 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<CoalesceDropFactory_kokkos> coalesceDropFact = rcp(new CoalesceDropFactory_kokkos());
-    TEST_EQUALITY(coalesceDropFact != Teuchos::null, true);
-
-    out << *coalesceDropFact << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<CoalesceDropFactory_kokkos> coalesceDropFact = rcp(new CoalesceDropFactory_kokkos());
+  TEST_EQUALITY(coalesceDropFact != Teuchos::null, true);
+
+  out << *coalesceDropFact << std::endl;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicScalarWithoutFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+  Level fineLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(36);
+  fineLevel.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
+  CoalesceDropFactory_kokkos dropFact;
+  dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  auto myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+
+  TEST_EQUALITY(as<int>(myDofsPerNode) == 1, true);
+
+  bool bCorrectGraph = false;
+  int reduction_val  = 0;
+  int comm_size = comm->getSize(), comm_rank = comm->getRank();
+  auto lclLWGraph = graph->getLocalLWGraph();
+  Kokkos::parallel_reduce(
+      "MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t>(0, 1),
+      KOKKOS_LAMBDA(const LO i, int &correct) {
+        if (comm_size == 1) {
+          auto v0 = lclLWGraph.getNeighborVertices(0);
+          auto v1 = lclLWGraph.getNeighborVertices(1);
+          auto v2 = lclLWGraph.getNeighborVertices(2);
+          if (v0.length == 2 && ((v0(0) == 0 && v0(1) == 1) || (v0(0) == 1 && v0(1) == 0)) &&
+              v1.length == 3 && v2.length == 3)
+            correct = true;
+        } else {
+          if (comm_rank == 0) {
+            if (lclLWGraph.getNeighborVertices(0).length == 2)
+              correct = true;
+
+          } else {
+            if (lclLWGraph.getNeighborVertices(0).length == 3)
+              correct = true;
+          }
+        }
+      },
+      reduction_val);
+  bCorrectGraph = reduction_val;
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  auto myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  auto myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(36 + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 35);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), 36);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicScalarWithFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  Level fineLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  auto dofMap = MapFactory::Build(lib, 3 * comm->getSize(), 0, comm);
+  auto mtx    = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 1.0, -1.0, -0.0001);
+
+  mtx->SetFixedBlockSize(1, 0);
+  fineLevel.Set("A", mtx);
+
+  RCP<AmalgamationFactory> amalgFact  = rcp(new AmalgamationFactory);
+  CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+  dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.5));
+
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  auto myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+
+  TEST_EQUALITY(as<int>(myDofsPerNode) == 1, true);
+  TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == 3 * comm->getSize(), true);
+
+  bool bCorrectGraph = false;
+  int reduction_val  = 0;
+  int comm_size = comm->getSize(), comm_rank = comm->getRank();
+  auto lclLWGraph = graph->getLocalLWGraph();
+  Kokkos::parallel_reduce(
+      "MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t>(0, 1),
+      KOKKOS_LAMBDA(const LO i, int &correct) {
+        if (comm_size == 1) {
+          auto v0 = lclLWGraph.getNeighborVertices(0);
+          auto v1 = lclLWGraph.getNeighborVertices(1);
+          auto v2 = lclLWGraph.getNeighborVertices(2);
+          if (v0.length == 1 && v0(0) == 0 &&
+              v1.length == 2 && ((v1(0) == 0 && v1(1) == 1) || (v1(0) == 1 && v1(1) == 0)) &&
+              v2.length == 2 && ((v2(0) == 1 && v2(1) == 2) || (v2(0) == 2 && v2(1) == 1)))
+            correct = true;
+        } else {
+          if (comm_rank == 0) {
+            if (lclLWGraph.getNeighborVertices(0).length == 1)
+              correct = true;
+
+          } else {
+            if (lclLWGraph.getNeighborVertices(0).length == 2)
+              correct = true;
+          }
+        }
+      },
+      reduction_val);
+  bCorrectGraph = reduction_val;
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  auto myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  auto myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), 3 * comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(3 * comm->getSize() + (comm->getSize() - 1) * 2));
+
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), 3 * comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), as<size_t>(3 * comm->getSize()));
+  TEST_EQUALITY(myDomainMap->getLocalNumElements(), 3);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicBlockWithoutFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  Level fineLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
+
+  int blockSize = 3;
+
+  auto dofMap = MapFactory::Build(lib, blockSize * comm->getSize(), 0, comm);
+  auto mtx    = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
+  mtx->SetFixedBlockSize(blockSize, 0);
+  fineLevel.Set("A", mtx);
+
+  RCP<AmalgamationFactory> amalgFact  = rcp(new AmalgamationFactory);
+  CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+  dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
+
+  dropFact.Build(fineLevel);
+
+  auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  auto myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
+
+  TEST_EQUALITY(as<int>(myDofsPerNode) == blockSize, true);
+  TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
+
+  bool bCorrectGraph = false;
+  int reduction_val  = 0;
+  int comm_size = comm->getSize(), comm_rank = comm->getRank();
+  auto lclLWGraph = graph->getLocalLWGraph();
+  Kokkos::parallel_reduce(
+      "MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t>(0, 1),
+      KOKKOS_LAMBDA(const LO i, int &correct) {
+        if (comm_size == 1 && lclLWGraph.getNeighborVertices(0).length == 1) {
+          correct = true;
+        } else {
+          if (comm_rank == 0 || comm_rank == comm_size - 1) {
+            if (lclLWGraph.getNeighborVertices(0).length == 2)
+              correct = true;
+
+          } else {
+            if (static_cast<int>(lclLWGraph.getNeighborVertices(0).length) == blockSize)
+              correct = true;
+          }
+        }
+      },
+      reduction_val);
+  bCorrectGraph = reduction_val;
+  TEST_EQUALITY(bCorrectGraph, true);
+
+  auto myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  auto myDomainMap = graph->GetDomainMap();
+
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts + 1);
+    } else {
+      TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts + 2);
+    }
   }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicScalarWithoutFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-    Level fineLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build1DPoisson(36);
-    fineLevel.Set("A", A);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-    CoalesceDropFactory_kokkos dropFact;
-    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
-
-    fineLevel.Request("Graph",       &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph",       &dropFact);
-    auto myDofsPerNode = fineLevel.Get<LO>                  ("DofsPerNode", &dropFact);
-
-    TEST_EQUALITY(as<int>(myDofsPerNode) == 1, true);
-
-    bool bCorrectGraph = false;
-    int reduction_val=0;
-    int comm_size=comm->getSize(), comm_rank=comm->getRank();
-    auto lclLWGraph = graph->getLocalLWGraph();
-    Kokkos::parallel_reduce("MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t> (0,1),
-			    KOKKOS_LAMBDA(const LO i, int &correct) {
-			      if (comm_size == 1) {
-				auto v0 = lclLWGraph.getNeighborVertices(0);
-				auto v1 = lclLWGraph.getNeighborVertices(1);
-				auto v2 = lclLWGraph.getNeighborVertices(2);
-				if (v0.length == 2 && ((v0(0) == 0 && v0(1) == 1) || (v0(0) == 1 && v0(1) == 0)) &&
-				    v1.length == 3 && v2.length == 3)
-				  correct = true;
-			      } else {
-				if (comm_rank == 0 ) {
-				  if (lclLWGraph.getNeighborVertices(0).length == 2)
-				    correct = true;
-				  
-				} else {
-				  if (lclLWGraph.getNeighborVertices(0).length == 3)
-				    correct = true;
-				}
-			      }
-			    }, reduction_val);
-    bCorrectGraph=reduction_val;
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    auto myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    auto myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(),  35);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(),  0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),      0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),  as<size_t>(36 + (comm->getSize()-1)*2));
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(),  35);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(),  0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),      0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),  36);
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t maxLocalIndex      = myImportMap->getMaxLocalIndex();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(maxLocalIndex, numLocalRowMapElts * blockSize - 2);
+    } else {
+      TEST_EQUALITY(maxLocalIndex, numLocalRowMapElts * blockSize - 1);
+    }
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicScalarWithFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    Level fineLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    auto dofMap = MapFactory::Build(lib, 3*comm->getSize(), 0, comm);
-    auto mtx    = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 1.0, -1.0, -0.0001);
-
-    mtx->SetFixedBlockSize(1, 0);
-    fineLevel.Set("A", mtx);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-    CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
-    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(0.5));
-
-    fineLevel.Request("Graph",       &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), as<size_t>(comm->getSize()));
+  TEST_EQUALITY(myDomainMap->getLocalNumElements(), 1);
+}
 
-    dropFact.Build(fineLevel);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicBlockWithFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph",       &dropFact);
-    auto myDofsPerNode = fineLevel.Get<LO>                  ("DofsPerNode", &dropFact);
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
 
-    TEST_EQUALITY(as<int>(myDofsPerNode) == 1, true);
-    TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == 3*comm->getSize(), true);
+  Level fineLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
-    bool bCorrectGraph = false;
-    int reduction_val=0;
-    int comm_size=comm->getSize(), comm_rank=comm->getRank();
-    auto lclLWGraph = graph->getLocalLWGraph();
-    Kokkos::parallel_reduce("MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t> (0,1),
-			    KOKKOS_LAMBDA(const LO i, int &correct) {
-			      if (comm_size == 1) {
-				auto v0 = lclLWGraph.getNeighborVertices(0);
-				auto v1 = lclLWGraph.getNeighborVertices(1);
-				auto v2 = lclLWGraph.getNeighborVertices(2);
-				if (v0.length == 1 &&   v0(0) == 0 &&
-				    v1.length == 2 && ((v1(0) == 0 && v1(1) == 1) || (v1(0) == 1 && v1(1) == 0)) &&
-				    v2.length == 2 && ((v2(0) == 1 && v2(1) == 2) || (v2(0) == 2 && v2(1) == 1)))
-				  correct = true;
-			      } else {
-				if (comm_rank == 0 ) {
-				  if (lclLWGraph.getNeighborVertices(0).length == 1)
-				    correct = true;
-				  
-				} else {
-				  if (lclLWGraph.getNeighborVertices(0).length == 2)
-				    correct = true;
-				}
-			      }
-			    }, reduction_val);
-    bCorrectGraph=reduction_val;			    
-    TEST_EQUALITY(bCorrectGraph, true);
+  auto dofMap = MapFactory::Build(lib, 3 * comm->getSize(), 0, comm);
+  auto mtx    = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, 0.00001);
 
-    auto myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    auto myDomainMap = graph->GetDomainMap();
+  mtx->SetFixedBlockSize(3, 0);
+  fineLevel.Set("A", mtx);
 
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(),  3*comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(),  0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),      0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(),  as<size_t>(3*comm->getSize()+(comm->getSize()-1)*2));
+  RCP<AmalgamationFactory> amalgFact  = rcp(new AmalgamationFactory);
+  CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+  dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+  dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(1.0));
 
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(),  3*comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(),  0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),      0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(),  as<size_t>(3*comm->getSize()));
-    TEST_EQUALITY(myDomainMap->getLocalNumElements(),    3);
-  }
+  fineLevel.Request("Graph", &dropFact);
+  fineLevel.Request("DofsPerNode", &dropFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicBlockWithoutFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+  dropFact.Build(fineLevel);
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
+  auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  auto myDofsPerNode = fineLevel.Get<LO>("DofsPerNode", &dropFact);
 
-    Level fineLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  TEST_EQUALITY(as<int>(myDofsPerNode) == 3, true);
+  TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
 
-    int blockSize = 3;
+  TEST_EQUALITY(graph->getNeighborVertices(0).size(), 1);
 
-    auto dofMap = MapFactory::Build(lib, blockSize*comm->getSize(), 0, comm);
-    auto mtx    = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-    mtx->SetFixedBlockSize(blockSize, 0);
-    fineLevel.Set("A", mtx);
+  auto myImportMap = graph->GetImportMap();  // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
+  auto myDomainMap = graph->GetDomainMap();
 
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-    CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
-
-    fineLevel.Request("Graph",       &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph",       &dropFact);
-    auto myDofsPerNode = fineLevel.Get<LO>                  ("DofsPerNode", &dropFact);
-
-    TEST_EQUALITY(as<int>(myDofsPerNode) == blockSize, true);
-    TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-
-    bool bCorrectGraph = false;
-    int reduction_val=0;
-    int comm_size=comm->getSize(), comm_rank=comm->getRank();
-    auto lclLWGraph = graph->getLocalLWGraph();
-    Kokkos::parallel_reduce("MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t> (0,1),
-			    KOKKOS_LAMBDA(const LO i, int &correct) {
-			      if (comm_size == 1 && lclLWGraph.getNeighborVertices(0).length == 1) {
-				correct = true;
-			      } else {
-				if (comm_rank == 0 || comm_rank == comm_size-1) {
-				  if (lclLWGraph.getNeighborVertices(0).length == 2)
-				    correct = true;
-				  
-				} else {
-				  if (static_cast<int>(lclLWGraph.getNeighborVertices(0).length) == blockSize)
-				    correct = true;
-				}
-			      }
-			    }, reduction_val);
-    bCorrectGraph=reduction_val;			    
-    TEST_EQUALITY(bCorrectGraph, true);
-
-    auto myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    auto myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),     0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-
-    if (comm->getSize() > 1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts+1);
-      } else {
-        TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts+2);
-      }
-    }
-    if (comm->getSize() > 1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t maxLocalIndex = myImportMap->getMaxLocalIndex();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(maxLocalIndex, numLocalRowMapElts*blockSize-2);
-      } else {
-        TEST_EQUALITY(maxLocalIndex, numLocalRowMapElts*blockSize-1);
-      }
+  TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(comm->getSize() + 2 * (comm->getSize() - 1)));
+  if (comm->getSize() > 1) {
+    size_t numLocalRowMapElts = graph->GetNodeNumVertices();
+    size_t numLocalImportElts = myImportMap->getLocalNumElements();
+    if (comm->getRank() == 0 || comm->getRank() == comm->getSize() - 1) {
+      TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts + 1);
+    } else {
+      TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts + 2);
     }
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),     0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),     0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(), as<size_t>(comm->getSize()));
-    TEST_EQUALITY(myDomainMap->getLocalNumElements(),   1);
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, ClassicBlockWithFiltering, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    Level fineLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
-
-    auto dofMap = MapFactory::Build(lib, 3*comm->getSize(), 0, comm);
-    auto mtx    = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, 0.00001);
-
-    mtx->SetFixedBlockSize(3, 0);
-    fineLevel.Set("A", mtx);
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory);
-    CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
-    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(1.0));
-
-    fineLevel.Request("Graph",       &dropFact);
-    fineLevel.Request("DofsPerNode", &dropFact);
-
-    dropFact.Build(fineLevel);
-
-    auto graph         = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph",       &dropFact);
-    auto myDofsPerNode = fineLevel.Get<LO>                  ("DofsPerNode", &dropFact);
-
-    TEST_EQUALITY(as<int>(myDofsPerNode) == 3, true);
-    TEST_EQUALITY(as<int>(graph->GetDomainMap()->getGlobalNumElements()) == comm->getSize(), true);
-
-    TEST_EQUALITY(graph->getNeighborVertices(0).size(), 1);
-
-    auto myImportMap = graph->GetImportMap(); // < note that the ImportMap is built from the column map of the matrix A WITHOUT dropping!
-    auto myDomainMap = graph->GetDomainMap();
-
-    TEST_EQUALITY(myImportMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myImportMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myImportMap->getMinLocalIndex(),     0);
-    TEST_EQUALITY(myImportMap->getGlobalNumElements(), as<size_t>(comm->getSize()+2*(comm->getSize()-1)));
-    if (comm->getSize() > 1) {
-      size_t numLocalRowMapElts = graph->GetNodeNumVertices();
-      size_t numLocalImportElts = myImportMap->getLocalNumElements();
-      if (comm->getRank() == 0 || comm->getRank() == comm->getSize()-1) {
-        TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts+1);
-      } else {
-        TEST_EQUALITY(numLocalImportElts, numLocalRowMapElts+2);
-      }
-    }
-
-    TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize()-1);
-    TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
-    TEST_EQUALITY(myDomainMap->getMaxLocalIndex(),     0);
-    TEST_EQUALITY(myDomainMap->getMinLocalIndex(),     0);
-    TEST_EQUALITY(myDomainMap->getGlobalNumElements(), as<size_t>(comm->getSize()));
-    TEST_EQUALITY(myDomainMap->getLocalNumElements(),   1);
-  }
+  TEST_EQUALITY(myDomainMap->getMaxAllGlobalIndex(), comm->getSize() - 1);
+  TEST_EQUALITY(myDomainMap->getMinAllGlobalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMaxLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getMinLocalIndex(), 0);
+  TEST_EQUALITY(myDomainMap->getGlobalNumElements(), as<size_t>(comm->getSize()));
+  TEST_EQUALITY(myDomainMap->getLocalNumElements(), 1);
+}
 
 #if 0
   TEUCHOS_UNIT_TEST(CoalesceDropFactory_kokkos, LaplacianScalarWithoutFiltering)
   {
-#   include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
     out << "version: " << MueLu::Version() << std::endl;
@@ -434,11 +435,10 @@ namespace MueLuTests {
   }
 #endif
 
-
 #if 0
   TEUCHOS_UNIT_TEST(CoalesceDropFactory, AmalgamationStridedLW)
   {
-#   include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
     out << "version: " << MueLu::Version() << std::endl;
@@ -538,7 +538,7 @@ namespace MueLuTests {
 
   TEUCHOS_UNIT_TEST(CoalesceDropFactory, AmalgamationStrided2LW)
   {
-#   include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
     out << "version: " << MueLu::Version() << std::endl;
@@ -743,109 +743,107 @@ namespace MueLuTests {
   } // AmalgamationStridedOffsetDropping2LW
 #endif
 
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, AggresiveDroppingIsMarkedAsBoundary, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+  // Test that when everything but the diagonal is dropped, the node is marked as boundary
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoalesceDropFactory_kokkos, AggresiveDroppingIsMarkedAsBoundary, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    // Test that when everything but the diagonal is dropped, the node is marked as boundary
-#   include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
 
-    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
+  RCP<const Map> dofMap    = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 12 * comm->getSize(), 0, comm);
+  Teuchos::RCP<Matrix> mtx = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
 
-    RCP<const Map> dofMap = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(lib, 12*comm->getSize(), 0, comm);
-    Teuchos::RCP<Matrix> mtx = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildTridiag(dofMap, 2.0, -1.0, -1.0);
-
-    {
-      Level fineLevel;
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  {
+    Level fineLevel;
+    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
-      mtx->SetFixedBlockSize(1);
-      fineLevel.Set("A", mtx);
+    mtx->SetFixedBlockSize(1);
+    fineLevel.Set("A", mtx);
 
-      CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-      dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-      fineLevel.Request("Graph", &dropFact);
-      fineLevel.Request("DofsPerNode", &dropFact);
+    CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+    RCP<AmalgamationFactory> amalgFact  = rcp(new AmalgamationFactory());
+    dropFact.SetFactory("UnAmalgamationInfo", amalgFact);
+    dropFact.SetParameter("aggregation: drop tol", Teuchos::ParameterEntry(4.1));
+    fineLevel.Request("Graph", &dropFact);
+    fineLevel.Request("DofsPerNode", &dropFact);
 
-      dropFact.Build(fineLevel);
+    dropFact.Build(fineLevel);
 
-      RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+    RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
 
-      auto boundaryNodes = graph->getLocalLWGraph().GetBoundaryNodeMap();
-      auto boundaryNodesHost = Kokkos::create_mirror_view(boundaryNodes);
-      Kokkos::deep_copy(boundaryNodesHost, boundaryNodes);
-      bool allNodesAreOnBoundary = true;
-      for (LO i = 0; i < Teuchos::as<LO>(boundaryNodesHost.size()); i++)
-        allNodesAreOnBoundary &= boundaryNodesHost(i);
-      TEST_EQUALITY(allNodesAreOnBoundary, true);
-    }
+    auto boundaryNodes     = graph->getLocalLWGraph().GetBoundaryNodeMap();
+    auto boundaryNodesHost = Kokkos::create_mirror_view(boundaryNodes);
+    Kokkos::deep_copy(boundaryNodesHost, boundaryNodes);
+    bool allNodesAreOnBoundary = true;
+    for (LO i = 0; i < Teuchos::as<LO>(boundaryNodesHost.size()); i++)
+      allNodesAreOnBoundary &= boundaryNodesHost(i);
+    TEST_EQUALITY(allNodesAreOnBoundary, true);
+  }
 
-    // {
-    //   Level fineLevel;
-    //   TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  // {
+  //   Level fineLevel;
+  //   TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
 
-    //   mtx->SetFixedBlockSize(2);
-    //   fineLevel.Set("A", mtx);
+  //   mtx->SetFixedBlockSize(2);
+  //   fineLevel.Set("A", mtx);
 
-    //   CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-    //   RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    //   dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    //   dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-    //   fineLevel.Request("Graph", &dropFact);
-    //   fineLevel.Request("DofsPerNode", &dropFact);
+  //   CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+  //   RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  //   dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
+  //   dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
+  //   fineLevel.Request("Graph", &dropFact);
+  //   fineLevel.Request("DofsPerNode", &dropFact);
 
-    //   dropFact.Build(fineLevel);
+  //   dropFact.Build(fineLevel);
 
-    //   RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  //   RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
 
-    //   auto boundaryNodes = graph->GetBoundaryNodeMap();
-    //   bool allNodesAreOnBoundary = true;
-    //   for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
-    //     allNodesAreOnBoundary &= boundaryNodes[i];
-    //   TEST_EQUALITY(allNodesAreOnBoundary, true);
-    // }
+  //   auto boundaryNodes = graph->GetBoundaryNodeMap();
+  //   bool allNodesAreOnBoundary = true;
+  //   for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
+  //     allNodesAreOnBoundary &= boundaryNodes[i];
+  //   TEST_EQUALITY(allNodesAreOnBoundary, true);
+  // }
 
-    // {
-    //   Level fineLevel;
-    //   TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  // {
+  //   Level fineLevel;
+  //   TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
 
-    //   mtx->SetFixedBlockSize(3);
-    //   fineLevel.Set("A", mtx);
+  //   mtx->SetFixedBlockSize(3);
+  //   fineLevel.Set("A", mtx);
 
-    //   CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
-    //   RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    //   dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
-    //   dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
-    //   fineLevel.Request("Graph", &dropFact);
-    //   fineLevel.Request("DofsPerNode", &dropFact);
+  //   CoalesceDropFactory_kokkos dropFact = CoalesceDropFactory_kokkos();
+  //   RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  //   dropFact.SetFactory("UnAmalgamationInfo",amalgFact);
+  //   dropFact.SetParameter("aggregation: drop tol",Teuchos::ParameterEntry(4.1));
+  //   fineLevel.Request("Graph", &dropFact);
+  //   fineLevel.Request("DofsPerNode", &dropFact);
 
-    //   dropFact.Build(fineLevel);
+  //   dropFact.Build(fineLevel);
 
-    //   RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
+  //   RCP<LWGraph_kokkos> graph = fineLevel.Get<RCP<LWGraph_kokkos> >("Graph", &dropFact);
 
-    //   auto boundaryNodes = graph->GetBoundaryNodeMap();
-    //   bool allNodesAreOnBoundary = true;
-    //   for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
-    //     allNodesAreOnBoundary &= boundaryNodes[i];
-    //   TEST_EQUALITY(allNodesAreOnBoundary, true);
-    // }
+  //   auto boundaryNodes = graph->GetBoundaryNodeMap();
+  //   bool allNodesAreOnBoundary = true;
+  //   for (LO i = 0; i < Teuchos::as<LO>(boundaryNodes.size()); i++)
+  //     allNodesAreOnBoundary &= boundaryNodes[i];
+  //   TEST_EQUALITY(allNodesAreOnBoundary, true);
+  // }
 
-  } // AggresiveDroppingIsMarkedAsBoundary
+}  // AggresiveDroppingIsMarkedAsBoundary
 
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, Constructor,                   SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                           \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, Constructor, SC, LO, GO, NO)                   \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicScalarWithoutFiltering, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicScalarWithFiltering,    SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicBlockWithoutFiltering,  SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, AggresiveDroppingIsMarkedAsBoundary,  SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicScalarWithFiltering, SC, LO, GO, NO)    \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicBlockWithoutFiltering, SC, LO, GO, NO)  \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, AggresiveDroppingIsMarkedAsBoundary, SC, LO, GO, NO)
 
-  //TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicBlockWithFiltering,     SC, LO, GO, NO) // not implemented yet
+// TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoalesceDropFactory_kokkos, ClassicBlockWithFiltering,     SC, LO, GO, NO) // not implemented yet
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/CoarseMapFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/CoarseMapFactory_kokkos.cpp
index ead5cbfb898d..55fda415d0b4 100644
--- a/packages/muelu/test/unit_tests_kokkos/CoarseMapFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/CoarseMapFactory_kokkos.cpp
@@ -55,45 +55,44 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMap_kokkos, StandardCase, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    out << "version: " << MueLu::Version() << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(CoarseMap_kokkos, StandardCase, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    Level fineLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(fineLevel);
+  Level fineLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(fineLevel);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
-    fineLevel.Set("A", A);
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(15);
+  fineLevel.Set("A", A);
 
-    // build dummy aggregate structure
-    RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
-    aggs->SetNumAggregates(10); // set (local!) number of aggregates
-    fineLevel.Set("Aggregates", aggs);
+  // build dummy aggregate structure
+  RCP<Aggregates> aggs = Teuchos::rcp(new Aggregates(A->getRowMap()));
+  aggs->SetNumAggregates(10);  // set (local!) number of aggregates
+  fineLevel.Set("Aggregates", aggs);
 
-    // build dummy nullspace vector
-    RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(),1);
-    nsp->putScalar(1.0);
-    fineLevel.Set("Nullspace", nsp);
+  // build dummy nullspace vector
+  RCP<MultiVector> nsp = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nsp->putScalar(1.0);
+  fineLevel.Set("Nullspace", nsp);
 
-    RCP<CoarseMapFactory> coarseMapFactory = Teuchos::rcp(new CoarseMapFactory());
-    coarseMapFactory->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
-    coarseMapFactory->SetFactory("Nullspace",  MueLu::NoFactory::getRCP());
+  RCP<CoarseMapFactory> coarseMapFactory = Teuchos::rcp(new CoarseMapFactory());
+  coarseMapFactory->SetFactory("Aggregates", MueLu::NoFactory::getRCP());
+  coarseMapFactory->SetFactory("Nullspace", MueLu::NoFactory::getRCP());
 
-    fineLevel.Request("CoarseMap", coarseMapFactory.get());
-    coarseMapFactory->Build(fineLevel);
+  fineLevel.Request("CoarseMap", coarseMapFactory.get());
+  coarseMapFactory->Build(fineLevel);
 
-    auto myCoarseMap = fineLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", coarseMapFactory.get());
+  auto myCoarseMap = fineLevel.Get<Teuchos::RCP<const Map> >("CoarseMap", coarseMapFactory.get());
 
-    TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
-    TEST_EQUALITY(myCoarseMap->getMaxLocalIndex()     == 9, true);
-  }
+  TEST_EQUALITY(myCoarseMap->getMinAllGlobalIndex() == 0, true);
+  TEST_EQUALITY(myCoarseMap->getMaxLocalIndex() == 9, true);
+}
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(CoarseMap_kokkos, StandardCase, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/IndexManager_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/IndexManager_kokkos.cpp
index d32f9e29951d..6febbead1b3c 100644
--- a/packages/muelu/test/unit_tests_kokkos/IndexManager_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/IndexManager_kokkos.cpp
@@ -55,163 +55,209 @@
 
 namespace MueLuTests {
 
-  template <class IndexManager_type, class LO>
-  struct testFunctor {
-    IndexManager_type IndexManager;
-
-    testFunctor(RCP<IndexManager_type> myIM) : IndexManager(*myIM) {}
-
-    KOKKOS_INLINE_FUNCTION
-    void operator() (const LO& LID) const {
-      LO fineNodeTuple[3];
-      IndexManager.getFineLID2FineTuple(LID, fineNodeTuple);
-      printf("LID=%d --> (i,j,k)=(%d, %d, %d)\n", static_cast<int>(LID),
-             static_cast<int>(fineNodeTuple[0]),
-             static_cast<int>(fineNodeTuple[1]),
-             static_cast<int>(fineNodeTuple[2]));
-    }
-  };
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager_kokkos, IndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    fout->setShowAllFrontMatter(false).setShowProcRank(true);
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers_kokkos::Parameters::getDefaultComm();
-
-    // Set global geometric data
-    const int numDimensions = 3;
-    const int interpolationOrder = 0;
-    Array<GO> meshData;
-    Array<GO> gNodesPerDir(3);
-    Array<LO> lNodesPerDir(3);
-    Array<LO> coarseRate(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 5;
-        coarseRate[dim]   = 2;
-      } else {
-        gNodesPerDir[dim] = -1;
-        coarseRate[dim]   =  1;
-      }
+template <class IndexManager_type, class LO>
+struct testFunctor {
+  IndexManager_type IndexManager;
+
+  testFunctor(RCP<IndexManager_type> myIM)
+    : IndexManager(*myIM) {}
+
+  KOKKOS_INLINE_FUNCTION
+  void operator()(const LO& LID) const {
+    LO fineNodeTuple[3];
+    IndexManager.getFineLID2FineTuple(LID, fineNodeTuple);
+    printf("LID=%d --> (i,j,k)=(%d, %d, %d)\n", static_cast<int>(LID),
+           static_cast<int>(fineNodeTuple[0]),
+           static_cast<int>(fineNodeTuple[1]),
+           static_cast<int>(fineNodeTuple[2]));
+  }
+};
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(IndexManager_kokkos, IndexManager, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<Teuchos::FancyOStream> fout = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
+  fout->setShowAllFrontMatter(false).setShowProcRank(true);
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  RCP<const Teuchos::Comm<int> > comm = MueLuTests::TestHelpers_kokkos::Parameters::getDefaultComm();
+
+  // Set global geometric data
+  const int numDimensions      = 3;
+  const int interpolationOrder = 0;
+  Array<GO> meshData;
+  Array<GO> gNodesPerDir(3);
+  Array<LO> lNodesPerDir(3);
+  Array<LO> coarseRate(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 5;
+      coarseRate[dim]   = 2;
+    } else {
+      gNodesPerDir[dim] = -1;
+      coarseRate[dim]   = 1;
     }
+  }
+
+  RCP<RealValuedMultiVector> coords =
+      MueLuTests::TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions,
+                                                                                       gNodesPerDir,
+                                                                                       lNodesPerDir,
+                                                                                       meshData,
+                                                                                       "Local Lexicographic");
 
-    RCP<RealValuedMultiVector> coords =
-      MueLuTests::TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions,
-                                                                                    gNodesPerDir,
-                                                                                    lNodesPerDir,
-                                                                                    meshData,
-                                                                                    "Local Lexicographic");
-
-    std::cout << "Hello!" << std::endl;
-    RCP<IndexManager_kokkos> myIndexManager = rcp(new IndexManager_kokkos(numDimensions,
-                                                                          interpolationOrder,
-                                                                          comm->getRank(),
-                                                                          lNodesPerDir,
-                                                                          coarseRate));
-
-    using execution_space = typename IndexManager_kokkos::execution_space;
-    using memory_space    = typename IndexManager_kokkos::memory_space;
-    using device_type     = typename IndexManager_kokkos::device_type;
-
-    Kokkos::fence();
-    std::cout << "Allocate host views before performing checks" << std::endl;
-
-    typename Kokkos::View<const int[3], device_type>::HostMirror coarseRate_h
-      = Kokkos::create_mirror_view(myIndexManager->getCoarseningRates());
-    Kokkos::deep_copy(coarseRate_h, myIndexManager->getCoarseningRates());
-
-    typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h
-      = Kokkos::create_mirror_view(myIndexManager->getLocalFineNodesPerDir());
-    Kokkos::deep_copy(lFineNodesPerDir_h, myIndexManager->getLocalFineNodesPerDir());
-
-    typename Kokkos::View<LO[3], device_type>::HostMirror lCoarseNodesPerDir_h
-      = Kokkos::create_mirror_view(myIndexManager->getCoarseNodesPerDir());
-    Kokkos::deep_copy(lCoarseNodesPerDir_h, myIndexManager->getCoarseNodesPerDir());
-
-    Kokkos::fence();
-    std::cout << "Start checking process" << std::endl;
-
-    int chk = 0;
-    if(myIndexManager->getInterpolationOrder()      != interpolationOrder)          {chk = -1;}
-    if(myIndexManager->getNumDimensions()           != numDimensions)               {chk = -1;}
-    for(int dim = 0; dim < 3; ++dim) {
-      if(coarseRate_h(dim)                              != coarseRate[dim])         {chk = -1;}
+  std::cout << "Hello!" << std::endl;
+  RCP<IndexManager_kokkos> myIndexManager = rcp(new IndexManager_kokkos(numDimensions,
+                                                                        interpolationOrder,
+                                                                        comm->getRank(),
+                                                                        lNodesPerDir,
+                                                                        coarseRate));
+
+  using execution_space = typename IndexManager_kokkos::execution_space;
+  using memory_space    = typename IndexManager_kokkos::memory_space;
+  using device_type     = typename IndexManager_kokkos::device_type;
+
+  Kokkos::fence();
+  std::cout << "Allocate host views before performing checks" << std::endl;
+
+  typename Kokkos::View<const int[3], device_type>::HostMirror coarseRate_h = Kokkos::create_mirror_view(myIndexManager->getCoarseningRates());
+  Kokkos::deep_copy(coarseRate_h, myIndexManager->getCoarseningRates());
+
+  typename Kokkos::View<LO[3], device_type>::HostMirror lFineNodesPerDir_h = Kokkos::create_mirror_view(myIndexManager->getLocalFineNodesPerDir());
+  Kokkos::deep_copy(lFineNodesPerDir_h, myIndexManager->getLocalFineNodesPerDir());
+
+  typename Kokkos::View<LO[3], device_type>::HostMirror lCoarseNodesPerDir_h = Kokkos::create_mirror_view(myIndexManager->getCoarseNodesPerDir());
+  Kokkos::deep_copy(lCoarseNodesPerDir_h, myIndexManager->getCoarseNodesPerDir());
+
+  Kokkos::fence();
+  std::cout << "Start checking process" << std::endl;
+
+  int chk = 0;
+  if (myIndexManager->getInterpolationOrder() != interpolationOrder) {
+    chk = -1;
+  }
+  if (myIndexManager->getNumDimensions() != numDimensions) {
+    chk = -1;
+  }
+  for (int dim = 0; dim < 3; ++dim) {
+    if (coarseRate_h(dim) != coarseRate[dim]) {
+      chk = -1;
+    }
+  }
+  if (comm->getSize() == 1) {
+    if (myIndexManager->getNumLocalFineNodes() != 125) {
+      chk = -1;
+    }
+    if (myIndexManager->getNumCoarseNodes() != 27) {
+      chk = -1;
     }
-    if(comm->getSize() == 1) {
-      if(myIndexManager->getNumLocalFineNodes()         != 125)                     {chk = -1;}
-      if(myIndexManager->getNumCoarseNodes()       != 27)                      {chk = -1;}
-      for(int dim = 0; dim < 3; ++dim) {
-        if(lFineNodesPerDir_h(dim)                        != lNodesPerDir[dim])     {chk = -1;}
-        if(lCoarseNodesPerDir_h(dim)                      != 3)                     {chk = -1;}
+    for (int dim = 0; dim < 3; ++dim) {
+      if (lFineNodesPerDir_h(dim) != lNodesPerDir[dim]) {
+        chk = -1;
       }
+      if (lCoarseNodesPerDir_h(dim) != 3) {
+        chk = -1;
+      }
+    }
 
-    } else if(comm->getSize() == 4) {
-      if(comm->getRank() == 0) {
-        if(myIndexManager->getNumLocalFineNodes()       != 45)                      {chk = -1;}
-        if(myIndexManager->getNumCoarseNodes()     != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,3,5}; LO lCNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(lFineNodesPerDir_h(dim)                        != lFNPD[dim])          {chk = -1;}
-          if(lCoarseNodesPerDir_h(dim)                      != lCNPD[dim])          {chk = -1;}
+  } else if (comm->getSize() == 4) {
+    if (comm->getRank() == 0) {
+      if (myIndexManager->getNumLocalFineNodes() != 45) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumCoarseNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 3, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (lFineNodesPerDir_h(dim) != lFNPD[dim]) {
+          chk = -1;
         }
-      } else if(comm->getRank() == 1) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumCoarseNodes()     != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,3,5}; LO lCNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(lFineNodesPerDir_h(dim)                        != lFNPD[dim])          {chk = -1;}
-          if(lCoarseNodesPerDir_h(dim)                      != lCNPD[dim])          {chk = -1;}
+        if (lCoarseNodesPerDir_h(dim) != lCNPD[dim]) {
+          chk = -1;
         }
-      } else if(comm->getRank() == 2) {
-        if(myIndexManager->getNumLocalFineNodes()       != 30)                      {chk = -1;}
-        if(myIndexManager->getNumCoarseNodes()     != 12)                      {chk = -1;}
-        LO lFNPD[3] = {3,2,5}; LO lCNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(lFineNodesPerDir_h(dim)                        != lFNPD[dim])          {chk = -1;}
-          if(lCoarseNodesPerDir_h(dim)                      != lCNPD[dim])          {chk = -1;}
+      }
+    } else if (comm->getRank() == 1) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumCoarseNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 3, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (lFineNodesPerDir_h(dim) != lFNPD[dim]) {
+          chk = -1;
         }
-      } else if(comm->getRank() == 3) {
-        if(myIndexManager->getNumLocalFineNodes()       != 20)                      {chk = -1;}
-        if(myIndexManager->getNumCoarseNodes()     != 12)                      {chk = -1;}
-        LO lFNPD[3] = {2,2,5}; LO lCNPD[3] = {2,2,3};
-        for(int dim = 0; dim < 3; ++dim) {
-          if(lFineNodesPerDir_h(dim)                        != lFNPD[dim])          {chk = -1;}
-          if(lCoarseNodesPerDir_h(dim)                      != lCNPD[dim])          {chk = -1;}
+        if (lCoarseNodesPerDir_h(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 2) {
+      if (myIndexManager->getNumLocalFineNodes() != 30) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumCoarseNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {3, 2, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (lFineNodesPerDir_h(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (lCoarseNodesPerDir_h(dim) != lCNPD[dim]) {
+          chk = -1;
+        }
+      }
+    } else if (comm->getRank() == 3) {
+      if (myIndexManager->getNumLocalFineNodes() != 20) {
+        chk = -1;
+      }
+      if (myIndexManager->getNumCoarseNodes() != 12) {
+        chk = -1;
+      }
+      LO lFNPD[3] = {2, 2, 5};
+      LO lCNPD[3] = {2, 2, 3};
+      for (int dim = 0; dim < 3; ++dim) {
+        if (lFineNodesPerDir_h(dim) != lFNPD[dim]) {
+          chk = -1;
+        }
+        if (lCoarseNodesPerDir_h(dim) != lCNPD[dim]) {
+          chk = -1;
         }
       }
     }
+  }
 
-    int gbl_chk[1] = {0};
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
-    TEST_EQUALITY(gbl_chk[0], 0);
+  int gbl_chk[1] = {0};
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &chk, gbl_chk);
+  TEST_EQUALITY(gbl_chk[0], 0);
 
-    std::cout << "Testing UIM methods call from device." << std::endl;
-
-    testFunctor<IndexManager_kokkos, LO> myTestFunctor(myIndexManager);
-    Kokkos::parallel_for("test UIM::getFineLID2FineTuple()",
-                         Kokkos::RangePolicy<execution_space>(0, 125),
-                         myTestFunctor);
-  } //Uncoupled
+  std::cout << "Testing UIM methods call from device." << std::endl;
 
+  testFunctor<IndexManager_kokkos, LO> myTestFunctor(myIndexManager);
+  Kokkos::parallel_for("test UIM::getFineLID2FineTuple()",
+                       Kokkos::RangePolicy<execution_space>(0, 125),
+                       myTestFunctor);
+}  // Uncoupled
 
 #ifdef HAVE_MUELU_SYCL
-  // Enabling this w/ SYCL yields the error "SYCL kernel cannot call a variadic function"
-#define MUELU_ETI_GROUP(SC,LO,GO,NO)
+// Enabling this w/ SYCL yields the error "SYCL kernel cannot call a variadic function"
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)
 #else
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(IndexManager_kokkos, IndexManager, SC, LO, GO, NO)
 #endif
-  
+
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/LWGraph_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/LWGraph_kokkos.cpp
index ae14f7496ebc..de5b32f58088 100644
--- a/packages/muelu/test/unit_tests_kokkos/LWGraph_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/LWGraph_kokkos.cpp
@@ -57,135 +57,138 @@
 
 namespace MueLuTests {
 
-  // Little utility to generate a LWGraph_kokkos.
-  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  Teuchos::RCP<MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node> >
-  gimmeLWGraph(const Teuchos::RCP<Xpetra::Matrix<Scalar,LocalOrdinal,GlobalOrdinal,Node> >& A,
-               Teuchos::RCP<MueLu::AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node> >& amalgInfo) {
-#   include "MueLu_UseShortNames.hpp"
+// Little utility to generate a LWGraph_kokkos.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+Teuchos::RCP<MueLu::LWGraph_kokkos<LocalOrdinal, GlobalOrdinal, Node> >
+gimmeLWGraph(const Teuchos::RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& A,
+             Teuchos::RCP<MueLu::AmalgamationInfo<LocalOrdinal, GlobalOrdinal, Node> >& amalgInfo) {
+#include "MueLu_UseShortNames.hpp"
+
+  Level level;
+
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(level);
+  level.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  level.Request("Graph", dropFact.get());
+  level.Request("UnAmalgamationInfo", amalgFact.get());
+
+  level.Request(*dropFact);
+  dropFact->Build(level);
+
+  auto graph = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
+  amalgInfo  = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo", amalgFact.get());
+  level.Release("UnAmalgamationInfo", amalgFact.get());
+  level.Release("Graph", dropFact.get());
+  return graph;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph_kokkos, CreateLWGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    Level level;
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
 
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createSingleLevelHierarchy(level);
-    level.Set("A", A);
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<LWGraph_kokkos> graph = gimmeLWGraph(A, amalgInfo);
 
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact  = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  const int numRanks = graph->GetComm()->getSize();
 
-    level.Request("Graph",              dropFact.get());
-    level.Request("UnAmalgamationInfo", amalgFact.get());
+  TEST_EQUALITY(graph != Teuchos::null, true);
+  if (numRanks == 1) {
+    TEST_EQUALITY(graph->GetNodeNumVertices() == 16, true);
+  } else if (numRanks == 4) {
+    TEST_EQUALITY(graph->GetNodeNumVertices() == 4, true);
+  }
+}  // CreateLWGraph
 
-    level.Request(*dropFact);
-    dropFact->Build(level);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph_kokkos, LocalGraphData, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    auto graph = level.Get<RCP<LWGraph_kokkos> >("Graph", dropFact.get());
-    amalgInfo = level.Get<RCP<AmalgamationInfo> >("UnAmalgamationInfo", amalgFact.get());
-    level.Release("UnAmalgamationInfo", amalgFact.get());
-    level.Release("Graph",              dropFact.get());
-    return graph;
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph_kokkos, CreateLWGraph, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
+  RCP<AmalgamationInfo> amalgInfo;
+  RCP<LWGraph_kokkos> graph = gimmeLWGraph(A, amalgInfo);
 
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<LWGraph_kokkos> graph = gimmeLWGraph(A, amalgInfo);
+  // const int numRanks = graph->GetComm()->getSize();
+  // const int myRank   = graph->GetComm()->getRank();
 
-    const int numRanks = graph->GetComm()->getSize();
+  using row_map_type = typename LWGraph_kokkos::local_graph_type::row_map_type;
+  using entries_type = typename LWGraph_kokkos::local_graph_type::entries_type;
 
-    TEST_EQUALITY(graph != Teuchos::null,                true);
-    if(numRanks == 1) {
-      TEST_EQUALITY(graph->GetNodeNumVertices() == 16,   true);
-    } else if(numRanks == 4) {
-      TEST_EQUALITY(graph->GetNodeNumVertices() == 4,    true);
-    }
-  } // CreateLWGraph
+  row_map_type rowPtrs                          = graph->getLocalLWGraph().getRowPtrs();
+  entries_type entries                          = graph->getLocalLWGraph().getEntries();
+  typename row_map_type::HostMirror rowPtrsHost = Kokkos::create_mirror_view(rowPtrs);
+  typename entries_type::HostMirror entriesHost = Kokkos::create_mirror_view(entries);
+  Kokkos::deep_copy(rowPtrsHost, rowPtrs);
+  Kokkos::deep_copy(entriesHost, entries);
+  rowPtrs = row_map_type();
+  entries = entries_type();
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(LWGraph_kokkos, LocalGraphData, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  // Check that the graph is not null
+  TEST_EQUALITY(graph != Teuchos::null, true);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16);
-
-    RCP<AmalgamationInfo> amalgInfo;
-    RCP<LWGraph_kokkos> graph = gimmeLWGraph(A, amalgInfo);
-
-    // const int numRanks = graph->GetComm()->getSize();
-    // const int myRank   = graph->GetComm()->getRank();
-
-    using row_map_type = typename LWGraph_kokkos::local_graph_type::row_map_type;
-    using entries_type = typename LWGraph_kokkos::local_graph_type::entries_type;
-
-    row_map_type rowPtrs = graph->getLocalLWGraph().getRowPtrs();
-    entries_type entries = graph->getLocalLWGraph().getEntries();
-    typename row_map_type::HostMirror rowPtrsHost = Kokkos::create_mirror_view(rowPtrs);
-    typename entries_type::HostMirror entriesHost = Kokkos::create_mirror_view(entries);
-    Kokkos::deep_copy(rowPtrsHost, rowPtrs);
-    Kokkos::deep_copy(entriesHost, entries);
-    rowPtrs = row_map_type();
-    entries = entries_type();
-
-    // Check that the graph is not null
-    TEST_EQUALITY(graph != Teuchos::null, true);
-
-    {
-      // Since this is a scalar problem and no entries are dropped
-      // we can simply compare the value from LWGraph with those in A
-      auto numrows=A->getLocalNumRows(), nument=A->getLocalNumEntries();
-      int result=0;
-      auto lclLWGraph = graph->getLocalLWGraph();
-      Kokkos::parallel_reduce("MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t> (0,1),
-			      KOKKOS_LAMBDA(const LO i, int &incorrect) {
-				if (lclLWGraph.GetNodeNumVertices() != numrows)
-				  incorrect++;
-				if (lclLWGraph.GetNodeNumEdges() != nument)
-				  incorrect++;
-			      }, result);
-      TEST_EQUALITY(result, 0);
+  {
+    // Since this is a scalar problem and no entries are dropped
+    // we can simply compare the value from LWGraph with those in A
+    auto numrows = A->getLocalNumRows(), nument = A->getLocalNumEntries();
+    int result      = 0;
+    auto lclLWGraph = graph->getLocalLWGraph();
+    Kokkos::parallel_reduce(
+        "MueLu:TentativePF:Build:compute_agg_sizes", Kokkos::RangePolicy<typename NO::execution_space, size_t>(0, 1),
+        KOKKOS_LAMBDA(const LO i, int& incorrect) {
+          if (lclLWGraph.GetNodeNumVertices() != numrows)
+            incorrect++;
+          if (lclLWGraph.GetNodeNumEdges() != nument)
+            incorrect++;
+        },
+        result);
+    TEST_EQUALITY(result, 0);
+  }
 
+  graph.reset();
+  Array<LO> indices(3);
+  Array<SC> values(3);
+  size_t numEntries;
+  bool chk_rowPtr = true;
+  bool chk_colInd = true;
+  for (size_t rowIdx = 0; rowIdx < A->getLocalNumRows(); ++rowIdx) {
+    A->getLocalRowCopy(rowIdx, indices(), values(), numEntries);
+    if (rowPtrsHost(rowIdx + 1) - rowPtrsHost(rowIdx) != numEntries) {
+      chk_rowPtr = false;
     }
-
-    graph.reset();
-    Array<LO> indices(3);
-    Array<SC> values(3);
-    size_t numEntries;
-    bool chk_rowPtr = true;
-    bool chk_colInd = true;
-    for(size_t rowIdx = 0; rowIdx < A->getLocalNumRows(); ++rowIdx) {
-      A->getLocalRowCopy(rowIdx, indices(), values(), numEntries);
-      if(rowPtrsHost(rowIdx + 1) - rowPtrsHost(rowIdx) != numEntries) {chk_rowPtr = false;}
-      for(size_t entryIdx = 0; entryIdx < numEntries; ++entryIdx) {
-        if(entriesHost(rowPtrsHost(rowIdx) + entryIdx) != indices[entryIdx]) {chk_colInd = false;}
+    for (size_t entryIdx = 0; entryIdx < numEntries; ++entryIdx) {
+      if (entriesHost(rowPtrsHost(rowIdx) + entryIdx) != indices[entryIdx]) {
+        chk_colInd = false;
       }
     }
-    TEST_EQUALITY(chk_rowPtr, true);
-    TEST_EQUALITY(chk_colInd, true);
-  } // LocalGraphData
+  }
+  TEST_EQUALITY(chk_rowPtr, true);
+  TEST_EQUALITY(chk_colInd, true);
+}  // LocalGraphData
 
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LWGraph_kokkos, CreateLWGraph,  SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                               \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LWGraph_kokkos, CreateLWGraph, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(LWGraph_kokkos, LocalGraphData, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/MatrixFree_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/MatrixFree_kokkos.cpp
index 6f81443e4036..34d2e12f9320 100644
--- a/packages/muelu/test/unit_tests_kokkos/MatrixFree_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/MatrixFree_kokkos.cpp
@@ -62,470 +62,460 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<MatrixFreeTentativePFactory> MFtentPFact = rcp(new MatrixFreeTentativePFactory);
-    TEST_EQUALITY(MFtentPFact != Teuchos::null, true);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
+  out << "version: " << MueLu::Version() << std::endl;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+  RCP<MatrixFreeTentativePFactory> MFtentPFact = rcp(new MatrixFreeTentativePFactory);
+  TEST_EQUALITY(MFtentPFact != Teuchos::null, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test with user-supplied nullspace" << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test with user-supplied nullspace" << std::endl;
 
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    fineLevel  .SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build1DPoisson(199);
-    fineLevel.Request("A");
-    fineLevel.Set    ("A", A);
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
 
-    // Only one NS vector -> exercises manual orthogonalization
-    LO NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace", nullSpace);
+  // Only one NS vector -> exercises manual orthogonalization
+  LO NSdim                   = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
 
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
 
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: deterministic",          true);
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: deterministic", true);
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    RCP<MatrixFreeTentativePFactory> MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
-    MFTentativePFact->SetFactory("Aggregates",         aggFact);
-    MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    MFTentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  RCP<MatrixFreeTentativePFactory> MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
+  MFTentativePFact->SetFactory("Aggregates", aggFact);
+  MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  MFTentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    coarseLevel.Request("P",            MFTentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",    MFTentativePFact.get());
-    coarseLevel.Request(*MFTentativePFact);
-    MFTentativePFact->Build(fineLevel, coarseLevel);
+  coarseLevel.Request("P", MFTentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", MFTentativePFact.get());
+  coarseLevel.Request(*MFTentativePFact);
+  MFTentativePFact->Build(fineLevel, coarseLevel);
 
-    RCP<Operator> Ptent;
-    coarseLevel.Get("P", Ptent, MFTentativePFact.get());
+  RCP<Operator> Ptent;
+  coarseLevel.Get("P", Ptent, MFTentativePFact.get());
 
-    RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
+  RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
 
-    // Check interpolation by computing ||fineNS - P*coarseNS||
-    // PtN = P*coarseNS
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
-    Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+  // Check interpolation by computing ||fineNS - P*coarseNS||
+  // PtN = P*coarseNS
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    diff->putScalar(0.0);
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    coarseLevel.Release("P",         MFTentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace", MFTentativePFact.get());
+  coarseLevel.Release("P", MFTentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", MFTentativePFact.get());
 
-    // diff = fineNS - PtN = fineNS - P*coarseNS
-    diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
+  // diff = fineNS - PtN = fineNS - P*coarseNS
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
 
-    Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LO i = 0; i < NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
-    }
+  Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LO i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_COMPARE_CONST(norms[i], <, 100 * TMT::eps());
+  }
 
-  } // MakeTentative
+}  // MakeTentative
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentativeVectorBasedUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentativeVectorBasedUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test when nullspace isn't supplied by user" << std::endl;
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test when nullspace isn't supplied by user" << std::endl;
 
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(200);
+  auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(200);
 
-    A->SetFixedBlockSize(2);
+  A->SetFixedBlockSize(2);
 
-    fineLevel.Set("A", A);
+  fineLevel.Set("A", A);
 
-    auto amalgFact = rcp(new AmalgamationFactory());
+  auto amalgFact = rcp(new AmalgamationFactory());
 
-    auto dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  auto dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: deterministic",         true);
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: deterministic", true);
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
-    auto coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  auto coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    auto MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
-    MFTentativePFact->SetFactory("Aggregates",         aggFact);
-    MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    MFTentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  auto MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
+  MFTentativePFact->SetFactory("Aggregates", aggFact);
+  MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  MFTentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
+  coarseLevel.Request("P", MFTentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", MFTentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*MFTentativePFact);
+  MFTentativePFact->Build(fineLevel, coarseLevel);
 
-    coarseLevel.Request("P",MFTentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace", MFTentativePFact.get());  // request coarse nullspace
-    coarseLevel.Request(*MFTentativePFact);
-    MFTentativePFact->Build(fineLevel,coarseLevel);
+  RCP<Operator> Ptent;
+  coarseLevel.Get("P", Ptent, MFTentativePFact.get());
 
-    RCP<Operator> Ptent;
-    coarseLevel.Get("P",Ptent,MFTentativePFact.get());
+  auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
 
-    auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
+  size_t NSdim = coarseNullspace->getNumVectors();
+  TEST_EQUALITY(NSdim, 2);
 
-    size_t NSdim = coarseNullspace->getNumVectors();
-    TEST_EQUALITY(NSdim, 2);
+  // coarseNullspace->describe(out, Teuchos::VERB_EXTREME);
 
-    //coarseNullspace->describe(out, Teuchos::VERB_EXTREME);
+  // Check interpolation by computing ||fineNS - P*coarseNS||
+  auto PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  // Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0); // TODO: don't support strides :)
 
-    // Check interpolation by computing ||fineNS - P*coarseNS||
-    auto PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
-    //Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0); // TODO: don't support strides :)
+  auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    diff->putScalar(0.0);
+  coarseLevel.Release("P", MFTentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", MFTentativePFact.get());
+
+  auto nspFact = Teuchos::rcp(new NullspaceFactory_kokkos());
+  fineLevel.Request("Nullspace", nspFact.get());
 
-    coarseLevel.Release("P",         MFTentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace", MFTentativePFact.get());
+  nspFact->Build(fineLevel);
 
-    auto nspFact = Teuchos::rcp(new NullspaceFactory_kokkos());
-    fineLevel.Request("Nullspace",nspFact.get());
+  auto fineNullspace = fineLevel.Get<RCP<MultiVector> >("Nullspace", nspFact.get());
 
-    nspFact->Build(fineLevel);
+  TEST_EQUALITY(fineNullspace->getNumVectors(), 2);
 
-    auto fineNullspace = fineLevel.Get<RCP<MultiVector> >("Nullspace", nspFact.get());
+  // diff = fineNS - (P*coarseNS)
+  diff->update(1.0, *fineNullspace, -1.0, *PtN, 0.0);
 
-    TEST_EQUALITY(fineNullspace->getNumVectors(), 2);
+  Array<magnitude_type> norms(NSdim);
+  diff->norm2(norms);
+  // for (decltype(NSdim) i = 0; i < NSdim; ++i)
+  //   TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
+}
 
-    // diff = fineNS - (P*coarseNS)
-    diff->update(1.0, *fineNullspace, -1.0, *PtN, 0.0);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentativeUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
+
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    Array<magnitude_type> norms(NSdim);
-    diff->norm2(norms);
-    // for (decltype(NSdim) i = 0; i < NSdim; ++i)
-    //   TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MakeTentativeUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test when nullspace isn't supplied by user" << std::endl;
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test when nullspace isn't supplied by user" << std::endl;
+  fineLevel.Set("A", A);
 
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  auto amalgFact = rcp(new AmalgamationFactory());
 
-    auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
+  auto dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    fineLevel.Set("A", A);
+  auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: deterministic", true);
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
-    auto amalgFact = rcp(new AmalgamationFactory());
+  auto coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    auto dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  auto MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
+  MFTentativePFact->SetFactory("Aggregates", aggFact);
+  MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  MFTentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: deterministic",         true);
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  coarseLevel.Request("P", MFTentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", MFTentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*MFTentativePFact);
+  MFTentativePFact->Build(fineLevel, coarseLevel);
 
-    auto coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  RCP<Operator> Ptent;
+  coarseLevel.Get("P", Ptent, MFTentativePFact.get());
 
-    auto MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
-    MFTentativePFact->SetFactory("Aggregates",         aggFact);
-    MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    MFTentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
 
+  coarseLevel.Release("P", MFTentativePFact.get());          // release Ptent
+  coarseLevel.Release("Nullspace", MFTentativePFact.get());  // release coarse nullspace
 
-    coarseLevel.Request("P",MFTentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace", MFTentativePFact.get());  // request coarse nullspace
-    coarseLevel.Request(*MFTentativePFact);
-    MFTentativePFact->Build(fineLevel,coarseLevel);
+  // grab default fine level nullspace (vector of all ones)
+  auto nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar(1.0);
 
-    RCP<Operator> Ptent;
-    coarseLevel.Get("P",Ptent,MFTentativePFact.get());
+  // check interpolation
+  LO NSdim = 1;
+  auto PtN = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",MFTentativePFact.get());
+  auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    coarseLevel.Release("P",MFTentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",MFTentativePFact.get());   // release coarse nullspace
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
 
-    //grab default fine level nullspace (vector of all ones)
-    auto nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
-    nullSpace->putScalar(1.0);
+  Teuchos::Array<magnitude_type> norms(NSdim);
+  diff->norm2(norms);
+  for (LO i = 0; i < NSdim; ++i)
+    TEST_COMPARE_CONST(norms[i], <, 100 * TMT::eps());
 
-    //check interpolation
-    LO NSdim = 1;
-    auto PtN = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    Ptent->apply(*coarseNullspace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
+}  // MakeTentativeUsingDefaultNullspace
 
-    auto diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MatrixVsMatrixFree, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
 
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    Teuchos::Array<magnitude_type> norms(NSdim);
-    diff->norm2(norms);
-    for (LO i=0; i<NSdim; ++i)
-      TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
+
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-  } //MakeTentativeUsingDefaultNullspace
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test matvec speed in matrix and matrix-free versions and compare results" << std::endl;
 
+  unsigned int numMatVecs = 100;
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(MatrixFreeTentativePFactory, MatrixVsMatrixFree, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
+  RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
+  RCP<MultiVector> Xcoarse;
+  RCP<MultiVector> MFXcoarse;
+  RCP<MultiVector> diff;
+  X->randomize();
+
+  // Matrix version
   {
-#include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+    Level fineLevel, coarseLevel;
+    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+    coarseLevel.SetFactoryManager(Teuchos::null);
+    fineLevel.Request("A");
+    fineLevel.Set("A", A);
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+    // Only one NS vector -> exercises manual orthogonalization
+    LO NSdim                   = 1;
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+    nullSpace->putScalar(1.0);
+    fineLevel.Set("Nullspace", nullSpace);
+
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+
+    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+    ParameterList aggParams;
+    aggParams.set("aggregation: ordering", "natural");
+    aggParams.set("aggregation: deterministic", true);
+    aggParams.set("aggregation: min agg size", 3);
+    aggParams.set("aggregation: max selected neighbors", 0);
+    aggFact->SetParameterList(aggParams);
+    aggFact->SetFactory("DofsPerNode", dropFact);
+    aggFact->SetFactory("Graph", dropFact);
 
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+    coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test matvec speed in matrix and matrix-free versions and compare results" << std::endl;
+    RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
+    TentativePFact->SetFactory("Aggregates", aggFact);
+    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    unsigned int numMatVecs = 100;
+    coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+    coarseLevel.Request("Nullspace", TentativePFact.get());
+    coarseLevel.Request(*TentativePFact);
+    TentativePFact->Build(fineLevel, coarseLevel);
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build1DPoisson(199);
-    RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-    RCP<MultiVector> Xcoarse;
-    RCP<MultiVector> MFXcoarse;
-    RCP<MultiVector> diff;
-    X->randomize();
+    RCP<Matrix> Ptent;
+    coarseLevel.Get("P", Ptent, TentativePFact.get());
+    Xcoarse = MultiVectorFactory::Build(Ptent->getDomainMap(), 1);
+    Ptent->apply(*X, *Xcoarse, Teuchos::TRANS);
+
+    RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+    RCP<MultiVector> PtN             = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+
+    // warm-up apply operations
+    for (unsigned int i = 0; i < 5; ++i)
+      Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    // Matrix version
     {
-      Level fineLevel, coarseLevel;
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-      fineLevel  .SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-      coarseLevel.SetFactoryManager(Teuchos::null);
-      fineLevel.Request("A");
-      fineLevel.Set    ("A", A);
-
-      // Only one NS vector -> exercises manual orthogonalization
-      LO NSdim = 1;
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-      nullSpace->putScalar(1.0);
-      fineLevel.Set("Nullspace", nullSpace);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-
-      RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      ParameterList aggParams;
-      aggParams.set("aggregation: ordering",              "natural");
-      aggParams.set("aggregation: deterministic",          true);
-      aggParams.set("aggregation: min agg size",           3);
-      aggParams.set("aggregation: max selected neighbors", 0);
-      aggFact->SetParameterList(aggParams);
-      aggFact->SetFactory("DofsPerNode",  dropFact);
-      aggFact->SetFactory("Graph",        dropFact);
-
-      RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-      coarseMapFact->SetFactory("Aggregates", aggFact);
-
-      RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
-      TentativePFact->SetFactory("Aggregates",         aggFact);
-      TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      TentativePFact->SetFactory("CoarseMap",          coarseMapFact);
-
-      coarseLevel.Request("P",            TentativePFact.get());  // request Ptent
-      coarseLevel.Request("Nullspace",    TentativePFact.get());
-      coarseLevel.Request(*TentativePFact);
-      TentativePFact->Build(fineLevel, coarseLevel);
-
-      RCP<Matrix> Ptent;
-      coarseLevel.Get("P", Ptent, TentativePFact.get());
-      Xcoarse = MultiVectorFactory::Build(Ptent->getDomainMap(),1);
-      Ptent->apply(*X,*Xcoarse,Teuchos::TRANS);
-
-      RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
-      RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
-
-      // warm-up apply operations
-      for(unsigned int i=0; i<5; ++i)
+      RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MatVec original")));
+      for (unsigned int i = 0; i < numMatVecs; ++i)
         Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
-
-      {
-        RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MatVec original")));
-        for(unsigned int i=0; i<numMatVecs; ++i)
-          Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
-      }
     }
+  }
 
-    // Matrix-free version
-    {
-      Level fineLevel, coarseLevel;
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-      fineLevel  .SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-      coarseLevel.SetFactoryManager(Teuchos::null);
-      fineLevel.Request("A");
-      fineLevel.Set    ("A", A);
-
-      // Only one NS vector -> exercises manual orthogonalization
-      LO NSdim = 1;
-      RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-      nullSpace->putScalar(1.0);
-      fineLevel.Set("Nullspace", nullSpace);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-
-      RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      ParameterList aggParams;
-      aggParams.set("aggregation: ordering",              "natural");
-      aggParams.set("aggregation: deterministic",          true);
-      aggParams.set("aggregation: min agg size",           3);
-      aggParams.set("aggregation: max selected neighbors", 0);
-      aggFact->SetParameterList(aggParams);
-      aggFact->SetFactory("DofsPerNode",  dropFact);
-      aggFact->SetFactory("Graph",        dropFact);
-
-      RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-      coarseMapFact->SetFactory("Aggregates", aggFact);
-
-      RCP<MatrixFreeTentativePFactory> MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
-      MFTentativePFact->SetFactory("Aggregates",         aggFact);
-      MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      MFTentativePFact->SetFactory("CoarseMap",          coarseMapFact);
-
-      coarseLevel.Request("P",            MFTentativePFact.get());  // request Ptent
-      coarseLevel.Request("Nullspace",    MFTentativePFact.get());
-      coarseLevel.Request(*MFTentativePFact);
-      MFTentativePFact->Build(fineLevel, coarseLevel);
-
-      RCP<Operator> MFPtent;
-      coarseLevel.Get("P", MFPtent, MFTentativePFact.get());
-      MFXcoarse = MultiVectorFactory::Build(MFPtent->getDomainMap(),1);
-      MFPtent->apply(*X,*MFXcoarse,Teuchos::TRANS);
-
-      RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
-      RCP<MultiVector> PtN = MultiVectorFactory::Build(MFPtent->getRangeMap(), NSdim);
-      // warm-up apply operations
-      for(unsigned int i=0; i<5; ++i)
-        MFPtent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+  // Matrix-free version
+  {
+    Level fineLevel, coarseLevel;
+    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+    coarseLevel.SetFactoryManager(Teuchos::null);
+    fineLevel.Request("A");
+    fineLevel.Set("A", A);
+
+    // Only one NS vector -> exercises manual orthogonalization
+    LO NSdim                   = 1;
+    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+    nullSpace->putScalar(1.0);
+    fineLevel.Set("Nullspace", nullSpace);
+
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+
+    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+    ParameterList aggParams;
+    aggParams.set("aggregation: ordering", "natural");
+    aggParams.set("aggregation: deterministic", true);
+    aggParams.set("aggregation: min agg size", 3);
+    aggParams.set("aggregation: max selected neighbors", 0);
+    aggFact->SetParameterList(aggParams);
+    aggFact->SetFactory("DofsPerNode", dropFact);
+    aggFact->SetFactory("Graph", dropFact);
+
+    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+    coarseMapFact->SetFactory("Aggregates", aggFact);
 
-      {
-        RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MatVec matrix-free")));
-        for(unsigned int i=0; i<numMatVecs; ++i)
-          MFPtent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
-      }
+    RCP<MatrixFreeTentativePFactory> MFTentativePFact = rcp(new MatrixFreeTentativePFactory());
+    MFTentativePFact->SetFactory("Aggregates", aggFact);
+    MFTentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+    MFTentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-      // need scope
-      diff = MultiVectorFactory::Build(MFPtent->getDomainMap(), 1);
-      diff->putScalar(0.0);
+    coarseLevel.Request("P", MFTentativePFact.get());  // request Ptent
+    coarseLevel.Request("Nullspace", MFTentativePFact.get());
+    coarseLevel.Request(*MFTentativePFact);
+    MFTentativePFact->Build(fineLevel, coarseLevel);
+
+    RCP<Operator> MFPtent;
+    coarseLevel.Get("P", MFPtent, MFTentativePFact.get());
+    MFXcoarse = MultiVectorFactory::Build(MFPtent->getDomainMap(), 1);
+    MFPtent->apply(*X, *MFXcoarse, Teuchos::TRANS);
+
+    RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", MFTentativePFact.get());
+    RCP<MultiVector> PtN             = MultiVectorFactory::Build(MFPtent->getRangeMap(), NSdim);
+    // warm-up apply operations
+    for (unsigned int i = 0; i < 5; ++i)
+      MFPtent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+
+    {
+      RCP<Teuchos::TimeMonitor> tm = rcp(new Teuchos::TimeMonitor(*Teuchos::TimeMonitor::getNewTimer("MatVec matrix-free")));
+      for (unsigned int i = 0; i < numMatVecs; ++i)
+        MFPtent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
     }
 
-    Teuchos::TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
-    Teuchos::TimeMonitor::zeroOutTimers();
+    // need scope
+    diff = MultiVectorFactory::Build(MFPtent->getDomainMap(), 1);
+    diff->putScalar(0.0);
+  }
+
+  Teuchos::TimeMonitor::summarize(A->getRowMap()->getComm().ptr(), std::cout, false, true, false, Teuchos::Union, "", true);
+  Teuchos::TimeMonitor::zeroOutTimers();
 
-    // Compare Xcoarse and MFXcoarse (diff = Xcoarse - MFXcoarse + 0*diff)
-    diff->update(1.0, *Xcoarse, -1.0, *MFXcoarse, 0.0);
-    Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
-    diff->norm2(norms);
-    out << "||diff||_2 = " << norms[0] << std::endl;
-    TEST_COMPARE_CONST(norms[0], <, 100*TMT::eps());
+  // Compare Xcoarse and MFXcoarse (diff = Xcoarse - MFXcoarse + 0*diff)
+  diff->update(1.0, *Xcoarse, -1.0, *MFXcoarse, 0.0);
+  Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(1);
+  diff->norm2(norms);
+  out << "||diff||_2 = " << norms[0] << std::endl;
+  TEST_COMPARE_CONST(norms[0], <, 100 * TMT::eps());
 
-  } // MatrixVsMatrixFree
+}  // MatrixVsMatrixFree
 
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, Constructor,   SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, MakeTentative, SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                                 \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, Constructor, SC, LO, GO, NO)                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, MakeTentative, SC, LO, GO, NO)                      \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, MakeTentativeUsingDefaultNullspace, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(MatrixFreeTentativePFactory, MatrixVsMatrixFree, SC, LO, GO, NO)
 
-
-
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_Common_kokkos.hpp b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_Common_kokkos.hpp
index ca45d16d4034..0aa54d8165ab 100644
--- a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_Common_kokkos.hpp
+++ b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_Common_kokkos.hpp
@@ -47,11 +47,13 @@
 #define MUELU_TEST_HELPERS_COMMON_KOKKOS_HPP
 
 #if 1
-#define MUELU_DEBUGGER_MACRO \
-  out << "debug: scalar = " << typeid(GO).name() << std::endl; \
-  out << "debug: node   = " << typeid(NO).name() << std::endl; \
-  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) out << "debug: linAlgebra = Epetra" << std::endl; \
-  else                                                               out << "debug: linAlgebra = Tpetra" << std::endl;
+#define MUELU_DEBUGGER_MACRO                                         \
+  out << "debug: scalar = " << typeid(GO).name() << std::endl;       \
+  out << "debug: node   = " << typeid(NO).name() << std::endl;       \
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) \
+    out << "debug: linAlgebra = Epetra" << std::endl;                \
+  else                                                               \
+    out << "debug: linAlgebra = Tpetra" << std::endl;
 #else
 #define MUELU_DEBUGGER_MACRO
 #endif
@@ -61,42 +63,72 @@
 
 #ifdef HAVE_MUELU_EPETRA
 
-# if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
+#if ((defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_OPENMP) || !defined(HAVE_TPETRA_INST_INT_INT))) || \
      (!defined(EPETRA_HAVE_OMP) && (!defined(HAVE_TPETRA_INST_SERIAL) || !defined(HAVE_TPETRA_INST_INT_INT))))
-  // <double,int,int,EpetraNode>: run Epetra, do not run Tpetra
-#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO) \
-  MUELU_DEBUGGER_MACRO \
-  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) { \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; } \
-  } else if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseTpetra) { \
+// <double,int,int,EpetraNode>: run Epetra, do not run Tpetra
+#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO)                                                  \
+  MUELU_DEBUGGER_MACRO                                                                         \
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {                         \
+    if (!TYPE_EQUAL(SC, double)) {                                                             \
+      out << "Skipping Epetra for SC != double" << std::endl;                                  \
+      return;                                                                                  \
+    }                                                                                          \
+    if (!TYPE_EQUAL(GO, int)) {                                                                \
+      out << "Skipping Epetra for GO != int" << std::endl;                                     \
+      return;                                                                                  \
+    }                                                                                          \
+    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                                                 \
+      out << "Skipping Epetra for NO != EpetraNode" << std::endl;                              \
+      return;                                                                                  \
+    }                                                                                          \
+  } else if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseTpetra) {                  \
     if (TYPE_EQUAL(SC, double) && TYPE_EQUAL(GO, int) && TYPE_EQUAL(NO, Xpetra::EpetraNode)) { \
-      out << "Skipping Tpetra for <double, int, EpetraNode>" << std::endl; \
-      return; \
-    } \
+      out << "Skipping Tpetra for <double, int, EpetraNode>" << std::endl;                     \
+      return;                                                                                  \
+    }                                                                                          \
   }
 
-#define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO) \
-  { out << "Skipping as we cannot run Epetra and Tpetra for the same combination of template args" << std::endl; return; }
+#define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO)                                           \
+  {                                                                                                              \
+    out << "Skipping as we cannot run Epetra and Tpetra for the same combination of template args" << std::endl; \
+    return;                                                                                                      \
+  }
 
-# else
+#else
 
-#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO) \
-  MUELU_DEBUGGER_MACRO \
+#define MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO)                          \
+  MUELU_DEBUGGER_MACRO                                                 \
   if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) { \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; } \
+    if (!TYPE_EQUAL(SC, double)) {                                     \
+      out << "Skipping Epetra for SC != double" << std::endl;          \
+      return;                                                          \
+    }                                                                  \
+    if (!TYPE_EQUAL(GO, int)) {                                        \
+      out << "Skipping Epetra for GO != int" << std::endl;             \
+      return;                                                          \
+    }                                                                  \
+    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                         \
+      out << "Skipping Epetra for NO != EpetraNode" << std::endl;      \
+      return;                                                          \
+    }                                                                  \
   }
 
 // If linAlgebra==Tpetra, but the test also requires Epetra, this macro will cause the test
 // to return early if SC!=double, GO!={int}, or NO!=Serial.
 #define MUELU_TESTING_LIMIT_EPETRA_SCOPE_TPETRA_IS_DEFAULT(SC, GO, NO) \
-    if (!TYPE_EQUAL(SC, double))                { out << "Skipping Epetra for SC != double"     << std::endl; return; } \
-    if (!TYPE_EQUAL(GO, int))                   { out << "Skipping Epetra for GO != int"        << std::endl; return; } \
-    if (!TYPE_EQUAL(NO, Xpetra::EpetraNode))    { out << "Skipping Epetra for NO != EpetraNode" << std::endl; return; }
-# endif
+  if (!TYPE_EQUAL(SC, double)) {                                       \
+    out << "Skipping Epetra for SC != double" << std::endl;            \
+    return;                                                            \
+  }                                                                    \
+  if (!TYPE_EQUAL(GO, int)) {                                          \
+    out << "Skipping Epetra for GO != int" << std::endl;               \
+    return;                                                            \
+  }                                                                    \
+  if (!TYPE_EQUAL(NO, Xpetra::EpetraNode)) {                           \
+    out << "Skipping Epetra for NO != EpetraNode" << std::endl;        \
+    return;                                                            \
+  }
+#endif
 
 #else  // HAVE_MUELU_EPETRA
 
@@ -105,16 +137,15 @@
 
 #endif  // HAVE_MUELU_EPETRA
 
-
 // Macro to set MueLu's internal oh-so FancyOStream to be the same as the one used by Teuchos' unit testing framework.
 // This prevents MueLu's output from intermingling with with the unit test pass/fail summary lines.
 #define MUELU_TESTING_SET_OSTREAM \
-   MueLu::VerboseObject::SetMueLuOStream(Teuchos::fancyOStream(out.getOStream()));
+  MueLu::VerboseObject::SetMueLuOStream(Teuchos::fancyOStream(out.getOStream()));
 
-#define MUELU_TESTING_DO_NOT_TEST(lib,packagesNotEnabled) \
-  if (TestHelpers_kokkos::Parameters::getLib() == lib) { \
+#define MUELU_TESTING_DO_NOT_TEST(lib, packagesNotEnabled)                                                                                                                                                                  \
+  if (TestHelpers_kokkos::Parameters::getLib() == lib) {                                                                                                                                                                    \
     out << "Skipping test for " << (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra ? "Epetra" : "Tpetra") << " because some required packages are not enabled (" << packagesNotEnabled << ")." << std::endl; \
-    return; \
+    return;                                                                                                                                                                                                                 \
   }
 
-#endif // ifndef MUELU_TEST_HELPERS_COMMON_KOKKOS_HPP
+#endif  // ifndef MUELU_TEST_HELPERS_COMMON_KOKKOS_HPP
diff --git a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.cpp
index 81957646d903..d02619819dd9 100644
--- a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.cpp
@@ -49,33 +49,33 @@
 
 namespace MueLuTests {
 
-  // static members initialization of the class TestHelpers::Parameters
-  Xpetra::Parameters TestHelpers_kokkos::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
+// static members initialization of the class TestHelpers::Parameters
+Xpetra::Parameters TestHelpers_kokkos::Parameters::xpetraParameters = Xpetra::Parameters(Teuchos::UnitTestRepository::getCLP());
 
-}
+}  // namespace MueLuTests
 
 namespace MueLuTests {
-  namespace TestHelpers_kokkos {
-
-    ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter) {
-      RCP<std::vector<std::string> > files = rcp(new std::vector<std::string>());
+namespace TestHelpers_kokkos {
 
-      DIR *dir = opendir(dirPath.c_str());
-      TEUCHOS_TEST_FOR_EXCEPTION(dir == NULL, MueLu::Exceptions::RuntimeError, "GetFileList(\"" + dirPath + "\") : " + strerror(errno));
+ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter) {
+  RCP<std::vector<std::string> > files = rcp(new std::vector<std::string>());
 
-      struct dirent *dirEntry;
-      while ((dirEntry = readdir(dir)) != NULL) {
-        std::string dirEntryS(dirEntry->d_name);
+  DIR* dir = opendir(dirPath.c_str());
+  TEUCHOS_TEST_FOR_EXCEPTION(dir == NULL, MueLu::Exceptions::RuntimeError, "GetFileList(\"" + dirPath + "\") : " + strerror(errno));
 
-        size_t pos = dirEntryS.rfind(filter);
-        if (pos != std::string::npos && pos+filter.size() == dirEntryS.size())
-          files->push_back(std::string(dirEntryS));
-      }
+  struct dirent* dirEntry;
+  while ((dirEntry = readdir(dir)) != NULL) {
+    std::string dirEntryS(dirEntry->d_name);
 
-      closedir(dir);
+    size_t pos = dirEntryS.rfind(filter);
+    if (pos != std::string::npos && pos + filter.size() == dirEntryS.size())
+      files->push_back(std::string(dirEntryS));
+  }
 
-      return arcp(files);
-    }
+  closedir(dir);
 
-  }
+  return arcp(files);
 }
+
+}  // namespace TestHelpers_kokkos
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.hpp b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.hpp
index 6f0229a607f6..7d2f4b2085e2 100644
--- a/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.hpp
+++ b/packages/muelu/test/unit_tests_kokkos/MueLu_TestHelpers_kokkos.hpp
@@ -81,8 +81,6 @@
 
 #include "MueLu_NoFactory.hpp"
 
-
-
 // Conditional Tpetra stuff
 #include <TpetraCore_config.h>
 #include <Xpetra_TpetraCrsGraph.hpp>
@@ -95,477 +93,467 @@
 #include <MueLu_TestHelpers_Common_kokkos.hpp>
 
 namespace MueLuTests {
-  using Teuchos::arcp;
-  using Teuchos::arcpFromArrayView;
-  using Teuchos::arcp_reinterpret_cast;
-  using Teuchos::Array;
-  using Teuchos::ArrayRCP;
-  using Teuchos::ArrayView;
-  using Teuchos::as;
-  using Teuchos::null;
-  using Teuchos::ParameterList;
-  using Teuchos::rcp;
-  using Teuchos::RCP;
-  using Teuchos::rcp_dynamic_cast;
-  using Teuchos::rcpFromRef;
-  using Teuchos::rcp_implicit_cast;
-
-# include <MueLu_TestHelpers_Common_kokkos.hpp>
-
-  namespace TestHelpers_kokkos {
+using Teuchos::arcp;
+using Teuchos::arcp_reinterpret_cast;
+using Teuchos::arcpFromArrayView;
+using Teuchos::Array;
+using Teuchos::ArrayRCP;
+using Teuchos::ArrayView;
+using Teuchos::as;
+using Teuchos::null;
+using Teuchos::ParameterList;
+using Teuchos::rcp;
+using Teuchos::RCP;
+using Teuchos::rcp_dynamic_cast;
+using Teuchos::rcp_implicit_cast;
+using Teuchos::rcpFromRef;
 
-    using Xpetra::global_size_t;
+#include <MueLu_TestHelpers_Common_kokkos.hpp>
 
-    class Parameters {
+namespace TestHelpers_kokkos {
 
-    private:
-      Parameters() {} // static class
+using Xpetra::global_size_t;
 
-    public:
+class Parameters {
+ private:
+  Parameters() {}  // static class
 
-      static Xpetra::Parameters xpetraParameters;
+ public:
+  static Xpetra::Parameters xpetraParameters;
 
-      inline static RCP<const Teuchos::Comm<int> > getDefaultComm() {
-        return Xpetra::DefaultPlatform::getDefaultPlatform().getComm();
-      }
+  inline static RCP<const Teuchos::Comm<int> > getDefaultComm() {
+    return Xpetra::DefaultPlatform::getDefaultPlatform().getComm();
+  }
 
-      inline static Xpetra::UnderlyingLib getLib() {
-        return TestHelpers_kokkos::Parameters::xpetraParameters.GetLib();
-      }
-    };
+  inline static Xpetra::UnderlyingLib getLib() {
+    return TestHelpers_kokkos::Parameters::xpetraParameters.GetLib();
+  }
+};
 
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class TestFactory {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TestFactory {
 #include "MueLu_UseShortNames.hpp"
 
-    private:
-      TestFactory() {} // static class
-
-    public:
+ private:
+  TestFactory() {}  // static class
+
+ public:
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+  typedef Xpetra::MultiVectorFactory<real_type, LO, GO, NO> RealValuedMultiVectorFactory;
+
+  // Create a map containing a specified number of local elements per process.
+  static const RCP<const Map> BuildMap(LO numElementsPerProc) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+    const global_size_t INVALID         = Teuchos::OrdinalTraits<global_size_t>::invalid();
+
+    return MapFactory::Build(TestHelpers_kokkos::Parameters::getLib(), INVALID, numElementsPerProc, 0, comm);
+
+  }  // BuildMap()
+
+  // Create a matrix as specified by parameter list options
+  static RCP<Matrix> BuildMatrix(ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers_kokkos::Parameters::getLib();
+
+    GO nx, ny, nz;
+    nx = ny = nz = 5;
+    nx           = matrixList.get("nx", nx);
+    ny           = matrixList.get("ny", ny);
+    nz           = matrixList.get("nz", nz);
+
+    std::string matrixType = matrixList.get("matrixType", "Laplace1D");
+    GO numGlobalElements;  // global_size_t
+    if (matrixType == "Laplace1D")
+      numGlobalElements = nx;
+    else if (matrixType == "Laplace2D" || matrixType == "Star2D")
+      numGlobalElements = nx * ny;
+    else if (matrixType == "Laplace3D")
+      numGlobalElements = nx * ny * nz;
+    else {
+      std::string msg = matrixType + " is unsupported (in unit testing)";
+      throw(MueLu::Exceptions::RuntimeError(msg));
+    }
+
+    RCP<const Map> map = MapFactory::Build(lib, numGlobalElements, 0, comm);
+    RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+        Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(matrixType, map, matrixList);
+    RCP<Matrix> Op = Pr->BuildMatrix();
+
+    return Op;
+  }
 
-      typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-      typedef Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-      typedef Xpetra::MultiVectorFactory<real_type,LO,GO,NO> RealValuedMultiVectorFactory;
+  // Create a 1D Poisson matrix with the specified number of rows
+  // nx: global number of rows
+  static RCP<Matrix> Build1DPoisson(GO nx, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {  // global_size_t
+    ParameterList matrixList;
+    matrixList.set("nx", nx);
+    matrixList.set("matrixType", "Laplace1D");
+    return BuildMatrix(matrixList, lib);
+  }
 
-      // Create a map containing a specified number of local elements per process.
-      static const RCP<const Map> BuildMap(LO numElementsPerProc) {
-        RCP<const Teuchos::Comm<int> > comm   = TestHelpers_kokkos::Parameters::getDefaultComm();
-        const global_size_t           INVALID = Teuchos::OrdinalTraits<global_size_t>::invalid();
+  // Create a 2D Poisson matrix with the specified number of rows
+  // nx: global number of rows
+  // ny: global number of rows
+  static RCP<Matrix> Build2DPoisson(GO nx, GO ny = -1, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {  // global_size_t
+    if (ny == -1) ny = nx;
 
-        return MapFactory::Build(TestHelpers_kokkos::Parameters::getLib(), INVALID, numElementsPerProc, 0, comm);
+    ParameterList matrixList;
+    matrixList.set("nx", nx);
+    matrixList.set("ny", ny);
+    matrixList.set("matrixType", "Laplace2D");
 
-      } // BuildMap()
+    return BuildMatrix(matrixList, lib);
+  }
 
-      // Create a matrix as specified by parameter list options
-      static RCP<Matrix> BuildMatrix(ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  // Create a tridiagonal matrix (stencil = [b,a,c]) with the specified number of rows
+  static RCP<Matrix> BuildTridiag(RCP<const Map> rowMap, SC a, SC b, SC c, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers_kokkos::Parameters::getLib();
+
+    RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
+
+    RCP<Matrix> mtx = MatrixFactory::Build(rowMap, 3);
+
+    ArrayView<const GO> MyGlobalElements = rowMap->getLocalElementList();
+    GO indexBase                         = rowMap->getIndexBase();
+
+    GO NumEntries;
+    LO nnz = 3;
+    std::vector<SC> Values(nnz);
+    std::vector<GO> Indices(nnz);
+
+    // access information from strided block map
+    std::vector<size_t> strInfo = std::vector<size_t>();
+    size_t blockSize            = 1;
+    LO blockId                  = -1;
+    GO offset                   = 0;
+
+    RCP<const StridedMap> strrowMap = rcp_dynamic_cast<const StridedMap>(rowMap);
+    if (strrowMap != Teuchos::null) {
+      strInfo   = strrowMap->getStridingData();
+      blockSize = strrowMap->getFixedBlockSize();
+      blockId   = strrowMap->getStridedBlockId();
+      offset    = strrowMap->getOffset();
+      TEUCHOS_TEST_FOR_EXCEPTION(blockId > -1 && strInfo[blockId] == 1, MueLu::Exceptions::RuntimeError,
+                                 "MueLu::TestHelpers_kokkos::BuildTridiag: strInfo block size must be > 1.");
+      // todo: write one more special case for a row being first and last bock row
+    } else {
+      // no striding information. emulate block matrix
+      blockId = 0;
+      strInfo.push_back(blockSize);  // default block size = 1
+    }
+
+    GO rrmax = (rowMap->getMaxAllGlobalIndex() - offset - indexBase) / blockSize;
+
+    // loop over all rows
+    LO numMyElements = rowMap->getLocalNumElements();
+    for (LO i = 0; i < numMyElements; i++) {
+      GO rr = (MyGlobalElements[i] - offset - indexBase) / blockSize + indexBase;  // node index
+
+      // distinguish 5 different cases
+
+      GO blockOffset = 0;
+      for (LO k = 0; k < blockId; k++)
+        blockOffset += Teuchos::as<GO>(strInfo[k]);
+
+      if (MyGlobalElements[i] == blockOffset + offset + indexBase) {
+        // very first row
+        Indices[0] = MyGlobalElements[i];
+        Values[0]  = a;
+        Indices[1] = MyGlobalElements[i] + 1;
+        Values[1]  = c;
+        NumEntries = 2;
+
+      } else if (MyGlobalElements[i] == rrmax * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
+        // very last row
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        NumEntries = 2;
+
+      } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
+        // last row in current node block
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = (rr + 1) * blockSize + blockOffset + offset + indexBase;
+        Values[2]  = c;
+        NumEntries = 3;
+
+      } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + offset + indexBase) {
+        // first row in current node block
+        Indices[0] = (rr - 1) * blockSize + blockOffset + strInfo[blockId] - 1 + offset + indexBase;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = MyGlobalElements[i] + 1;
+        Values[2]  = c;
+        NumEntries = 3;
+
+      } else {
+        // usual row entries in block rows
+        Indices[0] = MyGlobalElements[i] - 1;
+        Values[0]  = b;
+        Indices[1] = MyGlobalElements[i];
+        Values[1]  = a;
+        Indices[2] = MyGlobalElements[i] + 1;
+        Values[2]  = c;
+        NumEntries = 3;
+      }
 
-        if (lib == Xpetra::NotSpecified)
-          lib = TestHelpers_kokkos::Parameters::getLib();
+      // debug output
+      /*std::cout << "--------> Proc " << comm->getRank() << " lrow " << i << " grow " << MyGlobalElements[i] << " node " << rr << " values = " ;
+      for (size_t k = 0; k < NumEntries; k++) {
+        std::cout << " " << Indices[k] << "->" << Values[k] << "   ";
+      }
+      std::cout << std::endl;*/
 
-        GO nx, ny, nz;
-        nx = ny = nz = 5;
-        nx = matrixList.get("nx", nx);
-        ny = matrixList.get("ny", ny);
-        nz = matrixList.get("nz", nz);
+      // put the off-diagonal entries
+      // Xpetra wants ArrayViews (sigh)
+      ArrayView<SC> av(&Values[0], NumEntries);
+      ArrayView<GO> iv(&Indices[0], NumEntries);
+      mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
+    }
 
-        std::string matrixType = matrixList.get("matrixType", "Laplace1D");
-        GO numGlobalElements; //global_size_t
-        if (matrixType == "Laplace1D")
-          numGlobalElements = nx;
-        else if (matrixType == "Laplace2D" || matrixType == "Star2D")
-          numGlobalElements = nx*ny;
-        else if (matrixType == "Laplace3D")
-          numGlobalElements = nx*ny*nz;
-        else {
-          std::string msg = matrixType + " is unsupported (in unit testing)";
-          throw(MueLu::Exceptions::RuntimeError(msg));
-        }
+    mtx->fillComplete();
 
-        RCP<const Map> map = MapFactory::Build(lib, numGlobalElements, 0, comm);
-        RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-          Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(matrixType, map, matrixList);
-        RCP<Matrix> Op = Pr->BuildMatrix();
+    return mtx;
+  }
 
-        return Op;
+  static RCP<RealValuedMultiVector>
+  BuildGeoCoordinates(const int numDimensions, const Array<GO> gNodesPerDir,
+                      Array<LO>& lNodesPerDir, Array<GO>& meshData,
+                      const std::string meshLayout = "Global Lexicographic") {
+    // Get MPI infos
+    Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+    LO numRanks                         = comm->getSize();
+    LO myRank                           = comm->getRank();
+
+    meshData.resize(10 * numRanks);
+
+    ////////////////////////////////////
+    //                                //
+    //   Step 1: Compute map layout   //
+    //                                //
+    ////////////////////////////////////
+    if (numRanks == 1) {
+      if (meshLayout == "Local Lexicographic") {
+        meshData[0] = 0;  // Local Proc
+        meshData[1] = 0;  // Local Proc?
+        meshData[2] = 0;  // Parent Proc
       }
-
-      // Create a 1D Poisson matrix with the specified number of rows
-      // nx: global number of rows
-      static RCP<Matrix> Build1DPoisson(GO nx, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) { //global_size_t
-        ParameterList matrixList;
-        matrixList.set("nx",            nx);
-        matrixList.set("matrixType",    "Laplace1D");
-        return BuildMatrix(matrixList,lib);
+      for (int dim = 0; dim < 3; ++dim) {
+        lNodesPerDir[dim] = gNodesPerDir[dim];
+        if (meshLayout == "Local Lexicographic") {
+          meshData[3 + 2 * dim]     = 0;                      // Lowest  index in direction dim
+          meshData[3 + 2 * dim + 1] = gNodesPerDir[dim] - 1;  // Highest index in direction dim
+        }
       }
-
-      // Create a 2D Poisson matrix with the specified number of rows
-      // nx: global number of rows
-      // ny: global number of rows
-      static RCP<Matrix> Build2DPoisson(GO nx, GO ny = -1, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) { //global_size_t
-        if (ny == -1) ny = nx;
-
-        ParameterList matrixList;
-        matrixList.set("nx",            nx);
-        matrixList.set("ny",            ny);
-        matrixList.set("matrixType",    "Laplace2D");
-
-        return BuildMatrix(matrixList,lib);
+      if (meshLayout == "Local Lexicographic") {
+        meshData[9] = 0;
+      }
+    } else if (numRanks == 4) {
+      if (meshLayout == "Local Lexicographic") {
+        // First fill the rank related info in meshData
+        for (int rank = 0; rank < numRanks; ++rank) {
+          meshData[10 * rank + 0] = rank;  // Local Proc
+          meshData[10 * rank + 1] = rank;  // Local Proc?
+          meshData[10 * rank + 2] = 0;     // Parent Proc
+        }
       }
 
-      // Create a tridiagonal matrix (stencil = [b,a,c]) with the specified number of rows
-      static RCP<Matrix> BuildTridiag(RCP<const Map> rowMap, SC a, SC b, SC c, Xpetra::UnderlyingLib lib = Xpetra::NotSpecified) {
-        if (lib == Xpetra::NotSpecified)
-          lib = TestHelpers_kokkos::Parameters::getLib();
-
-        RCP<const Teuchos::Comm<int> > comm = Parameters::getDefaultComm();
-
-        RCP<Matrix> mtx = MatrixFactory::Build(rowMap, 3);
-
-        ArrayView<const GO> MyGlobalElements = rowMap->getLocalElementList();
-        GO indexBase = rowMap->getIndexBase();
-
-        GO NumEntries;
-        LO nnz = 3;
-        std::vector<SC> Values(nnz);
-        std::vector<GO> Indices(nnz);
-
-        // access information from strided block map
-        std::vector<size_t> strInfo = std::vector<size_t>();
-        size_t blockSize = 1;
-        LO blockId = -1;
-        GO offset = 0;
-
-        RCP<const StridedMap> strrowMap = rcp_dynamic_cast<const StridedMap>(rowMap);
-        if (strrowMap != Teuchos::null) {
-          strInfo   = strrowMap->getStridingData();
-          blockSize = strrowMap->getFixedBlockSize();
-          blockId   = strrowMap->getStridedBlockId();
-          offset    = strrowMap->getOffset();
-          TEUCHOS_TEST_FOR_EXCEPTION(blockId > -1 && strInfo[blockId]==1, MueLu::Exceptions::RuntimeError,
-                                     "MueLu::TestHelpers_kokkos::BuildTridiag: strInfo block size must be > 1.");
-          // todo: write one more special case for a row being first and last bock row
+      if (numDimensions == 1) {
+        LO numNodesPerRank = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
+        if (myRank == numRanks - 1) {
+          lNodesPerDir[0] = gNodesPerDir[0] - myRank * numNodesPerRank;
         } else {
-          // no striding information. emulate block matrix
-          blockId = 0;
-          strInfo.push_back(blockSize); // default block size = 1
+          lNodesPerDir[0] = numNodesPerRank;
         }
-
-        GO rrmax = (rowMap->getMaxAllGlobalIndex()-offset-indexBase) / blockSize;
-
-        // loop over all rows
-        LO numMyElements = rowMap->getLocalNumElements();
-        for (LO i = 0; i < numMyElements; i++) {
-          GO rr = (MyGlobalElements[i]-offset-indexBase) / blockSize + indexBase;  // node index
-
-          // distinguish 5 different cases
-
-          GO blockOffset = 0;
-          for (LO k = 0; k < blockId; k++)
-            blockOffset += Teuchos::as<GO>(strInfo[k]);
-
-          if (MyGlobalElements[i] == blockOffset + offset + indexBase) {
-            // very first row
-            Indices[0] = MyGlobalElements[i];
-            Values [0] = a;
-            Indices[1] = MyGlobalElements[i] + 1;
-            Values [1] = c;
-            NumEntries = 2;
-
-          } else if (MyGlobalElements[i] == rrmax * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
-            // very last row
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            NumEntries = 2;
-
-          } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + Teuchos::as<GO>(strInfo[blockId]) - 1 + offset + indexBase) {
-            // last row in current node block
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = (rr+1)*blockSize + blockOffset + offset + indexBase;
-            Values [2] = c;
-            NumEntries = 3;
-
-          } else if (MyGlobalElements[i] == rr * Teuchos::as<GO>(blockSize) + blockOffset + offset + indexBase) {
-            // first row in current node block
-            Indices[0] = (rr-1)*blockSize + blockOffset + strInfo[blockId] - 1 + offset + indexBase;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = MyGlobalElements[i] + 1;
-            Values [2] = c;
-            NumEntries = 3;
-
-          } else {
-            // usual row entries in block rows
-            Indices[0] = MyGlobalElements[i] - 1;
-            Values [0] = b;
-            Indices[1] = MyGlobalElements[i];
-            Values [1] = a;
-            Indices[2] = MyGlobalElements[i] + 1;
-            Values [2] = c;
-            NumEntries = 3;
-          }
-
-          // debug output
-          /*std::cout << "--------> Proc " << comm->getRank() << " lrow " << i << " grow " << MyGlobalElements[i] << " node " << rr << " values = " ;
-          for (size_t k = 0; k < NumEntries; k++) {
-            std::cout << " " << Indices[k] << "->" << Values[k] << "   ";
+        lNodesPerDir[1] = gNodesPerDir[1];
+
+        if (meshLayout == "Local Lexicographic") {
+          for (int rank = 0; rank < numRanks; ++rank) {
+            meshData[10 * rank + 3] = rank * numNodesPerRank;
+            meshData[10 * rank + 4] =
+                (rank == numRanks - 1) ? (gNodesPerDir[0] - 1) : ((rank + 1) * numNodesPerRank - 1);
+            meshData[10 * rank + 5] = 0;
+            meshData[10 * rank + 6] = 0;
           }
-          std::cout << std::endl;*/
-
-          // put the off-diagonal entries
-          // Xpetra wants ArrayViews (sigh)
-          ArrayView<SC> av(&Values [0], NumEntries);
-          ArrayView<GO> iv(&Indices[0], NumEntries);
-          mtx->insertGlobalValues(MyGlobalElements[i], iv, av);
         }
-
-        mtx->fillComplete();
-
-        return mtx;
-      }
-
-      static RCP<RealValuedMultiVector>
-      BuildGeoCoordinates(const int numDimensions, const Array<GO> gNodesPerDir,
-                          Array<LO>& lNodesPerDir, Array<GO>& meshData,
-                          const std::string meshLayout = "Global Lexicographic") {
-
-        // Get MPI infos
-        Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-        LO numRanks = comm->getSize();
-        LO myRank   = comm->getRank();
-
-        meshData.resize(10*numRanks);
-
-        ////////////////////////////////////
-        //                                //
-        //   Step 1: Compute map layout   //
-        //                                //
-        ////////////////////////////////////
-        if(numRanks == 1) {
-          if(meshLayout == "Local Lexicographic") {
-            meshData[0] = 0; // Local Proc
-            meshData[1] = 0; // Local Proc?
-            meshData[2] = 0; // Parent Proc
-          }
-          for(int dim = 0; dim < 3; ++dim) {
-            lNodesPerDir[dim] = gNodesPerDir[dim];
-            if(meshLayout == "Local Lexicographic") {
-              meshData[3 + 2*dim]    = 0;                        // Lowest  index in direction dim
-              meshData[3 + 2*dim +1] = gNodesPerDir[dim] - 1;    // Highest index in direction dim
-            }
-          }
-          if(meshLayout == "Local Lexicographic") {
-            meshData[9] = 0;
-          }
-        } else if(numRanks == 4) {
-          if(meshLayout == "Local Lexicographic") {
-            // First fill the rank related info in meshData
-            for(int rank = 0; rank < numRanks; ++rank) {
-              meshData[10*rank + 0] = rank; // Local Proc
-              meshData[10*rank + 1] = rank; // Local Proc?
-              meshData[10*rank + 2] = 0;        // Parent Proc
-            }
-          }
-
-          if(numDimensions == 1) {
-            LO numNodesPerRank = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
-            if(myRank == numRanks - 1) {
-              lNodesPerDir[0] = gNodesPerDir[0] - myRank*numNodesPerRank;
-            } else {
-              lNodesPerDir[0] = numNodesPerRank;
-            }
-            lNodesPerDir[1] = gNodesPerDir[1];
-
-            if(meshLayout == "Local Lexicographic") {
-              for(int rank = 0; rank < numRanks; ++rank) {
-                meshData[10*rank + 3] = rank*numNodesPerRank;
-                meshData[10*rank + 4] =
-                  (rank == numRanks - 1) ? (gNodesPerDir[0] - 1) : ((rank + 1)*numNodesPerRank - 1);
-                meshData[10*rank + 5] = 0;
-                meshData[10*rank + 6] = 0;
+      } else {
+        if (myRank == 0 || myRank == 2) {
+          lNodesPerDir[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        } else {
+          lNodesPerDir[0] = std::floor(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        }
+        if (myRank == 0 || myRank == 1) {
+          lNodesPerDir[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        } else {
+          lNodesPerDir[1] = std::floor(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        }
+        if (meshLayout == "Local Lexicographic") {
+          for (int j = 0; j < 2; ++j) {
+            for (int i = 0; i < 2; ++i) {
+              int rank = 2 * j + i;
+              if (i == 0) {
+                meshData[10 * rank + 3] = 0;
+                meshData[10 * rank + 4] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2) - 1;
+              } else if (i == 1) {
+                meshData[10 * rank + 3] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+                meshData[10 * rank + 4] = gNodesPerDir[0] - 1;
               }
-            }
-          } else {
-            if(myRank == 0 || myRank == 2) {
-              lNodesPerDir[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            } else {
-              lNodesPerDir[0] = std::floor(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            }
-            if(myRank == 0 || myRank == 1) {
-              lNodesPerDir[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            } else {
-              lNodesPerDir[1] = std::floor(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            }
-            if(meshLayout == "Local Lexicographic") {
-              for(int j = 0; j < 2; ++j) {
-                for(int i = 0; i < 2; ++i) {
-                  int rank = 2*j + i;
-                  if(i == 0) {
-                    meshData[10*rank + 3] = 0;
-                    meshData[10*rank + 4] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2) - 1;
-                  } else if(i == 1) {
-                    meshData[10*rank + 3] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-                    meshData[10*rank + 4] = gNodesPerDir[0] - 1;
-                  }
-                  if(j == 0) {
-                    meshData[10*rank + 5] = 0;
-                    meshData[10*rank + 6] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) - 1;
-                  } else if(j == 1) {
-                    meshData[10*rank + 5] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-                    meshData[10*rank + 6] = gNodesPerDir[1] - 1;
-                  }
-                }
+              if (j == 0) {
+                meshData[10 * rank + 5] = 0;
+                meshData[10 * rank + 6] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) - 1;
+              } else if (j == 1) {
+                meshData[10 * rank + 5] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+                meshData[10 * rank + 6] = gNodesPerDir[1] - 1;
               }
             }
           }
-          lNodesPerDir[2] = gNodesPerDir[2];
-          if(meshLayout == "Local Lexicographic") {
-            for(int rank = 0; rank < numRanks; ++rank) {
-              meshData[10*rank + 7] = 0;
-              meshData[10*rank + 8] = gNodesPerDir[2] - 1;
-            }
-            meshData[9] = 0;
-            for(int rank = 1; rank < numRanks; ++rank) {
-              meshData[10*rank + 9] = meshData[10*(rank - 1) + 9]
-                + (meshData[10*(rank - 1) + 4] - meshData[10*(rank - 1) + 3] + 1)
-                *(meshData[10*(rank - 1) + 6] - meshData[10*(rank - 1) + 5] + 1)
-                *(meshData[10*(rank - 1) + 8] - meshData[10*(rank - 1) + 7] + 1);
-            }
-          }
         }
-
-        GO gNumNodes = 1;
-        LO lNumNodes = 1;
-        for(int dim = 0; dim < 3; ++dim) {
-          gNumNodes = gNumNodes*gNodesPerDir[dim];
-          lNumNodes = lNumNodes*lNodesPerDir[dim];
+      }
+      lNodesPerDir[2] = gNodesPerDir[2];
+      if (meshLayout == "Local Lexicographic") {
+        for (int rank = 0; rank < numRanks; ++rank) {
+          meshData[10 * rank + 7] = 0;
+          meshData[10 * rank + 8] = gNodesPerDir[2] - 1;
         }
-
-        GO myGIDOffset = 0;
-        Array<GO> myLowestTuple(3);
-        if(numDimensions == 1) {
-          myGIDOffset = myRank*std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
-          myLowestTuple[0] = myGIDOffset;
-        } else {
-          if(myRank == 1) {
-            myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = myLowestTuple[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = myLowestTuple[0]*std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2)
-                *gNodesPerDir[2];
-            }
-          } else if(myRank == 2) {
-            myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2)*gNodesPerDir[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = myLowestTuple[1]*gNodesPerDir[0]*gNodesPerDir[2];
-            }
-          } else if(myRank == 3) {
-            myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
-            myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
-            if(meshLayout == "Global Lexicographic") {
-              myGIDOffset = myLowestTuple[1]*gNodesPerDir[0] + myLowestTuple[0];
-            } else if(meshLayout == "Local Lexicographic") {
-              myGIDOffset = (myLowestTuple[0]*(gNodesPerDir[1] - myLowestTuple[1])
-                             + myLowestTuple[1]*gNodesPerDir[0])*gNodesPerDir[2];
-            }
-          }
+        meshData[9] = 0;
+        for (int rank = 1; rank < numRanks; ++rank) {
+          meshData[10 * rank + 9] = meshData[10 * (rank - 1) + 9] + (meshData[10 * (rank - 1) + 4] - meshData[10 * (rank - 1) + 3] + 1) * (meshData[10 * (rank - 1) + 6] - meshData[10 * (rank - 1) + 5] + 1) * (meshData[10 * (rank - 1) + 8] - meshData[10 * (rank - 1) + 7] + 1);
         }
-
-        ////////////////////////////////////
-        //                                //
-        //    Step 2: Compute map GIDs    //
-        //                                //
-        ////////////////////////////////////
-        Array<GO> myGIDs(lNumNodes);
-        if(meshLayout == "Global Lexicographic") {
-          for(LO k = 0; k < lNodesPerDir[2]; ++k) {
-            for(LO j = 0; j < lNodesPerDir[1]; ++j) {
-              for(LO i = 0; i < lNodesPerDir[0]; ++i) {
-                myGIDs[k*lNodesPerDir[1]*lNodesPerDir[0] + j*lNodesPerDir[0] + i] =
-                  myGIDOffset + k*gNodesPerDir[1]*gNodesPerDir[0] + j*gNodesPerDir[0] + i;
-              }
-            }
-          }
-        } else if(meshLayout == "Local Lexicographic") {
-          for(LO nodeIdx = 0; nodeIdx < lNumNodes; ++nodeIdx) {
-            myGIDs[nodeIdx] = myGIDOffset + nodeIdx;
-          }
+      }
+    }
+
+    GO gNumNodes = 1;
+    LO lNumNodes = 1;
+    for (int dim = 0; dim < 3; ++dim) {
+      gNumNodes = gNumNodes * gNodesPerDir[dim];
+      lNumNodes = lNumNodes * lNodesPerDir[dim];
+    }
+
+    GO myGIDOffset = 0;
+    Array<GO> myLowestTuple(3);
+    if (numDimensions == 1) {
+      myGIDOffset      = myRank * std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / numRanks);
+      myLowestTuple[0] = myGIDOffset;
+    } else {
+      if (myRank == 1) {
+        myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = myLowestTuple[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = myLowestTuple[0] * std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) * gNodesPerDir[2];
         }
-
-        RCP<const Map> coordMap = MapFactory::Build(lib, gNumNodes, myGIDs(), 0, comm);
-
-
-        ///////////////////////////////////////
-        //                                   //
-        //    Step 3: Compute coordinates    //
-        //                                   //
-        ///////////////////////////////////////
-        RCP<RealValuedMultiVector> Coordinates = RealValuedMultiVectorFactory::Build(coordMap,
-                                                                                     numDimensions);
-        Array<ArrayRCP<real_type> > myCoords(numDimensions);
-        for(int dim = 0; dim < numDimensions; ++dim) {
-          myCoords[dim] = Coordinates->getDataNonConst(dim);
+      } else if (myRank == 2) {
+        myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2) * gNodesPerDir[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = myLowestTuple[1] * gNodesPerDir[0] * gNodesPerDir[2];
         }
-
-        LO nodeIdx = 0;
-        Array<LO> ijk(3);
-        for(ijk[2] = 0; ijk[2] < lNodesPerDir[2]; ++ijk[2]) {
-          for(ijk[1] = 0; ijk[1] < lNodesPerDir[1]; ++ijk[1]) {
-            for(ijk[0] = 0; ijk[0] < lNodesPerDir[0]; ++ijk[0]) {
-              nodeIdx = ijk[2]*lNodesPerDir[1]*lNodesPerDir[0] + ijk[1]*lNodesPerDir[0] + ijk[0];
-              for(int dim = 0; dim < numDimensions; ++dim) {
-                 if(gNodesPerDir[dim] == 1) {
-                  myCoords[dim][nodeIdx] = 0.0;
-                } else {
-                  myCoords[dim][nodeIdx] =
-                    Teuchos::as<real_type>(myLowestTuple[dim] + ijk[dim]) / (gNodesPerDir[dim] - 1);
-                }
-              }
+      } else if (myRank == 3) {
+        myLowestTuple[0] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[0]) / 2);
+        myLowestTuple[1] = std::ceil(Teuchos::as<real_type>(gNodesPerDir[1]) / 2);
+        if (meshLayout == "Global Lexicographic") {
+          myGIDOffset = myLowestTuple[1] * gNodesPerDir[0] + myLowestTuple[0];
+        } else if (meshLayout == "Local Lexicographic") {
+          myGIDOffset = (myLowestTuple[0] * (gNodesPerDir[1] - myLowestTuple[1]) + myLowestTuple[1] * gNodesPerDir[0]) * gNodesPerDir[2];
+        }
+      }
+    }
+
+    ////////////////////////////////////
+    //                                //
+    //    Step 2: Compute map GIDs    //
+    //                                //
+    ////////////////////////////////////
+    Array<GO> myGIDs(lNumNodes);
+    if (meshLayout == "Global Lexicographic") {
+      for (LO k = 0; k < lNodesPerDir[2]; ++k) {
+        for (LO j = 0; j < lNodesPerDir[1]; ++j) {
+          for (LO i = 0; i < lNodesPerDir[0]; ++i) {
+            myGIDs[k * lNodesPerDir[1] * lNodesPerDir[0] + j * lNodesPerDir[0] + i] =
+                myGIDOffset + k * gNodesPerDir[1] * gNodesPerDir[0] + j * gNodesPerDir[0] + i;
+          }
+        }
+      }
+    } else if (meshLayout == "Local Lexicographic") {
+      for (LO nodeIdx = 0; nodeIdx < lNumNodes; ++nodeIdx) {
+        myGIDs[nodeIdx] = myGIDOffset + nodeIdx;
+      }
+    }
+
+    RCP<const Map> coordMap = MapFactory::Build(lib, gNumNodes, myGIDs(), 0, comm);
+
+    ///////////////////////////////////////
+    //                                   //
+    //    Step 3: Compute coordinates    //
+    //                                   //
+    ///////////////////////////////////////
+    RCP<RealValuedMultiVector> Coordinates = RealValuedMultiVectorFactory::Build(coordMap,
+                                                                                 numDimensions);
+    Array<ArrayRCP<real_type> > myCoords(numDimensions);
+    for (int dim = 0; dim < numDimensions; ++dim) {
+      myCoords[dim] = Coordinates->getDataNonConst(dim);
+    }
+
+    LO nodeIdx = 0;
+    Array<LO> ijk(3);
+    for (ijk[2] = 0; ijk[2] < lNodesPerDir[2]; ++ijk[2]) {
+      for (ijk[1] = 0; ijk[1] < lNodesPerDir[1]; ++ijk[1]) {
+        for (ijk[0] = 0; ijk[0] < lNodesPerDir[0]; ++ijk[0]) {
+          nodeIdx = ijk[2] * lNodesPerDir[1] * lNodesPerDir[0] + ijk[1] * lNodesPerDir[0] + ijk[0];
+          for (int dim = 0; dim < numDimensions; ++dim) {
+            if (gNodesPerDir[dim] == 1) {
+              myCoords[dim][nodeIdx] = 0.0;
+            } else {
+              myCoords[dim][nodeIdx] =
+                  Teuchos::as<real_type>(myLowestTuple[dim] + ijk[dim]) / (gNodesPerDir[dim] - 1);
             }
           }
         }
-
-        return Coordinates;
       }
+    }
 
-      // Needed to initialize correctly a level used for testing SingleLevel factory Build() methods.
-      // This method initializes LevelID and linked list of level
-      static void createSingleLevelHierarchy(Level& currentLevel) {
-        RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-        currentLevel.SetFactoryManager(factoryHandler);
+    return Coordinates;
+  }
 
-        currentLevel.SetLevelID(0);
+  // Needed to initialize correctly a level used for testing SingleLevel factory Build() methods.
+  // This method initializes LevelID and linked list of level
+  static void createSingleLevelHierarchy(Level& currentLevel) {
+    RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+    currentLevel.SetFactoryManager(factoryHandler);
+
+    currentLevel.SetLevelID(0);
 #ifdef HAVE_MUELU_TIMER_SYNCHRONIZATION
-        currentLevel.SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
+    currentLevel.SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
 #endif
-      }
+  }
 
-      // Needed to initialize correctly levels used for testing TwoLevel factory Build() methods.
-      // This method initializes LevelID and linked list of level
-      static void createTwoLevelHierarchy(Level& fineLevel, Level& coarseLevel) {
-        RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
-        fineLevel  .SetFactoryManager(factoryHandler);
-        coarseLevel.SetFactoryManager(factoryHandler);
+  // Needed to initialize correctly levels used for testing TwoLevel factory Build() methods.
+  // This method initializes LevelID and linked list of level
+  static void createTwoLevelHierarchy(Level& fineLevel, Level& coarseLevel) {
+    RCP<MueLu::FactoryManagerBase> factoryHandler = rcp(new FactoryManager());
+    fineLevel.SetFactoryManager(factoryHandler);
+    coarseLevel.SetFactoryManager(factoryHandler);
 
-        coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
+    coarseLevel.SetPreviousLevel(rcpFromRef(fineLevel));
 
-        fineLevel  .SetLevelID(0);
-        coarseLevel.SetLevelID(1);
+    fineLevel.SetLevelID(0);
+    coarseLevel.SetLevelID(1);
 #ifdef HAVE_MUELU_TIMER_SYNCHRONIZATION
-        fineLevel  .SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
-        coarseLevel.SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
+    fineLevel.SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
+    coarseLevel.SetComm(TestHelpers_kokkos::Parameters::getDefaultComm());
 #endif
-      }
+  }
 
 #if 0
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_IFPACK)
@@ -578,133 +566,134 @@ namespace MueLuTests {
       }
 #endif
 #endif
-    }; // class TestFactory
-
-    // Helper class which has some Tpetra specific code inside
-    // We put this into an extra helper class as we need partial specializations and
-    // do not want to introduce partial specializations for the full TestFactory class
-    //
-    // The BuildBlockMatrix is only available with Teptra. However, if both Epetra
-    // and Tpetra are enabled it may be that Tpetra is not instantiated on either
-    // GO=int/long long and/or Node=Serial/OpenMP. We need partial specializations
-    // with an empty BuildBlockMatrix routine for all instantiations Teptra is not
-    // enabled for, but are existing in Xpetra due to Epetra enabled.
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-    class TpetraTestFactory {
-#include "MueLu_UseShortNames.hpp"
-    public:
-
-      // Create a matrix as specified by parameter list options
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) {
-        RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-        RCP<Matrix> Op;
-
-         if (lib == Xpetra::NotSpecified)
-           lib = TestHelpers_kokkos::Parameters::getLib();
-
-         // This only works for Tpetra
-         if (lib!=Xpetra::UseTpetra) return Op;
-
-         // Thanks for the code, Travis!
-
-         // Make the graph
-         RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > FirstMatrix = TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList,lib);
-         RCP<const Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > Graph = FirstMatrix->getCrsGraph();
-
-         int blocksize = 3;
-         RCP<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TGraph = rcp_dynamic_cast<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> >(Graph);
-         RCP<const Tpetra::CrsGraph<LocalOrdinal,GlobalOrdinal,Node> > TTGraph = TGraph->getTpetra_CrsGraph();
-
-         RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bcrsmatrix = rcp(new Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> (*TTGraph, blocksize));
+};  // class TestFactory
 
-         const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& meshRowMap = *bcrsmatrix->getRowMap();
-         const Scalar zero   = Teuchos::ScalarTraits<Scalar>::zero();
-         const Scalar one   = Teuchos::ScalarTraits<Scalar>::one();
-         const Scalar two   = one+one;
-         const Scalar three = two+one;
-
-         Teuchos::Array<Scalar> basematrix(blocksize*blocksize, zero);
-         basematrix[0] = two;
-         basematrix[2] = three;
-         basematrix[3] = three;
-         basematrix[4] = two;
-         basematrix[7] = three;
-         basematrix[8] = two;
-         Teuchos::Array<LocalOrdinal> lclColInds(1);
-         for (LocalOrdinal lclRowInd = meshRowMap.getMinLocalIndex (); lclRowInd <= meshRowMap.getMaxLocalIndex(); ++lclRowInd) {
-           lclColInds[0] = lclRowInd;
-           bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &basematrix[0], 1);
-         }
-
-         RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(bcrsmatrix));
-         Op = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(temp));
-
-         return Op;
-      } // BuildBlockMatrix()
-
-    private:
-      TpetraTestFactory() {} // static class
-
-    }; // class TpetraTestFactory
-
-    // TAW: 3/14/2016: If both Epetra and Tpetra are enabled we need partial specializations
-    //                 on GO=int/long long as well as NO=EpetraNode to disable BuildBlockMatrix
+// Helper class which has some Tpetra specific code inside
+// We put this into an extra helper class as we need partial specializations and
+// do not want to introduce partial specializations for the full TestFactory class
+//
+// The BuildBlockMatrix is only available with Teptra. However, if both Epetra
+// and Tpetra are enabled it may be that Tpetra is not instantiated on either
+// GO=int/long long and/or Node=Serial/OpenMP. We need partial specializations
+// with an empty BuildBlockMatrix routine for all instantiations Teptra is not
+// enabled for, but are existing in Xpetra due to Epetra enabled.
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class TpetraTestFactory {
+#include "MueLu_UseShortNames.hpp"
+ public:
+  // Create a matrix as specified by parameter list options
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) {
+    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+    RCP<Matrix> Op;
+
+    if (lib == Xpetra::NotSpecified)
+      lib = TestHelpers_kokkos::Parameters::getLib();
+
+    // This only works for Tpetra
+    if (lib != Xpetra::UseTpetra) return Op;
+
+    // Thanks for the code, Travis!
+
+    // Make the graph
+    RCP<Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > FirstMatrix = TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::BuildMatrix(matrixList, lib);
+    RCP<const Xpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > Graph       = FirstMatrix->getCrsGraph();
+
+    int blocksize                                                                = 3;
+    RCP<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TGraph = rcp_dynamic_cast<const Xpetra::TpetraCrsGraph<LocalOrdinal, GlobalOrdinal, Node> >(Graph);
+    RCP<const Tpetra::CrsGraph<LocalOrdinal, GlobalOrdinal, Node> > TTGraph      = TGraph->getTpetra_CrsGraph();
+
+    RCP<Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > bcrsmatrix = rcp(new Tpetra::BlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(*TTGraph, blocksize));
+
+    const Tpetra::Map<LocalOrdinal, GlobalOrdinal, Node>& meshRowMap = *bcrsmatrix->getRowMap();
+    const Scalar zero                                                = Teuchos::ScalarTraits<Scalar>::zero();
+    const Scalar one                                                 = Teuchos::ScalarTraits<Scalar>::one();
+    const Scalar two                                                 = one + one;
+    const Scalar three                                               = two + one;
+
+    Teuchos::Array<Scalar> basematrix(blocksize * blocksize, zero);
+    basematrix[0] = two;
+    basematrix[2] = three;
+    basematrix[3] = three;
+    basematrix[4] = two;
+    basematrix[7] = three;
+    basematrix[8] = two;
+    Teuchos::Array<LocalOrdinal> lclColInds(1);
+    for (LocalOrdinal lclRowInd = meshRowMap.getMinLocalIndex(); lclRowInd <= meshRowMap.getMaxLocalIndex(); ++lclRowInd) {
+      lclColInds[0] = lclRowInd;
+      bcrsmatrix->replaceLocalValues(lclRowInd, lclColInds.getRawPtr(), &basematrix[0], 1);
+    }
+
+    RCP<Xpetra::CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node> > temp = rcp(new Xpetra::TpetraBlockCrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>(bcrsmatrix));
+    Op                                                                      = rcp(new Xpetra::CrsMatrixWrap<Scalar, LocalOrdinal, GlobalOrdinal, Node>(temp));
+
+    return Op;
+  }  // BuildBlockMatrix()
+
+ private:
+  TpetraTestFactory() {}  // static class
+
+};  // class TpetraTestFactory
+
+// TAW: 3/14/2016: If both Epetra and Tpetra are enabled we need partial specializations
+//                 on GO=int/long long as well as NO=EpetraNode to disable BuildBlockMatrix
 #ifdef HAVE_MUELU_EPETRA
-    // partial specializations (GO=int not enabled with Tpetra)
+// partial specializations (GO=int not enabled with Tpetra)
 #if !defined(HAVE_TPETRA_INST_INT_INT)
-    template <class Scalar, class LocalOrdinal, class Node>
-    class TpetraTestFactory<Scalar, LocalOrdinal, int, Node> {
-      typedef int GlobalOrdinal;
+template <class Scalar, class LocalOrdinal, class Node>
+class TpetraTestFactory<Scalar, LocalOrdinal, int, Node> {
+  typedef int GlobalOrdinal;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
 #endif
 
-    // partial specializations (GO=long long not enabled with Tpetra)
+// partial specializations (GO=long long not enabled with Tpetra)
 #if !defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-    template <class Scalar, class LocalOrdinal, class Node>
-    class TpetraTestFactory<Scalar, LocalOrdinal, long long, Node> {
-      typedef long long GlobalOrdinal;
+template <class Scalar, class LocalOrdinal, class Node>
+class TpetraTestFactory<Scalar, LocalOrdinal, long long, Node> {
+  typedef long long GlobalOrdinal;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
+
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
 #endif
 
-    // partial specializations (NO=EpetraNode not enabled with Tpetra)
+// partial specializations (NO=EpetraNode not enabled with Tpetra)
 #if ((defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_OPENMP))) || \
-    (!defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_SERIAL))))
+     (!defined(EPETRA_HAVE_OMP) && !(defined(HAVE_TPETRA_INST_SERIAL))))
 
-    template <class Scalar, class LocalOrdinal, class GlobalOrdinal>
-    class TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Xpetra::EpetraNode> {
-      typedef Xpetra::EpetraNode Node;
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal>
+class TpetraTestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Xpetra::EpetraNode> {
+  typedef Xpetra::EpetraNode Node;
 #include "MueLu_UseShortNames.hpp"
-    public:
-      static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList &matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
-    private:
-      TpetraTestFactory() {} // static class
-    }; // class TpetraTestFactory
-#endif
-#endif // endif HAVE_MUELU_EPETRA
+ public:
+  static RCP<Matrix> BuildBlockMatrix(Teuchos::ParameterList& matrixList, Xpetra::UnderlyingLib lib) { return Teuchos::null; }
 
-    //! Return the list of files in the directory. Only files that are matching '*filter*' are returned.
-    ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string & filter);
+ private:
+  TpetraTestFactory() {}  // static class
+};                        // class TpetraTestFactory
+#endif
+#endif  // endif HAVE_MUELU_EPETRA
 
-  } // namespace TestHelpers_kokkos
+//! Return the list of files in the directory. Only files that are matching '*filter*' are returned.
+ArrayRCP<std::string> GetFileList(const std::string& dirPath, const std::string& filter);
 
-} // namespace MueLu
+}  // namespace TestHelpers_kokkos
 
+}  // namespace MueLuTests
 
 // Macro to skip a test when UnderlyingLib==Epetra or Tpetra
-#define MUELU_TEST_ONLY_FOR(UnderlyingLib) \
-  if (TestHelpers_kokkos::Parameters::getLib() != UnderlyingLib) { \
-    out << "Skipping test for " << ((TestHelpers_kokkos::Parameters::getLib()==Xpetra::UseEpetra) ? "Epetra" : "Tpetra") << std::endl; \
-    return; \
+#define MUELU_TEST_ONLY_FOR(UnderlyingLib)                                                                                               \
+  if (TestHelpers_kokkos::Parameters::getLib() != UnderlyingLib) {                                                                       \
+    out << "Skipping test for " << ((TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) ? "Epetra" : "Tpetra") << std::endl; \
+    return;                                                                                                                              \
   }
 
 // Macro to skip a test when Epetra is used with Ordinal != int
@@ -712,15 +701,14 @@ namespace MueLuTests {
   if (!(TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra && (Teuchos::OrdinalTraits<LocalOrdinal>::name() != string("int") || Teuchos::OrdinalTraits<GlobalOrdinal>::name() != string("int"))))
 
 // Macro to skip a test when Epetra is used with Scalar != double or Ordinal != int
-#define MUELU_TEST_EPETRA_ONLY_FOR_DOUBLE_AND_INT(Scalar, LocalOrdinal, GlobalOrdinal) \
+#define MUELU_TEST_EPETRA_ONLY_FOR_DOUBLE_AND_INT(Scalar, LocalOrdinal, GlobalOrdinal)                                               \
   if (!(TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra && Teuchos::ScalarTraits<Scalar>::name() != string("double"))) \
-    MUELU_TEST_EPETRA_ONLY_FOR_INT(LocalOrdinal, GlobalOrdinal)
-
+  MUELU_TEST_EPETRA_ONLY_FOR_INT(LocalOrdinal, GlobalOrdinal)
 
 //! Namespace for MueLu test classes
 namespace MueLuTests {
 
-  using namespace TestHelpers_kokkos;
+using namespace TestHelpers_kokkos;
 }
 
-#endif // ifndef MUELU_TEST_HELPERS_KOKKOS_H
+#endif  // ifndef MUELU_TEST_HELPERS_KOKKOS_H
diff --git a/packages/muelu/test/unit_tests_kokkos/MueLu_UnitTests_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/MueLu_UnitTests_kokkos.cpp
index 7ef58afd667e..6968ba5a81c4 100644
--- a/packages/muelu/test/unit_tests_kokkos/MueLu_UnitTests_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/MueLu_UnitTests_kokkos.cpp
@@ -44,7 +44,6 @@
 //
 // @HEADER
 
-
 #include <Teuchos_UnitTestRepository.hpp>
 #include <Teuchos_DefaultComm.hpp>
 #include <Teuchos_StandardCatchMacros.hpp>
@@ -70,7 +69,7 @@ int main(int argc, char* argv[]) {
 
   bool success = false;
   bool verbose = true;
-  int ierr = -1;
+  int ierr     = -1;
   try {
     // Note: the command line parameter --linAlgebra= is taken into account.
     // Xpetra parameters are added to the Teuchos::CommandLineProcessor of Teuchos::UnitTestRepository in MueLu_TestHelpers_kokkos.cpp
@@ -80,25 +79,26 @@ int main(int argc, char* argv[]) {
 
     int mypid = comm->getRank();
 
-    if (mypid  == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
-    for (int i = 0; i <comm->getSize(); i++) {
-      if (i == mypid ) {
+    if (mypid == 0) std::cout << "Host and Process Ids for tasks" << std::endl;
+    for (int i = 0; i < comm->getSize(); i++) {
+      if (i == mypid) {
         char buf[80];
         char hostname[80];
         gethostname(hostname, sizeof(hostname));
         int pid = getpid();
         sprintf(buf, "Host: %s\tMPI rank: %d,\tPID: %d\n\tattach %d\n\tcontinue\n",
                 hostname, mypid, pid, pid);
-        printf("%s\n",buf);
+        printf("%s\n", buf);
         fflush(stdout);
         sleep(1);
       }
     }
 
     if (mypid == 0) {
-      printf( "** Enter a character to continue > "); fflush(stdout);
+      printf("** Enter a character to continue > ");
+      fflush(stdout);
       char go = ' ';
-      scanf("%c",&go);
+      scanf("%c", &go);
     }
     comm->barrier();
 #endif
diff --git a/packages/muelu/test/unit_tests_kokkos/NotayAggregationFactory.cpp b/packages/muelu/test/unit_tests_kokkos/NotayAggregationFactory.cpp
index 81a5798d7efb..41e47668d738 100644
--- a/packages/muelu/test/unit_tests_kokkos/NotayAggregationFactory.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/NotayAggregationFactory.cpp
@@ -55,44 +55,43 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, InitialAggregation1D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    //    using TST                   = Teuchos::ScalarTraits<SC>;
-    //    using magnitude_type        = typename TST::magnitudeType;
-    //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    //    using real_type             = typename TST::coordinateType;
-    //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    int rank = comm->getRank();
-    int numproc = comm->getSize();
-    RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16*comm->getSize());
-    const int numRows = static_cast<int>(A->getLocalNumRows());
-    RCP<Aggregates> aggregates = rcp(new Aggregates(A->getMap()));
-    RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
-    std::vector<unsigned> aggStat(numRows, MueLu::READY);
-    LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
-
-    Array<LO> orderingVector(numRows);
-    for (LO i = 0; i < numRows; i++) {
-      orderingVector[i] = i;
-    }
-
-    Teuchos::ParameterList params;
-    params.set("aggregation: pairwise: tie threshold",1e-6);
-    NAF->BuildInitialAggregates(params, A, orderingVector(),
-                                Teuchos::ScalarTraits<SC>::magnitude(10.0),
-                                *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
-
-    Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-    std::cout << "p=" << rank << " | aggregate sizes="
-              << sizes.view(0, sizes.size()) << std::endl;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, InitialAggregation1D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  //    using TST                   = Teuchos::ScalarTraits<SC>;
+  //    using magnitude_type        = typename TST::magnitudeType;
+  //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  //    using real_type             = typename TST::coordinateType;
+  //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  int rank                            = comm->getRank();
+  int numproc                         = comm->getSize();
+  RCP<const Matrix> A                 = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(16 * comm->getSize());
+  const int numRows                   = static_cast<int>(A->getLocalNumRows());
+  RCP<Aggregates> aggregates          = rcp(new Aggregates(A->getMap()));
+  RCP<NotayAggregationFactory> NAF    = rcp(new NotayAggregationFactory());
+  std::vector<unsigned> aggStat(numRows, MueLu::READY);
+  LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
+
+  Array<LO> orderingVector(numRows);
+  for (LO i = 0; i < numRows; i++) {
+    orderingVector[i] = i;
+  }
+
+  Teuchos::ParameterList params;
+  params.set("aggregation: pairwise: tie threshold", 1e-6);
+  NAF->BuildInitialAggregates(params, A, orderingVector(),
+                              Teuchos::ScalarTraits<SC>::magnitude(10.0),
+                              *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+
+  Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
+  std::cout << "p=" << rank << " | aggregate sizes="
+            << sizes.view(0, sizes.size()) << std::endl;
 
 #if 0
     auto v2a = aggregates->GetVertex2AggId()->getData(0);
@@ -102,96 +101,98 @@ namespace MueLuTests {
     printf("\n");
 #endif
 
-    //    TEST_EQUALITY(numUnaggregatedNodes, 0);
-    if(numproc == 1) {
-      TEST_EQUALITY(aggregates->GetNumAggregates(),7);
-    }
-    else {
-      TEST_EQUALITY(aggregates->GetNumAggregates(),8);
-    }
-
-    // On proc 0 gets picked as a Dirichlet (so it does not get aggregated) and the last aggregate is a singleton
-    // All the other ranks wind up with strict pairs
-    int expected;
-    for(int i=0; i<(int)sizes.size(); i++) {
-      if(numproc == 1) expected = 2;
-      else expected = ((rank == 0 || rank == 3) && i == (int)sizes.size() -1) ? 1 : 2;
-      TEST_EQUALITY(sizes[i], expected);
-    }
-
-  } // InitialAggregation1D
-
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, InitialAggregation2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    //    using TST                   = Teuchos::ScalarTraits<SC>;
-    //    using magnitude_type        = typename TST::magnitudeType;
-    //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    //    using real_type             = typename TST::coordinateType;
-    //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    int rank = comm->getRank();
-    int numproc = comm->getSize();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    // Do a 2D Star2D Matrix with up/down as a "weak connection"
-    Teuchos::ParameterList mp;
-    mp.set("matrixType","Star2D");
-    const int nx = 3;
-    mp.set("nx",(GO)nx);
-    mp.set("ny",(GO)nx*comm->getSize());
-    mp.set("a",2.2);
-    mp.set("b",-0.1);  mp.set("c",-0.1);
-    mp.set("d",-1.0); mp.set("e",-1.0);
-    mp.set("z1",0.0);  mp.set("z2",0.0);  mp.set("z3",0.0);  mp.set("z4",0.0);
-    RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-    const int numRows = static_cast<int>(A->getLocalNumRows());
-
-    RCP<Aggregates> aggregates = rcp(new Aggregates(A->getMap()));
-    RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
-    std::vector<unsigned> aggStat(numRows, MueLu::READY);
-    LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
-
-    Array<LO> orderingVector(numRows);
-    for (LO i = 0; i < numRows; i++) {
-      orderingVector[i] = i;
-    }
-
-    Teuchos::ParameterList params;
-    params.set("aggregation: pairwise: tie threshold",1e-6);
-    NAF->BuildInitialAggregates(params, A, orderingVector(),
-                                Teuchos::ScalarTraits<SC>::magnitude(4.1),
-                                *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
-
-    Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-
-    TEST_EQUALITY(numUnaggregatedNodes, 0);
-
-    // For this problem, the four corners will be detected as "Dirichlet" and ignored, this will
-    // generate some number of singletons
-    Teuchos::Array<int> expected;
-    if(numproc == 1) {
-      expected = Teuchos::Array<int>({2, 1, 1, 1});
-    } else if(rank == 0 || rank == numproc - 1) {
-      expected = Teuchos::Array<int>({2, 2, 2, 1});
-    } else {
-      expected = Teuchos::Array<int>({2, 2, 2, 1, 1, 1});
-    }
-
-    TEST_EQUALITY(sizes.size(), expected.size());
-
-
-    for(int i = 0; i < static_cast<int>(sizes.size()); i++) {
-      TEST_EQUALITY(sizes[i], expected[i]);
-    }
+  //    TEST_EQUALITY(numUnaggregatedNodes, 0);
+  if (numproc == 1) {
+    TEST_EQUALITY(aggregates->GetNumAggregates(), 7);
+  } else {
+    TEST_EQUALITY(aggregates->GetNumAggregates(), 8);
+  }
+
+  // On proc 0 gets picked as a Dirichlet (so it does not get aggregated) and the last aggregate is a singleton
+  // All the other ranks wind up with strict pairs
+  int expected;
+  for (int i = 0; i < (int)sizes.size(); i++) {
+    if (numproc == 1)
+      expected = 2;
+    else
+      expected = ((rank == 0 || rank == 3) && i == (int)sizes.size() - 1) ? 1 : 2;
+    TEST_EQUALITY(sizes[i], expected);
+  }
+
+}  // InitialAggregation1D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, InitialAggregation2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  //    using TST                   = Teuchos::ScalarTraits<SC>;
+  //    using magnitude_type        = typename TST::magnitudeType;
+  //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  //    using real_type             = typename TST::coordinateType;
+  //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  int rank                            = comm->getRank();
+  int numproc                         = comm->getSize();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  // Do a 2D Star2D Matrix with up/down as a "weak connection"
+  Teuchos::ParameterList mp;
+  mp.set("matrixType", "Star2D");
+  const int nx = 3;
+  mp.set("nx", (GO)nx);
+  mp.set("ny", (GO)nx * comm->getSize());
+  mp.set("a", 2.2);
+  mp.set("b", -0.1);
+  mp.set("c", -0.1);
+  mp.set("d", -1.0);
+  mp.set("e", -1.0);
+  mp.set("z1", 0.0);
+  mp.set("z2", 0.0);
+  mp.set("z3", 0.0);
+  mp.set("z4", 0.0);
+  RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+  const int numRows   = static_cast<int>(A->getLocalNumRows());
+
+  RCP<Aggregates> aggregates       = rcp(new Aggregates(A->getMap()));
+  RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
+  std::vector<unsigned> aggStat(numRows, MueLu::READY);
+  LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
+
+  Array<LO> orderingVector(numRows);
+  for (LO i = 0; i < numRows; i++) {
+    orderingVector[i] = i;
+  }
+
+  Teuchos::ParameterList params;
+  params.set("aggregation: pairwise: tie threshold", 1e-6);
+  NAF->BuildInitialAggregates(params, A, orderingVector(),
+                              Teuchos::ScalarTraits<SC>::magnitude(4.1),
+                              *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+
+  Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
+
+  TEST_EQUALITY(numUnaggregatedNodes, 0);
+
+  // For this problem, the four corners will be detected as "Dirichlet" and ignored, this will
+  // generate some number of singletons
+  Teuchos::Array<int> expected;
+  if (numproc == 1) {
+    expected = Teuchos::Array<int>({2, 1, 1, 1});
+  } else if (rank == 0 || rank == numproc - 1) {
+    expected = Teuchos::Array<int>({2, 2, 2, 1});
+  } else {
+    expected = Teuchos::Array<int>({2, 2, 2, 1, 1, 1});
+  }
+
+  TEST_EQUALITY(sizes.size(), expected.size());
+
+  for (int i = 0; i < static_cast<int>(sizes.size()); i++) {
+    TEST_EQUALITY(sizes[i], expected[i]);
+  }
 
 #if 0
     auto v2a = aggregates->GetVertex2AggId()->getData(0);
@@ -201,265 +202,272 @@ namespace MueLuTests {
     printf("\n");
 #endif
 
-  } // InitialAggregation2D
+}  // InitialAggregation2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, IntermediateProlongator2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  //    using TST                   = Teuchos::ScalarTraits<SC>;
+  //    using magnitude_type        = typename TST::magnitudeType;
+  //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  //    using real_type             = typename TST::coordinateType;
+  //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  // Do a 2D Star2D Matrix with up/down as a "weak connection"
+  Teuchos::ParameterList mp;
+  mp.set("matrixType", "Star2D");
+  const int nx = 3;
+  mp.set("nx", (GO)nx);
+  mp.set("ny", (GO)nx * comm->getSize());
+  mp.set("a", 2.2);
+  mp.set("b", -0.1);
+  mp.set("c", -0.1);
+  mp.set("d", -1.0);
+  mp.set("e", -1.0);
+  mp.set("z1", 0.0);
+  mp.set("z2", 0.0);
+  mp.set("z3", 0.0);
+  mp.set("z4", 0.0);
+  RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+  const int numRows   = static_cast<int>(A->getLocalNumRows());
+
+  RCP<Aggregates> aggregates       = rcp(new Aggregates(A->getMap()));
+  RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
+  std::vector<unsigned> aggStat(numRows, MueLu::READY);
+  LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
+  typename Matrix::local_matrix_type intermediateP;
+
+  Array<LO> orderingVector(numRows);
+  for (LO i = 0; i < numRows; i++) {
+    orderingVector[i] = i;
+  }
+
+  Teuchos::ParameterList params;
+  params.set("aggregation: pairwise: tie threshold", 1e-6);
+  NAF->BuildInitialAggregates(params, A, orderingVector(),
+                              Teuchos::ScalarTraits<SC>::magnitude(4.1),
+                              *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+
+  out << "numDirichletNodes=" << numDirichletNodes << std::endl;
+
+  Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
+  auto v2a                    = aggregates->GetVertex2AggId()->getData(0);
+
+  TEST_EQUALITY(numUnaggregatedNodes, 0);
+
+  NAF->BuildIntermediateProlongator(A->getLocalNumRows(), numDirichletNodes,
+                                    aggregates->GetNumAggregates(),
+                                    v2a.view(0, A->getLocalNumRows()),
+                                    intermediateP);
+
+}  // IntermediateProlongator2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, CoarseLocalMatrix2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  // Do a 2D Star2D Matrix with up/down as a "weak connection"
+  Teuchos::ParameterList mp;
+  mp.set("matrixType", "Star2D");
+  const GO nx = 3;
+  mp.set("nx", nx);
+  mp.set("ny", nx * comm->getSize());
+  mp.set("a", 2.2);
+  mp.set("b", -0.1);
+  mp.set("c", -0.1);
+  mp.set("d", -1.0);
+  mp.set("e", -1.0);
+  mp.set("z1", 0.0);
+  mp.set("z2", 0.0);
+  mp.set("z3", 0.0);
+  mp.set("z4", 0.0);
+  RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+  const int numRows   = static_cast<int>(A->getLocalNumRows());
+
+  RCP<Aggregates> aggregates       = rcp(new Aggregates(A->getMap()));
+  RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
+  std::vector<unsigned> aggStat(numRows, MueLu::READY);
+  LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
+  typename Matrix::local_matrix_type intermediateP;
+
+  Array<LO> orderingVector(numRows);
+  for (LO i = 0; i < numRows; i++) {
+    orderingVector[i] = i;
+  }
+
+  Teuchos::ParameterList params;
+  params.set("aggregation: pairwise: tie threshold", 1e-6);
+  NAF->BuildInitialAggregates(params, A, orderingVector(),
+                              Teuchos::ScalarTraits<SC>::magnitude(4.1),
+                              *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+
+  Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
+  auto v2a                    = aggregates->GetVertex2AggId()->getData(0);
+
+  TEST_EQUALITY(numUnaggregatedNodes, 0);
+
+  typename Matrix::local_matrix_type coarseA = A->getLocalMatrixDevice();
+  NAF->BuildOnRankLocalMatrix(A->getLocalMatrixDevice(), coarseA);
+  NAF->BuildIntermediateProlongator(A->getLocalNumRows(), numDirichletNodes,
+                                    aggregates->GetNumAggregates(),
+                                    v2a.view(0, A->getLocalNumRows()),
+                                    intermediateP);
+  NAF->BuildCoarseLocalMatrix(intermediateP, coarseA);
+
+}  // CoarseLocalMatrix2D
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, BuildNotayAggregates2D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  Xpetra::UnderlyingLib lib           = TestHelpers_kokkos::Parameters::getLib();
+
+  // Do a 2D Star2D Matrix with up/down as a "weak connection"
+  Teuchos::ParameterList mp;
+  mp.set("matrixType", "Star2D");
+  const int nx = 9;
+  mp.set("nx", (GO)nx);
+  mp.set("ny", (GO)nx * comm->getSize());
+  mp.set("a", 2.2);
+  mp.set("b", -0.1);
+  mp.set("c", -0.1);
+  mp.set("d", -1.0);
+  mp.set("e", -1.0);
+  mp.set("z1", 0.0);
+  mp.set("z2", 0.0);
+  mp.set("z3", 0.0);
+  mp.set("z4", 0.0);
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+
+  // Use default ordering
+  {
+    MueLu::Level currentLevel;
+    test_factory::createSingleLevelHierarchy(currentLevel);
+    currentLevel.SetFactoryManager(Teuchos::null);
+    currentLevel.Request("A");
+    currentLevel.Set("A", A);
 
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, IntermediateProlongator2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    //    using TST                   = Teuchos::ScalarTraits<SC>;
-    //    using magnitude_type        = typename TST::magnitudeType;
-    //    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    //    using real_type             = typename TST::coordinateType;
-    //    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    // Do a 2D Star2D Matrix with up/down as a "weak connection"
-    Teuchos::ParameterList mp;
-    mp.set("matrixType","Star2D");
-    const int nx = 3;
-    mp.set("nx",(GO)nx);
-    mp.set("ny",(GO)nx*comm->getSize());
-    mp.set("a",2.2);
-    mp.set("b",-0.1);  mp.set("c",-0.1);
-    mp.set("d",-1.0); mp.set("e",-1.0);
-    mp.set("z1",0.0);  mp.set("z2",0.0);  mp.set("z3",0.0);  mp.set("z4",0.0);
-    RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-    const int numRows = static_cast<int>(A->getLocalNumRows());
-
-    RCP<Aggregates> aggregates = rcp(new Aggregates(A->getMap()));
     RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
-    std::vector<unsigned> aggStat(numRows, MueLu::READY);
-    LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
-    typename Matrix::local_matrix_type intermediateP;
+    NAF->SetFactory("Graph", dropFact);
+    NAF->SetFactory("DofsPerNode", dropFact);
+    NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
+    NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
+
+    currentLevel.Request("Aggregates", NAF.get());
+    currentLevel.Request(*NAF);
+    NAF->Build(currentLevel);
+    out << "Pairwise aggregation factory is done" << std::endl;
+    RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
+    out << "Recovered aggregates from muelu level" << std::endl;
+    currentLevel.Release("Aggregates", NAF.get());
+    std::cout << "Testing pairwise aggregates" << std::endl;
+
+    auto v2a                    = aggregates->GetVertex2AggId()->getData(0);
+    Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
 
-    Array<LO> orderingVector(numRows);
-    for (LO i = 0; i < numRows; i++) {
-      orderingVector[i] = i;
-    }
+    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  }
 
-    Teuchos::ParameterList params;
-    params.set("aggregation: pairwise: tie threshold",1e-6);
-    NAF->BuildInitialAggregates(params, A, orderingVector(),
-                                Teuchos::ScalarTraits<SC>::magnitude(4.1),
-                                *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+  // Use random ordering
+  {
+    MueLu::Level currentLevel;
+    test_factory::createSingleLevelHierarchy(currentLevel);
+    currentLevel.SetFactoryManager(Teuchos::null);
+    currentLevel.Request("A");
+    currentLevel.Set("A", A);
 
-    out << "numDirichletNodes=" << numDirichletNodes << std::endl;
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
+    RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
+    NAF->SetFactory("Graph", dropFact);
+    NAF->SetFactory("DofsPerNode", dropFact);
+    NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
+    NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
+    NAF->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("random")));
+
+    currentLevel.Request("Aggregates", NAF.get());
+    currentLevel.Request(*NAF);
+    NAF->Build(currentLevel);
+    out << "Pairwise aggregation factory is done" << std::endl;
+    RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
+    out << "Recovered aggregates from muelu level" << std::endl;
+    currentLevel.Release("Aggregates", NAF.get());
+    std::cout << "Testing pairwise aggregates" << std::endl;
+
+    auto v2a                    = aggregates->GetVertex2AggId()->getData(0);
     Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-    auto v2a = aggregates->GetVertex2AggId()->getData(0);
-
-    TEST_EQUALITY(numUnaggregatedNodes, 0);
-
-    NAF->BuildIntermediateProlongator(A->getLocalNumRows(), numDirichletNodes,
-                                      aggregates->GetNumAggregates(),
-                                      v2a.view(0, A->getLocalNumRows()),
-                                      intermediateP);
 
-  } // IntermediateProlongator2D
+    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  }
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, CoarseLocalMatrix2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
+  // Use Cuthill-McKee if we have it (requires Kokkos refactor)
   {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    // Do a 2D Star2D Matrix with up/down as a "weak connection"
-    Teuchos::ParameterList mp;
-    mp.set("matrixType","Star2D");
-    const GO nx = 3;
-    mp.set("nx", nx);
-    mp.set("ny", nx*comm->getSize());
-    mp.set("a",  2.2);
-    mp.set("b", -0.1); mp.set("c", -0.1);
-    mp.set("d", -1.0); mp.set("e", -1.0);
-    mp.set("z1", 0.0); mp.set("z2", 0.0);  mp.set("z3", 0.0);  mp.set("z4", 0.0);
-    RCP<const Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-    const int numRows = static_cast<int>(A->getLocalNumRows());
-
-    RCP<Aggregates> aggregates = rcp(new Aggregates(A->getMap()));
-    RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
-    std::vector<unsigned> aggStat(numRows, MueLu::READY);
-    LO numUnaggregatedNodes = numRows, numDirichletNodes = 0;
-    typename Matrix::local_matrix_type intermediateP;
-
-    Array<LO> orderingVector(numRows);
-    for (LO i = 0; i < numRows; i++) {
-      orderingVector[i] = i;
-    }
+    MueLu::Level currentLevel;
+    test_factory::createSingleLevelHierarchy(currentLevel);
+    currentLevel.SetFactoryManager(Teuchos::null);
+    currentLevel.Request("A");
+    currentLevel.Set("A", A);
 
-    Teuchos::ParameterList params;
-    params.set("aggregation: pairwise: tie threshold",1e-6);
-    NAF->BuildInitialAggregates(params, A, orderingVector(),
-                                Teuchos::ScalarTraits<SC>::magnitude(4.1),
-                                *aggregates, aggStat, numUnaggregatedNodes, numDirichletNodes);
+    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+    RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
+    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
+    RCP<NotayAggregationFactory> NAF = rcp(new NotayAggregationFactory());
+    NAF->SetFactory("Graph", dropFact);
+    NAF->SetFactory("DofsPerNode", dropFact);
+    NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
+    NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
+    NAF->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("cuthill-mckee")));
+
+    currentLevel.Request("Aggregates", NAF.get());
+    currentLevel.Request(*NAF);
+    NAF->Build(currentLevel);
+    out << "Pairwise aggregation factory is done" << std::endl;
+    RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
+    out << "Recovered aggregates from muelu level" << std::endl;
+    currentLevel.Release("Aggregates", NAF.get());
+    std::cout << "Testing pairwise aggregates" << std::endl;
+
+    auto v2a                    = aggregates->GetVertex2AggId()->getData(0);
     Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-    auto v2a = aggregates->GetVertex2AggId()->getData(0);
 
-    TEST_EQUALITY(numUnaggregatedNodes, 0);
+    TEST_EQUALITY(aggregates->AggregatesCrossProcessors(), false);
+  }
 
-    typename Matrix::local_matrix_type coarseA = A->getLocalMatrixDevice();
-    NAF->BuildOnRankLocalMatrix(A->getLocalMatrixDevice(), coarseA);
-    NAF->BuildIntermediateProlongator(A->getLocalNumRows(), numDirichletNodes,
-                                      aggregates->GetNumAggregates(),
-                                      v2a.view(0, A->getLocalNumRows()),
-                                      intermediateP);
-    NAF->BuildCoarseLocalMatrix(intermediateP, coarseA);
-
-  } // CoarseLocalMatrix2D
-
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(NotayAggregation, BuildNotayAggregates2D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-    RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-
-    // Do a 2D Star2D Matrix with up/down as a "weak connection"
-    Teuchos::ParameterList mp;
-    mp.set("matrixType","Star2D");
-    const int nx = 9;
-    mp.set("nx",(GO)nx);
-    mp.set("ny",(GO)nx*comm->getSize());
-    mp.set("a",2.2);
-    mp.set("b",-0.1);  mp.set("c",-0.1);
-    mp.set("d",-1.0); mp.set("e",-1.0);
-    mp.set("z1",0.0);  mp.set("z2",0.0);  mp.set("z3",0.0);  mp.set("z4",0.0);
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-
-    // Use default ordering
-    {
-      MueLu::Level currentLevel;
-      test_factory::createSingleLevelHierarchy(currentLevel);
-      currentLevel.SetFactoryManager(Teuchos::null);
-      currentLevel.Request("A");
-      currentLevel.Set("A", A);
-      
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      
-      RCP<NotayAggregationFactory> NAF   = rcp(new NotayAggregationFactory());
-      NAF->SetFactory("Graph", dropFact);
-      NAF->SetFactory("DofsPerNode", dropFact);
-      NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
-      NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
-      
-      currentLevel.Request("Aggregates", NAF.get());
-      currentLevel.Request(*NAF);
-      NAF->Build(currentLevel);
-      out << "Pairwise aggregation factory is done" << std::endl;
-      RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
-      out << "Recovered aggregates from muelu level" << std::endl;
-      currentLevel.Release("Aggregates", NAF.get());
-      std::cout << "Testing pairwise aggregates" << std::endl;
-      
-      auto v2a = aggregates->GetVertex2AggId()->getData(0);
-      Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-      
-      TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-    }
-
-    // Use random ordering
-    {
-      MueLu::Level currentLevel;
-      test_factory::createSingleLevelHierarchy(currentLevel);
-      currentLevel.SetFactoryManager(Teuchos::null);
-      currentLevel.Request("A");
-      currentLevel.Set("A", A);
-      
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      
-      RCP<NotayAggregationFactory> NAF   = rcp(new NotayAggregationFactory());
-      NAF->SetFactory("Graph", dropFact);
-      NAF->SetFactory("DofsPerNode", dropFact);
-      NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
-      NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
-      NAF->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("random")));
-      
-      currentLevel.Request("Aggregates", NAF.get());
-      currentLevel.Request(*NAF);
-      NAF->Build(currentLevel);
-      out << "Pairwise aggregation factory is done" << std::endl;
-      RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
-      out << "Recovered aggregates from muelu level" << std::endl;
-      currentLevel.Release("Aggregates", NAF.get());
-      std::cout << "Testing pairwise aggregates" << std::endl;
-      
-      auto v2a = aggregates->GetVertex2AggId()->getData(0);
-      Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-      
-      TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-    }
-
-    // Use Cuthill-McKee if we have it (requires Kokkos refactor)
-   {
-      MueLu::Level currentLevel;
-      test_factory::createSingleLevelHierarchy(currentLevel);
-      currentLevel.SetFactoryManager(Teuchos::null);
-      currentLevel.Request("A");
-      currentLevel.Set("A", A);
-      
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory> dropFact  = rcp(new CoalesceDropFactory());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-      
-      RCP<NotayAggregationFactory> NAF   = rcp(new NotayAggregationFactory());
-      NAF->SetFactory("Graph", dropFact);
-      NAF->SetFactory("DofsPerNode", dropFact);
-      NAF->SetParameter("aggregation: pairwise: size", Teuchos::ParameterEntry(4));
-      NAF->SetParameter("aggregation: Dirichlet threshold", Teuchos::ParameterEntry(4.1));
-      NAF->SetParameter("aggregation: ordering", Teuchos::ParameterEntry(std::string("cuthill-mckee")));
-      
-      currentLevel.Request("Aggregates", NAF.get());
-      currentLevel.Request(*NAF);
-      NAF->Build(currentLevel);
-      out << "Pairwise aggregation factory is done" << std::endl;
-      RCP<Aggregates> aggregates = currentLevel.Get<RCP<Aggregates> >("Aggregates", NAF.get());
-      out << "Recovered aggregates from muelu level" << std::endl;
-      currentLevel.Release("Aggregates", NAF.get());
-      std::cout << "Testing pairwise aggregates" << std::endl;
-      
-      auto v2a = aggregates->GetVertex2AggId()->getData(0);
-      Teuchos::ArrayRCP<LO> sizes = aggregates->ComputeAggregateSizesArrayRCP();
-      
-      TEST_EQUALITY(aggregates->AggregatesCrossProcessors(),false);
-    }
-
-  } // BuildNotayAggregates2D
-
-#  define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation,InitialAggregation1D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation,InitialAggregation2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation,IntermediateProlongator2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation,CoarseLocalMatrix2D,Scalar,LO,GO,Node) \
-      TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation,BuildNotayAggregates2D,Scalar,LO,GO,Node)
+}  // BuildNotayAggregates2D
 
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                             \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation, InitialAggregation1D, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation, InitialAggregation2D, Scalar, LO, GO, Node)      \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation, IntermediateProlongator2D, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation, CoarseLocalMatrix2D, Scalar, LO, GO, Node)       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(NotayAggregation, BuildNotayAggregates2D, Scalar, LO, GO, Node)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/SaPFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/SaPFactory_kokkos.cpp
index 900eee5c0aab..d9e03b6213de 100644
--- a/packages/muelu/test/unit_tests_kokkos/SaPFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/SaPFactory_kokkos.cpp
@@ -62,221 +62,228 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
-    TEST_EQUALITY(sapFactory != Teuchos::null, true);
-
-    out << *sapFactory << std::endl;
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // construct two levels
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // construct matrices
-    const typename Teuchos::ScalarTraits<Scalar>::magnitudeType lambdaMax = 5;
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build2DPoisson(27*comm->getSize());
-    A->SetMaxEigenvalueEstimate(lambdaMax);
-    RCP<Matrix> Ptent = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build2DPoisson(27*comm->getSize());
-
-    // set level matrices
-    fineLevel  .Set("A", A);
-    coarseLevel.Set("P", Ptent);
-
-    // construct the factory to be tested
-    const double dampingFactor = 0.5;
-    RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
-    ParameterList Pparams;
-    Pparams.set("sa: damping factor", dampingFactor);
-    sapFactory->SetParameterList(Pparams);
-    sapFactory->SetFactory("A", MueLu::NoFactory::getRCP());
-    sapFactory->SetFactory("P", MueLu::NoFactory::getRCP());
-
-    // build the data
-    coarseLevel.Request("P", sapFactory.get());
-    sapFactory->Build(fineLevel, coarseLevel);
-
-    // fetch the data
-    RCP<Matrix> Pfact = coarseLevel.Get<RCP<Matrix>>("P", sapFactory.get());
-
-    // construct the data to compare
-    SC omega = dampingFactor / lambdaMax;
-    RCP<Vector> invDiag = Utilities::GetMatrixDiagonalInverse(*A);
-    RCP<ParameterList> APparams = rcp(new ParameterList);
-    RCP<Matrix> Ptest   = Xpetra::IteratorOps<SC,LO,GO,NO>::Jacobi(omega, *invDiag, *A, *Ptent, Teuchos::null, out, "label", APparams);
-
-    // compare matrices by multiplying them by a random vector
-    RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-    X->setSeed(846930886);
-    X->randomize();
-
-    RCP<MultiVector> Bfact = MultiVectorFactory::Build(A->getRangeMap(),  1);
-    RCP<MultiVector> Btest = MultiVectorFactory::Build(A->getRangeMap(),  1);
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero(), one = STS::one();
-
-    Pfact->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
-    Ptest->apply(*X, *Btest, Teuchos::NO_TRANS, one, zero);
-    Btest->update(-one, *Bfact, one);
-
-    Array<typename STS::magnitudeType> norms(1);
-    Btest->norm2(norms);
-    out << "|| B_factory - B_test || = " << norms[0] << std::endl;
-    TEST_EQUALITY(norms[0] < 1e-12, true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
+  TEST_EQUALITY(sapFactory != Teuchos::null, true);
+
+  out << *sapFactory << std::endl;
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+
+  // construct two levels
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // construct matrices
+  const typename Teuchos::ScalarTraits<Scalar>::magnitudeType lambdaMax = 5;
+  RCP<Matrix> A                                                         = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build2DPoisson(27 * comm->getSize());
+  A->SetMaxEigenvalueEstimate(lambdaMax);
+  RCP<Matrix> Ptent = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build2DPoisson(27 * comm->getSize());
+
+  // set level matrices
+  fineLevel.Set("A", A);
+  coarseLevel.Set("P", Ptent);
+
+  // construct the factory to be tested
+  const double dampingFactor        = 0.5;
+  RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
+  ParameterList Pparams;
+  Pparams.set("sa: damping factor", dampingFactor);
+  sapFactory->SetParameterList(Pparams);
+  sapFactory->SetFactory("A", MueLu::NoFactory::getRCP());
+  sapFactory->SetFactory("P", MueLu::NoFactory::getRCP());
+
+  // build the data
+  coarseLevel.Request("P", sapFactory.get());
+  sapFactory->Build(fineLevel, coarseLevel);
+
+  // fetch the data
+  RCP<Matrix> Pfact = coarseLevel.Get<RCP<Matrix>>("P", sapFactory.get());
+
+  // construct the data to compare
+  SC omega                    = dampingFactor / lambdaMax;
+  RCP<Vector> invDiag         = Utilities::GetMatrixDiagonalInverse(*A);
+  RCP<ParameterList> APparams = rcp(new ParameterList);
+  RCP<Matrix> Ptest           = Xpetra::IteratorOps<SC, LO, GO, NO>::Jacobi(omega, *invDiag, *A, *Ptent, Teuchos::null, out, "label", APparams);
+
+  // compare matrices by multiplying them by a random vector
+  RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
+  X->setSeed(846930886);
+  X->randomize();
+
+  RCP<MultiVector> Bfact = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  RCP<MultiVector> Btest = MultiVectorFactory::Build(A->getRangeMap(), 1);
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero(), one = STS::one();
+
+  Pfact->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
+  Ptest->apply(*X, *Btest, Teuchos::NO_TRANS, one, zero);
+  Btest->update(-one, *Bfact, one);
+
+  Array<typename STS::magnitudeType> norms(1);
+  Btest->norm2(norms);
+  out << "|| B_factory - B_test || = " << norms[0] << std::endl;
+  TEST_EQUALITY(norms[0] < 1e-12, true);
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, EnforceConstraints, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero(), one = STS::one();
+
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+
+  // construct two levels
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+
+  // construct matrices
+  SC lambdaMax = 2;
+
+  Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
+  Teuchos::ParameterList mp;
+  mp.set("matrixType", "Star2D");
+  const GO nx = 6;
+  mp.set("nx", nx);
+  mp.set("ny", nx);
+  mp.set("a", 2.0);
+  mp.set("b", -1.0);
+  mp.set("c", -1.0);
+  mp.set("d", 0.5);
+  mp.set("e", 0.5);
+  mp.set("z1", -0.25);
+  mp.set("z2", -0.25);
+  mp.set("z3", -0.25);
+  mp.set("z4", -0.25);
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+  A->SetMaxEigenvalueEstimate(lambdaMax);
+  mp.set("a", 1.0);
+  mp.set("b", 1.0);
+  mp.set("c", 1.0);
+  mp.set("d", 1.0);
+  mp.set("e", 1.0);
+  mp.set("z1", 1.0);
+  mp.set("z2", 1.0);
+  mp.set("z3", 1.0);
+  mp.set("z4", 1.0);
+  RCP<Matrix> Ptent = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp, lib);
+
+  // set level matrices
+  fineLevel.Set("A", A);
+  coarseLevel.Set("P", Ptent);
+
+  // construct the factory to be tested
+  const double dampingFactor        = 0.5;
+  RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
+  ParameterList Pparams;
+  Pparams.set("sa: damping factor", dampingFactor);
+  Pparams.set("sa: enforce constraints", true);
+  Pparams.set("sa: max eigenvalue", (double)2.0);
+  Pparams.set("tentative: calculate qr", false);
+
+  sapFactory->SetParameterList(Pparams);
+  sapFactory->SetFactory("A", MueLu::NoFactory::getRCP());
+  sapFactory->SetFactory("P", MueLu::NoFactory::getRCP());
+
+  // build the data
+  coarseLevel.Request("P", sapFactory.get());
+  sapFactory->Build(fineLevel, coarseLevel);
+
+  // fetch the data
+  RCP<Matrix> Pfact = coarseLevel.Get<RCP<Matrix>>("P", sapFactory.get());
+
+  // check that row sums are all one by checking the norm of the vector
+  RCP<MultiVector> X     = MultiVectorFactory::Build(A->getDomainMap(), 1);
+  RCP<MultiVector> Bfact = MultiVectorFactory::Build(A->getRangeMap(), 1);
+  X->putScalar(one);
+  Pfact->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
+  Array<typename STS::magnitudeType> norms(1);
+  Bfact->norm2(norms);
+  out << "|| B_factory ones || = " << norms[0] << std::endl;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(as<Scalar>(6.0)), 100 * TMT::eps());
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, ConstrainRowOptimalScalarPDE, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef Teuchos::ScalarTraits<SC> STS;
+  SC zero = STS::zero(), one = STS::one();
+
+  RCP<const Teuchos::Comm<int>> comm = Teuchos::DefaultComm<int>::getComm();
+
+  // Don't test for complex - matrix reader won't work
+  if (STS::isComplex) {
+    success = true;
+    return;
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, EnforceConstraints, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero(), one = STS::one();
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // construct two levels
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
-
-    // construct matrices
-    SC lambdaMax = 2;
-
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-    Teuchos::ParameterList mp;
-    mp.set("matrixType","Star2D");
-    const GO nx = 6;
-    mp.set("nx", nx);
-    mp.set("ny", nx);
-    mp.set("a",  2.0);
-    mp.set("b", -1.0); mp.set("c", -1.0);
-    mp.set("d",  0.5); mp.set("e",  0.5);
-    mp.set("z1",-0.25); mp.set("z2",-0.25 );  mp.set("z3",-0.25);  mp.set("z4", -0.25);
-    RCP<Matrix> A= TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-    A->SetMaxEigenvalueEstimate(lambdaMax);
-    mp.set("a",  1.0);
-    mp.set("b",  1.0); mp.set("c",  1.0);
-    mp.set("d",  1.0); mp.set("e",  1.0);
-    mp.set("z1", 1.0); mp.set("z2", 1.0);  mp.set("z3", 1.0);  mp.set("z4", 1.0);
-    RCP<Matrix> Ptent = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildMatrix(mp,lib);
-
-    // set level matrices
-    fineLevel  .Set("A", A);
-    coarseLevel.Set("P", Ptent);
-
-    // construct the factory to be tested
-    const double dampingFactor = 0.5;
-    RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
-    ParameterList Pparams;
-    Pparams.set("sa: damping factor", dampingFactor);
-    Pparams.set("sa: enforce constraints", true);
-    Pparams.set("sa: max eigenvalue",(double) 2.0);
-    Pparams.set("tentative: calculate qr", false);
-
-    sapFactory->SetParameterList(Pparams);
-    sapFactory->SetFactory("A", MueLu::NoFactory::getRCP());
-    sapFactory->SetFactory("P", MueLu::NoFactory::getRCP());
-
-    // build the data
-    coarseLevel.Request("P", sapFactory.get());
-    sapFactory->Build(fineLevel, coarseLevel);
-
-    // fetch the data
-    RCP<Matrix> Pfact = coarseLevel.Get<RCP<Matrix>>("P", sapFactory.get());
-
-    // check that row sums are all one by checking the norm of the vector
-    RCP<MultiVector> X = MultiVectorFactory::Build(A->getDomainMap(), 1);
-    RCP<MultiVector> Bfact = MultiVectorFactory::Build(A->getRangeMap(),  1);
-    X->putScalar(one);
-    Pfact->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
-    Array<typename STS::magnitudeType> norms(1);
-    Bfact->norm2(norms);
-    out << "|| B_factory ones || = " << norms[0] << std::endl;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-    TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(as<Scalar>(6.0)), 100*TMT::eps());
-  }
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SaPFactory_kokkos, ConstrainRowOptimalScalarPDE, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef Teuchos::ScalarTraits<SC> STS;
-    SC zero = STS::zero(), one = STS::one();
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    // Don't test for complex - matrix reader won't work
-    if (STS::isComplex) {success=true; return;}
-
-    RCP<Matrix> P;
-    Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
-    P =  Xpetra::IO<SC, LO, GO, NO>::Read("../unit_tests/TestMatrices/SaP_constrainTest_P.mat", lib, comm);
-
-    RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
-    sapFactory->optimalSatisfyPConstraintsForScalarPDEs( P );
-
-    // check that row sums are all one by checking the norm of the vector
-    // note: optimalSatisfyPConstraintsForScalarPDEs preserves row sum of original P (one in this case),
-    //       but SatisfyPConstraints normalizes each row sum to one.
-    RCP<MultiVector> X = MultiVectorFactory::Build(P->getDomainMap(), 1);
-    RCP<MultiVector> Bfact = MultiVectorFactory::Build(P->getRangeMap(),  1);
-    X->putScalar(one);
-    P->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
-    Array<typename STS::magnitudeType> norms(1);
-    Bfact->norm2(norms);
-    out << "|| B_factory ones || = " << norms[0] << std::endl;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-    TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(6.0), 100*TMT::eps());
-
-    // check that the min and max of each row are in [0,1]
-    bool lowerViolation = false;
-    bool upperViolation = false;
-    for (size_t j = 0; j < as<size_t>(P->getRowMap()->getLocalNumElements()); j++) {
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> vals;
-      P->getLocalRowView((LocalOrdinal) j, indices, vals);
-      size_t nnz = indices.size();
-      for (LO i = 0; i < (LO) nnz; i++)  {
-        if (Teuchos::ScalarTraits<SC>::real(vals[i]) < Teuchos::ScalarTraits<SC>::real(zero)) lowerViolation = true;
-        if (Teuchos::ScalarTraits<SC>::real(vals[i]) > Teuchos::ScalarTraits<SC>::real(one)) upperViolation = true;
-      }
+  RCP<Matrix> P;
+  Xpetra::UnderlyingLib lib = TestHelpers_kokkos::Parameters::getLib();
+  P                         = Xpetra::IO<SC, LO, GO, NO>::Read("../unit_tests/TestMatrices/SaP_constrainTest_P.mat", lib, comm);
+
+  RCP<SaPFactory_kokkos> sapFactory = rcp(new SaPFactory_kokkos);
+  sapFactory->optimalSatisfyPConstraintsForScalarPDEs(P);
+
+  // check that row sums are all one by checking the norm of the vector
+  // note: optimalSatisfyPConstraintsForScalarPDEs preserves row sum of original P (one in this case),
+  //       but SatisfyPConstraints normalizes each row sum to one.
+  RCP<MultiVector> X     = MultiVectorFactory::Build(P->getDomainMap(), 1);
+  RCP<MultiVector> Bfact = MultiVectorFactory::Build(P->getRangeMap(), 1);
+  X->putScalar(one);
+  P->apply(*X, *Bfact, Teuchos::NO_TRANS, one, zero);
+  Array<typename STS::magnitudeType> norms(1);
+  Bfact->norm2(norms);
+  out << "|| B_factory ones || = " << norms[0] << std::endl;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  TEST_FLOATING_EQUALITY(STS::magnitude(norms[0]), STS::magnitude(6.0), 100 * TMT::eps());
+
+  // check that the min and max of each row are in [0,1]
+  bool lowerViolation = false;
+  bool upperViolation = false;
+  for (size_t j = 0; j < as<size_t>(P->getRowMap()->getLocalNumElements()); j++) {
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> vals;
+    P->getLocalRowView((LocalOrdinal)j, indices, vals);
+    size_t nnz = indices.size();
+    for (LO i = 0; i < (LO)nnz; i++) {
+      if (Teuchos::ScalarTraits<SC>::real(vals[i]) < Teuchos::ScalarTraits<SC>::real(zero)) lowerViolation = true;
+      if (Teuchos::ScalarTraits<SC>::real(vals[i]) > Teuchos::ScalarTraits<SC>::real(one)) upperViolation = true;
     }
-    TEST_EQUALITY(lowerViolation, false);
-    TEST_EQUALITY(upperViolation, false);
-
+  }
+  TEST_EQUALITY(lowerViolation, false);
+  TEST_EQUALITY(upperViolation, false);
 
-  } //SaPFactory_ConstrainRowOptimalScalarPDE
+}  // SaPFactory_ConstrainRowOptimalScalarPDE
 
-  // FIXME_KOKKOS: uncomment the test when we get all corresponding factories ported to kokkos
+// FIXME_KOKKOS: uncomment the test when we get all corresponding factories ported to kokkos
 #if 0
 #if defined(HAVE_MUELU_EPETRA) && defined(HAVE_MUELU_EPETRAEXT) && defined(HAVE_MUELU_IFPACK) && defined(HAVE_MUELU_IFPACK2)
   TEUCHOS_UNIT_TEST(SaPFactory_kokkos, EpetraVsTpetra)
   {
-#   include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
     MueLu::VerboseObject::SetMueLuOStream(Teuchos::rcpFromRef(out));
 
     out << "version: " << MueLu::Version() << std::endl;
@@ -476,13 +483,12 @@ namespace MueLuTests {
 #endif
 #endif
 
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, Constructor, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, Build,       SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                       \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, Constructor, SC, LO, GO, NO)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, Build, SC, LO, GO, NO)              \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, EnforceConstraints, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SaPFactory_kokkos, ConstrainRowOptimalScalarPDE, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-}//namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/SemiCoarsenPFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/SemiCoarsenPFactory_kokkos.cpp
index 15a187284cd9..c5629ea98f7d 100644
--- a/packages/muelu/test/unit_tests_kokkos/SemiCoarsenPFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/SemiCoarsenPFactory_kokkos.cpp
@@ -56,122 +56,122 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactory_kokkos, TestSemiCoarsenP, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Global Lexicographic";
-    const std::string coupling = "uncoupled";
-    const LO numDimensions = 3;
-    const LO numSize = 5;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(numDimensions);
-    Array<GO> gNodesPerDir(numDimensions);
-    for(int dim = 0; dim < numDimensions; ++dim)
-      gNodesPerDir[dim] = numSize;
-
-    typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coord_type;
-    typedef Xpetra::MultiVector<coord_type, LO, GO, NO> CoordMV;
-    RCP<CoordMV> fineCoords =
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                        lNodesPerDir, meshData,
-                                                                        meshLayout);
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("ny", gNodesPerDir[1]);
-    matrixList.set("nz", gNodesPerDir[2]);
-    matrixList.set("matrixType","Laplace3D");
-    matrixList.set("left boundary" , "Neumann");
-    matrixList.set("right boundary" , "Neumann");
-    matrixList.set("front boundary" , "Neumann");
-    matrixList.set("back boundary"  , "Neumann");
-    matrixList.set("bottom boundary", "Dirichlet");
-    matrixList.set("top boundary"   , "Dirichlet");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", fineCoords->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    nullSpace->putScalar(1.0);
-
-    fineLevel.Set("A", A);
-    fineLevel.Set("Nullspace", nullSpace);
-    fineLevel.Set("Coordinates", fineCoords);
-    fineLevel.Set("CoarseNumZLayers", numSize);
-
-    RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
-    LineDetectionFact->SetParameter("linedetection: orientation",
-                                    Teuchos::ParameterEntry(std::string("coordinates")));
-    LineDetectionFact->SetParameter("linedetection: num layers",
-                                    Teuchos::ParameterEntry(numSize));
-
-    RCP<SemiCoarsenPFactory_kokkos> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory_kokkos());
-    SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
-    SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
-    SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
-    coarseLevel.Request("P", SemiCoarsenPFact.get());
-    coarseLevel.Request("Coordinates", SemiCoarsenPFact.get());
-    SemiCoarsenPFact->Build(fineLevel, coarseLevel);
-
-    RCP<Matrix> P;
-    coarseLevel.Get("P", P, SemiCoarsenPFact.get());
-    RCP<CoordMV> coarseCoords = coarseLevel.Get<RCP<CoordMV>>("Coordinates", SemiCoarsenPFact.get());
-
-    coarseLevel.Release("P", SemiCoarsenPFact.get());
-    coarseLevel.Release("Coordinates", SemiCoarsenPFact.get());
-
-    // Prolongate coarse coords and compute difference
-    using STS = Teuchos::ScalarTraits<SC>;
-    const auto one = STS::one();
-    RCP<MultiVector> coarseCoordsSC = Utilities::RealValuedToScalarMultiVector(coarseCoords);
-    RCP<MultiVector> fineCoordsDiff = Utilities::RealValuedToScalarMultiVector(fineCoords);
-    P->apply(*coarseCoordsSC, *fineCoordsDiff, Teuchos::NO_TRANS, one, -one);
-
-    // check prolongation of coarse coordinates
-    // in this special case, the third layer will have the correct coordinates
-    using impl_SC = typename Kokkos::ArithTraits<SC>::val_type;
-    using impl_ATS = Kokkos::ArithTraits<impl_SC>;
-    const auto fineCoordsDiffView = fineCoordsDiff->getDeviceLocalView(Xpetra::Access::ReadOnly);
-    const int numNodes = fineCoordsDiff->getLocalLength();
-    const int numVectors = fineCoordsDiff->getNumVectors();
-    const auto fineMap = fineCoordsDiff->getMap()->getLocalMap();
-    const GO gStartLayer3 = 2 * numSize * numSize;
-    const GO gEndLayer3 = 3 * numSize * numSize;
-    using range_policy = Kokkos::RangePolicy<LO, typename SemiCoarsenPFactory_kokkos::execution_space>;
-    LO numBadCoords = 0;
-    Kokkos::parallel_reduce(
-        "Checking zeros on third layer",
-        range_policy(0, numNodes), KOKKOS_LAMBDA(const LO node, LO& lNumBadCoords) {
-          const GO gnode = fineMap.getGlobalElement(node);
-          if (gnode >= gStartLayer3 && gnode < gEndLayer3)
-            for (int k = 0; k < numVectors; ++k)
-              if (impl_ATS::magnitude(fineCoordsDiffView(node, k)) > 100*impl_ATS::eps())
-                lNumBadCoords += 1;
-        }, numBadCoords);
-    LO gNumBadCoords;
-    const auto comm = Teuchos::DefaultComm<int>::getComm();
-    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &numBadCoords, &gNumBadCoords);
-    TEST_EQUALITY(gNumBadCoords, 0);
-  }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(SemiCoarsenPFactory_kokkos, TestSemiCoarsenP, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Global Lexicographic";
+  const std::string coupling   = "uncoupled";
+  const LO numDimensions       = 3;
+  const LO numSize             = 5;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(numDimensions);
+  Array<GO> gNodesPerDir(numDimensions);
+  for (int dim = 0; dim < numDimensions; ++dim)
+    gNodesPerDir[dim] = numSize;
+
+  typedef typename Teuchos::ScalarTraits<Scalar>::coordinateType coord_type;
+  typedef Xpetra::MultiVector<coord_type, LO, GO, NO> CoordMV;
+  RCP<CoordMV> fineCoords =
+      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                           lNodesPerDir, meshData,
+                                                                           meshLayout);
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("ny", gNodesPerDir[1]);
+  matrixList.set("nz", gNodesPerDir[2]);
+  matrixList.set("matrixType", "Laplace3D");
+  matrixList.set("left boundary", "Neumann");
+  matrixList.set("right boundary", "Neumann");
+  matrixList.set("front boundary", "Neumann");
+  matrixList.set("back boundary", "Neumann");
+  matrixList.set("bottom boundary", "Dirichlet");
+  matrixList.set("top boundary", "Dirichlet");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector>> Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", fineCoords->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+
+  fineLevel.Set("A", A);
+  fineLevel.Set("Nullspace", nullSpace);
+  fineLevel.Set("Coordinates", fineCoords);
+  fineLevel.Set("CoarseNumZLayers", numSize);
+
+  RCP<LineDetectionFactory> LineDetectionFact = rcp(new LineDetectionFactory());
+  LineDetectionFact->SetParameter("linedetection: orientation",
+                                  Teuchos::ParameterEntry(std::string("coordinates")));
+  LineDetectionFact->SetParameter("linedetection: num layers",
+                                  Teuchos::ParameterEntry(numSize));
+
+  RCP<SemiCoarsenPFactory_kokkos> SemiCoarsenPFact = rcp(new SemiCoarsenPFactory_kokkos());
+  SemiCoarsenPFact->SetParameter("semicoarsen: coarsen rate", Teuchos::ParameterEntry(2));
+  SemiCoarsenPFact->SetFactory("LineDetection_VertLineIds", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("LineDetection_Layers", LineDetectionFact);
+  SemiCoarsenPFact->SetFactory("CoarseNumZLayers", LineDetectionFact);
+  coarseLevel.Request("P", SemiCoarsenPFact.get());
+  coarseLevel.Request("Coordinates", SemiCoarsenPFact.get());
+  SemiCoarsenPFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> P;
+  coarseLevel.Get("P", P, SemiCoarsenPFact.get());
+  RCP<CoordMV> coarseCoords = coarseLevel.Get<RCP<CoordMV>>("Coordinates", SemiCoarsenPFact.get());
+
+  coarseLevel.Release("P", SemiCoarsenPFact.get());
+  coarseLevel.Release("Coordinates", SemiCoarsenPFact.get());
+
+  // Prolongate coarse coords and compute difference
+  using STS                       = Teuchos::ScalarTraits<SC>;
+  const auto one                  = STS::one();
+  RCP<MultiVector> coarseCoordsSC = Utilities::RealValuedToScalarMultiVector(coarseCoords);
+  RCP<MultiVector> fineCoordsDiff = Utilities::RealValuedToScalarMultiVector(fineCoords);
+  P->apply(*coarseCoordsSC, *fineCoordsDiff, Teuchos::NO_TRANS, one, -one);
+
+  // check prolongation of coarse coordinates
+  // in this special case, the third layer will have the correct coordinates
+  using impl_SC                 = typename Kokkos::ArithTraits<SC>::val_type;
+  using impl_ATS                = Kokkos::ArithTraits<impl_SC>;
+  const auto fineCoordsDiffView = fineCoordsDiff->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  const int numNodes            = fineCoordsDiff->getLocalLength();
+  const int numVectors          = fineCoordsDiff->getNumVectors();
+  const auto fineMap            = fineCoordsDiff->getMap()->getLocalMap();
+  const GO gStartLayer3         = 2 * numSize * numSize;
+  const GO gEndLayer3           = 3 * numSize * numSize;
+  using range_policy            = Kokkos::RangePolicy<LO, typename SemiCoarsenPFactory_kokkos::execution_space>;
+  LO numBadCoords               = 0;
+  Kokkos::parallel_reduce(
+      "Checking zeros on third layer",
+      range_policy(0, numNodes), KOKKOS_LAMBDA(const LO node, LO& lNumBadCoords) {
+        const GO gnode = fineMap.getGlobalElement(node);
+        if (gnode >= gStartLayer3 && gnode < gEndLayer3)
+          for (int k = 0; k < numVectors; ++k)
+            if (impl_ATS::magnitude(fineCoordsDiffView(node, k)) > 100 * impl_ATS::eps())
+              lNumBadCoords += 1;
+      },
+      numBadCoords);
+  LO gNumBadCoords;
+  const auto comm = Teuchos::DefaultComm<int>::getComm();
+  Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &numBadCoords, &gNumBadCoords);
+  TEST_EQUALITY(gNumBadCoords, 0);
+}
 
 #define MUELU_ETI_GROUP(SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(SemiCoarsenPFactory_kokkos, TestSemiCoarsenP, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/StructuredAggregation_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/StructuredAggregation_kokkos.cpp
index 605a21b4a7cf..79a892987713 100644
--- a/packages/muelu/test/unit_tests_kokkos/StructuredAggregation_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/StructuredAggregation_kokkos.cpp
@@ -61,302 +61,296 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, CreateCrsGraphConstant, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, CreateCrsGraphConstant, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  typedef Teuchos::ScalarTraits<Scalar> TST;
+  typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level currentLevel;
+  test_factory::createSingleLevelHierarchy(currentLevel);
+  currentLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    typedef Teuchos::ScalarTraits<Scalar> TST;
-    typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level currentLevel;
-    test_factory::createSingleLevelHierarchy(currentLevel);
-    currentLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                        lNodesPerDir, meshData,
-                                                                        meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
-    }
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    currentLevel.Request("A");
-    currentLevel.Set("A", A);
-    currentLevel.Set("numDimensions", numDimensions);
-    currentLevel.Set("Coordinates", Coordinates);
-    currentLevel.Set("lNodesPerDim", lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    currentLevel.Set("Nullspace",nullSpace);
-
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
-    StructuredAggFact->SetFactory("Graph",       dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-    StructuredAggFact->SetParameter("aggregation: output type",
-                                    Teuchos::ParameterEntry(std::string("CrsGraph")));
-
-    currentLevel.Request("prolongatorGraph", StructuredAggFact.get());  // request prolongatorGraph
-    currentLevel.Request(*StructuredAggFact);
-    StructuredAggFact->Build(currentLevel);
-  } // CreateCrsGraphConstant
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, CreateCrsGraphLinear, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-
-    typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
-    typedef typename Xpetra::MultiVector<real_type,LO,GO,NO> RealValuedMultiVector;
-
-    typedef Teuchos::ScalarTraits<Scalar> TST;
-    typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level currentLevel;
-    test_factory::createSingleLevelHierarchy(currentLevel);
-    currentLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
-    }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                        lNodesPerDir, meshData,
-                                                                        meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                           lNodesPerDir, meshData,
+                                                                           meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  currentLevel.Request("A");
+  currentLevel.Set("A", A);
+  currentLevel.Set("numDimensions", numDimensions);
+  currentLevel.Set("Coordinates", Coordinates);
+  currentLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  currentLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+  StructuredAggFact->SetParameter("aggregation: output type",
+                                  Teuchos::ParameterEntry(std::string("CrsGraph")));
+
+  currentLevel.Request("prolongatorGraph", StructuredAggFact.get());  // request prolongatorGraph
+  currentLevel.Request(*StructuredAggFact);
+  StructuredAggFact->Build(currentLevel);
+}  // CreateCrsGraphConstant
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, CreateCrsGraphLinear, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
+
+  typedef typename Teuchos::ScalarTraits<SC>::magnitudeType real_type;
+  typedef typename Xpetra::MultiVector<real_type, LO, GO, NO> RealValuedMultiVector;
+
+  typedef Teuchos::ScalarTraits<Scalar> TST;
+  typedef TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node> test_factory;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level currentLevel;
+  test_factory::createSingleLevelHierarchy(currentLevel);
+  currentLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    currentLevel.Request("A");
-    currentLevel.Set("A", A);
-    currentLevel.Set("numDimensions", numDimensions);
-    currentLevel.Set("Coordinates", Coordinates);
-    currentLevel.Set("lNodesPerDim", lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    currentLevel.Set("Nullspace",nullSpace);
-
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
-    StructuredAggFact->SetFactory("Graph",       dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(1));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-    StructuredAggFact->SetParameter("aggregation: output type",
-                                    Teuchos::ParameterEntry(std::string("CrsGraph")));
-
-    currentLevel.Request("prolongatorGraph", StructuredAggFact.get());  // request prolongatorGraph
-    currentLevel.Request(*StructuredAggFact);
-    StructuredAggFact->Build(currentLevel);
-
-    RCP<CrsGraph> prolongatorGraph;
-    currentLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
-
-  } // CreateCrsGraphLinear
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, UncoupledTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-
-    using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
-    using TST                   = Teuchos::ScalarTraits<SC>;
-    using magnitude_type        = typename TST::magnitudeType;
-    using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
-    using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-
-    out << "version: " << MueLu::Version() << std::endl;
-
-    Level fineLevel, coarseLevel;
-    test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
-    fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
-
-    // Set global geometric data
-    const std::string meshLayout = "Local Lexicographic";
-    const std::string coupling = "uncoupled";
-    LO numDimensions = 3;
-    Array<GO> meshData;
-    Array<LO> lNodesPerDir(3);
-    Array<GO> gNodesPerDir(3);
-    for(int dim = 0; dim < 3; ++dim) {
-      if(dim < numDimensions) {
-        // Use more nodes in 1D to have a reasonable number of nodes per procs
-        gNodesPerDir[dim] = 6;
-      } else {
-        gNodesPerDir[dim] = 1;
-      }
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                           lNodesPerDir, meshData,
+                                                                           meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  currentLevel.Request("A");
+  currentLevel.Set("A", A);
+  currentLevel.Set("numDimensions", numDimensions);
+  currentLevel.Set("Coordinates", Coordinates);
+  currentLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  currentLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(1));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+  StructuredAggFact->SetParameter("aggregation: output type",
+                                  Teuchos::ParameterEntry(std::string("CrsGraph")));
+
+  currentLevel.Request("prolongatorGraph", StructuredAggFact.get());  // request prolongatorGraph
+  currentLevel.Request(*StructuredAggFact);
+  StructuredAggFact->Build(currentLevel);
+
+  RCP<CrsGraph> prolongatorGraph;
+  currentLevel.Get("prolongatorGraph", prolongatorGraph, StructuredAggFact.get());
+
+}  // CreateCrsGraphLinear
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(StructuredAggregation_kokkos, UncoupledTentative3D, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
+
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = typename Xpetra::MultiVector<real_type, LO, GO, NO>;
+  using TST                   = Teuchos::ScalarTraits<SC>;
+  using magnitude_type        = typename TST::magnitudeType;
+  using TMT                   = Teuchos::ScalarTraits<magnitude_type>;
+  using test_factory          = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+
+  out << "version: " << MueLu::Version() << std::endl;
+
+  Level fineLevel, coarseLevel;
+  test_factory::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
+
+  // Set global geometric data
+  const std::string meshLayout = "Local Lexicographic";
+  const std::string coupling   = "uncoupled";
+  LO numDimensions             = 3;
+  Array<GO> meshData;
+  Array<LO> lNodesPerDir(3);
+  Array<GO> gNodesPerDir(3);
+  for (int dim = 0; dim < 3; ++dim) {
+    if (dim < numDimensions) {
+      // Use more nodes in 1D to have a reasonable number of nodes per procs
+      gNodesPerDir[dim] = 6;
+    } else {
+      gNodesPerDir[dim] = 1;
     }
-
-    RCP<RealValuedMultiVector> Coordinates =
-      TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
-                                                                        lNodesPerDir, meshData,
-                                                                        meshLayout);
-
-    // Since we are doing uncoupled aggregation we fill meshData with -1
-    for(size_t idx = 0; idx < (size_t) meshData.size(); ++idx) {
-      meshData[idx] = -1;
-    }
-
-    Teuchos::ParameterList matrixList;
-    matrixList.set("nx", gNodesPerDir[0]);
-    matrixList.set("matrixType","Laplace1D");
-    RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr = Galeri::Xpetra::
-      BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>("Laplace3D", Coordinates->getMap(),
-                                                           matrixList);
-    RCP<Matrix> A = Pr->BuildMatrix();
-    fineLevel.Request("A");
-    fineLevel.Set("A", A);
-    fineLevel.Set("numDimensions", numDimensions);
-    fineLevel.Set("Coordinates", Coordinates);
-    fineLevel.Set("lNodesPerDim", lNodesPerDir);
-
-    // only one NS vector -> exercises manual orthogonalization
-    LocalOrdinal NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    nullSpace->putScalar(1.0);
-    fineLevel.Set("Nullspace",nullSpace);
-
-
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
-    StructuredAggFact->SetFactory("Graph",       dropFact);
-    StructuredAggFact->SetFactory("DofsPerNode", dropFact);
-    StructuredAggFact->SetParameter("aggregation: coarsening order",
-                                    Teuchos::ParameterEntry(0));
-    StructuredAggFact->SetParameter("aggregation: coarsening rate",
-                                    Teuchos::ParameterEntry(std::string("{3}")));
-
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
-    RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
-    TentativePFact->SetFactory("Aggregates", StructuredAggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap", coarseMapFact);
-
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
-
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P", Ptent, TentativePFact.get());
-
-    RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
-
-    coarseLevel.Release("P", TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace", TentativePFact.get());
-
-    // check normalization and orthogonality of prolongator columns
-    Teuchos::RCP<Xpetra::Matrix<SC,LO,GO,Node> > PtentTPtent = Xpetra::MatrixMatrix<SC,LO,GO,Node>::Multiply(*Ptent, true, *Ptent, false,out);
-    Teuchos::RCP<Xpetra::Vector<SC,LO,GO,Node> > diagVec = Xpetra::VectorFactory<SC,LO,GO,Node>::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (TST::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 1000*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(),  1000*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-
-  } // UncoupledTentative3D
-
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
+  }
+
+  RCP<RealValuedMultiVector> Coordinates =
+      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::BuildGeoCoordinates(numDimensions, gNodesPerDir,
+                                                                           lNodesPerDir, meshData,
+                                                                           meshLayout);
+
+  // Since we are doing uncoupled aggregation we fill meshData with -1
+  for (size_t idx = 0; idx < (size_t)meshData.size(); ++idx) {
+    meshData[idx] = -1;
+  }
+
+  Teuchos::ParameterList matrixList;
+  matrixList.set("nx", gNodesPerDir[0]);
+  matrixList.set("matrixType", "Laplace1D");
+  RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr = Galeri::Xpetra::
+      BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>("Laplace3D", Coordinates->getMap(),
+                                                                matrixList);
+  RCP<Matrix> A = Pr->BuildMatrix();
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
+  fineLevel.Set("numDimensions", numDimensions);
+  fineLevel.Set("Coordinates", Coordinates);
+  fineLevel.Set("lNodesPerDim", lNodesPerDir);
+
+  // only one NS vector -> exercises manual orthogonalization
+  LocalOrdinal NSdim         = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->putScalar(1.0);
+  fineLevel.Set("Nullspace", nullSpace);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  RCP<StructuredAggregationFactory_kokkos> StructuredAggFact = rcp(new StructuredAggregationFactory_kokkos());
+  StructuredAggFact->SetFactory("Graph", dropFact);
+  StructuredAggFact->SetFactory("DofsPerNode", dropFact);
+  StructuredAggFact->SetParameter("aggregation: coarsening order",
+                                  Teuchos::ParameterEntry(0));
+  StructuredAggFact->SetParameter("aggregation: coarsening rate",
+                                  Teuchos::ParameterEntry(std::string("{3}")));
+
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", StructuredAggFact);
+  RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
+  TentativePFact->SetFactory("Aggregates", StructuredAggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
+
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
+
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
+
+  RCP<MultiVector> coarseNullSpace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
+
+  // check normalization and orthogonality of prolongator columns
+  Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, Node> > PtentTPtent = Xpetra::MatrixMatrix<SC, LO, GO, Node>::Multiply(*Ptent, true, *Ptent, false, out);
+  Teuchos::RCP<Xpetra::Vector<SC, LO, GO, Node> > diagVec     = Xpetra::VectorFactory<SC, LO, GO, Node>::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (TST::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), Teuchos::as<magnitude_type>(diagVec->getGlobalLength()), 1000 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 1000 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+
+}  // UncoupledTentative3D
+
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                      \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation_kokkos, CreateCrsGraphConstant, SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation_kokkos, CreateCrsGraphLinear, SC, LO, GO, NO) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation_kokkos, CreateCrsGraphLinear, SC, LO, GO, NO)   \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(StructuredAggregation_kokkos, UncoupledTentative3D, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/TentativePFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/TentativePFactory_kokkos.cpp
index d4d7586c0dd4..a28abcd7fb10 100644
--- a/packages/muelu/test/unit_tests_kokkos/TentativePFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/TentativePFactory_kokkos.cpp
@@ -61,25 +61,24 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-    out << "version: " << MueLu::Version() << std::endl;
-
-    RCP<TentativePFactory_kokkos> tentPFact = rcp(new TentativePFactory_kokkos);
-    TEST_EQUALITY(tentPFact != Teuchos::null, true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, NO);
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
   }
+  out << "version: " << MueLu::Version() << std::endl;
+
+  RCP<TentativePFactory_kokkos> tentPFact = rcp(new TentativePFactory_kokkos);
+  TEST_EQUALITY(tentPFact != Teuchos::null, true);
+}
 
 #if 0
   TEUCHOS_UNIT_TEST(TentativePFactory, MakeTentative_LapackQR)
   {
-#   include "MueLu_UseShortNames.hpp"
+#include "MueLu_UseShortNames.hpp"
     MUELU_TESTING_SET_OSTREAM;
     MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,NO);
     out << "version: " << MueLu::Version() << std::endl;
@@ -159,296 +158,290 @@ namespace MueLuTests {
   } // MakeTentative  Lapack QR
 #endif
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
-
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
-
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
-
-    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
-
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR with user-supplied nullspace" << std::endl;
-
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentative, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
 
-    fineLevel  .SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
-    coarseLevel.SetFactoryManager(Teuchos::null);
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC,LO,GO,NO>::Build1DPoisson(199);
-    fineLevel.Request("A");
-    fineLevel.Set    ("A", A);
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    // Only one NS vector -> exercises manual orthogonalization
-    LO NSdim = 1;
-    RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    nullSpace->randomize();
-    fineLevel.Set("Nullspace", nullSpace);
+  RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
-    RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR with user-supplied nullspace" << std::endl;
 
-    RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  fineLevel.SetFactoryManager(Teuchos::null);  // factory manager is not used on this test
+  coarseLevel.SetFactoryManager(Teuchos::null);
 
-    RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
+  fineLevel.Request("A");
+  fineLevel.Set("A", A);
 
-    RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
-    TentativePFact->SetFactory("Aggregates",         aggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  // Only one NS vector -> exercises manual orthogonalization
+  LO NSdim                   = 1;
+  RCP<MultiVector> nullSpace = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  nullSpace->randomize();
+  fineLevel.Set("Nullspace", nullSpace);
 
-    coarseLevel.Request("P",            TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace",    TentativePFact.get());
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel, coarseLevel);
+  RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
 
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P", Ptent, TentativePFact.get());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
-    // Check interpolation by computing ||fineNS - P*coarseNS||
-    RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
-    Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+  RCP<CoarseMapFactory> coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    diff->putScalar(0.0);
+  RCP<TentativePFactory_kokkos> TentativePFact = rcp(new TentativePFactory_kokkos());
+  TentativePFact->SetFactory("Aggregates", aggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    coarseLevel.Release("P",         TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace", TentativePFact.get());
+  coarseLevel.Request("P", TentativePFact.get());  // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
 
-    // diff = fineNS - (P*coarseNS)
-    diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
 
-    Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(NSdim);
-    diff->norm2(norms);
-    for (LO i = 0; i < NSdim; ++i) {
-      out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
-      TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
-    }
+  RCP<MultiVector> coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
 
-    // Check normalization and orthogonality of prolongator columns by
-    // making sure that P^T*P = I
-    RCP<Matrix> PtentTPtent = MatrixMatrix::Multiply(*Ptent, true, *Ptent, false, out);
-    RCP<Vector> diagVec     = VectorFactory::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (STS::name().find("complex") == std::string::npos) { //skip check for Scalar=complex
-      TEST_EQUALITY(diagVec->norm1(), diagVec->getGlobalLength());
-      TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100*TMT::eps());
-    }
-    if (STS::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(STS::magnitude(diagVec->meanValue()), TMT::one(), 100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(),   diagVec->getGlobalLength());
-  } // MakeTentative
+  // Check interpolation by computing ||fineNS - P*coarseNS||
+  RCP<MultiVector> PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentativeVectorBasedUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+  RCP<MultiVector> diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+  // diff = fineNS - (P*coarseNS)
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR when nullspace isn't supplied by user" << std::endl;
+  Array<typename Teuchos::ScalarTraits<SC>::magnitudeType> norms(NSdim);
+  diff->norm2(norms);
+  for (LO i = 0; i < NSdim; ++i) {
+    out << "||diff_" << i << "||_2 = " << norms[i] << std::endl;
+    TEST_COMPARE_CONST(norms[i], <, 100 * TMT::eps());
+  }
 
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  // Check normalization and orthogonality of prolongator columns by
+  // making sure that P^T*P = I
+  RCP<Matrix> PtentTPtent = MatrixMatrix::Multiply(*Ptent, true, *Ptent, false, out);
+  RCP<Vector> diagVec     = VectorFactory::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (STS::name().find("complex") == std::string::npos) {  // skip check for Scalar=complex
+    TEST_EQUALITY(diagVec->norm1(), diagVec->getGlobalLength());
+    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  }
+  if (STS::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(STS::magnitude(diagVec->meanValue()), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+}  // MakeTentative
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentativeVectorBasedUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
+
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(200);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR when nullspace isn't supplied by user" << std::endl;
 
-    A->SetFixedBlockSize(2);
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
-    fineLevel.Set("A", A);
+  auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(200);
 
-    auto amalgFact = rcp(new AmalgamationFactory());
+  A->SetFixedBlockSize(2);
 
-    auto dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  fineLevel.Set("A", A);
 
-    auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  auto amalgFact = rcp(new AmalgamationFactory());
 
-    auto coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  auto dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    auto TentativePFact = rcp(new TentativePFactory_kokkos());
-    TentativePFact->SetFactory("Aggregates",         aggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
+  auto coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace", TentativePFact.get());  // request coarse nullspace
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
+  auto TentativePFact = rcp(new TentativePFactory_kokkos());
+  TentativePFact->SetFactory("Aggregates", aggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
+  coarseLevel.Request("P", TentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
 
-    auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
 
-    size_t NSdim = coarseNullspace->getNumVectors();
-    TEST_EQUALITY(NSdim, 2);
+  auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
 
-    //coarseNullspace->describe(out, Teuchos::VERB_EXTREME);
+  size_t NSdim = coarseNullspace->getNumVectors();
+  TEST_EQUALITY(NSdim, 2);
 
-    // Check interpolation by computing ||fineNS - P*coarseNS||
-    auto PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
-    Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
+  // coarseNullspace->describe(out, Teuchos::VERB_EXTREME);
 
-    auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
-    diff->putScalar(0.0);
+  // Check interpolation by computing ||fineNS - P*coarseNS||
+  auto PtN = MultiVectorFactory::Build(Ptent->getRangeMap(), NSdim);
+  Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    coarseLevel.Release("P",         TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace", TentativePFact.get());
+  auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    auto nspFact = Teuchos::rcp(new NullspaceFactory_kokkos());
-    fineLevel.Request("Nullspace",nspFact.get());
+  coarseLevel.Release("P", TentativePFact.get());  // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());
 
-    nspFact->Build(fineLevel);
+  auto nspFact = Teuchos::rcp(new NullspaceFactory_kokkos());
+  fineLevel.Request("Nullspace", nspFact.get());
 
-    auto fineNullspace = fineLevel.Get<RCP<MultiVector> >("Nullspace", nspFact.get());
+  nspFact->Build(fineLevel);
 
-    TEST_EQUALITY(fineNullspace->getNumVectors(), 2);
+  auto fineNullspace = fineLevel.Get<RCP<MultiVector> >("Nullspace", nspFact.get());
 
-    // diff = fineNS - (P*coarseNS)
-    diff->update(1.0, *fineNullspace, -1.0, *PtN, 0.0);
+  TEST_EQUALITY(fineNullspace->getNumVectors(), 2);
 
-    Array<magnitude_type> norms(NSdim);
-    diff->norm2(norms);
-    for (decltype(NSdim) i = 0; i < NSdim; ++i)
-      TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
+  // diff = fineNS - (P*coarseNS)
+  diff->update(1.0, *fineNullspace, -1.0, *PtN, 0.0);
 
-    // Check normalization and orthogonality of prolongator columns by
-    // making sure that P^T*P = I
-    RCP<Matrix> PtentTPtent = MatrixMatrix::Multiply(*Ptent, true, *Ptent, false, out);
-    RCP<Vector> diagVec     = VectorFactory::Build(PtentTPtent->getRowMap());
-    PtentTPtent->getLocalDiagCopy(*diagVec);
-    if (STS::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(diagVec->norm1(), STS::magnitude(diagVec->getGlobalLength()), 100*TMT::eps());
-    TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100*TMT::eps());
-    if (STS::name().find("complex") == std::string::npos) //skip check for Scalar=complex
-      TEST_FLOATING_EQUALITY(STS::magnitude(diagVec->meanValue()), TMT::one(), 100*TMT::eps());
-    TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
-  } // MakeTentativeVectorBasedUsingDefaultNullspace
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentativeUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(SC,GO,NO);
+  Array<magnitude_type> norms(NSdim);
+  diff->norm2(norms);
+  for (decltype(NSdim) i = 0; i < NSdim; ++i)
+    TEST_COMPARE_CONST(norms[i], <, 100 * TMT::eps());
 
-    using STS            = Teuchos::ScalarTraits<Scalar>;
-    using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using TMT            = Teuchos::ScalarTraits<magnitude_type>;
+  // Check normalization and orthogonality of prolongator columns by
+  // making sure that P^T*P = I
+  RCP<Matrix> PtentTPtent = MatrixMatrix::Multiply(*Ptent, true, *Ptent, false, out);
+  RCP<Vector> diagVec     = VectorFactory::Build(PtentTPtent->getRowMap());
+  PtentTPtent->getLocalDiagCopy(*diagVec);
+  if (STS::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(diagVec->norm1(), STS::magnitude(diagVec->getGlobalLength()), 100 * TMT::eps());
+  TEST_FLOATING_EQUALITY(diagVec->normInf(), TMT::one(), 100 * TMT::eps());
+  if (STS::name().find("complex") == std::string::npos)  // skip check for Scalar=complex
+    TEST_FLOATING_EQUALITY(STS::magnitude(diagVec->meanValue()), TMT::one(), 100 * TMT::eps());
+  TEST_EQUALITY(PtentTPtent->getGlobalNumEntries(), diagVec->getGlobalLength());
+}  // MakeTentativeVectorBasedUsingDefaultNullspace
 
-    if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
-      out << "skipping test for linAlgebra==UseEpetra" << std::endl;
-      return;
-    }
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(TentativePFactory_kokkos, MakeTentativeUsingDefaultNullspace, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(SC, GO, NO);
 
-    out << "version: " << MueLu::Version() << std::endl;
-    out << "Test QR when nullspace isn't supplied by user" << std::endl;
+  using STS            = Teuchos::ScalarTraits<Scalar>;
+  using magnitude_type = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using TMT            = Teuchos::ScalarTraits<magnitude_type>;
 
-    Level fineLevel, coarseLevel;
-    TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
+  if (TestHelpers_kokkos::Parameters::getLib() == Xpetra::UseEpetra) {
+    out << "skipping test for linAlgebra==UseEpetra" << std::endl;
+    return;
+  }
 
-    auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
+  out << "version: " << MueLu::Version() << std::endl;
+  out << "Test QR when nullspace isn't supplied by user" << std::endl;
 
-    fineLevel.Set("A", A);
+  Level fineLevel, coarseLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createTwoLevelHierarchy(fineLevel, coarseLevel);
 
+  auto A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(199);
 
-    auto amalgFact = rcp(new AmalgamationFactory());
+  fineLevel.Set("A", A);
 
-    auto dropFact = rcp(new CoalesceDropFactory_kokkos());
-    dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  auto amalgFact = rcp(new AmalgamationFactory());
 
-    auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-    ParameterList aggParams;
-    aggParams.set("aggregation: ordering",              "natural");
-    aggParams.set("aggregation: min agg size",           3);
-    aggParams.set("aggregation: max selected neighbors", 0);
-    aggFact->SetParameterList(aggParams);
-    aggFact->SetFactory("DofsPerNode",  dropFact);
-    aggFact->SetFactory("Graph",        dropFact);
+  auto dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-    auto coarseMapFact = rcp(new CoarseMapFactory());
-    coarseMapFact->SetFactory("Aggregates", aggFact);
+  auto aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  ParameterList aggParams;
+  aggParams.set("aggregation: ordering", "natural");
+  aggParams.set("aggregation: min agg size", 3);
+  aggParams.set("aggregation: max selected neighbors", 0);
+  aggFact->SetParameterList(aggParams);
+  aggFact->SetFactory("DofsPerNode", dropFact);
+  aggFact->SetFactory("Graph", dropFact);
 
-    auto TentativePFact = rcp(new TentativePFactory_kokkos());
-    TentativePFact->SetFactory("Aggregates",         aggFact);
-    TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
-    TentativePFact->SetFactory("CoarseMap",          coarseMapFact);
+  auto coarseMapFact = rcp(new CoarseMapFactory());
+  coarseMapFact->SetFactory("Aggregates", aggFact);
 
+  auto TentativePFact = rcp(new TentativePFactory_kokkos());
+  TentativePFact->SetFactory("Aggregates", aggFact);
+  TentativePFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  TentativePFact->SetFactory("CoarseMap", coarseMapFact);
 
-    coarseLevel.Request("P",TentativePFact.get());  // request Ptent
-    coarseLevel.Request("Nullspace", TentativePFact.get());  // request coarse nullspace
-    coarseLevel.Request(*TentativePFact);
-    TentativePFact->Build(fineLevel,coarseLevel);
+  coarseLevel.Request("P", TentativePFact.get());          // request Ptent
+  coarseLevel.Request("Nullspace", TentativePFact.get());  // request coarse nullspace
+  coarseLevel.Request(*TentativePFact);
+  TentativePFact->Build(fineLevel, coarseLevel);
 
-    RCP<Matrix> Ptent;
-    coarseLevel.Get("P",Ptent,TentativePFact.get());
+  RCP<Matrix> Ptent;
+  coarseLevel.Get("P", Ptent, TentativePFact.get());
 
-    auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace",TentativePFact.get());
+  auto coarseNullspace = coarseLevel.Get<RCP<MultiVector> >("Nullspace", TentativePFact.get());
 
-    coarseLevel.Release("P",TentativePFact.get()); // release Ptent
-    coarseLevel.Release("Nullspace",TentativePFact.get());   // release coarse nullspace
+  coarseLevel.Release("P", TentativePFact.get());          // release Ptent
+  coarseLevel.Release("Nullspace", TentativePFact.get());  // release coarse nullspace
 
-    //grab default fine level nullspace (vector of all ones)
-    auto nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
-    nullSpace->putScalar(1.0);
+  // grab default fine level nullspace (vector of all ones)
+  auto nullSpace = MultiVectorFactory::Build(A->getRowMap(), 1);
+  nullSpace->putScalar(1.0);
 
-    //check interpolation
-    LO NSdim = 1;
-    auto PtN = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    Ptent->apply(*coarseNullspace,*PtN,Teuchos::NO_TRANS,1.0,0.0);
+  // check interpolation
+  LO NSdim = 1;
+  auto PtN = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  Ptent->apply(*coarseNullspace, *PtN, Teuchos::NO_TRANS, 1.0, 0.0);
 
-    auto diff = MultiVectorFactory::Build(A->getRowMap(),NSdim);
-    diff->putScalar(0.0);
+  auto diff = MultiVectorFactory::Build(A->getRowMap(), NSdim);
+  diff->putScalar(0.0);
 
-    //diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
-    diff->update(1.0,*nullSpace,-1.0,*PtN,0.0);
+  // diff = fineNS + (-1.0)*(P*coarseNS) + 0*diff
+  diff->update(1.0, *nullSpace, -1.0, *PtN, 0.0);
 
-    Teuchos::Array<magnitude_type> norms(NSdim);
-    diff->norm2(norms);
-    for (LO i=0; i<NSdim; ++i)
-      TEST_COMPARE_CONST(norms[i], <, 100*TMT::eps());
+  Teuchos::Array<magnitude_type> norms(NSdim);
+  diff->norm2(norms);
+  for (LO i = 0; i < NSdim; ++i)
+    TEST_COMPARE_CONST(norms[i], <, 100 * TMT::eps());
 
-  } //MakeTentativeUsingDefaultNullspace
+}  // MakeTentativeUsingDefaultNullspace
 
 #if 0
   TEUCHOS_UNIT_TEST(TentativePFactory, NonStandardMaps)
@@ -771,14 +764,12 @@ namespace MueLuTests {
 #endif
 #endif
 
-#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, Constructor,   SC, LO, GO, NO) \
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, MakeTentative, SC, LO, GO, NO) \
+#define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, Constructor, SC, LO, GO, NO)                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, MakeTentative, SC, LO, GO, NO)                      \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, MakeTentativeUsingDefaultNullspace, SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(TentativePFactory_kokkos, MakeTentativeVectorBasedUsingDefaultNullspace, SC, LO, GO, NO)
 
-
-
 #include <MueLu_ETI_4arg.hpp>
 
-} // namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/UncoupledAggregationFactory_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/UncoupledAggregationFactory_kokkos.cpp
index bfc847577d77..6dc7fa14a87c 100644
--- a/packages/muelu/test/unit_tests_kokkos/UncoupledAggregationFactory_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/UncoupledAggregationFactory_kokkos.cpp
@@ -62,184 +62,175 @@
 
 namespace MueLuTests {
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-
-      out << "version: " << MueLu::Version() << std::endl;
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      TEST_EQUALITY(aggFact != Teuchos::null, true);
-      TEST_THROW(aggFact->SetOrdering("unknown_ordering"), Teuchos::Exceptions::InvalidParameterValue);
-      aggFact->SetOrdering("natural");
-      TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
-      aggFact->SetOrdering("graph");
-      TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
-      aggFact->SetOrdering("random");
-      TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Constructor, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  TEST_EQUALITY(aggFact != Teuchos::null, true);
+  TEST_THROW(aggFact->SetOrdering("unknown_ordering"), Teuchos::Exceptions::InvalidParameterValue);
+  aggFact->SetOrdering("natural");
+  TEST_EQUALITY(aggFact->GetOrdering() == "natural", true);
+  aggFact->SetOrdering("graph");
+  TEST_EQUALITY(aggFact->GetOrdering() == "graph", true);
+  aggFact->SetOrdering("random");
+  TEST_EQUALITY(aggFact->GetOrdering() == "random", true);
 
-      aggFact->SetMaxNeighAlreadySelected(100);
-      TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 100, true);
-      aggFact->SetMinNodesPerAggregate(10);
-      TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 10, true);
+  aggFact->SetMaxNeighAlreadySelected(100);
+  TEST_EQUALITY(aggFact->GetMaxNeighAlreadySelected() == 100, true);
+  aggFact->SetMinNodesPerAggregate(10);
+  TEST_EQUALITY(aggFact->GetMinNodesPerAggregate() == 10, true);
 
-      auto validParamList = aggFact->GetValidParameterList();
-      TEST_EQUALITY(validParamList != Teuchos::null, true);
+  auto validParamList = aggFact->GetValidParameterList();
+  TEST_EQUALITY(validParamList != Teuchos::null, true);
+
+}  // Constructor
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
 
+  // Make a Matrix with multiple degrees of freedom per node
+  GlobalOrdinal nx = 20, ny = 20;
+
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
+
+  const GO nxx       = 200;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+  Teuchos::CommandLineProcessor clp(false);
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
+
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
+  RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+  A->SetFixedBlockSize(2);
+
+  Level aLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Request("A");
+  aLevel.Set("A", A);
+
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
 
-  } // Constructor
+  aLevel.Request(*aggFact);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      out << "version: " << MueLu::Version() << std::endl;
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+  aggFact->Build(aLevel);
 
-      // Make a Matrix with multiple degrees of freedom per node
-      GlobalOrdinal nx = 20, ny = 20;
+}  // Build
 
-      // Describes the initial layout of matrix rows across processors.
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build_MIS2_Coarsen, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
 
-      const GO nxx = 200;
-      using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-      Teuchos::CommandLineProcessor clp(false);
-      Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-      Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+  // Make a Matrix with multiple degrees of freedom per node
+  GlobalOrdinal nx = 20, ny = 20;
 
-      RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
-      RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
 
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-      A->SetFixedBlockSize(2);
+  const GO nxx       = 200;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+  Teuchos::CommandLineProcessor clp(false);
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
-      Level aLevel;
-      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Request("A");
-      aLevel.Set("A",A);
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
+  RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
 
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+  A->SetFixedBlockSize(2);
 
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      aggFact->SetFactory("Graph", dropFact);
+  Level aLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Request("A");
+  aLevel.Set("A", A);
 
-      aLevel.Request(*aggFact);
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-      aggFact->Build(aLevel);
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: coloring algorithm",
+                        Teuchos::ParameterEntry(std::string("mis2 coarsening")));
 
+  aLevel.Request(*aggFact);
 
-  } // Build
+  aggFact->Build(aLevel);
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build_MIS2_Coarsen, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      out << "version: " << MueLu::Version() << std::endl;
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+}  // Build_MIS2_Coarsen
 
-      // Make a Matrix with multiple degrees of freedom per node
-      GlobalOrdinal nx = 20, ny = 20;
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build_MIS2_Aggregate, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include <MueLu_UseShortNames.hpp>
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+  RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
 
-      // Describes the initial layout of matrix rows across processors.
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
+  // Make a Matrix with multiple degrees of freedom per node
+  GlobalOrdinal nx = 20, ny = 20;
 
-      const GO nxx = 200;
-      using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-      Teuchos::CommandLineProcessor clp(false);
-      Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-      Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
+  // Describes the initial layout of matrix rows across processors.
+  Teuchos::ParameterList galeriList;
+  galeriList.set("nx", nx);
+  galeriList.set("ny", ny);
 
-      RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
-      RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
+  const GO nxx       = 200;
+  using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
+  Teuchos::CommandLineProcessor clp(false);
+  Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748);  // manage parameters of the test case
+  Xpetra::Parameters xpetraParameters(clp);                    // manage parameters of xpetra
 
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-      A->SetFixedBlockSize(2);
+  RCP<Matrix> A      = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
+  RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
 
-      Level aLevel;
-      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Request("A");
-      aLevel.Set("A",A);
+  map = Xpetra::MapFactory<LocalOrdinal, GlobalOrdinal, Node>::Build(map, 2);  // expand map for 2 DOFs per node
+  A->SetFixedBlockSize(2);
 
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
+  Level aLevel;
+  TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
+  aLevel.Request("A");
+  aLevel.Set("A", A);
 
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      aggFact->SetFactory("Graph", dropFact);
-      aggFact->SetParameter("aggregation: coloring algorithm",
-          Teuchos::ParameterEntry(std::string("mis2 coarsening")));
+  RCP<AmalgamationFactory> amalgFact       = rcp(new AmalgamationFactory());
+  RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
+  dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
 
-      aLevel.Request(*aggFact);
+  RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
+  aggFact->SetFactory("Graph", dropFact);
+  aggFact->SetParameter("aggregation: coloring algorithm",
+                        Teuchos::ParameterEntry(std::string("mis2 aggregation")));
 
-      aggFact->Build(aLevel);
+  aLevel.Request(*aggFact);
 
+  aggFact->Build(aLevel);
 
-  } // Build_MIS2_Coarsen
+}  // Build_MIS2_Aggregate
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UncoupledAggregationFactory_kokkos, Build_MIS2_Aggregate, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-    #   include <MueLu_UseShortNames.hpp>
-      MUELU_TESTING_SET_OSTREAM;
-      MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-      out << "version: " << MueLu::Version() << std::endl;
-      RCP<const Teuchos::Comm<int> > comm = TestHelpers_kokkos::Parameters::getDefaultComm();
+#define MUELU_ETI_GROUP(Scalar, LO, GO, Node)                                                                        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Constructor, Scalar, LO, GO, Node)        \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build, Scalar, LO, GO, Node)              \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build_MIS2_Coarsen, Scalar, LO, GO, Node) \
+  TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build_MIS2_Aggregate, Scalar, LO, GO, Node)
 
-      // Make a Matrix with multiple degrees of freedom per node
-      GlobalOrdinal nx = 20, ny = 20;
-
-      // Describes the initial layout of matrix rows across processors.
-      Teuchos::ParameterList galeriList;
-      galeriList.set("nx", nx);
-      galeriList.set("ny", ny);
-
-      const GO nxx = 200;
-      using test_factory = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>;
-      Teuchos::CommandLineProcessor clp(false);
-      Galeri::Xpetra::Parameters<GO> matrixParameters(clp, 8748); // manage parameters of the test case
-      Xpetra::Parameters xpetraParameters(clp);             // manage parameters of xpetra
-
-      RCP<Matrix> A = TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(nxx);
-      RCP<const Map> map = MapFactory::Build(xpetraParameters.GetLib(), matrixParameters.GetNumGlobalElements(), 0, comm);
-
-      map = Xpetra::MapFactory<LocalOrdinal,GlobalOrdinal,Node>::Build(map, 2); //expand map for 2 DOFs per node
-      A->SetFixedBlockSize(2);
-
-      Level aLevel;
-      TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::createSingleLevelHierarchy(aLevel);
-      aLevel.Request("A");
-      aLevel.Set("A",A);
-
-      RCP<AmalgamationFactory> amalgFact = rcp(new AmalgamationFactory());
-      RCP<CoalesceDropFactory_kokkos> dropFact = rcp(new CoalesceDropFactory_kokkos());
-      dropFact->SetFactory("UnAmalgamationInfo", amalgFact);
-
-      RCP<UncoupledAggregationFactory_kokkos> aggFact = rcp(new UncoupledAggregationFactory_kokkos());
-      aggFact->SetFactory("Graph", dropFact);
-      aggFact->SetParameter("aggregation: coloring algorithm",
-          Teuchos::ParameterEntry(std::string("mis2 aggregation")));
-
-      aLevel.Request(*aggFact);
-
-      aggFact->Build(aLevel);
-
-
-  } // Build_MIS2_Aggregate
-
-  # define MUELU_ETI_GROUP(Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Constructor, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build_MIS2_Coarsen, Scalar, LO, GO, Node) \
-    TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UncoupledAggregationFactory_kokkos, Build_MIS2_Aggregate, Scalar, LO, GO, Node)
-
-# include <MueLu_ETI_4arg.hpp>
-
-} // namespace MueLuTests
+#include <MueLu_ETI_4arg.hpp>
 
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/UseKokkos_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/UseKokkos_kokkos.cpp
index 82a9e1fd3915..d3c876b3f636 100644
--- a/packages/muelu/test/unit_tests_kokkos/UseKokkos_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/UseKokkos_kokkos.cpp
@@ -61,45 +61,41 @@
 
 namespace MueLuTests {
 
-
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UseKokkos_kokkos, FactoryManager, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
-#   include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar,GlobalOrdinal,Node);
-    out << "version: " << MueLu::Version() << std::endl;
-
-    using Teuchos::RCP;
-    using Teuchos::rcp;
-    using real_type = typename Teuchos::ScalarTraits<SC>::coordinateType;
-    using RealValuedMultiVector = Xpetra::MultiVector<real_type,LO,GO,NO>;
-
-    const bool useKokkos = false;
-
-    // Build the problem
-    RCP<Matrix> A = MueLuTests::TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2001);
-    RCP<const Map> map = A->getMap();
-    RCP<RealValuedMultiVector> coordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(map, 1);
-    RCP<MultiVector> nullspace = MultiVectorFactory::Build(map, 1);
-    nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
-
-    // Setting paramters for the hierarchy
-    Teuchos::ParameterList paramList;
-    paramList.set("verbosity", "test");
-    paramList.set("use kokkos refactor", useKokkos);
-    RCP<HierarchyManager> mueluFactory = rcp(new ParameterListInterpreter(paramList));
-    mueluFactory->CheckConfig();
-    const RCP<const FactoryManagerBase> manager = mueluFactory->GetFactoryManager(0);
-    RCP<const FactoryManager> factoryManager = Teuchos::rcp_dynamic_cast<const FactoryManager>(manager, true);
-
-    TEST_EQUALITY(factoryManager->GetKokkosRefactor() == useKokkos, true);
-  }
-
-
-#define MUELU_ETI_GROUP(SC,LO,GO,NO) \
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(UseKokkos_kokkos, FactoryManager, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
+#include "MueLu_UseShortNames.hpp"
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
+
+  using Teuchos::RCP;
+  using Teuchos::rcp;
+  using real_type             = typename Teuchos::ScalarTraits<SC>::coordinateType;
+  using RealValuedMultiVector = Xpetra::MultiVector<real_type, LO, GO, NO>;
+
+  const bool useKokkos = false;
+
+  // Build the problem
+  RCP<Matrix> A                          = MueLuTests::TestHelpers_kokkos::TestFactory<SC, LO, GO, NO>::Build1DPoisson(2001);
+  RCP<const Map> map                     = A->getMap();
+  RCP<RealValuedMultiVector> coordinates = Xpetra::MultiVectorFactory<real_type, LO, GO, NO>::Build(map, 1);
+  RCP<MultiVector> nullspace             = MultiVectorFactory::Build(map, 1);
+  nullspace->putScalar(Teuchos::ScalarTraits<SC>::one());
+
+  // Setting paramters for the hierarchy
+  Teuchos::ParameterList paramList;
+  paramList.set("verbosity", "test");
+  paramList.set("use kokkos refactor", useKokkos);
+  RCP<HierarchyManager> mueluFactory = rcp(new ParameterListInterpreter(paramList));
+  mueluFactory->CheckConfig();
+  const RCP<const FactoryManagerBase> manager = mueluFactory->GetFactoryManager(0);
+  RCP<const FactoryManager> factoryManager    = Teuchos::rcp_dynamic_cast<const FactoryManager>(manager, true);
+
+  TEST_EQUALITY(factoryManager->GetKokkosRefactor() == useKokkos, true);
+}
+
+#define MUELU_ETI_GROUP(SC, LO, GO, NO) \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(UseKokkos_kokkos, FactoryManager, SC, LO, GO, NO)
 
 #include <MueLu_ETI_4arg.hpp>
 
-
-} //namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/unit_tests_kokkos/Utilities_kokkos.cpp b/packages/muelu/test/unit_tests_kokkos/Utilities_kokkos.cpp
index cb83946e1a75..8773a52f1cac 100644
--- a/packages/muelu/test/unit_tests_kokkos/Utilities_kokkos.cpp
+++ b/packages/muelu/test/unit_tests_kokkos/Utilities_kokkos.cpp
@@ -56,622 +56,586 @@
 #include "MueLu_Utilities.hpp"
 #include "MueLu_Version.hpp"
 
-namespace MueLuTests
-{
+namespace MueLuTests {
 
-  template <typename SC, typename LO, typename GO, typename NO>
-  void CreateDirichletRow(Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, NO>> &A, LO localRowToZero, SC value = Teuchos::ScalarTraits<SC>::zero())
-  {
-    Teuchos::ArrayView<const LO> indices;
-    Teuchos::ArrayView<const SC> values;
+template <typename SC, typename LO, typename GO, typename NO>
+void CreateDirichletRow(Teuchos::RCP<Xpetra::Matrix<SC, LO, GO, NO>> &A, LO localRowToZero, SC value = Teuchos::ScalarTraits<SC>::zero()) {
+  Teuchos::ArrayView<const LO> indices;
+  Teuchos::ArrayView<const SC> values;
 
-    A->resumeFill();
-    A->getLocalRowView(localRowToZero, indices, values);
-    Array<SC> newValues(values.size(), value);
-
-    for (int j = 0; j < indices.size(); j++)
-    {
-      //keep diagonal
-      if (indices[j] == localRowToZero)
-      {
-        newValues[j] = values[j];
-      }
-    }
+  A->resumeFill();
+  A->getLocalRowView(localRowToZero, indices, values);
+  Array<SC> newValues(values.size(), value);
 
-    A->replaceLocalValues(localRowToZero, indices, newValues);
-    A->fillComplete();
+  for (int j = 0; j < indices.size(); j++) {
+    // keep diagonal
+    if (indices[j] == localRowToZero) {
+      newValues[j] = values[j];
+    }
   }
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, CuthillMcKee, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+  A->replaceLocalValues(localRowToZero, indices, newValues);
+  A->fillComplete();
+}
+
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, CuthillMcKee, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include "MueLu_UseShortNames.hpp"
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
-    out << "version: " << MueLu::Version() << std::endl;
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  out << "version: " << MueLu::Version() << std::endl;
 
-    // Build the problem
-    RCP<Matrix> A = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(100);
+  // Build the problem
+  RCP<Matrix> A = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build1DPoisson(100);
 
-    // CM Test
-    {
-      auto ordering = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CuthillMcKee(*A);
+  // CM Test
+  {
+    auto ordering = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::CuthillMcKee(*A);
 
-      TEST_EQUALITY(ordering->getLocalLength(), A->getLocalNumRows());
-      TEST_EQUALITY(ordering->getGlobalLength(), A->getGlobalNumRows());
-    }
+    TEST_EQUALITY(ordering->getLocalLength(), A->getLocalNumRows());
+    TEST_EQUALITY(ordering->getGlobalLength(), A->getGlobalNumRows());
+  }
 
-    // RCM Test
-    {
-      auto ordering = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReverseCuthillMcKee(*A);
+  // RCM Test
+  {
+    auto ordering = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>::ReverseCuthillMcKee(*A);
 
-      TEST_EQUALITY(ordering->getLocalLength(), A->getLocalNumRows());
-      TEST_EQUALITY(ordering->getGlobalLength(), A->getGlobalNumRows());
-    }
+    TEST_EQUALITY(ordering->getLocalLength(), A->getLocalNumRows());
+    TEST_EQUALITY(ordering->getGlobalLength(), A->getGlobalNumRows());
   }
+}
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    LocalOrdinal localRowToZero = 5;
+  auto A                      = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  LocalOrdinal localRowToZero = 5;
 
-    CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
+  CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
 
-    auto drows = Utils::DetectDirichletRows_kokkos(*A);
-    auto drowsHost = Kokkos::create_mirror_view(drows);
-    Kokkos::deep_copy(drowsHost, drows);
+  auto drows     = Utils::DetectDirichletRows_kokkos(*A);
+  auto drowsHost = Kokkos::create_mirror_view(drows);
+  Kokkos::deep_copy(drowsHost, drows);
 
-    TEST_EQUALITY(drowsHost(localRowToZero), true);
-    TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
+  TEST_EQUALITY(drowsHost(localRowToZero), true);
+  TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
 
-    CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero, Teuchos::as<Scalar>(0.25));
+  CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero, Teuchos::as<Scalar>(0.25));
 
-    // row 5 should not be Dirichlet
-    drows = Utils::DetectDirichletRows_kokkos(*A, TST::magnitude(0.24));
-    drowsHost = Kokkos::create_mirror_view(drows);
-    Kokkos::deep_copy(drowsHost, drows);
+  // row 5 should not be Dirichlet
+  drows     = Utils::DetectDirichletRows_kokkos(*A, TST::magnitude(0.24));
+  drowsHost = Kokkos::create_mirror_view(drows);
+  Kokkos::deep_copy(drowsHost, drows);
 
-    TEST_EQUALITY(drowsHost(localRowToZero), false);
-    TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
+  TEST_EQUALITY(drowsHost(localRowToZero), false);
+  TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
 
-    // row 5 should be Dirichlet
-    drows = Utils::DetectDirichletRows_kokkos(*A, TST::magnitude(0.26), true);
-    drowsHost = Kokkos::create_mirror_view(drows);
-    Kokkos::deep_copy(drowsHost, drows);
+  // row 5 should be Dirichlet
+  drows     = Utils::DetectDirichletRows_kokkos(*A, TST::magnitude(0.26), true);
+  drowsHost = Kokkos::create_mirror_view(drows);
+  Kokkos::deep_copy(drowsHost, drows);
 
-    TEST_EQUALITY(drowsHost(localRowToZero), true);
-    TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
-  } //DetectDirichletRows
+  TEST_EQUALITY(drowsHost(localRowToZero), true);
+  TEST_EQUALITY(drowsHost(localRowToZero - 1), false);
+}  // DetectDirichletRows
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletCols, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletCols, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    LocalOrdinal localRowToZero = 5;
-    CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
+  LocalOrdinal localRowToZero = 5;
+  CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
 
-    auto drows = Utils::DetectDirichletRows_kokkos(*A);
-    auto dcols = Utils::DetectDirichletCols(*A, drows);
+  auto drows = Utils::DetectDirichletRows_kokkos(*A);
+  auto dcols = Utils::DetectDirichletCols(*A, drows);
 
-    auto dcolsHost = Kokkos::create_mirror_view(dcols);
-    Kokkos::deep_copy(dcolsHost, dcols);
+  auto dcolsHost = Kokkos::create_mirror_view(dcols);
+  Kokkos::deep_copy(dcolsHost, dcols);
 
-    for (size_t col = 0; col < dcolsHost.extent(0); ++col)
-    {
-      const auto isDirchletCol = col >= 4 and col <= 6;
-      TEST_EQUALITY(dcolsHost(col), isDirchletCol);
-    }
-  } //DetectDirichletCols
+  for (size_t col = 0; col < dcolsHost.extent(0); ++col) {
+    const auto isDirchletCol = col >= 4 and col <= 6;
+    TEST_EQUALITY(dcolsHost(col), isDirchletCol);
+  }
+}  // DetectDirichletCols
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto diag = Utils::GetMatrixDiagonal(*A);
-    auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
+  auto A        = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto diag     = Utils::GetMatrixDiagonal(*A);
+  auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
 
-    TEST_EQUALITY(diagView.extent(0), A->getLocalNumRows());
+  TEST_EQUALITY(diagView.extent(0), A->getLocalNumRows());
 
-    for (size_t idx = 0; idx < diagView.extent(0); ++idx)
-    {
-      TEST_EQUALITY(diagView(idx, 0), Scalar(2));
-    }
+  for (size_t idx = 0; idx < diagView.extent(0); ++idx) {
+    TEST_EQUALITY(diagView(idx, 0), Scalar(2));
+  }
 
-  } //GetMatrixDiagonal
+}  // GetMatrixDiagonal
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixDiagonalInverse, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixDiagonalInverse, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto diag = Utils::GetMatrixDiagonalInverse(*A);
-    auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
+  auto A        = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto diag     = Utils::GetMatrixDiagonalInverse(*A);
+  auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
 
-    TEST_EQUALITY(diagView.extent(0), A->getLocalNumRows());
+  TEST_EQUALITY(diagView.extent(0), A->getLocalNumRows());
 
-    for (size_t idx = 0; idx < diagView.extent(0); ++idx)
-    {
-      TEST_EQUALITY(diagView(idx, 0), Kokkos::ArithTraits<Scalar>::one() / Scalar(2));
-    }
-  } //GetMatrixDiagonalInverse
+  for (size_t idx = 0; idx < diagView.extent(0); ++idx) {
+    TEST_EQUALITY(diagView(idx, 0), Kokkos::ArithTraits<Scalar>::one() / Scalar(2));
+  }
+}  // GetMatrixDiagonalInverse
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixOverlappedDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, GetMatrixOverlappedDiagonal, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto diag = Utils::GetMatrixOverlappedDiagonal(*A);
-    auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
+  auto A        = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto diag     = Utils::GetMatrixOverlappedDiagonal(*A);
+  auto diagView = diag->getHostLocalView(Xpetra::Access::ReadOnly);
 
-    for (size_t idx = 0; idx < diagView.extent(0); ++idx)
-    {
-      TEST_EQUALITY(diagView(idx, 0), Scalar(2));
-    }
-  } //GetMatrixOverlappedDiagonal
+  for (size_t idx = 0; idx < diagView.extent(0); ++idx) {
+    TEST_EQUALITY(diagView(idx, 0), Scalar(2));
+  }
+}  // GetMatrixOverlappedDiagonal
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, FindNonZeros, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, FindNonZeros, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using RangeType = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using RangeType                = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto map = A->getMap();
-    auto vector = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
-    auto vectorView = vector->getDeviceLocalView(Xpetra::Access::ReadOnly);
+  auto A          = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto map        = A->getMap();
+  auto vector     = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
+  auto vectorView = vector->getDeviceLocalView(Xpetra::Access::ReadOnly);
 
-    Kokkos::View<bool *, typename Node::device_type> nonZeros("", vectorView.extent(0));
-    unsigned int result = 0;
+  Kokkos::View<bool *, typename Node::device_type> nonZeros("", vectorView.extent(0));
+  unsigned int result = 0;
 
-    // Zero-ed out Vector
-    {
-      Utils::FindNonZeros(vector->getDeviceLocalView(Xpetra::Access::ReadOnly), nonZeros);
+  // Zero-ed out Vector
+  {
+    Utils::FindNonZeros(vector->getDeviceLocalView(Xpetra::Access::ReadOnly), nonZeros);
 
-      Kokkos::parallel_reduce(
-          "", RangeType(0, nonZeros.extent(0)),
-          KOKKOS_LAMBDA(const int i, unsigned int &r) { r += static_cast<unsigned int>(nonZeros(i)); }, result);
+    Kokkos::parallel_reduce(
+        "", RangeType(0, nonZeros.extent(0)),
+        KOKKOS_LAMBDA(const int i, unsigned int &r) { r += static_cast<unsigned int>(nonZeros(i)); }, result);
 
-      TEST_EQUALITY(result, 0);
-    }
+    TEST_EQUALITY(result, 0);
+  }
 
-    // Vector filled with ones
-    {
-      vector->putScalar(TST::one());
+  // Vector filled with ones
+  {
+    vector->putScalar(TST::one());
 
-      Utils::FindNonZeros(vector->getDeviceLocalView(Xpetra::Access::ReadOnly), nonZeros);
+    Utils::FindNonZeros(vector->getDeviceLocalView(Xpetra::Access::ReadOnly), nonZeros);
 
-      Kokkos::parallel_reduce(
-          "", RangeType(0, nonZeros.extent(0)),
-          KOKKOS_LAMBDA(const int i, unsigned int &r) { r += static_cast<unsigned int>(nonZeros(i)); }, result);
+    Kokkos::parallel_reduce(
+        "", RangeType(0, nonZeros.extent(0)),
+        KOKKOS_LAMBDA(const int i, unsigned int &r) { r += static_cast<unsigned int>(nonZeros(i)); }, result);
 
-      TEST_EQUALITY(result, nonZeros.extent(0));
-    }
+    TEST_EQUALITY(result, nonZeros.extent(0));
+  }
 
-  } //FindNonZeros
+}  // FindNonZeros
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletColsAndDomains, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, DetectDirichletColsAndDomains, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    LocalOrdinal localRowToZero = 5;
-    CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
+  LocalOrdinal localRowToZero = 5;
+  CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
 
-    auto drows = Utils::DetectDirichletRows_kokkos(*A);
-    Kokkos::View<bool *, typename Node::device_type> dirichletCols("dirichletCols", A->getColMap()->getLocalNumElements());
-    Kokkos::View<bool *, typename Node::device_type> dirichletDomain("dirichletDomain", A->getDomainMap()->getLocalNumElements());
+  auto drows = Utils::DetectDirichletRows_kokkos(*A);
+  Kokkos::View<bool *, typename Node::device_type> dirichletCols("dirichletCols", A->getColMap()->getLocalNumElements());
+  Kokkos::View<bool *, typename Node::device_type> dirichletDomain("dirichletDomain", A->getDomainMap()->getLocalNumElements());
 
-    Utils::DetectDirichletColsAndDomains(*A, drows, dirichletCols, dirichletDomain);
+  Utils::DetectDirichletColsAndDomains(*A, drows, dirichletCols, dirichletDomain);
 
-    auto dCol_host = Kokkos::create_mirror_view(dirichletCols);
-    Kokkos::deep_copy(dCol_host, dirichletCols);
+  auto dCol_host = Kokkos::create_mirror_view(dirichletCols);
+  Kokkos::deep_copy(dCol_host, dirichletCols);
 
-    for (size_t col = 0; col < dCol_host.extent(0); ++col)
-    {
-      const auto isDirchletCol = col >= 4 and col <= 6;
-      TEST_EQUALITY(dCol_host(col), isDirchletCol);
-    }
+  for (size_t col = 0; col < dCol_host.extent(0); ++col) {
+    const auto isDirchletCol = col >= 4 and col <= 6;
+    TEST_EQUALITY(dCol_host(col), isDirchletCol);
+  }
 
-  } //DetectDirichletColsAndDomains
+}  // DetectDirichletColsAndDomains
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ZeroDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ZeroDirichletRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    LocalOrdinal localRowToZero = 5;
-    CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
+  LocalOrdinal localRowToZero = 5;
+  CreateDirichletRow<Scalar, LocalOrdinal, GlobalOrdinal, Node>(A, localRowToZero);
 
-    auto drows = Utils::DetectDirichletRows_kokkos(*A);
-    auto dcols = Utils::DetectDirichletCols(*A, drows);
+  auto drows = Utils::DetectDirichletRows_kokkos(*A);
+  auto dcols = Utils::DetectDirichletCols(*A, drows);
 
-    // Matrix version
-    {
-      const auto zeroVal = Scalar(0.35);
-      Utils::ZeroDirichletRows(A, dcols, zeroVal);
+  // Matrix version
+  {
+    const auto zeroVal = Scalar(0.35);
+    Utils::ZeroDirichletRows(A, dcols, zeroVal);
 
-      Teuchos::ArrayView<const LocalOrdinal> indices;
-      Teuchos::ArrayView<const Scalar> values;
-      A->getLocalRowView(localRowToZero, indices, values);
+    Teuchos::ArrayView<const LocalOrdinal> indices;
+    Teuchos::ArrayView<const Scalar> values;
+    A->getLocalRowView(localRowToZero, indices, values);
 
-      for (int idx = 0; idx < indices.size(); ++idx)
-      {
-        TEST_EQUALITY(values[idx], zeroVal);
-      }
+    for (int idx = 0; idx < indices.size(); ++idx) {
+      TEST_EQUALITY(values[idx], zeroVal);
     }
+  }
 
-    // Multi-vector version
-    {
-      auto map = A->getMap();
-      auto vector = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, A->getGlobalNumCols());
+  // Multi-vector version
+  {
+    auto map    = A->getMap();
+    auto vector = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, A->getGlobalNumCols());
 
-      const auto zeroVal = Scalar(0.25);
-      Utils::ZeroDirichletRows(vector, dcols, zeroVal);
+    const auto zeroVal = Scalar(0.25);
+    Utils::ZeroDirichletRows(vector, dcols, zeroVal);
 
-      auto vecView = vector->getHostLocalView(Xpetra::Access::ReadOnly);
+    auto vecView = vector->getHostLocalView(Xpetra::Access::ReadOnly);
 
-      for (size_t idx = 0; idx < vecView.extent(0); ++idx)
-      {
-        TEST_EQUALITY(vecView(localRowToZero, idx), zeroVal);
-      }
+    for (size_t idx = 0; idx < vecView.extent(0); ++idx) {
+      TEST_EQUALITY(vecView(localRowToZero, idx), zeroVal);
     }
+  }
 
-  } //ZeroDirichletRows
+}  // ZeroDirichletRows
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ZeroDirichletCols, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ZeroDirichletCols, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
-
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using RangeType = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
-
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-
-    const auto numCols = A->getColMap()->getLocalNumElements();
-    const auto colToZero = 2;
-    Kokkos::View<bool *, typename Node::device_type> dCols("", numCols);
-    Kokkos::parallel_for(
-        RangeType(0, numCols), KOKKOS_LAMBDA(const int colIdx) {
-          dCols(colIdx) = colIdx == colToZero;
-        });
-
-    const auto zeroVal = Scalar(0.25);
-    Utils::ZeroDirichletCols(A, dCols, zeroVal);
-
-    const auto localMatrix = A->getLocalMatrixHost();
-    const auto numRows = A->getLocalNumRows();
-    for (size_t row = 0; row < numRows; ++row)
-    {
-      auto rowView = localMatrix.row(row);
-      auto length = rowView.length;
-
-      for (int colID = 0; colID < length; colID++)
-        if (rowView.colidx(colID) == colToZero)
-        {
-          TEST_EQUALITY(rowView.value(colID), zeroVal);
-        }
-    }
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using RangeType                = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+
+  const auto numCols   = A->getColMap()->getLocalNumElements();
+  const auto colToZero = 2;
+  Kokkos::View<bool *, typename Node::device_type> dCols("", numCols);
+  Kokkos::parallel_for(
+      RangeType(0, numCols), KOKKOS_LAMBDA(const int colIdx) {
+        dCols(colIdx) = colIdx == colToZero;
+      });
+
+  const auto zeroVal = Scalar(0.25);
+  Utils::ZeroDirichletCols(A, dCols, zeroVal);
+
+  const auto localMatrix = A->getLocalMatrixHost();
+  const auto numRows     = A->getLocalNumRows();
+  for (size_t row = 0; row < numRows; ++row) {
+    auto rowView = localMatrix.row(row);
+    auto length  = rowView.length;
+
+    for (int colID = 0; colID < length; colID++)
+      if (rowView.colidx(colID) == colToZero) {
+        TEST_EQUALITY(rowView.value(colID), zeroVal);
+      }
+  }
 
-  } //ZeroDirichletCols
+}  // ZeroDirichletCols
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ApplyRowSumCriterion, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ApplyRowSumCriterion, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using Magnitude = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Magnitude                = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    const auto numRows = A->getLocalNumRows();
-    Kokkos::View<bool *, typename Node::device_type> dRows("", numRows);
+  const auto numRows = A->getLocalNumRows();
+  Kokkos::View<bool *, typename Node::device_type> dRows("", numRows);
 
-    Utils::ApplyRowSumCriterion(*A, Magnitude(1.0), dRows);
-    auto dRowsHost = Kokkos::create_mirror_view(dRows);
-    Kokkos::deep_copy(dRowsHost, dRows);
+  Utils::ApplyRowSumCriterion(*A, Magnitude(1.0), dRows);
+  auto dRowsHost = Kokkos::create_mirror_view(dRows);
+  Kokkos::deep_copy(dRowsHost, dRows);
 
-    for (size_t idx = 0; idx < dRows.extent(0); ++idx)
-    {
-      TEST_EQUALITY(dRowsHost(idx), false);
-    }
+  for (size_t idx = 0; idx < dRows.extent(0); ++idx) {
+    TEST_EQUALITY(dRowsHost(idx), false);
+  }
 
-    Utils::ApplyRowSumCriterion(*A, Magnitude(-1.0), dRows);
-    Kokkos::deep_copy(dRowsHost, dRows);
+  Utils::ApplyRowSumCriterion(*A, Magnitude(-1.0), dRows);
+  Kokkos::deep_copy(dRowsHost, dRows);
 
-    for (size_t idx = 0; idx < dRows.extent(0); ++idx)
-    {
-      TEST_EQUALITY(dRowsHost(idx), true);
-    }
-  } //ApplyRowSumCriterion
+  for (size_t idx = 0; idx < dRows.extent(0); ++idx) {
+    TEST_EQUALITY(dRowsHost(idx), true);
+  }
+}  // ApplyRowSumCriterion
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, MyOldScaleMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, MyOldScaleMatrix, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto B = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto B = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    Teuchos::ArrayRCP<Scalar> arr(100, Scalar(2.0));
+  Teuchos::ArrayRCP<Scalar> arr(100, Scalar(2.0));
 
-    Utils::MyOldScaleMatrix(*A, arr, false);
+  Utils::MyOldScaleMatrix(*A, arr, false);
 
-    const auto localMatrixScaled = A->getLocalMatrixHost();
-    const auto localMatrixOriginal = B->getLocalMatrixHost();
-    const auto numRows = A->getLocalNumRows();
-    for (size_t row = 0; row < numRows; ++row)
-    {
-      auto scaledRowView = localMatrixScaled.row(row);
-      auto origRowView = localMatrixOriginal.row(row);
-      auto length = scaledRowView.length;
+  const auto localMatrixScaled   = A->getLocalMatrixHost();
+  const auto localMatrixOriginal = B->getLocalMatrixHost();
+  const auto numRows             = A->getLocalNumRows();
+  for (size_t row = 0; row < numRows; ++row) {
+    auto scaledRowView = localMatrixScaled.row(row);
+    auto origRowView   = localMatrixOriginal.row(row);
+    auto length        = scaledRowView.length;
 
-      for (int colID = 0; colID < length; colID++)
-      {
-        TEST_EQUALITY(scaledRowView.value(colID), Scalar(2.0) * origRowView.value(colID));
-      }
+    for (int colID = 0; colID < length; colID++) {
+      TEST_EQUALITY(scaledRowView.value(colID), Scalar(2.0) * origRowView.value(colID));
     }
+  }
 
 #ifdef HAVE_TPETRA_INST_INT_LONG_LONG
-    TEST_THROW(Utils::MyOldScaleMatrix_Epetra(*A, arr, false, false), MueLu::Exceptions::RuntimeError);
+  TEST_THROW(Utils::MyOldScaleMatrix_Epetra(*A, arr, false, false), MueLu::Exceptions::RuntimeError);
 #endif
 
-  } //MyOldScaleMatrix
+}  // MyOldScaleMatrix
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ApplyOAZToMatrixRows, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, ApplyOAZToMatrixRows, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using RangeType = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using RangeType                = Kokkos::RangePolicy<LocalOrdinal, typename Node::execution_space>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    Kokkos::View<bool *, typename Node::device_type> dRowsIn("", A->getLocalNumRows());
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  Kokkos::View<bool *, typename Node::device_type> dRowsIn("", A->getLocalNumRows());
 
-    Kokkos::parallel_for(
-        "", RangeType(0, dRowsIn.extent(0)), KOKKOS_LAMBDA(const int index) {
-          dRowsIn(index) = index % 2;
-        });
+  Kokkos::parallel_for(
+      "", RangeType(0, dRowsIn.extent(0)), KOKKOS_LAMBDA(const int index) {
+        dRowsIn(index) = index % 2;
+      });
 
-    Utils::ApplyOAZToMatrixRows(A, dRowsIn);
+  Utils::ApplyOAZToMatrixRows(A, dRowsIn);
 
-    auto dRowsOut = Utils::DetectDirichletRows_kokkos(*A);
-    auto dRowsOutHost = Kokkos::create_mirror_view(dRowsOut);
-    Kokkos::deep_copy(dRowsOutHost, dRowsOut);
+  auto dRowsOut     = Utils::DetectDirichletRows_kokkos(*A);
+  auto dRowsOutHost = Kokkos::create_mirror_view(dRowsOut);
+  Kokkos::deep_copy(dRowsOutHost, dRowsOut);
 
-    for (size_t row = 0; row < dRowsOutHost.extent(0); ++row)
-    {
-      TEST_EQUALITY(dRowsOutHost(row), row % 2);
-    }
-  } //ApplyOAZToMatrixRows
+  for (size_t row = 0; row < dRowsOutHost.extent(0); ++row) {
+    TEST_EQUALITY(dRowsOutHost(row), row % 2);
+  }
+}  // ApplyOAZToMatrixRows
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, TransformFunctions, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, TransformFunctions, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using TST = Teuchos::ScalarTraits<Scalar>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TST                      = Teuchos::ScalarTraits<Scalar>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
-    auto map = A->getMap();
-    auto vector = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
-    vector->randomize();
+  auto A      = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto map    = A->getMap();
+  auto vector = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
+  vector->randomize();
 
-    using MV = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using TpetraMV = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MV       = Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TpetraMV = Tpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto compareMV = [&](const MV &mv, const TpetraMV &tpetraMV)
-    {
-      const auto origMV = mv.getVector(0);
-      const auto newMV = tpetraMV.getVector(0);
+  auto compareMV = [&](const MV &mv, const TpetraMV &tpetraMV) {
+    const auto origMV = mv.getVector(0);
+    const auto newMV  = tpetraMV.getVector(0);
 
-      TEST_EQUALITY(origMV->meanValue(), newMV->meanValue());
-      TEST_EQUALITY(mv.getData(0).size(), tpetraMV.getData(0).size());
-    };
+    TEST_EQUALITY(origMV->meanValue(), newMV->meanValue());
+    TEST_EQUALITY(mv.getData(0).size(), tpetraMV.getData(0).size());
+  };
 
-    auto tpetraMV = Utils::MV2TpetraMV(vector);
-    compareMV(*vector, *tpetraMV);
+  auto tpetraMV = Utils::MV2TpetraMV(vector);
+  compareMV(*vector, *tpetraMV);
 
-    auto tpetraMV2 = Utils::MV2TpetraMV(*vector);
-    compareMV(*vector, tpetraMV2);
+  auto tpetraMV2 = Utils::MV2TpetraMV(*vector);
+  compareMV(*vector, tpetraMV2);
 
-    auto nonConstTpetraMV = Utils::MV2NonConstTpetraMV(vector);
-    compareMV(*vector, *nonConstTpetraMV);
+  auto nonConstTpetraMV = Utils::MV2NonConstTpetraMV(vector);
+  compareMV(*vector, *nonConstTpetraMV);
 
-    auto nonConstTpetraMV2 = Utils::MV2NonConstTpetraMV2(*vector);
-    compareMV(*vector, *nonConstTpetraMV2);
+  auto nonConstTpetraMV2 = Utils::MV2NonConstTpetraMV2(*vector);
+  compareMV(*vector, *nonConstTpetraMV2);
 
-    auto nonConstTpetraMV3 = Utils::MV2NonConstTpetraMV(*vector);
-    compareMV(*vector, nonConstTpetraMV3);
+  auto nonConstTpetraMV3 = Utils::MV2NonConstTpetraMV(*vector);
+  compareMV(*vector, nonConstTpetraMV3);
 
-    using TpetraMat = Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using Matrix = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using Operator = Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using CrsMatrixFactory = Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using TpetraMat        = Tpetra::RowMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Matrix           = Xpetra::Matrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Operator         = Xpetra::Operator<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using CrsMatrixFactory = Xpetra::CrsMatrixFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto compareMat = [&](const Matrix &xpetraMat, const TpetraMat &tpetraMat)
-    {
-      TEST_EQUALITY(xpetraMat.getGlobalNumRows(), tpetraMat.getGlobalNumRows());
-      TEST_EQUALITY(xpetraMat.getGlobalNumCols(), tpetraMat.getGlobalNumCols());
-      TEST_EQUALITY(xpetraMat.getLocalNumRows(), tpetraMat.getLocalNumRows());
-      TEST_EQUALITY(xpetraMat.getGlobalNumEntries(), tpetraMat.getGlobalNumEntries());
-    };
+  auto compareMat = [&](const Matrix &xpetraMat, const TpetraMat &tpetraMat) {
+    TEST_EQUALITY(xpetraMat.getGlobalNumRows(), tpetraMat.getGlobalNumRows());
+    TEST_EQUALITY(xpetraMat.getGlobalNumCols(), tpetraMat.getGlobalNumCols());
+    TEST_EQUALITY(xpetraMat.getLocalNumRows(), tpetraMat.getLocalNumRows());
+    TEST_EQUALITY(xpetraMat.getGlobalNumEntries(), tpetraMat.getGlobalNumEntries());
+  };
 
-    auto tpetraCrsMat = Utils::Op2TpetraCrs(A);
-    compareMat(*A, *tpetraCrsMat);
+  auto tpetraCrsMat = Utils::Op2TpetraCrs(A);
+  compareMat(*A, *tpetraCrsMat);
 
-    auto nonConstTpetraCrs = Utils::Op2NonConstTpetraCrs(A);
-    compareMat(*A, *nonConstTpetraCrs);
+  auto nonConstTpetraCrs = Utils::Op2NonConstTpetraCrs(A);
+  compareMat(*A, *nonConstTpetraCrs);
 
-    auto tpetraCrs = Utils::Op2TpetraCrs(*A);
-    compareMat(*A, tpetraCrs);
+  auto tpetraCrs = Utils::Op2TpetraCrs(*A);
+  compareMat(*A, tpetraCrs);
 
-    auto nonConstTpetraCrs2 = Utils::Op2NonConstTpetraCrs(*A);
-    compareMat(*A, nonConstTpetraCrs2);
+  auto nonConstTpetraCrs2 = Utils::Op2NonConstTpetraCrs(*A);
+  compareMat(*A, nonConstTpetraCrs2);
 
-    auto crsMat = CrsMatrixFactory::Build(map);
+  auto crsMat = CrsMatrixFactory::Build(map);
 
-    auto op = Utils::Crs2Op(crsMat);
+  auto op = Utils::Crs2Op(crsMat);
 
-    TEST_EQUALITY(crsMat->getGlobalNumRows(), op->getGlobalNumRows());
-    TEST_EQUALITY(crsMat->getLocalNumRows(), op->getLocalNumRows());
+  TEST_EQUALITY(crsMat->getGlobalNumRows(), op->getGlobalNumRows());
+  TEST_EQUALITY(crsMat->getLocalNumRows(), op->getLocalNumRows());
 
-    auto transposeRes = Utils::Transpose(*A);
-    TEST_EQUALITY(transposeRes->getGlobalNumRows(), A->getGlobalNumRows());
-    TEST_EQUALITY(transposeRes->getLocalNumRows(), A->getLocalNumRows());
+  auto transposeRes = Utils::Transpose(*A);
+  TEST_EQUALITY(transposeRes->getGlobalNumRows(), A->getGlobalNumRows());
+  TEST_EQUALITY(transposeRes->getLocalNumRows(), A->getLocalNumRows());
 
-    auto tpetraRow = Utils::Op2TpetraRow(A);
-    compareMat(*A, *tpetraRow);
+  auto tpetraRow = Utils::Op2TpetraRow(A);
+  compareMat(*A, *tpetraRow);
 
-    auto nonConstTpetraRow = Utils::Op2NonConstTpetraRow(A);
-    compareMat(*A, *nonConstTpetraRow);
+  auto nonConstTpetraRow = Utils::Op2NonConstTpetraRow(A);
+  compareMat(*A, *nonConstTpetraRow);
 
-    auto tpetraMap = Utils::Map2TpetraMap(*map);
-    TEST_INEQUALITY(tpetraMap, Teuchos::null);
-    TEST_EQUALITY_CONST(tpetraMap->getGlobalNumElements(), map->getGlobalNumElements());
+  auto tpetraMap = Utils::Map2TpetraMap(*map);
+  TEST_INEQUALITY(tpetraMap, Teuchos::null);
+  TEST_EQUALITY_CONST(tpetraMap->getGlobalNumElements(), map->getGlobalNumElements());
 
-    using VectorFactory = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MultiVectorFactory = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using VectorFactory      = Xpetra::VectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MultiVectorFactory = Xpetra::MultiVectorFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    // GetInverse
-    {
-      auto vec = Teuchos::RCP<Vector>(VectorFactory::Build(map));
-      vec->putScalar(Scalar(5.0));
+  // GetInverse
+  {
+    auto vec = Teuchos::RCP<Vector>(VectorFactory::Build(map));
+    vec->putScalar(Scalar(5.0));
 
-      auto inv = Utils::GetInverse(vec);
+    auto inv = Utils::GetInverse(vec);
 
-      auto invData = inv->getData(0);
+    auto invData = inv->getData(0);
 
-      TEST_EQUALITY(vec->getData(0).size(), invData.size());
-      for (int i = 0; i < invData.size(); ++i)
-      {
-        TEST_EQUALITY(Scalar(1.0 / 5.0), invData[i]);
-      }
+    TEST_EQUALITY(vec->getData(0).size(), invData.size());
+    for (int i = 0; i < invData.size(); ++i) {
+      TEST_EQUALITY(Scalar(1.0 / 5.0), invData[i]);
     }
+  }
 
-    // ResidualNorm
-    {
-      auto vector2 = MultiVectorFactory::Build(map, 1);
-      vector2->putScalar(Scalar(3.0));
-      vector->putScalar(Scalar(2.0));
+  // ResidualNorm
+  {
+    auto vector2 = MultiVectorFactory::Build(map, 1);
+    vector2->putScalar(Scalar(3.0));
+    vector->putScalar(Scalar(2.0));
 
-      auto residualNormRes = Utils::ResidualNorm((Operator &)(*A), *vector, *vector2);
-      TEST_EQUALITY(residualNormRes.size(), 1);
-    }
+    auto residualNormRes = Utils::ResidualNorm((Operator &)(*A), *vector, *vector2);
+    TEST_EQUALITY(residualNormRes.size(), 1);
+  }
 
-    // PowerMethod
-    {
-      auto powerRes = Utils::PowerMethod(*A);
+  // PowerMethod
+  {
+    auto powerRes = Utils::PowerMethod(*A);
 
-      auto inversDiag = Utils::GetMatrixDiagonalInverse(*A);
-      auto powerRes2 = Utils::PowerMethod(*A, inversDiag);
+    auto inversDiag = Utils::GetMatrixDiagonalInverse(*A);
+    auto powerRes2  = Utils::PowerMethod(*A, inversDiag);
 
-      TEST_INEQUALITY(powerRes, Scalar(0.0));
-      TEST_INEQUALITY(powerRes2, Scalar(0.0));
-    }
-  } //TransformFunctions
+    TEST_INEQUALITY(powerRes, Scalar(0.0));
+    TEST_INEQUALITY(powerRes2, Scalar(0.0));
+  }
+}  // TransformFunctions
 
-  TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, UtilsFunctions, Scalar, LocalOrdinal, GlobalOrdinal, Node)
-  {
+TEUCHOS_UNIT_TEST_TEMPLATE_4_DECL(Utilities_kokkos, UtilsFunctions, Scalar, LocalOrdinal, GlobalOrdinal, Node) {
 #include <MueLu_UseShortNames.hpp>
-    MUELU_TESTING_SET_OSTREAM;
-    MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
+  MUELU_TESTING_SET_OSTREAM;
+  MUELU_TESTING_LIMIT_SCOPE(Scalar, GlobalOrdinal, Node);
 
-    using Magnitude = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
-    using Utils = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
-    using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using Magnitude                = typename Teuchos::ScalarTraits<Scalar>::magnitudeType;
+  using Utils                    = MueLu::Utilities<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
+  using MueLu_TestHelper_Factory = MueLuTests::TestHelpers_kokkos::TestFactory<Scalar, LocalOrdinal, GlobalOrdinal, Node>;
 
-    auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
+  auto A = MueLu_TestHelper_Factory::Build1DPoisson(100);
 
-    ParameterList params;
+  ParameterList params;
 
-    auto coords = Utils::ExtractCoordinatesFromParameterList(params);
-    TEST_EQUALITY(coords, Teuchos::null);
+  auto coords = Utils::ExtractCoordinatesFromParameterList(params);
+  TEST_EQUALITY(coords, Teuchos::null);
 
-    auto map = A->getMap();
-    auto mvCoords = Xpetra::MultiVectorFactory<Magnitude, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
-    params.set("Coordinates", mvCoords);
-    coords = Utils::ExtractCoordinatesFromParameterList(params);
-    TEST_INEQUALITY(coords, Teuchos::null);
+  auto map      = A->getMap();
+  auto mvCoords = Xpetra::MultiVectorFactory<Magnitude, LocalOrdinal, GlobalOrdinal, Node>::Build(map, 1);
+  params.set("Coordinates", mvCoords);
+  coords = Utils::ExtractCoordinatesFromParameterList(params);
+  TEST_INEQUALITY(coords, Teuchos::null);
 
-    auto comm = Parameters::getDefaultComm();
-    TEST_NOTHROW(Utils::SetRandomSeed(*comm));
+  auto comm = Parameters::getDefaultComm();
+  TEST_NOTHROW(Utils::SetRandomSeed(*comm));
 
-    TEST_NOTHROW(Utils::MakeFancy(*(out.getOStream())));
-  } //UtilsFunctions
+  TEST_NOTHROW(Utils::MakeFancy(*(out.getOStream())));
+}  // UtilsFunctions
 
 #define MUELU_ETI_GROUP(SC, LO, GO, NO)                                                                 \
   TEUCHOS_UNIT_TEST_TEMPLATE_4_INSTANT(Utilities_kokkos, CuthillMcKee, SC, LO, GO, NO)                  \
@@ -692,4 +656,4 @@ namespace MueLuTests
 
 #include <MueLu_ETI_4arg.hpp>
 
-} //namespace MueLuTests
+}  // namespace MueLuTests
diff --git a/packages/muelu/test/vardofpernode/VarDofDriver.cpp b/packages/muelu/test/vardofpernode/VarDofDriver.cpp
index 0649d76c0cbb..55aba54e83a5 100644
--- a/packages/muelu/test/vardofpernode/VarDofDriver.cpp
+++ b/packages/muelu/test/vardofpernode/VarDofDriver.cpp
@@ -50,7 +50,7 @@
 #include <Teuchos_XMLParameterListHelpers.hpp>
 
 #include <MueLu.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 #include <MueLu_VisualizationHelpers.hpp>
 
 #include <MueLu_CreateXpetraPreconditioner.hpp>
@@ -60,94 +60,90 @@
 #include <BelosLinearProblem.hpp>
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-  template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-  class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
-  public:
-    ExportVTK() {};
-
-  public:
-
-    void writeFile(std::ofstream& fout, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol)
-    {
-      using namespace std;
-      typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
-      Teuchos::RCP<const Xpetra::Map<LocalOrdinal,GlobalOrdinal,Node> > nodeMap = coordinates->getMap();
-      std::vector<LocalOrdinal> vertices;
-      std::vector<LocalOrdinal> geomSize;
-      LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
-
-      vertices.reserve(numFineNodes);
-      geomSize.reserve(numFineNodes);
-      for(LocalOrdinal i = 0; i < numFineNodes; i++)
-      {
-        vertices.push_back(i);
-        geomSize.push_back(1);
-      }
-
-      Teuchos::ArrayRCP<const double> xCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(0));
-      Teuchos::ArrayRCP<const double> yCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(1));
-      Teuchos::ArrayRCP<const double> zCoords = Teuchos::null;
-      if(coordinates->getNumVectors() == 3) {
-        zCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(2));
-      }
-      Teuchos::ArrayRCP<const double> solData = Teuchos::arcp_reinterpret_cast<const double>(sol->getData(0));
-
-      std::vector<int> uniqueFine = this->makeUnique(vertices);
-      this->writeFileVTKOpening(fout, uniqueFine, geomSize);
-      this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
-
-      std::string indent = "      ";
-      fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      int myRank = coordinates->getMap()->getComm()->getRank();
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        fout << myRank << " ";
-        if(i % 20 == 19)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      // solution vector
-      fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
-      fout << indent;
-      for(int i = 0; i < int(uniqueFine.size()); i++)
-      {
-        size_t numVec = coordinates->getNumVectors();
-        for(int d=0; d<numVec; d++)
-          fout << solData[i*numVec+d] << " ";
-        if(i % 3 == 0)
-          fout << std::endl << indent;
-      }
-      fout << std::endl;
-      fout << "        </DataArray>" << std::endl;
-      fout << "      </PointData>" << std::endl; // that is annoying
-
-      this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
-      this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
-      this->writeFileVTKClosing(fout);
-      fout.close();
-
-    };
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+class ExportVTK : public MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> {
+ public:
+  ExportVTK(){};
+
+ public:
+  void writeFile(std::ofstream& fout, Teuchos::RCP<Xpetra::MultiVector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& coordinates, Teuchos::RCP<Xpetra::Vector<Scalar, LocalOrdinal, GlobalOrdinal, Node> >& sol) {
+    using namespace std;
+    typedef MueLu::VisualizationHelpers<Scalar, LocalOrdinal, GlobalOrdinal, Node> VH;
+    Teuchos::RCP<const Xpetra::Map<LocalOrdinal, GlobalOrdinal, Node> > nodeMap = coordinates->getMap();
+    std::vector<LocalOrdinal> vertices;
+    std::vector<LocalOrdinal> geomSize;
+    LocalOrdinal numFineNodes = Teuchos::as<LocalOrdinal>(coordinates->getLocalLength());
+
+    vertices.reserve(numFineNodes);
+    geomSize.reserve(numFineNodes);
+    for (LocalOrdinal i = 0; i < numFineNodes; i++) {
+      vertices.push_back(i);
+      geomSize.push_back(1);
+    }
 
+    Teuchos::ArrayRCP<const double> xCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(0));
+    Teuchos::ArrayRCP<const double> yCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(1));
+    Teuchos::ArrayRCP<const double> zCoords = Teuchos::null;
+    if (coordinates->getNumVectors() == 3) {
+      zCoords = Teuchos::arcp_reinterpret_cast<const double>(coordinates->getData(2));
+    }
+    Teuchos::ArrayRCP<const double> solData = Teuchos::arcp_reinterpret_cast<const double>(sol->getData(0));
+
+    std::vector<int> uniqueFine = this->makeUnique(vertices);
+    this->writeFileVTKOpening(fout, uniqueFine, geomSize);
+    this->writeFileVTKNodes(fout, uniqueFine, nodeMap);
+
+    std::string indent = "      ";
+    fout << "        <DataArray type=\"Int32\" Name=\"Processor\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    int myRank = coordinates->getMap()->getComm()->getRank();
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      fout << myRank << " ";
+      if (i % 20 == 19)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    // solution vector
+    fout << "        <DataArray type=\"Float64\" Name=\"Solution\" NumberOfComponents=\"3\" format=\"ascii\">" << std::endl;
+    fout << indent;
+    for (int i = 0; i < int(uniqueFine.size()); i++) {
+      size_t numVec = coordinates->getNumVectors();
+      for (int d = 0; d < numVec; d++)
+        fout << solData[i * numVec + d] << " ";
+      if (i % 3 == 0)
+        fout << std::endl
+             << indent;
+    }
+    fout << std::endl;
+    fout << "        </DataArray>" << std::endl;
+    fout << "      </PointData>" << std::endl;  // that is annoying
+
+    this->writeFileVTKCoordinates(fout, uniqueFine, xCoords, yCoords, zCoords, coordinates->getNumVectors());
+    this->writeFileVTKCells(fout, uniqueFine, vertices, geomSize);
+    this->writeFileVTKClosing(fout);
+    fout.close();
   };
+};
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
-  using Teuchos::RCP; using Teuchos::rcp;
+  using Teuchos::RCP;
+  using Teuchos::rcp;
   using Teuchos::TimeMonitor;
 
   bool success = true;
   bool verbose = true;
 
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -156,24 +152,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream& out       = *fancy;
     out.setOutputToRootOnly(0);
 
     // =========================================================================
     // Parameters initialization
     // =========================================================================
 
-    //GO nx = 100, ny = 100, nz = 100;
-    //Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
+    // GO nx = 100, ny = 100, nz = 100;
+    // Galeri::Xpetra::Parameters<GO> matrixParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);  // manage parameters of Xpetra
 
-    std::string xmlFileName = "driver.xml";      clp.setOption("xml",                   &xmlFileName,     "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
+    std::string xmlFileName = "driver.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file. Otherwise, this example uses by default 'scalingTest.xml'");
 
-    switch (clp.parse(argc,argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+    switch (clp.parse(argc, argv)) {
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     // =========================================================================
@@ -181,203 +178,200 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     // =========================================================================
     RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("MatrixRead: S - Global Time"))), tm;
 
-
     GlobalOrdinal nGlobalNodes = 0;
-    if(comm->getRank() == 0) {
+    if (comm->getRank() == 0) {
       std::ifstream data_file;
       data_file.open("xxx.mm");
-      TEUCHOS_TEST_FOR_EXCEPTION(data_file.good() == false, MueLu::Exceptions::RuntimeError,"Problem opening file xxx.mm");
+      TEUCHOS_TEST_FOR_EXCEPTION(data_file.good() == false, MueLu::Exceptions::RuntimeError, "Problem opening file xxx.mm");
       std::string line;
-      getline(data_file,line);
+      getline(data_file, line);
       data_file >> nGlobalNodes;
       data_file.close();
     }
-    Teuchos::broadcast(*comm,0,1,&nGlobalNodes);
+    Teuchos::broadcast(*comm, 0, 1, &nGlobalNodes);
 
     out << "Found " << nGlobalNodes << " nodes " << std::endl;
 
     // each processor reads in its dofPresent array
     LocalOrdinal nNodes = 0;
     LocalOrdinal nDofs  = 0;
-    int maxDofPerNode = -1;
+    int maxDofPerNode   = -1;
     Teuchos::ArrayRCP<LocalOrdinal> dofPresent;
     {
-      std::stringstream ss; ss << comm->getSize() << "dofPresent" << comm->getRank();
+      std::stringstream ss;
+      ss << comm->getSize() << "dofPresent" << comm->getRank();
 
-      try{
+      try {
         std::ifstream data_file(ss.str());
         // read in first line containing number of local nodes and maxDofPerNode
         data_file >> nNodes >> maxDofPerNode;
-        dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(nNodes * maxDofPerNode,0);
+        dofPresent = Teuchos::ArrayRCP<LocalOrdinal>(nNodes * maxDofPerNode, 0);
         // loop over all local nodes
-        for(LocalOrdinal i = 0; i < nNodes; i++) {
-          for(int j=0; j<maxDofPerNode; j++) {
+        for (LocalOrdinal i = 0; i < nNodes; i++) {
+          for (int j = 0; j < maxDofPerNode; j++) {
             int tmp = -1;
             data_file >> tmp;
-            if(tmp == 1) {
-              dofPresent[i*maxDofPerNode+j] = 1; nDofs++;
+            if (tmp == 1) {
+              dofPresent[i * maxDofPerNode + j] = 1;
+              nDofs++;
             } else {
-              dofPresent[i*maxDofPerNode+j] = 0;
+              dofPresent[i * maxDofPerNode + j] = 0;
             }
           }
         }
+      } catch (const std::exception&) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception&) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-
     }
 
     out << "PROC " << comm->getRank() << "/" << comm->getSize() << " " << nNodes << " " << maxDofPerNode << std::endl;
 
-
     /*for(LocalOrdinal i = 0; i < nNodes; i++) {
       for(int j=0; j<maxDofPerNode; j++) {
         std::cout << i << " " << dofPresent[i*maxDofPerNode+j] << std::endl;
       }
     }*/
 
-
     // read in dofs
 
     Teuchos::Array<GlobalOrdinal> dofGlobals;
     Teuchos::Array<GlobalOrdinal> nodalGlobals;  // nodal GIDs for laplacian (with holes)
     {
-      std::stringstream ss; ss << comm->getSize() << "proc" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "proc" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
 
         // read in first line containing number of local nodes and maxDofPerNode
         LocalOrdinal tmpDofs = 0;
         data_file >> tmpDofs;
-        TEUCHOS_TEST_FOR_EXCEPTION(tmpDofs != nDofs, MueLu::Exceptions::RuntimeError,"Number of gid entries in map file is " << tmpDofs << " but should be " << nDofs);
+        TEUCHOS_TEST_FOR_EXCEPTION(tmpDofs != nDofs, MueLu::Exceptions::RuntimeError, "Number of gid entries in map file is " << tmpDofs << " but should be " << nDofs);
 
         dofGlobals = Teuchos::Array<GlobalOrdinal>(nDofs);
 
         // loop over all local nodes
-        for(GlobalOrdinal i = 0; i < nDofs; i++) {
+        for (GlobalOrdinal i = 0; i < nDofs; i++) {
           int data;
           data_file >> data;
           dofGlobals[i] = Teuchos::as<GlobalOrdinal>(data);
         }
-      }
-      catch(const std::exception&) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
+      } catch (const std::exception&) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
 
       nodalGlobals = Teuchos::Array<GlobalOrdinal>(nNodes);
 
       GlobalOrdinal count = nDofs - 1;
-      for(GlobalOrdinal i = nNodes - 1; i>=0; i--) {
-        for(int j = maxDofPerNode-1; j >=0; j--) {
-          if(dofPresent[i*maxDofPerNode+j] == 1)
+      for (GlobalOrdinal i = nNodes - 1; i >= 0; i--) {
+        for (int j = maxDofPerNode - 1; j >= 0; j--) {
+          if (dofPresent[i * maxDofPerNode + j] == 1)
             nodalGlobals[i] = dofGlobals[count--];
         }
       }
     }
 
-    Teuchos::RCP<Map> LapMap = MapFactory::Build(lib,nGlobalNodes,nodalGlobals(),0,comm);
+    Teuchos::RCP<Map> LapMap = MapFactory::Build(lib, nGlobalNodes, nodalGlobals(), 0, comm);
 
-    //LapMap->describe(out, Teuchos::VERB_EXTREME);
+    // LapMap->describe(out, Teuchos::VERB_EXTREME);
 
     //
 
     {
-      std::stringstream ss; ss << comm->getSize() << "ProcLinear" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "ProcLinear" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
 
         // loop over all local nodes
-        for(GlobalOrdinal i = 0; i < nNodes; i++) {
+        for (GlobalOrdinal i = 0; i < nNodes; i++) {
           int data;
           data_file >> data;
           nodalGlobals[i] = Teuchos::as<GlobalOrdinal>(data);
         }
+      } catch (const std::exception&) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception&) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-      for(GlobalOrdinal i = 0; i < nNodes; i++) {
+      for (GlobalOrdinal i = 0; i < nNodes; i++) {
         nodalGlobals[i] = nodalGlobals[i] - 1;
       }
     }
 
-    Teuchos::RCP<Map> LinearMap = MapFactory::Build(lib,nGlobalNodes,nodalGlobals(),0,comm);
-    //LinearMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
+    Teuchos::RCP<Map> LinearMap = MapFactory::Build(lib, nGlobalNodes, nodalGlobals(), 0, comm);
+    // LinearMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
 
-    //LapMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
+    // LapMap->describe(out, Teuchos::VERB_EXTREME);LapMap->describe(out, Teuchos::VERB_EXTREME);
 
     Teuchos::RCP<Import> Importer = ImportFactory::Build(LinearMap, LapMap);
 
-    Teuchos::RCP<MultiVector> xpetraXXX = MultiVectorFactory::Build(LapMap,1);
-    Teuchos::RCP<MultiVector> xpetraYYY = MultiVectorFactory::Build(LapMap,1);
+    Teuchos::RCP<MultiVector> xpetraXXX = MultiVectorFactory::Build(LapMap, 1);
+    Teuchos::RCP<MultiVector> xpetraYYY = MultiVectorFactory::Build(LapMap, 1);
 
-    RCP<MultiVector> temp = IO::ReadMultiVector ("xxx.mm", LinearMap);
+    RCP<MultiVector> temp = IO::ReadMultiVector("xxx.mm", LinearMap);
     xpetraXXX->doImport(*temp, *Importer, Xpetra::INSERT);
-    temp = IO::ReadMultiVector ("yyy.mm", LinearMap);
+    temp = IO::ReadMultiVector("yyy.mm", LinearMap);
     xpetraYYY->doImport(*temp, *Importer, Xpetra::INSERT);
 
-    Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(LapMap,2);
+    Teuchos::RCP<MultiVector> coordinates = MultiVectorFactory::Build(LapMap, 2);
     // GH: scope data manipulation because we should not let a pointer to this data live once we call MueLu
-    {  
-      Teuchos::ArrayRCP< const Scalar > srcX = xpetraXXX->getData(0);
-      Teuchos::ArrayRCP< const Scalar > srcY = xpetraYYY->getData(0);
-      Teuchos::ArrayRCP< Scalar > dataX = coordinates->getDataNonConst(0);
-      Teuchos::ArrayRCP< Scalar > dataY = coordinates->getDataNonConst(1);
-      for(decltype(coordinates->getLocalLength()) i = 0; i < coordinates->getLocalLength(); i++) {
+    {
+      Teuchos::ArrayRCP<const Scalar> srcX = xpetraXXX->getData(0);
+      Teuchos::ArrayRCP<const Scalar> srcY = xpetraYYY->getData(0);
+      Teuchos::ArrayRCP<Scalar> dataX      = coordinates->getDataNonConst(0);
+      Teuchos::ArrayRCP<Scalar> dataY      = coordinates->getDataNonConst(1);
+      for (decltype(coordinates->getLocalLength()) i = 0; i < coordinates->getLocalLength(); i++) {
         dataX[i] = srcX[i];
         dataY[i] = srcY[i];
       }
     }
 
     // read in matrix
-    Teuchos::RCP<const Map>    DistributedMap  = Teuchos::null;
-    Teuchos::RCP<Matrix>       DistributedMatrix = Teuchos::null;
+    Teuchos::RCP<const Map> DistributedMap = Teuchos::null;
+    Teuchos::RCP<Matrix> DistributedMatrix = Teuchos::null;
 
     // read in matrix to determine number of rows
-    Teuchos::RCP<Matrix> SerialMatrix = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("theMatrix.m",lib,comm);
-    GlobalOrdinal NumRows = SerialMatrix->getRowMap()->getGlobalNumElements();
+    Teuchos::RCP<Matrix> SerialMatrix = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("theMatrix.m", lib, comm);
+    GlobalOrdinal NumRows             = SerialMatrix->getRowMap()->getGlobalNumElements();
 
     // re-read in matrix and distribute it using the user-given distribution over processors
-    DistributedMap    = MapFactory::Build(lib,NumRows,dofGlobals(),0,comm);
-    DistributedMatrix = Xpetra::IO<Scalar,LocalOrdinal,GlobalOrdinal,Node>::Read("theMatrix.m",
-              DistributedMap,
-              Teuchos::null,
-              DistributedMap,
-              DistributedMap,
-              true,   //           callFillComplete = true,
-              false,  //           binary           = false,
-              false,  //           tolerant         = false,
-              false); //           debug            = false)
+    DistributedMap    = MapFactory::Build(lib, NumRows, dofGlobals(), 0, comm);
+    DistributedMatrix = Xpetra::IO<Scalar, LocalOrdinal, GlobalOrdinal, Node>::Read("theMatrix.m",
+                                                                                    DistributedMap,
+                                                                                    Teuchos::null,
+                                                                                    DistributedMap,
+                                                                                    DistributedMap,
+                                                                                    true,    //           callFillComplete = true,
+                                                                                    false,   //           binary           = false,
+                                                                                    false,   //           tolerant         = false,
+                                                                                    false);  //           debug            = false)
 
     // read in global vectors (e.g. rhs)
     GlobalOrdinal nGlobalDof = 0;
     GlobalOrdinal nLocalDofs = Teuchos::as<GlobalOrdinal>(nDofs);
 
-    Teuchos::reduceAll(*comm,Teuchos::REDUCE_SUM,1,&nLocalDofs,&nGlobalDof);
+    Teuchos::reduceAll(*comm, Teuchos::REDUCE_SUM, 1, &nLocalDofs, &nGlobalDof);
 
     Teuchos::RCP<const Map> dofLinearMap = Teuchos::null;
     {
-      std::stringstream ss; ss << comm->getSize() << "ProcLinear" << comm->getRank();
-      try{
+      std::stringstream ss;
+      ss << comm->getSize() << "ProcLinear" << comm->getRank();
+      try {
         std::ifstream data_file(ss.str());
         // loop over all local nodes
         data_file >> dofGlobals[0];
+      } catch (const std::exception&) {
+        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError, "Problem opening/reading file " << ss.str());
       }
-      catch(const std::exception&) {
-        TEUCHOS_TEST_FOR_EXCEPTION(true, MueLu::Exceptions::RuntimeError,"Problem opening/reading file " << ss.str());
-      }
-      for(decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = i + dofGlobals[0];
-      for(decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = dofGlobals[i] - 1;
-      dofLinearMap = MapFactory::Build(lib,nGlobalDof,dofGlobals(),0,comm);
+      for (decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = i + dofGlobals[0];
+      for (decltype(nDofs) i = 0; i < nDofs; i++) dofGlobals[i] = dofGlobals[i] - 1;
+      dofLinearMap = MapFactory::Build(lib, nGlobalDof, dofGlobals(), 0, comm);
     }
 
     Teuchos::RCP<Import> dofImporter = ImportFactory::Build(dofLinearMap, DistributedMap);
 
-    RCP<MultiVector> RHS  = MultiVectorFactory::Build(DistributedMap,1);
-    RCP<MultiVector> LHS  = MultiVectorFactory::Build(DistributedMap,1);
+    RCP<MultiVector> RHS = MultiVectorFactory::Build(DistributedMap, 1);
+    RCP<MultiVector> LHS = MultiVectorFactory::Build(DistributedMap, 1);
 
-    temp = IO::ReadMultiVector ("rhs.mm", dofLinearMap);
+    temp = IO::ReadMultiVector("rhs.mm", dofLinearMap);
     RHS->doImport(*temp, *dofImporter, Xpetra::INSERT);
     LHS->putScalar(Teuchos::ScalarTraits<Scalar>::zero());
 
@@ -386,9 +380,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     Teuchos::ParameterList paramList;
     Teuchos::updateParametersFromXmlFileAndBroadcast(xmlFileName, Teuchos::Ptr<Teuchos::ParameterList>(&paramList), *comm);
 
-    const std::string userName = "user data";
+    const std::string userName            = "user data";
     Teuchos::ParameterList& userParamList = paramList.sublist(userName);
-    userParamList.set("multivector Coordinates",coordinates);
+    userParamList.set("multivector Coordinates", coordinates);
     userParamList.set("ArrayRCP<LO> DofPresent", dofPresent);
 
     RCP<Hierarchy> H;
@@ -398,15 +392,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
     {
       tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("Driver: 4 - Belos Solve")));
       // Operator and Multivector type that will be used with Belos
-      typedef MultiVector          MV;
+      typedef MultiVector MV;
       typedef Belos::OperatorT<MV> OP;
       H->IsPreconditioner(true);
 
       // Define Operator and Preconditioner
-      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(DistributedMatrix)); // Turns a Xpetra::Matrix object into a Belos operator
-      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));  // Turns a MueLu::Hierarchy object into a Belos operator
+      Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(DistributedMatrix));  // Turns a Xpetra::Matrix object into a Belos operator
+      Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));                   // Turns a MueLu::Hierarchy object into a Belos operator
       // Construct a Belos LinearProblem object
-      RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, LHS, RHS));
+      RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, LHS, RHS));
       belosProblem->setRightPrec(belosPrec);
       bool set = belosProblem->setProblem();
       if (set == false) {
@@ -416,14 +410,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
 
       // Belos parameter list
       Teuchos::ParameterList belosList;
-      belosList.set("Maximum Iterations",   60); // Maximum number of iterations allowed
-      belosList.set("Convergence Tolerance", 1e-5);    // Relative convergence tolerance requested
-      belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-      belosList.set("Output Frequency",      1);
-      belosList.set("Output Style",          Belos::Brief);
+      belosList.set("Maximum Iterations", 60);       // Maximum number of iterations allowed
+      belosList.set("Convergence Tolerance", 1e-5);  // Relative convergence tolerance requested
+      belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+      belosList.set("Output Frequency", 1);
+      belosList.set("Output Style", Belos::Brief);
 
       // Create an iterative solver manager
-      RCP< Belos::SolverManager<SC, MV, OP> > solver;
+      RCP<Belos::SolverManager<SC, MV, OP> > solver;
       solver = rcp(new Belos::PseudoBlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
 
       // Perform solve
@@ -434,31 +428,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib lib, int arg
         // Get the number of iterations for this solve.
         out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
-      } catch(...) {
-        out << std::endl << "ERROR:  Belos threw an error! " << std::endl;
+      } catch (...) {
+        out << std::endl
+            << "ERROR:  Belos threw an error! " << std::endl;
         success = false;
       }
       // Check convergence
       if (ret != Belos::Converged) {
-        out << std::endl << "ERROR:  Belos did not converge or a warning occured! " << std::endl;
+        out << std::endl
+            << "ERROR:  Belos did not converge or a warning occured! " << std::endl;
         success = false;
-      }
-      else
-        out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
+      } else
+        out << std::endl
+            << "SUCCESS:  Belos converged!" << std::endl;
     }
 #endif
-
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
-} //main
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
+}  // main
 
 int main(int argc, char* argv[]) {
   bool success = false;
   bool verbose = true;
 
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
   try {
     const bool throwExceptions     = false;
@@ -467,20 +462,21 @@ int main(int argc, char* argv[]) {
     Teuchos::CommandLineProcessor clp(throwExceptions, recogniseAllOptions);
     Xpetra::Parameters xpetraParameters(clp);
 
-    std::string node = "";  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+    std::string node = "";
+    clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
 
     switch (clp.parse(argc, argv, NULL)) {
-      case Teuchos::CommandLineProcessor::PARSE_ERROR:               return EXIT_FAILURE;
+      case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
 
     if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      return main_<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
+      return main_<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
 #else
       throw MueLu::Exceptions::RuntimeError("Epetra is not available");
 #endif
@@ -491,35 +487,35 @@ int main(int argc, char* argv[]) {
         typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
 #ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
 #else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#  else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#  endif
+#endif
 #endif
       } else if (node == "serial") {
 #ifdef KOKKOS_HAVE_SERIAL
         typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("Serial node type is disabled");
 #endif
@@ -527,19 +523,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_HAVE_OPENMP
         typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("OpenMP node type is disabled");
 #endif
@@ -547,19 +543,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_HAVE_CUDA
         typedef Tpetra::KokkosCompat::KokkosCudaWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("CUDA node type is disabled");
 #endif
@@ -567,19 +563,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_HAVE_HIP
         typedef Tpetra::KokkosCompat::KokkosHIPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("HIP node type is disabled");
 #endif
@@ -588,19 +584,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_HAVE_SYCL
         typedef Tpetra::KokkosCompat::KokkosSYCLWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("SYCL node type is disabled");
 #endif
@@ -611,6 +607,5 @@ int main(int argc, char* argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
-
diff --git a/packages/muelu/test/viz/AggExport.cpp b/packages/muelu/test/viz/AggExport.cpp
index 76d1d24b54e3..f22394a50f53 100644
--- a/packages/muelu/test/viz/AggExport.cpp
+++ b/packages/muelu/test/viz/AggExport.cpp
@@ -64,28 +64,28 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
 #include <MueLu_MutuallyExclusiveTime.hpp>
 
-bool compare_to_gold_all_ranks(int myRank, const std::string & baseFile) {
-  bool failed=false;
+bool compare_to_gold_all_ranks(int myRank, const std::string& baseFile) {
+  bool failed = false;
 
   // Create a copy of outputs
   std::string cmd = "cp -f ";
   int retVal;
   retVal = system((cmd + baseFile + ".gold " + baseFile + ".gold_filtered").c_str());
   TEUCHOS_TEST_FOR_EXCEPTION(retVal != 0, std::runtime_error, "goldfile copy failed!");
-  //system((cmd + baseFile + ".out "  + baseFile + ".out_filtered").c_str());
-  retVal = system((cmd + baseFile + ".vtu "  + baseFile + ".out_filtered").c_str());
+  // system((cmd + baseFile + ".out "  + baseFile + ".out_filtered").c_str());
+  retVal = system((cmd + baseFile + ".vtu " + baseFile + ".out_filtered").c_str());
   TEUCHOS_TEST_FOR_EXCEPTION(retVal != 0, std::runtime_error, "goldfile copy failed!");
 
   // Run comparison (ignoring whitespaces)
-  cmd = "diff -u -w -I\"^\\s*$\" " + baseFile + ".gold_filtered " + baseFile + ".out_filtered";
+  cmd     = "diff -u -w -I\"^\\s*$\" " + baseFile + ".gold_filtered " + baseFile + ".out_filtered";
   int ret = system(cmd.c_str());
-  std::cout<<ret<<std::endl;
+  std::cout << ret << std::endl;
   if (ret)
     failed = true;
 
@@ -94,9 +94,8 @@ bool compare_to_gold_all_ranks(int myRank, const std::string & baseFile) {
   return !failed;
 }
 
-
 std::string replaceAll(std::string result, const std::string& replaceWhat, const std::string& replaceWithWhat) {
-  while(1) {
+  while (1) {
     const int pos = result.find(replaceWhat);
     if (pos == -1)
       break;
@@ -105,25 +104,24 @@ std::string replaceAll(std::string result, const std::string& replaceWhat, const
   return result;
 }
 
-
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
   // =========================================================================
-  bool success = false;
+  bool success  = false;
   bool aggMatch = false;
-  bool verbose = true;
+  bool verbose  = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -135,24 +133,32 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName       = "viztest.xml";     clp.setOption("xml",                   &xmlFileName,       "read parameters from a file");
-
-    std::string mapFile;                               clp.setOption("map",                   &mapFile,           "map data file");
-    std::string colMapFile;                            clp.setOption("colmap",                &colMapFile,        "colmap data file");
-    std::string domainMapFile;                         clp.setOption("domainmap",             &domainMapFile,     "domainmap data file");
-    std::string rangeMapFile;                          clp.setOption("rangemap",              &rangeMapFile,      "rangemap data file");
-    std::string matrixFile;                            clp.setOption("matrix",                &matrixFile,        "matrix data file");
-    std::string coordFile;                             clp.setOption("coords",                &coordFile,         "coordinates data file");
-    std::string nullFile;                              clp.setOption("nullspace",             &nullFile,          "nullspace data file");
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+    std::string xmlFileName = "viztest.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file");
+
+    std::string mapFile;
+    clp.setOption("map", &mapFile, "map data file");
+    std::string colMapFile;
+    clp.setOption("colmap", &colMapFile, "colmap data file");
+    std::string domainMapFile;
+    clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+    std::string rangeMapFile;
+    clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+    std::string matrixFile;
+    clp.setOption("matrix", &matrixFile, "matrix data file");
+    std::string coordFile;
+    clp.setOption("coords", &coordFile, "coordinates data file");
+    std::string nullFile;
+    clp.setOption("nullspace", &nullFile, "nullspace data file");
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     ParameterList paramList;
@@ -170,13 +176,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
     RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
-    RCP<Matrix>      A;
-    RCP<const Map>   map;
+    RCP<Matrix> A;
+    RCP<const Map> map;
     RCP<MultiVector> coordinates;
     RCP<MultiVector> nullspace;
     int ndims = 2;
     if (matrixFile.empty()) {
-      galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+      galeriStream << "========================================================\n"
+                   << xpetraParameters << galeriParameters;
 
       // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
       //                                 d1  d2  d3
@@ -193,25 +200,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
       // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
       if (matrixType == "Laplace1D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", map, galeriList);
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, galeriList);
 
       } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
-          matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D", map, galeriList);
+                 matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, galeriList);
 
       } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D", map, galeriList);
-        ndims = 3;
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, galeriList);
+        ndims       = 3;
       }
 
       // Expand map to do multiple DOF per node for block problems
       if (matrixType == "Elasticity2D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
       if (matrixType == "Elasticity3D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
       galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                    << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -220,15 +227,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
         // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-        galeriList.set("right boundary" , "Neumann");
+        galeriList.set("right boundary", "Neumann");
         galeriList.set("bottom boundary", "Neumann");
-        galeriList.set("top boundary"   , "Neumann");
-        galeriList.set("front boundary" , "Neumann");
-        galeriList.set("back boundary"  , "Neumann");
+        galeriList.set("top boundary", "Neumann");
+        galeriList.set("front boundary", "Neumann");
+        galeriList.set("back boundary", "Neumann");
       }
 
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
       A = Pr->BuildMatrix();
 
       if (matrixType == "Elasticity2D" ||
@@ -239,19 +246,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     } else {
       if (!mapFile.empty())
-        map = Xpetra::IO<SC,LO,GO,Node>::ReadMap(mapFile, lib, comm);
+        map = Xpetra::IO<SC, LO, GO, Node>::ReadMap(mapFile, lib, comm);
       comm->barrier();
 
       const bool binaryFormat = false;
 
       if (!binaryFormat && !map.is_null()) {
-        RCP<const Map> colMap    = (!colMapFile.empty()    ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(colMapFile,    lib, comm) : Teuchos::null);
-        RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
-        RCP<const Map> rangeMap  = (!rangeMapFile.empty()  ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(rangeMapFile,  lib, comm) : Teuchos::null);
-        A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
+        RCP<const Map> colMap    = (!colMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(colMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> rangeMap  = (!rangeMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(rangeMapFile, lib, comm) : Teuchos::null);
+        A                        = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
 
       } else {
-        A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, lib, comm, binaryFormat);
+        A = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, lib, comm, binaryFormat);
 
         if (!map.is_null()) {
           RCP<Matrix> newMatrix = MatrixFactory::Build(map, 1);
@@ -269,11 +276,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       if (!coordFile.empty()) {
         // NOTE: currently we only allow reading scalar matrices, thus coordinate
         // map is same as matrix map
-        coordinates = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(coordFile, map);
+        coordinates = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(coordFile, map);
       }
 
       if (!nullFile.empty())
-        nullspace = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(nullFile, map);
+        nullspace = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(nullFile, map);
     }
 
     comm->barrier();
@@ -283,14 +290,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     std::string matrixType = galeriParameters.GetMatrixType();
     std::string aggVizType = paramList.get<std::string>("aggregation: output file: agg style");
-    aggVizType.erase(std::remove_if(aggVizType.begin(),aggVizType.end(), ::isspace), aggVizType.end());
-    if(ndims == 2)
+    aggVizType.erase(std::remove_if(aggVizType.begin(), aggVizType.end(), ::isspace), aggVizType.end());
+    if (ndims == 2)
       paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-2D-Level%LEVELID-Proc%PROCID");
-    else if(ndims == 3){
-      if(comm->getSize() > 1)
-        paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-"+matrixType+"-"+aggVizType+"-Level%LEVELID-Proc%PROCID");
+    else if (ndims == 3) {
+      if (comm->getSize() > 1)
+        paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-" + matrixType + "-" + aggVizType + "-Level%LEVELID-Proc%PROCID");
       else
-        paramList.set<std::string>("aggregation: output filename","MPI-Viz-Output-"+matrixType+"-"+aggVizType+"-Level%LEVELID");
+        paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-" + matrixType + "-" + aggVizType + "-Level%LEVELID");
     }
 
     if (nullspace.is_null()) {
@@ -309,13 +316,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       nullspace = MultiVectorFactory::Build(map, blkSize);
       for (int i = 0; i < blkSize; i++) {
         RCP<const Map> domainMap = A->getDomainMap();
-        GO             indexBase = domainMap->getIndexBase();
+        GO indexBase             = domainMap->getIndexBase();
 
         ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
         for (int j = 0; j < nsData.size(); j++) {
           GO GID = domainMap->getGlobalElement(j) - indexBase;
 
-          if ((GID-i) % blkSize == 0)
+          if ((GID - i) % blkSize == 0)
             nsData[j] = Teuchos::ScalarTraits<SC>::one();
         }
       }
@@ -323,7 +330,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     // Instead of checking each time for rank, create a rank 0 stream
     RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-    Teuchos::FancyOStream& out = *fancy;
+    Teuchos::FancyOStream& out       = *fancy;
     out.setOutputToRootOnly(0);
 
     out << galeriStream.str();
@@ -345,8 +352,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     A->SetMaxEigenvalueEstimate(-one);
 
     H = mueLuFactory->CreateHierarchy();
-    H->GetLevel(0)->Set("A",           A);
-    H->GetLevel(0)->Set("Nullspace",   nullspace);
+    H->GetLevel(0)->Set("A", A);
+    H->GetLevel(0)->Set("Nullspace", nullspace);
     if (!coordinates.is_null())
       H->GetLevel(0)->Set("Coordinates", coordinates);
     mueLuFactory->SetupHierarchy(*H);
@@ -358,30 +365,29 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // Check aggs
     // =========================================================================
     std::string filenameToWrite;
-    if(comm->getSize() > 1)
-      filenameToWrite = "MPI-Viz-Output-"+matrixType+"-"+aggVizType+"-Level0-Proc%PROCID";
+    if (comm->getSize() > 1)
+      filenameToWrite = "MPI-Viz-Output-" + matrixType + "-" + aggVizType + "-Level0-Proc%PROCID";
     else
-      filenameToWrite = "MPI-Viz-Output-"+matrixType+"-"+aggVizType+"-Level0";
-    std::string outfileName = replaceAll(filenameToWrite, "%PROCID", MueLu::toString(comm->getRank() ));
-    aggMatch = compare_to_gold_all_ranks( comm->getRank(), outfileName);
+      filenameToWrite = "MPI-Viz-Output-" + matrixType + "-" + aggVizType + "-Level0";
+    std::string outfileName = replaceAll(filenameToWrite, "%PROCID", MueLu::toString(comm->getRank()));
+    aggMatch                = compare_to_gold_all_ranks(comm->getRank(), outfileName);
 
     // =========================================================================
     // System solution (Ax = b)
     // =========================================================================
     comm->barrier();
-    
 
     success = aggMatch;
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 //- -- --------------------------------------------------------
 #define MUELU_AUTOMATIC_TEST_ETI_NAME main_
 #include "MueLu_Test_ETI.hpp"
 
-int main(int argc, char *argv[]) {
-  return Automatic_Test_ETI(argc,argv);
+int main(int argc, char* argv[]) {
+  return Automatic_Test_ETI(argc, argv);
 }
diff --git a/packages/muelu/test/viz/Viz.cpp b/packages/muelu/test/viz/Viz.cpp
index ec4c1819dd06..4dfc5db3cf7b 100644
--- a/packages/muelu/test/viz/Viz.cpp
+++ b/packages/muelu/test/viz/Viz.cpp
@@ -64,7 +64,7 @@
 #include <MueLu.hpp>
 #include <MueLu_Level.hpp>
 #include <MueLu_BaseClass.hpp>
-#include <MueLu_ParameterListInterpreter.hpp> // TODO: move into MueLu.hpp
+#include <MueLu_ParameterListInterpreter.hpp>  // TODO: move into MueLu.hpp
 
 #include <MueLu_Utilities.hpp>
 
@@ -76,19 +76,19 @@
 #include <BelosBlockCGSolMgr.hpp>
 #include <BelosPseudoBlockCGSolMgr.hpp>
 #include <BelosBlockGmresSolMgr.hpp>
-#include <BelosXpetraAdapter.hpp>     // => This header defines Belos::XpetraOp
-#include <BelosMueLuAdapter.hpp>      // => This header defines Belos::MueLuOp
+#include <BelosXpetraAdapter.hpp>  // => This header defines Belos::XpetraOp
+#include <BelosMueLuAdapter.hpp>   // => This header defines Belos::MueLuOp
 #endif
 
-template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
-int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int argc, char *argv[]) {
+template <class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node>
+int main_(Teuchos::CommandLineProcessor& clp, Xpetra::UnderlyingLib& lib, int argc, char* argv[]) {
 #include <MueLu_UseShortNames.hpp>
 
+  using Teuchos::ArrayRCP;
+  using Teuchos::ParameterList;
   using Teuchos::RCP;
   using Teuchos::rcp;
-  using Teuchos::ArrayRCP;
   using Teuchos::TimeMonitor;
-  using Teuchos::ParameterList;
 
   // =========================================================================
   // MPI initialization using Teuchos
@@ -96,7 +96,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   bool success = false;
   bool verbose = true;
   try {
-    RCP< const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
+    RCP<const Teuchos::Comm<int> > comm = Teuchos::DefaultComm<int>::getComm();
 
     // =========================================================================
     // Convenient definitions
@@ -108,31 +108,46 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     // Parameters initialization
     // =========================================================================
     GO nx = 100, ny = 100, nz = 100;
-    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D"); // manage parameters of the test case
-    Xpetra::Parameters             xpetraParameters(clp);                          // manage parameters of Xpetra
-
-    std::string xmlFileName       = "viztest.xml";     clp.setOption("xml",                   &xmlFileName,       "read parameters from a file");
-    bool        printTimings      = true;              clp.setOption("timings", "notimings",  &printTimings,      "print timings to screen");
-    int         writeMatricesOPT  = -2;                clp.setOption("write",                 &writeMatricesOPT,  "write matrices to file (-1 means all; i>=0 means level i)");
-    std::string dsolveType        = "cg", solveType;   clp.setOption("solver",                &dsolveType,        "solve type: (none | cg | gmres | standalone)");
-    double      dtol              = 1e-12, tol;        clp.setOption("tol",                   &dtol,              "solver convergence tolerance");
-
-    std::string mapFile;                               clp.setOption("map",                   &mapFile,           "map data file");
-    std::string colMapFile;                            clp.setOption("colmap",                &colMapFile,        "colmap data file");
-    std::string domainMapFile;                         clp.setOption("domainmap",             &domainMapFile,     "domainmap data file");
-    std::string rangeMapFile;                          clp.setOption("rangemap",              &rangeMapFile,      "rangemap data file");
-    std::string matrixFile;                            clp.setOption("matrix",                &matrixFile,        "matrix data file");
-    std::string coordFile;                             clp.setOption("coords",                &coordFile,         "coordinates data file");
-    std::string nullFile;                              clp.setOption("nullspace",             &nullFile,          "nullspace data file");
-    int         numRebuilds       = 0;                 clp.setOption("rebuild",               &numRebuilds,       "#times to rebuild hierarchy");
-    int         maxIts            = 200;               clp.setOption("its",                   &maxIts,            "maximum number of solver iterations");
-    bool        scaleResidualHist = true;              clp.setOption("scale", "noscale",      &scaleResidualHist, "scaled Krylov residual history");
+    Galeri::Xpetra::Parameters<GO> galeriParameters(clp, nx, ny, nz, "Laplace2D");  // manage parameters of the test case
+    Xpetra::Parameters xpetraParameters(clp);                                       // manage parameters of Xpetra
+
+    std::string xmlFileName = "viztest.xml";
+    clp.setOption("xml", &xmlFileName, "read parameters from a file");
+    bool printTimings = true;
+    clp.setOption("timings", "notimings", &printTimings, "print timings to screen");
+    int writeMatricesOPT = -2;
+    clp.setOption("write", &writeMatricesOPT, "write matrices to file (-1 means all; i>=0 means level i)");
+    std::string dsolveType = "cg", solveType;
+    clp.setOption("solver", &dsolveType, "solve type: (none | cg | gmres | standalone)");
+    double dtol = 1e-12, tol;
+    clp.setOption("tol", &dtol, "solver convergence tolerance");
+
+    std::string mapFile;
+    clp.setOption("map", &mapFile, "map data file");
+    std::string colMapFile;
+    clp.setOption("colmap", &colMapFile, "colmap data file");
+    std::string domainMapFile;
+    clp.setOption("domainmap", &domainMapFile, "domainmap data file");
+    std::string rangeMapFile;
+    clp.setOption("rangemap", &rangeMapFile, "rangemap data file");
+    std::string matrixFile;
+    clp.setOption("matrix", &matrixFile, "matrix data file");
+    std::string coordFile;
+    clp.setOption("coords", &coordFile, "coordinates data file");
+    std::string nullFile;
+    clp.setOption("nullspace", &nullFile, "nullspace data file");
+    int numRebuilds = 0;
+    clp.setOption("rebuild", &numRebuilds, "#times to rebuild hierarchy");
+    int maxIts = 200;
+    clp.setOption("its", &maxIts, "maximum number of solver iterations");
+    bool scaleResidualHist = true;
+    clp.setOption("scale", "noscale", &scaleResidualHist, "scaled Krylov residual history");
 
     switch (clp.parse(argc, argv)) {
-      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:        return EXIT_SUCCESS;
+      case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED: return EXIT_SUCCESS;
       case Teuchos::CommandLineProcessor::PARSE_ERROR:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION: return EXIT_FAILURE;
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     ParameterList paramList;
@@ -161,13 +176,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
     RCP<TimeMonitor> globalTimeMonitor = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: S - Global Time")));
     RCP<TimeMonitor> tm                = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 1 - Matrix Build")));
 
-    RCP<Matrix>      A;
-    RCP<const Map>   map;
+    RCP<Matrix> A;
+    RCP<const Map> map;
     RCP<MultiVector> coordinates;
     RCP<MultiVector> nullspace;
     int ndims = 2;
     if (matrixFile.empty()) {
-      galeriStream << "========================================================\n" << xpetraParameters << galeriParameters;
+      galeriStream << "========================================================\n"
+                   << xpetraParameters << galeriParameters;
 
       // Galeri will attempt to create a square-as-possible distribution of subdomains di, e.g.,
       //                                 d1  d2  d3
@@ -184,25 +200,25 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       // In the future, we hope to be able to first create a Galeri problem, and then request map and coordinates from it
       // At the moment, however, things are fragile as we hope that the Problem uses same map and coordinates inside
       if (matrixType == "Laplace1D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("1D", map, galeriList);
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian1D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("1D", map, galeriList);
 
       } else if (matrixType == "Laplace2D" || matrixType == "Star2D" ||
-          matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("2D", map, galeriList);
+                 matrixType == "BigStar2D" || matrixType == "Elasticity2D") {
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian2D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("2D", map, galeriList);
 
       } else if (matrixType == "Laplace3D" || matrixType == "Brick3D" || matrixType == "Elasticity3D") {
-        map = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
-        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC,LO,GO,Map,MultiVector>("3D", map, galeriList);
-        ndims = 3;
+        map         = Galeri::Xpetra::CreateMap<LO, GO, Node>(xpetraParameters.GetLib(), "Cartesian3D", comm, galeriList);
+        coordinates = Galeri::Xpetra::Utils::CreateCartesianCoordinates<SC, LO, GO, Map, MultiVector>("3D", map, galeriList);
+        ndims       = 3;
       }
 
       // Expand map to do multiple DOF per node for block problems
       if (matrixType == "Elasticity2D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 2);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 2);
       if (matrixType == "Elasticity3D")
-        map = Xpetra::MapFactory<LO,GO,Node>::Build(map, 3);
+        map = Xpetra::MapFactory<LO, GO, Node>::Build(map, 3);
 
       galeriStream << "Processor subdomains in x direction: " << galeriList.get<GO>("mx") << std::endl
                    << "Processor subdomains in y direction: " << galeriList.get<GO>("my") << std::endl
@@ -211,15 +227,15 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       if (matrixType == "Elasticity2D" || matrixType == "Elasticity3D") {
         // Our default test case for elasticity: all boundaries of a square/cube have Neumann b.c. except left which has Dirichlet
-        galeriList.set("right boundary" , "Neumann");
+        galeriList.set("right boundary", "Neumann");
         galeriList.set("bottom boundary", "Neumann");
-        galeriList.set("top boundary"   , "Neumann");
-        galeriList.set("front boundary" , "Neumann");
-        galeriList.set("back boundary"  , "Neumann");
+        galeriList.set("top boundary", "Neumann");
+        galeriList.set("front boundary", "Neumann");
+        galeriList.set("back boundary", "Neumann");
       }
 
-      RCP<Galeri::Xpetra::Problem<Map,CrsMatrixWrap,MultiVector> > Pr =
-        Galeri::Xpetra::BuildProblem<SC,LO,GO,Map,CrsMatrixWrap,MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
+      RCP<Galeri::Xpetra::Problem<Map, CrsMatrixWrap, MultiVector> > Pr =
+          Galeri::Xpetra::BuildProblem<SC, LO, GO, Map, CrsMatrixWrap, MultiVector>(galeriParameters.GetMatrixType(), map, galeriList);
       A = Pr->BuildMatrix();
 
       if (matrixType == "Elasticity2D" ||
@@ -230,19 +246,19 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     } else {
       if (!mapFile.empty())
-        map = Xpetra::IO<SC,LO,GO,Node>::ReadMap(mapFile, lib, comm);
+        map = Xpetra::IO<SC, LO, GO, Node>::ReadMap(mapFile, lib, comm);
       comm->barrier();
 
       const bool binaryFormat = false;
 
       if (!binaryFormat && !map.is_null()) {
-        RCP<const Map> colMap    = (!colMapFile.empty()    ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(colMapFile,    lib, comm) : Teuchos::null);
-        RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
-        RCP<const Map> rangeMap  = (!rangeMapFile.empty()  ? Xpetra::IO<SC,LO,GO,Node>::ReadMap(rangeMapFile,  lib, comm) : Teuchos::null);
-        A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
+        RCP<const Map> colMap    = (!colMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(colMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> domainMap = (!domainMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(domainMapFile, lib, comm) : Teuchos::null);
+        RCP<const Map> rangeMap  = (!rangeMapFile.empty() ? Xpetra::IO<SC, LO, GO, Node>::ReadMap(rangeMapFile, lib, comm) : Teuchos::null);
+        A                        = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, map, colMap, domainMap, rangeMap);
 
       } else {
-        A = Xpetra::IO<SC,LO,GO,Node>::Read(matrixFile, lib, comm, binaryFormat);
+        A = Xpetra::IO<SC, LO, GO, Node>::Read(matrixFile, lib, comm, binaryFormat);
 
         if (!map.is_null()) {
           RCP<Matrix> newMatrix = MatrixFactory::Build(map, 1);
@@ -260,11 +276,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
       if (!coordFile.empty()) {
         // NOTE: currently we only allow reading scalar matrices, thus coordinate
         // map is same as matrix map
-        coordinates = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(coordFile, map);
+        coordinates = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(coordFile, map);
       }
 
       if (!nullFile.empty())
-        nullspace = Xpetra::IO<SC,LO,GO,Node>::ReadMultiVector(nullFile, map);
+        nullspace = Xpetra::IO<SC, LO, GO, Node>::ReadMultiVector(nullFile, map);
     }
 
     comm->barrier();
@@ -272,9 +288,9 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
     galeriStream << "Galeri complete.\n========================================================" << std::endl;
 
-    if(ndims == 2)
+    if (ndims == 2)
       paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-2D-Level%LEVELID-Proc%PROCID");
-    else if(ndims == 3)
+    else if (ndims == 3)
       paramList.set<std::string>("aggregation: output filename", "MPI-Viz-Output-3D-Level%LEVELID-Proc%PROCID");
     int numReruns = 1;
     if (paramList.isParameter("number of reruns"))
@@ -286,11 +302,11 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
       bool stop = false;
       if (isDriver) {
-        runList   = paramList.sublist("Run1",  mustAlreadyExist);
-        mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+        runList   = paramList.sublist("Run1", mustAlreadyExist);
+        mueluList = runList.sublist("MueLu", mustAlreadyExist);
       } else {
         mueluList = paramList;
-        stop = true;
+        stop      = true;
       }
 
       mueluList.set("use kokkos refactor", false);
@@ -311,13 +327,13 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         nullspace = MultiVectorFactory::Build(map, blkSize);
         for (int i = 0; i < blkSize; i++) {
           RCP<const Map> domainMap = A->getDomainMap();
-          GO             indexBase = domainMap->getIndexBase();
+          GO indexBase             = domainMap->getIndexBase();
 
           ArrayRCP<SC> nsData = nullspace->getDataNonConst(i);
           for (int j = 0; j < nsData.size(); j++) {
             GO GID = domainMap->getGlobalElement(j) - indexBase;
 
-            if ((GID-i) % blkSize == 0)
+            if ((GID - i) % blkSize == 0)
               nsData[j] = Teuchos::ScalarTraits<SC>::one();
           }
         }
@@ -328,7 +344,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
         solveType = dsolveType;
         tol       = dtol;
 
-        int   savedOut  = -1;
+        int savedOut    = -1;
         FILE* openedOut = NULL;
         if (isDriver) {
           if (runList.isParameter("filename")) {
@@ -345,12 +361,12 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
             dup2(fileno(openedOut), STDOUT_FILENO);
           }
           if (runList.isParameter("solver")) solveType = runList.get<std::string>("solver");
-          if (runList.isParameter("tol"))    tol       = runList.get<double>     ("tol");
+          if (runList.isParameter("tol")) tol = runList.get<double>("tol");
         }
 
         // Instead of checking each time for rank, create a rank 0 stream
         RCP<Teuchos::FancyOStream> fancy = Teuchos::fancyOStream(Teuchos::rcpFromRef(std::cout));
-        Teuchos::FancyOStream& out = *fancy;
+        Teuchos::FancyOStream& out       = *fancy;
         out.setOutputToRootOnly(0);
 
         out << galeriStream.str();
@@ -373,8 +389,8 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           A->SetMaxEigenvalueEstimate(-one);
 
           H = mueLuFactory->CreateHierarchy();
-          H->GetLevel(0)->Set("A",           A);
-          H->GetLevel(0)->Set("Nullspace",   nullspace);
+          H->GetLevel(0)->Set("A", A);
+          H->GetLevel(0)->Set("Nullspace", nullspace);
           if (!coordinates.is_null())
             H->GetLevel(0)->Set("Coordinates", coordinates);
           mueLuFactory->SetupHierarchy(*H);
@@ -401,7 +417,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
           Teuchos::Array<typename STS::magnitudeType> norms(1);
           B->norm2(norms);
-          B->scale(one/norms[0]);
+          B->scale(one / norms[0]);
           X->putScalar(zero);
         }
         tm = Teuchos::null;
@@ -419,7 +435,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
         } else if (solveType == "standalone") {
           tm.reset();
-          tm = rcp (new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
+          tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 4 - Fixed Point Solve")));
 
           H->IsPreconditioner(false);
           H->Iterate(*B, *X, maxIts);
@@ -430,17 +446,17 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           tm = rcp(new TimeMonitor(*TimeMonitor::getNewTimer("ScalingTest: 5 - Belos Solve")));
 
           // Operator and Multivector type that will be used with Belos
-          typedef MultiVector          MV;
+          typedef MultiVector MV;
           typedef Belos::OperatorT<MV> OP;
 
           H->IsPreconditioner(true);
 
           // Define Operator and Preconditioner
-          Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A)); // Turns a Xpetra::Matrix object into a Belos operator
-          Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp <SC, LO, GO, NO>(H)); // Turns a MueLu::Hierarchy object into a Belos operator
+          Teuchos::RCP<OP> belosOp   = Teuchos::rcp(new Belos::XpetraOp<SC, LO, GO, NO>(A));  // Turns a Xpetra::Matrix object into a Belos operator
+          Teuchos::RCP<OP> belosPrec = Teuchos::rcp(new Belos::MueLuOp<SC, LO, GO, NO>(H));   // Turns a MueLu::Hierarchy object into a Belos operator
 
           // Construct a Belos LinearProblem object
-          RCP< Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
+          RCP<Belos::LinearProblem<SC, MV, OP> > belosProblem = rcp(new Belos::LinearProblem<SC, MV, OP>(belosOp, X, B));
           belosProblem->setRightPrec(belosPrec);
 
           bool set = belosProblem->setProblem();
@@ -451,40 +467,42 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
 
           // Belos parameter list
           Teuchos::ParameterList belosList;
-          belosList.set("Maximum Iterations",    maxIts); // Maximum number of iterations allowed
-          belosList.set("Convergence Tolerance", tol);    // Relative convergence tolerance requested
-          belosList.set("Verbosity",             Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
-          belosList.set("Output Frequency",      1);
-          belosList.set("Output Style",          Belos::Brief);
+          belosList.set("Maximum Iterations", maxIts);  // Maximum number of iterations allowed
+          belosList.set("Convergence Tolerance", tol);  // Relative convergence tolerance requested
+          belosList.set("Verbosity", Belos::Errors + Belos::Warnings + Belos::StatusTestDetails);
+          belosList.set("Output Frequency", 1);
+          belosList.set("Output Style", Belos::Brief);
           if (!scaleResidualHist)
             belosList.set("Implicit Residual Scaling", "None");
 
           // Create an iterative solver manager
-          RCP< Belos::SolverManager<SC, MV, OP> > solver;
+          RCP<Belos::SolverManager<SC, MV, OP> > solver;
           if (solveType == "cg") {
-            solver = rcp(new Belos::PseudoBlockCGSolMgr   <SC, MV, OP>(belosProblem, rcp(&belosList, false)));
+            solver = rcp(new Belos::PseudoBlockCGSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
           } else if (solveType == "gmres") {
             solver = rcp(new Belos::BlockGmresSolMgr<SC, MV, OP>(belosProblem, rcp(&belosList, false)));
           }
 
           // Perform solve
           Belos::ReturnType ret = Belos::Unconverged;
-          ret = solver->solve();
+          ret                   = solver->solve();
 
           // Get the number of iterations for this solve.
           out << "Number of iterations performed for this solve: " << solver->getNumIters() << std::endl;
 
           // Check convergence
           if (ret != Belos::Converged)
-            out << std::endl << "ERROR:  Belos did not converge! " << std::endl;
+            out << std::endl
+                << "ERROR:  Belos did not converge! " << std::endl;
           else
-            out << std::endl << "SUCCESS:  Belos converged!" << std::endl;
-#endif //ifdef HAVE_MUELU_BELOS
+            out << std::endl
+                << "SUCCESS:  Belos converged!" << std::endl;
+#endif  // ifdef HAVE_MUELU_BELOS
         } else {
           throw MueLu::Exceptions::RuntimeError("Unknown solver type: \"" + solveType + "\"");
         }
         comm->barrier();
-        tm = Teuchos::null;
+        tm                = Teuchos::null;
         globalTimeMonitor = Teuchos::null;
 
         if (printTimings) {
@@ -508,7 +526,7 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
           }
           try {
             runList   = paramList.sublist("Run" + MueLu::toString(++runCount), mustAlreadyExist);
-            mueluList = runList  .sublist("MueLu", mustAlreadyExist);
+            mueluList = runList.sublist("MueLu", mustAlreadyExist);
           } catch (Teuchos::Exceptions::InvalidParameterName& e) {
             stop = true;
           }
@@ -521,14 +539,14 @@ int main_(Teuchos::CommandLineProcessor &clp, Xpetra::UnderlyingLib& lib, int ar
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
 
 int main(int argc, char* argv[]) {
   bool success = false;
   bool verbose = true;
 
-  Teuchos::GlobalMPISession mpiSession(&argc,&argv);
+  Teuchos::GlobalMPISession mpiSession(&argc, &argv);
 
   try {
     const bool throwExceptions     = false;
@@ -537,20 +555,21 @@ int main(int argc, char* argv[]) {
     Teuchos::CommandLineProcessor clp(throwExceptions, recogniseAllOptions);
     Xpetra::Parameters xpetraParameters(clp);
 
-    std::string node = "";  clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
+    std::string node = "";
+    clp.setOption("node", &node, "node type (serial | openmp | cuda | hip)");
 
     switch (clp.parse(argc, argv, NULL)) {
-      case Teuchos::CommandLineProcessor::PARSE_ERROR:               return EXIT_FAILURE;
+      case Teuchos::CommandLineProcessor::PARSE_ERROR: return EXIT_FAILURE;
       case Teuchos::CommandLineProcessor::PARSE_HELP_PRINTED:
       case Teuchos::CommandLineProcessor::PARSE_UNRECOGNIZED_OPTION:
-      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL:          break;
+      case Teuchos::CommandLineProcessor::PARSE_SUCCESSFUL: break;
     }
 
     Xpetra::UnderlyingLib lib = xpetraParameters.GetLib();
 
     if (lib == Xpetra::UseEpetra) {
 #ifdef HAVE_MUELU_EPETRA
-      return main_<double,int,int,Xpetra::EpetraNode>(clp, lib, argc, argv);
+      return main_<double, int, int, Xpetra::EpetraNode>(clp, lib, argc, argv);
 #else
       throw MueLu::Exceptions::RuntimeError("Epetra is not available");
 #endif
@@ -561,35 +580,35 @@ int main(int argc, char* argv[]) {
         typedef Tpetra::KokkosClassic::DefaultNode::DefaultNodeType Node;
 
 #ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
 #else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#  else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#  endif
+#endif
 #endif
       } else if (node == "serial") {
 #ifdef KOKKOS_ENABLE_SERIAL
         typedef Tpetra::KokkosCompat::KokkosSerialWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SERIAL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("Serial node type is disabled");
 #endif
@@ -597,19 +616,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_OPENMP
         typedef Tpetra::KokkosCompat::KokkosOpenMPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if   defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_OPENMP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("OpenMP node type is disabled");
 #endif
@@ -617,19 +636,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_CUDA
         typedef Tpetra::KokkosCompat::KokkosCudaWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_CUDA) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("CUDA node type is disabled");
 #endif
@@ -637,19 +656,19 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_HIP
         typedef Tpetra::KokkosCompat::KokkosHIPWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_HIP) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("HIP node type is disabled");
 #endif
@@ -657,22 +676,22 @@ int main(int argc, char* argv[]) {
 #ifdef KOKKOS_ENABLE_SYCL
         typedef Tpetra::KokkosCompat::KokkosSYCLWrapperNode Node;
 
-#  ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#  else
-#    if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
-        return main_<double,int,int,Node> (clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
-        return main_<double,int,long,Node>(clp, lib, argc, argv);
-#    elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
-        return main_<double,int,long long,Node>(clp, lib, argc, argv);
-#    else
+#ifndef HAVE_MUELU_EXPLICIT_INSTANTIATION
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#else
+#if defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_INT)
+        return main_<double, int, int, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG)
+        return main_<double, int, long, Node>(clp, lib, argc, argv);
+#elif defined(HAVE_TPETRA_INST_DOUBLE) && defined(HAVE_TPETRA_INST_SYCL) && defined(HAVE_TPETRA_INST_INT_LONG_LONG)
+        return main_<double, int, long long, Node>(clp, lib, argc, argv);
+#else
         throw MueLu::Exceptions::RuntimeError("Found no suitable instantiation");
-#    endif
-#  endif
+#endif
+#endif
 #else
         throw MueLu::Exceptions::RuntimeError("SYCL node type is disabled");
-#endif	
+#endif
       } else {
         throw MueLu::Exceptions::RuntimeError("Unrecognized node type");
       }
@@ -680,5 +699,5 @@ int main(int argc, char* argv[]) {
   }
   TEUCHOS_STANDARD_CATCH_STATEMENTS(verbose, std::cerr, success);
 
-  return ( success ? EXIT_SUCCESS : EXIT_FAILURE );
+  return (success ? EXIT_SUCCESS : EXIT_FAILURE);
 }
diff --git a/packages/muelu/utils/map/t2e.cpp b/packages/muelu/utils/map/t2e.cpp
index 7a788ecce525..ca1cee701ac9 100644
--- a/packages/muelu/utils/map/t2e.cpp
+++ b/packages/muelu/utils/map/t2e.cpp
@@ -4,57 +4,57 @@
 
 using namespace std;
 int main(int argc, char* argv[]) {
-    if (argc < 3) {
-        cout << "Usage: ./a.out <in> <out>" << endl;
-        return 1;
-    }
-    ifstream ifs(argv[1]);
-    ofstream ofs(argv[2]);
-
-    std::cout << "Reading Tpetra map \"" << argv[1] << "\"" << std::endl;
-    // Skip % MatrixMarket header
-    char line[256];
-    while (true) {
-      ifs.getline(line, 256);
-      if (line[0] != '%')
-        break;
-    }
-
-    int n;
-    sscanf(line, "%d", &n);
-    std::cout << "n = " << n << std::endl;
-
-    std::vector<std::vector<int> > gids;
-    for (int i = 0; i < n; i += 2) {
-      int id, procid;
-      ifs >> id >> procid;
-
-      if (procid >= gids.size())
-        gids.resize(procid+1);
-      gids[procid].push_back(id);
-    }
-
-    ofs << "%%MatrixMarket matrix array integer general" << std::endl;
-    ofs << "%Format Version:" << std::endl;
-    ofs << "% 2 " << std::endl;
-    ofs << "%NumProc: Number of processors:" << std::endl;
-    ofs << "% " << gids.size() << " " << std::endl;
-    ofs << "%MaxElementSize: Maximum element size:" << std::endl;
-    ofs << "% 1 " << std::endl;
-    ofs << "%MinElementSize: Minimum element size:" << std::endl;
-    ofs << "% 1 " << std::endl;
-    ofs << "%IndexBase: Index base of map:" << std::endl;
-    ofs << "% 0 " << std::endl;
-    ofs << "%NumGlobalElements: Total number of GIDs in map:" << std::endl;
-    ofs << "% " << n/2 << " " << std::endl;
-    ofs << "%NumMyElements: BlockMap lengths per processor:" << std::endl;
-    for (int k = 0; k < gids.size(); k++)
-      ofs << "% " << gids[k].size() << std::endl;
-    ofs << n/2 << " 1" << std::endl;
-
-    for (int k = 0; k < gids.size(); k++)
-      for (int i = 0; i < gids[k].size(); i++)
-        ofs << gids[k][i] << std::endl;
-
-    return 0;
+  if (argc < 3) {
+    cout << "Usage: ./a.out <in> <out>" << endl;
+    return 1;
+  }
+  ifstream ifs(argv[1]);
+  ofstream ofs(argv[2]);
+
+  std::cout << "Reading Tpetra map \"" << argv[1] << "\"" << std::endl;
+  // Skip % MatrixMarket header
+  char line[256];
+  while (true) {
+    ifs.getline(line, 256);
+    if (line[0] != '%')
+      break;
+  }
+
+  int n;
+  sscanf(line, "%d", &n);
+  std::cout << "n = " << n << std::endl;
+
+  std::vector<std::vector<int> > gids;
+  for (int i = 0; i < n; i += 2) {
+    int id, procid;
+    ifs >> id >> procid;
+
+    if (procid >= gids.size())
+      gids.resize(procid + 1);
+    gids[procid].push_back(id);
+  }
+
+  ofs << "%%MatrixMarket matrix array integer general" << std::endl;
+  ofs << "%Format Version:" << std::endl;
+  ofs << "% 2 " << std::endl;
+  ofs << "%NumProc: Number of processors:" << std::endl;
+  ofs << "% " << gids.size() << " " << std::endl;
+  ofs << "%MaxElementSize: Maximum element size:" << std::endl;
+  ofs << "% 1 " << std::endl;
+  ofs << "%MinElementSize: Minimum element size:" << std::endl;
+  ofs << "% 1 " << std::endl;
+  ofs << "%IndexBase: Index base of map:" << std::endl;
+  ofs << "% 0 " << std::endl;
+  ofs << "%NumGlobalElements: Total number of GIDs in map:" << std::endl;
+  ofs << "% " << n / 2 << " " << std::endl;
+  ofs << "%NumMyElements: BlockMap lengths per processor:" << std::endl;
+  for (int k = 0; k < gids.size(); k++)
+    ofs << "% " << gids[k].size() << std::endl;
+  ofs << n / 2 << " 1" << std::endl;
+
+  for (int k = 0; k < gids.size(); k++)
+    for (int i = 0; i < gids[k].size(); i++)
+      ofs << gids[k][i] << std::endl;
+
+  return 0;
 }
diff --git a/packages/muelu/utils/matrix/ascii2binary.cpp b/packages/muelu/utils/matrix/ascii2binary.cpp
index 436bfcfe9c59..4e6635254925 100644
--- a/packages/muelu/utils/matrix/ascii2binary.cpp
+++ b/packages/muelu/utils/matrix/ascii2binary.cpp
@@ -8,87 +8,87 @@
 
 using namespace std;
 int main(int argc, char* argv[]) {
-    if (argc < 3) {
-        cout << "Usage: ./a.out <in> <out>" << endl;
-        return 1;
-    }
-    if (std::ifstream(argv[2])) {
-      std::cerr << "target file already exists" << std::endl;
-      return 1;
-    }
-    ifstream ifs(argv[1]);
-    ofstream ofs(argv[2], std::ios::binary);
+  if (argc < 3) {
+    cout << "Usage: ./a.out <in> <out>" << endl;
+    return 1;
+  }
+  if (std::ifstream(argv[2])) {
+    std::cerr << "target file already exists" << std::endl;
+    return 1;
+  }
+  ifstream ifs(argv[1]);
+  ofstream ofs(argv[2], std::ios::binary);
 
-    std::cout << "Reading matrix \"" << argv[1] << "\"" << std::endl;
-    // Skip %% MatrixMarket header and any comments.
-    char line[256];
-    char percent[1];
-    percent[0] = '%';
+  std::cout << "Reading matrix \"" << argv[1] << "\"" << std::endl;
+  // Skip %% MatrixMarket header and any comments.
+  char line[256];
+  char percent[1];
+  percent[0] = '%';
+  ifs.getline(line, 256);
+  while (strncmp(line, percent, 1) == 0) {
     ifs.getline(line, 256);
-    while (strncmp(line,percent,1) == 0) {
-      ifs.getline(line, 256);
-    }
+  }
 
-    int m, n, nnz;
-    int numConverted = sscanf(line,"%d %d %d",&m,&n,&nnz);
-
-    if (numConverted != 3) {
-      std::ostringstream errStr;
-      errStr << "Error reading matrix dimensions.  Expected 3 integers, found " << numConverted;
-      throw(std::runtime_error(errStr.str()));
-    }
-    ofs.write(reinterpret_cast<char*>(&m),   sizeof(m));
-    ofs.write(reinterpret_cast<char*>(&n),   sizeof(n));
-    ofs.write(reinterpret_cast<char*>(&nnz), sizeof(nnz));
+  int m, n, nnz;
+  int numConverted = sscanf(line, "%d %d %d", &m, &n, &nnz);
 
-    int    row = -99, i, j;
-    double v;
-    vector<int>    inds;
-    vector<double> vals;
-    vector<bool>   writtenRows(m, false);
-    size_t         writtenEntries = 0;
-    bool           done = false;
+  if (numConverted != 3) {
+    std::ostringstream errStr;
+    errStr << "Error reading matrix dimensions.  Expected 3 integers, found " << numConverted;
+    throw(std::runtime_error(errStr.str()));
+  }
+  ofs.write(reinterpret_cast<char*>(&m), sizeof(m));
+  ofs.write(reinterpret_cast<char*>(&n), sizeof(n));
+  ofs.write(reinterpret_cast<char*>(&nnz), sizeof(nnz));
 
-    ifs >> i >> j >> v;
-    i--; j--;
-    while (row != i && !done) {
-        row = i;
+  int row = -99, i, j;
+  double v;
+  vector<int> inds;
+  vector<double> vals;
+  vector<bool> writtenRows(m, false);
+  size_t writtenEntries = 0;
+  bool done             = false;
 
-        inds.resize(0);
-        vals.resize(0);
+  ifs >> i >> j >> v;
+  i--;
+  j--;
+  while (row != i && !done) {
+    row = i;
 
-        do {
-            inds.push_back(j);
-            vals.push_back(v);
+    inds.resize(0);
+    vals.resize(0);
 
-            if (ifs.good()) {
-              ifs >> i >> j >> v;
-              i--; j--;
-            } else {
-              i = -1;
-              done = true;
-            }
-        } while (row == i);
+    do {
+      inds.push_back(j);
+      vals.push_back(v);
 
+      if (ifs.good()) {
+        ifs >> i >> j >> v;
+        i--;
+        j--;
+      } else {
+        i    = -1;
+        done = true;
+      }
+    } while (row == i);
 
-        int rownnz = inds.size();
-        ofs.write(reinterpret_cast<char*>(&row),    sizeof(row));
-        ofs.write(reinterpret_cast<char*>(&rownnz), sizeof(rownnz));
-        for (int k = 0; k < rownnz; k++) ofs.write(reinterpret_cast<char*>(&inds[0] + k), sizeof(inds[k]));
-        for (int k = 0; k < rownnz; k++) ofs.write(reinterpret_cast<char*>(&vals[0] + k), sizeof(vals[k]));
-        writtenRows[row] = true;
-        writtenEntries += rownnz;
-
-    }
-    assert (writtenEntries == nnz);
+    int rownnz = inds.size();
+    ofs.write(reinterpret_cast<char*>(&row), sizeof(row));
+    ofs.write(reinterpret_cast<char*>(&rownnz), sizeof(rownnz));
+    for (int k = 0; k < rownnz; k++) ofs.write(reinterpret_cast<char*>(&inds[0] + k), sizeof(inds[k]));
+    for (int k = 0; k < rownnz; k++) ofs.write(reinterpret_cast<char*>(&vals[0] + k), sizeof(vals[k]));
+    writtenRows[row] = true;
+    writtenEntries += rownnz;
+  }
+  assert(writtenEntries == nnz);
 
-    int rownnz = 0;
-    for (row = 0; row < m; row++) {
-      if (!writtenRows[row]) {
-        ofs.write(reinterpret_cast<char*>(&row),    sizeof(row));
-        ofs.write(reinterpret_cast<char*>(&rownnz), sizeof(rownnz));
-      }
+  int rownnz = 0;
+  for (row = 0; row < m; row++) {
+    if (!writtenRows[row]) {
+      ofs.write(reinterpret_cast<char*>(&row), sizeof(row));
+      ofs.write(reinterpret_cast<char*>(&rownnz), sizeof(rownnz));
     }
+  }
 
-    return 0;
+  return 0;
 }
diff --git a/packages/muelu/utils/run_clang_format.sh b/packages/muelu/utils/run_clang_format.sh
new file mode 100755
index 000000000000..3f4b58bb8891
--- /dev/null
+++ b/packages/muelu/utils/run_clang_format.sh
@@ -0,0 +1,57 @@
+#!/bin/bash
+
+# assume this is run from MueLu's top level
+if [ ! -f "utils/run_clang_format.sh" ] ; then
+  echo "Error! This script should be run from the packages/muelu directory! Please use"
+  echo "  utils/run_clang_format.sh"
+  exit 1
+fi
+
+# check if clang-format exists on the user's machine. if not, formatting can't be performed
+if [ ! command -v clang-format &> /dev/null ] ; then
+  echo "Error! clang-format was not found! Please place clang-format in your PATH or current directory"
+  echo
+  echo "If you are looking to reformat your code due to the MueLu clang-format Github action, please use clang-format 14.0.0"
+  echo 'See https://github.com/llvm/llvm-project/releases/tag/llvmorg-14.0.0 to find the appropriate clang-format version'
+  echo "Linux x64 developers can find it here:"
+  echo '  https://github.com/llvm/llvm-project/releases/download/llvmorg-14.0.0/clang+llvm-14.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz'
+  exit 1
+else
+  if [[ ! "$@" =~ "-f" ]] ; then
+    # get the clang-format version by running clang-format --version
+    CLANGFORMAT_VERSION=$(clang-format --version)
+    # check if clang-format result contains the string 14.0.0. if not, error out
+    if [[ ! "${CLANGFORMAT_VERSION}" =~ "14.0.0" ]] ; then
+      echo "Error! clang-format is not version 14.0.0, so it may have conflicts with the MueLu style!"
+      echo "Detected clang-format version: ${CLANGFORMAT_VERSION}"
+      echo "If you would like to run this anyway, you can force run this with"
+      echo "  ./utils/run_clang_format.sh -f"
+      echo
+      echo "If you are looking to reformat your code due to the MueLu clang-format Github action, please use clang-format 14.0.0"
+      echo 'See https://github.com/llvm/llvm-project/releases/tag/llvmorg-14.0.0 to find the appropriate clang-format version'
+      echo "Linux x64 developers can find it here:"
+      echo '  https://github.com/llvm/llvm-project/releases/download/llvmorg-14.0.0/clang+llvm-14.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz'
+      exit 1
+    fi
+  fi
+fi
+
+# grab all *.hpp files and *.cpp files in the MueLu src directory
+HEADERS=$(find . -name "*.hpp")
+SOURCES=$(find . -name "*.cpp")
+
+echo "Running MueLu clang formatting..."
+
+# apply clang-format inline to each **.hpp, utilizing the style file
+for HEADER in ${HEADERS} ; do
+  echo ${HEADER}
+  clang-format -style=file -i ${HEADER}
+done
+
+# apply clang-format inline to each **.cpp, utilizing the style file
+for SOURCE in ${SOURCES} ; do
+  echo ${SOURCE}
+  clang-format -style=file -i ${SOURCE}
+done
+
+echo "Done running MueLu clang formatting!"